diff --git "a/trainer_state.json" "b/trainer_state.json" new file mode 100644--- /dev/null +++ "b/trainer_state.json" @@ -0,0 +1,99343 @@ +{ + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 1.0, + "global_step": 16553, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.0, + "learning_rate": 4.0241448692152927e-08, + "loss": 7.0893, + "step": 1 + }, + { + "epoch": 0.0, + "learning_rate": 8.048289738430585e-08, + "loss": 7.0402, + "step": 2 + }, + { + "epoch": 0.0, + "learning_rate": 1.2072434607645875e-07, + "loss": 7.1295, + "step": 3 + }, + { + "epoch": 0.0, + "learning_rate": 1.609657947686117e-07, + "loss": 7.0446, + "step": 4 + }, + { + "epoch": 0.0, + "learning_rate": 2.012072434607646e-07, + "loss": 7.1339, + "step": 5 + }, + { + "epoch": 0.0, + "learning_rate": 2.414486921529175e-07, + "loss": 7.0804, + "step": 6 + }, + { + "epoch": 0.0, + "learning_rate": 2.8169014084507043e-07, + "loss": 7.0268, + "step": 7 + }, + { + "epoch": 0.0, + "learning_rate": 3.219315895372234e-07, + "loss": 7.1652, + "step": 8 + }, + { + "epoch": 0.0, + "learning_rate": 3.621730382293763e-07, + "loss": 7.1696, + "step": 9 + }, + { + "epoch": 0.0, + "learning_rate": 4.024144869215292e-07, + "loss": 6.9821, + "step": 10 + }, + { + "epoch": 0.0, + "learning_rate": 4.4265593561368214e-07, + "loss": 6.8884, + "step": 11 + }, + { + "epoch": 0.0, + "learning_rate": 4.82897384305835e-07, + "loss": 6.5045, + "step": 12 + }, + { + "epoch": 0.0, + "learning_rate": 5.23138832997988e-07, + "loss": 6.5536, + "step": 13 + }, + { + "epoch": 0.0, + "learning_rate": 5.633802816901409e-07, + "loss": 6.3884, + "step": 14 + }, + { + "epoch": 0.0, + "learning_rate": 6.036217303822938e-07, + "loss": 6.317, + "step": 15 + }, + { + "epoch": 0.0, + "learning_rate": 6.438631790744468e-07, + "loss": 5.3705, + "step": 16 + }, + { + "epoch": 0.0, + "learning_rate": 6.841046277665996e-07, + "loss": 5.3393, + "step": 17 + }, + { + "epoch": 0.0, + "learning_rate": 7.243460764587526e-07, + "loss": 5.25, + "step": 18 + }, + { + "epoch": 0.0, + "learning_rate": 7.645875251509055e-07, + "loss": 5.1562, + "step": 19 + }, + { + "epoch": 0.0, + "learning_rate": 8.048289738430584e-07, + "loss": 5.1071, + "step": 20 + }, + { + "epoch": 0.0, + "learning_rate": 8.450704225352114e-07, + "loss": 4.5938, + "step": 21 + }, + { + "epoch": 0.0, + "learning_rate": 8.853118712273643e-07, + "loss": 3.8549, + "step": 22 + }, + { + "epoch": 0.0, + "learning_rate": 9.255533199195171e-07, + "loss": 3.7946, + "step": 23 + }, + { + "epoch": 0.0, + "learning_rate": 9.6579476861167e-07, + "loss": 3.8304, + "step": 24 + }, + { + "epoch": 0.0, + "learning_rate": 1.006036217303823e-06, + "loss": 3.6629, + "step": 25 + }, + { + "epoch": 0.0, + "learning_rate": 1.046277665995976e-06, + "loss": 3.5938, + "step": 26 + }, + { + "epoch": 0.0, + "learning_rate": 1.086519114688129e-06, + "loss": 3.5915, + "step": 27 + }, + { + "epoch": 0.0, + "learning_rate": 1.1267605633802817e-06, + "loss": 3.4397, + "step": 28 + }, + { + "epoch": 0.0, + "learning_rate": 1.1670020120724348e-06, + "loss": 3.0915, + "step": 29 + }, + { + "epoch": 0.0, + "learning_rate": 1.2072434607645876e-06, + "loss": 2.8058, + "step": 30 + }, + { + "epoch": 0.0, + "learning_rate": 1.2474849094567406e-06, + "loss": 2.6808, + "step": 31 + }, + { + "epoch": 0.0, + "learning_rate": 1.2877263581488936e-06, + "loss": 2.654, + "step": 32 + }, + { + "epoch": 0.0, + "learning_rate": 1.3279678068410465e-06, + "loss": 2.4866, + "step": 33 + }, + { + "epoch": 0.0, + "learning_rate": 1.3682092555331993e-06, + "loss": 2.4085, + "step": 34 + }, + { + "epoch": 0.0, + "learning_rate": 1.4084507042253523e-06, + "loss": 2.3638, + "step": 35 + }, + { + "epoch": 0.0, + "learning_rate": 1.4486921529175051e-06, + "loss": 2.2009, + "step": 36 + }, + { + "epoch": 0.0, + "learning_rate": 1.4889336016096582e-06, + "loss": 2.1607, + "step": 37 + }, + { + "epoch": 0.0, + "learning_rate": 1.529175050301811e-06, + "loss": 2.1161, + "step": 38 + }, + { + "epoch": 0.0, + "learning_rate": 1.5694164989939638e-06, + "loss": 2.0268, + "step": 39 + }, + { + "epoch": 0.0, + "learning_rate": 1.6096579476861169e-06, + "loss": 1.9275, + "step": 40 + }, + { + "epoch": 0.0, + "learning_rate": 1.6498993963782697e-06, + "loss": 1.9855, + "step": 41 + }, + { + "epoch": 0.0, + "learning_rate": 1.6901408450704227e-06, + "loss": 1.8058, + "step": 42 + }, + { + "epoch": 0.0, + "learning_rate": 1.7303822937625755e-06, + "loss": 1.7455, + "step": 43 + }, + { + "epoch": 0.0, + "learning_rate": 1.7706237424547286e-06, + "loss": 1.7388, + "step": 44 + }, + { + "epoch": 0.0, + "learning_rate": 1.8108651911468814e-06, + "loss": 1.6964, + "step": 45 + }, + { + "epoch": 0.0, + "learning_rate": 1.8511066398390342e-06, + "loss": 1.5692, + "step": 46 + }, + { + "epoch": 0.0, + "learning_rate": 1.8913480885311872e-06, + "loss": 1.4777, + "step": 47 + }, + { + "epoch": 0.0, + "learning_rate": 1.93158953722334e-06, + "loss": 1.5413, + "step": 48 + }, + { + "epoch": 0.0, + "learning_rate": 1.971830985915493e-06, + "loss": 1.4944, + "step": 49 + }, + { + "epoch": 0.0, + "learning_rate": 2.012072434607646e-06, + "loss": 1.5882, + "step": 50 + }, + { + "epoch": 0.0, + "learning_rate": 2.052313883299799e-06, + "loss": 1.4308, + "step": 51 + }, + { + "epoch": 0.0, + "learning_rate": 2.092555331991952e-06, + "loss": 1.3426, + "step": 52 + }, + { + "epoch": 0.0, + "learning_rate": 2.132796780684105e-06, + "loss": 1.3516, + "step": 53 + }, + { + "epoch": 0.0, + "learning_rate": 2.173038229376258e-06, + "loss": 1.3884, + "step": 54 + }, + { + "epoch": 0.0, + "learning_rate": 2.213279678068411e-06, + "loss": 1.3047, + "step": 55 + }, + { + "epoch": 0.0, + "learning_rate": 2.2535211267605635e-06, + "loss": 1.3147, + "step": 56 + }, + { + "epoch": 0.0, + "learning_rate": 2.2937625754527165e-06, + "loss": 1.2679, + "step": 57 + }, + { + "epoch": 0.0, + "learning_rate": 2.3340040241448695e-06, + "loss": 1.2567, + "step": 58 + }, + { + "epoch": 0.0, + "learning_rate": 2.374245472837022e-06, + "loss": 1.1373, + "step": 59 + }, + { + "epoch": 0.0, + "learning_rate": 2.414486921529175e-06, + "loss": 1.2388, + "step": 60 + }, + { + "epoch": 0.0, + "learning_rate": 2.454728370221328e-06, + "loss": 1.2042, + "step": 61 + }, + { + "epoch": 0.0, + "learning_rate": 2.4949698189134812e-06, + "loss": 1.1217, + "step": 62 + }, + { + "epoch": 0.0, + "learning_rate": 2.535211267605634e-06, + "loss": 1.1629, + "step": 63 + }, + { + "epoch": 0.0, + "learning_rate": 2.5754527162977873e-06, + "loss": 1.1462, + "step": 64 + }, + { + "epoch": 0.0, + "learning_rate": 2.61569416498994e-06, + "loss": 1.101, + "step": 65 + }, + { + "epoch": 0.0, + "learning_rate": 2.655935613682093e-06, + "loss": 1.1094, + "step": 66 + }, + { + "epoch": 0.0, + "learning_rate": 2.6961770623742455e-06, + "loss": 1.1228, + "step": 67 + }, + { + "epoch": 0.0, + "learning_rate": 2.7364185110663986e-06, + "loss": 1.0792, + "step": 68 + }, + { + "epoch": 0.0, + "learning_rate": 2.7766599597585516e-06, + "loss": 1.0826, + "step": 69 + }, + { + "epoch": 0.0, + "learning_rate": 2.8169014084507046e-06, + "loss": 1.0463, + "step": 70 + }, + { + "epoch": 0.0, + "learning_rate": 2.8571428571428573e-06, + "loss": 0.9565, + "step": 71 + }, + { + "epoch": 0.0, + "learning_rate": 2.8973843058350103e-06, + "loss": 1.0262, + "step": 72 + }, + { + "epoch": 0.0, + "learning_rate": 2.937625754527163e-06, + "loss": 1.0006, + "step": 73 + }, + { + "epoch": 0.0, + "learning_rate": 2.9778672032193164e-06, + "loss": 1.0033, + "step": 74 + }, + { + "epoch": 0.0, + "learning_rate": 3.018108651911469e-06, + "loss": 0.976, + "step": 75 + }, + { + "epoch": 0.0, + "learning_rate": 3.058350100603622e-06, + "loss": 0.9827, + "step": 76 + }, + { + "epoch": 0.0, + "learning_rate": 3.0985915492957746e-06, + "loss": 0.9453, + "step": 77 + }, + { + "epoch": 0.0, + "learning_rate": 3.1388329979879276e-06, + "loss": 0.9866, + "step": 78 + }, + { + "epoch": 0.0, + "learning_rate": 3.1790744466800807e-06, + "loss": 0.9794, + "step": 79 + }, + { + "epoch": 0.0, + "learning_rate": 3.2193158953722337e-06, + "loss": 0.9565, + "step": 80 + }, + { + "epoch": 0.0, + "learning_rate": 3.2595573440643867e-06, + "loss": 0.9732, + "step": 81 + }, + { + "epoch": 0.0, + "learning_rate": 3.2997987927565393e-06, + "loss": 0.9353, + "step": 82 + }, + { + "epoch": 0.01, + "learning_rate": 3.340040241448693e-06, + "loss": 1.0, + "step": 83 + }, + { + "epoch": 0.01, + "learning_rate": 3.3802816901408454e-06, + "loss": 0.9593, + "step": 84 + }, + { + "epoch": 0.01, + "learning_rate": 3.4205231388329984e-06, + "loss": 0.8722, + "step": 85 + }, + { + "epoch": 0.01, + "learning_rate": 3.460764587525151e-06, + "loss": 0.947, + "step": 86 + }, + { + "epoch": 0.01, + "learning_rate": 3.501006036217304e-06, + "loss": 0.9286, + "step": 87 + }, + { + "epoch": 0.01, + "learning_rate": 3.541247484909457e-06, + "loss": 0.9531, + "step": 88 + }, + { + "epoch": 0.01, + "learning_rate": 3.58148893360161e-06, + "loss": 0.8761, + "step": 89 + }, + { + "epoch": 0.01, + "learning_rate": 3.6217303822937628e-06, + "loss": 0.8694, + "step": 90 + }, + { + "epoch": 0.01, + "learning_rate": 3.6619718309859158e-06, + "loss": 0.8108, + "step": 91 + }, + { + "epoch": 0.01, + "learning_rate": 3.7022132796780684e-06, + "loss": 0.8761, + "step": 92 + }, + { + "epoch": 0.01, + "learning_rate": 3.742454728370222e-06, + "loss": 0.865, + "step": 93 + }, + { + "epoch": 0.01, + "learning_rate": 3.7826961770623745e-06, + "loss": 0.8901, + "step": 94 + }, + { + "epoch": 0.01, + "learning_rate": 3.8229376257545275e-06, + "loss": 0.846, + "step": 95 + }, + { + "epoch": 0.01, + "learning_rate": 3.86317907444668e-06, + "loss": 0.9375, + "step": 96 + }, + { + "epoch": 0.01, + "learning_rate": 3.9034205231388336e-06, + "loss": 0.8672, + "step": 97 + }, + { + "epoch": 0.01, + "learning_rate": 3.943661971830986e-06, + "loss": 0.8767, + "step": 98 + }, + { + "epoch": 0.01, + "learning_rate": 3.983903420523139e-06, + "loss": 0.8198, + "step": 99 + }, + { + "epoch": 0.01, + "learning_rate": 4.024144869215292e-06, + "loss": 0.7969, + "step": 100 + }, + { + "epoch": 0.01, + "learning_rate": 4.064386317907445e-06, + "loss": 0.8566, + "step": 101 + }, + { + "epoch": 0.01, + "learning_rate": 4.104627766599598e-06, + "loss": 0.899, + "step": 102 + }, + { + "epoch": 0.01, + "learning_rate": 4.144869215291751e-06, + "loss": 0.8583, + "step": 103 + }, + { + "epoch": 0.01, + "learning_rate": 4.185110663983904e-06, + "loss": 0.7974, + "step": 104 + }, + { + "epoch": 0.01, + "learning_rate": 4.225352112676057e-06, + "loss": 0.7824, + "step": 105 + }, + { + "epoch": 0.01, + "learning_rate": 4.26559356136821e-06, + "loss": 0.8192, + "step": 106 + }, + { + "epoch": 0.01, + "learning_rate": 4.305835010060362e-06, + "loss": 0.808, + "step": 107 + }, + { + "epoch": 0.01, + "learning_rate": 4.346076458752516e-06, + "loss": 0.8136, + "step": 108 + }, + { + "epoch": 0.01, + "learning_rate": 4.386317907444668e-06, + "loss": 0.7773, + "step": 109 + }, + { + "epoch": 0.01, + "learning_rate": 4.426559356136822e-06, + "loss": 0.8008, + "step": 110 + }, + { + "epoch": 0.01, + "learning_rate": 4.466800804828974e-06, + "loss": 0.7545, + "step": 111 + }, + { + "epoch": 0.01, + "learning_rate": 4.507042253521127e-06, + "loss": 0.7824, + "step": 112 + }, + { + "epoch": 0.01, + "learning_rate": 4.5472837022132795e-06, + "loss": 0.8504, + "step": 113 + }, + { + "epoch": 0.01, + "learning_rate": 4.587525150905433e-06, + "loss": 0.7729, + "step": 114 + }, + { + "epoch": 0.01, + "learning_rate": 4.627766599597586e-06, + "loss": 0.8086, + "step": 115 + }, + { + "epoch": 0.01, + "learning_rate": 4.668008048289739e-06, + "loss": 0.755, + "step": 116 + }, + { + "epoch": 0.01, + "learning_rate": 4.708249496981892e-06, + "loss": 0.7746, + "step": 117 + }, + { + "epoch": 0.01, + "learning_rate": 4.748490945674044e-06, + "loss": 0.7863, + "step": 118 + }, + { + "epoch": 0.01, + "learning_rate": 4.788732394366197e-06, + "loss": 0.8242, + "step": 119 + }, + { + "epoch": 0.01, + "learning_rate": 4.82897384305835e-06, + "loss": 0.7818, + "step": 120 + }, + { + "epoch": 0.01, + "learning_rate": 4.869215291750504e-06, + "loss": 0.7517, + "step": 121 + }, + { + "epoch": 0.01, + "learning_rate": 4.909456740442656e-06, + "loss": 0.774, + "step": 122 + }, + { + "epoch": 0.01, + "learning_rate": 4.94969818913481e-06, + "loss": 0.769, + "step": 123 + }, + { + "epoch": 0.01, + "learning_rate": 4.9899396378269625e-06, + "loss": 0.7338, + "step": 124 + }, + { + "epoch": 0.01, + "learning_rate": 5.030181086519115e-06, + "loss": 0.8002, + "step": 125 + }, + { + "epoch": 0.01, + "learning_rate": 5.070422535211268e-06, + "loss": 0.6881, + "step": 126 + }, + { + "epoch": 0.01, + "learning_rate": 5.110663983903421e-06, + "loss": 0.736, + "step": 127 + }, + { + "epoch": 0.01, + "learning_rate": 5.150905432595575e-06, + "loss": 0.7796, + "step": 128 + }, + { + "epoch": 0.01, + "learning_rate": 5.191146881287726e-06, + "loss": 0.7511, + "step": 129 + }, + { + "epoch": 0.01, + "learning_rate": 5.23138832997988e-06, + "loss": 0.6897, + "step": 130 + }, + { + "epoch": 0.01, + "learning_rate": 5.2716297786720324e-06, + "loss": 0.7617, + "step": 131 + }, + { + "epoch": 0.01, + "learning_rate": 5.311871227364186e-06, + "loss": 0.7182, + "step": 132 + }, + { + "epoch": 0.01, + "learning_rate": 5.352112676056338e-06, + "loss": 0.779, + "step": 133 + }, + { + "epoch": 0.01, + "learning_rate": 5.392354124748491e-06, + "loss": 0.7528, + "step": 134 + }, + { + "epoch": 0.01, + "learning_rate": 5.4325955734406446e-06, + "loss": 0.6652, + "step": 135 + }, + { + "epoch": 0.01, + "learning_rate": 5.472837022132797e-06, + "loss": 0.716, + "step": 136 + }, + { + "epoch": 0.01, + "learning_rate": 5.51307847082495e-06, + "loss": 0.707, + "step": 137 + }, + { + "epoch": 0.01, + "learning_rate": 5.553319919517103e-06, + "loss": 0.726, + "step": 138 + }, + { + "epoch": 0.01, + "learning_rate": 5.593561368209256e-06, + "loss": 0.7539, + "step": 139 + }, + { + "epoch": 0.01, + "learning_rate": 5.633802816901409e-06, + "loss": 0.6925, + "step": 140 + }, + { + "epoch": 0.01, + "learning_rate": 5.674044265593563e-06, + "loss": 0.7065, + "step": 141 + }, + { + "epoch": 0.01, + "learning_rate": 5.7142857142857145e-06, + "loss": 0.7109, + "step": 142 + }, + { + "epoch": 0.01, + "learning_rate": 5.754527162977868e-06, + "loss": 0.6886, + "step": 143 + }, + { + "epoch": 0.01, + "learning_rate": 5.7947686116700206e-06, + "loss": 0.7165, + "step": 144 + }, + { + "epoch": 0.01, + "learning_rate": 5.835010060362174e-06, + "loss": 0.7154, + "step": 145 + }, + { + "epoch": 0.01, + "learning_rate": 5.875251509054326e-06, + "loss": 0.6842, + "step": 146 + }, + { + "epoch": 0.01, + "learning_rate": 5.915492957746479e-06, + "loss": 0.7042, + "step": 147 + }, + { + "epoch": 0.01, + "learning_rate": 5.955734406438633e-06, + "loss": 0.6802, + "step": 148 + }, + { + "epoch": 0.01, + "learning_rate": 5.995975855130785e-06, + "loss": 0.7098, + "step": 149 + }, + { + "epoch": 0.01, + "learning_rate": 6.036217303822938e-06, + "loss": 0.6987, + "step": 150 + }, + { + "epoch": 0.01, + "learning_rate": 6.0764587525150905e-06, + "loss": 0.7154, + "step": 151 + }, + { + "epoch": 0.01, + "learning_rate": 6.116700201207244e-06, + "loss": 0.7093, + "step": 152 + }, + { + "epoch": 0.01, + "learning_rate": 6.1569416498993974e-06, + "loss": 0.6702, + "step": 153 + }, + { + "epoch": 0.01, + "learning_rate": 6.197183098591549e-06, + "loss": 0.6964, + "step": 154 + }, + { + "epoch": 0.01, + "learning_rate": 6.237424547283703e-06, + "loss": 0.7254, + "step": 155 + }, + { + "epoch": 0.01, + "learning_rate": 6.277665995975855e-06, + "loss": 0.7109, + "step": 156 + }, + { + "epoch": 0.01, + "learning_rate": 6.317907444668009e-06, + "loss": 0.6752, + "step": 157 + }, + { + "epoch": 0.01, + "learning_rate": 6.358148893360161e-06, + "loss": 0.7176, + "step": 158 + }, + { + "epoch": 0.01, + "learning_rate": 6.398390342052314e-06, + "loss": 0.6194, + "step": 159 + }, + { + "epoch": 0.01, + "learning_rate": 6.438631790744467e-06, + "loss": 0.7109, + "step": 160 + }, + { + "epoch": 0.01, + "learning_rate": 6.478873239436621e-06, + "loss": 0.6763, + "step": 161 + }, + { + "epoch": 0.01, + "learning_rate": 6.5191146881287735e-06, + "loss": 0.683, + "step": 162 + }, + { + "epoch": 0.01, + "learning_rate": 6.559356136820926e-06, + "loss": 0.639, + "step": 163 + }, + { + "epoch": 0.01, + "learning_rate": 6.599597585513079e-06, + "loss": 0.6897, + "step": 164 + }, + { + "epoch": 0.01, + "learning_rate": 6.639839034205232e-06, + "loss": 0.6161, + "step": 165 + }, + { + "epoch": 0.01, + "learning_rate": 6.680080482897386e-06, + "loss": 0.6925, + "step": 166 + }, + { + "epoch": 0.01, + "learning_rate": 6.720321931589537e-06, + "loss": 0.716, + "step": 167 + }, + { + "epoch": 0.01, + "learning_rate": 6.760563380281691e-06, + "loss": 0.6523, + "step": 168 + }, + { + "epoch": 0.01, + "learning_rate": 6.800804828973843e-06, + "loss": 0.6501, + "step": 169 + }, + { + "epoch": 0.01, + "learning_rate": 6.841046277665997e-06, + "loss": 0.6886, + "step": 170 + }, + { + "epoch": 0.01, + "learning_rate": 6.881287726358149e-06, + "loss": 0.6177, + "step": 171 + }, + { + "epoch": 0.01, + "learning_rate": 6.921529175050302e-06, + "loss": 0.6057, + "step": 172 + }, + { + "epoch": 0.01, + "learning_rate": 6.9617706237424556e-06, + "loss": 0.6624, + "step": 173 + }, + { + "epoch": 0.01, + "learning_rate": 7.002012072434608e-06, + "loss": 0.6727, + "step": 174 + }, + { + "epoch": 0.01, + "learning_rate": 7.042253521126761e-06, + "loss": 0.6392, + "step": 175 + }, + { + "epoch": 0.01, + "learning_rate": 7.082494969818914e-06, + "loss": 0.6278, + "step": 176 + }, + { + "epoch": 0.01, + "learning_rate": 7.122736418511067e-06, + "loss": 0.6283, + "step": 177 + }, + { + "epoch": 0.01, + "learning_rate": 7.16297786720322e-06, + "loss": 0.6401, + "step": 178 + }, + { + "epoch": 0.01, + "learning_rate": 7.203219315895374e-06, + "loss": 0.6334, + "step": 179 + }, + { + "epoch": 0.01, + "learning_rate": 7.2434607645875255e-06, + "loss": 0.5876, + "step": 180 + }, + { + "epoch": 0.01, + "learning_rate": 7.283702213279679e-06, + "loss": 0.6523, + "step": 181 + }, + { + "epoch": 0.01, + "learning_rate": 7.3239436619718316e-06, + "loss": 0.615, + "step": 182 + }, + { + "epoch": 0.01, + "learning_rate": 7.364185110663985e-06, + "loss": 0.6289, + "step": 183 + }, + { + "epoch": 0.01, + "learning_rate": 7.404426559356137e-06, + "loss": 0.6004, + "step": 184 + }, + { + "epoch": 0.01, + "learning_rate": 7.44466800804829e-06, + "loss": 0.644, + "step": 185 + }, + { + "epoch": 0.01, + "learning_rate": 7.484909456740444e-06, + "loss": 0.6741, + "step": 186 + }, + { + "epoch": 0.01, + "learning_rate": 7.525150905432596e-06, + "loss": 0.7015, + "step": 187 + }, + { + "epoch": 0.01, + "learning_rate": 7.565392354124749e-06, + "loss": 0.6367, + "step": 188 + }, + { + "epoch": 0.01, + "learning_rate": 7.6056338028169015e-06, + "loss": 0.63, + "step": 189 + }, + { + "epoch": 0.01, + "learning_rate": 7.645875251509055e-06, + "loss": 0.5898, + "step": 190 + }, + { + "epoch": 0.01, + "learning_rate": 7.686116700201208e-06, + "loss": 0.6328, + "step": 191 + }, + { + "epoch": 0.01, + "learning_rate": 7.72635814889336e-06, + "loss": 0.625, + "step": 192 + }, + { + "epoch": 0.01, + "learning_rate": 7.766599597585514e-06, + "loss": 0.6512, + "step": 193 + }, + { + "epoch": 0.01, + "learning_rate": 7.806841046277667e-06, + "loss": 0.6752, + "step": 194 + }, + { + "epoch": 0.01, + "learning_rate": 7.84708249496982e-06, + "loss": 0.6144, + "step": 195 + }, + { + "epoch": 0.01, + "learning_rate": 7.887323943661972e-06, + "loss": 0.6133, + "step": 196 + }, + { + "epoch": 0.01, + "learning_rate": 7.927565392354126e-06, + "loss": 0.6334, + "step": 197 + }, + { + "epoch": 0.01, + "learning_rate": 7.967806841046278e-06, + "loss": 0.6169, + "step": 198 + }, + { + "epoch": 0.01, + "learning_rate": 8.008048289738431e-06, + "loss": 0.6205, + "step": 199 + }, + { + "epoch": 0.01, + "learning_rate": 8.048289738430584e-06, + "loss": 0.5653, + "step": 200 + }, + { + "epoch": 0.01, + "learning_rate": 8.088531187122736e-06, + "loss": 0.6189, + "step": 201 + }, + { + "epoch": 0.01, + "learning_rate": 8.12877263581489e-06, + "loss": 0.6256, + "step": 202 + }, + { + "epoch": 0.01, + "learning_rate": 8.169014084507043e-06, + "loss": 0.6309, + "step": 203 + }, + { + "epoch": 0.01, + "learning_rate": 8.209255533199197e-06, + "loss": 0.6445, + "step": 204 + }, + { + "epoch": 0.01, + "learning_rate": 8.249496981891348e-06, + "loss": 0.6066, + "step": 205 + }, + { + "epoch": 0.01, + "learning_rate": 8.289738430583502e-06, + "loss": 0.5988, + "step": 206 + }, + { + "epoch": 0.01, + "learning_rate": 8.329979879275655e-06, + "loss": 0.6166, + "step": 207 + }, + { + "epoch": 0.01, + "learning_rate": 8.370221327967809e-06, + "loss": 0.606, + "step": 208 + }, + { + "epoch": 0.01, + "learning_rate": 8.41046277665996e-06, + "loss": 0.6004, + "step": 209 + }, + { + "epoch": 0.01, + "learning_rate": 8.450704225352114e-06, + "loss": 0.5926, + "step": 210 + }, + { + "epoch": 0.01, + "learning_rate": 8.490945674044266e-06, + "loss": 0.5742, + "step": 211 + }, + { + "epoch": 0.01, + "learning_rate": 8.53118712273642e-06, + "loss": 0.5871, + "step": 212 + }, + { + "epoch": 0.01, + "learning_rate": 8.571428571428571e-06, + "loss": 0.6306, + "step": 213 + }, + { + "epoch": 0.01, + "learning_rate": 8.611670020120724e-06, + "loss": 0.6384, + "step": 214 + }, + { + "epoch": 0.01, + "learning_rate": 8.651911468812878e-06, + "loss": 0.5871, + "step": 215 + }, + { + "epoch": 0.01, + "learning_rate": 8.692152917505031e-06, + "loss": 0.6311, + "step": 216 + }, + { + "epoch": 0.01, + "learning_rate": 8.732394366197183e-06, + "loss": 0.6016, + "step": 217 + }, + { + "epoch": 0.01, + "learning_rate": 8.772635814889337e-06, + "loss": 0.6362, + "step": 218 + }, + { + "epoch": 0.01, + "learning_rate": 8.81287726358149e-06, + "loss": 0.6016, + "step": 219 + }, + { + "epoch": 0.01, + "learning_rate": 8.853118712273643e-06, + "loss": 0.5949, + "step": 220 + }, + { + "epoch": 0.01, + "learning_rate": 8.893360160965795e-06, + "loss": 0.5876, + "step": 221 + }, + { + "epoch": 0.01, + "learning_rate": 8.933601609657949e-06, + "loss": 0.5603, + "step": 222 + }, + { + "epoch": 0.01, + "learning_rate": 8.973843058350102e-06, + "loss": 0.6189, + "step": 223 + }, + { + "epoch": 0.01, + "learning_rate": 9.014084507042254e-06, + "loss": 0.5795, + "step": 224 + }, + { + "epoch": 0.01, + "learning_rate": 9.054325955734407e-06, + "loss": 0.62, + "step": 225 + }, + { + "epoch": 0.01, + "learning_rate": 9.094567404426559e-06, + "loss": 0.5843, + "step": 226 + }, + { + "epoch": 0.01, + "learning_rate": 9.134808853118713e-06, + "loss": 0.6113, + "step": 227 + }, + { + "epoch": 0.01, + "learning_rate": 9.175050301810866e-06, + "loss": 0.6155, + "step": 228 + }, + { + "epoch": 0.01, + "learning_rate": 9.21529175050302e-06, + "loss": 0.6049, + "step": 229 + }, + { + "epoch": 0.01, + "learning_rate": 9.255533199195171e-06, + "loss": 0.6088, + "step": 230 + }, + { + "epoch": 0.01, + "learning_rate": 9.295774647887325e-06, + "loss": 0.5446, + "step": 231 + }, + { + "epoch": 0.01, + "learning_rate": 9.336016096579478e-06, + "loss": 0.5901, + "step": 232 + }, + { + "epoch": 0.01, + "learning_rate": 9.376257545271632e-06, + "loss": 0.5954, + "step": 233 + }, + { + "epoch": 0.01, + "learning_rate": 9.416498993963783e-06, + "loss": 0.5843, + "step": 234 + }, + { + "epoch": 0.01, + "learning_rate": 9.456740442655937e-06, + "loss": 0.6105, + "step": 235 + }, + { + "epoch": 0.01, + "learning_rate": 9.496981891348089e-06, + "loss": 0.582, + "step": 236 + }, + { + "epoch": 0.01, + "learning_rate": 9.537223340040242e-06, + "loss": 0.5421, + "step": 237 + }, + { + "epoch": 0.01, + "learning_rate": 9.577464788732394e-06, + "loss": 0.5896, + "step": 238 + }, + { + "epoch": 0.01, + "learning_rate": 9.617706237424547e-06, + "loss": 0.6228, + "step": 239 + }, + { + "epoch": 0.01, + "learning_rate": 9.6579476861167e-06, + "loss": 0.5625, + "step": 240 + }, + { + "epoch": 0.01, + "learning_rate": 9.698189134808854e-06, + "loss": 0.596, + "step": 241 + }, + { + "epoch": 0.01, + "learning_rate": 9.738430583501008e-06, + "loss": 0.5594, + "step": 242 + }, + { + "epoch": 0.01, + "learning_rate": 9.77867203219316e-06, + "loss": 0.5667, + "step": 243 + }, + { + "epoch": 0.01, + "learning_rate": 9.818913480885313e-06, + "loss": 0.6256, + "step": 244 + }, + { + "epoch": 0.01, + "learning_rate": 9.859154929577466e-06, + "loss": 0.5815, + "step": 245 + }, + { + "epoch": 0.01, + "learning_rate": 9.89939637826962e-06, + "loss": 0.5527, + "step": 246 + }, + { + "epoch": 0.01, + "learning_rate": 9.939637826961771e-06, + "loss": 0.6392, + "step": 247 + }, + { + "epoch": 0.01, + "learning_rate": 9.979879275653925e-06, + "loss": 0.5765, + "step": 248 + }, + { + "epoch": 0.02, + "learning_rate": 1.0020120724346077e-05, + "loss": 0.5943, + "step": 249 + }, + { + "epoch": 0.02, + "learning_rate": 1.006036217303823e-05, + "loss": 0.5564, + "step": 250 + }, + { + "epoch": 0.02, + "learning_rate": 1.0100603621730382e-05, + "loss": 0.5218, + "step": 251 + }, + { + "epoch": 0.02, + "learning_rate": 1.0140845070422535e-05, + "loss": 0.5658, + "step": 252 + }, + { + "epoch": 0.02, + "learning_rate": 1.0181086519114689e-05, + "loss": 0.601, + "step": 253 + }, + { + "epoch": 0.02, + "learning_rate": 1.0221327967806842e-05, + "loss": 0.5703, + "step": 254 + }, + { + "epoch": 0.02, + "learning_rate": 1.0261569416498996e-05, + "loss": 0.5787, + "step": 255 + }, + { + "epoch": 0.02, + "learning_rate": 1.030181086519115e-05, + "loss": 0.5285, + "step": 256 + }, + { + "epoch": 0.02, + "learning_rate": 1.03420523138833e-05, + "loss": 0.5647, + "step": 257 + }, + { + "epoch": 0.02, + "learning_rate": 1.0382293762575453e-05, + "loss": 0.5954, + "step": 258 + }, + { + "epoch": 0.02, + "learning_rate": 1.0422535211267606e-05, + "loss": 0.5391, + "step": 259 + }, + { + "epoch": 0.02, + "learning_rate": 1.046277665995976e-05, + "loss": 0.5642, + "step": 260 + }, + { + "epoch": 0.02, + "learning_rate": 1.0503018108651913e-05, + "loss": 0.5703, + "step": 261 + }, + { + "epoch": 0.02, + "learning_rate": 1.0543259557344065e-05, + "loss": 0.5698, + "step": 262 + }, + { + "epoch": 0.02, + "learning_rate": 1.0583501006036218e-05, + "loss": 0.5575, + "step": 263 + }, + { + "epoch": 0.02, + "learning_rate": 1.0623742454728372e-05, + "loss": 0.5474, + "step": 264 + }, + { + "epoch": 0.02, + "learning_rate": 1.0663983903420525e-05, + "loss": 0.5756, + "step": 265 + }, + { + "epoch": 0.02, + "learning_rate": 1.0704225352112675e-05, + "loss": 0.5511, + "step": 266 + }, + { + "epoch": 0.02, + "learning_rate": 1.0744466800804829e-05, + "loss": 0.5831, + "step": 267 + }, + { + "epoch": 0.02, + "learning_rate": 1.0784708249496982e-05, + "loss": 0.639, + "step": 268 + }, + { + "epoch": 0.02, + "learning_rate": 1.0824949698189136e-05, + "loss": 0.5352, + "step": 269 + }, + { + "epoch": 0.02, + "learning_rate": 1.0865191146881289e-05, + "loss": 0.5494, + "step": 270 + }, + { + "epoch": 0.02, + "learning_rate": 1.0905432595573443e-05, + "loss": 0.6256, + "step": 271 + }, + { + "epoch": 0.02, + "learning_rate": 1.0945674044265594e-05, + "loss": 0.5419, + "step": 272 + }, + { + "epoch": 0.02, + "learning_rate": 1.0985915492957748e-05, + "loss": 0.5405, + "step": 273 + }, + { + "epoch": 0.02, + "learning_rate": 1.10261569416499e-05, + "loss": 0.565, + "step": 274 + }, + { + "epoch": 0.02, + "learning_rate": 1.1066398390342053e-05, + "loss": 0.5664, + "step": 275 + }, + { + "epoch": 0.02, + "learning_rate": 1.1106639839034206e-05, + "loss": 0.5965, + "step": 276 + }, + { + "epoch": 0.02, + "learning_rate": 1.1146881287726358e-05, + "loss": 0.5179, + "step": 277 + }, + { + "epoch": 0.02, + "learning_rate": 1.1187122736418512e-05, + "loss": 0.548, + "step": 278 + }, + { + "epoch": 0.02, + "learning_rate": 1.1227364185110665e-05, + "loss": 0.5332, + "step": 279 + }, + { + "epoch": 0.02, + "learning_rate": 1.1267605633802819e-05, + "loss": 0.5709, + "step": 280 + }, + { + "epoch": 0.02, + "learning_rate": 1.1307847082494972e-05, + "loss": 0.6183, + "step": 281 + }, + { + "epoch": 0.02, + "learning_rate": 1.1348088531187125e-05, + "loss": 0.5552, + "step": 282 + }, + { + "epoch": 0.02, + "learning_rate": 1.1388329979879276e-05, + "loss": 0.5435, + "step": 283 + }, + { + "epoch": 0.02, + "learning_rate": 1.1428571428571429e-05, + "loss": 0.5965, + "step": 284 + }, + { + "epoch": 0.02, + "learning_rate": 1.1468812877263582e-05, + "loss": 0.5597, + "step": 285 + }, + { + "epoch": 0.02, + "learning_rate": 1.1509054325955736e-05, + "loss": 0.5435, + "step": 286 + }, + { + "epoch": 0.02, + "learning_rate": 1.1549295774647888e-05, + "loss": 0.5547, + "step": 287 + }, + { + "epoch": 0.02, + "learning_rate": 1.1589537223340041e-05, + "loss": 0.5494, + "step": 288 + }, + { + "epoch": 0.02, + "learning_rate": 1.1629778672032195e-05, + "loss": 0.5664, + "step": 289 + }, + { + "epoch": 0.02, + "learning_rate": 1.1670020120724348e-05, + "loss": 0.5765, + "step": 290 + }, + { + "epoch": 0.02, + "learning_rate": 1.1710261569416498e-05, + "loss": 0.5159, + "step": 291 + }, + { + "epoch": 0.02, + "learning_rate": 1.1750503018108652e-05, + "loss": 0.5304, + "step": 292 + }, + { + "epoch": 0.02, + "learning_rate": 1.1790744466800805e-05, + "loss": 0.5346, + "step": 293 + }, + { + "epoch": 0.02, + "learning_rate": 1.1830985915492958e-05, + "loss": 0.5812, + "step": 294 + }, + { + "epoch": 0.02, + "learning_rate": 1.1871227364185112e-05, + "loss": 0.5815, + "step": 295 + }, + { + "epoch": 0.02, + "learning_rate": 1.1911468812877265e-05, + "loss": 0.5566, + "step": 296 + }, + { + "epoch": 0.02, + "learning_rate": 1.1951710261569419e-05, + "loss": 0.5578, + "step": 297 + }, + { + "epoch": 0.02, + "learning_rate": 1.199195171026157e-05, + "loss": 0.5491, + "step": 298 + }, + { + "epoch": 0.02, + "learning_rate": 1.2032193158953724e-05, + "loss": 0.5299, + "step": 299 + }, + { + "epoch": 0.02, + "learning_rate": 1.2072434607645876e-05, + "loss": 0.546, + "step": 300 + }, + { + "epoch": 0.02, + "learning_rate": 1.211267605633803e-05, + "loss": 0.5547, + "step": 301 + }, + { + "epoch": 0.02, + "learning_rate": 1.2152917505030181e-05, + "loss": 0.5296, + "step": 302 + }, + { + "epoch": 0.02, + "learning_rate": 1.2193158953722335e-05, + "loss": 0.5458, + "step": 303 + }, + { + "epoch": 0.02, + "learning_rate": 1.2233400402414488e-05, + "loss": 0.5826, + "step": 304 + }, + { + "epoch": 0.02, + "learning_rate": 1.2273641851106641e-05, + "loss": 0.5265, + "step": 305 + }, + { + "epoch": 0.02, + "learning_rate": 1.2313883299798795e-05, + "loss": 0.5527, + "step": 306 + }, + { + "epoch": 0.02, + "learning_rate": 1.2354124748490948e-05, + "loss": 0.5282, + "step": 307 + }, + { + "epoch": 0.02, + "learning_rate": 1.2394366197183098e-05, + "loss": 0.5078, + "step": 308 + }, + { + "epoch": 0.02, + "learning_rate": 1.2434607645875252e-05, + "loss": 0.5287, + "step": 309 + }, + { + "epoch": 0.02, + "learning_rate": 1.2474849094567405e-05, + "loss": 0.5301, + "step": 310 + }, + { + "epoch": 0.02, + "learning_rate": 1.2515090543259559e-05, + "loss": 0.5114, + "step": 311 + }, + { + "epoch": 0.02, + "learning_rate": 1.255533199195171e-05, + "loss": 0.5522, + "step": 312 + }, + { + "epoch": 0.02, + "learning_rate": 1.2595573440643864e-05, + "loss": 0.5379, + "step": 313 + }, + { + "epoch": 0.02, + "learning_rate": 1.2635814889336017e-05, + "loss": 0.5279, + "step": 314 + }, + { + "epoch": 0.02, + "learning_rate": 1.2676056338028171e-05, + "loss": 0.534, + "step": 315 + }, + { + "epoch": 0.02, + "learning_rate": 1.2716297786720323e-05, + "loss": 0.4847, + "step": 316 + }, + { + "epoch": 0.02, + "learning_rate": 1.2756539235412474e-05, + "loss": 0.5466, + "step": 317 + }, + { + "epoch": 0.02, + "learning_rate": 1.2796780684104628e-05, + "loss": 0.5145, + "step": 318 + }, + { + "epoch": 0.02, + "learning_rate": 1.2837022132796781e-05, + "loss": 0.5661, + "step": 319 + }, + { + "epoch": 0.02, + "learning_rate": 1.2877263581488935e-05, + "loss": 0.5837, + "step": 320 + }, + { + "epoch": 0.02, + "learning_rate": 1.2917505030181088e-05, + "loss": 0.4986, + "step": 321 + }, + { + "epoch": 0.02, + "learning_rate": 1.2957746478873242e-05, + "loss": 0.5103, + "step": 322 + }, + { + "epoch": 0.02, + "learning_rate": 1.2997987927565393e-05, + "loss": 0.5257, + "step": 323 + }, + { + "epoch": 0.02, + "learning_rate": 1.3038229376257547e-05, + "loss": 0.5611, + "step": 324 + }, + { + "epoch": 0.02, + "learning_rate": 1.3078470824949699e-05, + "loss": 0.4824, + "step": 325 + }, + { + "epoch": 0.02, + "learning_rate": 1.3118712273641852e-05, + "loss": 0.488, + "step": 326 + }, + { + "epoch": 0.02, + "learning_rate": 1.3158953722334004e-05, + "loss": 0.4975, + "step": 327 + }, + { + "epoch": 0.02, + "learning_rate": 1.3199195171026157e-05, + "loss": 0.5212, + "step": 328 + }, + { + "epoch": 0.02, + "learning_rate": 1.323943661971831e-05, + "loss": 0.5603, + "step": 329 + }, + { + "epoch": 0.02, + "learning_rate": 1.3279678068410464e-05, + "loss": 0.5485, + "step": 330 + }, + { + "epoch": 0.02, + "learning_rate": 1.3319919517102618e-05, + "loss": 0.534, + "step": 331 + }, + { + "epoch": 0.02, + "learning_rate": 1.3360160965794771e-05, + "loss": 0.4877, + "step": 332 + }, + { + "epoch": 0.02, + "learning_rate": 1.3400402414486921e-05, + "loss": 0.498, + "step": 333 + }, + { + "epoch": 0.02, + "learning_rate": 1.3440643863179075e-05, + "loss": 0.5036, + "step": 334 + }, + { + "epoch": 0.02, + "learning_rate": 1.3480885311871228e-05, + "loss": 0.553, + "step": 335 + }, + { + "epoch": 0.02, + "learning_rate": 1.3521126760563382e-05, + "loss": 0.5022, + "step": 336 + }, + { + "epoch": 0.02, + "learning_rate": 1.3561368209255535e-05, + "loss": 0.4847, + "step": 337 + }, + { + "epoch": 0.02, + "learning_rate": 1.3601609657947687e-05, + "loss": 0.5226, + "step": 338 + }, + { + "epoch": 0.02, + "learning_rate": 1.364185110663984e-05, + "loss": 0.519, + "step": 339 + }, + { + "epoch": 0.02, + "learning_rate": 1.3682092555331994e-05, + "loss": 0.5684, + "step": 340 + }, + { + "epoch": 0.02, + "learning_rate": 1.3722334004024147e-05, + "loss": 0.5424, + "step": 341 + }, + { + "epoch": 0.02, + "learning_rate": 1.3762575452716297e-05, + "loss": 0.5326, + "step": 342 + }, + { + "epoch": 0.02, + "learning_rate": 1.380281690140845e-05, + "loss": 0.5246, + "step": 343 + }, + { + "epoch": 0.02, + "learning_rate": 1.3843058350100604e-05, + "loss": 0.5544, + "step": 344 + }, + { + "epoch": 0.02, + "learning_rate": 1.3883299798792758e-05, + "loss": 0.5561, + "step": 345 + }, + { + "epoch": 0.02, + "learning_rate": 1.3923541247484911e-05, + "loss": 0.495, + "step": 346 + }, + { + "epoch": 0.02, + "learning_rate": 1.3963782696177065e-05, + "loss": 0.5028, + "step": 347 + }, + { + "epoch": 0.02, + "learning_rate": 1.4004024144869216e-05, + "loss": 0.5324, + "step": 348 + }, + { + "epoch": 0.02, + "learning_rate": 1.404426559356137e-05, + "loss": 0.5022, + "step": 349 + }, + { + "epoch": 0.02, + "learning_rate": 1.4084507042253522e-05, + "loss": 0.5343, + "step": 350 + }, + { + "epoch": 0.02, + "learning_rate": 1.4124748490945675e-05, + "loss": 0.5483, + "step": 351 + }, + { + "epoch": 0.02, + "learning_rate": 1.4164989939637828e-05, + "loss": 0.5028, + "step": 352 + }, + { + "epoch": 0.02, + "learning_rate": 1.420523138832998e-05, + "loss": 0.5031, + "step": 353 + }, + { + "epoch": 0.02, + "learning_rate": 1.4245472837022134e-05, + "loss": 0.5106, + "step": 354 + }, + { + "epoch": 0.02, + "learning_rate": 1.4285714285714287e-05, + "loss": 0.5184, + "step": 355 + }, + { + "epoch": 0.02, + "learning_rate": 1.432595573440644e-05, + "loss": 0.4914, + "step": 356 + }, + { + "epoch": 0.02, + "learning_rate": 1.4366197183098594e-05, + "loss": 0.4986, + "step": 357 + }, + { + "epoch": 0.02, + "learning_rate": 1.4406438631790747e-05, + "loss": 0.5151, + "step": 358 + }, + { + "epoch": 0.02, + "learning_rate": 1.4446680080482898e-05, + "loss": 0.5223, + "step": 359 + }, + { + "epoch": 0.02, + "learning_rate": 1.4486921529175051e-05, + "loss": 0.5357, + "step": 360 + }, + { + "epoch": 0.02, + "learning_rate": 1.4527162977867204e-05, + "loss": 0.546, + "step": 361 + }, + { + "epoch": 0.02, + "learning_rate": 1.4567404426559358e-05, + "loss": 0.4986, + "step": 362 + }, + { + "epoch": 0.02, + "learning_rate": 1.460764587525151e-05, + "loss": 0.5388, + "step": 363 + }, + { + "epoch": 0.02, + "learning_rate": 1.4647887323943663e-05, + "loss": 0.5011, + "step": 364 + }, + { + "epoch": 0.02, + "learning_rate": 1.4688128772635817e-05, + "loss": 0.5156, + "step": 365 + }, + { + "epoch": 0.02, + "learning_rate": 1.472837022132797e-05, + "loss": 0.5271, + "step": 366 + }, + { + "epoch": 0.02, + "learning_rate": 1.476861167002012e-05, + "loss": 0.4916, + "step": 367 + }, + { + "epoch": 0.02, + "learning_rate": 1.4808853118712274e-05, + "loss": 0.4757, + "step": 368 + }, + { + "epoch": 0.02, + "learning_rate": 1.4849094567404427e-05, + "loss": 0.5223, + "step": 369 + }, + { + "epoch": 0.02, + "learning_rate": 1.488933601609658e-05, + "loss": 0.5006, + "step": 370 + }, + { + "epoch": 0.02, + "learning_rate": 1.4929577464788734e-05, + "loss": 0.5335, + "step": 371 + }, + { + "epoch": 0.02, + "learning_rate": 1.4969818913480887e-05, + "loss": 0.5491, + "step": 372 + }, + { + "epoch": 0.02, + "learning_rate": 1.5010060362173041e-05, + "loss": 0.536, + "step": 373 + }, + { + "epoch": 0.02, + "learning_rate": 1.5050301810865193e-05, + "loss": 0.5251, + "step": 374 + }, + { + "epoch": 0.02, + "learning_rate": 1.5090543259557344e-05, + "loss": 0.546, + "step": 375 + }, + { + "epoch": 0.02, + "learning_rate": 1.5130784708249498e-05, + "loss": 0.5017, + "step": 376 + }, + { + "epoch": 0.02, + "learning_rate": 1.5171026156941651e-05, + "loss": 0.5056, + "step": 377 + }, + { + "epoch": 0.02, + "learning_rate": 1.5211267605633803e-05, + "loss": 0.5162, + "step": 378 + }, + { + "epoch": 0.02, + "learning_rate": 1.5251509054325957e-05, + "loss": 0.5075, + "step": 379 + }, + { + "epoch": 0.02, + "learning_rate": 1.529175050301811e-05, + "loss": 0.5078, + "step": 380 + }, + { + "epoch": 0.02, + "learning_rate": 1.5331991951710262e-05, + "loss": 0.4944, + "step": 381 + }, + { + "epoch": 0.02, + "learning_rate": 1.5372233400402417e-05, + "loss": 0.5126, + "step": 382 + }, + { + "epoch": 0.02, + "learning_rate": 1.541247484909457e-05, + "loss": 0.5349, + "step": 383 + }, + { + "epoch": 0.02, + "learning_rate": 1.545271629778672e-05, + "loss": 0.4866, + "step": 384 + }, + { + "epoch": 0.02, + "learning_rate": 1.5492957746478872e-05, + "loss": 0.4877, + "step": 385 + }, + { + "epoch": 0.02, + "learning_rate": 1.5533199195171027e-05, + "loss": 0.4978, + "step": 386 + }, + { + "epoch": 0.02, + "learning_rate": 1.557344064386318e-05, + "loss": 0.4905, + "step": 387 + }, + { + "epoch": 0.02, + "learning_rate": 1.5613682092555334e-05, + "loss": 0.493, + "step": 388 + }, + { + "epoch": 0.02, + "learning_rate": 1.5653923541247486e-05, + "loss": 0.5028, + "step": 389 + }, + { + "epoch": 0.02, + "learning_rate": 1.569416498993964e-05, + "loss": 0.5165, + "step": 390 + }, + { + "epoch": 0.02, + "learning_rate": 1.5734406438631793e-05, + "loss": 0.5268, + "step": 391 + }, + { + "epoch": 0.02, + "learning_rate": 1.5774647887323945e-05, + "loss": 0.4916, + "step": 392 + }, + { + "epoch": 0.02, + "learning_rate": 1.5814889336016096e-05, + "loss": 0.5477, + "step": 393 + }, + { + "epoch": 0.02, + "learning_rate": 1.585513078470825e-05, + "loss": 0.5028, + "step": 394 + }, + { + "epoch": 0.02, + "learning_rate": 1.5895372233400403e-05, + "loss": 0.5594, + "step": 395 + }, + { + "epoch": 0.02, + "learning_rate": 1.5935613682092555e-05, + "loss": 0.5513, + "step": 396 + }, + { + "epoch": 0.02, + "learning_rate": 1.597585513078471e-05, + "loss": 0.4897, + "step": 397 + }, + { + "epoch": 0.02, + "learning_rate": 1.6016096579476862e-05, + "loss": 0.5318, + "step": 398 + }, + { + "epoch": 0.02, + "learning_rate": 1.6056338028169017e-05, + "loss": 0.5357, + "step": 399 + }, + { + "epoch": 0.02, + "learning_rate": 1.609657947686117e-05, + "loss": 0.5396, + "step": 400 + }, + { + "epoch": 0.02, + "learning_rate": 1.613682092555332e-05, + "loss": 0.4665, + "step": 401 + }, + { + "epoch": 0.02, + "learning_rate": 1.6177062374245472e-05, + "loss": 0.4916, + "step": 402 + }, + { + "epoch": 0.02, + "learning_rate": 1.6217303822937628e-05, + "loss": 0.5243, + "step": 403 + }, + { + "epoch": 0.02, + "learning_rate": 1.625754527162978e-05, + "loss": 0.4863, + "step": 404 + }, + { + "epoch": 0.02, + "learning_rate": 1.6297786720321935e-05, + "loss": 0.486, + "step": 405 + }, + { + "epoch": 0.02, + "learning_rate": 1.6338028169014086e-05, + "loss": 0.5128, + "step": 406 + }, + { + "epoch": 0.02, + "learning_rate": 1.6378269617706238e-05, + "loss": 0.4718, + "step": 407 + }, + { + "epoch": 0.02, + "learning_rate": 1.6418511066398393e-05, + "loss": 0.507, + "step": 408 + }, + { + "epoch": 0.02, + "learning_rate": 1.6458752515090545e-05, + "loss": 0.5151, + "step": 409 + }, + { + "epoch": 0.02, + "learning_rate": 1.6498993963782697e-05, + "loss": 0.5223, + "step": 410 + }, + { + "epoch": 0.02, + "learning_rate": 1.653923541247485e-05, + "loss": 0.5379, + "step": 411 + }, + { + "epoch": 0.02, + "learning_rate": 1.6579476861167004e-05, + "loss": 0.5036, + "step": 412 + }, + { + "epoch": 0.02, + "learning_rate": 1.6619718309859155e-05, + "loss": 0.5109, + "step": 413 + }, + { + "epoch": 0.03, + "learning_rate": 1.665995975855131e-05, + "loss": 0.5036, + "step": 414 + }, + { + "epoch": 0.03, + "learning_rate": 1.6700201207243462e-05, + "loss": 0.5243, + "step": 415 + }, + { + "epoch": 0.03, + "learning_rate": 1.6740442655935617e-05, + "loss": 0.5059, + "step": 416 + }, + { + "epoch": 0.03, + "learning_rate": 1.6780684104627766e-05, + "loss": 0.4897, + "step": 417 + }, + { + "epoch": 0.03, + "learning_rate": 1.682092555331992e-05, + "loss": 0.4551, + "step": 418 + }, + { + "epoch": 0.03, + "learning_rate": 1.6861167002012073e-05, + "loss": 0.5025, + "step": 419 + }, + { + "epoch": 0.03, + "learning_rate": 1.6901408450704228e-05, + "loss": 0.5006, + "step": 420 + }, + { + "epoch": 0.03, + "learning_rate": 1.694164989939638e-05, + "loss": 0.4819, + "step": 421 + }, + { + "epoch": 0.03, + "learning_rate": 1.698189134808853e-05, + "loss": 0.5131, + "step": 422 + }, + { + "epoch": 0.03, + "learning_rate": 1.7022132796780687e-05, + "loss": 0.493, + "step": 423 + }, + { + "epoch": 0.03, + "learning_rate": 1.706237424547284e-05, + "loss": 0.4662, + "step": 424 + }, + { + "epoch": 0.03, + "learning_rate": 1.7102615694164993e-05, + "loss": 0.5142, + "step": 425 + }, + { + "epoch": 0.03, + "learning_rate": 1.7142857142857142e-05, + "loss": 0.4763, + "step": 426 + }, + { + "epoch": 0.03, + "learning_rate": 1.7183098591549297e-05, + "loss": 0.5128, + "step": 427 + }, + { + "epoch": 0.03, + "learning_rate": 1.722334004024145e-05, + "loss": 0.486, + "step": 428 + }, + { + "epoch": 0.03, + "learning_rate": 1.7263581488933604e-05, + "loss": 0.5276, + "step": 429 + }, + { + "epoch": 0.03, + "learning_rate": 1.7303822937625756e-05, + "loss": 0.4986, + "step": 430 + }, + { + "epoch": 0.03, + "learning_rate": 1.734406438631791e-05, + "loss": 0.4888, + "step": 431 + }, + { + "epoch": 0.03, + "learning_rate": 1.7384305835010063e-05, + "loss": 0.4506, + "step": 432 + }, + { + "epoch": 0.03, + "learning_rate": 1.7424547283702214e-05, + "loss": 0.4844, + "step": 433 + }, + { + "epoch": 0.03, + "learning_rate": 1.7464788732394366e-05, + "loss": 0.4978, + "step": 434 + }, + { + "epoch": 0.03, + "learning_rate": 1.750503018108652e-05, + "loss": 0.495, + "step": 435 + }, + { + "epoch": 0.03, + "learning_rate": 1.7545271629778673e-05, + "loss": 0.5047, + "step": 436 + }, + { + "epoch": 0.03, + "learning_rate": 1.7585513078470825e-05, + "loss": 0.4855, + "step": 437 + }, + { + "epoch": 0.03, + "learning_rate": 1.762575452716298e-05, + "loss": 0.481, + "step": 438 + }, + { + "epoch": 0.03, + "learning_rate": 1.766599597585513e-05, + "loss": 0.4537, + "step": 439 + }, + { + "epoch": 0.03, + "learning_rate": 1.7706237424547287e-05, + "loss": 0.4565, + "step": 440 + }, + { + "epoch": 0.03, + "learning_rate": 1.774647887323944e-05, + "loss": 0.4386, + "step": 441 + }, + { + "epoch": 0.03, + "learning_rate": 1.778672032193159e-05, + "loss": 0.4807, + "step": 442 + }, + { + "epoch": 0.03, + "learning_rate": 1.7826961770623742e-05, + "loss": 0.519, + "step": 443 + }, + { + "epoch": 0.03, + "learning_rate": 1.7867203219315897e-05, + "loss": 0.5064, + "step": 444 + }, + { + "epoch": 0.03, + "learning_rate": 1.790744466800805e-05, + "loss": 0.4707, + "step": 445 + }, + { + "epoch": 0.03, + "learning_rate": 1.7947686116700204e-05, + "loss": 0.5259, + "step": 446 + }, + { + "epoch": 0.03, + "learning_rate": 1.7987927565392356e-05, + "loss": 0.4922, + "step": 447 + }, + { + "epoch": 0.03, + "learning_rate": 1.8028169014084508e-05, + "loss": 0.5056, + "step": 448 + }, + { + "epoch": 0.03, + "learning_rate": 1.8068410462776663e-05, + "loss": 0.5047, + "step": 449 + }, + { + "epoch": 0.03, + "learning_rate": 1.8108651911468815e-05, + "loss": 0.5039, + "step": 450 + }, + { + "epoch": 0.03, + "learning_rate": 1.8148893360160966e-05, + "loss": 0.5201, + "step": 451 + }, + { + "epoch": 0.03, + "learning_rate": 1.8189134808853118e-05, + "loss": 0.4693, + "step": 452 + }, + { + "epoch": 0.03, + "learning_rate": 1.8229376257545273e-05, + "loss": 0.4888, + "step": 453 + }, + { + "epoch": 0.03, + "learning_rate": 1.8269617706237425e-05, + "loss": 0.4872, + "step": 454 + }, + { + "epoch": 0.03, + "learning_rate": 1.830985915492958e-05, + "loss": 0.4914, + "step": 455 + }, + { + "epoch": 0.03, + "learning_rate": 1.8350100603621732e-05, + "loss": 0.5142, + "step": 456 + }, + { + "epoch": 0.03, + "learning_rate": 1.8390342052313884e-05, + "loss": 0.4621, + "step": 457 + }, + { + "epoch": 0.03, + "learning_rate": 1.843058350100604e-05, + "loss": 0.4964, + "step": 458 + }, + { + "epoch": 0.03, + "learning_rate": 1.847082494969819e-05, + "loss": 0.4941, + "step": 459 + }, + { + "epoch": 0.03, + "learning_rate": 1.8511066398390342e-05, + "loss": 0.4821, + "step": 460 + }, + { + "epoch": 0.03, + "learning_rate": 1.8551307847082498e-05, + "loss": 0.4888, + "step": 461 + }, + { + "epoch": 0.03, + "learning_rate": 1.859154929577465e-05, + "loss": 0.495, + "step": 462 + }, + { + "epoch": 0.03, + "learning_rate": 1.86317907444668e-05, + "loss": 0.4947, + "step": 463 + }, + { + "epoch": 0.03, + "learning_rate": 1.8672032193158956e-05, + "loss": 0.4975, + "step": 464 + }, + { + "epoch": 0.03, + "learning_rate": 1.8712273641851108e-05, + "loss": 0.4914, + "step": 465 + }, + { + "epoch": 0.03, + "learning_rate": 1.8752515090543263e-05, + "loss": 0.5025, + "step": 466 + }, + { + "epoch": 0.03, + "learning_rate": 1.8792756539235415e-05, + "loss": 0.5075, + "step": 467 + }, + { + "epoch": 0.03, + "learning_rate": 1.8832997987927567e-05, + "loss": 0.5017, + "step": 468 + }, + { + "epoch": 0.03, + "learning_rate": 1.887323943661972e-05, + "loss": 0.4905, + "step": 469 + }, + { + "epoch": 0.03, + "learning_rate": 1.8913480885311874e-05, + "loss": 0.4788, + "step": 470 + }, + { + "epoch": 0.03, + "learning_rate": 1.8953722334004025e-05, + "loss": 0.447, + "step": 471 + }, + { + "epoch": 0.03, + "learning_rate": 1.8993963782696177e-05, + "loss": 0.4827, + "step": 472 + }, + { + "epoch": 0.03, + "learning_rate": 1.9034205231388332e-05, + "loss": 0.4735, + "step": 473 + }, + { + "epoch": 0.03, + "learning_rate": 1.9074446680080484e-05, + "loss": 0.4735, + "step": 474 + }, + { + "epoch": 0.03, + "learning_rate": 1.911468812877264e-05, + "loss": 0.5114, + "step": 475 + }, + { + "epoch": 0.03, + "learning_rate": 1.9154929577464788e-05, + "loss": 0.4637, + "step": 476 + }, + { + "epoch": 0.03, + "learning_rate": 1.9195171026156943e-05, + "loss": 0.4835, + "step": 477 + }, + { + "epoch": 0.03, + "learning_rate": 1.9235412474849094e-05, + "loss": 0.4816, + "step": 478 + }, + { + "epoch": 0.03, + "learning_rate": 1.927565392354125e-05, + "loss": 0.4632, + "step": 479 + }, + { + "epoch": 0.03, + "learning_rate": 1.93158953722334e-05, + "loss": 0.5123, + "step": 480 + }, + { + "epoch": 0.03, + "learning_rate": 1.9356136820925557e-05, + "loss": 0.4693, + "step": 481 + }, + { + "epoch": 0.03, + "learning_rate": 1.9396378269617708e-05, + "loss": 0.5089, + "step": 482 + }, + { + "epoch": 0.03, + "learning_rate": 1.943661971830986e-05, + "loss": 0.5407, + "step": 483 + }, + { + "epoch": 0.03, + "learning_rate": 1.9476861167002015e-05, + "loss": 0.4835, + "step": 484 + }, + { + "epoch": 0.03, + "learning_rate": 1.9517102615694167e-05, + "loss": 0.5117, + "step": 485 + }, + { + "epoch": 0.03, + "learning_rate": 1.955734406438632e-05, + "loss": 0.4662, + "step": 486 + }, + { + "epoch": 0.03, + "learning_rate": 1.959758551307847e-05, + "loss": 0.4925, + "step": 487 + }, + { + "epoch": 0.03, + "learning_rate": 1.9637826961770626e-05, + "loss": 0.4746, + "step": 488 + }, + { + "epoch": 0.03, + "learning_rate": 1.9678068410462777e-05, + "loss": 0.4612, + "step": 489 + }, + { + "epoch": 0.03, + "learning_rate": 1.9718309859154933e-05, + "loss": 0.4403, + "step": 490 + }, + { + "epoch": 0.03, + "learning_rate": 1.9758551307847084e-05, + "loss": 0.4894, + "step": 491 + }, + { + "epoch": 0.03, + "learning_rate": 1.979879275653924e-05, + "loss": 0.4936, + "step": 492 + }, + { + "epoch": 0.03, + "learning_rate": 1.9839034205231388e-05, + "loss": 0.4869, + "step": 493 + }, + { + "epoch": 0.03, + "learning_rate": 1.9879275653923543e-05, + "loss": 0.4632, + "step": 494 + }, + { + "epoch": 0.03, + "learning_rate": 1.9919517102615695e-05, + "loss": 0.5209, + "step": 495 + }, + { + "epoch": 0.03, + "learning_rate": 1.995975855130785e-05, + "loss": 0.4461, + "step": 496 + }, + { + "epoch": 0.03, + "learning_rate": 2e-05, + "loss": 0.5405, + "step": 497 + }, + { + "epoch": 0.03, + "learning_rate": 1.99999998085766e-05, + "loss": 0.488, + "step": 498 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999999234306405e-05, + "loss": 0.4738, + "step": 499 + }, + { + "epoch": 0.03, + "learning_rate": 1.999999827718943e-05, + "loss": 0.4657, + "step": 500 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999996937225724e-05, + "loss": 0.507, + "step": 501 + }, + { + "epoch": 0.03, + "learning_rate": 1.999999521441533e-05, + "loss": 0.5028, + "step": 502 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999993108758315e-05, + "loss": 0.4665, + "step": 503 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999990620254757e-05, + "loss": 0.4905, + "step": 504 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999987748904762e-05, + "loss": 0.4874, + "step": 505 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999984494708432e-05, + "loss": 0.4729, + "step": 506 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999980857665888e-05, + "loss": 0.4361, + "step": 507 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999976837777275e-05, + "loss": 0.4805, + "step": 508 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999972435042747e-05, + "loss": 0.44, + "step": 509 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999967649462473e-05, + "loss": 0.4696, + "step": 510 + }, + { + "epoch": 0.03, + "learning_rate": 1.999996248103663e-05, + "loss": 0.4863, + "step": 511 + }, + { + "epoch": 0.03, + "learning_rate": 1.999995692976542e-05, + "loss": 0.466, + "step": 512 + }, + { + "epoch": 0.03, + "learning_rate": 1.999995099564906e-05, + "loss": 0.459, + "step": 513 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999944678687773e-05, + "loss": 0.476, + "step": 514 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999937978881797e-05, + "loss": 0.4939, + "step": 515 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999930896231396e-05, + "loss": 0.5279, + "step": 516 + }, + { + "epoch": 0.03, + "learning_rate": 1.999992343073683e-05, + "loss": 0.4371, + "step": 517 + }, + { + "epoch": 0.03, + "learning_rate": 1.99999155823984e-05, + "loss": 0.4838, + "step": 518 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999907351216398e-05, + "loss": 0.4629, + "step": 519 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999898737191137e-05, + "loss": 0.4545, + "step": 520 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999889740322952e-05, + "loss": 0.4824, + "step": 521 + }, + { + "epoch": 0.03, + "learning_rate": 1.999988036061218e-05, + "loss": 0.4732, + "step": 522 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999870598059192e-05, + "loss": 0.4858, + "step": 523 + }, + { + "epoch": 0.03, + "learning_rate": 1.999986045266435e-05, + "loss": 0.4339, + "step": 524 + }, + { + "epoch": 0.03, + "learning_rate": 1.999984992442805e-05, + "loss": 0.483, + "step": 525 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999839013350694e-05, + "loss": 0.4883, + "step": 526 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999827719432694e-05, + "loss": 0.512, + "step": 527 + }, + { + "epoch": 0.03, + "learning_rate": 1.999981604267449e-05, + "loss": 0.4604, + "step": 528 + }, + { + "epoch": 0.03, + "learning_rate": 1.999980398307652e-05, + "loss": 0.4784, + "step": 529 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999791540639256e-05, + "loss": 0.4774, + "step": 530 + }, + { + "epoch": 0.03, + "learning_rate": 1.999977871536317e-05, + "loss": 0.5346, + "step": 531 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999765507248752e-05, + "loss": 0.4604, + "step": 532 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999751916296506e-05, + "loss": 0.4595, + "step": 533 + }, + { + "epoch": 0.03, + "learning_rate": 1.999973794250696e-05, + "loss": 0.493, + "step": 534 + }, + { + "epoch": 0.03, + "learning_rate": 1.999972358588064e-05, + "loss": 0.4713, + "step": 535 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999708846418096e-05, + "loss": 0.4905, + "step": 536 + }, + { + "epoch": 0.03, + "learning_rate": 1.99996937241199e-05, + "loss": 0.4699, + "step": 537 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999678218986624e-05, + "loss": 0.4858, + "step": 538 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999662331018867e-05, + "loss": 0.4584, + "step": 539 + }, + { + "epoch": 0.03, + "learning_rate": 1.999964606021723e-05, + "loss": 0.4735, + "step": 540 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999629406582344e-05, + "loss": 0.4727, + "step": 541 + }, + { + "epoch": 0.03, + "learning_rate": 1.999961237011484e-05, + "loss": 0.5039, + "step": 542 + }, + { + "epoch": 0.03, + "learning_rate": 1.999959495081538e-05, + "loss": 0.4746, + "step": 543 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999577148684614e-05, + "loss": 0.5042, + "step": 544 + }, + { + "epoch": 0.03, + "learning_rate": 1.999955896372324e-05, + "loss": 0.486, + "step": 545 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999540395931944e-05, + "loss": 0.5106, + "step": 546 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999521445311446e-05, + "loss": 0.5, + "step": 547 + }, + { + "epoch": 0.03, + "learning_rate": 1.999950211186246e-05, + "loss": 0.4914, + "step": 548 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999482395585733e-05, + "loss": 0.4718, + "step": 549 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999462296482023e-05, + "loss": 0.486, + "step": 550 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999441814552092e-05, + "loss": 0.4838, + "step": 551 + }, + { + "epoch": 0.03, + "learning_rate": 1.999942094979673e-05, + "loss": 0.4255, + "step": 552 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999399702216735e-05, + "loss": 0.5117, + "step": 553 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999378071812917e-05, + "loss": 0.4914, + "step": 554 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999356058586104e-05, + "loss": 0.4576, + "step": 555 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999333662537147e-05, + "loss": 0.4721, + "step": 556 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999310883666893e-05, + "loss": 0.5424, + "step": 557 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999287721976218e-05, + "loss": 0.5212, + "step": 558 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999264177466008e-05, + "loss": 0.4729, + "step": 559 + }, + { + "epoch": 0.03, + "learning_rate": 1.999924025013717e-05, + "loss": 0.5223, + "step": 560 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999215939990613e-05, + "loss": 0.4727, + "step": 561 + }, + { + "epoch": 0.03, + "learning_rate": 1.999919124702727e-05, + "loss": 0.4933, + "step": 562 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999166171248087e-05, + "loss": 0.4782, + "step": 563 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999140712654025e-05, + "loss": 0.4688, + "step": 564 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999114871246056e-05, + "loss": 0.4805, + "step": 565 + }, + { + "epoch": 0.03, + "learning_rate": 1.999908864702517e-05, + "loss": 0.4634, + "step": 566 + }, + { + "epoch": 0.03, + "learning_rate": 1.999906203999237e-05, + "loss": 0.4528, + "step": 567 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999035050148682e-05, + "loss": 0.4598, + "step": 568 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999007677495127e-05, + "loss": 0.4978, + "step": 569 + }, + { + "epoch": 0.03, + "learning_rate": 1.9998979922032763e-05, + "loss": 0.4847, + "step": 570 + }, + { + "epoch": 0.03, + "learning_rate": 1.999895178376265e-05, + "loss": 0.4813, + "step": 571 + }, + { + "epoch": 0.03, + "learning_rate": 1.9998923262685865e-05, + "loss": 0.4953, + "step": 572 + }, + { + "epoch": 0.03, + "learning_rate": 1.9998894358803498e-05, + "loss": 0.4286, + "step": 573 + }, + { + "epoch": 0.03, + "learning_rate": 1.9998865072116656e-05, + "loss": 0.4685, + "step": 574 + }, + { + "epoch": 0.03, + "learning_rate": 1.999883540262646e-05, + "loss": 0.4894, + "step": 575 + }, + { + "epoch": 0.03, + "learning_rate": 1.9998805350334053e-05, + "loss": 0.4579, + "step": 576 + }, + { + "epoch": 0.03, + "learning_rate": 1.999877491524057e-05, + "loss": 0.4509, + "step": 577 + }, + { + "epoch": 0.03, + "learning_rate": 1.9998744097347195e-05, + "loss": 0.4674, + "step": 578 + }, + { + "epoch": 0.03, + "learning_rate": 1.9998712896655097e-05, + "loss": 0.4774, + "step": 579 + }, + { + "epoch": 0.04, + "learning_rate": 1.999868131316547e-05, + "loss": 0.43, + "step": 580 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998649346879524e-05, + "loss": 0.4531, + "step": 581 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998616997798485e-05, + "loss": 0.4459, + "step": 582 + }, + { + "epoch": 0.04, + "learning_rate": 1.999858426592359e-05, + "loss": 0.5098, + "step": 583 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998551151256094e-05, + "loss": 0.469, + "step": 584 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998517653797266e-05, + "loss": 0.4844, + "step": 585 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998483773548382e-05, + "loss": 0.4598, + "step": 586 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998449510510744e-05, + "loss": 0.4406, + "step": 587 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998414864685666e-05, + "loss": 0.4581, + "step": 588 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998379836074466e-05, + "loss": 0.488, + "step": 589 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998344424678496e-05, + "loss": 0.4802, + "step": 590 + }, + { + "epoch": 0.04, + "learning_rate": 1.99983086304991e-05, + "loss": 0.4679, + "step": 591 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998272453537657e-05, + "loss": 0.4693, + "step": 592 + }, + { + "epoch": 0.04, + "learning_rate": 1.999823589379555e-05, + "loss": 0.4877, + "step": 593 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998198951274177e-05, + "loss": 0.4707, + "step": 594 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998161625974954e-05, + "loss": 0.4738, + "step": 595 + }, + { + "epoch": 0.04, + "learning_rate": 1.999812391789931e-05, + "loss": 0.4668, + "step": 596 + }, + { + "epoch": 0.04, + "learning_rate": 1.999808582704869e-05, + "loss": 0.5109, + "step": 597 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998047353424544e-05, + "loss": 0.49, + "step": 598 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998008497028355e-05, + "loss": 0.4782, + "step": 599 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997969257861606e-05, + "loss": 0.4679, + "step": 600 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997929635925804e-05, + "loss": 0.4593, + "step": 601 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997889631222458e-05, + "loss": 0.4528, + "step": 602 + }, + { + "epoch": 0.04, + "learning_rate": 1.99978492437531e-05, + "loss": 0.4467, + "step": 603 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997808473519287e-05, + "loss": 0.445, + "step": 604 + }, + { + "epoch": 0.04, + "learning_rate": 1.999776732052257e-05, + "loss": 0.4816, + "step": 605 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997725784764528e-05, + "loss": 0.4676, + "step": 606 + }, + { + "epoch": 0.04, + "learning_rate": 1.999768386624675e-05, + "loss": 0.4417, + "step": 607 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997641564970842e-05, + "loss": 0.4459, + "step": 608 + }, + { + "epoch": 0.04, + "learning_rate": 1.999759888093842e-05, + "loss": 0.4735, + "step": 609 + }, + { + "epoch": 0.04, + "learning_rate": 1.999755581415112e-05, + "loss": 0.4763, + "step": 610 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997512364610596e-05, + "loss": 0.4997, + "step": 611 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997468532318508e-05, + "loss": 0.4754, + "step": 612 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997424317276528e-05, + "loss": 0.4344, + "step": 613 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997379719486355e-05, + "loss": 0.4551, + "step": 614 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997334738949702e-05, + "loss": 0.4855, + "step": 615 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997289375668276e-05, + "loss": 0.4523, + "step": 616 + }, + { + "epoch": 0.04, + "learning_rate": 1.999724362964383e-05, + "loss": 0.4972, + "step": 617 + }, + { + "epoch": 0.04, + "learning_rate": 1.99971975008781e-05, + "loss": 0.4417, + "step": 618 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997150989372862e-05, + "loss": 0.4347, + "step": 619 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997104095129897e-05, + "loss": 0.4718, + "step": 620 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997056818150992e-05, + "loss": 0.4718, + "step": 621 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997009158437967e-05, + "loss": 0.4559, + "step": 622 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996961115992643e-05, + "loss": 0.4894, + "step": 623 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996912690816857e-05, + "loss": 0.464, + "step": 624 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996863882912464e-05, + "loss": 0.4654, + "step": 625 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996814692281332e-05, + "loss": 0.4729, + "step": 626 + }, + { + "epoch": 0.04, + "learning_rate": 1.999676511892535e-05, + "loss": 0.4701, + "step": 627 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996715162846406e-05, + "loss": 0.5059, + "step": 628 + }, + { + "epoch": 0.04, + "learning_rate": 1.999666482404642e-05, + "loss": 0.4816, + "step": 629 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996614102527317e-05, + "loss": 0.4629, + "step": 630 + }, + { + "epoch": 0.04, + "learning_rate": 1.999656299829104e-05, + "loss": 0.4643, + "step": 631 + }, + { + "epoch": 0.04, + "learning_rate": 1.999651151133954e-05, + "loss": 0.4763, + "step": 632 + }, + { + "epoch": 0.04, + "learning_rate": 1.99964596416748e-05, + "loss": 0.4643, + "step": 633 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996407389298792e-05, + "loss": 0.4947, + "step": 634 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996354754213523e-05, + "loss": 0.4526, + "step": 635 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996301736421008e-05, + "loss": 0.4648, + "step": 636 + }, + { + "epoch": 0.04, + "learning_rate": 1.999624833592328e-05, + "loss": 0.454, + "step": 637 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996194552722378e-05, + "loss": 0.4556, + "step": 638 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996140386820363e-05, + "loss": 0.4727, + "step": 639 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996085838219307e-05, + "loss": 0.4238, + "step": 640 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996030906921302e-05, + "loss": 0.4255, + "step": 641 + }, + { + "epoch": 0.04, + "learning_rate": 1.999597559292845e-05, + "loss": 0.5025, + "step": 642 + }, + { + "epoch": 0.04, + "learning_rate": 1.999591989624287e-05, + "loss": 0.4152, + "step": 643 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995863816866687e-05, + "loss": 0.4877, + "step": 644 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995807354802056e-05, + "loss": 0.4484, + "step": 645 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995750510051137e-05, + "loss": 0.4752, + "step": 646 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995693282616104e-05, + "loss": 0.476, + "step": 647 + }, + { + "epoch": 0.04, + "learning_rate": 1.999563567249915e-05, + "loss": 0.471, + "step": 648 + }, + { + "epoch": 0.04, + "learning_rate": 1.999557767970248e-05, + "loss": 0.4964, + "step": 649 + }, + { + "epoch": 0.04, + "learning_rate": 1.999551930422831e-05, + "loss": 0.4255, + "step": 650 + }, + { + "epoch": 0.04, + "learning_rate": 1.999546054607888e-05, + "loss": 0.4428, + "step": 651 + }, + { + "epoch": 0.04, + "learning_rate": 1.999540140525644e-05, + "loss": 0.498, + "step": 652 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995341881763253e-05, + "loss": 0.4408, + "step": 653 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995281975601598e-05, + "loss": 0.4676, + "step": 654 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995221686773763e-05, + "loss": 0.4902, + "step": 655 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995161015282068e-05, + "loss": 0.4506, + "step": 656 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995099961128822e-05, + "loss": 0.4459, + "step": 657 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995038524316377e-05, + "loss": 0.4252, + "step": 658 + }, + { + "epoch": 0.04, + "learning_rate": 1.999497670484707e-05, + "loss": 0.4244, + "step": 659 + }, + { + "epoch": 0.04, + "learning_rate": 1.999491450272328e-05, + "loss": 0.4389, + "step": 660 + }, + { + "epoch": 0.04, + "learning_rate": 1.999485191794738e-05, + "loss": 0.4807, + "step": 661 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994788950521773e-05, + "loss": 0.4461, + "step": 662 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994725600448865e-05, + "loss": 0.4408, + "step": 663 + }, + { + "epoch": 0.04, + "learning_rate": 1.999466186773108e-05, + "loss": 0.4146, + "step": 664 + }, + { + "epoch": 0.04, + "learning_rate": 1.999459775237086e-05, + "loss": 0.4665, + "step": 665 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994533254370665e-05, + "loss": 0.4442, + "step": 666 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994468373732955e-05, + "loss": 0.4788, + "step": 667 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994403110460223e-05, + "loss": 0.4551, + "step": 668 + }, + { + "epoch": 0.04, + "learning_rate": 1.999433746455496e-05, + "loss": 0.4632, + "step": 669 + }, + { + "epoch": 0.04, + "learning_rate": 1.999427143601968e-05, + "loss": 0.4821, + "step": 670 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994205024856916e-05, + "loss": 0.4361, + "step": 671 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994138231069205e-05, + "loss": 0.4069, + "step": 672 + }, + { + "epoch": 0.04, + "learning_rate": 1.999407105465911e-05, + "loss": 0.442, + "step": 673 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994003495629196e-05, + "loss": 0.4439, + "step": 674 + }, + { + "epoch": 0.04, + "learning_rate": 1.999393555398206e-05, + "loss": 0.4849, + "step": 675 + }, + { + "epoch": 0.04, + "learning_rate": 1.999386722972029e-05, + "loss": 0.4771, + "step": 676 + }, + { + "epoch": 0.04, + "learning_rate": 1.999379852284651e-05, + "loss": 0.442, + "step": 677 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993729433363348e-05, + "loss": 0.4425, + "step": 678 + }, + { + "epoch": 0.04, + "learning_rate": 1.999365996127345e-05, + "loss": 0.4367, + "step": 679 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993590106579473e-05, + "loss": 0.4227, + "step": 680 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993519869284092e-05, + "loss": 0.4629, + "step": 681 + }, + { + "epoch": 0.04, + "learning_rate": 1.999344924939e-05, + "loss": 0.4657, + "step": 682 + }, + { + "epoch": 0.04, + "learning_rate": 1.99933782468999e-05, + "loss": 0.4601, + "step": 683 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993306861816508e-05, + "loss": 0.4771, + "step": 684 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993235094142555e-05, + "loss": 0.4403, + "step": 685 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993162943880793e-05, + "loss": 0.459, + "step": 686 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993090411033982e-05, + "loss": 0.418, + "step": 687 + }, + { + "epoch": 0.04, + "learning_rate": 1.99930174956049e-05, + "loss": 0.4481, + "step": 688 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992944197596337e-05, + "loss": 0.4372, + "step": 689 + }, + { + "epoch": 0.04, + "learning_rate": 1.99928705170111e-05, + "loss": 0.3951, + "step": 690 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992796453852012e-05, + "loss": 0.4355, + "step": 691 + }, + { + "epoch": 0.04, + "learning_rate": 1.99927220081219e-05, + "loss": 0.4208, + "step": 692 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992647179823625e-05, + "loss": 0.4118, + "step": 693 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992571968960046e-05, + "loss": 0.4802, + "step": 694 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992496375534044e-05, + "loss": 0.4579, + "step": 695 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992420399548514e-05, + "loss": 0.4344, + "step": 696 + }, + { + "epoch": 0.04, + "learning_rate": 1.999234404100636e-05, + "loss": 0.4559, + "step": 697 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992267299910512e-05, + "loss": 0.445, + "step": 698 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992190176263905e-05, + "loss": 0.4386, + "step": 699 + }, + { + "epoch": 0.04, + "learning_rate": 1.999211267006949e-05, + "loss": 0.4805, + "step": 700 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992034781330235e-05, + "loss": 0.4554, + "step": 701 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991956510049123e-05, + "loss": 0.4925, + "step": 702 + }, + { + "epoch": 0.04, + "learning_rate": 1.999187785622915e-05, + "loss": 0.4615, + "step": 703 + }, + { + "epoch": 0.04, + "learning_rate": 1.999179881987333e-05, + "loss": 0.4562, + "step": 704 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991719400984685e-05, + "loss": 0.4425, + "step": 705 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991639599566258e-05, + "loss": 0.4459, + "step": 706 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991559415621098e-05, + "loss": 0.4475, + "step": 707 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991478849152284e-05, + "loss": 0.4676, + "step": 708 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991397900162897e-05, + "loss": 0.4699, + "step": 709 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991316568656033e-05, + "loss": 0.4858, + "step": 710 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991234854634813e-05, + "loss": 0.478, + "step": 711 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991152758102353e-05, + "loss": 0.4191, + "step": 712 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991070279061808e-05, + "loss": 0.4629, + "step": 713 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990987417516333e-05, + "loss": 0.4439, + "step": 714 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990904173469095e-05, + "loss": 0.4668, + "step": 715 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990820546923285e-05, + "loss": 0.4314, + "step": 716 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990736537882107e-05, + "loss": 0.4336, + "step": 717 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990652146348774e-05, + "loss": 0.466, + "step": 718 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990567372326516e-05, + "loss": 0.4794, + "step": 719 + }, + { + "epoch": 0.04, + "learning_rate": 1.999048221581858e-05, + "loss": 0.4827, + "step": 720 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990396676828226e-05, + "loss": 0.466, + "step": 721 + }, + { + "epoch": 0.04, + "learning_rate": 1.999031075535873e-05, + "loss": 0.4428, + "step": 722 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990224451413378e-05, + "loss": 0.4492, + "step": 723 + }, + { + "epoch": 0.04, + "learning_rate": 1.999013776499548e-05, + "loss": 0.486, + "step": 724 + }, + { + "epoch": 0.04, + "learning_rate": 1.999005069610835e-05, + "loss": 0.464, + "step": 725 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989963244755322e-05, + "loss": 0.4082, + "step": 726 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989875410939743e-05, + "loss": 0.4782, + "step": 727 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989787194664984e-05, + "loss": 0.4107, + "step": 728 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989698595934408e-05, + "loss": 0.4632, + "step": 729 + }, + { + "epoch": 0.04, + "learning_rate": 1.998960961475142e-05, + "loss": 0.4378, + "step": 730 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989520251119418e-05, + "loss": 0.4743, + "step": 731 + }, + { + "epoch": 0.04, + "learning_rate": 1.998943050504183e-05, + "loss": 0.464, + "step": 732 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989340376522088e-05, + "loss": 0.4216, + "step": 733 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989249865563642e-05, + "loss": 0.4568, + "step": 734 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989158972169956e-05, + "loss": 0.4517, + "step": 735 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989067696344513e-05, + "loss": 0.4277, + "step": 736 + }, + { + "epoch": 0.04, + "learning_rate": 1.998897603809081e-05, + "loss": 0.4453, + "step": 737 + }, + { + "epoch": 0.04, + "learning_rate": 1.998888399741235e-05, + "loss": 0.4339, + "step": 738 + }, + { + "epoch": 0.04, + "learning_rate": 1.998879157431266e-05, + "loss": 0.4847, + "step": 739 + }, + { + "epoch": 0.04, + "learning_rate": 1.9988698768795277e-05, + "loss": 0.4662, + "step": 740 + }, + { + "epoch": 0.04, + "learning_rate": 1.9988605580863757e-05, + "loss": 0.4473, + "step": 741 + }, + { + "epoch": 0.04, + "learning_rate": 1.9988512010521663e-05, + "loss": 0.4459, + "step": 742 + }, + { + "epoch": 0.04, + "learning_rate": 1.998841805777258e-05, + "loss": 0.4492, + "step": 743 + }, + { + "epoch": 0.04, + "learning_rate": 1.9988323722620107e-05, + "loss": 0.4629, + "step": 744 + }, + { + "epoch": 0.05, + "learning_rate": 1.998822900506785e-05, + "loss": 0.4612, + "step": 745 + }, + { + "epoch": 0.05, + "learning_rate": 1.998813390511944e-05, + "loss": 0.4609, + "step": 746 + }, + { + "epoch": 0.05, + "learning_rate": 1.9988038422778518e-05, + "loss": 0.4665, + "step": 747 + }, + { + "epoch": 0.05, + "learning_rate": 1.9987942558048735e-05, + "loss": 0.4492, + "step": 748 + }, + { + "epoch": 0.05, + "learning_rate": 1.9987846310933768e-05, + "loss": 0.4654, + "step": 749 + }, + { + "epoch": 0.05, + "learning_rate": 1.9987749681437297e-05, + "loss": 0.4501, + "step": 750 + }, + { + "epoch": 0.05, + "learning_rate": 1.998765266956302e-05, + "loss": 0.4498, + "step": 751 + }, + { + "epoch": 0.05, + "learning_rate": 1.9987555275314658e-05, + "loss": 0.4191, + "step": 752 + }, + { + "epoch": 0.05, + "learning_rate": 1.9987457498695927e-05, + "loss": 0.4308, + "step": 753 + }, + { + "epoch": 0.05, + "learning_rate": 1.9987359339710586e-05, + "loss": 0.4252, + "step": 754 + }, + { + "epoch": 0.05, + "learning_rate": 1.998726079836238e-05, + "loss": 0.4439, + "step": 755 + }, + { + "epoch": 0.05, + "learning_rate": 1.9987161874655092e-05, + "loss": 0.442, + "step": 756 + }, + { + "epoch": 0.05, + "learning_rate": 1.9987062568592502e-05, + "loss": 0.4615, + "step": 757 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986962880178417e-05, + "loss": 0.4425, + "step": 758 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986862809416644e-05, + "loss": 0.4364, + "step": 759 + }, + { + "epoch": 0.05, + "learning_rate": 1.998676235631103e-05, + "loss": 0.4489, + "step": 760 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986661520865404e-05, + "loss": 0.4626, + "step": 761 + }, + { + "epoch": 0.05, + "learning_rate": 1.998656030308364e-05, + "loss": 0.4353, + "step": 762 + }, + { + "epoch": 0.05, + "learning_rate": 1.99864587029696e-05, + "loss": 0.4241, + "step": 763 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986356720527186e-05, + "loss": 0.4386, + "step": 764 + }, + { + "epoch": 0.05, + "learning_rate": 1.99862543557603e-05, + "loss": 0.4506, + "step": 765 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986151608672856e-05, + "loss": 0.4182, + "step": 766 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986048479268788e-05, + "loss": 0.4461, + "step": 767 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985944967552046e-05, + "loss": 0.4498, + "step": 768 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985841073526597e-05, + "loss": 0.406, + "step": 769 + }, + { + "epoch": 0.05, + "learning_rate": 1.998573679719641e-05, + "loss": 0.4754, + "step": 770 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985632138565484e-05, + "loss": 0.4595, + "step": 771 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985527097637828e-05, + "loss": 0.4467, + "step": 772 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985421674417452e-05, + "loss": 0.4554, + "step": 773 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985315868908404e-05, + "loss": 0.4467, + "step": 774 + }, + { + "epoch": 0.05, + "learning_rate": 1.998520968111473e-05, + "loss": 0.4311, + "step": 775 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985103111040494e-05, + "loss": 0.4336, + "step": 776 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984996158689778e-05, + "loss": 0.4378, + "step": 777 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984888824066675e-05, + "loss": 0.4526, + "step": 778 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984781107175296e-05, + "loss": 0.4297, + "step": 779 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984673008019767e-05, + "loss": 0.4601, + "step": 780 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984564526604222e-05, + "loss": 0.4425, + "step": 781 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984455662932815e-05, + "loss": 0.4397, + "step": 782 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984346417009716e-05, + "loss": 0.4568, + "step": 783 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984236788839103e-05, + "loss": 0.4314, + "step": 784 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984126778425178e-05, + "loss": 0.433, + "step": 785 + }, + { + "epoch": 0.05, + "learning_rate": 1.998401638577215e-05, + "loss": 0.4339, + "step": 786 + }, + { + "epoch": 0.05, + "learning_rate": 1.998390561088425e-05, + "loss": 0.488, + "step": 787 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983794453765714e-05, + "loss": 0.4375, + "step": 788 + }, + { + "epoch": 0.05, + "learning_rate": 1.99836829144208e-05, + "loss": 0.4216, + "step": 789 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983570992853773e-05, + "loss": 0.4593, + "step": 790 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983458689068927e-05, + "loss": 0.4238, + "step": 791 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983346003070553e-05, + "loss": 0.418, + "step": 792 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983232934862972e-05, + "loss": 0.428, + "step": 793 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983119484450506e-05, + "loss": 0.4919, + "step": 794 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983005651837503e-05, + "loss": 0.418, + "step": 795 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982891437028323e-05, + "loss": 0.447, + "step": 796 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982776840027333e-05, + "loss": 0.4333, + "step": 797 + }, + { + "epoch": 0.05, + "learning_rate": 1.998266186083892e-05, + "loss": 0.3887, + "step": 798 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982546499467495e-05, + "loss": 0.4478, + "step": 799 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982430755917463e-05, + "loss": 0.4581, + "step": 800 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982314630193263e-05, + "loss": 0.4838, + "step": 801 + }, + { + "epoch": 0.05, + "learning_rate": 1.998219812229934e-05, + "loss": 0.5114, + "step": 802 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982081232240152e-05, + "loss": 0.418, + "step": 803 + }, + { + "epoch": 0.05, + "learning_rate": 1.9981963960020178e-05, + "loss": 0.4696, + "step": 804 + }, + { + "epoch": 0.05, + "learning_rate": 1.99818463056439e-05, + "loss": 0.4222, + "step": 805 + }, + { + "epoch": 0.05, + "learning_rate": 1.9981728269115827e-05, + "loss": 0.4252, + "step": 806 + }, + { + "epoch": 0.05, + "learning_rate": 1.9981609850440483e-05, + "loss": 0.4464, + "step": 807 + }, + { + "epoch": 0.05, + "learning_rate": 1.9981491049622392e-05, + "loss": 0.4584, + "step": 808 + }, + { + "epoch": 0.05, + "learning_rate": 1.998137186666611e-05, + "loss": 0.4534, + "step": 809 + }, + { + "epoch": 0.05, + "learning_rate": 1.9981252301576198e-05, + "loss": 0.4068, + "step": 810 + }, + { + "epoch": 0.05, + "learning_rate": 1.9981132354357228e-05, + "loss": 0.4688, + "step": 811 + }, + { + "epoch": 0.05, + "learning_rate": 1.99810120250138e-05, + "loss": 0.4559, + "step": 812 + }, + { + "epoch": 0.05, + "learning_rate": 1.9980891313550515e-05, + "loss": 0.4336, + "step": 813 + }, + { + "epoch": 0.05, + "learning_rate": 1.9980770219971998e-05, + "loss": 0.4618, + "step": 814 + }, + { + "epoch": 0.05, + "learning_rate": 1.9980648744282885e-05, + "loss": 0.4501, + "step": 815 + }, + { + "epoch": 0.05, + "learning_rate": 1.9980526886487824e-05, + "loss": 0.4495, + "step": 816 + }, + { + "epoch": 0.05, + "learning_rate": 1.998040464659148e-05, + "loss": 0.4375, + "step": 817 + }, + { + "epoch": 0.05, + "learning_rate": 1.998028202459854e-05, + "loss": 0.4754, + "step": 818 + }, + { + "epoch": 0.05, + "learning_rate": 1.998015902051369e-05, + "loss": 0.4484, + "step": 819 + }, + { + "epoch": 0.05, + "learning_rate": 1.998003563434164e-05, + "loss": 0.4249, + "step": 820 + }, + { + "epoch": 0.05, + "learning_rate": 1.997991186608712e-05, + "loss": 0.4389, + "step": 821 + }, + { + "epoch": 0.05, + "learning_rate": 1.9979787715754864e-05, + "loss": 0.4322, + "step": 822 + }, + { + "epoch": 0.05, + "learning_rate": 1.9979663183349626e-05, + "loss": 0.421, + "step": 823 + }, + { + "epoch": 0.05, + "learning_rate": 1.997953826887617e-05, + "loss": 0.4584, + "step": 824 + }, + { + "epoch": 0.05, + "learning_rate": 1.997941297233928e-05, + "loss": 0.44, + "step": 825 + }, + { + "epoch": 0.05, + "learning_rate": 1.997928729374376e-05, + "loss": 0.4461, + "step": 826 + }, + { + "epoch": 0.05, + "learning_rate": 1.9979161233094414e-05, + "loss": 0.4609, + "step": 827 + }, + { + "epoch": 0.05, + "learning_rate": 1.997903479039607e-05, + "loss": 0.4618, + "step": 828 + }, + { + "epoch": 0.05, + "learning_rate": 1.9978907965653574e-05, + "loss": 0.4554, + "step": 829 + }, + { + "epoch": 0.05, + "learning_rate": 1.997878075887177e-05, + "loss": 0.4431, + "step": 830 + }, + { + "epoch": 0.05, + "learning_rate": 1.997865317005554e-05, + "loss": 0.4143, + "step": 831 + }, + { + "epoch": 0.05, + "learning_rate": 1.997852519920976e-05, + "loss": 0.4749, + "step": 832 + }, + { + "epoch": 0.05, + "learning_rate": 1.9978396846339333e-05, + "loss": 0.4568, + "step": 833 + }, + { + "epoch": 0.05, + "learning_rate": 1.9978268111449174e-05, + "loss": 0.4355, + "step": 834 + }, + { + "epoch": 0.05, + "learning_rate": 1.9978138994544208e-05, + "loss": 0.4222, + "step": 835 + }, + { + "epoch": 0.05, + "learning_rate": 1.997800949562938e-05, + "loss": 0.4302, + "step": 836 + }, + { + "epoch": 0.05, + "learning_rate": 1.997787961470965e-05, + "loss": 0.4729, + "step": 837 + }, + { + "epoch": 0.05, + "learning_rate": 1.997774935178999e-05, + "loss": 0.4007, + "step": 838 + }, + { + "epoch": 0.05, + "learning_rate": 1.9977618706875384e-05, + "loss": 0.4227, + "step": 839 + }, + { + "epoch": 0.05, + "learning_rate": 1.9977487679970834e-05, + "loss": 0.4395, + "step": 840 + }, + { + "epoch": 0.05, + "learning_rate": 1.997735627108136e-05, + "loss": 0.4235, + "step": 841 + }, + { + "epoch": 0.05, + "learning_rate": 1.997722448021199e-05, + "loss": 0.4322, + "step": 842 + }, + { + "epoch": 0.05, + "learning_rate": 1.9977092307367772e-05, + "loss": 0.4205, + "step": 843 + }, + { + "epoch": 0.05, + "learning_rate": 1.9976959752553764e-05, + "loss": 0.459, + "step": 844 + }, + { + "epoch": 0.05, + "learning_rate": 1.997682681577504e-05, + "loss": 0.4556, + "step": 845 + }, + { + "epoch": 0.05, + "learning_rate": 1.997669349703669e-05, + "loss": 0.4353, + "step": 846 + }, + { + "epoch": 0.05, + "learning_rate": 1.9976559796343818e-05, + "loss": 0.4576, + "step": 847 + }, + { + "epoch": 0.05, + "learning_rate": 1.9976425713701548e-05, + "loss": 0.4174, + "step": 848 + }, + { + "epoch": 0.05, + "learning_rate": 1.9976291249115005e-05, + "loss": 0.4503, + "step": 849 + }, + { + "epoch": 0.05, + "learning_rate": 1.9976156402589342e-05, + "loss": 0.4406, + "step": 850 + }, + { + "epoch": 0.05, + "learning_rate": 1.997602117412972e-05, + "loss": 0.421, + "step": 851 + }, + { + "epoch": 0.05, + "learning_rate": 1.9975885563741316e-05, + "loss": 0.4584, + "step": 852 + }, + { + "epoch": 0.05, + "learning_rate": 1.9975749571429324e-05, + "loss": 0.4835, + "step": 853 + }, + { + "epoch": 0.05, + "learning_rate": 1.9975613197198948e-05, + "loss": 0.4289, + "step": 854 + }, + { + "epoch": 0.05, + "learning_rate": 1.997547644105541e-05, + "loss": 0.4403, + "step": 855 + }, + { + "epoch": 0.05, + "learning_rate": 1.9975339303003945e-05, + "loss": 0.4741, + "step": 856 + }, + { + "epoch": 0.05, + "learning_rate": 1.9975201783049804e-05, + "loss": 0.4333, + "step": 857 + }, + { + "epoch": 0.05, + "learning_rate": 1.997506388119825e-05, + "loss": 0.4258, + "step": 858 + }, + { + "epoch": 0.05, + "learning_rate": 1.9974925597454567e-05, + "loss": 0.4442, + "step": 859 + }, + { + "epoch": 0.05, + "learning_rate": 1.9974786931824044e-05, + "loss": 0.4565, + "step": 860 + }, + { + "epoch": 0.05, + "learning_rate": 1.9974647884311994e-05, + "loss": 0.4157, + "step": 861 + }, + { + "epoch": 0.05, + "learning_rate": 1.997450845492374e-05, + "loss": 0.4249, + "step": 862 + }, + { + "epoch": 0.05, + "learning_rate": 1.997436864366461e-05, + "loss": 0.4528, + "step": 863 + }, + { + "epoch": 0.05, + "learning_rate": 1.9974228450539975e-05, + "loss": 0.4428, + "step": 864 + }, + { + "epoch": 0.05, + "learning_rate": 1.9974087875555186e-05, + "loss": 0.4439, + "step": 865 + }, + { + "epoch": 0.05, + "learning_rate": 1.9973946918715632e-05, + "loss": 0.3956, + "step": 866 + }, + { + "epoch": 0.05, + "learning_rate": 1.997380558002671e-05, + "loss": 0.4397, + "step": 867 + }, + { + "epoch": 0.05, + "learning_rate": 1.9973663859493832e-05, + "loss": 0.433, + "step": 868 + }, + { + "epoch": 0.05, + "learning_rate": 1.9973521757122418e-05, + "loss": 0.4489, + "step": 869 + }, + { + "epoch": 0.05, + "learning_rate": 1.9973379272917915e-05, + "loss": 0.3996, + "step": 870 + }, + { + "epoch": 0.05, + "learning_rate": 1.9973236406885773e-05, + "loss": 0.4308, + "step": 871 + }, + { + "epoch": 0.05, + "learning_rate": 1.9973093159031464e-05, + "loss": 0.4015, + "step": 872 + }, + { + "epoch": 0.05, + "learning_rate": 1.997294952936047e-05, + "loss": 0.4032, + "step": 873 + }, + { + "epoch": 0.05, + "learning_rate": 1.9972805517878292e-05, + "loss": 0.4249, + "step": 874 + }, + { + "epoch": 0.05, + "learning_rate": 1.997266112459044e-05, + "loss": 0.4487, + "step": 875 + }, + { + "epoch": 0.05, + "learning_rate": 1.9972516349502453e-05, + "loss": 0.4174, + "step": 876 + }, + { + "epoch": 0.05, + "learning_rate": 1.997237119261986e-05, + "loss": 0.4601, + "step": 877 + }, + { + "epoch": 0.05, + "learning_rate": 1.9972225653948226e-05, + "loss": 0.4163, + "step": 878 + }, + { + "epoch": 0.05, + "learning_rate": 1.9972079733493118e-05, + "loss": 0.4333, + "step": 879 + }, + { + "epoch": 0.05, + "learning_rate": 1.9971933431260128e-05, + "loss": 0.447, + "step": 880 + }, + { + "epoch": 0.05, + "learning_rate": 1.9971786747254854e-05, + "loss": 0.4431, + "step": 881 + }, + { + "epoch": 0.05, + "learning_rate": 1.997163968148291e-05, + "loss": 0.4612, + "step": 882 + }, + { + "epoch": 0.05, + "learning_rate": 1.997149223394993e-05, + "loss": 0.4805, + "step": 883 + }, + { + "epoch": 0.05, + "learning_rate": 1.9971344404661558e-05, + "loss": 0.423, + "step": 884 + }, + { + "epoch": 0.05, + "learning_rate": 1.9971196193623453e-05, + "loss": 0.43, + "step": 885 + }, + { + "epoch": 0.05, + "learning_rate": 1.997104760084129e-05, + "loss": 0.387, + "step": 886 + }, + { + "epoch": 0.05, + "learning_rate": 1.9970898626320752e-05, + "loss": 0.404, + "step": 887 + }, + { + "epoch": 0.05, + "learning_rate": 1.9970749270067554e-05, + "loss": 0.4406, + "step": 888 + }, + { + "epoch": 0.05, + "learning_rate": 1.9970599532087404e-05, + "loss": 0.4263, + "step": 889 + }, + { + "epoch": 0.05, + "learning_rate": 1.997044941238604e-05, + "loss": 0.4272, + "step": 890 + }, + { + "epoch": 0.05, + "learning_rate": 1.997029891096921e-05, + "loss": 0.4127, + "step": 891 + }, + { + "epoch": 0.05, + "learning_rate": 1.997014802784267e-05, + "loss": 0.4132, + "step": 892 + }, + { + "epoch": 0.05, + "learning_rate": 1.9969996763012198e-05, + "loss": 0.4607, + "step": 893 + }, + { + "epoch": 0.05, + "learning_rate": 1.9969845116483594e-05, + "loss": 0.4121, + "step": 894 + }, + { + "epoch": 0.05, + "learning_rate": 1.996969308826265e-05, + "loss": 0.435, + "step": 895 + }, + { + "epoch": 0.05, + "learning_rate": 1.9969540678355198e-05, + "loss": 0.4107, + "step": 896 + }, + { + "epoch": 0.05, + "learning_rate": 1.9969387886767063e-05, + "loss": 0.4431, + "step": 897 + }, + { + "epoch": 0.05, + "learning_rate": 1.9969234713504103e-05, + "loss": 0.406, + "step": 898 + }, + { + "epoch": 0.05, + "learning_rate": 1.996908115857218e-05, + "loss": 0.4057, + "step": 899 + }, + { + "epoch": 0.05, + "learning_rate": 1.9968927221977173e-05, + "loss": 0.4358, + "step": 900 + }, + { + "epoch": 0.05, + "learning_rate": 1.996877290372497e-05, + "loss": 0.4587, + "step": 901 + }, + { + "epoch": 0.05, + "learning_rate": 1.9968618203821487e-05, + "loss": 0.3876, + "step": 902 + }, + { + "epoch": 0.05, + "learning_rate": 1.9968463122272637e-05, + "loss": 0.4247, + "step": 903 + }, + { + "epoch": 0.05, + "learning_rate": 1.996830765908437e-05, + "loss": 0.4355, + "step": 904 + }, + { + "epoch": 0.05, + "learning_rate": 1.9968151814262627e-05, + "loss": 0.4286, + "step": 905 + }, + { + "epoch": 0.05, + "learning_rate": 1.996799558781338e-05, + "loss": 0.4233, + "step": 906 + }, + { + "epoch": 0.05, + "learning_rate": 1.9967838979742606e-05, + "loss": 0.4132, + "step": 907 + }, + { + "epoch": 0.05, + "learning_rate": 1.9967681990056307e-05, + "loss": 0.4459, + "step": 908 + }, + { + "epoch": 0.05, + "learning_rate": 1.996752461876049e-05, + "loss": 0.4526, + "step": 909 + }, + { + "epoch": 0.05, + "learning_rate": 1.9967366865861173e-05, + "loss": 0.4481, + "step": 910 + }, + { + "epoch": 0.06, + "learning_rate": 1.9967208731364408e-05, + "loss": 0.4342, + "step": 911 + }, + { + "epoch": 0.06, + "learning_rate": 1.996705021527624e-05, + "loss": 0.4224, + "step": 912 + }, + { + "epoch": 0.06, + "learning_rate": 1.9966891317602744e-05, + "loss": 0.4054, + "step": 913 + }, + { + "epoch": 0.06, + "learning_rate": 1.9966732038349997e-05, + "loss": 0.4099, + "step": 914 + }, + { + "epoch": 0.06, + "learning_rate": 1.99665723775241e-05, + "loss": 0.4132, + "step": 915 + }, + { + "epoch": 0.06, + "learning_rate": 1.9966412335131166e-05, + "loss": 0.4219, + "step": 916 + }, + { + "epoch": 0.06, + "learning_rate": 1.9966251911177324e-05, + "loss": 0.4436, + "step": 917 + }, + { + "epoch": 0.06, + "learning_rate": 1.9966091105668713e-05, + "loss": 0.4269, + "step": 918 + }, + { + "epoch": 0.06, + "learning_rate": 1.9965929918611488e-05, + "loss": 0.3936, + "step": 919 + }, + { + "epoch": 0.06, + "learning_rate": 1.9965768350011822e-05, + "loss": 0.4297, + "step": 920 + }, + { + "epoch": 0.06, + "learning_rate": 1.99656063998759e-05, + "loss": 0.4584, + "step": 921 + }, + { + "epoch": 0.06, + "learning_rate": 1.9965444068209927e-05, + "loss": 0.4043, + "step": 922 + }, + { + "epoch": 0.06, + "learning_rate": 1.996528135502011e-05, + "loss": 0.4506, + "step": 923 + }, + { + "epoch": 0.06, + "learning_rate": 1.9965118260312685e-05, + "loss": 0.4138, + "step": 924 + }, + { + "epoch": 0.06, + "learning_rate": 1.996495478409389e-05, + "loss": 0.4035, + "step": 925 + }, + { + "epoch": 0.06, + "learning_rate": 1.9964790926369985e-05, + "loss": 0.3973, + "step": 926 + }, + { + "epoch": 0.06, + "learning_rate": 1.996462668714725e-05, + "loss": 0.4369, + "step": 927 + }, + { + "epoch": 0.06, + "learning_rate": 1.9964462066431965e-05, + "loss": 0.4049, + "step": 928 + }, + { + "epoch": 0.06, + "learning_rate": 1.9964297064230437e-05, + "loss": 0.4342, + "step": 929 + }, + { + "epoch": 0.06, + "learning_rate": 1.9964131680548976e-05, + "loss": 0.3912, + "step": 930 + }, + { + "epoch": 0.06, + "learning_rate": 1.9963965915393925e-05, + "loss": 0.4141, + "step": 931 + }, + { + "epoch": 0.06, + "learning_rate": 1.9963799768771623e-05, + "loss": 0.4439, + "step": 932 + }, + { + "epoch": 0.06, + "learning_rate": 1.996363324068843e-05, + "loss": 0.4478, + "step": 933 + }, + { + "epoch": 0.06, + "learning_rate": 1.9963466331150726e-05, + "loss": 0.4143, + "step": 934 + }, + { + "epoch": 0.06, + "learning_rate": 1.9963299040164896e-05, + "loss": 0.4704, + "step": 935 + }, + { + "epoch": 0.06, + "learning_rate": 1.996313136773735e-05, + "loss": 0.4283, + "step": 936 + }, + { + "epoch": 0.06, + "learning_rate": 1.9962963313874503e-05, + "loss": 0.4403, + "step": 937 + }, + { + "epoch": 0.06, + "learning_rate": 1.996279487858279e-05, + "loss": 0.4289, + "step": 938 + }, + { + "epoch": 0.06, + "learning_rate": 1.9962626061868664e-05, + "loss": 0.4688, + "step": 939 + }, + { + "epoch": 0.06, + "learning_rate": 1.996245686373858e-05, + "loss": 0.4129, + "step": 940 + }, + { + "epoch": 0.06, + "learning_rate": 1.996228728419902e-05, + "loss": 0.4029, + "step": 941 + }, + { + "epoch": 0.06, + "learning_rate": 1.9962117323256477e-05, + "loss": 0.4155, + "step": 942 + }, + { + "epoch": 0.06, + "learning_rate": 1.9961946980917457e-05, + "loss": 0.4182, + "step": 943 + }, + { + "epoch": 0.06, + "learning_rate": 1.996177625718848e-05, + "loss": 0.4559, + "step": 944 + }, + { + "epoch": 0.06, + "learning_rate": 1.9961605152076084e-05, + "loss": 0.4353, + "step": 945 + }, + { + "epoch": 0.06, + "learning_rate": 1.9961433665586823e-05, + "loss": 0.3945, + "step": 946 + }, + { + "epoch": 0.06, + "learning_rate": 1.9961261797727256e-05, + "loss": 0.4269, + "step": 947 + }, + { + "epoch": 0.06, + "learning_rate": 1.9961089548503965e-05, + "loss": 0.4096, + "step": 948 + }, + { + "epoch": 0.06, + "learning_rate": 1.9960916917923545e-05, + "loss": 0.411, + "step": 949 + }, + { + "epoch": 0.06, + "learning_rate": 1.9960743905992605e-05, + "loss": 0.4562, + "step": 950 + }, + { + "epoch": 0.06, + "learning_rate": 1.996057051271777e-05, + "loss": 0.4579, + "step": 951 + }, + { + "epoch": 0.06, + "learning_rate": 1.9960396738105678e-05, + "loss": 0.4071, + "step": 952 + }, + { + "epoch": 0.06, + "learning_rate": 1.996022258216298e-05, + "loss": 0.4096, + "step": 953 + }, + { + "epoch": 0.06, + "learning_rate": 1.996004804489634e-05, + "loss": 0.476, + "step": 954 + }, + { + "epoch": 0.06, + "learning_rate": 1.995987312631245e-05, + "loss": 0.4565, + "step": 955 + }, + { + "epoch": 0.06, + "learning_rate": 1.9959697826417998e-05, + "loss": 0.4364, + "step": 956 + }, + { + "epoch": 0.06, + "learning_rate": 1.99595221452197e-05, + "loss": 0.4129, + "step": 957 + }, + { + "epoch": 0.06, + "learning_rate": 1.9959346082724283e-05, + "loss": 0.4821, + "step": 958 + }, + { + "epoch": 0.06, + "learning_rate": 1.9959169638938482e-05, + "loss": 0.4227, + "step": 959 + }, + { + "epoch": 0.06, + "learning_rate": 1.9958992813869057e-05, + "loss": 0.4065, + "step": 960 + }, + { + "epoch": 0.06, + "learning_rate": 1.9958815607522774e-05, + "loss": 0.3929, + "step": 961 + }, + { + "epoch": 0.06, + "learning_rate": 1.995863801990642e-05, + "loss": 0.4727, + "step": 962 + }, + { + "epoch": 0.06, + "learning_rate": 1.9958460051026795e-05, + "loss": 0.4174, + "step": 963 + }, + { + "epoch": 0.06, + "learning_rate": 1.995828170089071e-05, + "loss": 0.406, + "step": 964 + }, + { + "epoch": 0.06, + "learning_rate": 1.995810296950499e-05, + "loss": 0.3797, + "step": 965 + }, + { + "epoch": 0.06, + "learning_rate": 1.9957923856876484e-05, + "loss": 0.4074, + "step": 966 + }, + { + "epoch": 0.06, + "learning_rate": 1.9957744363012047e-05, + "loss": 0.4043, + "step": 967 + }, + { + "epoch": 0.06, + "learning_rate": 1.9957564487918545e-05, + "loss": 0.4138, + "step": 968 + }, + { + "epoch": 0.06, + "learning_rate": 1.9957384231602876e-05, + "loss": 0.4406, + "step": 969 + }, + { + "epoch": 0.06, + "learning_rate": 1.9957203594071935e-05, + "loss": 0.4294, + "step": 970 + }, + { + "epoch": 0.06, + "learning_rate": 1.9957022575332634e-05, + "loss": 0.4255, + "step": 971 + }, + { + "epoch": 0.06, + "learning_rate": 1.9956841175391908e-05, + "loss": 0.4598, + "step": 972 + }, + { + "epoch": 0.06, + "learning_rate": 1.9956659394256702e-05, + "loss": 0.4593, + "step": 973 + }, + { + "epoch": 0.06, + "learning_rate": 1.9956477231933974e-05, + "loss": 0.3998, + "step": 974 + }, + { + "epoch": 0.06, + "learning_rate": 1.9956294688430694e-05, + "loss": 0.4009, + "step": 975 + }, + { + "epoch": 0.06, + "learning_rate": 1.9956111763753857e-05, + "loss": 0.4593, + "step": 976 + }, + { + "epoch": 0.06, + "learning_rate": 1.9955928457910466e-05, + "loss": 0.4224, + "step": 977 + }, + { + "epoch": 0.06, + "learning_rate": 1.9955744770907535e-05, + "loss": 0.4185, + "step": 978 + }, + { + "epoch": 0.06, + "learning_rate": 1.99555607027521e-05, + "loss": 0.4491, + "step": 979 + }, + { + "epoch": 0.06, + "learning_rate": 1.9955376253451204e-05, + "loss": 0.4294, + "step": 980 + }, + { + "epoch": 0.06, + "learning_rate": 1.995519142301191e-05, + "loss": 0.4612, + "step": 981 + }, + { + "epoch": 0.06, + "learning_rate": 1.9955006211441295e-05, + "loss": 0.4302, + "step": 982 + }, + { + "epoch": 0.06, + "learning_rate": 1.995482061874645e-05, + "loss": 0.4517, + "step": 983 + }, + { + "epoch": 0.06, + "learning_rate": 1.9954634644934484e-05, + "loss": 0.4473, + "step": 984 + }, + { + "epoch": 0.06, + "learning_rate": 1.995444829001251e-05, + "loss": 0.4771, + "step": 985 + }, + { + "epoch": 0.06, + "learning_rate": 1.9954261553987662e-05, + "loss": 0.4685, + "step": 986 + }, + { + "epoch": 0.06, + "learning_rate": 1.9954074436867093e-05, + "loss": 0.4127, + "step": 987 + }, + { + "epoch": 0.06, + "learning_rate": 1.9953886938657972e-05, + "loss": 0.4283, + "step": 988 + }, + { + "epoch": 0.06, + "learning_rate": 1.9953699059367468e-05, + "loss": 0.4369, + "step": 989 + }, + { + "epoch": 0.06, + "learning_rate": 1.995351079900278e-05, + "loss": 0.3926, + "step": 990 + }, + { + "epoch": 0.06, + "learning_rate": 1.9953322157571108e-05, + "loss": 0.3984, + "step": 991 + }, + { + "epoch": 0.06, + "learning_rate": 1.9953133135079686e-05, + "loss": 0.4051, + "step": 992 + }, + { + "epoch": 0.06, + "learning_rate": 1.995294373153574e-05, + "loss": 0.4074, + "step": 993 + }, + { + "epoch": 0.06, + "learning_rate": 1.9952753946946522e-05, + "loss": 0.445, + "step": 994 + }, + { + "epoch": 0.06, + "learning_rate": 1.9952563781319303e-05, + "loss": 0.4612, + "step": 995 + }, + { + "epoch": 0.06, + "learning_rate": 1.9952373234661362e-05, + "loss": 0.4688, + "step": 996 + }, + { + "epoch": 0.06, + "learning_rate": 1.9952182306979996e-05, + "loss": 0.409, + "step": 997 + }, + { + "epoch": 0.06, + "learning_rate": 1.9951990998282505e-05, + "loss": 0.4001, + "step": 998 + }, + { + "epoch": 0.06, + "learning_rate": 1.995179930857623e-05, + "loss": 0.4116, + "step": 999 + }, + { + "epoch": 0.06, + "learning_rate": 1.995160723786849e-05, + "loss": 0.4191, + "step": 1000 + }, + { + "epoch": 0.06, + "learning_rate": 1.9951414786166656e-05, + "loss": 0.4244, + "step": 1001 + }, + { + "epoch": 0.06, + "learning_rate": 1.995122195347808e-05, + "loss": 0.4277, + "step": 1002 + }, + { + "epoch": 0.06, + "learning_rate": 1.995102873981016e-05, + "loss": 0.4464, + "step": 1003 + }, + { + "epoch": 0.06, + "learning_rate": 1.995083514517028e-05, + "loss": 0.4261, + "step": 1004 + }, + { + "epoch": 0.06, + "learning_rate": 1.9950641169565863e-05, + "loss": 0.4012, + "step": 1005 + }, + { + "epoch": 0.06, + "learning_rate": 1.9950446813004327e-05, + "loss": 0.4155, + "step": 1006 + }, + { + "epoch": 0.06, + "learning_rate": 1.9950252075493116e-05, + "loss": 0.435, + "step": 1007 + }, + { + "epoch": 0.06, + "learning_rate": 1.9950056957039688e-05, + "loss": 0.4473, + "step": 1008 + }, + { + "epoch": 0.06, + "learning_rate": 1.994986145765151e-05, + "loss": 0.4275, + "step": 1009 + }, + { + "epoch": 0.06, + "learning_rate": 1.994966557733606e-05, + "loss": 0.4406, + "step": 1010 + }, + { + "epoch": 0.06, + "learning_rate": 1.994946931610085e-05, + "loss": 0.445, + "step": 1011 + }, + { + "epoch": 0.06, + "learning_rate": 1.9949272673953385e-05, + "loss": 0.452, + "step": 1012 + }, + { + "epoch": 0.06, + "learning_rate": 1.9949075650901198e-05, + "loss": 0.4445, + "step": 1013 + }, + { + "epoch": 0.06, + "learning_rate": 1.994887824695183e-05, + "loss": 0.3984, + "step": 1014 + }, + { + "epoch": 0.06, + "learning_rate": 1.994868046211284e-05, + "loss": 0.4286, + "step": 1015 + }, + { + "epoch": 0.06, + "learning_rate": 1.9948482296391798e-05, + "loss": 0.4406, + "step": 1016 + }, + { + "epoch": 0.06, + "learning_rate": 1.9948283749796293e-05, + "loss": 0.4325, + "step": 1017 + }, + { + "epoch": 0.06, + "learning_rate": 1.994808482233392e-05, + "loss": 0.4475, + "step": 1018 + }, + { + "epoch": 0.06, + "learning_rate": 1.9947885514012303e-05, + "loss": 0.4227, + "step": 1019 + }, + { + "epoch": 0.06, + "learning_rate": 1.9947685824839072e-05, + "loss": 0.4182, + "step": 1020 + }, + { + "epoch": 0.06, + "learning_rate": 1.9947485754821865e-05, + "loss": 0.4116, + "step": 1021 + }, + { + "epoch": 0.06, + "learning_rate": 1.9947285303968348e-05, + "loss": 0.4146, + "step": 1022 + }, + { + "epoch": 0.06, + "learning_rate": 1.9947084472286195e-05, + "loss": 0.4319, + "step": 1023 + }, + { + "epoch": 0.06, + "learning_rate": 1.9946883259783088e-05, + "loss": 0.4174, + "step": 1024 + }, + { + "epoch": 0.06, + "learning_rate": 1.9946681666466737e-05, + "loss": 0.4283, + "step": 1025 + }, + { + "epoch": 0.06, + "learning_rate": 1.994647969234486e-05, + "loss": 0.4121, + "step": 1026 + }, + { + "epoch": 0.06, + "learning_rate": 1.9946277337425186e-05, + "loss": 0.4065, + "step": 1027 + }, + { + "epoch": 0.06, + "learning_rate": 1.9946074601715462e-05, + "loss": 0.4372, + "step": 1028 + }, + { + "epoch": 0.06, + "learning_rate": 1.9945871485223456e-05, + "loss": 0.4082, + "step": 1029 + }, + { + "epoch": 0.06, + "learning_rate": 1.9945667987956937e-05, + "loss": 0.3778, + "step": 1030 + }, + { + "epoch": 0.06, + "learning_rate": 1.9945464109923696e-05, + "loss": 0.4018, + "step": 1031 + }, + { + "epoch": 0.06, + "learning_rate": 1.9945259851131548e-05, + "loss": 0.4275, + "step": 1032 + }, + { + "epoch": 0.06, + "learning_rate": 1.9945055211588302e-05, + "loss": 0.4294, + "step": 1033 + }, + { + "epoch": 0.06, + "learning_rate": 1.9944850191301794e-05, + "loss": 0.4422, + "step": 1034 + }, + { + "epoch": 0.06, + "learning_rate": 1.994464479027988e-05, + "loss": 0.404, + "step": 1035 + }, + { + "epoch": 0.06, + "learning_rate": 1.9944439008530416e-05, + "loss": 0.3934, + "step": 1036 + }, + { + "epoch": 0.06, + "learning_rate": 1.9944232846061284e-05, + "loss": 0.421, + "step": 1037 + }, + { + "epoch": 0.06, + "learning_rate": 1.994402630288038e-05, + "loss": 0.435, + "step": 1038 + }, + { + "epoch": 0.06, + "learning_rate": 1.9943819378995604e-05, + "loss": 0.4032, + "step": 1039 + }, + { + "epoch": 0.06, + "learning_rate": 1.9943612074414883e-05, + "loss": 0.4545, + "step": 1040 + }, + { + "epoch": 0.06, + "learning_rate": 1.9943404389146154e-05, + "loss": 0.416, + "step": 1041 + }, + { + "epoch": 0.06, + "learning_rate": 1.9943196323197362e-05, + "loss": 0.4319, + "step": 1042 + }, + { + "epoch": 0.06, + "learning_rate": 1.9942987876576484e-05, + "loss": 0.4049, + "step": 1043 + }, + { + "epoch": 0.06, + "learning_rate": 1.994277904929149e-05, + "loss": 0.4141, + "step": 1044 + }, + { + "epoch": 0.06, + "learning_rate": 1.9942569841350376e-05, + "loss": 0.4489, + "step": 1045 + }, + { + "epoch": 0.06, + "learning_rate": 1.9942360252761158e-05, + "loss": 0.4294, + "step": 1046 + }, + { + "epoch": 0.06, + "learning_rate": 1.9942150283531855e-05, + "loss": 0.4601, + "step": 1047 + }, + { + "epoch": 0.06, + "learning_rate": 1.9941939933670506e-05, + "loss": 0.3823, + "step": 1048 + }, + { + "epoch": 0.06, + "learning_rate": 1.9941729203185168e-05, + "loss": 0.4448, + "step": 1049 + }, + { + "epoch": 0.06, + "learning_rate": 1.9941518092083904e-05, + "loss": 0.4043, + "step": 1050 + }, + { + "epoch": 0.06, + "learning_rate": 1.9941306600374796e-05, + "loss": 0.4143, + "step": 1051 + }, + { + "epoch": 0.06, + "learning_rate": 1.9941094728065945e-05, + "loss": 0.4422, + "step": 1052 + }, + { + "epoch": 0.06, + "learning_rate": 1.994088247516546e-05, + "loss": 0.4068, + "step": 1053 + }, + { + "epoch": 0.06, + "learning_rate": 1.9940669841681466e-05, + "loss": 0.44, + "step": 1054 + }, + { + "epoch": 0.06, + "learning_rate": 1.994045682762211e-05, + "loss": 0.404, + "step": 1055 + }, + { + "epoch": 0.06, + "learning_rate": 1.9940243432995535e-05, + "loss": 0.4369, + "step": 1056 + }, + { + "epoch": 0.06, + "learning_rate": 1.994002965780992e-05, + "loss": 0.3993, + "step": 1057 + }, + { + "epoch": 0.06, + "learning_rate": 1.993981550207345e-05, + "loss": 0.4372, + "step": 1058 + }, + { + "epoch": 0.06, + "learning_rate": 1.9939600965794317e-05, + "loss": 0.4266, + "step": 1059 + }, + { + "epoch": 0.06, + "learning_rate": 1.993938604898074e-05, + "loss": 0.4501, + "step": 1060 + }, + { + "epoch": 0.06, + "learning_rate": 1.993917075164095e-05, + "loss": 0.4403, + "step": 1061 + }, + { + "epoch": 0.06, + "learning_rate": 1.993895507378318e-05, + "loss": 0.4515, + "step": 1062 + }, + { + "epoch": 0.06, + "learning_rate": 1.9938739015415694e-05, + "loss": 0.4411, + "step": 1063 + }, + { + "epoch": 0.06, + "learning_rate": 1.993852257654676e-05, + "loss": 0.4149, + "step": 1064 + }, + { + "epoch": 0.06, + "learning_rate": 1.9938305757184668e-05, + "loss": 0.4481, + "step": 1065 + }, + { + "epoch": 0.06, + "learning_rate": 1.9938088557337717e-05, + "loss": 0.4227, + "step": 1066 + }, + { + "epoch": 0.06, + "learning_rate": 1.993787097701422e-05, + "loss": 0.4096, + "step": 1067 + }, + { + "epoch": 0.06, + "learning_rate": 1.9937653016222516e-05, + "loss": 0.3951, + "step": 1068 + }, + { + "epoch": 0.06, + "learning_rate": 1.993743467497094e-05, + "loss": 0.3956, + "step": 1069 + }, + { + "epoch": 0.06, + "learning_rate": 1.9937215953267854e-05, + "loss": 0.435, + "step": 1070 + }, + { + "epoch": 0.06, + "learning_rate": 1.993699685112163e-05, + "loss": 0.4297, + "step": 1071 + }, + { + "epoch": 0.06, + "learning_rate": 1.993677736854066e-05, + "loss": 0.4135, + "step": 1072 + }, + { + "epoch": 0.06, + "learning_rate": 1.9936557505533346e-05, + "loss": 0.4484, + "step": 1073 + }, + { + "epoch": 0.06, + "learning_rate": 1.9936337262108103e-05, + "loss": 0.4138, + "step": 1074 + }, + { + "epoch": 0.06, + "learning_rate": 1.9936116638273364e-05, + "loss": 0.4046, + "step": 1075 + }, + { + "epoch": 0.07, + "learning_rate": 1.9935895634037578e-05, + "loss": 0.416, + "step": 1076 + }, + { + "epoch": 0.07, + "learning_rate": 1.99356742494092e-05, + "loss": 0.382, + "step": 1077 + }, + { + "epoch": 0.07, + "learning_rate": 1.9935452484396714e-05, + "loss": 0.4208, + "step": 1078 + }, + { + "epoch": 0.07, + "learning_rate": 1.9935230339008604e-05, + "loss": 0.4163, + "step": 1079 + }, + { + "epoch": 0.07, + "learning_rate": 1.9935007813253376e-05, + "loss": 0.4068, + "step": 1080 + }, + { + "epoch": 0.07, + "learning_rate": 1.993478490713955e-05, + "loss": 0.4411, + "step": 1081 + }, + { + "epoch": 0.07, + "learning_rate": 1.993456162067566e-05, + "loss": 0.4116, + "step": 1082 + }, + { + "epoch": 0.07, + "learning_rate": 1.9934337953870254e-05, + "loss": 0.4252, + "step": 1083 + }, + { + "epoch": 0.07, + "learning_rate": 1.9934113906731895e-05, + "loss": 0.4258, + "step": 1084 + }, + { + "epoch": 0.07, + "learning_rate": 1.9933889479269164e-05, + "loss": 0.4026, + "step": 1085 + }, + { + "epoch": 0.07, + "learning_rate": 1.9933664671490648e-05, + "loss": 0.4408, + "step": 1086 + }, + { + "epoch": 0.07, + "learning_rate": 1.9933439483404954e-05, + "loss": 0.4272, + "step": 1087 + }, + { + "epoch": 0.07, + "learning_rate": 1.9933213915020707e-05, + "loss": 0.428, + "step": 1088 + }, + { + "epoch": 0.07, + "learning_rate": 1.9932987966346538e-05, + "loss": 0.4074, + "step": 1089 + }, + { + "epoch": 0.07, + "learning_rate": 1.9932761637391104e-05, + "loss": 0.3842, + "step": 1090 + }, + { + "epoch": 0.07, + "learning_rate": 1.9932534928163063e-05, + "loss": 0.4085, + "step": 1091 + }, + { + "epoch": 0.07, + "learning_rate": 1.99323078386711e-05, + "loss": 0.4121, + "step": 1092 + }, + { + "epoch": 0.07, + "learning_rate": 1.9932080368923903e-05, + "loss": 0.4222, + "step": 1093 + }, + { + "epoch": 0.07, + "learning_rate": 1.9931852518930185e-05, + "loss": 0.4177, + "step": 1094 + }, + { + "epoch": 0.07, + "learning_rate": 1.993162428869867e-05, + "loss": 0.4096, + "step": 1095 + }, + { + "epoch": 0.07, + "learning_rate": 1.99313956782381e-05, + "loss": 0.4319, + "step": 1096 + }, + { + "epoch": 0.07, + "learning_rate": 1.993116668755721e-05, + "loss": 0.4233, + "step": 1097 + }, + { + "epoch": 0.07, + "learning_rate": 1.9930937316664782e-05, + "loss": 0.38, + "step": 1098 + }, + { + "epoch": 0.07, + "learning_rate": 1.99307075655696e-05, + "loss": 0.4283, + "step": 1099 + }, + { + "epoch": 0.07, + "learning_rate": 1.9930477434280445e-05, + "loss": 0.4233, + "step": 1100 + }, + { + "epoch": 0.07, + "learning_rate": 1.993024692280614e-05, + "loss": 0.4085, + "step": 1101 + }, + { + "epoch": 0.07, + "learning_rate": 1.9930016031155506e-05, + "loss": 0.4395, + "step": 1102 + }, + { + "epoch": 0.07, + "learning_rate": 1.992978475933738e-05, + "loss": 0.4068, + "step": 1103 + }, + { + "epoch": 0.07, + "learning_rate": 1.992955310736062e-05, + "loss": 0.4188, + "step": 1104 + }, + { + "epoch": 0.07, + "learning_rate": 1.9929321075234096e-05, + "loss": 0.4157, + "step": 1105 + }, + { + "epoch": 0.07, + "learning_rate": 1.9929088662966687e-05, + "loss": 0.4266, + "step": 1106 + }, + { + "epoch": 0.07, + "learning_rate": 1.9928855870567295e-05, + "loss": 0.3929, + "step": 1107 + }, + { + "epoch": 0.07, + "learning_rate": 1.9928622698044827e-05, + "loss": 0.387, + "step": 1108 + }, + { + "epoch": 0.07, + "learning_rate": 1.9928389145408212e-05, + "loss": 0.406, + "step": 1109 + }, + { + "epoch": 0.07, + "learning_rate": 1.9928155212666395e-05, + "loss": 0.4224, + "step": 1110 + }, + { + "epoch": 0.07, + "learning_rate": 1.992792089982833e-05, + "loss": 0.4051, + "step": 1111 + }, + { + "epoch": 0.07, + "learning_rate": 1.9927686206902984e-05, + "loss": 0.4213, + "step": 1112 + }, + { + "epoch": 0.07, + "learning_rate": 1.992745113389935e-05, + "loss": 0.4604, + "step": 1113 + }, + { + "epoch": 0.07, + "learning_rate": 1.992721568082642e-05, + "loss": 0.4498, + "step": 1114 + }, + { + "epoch": 0.07, + "learning_rate": 1.992697984769321e-05, + "loss": 0.4088, + "step": 1115 + }, + { + "epoch": 0.07, + "learning_rate": 1.9926743634508753e-05, + "loss": 0.447, + "step": 1116 + }, + { + "epoch": 0.07, + "learning_rate": 1.9926507041282088e-05, + "loss": 0.4026, + "step": 1117 + }, + { + "epoch": 0.07, + "learning_rate": 1.9926270068022272e-05, + "loss": 0.4358, + "step": 1118 + }, + { + "epoch": 0.07, + "learning_rate": 1.992603271473838e-05, + "loss": 0.3809, + "step": 1119 + }, + { + "epoch": 0.07, + "learning_rate": 1.9925794981439504e-05, + "loss": 0.4355, + "step": 1120 + }, + { + "epoch": 0.07, + "learning_rate": 1.9925556868134738e-05, + "loss": 0.4333, + "step": 1121 + }, + { + "epoch": 0.07, + "learning_rate": 1.9925318374833197e-05, + "loss": 0.3898, + "step": 1122 + }, + { + "epoch": 0.07, + "learning_rate": 1.9925079501544016e-05, + "loss": 0.4135, + "step": 1123 + }, + { + "epoch": 0.07, + "learning_rate": 1.992484024827634e-05, + "loss": 0.4051, + "step": 1124 + }, + { + "epoch": 0.07, + "learning_rate": 1.992460061503933e-05, + "loss": 0.4082, + "step": 1125 + }, + { + "epoch": 0.07, + "learning_rate": 1.9924360601842156e-05, + "loss": 0.411, + "step": 1126 + }, + { + "epoch": 0.07, + "learning_rate": 1.992412020869401e-05, + "loss": 0.4238, + "step": 1127 + }, + { + "epoch": 0.07, + "learning_rate": 1.992387943560409e-05, + "loss": 0.3789, + "step": 1128 + }, + { + "epoch": 0.07, + "learning_rate": 1.9923638282581624e-05, + "loss": 0.4202, + "step": 1129 + }, + { + "epoch": 0.07, + "learning_rate": 1.9923396749635836e-05, + "loss": 0.4216, + "step": 1130 + }, + { + "epoch": 0.07, + "learning_rate": 1.9923154836775978e-05, + "loss": 0.4185, + "step": 1131 + }, + { + "epoch": 0.07, + "learning_rate": 1.9922912544011306e-05, + "loss": 0.3965, + "step": 1132 + }, + { + "epoch": 0.07, + "learning_rate": 1.99226698713511e-05, + "loss": 0.3901, + "step": 1133 + }, + { + "epoch": 0.07, + "learning_rate": 1.992242681880465e-05, + "loss": 0.4425, + "step": 1134 + }, + { + "epoch": 0.07, + "learning_rate": 1.9922183386381266e-05, + "loss": 0.4099, + "step": 1135 + }, + { + "epoch": 0.07, + "learning_rate": 1.9921939574090258e-05, + "loss": 0.4012, + "step": 1136 + }, + { + "epoch": 0.07, + "learning_rate": 1.9921695381940965e-05, + "loss": 0.4487, + "step": 1137 + }, + { + "epoch": 0.07, + "learning_rate": 1.992145080994274e-05, + "loss": 0.3892, + "step": 1138 + }, + { + "epoch": 0.07, + "learning_rate": 1.992120585810494e-05, + "loss": 0.4169, + "step": 1139 + }, + { + "epoch": 0.07, + "learning_rate": 1.9920960526436945e-05, + "loss": 0.4026, + "step": 1140 + }, + { + "epoch": 0.07, + "learning_rate": 1.9920714814948148e-05, + "loss": 0.4099, + "step": 1141 + }, + { + "epoch": 0.07, + "learning_rate": 1.9920468723647954e-05, + "loss": 0.3906, + "step": 1142 + }, + { + "epoch": 0.07, + "learning_rate": 1.992022225254579e-05, + "loss": 0.4406, + "step": 1143 + }, + { + "epoch": 0.07, + "learning_rate": 1.9919975401651084e-05, + "loss": 0.4157, + "step": 1144 + }, + { + "epoch": 0.07, + "learning_rate": 1.9919728170973297e-05, + "loss": 0.3951, + "step": 1145 + }, + { + "epoch": 0.07, + "learning_rate": 1.991948056052188e-05, + "loss": 0.4021, + "step": 1146 + }, + { + "epoch": 0.07, + "learning_rate": 1.991923257030633e-05, + "loss": 0.4043, + "step": 1147 + }, + { + "epoch": 0.07, + "learning_rate": 1.9918984200336126e-05, + "loss": 0.4269, + "step": 1148 + }, + { + "epoch": 0.07, + "learning_rate": 1.9918735450620785e-05, + "loss": 0.4118, + "step": 1149 + }, + { + "epoch": 0.07, + "learning_rate": 1.991848632116983e-05, + "loss": 0.4138, + "step": 1150 + }, + { + "epoch": 0.07, + "learning_rate": 1.991823681199279e-05, + "loss": 0.4219, + "step": 1151 + }, + { + "epoch": 0.07, + "learning_rate": 1.991798692309923e-05, + "loss": 0.4244, + "step": 1152 + }, + { + "epoch": 0.07, + "learning_rate": 1.9917736654498712e-05, + "loss": 0.4456, + "step": 1153 + }, + { + "epoch": 0.07, + "learning_rate": 1.9917486006200812e-05, + "loss": 0.394, + "step": 1154 + }, + { + "epoch": 0.07, + "learning_rate": 1.9917234978215135e-05, + "loss": 0.4294, + "step": 1155 + }, + { + "epoch": 0.07, + "learning_rate": 1.9916983570551286e-05, + "loss": 0.3912, + "step": 1156 + }, + { + "epoch": 0.07, + "learning_rate": 1.991673178321889e-05, + "loss": 0.3979, + "step": 1157 + }, + { + "epoch": 0.07, + "learning_rate": 1.991647961622759e-05, + "loss": 0.4222, + "step": 1158 + }, + { + "epoch": 0.07, + "learning_rate": 1.9916227069587034e-05, + "loss": 0.4235, + "step": 1159 + }, + { + "epoch": 0.07, + "learning_rate": 1.9915974143306898e-05, + "loss": 0.3976, + "step": 1160 + }, + { + "epoch": 0.07, + "learning_rate": 1.991572083739686e-05, + "loss": 0.4054, + "step": 1161 + }, + { + "epoch": 0.07, + "learning_rate": 1.9915467151866623e-05, + "loss": 0.4188, + "step": 1162 + }, + { + "epoch": 0.07, + "learning_rate": 1.9915213086725895e-05, + "loss": 0.4439, + "step": 1163 + }, + { + "epoch": 0.07, + "learning_rate": 1.99149586419844e-05, + "loss": 0.4397, + "step": 1164 + }, + { + "epoch": 0.07, + "learning_rate": 1.9914703817651887e-05, + "loss": 0.4121, + "step": 1165 + }, + { + "epoch": 0.07, + "learning_rate": 1.9914448613738107e-05, + "loss": 0.4155, + "step": 1166 + }, + { + "epoch": 0.07, + "learning_rate": 1.991419303025283e-05, + "loss": 0.4054, + "step": 1167 + }, + { + "epoch": 0.07, + "learning_rate": 1.991393706720584e-05, + "loss": 0.4464, + "step": 1168 + }, + { + "epoch": 0.07, + "learning_rate": 1.9913680724606946e-05, + "loss": 0.4475, + "step": 1169 + }, + { + "epoch": 0.07, + "learning_rate": 1.9913424002465947e-05, + "loss": 0.4353, + "step": 1170 + }, + { + "epoch": 0.07, + "learning_rate": 1.9913166900792683e-05, + "loss": 0.4277, + "step": 1171 + }, + { + "epoch": 0.07, + "learning_rate": 1.9912909419596993e-05, + "loss": 0.3976, + "step": 1172 + }, + { + "epoch": 0.07, + "learning_rate": 1.9912651558888733e-05, + "loss": 0.3929, + "step": 1173 + }, + { + "epoch": 0.07, + "learning_rate": 1.991239331867778e-05, + "loss": 0.4118, + "step": 1174 + }, + { + "epoch": 0.07, + "learning_rate": 1.9912134698974016e-05, + "loss": 0.4381, + "step": 1175 + }, + { + "epoch": 0.07, + "learning_rate": 1.991187569978734e-05, + "loss": 0.4096, + "step": 1176 + }, + { + "epoch": 0.07, + "learning_rate": 1.9911616321127678e-05, + "loss": 0.4319, + "step": 1177 + }, + { + "epoch": 0.07, + "learning_rate": 1.991135656300495e-05, + "loss": 0.454, + "step": 1178 + }, + { + "epoch": 0.07, + "learning_rate": 1.9911096425429107e-05, + "loss": 0.423, + "step": 1179 + }, + { + "epoch": 0.07, + "learning_rate": 1.99108359084101e-05, + "loss": 0.416, + "step": 1180 + }, + { + "epoch": 0.07, + "learning_rate": 1.9910575011957917e-05, + "loss": 0.4291, + "step": 1181 + }, + { + "epoch": 0.07, + "learning_rate": 1.9910313736082535e-05, + "loss": 0.4372, + "step": 1182 + }, + { + "epoch": 0.07, + "learning_rate": 1.9910052080793956e-05, + "loss": 0.421, + "step": 1183 + }, + { + "epoch": 0.07, + "learning_rate": 1.9909790046102204e-05, + "loss": 0.4459, + "step": 1184 + }, + { + "epoch": 0.07, + "learning_rate": 1.9909527632017307e-05, + "loss": 0.4057, + "step": 1185 + }, + { + "epoch": 0.07, + "learning_rate": 1.9909264838549313e-05, + "loss": 0.423, + "step": 1186 + }, + { + "epoch": 0.07, + "learning_rate": 1.9909001665708284e-05, + "loss": 0.404, + "step": 1187 + }, + { + "epoch": 0.07, + "learning_rate": 1.990873811350429e-05, + "loss": 0.4241, + "step": 1188 + }, + { + "epoch": 0.07, + "learning_rate": 1.9908474181947427e-05, + "loss": 0.4018, + "step": 1189 + }, + { + "epoch": 0.07, + "learning_rate": 1.99082098710478e-05, + "loss": 0.4244, + "step": 1190 + }, + { + "epoch": 0.07, + "learning_rate": 1.9907945180815523e-05, + "loss": 0.3973, + "step": 1191 + }, + { + "epoch": 0.07, + "learning_rate": 1.9907680111260733e-05, + "loss": 0.3797, + "step": 1192 + }, + { + "epoch": 0.07, + "learning_rate": 1.9907414662393575e-05, + "loss": 0.4294, + "step": 1193 + }, + { + "epoch": 0.07, + "learning_rate": 1.9907148834224216e-05, + "loss": 0.4501, + "step": 1194 + }, + { + "epoch": 0.07, + "learning_rate": 1.990688262676283e-05, + "loss": 0.4082, + "step": 1195 + }, + { + "epoch": 0.07, + "learning_rate": 1.9906616040019614e-05, + "loss": 0.3842, + "step": 1196 + }, + { + "epoch": 0.07, + "learning_rate": 1.9906349074004764e-05, + "loss": 0.3996, + "step": 1197 + }, + { + "epoch": 0.07, + "learning_rate": 1.9906081728728506e-05, + "loss": 0.3998, + "step": 1198 + }, + { + "epoch": 0.07, + "learning_rate": 1.990581400420108e-05, + "loss": 0.4076, + "step": 1199 + }, + { + "epoch": 0.07, + "learning_rate": 1.990554590043273e-05, + "loss": 0.3817, + "step": 1200 + }, + { + "epoch": 0.07, + "learning_rate": 1.9905277417433724e-05, + "loss": 0.3884, + "step": 1201 + }, + { + "epoch": 0.07, + "learning_rate": 1.9905008555214337e-05, + "loss": 0.375, + "step": 1202 + }, + { + "epoch": 0.07, + "learning_rate": 1.9904739313784862e-05, + "loss": 0.4261, + "step": 1203 + }, + { + "epoch": 0.07, + "learning_rate": 1.990446969315561e-05, + "loss": 0.4129, + "step": 1204 + }, + { + "epoch": 0.07, + "learning_rate": 1.99041996933369e-05, + "loss": 0.387, + "step": 1205 + }, + { + "epoch": 0.07, + "learning_rate": 1.9903929314339074e-05, + "loss": 0.4155, + "step": 1206 + }, + { + "epoch": 0.07, + "learning_rate": 1.990365855617248e-05, + "loss": 0.3881, + "step": 1207 + }, + { + "epoch": 0.07, + "learning_rate": 1.9903387418847483e-05, + "loss": 0.3965, + "step": 1208 + }, + { + "epoch": 0.07, + "learning_rate": 1.990311590237447e-05, + "loss": 0.3859, + "step": 1209 + }, + { + "epoch": 0.07, + "learning_rate": 1.9902844006763825e-05, + "loss": 0.4012, + "step": 1210 + }, + { + "epoch": 0.07, + "learning_rate": 1.990257173202596e-05, + "loss": 0.4088, + "step": 1211 + }, + { + "epoch": 0.07, + "learning_rate": 1.9902299078171307e-05, + "loss": 0.3987, + "step": 1212 + }, + { + "epoch": 0.07, + "learning_rate": 1.99020260452103e-05, + "loss": 0.4249, + "step": 1213 + }, + { + "epoch": 0.07, + "learning_rate": 1.990175263315339e-05, + "loss": 0.4495, + "step": 1214 + }, + { + "epoch": 0.07, + "learning_rate": 1.990147884201104e-05, + "loss": 0.3954, + "step": 1215 + }, + { + "epoch": 0.07, + "learning_rate": 1.9901204671793746e-05, + "loss": 0.4035, + "step": 1216 + }, + { + "epoch": 0.07, + "learning_rate": 1.9900930122511993e-05, + "loss": 0.4149, + "step": 1217 + }, + { + "epoch": 0.07, + "learning_rate": 1.9900655194176294e-05, + "loss": 0.356, + "step": 1218 + }, + { + "epoch": 0.07, + "learning_rate": 1.990037988679718e-05, + "loss": 0.3811, + "step": 1219 + }, + { + "epoch": 0.07, + "learning_rate": 1.9900104200385183e-05, + "loss": 0.4129, + "step": 1220 + }, + { + "epoch": 0.07, + "learning_rate": 1.9899828134950863e-05, + "loss": 0.4155, + "step": 1221 + }, + { + "epoch": 0.07, + "learning_rate": 1.9899551690504787e-05, + "loss": 0.3753, + "step": 1222 + }, + { + "epoch": 0.07, + "learning_rate": 1.9899274867057545e-05, + "loss": 0.4562, + "step": 1223 + }, + { + "epoch": 0.07, + "learning_rate": 1.9898997664619722e-05, + "loss": 0.3934, + "step": 1224 + }, + { + "epoch": 0.07, + "learning_rate": 1.989872008320194e-05, + "loss": 0.4015, + "step": 1225 + }, + { + "epoch": 0.07, + "learning_rate": 1.9898442122814825e-05, + "loss": 0.4065, + "step": 1226 + }, + { + "epoch": 0.07, + "learning_rate": 1.9898163783469014e-05, + "loss": 0.4224, + "step": 1227 + }, + { + "epoch": 0.07, + "learning_rate": 1.9897885065175174e-05, + "loss": 0.4302, + "step": 1228 + }, + { + "epoch": 0.07, + "learning_rate": 1.9897605967943964e-05, + "loss": 0.3758, + "step": 1229 + }, + { + "epoch": 0.07, + "learning_rate": 1.989732649178607e-05, + "loss": 0.4029, + "step": 1230 + }, + { + "epoch": 0.07, + "learning_rate": 1.9897046636712202e-05, + "loss": 0.435, + "step": 1231 + }, + { + "epoch": 0.07, + "learning_rate": 1.9896766402733063e-05, + "loss": 0.4121, + "step": 1232 + }, + { + "epoch": 0.07, + "learning_rate": 1.9896485789859386e-05, + "loss": 0.3862, + "step": 1233 + }, + { + "epoch": 0.07, + "learning_rate": 1.9896204798101916e-05, + "loss": 0.4436, + "step": 1234 + }, + { + "epoch": 0.07, + "learning_rate": 1.989592342747141e-05, + "loss": 0.3898, + "step": 1235 + }, + { + "epoch": 0.07, + "learning_rate": 1.9895641677978637e-05, + "loss": 0.3987, + "step": 1236 + }, + { + "epoch": 0.07, + "learning_rate": 1.9895359549634388e-05, + "loss": 0.4035, + "step": 1237 + }, + { + "epoch": 0.07, + "learning_rate": 1.9895077042449462e-05, + "loss": 0.4146, + "step": 1238 + }, + { + "epoch": 0.07, + "learning_rate": 1.989479415643467e-05, + "loss": 0.3931, + "step": 1239 + }, + { + "epoch": 0.07, + "learning_rate": 1.9894510891600854e-05, + "loss": 0.4127, + "step": 1240 + }, + { + "epoch": 0.07, + "learning_rate": 1.9894227247958847e-05, + "loss": 0.3778, + "step": 1241 + }, + { + "epoch": 0.08, + "learning_rate": 1.9893943225519514e-05, + "loss": 0.4347, + "step": 1242 + }, + { + "epoch": 0.08, + "learning_rate": 1.9893658824293725e-05, + "loss": 0.423, + "step": 1243 + }, + { + "epoch": 0.08, + "learning_rate": 1.9893374044292377e-05, + "loss": 0.4185, + "step": 1244 + }, + { + "epoch": 0.08, + "learning_rate": 1.9893088885526358e-05, + "loss": 0.4269, + "step": 1245 + }, + { + "epoch": 0.08, + "learning_rate": 1.9892803348006602e-05, + "loss": 0.4196, + "step": 1246 + }, + { + "epoch": 0.08, + "learning_rate": 1.9892517431744026e-05, + "loss": 0.3764, + "step": 1247 + }, + { + "epoch": 0.08, + "learning_rate": 1.9892231136749584e-05, + "loss": 0.4085, + "step": 1248 + }, + { + "epoch": 0.08, + "learning_rate": 1.989194446303424e-05, + "loss": 0.4141, + "step": 1249 + }, + { + "epoch": 0.08, + "learning_rate": 1.989165741060896e-05, + "loss": 0.4467, + "step": 1250 + }, + { + "epoch": 0.08, + "learning_rate": 1.989136997948474e-05, + "loss": 0.4037, + "step": 1251 + }, + { + "epoch": 0.08, + "learning_rate": 1.989108216967258e-05, + "loss": 0.3973, + "step": 1252 + }, + { + "epoch": 0.08, + "learning_rate": 1.98907939811835e-05, + "loss": 0.4079, + "step": 1253 + }, + { + "epoch": 0.08, + "learning_rate": 1.9890505414028538e-05, + "loss": 0.4417, + "step": 1254 + }, + { + "epoch": 0.08, + "learning_rate": 1.9890216468218738e-05, + "loss": 0.3823, + "step": 1255 + }, + { + "epoch": 0.08, + "learning_rate": 1.988992714376516e-05, + "loss": 0.4534, + "step": 1256 + }, + { + "epoch": 0.08, + "learning_rate": 1.9889637440678884e-05, + "loss": 0.4375, + "step": 1257 + }, + { + "epoch": 0.08, + "learning_rate": 1.9889347358971e-05, + "loss": 0.4222, + "step": 1258 + }, + { + "epoch": 0.08, + "learning_rate": 1.9889056898652614e-05, + "loss": 0.3926, + "step": 1259 + }, + { + "epoch": 0.08, + "learning_rate": 1.9888766059734846e-05, + "loss": 0.3951, + "step": 1260 + }, + { + "epoch": 0.08, + "learning_rate": 1.988847484222883e-05, + "loss": 0.4422, + "step": 1261 + }, + { + "epoch": 0.08, + "learning_rate": 1.988818324614572e-05, + "loss": 0.3915, + "step": 1262 + }, + { + "epoch": 0.08, + "learning_rate": 1.988789127149667e-05, + "loss": 0.4096, + "step": 1263 + }, + { + "epoch": 0.08, + "learning_rate": 1.9887598918292864e-05, + "loss": 0.3873, + "step": 1264 + }, + { + "epoch": 0.08, + "learning_rate": 1.9887306186545496e-05, + "loss": 0.4191, + "step": 1265 + }, + { + "epoch": 0.08, + "learning_rate": 1.9887013076265773e-05, + "loss": 0.3996, + "step": 1266 + }, + { + "epoch": 0.08, + "learning_rate": 1.9886719587464914e-05, + "loss": 0.4099, + "step": 1267 + }, + { + "epoch": 0.08, + "learning_rate": 1.9886425720154152e-05, + "loss": 0.375, + "step": 1268 + }, + { + "epoch": 0.08, + "learning_rate": 1.9886131474344746e-05, + "loss": 0.404, + "step": 1269 + }, + { + "epoch": 0.08, + "learning_rate": 1.9885836850047958e-05, + "loss": 0.416, + "step": 1270 + }, + { + "epoch": 0.08, + "learning_rate": 1.9885541847275067e-05, + "loss": 0.4127, + "step": 1271 + }, + { + "epoch": 0.08, + "learning_rate": 1.988524646603736e-05, + "loss": 0.3976, + "step": 1272 + }, + { + "epoch": 0.08, + "learning_rate": 1.988495070634616e-05, + "loss": 0.3996, + "step": 1273 + }, + { + "epoch": 0.08, + "learning_rate": 1.9884654568212775e-05, + "loss": 0.3853, + "step": 1274 + }, + { + "epoch": 0.08, + "learning_rate": 1.9884358051648553e-05, + "loss": 0.3845, + "step": 1275 + }, + { + "epoch": 0.08, + "learning_rate": 1.9884061156664844e-05, + "loss": 0.3711, + "step": 1276 + }, + { + "epoch": 0.08, + "learning_rate": 1.9883763883273014e-05, + "loss": 0.4263, + "step": 1277 + }, + { + "epoch": 0.08, + "learning_rate": 1.9883466231484442e-05, + "loss": 0.3783, + "step": 1278 + }, + { + "epoch": 0.08, + "learning_rate": 1.9883168201310525e-05, + "loss": 0.4222, + "step": 1279 + }, + { + "epoch": 0.08, + "learning_rate": 1.9882869792762673e-05, + "loss": 0.3775, + "step": 1280 + }, + { + "epoch": 0.08, + "learning_rate": 1.988257100585231e-05, + "loss": 0.4261, + "step": 1281 + }, + { + "epoch": 0.08, + "learning_rate": 1.9882271840590874e-05, + "loss": 0.4138, + "step": 1282 + }, + { + "epoch": 0.08, + "learning_rate": 1.9881972296989823e-05, + "loss": 0.3786, + "step": 1283 + }, + { + "epoch": 0.08, + "learning_rate": 1.988167237506062e-05, + "loss": 0.4079, + "step": 1284 + }, + { + "epoch": 0.08, + "learning_rate": 1.988137207481475e-05, + "loss": 0.4169, + "step": 1285 + }, + { + "epoch": 0.08, + "learning_rate": 1.9881071396263706e-05, + "loss": 0.4227, + "step": 1286 + }, + { + "epoch": 0.08, + "learning_rate": 1.9880770339419006e-05, + "loss": 0.4054, + "step": 1287 + }, + { + "epoch": 0.08, + "learning_rate": 1.988046890429217e-05, + "loss": 0.3962, + "step": 1288 + }, + { + "epoch": 0.08, + "learning_rate": 1.988016709089474e-05, + "loss": 0.4054, + "step": 1289 + }, + { + "epoch": 0.08, + "learning_rate": 1.9879864899238272e-05, + "loss": 0.3954, + "step": 1290 + }, + { + "epoch": 0.08, + "learning_rate": 1.9879562329334336e-05, + "loss": 0.4127, + "step": 1291 + }, + { + "epoch": 0.08, + "learning_rate": 1.9879259381194516e-05, + "loss": 0.4157, + "step": 1292 + }, + { + "epoch": 0.08, + "learning_rate": 1.9878956054830408e-05, + "loss": 0.4143, + "step": 1293 + }, + { + "epoch": 0.08, + "learning_rate": 1.987865235025362e-05, + "loss": 0.3943, + "step": 1294 + }, + { + "epoch": 0.08, + "learning_rate": 1.987834826747579e-05, + "loss": 0.4012, + "step": 1295 + }, + { + "epoch": 0.08, + "learning_rate": 1.9878043806508556e-05, + "loss": 0.3806, + "step": 1296 + }, + { + "epoch": 0.08, + "learning_rate": 1.9877738967363568e-05, + "loss": 0.3937, + "step": 1297 + }, + { + "epoch": 0.08, + "learning_rate": 1.9877433750052507e-05, + "loss": 0.4372, + "step": 1298 + }, + { + "epoch": 0.08, + "learning_rate": 1.9877128154587046e-05, + "loss": 0.3878, + "step": 1299 + }, + { + "epoch": 0.08, + "learning_rate": 1.9876822180978895e-05, + "loss": 0.4138, + "step": 1300 + }, + { + "epoch": 0.08, + "learning_rate": 1.9876515829239763e-05, + "loss": 0.3845, + "step": 1301 + }, + { + "epoch": 0.08, + "learning_rate": 1.9876209099381382e-05, + "loss": 0.4051, + "step": 1302 + }, + { + "epoch": 0.08, + "learning_rate": 1.9875901991415493e-05, + "loss": 0.4021, + "step": 1303 + }, + { + "epoch": 0.08, + "learning_rate": 1.987559450535385e-05, + "loss": 0.4425, + "step": 1304 + }, + { + "epoch": 0.08, + "learning_rate": 1.9875286641208228e-05, + "loss": 0.4141, + "step": 1305 + }, + { + "epoch": 0.08, + "learning_rate": 1.9874978398990416e-05, + "loss": 0.3884, + "step": 1306 + }, + { + "epoch": 0.08, + "learning_rate": 1.9874669778712215e-05, + "loss": 0.4051, + "step": 1307 + }, + { + "epoch": 0.08, + "learning_rate": 1.9874360780385435e-05, + "loss": 0.4121, + "step": 1308 + }, + { + "epoch": 0.08, + "learning_rate": 1.9874051404021908e-05, + "loss": 0.4088, + "step": 1309 + }, + { + "epoch": 0.08, + "learning_rate": 1.9873741649633486e-05, + "loss": 0.4261, + "step": 1310 + }, + { + "epoch": 0.08, + "learning_rate": 1.9873431517232013e-05, + "loss": 0.404, + "step": 1311 + }, + { + "epoch": 0.08, + "learning_rate": 1.9873121006829375e-05, + "loss": 0.4573, + "step": 1312 + }, + { + "epoch": 0.08, + "learning_rate": 1.9872810118437457e-05, + "loss": 0.4118, + "step": 1313 + }, + { + "epoch": 0.08, + "learning_rate": 1.987249885206816e-05, + "loss": 0.411, + "step": 1314 + }, + { + "epoch": 0.08, + "learning_rate": 1.9872187207733394e-05, + "loss": 0.4487, + "step": 1315 + }, + { + "epoch": 0.08, + "learning_rate": 1.98718751854451e-05, + "loss": 0.3937, + "step": 1316 + }, + { + "epoch": 0.08, + "learning_rate": 1.9871562785215223e-05, + "loss": 0.4054, + "step": 1317 + }, + { + "epoch": 0.08, + "learning_rate": 1.987125000705572e-05, + "loss": 0.4386, + "step": 1318 + }, + { + "epoch": 0.08, + "learning_rate": 1.9870936850978567e-05, + "loss": 0.4171, + "step": 1319 + }, + { + "epoch": 0.08, + "learning_rate": 1.987062331699575e-05, + "loss": 0.3873, + "step": 1320 + }, + { + "epoch": 0.08, + "learning_rate": 1.9870309405119274e-05, + "loss": 0.4135, + "step": 1321 + }, + { + "epoch": 0.08, + "learning_rate": 1.986999511536116e-05, + "loss": 0.4049, + "step": 1322 + }, + { + "epoch": 0.08, + "learning_rate": 1.9869680447733436e-05, + "loss": 0.4127, + "step": 1323 + }, + { + "epoch": 0.08, + "learning_rate": 1.986936540224815e-05, + "loss": 0.4054, + "step": 1324 + }, + { + "epoch": 0.08, + "learning_rate": 1.986904997891737e-05, + "loss": 0.4088, + "step": 1325 + }, + { + "epoch": 0.08, + "learning_rate": 1.9868734177753165e-05, + "loss": 0.4023, + "step": 1326 + }, + { + "epoch": 0.08, + "learning_rate": 1.9868417998767625e-05, + "loss": 0.4051, + "step": 1327 + }, + { + "epoch": 0.08, + "learning_rate": 1.9868101441972857e-05, + "loss": 0.3987, + "step": 1328 + }, + { + "epoch": 0.08, + "learning_rate": 1.9867784507380978e-05, + "loss": 0.4461, + "step": 1329 + }, + { + "epoch": 0.08, + "learning_rate": 1.9867467195004127e-05, + "loss": 0.418, + "step": 1330 + }, + { + "epoch": 0.08, + "learning_rate": 1.9867149504854448e-05, + "loss": 0.3881, + "step": 1331 + }, + { + "epoch": 0.08, + "learning_rate": 1.9866831436944107e-05, + "loss": 0.3878, + "step": 1332 + }, + { + "epoch": 0.08, + "learning_rate": 1.986651299128527e-05, + "loss": 0.4308, + "step": 1333 + }, + { + "epoch": 0.08, + "learning_rate": 1.9866194167890146e-05, + "loss": 0.3982, + "step": 1334 + }, + { + "epoch": 0.08, + "learning_rate": 1.986587496677093e-05, + "loss": 0.4076, + "step": 1335 + }, + { + "epoch": 0.08, + "learning_rate": 1.9865555387939845e-05, + "loss": 0.3987, + "step": 1336 + }, + { + "epoch": 0.08, + "learning_rate": 1.9865235431409123e-05, + "loss": 0.4219, + "step": 1337 + }, + { + "epoch": 0.08, + "learning_rate": 1.9864915097191018e-05, + "loss": 0.4263, + "step": 1338 + }, + { + "epoch": 0.08, + "learning_rate": 1.9864594385297795e-05, + "loss": 0.3792, + "step": 1339 + }, + { + "epoch": 0.08, + "learning_rate": 1.9864273295741725e-05, + "loss": 0.4088, + "step": 1340 + }, + { + "epoch": 0.08, + "learning_rate": 1.9863951828535107e-05, + "loss": 0.4171, + "step": 1341 + }, + { + "epoch": 0.08, + "learning_rate": 1.9863629983690248e-05, + "loss": 0.4291, + "step": 1342 + }, + { + "epoch": 0.08, + "learning_rate": 1.9863307761219466e-05, + "loss": 0.3962, + "step": 1343 + }, + { + "epoch": 0.08, + "learning_rate": 1.98629851611351e-05, + "loss": 0.3968, + "step": 1344 + }, + { + "epoch": 0.08, + "learning_rate": 1.98626621834495e-05, + "loss": 0.3968, + "step": 1345 + }, + { + "epoch": 0.08, + "learning_rate": 1.986233882817503e-05, + "loss": 0.3931, + "step": 1346 + }, + { + "epoch": 0.08, + "learning_rate": 1.9862015095324073e-05, + "loss": 0.3811, + "step": 1347 + }, + { + "epoch": 0.08, + "learning_rate": 1.9861690984909022e-05, + "loss": 0.409, + "step": 1348 + }, + { + "epoch": 0.08, + "learning_rate": 1.9861366496942283e-05, + "loss": 0.4026, + "step": 1349 + }, + { + "epoch": 0.08, + "learning_rate": 1.9861041631436276e-05, + "loss": 0.4082, + "step": 1350 + }, + { + "epoch": 0.08, + "learning_rate": 1.986071638840345e-05, + "loss": 0.4227, + "step": 1351 + }, + { + "epoch": 0.08, + "learning_rate": 1.9860390767856244e-05, + "loss": 0.4648, + "step": 1352 + }, + { + "epoch": 0.08, + "learning_rate": 1.9860064769807132e-05, + "loss": 0.4185, + "step": 1353 + }, + { + "epoch": 0.08, + "learning_rate": 1.985973839426859e-05, + "loss": 0.3749, + "step": 1354 + }, + { + "epoch": 0.08, + "learning_rate": 1.9859411641253118e-05, + "loss": 0.3984, + "step": 1355 + }, + { + "epoch": 0.08, + "learning_rate": 1.9859084510773224e-05, + "loss": 0.421, + "step": 1356 + }, + { + "epoch": 0.08, + "learning_rate": 1.985875700284143e-05, + "loss": 0.416, + "step": 1357 + }, + { + "epoch": 0.08, + "learning_rate": 1.9858429117470272e-05, + "loss": 0.4185, + "step": 1358 + }, + { + "epoch": 0.08, + "learning_rate": 1.9858100854672312e-05, + "loss": 0.3881, + "step": 1359 + }, + { + "epoch": 0.08, + "learning_rate": 1.985777221446011e-05, + "loss": 0.4012, + "step": 1360 + }, + { + "epoch": 0.08, + "learning_rate": 1.985744319684625e-05, + "loss": 0.3968, + "step": 1361 + }, + { + "epoch": 0.08, + "learning_rate": 1.985711380184333e-05, + "loss": 0.4118, + "step": 1362 + }, + { + "epoch": 0.08, + "learning_rate": 1.985678402946396e-05, + "loss": 0.4275, + "step": 1363 + }, + { + "epoch": 0.08, + "learning_rate": 1.9856453879720762e-05, + "loss": 0.4283, + "step": 1364 + }, + { + "epoch": 0.08, + "learning_rate": 1.9856123352626376e-05, + "loss": 0.3895, + "step": 1365 + }, + { + "epoch": 0.08, + "learning_rate": 1.9855792448193463e-05, + "loss": 0.3859, + "step": 1366 + }, + { + "epoch": 0.08, + "learning_rate": 1.9855461166434683e-05, + "loss": 0.387, + "step": 1367 + }, + { + "epoch": 0.08, + "learning_rate": 1.9855129507362725e-05, + "loss": 0.4046, + "step": 1368 + }, + { + "epoch": 0.08, + "learning_rate": 1.9854797470990282e-05, + "loss": 0.4294, + "step": 1369 + }, + { + "epoch": 0.08, + "learning_rate": 1.9854465057330072e-05, + "loss": 0.385, + "step": 1370 + }, + { + "epoch": 0.08, + "learning_rate": 1.9854132266394817e-05, + "loss": 0.3862, + "step": 1371 + }, + { + "epoch": 0.08, + "learning_rate": 1.9853799098197258e-05, + "loss": 0.3795, + "step": 1372 + }, + { + "epoch": 0.08, + "learning_rate": 1.9853465552750147e-05, + "loss": 0.3926, + "step": 1373 + }, + { + "epoch": 0.08, + "learning_rate": 1.9853131630066264e-05, + "loss": 0.4023, + "step": 1374 + }, + { + "epoch": 0.08, + "learning_rate": 1.985279733015838e-05, + "loss": 0.4071, + "step": 1375 + }, + { + "epoch": 0.08, + "learning_rate": 1.98524626530393e-05, + "loss": 0.4079, + "step": 1376 + }, + { + "epoch": 0.08, + "learning_rate": 1.9852127598721844e-05, + "loss": 0.3322, + "step": 1377 + }, + { + "epoch": 0.08, + "learning_rate": 1.9851792167218825e-05, + "loss": 0.4258, + "step": 1378 + }, + { + "epoch": 0.08, + "learning_rate": 1.9851456358543095e-05, + "loss": 0.4007, + "step": 1379 + }, + { + "epoch": 0.08, + "learning_rate": 1.985112017270751e-05, + "loss": 0.4255, + "step": 1380 + }, + { + "epoch": 0.08, + "learning_rate": 1.9850783609724934e-05, + "loss": 0.4032, + "step": 1381 + }, + { + "epoch": 0.08, + "learning_rate": 1.9850446669608257e-05, + "loss": 0.4012, + "step": 1382 + }, + { + "epoch": 0.08, + "learning_rate": 1.985010935237038e-05, + "loss": 0.3931, + "step": 1383 + }, + { + "epoch": 0.08, + "learning_rate": 1.9849771658024218e-05, + "loss": 0.4076, + "step": 1384 + }, + { + "epoch": 0.08, + "learning_rate": 1.9849433586582693e-05, + "loss": 0.3943, + "step": 1385 + }, + { + "epoch": 0.08, + "learning_rate": 1.9849095138058753e-05, + "loss": 0.4191, + "step": 1386 + }, + { + "epoch": 0.08, + "learning_rate": 1.9848756312465353e-05, + "loss": 0.4191, + "step": 1387 + }, + { + "epoch": 0.08, + "learning_rate": 1.984841710981547e-05, + "loss": 0.4456, + "step": 1388 + }, + { + "epoch": 0.08, + "learning_rate": 1.9848077530122083e-05, + "loss": 0.4249, + "step": 1389 + }, + { + "epoch": 0.08, + "learning_rate": 1.9847737573398196e-05, + "loss": 0.394, + "step": 1390 + }, + { + "epoch": 0.08, + "learning_rate": 1.9847397239656826e-05, + "loss": 0.4302, + "step": 1391 + }, + { + "epoch": 0.08, + "learning_rate": 1.9847056528910996e-05, + "loss": 0.404, + "step": 1392 + }, + { + "epoch": 0.08, + "learning_rate": 1.9846715441173758e-05, + "loss": 0.4308, + "step": 1393 + }, + { + "epoch": 0.08, + "learning_rate": 1.984637397645817e-05, + "loss": 0.4054, + "step": 1394 + }, + { + "epoch": 0.08, + "learning_rate": 1.9846032134777298e-05, + "loss": 0.4116, + "step": 1395 + }, + { + "epoch": 0.08, + "learning_rate": 1.9845689916144234e-05, + "loss": 0.3664, + "step": 1396 + }, + { + "epoch": 0.08, + "learning_rate": 1.984534732057208e-05, + "loss": 0.4378, + "step": 1397 + }, + { + "epoch": 0.08, + "learning_rate": 1.984500434807395e-05, + "loss": 0.3956, + "step": 1398 + }, + { + "epoch": 0.08, + "learning_rate": 1.984466099866298e-05, + "loss": 0.397, + "step": 1399 + }, + { + "epoch": 0.08, + "learning_rate": 1.9844317272352306e-05, + "loss": 0.4096, + "step": 1400 + }, + { + "epoch": 0.08, + "learning_rate": 1.984397316915509e-05, + "loss": 0.3878, + "step": 1401 + }, + { + "epoch": 0.08, + "learning_rate": 1.9843628689084516e-05, + "loss": 0.4375, + "step": 1402 + }, + { + "epoch": 0.08, + "learning_rate": 1.9843283832153757e-05, + "loss": 0.4328, + "step": 1403 + }, + { + "epoch": 0.08, + "learning_rate": 1.9842938598376027e-05, + "loss": 0.4177, + "step": 1404 + }, + { + "epoch": 0.08, + "learning_rate": 1.984259298776454e-05, + "loss": 0.3931, + "step": 1405 + }, + { + "epoch": 0.08, + "learning_rate": 1.9842247000332528e-05, + "loss": 0.4051, + "step": 1406 + }, + { + "epoch": 0.08, + "learning_rate": 1.9841900636093233e-05, + "loss": 0.3834, + "step": 1407 + }, + { + "epoch": 0.09, + "learning_rate": 1.9841553895059917e-05, + "loss": 0.3853, + "step": 1408 + }, + { + "epoch": 0.09, + "learning_rate": 1.9841206777245857e-05, + "loss": 0.3931, + "step": 1409 + }, + { + "epoch": 0.09, + "learning_rate": 1.9840859282664344e-05, + "loss": 0.344, + "step": 1410 + }, + { + "epoch": 0.09, + "learning_rate": 1.9840511411328677e-05, + "loss": 0.3884, + "step": 1411 + }, + { + "epoch": 0.09, + "learning_rate": 1.984016316325218e-05, + "loss": 0.4049, + "step": 1412 + }, + { + "epoch": 0.09, + "learning_rate": 1.983981453844818e-05, + "loss": 0.4012, + "step": 1413 + }, + { + "epoch": 0.09, + "learning_rate": 1.9839465536930024e-05, + "loss": 0.4129, + "step": 1414 + }, + { + "epoch": 0.09, + "learning_rate": 1.983911615871108e-05, + "loss": 0.3898, + "step": 1415 + }, + { + "epoch": 0.09, + "learning_rate": 1.9838766403804713e-05, + "loss": 0.3931, + "step": 1416 + }, + { + "epoch": 0.09, + "learning_rate": 1.9838416272224324e-05, + "loss": 0.4129, + "step": 1417 + }, + { + "epoch": 0.09, + "learning_rate": 1.983806576398331e-05, + "loss": 0.4188, + "step": 1418 + }, + { + "epoch": 0.09, + "learning_rate": 1.98377148790951e-05, + "loss": 0.406, + "step": 1419 + }, + { + "epoch": 0.09, + "learning_rate": 1.9837363617573114e-05, + "loss": 0.3998, + "step": 1420 + }, + { + "epoch": 0.09, + "learning_rate": 1.9837011979430807e-05, + "loss": 0.4035, + "step": 1421 + }, + { + "epoch": 0.09, + "learning_rate": 1.9836659964681643e-05, + "loss": 0.4169, + "step": 1422 + }, + { + "epoch": 0.09, + "learning_rate": 1.9836307573339096e-05, + "loss": 0.4124, + "step": 1423 + }, + { + "epoch": 0.09, + "learning_rate": 1.983595480541666e-05, + "loss": 0.4035, + "step": 1424 + }, + { + "epoch": 0.09, + "learning_rate": 1.9835601660927838e-05, + "loss": 0.4233, + "step": 1425 + }, + { + "epoch": 0.09, + "learning_rate": 1.983524813988615e-05, + "loss": 0.4124, + "step": 1426 + }, + { + "epoch": 0.09, + "learning_rate": 1.983489424230513e-05, + "loss": 0.3954, + "step": 1427 + }, + { + "epoch": 0.09, + "learning_rate": 1.9834539968198328e-05, + "loss": 0.3719, + "step": 1428 + }, + { + "epoch": 0.09, + "learning_rate": 1.983418531757931e-05, + "loss": 0.3862, + "step": 1429 + }, + { + "epoch": 0.09, + "learning_rate": 1.983383029046165e-05, + "loss": 0.3954, + "step": 1430 + }, + { + "epoch": 0.09, + "learning_rate": 1.983347488685894e-05, + "loss": 0.3761, + "step": 1431 + }, + { + "epoch": 0.09, + "learning_rate": 1.983311910678479e-05, + "loss": 0.3881, + "step": 1432 + }, + { + "epoch": 0.09, + "learning_rate": 1.9832762950252813e-05, + "loss": 0.3856, + "step": 1433 + }, + { + "epoch": 0.09, + "learning_rate": 1.9832406417276653e-05, + "loss": 0.3926, + "step": 1434 + }, + { + "epoch": 0.09, + "learning_rate": 1.9832049507869957e-05, + "loss": 0.3834, + "step": 1435 + }, + { + "epoch": 0.09, + "learning_rate": 1.983169222204639e-05, + "loss": 0.3973, + "step": 1436 + }, + { + "epoch": 0.09, + "learning_rate": 1.9831334559819627e-05, + "loss": 0.3903, + "step": 1437 + }, + { + "epoch": 0.09, + "learning_rate": 1.9830976521203365e-05, + "loss": 0.3753, + "step": 1438 + }, + { + "epoch": 0.09, + "learning_rate": 1.9830618106211307e-05, + "loss": 0.394, + "step": 1439 + }, + { + "epoch": 0.09, + "learning_rate": 1.983025931485718e-05, + "loss": 0.3895, + "step": 1440 + }, + { + "epoch": 0.09, + "learning_rate": 1.9829900147154717e-05, + "loss": 0.3756, + "step": 1441 + }, + { + "epoch": 0.09, + "learning_rate": 1.9829540603117667e-05, + "loss": 0.4191, + "step": 1442 + }, + { + "epoch": 0.09, + "learning_rate": 1.98291806827598e-05, + "loss": 0.399, + "step": 1443 + }, + { + "epoch": 0.09, + "learning_rate": 1.982882038609489e-05, + "loss": 0.3859, + "step": 1444 + }, + { + "epoch": 0.09, + "learning_rate": 1.982845971313674e-05, + "loss": 0.4252, + "step": 1445 + }, + { + "epoch": 0.09, + "learning_rate": 1.9828098663899147e-05, + "loss": 0.3917, + "step": 1446 + }, + { + "epoch": 0.09, + "learning_rate": 1.9827737238395937e-05, + "loss": 0.4235, + "step": 1447 + }, + { + "epoch": 0.09, + "learning_rate": 1.982737543664095e-05, + "loss": 0.3915, + "step": 1448 + }, + { + "epoch": 0.09, + "learning_rate": 1.9827013258648037e-05, + "loss": 0.4082, + "step": 1449 + }, + { + "epoch": 0.09, + "learning_rate": 1.9826650704431062e-05, + "loss": 0.3742, + "step": 1450 + }, + { + "epoch": 0.09, + "learning_rate": 1.9826287774003906e-05, + "loss": 0.3747, + "step": 1451 + }, + { + "epoch": 0.09, + "learning_rate": 1.9825924467380466e-05, + "loss": 0.4015, + "step": 1452 + }, + { + "epoch": 0.09, + "learning_rate": 1.9825560784574648e-05, + "loss": 0.423, + "step": 1453 + }, + { + "epoch": 0.09, + "learning_rate": 1.9825196725600376e-05, + "loss": 0.3948, + "step": 1454 + }, + { + "epoch": 0.09, + "learning_rate": 1.9824832290471587e-05, + "loss": 0.3677, + "step": 1455 + }, + { + "epoch": 0.09, + "learning_rate": 1.9824467479202235e-05, + "loss": 0.3733, + "step": 1456 + }, + { + "epoch": 0.09, + "learning_rate": 1.982410229180629e-05, + "loss": 0.4157, + "step": 1457 + }, + { + "epoch": 0.09, + "learning_rate": 1.9823736728297725e-05, + "loss": 0.3655, + "step": 1458 + }, + { + "epoch": 0.09, + "learning_rate": 1.9823370788690543e-05, + "loss": 0.4074, + "step": 1459 + }, + { + "epoch": 0.09, + "learning_rate": 1.9823004472998747e-05, + "loss": 0.3719, + "step": 1460 + }, + { + "epoch": 0.09, + "learning_rate": 1.9822637781236367e-05, + "loss": 0.3594, + "step": 1461 + }, + { + "epoch": 0.09, + "learning_rate": 1.9822270713417436e-05, + "loss": 0.3836, + "step": 1462 + }, + { + "epoch": 0.09, + "learning_rate": 1.9821903269556017e-05, + "loss": 0.3717, + "step": 1463 + }, + { + "epoch": 0.09, + "learning_rate": 1.982153544966617e-05, + "loss": 0.3792, + "step": 1464 + }, + { + "epoch": 0.09, + "learning_rate": 1.9821167253761976e-05, + "loss": 0.3881, + "step": 1465 + }, + { + "epoch": 0.09, + "learning_rate": 1.9820798681857534e-05, + "loss": 0.3753, + "step": 1466 + }, + { + "epoch": 0.09, + "learning_rate": 1.9820429733966954e-05, + "loss": 0.4074, + "step": 1467 + }, + { + "epoch": 0.09, + "learning_rate": 1.9820060410104358e-05, + "loss": 0.3725, + "step": 1468 + }, + { + "epoch": 0.09, + "learning_rate": 1.981969071028389e-05, + "loss": 0.4116, + "step": 1469 + }, + { + "epoch": 0.09, + "learning_rate": 1.9819320634519706e-05, + "loss": 0.3825, + "step": 1470 + }, + { + "epoch": 0.09, + "learning_rate": 1.9818950182825967e-05, + "loss": 0.4085, + "step": 1471 + }, + { + "epoch": 0.09, + "learning_rate": 1.9818579355216863e-05, + "loss": 0.3959, + "step": 1472 + }, + { + "epoch": 0.09, + "learning_rate": 1.981820815170658e-05, + "loss": 0.3931, + "step": 1473 + }, + { + "epoch": 0.09, + "learning_rate": 1.9817836572309344e-05, + "loss": 0.4129, + "step": 1474 + }, + { + "epoch": 0.09, + "learning_rate": 1.981746461703937e-05, + "loss": 0.3708, + "step": 1475 + }, + { + "epoch": 0.09, + "learning_rate": 1.98170922859109e-05, + "loss": 0.3926, + "step": 1476 + }, + { + "epoch": 0.09, + "learning_rate": 1.9816719578938195e-05, + "loss": 0.3842, + "step": 1477 + }, + { + "epoch": 0.09, + "learning_rate": 1.9816346496135516e-05, + "loss": 0.4194, + "step": 1478 + }, + { + "epoch": 0.09, + "learning_rate": 1.981597303751715e-05, + "loss": 0.4012, + "step": 1479 + }, + { + "epoch": 0.09, + "learning_rate": 1.9815599203097394e-05, + "loss": 0.4182, + "step": 1480 + }, + { + "epoch": 0.09, + "learning_rate": 1.981522499289056e-05, + "loss": 0.4252, + "step": 1481 + }, + { + "epoch": 0.09, + "learning_rate": 1.9814850406910978e-05, + "loss": 0.4004, + "step": 1482 + }, + { + "epoch": 0.09, + "learning_rate": 1.981447544517298e-05, + "loss": 0.3943, + "step": 1483 + }, + { + "epoch": 0.09, + "learning_rate": 1.9814100107690933e-05, + "loss": 0.4132, + "step": 1484 + }, + { + "epoch": 0.09, + "learning_rate": 1.98137243944792e-05, + "loss": 0.3753, + "step": 1485 + }, + { + "epoch": 0.09, + "learning_rate": 1.9813348305552164e-05, + "loss": 0.3919, + "step": 1486 + }, + { + "epoch": 0.09, + "learning_rate": 1.9812971840924222e-05, + "loss": 0.3906, + "step": 1487 + }, + { + "epoch": 0.09, + "learning_rate": 1.9812595000609793e-05, + "loss": 0.3867, + "step": 1488 + }, + { + "epoch": 0.09, + "learning_rate": 1.9812217784623302e-05, + "loss": 0.4177, + "step": 1489 + }, + { + "epoch": 0.09, + "learning_rate": 1.981184019297919e-05, + "loss": 0.3792, + "step": 1490 + }, + { + "epoch": 0.09, + "learning_rate": 1.981146222569191e-05, + "loss": 0.365, + "step": 1491 + }, + { + "epoch": 0.09, + "learning_rate": 1.981108388277594e-05, + "loss": 0.3956, + "step": 1492 + }, + { + "epoch": 0.09, + "learning_rate": 1.9810705164245757e-05, + "loss": 0.385, + "step": 1493 + }, + { + "epoch": 0.09, + "learning_rate": 1.9810326070115863e-05, + "loss": 0.4057, + "step": 1494 + }, + { + "epoch": 0.09, + "learning_rate": 1.9809946600400772e-05, + "loss": 0.4263, + "step": 1495 + }, + { + "epoch": 0.09, + "learning_rate": 1.980956675511501e-05, + "loss": 0.4484, + "step": 1496 + }, + { + "epoch": 0.09, + "learning_rate": 1.9809186534273125e-05, + "loss": 0.4166, + "step": 1497 + }, + { + "epoch": 0.09, + "learning_rate": 1.9808805937889665e-05, + "loss": 0.3867, + "step": 1498 + }, + { + "epoch": 0.09, + "learning_rate": 1.9808424965979206e-05, + "loss": 0.3775, + "step": 1499 + }, + { + "epoch": 0.09, + "learning_rate": 1.9808043618556335e-05, + "loss": 0.4411, + "step": 1500 + }, + { + "epoch": 0.09, + "learning_rate": 1.980766189563565e-05, + "loss": 0.3898, + "step": 1501 + }, + { + "epoch": 0.09, + "learning_rate": 1.9807279797231762e-05, + "loss": 0.3887, + "step": 1502 + }, + { + "epoch": 0.09, + "learning_rate": 1.9806897323359304e-05, + "loss": 0.3636, + "step": 1503 + }, + { + "epoch": 0.09, + "learning_rate": 1.9806514474032915e-05, + "loss": 0.3624, + "step": 1504 + }, + { + "epoch": 0.09, + "learning_rate": 1.9806131249267256e-05, + "loss": 0.3909, + "step": 1505 + }, + { + "epoch": 0.09, + "learning_rate": 1.9805747649076995e-05, + "loss": 0.3758, + "step": 1506 + }, + { + "epoch": 0.09, + "learning_rate": 1.9805363673476822e-05, + "loss": 0.4049, + "step": 1507 + }, + { + "epoch": 0.09, + "learning_rate": 1.9804979322481435e-05, + "loss": 0.3719, + "step": 1508 + }, + { + "epoch": 0.09, + "learning_rate": 1.980459459610555e-05, + "loss": 0.3817, + "step": 1509 + }, + { + "epoch": 0.09, + "learning_rate": 1.9804209494363892e-05, + "loss": 0.4063, + "step": 1510 + }, + { + "epoch": 0.09, + "learning_rate": 1.9803824017271213e-05, + "loss": 0.4021, + "step": 1511 + }, + { + "epoch": 0.09, + "learning_rate": 1.9803438164842262e-05, + "loss": 0.392, + "step": 1512 + }, + { + "epoch": 0.09, + "learning_rate": 1.9803051937091814e-05, + "loss": 0.3984, + "step": 1513 + }, + { + "epoch": 0.09, + "learning_rate": 1.9802665334034656e-05, + "loss": 0.3993, + "step": 1514 + }, + { + "epoch": 0.09, + "learning_rate": 1.9802278355685592e-05, + "loss": 0.4057, + "step": 1515 + }, + { + "epoch": 0.09, + "learning_rate": 1.9801891002059435e-05, + "loss": 0.4163, + "step": 1516 + }, + { + "epoch": 0.09, + "learning_rate": 1.9801503273171014e-05, + "loss": 0.3862, + "step": 1517 + }, + { + "epoch": 0.09, + "learning_rate": 1.980111516903517e-05, + "loss": 0.3823, + "step": 1518 + }, + { + "epoch": 0.09, + "learning_rate": 1.9800726689666768e-05, + "loss": 0.397, + "step": 1519 + }, + { + "epoch": 0.09, + "learning_rate": 1.9800337835080677e-05, + "loss": 0.423, + "step": 1520 + }, + { + "epoch": 0.09, + "learning_rate": 1.979994860529179e-05, + "loss": 0.4143, + "step": 1521 + }, + { + "epoch": 0.09, + "learning_rate": 1.9799559000314995e-05, + "loss": 0.3859, + "step": 1522 + }, + { + "epoch": 0.09, + "learning_rate": 1.9799169020165222e-05, + "loss": 0.389, + "step": 1523 + }, + { + "epoch": 0.09, + "learning_rate": 1.9798778664857394e-05, + "loss": 0.3781, + "step": 1524 + }, + { + "epoch": 0.09, + "learning_rate": 1.9798387934406457e-05, + "loss": 0.4163, + "step": 1525 + }, + { + "epoch": 0.09, + "learning_rate": 1.979799682882737e-05, + "loss": 0.4252, + "step": 1526 + }, + { + "epoch": 0.09, + "learning_rate": 1.9797605348135107e-05, + "loss": 0.4149, + "step": 1527 + }, + { + "epoch": 0.09, + "learning_rate": 1.9797213492344656e-05, + "loss": 0.3945, + "step": 1528 + }, + { + "epoch": 0.09, + "learning_rate": 1.9796821261471018e-05, + "loss": 0.3786, + "step": 1529 + }, + { + "epoch": 0.09, + "learning_rate": 1.979642865552921e-05, + "loss": 0.4107, + "step": 1530 + }, + { + "epoch": 0.09, + "learning_rate": 1.979603567453426e-05, + "loss": 0.4104, + "step": 1531 + }, + { + "epoch": 0.09, + "learning_rate": 1.979564231850122e-05, + "loss": 0.38, + "step": 1532 + }, + { + "epoch": 0.09, + "learning_rate": 1.9795248587445143e-05, + "loss": 0.4009, + "step": 1533 + }, + { + "epoch": 0.09, + "learning_rate": 1.9794854481381106e-05, + "loss": 0.4043, + "step": 1534 + }, + { + "epoch": 0.09, + "learning_rate": 1.97944600003242e-05, + "loss": 0.3717, + "step": 1535 + }, + { + "epoch": 0.09, + "learning_rate": 1.979406514428952e-05, + "loss": 0.3711, + "step": 1536 + }, + { + "epoch": 0.09, + "learning_rate": 1.9793669913292186e-05, + "loss": 0.3772, + "step": 1537 + }, + { + "epoch": 0.09, + "learning_rate": 1.979327430734733e-05, + "loss": 0.368, + "step": 1538 + }, + { + "epoch": 0.09, + "learning_rate": 1.97928783264701e-05, + "loss": 0.3903, + "step": 1539 + }, + { + "epoch": 0.09, + "learning_rate": 1.9792481970675653e-05, + "loss": 0.3817, + "step": 1540 + }, + { + "epoch": 0.09, + "learning_rate": 1.9792085239979164e-05, + "loss": 0.3664, + "step": 1541 + }, + { + "epoch": 0.09, + "learning_rate": 1.9791688134395824e-05, + "loss": 0.4079, + "step": 1542 + }, + { + "epoch": 0.09, + "learning_rate": 1.979129065394083e-05, + "loss": 0.3758, + "step": 1543 + }, + { + "epoch": 0.09, + "learning_rate": 1.9790892798629407e-05, + "loss": 0.4096, + "step": 1544 + }, + { + "epoch": 0.09, + "learning_rate": 1.979049456847678e-05, + "loss": 0.4135, + "step": 1545 + }, + { + "epoch": 0.09, + "learning_rate": 1.97900959634982e-05, + "loss": 0.3873, + "step": 1546 + }, + { + "epoch": 0.09, + "learning_rate": 1.9789696983708927e-05, + "loss": 0.3976, + "step": 1547 + }, + { + "epoch": 0.09, + "learning_rate": 1.9789297629124237e-05, + "loss": 0.3948, + "step": 1548 + }, + { + "epoch": 0.09, + "learning_rate": 1.9788897899759408e-05, + "loss": 0.3979, + "step": 1549 + }, + { + "epoch": 0.09, + "learning_rate": 1.978849779562976e-05, + "loss": 0.3725, + "step": 1550 + }, + { + "epoch": 0.09, + "learning_rate": 1.97880973167506e-05, + "loss": 0.421, + "step": 1551 + }, + { + "epoch": 0.09, + "learning_rate": 1.9787696463137266e-05, + "loss": 0.3677, + "step": 1552 + }, + { + "epoch": 0.09, + "learning_rate": 1.9787295234805098e-05, + "loss": 0.3655, + "step": 1553 + }, + { + "epoch": 0.09, + "learning_rate": 1.978689363176946e-05, + "loss": 0.4018, + "step": 1554 + }, + { + "epoch": 0.09, + "learning_rate": 1.9786491654045733e-05, + "loss": 0.4297, + "step": 1555 + }, + { + "epoch": 0.09, + "learning_rate": 1.9786089301649297e-05, + "loss": 0.4492, + "step": 1556 + }, + { + "epoch": 0.09, + "learning_rate": 1.9785686574595565e-05, + "loss": 0.4191, + "step": 1557 + }, + { + "epoch": 0.09, + "learning_rate": 1.9785283472899947e-05, + "loss": 0.404, + "step": 1558 + }, + { + "epoch": 0.09, + "learning_rate": 1.978487999657788e-05, + "loss": 0.3979, + "step": 1559 + }, + { + "epoch": 0.09, + "learning_rate": 1.9784476145644813e-05, + "loss": 0.3901, + "step": 1560 + }, + { + "epoch": 0.09, + "learning_rate": 1.97840719201162e-05, + "loss": 0.3895, + "step": 1561 + }, + { + "epoch": 0.09, + "learning_rate": 1.9783667320007524e-05, + "loss": 0.4085, + "step": 1562 + }, + { + "epoch": 0.09, + "learning_rate": 1.9783262345334276e-05, + "loss": 0.382, + "step": 1563 + }, + { + "epoch": 0.09, + "learning_rate": 1.978285699611195e-05, + "loss": 0.3903, + "step": 1564 + }, + { + "epoch": 0.09, + "learning_rate": 1.9782451272356077e-05, + "loss": 0.3954, + "step": 1565 + }, + { + "epoch": 0.09, + "learning_rate": 1.9782045174082182e-05, + "loss": 0.3834, + "step": 1566 + }, + { + "epoch": 0.09, + "learning_rate": 1.9781638701305812e-05, + "loss": 0.3884, + "step": 1567 + }, + { + "epoch": 0.09, + "learning_rate": 1.9781231854042536e-05, + "loss": 0.4088, + "step": 1568 + }, + { + "epoch": 0.09, + "learning_rate": 1.978082463230792e-05, + "loss": 0.3624, + "step": 1569 + }, + { + "epoch": 0.09, + "learning_rate": 1.9780417036117564e-05, + "loss": 0.3599, + "step": 1570 + }, + { + "epoch": 0.09, + "learning_rate": 1.9780009065487066e-05, + "loss": 0.3591, + "step": 1571 + }, + { + "epoch": 0.09, + "learning_rate": 1.9779600720432048e-05, + "loss": 0.3742, + "step": 1572 + }, + { + "epoch": 0.1, + "learning_rate": 1.9779192000968145e-05, + "loss": 0.38, + "step": 1573 + }, + { + "epoch": 0.1, + "learning_rate": 1.9778782907110997e-05, + "loss": 0.4068, + "step": 1574 + }, + { + "epoch": 0.1, + "learning_rate": 1.9778373438876277e-05, + "loss": 0.3697, + "step": 1575 + }, + { + "epoch": 0.1, + "learning_rate": 1.9777963596279653e-05, + "loss": 0.3641, + "step": 1576 + }, + { + "epoch": 0.1, + "learning_rate": 1.977755337933682e-05, + "loss": 0.4512, + "step": 1577 + }, + { + "epoch": 0.1, + "learning_rate": 1.977714278806348e-05, + "loss": 0.4046, + "step": 1578 + }, + { + "epoch": 0.1, + "learning_rate": 1.9776731822475356e-05, + "loss": 0.409, + "step": 1579 + }, + { + "epoch": 0.1, + "learning_rate": 1.977632048258818e-05, + "loss": 0.3931, + "step": 1580 + }, + { + "epoch": 0.1, + "learning_rate": 1.97759087684177e-05, + "loss": 0.4029, + "step": 1581 + }, + { + "epoch": 0.1, + "learning_rate": 1.9775496679979673e-05, + "loss": 0.3979, + "step": 1582 + }, + { + "epoch": 0.1, + "learning_rate": 1.9775084217289883e-05, + "loss": 0.3926, + "step": 1583 + }, + { + "epoch": 0.1, + "learning_rate": 1.9774671380364122e-05, + "loss": 0.3527, + "step": 1584 + }, + { + "epoch": 0.1, + "learning_rate": 1.977425816921819e-05, + "loss": 0.4157, + "step": 1585 + }, + { + "epoch": 0.1, + "learning_rate": 1.977384458386791e-05, + "loss": 0.4035, + "step": 1586 + }, + { + "epoch": 0.1, + "learning_rate": 1.9773430624329108e-05, + "loss": 0.4169, + "step": 1587 + }, + { + "epoch": 0.1, + "learning_rate": 1.9773016290617644e-05, + "loss": 0.3823, + "step": 1588 + }, + { + "epoch": 0.1, + "learning_rate": 1.9772601582749375e-05, + "loss": 0.4353, + "step": 1589 + }, + { + "epoch": 0.1, + "learning_rate": 1.977218650074018e-05, + "loss": 0.3772, + "step": 1590 + }, + { + "epoch": 0.1, + "learning_rate": 1.9771771044605948e-05, + "loss": 0.3926, + "step": 1591 + }, + { + "epoch": 0.1, + "learning_rate": 1.9771355214362584e-05, + "loss": 0.3892, + "step": 1592 + }, + { + "epoch": 0.1, + "learning_rate": 1.977093901002601e-05, + "loss": 0.3817, + "step": 1593 + }, + { + "epoch": 0.1, + "learning_rate": 1.977052243161216e-05, + "loss": 0.399, + "step": 1594 + }, + { + "epoch": 0.1, + "learning_rate": 1.977010547913698e-05, + "loss": 0.3571, + "step": 1595 + }, + { + "epoch": 0.1, + "learning_rate": 1.976968815261644e-05, + "loss": 0.3856, + "step": 1596 + }, + { + "epoch": 0.1, + "learning_rate": 1.976927045206651e-05, + "loss": 0.3753, + "step": 1597 + }, + { + "epoch": 0.1, + "learning_rate": 1.976885237750318e-05, + "loss": 0.4138, + "step": 1598 + }, + { + "epoch": 0.1, + "learning_rate": 1.9768433928942462e-05, + "loss": 0.3736, + "step": 1599 + }, + { + "epoch": 0.1, + "learning_rate": 1.9768015106400375e-05, + "loss": 0.3778, + "step": 1600 + }, + { + "epoch": 0.1, + "learning_rate": 1.9767595909892953e-05, + "loss": 0.4124, + "step": 1601 + }, + { + "epoch": 0.1, + "learning_rate": 1.9767176339436243e-05, + "loss": 0.3809, + "step": 1602 + }, + { + "epoch": 0.1, + "learning_rate": 1.976675639504631e-05, + "loss": 0.3454, + "step": 1603 + }, + { + "epoch": 0.1, + "learning_rate": 1.976633607673923e-05, + "loss": 0.3797, + "step": 1604 + }, + { + "epoch": 0.1, + "learning_rate": 1.9765915384531092e-05, + "loss": 0.3705, + "step": 1605 + }, + { + "epoch": 0.1, + "learning_rate": 1.976549431843801e-05, + "loss": 0.4156, + "step": 1606 + }, + { + "epoch": 0.1, + "learning_rate": 1.97650728784761e-05, + "loss": 0.3954, + "step": 1607 + }, + { + "epoch": 0.1, + "learning_rate": 1.9764651064661496e-05, + "loss": 0.3538, + "step": 1608 + }, + { + "epoch": 0.1, + "learning_rate": 1.9764228877010344e-05, + "loss": 0.4043, + "step": 1609 + }, + { + "epoch": 0.1, + "learning_rate": 1.9763806315538816e-05, + "loss": 0.4049, + "step": 1610 + }, + { + "epoch": 0.1, + "learning_rate": 1.976338338026308e-05, + "loss": 0.3915, + "step": 1611 + }, + { + "epoch": 0.1, + "learning_rate": 1.9762960071199334e-05, + "loss": 0.3845, + "step": 1612 + }, + { + "epoch": 0.1, + "learning_rate": 1.9762536388363784e-05, + "loss": 0.404, + "step": 1613 + }, + { + "epoch": 0.1, + "learning_rate": 1.976211233177265e-05, + "loss": 0.3842, + "step": 1614 + }, + { + "epoch": 0.1, + "learning_rate": 1.9761687901442165e-05, + "loss": 0.3898, + "step": 1615 + }, + { + "epoch": 0.1, + "learning_rate": 1.9761263097388578e-05, + "loss": 0.3647, + "step": 1616 + }, + { + "epoch": 0.1, + "learning_rate": 1.9760837919628157e-05, + "loss": 0.3901, + "step": 1617 + }, + { + "epoch": 0.1, + "learning_rate": 1.976041236817717e-05, + "loss": 0.3895, + "step": 1618 + }, + { + "epoch": 0.1, + "learning_rate": 1.9759986443051922e-05, + "loss": 0.4314, + "step": 1619 + }, + { + "epoch": 0.1, + "learning_rate": 1.9759560144268712e-05, + "loss": 0.4152, + "step": 1620 + }, + { + "epoch": 0.1, + "learning_rate": 1.975913347184386e-05, + "loss": 0.3836, + "step": 1621 + }, + { + "epoch": 0.1, + "learning_rate": 1.9758706425793702e-05, + "loss": 0.4071, + "step": 1622 + }, + { + "epoch": 0.1, + "learning_rate": 1.975827900613459e-05, + "loss": 0.4146, + "step": 1623 + }, + { + "epoch": 0.1, + "learning_rate": 1.9757851212882885e-05, + "loss": 0.3783, + "step": 1624 + }, + { + "epoch": 0.1, + "learning_rate": 1.9757423046054968e-05, + "loss": 0.3887, + "step": 1625 + }, + { + "epoch": 0.1, + "learning_rate": 1.975699450566723e-05, + "loss": 0.3817, + "step": 1626 + }, + { + "epoch": 0.1, + "learning_rate": 1.9756565591736073e-05, + "loss": 0.3697, + "step": 1627 + }, + { + "epoch": 0.1, + "learning_rate": 1.975613630427792e-05, + "loss": 0.394, + "step": 1628 + }, + { + "epoch": 0.1, + "learning_rate": 1.975570664330921e-05, + "loss": 0.3901, + "step": 1629 + }, + { + "epoch": 0.1, + "learning_rate": 1.975527660884639e-05, + "loss": 0.4007, + "step": 1630 + }, + { + "epoch": 0.1, + "learning_rate": 1.9754846200905924e-05, + "loss": 0.3959, + "step": 1631 + }, + { + "epoch": 0.1, + "learning_rate": 1.9754415419504288e-05, + "loss": 0.4035, + "step": 1632 + }, + { + "epoch": 0.1, + "learning_rate": 1.9753984264657976e-05, + "loss": 0.3414, + "step": 1633 + }, + { + "epoch": 0.1, + "learning_rate": 1.9753552736383497e-05, + "loss": 0.4132, + "step": 1634 + }, + { + "epoch": 0.1, + "learning_rate": 1.975312083469737e-05, + "loss": 0.397, + "step": 1635 + }, + { + "epoch": 0.1, + "learning_rate": 1.9752688559616124e-05, + "loss": 0.4076, + "step": 1636 + }, + { + "epoch": 0.1, + "learning_rate": 1.9752255911156317e-05, + "loss": 0.3864, + "step": 1637 + }, + { + "epoch": 0.1, + "learning_rate": 1.9751822889334514e-05, + "loss": 0.3876, + "step": 1638 + }, + { + "epoch": 0.1, + "learning_rate": 1.9751389494167284e-05, + "loss": 0.4219, + "step": 1639 + }, + { + "epoch": 0.1, + "learning_rate": 1.9750955725671227e-05, + "loss": 0.3619, + "step": 1640 + }, + { + "epoch": 0.1, + "learning_rate": 1.975052158386295e-05, + "loss": 0.4004, + "step": 1641 + }, + { + "epoch": 0.1, + "learning_rate": 1.9750087068759065e-05, + "loss": 0.3627, + "step": 1642 + }, + { + "epoch": 0.1, + "learning_rate": 1.9749652180376215e-05, + "loss": 0.3744, + "step": 1643 + }, + { + "epoch": 0.1, + "learning_rate": 1.9749216918731048e-05, + "loss": 0.377, + "step": 1644 + }, + { + "epoch": 0.1, + "learning_rate": 1.974878128384023e-05, + "loss": 0.3711, + "step": 1645 + }, + { + "epoch": 0.1, + "learning_rate": 1.974834527572044e-05, + "loss": 0.38, + "step": 1646 + }, + { + "epoch": 0.1, + "learning_rate": 1.974790889438836e-05, + "loss": 0.3638, + "step": 1647 + }, + { + "epoch": 0.1, + "learning_rate": 1.9747472139860712e-05, + "loss": 0.3619, + "step": 1648 + }, + { + "epoch": 0.1, + "learning_rate": 1.9747035012154203e-05, + "loss": 0.385, + "step": 1649 + }, + { + "epoch": 0.1, + "learning_rate": 1.9746597511285577e-05, + "loss": 0.3859, + "step": 1650 + }, + { + "epoch": 0.1, + "learning_rate": 1.9746159637271582e-05, + "loss": 0.3853, + "step": 1651 + }, + { + "epoch": 0.1, + "learning_rate": 1.974572139012898e-05, + "loss": 0.4004, + "step": 1652 + }, + { + "epoch": 0.1, + "learning_rate": 1.974528276987455e-05, + "loss": 0.3723, + "step": 1653 + }, + { + "epoch": 0.1, + "learning_rate": 1.9744843776525083e-05, + "loss": 0.3756, + "step": 1654 + }, + { + "epoch": 0.1, + "learning_rate": 1.974440441009739e-05, + "loss": 0.4026, + "step": 1655 + }, + { + "epoch": 0.1, + "learning_rate": 1.9743964670608288e-05, + "loss": 0.3809, + "step": 1656 + }, + { + "epoch": 0.1, + "learning_rate": 1.9743524558074614e-05, + "loss": 0.3675, + "step": 1657 + }, + { + "epoch": 0.1, + "learning_rate": 1.9743084072513217e-05, + "loss": 0.368, + "step": 1658 + }, + { + "epoch": 0.1, + "learning_rate": 1.974264321394096e-05, + "loss": 0.4116, + "step": 1659 + }, + { + "epoch": 0.1, + "learning_rate": 1.9742201982374723e-05, + "loss": 0.4093, + "step": 1660 + }, + { + "epoch": 0.1, + "learning_rate": 1.9741760377831398e-05, + "loss": 0.3915, + "step": 1661 + }, + { + "epoch": 0.1, + "learning_rate": 1.974131840032789e-05, + "loss": 0.3758, + "step": 1662 + }, + { + "epoch": 0.1, + "learning_rate": 1.974087604988112e-05, + "loss": 0.3915, + "step": 1663 + }, + { + "epoch": 0.1, + "learning_rate": 1.9740433326508025e-05, + "loss": 0.3976, + "step": 1664 + }, + { + "epoch": 0.1, + "learning_rate": 1.9739990230225557e-05, + "loss": 0.3943, + "step": 1665 + }, + { + "epoch": 0.1, + "learning_rate": 1.9739546761050676e-05, + "loss": 0.3836, + "step": 1666 + }, + { + "epoch": 0.1, + "learning_rate": 1.973910291900036e-05, + "loss": 0.3853, + "step": 1667 + }, + { + "epoch": 0.1, + "learning_rate": 1.97386587040916e-05, + "loss": 0.3783, + "step": 1668 + }, + { + "epoch": 0.1, + "learning_rate": 1.9738214116341407e-05, + "loss": 0.3742, + "step": 1669 + }, + { + "epoch": 0.1, + "learning_rate": 1.97377691557668e-05, + "loss": 0.3767, + "step": 1670 + }, + { + "epoch": 0.1, + "learning_rate": 1.973732382238481e-05, + "loss": 0.352, + "step": 1671 + }, + { + "epoch": 0.1, + "learning_rate": 1.9736878116212494e-05, + "loss": 0.3945, + "step": 1672 + }, + { + "epoch": 0.1, + "learning_rate": 1.9736432037266912e-05, + "loss": 0.3795, + "step": 1673 + }, + { + "epoch": 0.1, + "learning_rate": 1.973598558556514e-05, + "loss": 0.3859, + "step": 1674 + }, + { + "epoch": 0.1, + "learning_rate": 1.9735538761124272e-05, + "loss": 0.4272, + "step": 1675 + }, + { + "epoch": 0.1, + "learning_rate": 1.973509156396142e-05, + "loss": 0.3834, + "step": 1676 + }, + { + "epoch": 0.1, + "learning_rate": 1.9734643994093693e-05, + "loss": 0.3903, + "step": 1677 + }, + { + "epoch": 0.1, + "learning_rate": 1.9734196051538235e-05, + "loss": 0.4054, + "step": 1678 + }, + { + "epoch": 0.1, + "learning_rate": 1.9733747736312194e-05, + "loss": 0.3636, + "step": 1679 + }, + { + "epoch": 0.1, + "learning_rate": 1.973329904843273e-05, + "loss": 0.3892, + "step": 1680 + }, + { + "epoch": 0.1, + "learning_rate": 1.973284998791703e-05, + "loss": 0.3711, + "step": 1681 + }, + { + "epoch": 0.1, + "learning_rate": 1.9732400554782275e-05, + "loss": 0.392, + "step": 1682 + }, + { + "epoch": 0.1, + "learning_rate": 1.9731950749045676e-05, + "loss": 0.3622, + "step": 1683 + }, + { + "epoch": 0.1, + "learning_rate": 1.9731500570724453e-05, + "loss": 0.4227, + "step": 1684 + }, + { + "epoch": 0.1, + "learning_rate": 1.9731050019835843e-05, + "loss": 0.3613, + "step": 1685 + }, + { + "epoch": 0.1, + "learning_rate": 1.973059909639709e-05, + "loss": 0.382, + "step": 1686 + }, + { + "epoch": 0.1, + "learning_rate": 1.973014780042547e-05, + "loss": 0.4049, + "step": 1687 + }, + { + "epoch": 0.1, + "learning_rate": 1.9729696131938244e-05, + "loss": 0.3973, + "step": 1688 + }, + { + "epoch": 0.1, + "learning_rate": 1.9729244090952715e-05, + "loss": 0.361, + "step": 1689 + }, + { + "epoch": 0.1, + "learning_rate": 1.9728791677486186e-05, + "loss": 0.404, + "step": 1690 + }, + { + "epoch": 0.1, + "learning_rate": 1.9728338891555977e-05, + "loss": 0.3859, + "step": 1691 + }, + { + "epoch": 0.1, + "learning_rate": 1.9727885733179427e-05, + "loss": 0.3982, + "step": 1692 + }, + { + "epoch": 0.1, + "learning_rate": 1.9727432202373875e-05, + "loss": 0.4199, + "step": 1693 + }, + { + "epoch": 0.1, + "learning_rate": 1.9726978299156695e-05, + "loss": 0.3725, + "step": 1694 + }, + { + "epoch": 0.1, + "learning_rate": 1.972652402354526e-05, + "loss": 0.3761, + "step": 1695 + }, + { + "epoch": 0.1, + "learning_rate": 1.9726069375556963e-05, + "loss": 0.3823, + "step": 1696 + }, + { + "epoch": 0.1, + "learning_rate": 1.9725614355209207e-05, + "loss": 0.4266, + "step": 1697 + }, + { + "epoch": 0.1, + "learning_rate": 1.972515896251942e-05, + "loss": 0.373, + "step": 1698 + }, + { + "epoch": 0.1, + "learning_rate": 1.9724703197505023e-05, + "loss": 0.4076, + "step": 1699 + }, + { + "epoch": 0.1, + "learning_rate": 1.972424706018348e-05, + "loss": 0.3876, + "step": 1700 + }, + { + "epoch": 0.1, + "learning_rate": 1.972379055057224e-05, + "loss": 0.4088, + "step": 1701 + }, + { + "epoch": 0.1, + "learning_rate": 1.972333366868879e-05, + "loss": 0.392, + "step": 1702 + }, + { + "epoch": 0.1, + "learning_rate": 1.972287641455062e-05, + "loss": 0.4171, + "step": 1703 + }, + { + "epoch": 0.1, + "learning_rate": 1.9722418788175235e-05, + "loss": 0.377, + "step": 1704 + }, + { + "epoch": 0.1, + "learning_rate": 1.972196078958015e-05, + "loss": 0.4188, + "step": 1705 + }, + { + "epoch": 0.1, + "learning_rate": 1.972150241878291e-05, + "loss": 0.4199, + "step": 1706 + }, + { + "epoch": 0.1, + "learning_rate": 1.972104367580106e-05, + "loss": 0.3803, + "step": 1707 + }, + { + "epoch": 0.1, + "learning_rate": 1.9720584560652157e-05, + "loss": 0.3878, + "step": 1708 + }, + { + "epoch": 0.1, + "learning_rate": 1.9720125073353778e-05, + "loss": 0.3873, + "step": 1709 + }, + { + "epoch": 0.1, + "learning_rate": 1.9719665213923522e-05, + "loss": 0.3915, + "step": 1710 + }, + { + "epoch": 0.1, + "learning_rate": 1.971920498237899e-05, + "loss": 0.3658, + "step": 1711 + }, + { + "epoch": 0.1, + "learning_rate": 1.97187443787378e-05, + "loss": 0.3783, + "step": 1712 + }, + { + "epoch": 0.1, + "learning_rate": 1.971828340301759e-05, + "loss": 0.3993, + "step": 1713 + }, + { + "epoch": 0.1, + "learning_rate": 1.9717822055236008e-05, + "loss": 0.373, + "step": 1714 + }, + { + "epoch": 0.1, + "learning_rate": 1.9717360335410715e-05, + "loss": 0.377, + "step": 1715 + }, + { + "epoch": 0.1, + "learning_rate": 1.9716898243559383e-05, + "loss": 0.3937, + "step": 1716 + }, + { + "epoch": 0.1, + "learning_rate": 1.9716435779699715e-05, + "loss": 0.4358, + "step": 1717 + }, + { + "epoch": 0.1, + "learning_rate": 1.9715972943849405e-05, + "loss": 0.399, + "step": 1718 + }, + { + "epoch": 0.1, + "learning_rate": 1.971550973602618e-05, + "loss": 0.3435, + "step": 1719 + }, + { + "epoch": 0.1, + "learning_rate": 1.9715046156247766e-05, + "loss": 0.385, + "step": 1720 + }, + { + "epoch": 0.1, + "learning_rate": 1.971458220453192e-05, + "loss": 0.3666, + "step": 1721 + }, + { + "epoch": 0.1, + "learning_rate": 1.9714117880896398e-05, + "loss": 0.3996, + "step": 1722 + }, + { + "epoch": 0.1, + "learning_rate": 1.971365318535898e-05, + "loss": 0.3831, + "step": 1723 + }, + { + "epoch": 0.1, + "learning_rate": 1.9713188117937452e-05, + "loss": 0.4004, + "step": 1724 + }, + { + "epoch": 0.1, + "learning_rate": 1.9712722678649623e-05, + "loss": 0.4233, + "step": 1725 + }, + { + "epoch": 0.1, + "learning_rate": 1.9712256867513313e-05, + "loss": 0.3825, + "step": 1726 + }, + { + "epoch": 0.1, + "learning_rate": 1.9711790684546352e-05, + "loss": 0.3945, + "step": 1727 + }, + { + "epoch": 0.1, + "learning_rate": 1.971132412976659e-05, + "loss": 0.4068, + "step": 1728 + }, + { + "epoch": 0.1, + "learning_rate": 1.9710857203191888e-05, + "loss": 0.3917, + "step": 1729 + }, + { + "epoch": 0.1, + "learning_rate": 1.9710389904840124e-05, + "loss": 0.3753, + "step": 1730 + }, + { + "epoch": 0.1, + "learning_rate": 1.9709922234729183e-05, + "loss": 0.3669, + "step": 1731 + }, + { + "epoch": 0.1, + "learning_rate": 1.9709454192876977e-05, + "loss": 0.4071, + "step": 1732 + }, + { + "epoch": 0.1, + "learning_rate": 1.9708985779301417e-05, + "loss": 0.3666, + "step": 1733 + }, + { + "epoch": 0.1, + "learning_rate": 1.9708516994020445e-05, + "loss": 0.3956, + "step": 1734 + }, + { + "epoch": 0.1, + "learning_rate": 1.9708047837052003e-05, + "loss": 0.4012, + "step": 1735 + }, + { + "epoch": 0.1, + "learning_rate": 1.970757830841405e-05, + "loss": 0.4235, + "step": 1736 + }, + { + "epoch": 0.1, + "learning_rate": 1.9707108408124565e-05, + "loss": 0.3845, + "step": 1737 + }, + { + "epoch": 0.1, + "learning_rate": 1.970663813620154e-05, + "loss": 0.3864, + "step": 1738 + }, + { + "epoch": 0.11, + "learning_rate": 1.9706167492662976e-05, + "loss": 0.4029, + "step": 1739 + }, + { + "epoch": 0.11, + "learning_rate": 1.970569647752689e-05, + "loss": 0.3733, + "step": 1740 + }, + { + "epoch": 0.11, + "learning_rate": 1.970522509081132e-05, + "loss": 0.4088, + "step": 1741 + }, + { + "epoch": 0.11, + "learning_rate": 1.9704753332534313e-05, + "loss": 0.4188, + "step": 1742 + }, + { + "epoch": 0.11, + "learning_rate": 1.9704281202713922e-05, + "loss": 0.3694, + "step": 1743 + }, + { + "epoch": 0.11, + "learning_rate": 1.970380870136823e-05, + "loss": 0.3842, + "step": 1744 + }, + { + "epoch": 0.11, + "learning_rate": 1.9703335828515324e-05, + "loss": 0.3488, + "step": 1745 + }, + { + "epoch": 0.11, + "learning_rate": 1.9702862584173304e-05, + "loss": 0.3722, + "step": 1746 + }, + { + "epoch": 0.11, + "learning_rate": 1.9702388968360293e-05, + "loss": 0.385, + "step": 1747 + }, + { + "epoch": 0.11, + "learning_rate": 1.9701914981094424e-05, + "loss": 0.3814, + "step": 1748 + }, + { + "epoch": 0.11, + "learning_rate": 1.970144062239384e-05, + "loss": 0.3864, + "step": 1749 + }, + { + "epoch": 0.11, + "learning_rate": 1.9700965892276708e-05, + "loss": 0.3951, + "step": 1750 + }, + { + "epoch": 0.11, + "learning_rate": 1.9700490790761192e-05, + "loss": 0.4035, + "step": 1751 + }, + { + "epoch": 0.11, + "learning_rate": 1.9700015317865487e-05, + "loss": 0.3764, + "step": 1752 + }, + { + "epoch": 0.11, + "learning_rate": 1.96995394736078e-05, + "loss": 0.3661, + "step": 1753 + }, + { + "epoch": 0.11, + "learning_rate": 1.9699063258006345e-05, + "loss": 0.382, + "step": 1754 + }, + { + "epoch": 0.11, + "learning_rate": 1.9698586671079352e-05, + "loss": 0.3956, + "step": 1755 + }, + { + "epoch": 0.11, + "learning_rate": 1.969810971284507e-05, + "loss": 0.4196, + "step": 1756 + }, + { + "epoch": 0.11, + "learning_rate": 1.9697632383321755e-05, + "loss": 0.3705, + "step": 1757 + }, + { + "epoch": 0.11, + "learning_rate": 1.969715468252769e-05, + "loss": 0.3753, + "step": 1758 + }, + { + "epoch": 0.11, + "learning_rate": 1.9696676610481153e-05, + "loss": 0.3825, + "step": 1759 + }, + { + "epoch": 0.11, + "learning_rate": 1.9696198167200452e-05, + "loss": 0.365, + "step": 1760 + }, + { + "epoch": 0.11, + "learning_rate": 1.9695719352703906e-05, + "loss": 0.3767, + "step": 1761 + }, + { + "epoch": 0.11, + "learning_rate": 1.9695240167009842e-05, + "loss": 0.3733, + "step": 1762 + }, + { + "epoch": 0.11, + "learning_rate": 1.969476061013661e-05, + "loss": 0.4157, + "step": 1763 + }, + { + "epoch": 0.11, + "learning_rate": 1.9694280682102565e-05, + "loss": 0.365, + "step": 1764 + }, + { + "epoch": 0.11, + "learning_rate": 1.9693800382926083e-05, + "loss": 0.4009, + "step": 1765 + }, + { + "epoch": 0.11, + "learning_rate": 1.9693319712625554e-05, + "loss": 0.3228, + "step": 1766 + }, + { + "epoch": 0.11, + "learning_rate": 1.969283867121938e-05, + "loss": 0.3806, + "step": 1767 + }, + { + "epoch": 0.11, + "learning_rate": 1.9692357258725977e-05, + "loss": 0.4227, + "step": 1768 + }, + { + "epoch": 0.11, + "learning_rate": 1.969187547516377e-05, + "loss": 0.4498, + "step": 1769 + }, + { + "epoch": 0.11, + "learning_rate": 1.969139332055121e-05, + "loss": 0.3585, + "step": 1770 + }, + { + "epoch": 0.11, + "learning_rate": 1.9690910794906755e-05, + "loss": 0.3708, + "step": 1771 + }, + { + "epoch": 0.11, + "learning_rate": 1.9690427898248877e-05, + "loss": 0.4333, + "step": 1772 + }, + { + "epoch": 0.11, + "learning_rate": 1.968994463059607e-05, + "loss": 0.4018, + "step": 1773 + }, + { + "epoch": 0.11, + "learning_rate": 1.9689460991966826e-05, + "loss": 0.3597, + "step": 1774 + }, + { + "epoch": 0.11, + "learning_rate": 1.968897698237967e-05, + "loss": 0.3552, + "step": 1775 + }, + { + "epoch": 0.11, + "learning_rate": 1.9688492601853122e-05, + "loss": 0.4096, + "step": 1776 + }, + { + "epoch": 0.11, + "learning_rate": 1.9688007850405735e-05, + "loss": 0.3867, + "step": 1777 + }, + { + "epoch": 0.11, + "learning_rate": 1.968752272805606e-05, + "loss": 0.382, + "step": 1778 + }, + { + "epoch": 0.11, + "learning_rate": 1.968703723482268e-05, + "loss": 0.3599, + "step": 1779 + }, + { + "epoch": 0.11, + "learning_rate": 1.9686551370724174e-05, + "loss": 0.3998, + "step": 1780 + }, + { + "epoch": 0.11, + "learning_rate": 1.9686065135779143e-05, + "loss": 0.3806, + "step": 1781 + }, + { + "epoch": 0.11, + "learning_rate": 1.968557853000621e-05, + "loss": 0.3781, + "step": 1782 + }, + { + "epoch": 0.11, + "learning_rate": 1.9685091553423997e-05, + "loss": 0.4289, + "step": 1783 + }, + { + "epoch": 0.11, + "learning_rate": 1.9684604206051148e-05, + "loss": 0.4222, + "step": 1784 + }, + { + "epoch": 0.11, + "learning_rate": 1.9684116487906326e-05, + "loss": 0.3864, + "step": 1785 + }, + { + "epoch": 0.11, + "learning_rate": 1.9683628399008198e-05, + "loss": 0.3652, + "step": 1786 + }, + { + "epoch": 0.11, + "learning_rate": 1.9683139939375454e-05, + "loss": 0.397, + "step": 1787 + }, + { + "epoch": 0.11, + "learning_rate": 1.968265110902679e-05, + "loss": 0.3605, + "step": 1788 + }, + { + "epoch": 0.11, + "learning_rate": 1.9682161907980924e-05, + "loss": 0.37, + "step": 1789 + }, + { + "epoch": 0.11, + "learning_rate": 1.9681672336256588e-05, + "loss": 0.3683, + "step": 1790 + }, + { + "epoch": 0.11, + "learning_rate": 1.968118239387252e-05, + "loss": 0.3906, + "step": 1791 + }, + { + "epoch": 0.11, + "learning_rate": 1.968069208084748e-05, + "loss": 0.3672, + "step": 1792 + }, + { + "epoch": 0.11, + "learning_rate": 1.968020139720024e-05, + "loss": 0.3931, + "step": 1793 + }, + { + "epoch": 0.11, + "learning_rate": 1.9679710342949578e-05, + "loss": 0.3892, + "step": 1794 + }, + { + "epoch": 0.11, + "learning_rate": 1.96792189181143e-05, + "loss": 0.3756, + "step": 1795 + }, + { + "epoch": 0.11, + "learning_rate": 1.9678727122713226e-05, + "loss": 0.3638, + "step": 1796 + }, + { + "epoch": 0.11, + "learning_rate": 1.9678234956765175e-05, + "loss": 0.3739, + "step": 1797 + }, + { + "epoch": 0.11, + "learning_rate": 1.967774242028899e-05, + "loss": 0.394, + "step": 1798 + }, + { + "epoch": 0.11, + "learning_rate": 1.9677249513303537e-05, + "loss": 0.4032, + "step": 1799 + }, + { + "epoch": 0.11, + "learning_rate": 1.9676756235827675e-05, + "loss": 0.3705, + "step": 1800 + }, + { + "epoch": 0.11, + "learning_rate": 1.9676262587880296e-05, + "loss": 0.387, + "step": 1801 + }, + { + "epoch": 0.11, + "learning_rate": 1.96757685694803e-05, + "loss": 0.3867, + "step": 1802 + }, + { + "epoch": 0.11, + "learning_rate": 1.967527418064659e-05, + "loss": 0.3552, + "step": 1803 + }, + { + "epoch": 0.11, + "learning_rate": 1.9674779421398104e-05, + "loss": 0.3853, + "step": 1804 + }, + { + "epoch": 0.11, + "learning_rate": 1.9674284291753785e-05, + "loss": 0.4043, + "step": 1805 + }, + { + "epoch": 0.11, + "learning_rate": 1.967378879173258e-05, + "loss": 0.4113, + "step": 1806 + }, + { + "epoch": 0.11, + "learning_rate": 1.9673292921353463e-05, + "loss": 0.4132, + "step": 1807 + }, + { + "epoch": 0.11, + "learning_rate": 1.9672796680635424e-05, + "loss": 0.4252, + "step": 1808 + }, + { + "epoch": 0.11, + "learning_rate": 1.9672300069597453e-05, + "loss": 0.3892, + "step": 1809 + }, + { + "epoch": 0.11, + "learning_rate": 1.967180308825856e-05, + "loss": 0.3689, + "step": 1810 + }, + { + "epoch": 0.11, + "learning_rate": 1.9671305736637788e-05, + "loss": 0.3661, + "step": 1811 + }, + { + "epoch": 0.11, + "learning_rate": 1.9670808014754162e-05, + "loss": 0.4113, + "step": 1812 + }, + { + "epoch": 0.11, + "learning_rate": 1.9670309922626743e-05, + "loss": 0.4143, + "step": 1813 + }, + { + "epoch": 0.11, + "learning_rate": 1.9669811460274603e-05, + "loss": 0.3828, + "step": 1814 + }, + { + "epoch": 0.11, + "learning_rate": 1.966931262771682e-05, + "loss": 0.3862, + "step": 1815 + }, + { + "epoch": 0.11, + "learning_rate": 1.9668813424972492e-05, + "loss": 0.373, + "step": 1816 + }, + { + "epoch": 0.11, + "learning_rate": 1.9668313852060737e-05, + "loss": 0.3689, + "step": 1817 + }, + { + "epoch": 0.11, + "learning_rate": 1.9667813909000675e-05, + "loss": 0.3594, + "step": 1818 + }, + { + "epoch": 0.11, + "learning_rate": 1.9667313595811448e-05, + "loss": 0.3535, + "step": 1819 + }, + { + "epoch": 0.11, + "learning_rate": 1.9666812912512208e-05, + "loss": 0.3831, + "step": 1820 + }, + { + "epoch": 0.11, + "learning_rate": 1.9666311859122132e-05, + "loss": 0.3736, + "step": 1821 + }, + { + "epoch": 0.11, + "learning_rate": 1.9665810435660397e-05, + "loss": 0.4308, + "step": 1822 + }, + { + "epoch": 0.11, + "learning_rate": 1.9665308642146192e-05, + "loss": 0.3608, + "step": 1823 + }, + { + "epoch": 0.11, + "learning_rate": 1.966480647859874e-05, + "loss": 0.3817, + "step": 1824 + }, + { + "epoch": 0.11, + "learning_rate": 1.966430394503726e-05, + "loss": 0.4336, + "step": 1825 + }, + { + "epoch": 0.11, + "learning_rate": 1.9663801041480995e-05, + "loss": 0.3859, + "step": 1826 + }, + { + "epoch": 0.11, + "learning_rate": 1.9663297767949198e-05, + "loss": 0.4249, + "step": 1827 + }, + { + "epoch": 0.11, + "learning_rate": 1.9662794124461132e-05, + "loss": 0.3929, + "step": 1828 + }, + { + "epoch": 0.11, + "learning_rate": 1.966229011103608e-05, + "loss": 0.3747, + "step": 1829 + }, + { + "epoch": 0.11, + "learning_rate": 1.966178572769334e-05, + "loss": 0.4007, + "step": 1830 + }, + { + "epoch": 0.11, + "learning_rate": 1.9661280974452225e-05, + "loss": 0.3859, + "step": 1831 + }, + { + "epoch": 0.11, + "learning_rate": 1.9660775851332058e-05, + "loss": 0.3744, + "step": 1832 + }, + { + "epoch": 0.11, + "learning_rate": 1.9660270358352172e-05, + "loss": 0.411, + "step": 1833 + }, + { + "epoch": 0.11, + "learning_rate": 1.9659764495531926e-05, + "loss": 0.4155, + "step": 1834 + }, + { + "epoch": 0.11, + "learning_rate": 1.9659258262890683e-05, + "loss": 0.3761, + "step": 1835 + }, + { + "epoch": 0.11, + "learning_rate": 1.9658751660447826e-05, + "loss": 0.3633, + "step": 1836 + }, + { + "epoch": 0.11, + "learning_rate": 1.9658244688222748e-05, + "loss": 0.3864, + "step": 1837 + }, + { + "epoch": 0.11, + "learning_rate": 1.9657737346234866e-05, + "loss": 0.3675, + "step": 1838 + }, + { + "epoch": 0.11, + "learning_rate": 1.965722963450359e-05, + "loss": 0.4004, + "step": 1839 + }, + { + "epoch": 0.11, + "learning_rate": 1.965672155304837e-05, + "loss": 0.3689, + "step": 1840 + }, + { + "epoch": 0.11, + "learning_rate": 1.9656213101888647e-05, + "loss": 0.3811, + "step": 1841 + }, + { + "epoch": 0.11, + "learning_rate": 1.9655704281043895e-05, + "loss": 0.3493, + "step": 1842 + }, + { + "epoch": 0.11, + "learning_rate": 1.965519509053359e-05, + "loss": 0.3479, + "step": 1843 + }, + { + "epoch": 0.11, + "learning_rate": 1.9654685530377228e-05, + "loss": 0.4155, + "step": 1844 + }, + { + "epoch": 0.11, + "learning_rate": 1.965417560059432e-05, + "loss": 0.3797, + "step": 1845 + }, + { + "epoch": 0.11, + "learning_rate": 1.965366530120438e-05, + "loss": 0.3825, + "step": 1846 + }, + { + "epoch": 0.11, + "learning_rate": 1.965315463222695e-05, + "loss": 0.3823, + "step": 1847 + }, + { + "epoch": 0.11, + "learning_rate": 1.9652643593681588e-05, + "loss": 0.3783, + "step": 1848 + }, + { + "epoch": 0.11, + "learning_rate": 1.9652132185587845e-05, + "loss": 0.3753, + "step": 1849 + }, + { + "epoch": 0.11, + "learning_rate": 1.965162040796531e-05, + "loss": 0.3873, + "step": 1850 + }, + { + "epoch": 0.11, + "learning_rate": 1.965110826083357e-05, + "loss": 0.3915, + "step": 1851 + }, + { + "epoch": 0.11, + "learning_rate": 1.965059574421224e-05, + "loss": 0.4143, + "step": 1852 + }, + { + "epoch": 0.11, + "learning_rate": 1.9650082858120935e-05, + "loss": 0.3912, + "step": 1853 + }, + { + "epoch": 0.11, + "learning_rate": 1.9649569602579294e-05, + "loss": 0.4182, + "step": 1854 + }, + { + "epoch": 0.11, + "learning_rate": 1.964905597760696e-05, + "loss": 0.3806, + "step": 1855 + }, + { + "epoch": 0.11, + "learning_rate": 1.964854198322361e-05, + "loss": 0.356, + "step": 1856 + }, + { + "epoch": 0.11, + "learning_rate": 1.964802761944891e-05, + "loss": 0.3764, + "step": 1857 + }, + { + "epoch": 0.11, + "learning_rate": 1.9647512886302558e-05, + "loss": 0.4018, + "step": 1858 + }, + { + "epoch": 0.11, + "learning_rate": 1.964699778380426e-05, + "loss": 0.404, + "step": 1859 + }, + { + "epoch": 0.11, + "learning_rate": 1.9646482311973734e-05, + "loss": 0.3934, + "step": 1860 + }, + { + "epoch": 0.11, + "learning_rate": 1.964596647083072e-05, + "loss": 0.3644, + "step": 1861 + }, + { + "epoch": 0.11, + "learning_rate": 1.9645450260394957e-05, + "loss": 0.3672, + "step": 1862 + }, + { + "epoch": 0.11, + "learning_rate": 1.964493368068622e-05, + "loss": 0.3703, + "step": 1863 + }, + { + "epoch": 0.11, + "learning_rate": 1.964441673172428e-05, + "loss": 0.3996, + "step": 1864 + }, + { + "epoch": 0.11, + "learning_rate": 1.9643899413528926e-05, + "loss": 0.3887, + "step": 1865 + }, + { + "epoch": 0.11, + "learning_rate": 1.9643381726119967e-05, + "loss": 0.4113, + "step": 1866 + }, + { + "epoch": 0.11, + "learning_rate": 1.964286366951722e-05, + "loss": 0.3984, + "step": 1867 + }, + { + "epoch": 0.11, + "learning_rate": 1.9642345243740523e-05, + "loss": 0.3862, + "step": 1868 + }, + { + "epoch": 0.11, + "learning_rate": 1.964182644880972e-05, + "loss": 0.3789, + "step": 1869 + }, + { + "epoch": 0.11, + "learning_rate": 1.9641307284744674e-05, + "loss": 0.3733, + "step": 1870 + }, + { + "epoch": 0.11, + "learning_rate": 1.964078775156526e-05, + "loss": 0.3683, + "step": 1871 + }, + { + "epoch": 0.11, + "learning_rate": 1.964026784929137e-05, + "loss": 0.4074, + "step": 1872 + }, + { + "epoch": 0.11, + "learning_rate": 1.9639747577942906e-05, + "loss": 0.4216, + "step": 1873 + }, + { + "epoch": 0.11, + "learning_rate": 1.9639226937539788e-05, + "loss": 0.3691, + "step": 1874 + }, + { + "epoch": 0.11, + "learning_rate": 1.9638705928101945e-05, + "loss": 0.3954, + "step": 1875 + }, + { + "epoch": 0.11, + "learning_rate": 1.963818454964933e-05, + "loss": 0.3722, + "step": 1876 + }, + { + "epoch": 0.11, + "learning_rate": 1.96376628022019e-05, + "loss": 0.3814, + "step": 1877 + }, + { + "epoch": 0.11, + "learning_rate": 1.963714068577963e-05, + "loss": 0.392, + "step": 1878 + }, + { + "epoch": 0.11, + "learning_rate": 1.963661820040251e-05, + "loss": 0.4255, + "step": 1879 + }, + { + "epoch": 0.11, + "learning_rate": 1.9636095346090543e-05, + "loss": 0.3806, + "step": 1880 + }, + { + "epoch": 0.11, + "learning_rate": 1.9635572122863747e-05, + "loss": 0.4249, + "step": 1881 + }, + { + "epoch": 0.11, + "learning_rate": 1.9635048530742153e-05, + "loss": 0.3686, + "step": 1882 + }, + { + "epoch": 0.11, + "learning_rate": 1.9634524569745804e-05, + "loss": 0.3546, + "step": 1883 + }, + { + "epoch": 0.11, + "learning_rate": 1.9634000239894763e-05, + "loss": 0.3848, + "step": 1884 + }, + { + "epoch": 0.11, + "learning_rate": 1.9633475541209102e-05, + "loss": 0.3834, + "step": 1885 + }, + { + "epoch": 0.11, + "learning_rate": 1.963295047370891e-05, + "loss": 0.3817, + "step": 1886 + }, + { + "epoch": 0.11, + "learning_rate": 1.963242503741429e-05, + "loss": 0.3675, + "step": 1887 + }, + { + "epoch": 0.11, + "learning_rate": 1.9631899232345354e-05, + "loss": 0.3677, + "step": 1888 + }, + { + "epoch": 0.11, + "learning_rate": 1.9631373058522237e-05, + "loss": 0.3859, + "step": 1889 + }, + { + "epoch": 0.11, + "learning_rate": 1.9630846515965083e-05, + "loss": 0.3739, + "step": 1890 + }, + { + "epoch": 0.11, + "learning_rate": 1.9630319604694046e-05, + "loss": 0.3979, + "step": 1891 + }, + { + "epoch": 0.11, + "learning_rate": 1.9629792324729302e-05, + "loss": 0.3968, + "step": 1892 + }, + { + "epoch": 0.11, + "learning_rate": 1.9629264676091036e-05, + "loss": 0.3931, + "step": 1893 + }, + { + "epoch": 0.11, + "learning_rate": 1.9628736658799454e-05, + "loss": 0.3477, + "step": 1894 + }, + { + "epoch": 0.11, + "learning_rate": 1.9628208272874766e-05, + "loss": 0.399, + "step": 1895 + }, + { + "epoch": 0.11, + "learning_rate": 1.96276795183372e-05, + "loss": 0.4063, + "step": 1896 + }, + { + "epoch": 0.11, + "learning_rate": 1.9627150395207003e-05, + "loss": 0.3817, + "step": 1897 + }, + { + "epoch": 0.11, + "learning_rate": 1.962662090350443e-05, + "loss": 0.3817, + "step": 1898 + }, + { + "epoch": 0.11, + "learning_rate": 1.9626091043249757e-05, + "loss": 0.3878, + "step": 1899 + }, + { + "epoch": 0.11, + "learning_rate": 1.962556081446326e-05, + "loss": 0.387, + "step": 1900 + }, + { + "epoch": 0.11, + "learning_rate": 1.9625030217165246e-05, + "loss": 0.3549, + "step": 1901 + }, + { + "epoch": 0.11, + "learning_rate": 1.9624499251376027e-05, + "loss": 0.3973, + "step": 1902 + }, + { + "epoch": 0.11, + "learning_rate": 1.9623967917115935e-05, + "loss": 0.3778, + "step": 1903 + }, + { + "epoch": 0.12, + "learning_rate": 1.96234362144053e-05, + "loss": 0.4018, + "step": 1904 + }, + { + "epoch": 0.12, + "learning_rate": 1.9622904143264492e-05, + "loss": 0.4026, + "step": 1905 + }, + { + "epoch": 0.12, + "learning_rate": 1.9622371703713872e-05, + "loss": 0.3613, + "step": 1906 + }, + { + "epoch": 0.12, + "learning_rate": 1.962183889577383e-05, + "loss": 0.4381, + "step": 1907 + }, + { + "epoch": 0.12, + "learning_rate": 1.9621305719464762e-05, + "loss": 0.3775, + "step": 1908 + }, + { + "epoch": 0.12, + "learning_rate": 1.962077217480708e-05, + "loss": 0.3803, + "step": 1909 + }, + { + "epoch": 0.12, + "learning_rate": 1.962023826182121e-05, + "loss": 0.3873, + "step": 1910 + }, + { + "epoch": 0.12, + "learning_rate": 1.9619703980527592e-05, + "loss": 0.4124, + "step": 1911 + }, + { + "epoch": 0.12, + "learning_rate": 1.9619169330946684e-05, + "loss": 0.3912, + "step": 1912 + }, + { + "epoch": 0.12, + "learning_rate": 1.961863431309895e-05, + "loss": 0.3574, + "step": 1913 + }, + { + "epoch": 0.12, + "learning_rate": 1.961809892700488e-05, + "loss": 0.3477, + "step": 1914 + }, + { + "epoch": 0.12, + "learning_rate": 1.9617563172684968e-05, + "loss": 0.3644, + "step": 1915 + }, + { + "epoch": 0.12, + "learning_rate": 1.961702705015972e-05, + "loss": 0.3976, + "step": 1916 + }, + { + "epoch": 0.12, + "learning_rate": 1.9616490559449673e-05, + "loss": 0.3931, + "step": 1917 + }, + { + "epoch": 0.12, + "learning_rate": 1.9615953700575355e-05, + "loss": 0.3864, + "step": 1918 + }, + { + "epoch": 0.12, + "learning_rate": 1.9615416473557322e-05, + "loss": 0.3588, + "step": 1919 + }, + { + "epoch": 0.12, + "learning_rate": 1.9614878878416145e-05, + "loss": 0.397, + "step": 1920 + }, + { + "epoch": 0.12, + "learning_rate": 1.9614340915172403e-05, + "loss": 0.3834, + "step": 1921 + }, + { + "epoch": 0.12, + "learning_rate": 1.9613802583846696e-05, + "loss": 0.3722, + "step": 1922 + }, + { + "epoch": 0.12, + "learning_rate": 1.9613263884459626e-05, + "loss": 0.3516, + "step": 1923 + }, + { + "epoch": 0.12, + "learning_rate": 1.961272481703182e-05, + "loss": 0.3691, + "step": 1924 + }, + { + "epoch": 0.12, + "learning_rate": 1.9612185381583925e-05, + "loss": 0.3853, + "step": 1925 + }, + { + "epoch": 0.12, + "learning_rate": 1.961164557813658e-05, + "loss": 0.3697, + "step": 1926 + }, + { + "epoch": 0.12, + "learning_rate": 1.9611105406710455e-05, + "loss": 0.3443, + "step": 1927 + }, + { + "epoch": 0.12, + "learning_rate": 1.9610564867326237e-05, + "loss": 0.3912, + "step": 1928 + }, + { + "epoch": 0.12, + "learning_rate": 1.961002396000461e-05, + "loss": 0.3577, + "step": 1929 + }, + { + "epoch": 0.12, + "learning_rate": 1.9609482684766293e-05, + "loss": 0.3862, + "step": 1930 + }, + { + "epoch": 0.12, + "learning_rate": 1.9608941041632e-05, + "loss": 0.3836, + "step": 1931 + }, + { + "epoch": 0.12, + "learning_rate": 1.9608399030622475e-05, + "loss": 0.3934, + "step": 1932 + }, + { + "epoch": 0.12, + "learning_rate": 1.960785665175846e-05, + "loss": 0.3725, + "step": 1933 + }, + { + "epoch": 0.12, + "learning_rate": 1.9607313905060725e-05, + "loss": 0.3965, + "step": 1934 + }, + { + "epoch": 0.12, + "learning_rate": 1.960677079055005e-05, + "loss": 0.3675, + "step": 1935 + }, + { + "epoch": 0.12, + "learning_rate": 1.960622730824723e-05, + "loss": 0.3426, + "step": 1936 + }, + { + "epoch": 0.12, + "learning_rate": 1.960568345817306e-05, + "loss": 0.3867, + "step": 1937 + }, + { + "epoch": 0.12, + "learning_rate": 1.9605139240348377e-05, + "loss": 0.3666, + "step": 1938 + }, + { + "epoch": 0.12, + "learning_rate": 1.9604594654794004e-05, + "loss": 0.4012, + "step": 1939 + }, + { + "epoch": 0.12, + "learning_rate": 1.9604049701530798e-05, + "loss": 0.387, + "step": 1940 + }, + { + "epoch": 0.12, + "learning_rate": 1.960350438057962e-05, + "loss": 0.3703, + "step": 1941 + }, + { + "epoch": 0.12, + "learning_rate": 1.9602958691961344e-05, + "loss": 0.3421, + "step": 1942 + }, + { + "epoch": 0.12, + "learning_rate": 1.9602412635696865e-05, + "loss": 0.3733, + "step": 1943 + }, + { + "epoch": 0.12, + "learning_rate": 1.960186621180709e-05, + "loss": 0.3915, + "step": 1944 + }, + { + "epoch": 0.12, + "learning_rate": 1.9601319420312934e-05, + "loss": 0.3817, + "step": 1945 + }, + { + "epoch": 0.12, + "learning_rate": 1.9600772261235336e-05, + "loss": 0.397, + "step": 1946 + }, + { + "epoch": 0.12, + "learning_rate": 1.9600224734595237e-05, + "loss": 0.3594, + "step": 1947 + }, + { + "epoch": 0.12, + "learning_rate": 1.9599676840413607e-05, + "loss": 0.3778, + "step": 1948 + }, + { + "epoch": 0.12, + "learning_rate": 1.9599128578711416e-05, + "loss": 0.3666, + "step": 1949 + }, + { + "epoch": 0.12, + "learning_rate": 1.9598579949509658e-05, + "loss": 0.3797, + "step": 1950 + }, + { + "epoch": 0.12, + "learning_rate": 1.959803095282933e-05, + "loss": 0.3903, + "step": 1951 + }, + { + "epoch": 0.12, + "learning_rate": 1.9597481588691464e-05, + "loss": 0.3828, + "step": 1952 + }, + { + "epoch": 0.12, + "learning_rate": 1.9596931857117078e-05, + "loss": 0.365, + "step": 1953 + }, + { + "epoch": 0.12, + "learning_rate": 1.9596381758127223e-05, + "loss": 0.3563, + "step": 1954 + }, + { + "epoch": 0.12, + "learning_rate": 1.9595831291742963e-05, + "loss": 0.3945, + "step": 1955 + }, + { + "epoch": 0.12, + "learning_rate": 1.959528045798537e-05, + "loss": 0.404, + "step": 1956 + }, + { + "epoch": 0.12, + "learning_rate": 1.9594729256875528e-05, + "loss": 0.356, + "step": 1957 + }, + { + "epoch": 0.12, + "learning_rate": 1.9594177688434546e-05, + "loss": 0.3803, + "step": 1958 + }, + { + "epoch": 0.12, + "learning_rate": 1.9593625752683536e-05, + "loss": 0.3998, + "step": 1959 + }, + { + "epoch": 0.12, + "learning_rate": 1.9593073449643635e-05, + "loss": 0.3594, + "step": 1960 + }, + { + "epoch": 0.12, + "learning_rate": 1.959252077933598e-05, + "loss": 0.4071, + "step": 1961 + }, + { + "epoch": 0.12, + "learning_rate": 1.9591967741781736e-05, + "loss": 0.3616, + "step": 1962 + }, + { + "epoch": 0.12, + "learning_rate": 1.9591414337002072e-05, + "loss": 0.397, + "step": 1963 + }, + { + "epoch": 0.12, + "learning_rate": 1.9590860565018176e-05, + "loss": 0.4138, + "step": 1964 + }, + { + "epoch": 0.12, + "learning_rate": 1.9590306425851253e-05, + "loss": 0.3675, + "step": 1965 + }, + { + "epoch": 0.12, + "learning_rate": 1.9589751919522514e-05, + "loss": 0.4012, + "step": 1966 + }, + { + "epoch": 0.12, + "learning_rate": 1.9589197046053182e-05, + "loss": 0.3901, + "step": 1967 + }, + { + "epoch": 0.12, + "learning_rate": 1.9588641805464514e-05, + "loss": 0.404, + "step": 1968 + }, + { + "epoch": 0.12, + "learning_rate": 1.9588086197777753e-05, + "loss": 0.3853, + "step": 1969 + }, + { + "epoch": 0.12, + "learning_rate": 1.958753022301418e-05, + "loss": 0.3943, + "step": 1970 + }, + { + "epoch": 0.12, + "learning_rate": 1.958697388119508e-05, + "loss": 0.4093, + "step": 1971 + }, + { + "epoch": 0.12, + "learning_rate": 1.9586417172341748e-05, + "loss": 0.3672, + "step": 1972 + }, + { + "epoch": 0.12, + "learning_rate": 1.95858600964755e-05, + "loss": 0.3926, + "step": 1973 + }, + { + "epoch": 0.12, + "learning_rate": 1.958530265361766e-05, + "loss": 0.4129, + "step": 1974 + }, + { + "epoch": 0.12, + "learning_rate": 1.9584744843789573e-05, + "loss": 0.3758, + "step": 1975 + }, + { + "epoch": 0.12, + "learning_rate": 1.9584186667012594e-05, + "loss": 0.3795, + "step": 1976 + }, + { + "epoch": 0.12, + "learning_rate": 1.9583628123308092e-05, + "loss": 0.3585, + "step": 1977 + }, + { + "epoch": 0.12, + "learning_rate": 1.958306921269745e-05, + "loss": 0.3753, + "step": 1978 + }, + { + "epoch": 0.12, + "learning_rate": 1.9582509935202073e-05, + "loss": 0.3686, + "step": 1979 + }, + { + "epoch": 0.12, + "learning_rate": 1.9581950290843358e-05, + "loss": 0.3772, + "step": 1980 + }, + { + "epoch": 0.12, + "learning_rate": 1.9581390279642745e-05, + "loss": 0.3775, + "step": 1981 + }, + { + "epoch": 0.12, + "learning_rate": 1.9580829901621666e-05, + "loss": 0.3633, + "step": 1982 + }, + { + "epoch": 0.12, + "learning_rate": 1.9580269156801575e-05, + "loss": 0.3903, + "step": 1983 + }, + { + "epoch": 0.12, + "learning_rate": 1.9579708045203943e-05, + "loss": 0.3585, + "step": 1984 + }, + { + "epoch": 0.12, + "learning_rate": 1.9579146566850253e-05, + "loss": 0.3705, + "step": 1985 + }, + { + "epoch": 0.12, + "learning_rate": 1.9578584721762e-05, + "loss": 0.3968, + "step": 1986 + }, + { + "epoch": 0.12, + "learning_rate": 1.957802250996069e-05, + "loss": 0.387, + "step": 1987 + }, + { + "epoch": 0.12, + "learning_rate": 1.957745993146785e-05, + "loss": 0.3783, + "step": 1988 + }, + { + "epoch": 0.12, + "learning_rate": 1.9576896986305023e-05, + "loss": 0.3831, + "step": 1989 + }, + { + "epoch": 0.12, + "learning_rate": 1.957633367449375e-05, + "loss": 0.3655, + "step": 1990 + }, + { + "epoch": 0.12, + "learning_rate": 1.9575769996055606e-05, + "loss": 0.3856, + "step": 1991 + }, + { + "epoch": 0.12, + "learning_rate": 1.957520595101217e-05, + "loss": 0.387, + "step": 1992 + }, + { + "epoch": 0.12, + "learning_rate": 1.9574641539385034e-05, + "loss": 0.385, + "step": 1993 + }, + { + "epoch": 0.12, + "learning_rate": 1.9574076761195808e-05, + "loss": 0.3823, + "step": 1994 + }, + { + "epoch": 0.12, + "learning_rate": 1.9573511616466115e-05, + "loss": 0.3842, + "step": 1995 + }, + { + "epoch": 0.12, + "learning_rate": 1.957294610521759e-05, + "loss": 0.4166, + "step": 1996 + }, + { + "epoch": 0.12, + "learning_rate": 1.957238022747188e-05, + "loss": 0.3937, + "step": 1997 + }, + { + "epoch": 0.12, + "learning_rate": 1.9571813983250658e-05, + "loss": 0.4166, + "step": 1998 + }, + { + "epoch": 0.12, + "learning_rate": 1.9571247372575596e-05, + "loss": 0.3708, + "step": 1999 + }, + { + "epoch": 0.12, + "learning_rate": 1.9570680395468388e-05, + "loss": 0.4171, + "step": 2000 + }, + { + "epoch": 0.12, + "learning_rate": 1.957011305195074e-05, + "loss": 0.3823, + "step": 2001 + }, + { + "epoch": 0.12, + "learning_rate": 1.9569545342044375e-05, + "loss": 0.3873, + "step": 2002 + }, + { + "epoch": 0.12, + "learning_rate": 1.9568977265771023e-05, + "loss": 0.3884, + "step": 2003 + }, + { + "epoch": 0.12, + "learning_rate": 1.9568408823152435e-05, + "loss": 0.3407, + "step": 2004 + }, + { + "epoch": 0.12, + "learning_rate": 1.9567840014210376e-05, + "loss": 0.3775, + "step": 2005 + }, + { + "epoch": 0.12, + "learning_rate": 1.956727083896662e-05, + "loss": 0.3619, + "step": 2006 + }, + { + "epoch": 0.12, + "learning_rate": 1.9566701297442958e-05, + "loss": 0.4127, + "step": 2007 + }, + { + "epoch": 0.12, + "learning_rate": 1.9566131389661197e-05, + "loss": 0.3546, + "step": 2008 + }, + { + "epoch": 0.12, + "learning_rate": 1.9565561115643153e-05, + "loss": 0.3429, + "step": 2009 + }, + { + "epoch": 0.12, + "learning_rate": 1.9564990475410658e-05, + "loss": 0.3719, + "step": 2010 + }, + { + "epoch": 0.12, + "learning_rate": 1.9564419468985562e-05, + "loss": 0.3739, + "step": 2011 + }, + { + "epoch": 0.12, + "learning_rate": 1.9563848096389722e-05, + "loss": 0.3962, + "step": 2012 + }, + { + "epoch": 0.12, + "learning_rate": 1.956327635764502e-05, + "loss": 0.3758, + "step": 2013 + }, + { + "epoch": 0.12, + "learning_rate": 1.9562704252773333e-05, + "loss": 0.3845, + "step": 2014 + }, + { + "epoch": 0.12, + "learning_rate": 1.956213178179658e-05, + "loss": 0.3912, + "step": 2015 + }, + { + "epoch": 0.12, + "learning_rate": 1.9561558944736657e-05, + "loss": 0.368, + "step": 2016 + }, + { + "epoch": 0.12, + "learning_rate": 1.9560985741615516e-05, + "loss": 0.3284, + "step": 2017 + }, + { + "epoch": 0.12, + "learning_rate": 1.9560412172455088e-05, + "loss": 0.3781, + "step": 2018 + }, + { + "epoch": 0.12, + "learning_rate": 1.9559838237277336e-05, + "loss": 0.4129, + "step": 2019 + }, + { + "epoch": 0.12, + "learning_rate": 1.9559263936104235e-05, + "loss": 0.3725, + "step": 2020 + }, + { + "epoch": 0.12, + "learning_rate": 1.9558689268957768e-05, + "loss": 0.3256, + "step": 2021 + }, + { + "epoch": 0.12, + "learning_rate": 1.9558114235859943e-05, + "loss": 0.3655, + "step": 2022 + }, + { + "epoch": 0.12, + "learning_rate": 1.9557538836832764e-05, + "loss": 0.3594, + "step": 2023 + }, + { + "epoch": 0.12, + "learning_rate": 1.955696307189827e-05, + "loss": 0.392, + "step": 2024 + }, + { + "epoch": 0.12, + "learning_rate": 1.9556386941078497e-05, + "loss": 0.3711, + "step": 2025 + }, + { + "epoch": 0.12, + "learning_rate": 1.9555810444395508e-05, + "loss": 0.3943, + "step": 2026 + }, + { + "epoch": 0.12, + "learning_rate": 1.9555233581871366e-05, + "loss": 0.3956, + "step": 2027 + }, + { + "epoch": 0.12, + "learning_rate": 1.9554656353528165e-05, + "loss": 0.3647, + "step": 2028 + }, + { + "epoch": 0.12, + "learning_rate": 1.9554078759388e-05, + "loss": 0.3624, + "step": 2029 + }, + { + "epoch": 0.12, + "learning_rate": 1.9553500799472984e-05, + "loss": 0.3744, + "step": 2030 + }, + { + "epoch": 0.12, + "learning_rate": 1.9552922473805244e-05, + "loss": 0.3783, + "step": 2031 + }, + { + "epoch": 0.12, + "learning_rate": 1.955234378240692e-05, + "loss": 0.4152, + "step": 2032 + }, + { + "epoch": 0.12, + "learning_rate": 1.9551764725300167e-05, + "loss": 0.3968, + "step": 2033 + }, + { + "epoch": 0.12, + "learning_rate": 1.9551185302507157e-05, + "loss": 0.3689, + "step": 2034 + }, + { + "epoch": 0.12, + "learning_rate": 1.9550605514050068e-05, + "loss": 0.3878, + "step": 2035 + }, + { + "epoch": 0.12, + "learning_rate": 1.9550025359951102e-05, + "loss": 0.3797, + "step": 2036 + }, + { + "epoch": 0.12, + "learning_rate": 1.954944484023247e-05, + "loss": 0.4018, + "step": 2037 + }, + { + "epoch": 0.12, + "learning_rate": 1.9548863954916394e-05, + "loss": 0.377, + "step": 2038 + }, + { + "epoch": 0.12, + "learning_rate": 1.9548282704025113e-05, + "loss": 0.3987, + "step": 2039 + }, + { + "epoch": 0.12, + "learning_rate": 1.9547701087580878e-05, + "loss": 0.3566, + "step": 2040 + }, + { + "epoch": 0.12, + "learning_rate": 1.9547119105605964e-05, + "loss": 0.4009, + "step": 2041 + }, + { + "epoch": 0.12, + "learning_rate": 1.9546536758122645e-05, + "loss": 0.3976, + "step": 2042 + }, + { + "epoch": 0.12, + "learning_rate": 1.954595404515322e-05, + "loss": 0.4224, + "step": 2043 + }, + { + "epoch": 0.12, + "learning_rate": 1.9545370966719992e-05, + "loss": 0.3527, + "step": 2044 + }, + { + "epoch": 0.12, + "learning_rate": 1.954478752284529e-05, + "loss": 0.3438, + "step": 2045 + }, + { + "epoch": 0.12, + "learning_rate": 1.9544203713551454e-05, + "loss": 0.4076, + "step": 2046 + }, + { + "epoch": 0.12, + "learning_rate": 1.954361953886082e-05, + "loss": 0.3471, + "step": 2047 + }, + { + "epoch": 0.12, + "learning_rate": 1.954303499879577e-05, + "loss": 0.3672, + "step": 2048 + }, + { + "epoch": 0.12, + "learning_rate": 1.9542450093378675e-05, + "loss": 0.3541, + "step": 2049 + }, + { + "epoch": 0.12, + "learning_rate": 1.954186482263193e-05, + "loss": 0.3686, + "step": 2050 + }, + { + "epoch": 0.12, + "learning_rate": 1.9541279186577937e-05, + "loss": 0.3934, + "step": 2051 + }, + { + "epoch": 0.12, + "learning_rate": 1.9540693185239124e-05, + "loss": 0.3834, + "step": 2052 + }, + { + "epoch": 0.12, + "learning_rate": 1.954010681863792e-05, + "loss": 0.3658, + "step": 2053 + }, + { + "epoch": 0.12, + "learning_rate": 1.9539520086796773e-05, + "loss": 0.3962, + "step": 2054 + }, + { + "epoch": 0.12, + "learning_rate": 1.9538932989738156e-05, + "loss": 0.3906, + "step": 2055 + }, + { + "epoch": 0.12, + "learning_rate": 1.9538345527484535e-05, + "loss": 0.3624, + "step": 2056 + }, + { + "epoch": 0.12, + "learning_rate": 1.9537757700058403e-05, + "loss": 0.3996, + "step": 2057 + }, + { + "epoch": 0.12, + "learning_rate": 1.953716950748227e-05, + "loss": 0.316, + "step": 2058 + }, + { + "epoch": 0.12, + "learning_rate": 1.953658094977865e-05, + "loss": 0.4155, + "step": 2059 + }, + { + "epoch": 0.12, + "learning_rate": 1.953599202697008e-05, + "loss": 0.3468, + "step": 2060 + }, + { + "epoch": 0.12, + "learning_rate": 1.9535402739079098e-05, + "loss": 0.3839, + "step": 2061 + }, + { + "epoch": 0.12, + "learning_rate": 1.9534813086128275e-05, + "loss": 0.3809, + "step": 2062 + }, + { + "epoch": 0.12, + "learning_rate": 1.953422306814018e-05, + "loss": 0.3811, + "step": 2063 + }, + { + "epoch": 0.12, + "learning_rate": 1.9533632685137402e-05, + "loss": 0.3898, + "step": 2064 + }, + { + "epoch": 0.12, + "learning_rate": 1.953304193714254e-05, + "loss": 0.4046, + "step": 2065 + }, + { + "epoch": 0.12, + "learning_rate": 1.9532450824178223e-05, + "loss": 0.3809, + "step": 2066 + }, + { + "epoch": 0.12, + "learning_rate": 1.953185934626707e-05, + "loss": 0.3976, + "step": 2067 + }, + { + "epoch": 0.12, + "learning_rate": 1.953126750343173e-05, + "loss": 0.4043, + "step": 2068 + }, + { + "epoch": 0.12, + "learning_rate": 1.9530675295694856e-05, + "loss": 0.3619, + "step": 2069 + }, + { + "epoch": 0.13, + "learning_rate": 1.953008272307913e-05, + "loss": 0.3535, + "step": 2070 + }, + { + "epoch": 0.13, + "learning_rate": 1.9529489785607232e-05, + "loss": 0.3597, + "step": 2071 + }, + { + "epoch": 0.13, + "learning_rate": 1.9528896483301866e-05, + "loss": 0.382, + "step": 2072 + }, + { + "epoch": 0.13, + "learning_rate": 1.952830281618574e-05, + "loss": 0.3689, + "step": 2073 + }, + { + "epoch": 0.13, + "learning_rate": 1.952770878428159e-05, + "loss": 0.3736, + "step": 2074 + }, + { + "epoch": 0.13, + "learning_rate": 1.9527114387612153e-05, + "loss": 0.3764, + "step": 2075 + }, + { + "epoch": 0.13, + "learning_rate": 1.9526519626200188e-05, + "loss": 0.3792, + "step": 2076 + }, + { + "epoch": 0.13, + "learning_rate": 1.9525924500068467e-05, + "loss": 0.3862, + "step": 2077 + }, + { + "epoch": 0.13, + "learning_rate": 1.952532900923977e-05, + "loss": 0.3814, + "step": 2078 + }, + { + "epoch": 0.13, + "learning_rate": 1.9524733153736897e-05, + "loss": 0.3767, + "step": 2079 + }, + { + "epoch": 0.13, + "learning_rate": 1.952413693358266e-05, + "loss": 0.3834, + "step": 2080 + }, + { + "epoch": 0.13, + "learning_rate": 1.9523540348799887e-05, + "loss": 0.375, + "step": 2081 + }, + { + "epoch": 0.13, + "learning_rate": 1.9522943399411414e-05, + "loss": 0.3666, + "step": 2082 + }, + { + "epoch": 0.13, + "learning_rate": 1.9522346085440095e-05, + "loss": 0.3491, + "step": 2083 + }, + { + "epoch": 0.13, + "learning_rate": 1.9521748406908805e-05, + "loss": 0.3792, + "step": 2084 + }, + { + "epoch": 0.13, + "learning_rate": 1.9521150363840416e-05, + "loss": 0.3488, + "step": 2085 + }, + { + "epoch": 0.13, + "learning_rate": 1.9520551956257832e-05, + "loss": 0.3839, + "step": 2086 + }, + { + "epoch": 0.13, + "learning_rate": 1.951995318418396e-05, + "loss": 0.3502, + "step": 2087 + }, + { + "epoch": 0.13, + "learning_rate": 1.9519354047641726e-05, + "loss": 0.3619, + "step": 2088 + }, + { + "epoch": 0.13, + "learning_rate": 1.9518754546654062e-05, + "loss": 0.3652, + "step": 2089 + }, + { + "epoch": 0.13, + "learning_rate": 1.9518154681243927e-05, + "loss": 0.3708, + "step": 2090 + }, + { + "epoch": 0.13, + "learning_rate": 1.9517554451434278e-05, + "loss": 0.3884, + "step": 2091 + }, + { + "epoch": 0.13, + "learning_rate": 1.95169538572481e-05, + "loss": 0.356, + "step": 2092 + }, + { + "epoch": 0.13, + "learning_rate": 1.9516352898708388e-05, + "loss": 0.3664, + "step": 2093 + }, + { + "epoch": 0.13, + "learning_rate": 1.9515751575838147e-05, + "loss": 0.37, + "step": 2094 + }, + { + "epoch": 0.13, + "learning_rate": 1.9515149888660398e-05, + "loss": 0.3839, + "step": 2095 + }, + { + "epoch": 0.13, + "learning_rate": 1.951454783719818e-05, + "loss": 0.3756, + "step": 2096 + }, + { + "epoch": 0.13, + "learning_rate": 1.9513945421474537e-05, + "loss": 0.3758, + "step": 2097 + }, + { + "epoch": 0.13, + "learning_rate": 1.9513342641512535e-05, + "loss": 0.4021, + "step": 2098 + }, + { + "epoch": 0.13, + "learning_rate": 1.9512739497335253e-05, + "loss": 0.3758, + "step": 2099 + }, + { + "epoch": 0.13, + "learning_rate": 1.951213598896578e-05, + "loss": 0.3638, + "step": 2100 + }, + { + "epoch": 0.13, + "learning_rate": 1.951153211642722e-05, + "loss": 0.37, + "step": 2101 + }, + { + "epoch": 0.13, + "learning_rate": 1.9510927879742694e-05, + "loss": 0.4049, + "step": 2102 + }, + { + "epoch": 0.13, + "learning_rate": 1.9510323278935334e-05, + "loss": 0.3672, + "step": 2103 + }, + { + "epoch": 0.13, + "learning_rate": 1.950971831402829e-05, + "loss": 0.3664, + "step": 2104 + }, + { + "epoch": 0.13, + "learning_rate": 1.9509112985044717e-05, + "loss": 0.3915, + "step": 2105 + }, + { + "epoch": 0.13, + "learning_rate": 1.9508507292007796e-05, + "loss": 0.394, + "step": 2106 + }, + { + "epoch": 0.13, + "learning_rate": 1.950790123494071e-05, + "loss": 0.3767, + "step": 2107 + }, + { + "epoch": 0.13, + "learning_rate": 1.950729481386667e-05, + "loss": 0.392, + "step": 2108 + }, + { + "epoch": 0.13, + "learning_rate": 1.9506688028808886e-05, + "loss": 0.3926, + "step": 2109 + }, + { + "epoch": 0.13, + "learning_rate": 1.9506080879790586e-05, + "loss": 0.3669, + "step": 2110 + }, + { + "epoch": 0.13, + "learning_rate": 1.950547336683502e-05, + "loss": 0.377, + "step": 2111 + }, + { + "epoch": 0.13, + "learning_rate": 1.9504865489965447e-05, + "loss": 0.3694, + "step": 2112 + }, + { + "epoch": 0.13, + "learning_rate": 1.9504257249205138e-05, + "loss": 0.3747, + "step": 2113 + }, + { + "epoch": 0.13, + "learning_rate": 1.9503648644577374e-05, + "loss": 0.3661, + "step": 2114 + }, + { + "epoch": 0.13, + "learning_rate": 1.950303967610546e-05, + "loss": 0.3725, + "step": 2115 + }, + { + "epoch": 0.13, + "learning_rate": 1.9502430343812713e-05, + "loss": 0.3714, + "step": 2116 + }, + { + "epoch": 0.13, + "learning_rate": 1.9501820647722458e-05, + "loss": 0.4004, + "step": 2117 + }, + { + "epoch": 0.13, + "learning_rate": 1.9501210587858036e-05, + "loss": 0.3767, + "step": 2118 + }, + { + "epoch": 0.13, + "learning_rate": 1.9500600164242803e-05, + "loss": 0.3471, + "step": 2119 + }, + { + "epoch": 0.13, + "learning_rate": 1.9499989376900134e-05, + "loss": 0.3965, + "step": 2120 + }, + { + "epoch": 0.13, + "learning_rate": 1.9499378225853402e-05, + "loss": 0.3962, + "step": 2121 + }, + { + "epoch": 0.13, + "learning_rate": 1.9498766711126017e-05, + "loss": 0.3881, + "step": 2122 + }, + { + "epoch": 0.13, + "learning_rate": 1.9498154832741384e-05, + "loss": 0.3524, + "step": 2123 + }, + { + "epoch": 0.13, + "learning_rate": 1.9497542590722928e-05, + "loss": 0.373, + "step": 2124 + }, + { + "epoch": 0.13, + "learning_rate": 1.949692998509409e-05, + "loss": 0.3931, + "step": 2125 + }, + { + "epoch": 0.13, + "learning_rate": 1.9496317015878322e-05, + "loss": 0.4224, + "step": 2126 + }, + { + "epoch": 0.13, + "learning_rate": 1.9495703683099097e-05, + "loss": 0.3845, + "step": 2127 + }, + { + "epoch": 0.13, + "learning_rate": 1.9495089986779888e-05, + "loss": 0.3901, + "step": 2128 + }, + { + "epoch": 0.13, + "learning_rate": 1.9494475926944197e-05, + "loss": 0.3951, + "step": 2129 + }, + { + "epoch": 0.13, + "learning_rate": 1.949386150361553e-05, + "loss": 0.3982, + "step": 2130 + }, + { + "epoch": 0.13, + "learning_rate": 1.949324671681741e-05, + "loss": 0.3728, + "step": 2131 + }, + { + "epoch": 0.13, + "learning_rate": 1.9492631566573374e-05, + "loss": 0.3518, + "step": 2132 + }, + { + "epoch": 0.13, + "learning_rate": 1.9492016052906975e-05, + "loss": 0.3959, + "step": 2133 + }, + { + "epoch": 0.13, + "learning_rate": 1.949140017584177e-05, + "loss": 0.3689, + "step": 2134 + }, + { + "epoch": 0.13, + "learning_rate": 1.949078393540135e-05, + "loss": 0.4143, + "step": 2135 + }, + { + "epoch": 0.13, + "learning_rate": 1.94901673316093e-05, + "loss": 0.3892, + "step": 2136 + }, + { + "epoch": 0.13, + "learning_rate": 1.9489550364489228e-05, + "loss": 0.3803, + "step": 2137 + }, + { + "epoch": 0.13, + "learning_rate": 1.948893303406475e-05, + "loss": 0.3873, + "step": 2138 + }, + { + "epoch": 0.13, + "learning_rate": 1.9488315340359507e-05, + "loss": 0.3993, + "step": 2139 + }, + { + "epoch": 0.13, + "learning_rate": 1.9487697283397144e-05, + "loss": 0.3675, + "step": 2140 + }, + { + "epoch": 0.13, + "learning_rate": 1.9487078863201324e-05, + "loss": 0.3906, + "step": 2141 + }, + { + "epoch": 0.13, + "learning_rate": 1.9486460079795722e-05, + "loss": 0.3781, + "step": 2142 + }, + { + "epoch": 0.13, + "learning_rate": 1.9485840933204027e-05, + "loss": 0.3884, + "step": 2143 + }, + { + "epoch": 0.13, + "learning_rate": 1.9485221423449946e-05, + "loss": 0.3412, + "step": 2144 + }, + { + "epoch": 0.13, + "learning_rate": 1.9484601550557195e-05, + "loss": 0.3789, + "step": 2145 + }, + { + "epoch": 0.13, + "learning_rate": 1.9483981314549507e-05, + "loss": 0.37, + "step": 2146 + }, + { + "epoch": 0.13, + "learning_rate": 1.9483360715450622e-05, + "loss": 0.3345, + "step": 2147 + }, + { + "epoch": 0.13, + "learning_rate": 1.948273975328431e-05, + "loss": 0.373, + "step": 2148 + }, + { + "epoch": 0.13, + "learning_rate": 1.948211842807433e-05, + "loss": 0.3669, + "step": 2149 + }, + { + "epoch": 0.13, + "learning_rate": 1.9481496739844484e-05, + "loss": 0.3557, + "step": 2150 + }, + { + "epoch": 0.13, + "learning_rate": 1.9480874688618563e-05, + "loss": 0.3795, + "step": 2151 + }, + { + "epoch": 0.13, + "learning_rate": 1.948025227442039e-05, + "loss": 0.3711, + "step": 2152 + }, + { + "epoch": 0.13, + "learning_rate": 1.9479629497273783e-05, + "loss": 0.361, + "step": 2153 + }, + { + "epoch": 0.13, + "learning_rate": 1.9479006357202594e-05, + "loss": 0.3594, + "step": 2154 + }, + { + "epoch": 0.13, + "learning_rate": 1.9478382854230678e-05, + "loss": 0.3438, + "step": 2155 + }, + { + "epoch": 0.13, + "learning_rate": 1.9477758988381903e-05, + "loss": 0.4057, + "step": 2156 + }, + { + "epoch": 0.13, + "learning_rate": 1.9477134759680154e-05, + "loss": 0.3968, + "step": 2157 + }, + { + "epoch": 0.13, + "learning_rate": 1.947651016814933e-05, + "loss": 0.3744, + "step": 2158 + }, + { + "epoch": 0.13, + "learning_rate": 1.947588521381335e-05, + "loss": 0.4277, + "step": 2159 + }, + { + "epoch": 0.13, + "learning_rate": 1.947525989669613e-05, + "loss": 0.3968, + "step": 2160 + }, + { + "epoch": 0.13, + "learning_rate": 1.9474634216821608e-05, + "loss": 0.3987, + "step": 2161 + }, + { + "epoch": 0.13, + "learning_rate": 1.947400817421375e-05, + "loss": 0.3717, + "step": 2162 + }, + { + "epoch": 0.13, + "learning_rate": 1.9473381768896516e-05, + "loss": 0.3878, + "step": 2163 + }, + { + "epoch": 0.13, + "learning_rate": 1.9472755000893892e-05, + "loss": 0.4102, + "step": 2164 + }, + { + "epoch": 0.13, + "learning_rate": 1.947212787022987e-05, + "loss": 0.3661, + "step": 2165 + }, + { + "epoch": 0.13, + "learning_rate": 1.9471500376928457e-05, + "loss": 0.4029, + "step": 2166 + }, + { + "epoch": 0.13, + "learning_rate": 1.9470872521013683e-05, + "loss": 0.3811, + "step": 2167 + }, + { + "epoch": 0.13, + "learning_rate": 1.9470244302509583e-05, + "loss": 0.382, + "step": 2168 + }, + { + "epoch": 0.13, + "learning_rate": 1.9469615721440206e-05, + "loss": 0.3786, + "step": 2169 + }, + { + "epoch": 0.13, + "learning_rate": 1.9468986777829618e-05, + "loss": 0.3828, + "step": 2170 + }, + { + "epoch": 0.13, + "learning_rate": 1.94683574717019e-05, + "loss": 0.382, + "step": 2171 + }, + { + "epoch": 0.13, + "learning_rate": 1.946772780308114e-05, + "loss": 0.3747, + "step": 2172 + }, + { + "epoch": 0.13, + "learning_rate": 1.946709777199145e-05, + "loss": 0.3563, + "step": 2173 + }, + { + "epoch": 0.13, + "learning_rate": 1.9466467378456947e-05, + "loss": 0.3451, + "step": 2174 + }, + { + "epoch": 0.13, + "learning_rate": 1.946583662250177e-05, + "loss": 0.3703, + "step": 2175 + }, + { + "epoch": 0.13, + "learning_rate": 1.946520550415006e-05, + "loss": 0.3825, + "step": 2176 + }, + { + "epoch": 0.13, + "learning_rate": 1.9464574023425982e-05, + "loss": 0.3906, + "step": 2177 + }, + { + "epoch": 0.13, + "learning_rate": 1.9463942180353718e-05, + "loss": 0.3764, + "step": 2178 + }, + { + "epoch": 0.13, + "learning_rate": 1.946330997495745e-05, + "loss": 0.3923, + "step": 2179 + }, + { + "epoch": 0.13, + "learning_rate": 1.9462677407261386e-05, + "loss": 0.3831, + "step": 2180 + }, + { + "epoch": 0.13, + "learning_rate": 1.946204447728974e-05, + "loss": 0.3736, + "step": 2181 + }, + { + "epoch": 0.13, + "learning_rate": 1.946141118506675e-05, + "loss": 0.3588, + "step": 2182 + }, + { + "epoch": 0.13, + "learning_rate": 1.9460777530616653e-05, + "loss": 0.37, + "step": 2183 + }, + { + "epoch": 0.13, + "learning_rate": 1.9460143513963717e-05, + "loss": 0.404, + "step": 2184 + }, + { + "epoch": 0.13, + "learning_rate": 1.9459509135132208e-05, + "loss": 0.3546, + "step": 2185 + }, + { + "epoch": 0.13, + "learning_rate": 1.9458874394146416e-05, + "loss": 0.346, + "step": 2186 + }, + { + "epoch": 0.13, + "learning_rate": 1.9458239291030644e-05, + "loss": 0.3426, + "step": 2187 + }, + { + "epoch": 0.13, + "learning_rate": 1.9457603825809205e-05, + "loss": 0.3703, + "step": 2188 + }, + { + "epoch": 0.13, + "learning_rate": 1.9456967998506423e-05, + "loss": 0.3703, + "step": 2189 + }, + { + "epoch": 0.13, + "learning_rate": 1.9456331809146648e-05, + "loss": 0.3959, + "step": 2190 + }, + { + "epoch": 0.13, + "learning_rate": 1.9455695257754227e-05, + "loss": 0.4054, + "step": 2191 + }, + { + "epoch": 0.13, + "learning_rate": 1.945505834435354e-05, + "loss": 0.3652, + "step": 2192 + }, + { + "epoch": 0.13, + "learning_rate": 1.945442106896897e-05, + "loss": 0.3624, + "step": 2193 + }, + { + "epoch": 0.13, + "learning_rate": 1.9453783431624903e-05, + "loss": 0.3901, + "step": 2194 + }, + { + "epoch": 0.13, + "learning_rate": 1.9453145432345767e-05, + "loss": 0.3761, + "step": 2195 + }, + { + "epoch": 0.13, + "learning_rate": 1.9452507071155978e-05, + "loss": 0.3845, + "step": 2196 + }, + { + "epoch": 0.13, + "learning_rate": 1.9451868348079975e-05, + "loss": 0.4049, + "step": 2197 + }, + { + "epoch": 0.13, + "learning_rate": 1.945122926314222e-05, + "loss": 0.3456, + "step": 2198 + }, + { + "epoch": 0.13, + "learning_rate": 1.945058981636717e-05, + "loss": 0.3424, + "step": 2199 + }, + { + "epoch": 0.13, + "learning_rate": 1.944995000777931e-05, + "loss": 0.4096, + "step": 2200 + }, + { + "epoch": 0.13, + "learning_rate": 1.9449309837403136e-05, + "loss": 0.3717, + "step": 2201 + }, + { + "epoch": 0.13, + "learning_rate": 1.9448669305263153e-05, + "loss": 0.3956, + "step": 2202 + }, + { + "epoch": 0.13, + "learning_rate": 1.944802841138389e-05, + "loss": 0.3836, + "step": 2203 + }, + { + "epoch": 0.13, + "learning_rate": 1.9447387155789878e-05, + "loss": 0.385, + "step": 2204 + }, + { + "epoch": 0.13, + "learning_rate": 1.944674553850567e-05, + "loss": 0.3479, + "step": 2205 + }, + { + "epoch": 0.13, + "learning_rate": 1.9446103559555825e-05, + "loss": 0.3867, + "step": 2206 + }, + { + "epoch": 0.13, + "learning_rate": 1.944546121896493e-05, + "loss": 0.389, + "step": 2207 + }, + { + "epoch": 0.13, + "learning_rate": 1.944481851675757e-05, + "loss": 0.3597, + "step": 2208 + }, + { + "epoch": 0.13, + "learning_rate": 1.944417545295835e-05, + "loss": 0.3518, + "step": 2209 + }, + { + "epoch": 0.13, + "learning_rate": 1.944353202759189e-05, + "loss": 0.3767, + "step": 2210 + }, + { + "epoch": 0.13, + "learning_rate": 1.9442888240682833e-05, + "loss": 0.411, + "step": 2211 + }, + { + "epoch": 0.13, + "learning_rate": 1.9442244092255812e-05, + "loss": 0.363, + "step": 2212 + }, + { + "epoch": 0.13, + "learning_rate": 1.94415995823355e-05, + "loss": 0.3892, + "step": 2213 + }, + { + "epoch": 0.13, + "learning_rate": 1.9440954710946564e-05, + "loss": 0.3602, + "step": 2214 + }, + { + "epoch": 0.13, + "learning_rate": 1.9440309478113692e-05, + "loss": 0.3468, + "step": 2215 + }, + { + "epoch": 0.13, + "learning_rate": 1.943966388386159e-05, + "loss": 0.3753, + "step": 2216 + }, + { + "epoch": 0.13, + "learning_rate": 1.9439017928214978e-05, + "loss": 0.3549, + "step": 2217 + }, + { + "epoch": 0.13, + "learning_rate": 1.943837161119858e-05, + "loss": 0.3797, + "step": 2218 + }, + { + "epoch": 0.13, + "learning_rate": 1.9437724932837142e-05, + "loss": 0.3634, + "step": 2219 + }, + { + "epoch": 0.13, + "learning_rate": 1.9437077893155417e-05, + "loss": 0.339, + "step": 2220 + }, + { + "epoch": 0.13, + "learning_rate": 1.943643049217819e-05, + "loss": 0.368, + "step": 2221 + }, + { + "epoch": 0.13, + "learning_rate": 1.9435782729930234e-05, + "loss": 0.3616, + "step": 2222 + }, + { + "epoch": 0.13, + "learning_rate": 1.943513460643635e-05, + "loss": 0.3747, + "step": 2223 + }, + { + "epoch": 0.13, + "learning_rate": 1.943448612172136e-05, + "loss": 0.3661, + "step": 2224 + }, + { + "epoch": 0.13, + "learning_rate": 1.9433837275810084e-05, + "loss": 0.4116, + "step": 2225 + }, + { + "epoch": 0.13, + "learning_rate": 1.943318806872736e-05, + "loss": 0.4082, + "step": 2226 + }, + { + "epoch": 0.13, + "learning_rate": 1.9432538500498048e-05, + "loss": 0.3747, + "step": 2227 + }, + { + "epoch": 0.13, + "learning_rate": 1.9431888571147015e-05, + "loss": 0.3605, + "step": 2228 + }, + { + "epoch": 0.13, + "learning_rate": 1.943123828069914e-05, + "loss": 0.4104, + "step": 2229 + }, + { + "epoch": 0.13, + "learning_rate": 1.943058762917933e-05, + "loss": 0.3368, + "step": 2230 + }, + { + "epoch": 0.13, + "learning_rate": 1.942993661661248e-05, + "loss": 0.3747, + "step": 2231 + }, + { + "epoch": 0.13, + "learning_rate": 1.9429285243023527e-05, + "loss": 0.339, + "step": 2232 + }, + { + "epoch": 0.13, + "learning_rate": 1.94286335084374e-05, + "loss": 0.3306, + "step": 2233 + }, + { + "epoch": 0.13, + "learning_rate": 1.9427981412879053e-05, + "loss": 0.3853, + "step": 2234 + }, + { + "epoch": 0.14, + "learning_rate": 1.942732895637345e-05, + "loss": 0.3588, + "step": 2235 + }, + { + "epoch": 0.14, + "learning_rate": 1.9426676138945576e-05, + "loss": 0.3778, + "step": 2236 + }, + { + "epoch": 0.14, + "learning_rate": 1.9426022960620418e-05, + "loss": 0.3616, + "step": 2237 + }, + { + "epoch": 0.14, + "learning_rate": 1.942536942142298e-05, + "loss": 0.3619, + "step": 2238 + }, + { + "epoch": 0.14, + "learning_rate": 1.942471552137829e-05, + "loss": 0.353, + "step": 2239 + }, + { + "epoch": 0.14, + "learning_rate": 1.9424061260511382e-05, + "loss": 0.4082, + "step": 2240 + }, + { + "epoch": 0.14, + "learning_rate": 1.9423406638847296e-05, + "loss": 0.3566, + "step": 2241 + }, + { + "epoch": 0.14, + "learning_rate": 1.9422751656411102e-05, + "loss": 0.353, + "step": 2242 + }, + { + "epoch": 0.14, + "learning_rate": 1.9422096313227875e-05, + "loss": 0.3842, + "step": 2243 + }, + { + "epoch": 0.14, + "learning_rate": 1.94214406093227e-05, + "loss": 0.382, + "step": 2244 + }, + { + "epoch": 0.14, + "learning_rate": 1.9420784544720686e-05, + "loss": 0.4049, + "step": 2245 + }, + { + "epoch": 0.14, + "learning_rate": 1.9420128119446944e-05, + "loss": 0.3591, + "step": 2246 + }, + { + "epoch": 0.14, + "learning_rate": 1.941947133352661e-05, + "loss": 0.3532, + "step": 2247 + }, + { + "epoch": 0.14, + "learning_rate": 1.9418814186984826e-05, + "loss": 0.3521, + "step": 2248 + }, + { + "epoch": 0.14, + "learning_rate": 1.9418156679846753e-05, + "loss": 0.3532, + "step": 2249 + }, + { + "epoch": 0.14, + "learning_rate": 1.9417498812137564e-05, + "loss": 0.3753, + "step": 2250 + }, + { + "epoch": 0.14, + "learning_rate": 1.9416840583882443e-05, + "loss": 0.385, + "step": 2251 + }, + { + "epoch": 0.14, + "learning_rate": 1.9416181995106585e-05, + "loss": 0.3993, + "step": 2252 + }, + { + "epoch": 0.14, + "learning_rate": 1.9415523045835213e-05, + "loss": 0.3379, + "step": 2253 + }, + { + "epoch": 0.14, + "learning_rate": 1.9414863736093554e-05, + "loss": 0.3622, + "step": 2254 + }, + { + "epoch": 0.14, + "learning_rate": 1.9414204065906846e-05, + "loss": 0.3753, + "step": 2255 + }, + { + "epoch": 0.14, + "learning_rate": 1.9413544035300342e-05, + "loss": 0.3912, + "step": 2256 + }, + { + "epoch": 0.14, + "learning_rate": 1.9412883644299317e-05, + "loss": 0.3694, + "step": 2257 + }, + { + "epoch": 0.14, + "learning_rate": 1.941222289292905e-05, + "loss": 0.3491, + "step": 2258 + }, + { + "epoch": 0.14, + "learning_rate": 1.9411561781214837e-05, + "loss": 0.387, + "step": 2259 + }, + { + "epoch": 0.14, + "learning_rate": 1.941090030918199e-05, + "loss": 0.3563, + "step": 2260 + }, + { + "epoch": 0.14, + "learning_rate": 1.9410238476855832e-05, + "loss": 0.3839, + "step": 2261 + }, + { + "epoch": 0.14, + "learning_rate": 1.9409576284261705e-05, + "loss": 0.4074, + "step": 2262 + }, + { + "epoch": 0.14, + "learning_rate": 1.9408913731424956e-05, + "loss": 0.3655, + "step": 2263 + }, + { + "epoch": 0.14, + "learning_rate": 1.9408250818370953e-05, + "loss": 0.385, + "step": 2264 + }, + { + "epoch": 0.14, + "learning_rate": 1.9407587545125077e-05, + "loss": 0.3691, + "step": 2265 + }, + { + "epoch": 0.14, + "learning_rate": 1.9406923911712717e-05, + "loss": 0.3697, + "step": 2266 + }, + { + "epoch": 0.14, + "learning_rate": 1.9406259918159285e-05, + "loss": 0.373, + "step": 2267 + }, + { + "epoch": 0.14, + "learning_rate": 1.9405595564490196e-05, + "loss": 0.3758, + "step": 2268 + }, + { + "epoch": 0.14, + "learning_rate": 1.9404930850730886e-05, + "loss": 0.3616, + "step": 2269 + }, + { + "epoch": 0.14, + "learning_rate": 1.940426577690681e-05, + "loss": 0.4032, + "step": 2270 + }, + { + "epoch": 0.14, + "learning_rate": 1.9403600343043422e-05, + "loss": 0.3658, + "step": 2271 + }, + { + "epoch": 0.14, + "learning_rate": 1.94029345491662e-05, + "loss": 0.4118, + "step": 2272 + }, + { + "epoch": 0.14, + "learning_rate": 1.940226839530064e-05, + "loss": 0.4085, + "step": 2273 + }, + { + "epoch": 0.14, + "learning_rate": 1.9401601881472236e-05, + "loss": 0.389, + "step": 2274 + }, + { + "epoch": 0.14, + "learning_rate": 1.940093500770651e-05, + "loss": 0.3834, + "step": 2275 + }, + { + "epoch": 0.14, + "learning_rate": 1.940026777402899e-05, + "loss": 0.3781, + "step": 2276 + }, + { + "epoch": 0.14, + "learning_rate": 1.9399600180465228e-05, + "loss": 0.4001, + "step": 2277 + }, + { + "epoch": 0.14, + "learning_rate": 1.9398932227040777e-05, + "loss": 0.3675, + "step": 2278 + }, + { + "epoch": 0.14, + "learning_rate": 1.939826391378121e-05, + "loss": 0.3666, + "step": 2279 + }, + { + "epoch": 0.14, + "learning_rate": 1.9397595240712112e-05, + "loss": 0.368, + "step": 2280 + }, + { + "epoch": 0.14, + "learning_rate": 1.9396926207859085e-05, + "loss": 0.3895, + "step": 2281 + }, + { + "epoch": 0.14, + "learning_rate": 1.9396256815247743e-05, + "loss": 0.3848, + "step": 2282 + }, + { + "epoch": 0.14, + "learning_rate": 1.939558706290371e-05, + "loss": 0.3725, + "step": 2283 + }, + { + "epoch": 0.14, + "learning_rate": 1.9394916950852634e-05, + "loss": 0.3876, + "step": 2284 + }, + { + "epoch": 0.14, + "learning_rate": 1.9394246479120165e-05, + "loss": 0.3552, + "step": 2285 + }, + { + "epoch": 0.14, + "learning_rate": 1.939357564773197e-05, + "loss": 0.3792, + "step": 2286 + }, + { + "epoch": 0.14, + "learning_rate": 1.9392904456713737e-05, + "loss": 0.3225, + "step": 2287 + }, + { + "epoch": 0.14, + "learning_rate": 1.939223290609116e-05, + "loss": 0.3733, + "step": 2288 + }, + { + "epoch": 0.14, + "learning_rate": 1.9391560995889948e-05, + "loss": 0.3836, + "step": 2289 + }, + { + "epoch": 0.14, + "learning_rate": 1.9390888726135824e-05, + "loss": 0.4129, + "step": 2290 + }, + { + "epoch": 0.14, + "learning_rate": 1.9390216096854528e-05, + "loss": 0.3549, + "step": 2291 + }, + { + "epoch": 0.14, + "learning_rate": 1.9389543108071808e-05, + "loss": 0.38, + "step": 2292 + }, + { + "epoch": 0.14, + "learning_rate": 1.9388869759813435e-05, + "loss": 0.3881, + "step": 2293 + }, + { + "epoch": 0.14, + "learning_rate": 1.9388196052105185e-05, + "loss": 0.3694, + "step": 2294 + }, + { + "epoch": 0.14, + "learning_rate": 1.9387521984972847e-05, + "loss": 0.3809, + "step": 2295 + }, + { + "epoch": 0.14, + "learning_rate": 1.9386847558442233e-05, + "loss": 0.3585, + "step": 2296 + }, + { + "epoch": 0.14, + "learning_rate": 1.9386172772539162e-05, + "loss": 0.3619, + "step": 2297 + }, + { + "epoch": 0.14, + "learning_rate": 1.9385497627289468e-05, + "loss": 0.3884, + "step": 2298 + }, + { + "epoch": 0.14, + "learning_rate": 1.9384822122718995e-05, + "loss": 0.3362, + "step": 2299 + }, + { + "epoch": 0.14, + "learning_rate": 1.938414625885361e-05, + "loss": 0.3686, + "step": 2300 + }, + { + "epoch": 0.14, + "learning_rate": 1.9383470035719182e-05, + "loss": 0.3714, + "step": 2301 + }, + { + "epoch": 0.14, + "learning_rate": 1.9382793453341602e-05, + "loss": 0.3828, + "step": 2302 + }, + { + "epoch": 0.14, + "learning_rate": 1.938211651174678e-05, + "loss": 0.3717, + "step": 2303 + }, + { + "epoch": 0.14, + "learning_rate": 1.9381439210960624e-05, + "loss": 0.3717, + "step": 2304 + }, + { + "epoch": 0.14, + "learning_rate": 1.9380761551009067e-05, + "loss": 0.4132, + "step": 2305 + }, + { + "epoch": 0.14, + "learning_rate": 1.9380083531918054e-05, + "loss": 0.4099, + "step": 2306 + }, + { + "epoch": 0.14, + "learning_rate": 1.937940515371354e-05, + "loss": 0.3761, + "step": 2307 + }, + { + "epoch": 0.14, + "learning_rate": 1.9378726416421494e-05, + "loss": 0.3421, + "step": 2308 + }, + { + "epoch": 0.14, + "learning_rate": 1.937804732006791e-05, + "loss": 0.3694, + "step": 2309 + }, + { + "epoch": 0.14, + "learning_rate": 1.9377367864678783e-05, + "loss": 0.3792, + "step": 2310 + }, + { + "epoch": 0.14, + "learning_rate": 1.9376688050280124e-05, + "loss": 0.3845, + "step": 2311 + }, + { + "epoch": 0.14, + "learning_rate": 1.937600787689796e-05, + "loss": 0.4065, + "step": 2312 + }, + { + "epoch": 0.14, + "learning_rate": 1.9375327344558334e-05, + "loss": 0.3786, + "step": 2313 + }, + { + "epoch": 0.14, + "learning_rate": 1.9374646453287295e-05, + "loss": 0.3853, + "step": 2314 + }, + { + "epoch": 0.14, + "learning_rate": 1.9373965203110913e-05, + "loss": 0.3717, + "step": 2315 + }, + { + "epoch": 0.14, + "learning_rate": 1.937328359405527e-05, + "loss": 0.3934, + "step": 2316 + }, + { + "epoch": 0.14, + "learning_rate": 1.9372601626146464e-05, + "loss": 0.3675, + "step": 2317 + }, + { + "epoch": 0.14, + "learning_rate": 1.93719192994106e-05, + "loss": 0.3552, + "step": 2318 + }, + { + "epoch": 0.14, + "learning_rate": 1.93712366138738e-05, + "loss": 0.3887, + "step": 2319 + }, + { + "epoch": 0.14, + "learning_rate": 1.93705535695622e-05, + "loss": 0.3778, + "step": 2320 + }, + { + "epoch": 0.14, + "learning_rate": 1.9369870166501958e-05, + "loss": 0.4065, + "step": 2321 + }, + { + "epoch": 0.14, + "learning_rate": 1.9369186404719233e-05, + "loss": 0.3739, + "step": 2322 + }, + { + "epoch": 0.14, + "learning_rate": 1.9368502284240196e-05, + "loss": 0.3728, + "step": 2323 + }, + { + "epoch": 0.14, + "learning_rate": 1.9367817805091045e-05, + "loss": 0.3686, + "step": 2324 + }, + { + "epoch": 0.14, + "learning_rate": 1.9367132967297982e-05, + "loss": 0.3569, + "step": 2325 + }, + { + "epoch": 0.14, + "learning_rate": 1.9366447770887233e-05, + "loss": 0.3792, + "step": 2326 + }, + { + "epoch": 0.14, + "learning_rate": 1.9365762215885025e-05, + "loss": 0.3878, + "step": 2327 + }, + { + "epoch": 0.14, + "learning_rate": 1.9365076302317602e-05, + "loss": 0.3504, + "step": 2328 + }, + { + "epoch": 0.14, + "learning_rate": 1.9364390030211226e-05, + "loss": 0.3742, + "step": 2329 + }, + { + "epoch": 0.14, + "learning_rate": 1.9363703399592174e-05, + "loss": 0.351, + "step": 2330 + }, + { + "epoch": 0.14, + "learning_rate": 1.936301641048673e-05, + "loss": 0.3664, + "step": 2331 + }, + { + "epoch": 0.14, + "learning_rate": 1.9362329062921194e-05, + "loss": 0.358, + "step": 2332 + }, + { + "epoch": 0.14, + "learning_rate": 1.9361641356921883e-05, + "loss": 0.3373, + "step": 2333 + }, + { + "epoch": 0.14, + "learning_rate": 1.936095329251513e-05, + "loss": 0.3691, + "step": 2334 + }, + { + "epoch": 0.14, + "learning_rate": 1.9360264869727267e-05, + "loss": 0.3481, + "step": 2335 + }, + { + "epoch": 0.14, + "learning_rate": 1.935957608858466e-05, + "loss": 0.4182, + "step": 2336 + }, + { + "epoch": 0.14, + "learning_rate": 1.935888694911367e-05, + "loss": 0.3717, + "step": 2337 + }, + { + "epoch": 0.14, + "learning_rate": 1.9358197451340688e-05, + "loss": 0.373, + "step": 2338 + }, + { + "epoch": 0.14, + "learning_rate": 1.9357507595292104e-05, + "loss": 0.3577, + "step": 2339 + }, + { + "epoch": 0.14, + "learning_rate": 1.935681738099434e-05, + "loss": 0.3608, + "step": 2340 + }, + { + "epoch": 0.14, + "learning_rate": 1.935612680847381e-05, + "loss": 0.3555, + "step": 2341 + }, + { + "epoch": 0.14, + "learning_rate": 1.9355435877756957e-05, + "loss": 0.3987, + "step": 2342 + }, + { + "epoch": 0.14, + "learning_rate": 1.935474458887023e-05, + "loss": 0.3471, + "step": 2343 + }, + { + "epoch": 0.14, + "learning_rate": 1.9354052941840102e-05, + "loss": 0.38, + "step": 2344 + }, + { + "epoch": 0.14, + "learning_rate": 1.9353360936693042e-05, + "loss": 0.3672, + "step": 2345 + }, + { + "epoch": 0.14, + "learning_rate": 1.935266857345555e-05, + "loss": 0.4166, + "step": 2346 + }, + { + "epoch": 0.14, + "learning_rate": 1.935197585215413e-05, + "loss": 0.356, + "step": 2347 + }, + { + "epoch": 0.14, + "learning_rate": 1.935128277281531e-05, + "loss": 0.3806, + "step": 2348 + }, + { + "epoch": 0.14, + "learning_rate": 1.9350589335465614e-05, + "loss": 0.3708, + "step": 2349 + }, + { + "epoch": 0.14, + "learning_rate": 1.9349895540131595e-05, + "loss": 0.3541, + "step": 2350 + }, + { + "epoch": 0.14, + "learning_rate": 1.9349201386839815e-05, + "loss": 0.368, + "step": 2351 + }, + { + "epoch": 0.14, + "learning_rate": 1.934850687561685e-05, + "loss": 0.3532, + "step": 2352 + }, + { + "epoch": 0.14, + "learning_rate": 1.9347812006489283e-05, + "loss": 0.3691, + "step": 2353 + }, + { + "epoch": 0.14, + "learning_rate": 1.9347116779483727e-05, + "loss": 0.3647, + "step": 2354 + }, + { + "epoch": 0.14, + "learning_rate": 1.934642119462679e-05, + "loss": 0.3962, + "step": 2355 + }, + { + "epoch": 0.14, + "learning_rate": 1.9345725251945107e-05, + "loss": 0.3491, + "step": 2356 + }, + { + "epoch": 0.14, + "learning_rate": 1.934502895146532e-05, + "loss": 0.3546, + "step": 2357 + }, + { + "epoch": 0.14, + "learning_rate": 1.9344332293214086e-05, + "loss": 0.3856, + "step": 2358 + }, + { + "epoch": 0.14, + "learning_rate": 1.934363527721808e-05, + "loss": 0.3622, + "step": 2359 + }, + { + "epoch": 0.14, + "learning_rate": 1.934293790350398e-05, + "loss": 0.3912, + "step": 2360 + }, + { + "epoch": 0.14, + "learning_rate": 1.9342240172098493e-05, + "loss": 0.375, + "step": 2361 + }, + { + "epoch": 0.14, + "learning_rate": 1.9341542083028327e-05, + "loss": 0.377, + "step": 2362 + }, + { + "epoch": 0.14, + "learning_rate": 1.9340843636320207e-05, + "loss": 0.3744, + "step": 2363 + }, + { + "epoch": 0.14, + "learning_rate": 1.9340144832000876e-05, + "loss": 0.3736, + "step": 2364 + }, + { + "epoch": 0.14, + "learning_rate": 1.9339445670097085e-05, + "loss": 0.3557, + "step": 2365 + }, + { + "epoch": 0.14, + "learning_rate": 1.93387461506356e-05, + "loss": 0.3407, + "step": 2366 + }, + { + "epoch": 0.14, + "learning_rate": 1.933804627364321e-05, + "loss": 0.3624, + "step": 2367 + }, + { + "epoch": 0.14, + "learning_rate": 1.93373460391467e-05, + "loss": 0.4269, + "step": 2368 + }, + { + "epoch": 0.14, + "learning_rate": 1.933664544717288e-05, + "loss": 0.3613, + "step": 2369 + }, + { + "epoch": 0.14, + "learning_rate": 1.9335944497748577e-05, + "loss": 0.3711, + "step": 2370 + }, + { + "epoch": 0.14, + "learning_rate": 1.9335243190900625e-05, + "loss": 0.3602, + "step": 2371 + }, + { + "epoch": 0.14, + "learning_rate": 1.933454152665587e-05, + "loss": 0.3728, + "step": 2372 + }, + { + "epoch": 0.14, + "learning_rate": 1.9333839505041173e-05, + "loss": 0.382, + "step": 2373 + }, + { + "epoch": 0.14, + "learning_rate": 1.9333137126083413e-05, + "loss": 0.3937, + "step": 2374 + }, + { + "epoch": 0.14, + "learning_rate": 1.933243438980949e-05, + "loss": 0.3627, + "step": 2375 + }, + { + "epoch": 0.14, + "learning_rate": 1.9331731296246295e-05, + "loss": 0.3814, + "step": 2376 + }, + { + "epoch": 0.14, + "learning_rate": 1.933102784542075e-05, + "loss": 0.3518, + "step": 2377 + }, + { + "epoch": 0.14, + "learning_rate": 1.933032403735979e-05, + "loss": 0.3504, + "step": 2378 + }, + { + "epoch": 0.14, + "learning_rate": 1.9329619872090354e-05, + "loss": 0.3705, + "step": 2379 + }, + { + "epoch": 0.14, + "learning_rate": 1.9328915349639407e-05, + "loss": 0.3809, + "step": 2380 + }, + { + "epoch": 0.14, + "learning_rate": 1.9328210470033913e-05, + "loss": 0.3703, + "step": 2381 + }, + { + "epoch": 0.14, + "learning_rate": 1.9327505233300864e-05, + "loss": 0.3613, + "step": 2382 + }, + { + "epoch": 0.14, + "learning_rate": 1.9326799639467263e-05, + "loss": 0.3384, + "step": 2383 + }, + { + "epoch": 0.14, + "learning_rate": 1.932609368856012e-05, + "loss": 0.3655, + "step": 2384 + }, + { + "epoch": 0.14, + "learning_rate": 1.9325387380606456e-05, + "loss": 0.3739, + "step": 2385 + }, + { + "epoch": 0.14, + "learning_rate": 1.932468071563332e-05, + "loss": 0.3739, + "step": 2386 + }, + { + "epoch": 0.14, + "learning_rate": 1.9323973693667762e-05, + "loss": 0.3482, + "step": 2387 + }, + { + "epoch": 0.14, + "learning_rate": 1.9323266314736856e-05, + "loss": 0.4063, + "step": 2388 + }, + { + "epoch": 0.14, + "learning_rate": 1.9322558578867675e-05, + "loss": 0.3834, + "step": 2389 + }, + { + "epoch": 0.14, + "learning_rate": 1.9321850486087325e-05, + "loss": 0.3725, + "step": 2390 + }, + { + "epoch": 0.14, + "learning_rate": 1.9321142036422903e-05, + "loss": 0.3382, + "step": 2391 + }, + { + "epoch": 0.14, + "learning_rate": 1.932043322990154e-05, + "loss": 0.3569, + "step": 2392 + }, + { + "epoch": 0.14, + "learning_rate": 1.9319724066550375e-05, + "loss": 0.4007, + "step": 2393 + }, + { + "epoch": 0.14, + "learning_rate": 1.9319014546396547e-05, + "loss": 0.3823, + "step": 2394 + }, + { + "epoch": 0.14, + "learning_rate": 1.931830466946723e-05, + "loss": 0.3717, + "step": 2395 + }, + { + "epoch": 0.14, + "learning_rate": 1.9317594435789595e-05, + "loss": 0.377, + "step": 2396 + }, + { + "epoch": 0.14, + "learning_rate": 1.9316883845390836e-05, + "loss": 0.3691, + "step": 2397 + }, + { + "epoch": 0.14, + "learning_rate": 1.931617289829816e-05, + "loss": 0.3689, + "step": 2398 + }, + { + "epoch": 0.14, + "learning_rate": 1.9315461594538777e-05, + "loss": 0.3959, + "step": 2399 + }, + { + "epoch": 0.14, + "learning_rate": 1.9314749934139932e-05, + "loss": 0.3432, + "step": 2400 + }, + { + "epoch": 0.15, + "learning_rate": 1.931403791712886e-05, + "loss": 0.3806, + "step": 2401 + }, + { + "epoch": 0.15, + "learning_rate": 1.9313325543532825e-05, + "loss": 0.3761, + "step": 2402 + }, + { + "epoch": 0.15, + "learning_rate": 1.9312612813379097e-05, + "loss": 0.392, + "step": 2403 + }, + { + "epoch": 0.15, + "learning_rate": 1.9311899726694964e-05, + "loss": 0.3881, + "step": 2404 + }, + { + "epoch": 0.15, + "learning_rate": 1.931118628350773e-05, + "loss": 0.3557, + "step": 2405 + }, + { + "epoch": 0.15, + "learning_rate": 1.9310472483844706e-05, + "loss": 0.3887, + "step": 2406 + }, + { + "epoch": 0.15, + "learning_rate": 1.9309758327733213e-05, + "loss": 0.3786, + "step": 2407 + }, + { + "epoch": 0.15, + "learning_rate": 1.9309043815200604e-05, + "loss": 0.3577, + "step": 2408 + }, + { + "epoch": 0.15, + "learning_rate": 1.9308328946274227e-05, + "loss": 0.363, + "step": 2409 + }, + { + "epoch": 0.15, + "learning_rate": 1.930761372098145e-05, + "loss": 0.3689, + "step": 2410 + }, + { + "epoch": 0.15, + "learning_rate": 1.9306898139349664e-05, + "loss": 0.3569, + "step": 2411 + }, + { + "epoch": 0.15, + "learning_rate": 1.9306182201406253e-05, + "loss": 0.4076, + "step": 2412 + }, + { + "epoch": 0.15, + "learning_rate": 1.9305465907178633e-05, + "loss": 0.3982, + "step": 2413 + }, + { + "epoch": 0.15, + "learning_rate": 1.9304749256694225e-05, + "loss": 0.3998, + "step": 2414 + }, + { + "epoch": 0.15, + "learning_rate": 1.9304032249980464e-05, + "loss": 0.3552, + "step": 2415 + }, + { + "epoch": 0.15, + "learning_rate": 1.9303314887064805e-05, + "loss": 0.3616, + "step": 2416 + }, + { + "epoch": 0.15, + "learning_rate": 1.930259716797471e-05, + "loss": 0.3951, + "step": 2417 + }, + { + "epoch": 0.15, + "learning_rate": 1.9301879092737655e-05, + "loss": 0.3351, + "step": 2418 + }, + { + "epoch": 0.15, + "learning_rate": 1.9301160661381135e-05, + "loss": 0.3541, + "step": 2419 + }, + { + "epoch": 0.15, + "learning_rate": 1.9300441873932648e-05, + "loss": 0.3733, + "step": 2420 + }, + { + "epoch": 0.15, + "learning_rate": 1.929972273041972e-05, + "loss": 0.3711, + "step": 2421 + }, + { + "epoch": 0.15, + "learning_rate": 1.929900323086988e-05, + "loss": 0.3315, + "step": 2422 + }, + { + "epoch": 0.15, + "learning_rate": 1.9298283375310675e-05, + "loss": 0.3834, + "step": 2423 + }, + { + "epoch": 0.15, + "learning_rate": 1.9297563163769658e-05, + "loss": 0.3959, + "step": 2424 + }, + { + "epoch": 0.15, + "learning_rate": 1.9296842596274413e-05, + "loss": 0.3382, + "step": 2425 + }, + { + "epoch": 0.15, + "learning_rate": 1.929612167285252e-05, + "loss": 0.3499, + "step": 2426 + }, + { + "epoch": 0.15, + "learning_rate": 1.929540039353158e-05, + "loss": 0.3435, + "step": 2427 + }, + { + "epoch": 0.15, + "learning_rate": 1.929467875833921e-05, + "loss": 0.3655, + "step": 2428 + }, + { + "epoch": 0.15, + "learning_rate": 1.9293956767303032e-05, + "loss": 0.3898, + "step": 2429 + }, + { + "epoch": 0.15, + "learning_rate": 1.9293234420450692e-05, + "loss": 0.3619, + "step": 2430 + }, + { + "epoch": 0.15, + "learning_rate": 1.929251171780984e-05, + "loss": 0.3892, + "step": 2431 + }, + { + "epoch": 0.15, + "learning_rate": 1.929178865940815e-05, + "loss": 0.3705, + "step": 2432 + }, + { + "epoch": 0.15, + "learning_rate": 1.92910652452733e-05, + "loss": 0.3407, + "step": 2433 + }, + { + "epoch": 0.15, + "learning_rate": 1.929034147543299e-05, + "loss": 0.3465, + "step": 2434 + }, + { + "epoch": 0.15, + "learning_rate": 1.9289617349914924e-05, + "loss": 0.3446, + "step": 2435 + }, + { + "epoch": 0.15, + "learning_rate": 1.9288892868746832e-05, + "loss": 0.3703, + "step": 2436 + }, + { + "epoch": 0.15, + "learning_rate": 1.928816803195644e-05, + "loss": 0.4046, + "step": 2437 + }, + { + "epoch": 0.15, + "learning_rate": 1.9287442839571508e-05, + "loss": 0.3753, + "step": 2438 + }, + { + "epoch": 0.15, + "learning_rate": 1.9286717291619796e-05, + "loss": 0.3602, + "step": 2439 + }, + { + "epoch": 0.15, + "learning_rate": 1.9285991388129077e-05, + "loss": 0.3521, + "step": 2440 + }, + { + "epoch": 0.15, + "learning_rate": 1.928526512912715e-05, + "loss": 0.332, + "step": 2441 + }, + { + "epoch": 0.15, + "learning_rate": 1.9284538514641815e-05, + "loss": 0.3831, + "step": 2442 + }, + { + "epoch": 0.15, + "learning_rate": 1.928381154470089e-05, + "loss": 0.3968, + "step": 2443 + }, + { + "epoch": 0.15, + "learning_rate": 1.928308421933221e-05, + "loss": 0.3756, + "step": 2444 + }, + { + "epoch": 0.15, + "learning_rate": 1.928235653856362e-05, + "loss": 0.3739, + "step": 2445 + }, + { + "epoch": 0.15, + "learning_rate": 1.9281628502422975e-05, + "loss": 0.3471, + "step": 2446 + }, + { + "epoch": 0.15, + "learning_rate": 1.928090011093815e-05, + "loss": 0.358, + "step": 2447 + }, + { + "epoch": 0.15, + "learning_rate": 1.9280171364137027e-05, + "loss": 0.3998, + "step": 2448 + }, + { + "epoch": 0.15, + "learning_rate": 1.9279442262047517e-05, + "loss": 0.3809, + "step": 2449 + }, + { + "epoch": 0.15, + "learning_rate": 1.927871280469752e-05, + "loss": 0.3945, + "step": 2450 + }, + { + "epoch": 0.15, + "learning_rate": 1.9277982992114975e-05, + "loss": 0.3362, + "step": 2451 + }, + { + "epoch": 0.15, + "learning_rate": 1.9277252824327814e-05, + "loss": 0.3599, + "step": 2452 + }, + { + "epoch": 0.15, + "learning_rate": 1.9276522301363997e-05, + "loss": 0.3633, + "step": 2453 + }, + { + "epoch": 0.15, + "learning_rate": 1.9275791423251486e-05, + "loss": 0.3711, + "step": 2454 + }, + { + "epoch": 0.15, + "learning_rate": 1.9275060190018267e-05, + "loss": 0.3853, + "step": 2455 + }, + { + "epoch": 0.15, + "learning_rate": 1.9274328601692334e-05, + "loss": 0.3557, + "step": 2456 + }, + { + "epoch": 0.15, + "learning_rate": 1.9273596658301694e-05, + "loss": 0.3569, + "step": 2457 + }, + { + "epoch": 0.15, + "learning_rate": 1.9272864359874373e-05, + "loss": 0.3761, + "step": 2458 + }, + { + "epoch": 0.15, + "learning_rate": 1.9272131706438403e-05, + "loss": 0.3862, + "step": 2459 + }, + { + "epoch": 0.15, + "learning_rate": 1.927139869802183e-05, + "loss": 0.3421, + "step": 2460 + }, + { + "epoch": 0.15, + "learning_rate": 1.927066533465273e-05, + "loss": 0.3956, + "step": 2461 + }, + { + "epoch": 0.15, + "learning_rate": 1.926993161635917e-05, + "loss": 0.3518, + "step": 2462 + }, + { + "epoch": 0.15, + "learning_rate": 1.926919754316924e-05, + "loss": 0.356, + "step": 2463 + }, + { + "epoch": 0.15, + "learning_rate": 1.926846311511104e-05, + "loss": 0.3686, + "step": 2464 + }, + { + "epoch": 0.15, + "learning_rate": 1.92677283322127e-05, + "loss": 0.3691, + "step": 2465 + }, + { + "epoch": 0.15, + "learning_rate": 1.926699319450234e-05, + "loss": 0.3658, + "step": 2466 + }, + { + "epoch": 0.15, + "learning_rate": 1.9266257702008108e-05, + "loss": 0.3599, + "step": 2467 + }, + { + "epoch": 0.15, + "learning_rate": 1.9265521854758163e-05, + "loss": 0.3502, + "step": 2468 + }, + { + "epoch": 0.15, + "learning_rate": 1.9264785652780677e-05, + "loss": 0.3488, + "step": 2469 + }, + { + "epoch": 0.15, + "learning_rate": 1.926404909610383e-05, + "loss": 0.3465, + "step": 2470 + }, + { + "epoch": 0.15, + "learning_rate": 1.9263312184755828e-05, + "loss": 0.3578, + "step": 2471 + }, + { + "epoch": 0.15, + "learning_rate": 1.926257491876488e-05, + "loss": 0.3514, + "step": 2472 + }, + { + "epoch": 0.15, + "learning_rate": 1.9261837298159213e-05, + "loss": 0.3689, + "step": 2473 + }, + { + "epoch": 0.15, + "learning_rate": 1.9261099322967064e-05, + "loss": 0.3892, + "step": 2474 + }, + { + "epoch": 0.15, + "learning_rate": 1.9260360993216692e-05, + "loss": 0.3694, + "step": 2475 + }, + { + "epoch": 0.15, + "learning_rate": 1.9259622308936353e-05, + "loss": 0.3251, + "step": 2476 + }, + { + "epoch": 0.15, + "learning_rate": 1.925888327015434e-05, + "loss": 0.3987, + "step": 2477 + }, + { + "epoch": 0.15, + "learning_rate": 1.9258143876898938e-05, + "loss": 0.3806, + "step": 2478 + }, + { + "epoch": 0.15, + "learning_rate": 1.9257404129198458e-05, + "loss": 0.3658, + "step": 2479 + }, + { + "epoch": 0.15, + "learning_rate": 1.9256664027081222e-05, + "loss": 0.3563, + "step": 2480 + }, + { + "epoch": 0.15, + "learning_rate": 1.9255923570575562e-05, + "loss": 0.3717, + "step": 2481 + }, + { + "epoch": 0.15, + "learning_rate": 1.9255182759709826e-05, + "loss": 0.3881, + "step": 2482 + }, + { + "epoch": 0.15, + "learning_rate": 1.9254441594512373e-05, + "loss": 0.3471, + "step": 2483 + }, + { + "epoch": 0.15, + "learning_rate": 1.9253700075011586e-05, + "loss": 0.382, + "step": 2484 + }, + { + "epoch": 0.15, + "learning_rate": 1.9252958201235853e-05, + "loss": 0.3783, + "step": 2485 + }, + { + "epoch": 0.15, + "learning_rate": 1.925221597321357e-05, + "loss": 0.3792, + "step": 2486 + }, + { + "epoch": 0.15, + "learning_rate": 1.9251473390973156e-05, + "loss": 0.3544, + "step": 2487 + }, + { + "epoch": 0.15, + "learning_rate": 1.925073045454304e-05, + "loss": 0.3608, + "step": 2488 + }, + { + "epoch": 0.15, + "learning_rate": 1.9249987163951668e-05, + "loss": 0.3761, + "step": 2489 + }, + { + "epoch": 0.15, + "learning_rate": 1.9249243519227493e-05, + "loss": 0.3917, + "step": 2490 + }, + { + "epoch": 0.15, + "learning_rate": 1.9248499520398985e-05, + "loss": 0.377, + "step": 2491 + }, + { + "epoch": 0.15, + "learning_rate": 1.9247755167494632e-05, + "loss": 0.3742, + "step": 2492 + }, + { + "epoch": 0.15, + "learning_rate": 1.924701046054293e-05, + "loss": 0.3538, + "step": 2493 + }, + { + "epoch": 0.15, + "learning_rate": 1.9246265399572387e-05, + "loss": 0.3703, + "step": 2494 + }, + { + "epoch": 0.15, + "learning_rate": 1.9245519984611528e-05, + "loss": 0.3661, + "step": 2495 + }, + { + "epoch": 0.15, + "learning_rate": 1.924477421568889e-05, + "loss": 0.3996, + "step": 2496 + }, + { + "epoch": 0.15, + "learning_rate": 1.9244028092833033e-05, + "loss": 0.3753, + "step": 2497 + }, + { + "epoch": 0.15, + "learning_rate": 1.9243281616072515e-05, + "loss": 0.3951, + "step": 2498 + }, + { + "epoch": 0.15, + "learning_rate": 1.924253478543591e-05, + "loss": 0.3345, + "step": 2499 + }, + { + "epoch": 0.15, + "learning_rate": 1.9241787600951814e-05, + "loss": 0.3569, + "step": 2500 + }, + { + "epoch": 0.15, + "learning_rate": 1.924104006264884e-05, + "loss": 0.3371, + "step": 2501 + }, + { + "epoch": 0.15, + "learning_rate": 1.9240292170555597e-05, + "loss": 0.3929, + "step": 2502 + }, + { + "epoch": 0.15, + "learning_rate": 1.9239543924700727e-05, + "loss": 0.3563, + "step": 2503 + }, + { + "epoch": 0.15, + "learning_rate": 1.9238795325112867e-05, + "loss": 0.3661, + "step": 2504 + }, + { + "epoch": 0.15, + "learning_rate": 1.9238046371820688e-05, + "loss": 0.3909, + "step": 2505 + }, + { + "epoch": 0.15, + "learning_rate": 1.923729706485285e-05, + "loss": 0.3658, + "step": 2506 + }, + { + "epoch": 0.15, + "learning_rate": 1.9236547404238048e-05, + "loss": 0.3742, + "step": 2507 + }, + { + "epoch": 0.15, + "learning_rate": 1.9235797390004985e-05, + "loss": 0.3373, + "step": 2508 + }, + { + "epoch": 0.15, + "learning_rate": 1.9235047022182366e-05, + "loss": 0.3823, + "step": 2509 + }, + { + "epoch": 0.15, + "learning_rate": 1.923429630079893e-05, + "loss": 0.3742, + "step": 2510 + }, + { + "epoch": 0.15, + "learning_rate": 1.9233545225883407e-05, + "loss": 0.3722, + "step": 2511 + }, + { + "epoch": 0.15, + "learning_rate": 1.923279379746456e-05, + "loss": 0.3965, + "step": 2512 + }, + { + "epoch": 0.15, + "learning_rate": 1.9232042015571152e-05, + "loss": 0.3602, + "step": 2513 + }, + { + "epoch": 0.15, + "learning_rate": 1.923128988023197e-05, + "loss": 0.3566, + "step": 2514 + }, + { + "epoch": 0.15, + "learning_rate": 1.9230537391475802e-05, + "loss": 0.3728, + "step": 2515 + }, + { + "epoch": 0.15, + "learning_rate": 1.922978454933146e-05, + "loss": 0.3848, + "step": 2516 + }, + { + "epoch": 0.15, + "learning_rate": 1.9229031353827773e-05, + "loss": 0.3677, + "step": 2517 + }, + { + "epoch": 0.15, + "learning_rate": 1.9228277804993568e-05, + "loss": 0.353, + "step": 2518 + }, + { + "epoch": 0.15, + "learning_rate": 1.9227523902857695e-05, + "loss": 0.3717, + "step": 2519 + }, + { + "epoch": 0.15, + "learning_rate": 1.922676964744902e-05, + "loss": 0.3917, + "step": 2520 + }, + { + "epoch": 0.15, + "learning_rate": 1.922601503879642e-05, + "loss": 0.3616, + "step": 2521 + }, + { + "epoch": 0.15, + "learning_rate": 1.9225260076928783e-05, + "loss": 0.3474, + "step": 2522 + }, + { + "epoch": 0.15, + "learning_rate": 1.9224504761875016e-05, + "loss": 0.3677, + "step": 2523 + }, + { + "epoch": 0.15, + "learning_rate": 1.9223749093664027e-05, + "loss": 0.373, + "step": 2524 + }, + { + "epoch": 0.15, + "learning_rate": 1.9222993072324758e-05, + "loss": 0.3513, + "step": 2525 + }, + { + "epoch": 0.15, + "learning_rate": 1.9222236697886148e-05, + "loss": 0.3664, + "step": 2526 + }, + { + "epoch": 0.15, + "learning_rate": 1.922147997037715e-05, + "loss": 0.37, + "step": 2527 + }, + { + "epoch": 0.15, + "learning_rate": 1.9220722889826743e-05, + "loss": 0.3831, + "step": 2528 + }, + { + "epoch": 0.15, + "learning_rate": 1.9219965456263906e-05, + "loss": 0.3446, + "step": 2529 + }, + { + "epoch": 0.15, + "learning_rate": 1.9219207669717638e-05, + "loss": 0.3697, + "step": 2530 + }, + { + "epoch": 0.15, + "learning_rate": 1.9218449530216953e-05, + "loss": 0.361, + "step": 2531 + }, + { + "epoch": 0.15, + "learning_rate": 1.9217691037790872e-05, + "loss": 0.3853, + "step": 2532 + }, + { + "epoch": 0.15, + "learning_rate": 1.921693219246844e-05, + "loss": 0.3597, + "step": 2533 + }, + { + "epoch": 0.15, + "learning_rate": 1.9216172994278706e-05, + "loss": 0.387, + "step": 2534 + }, + { + "epoch": 0.15, + "learning_rate": 1.9215413443250733e-05, + "loss": 0.3689, + "step": 2535 + }, + { + "epoch": 0.15, + "learning_rate": 1.9214653539413602e-05, + "loss": 0.3599, + "step": 2536 + }, + { + "epoch": 0.15, + "learning_rate": 1.9213893282796407e-05, + "loss": 0.3739, + "step": 2537 + }, + { + "epoch": 0.15, + "learning_rate": 1.921313267342825e-05, + "loss": 0.373, + "step": 2538 + }, + { + "epoch": 0.15, + "learning_rate": 1.921237171133826e-05, + "loss": 0.3463, + "step": 2539 + }, + { + "epoch": 0.15, + "learning_rate": 1.921161039655556e-05, + "loss": 0.3853, + "step": 2540 + }, + { + "epoch": 0.15, + "learning_rate": 1.92108487291093e-05, + "loss": 0.3387, + "step": 2541 + }, + { + "epoch": 0.15, + "learning_rate": 1.921008670902864e-05, + "loss": 0.3917, + "step": 2542 + }, + { + "epoch": 0.15, + "learning_rate": 1.920932433634276e-05, + "loss": 0.3566, + "step": 2543 + }, + { + "epoch": 0.15, + "learning_rate": 1.920856161108084e-05, + "loss": 0.377, + "step": 2544 + }, + { + "epoch": 0.15, + "learning_rate": 1.9207798533272083e-05, + "loss": 0.3775, + "step": 2545 + }, + { + "epoch": 0.15, + "learning_rate": 1.9207035102945698e-05, + "loss": 0.3915, + "step": 2546 + }, + { + "epoch": 0.15, + "learning_rate": 1.920627132013092e-05, + "loss": 0.3479, + "step": 2547 + }, + { + "epoch": 0.15, + "learning_rate": 1.920550718485699e-05, + "loss": 0.3853, + "step": 2548 + }, + { + "epoch": 0.15, + "learning_rate": 1.9204742697153155e-05, + "loss": 0.4054, + "step": 2549 + }, + { + "epoch": 0.15, + "learning_rate": 1.9203977857048693e-05, + "loss": 0.385, + "step": 2550 + }, + { + "epoch": 0.15, + "learning_rate": 1.920321266457288e-05, + "loss": 0.3549, + "step": 2551 + }, + { + "epoch": 0.15, + "learning_rate": 1.9202447119755012e-05, + "loss": 0.3364, + "step": 2552 + }, + { + "epoch": 0.15, + "learning_rate": 1.9201681222624396e-05, + "loss": 0.3624, + "step": 2553 + }, + { + "epoch": 0.15, + "learning_rate": 1.920091497321036e-05, + "loss": 0.4138, + "step": 2554 + }, + { + "epoch": 0.15, + "learning_rate": 1.9200148371542235e-05, + "loss": 0.3848, + "step": 2555 + }, + { + "epoch": 0.15, + "learning_rate": 1.919938141764937e-05, + "loss": 0.3842, + "step": 2556 + }, + { + "epoch": 0.15, + "learning_rate": 1.9198614111561125e-05, + "loss": 0.3856, + "step": 2557 + }, + { + "epoch": 0.15, + "learning_rate": 1.919784645330688e-05, + "loss": 0.3426, + "step": 2558 + }, + { + "epoch": 0.15, + "learning_rate": 1.9197078442916028e-05, + "loss": 0.3474, + "step": 2559 + }, + { + "epoch": 0.15, + "learning_rate": 1.9196310080417967e-05, + "loss": 0.3463, + "step": 2560 + }, + { + "epoch": 0.15, + "learning_rate": 1.919554136584211e-05, + "loss": 0.358, + "step": 2561 + }, + { + "epoch": 0.15, + "learning_rate": 1.9194772299217897e-05, + "loss": 0.3744, + "step": 2562 + }, + { + "epoch": 0.15, + "learning_rate": 1.9194002880574762e-05, + "loss": 0.373, + "step": 2563 + }, + { + "epoch": 0.15, + "learning_rate": 1.9193233109942167e-05, + "loss": 0.3463, + "step": 2564 + }, + { + "epoch": 0.15, + "learning_rate": 1.919246298734958e-05, + "loss": 0.3524, + "step": 2565 + }, + { + "epoch": 0.16, + "learning_rate": 1.9191692512826492e-05, + "loss": 0.3747, + "step": 2566 + }, + { + "epoch": 0.16, + "learning_rate": 1.919092168640239e-05, + "loss": 0.3839, + "step": 2567 + }, + { + "epoch": 0.16, + "learning_rate": 1.919015050810679e-05, + "loss": 0.3697, + "step": 2568 + }, + { + "epoch": 0.16, + "learning_rate": 1.9189378977969214e-05, + "loss": 0.3683, + "step": 2569 + }, + { + "epoch": 0.16, + "learning_rate": 1.9188607096019205e-05, + "loss": 0.3733, + "step": 2570 + }, + { + "epoch": 0.16, + "learning_rate": 1.918783486228631e-05, + "loss": 0.3767, + "step": 2571 + }, + { + "epoch": 0.16, + "learning_rate": 1.918706227680009e-05, + "loss": 0.3756, + "step": 2572 + }, + { + "epoch": 0.16, + "learning_rate": 1.918628933959013e-05, + "loss": 0.3781, + "step": 2573 + }, + { + "epoch": 0.16, + "learning_rate": 1.9185516050686023e-05, + "loss": 0.3722, + "step": 2574 + }, + { + "epoch": 0.16, + "learning_rate": 1.918474241011737e-05, + "loss": 0.3468, + "step": 2575 + }, + { + "epoch": 0.16, + "learning_rate": 1.9183968417913785e-05, + "loss": 0.3566, + "step": 2576 + }, + { + "epoch": 0.16, + "learning_rate": 1.9183194074104907e-05, + "loss": 0.3848, + "step": 2577 + }, + { + "epoch": 0.16, + "learning_rate": 1.918241937872038e-05, + "loss": 0.3641, + "step": 2578 + }, + { + "epoch": 0.16, + "learning_rate": 1.9181644331789863e-05, + "loss": 0.3703, + "step": 2579 + }, + { + "epoch": 0.16, + "learning_rate": 1.918086893334303e-05, + "loss": 0.3993, + "step": 2580 + }, + { + "epoch": 0.16, + "learning_rate": 1.918009318340956e-05, + "loss": 0.358, + "step": 2581 + }, + { + "epoch": 0.16, + "learning_rate": 1.917931708201916e-05, + "loss": 0.3828, + "step": 2582 + }, + { + "epoch": 0.16, + "learning_rate": 1.9178540629201537e-05, + "loss": 0.3666, + "step": 2583 + }, + { + "epoch": 0.16, + "learning_rate": 1.9177763824986425e-05, + "loss": 0.3474, + "step": 2584 + }, + { + "epoch": 0.16, + "learning_rate": 1.9176986669403556e-05, + "loss": 0.3569, + "step": 2585 + }, + { + "epoch": 0.16, + "learning_rate": 1.9176209162482687e-05, + "loss": 0.3552, + "step": 2586 + }, + { + "epoch": 0.16, + "learning_rate": 1.9175431304253585e-05, + "loss": 0.3491, + "step": 2587 + }, + { + "epoch": 0.16, + "learning_rate": 1.917465309474603e-05, + "loss": 0.3867, + "step": 2588 + }, + { + "epoch": 0.16, + "learning_rate": 1.917387453398981e-05, + "loss": 0.3691, + "step": 2589 + }, + { + "epoch": 0.16, + "learning_rate": 1.9173095622014736e-05, + "loss": 0.3764, + "step": 2590 + }, + { + "epoch": 0.16, + "learning_rate": 1.917231635885063e-05, + "loss": 0.3898, + "step": 2591 + }, + { + "epoch": 0.16, + "learning_rate": 1.9171536744527327e-05, + "loss": 0.377, + "step": 2592 + }, + { + "epoch": 0.16, + "learning_rate": 1.917075677907467e-05, + "loss": 0.3633, + "step": 2593 + }, + { + "epoch": 0.16, + "learning_rate": 1.916997646252252e-05, + "loss": 0.3714, + "step": 2594 + }, + { + "epoch": 0.16, + "learning_rate": 1.9169195794900752e-05, + "loss": 0.37, + "step": 2595 + }, + { + "epoch": 0.16, + "learning_rate": 1.9168414776239258e-05, + "loss": 0.3652, + "step": 2596 + }, + { + "epoch": 0.16, + "learning_rate": 1.916763340656793e-05, + "loss": 0.3599, + "step": 2597 + }, + { + "epoch": 0.16, + "learning_rate": 1.9166851685916693e-05, + "loss": 0.3683, + "step": 2598 + }, + { + "epoch": 0.16, + "learning_rate": 1.9166069614315468e-05, + "loss": 0.3329, + "step": 2599 + }, + { + "epoch": 0.16, + "learning_rate": 1.9165287191794194e-05, + "loss": 0.3661, + "step": 2600 + }, + { + "epoch": 0.16, + "learning_rate": 1.9164504418382834e-05, + "loss": 0.3407, + "step": 2601 + }, + { + "epoch": 0.16, + "learning_rate": 1.9163721294111352e-05, + "loss": 0.3825, + "step": 2602 + }, + { + "epoch": 0.16, + "learning_rate": 1.916293781900973e-05, + "loss": 0.3856, + "step": 2603 + }, + { + "epoch": 0.16, + "learning_rate": 1.9162153993107962e-05, + "loss": 0.3577, + "step": 2604 + }, + { + "epoch": 0.16, + "learning_rate": 1.9161369816436055e-05, + "loss": 0.3449, + "step": 2605 + }, + { + "epoch": 0.16, + "learning_rate": 1.9160585289024037e-05, + "loss": 0.3736, + "step": 2606 + }, + { + "epoch": 0.16, + "learning_rate": 1.915980041090194e-05, + "loss": 0.3555, + "step": 2607 + }, + { + "epoch": 0.16, + "learning_rate": 1.9159015182099807e-05, + "loss": 0.3532, + "step": 2608 + }, + { + "epoch": 0.16, + "learning_rate": 1.9158229602647708e-05, + "loss": 0.3633, + "step": 2609 + }, + { + "epoch": 0.16, + "learning_rate": 1.915744367257572e-05, + "loss": 0.3622, + "step": 2610 + }, + { + "epoch": 0.16, + "learning_rate": 1.9156657391913927e-05, + "loss": 0.3597, + "step": 2611 + }, + { + "epoch": 0.16, + "learning_rate": 1.915587076069243e-05, + "loss": 0.3415, + "step": 2612 + }, + { + "epoch": 0.16, + "learning_rate": 1.9155083778941353e-05, + "loss": 0.3524, + "step": 2613 + }, + { + "epoch": 0.16, + "learning_rate": 1.9154296446690814e-05, + "loss": 0.3689, + "step": 2614 + }, + { + "epoch": 0.16, + "learning_rate": 1.9153508763970966e-05, + "loss": 0.3541, + "step": 2615 + }, + { + "epoch": 0.16, + "learning_rate": 1.915272073081196e-05, + "loss": 0.4202, + "step": 2616 + }, + { + "epoch": 0.16, + "learning_rate": 1.915193234724397e-05, + "loss": 0.3382, + "step": 2617 + }, + { + "epoch": 0.16, + "learning_rate": 1.9151143613297168e-05, + "loss": 0.3742, + "step": 2618 + }, + { + "epoch": 0.16, + "learning_rate": 1.9150354529001765e-05, + "loss": 0.3679, + "step": 2619 + }, + { + "epoch": 0.16, + "learning_rate": 1.9149565094387962e-05, + "loss": 0.3775, + "step": 2620 + }, + { + "epoch": 0.16, + "learning_rate": 1.9148775309485985e-05, + "loss": 0.3789, + "step": 2621 + }, + { + "epoch": 0.16, + "learning_rate": 1.9147985174326068e-05, + "loss": 0.3898, + "step": 2622 + }, + { + "epoch": 0.16, + "learning_rate": 1.9147194688938465e-05, + "loss": 0.3463, + "step": 2623 + }, + { + "epoch": 0.16, + "learning_rate": 1.9146403853353438e-05, + "loss": 0.3602, + "step": 2624 + }, + { + "epoch": 0.16, + "learning_rate": 1.9145612667601263e-05, + "loss": 0.373, + "step": 2625 + }, + { + "epoch": 0.16, + "learning_rate": 1.914482113171223e-05, + "loss": 0.3982, + "step": 2626 + }, + { + "epoch": 0.16, + "learning_rate": 1.9144029245716642e-05, + "loss": 0.3739, + "step": 2627 + }, + { + "epoch": 0.16, + "learning_rate": 1.9143237009644823e-05, + "loss": 0.3789, + "step": 2628 + }, + { + "epoch": 0.16, + "learning_rate": 1.914244442352709e-05, + "loss": 0.368, + "step": 2629 + }, + { + "epoch": 0.16, + "learning_rate": 1.9141651487393802e-05, + "loss": 0.3624, + "step": 2630 + }, + { + "epoch": 0.16, + "learning_rate": 1.9140858201275306e-05, + "loss": 0.3544, + "step": 2631 + }, + { + "epoch": 0.16, + "learning_rate": 1.9140064565201973e-05, + "loss": 0.3811, + "step": 2632 + }, + { + "epoch": 0.16, + "learning_rate": 1.9139270579204195e-05, + "loss": 0.3502, + "step": 2633 + }, + { + "epoch": 0.16, + "learning_rate": 1.9138476243312363e-05, + "loss": 0.363, + "step": 2634 + }, + { + "epoch": 0.16, + "learning_rate": 1.9137681557556886e-05, + "loss": 0.3622, + "step": 2635 + }, + { + "epoch": 0.16, + "learning_rate": 1.9136886521968195e-05, + "loss": 0.3641, + "step": 2636 + }, + { + "epoch": 0.16, + "learning_rate": 1.9136091136576727e-05, + "loss": 0.3421, + "step": 2637 + }, + { + "epoch": 0.16, + "learning_rate": 1.9135295401412928e-05, + "loss": 0.3624, + "step": 2638 + }, + { + "epoch": 0.16, + "learning_rate": 1.9134499316507266e-05, + "loss": 0.3499, + "step": 2639 + }, + { + "epoch": 0.16, + "learning_rate": 1.913370288189021e-05, + "loss": 0.3504, + "step": 2640 + }, + { + "epoch": 0.16, + "learning_rate": 1.9132906097592267e-05, + "loss": 0.328, + "step": 2641 + }, + { + "epoch": 0.16, + "learning_rate": 1.9132108963643934e-05, + "loss": 0.3563, + "step": 2642 + }, + { + "epoch": 0.16, + "learning_rate": 1.9131311480075724e-05, + "loss": 0.3404, + "step": 2643 + }, + { + "epoch": 0.16, + "learning_rate": 1.913051364691818e-05, + "loss": 0.373, + "step": 2644 + }, + { + "epoch": 0.16, + "learning_rate": 1.9129715464201833e-05, + "loss": 0.3602, + "step": 2645 + }, + { + "epoch": 0.16, + "learning_rate": 1.912891693195725e-05, + "loss": 0.3672, + "step": 2646 + }, + { + "epoch": 0.16, + "learning_rate": 1.9128118050215003e-05, + "loss": 0.3647, + "step": 2647 + }, + { + "epoch": 0.16, + "learning_rate": 1.9127318819005672e-05, + "loss": 0.3647, + "step": 2648 + }, + { + "epoch": 0.16, + "learning_rate": 1.9126519238359858e-05, + "loss": 0.3373, + "step": 2649 + }, + { + "epoch": 0.16, + "learning_rate": 1.9125719308308176e-05, + "loss": 0.382, + "step": 2650 + }, + { + "epoch": 0.16, + "learning_rate": 1.9124919028881244e-05, + "loss": 0.3739, + "step": 2651 + }, + { + "epoch": 0.16, + "learning_rate": 1.9124118400109702e-05, + "loss": 0.3644, + "step": 2652 + }, + { + "epoch": 0.16, + "learning_rate": 1.9123317422024212e-05, + "loss": 0.3524, + "step": 2653 + }, + { + "epoch": 0.16, + "learning_rate": 1.9122516094655423e-05, + "loss": 0.3708, + "step": 2654 + }, + { + "epoch": 0.16, + "learning_rate": 1.9121714418034024e-05, + "loss": 0.3354, + "step": 2655 + }, + { + "epoch": 0.16, + "learning_rate": 1.9120912392190706e-05, + "loss": 0.351, + "step": 2656 + }, + { + "epoch": 0.16, + "learning_rate": 1.9120110017156172e-05, + "loss": 0.4121, + "step": 2657 + }, + { + "epoch": 0.16, + "learning_rate": 1.911930729296114e-05, + "loss": 0.3571, + "step": 2658 + }, + { + "epoch": 0.16, + "learning_rate": 1.9118504219636346e-05, + "loss": 0.3516, + "step": 2659 + }, + { + "epoch": 0.16, + "learning_rate": 1.911770079721253e-05, + "loss": 0.3284, + "step": 2660 + }, + { + "epoch": 0.16, + "learning_rate": 1.9116897025720455e-05, + "loss": 0.3267, + "step": 2661 + }, + { + "epoch": 0.16, + "learning_rate": 1.9116092905190893e-05, + "loss": 0.3806, + "step": 2662 + }, + { + "epoch": 0.16, + "learning_rate": 1.911528843565463e-05, + "loss": 0.3594, + "step": 2663 + }, + { + "epoch": 0.16, + "learning_rate": 1.9114483617142457e-05, + "loss": 0.3876, + "step": 2664 + }, + { + "epoch": 0.16, + "learning_rate": 1.91136784496852e-05, + "loss": 0.3683, + "step": 2665 + }, + { + "epoch": 0.16, + "learning_rate": 1.911287293331367e-05, + "loss": 0.3613, + "step": 2666 + }, + { + "epoch": 0.16, + "learning_rate": 1.9112067068058717e-05, + "loss": 0.385, + "step": 2667 + }, + { + "epoch": 0.16, + "learning_rate": 1.911126085395119e-05, + "loss": 0.346, + "step": 2668 + }, + { + "epoch": 0.16, + "learning_rate": 1.9110454291021954e-05, + "loss": 0.4018, + "step": 2669 + }, + { + "epoch": 0.16, + "learning_rate": 1.9109647379301885e-05, + "loss": 0.3845, + "step": 2670 + }, + { + "epoch": 0.16, + "learning_rate": 1.9108840118821882e-05, + "loss": 0.3588, + "step": 2671 + }, + { + "epoch": 0.16, + "learning_rate": 1.9108032509612844e-05, + "loss": 0.3836, + "step": 2672 + }, + { + "epoch": 0.16, + "learning_rate": 1.9107224551705692e-05, + "loss": 0.3867, + "step": 2673 + }, + { + "epoch": 0.16, + "learning_rate": 1.910641624513136e-05, + "loss": 0.353, + "step": 2674 + }, + { + "epoch": 0.16, + "learning_rate": 1.9105607589920797e-05, + "loss": 0.368, + "step": 2675 + }, + { + "epoch": 0.16, + "learning_rate": 1.9104798586104954e-05, + "loss": 0.382, + "step": 2676 + }, + { + "epoch": 0.16, + "learning_rate": 1.910398923371481e-05, + "loss": 0.3624, + "step": 2677 + }, + { + "epoch": 0.16, + "learning_rate": 1.9103179532781348e-05, + "loss": 0.3474, + "step": 2678 + }, + { + "epoch": 0.16, + "learning_rate": 1.9102369483335567e-05, + "loss": 0.3571, + "step": 2679 + }, + { + "epoch": 0.16, + "learning_rate": 1.910155908540848e-05, + "loss": 0.3517, + "step": 2680 + }, + { + "epoch": 0.16, + "learning_rate": 1.9100748339031113e-05, + "loss": 0.3892, + "step": 2681 + }, + { + "epoch": 0.16, + "learning_rate": 1.9099937244234508e-05, + "loss": 0.3371, + "step": 2682 + }, + { + "epoch": 0.16, + "learning_rate": 1.9099125801049714e-05, + "loss": 0.3387, + "step": 2683 + }, + { + "epoch": 0.16, + "learning_rate": 1.9098314009507794e-05, + "loss": 0.3691, + "step": 2684 + }, + { + "epoch": 0.16, + "learning_rate": 1.9097501869639834e-05, + "loss": 0.3778, + "step": 2685 + }, + { + "epoch": 0.16, + "learning_rate": 1.9096689381476924e-05, + "loss": 0.3421, + "step": 2686 + }, + { + "epoch": 0.16, + "learning_rate": 1.9095876545050166e-05, + "loss": 0.3647, + "step": 2687 + }, + { + "epoch": 0.16, + "learning_rate": 1.909506336039068e-05, + "loss": 0.3831, + "step": 2688 + }, + { + "epoch": 0.16, + "learning_rate": 1.909424982752961e-05, + "loss": 0.3474, + "step": 2689 + }, + { + "epoch": 0.16, + "learning_rate": 1.9093435946498083e-05, + "loss": 0.3326, + "step": 2690 + }, + { + "epoch": 0.16, + "learning_rate": 1.9092621717327275e-05, + "loss": 0.363, + "step": 2691 + }, + { + "epoch": 0.16, + "learning_rate": 1.9091807140048348e-05, + "loss": 0.3524, + "step": 2692 + }, + { + "epoch": 0.16, + "learning_rate": 1.909099221469249e-05, + "loss": 0.3878, + "step": 2693 + }, + { + "epoch": 0.16, + "learning_rate": 1.9090176941290902e-05, + "loss": 0.3809, + "step": 2694 + }, + { + "epoch": 0.16, + "learning_rate": 1.9089361319874797e-05, + "loss": 0.3292, + "step": 2695 + }, + { + "epoch": 0.16, + "learning_rate": 1.9088545350475398e-05, + "loss": 0.3527, + "step": 2696 + }, + { + "epoch": 0.16, + "learning_rate": 1.9087729033123948e-05, + "loss": 0.3524, + "step": 2697 + }, + { + "epoch": 0.16, + "learning_rate": 1.9086912367851692e-05, + "loss": 0.3591, + "step": 2698 + }, + { + "epoch": 0.16, + "learning_rate": 1.9086095354689907e-05, + "loss": 0.392, + "step": 2699 + }, + { + "epoch": 0.16, + "learning_rate": 1.9085277993669863e-05, + "loss": 0.358, + "step": 2700 + }, + { + "epoch": 0.16, + "learning_rate": 1.9084460284822856e-05, + "loss": 0.3926, + "step": 2701 + }, + { + "epoch": 0.16, + "learning_rate": 1.908364222818019e-05, + "loss": 0.3658, + "step": 2702 + }, + { + "epoch": 0.16, + "learning_rate": 1.9082823823773188e-05, + "loss": 0.3638, + "step": 2703 + }, + { + "epoch": 0.16, + "learning_rate": 1.9082005071633175e-05, + "loss": 0.3438, + "step": 2704 + }, + { + "epoch": 0.16, + "learning_rate": 1.9081185971791504e-05, + "loss": 0.3362, + "step": 2705 + }, + { + "epoch": 0.16, + "learning_rate": 1.908036652427953e-05, + "loss": 0.3613, + "step": 2706 + }, + { + "epoch": 0.16, + "learning_rate": 1.907954672912863e-05, + "loss": 0.3323, + "step": 2707 + }, + { + "epoch": 0.16, + "learning_rate": 1.907872658637018e-05, + "loss": 0.3535, + "step": 2708 + }, + { + "epoch": 0.16, + "learning_rate": 1.907790609603559e-05, + "loss": 0.382, + "step": 2709 + }, + { + "epoch": 0.16, + "learning_rate": 1.907708525815627e-05, + "loss": 0.3683, + "step": 2710 + }, + { + "epoch": 0.16, + "learning_rate": 1.9076264072763634e-05, + "loss": 0.3415, + "step": 2711 + }, + { + "epoch": 0.16, + "learning_rate": 1.9075442539889133e-05, + "loss": 0.3954, + "step": 2712 + }, + { + "epoch": 0.16, + "learning_rate": 1.9074620659564217e-05, + "loss": 0.3725, + "step": 2713 + }, + { + "epoch": 0.16, + "learning_rate": 1.907379843182035e-05, + "loss": 0.3371, + "step": 2714 + }, + { + "epoch": 0.16, + "learning_rate": 1.9072975856689008e-05, + "loss": 0.3491, + "step": 2715 + }, + { + "epoch": 0.16, + "learning_rate": 1.907215293420169e-05, + "loss": 0.3577, + "step": 2716 + }, + { + "epoch": 0.16, + "learning_rate": 1.9071329664389893e-05, + "loss": 0.3636, + "step": 2717 + }, + { + "epoch": 0.16, + "learning_rate": 1.9070506047285143e-05, + "loss": 0.3761, + "step": 2718 + }, + { + "epoch": 0.16, + "learning_rate": 1.9069682082918968e-05, + "loss": 0.3728, + "step": 2719 + }, + { + "epoch": 0.16, + "learning_rate": 1.9068857771322912e-05, + "loss": 0.3775, + "step": 2720 + }, + { + "epoch": 0.16, + "learning_rate": 1.9068033112528535e-05, + "loss": 0.377, + "step": 2721 + }, + { + "epoch": 0.16, + "learning_rate": 1.9067208106567412e-05, + "loss": 0.3454, + "step": 2722 + }, + { + "epoch": 0.16, + "learning_rate": 1.9066382753471123e-05, + "loss": 0.3689, + "step": 2723 + }, + { + "epoch": 0.16, + "learning_rate": 1.9065557053271272e-05, + "loss": 0.3761, + "step": 2724 + }, + { + "epoch": 0.16, + "learning_rate": 1.9064731005999465e-05, + "loss": 0.3362, + "step": 2725 + }, + { + "epoch": 0.16, + "learning_rate": 1.9063904611687326e-05, + "loss": 0.344, + "step": 2726 + }, + { + "epoch": 0.16, + "learning_rate": 1.9063077870366504e-05, + "loss": 0.3382, + "step": 2727 + }, + { + "epoch": 0.16, + "learning_rate": 1.9062250782068636e-05, + "loss": 0.3658, + "step": 2728 + }, + { + "epoch": 0.16, + "learning_rate": 1.9061423346825395e-05, + "loss": 0.3761, + "step": 2729 + }, + { + "epoch": 0.16, + "learning_rate": 1.906059556466846e-05, + "loss": 0.3781, + "step": 2730 + }, + { + "epoch": 0.16, + "learning_rate": 1.9059767435629522e-05, + "loss": 0.3463, + "step": 2731 + }, + { + "epoch": 0.17, + "learning_rate": 1.905893895974028e-05, + "loss": 0.3859, + "step": 2732 + }, + { + "epoch": 0.17, + "learning_rate": 1.905811013703246e-05, + "loss": 0.3862, + "step": 2733 + }, + { + "epoch": 0.17, + "learning_rate": 1.9057280967537783e-05, + "loss": 0.3421, + "step": 2734 + }, + { + "epoch": 0.17, + "learning_rate": 1.9056451451288006e-05, + "loss": 0.4088, + "step": 2735 + }, + { + "epoch": 0.17, + "learning_rate": 1.9055621588314877e-05, + "loss": 0.3227, + "step": 2736 + }, + { + "epoch": 0.17, + "learning_rate": 1.905479137865017e-05, + "loss": 0.3557, + "step": 2737 + }, + { + "epoch": 0.17, + "learning_rate": 1.9053960822325672e-05, + "loss": 0.394, + "step": 2738 + }, + { + "epoch": 0.17, + "learning_rate": 1.9053129919373178e-05, + "loss": 0.3538, + "step": 2739 + }, + { + "epoch": 0.17, + "learning_rate": 1.9052298669824495e-05, + "loss": 0.3834, + "step": 2740 + }, + { + "epoch": 0.17, + "learning_rate": 1.9051467073711455e-05, + "loss": 0.3518, + "step": 2741 + }, + { + "epoch": 0.17, + "learning_rate": 1.905063513106589e-05, + "loss": 0.351, + "step": 2742 + }, + { + "epoch": 0.17, + "learning_rate": 1.9049802841919657e-05, + "loss": 0.356, + "step": 2743 + }, + { + "epoch": 0.17, + "learning_rate": 1.904897020630461e-05, + "loss": 0.3495, + "step": 2744 + }, + { + "epoch": 0.17, + "learning_rate": 1.9048137224252633e-05, + "loss": 0.3915, + "step": 2745 + }, + { + "epoch": 0.17, + "learning_rate": 1.9047303895795615e-05, + "loss": 0.3557, + "step": 2746 + }, + { + "epoch": 0.17, + "learning_rate": 1.9046470220965457e-05, + "loss": 0.3792, + "step": 2747 + }, + { + "epoch": 0.17, + "learning_rate": 1.9045636199794084e-05, + "loss": 0.3552, + "step": 2748 + }, + { + "epoch": 0.17, + "learning_rate": 1.904480183231342e-05, + "loss": 0.3783, + "step": 2749 + }, + { + "epoch": 0.17, + "learning_rate": 1.9043967118555406e-05, + "loss": 0.3599, + "step": 2750 + }, + { + "epoch": 0.17, + "learning_rate": 1.9043132058552004e-05, + "loss": 0.3602, + "step": 2751 + }, + { + "epoch": 0.17, + "learning_rate": 1.904229665233518e-05, + "loss": 0.3557, + "step": 2752 + }, + { + "epoch": 0.17, + "learning_rate": 1.9041460899936922e-05, + "loss": 0.3672, + "step": 2753 + }, + { + "epoch": 0.17, + "learning_rate": 1.9040624801389223e-05, + "loss": 0.3457, + "step": 2754 + }, + { + "epoch": 0.17, + "learning_rate": 1.903978835672409e-05, + "loss": 0.3644, + "step": 2755 + }, + { + "epoch": 0.17, + "learning_rate": 1.9038951565973552e-05, + "loss": 0.3753, + "step": 2756 + }, + { + "epoch": 0.17, + "learning_rate": 1.9038114429169644e-05, + "loss": 0.3627, + "step": 2757 + }, + { + "epoch": 0.17, + "learning_rate": 1.9037276946344415e-05, + "loss": 0.3675, + "step": 2758 + }, + { + "epoch": 0.17, + "learning_rate": 1.9036439117529924e-05, + "loss": 0.356, + "step": 2759 + }, + { + "epoch": 0.17, + "learning_rate": 1.9035600942758256e-05, + "loss": 0.4071, + "step": 2760 + }, + { + "epoch": 0.17, + "learning_rate": 1.903476242206149e-05, + "loss": 0.3898, + "step": 2761 + }, + { + "epoch": 0.17, + "learning_rate": 1.9033923555471736e-05, + "loss": 0.377, + "step": 2762 + }, + { + "epoch": 0.17, + "learning_rate": 1.90330843430211e-05, + "loss": 0.3497, + "step": 2763 + }, + { + "epoch": 0.17, + "learning_rate": 1.903224478474172e-05, + "loss": 0.3876, + "step": 2764 + }, + { + "epoch": 0.17, + "learning_rate": 1.9031404880665742e-05, + "loss": 0.3571, + "step": 2765 + }, + { + "epoch": 0.17, + "learning_rate": 1.903056463082531e-05, + "loss": 0.3574, + "step": 2766 + }, + { + "epoch": 0.17, + "learning_rate": 1.90297240352526e-05, + "loss": 0.3616, + "step": 2767 + }, + { + "epoch": 0.17, + "learning_rate": 1.902888309397979e-05, + "loss": 0.3521, + "step": 2768 + }, + { + "epoch": 0.17, + "learning_rate": 1.9028041807039077e-05, + "loss": 0.3524, + "step": 2769 + }, + { + "epoch": 0.17, + "learning_rate": 1.902720017446267e-05, + "loss": 0.3677, + "step": 2770 + }, + { + "epoch": 0.17, + "learning_rate": 1.9026358196282792e-05, + "loss": 0.3647, + "step": 2771 + }, + { + "epoch": 0.17, + "learning_rate": 1.9025515872531675e-05, + "loss": 0.3569, + "step": 2772 + }, + { + "epoch": 0.17, + "learning_rate": 1.902467320324157e-05, + "loss": 0.3451, + "step": 2773 + }, + { + "epoch": 0.17, + "learning_rate": 1.9023830188444733e-05, + "loss": 0.3622, + "step": 2774 + }, + { + "epoch": 0.17, + "learning_rate": 1.9022986828173443e-05, + "loss": 0.3903, + "step": 2775 + }, + { + "epoch": 0.17, + "learning_rate": 1.9022143122459988e-05, + "loss": 0.3636, + "step": 2776 + }, + { + "epoch": 0.17, + "learning_rate": 1.9021299071336665e-05, + "loss": 0.3552, + "step": 2777 + }, + { + "epoch": 0.17, + "learning_rate": 1.9020454674835794e-05, + "loss": 0.3535, + "step": 2778 + }, + { + "epoch": 0.17, + "learning_rate": 1.9019609932989696e-05, + "loss": 0.3482, + "step": 2779 + }, + { + "epoch": 0.17, + "learning_rate": 1.901876484583072e-05, + "loss": 0.3744, + "step": 2780 + }, + { + "epoch": 0.17, + "learning_rate": 1.9017919413391213e-05, + "loss": 0.3588, + "step": 2781 + }, + { + "epoch": 0.17, + "learning_rate": 1.9017073635703544e-05, + "loss": 0.4004, + "step": 2782 + }, + { + "epoch": 0.17, + "learning_rate": 1.9016227512800092e-05, + "loss": 0.3499, + "step": 2783 + }, + { + "epoch": 0.17, + "learning_rate": 1.9015381044713252e-05, + "loss": 0.3842, + "step": 2784 + }, + { + "epoch": 0.17, + "learning_rate": 1.901453423147543e-05, + "loss": 0.3722, + "step": 2785 + }, + { + "epoch": 0.17, + "learning_rate": 1.901368707311905e-05, + "loss": 0.3809, + "step": 2786 + }, + { + "epoch": 0.17, + "learning_rate": 1.901283956967654e-05, + "loss": 0.3418, + "step": 2787 + }, + { + "epoch": 0.17, + "learning_rate": 1.901199172118035e-05, + "loss": 0.3532, + "step": 2788 + }, + { + "epoch": 0.17, + "learning_rate": 1.9011143527662938e-05, + "loss": 0.3675, + "step": 2789 + }, + { + "epoch": 0.17, + "learning_rate": 1.901029498915677e-05, + "loss": 0.3499, + "step": 2790 + }, + { + "epoch": 0.17, + "learning_rate": 1.9009446105694347e-05, + "loss": 0.373, + "step": 2791 + }, + { + "epoch": 0.17, + "learning_rate": 1.9008596877308157e-05, + "loss": 0.3499, + "step": 2792 + }, + { + "epoch": 0.17, + "learning_rate": 1.9007747304030716e-05, + "loss": 0.3622, + "step": 2793 + }, + { + "epoch": 0.17, + "learning_rate": 1.900689738589455e-05, + "loss": 0.3644, + "step": 2794 + }, + { + "epoch": 0.17, + "learning_rate": 1.900604712293219e-05, + "loss": 0.3549, + "step": 2795 + }, + { + "epoch": 0.17, + "learning_rate": 1.9005196515176204e-05, + "loss": 0.3449, + "step": 2796 + }, + { + "epoch": 0.17, + "learning_rate": 1.9004345562659145e-05, + "loss": 0.3786, + "step": 2797 + }, + { + "epoch": 0.17, + "learning_rate": 1.9003494265413595e-05, + "loss": 0.3929, + "step": 2798 + }, + { + "epoch": 0.17, + "learning_rate": 1.900264262347214e-05, + "loss": 0.3772, + "step": 2799 + }, + { + "epoch": 0.17, + "learning_rate": 1.9001790636867398e-05, + "loss": 0.3491, + "step": 2800 + }, + { + "epoch": 0.17, + "learning_rate": 1.9000938305631975e-05, + "loss": 0.3597, + "step": 2801 + }, + { + "epoch": 0.17, + "learning_rate": 1.9000085629798507e-05, + "loss": 0.3923, + "step": 2802 + }, + { + "epoch": 0.17, + "learning_rate": 1.899923260939964e-05, + "loss": 0.3638, + "step": 2803 + }, + { + "epoch": 0.17, + "learning_rate": 1.8998379244468027e-05, + "loss": 0.3557, + "step": 2804 + }, + { + "epoch": 0.17, + "learning_rate": 1.899752553503634e-05, + "loss": 0.409, + "step": 2805 + }, + { + "epoch": 0.17, + "learning_rate": 1.8996671481137267e-05, + "loss": 0.3917, + "step": 2806 + }, + { + "epoch": 0.17, + "learning_rate": 1.89958170828035e-05, + "loss": 0.3638, + "step": 2807 + }, + { + "epoch": 0.17, + "learning_rate": 1.8994962340067756e-05, + "loss": 0.3404, + "step": 2808 + }, + { + "epoch": 0.17, + "learning_rate": 1.899410725296275e-05, + "loss": 0.3557, + "step": 2809 + }, + { + "epoch": 0.17, + "learning_rate": 1.8993251821521225e-05, + "loss": 0.365, + "step": 2810 + }, + { + "epoch": 0.17, + "learning_rate": 1.8992396045775927e-05, + "loss": 0.3775, + "step": 2811 + }, + { + "epoch": 0.17, + "learning_rate": 1.8991539925759624e-05, + "loss": 0.3532, + "step": 2812 + }, + { + "epoch": 0.17, + "learning_rate": 1.8990683461505086e-05, + "loss": 0.3856, + "step": 2813 + }, + { + "epoch": 0.17, + "learning_rate": 1.898982665304511e-05, + "loss": 0.389, + "step": 2814 + }, + { + "epoch": 0.17, + "learning_rate": 1.8988969500412492e-05, + "loss": 0.3719, + "step": 2815 + }, + { + "epoch": 0.17, + "learning_rate": 1.8988112003640048e-05, + "loss": 0.3836, + "step": 2816 + }, + { + "epoch": 0.17, + "learning_rate": 1.8987254162760614e-05, + "loss": 0.3705, + "step": 2817 + }, + { + "epoch": 0.17, + "learning_rate": 1.8986395977807026e-05, + "loss": 0.3809, + "step": 2818 + }, + { + "epoch": 0.17, + "learning_rate": 1.8985537448812138e-05, + "loss": 0.3758, + "step": 2819 + }, + { + "epoch": 0.17, + "learning_rate": 1.8984678575808825e-05, + "loss": 0.3705, + "step": 2820 + }, + { + "epoch": 0.17, + "learning_rate": 1.898381935882996e-05, + "loss": 0.38, + "step": 2821 + }, + { + "epoch": 0.17, + "learning_rate": 1.8982959797908447e-05, + "loss": 0.3468, + "step": 2822 + }, + { + "epoch": 0.17, + "learning_rate": 1.898209989307719e-05, + "loss": 0.3398, + "step": 2823 + }, + { + "epoch": 0.17, + "learning_rate": 1.898123964436911e-05, + "loss": 0.3627, + "step": 2824 + }, + { + "epoch": 0.17, + "learning_rate": 1.898037905181714e-05, + "loss": 0.3825, + "step": 2825 + }, + { + "epoch": 0.17, + "learning_rate": 1.897951811545423e-05, + "loss": 0.3502, + "step": 2826 + }, + { + "epoch": 0.17, + "learning_rate": 1.897865683531334e-05, + "loss": 0.3414, + "step": 2827 + }, + { + "epoch": 0.17, + "learning_rate": 1.897779521142744e-05, + "loss": 0.3379, + "step": 2828 + }, + { + "epoch": 0.17, + "learning_rate": 1.897693324382952e-05, + "loss": 0.3931, + "step": 2829 + }, + { + "epoch": 0.17, + "learning_rate": 1.8976070932552583e-05, + "loss": 0.3527, + "step": 2830 + }, + { + "epoch": 0.17, + "learning_rate": 1.897520827762964e-05, + "loss": 0.3544, + "step": 2831 + }, + { + "epoch": 0.17, + "learning_rate": 1.897434527909372e-05, + "loss": 0.3318, + "step": 2832 + }, + { + "epoch": 0.17, + "learning_rate": 1.8973481936977857e-05, + "loss": 0.3549, + "step": 2833 + }, + { + "epoch": 0.17, + "learning_rate": 1.8972618251315104e-05, + "loss": 0.3479, + "step": 2834 + }, + { + "epoch": 0.17, + "learning_rate": 1.8971754222138532e-05, + "loss": 0.3828, + "step": 2835 + }, + { + "epoch": 0.17, + "learning_rate": 1.8970889849481216e-05, + "loss": 0.346, + "step": 2836 + }, + { + "epoch": 0.17, + "learning_rate": 1.8970025133376252e-05, + "loss": 0.3669, + "step": 2837 + }, + { + "epoch": 0.17, + "learning_rate": 1.8969160073856745e-05, + "loss": 0.3901, + "step": 2838 + }, + { + "epoch": 0.17, + "learning_rate": 1.896829467095581e-05, + "loss": 0.3664, + "step": 2839 + }, + { + "epoch": 0.17, + "learning_rate": 1.896742892470658e-05, + "loss": 0.3711, + "step": 2840 + }, + { + "epoch": 0.17, + "learning_rate": 1.89665628351422e-05, + "loss": 0.3717, + "step": 2841 + }, + { + "epoch": 0.17, + "learning_rate": 1.896569640229583e-05, + "loss": 0.3694, + "step": 2842 + }, + { + "epoch": 0.17, + "learning_rate": 1.8964829626200637e-05, + "loss": 0.3666, + "step": 2843 + }, + { + "epoch": 0.17, + "learning_rate": 1.896396250688981e-05, + "loss": 0.3103, + "step": 2844 + }, + { + "epoch": 0.17, + "learning_rate": 1.8963095044396542e-05, + "loss": 0.3594, + "step": 2845 + }, + { + "epoch": 0.17, + "learning_rate": 1.8962227238754048e-05, + "loss": 0.3535, + "step": 2846 + }, + { + "epoch": 0.17, + "learning_rate": 1.896135908999555e-05, + "loss": 0.3574, + "step": 2847 + }, + { + "epoch": 0.17, + "learning_rate": 1.896049059815428e-05, + "loss": 0.3482, + "step": 2848 + }, + { + "epoch": 0.17, + "learning_rate": 1.8959621763263494e-05, + "loss": 0.3664, + "step": 2849 + }, + { + "epoch": 0.17, + "learning_rate": 1.895875258535645e-05, + "loss": 0.3624, + "step": 2850 + }, + { + "epoch": 0.17, + "learning_rate": 1.8957883064466435e-05, + "loss": 0.3622, + "step": 2851 + }, + { + "epoch": 0.17, + "learning_rate": 1.8957013200626723e-05, + "loss": 0.3345, + "step": 2852 + }, + { + "epoch": 0.17, + "learning_rate": 1.8956142993870626e-05, + "loss": 0.3809, + "step": 2853 + }, + { + "epoch": 0.17, + "learning_rate": 1.895527244423146e-05, + "loss": 0.3078, + "step": 2854 + }, + { + "epoch": 0.17, + "learning_rate": 1.8954401551742546e-05, + "loss": 0.3619, + "step": 2855 + }, + { + "epoch": 0.17, + "learning_rate": 1.8953530316437238e-05, + "loss": 0.3304, + "step": 2856 + }, + { + "epoch": 0.17, + "learning_rate": 1.895265873834888e-05, + "loss": 0.3959, + "step": 2857 + }, + { + "epoch": 0.17, + "learning_rate": 1.8951786817510846e-05, + "loss": 0.3633, + "step": 2858 + }, + { + "epoch": 0.17, + "learning_rate": 1.8950914553956515e-05, + "loss": 0.3641, + "step": 2859 + }, + { + "epoch": 0.17, + "learning_rate": 1.895004194771928e-05, + "loss": 0.3242, + "step": 2860 + }, + { + "epoch": 0.17, + "learning_rate": 1.894916899883255e-05, + "loss": 0.3344, + "step": 2861 + }, + { + "epoch": 0.17, + "learning_rate": 1.894829570732975e-05, + "loss": 0.3189, + "step": 2862 + }, + { + "epoch": 0.17, + "learning_rate": 1.8947422073244305e-05, + "loss": 0.3705, + "step": 2863 + }, + { + "epoch": 0.17, + "learning_rate": 1.8946548096609668e-05, + "loss": 0.3527, + "step": 2864 + }, + { + "epoch": 0.17, + "learning_rate": 1.8945673777459297e-05, + "loss": 0.3744, + "step": 2865 + }, + { + "epoch": 0.17, + "learning_rate": 1.8944799115826666e-05, + "loss": 0.3616, + "step": 2866 + }, + { + "epoch": 0.17, + "learning_rate": 1.8943924111745258e-05, + "loss": 0.358, + "step": 2867 + }, + { + "epoch": 0.17, + "learning_rate": 1.8943048765248578e-05, + "loss": 0.3647, + "step": 2868 + }, + { + "epoch": 0.17, + "learning_rate": 1.8942173076370133e-05, + "loss": 0.334, + "step": 2869 + }, + { + "epoch": 0.17, + "learning_rate": 1.8941297045143447e-05, + "loss": 0.365, + "step": 2870 + }, + { + "epoch": 0.17, + "learning_rate": 1.8940420671602067e-05, + "loss": 0.3463, + "step": 2871 + }, + { + "epoch": 0.17, + "learning_rate": 1.8939543955779535e-05, + "loss": 0.3691, + "step": 2872 + }, + { + "epoch": 0.17, + "learning_rate": 1.8938666897709427e-05, + "loss": 0.3711, + "step": 2873 + }, + { + "epoch": 0.17, + "learning_rate": 1.893778949742531e-05, + "loss": 0.3613, + "step": 2874 + }, + { + "epoch": 0.17, + "learning_rate": 1.893691175496078e-05, + "loss": 0.3683, + "step": 2875 + }, + { + "epoch": 0.17, + "learning_rate": 1.893603367034944e-05, + "loss": 0.3828, + "step": 2876 + }, + { + "epoch": 0.17, + "learning_rate": 1.893515524362491e-05, + "loss": 0.37, + "step": 2877 + }, + { + "epoch": 0.17, + "learning_rate": 1.8934276474820816e-05, + "loss": 0.339, + "step": 2878 + }, + { + "epoch": 0.17, + "learning_rate": 1.8933397363970802e-05, + "loss": 0.3683, + "step": 2879 + }, + { + "epoch": 0.17, + "learning_rate": 1.893251791110853e-05, + "loss": 0.3527, + "step": 2880 + }, + { + "epoch": 0.17, + "learning_rate": 1.893163811626766e-05, + "loss": 0.3491, + "step": 2881 + }, + { + "epoch": 0.17, + "learning_rate": 1.893075797948188e-05, + "loss": 0.3421, + "step": 2882 + }, + { + "epoch": 0.17, + "learning_rate": 1.892987750078489e-05, + "loss": 0.3348, + "step": 2883 + }, + { + "epoch": 0.17, + "learning_rate": 1.8928996680210395e-05, + "loss": 0.3934, + "step": 2884 + }, + { + "epoch": 0.17, + "learning_rate": 1.8928115517792112e-05, + "loss": 0.3446, + "step": 2885 + }, + { + "epoch": 0.17, + "learning_rate": 1.8927234013563783e-05, + "loss": 0.3446, + "step": 2886 + }, + { + "epoch": 0.17, + "learning_rate": 1.8926352167559154e-05, + "loss": 0.3435, + "step": 2887 + }, + { + "epoch": 0.17, + "learning_rate": 1.8925469979811984e-05, + "loss": 0.3644, + "step": 2888 + }, + { + "epoch": 0.17, + "learning_rate": 1.892458745035605e-05, + "loss": 0.3549, + "step": 2889 + }, + { + "epoch": 0.17, + "learning_rate": 1.892370457922514e-05, + "loss": 0.3527, + "step": 2890 + }, + { + "epoch": 0.17, + "learning_rate": 1.8922821366453052e-05, + "loss": 0.3435, + "step": 2891 + }, + { + "epoch": 0.17, + "learning_rate": 1.89219378120736e-05, + "loss": 0.3683, + "step": 2892 + }, + { + "epoch": 0.17, + "learning_rate": 1.892105391612061e-05, + "loss": 0.3343, + "step": 2893 + }, + { + "epoch": 0.17, + "learning_rate": 1.8920169678627928e-05, + "loss": 0.3468, + "step": 2894 + }, + { + "epoch": 0.17, + "learning_rate": 1.8919285099629397e-05, + "loss": 0.3669, + "step": 2895 + }, + { + "epoch": 0.17, + "learning_rate": 1.8918400179158888e-05, + "loss": 0.3652, + "step": 2896 + }, + { + "epoch": 0.18, + "learning_rate": 1.8917514917250276e-05, + "loss": 0.346, + "step": 2897 + }, + { + "epoch": 0.18, + "learning_rate": 1.891662931393746e-05, + "loss": 0.361, + "step": 2898 + }, + { + "epoch": 0.18, + "learning_rate": 1.891574336925434e-05, + "loss": 0.3638, + "step": 2899 + }, + { + "epoch": 0.18, + "learning_rate": 1.8914857083234838e-05, + "loss": 0.3451, + "step": 2900 + }, + { + "epoch": 0.18, + "learning_rate": 1.891397045591288e-05, + "loss": 0.3451, + "step": 2901 + }, + { + "epoch": 0.18, + "learning_rate": 1.891308348732241e-05, + "loss": 0.3438, + "step": 2902 + }, + { + "epoch": 0.18, + "learning_rate": 1.891219617749739e-05, + "loss": 0.3265, + "step": 2903 + }, + { + "epoch": 0.18, + "learning_rate": 1.8911308526471792e-05, + "loss": 0.3686, + "step": 2904 + }, + { + "epoch": 0.18, + "learning_rate": 1.8910420534279593e-05, + "loss": 0.3722, + "step": 2905 + }, + { + "epoch": 0.18, + "learning_rate": 1.8909532200954792e-05, + "loss": 0.389, + "step": 2906 + }, + { + "epoch": 0.18, + "learning_rate": 1.89086435265314e-05, + "loss": 0.3772, + "step": 2907 + }, + { + "epoch": 0.18, + "learning_rate": 1.8907754511043437e-05, + "loss": 0.3555, + "step": 2908 + }, + { + "epoch": 0.18, + "learning_rate": 1.8906865154524943e-05, + "loss": 0.3828, + "step": 2909 + }, + { + "epoch": 0.18, + "learning_rate": 1.890597545700996e-05, + "loss": 0.3842, + "step": 2910 + }, + { + "epoch": 0.18, + "learning_rate": 1.8905085418532557e-05, + "loss": 0.3878, + "step": 2911 + }, + { + "epoch": 0.18, + "learning_rate": 1.8904195039126804e-05, + "loss": 0.3809, + "step": 2912 + }, + { + "epoch": 0.18, + "learning_rate": 1.890330431882679e-05, + "loss": 0.3242, + "step": 2913 + }, + { + "epoch": 0.18, + "learning_rate": 1.8902413257666616e-05, + "loss": 0.3636, + "step": 2914 + }, + { + "epoch": 0.18, + "learning_rate": 1.8901521855680397e-05, + "loss": 0.3435, + "step": 2915 + }, + { + "epoch": 0.18, + "learning_rate": 1.890063011290226e-05, + "loss": 0.3351, + "step": 2916 + }, + { + "epoch": 0.18, + "learning_rate": 1.8899738029366346e-05, + "loss": 0.3203, + "step": 2917 + }, + { + "epoch": 0.18, + "learning_rate": 1.8898845605106804e-05, + "loss": 0.3585, + "step": 2918 + }, + { + "epoch": 0.18, + "learning_rate": 1.8897952840157804e-05, + "loss": 0.3736, + "step": 2919 + }, + { + "epoch": 0.18, + "learning_rate": 1.889705973455352e-05, + "loss": 0.329, + "step": 2920 + }, + { + "epoch": 0.18, + "learning_rate": 1.8896166288328156e-05, + "loss": 0.3652, + "step": 2921 + }, + { + "epoch": 0.18, + "learning_rate": 1.8895272501515904e-05, + "loss": 0.3496, + "step": 2922 + }, + { + "epoch": 0.18, + "learning_rate": 1.889437837415099e-05, + "loss": 0.3521, + "step": 2923 + }, + { + "epoch": 0.18, + "learning_rate": 1.8893483906267643e-05, + "loss": 0.368, + "step": 2924 + }, + { + "epoch": 0.18, + "learning_rate": 1.8892589097900108e-05, + "loss": 0.3588, + "step": 2925 + }, + { + "epoch": 0.18, + "learning_rate": 1.889169394908264e-05, + "loss": 0.3744, + "step": 2926 + }, + { + "epoch": 0.18, + "learning_rate": 1.889079845984951e-05, + "loss": 0.3315, + "step": 2927 + }, + { + "epoch": 0.18, + "learning_rate": 1.8889902630235012e-05, + "loss": 0.361, + "step": 2928 + }, + { + "epoch": 0.18, + "learning_rate": 1.8889006460273426e-05, + "loss": 0.3666, + "step": 2929 + }, + { + "epoch": 0.18, + "learning_rate": 1.8888109949999073e-05, + "loss": 0.3797, + "step": 2930 + }, + { + "epoch": 0.18, + "learning_rate": 1.888721309944627e-05, + "loss": 0.3574, + "step": 2931 + }, + { + "epoch": 0.18, + "learning_rate": 1.8886315908649356e-05, + "loss": 0.3488, + "step": 2932 + }, + { + "epoch": 0.18, + "learning_rate": 1.8885418377642673e-05, + "loss": 0.3619, + "step": 2933 + }, + { + "epoch": 0.18, + "learning_rate": 1.8884520506460596e-05, + "loss": 0.3379, + "step": 2934 + }, + { + "epoch": 0.18, + "learning_rate": 1.8883622295137487e-05, + "loss": 0.3432, + "step": 2935 + }, + { + "epoch": 0.18, + "learning_rate": 1.888272374370774e-05, + "loss": 0.3739, + "step": 2936 + }, + { + "epoch": 0.18, + "learning_rate": 1.8881824852205752e-05, + "loss": 0.3365, + "step": 2937 + }, + { + "epoch": 0.18, + "learning_rate": 1.8880925620665943e-05, + "loss": 0.3471, + "step": 2938 + }, + { + "epoch": 0.18, + "learning_rate": 1.888002604912273e-05, + "loss": 0.3672, + "step": 2939 + }, + { + "epoch": 0.18, + "learning_rate": 1.8879126137610563e-05, + "loss": 0.358, + "step": 2940 + }, + { + "epoch": 0.18, + "learning_rate": 1.8878225886163884e-05, + "loss": 0.368, + "step": 2941 + }, + { + "epoch": 0.18, + "learning_rate": 1.8877325294817174e-05, + "loss": 0.3708, + "step": 2942 + }, + { + "epoch": 0.18, + "learning_rate": 1.8876424363604898e-05, + "loss": 0.3708, + "step": 2943 + }, + { + "epoch": 0.18, + "learning_rate": 1.8875523092561553e-05, + "loss": 0.3811, + "step": 2944 + }, + { + "epoch": 0.18, + "learning_rate": 1.8874621481721645e-05, + "loss": 0.365, + "step": 2945 + }, + { + "epoch": 0.18, + "learning_rate": 1.887371953111969e-05, + "loss": 0.38, + "step": 2946 + }, + { + "epoch": 0.18, + "learning_rate": 1.887281724079022e-05, + "loss": 0.3686, + "step": 2947 + }, + { + "epoch": 0.18, + "learning_rate": 1.887191461076778e-05, + "loss": 0.3809, + "step": 2948 + }, + { + "epoch": 0.18, + "learning_rate": 1.887101164108692e-05, + "loss": 0.3549, + "step": 2949 + }, + { + "epoch": 0.18, + "learning_rate": 1.887010833178222e-05, + "loss": 0.3242, + "step": 2950 + }, + { + "epoch": 0.18, + "learning_rate": 1.8869204682888255e-05, + "loss": 0.353, + "step": 2951 + }, + { + "epoch": 0.18, + "learning_rate": 1.8868300694439627e-05, + "loss": 0.3449, + "step": 2952 + }, + { + "epoch": 0.18, + "learning_rate": 1.8867396366470936e-05, + "loss": 0.3711, + "step": 2953 + }, + { + "epoch": 0.18, + "learning_rate": 1.8866491699016818e-05, + "loss": 0.3535, + "step": 2954 + }, + { + "epoch": 0.18, + "learning_rate": 1.8865586692111898e-05, + "loss": 0.3465, + "step": 2955 + }, + { + "epoch": 0.18, + "learning_rate": 1.886468134579082e-05, + "loss": 0.3867, + "step": 2956 + }, + { + "epoch": 0.18, + "learning_rate": 1.8863775660088258e-05, + "loss": 0.3499, + "step": 2957 + }, + { + "epoch": 0.18, + "learning_rate": 1.8862869635038873e-05, + "loss": 0.3267, + "step": 2958 + }, + { + "epoch": 0.18, + "learning_rate": 1.8861963270677356e-05, + "loss": 0.3502, + "step": 2959 + }, + { + "epoch": 0.18, + "learning_rate": 1.8861056567038415e-05, + "loss": 0.346, + "step": 2960 + }, + { + "epoch": 0.18, + "learning_rate": 1.886014952415675e-05, + "loss": 0.3583, + "step": 2961 + }, + { + "epoch": 0.18, + "learning_rate": 1.8859242142067096e-05, + "loss": 0.3585, + "step": 2962 + }, + { + "epoch": 0.18, + "learning_rate": 1.8858334420804188e-05, + "loss": 0.3739, + "step": 2963 + }, + { + "epoch": 0.18, + "learning_rate": 1.885742636040278e-05, + "loss": 0.341, + "step": 2964 + }, + { + "epoch": 0.18, + "learning_rate": 1.8856517960897633e-05, + "loss": 0.3574, + "step": 2965 + }, + { + "epoch": 0.18, + "learning_rate": 1.8855609222323527e-05, + "loss": 0.3563, + "step": 2966 + }, + { + "epoch": 0.18, + "learning_rate": 1.885470014471525e-05, + "loss": 0.2963, + "step": 2967 + }, + { + "epoch": 0.18, + "learning_rate": 1.8853790728107616e-05, + "loss": 0.3597, + "step": 2968 + }, + { + "epoch": 0.18, + "learning_rate": 1.8852880972535433e-05, + "loss": 0.3669, + "step": 2969 + }, + { + "epoch": 0.18, + "learning_rate": 1.8851970878033526e-05, + "loss": 0.3327, + "step": 2970 + }, + { + "epoch": 0.18, + "learning_rate": 1.885106044463675e-05, + "loss": 0.3644, + "step": 2971 + }, + { + "epoch": 0.18, + "learning_rate": 1.8850149672379953e-05, + "loss": 0.3588, + "step": 2972 + }, + { + "epoch": 0.18, + "learning_rate": 1.8849238561298003e-05, + "loss": 0.3331, + "step": 2973 + }, + { + "epoch": 0.18, + "learning_rate": 1.8848327111425783e-05, + "loss": 0.3733, + "step": 2974 + }, + { + "epoch": 0.18, + "learning_rate": 1.884741532279819e-05, + "loss": 0.3404, + "step": 2975 + }, + { + "epoch": 0.18, + "learning_rate": 1.884650319545013e-05, + "loss": 0.3471, + "step": 2976 + }, + { + "epoch": 0.18, + "learning_rate": 1.8845590729416524e-05, + "loss": 0.3714, + "step": 2977 + }, + { + "epoch": 0.18, + "learning_rate": 1.8844677924732302e-05, + "loss": 0.351, + "step": 2978 + }, + { + "epoch": 0.18, + "learning_rate": 1.8843764781432413e-05, + "loss": 0.3842, + "step": 2979 + }, + { + "epoch": 0.18, + "learning_rate": 1.884285129955182e-05, + "loss": 0.3451, + "step": 2980 + }, + { + "epoch": 0.18, + "learning_rate": 1.884193747912549e-05, + "loss": 0.3569, + "step": 2981 + }, + { + "epoch": 0.18, + "learning_rate": 1.884102332018841e-05, + "loss": 0.365, + "step": 2982 + }, + { + "epoch": 0.18, + "learning_rate": 1.884010882277558e-05, + "loss": 0.356, + "step": 2983 + }, + { + "epoch": 0.18, + "learning_rate": 1.8839193986922005e-05, + "loss": 0.3742, + "step": 2984 + }, + { + "epoch": 0.18, + "learning_rate": 1.8838278812662717e-05, + "loss": 0.3753, + "step": 2985 + }, + { + "epoch": 0.18, + "learning_rate": 1.8837363300032754e-05, + "loss": 0.3661, + "step": 2986 + }, + { + "epoch": 0.18, + "learning_rate": 1.8836447449067158e-05, + "loss": 0.3641, + "step": 2987 + }, + { + "epoch": 0.18, + "learning_rate": 1.8835531259800994e-05, + "loss": 0.3856, + "step": 2988 + }, + { + "epoch": 0.18, + "learning_rate": 1.8834614732269343e-05, + "loss": 0.3633, + "step": 2989 + }, + { + "epoch": 0.18, + "learning_rate": 1.8833697866507293e-05, + "loss": 0.4068, + "step": 2990 + }, + { + "epoch": 0.18, + "learning_rate": 1.8832780662549943e-05, + "loss": 0.3549, + "step": 2991 + }, + { + "epoch": 0.18, + "learning_rate": 1.883186312043241e-05, + "loss": 0.4032, + "step": 2992 + }, + { + "epoch": 0.18, + "learning_rate": 1.883094524018982e-05, + "loss": 0.3658, + "step": 2993 + }, + { + "epoch": 0.18, + "learning_rate": 1.883002702185731e-05, + "loss": 0.3602, + "step": 2994 + }, + { + "epoch": 0.18, + "learning_rate": 1.8829108465470044e-05, + "loss": 0.4051, + "step": 2995 + }, + { + "epoch": 0.18, + "learning_rate": 1.882818957106318e-05, + "loss": 0.3401, + "step": 2996 + }, + { + "epoch": 0.18, + "learning_rate": 1.88272703386719e-05, + "loss": 0.3842, + "step": 2997 + }, + { + "epoch": 0.18, + "learning_rate": 1.88263507683314e-05, + "loss": 0.327, + "step": 2998 + }, + { + "epoch": 0.18, + "learning_rate": 1.8825430860076882e-05, + "loss": 0.3725, + "step": 2999 + }, + { + "epoch": 0.18, + "learning_rate": 1.882451061394356e-05, + "loss": 0.3658, + "step": 3000 + }, + { + "epoch": 0.18, + "learning_rate": 1.8823590029966677e-05, + "loss": 0.3619, + "step": 3001 + }, + { + "epoch": 0.18, + "learning_rate": 1.882266910818147e-05, + "loss": 0.3739, + "step": 3002 + }, + { + "epoch": 0.18, + "learning_rate": 1.8821747848623192e-05, + "loss": 0.356, + "step": 3003 + }, + { + "epoch": 0.18, + "learning_rate": 1.882082625132712e-05, + "loss": 0.3507, + "step": 3004 + }, + { + "epoch": 0.18, + "learning_rate": 1.8819904316328534e-05, + "loss": 0.3454, + "step": 3005 + }, + { + "epoch": 0.18, + "learning_rate": 1.881898204366273e-05, + "loss": 0.3516, + "step": 3006 + }, + { + "epoch": 0.18, + "learning_rate": 1.881805943336502e-05, + "loss": 0.3809, + "step": 3007 + }, + { + "epoch": 0.18, + "learning_rate": 1.881713648547072e-05, + "loss": 0.3624, + "step": 3008 + }, + { + "epoch": 0.18, + "learning_rate": 1.8816213200015175e-05, + "loss": 0.3566, + "step": 3009 + }, + { + "epoch": 0.18, + "learning_rate": 1.881528957703372e-05, + "loss": 0.3747, + "step": 3010 + }, + { + "epoch": 0.18, + "learning_rate": 1.881436561656172e-05, + "loss": 0.356, + "step": 3011 + }, + { + "epoch": 0.18, + "learning_rate": 1.8813441318634558e-05, + "loss": 0.3809, + "step": 3012 + }, + { + "epoch": 0.18, + "learning_rate": 1.8812516683287608e-05, + "loss": 0.3627, + "step": 3013 + }, + { + "epoch": 0.18, + "learning_rate": 1.8811591710556275e-05, + "loss": 0.3585, + "step": 3014 + }, + { + "epoch": 0.18, + "learning_rate": 1.881066640047597e-05, + "loss": 0.3644, + "step": 3015 + }, + { + "epoch": 0.18, + "learning_rate": 1.8809740753082116e-05, + "loss": 0.38, + "step": 3016 + }, + { + "epoch": 0.18, + "learning_rate": 1.8808814768410157e-05, + "loss": 0.3569, + "step": 3017 + }, + { + "epoch": 0.18, + "learning_rate": 1.880788844649554e-05, + "loss": 0.3541, + "step": 3018 + }, + { + "epoch": 0.18, + "learning_rate": 1.8806961787373728e-05, + "loss": 0.3742, + "step": 3019 + }, + { + "epoch": 0.18, + "learning_rate": 1.88060347910802e-05, + "loss": 0.3792, + "step": 3020 + }, + { + "epoch": 0.18, + "learning_rate": 1.8805107457650444e-05, + "loss": 0.3591, + "step": 3021 + }, + { + "epoch": 0.18, + "learning_rate": 1.8804179787119964e-05, + "loss": 0.3689, + "step": 3022 + }, + { + "epoch": 0.18, + "learning_rate": 1.880325177952428e-05, + "loss": 0.3682, + "step": 3023 + }, + { + "epoch": 0.18, + "learning_rate": 1.880232343489891e-05, + "loss": 0.3446, + "step": 3024 + }, + { + "epoch": 0.18, + "learning_rate": 1.8801394753279407e-05, + "loss": 0.2949, + "step": 3025 + }, + { + "epoch": 0.18, + "learning_rate": 1.8800465734701313e-05, + "loss": 0.368, + "step": 3026 + }, + { + "epoch": 0.18, + "learning_rate": 1.8799536379200205e-05, + "loss": 0.3619, + "step": 3027 + }, + { + "epoch": 0.18, + "learning_rate": 1.879860668681166e-05, + "loss": 0.3717, + "step": 3028 + }, + { + "epoch": 0.18, + "learning_rate": 1.879767665757127e-05, + "loss": 0.3371, + "step": 3029 + }, + { + "epoch": 0.18, + "learning_rate": 1.879674629151464e-05, + "loss": 0.3583, + "step": 3030 + }, + { + "epoch": 0.18, + "learning_rate": 1.8795815588677393e-05, + "loss": 0.353, + "step": 3031 + }, + { + "epoch": 0.18, + "learning_rate": 1.8794884549095157e-05, + "loss": 0.3552, + "step": 3032 + }, + { + "epoch": 0.18, + "learning_rate": 1.879395317280358e-05, + "loss": 0.3694, + "step": 3033 + }, + { + "epoch": 0.18, + "learning_rate": 1.8793021459838316e-05, + "loss": 0.3828, + "step": 3034 + }, + { + "epoch": 0.18, + "learning_rate": 1.8792089410235033e-05, + "loss": 0.3384, + "step": 3035 + }, + { + "epoch": 0.18, + "learning_rate": 1.879115702402942e-05, + "loss": 0.3331, + "step": 3036 + }, + { + "epoch": 0.18, + "learning_rate": 1.879022430125717e-05, + "loss": 0.358, + "step": 3037 + }, + { + "epoch": 0.18, + "learning_rate": 1.8789291241953994e-05, + "loss": 0.3357, + "step": 3038 + }, + { + "epoch": 0.18, + "learning_rate": 1.878835784615561e-05, + "loss": 0.3267, + "step": 3039 + }, + { + "epoch": 0.18, + "learning_rate": 1.8787424113897757e-05, + "loss": 0.3677, + "step": 3040 + }, + { + "epoch": 0.18, + "learning_rate": 1.8786490045216184e-05, + "loss": 0.3973, + "step": 3041 + }, + { + "epoch": 0.18, + "learning_rate": 1.8785555640146643e-05, + "loss": 0.3853, + "step": 3042 + }, + { + "epoch": 0.18, + "learning_rate": 1.8784620898724916e-05, + "loss": 0.3602, + "step": 3043 + }, + { + "epoch": 0.18, + "learning_rate": 1.878368582098679e-05, + "loss": 0.3825, + "step": 3044 + }, + { + "epoch": 0.18, + "learning_rate": 1.8782750406968054e-05, + "loss": 0.3488, + "step": 3045 + }, + { + "epoch": 0.18, + "learning_rate": 1.8781814656704527e-05, + "loss": 0.368, + "step": 3046 + }, + { + "epoch": 0.18, + "learning_rate": 1.8780878570232035e-05, + "loss": 0.3641, + "step": 3047 + }, + { + "epoch": 0.18, + "learning_rate": 1.8779942147586415e-05, + "loss": 0.3775, + "step": 3048 + }, + { + "epoch": 0.18, + "learning_rate": 1.8779005388803516e-05, + "loss": 0.3457, + "step": 3049 + }, + { + "epoch": 0.18, + "learning_rate": 1.8778068293919198e-05, + "loss": 0.3566, + "step": 3050 + }, + { + "epoch": 0.18, + "learning_rate": 1.8777130862969346e-05, + "loss": 0.3373, + "step": 3051 + }, + { + "epoch": 0.18, + "learning_rate": 1.8776193095989845e-05, + "loss": 0.3435, + "step": 3052 + }, + { + "epoch": 0.18, + "learning_rate": 1.8775254993016596e-05, + "loss": 0.3474, + "step": 3053 + }, + { + "epoch": 0.18, + "learning_rate": 1.8774316554085513e-05, + "loss": 0.3521, + "step": 3054 + }, + { + "epoch": 0.18, + "learning_rate": 1.877337777923253e-05, + "loss": 0.3499, + "step": 3055 + }, + { + "epoch": 0.18, + "learning_rate": 1.8772438668493578e-05, + "loss": 0.3507, + "step": 3056 + }, + { + "epoch": 0.18, + "learning_rate": 1.877149922190462e-05, + "loss": 0.3633, + "step": 3057 + }, + { + "epoch": 0.18, + "learning_rate": 1.8770559439501616e-05, + "loss": 0.3895, + "step": 3058 + }, + { + "epoch": 0.18, + "learning_rate": 1.876961932132055e-05, + "loss": 0.3878, + "step": 3059 + }, + { + "epoch": 0.18, + "learning_rate": 1.876867886739741e-05, + "loss": 0.3714, + "step": 3060 + }, + { + "epoch": 0.18, + "learning_rate": 1.8767738077768203e-05, + "loss": 0.3809, + "step": 3061 + }, + { + "epoch": 0.18, + "learning_rate": 1.8766796952468946e-05, + "loss": 0.3836, + "step": 3062 + }, + { + "epoch": 0.19, + "learning_rate": 1.8765855491535672e-05, + "loss": 0.3619, + "step": 3063 + }, + { + "epoch": 0.19, + "learning_rate": 1.876491369500442e-05, + "loss": 0.3873, + "step": 3064 + }, + { + "epoch": 0.19, + "learning_rate": 1.8763971562911252e-05, + "loss": 0.3493, + "step": 3065 + }, + { + "epoch": 0.19, + "learning_rate": 1.8763029095292232e-05, + "loss": 0.3641, + "step": 3066 + }, + { + "epoch": 0.19, + "learning_rate": 1.876208629218345e-05, + "loss": 0.3669, + "step": 3067 + }, + { + "epoch": 0.19, + "learning_rate": 1.8761143153620985e-05, + "loss": 0.3675, + "step": 3068 + }, + { + "epoch": 0.19, + "learning_rate": 1.8760199679640966e-05, + "loss": 0.3527, + "step": 3069 + }, + { + "epoch": 0.19, + "learning_rate": 1.8759255870279494e-05, + "loss": 0.3401, + "step": 3070 + }, + { + "epoch": 0.19, + "learning_rate": 1.875831172557272e-05, + "loss": 0.3474, + "step": 3071 + }, + { + "epoch": 0.19, + "learning_rate": 1.8757367245556773e-05, + "loss": 0.3817, + "step": 3072 + }, + { + "epoch": 0.19, + "learning_rate": 1.8756422430267823e-05, + "loss": 0.3697, + "step": 3073 + }, + { + "epoch": 0.19, + "learning_rate": 1.8755477279742044e-05, + "loss": 0.3292, + "step": 3074 + }, + { + "epoch": 0.19, + "learning_rate": 1.875453179401561e-05, + "loss": 0.3438, + "step": 3075 + }, + { + "epoch": 0.19, + "learning_rate": 1.8753585973124727e-05, + "loss": 0.3446, + "step": 3076 + }, + { + "epoch": 0.19, + "learning_rate": 1.8752639817105607e-05, + "loss": 0.356, + "step": 3077 + }, + { + "epoch": 0.19, + "learning_rate": 1.8751693325994466e-05, + "loss": 0.3717, + "step": 3078 + }, + { + "epoch": 0.19, + "learning_rate": 1.8750746499827545e-05, + "loss": 0.3393, + "step": 3079 + }, + { + "epoch": 0.19, + "learning_rate": 1.874979933864109e-05, + "loss": 0.37, + "step": 3080 + }, + { + "epoch": 0.19, + "learning_rate": 1.874885184247137e-05, + "loss": 0.3753, + "step": 3081 + }, + { + "epoch": 0.19, + "learning_rate": 1.874790401135465e-05, + "loss": 0.3318, + "step": 3082 + }, + { + "epoch": 0.19, + "learning_rate": 1.8746955845327218e-05, + "loss": 0.3598, + "step": 3083 + }, + { + "epoch": 0.19, + "learning_rate": 1.874600734442538e-05, + "loss": 0.3502, + "step": 3084 + }, + { + "epoch": 0.19, + "learning_rate": 1.8745058508685453e-05, + "loss": 0.3569, + "step": 3085 + }, + { + "epoch": 0.19, + "learning_rate": 1.874410933814375e-05, + "loss": 0.3956, + "step": 3086 + }, + { + "epoch": 0.19, + "learning_rate": 1.8743159832836616e-05, + "loss": 0.3379, + "step": 3087 + }, + { + "epoch": 0.19, + "learning_rate": 1.8742209992800405e-05, + "loss": 0.3655, + "step": 3088 + }, + { + "epoch": 0.19, + "learning_rate": 1.874125981807148e-05, + "loss": 0.351, + "step": 3089 + }, + { + "epoch": 0.19, + "learning_rate": 1.8740309308686215e-05, + "loss": 0.3747, + "step": 3090 + }, + { + "epoch": 0.19, + "learning_rate": 1.8739358464681005e-05, + "loss": 0.3468, + "step": 3091 + }, + { + "epoch": 0.19, + "learning_rate": 1.8738407286092245e-05, + "loss": 0.375, + "step": 3092 + }, + { + "epoch": 0.19, + "learning_rate": 1.873745577295636e-05, + "loss": 0.3845, + "step": 3093 + }, + { + "epoch": 0.19, + "learning_rate": 1.8736503925309773e-05, + "loss": 0.3884, + "step": 3094 + }, + { + "epoch": 0.19, + "learning_rate": 1.8735551743188924e-05, + "loss": 0.3393, + "step": 3095 + }, + { + "epoch": 0.19, + "learning_rate": 1.8734599226630274e-05, + "loss": 0.3454, + "step": 3096 + }, + { + "epoch": 0.19, + "learning_rate": 1.873364637567028e-05, + "loss": 0.4026, + "step": 3097 + }, + { + "epoch": 0.19, + "learning_rate": 1.873269319034543e-05, + "loss": 0.3521, + "step": 3098 + }, + { + "epoch": 0.19, + "learning_rate": 1.873173967069221e-05, + "loss": 0.3463, + "step": 3099 + }, + { + "epoch": 0.19, + "learning_rate": 1.873078581674713e-05, + "loss": 0.3237, + "step": 3100 + }, + { + "epoch": 0.19, + "learning_rate": 1.8729831628546704e-05, + "loss": 0.3945, + "step": 3101 + }, + { + "epoch": 0.19, + "learning_rate": 1.8728877106127465e-05, + "loss": 0.3686, + "step": 3102 + }, + { + "epoch": 0.19, + "learning_rate": 1.872792224952596e-05, + "loss": 0.3507, + "step": 3103 + }, + { + "epoch": 0.19, + "learning_rate": 1.872696705877874e-05, + "loss": 0.341, + "step": 3104 + }, + { + "epoch": 0.19, + "learning_rate": 1.8726011533922373e-05, + "loss": 0.3577, + "step": 3105 + }, + { + "epoch": 0.19, + "learning_rate": 1.872505567499345e-05, + "loss": 0.3474, + "step": 3106 + }, + { + "epoch": 0.19, + "learning_rate": 1.8724099482028554e-05, + "loss": 0.3739, + "step": 3107 + }, + { + "epoch": 0.19, + "learning_rate": 1.87231429550643e-05, + "loss": 0.3633, + "step": 3108 + }, + { + "epoch": 0.19, + "learning_rate": 1.8722186094137307e-05, + "loss": 0.3566, + "step": 3109 + }, + { + "epoch": 0.19, + "learning_rate": 1.872122889928421e-05, + "loss": 0.38, + "step": 3110 + }, + { + "epoch": 0.19, + "learning_rate": 1.8720271370541646e-05, + "loss": 0.3666, + "step": 3111 + }, + { + "epoch": 0.19, + "learning_rate": 1.8719313507946286e-05, + "loss": 0.3661, + "step": 3112 + }, + { + "epoch": 0.19, + "learning_rate": 1.8718355311534795e-05, + "loss": 0.3948, + "step": 3113 + }, + { + "epoch": 0.19, + "learning_rate": 1.871739678134386e-05, + "loss": 0.3633, + "step": 3114 + }, + { + "epoch": 0.19, + "learning_rate": 1.871643791741017e-05, + "loss": 0.3357, + "step": 3115 + }, + { + "epoch": 0.19, + "learning_rate": 1.8715478719770445e-05, + "loss": 0.3571, + "step": 3116 + }, + { + "epoch": 0.19, + "learning_rate": 1.8714519188461402e-05, + "loss": 0.3491, + "step": 3117 + }, + { + "epoch": 0.19, + "learning_rate": 1.871355932351978e-05, + "loss": 0.3705, + "step": 3118 + }, + { + "epoch": 0.19, + "learning_rate": 1.8712599124982322e-05, + "loss": 0.375, + "step": 3119 + }, + { + "epoch": 0.19, + "learning_rate": 1.8711638592885793e-05, + "loss": 0.3714, + "step": 3120 + }, + { + "epoch": 0.19, + "learning_rate": 1.8710677727266965e-05, + "loss": 0.3655, + "step": 3121 + }, + { + "epoch": 0.19, + "learning_rate": 1.8709716528162626e-05, + "loss": 0.3507, + "step": 3122 + }, + { + "epoch": 0.19, + "learning_rate": 1.870875499560957e-05, + "loss": 0.3753, + "step": 3123 + }, + { + "epoch": 0.19, + "learning_rate": 1.8707793129644618e-05, + "loss": 0.3446, + "step": 3124 + }, + { + "epoch": 0.19, + "learning_rate": 1.8706830930304587e-05, + "loss": 0.3415, + "step": 3125 + }, + { + "epoch": 0.19, + "learning_rate": 1.8705868397626318e-05, + "loss": 0.3323, + "step": 3126 + }, + { + "epoch": 0.19, + "learning_rate": 1.870490553164666e-05, + "loss": 0.3926, + "step": 3127 + }, + { + "epoch": 0.19, + "learning_rate": 1.8703942332402478e-05, + "loss": 0.3479, + "step": 3128 + }, + { + "epoch": 0.19, + "learning_rate": 1.8702978799930648e-05, + "loss": 0.4026, + "step": 3129 + }, + { + "epoch": 0.19, + "learning_rate": 1.870201493426805e-05, + "loss": 0.339, + "step": 3130 + }, + { + "epoch": 0.19, + "learning_rate": 1.87010507354516e-05, + "loss": 0.3348, + "step": 3131 + }, + { + "epoch": 0.19, + "learning_rate": 1.87000862035182e-05, + "loss": 0.361, + "step": 3132 + }, + { + "epoch": 0.19, + "learning_rate": 1.8699121338504784e-05, + "loss": 0.3585, + "step": 3133 + }, + { + "epoch": 0.19, + "learning_rate": 1.869815614044828e-05, + "loss": 0.3541, + "step": 3134 + }, + { + "epoch": 0.19, + "learning_rate": 1.8697190609385657e-05, + "loss": 0.3295, + "step": 3135 + }, + { + "epoch": 0.19, + "learning_rate": 1.869622474535387e-05, + "loss": 0.3541, + "step": 3136 + }, + { + "epoch": 0.19, + "learning_rate": 1.8695258548389898e-05, + "loss": 0.3772, + "step": 3137 + }, + { + "epoch": 0.19, + "learning_rate": 1.8694292018530733e-05, + "loss": 0.3602, + "step": 3138 + }, + { + "epoch": 0.19, + "learning_rate": 1.869332515581338e-05, + "loss": 0.356, + "step": 3139 + }, + { + "epoch": 0.19, + "learning_rate": 1.869235796027485e-05, + "loss": 0.361, + "step": 3140 + }, + { + "epoch": 0.19, + "learning_rate": 1.869139043195217e-05, + "loss": 0.3859, + "step": 3141 + }, + { + "epoch": 0.19, + "learning_rate": 1.8690422570882388e-05, + "loss": 0.3719, + "step": 3142 + }, + { + "epoch": 0.19, + "learning_rate": 1.868945437710256e-05, + "loss": 0.3669, + "step": 3143 + }, + { + "epoch": 0.19, + "learning_rate": 1.8688485850649745e-05, + "loss": 0.3689, + "step": 3144 + }, + { + "epoch": 0.19, + "learning_rate": 1.868751699156103e-05, + "loss": 0.411, + "step": 3145 + }, + { + "epoch": 0.19, + "learning_rate": 1.8686547799873504e-05, + "loss": 0.3371, + "step": 3146 + }, + { + "epoch": 0.19, + "learning_rate": 1.8685578275624266e-05, + "loss": 0.3725, + "step": 3147 + }, + { + "epoch": 0.19, + "learning_rate": 1.8684608418850446e-05, + "loss": 0.3485, + "step": 3148 + }, + { + "epoch": 0.19, + "learning_rate": 1.868363822958917e-05, + "loss": 0.3767, + "step": 3149 + }, + { + "epoch": 0.19, + "learning_rate": 1.868266770787758e-05, + "loss": 0.351, + "step": 3150 + }, + { + "epoch": 0.19, + "learning_rate": 1.868169685375283e-05, + "loss": 0.358, + "step": 3151 + }, + { + "epoch": 0.19, + "learning_rate": 1.8680725667252092e-05, + "loss": 0.3736, + "step": 3152 + }, + { + "epoch": 0.19, + "learning_rate": 1.8679754148412547e-05, + "loss": 0.3686, + "step": 3153 + }, + { + "epoch": 0.19, + "learning_rate": 1.867878229727139e-05, + "loss": 0.3521, + "step": 3154 + }, + { + "epoch": 0.19, + "learning_rate": 1.8677810113865826e-05, + "loss": 0.3255, + "step": 3155 + }, + { + "epoch": 0.19, + "learning_rate": 1.8676837598233077e-05, + "loss": 0.3613, + "step": 3156 + }, + { + "epoch": 0.19, + "learning_rate": 1.8675864750410374e-05, + "loss": 0.3108, + "step": 3157 + }, + { + "epoch": 0.19, + "learning_rate": 1.867489157043496e-05, + "loss": 0.3453, + "step": 3158 + }, + { + "epoch": 0.19, + "learning_rate": 1.86739180583441e-05, + "loss": 0.3823, + "step": 3159 + }, + { + "epoch": 0.19, + "learning_rate": 1.8672944214175058e-05, + "loss": 0.3633, + "step": 3160 + }, + { + "epoch": 0.19, + "learning_rate": 1.867197003796512e-05, + "loss": 0.3931, + "step": 3161 + }, + { + "epoch": 0.19, + "learning_rate": 1.8670995529751577e-05, + "loss": 0.3359, + "step": 3162 + }, + { + "epoch": 0.19, + "learning_rate": 1.867002068957175e-05, + "loss": 0.3488, + "step": 3163 + }, + { + "epoch": 0.19, + "learning_rate": 1.8669045517462948e-05, + "loss": 0.3711, + "step": 3164 + }, + { + "epoch": 0.19, + "learning_rate": 1.866807001346251e-05, + "loss": 0.3563, + "step": 3165 + }, + { + "epoch": 0.19, + "learning_rate": 1.8667094177607782e-05, + "loss": 0.3504, + "step": 3166 + }, + { + "epoch": 0.19, + "learning_rate": 1.8666118009936123e-05, + "loss": 0.3518, + "step": 3167 + }, + { + "epoch": 0.19, + "learning_rate": 1.8665141510484908e-05, + "loss": 0.3783, + "step": 3168 + }, + { + "epoch": 0.19, + "learning_rate": 1.866416467929152e-05, + "loss": 0.411, + "step": 3169 + }, + { + "epoch": 0.19, + "learning_rate": 1.866318751639336e-05, + "loss": 0.3235, + "step": 3170 + }, + { + "epoch": 0.19, + "learning_rate": 1.8662210021827832e-05, + "loss": 0.3281, + "step": 3171 + }, + { + "epoch": 0.19, + "learning_rate": 1.866123219563236e-05, + "loss": 0.3984, + "step": 3172 + }, + { + "epoch": 0.19, + "learning_rate": 1.866025403784439e-05, + "loss": 0.3817, + "step": 3173 + }, + { + "epoch": 0.19, + "learning_rate": 1.8659275548501356e-05, + "loss": 0.3591, + "step": 3174 + }, + { + "epoch": 0.19, + "learning_rate": 1.865829672764073e-05, + "loss": 0.3345, + "step": 3175 + }, + { + "epoch": 0.19, + "learning_rate": 1.865731757529998e-05, + "loss": 0.3647, + "step": 3176 + }, + { + "epoch": 0.19, + "learning_rate": 1.865633809151659e-05, + "loss": 0.3705, + "step": 3177 + }, + { + "epoch": 0.19, + "learning_rate": 1.865535827632807e-05, + "loss": 0.3496, + "step": 3178 + }, + { + "epoch": 0.19, + "learning_rate": 1.865437812977192e-05, + "loss": 0.3823, + "step": 3179 + }, + { + "epoch": 0.19, + "learning_rate": 1.8653397651885675e-05, + "loss": 0.37, + "step": 3180 + }, + { + "epoch": 0.19, + "learning_rate": 1.8652416842706864e-05, + "loss": 0.3647, + "step": 3181 + }, + { + "epoch": 0.19, + "learning_rate": 1.865143570227304e-05, + "loss": 0.3781, + "step": 3182 + }, + { + "epoch": 0.19, + "learning_rate": 1.8650454230621767e-05, + "loss": 0.3164, + "step": 3183 + }, + { + "epoch": 0.19, + "learning_rate": 1.8649472427790617e-05, + "loss": 0.3454, + "step": 3184 + }, + { + "epoch": 0.19, + "learning_rate": 1.8648490293817185e-05, + "loss": 0.3753, + "step": 3185 + }, + { + "epoch": 0.19, + "learning_rate": 1.864750782873906e-05, + "loss": 0.3661, + "step": 3186 + }, + { + "epoch": 0.19, + "learning_rate": 1.864652503259387e-05, + "loss": 0.3438, + "step": 3187 + }, + { + "epoch": 0.19, + "learning_rate": 1.8645541905419228e-05, + "loss": 0.3493, + "step": 3188 + }, + { + "epoch": 0.19, + "learning_rate": 1.864455844725278e-05, + "loss": 0.3475, + "step": 3189 + }, + { + "epoch": 0.19, + "learning_rate": 1.8643574658132177e-05, + "loss": 0.351, + "step": 3190 + }, + { + "epoch": 0.19, + "learning_rate": 1.8642590538095077e-05, + "loss": 0.3725, + "step": 3191 + }, + { + "epoch": 0.19, + "learning_rate": 1.8641606087179163e-05, + "loss": 0.3691, + "step": 3192 + }, + { + "epoch": 0.19, + "learning_rate": 1.8640621305422124e-05, + "loss": 0.3267, + "step": 3193 + }, + { + "epoch": 0.19, + "learning_rate": 1.863963619286166e-05, + "loss": 0.3622, + "step": 3194 + }, + { + "epoch": 0.19, + "learning_rate": 1.8638650749535486e-05, + "loss": 0.3842, + "step": 3195 + }, + { + "epoch": 0.19, + "learning_rate": 1.8637664975481327e-05, + "loss": 0.3697, + "step": 3196 + }, + { + "epoch": 0.19, + "learning_rate": 1.8636678870736928e-05, + "loss": 0.3627, + "step": 3197 + }, + { + "epoch": 0.19, + "learning_rate": 1.863569243534004e-05, + "loss": 0.3549, + "step": 3198 + }, + { + "epoch": 0.19, + "learning_rate": 1.8634705669328427e-05, + "loss": 0.3691, + "step": 3199 + }, + { + "epoch": 0.19, + "learning_rate": 1.863371857273987e-05, + "loss": 0.3987, + "step": 3200 + }, + { + "epoch": 0.19, + "learning_rate": 1.8632731145612153e-05, + "loss": 0.3633, + "step": 3201 + }, + { + "epoch": 0.19, + "learning_rate": 1.8631743387983087e-05, + "loss": 0.3454, + "step": 3202 + }, + { + "epoch": 0.19, + "learning_rate": 1.8630755299890485e-05, + "loss": 0.3825, + "step": 3203 + }, + { + "epoch": 0.19, + "learning_rate": 1.8629766881372174e-05, + "loss": 0.3507, + "step": 3204 + }, + { + "epoch": 0.19, + "learning_rate": 1.8628778132465995e-05, + "loss": 0.3496, + "step": 3205 + }, + { + "epoch": 0.19, + "learning_rate": 1.8627789053209808e-05, + "loss": 0.3772, + "step": 3206 + }, + { + "epoch": 0.19, + "learning_rate": 1.8626799643641475e-05, + "loss": 0.3599, + "step": 3207 + }, + { + "epoch": 0.19, + "learning_rate": 1.862580990379887e-05, + "loss": 0.3524, + "step": 3208 + }, + { + "epoch": 0.19, + "learning_rate": 1.8624819833719897e-05, + "loss": 0.3384, + "step": 3209 + }, + { + "epoch": 0.19, + "learning_rate": 1.862382943344245e-05, + "loss": 0.3725, + "step": 3210 + }, + { + "epoch": 0.19, + "learning_rate": 1.862283870300445e-05, + "loss": 0.3237, + "step": 3211 + }, + { + "epoch": 0.19, + "learning_rate": 1.8621847642443825e-05, + "loss": 0.3396, + "step": 3212 + }, + { + "epoch": 0.19, + "learning_rate": 1.8620856251798524e-05, + "loss": 0.3571, + "step": 3213 + }, + { + "epoch": 0.19, + "learning_rate": 1.8619864531106496e-05, + "loss": 0.3418, + "step": 3214 + }, + { + "epoch": 0.19, + "learning_rate": 1.861887248040571e-05, + "loss": 0.3211, + "step": 3215 + }, + { + "epoch": 0.19, + "learning_rate": 1.8617880099734144e-05, + "loss": 0.3613, + "step": 3216 + }, + { + "epoch": 0.19, + "learning_rate": 1.8616887389129795e-05, + "loss": 0.3669, + "step": 3217 + }, + { + "epoch": 0.19, + "learning_rate": 1.8615894348630666e-05, + "loss": 0.3597, + "step": 3218 + }, + { + "epoch": 0.19, + "learning_rate": 1.8614900978274777e-05, + "loss": 0.3814, + "step": 3219 + }, + { + "epoch": 0.19, + "learning_rate": 1.8613907278100162e-05, + "loss": 0.3725, + "step": 3220 + }, + { + "epoch": 0.19, + "learning_rate": 1.8612913248144853e-05, + "loss": 0.3661, + "step": 3221 + }, + { + "epoch": 0.19, + "learning_rate": 1.861191888844692e-05, + "loss": 0.3516, + "step": 3222 + }, + { + "epoch": 0.19, + "learning_rate": 1.8610924199044422e-05, + "loss": 0.3223, + "step": 3223 + }, + { + "epoch": 0.19, + "learning_rate": 1.8609929179975442e-05, + "loss": 0.3164, + "step": 3224 + }, + { + "epoch": 0.19, + "learning_rate": 1.860893383127808e-05, + "loss": 0.3429, + "step": 3225 + }, + { + "epoch": 0.19, + "learning_rate": 1.8607938152990436e-05, + "loss": 0.3496, + "step": 3226 + }, + { + "epoch": 0.19, + "learning_rate": 1.8606942145150633e-05, + "loss": 0.3424, + "step": 3227 + }, + { + "epoch": 0.2, + "learning_rate": 1.8605945807796803e-05, + "loss": 0.3549, + "step": 3228 + }, + { + "epoch": 0.2, + "learning_rate": 1.8604949140967085e-05, + "loss": 0.3488, + "step": 3229 + }, + { + "epoch": 0.2, + "learning_rate": 1.8603952144699643e-05, + "loss": 0.3569, + "step": 3230 + }, + { + "epoch": 0.2, + "learning_rate": 1.8602954819032642e-05, + "loss": 0.3446, + "step": 3231 + }, + { + "epoch": 0.2, + "learning_rate": 1.8601957164004266e-05, + "loss": 0.3792, + "step": 3232 + }, + { + "epoch": 0.2, + "learning_rate": 1.8600959179652708e-05, + "loss": 0.3309, + "step": 3233 + }, + { + "epoch": 0.2, + "learning_rate": 1.859996086601618e-05, + "loss": 0.3686, + "step": 3234 + }, + { + "epoch": 0.2, + "learning_rate": 1.8598962223132902e-05, + "loss": 0.3638, + "step": 3235 + }, + { + "epoch": 0.2, + "learning_rate": 1.8597963251041097e-05, + "loss": 0.3705, + "step": 3236 + }, + { + "epoch": 0.2, + "learning_rate": 1.8596963949779027e-05, + "loss": 0.3599, + "step": 3237 + }, + { + "epoch": 0.2, + "learning_rate": 1.8595964319384934e-05, + "loss": 0.3555, + "step": 3238 + }, + { + "epoch": 0.2, + "learning_rate": 1.8594964359897095e-05, + "loss": 0.3538, + "step": 3239 + }, + { + "epoch": 0.2, + "learning_rate": 1.8593964071353795e-05, + "loss": 0.3555, + "step": 3240 + }, + { + "epoch": 0.2, + "learning_rate": 1.8592963453793326e-05, + "loss": 0.3532, + "step": 3241 + }, + { + "epoch": 0.2, + "learning_rate": 1.8591962507254e-05, + "loss": 0.346, + "step": 3242 + }, + { + "epoch": 0.2, + "learning_rate": 1.8590961231774132e-05, + "loss": 0.3463, + "step": 3243 + }, + { + "epoch": 0.2, + "learning_rate": 1.8589959627392064e-05, + "loss": 0.3717, + "step": 3244 + }, + { + "epoch": 0.2, + "learning_rate": 1.858895769414614e-05, + "loss": 0.365, + "step": 3245 + }, + { + "epoch": 0.2, + "learning_rate": 1.8587955432074714e-05, + "loss": 0.3834, + "step": 3246 + }, + { + "epoch": 0.2, + "learning_rate": 1.8586952841216157e-05, + "loss": 0.3708, + "step": 3247 + }, + { + "epoch": 0.2, + "learning_rate": 1.8585949921608857e-05, + "loss": 0.3597, + "step": 3248 + }, + { + "epoch": 0.2, + "learning_rate": 1.8584946673291207e-05, + "loss": 0.3284, + "step": 3249 + }, + { + "epoch": 0.2, + "learning_rate": 1.8583943096301623e-05, + "loss": 0.3767, + "step": 3250 + }, + { + "epoch": 0.2, + "learning_rate": 1.858293919067852e-05, + "loss": 0.3574, + "step": 3251 + }, + { + "epoch": 0.2, + "learning_rate": 1.8581934956460333e-05, + "loss": 0.3906, + "step": 3252 + }, + { + "epoch": 0.2, + "learning_rate": 1.858093039368551e-05, + "loss": 0.3401, + "step": 3253 + }, + { + "epoch": 0.2, + "learning_rate": 1.857992550239251e-05, + "loss": 0.3535, + "step": 3254 + }, + { + "epoch": 0.2, + "learning_rate": 1.8578920282619804e-05, + "loss": 0.3736, + "step": 3255 + }, + { + "epoch": 0.2, + "learning_rate": 1.857791473440588e-05, + "loss": 0.37, + "step": 3256 + }, + { + "epoch": 0.2, + "learning_rate": 1.8576908857789232e-05, + "loss": 0.3502, + "step": 3257 + }, + { + "epoch": 0.2, + "learning_rate": 1.8575902652808365e-05, + "loss": 0.3566, + "step": 3258 + }, + { + "epoch": 0.2, + "learning_rate": 1.857489611950181e-05, + "loss": 0.3742, + "step": 3259 + }, + { + "epoch": 0.2, + "learning_rate": 1.8573889257908104e-05, + "loss": 0.3513, + "step": 3260 + }, + { + "epoch": 0.2, + "learning_rate": 1.857288206806578e-05, + "loss": 0.3605, + "step": 3261 + }, + { + "epoch": 0.2, + "learning_rate": 1.857187455001341e-05, + "loss": 0.3393, + "step": 3262 + }, + { + "epoch": 0.2, + "learning_rate": 1.857086670378956e-05, + "loss": 0.3507, + "step": 3263 + }, + { + "epoch": 0.2, + "learning_rate": 1.8569858529432826e-05, + "loss": 0.3465, + "step": 3264 + }, + { + "epoch": 0.2, + "learning_rate": 1.856885002698179e-05, + "loss": 0.3976, + "step": 3265 + }, + { + "epoch": 0.2, + "learning_rate": 1.8567841196475076e-05, + "loss": 0.3384, + "step": 3266 + }, + { + "epoch": 0.2, + "learning_rate": 1.85668320379513e-05, + "loss": 0.3468, + "step": 3267 + }, + { + "epoch": 0.2, + "learning_rate": 1.8565822551449092e-05, + "loss": 0.3407, + "step": 3268 + }, + { + "epoch": 0.2, + "learning_rate": 1.8564812737007113e-05, + "loss": 0.3719, + "step": 3269 + }, + { + "epoch": 0.2, + "learning_rate": 1.856380259466401e-05, + "loss": 0.3524, + "step": 3270 + }, + { + "epoch": 0.2, + "learning_rate": 1.8562792124458467e-05, + "loss": 0.3513, + "step": 3271 + }, + { + "epoch": 0.2, + "learning_rate": 1.8561781326429165e-05, + "loss": 0.3345, + "step": 3272 + }, + { + "epoch": 0.2, + "learning_rate": 1.85607702006148e-05, + "loss": 0.3276, + "step": 3273 + }, + { + "epoch": 0.2, + "learning_rate": 1.8559758747054087e-05, + "loss": 0.3605, + "step": 3274 + }, + { + "epoch": 0.2, + "learning_rate": 1.8558746965785745e-05, + "loss": 0.3432, + "step": 3275 + }, + { + "epoch": 0.2, + "learning_rate": 1.855773485684851e-05, + "loss": 0.3705, + "step": 3276 + }, + { + "epoch": 0.2, + "learning_rate": 1.8556722420281135e-05, + "loss": 0.3326, + "step": 3277 + }, + { + "epoch": 0.2, + "learning_rate": 1.8555709656122374e-05, + "loss": 0.3719, + "step": 3278 + }, + { + "epoch": 0.2, + "learning_rate": 1.8554696564411005e-05, + "loss": 0.3532, + "step": 3279 + }, + { + "epoch": 0.2, + "learning_rate": 1.8553683145185816e-05, + "loss": 0.3689, + "step": 3280 + }, + { + "epoch": 0.2, + "learning_rate": 1.8552669398485598e-05, + "loss": 0.3521, + "step": 3281 + }, + { + "epoch": 0.2, + "learning_rate": 1.8551655324349167e-05, + "loss": 0.3915, + "step": 3282 + }, + { + "epoch": 0.2, + "learning_rate": 1.855064092281535e-05, + "loss": 0.3451, + "step": 3283 + }, + { + "epoch": 0.2, + "learning_rate": 1.8549626193922973e-05, + "loss": 0.3569, + "step": 3284 + }, + { + "epoch": 0.2, + "learning_rate": 1.854861113771089e-05, + "loss": 0.385, + "step": 3285 + }, + { + "epoch": 0.2, + "learning_rate": 1.8547595754217965e-05, + "loss": 0.3488, + "step": 3286 + }, + { + "epoch": 0.2, + "learning_rate": 1.8546580043483075e-05, + "loss": 0.3393, + "step": 3287 + }, + { + "epoch": 0.2, + "learning_rate": 1.854556400554509e-05, + "loss": 0.3382, + "step": 3288 + }, + { + "epoch": 0.2, + "learning_rate": 1.8544547640442926e-05, + "loss": 0.3647, + "step": 3289 + }, + { + "epoch": 0.2, + "learning_rate": 1.8543530948215483e-05, + "loss": 0.4076, + "step": 3290 + }, + { + "epoch": 0.2, + "learning_rate": 1.854251392890169e-05, + "loss": 0.3583, + "step": 3291 + }, + { + "epoch": 0.2, + "learning_rate": 1.8541496582540483e-05, + "loss": 0.3664, + "step": 3292 + }, + { + "epoch": 0.2, + "learning_rate": 1.854047890917081e-05, + "loss": 0.3217, + "step": 3293 + }, + { + "epoch": 0.2, + "learning_rate": 1.853946090883163e-05, + "loss": 0.3783, + "step": 3294 + }, + { + "epoch": 0.2, + "learning_rate": 1.8538442581561922e-05, + "loss": 0.3499, + "step": 3295 + }, + { + "epoch": 0.2, + "learning_rate": 1.8537423927400666e-05, + "loss": 0.3619, + "step": 3296 + }, + { + "epoch": 0.2, + "learning_rate": 1.8536404946386865e-05, + "loss": 0.3451, + "step": 3297 + }, + { + "epoch": 0.2, + "learning_rate": 1.8535385638559532e-05, + "loss": 0.3306, + "step": 3298 + }, + { + "epoch": 0.2, + "learning_rate": 1.8534366003957687e-05, + "loss": 0.3636, + "step": 3299 + }, + { + "epoch": 0.2, + "learning_rate": 1.8533346042620367e-05, + "loss": 0.3691, + "step": 3300 + }, + { + "epoch": 0.2, + "learning_rate": 1.8532325754586623e-05, + "loss": 0.3686, + "step": 3301 + }, + { + "epoch": 0.2, + "learning_rate": 1.8531305139895518e-05, + "loss": 0.3867, + "step": 3302 + }, + { + "epoch": 0.2, + "learning_rate": 1.8530284198586116e-05, + "loss": 0.3591, + "step": 3303 + }, + { + "epoch": 0.2, + "learning_rate": 1.8529262930697514e-05, + "loss": 0.3764, + "step": 3304 + }, + { + "epoch": 0.2, + "learning_rate": 1.852824133626881e-05, + "loss": 0.3636, + "step": 3305 + }, + { + "epoch": 0.2, + "learning_rate": 1.852721941533911e-05, + "loss": 0.3557, + "step": 3306 + }, + { + "epoch": 0.2, + "learning_rate": 1.852619716794754e-05, + "loss": 0.3463, + "step": 3307 + }, + { + "epoch": 0.2, + "learning_rate": 1.852517459413324e-05, + "loss": 0.3597, + "step": 3308 + }, + { + "epoch": 0.2, + "learning_rate": 1.852415169393535e-05, + "loss": 0.3864, + "step": 3309 + }, + { + "epoch": 0.2, + "learning_rate": 1.852312846739304e-05, + "loss": 0.3683, + "step": 3310 + }, + { + "epoch": 0.2, + "learning_rate": 1.8522104914545486e-05, + "loss": 0.339, + "step": 3311 + }, + { + "epoch": 0.2, + "learning_rate": 1.8521081035431865e-05, + "loss": 0.3619, + "step": 3312 + }, + { + "epoch": 0.2, + "learning_rate": 1.8520056830091382e-05, + "loss": 0.3415, + "step": 3313 + }, + { + "epoch": 0.2, + "learning_rate": 1.8519032298563245e-05, + "loss": 0.3287, + "step": 3314 + }, + { + "epoch": 0.2, + "learning_rate": 1.8518007440886684e-05, + "loss": 0.3549, + "step": 3315 + }, + { + "epoch": 0.2, + "learning_rate": 1.8516982257100927e-05, + "loss": 0.3365, + "step": 3316 + }, + { + "epoch": 0.2, + "learning_rate": 1.8515956747245232e-05, + "loss": 0.3697, + "step": 3317 + }, + { + "epoch": 0.2, + "learning_rate": 1.8514930911358852e-05, + "loss": 0.3811, + "step": 3318 + }, + { + "epoch": 0.2, + "learning_rate": 1.851390474948106e-05, + "loss": 0.341, + "step": 3319 + }, + { + "epoch": 0.2, + "learning_rate": 1.8512878261651154e-05, + "loss": 0.3638, + "step": 3320 + }, + { + "epoch": 0.2, + "learning_rate": 1.851185144790842e-05, + "loss": 0.3474, + "step": 3321 + }, + { + "epoch": 0.2, + "learning_rate": 1.8510824308292178e-05, + "loss": 0.3209, + "step": 3322 + }, + { + "epoch": 0.2, + "learning_rate": 1.8509796842841746e-05, + "loss": 0.3574, + "step": 3323 + }, + { + "epoch": 0.2, + "learning_rate": 1.850876905159646e-05, + "loss": 0.3326, + "step": 3324 + }, + { + "epoch": 0.2, + "learning_rate": 1.8507740934595675e-05, + "loss": 0.3633, + "step": 3325 + }, + { + "epoch": 0.2, + "learning_rate": 1.8506712491878747e-05, + "loss": 0.4021, + "step": 3326 + }, + { + "epoch": 0.2, + "learning_rate": 1.8505683723485052e-05, + "loss": 0.3057, + "step": 3327 + }, + { + "epoch": 0.2, + "learning_rate": 1.850465462945397e-05, + "loss": 0.3354, + "step": 3328 + }, + { + "epoch": 0.2, + "learning_rate": 1.8503625209824906e-05, + "loss": 0.3527, + "step": 3329 + }, + { + "epoch": 0.2, + "learning_rate": 1.850259546463727e-05, + "loss": 0.3624, + "step": 3330 + }, + { + "epoch": 0.2, + "learning_rate": 1.8501565393930487e-05, + "loss": 0.368, + "step": 3331 + }, + { + "epoch": 0.2, + "learning_rate": 1.850053499774399e-05, + "loss": 0.3463, + "step": 3332 + }, + { + "epoch": 0.2, + "learning_rate": 1.849950427611723e-05, + "loss": 0.334, + "step": 3333 + }, + { + "epoch": 0.2, + "learning_rate": 1.849847322908966e-05, + "loss": 0.4107, + "step": 3334 + }, + { + "epoch": 0.2, + "learning_rate": 1.8497441856700763e-05, + "loss": 0.3677, + "step": 3335 + }, + { + "epoch": 0.2, + "learning_rate": 1.8496410158990017e-05, + "loss": 0.365, + "step": 3336 + }, + { + "epoch": 0.2, + "learning_rate": 1.8495378135996932e-05, + "loss": 0.3474, + "step": 3337 + }, + { + "epoch": 0.2, + "learning_rate": 1.8494345787761007e-05, + "loss": 0.3733, + "step": 3338 + }, + { + "epoch": 0.2, + "learning_rate": 1.8493313114321767e-05, + "loss": 0.365, + "step": 3339 + }, + { + "epoch": 0.2, + "learning_rate": 1.849228011571875e-05, + "loss": 0.3619, + "step": 3340 + }, + { + "epoch": 0.2, + "learning_rate": 1.8491246791991506e-05, + "loss": 0.365, + "step": 3341 + }, + { + "epoch": 0.2, + "learning_rate": 1.849021314317959e-05, + "loss": 0.3446, + "step": 3342 + }, + { + "epoch": 0.2, + "learning_rate": 1.848917916932258e-05, + "loss": 0.375, + "step": 3343 + }, + { + "epoch": 0.2, + "learning_rate": 1.848814487046006e-05, + "loss": 0.3326, + "step": 3344 + }, + { + "epoch": 0.2, + "learning_rate": 1.8487110246631625e-05, + "loss": 0.3761, + "step": 3345 + }, + { + "epoch": 0.2, + "learning_rate": 1.8486075297876887e-05, + "loss": 0.3705, + "step": 3346 + }, + { + "epoch": 0.2, + "learning_rate": 1.8485040024235472e-05, + "loss": 0.368, + "step": 3347 + }, + { + "epoch": 0.2, + "learning_rate": 1.848400442574701e-05, + "loss": 0.3471, + "step": 3348 + }, + { + "epoch": 0.2, + "learning_rate": 1.8482968502451154e-05, + "loss": 0.3636, + "step": 3349 + }, + { + "epoch": 0.2, + "learning_rate": 1.8481932254387556e-05, + "loss": 0.3608, + "step": 3350 + }, + { + "epoch": 0.2, + "learning_rate": 1.8480895681595897e-05, + "loss": 0.3474, + "step": 3351 + }, + { + "epoch": 0.2, + "learning_rate": 1.8479858784115852e-05, + "loss": 0.3664, + "step": 3352 + }, + { + "epoch": 0.2, + "learning_rate": 1.847882156198713e-05, + "loss": 0.3753, + "step": 3353 + }, + { + "epoch": 0.2, + "learning_rate": 1.8477784015249433e-05, + "loss": 0.3661, + "step": 3354 + }, + { + "epoch": 0.2, + "learning_rate": 1.8476746143942487e-05, + "loss": 0.3605, + "step": 3355 + }, + { + "epoch": 0.2, + "learning_rate": 1.8475707948106024e-05, + "loss": 0.3898, + "step": 3356 + }, + { + "epoch": 0.2, + "learning_rate": 1.847466942777979e-05, + "loss": 0.368, + "step": 3357 + }, + { + "epoch": 0.2, + "learning_rate": 1.847363058300355e-05, + "loss": 0.3488, + "step": 3358 + }, + { + "epoch": 0.2, + "learning_rate": 1.8472591413817068e-05, + "loss": 0.3513, + "step": 3359 + }, + { + "epoch": 0.2, + "learning_rate": 1.8471551920260132e-05, + "loss": 0.3493, + "step": 3360 + }, + { + "epoch": 0.2, + "learning_rate": 1.847051210237254e-05, + "loss": 0.3836, + "step": 3361 + }, + { + "epoch": 0.2, + "learning_rate": 1.8469471960194103e-05, + "loss": 0.3432, + "step": 3362 + }, + { + "epoch": 0.2, + "learning_rate": 1.8468431493764635e-05, + "loss": 0.3577, + "step": 3363 + }, + { + "epoch": 0.2, + "learning_rate": 1.846739070312398e-05, + "loss": 0.3479, + "step": 3364 + }, + { + "epoch": 0.2, + "learning_rate": 1.8466349588311972e-05, + "loss": 0.3655, + "step": 3365 + }, + { + "epoch": 0.2, + "learning_rate": 1.846530814936848e-05, + "loss": 0.3454, + "step": 3366 + }, + { + "epoch": 0.2, + "learning_rate": 1.8464266386333373e-05, + "loss": 0.3504, + "step": 3367 + }, + { + "epoch": 0.2, + "learning_rate": 1.846322429924653e-05, + "loss": 0.3619, + "step": 3368 + }, + { + "epoch": 0.2, + "learning_rate": 1.846218188814785e-05, + "loss": 0.3479, + "step": 3369 + }, + { + "epoch": 0.2, + "learning_rate": 1.8461139153077246e-05, + "loss": 0.3814, + "step": 3370 + }, + { + "epoch": 0.2, + "learning_rate": 1.8460096094074632e-05, + "loss": 0.3424, + "step": 3371 + }, + { + "epoch": 0.2, + "learning_rate": 1.8459052711179945e-05, + "loss": 0.3496, + "step": 3372 + }, + { + "epoch": 0.2, + "learning_rate": 1.8458009004433126e-05, + "loss": 0.3594, + "step": 3373 + }, + { + "epoch": 0.2, + "learning_rate": 1.845696497387414e-05, + "loss": 0.3315, + "step": 3374 + }, + { + "epoch": 0.2, + "learning_rate": 1.845592061954295e-05, + "loss": 0.3638, + "step": 3375 + }, + { + "epoch": 0.2, + "learning_rate": 1.8454875941479543e-05, + "loss": 0.3597, + "step": 3376 + }, + { + "epoch": 0.2, + "learning_rate": 1.8453830939723913e-05, + "loss": 0.3677, + "step": 3377 + }, + { + "epoch": 0.2, + "learning_rate": 1.8452785614316072e-05, + "loss": 0.3454, + "step": 3378 + }, + { + "epoch": 0.2, + "learning_rate": 1.8451739965296028e-05, + "loss": 0.3502, + "step": 3379 + }, + { + "epoch": 0.2, + "learning_rate": 1.8450693992703827e-05, + "loss": 0.3588, + "step": 3380 + }, + { + "epoch": 0.2, + "learning_rate": 1.844964769657951e-05, + "loss": 0.353, + "step": 3381 + }, + { + "epoch": 0.2, + "learning_rate": 1.844860107696313e-05, + "loss": 0.3823, + "step": 3382 + }, + { + "epoch": 0.2, + "learning_rate": 1.8447554133894757e-05, + "loss": 0.3664, + "step": 3383 + }, + { + "epoch": 0.2, + "learning_rate": 1.8446506867414474e-05, + "loss": 0.3633, + "step": 3384 + }, + { + "epoch": 0.2, + "learning_rate": 1.844545927756238e-05, + "loss": 0.3594, + "step": 3385 + }, + { + "epoch": 0.2, + "learning_rate": 1.8444411364378574e-05, + "loss": 0.3591, + "step": 3386 + }, + { + "epoch": 0.2, + "learning_rate": 1.844336312790318e-05, + "loss": 0.3691, + "step": 3387 + }, + { + "epoch": 0.2, + "learning_rate": 1.8442314568176324e-05, + "loss": 0.3538, + "step": 3388 + }, + { + "epoch": 0.2, + "learning_rate": 1.844126568523816e-05, + "loss": 0.3465, + "step": 3389 + }, + { + "epoch": 0.2, + "learning_rate": 1.8440216479128832e-05, + "loss": 0.3373, + "step": 3390 + }, + { + "epoch": 0.2, + "learning_rate": 1.8439166949888515e-05, + "loss": 0.3382, + "step": 3391 + }, + { + "epoch": 0.2, + "learning_rate": 1.8438117097557394e-05, + "loss": 0.3446, + "step": 3392 + }, + { + "epoch": 0.2, + "learning_rate": 1.8437066922175653e-05, + "loss": 0.3915, + "step": 3393 + }, + { + "epoch": 0.21, + "learning_rate": 1.8436016423783504e-05, + "loss": 0.3602, + "step": 3394 + }, + { + "epoch": 0.21, + "learning_rate": 1.8434965602421162e-05, + "loss": 0.3686, + "step": 3395 + }, + { + "epoch": 0.21, + "learning_rate": 1.843391445812886e-05, + "loss": 0.3569, + "step": 3396 + }, + { + "epoch": 0.21, + "learning_rate": 1.8432862990946837e-05, + "loss": 0.3488, + "step": 3397 + }, + { + "epoch": 0.21, + "learning_rate": 1.843181120091535e-05, + "loss": 0.3334, + "step": 3398 + }, + { + "epoch": 0.21, + "learning_rate": 1.8430759088074666e-05, + "loss": 0.3878, + "step": 3399 + }, + { + "epoch": 0.21, + "learning_rate": 1.842970665246507e-05, + "loss": 0.3549, + "step": 3400 + }, + { + "epoch": 0.21, + "learning_rate": 1.8428653894126848e-05, + "loss": 0.377, + "step": 3401 + }, + { + "epoch": 0.21, + "learning_rate": 1.8427600813100302e-05, + "loss": 0.3371, + "step": 3402 + }, + { + "epoch": 0.21, + "learning_rate": 1.8426547409425758e-05, + "loss": 0.3432, + "step": 3403 + }, + { + "epoch": 0.21, + "learning_rate": 1.8425493683143537e-05, + "loss": 0.3516, + "step": 3404 + }, + { + "epoch": 0.21, + "learning_rate": 1.8424439634293985e-05, + "loss": 0.3276, + "step": 3405 + }, + { + "epoch": 0.21, + "learning_rate": 1.8423385262917454e-05, + "loss": 0.3491, + "step": 3406 + }, + { + "epoch": 0.21, + "learning_rate": 1.8422330569054315e-05, + "loss": 0.3666, + "step": 3407 + }, + { + "epoch": 0.21, + "learning_rate": 1.8421275552744937e-05, + "loss": 0.3689, + "step": 3408 + }, + { + "epoch": 0.21, + "learning_rate": 1.842022021402972e-05, + "loss": 0.3636, + "step": 3409 + }, + { + "epoch": 0.21, + "learning_rate": 1.8419164552949064e-05, + "loss": 0.3627, + "step": 3410 + }, + { + "epoch": 0.21, + "learning_rate": 1.841810856954338e-05, + "loss": 0.351, + "step": 3411 + }, + { + "epoch": 0.21, + "learning_rate": 1.8417052263853107e-05, + "loss": 0.3993, + "step": 3412 + }, + { + "epoch": 0.21, + "learning_rate": 1.8415995635918677e-05, + "loss": 0.368, + "step": 3413 + }, + { + "epoch": 0.21, + "learning_rate": 1.841493868578054e-05, + "loss": 0.3672, + "step": 3414 + }, + { + "epoch": 0.21, + "learning_rate": 1.841388141347917e-05, + "loss": 0.3521, + "step": 3415 + }, + { + "epoch": 0.21, + "learning_rate": 1.841282381905504e-05, + "loss": 0.3619, + "step": 3416 + }, + { + "epoch": 0.21, + "learning_rate": 1.8411765902548638e-05, + "loss": 0.3237, + "step": 3417 + }, + { + "epoch": 0.21, + "learning_rate": 1.8410707664000463e-05, + "loss": 0.3597, + "step": 3418 + }, + { + "epoch": 0.21, + "learning_rate": 1.8409649103451042e-05, + "loss": 0.3541, + "step": 3419 + }, + { + "epoch": 0.21, + "learning_rate": 1.840859022094089e-05, + "loss": 0.3331, + "step": 3420 + }, + { + "epoch": 0.21, + "learning_rate": 1.8407531016510547e-05, + "loss": 0.3418, + "step": 3421 + }, + { + "epoch": 0.21, + "learning_rate": 1.840647149020057e-05, + "loss": 0.3552, + "step": 3422 + }, + { + "epoch": 0.21, + "learning_rate": 1.8405411642051518e-05, + "loss": 0.3655, + "step": 3423 + }, + { + "epoch": 0.21, + "learning_rate": 1.840435147210397e-05, + "loss": 0.3345, + "step": 3424 + }, + { + "epoch": 0.21, + "learning_rate": 1.840329098039851e-05, + "loss": 0.3739, + "step": 3425 + }, + { + "epoch": 0.21, + "learning_rate": 1.8402230166975744e-05, + "loss": 0.3817, + "step": 3426 + }, + { + "epoch": 0.21, + "learning_rate": 1.840116903187628e-05, + "loss": 0.3302, + "step": 3427 + }, + { + "epoch": 0.21, + "learning_rate": 1.8400107575140748e-05, + "loss": 0.373, + "step": 3428 + }, + { + "epoch": 0.21, + "learning_rate": 1.839904579680978e-05, + "loss": 0.3803, + "step": 3429 + }, + { + "epoch": 0.21, + "learning_rate": 1.8397983696924027e-05, + "loss": 0.3644, + "step": 3430 + }, + { + "epoch": 0.21, + "learning_rate": 1.839692127552416e-05, + "loss": 0.3644, + "step": 3431 + }, + { + "epoch": 0.21, + "learning_rate": 1.839585853265084e-05, + "loss": 0.3655, + "step": 3432 + }, + { + "epoch": 0.21, + "learning_rate": 1.8394795468344762e-05, + "loss": 0.3532, + "step": 3433 + }, + { + "epoch": 0.21, + "learning_rate": 1.839373208264662e-05, + "loss": 0.3504, + "step": 3434 + }, + { + "epoch": 0.21, + "learning_rate": 1.8392668375597133e-05, + "loss": 0.3585, + "step": 3435 + }, + { + "epoch": 0.21, + "learning_rate": 1.839160434723702e-05, + "loss": 0.3636, + "step": 3436 + }, + { + "epoch": 0.21, + "learning_rate": 1.8390539997607016e-05, + "loss": 0.3365, + "step": 3437 + }, + { + "epoch": 0.21, + "learning_rate": 1.838947532674787e-05, + "loss": 0.3758, + "step": 3438 + }, + { + "epoch": 0.21, + "learning_rate": 1.8388410334700345e-05, + "loss": 0.3658, + "step": 3439 + }, + { + "epoch": 0.21, + "learning_rate": 1.838734502150521e-05, + "loss": 0.3733, + "step": 3440 + }, + { + "epoch": 0.21, + "learning_rate": 1.838627938720325e-05, + "loss": 0.3449, + "step": 3441 + }, + { + "epoch": 0.21, + "learning_rate": 1.8385213431835266e-05, + "loss": 0.3624, + "step": 3442 + }, + { + "epoch": 0.21, + "learning_rate": 1.838414715544207e-05, + "loss": 0.3532, + "step": 3443 + }, + { + "epoch": 0.21, + "learning_rate": 1.8383080558064478e-05, + "loss": 0.3396, + "step": 3444 + }, + { + "epoch": 0.21, + "learning_rate": 1.8382013639743324e-05, + "loss": 0.3652, + "step": 3445 + }, + { + "epoch": 0.21, + "learning_rate": 1.8380946400519463e-05, + "loss": 0.3251, + "step": 3446 + }, + { + "epoch": 0.21, + "learning_rate": 1.8379878840433744e-05, + "loss": 0.3555, + "step": 3447 + }, + { + "epoch": 0.21, + "learning_rate": 1.8378810959527046e-05, + "loss": 0.3477, + "step": 3448 + }, + { + "epoch": 0.21, + "learning_rate": 1.8377742757840246e-05, + "loss": 0.3666, + "step": 3449 + }, + { + "epoch": 0.21, + "learning_rate": 1.837667423541424e-05, + "loss": 0.3792, + "step": 3450 + }, + { + "epoch": 0.21, + "learning_rate": 1.8375605392289944e-05, + "loss": 0.361, + "step": 3451 + }, + { + "epoch": 0.21, + "learning_rate": 1.837453622850827e-05, + "loss": 0.3552, + "step": 3452 + }, + { + "epoch": 0.21, + "learning_rate": 1.8373466744110157e-05, + "loss": 0.3613, + "step": 3453 + }, + { + "epoch": 0.21, + "learning_rate": 1.8372396939136545e-05, + "loss": 0.3697, + "step": 3454 + }, + { + "epoch": 0.21, + "learning_rate": 1.8371326813628393e-05, + "loss": 0.3597, + "step": 3455 + }, + { + "epoch": 0.21, + "learning_rate": 1.8370256367626667e-05, + "loss": 0.3574, + "step": 3456 + }, + { + "epoch": 0.21, + "learning_rate": 1.8369185601172353e-05, + "loss": 0.3432, + "step": 3457 + }, + { + "epoch": 0.21, + "learning_rate": 1.8368114514306445e-05, + "loss": 0.3764, + "step": 3458 + }, + { + "epoch": 0.21, + "learning_rate": 1.8367043107069948e-05, + "loss": 0.3535, + "step": 3459 + }, + { + "epoch": 0.21, + "learning_rate": 1.836597137950388e-05, + "loss": 0.3585, + "step": 3460 + }, + { + "epoch": 0.21, + "learning_rate": 1.8364899331649274e-05, + "loss": 0.3658, + "step": 3461 + }, + { + "epoch": 0.21, + "learning_rate": 1.836382696354717e-05, + "loss": 0.3454, + "step": 3462 + }, + { + "epoch": 0.21, + "learning_rate": 1.8362754275238624e-05, + "loss": 0.3518, + "step": 3463 + }, + { + "epoch": 0.21, + "learning_rate": 1.83616812667647e-05, + "loss": 0.3717, + "step": 3464 + }, + { + "epoch": 0.21, + "learning_rate": 1.8360607938166487e-05, + "loss": 0.3585, + "step": 3465 + }, + { + "epoch": 0.21, + "learning_rate": 1.835953428948507e-05, + "loss": 0.3705, + "step": 3466 + }, + { + "epoch": 0.21, + "learning_rate": 1.8358460320761556e-05, + "loss": 0.3449, + "step": 3467 + }, + { + "epoch": 0.21, + "learning_rate": 1.8357386032037063e-05, + "loss": 0.3823, + "step": 3468 + }, + { + "epoch": 0.21, + "learning_rate": 1.8356311423352714e-05, + "loss": 0.329, + "step": 3469 + }, + { + "epoch": 0.21, + "learning_rate": 1.8355236494749653e-05, + "loss": 0.3371, + "step": 3470 + }, + { + "epoch": 0.21, + "learning_rate": 1.8354161246269037e-05, + "loss": 0.3499, + "step": 3471 + }, + { + "epoch": 0.21, + "learning_rate": 1.8353085677952026e-05, + "loss": 0.3689, + "step": 3472 + }, + { + "epoch": 0.21, + "learning_rate": 1.83520097898398e-05, + "loss": 0.3665, + "step": 3473 + }, + { + "epoch": 0.21, + "learning_rate": 1.835093358197355e-05, + "loss": 0.3504, + "step": 3474 + }, + { + "epoch": 0.21, + "learning_rate": 1.8349857054394475e-05, + "loss": 0.3418, + "step": 3475 + }, + { + "epoch": 0.21, + "learning_rate": 1.8348780207143796e-05, + "loss": 0.3451, + "step": 3476 + }, + { + "epoch": 0.21, + "learning_rate": 1.8347703040262734e-05, + "loss": 0.3803, + "step": 3477 + }, + { + "epoch": 0.21, + "learning_rate": 1.834662555379253e-05, + "loss": 0.3781, + "step": 3478 + }, + { + "epoch": 0.21, + "learning_rate": 1.8345547747774436e-05, + "loss": 0.3839, + "step": 3479 + }, + { + "epoch": 0.21, + "learning_rate": 1.834446962224971e-05, + "loss": 0.3418, + "step": 3480 + }, + { + "epoch": 0.21, + "learning_rate": 1.8343391177259635e-05, + "loss": 0.3717, + "step": 3481 + }, + { + "epoch": 0.21, + "learning_rate": 1.8342312412845497e-05, + "loss": 0.3398, + "step": 3482 + }, + { + "epoch": 0.21, + "learning_rate": 1.8341233329048593e-05, + "loss": 0.3563, + "step": 3483 + }, + { + "epoch": 0.21, + "learning_rate": 1.8340153925910237e-05, + "loss": 0.3256, + "step": 3484 + }, + { + "epoch": 0.21, + "learning_rate": 1.833907420347176e-05, + "loss": 0.3797, + "step": 3485 + }, + { + "epoch": 0.21, + "learning_rate": 1.8337994161774487e-05, + "loss": 0.3231, + "step": 3486 + }, + { + "epoch": 0.21, + "learning_rate": 1.8336913800859775e-05, + "loss": 0.3742, + "step": 3487 + }, + { + "epoch": 0.21, + "learning_rate": 1.8335833120768984e-05, + "loss": 0.3265, + "step": 3488 + }, + { + "epoch": 0.21, + "learning_rate": 1.8334752121543483e-05, + "loss": 0.3742, + "step": 3489 + }, + { + "epoch": 0.21, + "learning_rate": 1.8333670803224666e-05, + "loss": 0.3412, + "step": 3490 + }, + { + "epoch": 0.21, + "learning_rate": 1.8332589165853924e-05, + "loss": 0.3421, + "step": 3491 + }, + { + "epoch": 0.21, + "learning_rate": 1.833150720947267e-05, + "loss": 0.3541, + "step": 3492 + }, + { + "epoch": 0.21, + "learning_rate": 1.8330424934122326e-05, + "loss": 0.3683, + "step": 3493 + }, + { + "epoch": 0.21, + "learning_rate": 1.832934233984433e-05, + "loss": 0.3507, + "step": 3494 + }, + { + "epoch": 0.21, + "learning_rate": 1.8328259426680118e-05, + "loss": 0.3591, + "step": 3495 + }, + { + "epoch": 0.21, + "learning_rate": 1.832717619467116e-05, + "loss": 0.3599, + "step": 3496 + }, + { + "epoch": 0.21, + "learning_rate": 1.8326092643858924e-05, + "loss": 0.3624, + "step": 3497 + }, + { + "epoch": 0.21, + "learning_rate": 1.8325008774284892e-05, + "loss": 0.3711, + "step": 3498 + }, + { + "epoch": 0.21, + "learning_rate": 1.832392458599056e-05, + "loss": 0.3641, + "step": 3499 + }, + { + "epoch": 0.21, + "learning_rate": 1.832284007901744e-05, + "loss": 0.3488, + "step": 3500 + }, + { + "epoch": 0.21, + "learning_rate": 1.832175525340704e-05, + "loss": 0.3424, + "step": 3501 + }, + { + "epoch": 0.21, + "learning_rate": 1.8320670109200907e-05, + "loss": 0.3641, + "step": 3502 + }, + { + "epoch": 0.21, + "learning_rate": 1.831958464644058e-05, + "loss": 0.3521, + "step": 3503 + }, + { + "epoch": 0.21, + "learning_rate": 1.831849886516761e-05, + "loss": 0.3507, + "step": 3504 + }, + { + "epoch": 0.21, + "learning_rate": 1.8317412765423574e-05, + "loss": 0.3386, + "step": 3505 + }, + { + "epoch": 0.21, + "learning_rate": 1.8316326347250048e-05, + "loss": 0.353, + "step": 3506 + }, + { + "epoch": 0.21, + "learning_rate": 1.8315239610688624e-05, + "loss": 0.3686, + "step": 3507 + }, + { + "epoch": 0.21, + "learning_rate": 1.831415255578091e-05, + "loss": 0.365, + "step": 3508 + }, + { + "epoch": 0.21, + "learning_rate": 1.831306518256853e-05, + "loss": 0.363, + "step": 3509 + }, + { + "epoch": 0.21, + "learning_rate": 1.8311977491093098e-05, + "loss": 0.3686, + "step": 3510 + }, + { + "epoch": 0.21, + "learning_rate": 1.831088948139627e-05, + "loss": 0.3407, + "step": 3511 + }, + { + "epoch": 0.21, + "learning_rate": 1.8309801153519692e-05, + "loss": 0.3613, + "step": 3512 + }, + { + "epoch": 0.21, + "learning_rate": 1.8308712507505038e-05, + "loss": 0.3633, + "step": 3513 + }, + { + "epoch": 0.21, + "learning_rate": 1.8307623543393974e-05, + "loss": 0.3717, + "step": 3514 + }, + { + "epoch": 0.21, + "learning_rate": 1.8306534261228205e-05, + "loss": 0.358, + "step": 3515 + }, + { + "epoch": 0.21, + "learning_rate": 1.8305444661049425e-05, + "loss": 0.3666, + "step": 3516 + }, + { + "epoch": 0.21, + "learning_rate": 1.830435474289935e-05, + "loss": 0.3521, + "step": 3517 + }, + { + "epoch": 0.21, + "learning_rate": 1.8303264506819714e-05, + "loss": 0.3326, + "step": 3518 + }, + { + "epoch": 0.21, + "learning_rate": 1.8302173952852243e-05, + "loss": 0.3594, + "step": 3519 + }, + { + "epoch": 0.21, + "learning_rate": 1.8301083081038702e-05, + "loss": 0.3574, + "step": 3520 + }, + { + "epoch": 0.21, + "learning_rate": 1.8299991891420848e-05, + "loss": 0.3231, + "step": 3521 + }, + { + "epoch": 0.21, + "learning_rate": 1.8298900384040458e-05, + "loss": 0.3496, + "step": 3522 + }, + { + "epoch": 0.21, + "learning_rate": 1.829780855893932e-05, + "loss": 0.3393, + "step": 3523 + }, + { + "epoch": 0.21, + "learning_rate": 1.829671641615923e-05, + "loss": 0.3262, + "step": 3524 + }, + { + "epoch": 0.21, + "learning_rate": 1.8295623955742007e-05, + "loss": 0.3697, + "step": 3525 + }, + { + "epoch": 0.21, + "learning_rate": 1.8294531177729475e-05, + "loss": 0.3206, + "step": 3526 + }, + { + "epoch": 0.21, + "learning_rate": 1.8293438082163464e-05, + "loss": 0.3218, + "step": 3527 + }, + { + "epoch": 0.21, + "learning_rate": 1.829234466908583e-05, + "loss": 0.3429, + "step": 3528 + }, + { + "epoch": 0.21, + "learning_rate": 1.829125093853843e-05, + "loss": 0.3234, + "step": 3529 + }, + { + "epoch": 0.21, + "learning_rate": 1.829015689056314e-05, + "loss": 0.3493, + "step": 3530 + }, + { + "epoch": 0.21, + "learning_rate": 1.828906252520184e-05, + "loss": 0.3373, + "step": 3531 + }, + { + "epoch": 0.21, + "learning_rate": 1.8287967842496437e-05, + "loss": 0.3658, + "step": 3532 + }, + { + "epoch": 0.21, + "learning_rate": 1.8286872842488832e-05, + "loss": 0.3541, + "step": 3533 + }, + { + "epoch": 0.21, + "learning_rate": 1.8285777525220947e-05, + "loss": 0.3359, + "step": 3534 + }, + { + "epoch": 0.21, + "learning_rate": 1.828468189073472e-05, + "loss": 0.3655, + "step": 3535 + }, + { + "epoch": 0.21, + "learning_rate": 1.8283585939072094e-05, + "loss": 0.3636, + "step": 3536 + }, + { + "epoch": 0.21, + "learning_rate": 1.828248967027503e-05, + "loss": 0.3502, + "step": 3537 + }, + { + "epoch": 0.21, + "learning_rate": 1.8281393084385496e-05, + "loss": 0.3563, + "step": 3538 + }, + { + "epoch": 0.21, + "learning_rate": 1.828029618144548e-05, + "loss": 0.3479, + "step": 3539 + }, + { + "epoch": 0.21, + "learning_rate": 1.8279198961496965e-05, + "loss": 0.3583, + "step": 3540 + }, + { + "epoch": 0.21, + "learning_rate": 1.8278101424581966e-05, + "loss": 0.3449, + "step": 3541 + }, + { + "epoch": 0.21, + "learning_rate": 1.8277003570742502e-05, + "loss": 0.3602, + "step": 3542 + }, + { + "epoch": 0.21, + "learning_rate": 1.8275905400020602e-05, + "loss": 0.3304, + "step": 3543 + }, + { + "epoch": 0.21, + "learning_rate": 1.827480691245831e-05, + "loss": 0.3446, + "step": 3544 + }, + { + "epoch": 0.21, + "learning_rate": 1.8273708108097677e-05, + "loss": 0.3597, + "step": 3545 + }, + { + "epoch": 0.21, + "learning_rate": 1.8272608986980778e-05, + "loss": 0.3605, + "step": 3546 + }, + { + "epoch": 0.21, + "learning_rate": 1.8271509549149685e-05, + "loss": 0.3359, + "step": 3547 + }, + { + "epoch": 0.21, + "learning_rate": 1.82704097946465e-05, + "loss": 0.3418, + "step": 3548 + }, + { + "epoch": 0.21, + "learning_rate": 1.826930972351331e-05, + "loss": 0.3708, + "step": 3549 + }, + { + "epoch": 0.21, + "learning_rate": 1.826820933579225e-05, + "loss": 0.3895, + "step": 3550 + }, + { + "epoch": 0.21, + "learning_rate": 1.8267108631525433e-05, + "loss": 0.358, + "step": 3551 + }, + { + "epoch": 0.21, + "learning_rate": 1.8266007610755004e-05, + "loss": 0.3652, + "step": 3552 + }, + { + "epoch": 0.21, + "learning_rate": 1.8264906273523122e-05, + "loss": 0.3691, + "step": 3553 + }, + { + "epoch": 0.21, + "learning_rate": 1.8263804619871942e-05, + "loss": 0.3552, + "step": 3554 + }, + { + "epoch": 0.21, + "learning_rate": 1.8262702649843646e-05, + "loss": 0.3915, + "step": 3555 + }, + { + "epoch": 0.21, + "learning_rate": 1.8261600363480416e-05, + "loss": 0.3555, + "step": 3556 + }, + { + "epoch": 0.21, + "learning_rate": 1.826049776082446e-05, + "loss": 0.3354, + "step": 3557 + }, + { + "epoch": 0.21, + "learning_rate": 1.8259394841917987e-05, + "loss": 0.3736, + "step": 3558 + }, + { + "epoch": 0.22, + "learning_rate": 1.8258291606803222e-05, + "loss": 0.387, + "step": 3559 + }, + { + "epoch": 0.22, + "learning_rate": 1.8257188055522406e-05, + "loss": 0.3276, + "step": 3560 + }, + { + "epoch": 0.22, + "learning_rate": 1.8256084188117784e-05, + "loss": 0.3552, + "step": 3561 + }, + { + "epoch": 0.22, + "learning_rate": 1.8254980004631616e-05, + "loss": 0.3772, + "step": 3562 + }, + { + "epoch": 0.22, + "learning_rate": 1.825387550510618e-05, + "loss": 0.3809, + "step": 3563 + }, + { + "epoch": 0.22, + "learning_rate": 1.825277068958376e-05, + "loss": 0.3371, + "step": 3564 + }, + { + "epoch": 0.22, + "learning_rate": 1.825166555810665e-05, + "loss": 0.3641, + "step": 3565 + }, + { + "epoch": 0.22, + "learning_rate": 1.8250560110717162e-05, + "loss": 0.3839, + "step": 3566 + }, + { + "epoch": 0.22, + "learning_rate": 1.8249454347457617e-05, + "loss": 0.3627, + "step": 3567 + }, + { + "epoch": 0.22, + "learning_rate": 1.8248348268370352e-05, + "loss": 0.3304, + "step": 3568 + }, + { + "epoch": 0.22, + "learning_rate": 1.8247241873497708e-05, + "loss": 0.3373, + "step": 3569 + }, + { + "epoch": 0.22, + "learning_rate": 1.8246135162882044e-05, + "loss": 0.3496, + "step": 3570 + }, + { + "epoch": 0.22, + "learning_rate": 1.8245028136565732e-05, + "loss": 0.3563, + "step": 3571 + }, + { + "epoch": 0.22, + "learning_rate": 1.824392079459116e-05, + "loss": 0.3304, + "step": 3572 + }, + { + "epoch": 0.22, + "learning_rate": 1.8242813137000708e-05, + "loss": 0.3438, + "step": 3573 + }, + { + "epoch": 0.22, + "learning_rate": 1.8241705163836794e-05, + "loss": 0.3644, + "step": 3574 + }, + { + "epoch": 0.22, + "learning_rate": 1.8240596875141832e-05, + "loss": 0.3504, + "step": 3575 + }, + { + "epoch": 0.22, + "learning_rate": 1.823948827095825e-05, + "loss": 0.3898, + "step": 3576 + }, + { + "epoch": 0.22, + "learning_rate": 1.8238379351328498e-05, + "loss": 0.3415, + "step": 3577 + }, + { + "epoch": 0.22, + "learning_rate": 1.8237270116295022e-05, + "loss": 0.3535, + "step": 3578 + }, + { + "epoch": 0.22, + "learning_rate": 1.8236160565900297e-05, + "loss": 0.3345, + "step": 3579 + }, + { + "epoch": 0.22, + "learning_rate": 1.8235050700186797e-05, + "loss": 0.3493, + "step": 3580 + }, + { + "epoch": 0.22, + "learning_rate": 1.823394051919701e-05, + "loss": 0.3597, + "step": 3581 + }, + { + "epoch": 0.22, + "learning_rate": 1.8232830022973446e-05, + "loss": 0.3613, + "step": 3582 + }, + { + "epoch": 0.22, + "learning_rate": 1.8231719211558617e-05, + "loss": 0.3253, + "step": 3583 + }, + { + "epoch": 0.22, + "learning_rate": 1.8230608084995047e-05, + "loss": 0.3426, + "step": 3584 + }, + { + "epoch": 0.22, + "learning_rate": 1.822949664332528e-05, + "loss": 0.3613, + "step": 3585 + }, + { + "epoch": 0.22, + "learning_rate": 1.8228384886591862e-05, + "loss": 0.3705, + "step": 3586 + }, + { + "epoch": 0.22, + "learning_rate": 1.822727281483736e-05, + "loss": 0.3959, + "step": 3587 + }, + { + "epoch": 0.22, + "learning_rate": 1.822616042810435e-05, + "loss": 0.3703, + "step": 3588 + }, + { + "epoch": 0.22, + "learning_rate": 1.8225047726435416e-05, + "loss": 0.3591, + "step": 3589 + }, + { + "epoch": 0.22, + "learning_rate": 1.822393470987316e-05, + "loss": 0.3566, + "step": 3590 + }, + { + "epoch": 0.22, + "learning_rate": 1.822282137846019e-05, + "loss": 0.3719, + "step": 3591 + }, + { + "epoch": 0.22, + "learning_rate": 1.8221707732239137e-05, + "loss": 0.3491, + "step": 3592 + }, + { + "epoch": 0.22, + "learning_rate": 1.822059377125263e-05, + "loss": 0.3599, + "step": 3593 + }, + { + "epoch": 0.22, + "learning_rate": 1.8219479495543314e-05, + "loss": 0.3717, + "step": 3594 + }, + { + "epoch": 0.22, + "learning_rate": 1.821836490515386e-05, + "loss": 0.334, + "step": 3595 + }, + { + "epoch": 0.22, + "learning_rate": 1.821725000012693e-05, + "loss": 0.3714, + "step": 3596 + }, + { + "epoch": 0.22, + "learning_rate": 1.821613478050521e-05, + "loss": 0.3211, + "step": 3597 + }, + { + "epoch": 0.22, + "learning_rate": 1.8215019246331397e-05, + "loss": 0.3836, + "step": 3598 + }, + { + "epoch": 0.22, + "learning_rate": 1.82139033976482e-05, + "loss": 0.3666, + "step": 3599 + }, + { + "epoch": 0.22, + "learning_rate": 1.8212787234498334e-05, + "loss": 0.3086, + "step": 3600 + }, + { + "epoch": 0.22, + "learning_rate": 1.8211670756924534e-05, + "loss": 0.3677, + "step": 3601 + }, + { + "epoch": 0.22, + "learning_rate": 1.8210553964969545e-05, + "loss": 0.3513, + "step": 3602 + }, + { + "epoch": 0.22, + "learning_rate": 1.8209436858676124e-05, + "loss": 0.3443, + "step": 3603 + }, + { + "epoch": 0.22, + "learning_rate": 1.8208319438087038e-05, + "loss": 0.3485, + "step": 3604 + }, + { + "epoch": 0.22, + "learning_rate": 1.8207201703245063e-05, + "loss": 0.3256, + "step": 3605 + }, + { + "epoch": 0.22, + "learning_rate": 1.8206083654193e-05, + "loss": 0.3669, + "step": 3606 + }, + { + "epoch": 0.22, + "learning_rate": 1.8204965290973643e-05, + "loss": 0.3284, + "step": 3607 + }, + { + "epoch": 0.22, + "learning_rate": 1.8203846613629813e-05, + "loss": 0.3599, + "step": 3608 + }, + { + "epoch": 0.22, + "learning_rate": 1.8202727622204336e-05, + "loss": 0.3661, + "step": 3609 + }, + { + "epoch": 0.22, + "learning_rate": 1.8201608316740057e-05, + "loss": 0.3443, + "step": 3610 + }, + { + "epoch": 0.22, + "learning_rate": 1.8200488697279826e-05, + "loss": 0.3719, + "step": 3611 + }, + { + "epoch": 0.22, + "learning_rate": 1.819936876386651e-05, + "loss": 0.3694, + "step": 3612 + }, + { + "epoch": 0.22, + "learning_rate": 1.8198248516542974e-05, + "loss": 0.3622, + "step": 3613 + }, + { + "epoch": 0.22, + "learning_rate": 1.819712795535212e-05, + "loss": 0.3549, + "step": 3614 + }, + { + "epoch": 0.22, + "learning_rate": 1.8196007080336842e-05, + "loss": 0.3334, + "step": 3615 + }, + { + "epoch": 0.22, + "learning_rate": 1.819488589154005e-05, + "loss": 0.361, + "step": 3616 + }, + { + "epoch": 0.22, + "learning_rate": 1.8193764389004672e-05, + "loss": 0.3532, + "step": 3617 + }, + { + "epoch": 0.22, + "learning_rate": 1.8192642572773648e-05, + "loss": 0.3343, + "step": 3618 + }, + { + "epoch": 0.22, + "learning_rate": 1.819152044288992e-05, + "loss": 0.3747, + "step": 3619 + }, + { + "epoch": 0.22, + "learning_rate": 1.819039799939645e-05, + "loss": 0.3477, + "step": 3620 + }, + { + "epoch": 0.22, + "learning_rate": 1.818927524233621e-05, + "loss": 0.3789, + "step": 3621 + }, + { + "epoch": 0.22, + "learning_rate": 1.8188152171752188e-05, + "loss": 0.4037, + "step": 3622 + }, + { + "epoch": 0.22, + "learning_rate": 1.8187028787687377e-05, + "loss": 0.3156, + "step": 3623 + }, + { + "epoch": 0.22, + "learning_rate": 1.8185905090184782e-05, + "loss": 0.3292, + "step": 3624 + }, + { + "epoch": 0.22, + "learning_rate": 1.818478107928743e-05, + "loss": 0.358, + "step": 3625 + }, + { + "epoch": 0.22, + "learning_rate": 1.8183656755038352e-05, + "loss": 0.3552, + "step": 3626 + }, + { + "epoch": 0.22, + "learning_rate": 1.818253211748059e-05, + "loss": 0.3691, + "step": 3627 + }, + { + "epoch": 0.22, + "learning_rate": 1.8181407166657203e-05, + "loss": 0.3633, + "step": 3628 + }, + { + "epoch": 0.22, + "learning_rate": 1.818028190261126e-05, + "loss": 0.3443, + "step": 3629 + }, + { + "epoch": 0.22, + "learning_rate": 1.8179156325385836e-05, + "loss": 0.3694, + "step": 3630 + }, + { + "epoch": 0.22, + "learning_rate": 1.817803043502403e-05, + "loss": 0.3633, + "step": 3631 + }, + { + "epoch": 0.22, + "learning_rate": 1.8176904231568938e-05, + "loss": 0.3284, + "step": 3632 + }, + { + "epoch": 0.22, + "learning_rate": 1.8175777715063686e-05, + "loss": 0.3783, + "step": 3633 + }, + { + "epoch": 0.22, + "learning_rate": 1.8174650885551395e-05, + "loss": 0.3664, + "step": 3634 + }, + { + "epoch": 0.22, + "learning_rate": 1.817352374307521e-05, + "loss": 0.3691, + "step": 3635 + }, + { + "epoch": 0.22, + "learning_rate": 1.817239628767828e-05, + "loss": 0.3429, + "step": 3636 + }, + { + "epoch": 0.22, + "learning_rate": 1.817126851940377e-05, + "loss": 0.327, + "step": 3637 + }, + { + "epoch": 0.22, + "learning_rate": 1.8170140438294858e-05, + "loss": 0.3184, + "step": 3638 + }, + { + "epoch": 0.22, + "learning_rate": 1.8169012044394733e-05, + "loss": 0.3613, + "step": 3639 + }, + { + "epoch": 0.22, + "learning_rate": 1.816788333774659e-05, + "loss": 0.3876, + "step": 3640 + }, + { + "epoch": 0.22, + "learning_rate": 1.816675431839365e-05, + "loss": 0.3638, + "step": 3641 + }, + { + "epoch": 0.22, + "learning_rate": 1.8165624986379127e-05, + "loss": 0.3569, + "step": 3642 + }, + { + "epoch": 0.22, + "learning_rate": 1.8164495341746263e-05, + "loss": 0.4255, + "step": 3643 + }, + { + "epoch": 0.22, + "learning_rate": 1.81633653845383e-05, + "loss": 0.3446, + "step": 3644 + }, + { + "epoch": 0.22, + "learning_rate": 1.8162235114798514e-05, + "loss": 0.3666, + "step": 3645 + }, + { + "epoch": 0.22, + "learning_rate": 1.816110453257016e-05, + "loss": 0.3372, + "step": 3646 + }, + { + "epoch": 0.22, + "learning_rate": 1.8159973637896527e-05, + "loss": 0.3761, + "step": 3647 + }, + { + "epoch": 0.22, + "learning_rate": 1.815884243082091e-05, + "loss": 0.3641, + "step": 3648 + }, + { + "epoch": 0.22, + "learning_rate": 1.8157710911386622e-05, + "loss": 0.3733, + "step": 3649 + }, + { + "epoch": 0.22, + "learning_rate": 1.815657907963698e-05, + "loss": 0.3449, + "step": 3650 + }, + { + "epoch": 0.22, + "learning_rate": 1.8155446935615312e-05, + "loss": 0.3806, + "step": 3651 + }, + { + "epoch": 0.22, + "learning_rate": 1.8154314479364966e-05, + "loss": 0.3641, + "step": 3652 + }, + { + "epoch": 0.22, + "learning_rate": 1.81531817109293e-05, + "loss": 0.334, + "step": 3653 + }, + { + "epoch": 0.22, + "learning_rate": 1.8152048630351677e-05, + "loss": 0.3605, + "step": 3654 + }, + { + "epoch": 0.22, + "learning_rate": 1.815091523767548e-05, + "loss": 0.3948, + "step": 3655 + }, + { + "epoch": 0.22, + "learning_rate": 1.8149781532944096e-05, + "loss": 0.3588, + "step": 3656 + }, + { + "epoch": 0.22, + "learning_rate": 1.8148647516200932e-05, + "loss": 0.3351, + "step": 3657 + }, + { + "epoch": 0.22, + "learning_rate": 1.8147513187489402e-05, + "loss": 0.3334, + "step": 3658 + }, + { + "epoch": 0.22, + "learning_rate": 1.8146378546852937e-05, + "loss": 0.3429, + "step": 3659 + }, + { + "epoch": 0.22, + "learning_rate": 1.814524359433497e-05, + "loss": 0.3384, + "step": 3660 + }, + { + "epoch": 0.22, + "learning_rate": 1.814410832997896e-05, + "loss": 0.3588, + "step": 3661 + }, + { + "epoch": 0.22, + "learning_rate": 1.814297275382836e-05, + "loss": 0.3312, + "step": 3662 + }, + { + "epoch": 0.22, + "learning_rate": 1.814183686592666e-05, + "loss": 0.3641, + "step": 3663 + }, + { + "epoch": 0.22, + "learning_rate": 1.8140700666317334e-05, + "loss": 0.351, + "step": 3664 + }, + { + "epoch": 0.22, + "learning_rate": 1.8139564155043885e-05, + "loss": 0.3337, + "step": 3665 + }, + { + "epoch": 0.22, + "learning_rate": 1.8138427332149824e-05, + "loss": 0.3362, + "step": 3666 + }, + { + "epoch": 0.22, + "learning_rate": 1.8137290197678675e-05, + "loss": 0.3665, + "step": 3667 + }, + { + "epoch": 0.22, + "learning_rate": 1.813615275167397e-05, + "loss": 0.365, + "step": 3668 + }, + { + "epoch": 0.22, + "learning_rate": 1.813501499417926e-05, + "loss": 0.3351, + "step": 3669 + }, + { + "epoch": 0.22, + "learning_rate": 1.81338769252381e-05, + "loss": 0.3633, + "step": 3670 + }, + { + "epoch": 0.22, + "learning_rate": 1.8132738544894063e-05, + "loss": 0.3351, + "step": 3671 + }, + { + "epoch": 0.22, + "learning_rate": 1.8131599853190733e-05, + "loss": 0.3535, + "step": 3672 + }, + { + "epoch": 0.22, + "learning_rate": 1.8130460850171698e-05, + "loss": 0.3368, + "step": 3673 + }, + { + "epoch": 0.22, + "learning_rate": 1.812932153588057e-05, + "loss": 0.3599, + "step": 3674 + }, + { + "epoch": 0.22, + "learning_rate": 1.8128181910360967e-05, + "loss": 0.3599, + "step": 3675 + }, + { + "epoch": 0.22, + "learning_rate": 1.812704197365652e-05, + "loss": 0.3795, + "step": 3676 + }, + { + "epoch": 0.22, + "learning_rate": 1.8125901725810865e-05, + "loss": 0.3376, + "step": 3677 + }, + { + "epoch": 0.22, + "learning_rate": 1.8124761166867662e-05, + "loss": 0.3571, + "step": 3678 + }, + { + "epoch": 0.22, + "learning_rate": 1.8123620296870576e-05, + "loss": 0.3532, + "step": 3679 + }, + { + "epoch": 0.22, + "learning_rate": 1.812247911586328e-05, + "loss": 0.3848, + "step": 3680 + }, + { + "epoch": 0.22, + "learning_rate": 1.8121337623889473e-05, + "loss": 0.351, + "step": 3681 + }, + { + "epoch": 0.22, + "learning_rate": 1.8120195820992847e-05, + "loss": 0.3262, + "step": 3682 + }, + { + "epoch": 0.22, + "learning_rate": 1.8119053707217124e-05, + "loss": 0.3446, + "step": 3683 + }, + { + "epoch": 0.22, + "learning_rate": 1.8117911282606022e-05, + "loss": 0.3535, + "step": 3684 + }, + { + "epoch": 0.22, + "learning_rate": 1.8116768547203282e-05, + "loss": 0.377, + "step": 3685 + }, + { + "epoch": 0.22, + "learning_rate": 1.8115625501052655e-05, + "loss": 0.3471, + "step": 3686 + }, + { + "epoch": 0.22, + "learning_rate": 1.8114482144197894e-05, + "loss": 0.3597, + "step": 3687 + }, + { + "epoch": 0.22, + "learning_rate": 1.8113338476682786e-05, + "loss": 0.3415, + "step": 3688 + }, + { + "epoch": 0.22, + "learning_rate": 1.8112194498551107e-05, + "loss": 0.3577, + "step": 3689 + }, + { + "epoch": 0.22, + "learning_rate": 1.811105020984665e-05, + "loss": 0.3521, + "step": 3690 + }, + { + "epoch": 0.22, + "learning_rate": 1.8109905610613232e-05, + "loss": 0.3493, + "step": 3691 + }, + { + "epoch": 0.22, + "learning_rate": 1.810876070089467e-05, + "loss": 0.3454, + "step": 3692 + }, + { + "epoch": 0.22, + "learning_rate": 1.8107615480734796e-05, + "loss": 0.3608, + "step": 3693 + }, + { + "epoch": 0.22, + "learning_rate": 1.8106469950177456e-05, + "loss": 0.3396, + "step": 3694 + }, + { + "epoch": 0.22, + "learning_rate": 1.8105324109266505e-05, + "loss": 0.3703, + "step": 3695 + }, + { + "epoch": 0.22, + "learning_rate": 1.8104177958045813e-05, + "loss": 0.3429, + "step": 3696 + }, + { + "epoch": 0.22, + "learning_rate": 1.8103031496559256e-05, + "loss": 0.3348, + "step": 3697 + }, + { + "epoch": 0.22, + "learning_rate": 1.8101884724850728e-05, + "loss": 0.3828, + "step": 3698 + }, + { + "epoch": 0.22, + "learning_rate": 1.8100737642964135e-05, + "loss": 0.3569, + "step": 3699 + }, + { + "epoch": 0.22, + "learning_rate": 1.8099590250943392e-05, + "loss": 0.3728, + "step": 3700 + }, + { + "epoch": 0.22, + "learning_rate": 1.8098442548832428e-05, + "loss": 0.3287, + "step": 3701 + }, + { + "epoch": 0.22, + "learning_rate": 1.8097294536675173e-05, + "loss": 0.3585, + "step": 3702 + }, + { + "epoch": 0.22, + "learning_rate": 1.809614621451559e-05, + "loss": 0.3348, + "step": 3703 + }, + { + "epoch": 0.22, + "learning_rate": 1.809499758239764e-05, + "loss": 0.3655, + "step": 3704 + }, + { + "epoch": 0.22, + "learning_rate": 1.8093848640365288e-05, + "loss": 0.3214, + "step": 3705 + }, + { + "epoch": 0.22, + "learning_rate": 1.809269938846253e-05, + "loss": 0.3143, + "step": 3706 + }, + { + "epoch": 0.22, + "learning_rate": 1.8091549826733364e-05, + "loss": 0.368, + "step": 3707 + }, + { + "epoch": 0.22, + "learning_rate": 1.8090399955221802e-05, + "loss": 0.3343, + "step": 3708 + }, + { + "epoch": 0.22, + "learning_rate": 1.8089249773971863e-05, + "loss": 0.358, + "step": 3709 + }, + { + "epoch": 0.22, + "learning_rate": 1.8088099283027583e-05, + "loss": 0.3337, + "step": 3710 + }, + { + "epoch": 0.22, + "learning_rate": 1.8086948482433008e-05, + "loss": 0.3783, + "step": 3711 + }, + { + "epoch": 0.22, + "learning_rate": 1.808579737223219e-05, + "loss": 0.3493, + "step": 3712 + }, + { + "epoch": 0.22, + "learning_rate": 1.8084645952469212e-05, + "loss": 0.3429, + "step": 3713 + }, + { + "epoch": 0.22, + "learning_rate": 1.8083494223188144e-05, + "loss": 0.3407, + "step": 3714 + }, + { + "epoch": 0.22, + "learning_rate": 1.8082342184433082e-05, + "loss": 0.334, + "step": 3715 + }, + { + "epoch": 0.22, + "learning_rate": 1.8081189836248136e-05, + "loss": 0.3206, + "step": 3716 + }, + { + "epoch": 0.22, + "learning_rate": 1.808003717867742e-05, + "loss": 0.3479, + "step": 3717 + }, + { + "epoch": 0.22, + "learning_rate": 1.8078884211765064e-05, + "loss": 0.4015, + "step": 3718 + }, + { + "epoch": 0.22, + "learning_rate": 1.807773093555521e-05, + "loss": 0.3647, + "step": 3719 + }, + { + "epoch": 0.22, + "learning_rate": 1.8076577350092005e-05, + "loss": 0.3331, + "step": 3720 + }, + { + "epoch": 0.22, + "learning_rate": 1.8075423455419623e-05, + "loss": 0.3273, + "step": 3721 + }, + { + "epoch": 0.22, + "learning_rate": 1.8074269251582235e-05, + "loss": 0.3552, + "step": 3722 + }, + { + "epoch": 0.22, + "learning_rate": 1.8073114738624026e-05, + "loss": 0.3359, + "step": 3723 + }, + { + "epoch": 0.22, + "learning_rate": 1.8071959916589203e-05, + "loss": 0.3295, + "step": 3724 + }, + { + "epoch": 0.23, + "learning_rate": 1.8070804785521974e-05, + "loss": 0.3549, + "step": 3725 + }, + { + "epoch": 0.23, + "learning_rate": 1.8069649345466566e-05, + "loss": 0.3071, + "step": 3726 + }, + { + "epoch": 0.23, + "learning_rate": 1.806849359646721e-05, + "loss": 0.3609, + "step": 3727 + }, + { + "epoch": 0.23, + "learning_rate": 1.8067337538568156e-05, + "loss": 0.332, + "step": 3728 + }, + { + "epoch": 0.23, + "learning_rate": 1.8066181171813665e-05, + "loss": 0.3189, + "step": 3729 + }, + { + "epoch": 0.23, + "learning_rate": 1.8065024496248008e-05, + "loss": 0.3384, + "step": 3730 + }, + { + "epoch": 0.23, + "learning_rate": 1.8063867511915463e-05, + "loss": 0.3619, + "step": 3731 + }, + { + "epoch": 0.23, + "learning_rate": 1.806271021886033e-05, + "loss": 0.3661, + "step": 3732 + }, + { + "epoch": 0.23, + "learning_rate": 1.8061552617126913e-05, + "loss": 0.3306, + "step": 3733 + }, + { + "epoch": 0.23, + "learning_rate": 1.8060394706759532e-05, + "loss": 0.3708, + "step": 3734 + }, + { + "epoch": 0.23, + "learning_rate": 1.8059236487802513e-05, + "loss": 0.3412, + "step": 3735 + }, + { + "epoch": 0.23, + "learning_rate": 1.8058077960300205e-05, + "loss": 0.3786, + "step": 3736 + }, + { + "epoch": 0.23, + "learning_rate": 1.8056919124296957e-05, + "loss": 0.3304, + "step": 3737 + }, + { + "epoch": 0.23, + "learning_rate": 1.8055759979837138e-05, + "loss": 0.3192, + "step": 3738 + }, + { + "epoch": 0.23, + "learning_rate": 1.805460052696512e-05, + "loss": 0.3396, + "step": 3739 + }, + { + "epoch": 0.23, + "learning_rate": 1.8053440765725298e-05, + "loss": 0.377, + "step": 3740 + }, + { + "epoch": 0.23, + "learning_rate": 1.805228069616207e-05, + "loss": 0.3265, + "step": 3741 + }, + { + "epoch": 0.23, + "learning_rate": 1.805112031831985e-05, + "loss": 0.375, + "step": 3742 + }, + { + "epoch": 0.23, + "learning_rate": 1.804995963224306e-05, + "loss": 0.3753, + "step": 3743 + }, + { + "epoch": 0.23, + "learning_rate": 1.8048798637976143e-05, + "loss": 0.339, + "step": 3744 + }, + { + "epoch": 0.23, + "learning_rate": 1.8047637335563545e-05, + "loss": 0.351, + "step": 3745 + }, + { + "epoch": 0.23, + "learning_rate": 1.804647572504972e-05, + "loss": 0.3354, + "step": 3746 + }, + { + "epoch": 0.23, + "learning_rate": 1.8045313806479145e-05, + "loss": 0.3544, + "step": 3747 + }, + { + "epoch": 0.23, + "learning_rate": 1.80441515798963e-05, + "loss": 0.3521, + "step": 3748 + }, + { + "epoch": 0.23, + "learning_rate": 1.804298904534569e-05, + "loss": 0.3496, + "step": 3749 + }, + { + "epoch": 0.23, + "learning_rate": 1.804182620287181e-05, + "loss": 0.3424, + "step": 3750 + }, + { + "epoch": 0.23, + "learning_rate": 1.8040663052519187e-05, + "loss": 0.3457, + "step": 3751 + }, + { + "epoch": 0.23, + "learning_rate": 1.803949959433235e-05, + "loss": 0.3647, + "step": 3752 + }, + { + "epoch": 0.23, + "learning_rate": 1.803833582835584e-05, + "loss": 0.3438, + "step": 3753 + }, + { + "epoch": 0.23, + "learning_rate": 1.803717175463421e-05, + "loss": 0.3438, + "step": 3754 + }, + { + "epoch": 0.23, + "learning_rate": 1.803600737321203e-05, + "loss": 0.3555, + "step": 3755 + }, + { + "epoch": 0.23, + "learning_rate": 1.803484268413388e-05, + "loss": 0.3993, + "step": 3756 + }, + { + "epoch": 0.23, + "learning_rate": 1.803367768744434e-05, + "loss": 0.3683, + "step": 3757 + }, + { + "epoch": 0.23, + "learning_rate": 1.8032512383188025e-05, + "loss": 0.3351, + "step": 3758 + }, + { + "epoch": 0.23, + "learning_rate": 1.8031346771409538e-05, + "loss": 0.3945, + "step": 3759 + }, + { + "epoch": 0.23, + "learning_rate": 1.803018085215351e-05, + "loss": 0.3672, + "step": 3760 + }, + { + "epoch": 0.23, + "learning_rate": 1.8029014625464573e-05, + "loss": 0.3451, + "step": 3761 + }, + { + "epoch": 0.23, + "learning_rate": 1.8027848091387376e-05, + "loss": 0.3719, + "step": 3762 + }, + { + "epoch": 0.23, + "learning_rate": 1.802668124996658e-05, + "loss": 0.3287, + "step": 3763 + }, + { + "epoch": 0.23, + "learning_rate": 1.8025514101246865e-05, + "loss": 0.3429, + "step": 3764 + }, + { + "epoch": 0.23, + "learning_rate": 1.8024346645272906e-05, + "loss": 0.3398, + "step": 3765 + }, + { + "epoch": 0.23, + "learning_rate": 1.80231788820894e-05, + "loss": 0.3323, + "step": 3766 + }, + { + "epoch": 0.23, + "learning_rate": 1.8022010811741054e-05, + "loss": 0.3583, + "step": 3767 + }, + { + "epoch": 0.23, + "learning_rate": 1.8020842434272586e-05, + "loss": 0.3644, + "step": 3768 + }, + { + "epoch": 0.23, + "learning_rate": 1.8019673749728735e-05, + "loss": 0.3691, + "step": 3769 + }, + { + "epoch": 0.23, + "learning_rate": 1.8018504758154235e-05, + "loss": 0.3574, + "step": 3770 + }, + { + "epoch": 0.23, + "learning_rate": 1.8017335459593847e-05, + "loss": 0.3507, + "step": 3771 + }, + { + "epoch": 0.23, + "learning_rate": 1.801616585409233e-05, + "loss": 0.3387, + "step": 3772 + }, + { + "epoch": 0.23, + "learning_rate": 1.801499594169447e-05, + "loss": 0.3428, + "step": 3773 + }, + { + "epoch": 0.23, + "learning_rate": 1.8013825722445045e-05, + "loss": 0.3518, + "step": 3774 + }, + { + "epoch": 0.23, + "learning_rate": 1.801265519638887e-05, + "loss": 0.3535, + "step": 3775 + }, + { + "epoch": 0.23, + "learning_rate": 1.8011484363570754e-05, + "loss": 0.3398, + "step": 3776 + }, + { + "epoch": 0.23, + "learning_rate": 1.8010313224035517e-05, + "loss": 0.3265, + "step": 3777 + }, + { + "epoch": 0.23, + "learning_rate": 1.8009141777828e-05, + "loss": 0.3368, + "step": 3778 + }, + { + "epoch": 0.23, + "learning_rate": 1.800797002499305e-05, + "loss": 0.3783, + "step": 3779 + }, + { + "epoch": 0.23, + "learning_rate": 1.800679796557553e-05, + "loss": 0.351, + "step": 3780 + }, + { + "epoch": 0.23, + "learning_rate": 1.8005625599620303e-05, + "loss": 0.32, + "step": 3781 + }, + { + "epoch": 0.23, + "learning_rate": 1.8004452927172265e-05, + "loss": 0.3594, + "step": 3782 + }, + { + "epoch": 0.23, + "learning_rate": 1.8003279948276305e-05, + "loss": 0.3541, + "step": 3783 + }, + { + "epoch": 0.23, + "learning_rate": 1.800210666297733e-05, + "loss": 0.3571, + "step": 3784 + }, + { + "epoch": 0.23, + "learning_rate": 1.8000933071320258e-05, + "loss": 0.3398, + "step": 3785 + }, + { + "epoch": 0.23, + "learning_rate": 1.7999759173350024e-05, + "loss": 0.3574, + "step": 3786 + }, + { + "epoch": 0.23, + "learning_rate": 1.7998584969111564e-05, + "loss": 0.3168, + "step": 3787 + }, + { + "epoch": 0.23, + "learning_rate": 1.7997410458649837e-05, + "loss": 0.3412, + "step": 3788 + }, + { + "epoch": 0.23, + "learning_rate": 1.7996235642009806e-05, + "loss": 0.3814, + "step": 3789 + }, + { + "epoch": 0.23, + "learning_rate": 1.7995060519236452e-05, + "loss": 0.3906, + "step": 3790 + }, + { + "epoch": 0.23, + "learning_rate": 1.7993885090374763e-05, + "loss": 0.3471, + "step": 3791 + }, + { + "epoch": 0.23, + "learning_rate": 1.799270935546974e-05, + "loss": 0.3304, + "step": 3792 + }, + { + "epoch": 0.23, + "learning_rate": 1.799153331456639e-05, + "loss": 0.3376, + "step": 3793 + }, + { + "epoch": 0.23, + "learning_rate": 1.7990356967709737e-05, + "loss": 0.3624, + "step": 3794 + }, + { + "epoch": 0.23, + "learning_rate": 1.7989180314944833e-05, + "loss": 0.3597, + "step": 3795 + }, + { + "epoch": 0.23, + "learning_rate": 1.7988003356316708e-05, + "loss": 0.341, + "step": 3796 + }, + { + "epoch": 0.23, + "learning_rate": 1.798682609187043e-05, + "loss": 0.3265, + "step": 3797 + }, + { + "epoch": 0.23, + "learning_rate": 1.7985648521651065e-05, + "loss": 0.351, + "step": 3798 + }, + { + "epoch": 0.23, + "learning_rate": 1.7984470645703704e-05, + "loss": 0.3142, + "step": 3799 + }, + { + "epoch": 0.23, + "learning_rate": 1.7983292464073434e-05, + "loss": 0.3588, + "step": 3800 + }, + { + "epoch": 0.23, + "learning_rate": 1.7982113976805365e-05, + "loss": 0.3357, + "step": 3801 + }, + { + "epoch": 0.23, + "learning_rate": 1.7980935183944614e-05, + "loss": 0.3644, + "step": 3802 + }, + { + "epoch": 0.23, + "learning_rate": 1.7979756085536314e-05, + "loss": 0.3583, + "step": 3803 + }, + { + "epoch": 0.23, + "learning_rate": 1.7978576681625595e-05, + "loss": 0.3421, + "step": 3804 + }, + { + "epoch": 0.23, + "learning_rate": 1.7977396972257628e-05, + "loss": 0.3549, + "step": 3805 + }, + { + "epoch": 0.23, + "learning_rate": 1.7976216957477563e-05, + "loss": 0.3412, + "step": 3806 + }, + { + "epoch": 0.23, + "learning_rate": 1.7975036637330583e-05, + "loss": 0.3284, + "step": 3807 + }, + { + "epoch": 0.23, + "learning_rate": 1.7973856011861874e-05, + "loss": 0.3574, + "step": 3808 + }, + { + "epoch": 0.23, + "learning_rate": 1.797267508111664e-05, + "loss": 0.3242, + "step": 3809 + }, + { + "epoch": 0.23, + "learning_rate": 1.7971493845140086e-05, + "loss": 0.3636, + "step": 3810 + }, + { + "epoch": 0.23, + "learning_rate": 1.7970312303977434e-05, + "loss": 0.3298, + "step": 3811 + }, + { + "epoch": 0.23, + "learning_rate": 1.796913045767393e-05, + "loss": 0.3544, + "step": 3812 + }, + { + "epoch": 0.23, + "learning_rate": 1.796794830627481e-05, + "loss": 0.3691, + "step": 3813 + }, + { + "epoch": 0.23, + "learning_rate": 1.7966765849825337e-05, + "loss": 0.3407, + "step": 3814 + }, + { + "epoch": 0.23, + "learning_rate": 1.7965583088370784e-05, + "loss": 0.3605, + "step": 3815 + }, + { + "epoch": 0.23, + "learning_rate": 1.7964400021956426e-05, + "loss": 0.3591, + "step": 3816 + }, + { + "epoch": 0.23, + "learning_rate": 1.796321665062756e-05, + "loss": 0.3641, + "step": 3817 + }, + { + "epoch": 0.23, + "learning_rate": 1.7962032974429487e-05, + "loss": 0.3479, + "step": 3818 + }, + { + "epoch": 0.23, + "learning_rate": 1.7960848993407532e-05, + "loss": 0.3451, + "step": 3819 + }, + { + "epoch": 0.23, + "learning_rate": 1.7959664707607016e-05, + "loss": 0.3544, + "step": 3820 + }, + { + "epoch": 0.23, + "learning_rate": 1.7958480117073282e-05, + "loss": 0.356, + "step": 3821 + }, + { + "epoch": 0.23, + "learning_rate": 1.795729522185168e-05, + "loss": 0.3521, + "step": 3822 + }, + { + "epoch": 0.23, + "learning_rate": 1.7956110021987572e-05, + "loss": 0.339, + "step": 3823 + }, + { + "epoch": 0.23, + "learning_rate": 1.795492451752634e-05, + "loss": 0.3775, + "step": 3824 + }, + { + "epoch": 0.23, + "learning_rate": 1.795373870851336e-05, + "loss": 0.3781, + "step": 3825 + }, + { + "epoch": 0.23, + "learning_rate": 1.7952552594994042e-05, + "loss": 0.351, + "step": 3826 + }, + { + "epoch": 0.23, + "learning_rate": 1.7951366177013788e-05, + "loss": 0.3496, + "step": 3827 + }, + { + "epoch": 0.23, + "learning_rate": 1.795017945461802e-05, + "loss": 0.3549, + "step": 3828 + }, + { + "epoch": 0.23, + "learning_rate": 1.7948992427852173e-05, + "loss": 0.3382, + "step": 3829 + }, + { + "epoch": 0.23, + "learning_rate": 1.7947805096761698e-05, + "loss": 0.38, + "step": 3830 + }, + { + "epoch": 0.23, + "learning_rate": 1.7946617461392038e-05, + "loss": 0.3786, + "step": 3831 + }, + { + "epoch": 0.23, + "learning_rate": 1.7945429521788672e-05, + "loss": 0.3357, + "step": 3832 + }, + { + "epoch": 0.23, + "learning_rate": 1.794424127799708e-05, + "loss": 0.3803, + "step": 3833 + }, + { + "epoch": 0.23, + "learning_rate": 1.7943052730062745e-05, + "loss": 0.344, + "step": 3834 + }, + { + "epoch": 0.23, + "learning_rate": 1.794186387803118e-05, + "loss": 0.3641, + "step": 3835 + }, + { + "epoch": 0.23, + "learning_rate": 1.7940674721947894e-05, + "loss": 0.3223, + "step": 3836 + }, + { + "epoch": 0.23, + "learning_rate": 1.7939485261858412e-05, + "loss": 0.3602, + "step": 3837 + }, + { + "epoch": 0.23, + "learning_rate": 1.7938295497808277e-05, + "loss": 0.3365, + "step": 3838 + }, + { + "epoch": 0.23, + "learning_rate": 1.7937105429843036e-05, + "loss": 0.3292, + "step": 3839 + }, + { + "epoch": 0.23, + "learning_rate": 1.793591505800825e-05, + "loss": 0.3371, + "step": 3840 + }, + { + "epoch": 0.23, + "learning_rate": 1.7934724382349495e-05, + "loss": 0.3602, + "step": 3841 + }, + { + "epoch": 0.23, + "learning_rate": 1.7933533402912354e-05, + "loss": 0.3496, + "step": 3842 + }, + { + "epoch": 0.23, + "learning_rate": 1.793234211974242e-05, + "loss": 0.322, + "step": 3843 + }, + { + "epoch": 0.23, + "learning_rate": 1.7931150532885305e-05, + "loss": 0.3594, + "step": 3844 + }, + { + "epoch": 0.23, + "learning_rate": 1.7929958642386627e-05, + "loss": 0.3647, + "step": 3845 + }, + { + "epoch": 0.23, + "learning_rate": 1.792876644829202e-05, + "loss": 0.3728, + "step": 3846 + }, + { + "epoch": 0.23, + "learning_rate": 1.792757395064712e-05, + "loss": 0.3502, + "step": 3847 + }, + { + "epoch": 0.23, + "learning_rate": 1.7926381149497588e-05, + "loss": 0.3583, + "step": 3848 + }, + { + "epoch": 0.23, + "learning_rate": 1.7925188044889087e-05, + "loss": 0.3739, + "step": 3849 + }, + { + "epoch": 0.23, + "learning_rate": 1.79239946368673e-05, + "loss": 0.3664, + "step": 3850 + }, + { + "epoch": 0.23, + "learning_rate": 1.7922800925477906e-05, + "loss": 0.3722, + "step": 3851 + }, + { + "epoch": 0.23, + "learning_rate": 1.7921606910766614e-05, + "loss": 0.3471, + "step": 3852 + }, + { + "epoch": 0.23, + "learning_rate": 1.792041259277913e-05, + "loss": 0.3376, + "step": 3853 + }, + { + "epoch": 0.23, + "learning_rate": 1.7919217971561186e-05, + "loss": 0.3594, + "step": 3854 + }, + { + "epoch": 0.23, + "learning_rate": 1.7918023047158512e-05, + "loss": 0.3397, + "step": 3855 + }, + { + "epoch": 0.23, + "learning_rate": 1.7916827819616854e-05, + "loss": 0.3744, + "step": 3856 + }, + { + "epoch": 0.23, + "learning_rate": 1.791563228898198e-05, + "loss": 0.3477, + "step": 3857 + }, + { + "epoch": 0.23, + "learning_rate": 1.7914436455299648e-05, + "loss": 0.3493, + "step": 3858 + }, + { + "epoch": 0.23, + "learning_rate": 1.791324031861565e-05, + "loss": 0.344, + "step": 3859 + }, + { + "epoch": 0.23, + "learning_rate": 1.7912043878975774e-05, + "loss": 0.3719, + "step": 3860 + }, + { + "epoch": 0.23, + "learning_rate": 1.791084713642583e-05, + "loss": 0.3764, + "step": 3861 + }, + { + "epoch": 0.23, + "learning_rate": 1.790965009101163e-05, + "loss": 0.3507, + "step": 3862 + }, + { + "epoch": 0.23, + "learning_rate": 1.7908452742779006e-05, + "loss": 0.3602, + "step": 3863 + }, + { + "epoch": 0.23, + "learning_rate": 1.7907255091773798e-05, + "loss": 0.3644, + "step": 3864 + }, + { + "epoch": 0.23, + "learning_rate": 1.7906057138041854e-05, + "loss": 0.3379, + "step": 3865 + }, + { + "epoch": 0.23, + "learning_rate": 1.7904858881629043e-05, + "loss": 0.3627, + "step": 3866 + }, + { + "epoch": 0.23, + "learning_rate": 1.7903660322581236e-05, + "loss": 0.3432, + "step": 3867 + }, + { + "epoch": 0.23, + "learning_rate": 1.7902461460944315e-05, + "loss": 0.3477, + "step": 3868 + }, + { + "epoch": 0.23, + "learning_rate": 1.790126229676419e-05, + "loss": 0.3446, + "step": 3869 + }, + { + "epoch": 0.23, + "learning_rate": 1.7900062830086763e-05, + "loss": 0.375, + "step": 3870 + }, + { + "epoch": 0.23, + "learning_rate": 1.789886306095795e-05, + "loss": 0.3594, + "step": 3871 + }, + { + "epoch": 0.23, + "learning_rate": 1.7897662989423698e-05, + "loss": 0.4135, + "step": 3872 + }, + { + "epoch": 0.23, + "learning_rate": 1.789646261552994e-05, + "loss": 0.3167, + "step": 3873 + }, + { + "epoch": 0.23, + "learning_rate": 1.7895261939322633e-05, + "loss": 0.3438, + "step": 3874 + }, + { + "epoch": 0.23, + "learning_rate": 1.7894060960847747e-05, + "loss": 0.3711, + "step": 3875 + }, + { + "epoch": 0.23, + "learning_rate": 1.7892859680151262e-05, + "loss": 0.3502, + "step": 3876 + }, + { + "epoch": 0.23, + "learning_rate": 1.7891658097279168e-05, + "loss": 0.3324, + "step": 3877 + }, + { + "epoch": 0.23, + "learning_rate": 1.7890456212277465e-05, + "loss": 0.3555, + "step": 3878 + }, + { + "epoch": 0.23, + "learning_rate": 1.7889254025192172e-05, + "loss": 0.3492, + "step": 3879 + }, + { + "epoch": 0.23, + "learning_rate": 1.7888051536069305e-05, + "loss": 0.3513, + "step": 3880 + }, + { + "epoch": 0.23, + "learning_rate": 1.788684874495491e-05, + "loss": 0.322, + "step": 3881 + }, + { + "epoch": 0.23, + "learning_rate": 1.788564565189503e-05, + "loss": 0.3217, + "step": 3882 + }, + { + "epoch": 0.23, + "learning_rate": 1.788444225693573e-05, + "loss": 0.3426, + "step": 3883 + }, + { + "epoch": 0.23, + "learning_rate": 1.7883238560123077e-05, + "loss": 0.3585, + "step": 3884 + }, + { + "epoch": 0.23, + "learning_rate": 1.7882034561503157e-05, + "loss": 0.3711, + "step": 3885 + }, + { + "epoch": 0.23, + "learning_rate": 1.7880830261122062e-05, + "loss": 0.37, + "step": 3886 + }, + { + "epoch": 0.23, + "learning_rate": 1.78796256590259e-05, + "loss": 0.3457, + "step": 3887 + }, + { + "epoch": 0.23, + "learning_rate": 1.787842075526079e-05, + "loss": 0.3647, + "step": 3888 + }, + { + "epoch": 0.23, + "learning_rate": 1.7877215549872857e-05, + "loss": 0.3401, + "step": 3889 + }, + { + "epoch": 0.24, + "learning_rate": 1.7876010042908247e-05, + "loss": 0.3728, + "step": 3890 + }, + { + "epoch": 0.24, + "learning_rate": 1.787480423441311e-05, + "loss": 0.3538, + "step": 3891 + }, + { + "epoch": 0.24, + "learning_rate": 1.787359812443361e-05, + "loss": 0.3482, + "step": 3892 + }, + { + "epoch": 0.24, + "learning_rate": 1.7872391713015923e-05, + "loss": 0.3443, + "step": 3893 + }, + { + "epoch": 0.24, + "learning_rate": 1.7871185000206237e-05, + "loss": 0.3396, + "step": 3894 + }, + { + "epoch": 0.24, + "learning_rate": 1.786997798605075e-05, + "loss": 0.3492, + "step": 3895 + }, + { + "epoch": 0.24, + "learning_rate": 1.786877067059567e-05, + "loss": 0.3577, + "step": 3896 + }, + { + "epoch": 0.24, + "learning_rate": 1.786756305388722e-05, + "loss": 0.3744, + "step": 3897 + }, + { + "epoch": 0.24, + "learning_rate": 1.7866355135971637e-05, + "loss": 0.3318, + "step": 3898 + }, + { + "epoch": 0.24, + "learning_rate": 1.7865146916895156e-05, + "loss": 0.3677, + "step": 3899 + }, + { + "epoch": 0.24, + "learning_rate": 1.7863938396704046e-05, + "loss": 0.3323, + "step": 3900 + }, + { + "epoch": 0.24, + "learning_rate": 1.7862729575444566e-05, + "loss": 0.3714, + "step": 3901 + }, + { + "epoch": 0.24, + "learning_rate": 1.7861520453162996e-05, + "loss": 0.3362, + "step": 3902 + }, + { + "epoch": 0.24, + "learning_rate": 1.7860311029905632e-05, + "loss": 0.3722, + "step": 3903 + }, + { + "epoch": 0.24, + "learning_rate": 1.7859101305718767e-05, + "loss": 0.3474, + "step": 3904 + }, + { + "epoch": 0.24, + "learning_rate": 1.785789128064873e-05, + "loss": 0.3225, + "step": 3905 + }, + { + "epoch": 0.24, + "learning_rate": 1.7856680954741833e-05, + "loss": 0.3563, + "step": 3906 + }, + { + "epoch": 0.24, + "learning_rate": 1.7855470328044412e-05, + "loss": 0.3797, + "step": 3907 + }, + { + "epoch": 0.24, + "learning_rate": 1.7854259400602827e-05, + "loss": 0.3689, + "step": 3908 + }, + { + "epoch": 0.24, + "learning_rate": 1.785304817246343e-05, + "loss": 0.341, + "step": 3909 + }, + { + "epoch": 0.24, + "learning_rate": 1.7851836643672594e-05, + "loss": 0.3426, + "step": 3910 + }, + { + "epoch": 0.24, + "learning_rate": 1.7850624814276703e-05, + "loss": 0.3379, + "step": 3911 + }, + { + "epoch": 0.24, + "learning_rate": 1.784941268432215e-05, + "loss": 0.3248, + "step": 3912 + }, + { + "epoch": 0.24, + "learning_rate": 1.7848200253855344e-05, + "loss": 0.3438, + "step": 3913 + }, + { + "epoch": 0.24, + "learning_rate": 1.7846987522922697e-05, + "loss": 0.3287, + "step": 3914 + }, + { + "epoch": 0.24, + "learning_rate": 1.784577449157064e-05, + "loss": 0.3583, + "step": 3915 + }, + { + "epoch": 0.24, + "learning_rate": 1.7844561159845622e-05, + "loss": 0.3633, + "step": 3916 + }, + { + "epoch": 0.24, + "learning_rate": 1.784334752779408e-05, + "loss": 0.3694, + "step": 3917 + }, + { + "epoch": 0.24, + "learning_rate": 1.7842133595462488e-05, + "loss": 0.3647, + "step": 3918 + }, + { + "epoch": 0.24, + "learning_rate": 1.7840919362897317e-05, + "loss": 0.3178, + "step": 3919 + }, + { + "epoch": 0.24, + "learning_rate": 1.783970483014506e-05, + "loss": 0.3407, + "step": 3920 + }, + { + "epoch": 0.24, + "learning_rate": 1.7838489997252203e-05, + "loss": 0.3379, + "step": 3921 + }, + { + "epoch": 0.24, + "learning_rate": 1.7837274864265265e-05, + "loss": 0.3594, + "step": 3922 + }, + { + "epoch": 0.24, + "learning_rate": 1.7836059431230767e-05, + "loss": 0.373, + "step": 3923 + }, + { + "epoch": 0.24, + "learning_rate": 1.7834843698195233e-05, + "loss": 0.3915, + "step": 3924 + }, + { + "epoch": 0.24, + "learning_rate": 1.7833627665205218e-05, + "loss": 0.3583, + "step": 3925 + }, + { + "epoch": 0.24, + "learning_rate": 1.783241133230727e-05, + "loss": 0.3231, + "step": 3926 + }, + { + "epoch": 0.24, + "learning_rate": 1.7831194699547957e-05, + "loss": 0.3421, + "step": 3927 + }, + { + "epoch": 0.24, + "learning_rate": 1.782997776697386e-05, + "loss": 0.3499, + "step": 3928 + }, + { + "epoch": 0.24, + "learning_rate": 1.7828760534631565e-05, + "loss": 0.3516, + "step": 3929 + }, + { + "epoch": 0.24, + "learning_rate": 1.7827543002567678e-05, + "loss": 0.3443, + "step": 3930 + }, + { + "epoch": 0.24, + "learning_rate": 1.782632517082881e-05, + "loss": 0.3906, + "step": 3931 + }, + { + "epoch": 0.24, + "learning_rate": 1.7825107039461583e-05, + "loss": 0.3797, + "step": 3932 + }, + { + "epoch": 0.24, + "learning_rate": 1.7823888608512634e-05, + "loss": 0.3304, + "step": 3933 + }, + { + "epoch": 0.24, + "learning_rate": 1.7822669878028613e-05, + "loss": 0.3379, + "step": 3934 + }, + { + "epoch": 0.24, + "learning_rate": 1.7821450848056175e-05, + "loss": 0.3348, + "step": 3935 + }, + { + "epoch": 0.24, + "learning_rate": 1.782023151864199e-05, + "loss": 0.3387, + "step": 3936 + }, + { + "epoch": 0.24, + "learning_rate": 1.7819011889832746e-05, + "loss": 0.3479, + "step": 3937 + }, + { + "epoch": 0.24, + "learning_rate": 1.781779196167513e-05, + "loss": 0.334, + "step": 3938 + }, + { + "epoch": 0.24, + "learning_rate": 1.7816571734215845e-05, + "loss": 0.363, + "step": 3939 + }, + { + "epoch": 0.24, + "learning_rate": 1.781535120750161e-05, + "loss": 0.3387, + "step": 3940 + }, + { + "epoch": 0.24, + "learning_rate": 1.7814130381579156e-05, + "loss": 0.3479, + "step": 3941 + }, + { + "epoch": 0.24, + "learning_rate": 1.7812909256495215e-05, + "loss": 0.3647, + "step": 3942 + }, + { + "epoch": 0.24, + "learning_rate": 1.781168783229654e-05, + "loss": 0.3493, + "step": 3943 + }, + { + "epoch": 0.24, + "learning_rate": 1.7810466109029897e-05, + "loss": 0.3451, + "step": 3944 + }, + { + "epoch": 0.24, + "learning_rate": 1.7809244086742055e-05, + "loss": 0.3507, + "step": 3945 + }, + { + "epoch": 0.24, + "learning_rate": 1.78080217654798e-05, + "loss": 0.3145, + "step": 3946 + }, + { + "epoch": 0.24, + "learning_rate": 1.780679914528993e-05, + "loss": 0.3677, + "step": 3947 + }, + { + "epoch": 0.24, + "learning_rate": 1.7805576226219243e-05, + "loss": 0.3502, + "step": 3948 + }, + { + "epoch": 0.24, + "learning_rate": 1.780435300831457e-05, + "loss": 0.3817, + "step": 3949 + }, + { + "epoch": 0.24, + "learning_rate": 1.7803129491622735e-05, + "loss": 0.3016, + "step": 3950 + }, + { + "epoch": 0.24, + "learning_rate": 1.7801905676190586e-05, + "loss": 0.3366, + "step": 3951 + }, + { + "epoch": 0.24, + "learning_rate": 1.780068156206497e-05, + "loss": 0.3471, + "step": 3952 + }, + { + "epoch": 0.24, + "learning_rate": 1.7799457149292752e-05, + "loss": 0.3839, + "step": 3953 + }, + { + "epoch": 0.24, + "learning_rate": 1.7798232437920814e-05, + "loss": 0.3376, + "step": 3954 + }, + { + "epoch": 0.24, + "learning_rate": 1.779700742799604e-05, + "loss": 0.3384, + "step": 3955 + }, + { + "epoch": 0.24, + "learning_rate": 1.779578211956533e-05, + "loss": 0.3304, + "step": 3956 + }, + { + "epoch": 0.24, + "learning_rate": 1.7794556512675585e-05, + "loss": 0.3424, + "step": 3957 + }, + { + "epoch": 0.24, + "learning_rate": 1.7793330607373747e-05, + "loss": 0.3792, + "step": 3958 + }, + { + "epoch": 0.24, + "learning_rate": 1.779210440370673e-05, + "loss": 0.3345, + "step": 3959 + }, + { + "epoch": 0.24, + "learning_rate": 1.7790877901721494e-05, + "loss": 0.3334, + "step": 3960 + }, + { + "epoch": 0.24, + "learning_rate": 1.7789651101464982e-05, + "loss": 0.3343, + "step": 3961 + }, + { + "epoch": 0.24, + "learning_rate": 1.7788424002984168e-05, + "loss": 0.3198, + "step": 3962 + }, + { + "epoch": 0.24, + "learning_rate": 1.7787196606326035e-05, + "loss": 0.3772, + "step": 3963 + }, + { + "epoch": 0.24, + "learning_rate": 1.7785968911537567e-05, + "loss": 0.3147, + "step": 3964 + }, + { + "epoch": 0.24, + "learning_rate": 1.7784740918665767e-05, + "loss": 0.3507, + "step": 3965 + }, + { + "epoch": 0.24, + "learning_rate": 1.7783512627757652e-05, + "loss": 0.3786, + "step": 3966 + }, + { + "epoch": 0.24, + "learning_rate": 1.7782284038860242e-05, + "loss": 0.3443, + "step": 3967 + }, + { + "epoch": 0.24, + "learning_rate": 1.778105515202058e-05, + "loss": 0.3404, + "step": 3968 + }, + { + "epoch": 0.24, + "learning_rate": 1.7779825967285704e-05, + "loss": 0.3672, + "step": 3969 + }, + { + "epoch": 0.24, + "learning_rate": 1.7778596484702683e-05, + "loss": 0.3415, + "step": 3970 + }, + { + "epoch": 0.24, + "learning_rate": 1.7777366704318576e-05, + "loss": 0.3507, + "step": 3971 + }, + { + "epoch": 0.24, + "learning_rate": 1.7776136626180473e-05, + "loss": 0.3362, + "step": 3972 + }, + { + "epoch": 0.24, + "learning_rate": 1.7774906250335466e-05, + "loss": 0.3461, + "step": 3973 + }, + { + "epoch": 0.24, + "learning_rate": 1.777367557683066e-05, + "loss": 0.3281, + "step": 3974 + }, + { + "epoch": 0.24, + "learning_rate": 1.777244460571317e-05, + "loss": 0.3415, + "step": 3975 + }, + { + "epoch": 0.24, + "learning_rate": 1.7771213337030116e-05, + "loss": 0.3259, + "step": 3976 + }, + { + "epoch": 0.24, + "learning_rate": 1.776998177082865e-05, + "loss": 0.3396, + "step": 3977 + }, + { + "epoch": 0.24, + "learning_rate": 1.7768749907155917e-05, + "loss": 0.3351, + "step": 3978 + }, + { + "epoch": 0.24, + "learning_rate": 1.7767517746059077e-05, + "loss": 0.358, + "step": 3979 + }, + { + "epoch": 0.24, + "learning_rate": 1.77662852875853e-05, + "loss": 0.3496, + "step": 3980 + }, + { + "epoch": 0.24, + "learning_rate": 1.7765052531781774e-05, + "loss": 0.3496, + "step": 3981 + }, + { + "epoch": 0.24, + "learning_rate": 1.7763819478695693e-05, + "loss": 0.3223, + "step": 3982 + }, + { + "epoch": 0.24, + "learning_rate": 1.7762586128374267e-05, + "loss": 0.3616, + "step": 3983 + }, + { + "epoch": 0.24, + "learning_rate": 1.7761352480864715e-05, + "loss": 0.3499, + "step": 3984 + }, + { + "epoch": 0.24, + "learning_rate": 1.776011853621426e-05, + "loss": 0.3585, + "step": 3985 + }, + { + "epoch": 0.24, + "learning_rate": 1.775888429447015e-05, + "loss": 0.3318, + "step": 3986 + }, + { + "epoch": 0.24, + "learning_rate": 1.7757649755679636e-05, + "loss": 0.3471, + "step": 3987 + }, + { + "epoch": 0.24, + "learning_rate": 1.775641491988998e-05, + "loss": 0.3398, + "step": 3988 + }, + { + "epoch": 0.24, + "learning_rate": 1.775517978714846e-05, + "loss": 0.3161, + "step": 3989 + }, + { + "epoch": 0.24, + "learning_rate": 1.775394435750236e-05, + "loss": 0.3739, + "step": 3990 + }, + { + "epoch": 0.24, + "learning_rate": 1.775270863099898e-05, + "loss": 0.3312, + "step": 3991 + }, + { + "epoch": 0.24, + "learning_rate": 1.7751472607685625e-05, + "loss": 0.3566, + "step": 3992 + }, + { + "epoch": 0.24, + "learning_rate": 1.775023628760962e-05, + "loss": 0.3502, + "step": 3993 + }, + { + "epoch": 0.24, + "learning_rate": 1.77489996708183e-05, + "loss": 0.365, + "step": 3994 + }, + { + "epoch": 0.24, + "learning_rate": 1.7747762757358998e-05, + "loss": 0.3301, + "step": 3995 + }, + { + "epoch": 0.24, + "learning_rate": 1.7746525547279082e-05, + "loss": 0.3672, + "step": 3996 + }, + { + "epoch": 0.24, + "learning_rate": 1.774528804062591e-05, + "loss": 0.3424, + "step": 3997 + }, + { + "epoch": 0.24, + "learning_rate": 1.774405023744686e-05, + "loss": 0.3691, + "step": 3998 + }, + { + "epoch": 0.24, + "learning_rate": 1.7742812137789325e-05, + "loss": 0.3616, + "step": 3999 + }, + { + "epoch": 0.24, + "learning_rate": 1.77415737417007e-05, + "loss": 0.3298, + "step": 4000 + }, + { + "epoch": 0.24, + "learning_rate": 1.7740335049228402e-05, + "loss": 0.3362, + "step": 4001 + }, + { + "epoch": 0.24, + "learning_rate": 1.773909606041985e-05, + "loss": 0.3245, + "step": 4002 + }, + { + "epoch": 0.24, + "learning_rate": 1.7737856775322475e-05, + "loss": 0.3627, + "step": 4003 + }, + { + "epoch": 0.24, + "learning_rate": 1.773661719398373e-05, + "loss": 0.3451, + "step": 4004 + }, + { + "epoch": 0.24, + "learning_rate": 1.7735377316451072e-05, + "loss": 0.3463, + "step": 4005 + }, + { + "epoch": 0.24, + "learning_rate": 1.7734137142771962e-05, + "loss": 0.3876, + "step": 4006 + }, + { + "epoch": 0.24, + "learning_rate": 1.773289667299389e-05, + "loss": 0.3613, + "step": 4007 + }, + { + "epoch": 0.24, + "learning_rate": 1.7731655907164333e-05, + "loss": 0.3371, + "step": 4008 + }, + { + "epoch": 0.24, + "learning_rate": 1.7730414845330805e-05, + "loss": 0.3304, + "step": 4009 + }, + { + "epoch": 0.24, + "learning_rate": 1.772917348754082e-05, + "loss": 0.3792, + "step": 4010 + }, + { + "epoch": 0.24, + "learning_rate": 1.7727931833841895e-05, + "loss": 0.3181, + "step": 4011 + }, + { + "epoch": 0.24, + "learning_rate": 1.7726689884281567e-05, + "loss": 0.3566, + "step": 4012 + }, + { + "epoch": 0.24, + "learning_rate": 1.772544763890739e-05, + "loss": 0.3373, + "step": 4013 + }, + { + "epoch": 0.24, + "learning_rate": 1.772420509776692e-05, + "loss": 0.341, + "step": 4014 + }, + { + "epoch": 0.24, + "learning_rate": 1.772296226090773e-05, + "loss": 0.3594, + "step": 4015 + }, + { + "epoch": 0.24, + "learning_rate": 1.7721719128377393e-05, + "loss": 0.346, + "step": 4016 + }, + { + "epoch": 0.24, + "learning_rate": 1.772047570022351e-05, + "loss": 0.3661, + "step": 4017 + }, + { + "epoch": 0.24, + "learning_rate": 1.7719231976493683e-05, + "loss": 0.3354, + "step": 4018 + }, + { + "epoch": 0.24, + "learning_rate": 1.771798795723553e-05, + "loss": 0.3845, + "step": 4019 + }, + { + "epoch": 0.24, + "learning_rate": 1.771674364249667e-05, + "loss": 0.3318, + "step": 4020 + }, + { + "epoch": 0.24, + "learning_rate": 1.771549903232475e-05, + "loss": 0.3393, + "step": 4021 + }, + { + "epoch": 0.24, + "learning_rate": 1.7714254126767417e-05, + "loss": 0.3393, + "step": 4022 + }, + { + "epoch": 0.24, + "learning_rate": 1.7713008925872328e-05, + "loss": 0.3237, + "step": 4023 + }, + { + "epoch": 0.24, + "learning_rate": 1.771176342968716e-05, + "loss": 0.3658, + "step": 4024 + }, + { + "epoch": 0.24, + "learning_rate": 1.7710517638259593e-05, + "loss": 0.358, + "step": 4025 + }, + { + "epoch": 0.24, + "learning_rate": 1.7709271551637323e-05, + "loss": 0.339, + "step": 4026 + }, + { + "epoch": 0.24, + "learning_rate": 1.770802516986806e-05, + "loss": 0.3438, + "step": 4027 + }, + { + "epoch": 0.24, + "learning_rate": 1.770677849299951e-05, + "loss": 0.3499, + "step": 4028 + }, + { + "epoch": 0.24, + "learning_rate": 1.7705531521079416e-05, + "loss": 0.3451, + "step": 4029 + }, + { + "epoch": 0.24, + "learning_rate": 1.7704284254155508e-05, + "loss": 0.3538, + "step": 4030 + }, + { + "epoch": 0.24, + "learning_rate": 1.770303669227554e-05, + "loss": 0.3783, + "step": 4031 + }, + { + "epoch": 0.24, + "learning_rate": 1.7701788835487276e-05, + "loss": 0.3641, + "step": 4032 + }, + { + "epoch": 0.24, + "learning_rate": 1.770054068383849e-05, + "loss": 0.346, + "step": 4033 + }, + { + "epoch": 0.24, + "learning_rate": 1.7699292237376965e-05, + "loss": 0.3337, + "step": 4034 + }, + { + "epoch": 0.24, + "learning_rate": 1.7698043496150494e-05, + "loss": 0.3641, + "step": 4035 + }, + { + "epoch": 0.24, + "learning_rate": 1.769679446020689e-05, + "loss": 0.3463, + "step": 4036 + }, + { + "epoch": 0.24, + "learning_rate": 1.7695545129593974e-05, + "loss": 0.3357, + "step": 4037 + }, + { + "epoch": 0.24, + "learning_rate": 1.7694295504359572e-05, + "loss": 0.3636, + "step": 4038 + }, + { + "epoch": 0.24, + "learning_rate": 1.7693045584551526e-05, + "loss": 0.3454, + "step": 4039 + }, + { + "epoch": 0.24, + "learning_rate": 1.7691795370217686e-05, + "loss": 0.3987, + "step": 4040 + }, + { + "epoch": 0.24, + "learning_rate": 1.7690544861405924e-05, + "loss": 0.3093, + "step": 4041 + }, + { + "epoch": 0.24, + "learning_rate": 1.7689294058164106e-05, + "loss": 0.3477, + "step": 4042 + }, + { + "epoch": 0.24, + "learning_rate": 1.7688042960540125e-05, + "loss": 0.3454, + "step": 4043 + }, + { + "epoch": 0.24, + "learning_rate": 1.7686791568581877e-05, + "loss": 0.3438, + "step": 4044 + }, + { + "epoch": 0.24, + "learning_rate": 1.768553988233727e-05, + "loss": 0.3711, + "step": 4045 + }, + { + "epoch": 0.24, + "learning_rate": 1.7684287901854228e-05, + "loss": 0.3708, + "step": 4046 + }, + { + "epoch": 0.24, + "learning_rate": 1.768303562718068e-05, + "loss": 0.3273, + "step": 4047 + }, + { + "epoch": 0.24, + "learning_rate": 1.768178305836457e-05, + "loss": 0.3357, + "step": 4048 + }, + { + "epoch": 0.24, + "learning_rate": 1.7680530195453845e-05, + "loss": 0.3583, + "step": 4049 + }, + { + "epoch": 0.24, + "learning_rate": 1.7679277038496484e-05, + "loss": 0.3273, + "step": 4050 + }, + { + "epoch": 0.24, + "learning_rate": 1.7678023587540455e-05, + "loss": 0.3309, + "step": 4051 + }, + { + "epoch": 0.24, + "learning_rate": 1.7676769842633745e-05, + "loss": 0.3485, + "step": 4052 + }, + { + "epoch": 0.24, + "learning_rate": 1.7675515803824358e-05, + "loss": 0.3318, + "step": 4053 + }, + { + "epoch": 0.24, + "learning_rate": 1.7674261471160296e-05, + "loss": 0.3244, + "step": 4054 + }, + { + "epoch": 0.24, + "learning_rate": 1.7673006844689595e-05, + "loss": 0.3465, + "step": 4055 + }, + { + "epoch": 0.25, + "learning_rate": 1.7671751924460275e-05, + "loss": 0.3265, + "step": 4056 + }, + { + "epoch": 0.25, + "learning_rate": 1.767049671052039e-05, + "loss": 0.3711, + "step": 4057 + }, + { + "epoch": 0.25, + "learning_rate": 1.7669241202917987e-05, + "loss": 0.3742, + "step": 4058 + }, + { + "epoch": 0.25, + "learning_rate": 1.7667985401701137e-05, + "loss": 0.3521, + "step": 4059 + }, + { + "epoch": 0.25, + "learning_rate": 1.7666729306917917e-05, + "loss": 0.365, + "step": 4060 + }, + { + "epoch": 0.25, + "learning_rate": 1.7665472918616412e-05, + "loss": 0.3571, + "step": 4061 + }, + { + "epoch": 0.25, + "learning_rate": 1.7664216236844733e-05, + "loss": 0.3396, + "step": 4062 + }, + { + "epoch": 0.25, + "learning_rate": 1.7662959261650986e-05, + "loss": 0.3278, + "step": 4063 + }, + { + "epoch": 0.25, + "learning_rate": 1.7661701993083288e-05, + "loss": 0.3181, + "step": 4064 + }, + { + "epoch": 0.25, + "learning_rate": 1.766044443118978e-05, + "loss": 0.3555, + "step": 4065 + }, + { + "epoch": 0.25, + "learning_rate": 1.7659186576018607e-05, + "loss": 0.368, + "step": 4066 + }, + { + "epoch": 0.25, + "learning_rate": 1.7657928427617928e-05, + "loss": 0.353, + "step": 4067 + }, + { + "epoch": 0.25, + "learning_rate": 1.76566699860359e-05, + "loss": 0.385, + "step": 4068 + }, + { + "epoch": 0.25, + "learning_rate": 1.7655411251320717e-05, + "loss": 0.3468, + "step": 4069 + }, + { + "epoch": 0.25, + "learning_rate": 1.7654152223520556e-05, + "loss": 0.3276, + "step": 4070 + }, + { + "epoch": 0.25, + "learning_rate": 1.765289290268363e-05, + "loss": 0.3524, + "step": 4071 + }, + { + "epoch": 0.25, + "learning_rate": 1.7651633288858136e-05, + "loss": 0.3546, + "step": 4072 + }, + { + "epoch": 0.25, + "learning_rate": 1.7650373382092313e-05, + "loss": 0.3365, + "step": 4073 + }, + { + "epoch": 0.25, + "learning_rate": 1.7649113182434396e-05, + "loss": 0.3376, + "step": 4074 + }, + { + "epoch": 0.25, + "learning_rate": 1.7647852689932618e-05, + "loss": 0.3491, + "step": 4075 + }, + { + "epoch": 0.25, + "learning_rate": 1.7646591904635247e-05, + "loss": 0.3708, + "step": 4076 + }, + { + "epoch": 0.25, + "learning_rate": 1.764533082659055e-05, + "loss": 0.3178, + "step": 4077 + }, + { + "epoch": 0.25, + "learning_rate": 1.7644069455846808e-05, + "loss": 0.3312, + "step": 4078 + }, + { + "epoch": 0.25, + "learning_rate": 1.7642807792452305e-05, + "loss": 0.3655, + "step": 4079 + }, + { + "epoch": 0.25, + "learning_rate": 1.7641545836455356e-05, + "loss": 0.3641, + "step": 4080 + }, + { + "epoch": 0.25, + "learning_rate": 1.7640283587904262e-05, + "loss": 0.3309, + "step": 4081 + }, + { + "epoch": 0.25, + "learning_rate": 1.763902104684736e-05, + "loss": 0.3376, + "step": 4082 + }, + { + "epoch": 0.25, + "learning_rate": 1.763775821333297e-05, + "loss": 0.335, + "step": 4083 + }, + { + "epoch": 0.25, + "learning_rate": 1.7636495087409456e-05, + "loss": 0.3362, + "step": 4084 + }, + { + "epoch": 0.25, + "learning_rate": 1.7635231669125164e-05, + "loss": 0.3368, + "step": 4085 + }, + { + "epoch": 0.25, + "learning_rate": 1.763396795852847e-05, + "loss": 0.3602, + "step": 4086 + }, + { + "epoch": 0.25, + "learning_rate": 1.7632703955667755e-05, + "loss": 0.3323, + "step": 4087 + }, + { + "epoch": 0.25, + "learning_rate": 1.7631439660591406e-05, + "loss": 0.385, + "step": 4088 + }, + { + "epoch": 0.25, + "learning_rate": 1.7630175073347833e-05, + "loss": 0.3198, + "step": 4089 + }, + { + "epoch": 0.25, + "learning_rate": 1.7628910193985443e-05, + "loss": 0.3164, + "step": 4090 + }, + { + "epoch": 0.25, + "learning_rate": 1.7627645022552666e-05, + "loss": 0.341, + "step": 4091 + }, + { + "epoch": 0.25, + "learning_rate": 1.7626379559097938e-05, + "loss": 0.3248, + "step": 4092 + }, + { + "epoch": 0.25, + "learning_rate": 1.762511380366971e-05, + "loss": 0.2966, + "step": 4093 + }, + { + "epoch": 0.25, + "learning_rate": 1.7623847756316434e-05, + "loss": 0.3636, + "step": 4094 + }, + { + "epoch": 0.25, + "learning_rate": 1.7622581417086583e-05, + "loss": 0.3309, + "step": 4095 + }, + { + "epoch": 0.25, + "learning_rate": 1.7621314786028643e-05, + "loss": 0.3691, + "step": 4096 + }, + { + "epoch": 0.25, + "learning_rate": 1.76200478631911e-05, + "loss": 0.353, + "step": 4097 + }, + { + "epoch": 0.25, + "learning_rate": 1.761878064862246e-05, + "loss": 0.3845, + "step": 4098 + }, + { + "epoch": 0.25, + "learning_rate": 1.761751314237124e-05, + "loss": 0.3795, + "step": 4099 + }, + { + "epoch": 0.25, + "learning_rate": 1.7616245344485964e-05, + "loss": 0.3251, + "step": 4100 + }, + { + "epoch": 0.25, + "learning_rate": 1.7614977255015166e-05, + "loss": 0.3329, + "step": 4101 + }, + { + "epoch": 0.25, + "learning_rate": 1.7613708874007405e-05, + "loss": 0.3443, + "step": 4102 + }, + { + "epoch": 0.25, + "learning_rate": 1.761244020151123e-05, + "loss": 0.3189, + "step": 4103 + }, + { + "epoch": 0.25, + "learning_rate": 1.7611171237575214e-05, + "loss": 0.3415, + "step": 4104 + }, + { + "epoch": 0.25, + "learning_rate": 1.7609901982247942e-05, + "loss": 0.3532, + "step": 4105 + }, + { + "epoch": 0.25, + "learning_rate": 1.760863243557801e-05, + "loss": 0.3195, + "step": 4106 + }, + { + "epoch": 0.25, + "learning_rate": 1.760736259761401e-05, + "loss": 0.3541, + "step": 4107 + }, + { + "epoch": 0.25, + "learning_rate": 1.7606092468404567e-05, + "loss": 0.3362, + "step": 4108 + }, + { + "epoch": 0.25, + "learning_rate": 1.7604822047998306e-05, + "loss": 0.3521, + "step": 4109 + }, + { + "epoch": 0.25, + "learning_rate": 1.7603551336443866e-05, + "loss": 0.3373, + "step": 4110 + }, + { + "epoch": 0.25, + "learning_rate": 1.7602280333789892e-05, + "loss": 0.344, + "step": 4111 + }, + { + "epoch": 0.25, + "learning_rate": 1.7601009040085046e-05, + "loss": 0.3273, + "step": 4112 + }, + { + "epoch": 0.25, + "learning_rate": 1.7599737455378e-05, + "loss": 0.3284, + "step": 4113 + }, + { + "epoch": 0.25, + "learning_rate": 1.7598465579717432e-05, + "loss": 0.3443, + "step": 4114 + }, + { + "epoch": 0.25, + "learning_rate": 1.759719341315204e-05, + "loss": 0.3571, + "step": 4115 + }, + { + "epoch": 0.25, + "learning_rate": 1.7595920955730527e-05, + "loss": 0.341, + "step": 4116 + }, + { + "epoch": 0.25, + "learning_rate": 1.759464820750161e-05, + "loss": 0.3373, + "step": 4117 + }, + { + "epoch": 0.25, + "learning_rate": 1.759337516851401e-05, + "loss": 0.3563, + "step": 4118 + }, + { + "epoch": 0.25, + "learning_rate": 1.7592101838816473e-05, + "loss": 0.3544, + "step": 4119 + }, + { + "epoch": 0.25, + "learning_rate": 1.7590828218457743e-05, + "loss": 0.3518, + "step": 4120 + }, + { + "epoch": 0.25, + "learning_rate": 1.758955430748658e-05, + "loss": 0.3742, + "step": 4121 + }, + { + "epoch": 0.25, + "learning_rate": 1.7588280105951758e-05, + "loss": 0.3627, + "step": 4122 + }, + { + "epoch": 0.25, + "learning_rate": 1.7587005613902056e-05, + "loss": 0.3368, + "step": 4123 + }, + { + "epoch": 0.25, + "learning_rate": 1.7585730831386276e-05, + "loss": 0.3504, + "step": 4124 + }, + { + "epoch": 0.25, + "learning_rate": 1.758445575845321e-05, + "loss": 0.3624, + "step": 4125 + }, + { + "epoch": 0.25, + "learning_rate": 1.758318039515168e-05, + "loss": 0.3585, + "step": 4126 + }, + { + "epoch": 0.25, + "learning_rate": 1.7581904741530514e-05, + "loss": 0.3373, + "step": 4127 + }, + { + "epoch": 0.25, + "learning_rate": 1.7580628797638552e-05, + "loss": 0.365, + "step": 4128 + }, + { + "epoch": 0.25, + "learning_rate": 1.7579352563524635e-05, + "loss": 0.3304, + "step": 4129 + }, + { + "epoch": 0.25, + "learning_rate": 1.7578076039237627e-05, + "loss": 0.334, + "step": 4130 + }, + { + "epoch": 0.25, + "learning_rate": 1.7576799224826405e-05, + "loss": 0.3417, + "step": 4131 + }, + { + "epoch": 0.25, + "learning_rate": 1.7575522120339846e-05, + "loss": 0.3235, + "step": 4132 + }, + { + "epoch": 0.25, + "learning_rate": 1.757424472582684e-05, + "loss": 0.3831, + "step": 4133 + }, + { + "epoch": 0.25, + "learning_rate": 1.75729670413363e-05, + "loss": 0.3117, + "step": 4134 + }, + { + "epoch": 0.25, + "learning_rate": 1.757168906691714e-05, + "loss": 0.3281, + "step": 4135 + }, + { + "epoch": 0.25, + "learning_rate": 1.757041080261828e-05, + "loss": 0.3418, + "step": 4136 + }, + { + "epoch": 0.25, + "learning_rate": 1.7569132248488664e-05, + "loss": 0.3396, + "step": 4137 + }, + { + "epoch": 0.25, + "learning_rate": 1.7567853404577244e-05, + "loss": 0.3689, + "step": 4138 + }, + { + "epoch": 0.25, + "learning_rate": 1.756657427093297e-05, + "loss": 0.3488, + "step": 4139 + }, + { + "epoch": 0.25, + "learning_rate": 1.7565294847604816e-05, + "loss": 0.351, + "step": 4140 + }, + { + "epoch": 0.25, + "learning_rate": 1.7564015134641773e-05, + "loss": 0.3845, + "step": 4141 + }, + { + "epoch": 0.25, + "learning_rate": 1.756273513209283e-05, + "loss": 0.3331, + "step": 4142 + }, + { + "epoch": 0.25, + "learning_rate": 1.7561454840006987e-05, + "loss": 0.3647, + "step": 4143 + }, + { + "epoch": 0.25, + "learning_rate": 1.7560174258433263e-05, + "loss": 0.3736, + "step": 4144 + }, + { + "epoch": 0.25, + "learning_rate": 1.7558893387420683e-05, + "loss": 0.3781, + "step": 4145 + }, + { + "epoch": 0.25, + "learning_rate": 1.7557612227018286e-05, + "loss": 0.3722, + "step": 4146 + }, + { + "epoch": 0.25, + "learning_rate": 1.7556330777275124e-05, + "loss": 0.3622, + "step": 4147 + }, + { + "epoch": 0.25, + "learning_rate": 1.7555049038240252e-05, + "loss": 0.3265, + "step": 4148 + }, + { + "epoch": 0.25, + "learning_rate": 1.7553767009962745e-05, + "loss": 0.3468, + "step": 4149 + }, + { + "epoch": 0.25, + "learning_rate": 1.755248469249168e-05, + "loss": 0.3697, + "step": 4150 + }, + { + "epoch": 0.25, + "learning_rate": 1.7551202085876152e-05, + "loss": 0.3411, + "step": 4151 + }, + { + "epoch": 0.25, + "learning_rate": 1.754991919016527e-05, + "loss": 0.3465, + "step": 4152 + }, + { + "epoch": 0.25, + "learning_rate": 1.7548636005408145e-05, + "loss": 0.3426, + "step": 4153 + }, + { + "epoch": 0.25, + "learning_rate": 1.75473525316539e-05, + "loss": 0.3767, + "step": 4154 + }, + { + "epoch": 0.25, + "learning_rate": 1.754606876895168e-05, + "loss": 0.3143, + "step": 4155 + }, + { + "epoch": 0.25, + "learning_rate": 1.7544784717350626e-05, + "loss": 0.3209, + "step": 4156 + }, + { + "epoch": 0.25, + "learning_rate": 1.7543500376899903e-05, + "loss": 0.3491, + "step": 4157 + }, + { + "epoch": 0.25, + "learning_rate": 1.7542215747648678e-05, + "loss": 0.373, + "step": 4158 + }, + { + "epoch": 0.25, + "learning_rate": 1.7540930829646138e-05, + "loss": 0.3411, + "step": 4159 + }, + { + "epoch": 0.25, + "learning_rate": 1.7539645622941467e-05, + "loss": 0.3465, + "step": 4160 + }, + { + "epoch": 0.25, + "learning_rate": 1.7538360127583875e-05, + "loss": 0.3521, + "step": 4161 + }, + { + "epoch": 0.25, + "learning_rate": 1.753707434362258e-05, + "loss": 0.3415, + "step": 4162 + }, + { + "epoch": 0.25, + "learning_rate": 1.7535788271106796e-05, + "loss": 0.3502, + "step": 4163 + }, + { + "epoch": 0.25, + "learning_rate": 1.753450191008577e-05, + "loss": 0.3295, + "step": 4164 + }, + { + "epoch": 0.25, + "learning_rate": 1.753321526060875e-05, + "loss": 0.3248, + "step": 4165 + }, + { + "epoch": 0.25, + "learning_rate": 1.7531928322724988e-05, + "loss": 0.3482, + "step": 4166 + }, + { + "epoch": 0.25, + "learning_rate": 1.753064109648376e-05, + "loss": 0.3491, + "step": 4167 + }, + { + "epoch": 0.25, + "learning_rate": 1.7529353581934345e-05, + "loss": 0.3544, + "step": 4168 + }, + { + "epoch": 0.25, + "learning_rate": 1.7528065779126035e-05, + "loss": 0.3161, + "step": 4169 + }, + { + "epoch": 0.25, + "learning_rate": 1.7526777688108133e-05, + "loss": 0.3474, + "step": 4170 + }, + { + "epoch": 0.25, + "learning_rate": 1.752548930892995e-05, + "loss": 0.3435, + "step": 4171 + }, + { + "epoch": 0.25, + "learning_rate": 1.752420064164082e-05, + "loss": 0.3365, + "step": 4172 + }, + { + "epoch": 0.25, + "learning_rate": 1.7522911686290073e-05, + "loss": 0.3438, + "step": 4173 + }, + { + "epoch": 0.25, + "learning_rate": 1.7521622442927056e-05, + "loss": 0.37, + "step": 4174 + }, + { + "epoch": 0.25, + "learning_rate": 1.7520332911601127e-05, + "loss": 0.3449, + "step": 4175 + }, + { + "epoch": 0.25, + "learning_rate": 1.751904309236166e-05, + "loss": 0.3845, + "step": 4176 + }, + { + "epoch": 0.25, + "learning_rate": 1.7517752985258027e-05, + "loss": 0.3396, + "step": 4177 + }, + { + "epoch": 0.25, + "learning_rate": 1.7516462590339626e-05, + "loss": 0.3488, + "step": 4178 + }, + { + "epoch": 0.25, + "learning_rate": 1.751517190765586e-05, + "loss": 0.3285, + "step": 4179 + }, + { + "epoch": 0.25, + "learning_rate": 1.7513880937256136e-05, + "loss": 0.3357, + "step": 4180 + }, + { + "epoch": 0.25, + "learning_rate": 1.7512589679189888e-05, + "loss": 0.3432, + "step": 4181 + }, + { + "epoch": 0.25, + "learning_rate": 1.751129813350654e-05, + "loss": 0.3541, + "step": 4182 + }, + { + "epoch": 0.25, + "learning_rate": 1.7510006300255548e-05, + "loss": 0.3577, + "step": 4183 + }, + { + "epoch": 0.25, + "learning_rate": 1.7508714179486365e-05, + "loss": 0.3415, + "step": 4184 + }, + { + "epoch": 0.25, + "learning_rate": 1.7507421771248457e-05, + "loss": 0.3451, + "step": 4185 + }, + { + "epoch": 0.25, + "learning_rate": 1.7506129075591308e-05, + "loss": 0.3485, + "step": 4186 + }, + { + "epoch": 0.25, + "learning_rate": 1.7504836092564408e-05, + "loss": 0.351, + "step": 4187 + }, + { + "epoch": 0.25, + "learning_rate": 1.7503542822217258e-05, + "loss": 0.3426, + "step": 4188 + }, + { + "epoch": 0.25, + "learning_rate": 1.7502249264599368e-05, + "loss": 0.3544, + "step": 4189 + }, + { + "epoch": 0.25, + "learning_rate": 1.7500955419760263e-05, + "loss": 0.3343, + "step": 4190 + }, + { + "epoch": 0.25, + "learning_rate": 1.749966128774948e-05, + "loss": 0.3348, + "step": 4191 + }, + { + "epoch": 0.25, + "learning_rate": 1.749836686861656e-05, + "loss": 0.3585, + "step": 4192 + }, + { + "epoch": 0.25, + "learning_rate": 1.7497072162411062e-05, + "loss": 0.3588, + "step": 4193 + }, + { + "epoch": 0.25, + "learning_rate": 1.7495777169182552e-05, + "loss": 0.3443, + "step": 4194 + }, + { + "epoch": 0.25, + "learning_rate": 1.749448188898061e-05, + "loss": 0.3507, + "step": 4195 + }, + { + "epoch": 0.25, + "learning_rate": 1.7493186321854826e-05, + "loss": 0.3359, + "step": 4196 + }, + { + "epoch": 0.25, + "learning_rate": 1.74918904678548e-05, + "loss": 0.3334, + "step": 4197 + }, + { + "epoch": 0.25, + "learning_rate": 1.749059432703014e-05, + "loss": 0.3836, + "step": 4198 + }, + { + "epoch": 0.25, + "learning_rate": 1.748929789943047e-05, + "loss": 0.375, + "step": 4199 + }, + { + "epoch": 0.25, + "learning_rate": 1.7488001185105426e-05, + "loss": 0.334, + "step": 4200 + }, + { + "epoch": 0.25, + "learning_rate": 1.748670418410465e-05, + "loss": 0.3477, + "step": 4201 + }, + { + "epoch": 0.25, + "learning_rate": 1.7485406896477798e-05, + "loss": 0.3573, + "step": 4202 + }, + { + "epoch": 0.25, + "learning_rate": 1.7484109322274538e-05, + "loss": 0.3432, + "step": 4203 + }, + { + "epoch": 0.25, + "learning_rate": 1.748281146154454e-05, + "loss": 0.3831, + "step": 4204 + }, + { + "epoch": 0.25, + "learning_rate": 1.74815133143375e-05, + "loss": 0.3488, + "step": 4205 + }, + { + "epoch": 0.25, + "learning_rate": 1.7480214880703114e-05, + "loss": 0.3518, + "step": 4206 + }, + { + "epoch": 0.25, + "learning_rate": 1.747891616069109e-05, + "loss": 0.3622, + "step": 4207 + }, + { + "epoch": 0.25, + "learning_rate": 1.7477617154351158e-05, + "loss": 0.3287, + "step": 4208 + }, + { + "epoch": 0.25, + "learning_rate": 1.7476317861733042e-05, + "loss": 0.3471, + "step": 4209 + }, + { + "epoch": 0.25, + "learning_rate": 1.7475018282886484e-05, + "loss": 0.3334, + "step": 4210 + }, + { + "epoch": 0.25, + "learning_rate": 1.7473718417861242e-05, + "loss": 0.339, + "step": 4211 + }, + { + "epoch": 0.25, + "learning_rate": 1.7472418266707078e-05, + "loss": 0.3368, + "step": 4212 + }, + { + "epoch": 0.25, + "learning_rate": 1.747111782947377e-05, + "loss": 0.3641, + "step": 4213 + }, + { + "epoch": 0.25, + "learning_rate": 1.7469817106211108e-05, + "loss": 0.3334, + "step": 4214 + }, + { + "epoch": 0.25, + "learning_rate": 1.746851609696888e-05, + "loss": 0.3566, + "step": 4215 + }, + { + "epoch": 0.25, + "learning_rate": 1.7467214801796905e-05, + "loss": 0.3976, + "step": 4216 + }, + { + "epoch": 0.25, + "learning_rate": 1.7465913220745e-05, + "loss": 0.3571, + "step": 4217 + }, + { + "epoch": 0.25, + "learning_rate": 1.746461135386299e-05, + "loss": 0.3619, + "step": 4218 + }, + { + "epoch": 0.25, + "learning_rate": 1.746330920120072e-05, + "loss": 0.3488, + "step": 4219 + }, + { + "epoch": 0.25, + "learning_rate": 1.746200676280805e-05, + "loss": 0.3686, + "step": 4220 + }, + { + "epoch": 0.25, + "learning_rate": 1.7460704038734828e-05, + "loss": 0.3387, + "step": 4221 + }, + { + "epoch": 0.26, + "learning_rate": 1.7459401029030943e-05, + "loss": 0.3552, + "step": 4222 + }, + { + "epoch": 0.26, + "learning_rate": 1.745809773374627e-05, + "loss": 0.3175, + "step": 4223 + }, + { + "epoch": 0.26, + "learning_rate": 1.7456794152930713e-05, + "loss": 0.3454, + "step": 4224 + }, + { + "epoch": 0.26, + "learning_rate": 1.7455490286634172e-05, + "loss": 0.3251, + "step": 4225 + }, + { + "epoch": 0.26, + "learning_rate": 1.745418613490657e-05, + "loss": 0.3379, + "step": 4226 + }, + { + "epoch": 0.26, + "learning_rate": 1.7452881697797836e-05, + "loss": 0.3412, + "step": 4227 + }, + { + "epoch": 0.26, + "learning_rate": 1.745157697535791e-05, + "loss": 0.3382, + "step": 4228 + }, + { + "epoch": 0.26, + "learning_rate": 1.745027196763674e-05, + "loss": 0.3304, + "step": 4229 + }, + { + "epoch": 0.26, + "learning_rate": 1.7448966674684286e-05, + "loss": 0.3306, + "step": 4230 + }, + { + "epoch": 0.26, + "learning_rate": 1.7447661096550527e-05, + "loss": 0.2999, + "step": 4231 + }, + { + "epoch": 0.26, + "learning_rate": 1.7446355233285447e-05, + "loss": 0.3471, + "step": 4232 + }, + { + "epoch": 0.26, + "learning_rate": 1.7445049084939035e-05, + "loss": 0.3379, + "step": 4233 + }, + { + "epoch": 0.26, + "learning_rate": 1.7443742651561296e-05, + "loss": 0.375, + "step": 4234 + }, + { + "epoch": 0.26, + "learning_rate": 1.744243593320225e-05, + "loss": 0.3636, + "step": 4235 + }, + { + "epoch": 0.26, + "learning_rate": 1.7441128929911924e-05, + "loss": 0.3666, + "step": 4236 + }, + { + "epoch": 0.26, + "learning_rate": 1.743982164174036e-05, + "loss": 0.3362, + "step": 4237 + }, + { + "epoch": 0.26, + "learning_rate": 1.74385140687376e-05, + "loss": 0.3331, + "step": 4238 + }, + { + "epoch": 0.26, + "learning_rate": 1.7437206210953704e-05, + "loss": 0.3493, + "step": 4239 + }, + { + "epoch": 0.26, + "learning_rate": 1.743589806843875e-05, + "loss": 0.3468, + "step": 4240 + }, + { + "epoch": 0.26, + "learning_rate": 1.7434589641242814e-05, + "loss": 0.3418, + "step": 4241 + }, + { + "epoch": 0.26, + "learning_rate": 1.7433280929415988e-05, + "loss": 0.3758, + "step": 4242 + }, + { + "epoch": 0.26, + "learning_rate": 1.743197193300838e-05, + "loss": 0.3196, + "step": 4243 + }, + { + "epoch": 0.26, + "learning_rate": 1.7430662652070105e-05, + "loss": 0.3739, + "step": 4244 + }, + { + "epoch": 0.26, + "learning_rate": 1.7429353086651282e-05, + "loss": 0.37, + "step": 4245 + }, + { + "epoch": 0.26, + "learning_rate": 1.742804323680205e-05, + "loss": 0.3465, + "step": 4246 + }, + { + "epoch": 0.26, + "learning_rate": 1.7426733102572563e-05, + "loss": 0.3435, + "step": 4247 + }, + { + "epoch": 0.26, + "learning_rate": 1.7425422684012967e-05, + "loss": 0.3647, + "step": 4248 + }, + { + "epoch": 0.26, + "learning_rate": 1.742411198117344e-05, + "loss": 0.3255, + "step": 4249 + }, + { + "epoch": 0.26, + "learning_rate": 1.742280099410416e-05, + "loss": 0.3267, + "step": 4250 + }, + { + "epoch": 0.26, + "learning_rate": 1.7421489722855318e-05, + "loss": 0.3262, + "step": 4251 + }, + { + "epoch": 0.26, + "learning_rate": 1.7420178167477112e-05, + "loss": 0.344, + "step": 4252 + }, + { + "epoch": 0.26, + "learning_rate": 1.741886632801976e-05, + "loss": 0.347, + "step": 4253 + }, + { + "epoch": 0.26, + "learning_rate": 1.7417554204533478e-05, + "loss": 0.3557, + "step": 4254 + }, + { + "epoch": 0.26, + "learning_rate": 1.7416241797068506e-05, + "loss": 0.3527, + "step": 4255 + }, + { + "epoch": 0.26, + "learning_rate": 1.7414929105675086e-05, + "loss": 0.3438, + "step": 4256 + }, + { + "epoch": 0.26, + "learning_rate": 1.741361613040348e-05, + "loss": 0.3786, + "step": 4257 + }, + { + "epoch": 0.26, + "learning_rate": 1.7412302871303946e-05, + "loss": 0.3396, + "step": 4258 + }, + { + "epoch": 0.26, + "learning_rate": 1.741098932842677e-05, + "loss": 0.3281, + "step": 4259 + }, + { + "epoch": 0.26, + "learning_rate": 1.7409675501822236e-05, + "loss": 0.3443, + "step": 4260 + }, + { + "epoch": 0.26, + "learning_rate": 1.7408361391540642e-05, + "loss": 0.3479, + "step": 4261 + }, + { + "epoch": 0.26, + "learning_rate": 1.7407046997632302e-05, + "loss": 0.361, + "step": 4262 + }, + { + "epoch": 0.26, + "learning_rate": 1.7405732320147533e-05, + "loss": 0.3638, + "step": 4263 + }, + { + "epoch": 0.26, + "learning_rate": 1.7404417359136673e-05, + "loss": 0.3443, + "step": 4264 + }, + { + "epoch": 0.26, + "learning_rate": 1.740310211465006e-05, + "loss": 0.3594, + "step": 4265 + }, + { + "epoch": 0.26, + "learning_rate": 1.7401786586738047e-05, + "loss": 0.3092, + "step": 4266 + }, + { + "epoch": 0.26, + "learning_rate": 1.7400470775451004e-05, + "loss": 0.3524, + "step": 4267 + }, + { + "epoch": 0.26, + "learning_rate": 1.73991546808393e-05, + "loss": 0.3499, + "step": 4268 + }, + { + "epoch": 0.26, + "learning_rate": 1.7397838302953333e-05, + "loss": 0.3418, + "step": 4269 + }, + { + "epoch": 0.26, + "learning_rate": 1.739652164184348e-05, + "loss": 0.3128, + "step": 4270 + }, + { + "epoch": 0.26, + "learning_rate": 1.7395204697560167e-05, + "loss": 0.351, + "step": 4271 + }, + { + "epoch": 0.26, + "learning_rate": 1.7393887470153803e-05, + "loss": 0.3599, + "step": 4272 + }, + { + "epoch": 0.26, + "learning_rate": 1.7392569959674826e-05, + "loss": 0.3329, + "step": 4273 + }, + { + "epoch": 0.26, + "learning_rate": 1.7391252166173667e-05, + "loss": 0.3359, + "step": 4274 + }, + { + "epoch": 0.26, + "learning_rate": 1.7389934089700783e-05, + "loss": 0.3295, + "step": 4275 + }, + { + "epoch": 0.26, + "learning_rate": 1.7388615730306636e-05, + "loss": 0.332, + "step": 4276 + }, + { + "epoch": 0.26, + "learning_rate": 1.7387297088041696e-05, + "loss": 0.3415, + "step": 4277 + }, + { + "epoch": 0.26, + "learning_rate": 1.738597816295645e-05, + "loss": 0.3482, + "step": 4278 + }, + { + "epoch": 0.26, + "learning_rate": 1.7384658955101387e-05, + "loss": 0.3491, + "step": 4279 + }, + { + "epoch": 0.26, + "learning_rate": 1.738333946452702e-05, + "loss": 0.3544, + "step": 4280 + }, + { + "epoch": 0.26, + "learning_rate": 1.7382019691283862e-05, + "loss": 0.3524, + "step": 4281 + }, + { + "epoch": 0.26, + "learning_rate": 1.7380699635422438e-05, + "loss": 0.3161, + "step": 4282 + }, + { + "epoch": 0.26, + "learning_rate": 1.7379379296993293e-05, + "loss": 0.3412, + "step": 4283 + }, + { + "epoch": 0.26, + "learning_rate": 1.7378058676046964e-05, + "loss": 0.3468, + "step": 4284 + }, + { + "epoch": 0.26, + "learning_rate": 1.7376737772634022e-05, + "loss": 0.3535, + "step": 4285 + }, + { + "epoch": 0.26, + "learning_rate": 1.737541658680503e-05, + "loss": 0.344, + "step": 4286 + }, + { + "epoch": 0.26, + "learning_rate": 1.7374095118610572e-05, + "loss": 0.3482, + "step": 4287 + }, + { + "epoch": 0.26, + "learning_rate": 1.737277336810124e-05, + "loss": 0.3365, + "step": 4288 + }, + { + "epoch": 0.26, + "learning_rate": 1.737145133532764e-05, + "loss": 0.3549, + "step": 4289 + }, + { + "epoch": 0.26, + "learning_rate": 1.737012902034038e-05, + "loss": 0.3312, + "step": 4290 + }, + { + "epoch": 0.26, + "learning_rate": 1.7368806423190083e-05, + "loss": 0.3451, + "step": 4291 + }, + { + "epoch": 0.26, + "learning_rate": 1.7367483543927394e-05, + "loss": 0.3739, + "step": 4292 + }, + { + "epoch": 0.26, + "learning_rate": 1.7366160382602946e-05, + "loss": 0.346, + "step": 4293 + }, + { + "epoch": 0.26, + "learning_rate": 1.7364836939267407e-05, + "loss": 0.3496, + "step": 4294 + }, + { + "epoch": 0.26, + "learning_rate": 1.736351321397144e-05, + "loss": 0.3664, + "step": 4295 + }, + { + "epoch": 0.26, + "learning_rate": 1.736218920676572e-05, + "loss": 0.3295, + "step": 4296 + }, + { + "epoch": 0.26, + "learning_rate": 1.7360864917700943e-05, + "loss": 0.3853, + "step": 4297 + }, + { + "epoch": 0.26, + "learning_rate": 1.7359540346827806e-05, + "loss": 0.3538, + "step": 4298 + }, + { + "epoch": 0.26, + "learning_rate": 1.735821549419702e-05, + "loss": 0.3574, + "step": 4299 + }, + { + "epoch": 0.26, + "learning_rate": 1.73568903598593e-05, + "loss": 0.3432, + "step": 4300 + }, + { + "epoch": 0.26, + "learning_rate": 1.7355564943865393e-05, + "loss": 0.3412, + "step": 4301 + }, + { + "epoch": 0.26, + "learning_rate": 1.7354239246266028e-05, + "loss": 0.3407, + "step": 4302 + }, + { + "epoch": 0.26, + "learning_rate": 1.7352913267111965e-05, + "loss": 0.3633, + "step": 4303 + }, + { + "epoch": 0.26, + "learning_rate": 1.735158700645397e-05, + "loss": 0.3538, + "step": 4304 + }, + { + "epoch": 0.26, + "learning_rate": 1.7350260464342817e-05, + "loss": 0.3496, + "step": 4305 + }, + { + "epoch": 0.26, + "learning_rate": 1.734893364082929e-05, + "loss": 0.3225, + "step": 4306 + }, + { + "epoch": 0.26, + "learning_rate": 1.734760653596419e-05, + "loss": 0.3518, + "step": 4307 + }, + { + "epoch": 0.26, + "learning_rate": 1.734627914979832e-05, + "loss": 0.3778, + "step": 4308 + }, + { + "epoch": 0.26, + "learning_rate": 1.7344951482382502e-05, + "loss": 0.3421, + "step": 4309 + }, + { + "epoch": 0.26, + "learning_rate": 1.7343623533767567e-05, + "loss": 0.38, + "step": 4310 + }, + { + "epoch": 0.26, + "learning_rate": 1.7342295304004348e-05, + "loss": 0.3442, + "step": 4311 + }, + { + "epoch": 0.26, + "learning_rate": 1.7340966793143704e-05, + "loss": 0.368, + "step": 4312 + }, + { + "epoch": 0.26, + "learning_rate": 1.7339638001236495e-05, + "loss": 0.3744, + "step": 4313 + }, + { + "epoch": 0.26, + "learning_rate": 1.7338308928333587e-05, + "loss": 0.3241, + "step": 4314 + }, + { + "epoch": 0.26, + "learning_rate": 1.733697957448587e-05, + "loss": 0.3711, + "step": 4315 + }, + { + "epoch": 0.26, + "learning_rate": 1.733564993974423e-05, + "loss": 0.3672, + "step": 4316 + }, + { + "epoch": 0.26, + "learning_rate": 1.7334320024159584e-05, + "loss": 0.3638, + "step": 4317 + }, + { + "epoch": 0.26, + "learning_rate": 1.733298982778284e-05, + "loss": 0.3742, + "step": 4318 + }, + { + "epoch": 0.26, + "learning_rate": 1.7331659350664925e-05, + "loss": 0.3831, + "step": 4319 + }, + { + "epoch": 0.26, + "learning_rate": 1.733032859285677e-05, + "loss": 0.3237, + "step": 4320 + }, + { + "epoch": 0.26, + "learning_rate": 1.7328997554409334e-05, + "loss": 0.3809, + "step": 4321 + }, + { + "epoch": 0.26, + "learning_rate": 1.7327666235373567e-05, + "loss": 0.3306, + "step": 4322 + }, + { + "epoch": 0.26, + "learning_rate": 1.732633463580044e-05, + "loss": 0.3591, + "step": 4323 + }, + { + "epoch": 0.26, + "learning_rate": 1.7325002755740935e-05, + "loss": 0.3384, + "step": 4324 + }, + { + "epoch": 0.26, + "learning_rate": 1.732367059524604e-05, + "loss": 0.3474, + "step": 4325 + }, + { + "epoch": 0.26, + "learning_rate": 1.732233815436676e-05, + "loss": 0.339, + "step": 4326 + }, + { + "epoch": 0.26, + "learning_rate": 1.7321005433154102e-05, + "loss": 0.3532, + "step": 4327 + }, + { + "epoch": 0.26, + "learning_rate": 1.7319672431659092e-05, + "loss": 0.3524, + "step": 4328 + }, + { + "epoch": 0.26, + "learning_rate": 1.7318339149932763e-05, + "loss": 0.3491, + "step": 4329 + }, + { + "epoch": 0.26, + "learning_rate": 1.731700558802616e-05, + "loss": 0.3376, + "step": 4330 + }, + { + "epoch": 0.26, + "learning_rate": 1.7315671745990338e-05, + "loss": 0.356, + "step": 4331 + }, + { + "epoch": 0.26, + "learning_rate": 1.731433762387636e-05, + "loss": 0.3666, + "step": 4332 + }, + { + "epoch": 0.26, + "learning_rate": 1.7313003221735306e-05, + "loss": 0.3535, + "step": 4333 + }, + { + "epoch": 0.26, + "learning_rate": 1.7311668539618258e-05, + "loss": 0.3806, + "step": 4334 + }, + { + "epoch": 0.26, + "learning_rate": 1.7310333577576323e-05, + "loss": 0.3326, + "step": 4335 + }, + { + "epoch": 0.26, + "learning_rate": 1.7308998335660598e-05, + "loss": 0.339, + "step": 4336 + }, + { + "epoch": 0.26, + "learning_rate": 1.730766281392221e-05, + "loss": 0.3636, + "step": 4337 + }, + { + "epoch": 0.26, + "learning_rate": 1.730632701241229e-05, + "loss": 0.3465, + "step": 4338 + }, + { + "epoch": 0.26, + "learning_rate": 1.7304990931181975e-05, + "loss": 0.341, + "step": 4339 + }, + { + "epoch": 0.26, + "learning_rate": 1.7303654570282417e-05, + "loss": 0.3323, + "step": 4340 + }, + { + "epoch": 0.26, + "learning_rate": 1.7302317929764775e-05, + "loss": 0.334, + "step": 4341 + }, + { + "epoch": 0.26, + "learning_rate": 1.730098100968023e-05, + "loss": 0.3507, + "step": 4342 + }, + { + "epoch": 0.26, + "learning_rate": 1.7299643810079956e-05, + "loss": 0.3334, + "step": 4343 + }, + { + "epoch": 0.26, + "learning_rate": 1.729830633101516e-05, + "loss": 0.3544, + "step": 4344 + }, + { + "epoch": 0.26, + "learning_rate": 1.7296968572537033e-05, + "loss": 0.3647, + "step": 4345 + }, + { + "epoch": 0.26, + "learning_rate": 1.72956305346968e-05, + "loss": 0.3566, + "step": 4346 + }, + { + "epoch": 0.26, + "learning_rate": 1.7294292217545684e-05, + "loss": 0.332, + "step": 4347 + }, + { + "epoch": 0.26, + "learning_rate": 1.7292953621134922e-05, + "loss": 0.3376, + "step": 4348 + }, + { + "epoch": 0.26, + "learning_rate": 1.729161474551576e-05, + "loss": 0.3368, + "step": 4349 + }, + { + "epoch": 0.26, + "learning_rate": 1.729027559073946e-05, + "loss": 0.3803, + "step": 4350 + }, + { + "epoch": 0.26, + "learning_rate": 1.728893615685729e-05, + "loss": 0.3323, + "step": 4351 + }, + { + "epoch": 0.26, + "learning_rate": 1.7287596443920528e-05, + "loss": 0.3379, + "step": 4352 + }, + { + "epoch": 0.26, + "learning_rate": 1.7286256451980467e-05, + "loss": 0.3248, + "step": 4353 + }, + { + "epoch": 0.26, + "learning_rate": 1.728491618108841e-05, + "loss": 0.3683, + "step": 4354 + }, + { + "epoch": 0.26, + "learning_rate": 1.728357563129566e-05, + "loss": 0.3507, + "step": 4355 + }, + { + "epoch": 0.26, + "learning_rate": 1.728223480265355e-05, + "loss": 0.3345, + "step": 4356 + }, + { + "epoch": 0.26, + "learning_rate": 1.7280893695213405e-05, + "loss": 0.3566, + "step": 4357 + }, + { + "epoch": 0.26, + "learning_rate": 1.7279552309026576e-05, + "loss": 0.3465, + "step": 4358 + }, + { + "epoch": 0.26, + "learning_rate": 1.727821064414441e-05, + "loss": 0.3443, + "step": 4359 + }, + { + "epoch": 0.26, + "learning_rate": 1.7276868700618277e-05, + "loss": 0.368, + "step": 4360 + }, + { + "epoch": 0.26, + "learning_rate": 1.7275526478499556e-05, + "loss": 0.3354, + "step": 4361 + }, + { + "epoch": 0.26, + "learning_rate": 1.7274183977839627e-05, + "loss": 0.3407, + "step": 4362 + }, + { + "epoch": 0.26, + "learning_rate": 1.727284119868989e-05, + "loss": 0.3544, + "step": 4363 + }, + { + "epoch": 0.26, + "learning_rate": 1.7271498141101754e-05, + "loss": 0.3535, + "step": 4364 + }, + { + "epoch": 0.26, + "learning_rate": 1.7270154805126633e-05, + "loss": 0.3725, + "step": 4365 + }, + { + "epoch": 0.26, + "learning_rate": 1.726881119081596e-05, + "loss": 0.3485, + "step": 4366 + }, + { + "epoch": 0.26, + "learning_rate": 1.7267467298221178e-05, + "loss": 0.3504, + "step": 4367 + }, + { + "epoch": 0.26, + "learning_rate": 1.7266123127393732e-05, + "loss": 0.3139, + "step": 4368 + }, + { + "epoch": 0.26, + "learning_rate": 1.7264778678385083e-05, + "loss": 0.3301, + "step": 4369 + }, + { + "epoch": 0.26, + "learning_rate": 1.7263433951246706e-05, + "loss": 0.3549, + "step": 4370 + }, + { + "epoch": 0.26, + "learning_rate": 1.726208894603008e-05, + "loss": 0.3524, + "step": 4371 + }, + { + "epoch": 0.26, + "learning_rate": 1.7260743662786703e-05, + "loss": 0.3244, + "step": 4372 + }, + { + "epoch": 0.26, + "learning_rate": 1.7259398101568077e-05, + "loss": 0.3111, + "step": 4373 + }, + { + "epoch": 0.26, + "learning_rate": 1.7258052262425714e-05, + "loss": 0.3404, + "step": 4374 + }, + { + "epoch": 0.26, + "learning_rate": 1.725670614541114e-05, + "loss": 0.334, + "step": 4375 + }, + { + "epoch": 0.26, + "learning_rate": 1.725535975057589e-05, + "loss": 0.3856, + "step": 4376 + }, + { + "epoch": 0.26, + "learning_rate": 1.725401307797151e-05, + "loss": 0.3362, + "step": 4377 + }, + { + "epoch": 0.26, + "learning_rate": 1.725266612764956e-05, + "loss": 0.3669, + "step": 4378 + }, + { + "epoch": 0.26, + "learning_rate": 1.725131889966161e-05, + "loss": 0.3209, + "step": 4379 + }, + { + "epoch": 0.26, + "learning_rate": 1.724997139405923e-05, + "loss": 0.3594, + "step": 4380 + }, + { + "epoch": 0.26, + "learning_rate": 1.7248623610894015e-05, + "loss": 0.358, + "step": 4381 + }, + { + "epoch": 0.26, + "learning_rate": 1.724727555021756e-05, + "loss": 0.3301, + "step": 4382 + }, + { + "epoch": 0.26, + "learning_rate": 1.724592721208148e-05, + "loss": 0.3605, + "step": 4383 + }, + { + "epoch": 0.26, + "learning_rate": 1.7244578596537387e-05, + "loss": 0.3722, + "step": 4384 + }, + { + "epoch": 0.26, + "learning_rate": 1.7243229703636924e-05, + "loss": 0.3359, + "step": 4385 + }, + { + "epoch": 0.26, + "learning_rate": 1.7241880533431725e-05, + "loss": 0.3736, + "step": 4386 + }, + { + "epoch": 0.27, + "learning_rate": 1.7240531085973448e-05, + "loss": 0.3608, + "step": 4387 + }, + { + "epoch": 0.27, + "learning_rate": 1.723918136131375e-05, + "loss": 0.3666, + "step": 4388 + }, + { + "epoch": 0.27, + "learning_rate": 1.7237831359504308e-05, + "loss": 0.3544, + "step": 4389 + }, + { + "epoch": 0.27, + "learning_rate": 1.7236481080596806e-05, + "loss": 0.3178, + "step": 4390 + }, + { + "epoch": 0.27, + "learning_rate": 1.723513052464294e-05, + "loss": 0.3371, + "step": 4391 + }, + { + "epoch": 0.27, + "learning_rate": 1.7233779691694413e-05, + "loss": 0.3371, + "step": 4392 + }, + { + "epoch": 0.27, + "learning_rate": 1.7232428581802948e-05, + "loss": 0.3382, + "step": 4393 + }, + { + "epoch": 0.27, + "learning_rate": 1.723107719502026e-05, + "loss": 0.345, + "step": 4394 + }, + { + "epoch": 0.27, + "learning_rate": 1.72297255313981e-05, + "loss": 0.3499, + "step": 4395 + }, + { + "epoch": 0.27, + "learning_rate": 1.7228373590988205e-05, + "loss": 0.3569, + "step": 4396 + }, + { + "epoch": 0.27, + "learning_rate": 1.722702137384234e-05, + "loss": 0.3161, + "step": 4397 + }, + { + "epoch": 0.27, + "learning_rate": 1.722566888001227e-05, + "loss": 0.3527, + "step": 4398 + }, + { + "epoch": 0.27, + "learning_rate": 1.7224316109549777e-05, + "loss": 0.363, + "step": 4399 + }, + { + "epoch": 0.27, + "learning_rate": 1.7222963062506654e-05, + "loss": 0.3426, + "step": 4400 + }, + { + "epoch": 0.27, + "learning_rate": 1.7221609738934697e-05, + "loss": 0.3401, + "step": 4401 + }, + { + "epoch": 0.27, + "learning_rate": 1.722025613888572e-05, + "loss": 0.3523, + "step": 4402 + }, + { + "epoch": 0.27, + "learning_rate": 1.7218902262411547e-05, + "loss": 0.361, + "step": 4403 + }, + { + "epoch": 0.27, + "learning_rate": 1.7217548109564003e-05, + "loss": 0.3382, + "step": 4404 + }, + { + "epoch": 0.27, + "learning_rate": 1.7216193680394943e-05, + "loss": 0.3489, + "step": 4405 + }, + { + "epoch": 0.27, + "learning_rate": 1.721483897495621e-05, + "loss": 0.3055, + "step": 4406 + }, + { + "epoch": 0.27, + "learning_rate": 1.7213483993299675e-05, + "loss": 0.3599, + "step": 4407 + }, + { + "epoch": 0.27, + "learning_rate": 1.7212128735477215e-05, + "loss": 0.332, + "step": 4408 + }, + { + "epoch": 0.27, + "learning_rate": 1.7210773201540706e-05, + "loss": 0.3329, + "step": 4409 + }, + { + "epoch": 0.27, + "learning_rate": 1.7209417391542055e-05, + "loss": 0.3594, + "step": 4410 + }, + { + "epoch": 0.27, + "learning_rate": 1.7208061305533158e-05, + "loss": 0.3714, + "step": 4411 + }, + { + "epoch": 0.27, + "learning_rate": 1.7206704943565945e-05, + "loss": 0.3454, + "step": 4412 + }, + { + "epoch": 0.27, + "learning_rate": 1.7205348305692335e-05, + "loss": 0.3566, + "step": 4413 + }, + { + "epoch": 0.27, + "learning_rate": 1.7203991391964265e-05, + "loss": 0.3337, + "step": 4414 + }, + { + "epoch": 0.27, + "learning_rate": 1.720263420243369e-05, + "loss": 0.3111, + "step": 4415 + }, + { + "epoch": 0.27, + "learning_rate": 1.720127673715257e-05, + "loss": 0.3318, + "step": 4416 + }, + { + "epoch": 0.27, + "learning_rate": 1.7199918996172868e-05, + "loss": 0.356, + "step": 4417 + }, + { + "epoch": 0.27, + "learning_rate": 1.7198560979546572e-05, + "loss": 0.3477, + "step": 4418 + }, + { + "epoch": 0.27, + "learning_rate": 1.719720268732567e-05, + "loss": 0.3225, + "step": 4419 + }, + { + "epoch": 0.27, + "learning_rate": 1.7195844119562156e-05, + "loss": 0.3345, + "step": 4420 + }, + { + "epoch": 0.27, + "learning_rate": 1.7194485276308058e-05, + "loss": 0.3608, + "step": 4421 + }, + { + "epoch": 0.27, + "learning_rate": 1.719312615761539e-05, + "loss": 0.3371, + "step": 4422 + }, + { + "epoch": 0.27, + "learning_rate": 1.7191766763536187e-05, + "loss": 0.341, + "step": 4423 + }, + { + "epoch": 0.27, + "learning_rate": 1.719040709412249e-05, + "loss": 0.3207, + "step": 4424 + }, + { + "epoch": 0.27, + "learning_rate": 1.7189047149426362e-05, + "loss": 0.3304, + "step": 4425 + }, + { + "epoch": 0.27, + "learning_rate": 1.7187686929499855e-05, + "loss": 0.3544, + "step": 4426 + }, + { + "epoch": 0.27, + "learning_rate": 1.7186326434395056e-05, + "loss": 0.3245, + "step": 4427 + }, + { + "epoch": 0.27, + "learning_rate": 1.7184965664164046e-05, + "loss": 0.3465, + "step": 4428 + }, + { + "epoch": 0.27, + "learning_rate": 1.718360461885892e-05, + "loss": 0.3348, + "step": 4429 + }, + { + "epoch": 0.27, + "learning_rate": 1.7182243298531792e-05, + "loss": 0.3354, + "step": 4430 + }, + { + "epoch": 0.27, + "learning_rate": 1.718088170323477e-05, + "loss": 0.3086, + "step": 4431 + }, + { + "epoch": 0.27, + "learning_rate": 1.7179519833019992e-05, + "loss": 0.3326, + "step": 4432 + }, + { + "epoch": 0.27, + "learning_rate": 1.717815768793959e-05, + "loss": 0.3312, + "step": 4433 + }, + { + "epoch": 0.27, + "learning_rate": 1.717679526804572e-05, + "loss": 0.3398, + "step": 4434 + }, + { + "epoch": 0.27, + "learning_rate": 1.717543257339053e-05, + "loss": 0.3474, + "step": 4435 + }, + { + "epoch": 0.27, + "learning_rate": 1.7174069604026202e-05, + "loss": 0.3359, + "step": 4436 + }, + { + "epoch": 0.27, + "learning_rate": 1.7172706360004912e-05, + "loss": 0.3432, + "step": 4437 + }, + { + "epoch": 0.27, + "learning_rate": 1.7171342841378854e-05, + "loss": 0.3449, + "step": 4438 + }, + { + "epoch": 0.27, + "learning_rate": 1.7169979048200226e-05, + "loss": 0.3323, + "step": 4439 + }, + { + "epoch": 0.27, + "learning_rate": 1.7168614980521242e-05, + "loss": 0.3454, + "step": 4440 + }, + { + "epoch": 0.27, + "learning_rate": 1.7167250638394127e-05, + "loss": 0.3412, + "step": 4441 + }, + { + "epoch": 0.27, + "learning_rate": 1.716588602187111e-05, + "loss": 0.3343, + "step": 4442 + }, + { + "epoch": 0.27, + "learning_rate": 1.716452113100444e-05, + "loss": 0.3703, + "step": 4443 + }, + { + "epoch": 0.27, + "learning_rate": 1.7163155965846366e-05, + "loss": 0.3638, + "step": 4444 + }, + { + "epoch": 0.27, + "learning_rate": 1.7161790526449156e-05, + "loss": 0.3435, + "step": 4445 + }, + { + "epoch": 0.27, + "learning_rate": 1.716042481286509e-05, + "loss": 0.3412, + "step": 4446 + }, + { + "epoch": 0.27, + "learning_rate": 1.7159058825146444e-05, + "loss": 0.3315, + "step": 4447 + }, + { + "epoch": 0.27, + "learning_rate": 1.715769256334552e-05, + "loss": 0.3563, + "step": 4448 + }, + { + "epoch": 0.27, + "learning_rate": 1.7156326027514623e-05, + "loss": 0.3527, + "step": 4449 + }, + { + "epoch": 0.27, + "learning_rate": 1.7154959217706077e-05, + "loss": 0.3256, + "step": 4450 + }, + { + "epoch": 0.27, + "learning_rate": 1.7153592133972202e-05, + "loss": 0.3301, + "step": 4451 + }, + { + "epoch": 0.27, + "learning_rate": 1.7152224776365338e-05, + "loss": 0.3571, + "step": 4452 + }, + { + "epoch": 0.27, + "learning_rate": 1.7150857144937835e-05, + "loss": 0.3541, + "step": 4453 + }, + { + "epoch": 0.27, + "learning_rate": 1.7149489239742053e-05, + "loss": 0.3401, + "step": 4454 + }, + { + "epoch": 0.27, + "learning_rate": 1.714812106083036e-05, + "loss": 0.3661, + "step": 4455 + }, + { + "epoch": 0.27, + "learning_rate": 1.714675260825514e-05, + "loss": 0.3304, + "step": 4456 + }, + { + "epoch": 0.27, + "learning_rate": 1.714538388206878e-05, + "loss": 0.353, + "step": 4457 + }, + { + "epoch": 0.27, + "learning_rate": 1.714401488232368e-05, + "loss": 0.3365, + "step": 4458 + }, + { + "epoch": 0.27, + "learning_rate": 1.714264560907226e-05, + "loss": 0.3451, + "step": 4459 + }, + { + "epoch": 0.27, + "learning_rate": 1.7141276062366927e-05, + "loss": 0.3404, + "step": 4460 + }, + { + "epoch": 0.27, + "learning_rate": 1.713990624226013e-05, + "loss": 0.3384, + "step": 4461 + }, + { + "epoch": 0.27, + "learning_rate": 1.7138536148804302e-05, + "loss": 0.3359, + "step": 4462 + }, + { + "epoch": 0.27, + "learning_rate": 1.71371657820519e-05, + "loss": 0.3747, + "step": 4463 + }, + { + "epoch": 0.27, + "learning_rate": 1.7135795142055386e-05, + "loss": 0.3097, + "step": 4464 + }, + { + "epoch": 0.27, + "learning_rate": 1.713442422886724e-05, + "loss": 0.346, + "step": 4465 + }, + { + "epoch": 0.27, + "learning_rate": 1.713305304253994e-05, + "loss": 0.3717, + "step": 4466 + }, + { + "epoch": 0.27, + "learning_rate": 1.7131681583125987e-05, + "loss": 0.3521, + "step": 4467 + }, + { + "epoch": 0.27, + "learning_rate": 1.713030985067788e-05, + "loss": 0.3588, + "step": 4468 + }, + { + "epoch": 0.27, + "learning_rate": 1.7128937845248146e-05, + "loss": 0.3636, + "step": 4469 + }, + { + "epoch": 0.27, + "learning_rate": 1.7127565566889305e-05, + "loss": 0.3086, + "step": 4470 + }, + { + "epoch": 0.27, + "learning_rate": 1.712619301565389e-05, + "loss": 0.365, + "step": 4471 + }, + { + "epoch": 0.27, + "learning_rate": 1.7124820191594452e-05, + "loss": 0.3128, + "step": 4472 + }, + { + "epoch": 0.27, + "learning_rate": 1.712344709476356e-05, + "loss": 0.3438, + "step": 4473 + }, + { + "epoch": 0.27, + "learning_rate": 1.7122073725213766e-05, + "loss": 0.3429, + "step": 4474 + }, + { + "epoch": 0.27, + "learning_rate": 1.7120700082997655e-05, + "loss": 0.3251, + "step": 4475 + }, + { + "epoch": 0.27, + "learning_rate": 1.7119326168167822e-05, + "loss": 0.3496, + "step": 4476 + }, + { + "epoch": 0.27, + "learning_rate": 1.7117951980776865e-05, + "loss": 0.3418, + "step": 4477 + }, + { + "epoch": 0.27, + "learning_rate": 1.7116577520877387e-05, + "loss": 0.3468, + "step": 4478 + }, + { + "epoch": 0.27, + "learning_rate": 1.7115202788522012e-05, + "loss": 0.3583, + "step": 4479 + }, + { + "epoch": 0.27, + "learning_rate": 1.7113827783763377e-05, + "loss": 0.3608, + "step": 4480 + }, + { + "epoch": 0.27, + "learning_rate": 1.7112452506654117e-05, + "loss": 0.3474, + "step": 4481 + }, + { + "epoch": 0.27, + "learning_rate": 1.711107695724689e-05, + "loss": 0.3407, + "step": 4482 + }, + { + "epoch": 0.27, + "learning_rate": 1.710970113559435e-05, + "loss": 0.3809, + "step": 4483 + }, + { + "epoch": 0.27, + "learning_rate": 1.710832504174918e-05, + "loss": 0.3429, + "step": 4484 + }, + { + "epoch": 0.27, + "learning_rate": 1.7106948675764056e-05, + "loss": 0.351, + "step": 4485 + }, + { + "epoch": 0.27, + "learning_rate": 1.710557203769167e-05, + "loss": 0.3429, + "step": 4486 + }, + { + "epoch": 0.27, + "learning_rate": 1.7104195127584735e-05, + "loss": 0.3588, + "step": 4487 + }, + { + "epoch": 0.27, + "learning_rate": 1.7102817945495958e-05, + "loss": 0.3262, + "step": 4488 + }, + { + "epoch": 0.27, + "learning_rate": 1.710144049147807e-05, + "loss": 0.3443, + "step": 4489 + }, + { + "epoch": 0.27, + "learning_rate": 1.71000627655838e-05, + "loss": 0.3574, + "step": 4490 + }, + { + "epoch": 0.27, + "learning_rate": 1.7098684767865896e-05, + "loss": 0.3689, + "step": 4491 + }, + { + "epoch": 0.27, + "learning_rate": 1.7097306498377116e-05, + "loss": 0.3569, + "step": 4492 + }, + { + "epoch": 0.27, + "learning_rate": 1.709592795717023e-05, + "loss": 0.3396, + "step": 4493 + }, + { + "epoch": 0.27, + "learning_rate": 1.7094549144298006e-05, + "loss": 0.3471, + "step": 4494 + }, + { + "epoch": 0.27, + "learning_rate": 1.7093170059813237e-05, + "loss": 0.3376, + "step": 4495 + }, + { + "epoch": 0.27, + "learning_rate": 1.709179070376872e-05, + "loss": 0.3415, + "step": 4496 + }, + { + "epoch": 0.27, + "learning_rate": 1.709041107621726e-05, + "loss": 0.3499, + "step": 4497 + }, + { + "epoch": 0.27, + "learning_rate": 1.708903117721168e-05, + "loss": 0.3284, + "step": 4498 + }, + { + "epoch": 0.27, + "learning_rate": 1.708765100680481e-05, + "loss": 0.3468, + "step": 4499 + }, + { + "epoch": 0.27, + "learning_rate": 1.7086270565049486e-05, + "loss": 0.3538, + "step": 4500 + }, + { + "epoch": 0.27, + "learning_rate": 1.7084889851998556e-05, + "loss": 0.3569, + "step": 4501 + }, + { + "epoch": 0.27, + "learning_rate": 1.7083508867704887e-05, + "loss": 0.3538, + "step": 4502 + }, + { + "epoch": 0.27, + "learning_rate": 1.7082127612221344e-05, + "loss": 0.3156, + "step": 4503 + }, + { + "epoch": 0.27, + "learning_rate": 1.708074608560081e-05, + "loss": 0.3694, + "step": 4504 + }, + { + "epoch": 0.27, + "learning_rate": 1.7079364287896174e-05, + "loss": 0.3407, + "step": 4505 + }, + { + "epoch": 0.27, + "learning_rate": 1.707798221916034e-05, + "loss": 0.3744, + "step": 4506 + }, + { + "epoch": 0.27, + "learning_rate": 1.7076599879446217e-05, + "loss": 0.3585, + "step": 4507 + }, + { + "epoch": 0.27, + "learning_rate": 1.7075217268806734e-05, + "loss": 0.3203, + "step": 4508 + }, + { + "epoch": 0.27, + "learning_rate": 1.707383438729482e-05, + "loss": 0.3191, + "step": 4509 + }, + { + "epoch": 0.27, + "learning_rate": 1.7072451234963415e-05, + "loss": 0.3354, + "step": 4510 + }, + { + "epoch": 0.27, + "learning_rate": 1.7071067811865477e-05, + "loss": 0.3691, + "step": 4511 + }, + { + "epoch": 0.27, + "learning_rate": 1.706968411805397e-05, + "loss": 0.3549, + "step": 4512 + }, + { + "epoch": 0.27, + "learning_rate": 1.706830015358186e-05, + "loss": 0.3429, + "step": 4513 + }, + { + "epoch": 0.27, + "learning_rate": 1.7066915918502144e-05, + "loss": 0.3535, + "step": 4514 + }, + { + "epoch": 0.27, + "learning_rate": 1.7065531412867812e-05, + "loss": 0.3362, + "step": 4515 + }, + { + "epoch": 0.27, + "learning_rate": 1.7064146636731865e-05, + "loss": 0.3371, + "step": 4516 + }, + { + "epoch": 0.27, + "learning_rate": 1.7062761590147325e-05, + "loss": 0.3531, + "step": 4517 + }, + { + "epoch": 0.27, + "learning_rate": 1.7061376273167214e-05, + "loss": 0.3449, + "step": 4518 + }, + { + "epoch": 0.27, + "learning_rate": 1.7059990685844568e-05, + "loss": 0.3186, + "step": 4519 + }, + { + "epoch": 0.27, + "learning_rate": 1.7058604828232438e-05, + "loss": 0.3691, + "step": 4520 + }, + { + "epoch": 0.27, + "learning_rate": 1.7057218700383876e-05, + "loss": 0.3585, + "step": 4521 + }, + { + "epoch": 0.27, + "learning_rate": 1.7055832302351956e-05, + "loss": 0.3396, + "step": 4522 + }, + { + "epoch": 0.27, + "learning_rate": 1.705444563418975e-05, + "loss": 0.3605, + "step": 4523 + }, + { + "epoch": 0.27, + "learning_rate": 1.705305869595035e-05, + "loss": 0.3764, + "step": 4524 + }, + { + "epoch": 0.27, + "learning_rate": 1.7051671487686848e-05, + "loss": 0.3479, + "step": 4525 + }, + { + "epoch": 0.27, + "learning_rate": 1.705028400945236e-05, + "loss": 0.358, + "step": 4526 + }, + { + "epoch": 0.27, + "learning_rate": 1.7048896261300002e-05, + "loss": 0.3619, + "step": 4527 + }, + { + "epoch": 0.27, + "learning_rate": 1.7047508243282903e-05, + "loss": 0.3491, + "step": 4528 + }, + { + "epoch": 0.27, + "learning_rate": 1.704611995545421e-05, + "loss": 0.3393, + "step": 4529 + }, + { + "epoch": 0.27, + "learning_rate": 1.704473139786706e-05, + "loss": 0.3451, + "step": 4530 + }, + { + "epoch": 0.27, + "learning_rate": 1.7043342570574622e-05, + "loss": 0.3076, + "step": 4531 + }, + { + "epoch": 0.27, + "learning_rate": 1.704195347363007e-05, + "loss": 0.3139, + "step": 4532 + }, + { + "epoch": 0.27, + "learning_rate": 1.7040564107086573e-05, + "loss": 0.3259, + "step": 4533 + }, + { + "epoch": 0.27, + "learning_rate": 1.7039174470997336e-05, + "loss": 0.3465, + "step": 4534 + }, + { + "epoch": 0.27, + "learning_rate": 1.703778456541555e-05, + "loss": 0.3705, + "step": 4535 + }, + { + "epoch": 0.27, + "learning_rate": 1.7036394390394435e-05, + "loss": 0.3493, + "step": 4536 + }, + { + "epoch": 0.27, + "learning_rate": 1.703500394598721e-05, + "loss": 0.3468, + "step": 4537 + }, + { + "epoch": 0.27, + "learning_rate": 1.7033613232247104e-05, + "loss": 0.3239, + "step": 4538 + }, + { + "epoch": 0.27, + "learning_rate": 1.703222224922737e-05, + "loss": 0.3725, + "step": 4539 + }, + { + "epoch": 0.27, + "learning_rate": 1.7030830996981254e-05, + "loss": 0.3622, + "step": 4540 + }, + { + "epoch": 0.27, + "learning_rate": 1.7029439475562017e-05, + "loss": 0.3507, + "step": 4541 + }, + { + "epoch": 0.27, + "learning_rate": 1.702804768502294e-05, + "loss": 0.3234, + "step": 4542 + }, + { + "epoch": 0.27, + "learning_rate": 1.70266556254173e-05, + "loss": 0.3404, + "step": 4543 + }, + { + "epoch": 0.27, + "learning_rate": 1.7025263296798403e-05, + "loss": 0.3211, + "step": 4544 + }, + { + "epoch": 0.27, + "learning_rate": 1.702387069921954e-05, + "loss": 0.3599, + "step": 4545 + }, + { + "epoch": 0.27, + "learning_rate": 1.7022477832734034e-05, + "loss": 0.3292, + "step": 4546 + }, + { + "epoch": 0.27, + "learning_rate": 1.702108469739521e-05, + "loss": 0.3398, + "step": 4547 + }, + { + "epoch": 0.27, + "learning_rate": 1.70196912932564e-05, + "loss": 0.3338, + "step": 4548 + }, + { + "epoch": 0.27, + "learning_rate": 1.7018297620370957e-05, + "loss": 0.3485, + "step": 4549 + }, + { + "epoch": 0.27, + "learning_rate": 1.7016903678792232e-05, + "loss": 0.3538, + "step": 4550 + }, + { + "epoch": 0.27, + "learning_rate": 1.7015509468573593e-05, + "loss": 0.3532, + "step": 4551 + }, + { + "epoch": 0.27, + "learning_rate": 1.701411498976842e-05, + "loss": 0.3396, + "step": 4552 + }, + { + "epoch": 0.28, + "learning_rate": 1.701272024243009e-05, + "loss": 0.3458, + "step": 4553 + }, + { + "epoch": 0.28, + "learning_rate": 1.7011325226612016e-05, + "loss": 0.361, + "step": 4554 + }, + { + "epoch": 0.28, + "learning_rate": 1.700992994236759e-05, + "loss": 0.329, + "step": 4555 + }, + { + "epoch": 0.28, + "learning_rate": 1.7008534389750238e-05, + "loss": 0.3541, + "step": 4556 + }, + { + "epoch": 0.28, + "learning_rate": 1.7007138568813387e-05, + "loss": 0.3834, + "step": 4557 + }, + { + "epoch": 0.28, + "learning_rate": 1.700574247961048e-05, + "loss": 0.3072, + "step": 4558 + }, + { + "epoch": 0.28, + "learning_rate": 1.7004346122194957e-05, + "loss": 0.3373, + "step": 4559 + }, + { + "epoch": 0.28, + "learning_rate": 1.7002949496620286e-05, + "loss": 0.3697, + "step": 4560 + }, + { + "epoch": 0.28, + "learning_rate": 1.7001552602939928e-05, + "loss": 0.3661, + "step": 4561 + }, + { + "epoch": 0.28, + "learning_rate": 1.700015544120737e-05, + "loss": 0.3569, + "step": 4562 + }, + { + "epoch": 0.28, + "learning_rate": 1.6998758011476096e-05, + "loss": 0.3354, + "step": 4563 + }, + { + "epoch": 0.28, + "learning_rate": 1.6997360313799613e-05, + "loss": 0.3516, + "step": 4564 + }, + { + "epoch": 0.28, + "learning_rate": 1.6995962348231424e-05, + "loss": 0.3348, + "step": 4565 + }, + { + "epoch": 0.28, + "learning_rate": 1.6994564114825057e-05, + "loss": 0.3382, + "step": 4566 + }, + { + "epoch": 0.28, + "learning_rate": 1.699316561363403e-05, + "loss": 0.3284, + "step": 4567 + }, + { + "epoch": 0.28, + "learning_rate": 1.69917668447119e-05, + "loss": 0.3633, + "step": 4568 + }, + { + "epoch": 0.28, + "learning_rate": 1.6990367808112212e-05, + "loss": 0.3371, + "step": 4569 + }, + { + "epoch": 0.28, + "learning_rate": 1.698896850388853e-05, + "loss": 0.3362, + "step": 4570 + }, + { + "epoch": 0.28, + "learning_rate": 1.6987568932094417e-05, + "loss": 0.339, + "step": 4571 + }, + { + "epoch": 0.28, + "learning_rate": 1.6986169092783462e-05, + "loss": 0.358, + "step": 4572 + }, + { + "epoch": 0.28, + "learning_rate": 1.6984768986009258e-05, + "loss": 0.377, + "step": 4573 + }, + { + "epoch": 0.28, + "learning_rate": 1.6983368611825403e-05, + "loss": 0.385, + "step": 4574 + }, + { + "epoch": 0.28, + "learning_rate": 1.698196797028552e-05, + "loss": 0.3516, + "step": 4575 + }, + { + "epoch": 0.28, + "learning_rate": 1.6980567061443222e-05, + "loss": 0.3376, + "step": 4576 + }, + { + "epoch": 0.28, + "learning_rate": 1.697916588535214e-05, + "loss": 0.3583, + "step": 4577 + }, + { + "epoch": 0.28, + "learning_rate": 1.697776444206593e-05, + "loss": 0.3504, + "step": 4578 + }, + { + "epoch": 0.28, + "learning_rate": 1.697636273163824e-05, + "loss": 0.3248, + "step": 4579 + }, + { + "epoch": 0.28, + "learning_rate": 1.6974960754122726e-05, + "loss": 0.3527, + "step": 4580 + }, + { + "epoch": 0.28, + "learning_rate": 1.6973558509573074e-05, + "loss": 0.3393, + "step": 4581 + }, + { + "epoch": 0.28, + "learning_rate": 1.6972155998042964e-05, + "loss": 0.3301, + "step": 4582 + }, + { + "epoch": 0.28, + "learning_rate": 1.6970753219586087e-05, + "loss": 0.3368, + "step": 4583 + }, + { + "epoch": 0.28, + "learning_rate": 1.6969350174256155e-05, + "loss": 0.3817, + "step": 4584 + }, + { + "epoch": 0.28, + "learning_rate": 1.6967946862106878e-05, + "loss": 0.3703, + "step": 4585 + }, + { + "epoch": 0.28, + "learning_rate": 1.6966543283191983e-05, + "loss": 0.3524, + "step": 4586 + }, + { + "epoch": 0.28, + "learning_rate": 1.6965139437565204e-05, + "loss": 0.3418, + "step": 4587 + }, + { + "epoch": 0.28, + "learning_rate": 1.6963735325280293e-05, + "loss": 0.3789, + "step": 4588 + }, + { + "epoch": 0.28, + "learning_rate": 1.6962330946391e-05, + "loss": 0.3078, + "step": 4589 + }, + { + "epoch": 0.28, + "learning_rate": 1.696092630095109e-05, + "loss": 0.3245, + "step": 4590 + }, + { + "epoch": 0.28, + "learning_rate": 1.695952138901434e-05, + "loss": 0.3354, + "step": 4591 + }, + { + "epoch": 0.28, + "learning_rate": 1.6958116210634546e-05, + "loss": 0.3245, + "step": 4592 + }, + { + "epoch": 0.28, + "learning_rate": 1.695671076586549e-05, + "loss": 0.3394, + "step": 4593 + }, + { + "epoch": 0.28, + "learning_rate": 1.695530505476099e-05, + "loss": 0.3387, + "step": 4594 + }, + { + "epoch": 0.28, + "learning_rate": 1.6953899077374858e-05, + "loss": 0.3474, + "step": 4595 + }, + { + "epoch": 0.28, + "learning_rate": 1.6952492833760922e-05, + "loss": 0.3694, + "step": 4596 + }, + { + "epoch": 0.28, + "learning_rate": 1.6951086323973022e-05, + "loss": 0.3404, + "step": 4597 + }, + { + "epoch": 0.28, + "learning_rate": 1.6949679548065002e-05, + "loss": 0.3262, + "step": 4598 + }, + { + "epoch": 0.28, + "learning_rate": 1.6948272506090722e-05, + "loss": 0.3435, + "step": 4599 + }, + { + "epoch": 0.28, + "learning_rate": 1.6946865198104055e-05, + "loss": 0.3292, + "step": 4600 + }, + { + "epoch": 0.28, + "learning_rate": 1.694545762415887e-05, + "loss": 0.3401, + "step": 4601 + }, + { + "epoch": 0.28, + "learning_rate": 1.6944049784309063e-05, + "loss": 0.356, + "step": 4602 + }, + { + "epoch": 0.28, + "learning_rate": 1.6942641678608528e-05, + "loss": 0.3725, + "step": 4603 + }, + { + "epoch": 0.28, + "learning_rate": 1.6941233307111175e-05, + "loss": 0.3075, + "step": 4604 + }, + { + "epoch": 0.28, + "learning_rate": 1.6939824669870926e-05, + "loss": 0.3323, + "step": 4605 + }, + { + "epoch": 0.28, + "learning_rate": 1.693841576694171e-05, + "loss": 0.3398, + "step": 4606 + }, + { + "epoch": 0.28, + "learning_rate": 1.693700659837746e-05, + "loss": 0.3246, + "step": 4607 + }, + { + "epoch": 0.28, + "learning_rate": 1.6935597164232132e-05, + "loss": 0.3504, + "step": 4608 + }, + { + "epoch": 0.28, + "learning_rate": 1.6934187464559684e-05, + "loss": 0.3675, + "step": 4609 + }, + { + "epoch": 0.28, + "learning_rate": 1.6932777499414087e-05, + "loss": 0.3368, + "step": 4610 + }, + { + "epoch": 0.28, + "learning_rate": 1.6931367268849316e-05, + "loss": 0.3583, + "step": 4611 + }, + { + "epoch": 0.28, + "learning_rate": 1.6929956772919365e-05, + "loss": 0.3524, + "step": 4612 + }, + { + "epoch": 0.28, + "learning_rate": 1.692854601167824e-05, + "loss": 0.3312, + "step": 4613 + }, + { + "epoch": 0.28, + "learning_rate": 1.692713498517994e-05, + "loss": 0.3301, + "step": 4614 + }, + { + "epoch": 0.28, + "learning_rate": 1.6925723693478494e-05, + "loss": 0.3217, + "step": 4615 + }, + { + "epoch": 0.28, + "learning_rate": 1.692431213662793e-05, + "loss": 0.3185, + "step": 4616 + }, + { + "epoch": 0.28, + "learning_rate": 1.6922900314682288e-05, + "loss": 0.3446, + "step": 4617 + }, + { + "epoch": 0.28, + "learning_rate": 1.6921488227695624e-05, + "loss": 0.3384, + "step": 4618 + }, + { + "epoch": 0.28, + "learning_rate": 1.6920075875721996e-05, + "loss": 0.3658, + "step": 4619 + }, + { + "epoch": 0.28, + "learning_rate": 1.6918663258815474e-05, + "loss": 0.3295, + "step": 4620 + }, + { + "epoch": 0.28, + "learning_rate": 1.6917250377030138e-05, + "loss": 0.3491, + "step": 4621 + }, + { + "epoch": 0.28, + "learning_rate": 1.6915837230420086e-05, + "loss": 0.3516, + "step": 4622 + }, + { + "epoch": 0.28, + "learning_rate": 1.6914423819039414e-05, + "loss": 0.3627, + "step": 4623 + }, + { + "epoch": 0.28, + "learning_rate": 1.6913010142942237e-05, + "loss": 0.3387, + "step": 4624 + }, + { + "epoch": 0.28, + "learning_rate": 1.691159620218268e-05, + "loss": 0.3411, + "step": 4625 + }, + { + "epoch": 0.28, + "learning_rate": 1.691018199681487e-05, + "loss": 0.3253, + "step": 4626 + }, + { + "epoch": 0.28, + "learning_rate": 1.6908767526892948e-05, + "loss": 0.3457, + "step": 4627 + }, + { + "epoch": 0.28, + "learning_rate": 1.6907352792471075e-05, + "loss": 0.3331, + "step": 4628 + }, + { + "epoch": 0.28, + "learning_rate": 1.6905937793603405e-05, + "loss": 0.3557, + "step": 4629 + }, + { + "epoch": 0.28, + "learning_rate": 1.6904522530344114e-05, + "loss": 0.3552, + "step": 4630 + }, + { + "epoch": 0.28, + "learning_rate": 1.690310700274739e-05, + "loss": 0.3373, + "step": 4631 + }, + { + "epoch": 0.28, + "learning_rate": 1.6901691210867416e-05, + "loss": 0.3532, + "step": 4632 + }, + { + "epoch": 0.28, + "learning_rate": 1.69002751547584e-05, + "loss": 0.3446, + "step": 4633 + }, + { + "epoch": 0.28, + "learning_rate": 1.689885883447456e-05, + "loss": 0.3591, + "step": 4634 + }, + { + "epoch": 0.28, + "learning_rate": 1.689744225007011e-05, + "loss": 0.351, + "step": 4635 + }, + { + "epoch": 0.28, + "learning_rate": 1.6896025401599293e-05, + "loss": 0.3318, + "step": 4636 + }, + { + "epoch": 0.28, + "learning_rate": 1.6894608289116344e-05, + "loss": 0.3418, + "step": 4637 + }, + { + "epoch": 0.28, + "learning_rate": 1.6893190912675523e-05, + "loss": 0.3338, + "step": 4638 + }, + { + "epoch": 0.28, + "learning_rate": 1.689177327233109e-05, + "loss": 0.3203, + "step": 4639 + }, + { + "epoch": 0.28, + "learning_rate": 1.6890355368137324e-05, + "loss": 0.3465, + "step": 4640 + }, + { + "epoch": 0.28, + "learning_rate": 1.6888937200148503e-05, + "loss": 0.3446, + "step": 4641 + }, + { + "epoch": 0.28, + "learning_rate": 1.6887518768418925e-05, + "loss": 0.3298, + "step": 4642 + }, + { + "epoch": 0.28, + "learning_rate": 1.688610007300289e-05, + "loss": 0.3301, + "step": 4643 + }, + { + "epoch": 0.28, + "learning_rate": 1.6884681113954716e-05, + "loss": 0.3119, + "step": 4644 + }, + { + "epoch": 0.28, + "learning_rate": 1.6883261891328726e-05, + "loss": 0.3331, + "step": 4645 + }, + { + "epoch": 0.28, + "learning_rate": 1.6881842405179257e-05, + "loss": 0.361, + "step": 4646 + }, + { + "epoch": 0.28, + "learning_rate": 1.688042265556065e-05, + "loss": 0.3189, + "step": 4647 + }, + { + "epoch": 0.28, + "learning_rate": 1.687900264252726e-05, + "loss": 0.4174, + "step": 4648 + }, + { + "epoch": 0.28, + "learning_rate": 1.6877582366133457e-05, + "loss": 0.3331, + "step": 4649 + }, + { + "epoch": 0.28, + "learning_rate": 1.6876161826433607e-05, + "loss": 0.3295, + "step": 4650 + }, + { + "epoch": 0.28, + "learning_rate": 1.6874741023482103e-05, + "loss": 0.3357, + "step": 4651 + }, + { + "epoch": 0.28, + "learning_rate": 1.6873319957333336e-05, + "loss": 0.3384, + "step": 4652 + }, + { + "epoch": 0.28, + "learning_rate": 1.687189862804171e-05, + "loss": 0.3619, + "step": 4653 + }, + { + "epoch": 0.28, + "learning_rate": 1.6870477035661643e-05, + "loss": 0.3343, + "step": 4654 + }, + { + "epoch": 0.28, + "learning_rate": 1.686905518024756e-05, + "loss": 0.3188, + "step": 4655 + }, + { + "epoch": 0.28, + "learning_rate": 1.6867633061853894e-05, + "loss": 0.3449, + "step": 4656 + }, + { + "epoch": 0.28, + "learning_rate": 1.6866210680535095e-05, + "loss": 0.3853, + "step": 4657 + }, + { + "epoch": 0.28, + "learning_rate": 1.686478803634561e-05, + "loss": 0.3516, + "step": 4658 + }, + { + "epoch": 0.28, + "learning_rate": 1.686336512933991e-05, + "loss": 0.3424, + "step": 4659 + }, + { + "epoch": 0.28, + "learning_rate": 1.6861941959572473e-05, + "loss": 0.3454, + "step": 4660 + }, + { + "epoch": 0.28, + "learning_rate": 1.686051852709778e-05, + "loss": 0.3245, + "step": 4661 + }, + { + "epoch": 0.28, + "learning_rate": 1.685909483197033e-05, + "loss": 0.3262, + "step": 4662 + }, + { + "epoch": 0.28, + "learning_rate": 1.6857670874244624e-05, + "loss": 0.3535, + "step": 4663 + }, + { + "epoch": 0.28, + "learning_rate": 1.6856246653975184e-05, + "loss": 0.3306, + "step": 4664 + }, + { + "epoch": 0.28, + "learning_rate": 1.6854822171216533e-05, + "loss": 0.3407, + "step": 4665 + }, + { + "epoch": 0.28, + "learning_rate": 1.6853397426023202e-05, + "loss": 0.3479, + "step": 4666 + }, + { + "epoch": 0.28, + "learning_rate": 1.6851972418449743e-05, + "loss": 0.3538, + "step": 4667 + }, + { + "epoch": 0.28, + "learning_rate": 1.6850547148550713e-05, + "loss": 0.3265, + "step": 4668 + }, + { + "epoch": 0.28, + "learning_rate": 1.6849121616380677e-05, + "loss": 0.3041, + "step": 4669 + }, + { + "epoch": 0.28, + "learning_rate": 1.6847695821994207e-05, + "loss": 0.3354, + "step": 4670 + }, + { + "epoch": 0.28, + "learning_rate": 1.684626976544589e-05, + "loss": 0.3566, + "step": 4671 + }, + { + "epoch": 0.28, + "learning_rate": 1.6844843446790323e-05, + "loss": 0.3278, + "step": 4672 + }, + { + "epoch": 0.28, + "learning_rate": 1.6843416866082118e-05, + "loss": 0.3507, + "step": 4673 + }, + { + "epoch": 0.28, + "learning_rate": 1.684199002337588e-05, + "loss": 0.3195, + "step": 4674 + }, + { + "epoch": 0.28, + "learning_rate": 1.6840562918726245e-05, + "loss": 0.3624, + "step": 4675 + }, + { + "epoch": 0.28, + "learning_rate": 1.6839135552187844e-05, + "loss": 0.3214, + "step": 4676 + }, + { + "epoch": 0.28, + "learning_rate": 1.6837707923815324e-05, + "loss": 0.3362, + "step": 4677 + }, + { + "epoch": 0.28, + "learning_rate": 1.6836280033663344e-05, + "loss": 0.3396, + "step": 4678 + }, + { + "epoch": 0.28, + "learning_rate": 1.683485188178657e-05, + "loss": 0.3432, + "step": 4679 + }, + { + "epoch": 0.28, + "learning_rate": 1.683342346823967e-05, + "loss": 0.3477, + "step": 4680 + }, + { + "epoch": 0.28, + "learning_rate": 1.6831994793077343e-05, + "loss": 0.3281, + "step": 4681 + }, + { + "epoch": 0.28, + "learning_rate": 1.6830565856354276e-05, + "loss": 0.3502, + "step": 4682 + }, + { + "epoch": 0.28, + "learning_rate": 1.6829136658125182e-05, + "loss": 0.3192, + "step": 4683 + }, + { + "epoch": 0.28, + "learning_rate": 1.682770719844477e-05, + "loss": 0.3412, + "step": 4684 + }, + { + "epoch": 0.28, + "learning_rate": 1.6826277477367775e-05, + "loss": 0.3398, + "step": 4685 + }, + { + "epoch": 0.28, + "learning_rate": 1.6824847494948928e-05, + "loss": 0.3334, + "step": 4686 + }, + { + "epoch": 0.28, + "learning_rate": 1.6823417251242975e-05, + "loss": 0.341, + "step": 4687 + }, + { + "epoch": 0.28, + "learning_rate": 1.682198674630467e-05, + "loss": 0.3557, + "step": 4688 + }, + { + "epoch": 0.28, + "learning_rate": 1.6820555980188793e-05, + "loss": 0.3262, + "step": 4689 + }, + { + "epoch": 0.28, + "learning_rate": 1.6819124952950106e-05, + "loss": 0.3658, + "step": 4690 + }, + { + "epoch": 0.28, + "learning_rate": 1.68176936646434e-05, + "loss": 0.3318, + "step": 4691 + }, + { + "epoch": 0.28, + "learning_rate": 1.681626211532347e-05, + "loss": 0.3532, + "step": 4692 + }, + { + "epoch": 0.28, + "learning_rate": 1.6814830305045125e-05, + "loss": 0.3597, + "step": 4693 + }, + { + "epoch": 0.28, + "learning_rate": 1.681339823386318e-05, + "loss": 0.3384, + "step": 4694 + }, + { + "epoch": 0.28, + "learning_rate": 1.6811965901832464e-05, + "loss": 0.3502, + "step": 4695 + }, + { + "epoch": 0.28, + "learning_rate": 1.681053330900781e-05, + "loss": 0.3532, + "step": 4696 + }, + { + "epoch": 0.28, + "learning_rate": 1.680910045544406e-05, + "loss": 0.361, + "step": 4697 + }, + { + "epoch": 0.28, + "learning_rate": 1.680766734119608e-05, + "loss": 0.334, + "step": 4698 + }, + { + "epoch": 0.28, + "learning_rate": 1.6806233966318736e-05, + "loss": 0.3493, + "step": 4699 + }, + { + "epoch": 0.28, + "learning_rate": 1.6804800330866897e-05, + "loss": 0.3273, + "step": 4700 + }, + { + "epoch": 0.28, + "learning_rate": 1.6803366434895453e-05, + "loss": 0.3652, + "step": 4701 + }, + { + "epoch": 0.28, + "learning_rate": 1.68019322784593e-05, + "loss": 0.3616, + "step": 4702 + }, + { + "epoch": 0.28, + "learning_rate": 1.6800497861613343e-05, + "loss": 0.3454, + "step": 4703 + }, + { + "epoch": 0.28, + "learning_rate": 1.67990631844125e-05, + "loss": 0.3705, + "step": 4704 + }, + { + "epoch": 0.28, + "learning_rate": 1.6797628246911696e-05, + "loss": 0.3368, + "step": 4705 + }, + { + "epoch": 0.28, + "learning_rate": 1.6796193049165866e-05, + "loss": 0.3359, + "step": 4706 + }, + { + "epoch": 0.28, + "learning_rate": 1.6794757591229962e-05, + "loss": 0.3594, + "step": 4707 + }, + { + "epoch": 0.28, + "learning_rate": 1.6793321873158935e-05, + "loss": 0.3591, + "step": 4708 + }, + { + "epoch": 0.28, + "learning_rate": 1.6791885895007748e-05, + "loss": 0.3477, + "step": 4709 + }, + { + "epoch": 0.28, + "learning_rate": 1.6790449656831386e-05, + "loss": 0.3345, + "step": 4710 + }, + { + "epoch": 0.28, + "learning_rate": 1.678901315868483e-05, + "loss": 0.3527, + "step": 4711 + }, + { + "epoch": 0.28, + "learning_rate": 1.678757640062307e-05, + "loss": 0.37, + "step": 4712 + }, + { + "epoch": 0.28, + "learning_rate": 1.678613938270112e-05, + "loss": 0.3652, + "step": 4713 + }, + { + "epoch": 0.28, + "learning_rate": 1.678470210497399e-05, + "loss": 0.3267, + "step": 4714 + }, + { + "epoch": 0.28, + "learning_rate": 1.6783264567496713e-05, + "loss": 0.3454, + "step": 4715 + }, + { + "epoch": 0.28, + "learning_rate": 1.6781826770324322e-05, + "loss": 0.3276, + "step": 4716 + }, + { + "epoch": 0.28, + "learning_rate": 1.6780388713511854e-05, + "loss": 0.3569, + "step": 4717 + }, + { + "epoch": 0.29, + "learning_rate": 1.677895039711438e-05, + "loss": 0.3281, + "step": 4718 + }, + { + "epoch": 0.29, + "learning_rate": 1.677751182118695e-05, + "loss": 0.3404, + "step": 4719 + }, + { + "epoch": 0.29, + "learning_rate": 1.6776072985784652e-05, + "loss": 0.3859, + "step": 4720 + }, + { + "epoch": 0.29, + "learning_rate": 1.6774633890962563e-05, + "loss": 0.3398, + "step": 4721 + }, + { + "epoch": 0.29, + "learning_rate": 1.677319453677578e-05, + "loss": 0.34, + "step": 4722 + }, + { + "epoch": 0.29, + "learning_rate": 1.6771754923279413e-05, + "loss": 0.356, + "step": 4723 + }, + { + "epoch": 0.29, + "learning_rate": 1.6770315050528573e-05, + "loss": 0.3552, + "step": 4724 + }, + { + "epoch": 0.29, + "learning_rate": 1.6768874918578385e-05, + "loss": 0.3418, + "step": 4725 + }, + { + "epoch": 0.29, + "learning_rate": 1.6767434527483985e-05, + "loss": 0.3719, + "step": 4726 + }, + { + "epoch": 0.29, + "learning_rate": 1.6765993877300517e-05, + "loss": 0.3195, + "step": 4727 + }, + { + "epoch": 0.29, + "learning_rate": 1.6764552968083135e-05, + "loss": 0.3605, + "step": 4728 + }, + { + "epoch": 0.29, + "learning_rate": 1.6763111799887008e-05, + "loss": 0.344, + "step": 4729 + }, + { + "epoch": 0.29, + "learning_rate": 1.67616703727673e-05, + "loss": 0.3304, + "step": 4730 + }, + { + "epoch": 0.29, + "learning_rate": 1.6760228686779215e-05, + "loss": 0.3037, + "step": 4731 + }, + { + "epoch": 0.29, + "learning_rate": 1.675878674197793e-05, + "loss": 0.3111, + "step": 4732 + }, + { + "epoch": 0.29, + "learning_rate": 1.6757344538418655e-05, + "loss": 0.3742, + "step": 4733 + }, + { + "epoch": 0.29, + "learning_rate": 1.6755902076156606e-05, + "loss": 0.3479, + "step": 4734 + }, + { + "epoch": 0.29, + "learning_rate": 1.6754459355247005e-05, + "loss": 0.3278, + "step": 4735 + }, + { + "epoch": 0.29, + "learning_rate": 1.6753016375745087e-05, + "loss": 0.3672, + "step": 4736 + }, + { + "epoch": 0.29, + "learning_rate": 1.6751573137706094e-05, + "loss": 0.3457, + "step": 4737 + }, + { + "epoch": 0.29, + "learning_rate": 1.675012964118528e-05, + "loss": 0.3471, + "step": 4738 + }, + { + "epoch": 0.29, + "learning_rate": 1.6748685886237918e-05, + "loss": 0.3546, + "step": 4739 + }, + { + "epoch": 0.29, + "learning_rate": 1.674724187291927e-05, + "loss": 0.3775, + "step": 4740 + }, + { + "epoch": 0.29, + "learning_rate": 1.6745797601284623e-05, + "loss": 0.3555, + "step": 4741 + }, + { + "epoch": 0.29, + "learning_rate": 1.674435307138927e-05, + "loss": 0.3689, + "step": 4742 + }, + { + "epoch": 0.29, + "learning_rate": 1.674290828328852e-05, + "loss": 0.3368, + "step": 4743 + }, + { + "epoch": 0.29, + "learning_rate": 1.674146323703768e-05, + "loss": 0.3465, + "step": 4744 + }, + { + "epoch": 0.29, + "learning_rate": 1.6740017932692073e-05, + "loss": 0.3354, + "step": 4745 + }, + { + "epoch": 0.29, + "learning_rate": 1.673857237030704e-05, + "loss": 0.361, + "step": 4746 + }, + { + "epoch": 0.29, + "learning_rate": 1.6737126549937914e-05, + "loss": 0.3147, + "step": 4747 + }, + { + "epoch": 0.29, + "learning_rate": 1.673568047164005e-05, + "loss": 0.3373, + "step": 4748 + }, + { + "epoch": 0.29, + "learning_rate": 1.6734234135468815e-05, + "loss": 0.3739, + "step": 4749 + }, + { + "epoch": 0.29, + "learning_rate": 1.673278754147958e-05, + "loss": 0.3541, + "step": 4750 + }, + { + "epoch": 0.29, + "learning_rate": 1.673134068972772e-05, + "loss": 0.3415, + "step": 4751 + }, + { + "epoch": 0.29, + "learning_rate": 1.672989358026864e-05, + "loss": 0.3485, + "step": 4752 + }, + { + "epoch": 0.29, + "learning_rate": 1.672844621315773e-05, + "loss": 0.322, + "step": 4753 + }, + { + "epoch": 0.29, + "learning_rate": 1.6726998588450413e-05, + "loss": 0.3373, + "step": 4754 + }, + { + "epoch": 0.29, + "learning_rate": 1.6725550706202106e-05, + "loss": 0.3239, + "step": 4755 + }, + { + "epoch": 0.29, + "learning_rate": 1.6724102566468235e-05, + "loss": 0.3309, + "step": 4756 + }, + { + "epoch": 0.29, + "learning_rate": 1.6722654169304253e-05, + "loss": 0.3513, + "step": 4757 + }, + { + "epoch": 0.29, + "learning_rate": 1.6721205514765604e-05, + "loss": 0.3223, + "step": 4758 + }, + { + "epoch": 0.29, + "learning_rate": 1.6719756602907747e-05, + "loss": 0.341, + "step": 4759 + }, + { + "epoch": 0.29, + "learning_rate": 1.671830743378616e-05, + "loss": 0.3468, + "step": 4760 + }, + { + "epoch": 0.29, + "learning_rate": 1.6716858007456317e-05, + "loss": 0.3119, + "step": 4761 + }, + { + "epoch": 0.29, + "learning_rate": 1.6715408323973715e-05, + "loss": 0.3485, + "step": 4762 + }, + { + "epoch": 0.29, + "learning_rate": 1.6713958383393852e-05, + "loss": 0.3361, + "step": 4763 + }, + { + "epoch": 0.29, + "learning_rate": 1.6712508185772237e-05, + "loss": 0.32, + "step": 4764 + }, + { + "epoch": 0.29, + "learning_rate": 1.6711057731164392e-05, + "loss": 0.3348, + "step": 4765 + }, + { + "epoch": 0.29, + "learning_rate": 1.670960701962585e-05, + "loss": 0.3655, + "step": 4766 + }, + { + "epoch": 0.29, + "learning_rate": 1.6708156051212147e-05, + "loss": 0.3634, + "step": 4767 + }, + { + "epoch": 0.29, + "learning_rate": 1.6706704825978833e-05, + "loss": 0.3747, + "step": 4768 + }, + { + "epoch": 0.29, + "learning_rate": 1.670525334398147e-05, + "loss": 0.3513, + "step": 4769 + }, + { + "epoch": 0.29, + "learning_rate": 1.6703801605275626e-05, + "loss": 0.3396, + "step": 4770 + }, + { + "epoch": 0.29, + "learning_rate": 1.670234960991688e-05, + "loss": 0.3195, + "step": 4771 + }, + { + "epoch": 0.29, + "learning_rate": 1.6700897357960823e-05, + "loss": 0.3318, + "step": 4772 + }, + { + "epoch": 0.29, + "learning_rate": 1.6699444849463053e-05, + "loss": 0.3583, + "step": 4773 + }, + { + "epoch": 0.29, + "learning_rate": 1.6697992084479176e-05, + "loss": 0.3424, + "step": 4774 + }, + { + "epoch": 0.29, + "learning_rate": 1.6696539063064813e-05, + "loss": 0.358, + "step": 4775 + }, + { + "epoch": 0.29, + "learning_rate": 1.6695085785275597e-05, + "loss": 0.3018, + "step": 4776 + }, + { + "epoch": 0.29, + "learning_rate": 1.669363225116716e-05, + "loss": 0.3323, + "step": 4777 + }, + { + "epoch": 0.29, + "learning_rate": 1.6692178460795148e-05, + "loss": 0.3387, + "step": 4778 + }, + { + "epoch": 0.29, + "learning_rate": 1.669072441421523e-05, + "loss": 0.3683, + "step": 4779 + }, + { + "epoch": 0.29, + "learning_rate": 1.668927011148306e-05, + "loss": 0.3424, + "step": 4780 + }, + { + "epoch": 0.29, + "learning_rate": 1.6687815552654328e-05, + "loss": 0.3308, + "step": 4781 + }, + { + "epoch": 0.29, + "learning_rate": 1.6686360737784712e-05, + "loss": 0.3544, + "step": 4782 + }, + { + "epoch": 0.29, + "learning_rate": 1.6684905666929914e-05, + "loss": 0.3563, + "step": 4783 + }, + { + "epoch": 0.29, + "learning_rate": 1.6683450340145637e-05, + "loss": 0.3502, + "step": 4784 + }, + { + "epoch": 0.29, + "learning_rate": 1.6681994757487606e-05, + "loss": 0.3276, + "step": 4785 + }, + { + "epoch": 0.29, + "learning_rate": 1.6680538919011535e-05, + "loss": 0.3404, + "step": 4786 + }, + { + "epoch": 0.29, + "learning_rate": 1.667908282477317e-05, + "loss": 0.3253, + "step": 4787 + }, + { + "epoch": 0.29, + "learning_rate": 1.6677626474828255e-05, + "loss": 0.3585, + "step": 4788 + }, + { + "epoch": 0.29, + "learning_rate": 1.6676169869232546e-05, + "loss": 0.3524, + "step": 4789 + }, + { + "epoch": 0.29, + "learning_rate": 1.6674713008041806e-05, + "loss": 0.3719, + "step": 4790 + }, + { + "epoch": 0.29, + "learning_rate": 1.6673255891311813e-05, + "loss": 0.3368, + "step": 4791 + }, + { + "epoch": 0.29, + "learning_rate": 1.667179851909835e-05, + "loss": 0.3535, + "step": 4792 + }, + { + "epoch": 0.29, + "learning_rate": 1.6670340891457218e-05, + "loss": 0.3033, + "step": 4793 + }, + { + "epoch": 0.29, + "learning_rate": 1.6668883008444213e-05, + "loss": 0.3301, + "step": 4794 + }, + { + "epoch": 0.29, + "learning_rate": 1.6667424870115155e-05, + "loss": 0.3429, + "step": 4795 + }, + { + "epoch": 0.29, + "learning_rate": 1.666596647652587e-05, + "loss": 0.3217, + "step": 4796 + }, + { + "epoch": 0.29, + "learning_rate": 1.6664507827732185e-05, + "loss": 0.339, + "step": 4797 + }, + { + "epoch": 0.29, + "learning_rate": 1.6663048923789952e-05, + "loss": 0.3426, + "step": 4798 + }, + { + "epoch": 0.29, + "learning_rate": 1.6661589764755016e-05, + "loss": 0.3136, + "step": 4799 + }, + { + "epoch": 0.29, + "learning_rate": 1.6660130350683247e-05, + "loss": 0.3477, + "step": 4800 + }, + { + "epoch": 0.29, + "learning_rate": 1.665867068163052e-05, + "loss": 0.3463, + "step": 4801 + }, + { + "epoch": 0.29, + "learning_rate": 1.6657210757652714e-05, + "loss": 0.3331, + "step": 4802 + }, + { + "epoch": 0.29, + "learning_rate": 1.665575057880572e-05, + "loss": 0.3619, + "step": 4803 + }, + { + "epoch": 0.29, + "learning_rate": 1.6654290145145445e-05, + "loss": 0.3424, + "step": 4804 + }, + { + "epoch": 0.29, + "learning_rate": 1.6652829456727795e-05, + "loss": 0.3638, + "step": 4805 + }, + { + "epoch": 0.29, + "learning_rate": 1.6651368513608704e-05, + "loss": 0.3256, + "step": 4806 + }, + { + "epoch": 0.29, + "learning_rate": 1.664990731584409e-05, + "loss": 0.3376, + "step": 4807 + }, + { + "epoch": 0.29, + "learning_rate": 1.66484458634899e-05, + "loss": 0.3384, + "step": 4808 + }, + { + "epoch": 0.29, + "learning_rate": 1.6646984156602085e-05, + "loss": 0.3373, + "step": 4809 + }, + { + "epoch": 0.29, + "learning_rate": 1.664552219523661e-05, + "loss": 0.3477, + "step": 4810 + }, + { + "epoch": 0.29, + "learning_rate": 1.6644059979449442e-05, + "loss": 0.3306, + "step": 4811 + }, + { + "epoch": 0.29, + "learning_rate": 1.6642597509296558e-05, + "loss": 0.3438, + "step": 4812 + }, + { + "epoch": 0.29, + "learning_rate": 1.6641134784833955e-05, + "loss": 0.3516, + "step": 4813 + }, + { + "epoch": 0.29, + "learning_rate": 1.663967180611763e-05, + "loss": 0.3491, + "step": 4814 + }, + { + "epoch": 0.29, + "learning_rate": 1.663820857320359e-05, + "loss": 0.3502, + "step": 4815 + }, + { + "epoch": 0.29, + "learning_rate": 1.6636745086147858e-05, + "loss": 0.3426, + "step": 4816 + }, + { + "epoch": 0.29, + "learning_rate": 1.663528134500646e-05, + "loss": 0.333, + "step": 4817 + }, + { + "epoch": 0.29, + "learning_rate": 1.6633817349835442e-05, + "loss": 0.3574, + "step": 4818 + }, + { + "epoch": 0.29, + "learning_rate": 1.663235310069084e-05, + "loss": 0.3535, + "step": 4819 + }, + { + "epoch": 0.29, + "learning_rate": 1.663088859762873e-05, + "loss": 0.3722, + "step": 4820 + }, + { + "epoch": 0.29, + "learning_rate": 1.662942384070516e-05, + "loss": 0.346, + "step": 4821 + }, + { + "epoch": 0.29, + "learning_rate": 1.6627958829976224e-05, + "loss": 0.3287, + "step": 4822 + }, + { + "epoch": 0.29, + "learning_rate": 1.6626493565497998e-05, + "loss": 0.3585, + "step": 4823 + }, + { + "epoch": 0.29, + "learning_rate": 1.6625028047326588e-05, + "loss": 0.339, + "step": 4824 + }, + { + "epoch": 0.29, + "learning_rate": 1.6623562275518094e-05, + "loss": 0.344, + "step": 4825 + }, + { + "epoch": 0.29, + "learning_rate": 1.662209625012864e-05, + "loss": 0.3571, + "step": 4826 + }, + { + "epoch": 0.29, + "learning_rate": 1.6620629971214344e-05, + "loss": 0.3345, + "step": 4827 + }, + { + "epoch": 0.29, + "learning_rate": 1.6619163438831348e-05, + "loss": 0.3591, + "step": 4828 + }, + { + "epoch": 0.29, + "learning_rate": 1.6617696653035796e-05, + "loss": 0.3499, + "step": 4829 + }, + { + "epoch": 0.29, + "learning_rate": 1.661622961388384e-05, + "loss": 0.2962, + "step": 4830 + }, + { + "epoch": 0.29, + "learning_rate": 1.661476232143165e-05, + "loss": 0.3535, + "step": 4831 + }, + { + "epoch": 0.29, + "learning_rate": 1.66132947757354e-05, + "loss": 0.3736, + "step": 4832 + }, + { + "epoch": 0.29, + "learning_rate": 1.6611826976851277e-05, + "loss": 0.3675, + "step": 4833 + }, + { + "epoch": 0.29, + "learning_rate": 1.661035892483547e-05, + "loss": 0.3446, + "step": 4834 + }, + { + "epoch": 0.29, + "learning_rate": 1.6608890619744184e-05, + "loss": 0.3248, + "step": 4835 + }, + { + "epoch": 0.29, + "learning_rate": 1.660742206163363e-05, + "loss": 0.3119, + "step": 4836 + }, + { + "epoch": 0.29, + "learning_rate": 1.6605953250560038e-05, + "loss": 0.3491, + "step": 4837 + }, + { + "epoch": 0.29, + "learning_rate": 1.6604484186579637e-05, + "loss": 0.3555, + "step": 4838 + }, + { + "epoch": 0.29, + "learning_rate": 1.660301486974867e-05, + "loss": 0.3299, + "step": 4839 + }, + { + "epoch": 0.29, + "learning_rate": 1.660154530012339e-05, + "loss": 0.3078, + "step": 4840 + }, + { + "epoch": 0.29, + "learning_rate": 1.660007547776006e-05, + "loss": 0.3569, + "step": 4841 + }, + { + "epoch": 0.29, + "learning_rate": 1.6598605402714945e-05, + "loss": 0.3276, + "step": 4842 + }, + { + "epoch": 0.29, + "learning_rate": 1.6597135075044337e-05, + "loss": 0.3518, + "step": 4843 + }, + { + "epoch": 0.29, + "learning_rate": 1.659566449480452e-05, + "loss": 0.3304, + "step": 4844 + }, + { + "epoch": 0.29, + "learning_rate": 1.6594193662051796e-05, + "loss": 0.3393, + "step": 4845 + }, + { + "epoch": 0.29, + "learning_rate": 1.6592722576842476e-05, + "loss": 0.341, + "step": 4846 + }, + { + "epoch": 0.29, + "learning_rate": 1.659125123923288e-05, + "loss": 0.3276, + "step": 4847 + }, + { + "epoch": 0.29, + "learning_rate": 1.6589779649279337e-05, + "loss": 0.3507, + "step": 4848 + }, + { + "epoch": 0.29, + "learning_rate": 1.6588307807038187e-05, + "loss": 0.3343, + "step": 4849 + }, + { + "epoch": 0.29, + "learning_rate": 1.6586835712565777e-05, + "loss": 0.3326, + "step": 4850 + }, + { + "epoch": 0.29, + "learning_rate": 1.6585363365918468e-05, + "loss": 0.3504, + "step": 4851 + }, + { + "epoch": 0.29, + "learning_rate": 1.658389076715263e-05, + "loss": 0.3518, + "step": 4852 + }, + { + "epoch": 0.29, + "learning_rate": 1.6582417916324636e-05, + "loss": 0.3474, + "step": 4853 + }, + { + "epoch": 0.29, + "learning_rate": 1.658094481349088e-05, + "loss": 0.3599, + "step": 4854 + }, + { + "epoch": 0.29, + "learning_rate": 1.6579471458707752e-05, + "loss": 0.3157, + "step": 4855 + }, + { + "epoch": 0.29, + "learning_rate": 1.6577997852031663e-05, + "loss": 0.3756, + "step": 4856 + }, + { + "epoch": 0.29, + "learning_rate": 1.6576523993519028e-05, + "loss": 0.3485, + "step": 4857 + }, + { + "epoch": 0.29, + "learning_rate": 1.6575049883226276e-05, + "loss": 0.3249, + "step": 4858 + }, + { + "epoch": 0.29, + "learning_rate": 1.6573575521209844e-05, + "loss": 0.3571, + "step": 4859 + }, + { + "epoch": 0.29, + "learning_rate": 1.657210090752617e-05, + "loss": 0.3404, + "step": 4860 + }, + { + "epoch": 0.29, + "learning_rate": 1.657062604223172e-05, + "loss": 0.3206, + "step": 4861 + }, + { + "epoch": 0.29, + "learning_rate": 1.6569150925382946e-05, + "loss": 0.3234, + "step": 4862 + }, + { + "epoch": 0.29, + "learning_rate": 1.6567675557036333e-05, + "loss": 0.3825, + "step": 4863 + }, + { + "epoch": 0.29, + "learning_rate": 1.6566199937248358e-05, + "loss": 0.3357, + "step": 4864 + }, + { + "epoch": 0.29, + "learning_rate": 1.656472406607552e-05, + "loss": 0.3446, + "step": 4865 + }, + { + "epoch": 0.29, + "learning_rate": 1.6563247943574316e-05, + "loss": 0.3298, + "step": 4866 + }, + { + "epoch": 0.29, + "learning_rate": 1.6561771569801264e-05, + "loss": 0.3624, + "step": 4867 + }, + { + "epoch": 0.29, + "learning_rate": 1.6560294944812887e-05, + "loss": 0.3122, + "step": 4868 + }, + { + "epoch": 0.29, + "learning_rate": 1.6558818068665712e-05, + "loss": 0.334, + "step": 4869 + }, + { + "epoch": 0.29, + "learning_rate": 1.6557340941416288e-05, + "loss": 0.3664, + "step": 4870 + }, + { + "epoch": 0.29, + "learning_rate": 1.6555863563121158e-05, + "loss": 0.3384, + "step": 4871 + }, + { + "epoch": 0.29, + "learning_rate": 1.655438593383689e-05, + "loss": 0.3164, + "step": 4872 + }, + { + "epoch": 0.29, + "learning_rate": 1.6552908053620053e-05, + "loss": 0.3101, + "step": 4873 + }, + { + "epoch": 0.29, + "learning_rate": 1.6551429922527223e-05, + "loss": 0.3256, + "step": 4874 + }, + { + "epoch": 0.29, + "learning_rate": 1.6549951540614996e-05, + "loss": 0.3298, + "step": 4875 + }, + { + "epoch": 0.29, + "learning_rate": 1.6548472907939965e-05, + "loss": 0.3744, + "step": 4876 + }, + { + "epoch": 0.29, + "learning_rate": 1.6546994024558744e-05, + "loss": 0.3482, + "step": 4877 + }, + { + "epoch": 0.29, + "learning_rate": 1.6545514890527948e-05, + "loss": 0.358, + "step": 4878 + }, + { + "epoch": 0.29, + "learning_rate": 1.6544035505904213e-05, + "loss": 0.3228, + "step": 4879 + }, + { + "epoch": 0.29, + "learning_rate": 1.6542555870744164e-05, + "loss": 0.3287, + "step": 4880 + }, + { + "epoch": 0.29, + "learning_rate": 1.6541075985104456e-05, + "loss": 0.3165, + "step": 4881 + }, + { + "epoch": 0.29, + "learning_rate": 1.6539595849041747e-05, + "loss": 0.3454, + "step": 4882 + }, + { + "epoch": 0.29, + "learning_rate": 1.6538115462612697e-05, + "loss": 0.3557, + "step": 4883 + }, + { + "epoch": 0.3, + "learning_rate": 1.6536634825873993e-05, + "loss": 0.3284, + "step": 4884 + }, + { + "epoch": 0.3, + "learning_rate": 1.6535153938882312e-05, + "loss": 0.3133, + "step": 4885 + }, + { + "epoch": 0.3, + "learning_rate": 1.653367280169435e-05, + "loss": 0.3451, + "step": 4886 + }, + { + "epoch": 0.3, + "learning_rate": 1.6532191414366813e-05, + "loss": 0.3591, + "step": 4887 + }, + { + "epoch": 0.3, + "learning_rate": 1.653070977695642e-05, + "loss": 0.3571, + "step": 4888 + }, + { + "epoch": 0.3, + "learning_rate": 1.6529227889519884e-05, + "loss": 0.3597, + "step": 4889 + }, + { + "epoch": 0.3, + "learning_rate": 1.652774575211395e-05, + "loss": 0.317, + "step": 4890 + }, + { + "epoch": 0.3, + "learning_rate": 1.652626336479536e-05, + "loss": 0.3479, + "step": 4891 + }, + { + "epoch": 0.3, + "learning_rate": 1.6524780727620858e-05, + "loss": 0.3292, + "step": 4892 + }, + { + "epoch": 0.3, + "learning_rate": 1.6523297840647213e-05, + "loss": 0.3538, + "step": 4893 + }, + { + "epoch": 0.3, + "learning_rate": 1.65218147039312e-05, + "loss": 0.3396, + "step": 4894 + }, + { + "epoch": 0.3, + "learning_rate": 1.652033131752959e-05, + "loss": 0.3482, + "step": 4895 + }, + { + "epoch": 0.3, + "learning_rate": 1.6518847681499183e-05, + "loss": 0.3605, + "step": 4896 + }, + { + "epoch": 0.3, + "learning_rate": 1.6517363795896776e-05, + "loss": 0.363, + "step": 4897 + }, + { + "epoch": 0.3, + "learning_rate": 1.6515879660779175e-05, + "loss": 0.351, + "step": 4898 + }, + { + "epoch": 0.3, + "learning_rate": 1.651439527620321e-05, + "loss": 0.3753, + "step": 4899 + }, + { + "epoch": 0.3, + "learning_rate": 1.65129106422257e-05, + "loss": 0.3404, + "step": 4900 + }, + { + "epoch": 0.3, + "learning_rate": 1.6511425758903493e-05, + "loss": 0.3421, + "step": 4901 + }, + { + "epoch": 0.3, + "learning_rate": 1.650994062629343e-05, + "loss": 0.3454, + "step": 4902 + }, + { + "epoch": 0.3, + "learning_rate": 1.650845524445237e-05, + "loss": 0.3532, + "step": 4903 + }, + { + "epoch": 0.3, + "learning_rate": 1.6506969613437186e-05, + "loss": 0.3504, + "step": 4904 + }, + { + "epoch": 0.3, + "learning_rate": 1.6505483733304744e-05, + "loss": 0.3633, + "step": 4905 + }, + { + "epoch": 0.3, + "learning_rate": 1.6503997604111943e-05, + "loss": 0.3365, + "step": 4906 + }, + { + "epoch": 0.3, + "learning_rate": 1.6502511225915668e-05, + "loss": 0.363, + "step": 4907 + }, + { + "epoch": 0.3, + "learning_rate": 1.650102459877283e-05, + "loss": 0.3912, + "step": 4908 + }, + { + "epoch": 0.3, + "learning_rate": 1.6499537722740347e-05, + "loss": 0.3622, + "step": 4909 + }, + { + "epoch": 0.3, + "learning_rate": 1.649805059787514e-05, + "loss": 0.3329, + "step": 4910 + }, + { + "epoch": 0.3, + "learning_rate": 1.649656322423414e-05, + "loss": 0.3121, + "step": 4911 + }, + { + "epoch": 0.3, + "learning_rate": 1.6495075601874296e-05, + "loss": 0.3284, + "step": 4912 + }, + { + "epoch": 0.3, + "learning_rate": 1.6493587730852558e-05, + "loss": 0.3496, + "step": 4913 + }, + { + "epoch": 0.3, + "learning_rate": 1.649209961122589e-05, + "loss": 0.3917, + "step": 4914 + }, + { + "epoch": 0.3, + "learning_rate": 1.6490611243051264e-05, + "loss": 0.3368, + "step": 4915 + }, + { + "epoch": 0.3, + "learning_rate": 1.648912262638566e-05, + "loss": 0.3613, + "step": 4916 + }, + { + "epoch": 0.3, + "learning_rate": 1.6487633761286072e-05, + "loss": 0.3538, + "step": 4917 + }, + { + "epoch": 0.3, + "learning_rate": 1.64861446478095e-05, + "loss": 0.3125, + "step": 4918 + }, + { + "epoch": 0.3, + "learning_rate": 1.648465528601295e-05, + "loss": 0.3544, + "step": 4919 + }, + { + "epoch": 0.3, + "learning_rate": 1.648316567595345e-05, + "loss": 0.3471, + "step": 4920 + }, + { + "epoch": 0.3, + "learning_rate": 1.6481675817688023e-05, + "loss": 0.3178, + "step": 4921 + }, + { + "epoch": 0.3, + "learning_rate": 1.648018571127371e-05, + "loss": 0.3446, + "step": 4922 + }, + { + "epoch": 0.3, + "learning_rate": 1.6478695356767555e-05, + "loss": 0.3309, + "step": 4923 + }, + { + "epoch": 0.3, + "learning_rate": 1.6477204754226623e-05, + "loss": 0.3647, + "step": 4924 + }, + { + "epoch": 0.3, + "learning_rate": 1.6475713903707977e-05, + "loss": 0.3245, + "step": 4925 + }, + { + "epoch": 0.3, + "learning_rate": 1.6474222805268696e-05, + "loss": 0.3677, + "step": 4926 + }, + { + "epoch": 0.3, + "learning_rate": 1.6472731458965863e-05, + "loss": 0.3239, + "step": 4927 + }, + { + "epoch": 0.3, + "learning_rate": 1.6471239864856573e-05, + "loss": 0.363, + "step": 4928 + }, + { + "epoch": 0.3, + "learning_rate": 1.6469748022997934e-05, + "loss": 0.3591, + "step": 4929 + }, + { + "epoch": 0.3, + "learning_rate": 1.6468255933447065e-05, + "loss": 0.339, + "step": 4930 + }, + { + "epoch": 0.3, + "learning_rate": 1.6466763596261083e-05, + "loss": 0.3675, + "step": 4931 + }, + { + "epoch": 0.3, + "learning_rate": 1.646527101149712e-05, + "loss": 0.3541, + "step": 4932 + }, + { + "epoch": 0.3, + "learning_rate": 1.6463778179212327e-05, + "loss": 0.3465, + "step": 4933 + }, + { + "epoch": 0.3, + "learning_rate": 1.6462285099463852e-05, + "loss": 0.3276, + "step": 4934 + }, + { + "epoch": 0.3, + "learning_rate": 1.6460791772308858e-05, + "loss": 0.3658, + "step": 4935 + }, + { + "epoch": 0.3, + "learning_rate": 1.645929819780452e-05, + "loss": 0.3379, + "step": 4936 + }, + { + "epoch": 0.3, + "learning_rate": 1.645780437600801e-05, + "loss": 0.3617, + "step": 4937 + }, + { + "epoch": 0.3, + "learning_rate": 1.6456310306976525e-05, + "loss": 0.344, + "step": 4938 + }, + { + "epoch": 0.3, + "learning_rate": 1.6454815990767264e-05, + "loss": 0.3491, + "step": 4939 + }, + { + "epoch": 0.3, + "learning_rate": 1.6453321427437438e-05, + "loss": 0.3365, + "step": 4940 + }, + { + "epoch": 0.3, + "learning_rate": 1.6451826617044263e-05, + "loss": 0.3463, + "step": 4941 + }, + { + "epoch": 0.3, + "learning_rate": 1.6450331559644967e-05, + "loss": 0.3599, + "step": 4942 + }, + { + "epoch": 0.3, + "learning_rate": 1.644883625529679e-05, + "loss": 0.3605, + "step": 4943 + }, + { + "epoch": 0.3, + "learning_rate": 1.644734070405698e-05, + "loss": 0.3499, + "step": 4944 + }, + { + "epoch": 0.3, + "learning_rate": 1.644584490598279e-05, + "loss": 0.361, + "step": 4945 + }, + { + "epoch": 0.3, + "learning_rate": 1.644434886113149e-05, + "loss": 0.3451, + "step": 4946 + }, + { + "epoch": 0.3, + "learning_rate": 1.6442852569560356e-05, + "loss": 0.3577, + "step": 4947 + }, + { + "epoch": 0.3, + "learning_rate": 1.644135603132667e-05, + "loss": 0.3524, + "step": 4948 + }, + { + "epoch": 0.3, + "learning_rate": 1.6439859246487725e-05, + "loss": 0.3502, + "step": 4949 + }, + { + "epoch": 0.3, + "learning_rate": 1.6438362215100826e-05, + "loss": 0.3429, + "step": 4950 + }, + { + "epoch": 0.3, + "learning_rate": 1.643686493722329e-05, + "loss": 0.3122, + "step": 4951 + }, + { + "epoch": 0.3, + "learning_rate": 1.6435367412912443e-05, + "loss": 0.3361, + "step": 4952 + }, + { + "epoch": 0.3, + "learning_rate": 1.6433869642225606e-05, + "loss": 0.3636, + "step": 4953 + }, + { + "epoch": 0.3, + "learning_rate": 1.6432371625220124e-05, + "loss": 0.3563, + "step": 4954 + }, + { + "epoch": 0.3, + "learning_rate": 1.643087336195336e-05, + "loss": 0.3468, + "step": 4955 + }, + { + "epoch": 0.3, + "learning_rate": 1.6429374852482657e-05, + "loss": 0.3446, + "step": 4956 + }, + { + "epoch": 0.3, + "learning_rate": 1.6427876096865394e-05, + "loss": 0.3535, + "step": 4957 + }, + { + "epoch": 0.3, + "learning_rate": 1.642637709515895e-05, + "loss": 0.3298, + "step": 4958 + }, + { + "epoch": 0.3, + "learning_rate": 1.6424877847420715e-05, + "loss": 0.3488, + "step": 4959 + }, + { + "epoch": 0.3, + "learning_rate": 1.6423378353708086e-05, + "loss": 0.3429, + "step": 4960 + }, + { + "epoch": 0.3, + "learning_rate": 1.642187861407847e-05, + "loss": 0.3518, + "step": 4961 + }, + { + "epoch": 0.3, + "learning_rate": 1.6420378628589282e-05, + "loss": 0.3577, + "step": 4962 + }, + { + "epoch": 0.3, + "learning_rate": 1.6418878397297954e-05, + "loss": 0.3251, + "step": 4963 + }, + { + "epoch": 0.3, + "learning_rate": 1.6417377920261914e-05, + "loss": 0.3337, + "step": 4964 + }, + { + "epoch": 0.3, + "learning_rate": 1.6415877197538614e-05, + "loss": 0.373, + "step": 4965 + }, + { + "epoch": 0.3, + "learning_rate": 1.641437622918551e-05, + "loss": 0.3616, + "step": 4966 + }, + { + "epoch": 0.3, + "learning_rate": 1.641287501526006e-05, + "loss": 0.3253, + "step": 4967 + }, + { + "epoch": 0.3, + "learning_rate": 1.641137355581974e-05, + "loss": 0.334, + "step": 4968 + }, + { + "epoch": 0.3, + "learning_rate": 1.640987185092203e-05, + "loss": 0.3108, + "step": 4969 + }, + { + "epoch": 0.3, + "learning_rate": 1.6408369900624426e-05, + "loss": 0.3438, + "step": 4970 + }, + { + "epoch": 0.3, + "learning_rate": 1.640686770498443e-05, + "loss": 0.3424, + "step": 4971 + }, + { + "epoch": 0.3, + "learning_rate": 1.6405365264059555e-05, + "loss": 0.3703, + "step": 4972 + }, + { + "epoch": 0.3, + "learning_rate": 1.6403862577907317e-05, + "loss": 0.353, + "step": 4973 + }, + { + "epoch": 0.3, + "learning_rate": 1.6402359646585245e-05, + "loss": 0.3485, + "step": 4974 + }, + { + "epoch": 0.3, + "learning_rate": 1.6400856470150883e-05, + "loss": 0.3184, + "step": 4975 + }, + { + "epoch": 0.3, + "learning_rate": 1.6399353048661778e-05, + "loss": 0.3549, + "step": 4976 + }, + { + "epoch": 0.3, + "learning_rate": 1.6397849382175487e-05, + "loss": 0.3555, + "step": 4977 + }, + { + "epoch": 0.3, + "learning_rate": 1.6396345470749576e-05, + "loss": 0.339, + "step": 4978 + }, + { + "epoch": 0.3, + "learning_rate": 1.6394841314441625e-05, + "loss": 0.3368, + "step": 4979 + }, + { + "epoch": 0.3, + "learning_rate": 1.6393336913309216e-05, + "loss": 0.3371, + "step": 4980 + }, + { + "epoch": 0.3, + "learning_rate": 1.639183226740995e-05, + "loss": 0.3552, + "step": 4981 + }, + { + "epoch": 0.3, + "learning_rate": 1.6390327376801426e-05, + "loss": 0.3493, + "step": 4982 + }, + { + "epoch": 0.3, + "learning_rate": 1.6388822241541265e-05, + "loss": 0.317, + "step": 4983 + }, + { + "epoch": 0.3, + "learning_rate": 1.6387316861687087e-05, + "loss": 0.3348, + "step": 4984 + }, + { + "epoch": 0.3, + "learning_rate": 1.638581123729652e-05, + "loss": 0.3365, + "step": 4985 + }, + { + "epoch": 0.3, + "learning_rate": 1.6384305368427215e-05, + "loss": 0.3609, + "step": 4986 + }, + { + "epoch": 0.3, + "learning_rate": 1.638279925513682e-05, + "loss": 0.3619, + "step": 4987 + }, + { + "epoch": 0.3, + "learning_rate": 1.6381292897482995e-05, + "loss": 0.3669, + "step": 4988 + }, + { + "epoch": 0.3, + "learning_rate": 1.637978629552341e-05, + "loss": 0.2979, + "step": 4989 + }, + { + "epoch": 0.3, + "learning_rate": 1.6378279449315747e-05, + "loss": 0.3719, + "step": 4990 + }, + { + "epoch": 0.3, + "learning_rate": 1.6376772358917694e-05, + "loss": 0.3128, + "step": 4991 + }, + { + "epoch": 0.3, + "learning_rate": 1.637526502438695e-05, + "loss": 0.344, + "step": 4992 + }, + { + "epoch": 0.3, + "learning_rate": 1.6373757445781224e-05, + "loss": 0.3521, + "step": 4993 + }, + { + "epoch": 0.3, + "learning_rate": 1.637224962315823e-05, + "loss": 0.3033, + "step": 4994 + }, + { + "epoch": 0.3, + "learning_rate": 1.6370741556575698e-05, + "loss": 0.3253, + "step": 4995 + }, + { + "epoch": 0.3, + "learning_rate": 1.636923324609136e-05, + "loss": 0.3546, + "step": 4996 + }, + { + "epoch": 0.3, + "learning_rate": 1.6367724691762967e-05, + "loss": 0.3549, + "step": 4997 + }, + { + "epoch": 0.3, + "learning_rate": 1.6366215893648267e-05, + "loss": 0.3379, + "step": 4998 + }, + { + "epoch": 0.3, + "learning_rate": 1.636470685180503e-05, + "loss": 0.3488, + "step": 4999 + }, + { + "epoch": 0.3, + "learning_rate": 1.636319756629102e-05, + "loss": 0.3502, + "step": 5000 + }, + { + "epoch": 0.3, + "learning_rate": 1.6361688037164028e-05, + "loss": 0.361, + "step": 5001 + }, + { + "epoch": 0.3, + "learning_rate": 1.6360178264481846e-05, + "loss": 0.3315, + "step": 5002 + }, + { + "epoch": 0.3, + "learning_rate": 1.6358668248302266e-05, + "loss": 0.3037, + "step": 5003 + }, + { + "epoch": 0.3, + "learning_rate": 1.635715798868311e-05, + "loss": 0.327, + "step": 5004 + }, + { + "epoch": 0.3, + "learning_rate": 1.6355647485682192e-05, + "loss": 0.3418, + "step": 5005 + }, + { + "epoch": 0.3, + "learning_rate": 1.635413673935734e-05, + "loss": 0.3209, + "step": 5006 + }, + { + "epoch": 0.3, + "learning_rate": 1.6352625749766394e-05, + "loss": 0.3186, + "step": 5007 + }, + { + "epoch": 0.3, + "learning_rate": 1.6351114516967205e-05, + "loss": 0.3323, + "step": 5008 + }, + { + "epoch": 0.3, + "learning_rate": 1.6349603041017626e-05, + "loss": 0.341, + "step": 5009 + }, + { + "epoch": 0.3, + "learning_rate": 1.6348091321975525e-05, + "loss": 0.328, + "step": 5010 + }, + { + "epoch": 0.3, + "learning_rate": 1.6346579359898776e-05, + "loss": 0.3546, + "step": 5011 + }, + { + "epoch": 0.3, + "learning_rate": 1.6345067154845263e-05, + "loss": 0.3619, + "step": 5012 + }, + { + "epoch": 0.3, + "learning_rate": 1.6343554706872886e-05, + "loss": 0.3569, + "step": 5013 + }, + { + "epoch": 0.3, + "learning_rate": 1.634204201603954e-05, + "loss": 0.3132, + "step": 5014 + }, + { + "epoch": 0.3, + "learning_rate": 1.6340529082403146e-05, + "loss": 0.358, + "step": 5015 + }, + { + "epoch": 0.3, + "learning_rate": 1.6339015906021628e-05, + "loss": 0.3387, + "step": 5016 + }, + { + "epoch": 0.3, + "learning_rate": 1.6337502486952905e-05, + "loss": 0.3585, + "step": 5017 + }, + { + "epoch": 0.3, + "learning_rate": 1.633598882525493e-05, + "loss": 0.3379, + "step": 5018 + }, + { + "epoch": 0.3, + "learning_rate": 1.6334474920985645e-05, + "loss": 0.2941, + "step": 5019 + }, + { + "epoch": 0.3, + "learning_rate": 1.6332960774203013e-05, + "loss": 0.3524, + "step": 5020 + }, + { + "epoch": 0.3, + "learning_rate": 1.6331446384965002e-05, + "loss": 0.3142, + "step": 5021 + }, + { + "epoch": 0.3, + "learning_rate": 1.6329931753329595e-05, + "loss": 0.3597, + "step": 5022 + }, + { + "epoch": 0.3, + "learning_rate": 1.632841687935477e-05, + "loss": 0.3365, + "step": 5023 + }, + { + "epoch": 0.3, + "learning_rate": 1.632690176309853e-05, + "loss": 0.3717, + "step": 5024 + }, + { + "epoch": 0.3, + "learning_rate": 1.6325386404618873e-05, + "loss": 0.3418, + "step": 5025 + }, + { + "epoch": 0.3, + "learning_rate": 1.6323870803973827e-05, + "loss": 0.3424, + "step": 5026 + }, + { + "epoch": 0.3, + "learning_rate": 1.6322354961221404e-05, + "loss": 0.3591, + "step": 5027 + }, + { + "epoch": 0.3, + "learning_rate": 1.6320838876419646e-05, + "loss": 0.3597, + "step": 5028 + }, + { + "epoch": 0.3, + "learning_rate": 1.631932254962659e-05, + "loss": 0.3382, + "step": 5029 + }, + { + "epoch": 0.3, + "learning_rate": 1.6317805980900293e-05, + "loss": 0.358, + "step": 5030 + }, + { + "epoch": 0.3, + "learning_rate": 1.6316289170298808e-05, + "loss": 0.3758, + "step": 5031 + }, + { + "epoch": 0.3, + "learning_rate": 1.6314772117880215e-05, + "loss": 0.3463, + "step": 5032 + }, + { + "epoch": 0.3, + "learning_rate": 1.631325482370259e-05, + "loss": 0.3553, + "step": 5033 + }, + { + "epoch": 0.3, + "learning_rate": 1.6311737287824024e-05, + "loss": 0.3223, + "step": 5034 + }, + { + "epoch": 0.3, + "learning_rate": 1.631021951030261e-05, + "loss": 0.3105, + "step": 5035 + }, + { + "epoch": 0.3, + "learning_rate": 1.630870149119646e-05, + "loss": 0.365, + "step": 5036 + }, + { + "epoch": 0.3, + "learning_rate": 1.6307183230563692e-05, + "loss": 0.341, + "step": 5037 + }, + { + "epoch": 0.3, + "learning_rate": 1.6305664728462433e-05, + "loss": 0.3675, + "step": 5038 + }, + { + "epoch": 0.3, + "learning_rate": 1.630414598495081e-05, + "loss": 0.3457, + "step": 5039 + }, + { + "epoch": 0.3, + "learning_rate": 1.6302627000086976e-05, + "loss": 0.3555, + "step": 5040 + }, + { + "epoch": 0.3, + "learning_rate": 1.6301107773929082e-05, + "loss": 0.3471, + "step": 5041 + }, + { + "epoch": 0.3, + "learning_rate": 1.629958830653529e-05, + "loss": 0.3488, + "step": 5042 + }, + { + "epoch": 0.3, + "learning_rate": 1.6298068597963778e-05, + "loss": 0.3119, + "step": 5043 + }, + { + "epoch": 0.3, + "learning_rate": 1.6296548648272716e-05, + "loss": 0.3319, + "step": 5044 + }, + { + "epoch": 0.3, + "learning_rate": 1.629502845752031e-05, + "loss": 0.3348, + "step": 5045 + }, + { + "epoch": 0.3, + "learning_rate": 1.6293508025764744e-05, + "loss": 0.3451, + "step": 5046 + }, + { + "epoch": 0.3, + "learning_rate": 1.629198735306424e-05, + "loss": 0.3044, + "step": 5047 + }, + { + "epoch": 0.3, + "learning_rate": 1.629046643947701e-05, + "loss": 0.3563, + "step": 5048 + }, + { + "epoch": 0.31, + "learning_rate": 1.6288945285061284e-05, + "loss": 0.344, + "step": 5049 + }, + { + "epoch": 0.31, + "learning_rate": 1.6287423889875297e-05, + "loss": 0.3221, + "step": 5050 + }, + { + "epoch": 0.31, + "learning_rate": 1.6285902253977298e-05, + "loss": 0.3426, + "step": 5051 + }, + { + "epoch": 0.31, + "learning_rate": 1.6284380377425536e-05, + "loss": 0.3354, + "step": 5052 + }, + { + "epoch": 0.31, + "learning_rate": 1.6282858260278288e-05, + "loss": 0.3178, + "step": 5053 + }, + { + "epoch": 0.31, + "learning_rate": 1.6281335902593814e-05, + "loss": 0.3371, + "step": 5054 + }, + { + "epoch": 0.31, + "learning_rate": 1.6279813304430406e-05, + "loss": 0.3415, + "step": 5055 + }, + { + "epoch": 0.31, + "learning_rate": 1.627829046584635e-05, + "loss": 0.3421, + "step": 5056 + }, + { + "epoch": 0.31, + "learning_rate": 1.6276767386899957e-05, + "loss": 0.3371, + "step": 5057 + }, + { + "epoch": 0.31, + "learning_rate": 1.6275244067649526e-05, + "loss": 0.3214, + "step": 5058 + }, + { + "epoch": 0.31, + "learning_rate": 1.6273720508153382e-05, + "loss": 0.3465, + "step": 5059 + }, + { + "epoch": 0.31, + "learning_rate": 1.6272196708469855e-05, + "loss": 0.3273, + "step": 5060 + }, + { + "epoch": 0.31, + "learning_rate": 1.6270672668657282e-05, + "loss": 0.3316, + "step": 5061 + }, + { + "epoch": 0.31, + "learning_rate": 1.626914838877401e-05, + "loss": 0.3016, + "step": 5062 + }, + { + "epoch": 0.31, + "learning_rate": 1.6267623868878394e-05, + "loss": 0.3256, + "step": 5063 + }, + { + "epoch": 0.31, + "learning_rate": 1.6266099109028808e-05, + "loss": 0.3535, + "step": 5064 + }, + { + "epoch": 0.31, + "learning_rate": 1.6264574109283617e-05, + "loss": 0.3398, + "step": 5065 + }, + { + "epoch": 0.31, + "learning_rate": 1.6263048869701207e-05, + "loss": 0.327, + "step": 5066 + }, + { + "epoch": 0.31, + "learning_rate": 1.6261523390339978e-05, + "loss": 0.3273, + "step": 5067 + }, + { + "epoch": 0.31, + "learning_rate": 1.6259997671258325e-05, + "loss": 0.3432, + "step": 5068 + }, + { + "epoch": 0.31, + "learning_rate": 1.625847171251466e-05, + "loss": 0.3326, + "step": 5069 + }, + { + "epoch": 0.31, + "learning_rate": 1.625694551416741e-05, + "loss": 0.3538, + "step": 5070 + }, + { + "epoch": 0.31, + "learning_rate": 1.6255419076274997e-05, + "loss": 0.351, + "step": 5071 + }, + { + "epoch": 0.31, + "learning_rate": 1.625389239889587e-05, + "loss": 0.3541, + "step": 5072 + }, + { + "epoch": 0.31, + "learning_rate": 1.6252365482088468e-05, + "loss": 0.3594, + "step": 5073 + }, + { + "epoch": 0.31, + "learning_rate": 1.6250838325911252e-05, + "loss": 0.3655, + "step": 5074 + }, + { + "epoch": 0.31, + "learning_rate": 1.624931093042269e-05, + "loss": 0.3499, + "step": 5075 + }, + { + "epoch": 0.31, + "learning_rate": 1.624778329568126e-05, + "loss": 0.3337, + "step": 5076 + }, + { + "epoch": 0.31, + "learning_rate": 1.6246255421745437e-05, + "loss": 0.3177, + "step": 5077 + }, + { + "epoch": 0.31, + "learning_rate": 1.624472730867373e-05, + "loss": 0.346, + "step": 5078 + }, + { + "epoch": 0.31, + "learning_rate": 1.6243198956524626e-05, + "loss": 0.32, + "step": 5079 + }, + { + "epoch": 0.31, + "learning_rate": 1.624167036535665e-05, + "loss": 0.3273, + "step": 5080 + }, + { + "epoch": 0.31, + "learning_rate": 1.624014153522832e-05, + "loss": 0.3295, + "step": 5081 + }, + { + "epoch": 0.31, + "learning_rate": 1.6238612466198166e-05, + "loss": 0.3357, + "step": 5082 + }, + { + "epoch": 0.31, + "learning_rate": 1.623708315832473e-05, + "loss": 0.2946, + "step": 5083 + }, + { + "epoch": 0.31, + "learning_rate": 1.6235553611666556e-05, + "loss": 0.3329, + "step": 5084 + }, + { + "epoch": 0.31, + "learning_rate": 1.6234023826282202e-05, + "loss": 0.3365, + "step": 5085 + }, + { + "epoch": 0.31, + "learning_rate": 1.6232493802230246e-05, + "loss": 0.344, + "step": 5086 + }, + { + "epoch": 0.31, + "learning_rate": 1.623096353956925e-05, + "loss": 0.356, + "step": 5087 + }, + { + "epoch": 0.31, + "learning_rate": 1.6229433038357814e-05, + "loss": 0.3432, + "step": 5088 + }, + { + "epoch": 0.31, + "learning_rate": 1.622790229865452e-05, + "loss": 0.3351, + "step": 5089 + }, + { + "epoch": 0.31, + "learning_rate": 1.622637132051798e-05, + "loss": 0.3161, + "step": 5090 + }, + { + "epoch": 0.31, + "learning_rate": 1.6224840104006803e-05, + "loss": 0.3365, + "step": 5091 + }, + { + "epoch": 0.31, + "learning_rate": 1.6223308649179616e-05, + "loss": 0.3398, + "step": 5092 + }, + { + "epoch": 0.31, + "learning_rate": 1.6221776956095045e-05, + "loss": 0.3354, + "step": 5093 + }, + { + "epoch": 0.31, + "learning_rate": 1.622024502481173e-05, + "loss": 0.3541, + "step": 5094 + }, + { + "epoch": 0.31, + "learning_rate": 1.6218712855388325e-05, + "loss": 0.3711, + "step": 5095 + }, + { + "epoch": 0.31, + "learning_rate": 1.6217180447883484e-05, + "loss": 0.353, + "step": 5096 + }, + { + "epoch": 0.31, + "learning_rate": 1.621564780235588e-05, + "loss": 0.3404, + "step": 5097 + }, + { + "epoch": 0.31, + "learning_rate": 1.6214114918864187e-05, + "loss": 0.3323, + "step": 5098 + }, + { + "epoch": 0.31, + "learning_rate": 1.621258179746709e-05, + "loss": 0.3351, + "step": 5099 + }, + { + "epoch": 0.31, + "learning_rate": 1.6211048438223286e-05, + "loss": 0.346, + "step": 5100 + }, + { + "epoch": 0.31, + "learning_rate": 1.6209514841191472e-05, + "loss": 0.3203, + "step": 5101 + }, + { + "epoch": 0.31, + "learning_rate": 1.6207981006430372e-05, + "loss": 0.3521, + "step": 5102 + }, + { + "epoch": 0.31, + "learning_rate": 1.6206446933998708e-05, + "loss": 0.341, + "step": 5103 + }, + { + "epoch": 0.31, + "learning_rate": 1.6204912623955204e-05, + "loss": 0.332, + "step": 5104 + }, + { + "epoch": 0.31, + "learning_rate": 1.6203378076358602e-05, + "loss": 0.3569, + "step": 5105 + }, + { + "epoch": 0.31, + "learning_rate": 1.6201843291267653e-05, + "loss": 0.3516, + "step": 5106 + }, + { + "epoch": 0.31, + "learning_rate": 1.6200308268741118e-05, + "loss": 0.3214, + "step": 5107 + }, + { + "epoch": 0.31, + "learning_rate": 1.619877300883776e-05, + "loss": 0.3386, + "step": 5108 + }, + { + "epoch": 0.31, + "learning_rate": 1.6197237511616363e-05, + "loss": 0.3253, + "step": 5109 + }, + { + "epoch": 0.31, + "learning_rate": 1.6195701777135705e-05, + "loss": 0.3111, + "step": 5110 + }, + { + "epoch": 0.31, + "learning_rate": 1.619416580545459e-05, + "loss": 0.3655, + "step": 5111 + }, + { + "epoch": 0.31, + "learning_rate": 1.6192629596631815e-05, + "loss": 0.3156, + "step": 5112 + }, + { + "epoch": 0.31, + "learning_rate": 1.6191093150726195e-05, + "loss": 0.3379, + "step": 5113 + }, + { + "epoch": 0.31, + "learning_rate": 1.6189556467796555e-05, + "loss": 0.3295, + "step": 5114 + }, + { + "epoch": 0.31, + "learning_rate": 1.6188019547901724e-05, + "loss": 0.339, + "step": 5115 + }, + { + "epoch": 0.31, + "learning_rate": 1.618648239110054e-05, + "loss": 0.3214, + "step": 5116 + }, + { + "epoch": 0.31, + "learning_rate": 1.6184944997451855e-05, + "loss": 0.3295, + "step": 5117 + }, + { + "epoch": 0.31, + "learning_rate": 1.6183407367014532e-05, + "loss": 0.3032, + "step": 5118 + }, + { + "epoch": 0.31, + "learning_rate": 1.618186949984743e-05, + "loss": 0.3488, + "step": 5119 + }, + { + "epoch": 0.31, + "learning_rate": 1.6180331396009434e-05, + "loss": 0.3221, + "step": 5120 + }, + { + "epoch": 0.31, + "learning_rate": 1.6178793055559425e-05, + "loss": 0.3265, + "step": 5121 + }, + { + "epoch": 0.31, + "learning_rate": 1.6177254478556298e-05, + "loss": 0.3446, + "step": 5122 + }, + { + "epoch": 0.31, + "learning_rate": 1.617571566505896e-05, + "loss": 0.3446, + "step": 5123 + }, + { + "epoch": 0.31, + "learning_rate": 1.617417661512632e-05, + "loss": 0.334, + "step": 5124 + }, + { + "epoch": 0.31, + "learning_rate": 1.6172637328817305e-05, + "loss": 0.3343, + "step": 5125 + }, + { + "epoch": 0.31, + "learning_rate": 1.617109780619084e-05, + "loss": 0.3097, + "step": 5126 + }, + { + "epoch": 0.31, + "learning_rate": 1.6169558047305872e-05, + "loss": 0.3379, + "step": 5127 + }, + { + "epoch": 0.31, + "learning_rate": 1.616801805222134e-05, + "loss": 0.3502, + "step": 5128 + }, + { + "epoch": 0.31, + "learning_rate": 1.6166477820996217e-05, + "loss": 0.32, + "step": 5129 + }, + { + "epoch": 0.31, + "learning_rate": 1.6164937353689455e-05, + "loss": 0.3334, + "step": 5130 + }, + { + "epoch": 0.31, + "learning_rate": 1.6163396650360043e-05, + "loss": 0.3507, + "step": 5131 + }, + { + "epoch": 0.31, + "learning_rate": 1.6161855711066956e-05, + "loss": 0.3647, + "step": 5132 + }, + { + "epoch": 0.31, + "learning_rate": 1.6160314535869194e-05, + "loss": 0.3023, + "step": 5133 + }, + { + "epoch": 0.31, + "learning_rate": 1.6158773124825763e-05, + "loss": 0.332, + "step": 5134 + }, + { + "epoch": 0.31, + "learning_rate": 1.6157231477995668e-05, + "loss": 0.3504, + "step": 5135 + }, + { + "epoch": 0.31, + "learning_rate": 1.6155689595437933e-05, + "loss": 0.3463, + "step": 5136 + }, + { + "epoch": 0.31, + "learning_rate": 1.6154147477211593e-05, + "loss": 0.3114, + "step": 5137 + }, + { + "epoch": 0.31, + "learning_rate": 1.615260512337568e-05, + "loss": 0.3535, + "step": 5138 + }, + { + "epoch": 0.31, + "learning_rate": 1.615106253398925e-05, + "loss": 0.3351, + "step": 5139 + }, + { + "epoch": 0.31, + "learning_rate": 1.6149519709111356e-05, + "loss": 0.3122, + "step": 5140 + }, + { + "epoch": 0.31, + "learning_rate": 1.614797664880107e-05, + "loss": 0.3351, + "step": 5141 + }, + { + "epoch": 0.31, + "learning_rate": 1.6146433353117457e-05, + "loss": 0.3471, + "step": 5142 + }, + { + "epoch": 0.31, + "learning_rate": 1.6144889822119613e-05, + "loss": 0.3225, + "step": 5143 + }, + { + "epoch": 0.31, + "learning_rate": 1.614334605586662e-05, + "loss": 0.3418, + "step": 5144 + }, + { + "epoch": 0.31, + "learning_rate": 1.6141802054417595e-05, + "loss": 0.3471, + "step": 5145 + }, + { + "epoch": 0.31, + "learning_rate": 1.6140257817831636e-05, + "loss": 0.3591, + "step": 5146 + }, + { + "epoch": 0.31, + "learning_rate": 1.6138713346167875e-05, + "loss": 0.3295, + "step": 5147 + }, + { + "epoch": 0.31, + "learning_rate": 1.613716863948543e-05, + "loss": 0.3125, + "step": 5148 + }, + { + "epoch": 0.31, + "learning_rate": 1.613562369784345e-05, + "loss": 0.3491, + "step": 5149 + }, + { + "epoch": 0.31, + "learning_rate": 1.6134078521301076e-05, + "loss": 0.3624, + "step": 5150 + }, + { + "epoch": 0.31, + "learning_rate": 1.613253310991747e-05, + "loss": 0.3387, + "step": 5151 + }, + { + "epoch": 0.31, + "learning_rate": 1.6130987463751792e-05, + "loss": 0.3569, + "step": 5152 + }, + { + "epoch": 0.31, + "learning_rate": 1.6129441582863216e-05, + "loss": 0.3491, + "step": 5153 + }, + { + "epoch": 0.31, + "learning_rate": 1.612789546731093e-05, + "loss": 0.3482, + "step": 5154 + }, + { + "epoch": 0.31, + "learning_rate": 1.6126349117154126e-05, + "loss": 0.3287, + "step": 5155 + }, + { + "epoch": 0.31, + "learning_rate": 1.6124802532452006e-05, + "loss": 0.3304, + "step": 5156 + }, + { + "epoch": 0.31, + "learning_rate": 1.6123255713263777e-05, + "loss": 0.3211, + "step": 5157 + }, + { + "epoch": 0.31, + "learning_rate": 1.612170865964866e-05, + "loss": 0.3326, + "step": 5158 + }, + { + "epoch": 0.31, + "learning_rate": 1.6120161371665886e-05, + "loss": 0.3929, + "step": 5159 + }, + { + "epoch": 0.31, + "learning_rate": 1.611861384937469e-05, + "loss": 0.3451, + "step": 5160 + }, + { + "epoch": 0.31, + "learning_rate": 1.611706609283432e-05, + "loss": 0.3532, + "step": 5161 + }, + { + "epoch": 0.31, + "learning_rate": 1.6115518102104026e-05, + "loss": 0.3337, + "step": 5162 + }, + { + "epoch": 0.31, + "learning_rate": 1.611396987724308e-05, + "loss": 0.3555, + "step": 5163 + }, + { + "epoch": 0.31, + "learning_rate": 1.6112421418310747e-05, + "loss": 0.3479, + "step": 5164 + }, + { + "epoch": 0.31, + "learning_rate": 1.611087272536632e-05, + "loss": 0.3521, + "step": 5165 + }, + { + "epoch": 0.31, + "learning_rate": 1.610932379846908e-05, + "loss": 0.3446, + "step": 5166 + }, + { + "epoch": 0.31, + "learning_rate": 1.6107774637678333e-05, + "loss": 0.3552, + "step": 5167 + }, + { + "epoch": 0.31, + "learning_rate": 1.6106225243053387e-05, + "loss": 0.3411, + "step": 5168 + }, + { + "epoch": 0.31, + "learning_rate": 1.610467561465356e-05, + "loss": 0.3806, + "step": 5169 + }, + { + "epoch": 0.31, + "learning_rate": 1.610312575253818e-05, + "loss": 0.3382, + "step": 5170 + }, + { + "epoch": 0.31, + "learning_rate": 1.6101575656766576e-05, + "loss": 0.314, + "step": 5171 + }, + { + "epoch": 0.31, + "learning_rate": 1.6100025327398102e-05, + "loss": 0.3585, + "step": 5172 + }, + { + "epoch": 0.31, + "learning_rate": 1.609847476449211e-05, + "loss": 0.3564, + "step": 5173 + }, + { + "epoch": 0.31, + "learning_rate": 1.6096923968107957e-05, + "loss": 0.3524, + "step": 5174 + }, + { + "epoch": 0.31, + "learning_rate": 1.609537293830502e-05, + "loss": 0.2939, + "step": 5175 + }, + { + "epoch": 0.31, + "learning_rate": 1.6093821675142683e-05, + "loss": 0.37, + "step": 5176 + }, + { + "epoch": 0.31, + "learning_rate": 1.609227017868033e-05, + "loss": 0.3211, + "step": 5177 + }, + { + "epoch": 0.31, + "learning_rate": 1.609071844897736e-05, + "loss": 0.3192, + "step": 5178 + }, + { + "epoch": 0.31, + "learning_rate": 1.608916648609318e-05, + "loss": 0.3546, + "step": 5179 + }, + { + "epoch": 0.31, + "learning_rate": 1.608761429008721e-05, + "loss": 0.3267, + "step": 5180 + }, + { + "epoch": 0.31, + "learning_rate": 1.608606186101887e-05, + "loss": 0.3446, + "step": 5181 + }, + { + "epoch": 0.31, + "learning_rate": 1.6084509198947598e-05, + "loss": 0.3407, + "step": 5182 + }, + { + "epoch": 0.31, + "learning_rate": 1.6082956303932838e-05, + "loss": 0.3295, + "step": 5183 + }, + { + "epoch": 0.31, + "learning_rate": 1.608140317603404e-05, + "loss": 0.3658, + "step": 5184 + }, + { + "epoch": 0.31, + "learning_rate": 1.6079849815310664e-05, + "loss": 0.3471, + "step": 5185 + }, + { + "epoch": 0.31, + "learning_rate": 1.607829622182218e-05, + "loss": 0.3248, + "step": 5186 + }, + { + "epoch": 0.31, + "learning_rate": 1.6076742395628073e-05, + "loss": 0.3484, + "step": 5187 + }, + { + "epoch": 0.31, + "learning_rate": 1.6075188336787824e-05, + "loss": 0.3357, + "step": 5188 + }, + { + "epoch": 0.31, + "learning_rate": 1.607363404536093e-05, + "loss": 0.3569, + "step": 5189 + }, + { + "epoch": 0.31, + "learning_rate": 1.60720795214069e-05, + "loss": 0.3384, + "step": 5190 + }, + { + "epoch": 0.31, + "learning_rate": 1.6070524764985247e-05, + "loss": 0.3404, + "step": 5191 + }, + { + "epoch": 0.31, + "learning_rate": 1.6068969776155494e-05, + "loss": 0.332, + "step": 5192 + }, + { + "epoch": 0.31, + "learning_rate": 1.606741455497717e-05, + "loss": 0.3597, + "step": 5193 + }, + { + "epoch": 0.31, + "learning_rate": 1.6065859101509824e-05, + "loss": 0.3725, + "step": 5194 + }, + { + "epoch": 0.31, + "learning_rate": 1.6064303415813e-05, + "loss": 0.3287, + "step": 5195 + }, + { + "epoch": 0.31, + "learning_rate": 1.6062747497946255e-05, + "loss": 0.3278, + "step": 5196 + }, + { + "epoch": 0.31, + "learning_rate": 1.606119134796916e-05, + "loss": 0.3624, + "step": 5197 + }, + { + "epoch": 0.31, + "learning_rate": 1.6059634965941297e-05, + "loss": 0.3329, + "step": 5198 + }, + { + "epoch": 0.31, + "learning_rate": 1.6058078351922245e-05, + "loss": 0.3365, + "step": 5199 + }, + { + "epoch": 0.31, + "learning_rate": 1.60565215059716e-05, + "loss": 0.3604, + "step": 5200 + }, + { + "epoch": 0.31, + "learning_rate": 1.6054964428148964e-05, + "loss": 0.3599, + "step": 5201 + }, + { + "epoch": 0.31, + "learning_rate": 1.605340711851395e-05, + "loss": 0.3491, + "step": 5202 + }, + { + "epoch": 0.31, + "learning_rate": 1.605184957712618e-05, + "loss": 0.3245, + "step": 5203 + }, + { + "epoch": 0.31, + "learning_rate": 1.6050291804045283e-05, + "loss": 0.3432, + "step": 5204 + }, + { + "epoch": 0.31, + "learning_rate": 1.60487337993309e-05, + "loss": 0.361, + "step": 5205 + }, + { + "epoch": 0.31, + "learning_rate": 1.6047175563042678e-05, + "loss": 0.3329, + "step": 5206 + }, + { + "epoch": 0.31, + "learning_rate": 1.604561709524027e-05, + "loss": 0.3424, + "step": 5207 + }, + { + "epoch": 0.31, + "learning_rate": 1.604405839598335e-05, + "loss": 0.3329, + "step": 5208 + }, + { + "epoch": 0.31, + "learning_rate": 1.604249946533158e-05, + "loss": 0.3103, + "step": 5209 + }, + { + "epoch": 0.31, + "learning_rate": 1.6040940303344654e-05, + "loss": 0.353, + "step": 5210 + }, + { + "epoch": 0.31, + "learning_rate": 1.6039380910082254e-05, + "loss": 0.3225, + "step": 5211 + }, + { + "epoch": 0.31, + "learning_rate": 1.603782128560409e-05, + "loss": 0.3485, + "step": 5212 + }, + { + "epoch": 0.31, + "learning_rate": 1.603626142996987e-05, + "loss": 0.32, + "step": 5213 + }, + { + "epoch": 0.31, + "learning_rate": 1.6034701343239306e-05, + "loss": 0.3633, + "step": 5214 + }, + { + "epoch": 0.32, + "learning_rate": 1.6033141025472134e-05, + "loss": 0.3315, + "step": 5215 + }, + { + "epoch": 0.32, + "learning_rate": 1.6031580476728086e-05, + "loss": 0.3202, + "step": 5216 + }, + { + "epoch": 0.32, + "learning_rate": 1.603001969706691e-05, + "loss": 0.3089, + "step": 5217 + }, + { + "epoch": 0.32, + "learning_rate": 1.6028458686548355e-05, + "loss": 0.3471, + "step": 5218 + }, + { + "epoch": 0.32, + "learning_rate": 1.6026897445232185e-05, + "loss": 0.341, + "step": 5219 + }, + { + "epoch": 0.32, + "learning_rate": 1.6025335973178172e-05, + "loss": 0.3278, + "step": 5220 + }, + { + "epoch": 0.32, + "learning_rate": 1.6023774270446103e-05, + "loss": 0.317, + "step": 5221 + }, + { + "epoch": 0.32, + "learning_rate": 1.602221233709576e-05, + "loss": 0.3337, + "step": 5222 + }, + { + "epoch": 0.32, + "learning_rate": 1.602065017318694e-05, + "loss": 0.3396, + "step": 5223 + }, + { + "epoch": 0.32, + "learning_rate": 1.601908777877945e-05, + "loss": 0.3465, + "step": 5224 + }, + { + "epoch": 0.32, + "learning_rate": 1.6017525153933116e-05, + "loss": 0.3513, + "step": 5225 + }, + { + "epoch": 0.32, + "learning_rate": 1.601596229870775e-05, + "loss": 0.3178, + "step": 5226 + }, + { + "epoch": 0.32, + "learning_rate": 1.6014399213163194e-05, + "loss": 0.3368, + "step": 5227 + }, + { + "epoch": 0.32, + "learning_rate": 1.6012835897359287e-05, + "loss": 0.3602, + "step": 5228 + }, + { + "epoch": 0.32, + "learning_rate": 1.6011272351355874e-05, + "loss": 0.303, + "step": 5229 + }, + { + "epoch": 0.32, + "learning_rate": 1.600970857521283e-05, + "loss": 0.3703, + "step": 5230 + }, + { + "epoch": 0.32, + "learning_rate": 1.600814456899001e-05, + "loss": 0.308, + "step": 5231 + }, + { + "epoch": 0.32, + "learning_rate": 1.6006580332747294e-05, + "loss": 0.3722, + "step": 5232 + }, + { + "epoch": 0.32, + "learning_rate": 1.6005015866544573e-05, + "loss": 0.3655, + "step": 5233 + }, + { + "epoch": 0.32, + "learning_rate": 1.6003451170441734e-05, + "loss": 0.363, + "step": 5234 + }, + { + "epoch": 0.32, + "learning_rate": 1.600188624449869e-05, + "loss": 0.3281, + "step": 5235 + }, + { + "epoch": 0.32, + "learning_rate": 1.600032108877535e-05, + "loss": 0.329, + "step": 5236 + }, + { + "epoch": 0.32, + "learning_rate": 1.5998755703331633e-05, + "loss": 0.3324, + "step": 5237 + }, + { + "epoch": 0.32, + "learning_rate": 1.5997190088227473e-05, + "loss": 0.3499, + "step": 5238 + }, + { + "epoch": 0.32, + "learning_rate": 1.599562424352281e-05, + "loss": 0.361, + "step": 5239 + }, + { + "epoch": 0.32, + "learning_rate": 1.5994058169277588e-05, + "loss": 0.3378, + "step": 5240 + }, + { + "epoch": 0.32, + "learning_rate": 1.5992491865551766e-05, + "loss": 0.3571, + "step": 5241 + }, + { + "epoch": 0.32, + "learning_rate": 1.5990925332405308e-05, + "loss": 0.3228, + "step": 5242 + }, + { + "epoch": 0.32, + "learning_rate": 1.598935856989819e-05, + "loss": 0.3836, + "step": 5243 + }, + { + "epoch": 0.32, + "learning_rate": 1.598779157809039e-05, + "loss": 0.3345, + "step": 5244 + }, + { + "epoch": 0.32, + "learning_rate": 1.5986224357041902e-05, + "loss": 0.3571, + "step": 5245 + }, + { + "epoch": 0.32, + "learning_rate": 1.5984656906812734e-05, + "loss": 0.3181, + "step": 5246 + }, + { + "epoch": 0.32, + "learning_rate": 1.5983089227462888e-05, + "loss": 0.3351, + "step": 5247 + }, + { + "epoch": 0.32, + "learning_rate": 1.598152131905238e-05, + "loss": 0.3186, + "step": 5248 + }, + { + "epoch": 0.32, + "learning_rate": 1.5979953181641246e-05, + "loss": 0.3457, + "step": 5249 + }, + { + "epoch": 0.32, + "learning_rate": 1.5978384815289512e-05, + "loss": 0.3253, + "step": 5250 + }, + { + "epoch": 0.32, + "learning_rate": 1.597681622005723e-05, + "loss": 0.3527, + "step": 5251 + }, + { + "epoch": 0.32, + "learning_rate": 1.5975247396004448e-05, + "loss": 0.3318, + "step": 5252 + }, + { + "epoch": 0.32, + "learning_rate": 1.597367834319123e-05, + "loss": 0.3549, + "step": 5253 + }, + { + "epoch": 0.32, + "learning_rate": 1.5972109061677646e-05, + "loss": 0.3415, + "step": 5254 + }, + { + "epoch": 0.32, + "learning_rate": 1.5970539551523774e-05, + "loss": 0.3532, + "step": 5255 + }, + { + "epoch": 0.32, + "learning_rate": 1.5968969812789707e-05, + "loss": 0.363, + "step": 5256 + }, + { + "epoch": 0.32, + "learning_rate": 1.5967399845535535e-05, + "loss": 0.3164, + "step": 5257 + }, + { + "epoch": 0.32, + "learning_rate": 1.5965829649821368e-05, + "loss": 0.3318, + "step": 5258 + }, + { + "epoch": 0.32, + "learning_rate": 1.5964259225707325e-05, + "loss": 0.3661, + "step": 5259 + }, + { + "epoch": 0.32, + "learning_rate": 1.596268857325352e-05, + "loss": 0.3542, + "step": 5260 + }, + { + "epoch": 0.32, + "learning_rate": 1.596111769252009e-05, + "loss": 0.3407, + "step": 5261 + }, + { + "epoch": 0.32, + "learning_rate": 1.5959546583567175e-05, + "loss": 0.3343, + "step": 5262 + }, + { + "epoch": 0.32, + "learning_rate": 1.5957975246454926e-05, + "loss": 0.3594, + "step": 5263 + }, + { + "epoch": 0.32, + "learning_rate": 1.5956403681243497e-05, + "loss": 0.3259, + "step": 5264 + }, + { + "epoch": 0.32, + "learning_rate": 1.5954831887993055e-05, + "loss": 0.3438, + "step": 5265 + }, + { + "epoch": 0.32, + "learning_rate": 1.5953259866763783e-05, + "loss": 0.332, + "step": 5266 + }, + { + "epoch": 0.32, + "learning_rate": 1.5951687617615855e-05, + "loss": 0.3591, + "step": 5267 + }, + { + "epoch": 0.32, + "learning_rate": 1.5950115140609474e-05, + "loss": 0.3597, + "step": 5268 + }, + { + "epoch": 0.32, + "learning_rate": 1.5948542435804838e-05, + "loss": 0.3683, + "step": 5269 + }, + { + "epoch": 0.32, + "learning_rate": 1.5946969503262152e-05, + "loss": 0.3432, + "step": 5270 + }, + { + "epoch": 0.32, + "learning_rate": 1.594539634304164e-05, + "loss": 0.3345, + "step": 5271 + }, + { + "epoch": 0.32, + "learning_rate": 1.594382295520353e-05, + "loss": 0.3326, + "step": 5272 + }, + { + "epoch": 0.32, + "learning_rate": 1.594224933980806e-05, + "loss": 0.3783, + "step": 5273 + }, + { + "epoch": 0.32, + "learning_rate": 1.5940675496915473e-05, + "loss": 0.3465, + "step": 5274 + }, + { + "epoch": 0.32, + "learning_rate": 1.593910142658602e-05, + "loss": 0.3351, + "step": 5275 + }, + { + "epoch": 0.32, + "learning_rate": 1.593752712887997e-05, + "loss": 0.3418, + "step": 5276 + }, + { + "epoch": 0.32, + "learning_rate": 1.5935952603857594e-05, + "loss": 0.3347, + "step": 5277 + }, + { + "epoch": 0.32, + "learning_rate": 1.593437785157917e-05, + "loss": 0.3315, + "step": 5278 + }, + { + "epoch": 0.32, + "learning_rate": 1.5932802872104987e-05, + "loss": 0.3237, + "step": 5279 + }, + { + "epoch": 0.32, + "learning_rate": 1.593122766549534e-05, + "loss": 0.3398, + "step": 5280 + }, + { + "epoch": 0.32, + "learning_rate": 1.5929652231810542e-05, + "loss": 0.341, + "step": 5281 + }, + { + "epoch": 0.32, + "learning_rate": 1.59280765711109e-05, + "loss": 0.3276, + "step": 5282 + }, + { + "epoch": 0.32, + "learning_rate": 1.5926500683456744e-05, + "loss": 0.3585, + "step": 5283 + }, + { + "epoch": 0.32, + "learning_rate": 1.59249245689084e-05, + "loss": 0.353, + "step": 5284 + }, + { + "epoch": 0.32, + "learning_rate": 1.5923348227526218e-05, + "loss": 0.3546, + "step": 5285 + }, + { + "epoch": 0.32, + "learning_rate": 1.5921771659370543e-05, + "loss": 0.3351, + "step": 5286 + }, + { + "epoch": 0.32, + "learning_rate": 1.592019486450173e-05, + "loss": 0.3306, + "step": 5287 + }, + { + "epoch": 0.32, + "learning_rate": 1.591861784298015e-05, + "loss": 0.3117, + "step": 5288 + }, + { + "epoch": 0.32, + "learning_rate": 1.5917040594866177e-05, + "loss": 0.358, + "step": 5289 + }, + { + "epoch": 0.32, + "learning_rate": 1.5915463120220195e-05, + "loss": 0.3064, + "step": 5290 + }, + { + "epoch": 0.32, + "learning_rate": 1.59138854191026e-05, + "loss": 0.3783, + "step": 5291 + }, + { + "epoch": 0.32, + "learning_rate": 1.5912307491573795e-05, + "loss": 0.315, + "step": 5292 + }, + { + "epoch": 0.32, + "learning_rate": 1.5910729337694187e-05, + "loss": 0.3276, + "step": 5293 + }, + { + "epoch": 0.32, + "learning_rate": 1.5909150957524194e-05, + "loss": 0.3524, + "step": 5294 + }, + { + "epoch": 0.32, + "learning_rate": 1.5907572351124247e-05, + "loss": 0.334, + "step": 5295 + }, + { + "epoch": 0.32, + "learning_rate": 1.5905993518554777e-05, + "loss": 0.3252, + "step": 5296 + }, + { + "epoch": 0.32, + "learning_rate": 1.590441445987624e-05, + "loss": 0.3022, + "step": 5297 + }, + { + "epoch": 0.32, + "learning_rate": 1.590283517514908e-05, + "loss": 0.3337, + "step": 5298 + }, + { + "epoch": 0.32, + "learning_rate": 1.590125566443376e-05, + "loss": 0.3544, + "step": 5299 + }, + { + "epoch": 0.32, + "learning_rate": 1.5899675927790754e-05, + "loss": 0.329, + "step": 5300 + }, + { + "epoch": 0.32, + "learning_rate": 1.5898095965280545e-05, + "loss": 0.3298, + "step": 5301 + }, + { + "epoch": 0.32, + "learning_rate": 1.5896515776963614e-05, + "loss": 0.3203, + "step": 5302 + }, + { + "epoch": 0.32, + "learning_rate": 1.5894935362900463e-05, + "loss": 0.3298, + "step": 5303 + }, + { + "epoch": 0.32, + "learning_rate": 1.5893354723151594e-05, + "loss": 0.3449, + "step": 5304 + }, + { + "epoch": 0.32, + "learning_rate": 1.5891773857777525e-05, + "loss": 0.3156, + "step": 5305 + }, + { + "epoch": 0.32, + "learning_rate": 1.5890192766838775e-05, + "loss": 0.2965, + "step": 5306 + }, + { + "epoch": 0.32, + "learning_rate": 1.5888611450395884e-05, + "loss": 0.3304, + "step": 5307 + }, + { + "epoch": 0.32, + "learning_rate": 1.588702990850938e-05, + "loss": 0.3569, + "step": 5308 + }, + { + "epoch": 0.32, + "learning_rate": 1.5885448141239823e-05, + "loss": 0.3214, + "step": 5309 + }, + { + "epoch": 0.32, + "learning_rate": 1.588386614864776e-05, + "loss": 0.3178, + "step": 5310 + }, + { + "epoch": 0.32, + "learning_rate": 1.5882283930793768e-05, + "loss": 0.3502, + "step": 5311 + }, + { + "epoch": 0.32, + "learning_rate": 1.588070148773841e-05, + "loss": 0.3552, + "step": 5312 + }, + { + "epoch": 0.32, + "learning_rate": 1.587911881954228e-05, + "loss": 0.4051, + "step": 5313 + }, + { + "epoch": 0.32, + "learning_rate": 1.5877535926265965e-05, + "loss": 0.3449, + "step": 5314 + }, + { + "epoch": 0.32, + "learning_rate": 1.587595280797006e-05, + "loss": 0.3362, + "step": 5315 + }, + { + "epoch": 0.32, + "learning_rate": 1.587436946471519e-05, + "loss": 0.3499, + "step": 5316 + }, + { + "epoch": 0.32, + "learning_rate": 1.5872785896561956e-05, + "loss": 0.3401, + "step": 5317 + }, + { + "epoch": 0.32, + "learning_rate": 1.5871202103570995e-05, + "loss": 0.3184, + "step": 5318 + }, + { + "epoch": 0.32, + "learning_rate": 1.5869618085802937e-05, + "loss": 0.3365, + "step": 5319 + }, + { + "epoch": 0.32, + "learning_rate": 1.5868033843318428e-05, + "loss": 0.3524, + "step": 5320 + }, + { + "epoch": 0.32, + "learning_rate": 1.5866449376178118e-05, + "loss": 0.3295, + "step": 5321 + }, + { + "epoch": 0.32, + "learning_rate": 1.586486468444267e-05, + "loss": 0.3323, + "step": 5322 + }, + { + "epoch": 0.32, + "learning_rate": 1.586327976817275e-05, + "loss": 0.3797, + "step": 5323 + }, + { + "epoch": 0.32, + "learning_rate": 1.5861694627429042e-05, + "loss": 0.3535, + "step": 5324 + }, + { + "epoch": 0.32, + "learning_rate": 1.5860109262272224e-05, + "loss": 0.3786, + "step": 5325 + }, + { + "epoch": 0.32, + "learning_rate": 1.5858523672763e-05, + "loss": 0.3301, + "step": 5326 + }, + { + "epoch": 0.32, + "learning_rate": 1.585693785896207e-05, + "loss": 0.3331, + "step": 5327 + }, + { + "epoch": 0.32, + "learning_rate": 1.5855351820930146e-05, + "loss": 0.3329, + "step": 5328 + }, + { + "epoch": 0.32, + "learning_rate": 1.5853765558727948e-05, + "loss": 0.3675, + "step": 5329 + }, + { + "epoch": 0.32, + "learning_rate": 1.5852179072416207e-05, + "loss": 0.3284, + "step": 5330 + }, + { + "epoch": 0.32, + "learning_rate": 1.5850592362055663e-05, + "loss": 0.3217, + "step": 5331 + }, + { + "epoch": 0.32, + "learning_rate": 1.584900542770706e-05, + "loss": 0.3404, + "step": 5332 + }, + { + "epoch": 0.32, + "learning_rate": 1.5847418269431154e-05, + "loss": 0.382, + "step": 5333 + }, + { + "epoch": 0.32, + "learning_rate": 1.584583088728871e-05, + "loss": 0.322, + "step": 5334 + }, + { + "epoch": 0.32, + "learning_rate": 1.5844243281340497e-05, + "loss": 0.315, + "step": 5335 + }, + { + "epoch": 0.32, + "learning_rate": 1.5842655451647302e-05, + "loss": 0.3362, + "step": 5336 + }, + { + "epoch": 0.32, + "learning_rate": 1.5841067398269907e-05, + "loss": 0.3412, + "step": 5337 + }, + { + "epoch": 0.32, + "learning_rate": 1.5839479121269113e-05, + "loss": 0.3376, + "step": 5338 + }, + { + "epoch": 0.32, + "learning_rate": 1.5837890620705732e-05, + "loss": 0.3398, + "step": 5339 + }, + { + "epoch": 0.32, + "learning_rate": 1.5836301896640572e-05, + "loss": 0.3465, + "step": 5340 + }, + { + "epoch": 0.32, + "learning_rate": 1.583471294913446e-05, + "loss": 0.3287, + "step": 5341 + }, + { + "epoch": 0.32, + "learning_rate": 1.5833123778248226e-05, + "loss": 0.3192, + "step": 5342 + }, + { + "epoch": 0.32, + "learning_rate": 1.5831534384042714e-05, + "loss": 0.3284, + "step": 5343 + }, + { + "epoch": 0.32, + "learning_rate": 1.5829944766578776e-05, + "loss": 0.3396, + "step": 5344 + }, + { + "epoch": 0.32, + "learning_rate": 1.5828354925917264e-05, + "loss": 0.3245, + "step": 5345 + }, + { + "epoch": 0.32, + "learning_rate": 1.5826764862119045e-05, + "loss": 0.3574, + "step": 5346 + }, + { + "epoch": 0.32, + "learning_rate": 1.5825174575245e-05, + "loss": 0.3312, + "step": 5347 + }, + { + "epoch": 0.32, + "learning_rate": 1.5823584065356007e-05, + "loss": 0.3362, + "step": 5348 + }, + { + "epoch": 0.32, + "learning_rate": 1.5821993332512957e-05, + "loss": 0.3273, + "step": 5349 + }, + { + "epoch": 0.32, + "learning_rate": 1.5820402376776756e-05, + "loss": 0.3451, + "step": 5350 + }, + { + "epoch": 0.32, + "learning_rate": 1.5818811198208307e-05, + "loss": 0.3175, + "step": 5351 + }, + { + "epoch": 0.32, + "learning_rate": 1.5817219796868536e-05, + "loss": 0.3351, + "step": 5352 + }, + { + "epoch": 0.32, + "learning_rate": 1.5815628172818365e-05, + "loss": 0.3013, + "step": 5353 + }, + { + "epoch": 0.32, + "learning_rate": 1.5814036326118726e-05, + "loss": 0.3664, + "step": 5354 + }, + { + "epoch": 0.32, + "learning_rate": 1.5812444256830566e-05, + "loss": 0.3292, + "step": 5355 + }, + { + "epoch": 0.32, + "learning_rate": 1.5810851965014836e-05, + "loss": 0.3457, + "step": 5356 + }, + { + "epoch": 0.32, + "learning_rate": 1.5809259450732495e-05, + "loss": 0.344, + "step": 5357 + }, + { + "epoch": 0.32, + "learning_rate": 1.5807666714044514e-05, + "loss": 0.3415, + "step": 5358 + }, + { + "epoch": 0.32, + "learning_rate": 1.5806073755011868e-05, + "loss": 0.3175, + "step": 5359 + }, + { + "epoch": 0.32, + "learning_rate": 1.5804480573695547e-05, + "loss": 0.3552, + "step": 5360 + }, + { + "epoch": 0.32, + "learning_rate": 1.580288717015654e-05, + "loss": 0.3306, + "step": 5361 + }, + { + "epoch": 0.32, + "learning_rate": 1.5801293544455853e-05, + "loss": 0.3185, + "step": 5362 + }, + { + "epoch": 0.32, + "learning_rate": 1.5799699696654497e-05, + "loss": 0.3323, + "step": 5363 + }, + { + "epoch": 0.32, + "learning_rate": 1.5798105626813492e-05, + "loss": 0.3368, + "step": 5364 + }, + { + "epoch": 0.32, + "learning_rate": 1.579651133499387e-05, + "loss": 0.3384, + "step": 5365 + }, + { + "epoch": 0.32, + "learning_rate": 1.579491682125666e-05, + "loss": 0.3719, + "step": 5366 + }, + { + "epoch": 0.32, + "learning_rate": 1.5793322085662916e-05, + "loss": 0.3552, + "step": 5367 + }, + { + "epoch": 0.32, + "learning_rate": 1.5791727128273687e-05, + "loss": 0.3211, + "step": 5368 + }, + { + "epoch": 0.32, + "learning_rate": 1.5790131949150035e-05, + "loss": 0.317, + "step": 5369 + }, + { + "epoch": 0.32, + "learning_rate": 1.5788536548353035e-05, + "loss": 0.3214, + "step": 5370 + }, + { + "epoch": 0.32, + "learning_rate": 1.5786940925943762e-05, + "loss": 0.3435, + "step": 5371 + }, + { + "epoch": 0.32, + "learning_rate": 1.5785345081983305e-05, + "loss": 0.3262, + "step": 5372 + }, + { + "epoch": 0.32, + "learning_rate": 1.5783749016532762e-05, + "loss": 0.3438, + "step": 5373 + }, + { + "epoch": 0.32, + "learning_rate": 1.5782152729653237e-05, + "loss": 0.3114, + "step": 5374 + }, + { + "epoch": 0.32, + "learning_rate": 1.5780556221405843e-05, + "loss": 0.3454, + "step": 5375 + }, + { + "epoch": 0.32, + "learning_rate": 1.57789594918517e-05, + "loss": 0.3555, + "step": 5376 + }, + { + "epoch": 0.32, + "learning_rate": 1.5777362541051943e-05, + "loss": 0.3189, + "step": 5377 + }, + { + "epoch": 0.32, + "learning_rate": 1.5775765369067708e-05, + "loss": 0.3415, + "step": 5378 + }, + { + "epoch": 0.32, + "learning_rate": 1.5774167975960142e-05, + "loss": 0.3549, + "step": 5379 + }, + { + "epoch": 0.33, + "learning_rate": 1.5772570361790398e-05, + "loss": 0.3142, + "step": 5380 + }, + { + "epoch": 0.33, + "learning_rate": 1.5770972526619647e-05, + "loss": 0.3117, + "step": 5381 + }, + { + "epoch": 0.33, + "learning_rate": 1.5769374470509052e-05, + "loss": 0.3418, + "step": 5382 + }, + { + "epoch": 0.33, + "learning_rate": 1.5767776193519804e-05, + "loss": 0.3348, + "step": 5383 + }, + { + "epoch": 0.33, + "learning_rate": 1.5766177695713088e-05, + "loss": 0.3175, + "step": 5384 + }, + { + "epoch": 0.33, + "learning_rate": 1.5764578977150104e-05, + "loss": 0.356, + "step": 5385 + }, + { + "epoch": 0.33, + "learning_rate": 1.5762980037892056e-05, + "loss": 0.332, + "step": 5386 + }, + { + "epoch": 0.33, + "learning_rate": 1.576138087800016e-05, + "loss": 0.327, + "step": 5387 + }, + { + "epoch": 0.33, + "learning_rate": 1.5759781497535636e-05, + "loss": 0.3276, + "step": 5388 + }, + { + "epoch": 0.33, + "learning_rate": 1.5758181896559718e-05, + "loss": 0.3262, + "step": 5389 + }, + { + "epoch": 0.33, + "learning_rate": 1.575658207513365e-05, + "loss": 0.3298, + "step": 5390 + }, + { + "epoch": 0.33, + "learning_rate": 1.5754982033318676e-05, + "loss": 0.3396, + "step": 5391 + }, + { + "epoch": 0.33, + "learning_rate": 1.5753381771176057e-05, + "loss": 0.3616, + "step": 5392 + }, + { + "epoch": 0.33, + "learning_rate": 1.5751781288767052e-05, + "loss": 0.3295, + "step": 5393 + }, + { + "epoch": 0.33, + "learning_rate": 1.575018058615294e-05, + "loss": 0.322, + "step": 5394 + }, + { + "epoch": 0.33, + "learning_rate": 1.5748579663395004e-05, + "loss": 0.3518, + "step": 5395 + }, + { + "epoch": 0.33, + "learning_rate": 1.5746978520554533e-05, + "loss": 0.3549, + "step": 5396 + }, + { + "epoch": 0.33, + "learning_rate": 1.5745377157692826e-05, + "loss": 0.3398, + "step": 5397 + }, + { + "epoch": 0.33, + "learning_rate": 1.5743775574871193e-05, + "loss": 0.3583, + "step": 5398 + }, + { + "epoch": 0.33, + "learning_rate": 1.5742173772150948e-05, + "loss": 0.3622, + "step": 5399 + }, + { + "epoch": 0.33, + "learning_rate": 1.5740571749593415e-05, + "loss": 0.3331, + "step": 5400 + }, + { + "epoch": 0.33, + "learning_rate": 1.5738969507259926e-05, + "loss": 0.3323, + "step": 5401 + }, + { + "epoch": 0.33, + "learning_rate": 1.5737367045211826e-05, + "loss": 0.3117, + "step": 5402 + }, + { + "epoch": 0.33, + "learning_rate": 1.573576436351046e-05, + "loss": 0.3329, + "step": 5403 + }, + { + "epoch": 0.33, + "learning_rate": 1.5734161462217195e-05, + "loss": 0.3577, + "step": 5404 + }, + { + "epoch": 0.33, + "learning_rate": 1.5732558341393388e-05, + "loss": 0.358, + "step": 5405 + }, + { + "epoch": 0.33, + "learning_rate": 1.5730955001100417e-05, + "loss": 0.373, + "step": 5406 + }, + { + "epoch": 0.33, + "learning_rate": 1.5729351441399665e-05, + "loss": 0.3032, + "step": 5407 + }, + { + "epoch": 0.33, + "learning_rate": 1.5727747662352525e-05, + "loss": 0.3549, + "step": 5408 + }, + { + "epoch": 0.33, + "learning_rate": 1.57261436640204e-05, + "loss": 0.3097, + "step": 5409 + }, + { + "epoch": 0.33, + "learning_rate": 1.5724539446464692e-05, + "loss": 0.3351, + "step": 5410 + }, + { + "epoch": 0.33, + "learning_rate": 1.5722935009746822e-05, + "loss": 0.3164, + "step": 5411 + }, + { + "epoch": 0.33, + "learning_rate": 1.5721330353928214e-05, + "loss": 0.3203, + "step": 5412 + }, + { + "epoch": 0.33, + "learning_rate": 1.5719725479070305e-05, + "loss": 0.3073, + "step": 5413 + }, + { + "epoch": 0.33, + "learning_rate": 1.571812038523453e-05, + "loss": 0.3227, + "step": 5414 + }, + { + "epoch": 0.33, + "learning_rate": 1.571651507248235e-05, + "loss": 0.3365, + "step": 5415 + }, + { + "epoch": 0.33, + "learning_rate": 1.5714909540875215e-05, + "loss": 0.3418, + "step": 5416 + }, + { + "epoch": 0.33, + "learning_rate": 1.5713303790474595e-05, + "loss": 0.3287, + "step": 5417 + }, + { + "epoch": 0.33, + "learning_rate": 1.5711697821341966e-05, + "loss": 0.3527, + "step": 5418 + }, + { + "epoch": 0.33, + "learning_rate": 1.5710091633538807e-05, + "loss": 0.3309, + "step": 5419 + }, + { + "epoch": 0.33, + "learning_rate": 1.5708485227126626e-05, + "loss": 0.3471, + "step": 5420 + }, + { + "epoch": 0.33, + "learning_rate": 1.5706878602166903e-05, + "loss": 0.3237, + "step": 5421 + }, + { + "epoch": 0.33, + "learning_rate": 1.5705271758721164e-05, + "loss": 0.3278, + "step": 5422 + }, + { + "epoch": 0.33, + "learning_rate": 1.5703664696850914e-05, + "loss": 0.3315, + "step": 5423 + }, + { + "epoch": 0.33, + "learning_rate": 1.5702057416617687e-05, + "loss": 0.3471, + "step": 5424 + }, + { + "epoch": 0.33, + "learning_rate": 1.5700449918083017e-05, + "loss": 0.3054, + "step": 5425 + }, + { + "epoch": 0.33, + "learning_rate": 1.569884220130844e-05, + "loss": 0.3241, + "step": 5426 + }, + { + "epoch": 0.33, + "learning_rate": 1.5697234266355513e-05, + "loss": 0.3287, + "step": 5427 + }, + { + "epoch": 0.33, + "learning_rate": 1.5695626113285792e-05, + "loss": 0.3739, + "step": 5428 + }, + { + "epoch": 0.33, + "learning_rate": 1.569401774216085e-05, + "loss": 0.356, + "step": 5429 + }, + { + "epoch": 0.33, + "learning_rate": 1.5692409153042255e-05, + "loss": 0.3435, + "step": 5430 + }, + { + "epoch": 0.33, + "learning_rate": 1.5690800345991596e-05, + "loss": 0.3139, + "step": 5431 + }, + { + "epoch": 0.33, + "learning_rate": 1.5689191321070465e-05, + "loss": 0.3446, + "step": 5432 + }, + { + "epoch": 0.33, + "learning_rate": 1.5687582078340464e-05, + "loss": 0.3196, + "step": 5433 + }, + { + "epoch": 0.33, + "learning_rate": 1.5685972617863205e-05, + "loss": 0.3309, + "step": 5434 + }, + { + "epoch": 0.33, + "learning_rate": 1.5684362939700294e-05, + "loss": 0.3297, + "step": 5435 + }, + { + "epoch": 0.33, + "learning_rate": 1.568275304391337e-05, + "loss": 0.344, + "step": 5436 + }, + { + "epoch": 0.33, + "learning_rate": 1.5681142930564063e-05, + "loss": 0.329, + "step": 5437 + }, + { + "epoch": 0.33, + "learning_rate": 1.5679532599714013e-05, + "loss": 0.3371, + "step": 5438 + }, + { + "epoch": 0.33, + "learning_rate": 1.5677922051424876e-05, + "loss": 0.3544, + "step": 5439 + }, + { + "epoch": 0.33, + "learning_rate": 1.5676311285758306e-05, + "loss": 0.346, + "step": 5440 + }, + { + "epoch": 0.33, + "learning_rate": 1.567470030277597e-05, + "loss": 0.304, + "step": 5441 + }, + { + "epoch": 0.33, + "learning_rate": 1.5673089102539553e-05, + "loss": 0.3242, + "step": 5442 + }, + { + "epoch": 0.33, + "learning_rate": 1.567147768511073e-05, + "loss": 0.3398, + "step": 5443 + }, + { + "epoch": 0.33, + "learning_rate": 1.5669866050551195e-05, + "loss": 0.3488, + "step": 5444 + }, + { + "epoch": 0.33, + "learning_rate": 1.5668254198922653e-05, + "loss": 0.3217, + "step": 5445 + }, + { + "epoch": 0.33, + "learning_rate": 1.566664213028681e-05, + "loss": 0.317, + "step": 5446 + }, + { + "epoch": 0.33, + "learning_rate": 1.5665029844705385e-05, + "loss": 0.3622, + "step": 5447 + }, + { + "epoch": 0.33, + "learning_rate": 1.5663417342240104e-05, + "loss": 0.298, + "step": 5448 + }, + { + "epoch": 0.33, + "learning_rate": 1.5661804622952703e-05, + "loss": 0.3273, + "step": 5449 + }, + { + "epoch": 0.33, + "learning_rate": 1.5660191686904918e-05, + "loss": 0.3357, + "step": 5450 + }, + { + "epoch": 0.33, + "learning_rate": 1.5658578534158505e-05, + "loss": 0.3295, + "step": 5451 + }, + { + "epoch": 0.33, + "learning_rate": 1.565696516477522e-05, + "loss": 0.3354, + "step": 5452 + }, + { + "epoch": 0.33, + "learning_rate": 1.5655351578816836e-05, + "loss": 0.3192, + "step": 5453 + }, + { + "epoch": 0.33, + "learning_rate": 1.5653737776345124e-05, + "loss": 0.3087, + "step": 5454 + }, + { + "epoch": 0.33, + "learning_rate": 1.5652123757421867e-05, + "loss": 0.3518, + "step": 5455 + }, + { + "epoch": 0.33, + "learning_rate": 1.5650509522108862e-05, + "loss": 0.3239, + "step": 5456 + }, + { + "epoch": 0.33, + "learning_rate": 1.56488950704679e-05, + "loss": 0.3193, + "step": 5457 + }, + { + "epoch": 0.33, + "learning_rate": 1.56472804025608e-05, + "loss": 0.3541, + "step": 5458 + }, + { + "epoch": 0.33, + "learning_rate": 1.5645665518449374e-05, + "loss": 0.3393, + "step": 5459 + }, + { + "epoch": 0.33, + "learning_rate": 1.564405041819545e-05, + "loss": 0.3398, + "step": 5460 + }, + { + "epoch": 0.33, + "learning_rate": 1.564243510186086e-05, + "loss": 0.3239, + "step": 5461 + }, + { + "epoch": 0.33, + "learning_rate": 1.5640819569507443e-05, + "loss": 0.3271, + "step": 5462 + }, + { + "epoch": 0.33, + "learning_rate": 1.5639203821197056e-05, + "loss": 0.3398, + "step": 5463 + }, + { + "epoch": 0.33, + "learning_rate": 1.5637587856991553e-05, + "loss": 0.3546, + "step": 5464 + }, + { + "epoch": 0.33, + "learning_rate": 1.56359716769528e-05, + "loss": 0.3557, + "step": 5465 + }, + { + "epoch": 0.33, + "learning_rate": 1.563435528114267e-05, + "loss": 0.3387, + "step": 5466 + }, + { + "epoch": 0.33, + "learning_rate": 1.5632738669623053e-05, + "loss": 0.3588, + "step": 5467 + }, + { + "epoch": 0.33, + "learning_rate": 1.5631121842455836e-05, + "loss": 0.3181, + "step": 5468 + }, + { + "epoch": 0.33, + "learning_rate": 1.562950479970292e-05, + "loss": 0.3795, + "step": 5469 + }, + { + "epoch": 0.33, + "learning_rate": 1.562788754142621e-05, + "loss": 0.377, + "step": 5470 + }, + { + "epoch": 0.33, + "learning_rate": 1.5626270067687627e-05, + "loss": 0.3382, + "step": 5471 + }, + { + "epoch": 0.33, + "learning_rate": 1.5624652378549092e-05, + "loss": 0.3362, + "step": 5472 + }, + { + "epoch": 0.33, + "learning_rate": 1.562303447407254e-05, + "loss": 0.3524, + "step": 5473 + }, + { + "epoch": 0.33, + "learning_rate": 1.562141635431991e-05, + "loss": 0.3292, + "step": 5474 + }, + { + "epoch": 0.33, + "learning_rate": 1.561979801935315e-05, + "loss": 0.3499, + "step": 5475 + }, + { + "epoch": 0.33, + "learning_rate": 1.561817946923422e-05, + "loss": 0.3599, + "step": 5476 + }, + { + "epoch": 0.33, + "learning_rate": 1.5616560704025088e-05, + "loss": 0.3133, + "step": 5477 + }, + { + "epoch": 0.33, + "learning_rate": 1.5614941723787723e-05, + "loss": 0.329, + "step": 5478 + }, + { + "epoch": 0.33, + "learning_rate": 1.561332252858411e-05, + "loss": 0.309, + "step": 5479 + }, + { + "epoch": 0.33, + "learning_rate": 1.5611703118476236e-05, + "loss": 0.3345, + "step": 5480 + }, + { + "epoch": 0.33, + "learning_rate": 1.5610083493526103e-05, + "loss": 0.3438, + "step": 5481 + }, + { + "epoch": 0.33, + "learning_rate": 1.560846365379572e-05, + "loss": 0.3393, + "step": 5482 + }, + { + "epoch": 0.33, + "learning_rate": 1.5606843599347095e-05, + "loss": 0.363, + "step": 5483 + }, + { + "epoch": 0.33, + "learning_rate": 1.5605223330242256e-05, + "loss": 0.3295, + "step": 5484 + }, + { + "epoch": 0.33, + "learning_rate": 1.5603602846543236e-05, + "loss": 0.3368, + "step": 5485 + }, + { + "epoch": 0.33, + "learning_rate": 1.560198214831207e-05, + "loss": 0.329, + "step": 5486 + }, + { + "epoch": 0.33, + "learning_rate": 1.5600361235610807e-05, + "loss": 0.32, + "step": 5487 + }, + { + "epoch": 0.33, + "learning_rate": 1.559874010850151e-05, + "loss": 0.3546, + "step": 5488 + }, + { + "epoch": 0.33, + "learning_rate": 1.5597118767046233e-05, + "loss": 0.3652, + "step": 5489 + }, + { + "epoch": 0.33, + "learning_rate": 1.5595497211307055e-05, + "loss": 0.3383, + "step": 5490 + }, + { + "epoch": 0.33, + "learning_rate": 1.5593875441346055e-05, + "loss": 0.3258, + "step": 5491 + }, + { + "epoch": 0.33, + "learning_rate": 1.559225345722532e-05, + "loss": 0.3295, + "step": 5492 + }, + { + "epoch": 0.33, + "learning_rate": 1.559063125900695e-05, + "loss": 0.3373, + "step": 5493 + }, + { + "epoch": 0.33, + "learning_rate": 1.558900884675305e-05, + "loss": 0.3502, + "step": 5494 + }, + { + "epoch": 0.33, + "learning_rate": 1.5587386220525737e-05, + "loss": 0.3396, + "step": 5495 + }, + { + "epoch": 0.33, + "learning_rate": 1.5585763380387122e-05, + "loss": 0.3248, + "step": 5496 + }, + { + "epoch": 0.33, + "learning_rate": 1.5584140326399343e-05, + "loss": 0.3496, + "step": 5497 + }, + { + "epoch": 0.33, + "learning_rate": 1.558251705862454e-05, + "loss": 0.334, + "step": 5498 + }, + { + "epoch": 0.33, + "learning_rate": 1.5580893577124854e-05, + "loss": 0.3583, + "step": 5499 + }, + { + "epoch": 0.33, + "learning_rate": 1.557926988196244e-05, + "loss": 0.3781, + "step": 5500 + }, + { + "epoch": 0.33, + "learning_rate": 1.5577645973199468e-05, + "loss": 0.3326, + "step": 5501 + }, + { + "epoch": 0.33, + "learning_rate": 1.5576021850898096e-05, + "loss": 0.3011, + "step": 5502 + }, + { + "epoch": 0.33, + "learning_rate": 1.5574397515120515e-05, + "loss": 0.3117, + "step": 5503 + }, + { + "epoch": 0.33, + "learning_rate": 1.5572772965928903e-05, + "loss": 0.3362, + "step": 5504 + }, + { + "epoch": 0.33, + "learning_rate": 1.5571148203385464e-05, + "loss": 0.344, + "step": 5505 + }, + { + "epoch": 0.33, + "learning_rate": 1.5569523227552393e-05, + "loss": 0.3443, + "step": 5506 + }, + { + "epoch": 0.33, + "learning_rate": 1.5567898038491906e-05, + "loss": 0.3348, + "step": 5507 + }, + { + "epoch": 0.33, + "learning_rate": 1.5566272636266227e-05, + "loss": 0.3697, + "step": 5508 + }, + { + "epoch": 0.33, + "learning_rate": 1.5564647020937576e-05, + "loss": 0.3362, + "step": 5509 + }, + { + "epoch": 0.33, + "learning_rate": 1.5563021192568194e-05, + "loss": 0.3306, + "step": 5510 + }, + { + "epoch": 0.33, + "learning_rate": 1.556139515122032e-05, + "loss": 0.3566, + "step": 5511 + }, + { + "epoch": 0.33, + "learning_rate": 1.5559768896956214e-05, + "loss": 0.3295, + "step": 5512 + }, + { + "epoch": 0.33, + "learning_rate": 1.5558142429838136e-05, + "loss": 0.3119, + "step": 5513 + }, + { + "epoch": 0.33, + "learning_rate": 1.5556515749928345e-05, + "loss": 0.3454, + "step": 5514 + }, + { + "epoch": 0.33, + "learning_rate": 1.555488885728913e-05, + "loss": 0.3097, + "step": 5515 + }, + { + "epoch": 0.33, + "learning_rate": 1.5553261751982768e-05, + "loss": 0.3136, + "step": 5516 + }, + { + "epoch": 0.33, + "learning_rate": 1.555163443407156e-05, + "loss": 0.3312, + "step": 5517 + }, + { + "epoch": 0.33, + "learning_rate": 1.5550006903617796e-05, + "loss": 0.3432, + "step": 5518 + }, + { + "epoch": 0.33, + "learning_rate": 1.5548379160683794e-05, + "loss": 0.3214, + "step": 5519 + }, + { + "epoch": 0.33, + "learning_rate": 1.554675120533187e-05, + "loss": 0.3203, + "step": 5520 + }, + { + "epoch": 0.33, + "learning_rate": 1.554512303762435e-05, + "loss": 0.2984, + "step": 5521 + }, + { + "epoch": 0.33, + "learning_rate": 1.5543494657623568e-05, + "loss": 0.3111, + "step": 5522 + }, + { + "epoch": 0.33, + "learning_rate": 1.5541866065391863e-05, + "loss": 0.3513, + "step": 5523 + }, + { + "epoch": 0.33, + "learning_rate": 1.5540237260991592e-05, + "loss": 0.3509, + "step": 5524 + }, + { + "epoch": 0.33, + "learning_rate": 1.5538608244485102e-05, + "loss": 0.346, + "step": 5525 + }, + { + "epoch": 0.33, + "learning_rate": 1.553697901593477e-05, + "loss": 0.3585, + "step": 5526 + }, + { + "epoch": 0.33, + "learning_rate": 1.5535349575402963e-05, + "loss": 0.3265, + "step": 5527 + }, + { + "epoch": 0.33, + "learning_rate": 1.5533719922952072e-05, + "loss": 0.3245, + "step": 5528 + }, + { + "epoch": 0.33, + "learning_rate": 1.553209005864448e-05, + "loss": 0.326, + "step": 5529 + }, + { + "epoch": 0.33, + "learning_rate": 1.553045998254259e-05, + "loss": 0.3025, + "step": 5530 + }, + { + "epoch": 0.33, + "learning_rate": 1.5528829694708805e-05, + "loss": 0.308, + "step": 5531 + }, + { + "epoch": 0.33, + "learning_rate": 1.5527199195205543e-05, + "loss": 0.3192, + "step": 5532 + }, + { + "epoch": 0.33, + "learning_rate": 1.5525568484095228e-05, + "loss": 0.3379, + "step": 5533 + }, + { + "epoch": 0.33, + "learning_rate": 1.552393756144029e-05, + "loss": 0.3145, + "step": 5534 + }, + { + "epoch": 0.33, + "learning_rate": 1.552230642730317e-05, + "loss": 0.3638, + "step": 5535 + }, + { + "epoch": 0.33, + "learning_rate": 1.552067508174631e-05, + "loss": 0.3454, + "step": 5536 + }, + { + "epoch": 0.33, + "learning_rate": 1.551904352483217e-05, + "loss": 0.3165, + "step": 5537 + }, + { + "epoch": 0.33, + "learning_rate": 1.5517411756623214e-05, + "loss": 0.3287, + "step": 5538 + }, + { + "epoch": 0.33, + "learning_rate": 1.5515779777181915e-05, + "loss": 0.3351, + "step": 5539 + }, + { + "epoch": 0.33, + "learning_rate": 1.5514147586570746e-05, + "loss": 0.3278, + "step": 5540 + }, + { + "epoch": 0.33, + "learning_rate": 1.5512515184852203e-05, + "loss": 0.3566, + "step": 5541 + }, + { + "epoch": 0.33, + "learning_rate": 1.5510882572088778e-05, + "loss": 0.2999, + "step": 5542 + }, + { + "epoch": 0.33, + "learning_rate": 1.5509249748342975e-05, + "loss": 0.3265, + "step": 5543 + }, + { + "epoch": 0.33, + "learning_rate": 1.5507616713677304e-05, + "loss": 0.32, + "step": 5544 + }, + { + "epoch": 0.33, + "learning_rate": 1.5505983468154294e-05, + "loss": 0.3463, + "step": 5545 + }, + { + "epoch": 0.34, + "learning_rate": 1.550435001183646e-05, + "loss": 0.3301, + "step": 5546 + }, + { + "epoch": 0.34, + "learning_rate": 1.550271634478635e-05, + "loss": 0.3301, + "step": 5547 + }, + { + "epoch": 0.34, + "learning_rate": 1.55010824670665e-05, + "loss": 0.3541, + "step": 5548 + }, + { + "epoch": 0.34, + "learning_rate": 1.549944837873947e-05, + "loss": 0.3474, + "step": 5549 + }, + { + "epoch": 0.34, + "learning_rate": 1.549781407986782e-05, + "loss": 0.3341, + "step": 5550 + }, + { + "epoch": 0.34, + "learning_rate": 1.5496179570514107e-05, + "loss": 0.3418, + "step": 5551 + }, + { + "epoch": 0.34, + "learning_rate": 1.549454485074092e-05, + "loss": 0.3323, + "step": 5552 + }, + { + "epoch": 0.34, + "learning_rate": 1.549290992061084e-05, + "loss": 0.3276, + "step": 5553 + }, + { + "epoch": 0.34, + "learning_rate": 1.549127478018646e-05, + "loss": 0.3083, + "step": 5554 + }, + { + "epoch": 0.34, + "learning_rate": 1.5489639429530382e-05, + "loss": 0.3161, + "step": 5555 + }, + { + "epoch": 0.34, + "learning_rate": 1.548800386870521e-05, + "loss": 0.3311, + "step": 5556 + }, + { + "epoch": 0.34, + "learning_rate": 1.5486368097773566e-05, + "loss": 0.3189, + "step": 5557 + }, + { + "epoch": 0.34, + "learning_rate": 1.5484732116798076e-05, + "loss": 0.3131, + "step": 5558 + }, + { + "epoch": 0.34, + "learning_rate": 1.5483095925841368e-05, + "loss": 0.3864, + "step": 5559 + }, + { + "epoch": 0.34, + "learning_rate": 1.5481459524966083e-05, + "loss": 0.361, + "step": 5560 + }, + { + "epoch": 0.34, + "learning_rate": 1.5479822914234875e-05, + "loss": 0.3424, + "step": 5561 + }, + { + "epoch": 0.34, + "learning_rate": 1.5478186093710398e-05, + "loss": 0.3809, + "step": 5562 + }, + { + "epoch": 0.34, + "learning_rate": 1.5476549063455318e-05, + "loss": 0.3541, + "step": 5563 + }, + { + "epoch": 0.34, + "learning_rate": 1.5474911823532306e-05, + "loss": 0.3463, + "step": 5564 + }, + { + "epoch": 0.34, + "learning_rate": 1.5473274374004046e-05, + "loss": 0.317, + "step": 5565 + }, + { + "epoch": 0.34, + "learning_rate": 1.5471636714933227e-05, + "loss": 0.3298, + "step": 5566 + }, + { + "epoch": 0.34, + "learning_rate": 1.5469998846382545e-05, + "loss": 0.3504, + "step": 5567 + }, + { + "epoch": 0.34, + "learning_rate": 1.5468360768414706e-05, + "loss": 0.3493, + "step": 5568 + }, + { + "epoch": 0.34, + "learning_rate": 1.5466722481092423e-05, + "loss": 0.3703, + "step": 5569 + }, + { + "epoch": 0.34, + "learning_rate": 1.5465083984478418e-05, + "loss": 0.3828, + "step": 5570 + }, + { + "epoch": 0.34, + "learning_rate": 1.546344527863542e-05, + "loss": 0.3655, + "step": 5571 + }, + { + "epoch": 0.34, + "learning_rate": 1.5461806363626163e-05, + "loss": 0.3206, + "step": 5572 + }, + { + "epoch": 0.34, + "learning_rate": 1.5460167239513397e-05, + "loss": 0.358, + "step": 5573 + }, + { + "epoch": 0.34, + "learning_rate": 1.545852790635988e-05, + "loss": 0.3443, + "step": 5574 + }, + { + "epoch": 0.34, + "learning_rate": 1.5456888364228357e-05, + "loss": 0.3131, + "step": 5575 + }, + { + "epoch": 0.34, + "learning_rate": 1.5455248613181616e-05, + "loss": 0.3563, + "step": 5576 + }, + { + "epoch": 0.34, + "learning_rate": 1.545360865328242e-05, + "loss": 0.3228, + "step": 5577 + }, + { + "epoch": 0.34, + "learning_rate": 1.545196848459356e-05, + "loss": 0.3087, + "step": 5578 + }, + { + "epoch": 0.34, + "learning_rate": 1.5450328107177833e-05, + "loss": 0.3429, + "step": 5579 + }, + { + "epoch": 0.34, + "learning_rate": 1.5448687521098035e-05, + "loss": 0.3474, + "step": 5580 + }, + { + "epoch": 0.34, + "learning_rate": 1.544704672641698e-05, + "loss": 0.3373, + "step": 5581 + }, + { + "epoch": 0.34, + "learning_rate": 1.544540572319748e-05, + "loss": 0.3557, + "step": 5582 + }, + { + "epoch": 0.34, + "learning_rate": 1.544376451150236e-05, + "loss": 0.3133, + "step": 5583 + }, + { + "epoch": 0.34, + "learning_rate": 1.5442123091394457e-05, + "loss": 0.37, + "step": 5584 + }, + { + "epoch": 0.34, + "learning_rate": 1.5440481462936615e-05, + "loss": 0.2843, + "step": 5585 + }, + { + "epoch": 0.34, + "learning_rate": 1.5438839626191677e-05, + "loss": 0.3265, + "step": 5586 + }, + { + "epoch": 0.34, + "learning_rate": 1.54371975812225e-05, + "loss": 0.3281, + "step": 5587 + }, + { + "epoch": 0.34, + "learning_rate": 1.5435555328091956e-05, + "loss": 0.3468, + "step": 5588 + }, + { + "epoch": 0.34, + "learning_rate": 1.543391286686291e-05, + "loss": 0.3228, + "step": 5589 + }, + { + "epoch": 0.34, + "learning_rate": 1.543227019759825e-05, + "loss": 0.3287, + "step": 5590 + }, + { + "epoch": 0.34, + "learning_rate": 1.543062732036086e-05, + "loss": 0.3312, + "step": 5591 + }, + { + "epoch": 0.34, + "learning_rate": 1.5428984235213643e-05, + "loss": 0.3429, + "step": 5592 + }, + { + "epoch": 0.34, + "learning_rate": 1.5427340942219498e-05, + "loss": 0.3354, + "step": 5593 + }, + { + "epoch": 0.34, + "learning_rate": 1.542569744144134e-05, + "loss": 0.3145, + "step": 5594 + }, + { + "epoch": 0.34, + "learning_rate": 1.5424053732942087e-05, + "loss": 0.3717, + "step": 5595 + }, + { + "epoch": 0.34, + "learning_rate": 1.5422409816784677e-05, + "loss": 0.2983, + "step": 5596 + }, + { + "epoch": 0.34, + "learning_rate": 1.5420765693032035e-05, + "loss": 0.3544, + "step": 5597 + }, + { + "epoch": 0.34, + "learning_rate": 1.541912136174711e-05, + "loss": 0.3373, + "step": 5598 + }, + { + "epoch": 0.34, + "learning_rate": 1.5417476822992862e-05, + "loss": 0.3265, + "step": 5599 + }, + { + "epoch": 0.34, + "learning_rate": 1.541583207683224e-05, + "loss": 0.3253, + "step": 5600 + }, + { + "epoch": 0.34, + "learning_rate": 1.5414187123328224e-05, + "loss": 0.3376, + "step": 5601 + }, + { + "epoch": 0.34, + "learning_rate": 1.541254196254378e-05, + "loss": 0.315, + "step": 5602 + }, + { + "epoch": 0.34, + "learning_rate": 1.54108965945419e-05, + "loss": 0.3265, + "step": 5603 + }, + { + "epoch": 0.34, + "learning_rate": 1.5409251019385573e-05, + "loss": 0.3463, + "step": 5604 + }, + { + "epoch": 0.34, + "learning_rate": 1.5407605237137802e-05, + "loss": 0.3362, + "step": 5605 + }, + { + "epoch": 0.34, + "learning_rate": 1.540595924786159e-05, + "loss": 0.3719, + "step": 5606 + }, + { + "epoch": 0.34, + "learning_rate": 1.5404313051619956e-05, + "loss": 0.3157, + "step": 5607 + }, + { + "epoch": 0.34, + "learning_rate": 1.5402666648475927e-05, + "loss": 0.339, + "step": 5608 + }, + { + "epoch": 0.34, + "learning_rate": 1.540102003849253e-05, + "loss": 0.3644, + "step": 5609 + }, + { + "epoch": 0.34, + "learning_rate": 1.5399373221732812e-05, + "loss": 0.3348, + "step": 5610 + }, + { + "epoch": 0.34, + "learning_rate": 1.5397726198259812e-05, + "loss": 0.3532, + "step": 5611 + }, + { + "epoch": 0.34, + "learning_rate": 1.539607896813659e-05, + "loss": 0.3323, + "step": 5612 + }, + { + "epoch": 0.34, + "learning_rate": 1.5394431531426212e-05, + "loss": 0.3443, + "step": 5613 + }, + { + "epoch": 0.34, + "learning_rate": 1.5392783888191748e-05, + "loss": 0.3365, + "step": 5614 + }, + { + "epoch": 0.34, + "learning_rate": 1.5391136038496276e-05, + "loss": 0.3697, + "step": 5615 + }, + { + "epoch": 0.34, + "learning_rate": 1.5389487982402883e-05, + "loss": 0.3627, + "step": 5616 + }, + { + "epoch": 0.34, + "learning_rate": 1.5387839719974663e-05, + "loss": 0.3384, + "step": 5617 + }, + { + "epoch": 0.34, + "learning_rate": 1.5386191251274727e-05, + "loss": 0.3479, + "step": 5618 + }, + { + "epoch": 0.34, + "learning_rate": 1.5384542576366178e-05, + "loss": 0.351, + "step": 5619 + }, + { + "epoch": 0.34, + "learning_rate": 1.5382893695312137e-05, + "loss": 0.3504, + "step": 5620 + }, + { + "epoch": 0.34, + "learning_rate": 1.5381244608175732e-05, + "loss": 0.3571, + "step": 5621 + }, + { + "epoch": 0.34, + "learning_rate": 1.53795953150201e-05, + "loss": 0.3083, + "step": 5622 + }, + { + "epoch": 0.34, + "learning_rate": 1.5377945815908376e-05, + "loss": 0.341, + "step": 5623 + }, + { + "epoch": 0.34, + "learning_rate": 1.537629611090372e-05, + "loss": 0.332, + "step": 5624 + }, + { + "epoch": 0.34, + "learning_rate": 1.5374646200069286e-05, + "loss": 0.329, + "step": 5625 + }, + { + "epoch": 0.34, + "learning_rate": 1.5372996083468242e-05, + "loss": 0.3343, + "step": 5626 + }, + { + "epoch": 0.34, + "learning_rate": 1.5371345761163755e-05, + "loss": 0.3206, + "step": 5627 + }, + { + "epoch": 0.34, + "learning_rate": 1.5369695233219015e-05, + "loss": 0.3172, + "step": 5628 + }, + { + "epoch": 0.34, + "learning_rate": 1.5368044499697214e-05, + "loss": 0.3253, + "step": 5629 + }, + { + "epoch": 0.34, + "learning_rate": 1.536639356066154e-05, + "loss": 0.3365, + "step": 5630 + }, + { + "epoch": 0.34, + "learning_rate": 1.5364742416175206e-05, + "loss": 0.3274, + "step": 5631 + }, + { + "epoch": 0.34, + "learning_rate": 1.5363091066301423e-05, + "loss": 0.3072, + "step": 5632 + }, + { + "epoch": 0.34, + "learning_rate": 1.5361439511103412e-05, + "loss": 0.3253, + "step": 5633 + }, + { + "epoch": 0.34, + "learning_rate": 1.5359787750644405e-05, + "loss": 0.3647, + "step": 5634 + }, + { + "epoch": 0.34, + "learning_rate": 1.535813578498764e-05, + "loss": 0.3105, + "step": 5635 + }, + { + "epoch": 0.34, + "learning_rate": 1.5356483614196354e-05, + "loss": 0.3086, + "step": 5636 + }, + { + "epoch": 0.34, + "learning_rate": 1.5354831238333806e-05, + "loss": 0.3418, + "step": 5637 + }, + { + "epoch": 0.34, + "learning_rate": 1.535317865746326e-05, + "loss": 0.3513, + "step": 5638 + }, + { + "epoch": 0.34, + "learning_rate": 1.5351525871647977e-05, + "loss": 0.3245, + "step": 5639 + }, + { + "epoch": 0.34, + "learning_rate": 1.534987288095124e-05, + "loss": 0.3504, + "step": 5640 + }, + { + "epoch": 0.34, + "learning_rate": 1.5348219685436325e-05, + "loss": 0.3239, + "step": 5641 + }, + { + "epoch": 0.34, + "learning_rate": 1.534656628516653e-05, + "loss": 0.328, + "step": 5642 + }, + { + "epoch": 0.34, + "learning_rate": 1.534491268020516e-05, + "loss": 0.3373, + "step": 5643 + }, + { + "epoch": 0.34, + "learning_rate": 1.534325887061551e-05, + "loss": 0.3326, + "step": 5644 + }, + { + "epoch": 0.34, + "learning_rate": 1.5341604856460905e-05, + "loss": 0.3577, + "step": 5645 + }, + { + "epoch": 0.34, + "learning_rate": 1.5339950637804664e-05, + "loss": 0.3555, + "step": 5646 + }, + { + "epoch": 0.34, + "learning_rate": 1.5338296214710124e-05, + "loss": 0.3094, + "step": 5647 + }, + { + "epoch": 0.34, + "learning_rate": 1.5336641587240614e-05, + "loss": 0.3225, + "step": 5648 + }, + { + "epoch": 0.34, + "learning_rate": 1.5334986755459493e-05, + "loss": 0.3117, + "step": 5649 + }, + { + "epoch": 0.34, + "learning_rate": 1.5333331719430106e-05, + "loss": 0.3532, + "step": 5650 + }, + { + "epoch": 0.34, + "learning_rate": 1.533167647921582e-05, + "loss": 0.322, + "step": 5651 + }, + { + "epoch": 0.34, + "learning_rate": 1.5330021034880005e-05, + "loss": 0.3451, + "step": 5652 + }, + { + "epoch": 0.34, + "learning_rate": 1.5328365386486038e-05, + "loss": 0.351, + "step": 5653 + }, + { + "epoch": 0.34, + "learning_rate": 1.5326709534097304e-05, + "loss": 0.3105, + "step": 5654 + }, + { + "epoch": 0.34, + "learning_rate": 1.5325053477777202e-05, + "loss": 0.3315, + "step": 5655 + }, + { + "epoch": 0.34, + "learning_rate": 1.5323397217589125e-05, + "loss": 0.3186, + "step": 5656 + }, + { + "epoch": 0.34, + "learning_rate": 1.532174075359649e-05, + "loss": 0.3499, + "step": 5657 + }, + { + "epoch": 0.34, + "learning_rate": 1.5320084085862712e-05, + "loss": 0.3273, + "step": 5658 + }, + { + "epoch": 0.34, + "learning_rate": 1.531842721445121e-05, + "loss": 0.3069, + "step": 5659 + }, + { + "epoch": 0.34, + "learning_rate": 1.5316770139425435e-05, + "loss": 0.3513, + "step": 5660 + }, + { + "epoch": 0.34, + "learning_rate": 1.5315112860848805e-05, + "loss": 0.3438, + "step": 5661 + }, + { + "epoch": 0.34, + "learning_rate": 1.531345537878478e-05, + "loss": 0.3326, + "step": 5662 + }, + { + "epoch": 0.34, + "learning_rate": 1.5311797693296813e-05, + "loss": 0.3178, + "step": 5663 + }, + { + "epoch": 0.34, + "learning_rate": 1.531013980444837e-05, + "loss": 0.394, + "step": 5664 + }, + { + "epoch": 0.34, + "learning_rate": 1.5308481712302927e-05, + "loss": 0.3125, + "step": 5665 + }, + { + "epoch": 0.34, + "learning_rate": 1.5306823416923955e-05, + "loss": 0.353, + "step": 5666 + }, + { + "epoch": 0.34, + "learning_rate": 1.5305164918374943e-05, + "loss": 0.3198, + "step": 5667 + }, + { + "epoch": 0.34, + "learning_rate": 1.5303506216719393e-05, + "loss": 0.3251, + "step": 5668 + }, + { + "epoch": 0.34, + "learning_rate": 1.5301847312020797e-05, + "loss": 0.3541, + "step": 5669 + }, + { + "epoch": 0.34, + "learning_rate": 1.5300188204342674e-05, + "loss": 0.3203, + "step": 5670 + }, + { + "epoch": 0.34, + "learning_rate": 1.529852889374854e-05, + "loss": 0.3147, + "step": 5671 + }, + { + "epoch": 0.34, + "learning_rate": 1.529686938030192e-05, + "loss": 0.3326, + "step": 5672 + }, + { + "epoch": 0.34, + "learning_rate": 1.5295209664066357e-05, + "loss": 0.3373, + "step": 5673 + }, + { + "epoch": 0.34, + "learning_rate": 1.5293549745105372e-05, + "loss": 0.3117, + "step": 5674 + }, + { + "epoch": 0.34, + "learning_rate": 1.5291889623482536e-05, + "loss": 0.3424, + "step": 5675 + }, + { + "epoch": 0.34, + "learning_rate": 1.5290229299261396e-05, + "loss": 0.3124, + "step": 5676 + }, + { + "epoch": 0.34, + "learning_rate": 1.5288568772505517e-05, + "loss": 0.3549, + "step": 5677 + }, + { + "epoch": 0.34, + "learning_rate": 1.5286908043278474e-05, + "loss": 0.3228, + "step": 5678 + }, + { + "epoch": 0.34, + "learning_rate": 1.5285247111643848e-05, + "loss": 0.3708, + "step": 5679 + }, + { + "epoch": 0.34, + "learning_rate": 1.5283585977665224e-05, + "loss": 0.3281, + "step": 5680 + }, + { + "epoch": 0.34, + "learning_rate": 1.52819246414062e-05, + "loss": 0.3288, + "step": 5681 + }, + { + "epoch": 0.34, + "learning_rate": 1.5280263102930382e-05, + "loss": 0.3372, + "step": 5682 + }, + { + "epoch": 0.34, + "learning_rate": 1.5278601362301375e-05, + "loss": 0.3304, + "step": 5683 + }, + { + "epoch": 0.34, + "learning_rate": 1.5276939419582805e-05, + "loss": 0.3432, + "step": 5684 + }, + { + "epoch": 0.34, + "learning_rate": 1.5275277274838295e-05, + "loss": 0.3658, + "step": 5685 + }, + { + "epoch": 0.34, + "learning_rate": 1.527361492813148e-05, + "loss": 0.3652, + "step": 5686 + }, + { + "epoch": 0.34, + "learning_rate": 1.5271952379526008e-05, + "loss": 0.3259, + "step": 5687 + }, + { + "epoch": 0.34, + "learning_rate": 1.527028962908552e-05, + "loss": 0.3199, + "step": 5688 + }, + { + "epoch": 0.34, + "learning_rate": 1.526862667687368e-05, + "loss": 0.3407, + "step": 5689 + }, + { + "epoch": 0.34, + "learning_rate": 1.5266963522954154e-05, + "loss": 0.3315, + "step": 5690 + }, + { + "epoch": 0.34, + "learning_rate": 1.5265300167390608e-05, + "loss": 0.3541, + "step": 5691 + }, + { + "epoch": 0.34, + "learning_rate": 1.5263636610246733e-05, + "loss": 0.3449, + "step": 5692 + }, + { + "epoch": 0.34, + "learning_rate": 1.526197285158621e-05, + "loss": 0.368, + "step": 5693 + }, + { + "epoch": 0.34, + "learning_rate": 1.5260308891472743e-05, + "loss": 0.3479, + "step": 5694 + }, + { + "epoch": 0.34, + "learning_rate": 1.525864472997003e-05, + "loss": 0.3633, + "step": 5695 + }, + { + "epoch": 0.34, + "learning_rate": 1.5256980367141781e-05, + "loss": 0.3376, + "step": 5696 + }, + { + "epoch": 0.34, + "learning_rate": 1.525531580305172e-05, + "loss": 0.3029, + "step": 5697 + }, + { + "epoch": 0.34, + "learning_rate": 1.5253651037763575e-05, + "loss": 0.3482, + "step": 5698 + }, + { + "epoch": 0.34, + "learning_rate": 1.5251986071341081e-05, + "loss": 0.3449, + "step": 5699 + }, + { + "epoch": 0.34, + "learning_rate": 1.5250320903847977e-05, + "loss": 0.3365, + "step": 5700 + }, + { + "epoch": 0.34, + "learning_rate": 1.5248655535348017e-05, + "loss": 0.317, + "step": 5701 + }, + { + "epoch": 0.34, + "learning_rate": 1.5246989965904955e-05, + "loss": 0.3308, + "step": 5702 + }, + { + "epoch": 0.34, + "learning_rate": 1.5245324195582564e-05, + "loss": 0.3052, + "step": 5703 + }, + { + "epoch": 0.34, + "learning_rate": 1.5243658224444608e-05, + "loss": 0.3306, + "step": 5704 + }, + { + "epoch": 0.34, + "learning_rate": 1.5241992052554876e-05, + "loss": 0.3191, + "step": 5705 + }, + { + "epoch": 0.34, + "learning_rate": 1.5240325679977153e-05, + "loss": 0.3256, + "step": 5706 + }, + { + "epoch": 0.34, + "learning_rate": 1.523865910677524e-05, + "loss": 0.3474, + "step": 5707 + }, + { + "epoch": 0.34, + "learning_rate": 1.5236992333012934e-05, + "loss": 0.3357, + "step": 5708 + }, + { + "epoch": 0.34, + "learning_rate": 1.523532535875405e-05, + "loss": 0.3917, + "step": 5709 + }, + { + "epoch": 0.34, + "learning_rate": 1.5233658184062412e-05, + "loss": 0.3393, + "step": 5710 + }, + { + "epoch": 0.35, + "learning_rate": 1.5231990809001842e-05, + "loss": 0.3373, + "step": 5711 + }, + { + "epoch": 0.35, + "learning_rate": 1.5230323233636178e-05, + "loss": 0.308, + "step": 5712 + }, + { + "epoch": 0.35, + "learning_rate": 1.5228655458029255e-05, + "loss": 0.3398, + "step": 5713 + }, + { + "epoch": 0.35, + "learning_rate": 1.5226987482244937e-05, + "loss": 0.32, + "step": 5714 + }, + { + "epoch": 0.35, + "learning_rate": 1.522531930634707e-05, + "loss": 0.3013, + "step": 5715 + }, + { + "epoch": 0.35, + "learning_rate": 1.5223650930399524e-05, + "loss": 0.3265, + "step": 5716 + }, + { + "epoch": 0.35, + "learning_rate": 1.5221982354466172e-05, + "loss": 0.3292, + "step": 5717 + }, + { + "epoch": 0.35, + "learning_rate": 1.5220313578610894e-05, + "loss": 0.3557, + "step": 5718 + }, + { + "epoch": 0.35, + "learning_rate": 1.5218644602897583e-05, + "loss": 0.3379, + "step": 5719 + }, + { + "epoch": 0.35, + "learning_rate": 1.521697542739013e-05, + "loss": 0.3232, + "step": 5720 + }, + { + "epoch": 0.35, + "learning_rate": 1.5215306052152436e-05, + "loss": 0.3181, + "step": 5721 + }, + { + "epoch": 0.35, + "learning_rate": 1.5213636477248423e-05, + "loss": 0.3655, + "step": 5722 + }, + { + "epoch": 0.35, + "learning_rate": 1.5211966702742004e-05, + "loss": 0.3491, + "step": 5723 + }, + { + "epoch": 0.35, + "learning_rate": 1.5210296728697103e-05, + "loss": 0.334, + "step": 5724 + }, + { + "epoch": 0.35, + "learning_rate": 1.5208626555177657e-05, + "loss": 0.3527, + "step": 5725 + }, + { + "epoch": 0.35, + "learning_rate": 1.520695618224761e-05, + "loss": 0.3362, + "step": 5726 + }, + { + "epoch": 0.35, + "learning_rate": 1.5205285609970913e-05, + "loss": 0.353, + "step": 5727 + }, + { + "epoch": 0.35, + "learning_rate": 1.5203614838411515e-05, + "loss": 0.3407, + "step": 5728 + }, + { + "epoch": 0.35, + "learning_rate": 1.520194386763339e-05, + "loss": 0.3393, + "step": 5729 + }, + { + "epoch": 0.35, + "learning_rate": 1.5200272697700506e-05, + "loss": 0.329, + "step": 5730 + }, + { + "epoch": 0.35, + "learning_rate": 1.5198601328676846e-05, + "loss": 0.3499, + "step": 5731 + }, + { + "epoch": 0.35, + "learning_rate": 1.5196929760626394e-05, + "loss": 0.3518, + "step": 5732 + }, + { + "epoch": 0.35, + "learning_rate": 1.5195257993613145e-05, + "loss": 0.3009, + "step": 5733 + }, + { + "epoch": 0.35, + "learning_rate": 1.5193586027701108e-05, + "loss": 0.3248, + "step": 5734 + }, + { + "epoch": 0.35, + "learning_rate": 1.519191386295429e-05, + "loss": 0.3281, + "step": 5735 + }, + { + "epoch": 0.35, + "learning_rate": 1.5190241499436709e-05, + "loss": 0.3312, + "step": 5736 + }, + { + "epoch": 0.35, + "learning_rate": 1.5188568937212391e-05, + "loss": 0.2902, + "step": 5737 + }, + { + "epoch": 0.35, + "learning_rate": 1.5186896176345369e-05, + "loss": 0.3287, + "step": 5738 + }, + { + "epoch": 0.35, + "learning_rate": 1.5185223216899687e-05, + "loss": 0.3273, + "step": 5739 + }, + { + "epoch": 0.35, + "learning_rate": 1.5183550058939391e-05, + "loss": 0.3368, + "step": 5740 + }, + { + "epoch": 0.35, + "learning_rate": 1.5181876702528537e-05, + "loss": 0.3224, + "step": 5741 + }, + { + "epoch": 0.35, + "learning_rate": 1.5180203147731191e-05, + "loss": 0.3119, + "step": 5742 + }, + { + "epoch": 0.35, + "learning_rate": 1.5178529394611425e-05, + "loss": 0.3223, + "step": 5743 + }, + { + "epoch": 0.35, + "learning_rate": 1.5176855443233316e-05, + "loss": 0.3415, + "step": 5744 + }, + { + "epoch": 0.35, + "learning_rate": 1.5175181293660952e-05, + "loss": 0.3396, + "step": 5745 + }, + { + "epoch": 0.35, + "learning_rate": 1.5173506945958423e-05, + "loss": 0.3284, + "step": 5746 + }, + { + "epoch": 0.35, + "learning_rate": 1.5171832400189837e-05, + "loss": 0.3435, + "step": 5747 + }, + { + "epoch": 0.35, + "learning_rate": 1.51701576564193e-05, + "loss": 0.3298, + "step": 5748 + }, + { + "epoch": 0.35, + "learning_rate": 1.5168482714710932e-05, + "loss": 0.3119, + "step": 5749 + }, + { + "epoch": 0.35, + "learning_rate": 1.5166807575128855e-05, + "loss": 0.3443, + "step": 5750 + }, + { + "epoch": 0.35, + "learning_rate": 1.5165132237737199e-05, + "loss": 0.303, + "step": 5751 + }, + { + "epoch": 0.35, + "learning_rate": 1.5163456702600107e-05, + "loss": 0.3426, + "step": 5752 + }, + { + "epoch": 0.35, + "learning_rate": 1.5161780969781728e-05, + "loss": 0.3136, + "step": 5753 + }, + { + "epoch": 0.35, + "learning_rate": 1.5160105039346213e-05, + "loss": 0.3429, + "step": 5754 + }, + { + "epoch": 0.35, + "learning_rate": 1.5158428911357726e-05, + "loss": 0.308, + "step": 5755 + }, + { + "epoch": 0.35, + "learning_rate": 1.5156752585880439e-05, + "loss": 0.3407, + "step": 5756 + }, + { + "epoch": 0.35, + "learning_rate": 1.5155076062978527e-05, + "loss": 0.3195, + "step": 5757 + }, + { + "epoch": 0.35, + "learning_rate": 1.5153399342716175e-05, + "loss": 0.3318, + "step": 5758 + }, + { + "epoch": 0.35, + "learning_rate": 1.5151722425157576e-05, + "loss": 0.3588, + "step": 5759 + }, + { + "epoch": 0.35, + "learning_rate": 1.5150045310366931e-05, + "loss": 0.3124, + "step": 5760 + }, + { + "epoch": 0.35, + "learning_rate": 1.514836799840845e-05, + "loss": 0.3265, + "step": 5761 + }, + { + "epoch": 0.35, + "learning_rate": 1.5146690489346347e-05, + "loss": 0.3108, + "step": 5762 + }, + { + "epoch": 0.35, + "learning_rate": 1.5145012783244842e-05, + "loss": 0.353, + "step": 5763 + }, + { + "epoch": 0.35, + "learning_rate": 1.5143334880168166e-05, + "loss": 0.3602, + "step": 5764 + }, + { + "epoch": 0.35, + "learning_rate": 1.5141656780180558e-05, + "loss": 0.3105, + "step": 5765 + }, + { + "epoch": 0.35, + "learning_rate": 1.5139978483346265e-05, + "loss": 0.3246, + "step": 5766 + }, + { + "epoch": 0.35, + "learning_rate": 1.5138299989729541e-05, + "loss": 0.3407, + "step": 5767 + }, + { + "epoch": 0.35, + "learning_rate": 1.513662129939464e-05, + "loss": 0.3426, + "step": 5768 + }, + { + "epoch": 0.35, + "learning_rate": 1.5134942412405837e-05, + "loss": 0.3315, + "step": 5769 + }, + { + "epoch": 0.35, + "learning_rate": 1.5133263328827408e-05, + "loss": 0.3344, + "step": 5770 + }, + { + "epoch": 0.35, + "learning_rate": 1.5131584048723631e-05, + "loss": 0.3619, + "step": 5771 + }, + { + "epoch": 0.35, + "learning_rate": 1.5129904572158797e-05, + "loss": 0.3761, + "step": 5772 + }, + { + "epoch": 0.35, + "learning_rate": 1.5128224899197208e-05, + "loss": 0.3527, + "step": 5773 + }, + { + "epoch": 0.35, + "learning_rate": 1.512654502990317e-05, + "loss": 0.368, + "step": 5774 + }, + { + "epoch": 0.35, + "learning_rate": 1.5124864964340993e-05, + "loss": 0.3239, + "step": 5775 + }, + { + "epoch": 0.35, + "learning_rate": 1.5123184702574997e-05, + "loss": 0.3036, + "step": 5776 + }, + { + "epoch": 0.35, + "learning_rate": 1.5121504244669516e-05, + "loss": 0.3424, + "step": 5777 + }, + { + "epoch": 0.35, + "learning_rate": 1.5119823590688882e-05, + "loss": 0.3365, + "step": 5778 + }, + { + "epoch": 0.35, + "learning_rate": 1.5118142740697439e-05, + "loss": 0.351, + "step": 5779 + }, + { + "epoch": 0.35, + "learning_rate": 1.5116461694759537e-05, + "loss": 0.3331, + "step": 5780 + }, + { + "epoch": 0.35, + "learning_rate": 1.511478045293953e-05, + "loss": 0.3136, + "step": 5781 + }, + { + "epoch": 0.35, + "learning_rate": 1.5113099015301792e-05, + "loss": 0.3248, + "step": 5782 + }, + { + "epoch": 0.35, + "learning_rate": 1.5111417381910694e-05, + "loss": 0.3418, + "step": 5783 + }, + { + "epoch": 0.35, + "learning_rate": 1.5109735552830613e-05, + "loss": 0.3535, + "step": 5784 + }, + { + "epoch": 0.35, + "learning_rate": 1.5108053528125944e-05, + "loss": 0.339, + "step": 5785 + }, + { + "epoch": 0.35, + "learning_rate": 1.5106371307861074e-05, + "loss": 0.3465, + "step": 5786 + }, + { + "epoch": 0.35, + "learning_rate": 1.5104688892100413e-05, + "loss": 0.3139, + "step": 5787 + }, + { + "epoch": 0.35, + "learning_rate": 1.5103006280908367e-05, + "loss": 0.3418, + "step": 5788 + }, + { + "epoch": 0.35, + "learning_rate": 1.510132347434936e-05, + "loss": 0.317, + "step": 5789 + }, + { + "epoch": 0.35, + "learning_rate": 1.5099640472487815e-05, + "loss": 0.3373, + "step": 5790 + }, + { + "epoch": 0.35, + "learning_rate": 1.5097957275388162e-05, + "loss": 0.3161, + "step": 5791 + }, + { + "epoch": 0.35, + "learning_rate": 1.5096273883114848e-05, + "loss": 0.3304, + "step": 5792 + }, + { + "epoch": 0.35, + "learning_rate": 1.5094590295732315e-05, + "loss": 0.3316, + "step": 5793 + }, + { + "epoch": 0.35, + "learning_rate": 1.5092906513305021e-05, + "loss": 0.3485, + "step": 5794 + }, + { + "epoch": 0.35, + "learning_rate": 1.509122253589743e-05, + "loss": 0.334, + "step": 5795 + }, + { + "epoch": 0.35, + "learning_rate": 1.5089538363574013e-05, + "loss": 0.3295, + "step": 5796 + }, + { + "epoch": 0.35, + "learning_rate": 1.5087853996399245e-05, + "loss": 0.3177, + "step": 5797 + }, + { + "epoch": 0.35, + "learning_rate": 1.5086169434437611e-05, + "loss": 0.3348, + "step": 5798 + }, + { + "epoch": 0.35, + "learning_rate": 1.508448467775361e-05, + "loss": 0.3418, + "step": 5799 + }, + { + "epoch": 0.35, + "learning_rate": 1.5082799726411738e-05, + "loss": 0.2858, + "step": 5800 + }, + { + "epoch": 0.35, + "learning_rate": 1.5081114580476502e-05, + "loss": 0.3072, + "step": 5801 + }, + { + "epoch": 0.35, + "learning_rate": 1.507942924001242e-05, + "loss": 0.3557, + "step": 5802 + }, + { + "epoch": 0.35, + "learning_rate": 1.5077743705084012e-05, + "loss": 0.335, + "step": 5803 + }, + { + "epoch": 0.35, + "learning_rate": 1.5076057975755811e-05, + "loss": 0.3563, + "step": 5804 + }, + { + "epoch": 0.35, + "learning_rate": 1.5074372052092353e-05, + "loss": 0.3306, + "step": 5805 + }, + { + "epoch": 0.35, + "learning_rate": 1.507268593415818e-05, + "loss": 0.3541, + "step": 5806 + }, + { + "epoch": 0.35, + "learning_rate": 1.5070999622017853e-05, + "loss": 0.3379, + "step": 5807 + }, + { + "epoch": 0.35, + "learning_rate": 1.5069313115735921e-05, + "loss": 0.3446, + "step": 5808 + }, + { + "epoch": 0.35, + "learning_rate": 1.5067626415376959e-05, + "loss": 0.317, + "step": 5809 + }, + { + "epoch": 0.35, + "learning_rate": 1.5065939521005542e-05, + "loss": 0.3387, + "step": 5810 + }, + { + "epoch": 0.35, + "learning_rate": 1.5064252432686247e-05, + "loss": 0.3719, + "step": 5811 + }, + { + "epoch": 0.35, + "learning_rate": 1.5062565150483668e-05, + "loss": 0.3828, + "step": 5812 + }, + { + "epoch": 0.35, + "learning_rate": 1.5060877674462399e-05, + "loss": 0.3329, + "step": 5813 + }, + { + "epoch": 0.35, + "learning_rate": 1.5059190004687049e-05, + "loss": 0.3633, + "step": 5814 + }, + { + "epoch": 0.35, + "learning_rate": 1.5057502141222227e-05, + "loss": 0.3633, + "step": 5815 + }, + { + "epoch": 0.35, + "learning_rate": 1.505581408413255e-05, + "loss": 0.3334, + "step": 5816 + }, + { + "epoch": 0.35, + "learning_rate": 1.5054125833482647e-05, + "loss": 0.3149, + "step": 5817 + }, + { + "epoch": 0.35, + "learning_rate": 1.5052437389337154e-05, + "loss": 0.3329, + "step": 5818 + }, + { + "epoch": 0.35, + "learning_rate": 1.5050748751760709e-05, + "loss": 0.3387, + "step": 5819 + }, + { + "epoch": 0.35, + "learning_rate": 1.5049059920817963e-05, + "loss": 0.3538, + "step": 5820 + }, + { + "epoch": 0.35, + "learning_rate": 1.5047370896573572e-05, + "loss": 0.3287, + "step": 5821 + }, + { + "epoch": 0.35, + "learning_rate": 1.5045681679092201e-05, + "loss": 0.346, + "step": 5822 + }, + { + "epoch": 0.35, + "learning_rate": 1.5043992268438519e-05, + "loss": 0.3059, + "step": 5823 + }, + { + "epoch": 0.35, + "learning_rate": 1.5042302664677207e-05, + "loss": 0.3493, + "step": 5824 + }, + { + "epoch": 0.35, + "learning_rate": 1.5040612867872945e-05, + "loss": 0.361, + "step": 5825 + }, + { + "epoch": 0.35, + "learning_rate": 1.5038922878090435e-05, + "loss": 0.3454, + "step": 5826 + }, + { + "epoch": 0.35, + "learning_rate": 1.5037232695394373e-05, + "loss": 0.3976, + "step": 5827 + }, + { + "epoch": 0.35, + "learning_rate": 1.5035542319849468e-05, + "loss": 0.353, + "step": 5828 + }, + { + "epoch": 0.35, + "learning_rate": 1.5033851751520429e-05, + "loss": 0.3245, + "step": 5829 + }, + { + "epoch": 0.35, + "learning_rate": 1.5032160990471996e-05, + "loss": 0.3504, + "step": 5830 + }, + { + "epoch": 0.35, + "learning_rate": 1.5030470036768878e-05, + "loss": 0.351, + "step": 5831 + }, + { + "epoch": 0.35, + "learning_rate": 1.5028778890475828e-05, + "loss": 0.3334, + "step": 5832 + }, + { + "epoch": 0.35, + "learning_rate": 1.5027087551657581e-05, + "loss": 0.3348, + "step": 5833 + }, + { + "epoch": 0.35, + "learning_rate": 1.5025396020378899e-05, + "loss": 0.353, + "step": 5834 + }, + { + "epoch": 0.35, + "learning_rate": 1.5023704296704538e-05, + "loss": 0.3569, + "step": 5835 + }, + { + "epoch": 0.35, + "learning_rate": 1.502201238069926e-05, + "loss": 0.329, + "step": 5836 + }, + { + "epoch": 0.35, + "learning_rate": 1.5020320272427845e-05, + "loss": 0.3591, + "step": 5837 + }, + { + "epoch": 0.35, + "learning_rate": 1.5018627971955072e-05, + "loss": 0.3198, + "step": 5838 + }, + { + "epoch": 0.35, + "learning_rate": 1.5016935479345734e-05, + "loss": 0.3281, + "step": 5839 + }, + { + "epoch": 0.35, + "learning_rate": 1.5015242794664623e-05, + "loss": 0.3186, + "step": 5840 + }, + { + "epoch": 0.35, + "learning_rate": 1.5013549917976544e-05, + "loss": 0.3733, + "step": 5841 + }, + { + "epoch": 0.35, + "learning_rate": 1.5011856849346314e-05, + "loss": 0.3789, + "step": 5842 + }, + { + "epoch": 0.35, + "learning_rate": 1.5010163588838743e-05, + "loss": 0.3209, + "step": 5843 + }, + { + "epoch": 0.35, + "learning_rate": 1.5008470136518658e-05, + "loss": 0.3387, + "step": 5844 + }, + { + "epoch": 0.35, + "learning_rate": 1.50067764924509e-05, + "loss": 0.3248, + "step": 5845 + }, + { + "epoch": 0.35, + "learning_rate": 1.5005082656700301e-05, + "loss": 0.3265, + "step": 5846 + }, + { + "epoch": 0.35, + "learning_rate": 1.5003388629331717e-05, + "loss": 0.3117, + "step": 5847 + }, + { + "epoch": 0.35, + "learning_rate": 1.5001694410409995e-05, + "loss": 0.3379, + "step": 5848 + }, + { + "epoch": 0.35, + "learning_rate": 1.5000000000000002e-05, + "loss": 0.3354, + "step": 5849 + }, + { + "epoch": 0.35, + "learning_rate": 1.4998305398166607e-05, + "loss": 0.3563, + "step": 5850 + }, + { + "epoch": 0.35, + "learning_rate": 1.4996610604974689e-05, + "loss": 0.3594, + "step": 5851 + }, + { + "epoch": 0.35, + "learning_rate": 1.4994915620489132e-05, + "loss": 0.3239, + "step": 5852 + }, + { + "epoch": 0.35, + "learning_rate": 1.4993220444774822e-05, + "loss": 0.3343, + "step": 5853 + }, + { + "epoch": 0.35, + "learning_rate": 1.4991525077896669e-05, + "loss": 0.3125, + "step": 5854 + }, + { + "epoch": 0.35, + "learning_rate": 1.4989829519919575e-05, + "loss": 0.3083, + "step": 5855 + }, + { + "epoch": 0.35, + "learning_rate": 1.4988133770908448e-05, + "loss": 0.3362, + "step": 5856 + }, + { + "epoch": 0.35, + "learning_rate": 1.4986437830928217e-05, + "loss": 0.3359, + "step": 5857 + }, + { + "epoch": 0.35, + "learning_rate": 1.4984741700043808e-05, + "loss": 0.3175, + "step": 5858 + }, + { + "epoch": 0.35, + "learning_rate": 1.4983045378320156e-05, + "loss": 0.3557, + "step": 5859 + }, + { + "epoch": 0.35, + "learning_rate": 1.4981348865822204e-05, + "loss": 0.3122, + "step": 5860 + }, + { + "epoch": 0.35, + "learning_rate": 1.4979652162614903e-05, + "loss": 0.3597, + "step": 5861 + }, + { + "epoch": 0.35, + "learning_rate": 1.4977955268763213e-05, + "loss": 0.351, + "step": 5862 + }, + { + "epoch": 0.35, + "learning_rate": 1.4976258184332094e-05, + "loss": 0.3504, + "step": 5863 + }, + { + "epoch": 0.35, + "learning_rate": 1.4974560909386523e-05, + "loss": 0.3139, + "step": 5864 + }, + { + "epoch": 0.35, + "learning_rate": 1.4972863443991478e-05, + "loss": 0.3089, + "step": 5865 + }, + { + "epoch": 0.35, + "learning_rate": 1.4971165788211946e-05, + "loss": 0.3415, + "step": 5866 + }, + { + "epoch": 0.35, + "learning_rate": 1.496946794211292e-05, + "loss": 0.3661, + "step": 5867 + }, + { + "epoch": 0.35, + "learning_rate": 1.4967769905759401e-05, + "loss": 0.3306, + "step": 5868 + }, + { + "epoch": 0.35, + "learning_rate": 1.4966071679216403e-05, + "loss": 0.3359, + "step": 5869 + }, + { + "epoch": 0.35, + "learning_rate": 1.4964373262548938e-05, + "loss": 0.351, + "step": 5870 + }, + { + "epoch": 0.35, + "learning_rate": 1.4962674655822028e-05, + "loss": 0.3412, + "step": 5871 + }, + { + "epoch": 0.35, + "learning_rate": 1.4960975859100703e-05, + "loss": 0.3538, + "step": 5872 + }, + { + "epoch": 0.35, + "learning_rate": 1.4959276872450006e-05, + "loss": 0.3404, + "step": 5873 + }, + { + "epoch": 0.35, + "learning_rate": 1.4957577695934978e-05, + "loss": 0.3354, + "step": 5874 + }, + { + "epoch": 0.35, + "learning_rate": 1.4955878329620674e-05, + "loss": 0.327, + "step": 5875 + }, + { + "epoch": 0.35, + "learning_rate": 1.4954178773572153e-05, + "loss": 0.3446, + "step": 5876 + }, + { + "epoch": 0.36, + "learning_rate": 1.4952479027854479e-05, + "loss": 0.365, + "step": 5877 + }, + { + "epoch": 0.36, + "learning_rate": 1.495077909253273e-05, + "loss": 0.3401, + "step": 5878 + }, + { + "epoch": 0.36, + "learning_rate": 1.4949078967671986e-05, + "loss": 0.3432, + "step": 5879 + }, + { + "epoch": 0.36, + "learning_rate": 1.4947378653337337e-05, + "loss": 0.339, + "step": 5880 + }, + { + "epoch": 0.36, + "learning_rate": 1.4945678149593877e-05, + "loss": 0.332, + "step": 5881 + }, + { + "epoch": 0.36, + "learning_rate": 1.4943977456506711e-05, + "loss": 0.3295, + "step": 5882 + }, + { + "epoch": 0.36, + "learning_rate": 1.4942276574140946e-05, + "loss": 0.3401, + "step": 5883 + }, + { + "epoch": 0.36, + "learning_rate": 1.4940575502561706e-05, + "loss": 0.3058, + "step": 5884 + }, + { + "epoch": 0.36, + "learning_rate": 1.493887424183411e-05, + "loss": 0.3443, + "step": 5885 + }, + { + "epoch": 0.36, + "learning_rate": 1.4937172792023292e-05, + "loss": 0.3269, + "step": 5886 + }, + { + "epoch": 0.36, + "learning_rate": 1.4935471153194393e-05, + "loss": 0.3504, + "step": 5887 + }, + { + "epoch": 0.36, + "learning_rate": 1.4933769325412555e-05, + "loss": 0.3132, + "step": 5888 + }, + { + "epoch": 0.36, + "learning_rate": 1.4932067308742939e-05, + "loss": 0.3387, + "step": 5889 + }, + { + "epoch": 0.36, + "learning_rate": 1.4930365103250703e-05, + "loss": 0.327, + "step": 5890 + }, + { + "epoch": 0.36, + "learning_rate": 1.4928662709001013e-05, + "loss": 0.3599, + "step": 5891 + }, + { + "epoch": 0.36, + "learning_rate": 1.4926960126059052e-05, + "loss": 0.341, + "step": 5892 + }, + { + "epoch": 0.36, + "learning_rate": 1.4925257354489992e-05, + "loss": 0.3076, + "step": 5893 + }, + { + "epoch": 0.36, + "learning_rate": 1.4923554394359033e-05, + "loss": 0.3594, + "step": 5894 + }, + { + "epoch": 0.36, + "learning_rate": 1.4921851245731364e-05, + "loss": 0.3506, + "step": 5895 + }, + { + "epoch": 0.36, + "learning_rate": 1.4920147908672195e-05, + "loss": 0.3426, + "step": 5896 + }, + { + "epoch": 0.36, + "learning_rate": 1.4918444383246738e-05, + "loss": 0.3228, + "step": 5897 + }, + { + "epoch": 0.36, + "learning_rate": 1.4916740669520209e-05, + "loss": 0.3412, + "step": 5898 + }, + { + "epoch": 0.36, + "learning_rate": 1.4915036767557835e-05, + "loss": 0.3198, + "step": 5899 + }, + { + "epoch": 0.36, + "learning_rate": 1.4913332677424848e-05, + "loss": 0.3563, + "step": 5900 + }, + { + "epoch": 0.36, + "learning_rate": 1.4911628399186494e-05, + "loss": 0.3072, + "step": 5901 + }, + { + "epoch": 0.36, + "learning_rate": 1.4909923932908014e-05, + "loss": 0.3485, + "step": 5902 + }, + { + "epoch": 0.36, + "learning_rate": 1.4908219278654666e-05, + "loss": 0.3504, + "step": 5903 + }, + { + "epoch": 0.36, + "learning_rate": 1.4906514436491713e-05, + "loss": 0.322, + "step": 5904 + }, + { + "epoch": 0.36, + "learning_rate": 1.4904809406484421e-05, + "loss": 0.3251, + "step": 5905 + }, + { + "epoch": 0.36, + "learning_rate": 1.4903104188698072e-05, + "loss": 0.3577, + "step": 5906 + }, + { + "epoch": 0.36, + "learning_rate": 1.4901398783197944e-05, + "loss": 0.3585, + "step": 5907 + }, + { + "epoch": 0.36, + "learning_rate": 1.4899693190049331e-05, + "loss": 0.3298, + "step": 5908 + }, + { + "epoch": 0.36, + "learning_rate": 1.4897987409317531e-05, + "loss": 0.3396, + "step": 5909 + }, + { + "epoch": 0.36, + "learning_rate": 1.489628144106785e-05, + "loss": 0.3199, + "step": 5910 + }, + { + "epoch": 0.36, + "learning_rate": 1.4894575285365598e-05, + "loss": 0.3242, + "step": 5911 + }, + { + "epoch": 0.36, + "learning_rate": 1.4892868942276097e-05, + "loss": 0.3318, + "step": 5912 + }, + { + "epoch": 0.36, + "learning_rate": 1.489116241186467e-05, + "loss": 0.3065, + "step": 5913 + }, + { + "epoch": 0.36, + "learning_rate": 1.4889455694196656e-05, + "loss": 0.3234, + "step": 5914 + }, + { + "epoch": 0.36, + "learning_rate": 1.4887748789337396e-05, + "loss": 0.3371, + "step": 5915 + }, + { + "epoch": 0.36, + "learning_rate": 1.4886041697352234e-05, + "loss": 0.3147, + "step": 5916 + }, + { + "epoch": 0.36, + "learning_rate": 1.4884334418306525e-05, + "loss": 0.3304, + "step": 5917 + }, + { + "epoch": 0.36, + "learning_rate": 1.4882626952265635e-05, + "loss": 0.3401, + "step": 5918 + }, + { + "epoch": 0.36, + "learning_rate": 1.4880919299294935e-05, + "loss": 0.3468, + "step": 5919 + }, + { + "epoch": 0.36, + "learning_rate": 1.4879211459459802e-05, + "loss": 0.3463, + "step": 5920 + }, + { + "epoch": 0.36, + "learning_rate": 1.4877503432825612e-05, + "loss": 0.3479, + "step": 5921 + }, + { + "epoch": 0.36, + "learning_rate": 1.4875795219457767e-05, + "loss": 0.3524, + "step": 5922 + }, + { + "epoch": 0.36, + "learning_rate": 1.4874086819421657e-05, + "loss": 0.3425, + "step": 5923 + }, + { + "epoch": 0.36, + "learning_rate": 1.4872378232782696e-05, + "loss": 0.3368, + "step": 5924 + }, + { + "epoch": 0.36, + "learning_rate": 1.4870669459606291e-05, + "loss": 0.3178, + "step": 5925 + }, + { + "epoch": 0.36, + "learning_rate": 1.4868960499957858e-05, + "loss": 0.3438, + "step": 5926 + }, + { + "epoch": 0.36, + "learning_rate": 1.4867251353902834e-05, + "loss": 0.3326, + "step": 5927 + }, + { + "epoch": 0.36, + "learning_rate": 1.4865542021506644e-05, + "loss": 0.3371, + "step": 5928 + }, + { + "epoch": 0.36, + "learning_rate": 1.4863832502834737e-05, + "loss": 0.3329, + "step": 5929 + }, + { + "epoch": 0.36, + "learning_rate": 1.4862122797952556e-05, + "loss": 0.3479, + "step": 5930 + }, + { + "epoch": 0.36, + "learning_rate": 1.4860412906925556e-05, + "loss": 0.3343, + "step": 5931 + }, + { + "epoch": 0.36, + "learning_rate": 1.4858702829819204e-05, + "loss": 0.3304, + "step": 5932 + }, + { + "epoch": 0.36, + "learning_rate": 1.4856992566698967e-05, + "loss": 0.3301, + "step": 5933 + }, + { + "epoch": 0.36, + "learning_rate": 1.4855282117630322e-05, + "loss": 0.3057, + "step": 5934 + }, + { + "epoch": 0.36, + "learning_rate": 1.4853571482678752e-05, + "loss": 0.3158, + "step": 5935 + }, + { + "epoch": 0.36, + "learning_rate": 1.4851860661909751e-05, + "loss": 0.3287, + "step": 5936 + }, + { + "epoch": 0.36, + "learning_rate": 1.4850149655388817e-05, + "loss": 0.3412, + "step": 5937 + }, + { + "epoch": 0.36, + "learning_rate": 1.4848438463181452e-05, + "loss": 0.3315, + "step": 5938 + }, + { + "epoch": 0.36, + "learning_rate": 1.4846727085353169e-05, + "loss": 0.3209, + "step": 5939 + }, + { + "epoch": 0.36, + "learning_rate": 1.4845015521969492e-05, + "loss": 0.3429, + "step": 5940 + }, + { + "epoch": 0.36, + "learning_rate": 1.4843303773095943e-05, + "loss": 0.3499, + "step": 5941 + }, + { + "epoch": 0.36, + "learning_rate": 1.4841591838798058e-05, + "loss": 0.3443, + "step": 5942 + }, + { + "epoch": 0.36, + "learning_rate": 1.4839879719141375e-05, + "loss": 0.3172, + "step": 5943 + }, + { + "epoch": 0.36, + "learning_rate": 1.4838167414191447e-05, + "loss": 0.3239, + "step": 5944 + }, + { + "epoch": 0.36, + "learning_rate": 1.4836454924013826e-05, + "loss": 0.3262, + "step": 5945 + }, + { + "epoch": 0.36, + "learning_rate": 1.4834742248674073e-05, + "loss": 0.3256, + "step": 5946 + }, + { + "epoch": 0.36, + "learning_rate": 1.483302938823776e-05, + "loss": 0.3672, + "step": 5947 + }, + { + "epoch": 0.36, + "learning_rate": 1.4831316342770456e-05, + "loss": 0.3228, + "step": 5948 + }, + { + "epoch": 0.36, + "learning_rate": 1.4829603112337761e-05, + "loss": 0.3451, + "step": 5949 + }, + { + "epoch": 0.36, + "learning_rate": 1.4827889697005246e-05, + "loss": 0.3633, + "step": 5950 + }, + { + "epoch": 0.36, + "learning_rate": 1.482617609683852e-05, + "loss": 0.3497, + "step": 5951 + }, + { + "epoch": 0.36, + "learning_rate": 1.4824462311903187e-05, + "loss": 0.3479, + "step": 5952 + }, + { + "epoch": 0.36, + "learning_rate": 1.4822748342264853e-05, + "loss": 0.3421, + "step": 5953 + }, + { + "epoch": 0.36, + "learning_rate": 1.4821034187989143e-05, + "loss": 0.3412, + "step": 5954 + }, + { + "epoch": 0.36, + "learning_rate": 1.481931984914168e-05, + "loss": 0.2913, + "step": 5955 + }, + { + "epoch": 0.36, + "learning_rate": 1.4817605325788096e-05, + "loss": 0.3163, + "step": 5956 + }, + { + "epoch": 0.36, + "learning_rate": 1.4815890617994035e-05, + "loss": 0.3312, + "step": 5957 + }, + { + "epoch": 0.36, + "learning_rate": 1.4814175725825137e-05, + "loss": 0.3516, + "step": 5958 + }, + { + "epoch": 0.36, + "learning_rate": 1.4812460649347063e-05, + "loss": 0.3477, + "step": 5959 + }, + { + "epoch": 0.36, + "learning_rate": 1.4810745388625472e-05, + "loss": 0.341, + "step": 5960 + }, + { + "epoch": 0.36, + "learning_rate": 1.480902994372603e-05, + "loss": 0.3479, + "step": 5961 + }, + { + "epoch": 0.36, + "learning_rate": 1.4807314314714415e-05, + "loss": 0.3034, + "step": 5962 + }, + { + "epoch": 0.36, + "learning_rate": 1.4805598501656309e-05, + "loss": 0.3298, + "step": 5963 + }, + { + "epoch": 0.36, + "learning_rate": 1.48038825046174e-05, + "loss": 0.3295, + "step": 5964 + }, + { + "epoch": 0.36, + "learning_rate": 1.4802166323663385e-05, + "loss": 0.377, + "step": 5965 + }, + { + "epoch": 0.36, + "learning_rate": 1.4800449958859969e-05, + "loss": 0.3541, + "step": 5966 + }, + { + "epoch": 0.36, + "learning_rate": 1.4798733410272859e-05, + "loss": 0.32, + "step": 5967 + }, + { + "epoch": 0.36, + "learning_rate": 1.4797016677967773e-05, + "loss": 0.3362, + "step": 5968 + }, + { + "epoch": 0.36, + "learning_rate": 1.479529976201044e-05, + "loss": 0.3033, + "step": 5969 + }, + { + "epoch": 0.36, + "learning_rate": 1.4793582662466587e-05, + "loss": 0.3691, + "step": 5970 + }, + { + "epoch": 0.36, + "learning_rate": 1.4791865379401953e-05, + "loss": 0.3079, + "step": 5971 + }, + { + "epoch": 0.36, + "learning_rate": 1.4790147912882286e-05, + "loss": 0.3555, + "step": 5972 + }, + { + "epoch": 0.36, + "learning_rate": 1.4788430262973336e-05, + "loss": 0.3284, + "step": 5973 + }, + { + "epoch": 0.36, + "learning_rate": 1.4786712429740866e-05, + "loss": 0.315, + "step": 5974 + }, + { + "epoch": 0.36, + "learning_rate": 1.4784994413250641e-05, + "loss": 0.3507, + "step": 5975 + }, + { + "epoch": 0.36, + "learning_rate": 1.4783276213568433e-05, + "loss": 0.3287, + "step": 5976 + }, + { + "epoch": 0.36, + "learning_rate": 1.4781557830760026e-05, + "loss": 0.3398, + "step": 5977 + }, + { + "epoch": 0.36, + "learning_rate": 1.4779839264891206e-05, + "loss": 0.3136, + "step": 5978 + }, + { + "epoch": 0.36, + "learning_rate": 1.4778120516027767e-05, + "loss": 0.3549, + "step": 5979 + }, + { + "epoch": 0.36, + "learning_rate": 1.4776401584235513e-05, + "loss": 0.3504, + "step": 5980 + }, + { + "epoch": 0.36, + "learning_rate": 1.4774682469580248e-05, + "loss": 0.3158, + "step": 5981 + }, + { + "epoch": 0.36, + "learning_rate": 1.4772963172127794e-05, + "loss": 0.3225, + "step": 5982 + }, + { + "epoch": 0.36, + "learning_rate": 1.477124369194397e-05, + "loss": 0.3168, + "step": 5983 + }, + { + "epoch": 0.36, + "learning_rate": 1.4769524029094608e-05, + "loss": 0.3379, + "step": 5984 + }, + { + "epoch": 0.36, + "learning_rate": 1.476780418364554e-05, + "loss": 0.3393, + "step": 5985 + }, + { + "epoch": 0.36, + "learning_rate": 1.4766084155662616e-05, + "loss": 0.3251, + "step": 5986 + }, + { + "epoch": 0.36, + "learning_rate": 1.4764363945211683e-05, + "loss": 0.3309, + "step": 5987 + }, + { + "epoch": 0.36, + "learning_rate": 1.4762643552358597e-05, + "loss": 0.3803, + "step": 5988 + }, + { + "epoch": 0.36, + "learning_rate": 1.4760922977169227e-05, + "loss": 0.3089, + "step": 5989 + }, + { + "epoch": 0.36, + "learning_rate": 1.4759202219709445e-05, + "loss": 0.3284, + "step": 5990 + }, + { + "epoch": 0.36, + "learning_rate": 1.4757481280045124e-05, + "loss": 0.3062, + "step": 5991 + }, + { + "epoch": 0.36, + "learning_rate": 1.4755760158242156e-05, + "loss": 0.3248, + "step": 5992 + }, + { + "epoch": 0.36, + "learning_rate": 1.4754038854366427e-05, + "loss": 0.3624, + "step": 5993 + }, + { + "epoch": 0.36, + "learning_rate": 1.4752317368483841e-05, + "loss": 0.329, + "step": 5994 + }, + { + "epoch": 0.36, + "learning_rate": 1.4750595700660306e-05, + "loss": 0.3292, + "step": 5995 + }, + { + "epoch": 0.36, + "learning_rate": 1.4748873850961732e-05, + "loss": 0.3331, + "step": 5996 + }, + { + "epoch": 0.36, + "learning_rate": 1.4747151819454042e-05, + "loss": 0.3265, + "step": 5997 + }, + { + "epoch": 0.36, + "learning_rate": 1.474542960620316e-05, + "loss": 0.3468, + "step": 5998 + }, + { + "epoch": 0.36, + "learning_rate": 1.4743707211275023e-05, + "loss": 0.3359, + "step": 5999 + }, + { + "epoch": 0.36, + "learning_rate": 1.4741984634735574e-05, + "loss": 0.3382, + "step": 6000 + }, + { + "epoch": 0.36, + "learning_rate": 1.4740261876650759e-05, + "loss": 0.3504, + "step": 6001 + }, + { + "epoch": 0.36, + "learning_rate": 1.4738538937086533e-05, + "loss": 0.3227, + "step": 6002 + }, + { + "epoch": 0.36, + "learning_rate": 1.4736815816108855e-05, + "loss": 0.3354, + "step": 6003 + }, + { + "epoch": 0.36, + "learning_rate": 1.4735092513783701e-05, + "loss": 0.3588, + "step": 6004 + }, + { + "epoch": 0.36, + "learning_rate": 1.4733369030177043e-05, + "loss": 0.3078, + "step": 6005 + }, + { + "epoch": 0.36, + "learning_rate": 1.4731645365354866e-05, + "loss": 0.3482, + "step": 6006 + }, + { + "epoch": 0.36, + "learning_rate": 1.472992151938316e-05, + "loss": 0.3078, + "step": 6007 + }, + { + "epoch": 0.36, + "learning_rate": 1.4728197492327917e-05, + "loss": 0.3387, + "step": 6008 + }, + { + "epoch": 0.36, + "learning_rate": 1.4726473284255148e-05, + "loss": 0.3281, + "step": 6009 + }, + { + "epoch": 0.36, + "learning_rate": 1.4724748895230856e-05, + "loss": 0.3177, + "step": 6010 + }, + { + "epoch": 0.36, + "learning_rate": 1.4723024325321064e-05, + "loss": 0.3465, + "step": 6011 + }, + { + "epoch": 0.36, + "learning_rate": 1.47212995745918e-05, + "loss": 0.3521, + "step": 6012 + }, + { + "epoch": 0.36, + "learning_rate": 1.4719574643109087e-05, + "loss": 0.3178, + "step": 6013 + }, + { + "epoch": 0.36, + "learning_rate": 1.4717849530938968e-05, + "loss": 0.3181, + "step": 6014 + }, + { + "epoch": 0.36, + "learning_rate": 1.471612423814749e-05, + "loss": 0.3304, + "step": 6015 + }, + { + "epoch": 0.36, + "learning_rate": 1.4714398764800699e-05, + "loss": 0.358, + "step": 6016 + }, + { + "epoch": 0.36, + "learning_rate": 1.4712673110964665e-05, + "loss": 0.3181, + "step": 6017 + }, + { + "epoch": 0.36, + "learning_rate": 1.4710947276705442e-05, + "loss": 0.3198, + "step": 6018 + }, + { + "epoch": 0.36, + "learning_rate": 1.4709221262089111e-05, + "loss": 0.3309, + "step": 6019 + }, + { + "epoch": 0.36, + "learning_rate": 1.4707495067181748e-05, + "loss": 0.341, + "step": 6020 + }, + { + "epoch": 0.36, + "learning_rate": 1.4705768692049443e-05, + "loss": 0.3384, + "step": 6021 + }, + { + "epoch": 0.36, + "learning_rate": 1.4704042136758286e-05, + "loss": 0.3343, + "step": 6022 + }, + { + "epoch": 0.36, + "learning_rate": 1.470231540137438e-05, + "loss": 0.3239, + "step": 6023 + }, + { + "epoch": 0.36, + "learning_rate": 1.4700588485963834e-05, + "loss": 0.352, + "step": 6024 + }, + { + "epoch": 0.36, + "learning_rate": 1.4698861390592756e-05, + "loss": 0.3532, + "step": 6025 + }, + { + "epoch": 0.36, + "learning_rate": 1.4697134115327276e-05, + "loss": 0.3329, + "step": 6026 + }, + { + "epoch": 0.36, + "learning_rate": 1.4695406660233518e-05, + "loss": 0.3323, + "step": 6027 + }, + { + "epoch": 0.36, + "learning_rate": 1.4693679025377614e-05, + "loss": 0.3424, + "step": 6028 + }, + { + "epoch": 0.36, + "learning_rate": 1.4691951210825713e-05, + "loss": 0.3108, + "step": 6029 + }, + { + "epoch": 0.36, + "learning_rate": 1.4690223216643957e-05, + "loss": 0.3242, + "step": 6030 + }, + { + "epoch": 0.36, + "learning_rate": 1.4688495042898503e-05, + "loss": 0.3359, + "step": 6031 + }, + { + "epoch": 0.36, + "learning_rate": 1.4686766689655518e-05, + "loss": 0.2988, + "step": 6032 + }, + { + "epoch": 0.36, + "learning_rate": 1.4685038156981166e-05, + "loss": 0.3407, + "step": 6033 + }, + { + "epoch": 0.36, + "learning_rate": 1.4683309444941632e-05, + "loss": 0.3616, + "step": 6034 + }, + { + "epoch": 0.36, + "learning_rate": 1.4681580553603088e-05, + "loss": 0.3198, + "step": 6035 + }, + { + "epoch": 0.36, + "learning_rate": 1.467985148303173e-05, + "loss": 0.3323, + "step": 6036 + }, + { + "epoch": 0.36, + "learning_rate": 1.4678122233293754e-05, + "loss": 0.3398, + "step": 6037 + }, + { + "epoch": 0.36, + "learning_rate": 1.4676392804455361e-05, + "loss": 0.3153, + "step": 6038 + }, + { + "epoch": 0.36, + "learning_rate": 1.4674663196582768e-05, + "loss": 0.3052, + "step": 6039 + }, + { + "epoch": 0.36, + "learning_rate": 1.4672933409742186e-05, + "loss": 0.3312, + "step": 6040 + }, + { + "epoch": 0.36, + "learning_rate": 1.4671203443999847e-05, + "loss": 0.332, + "step": 6041 + }, + { + "epoch": 0.37, + "learning_rate": 1.4669473299421975e-05, + "loss": 0.334, + "step": 6042 + }, + { + "epoch": 0.37, + "learning_rate": 1.466774297607481e-05, + "loss": 0.3147, + "step": 6043 + }, + { + "epoch": 0.37, + "learning_rate": 1.4666012474024599e-05, + "loss": 0.3142, + "step": 6044 + }, + { + "epoch": 0.37, + "learning_rate": 1.466428179333759e-05, + "loss": 0.3443, + "step": 6045 + }, + { + "epoch": 0.37, + "learning_rate": 1.4662550934080044e-05, + "loss": 0.3387, + "step": 6046 + }, + { + "epoch": 0.37, + "learning_rate": 1.4660819896318226e-05, + "loss": 0.3725, + "step": 6047 + }, + { + "epoch": 0.37, + "learning_rate": 1.4659088680118407e-05, + "loss": 0.3742, + "step": 6048 + }, + { + "epoch": 0.37, + "learning_rate": 1.4657357285546869e-05, + "loss": 0.3468, + "step": 6049 + }, + { + "epoch": 0.37, + "learning_rate": 1.4655625712669897e-05, + "loss": 0.3538, + "step": 6050 + }, + { + "epoch": 0.37, + "learning_rate": 1.4653893961553782e-05, + "loss": 0.3265, + "step": 6051 + }, + { + "epoch": 0.37, + "learning_rate": 1.4652162032264828e-05, + "loss": 0.3504, + "step": 6052 + }, + { + "epoch": 0.37, + "learning_rate": 1.4650429924869331e-05, + "loss": 0.3412, + "step": 6053 + }, + { + "epoch": 0.37, + "learning_rate": 1.4648697639433616e-05, + "loss": 0.3488, + "step": 6054 + }, + { + "epoch": 0.37, + "learning_rate": 1.4646965176023993e-05, + "loss": 0.3228, + "step": 6055 + }, + { + "epoch": 0.37, + "learning_rate": 1.46452325347068e-05, + "loss": 0.3622, + "step": 6056 + }, + { + "epoch": 0.37, + "learning_rate": 1.4643499715548362e-05, + "loss": 0.3474, + "step": 6057 + }, + { + "epoch": 0.37, + "learning_rate": 1.4641766718615019e-05, + "loss": 0.3387, + "step": 6058 + }, + { + "epoch": 0.37, + "learning_rate": 1.4640033543973123e-05, + "loss": 0.3407, + "step": 6059 + }, + { + "epoch": 0.37, + "learning_rate": 1.4638300191689026e-05, + "loss": 0.3209, + "step": 6060 + }, + { + "epoch": 0.37, + "learning_rate": 1.4636566661829088e-05, + "loss": 0.3259, + "step": 6061 + }, + { + "epoch": 0.37, + "learning_rate": 1.4634832954459681e-05, + "loss": 0.3027, + "step": 6062 + }, + { + "epoch": 0.37, + "learning_rate": 1.4633099069647172e-05, + "loss": 0.3184, + "step": 6063 + }, + { + "epoch": 0.37, + "learning_rate": 1.463136500745795e-05, + "loss": 0.3051, + "step": 6064 + }, + { + "epoch": 0.37, + "learning_rate": 1.4629630767958397e-05, + "loss": 0.3424, + "step": 6065 + }, + { + "epoch": 0.37, + "learning_rate": 1.462789635121491e-05, + "loss": 0.3616, + "step": 6066 + }, + { + "epoch": 0.37, + "learning_rate": 1.4626161757293892e-05, + "loss": 0.3432, + "step": 6067 + }, + { + "epoch": 0.37, + "learning_rate": 1.4624426986261748e-05, + "loss": 0.3477, + "step": 6068 + }, + { + "epoch": 0.37, + "learning_rate": 1.4622692038184899e-05, + "loss": 0.3541, + "step": 6069 + }, + { + "epoch": 0.37, + "learning_rate": 1.4620956913129756e-05, + "loss": 0.332, + "step": 6070 + }, + { + "epoch": 0.37, + "learning_rate": 1.4619221611162762e-05, + "loss": 0.3652, + "step": 6071 + }, + { + "epoch": 0.37, + "learning_rate": 1.4617486132350343e-05, + "loss": 0.3145, + "step": 6072 + }, + { + "epoch": 0.37, + "learning_rate": 1.461575047675894e-05, + "loss": 0.3265, + "step": 6073 + }, + { + "epoch": 0.37, + "learning_rate": 1.461401464445501e-05, + "loss": 0.3373, + "step": 6074 + }, + { + "epoch": 0.37, + "learning_rate": 1.4612278635505002e-05, + "loss": 0.3086, + "step": 6075 + }, + { + "epoch": 0.37, + "learning_rate": 1.4610542449975382e-05, + "loss": 0.308, + "step": 6076 + }, + { + "epoch": 0.37, + "learning_rate": 1.460880608793262e-05, + "loss": 0.3504, + "step": 6077 + }, + { + "epoch": 0.37, + "learning_rate": 1.4607069549443184e-05, + "loss": 0.3267, + "step": 6078 + }, + { + "epoch": 0.37, + "learning_rate": 1.4605332834573568e-05, + "loss": 0.3474, + "step": 6079 + }, + { + "epoch": 0.37, + "learning_rate": 1.4603595943390258e-05, + "loss": 0.3253, + "step": 6080 + }, + { + "epoch": 0.37, + "learning_rate": 1.4601858875959748e-05, + "loss": 0.3446, + "step": 6081 + }, + { + "epoch": 0.37, + "learning_rate": 1.460012163234854e-05, + "loss": 0.3251, + "step": 6082 + }, + { + "epoch": 0.37, + "learning_rate": 1.4598384212623144e-05, + "loss": 0.3354, + "step": 6083 + }, + { + "epoch": 0.37, + "learning_rate": 1.4596646616850083e-05, + "loss": 0.3295, + "step": 6084 + }, + { + "epoch": 0.37, + "learning_rate": 1.4594908845095874e-05, + "loss": 0.3382, + "step": 6085 + }, + { + "epoch": 0.37, + "learning_rate": 1.459317089742705e-05, + "loss": 0.3644, + "step": 6086 + }, + { + "epoch": 0.37, + "learning_rate": 1.4591432773910146e-05, + "loss": 0.3594, + "step": 6087 + }, + { + "epoch": 0.37, + "learning_rate": 1.4589694474611704e-05, + "loss": 0.3016, + "step": 6088 + }, + { + "epoch": 0.37, + "learning_rate": 1.458795599959828e-05, + "loss": 0.3122, + "step": 6089 + }, + { + "epoch": 0.37, + "learning_rate": 1.4586217348936423e-05, + "loss": 0.3209, + "step": 6090 + }, + { + "epoch": 0.37, + "learning_rate": 1.4584478522692704e-05, + "loss": 0.3245, + "step": 6091 + }, + { + "epoch": 0.37, + "learning_rate": 1.4582739520933689e-05, + "loss": 0.3211, + "step": 6092 + }, + { + "epoch": 0.37, + "learning_rate": 1.4581000343725955e-05, + "loss": 0.3379, + "step": 6093 + }, + { + "epoch": 0.37, + "learning_rate": 1.4579260991136091e-05, + "loss": 0.3214, + "step": 6094 + }, + { + "epoch": 0.37, + "learning_rate": 1.4577521463230679e-05, + "loss": 0.3415, + "step": 6095 + }, + { + "epoch": 0.37, + "learning_rate": 1.4575781760076325e-05, + "loss": 0.3248, + "step": 6096 + }, + { + "epoch": 0.37, + "learning_rate": 1.457404188173963e-05, + "loss": 0.3178, + "step": 6097 + }, + { + "epoch": 0.37, + "learning_rate": 1.4572301828287199e-05, + "loss": 0.3359, + "step": 6098 + }, + { + "epoch": 0.37, + "learning_rate": 1.4570561599785657e-05, + "loss": 0.305, + "step": 6099 + }, + { + "epoch": 0.37, + "learning_rate": 1.4568821196301626e-05, + "loss": 0.3647, + "step": 6100 + }, + { + "epoch": 0.37, + "learning_rate": 1.4567080617901734e-05, + "loss": 0.3186, + "step": 6101 + }, + { + "epoch": 0.37, + "learning_rate": 1.4565339864652626e-05, + "loss": 0.3251, + "step": 6102 + }, + { + "epoch": 0.37, + "learning_rate": 1.4563598936620935e-05, + "loss": 0.3343, + "step": 6103 + }, + { + "epoch": 0.37, + "learning_rate": 1.456185783387332e-05, + "loss": 0.3105, + "step": 6104 + }, + { + "epoch": 0.37, + "learning_rate": 1.4560116556476433e-05, + "loss": 0.3203, + "step": 6105 + }, + { + "epoch": 0.37, + "learning_rate": 1.4558375104496945e-05, + "loss": 0.3193, + "step": 6106 + }, + { + "epoch": 0.37, + "learning_rate": 1.4556633478001523e-05, + "loss": 0.3482, + "step": 6107 + }, + { + "epoch": 0.37, + "learning_rate": 1.4554891677056844e-05, + "loss": 0.3265, + "step": 6108 + }, + { + "epoch": 0.37, + "learning_rate": 1.4553149701729594e-05, + "loss": 0.332, + "step": 6109 + }, + { + "epoch": 0.37, + "learning_rate": 1.4551407552086461e-05, + "loss": 0.3158, + "step": 6110 + }, + { + "epoch": 0.37, + "learning_rate": 1.4549665228194147e-05, + "loss": 0.3544, + "step": 6111 + }, + { + "epoch": 0.37, + "learning_rate": 1.4547922730119355e-05, + "loss": 0.3597, + "step": 6112 + }, + { + "epoch": 0.37, + "learning_rate": 1.4546180057928792e-05, + "loss": 0.3098, + "step": 6113 + }, + { + "epoch": 0.37, + "learning_rate": 1.454443721168918e-05, + "loss": 0.3608, + "step": 6114 + }, + { + "epoch": 0.37, + "learning_rate": 1.4542694191467243e-05, + "loss": 0.296, + "step": 6115 + }, + { + "epoch": 0.37, + "learning_rate": 1.454095099732971e-05, + "loss": 0.3026, + "step": 6116 + }, + { + "epoch": 0.37, + "learning_rate": 1.4539207629343322e-05, + "loss": 0.3278, + "step": 6117 + }, + { + "epoch": 0.37, + "learning_rate": 1.453746408757482e-05, + "loss": 0.3566, + "step": 6118 + }, + { + "epoch": 0.37, + "learning_rate": 1.4535720372090957e-05, + "loss": 0.3357, + "step": 6119 + }, + { + "epoch": 0.37, + "learning_rate": 1.4533976482958486e-05, + "loss": 0.3192, + "step": 6120 + }, + { + "epoch": 0.37, + "learning_rate": 1.4532232420244179e-05, + "loss": 0.3158, + "step": 6121 + }, + { + "epoch": 0.37, + "learning_rate": 1.4530488184014799e-05, + "loss": 0.3301, + "step": 6122 + }, + { + "epoch": 0.37, + "learning_rate": 1.4528743774337129e-05, + "loss": 0.3432, + "step": 6123 + }, + { + "epoch": 0.37, + "learning_rate": 1.4526999191277951e-05, + "loss": 0.3516, + "step": 6124 + }, + { + "epoch": 0.37, + "learning_rate": 1.4525254434904055e-05, + "loss": 0.2928, + "step": 6125 + }, + { + "epoch": 0.37, + "learning_rate": 1.4523509505282243e-05, + "loss": 0.3396, + "step": 6126 + }, + { + "epoch": 0.37, + "learning_rate": 1.4521764402479313e-05, + "loss": 0.3329, + "step": 6127 + }, + { + "epoch": 0.37, + "learning_rate": 1.4520019126562078e-05, + "loss": 0.3178, + "step": 6128 + }, + { + "epoch": 0.37, + "learning_rate": 1.4518273677597356e-05, + "loss": 0.3315, + "step": 6129 + }, + { + "epoch": 0.37, + "learning_rate": 1.4516528055651971e-05, + "loss": 0.3312, + "step": 6130 + }, + { + "epoch": 0.37, + "learning_rate": 1.4514782260792752e-05, + "loss": 0.3044, + "step": 6131 + }, + { + "epoch": 0.37, + "learning_rate": 1.451303629308654e-05, + "loss": 0.3449, + "step": 6132 + }, + { + "epoch": 0.37, + "learning_rate": 1.4511290152600176e-05, + "loss": 0.3354, + "step": 6133 + }, + { + "epoch": 0.37, + "learning_rate": 1.4509543839400508e-05, + "loss": 0.3454, + "step": 6134 + }, + { + "epoch": 0.37, + "learning_rate": 1.45077973535544e-05, + "loss": 0.3357, + "step": 6135 + }, + { + "epoch": 0.37, + "learning_rate": 1.4506050695128708e-05, + "loss": 0.3446, + "step": 6136 + }, + { + "epoch": 0.37, + "learning_rate": 1.4504303864190307e-05, + "loss": 0.3281, + "step": 6137 + }, + { + "epoch": 0.37, + "learning_rate": 1.4502556860806073e-05, + "loss": 0.3549, + "step": 6138 + }, + { + "epoch": 0.37, + "learning_rate": 1.4500809685042887e-05, + "loss": 0.317, + "step": 6139 + }, + { + "epoch": 0.37, + "learning_rate": 1.4499062336967642e-05, + "loss": 0.3359, + "step": 6140 + }, + { + "epoch": 0.37, + "learning_rate": 1.4497314816647234e-05, + "loss": 0.3535, + "step": 6141 + }, + { + "epoch": 0.37, + "learning_rate": 1.4495567124148566e-05, + "loss": 0.3027, + "step": 6142 + }, + { + "epoch": 0.37, + "learning_rate": 1.4493819259538546e-05, + "loss": 0.3092, + "step": 6143 + }, + { + "epoch": 0.37, + "learning_rate": 1.4492071222884095e-05, + "loss": 0.315, + "step": 6144 + }, + { + "epoch": 0.37, + "learning_rate": 1.449032301425213e-05, + "loss": 0.3491, + "step": 6145 + }, + { + "epoch": 0.37, + "learning_rate": 1.4488574633709586e-05, + "loss": 0.3351, + "step": 6146 + }, + { + "epoch": 0.37, + "learning_rate": 1.4486826081323399e-05, + "loss": 0.3532, + "step": 6147 + }, + { + "epoch": 0.37, + "learning_rate": 1.4485077357160507e-05, + "loss": 0.3262, + "step": 6148 + }, + { + "epoch": 0.37, + "learning_rate": 1.4483328461287862e-05, + "loss": 0.3398, + "step": 6149 + }, + { + "epoch": 0.37, + "learning_rate": 1.448157939377242e-05, + "loss": 0.3577, + "step": 6150 + }, + { + "epoch": 0.37, + "learning_rate": 1.4479830154681146e-05, + "loss": 0.3203, + "step": 6151 + }, + { + "epoch": 0.37, + "learning_rate": 1.4478080744081005e-05, + "loss": 0.3318, + "step": 6152 + }, + { + "epoch": 0.37, + "learning_rate": 1.4476331162038975e-05, + "loss": 0.3345, + "step": 6153 + }, + { + "epoch": 0.37, + "learning_rate": 1.4474581408622038e-05, + "loss": 0.3393, + "step": 6154 + }, + { + "epoch": 0.37, + "learning_rate": 1.4472831483897181e-05, + "loss": 0.3181, + "step": 6155 + }, + { + "epoch": 0.37, + "learning_rate": 1.44710813879314e-05, + "loss": 0.327, + "step": 6156 + }, + { + "epoch": 0.37, + "learning_rate": 1.4469331120791698e-05, + "loss": 0.3251, + "step": 6157 + }, + { + "epoch": 0.37, + "learning_rate": 1.4467580682545082e-05, + "loss": 0.3555, + "step": 6158 + }, + { + "epoch": 0.37, + "learning_rate": 1.4465830073258568e-05, + "loss": 0.3083, + "step": 6159 + }, + { + "epoch": 0.37, + "learning_rate": 1.4464079292999178e-05, + "loss": 0.3345, + "step": 6160 + }, + { + "epoch": 0.37, + "learning_rate": 1.446232834183394e-05, + "loss": 0.3797, + "step": 6161 + }, + { + "epoch": 0.37, + "learning_rate": 1.4460577219829887e-05, + "loss": 0.3318, + "step": 6162 + }, + { + "epoch": 0.37, + "learning_rate": 1.445882592705406e-05, + "loss": 0.2642, + "step": 6163 + }, + { + "epoch": 0.37, + "learning_rate": 1.445707446357351e-05, + "loss": 0.3078, + "step": 6164 + }, + { + "epoch": 0.37, + "learning_rate": 1.4455322829455288e-05, + "loss": 0.3139, + "step": 6165 + }, + { + "epoch": 0.37, + "learning_rate": 1.4453571024766455e-05, + "loss": 0.3214, + "step": 6166 + }, + { + "epoch": 0.37, + "learning_rate": 1.445181904957408e-05, + "loss": 0.3114, + "step": 6167 + }, + { + "epoch": 0.37, + "learning_rate": 1.4450066903945236e-05, + "loss": 0.3507, + "step": 6168 + }, + { + "epoch": 0.37, + "learning_rate": 1.4448314587947004e-05, + "loss": 0.3172, + "step": 6169 + }, + { + "epoch": 0.37, + "learning_rate": 1.4446562101646469e-05, + "loss": 0.3421, + "step": 6170 + }, + { + "epoch": 0.37, + "learning_rate": 1.4444809445110725e-05, + "loss": 0.3368, + "step": 6171 + }, + { + "epoch": 0.37, + "learning_rate": 1.4443056618406875e-05, + "loss": 0.3064, + "step": 6172 + }, + { + "epoch": 0.37, + "learning_rate": 1.4441303621602018e-05, + "loss": 0.3471, + "step": 6173 + }, + { + "epoch": 0.37, + "learning_rate": 1.4439550454763276e-05, + "loss": 0.351, + "step": 6174 + }, + { + "epoch": 0.37, + "learning_rate": 1.4437797117957758e-05, + "loss": 0.3186, + "step": 6175 + }, + { + "epoch": 0.37, + "learning_rate": 1.4436043611252603e-05, + "loss": 0.3281, + "step": 6176 + }, + { + "epoch": 0.37, + "learning_rate": 1.4434289934714935e-05, + "loss": 0.3178, + "step": 6177 + }, + { + "epoch": 0.37, + "learning_rate": 1.4432536088411891e-05, + "loss": 0.3231, + "step": 6178 + }, + { + "epoch": 0.37, + "learning_rate": 1.4430782072410622e-05, + "loss": 0.3082, + "step": 6179 + }, + { + "epoch": 0.37, + "learning_rate": 1.4429027886778277e-05, + "loss": 0.3267, + "step": 6180 + }, + { + "epoch": 0.37, + "learning_rate": 1.4427273531582016e-05, + "loss": 0.3758, + "step": 6181 + }, + { + "epoch": 0.37, + "learning_rate": 1.4425519006889003e-05, + "loss": 0.3368, + "step": 6182 + }, + { + "epoch": 0.37, + "learning_rate": 1.442376431276641e-05, + "loss": 0.3119, + "step": 6183 + }, + { + "epoch": 0.37, + "learning_rate": 1.4422009449281413e-05, + "loss": 0.2932, + "step": 6184 + }, + { + "epoch": 0.37, + "learning_rate": 1.4420254416501198e-05, + "loss": 0.3348, + "step": 6185 + }, + { + "epoch": 0.37, + "learning_rate": 1.4418499214492959e-05, + "loss": 0.3287, + "step": 6186 + }, + { + "epoch": 0.37, + "learning_rate": 1.4416743843323888e-05, + "loss": 0.356, + "step": 6187 + }, + { + "epoch": 0.37, + "learning_rate": 1.4414988303061193e-05, + "loss": 0.3594, + "step": 6188 + }, + { + "epoch": 0.37, + "learning_rate": 1.441323259377208e-05, + "loss": 0.3284, + "step": 6189 + }, + { + "epoch": 0.37, + "learning_rate": 1.4411476715523764e-05, + "loss": 0.2913, + "step": 6190 + }, + { + "epoch": 0.37, + "learning_rate": 1.4409720668383478e-05, + "loss": 0.3613, + "step": 6191 + }, + { + "epoch": 0.37, + "learning_rate": 1.4407964452418445e-05, + "loss": 0.3278, + "step": 6192 + }, + { + "epoch": 0.37, + "learning_rate": 1.44062080676959e-05, + "loss": 0.3278, + "step": 6193 + }, + { + "epoch": 0.37, + "learning_rate": 1.440445151428309e-05, + "loss": 0.3315, + "step": 6194 + }, + { + "epoch": 0.37, + "learning_rate": 1.4402694792247258e-05, + "loss": 0.3177, + "step": 6195 + }, + { + "epoch": 0.37, + "learning_rate": 1.4400937901655667e-05, + "loss": 0.3521, + "step": 6196 + }, + { + "epoch": 0.37, + "learning_rate": 1.4399180842575575e-05, + "loss": 0.3135, + "step": 6197 + }, + { + "epoch": 0.37, + "learning_rate": 1.4397423615074249e-05, + "loss": 0.3161, + "step": 6198 + }, + { + "epoch": 0.37, + "learning_rate": 1.4395666219218966e-05, + "loss": 0.3287, + "step": 6199 + }, + { + "epoch": 0.37, + "learning_rate": 1.4393908655077007e-05, + "loss": 0.3334, + "step": 6200 + }, + { + "epoch": 0.37, + "learning_rate": 1.4392150922715662e-05, + "loss": 0.3421, + "step": 6201 + }, + { + "epoch": 0.37, + "learning_rate": 1.4390393022202222e-05, + "loss": 0.327, + "step": 6202 + }, + { + "epoch": 0.37, + "learning_rate": 1.4388634953603986e-05, + "loss": 0.2785, + "step": 6203 + }, + { + "epoch": 0.37, + "learning_rate": 1.4386876716988269e-05, + "loss": 0.358, + "step": 6204 + }, + { + "epoch": 0.37, + "learning_rate": 1.4385118312422375e-05, + "loss": 0.3371, + "step": 6205 + }, + { + "epoch": 0.37, + "learning_rate": 1.438335973997363e-05, + "loss": 0.3203, + "step": 6206 + }, + { + "epoch": 0.37, + "learning_rate": 1.438160099970936e-05, + "loss": 0.3432, + "step": 6207 + }, + { + "epoch": 0.38, + "learning_rate": 1.4379842091696894e-05, + "loss": 0.3524, + "step": 6208 + }, + { + "epoch": 0.38, + "learning_rate": 1.4378083016003572e-05, + "loss": 0.3502, + "step": 6209 + }, + { + "epoch": 0.38, + "learning_rate": 1.4376323772696742e-05, + "loss": 0.3184, + "step": 6210 + }, + { + "epoch": 0.38, + "learning_rate": 1.4374564361843759e-05, + "loss": 0.3337, + "step": 6211 + }, + { + "epoch": 0.38, + "learning_rate": 1.4372804783511978e-05, + "loss": 0.291, + "step": 6212 + }, + { + "epoch": 0.38, + "learning_rate": 1.437104503776876e-05, + "loss": 0.3111, + "step": 6213 + }, + { + "epoch": 0.38, + "learning_rate": 1.436928512468148e-05, + "loss": 0.3396, + "step": 6214 + }, + { + "epoch": 0.38, + "learning_rate": 1.4367525044317519e-05, + "loss": 0.3206, + "step": 6215 + }, + { + "epoch": 0.38, + "learning_rate": 1.4365764796744254e-05, + "loss": 0.3178, + "step": 6216 + }, + { + "epoch": 0.38, + "learning_rate": 1.4364004382029084e-05, + "loss": 0.3376, + "step": 6217 + }, + { + "epoch": 0.38, + "learning_rate": 1.4362243800239397e-05, + "loss": 0.3577, + "step": 6218 + }, + { + "epoch": 0.38, + "learning_rate": 1.4360483051442604e-05, + "loss": 0.3594, + "step": 6219 + }, + { + "epoch": 0.38, + "learning_rate": 1.4358722135706108e-05, + "loss": 0.3164, + "step": 6220 + }, + { + "epoch": 0.38, + "learning_rate": 1.4356961053097334e-05, + "loss": 0.3211, + "step": 6221 + }, + { + "epoch": 0.38, + "learning_rate": 1.4355199803683695e-05, + "loss": 0.3147, + "step": 6222 + }, + { + "epoch": 0.38, + "learning_rate": 1.4353438387532623e-05, + "loss": 0.3041, + "step": 6223 + }, + { + "epoch": 0.38, + "learning_rate": 1.4351676804711555e-05, + "loss": 0.3468, + "step": 6224 + }, + { + "epoch": 0.38, + "learning_rate": 1.4349915055287933e-05, + "loss": 0.338, + "step": 6225 + }, + { + "epoch": 0.38, + "learning_rate": 1.4348153139329202e-05, + "loss": 0.3211, + "step": 6226 + }, + { + "epoch": 0.38, + "learning_rate": 1.434639105690282e-05, + "loss": 0.3421, + "step": 6227 + }, + { + "epoch": 0.38, + "learning_rate": 1.4344628808076242e-05, + "loss": 0.31, + "step": 6228 + }, + { + "epoch": 0.38, + "learning_rate": 1.4342866392916942e-05, + "loss": 0.3276, + "step": 6229 + }, + { + "epoch": 0.38, + "learning_rate": 1.4341103811492389e-05, + "loss": 0.316, + "step": 6230 + }, + { + "epoch": 0.38, + "learning_rate": 1.4339341063870068e-05, + "loss": 0.3037, + "step": 6231 + }, + { + "epoch": 0.38, + "learning_rate": 1.4337578150117457e-05, + "loss": 0.3318, + "step": 6232 + }, + { + "epoch": 0.38, + "learning_rate": 1.4335815070302054e-05, + "loss": 0.3552, + "step": 6233 + }, + { + "epoch": 0.38, + "learning_rate": 1.4334051824491359e-05, + "loss": 0.3048, + "step": 6234 + }, + { + "epoch": 0.38, + "learning_rate": 1.4332288412752876e-05, + "loss": 0.3424, + "step": 6235 + }, + { + "epoch": 0.38, + "learning_rate": 1.4330524835154113e-05, + "loss": 0.3139, + "step": 6236 + }, + { + "epoch": 0.38, + "learning_rate": 1.4328761091762593e-05, + "loss": 0.3114, + "step": 6237 + }, + { + "epoch": 0.38, + "learning_rate": 1.4326997182645838e-05, + "loss": 0.3359, + "step": 6238 + }, + { + "epoch": 0.38, + "learning_rate": 1.432523310787138e-05, + "loss": 0.3426, + "step": 6239 + }, + { + "epoch": 0.38, + "learning_rate": 1.4323468867506755e-05, + "loss": 0.3253, + "step": 6240 + }, + { + "epoch": 0.38, + "learning_rate": 1.4321704461619505e-05, + "loss": 0.315, + "step": 6241 + }, + { + "epoch": 0.38, + "learning_rate": 1.4319939890277184e-05, + "loss": 0.3256, + "step": 6242 + }, + { + "epoch": 0.38, + "learning_rate": 1.4318175153547342e-05, + "loss": 0.3133, + "step": 6243 + }, + { + "epoch": 0.38, + "learning_rate": 1.4316410251497548e-05, + "loss": 0.3499, + "step": 6244 + }, + { + "epoch": 0.38, + "learning_rate": 1.4314645184195364e-05, + "loss": 0.3479, + "step": 6245 + }, + { + "epoch": 0.38, + "learning_rate": 1.431287995170837e-05, + "loss": 0.3404, + "step": 6246 + }, + { + "epoch": 0.38, + "learning_rate": 1.4311114554104147e-05, + "loss": 0.3496, + "step": 6247 + }, + { + "epoch": 0.38, + "learning_rate": 1.4309348991450278e-05, + "loss": 0.3521, + "step": 6248 + }, + { + "epoch": 0.38, + "learning_rate": 1.4307583263814365e-05, + "loss": 0.3292, + "step": 6249 + }, + { + "epoch": 0.38, + "learning_rate": 1.4305817371264e-05, + "loss": 0.332, + "step": 6250 + }, + { + "epoch": 0.38, + "learning_rate": 1.4304051313866793e-05, + "loss": 0.3382, + "step": 6251 + }, + { + "epoch": 0.38, + "learning_rate": 1.4302285091690362e-05, + "loss": 0.3334, + "step": 6252 + }, + { + "epoch": 0.38, + "learning_rate": 1.4300518704802318e-05, + "loss": 0.3209, + "step": 6253 + }, + { + "epoch": 0.38, + "learning_rate": 1.4298752153270291e-05, + "loss": 0.3412, + "step": 6254 + }, + { + "epoch": 0.38, + "learning_rate": 1.4296985437161915e-05, + "loss": 0.3359, + "step": 6255 + }, + { + "epoch": 0.38, + "learning_rate": 1.4295218556544822e-05, + "loss": 0.344, + "step": 6256 + }, + { + "epoch": 0.38, + "learning_rate": 1.4293451511486658e-05, + "loss": 0.3708, + "step": 6257 + }, + { + "epoch": 0.38, + "learning_rate": 1.4291684302055078e-05, + "loss": 0.3055, + "step": 6258 + }, + { + "epoch": 0.38, + "learning_rate": 1.4289916928317738e-05, + "loss": 0.3496, + "step": 6259 + }, + { + "epoch": 0.38, + "learning_rate": 1.4288149390342296e-05, + "loss": 0.3249, + "step": 6260 + }, + { + "epoch": 0.38, + "learning_rate": 1.4286381688196427e-05, + "loss": 0.3597, + "step": 6261 + }, + { + "epoch": 0.38, + "learning_rate": 1.4284613821947807e-05, + "loss": 0.3218, + "step": 6262 + }, + { + "epoch": 0.38, + "learning_rate": 1.4282845791664112e-05, + "loss": 0.3202, + "step": 6263 + }, + { + "epoch": 0.38, + "learning_rate": 1.428107759741304e-05, + "loss": 0.3262, + "step": 6264 + }, + { + "epoch": 0.38, + "learning_rate": 1.4279309239262278e-05, + "loss": 0.3195, + "step": 6265 + }, + { + "epoch": 0.38, + "learning_rate": 1.427754071727953e-05, + "loss": 0.326, + "step": 6266 + }, + { + "epoch": 0.38, + "learning_rate": 1.4275772031532504e-05, + "loss": 0.3387, + "step": 6267 + }, + { + "epoch": 0.38, + "learning_rate": 1.4274003182088911e-05, + "loss": 0.3281, + "step": 6268 + }, + { + "epoch": 0.38, + "learning_rate": 1.4272234169016475e-05, + "loss": 0.327, + "step": 6269 + }, + { + "epoch": 0.38, + "learning_rate": 1.4270464992382918e-05, + "loss": 0.3479, + "step": 6270 + }, + { + "epoch": 0.38, + "learning_rate": 1.4268695652255975e-05, + "loss": 0.3276, + "step": 6271 + }, + { + "epoch": 0.38, + "learning_rate": 1.4266926148703383e-05, + "loss": 0.3242, + "step": 6272 + }, + { + "epoch": 0.38, + "learning_rate": 1.4265156481792887e-05, + "loss": 0.3337, + "step": 6273 + }, + { + "epoch": 0.38, + "learning_rate": 1.426338665159224e-05, + "loss": 0.3152, + "step": 6274 + }, + { + "epoch": 0.38, + "learning_rate": 1.4261616658169196e-05, + "loss": 0.3524, + "step": 6275 + }, + { + "epoch": 0.38, + "learning_rate": 1.4259846501591523e-05, + "loss": 0.3047, + "step": 6276 + }, + { + "epoch": 0.38, + "learning_rate": 1.4258076181926988e-05, + "loss": 0.3262, + "step": 6277 + }, + { + "epoch": 0.38, + "learning_rate": 1.4256305699243365e-05, + "loss": 0.3373, + "step": 6278 + }, + { + "epoch": 0.38, + "learning_rate": 1.425453505360844e-05, + "loss": 0.3315, + "step": 6279 + }, + { + "epoch": 0.38, + "learning_rate": 1.425276424509e-05, + "loss": 0.3479, + "step": 6280 + }, + { + "epoch": 0.38, + "learning_rate": 1.4250993273755845e-05, + "loss": 0.3557, + "step": 6281 + }, + { + "epoch": 0.38, + "learning_rate": 1.4249222139673765e-05, + "loss": 0.3368, + "step": 6282 + }, + { + "epoch": 0.38, + "learning_rate": 1.4247450842911577e-05, + "loss": 0.3493, + "step": 6283 + }, + { + "epoch": 0.38, + "learning_rate": 1.4245679383537092e-05, + "loss": 0.31, + "step": 6284 + }, + { + "epoch": 0.38, + "learning_rate": 1.4243907761618128e-05, + "loss": 0.3449, + "step": 6285 + }, + { + "epoch": 0.38, + "learning_rate": 1.4242135977222514e-05, + "loss": 0.3343, + "step": 6286 + }, + { + "epoch": 0.38, + "learning_rate": 1.4240364030418076e-05, + "loss": 0.3278, + "step": 6287 + }, + { + "epoch": 0.38, + "learning_rate": 1.423859192127266e-05, + "loss": 0.3267, + "step": 6288 + }, + { + "epoch": 0.38, + "learning_rate": 1.4236819649854108e-05, + "loss": 0.3156, + "step": 6289 + }, + { + "epoch": 0.38, + "learning_rate": 1.4235047216230266e-05, + "loss": 0.3435, + "step": 6290 + }, + { + "epoch": 0.38, + "learning_rate": 1.4233274620469e-05, + "loss": 0.332, + "step": 6291 + }, + { + "epoch": 0.38, + "learning_rate": 1.4231501862638164e-05, + "loss": 0.3557, + "step": 6292 + }, + { + "epoch": 0.38, + "learning_rate": 1.4229728942805636e-05, + "loss": 0.3235, + "step": 6293 + }, + { + "epoch": 0.38, + "learning_rate": 1.4227955861039284e-05, + "loss": 0.3518, + "step": 6294 + }, + { + "epoch": 0.38, + "learning_rate": 1.4226182617406996e-05, + "loss": 0.3368, + "step": 6295 + }, + { + "epoch": 0.38, + "learning_rate": 1.4224409211976658e-05, + "loss": 0.3331, + "step": 6296 + }, + { + "epoch": 0.38, + "learning_rate": 1.4222635644816163e-05, + "loss": 0.298, + "step": 6297 + }, + { + "epoch": 0.38, + "learning_rate": 1.422086191599341e-05, + "loss": 0.3675, + "step": 6298 + }, + { + "epoch": 0.38, + "learning_rate": 1.4219088025576309e-05, + "loss": 0.3387, + "step": 6299 + }, + { + "epoch": 0.38, + "learning_rate": 1.4217313973632771e-05, + "loss": 0.3384, + "step": 6300 + }, + { + "epoch": 0.38, + "learning_rate": 1.4215539760230718e-05, + "loss": 0.3387, + "step": 6301 + }, + { + "epoch": 0.38, + "learning_rate": 1.4213765385438074e-05, + "loss": 0.334, + "step": 6302 + }, + { + "epoch": 0.38, + "learning_rate": 1.4211990849322765e-05, + "loss": 0.3114, + "step": 6303 + }, + { + "epoch": 0.38, + "learning_rate": 1.4210216151952736e-05, + "loss": 0.344, + "step": 6304 + }, + { + "epoch": 0.38, + "learning_rate": 1.4208441293395924e-05, + "loss": 0.3619, + "step": 6305 + }, + { + "epoch": 0.38, + "learning_rate": 1.4206666273720289e-05, + "loss": 0.3463, + "step": 6306 + }, + { + "epoch": 0.38, + "learning_rate": 1.4204891092993776e-05, + "loss": 0.3555, + "step": 6307 + }, + { + "epoch": 0.38, + "learning_rate": 1.420311575128435e-05, + "loss": 0.3203, + "step": 6308 + }, + { + "epoch": 0.38, + "learning_rate": 1.4201340248659985e-05, + "loss": 0.3502, + "step": 6309 + }, + { + "epoch": 0.38, + "learning_rate": 1.4199564585188648e-05, + "loss": 0.3418, + "step": 6310 + }, + { + "epoch": 0.38, + "learning_rate": 1.4197788760938327e-05, + "loss": 0.3373, + "step": 6311 + }, + { + "epoch": 0.38, + "learning_rate": 1.4196012775977002e-05, + "loss": 0.3217, + "step": 6312 + }, + { + "epoch": 0.38, + "learning_rate": 1.419423663037267e-05, + "loss": 0.334, + "step": 6313 + }, + { + "epoch": 0.38, + "learning_rate": 1.4192460324193333e-05, + "loss": 0.3245, + "step": 6314 + }, + { + "epoch": 0.38, + "learning_rate": 1.4190683857506989e-05, + "loss": 0.3231, + "step": 6315 + }, + { + "epoch": 0.38, + "learning_rate": 1.4188907230381655e-05, + "loss": 0.315, + "step": 6316 + }, + { + "epoch": 0.38, + "learning_rate": 1.4187130442885346e-05, + "loss": 0.3228, + "step": 6317 + }, + { + "epoch": 0.38, + "learning_rate": 1.4185353495086085e-05, + "loss": 0.3326, + "step": 6318 + }, + { + "epoch": 0.38, + "learning_rate": 1.4183576387051905e-05, + "loss": 0.3295, + "step": 6319 + }, + { + "epoch": 0.38, + "learning_rate": 1.4181799118850841e-05, + "loss": 0.3276, + "step": 6320 + }, + { + "epoch": 0.38, + "learning_rate": 1.4180021690550934e-05, + "loss": 0.3312, + "step": 6321 + }, + { + "epoch": 0.38, + "learning_rate": 1.4178244102220234e-05, + "loss": 0.3463, + "step": 6322 + }, + { + "epoch": 0.38, + "learning_rate": 1.4176466353926794e-05, + "loss": 0.2945, + "step": 6323 + }, + { + "epoch": 0.38, + "learning_rate": 1.4174688445738673e-05, + "loss": 0.3365, + "step": 6324 + }, + { + "epoch": 0.38, + "learning_rate": 1.4172910377723938e-05, + "loss": 0.3178, + "step": 6325 + }, + { + "epoch": 0.38, + "learning_rate": 1.4171132149950668e-05, + "loss": 0.3624, + "step": 6326 + }, + { + "epoch": 0.38, + "learning_rate": 1.4169353762486933e-05, + "loss": 0.3329, + "step": 6327 + }, + { + "epoch": 0.38, + "learning_rate": 1.4167575215400824e-05, + "loss": 0.3036, + "step": 6328 + }, + { + "epoch": 0.38, + "learning_rate": 1.416579650876043e-05, + "loss": 0.3078, + "step": 6329 + }, + { + "epoch": 0.38, + "learning_rate": 1.4164017642633848e-05, + "loss": 0.3011, + "step": 6330 + }, + { + "epoch": 0.38, + "learning_rate": 1.4162238617089181e-05, + "loss": 0.3442, + "step": 6331 + }, + { + "epoch": 0.38, + "learning_rate": 1.4160459432194542e-05, + "loss": 0.3164, + "step": 6332 + }, + { + "epoch": 0.38, + "learning_rate": 1.415868008801804e-05, + "loss": 0.3198, + "step": 6333 + }, + { + "epoch": 0.38, + "learning_rate": 1.4156900584627802e-05, + "loss": 0.3364, + "step": 6334 + }, + { + "epoch": 0.38, + "learning_rate": 1.4155120922091954e-05, + "loss": 0.3404, + "step": 6335 + }, + { + "epoch": 0.38, + "learning_rate": 1.415334110047863e-05, + "loss": 0.3117, + "step": 6336 + }, + { + "epoch": 0.38, + "learning_rate": 1.415156111985597e-05, + "loss": 0.3518, + "step": 6337 + }, + { + "epoch": 0.38, + "learning_rate": 1.4149780980292118e-05, + "loss": 0.3345, + "step": 6338 + }, + { + "epoch": 0.38, + "learning_rate": 1.414800068185523e-05, + "loss": 0.3482, + "step": 6339 + }, + { + "epoch": 0.38, + "learning_rate": 1.4146220224613463e-05, + "loss": 0.3298, + "step": 6340 + }, + { + "epoch": 0.38, + "learning_rate": 1.4144439608634977e-05, + "loss": 0.3262, + "step": 6341 + }, + { + "epoch": 0.38, + "learning_rate": 1.4142658833987946e-05, + "loss": 0.3613, + "step": 6342 + }, + { + "epoch": 0.38, + "learning_rate": 1.4140877900740546e-05, + "loss": 0.3172, + "step": 6343 + }, + { + "epoch": 0.38, + "learning_rate": 1.4139096808960962e-05, + "loss": 0.322, + "step": 6344 + }, + { + "epoch": 0.38, + "learning_rate": 1.4137315558717374e-05, + "loss": 0.3227, + "step": 6345 + }, + { + "epoch": 0.38, + "learning_rate": 1.4135534150077987e-05, + "loss": 0.3234, + "step": 6346 + }, + { + "epoch": 0.38, + "learning_rate": 1.4133752583110995e-05, + "loss": 0.3304, + "step": 6347 + }, + { + "epoch": 0.38, + "learning_rate": 1.4131970857884609e-05, + "loss": 0.3384, + "step": 6348 + }, + { + "epoch": 0.38, + "learning_rate": 1.4130188974467038e-05, + "loss": 0.3521, + "step": 6349 + }, + { + "epoch": 0.38, + "learning_rate": 1.4128406932926504e-05, + "loss": 0.3262, + "step": 6350 + }, + { + "epoch": 0.38, + "learning_rate": 1.4126624733331229e-05, + "loss": 0.3426, + "step": 6351 + }, + { + "epoch": 0.38, + "learning_rate": 1.4124842375749449e-05, + "loss": 0.3309, + "step": 6352 + }, + { + "epoch": 0.38, + "learning_rate": 1.412305986024939e-05, + "loss": 0.308, + "step": 6353 + }, + { + "epoch": 0.38, + "learning_rate": 1.4121277186899307e-05, + "loss": 0.3343, + "step": 6354 + }, + { + "epoch": 0.38, + "learning_rate": 1.4119494355767442e-05, + "loss": 0.3069, + "step": 6355 + }, + { + "epoch": 0.38, + "learning_rate": 1.4117711366922056e-05, + "loss": 0.3636, + "step": 6356 + }, + { + "epoch": 0.38, + "learning_rate": 1.4115928220431405e-05, + "loss": 0.3633, + "step": 6357 + }, + { + "epoch": 0.38, + "learning_rate": 1.4114144916363759e-05, + "loss": 0.3278, + "step": 6358 + }, + { + "epoch": 0.38, + "learning_rate": 1.4112361454787389e-05, + "loss": 0.3312, + "step": 6359 + }, + { + "epoch": 0.38, + "learning_rate": 1.4110577835770574e-05, + "loss": 0.3312, + "step": 6360 + }, + { + "epoch": 0.38, + "learning_rate": 1.4108794059381604e-05, + "loss": 0.3386, + "step": 6361 + }, + { + "epoch": 0.38, + "learning_rate": 1.4107010125688762e-05, + "loss": 0.2806, + "step": 6362 + }, + { + "epoch": 0.38, + "learning_rate": 1.4105226034760353e-05, + "loss": 0.3089, + "step": 6363 + }, + { + "epoch": 0.38, + "learning_rate": 1.4103441786664677e-05, + "loss": 0.3569, + "step": 6364 + }, + { + "epoch": 0.38, + "learning_rate": 1.4101657381470046e-05, + "loss": 0.3172, + "step": 6365 + }, + { + "epoch": 0.38, + "learning_rate": 1.4099872819244773e-05, + "loss": 0.3234, + "step": 6366 + }, + { + "epoch": 0.38, + "learning_rate": 1.4098088100057178e-05, + "loss": 0.3203, + "step": 6367 + }, + { + "epoch": 0.38, + "learning_rate": 1.409630322397559e-05, + "loss": 0.3244, + "step": 6368 + }, + { + "epoch": 0.38, + "learning_rate": 1.4094518191068345e-05, + "loss": 0.3105, + "step": 6369 + }, + { + "epoch": 0.38, + "learning_rate": 1.4092733001403778e-05, + "loss": 0.3161, + "step": 6370 + }, + { + "epoch": 0.38, + "learning_rate": 1.4090947655050236e-05, + "loss": 0.373, + "step": 6371 + }, + { + "epoch": 0.38, + "learning_rate": 1.4089162152076072e-05, + "loss": 0.3163, + "step": 6372 + }, + { + "epoch": 0.39, + "learning_rate": 1.408737649254964e-05, + "loss": 0.3278, + "step": 6373 + }, + { + "epoch": 0.39, + "learning_rate": 1.4085590676539312e-05, + "loss": 0.3228, + "step": 6374 + }, + { + "epoch": 0.39, + "learning_rate": 1.4083804704113444e-05, + "loss": 0.3103, + "step": 6375 + }, + { + "epoch": 0.39, + "learning_rate": 1.4082018575340422e-05, + "loss": 0.3145, + "step": 6376 + }, + { + "epoch": 0.39, + "learning_rate": 1.4080232290288622e-05, + "loss": 0.351, + "step": 6377 + }, + { + "epoch": 0.39, + "learning_rate": 1.4078445849026435e-05, + "loss": 0.322, + "step": 6378 + }, + { + "epoch": 0.39, + "learning_rate": 1.4076659251622252e-05, + "loss": 0.3323, + "step": 6379 + }, + { + "epoch": 0.39, + "learning_rate": 1.4074872498144472e-05, + "loss": 0.3034, + "step": 6380 + }, + { + "epoch": 0.39, + "learning_rate": 1.4073085588661501e-05, + "loss": 0.3343, + "step": 6381 + }, + { + "epoch": 0.39, + "learning_rate": 1.407129852324175e-05, + "loss": 0.3359, + "step": 6382 + }, + { + "epoch": 0.39, + "learning_rate": 1.4069511301953637e-05, + "loss": 0.2944, + "step": 6383 + }, + { + "epoch": 0.39, + "learning_rate": 1.4067723924865585e-05, + "loss": 0.3273, + "step": 6384 + }, + { + "epoch": 0.39, + "learning_rate": 1.4065936392046022e-05, + "loss": 0.3231, + "step": 6385 + }, + { + "epoch": 0.39, + "learning_rate": 1.4064148703563385e-05, + "loss": 0.2955, + "step": 6386 + }, + { + "epoch": 0.39, + "learning_rate": 1.4062360859486115e-05, + "loss": 0.3228, + "step": 6387 + }, + { + "epoch": 0.39, + "learning_rate": 1.4060572859882658e-05, + "loss": 0.3069, + "step": 6388 + }, + { + "epoch": 0.39, + "learning_rate": 1.4058784704821465e-05, + "loss": 0.3064, + "step": 6389 + }, + { + "epoch": 0.39, + "learning_rate": 1.4056996394370997e-05, + "loss": 0.341, + "step": 6390 + }, + { + "epoch": 0.39, + "learning_rate": 1.4055207928599723e-05, + "loss": 0.3184, + "step": 6391 + }, + { + "epoch": 0.39, + "learning_rate": 1.4053419307576107e-05, + "loss": 0.3117, + "step": 6392 + }, + { + "epoch": 0.39, + "learning_rate": 1.405163053136863e-05, + "loss": 0.3468, + "step": 6393 + }, + { + "epoch": 0.39, + "learning_rate": 1.4049841600045771e-05, + "loss": 0.2941, + "step": 6394 + }, + { + "epoch": 0.39, + "learning_rate": 1.404805251367602e-05, + "loss": 0.3424, + "step": 6395 + }, + { + "epoch": 0.39, + "learning_rate": 1.4046263272327876e-05, + "loss": 0.2951, + "step": 6396 + }, + { + "epoch": 0.39, + "learning_rate": 1.4044473876069833e-05, + "loss": 0.3502, + "step": 6397 + }, + { + "epoch": 0.39, + "learning_rate": 1.4042684324970404e-05, + "loss": 0.3351, + "step": 6398 + }, + { + "epoch": 0.39, + "learning_rate": 1.4040894619098097e-05, + "loss": 0.332, + "step": 6399 + }, + { + "epoch": 0.39, + "learning_rate": 1.403910475852143e-05, + "loss": 0.3337, + "step": 6400 + }, + { + "epoch": 0.39, + "learning_rate": 1.403731474330893e-05, + "loss": 0.3435, + "step": 6401 + }, + { + "epoch": 0.39, + "learning_rate": 1.4035524573529124e-05, + "loss": 0.3301, + "step": 6402 + }, + { + "epoch": 0.39, + "learning_rate": 1.403373424925055e-05, + "loss": 0.3092, + "step": 6403 + }, + { + "epoch": 0.39, + "learning_rate": 1.4031943770541752e-05, + "loss": 0.3438, + "step": 6404 + }, + { + "epoch": 0.39, + "learning_rate": 1.4030153137471274e-05, + "loss": 0.3563, + "step": 6405 + }, + { + "epoch": 0.39, + "learning_rate": 1.402836235010767e-05, + "loss": 0.3338, + "step": 6406 + }, + { + "epoch": 0.39, + "learning_rate": 1.4026571408519504e-05, + "loss": 0.327, + "step": 6407 + }, + { + "epoch": 0.39, + "learning_rate": 1.4024780312775339e-05, + "loss": 0.3217, + "step": 6408 + }, + { + "epoch": 0.39, + "learning_rate": 1.4022989062943746e-05, + "loss": 0.3343, + "step": 6409 + }, + { + "epoch": 0.39, + "learning_rate": 1.4021197659093301e-05, + "loss": 0.3426, + "step": 6410 + }, + { + "epoch": 0.39, + "learning_rate": 1.4019406101292589e-05, + "loss": 0.2727, + "step": 6411 + }, + { + "epoch": 0.39, + "learning_rate": 1.4017614389610202e-05, + "loss": 0.3546, + "step": 6412 + }, + { + "epoch": 0.39, + "learning_rate": 1.401582252411473e-05, + "loss": 0.3057, + "step": 6413 + }, + { + "epoch": 0.39, + "learning_rate": 1.4014030504874778e-05, + "loss": 0.3287, + "step": 6414 + }, + { + "epoch": 0.39, + "learning_rate": 1.401223833195895e-05, + "loss": 0.3206, + "step": 6415 + }, + { + "epoch": 0.39, + "learning_rate": 1.401044600543586e-05, + "loss": 0.3309, + "step": 6416 + }, + { + "epoch": 0.39, + "learning_rate": 1.4008653525374127e-05, + "loss": 0.3087, + "step": 6417 + }, + { + "epoch": 0.39, + "learning_rate": 1.4006860891842378e-05, + "loss": 0.3107, + "step": 6418 + }, + { + "epoch": 0.39, + "learning_rate": 1.4005068104909239e-05, + "loss": 0.346, + "step": 6419 + }, + { + "epoch": 0.39, + "learning_rate": 1.4003275164643344e-05, + "loss": 0.3451, + "step": 6420 + }, + { + "epoch": 0.39, + "learning_rate": 1.4001482071113344e-05, + "loss": 0.3471, + "step": 6421 + }, + { + "epoch": 0.39, + "learning_rate": 1.3999688824387882e-05, + "loss": 0.3092, + "step": 6422 + }, + { + "epoch": 0.39, + "learning_rate": 1.3997895424535608e-05, + "loss": 0.3506, + "step": 6423 + }, + { + "epoch": 0.39, + "learning_rate": 1.399610187162519e-05, + "loss": 0.3588, + "step": 6424 + }, + { + "epoch": 0.39, + "learning_rate": 1.3994308165725287e-05, + "loss": 0.3301, + "step": 6425 + }, + { + "epoch": 0.39, + "learning_rate": 1.3992514306904574e-05, + "loss": 0.3245, + "step": 6426 + }, + { + "epoch": 0.39, + "learning_rate": 1.3990720295231726e-05, + "loss": 0.3189, + "step": 6427 + }, + { + "epoch": 0.39, + "learning_rate": 1.3988926130775428e-05, + "loss": 0.3111, + "step": 6428 + }, + { + "epoch": 0.39, + "learning_rate": 1.3987131813604369e-05, + "loss": 0.3292, + "step": 6429 + }, + { + "epoch": 0.39, + "learning_rate": 1.3985337343787242e-05, + "loss": 0.3209, + "step": 6430 + }, + { + "epoch": 0.39, + "learning_rate": 1.398354272139275e-05, + "loss": 0.3387, + "step": 6431 + }, + { + "epoch": 0.39, + "learning_rate": 1.3981747946489595e-05, + "loss": 0.3301, + "step": 6432 + }, + { + "epoch": 0.39, + "learning_rate": 1.3979953019146495e-05, + "loss": 0.3278, + "step": 6433 + }, + { + "epoch": 0.39, + "learning_rate": 1.3978157939432167e-05, + "loss": 0.3605, + "step": 6434 + }, + { + "epoch": 0.39, + "learning_rate": 1.3976362707415333e-05, + "loss": 0.3131, + "step": 6435 + }, + { + "epoch": 0.39, + "learning_rate": 1.3974567323164723e-05, + "loss": 0.3577, + "step": 6436 + }, + { + "epoch": 0.39, + "learning_rate": 1.3972771786749074e-05, + "loss": 0.2924, + "step": 6437 + }, + { + "epoch": 0.39, + "learning_rate": 1.3970976098237128e-05, + "loss": 0.3284, + "step": 6438 + }, + { + "epoch": 0.39, + "learning_rate": 1.3969180257697631e-05, + "loss": 0.3083, + "step": 6439 + }, + { + "epoch": 0.39, + "learning_rate": 1.3967384265199337e-05, + "loss": 0.3491, + "step": 6440 + }, + { + "epoch": 0.39, + "learning_rate": 1.3965588120811004e-05, + "loss": 0.3301, + "step": 6441 + }, + { + "epoch": 0.39, + "learning_rate": 1.3963791824601401e-05, + "loss": 0.3203, + "step": 6442 + }, + { + "epoch": 0.39, + "learning_rate": 1.3961995376639293e-05, + "loss": 0.3276, + "step": 6443 + }, + { + "epoch": 0.39, + "learning_rate": 1.3960198776993457e-05, + "loss": 0.3122, + "step": 6444 + }, + { + "epoch": 0.39, + "learning_rate": 1.3958402025732679e-05, + "loss": 0.3228, + "step": 6445 + }, + { + "epoch": 0.39, + "learning_rate": 1.3956605122925741e-05, + "loss": 0.3239, + "step": 6446 + }, + { + "epoch": 0.39, + "learning_rate": 1.3954808068641443e-05, + "loss": 0.3125, + "step": 6447 + }, + { + "epoch": 0.39, + "learning_rate": 1.3953010862948584e-05, + "loss": 0.3225, + "step": 6448 + }, + { + "epoch": 0.39, + "learning_rate": 1.3951213505915968e-05, + "loss": 0.3438, + "step": 6449 + }, + { + "epoch": 0.39, + "learning_rate": 1.3949415997612403e-05, + "loss": 0.3309, + "step": 6450 + }, + { + "epoch": 0.39, + "learning_rate": 1.3947618338106712e-05, + "loss": 0.3557, + "step": 6451 + }, + { + "epoch": 0.39, + "learning_rate": 1.394582052746771e-05, + "loss": 0.3097, + "step": 6452 + }, + { + "epoch": 0.39, + "learning_rate": 1.3944022565764237e-05, + "loss": 0.3276, + "step": 6453 + }, + { + "epoch": 0.39, + "learning_rate": 1.3942224453065119e-05, + "loss": 0.3474, + "step": 6454 + }, + { + "epoch": 0.39, + "learning_rate": 1.3940426189439195e-05, + "loss": 0.3209, + "step": 6455 + }, + { + "epoch": 0.39, + "learning_rate": 1.3938627774955315e-05, + "loss": 0.346, + "step": 6456 + }, + { + "epoch": 0.39, + "learning_rate": 1.3936829209682331e-05, + "loss": 0.2966, + "step": 6457 + }, + { + "epoch": 0.39, + "learning_rate": 1.39350304936891e-05, + "loss": 0.3683, + "step": 6458 + }, + { + "epoch": 0.39, + "learning_rate": 1.3933231627044485e-05, + "loss": 0.3234, + "step": 6459 + }, + { + "epoch": 0.39, + "learning_rate": 1.3931432609817353e-05, + "loss": 0.3622, + "step": 6460 + }, + { + "epoch": 0.39, + "learning_rate": 1.392963344207658e-05, + "loss": 0.3156, + "step": 6461 + }, + { + "epoch": 0.39, + "learning_rate": 1.3927834123891046e-05, + "loss": 0.3026, + "step": 6462 + }, + { + "epoch": 0.39, + "learning_rate": 1.3926034655329642e-05, + "loss": 0.3365, + "step": 6463 + }, + { + "epoch": 0.39, + "learning_rate": 1.3924235036461254e-05, + "loss": 0.344, + "step": 6464 + }, + { + "epoch": 0.39, + "learning_rate": 1.3922435267354779e-05, + "loss": 0.3098, + "step": 6465 + }, + { + "epoch": 0.39, + "learning_rate": 1.3920635348079127e-05, + "loss": 0.3516, + "step": 6466 + }, + { + "epoch": 0.39, + "learning_rate": 1.3918835278703205e-05, + "loss": 0.3518, + "step": 6467 + }, + { + "epoch": 0.39, + "learning_rate": 1.3917035059295925e-05, + "loss": 0.3407, + "step": 6468 + }, + { + "epoch": 0.39, + "learning_rate": 1.3915234689926211e-05, + "loss": 0.3348, + "step": 6469 + }, + { + "epoch": 0.39, + "learning_rate": 1.3913434170662988e-05, + "loss": 0.3418, + "step": 6470 + }, + { + "epoch": 0.39, + "learning_rate": 1.391163350157519e-05, + "loss": 0.3248, + "step": 6471 + }, + { + "epoch": 0.39, + "learning_rate": 1.3909832682731752e-05, + "loss": 0.3329, + "step": 6472 + }, + { + "epoch": 0.39, + "learning_rate": 1.3908031714201621e-05, + "loss": 0.3206, + "step": 6473 + }, + { + "epoch": 0.39, + "learning_rate": 1.3906230596053745e-05, + "loss": 0.3218, + "step": 6474 + }, + { + "epoch": 0.39, + "learning_rate": 1.3904429328357078e-05, + "loss": 0.3468, + "step": 6475 + }, + { + "epoch": 0.39, + "learning_rate": 1.3902627911180586e-05, + "loss": 0.3326, + "step": 6476 + }, + { + "epoch": 0.39, + "learning_rate": 1.3900826344593229e-05, + "loss": 0.3365, + "step": 6477 + }, + { + "epoch": 0.39, + "learning_rate": 1.3899024628663983e-05, + "loss": 0.3302, + "step": 6478 + }, + { + "epoch": 0.39, + "learning_rate": 1.3897222763461825e-05, + "loss": 0.3278, + "step": 6479 + }, + { + "epoch": 0.39, + "learning_rate": 1.3895420749055739e-05, + "loss": 0.3064, + "step": 6480 + }, + { + "epoch": 0.39, + "learning_rate": 1.389361858551472e-05, + "loss": 0.3451, + "step": 6481 + }, + { + "epoch": 0.39, + "learning_rate": 1.389181627290775e-05, + "loss": 0.3449, + "step": 6482 + }, + { + "epoch": 0.39, + "learning_rate": 1.3890013811303842e-05, + "loss": 0.3218, + "step": 6483 + }, + { + "epoch": 0.39, + "learning_rate": 1.3888211200772e-05, + "loss": 0.3178, + "step": 6484 + }, + { + "epoch": 0.39, + "learning_rate": 1.3886408441381234e-05, + "loss": 0.3158, + "step": 6485 + }, + { + "epoch": 0.39, + "learning_rate": 1.3884605533200563e-05, + "loss": 0.3624, + "step": 6486 + }, + { + "epoch": 0.39, + "learning_rate": 1.388280247629901e-05, + "loss": 0.329, + "step": 6487 + }, + { + "epoch": 0.39, + "learning_rate": 1.3880999270745608e-05, + "loss": 0.3393, + "step": 6488 + }, + { + "epoch": 0.39, + "learning_rate": 1.387919591660939e-05, + "loss": 0.341, + "step": 6489 + }, + { + "epoch": 0.39, + "learning_rate": 1.3877392413959393e-05, + "loss": 0.3251, + "step": 6490 + }, + { + "epoch": 0.39, + "learning_rate": 1.387558876286467e-05, + "loss": 0.305, + "step": 6491 + }, + { + "epoch": 0.39, + "learning_rate": 1.3873784963394268e-05, + "loss": 0.3097, + "step": 6492 + }, + { + "epoch": 0.39, + "learning_rate": 1.3871981015617248e-05, + "loss": 0.3136, + "step": 6493 + }, + { + "epoch": 0.39, + "learning_rate": 1.3870176919602674e-05, + "loss": 0.3312, + "step": 6494 + }, + { + "epoch": 0.39, + "learning_rate": 1.3868372675419611e-05, + "loss": 0.3179, + "step": 6495 + }, + { + "epoch": 0.39, + "learning_rate": 1.3866568283137139e-05, + "loss": 0.3309, + "step": 6496 + }, + { + "epoch": 0.39, + "learning_rate": 1.3864763742824334e-05, + "loss": 0.2906, + "step": 6497 + }, + { + "epoch": 0.39, + "learning_rate": 1.3862959054550286e-05, + "loss": 0.2946, + "step": 6498 + }, + { + "epoch": 0.39, + "learning_rate": 1.3861154218384087e-05, + "loss": 0.3103, + "step": 6499 + }, + { + "epoch": 0.39, + "learning_rate": 1.385934923439483e-05, + "loss": 0.3424, + "step": 6500 + }, + { + "epoch": 0.39, + "learning_rate": 1.3857544102651624e-05, + "loss": 0.3532, + "step": 6501 + }, + { + "epoch": 0.39, + "learning_rate": 1.3855738823223575e-05, + "loss": 0.3119, + "step": 6502 + }, + { + "epoch": 0.39, + "learning_rate": 1.3853933396179798e-05, + "loss": 0.353, + "step": 6503 + }, + { + "epoch": 0.39, + "learning_rate": 1.385212782158941e-05, + "loss": 0.3198, + "step": 6504 + }, + { + "epoch": 0.39, + "learning_rate": 1.385032209952154e-05, + "loss": 0.3474, + "step": 6505 + }, + { + "epoch": 0.39, + "learning_rate": 1.3848516230045324e-05, + "loss": 0.32, + "step": 6506 + }, + { + "epoch": 0.39, + "learning_rate": 1.3846710213229889e-05, + "loss": 0.3181, + "step": 6507 + }, + { + "epoch": 0.39, + "learning_rate": 1.3844904049144385e-05, + "loss": 0.3524, + "step": 6508 + }, + { + "epoch": 0.39, + "learning_rate": 1.384309773785796e-05, + "loss": 0.3242, + "step": 6509 + }, + { + "epoch": 0.39, + "learning_rate": 1.3841291279439769e-05, + "loss": 0.3171, + "step": 6510 + }, + { + "epoch": 0.39, + "learning_rate": 1.3839484673958966e-05, + "loss": 0.3223, + "step": 6511 + }, + { + "epoch": 0.39, + "learning_rate": 1.3837677921484721e-05, + "loss": 0.326, + "step": 6512 + }, + { + "epoch": 0.39, + "learning_rate": 1.3835871022086203e-05, + "loss": 0.3334, + "step": 6513 + }, + { + "epoch": 0.39, + "learning_rate": 1.3834063975832588e-05, + "loss": 0.3424, + "step": 6514 + }, + { + "epoch": 0.39, + "learning_rate": 1.3832256782793058e-05, + "loss": 0.3541, + "step": 6515 + }, + { + "epoch": 0.39, + "learning_rate": 1.3830449443036807e-05, + "loss": 0.3214, + "step": 6516 + }, + { + "epoch": 0.39, + "learning_rate": 1.3828641956633018e-05, + "loss": 0.3242, + "step": 6517 + }, + { + "epoch": 0.39, + "learning_rate": 1.3826834323650899e-05, + "loss": 0.32, + "step": 6518 + }, + { + "epoch": 0.39, + "learning_rate": 1.3825026544159653e-05, + "loss": 0.3318, + "step": 6519 + }, + { + "epoch": 0.39, + "learning_rate": 1.3823218618228485e-05, + "loss": 0.3128, + "step": 6520 + }, + { + "epoch": 0.39, + "learning_rate": 1.3821410545926614e-05, + "loss": 0.3318, + "step": 6521 + }, + { + "epoch": 0.39, + "learning_rate": 1.3819602327323263e-05, + "loss": 0.3013, + "step": 6522 + }, + { + "epoch": 0.39, + "learning_rate": 1.381779396248766e-05, + "loss": 0.3415, + "step": 6523 + }, + { + "epoch": 0.39, + "learning_rate": 1.3815985451489036e-05, + "loss": 0.3126, + "step": 6524 + }, + { + "epoch": 0.39, + "learning_rate": 1.3814176794396626e-05, + "loss": 0.3881, + "step": 6525 + }, + { + "epoch": 0.39, + "learning_rate": 1.3812367991279678e-05, + "loss": 0.3301, + "step": 6526 + }, + { + "epoch": 0.39, + "learning_rate": 1.381055904220744e-05, + "loss": 0.3147, + "step": 6527 + }, + { + "epoch": 0.39, + "learning_rate": 1.3808749947249167e-05, + "loss": 0.3147, + "step": 6528 + }, + { + "epoch": 0.39, + "learning_rate": 1.380694070647412e-05, + "loss": 0.3153, + "step": 6529 + }, + { + "epoch": 0.39, + "learning_rate": 1.3805131319951565e-05, + "loss": 0.3457, + "step": 6530 + }, + { + "epoch": 0.39, + "learning_rate": 1.3803321787750776e-05, + "loss": 0.3426, + "step": 6531 + }, + { + "epoch": 0.39, + "learning_rate": 1.3801512109941024e-05, + "loss": 0.3393, + "step": 6532 + }, + { + "epoch": 0.39, + "learning_rate": 1.3799702286591599e-05, + "loss": 0.3337, + "step": 6533 + }, + { + "epoch": 0.39, + "learning_rate": 1.3797892317771786e-05, + "loss": 0.3331, + "step": 6534 + }, + { + "epoch": 0.39, + "learning_rate": 1.379608220355088e-05, + "loss": 0.3128, + "step": 6535 + }, + { + "epoch": 0.39, + "learning_rate": 1.379427194399818e-05, + "loss": 0.2786, + "step": 6536 + }, + { + "epoch": 0.39, + "learning_rate": 1.379246153918299e-05, + "loss": 0.3277, + "step": 6537 + }, + { + "epoch": 0.39, + "learning_rate": 1.3790650989174625e-05, + "loss": 0.346, + "step": 6538 + }, + { + "epoch": 0.4, + "learning_rate": 1.3788840294042399e-05, + "loss": 0.3089, + "step": 6539 + }, + { + "epoch": 0.4, + "learning_rate": 1.3787029453855631e-05, + "loss": 0.3652, + "step": 6540 + }, + { + "epoch": 0.4, + "learning_rate": 1.3785218468683654e-05, + "loss": 0.3253, + "step": 6541 + }, + { + "epoch": 0.4, + "learning_rate": 1.3783407338595796e-05, + "loss": 0.2981, + "step": 6542 + }, + { + "epoch": 0.4, + "learning_rate": 1.3781596063661398e-05, + "loss": 0.3167, + "step": 6543 + }, + { + "epoch": 0.4, + "learning_rate": 1.3779784643949803e-05, + "loss": 0.3446, + "step": 6544 + }, + { + "epoch": 0.4, + "learning_rate": 1.3777973079530362e-05, + "loss": 0.3323, + "step": 6545 + }, + { + "epoch": 0.4, + "learning_rate": 1.377616137047243e-05, + "loss": 0.3259, + "step": 6546 + }, + { + "epoch": 0.4, + "learning_rate": 1.3774349516845364e-05, + "loss": 0.3184, + "step": 6547 + }, + { + "epoch": 0.4, + "learning_rate": 1.3772537518718539e-05, + "loss": 0.3301, + "step": 6548 + }, + { + "epoch": 0.4, + "learning_rate": 1.3770725376161316e-05, + "loss": 0.3421, + "step": 6549 + }, + { + "epoch": 0.4, + "learning_rate": 1.3768913089243078e-05, + "loss": 0.3181, + "step": 6550 + }, + { + "epoch": 0.4, + "learning_rate": 1.3767100658033207e-05, + "loss": 0.3566, + "step": 6551 + }, + { + "epoch": 0.4, + "learning_rate": 1.3765288082601093e-05, + "loss": 0.3205, + "step": 6552 + }, + { + "epoch": 0.4, + "learning_rate": 1.376347536301613e-05, + "loss": 0.3242, + "step": 6553 + }, + { + "epoch": 0.4, + "learning_rate": 1.3761662499347715e-05, + "loss": 0.3108, + "step": 6554 + }, + { + "epoch": 0.4, + "learning_rate": 1.3759849491665252e-05, + "loss": 0.2979, + "step": 6555 + }, + { + "epoch": 0.4, + "learning_rate": 1.3758036340038152e-05, + "loss": 0.2868, + "step": 6556 + }, + { + "epoch": 0.4, + "learning_rate": 1.3756223044535835e-05, + "loss": 0.3329, + "step": 6557 + }, + { + "epoch": 0.4, + "learning_rate": 1.3754409605227719e-05, + "loss": 0.3368, + "step": 6558 + }, + { + "epoch": 0.4, + "learning_rate": 1.375259602218323e-05, + "loss": 0.3128, + "step": 6559 + }, + { + "epoch": 0.4, + "learning_rate": 1.3750782295471804e-05, + "loss": 0.3129, + "step": 6560 + }, + { + "epoch": 0.4, + "learning_rate": 1.3748968425162878e-05, + "loss": 0.3499, + "step": 6561 + }, + { + "epoch": 0.4, + "learning_rate": 1.374715441132589e-05, + "loss": 0.2977, + "step": 6562 + }, + { + "epoch": 0.4, + "learning_rate": 1.3745340254030297e-05, + "loss": 0.3371, + "step": 6563 + }, + { + "epoch": 0.4, + "learning_rate": 1.3743525953345548e-05, + "loss": 0.3192, + "step": 6564 + }, + { + "epoch": 0.4, + "learning_rate": 1.3741711509341105e-05, + "loss": 0.3518, + "step": 6565 + }, + { + "epoch": 0.4, + "learning_rate": 1.3739896922086434e-05, + "loss": 0.3158, + "step": 6566 + }, + { + "epoch": 0.4, + "learning_rate": 1.3738082191651002e-05, + "loss": 0.3602, + "step": 6567 + }, + { + "epoch": 0.4, + "learning_rate": 1.3736267318104291e-05, + "loss": 0.3216, + "step": 6568 + }, + { + "epoch": 0.4, + "learning_rate": 1.3734452301515778e-05, + "loss": 0.3323, + "step": 6569 + }, + { + "epoch": 0.4, + "learning_rate": 1.3732637141954952e-05, + "loss": 0.3178, + "step": 6570 + }, + { + "epoch": 0.4, + "learning_rate": 1.3730821839491308e-05, + "loss": 0.2949, + "step": 6571 + }, + { + "epoch": 0.4, + "learning_rate": 1.3729006394194343e-05, + "loss": 0.3057, + "step": 6572 + }, + { + "epoch": 0.4, + "learning_rate": 1.3727190806133559e-05, + "loss": 0.3376, + "step": 6573 + }, + { + "epoch": 0.4, + "learning_rate": 1.3725375075378469e-05, + "loss": 0.3108, + "step": 6574 + }, + { + "epoch": 0.4, + "learning_rate": 1.3723559201998582e-05, + "loss": 0.3186, + "step": 6575 + }, + { + "epoch": 0.4, + "learning_rate": 1.3721743186063423e-05, + "loss": 0.3415, + "step": 6576 + }, + { + "epoch": 0.4, + "learning_rate": 1.3719927027642515e-05, + "loss": 0.3329, + "step": 6577 + }, + { + "epoch": 0.4, + "learning_rate": 1.3718110726805391e-05, + "loss": 0.3447, + "step": 6578 + }, + { + "epoch": 0.4, + "learning_rate": 1.3716294283621587e-05, + "loss": 0.3454, + "step": 6579 + }, + { + "epoch": 0.4, + "learning_rate": 1.3714477698160641e-05, + "loss": 0.3371, + "step": 6580 + }, + { + "epoch": 0.4, + "learning_rate": 1.3712660970492108e-05, + "loss": 0.3184, + "step": 6581 + }, + { + "epoch": 0.4, + "learning_rate": 1.3710844100685534e-05, + "loss": 0.3489, + "step": 6582 + }, + { + "epoch": 0.4, + "learning_rate": 1.3709027088810482e-05, + "loss": 0.3516, + "step": 6583 + }, + { + "epoch": 0.4, + "learning_rate": 1.3707209934936511e-05, + "loss": 0.3312, + "step": 6584 + }, + { + "epoch": 0.4, + "learning_rate": 1.3705392639133195e-05, + "loss": 0.3072, + "step": 6585 + }, + { + "epoch": 0.4, + "learning_rate": 1.3703575201470106e-05, + "loss": 0.3029, + "step": 6586 + }, + { + "epoch": 0.4, + "learning_rate": 1.3701757622016821e-05, + "loss": 0.3164, + "step": 6587 + }, + { + "epoch": 0.4, + "learning_rate": 1.3699939900842934e-05, + "loss": 0.3228, + "step": 6588 + }, + { + "epoch": 0.4, + "learning_rate": 1.3698122038018027e-05, + "loss": 0.3125, + "step": 6589 + }, + { + "epoch": 0.4, + "learning_rate": 1.3696304033611701e-05, + "loss": 0.3184, + "step": 6590 + }, + { + "epoch": 0.4, + "learning_rate": 1.3694485887693557e-05, + "loss": 0.3251, + "step": 6591 + }, + { + "epoch": 0.4, + "learning_rate": 1.36926676003332e-05, + "loss": 0.3384, + "step": 6592 + }, + { + "epoch": 0.4, + "learning_rate": 1.3690849171600247e-05, + "loss": 0.315, + "step": 6593 + }, + { + "epoch": 0.4, + "learning_rate": 1.3689030601564315e-05, + "loss": 0.3237, + "step": 6594 + }, + { + "epoch": 0.4, + "learning_rate": 1.3687211890295023e-05, + "loss": 0.3039, + "step": 6595 + }, + { + "epoch": 0.4, + "learning_rate": 1.3685393037862002e-05, + "loss": 0.2999, + "step": 6596 + }, + { + "epoch": 0.4, + "learning_rate": 1.3683574044334886e-05, + "loss": 0.3518, + "step": 6597 + }, + { + "epoch": 0.4, + "learning_rate": 1.3681754909783317e-05, + "loss": 0.3276, + "step": 6598 + }, + { + "epoch": 0.4, + "learning_rate": 1.367993563427694e-05, + "loss": 0.3546, + "step": 6599 + }, + { + "epoch": 0.4, + "learning_rate": 1.36781162178854e-05, + "loss": 0.3407, + "step": 6600 + }, + { + "epoch": 0.4, + "learning_rate": 1.3676296660678359e-05, + "loss": 0.3454, + "step": 6601 + }, + { + "epoch": 0.4, + "learning_rate": 1.3674476962725473e-05, + "loss": 0.3326, + "step": 6602 + }, + { + "epoch": 0.4, + "learning_rate": 1.3672657124096413e-05, + "loss": 0.3337, + "step": 6603 + }, + { + "epoch": 0.4, + "learning_rate": 1.3670837144860851e-05, + "loss": 0.3337, + "step": 6604 + }, + { + "epoch": 0.4, + "learning_rate": 1.3669017025088457e-05, + "loss": 0.3591, + "step": 6605 + }, + { + "epoch": 0.4, + "learning_rate": 1.3667196764848924e-05, + "loss": 0.3043, + "step": 6606 + }, + { + "epoch": 0.4, + "learning_rate": 1.3665376364211933e-05, + "loss": 0.3401, + "step": 6607 + }, + { + "epoch": 0.4, + "learning_rate": 1.366355582324718e-05, + "loss": 0.3348, + "step": 6608 + }, + { + "epoch": 0.4, + "learning_rate": 1.3661735142024366e-05, + "loss": 0.3276, + "step": 6609 + }, + { + "epoch": 0.4, + "learning_rate": 1.3659914320613188e-05, + "loss": 0.3147, + "step": 6610 + }, + { + "epoch": 0.4, + "learning_rate": 1.3658093359083365e-05, + "loss": 0.3273, + "step": 6611 + }, + { + "epoch": 0.4, + "learning_rate": 1.3656272257504604e-05, + "loss": 0.3284, + "step": 6612 + }, + { + "epoch": 0.4, + "learning_rate": 1.365445101594663e-05, + "loss": 0.3429, + "step": 6613 + }, + { + "epoch": 0.4, + "learning_rate": 1.3652629634479167e-05, + "loss": 0.3675, + "step": 6614 + }, + { + "epoch": 0.4, + "learning_rate": 1.3650808113171945e-05, + "loss": 0.3451, + "step": 6615 + }, + { + "epoch": 0.4, + "learning_rate": 1.36489864520947e-05, + "loss": 0.3062, + "step": 6616 + }, + { + "epoch": 0.4, + "learning_rate": 1.3647164651317178e-05, + "loss": 0.3479, + "step": 6617 + }, + { + "epoch": 0.4, + "learning_rate": 1.3645342710909124e-05, + "loss": 0.3178, + "step": 6618 + }, + { + "epoch": 0.4, + "learning_rate": 1.3643520630940287e-05, + "loss": 0.31, + "step": 6619 + }, + { + "epoch": 0.4, + "learning_rate": 1.3641698411480428e-05, + "loss": 0.2967, + "step": 6620 + }, + { + "epoch": 0.4, + "learning_rate": 1.363987605259931e-05, + "loss": 0.3379, + "step": 6621 + }, + { + "epoch": 0.4, + "learning_rate": 1.3638053554366702e-05, + "loss": 0.3396, + "step": 6622 + }, + { + "epoch": 0.4, + "learning_rate": 1.3636230916852373e-05, + "loss": 0.344, + "step": 6623 + }, + { + "epoch": 0.4, + "learning_rate": 1.363440814012611e-05, + "loss": 0.3333, + "step": 6624 + }, + { + "epoch": 0.4, + "learning_rate": 1.363258522425769e-05, + "loss": 0.3504, + "step": 6625 + }, + { + "epoch": 0.4, + "learning_rate": 1.3630762169316908e-05, + "loss": 0.3234, + "step": 6626 + }, + { + "epoch": 0.4, + "learning_rate": 1.3628938975373556e-05, + "loss": 0.3167, + "step": 6627 + }, + { + "epoch": 0.4, + "learning_rate": 1.3627115642497436e-05, + "loss": 0.3326, + "step": 6628 + }, + { + "epoch": 0.4, + "learning_rate": 1.3625292170758355e-05, + "loss": 0.2938, + "step": 6629 + }, + { + "epoch": 0.4, + "learning_rate": 1.3623468560226121e-05, + "loss": 0.3209, + "step": 6630 + }, + { + "epoch": 0.4, + "learning_rate": 1.362164481097055e-05, + "loss": 0.3278, + "step": 6631 + }, + { + "epoch": 0.4, + "learning_rate": 1.3619820923061467e-05, + "loss": 0.3499, + "step": 6632 + }, + { + "epoch": 0.4, + "learning_rate": 1.3617996896568694e-05, + "loss": 0.3237, + "step": 6633 + }, + { + "epoch": 0.4, + "learning_rate": 1.361617273156207e-05, + "loss": 0.3371, + "step": 6634 + }, + { + "epoch": 0.4, + "learning_rate": 1.3614348428111428e-05, + "loss": 0.3184, + "step": 6635 + }, + { + "epoch": 0.4, + "learning_rate": 1.3612523986286612e-05, + "loss": 0.3387, + "step": 6636 + }, + { + "epoch": 0.4, + "learning_rate": 1.3610699406157467e-05, + "loss": 0.3259, + "step": 6637 + }, + { + "epoch": 0.4, + "learning_rate": 1.3608874687793853e-05, + "loss": 0.3404, + "step": 6638 + }, + { + "epoch": 0.4, + "learning_rate": 1.3607049831265627e-05, + "loss": 0.3231, + "step": 6639 + }, + { + "epoch": 0.4, + "learning_rate": 1.3605224836642645e-05, + "loss": 0.305, + "step": 6640 + }, + { + "epoch": 0.4, + "learning_rate": 1.3603399703994788e-05, + "loss": 0.2974, + "step": 6641 + }, + { + "epoch": 0.4, + "learning_rate": 1.3601574433391922e-05, + "loss": 0.3164, + "step": 6642 + }, + { + "epoch": 0.4, + "learning_rate": 1.3599749024903933e-05, + "loss": 0.3488, + "step": 6643 + }, + { + "epoch": 0.4, + "learning_rate": 1.3597923478600704e-05, + "loss": 0.3135, + "step": 6644 + }, + { + "epoch": 0.4, + "learning_rate": 1.359609779455212e-05, + "loss": 0.3234, + "step": 6645 + }, + { + "epoch": 0.4, + "learning_rate": 1.3594271972828086e-05, + "loss": 0.3214, + "step": 6646 + }, + { + "epoch": 0.4, + "learning_rate": 1.3592446013498495e-05, + "loss": 0.3259, + "step": 6647 + }, + { + "epoch": 0.4, + "learning_rate": 1.359061991663326e-05, + "loss": 0.3061, + "step": 6648 + }, + { + "epoch": 0.4, + "learning_rate": 1.3588793682302287e-05, + "loss": 0.3234, + "step": 6649 + }, + { + "epoch": 0.4, + "learning_rate": 1.3586967310575496e-05, + "loss": 0.3401, + "step": 6650 + }, + { + "epoch": 0.4, + "learning_rate": 1.358514080152281e-05, + "loss": 0.3214, + "step": 6651 + }, + { + "epoch": 0.4, + "learning_rate": 1.3583314155214152e-05, + "loss": 0.3251, + "step": 6652 + }, + { + "epoch": 0.4, + "learning_rate": 1.3581487371719457e-05, + "loss": 0.3421, + "step": 6653 + }, + { + "epoch": 0.4, + "learning_rate": 1.3579660451108665e-05, + "loss": 0.3228, + "step": 6654 + }, + { + "epoch": 0.4, + "learning_rate": 1.3577833393451716e-05, + "loss": 0.3234, + "step": 6655 + }, + { + "epoch": 0.4, + "learning_rate": 1.3576006198818558e-05, + "loss": 0.332, + "step": 6656 + }, + { + "epoch": 0.4, + "learning_rate": 1.3574178867279147e-05, + "loss": 0.3318, + "step": 6657 + }, + { + "epoch": 0.4, + "learning_rate": 1.3572351398903441e-05, + "loss": 0.3175, + "step": 6658 + }, + { + "epoch": 0.4, + "learning_rate": 1.3570523793761405e-05, + "loss": 0.3371, + "step": 6659 + }, + { + "epoch": 0.4, + "learning_rate": 1.3568696051923005e-05, + "loss": 0.3184, + "step": 6660 + }, + { + "epoch": 0.4, + "learning_rate": 1.3566868173458217e-05, + "loss": 0.3184, + "step": 6661 + }, + { + "epoch": 0.4, + "learning_rate": 1.3565040158437022e-05, + "loss": 0.3301, + "step": 6662 + }, + { + "epoch": 0.4, + "learning_rate": 1.3563212006929404e-05, + "loss": 0.3107, + "step": 6663 + }, + { + "epoch": 0.4, + "learning_rate": 1.3561383719005354e-05, + "loss": 0.3407, + "step": 6664 + }, + { + "epoch": 0.4, + "learning_rate": 1.3559555294734867e-05, + "loss": 0.3546, + "step": 6665 + }, + { + "epoch": 0.4, + "learning_rate": 1.3557726734187943e-05, + "loss": 0.3373, + "step": 6666 + }, + { + "epoch": 0.4, + "learning_rate": 1.3555898037434587e-05, + "loss": 0.2903, + "step": 6667 + }, + { + "epoch": 0.4, + "learning_rate": 1.3554069204544816e-05, + "loss": 0.3242, + "step": 6668 + }, + { + "epoch": 0.4, + "learning_rate": 1.3552240235588636e-05, + "loss": 0.3089, + "step": 6669 + }, + { + "epoch": 0.4, + "learning_rate": 1.3550411130636079e-05, + "loss": 0.3301, + "step": 6670 + }, + { + "epoch": 0.4, + "learning_rate": 1.3548581889757164e-05, + "loss": 0.3306, + "step": 6671 + }, + { + "epoch": 0.4, + "learning_rate": 1.3546752513021924e-05, + "loss": 0.3373, + "step": 6672 + }, + { + "epoch": 0.4, + "learning_rate": 1.3544923000500401e-05, + "loss": 0.3574, + "step": 6673 + }, + { + "epoch": 0.4, + "learning_rate": 1.3543093352262636e-05, + "loss": 0.3532, + "step": 6674 + }, + { + "epoch": 0.4, + "learning_rate": 1.3541263568378668e-05, + "loss": 0.3175, + "step": 6675 + }, + { + "epoch": 0.4, + "learning_rate": 1.3539433648918561e-05, + "loss": 0.3239, + "step": 6676 + }, + { + "epoch": 0.4, + "learning_rate": 1.3537603593952366e-05, + "loss": 0.3479, + "step": 6677 + }, + { + "epoch": 0.4, + "learning_rate": 1.353577340355015e-05, + "loss": 0.3248, + "step": 6678 + }, + { + "epoch": 0.4, + "learning_rate": 1.3533943077781982e-05, + "loss": 0.3083, + "step": 6679 + }, + { + "epoch": 0.4, + "learning_rate": 1.353211261671793e-05, + "loss": 0.3145, + "step": 6680 + }, + { + "epoch": 0.4, + "learning_rate": 1.3530282020428075e-05, + "loss": 0.3153, + "step": 6681 + }, + { + "epoch": 0.4, + "learning_rate": 1.3528451288982502e-05, + "loss": 0.3387, + "step": 6682 + }, + { + "epoch": 0.4, + "learning_rate": 1.3526620422451303e-05, + "loss": 0.3145, + "step": 6683 + }, + { + "epoch": 0.4, + "learning_rate": 1.3524789420904562e-05, + "loss": 0.3062, + "step": 6684 + }, + { + "epoch": 0.4, + "learning_rate": 1.352295828441239e-05, + "loss": 0.3027, + "step": 6685 + }, + { + "epoch": 0.4, + "learning_rate": 1.3521127013044884e-05, + "loss": 0.3047, + "step": 6686 + }, + { + "epoch": 0.4, + "learning_rate": 1.3519295606872157e-05, + "loss": 0.3309, + "step": 6687 + }, + { + "epoch": 0.4, + "learning_rate": 1.3517464065964321e-05, + "loss": 0.3329, + "step": 6688 + }, + { + "epoch": 0.4, + "learning_rate": 1.35156323903915e-05, + "loss": 0.3223, + "step": 6689 + }, + { + "epoch": 0.4, + "learning_rate": 1.3513800580223814e-05, + "loss": 0.3428, + "step": 6690 + }, + { + "epoch": 0.4, + "learning_rate": 1.3511968635531398e-05, + "loss": 0.3184, + "step": 6691 + }, + { + "epoch": 0.4, + "learning_rate": 1.3510136556384384e-05, + "loss": 0.3237, + "step": 6692 + }, + { + "epoch": 0.4, + "learning_rate": 1.3508304342852916e-05, + "loss": 0.3379, + "step": 6693 + }, + { + "epoch": 0.4, + "learning_rate": 1.3506471995007137e-05, + "loss": 0.3343, + "step": 6694 + }, + { + "epoch": 0.4, + "learning_rate": 1.3504639512917199e-05, + "loss": 0.3064, + "step": 6695 + }, + { + "epoch": 0.4, + "learning_rate": 1.3502806896653257e-05, + "loss": 0.3481, + "step": 6696 + }, + { + "epoch": 0.4, + "learning_rate": 1.3500974146285476e-05, + "loss": 0.3136, + "step": 6697 + }, + { + "epoch": 0.4, + "learning_rate": 1.3499141261884015e-05, + "loss": 0.3167, + "step": 6698 + }, + { + "epoch": 0.4, + "learning_rate": 1.3497308243519049e-05, + "loss": 0.3535, + "step": 6699 + }, + { + "epoch": 0.4, + "learning_rate": 1.3495475091260758e-05, + "loss": 0.3298, + "step": 6700 + }, + { + "epoch": 0.4, + "learning_rate": 1.3493641805179318e-05, + "loss": 0.3364, + "step": 6701 + }, + { + "epoch": 0.4, + "learning_rate": 1.3491808385344917e-05, + "loss": 0.3396, + "step": 6702 + }, + { + "epoch": 0.4, + "learning_rate": 1.3489974831827752e-05, + "loss": 0.3535, + "step": 6703 + }, + { + "epoch": 0.41, + "learning_rate": 1.3488141144698014e-05, + "loss": 0.3468, + "step": 6704 + }, + { + "epoch": 0.41, + "learning_rate": 1.3486307324025909e-05, + "loss": 0.3105, + "step": 6705 + }, + { + "epoch": 0.41, + "learning_rate": 1.3484473369881643e-05, + "loss": 0.3052, + "step": 6706 + }, + { + "epoch": 0.41, + "learning_rate": 1.3482639282335423e-05, + "loss": 0.2979, + "step": 6707 + }, + { + "epoch": 0.41, + "learning_rate": 1.3480805061457478e-05, + "loss": 0.3301, + "step": 6708 + }, + { + "epoch": 0.41, + "learning_rate": 1.3478970707318023e-05, + "loss": 0.327, + "step": 6709 + }, + { + "epoch": 0.41, + "learning_rate": 1.347713621998728e-05, + "loss": 0.3435, + "step": 6710 + }, + { + "epoch": 0.41, + "learning_rate": 1.3475301599535496e-05, + "loss": 0.326, + "step": 6711 + }, + { + "epoch": 0.41, + "learning_rate": 1.3473466846032897e-05, + "loss": 0.3172, + "step": 6712 + }, + { + "epoch": 0.41, + "learning_rate": 1.347163195954973e-05, + "loss": 0.3337, + "step": 6713 + }, + { + "epoch": 0.41, + "learning_rate": 1.3469796940156248e-05, + "loss": 0.3205, + "step": 6714 + }, + { + "epoch": 0.41, + "learning_rate": 1.3467961787922696e-05, + "loss": 0.3142, + "step": 6715 + }, + { + "epoch": 0.41, + "learning_rate": 1.3466126502919337e-05, + "loss": 0.3345, + "step": 6716 + }, + { + "epoch": 0.41, + "learning_rate": 1.346429108521643e-05, + "loss": 0.3097, + "step": 6717 + }, + { + "epoch": 0.41, + "learning_rate": 1.3462455534884247e-05, + "loss": 0.3256, + "step": 6718 + }, + { + "epoch": 0.41, + "learning_rate": 1.3460619851993062e-05, + "loss": 0.3541, + "step": 6719 + }, + { + "epoch": 0.41, + "learning_rate": 1.345878403661315e-05, + "loss": 0.3085, + "step": 6720 + }, + { + "epoch": 0.41, + "learning_rate": 1.34569480888148e-05, + "loss": 0.3454, + "step": 6721 + }, + { + "epoch": 0.41, + "learning_rate": 1.3455112008668297e-05, + "loss": 0.3119, + "step": 6722 + }, + { + "epoch": 0.41, + "learning_rate": 1.3453275796243934e-05, + "loss": 0.3066, + "step": 6723 + }, + { + "epoch": 0.41, + "learning_rate": 1.3451439451612012e-05, + "loss": 0.328, + "step": 6724 + }, + { + "epoch": 0.41, + "learning_rate": 1.3449602974842832e-05, + "loss": 0.3485, + "step": 6725 + }, + { + "epoch": 0.41, + "learning_rate": 1.344776636600671e-05, + "loss": 0.3345, + "step": 6726 + }, + { + "epoch": 0.41, + "learning_rate": 1.3445929625173948e-05, + "loss": 0.2952, + "step": 6727 + }, + { + "epoch": 0.41, + "learning_rate": 1.3444092752414876e-05, + "loss": 0.3496, + "step": 6728 + }, + { + "epoch": 0.41, + "learning_rate": 1.3442255747799812e-05, + "loss": 0.3398, + "step": 6729 + }, + { + "epoch": 0.41, + "learning_rate": 1.3440418611399088e-05, + "loss": 0.3125, + "step": 6730 + }, + { + "epoch": 0.41, + "learning_rate": 1.3438581343283041e-05, + "loss": 0.3217, + "step": 6731 + }, + { + "epoch": 0.41, + "learning_rate": 1.3436743943522001e-05, + "loss": 0.298, + "step": 6732 + }, + { + "epoch": 0.41, + "learning_rate": 1.3434906412186321e-05, + "loss": 0.3161, + "step": 6733 + }, + { + "epoch": 0.41, + "learning_rate": 1.3433068749346344e-05, + "loss": 0.3415, + "step": 6734 + }, + { + "epoch": 0.41, + "learning_rate": 1.3431230955072431e-05, + "loss": 0.3382, + "step": 6735 + }, + { + "epoch": 0.41, + "learning_rate": 1.3429393029434938e-05, + "loss": 0.2949, + "step": 6736 + }, + { + "epoch": 0.41, + "learning_rate": 1.3427554972504227e-05, + "loss": 0.3052, + "step": 6737 + }, + { + "epoch": 0.41, + "learning_rate": 1.3425716784350669e-05, + "loss": 0.3147, + "step": 6738 + }, + { + "epoch": 0.41, + "learning_rate": 1.3423878465044639e-05, + "loss": 0.3175, + "step": 6739 + }, + { + "epoch": 0.41, + "learning_rate": 1.3422040014656518e-05, + "loss": 0.3477, + "step": 6740 + }, + { + "epoch": 0.41, + "learning_rate": 1.342020143325669e-05, + "loss": 0.3312, + "step": 6741 + }, + { + "epoch": 0.41, + "learning_rate": 1.3418362720915541e-05, + "loss": 0.3041, + "step": 6742 + }, + { + "epoch": 0.41, + "learning_rate": 1.3416523877703469e-05, + "loss": 0.3203, + "step": 6743 + }, + { + "epoch": 0.41, + "learning_rate": 1.3414684903690873e-05, + "loss": 0.3267, + "step": 6744 + }, + { + "epoch": 0.41, + "learning_rate": 1.3412845798948155e-05, + "loss": 0.3044, + "step": 6745 + }, + { + "epoch": 0.41, + "learning_rate": 1.341100656354573e-05, + "loss": 0.3052, + "step": 6746 + }, + { + "epoch": 0.41, + "learning_rate": 1.3409167197554004e-05, + "loss": 0.3205, + "step": 6747 + }, + { + "epoch": 0.41, + "learning_rate": 1.3407327701043409e-05, + "loss": 0.3253, + "step": 6748 + }, + { + "epoch": 0.41, + "learning_rate": 1.3405488074084359e-05, + "loss": 0.3075, + "step": 6749 + }, + { + "epoch": 0.41, + "learning_rate": 1.3403648316747284e-05, + "loss": 0.2992, + "step": 6750 + }, + { + "epoch": 0.41, + "learning_rate": 1.3401808429102625e-05, + "loss": 0.3259, + "step": 6751 + }, + { + "epoch": 0.41, + "learning_rate": 1.3399968411220816e-05, + "loss": 0.3306, + "step": 6752 + }, + { + "epoch": 0.41, + "learning_rate": 1.3398128263172303e-05, + "loss": 0.3228, + "step": 6753 + }, + { + "epoch": 0.41, + "learning_rate": 1.3396287985027537e-05, + "loss": 0.2905, + "step": 6754 + }, + { + "epoch": 0.41, + "learning_rate": 1.3394447576856968e-05, + "loss": 0.3262, + "step": 6755 + }, + { + "epoch": 0.41, + "learning_rate": 1.3392607038731063e-05, + "loss": 0.3131, + "step": 6756 + }, + { + "epoch": 0.41, + "learning_rate": 1.3390766370720279e-05, + "loss": 0.3379, + "step": 6757 + }, + { + "epoch": 0.41, + "learning_rate": 1.3388925572895091e-05, + "loss": 0.3223, + "step": 6758 + }, + { + "epoch": 0.41, + "learning_rate": 1.3387084645325968e-05, + "loss": 0.3315, + "step": 6759 + }, + { + "epoch": 0.41, + "learning_rate": 1.3385243588083394e-05, + "loss": 0.3231, + "step": 6760 + }, + { + "epoch": 0.41, + "learning_rate": 1.338340240123785e-05, + "loss": 0.3111, + "step": 6761 + }, + { + "epoch": 0.41, + "learning_rate": 1.3381561084859827e-05, + "loss": 0.3376, + "step": 6762 + }, + { + "epoch": 0.41, + "learning_rate": 1.3379719639019821e-05, + "loss": 0.3323, + "step": 6763 + }, + { + "epoch": 0.41, + "learning_rate": 1.3377878063788328e-05, + "loss": 0.3496, + "step": 6764 + }, + { + "epoch": 0.41, + "learning_rate": 1.3376036359235854e-05, + "loss": 0.2885, + "step": 6765 + }, + { + "epoch": 0.41, + "learning_rate": 1.3374194525432906e-05, + "loss": 0.3373, + "step": 6766 + }, + { + "epoch": 0.41, + "learning_rate": 1.3372352562449997e-05, + "loss": 0.3477, + "step": 6767 + }, + { + "epoch": 0.41, + "learning_rate": 1.3370510470357654e-05, + "loss": 0.3569, + "step": 6768 + }, + { + "epoch": 0.41, + "learning_rate": 1.3368668249226389e-05, + "loss": 0.3143, + "step": 6769 + }, + { + "epoch": 0.41, + "learning_rate": 1.3366825899126741e-05, + "loss": 0.3326, + "step": 6770 + }, + { + "epoch": 0.41, + "learning_rate": 1.3364983420129238e-05, + "loss": 0.2853, + "step": 6771 + }, + { + "epoch": 0.41, + "learning_rate": 1.336314081230442e-05, + "loss": 0.3025, + "step": 6772 + }, + { + "epoch": 0.41, + "learning_rate": 1.3361298075722833e-05, + "loss": 0.3092, + "step": 6773 + }, + { + "epoch": 0.41, + "learning_rate": 1.335945521045502e-05, + "loss": 0.3549, + "step": 6774 + }, + { + "epoch": 0.41, + "learning_rate": 1.3357612216571542e-05, + "loss": 0.3211, + "step": 6775 + }, + { + "epoch": 0.41, + "learning_rate": 1.3355769094142954e-05, + "loss": 0.3248, + "step": 6776 + }, + { + "epoch": 0.41, + "learning_rate": 1.3353925843239815e-05, + "loss": 0.3471, + "step": 6777 + }, + { + "epoch": 0.41, + "learning_rate": 1.3352082463932701e-05, + "loss": 0.2995, + "step": 6778 + }, + { + "epoch": 0.41, + "learning_rate": 1.3350238956292181e-05, + "loss": 0.3033, + "step": 6779 + }, + { + "epoch": 0.41, + "learning_rate": 1.3348395320388832e-05, + "loss": 0.3359, + "step": 6780 + }, + { + "epoch": 0.41, + "learning_rate": 1.334655155629324e-05, + "loss": 0.3306, + "step": 6781 + }, + { + "epoch": 0.41, + "learning_rate": 1.3344707664075993e-05, + "loss": 0.3359, + "step": 6782 + }, + { + "epoch": 0.41, + "learning_rate": 1.3342863643807679e-05, + "loss": 0.3161, + "step": 6783 + }, + { + "epoch": 0.41, + "learning_rate": 1.33410194955589e-05, + "loss": 0.3092, + "step": 6784 + }, + { + "epoch": 0.41, + "learning_rate": 1.3339175219400257e-05, + "loss": 0.3329, + "step": 6785 + }, + { + "epoch": 0.41, + "learning_rate": 1.333733081540236e-05, + "loss": 0.3186, + "step": 6786 + }, + { + "epoch": 0.41, + "learning_rate": 1.333548628363582e-05, + "loss": 0.3345, + "step": 6787 + }, + { + "epoch": 0.41, + "learning_rate": 1.3333641624171251e-05, + "loss": 0.3306, + "step": 6788 + }, + { + "epoch": 0.41, + "learning_rate": 1.3331796837079281e-05, + "loss": 0.3262, + "step": 6789 + }, + { + "epoch": 0.41, + "learning_rate": 1.3329951922430533e-05, + "loss": 0.3446, + "step": 6790 + }, + { + "epoch": 0.41, + "learning_rate": 1.332810688029564e-05, + "loss": 0.3175, + "step": 6791 + }, + { + "epoch": 0.41, + "learning_rate": 1.3326261710745239e-05, + "loss": 0.3379, + "step": 6792 + }, + { + "epoch": 0.41, + "learning_rate": 1.3324416413849972e-05, + "loss": 0.3234, + "step": 6793 + }, + { + "epoch": 0.41, + "learning_rate": 1.3322570989680483e-05, + "loss": 0.3424, + "step": 6794 + }, + { + "epoch": 0.41, + "learning_rate": 1.3320725438307428e-05, + "loss": 0.3189, + "step": 6795 + }, + { + "epoch": 0.41, + "learning_rate": 1.3318879759801464e-05, + "loss": 0.3128, + "step": 6796 + }, + { + "epoch": 0.41, + "learning_rate": 1.3317033954233246e-05, + "loss": 0.3089, + "step": 6797 + }, + { + "epoch": 0.41, + "learning_rate": 1.3315188021673443e-05, + "loss": 0.3234, + "step": 6798 + }, + { + "epoch": 0.41, + "learning_rate": 1.331334196219273e-05, + "loss": 0.3248, + "step": 6799 + }, + { + "epoch": 0.41, + "learning_rate": 1.3311495775861776e-05, + "loss": 0.3345, + "step": 6800 + }, + { + "epoch": 0.41, + "learning_rate": 1.3309649462751266e-05, + "loss": 0.3089, + "step": 6801 + }, + { + "epoch": 0.41, + "learning_rate": 1.3307803022931884e-05, + "loss": 0.3566, + "step": 6802 + }, + { + "epoch": 0.41, + "learning_rate": 1.3305956456474321e-05, + "loss": 0.3315, + "step": 6803 + }, + { + "epoch": 0.41, + "learning_rate": 1.3304109763449269e-05, + "loss": 0.3326, + "step": 6804 + }, + { + "epoch": 0.41, + "learning_rate": 1.3302262943927435e-05, + "loss": 0.3304, + "step": 6805 + }, + { + "epoch": 0.41, + "learning_rate": 1.3300415997979519e-05, + "loss": 0.3351, + "step": 6806 + }, + { + "epoch": 0.41, + "learning_rate": 1.329856892567623e-05, + "loss": 0.3186, + "step": 6807 + }, + { + "epoch": 0.41, + "learning_rate": 1.3296721727088286e-05, + "loss": 0.3192, + "step": 6808 + }, + { + "epoch": 0.41, + "learning_rate": 1.3294874402286402e-05, + "loss": 0.3278, + "step": 6809 + }, + { + "epoch": 0.41, + "learning_rate": 1.3293026951341308e-05, + "loss": 0.2983, + "step": 6810 + }, + { + "epoch": 0.41, + "learning_rate": 1.3291179374323729e-05, + "loss": 0.3318, + "step": 6811 + }, + { + "epoch": 0.41, + "learning_rate": 1.3289331671304397e-05, + "loss": 0.317, + "step": 6812 + }, + { + "epoch": 0.41, + "learning_rate": 1.3287483842354057e-05, + "loss": 0.3359, + "step": 6813 + }, + { + "epoch": 0.41, + "learning_rate": 1.328563588754345e-05, + "loss": 0.3209, + "step": 6814 + }, + { + "epoch": 0.41, + "learning_rate": 1.3283787806943318e-05, + "loss": 0.351, + "step": 6815 + }, + { + "epoch": 0.41, + "learning_rate": 1.3281939600624428e-05, + "loss": 0.3359, + "step": 6816 + }, + { + "epoch": 0.41, + "learning_rate": 1.328009126865752e-05, + "loss": 0.3189, + "step": 6817 + }, + { + "epoch": 0.41, + "learning_rate": 1.3278242811113371e-05, + "loss": 0.3008, + "step": 6818 + }, + { + "epoch": 0.41, + "learning_rate": 1.3276394228062743e-05, + "loss": 0.3061, + "step": 6819 + }, + { + "epoch": 0.41, + "learning_rate": 1.3274545519576411e-05, + "loss": 0.3438, + "step": 6820 + }, + { + "epoch": 0.41, + "learning_rate": 1.327269668572515e-05, + "loss": 0.3209, + "step": 6821 + }, + { + "epoch": 0.41, + "learning_rate": 1.3270847726579741e-05, + "loss": 0.3334, + "step": 6822 + }, + { + "epoch": 0.41, + "learning_rate": 1.3268998642210976e-05, + "loss": 0.3443, + "step": 6823 + }, + { + "epoch": 0.41, + "learning_rate": 1.3267149432689638e-05, + "loss": 0.3382, + "step": 6824 + }, + { + "epoch": 0.41, + "learning_rate": 1.326530009808653e-05, + "loss": 0.3295, + "step": 6825 + }, + { + "epoch": 0.41, + "learning_rate": 1.3263450638472456e-05, + "loss": 0.3094, + "step": 6826 + }, + { + "epoch": 0.41, + "learning_rate": 1.3261601053918212e-05, + "loss": 0.3192, + "step": 6827 + }, + { + "epoch": 0.41, + "learning_rate": 1.3259751344494616e-05, + "loss": 0.3192, + "step": 6828 + }, + { + "epoch": 0.41, + "learning_rate": 1.3257901510272482e-05, + "loss": 0.3242, + "step": 6829 + }, + { + "epoch": 0.41, + "learning_rate": 1.325605155132263e-05, + "loss": 0.3488, + "step": 6830 + }, + { + "epoch": 0.41, + "learning_rate": 1.3254201467715888e-05, + "loss": 0.3451, + "step": 6831 + }, + { + "epoch": 0.41, + "learning_rate": 1.325235125952308e-05, + "loss": 0.3351, + "step": 6832 + }, + { + "epoch": 0.41, + "learning_rate": 1.3250500926815046e-05, + "loss": 0.3345, + "step": 6833 + }, + { + "epoch": 0.41, + "learning_rate": 1.324865046966262e-05, + "loss": 0.2974, + "step": 6834 + }, + { + "epoch": 0.41, + "learning_rate": 1.3246799888136655e-05, + "loss": 0.3336, + "step": 6835 + }, + { + "epoch": 0.41, + "learning_rate": 1.3244949182307989e-05, + "loss": 0.3265, + "step": 6836 + }, + { + "epoch": 0.41, + "learning_rate": 1.324309835224748e-05, + "loss": 0.3105, + "step": 6837 + }, + { + "epoch": 0.41, + "learning_rate": 1.3241247398025991e-05, + "loss": 0.3248, + "step": 6838 + }, + { + "epoch": 0.41, + "learning_rate": 1.323939631971438e-05, + "loss": 0.3337, + "step": 6839 + }, + { + "epoch": 0.41, + "learning_rate": 1.3237545117383516e-05, + "loss": 0.3251, + "step": 6840 + }, + { + "epoch": 0.41, + "learning_rate": 1.3235693791104273e-05, + "loss": 0.3223, + "step": 6841 + }, + { + "epoch": 0.41, + "learning_rate": 1.3233842340947528e-05, + "loss": 0.3122, + "step": 6842 + }, + { + "epoch": 0.41, + "learning_rate": 1.3231990766984162e-05, + "loss": 0.3318, + "step": 6843 + }, + { + "epoch": 0.41, + "learning_rate": 1.3230139069285061e-05, + "loss": 0.353, + "step": 6844 + }, + { + "epoch": 0.41, + "learning_rate": 1.322828724792112e-05, + "loss": 0.3004, + "step": 6845 + }, + { + "epoch": 0.41, + "learning_rate": 1.3226435302963236e-05, + "loss": 0.3714, + "step": 6846 + }, + { + "epoch": 0.41, + "learning_rate": 1.3224583234482305e-05, + "loss": 0.3191, + "step": 6847 + }, + { + "epoch": 0.41, + "learning_rate": 1.3222731042549234e-05, + "loss": 0.3535, + "step": 6848 + }, + { + "epoch": 0.41, + "learning_rate": 1.322087872723494e-05, + "loss": 0.3343, + "step": 6849 + }, + { + "epoch": 0.41, + "learning_rate": 1.3219026288610333e-05, + "loss": 0.3331, + "step": 6850 + }, + { + "epoch": 0.41, + "learning_rate": 1.3217173726746333e-05, + "loss": 0.308, + "step": 6851 + }, + { + "epoch": 0.41, + "learning_rate": 1.3215321041713861e-05, + "loss": 0.3139, + "step": 6852 + }, + { + "epoch": 0.41, + "learning_rate": 1.3213468233583857e-05, + "loss": 0.3027, + "step": 6853 + }, + { + "epoch": 0.41, + "learning_rate": 1.3211615302427244e-05, + "loss": 0.3071, + "step": 6854 + }, + { + "epoch": 0.41, + "learning_rate": 1.3209762248314967e-05, + "loss": 0.3273, + "step": 6855 + }, + { + "epoch": 0.41, + "learning_rate": 1.3207909071317971e-05, + "loss": 0.3343, + "step": 6856 + }, + { + "epoch": 0.41, + "learning_rate": 1.3206055771507198e-05, + "loss": 0.3384, + "step": 6857 + }, + { + "epoch": 0.41, + "learning_rate": 1.3204202348953607e-05, + "loss": 0.305, + "step": 6858 + }, + { + "epoch": 0.41, + "learning_rate": 1.3202348803728153e-05, + "loss": 0.3384, + "step": 6859 + }, + { + "epoch": 0.41, + "learning_rate": 1.3200495135901798e-05, + "loss": 0.3454, + "step": 6860 + }, + { + "epoch": 0.41, + "learning_rate": 1.3198641345545513e-05, + "loss": 0.3172, + "step": 6861 + }, + { + "epoch": 0.41, + "learning_rate": 1.3196787432730262e-05, + "loss": 0.3008, + "step": 6862 + }, + { + "epoch": 0.41, + "learning_rate": 1.319493339752703e-05, + "loss": 0.3387, + "step": 6863 + }, + { + "epoch": 0.41, + "learning_rate": 1.3193079240006792e-05, + "loss": 0.31, + "step": 6864 + }, + { + "epoch": 0.41, + "learning_rate": 1.3191224960240536e-05, + "loss": 0.3016, + "step": 6865 + }, + { + "epoch": 0.41, + "learning_rate": 1.3189370558299255e-05, + "loss": 0.3259, + "step": 6866 + }, + { + "epoch": 0.41, + "learning_rate": 1.318751603425394e-05, + "loss": 0.3329, + "step": 6867 + }, + { + "epoch": 0.41, + "learning_rate": 1.3185661388175594e-05, + "loss": 0.3062, + "step": 6868 + }, + { + "epoch": 0.41, + "learning_rate": 1.3183806620135218e-05, + "loss": 0.3206, + "step": 6869 + }, + { + "epoch": 0.42, + "learning_rate": 1.3181951730203823e-05, + "loss": 0.2994, + "step": 6870 + }, + { + "epoch": 0.42, + "learning_rate": 1.3180096718452426e-05, + "loss": 0.3131, + "step": 6871 + }, + { + "epoch": 0.42, + "learning_rate": 1.317824158495204e-05, + "loss": 0.3417, + "step": 6872 + }, + { + "epoch": 0.42, + "learning_rate": 1.3176386329773694e-05, + "loss": 0.296, + "step": 6873 + }, + { + "epoch": 0.42, + "learning_rate": 1.317453095298841e-05, + "loss": 0.3078, + "step": 6874 + }, + { + "epoch": 0.42, + "learning_rate": 1.3172675454667223e-05, + "loss": 0.3156, + "step": 6875 + }, + { + "epoch": 0.42, + "learning_rate": 1.3170819834881172e-05, + "loss": 0.3178, + "step": 6876 + }, + { + "epoch": 0.42, + "learning_rate": 1.3168964093701297e-05, + "loss": 0.3178, + "step": 6877 + }, + { + "epoch": 0.42, + "learning_rate": 1.3167108231198644e-05, + "loss": 0.2983, + "step": 6878 + }, + { + "epoch": 0.42, + "learning_rate": 1.3165252247444265e-05, + "loss": 0.3228, + "step": 6879 + }, + { + "epoch": 0.42, + "learning_rate": 1.3163396142509217e-05, + "loss": 0.3351, + "step": 6880 + }, + { + "epoch": 0.42, + "learning_rate": 1.3161539916464559e-05, + "loss": 0.2955, + "step": 6881 + }, + { + "epoch": 0.42, + "learning_rate": 1.3159683569381353e-05, + "loss": 0.3156, + "step": 6882 + }, + { + "epoch": 0.42, + "learning_rate": 1.3157827101330674e-05, + "loss": 0.3178, + "step": 6883 + }, + { + "epoch": 0.42, + "learning_rate": 1.3155970512383593e-05, + "loss": 0.3117, + "step": 6884 + }, + { + "epoch": 0.42, + "learning_rate": 1.3154113802611193e-05, + "loss": 0.2871, + "step": 6885 + }, + { + "epoch": 0.42, + "learning_rate": 1.3152256972084552e-05, + "loss": 0.3259, + "step": 6886 + }, + { + "epoch": 0.42, + "learning_rate": 1.3150400020874758e-05, + "loss": 0.3463, + "step": 6887 + }, + { + "epoch": 0.42, + "learning_rate": 1.3148542949052908e-05, + "loss": 0.3362, + "step": 6888 + }, + { + "epoch": 0.42, + "learning_rate": 1.3146685756690098e-05, + "loss": 0.3096, + "step": 6889 + }, + { + "epoch": 0.42, + "learning_rate": 1.3144828443857431e-05, + "loss": 0.3245, + "step": 6890 + }, + { + "epoch": 0.42, + "learning_rate": 1.3142971010626013e-05, + "loss": 0.3315, + "step": 6891 + }, + { + "epoch": 0.42, + "learning_rate": 1.3141113457066951e-05, + "loss": 0.3195, + "step": 6892 + }, + { + "epoch": 0.42, + "learning_rate": 1.3139255783251368e-05, + "loss": 0.315, + "step": 6893 + }, + { + "epoch": 0.42, + "learning_rate": 1.3137397989250378e-05, + "loss": 0.3677, + "step": 6894 + }, + { + "epoch": 0.42, + "learning_rate": 1.3135540075135112e-05, + "loss": 0.3206, + "step": 6895 + }, + { + "epoch": 0.42, + "learning_rate": 1.3133682040976696e-05, + "loss": 0.3228, + "step": 6896 + }, + { + "epoch": 0.42, + "learning_rate": 1.3131823886846262e-05, + "loss": 0.3368, + "step": 6897 + }, + { + "epoch": 0.42, + "learning_rate": 1.3129965612814951e-05, + "loss": 0.3627, + "step": 6898 + }, + { + "epoch": 0.42, + "learning_rate": 1.312810721895391e-05, + "loss": 0.3273, + "step": 6899 + }, + { + "epoch": 0.42, + "learning_rate": 1.3126248705334286e-05, + "loss": 0.3549, + "step": 6900 + }, + { + "epoch": 0.42, + "learning_rate": 1.3124390072027227e-05, + "loss": 0.3301, + "step": 6901 + }, + { + "epoch": 0.42, + "learning_rate": 1.3122531319103894e-05, + "loss": 0.2932, + "step": 6902 + }, + { + "epoch": 0.42, + "learning_rate": 1.3120672446635448e-05, + "loss": 0.3234, + "step": 6903 + }, + { + "epoch": 0.42, + "learning_rate": 1.3118813454693054e-05, + "loss": 0.3186, + "step": 6904 + }, + { + "epoch": 0.42, + "learning_rate": 1.3116954343347882e-05, + "loss": 0.358, + "step": 6905 + }, + { + "epoch": 0.42, + "learning_rate": 1.3115095112671115e-05, + "loss": 0.3351, + "step": 6906 + }, + { + "epoch": 0.42, + "learning_rate": 1.311323576273392e-05, + "loss": 0.3429, + "step": 6907 + }, + { + "epoch": 0.42, + "learning_rate": 1.3111376293607494e-05, + "loss": 0.3133, + "step": 6908 + }, + { + "epoch": 0.42, + "learning_rate": 1.3109516705363018e-05, + "loss": 0.2888, + "step": 6909 + }, + { + "epoch": 0.42, + "learning_rate": 1.3107656998071691e-05, + "loss": 0.3206, + "step": 6910 + }, + { + "epoch": 0.42, + "learning_rate": 1.3105797171804713e-05, + "loss": 0.3172, + "step": 6911 + }, + { + "epoch": 0.42, + "learning_rate": 1.3103937226633277e-05, + "loss": 0.3359, + "step": 6912 + }, + { + "epoch": 0.42, + "learning_rate": 1.31020771626286e-05, + "loss": 0.2955, + "step": 6913 + }, + { + "epoch": 0.42, + "learning_rate": 1.3100216979861888e-05, + "loss": 0.316, + "step": 6914 + }, + { + "epoch": 0.42, + "learning_rate": 1.309835667840436e-05, + "loss": 0.3178, + "step": 6915 + }, + { + "epoch": 0.42, + "learning_rate": 1.309649625832724e-05, + "loss": 0.3284, + "step": 6916 + }, + { + "epoch": 0.42, + "learning_rate": 1.309463571970175e-05, + "loss": 0.3326, + "step": 6917 + }, + { + "epoch": 0.42, + "learning_rate": 1.309277506259912e-05, + "loss": 0.3013, + "step": 6918 + }, + { + "epoch": 0.42, + "learning_rate": 1.3090914287090586e-05, + "loss": 0.3301, + "step": 6919 + }, + { + "epoch": 0.42, + "learning_rate": 1.3089053393247387e-05, + "loss": 0.3304, + "step": 6920 + }, + { + "epoch": 0.42, + "learning_rate": 1.3087192381140763e-05, + "loss": 0.3175, + "step": 6921 + }, + { + "epoch": 0.42, + "learning_rate": 1.3085331250841968e-05, + "loss": 0.3306, + "step": 6922 + }, + { + "epoch": 0.42, + "learning_rate": 1.3083470002422254e-05, + "loss": 0.3446, + "step": 6923 + }, + { + "epoch": 0.42, + "learning_rate": 1.3081608635952874e-05, + "loss": 0.3178, + "step": 6924 + }, + { + "epoch": 0.42, + "learning_rate": 1.3079747151505093e-05, + "loss": 0.3354, + "step": 6925 + }, + { + "epoch": 0.42, + "learning_rate": 1.307788554915018e-05, + "loss": 0.3281, + "step": 6926 + }, + { + "epoch": 0.42, + "learning_rate": 1.3076023828959396e-05, + "loss": 0.3306, + "step": 6927 + }, + { + "epoch": 0.42, + "learning_rate": 1.307416199100403e-05, + "loss": 0.3563, + "step": 6928 + }, + { + "epoch": 0.42, + "learning_rate": 1.307230003535535e-05, + "loss": 0.3401, + "step": 6929 + }, + { + "epoch": 0.42, + "learning_rate": 1.3070437962084647e-05, + "loss": 0.2896, + "step": 6930 + }, + { + "epoch": 0.42, + "learning_rate": 1.306857577126321e-05, + "loss": 0.3541, + "step": 6931 + }, + { + "epoch": 0.42, + "learning_rate": 1.3066713462962329e-05, + "loss": 0.3075, + "step": 6932 + }, + { + "epoch": 0.42, + "learning_rate": 1.3064851037253304e-05, + "loss": 0.3354, + "step": 6933 + }, + { + "epoch": 0.42, + "learning_rate": 1.3062988494207435e-05, + "loss": 0.3065, + "step": 6934 + }, + { + "epoch": 0.42, + "learning_rate": 1.3061125833896035e-05, + "loss": 0.3368, + "step": 6935 + }, + { + "epoch": 0.42, + "learning_rate": 1.3059263056390408e-05, + "loss": 0.3273, + "step": 6936 + }, + { + "epoch": 0.42, + "learning_rate": 1.3057400161761873e-05, + "loss": 0.3354, + "step": 6937 + }, + { + "epoch": 0.42, + "learning_rate": 1.305553715008175e-05, + "loss": 0.327, + "step": 6938 + }, + { + "epoch": 0.42, + "learning_rate": 1.3053674021421364e-05, + "loss": 0.3301, + "step": 6939 + }, + { + "epoch": 0.42, + "learning_rate": 1.3051810775852042e-05, + "loss": 0.3345, + "step": 6940 + }, + { + "epoch": 0.42, + "learning_rate": 1.3049947413445125e-05, + "loss": 0.3407, + "step": 6941 + }, + { + "epoch": 0.42, + "learning_rate": 1.3048083934271942e-05, + "loss": 0.3265, + "step": 6942 + }, + { + "epoch": 0.42, + "learning_rate": 1.3046220338403844e-05, + "loss": 0.3195, + "step": 6943 + }, + { + "epoch": 0.42, + "learning_rate": 1.3044356625912169e-05, + "loss": 0.3059, + "step": 6944 + }, + { + "epoch": 0.42, + "learning_rate": 1.3042492796868277e-05, + "loss": 0.3209, + "step": 6945 + }, + { + "epoch": 0.42, + "learning_rate": 1.3040628851343522e-05, + "loss": 0.3114, + "step": 6946 + }, + { + "epoch": 0.42, + "learning_rate": 1.3038764789409261e-05, + "loss": 0.3125, + "step": 6947 + }, + { + "epoch": 0.42, + "learning_rate": 1.3036900611136862e-05, + "loss": 0.3044, + "step": 6948 + }, + { + "epoch": 0.42, + "learning_rate": 1.3035036316597695e-05, + "loss": 0.3424, + "step": 6949 + }, + { + "epoch": 0.42, + "learning_rate": 1.3033171905863134e-05, + "loss": 0.3094, + "step": 6950 + }, + { + "epoch": 0.42, + "learning_rate": 1.3031307379004554e-05, + "loss": 0.322, + "step": 6951 + }, + { + "epoch": 0.42, + "learning_rate": 1.3029442736093341e-05, + "loss": 0.2946, + "step": 6952 + }, + { + "epoch": 0.42, + "learning_rate": 1.3027577977200882e-05, + "loss": 0.2953, + "step": 6953 + }, + { + "epoch": 0.42, + "learning_rate": 1.3025713102398565e-05, + "loss": 0.3301, + "step": 6954 + }, + { + "epoch": 0.42, + "learning_rate": 1.3023848111757792e-05, + "loss": 0.3387, + "step": 6955 + }, + { + "epoch": 0.42, + "learning_rate": 1.3021983005349959e-05, + "loss": 0.3309, + "step": 6956 + }, + { + "epoch": 0.42, + "learning_rate": 1.3020117783246473e-05, + "loss": 0.3393, + "step": 6957 + }, + { + "epoch": 0.42, + "learning_rate": 1.3018252445518745e-05, + "loss": 0.3273, + "step": 6958 + }, + { + "epoch": 0.42, + "learning_rate": 1.3016386992238184e-05, + "loss": 0.3398, + "step": 6959 + }, + { + "epoch": 0.42, + "learning_rate": 1.3014521423476214e-05, + "loss": 0.3052, + "step": 6960 + }, + { + "epoch": 0.42, + "learning_rate": 1.3012655739304255e-05, + "loss": 0.3281, + "step": 6961 + }, + { + "epoch": 0.42, + "learning_rate": 1.3010789939793732e-05, + "loss": 0.2821, + "step": 6962 + }, + { + "epoch": 0.42, + "learning_rate": 1.3008924025016079e-05, + "loss": 0.3315, + "step": 6963 + }, + { + "epoch": 0.42, + "learning_rate": 1.300705799504273e-05, + "loss": 0.3214, + "step": 6964 + }, + { + "epoch": 0.42, + "learning_rate": 1.3005191849945132e-05, + "loss": 0.315, + "step": 6965 + }, + { + "epoch": 0.42, + "learning_rate": 1.3003325589794721e-05, + "loss": 0.3535, + "step": 6966 + }, + { + "epoch": 0.42, + "learning_rate": 1.300145921466295e-05, + "loss": 0.2991, + "step": 6967 + }, + { + "epoch": 0.42, + "learning_rate": 1.2999592724621275e-05, + "loss": 0.3163, + "step": 6968 + }, + { + "epoch": 0.42, + "learning_rate": 1.299772611974115e-05, + "loss": 0.3401, + "step": 6969 + }, + { + "epoch": 0.42, + "learning_rate": 1.2995859400094039e-05, + "loss": 0.3158, + "step": 6970 + }, + { + "epoch": 0.42, + "learning_rate": 1.2993992565751408e-05, + "loss": 0.3105, + "step": 6971 + }, + { + "epoch": 0.42, + "learning_rate": 1.299212561678473e-05, + "loss": 0.346, + "step": 6972 + }, + { + "epoch": 0.42, + "learning_rate": 1.2990258553265479e-05, + "loss": 0.3401, + "step": 6973 + }, + { + "epoch": 0.42, + "learning_rate": 1.2988391375265133e-05, + "loss": 0.3265, + "step": 6974 + }, + { + "epoch": 0.42, + "learning_rate": 1.2986524082855181e-05, + "loss": 0.3412, + "step": 6975 + }, + { + "epoch": 0.42, + "learning_rate": 1.2984656676107109e-05, + "loss": 0.3225, + "step": 6976 + }, + { + "epoch": 0.42, + "learning_rate": 1.2982789155092407e-05, + "loss": 0.3267, + "step": 6977 + }, + { + "epoch": 0.42, + "learning_rate": 1.2980921519882579e-05, + "loss": 0.2782, + "step": 6978 + }, + { + "epoch": 0.42, + "learning_rate": 1.2979053770549122e-05, + "loss": 0.3401, + "step": 6979 + }, + { + "epoch": 0.42, + "learning_rate": 1.2977185907163542e-05, + "loss": 0.3354, + "step": 6980 + }, + { + "epoch": 0.42, + "learning_rate": 1.2975317929797357e-05, + "loss": 0.3281, + "step": 6981 + }, + { + "epoch": 0.42, + "learning_rate": 1.2973449838522071e-05, + "loss": 0.3122, + "step": 6982 + }, + { + "epoch": 0.42, + "learning_rate": 1.2971581633409211e-05, + "loss": 0.3047, + "step": 6983 + }, + { + "epoch": 0.42, + "learning_rate": 1.2969713314530297e-05, + "loss": 0.3248, + "step": 6984 + }, + { + "epoch": 0.42, + "learning_rate": 1.296784488195686e-05, + "loss": 0.3424, + "step": 6985 + }, + { + "epoch": 0.42, + "learning_rate": 1.2965976335760428e-05, + "loss": 0.296, + "step": 6986 + }, + { + "epoch": 0.42, + "learning_rate": 1.2964107676012542e-05, + "loss": 0.31, + "step": 6987 + }, + { + "epoch": 0.42, + "learning_rate": 1.2962238902784742e-05, + "loss": 0.3242, + "step": 6988 + }, + { + "epoch": 0.42, + "learning_rate": 1.296037001614857e-05, + "loss": 0.3304, + "step": 6989 + }, + { + "epoch": 0.42, + "learning_rate": 1.295850101617558e-05, + "loss": 0.3262, + "step": 6990 + }, + { + "epoch": 0.42, + "learning_rate": 1.2956631902937322e-05, + "loss": 0.3373, + "step": 6991 + }, + { + "epoch": 0.42, + "learning_rate": 1.295476267650536e-05, + "loss": 0.3251, + "step": 6992 + }, + { + "epoch": 0.42, + "learning_rate": 1.2952893336951253e-05, + "loss": 0.3371, + "step": 6993 + }, + { + "epoch": 0.42, + "learning_rate": 1.2951023884346564e-05, + "loss": 0.3396, + "step": 6994 + }, + { + "epoch": 0.42, + "learning_rate": 1.2949154318762876e-05, + "loss": 0.2965, + "step": 6995 + }, + { + "epoch": 0.42, + "learning_rate": 1.2947284640271756e-05, + "loss": 0.3292, + "step": 6996 + }, + { + "epoch": 0.42, + "learning_rate": 1.2945414848944782e-05, + "loss": 0.3013, + "step": 6997 + }, + { + "epoch": 0.42, + "learning_rate": 1.2943544944853543e-05, + "loss": 0.303, + "step": 6998 + }, + { + "epoch": 0.42, + "learning_rate": 1.2941674928069627e-05, + "loss": 0.3357, + "step": 6999 + }, + { + "epoch": 0.42, + "learning_rate": 1.293980479866463e-05, + "loss": 0.3239, + "step": 7000 + }, + { + "epoch": 0.42, + "learning_rate": 1.2937934556710143e-05, + "loss": 0.3318, + "step": 7001 + }, + { + "epoch": 0.42, + "learning_rate": 1.2936064202277771e-05, + "loss": 0.3228, + "step": 7002 + }, + { + "epoch": 0.42, + "learning_rate": 1.2934193735439123e-05, + "loss": 0.327, + "step": 7003 + }, + { + "epoch": 0.42, + "learning_rate": 1.2932323156265803e-05, + "loss": 0.327, + "step": 7004 + }, + { + "epoch": 0.42, + "learning_rate": 1.2930452464829428e-05, + "loss": 0.3465, + "step": 7005 + }, + { + "epoch": 0.42, + "learning_rate": 1.2928581661201616e-05, + "loss": 0.31, + "step": 7006 + }, + { + "epoch": 0.42, + "learning_rate": 1.2926710745453992e-05, + "loss": 0.3117, + "step": 7007 + }, + { + "epoch": 0.42, + "learning_rate": 1.2924839717658184e-05, + "loss": 0.2902, + "step": 7008 + }, + { + "epoch": 0.42, + "learning_rate": 1.2922968577885822e-05, + "loss": 0.3005, + "step": 7009 + }, + { + "epoch": 0.42, + "learning_rate": 1.2921097326208543e-05, + "loss": 0.3354, + "step": 7010 + }, + { + "epoch": 0.42, + "learning_rate": 1.2919225962697985e-05, + "loss": 0.3071, + "step": 7011 + }, + { + "epoch": 0.42, + "learning_rate": 1.2917354487425796e-05, + "loss": 0.3234, + "step": 7012 + }, + { + "epoch": 0.42, + "learning_rate": 1.2915482900463622e-05, + "loss": 0.3315, + "step": 7013 + }, + { + "epoch": 0.42, + "learning_rate": 1.2913611201883116e-05, + "loss": 0.3033, + "step": 7014 + }, + { + "epoch": 0.42, + "learning_rate": 1.291173939175594e-05, + "loss": 0.3083, + "step": 7015 + }, + { + "epoch": 0.42, + "learning_rate": 1.2909867470153752e-05, + "loss": 0.3357, + "step": 7016 + }, + { + "epoch": 0.42, + "learning_rate": 1.2907995437148216e-05, + "loss": 0.3089, + "step": 7017 + }, + { + "epoch": 0.42, + "learning_rate": 1.2906123292811004e-05, + "loss": 0.305, + "step": 7018 + }, + { + "epoch": 0.42, + "learning_rate": 1.2904251037213792e-05, + "loss": 0.3248, + "step": 7019 + }, + { + "epoch": 0.42, + "learning_rate": 1.2902378670428257e-05, + "loss": 0.3569, + "step": 7020 + }, + { + "epoch": 0.42, + "learning_rate": 1.290050619252608e-05, + "loss": 0.3114, + "step": 7021 + }, + { + "epoch": 0.42, + "learning_rate": 1.2898633603578957e-05, + "loss": 0.3189, + "step": 7022 + }, + { + "epoch": 0.42, + "learning_rate": 1.2896760903658567e-05, + "loss": 0.3382, + "step": 7023 + }, + { + "epoch": 0.42, + "learning_rate": 1.2894888092836615e-05, + "loss": 0.3245, + "step": 7024 + }, + { + "epoch": 0.42, + "learning_rate": 1.2893015171184797e-05, + "loss": 0.3206, + "step": 7025 + }, + { + "epoch": 0.42, + "learning_rate": 1.2891142138774818e-05, + "loss": 0.308, + "step": 7026 + }, + { + "epoch": 0.42, + "learning_rate": 1.2889268995678386e-05, + "loss": 0.344, + "step": 7027 + }, + { + "epoch": 0.42, + "learning_rate": 1.2887395741967215e-05, + "loss": 0.3125, + "step": 7028 + }, + { + "epoch": 0.42, + "learning_rate": 1.288552237771302e-05, + "loss": 0.332, + "step": 7029 + }, + { + "epoch": 0.42, + "learning_rate": 1.2883648902987523e-05, + "loss": 0.2946, + "step": 7030 + }, + { + "epoch": 0.42, + "learning_rate": 1.288177531786245e-05, + "loss": 0.3203, + "step": 7031 + }, + { + "epoch": 0.42, + "learning_rate": 1.287990162240953e-05, + "loss": 0.3541, + "step": 7032 + }, + { + "epoch": 0.42, + "learning_rate": 1.2878027816700497e-05, + "loss": 0.299, + "step": 7033 + }, + { + "epoch": 0.42, + "learning_rate": 1.2876153900807088e-05, + "loss": 0.3244, + "step": 7034 + }, + { + "epoch": 0.42, + "learning_rate": 1.287427987480105e-05, + "loss": 0.3365, + "step": 7035 + }, + { + "epoch": 0.43, + "learning_rate": 1.2872405738754126e-05, + "loss": 0.3295, + "step": 7036 + }, + { + "epoch": 0.43, + "learning_rate": 1.2870531492738066e-05, + "loss": 0.3251, + "step": 7037 + }, + { + "epoch": 0.43, + "learning_rate": 1.2868657136824623e-05, + "loss": 0.3432, + "step": 7038 + }, + { + "epoch": 0.43, + "learning_rate": 1.2866782671085559e-05, + "loss": 0.3354, + "step": 7039 + }, + { + "epoch": 0.43, + "learning_rate": 1.2864908095592638e-05, + "loss": 0.3178, + "step": 7040 + }, + { + "epoch": 0.43, + "learning_rate": 1.2863033410417625e-05, + "loss": 0.3304, + "step": 7041 + }, + { + "epoch": 0.43, + "learning_rate": 1.2861158615632294e-05, + "loss": 0.3125, + "step": 7042 + }, + { + "epoch": 0.43, + "learning_rate": 1.285928371130842e-05, + "loss": 0.341, + "step": 7043 + }, + { + "epoch": 0.43, + "learning_rate": 1.2857408697517783e-05, + "loss": 0.3253, + "step": 7044 + }, + { + "epoch": 0.43, + "learning_rate": 1.2855533574332168e-05, + "loss": 0.3348, + "step": 7045 + }, + { + "epoch": 0.43, + "learning_rate": 1.285365834182336e-05, + "loss": 0.3251, + "step": 7046 + }, + { + "epoch": 0.43, + "learning_rate": 1.2851783000063161e-05, + "loss": 0.3022, + "step": 7047 + }, + { + "epoch": 0.43, + "learning_rate": 1.2849907549123357e-05, + "loss": 0.3111, + "step": 7048 + }, + { + "epoch": 0.43, + "learning_rate": 1.2848031989075754e-05, + "loss": 0.3172, + "step": 7049 + }, + { + "epoch": 0.43, + "learning_rate": 1.2846156319992156e-05, + "loss": 0.3368, + "step": 7050 + }, + { + "epoch": 0.43, + "learning_rate": 1.2844280541944377e-05, + "loss": 0.3133, + "step": 7051 + }, + { + "epoch": 0.43, + "learning_rate": 1.2842404655004222e-05, + "loss": 0.3055, + "step": 7052 + }, + { + "epoch": 0.43, + "learning_rate": 1.2840528659243514e-05, + "loss": 0.3259, + "step": 7053 + }, + { + "epoch": 0.43, + "learning_rate": 1.2838652554734076e-05, + "loss": 0.3426, + "step": 7054 + }, + { + "epoch": 0.43, + "learning_rate": 1.2836776341547735e-05, + "loss": 0.3442, + "step": 7055 + }, + { + "epoch": 0.43, + "learning_rate": 1.2834900019756313e-05, + "loss": 0.3401, + "step": 7056 + }, + { + "epoch": 0.43, + "learning_rate": 1.2833023589431654e-05, + "loss": 0.3147, + "step": 7057 + }, + { + "epoch": 0.43, + "learning_rate": 1.283114705064559e-05, + "loss": 0.3251, + "step": 7058 + }, + { + "epoch": 0.43, + "learning_rate": 1.2829270403469969e-05, + "loss": 0.3382, + "step": 7059 + }, + { + "epoch": 0.43, + "learning_rate": 1.2827393647976633e-05, + "loss": 0.3295, + "step": 7060 + }, + { + "epoch": 0.43, + "learning_rate": 1.2825516784237435e-05, + "loss": 0.3516, + "step": 7061 + }, + { + "epoch": 0.43, + "learning_rate": 1.2823639812324231e-05, + "loss": 0.3133, + "step": 7062 + }, + { + "epoch": 0.43, + "learning_rate": 1.282176273230888e-05, + "loss": 0.3302, + "step": 7063 + }, + { + "epoch": 0.43, + "learning_rate": 1.2819885544263243e-05, + "loss": 0.3432, + "step": 7064 + }, + { + "epoch": 0.43, + "learning_rate": 1.281800824825919e-05, + "loss": 0.2991, + "step": 7065 + }, + { + "epoch": 0.43, + "learning_rate": 1.2816130844368595e-05, + "loss": 0.3022, + "step": 7066 + }, + { + "epoch": 0.43, + "learning_rate": 1.2814253332663327e-05, + "loss": 0.2972, + "step": 7067 + }, + { + "epoch": 0.43, + "learning_rate": 1.281237571321527e-05, + "loss": 0.2963, + "step": 7068 + }, + { + "epoch": 0.43, + "learning_rate": 1.2810497986096308e-05, + "loss": 0.3097, + "step": 7069 + }, + { + "epoch": 0.43, + "learning_rate": 1.2808620151378333e-05, + "loss": 0.3108, + "step": 7070 + }, + { + "epoch": 0.43, + "learning_rate": 1.2806742209133231e-05, + "loss": 0.3295, + "step": 7071 + }, + { + "epoch": 0.43, + "learning_rate": 1.2804864159432902e-05, + "loss": 0.3128, + "step": 7072 + }, + { + "epoch": 0.43, + "learning_rate": 1.2802986002349242e-05, + "loss": 0.3287, + "step": 7073 + }, + { + "epoch": 0.43, + "learning_rate": 1.2801107737954161e-05, + "loss": 0.3069, + "step": 7074 + }, + { + "epoch": 0.43, + "learning_rate": 1.2799229366319566e-05, + "loss": 0.3037, + "step": 7075 + }, + { + "epoch": 0.43, + "learning_rate": 1.2797350887517368e-05, + "loss": 0.3139, + "step": 7076 + }, + { + "epoch": 0.43, + "learning_rate": 1.2795472301619488e-05, + "loss": 0.3396, + "step": 7077 + }, + { + "epoch": 0.43, + "learning_rate": 1.2793593608697844e-05, + "loss": 0.3295, + "step": 7078 + }, + { + "epoch": 0.43, + "learning_rate": 1.2791714808824358e-05, + "loss": 0.3334, + "step": 7079 + }, + { + "epoch": 0.43, + "learning_rate": 1.2789835902070966e-05, + "loss": 0.3359, + "step": 7080 + }, + { + "epoch": 0.43, + "learning_rate": 1.2787956888509596e-05, + "loss": 0.3345, + "step": 7081 + }, + { + "epoch": 0.43, + "learning_rate": 1.2786077768212193e-05, + "loss": 0.305, + "step": 7082 + }, + { + "epoch": 0.43, + "learning_rate": 1.278419854125069e-05, + "loss": 0.2991, + "step": 7083 + }, + { + "epoch": 0.43, + "learning_rate": 1.2782319207697036e-05, + "loss": 0.305, + "step": 7084 + }, + { + "epoch": 0.43, + "learning_rate": 1.278043976762318e-05, + "loss": 0.3326, + "step": 7085 + }, + { + "epoch": 0.43, + "learning_rate": 1.277856022110108e-05, + "loss": 0.3365, + "step": 7086 + }, + { + "epoch": 0.43, + "learning_rate": 1.2776680568202684e-05, + "loss": 0.3253, + "step": 7087 + }, + { + "epoch": 0.43, + "learning_rate": 1.2774800808999967e-05, + "loss": 0.3287, + "step": 7088 + }, + { + "epoch": 0.43, + "learning_rate": 1.2772920943564882e-05, + "loss": 0.3092, + "step": 7089 + }, + { + "epoch": 0.43, + "learning_rate": 1.2771040971969409e-05, + "loss": 0.3072, + "step": 7090 + }, + { + "epoch": 0.43, + "learning_rate": 1.2769160894285516e-05, + "loss": 0.3312, + "step": 7091 + }, + { + "epoch": 0.43, + "learning_rate": 1.2767280710585183e-05, + "loss": 0.3585, + "step": 7092 + }, + { + "epoch": 0.43, + "learning_rate": 1.2765400420940397e-05, + "loss": 0.3387, + "step": 7093 + }, + { + "epoch": 0.43, + "learning_rate": 1.2763520025423135e-05, + "loss": 0.3334, + "step": 7094 + }, + { + "epoch": 0.43, + "learning_rate": 1.2761639524105393e-05, + "loss": 0.3329, + "step": 7095 + }, + { + "epoch": 0.43, + "learning_rate": 1.2759758917059166e-05, + "loss": 0.3085, + "step": 7096 + }, + { + "epoch": 0.43, + "learning_rate": 1.2757878204356451e-05, + "loss": 0.3069, + "step": 7097 + }, + { + "epoch": 0.43, + "learning_rate": 1.2755997386069252e-05, + "loss": 0.3195, + "step": 7098 + }, + { + "epoch": 0.43, + "learning_rate": 1.275411646226957e-05, + "loss": 0.2775, + "step": 7099 + }, + { + "epoch": 0.43, + "learning_rate": 1.2752235433029423e-05, + "loss": 0.3278, + "step": 7100 + }, + { + "epoch": 0.43, + "learning_rate": 1.2750354298420824e-05, + "loss": 0.3195, + "step": 7101 + }, + { + "epoch": 0.43, + "learning_rate": 1.2748473058515785e-05, + "loss": 0.3223, + "step": 7102 + }, + { + "epoch": 0.43, + "learning_rate": 1.2746591713386337e-05, + "loss": 0.3245, + "step": 7103 + }, + { + "epoch": 0.43, + "learning_rate": 1.27447102631045e-05, + "loss": 0.3251, + "step": 7104 + }, + { + "epoch": 0.43, + "learning_rate": 1.2742828707742314e-05, + "loss": 0.3401, + "step": 7105 + }, + { + "epoch": 0.43, + "learning_rate": 1.27409470473718e-05, + "loss": 0.3111, + "step": 7106 + }, + { + "epoch": 0.43, + "learning_rate": 1.2739065282065011e-05, + "loss": 0.3287, + "step": 7107 + }, + { + "epoch": 0.43, + "learning_rate": 1.2737183411893983e-05, + "loss": 0.2963, + "step": 7108 + }, + { + "epoch": 0.43, + "learning_rate": 1.2735301436930757e-05, + "loss": 0.3111, + "step": 7109 + }, + { + "epoch": 0.43, + "learning_rate": 1.2733419357247397e-05, + "loss": 0.3156, + "step": 7110 + }, + { + "epoch": 0.43, + "learning_rate": 1.2731537172915945e-05, + "loss": 0.3499, + "step": 7111 + }, + { + "epoch": 0.43, + "learning_rate": 1.2729654884008468e-05, + "loss": 0.3297, + "step": 7112 + }, + { + "epoch": 0.43, + "learning_rate": 1.2727772490597028e-05, + "loss": 0.2972, + "step": 7113 + }, + { + "epoch": 0.43, + "learning_rate": 1.2725889992753688e-05, + "loss": 0.3013, + "step": 7114 + }, + { + "epoch": 0.43, + "learning_rate": 1.2724007390550524e-05, + "loss": 0.3172, + "step": 7115 + }, + { + "epoch": 0.43, + "learning_rate": 1.2722124684059602e-05, + "loss": 0.3485, + "step": 7116 + }, + { + "epoch": 0.43, + "learning_rate": 1.2720241873353012e-05, + "loss": 0.353, + "step": 7117 + }, + { + "epoch": 0.43, + "learning_rate": 1.2718358958502832e-05, + "loss": 0.287, + "step": 7118 + }, + { + "epoch": 0.43, + "learning_rate": 1.2716475939581144e-05, + "loss": 0.2793, + "step": 7119 + }, + { + "epoch": 0.43, + "learning_rate": 1.2714592816660045e-05, + "loss": 0.3435, + "step": 7120 + }, + { + "epoch": 0.43, + "learning_rate": 1.2712709589811629e-05, + "loss": 0.3276, + "step": 7121 + }, + { + "epoch": 0.43, + "learning_rate": 1.2710826259107993e-05, + "loss": 0.3052, + "step": 7122 + }, + { + "epoch": 0.43, + "learning_rate": 1.2708942824621243e-05, + "loss": 0.3482, + "step": 7123 + }, + { + "epoch": 0.43, + "learning_rate": 1.2707059286423475e-05, + "loss": 0.3066, + "step": 7124 + }, + { + "epoch": 0.43, + "learning_rate": 1.2705175644586816e-05, + "loss": 0.304, + "step": 7125 + }, + { + "epoch": 0.43, + "learning_rate": 1.2703291899183366e-05, + "loss": 0.3267, + "step": 7126 + }, + { + "epoch": 0.43, + "learning_rate": 1.2701408050285253e-05, + "loss": 0.3195, + "step": 7127 + }, + { + "epoch": 0.43, + "learning_rate": 1.2699524097964597e-05, + "loss": 0.346, + "step": 7128 + }, + { + "epoch": 0.43, + "learning_rate": 1.2697640042293522e-05, + "loss": 0.3153, + "step": 7129 + }, + { + "epoch": 0.43, + "learning_rate": 1.269575588334416e-05, + "loss": 0.3404, + "step": 7130 + }, + { + "epoch": 0.43, + "learning_rate": 1.2693871621188647e-05, + "loss": 0.2705, + "step": 7131 + }, + { + "epoch": 0.43, + "learning_rate": 1.269198725589912e-05, + "loss": 0.3136, + "step": 7132 + }, + { + "epoch": 0.43, + "learning_rate": 1.2690102787547722e-05, + "loss": 0.3139, + "step": 7133 + }, + { + "epoch": 0.43, + "learning_rate": 1.2688218216206599e-05, + "loss": 0.3488, + "step": 7134 + }, + { + "epoch": 0.43, + "learning_rate": 1.26863335419479e-05, + "loss": 0.3329, + "step": 7135 + }, + { + "epoch": 0.43, + "learning_rate": 1.2684448764843782e-05, + "loss": 0.3373, + "step": 7136 + }, + { + "epoch": 0.43, + "learning_rate": 1.2682563884966396e-05, + "loss": 0.3315, + "step": 7137 + }, + { + "epoch": 0.43, + "learning_rate": 1.2680678902387914e-05, + "loss": 0.304, + "step": 7138 + }, + { + "epoch": 0.43, + "learning_rate": 1.2678793817180496e-05, + "loss": 0.3209, + "step": 7139 + }, + { + "epoch": 0.43, + "learning_rate": 1.2676908629416316e-05, + "loss": 0.3368, + "step": 7140 + }, + { + "epoch": 0.43, + "learning_rate": 1.267502333916754e-05, + "loss": 0.3306, + "step": 7141 + }, + { + "epoch": 0.43, + "learning_rate": 1.2673137946506354e-05, + "loss": 0.3075, + "step": 7142 + }, + { + "epoch": 0.43, + "learning_rate": 1.2671252451504935e-05, + "loss": 0.3354, + "step": 7143 + }, + { + "epoch": 0.43, + "learning_rate": 1.2669366854235468e-05, + "loss": 0.2902, + "step": 7144 + }, + { + "epoch": 0.43, + "learning_rate": 1.2667481154770146e-05, + "loss": 0.3488, + "step": 7145 + }, + { + "epoch": 0.43, + "learning_rate": 1.2665595353181159e-05, + "loss": 0.2995, + "step": 7146 + }, + { + "epoch": 0.43, + "learning_rate": 1.2663709449540709e-05, + "loss": 0.3432, + "step": 7147 + }, + { + "epoch": 0.43, + "learning_rate": 1.2661823443920995e-05, + "loss": 0.3055, + "step": 7148 + }, + { + "epoch": 0.43, + "learning_rate": 1.2659937336394217e-05, + "loss": 0.3174, + "step": 7149 + }, + { + "epoch": 0.43, + "learning_rate": 1.2658051127032592e-05, + "loss": 0.3068, + "step": 7150 + }, + { + "epoch": 0.43, + "learning_rate": 1.2656164815908328e-05, + "loss": 0.3235, + "step": 7151 + }, + { + "epoch": 0.43, + "learning_rate": 1.2654278403093644e-05, + "loss": 0.3242, + "step": 7152 + }, + { + "epoch": 0.43, + "learning_rate": 1.265239188866076e-05, + "loss": 0.3061, + "step": 7153 + }, + { + "epoch": 0.43, + "learning_rate": 1.2650505272681897e-05, + "loss": 0.3147, + "step": 7154 + }, + { + "epoch": 0.43, + "learning_rate": 1.2648618555229291e-05, + "loss": 0.3429, + "step": 7155 + }, + { + "epoch": 0.43, + "learning_rate": 1.2646731736375169e-05, + "loss": 0.2955, + "step": 7156 + }, + { + "epoch": 0.43, + "learning_rate": 1.264484481619177e-05, + "loss": 0.339, + "step": 7157 + }, + { + "epoch": 0.43, + "learning_rate": 1.264295779475133e-05, + "loss": 0.3415, + "step": 7158 + }, + { + "epoch": 0.43, + "learning_rate": 1.2641070672126096e-05, + "loss": 0.3251, + "step": 7159 + }, + { + "epoch": 0.43, + "learning_rate": 1.2639183448388316e-05, + "loss": 0.3234, + "step": 7160 + }, + { + "epoch": 0.43, + "learning_rate": 1.2637296123610243e-05, + "loss": 0.3373, + "step": 7161 + }, + { + "epoch": 0.43, + "learning_rate": 1.2635408697864131e-05, + "loss": 0.3276, + "step": 7162 + }, + { + "epoch": 0.43, + "learning_rate": 1.2633521171222238e-05, + "loss": 0.3224, + "step": 7163 + }, + { + "epoch": 0.43, + "learning_rate": 1.2631633543756827e-05, + "loss": 0.3351, + "step": 7164 + }, + { + "epoch": 0.43, + "learning_rate": 1.2629745815540171e-05, + "loss": 0.305, + "step": 7165 + }, + { + "epoch": 0.43, + "learning_rate": 1.2627857986644533e-05, + "loss": 0.3186, + "step": 7166 + }, + { + "epoch": 0.43, + "learning_rate": 1.2625970057142197e-05, + "loss": 0.3256, + "step": 7167 + }, + { + "epoch": 0.43, + "learning_rate": 1.2624082027105436e-05, + "loss": 0.3164, + "step": 7168 + }, + { + "epoch": 0.43, + "learning_rate": 1.262219389660653e-05, + "loss": 0.3411, + "step": 7169 + }, + { + "epoch": 0.43, + "learning_rate": 1.2620305665717771e-05, + "loss": 0.3026, + "step": 7170 + }, + { + "epoch": 0.43, + "learning_rate": 1.2618417334511445e-05, + "loss": 0.3239, + "step": 7171 + }, + { + "epoch": 0.43, + "learning_rate": 1.2616528903059852e-05, + "loss": 0.3412, + "step": 7172 + }, + { + "epoch": 0.43, + "learning_rate": 1.2614640371435288e-05, + "loss": 0.3348, + "step": 7173 + }, + { + "epoch": 0.43, + "learning_rate": 1.2612751739710048e-05, + "loss": 0.3231, + "step": 7174 + }, + { + "epoch": 0.43, + "learning_rate": 1.2610863007956447e-05, + "loss": 0.3164, + "step": 7175 + }, + { + "epoch": 0.43, + "learning_rate": 1.2608974176246785e-05, + "loss": 0.317, + "step": 7176 + }, + { + "epoch": 0.43, + "learning_rate": 1.2607085244653385e-05, + "loss": 0.3186, + "step": 7177 + }, + { + "epoch": 0.43, + "learning_rate": 1.260519621324856e-05, + "loss": 0.3563, + "step": 7178 + }, + { + "epoch": 0.43, + "learning_rate": 1.2603307082104627e-05, + "loss": 0.305, + "step": 7179 + }, + { + "epoch": 0.43, + "learning_rate": 1.2601417851293919e-05, + "loss": 0.3119, + "step": 7180 + }, + { + "epoch": 0.43, + "learning_rate": 1.2599528520888757e-05, + "loss": 0.3273, + "step": 7181 + }, + { + "epoch": 0.43, + "learning_rate": 1.2597639090961477e-05, + "loss": 0.3189, + "step": 7182 + }, + { + "epoch": 0.43, + "learning_rate": 1.2595749561584419e-05, + "loss": 0.311, + "step": 7183 + }, + { + "epoch": 0.43, + "learning_rate": 1.2593859932829915e-05, + "loss": 0.3083, + "step": 7184 + }, + { + "epoch": 0.43, + "learning_rate": 1.2591970204770312e-05, + "loss": 0.3092, + "step": 7185 + }, + { + "epoch": 0.43, + "learning_rate": 1.2590080377477958e-05, + "loss": 0.3329, + "step": 7186 + }, + { + "epoch": 0.43, + "learning_rate": 1.2588190451025209e-05, + "loss": 0.3814, + "step": 7187 + }, + { + "epoch": 0.43, + "learning_rate": 1.2586300425484414e-05, + "loss": 0.3248, + "step": 7188 + }, + { + "epoch": 0.43, + "learning_rate": 1.2584410300927931e-05, + "loss": 0.2977, + "step": 7189 + }, + { + "epoch": 0.43, + "learning_rate": 1.2582520077428134e-05, + "loss": 0.2803, + "step": 7190 + }, + { + "epoch": 0.43, + "learning_rate": 1.2580629755057372e-05, + "loss": 0.3122, + "step": 7191 + }, + { + "epoch": 0.43, + "learning_rate": 1.2578739333888028e-05, + "loss": 0.3029, + "step": 7192 + }, + { + "epoch": 0.43, + "learning_rate": 1.2576848813992475e-05, + "loss": 0.3239, + "step": 7193 + }, + { + "epoch": 0.43, + "learning_rate": 1.2574958195443085e-05, + "loss": 0.3323, + "step": 7194 + }, + { + "epoch": 0.43, + "learning_rate": 1.257306747831225e-05, + "loss": 0.3256, + "step": 7195 + }, + { + "epoch": 0.43, + "learning_rate": 1.2571176662672342e-05, + "loss": 0.3015, + "step": 7196 + }, + { + "epoch": 0.43, + "learning_rate": 1.2569285748595761e-05, + "loss": 0.3157, + "step": 7197 + }, + { + "epoch": 0.43, + "learning_rate": 1.2567394736154896e-05, + "loss": 0.3242, + "step": 7198 + }, + { + "epoch": 0.43, + "learning_rate": 1.2565503625422144e-05, + "loss": 0.3415, + "step": 7199 + }, + { + "epoch": 0.43, + "learning_rate": 1.2563612416469906e-05, + "loss": 0.2969, + "step": 7200 + }, + { + "epoch": 0.44, + "learning_rate": 1.2561721109370586e-05, + "loss": 0.2907, + "step": 7201 + }, + { + "epoch": 0.44, + "learning_rate": 1.2559829704196593e-05, + "loss": 0.3245, + "step": 7202 + }, + { + "epoch": 0.44, + "learning_rate": 1.2557938201020339e-05, + "loss": 0.2913, + "step": 7203 + }, + { + "epoch": 0.44, + "learning_rate": 1.2556046599914236e-05, + "loss": 0.3398, + "step": 7204 + }, + { + "epoch": 0.44, + "learning_rate": 1.2554154900950708e-05, + "loss": 0.3256, + "step": 7205 + }, + { + "epoch": 0.44, + "learning_rate": 1.2552263104202173e-05, + "loss": 0.3312, + "step": 7206 + }, + { + "epoch": 0.44, + "learning_rate": 1.2550371209741068e-05, + "loss": 0.3262, + "step": 7207 + }, + { + "epoch": 0.44, + "learning_rate": 1.2548479217639811e-05, + "loss": 0.3404, + "step": 7208 + }, + { + "epoch": 0.44, + "learning_rate": 1.2546587127970841e-05, + "loss": 0.3094, + "step": 7209 + }, + { + "epoch": 0.44, + "learning_rate": 1.2544694940806602e-05, + "loss": 0.332, + "step": 7210 + }, + { + "epoch": 0.44, + "learning_rate": 1.2542802656219526e-05, + "loss": 0.339, + "step": 7211 + }, + { + "epoch": 0.44, + "learning_rate": 1.2540910274282065e-05, + "loss": 0.3267, + "step": 7212 + }, + { + "epoch": 0.44, + "learning_rate": 1.2539017795066667e-05, + "loss": 0.3271, + "step": 7213 + }, + { + "epoch": 0.44, + "learning_rate": 1.2537125218645783e-05, + "loss": 0.3379, + "step": 7214 + }, + { + "epoch": 0.44, + "learning_rate": 1.253523254509187e-05, + "loss": 0.2875, + "step": 7215 + }, + { + "epoch": 0.44, + "learning_rate": 1.2533339774477392e-05, + "loss": 0.3382, + "step": 7216 + }, + { + "epoch": 0.44, + "learning_rate": 1.253144690687481e-05, + "loss": 0.3384, + "step": 7217 + }, + { + "epoch": 0.44, + "learning_rate": 1.2529553942356594e-05, + "loss": 0.3036, + "step": 7218 + }, + { + "epoch": 0.44, + "learning_rate": 1.2527660880995209e-05, + "loss": 0.3005, + "step": 7219 + }, + { + "epoch": 0.44, + "learning_rate": 1.2525767722863141e-05, + "loss": 0.3237, + "step": 7220 + }, + { + "epoch": 0.44, + "learning_rate": 1.2523874468032857e-05, + "loss": 0.3343, + "step": 7221 + }, + { + "epoch": 0.44, + "learning_rate": 1.252198111657685e-05, + "loss": 0.3284, + "step": 7222 + }, + { + "epoch": 0.44, + "learning_rate": 1.2520087668567601e-05, + "loss": 0.3065, + "step": 7223 + }, + { + "epoch": 0.44, + "learning_rate": 1.2518194124077603e-05, + "loss": 0.3242, + "step": 7224 + }, + { + "epoch": 0.44, + "learning_rate": 1.2516300483179349e-05, + "loss": 0.3153, + "step": 7225 + }, + { + "epoch": 0.44, + "learning_rate": 1.2514406745945331e-05, + "loss": 0.3057, + "step": 7226 + }, + { + "epoch": 0.44, + "learning_rate": 1.2512512912448055e-05, + "loss": 0.3192, + "step": 7227 + }, + { + "epoch": 0.44, + "learning_rate": 1.251061898276003e-05, + "loss": 0.3245, + "step": 7228 + }, + { + "epoch": 0.44, + "learning_rate": 1.2508724956953755e-05, + "loss": 0.3225, + "step": 7229 + }, + { + "epoch": 0.44, + "learning_rate": 1.2506830835101749e-05, + "loss": 0.2959, + "step": 7230 + }, + { + "epoch": 0.44, + "learning_rate": 1.2504936617276524e-05, + "loss": 0.3248, + "step": 7231 + }, + { + "epoch": 0.44, + "learning_rate": 1.2503042303550605e-05, + "loss": 0.3111, + "step": 7232 + }, + { + "epoch": 0.44, + "learning_rate": 1.2501147893996511e-05, + "loss": 0.3217, + "step": 7233 + }, + { + "epoch": 0.44, + "learning_rate": 1.2499253388686766e-05, + "loss": 0.3052, + "step": 7234 + }, + { + "epoch": 0.44, + "learning_rate": 1.2497358787693907e-05, + "loss": 0.3209, + "step": 7235 + }, + { + "epoch": 0.44, + "learning_rate": 1.2495464091090463e-05, + "loss": 0.3524, + "step": 7236 + }, + { + "epoch": 0.44, + "learning_rate": 1.2493569298948976e-05, + "loss": 0.3145, + "step": 7237 + }, + { + "epoch": 0.44, + "learning_rate": 1.2491674411341986e-05, + "loss": 0.2955, + "step": 7238 + }, + { + "epoch": 0.44, + "learning_rate": 1.2489779428342035e-05, + "loss": 0.3345, + "step": 7239 + }, + { + "epoch": 0.44, + "learning_rate": 1.2487884350021677e-05, + "loss": 0.3245, + "step": 7240 + }, + { + "epoch": 0.44, + "learning_rate": 1.2485989176453462e-05, + "loss": 0.2951, + "step": 7241 + }, + { + "epoch": 0.44, + "learning_rate": 1.2484093907709945e-05, + "loss": 0.3198, + "step": 7242 + }, + { + "epoch": 0.44, + "learning_rate": 1.2482198543863686e-05, + "loss": 0.3304, + "step": 7243 + }, + { + "epoch": 0.44, + "learning_rate": 1.248030308498725e-05, + "loss": 0.3108, + "step": 7244 + }, + { + "epoch": 0.44, + "learning_rate": 1.2478407531153206e-05, + "loss": 0.3189, + "step": 7245 + }, + { + "epoch": 0.44, + "learning_rate": 1.2476511882434117e-05, + "loss": 0.3117, + "step": 7246 + }, + { + "epoch": 0.44, + "learning_rate": 1.2474616138902566e-05, + "loss": 0.2932, + "step": 7247 + }, + { + "epoch": 0.44, + "learning_rate": 1.2472720300631127e-05, + "loss": 0.3376, + "step": 7248 + }, + { + "epoch": 0.44, + "learning_rate": 1.2470824367692378e-05, + "loss": 0.2967, + "step": 7249 + }, + { + "epoch": 0.44, + "learning_rate": 1.2468928340158913e-05, + "loss": 0.3301, + "step": 7250 + }, + { + "epoch": 0.44, + "learning_rate": 1.2467032218103312e-05, + "loss": 0.3206, + "step": 7251 + }, + { + "epoch": 0.44, + "learning_rate": 1.2465136001598172e-05, + "loss": 0.3206, + "step": 7252 + }, + { + "epoch": 0.44, + "learning_rate": 1.2463239690716091e-05, + "loss": 0.3214, + "step": 7253 + }, + { + "epoch": 0.44, + "learning_rate": 1.2461343285529661e-05, + "loss": 0.3412, + "step": 7254 + }, + { + "epoch": 0.44, + "learning_rate": 1.2459446786111496e-05, + "loss": 0.3368, + "step": 7255 + }, + { + "epoch": 0.44, + "learning_rate": 1.2457550192534191e-05, + "loss": 0.3438, + "step": 7256 + }, + { + "epoch": 0.44, + "learning_rate": 1.2455653504870366e-05, + "loss": 0.3309, + "step": 7257 + }, + { + "epoch": 0.44, + "learning_rate": 1.2453756723192631e-05, + "loss": 0.3242, + "step": 7258 + }, + { + "epoch": 0.44, + "learning_rate": 1.2451859847573606e-05, + "loss": 0.361, + "step": 7259 + }, + { + "epoch": 0.44, + "learning_rate": 1.244996287808591e-05, + "loss": 0.3206, + "step": 7260 + }, + { + "epoch": 0.44, + "learning_rate": 1.2448065814802166e-05, + "loss": 0.3122, + "step": 7261 + }, + { + "epoch": 0.44, + "learning_rate": 1.2446168657795008e-05, + "loss": 0.32, + "step": 7262 + }, + { + "epoch": 0.44, + "learning_rate": 1.2444271407137063e-05, + "loss": 0.2958, + "step": 7263 + }, + { + "epoch": 0.44, + "learning_rate": 1.2442374062900968e-05, + "loss": 0.3111, + "step": 7264 + }, + { + "epoch": 0.44, + "learning_rate": 1.2440476625159363e-05, + "loss": 0.3103, + "step": 7265 + }, + { + "epoch": 0.44, + "learning_rate": 1.2438579093984891e-05, + "loss": 0.3404, + "step": 7266 + }, + { + "epoch": 0.44, + "learning_rate": 1.2436681469450198e-05, + "loss": 0.3292, + "step": 7267 + }, + { + "epoch": 0.44, + "learning_rate": 1.2434783751627935e-05, + "loss": 0.3184, + "step": 7268 + }, + { + "epoch": 0.44, + "learning_rate": 1.2432885940590752e-05, + "loss": 0.3278, + "step": 7269 + }, + { + "epoch": 0.44, + "learning_rate": 1.2430988036411311e-05, + "loss": 0.3262, + "step": 7270 + }, + { + "epoch": 0.44, + "learning_rate": 1.2429090039162265e-05, + "loss": 0.3178, + "step": 7271 + }, + { + "epoch": 0.44, + "learning_rate": 1.242719194891629e-05, + "loss": 0.2888, + "step": 7272 + }, + { + "epoch": 0.44, + "learning_rate": 1.2425293765746045e-05, + "loss": 0.3284, + "step": 7273 + }, + { + "epoch": 0.44, + "learning_rate": 1.2423395489724201e-05, + "loss": 0.3044, + "step": 7274 + }, + { + "epoch": 0.44, + "learning_rate": 1.2421497120923443e-05, + "loss": 0.3552, + "step": 7275 + }, + { + "epoch": 0.44, + "learning_rate": 1.2419598659416434e-05, + "loss": 0.3457, + "step": 7276 + }, + { + "epoch": 0.44, + "learning_rate": 1.2417700105275868e-05, + "loss": 0.3234, + "step": 7277 + }, + { + "epoch": 0.44, + "learning_rate": 1.2415801458574422e-05, + "loss": 0.2945, + "step": 7278 + }, + { + "epoch": 0.44, + "learning_rate": 1.2413902719384791e-05, + "loss": 0.317, + "step": 7279 + }, + { + "epoch": 0.44, + "learning_rate": 1.241200388777967e-05, + "loss": 0.3479, + "step": 7280 + }, + { + "epoch": 0.44, + "learning_rate": 1.2410104963831746e-05, + "loss": 0.2999, + "step": 7281 + }, + { + "epoch": 0.44, + "learning_rate": 1.2408205947613727e-05, + "loss": 0.2888, + "step": 7282 + }, + { + "epoch": 0.44, + "learning_rate": 1.2406306839198312e-05, + "loss": 0.3103, + "step": 7283 + }, + { + "epoch": 0.44, + "learning_rate": 1.2404407638658209e-05, + "loss": 0.3376, + "step": 7284 + }, + { + "epoch": 0.44, + "learning_rate": 1.240250834606613e-05, + "loss": 0.327, + "step": 7285 + }, + { + "epoch": 0.44, + "learning_rate": 1.2400608961494783e-05, + "loss": 0.3147, + "step": 7286 + }, + { + "epoch": 0.44, + "learning_rate": 1.2398709485016892e-05, + "loss": 0.3073, + "step": 7287 + }, + { + "epoch": 0.44, + "learning_rate": 1.2396809916705177e-05, + "loss": 0.3242, + "step": 7288 + }, + { + "epoch": 0.44, + "learning_rate": 1.2394910256632355e-05, + "loss": 0.3248, + "step": 7289 + }, + { + "epoch": 0.44, + "learning_rate": 1.2393010504871166e-05, + "loss": 0.3407, + "step": 7290 + }, + { + "epoch": 0.44, + "learning_rate": 1.239111066149433e-05, + "loss": 0.344, + "step": 7291 + }, + { + "epoch": 0.44, + "learning_rate": 1.238921072657459e-05, + "loss": 0.3244, + "step": 7292 + }, + { + "epoch": 0.44, + "learning_rate": 1.2387310700184678e-05, + "loss": 0.3104, + "step": 7293 + }, + { + "epoch": 0.44, + "learning_rate": 1.2385410582397341e-05, + "loss": 0.3066, + "step": 7294 + }, + { + "epoch": 0.44, + "learning_rate": 1.2383510373285324e-05, + "loss": 0.3214, + "step": 7295 + }, + { + "epoch": 0.44, + "learning_rate": 1.238161007292137e-05, + "loss": 0.3373, + "step": 7296 + }, + { + "epoch": 0.44, + "learning_rate": 1.237970968137824e-05, + "loss": 0.3228, + "step": 7297 + }, + { + "epoch": 0.44, + "learning_rate": 1.2377809198728683e-05, + "loss": 0.297, + "step": 7298 + }, + { + "epoch": 0.44, + "learning_rate": 1.2375908625045461e-05, + "loss": 0.3253, + "step": 7299 + }, + { + "epoch": 0.44, + "learning_rate": 1.2374007960401337e-05, + "loss": 0.3214, + "step": 7300 + }, + { + "epoch": 0.44, + "learning_rate": 1.2372107204869078e-05, + "loss": 0.3217, + "step": 7301 + }, + { + "epoch": 0.44, + "learning_rate": 1.2370206358521452e-05, + "loss": 0.3237, + "step": 7302 + }, + { + "epoch": 0.44, + "learning_rate": 1.2368305421431235e-05, + "loss": 0.3371, + "step": 7303 + }, + { + "epoch": 0.44, + "learning_rate": 1.2366404393671202e-05, + "loss": 0.3516, + "step": 7304 + }, + { + "epoch": 0.44, + "learning_rate": 1.236450327531413e-05, + "loss": 0.3326, + "step": 7305 + }, + { + "epoch": 0.44, + "learning_rate": 1.2362602066432807e-05, + "loss": 0.3298, + "step": 7306 + }, + { + "epoch": 0.44, + "learning_rate": 1.2360700767100022e-05, + "loss": 0.2919, + "step": 7307 + }, + { + "epoch": 0.44, + "learning_rate": 1.2358799377388562e-05, + "loss": 0.3284, + "step": 7308 + }, + { + "epoch": 0.44, + "learning_rate": 1.2356897897371221e-05, + "loss": 0.3438, + "step": 7309 + }, + { + "epoch": 0.44, + "learning_rate": 1.2354996327120799e-05, + "loss": 0.3359, + "step": 7310 + }, + { + "epoch": 0.44, + "learning_rate": 1.2353094666710093e-05, + "loss": 0.3471, + "step": 7311 + }, + { + "epoch": 0.44, + "learning_rate": 1.2351192916211911e-05, + "loss": 0.3033, + "step": 7312 + }, + { + "epoch": 0.44, + "learning_rate": 1.234929107569906e-05, + "loss": 0.3362, + "step": 7313 + }, + { + "epoch": 0.44, + "learning_rate": 1.234738914524435e-05, + "loss": 0.3298, + "step": 7314 + }, + { + "epoch": 0.44, + "learning_rate": 1.2345487124920599e-05, + "loss": 0.3189, + "step": 7315 + }, + { + "epoch": 0.44, + "learning_rate": 1.234358501480062e-05, + "loss": 0.3245, + "step": 7316 + }, + { + "epoch": 0.44, + "learning_rate": 1.2341682814957241e-05, + "loss": 0.3326, + "step": 7317 + }, + { + "epoch": 0.44, + "learning_rate": 1.2339780525463283e-05, + "loss": 0.3298, + "step": 7318 + }, + { + "epoch": 0.44, + "learning_rate": 1.2337878146391575e-05, + "loss": 0.3273, + "step": 7319 + }, + { + "epoch": 0.44, + "learning_rate": 1.2335975677814951e-05, + "loss": 0.3005, + "step": 7320 + }, + { + "epoch": 0.44, + "learning_rate": 1.2334073119806243e-05, + "loss": 0.2999, + "step": 7321 + }, + { + "epoch": 0.44, + "learning_rate": 1.2332170472438295e-05, + "loss": 0.308, + "step": 7322 + }, + { + "epoch": 0.44, + "learning_rate": 1.2330267735783942e-05, + "loss": 0.3228, + "step": 7323 + }, + { + "epoch": 0.44, + "learning_rate": 1.2328364909916035e-05, + "loss": 0.3354, + "step": 7324 + }, + { + "epoch": 0.44, + "learning_rate": 1.2326461994907423e-05, + "loss": 0.2932, + "step": 7325 + }, + { + "epoch": 0.44, + "learning_rate": 1.2324558990830956e-05, + "loss": 0.3055, + "step": 7326 + }, + { + "epoch": 0.44, + "learning_rate": 1.2322655897759494e-05, + "loss": 0.353, + "step": 7327 + }, + { + "epoch": 0.44, + "learning_rate": 1.232075271576589e-05, + "loss": 0.3066, + "step": 7328 + }, + { + "epoch": 0.44, + "learning_rate": 1.2318849444923011e-05, + "loss": 0.3025, + "step": 7329 + }, + { + "epoch": 0.44, + "learning_rate": 1.2316946085303726e-05, + "loss": 0.3011, + "step": 7330 + }, + { + "epoch": 0.44, + "learning_rate": 1.2315042636980894e-05, + "loss": 0.3198, + "step": 7331 + }, + { + "epoch": 0.44, + "learning_rate": 1.23131391000274e-05, + "loss": 0.3231, + "step": 7332 + }, + { + "epoch": 0.44, + "learning_rate": 1.231123547451611e-05, + "loss": 0.3016, + "step": 7333 + }, + { + "epoch": 0.44, + "learning_rate": 1.2309331760519913e-05, + "loss": 0.2972, + "step": 7334 + }, + { + "epoch": 0.44, + "learning_rate": 1.2307427958111688e-05, + "loss": 0.3122, + "step": 7335 + }, + { + "epoch": 0.44, + "learning_rate": 1.2305524067364319e-05, + "loss": 0.3306, + "step": 7336 + }, + { + "epoch": 0.44, + "learning_rate": 1.23036200883507e-05, + "loss": 0.3248, + "step": 7337 + }, + { + "epoch": 0.44, + "learning_rate": 1.2301716021143721e-05, + "loss": 0.3315, + "step": 7338 + }, + { + "epoch": 0.44, + "learning_rate": 1.229981186581628e-05, + "loss": 0.3111, + "step": 7339 + }, + { + "epoch": 0.44, + "learning_rate": 1.2297907622441279e-05, + "loss": 0.3359, + "step": 7340 + }, + { + "epoch": 0.44, + "learning_rate": 1.2296003291091616e-05, + "loss": 0.3237, + "step": 7341 + }, + { + "epoch": 0.44, + "learning_rate": 1.2294098871840204e-05, + "loss": 0.3036, + "step": 7342 + }, + { + "epoch": 0.44, + "learning_rate": 1.2292194364759949e-05, + "loss": 0.3265, + "step": 7343 + }, + { + "epoch": 0.44, + "learning_rate": 1.2290289769923765e-05, + "loss": 0.3172, + "step": 7344 + }, + { + "epoch": 0.44, + "learning_rate": 1.228838508740457e-05, + "loss": 0.3438, + "step": 7345 + }, + { + "epoch": 0.44, + "learning_rate": 1.2286480317275283e-05, + "loss": 0.3149, + "step": 7346 + }, + { + "epoch": 0.44, + "learning_rate": 1.2284575459608828e-05, + "loss": 0.3125, + "step": 7347 + }, + { + "epoch": 0.44, + "learning_rate": 1.228267051447813e-05, + "loss": 0.3145, + "step": 7348 + }, + { + "epoch": 0.44, + "learning_rate": 1.2280765481956123e-05, + "loss": 0.3298, + "step": 7349 + }, + { + "epoch": 0.44, + "learning_rate": 1.2278860362115739e-05, + "loss": 0.3315, + "step": 7350 + }, + { + "epoch": 0.44, + "learning_rate": 1.2276955155029913e-05, + "loss": 0.303, + "step": 7351 + }, + { + "epoch": 0.44, + "learning_rate": 1.2275049860771589e-05, + "loss": 0.3111, + "step": 7352 + }, + { + "epoch": 0.44, + "learning_rate": 1.2273144479413707e-05, + "loss": 0.323, + "step": 7353 + }, + { + "epoch": 0.44, + "learning_rate": 1.2271239011029215e-05, + "loss": 0.3482, + "step": 7354 + }, + { + "epoch": 0.44, + "learning_rate": 1.2269333455691066e-05, + "loss": 0.3145, + "step": 7355 + }, + { + "epoch": 0.44, + "learning_rate": 1.2267427813472205e-05, + "loss": 0.3273, + "step": 7356 + }, + { + "epoch": 0.44, + "learning_rate": 1.2265522084445602e-05, + "loss": 0.3803, + "step": 7357 + }, + { + "epoch": 0.44, + "learning_rate": 1.2263616268684207e-05, + "loss": 0.3133, + "step": 7358 + }, + { + "epoch": 0.44, + "learning_rate": 1.2261710366260986e-05, + "loss": 0.292, + "step": 7359 + }, + { + "epoch": 0.44, + "learning_rate": 1.2259804377248912e-05, + "loss": 0.3446, + "step": 7360 + }, + { + "epoch": 0.44, + "learning_rate": 1.2257898301720943e-05, + "loss": 0.3262, + "step": 7361 + }, + { + "epoch": 0.44, + "learning_rate": 1.2255992139750064e-05, + "loss": 0.3125, + "step": 7362 + }, + { + "epoch": 0.44, + "learning_rate": 1.2254085891409241e-05, + "loss": 0.3326, + "step": 7363 + }, + { + "epoch": 0.44, + "learning_rate": 1.2252179556771467e-05, + "loss": 0.3312, + "step": 7364 + }, + { + "epoch": 0.44, + "learning_rate": 1.2250273135909716e-05, + "loss": 0.298, + "step": 7365 + }, + { + "epoch": 0.44, + "learning_rate": 1.2248366628896977e-05, + "loss": 0.3401, + "step": 7366 + }, + { + "epoch": 0.45, + "learning_rate": 1.2246460035806242e-05, + "loss": 0.3189, + "step": 7367 + }, + { + "epoch": 0.45, + "learning_rate": 1.22445533567105e-05, + "loss": 0.32, + "step": 7368 + }, + { + "epoch": 0.45, + "learning_rate": 1.2242646591682752e-05, + "loss": 0.305, + "step": 7369 + }, + { + "epoch": 0.45, + "learning_rate": 1.2240739740795998e-05, + "loss": 0.3398, + "step": 7370 + }, + { + "epoch": 0.45, + "learning_rate": 1.2238832804123235e-05, + "loss": 0.2868, + "step": 7371 + }, + { + "epoch": 0.45, + "learning_rate": 1.2236925781737476e-05, + "loss": 0.3069, + "step": 7372 + }, + { + "epoch": 0.45, + "learning_rate": 1.2235018673711729e-05, + "loss": 0.3075, + "step": 7373 + }, + { + "epoch": 0.45, + "learning_rate": 1.2233111480119005e-05, + "loss": 0.3218, + "step": 7374 + }, + { + "epoch": 0.45, + "learning_rate": 1.2231204201032325e-05, + "loss": 0.3251, + "step": 7375 + }, + { + "epoch": 0.45, + "learning_rate": 1.22292968365247e-05, + "loss": 0.3111, + "step": 7376 + }, + { + "epoch": 0.45, + "learning_rate": 1.2227389386669166e-05, + "loss": 0.3055, + "step": 7377 + }, + { + "epoch": 0.45, + "learning_rate": 1.2225481851538737e-05, + "loss": 0.3365, + "step": 7378 + }, + { + "epoch": 0.45, + "learning_rate": 1.2223574231206445e-05, + "loss": 0.3136, + "step": 7379 + }, + { + "epoch": 0.45, + "learning_rate": 1.2221666525745328e-05, + "loss": 0.3521, + "step": 7380 + }, + { + "epoch": 0.45, + "learning_rate": 1.2219758735228415e-05, + "loss": 0.3175, + "step": 7381 + }, + { + "epoch": 0.45, + "learning_rate": 1.221785085972875e-05, + "loss": 0.3323, + "step": 7382 + }, + { + "epoch": 0.45, + "learning_rate": 1.2215942899319372e-05, + "loss": 0.3281, + "step": 7383 + }, + { + "epoch": 0.45, + "learning_rate": 1.2214034854073333e-05, + "loss": 0.3161, + "step": 7384 + }, + { + "epoch": 0.45, + "learning_rate": 1.2212126724063676e-05, + "loss": 0.3443, + "step": 7385 + }, + { + "epoch": 0.45, + "learning_rate": 1.2210218509363452e-05, + "loss": 0.3337, + "step": 7386 + }, + { + "epoch": 0.45, + "learning_rate": 1.220831021004572e-05, + "loss": 0.3125, + "step": 7387 + }, + { + "epoch": 0.45, + "learning_rate": 1.2206401826183538e-05, + "loss": 0.2905, + "step": 7388 + }, + { + "epoch": 0.45, + "learning_rate": 1.2204493357849967e-05, + "loss": 0.3223, + "step": 7389 + }, + { + "epoch": 0.45, + "learning_rate": 1.2202584805118071e-05, + "loss": 0.3583, + "step": 7390 + }, + { + "epoch": 0.45, + "learning_rate": 1.2200676168060921e-05, + "loss": 0.3616, + "step": 7391 + }, + { + "epoch": 0.45, + "learning_rate": 1.2198767446751589e-05, + "loss": 0.2949, + "step": 7392 + }, + { + "epoch": 0.45, + "learning_rate": 1.2196858641263147e-05, + "loss": 0.3225, + "step": 7393 + }, + { + "epoch": 0.45, + "learning_rate": 1.2194949751668677e-05, + "loss": 0.3079, + "step": 7394 + }, + { + "epoch": 0.45, + "learning_rate": 1.2193040778041256e-05, + "loss": 0.2938, + "step": 7395 + }, + { + "epoch": 0.45, + "learning_rate": 1.2191131720453966e-05, + "loss": 0.3186, + "step": 7396 + }, + { + "epoch": 0.45, + "learning_rate": 1.2189222578979902e-05, + "loss": 0.3619, + "step": 7397 + }, + { + "epoch": 0.45, + "learning_rate": 1.218731335369215e-05, + "loss": 0.356, + "step": 7398 + }, + { + "epoch": 0.45, + "learning_rate": 1.2185404044663806e-05, + "loss": 0.3164, + "step": 7399 + }, + { + "epoch": 0.45, + "learning_rate": 1.218349465196797e-05, + "loss": 0.3195, + "step": 7400 + }, + { + "epoch": 0.45, + "learning_rate": 1.2181585175677734e-05, + "loss": 0.3105, + "step": 7401 + }, + { + "epoch": 0.45, + "learning_rate": 1.2179675615866209e-05, + "loss": 0.3016, + "step": 7402 + }, + { + "epoch": 0.45, + "learning_rate": 1.21777659726065e-05, + "loss": 0.3011, + "step": 7403 + }, + { + "epoch": 0.45, + "learning_rate": 1.2175856245971718e-05, + "loss": 0.2983, + "step": 7404 + }, + { + "epoch": 0.45, + "learning_rate": 1.2173946436034973e-05, + "loss": 0.3073, + "step": 7405 + }, + { + "epoch": 0.45, + "learning_rate": 1.2172036542869386e-05, + "loss": 0.2927, + "step": 7406 + }, + { + "epoch": 0.45, + "learning_rate": 1.2170126566548073e-05, + "loss": 0.293, + "step": 7407 + }, + { + "epoch": 0.45, + "learning_rate": 1.2168216507144159e-05, + "loss": 0.3334, + "step": 7408 + }, + { + "epoch": 0.45, + "learning_rate": 1.2166306364730766e-05, + "loss": 0.3451, + "step": 7409 + }, + { + "epoch": 0.45, + "learning_rate": 1.2164396139381029e-05, + "loss": 0.3306, + "step": 7410 + }, + { + "epoch": 0.45, + "learning_rate": 1.2162485831168081e-05, + "loss": 0.303, + "step": 7411 + }, + { + "epoch": 0.45, + "learning_rate": 1.216057544016505e-05, + "loss": 0.3245, + "step": 7412 + }, + { + "epoch": 0.45, + "learning_rate": 1.215866496644508e-05, + "loss": 0.3178, + "step": 7413 + }, + { + "epoch": 0.45, + "learning_rate": 1.215675441008131e-05, + "loss": 0.3161, + "step": 7414 + }, + { + "epoch": 0.45, + "learning_rate": 1.215484377114689e-05, + "loss": 0.3351, + "step": 7415 + }, + { + "epoch": 0.45, + "learning_rate": 1.2152933049714961e-05, + "loss": 0.3242, + "step": 7416 + }, + { + "epoch": 0.45, + "learning_rate": 1.2151022245858683e-05, + "loss": 0.2987, + "step": 7417 + }, + { + "epoch": 0.45, + "learning_rate": 1.2149111359651203e-05, + "loss": 0.3379, + "step": 7418 + }, + { + "epoch": 0.45, + "learning_rate": 1.2147200391165683e-05, + "loss": 0.3072, + "step": 7419 + }, + { + "epoch": 0.45, + "learning_rate": 1.2145289340475283e-05, + "loss": 0.3039, + "step": 7420 + }, + { + "epoch": 0.45, + "learning_rate": 1.2143378207653164e-05, + "loss": 0.3211, + "step": 7421 + }, + { + "epoch": 0.45, + "learning_rate": 1.2141466992772497e-05, + "loss": 0.2983, + "step": 7422 + }, + { + "epoch": 0.45, + "learning_rate": 1.2139555695906454e-05, + "loss": 0.3114, + "step": 7423 + }, + { + "epoch": 0.45, + "learning_rate": 1.21376443171282e-05, + "loss": 0.3521, + "step": 7424 + }, + { + "epoch": 0.45, + "learning_rate": 1.2135732856510918e-05, + "loss": 0.3105, + "step": 7425 + }, + { + "epoch": 0.45, + "learning_rate": 1.2133821314127786e-05, + "loss": 0.346, + "step": 7426 + }, + { + "epoch": 0.45, + "learning_rate": 1.2131909690051988e-05, + "loss": 0.3207, + "step": 7427 + }, + { + "epoch": 0.45, + "learning_rate": 1.212999798435671e-05, + "loss": 0.3465, + "step": 7428 + }, + { + "epoch": 0.45, + "learning_rate": 1.2128086197115141e-05, + "loss": 0.32, + "step": 7429 + }, + { + "epoch": 0.45, + "learning_rate": 1.212617432840047e-05, + "loss": 0.3479, + "step": 7430 + }, + { + "epoch": 0.45, + "learning_rate": 1.2124262378285893e-05, + "loss": 0.3407, + "step": 7431 + }, + { + "epoch": 0.45, + "learning_rate": 1.2122350346844614e-05, + "loss": 0.3393, + "step": 7432 + }, + { + "epoch": 0.45, + "learning_rate": 1.2120438234149828e-05, + "loss": 0.3186, + "step": 7433 + }, + { + "epoch": 0.45, + "learning_rate": 1.2118526040274743e-05, + "loss": 0.3443, + "step": 7434 + }, + { + "epoch": 0.45, + "learning_rate": 1.2116613765292564e-05, + "loss": 0.3412, + "step": 7435 + }, + { + "epoch": 0.45, + "learning_rate": 1.2114701409276503e-05, + "loss": 0.339, + "step": 7436 + }, + { + "epoch": 0.45, + "learning_rate": 1.2112788972299778e-05, + "loss": 0.3368, + "step": 7437 + }, + { + "epoch": 0.45, + "learning_rate": 1.2110876454435601e-05, + "loss": 0.3002, + "step": 7438 + }, + { + "epoch": 0.45, + "learning_rate": 1.2108963855757194e-05, + "loss": 0.3203, + "step": 7439 + }, + { + "epoch": 0.45, + "learning_rate": 1.2107051176337781e-05, + "loss": 0.2945, + "step": 7440 + }, + { + "epoch": 0.45, + "learning_rate": 1.2105138416250584e-05, + "loss": 0.3178, + "step": 7441 + }, + { + "epoch": 0.45, + "learning_rate": 1.2103225575568838e-05, + "loss": 0.3587, + "step": 7442 + }, + { + "epoch": 0.45, + "learning_rate": 1.2101312654365773e-05, + "loss": 0.308, + "step": 7443 + }, + { + "epoch": 0.45, + "learning_rate": 1.2099399652714624e-05, + "loss": 0.3006, + "step": 7444 + }, + { + "epoch": 0.45, + "learning_rate": 1.2097486570688635e-05, + "loss": 0.3055, + "step": 7445 + }, + { + "epoch": 0.45, + "learning_rate": 1.2095573408361036e-05, + "loss": 0.2999, + "step": 7446 + }, + { + "epoch": 0.45, + "learning_rate": 1.2093660165805086e-05, + "loss": 0.3267, + "step": 7447 + }, + { + "epoch": 0.45, + "learning_rate": 1.209174684309402e-05, + "loss": 0.3142, + "step": 7448 + }, + { + "epoch": 0.45, + "learning_rate": 1.2089833440301098e-05, + "loss": 0.3205, + "step": 7449 + }, + { + "epoch": 0.45, + "learning_rate": 1.208791995749957e-05, + "loss": 0.3485, + "step": 7450 + }, + { + "epoch": 0.45, + "learning_rate": 1.2086006394762692e-05, + "loss": 0.3136, + "step": 7451 + }, + { + "epoch": 0.45, + "learning_rate": 1.2084092752163729e-05, + "loss": 0.3174, + "step": 7452 + }, + { + "epoch": 0.45, + "learning_rate": 1.2082179029775937e-05, + "loss": 0.3128, + "step": 7453 + }, + { + "epoch": 0.45, + "learning_rate": 1.2080265227672591e-05, + "loss": 0.3189, + "step": 7454 + }, + { + "epoch": 0.45, + "learning_rate": 1.2078351345926955e-05, + "loss": 0.3393, + "step": 7455 + }, + { + "epoch": 0.45, + "learning_rate": 1.20764373846123e-05, + "loss": 0.3485, + "step": 7456 + }, + { + "epoch": 0.45, + "learning_rate": 1.2074523343801907e-05, + "loss": 0.2917, + "step": 7457 + }, + { + "epoch": 0.45, + "learning_rate": 1.2072609223569047e-05, + "loss": 0.317, + "step": 7458 + }, + { + "epoch": 0.45, + "learning_rate": 1.2070695023987007e-05, + "loss": 0.2944, + "step": 7459 + }, + { + "epoch": 0.45, + "learning_rate": 1.2068780745129072e-05, + "loss": 0.3513, + "step": 7460 + }, + { + "epoch": 0.45, + "learning_rate": 1.2066866387068524e-05, + "loss": 0.3008, + "step": 7461 + }, + { + "epoch": 0.45, + "learning_rate": 1.2064951949878663e-05, + "loss": 0.3295, + "step": 7462 + }, + { + "epoch": 0.45, + "learning_rate": 1.206303743363277e-05, + "loss": 0.3125, + "step": 7463 + }, + { + "epoch": 0.45, + "learning_rate": 1.206112283840415e-05, + "loss": 0.3047, + "step": 7464 + }, + { + "epoch": 0.45, + "learning_rate": 1.2059208164266106e-05, + "loss": 0.3158, + "step": 7465 + }, + { + "epoch": 0.45, + "learning_rate": 1.2057293411291928e-05, + "loss": 0.2938, + "step": 7466 + }, + { + "epoch": 0.45, + "learning_rate": 1.2055378579554935e-05, + "loss": 0.346, + "step": 7467 + }, + { + "epoch": 0.45, + "learning_rate": 1.2053463669128426e-05, + "loss": 0.3072, + "step": 7468 + }, + { + "epoch": 0.45, + "learning_rate": 1.205154868008572e-05, + "loss": 0.3331, + "step": 7469 + }, + { + "epoch": 0.45, + "learning_rate": 1.2049633612500128e-05, + "loss": 0.293, + "step": 7470 + }, + { + "epoch": 0.45, + "learning_rate": 1.2047718466444968e-05, + "loss": 0.3034, + "step": 7471 + }, + { + "epoch": 0.45, + "learning_rate": 1.2045803241993563e-05, + "loss": 0.3329, + "step": 7472 + }, + { + "epoch": 0.45, + "learning_rate": 1.2043887939219232e-05, + "loss": 0.3421, + "step": 7473 + }, + { + "epoch": 0.45, + "learning_rate": 1.2041972558195304e-05, + "loss": 0.3108, + "step": 7474 + }, + { + "epoch": 0.45, + "learning_rate": 1.2040057098995113e-05, + "loss": 0.3168, + "step": 7475 + }, + { + "epoch": 0.45, + "learning_rate": 1.2038141561691986e-05, + "loss": 0.332, + "step": 7476 + }, + { + "epoch": 0.45, + "learning_rate": 1.2036225946359261e-05, + "loss": 0.3228, + "step": 7477 + }, + { + "epoch": 0.45, + "learning_rate": 1.2034310253070278e-05, + "loss": 0.3298, + "step": 7478 + }, + { + "epoch": 0.45, + "learning_rate": 1.2032394481898378e-05, + "loss": 0.332, + "step": 7479 + }, + { + "epoch": 0.45, + "learning_rate": 1.2030478632916902e-05, + "loss": 0.3248, + "step": 7480 + }, + { + "epoch": 0.45, + "learning_rate": 1.2028562706199201e-05, + "loss": 0.3145, + "step": 7481 + }, + { + "epoch": 0.45, + "learning_rate": 1.2026646701818627e-05, + "loss": 0.3334, + "step": 7482 + }, + { + "epoch": 0.45, + "learning_rate": 1.202473061984853e-05, + "loss": 0.3298, + "step": 7483 + }, + { + "epoch": 0.45, + "learning_rate": 1.2022814460362269e-05, + "loss": 0.3502, + "step": 7484 + }, + { + "epoch": 0.45, + "learning_rate": 1.2020898223433207e-05, + "loss": 0.293, + "step": 7485 + }, + { + "epoch": 0.45, + "learning_rate": 1.2018981909134695e-05, + "loss": 0.3092, + "step": 7486 + }, + { + "epoch": 0.45, + "learning_rate": 1.2017065517540112e-05, + "loss": 0.3103, + "step": 7487 + }, + { + "epoch": 0.45, + "learning_rate": 1.2015149048722816e-05, + "loss": 0.3312, + "step": 7488 + }, + { + "epoch": 0.45, + "learning_rate": 1.2013232502756187e-05, + "loss": 0.3117, + "step": 7489 + }, + { + "epoch": 0.45, + "learning_rate": 1.2011315879713595e-05, + "loss": 0.3156, + "step": 7490 + }, + { + "epoch": 0.45, + "learning_rate": 1.2009399179668413e-05, + "loss": 0.3627, + "step": 7491 + }, + { + "epoch": 0.45, + "learning_rate": 1.2007482402694031e-05, + "loss": 0.3089, + "step": 7492 + }, + { + "epoch": 0.45, + "learning_rate": 1.2005565548863823e-05, + "loss": 0.3189, + "step": 7493 + }, + { + "epoch": 0.45, + "learning_rate": 1.2003648618251182e-05, + "loss": 0.3292, + "step": 7494 + }, + { + "epoch": 0.45, + "learning_rate": 1.2001731610929495e-05, + "loss": 0.3669, + "step": 7495 + }, + { + "epoch": 0.45, + "learning_rate": 1.1999814526972151e-05, + "loss": 0.2934, + "step": 7496 + }, + { + "epoch": 0.45, + "learning_rate": 1.199789736645255e-05, + "loss": 0.3265, + "step": 7497 + }, + { + "epoch": 0.45, + "learning_rate": 1.199598012944408e-05, + "loss": 0.3044, + "step": 7498 + }, + { + "epoch": 0.45, + "learning_rate": 1.1994062816020156e-05, + "loss": 0.3334, + "step": 7499 + }, + { + "epoch": 0.45, + "learning_rate": 1.1992145426254172e-05, + "loss": 0.3234, + "step": 7500 + }, + { + "epoch": 0.45, + "learning_rate": 1.1990227960219535e-05, + "loss": 0.3206, + "step": 7501 + }, + { + "epoch": 0.45, + "learning_rate": 1.1988310417989659e-05, + "loss": 0.3184, + "step": 7502 + }, + { + "epoch": 0.45, + "learning_rate": 1.1986392799637953e-05, + "loss": 0.3613, + "step": 7503 + }, + { + "epoch": 0.45, + "learning_rate": 1.1984475105237837e-05, + "loss": 0.2977, + "step": 7504 + }, + { + "epoch": 0.45, + "learning_rate": 1.1982557334862724e-05, + "loss": 0.3225, + "step": 7505 + }, + { + "epoch": 0.45, + "learning_rate": 1.1980639488586036e-05, + "loss": 0.3237, + "step": 7506 + }, + { + "epoch": 0.45, + "learning_rate": 1.19787215664812e-05, + "loss": 0.334, + "step": 7507 + }, + { + "epoch": 0.45, + "learning_rate": 1.197680356862164e-05, + "loss": 0.315, + "step": 7508 + }, + { + "epoch": 0.45, + "learning_rate": 1.1974885495080788e-05, + "loss": 0.3058, + "step": 7509 + }, + { + "epoch": 0.45, + "learning_rate": 1.1972967345932076e-05, + "loss": 0.3145, + "step": 7510 + }, + { + "epoch": 0.45, + "learning_rate": 1.197104912124894e-05, + "loss": 0.3146, + "step": 7511 + }, + { + "epoch": 0.45, + "learning_rate": 1.1969130821104819e-05, + "loss": 0.3172, + "step": 7512 + }, + { + "epoch": 0.45, + "learning_rate": 1.1967212445573155e-05, + "loss": 0.3493, + "step": 7513 + }, + { + "epoch": 0.45, + "learning_rate": 1.196529399472739e-05, + "loss": 0.3228, + "step": 7514 + }, + { + "epoch": 0.45, + "learning_rate": 1.1963375468640974e-05, + "loss": 0.2914, + "step": 7515 + }, + { + "epoch": 0.45, + "learning_rate": 1.1961456867387352e-05, + "loss": 0.3393, + "step": 7516 + }, + { + "epoch": 0.45, + "learning_rate": 1.1959538191039986e-05, + "loss": 0.3139, + "step": 7517 + }, + { + "epoch": 0.45, + "learning_rate": 1.1957619439672322e-05, + "loss": 0.3371, + "step": 7518 + }, + { + "epoch": 0.45, + "learning_rate": 1.1955700613357826e-05, + "loss": 0.3538, + "step": 7519 + }, + { + "epoch": 0.45, + "learning_rate": 1.195378171216996e-05, + "loss": 0.332, + "step": 7520 + }, + { + "epoch": 0.45, + "learning_rate": 1.195186273618218e-05, + "loss": 0.31, + "step": 7521 + }, + { + "epoch": 0.45, + "learning_rate": 1.1949943685467962e-05, + "loss": 0.3273, + "step": 7522 + }, + { + "epoch": 0.45, + "learning_rate": 1.1948024560100772e-05, + "loss": 0.3306, + "step": 7523 + }, + { + "epoch": 0.45, + "learning_rate": 1.1946105360154085e-05, + "loss": 0.3477, + "step": 7524 + }, + { + "epoch": 0.45, + "learning_rate": 1.1944186085701379e-05, + "loss": 0.322, + "step": 7525 + }, + { + "epoch": 0.45, + "learning_rate": 1.1942266736816129e-05, + "loss": 0.3005, + "step": 7526 + }, + { + "epoch": 0.45, + "learning_rate": 1.1940347313571816e-05, + "loss": 0.3218, + "step": 7527 + }, + { + "epoch": 0.45, + "learning_rate": 1.1938427816041927e-05, + "loss": 0.3175, + "step": 7528 + }, + { + "epoch": 0.45, + "learning_rate": 1.1936508244299949e-05, + "loss": 0.3284, + "step": 7529 + }, + { + "epoch": 0.45, + "learning_rate": 1.1934588598419375e-05, + "loss": 0.3186, + "step": 7530 + }, + { + "epoch": 0.45, + "learning_rate": 1.1932668878473691e-05, + "loss": 0.3278, + "step": 7531 + }, + { + "epoch": 0.46, + "learning_rate": 1.1930749084536399e-05, + "loss": 0.2955, + "step": 7532 + }, + { + "epoch": 0.46, + "learning_rate": 1.1928829216680993e-05, + "loss": 0.3394, + "step": 7533 + }, + { + "epoch": 0.46, + "learning_rate": 1.192690927498098e-05, + "loss": 0.3125, + "step": 7534 + }, + { + "epoch": 0.46, + "learning_rate": 1.192498925950986e-05, + "loss": 0.3039, + "step": 7535 + }, + { + "epoch": 0.46, + "learning_rate": 1.192306917034114e-05, + "loss": 0.305, + "step": 7536 + }, + { + "epoch": 0.46, + "learning_rate": 1.1921149007548333e-05, + "loss": 0.3354, + "step": 7537 + }, + { + "epoch": 0.46, + "learning_rate": 1.1919228771204951e-05, + "loss": 0.2623, + "step": 7538 + }, + { + "epoch": 0.46, + "learning_rate": 1.1917308461384508e-05, + "loss": 0.3147, + "step": 7539 + }, + { + "epoch": 0.46, + "learning_rate": 1.1915388078160526e-05, + "loss": 0.3552, + "step": 7540 + }, + { + "epoch": 0.46, + "learning_rate": 1.1913467621606521e-05, + "loss": 0.3242, + "step": 7541 + }, + { + "epoch": 0.46, + "learning_rate": 1.1911547091796021e-05, + "loss": 0.3518, + "step": 7542 + }, + { + "epoch": 0.46, + "learning_rate": 1.190962648880255e-05, + "loss": 0.3057, + "step": 7543 + }, + { + "epoch": 0.46, + "learning_rate": 1.190770581269964e-05, + "loss": 0.3451, + "step": 7544 + }, + { + "epoch": 0.46, + "learning_rate": 1.1905785063560827e-05, + "loss": 0.3086, + "step": 7545 + }, + { + "epoch": 0.46, + "learning_rate": 1.1903864241459635e-05, + "loss": 0.3045, + "step": 7546 + }, + { + "epoch": 0.46, + "learning_rate": 1.1901943346469613e-05, + "loss": 0.3147, + "step": 7547 + }, + { + "epoch": 0.46, + "learning_rate": 1.19000223786643e-05, + "loss": 0.3281, + "step": 7548 + }, + { + "epoch": 0.46, + "learning_rate": 1.1898101338117236e-05, + "loss": 0.2994, + "step": 7549 + }, + { + "epoch": 0.46, + "learning_rate": 1.1896180224901969e-05, + "loss": 0.296, + "step": 7550 + }, + { + "epoch": 0.46, + "learning_rate": 1.1894259039092046e-05, + "loss": 0.2867, + "step": 7551 + }, + { + "epoch": 0.46, + "learning_rate": 1.1892337780761026e-05, + "loss": 0.2951, + "step": 7552 + }, + { + "epoch": 0.46, + "learning_rate": 1.1890416449982451e-05, + "loss": 0.3329, + "step": 7553 + }, + { + "epoch": 0.46, + "learning_rate": 1.1888495046829893e-05, + "loss": 0.3449, + "step": 7554 + }, + { + "epoch": 0.46, + "learning_rate": 1.1886573571376907e-05, + "loss": 0.3223, + "step": 7555 + }, + { + "epoch": 0.46, + "learning_rate": 1.188465202369705e-05, + "loss": 0.3142, + "step": 7556 + }, + { + "epoch": 0.46, + "learning_rate": 1.1882730403863896e-05, + "loss": 0.3357, + "step": 7557 + }, + { + "epoch": 0.46, + "learning_rate": 1.1880808711951008e-05, + "loss": 0.3376, + "step": 7558 + }, + { + "epoch": 0.46, + "learning_rate": 1.187888694803196e-05, + "loss": 0.3175, + "step": 7559 + }, + { + "epoch": 0.46, + "learning_rate": 1.187696511218033e-05, + "loss": 0.327, + "step": 7560 + }, + { + "epoch": 0.46, + "learning_rate": 1.1875043204469684e-05, + "loss": 0.3125, + "step": 7561 + }, + { + "epoch": 0.46, + "learning_rate": 1.1873121224973613e-05, + "loss": 0.3011, + "step": 7562 + }, + { + "epoch": 0.46, + "learning_rate": 1.1871199173765693e-05, + "loss": 0.296, + "step": 7563 + }, + { + "epoch": 0.46, + "learning_rate": 1.1869277050919511e-05, + "loss": 0.3211, + "step": 7564 + }, + { + "epoch": 0.46, + "learning_rate": 1.1867354856508657e-05, + "loss": 0.2949, + "step": 7565 + }, + { + "epoch": 0.46, + "learning_rate": 1.1865432590606718e-05, + "loss": 0.3248, + "step": 7566 + }, + { + "epoch": 0.46, + "learning_rate": 1.186351025328729e-05, + "loss": 0.3198, + "step": 7567 + }, + { + "epoch": 0.46, + "learning_rate": 1.1861587844623963e-05, + "loss": 0.3025, + "step": 7568 + }, + { + "epoch": 0.46, + "learning_rate": 1.1859665364690344e-05, + "loss": 0.3036, + "step": 7569 + }, + { + "epoch": 0.46, + "learning_rate": 1.1857742813560031e-05, + "loss": 0.3189, + "step": 7570 + }, + { + "epoch": 0.46, + "learning_rate": 1.1855820191306627e-05, + "loss": 0.3418, + "step": 7571 + }, + { + "epoch": 0.46, + "learning_rate": 1.1853897498003742e-05, + "loss": 0.3477, + "step": 7572 + }, + { + "epoch": 0.46, + "learning_rate": 1.185197473372498e-05, + "loss": 0.3004, + "step": 7573 + }, + { + "epoch": 0.46, + "learning_rate": 1.1850051898543961e-05, + "loss": 0.3158, + "step": 7574 + }, + { + "epoch": 0.46, + "learning_rate": 1.1848128992534296e-05, + "loss": 0.3133, + "step": 7575 + }, + { + "epoch": 0.46, + "learning_rate": 1.1846206015769602e-05, + "loss": 0.3111, + "step": 7576 + }, + { + "epoch": 0.46, + "learning_rate": 1.1844282968323502e-05, + "loss": 0.2935, + "step": 7577 + }, + { + "epoch": 0.46, + "learning_rate": 1.1842359850269616e-05, + "loss": 0.2882, + "step": 7578 + }, + { + "epoch": 0.46, + "learning_rate": 1.1840436661681575e-05, + "loss": 0.3072, + "step": 7579 + }, + { + "epoch": 0.46, + "learning_rate": 1.1838513402633006e-05, + "loss": 0.3371, + "step": 7580 + }, + { + "epoch": 0.46, + "learning_rate": 1.1836590073197533e-05, + "loss": 0.3524, + "step": 7581 + }, + { + "epoch": 0.46, + "learning_rate": 1.1834666673448804e-05, + "loss": 0.3139, + "step": 7582 + }, + { + "epoch": 0.46, + "learning_rate": 1.1832743203460443e-05, + "loss": 0.3552, + "step": 7583 + }, + { + "epoch": 0.46, + "learning_rate": 1.1830819663306098e-05, + "loss": 0.3147, + "step": 7584 + }, + { + "epoch": 0.46, + "learning_rate": 1.1828896053059404e-05, + "loss": 0.3066, + "step": 7585 + }, + { + "epoch": 0.46, + "learning_rate": 1.1826972372794008e-05, + "loss": 0.317, + "step": 7586 + }, + { + "epoch": 0.46, + "learning_rate": 1.182504862258356e-05, + "loss": 0.3036, + "step": 7587 + }, + { + "epoch": 0.46, + "learning_rate": 1.182312480250171e-05, + "loss": 0.353, + "step": 7588 + }, + { + "epoch": 0.46, + "learning_rate": 1.1821200912622111e-05, + "loss": 0.3186, + "step": 7589 + }, + { + "epoch": 0.46, + "learning_rate": 1.1819276953018416e-05, + "loss": 0.315, + "step": 7590 + }, + { + "epoch": 0.46, + "learning_rate": 1.1817352923764283e-05, + "loss": 0.3451, + "step": 7591 + }, + { + "epoch": 0.46, + "learning_rate": 1.1815428824933375e-05, + "loss": 0.2963, + "step": 7592 + }, + { + "epoch": 0.46, + "learning_rate": 1.1813504656599355e-05, + "loss": 0.3418, + "step": 7593 + }, + { + "epoch": 0.46, + "learning_rate": 1.181158041883589e-05, + "loss": 0.3117, + "step": 7594 + }, + { + "epoch": 0.46, + "learning_rate": 1.1809656111716647e-05, + "loss": 0.2972, + "step": 7595 + }, + { + "epoch": 0.46, + "learning_rate": 1.1807731735315297e-05, + "loss": 0.3224, + "step": 7596 + }, + { + "epoch": 0.46, + "learning_rate": 1.1805807289705518e-05, + "loss": 0.2972, + "step": 7597 + }, + { + "epoch": 0.46, + "learning_rate": 1.1803882774960982e-05, + "loss": 0.3259, + "step": 7598 + }, + { + "epoch": 0.46, + "learning_rate": 1.1801958191155373e-05, + "loss": 0.3577, + "step": 7599 + }, + { + "epoch": 0.46, + "learning_rate": 1.180003353836237e-05, + "loss": 0.3033, + "step": 7600 + }, + { + "epoch": 0.46, + "learning_rate": 1.1798108816655657e-05, + "loss": 0.3295, + "step": 7601 + }, + { + "epoch": 0.46, + "learning_rate": 1.1796184026108923e-05, + "loss": 0.3033, + "step": 7602 + }, + { + "epoch": 0.46, + "learning_rate": 1.1794259166795859e-05, + "loss": 0.3005, + "step": 7603 + }, + { + "epoch": 0.46, + "learning_rate": 1.1792334238790155e-05, + "loss": 0.3262, + "step": 7604 + }, + { + "epoch": 0.46, + "learning_rate": 1.179040924216551e-05, + "loss": 0.3624, + "step": 7605 + }, + { + "epoch": 0.46, + "learning_rate": 1.1788484176995615e-05, + "loss": 0.3136, + "step": 7606 + }, + { + "epoch": 0.46, + "learning_rate": 1.178655904335418e-05, + "loss": 0.3251, + "step": 7607 + }, + { + "epoch": 0.46, + "learning_rate": 1.1784633841314898e-05, + "loss": 0.3315, + "step": 7608 + }, + { + "epoch": 0.46, + "learning_rate": 1.1782708570951484e-05, + "loss": 0.3097, + "step": 7609 + }, + { + "epoch": 0.46, + "learning_rate": 1.1780783232337642e-05, + "loss": 0.3033, + "step": 7610 + }, + { + "epoch": 0.46, + "learning_rate": 1.1778857825547081e-05, + "loss": 0.3103, + "step": 7611 + }, + { + "epoch": 0.46, + "learning_rate": 1.1776932350653519e-05, + "loss": 0.334, + "step": 7612 + }, + { + "epoch": 0.46, + "learning_rate": 1.1775006807730667e-05, + "loss": 0.3203, + "step": 7613 + }, + { + "epoch": 0.46, + "learning_rate": 1.177308119685225e-05, + "loss": 0.3136, + "step": 7614 + }, + { + "epoch": 0.46, + "learning_rate": 1.1771155518091985e-05, + "loss": 0.2881, + "step": 7615 + }, + { + "epoch": 0.46, + "learning_rate": 1.1769229771523599e-05, + "loss": 0.3287, + "step": 7616 + }, + { + "epoch": 0.46, + "learning_rate": 1.1767303957220817e-05, + "loss": 0.3331, + "step": 7617 + }, + { + "epoch": 0.46, + "learning_rate": 1.1765378075257364e-05, + "loss": 0.2872, + "step": 7618 + }, + { + "epoch": 0.46, + "learning_rate": 1.1763452125706978e-05, + "loss": 0.3329, + "step": 7619 + }, + { + "epoch": 0.46, + "learning_rate": 1.176152610864339e-05, + "loss": 0.303, + "step": 7620 + }, + { + "epoch": 0.46, + "learning_rate": 1.1759600024140339e-05, + "loss": 0.3181, + "step": 7621 + }, + { + "epoch": 0.46, + "learning_rate": 1.1757673872271565e-05, + "loss": 0.3323, + "step": 7622 + }, + { + "epoch": 0.46, + "learning_rate": 1.1755747653110805e-05, + "loss": 0.3284, + "step": 7623 + }, + { + "epoch": 0.46, + "learning_rate": 1.1753821366731812e-05, + "loss": 0.3131, + "step": 7624 + }, + { + "epoch": 0.46, + "learning_rate": 1.1751895013208325e-05, + "loss": 0.3198, + "step": 7625 + }, + { + "epoch": 0.46, + "learning_rate": 1.1749968592614098e-05, + "loss": 0.31, + "step": 7626 + }, + { + "epoch": 0.46, + "learning_rate": 1.1748042105022885e-05, + "loss": 0.3161, + "step": 7627 + }, + { + "epoch": 0.46, + "learning_rate": 1.1746115550508437e-05, + "loss": 0.3211, + "step": 7628 + }, + { + "epoch": 0.46, + "learning_rate": 1.1744188929144513e-05, + "loss": 0.3362, + "step": 7629 + }, + { + "epoch": 0.46, + "learning_rate": 1.1742262241004875e-05, + "loss": 0.3245, + "step": 7630 + }, + { + "epoch": 0.46, + "learning_rate": 1.1740335486163284e-05, + "loss": 0.3242, + "step": 7631 + }, + { + "epoch": 0.46, + "learning_rate": 1.1738408664693504e-05, + "loss": 0.3304, + "step": 7632 + }, + { + "epoch": 0.46, + "learning_rate": 1.1736481776669307e-05, + "loss": 0.3591, + "step": 7633 + }, + { + "epoch": 0.46, + "learning_rate": 1.1734554822164457e-05, + "loss": 0.3189, + "step": 7634 + }, + { + "epoch": 0.46, + "learning_rate": 1.173262780125273e-05, + "loss": 0.3371, + "step": 7635 + }, + { + "epoch": 0.46, + "learning_rate": 1.1730700714007903e-05, + "loss": 0.3008, + "step": 7636 + }, + { + "epoch": 0.46, + "learning_rate": 1.1728773560503753e-05, + "loss": 0.2997, + "step": 7637 + }, + { + "epoch": 0.46, + "learning_rate": 1.1726846340814057e-05, + "loss": 0.3186, + "step": 7638 + }, + { + "epoch": 0.46, + "learning_rate": 1.1724919055012602e-05, + "loss": 0.308, + "step": 7639 + }, + { + "epoch": 0.46, + "learning_rate": 1.1722991703173175e-05, + "loss": 0.3421, + "step": 7640 + }, + { + "epoch": 0.46, + "learning_rate": 1.1721064285369559e-05, + "loss": 0.3125, + "step": 7641 + }, + { + "epoch": 0.46, + "learning_rate": 1.1719136801675549e-05, + "loss": 0.3457, + "step": 7642 + }, + { + "epoch": 0.46, + "learning_rate": 1.1717209252164932e-05, + "loss": 0.3192, + "step": 7643 + }, + { + "epoch": 0.46, + "learning_rate": 1.1715281636911511e-05, + "loss": 0.293, + "step": 7644 + }, + { + "epoch": 0.46, + "learning_rate": 1.1713353955989082e-05, + "loss": 0.2924, + "step": 7645 + }, + { + "epoch": 0.46, + "learning_rate": 1.1711426209471441e-05, + "loss": 0.2924, + "step": 7646 + }, + { + "epoch": 0.46, + "learning_rate": 1.1709498397432397e-05, + "loss": 0.3502, + "step": 7647 + }, + { + "epoch": 0.46, + "learning_rate": 1.1707570519945753e-05, + "loss": 0.3471, + "step": 7648 + }, + { + "epoch": 0.46, + "learning_rate": 1.1705642577085317e-05, + "loss": 0.3036, + "step": 7649 + }, + { + "epoch": 0.46, + "learning_rate": 1.1703714568924901e-05, + "loss": 0.2967, + "step": 7650 + }, + { + "epoch": 0.46, + "learning_rate": 1.1701786495538318e-05, + "loss": 0.3168, + "step": 7651 + }, + { + "epoch": 0.46, + "learning_rate": 1.1699858356999383e-05, + "loss": 0.3477, + "step": 7652 + }, + { + "epoch": 0.46, + "learning_rate": 1.1697930153381911e-05, + "loss": 0.314, + "step": 7653 + }, + { + "epoch": 0.46, + "learning_rate": 1.169600188475973e-05, + "loss": 0.3136, + "step": 7654 + }, + { + "epoch": 0.46, + "learning_rate": 1.1694073551206655e-05, + "loss": 0.3198, + "step": 7655 + }, + { + "epoch": 0.46, + "learning_rate": 1.169214515279652e-05, + "loss": 0.3117, + "step": 7656 + }, + { + "epoch": 0.46, + "learning_rate": 1.1690216689603147e-05, + "loss": 0.3309, + "step": 7657 + }, + { + "epoch": 0.46, + "learning_rate": 1.1688288161700365e-05, + "loss": 0.3265, + "step": 7658 + }, + { + "epoch": 0.46, + "learning_rate": 1.1686359569162015e-05, + "loss": 0.2928, + "step": 7659 + }, + { + "epoch": 0.46, + "learning_rate": 1.1684430912061928e-05, + "loss": 0.3181, + "step": 7660 + }, + { + "epoch": 0.46, + "learning_rate": 1.168250219047394e-05, + "loss": 0.3329, + "step": 7661 + }, + { + "epoch": 0.46, + "learning_rate": 1.1680573404471894e-05, + "loss": 0.3331, + "step": 7662 + }, + { + "epoch": 0.46, + "learning_rate": 1.167864455412963e-05, + "loss": 0.3066, + "step": 7663 + }, + { + "epoch": 0.46, + "learning_rate": 1.1676715639520999e-05, + "loss": 0.356, + "step": 7664 + }, + { + "epoch": 0.46, + "learning_rate": 1.1674786660719845e-05, + "loss": 0.2829, + "step": 7665 + }, + { + "epoch": 0.46, + "learning_rate": 1.1672857617800018e-05, + "loss": 0.3097, + "step": 7666 + }, + { + "epoch": 0.46, + "learning_rate": 1.1670928510835376e-05, + "loss": 0.3393, + "step": 7667 + }, + { + "epoch": 0.46, + "learning_rate": 1.1668999339899765e-05, + "loss": 0.3357, + "step": 7668 + }, + { + "epoch": 0.46, + "learning_rate": 1.1667070105067051e-05, + "loss": 0.2882, + "step": 7669 + }, + { + "epoch": 0.46, + "learning_rate": 1.1665140806411087e-05, + "loss": 0.2988, + "step": 7670 + }, + { + "epoch": 0.46, + "learning_rate": 1.1663211444005744e-05, + "loss": 0.3253, + "step": 7671 + }, + { + "epoch": 0.46, + "learning_rate": 1.1661282017924883e-05, + "loss": 0.3025, + "step": 7672 + }, + { + "epoch": 0.46, + "learning_rate": 1.1659352528242368e-05, + "loss": 0.3359, + "step": 7673 + }, + { + "epoch": 0.46, + "learning_rate": 1.1657422975032073e-05, + "loss": 0.32, + "step": 7674 + }, + { + "epoch": 0.46, + "learning_rate": 1.165549335836787e-05, + "loss": 0.3065, + "step": 7675 + }, + { + "epoch": 0.46, + "learning_rate": 1.1653563678323629e-05, + "loss": 0.3047, + "step": 7676 + }, + { + "epoch": 0.46, + "learning_rate": 1.1651633934973237e-05, + "loss": 0.3239, + "step": 7677 + }, + { + "epoch": 0.46, + "learning_rate": 1.1649704128390563e-05, + "loss": 0.3097, + "step": 7678 + }, + { + "epoch": 0.46, + "learning_rate": 1.1647774258649496e-05, + "loss": 0.2928, + "step": 7679 + }, + { + "epoch": 0.46, + "learning_rate": 1.164584432582392e-05, + "loss": 0.3242, + "step": 7680 + }, + { + "epoch": 0.46, + "learning_rate": 1.1643914329987718e-05, + "loss": 0.3334, + "step": 7681 + }, + { + "epoch": 0.46, + "learning_rate": 1.1641984271214782e-05, + "loss": 0.308, + "step": 7682 + }, + { + "epoch": 0.46, + "learning_rate": 1.1640054149579001e-05, + "loss": 0.3156, + "step": 7683 + }, + { + "epoch": 0.46, + "learning_rate": 1.1638123965154276e-05, + "loss": 0.3345, + "step": 7684 + }, + { + "epoch": 0.46, + "learning_rate": 1.1636193718014494e-05, + "loss": 0.3036, + "step": 7685 + }, + { + "epoch": 0.46, + "learning_rate": 1.1634263408233562e-05, + "loss": 0.3164, + "step": 7686 + }, + { + "epoch": 0.46, + "learning_rate": 1.1632333035885377e-05, + "loss": 0.3164, + "step": 7687 + }, + { + "epoch": 0.46, + "learning_rate": 1.1630402601043839e-05, + "loss": 0.3036, + "step": 7688 + }, + { + "epoch": 0.46, + "learning_rate": 1.1628472103782863e-05, + "loss": 0.3359, + "step": 7689 + }, + { + "epoch": 0.46, + "learning_rate": 1.1626541544176353e-05, + "loss": 0.3203, + "step": 7690 + }, + { + "epoch": 0.46, + "learning_rate": 1.1624610922298221e-05, + "loss": 0.2864, + "step": 7691 + }, + { + "epoch": 0.46, + "learning_rate": 1.1622680238222377e-05, + "loss": 0.3457, + "step": 7692 + }, + { + "epoch": 0.46, + "learning_rate": 1.1620749492022738e-05, + "loss": 0.2999, + "step": 7693 + }, + { + "epoch": 0.46, + "learning_rate": 1.1618818683773226e-05, + "loss": 0.2952, + "step": 7694 + }, + { + "epoch": 0.46, + "learning_rate": 1.1616887813547754e-05, + "loss": 0.2934, + "step": 7695 + }, + { + "epoch": 0.46, + "learning_rate": 1.1614956881420252e-05, + "loss": 0.3097, + "step": 7696 + }, + { + "epoch": 0.46, + "learning_rate": 1.1613025887464642e-05, + "loss": 0.2969, + "step": 7697 + }, + { + "epoch": 0.47, + "learning_rate": 1.161109483175485e-05, + "loss": 0.339, + "step": 7698 + }, + { + "epoch": 0.47, + "learning_rate": 1.1609163714364809e-05, + "loss": 0.3114, + "step": 7699 + }, + { + "epoch": 0.47, + "learning_rate": 1.1607232535368449e-05, + "loss": 0.2927, + "step": 7700 + }, + { + "epoch": 0.47, + "learning_rate": 1.1605301294839706e-05, + "loss": 0.2836, + "step": 7701 + }, + { + "epoch": 0.47, + "learning_rate": 1.1603369992852515e-05, + "loss": 0.3064, + "step": 7702 + }, + { + "epoch": 0.47, + "learning_rate": 1.1601438629480816e-05, + "loss": 0.3019, + "step": 7703 + }, + { + "epoch": 0.47, + "learning_rate": 1.1599507204798552e-05, + "loss": 0.3039, + "step": 7704 + }, + { + "epoch": 0.47, + "learning_rate": 1.1597575718879665e-05, + "loss": 0.3047, + "step": 7705 + }, + { + "epoch": 0.47, + "learning_rate": 1.1595644171798104e-05, + "loss": 0.3139, + "step": 7706 + }, + { + "epoch": 0.47, + "learning_rate": 1.1593712563627815e-05, + "loss": 0.3502, + "step": 7707 + }, + { + "epoch": 0.47, + "learning_rate": 1.159178089444275e-05, + "loss": 0.2938, + "step": 7708 + }, + { + "epoch": 0.47, + "learning_rate": 1.1589849164316863e-05, + "loss": 0.3195, + "step": 7709 + }, + { + "epoch": 0.47, + "learning_rate": 1.158791737332411e-05, + "loss": 0.3337, + "step": 7710 + }, + { + "epoch": 0.47, + "learning_rate": 1.1585985521538443e-05, + "loss": 0.316, + "step": 7711 + }, + { + "epoch": 0.47, + "learning_rate": 1.1584053609033833e-05, + "loss": 0.3231, + "step": 7712 + }, + { + "epoch": 0.47, + "learning_rate": 1.1582121635884232e-05, + "loss": 0.32, + "step": 7713 + }, + { + "epoch": 0.47, + "learning_rate": 1.1580189602163611e-05, + "loss": 0.3368, + "step": 7714 + }, + { + "epoch": 0.47, + "learning_rate": 1.1578257507945941e-05, + "loss": 0.2875, + "step": 7715 + }, + { + "epoch": 0.47, + "learning_rate": 1.1576325353305182e-05, + "loss": 0.3237, + "step": 7716 + }, + { + "epoch": 0.47, + "learning_rate": 1.1574393138315311e-05, + "loss": 0.3175, + "step": 7717 + }, + { + "epoch": 0.47, + "learning_rate": 1.1572460863050305e-05, + "loss": 0.3304, + "step": 7718 + }, + { + "epoch": 0.47, + "learning_rate": 1.1570528527584138e-05, + "loss": 0.3143, + "step": 7719 + }, + { + "epoch": 0.47, + "learning_rate": 1.1568596131990783e-05, + "loss": 0.3309, + "step": 7720 + }, + { + "epoch": 0.47, + "learning_rate": 1.1566663676344232e-05, + "loss": 0.3019, + "step": 7721 + }, + { + "epoch": 0.47, + "learning_rate": 1.156473116071846e-05, + "loss": 0.3147, + "step": 7722 + }, + { + "epoch": 0.47, + "learning_rate": 1.1562798585187455e-05, + "loss": 0.3228, + "step": 7723 + }, + { + "epoch": 0.47, + "learning_rate": 1.1560865949825209e-05, + "loss": 0.2868, + "step": 7724 + }, + { + "epoch": 0.47, + "learning_rate": 1.1558933254705706e-05, + "loss": 0.3259, + "step": 7725 + }, + { + "epoch": 0.47, + "learning_rate": 1.1557000499902945e-05, + "loss": 0.2888, + "step": 7726 + }, + { + "epoch": 0.47, + "learning_rate": 1.1555067685490915e-05, + "loss": 0.3245, + "step": 7727 + }, + { + "epoch": 0.47, + "learning_rate": 1.1553134811543616e-05, + "loss": 0.3379, + "step": 7728 + }, + { + "epoch": 0.47, + "learning_rate": 1.1551201878135048e-05, + "loss": 0.3192, + "step": 7729 + }, + { + "epoch": 0.47, + "learning_rate": 1.1549268885339213e-05, + "loss": 0.3281, + "step": 7730 + }, + { + "epoch": 0.47, + "learning_rate": 1.1547335833230111e-05, + "loss": 0.3306, + "step": 7731 + }, + { + "epoch": 0.47, + "learning_rate": 1.1545402721881752e-05, + "loss": 0.3186, + "step": 7732 + }, + { + "epoch": 0.47, + "learning_rate": 1.1543469551368146e-05, + "loss": 0.3039, + "step": 7733 + }, + { + "epoch": 0.47, + "learning_rate": 1.15415363217633e-05, + "loss": 0.2916, + "step": 7734 + }, + { + "epoch": 0.47, + "learning_rate": 1.1539603033141228e-05, + "loss": 0.3315, + "step": 7735 + }, + { + "epoch": 0.47, + "learning_rate": 1.153766968557595e-05, + "loss": 0.3337, + "step": 7736 + }, + { + "epoch": 0.47, + "learning_rate": 1.1535736279141477e-05, + "loss": 0.2757, + "step": 7737 + }, + { + "epoch": 0.47, + "learning_rate": 1.153380281391183e-05, + "loss": 0.3253, + "step": 7738 + }, + { + "epoch": 0.47, + "learning_rate": 1.1531869289961035e-05, + "loss": 0.321, + "step": 7739 + }, + { + "epoch": 0.47, + "learning_rate": 1.1529935707363112e-05, + "loss": 0.3111, + "step": 7740 + }, + { + "epoch": 0.47, + "learning_rate": 1.152800206619209e-05, + "loss": 0.3337, + "step": 7741 + }, + { + "epoch": 0.47, + "learning_rate": 1.1526068366521999e-05, + "loss": 0.3276, + "step": 7742 + }, + { + "epoch": 0.47, + "learning_rate": 1.1524134608426866e-05, + "loss": 0.3133, + "step": 7743 + }, + { + "epoch": 0.47, + "learning_rate": 1.1522200791980728e-05, + "loss": 0.3357, + "step": 7744 + }, + { + "epoch": 0.47, + "learning_rate": 1.1520266917257619e-05, + "loss": 0.3234, + "step": 7745 + }, + { + "epoch": 0.47, + "learning_rate": 1.1518332984331578e-05, + "loss": 0.3131, + "step": 7746 + }, + { + "epoch": 0.47, + "learning_rate": 1.1516398993276645e-05, + "loss": 0.2913, + "step": 7747 + }, + { + "epoch": 0.47, + "learning_rate": 1.1514464944166857e-05, + "loss": 0.317, + "step": 7748 + }, + { + "epoch": 0.47, + "learning_rate": 1.1512530837076267e-05, + "loss": 0.3301, + "step": 7749 + }, + { + "epoch": 0.47, + "learning_rate": 1.1510596672078917e-05, + "loss": 0.3253, + "step": 7750 + }, + { + "epoch": 0.47, + "learning_rate": 1.1508662449248853e-05, + "loss": 0.3094, + "step": 7751 + }, + { + "epoch": 0.47, + "learning_rate": 1.1506728168660134e-05, + "loss": 0.3318, + "step": 7752 + }, + { + "epoch": 0.47, + "learning_rate": 1.1504793830386806e-05, + "loss": 0.2946, + "step": 7753 + }, + { + "epoch": 0.47, + "learning_rate": 1.1502859434502929e-05, + "loss": 0.3122, + "step": 7754 + }, + { + "epoch": 0.47, + "learning_rate": 1.1500924981082555e-05, + "loss": 0.3054, + "step": 7755 + }, + { + "epoch": 0.47, + "learning_rate": 1.1498990470199751e-05, + "loss": 0.3211, + "step": 7756 + }, + { + "epoch": 0.47, + "learning_rate": 1.1497055901928577e-05, + "loss": 0.3301, + "step": 7757 + }, + { + "epoch": 0.47, + "learning_rate": 1.1495121276343093e-05, + "loss": 0.3265, + "step": 7758 + }, + { + "epoch": 0.47, + "learning_rate": 1.1493186593517374e-05, + "loss": 0.3351, + "step": 7759 + }, + { + "epoch": 0.47, + "learning_rate": 1.1491251853525477e-05, + "loss": 0.3043, + "step": 7760 + }, + { + "epoch": 0.47, + "learning_rate": 1.1489317056441485e-05, + "loss": 0.3206, + "step": 7761 + }, + { + "epoch": 0.47, + "learning_rate": 1.1487382202339463e-05, + "loss": 0.3013, + "step": 7762 + }, + { + "epoch": 0.47, + "learning_rate": 1.1485447291293487e-05, + "loss": 0.3135, + "step": 7763 + }, + { + "epoch": 0.47, + "learning_rate": 1.148351232337764e-05, + "loss": 0.3253, + "step": 7764 + }, + { + "epoch": 0.47, + "learning_rate": 1.1481577298666e-05, + "loss": 0.3217, + "step": 7765 + }, + { + "epoch": 0.47, + "learning_rate": 1.147964221723264e-05, + "loss": 0.3147, + "step": 7766 + }, + { + "epoch": 0.47, + "learning_rate": 1.1477707079151657e-05, + "loss": 0.291, + "step": 7767 + }, + { + "epoch": 0.47, + "learning_rate": 1.1475771884497127e-05, + "loss": 0.3161, + "step": 7768 + }, + { + "epoch": 0.47, + "learning_rate": 1.1473836633343145e-05, + "loss": 0.3189, + "step": 7769 + }, + { + "epoch": 0.47, + "learning_rate": 1.1471901325763795e-05, + "loss": 0.3092, + "step": 7770 + }, + { + "epoch": 0.47, + "learning_rate": 1.1469965961833175e-05, + "loss": 0.3214, + "step": 7771 + }, + { + "epoch": 0.47, + "learning_rate": 1.146803054162538e-05, + "loss": 0.3214, + "step": 7772 + }, + { + "epoch": 0.47, + "learning_rate": 1.1466095065214501e-05, + "loss": 0.3401, + "step": 7773 + }, + { + "epoch": 0.47, + "learning_rate": 1.1464159532674645e-05, + "loss": 0.3334, + "step": 7774 + }, + { + "epoch": 0.47, + "learning_rate": 1.1462223944079907e-05, + "loss": 0.3171, + "step": 7775 + }, + { + "epoch": 0.47, + "learning_rate": 1.1460288299504394e-05, + "loss": 0.3189, + "step": 7776 + }, + { + "epoch": 0.47, + "learning_rate": 1.1458352599022211e-05, + "loss": 0.3253, + "step": 7777 + }, + { + "epoch": 0.47, + "learning_rate": 1.1456416842707463e-05, + "loss": 0.3276, + "step": 7778 + }, + { + "epoch": 0.47, + "learning_rate": 1.1454481030634263e-05, + "loss": 0.356, + "step": 7779 + }, + { + "epoch": 0.47, + "learning_rate": 1.1452545162876718e-05, + "loss": 0.3016, + "step": 7780 + }, + { + "epoch": 0.47, + "learning_rate": 1.1450609239508952e-05, + "loss": 0.3195, + "step": 7781 + }, + { + "epoch": 0.47, + "learning_rate": 1.1448673260605072e-05, + "loss": 0.3491, + "step": 7782 + }, + { + "epoch": 0.47, + "learning_rate": 1.1446737226239198e-05, + "loss": 0.296, + "step": 7783 + }, + { + "epoch": 0.47, + "learning_rate": 1.1444801136485453e-05, + "loss": 0.3507, + "step": 7784 + }, + { + "epoch": 0.47, + "learning_rate": 1.144286499141796e-05, + "loss": 0.341, + "step": 7785 + }, + { + "epoch": 0.47, + "learning_rate": 1.1440928791110842e-05, + "loss": 0.334, + "step": 7786 + }, + { + "epoch": 0.47, + "learning_rate": 1.1438992535638225e-05, + "loss": 0.3354, + "step": 7787 + }, + { + "epoch": 0.47, + "learning_rate": 1.1437056225074237e-05, + "loss": 0.2955, + "step": 7788 + }, + { + "epoch": 0.47, + "learning_rate": 1.143511985949301e-05, + "loss": 0.3125, + "step": 7789 + }, + { + "epoch": 0.47, + "learning_rate": 1.143318343896868e-05, + "loss": 0.2999, + "step": 7790 + }, + { + "epoch": 0.47, + "learning_rate": 1.1431246963575383e-05, + "loss": 0.3599, + "step": 7791 + }, + { + "epoch": 0.47, + "learning_rate": 1.142931043338725e-05, + "loss": 0.3211, + "step": 7792 + }, + { + "epoch": 0.47, + "learning_rate": 1.1427373848478422e-05, + "loss": 0.3108, + "step": 7793 + }, + { + "epoch": 0.47, + "learning_rate": 1.1425437208923047e-05, + "loss": 0.3122, + "step": 7794 + }, + { + "epoch": 0.47, + "learning_rate": 1.142350051479526e-05, + "loss": 0.3178, + "step": 7795 + }, + { + "epoch": 0.47, + "learning_rate": 1.1421563766169215e-05, + "loss": 0.2999, + "step": 7796 + }, + { + "epoch": 0.47, + "learning_rate": 1.1419626963119054e-05, + "loss": 0.3016, + "step": 7797 + }, + { + "epoch": 0.47, + "learning_rate": 1.1417690105718928e-05, + "loss": 0.3083, + "step": 7798 + }, + { + "epoch": 0.47, + "learning_rate": 1.1415753194042991e-05, + "loss": 0.3115, + "step": 7799 + }, + { + "epoch": 0.47, + "learning_rate": 1.1413816228165395e-05, + "loss": 0.2644, + "step": 7800 + }, + { + "epoch": 0.47, + "learning_rate": 1.1411879208160294e-05, + "loss": 0.3075, + "step": 7801 + }, + { + "epoch": 0.47, + "learning_rate": 1.1409942134101852e-05, + "loss": 0.2983, + "step": 7802 + }, + { + "epoch": 0.47, + "learning_rate": 1.1408005006064226e-05, + "loss": 0.298, + "step": 7803 + }, + { + "epoch": 0.47, + "learning_rate": 1.1406067824121579e-05, + "loss": 0.3273, + "step": 7804 + }, + { + "epoch": 0.47, + "learning_rate": 1.1404130588348072e-05, + "loss": 0.3287, + "step": 7805 + }, + { + "epoch": 0.47, + "learning_rate": 1.1402193298817877e-05, + "loss": 0.3225, + "step": 7806 + }, + { + "epoch": 0.47, + "learning_rate": 1.140025595560516e-05, + "loss": 0.3192, + "step": 7807 + }, + { + "epoch": 0.47, + "learning_rate": 1.1398318558784088e-05, + "loss": 0.2899, + "step": 7808 + }, + { + "epoch": 0.47, + "learning_rate": 1.139638110842884e-05, + "loss": 0.3142, + "step": 7809 + }, + { + "epoch": 0.47, + "learning_rate": 1.1394443604613583e-05, + "loss": 0.3359, + "step": 7810 + }, + { + "epoch": 0.47, + "learning_rate": 1.1392506047412504e-05, + "loss": 0.31, + "step": 7811 + }, + { + "epoch": 0.47, + "learning_rate": 1.1390568436899775e-05, + "loss": 0.3078, + "step": 7812 + }, + { + "epoch": 0.47, + "learning_rate": 1.1388630773149575e-05, + "loss": 0.3267, + "step": 7813 + }, + { + "epoch": 0.47, + "learning_rate": 1.1386693056236093e-05, + "loss": 0.3206, + "step": 7814 + }, + { + "epoch": 0.47, + "learning_rate": 1.1384755286233509e-05, + "loss": 0.327, + "step": 7815 + }, + { + "epoch": 0.47, + "learning_rate": 1.1382817463216012e-05, + "loss": 0.2759, + "step": 7816 + }, + { + "epoch": 0.47, + "learning_rate": 1.1380879587257792e-05, + "loss": 0.2829, + "step": 7817 + }, + { + "epoch": 0.47, + "learning_rate": 1.1378941658433034e-05, + "loss": 0.2999, + "step": 7818 + }, + { + "epoch": 0.47, + "learning_rate": 1.1377003676815938e-05, + "loss": 0.2912, + "step": 7819 + }, + { + "epoch": 0.47, + "learning_rate": 1.1375065642480699e-05, + "loss": 0.3016, + "step": 7820 + }, + { + "epoch": 0.47, + "learning_rate": 1.137312755550151e-05, + "loss": 0.3389, + "step": 7821 + }, + { + "epoch": 0.47, + "learning_rate": 1.1371189415952573e-05, + "loss": 0.2902, + "step": 7822 + }, + { + "epoch": 0.47, + "learning_rate": 1.1369251223908081e-05, + "loss": 0.3184, + "step": 7823 + }, + { + "epoch": 0.47, + "learning_rate": 1.136731297944225e-05, + "loss": 0.3013, + "step": 7824 + }, + { + "epoch": 0.47, + "learning_rate": 1.1365374682629275e-05, + "loss": 0.3002, + "step": 7825 + }, + { + "epoch": 0.47, + "learning_rate": 1.1363436333543369e-05, + "loss": 0.3591, + "step": 7826 + }, + { + "epoch": 0.47, + "learning_rate": 1.136149793225874e-05, + "loss": 0.3362, + "step": 7827 + }, + { + "epoch": 0.47, + "learning_rate": 1.1359559478849595e-05, + "loss": 0.3011, + "step": 7828 + }, + { + "epoch": 0.47, + "learning_rate": 1.135762097339015e-05, + "loss": 0.3337, + "step": 7829 + }, + { + "epoch": 0.47, + "learning_rate": 1.135568241595462e-05, + "loss": 0.3465, + "step": 7830 + }, + { + "epoch": 0.47, + "learning_rate": 1.1353743806617226e-05, + "loss": 0.2793, + "step": 7831 + }, + { + "epoch": 0.47, + "learning_rate": 1.135180514545218e-05, + "loss": 0.3485, + "step": 7832 + }, + { + "epoch": 0.47, + "learning_rate": 1.1349866432533705e-05, + "loss": 0.3234, + "step": 7833 + }, + { + "epoch": 0.47, + "learning_rate": 1.1347927667936027e-05, + "loss": 0.3009, + "step": 7834 + }, + { + "epoch": 0.47, + "learning_rate": 1.134598885173337e-05, + "loss": 0.3319, + "step": 7835 + }, + { + "epoch": 0.47, + "learning_rate": 1.1344049983999956e-05, + "loss": 0.3411, + "step": 7836 + }, + { + "epoch": 0.47, + "learning_rate": 1.1342111064810026e-05, + "loss": 0.3343, + "step": 7837 + }, + { + "epoch": 0.47, + "learning_rate": 1.1340172094237796e-05, + "loss": 0.3384, + "step": 7838 + }, + { + "epoch": 0.47, + "learning_rate": 1.1338233072357508e-05, + "loss": 0.3025, + "step": 7839 + }, + { + "epoch": 0.47, + "learning_rate": 1.1336293999243393e-05, + "loss": 0.3178, + "step": 7840 + }, + { + "epoch": 0.47, + "learning_rate": 1.133435487496969e-05, + "loss": 0.3245, + "step": 7841 + }, + { + "epoch": 0.47, + "learning_rate": 1.133241569961064e-05, + "loss": 0.3133, + "step": 7842 + }, + { + "epoch": 0.47, + "learning_rate": 1.1330476473240478e-05, + "loss": 0.3496, + "step": 7843 + }, + { + "epoch": 0.47, + "learning_rate": 1.1328537195933452e-05, + "loss": 0.3209, + "step": 7844 + }, + { + "epoch": 0.47, + "learning_rate": 1.13265978677638e-05, + "loss": 0.3217, + "step": 7845 + }, + { + "epoch": 0.47, + "learning_rate": 1.1324658488805775e-05, + "loss": 0.3245, + "step": 7846 + }, + { + "epoch": 0.47, + "learning_rate": 1.1322719059133624e-05, + "loss": 0.3118, + "step": 7847 + }, + { + "epoch": 0.47, + "learning_rate": 1.1320779578821597e-05, + "loss": 0.3273, + "step": 7848 + }, + { + "epoch": 0.47, + "learning_rate": 1.1318840047943944e-05, + "loss": 0.3041, + "step": 7849 + }, + { + "epoch": 0.47, + "learning_rate": 1.1316900466574921e-05, + "loss": 0.3027, + "step": 7850 + }, + { + "epoch": 0.47, + "learning_rate": 1.1314960834788788e-05, + "loss": 0.3094, + "step": 7851 + }, + { + "epoch": 0.47, + "learning_rate": 1.1313021152659798e-05, + "loss": 0.3164, + "step": 7852 + }, + { + "epoch": 0.47, + "learning_rate": 1.1311081420262211e-05, + "loss": 0.3083, + "step": 7853 + }, + { + "epoch": 0.47, + "learning_rate": 1.1309141637670296e-05, + "loss": 0.2902, + "step": 7854 + }, + { + "epoch": 0.47, + "learning_rate": 1.1307201804958307e-05, + "loss": 0.3097, + "step": 7855 + }, + { + "epoch": 0.47, + "learning_rate": 1.130526192220052e-05, + "loss": 0.2953, + "step": 7856 + }, + { + "epoch": 0.47, + "learning_rate": 1.1303321989471192e-05, + "loss": 0.3393, + "step": 7857 + }, + { + "epoch": 0.47, + "learning_rate": 1.13013820068446e-05, + "loss": 0.3064, + "step": 7858 + }, + { + "epoch": 0.47, + "learning_rate": 1.1299441974395016e-05, + "loss": 0.3114, + "step": 7859 + }, + { + "epoch": 0.47, + "learning_rate": 1.1297501892196708e-05, + "loss": 0.3443, + "step": 7860 + }, + { + "epoch": 0.47, + "learning_rate": 1.1295561760323957e-05, + "loss": 0.2852, + "step": 7861 + }, + { + "epoch": 0.47, + "learning_rate": 1.1293621578851038e-05, + "loss": 0.3026, + "step": 7862 + }, + { + "epoch": 0.48, + "learning_rate": 1.1291681347852229e-05, + "loss": 0.3055, + "step": 7863 + }, + { + "epoch": 0.48, + "learning_rate": 1.1289741067401813e-05, + "loss": 0.3189, + "step": 7864 + }, + { + "epoch": 0.48, + "learning_rate": 1.1287800737574073e-05, + "loss": 0.3075, + "step": 7865 + }, + { + "epoch": 0.48, + "learning_rate": 1.1285860358443295e-05, + "loss": 0.3198, + "step": 7866 + }, + { + "epoch": 0.48, + "learning_rate": 1.1283919930083763e-05, + "loss": 0.2934, + "step": 7867 + }, + { + "epoch": 0.48, + "learning_rate": 1.1281979452569765e-05, + "loss": 0.2974, + "step": 7868 + }, + { + "epoch": 0.48, + "learning_rate": 1.1280038925975596e-05, + "loss": 0.3078, + "step": 7869 + }, + { + "epoch": 0.48, + "learning_rate": 1.1278098350375544e-05, + "loss": 0.2934, + "step": 7870 + }, + { + "epoch": 0.48, + "learning_rate": 1.127615772584391e-05, + "loss": 0.3156, + "step": 7871 + }, + { + "epoch": 0.48, + "learning_rate": 1.1274217052454981e-05, + "loss": 0.2985, + "step": 7872 + }, + { + "epoch": 0.48, + "learning_rate": 1.127227633028306e-05, + "loss": 0.3016, + "step": 7873 + }, + { + "epoch": 0.48, + "learning_rate": 1.1270335559402448e-05, + "loss": 0.3213, + "step": 7874 + }, + { + "epoch": 0.48, + "learning_rate": 1.1268394739887442e-05, + "loss": 0.3541, + "step": 7875 + }, + { + "epoch": 0.48, + "learning_rate": 1.126645387181235e-05, + "loss": 0.3013, + "step": 7876 + }, + { + "epoch": 0.48, + "learning_rate": 1.1264512955251479e-05, + "loss": 0.3246, + "step": 7877 + }, + { + "epoch": 0.48, + "learning_rate": 1.1262571990279133e-05, + "loss": 0.3041, + "step": 7878 + }, + { + "epoch": 0.48, + "learning_rate": 1.126063097696962e-05, + "loss": 0.3387, + "step": 7879 + }, + { + "epoch": 0.48, + "learning_rate": 1.1258689915397252e-05, + "loss": 0.32, + "step": 7880 + }, + { + "epoch": 0.48, + "learning_rate": 1.1256748805636346e-05, + "loss": 0.3125, + "step": 7881 + }, + { + "epoch": 0.48, + "learning_rate": 1.1254807647761215e-05, + "loss": 0.31, + "step": 7882 + }, + { + "epoch": 0.48, + "learning_rate": 1.1252866441846169e-05, + "loss": 0.3299, + "step": 7883 + }, + { + "epoch": 0.48, + "learning_rate": 1.1250925187965536e-05, + "loss": 0.3172, + "step": 7884 + }, + { + "epoch": 0.48, + "learning_rate": 1.1248983886193629e-05, + "loss": 0.3231, + "step": 7885 + }, + { + "epoch": 0.48, + "learning_rate": 1.1247042536604775e-05, + "loss": 0.3206, + "step": 7886 + }, + { + "epoch": 0.48, + "learning_rate": 1.1245101139273297e-05, + "loss": 0.3029, + "step": 7887 + }, + { + "epoch": 0.48, + "learning_rate": 1.124315969427352e-05, + "loss": 0.3189, + "step": 7888 + }, + { + "epoch": 0.48, + "learning_rate": 1.1241218201679774e-05, + "loss": 0.3242, + "step": 7889 + }, + { + "epoch": 0.48, + "learning_rate": 1.1239276661566381e-05, + "loss": 0.2799, + "step": 7890 + }, + { + "epoch": 0.48, + "learning_rate": 1.1237335074007678e-05, + "loss": 0.2948, + "step": 7891 + }, + { + "epoch": 0.48, + "learning_rate": 1.1235393439077999e-05, + "loss": 0.3152, + "step": 7892 + }, + { + "epoch": 0.48, + "learning_rate": 1.1233451756851675e-05, + "loss": 0.3574, + "step": 7893 + }, + { + "epoch": 0.48, + "learning_rate": 1.1231510027403047e-05, + "loss": 0.2898, + "step": 7894 + }, + { + "epoch": 0.48, + "learning_rate": 1.1229568250806448e-05, + "loss": 0.3036, + "step": 7895 + }, + { + "epoch": 0.48, + "learning_rate": 1.1227626427136225e-05, + "loss": 0.3107, + "step": 7896 + }, + { + "epoch": 0.48, + "learning_rate": 1.1225684556466715e-05, + "loss": 0.2962, + "step": 7897 + }, + { + "epoch": 0.48, + "learning_rate": 1.1223742638872263e-05, + "loss": 0.3086, + "step": 7898 + }, + { + "epoch": 0.48, + "learning_rate": 1.122180067442722e-05, + "loss": 0.3119, + "step": 7899 + }, + { + "epoch": 0.48, + "learning_rate": 1.1219858663205923e-05, + "loss": 0.3206, + "step": 7900 + }, + { + "epoch": 0.48, + "learning_rate": 1.1217916605282728e-05, + "loss": 0.3125, + "step": 7901 + }, + { + "epoch": 0.48, + "learning_rate": 1.1215974500731988e-05, + "loss": 0.3291, + "step": 7902 + }, + { + "epoch": 0.48, + "learning_rate": 1.121403234962805e-05, + "loss": 0.3259, + "step": 7903 + }, + { + "epoch": 0.48, + "learning_rate": 1.1212090152045274e-05, + "loss": 0.3083, + "step": 7904 + }, + { + "epoch": 0.48, + "learning_rate": 1.1210147908058013e-05, + "loss": 0.2744, + "step": 7905 + }, + { + "epoch": 0.48, + "learning_rate": 1.1208205617740626e-05, + "loss": 0.3119, + "step": 7906 + }, + { + "epoch": 0.48, + "learning_rate": 1.1206263281167473e-05, + "loss": 0.3337, + "step": 7907 + }, + { + "epoch": 0.48, + "learning_rate": 1.1204320898412917e-05, + "loss": 0.3214, + "step": 7908 + }, + { + "epoch": 0.48, + "learning_rate": 1.1202378469551321e-05, + "loss": 0.3133, + "step": 7909 + }, + { + "epoch": 0.48, + "learning_rate": 1.1200435994657045e-05, + "loss": 0.3211, + "step": 7910 + }, + { + "epoch": 0.48, + "learning_rate": 1.1198493473804466e-05, + "loss": 0.3178, + "step": 7911 + }, + { + "epoch": 0.48, + "learning_rate": 1.1196550907067944e-05, + "loss": 0.3298, + "step": 7912 + }, + { + "epoch": 0.48, + "learning_rate": 1.1194608294521853e-05, + "loss": 0.3418, + "step": 7913 + }, + { + "epoch": 0.48, + "learning_rate": 1.119266563624057e-05, + "loss": 0.3061, + "step": 7914 + }, + { + "epoch": 0.48, + "learning_rate": 1.119072293229846e-05, + "loss": 0.3354, + "step": 7915 + }, + { + "epoch": 0.48, + "learning_rate": 1.1188780182769902e-05, + "loss": 0.3146, + "step": 7916 + }, + { + "epoch": 0.48, + "learning_rate": 1.118683738772928e-05, + "loss": 0.3019, + "step": 7917 + }, + { + "epoch": 0.48, + "learning_rate": 1.1184894547250964e-05, + "loss": 0.3142, + "step": 7918 + }, + { + "epoch": 0.48, + "learning_rate": 1.118295166140934e-05, + "loss": 0.3075, + "step": 7919 + }, + { + "epoch": 0.48, + "learning_rate": 1.1181008730278788e-05, + "loss": 0.3278, + "step": 7920 + }, + { + "epoch": 0.48, + "learning_rate": 1.1179065753933698e-05, + "loss": 0.3384, + "step": 7921 + }, + { + "epoch": 0.48, + "learning_rate": 1.1177122732448451e-05, + "loss": 0.3292, + "step": 7922 + }, + { + "epoch": 0.48, + "learning_rate": 1.1175179665897437e-05, + "loss": 0.3145, + "step": 7923 + }, + { + "epoch": 0.48, + "learning_rate": 1.1173236554355044e-05, + "loss": 0.3295, + "step": 7924 + }, + { + "epoch": 0.48, + "learning_rate": 1.1171293397895667e-05, + "loss": 0.3115, + "step": 7925 + }, + { + "epoch": 0.48, + "learning_rate": 1.1169350196593694e-05, + "loss": 0.2843, + "step": 7926 + }, + { + "epoch": 0.48, + "learning_rate": 1.1167406950523525e-05, + "loss": 0.303, + "step": 7927 + }, + { + "epoch": 0.48, + "learning_rate": 1.116546365975955e-05, + "loss": 0.3089, + "step": 7928 + }, + { + "epoch": 0.48, + "learning_rate": 1.1163520324376174e-05, + "loss": 0.3337, + "step": 7929 + }, + { + "epoch": 0.48, + "learning_rate": 1.1161576944447794e-05, + "loss": 0.3384, + "step": 7930 + }, + { + "epoch": 0.48, + "learning_rate": 1.1159633520048811e-05, + "loss": 0.3285, + "step": 7931 + }, + { + "epoch": 0.48, + "learning_rate": 1.115769005125363e-05, + "loss": 0.2916, + "step": 7932 + }, + { + "epoch": 0.48, + "learning_rate": 1.1155746538136655e-05, + "loss": 0.3362, + "step": 7933 + }, + { + "epoch": 0.48, + "learning_rate": 1.1153802980772298e-05, + "loss": 0.286, + "step": 7934 + }, + { + "epoch": 0.48, + "learning_rate": 1.1151859379234957e-05, + "loss": 0.3066, + "step": 7935 + }, + { + "epoch": 0.48, + "learning_rate": 1.114991573359905e-05, + "loss": 0.3069, + "step": 7936 + }, + { + "epoch": 0.48, + "learning_rate": 1.1147972043938988e-05, + "loss": 0.3019, + "step": 7937 + }, + { + "epoch": 0.48, + "learning_rate": 1.114602831032918e-05, + "loss": 0.3153, + "step": 7938 + }, + { + "epoch": 0.48, + "learning_rate": 1.1144084532844048e-05, + "loss": 0.311, + "step": 7939 + }, + { + "epoch": 0.48, + "learning_rate": 1.1142140711558002e-05, + "loss": 0.3206, + "step": 7940 + }, + { + "epoch": 0.48, + "learning_rate": 1.1140196846545467e-05, + "loss": 0.284, + "step": 7941 + }, + { + "epoch": 0.48, + "learning_rate": 1.1138252937880858e-05, + "loss": 0.3136, + "step": 7942 + }, + { + "epoch": 0.48, + "learning_rate": 1.1136308985638602e-05, + "loss": 0.2944, + "step": 7943 + }, + { + "epoch": 0.48, + "learning_rate": 1.1134364989893119e-05, + "loss": 0.3457, + "step": 7944 + }, + { + "epoch": 0.48, + "learning_rate": 1.1132420950718831e-05, + "loss": 0.3094, + "step": 7945 + }, + { + "epoch": 0.48, + "learning_rate": 1.1130476868190174e-05, + "loss": 0.3044, + "step": 7946 + }, + { + "epoch": 0.48, + "learning_rate": 1.112853274238157e-05, + "loss": 0.3136, + "step": 7947 + }, + { + "epoch": 0.48, + "learning_rate": 1.1126588573367447e-05, + "loss": 0.2896, + "step": 7948 + }, + { + "epoch": 0.48, + "learning_rate": 1.1124644361222246e-05, + "loss": 0.2785, + "step": 7949 + }, + { + "epoch": 0.48, + "learning_rate": 1.112270010602039e-05, + "loss": 0.3128, + "step": 7950 + }, + { + "epoch": 0.48, + "learning_rate": 1.1120755807836324e-05, + "loss": 0.2953, + "step": 7951 + }, + { + "epoch": 0.48, + "learning_rate": 1.111881146674448e-05, + "loss": 0.3012, + "step": 7952 + }, + { + "epoch": 0.48, + "learning_rate": 1.1116867082819292e-05, + "loss": 0.3136, + "step": 7953 + }, + { + "epoch": 0.48, + "learning_rate": 1.111492265613521e-05, + "loss": 0.3231, + "step": 7954 + }, + { + "epoch": 0.48, + "learning_rate": 1.1112978186766665e-05, + "loss": 0.314, + "step": 7955 + }, + { + "epoch": 0.48, + "learning_rate": 1.1111033674788112e-05, + "loss": 0.3092, + "step": 7956 + }, + { + "epoch": 0.48, + "learning_rate": 1.1109089120273986e-05, + "loss": 0.2963, + "step": 7957 + }, + { + "epoch": 0.48, + "learning_rate": 1.1107144523298741e-05, + "loss": 0.3122, + "step": 7958 + }, + { + "epoch": 0.48, + "learning_rate": 1.110519988393682e-05, + "loss": 0.3072, + "step": 7959 + }, + { + "epoch": 0.48, + "learning_rate": 1.1103255202262673e-05, + "loss": 0.3382, + "step": 7960 + }, + { + "epoch": 0.48, + "learning_rate": 1.1101310478350755e-05, + "loss": 0.29, + "step": 7961 + }, + { + "epoch": 0.48, + "learning_rate": 1.1099365712275518e-05, + "loss": 0.3569, + "step": 7962 + }, + { + "epoch": 0.48, + "learning_rate": 1.1097420904111414e-05, + "loss": 0.3072, + "step": 7963 + }, + { + "epoch": 0.48, + "learning_rate": 1.1095476053932904e-05, + "loss": 0.3108, + "step": 7964 + }, + { + "epoch": 0.48, + "learning_rate": 1.109353116181444e-05, + "loss": 0.3457, + "step": 7965 + }, + { + "epoch": 0.48, + "learning_rate": 1.1091586227830486e-05, + "loss": 0.3357, + "step": 7966 + }, + { + "epoch": 0.48, + "learning_rate": 1.1089641252055506e-05, + "loss": 0.3281, + "step": 7967 + }, + { + "epoch": 0.48, + "learning_rate": 1.1087696234563953e-05, + "loss": 0.2902, + "step": 7968 + }, + { + "epoch": 0.48, + "learning_rate": 1.1085751175430304e-05, + "loss": 0.3033, + "step": 7969 + }, + { + "epoch": 0.48, + "learning_rate": 1.1083806074729013e-05, + "loss": 0.3111, + "step": 7970 + }, + { + "epoch": 0.48, + "learning_rate": 1.1081860932534556e-05, + "loss": 0.3379, + "step": 7971 + }, + { + "epoch": 0.48, + "learning_rate": 1.1079915748921399e-05, + "loss": 0.3253, + "step": 7972 + }, + { + "epoch": 0.48, + "learning_rate": 1.1077970523964013e-05, + "loss": 0.3368, + "step": 7973 + }, + { + "epoch": 0.48, + "learning_rate": 1.107602525773687e-05, + "loss": 0.3195, + "step": 7974 + }, + { + "epoch": 0.48, + "learning_rate": 1.107407995031444e-05, + "loss": 0.3008, + "step": 7975 + }, + { + "epoch": 0.48, + "learning_rate": 1.1072134601771206e-05, + "loss": 0.3146, + "step": 7976 + }, + { + "epoch": 0.48, + "learning_rate": 1.1070189212181641e-05, + "loss": 0.2907, + "step": 7977 + }, + { + "epoch": 0.48, + "learning_rate": 1.1068243781620224e-05, + "loss": 0.2941, + "step": 7978 + }, + { + "epoch": 0.48, + "learning_rate": 1.1066298310161436e-05, + "loss": 0.3055, + "step": 7979 + }, + { + "epoch": 0.48, + "learning_rate": 1.1064352797879756e-05, + "loss": 0.2931, + "step": 7980 + }, + { + "epoch": 0.48, + "learning_rate": 1.1062407244849673e-05, + "loss": 0.2824, + "step": 7981 + }, + { + "epoch": 0.48, + "learning_rate": 1.1060461651145667e-05, + "loss": 0.2779, + "step": 7982 + }, + { + "epoch": 0.48, + "learning_rate": 1.1058516016842222e-05, + "loss": 0.3142, + "step": 7983 + }, + { + "epoch": 0.48, + "learning_rate": 1.1056570342013835e-05, + "loss": 0.3263, + "step": 7984 + }, + { + "epoch": 0.48, + "learning_rate": 1.1054624626734985e-05, + "loss": 0.332, + "step": 7985 + }, + { + "epoch": 0.48, + "learning_rate": 1.1052678871080172e-05, + "loss": 0.3273, + "step": 7986 + }, + { + "epoch": 0.48, + "learning_rate": 1.1050733075123885e-05, + "loss": 0.3262, + "step": 7987 + }, + { + "epoch": 0.48, + "learning_rate": 1.1048787238940614e-05, + "loss": 0.3689, + "step": 7988 + }, + { + "epoch": 0.48, + "learning_rate": 1.1046841362604863e-05, + "loss": 0.3538, + "step": 7989 + }, + { + "epoch": 0.48, + "learning_rate": 1.1044895446191127e-05, + "loss": 0.3237, + "step": 7990 + }, + { + "epoch": 0.48, + "learning_rate": 1.10429494897739e-05, + "loss": 0.303, + "step": 7991 + }, + { + "epoch": 0.48, + "learning_rate": 1.1041003493427683e-05, + "loss": 0.3295, + "step": 7992 + }, + { + "epoch": 0.48, + "learning_rate": 1.1039057457226982e-05, + "loss": 0.327, + "step": 7993 + }, + { + "epoch": 0.48, + "learning_rate": 1.1037111381246299e-05, + "loss": 0.2995, + "step": 7994 + }, + { + "epoch": 0.48, + "learning_rate": 1.1035165265560136e-05, + "loss": 0.3292, + "step": 7995 + }, + { + "epoch": 0.48, + "learning_rate": 1.1033219110243005e-05, + "loss": 0.2913, + "step": 7996 + }, + { + "epoch": 0.48, + "learning_rate": 1.1031272915369411e-05, + "loss": 0.3211, + "step": 7997 + }, + { + "epoch": 0.48, + "learning_rate": 1.1029326681013858e-05, + "loss": 0.32, + "step": 7998 + }, + { + "epoch": 0.48, + "learning_rate": 1.1027380407250867e-05, + "loss": 0.3075, + "step": 7999 + }, + { + "epoch": 0.48, + "learning_rate": 1.1025434094154943e-05, + "loss": 0.3133, + "step": 8000 + }, + { + "epoch": 0.48, + "learning_rate": 1.1023487741800604e-05, + "loss": 0.3131, + "step": 8001 + }, + { + "epoch": 0.48, + "learning_rate": 1.1021541350262366e-05, + "loss": 0.3058, + "step": 8002 + }, + { + "epoch": 0.48, + "learning_rate": 1.101959491961474e-05, + "loss": 0.3065, + "step": 8003 + }, + { + "epoch": 0.48, + "learning_rate": 1.101764844993225e-05, + "loss": 0.296, + "step": 8004 + }, + { + "epoch": 0.48, + "learning_rate": 1.1015701941289411e-05, + "loss": 0.3195, + "step": 8005 + }, + { + "epoch": 0.48, + "learning_rate": 1.1013755393760755e-05, + "loss": 0.3362, + "step": 8006 + }, + { + "epoch": 0.48, + "learning_rate": 1.1011808807420795e-05, + "loss": 0.3089, + "step": 8007 + }, + { + "epoch": 0.48, + "learning_rate": 1.1009862182344058e-05, + "loss": 0.3117, + "step": 8008 + }, + { + "epoch": 0.48, + "learning_rate": 1.1007915518605068e-05, + "loss": 0.3001, + "step": 8009 + }, + { + "epoch": 0.48, + "learning_rate": 1.1005968816278357e-05, + "loss": 0.3131, + "step": 8010 + }, + { + "epoch": 0.48, + "learning_rate": 1.100402207543845e-05, + "loss": 0.3206, + "step": 8011 + }, + { + "epoch": 0.48, + "learning_rate": 1.1002075296159877e-05, + "loss": 0.3036, + "step": 8012 + }, + { + "epoch": 0.48, + "learning_rate": 1.1000128478517176e-05, + "loss": 0.3103, + "step": 8013 + }, + { + "epoch": 0.48, + "learning_rate": 1.0998181622584873e-05, + "loss": 0.3309, + "step": 8014 + }, + { + "epoch": 0.48, + "learning_rate": 1.0996234728437507e-05, + "loss": 0.3061, + "step": 8015 + }, + { + "epoch": 0.48, + "learning_rate": 1.0994287796149612e-05, + "loss": 0.3105, + "step": 8016 + }, + { + "epoch": 0.48, + "learning_rate": 1.0992340825795728e-05, + "loss": 0.3097, + "step": 8017 + }, + { + "epoch": 0.48, + "learning_rate": 1.0990393817450391e-05, + "loss": 0.2845, + "step": 8018 + }, + { + "epoch": 0.48, + "learning_rate": 1.0988446771188144e-05, + "loss": 0.2902, + "step": 8019 + }, + { + "epoch": 0.48, + "learning_rate": 1.0986499687083528e-05, + "loss": 0.3044, + "step": 8020 + }, + { + "epoch": 0.48, + "learning_rate": 1.0984552565211088e-05, + "loss": 0.3267, + "step": 8021 + }, + { + "epoch": 0.48, + "learning_rate": 1.0982605405645368e-05, + "loss": 0.2949, + "step": 8022 + }, + { + "epoch": 0.48, + "learning_rate": 1.098065820846091e-05, + "loss": 0.3111, + "step": 8023 + }, + { + "epoch": 0.48, + "learning_rate": 1.0978710973732274e-05, + "loss": 0.3217, + "step": 8024 + }, + { + "epoch": 0.48, + "learning_rate": 1.0976763701533995e-05, + "loss": 0.3262, + "step": 8025 + }, + { + "epoch": 0.48, + "learning_rate": 1.0974816391940631e-05, + "loss": 0.3304, + "step": 8026 + }, + { + "epoch": 0.48, + "learning_rate": 1.0972869045026733e-05, + "loss": 0.2972, + "step": 8027 + }, + { + "epoch": 0.48, + "learning_rate": 1.0970921660866855e-05, + "loss": 0.3094, + "step": 8028 + }, + { + "epoch": 0.49, + "learning_rate": 1.0968974239535552e-05, + "loss": 0.3318, + "step": 8029 + }, + { + "epoch": 0.49, + "learning_rate": 1.096702678110738e-05, + "loss": 0.3087, + "step": 8030 + }, + { + "epoch": 0.49, + "learning_rate": 1.0965079285656896e-05, + "loss": 0.2987, + "step": 8031 + }, + { + "epoch": 0.49, + "learning_rate": 1.096313175325866e-05, + "loss": 0.3022, + "step": 8032 + }, + { + "epoch": 0.49, + "learning_rate": 1.0961184183987232e-05, + "loss": 0.2893, + "step": 8033 + }, + { + "epoch": 0.49, + "learning_rate": 1.0959236577917178e-05, + "loss": 0.3083, + "step": 8034 + }, + { + "epoch": 0.49, + "learning_rate": 1.0957288935123056e-05, + "loss": 0.2872, + "step": 8035 + }, + { + "epoch": 0.49, + "learning_rate": 1.0955341255679433e-05, + "loss": 0.3451, + "step": 8036 + }, + { + "epoch": 0.49, + "learning_rate": 1.0953393539660878e-05, + "loss": 0.3122, + "step": 8037 + }, + { + "epoch": 0.49, + "learning_rate": 1.0951445787141953e-05, + "loss": 0.3086, + "step": 8038 + }, + { + "epoch": 0.49, + "learning_rate": 1.0949497998197231e-05, + "loss": 0.3262, + "step": 8039 + }, + { + "epoch": 0.49, + "learning_rate": 1.094755017290128e-05, + "loss": 0.3062, + "step": 8040 + }, + { + "epoch": 0.49, + "learning_rate": 1.0945602311328679e-05, + "loss": 0.3002, + "step": 8041 + }, + { + "epoch": 0.49, + "learning_rate": 1.094365441355399e-05, + "loss": 0.3156, + "step": 8042 + }, + { + "epoch": 0.49, + "learning_rate": 1.0941706479651797e-05, + "loss": 0.339, + "step": 8043 + }, + { + "epoch": 0.49, + "learning_rate": 1.0939758509696672e-05, + "loss": 0.2932, + "step": 8044 + }, + { + "epoch": 0.49, + "learning_rate": 1.093781050376319e-05, + "loss": 0.3251, + "step": 8045 + }, + { + "epoch": 0.49, + "learning_rate": 1.0935862461925937e-05, + "loss": 0.3384, + "step": 8046 + }, + { + "epoch": 0.49, + "learning_rate": 1.0933914384259484e-05, + "loss": 0.3139, + "step": 8047 + }, + { + "epoch": 0.49, + "learning_rate": 1.093196627083842e-05, + "loss": 0.3223, + "step": 8048 + }, + { + "epoch": 0.49, + "learning_rate": 1.0930018121737326e-05, + "loss": 0.3449, + "step": 8049 + }, + { + "epoch": 0.49, + "learning_rate": 1.0928069937030782e-05, + "loss": 0.2955, + "step": 8050 + }, + { + "epoch": 0.49, + "learning_rate": 1.0926121716793383e-05, + "loss": 0.2902, + "step": 8051 + }, + { + "epoch": 0.49, + "learning_rate": 1.0924173461099707e-05, + "loss": 0.3327, + "step": 8052 + }, + { + "epoch": 0.49, + "learning_rate": 1.0922225170024343e-05, + "loss": 0.3059, + "step": 8053 + }, + { + "epoch": 0.49, + "learning_rate": 1.0920276843641889e-05, + "loss": 0.3178, + "step": 8054 + }, + { + "epoch": 0.49, + "learning_rate": 1.0918328482026923e-05, + "loss": 0.3306, + "step": 8055 + }, + { + "epoch": 0.49, + "learning_rate": 1.091638008525405e-05, + "loss": 0.2972, + "step": 8056 + }, + { + "epoch": 0.49, + "learning_rate": 1.0914431653397856e-05, + "loss": 0.2944, + "step": 8057 + }, + { + "epoch": 0.49, + "learning_rate": 1.0912483186532941e-05, + "loss": 0.3248, + "step": 8058 + }, + { + "epoch": 0.49, + "learning_rate": 1.0910534684733897e-05, + "loss": 0.3072, + "step": 8059 + }, + { + "epoch": 0.49, + "learning_rate": 1.0908586148075324e-05, + "loss": 0.3186, + "step": 8060 + }, + { + "epoch": 0.49, + "learning_rate": 1.0906637576631822e-05, + "loss": 0.3181, + "step": 8061 + }, + { + "epoch": 0.49, + "learning_rate": 1.0904688970477989e-05, + "loss": 0.3027, + "step": 8062 + }, + { + "epoch": 0.49, + "learning_rate": 1.0902740329688426e-05, + "loss": 0.2983, + "step": 8063 + }, + { + "epoch": 0.49, + "learning_rate": 1.0900791654337742e-05, + "loss": 0.3262, + "step": 8064 + }, + { + "epoch": 0.49, + "learning_rate": 1.0898842944500533e-05, + "loss": 0.3365, + "step": 8065 + }, + { + "epoch": 0.49, + "learning_rate": 1.0896894200251414e-05, + "loss": 0.339, + "step": 8066 + }, + { + "epoch": 0.49, + "learning_rate": 1.0894945421664983e-05, + "loss": 0.3239, + "step": 8067 + }, + { + "epoch": 0.49, + "learning_rate": 1.0892996608815855e-05, + "loss": 0.3235, + "step": 8068 + }, + { + "epoch": 0.49, + "learning_rate": 1.0891047761778637e-05, + "loss": 0.2899, + "step": 8069 + }, + { + "epoch": 0.49, + "learning_rate": 1.088909888062794e-05, + "loss": 0.3027, + "step": 8070 + }, + { + "epoch": 0.49, + "learning_rate": 1.0887149965438377e-05, + "loss": 0.3217, + "step": 8071 + }, + { + "epoch": 0.49, + "learning_rate": 1.088520101628456e-05, + "loss": 0.2983, + "step": 8072 + }, + { + "epoch": 0.49, + "learning_rate": 1.0883252033241105e-05, + "loss": 0.2846, + "step": 8073 + }, + { + "epoch": 0.49, + "learning_rate": 1.0881303016382629e-05, + "loss": 0.3309, + "step": 8074 + }, + { + "epoch": 0.49, + "learning_rate": 1.0879353965783752e-05, + "loss": 0.3052, + "step": 8075 + }, + { + "epoch": 0.49, + "learning_rate": 1.0877404881519087e-05, + "loss": 0.2969, + "step": 8076 + }, + { + "epoch": 0.49, + "learning_rate": 1.0875455763663255e-05, + "loss": 0.2977, + "step": 8077 + }, + { + "epoch": 0.49, + "learning_rate": 1.0873506612290881e-05, + "loss": 0.3122, + "step": 8078 + }, + { + "epoch": 0.49, + "learning_rate": 1.0871557427476585e-05, + "loss": 0.2868, + "step": 8079 + }, + { + "epoch": 0.49, + "learning_rate": 1.0869608209294989e-05, + "loss": 0.298, + "step": 8080 + }, + { + "epoch": 0.49, + "learning_rate": 1.0867658957820724e-05, + "loss": 0.3061, + "step": 8081 + }, + { + "epoch": 0.49, + "learning_rate": 1.0865709673128411e-05, + "loss": 0.304, + "step": 8082 + }, + { + "epoch": 0.49, + "learning_rate": 1.086376035529268e-05, + "loss": 0.3209, + "step": 8083 + }, + { + "epoch": 0.49, + "learning_rate": 1.0861811004388163e-05, + "loss": 0.3323, + "step": 8084 + }, + { + "epoch": 0.49, + "learning_rate": 1.0859861620489482e-05, + "loss": 0.3454, + "step": 8085 + }, + { + "epoch": 0.49, + "learning_rate": 1.0857912203671279e-05, + "loss": 0.3164, + "step": 8086 + }, + { + "epoch": 0.49, + "learning_rate": 1.085596275400818e-05, + "loss": 0.3259, + "step": 8087 + }, + { + "epoch": 0.49, + "learning_rate": 1.085401327157482e-05, + "loss": 0.3253, + "step": 8088 + }, + { + "epoch": 0.49, + "learning_rate": 1.0852063756445834e-05, + "loss": 0.3013, + "step": 8089 + }, + { + "epoch": 0.49, + "learning_rate": 1.0850114208695858e-05, + "loss": 0.3103, + "step": 8090 + }, + { + "epoch": 0.49, + "learning_rate": 1.0848164628399536e-05, + "loss": 0.3225, + "step": 8091 + }, + { + "epoch": 0.49, + "learning_rate": 1.08462150156315e-05, + "loss": 0.322, + "step": 8092 + }, + { + "epoch": 0.49, + "learning_rate": 1.0844265370466393e-05, + "loss": 0.3131, + "step": 8093 + }, + { + "epoch": 0.49, + "learning_rate": 1.0842315692978858e-05, + "loss": 0.3384, + "step": 8094 + }, + { + "epoch": 0.49, + "learning_rate": 1.0840365983243533e-05, + "loss": 0.3184, + "step": 8095 + }, + { + "epoch": 0.49, + "learning_rate": 1.0838416241335065e-05, + "loss": 0.3052, + "step": 8096 + }, + { + "epoch": 0.49, + "learning_rate": 1.08364664673281e-05, + "loss": 0.2872, + "step": 8097 + }, + { + "epoch": 0.49, + "learning_rate": 1.0834516661297284e-05, + "loss": 0.3267, + "step": 8098 + }, + { + "epoch": 0.49, + "learning_rate": 1.0832566823317264e-05, + "loss": 0.3253, + "step": 8099 + }, + { + "epoch": 0.49, + "learning_rate": 1.083061695346269e-05, + "loss": 0.3421, + "step": 8100 + }, + { + "epoch": 0.49, + "learning_rate": 1.0828667051808211e-05, + "loss": 0.3078, + "step": 8101 + }, + { + "epoch": 0.49, + "learning_rate": 1.0826717118428476e-05, + "loss": 0.3052, + "step": 8102 + }, + { + "epoch": 0.49, + "learning_rate": 1.0824767153398144e-05, + "loss": 0.3348, + "step": 8103 + }, + { + "epoch": 0.49, + "learning_rate": 1.0822817156791866e-05, + "loss": 0.3167, + "step": 8104 + }, + { + "epoch": 0.49, + "learning_rate": 1.0820867128684291e-05, + "loss": 0.3044, + "step": 8105 + }, + { + "epoch": 0.49, + "learning_rate": 1.0818917069150085e-05, + "loss": 0.2896, + "step": 8106 + }, + { + "epoch": 0.49, + "learning_rate": 1.0816966978263901e-05, + "loss": 0.3331, + "step": 8107 + }, + { + "epoch": 0.49, + "learning_rate": 1.0815016856100395e-05, + "loss": 0.3178, + "step": 8108 + }, + { + "epoch": 0.49, + "learning_rate": 1.0813066702734234e-05, + "loss": 0.3246, + "step": 8109 + }, + { + "epoch": 0.49, + "learning_rate": 1.0811116518240067e-05, + "loss": 0.2882, + "step": 8110 + }, + { + "epoch": 0.49, + "learning_rate": 1.0809166302692569e-05, + "loss": 0.3415, + "step": 8111 + }, + { + "epoch": 0.49, + "learning_rate": 1.0807216056166395e-05, + "loss": 0.3237, + "step": 8112 + }, + { + "epoch": 0.49, + "learning_rate": 1.0805265778736214e-05, + "loss": 0.3369, + "step": 8113 + }, + { + "epoch": 0.49, + "learning_rate": 1.080331547047669e-05, + "loss": 0.3239, + "step": 8114 + }, + { + "epoch": 0.49, + "learning_rate": 1.0801365131462487e-05, + "loss": 0.3167, + "step": 8115 + }, + { + "epoch": 0.49, + "learning_rate": 1.0799414761768278e-05, + "loss": 0.332, + "step": 8116 + }, + { + "epoch": 0.49, + "learning_rate": 1.079746436146873e-05, + "loss": 0.3276, + "step": 8117 + }, + { + "epoch": 0.49, + "learning_rate": 1.0795513930638515e-05, + "loss": 0.3058, + "step": 8118 + }, + { + "epoch": 0.49, + "learning_rate": 1.0793563469352305e-05, + "loss": 0.3309, + "step": 8119 + }, + { + "epoch": 0.49, + "learning_rate": 1.0791612977684767e-05, + "loss": 0.3343, + "step": 8120 + }, + { + "epoch": 0.49, + "learning_rate": 1.0789662455710582e-05, + "loss": 0.3267, + "step": 8121 + }, + { + "epoch": 0.49, + "learning_rate": 1.0787711903504424e-05, + "loss": 0.3128, + "step": 8122 + }, + { + "epoch": 0.49, + "learning_rate": 1.0785761321140962e-05, + "loss": 0.2931, + "step": 8123 + }, + { + "epoch": 0.49, + "learning_rate": 1.0783810708694883e-05, + "loss": 0.3438, + "step": 8124 + }, + { + "epoch": 0.49, + "learning_rate": 1.0781860066240861e-05, + "loss": 0.2997, + "step": 8125 + }, + { + "epoch": 0.49, + "learning_rate": 1.0779909393853578e-05, + "loss": 0.3128, + "step": 8126 + }, + { + "epoch": 0.49, + "learning_rate": 1.077795869160771e-05, + "loss": 0.3329, + "step": 8127 + }, + { + "epoch": 0.49, + "learning_rate": 1.0776007959577946e-05, + "loss": 0.2706, + "step": 8128 + }, + { + "epoch": 0.49, + "learning_rate": 1.0774057197838963e-05, + "loss": 0.3218, + "step": 8129 + }, + { + "epoch": 0.49, + "learning_rate": 1.0772106406465446e-05, + "loss": 0.3513, + "step": 8130 + }, + { + "epoch": 0.49, + "learning_rate": 1.0770155585532085e-05, + "loss": 0.3273, + "step": 8131 + }, + { + "epoch": 0.49, + "learning_rate": 1.0768204735113561e-05, + "loss": 0.3179, + "step": 8132 + }, + { + "epoch": 0.49, + "learning_rate": 1.0766253855284565e-05, + "loss": 0.3234, + "step": 8133 + }, + { + "epoch": 0.49, + "learning_rate": 1.0764302946119786e-05, + "loss": 0.3171, + "step": 8134 + }, + { + "epoch": 0.49, + "learning_rate": 1.0762352007693913e-05, + "loss": 0.3198, + "step": 8135 + }, + { + "epoch": 0.49, + "learning_rate": 1.0760401040081639e-05, + "loss": 0.2905, + "step": 8136 + }, + { + "epoch": 0.49, + "learning_rate": 1.0758450043357648e-05, + "loss": 0.2924, + "step": 8137 + }, + { + "epoch": 0.49, + "learning_rate": 1.0756499017596644e-05, + "loss": 0.308, + "step": 8138 + }, + { + "epoch": 0.49, + "learning_rate": 1.075454796287332e-05, + "loss": 0.3125, + "step": 8139 + }, + { + "epoch": 0.49, + "learning_rate": 1.0752596879262363e-05, + "loss": 0.3027, + "step": 8140 + }, + { + "epoch": 0.49, + "learning_rate": 1.0750645766838476e-05, + "loss": 0.3005, + "step": 8141 + }, + { + "epoch": 0.49, + "learning_rate": 1.074869462567636e-05, + "loss": 0.3401, + "step": 8142 + }, + { + "epoch": 0.49, + "learning_rate": 1.0746743455850705e-05, + "loss": 0.3105, + "step": 8143 + }, + { + "epoch": 0.49, + "learning_rate": 1.0744792257436219e-05, + "loss": 0.3351, + "step": 8144 + }, + { + "epoch": 0.49, + "learning_rate": 1.0742841030507596e-05, + "loss": 0.3136, + "step": 8145 + }, + { + "epoch": 0.49, + "learning_rate": 1.0740889775139544e-05, + "loss": 0.3078, + "step": 8146 + }, + { + "epoch": 0.49, + "learning_rate": 1.0738938491406761e-05, + "loss": 0.3343, + "step": 8147 + }, + { + "epoch": 0.49, + "learning_rate": 1.0736987179383957e-05, + "loss": 0.2836, + "step": 8148 + }, + { + "epoch": 0.49, + "learning_rate": 1.0735035839145836e-05, + "loss": 0.3287, + "step": 8149 + }, + { + "epoch": 0.49, + "learning_rate": 1.0733084470767099e-05, + "loss": 0.3318, + "step": 8150 + }, + { + "epoch": 0.49, + "learning_rate": 1.0731133074322457e-05, + "loss": 0.322, + "step": 8151 + }, + { + "epoch": 0.49, + "learning_rate": 1.072918164988662e-05, + "loss": 0.3206, + "step": 8152 + }, + { + "epoch": 0.49, + "learning_rate": 1.0727230197534299e-05, + "loss": 0.2916, + "step": 8153 + }, + { + "epoch": 0.49, + "learning_rate": 1.0725278717340203e-05, + "loss": 0.2972, + "step": 8154 + }, + { + "epoch": 0.49, + "learning_rate": 1.0723327209379038e-05, + "loss": 0.3002, + "step": 8155 + }, + { + "epoch": 0.49, + "learning_rate": 1.0721375673725524e-05, + "loss": 0.2974, + "step": 8156 + }, + { + "epoch": 0.49, + "learning_rate": 1.0719424110454372e-05, + "loss": 0.3192, + "step": 8157 + }, + { + "epoch": 0.49, + "learning_rate": 1.0717472519640302e-05, + "loss": 0.3351, + "step": 8158 + }, + { + "epoch": 0.49, + "learning_rate": 1.0715520901358026e-05, + "loss": 0.3103, + "step": 8159 + }, + { + "epoch": 0.49, + "learning_rate": 1.0713569255682259e-05, + "loss": 0.3002, + "step": 8160 + }, + { + "epoch": 0.49, + "learning_rate": 1.0711617582687721e-05, + "loss": 0.3287, + "step": 8161 + }, + { + "epoch": 0.49, + "learning_rate": 1.0709665882449129e-05, + "loss": 0.3079, + "step": 8162 + }, + { + "epoch": 0.49, + "learning_rate": 1.070771415504121e-05, + "loss": 0.3241, + "step": 8163 + }, + { + "epoch": 0.49, + "learning_rate": 1.070576240053868e-05, + "loss": 0.2754, + "step": 8164 + }, + { + "epoch": 0.49, + "learning_rate": 1.070381061901626e-05, + "loss": 0.2893, + "step": 8165 + }, + { + "epoch": 0.49, + "learning_rate": 1.0701858810548675e-05, + "loss": 0.3299, + "step": 8166 + }, + { + "epoch": 0.49, + "learning_rate": 1.069990697521065e-05, + "loss": 0.3045, + "step": 8167 + }, + { + "epoch": 0.49, + "learning_rate": 1.0697955113076911e-05, + "loss": 0.2991, + "step": 8168 + }, + { + "epoch": 0.49, + "learning_rate": 1.0696003224222186e-05, + "loss": 0.3131, + "step": 8169 + }, + { + "epoch": 0.49, + "learning_rate": 1.0694051308721196e-05, + "loss": 0.3078, + "step": 8170 + }, + { + "epoch": 0.49, + "learning_rate": 1.0692099366648677e-05, + "loss": 0.3117, + "step": 8171 + }, + { + "epoch": 0.49, + "learning_rate": 1.069014739807935e-05, + "loss": 0.3167, + "step": 8172 + }, + { + "epoch": 0.49, + "learning_rate": 1.0688195403087955e-05, + "loss": 0.3231, + "step": 8173 + }, + { + "epoch": 0.49, + "learning_rate": 1.068624338174922e-05, + "loss": 0.3011, + "step": 8174 + }, + { + "epoch": 0.49, + "learning_rate": 1.0684291334137873e-05, + "loss": 0.3111, + "step": 8175 + }, + { + "epoch": 0.49, + "learning_rate": 1.0682339260328654e-05, + "loss": 0.2847, + "step": 8176 + }, + { + "epoch": 0.49, + "learning_rate": 1.0680387160396293e-05, + "loss": 0.3242, + "step": 8177 + }, + { + "epoch": 0.49, + "learning_rate": 1.0678435034415528e-05, + "loss": 0.2818, + "step": 8178 + }, + { + "epoch": 0.49, + "learning_rate": 1.0676482882461094e-05, + "loss": 0.3211, + "step": 8179 + }, + { + "epoch": 0.49, + "learning_rate": 1.0674530704607726e-05, + "loss": 0.3287, + "step": 8180 + }, + { + "epoch": 0.49, + "learning_rate": 1.067257850093017e-05, + "loss": 0.3147, + "step": 8181 + }, + { + "epoch": 0.49, + "learning_rate": 1.0670626271503157e-05, + "loss": 0.3125, + "step": 8182 + }, + { + "epoch": 0.49, + "learning_rate": 1.0668674016401437e-05, + "loss": 0.327, + "step": 8183 + }, + { + "epoch": 0.49, + "learning_rate": 1.0666721735699743e-05, + "loss": 0.32, + "step": 8184 + }, + { + "epoch": 0.49, + "learning_rate": 1.066476942947282e-05, + "loss": 0.3239, + "step": 8185 + }, + { + "epoch": 0.49, + "learning_rate": 1.0662817097795413e-05, + "loss": 0.3161, + "step": 8186 + }, + { + "epoch": 0.49, + "learning_rate": 1.0660864740742262e-05, + "loss": 0.3186, + "step": 8187 + }, + { + "epoch": 0.49, + "learning_rate": 1.065891235838812e-05, + "loss": 0.3122, + "step": 8188 + }, + { + "epoch": 0.49, + "learning_rate": 1.0656959950807728e-05, + "loss": 0.3234, + "step": 8189 + }, + { + "epoch": 0.49, + "learning_rate": 1.0655007518075833e-05, + "loss": 0.3121, + "step": 8190 + }, + { + "epoch": 0.49, + "learning_rate": 1.0653055060267185e-05, + "loss": 0.3041, + "step": 8191 + }, + { + "epoch": 0.49, + "learning_rate": 1.0651102577456532e-05, + "loss": 0.3001, + "step": 8192 + }, + { + "epoch": 0.49, + "learning_rate": 1.0649150069718627e-05, + "loss": 0.2969, + "step": 8193 + }, + { + "epoch": 0.5, + "learning_rate": 1.064719753712822e-05, + "loss": 0.3147, + "step": 8194 + }, + { + "epoch": 0.5, + "learning_rate": 1.064524497976006e-05, + "loss": 0.3262, + "step": 8195 + }, + { + "epoch": 0.5, + "learning_rate": 1.0643292397688904e-05, + "loss": 0.3393, + "step": 8196 + }, + { + "epoch": 0.5, + "learning_rate": 1.06413397909895e-05, + "loss": 0.3055, + "step": 8197 + }, + { + "epoch": 0.5, + "learning_rate": 1.0639387159736612e-05, + "loss": 0.3237, + "step": 8198 + }, + { + "epoch": 0.5, + "learning_rate": 1.0637434504004989e-05, + "loss": 0.3158, + "step": 8199 + }, + { + "epoch": 0.5, + "learning_rate": 1.063548182386939e-05, + "loss": 0.3094, + "step": 8200 + }, + { + "epoch": 0.5, + "learning_rate": 1.0633529119404571e-05, + "loss": 0.3245, + "step": 8201 + }, + { + "epoch": 0.5, + "learning_rate": 1.0631576390685291e-05, + "loss": 0.3033, + "step": 8202 + }, + { + "epoch": 0.5, + "learning_rate": 1.0629623637786314e-05, + "loss": 0.3036, + "step": 8203 + }, + { + "epoch": 0.5, + "learning_rate": 1.0627670860782397e-05, + "loss": 0.3181, + "step": 8204 + }, + { + "epoch": 0.5, + "learning_rate": 1.0625718059748303e-05, + "loss": 0.3214, + "step": 8205 + }, + { + "epoch": 0.5, + "learning_rate": 1.0623765234758792e-05, + "loss": 0.3398, + "step": 8206 + }, + { + "epoch": 0.5, + "learning_rate": 1.0621812385888626e-05, + "loss": 0.3292, + "step": 8207 + }, + { + "epoch": 0.5, + "learning_rate": 1.0619859513212575e-05, + "loss": 0.3224, + "step": 8208 + }, + { + "epoch": 0.5, + "learning_rate": 1.0617906616805398e-05, + "loss": 0.3045, + "step": 8209 + }, + { + "epoch": 0.5, + "learning_rate": 1.0615953696741863e-05, + "loss": 0.3136, + "step": 8210 + }, + { + "epoch": 0.5, + "learning_rate": 1.0614000753096743e-05, + "loss": 0.3058, + "step": 8211 + }, + { + "epoch": 0.5, + "learning_rate": 1.0612047785944795e-05, + "loss": 0.3192, + "step": 8212 + }, + { + "epoch": 0.5, + "learning_rate": 1.0610094795360796e-05, + "loss": 0.3271, + "step": 8213 + }, + { + "epoch": 0.5, + "learning_rate": 1.0608141781419512e-05, + "loss": 0.3047, + "step": 8214 + }, + { + "epoch": 0.5, + "learning_rate": 1.0606188744195713e-05, + "loss": 0.334, + "step": 8215 + }, + { + "epoch": 0.5, + "learning_rate": 1.0604235683764172e-05, + "loss": 0.3083, + "step": 8216 + }, + { + "epoch": 0.5, + "learning_rate": 1.0602282600199663e-05, + "loss": 0.3055, + "step": 8217 + }, + { + "epoch": 0.5, + "learning_rate": 1.0600329493576956e-05, + "loss": 0.3265, + "step": 8218 + }, + { + "epoch": 0.5, + "learning_rate": 1.0598376363970826e-05, + "loss": 0.3256, + "step": 8219 + }, + { + "epoch": 0.5, + "learning_rate": 1.0596423211456048e-05, + "loss": 0.2966, + "step": 8220 + }, + { + "epoch": 0.5, + "learning_rate": 1.0594470036107398e-05, + "loss": 0.2828, + "step": 8221 + }, + { + "epoch": 0.5, + "learning_rate": 1.0592516837999654e-05, + "loss": 0.2979, + "step": 8222 + }, + { + "epoch": 0.5, + "learning_rate": 1.059056361720759e-05, + "loss": 0.3345, + "step": 8223 + }, + { + "epoch": 0.5, + "learning_rate": 1.058861037380599e-05, + "loss": 0.3323, + "step": 8224 + }, + { + "epoch": 0.5, + "learning_rate": 1.0586657107869626e-05, + "loss": 0.3009, + "step": 8225 + }, + { + "epoch": 0.5, + "learning_rate": 1.0584703819473286e-05, + "loss": 0.2776, + "step": 8226 + }, + { + "epoch": 0.5, + "learning_rate": 1.0582750508691743e-05, + "loss": 0.3147, + "step": 8227 + }, + { + "epoch": 0.5, + "learning_rate": 1.0580797175599788e-05, + "loss": 0.3552, + "step": 8228 + }, + { + "epoch": 0.5, + "learning_rate": 1.0578843820272195e-05, + "loss": 0.2934, + "step": 8229 + }, + { + "epoch": 0.5, + "learning_rate": 1.0576890442783756e-05, + "loss": 0.3161, + "step": 8230 + }, + { + "epoch": 0.5, + "learning_rate": 1.0574937043209248e-05, + "loss": 0.3172, + "step": 8231 + }, + { + "epoch": 0.5, + "learning_rate": 1.0572983621623455e-05, + "loss": 0.2974, + "step": 8232 + }, + { + "epoch": 0.5, + "learning_rate": 1.0571030178101174e-05, + "loss": 0.3142, + "step": 8233 + }, + { + "epoch": 0.5, + "learning_rate": 1.056907671271718e-05, + "loss": 0.2992, + "step": 8234 + }, + { + "epoch": 0.5, + "learning_rate": 1.056712322554627e-05, + "loss": 0.2972, + "step": 8235 + }, + { + "epoch": 0.5, + "learning_rate": 1.0565169716663227e-05, + "loss": 0.3002, + "step": 8236 + }, + { + "epoch": 0.5, + "learning_rate": 1.056321618614284e-05, + "loss": 0.3252, + "step": 8237 + }, + { + "epoch": 0.5, + "learning_rate": 1.0561262634059904e-05, + "loss": 0.3108, + "step": 8238 + }, + { + "epoch": 0.5, + "learning_rate": 1.0559309060489207e-05, + "loss": 0.3136, + "step": 8239 + }, + { + "epoch": 0.5, + "learning_rate": 1.0557355465505539e-05, + "loss": 0.31, + "step": 8240 + }, + { + "epoch": 0.5, + "learning_rate": 1.0555401849183698e-05, + "loss": 0.334, + "step": 8241 + }, + { + "epoch": 0.5, + "learning_rate": 1.0553448211598474e-05, + "loss": 0.3384, + "step": 8242 + }, + { + "epoch": 0.5, + "learning_rate": 1.0551494552824661e-05, + "loss": 0.3072, + "step": 8243 + }, + { + "epoch": 0.5, + "learning_rate": 1.0549540872937059e-05, + "loss": 0.3309, + "step": 8244 + }, + { + "epoch": 0.5, + "learning_rate": 1.0547587172010458e-05, + "loss": 0.3136, + "step": 8245 + }, + { + "epoch": 0.5, + "learning_rate": 1.0545633450119657e-05, + "loss": 0.3256, + "step": 8246 + }, + { + "epoch": 0.5, + "learning_rate": 1.0543679707339453e-05, + "loss": 0.3298, + "step": 8247 + }, + { + "epoch": 0.5, + "learning_rate": 1.0541725943744648e-05, + "loss": 0.2991, + "step": 8248 + }, + { + "epoch": 0.5, + "learning_rate": 1.0539772159410036e-05, + "loss": 0.3262, + "step": 8249 + }, + { + "epoch": 0.5, + "learning_rate": 1.0537818354410418e-05, + "loss": 0.3211, + "step": 8250 + }, + { + "epoch": 0.5, + "learning_rate": 1.0535864528820599e-05, + "loss": 0.3086, + "step": 8251 + }, + { + "epoch": 0.5, + "learning_rate": 1.0533910682715376e-05, + "loss": 0.284, + "step": 8252 + }, + { + "epoch": 0.5, + "learning_rate": 1.0531956816169555e-05, + "loss": 0.3276, + "step": 8253 + }, + { + "epoch": 0.5, + "learning_rate": 1.0530002929257937e-05, + "loss": 0.3209, + "step": 8254 + }, + { + "epoch": 0.5, + "learning_rate": 1.0528049022055324e-05, + "loss": 0.3418, + "step": 8255 + }, + { + "epoch": 0.5, + "learning_rate": 1.0526095094636527e-05, + "loss": 0.3124, + "step": 8256 + }, + { + "epoch": 0.5, + "learning_rate": 1.0524141147076341e-05, + "loss": 0.2972, + "step": 8257 + }, + { + "epoch": 0.5, + "learning_rate": 1.0522187179449583e-05, + "loss": 0.3287, + "step": 8258 + }, + { + "epoch": 0.5, + "learning_rate": 1.0520233191831058e-05, + "loss": 0.3078, + "step": 8259 + }, + { + "epoch": 0.5, + "learning_rate": 1.0518279184295568e-05, + "loss": 0.3184, + "step": 8260 + }, + { + "epoch": 0.5, + "learning_rate": 1.0516325156917925e-05, + "loss": 0.3235, + "step": 8261 + }, + { + "epoch": 0.5, + "learning_rate": 1.0514371109772944e-05, + "loss": 0.3217, + "step": 8262 + }, + { + "epoch": 0.5, + "learning_rate": 1.0512417042935425e-05, + "loss": 0.3041, + "step": 8263 + }, + { + "epoch": 0.5, + "learning_rate": 1.0510462956480182e-05, + "loss": 0.3132, + "step": 8264 + }, + { + "epoch": 0.5, + "learning_rate": 1.0508508850482031e-05, + "loss": 0.317, + "step": 8265 + }, + { + "epoch": 0.5, + "learning_rate": 1.0506554725015779e-05, + "loss": 0.3128, + "step": 8266 + }, + { + "epoch": 0.5, + "learning_rate": 1.0504600580156241e-05, + "loss": 0.3094, + "step": 8267 + }, + { + "epoch": 0.5, + "learning_rate": 1.0502646415978232e-05, + "loss": 0.3156, + "step": 8268 + }, + { + "epoch": 0.5, + "learning_rate": 1.0500692232556569e-05, + "loss": 0.3158, + "step": 8269 + }, + { + "epoch": 0.5, + "learning_rate": 1.0498738029966058e-05, + "loss": 0.3108, + "step": 8270 + }, + { + "epoch": 0.5, + "learning_rate": 1.0496783808281523e-05, + "loss": 0.3359, + "step": 8271 + }, + { + "epoch": 0.5, + "learning_rate": 1.049482956757778e-05, + "loss": 0.3401, + "step": 8272 + }, + { + "epoch": 0.5, + "learning_rate": 1.0492875307929643e-05, + "loss": 0.3209, + "step": 8273 + }, + { + "epoch": 0.5, + "learning_rate": 1.0490921029411936e-05, + "loss": 0.305, + "step": 8274 + }, + { + "epoch": 0.5, + "learning_rate": 1.0488966732099471e-05, + "loss": 0.2854, + "step": 8275 + }, + { + "epoch": 0.5, + "learning_rate": 1.0487012416067072e-05, + "loss": 0.304, + "step": 8276 + }, + { + "epoch": 0.5, + "learning_rate": 1.0485058081389559e-05, + "loss": 0.3142, + "step": 8277 + }, + { + "epoch": 0.5, + "learning_rate": 1.0483103728141752e-05, + "loss": 0.3304, + "step": 8278 + }, + { + "epoch": 0.5, + "learning_rate": 1.0481149356398475e-05, + "loss": 0.3209, + "step": 8279 + }, + { + "epoch": 0.5, + "learning_rate": 1.0479194966234547e-05, + "loss": 0.3036, + "step": 8280 + }, + { + "epoch": 0.5, + "learning_rate": 1.0477240557724793e-05, + "loss": 0.3281, + "step": 8281 + }, + { + "epoch": 0.5, + "learning_rate": 1.0475286130944036e-05, + "loss": 0.3231, + "step": 8282 + }, + { + "epoch": 0.5, + "learning_rate": 1.0473331685967103e-05, + "loss": 0.2977, + "step": 8283 + }, + { + "epoch": 0.5, + "learning_rate": 1.0471377222868817e-05, + "loss": 0.3245, + "step": 8284 + }, + { + "epoch": 0.5, + "learning_rate": 1.0469422741724003e-05, + "loss": 0.2972, + "step": 8285 + }, + { + "epoch": 0.5, + "learning_rate": 1.0467468242607492e-05, + "loss": 0.3147, + "step": 8286 + }, + { + "epoch": 0.5, + "learning_rate": 1.0465513725594105e-05, + "loss": 0.3119, + "step": 8287 + }, + { + "epoch": 0.5, + "learning_rate": 1.046355919075868e-05, + "loss": 0.3373, + "step": 8288 + }, + { + "epoch": 0.5, + "learning_rate": 1.0461604638176035e-05, + "loss": 0.3239, + "step": 8289 + }, + { + "epoch": 0.5, + "learning_rate": 1.0459650067921003e-05, + "loss": 0.3092, + "step": 8290 + }, + { + "epoch": 0.5, + "learning_rate": 1.045769548006842e-05, + "loss": 0.3136, + "step": 8291 + }, + { + "epoch": 0.5, + "learning_rate": 1.0455740874693106e-05, + "loss": 0.2953, + "step": 8292 + }, + { + "epoch": 0.5, + "learning_rate": 1.0453786251869902e-05, + "loss": 0.3223, + "step": 8293 + }, + { + "epoch": 0.5, + "learning_rate": 1.0451831611673637e-05, + "loss": 0.334, + "step": 8294 + }, + { + "epoch": 0.5, + "learning_rate": 1.0449876954179141e-05, + "loss": 0.3248, + "step": 8295 + }, + { + "epoch": 0.5, + "learning_rate": 1.0447922279461254e-05, + "loss": 0.3143, + "step": 8296 + }, + { + "epoch": 0.5, + "learning_rate": 1.04459675875948e-05, + "loss": 0.291, + "step": 8297 + }, + { + "epoch": 0.5, + "learning_rate": 1.0444012878654623e-05, + "loss": 0.3248, + "step": 8298 + }, + { + "epoch": 0.5, + "learning_rate": 1.044205815271555e-05, + "loss": 0.3161, + "step": 8299 + }, + { + "epoch": 0.5, + "learning_rate": 1.044010340985243e-05, + "loss": 0.3189, + "step": 8300 + }, + { + "epoch": 0.5, + "learning_rate": 1.0438148650140085e-05, + "loss": 0.3223, + "step": 8301 + }, + { + "epoch": 0.5, + "learning_rate": 1.0436193873653362e-05, + "loss": 0.3119, + "step": 8302 + }, + { + "epoch": 0.5, + "learning_rate": 1.0434239080467094e-05, + "loss": 0.2934, + "step": 8303 + }, + { + "epoch": 0.5, + "learning_rate": 1.0432284270656125e-05, + "loss": 0.3029, + "step": 8304 + }, + { + "epoch": 0.5, + "learning_rate": 1.0430329444295286e-05, + "loss": 0.2983, + "step": 8305 + }, + { + "epoch": 0.5, + "learning_rate": 1.0428374601459426e-05, + "loss": 0.3157, + "step": 8306 + }, + { + "epoch": 0.5, + "learning_rate": 1.042641974222338e-05, + "loss": 0.3278, + "step": 8307 + }, + { + "epoch": 0.5, + "learning_rate": 1.042446486666199e-05, + "loss": 0.3284, + "step": 8308 + }, + { + "epoch": 0.5, + "learning_rate": 1.0422509974850099e-05, + "loss": 0.2946, + "step": 8309 + }, + { + "epoch": 0.5, + "learning_rate": 1.0420555066862548e-05, + "loss": 0.2999, + "step": 8310 + }, + { + "epoch": 0.5, + "learning_rate": 1.041860014277418e-05, + "loss": 0.3195, + "step": 8311 + }, + { + "epoch": 0.5, + "learning_rate": 1.0416645202659838e-05, + "loss": 0.3206, + "step": 8312 + }, + { + "epoch": 0.5, + "learning_rate": 1.0414690246594374e-05, + "loss": 0.3078, + "step": 8313 + }, + { + "epoch": 0.5, + "learning_rate": 1.041273527465262e-05, + "loss": 0.3107, + "step": 8314 + }, + { + "epoch": 0.5, + "learning_rate": 1.0410780286909432e-05, + "loss": 0.3365, + "step": 8315 + }, + { + "epoch": 0.5, + "learning_rate": 1.0408825283439651e-05, + "loss": 0.3128, + "step": 8316 + }, + { + "epoch": 0.5, + "learning_rate": 1.0406870264318124e-05, + "loss": 0.3253, + "step": 8317 + }, + { + "epoch": 0.5, + "learning_rate": 1.0404915229619699e-05, + "loss": 0.3379, + "step": 8318 + }, + { + "epoch": 0.5, + "learning_rate": 1.0402960179419224e-05, + "loss": 0.2919, + "step": 8319 + }, + { + "epoch": 0.5, + "learning_rate": 1.0401005113791547e-05, + "loss": 0.317, + "step": 8320 + }, + { + "epoch": 0.5, + "learning_rate": 1.039905003281152e-05, + "loss": 0.3145, + "step": 8321 + }, + { + "epoch": 0.5, + "learning_rate": 1.0397094936553987e-05, + "loss": 0.3122, + "step": 8322 + }, + { + "epoch": 0.5, + "learning_rate": 1.0395139825093801e-05, + "loss": 0.3147, + "step": 8323 + }, + { + "epoch": 0.5, + "learning_rate": 1.0393184698505817e-05, + "loss": 0.3089, + "step": 8324 + }, + { + "epoch": 0.5, + "learning_rate": 1.0391229556864876e-05, + "loss": 0.3331, + "step": 8325 + }, + { + "epoch": 0.5, + "learning_rate": 1.038927440024584e-05, + "loss": 0.3175, + "step": 8326 + }, + { + "epoch": 0.5, + "learning_rate": 1.0387319228723558e-05, + "loss": 0.3119, + "step": 8327 + }, + { + "epoch": 0.5, + "learning_rate": 1.0385364042372885e-05, + "loss": 0.3234, + "step": 8328 + }, + { + "epoch": 0.5, + "learning_rate": 1.0383408841268671e-05, + "loss": 0.3037, + "step": 8329 + }, + { + "epoch": 0.5, + "learning_rate": 1.0381453625485773e-05, + "loss": 0.3058, + "step": 8330 + }, + { + "epoch": 0.5, + "learning_rate": 1.0379498395099046e-05, + "loss": 0.3118, + "step": 8331 + }, + { + "epoch": 0.5, + "learning_rate": 1.037754315018334e-05, + "loss": 0.3217, + "step": 8332 + }, + { + "epoch": 0.5, + "learning_rate": 1.037558789081352e-05, + "loss": 0.3213, + "step": 8333 + }, + { + "epoch": 0.5, + "learning_rate": 1.0373632617064433e-05, + "loss": 0.3062, + "step": 8334 + }, + { + "epoch": 0.5, + "learning_rate": 1.0371677329010946e-05, + "loss": 0.3061, + "step": 8335 + }, + { + "epoch": 0.5, + "learning_rate": 1.0369722026727906e-05, + "loss": 0.3223, + "step": 8336 + }, + { + "epoch": 0.5, + "learning_rate": 1.0367766710290179e-05, + "loss": 0.3265, + "step": 8337 + }, + { + "epoch": 0.5, + "learning_rate": 1.0365811379772621e-05, + "loss": 0.3152, + "step": 8338 + }, + { + "epoch": 0.5, + "learning_rate": 1.0363856035250093e-05, + "loss": 0.2769, + "step": 8339 + }, + { + "epoch": 0.5, + "learning_rate": 1.036190067679745e-05, + "loss": 0.329, + "step": 8340 + }, + { + "epoch": 0.5, + "learning_rate": 1.0359945304489556e-05, + "loss": 0.3082, + "step": 8341 + }, + { + "epoch": 0.5, + "learning_rate": 1.035798991840127e-05, + "loss": 0.2872, + "step": 8342 + }, + { + "epoch": 0.5, + "learning_rate": 1.0356034518607457e-05, + "loss": 0.285, + "step": 8343 + }, + { + "epoch": 0.5, + "learning_rate": 1.0354079105182975e-05, + "loss": 0.3351, + "step": 8344 + }, + { + "epoch": 0.5, + "learning_rate": 1.0352123678202686e-05, + "loss": 0.2807, + "step": 8345 + }, + { + "epoch": 0.5, + "learning_rate": 1.0350168237741457e-05, + "loss": 0.3122, + "step": 8346 + }, + { + "epoch": 0.5, + "learning_rate": 1.034821278387415e-05, + "loss": 0.3015, + "step": 8347 + }, + { + "epoch": 0.5, + "learning_rate": 1.0346257316675627e-05, + "loss": 0.3089, + "step": 8348 + }, + { + "epoch": 0.5, + "learning_rate": 1.0344301836220751e-05, + "loss": 0.3142, + "step": 8349 + }, + { + "epoch": 0.5, + "learning_rate": 1.034234634258439e-05, + "loss": 0.3186, + "step": 8350 + }, + { + "epoch": 0.5, + "learning_rate": 1.0340390835841412e-05, + "loss": 0.3435, + "step": 8351 + }, + { + "epoch": 0.5, + "learning_rate": 1.0338435316066675e-05, + "loss": 0.3195, + "step": 8352 + }, + { + "epoch": 0.5, + "learning_rate": 1.0336479783335054e-05, + "loss": 0.3062, + "step": 8353 + }, + { + "epoch": 0.5, + "learning_rate": 1.0334524237721409e-05, + "loss": 0.2914, + "step": 8354 + }, + { + "epoch": 0.5, + "learning_rate": 1.0332568679300612e-05, + "loss": 0.3265, + "step": 8355 + }, + { + "epoch": 0.5, + "learning_rate": 1.033061310814753e-05, + "loss": 0.3125, + "step": 8356 + }, + { + "epoch": 0.5, + "learning_rate": 1.0328657524337029e-05, + "loss": 0.2972, + "step": 8357 + }, + { + "epoch": 0.5, + "learning_rate": 1.032670192794398e-05, + "loss": 0.351, + "step": 8358 + }, + { + "epoch": 0.5, + "learning_rate": 1.0324746319043256e-05, + "loss": 0.3287, + "step": 8359 + }, + { + "epoch": 0.51, + "learning_rate": 1.0322790697709717e-05, + "loss": 0.2945, + "step": 8360 + }, + { + "epoch": 0.51, + "learning_rate": 1.0320835064018244e-05, + "loss": 0.341, + "step": 8361 + }, + { + "epoch": 0.51, + "learning_rate": 1.0318879418043699e-05, + "loss": 0.2838, + "step": 8362 + }, + { + "epoch": 0.51, + "learning_rate": 1.0316923759860962e-05, + "loss": 0.2958, + "step": 8363 + }, + { + "epoch": 0.51, + "learning_rate": 1.0314968089544899e-05, + "loss": 0.3223, + "step": 8364 + }, + { + "epoch": 0.51, + "learning_rate": 1.0313012407170382e-05, + "loss": 0.3019, + "step": 8365 + }, + { + "epoch": 0.51, + "learning_rate": 1.0311056712812286e-05, + "loss": 0.3037, + "step": 8366 + }, + { + "epoch": 0.51, + "learning_rate": 1.0309101006545481e-05, + "loss": 0.3387, + "step": 8367 + }, + { + "epoch": 0.51, + "learning_rate": 1.0307145288444847e-05, + "loss": 0.2871, + "step": 8368 + }, + { + "epoch": 0.51, + "learning_rate": 1.030518955858525e-05, + "loss": 0.3131, + "step": 8369 + }, + { + "epoch": 0.51, + "learning_rate": 1.0303233817041569e-05, + "loss": 0.3011, + "step": 8370 + }, + { + "epoch": 0.51, + "learning_rate": 1.030127806388868e-05, + "loss": 0.2946, + "step": 8371 + }, + { + "epoch": 0.51, + "learning_rate": 1.0299322299201453e-05, + "loss": 0.3253, + "step": 8372 + }, + { + "epoch": 0.51, + "learning_rate": 1.029736652305477e-05, + "loss": 0.3228, + "step": 8373 + }, + { + "epoch": 0.51, + "learning_rate": 1.0295410735523506e-05, + "loss": 0.2902, + "step": 8374 + }, + { + "epoch": 0.51, + "learning_rate": 1.0293454936682535e-05, + "loss": 0.3094, + "step": 8375 + }, + { + "epoch": 0.51, + "learning_rate": 1.0291499126606737e-05, + "loss": 0.2783, + "step": 8376 + }, + { + "epoch": 0.51, + "learning_rate": 1.0289543305370983e-05, + "loss": 0.3262, + "step": 8377 + }, + { + "epoch": 0.51, + "learning_rate": 1.0287587473050163e-05, + "loss": 0.3018, + "step": 8378 + }, + { + "epoch": 0.51, + "learning_rate": 1.0285631629719143e-05, + "loss": 0.3549, + "step": 8379 + }, + { + "epoch": 0.51, + "learning_rate": 1.0283675775452811e-05, + "loss": 0.3496, + "step": 8380 + }, + { + "epoch": 0.51, + "learning_rate": 1.0281719910326043e-05, + "loss": 0.3172, + "step": 8381 + }, + { + "epoch": 0.51, + "learning_rate": 1.0279764034413715e-05, + "loss": 0.2895, + "step": 8382 + }, + { + "epoch": 0.51, + "learning_rate": 1.0277808147790714e-05, + "loss": 0.3284, + "step": 8383 + }, + { + "epoch": 0.51, + "learning_rate": 1.0275852250531912e-05, + "loss": 0.3019, + "step": 8384 + }, + { + "epoch": 0.51, + "learning_rate": 1.02738963427122e-05, + "loss": 0.3211, + "step": 8385 + }, + { + "epoch": 0.51, + "learning_rate": 1.0271940424406453e-05, + "loss": 0.3167, + "step": 8386 + }, + { + "epoch": 0.51, + "learning_rate": 1.0269984495689552e-05, + "loss": 0.3304, + "step": 8387 + }, + { + "epoch": 0.51, + "learning_rate": 1.026802855663638e-05, + "loss": 0.2811, + "step": 8388 + }, + { + "epoch": 0.51, + "learning_rate": 1.026607260732182e-05, + "loss": 0.2874, + "step": 8389 + }, + { + "epoch": 0.51, + "learning_rate": 1.0264116647820758e-05, + "loss": 0.3192, + "step": 8390 + }, + { + "epoch": 0.51, + "learning_rate": 1.0262160678208075e-05, + "loss": 0.3262, + "step": 8391 + }, + { + "epoch": 0.51, + "learning_rate": 1.026020469855865e-05, + "loss": 0.3108, + "step": 8392 + }, + { + "epoch": 0.51, + "learning_rate": 1.0258248708947375e-05, + "loss": 0.3167, + "step": 8393 + }, + { + "epoch": 0.51, + "learning_rate": 1.0256292709449132e-05, + "loss": 0.285, + "step": 8394 + }, + { + "epoch": 0.51, + "learning_rate": 1.02543367001388e-05, + "loss": 0.3186, + "step": 8395 + }, + { + "epoch": 0.51, + "learning_rate": 1.0252380681091274e-05, + "loss": 0.3142, + "step": 8396 + }, + { + "epoch": 0.51, + "learning_rate": 1.025042465238143e-05, + "loss": 0.3055, + "step": 8397 + }, + { + "epoch": 0.51, + "learning_rate": 1.0248468614084161e-05, + "loss": 0.3008, + "step": 8398 + }, + { + "epoch": 0.51, + "learning_rate": 1.024651256627435e-05, + "loss": 0.3158, + "step": 8399 + }, + { + "epoch": 0.51, + "learning_rate": 1.0244556509026884e-05, + "loss": 0.2687, + "step": 8400 + }, + { + "epoch": 0.51, + "learning_rate": 1.0242600442416651e-05, + "loss": 0.3142, + "step": 8401 + }, + { + "epoch": 0.51, + "learning_rate": 1.0240644366518536e-05, + "loss": 0.3203, + "step": 8402 + }, + { + "epoch": 0.51, + "learning_rate": 1.0238688281407428e-05, + "loss": 0.3228, + "step": 8403 + }, + { + "epoch": 0.51, + "learning_rate": 1.0236732187158214e-05, + "loss": 0.2833, + "step": 8404 + }, + { + "epoch": 0.51, + "learning_rate": 1.0234776083845787e-05, + "loss": 0.3418, + "step": 8405 + }, + { + "epoch": 0.51, + "learning_rate": 1.0232819971545034e-05, + "loss": 0.291, + "step": 8406 + }, + { + "epoch": 0.51, + "learning_rate": 1.0230863850330839e-05, + "loss": 0.3313, + "step": 8407 + }, + { + "epoch": 0.51, + "learning_rate": 1.0228907720278096e-05, + "loss": 0.2669, + "step": 8408 + }, + { + "epoch": 0.51, + "learning_rate": 1.0226951581461697e-05, + "loss": 0.3256, + "step": 8409 + }, + { + "epoch": 0.51, + "learning_rate": 1.0224995433956527e-05, + "loss": 0.3489, + "step": 8410 + }, + { + "epoch": 0.51, + "learning_rate": 1.022303927783748e-05, + "loss": 0.2764, + "step": 8411 + }, + { + "epoch": 0.51, + "learning_rate": 1.0221083113179443e-05, + "loss": 0.3502, + "step": 8412 + }, + { + "epoch": 0.51, + "learning_rate": 1.0219126940057312e-05, + "loss": 0.3343, + "step": 8413 + }, + { + "epoch": 0.51, + "learning_rate": 1.0217170758545976e-05, + "loss": 0.3239, + "step": 8414 + }, + { + "epoch": 0.51, + "learning_rate": 1.0215214568720327e-05, + "loss": 0.298, + "step": 8415 + }, + { + "epoch": 0.51, + "learning_rate": 1.0213258370655258e-05, + "loss": 0.3157, + "step": 8416 + }, + { + "epoch": 0.51, + "learning_rate": 1.0211302164425657e-05, + "loss": 0.2995, + "step": 8417 + }, + { + "epoch": 0.51, + "learning_rate": 1.0209345950106422e-05, + "loss": 0.2715, + "step": 8418 + }, + { + "epoch": 0.51, + "learning_rate": 1.0207389727772441e-05, + "loss": 0.2864, + "step": 8419 + }, + { + "epoch": 0.51, + "learning_rate": 1.0205433497498616e-05, + "loss": 0.3211, + "step": 8420 + }, + { + "epoch": 0.51, + "learning_rate": 1.0203477259359834e-05, + "loss": 0.3231, + "step": 8421 + }, + { + "epoch": 0.51, + "learning_rate": 1.0201521013430986e-05, + "loss": 0.2879, + "step": 8422 + }, + { + "epoch": 0.51, + "learning_rate": 1.0199564759786974e-05, + "loss": 0.2793, + "step": 8423 + }, + { + "epoch": 0.51, + "learning_rate": 1.0197608498502688e-05, + "loss": 0.3136, + "step": 8424 + }, + { + "epoch": 0.51, + "learning_rate": 1.0195652229653026e-05, + "loss": 0.3117, + "step": 8425 + }, + { + "epoch": 0.51, + "learning_rate": 1.0193695953312878e-05, + "loss": 0.2888, + "step": 8426 + }, + { + "epoch": 0.51, + "learning_rate": 1.0191739669557144e-05, + "loss": 0.3075, + "step": 8427 + }, + { + "epoch": 0.51, + "learning_rate": 1.0189783378460718e-05, + "loss": 0.3114, + "step": 8428 + }, + { + "epoch": 0.51, + "learning_rate": 1.0187827080098498e-05, + "loss": 0.3154, + "step": 8429 + }, + { + "epoch": 0.51, + "learning_rate": 1.0185870774545375e-05, + "loss": 0.31, + "step": 8430 + }, + { + "epoch": 0.51, + "learning_rate": 1.018391446187625e-05, + "loss": 0.2952, + "step": 8431 + }, + { + "epoch": 0.51, + "learning_rate": 1.0181958142166017e-05, + "loss": 0.2963, + "step": 8432 + }, + { + "epoch": 0.51, + "learning_rate": 1.0180001815489578e-05, + "loss": 0.3062, + "step": 8433 + }, + { + "epoch": 0.51, + "learning_rate": 1.0178045481921822e-05, + "loss": 0.296, + "step": 8434 + }, + { + "epoch": 0.51, + "learning_rate": 1.0176089141537658e-05, + "loss": 0.3158, + "step": 8435 + }, + { + "epoch": 0.51, + "learning_rate": 1.0174132794411975e-05, + "loss": 0.2959, + "step": 8436 + }, + { + "epoch": 0.51, + "learning_rate": 1.0172176440619671e-05, + "loss": 0.2815, + "step": 8437 + }, + { + "epoch": 0.51, + "learning_rate": 1.0170220080235651e-05, + "loss": 0.3178, + "step": 8438 + }, + { + "epoch": 0.51, + "learning_rate": 1.0168263713334805e-05, + "loss": 0.3228, + "step": 8439 + }, + { + "epoch": 0.51, + "learning_rate": 1.016630733999204e-05, + "loss": 0.346, + "step": 8440 + }, + { + "epoch": 0.51, + "learning_rate": 1.0164350960282252e-05, + "loss": 0.308, + "step": 8441 + }, + { + "epoch": 0.51, + "learning_rate": 1.0162394574280337e-05, + "loss": 0.3203, + "step": 8442 + }, + { + "epoch": 0.51, + "learning_rate": 1.0160438182061201e-05, + "loss": 0.3104, + "step": 8443 + }, + { + "epoch": 0.51, + "learning_rate": 1.0158481783699736e-05, + "loss": 0.327, + "step": 8444 + }, + { + "epoch": 0.51, + "learning_rate": 1.0156525379270852e-05, + "loss": 0.3278, + "step": 8445 + }, + { + "epoch": 0.51, + "learning_rate": 1.0154568968849441e-05, + "loss": 0.305, + "step": 8446 + }, + { + "epoch": 0.51, + "learning_rate": 1.0152612552510406e-05, + "loss": 0.3156, + "step": 8447 + }, + { + "epoch": 0.51, + "learning_rate": 1.015065613032865e-05, + "loss": 0.2997, + "step": 8448 + }, + { + "epoch": 0.51, + "learning_rate": 1.0148699702379072e-05, + "loss": 0.291, + "step": 8449 + }, + { + "epoch": 0.51, + "learning_rate": 1.0146743268736574e-05, + "loss": 0.2836, + "step": 8450 + }, + { + "epoch": 0.51, + "learning_rate": 1.0144786829476056e-05, + "loss": 0.2875, + "step": 8451 + }, + { + "epoch": 0.51, + "learning_rate": 1.014283038467242e-05, + "loss": 0.3189, + "step": 8452 + }, + { + "epoch": 0.51, + "learning_rate": 1.0140873934400568e-05, + "loss": 0.2969, + "step": 8453 + }, + { + "epoch": 0.51, + "learning_rate": 1.0138917478735402e-05, + "loss": 0.3032, + "step": 8454 + }, + { + "epoch": 0.51, + "learning_rate": 1.0136961017751825e-05, + "loss": 0.3164, + "step": 8455 + }, + { + "epoch": 0.51, + "learning_rate": 1.013500455152474e-05, + "loss": 0.3008, + "step": 8456 + }, + { + "epoch": 0.51, + "learning_rate": 1.0133048080129048e-05, + "loss": 0.3117, + "step": 8457 + }, + { + "epoch": 0.51, + "learning_rate": 1.0131091603639655e-05, + "loss": 0.2828, + "step": 8458 + }, + { + "epoch": 0.51, + "learning_rate": 1.012913512213146e-05, + "loss": 0.3318, + "step": 8459 + }, + { + "epoch": 0.51, + "learning_rate": 1.0127178635679369e-05, + "loss": 0.3128, + "step": 8460 + }, + { + "epoch": 0.51, + "learning_rate": 1.0125222144358285e-05, + "loss": 0.3273, + "step": 8461 + }, + { + "epoch": 0.51, + "learning_rate": 1.0123265648243108e-05, + "loss": 0.3373, + "step": 8462 + }, + { + "epoch": 0.51, + "learning_rate": 1.012130914740875e-05, + "loss": 0.3284, + "step": 8463 + }, + { + "epoch": 0.51, + "learning_rate": 1.0119352641930106e-05, + "loss": 0.298, + "step": 8464 + }, + { + "epoch": 0.51, + "learning_rate": 1.0117396131882086e-05, + "loss": 0.2951, + "step": 8465 + }, + { + "epoch": 0.51, + "learning_rate": 1.0115439617339596e-05, + "loss": 0.3128, + "step": 8466 + }, + { + "epoch": 0.51, + "learning_rate": 1.0113483098377533e-05, + "loss": 0.3016, + "step": 8467 + }, + { + "epoch": 0.51, + "learning_rate": 1.0111526575070804e-05, + "loss": 0.3147, + "step": 8468 + }, + { + "epoch": 0.51, + "learning_rate": 1.010957004749432e-05, + "loss": 0.2881, + "step": 8469 + }, + { + "epoch": 0.51, + "learning_rate": 1.0107613515722978e-05, + "loss": 0.2747, + "step": 8470 + }, + { + "epoch": 0.51, + "learning_rate": 1.0105656979831691e-05, + "loss": 0.275, + "step": 8471 + }, + { + "epoch": 0.51, + "learning_rate": 1.0103700439895356e-05, + "loss": 0.3117, + "step": 8472 + }, + { + "epoch": 0.51, + "learning_rate": 1.0101743895988886e-05, + "loss": 0.293, + "step": 8473 + }, + { + "epoch": 0.51, + "learning_rate": 1.0099787348187183e-05, + "loss": 0.2944, + "step": 8474 + }, + { + "epoch": 0.51, + "learning_rate": 1.0097830796565151e-05, + "loss": 0.3195, + "step": 8475 + }, + { + "epoch": 0.51, + "learning_rate": 1.00958742411977e-05, + "loss": 0.2875, + "step": 8476 + }, + { + "epoch": 0.51, + "learning_rate": 1.0093917682159736e-05, + "loss": 0.315, + "step": 8477 + }, + { + "epoch": 0.51, + "learning_rate": 1.0091961119526162e-05, + "loss": 0.2793, + "step": 8478 + }, + { + "epoch": 0.51, + "learning_rate": 1.0090004553371884e-05, + "loss": 0.3184, + "step": 8479 + }, + { + "epoch": 0.51, + "learning_rate": 1.0088047983771811e-05, + "loss": 0.2829, + "step": 8480 + }, + { + "epoch": 0.51, + "learning_rate": 1.0086091410800854e-05, + "loss": 0.3078, + "step": 8481 + }, + { + "epoch": 0.51, + "learning_rate": 1.0084134834533907e-05, + "loss": 0.298, + "step": 8482 + }, + { + "epoch": 0.51, + "learning_rate": 1.0082178255045894e-05, + "loss": 0.3097, + "step": 8483 + }, + { + "epoch": 0.51, + "learning_rate": 1.0080221672411706e-05, + "loss": 0.3004, + "step": 8484 + }, + { + "epoch": 0.51, + "learning_rate": 1.0078265086706259e-05, + "loss": 0.3027, + "step": 8485 + }, + { + "epoch": 0.51, + "learning_rate": 1.0076308498004458e-05, + "loss": 0.3061, + "step": 8486 + }, + { + "epoch": 0.51, + "learning_rate": 1.0074351906381211e-05, + "loss": 0.3047, + "step": 8487 + }, + { + "epoch": 0.51, + "learning_rate": 1.0072395311911423e-05, + "loss": 0.3033, + "step": 8488 + }, + { + "epoch": 0.51, + "learning_rate": 1.0070438714670004e-05, + "loss": 0.3178, + "step": 8489 + }, + { + "epoch": 0.51, + "learning_rate": 1.0068482114731864e-05, + "loss": 0.284, + "step": 8490 + }, + { + "epoch": 0.51, + "learning_rate": 1.0066525512171908e-05, + "loss": 0.3415, + "step": 8491 + }, + { + "epoch": 0.51, + "learning_rate": 1.006456890706504e-05, + "loss": 0.3189, + "step": 8492 + }, + { + "epoch": 0.51, + "learning_rate": 1.0062612299486175e-05, + "loss": 0.339, + "step": 8493 + }, + { + "epoch": 0.51, + "learning_rate": 1.0060655689510217e-05, + "loss": 0.297, + "step": 8494 + }, + { + "epoch": 0.51, + "learning_rate": 1.0058699077212076e-05, + "loss": 0.3065, + "step": 8495 + }, + { + "epoch": 0.51, + "learning_rate": 1.005674246266666e-05, + "loss": 0.3477, + "step": 8496 + }, + { + "epoch": 0.51, + "learning_rate": 1.0054785845948875e-05, + "loss": 0.3083, + "step": 8497 + }, + { + "epoch": 0.51, + "learning_rate": 1.0052829227133634e-05, + "loss": 0.3125, + "step": 8498 + }, + { + "epoch": 0.51, + "learning_rate": 1.0050872606295842e-05, + "loss": 0.303, + "step": 8499 + }, + { + "epoch": 0.51, + "learning_rate": 1.004891598351041e-05, + "loss": 0.3329, + "step": 8500 + }, + { + "epoch": 0.51, + "learning_rate": 1.0046959358852244e-05, + "loss": 0.3072, + "step": 8501 + }, + { + "epoch": 0.51, + "learning_rate": 1.0045002732396252e-05, + "loss": 0.3156, + "step": 8502 + }, + { + "epoch": 0.51, + "learning_rate": 1.0043046104217346e-05, + "loss": 0.3263, + "step": 8503 + }, + { + "epoch": 0.51, + "learning_rate": 1.0041089474390432e-05, + "loss": 0.2829, + "step": 8504 + }, + { + "epoch": 0.51, + "learning_rate": 1.0039132842990424e-05, + "loss": 0.3348, + "step": 8505 + }, + { + "epoch": 0.51, + "learning_rate": 1.0037176210092228e-05, + "loss": 0.3001, + "step": 8506 + }, + { + "epoch": 0.51, + "learning_rate": 1.003521957577075e-05, + "loss": 0.323, + "step": 8507 + }, + { + "epoch": 0.51, + "learning_rate": 1.0033262940100903e-05, + "loss": 0.3055, + "step": 8508 + }, + { + "epoch": 0.51, + "learning_rate": 1.0031306303157593e-05, + "loss": 0.2962, + "step": 8509 + }, + { + "epoch": 0.51, + "learning_rate": 1.0029349665015734e-05, + "loss": 0.315, + "step": 8510 + }, + { + "epoch": 0.51, + "learning_rate": 1.0027393025750231e-05, + "loss": 0.3326, + "step": 8511 + }, + { + "epoch": 0.51, + "learning_rate": 1.0025436385435996e-05, + "loss": 0.3089, + "step": 8512 + }, + { + "epoch": 0.51, + "learning_rate": 1.0023479744147937e-05, + "loss": 0.3329, + "step": 8513 + }, + { + "epoch": 0.51, + "learning_rate": 1.002152310196096e-05, + "loss": 0.267, + "step": 8514 + }, + { + "epoch": 0.51, + "learning_rate": 1.0019566458949981e-05, + "loss": 0.2979, + "step": 8515 + }, + { + "epoch": 0.51, + "learning_rate": 1.0017609815189907e-05, + "loss": 0.3001, + "step": 8516 + }, + { + "epoch": 0.51, + "learning_rate": 1.0015653170755648e-05, + "loss": 0.3499, + "step": 8517 + }, + { + "epoch": 0.51, + "learning_rate": 1.001369652572211e-05, + "loss": 0.3209, + "step": 8518 + }, + { + "epoch": 0.51, + "learning_rate": 1.0011739880164203e-05, + "loss": 0.3179, + "step": 8519 + }, + { + "epoch": 0.51, + "learning_rate": 1.0009783234156843e-05, + "loss": 0.2985, + "step": 8520 + }, + { + "epoch": 0.51, + "learning_rate": 1.0007826587774934e-05, + "loss": 0.3087, + "step": 8521 + }, + { + "epoch": 0.51, + "learning_rate": 1.0005869941093383e-05, + "loss": 0.2804, + "step": 8522 + }, + { + "epoch": 0.51, + "learning_rate": 1.0003913294187107e-05, + "loss": 0.3161, + "step": 8523 + }, + { + "epoch": 0.51, + "learning_rate": 1.0001956647131008e-05, + "loss": 0.3203, + "step": 8524 + }, + { + "epoch": 0.52, + "learning_rate": 1e-05, + "loss": 0.3206, + "step": 8525 + }, + { + "epoch": 0.52, + "learning_rate": 9.998043352868995e-06, + "loss": 0.2871, + "step": 8526 + }, + { + "epoch": 0.52, + "learning_rate": 9.996086705812898e-06, + "loss": 0.3108, + "step": 8527 + }, + { + "epoch": 0.52, + "learning_rate": 9.99413005890662e-06, + "loss": 0.3079, + "step": 8528 + }, + { + "epoch": 0.52, + "learning_rate": 9.99217341222507e-06, + "loss": 0.2914, + "step": 8529 + }, + { + "epoch": 0.52, + "learning_rate": 9.99021676584316e-06, + "loss": 0.3278, + "step": 8530 + }, + { + "epoch": 0.52, + "learning_rate": 9.988260119835798e-06, + "loss": 0.3616, + "step": 8531 + }, + { + "epoch": 0.52, + "learning_rate": 9.986303474277894e-06, + "loss": 0.3005, + "step": 8532 + }, + { + "epoch": 0.52, + "learning_rate": 9.984346829244355e-06, + "loss": 0.3352, + "step": 8533 + }, + { + "epoch": 0.52, + "learning_rate": 9.982390184810096e-06, + "loss": 0.3062, + "step": 8534 + }, + { + "epoch": 0.52, + "learning_rate": 9.980433541050019e-06, + "loss": 0.3175, + "step": 8535 + }, + { + "epoch": 0.52, + "learning_rate": 9.97847689803904e-06, + "loss": 0.2991, + "step": 8536 + }, + { + "epoch": 0.52, + "learning_rate": 9.976520255852065e-06, + "loss": 0.3076, + "step": 8537 + }, + { + "epoch": 0.52, + "learning_rate": 9.974563614564006e-06, + "loss": 0.3524, + "step": 8538 + }, + { + "epoch": 0.52, + "learning_rate": 9.97260697424977e-06, + "loss": 0.3382, + "step": 8539 + }, + { + "epoch": 0.52, + "learning_rate": 9.970650334984267e-06, + "loss": 0.2938, + "step": 8540 + }, + { + "epoch": 0.52, + "learning_rate": 9.968693696842407e-06, + "loss": 0.2836, + "step": 8541 + }, + { + "epoch": 0.52, + "learning_rate": 9.966737059899097e-06, + "loss": 0.327, + "step": 8542 + }, + { + "epoch": 0.52, + "learning_rate": 9.964780424229255e-06, + "loss": 0.2914, + "step": 8543 + }, + { + "epoch": 0.52, + "learning_rate": 9.962823789907777e-06, + "loss": 0.3245, + "step": 8544 + }, + { + "epoch": 0.52, + "learning_rate": 9.960867157009579e-06, + "loss": 0.3117, + "step": 8545 + }, + { + "epoch": 0.52, + "learning_rate": 9.95891052560957e-06, + "loss": 0.3052, + "step": 8546 + }, + { + "epoch": 0.52, + "learning_rate": 9.956953895782657e-06, + "loss": 0.2775, + "step": 8547 + }, + { + "epoch": 0.52, + "learning_rate": 9.954997267603753e-06, + "loss": 0.3251, + "step": 8548 + }, + { + "epoch": 0.52, + "learning_rate": 9.953040641147761e-06, + "loss": 0.2958, + "step": 8549 + }, + { + "epoch": 0.52, + "learning_rate": 9.951084016489595e-06, + "loss": 0.3039, + "step": 8550 + }, + { + "epoch": 0.52, + "learning_rate": 9.94912739370416e-06, + "loss": 0.3069, + "step": 8551 + }, + { + "epoch": 0.52, + "learning_rate": 9.94717077286637e-06, + "loss": 0.3103, + "step": 8552 + }, + { + "epoch": 0.52, + "learning_rate": 9.945214154051126e-06, + "loss": 0.3594, + "step": 8553 + }, + { + "epoch": 0.52, + "learning_rate": 9.943257537333344e-06, + "loss": 0.3129, + "step": 8554 + }, + { + "epoch": 0.52, + "learning_rate": 9.941300922787928e-06, + "loss": 0.327, + "step": 8555 + }, + { + "epoch": 0.52, + "learning_rate": 9.939344310489787e-06, + "loss": 0.3136, + "step": 8556 + }, + { + "epoch": 0.52, + "learning_rate": 9.937387700513827e-06, + "loss": 0.2867, + "step": 8557 + }, + { + "epoch": 0.52, + "learning_rate": 9.935431092934962e-06, + "loss": 0.3065, + "step": 8558 + }, + { + "epoch": 0.52, + "learning_rate": 9.933474487828097e-06, + "loss": 0.3135, + "step": 8559 + }, + { + "epoch": 0.52, + "learning_rate": 9.93151788526814e-06, + "loss": 0.3112, + "step": 8560 + }, + { + "epoch": 0.52, + "learning_rate": 9.929561285329998e-06, + "loss": 0.3072, + "step": 8561 + }, + { + "epoch": 0.52, + "learning_rate": 9.927604688088579e-06, + "loss": 0.317, + "step": 8562 + }, + { + "epoch": 0.52, + "learning_rate": 9.925648093618794e-06, + "loss": 0.327, + "step": 8563 + }, + { + "epoch": 0.52, + "learning_rate": 9.923691501995545e-06, + "loss": 0.3164, + "step": 8564 + }, + { + "epoch": 0.52, + "learning_rate": 9.921734913293744e-06, + "loss": 0.3103, + "step": 8565 + }, + { + "epoch": 0.52, + "learning_rate": 9.919778327588298e-06, + "loss": 0.3152, + "step": 8566 + }, + { + "epoch": 0.52, + "learning_rate": 9.917821744954111e-06, + "loss": 0.3037, + "step": 8567 + }, + { + "epoch": 0.52, + "learning_rate": 9.915865165466093e-06, + "loss": 0.2949, + "step": 8568 + }, + { + "epoch": 0.52, + "learning_rate": 9.913908589199147e-06, + "loss": 0.2977, + "step": 8569 + }, + { + "epoch": 0.52, + "learning_rate": 9.911952016228189e-06, + "loss": 0.2797, + "step": 8570 + }, + { + "epoch": 0.52, + "learning_rate": 9.909995446628118e-06, + "loss": 0.3276, + "step": 8571 + }, + { + "epoch": 0.52, + "learning_rate": 9.90803888047384e-06, + "loss": 0.317, + "step": 8572 + }, + { + "epoch": 0.52, + "learning_rate": 9.906082317840266e-06, + "loss": 0.2838, + "step": 8573 + }, + { + "epoch": 0.52, + "learning_rate": 9.904125758802301e-06, + "loss": 0.3171, + "step": 8574 + }, + { + "epoch": 0.52, + "learning_rate": 9.902169203434849e-06, + "loss": 0.2832, + "step": 8575 + }, + { + "epoch": 0.52, + "learning_rate": 9.900212651812819e-06, + "loss": 0.3329, + "step": 8576 + }, + { + "epoch": 0.52, + "learning_rate": 9.898256104011119e-06, + "loss": 0.3128, + "step": 8577 + }, + { + "epoch": 0.52, + "learning_rate": 9.896299560104649e-06, + "loss": 0.3041, + "step": 8578 + }, + { + "epoch": 0.52, + "learning_rate": 9.894343020168314e-06, + "loss": 0.3131, + "step": 8579 + }, + { + "epoch": 0.52, + "learning_rate": 9.892386484277026e-06, + "loss": 0.2966, + "step": 8580 + }, + { + "epoch": 0.52, + "learning_rate": 9.890429952505685e-06, + "loss": 0.2935, + "step": 8581 + }, + { + "epoch": 0.52, + "learning_rate": 9.888473424929199e-06, + "loss": 0.3228, + "step": 8582 + }, + { + "epoch": 0.52, + "learning_rate": 9.886516901622474e-06, + "loss": 0.3128, + "step": 8583 + }, + { + "epoch": 0.52, + "learning_rate": 9.88456038266041e-06, + "loss": 0.3214, + "step": 8584 + }, + { + "epoch": 0.52, + "learning_rate": 9.882603868117917e-06, + "loss": 0.2914, + "step": 8585 + }, + { + "epoch": 0.52, + "learning_rate": 9.880647358069895e-06, + "loss": 0.3097, + "step": 8586 + }, + { + "epoch": 0.52, + "learning_rate": 9.878690852591254e-06, + "loss": 0.3161, + "step": 8587 + }, + { + "epoch": 0.52, + "learning_rate": 9.876734351756894e-06, + "loss": 0.31, + "step": 8588 + }, + { + "epoch": 0.52, + "learning_rate": 9.874777855641719e-06, + "loss": 0.3242, + "step": 8589 + }, + { + "epoch": 0.52, + "learning_rate": 9.872821364320633e-06, + "loss": 0.32, + "step": 8590 + }, + { + "epoch": 0.52, + "learning_rate": 9.870864877868544e-06, + "loss": 0.3373, + "step": 8591 + }, + { + "epoch": 0.52, + "learning_rate": 9.868908396360346e-06, + "loss": 0.315, + "step": 8592 + }, + { + "epoch": 0.52, + "learning_rate": 9.866951919870953e-06, + "loss": 0.3061, + "step": 8593 + }, + { + "epoch": 0.52, + "learning_rate": 9.864995448475261e-06, + "loss": 0.3446, + "step": 8594 + }, + { + "epoch": 0.52, + "learning_rate": 9.863038982248177e-06, + "loss": 0.3061, + "step": 8595 + }, + { + "epoch": 0.52, + "learning_rate": 9.861082521264601e-06, + "loss": 0.3044, + "step": 8596 + }, + { + "epoch": 0.52, + "learning_rate": 9.859126065599435e-06, + "loss": 0.2991, + "step": 8597 + }, + { + "epoch": 0.52, + "learning_rate": 9.857169615327584e-06, + "loss": 0.3029, + "step": 8598 + }, + { + "epoch": 0.52, + "learning_rate": 9.855213170523947e-06, + "loss": 0.2994, + "step": 8599 + }, + { + "epoch": 0.52, + "learning_rate": 9.853256731263429e-06, + "loss": 0.2867, + "step": 8600 + }, + { + "epoch": 0.52, + "learning_rate": 9.85130029762093e-06, + "loss": 0.2846, + "step": 8601 + }, + { + "epoch": 0.52, + "learning_rate": 9.849343869671351e-06, + "loss": 0.2843, + "step": 8602 + }, + { + "epoch": 0.52, + "learning_rate": 9.847387447489594e-06, + "loss": 0.3225, + "step": 8603 + }, + { + "epoch": 0.52, + "learning_rate": 9.845431031150559e-06, + "loss": 0.3171, + "step": 8604 + }, + { + "epoch": 0.52, + "learning_rate": 9.84347462072915e-06, + "loss": 0.2991, + "step": 8605 + }, + { + "epoch": 0.52, + "learning_rate": 9.841518216300263e-06, + "loss": 0.2863, + "step": 8606 + }, + { + "epoch": 0.52, + "learning_rate": 9.8395618179388e-06, + "loss": 0.3192, + "step": 8607 + }, + { + "epoch": 0.52, + "learning_rate": 9.837605425719663e-06, + "loss": 0.3633, + "step": 8608 + }, + { + "epoch": 0.52, + "learning_rate": 9.835649039717751e-06, + "loss": 0.3054, + "step": 8609 + }, + { + "epoch": 0.52, + "learning_rate": 9.833692660007961e-06, + "loss": 0.31, + "step": 8610 + }, + { + "epoch": 0.52, + "learning_rate": 9.831736286665197e-06, + "loss": 0.2941, + "step": 8611 + }, + { + "epoch": 0.52, + "learning_rate": 9.829779919764356e-06, + "loss": 0.3013, + "step": 8612 + }, + { + "epoch": 0.52, + "learning_rate": 9.827823559380334e-06, + "loss": 0.3189, + "step": 8613 + }, + { + "epoch": 0.52, + "learning_rate": 9.82586720558803e-06, + "loss": 0.3217, + "step": 8614 + }, + { + "epoch": 0.52, + "learning_rate": 9.823910858462347e-06, + "loss": 0.2905, + "step": 8615 + }, + { + "epoch": 0.52, + "learning_rate": 9.821954518078179e-06, + "loss": 0.2758, + "step": 8616 + }, + { + "epoch": 0.52, + "learning_rate": 9.819998184510426e-06, + "loss": 0.2981, + "step": 8617 + }, + { + "epoch": 0.52, + "learning_rate": 9.818041857833986e-06, + "loss": 0.315, + "step": 8618 + }, + { + "epoch": 0.52, + "learning_rate": 9.816085538123755e-06, + "loss": 0.3145, + "step": 8619 + }, + { + "epoch": 0.52, + "learning_rate": 9.81412922545463e-06, + "loss": 0.2886, + "step": 8620 + }, + { + "epoch": 0.52, + "learning_rate": 9.812172919901505e-06, + "loss": 0.3334, + "step": 8621 + }, + { + "epoch": 0.52, + "learning_rate": 9.810216621539285e-06, + "loss": 0.2829, + "step": 8622 + }, + { + "epoch": 0.52, + "learning_rate": 9.80826033044286e-06, + "loss": 0.3362, + "step": 8623 + }, + { + "epoch": 0.52, + "learning_rate": 9.806304046687123e-06, + "loss": 0.2969, + "step": 8624 + }, + { + "epoch": 0.52, + "learning_rate": 9.804347770346977e-06, + "loss": 0.303, + "step": 8625 + }, + { + "epoch": 0.52, + "learning_rate": 9.802391501497315e-06, + "loss": 0.3055, + "step": 8626 + }, + { + "epoch": 0.52, + "learning_rate": 9.800435240213027e-06, + "loss": 0.3211, + "step": 8627 + }, + { + "epoch": 0.52, + "learning_rate": 9.798478986569016e-06, + "loss": 0.3034, + "step": 8628 + }, + { + "epoch": 0.52, + "learning_rate": 9.79652274064017e-06, + "loss": 0.3223, + "step": 8629 + }, + { + "epoch": 0.52, + "learning_rate": 9.794566502501388e-06, + "loss": 0.2807, + "step": 8630 + }, + { + "epoch": 0.52, + "learning_rate": 9.79261027222756e-06, + "loss": 0.2976, + "step": 8631 + }, + { + "epoch": 0.52, + "learning_rate": 9.790654049893581e-06, + "loss": 0.299, + "step": 8632 + }, + { + "epoch": 0.52, + "learning_rate": 9.788697835574348e-06, + "loss": 0.2847, + "step": 8633 + }, + { + "epoch": 0.52, + "learning_rate": 9.786741629344747e-06, + "loss": 0.3195, + "step": 8634 + }, + { + "epoch": 0.52, + "learning_rate": 9.784785431279677e-06, + "loss": 0.2838, + "step": 8635 + }, + { + "epoch": 0.52, + "learning_rate": 9.782829241454028e-06, + "loss": 0.304, + "step": 8636 + }, + { + "epoch": 0.52, + "learning_rate": 9.78087305994269e-06, + "loss": 0.2984, + "step": 8637 + }, + { + "epoch": 0.52, + "learning_rate": 9.778916886820559e-06, + "loss": 0.3131, + "step": 8638 + }, + { + "epoch": 0.52, + "learning_rate": 9.776960722162522e-06, + "loss": 0.3013, + "step": 8639 + }, + { + "epoch": 0.52, + "learning_rate": 9.775004566043475e-06, + "loss": 0.3057, + "step": 8640 + }, + { + "epoch": 0.52, + "learning_rate": 9.773048418538306e-06, + "loss": 0.2759, + "step": 8641 + }, + { + "epoch": 0.52, + "learning_rate": 9.771092279721902e-06, + "loss": 0.317, + "step": 8642 + }, + { + "epoch": 0.52, + "learning_rate": 9.769136149669163e-06, + "loss": 0.332, + "step": 8643 + }, + { + "epoch": 0.52, + "learning_rate": 9.76718002845497e-06, + "loss": 0.308, + "step": 8644 + }, + { + "epoch": 0.52, + "learning_rate": 9.765223916154218e-06, + "loss": 0.298, + "step": 8645 + }, + { + "epoch": 0.52, + "learning_rate": 9.763267812841788e-06, + "loss": 0.2913, + "step": 8646 + }, + { + "epoch": 0.52, + "learning_rate": 9.761311718592577e-06, + "loss": 0.3013, + "step": 8647 + }, + { + "epoch": 0.52, + "learning_rate": 9.759355633481471e-06, + "loss": 0.3324, + "step": 8648 + }, + { + "epoch": 0.52, + "learning_rate": 9.757399557583354e-06, + "loss": 0.2884, + "step": 8649 + }, + { + "epoch": 0.52, + "learning_rate": 9.755443490973121e-06, + "loss": 0.2935, + "step": 8650 + }, + { + "epoch": 0.52, + "learning_rate": 9.753487433725653e-06, + "loss": 0.3066, + "step": 8651 + }, + { + "epoch": 0.52, + "learning_rate": 9.751531385915842e-06, + "loss": 0.3066, + "step": 8652 + }, + { + "epoch": 0.52, + "learning_rate": 9.749575347618573e-06, + "loss": 0.2803, + "step": 8653 + }, + { + "epoch": 0.52, + "learning_rate": 9.74761931890873e-06, + "loss": 0.3329, + "step": 8654 + }, + { + "epoch": 0.52, + "learning_rate": 9.745663299861201e-06, + "loss": 0.3058, + "step": 8655 + }, + { + "epoch": 0.52, + "learning_rate": 9.74370729055087e-06, + "loss": 0.3012, + "step": 8656 + }, + { + "epoch": 0.52, + "learning_rate": 9.741751291052626e-06, + "loss": 0.2854, + "step": 8657 + }, + { + "epoch": 0.52, + "learning_rate": 9.739795301441351e-06, + "loss": 0.3318, + "step": 8658 + }, + { + "epoch": 0.52, + "learning_rate": 9.737839321791928e-06, + "loss": 0.2985, + "step": 8659 + }, + { + "epoch": 0.52, + "learning_rate": 9.735883352179243e-06, + "loss": 0.3267, + "step": 8660 + }, + { + "epoch": 0.52, + "learning_rate": 9.733927392678182e-06, + "loss": 0.3126, + "step": 8661 + }, + { + "epoch": 0.52, + "learning_rate": 9.731971443363623e-06, + "loss": 0.3304, + "step": 8662 + }, + { + "epoch": 0.52, + "learning_rate": 9.730015504310453e-06, + "loss": 0.2854, + "step": 8663 + }, + { + "epoch": 0.52, + "learning_rate": 9.728059575593552e-06, + "loss": 0.3097, + "step": 8664 + }, + { + "epoch": 0.52, + "learning_rate": 9.726103657287804e-06, + "loss": 0.3147, + "step": 8665 + }, + { + "epoch": 0.52, + "learning_rate": 9.72414774946809e-06, + "loss": 0.3463, + "step": 8666 + }, + { + "epoch": 0.52, + "learning_rate": 9.72219185220929e-06, + "loss": 0.2907, + "step": 8667 + }, + { + "epoch": 0.52, + "learning_rate": 9.720235965586287e-06, + "loss": 0.3167, + "step": 8668 + }, + { + "epoch": 0.52, + "learning_rate": 9.718280089673959e-06, + "loss": 0.3121, + "step": 8669 + }, + { + "epoch": 0.52, + "learning_rate": 9.71632422454719e-06, + "loss": 0.3047, + "step": 8670 + }, + { + "epoch": 0.52, + "learning_rate": 9.714368370280858e-06, + "loss": 0.2945, + "step": 8671 + }, + { + "epoch": 0.52, + "learning_rate": 9.712412526949838e-06, + "loss": 0.3128, + "step": 8672 + }, + { + "epoch": 0.52, + "learning_rate": 9.710456694629017e-06, + "loss": 0.3184, + "step": 8673 + }, + { + "epoch": 0.52, + "learning_rate": 9.708500873393265e-06, + "loss": 0.3131, + "step": 8674 + }, + { + "epoch": 0.52, + "learning_rate": 9.706545063317467e-06, + "loss": 0.3108, + "step": 8675 + }, + { + "epoch": 0.52, + "learning_rate": 9.704589264476495e-06, + "loss": 0.3287, + "step": 8676 + }, + { + "epoch": 0.52, + "learning_rate": 9.70263347694523e-06, + "loss": 0.3298, + "step": 8677 + }, + { + "epoch": 0.52, + "learning_rate": 9.700677700798547e-06, + "loss": 0.3136, + "step": 8678 + }, + { + "epoch": 0.52, + "learning_rate": 9.698721936111326e-06, + "loss": 0.3348, + "step": 8679 + }, + { + "epoch": 0.52, + "learning_rate": 9.696766182958436e-06, + "loss": 0.3376, + "step": 8680 + }, + { + "epoch": 0.52, + "learning_rate": 9.694810441414756e-06, + "loss": 0.2754, + "step": 8681 + }, + { + "epoch": 0.52, + "learning_rate": 9.69285471155516e-06, + "loss": 0.2733, + "step": 8682 + }, + { + "epoch": 0.52, + "learning_rate": 9.690898993454524e-06, + "loss": 0.3156, + "step": 8683 + }, + { + "epoch": 0.52, + "learning_rate": 9.688943287187719e-06, + "loss": 0.3281, + "step": 8684 + }, + { + "epoch": 0.52, + "learning_rate": 9.686987592829624e-06, + "loss": 0.3246, + "step": 8685 + }, + { + "epoch": 0.52, + "learning_rate": 9.685031910455104e-06, + "loss": 0.3245, + "step": 8686 + }, + { + "epoch": 0.52, + "learning_rate": 9.683076240139043e-06, + "loss": 0.2924, + "step": 8687 + }, + { + "epoch": 0.52, + "learning_rate": 9.681120581956303e-06, + "loss": 0.3005, + "step": 8688 + }, + { + "epoch": 0.52, + "learning_rate": 9.67916493598176e-06, + "loss": 0.3105, + "step": 8689 + }, + { + "epoch": 0.52, + "learning_rate": 9.677209302290286e-06, + "loss": 0.2912, + "step": 8690 + }, + { + "epoch": 0.53, + "learning_rate": 9.675253680956747e-06, + "loss": 0.3352, + "step": 8691 + }, + { + "epoch": 0.53, + "learning_rate": 9.673298072056023e-06, + "loss": 0.3147, + "step": 8692 + }, + { + "epoch": 0.53, + "learning_rate": 9.671342475662974e-06, + "loss": 0.3119, + "step": 8693 + }, + { + "epoch": 0.53, + "learning_rate": 9.669386891852472e-06, + "loss": 0.2681, + "step": 8694 + }, + { + "epoch": 0.53, + "learning_rate": 9.667431320699391e-06, + "loss": 0.2988, + "step": 8695 + }, + { + "epoch": 0.53, + "learning_rate": 9.665475762278594e-06, + "loss": 0.3004, + "step": 8696 + }, + { + "epoch": 0.53, + "learning_rate": 9.66352021666495e-06, + "loss": 0.3301, + "step": 8697 + }, + { + "epoch": 0.53, + "learning_rate": 9.661564683933329e-06, + "loss": 0.3315, + "step": 8698 + }, + { + "epoch": 0.53, + "learning_rate": 9.659609164158591e-06, + "loss": 0.3198, + "step": 8699 + }, + { + "epoch": 0.53, + "learning_rate": 9.657653657415612e-06, + "loss": 0.3326, + "step": 8700 + }, + { + "epoch": 0.53, + "learning_rate": 9.655698163779252e-06, + "loss": 0.3128, + "step": 8701 + }, + { + "epoch": 0.53, + "learning_rate": 9.653742683324376e-06, + "loss": 0.304, + "step": 8702 + }, + { + "epoch": 0.53, + "learning_rate": 9.651787216125853e-06, + "loss": 0.3128, + "step": 8703 + }, + { + "epoch": 0.53, + "learning_rate": 9.649831762258543e-06, + "loss": 0.305, + "step": 8704 + }, + { + "epoch": 0.53, + "learning_rate": 9.647876321797314e-06, + "loss": 0.3235, + "step": 8705 + }, + { + "epoch": 0.53, + "learning_rate": 9.645920894817029e-06, + "loss": 0.3078, + "step": 8706 + }, + { + "epoch": 0.53, + "learning_rate": 9.643965481392545e-06, + "loss": 0.3001, + "step": 8707 + }, + { + "epoch": 0.53, + "learning_rate": 9.64201008159873e-06, + "loss": 0.308, + "step": 8708 + }, + { + "epoch": 0.53, + "learning_rate": 9.640054695510444e-06, + "loss": 0.2967, + "step": 8709 + }, + { + "epoch": 0.53, + "learning_rate": 9.638099323202552e-06, + "loss": 0.3267, + "step": 8710 + }, + { + "epoch": 0.53, + "learning_rate": 9.636143964749912e-06, + "loss": 0.296, + "step": 8711 + }, + { + "epoch": 0.53, + "learning_rate": 9.634188620227379e-06, + "loss": 0.3583, + "step": 8712 + }, + { + "epoch": 0.53, + "learning_rate": 9.632233289709824e-06, + "loss": 0.35, + "step": 8713 + }, + { + "epoch": 0.53, + "learning_rate": 9.630277973272097e-06, + "loss": 0.327, + "step": 8714 + }, + { + "epoch": 0.53, + "learning_rate": 9.628322670989061e-06, + "loss": 0.2973, + "step": 8715 + }, + { + "epoch": 0.53, + "learning_rate": 9.62636738293557e-06, + "loss": 0.308, + "step": 8716 + }, + { + "epoch": 0.53, + "learning_rate": 9.624412109186486e-06, + "loss": 0.317, + "step": 8717 + }, + { + "epoch": 0.53, + "learning_rate": 9.622456849816665e-06, + "loss": 0.3129, + "step": 8718 + }, + { + "epoch": 0.53, + "learning_rate": 9.62050160490096e-06, + "loss": 0.2951, + "step": 8719 + }, + { + "epoch": 0.53, + "learning_rate": 9.618546374514232e-06, + "loss": 0.2966, + "step": 8720 + }, + { + "epoch": 0.53, + "learning_rate": 9.61659115873133e-06, + "loss": 0.3142, + "step": 8721 + }, + { + "epoch": 0.53, + "learning_rate": 9.614635957627119e-06, + "loss": 0.3203, + "step": 8722 + }, + { + "epoch": 0.53, + "learning_rate": 9.612680771276445e-06, + "loss": 0.3273, + "step": 8723 + }, + { + "epoch": 0.53, + "learning_rate": 9.610725599754161e-06, + "loss": 0.2818, + "step": 8724 + }, + { + "epoch": 0.53, + "learning_rate": 9.608770443135125e-06, + "loss": 0.3211, + "step": 8725 + }, + { + "epoch": 0.53, + "learning_rate": 9.606815301494188e-06, + "loss": 0.2906, + "step": 8726 + }, + { + "epoch": 0.53, + "learning_rate": 9.604860174906202e-06, + "loss": 0.3092, + "step": 8727 + }, + { + "epoch": 0.53, + "learning_rate": 9.602905063446018e-06, + "loss": 0.3117, + "step": 8728 + }, + { + "epoch": 0.53, + "learning_rate": 9.600949967188485e-06, + "loss": 0.2985, + "step": 8729 + }, + { + "epoch": 0.53, + "learning_rate": 9.598994886208456e-06, + "loss": 0.3122, + "step": 8730 + }, + { + "epoch": 0.53, + "learning_rate": 9.59703982058078e-06, + "loss": 0.2958, + "step": 8731 + }, + { + "epoch": 0.53, + "learning_rate": 9.595084770380303e-06, + "loss": 0.2997, + "step": 8732 + }, + { + "epoch": 0.53, + "learning_rate": 9.593129735681878e-06, + "loss": 0.3223, + "step": 8733 + }, + { + "epoch": 0.53, + "learning_rate": 9.59117471656035e-06, + "loss": 0.2959, + "step": 8734 + }, + { + "epoch": 0.53, + "learning_rate": 9.58921971309057e-06, + "loss": 0.2903, + "step": 8735 + }, + { + "epoch": 0.53, + "learning_rate": 9.587264725347381e-06, + "loss": 0.3064, + "step": 8736 + }, + { + "epoch": 0.53, + "learning_rate": 9.585309753405628e-06, + "loss": 0.2688, + "step": 8737 + }, + { + "epoch": 0.53, + "learning_rate": 9.583354797340161e-06, + "loss": 0.3181, + "step": 8738 + }, + { + "epoch": 0.53, + "learning_rate": 9.58139985722582e-06, + "loss": 0.2893, + "step": 8739 + }, + { + "epoch": 0.53, + "learning_rate": 9.579444933137454e-06, + "loss": 0.2782, + "step": 8740 + }, + { + "epoch": 0.53, + "learning_rate": 9.577490025149904e-06, + "loss": 0.32, + "step": 8741 + }, + { + "epoch": 0.53, + "learning_rate": 9.575535133338011e-06, + "loss": 0.2924, + "step": 8742 + }, + { + "epoch": 0.53, + "learning_rate": 9.573580257776622e-06, + "loss": 0.3002, + "step": 8743 + }, + { + "epoch": 0.53, + "learning_rate": 9.571625398540576e-06, + "loss": 0.3276, + "step": 8744 + }, + { + "epoch": 0.53, + "learning_rate": 9.569670555704714e-06, + "loss": 0.3089, + "step": 8745 + }, + { + "epoch": 0.53, + "learning_rate": 9.567715729343878e-06, + "loss": 0.2927, + "step": 8746 + }, + { + "epoch": 0.53, + "learning_rate": 9.565760919532911e-06, + "loss": 0.3061, + "step": 8747 + }, + { + "epoch": 0.53, + "learning_rate": 9.563806126346643e-06, + "loss": 0.3008, + "step": 8748 + }, + { + "epoch": 0.53, + "learning_rate": 9.561851349859918e-06, + "loss": 0.3181, + "step": 8749 + }, + { + "epoch": 0.53, + "learning_rate": 9.559896590147577e-06, + "loss": 0.3178, + "step": 8750 + }, + { + "epoch": 0.53, + "learning_rate": 9.557941847284451e-06, + "loss": 0.2997, + "step": 8751 + }, + { + "epoch": 0.53, + "learning_rate": 9.555987121345382e-06, + "loss": 0.3181, + "step": 8752 + }, + { + "epoch": 0.53, + "learning_rate": 9.554032412405205e-06, + "loss": 0.3359, + "step": 8753 + }, + { + "epoch": 0.53, + "learning_rate": 9.552077720538751e-06, + "loss": 0.278, + "step": 8754 + }, + { + "epoch": 0.53, + "learning_rate": 9.550123045820862e-06, + "loss": 0.2835, + "step": 8755 + }, + { + "epoch": 0.53, + "learning_rate": 9.548168388326365e-06, + "loss": 0.3432, + "step": 8756 + }, + { + "epoch": 0.53, + "learning_rate": 9.5462137481301e-06, + "loss": 0.3368, + "step": 8757 + }, + { + "epoch": 0.53, + "learning_rate": 9.544259125306895e-06, + "loss": 0.2867, + "step": 8758 + }, + { + "epoch": 0.53, + "learning_rate": 9.542304519931584e-06, + "loss": 0.2803, + "step": 8759 + }, + { + "epoch": 0.53, + "learning_rate": 9.540349932078999e-06, + "loss": 0.3139, + "step": 8760 + }, + { + "epoch": 0.53, + "learning_rate": 9.538395361823967e-06, + "loss": 0.3172, + "step": 8761 + }, + { + "epoch": 0.53, + "learning_rate": 9.536440809241324e-06, + "loss": 0.3206, + "step": 8762 + }, + { + "epoch": 0.53, + "learning_rate": 9.534486274405896e-06, + "loss": 0.3167, + "step": 8763 + }, + { + "epoch": 0.53, + "learning_rate": 9.532531757392511e-06, + "loss": 0.3147, + "step": 8764 + }, + { + "epoch": 0.53, + "learning_rate": 9.530577258275999e-06, + "loss": 0.2867, + "step": 8765 + }, + { + "epoch": 0.53, + "learning_rate": 9.528622777131186e-06, + "loss": 0.3131, + "step": 8766 + }, + { + "epoch": 0.53, + "learning_rate": 9.5266683140329e-06, + "loss": 0.3164, + "step": 8767 + }, + { + "epoch": 0.53, + "learning_rate": 9.524713869055968e-06, + "loss": 0.3108, + "step": 8768 + }, + { + "epoch": 0.53, + "learning_rate": 9.522759442275208e-06, + "loss": 0.3485, + "step": 8769 + }, + { + "epoch": 0.53, + "learning_rate": 9.520805033765456e-06, + "loss": 0.2766, + "step": 8770 + }, + { + "epoch": 0.53, + "learning_rate": 9.518850643601528e-06, + "loss": 0.3078, + "step": 8771 + }, + { + "epoch": 0.53, + "learning_rate": 9.516896271858248e-06, + "loss": 0.3248, + "step": 8772 + }, + { + "epoch": 0.53, + "learning_rate": 9.514941918610441e-06, + "loss": 0.3164, + "step": 8773 + }, + { + "epoch": 0.53, + "learning_rate": 9.512987583932928e-06, + "loss": 0.2783, + "step": 8774 + }, + { + "epoch": 0.53, + "learning_rate": 9.511033267900529e-06, + "loss": 0.305, + "step": 8775 + }, + { + "epoch": 0.53, + "learning_rate": 9.509078970588068e-06, + "loss": 0.2852, + "step": 8776 + }, + { + "epoch": 0.53, + "learning_rate": 9.507124692070356e-06, + "loss": 0.297, + "step": 8777 + }, + { + "epoch": 0.53, + "learning_rate": 9.505170432422222e-06, + "loss": 0.3147, + "step": 8778 + }, + { + "epoch": 0.53, + "learning_rate": 9.503216191718477e-06, + "loss": 0.317, + "step": 8779 + }, + { + "epoch": 0.53, + "learning_rate": 9.501261970033942e-06, + "loss": 0.3225, + "step": 8780 + }, + { + "epoch": 0.53, + "learning_rate": 9.499307767443438e-06, + "loss": 0.3318, + "step": 8781 + }, + { + "epoch": 0.53, + "learning_rate": 9.497353584021773e-06, + "loss": 0.3052, + "step": 8782 + }, + { + "epoch": 0.53, + "learning_rate": 9.495399419843762e-06, + "loss": 0.3117, + "step": 8783 + }, + { + "epoch": 0.53, + "learning_rate": 9.493445274984226e-06, + "loss": 0.2888, + "step": 8784 + }, + { + "epoch": 0.53, + "learning_rate": 9.491491149517976e-06, + "loss": 0.3071, + "step": 8785 + }, + { + "epoch": 0.53, + "learning_rate": 9.48953704351982e-06, + "loss": 0.2852, + "step": 8786 + }, + { + "epoch": 0.53, + "learning_rate": 9.48758295706458e-06, + "loss": 0.32, + "step": 8787 + }, + { + "epoch": 0.53, + "learning_rate": 9.485628890227063e-06, + "loss": 0.3083, + "step": 8788 + }, + { + "epoch": 0.53, + "learning_rate": 9.483674843082076e-06, + "loss": 0.293, + "step": 8789 + }, + { + "epoch": 0.53, + "learning_rate": 9.481720815704435e-06, + "loss": 0.3039, + "step": 8790 + }, + { + "epoch": 0.53, + "learning_rate": 9.479766808168945e-06, + "loss": 0.3075, + "step": 8791 + }, + { + "epoch": 0.53, + "learning_rate": 9.477812820550419e-06, + "loss": 0.2941, + "step": 8792 + }, + { + "epoch": 0.53, + "learning_rate": 9.47585885292366e-06, + "loss": 0.3143, + "step": 8793 + }, + { + "epoch": 0.53, + "learning_rate": 9.473904905363478e-06, + "loss": 0.29, + "step": 8794 + }, + { + "epoch": 0.53, + "learning_rate": 9.47195097794468e-06, + "loss": 0.3242, + "step": 8795 + }, + { + "epoch": 0.53, + "learning_rate": 9.469997070742067e-06, + "loss": 0.303, + "step": 8796 + }, + { + "epoch": 0.53, + "learning_rate": 9.468043183830448e-06, + "loss": 0.3055, + "step": 8797 + }, + { + "epoch": 0.53, + "learning_rate": 9.466089317284628e-06, + "loss": 0.3097, + "step": 8798 + }, + { + "epoch": 0.53, + "learning_rate": 9.464135471179404e-06, + "loss": 0.3232, + "step": 8799 + }, + { + "epoch": 0.53, + "learning_rate": 9.462181645589585e-06, + "loss": 0.2966, + "step": 8800 + }, + { + "epoch": 0.53, + "learning_rate": 9.460227840589968e-06, + "loss": 0.2994, + "step": 8801 + }, + { + "epoch": 0.53, + "learning_rate": 9.458274056255357e-06, + "loss": 0.3126, + "step": 8802 + }, + { + "epoch": 0.53, + "learning_rate": 9.45632029266055e-06, + "loss": 0.2921, + "step": 8803 + }, + { + "epoch": 0.53, + "learning_rate": 9.454366549880346e-06, + "loss": 0.3267, + "step": 8804 + }, + { + "epoch": 0.53, + "learning_rate": 9.452412827989545e-06, + "loss": 0.2806, + "step": 8805 + }, + { + "epoch": 0.53, + "learning_rate": 9.450459127062945e-06, + "loss": 0.3016, + "step": 8806 + }, + { + "epoch": 0.53, + "learning_rate": 9.448505447175339e-06, + "loss": 0.2771, + "step": 8807 + }, + { + "epoch": 0.53, + "learning_rate": 9.446551788401528e-06, + "loss": 0.3108, + "step": 8808 + }, + { + "epoch": 0.53, + "learning_rate": 9.444598150816302e-06, + "loss": 0.2969, + "step": 8809 + }, + { + "epoch": 0.53, + "learning_rate": 9.442644534494461e-06, + "loss": 0.2924, + "step": 8810 + }, + { + "epoch": 0.53, + "learning_rate": 9.440690939510797e-06, + "loss": 0.3273, + "step": 8811 + }, + { + "epoch": 0.53, + "learning_rate": 9.438737365940096e-06, + "loss": 0.3019, + "step": 8812 + }, + { + "epoch": 0.53, + "learning_rate": 9.436783813857161e-06, + "loss": 0.3253, + "step": 8813 + }, + { + "epoch": 0.53, + "learning_rate": 9.434830283336775e-06, + "loss": 0.2969, + "step": 8814 + }, + { + "epoch": 0.53, + "learning_rate": 9.432876774453737e-06, + "loss": 0.2905, + "step": 8815 + }, + { + "epoch": 0.53, + "learning_rate": 9.430923287282823e-06, + "loss": 0.3149, + "step": 8816 + }, + { + "epoch": 0.53, + "learning_rate": 9.428969821898833e-06, + "loss": 0.2919, + "step": 8817 + }, + { + "epoch": 0.53, + "learning_rate": 9.427016378376546e-06, + "loss": 0.3112, + "step": 8818 + }, + { + "epoch": 0.53, + "learning_rate": 9.425062956790758e-06, + "loss": 0.3387, + "step": 8819 + }, + { + "epoch": 0.53, + "learning_rate": 9.42310955721625e-06, + "loss": 0.327, + "step": 8820 + }, + { + "epoch": 0.53, + "learning_rate": 9.421156179727806e-06, + "loss": 0.3022, + "step": 8821 + }, + { + "epoch": 0.53, + "learning_rate": 9.419202824400215e-06, + "loss": 0.3203, + "step": 8822 + }, + { + "epoch": 0.53, + "learning_rate": 9.417249491308258e-06, + "loss": 0.2801, + "step": 8823 + }, + { + "epoch": 0.53, + "learning_rate": 9.415296180526718e-06, + "loss": 0.3135, + "step": 8824 + }, + { + "epoch": 0.53, + "learning_rate": 9.413342892130376e-06, + "loss": 0.2999, + "step": 8825 + }, + { + "epoch": 0.53, + "learning_rate": 9.411389626194013e-06, + "loss": 0.3331, + "step": 8826 + }, + { + "epoch": 0.53, + "learning_rate": 9.409436382792412e-06, + "loss": 0.3027, + "step": 8827 + }, + { + "epoch": 0.53, + "learning_rate": 9.40748316200035e-06, + "loss": 0.2967, + "step": 8828 + }, + { + "epoch": 0.53, + "learning_rate": 9.405529963892604e-06, + "loss": 0.2871, + "step": 8829 + }, + { + "epoch": 0.53, + "learning_rate": 9.403576788543955e-06, + "loss": 0.3041, + "step": 8830 + }, + { + "epoch": 0.53, + "learning_rate": 9.401623636029176e-06, + "loss": 0.281, + "step": 8831 + }, + { + "epoch": 0.53, + "learning_rate": 9.399670506423048e-06, + "loss": 0.2988, + "step": 8832 + }, + { + "epoch": 0.53, + "learning_rate": 9.397717399800342e-06, + "loss": 0.3265, + "step": 8833 + }, + { + "epoch": 0.53, + "learning_rate": 9.39576431623583e-06, + "loss": 0.3228, + "step": 8834 + }, + { + "epoch": 0.53, + "learning_rate": 9.39381125580429e-06, + "loss": 0.3189, + "step": 8835 + }, + { + "epoch": 0.53, + "learning_rate": 9.39185821858049e-06, + "loss": 0.3191, + "step": 8836 + }, + { + "epoch": 0.53, + "learning_rate": 9.389905204639206e-06, + "loss": 0.3323, + "step": 8837 + }, + { + "epoch": 0.53, + "learning_rate": 9.387952214055208e-06, + "loss": 0.3136, + "step": 8838 + }, + { + "epoch": 0.53, + "learning_rate": 9.38599924690326e-06, + "loss": 0.3493, + "step": 8839 + }, + { + "epoch": 0.53, + "learning_rate": 9.384046303258137e-06, + "loss": 0.3111, + "step": 8840 + }, + { + "epoch": 0.53, + "learning_rate": 9.382093383194606e-06, + "loss": 0.3048, + "step": 8841 + }, + { + "epoch": 0.53, + "learning_rate": 9.380140486787427e-06, + "loss": 0.292, + "step": 8842 + }, + { + "epoch": 0.53, + "learning_rate": 9.378187614111376e-06, + "loss": 0.3223, + "step": 8843 + }, + { + "epoch": 0.53, + "learning_rate": 9.37623476524121e-06, + "loss": 0.3083, + "step": 8844 + }, + { + "epoch": 0.53, + "learning_rate": 9.374281940251699e-06, + "loss": 0.2921, + "step": 8845 + }, + { + "epoch": 0.53, + "learning_rate": 9.372329139217604e-06, + "loss": 0.3175, + "step": 8846 + }, + { + "epoch": 0.53, + "learning_rate": 9.370376362213686e-06, + "loss": 0.3086, + "step": 8847 + }, + { + "epoch": 0.53, + "learning_rate": 9.368423609314709e-06, + "loss": 0.2919, + "step": 8848 + }, + { + "epoch": 0.53, + "learning_rate": 9.366470880595434e-06, + "loss": 0.3058, + "step": 8849 + }, + { + "epoch": 0.53, + "learning_rate": 9.364518176130617e-06, + "loss": 0.2753, + "step": 8850 + }, + { + "epoch": 0.53, + "learning_rate": 9.362565495995016e-06, + "loss": 0.2937, + "step": 8851 + }, + { + "epoch": 0.53, + "learning_rate": 9.360612840263393e-06, + "loss": 0.3098, + "step": 8852 + }, + { + "epoch": 0.53, + "learning_rate": 9.358660209010502e-06, + "loss": 0.3005, + "step": 8853 + }, + { + "epoch": 0.53, + "learning_rate": 9.3567076023111e-06, + "loss": 0.3019, + "step": 8854 + }, + { + "epoch": 0.53, + "learning_rate": 9.354755020239945e-06, + "loss": 0.3052, + "step": 8855 + }, + { + "epoch": 0.54, + "learning_rate": 9.352802462871784e-06, + "loss": 0.3143, + "step": 8856 + }, + { + "epoch": 0.54, + "learning_rate": 9.350849930281376e-06, + "loss": 0.2985, + "step": 8857 + }, + { + "epoch": 0.54, + "learning_rate": 9.34889742254347e-06, + "loss": 0.3136, + "step": 8858 + }, + { + "epoch": 0.54, + "learning_rate": 9.346944939732816e-06, + "loss": 0.3186, + "step": 8859 + }, + { + "epoch": 0.54, + "learning_rate": 9.34499248192417e-06, + "loss": 0.3195, + "step": 8860 + }, + { + "epoch": 0.54, + "learning_rate": 9.343040049192275e-06, + "loss": 0.2888, + "step": 8861 + }, + { + "epoch": 0.54, + "learning_rate": 9.341087641611883e-06, + "loss": 0.3073, + "step": 8862 + }, + { + "epoch": 0.54, + "learning_rate": 9.33913525925774e-06, + "loss": 0.2913, + "step": 8863 + }, + { + "epoch": 0.54, + "learning_rate": 9.33718290220459e-06, + "loss": 0.3154, + "step": 8864 + }, + { + "epoch": 0.54, + "learning_rate": 9.335230570527184e-06, + "loss": 0.3078, + "step": 8865 + }, + { + "epoch": 0.54, + "learning_rate": 9.333278264300259e-06, + "loss": 0.3195, + "step": 8866 + }, + { + "epoch": 0.54, + "learning_rate": 9.331325983598567e-06, + "loss": 0.3225, + "step": 8867 + }, + { + "epoch": 0.54, + "learning_rate": 9.329373728496845e-06, + "loss": 0.3273, + "step": 8868 + }, + { + "epoch": 0.54, + "learning_rate": 9.327421499069833e-06, + "loss": 0.3418, + "step": 8869 + }, + { + "epoch": 0.54, + "learning_rate": 9.325469295392275e-06, + "loss": 0.3139, + "step": 8870 + }, + { + "epoch": 0.54, + "learning_rate": 9.32351711753891e-06, + "loss": 0.2997, + "step": 8871 + }, + { + "epoch": 0.54, + "learning_rate": 9.321564965584475e-06, + "loss": 0.3181, + "step": 8872 + }, + { + "epoch": 0.54, + "learning_rate": 9.31961283960371e-06, + "loss": 0.3203, + "step": 8873 + }, + { + "epoch": 0.54, + "learning_rate": 9.317660739671348e-06, + "loss": 0.3025, + "step": 8874 + }, + { + "epoch": 0.54, + "learning_rate": 9.315708665862127e-06, + "loss": 0.2966, + "step": 8875 + }, + { + "epoch": 0.54, + "learning_rate": 9.313756618250783e-06, + "loss": 0.3037, + "step": 8876 + }, + { + "epoch": 0.54, + "learning_rate": 9.311804596912043e-06, + "loss": 0.3195, + "step": 8877 + }, + { + "epoch": 0.54, + "learning_rate": 9.30985260192065e-06, + "loss": 0.3039, + "step": 8878 + }, + { + "epoch": 0.54, + "learning_rate": 9.307900633351325e-06, + "loss": 0.308, + "step": 8879 + }, + { + "epoch": 0.54, + "learning_rate": 9.305948691278804e-06, + "loss": 0.3175, + "step": 8880 + }, + { + "epoch": 0.54, + "learning_rate": 9.303996775777817e-06, + "loss": 0.2819, + "step": 8881 + }, + { + "epoch": 0.54, + "learning_rate": 9.302044886923088e-06, + "loss": 0.2726, + "step": 8882 + }, + { + "epoch": 0.54, + "learning_rate": 9.300093024789353e-06, + "loss": 0.2787, + "step": 8883 + }, + { + "epoch": 0.54, + "learning_rate": 9.29814118945133e-06, + "loss": 0.3203, + "step": 8884 + }, + { + "epoch": 0.54, + "learning_rate": 9.296189380983747e-06, + "loss": 0.3128, + "step": 8885 + }, + { + "epoch": 0.54, + "learning_rate": 9.294237599461327e-06, + "loss": 0.2999, + "step": 8886 + }, + { + "epoch": 0.54, + "learning_rate": 9.292285844958795e-06, + "loss": 0.3343, + "step": 8887 + }, + { + "epoch": 0.54, + "learning_rate": 9.290334117550873e-06, + "loss": 0.3161, + "step": 8888 + }, + { + "epoch": 0.54, + "learning_rate": 9.288382417312284e-06, + "loss": 0.3278, + "step": 8889 + }, + { + "epoch": 0.54, + "learning_rate": 9.286430744317748e-06, + "loss": 0.2979, + "step": 8890 + }, + { + "epoch": 0.54, + "learning_rate": 9.284479098641977e-06, + "loss": 0.3239, + "step": 8891 + }, + { + "epoch": 0.54, + "learning_rate": 9.2825274803597e-06, + "loss": 0.2903, + "step": 8892 + }, + { + "epoch": 0.54, + "learning_rate": 9.28057588954563e-06, + "loss": 0.3209, + "step": 8893 + }, + { + "epoch": 0.54, + "learning_rate": 9.278624326274478e-06, + "loss": 0.2945, + "step": 8894 + }, + { + "epoch": 0.54, + "learning_rate": 9.276672790620965e-06, + "loss": 0.2746, + "step": 8895 + }, + { + "epoch": 0.54, + "learning_rate": 9.274721282659802e-06, + "loss": 0.2769, + "step": 8896 + }, + { + "epoch": 0.54, + "learning_rate": 9.272769802465705e-06, + "loss": 0.3218, + "step": 8897 + }, + { + "epoch": 0.54, + "learning_rate": 9.270818350113381e-06, + "loss": 0.2958, + "step": 8898 + }, + { + "epoch": 0.54, + "learning_rate": 9.268866925677545e-06, + "loss": 0.2796, + "step": 8899 + }, + { + "epoch": 0.54, + "learning_rate": 9.266915529232904e-06, + "loss": 0.3114, + "step": 8900 + }, + { + "epoch": 0.54, + "learning_rate": 9.264964160854169e-06, + "loss": 0.2946, + "step": 8901 + }, + { + "epoch": 0.54, + "learning_rate": 9.263012820616045e-06, + "loss": 0.3167, + "step": 8902 + }, + { + "epoch": 0.54, + "learning_rate": 9.26106150859324e-06, + "loss": 0.2956, + "step": 8903 + }, + { + "epoch": 0.54, + "learning_rate": 9.259110224860459e-06, + "loss": 0.344, + "step": 8904 + }, + { + "epoch": 0.54, + "learning_rate": 9.257158969492407e-06, + "loss": 0.2825, + "step": 8905 + }, + { + "epoch": 0.54, + "learning_rate": 9.255207742563783e-06, + "loss": 0.3062, + "step": 8906 + }, + { + "epoch": 0.54, + "learning_rate": 9.253256544149296e-06, + "loss": 0.309, + "step": 8907 + }, + { + "epoch": 0.54, + "learning_rate": 9.251305374323644e-06, + "loss": 0.3189, + "step": 8908 + }, + { + "epoch": 0.54, + "learning_rate": 9.249354233161523e-06, + "loss": 0.3153, + "step": 8909 + }, + { + "epoch": 0.54, + "learning_rate": 9.247403120737639e-06, + "loss": 0.2868, + "step": 8910 + }, + { + "epoch": 0.54, + "learning_rate": 9.245452037126684e-06, + "loss": 0.2945, + "step": 8911 + }, + { + "epoch": 0.54, + "learning_rate": 9.243500982403355e-06, + "loss": 0.2924, + "step": 8912 + }, + { + "epoch": 0.54, + "learning_rate": 9.241549956642352e-06, + "loss": 0.3083, + "step": 8913 + }, + { + "epoch": 0.54, + "learning_rate": 9.239598959918363e-06, + "loss": 0.2877, + "step": 8914 + }, + { + "epoch": 0.54, + "learning_rate": 9.237647992306089e-06, + "loss": 0.3192, + "step": 8915 + }, + { + "epoch": 0.54, + "learning_rate": 9.235697053880215e-06, + "loss": 0.3131, + "step": 8916 + }, + { + "epoch": 0.54, + "learning_rate": 9.233746144715439e-06, + "loss": 0.3248, + "step": 8917 + }, + { + "epoch": 0.54, + "learning_rate": 9.231795264886442e-06, + "loss": 0.3184, + "step": 8918 + }, + { + "epoch": 0.54, + "learning_rate": 9.22984441446792e-06, + "loss": 0.3251, + "step": 8919 + }, + { + "epoch": 0.54, + "learning_rate": 9.22789359353456e-06, + "loss": 0.3242, + "step": 8920 + }, + { + "epoch": 0.54, + "learning_rate": 9.225942802161042e-06, + "loss": 0.2991, + "step": 8921 + }, + { + "epoch": 0.54, + "learning_rate": 9.223992040422059e-06, + "loss": 0.3086, + "step": 8922 + }, + { + "epoch": 0.54, + "learning_rate": 9.222041308392292e-06, + "loss": 0.2854, + "step": 8923 + }, + { + "epoch": 0.54, + "learning_rate": 9.220090606146427e-06, + "loss": 0.3108, + "step": 8924 + }, + { + "epoch": 0.54, + "learning_rate": 9.218139933759142e-06, + "loss": 0.3256, + "step": 8925 + }, + { + "epoch": 0.54, + "learning_rate": 9.216189291305118e-06, + "loss": 0.3182, + "step": 8926 + }, + { + "epoch": 0.54, + "learning_rate": 9.21423867885904e-06, + "loss": 0.3025, + "step": 8927 + }, + { + "epoch": 0.54, + "learning_rate": 9.212288096495581e-06, + "loss": 0.2991, + "step": 8928 + }, + { + "epoch": 0.54, + "learning_rate": 9.21033754428942e-06, + "loss": 0.2994, + "step": 8929 + }, + { + "epoch": 0.54, + "learning_rate": 9.208387022315236e-06, + "loss": 0.2713, + "step": 8930 + }, + { + "epoch": 0.54, + "learning_rate": 9.206436530647699e-06, + "loss": 0.3071, + "step": 8931 + }, + { + "epoch": 0.54, + "learning_rate": 9.204486069361488e-06, + "loss": 0.3248, + "step": 8932 + }, + { + "epoch": 0.54, + "learning_rate": 9.202535638531272e-06, + "loss": 0.2974, + "step": 8933 + }, + { + "epoch": 0.54, + "learning_rate": 9.200585238231723e-06, + "loss": 0.2797, + "step": 8934 + }, + { + "epoch": 0.54, + "learning_rate": 9.198634868537516e-06, + "loss": 0.3093, + "step": 8935 + }, + { + "epoch": 0.54, + "learning_rate": 9.196684529523314e-06, + "loss": 0.2974, + "step": 8936 + }, + { + "epoch": 0.54, + "learning_rate": 9.19473422126379e-06, + "loss": 0.3001, + "step": 8937 + }, + { + "epoch": 0.54, + "learning_rate": 9.192783943833609e-06, + "loss": 0.3061, + "step": 8938 + }, + { + "epoch": 0.54, + "learning_rate": 9.190833697307435e-06, + "loss": 0.2995, + "step": 8939 + }, + { + "epoch": 0.54, + "learning_rate": 9.188883481759934e-06, + "loss": 0.2783, + "step": 8940 + }, + { + "epoch": 0.54, + "learning_rate": 9.18693329726577e-06, + "loss": 0.3138, + "step": 8941 + }, + { + "epoch": 0.54, + "learning_rate": 9.184983143899606e-06, + "loss": 0.2991, + "step": 8942 + }, + { + "epoch": 0.54, + "learning_rate": 9.183033021736102e-06, + "loss": 0.317, + "step": 8943 + }, + { + "epoch": 0.54, + "learning_rate": 9.181082930849915e-06, + "loss": 0.3066, + "step": 8944 + }, + { + "epoch": 0.54, + "learning_rate": 9.179132871315707e-06, + "loss": 0.3061, + "step": 8945 + }, + { + "epoch": 0.54, + "learning_rate": 9.177182843208138e-06, + "loss": 0.3217, + "step": 8946 + }, + { + "epoch": 0.54, + "learning_rate": 9.175232846601857e-06, + "loss": 0.305, + "step": 8947 + }, + { + "epoch": 0.54, + "learning_rate": 9.173282881571525e-06, + "loss": 0.3234, + "step": 8948 + }, + { + "epoch": 0.54, + "learning_rate": 9.17133294819179e-06, + "loss": 0.3057, + "step": 8949 + }, + { + "epoch": 0.54, + "learning_rate": 9.169383046537312e-06, + "loss": 0.2919, + "step": 8950 + }, + { + "epoch": 0.54, + "learning_rate": 9.167433176682741e-06, + "loss": 0.3068, + "step": 8951 + }, + { + "epoch": 0.54, + "learning_rate": 9.165483338702721e-06, + "loss": 0.2987, + "step": 8952 + }, + { + "epoch": 0.54, + "learning_rate": 9.163533532671905e-06, + "loss": 0.2822, + "step": 8953 + }, + { + "epoch": 0.54, + "learning_rate": 9.16158375866494e-06, + "loss": 0.2863, + "step": 8954 + }, + { + "epoch": 0.54, + "learning_rate": 9.159634016756474e-06, + "loss": 0.3248, + "step": 8955 + }, + { + "epoch": 0.54, + "learning_rate": 9.157684307021147e-06, + "loss": 0.3463, + "step": 8956 + }, + { + "epoch": 0.54, + "learning_rate": 9.155734629533612e-06, + "loss": 0.3119, + "step": 8957 + }, + { + "epoch": 0.54, + "learning_rate": 9.153784984368503e-06, + "loss": 0.3246, + "step": 8958 + }, + { + "epoch": 0.54, + "learning_rate": 9.151835371600467e-06, + "loss": 0.3158, + "step": 8959 + }, + { + "epoch": 0.54, + "learning_rate": 9.149885791304143e-06, + "loss": 0.2909, + "step": 8960 + }, + { + "epoch": 0.54, + "learning_rate": 9.14793624355417e-06, + "loss": 0.2983, + "step": 8961 + }, + { + "epoch": 0.54, + "learning_rate": 9.145986728425185e-06, + "loss": 0.3205, + "step": 8962 + }, + { + "epoch": 0.54, + "learning_rate": 9.144037245991823e-06, + "loss": 0.2944, + "step": 8963 + }, + { + "epoch": 0.54, + "learning_rate": 9.142087796328723e-06, + "loss": 0.3083, + "step": 8964 + }, + { + "epoch": 0.54, + "learning_rate": 9.14013837951052e-06, + "loss": 0.3253, + "step": 8965 + }, + { + "epoch": 0.54, + "learning_rate": 9.138188995611839e-06, + "loss": 0.3047, + "step": 8966 + }, + { + "epoch": 0.54, + "learning_rate": 9.136239644707323e-06, + "loss": 0.3061, + "step": 8967 + }, + { + "epoch": 0.54, + "learning_rate": 9.134290326871592e-06, + "loss": 0.2916, + "step": 8968 + }, + { + "epoch": 0.54, + "learning_rate": 9.132341042179279e-06, + "loss": 0.316, + "step": 8969 + }, + { + "epoch": 0.54, + "learning_rate": 9.130391790705013e-06, + "loss": 0.3094, + "step": 8970 + }, + { + "epoch": 0.54, + "learning_rate": 9.128442572523418e-06, + "loss": 0.2857, + "step": 8971 + }, + { + "epoch": 0.54, + "learning_rate": 9.126493387709124e-06, + "loss": 0.3259, + "step": 8972 + }, + { + "epoch": 0.54, + "learning_rate": 9.124544236336749e-06, + "loss": 0.309, + "step": 8973 + }, + { + "epoch": 0.54, + "learning_rate": 9.122595118480916e-06, + "loss": 0.3164, + "step": 8974 + }, + { + "epoch": 0.54, + "learning_rate": 9.120646034216251e-06, + "loss": 0.2899, + "step": 8975 + }, + { + "epoch": 0.54, + "learning_rate": 9.11869698361737e-06, + "loss": 0.3146, + "step": 8976 + }, + { + "epoch": 0.54, + "learning_rate": 9.116747966758895e-06, + "loss": 0.3105, + "step": 8977 + }, + { + "epoch": 0.54, + "learning_rate": 9.114798983715441e-06, + "loss": 0.3426, + "step": 8978 + }, + { + "epoch": 0.54, + "learning_rate": 9.112850034561625e-06, + "loss": 0.3094, + "step": 8979 + }, + { + "epoch": 0.54, + "learning_rate": 9.110901119372062e-06, + "loss": 0.3172, + "step": 8980 + }, + { + "epoch": 0.54, + "learning_rate": 9.108952238221366e-06, + "loss": 0.339, + "step": 8981 + }, + { + "epoch": 0.54, + "learning_rate": 9.107003391184146e-06, + "loss": 0.3131, + "step": 8982 + }, + { + "epoch": 0.54, + "learning_rate": 9.105054578335019e-06, + "loss": 0.3142, + "step": 8983 + }, + { + "epoch": 0.54, + "learning_rate": 9.103105799748588e-06, + "loss": 0.2863, + "step": 8984 + }, + { + "epoch": 0.54, + "learning_rate": 9.101157055499472e-06, + "loss": 0.3131, + "step": 8985 + }, + { + "epoch": 0.54, + "learning_rate": 9.099208345662263e-06, + "loss": 0.2888, + "step": 8986 + }, + { + "epoch": 0.54, + "learning_rate": 9.097259670311579e-06, + "loss": 0.2849, + "step": 8987 + }, + { + "epoch": 0.54, + "learning_rate": 9.095311029522016e-06, + "loss": 0.3203, + "step": 8988 + }, + { + "epoch": 0.54, + "learning_rate": 9.093362423368185e-06, + "loss": 0.3027, + "step": 8989 + }, + { + "epoch": 0.54, + "learning_rate": 9.09141385192468e-06, + "loss": 0.3105, + "step": 8990 + }, + { + "epoch": 0.54, + "learning_rate": 9.089465315266106e-06, + "loss": 0.3117, + "step": 8991 + }, + { + "epoch": 0.54, + "learning_rate": 9.087516813467064e-06, + "loss": 0.2919, + "step": 8992 + }, + { + "epoch": 0.54, + "learning_rate": 9.085568346602146e-06, + "loss": 0.3061, + "step": 8993 + }, + { + "epoch": 0.54, + "learning_rate": 9.083619914745953e-06, + "loss": 0.3094, + "step": 8994 + }, + { + "epoch": 0.54, + "learning_rate": 9.08167151797308e-06, + "loss": 0.3073, + "step": 8995 + }, + { + "epoch": 0.54, + "learning_rate": 9.079723156358116e-06, + "loss": 0.2966, + "step": 8996 + }, + { + "epoch": 0.54, + "learning_rate": 9.077774829975659e-06, + "loss": 0.3058, + "step": 8997 + }, + { + "epoch": 0.54, + "learning_rate": 9.075826538900296e-06, + "loss": 0.2888, + "step": 8998 + }, + { + "epoch": 0.54, + "learning_rate": 9.07387828320662e-06, + "loss": 0.3234, + "step": 8999 + }, + { + "epoch": 0.54, + "learning_rate": 9.07193006296922e-06, + "loss": 0.3227, + "step": 9000 + }, + { + "epoch": 0.54, + "learning_rate": 9.069981878262677e-06, + "loss": 0.2891, + "step": 9001 + }, + { + "epoch": 0.54, + "learning_rate": 9.068033729161583e-06, + "loss": 0.2946, + "step": 9002 + }, + { + "epoch": 0.54, + "learning_rate": 9.066085615740518e-06, + "loss": 0.3119, + "step": 9003 + }, + { + "epoch": 0.54, + "learning_rate": 9.064137538074068e-06, + "loss": 0.3158, + "step": 9004 + }, + { + "epoch": 0.54, + "learning_rate": 9.062189496236813e-06, + "loss": 0.3083, + "step": 9005 + }, + { + "epoch": 0.54, + "learning_rate": 9.060241490303331e-06, + "loss": 0.3103, + "step": 9006 + }, + { + "epoch": 0.54, + "learning_rate": 9.058293520348206e-06, + "loss": 0.3002, + "step": 9007 + }, + { + "epoch": 0.54, + "learning_rate": 9.056345586446012e-06, + "loss": 0.305, + "step": 9008 + }, + { + "epoch": 0.54, + "learning_rate": 9.054397688671325e-06, + "loss": 0.2825, + "step": 9009 + }, + { + "epoch": 0.54, + "learning_rate": 9.05244982709872e-06, + "loss": 0.3094, + "step": 9010 + }, + { + "epoch": 0.54, + "learning_rate": 9.05050200180277e-06, + "loss": 0.2977, + "step": 9011 + }, + { + "epoch": 0.54, + "learning_rate": 9.048554212858049e-06, + "loss": 0.2716, + "step": 9012 + }, + { + "epoch": 0.54, + "learning_rate": 9.046606460339126e-06, + "loss": 0.322, + "step": 9013 + }, + { + "epoch": 0.54, + "learning_rate": 9.044658744320567e-06, + "loss": 0.3301, + "step": 9014 + }, + { + "epoch": 0.54, + "learning_rate": 9.042711064876946e-06, + "loss": 0.2963, + "step": 9015 + }, + { + "epoch": 0.54, + "learning_rate": 9.040763422082822e-06, + "loss": 0.3167, + "step": 9016 + }, + { + "epoch": 0.54, + "learning_rate": 9.038815816012768e-06, + "loss": 0.2994, + "step": 9017 + }, + { + "epoch": 0.54, + "learning_rate": 9.036868246741342e-06, + "loss": 0.3011, + "step": 9018 + }, + { + "epoch": 0.54, + "learning_rate": 9.03492071434311e-06, + "loss": 0.2944, + "step": 9019 + }, + { + "epoch": 0.54, + "learning_rate": 9.032973218892627e-06, + "loss": 0.2878, + "step": 9020 + }, + { + "epoch": 0.54, + "learning_rate": 9.031025760464453e-06, + "loss": 0.2779, + "step": 9021 + }, + { + "epoch": 0.55, + "learning_rate": 9.02907833913315e-06, + "loss": 0.3373, + "step": 9022 + }, + { + "epoch": 0.55, + "learning_rate": 9.02713095497327e-06, + "loss": 0.3104, + "step": 9023 + }, + { + "epoch": 0.55, + "learning_rate": 9.025183608059374e-06, + "loss": 0.2999, + "step": 9024 + }, + { + "epoch": 0.55, + "learning_rate": 9.02323629846601e-06, + "loss": 0.3019, + "step": 9025 + }, + { + "epoch": 0.55, + "learning_rate": 9.021289026267733e-06, + "loss": 0.3205, + "step": 9026 + }, + { + "epoch": 0.55, + "learning_rate": 9.019341791539091e-06, + "loss": 0.3175, + "step": 9027 + }, + { + "epoch": 0.55, + "learning_rate": 9.017394594354637e-06, + "loss": 0.3041, + "step": 9028 + }, + { + "epoch": 0.55, + "learning_rate": 9.015447434788915e-06, + "loss": 0.3158, + "step": 9029 + }, + { + "epoch": 0.55, + "learning_rate": 9.013500312916476e-06, + "loss": 0.304, + "step": 9030 + }, + { + "epoch": 0.55, + "learning_rate": 9.011553228811858e-06, + "loss": 0.2909, + "step": 9031 + }, + { + "epoch": 0.55, + "learning_rate": 9.009606182549612e-06, + "loss": 0.3231, + "step": 9032 + }, + { + "epoch": 0.55, + "learning_rate": 9.007659174204275e-06, + "loss": 0.3075, + "step": 9033 + }, + { + "epoch": 0.55, + "learning_rate": 9.00571220385039e-06, + "loss": 0.3203, + "step": 9034 + }, + { + "epoch": 0.55, + "learning_rate": 9.003765271562498e-06, + "loss": 0.3203, + "step": 9035 + }, + { + "epoch": 0.55, + "learning_rate": 9.001818377415128e-06, + "loss": 0.2787, + "step": 9036 + }, + { + "epoch": 0.55, + "learning_rate": 8.999871521482828e-06, + "loss": 0.3142, + "step": 9037 + }, + { + "epoch": 0.55, + "learning_rate": 8.997924703840125e-06, + "loss": 0.3276, + "step": 9038 + }, + { + "epoch": 0.55, + "learning_rate": 8.995977924561553e-06, + "loss": 0.2966, + "step": 9039 + }, + { + "epoch": 0.55, + "learning_rate": 8.994031183721648e-06, + "loss": 0.2997, + "step": 9040 + }, + { + "epoch": 0.55, + "learning_rate": 8.992084481394934e-06, + "loss": 0.3039, + "step": 9041 + }, + { + "epoch": 0.55, + "learning_rate": 8.990137817655946e-06, + "loss": 0.3022, + "step": 9042 + }, + { + "epoch": 0.55, + "learning_rate": 8.98819119257921e-06, + "loss": 0.2946, + "step": 9043 + }, + { + "epoch": 0.55, + "learning_rate": 8.986244606239247e-06, + "loss": 0.3175, + "step": 9044 + }, + { + "epoch": 0.55, + "learning_rate": 8.984298058710587e-06, + "loss": 0.2977, + "step": 9045 + }, + { + "epoch": 0.55, + "learning_rate": 8.982351550067751e-06, + "loss": 0.2864, + "step": 9046 + }, + { + "epoch": 0.55, + "learning_rate": 8.980405080385261e-06, + "loss": 0.303, + "step": 9047 + }, + { + "epoch": 0.55, + "learning_rate": 8.978458649737638e-06, + "loss": 0.2716, + "step": 9048 + }, + { + "epoch": 0.55, + "learning_rate": 8.976512258199396e-06, + "loss": 0.315, + "step": 9049 + }, + { + "epoch": 0.55, + "learning_rate": 8.974565905845057e-06, + "loss": 0.3145, + "step": 9050 + }, + { + "epoch": 0.55, + "learning_rate": 8.972619592749131e-06, + "loss": 0.2863, + "step": 9051 + }, + { + "epoch": 0.55, + "learning_rate": 8.97067331898614e-06, + "loss": 0.3178, + "step": 9052 + }, + { + "epoch": 0.55, + "learning_rate": 8.968727084630594e-06, + "loss": 0.3171, + "step": 9053 + }, + { + "epoch": 0.55, + "learning_rate": 8.966780889756998e-06, + "loss": 0.3181, + "step": 9054 + }, + { + "epoch": 0.55, + "learning_rate": 8.964834734439869e-06, + "loss": 0.317, + "step": 9055 + }, + { + "epoch": 0.55, + "learning_rate": 8.962888618753705e-06, + "loss": 0.3451, + "step": 9056 + }, + { + "epoch": 0.55, + "learning_rate": 8.960942542773023e-06, + "loss": 0.3033, + "step": 9057 + }, + { + "epoch": 0.55, + "learning_rate": 8.958996506572319e-06, + "loss": 0.3117, + "step": 9058 + }, + { + "epoch": 0.55, + "learning_rate": 8.957050510226106e-06, + "loss": 0.3119, + "step": 9059 + }, + { + "epoch": 0.55, + "learning_rate": 8.95510455380888e-06, + "loss": 0.3092, + "step": 9060 + }, + { + "epoch": 0.55, + "learning_rate": 8.953158637395138e-06, + "loss": 0.2932, + "step": 9061 + }, + { + "epoch": 0.55, + "learning_rate": 8.951212761059387e-06, + "loss": 0.2955, + "step": 9062 + }, + { + "epoch": 0.55, + "learning_rate": 8.949266924876118e-06, + "loss": 0.3178, + "step": 9063 + }, + { + "epoch": 0.55, + "learning_rate": 8.947321128919831e-06, + "loss": 0.3175, + "step": 9064 + }, + { + "epoch": 0.55, + "learning_rate": 8.945375373265017e-06, + "loss": 0.3033, + "step": 9065 + }, + { + "epoch": 0.55, + "learning_rate": 8.943429657986168e-06, + "loss": 0.2946, + "step": 9066 + }, + { + "epoch": 0.55, + "learning_rate": 8.94148398315778e-06, + "loss": 0.2944, + "step": 9067 + }, + { + "epoch": 0.55, + "learning_rate": 8.939538348854336e-06, + "loss": 0.2977, + "step": 9068 + }, + { + "epoch": 0.55, + "learning_rate": 8.93759275515033e-06, + "loss": 0.3365, + "step": 9069 + }, + { + "epoch": 0.55, + "learning_rate": 8.935647202120247e-06, + "loss": 0.2906, + "step": 9070 + }, + { + "epoch": 0.55, + "learning_rate": 8.933701689838567e-06, + "loss": 0.3163, + "step": 9071 + }, + { + "epoch": 0.55, + "learning_rate": 8.931756218379779e-06, + "loss": 0.3064, + "step": 9072 + }, + { + "epoch": 0.55, + "learning_rate": 8.929810787818362e-06, + "loss": 0.3253, + "step": 9073 + }, + { + "epoch": 0.55, + "learning_rate": 8.927865398228795e-06, + "loss": 0.3044, + "step": 9074 + }, + { + "epoch": 0.55, + "learning_rate": 8.925920049685562e-06, + "loss": 0.2896, + "step": 9075 + }, + { + "epoch": 0.55, + "learning_rate": 8.923974742263134e-06, + "loss": 0.3015, + "step": 9076 + }, + { + "epoch": 0.55, + "learning_rate": 8.92202947603599e-06, + "loss": 0.298, + "step": 9077 + }, + { + "epoch": 0.55, + "learning_rate": 8.920084251078605e-06, + "loss": 0.3156, + "step": 9078 + }, + { + "epoch": 0.55, + "learning_rate": 8.918139067465444e-06, + "loss": 0.3309, + "step": 9079 + }, + { + "epoch": 0.55, + "learning_rate": 8.916193925270987e-06, + "loss": 0.2822, + "step": 9080 + }, + { + "epoch": 0.55, + "learning_rate": 8.914248824569698e-06, + "loss": 0.2899, + "step": 9081 + }, + { + "epoch": 0.55, + "learning_rate": 8.912303765436045e-06, + "loss": 0.2839, + "step": 9082 + }, + { + "epoch": 0.55, + "learning_rate": 8.910358747944497e-06, + "loss": 0.314, + "step": 9083 + }, + { + "epoch": 0.55, + "learning_rate": 8.908413772169512e-06, + "loss": 0.3225, + "step": 9084 + }, + { + "epoch": 0.55, + "learning_rate": 8.906468838185562e-06, + "loss": 0.3189, + "step": 9085 + }, + { + "epoch": 0.55, + "learning_rate": 8.904523946067098e-06, + "loss": 0.3114, + "step": 9086 + }, + { + "epoch": 0.55, + "learning_rate": 8.902579095888591e-06, + "loss": 0.339, + "step": 9087 + }, + { + "epoch": 0.55, + "learning_rate": 8.900634287724487e-06, + "loss": 0.3103, + "step": 9088 + }, + { + "epoch": 0.55, + "learning_rate": 8.89868952164925e-06, + "loss": 0.3122, + "step": 9089 + }, + { + "epoch": 0.55, + "learning_rate": 8.89674479773733e-06, + "loss": 0.3075, + "step": 9090 + }, + { + "epoch": 0.55, + "learning_rate": 8.894800116063185e-06, + "loss": 0.3114, + "step": 9091 + }, + { + "epoch": 0.55, + "learning_rate": 8.892855476701265e-06, + "loss": 0.3136, + "step": 9092 + }, + { + "epoch": 0.55, + "learning_rate": 8.890910879726016e-06, + "loss": 0.3115, + "step": 9093 + }, + { + "epoch": 0.55, + "learning_rate": 8.888966325211893e-06, + "loss": 0.3181, + "step": 9094 + }, + { + "epoch": 0.55, + "learning_rate": 8.887021813233338e-06, + "loss": 0.2902, + "step": 9095 + }, + { + "epoch": 0.55, + "learning_rate": 8.885077343864795e-06, + "loss": 0.339, + "step": 9096 + }, + { + "epoch": 0.55, + "learning_rate": 8.88313291718071e-06, + "loss": 0.2864, + "step": 9097 + }, + { + "epoch": 0.55, + "learning_rate": 8.881188533255524e-06, + "loss": 0.274, + "step": 9098 + }, + { + "epoch": 0.55, + "learning_rate": 8.87924419216368e-06, + "loss": 0.293, + "step": 9099 + }, + { + "epoch": 0.55, + "learning_rate": 8.877299893979612e-06, + "loss": 0.3544, + "step": 9100 + }, + { + "epoch": 0.55, + "learning_rate": 8.875355638777757e-06, + "loss": 0.2826, + "step": 9101 + }, + { + "epoch": 0.55, + "learning_rate": 8.873411426632554e-06, + "loss": 0.2985, + "step": 9102 + }, + { + "epoch": 0.55, + "learning_rate": 8.871467257618433e-06, + "loss": 0.3129, + "step": 9103 + }, + { + "epoch": 0.55, + "learning_rate": 8.86952313180983e-06, + "loss": 0.2878, + "step": 9104 + }, + { + "epoch": 0.55, + "learning_rate": 8.86757904928117e-06, + "loss": 0.3251, + "step": 9105 + }, + { + "epoch": 0.55, + "learning_rate": 8.865635010106885e-06, + "loss": 0.2984, + "step": 9106 + }, + { + "epoch": 0.55, + "learning_rate": 8.863691014361401e-06, + "loss": 0.2853, + "step": 9107 + }, + { + "epoch": 0.55, + "learning_rate": 8.861747062119144e-06, + "loss": 0.302, + "step": 9108 + }, + { + "epoch": 0.55, + "learning_rate": 8.859803153454534e-06, + "loss": 0.3129, + "step": 9109 + }, + { + "epoch": 0.55, + "learning_rate": 8.857859288442e-06, + "loss": 0.2949, + "step": 9110 + }, + { + "epoch": 0.55, + "learning_rate": 8.855915467155954e-06, + "loss": 0.293, + "step": 9111 + }, + { + "epoch": 0.55, + "learning_rate": 8.85397168967082e-06, + "loss": 0.2856, + "step": 9112 + }, + { + "epoch": 0.55, + "learning_rate": 8.852027956061015e-06, + "loss": 0.3323, + "step": 9113 + }, + { + "epoch": 0.55, + "learning_rate": 8.850084266400952e-06, + "loss": 0.2857, + "step": 9114 + }, + { + "epoch": 0.55, + "learning_rate": 8.848140620765045e-06, + "loss": 0.3016, + "step": 9115 + }, + { + "epoch": 0.55, + "learning_rate": 8.846197019227704e-06, + "loss": 0.2944, + "step": 9116 + }, + { + "epoch": 0.55, + "learning_rate": 8.844253461863343e-06, + "loss": 0.3521, + "step": 9117 + }, + { + "epoch": 0.55, + "learning_rate": 8.84230994874637e-06, + "loss": 0.3086, + "step": 9118 + }, + { + "epoch": 0.55, + "learning_rate": 8.840366479951189e-06, + "loss": 0.2833, + "step": 9119 + }, + { + "epoch": 0.55, + "learning_rate": 8.838423055552208e-06, + "loss": 0.2825, + "step": 9120 + }, + { + "epoch": 0.55, + "learning_rate": 8.83647967562383e-06, + "loss": 0.309, + "step": 9121 + }, + { + "epoch": 0.55, + "learning_rate": 8.834536340240455e-06, + "loss": 0.3064, + "step": 9122 + }, + { + "epoch": 0.55, + "learning_rate": 8.832593049476482e-06, + "loss": 0.2812, + "step": 9123 + }, + { + "epoch": 0.55, + "learning_rate": 8.83064980340631e-06, + "loss": 0.3108, + "step": 9124 + }, + { + "epoch": 0.55, + "learning_rate": 8.828706602104338e-06, + "loss": 0.3016, + "step": 9125 + }, + { + "epoch": 0.55, + "learning_rate": 8.826763445644959e-06, + "loss": 0.2825, + "step": 9126 + }, + { + "epoch": 0.55, + "learning_rate": 8.824820334102568e-06, + "loss": 0.3211, + "step": 9127 + }, + { + "epoch": 0.55, + "learning_rate": 8.822877267551552e-06, + "loss": 0.2699, + "step": 9128 + }, + { + "epoch": 0.55, + "learning_rate": 8.820934246066305e-06, + "loss": 0.2912, + "step": 9129 + }, + { + "epoch": 0.55, + "learning_rate": 8.818991269721215e-06, + "loss": 0.3359, + "step": 9130 + }, + { + "epoch": 0.55, + "learning_rate": 8.817048338590662e-06, + "loss": 0.3128, + "step": 9131 + }, + { + "epoch": 0.55, + "learning_rate": 8.81510545274904e-06, + "loss": 0.293, + "step": 9132 + }, + { + "epoch": 0.55, + "learning_rate": 8.813162612270722e-06, + "loss": 0.2917, + "step": 9133 + }, + { + "epoch": 0.55, + "learning_rate": 8.8112198172301e-06, + "loss": 0.31, + "step": 9134 + }, + { + "epoch": 0.55, + "learning_rate": 8.809277067701544e-06, + "loss": 0.2967, + "step": 9135 + }, + { + "epoch": 0.55, + "learning_rate": 8.807334363759435e-06, + "loss": 0.303, + "step": 9136 + }, + { + "epoch": 0.55, + "learning_rate": 8.805391705478149e-06, + "loss": 0.2868, + "step": 9137 + }, + { + "epoch": 0.55, + "learning_rate": 8.803449092932057e-06, + "loss": 0.31, + "step": 9138 + }, + { + "epoch": 0.55, + "learning_rate": 8.801506526195537e-06, + "loss": 0.3223, + "step": 9139 + }, + { + "epoch": 0.55, + "learning_rate": 8.799564005342958e-06, + "loss": 0.3484, + "step": 9140 + }, + { + "epoch": 0.55, + "learning_rate": 8.797621530448684e-06, + "loss": 0.2843, + "step": 9141 + }, + { + "epoch": 0.55, + "learning_rate": 8.795679101587088e-06, + "loss": 0.2939, + "step": 9142 + }, + { + "epoch": 0.55, + "learning_rate": 8.79373671883253e-06, + "loss": 0.274, + "step": 9143 + }, + { + "epoch": 0.55, + "learning_rate": 8.791794382259377e-06, + "loss": 0.322, + "step": 9144 + }, + { + "epoch": 0.55, + "learning_rate": 8.78985209194199e-06, + "loss": 0.2801, + "step": 9145 + }, + { + "epoch": 0.55, + "learning_rate": 8.787909847954728e-06, + "loss": 0.3237, + "step": 9146 + }, + { + "epoch": 0.55, + "learning_rate": 8.785967650371951e-06, + "loss": 0.3008, + "step": 9147 + }, + { + "epoch": 0.55, + "learning_rate": 8.784025499268015e-06, + "loss": 0.3168, + "step": 9148 + }, + { + "epoch": 0.55, + "learning_rate": 8.782083394717272e-06, + "loss": 0.3075, + "step": 9149 + }, + { + "epoch": 0.55, + "learning_rate": 8.780141336794079e-06, + "loss": 0.2838, + "step": 9150 + }, + { + "epoch": 0.55, + "learning_rate": 8.778199325572782e-06, + "loss": 0.2656, + "step": 9151 + }, + { + "epoch": 0.55, + "learning_rate": 8.776257361127737e-06, + "loss": 0.3019, + "step": 9152 + }, + { + "epoch": 0.55, + "learning_rate": 8.774315443533286e-06, + "loss": 0.2966, + "step": 9153 + }, + { + "epoch": 0.55, + "learning_rate": 8.772373572863777e-06, + "loss": 0.2949, + "step": 9154 + }, + { + "epoch": 0.55, + "learning_rate": 8.770431749193554e-06, + "loss": 0.2681, + "step": 9155 + }, + { + "epoch": 0.55, + "learning_rate": 8.768489972596958e-06, + "loss": 0.2746, + "step": 9156 + }, + { + "epoch": 0.55, + "learning_rate": 8.76654824314833e-06, + "loss": 0.3309, + "step": 9157 + }, + { + "epoch": 0.55, + "learning_rate": 8.764606560922006e-06, + "loss": 0.2988, + "step": 9158 + }, + { + "epoch": 0.55, + "learning_rate": 8.762664925992327e-06, + "loss": 0.2893, + "step": 9159 + }, + { + "epoch": 0.55, + "learning_rate": 8.760723338433624e-06, + "loss": 0.3231, + "step": 9160 + }, + { + "epoch": 0.55, + "learning_rate": 8.758781798320233e-06, + "loss": 0.2967, + "step": 9161 + }, + { + "epoch": 0.55, + "learning_rate": 8.756840305726485e-06, + "loss": 0.3248, + "step": 9162 + }, + { + "epoch": 0.55, + "learning_rate": 8.754898860726705e-06, + "loss": 0.3312, + "step": 9163 + }, + { + "epoch": 0.55, + "learning_rate": 8.752957463395227e-06, + "loss": 0.2907, + "step": 9164 + }, + { + "epoch": 0.55, + "learning_rate": 8.751016113806373e-06, + "loss": 0.3005, + "step": 9165 + }, + { + "epoch": 0.55, + "learning_rate": 8.749074812034467e-06, + "loss": 0.3198, + "step": 9166 + }, + { + "epoch": 0.55, + "learning_rate": 8.747133558153833e-06, + "loss": 0.3078, + "step": 9167 + }, + { + "epoch": 0.55, + "learning_rate": 8.745192352238789e-06, + "loss": 0.2959, + "step": 9168 + }, + { + "epoch": 0.55, + "learning_rate": 8.743251194363657e-06, + "loss": 0.3069, + "step": 9169 + }, + { + "epoch": 0.55, + "learning_rate": 8.74131008460275e-06, + "loss": 0.2792, + "step": 9170 + }, + { + "epoch": 0.55, + "learning_rate": 8.739369023030384e-06, + "loss": 0.3019, + "step": 9171 + }, + { + "epoch": 0.55, + "learning_rate": 8.737428009720872e-06, + "loss": 0.2902, + "step": 9172 + }, + { + "epoch": 0.55, + "learning_rate": 8.735487044748523e-06, + "loss": 0.3189, + "step": 9173 + }, + { + "epoch": 0.55, + "learning_rate": 8.733546128187652e-06, + "loss": 0.3343, + "step": 9174 + }, + { + "epoch": 0.55, + "learning_rate": 8.731605260112561e-06, + "loss": 0.3025, + "step": 9175 + }, + { + "epoch": 0.55, + "learning_rate": 8.729664440597556e-06, + "loss": 0.2941, + "step": 9176 + }, + { + "epoch": 0.55, + "learning_rate": 8.727723669716943e-06, + "loss": 0.2863, + "step": 9177 + }, + { + "epoch": 0.55, + "learning_rate": 8.72578294754502e-06, + "loss": 0.2997, + "step": 9178 + }, + { + "epoch": 0.55, + "learning_rate": 8.723842274156094e-06, + "loss": 0.3101, + "step": 9179 + }, + { + "epoch": 0.55, + "learning_rate": 8.721901649624456e-06, + "loss": 0.2966, + "step": 9180 + }, + { + "epoch": 0.55, + "learning_rate": 8.719961074024404e-06, + "loss": 0.2963, + "step": 9181 + }, + { + "epoch": 0.55, + "learning_rate": 8.718020547430235e-06, + "loss": 0.291, + "step": 9182 + }, + { + "epoch": 0.55, + "learning_rate": 8.716080069916239e-06, + "loss": 0.3133, + "step": 9183 + }, + { + "epoch": 0.55, + "learning_rate": 8.714139641556707e-06, + "loss": 0.3083, + "step": 9184 + }, + { + "epoch": 0.55, + "learning_rate": 8.712199262425928e-06, + "loss": 0.3023, + "step": 9185 + }, + { + "epoch": 0.55, + "learning_rate": 8.710258932598186e-06, + "loss": 0.3177, + "step": 9186 + }, + { + "epoch": 0.56, + "learning_rate": 8.708318652147771e-06, + "loss": 0.3039, + "step": 9187 + }, + { + "epoch": 0.56, + "learning_rate": 8.706378421148964e-06, + "loss": 0.3075, + "step": 9188 + }, + { + "epoch": 0.56, + "learning_rate": 8.704438239676048e-06, + "loss": 0.286, + "step": 9189 + }, + { + "epoch": 0.56, + "learning_rate": 8.702498107803296e-06, + "loss": 0.2896, + "step": 9190 + }, + { + "epoch": 0.56, + "learning_rate": 8.70055802560499e-06, + "loss": 0.315, + "step": 9191 + }, + { + "epoch": 0.56, + "learning_rate": 8.698617993155406e-06, + "loss": 0.3072, + "step": 9192 + }, + { + "epoch": 0.56, + "learning_rate": 8.696678010528812e-06, + "loss": 0.293, + "step": 9193 + }, + { + "epoch": 0.56, + "learning_rate": 8.694738077799487e-06, + "loss": 0.3052, + "step": 9194 + }, + { + "epoch": 0.56, + "learning_rate": 8.692798195041695e-06, + "loss": 0.2799, + "step": 9195 + }, + { + "epoch": 0.56, + "learning_rate": 8.690858362329709e-06, + "loss": 0.3209, + "step": 9196 + }, + { + "epoch": 0.56, + "learning_rate": 8.68891857973779e-06, + "loss": 0.2941, + "step": 9197 + }, + { + "epoch": 0.56, + "learning_rate": 8.686978847340203e-06, + "loss": 0.3184, + "step": 9198 + }, + { + "epoch": 0.56, + "learning_rate": 8.685039165211216e-06, + "loss": 0.3036, + "step": 9199 + }, + { + "epoch": 0.56, + "learning_rate": 8.68309953342508e-06, + "loss": 0.2847, + "step": 9200 + }, + { + "epoch": 0.56, + "learning_rate": 8.681159952056058e-06, + "loss": 0.2725, + "step": 9201 + }, + { + "epoch": 0.56, + "learning_rate": 8.679220421178406e-06, + "loss": 0.3276, + "step": 9202 + }, + { + "epoch": 0.56, + "learning_rate": 8.677280940866379e-06, + "loss": 0.2907, + "step": 9203 + }, + { + "epoch": 0.56, + "learning_rate": 8.675341511194228e-06, + "loss": 0.2913, + "step": 9204 + }, + { + "epoch": 0.56, + "learning_rate": 8.673402132236203e-06, + "loss": 0.3058, + "step": 9205 + }, + { + "epoch": 0.56, + "learning_rate": 8.671462804066553e-06, + "loss": 0.2807, + "step": 9206 + }, + { + "epoch": 0.56, + "learning_rate": 8.669523526759524e-06, + "loss": 0.284, + "step": 9207 + }, + { + "epoch": 0.56, + "learning_rate": 8.667584300389362e-06, + "loss": 0.3108, + "step": 9208 + }, + { + "epoch": 0.56, + "learning_rate": 8.665645125030312e-06, + "loss": 0.284, + "step": 9209 + }, + { + "epoch": 0.56, + "learning_rate": 8.66370600075661e-06, + "loss": 0.3022, + "step": 9210 + }, + { + "epoch": 0.56, + "learning_rate": 8.661766927642494e-06, + "loss": 0.3041, + "step": 9211 + }, + { + "epoch": 0.56, + "learning_rate": 8.659827905762207e-06, + "loss": 0.2921, + "step": 9212 + }, + { + "epoch": 0.56, + "learning_rate": 8.657888935189978e-06, + "loss": 0.3156, + "step": 9213 + }, + { + "epoch": 0.56, + "learning_rate": 8.655950016000044e-06, + "loss": 0.2891, + "step": 9214 + }, + { + "epoch": 0.56, + "learning_rate": 8.654011148266631e-06, + "loss": 0.2863, + "step": 9215 + }, + { + "epoch": 0.56, + "learning_rate": 8.652072332063973e-06, + "loss": 0.3015, + "step": 9216 + }, + { + "epoch": 0.56, + "learning_rate": 8.650133567466295e-06, + "loss": 0.3086, + "step": 9217 + }, + { + "epoch": 0.56, + "learning_rate": 8.648194854547822e-06, + "loss": 0.3167, + "step": 9218 + }, + { + "epoch": 0.56, + "learning_rate": 8.646256193382775e-06, + "loss": 0.3036, + "step": 9219 + }, + { + "epoch": 0.56, + "learning_rate": 8.64431758404538e-06, + "loss": 0.2634, + "step": 9220 + }, + { + "epoch": 0.56, + "learning_rate": 8.64237902660985e-06, + "loss": 0.2987, + "step": 9221 + }, + { + "epoch": 0.56, + "learning_rate": 8.640440521150406e-06, + "loss": 0.3228, + "step": 9222 + }, + { + "epoch": 0.56, + "learning_rate": 8.638502067741266e-06, + "loss": 0.3002, + "step": 9223 + }, + { + "epoch": 0.56, + "learning_rate": 8.636563666456636e-06, + "loss": 0.2856, + "step": 9224 + }, + { + "epoch": 0.56, + "learning_rate": 8.634625317370727e-06, + "loss": 0.292, + "step": 9225 + }, + { + "epoch": 0.56, + "learning_rate": 8.632687020557755e-06, + "loss": 0.3245, + "step": 9226 + }, + { + "epoch": 0.56, + "learning_rate": 8.630748776091922e-06, + "loss": 0.2766, + "step": 9227 + }, + { + "epoch": 0.56, + "learning_rate": 8.628810584047434e-06, + "loss": 0.351, + "step": 9228 + }, + { + "epoch": 0.56, + "learning_rate": 8.626872444498495e-06, + "loss": 0.3036, + "step": 9229 + }, + { + "epoch": 0.56, + "learning_rate": 8.624934357519303e-06, + "loss": 0.2815, + "step": 9230 + }, + { + "epoch": 0.56, + "learning_rate": 8.622996323184063e-06, + "loss": 0.3016, + "step": 9231 + }, + { + "epoch": 0.56, + "learning_rate": 8.621058341566967e-06, + "loss": 0.284, + "step": 9232 + }, + { + "epoch": 0.56, + "learning_rate": 8.619120412742213e-06, + "loss": 0.3167, + "step": 9233 + }, + { + "epoch": 0.56, + "learning_rate": 8.617182536783991e-06, + "loss": 0.3066, + "step": 9234 + }, + { + "epoch": 0.56, + "learning_rate": 8.615244713766495e-06, + "loss": 0.297, + "step": 9235 + }, + { + "epoch": 0.56, + "learning_rate": 8.613306943763909e-06, + "loss": 0.3357, + "step": 9236 + }, + { + "epoch": 0.56, + "learning_rate": 8.611369226850428e-06, + "loss": 0.2748, + "step": 9237 + }, + { + "epoch": 0.56, + "learning_rate": 8.609431563100228e-06, + "loss": 0.3192, + "step": 9238 + }, + { + "epoch": 0.56, + "learning_rate": 8.607493952587499e-06, + "loss": 0.3184, + "step": 9239 + }, + { + "epoch": 0.56, + "learning_rate": 8.605556395386419e-06, + "loss": 0.322, + "step": 9240 + }, + { + "epoch": 0.56, + "learning_rate": 8.603618891571164e-06, + "loss": 0.3217, + "step": 9241 + }, + { + "epoch": 0.56, + "learning_rate": 8.601681441215916e-06, + "loss": 0.2905, + "step": 9242 + }, + { + "epoch": 0.56, + "learning_rate": 8.599744044394844e-06, + "loss": 0.3034, + "step": 9243 + }, + { + "epoch": 0.56, + "learning_rate": 8.597806701182126e-06, + "loss": 0.281, + "step": 9244 + }, + { + "epoch": 0.56, + "learning_rate": 8.595869411651931e-06, + "loss": 0.315, + "step": 9245 + }, + { + "epoch": 0.56, + "learning_rate": 8.593932175878423e-06, + "loss": 0.3276, + "step": 9246 + }, + { + "epoch": 0.56, + "learning_rate": 8.591994993935776e-06, + "loss": 0.296, + "step": 9247 + }, + { + "epoch": 0.56, + "learning_rate": 8.590057865898148e-06, + "loss": 0.3301, + "step": 9248 + }, + { + "epoch": 0.56, + "learning_rate": 8.588120791839706e-06, + "loss": 0.2928, + "step": 9249 + }, + { + "epoch": 0.56, + "learning_rate": 8.586183771834607e-06, + "loss": 0.2852, + "step": 9250 + }, + { + "epoch": 0.56, + "learning_rate": 8.58424680595701e-06, + "loss": 0.3131, + "step": 9251 + }, + { + "epoch": 0.56, + "learning_rate": 8.582309894281072e-06, + "loss": 0.3145, + "step": 9252 + }, + { + "epoch": 0.56, + "learning_rate": 8.580373036880947e-06, + "loss": 0.3131, + "step": 9253 + }, + { + "epoch": 0.56, + "learning_rate": 8.578436233830785e-06, + "loss": 0.3096, + "step": 9254 + }, + { + "epoch": 0.56, + "learning_rate": 8.576499485204739e-06, + "loss": 0.3087, + "step": 9255 + }, + { + "epoch": 0.56, + "learning_rate": 8.574562791076953e-06, + "loss": 0.3175, + "step": 9256 + }, + { + "epoch": 0.56, + "learning_rate": 8.572626151521581e-06, + "loss": 0.3343, + "step": 9257 + }, + { + "epoch": 0.56, + "learning_rate": 8.570689566612755e-06, + "loss": 0.2955, + "step": 9258 + }, + { + "epoch": 0.56, + "learning_rate": 8.568753036424624e-06, + "loss": 0.3147, + "step": 9259 + }, + { + "epoch": 0.56, + "learning_rate": 8.566816561031322e-06, + "loss": 0.2766, + "step": 9260 + }, + { + "epoch": 0.56, + "learning_rate": 8.564880140506993e-06, + "loss": 0.2812, + "step": 9261 + }, + { + "epoch": 0.56, + "learning_rate": 8.56294377492577e-06, + "loss": 0.3242, + "step": 9262 + }, + { + "epoch": 0.56, + "learning_rate": 8.561007464361782e-06, + "loss": 0.3087, + "step": 9263 + }, + { + "epoch": 0.56, + "learning_rate": 8.559071208889163e-06, + "loss": 0.2831, + "step": 9264 + }, + { + "epoch": 0.56, + "learning_rate": 8.557135008582042e-06, + "loss": 0.2779, + "step": 9265 + }, + { + "epoch": 0.56, + "learning_rate": 8.555198863514548e-06, + "loss": 0.3083, + "step": 9266 + }, + { + "epoch": 0.56, + "learning_rate": 8.553262773760805e-06, + "loss": 0.31, + "step": 9267 + }, + { + "epoch": 0.56, + "learning_rate": 8.551326739394931e-06, + "loss": 0.287, + "step": 9268 + }, + { + "epoch": 0.56, + "learning_rate": 8.549390760491051e-06, + "loss": 0.3271, + "step": 9269 + }, + { + "epoch": 0.56, + "learning_rate": 8.547454837123283e-06, + "loss": 0.3507, + "step": 9270 + }, + { + "epoch": 0.56, + "learning_rate": 8.54551896936574e-06, + "loss": 0.2793, + "step": 9271 + }, + { + "epoch": 0.56, + "learning_rate": 8.54358315729254e-06, + "loss": 0.3069, + "step": 9272 + }, + { + "epoch": 0.56, + "learning_rate": 8.541647400977792e-06, + "loss": 0.2888, + "step": 9273 + }, + { + "epoch": 0.56, + "learning_rate": 8.53971170049561e-06, + "loss": 0.3178, + "step": 9274 + }, + { + "epoch": 0.56, + "learning_rate": 8.537776055920096e-06, + "loss": 0.2877, + "step": 9275 + }, + { + "epoch": 0.56, + "learning_rate": 8.535840467325356e-06, + "loss": 0.2952, + "step": 9276 + }, + { + "epoch": 0.56, + "learning_rate": 8.5339049347855e-06, + "loss": 0.3304, + "step": 9277 + }, + { + "epoch": 0.56, + "learning_rate": 8.531969458374623e-06, + "loss": 0.3118, + "step": 9278 + }, + { + "epoch": 0.56, + "learning_rate": 8.530034038166826e-06, + "loss": 0.2786, + "step": 9279 + }, + { + "epoch": 0.56, + "learning_rate": 8.528098674236207e-06, + "loss": 0.3135, + "step": 9280 + }, + { + "epoch": 0.56, + "learning_rate": 8.526163366656858e-06, + "loss": 0.2927, + "step": 9281 + }, + { + "epoch": 0.56, + "learning_rate": 8.524228115502875e-06, + "loss": 0.3105, + "step": 9282 + }, + { + "epoch": 0.56, + "learning_rate": 8.522292920848344e-06, + "loss": 0.2734, + "step": 9283 + }, + { + "epoch": 0.56, + "learning_rate": 8.52035778276736e-06, + "loss": 0.314, + "step": 9284 + }, + { + "epoch": 0.56, + "learning_rate": 8.518422701334003e-06, + "loss": 0.3278, + "step": 9285 + }, + { + "epoch": 0.56, + "learning_rate": 8.51648767662236e-06, + "loss": 0.321, + "step": 9286 + }, + { + "epoch": 0.56, + "learning_rate": 8.514552708706512e-06, + "loss": 0.3276, + "step": 9287 + }, + { + "epoch": 0.56, + "learning_rate": 8.512617797660539e-06, + "loss": 0.2829, + "step": 9288 + }, + { + "epoch": 0.56, + "learning_rate": 8.510682943558515e-06, + "loss": 0.2877, + "step": 9289 + }, + { + "epoch": 0.56, + "learning_rate": 8.508748146474524e-06, + "loss": 0.3308, + "step": 9290 + }, + { + "epoch": 0.56, + "learning_rate": 8.506813406482633e-06, + "loss": 0.3034, + "step": 9291 + }, + { + "epoch": 0.56, + "learning_rate": 8.50487872365691e-06, + "loss": 0.303, + "step": 9292 + }, + { + "epoch": 0.56, + "learning_rate": 8.502944098071428e-06, + "loss": 0.305, + "step": 9293 + }, + { + "epoch": 0.56, + "learning_rate": 8.501009529800252e-06, + "loss": 0.2958, + "step": 9294 + }, + { + "epoch": 0.56, + "learning_rate": 8.499075018917447e-06, + "loss": 0.2994, + "step": 9295 + }, + { + "epoch": 0.56, + "learning_rate": 8.497140565497078e-06, + "loss": 0.3143, + "step": 9296 + }, + { + "epoch": 0.56, + "learning_rate": 8.495206169613199e-06, + "loss": 0.3015, + "step": 9297 + }, + { + "epoch": 0.56, + "learning_rate": 8.493271831339869e-06, + "loss": 0.3111, + "step": 9298 + }, + { + "epoch": 0.56, + "learning_rate": 8.49133755075115e-06, + "loss": 0.3323, + "step": 9299 + }, + { + "epoch": 0.56, + "learning_rate": 8.489403327921086e-06, + "loss": 0.2891, + "step": 9300 + }, + { + "epoch": 0.56, + "learning_rate": 8.487469162923736e-06, + "loss": 0.3172, + "step": 9301 + }, + { + "epoch": 0.56, + "learning_rate": 8.485535055833145e-06, + "loss": 0.2983, + "step": 9302 + }, + { + "epoch": 0.56, + "learning_rate": 8.483601006723359e-06, + "loss": 0.3251, + "step": 9303 + }, + { + "epoch": 0.56, + "learning_rate": 8.481667015668424e-06, + "loss": 0.3242, + "step": 9304 + }, + { + "epoch": 0.56, + "learning_rate": 8.479733082742384e-06, + "loss": 0.3318, + "step": 9305 + }, + { + "epoch": 0.56, + "learning_rate": 8.477799208019273e-06, + "loss": 0.3022, + "step": 9306 + }, + { + "epoch": 0.56, + "learning_rate": 8.475865391573137e-06, + "loss": 0.2946, + "step": 9307 + }, + { + "epoch": 0.56, + "learning_rate": 8.473931633478003e-06, + "loss": 0.2828, + "step": 9308 + }, + { + "epoch": 0.56, + "learning_rate": 8.471997933807911e-06, + "loss": 0.3115, + "step": 9309 + }, + { + "epoch": 0.56, + "learning_rate": 8.470064292636892e-06, + "loss": 0.2946, + "step": 9310 + }, + { + "epoch": 0.56, + "learning_rate": 8.468130710038968e-06, + "loss": 0.2944, + "step": 9311 + }, + { + "epoch": 0.56, + "learning_rate": 8.466197186088172e-06, + "loss": 0.2863, + "step": 9312 + }, + { + "epoch": 0.56, + "learning_rate": 8.464263720858525e-06, + "loss": 0.3133, + "step": 9313 + }, + { + "epoch": 0.56, + "learning_rate": 8.462330314424053e-06, + "loss": 0.3002, + "step": 9314 + }, + { + "epoch": 0.56, + "learning_rate": 8.460396966858773e-06, + "loss": 0.2919, + "step": 9315 + }, + { + "epoch": 0.56, + "learning_rate": 8.458463678236702e-06, + "loss": 0.2973, + "step": 9316 + }, + { + "epoch": 0.56, + "learning_rate": 8.456530448631856e-06, + "loss": 0.2963, + "step": 9317 + }, + { + "epoch": 0.56, + "learning_rate": 8.454597278118247e-06, + "loss": 0.3086, + "step": 9318 + }, + { + "epoch": 0.56, + "learning_rate": 8.45266416676989e-06, + "loss": 0.3147, + "step": 9319 + }, + { + "epoch": 0.56, + "learning_rate": 8.45073111466079e-06, + "loss": 0.3234, + "step": 9320 + }, + { + "epoch": 0.56, + "learning_rate": 8.448798121864953e-06, + "loss": 0.255, + "step": 9321 + }, + { + "epoch": 0.56, + "learning_rate": 8.446865188456385e-06, + "loss": 0.3027, + "step": 9322 + }, + { + "epoch": 0.56, + "learning_rate": 8.444932314509088e-06, + "loss": 0.3122, + "step": 9323 + }, + { + "epoch": 0.56, + "learning_rate": 8.442999500097057e-06, + "loss": 0.2732, + "step": 9324 + }, + { + "epoch": 0.56, + "learning_rate": 8.441066745294297e-06, + "loss": 0.3105, + "step": 9325 + }, + { + "epoch": 0.56, + "learning_rate": 8.439134050174796e-06, + "loss": 0.3138, + "step": 9326 + }, + { + "epoch": 0.56, + "learning_rate": 8.43720141481255e-06, + "loss": 0.3078, + "step": 9327 + }, + { + "epoch": 0.56, + "learning_rate": 8.435268839281543e-06, + "loss": 0.3019, + "step": 9328 + }, + { + "epoch": 0.56, + "learning_rate": 8.433336323655775e-06, + "loss": 0.3016, + "step": 9329 + }, + { + "epoch": 0.56, + "learning_rate": 8.431403868009219e-06, + "loss": 0.2874, + "step": 9330 + }, + { + "epoch": 0.56, + "learning_rate": 8.429471472415867e-06, + "loss": 0.3119, + "step": 9331 + }, + { + "epoch": 0.56, + "learning_rate": 8.4275391369497e-06, + "loss": 0.3122, + "step": 9332 + }, + { + "epoch": 0.56, + "learning_rate": 8.42560686168469e-06, + "loss": 0.2764, + "step": 9333 + }, + { + "epoch": 0.56, + "learning_rate": 8.423674646694821e-06, + "loss": 0.296, + "step": 9334 + }, + { + "epoch": 0.56, + "learning_rate": 8.421742492054062e-06, + "loss": 0.3298, + "step": 9335 + }, + { + "epoch": 0.56, + "learning_rate": 8.41981039783639e-06, + "loss": 0.298, + "step": 9336 + }, + { + "epoch": 0.56, + "learning_rate": 8.417878364115772e-06, + "loss": 0.2944, + "step": 9337 + }, + { + "epoch": 0.56, + "learning_rate": 8.415946390966172e-06, + "loss": 0.3128, + "step": 9338 + }, + { + "epoch": 0.56, + "learning_rate": 8.414014478461559e-06, + "loss": 0.2913, + "step": 9339 + }, + { + "epoch": 0.56, + "learning_rate": 8.412082626675896e-06, + "loss": 0.2863, + "step": 9340 + }, + { + "epoch": 0.56, + "learning_rate": 8.41015083568314e-06, + "loss": 0.2874, + "step": 9341 + }, + { + "epoch": 0.56, + "learning_rate": 8.408219105557253e-06, + "loss": 0.2939, + "step": 9342 + }, + { + "epoch": 0.56, + "learning_rate": 8.406287436372186e-06, + "loss": 0.2755, + "step": 9343 + }, + { + "epoch": 0.56, + "learning_rate": 8.4043558282019e-06, + "loss": 0.2941, + "step": 9344 + }, + { + "epoch": 0.56, + "learning_rate": 8.402424281120338e-06, + "loss": 0.3012, + "step": 9345 + }, + { + "epoch": 0.56, + "learning_rate": 8.40049279520145e-06, + "loss": 0.2999, + "step": 9346 + }, + { + "epoch": 0.56, + "learning_rate": 8.398561370519187e-06, + "loss": 0.3128, + "step": 9347 + }, + { + "epoch": 0.56, + "learning_rate": 8.396630007147487e-06, + "loss": 0.2617, + "step": 9348 + }, + { + "epoch": 0.56, + "learning_rate": 8.394698705160296e-06, + "loss": 0.3078, + "step": 9349 + }, + { + "epoch": 0.56, + "learning_rate": 8.392767464631555e-06, + "loss": 0.3186, + "step": 9350 + }, + { + "epoch": 0.56, + "learning_rate": 8.390836285635193e-06, + "loss": 0.2812, + "step": 9351 + }, + { + "epoch": 0.56, + "learning_rate": 8.388905168245152e-06, + "loss": 0.2909, + "step": 9352 + }, + { + "epoch": 0.57, + "learning_rate": 8.38697411253536e-06, + "loss": 0.3089, + "step": 9353 + }, + { + "epoch": 0.57, + "learning_rate": 8.385043118579748e-06, + "loss": 0.3119, + "step": 9354 + }, + { + "epoch": 0.57, + "learning_rate": 8.383112186452247e-06, + "loss": 0.3061, + "step": 9355 + }, + { + "epoch": 0.57, + "learning_rate": 8.381181316226776e-06, + "loss": 0.303, + "step": 9356 + }, + { + "epoch": 0.57, + "learning_rate": 8.379250507977264e-06, + "loss": 0.3078, + "step": 9357 + }, + { + "epoch": 0.57, + "learning_rate": 8.377319761777626e-06, + "loss": 0.303, + "step": 9358 + }, + { + "epoch": 0.57, + "learning_rate": 8.375389077701786e-06, + "loss": 0.303, + "step": 9359 + }, + { + "epoch": 0.57, + "learning_rate": 8.37345845582365e-06, + "loss": 0.3064, + "step": 9360 + }, + { + "epoch": 0.57, + "learning_rate": 8.37152789621714e-06, + "loss": 0.3298, + "step": 9361 + }, + { + "epoch": 0.57, + "learning_rate": 8.369597398956164e-06, + "loss": 0.2785, + "step": 9362 + }, + { + "epoch": 0.57, + "learning_rate": 8.367666964114629e-06, + "loss": 0.3041, + "step": 9363 + }, + { + "epoch": 0.57, + "learning_rate": 8.365736591766443e-06, + "loss": 0.2801, + "step": 9364 + }, + { + "epoch": 0.57, + "learning_rate": 8.363806281985509e-06, + "loss": 0.2994, + "step": 9365 + }, + { + "epoch": 0.57, + "learning_rate": 8.361876034845729e-06, + "loss": 0.299, + "step": 9366 + }, + { + "epoch": 0.57, + "learning_rate": 8.359945850421002e-06, + "loss": 0.3041, + "step": 9367 + }, + { + "epoch": 0.57, + "learning_rate": 8.358015728785221e-06, + "loss": 0.2959, + "step": 9368 + }, + { + "epoch": 0.57, + "learning_rate": 8.356085670012286e-06, + "loss": 0.3027, + "step": 9369 + }, + { + "epoch": 0.57, + "learning_rate": 8.354155674176083e-06, + "loss": 0.3368, + "step": 9370 + }, + { + "epoch": 0.57, + "learning_rate": 8.352225741350506e-06, + "loss": 0.2801, + "step": 9371 + }, + { + "epoch": 0.57, + "learning_rate": 8.350295871609439e-06, + "loss": 0.2965, + "step": 9372 + }, + { + "epoch": 0.57, + "learning_rate": 8.348366065026767e-06, + "loss": 0.3051, + "step": 9373 + }, + { + "epoch": 0.57, + "learning_rate": 8.346436321676373e-06, + "loss": 0.2994, + "step": 9374 + }, + { + "epoch": 0.57, + "learning_rate": 8.344506641632132e-06, + "loss": 0.2966, + "step": 9375 + }, + { + "epoch": 0.57, + "learning_rate": 8.34257702496793e-06, + "loss": 0.296, + "step": 9376 + }, + { + "epoch": 0.57, + "learning_rate": 8.340647471757636e-06, + "loss": 0.315, + "step": 9377 + }, + { + "epoch": 0.57, + "learning_rate": 8.338717982075119e-06, + "loss": 0.2806, + "step": 9378 + }, + { + "epoch": 0.57, + "learning_rate": 8.336788555994258e-06, + "loss": 0.3101, + "step": 9379 + }, + { + "epoch": 0.57, + "learning_rate": 8.334859193588914e-06, + "loss": 0.3156, + "step": 9380 + }, + { + "epoch": 0.57, + "learning_rate": 8.332929894932952e-06, + "loss": 0.3147, + "step": 9381 + }, + { + "epoch": 0.57, + "learning_rate": 8.331000660100237e-06, + "loss": 0.3214, + "step": 9382 + }, + { + "epoch": 0.57, + "learning_rate": 8.329071489164626e-06, + "loss": 0.2945, + "step": 9383 + }, + { + "epoch": 0.57, + "learning_rate": 8.327142382199984e-06, + "loss": 0.3198, + "step": 9384 + }, + { + "epoch": 0.57, + "learning_rate": 8.325213339280157e-06, + "loss": 0.2955, + "step": 9385 + }, + { + "epoch": 0.57, + "learning_rate": 8.323284360479003e-06, + "loss": 0.2909, + "step": 9386 + }, + { + "epoch": 0.57, + "learning_rate": 8.321355445870371e-06, + "loss": 0.2879, + "step": 9387 + }, + { + "epoch": 0.57, + "learning_rate": 8.319426595528108e-06, + "loss": 0.2764, + "step": 9388 + }, + { + "epoch": 0.57, + "learning_rate": 8.317497809526062e-06, + "loss": 0.2881, + "step": 9389 + }, + { + "epoch": 0.57, + "learning_rate": 8.315569087938075e-06, + "loss": 0.2976, + "step": 9390 + }, + { + "epoch": 0.57, + "learning_rate": 8.313640430837986e-06, + "loss": 0.287, + "step": 9391 + }, + { + "epoch": 0.57, + "learning_rate": 8.311711838299635e-06, + "loss": 0.3019, + "step": 9392 + }, + { + "epoch": 0.57, + "learning_rate": 8.309783310396858e-06, + "loss": 0.3273, + "step": 9393 + }, + { + "epoch": 0.57, + "learning_rate": 8.307854847203485e-06, + "loss": 0.2994, + "step": 9394 + }, + { + "epoch": 0.57, + "learning_rate": 8.305926448793347e-06, + "loss": 0.3239, + "step": 9395 + }, + { + "epoch": 0.57, + "learning_rate": 8.303998115240274e-06, + "loss": 0.3125, + "step": 9396 + }, + { + "epoch": 0.57, + "learning_rate": 8.302069846618092e-06, + "loss": 0.3094, + "step": 9397 + }, + { + "epoch": 0.57, + "learning_rate": 8.300141643000622e-06, + "loss": 0.2885, + "step": 9398 + }, + { + "epoch": 0.57, + "learning_rate": 8.298213504461687e-06, + "loss": 0.3006, + "step": 9399 + }, + { + "epoch": 0.57, + "learning_rate": 8.296285431075102e-06, + "loss": 0.3192, + "step": 9400 + }, + { + "epoch": 0.57, + "learning_rate": 8.294357422914686e-06, + "loss": 0.2997, + "step": 9401 + }, + { + "epoch": 0.57, + "learning_rate": 8.292429480054252e-06, + "loss": 0.2757, + "step": 9402 + }, + { + "epoch": 0.57, + "learning_rate": 8.290501602567606e-06, + "loss": 0.3195, + "step": 9403 + }, + { + "epoch": 0.57, + "learning_rate": 8.288573790528562e-06, + "loss": 0.2948, + "step": 9404 + }, + { + "epoch": 0.57, + "learning_rate": 8.286646044010923e-06, + "loss": 0.2991, + "step": 9405 + }, + { + "epoch": 0.57, + "learning_rate": 8.28471836308849e-06, + "loss": 0.3076, + "step": 9406 + }, + { + "epoch": 0.57, + "learning_rate": 8.282790747835071e-06, + "loss": 0.2963, + "step": 9407 + }, + { + "epoch": 0.57, + "learning_rate": 8.280863198324455e-06, + "loss": 0.2935, + "step": 9408 + }, + { + "epoch": 0.57, + "learning_rate": 8.278935714630444e-06, + "loss": 0.2733, + "step": 9409 + }, + { + "epoch": 0.57, + "learning_rate": 8.277008296826827e-06, + "loss": 0.3145, + "step": 9410 + }, + { + "epoch": 0.57, + "learning_rate": 8.2750809449874e-06, + "loss": 0.3052, + "step": 9411 + }, + { + "epoch": 0.57, + "learning_rate": 8.273153659185946e-06, + "loss": 0.3242, + "step": 9412 + }, + { + "epoch": 0.57, + "learning_rate": 8.27122643949625e-06, + "loss": 0.298, + "step": 9413 + }, + { + "epoch": 0.57, + "learning_rate": 8.2692992859921e-06, + "loss": 0.266, + "step": 9414 + }, + { + "epoch": 0.57, + "learning_rate": 8.267372198747274e-06, + "loss": 0.2953, + "step": 9415 + }, + { + "epoch": 0.57, + "learning_rate": 8.265445177835547e-06, + "loss": 0.3039, + "step": 9416 + }, + { + "epoch": 0.57, + "learning_rate": 8.263518223330698e-06, + "loss": 0.3145, + "step": 9417 + }, + { + "epoch": 0.57, + "learning_rate": 8.261591335306497e-06, + "loss": 0.2931, + "step": 9418 + }, + { + "epoch": 0.57, + "learning_rate": 8.259664513836718e-06, + "loss": 0.3025, + "step": 9419 + }, + { + "epoch": 0.57, + "learning_rate": 8.257737758995126e-06, + "loss": 0.298, + "step": 9420 + }, + { + "epoch": 0.57, + "learning_rate": 8.255811070855487e-06, + "loss": 0.3026, + "step": 9421 + }, + { + "epoch": 0.57, + "learning_rate": 8.253884449491564e-06, + "loss": 0.3078, + "step": 9422 + }, + { + "epoch": 0.57, + "learning_rate": 8.251957894977117e-06, + "loss": 0.2713, + "step": 9423 + }, + { + "epoch": 0.57, + "learning_rate": 8.250031407385902e-06, + "loss": 0.3039, + "step": 9424 + }, + { + "epoch": 0.57, + "learning_rate": 8.248104986791677e-06, + "loss": 0.2994, + "step": 9425 + }, + { + "epoch": 0.57, + "learning_rate": 8.24617863326819e-06, + "loss": 0.3292, + "step": 9426 + }, + { + "epoch": 0.57, + "learning_rate": 8.244252346889196e-06, + "loss": 0.2891, + "step": 9427 + }, + { + "epoch": 0.57, + "learning_rate": 8.24232612772844e-06, + "loss": 0.3019, + "step": 9428 + }, + { + "epoch": 0.57, + "learning_rate": 8.240399975859666e-06, + "loss": 0.2815, + "step": 9429 + }, + { + "epoch": 0.57, + "learning_rate": 8.238473891356613e-06, + "loss": 0.3181, + "step": 9430 + }, + { + "epoch": 0.57, + "learning_rate": 8.236547874293026e-06, + "loss": 0.3234, + "step": 9431 + }, + { + "epoch": 0.57, + "learning_rate": 8.234621924742641e-06, + "loss": 0.3287, + "step": 9432 + }, + { + "epoch": 0.57, + "learning_rate": 8.23269604277919e-06, + "loss": 0.3037, + "step": 9433 + }, + { + "epoch": 0.57, + "learning_rate": 8.230770228476406e-06, + "loss": 0.2963, + "step": 9434 + }, + { + "epoch": 0.57, + "learning_rate": 8.228844481908017e-06, + "loss": 0.3287, + "step": 9435 + }, + { + "epoch": 0.57, + "learning_rate": 8.226918803147752e-06, + "loss": 0.3118, + "step": 9436 + }, + { + "epoch": 0.57, + "learning_rate": 8.224993192269334e-06, + "loss": 0.3242, + "step": 9437 + }, + { + "epoch": 0.57, + "learning_rate": 8.223067649346483e-06, + "loss": 0.2945, + "step": 9438 + }, + { + "epoch": 0.57, + "learning_rate": 8.221142174452922e-06, + "loss": 0.2965, + "step": 9439 + }, + { + "epoch": 0.57, + "learning_rate": 8.219216767662362e-06, + "loss": 0.3103, + "step": 9440 + }, + { + "epoch": 0.57, + "learning_rate": 8.217291429048519e-06, + "loss": 0.3126, + "step": 9441 + }, + { + "epoch": 0.57, + "learning_rate": 8.215366158685104e-06, + "loss": 0.2917, + "step": 9442 + }, + { + "epoch": 0.57, + "learning_rate": 8.213440956645824e-06, + "loss": 0.3039, + "step": 9443 + }, + { + "epoch": 0.57, + "learning_rate": 8.211515823004388e-06, + "loss": 0.2934, + "step": 9444 + }, + { + "epoch": 0.57, + "learning_rate": 8.209590757834494e-06, + "loss": 0.3167, + "step": 9445 + }, + { + "epoch": 0.57, + "learning_rate": 8.207665761209848e-06, + "loss": 0.2877, + "step": 9446 + }, + { + "epoch": 0.57, + "learning_rate": 8.205740833204146e-06, + "loss": 0.3418, + "step": 9447 + }, + { + "epoch": 0.57, + "learning_rate": 8.203815973891078e-06, + "loss": 0.3256, + "step": 9448 + }, + { + "epoch": 0.57, + "learning_rate": 8.201891183344346e-06, + "loss": 0.3064, + "step": 9449 + }, + { + "epoch": 0.57, + "learning_rate": 8.199966461637633e-06, + "loss": 0.2963, + "step": 9450 + }, + { + "epoch": 0.57, + "learning_rate": 8.198041808844628e-06, + "loss": 0.3041, + "step": 9451 + }, + { + "epoch": 0.57, + "learning_rate": 8.19611722503902e-06, + "loss": 0.2807, + "step": 9452 + }, + { + "epoch": 0.57, + "learning_rate": 8.194192710294482e-06, + "loss": 0.2924, + "step": 9453 + }, + { + "epoch": 0.57, + "learning_rate": 8.192268264684703e-06, + "loss": 0.2857, + "step": 9454 + }, + { + "epoch": 0.57, + "learning_rate": 8.190343888283355e-06, + "loss": 0.2779, + "step": 9455 + }, + { + "epoch": 0.57, + "learning_rate": 8.188419581164111e-06, + "loss": 0.2765, + "step": 9456 + }, + { + "epoch": 0.57, + "learning_rate": 8.186495343400645e-06, + "loss": 0.2796, + "step": 9457 + }, + { + "epoch": 0.57, + "learning_rate": 8.184571175066625e-06, + "loss": 0.2984, + "step": 9458 + }, + { + "epoch": 0.57, + "learning_rate": 8.182647076235719e-06, + "loss": 0.3315, + "step": 9459 + }, + { + "epoch": 0.57, + "learning_rate": 8.180723046981586e-06, + "loss": 0.2723, + "step": 9460 + }, + { + "epoch": 0.57, + "learning_rate": 8.178799087377894e-06, + "loss": 0.3055, + "step": 9461 + }, + { + "epoch": 0.57, + "learning_rate": 8.176875197498293e-06, + "loss": 0.3125, + "step": 9462 + }, + { + "epoch": 0.57, + "learning_rate": 8.174951377416443e-06, + "loss": 0.3025, + "step": 9463 + }, + { + "epoch": 0.57, + "learning_rate": 8.173027627205997e-06, + "loss": 0.279, + "step": 9464 + }, + { + "epoch": 0.57, + "learning_rate": 8.171103946940602e-06, + "loss": 0.3273, + "step": 9465 + }, + { + "epoch": 0.57, + "learning_rate": 8.16918033669391e-06, + "loss": 0.3092, + "step": 9466 + }, + { + "epoch": 0.57, + "learning_rate": 8.167256796539562e-06, + "loss": 0.3145, + "step": 9467 + }, + { + "epoch": 0.57, + "learning_rate": 8.1653333265512e-06, + "loss": 0.3089, + "step": 9468 + }, + { + "epoch": 0.57, + "learning_rate": 8.163409926802468e-06, + "loss": 0.305, + "step": 9469 + }, + { + "epoch": 0.57, + "learning_rate": 8.161486597366997e-06, + "loss": 0.2972, + "step": 9470 + }, + { + "epoch": 0.57, + "learning_rate": 8.159563338318426e-06, + "loss": 0.296, + "step": 9471 + }, + { + "epoch": 0.57, + "learning_rate": 8.157640149730386e-06, + "loss": 0.3011, + "step": 9472 + }, + { + "epoch": 0.57, + "learning_rate": 8.1557170316765e-06, + "loss": 0.2898, + "step": 9473 + }, + { + "epoch": 0.57, + "learning_rate": 8.1537939842304e-06, + "loss": 0.2952, + "step": 9474 + }, + { + "epoch": 0.57, + "learning_rate": 8.151871007465706e-06, + "loss": 0.3105, + "step": 9475 + }, + { + "epoch": 0.57, + "learning_rate": 8.149948101456042e-06, + "loss": 0.3154, + "step": 9476 + }, + { + "epoch": 0.57, + "learning_rate": 8.148025266275023e-06, + "loss": 0.3198, + "step": 9477 + }, + { + "epoch": 0.57, + "learning_rate": 8.146102501996262e-06, + "loss": 0.3228, + "step": 9478 + }, + { + "epoch": 0.57, + "learning_rate": 8.144179808693376e-06, + "loss": 0.2924, + "step": 9479 + }, + { + "epoch": 0.57, + "learning_rate": 8.142257186439972e-06, + "loss": 0.3103, + "step": 9480 + }, + { + "epoch": 0.57, + "learning_rate": 8.140334635309658e-06, + "loss": 0.2886, + "step": 9481 + }, + { + "epoch": 0.57, + "learning_rate": 8.138412155376039e-06, + "loss": 0.3064, + "step": 9482 + }, + { + "epoch": 0.57, + "learning_rate": 8.136489746712714e-06, + "loss": 0.2931, + "step": 9483 + }, + { + "epoch": 0.57, + "learning_rate": 8.134567409393285e-06, + "loss": 0.3196, + "step": 9484 + }, + { + "epoch": 0.57, + "learning_rate": 8.132645143491346e-06, + "loss": 0.2665, + "step": 9485 + }, + { + "epoch": 0.57, + "learning_rate": 8.130722949080489e-06, + "loss": 0.3048, + "step": 9486 + }, + { + "epoch": 0.57, + "learning_rate": 8.128800826234308e-06, + "loss": 0.2917, + "step": 9487 + }, + { + "epoch": 0.57, + "learning_rate": 8.126878775026387e-06, + "loss": 0.3013, + "step": 9488 + }, + { + "epoch": 0.57, + "learning_rate": 8.124956795530316e-06, + "loss": 0.3033, + "step": 9489 + }, + { + "epoch": 0.57, + "learning_rate": 8.123034887819674e-06, + "loss": 0.2663, + "step": 9490 + }, + { + "epoch": 0.57, + "learning_rate": 8.12111305196804e-06, + "loss": 0.285, + "step": 9491 + }, + { + "epoch": 0.57, + "learning_rate": 8.119191288048994e-06, + "loss": 0.3306, + "step": 9492 + }, + { + "epoch": 0.57, + "learning_rate": 8.117269596136104e-06, + "loss": 0.2871, + "step": 9493 + }, + { + "epoch": 0.57, + "learning_rate": 8.11534797630295e-06, + "loss": 0.31, + "step": 9494 + }, + { + "epoch": 0.57, + "learning_rate": 8.113426428623098e-06, + "loss": 0.2907, + "step": 9495 + }, + { + "epoch": 0.57, + "learning_rate": 8.11150495317011e-06, + "loss": 0.3039, + "step": 9496 + }, + { + "epoch": 0.57, + "learning_rate": 8.10958355001755e-06, + "loss": 0.278, + "step": 9497 + }, + { + "epoch": 0.57, + "learning_rate": 8.10766221923898e-06, + "loss": 0.2695, + "step": 9498 + }, + { + "epoch": 0.57, + "learning_rate": 8.105740960907959e-06, + "loss": 0.3089, + "step": 9499 + }, + { + "epoch": 0.57, + "learning_rate": 8.103819775098036e-06, + "loss": 0.3117, + "step": 9500 + }, + { + "epoch": 0.57, + "learning_rate": 8.101898661882769e-06, + "loss": 0.3036, + "step": 9501 + }, + { + "epoch": 0.57, + "learning_rate": 8.099977621335703e-06, + "loss": 0.3158, + "step": 9502 + }, + { + "epoch": 0.57, + "learning_rate": 8.098056653530389e-06, + "loss": 0.3242, + "step": 9503 + }, + { + "epoch": 0.57, + "learning_rate": 8.096135758540367e-06, + "loss": 0.3186, + "step": 9504 + }, + { + "epoch": 0.57, + "learning_rate": 8.094214936439177e-06, + "loss": 0.3064, + "step": 9505 + }, + { + "epoch": 0.57, + "learning_rate": 8.092294187300361e-06, + "loss": 0.2913, + "step": 9506 + }, + { + "epoch": 0.57, + "learning_rate": 8.090373511197454e-06, + "loss": 0.3114, + "step": 9507 + }, + { + "epoch": 0.57, + "learning_rate": 8.088452908203982e-06, + "loss": 0.2896, + "step": 9508 + }, + { + "epoch": 0.57, + "learning_rate": 8.086532378393482e-06, + "loss": 0.2824, + "step": 9509 + }, + { + "epoch": 0.57, + "learning_rate": 8.084611921839477e-06, + "loss": 0.3054, + "step": 9510 + }, + { + "epoch": 0.57, + "learning_rate": 8.082691538615493e-06, + "loss": 0.2914, + "step": 9511 + }, + { + "epoch": 0.57, + "learning_rate": 8.080771228795052e-06, + "loss": 0.2988, + "step": 9512 + }, + { + "epoch": 0.57, + "learning_rate": 8.078850992451668e-06, + "loss": 0.3112, + "step": 9513 + }, + { + "epoch": 0.57, + "learning_rate": 8.076930829658862e-06, + "loss": 0.3025, + "step": 9514 + }, + { + "epoch": 0.57, + "learning_rate": 8.075010740490142e-06, + "loss": 0.2891, + "step": 9515 + }, + { + "epoch": 0.57, + "learning_rate": 8.073090725019023e-06, + "loss": 0.2999, + "step": 9516 + }, + { + "epoch": 0.57, + "learning_rate": 8.071170783319008e-06, + "loss": 0.3184, + "step": 9517 + }, + { + "epoch": 0.58, + "learning_rate": 8.069250915463603e-06, + "loss": 0.3161, + "step": 9518 + }, + { + "epoch": 0.58, + "learning_rate": 8.067331121526312e-06, + "loss": 0.3122, + "step": 9519 + }, + { + "epoch": 0.58, + "learning_rate": 8.06541140158063e-06, + "loss": 0.3133, + "step": 9520 + }, + { + "epoch": 0.58, + "learning_rate": 8.063491755700051e-06, + "loss": 0.2952, + "step": 9521 + }, + { + "epoch": 0.58, + "learning_rate": 8.061572183958074e-06, + "loss": 0.3133, + "step": 9522 + }, + { + "epoch": 0.58, + "learning_rate": 8.059652686428184e-06, + "loss": 0.2909, + "step": 9523 + }, + { + "epoch": 0.58, + "learning_rate": 8.057733263183873e-06, + "loss": 0.3032, + "step": 9524 + }, + { + "epoch": 0.58, + "learning_rate": 8.055813914298624e-06, + "loss": 0.3305, + "step": 9525 + }, + { + "epoch": 0.58, + "learning_rate": 8.053894639845915e-06, + "loss": 0.2969, + "step": 9526 + }, + { + "epoch": 0.58, + "learning_rate": 8.05197543989923e-06, + "loss": 0.3076, + "step": 9527 + }, + { + "epoch": 0.58, + "learning_rate": 8.050056314532038e-06, + "loss": 0.3058, + "step": 9528 + }, + { + "epoch": 0.58, + "learning_rate": 8.048137263817825e-06, + "loss": 0.286, + "step": 9529 + }, + { + "epoch": 0.58, + "learning_rate": 8.046218287830046e-06, + "loss": 0.3284, + "step": 9530 + }, + { + "epoch": 0.58, + "learning_rate": 8.044299386642177e-06, + "loss": 0.3444, + "step": 9531 + }, + { + "epoch": 0.58, + "learning_rate": 8.04238056032768e-06, + "loss": 0.308, + "step": 9532 + }, + { + "epoch": 0.58, + "learning_rate": 8.04046180896002e-06, + "loss": 0.3001, + "step": 9533 + }, + { + "epoch": 0.58, + "learning_rate": 8.038543132612653e-06, + "loss": 0.3312, + "step": 9534 + }, + { + "epoch": 0.58, + "learning_rate": 8.03662453135903e-06, + "loss": 0.299, + "step": 9535 + }, + { + "epoch": 0.58, + "learning_rate": 8.034706005272615e-06, + "loss": 0.2902, + "step": 9536 + }, + { + "epoch": 0.58, + "learning_rate": 8.032787554426849e-06, + "loss": 0.2765, + "step": 9537 + }, + { + "epoch": 0.58, + "learning_rate": 8.030869178895184e-06, + "loss": 0.2711, + "step": 9538 + }, + { + "epoch": 0.58, + "learning_rate": 8.028950878751062e-06, + "loss": 0.2935, + "step": 9539 + }, + { + "epoch": 0.58, + "learning_rate": 8.027032654067926e-06, + "loss": 0.2946, + "step": 9540 + }, + { + "epoch": 0.58, + "learning_rate": 8.025114504919215e-06, + "loss": 0.2879, + "step": 9541 + }, + { + "epoch": 0.58, + "learning_rate": 8.023196431378365e-06, + "loss": 0.3198, + "step": 9542 + }, + { + "epoch": 0.58, + "learning_rate": 8.021278433518803e-06, + "loss": 0.3019, + "step": 9543 + }, + { + "epoch": 0.58, + "learning_rate": 8.019360511413967e-06, + "loss": 0.3094, + "step": 9544 + }, + { + "epoch": 0.58, + "learning_rate": 8.017442665137278e-06, + "loss": 0.3312, + "step": 9545 + }, + { + "epoch": 0.58, + "learning_rate": 8.015524894762167e-06, + "loss": 0.3504, + "step": 9546 + }, + { + "epoch": 0.58, + "learning_rate": 8.013607200362049e-06, + "loss": 0.284, + "step": 9547 + }, + { + "epoch": 0.58, + "learning_rate": 8.011689582010343e-06, + "loss": 0.3234, + "step": 9548 + }, + { + "epoch": 0.58, + "learning_rate": 8.009772039780468e-06, + "loss": 0.3223, + "step": 9549 + }, + { + "epoch": 0.58, + "learning_rate": 8.007854573745831e-06, + "loss": 0.2893, + "step": 9550 + }, + { + "epoch": 0.58, + "learning_rate": 8.005937183979847e-06, + "loss": 0.2722, + "step": 9551 + }, + { + "epoch": 0.58, + "learning_rate": 8.004019870555922e-06, + "loss": 0.3001, + "step": 9552 + }, + { + "epoch": 0.58, + "learning_rate": 8.002102633547456e-06, + "loss": 0.341, + "step": 9553 + }, + { + "epoch": 0.58, + "learning_rate": 8.000185473027852e-06, + "loss": 0.2969, + "step": 9554 + }, + { + "epoch": 0.58, + "learning_rate": 7.99826838907051e-06, + "loss": 0.2782, + "step": 9555 + }, + { + "epoch": 0.58, + "learning_rate": 7.99635138174882e-06, + "loss": 0.2988, + "step": 9556 + }, + { + "epoch": 0.58, + "learning_rate": 7.994434451136177e-06, + "loss": 0.2888, + "step": 9557 + }, + { + "epoch": 0.58, + "learning_rate": 7.99251759730597e-06, + "loss": 0.3284, + "step": 9558 + }, + { + "epoch": 0.58, + "learning_rate": 7.990600820331587e-06, + "loss": 0.2966, + "step": 9559 + }, + { + "epoch": 0.58, + "learning_rate": 7.988684120286408e-06, + "loss": 0.3066, + "step": 9560 + }, + { + "epoch": 0.58, + "learning_rate": 7.986767497243813e-06, + "loss": 0.2871, + "step": 9561 + }, + { + "epoch": 0.58, + "learning_rate": 7.984850951277184e-06, + "loss": 0.2988, + "step": 9562 + }, + { + "epoch": 0.58, + "learning_rate": 7.982934482459895e-06, + "loss": 0.3027, + "step": 9563 + }, + { + "epoch": 0.58, + "learning_rate": 7.981018090865308e-06, + "loss": 0.3211, + "step": 9564 + }, + { + "epoch": 0.58, + "learning_rate": 7.9791017765668e-06, + "loss": 0.3198, + "step": 9565 + }, + { + "epoch": 0.58, + "learning_rate": 7.977185539637734e-06, + "loss": 0.3248, + "step": 9566 + }, + { + "epoch": 0.58, + "learning_rate": 7.975269380151473e-06, + "loss": 0.3012, + "step": 9567 + }, + { + "epoch": 0.58, + "learning_rate": 7.973353298181378e-06, + "loss": 0.2966, + "step": 9568 + }, + { + "epoch": 0.58, + "learning_rate": 7.971437293800804e-06, + "loss": 0.317, + "step": 9569 + }, + { + "epoch": 0.58, + "learning_rate": 7.969521367083101e-06, + "loss": 0.3043, + "step": 9570 + }, + { + "epoch": 0.58, + "learning_rate": 7.967605518101628e-06, + "loss": 0.3158, + "step": 9571 + }, + { + "epoch": 0.58, + "learning_rate": 7.965689746929725e-06, + "loss": 0.2892, + "step": 9572 + }, + { + "epoch": 0.58, + "learning_rate": 7.963774053640742e-06, + "loss": 0.2914, + "step": 9573 + }, + { + "epoch": 0.58, + "learning_rate": 7.961858438308016e-06, + "loss": 0.3131, + "step": 9574 + }, + { + "epoch": 0.58, + "learning_rate": 7.959942901004889e-06, + "loss": 0.2893, + "step": 9575 + }, + { + "epoch": 0.58, + "learning_rate": 7.958027441804697e-06, + "loss": 0.3234, + "step": 9576 + }, + { + "epoch": 0.58, + "learning_rate": 7.956112060780771e-06, + "loss": 0.2958, + "step": 9577 + }, + { + "epoch": 0.58, + "learning_rate": 7.95419675800644e-06, + "loss": 0.2631, + "step": 9578 + }, + { + "epoch": 0.58, + "learning_rate": 7.952281533555035e-06, + "loss": 0.3059, + "step": 9579 + }, + { + "epoch": 0.58, + "learning_rate": 7.950366387499873e-06, + "loss": 0.2941, + "step": 9580 + }, + { + "epoch": 0.58, + "learning_rate": 7.948451319914283e-06, + "loss": 0.2818, + "step": 9581 + }, + { + "epoch": 0.58, + "learning_rate": 7.946536330871575e-06, + "loss": 0.3451, + "step": 9582 + }, + { + "epoch": 0.58, + "learning_rate": 7.944621420445067e-06, + "loss": 0.3043, + "step": 9583 + }, + { + "epoch": 0.58, + "learning_rate": 7.942706588708074e-06, + "loss": 0.2959, + "step": 9584 + }, + { + "epoch": 0.58, + "learning_rate": 7.9407918357339e-06, + "loss": 0.2959, + "step": 9585 + }, + { + "epoch": 0.58, + "learning_rate": 7.938877161595851e-06, + "loss": 0.2772, + "step": 9586 + }, + { + "epoch": 0.58, + "learning_rate": 7.936962566367231e-06, + "loss": 0.2958, + "step": 9587 + }, + { + "epoch": 0.58, + "learning_rate": 7.93504805012134e-06, + "loss": 0.305, + "step": 9588 + }, + { + "epoch": 0.58, + "learning_rate": 7.933133612931476e-06, + "loss": 0.3234, + "step": 9589 + }, + { + "epoch": 0.58, + "learning_rate": 7.93121925487093e-06, + "loss": 0.2988, + "step": 9590 + }, + { + "epoch": 0.58, + "learning_rate": 7.929304976012993e-06, + "loss": 0.3094, + "step": 9591 + }, + { + "epoch": 0.58, + "learning_rate": 7.927390776430954e-06, + "loss": 0.298, + "step": 9592 + }, + { + "epoch": 0.58, + "learning_rate": 7.925476656198095e-06, + "loss": 0.2893, + "step": 9593 + }, + { + "epoch": 0.58, + "learning_rate": 7.923562615387699e-06, + "loss": 0.3132, + "step": 9594 + }, + { + "epoch": 0.58, + "learning_rate": 7.921648654073048e-06, + "loss": 0.2944, + "step": 9595 + }, + { + "epoch": 0.58, + "learning_rate": 7.91973477232741e-06, + "loss": 0.2821, + "step": 9596 + }, + { + "epoch": 0.58, + "learning_rate": 7.917820970224064e-06, + "loss": 0.2863, + "step": 9597 + }, + { + "epoch": 0.58, + "learning_rate": 7.915907247836276e-06, + "loss": 0.3111, + "step": 9598 + }, + { + "epoch": 0.58, + "learning_rate": 7.913993605237313e-06, + "loss": 0.3103, + "step": 9599 + }, + { + "epoch": 0.58, + "learning_rate": 7.912080042500436e-06, + "loss": 0.2934, + "step": 9600 + }, + { + "epoch": 0.58, + "learning_rate": 7.910166559698907e-06, + "loss": 0.3195, + "step": 9601 + }, + { + "epoch": 0.58, + "learning_rate": 7.908253156905982e-06, + "loss": 0.293, + "step": 9602 + }, + { + "epoch": 0.58, + "learning_rate": 7.90633983419492e-06, + "loss": 0.3351, + "step": 9603 + }, + { + "epoch": 0.58, + "learning_rate": 7.904426591638967e-06, + "loss": 0.2988, + "step": 9604 + }, + { + "epoch": 0.58, + "learning_rate": 7.90251342931137e-06, + "loss": 0.2969, + "step": 9605 + }, + { + "epoch": 0.58, + "learning_rate": 7.900600347285379e-06, + "loss": 0.3154, + "step": 9606 + }, + { + "epoch": 0.58, + "learning_rate": 7.898687345634229e-06, + "loss": 0.3097, + "step": 9607 + }, + { + "epoch": 0.58, + "learning_rate": 7.896774424431163e-06, + "loss": 0.3016, + "step": 9608 + }, + { + "epoch": 0.58, + "learning_rate": 7.89486158374942e-06, + "loss": 0.3122, + "step": 9609 + }, + { + "epoch": 0.58, + "learning_rate": 7.892948823662224e-06, + "loss": 0.3146, + "step": 9610 + }, + { + "epoch": 0.58, + "learning_rate": 7.891036144242809e-06, + "loss": 0.3265, + "step": 9611 + }, + { + "epoch": 0.58, + "learning_rate": 7.889123545564404e-06, + "loss": 0.3044, + "step": 9612 + }, + { + "epoch": 0.58, + "learning_rate": 7.887211027700225e-06, + "loss": 0.2806, + "step": 9613 + }, + { + "epoch": 0.58, + "learning_rate": 7.885298590723498e-06, + "loss": 0.3133, + "step": 9614 + }, + { + "epoch": 0.58, + "learning_rate": 7.883386234707437e-06, + "loss": 0.2906, + "step": 9615 + }, + { + "epoch": 0.58, + "learning_rate": 7.881473959725262e-06, + "loss": 0.3061, + "step": 9616 + }, + { + "epoch": 0.58, + "learning_rate": 7.879561765850177e-06, + "loss": 0.2891, + "step": 9617 + }, + { + "epoch": 0.58, + "learning_rate": 7.87764965315539e-06, + "loss": 0.3209, + "step": 9618 + }, + { + "epoch": 0.58, + "learning_rate": 7.875737621714109e-06, + "loss": 0.3071, + "step": 9619 + }, + { + "epoch": 0.58, + "learning_rate": 7.873825671599533e-06, + "loss": 0.3008, + "step": 9620 + }, + { + "epoch": 0.58, + "learning_rate": 7.871913802884862e-06, + "loss": 0.2779, + "step": 9621 + }, + { + "epoch": 0.58, + "learning_rate": 7.870002015643292e-06, + "loss": 0.3114, + "step": 9622 + }, + { + "epoch": 0.58, + "learning_rate": 7.868090309948012e-06, + "loss": 0.2924, + "step": 9623 + }, + { + "epoch": 0.58, + "learning_rate": 7.866178685872216e-06, + "loss": 0.3023, + "step": 9624 + }, + { + "epoch": 0.58, + "learning_rate": 7.864267143489082e-06, + "loss": 0.2955, + "step": 9625 + }, + { + "epoch": 0.58, + "learning_rate": 7.862355682871802e-06, + "loss": 0.2861, + "step": 9626 + }, + { + "epoch": 0.58, + "learning_rate": 7.860444304093551e-06, + "loss": 0.2846, + "step": 9627 + }, + { + "epoch": 0.58, + "learning_rate": 7.858533007227503e-06, + "loss": 0.3237, + "step": 9628 + }, + { + "epoch": 0.58, + "learning_rate": 7.856621792346838e-06, + "loss": 0.2846, + "step": 9629 + }, + { + "epoch": 0.58, + "learning_rate": 7.85471065952472e-06, + "loss": 0.3172, + "step": 9630 + }, + { + "epoch": 0.58, + "learning_rate": 7.852799608834322e-06, + "loss": 0.3036, + "step": 9631 + }, + { + "epoch": 0.58, + "learning_rate": 7.8508886403488e-06, + "loss": 0.3002, + "step": 9632 + }, + { + "epoch": 0.58, + "learning_rate": 7.848977754141323e-06, + "loss": 0.3396, + "step": 9633 + }, + { + "epoch": 0.58, + "learning_rate": 7.847066950285042e-06, + "loss": 0.2965, + "step": 9634 + }, + { + "epoch": 0.58, + "learning_rate": 7.845156228853115e-06, + "loss": 0.3025, + "step": 9635 + }, + { + "epoch": 0.58, + "learning_rate": 7.843245589918694e-06, + "loss": 0.2812, + "step": 9636 + }, + { + "epoch": 0.58, + "learning_rate": 7.841335033554924e-06, + "loss": 0.3139, + "step": 9637 + }, + { + "epoch": 0.58, + "learning_rate": 7.839424559834955e-06, + "loss": 0.2772, + "step": 9638 + }, + { + "epoch": 0.58, + "learning_rate": 7.837514168831926e-06, + "loss": 0.3012, + "step": 9639 + }, + { + "epoch": 0.58, + "learning_rate": 7.835603860618973e-06, + "loss": 0.3092, + "step": 9640 + }, + { + "epoch": 0.58, + "learning_rate": 7.833693635269235e-06, + "loss": 0.2882, + "step": 9641 + }, + { + "epoch": 0.58, + "learning_rate": 7.831783492855845e-06, + "loss": 0.2955, + "step": 9642 + }, + { + "epoch": 0.58, + "learning_rate": 7.829873433451932e-06, + "loss": 0.3125, + "step": 9643 + }, + { + "epoch": 0.58, + "learning_rate": 7.827963457130618e-06, + "loss": 0.2994, + "step": 9644 + }, + { + "epoch": 0.58, + "learning_rate": 7.826053563965029e-06, + "loss": 0.3069, + "step": 9645 + }, + { + "epoch": 0.58, + "learning_rate": 7.824143754028285e-06, + "loss": 0.3025, + "step": 9646 + }, + { + "epoch": 0.58, + "learning_rate": 7.822234027393504e-06, + "loss": 0.2783, + "step": 9647 + }, + { + "epoch": 0.58, + "learning_rate": 7.820324384133793e-06, + "loss": 0.3192, + "step": 9648 + }, + { + "epoch": 0.58, + "learning_rate": 7.818414824322269e-06, + "loss": 0.2875, + "step": 9649 + }, + { + "epoch": 0.58, + "learning_rate": 7.816505348032034e-06, + "loss": 0.292, + "step": 9650 + }, + { + "epoch": 0.58, + "learning_rate": 7.814595955336195e-06, + "loss": 0.3153, + "step": 9651 + }, + { + "epoch": 0.58, + "learning_rate": 7.812686646307852e-06, + "loss": 0.2769, + "step": 9652 + }, + { + "epoch": 0.58, + "learning_rate": 7.8107774210201e-06, + "loss": 0.2886, + "step": 9653 + }, + { + "epoch": 0.58, + "learning_rate": 7.808868279546036e-06, + "loss": 0.3044, + "step": 9654 + }, + { + "epoch": 0.58, + "learning_rate": 7.806959221958747e-06, + "loss": 0.2814, + "step": 9655 + }, + { + "epoch": 0.58, + "learning_rate": 7.805050248331326e-06, + "loss": 0.3292, + "step": 9656 + }, + { + "epoch": 0.58, + "learning_rate": 7.803141358736854e-06, + "loss": 0.297, + "step": 9657 + }, + { + "epoch": 0.58, + "learning_rate": 7.80123255324841e-06, + "loss": 0.2638, + "step": 9658 + }, + { + "epoch": 0.58, + "learning_rate": 7.799323831939079e-06, + "loss": 0.2864, + "step": 9659 + }, + { + "epoch": 0.58, + "learning_rate": 7.797415194881927e-06, + "loss": 0.2797, + "step": 9660 + }, + { + "epoch": 0.58, + "learning_rate": 7.795506642150034e-06, + "loss": 0.2877, + "step": 9661 + }, + { + "epoch": 0.58, + "learning_rate": 7.793598173816464e-06, + "loss": 0.3022, + "step": 9662 + }, + { + "epoch": 0.58, + "learning_rate": 7.79168978995428e-06, + "loss": 0.2926, + "step": 9663 + }, + { + "epoch": 0.58, + "learning_rate": 7.78978149063655e-06, + "loss": 0.3086, + "step": 9664 + }, + { + "epoch": 0.58, + "learning_rate": 7.787873275936329e-06, + "loss": 0.2988, + "step": 9665 + }, + { + "epoch": 0.58, + "learning_rate": 7.785965145926672e-06, + "loss": 0.2984, + "step": 9666 + }, + { + "epoch": 0.58, + "learning_rate": 7.78405710068063e-06, + "loss": 0.315, + "step": 9667 + }, + { + "epoch": 0.58, + "learning_rate": 7.782149140271253e-06, + "loss": 0.3027, + "step": 9668 + }, + { + "epoch": 0.58, + "learning_rate": 7.78024126477159e-06, + "loss": 0.3287, + "step": 9669 + }, + { + "epoch": 0.58, + "learning_rate": 7.778333474254677e-06, + "loss": 0.2905, + "step": 9670 + }, + { + "epoch": 0.58, + "learning_rate": 7.776425768793558e-06, + "loss": 0.2854, + "step": 9671 + }, + { + "epoch": 0.58, + "learning_rate": 7.774518148461268e-06, + "loss": 0.3415, + "step": 9672 + }, + { + "epoch": 0.58, + "learning_rate": 7.77261061333084e-06, + "loss": 0.3065, + "step": 9673 + }, + { + "epoch": 0.58, + "learning_rate": 7.770703163475302e-06, + "loss": 0.3001, + "step": 9674 + }, + { + "epoch": 0.58, + "learning_rate": 7.768795798967679e-06, + "loss": 0.2994, + "step": 9675 + }, + { + "epoch": 0.58, + "learning_rate": 7.766888519880997e-06, + "loss": 0.2893, + "step": 9676 + }, + { + "epoch": 0.58, + "learning_rate": 7.764981326288273e-06, + "loss": 0.3033, + "step": 9677 + }, + { + "epoch": 0.58, + "learning_rate": 7.763074218262526e-06, + "loss": 0.3312, + "step": 9678 + }, + { + "epoch": 0.58, + "learning_rate": 7.761167195876768e-06, + "loss": 0.2905, + "step": 9679 + }, + { + "epoch": 0.58, + "learning_rate": 7.759260259204007e-06, + "loss": 0.3079, + "step": 9680 + }, + { + "epoch": 0.58, + "learning_rate": 7.757353408317251e-06, + "loss": 0.2838, + "step": 9681 + }, + { + "epoch": 0.58, + "learning_rate": 7.755446643289503e-06, + "loss": 0.3059, + "step": 9682 + }, + { + "epoch": 0.58, + "learning_rate": 7.753539964193761e-06, + "loss": 0.2937, + "step": 9683 + }, + { + "epoch": 0.59, + "learning_rate": 7.751633371103026e-06, + "loss": 0.334, + "step": 9684 + }, + { + "epoch": 0.59, + "learning_rate": 7.749726864090286e-06, + "loss": 0.3507, + "step": 9685 + }, + { + "epoch": 0.59, + "learning_rate": 7.747820443228535e-06, + "loss": 0.2659, + "step": 9686 + }, + { + "epoch": 0.59, + "learning_rate": 7.74591410859076e-06, + "loss": 0.2921, + "step": 9687 + }, + { + "epoch": 0.59, + "learning_rate": 7.74400786024994e-06, + "loss": 0.31, + "step": 9688 + }, + { + "epoch": 0.59, + "learning_rate": 7.74210169827906e-06, + "loss": 0.2891, + "step": 9689 + }, + { + "epoch": 0.59, + "learning_rate": 7.740195622751092e-06, + "loss": 0.2984, + "step": 9690 + }, + { + "epoch": 0.59, + "learning_rate": 7.738289633739014e-06, + "loss": 0.3013, + "step": 9691 + }, + { + "epoch": 0.59, + "learning_rate": 7.736383731315795e-06, + "loss": 0.3097, + "step": 9692 + }, + { + "epoch": 0.59, + "learning_rate": 7.7344779155544e-06, + "loss": 0.2914, + "step": 9693 + }, + { + "epoch": 0.59, + "learning_rate": 7.732572186527793e-06, + "loss": 0.2913, + "step": 9694 + }, + { + "epoch": 0.59, + "learning_rate": 7.730666544308936e-06, + "loss": 0.3223, + "step": 9695 + }, + { + "epoch": 0.59, + "learning_rate": 7.728760988970785e-06, + "loss": 0.3043, + "step": 9696 + }, + { + "epoch": 0.59, + "learning_rate": 7.726855520586294e-06, + "loss": 0.281, + "step": 9697 + }, + { + "epoch": 0.59, + "learning_rate": 7.724950139228413e-06, + "loss": 0.2874, + "step": 9698 + }, + { + "epoch": 0.59, + "learning_rate": 7.723044844970089e-06, + "loss": 0.2712, + "step": 9699 + }, + { + "epoch": 0.59, + "learning_rate": 7.721139637884264e-06, + "loss": 0.341, + "step": 9700 + }, + { + "epoch": 0.59, + "learning_rate": 7.719234518043882e-06, + "loss": 0.2867, + "step": 9701 + }, + { + "epoch": 0.59, + "learning_rate": 7.717329485521875e-06, + "loss": 0.3206, + "step": 9702 + }, + { + "epoch": 0.59, + "learning_rate": 7.715424540391177e-06, + "loss": 0.2882, + "step": 9703 + }, + { + "epoch": 0.59, + "learning_rate": 7.713519682724724e-06, + "loss": 0.2984, + "step": 9704 + }, + { + "epoch": 0.59, + "learning_rate": 7.711614912595436e-06, + "loss": 0.3178, + "step": 9705 + }, + { + "epoch": 0.59, + "learning_rate": 7.70971023007624e-06, + "loss": 0.2926, + "step": 9706 + }, + { + "epoch": 0.59, + "learning_rate": 7.707805635240056e-06, + "loss": 0.2653, + "step": 9707 + }, + { + "epoch": 0.59, + "learning_rate": 7.705901128159801e-06, + "loss": 0.3025, + "step": 9708 + }, + { + "epoch": 0.59, + "learning_rate": 7.703996708908388e-06, + "loss": 0.2829, + "step": 9709 + }, + { + "epoch": 0.59, + "learning_rate": 7.702092377558726e-06, + "loss": 0.3055, + "step": 9710 + }, + { + "epoch": 0.59, + "learning_rate": 7.700188134183722e-06, + "loss": 0.2812, + "step": 9711 + }, + { + "epoch": 0.59, + "learning_rate": 7.69828397885628e-06, + "loss": 0.3006, + "step": 9712 + }, + { + "epoch": 0.59, + "learning_rate": 7.696379911649303e-06, + "loss": 0.2919, + "step": 9713 + }, + { + "epoch": 0.59, + "learning_rate": 7.694475932635685e-06, + "loss": 0.3122, + "step": 9714 + }, + { + "epoch": 0.59, + "learning_rate": 7.692572041888313e-06, + "loss": 0.3012, + "step": 9715 + }, + { + "epoch": 0.59, + "learning_rate": 7.690668239480089e-06, + "loss": 0.3064, + "step": 9716 + }, + { + "epoch": 0.59, + "learning_rate": 7.688764525483892e-06, + "loss": 0.2849, + "step": 9717 + }, + { + "epoch": 0.59, + "learning_rate": 7.686860899972603e-06, + "loss": 0.2853, + "step": 9718 + }, + { + "epoch": 0.59, + "learning_rate": 7.684957363019108e-06, + "loss": 0.298, + "step": 9719 + }, + { + "epoch": 0.59, + "learning_rate": 7.68305391469628e-06, + "loss": 0.2944, + "step": 9720 + }, + { + "epoch": 0.59, + "learning_rate": 7.68115055507699e-06, + "loss": 0.3237, + "step": 9721 + }, + { + "epoch": 0.59, + "learning_rate": 7.679247284234112e-06, + "loss": 0.3069, + "step": 9722 + }, + { + "epoch": 0.59, + "learning_rate": 7.67734410224051e-06, + "loss": 0.3055, + "step": 9723 + }, + { + "epoch": 0.59, + "learning_rate": 7.675441009169045e-06, + "loss": 0.2913, + "step": 9724 + }, + { + "epoch": 0.59, + "learning_rate": 7.673538005092578e-06, + "loss": 0.2937, + "step": 9725 + }, + { + "epoch": 0.59, + "learning_rate": 7.671635090083965e-06, + "loss": 0.2824, + "step": 9726 + }, + { + "epoch": 0.59, + "learning_rate": 7.66973226421606e-06, + "loss": 0.2926, + "step": 9727 + }, + { + "epoch": 0.59, + "learning_rate": 7.667829527561707e-06, + "loss": 0.2958, + "step": 9728 + }, + { + "epoch": 0.59, + "learning_rate": 7.665926880193759e-06, + "loss": 0.32, + "step": 9729 + }, + { + "epoch": 0.59, + "learning_rate": 7.66402432218505e-06, + "loss": 0.3094, + "step": 9730 + }, + { + "epoch": 0.59, + "learning_rate": 7.662121853608425e-06, + "loss": 0.2706, + "step": 9731 + }, + { + "epoch": 0.59, + "learning_rate": 7.660219474536719e-06, + "loss": 0.3136, + "step": 9732 + }, + { + "epoch": 0.59, + "learning_rate": 7.658317185042764e-06, + "loss": 0.3298, + "step": 9733 + }, + { + "epoch": 0.59, + "learning_rate": 7.656414985199383e-06, + "loss": 0.3217, + "step": 9734 + }, + { + "epoch": 0.59, + "learning_rate": 7.654512875079406e-06, + "loss": 0.32, + "step": 9735 + }, + { + "epoch": 0.59, + "learning_rate": 7.652610854755656e-06, + "loss": 0.2872, + "step": 9736 + }, + { + "epoch": 0.59, + "learning_rate": 7.650708924300944e-06, + "loss": 0.3318, + "step": 9737 + }, + { + "epoch": 0.59, + "learning_rate": 7.648807083788094e-06, + "loss": 0.2875, + "step": 9738 + }, + { + "epoch": 0.59, + "learning_rate": 7.646905333289912e-06, + "loss": 0.3153, + "step": 9739 + }, + { + "epoch": 0.59, + "learning_rate": 7.645003672879206e-06, + "loss": 0.2963, + "step": 9740 + }, + { + "epoch": 0.59, + "learning_rate": 7.643102102628784e-06, + "loss": 0.3153, + "step": 9741 + }, + { + "epoch": 0.59, + "learning_rate": 7.641200622611441e-06, + "loss": 0.2944, + "step": 9742 + }, + { + "epoch": 0.59, + "learning_rate": 7.639299232899981e-06, + "loss": 0.305, + "step": 9743 + }, + { + "epoch": 0.59, + "learning_rate": 7.637397933567196e-06, + "loss": 0.2832, + "step": 9744 + }, + { + "epoch": 0.59, + "learning_rate": 7.635496724685871e-06, + "loss": 0.3135, + "step": 9745 + }, + { + "epoch": 0.59, + "learning_rate": 7.633595606328803e-06, + "loss": 0.2916, + "step": 9746 + }, + { + "epoch": 0.59, + "learning_rate": 7.631694578568767e-06, + "loss": 0.31, + "step": 9747 + }, + { + "epoch": 0.59, + "learning_rate": 7.62979364147855e-06, + "loss": 0.2577, + "step": 9748 + }, + { + "epoch": 0.59, + "learning_rate": 7.6278927951309255e-06, + "loss": 0.2962, + "step": 9749 + }, + { + "epoch": 0.59, + "learning_rate": 7.625992039598664e-06, + "loss": 0.3206, + "step": 9750 + }, + { + "epoch": 0.59, + "learning_rate": 7.624091374954542e-06, + "loss": 0.2759, + "step": 9751 + }, + { + "epoch": 0.59, + "learning_rate": 7.622190801271319e-06, + "loss": 0.2924, + "step": 9752 + }, + { + "epoch": 0.59, + "learning_rate": 7.620290318621763e-06, + "loss": 0.2994, + "step": 9753 + }, + { + "epoch": 0.59, + "learning_rate": 7.618389927078632e-06, + "loss": 0.3267, + "step": 9754 + }, + { + "epoch": 0.59, + "learning_rate": 7.616489626714679e-06, + "loss": 0.3039, + "step": 9755 + }, + { + "epoch": 0.59, + "learning_rate": 7.6145894176026605e-06, + "loss": 0.2976, + "step": 9756 + }, + { + "epoch": 0.59, + "learning_rate": 7.6126892998153245e-06, + "loss": 0.2893, + "step": 9757 + }, + { + "epoch": 0.59, + "learning_rate": 7.610789273425412e-06, + "loss": 0.298, + "step": 9758 + }, + { + "epoch": 0.59, + "learning_rate": 7.608889338505672e-06, + "loss": 0.3145, + "step": 9759 + }, + { + "epoch": 0.59, + "learning_rate": 7.606989495128836e-06, + "loss": 0.2905, + "step": 9760 + }, + { + "epoch": 0.59, + "learning_rate": 7.605089743367644e-06, + "loss": 0.2884, + "step": 9761 + }, + { + "epoch": 0.59, + "learning_rate": 7.603190083294827e-06, + "loss": 0.3124, + "step": 9762 + }, + { + "epoch": 0.59, + "learning_rate": 7.6012905149831084e-06, + "loss": 0.3471, + "step": 9763 + }, + { + "epoch": 0.59, + "learning_rate": 7.599391038505218e-06, + "loss": 0.2912, + "step": 9764 + }, + { + "epoch": 0.59, + "learning_rate": 7.597491653933871e-06, + "loss": 0.3069, + "step": 9765 + }, + { + "epoch": 0.59, + "learning_rate": 7.5955923613417925e-06, + "loss": 0.2711, + "step": 9766 + }, + { + "epoch": 0.59, + "learning_rate": 7.593693160801692e-06, + "loss": 0.293, + "step": 9767 + }, + { + "epoch": 0.59, + "learning_rate": 7.591794052386279e-06, + "loss": 0.2949, + "step": 9768 + }, + { + "epoch": 0.59, + "learning_rate": 7.589895036168257e-06, + "loss": 0.3198, + "step": 9769 + }, + { + "epoch": 0.59, + "learning_rate": 7.5879961122203365e-06, + "loss": 0.3203, + "step": 9770 + }, + { + "epoch": 0.59, + "learning_rate": 7.586097280615212e-06, + "loss": 0.3022, + "step": 9771 + }, + { + "epoch": 0.59, + "learning_rate": 7.584198541425581e-06, + "loss": 0.2889, + "step": 9772 + }, + { + "epoch": 0.59, + "learning_rate": 7.582299894724139e-06, + "loss": 0.2807, + "step": 9773 + }, + { + "epoch": 0.59, + "learning_rate": 7.580401340583572e-06, + "loss": 0.3172, + "step": 9774 + }, + { + "epoch": 0.59, + "learning_rate": 7.578502879076564e-06, + "loss": 0.2748, + "step": 9775 + }, + { + "epoch": 0.59, + "learning_rate": 7.5766045102758e-06, + "loss": 0.2868, + "step": 9776 + }, + { + "epoch": 0.59, + "learning_rate": 7.574706234253957e-06, + "loss": 0.2909, + "step": 9777 + }, + { + "epoch": 0.59, + "learning_rate": 7.572808051083713e-06, + "loss": 0.2759, + "step": 9778 + }, + { + "epoch": 0.59, + "learning_rate": 7.570909960837735e-06, + "loss": 0.3139, + "step": 9779 + }, + { + "epoch": 0.59, + "learning_rate": 7.569011963588692e-06, + "loss": 0.2744, + "step": 9780 + }, + { + "epoch": 0.59, + "learning_rate": 7.567114059409251e-06, + "loss": 0.272, + "step": 9781 + }, + { + "epoch": 0.59, + "learning_rate": 7.565216248372068e-06, + "loss": 0.3044, + "step": 9782 + }, + { + "epoch": 0.59, + "learning_rate": 7.563318530549805e-06, + "loss": 0.3149, + "step": 9783 + }, + { + "epoch": 0.59, + "learning_rate": 7.561420906015112e-06, + "loss": 0.2893, + "step": 9784 + }, + { + "epoch": 0.59, + "learning_rate": 7.559523374840639e-06, + "loss": 0.2713, + "step": 9785 + }, + { + "epoch": 0.59, + "learning_rate": 7.557625937099036e-06, + "loss": 0.2801, + "step": 9786 + }, + { + "epoch": 0.59, + "learning_rate": 7.55572859286294e-06, + "loss": 0.3292, + "step": 9787 + }, + { + "epoch": 0.59, + "learning_rate": 7.5538313422049955e-06, + "loss": 0.3172, + "step": 9788 + }, + { + "epoch": 0.59, + "learning_rate": 7.551934185197837e-06, + "loss": 0.3128, + "step": 9789 + }, + { + "epoch": 0.59, + "learning_rate": 7.5500371219140926e-06, + "loss": 0.2879, + "step": 9790 + }, + { + "epoch": 0.59, + "learning_rate": 7.548140152426396e-06, + "loss": 0.3198, + "step": 9791 + }, + { + "epoch": 0.59, + "learning_rate": 7.54624327680737e-06, + "loss": 0.2999, + "step": 9792 + }, + { + "epoch": 0.59, + "learning_rate": 7.544346495129634e-06, + "loss": 0.2614, + "step": 9793 + }, + { + "epoch": 0.59, + "learning_rate": 7.54244980746581e-06, + "loss": 0.3069, + "step": 9794 + }, + { + "epoch": 0.59, + "learning_rate": 7.540553213888506e-06, + "loss": 0.2914, + "step": 9795 + }, + { + "epoch": 0.59, + "learning_rate": 7.5386567144703386e-06, + "loss": 0.2997, + "step": 9796 + }, + { + "epoch": 0.59, + "learning_rate": 7.536760309283912e-06, + "loss": 0.3044, + "step": 9797 + }, + { + "epoch": 0.59, + "learning_rate": 7.534863998401828e-06, + "loss": 0.2776, + "step": 9798 + }, + { + "epoch": 0.59, + "learning_rate": 7.53296778189669e-06, + "loss": 0.3005, + "step": 9799 + }, + { + "epoch": 0.59, + "learning_rate": 7.531071659841088e-06, + "loss": 0.2801, + "step": 9800 + }, + { + "epoch": 0.59, + "learning_rate": 7.529175632307626e-06, + "loss": 0.3225, + "step": 9801 + }, + { + "epoch": 0.59, + "learning_rate": 7.527279699368879e-06, + "loss": 0.3147, + "step": 9802 + }, + { + "epoch": 0.59, + "learning_rate": 7.525383861097439e-06, + "loss": 0.2859, + "step": 9803 + }, + { + "epoch": 0.59, + "learning_rate": 7.523488117565886e-06, + "loss": 0.3055, + "step": 9804 + }, + { + "epoch": 0.59, + "learning_rate": 7.5215924688468e-06, + "loss": 0.3008, + "step": 9805 + }, + { + "epoch": 0.59, + "learning_rate": 7.519696915012754e-06, + "loss": 0.2906, + "step": 9806 + }, + { + "epoch": 0.59, + "learning_rate": 7.517801456136318e-06, + "loss": 0.305, + "step": 9807 + }, + { + "epoch": 0.59, + "learning_rate": 7.51590609229006e-06, + "loss": 0.2917, + "step": 9808 + }, + { + "epoch": 0.59, + "learning_rate": 7.514010823546543e-06, + "loss": 0.3086, + "step": 9809 + }, + { + "epoch": 0.59, + "learning_rate": 7.512115649978326e-06, + "loss": 0.2932, + "step": 9810 + }, + { + "epoch": 0.59, + "learning_rate": 7.510220571657968e-06, + "loss": 0.3189, + "step": 9811 + }, + { + "epoch": 0.59, + "learning_rate": 7.508325588658017e-06, + "loss": 0.291, + "step": 9812 + }, + { + "epoch": 0.59, + "learning_rate": 7.506430701051026e-06, + "loss": 0.2972, + "step": 9813 + }, + { + "epoch": 0.59, + "learning_rate": 7.504535908909541e-06, + "loss": 0.2913, + "step": 9814 + }, + { + "epoch": 0.59, + "learning_rate": 7.502641212306096e-06, + "loss": 0.2969, + "step": 9815 + }, + { + "epoch": 0.59, + "learning_rate": 7.500746611313237e-06, + "loss": 0.2877, + "step": 9816 + }, + { + "epoch": 0.59, + "learning_rate": 7.498852106003492e-06, + "loss": 0.2927, + "step": 9817 + }, + { + "epoch": 0.59, + "learning_rate": 7.496957696449399e-06, + "loss": 0.2708, + "step": 9818 + }, + { + "epoch": 0.59, + "learning_rate": 7.495063382723478e-06, + "loss": 0.2985, + "step": 9819 + }, + { + "epoch": 0.59, + "learning_rate": 7.493169164898253e-06, + "loss": 0.2836, + "step": 9820 + }, + { + "epoch": 0.59, + "learning_rate": 7.491275043046246e-06, + "loss": 0.2916, + "step": 9821 + }, + { + "epoch": 0.59, + "learning_rate": 7.489381017239973e-06, + "loss": 0.2775, + "step": 9822 + }, + { + "epoch": 0.59, + "learning_rate": 7.4874870875519455e-06, + "loss": 0.2938, + "step": 9823 + }, + { + "epoch": 0.59, + "learning_rate": 7.485593254054673e-06, + "loss": 0.3147, + "step": 9824 + }, + { + "epoch": 0.59, + "learning_rate": 7.483699516820655e-06, + "loss": 0.2902, + "step": 9825 + }, + { + "epoch": 0.59, + "learning_rate": 7.481805875922399e-06, + "loss": 0.2946, + "step": 9826 + }, + { + "epoch": 0.59, + "learning_rate": 7.479912331432401e-06, + "loss": 0.2832, + "step": 9827 + }, + { + "epoch": 0.59, + "learning_rate": 7.478018883423151e-06, + "loss": 0.2912, + "step": 9828 + }, + { + "epoch": 0.59, + "learning_rate": 7.476125531967143e-06, + "loss": 0.3231, + "step": 9829 + }, + { + "epoch": 0.59, + "learning_rate": 7.4742322771368615e-06, + "loss": 0.3227, + "step": 9830 + }, + { + "epoch": 0.59, + "learning_rate": 7.472339119004792e-06, + "loss": 0.2807, + "step": 9831 + }, + { + "epoch": 0.59, + "learning_rate": 7.47044605764341e-06, + "loss": 0.3075, + "step": 9832 + }, + { + "epoch": 0.59, + "learning_rate": 7.4685530931251906e-06, + "loss": 0.3244, + "step": 9833 + }, + { + "epoch": 0.59, + "learning_rate": 7.466660225522609e-06, + "loss": 0.3083, + "step": 9834 + }, + { + "epoch": 0.59, + "learning_rate": 7.464767454908133e-06, + "loss": 0.2787, + "step": 9835 + }, + { + "epoch": 0.59, + "learning_rate": 7.462874781354223e-06, + "loss": 0.2999, + "step": 9836 + }, + { + "epoch": 0.59, + "learning_rate": 7.460982204933338e-06, + "loss": 0.293, + "step": 9837 + }, + { + "epoch": 0.59, + "learning_rate": 7.459089725717941e-06, + "loss": 0.2761, + "step": 9838 + }, + { + "epoch": 0.59, + "learning_rate": 7.457197343780478e-06, + "loss": 0.2966, + "step": 9839 + }, + { + "epoch": 0.59, + "learning_rate": 7.4553050591934024e-06, + "loss": 0.2878, + "step": 9840 + }, + { + "epoch": 0.59, + "learning_rate": 7.453412872029162e-06, + "loss": 0.3323, + "step": 9841 + }, + { + "epoch": 0.59, + "learning_rate": 7.451520782360193e-06, + "loss": 0.3223, + "step": 9842 + }, + { + "epoch": 0.59, + "learning_rate": 7.449628790258938e-06, + "loss": 0.3426, + "step": 9843 + }, + { + "epoch": 0.59, + "learning_rate": 7.447736895797828e-06, + "loss": 0.3108, + "step": 9844 + }, + { + "epoch": 0.59, + "learning_rate": 7.445845099049295e-06, + "loss": 0.2896, + "step": 9845 + }, + { + "epoch": 0.59, + "learning_rate": 7.443953400085767e-06, + "loss": 0.2833, + "step": 9846 + }, + { + "epoch": 0.59, + "learning_rate": 7.442061798979664e-06, + "loss": 0.3172, + "step": 9847 + }, + { + "epoch": 0.59, + "learning_rate": 7.44017029580341e-06, + "loss": 0.2882, + "step": 9848 + }, + { + "epoch": 0.59, + "learning_rate": 7.4382788906294165e-06, + "loss": 0.3012, + "step": 9849 + }, + { + "epoch": 0.6, + "learning_rate": 7.436387583530096e-06, + "loss": 0.2807, + "step": 9850 + }, + { + "epoch": 0.6, + "learning_rate": 7.434496374577859e-06, + "loss": 0.3019, + "step": 9851 + }, + { + "epoch": 0.6, + "learning_rate": 7.432605263845105e-06, + "loss": 0.287, + "step": 9852 + }, + { + "epoch": 0.6, + "learning_rate": 7.430714251404242e-06, + "loss": 0.3057, + "step": 9853 + }, + { + "epoch": 0.6, + "learning_rate": 7.4288233373276605e-06, + "loss": 0.2716, + "step": 9854 + }, + { + "epoch": 0.6, + "learning_rate": 7.426932521687755e-06, + "loss": 0.2673, + "step": 9855 + }, + { + "epoch": 0.6, + "learning_rate": 7.425041804556916e-06, + "loss": 0.292, + "step": 9856 + }, + { + "epoch": 0.6, + "learning_rate": 7.423151186007527e-06, + "loss": 0.2875, + "step": 9857 + }, + { + "epoch": 0.6, + "learning_rate": 7.421260666111973e-06, + "loss": 0.3136, + "step": 9858 + }, + { + "epoch": 0.6, + "learning_rate": 7.4193702449426305e-06, + "loss": 0.2789, + "step": 9859 + }, + { + "epoch": 0.6, + "learning_rate": 7.41747992257187e-06, + "loss": 0.317, + "step": 9860 + }, + { + "epoch": 0.6, + "learning_rate": 7.415589699072069e-06, + "loss": 0.317, + "step": 9861 + }, + { + "epoch": 0.6, + "learning_rate": 7.413699574515589e-06, + "loss": 0.3128, + "step": 9862 + }, + { + "epoch": 0.6, + "learning_rate": 7.411809548974792e-06, + "loss": 0.2845, + "step": 9863 + }, + { + "epoch": 0.6, + "learning_rate": 7.409919622522042e-06, + "loss": 0.3075, + "step": 9864 + }, + { + "epoch": 0.6, + "learning_rate": 7.408029795229687e-06, + "loss": 0.2835, + "step": 9865 + }, + { + "epoch": 0.6, + "learning_rate": 7.406140067170087e-06, + "loss": 0.2793, + "step": 9866 + }, + { + "epoch": 0.6, + "learning_rate": 7.404250438415584e-06, + "loss": 0.3331, + "step": 9867 + }, + { + "epoch": 0.6, + "learning_rate": 7.402360909038521e-06, + "loss": 0.3039, + "step": 9868 + }, + { + "epoch": 0.6, + "learning_rate": 7.400471479111246e-06, + "loss": 0.3105, + "step": 9869 + }, + { + "epoch": 0.6, + "learning_rate": 7.398582148706086e-06, + "loss": 0.3164, + "step": 9870 + }, + { + "epoch": 0.6, + "learning_rate": 7.396692917895376e-06, + "loss": 0.3016, + "step": 9871 + }, + { + "epoch": 0.6, + "learning_rate": 7.394803786751445e-06, + "loss": 0.279, + "step": 9872 + }, + { + "epoch": 0.6, + "learning_rate": 7.39291475534662e-06, + "loss": 0.3178, + "step": 9873 + }, + { + "epoch": 0.6, + "learning_rate": 7.3910258237532175e-06, + "loss": 0.3284, + "step": 9874 + }, + { + "epoch": 0.6, + "learning_rate": 7.38913699204356e-06, + "loss": 0.3061, + "step": 9875 + }, + { + "epoch": 0.6, + "learning_rate": 7.387248260289957e-06, + "loss": 0.2951, + "step": 9876 + }, + { + "epoch": 0.6, + "learning_rate": 7.385359628564718e-06, + "loss": 0.3078, + "step": 9877 + }, + { + "epoch": 0.6, + "learning_rate": 7.383471096940151e-06, + "loss": 0.3273, + "step": 9878 + }, + { + "epoch": 0.6, + "learning_rate": 7.381582665488558e-06, + "loss": 0.2895, + "step": 9879 + }, + { + "epoch": 0.6, + "learning_rate": 7.379694334282233e-06, + "loss": 0.2932, + "step": 9880 + }, + { + "epoch": 0.6, + "learning_rate": 7.377806103393474e-06, + "loss": 0.2803, + "step": 9881 + }, + { + "epoch": 0.6, + "learning_rate": 7.375917972894569e-06, + "loss": 0.2916, + "step": 9882 + }, + { + "epoch": 0.6, + "learning_rate": 7.374029942857805e-06, + "loss": 0.2903, + "step": 9883 + }, + { + "epoch": 0.6, + "learning_rate": 7.3721420133554675e-06, + "loss": 0.2988, + "step": 9884 + }, + { + "epoch": 0.6, + "learning_rate": 7.370254184459831e-06, + "loss": 0.2969, + "step": 9885 + }, + { + "epoch": 0.6, + "learning_rate": 7.368366456243175e-06, + "loss": 0.2921, + "step": 9886 + }, + { + "epoch": 0.6, + "learning_rate": 7.366478828777765e-06, + "loss": 0.3145, + "step": 9887 + }, + { + "epoch": 0.6, + "learning_rate": 7.364591302135873e-06, + "loss": 0.2812, + "step": 9888 + }, + { + "epoch": 0.6, + "learning_rate": 7.3627038763897604e-06, + "loss": 0.2899, + "step": 9889 + }, + { + "epoch": 0.6, + "learning_rate": 7.360816551611684e-06, + "loss": 0.3005, + "step": 9890 + }, + { + "epoch": 0.6, + "learning_rate": 7.358929327873906e-06, + "loss": 0.3172, + "step": 9891 + }, + { + "epoch": 0.6, + "learning_rate": 7.3570422052486725e-06, + "loss": 0.3023, + "step": 9892 + }, + { + "epoch": 0.6, + "learning_rate": 7.355155183808234e-06, + "loss": 0.2831, + "step": 9893 + }, + { + "epoch": 0.6, + "learning_rate": 7.3532682636248335e-06, + "loss": 0.274, + "step": 9894 + }, + { + "epoch": 0.6, + "learning_rate": 7.351381444770711e-06, + "loss": 0.3153, + "step": 9895 + }, + { + "epoch": 0.6, + "learning_rate": 7.349494727318104e-06, + "loss": 0.2892, + "step": 9896 + }, + { + "epoch": 0.6, + "learning_rate": 7.3476081113392435e-06, + "loss": 0.3178, + "step": 9897 + }, + { + "epoch": 0.6, + "learning_rate": 7.345721596906358e-06, + "loss": 0.2856, + "step": 9898 + }, + { + "epoch": 0.6, + "learning_rate": 7.343835184091674e-06, + "loss": 0.3002, + "step": 9899 + }, + { + "epoch": 0.6, + "learning_rate": 7.341948872967408e-06, + "loss": 0.3259, + "step": 9900 + }, + { + "epoch": 0.6, + "learning_rate": 7.340062663605783e-06, + "loss": 0.305, + "step": 9901 + }, + { + "epoch": 0.6, + "learning_rate": 7.338176556079008e-06, + "loss": 0.2893, + "step": 9902 + }, + { + "epoch": 0.6, + "learning_rate": 7.336290550459295e-06, + "loss": 0.2819, + "step": 9903 + }, + { + "epoch": 0.6, + "learning_rate": 7.3344046468188426e-06, + "loss": 0.302, + "step": 9904 + }, + { + "epoch": 0.6, + "learning_rate": 7.332518845229859e-06, + "loss": 0.3092, + "step": 9905 + }, + { + "epoch": 0.6, + "learning_rate": 7.330633145764536e-06, + "loss": 0.3006, + "step": 9906 + }, + { + "epoch": 0.6, + "learning_rate": 7.3287475484950695e-06, + "loss": 0.2958, + "step": 9907 + }, + { + "epoch": 0.6, + "learning_rate": 7.326862053493652e-06, + "loss": 0.3133, + "step": 9908 + }, + { + "epoch": 0.6, + "learning_rate": 7.324976660832464e-06, + "loss": 0.3131, + "step": 9909 + }, + { + "epoch": 0.6, + "learning_rate": 7.3230913705836905e-06, + "loss": 0.2821, + "step": 9910 + }, + { + "epoch": 0.6, + "learning_rate": 7.321206182819507e-06, + "loss": 0.3203, + "step": 9911 + }, + { + "epoch": 0.6, + "learning_rate": 7.319321097612089e-06, + "loss": 0.2988, + "step": 9912 + }, + { + "epoch": 0.6, + "learning_rate": 7.317436115033606e-06, + "loss": 0.2818, + "step": 9913 + }, + { + "epoch": 0.6, + "learning_rate": 7.315551235156222e-06, + "loss": 0.303, + "step": 9914 + }, + { + "epoch": 0.6, + "learning_rate": 7.313666458052104e-06, + "loss": 0.2914, + "step": 9915 + }, + { + "epoch": 0.6, + "learning_rate": 7.311781783793406e-06, + "loss": 0.2746, + "step": 9916 + }, + { + "epoch": 0.6, + "learning_rate": 7.30989721245228e-06, + "loss": 0.2754, + "step": 9917 + }, + { + "epoch": 0.6, + "learning_rate": 7.308012744100883e-06, + "loss": 0.2995, + "step": 9918 + }, + { + "epoch": 0.6, + "learning_rate": 7.306128378811357e-06, + "loss": 0.2913, + "step": 9919 + }, + { + "epoch": 0.6, + "learning_rate": 7.304244116655842e-06, + "loss": 0.2953, + "step": 9920 + }, + { + "epoch": 0.6, + "learning_rate": 7.3023599577064815e-06, + "loss": 0.3133, + "step": 9921 + }, + { + "epoch": 0.6, + "learning_rate": 7.300475902035407e-06, + "loss": 0.3231, + "step": 9922 + }, + { + "epoch": 0.6, + "learning_rate": 7.2985919497147495e-06, + "loss": 0.3206, + "step": 9923 + }, + { + "epoch": 0.6, + "learning_rate": 7.2967081008166364e-06, + "loss": 0.2921, + "step": 9924 + }, + { + "epoch": 0.6, + "learning_rate": 7.294824355413188e-06, + "loss": 0.3198, + "step": 9925 + }, + { + "epoch": 0.6, + "learning_rate": 7.292940713576526e-06, + "loss": 0.2818, + "step": 9926 + }, + { + "epoch": 0.6, + "learning_rate": 7.2910571753787615e-06, + "loss": 0.2852, + "step": 9927 + }, + { + "epoch": 0.6, + "learning_rate": 7.289173740892009e-06, + "loss": 0.2775, + "step": 9928 + }, + { + "epoch": 0.6, + "learning_rate": 7.287290410188374e-06, + "loss": 0.2932, + "step": 9929 + }, + { + "epoch": 0.6, + "learning_rate": 7.285407183339955e-06, + "loss": 0.2849, + "step": 9930 + }, + { + "epoch": 0.6, + "learning_rate": 7.283524060418857e-06, + "loss": 0.3175, + "step": 9931 + }, + { + "epoch": 0.6, + "learning_rate": 7.281641041497172e-06, + "loss": 0.2967, + "step": 9932 + }, + { + "epoch": 0.6, + "learning_rate": 7.2797581266469884e-06, + "loss": 0.3103, + "step": 9933 + }, + { + "epoch": 0.6, + "learning_rate": 7.277875315940397e-06, + "loss": 0.2999, + "step": 9934 + }, + { + "epoch": 0.6, + "learning_rate": 7.275992609449478e-06, + "loss": 0.2829, + "step": 9935 + }, + { + "epoch": 0.6, + "learning_rate": 7.274110007246312e-06, + "loss": 0.3121, + "step": 9936 + }, + { + "epoch": 0.6, + "learning_rate": 7.272227509402976e-06, + "loss": 0.2969, + "step": 9937 + }, + { + "epoch": 0.6, + "learning_rate": 7.270345115991535e-06, + "loss": 0.2987, + "step": 9938 + }, + { + "epoch": 0.6, + "learning_rate": 7.268462827084058e-06, + "loss": 0.284, + "step": 9939 + }, + { + "epoch": 0.6, + "learning_rate": 7.266580642752609e-06, + "loss": 0.2916, + "step": 9940 + }, + { + "epoch": 0.6, + "learning_rate": 7.264698563069246e-06, + "loss": 0.2811, + "step": 9941 + }, + { + "epoch": 0.6, + "learning_rate": 7.262816588106023e-06, + "loss": 0.2872, + "step": 9942 + }, + { + "epoch": 0.6, + "learning_rate": 7.260934717934993e-06, + "loss": 0.3231, + "step": 9943 + }, + { + "epoch": 0.6, + "learning_rate": 7.2590529526282e-06, + "loss": 0.3192, + "step": 9944 + }, + { + "epoch": 0.6, + "learning_rate": 7.257171292257692e-06, + "loss": 0.3119, + "step": 9945 + }, + { + "epoch": 0.6, + "learning_rate": 7.255289736895503e-06, + "loss": 0.3025, + "step": 9946 + }, + { + "epoch": 0.6, + "learning_rate": 7.253408286613666e-06, + "loss": 0.2845, + "step": 9947 + }, + { + "epoch": 0.6, + "learning_rate": 7.251526941484218e-06, + "loss": 0.3228, + "step": 9948 + }, + { + "epoch": 0.6, + "learning_rate": 7.24964570157918e-06, + "loss": 0.3178, + "step": 9949 + }, + { + "epoch": 0.6, + "learning_rate": 7.2477645669705785e-06, + "loss": 0.2981, + "step": 9950 + }, + { + "epoch": 0.6, + "learning_rate": 7.245883537730432e-06, + "loss": 0.2739, + "step": 9951 + }, + { + "epoch": 0.6, + "learning_rate": 7.244002613930752e-06, + "loss": 0.3158, + "step": 9952 + }, + { + "epoch": 0.6, + "learning_rate": 7.242121795643552e-06, + "loss": 0.2891, + "step": 9953 + }, + { + "epoch": 0.6, + "learning_rate": 7.240241082940838e-06, + "loss": 0.2931, + "step": 9954 + }, + { + "epoch": 0.6, + "learning_rate": 7.238360475894608e-06, + "loss": 0.2937, + "step": 9955 + }, + { + "epoch": 0.6, + "learning_rate": 7.236479974576869e-06, + "loss": 0.3075, + "step": 9956 + }, + { + "epoch": 0.6, + "learning_rate": 7.234599579059607e-06, + "loss": 0.2688, + "step": 9957 + }, + { + "epoch": 0.6, + "learning_rate": 7.232719289414818e-06, + "loss": 0.2998, + "step": 9958 + }, + { + "epoch": 0.6, + "learning_rate": 7.230839105714488e-06, + "loss": 0.2952, + "step": 9959 + }, + { + "epoch": 0.6, + "learning_rate": 7.228959028030593e-06, + "loss": 0.3181, + "step": 9960 + }, + { + "epoch": 0.6, + "learning_rate": 7.22707905643512e-06, + "loss": 0.3111, + "step": 9961 + }, + { + "epoch": 0.6, + "learning_rate": 7.2251991910000364e-06, + "loss": 0.3326, + "step": 9962 + }, + { + "epoch": 0.6, + "learning_rate": 7.223319431797316e-06, + "loss": 0.2864, + "step": 9963 + }, + { + "epoch": 0.6, + "learning_rate": 7.221439778898924e-06, + "loss": 0.3181, + "step": 9964 + }, + { + "epoch": 0.6, + "learning_rate": 7.219560232376821e-06, + "loss": 0.3326, + "step": 9965 + }, + { + "epoch": 0.6, + "learning_rate": 7.217680792302966e-06, + "loss": 0.3107, + "step": 9966 + }, + { + "epoch": 0.6, + "learning_rate": 7.215801458749312e-06, + "loss": 0.2963, + "step": 9967 + }, + { + "epoch": 0.6, + "learning_rate": 7.213922231787809e-06, + "loss": 0.316, + "step": 9968 + }, + { + "epoch": 0.6, + "learning_rate": 7.212043111490403e-06, + "loss": 0.3012, + "step": 9969 + }, + { + "epoch": 0.6, + "learning_rate": 7.210164097929033e-06, + "loss": 0.2987, + "step": 9970 + }, + { + "epoch": 0.6, + "learning_rate": 7.208285191175647e-06, + "loss": 0.3122, + "step": 9971 + }, + { + "epoch": 0.6, + "learning_rate": 7.206406391302163e-06, + "loss": 0.2898, + "step": 9972 + }, + { + "epoch": 0.6, + "learning_rate": 7.204527698380518e-06, + "loss": 0.2744, + "step": 9973 + }, + { + "epoch": 0.6, + "learning_rate": 7.202649112482636e-06, + "loss": 0.2937, + "step": 9974 + }, + { + "epoch": 0.6, + "learning_rate": 7.200770633680439e-06, + "loss": 0.2995, + "step": 9975 + }, + { + "epoch": 0.6, + "learning_rate": 7.198892262045845e-06, + "loss": 0.2786, + "step": 9976 + }, + { + "epoch": 0.6, + "learning_rate": 7.197013997650762e-06, + "loss": 0.3145, + "step": 9977 + }, + { + "epoch": 0.6, + "learning_rate": 7.195135840567104e-06, + "loss": 0.2828, + "step": 9978 + }, + { + "epoch": 0.6, + "learning_rate": 7.193257790866772e-06, + "loss": 0.2864, + "step": 9979 + }, + { + "epoch": 0.6, + "learning_rate": 7.1913798486216706e-06, + "loss": 0.3082, + "step": 9980 + }, + { + "epoch": 0.6, + "learning_rate": 7.1895020139036935e-06, + "loss": 0.2694, + "step": 9981 + }, + { + "epoch": 0.6, + "learning_rate": 7.187624286784731e-06, + "loss": 0.2826, + "step": 9982 + }, + { + "epoch": 0.6, + "learning_rate": 7.185746667336676e-06, + "loss": 0.302, + "step": 9983 + }, + { + "epoch": 0.6, + "learning_rate": 7.183869155631409e-06, + "loss": 0.3023, + "step": 9984 + }, + { + "epoch": 0.6, + "learning_rate": 7.181991751740811e-06, + "loss": 0.2863, + "step": 9985 + }, + { + "epoch": 0.6, + "learning_rate": 7.18011445573676e-06, + "loss": 0.3027, + "step": 9986 + }, + { + "epoch": 0.6, + "learning_rate": 7.178237267691123e-06, + "loss": 0.308, + "step": 9987 + }, + { + "epoch": 0.6, + "learning_rate": 7.1763601876757715e-06, + "loss": 0.3027, + "step": 9988 + }, + { + "epoch": 0.6, + "learning_rate": 7.174483215762569e-06, + "loss": 0.3231, + "step": 9989 + }, + { + "epoch": 0.6, + "learning_rate": 7.172606352023369e-06, + "loss": 0.2849, + "step": 9990 + }, + { + "epoch": 0.6, + "learning_rate": 7.170729596530035e-06, + "loss": 0.2972, + "step": 9991 + }, + { + "epoch": 0.6, + "learning_rate": 7.16885294935441e-06, + "loss": 0.3019, + "step": 9992 + }, + { + "epoch": 0.6, + "learning_rate": 7.166976410568349e-06, + "loss": 0.3108, + "step": 9993 + }, + { + "epoch": 0.6, + "learning_rate": 7.165099980243689e-06, + "loss": 0.2921, + "step": 9994 + }, + { + "epoch": 0.6, + "learning_rate": 7.1632236584522685e-06, + "loss": 0.3055, + "step": 9995 + }, + { + "epoch": 0.6, + "learning_rate": 7.161347445265925e-06, + "loss": 0.3105, + "step": 9996 + }, + { + "epoch": 0.6, + "learning_rate": 7.159471340756486e-06, + "loss": 0.298, + "step": 9997 + }, + { + "epoch": 0.6, + "learning_rate": 7.15759534499578e-06, + "loss": 0.2863, + "step": 9998 + }, + { + "epoch": 0.6, + "learning_rate": 7.155719458055628e-06, + "loss": 0.2952, + "step": 9999 + }, + { + "epoch": 0.6, + "learning_rate": 7.153843680007844e-06, + "loss": 0.3117, + "step": 10000 + }, + { + "epoch": 0.6, + "learning_rate": 7.1519680109242486e-06, + "loss": 0.2886, + "step": 10001 + }, + { + "epoch": 0.6, + "learning_rate": 7.150092450876644e-06, + "loss": 0.3044, + "step": 10002 + }, + { + "epoch": 0.6, + "learning_rate": 7.148216999936841e-06, + "loss": 0.299, + "step": 10003 + }, + { + "epoch": 0.6, + "learning_rate": 7.146341658176638e-06, + "loss": 0.3145, + "step": 10004 + }, + { + "epoch": 0.6, + "learning_rate": 7.144466425667838e-06, + "loss": 0.2779, + "step": 10005 + }, + { + "epoch": 0.6, + "learning_rate": 7.142591302482223e-06, + "loss": 0.292, + "step": 10006 + }, + { + "epoch": 0.6, + "learning_rate": 7.1407162886915845e-06, + "loss": 0.3231, + "step": 10007 + }, + { + "epoch": 0.6, + "learning_rate": 7.138841384367711e-06, + "loss": 0.2698, + "step": 10008 + }, + { + "epoch": 0.6, + "learning_rate": 7.1369665895823795e-06, + "loss": 0.2733, + "step": 10009 + }, + { + "epoch": 0.6, + "learning_rate": 7.135091904407368e-06, + "loss": 0.2863, + "step": 10010 + }, + { + "epoch": 0.6, + "learning_rate": 7.133217328914446e-06, + "loss": 0.2891, + "step": 10011 + }, + { + "epoch": 0.6, + "learning_rate": 7.1313428631753815e-06, + "loss": 0.2713, + "step": 10012 + }, + { + "epoch": 0.6, + "learning_rate": 7.129468507261939e-06, + "loss": 0.3048, + "step": 10013 + }, + { + "epoch": 0.6, + "learning_rate": 7.127594261245877e-06, + "loss": 0.2683, + "step": 10014 + }, + { + "epoch": 0.61, + "learning_rate": 7.125720125198951e-06, + "loss": 0.2879, + "step": 10015 + }, + { + "epoch": 0.61, + "learning_rate": 7.1238460991929125e-06, + "loss": 0.291, + "step": 10016 + }, + { + "epoch": 0.61, + "learning_rate": 7.121972183299503e-06, + "loss": 0.2941, + "step": 10017 + }, + { + "epoch": 0.61, + "learning_rate": 7.120098377590472e-06, + "loss": 0.2921, + "step": 10018 + }, + { + "epoch": 0.61, + "learning_rate": 7.118224682137551e-06, + "loss": 0.3315, + "step": 10019 + }, + { + "epoch": 0.61, + "learning_rate": 7.1163510970124795e-06, + "loss": 0.2992, + "step": 10020 + }, + { + "epoch": 0.61, + "learning_rate": 7.114477622286984e-06, + "loss": 0.3006, + "step": 10021 + }, + { + "epoch": 0.61, + "learning_rate": 7.112604258032788e-06, + "loss": 0.287, + "step": 10022 + }, + { + "epoch": 0.61, + "learning_rate": 7.110731004321617e-06, + "loss": 0.323, + "step": 10023 + }, + { + "epoch": 0.61, + "learning_rate": 7.108857861225187e-06, + "loss": 0.3142, + "step": 10024 + }, + { + "epoch": 0.61, + "learning_rate": 7.106984828815207e-06, + "loss": 0.3013, + "step": 10025 + }, + { + "epoch": 0.61, + "learning_rate": 7.105111907163389e-06, + "loss": 0.2704, + "step": 10026 + }, + { + "epoch": 0.61, + "learning_rate": 7.103239096341434e-06, + "loss": 0.2836, + "step": 10027 + }, + { + "epoch": 0.61, + "learning_rate": 7.101366396421047e-06, + "loss": 0.2812, + "step": 10028 + }, + { + "epoch": 0.61, + "learning_rate": 7.0994938074739205e-06, + "loss": 0.3043, + "step": 10029 + }, + { + "epoch": 0.61, + "learning_rate": 7.097621329571744e-06, + "loss": 0.3041, + "step": 10030 + }, + { + "epoch": 0.61, + "learning_rate": 7.095748962786211e-06, + "loss": 0.2902, + "step": 10031 + }, + { + "epoch": 0.61, + "learning_rate": 7.093876707188997e-06, + "loss": 0.2976, + "step": 10032 + }, + { + "epoch": 0.61, + "learning_rate": 7.092004562851787e-06, + "loss": 0.3052, + "step": 10033 + }, + { + "epoch": 0.61, + "learning_rate": 7.090132529846252e-06, + "loss": 0.3066, + "step": 10034 + }, + { + "epoch": 0.61, + "learning_rate": 7.0882606082440605e-06, + "loss": 0.2907, + "step": 10035 + }, + { + "epoch": 0.61, + "learning_rate": 7.086388798116884e-06, + "loss": 0.3033, + "step": 10036 + }, + { + "epoch": 0.61, + "learning_rate": 7.084517099536378e-06, + "loss": 0.2766, + "step": 10037 + }, + { + "epoch": 0.61, + "learning_rate": 7.082645512574204e-06, + "loss": 0.2792, + "step": 10038 + }, + { + "epoch": 0.61, + "learning_rate": 7.080774037302019e-06, + "loss": 0.3055, + "step": 10039 + }, + { + "epoch": 0.61, + "learning_rate": 7.078902673791462e-06, + "loss": 0.2686, + "step": 10040 + }, + { + "epoch": 0.61, + "learning_rate": 7.0770314221141834e-06, + "loss": 0.2545, + "step": 10041 + }, + { + "epoch": 0.61, + "learning_rate": 7.0751602823418195e-06, + "loss": 0.2868, + "step": 10042 + }, + { + "epoch": 0.61, + "learning_rate": 7.073289254546012e-06, + "loss": 0.3105, + "step": 10043 + }, + { + "epoch": 0.61, + "learning_rate": 7.071418338798388e-06, + "loss": 0.3029, + "step": 10044 + }, + { + "epoch": 0.61, + "learning_rate": 7.069547535170578e-06, + "loss": 0.2772, + "step": 10045 + }, + { + "epoch": 0.61, + "learning_rate": 7.067676843734204e-06, + "loss": 0.315, + "step": 10046 + }, + { + "epoch": 0.61, + "learning_rate": 7.0658062645608815e-06, + "loss": 0.3101, + "step": 10047 + }, + { + "epoch": 0.61, + "learning_rate": 7.063935797722232e-06, + "loss": 0.308, + "step": 10048 + }, + { + "epoch": 0.61, + "learning_rate": 7.0620654432898596e-06, + "loss": 0.3345, + "step": 10049 + }, + { + "epoch": 0.61, + "learning_rate": 7.060195201335374e-06, + "loss": 0.2843, + "step": 10050 + }, + { + "epoch": 0.61, + "learning_rate": 7.058325071930375e-06, + "loss": 0.2867, + "step": 10051 + }, + { + "epoch": 0.61, + "learning_rate": 7.056455055146458e-06, + "loss": 0.3125, + "step": 10052 + }, + { + "epoch": 0.61, + "learning_rate": 7.0545851510552225e-06, + "loss": 0.2792, + "step": 10053 + }, + { + "epoch": 0.61, + "learning_rate": 7.052715359728248e-06, + "loss": 0.335, + "step": 10054 + }, + { + "epoch": 0.61, + "learning_rate": 7.050845681237127e-06, + "loss": 0.2974, + "step": 10055 + }, + { + "epoch": 0.61, + "learning_rate": 7.048976115653436e-06, + "loss": 0.2725, + "step": 10056 + }, + { + "epoch": 0.61, + "learning_rate": 7.047106663048751e-06, + "loss": 0.2867, + "step": 10057 + }, + { + "epoch": 0.61, + "learning_rate": 7.0452373234946425e-06, + "loss": 0.2941, + "step": 10058 + }, + { + "epoch": 0.61, + "learning_rate": 7.0433680970626796e-06, + "loss": 0.329, + "step": 10059 + }, + { + "epoch": 0.61, + "learning_rate": 7.041498983824422e-06, + "loss": 0.3013, + "step": 10060 + }, + { + "epoch": 0.61, + "learning_rate": 7.039629983851433e-06, + "loss": 0.2927, + "step": 10061 + }, + { + "epoch": 0.61, + "learning_rate": 7.037761097215261e-06, + "loss": 0.2808, + "step": 10062 + }, + { + "epoch": 0.61, + "learning_rate": 7.035892323987459e-06, + "loss": 0.2762, + "step": 10063 + }, + { + "epoch": 0.61, + "learning_rate": 7.034023664239573e-06, + "loss": 0.3022, + "step": 10064 + }, + { + "epoch": 0.61, + "learning_rate": 7.032155118043141e-06, + "loss": 0.3133, + "step": 10065 + }, + { + "epoch": 0.61, + "learning_rate": 7.030286685469704e-06, + "loss": 0.2988, + "step": 10066 + }, + { + "epoch": 0.61, + "learning_rate": 7.028418366590789e-06, + "loss": 0.2909, + "step": 10067 + }, + { + "epoch": 0.61, + "learning_rate": 7.02655016147793e-06, + "loss": 0.3348, + "step": 10068 + }, + { + "epoch": 0.61, + "learning_rate": 7.024682070202646e-06, + "loss": 0.2789, + "step": 10069 + }, + { + "epoch": 0.61, + "learning_rate": 7.022814092836457e-06, + "loss": 0.3039, + "step": 10070 + }, + { + "epoch": 0.61, + "learning_rate": 7.02094622945088e-06, + "loss": 0.2662, + "step": 10071 + }, + { + "epoch": 0.61, + "learning_rate": 7.019078480117423e-06, + "loss": 0.3438, + "step": 10072 + }, + { + "epoch": 0.61, + "learning_rate": 7.017210844907598e-06, + "loss": 0.3078, + "step": 10073 + }, + { + "epoch": 0.61, + "learning_rate": 7.015343323892898e-06, + "loss": 0.2815, + "step": 10074 + }, + { + "epoch": 0.61, + "learning_rate": 7.013475917144825e-06, + "loss": 0.3101, + "step": 10075 + }, + { + "epoch": 0.61, + "learning_rate": 7.011608624734871e-06, + "loss": 0.2896, + "step": 10076 + }, + { + "epoch": 0.61, + "learning_rate": 7.009741446734527e-06, + "loss": 0.2715, + "step": 10077 + }, + { + "epoch": 0.61, + "learning_rate": 7.007874383215276e-06, + "loss": 0.3126, + "step": 10078 + }, + { + "epoch": 0.61, + "learning_rate": 7.0060074342485965e-06, + "loss": 0.3185, + "step": 10079 + }, + { + "epoch": 0.61, + "learning_rate": 7.004140599905965e-06, + "loss": 0.2822, + "step": 10080 + }, + { + "epoch": 0.61, + "learning_rate": 7.002273880258854e-06, + "loss": 0.267, + "step": 10081 + }, + { + "epoch": 0.61, + "learning_rate": 7.000407275378729e-06, + "loss": 0.3198, + "step": 10082 + }, + { + "epoch": 0.61, + "learning_rate": 6.998540785337052e-06, + "loss": 0.32, + "step": 10083 + }, + { + "epoch": 0.61, + "learning_rate": 6.9966744102052805e-06, + "loss": 0.3041, + "step": 10084 + }, + { + "epoch": 0.61, + "learning_rate": 6.994808150054871e-06, + "loss": 0.2997, + "step": 10085 + }, + { + "epoch": 0.61, + "learning_rate": 6.992942004957271e-06, + "loss": 0.3128, + "step": 10086 + }, + { + "epoch": 0.61, + "learning_rate": 6.991075974983923e-06, + "loss": 0.3022, + "step": 10087 + }, + { + "epoch": 0.61, + "learning_rate": 6.989210060206271e-06, + "loss": 0.3025, + "step": 10088 + }, + { + "epoch": 0.61, + "learning_rate": 6.987344260695748e-06, + "loss": 0.3195, + "step": 10089 + }, + { + "epoch": 0.61, + "learning_rate": 6.985478576523789e-06, + "loss": 0.3111, + "step": 10090 + }, + { + "epoch": 0.61, + "learning_rate": 6.983613007761818e-06, + "loss": 0.291, + "step": 10091 + }, + { + "epoch": 0.61, + "learning_rate": 6.981747554481258e-06, + "loss": 0.3013, + "step": 10092 + }, + { + "epoch": 0.61, + "learning_rate": 6.979882216753529e-06, + "loss": 0.3424, + "step": 10093 + }, + { + "epoch": 0.61, + "learning_rate": 6.978016994650044e-06, + "loss": 0.2941, + "step": 10094 + }, + { + "epoch": 0.61, + "learning_rate": 6.976151888242211e-06, + "loss": 0.3097, + "step": 10095 + }, + { + "epoch": 0.61, + "learning_rate": 6.974286897601438e-06, + "loss": 0.2741, + "step": 10096 + }, + { + "epoch": 0.61, + "learning_rate": 6.972422022799121e-06, + "loss": 0.3145, + "step": 10097 + }, + { + "epoch": 0.61, + "learning_rate": 6.970557263906661e-06, + "loss": 0.2924, + "step": 10098 + }, + { + "epoch": 0.61, + "learning_rate": 6.96869262099545e-06, + "loss": 0.2888, + "step": 10099 + }, + { + "epoch": 0.61, + "learning_rate": 6.966828094136868e-06, + "loss": 0.3167, + "step": 10100 + }, + { + "epoch": 0.61, + "learning_rate": 6.964963683402306e-06, + "loss": 0.3011, + "step": 10101 + }, + { + "epoch": 0.61, + "learning_rate": 6.963099388863137e-06, + "loss": 0.3158, + "step": 10102 + }, + { + "epoch": 0.61, + "learning_rate": 6.96123521059074e-06, + "loss": 0.3062, + "step": 10103 + }, + { + "epoch": 0.61, + "learning_rate": 6.959371148656481e-06, + "loss": 0.2826, + "step": 10104 + }, + { + "epoch": 0.61, + "learning_rate": 6.957507203131723e-06, + "loss": 0.2875, + "step": 10105 + }, + { + "epoch": 0.61, + "learning_rate": 6.955643374087831e-06, + "loss": 0.2852, + "step": 10106 + }, + { + "epoch": 0.61, + "learning_rate": 6.953779661596163e-06, + "loss": 0.296, + "step": 10107 + }, + { + "epoch": 0.61, + "learning_rate": 6.9519160657280625e-06, + "loss": 0.2824, + "step": 10108 + }, + { + "epoch": 0.61, + "learning_rate": 6.950052586554881e-06, + "loss": 0.2635, + "step": 10109 + }, + { + "epoch": 0.61, + "learning_rate": 6.948189224147961e-06, + "loss": 0.2672, + "step": 10110 + }, + { + "epoch": 0.61, + "learning_rate": 6.946325978578641e-06, + "loss": 0.3089, + "step": 10111 + }, + { + "epoch": 0.61, + "learning_rate": 6.944462849918254e-06, + "loss": 0.2755, + "step": 10112 + }, + { + "epoch": 0.61, + "learning_rate": 6.942599838238133e-06, + "loss": 0.2778, + "step": 10113 + }, + { + "epoch": 0.61, + "learning_rate": 6.940736943609596e-06, + "loss": 0.2995, + "step": 10114 + }, + { + "epoch": 0.61, + "learning_rate": 6.93887416610397e-06, + "loss": 0.3124, + "step": 10115 + }, + { + "epoch": 0.61, + "learning_rate": 6.937011505792568e-06, + "loss": 0.3115, + "step": 10116 + }, + { + "epoch": 0.61, + "learning_rate": 6.935148962746699e-06, + "loss": 0.304, + "step": 10117 + }, + { + "epoch": 0.61, + "learning_rate": 6.933286537037674e-06, + "loss": 0.2852, + "step": 10118 + }, + { + "epoch": 0.61, + "learning_rate": 6.931424228736792e-06, + "loss": 0.2691, + "step": 10119 + }, + { + "epoch": 0.61, + "learning_rate": 6.929562037915354e-06, + "loss": 0.2624, + "step": 10120 + }, + { + "epoch": 0.61, + "learning_rate": 6.927699964644652e-06, + "loss": 0.2727, + "step": 10121 + }, + { + "epoch": 0.61, + "learning_rate": 6.925838008995973e-06, + "loss": 0.2771, + "step": 10122 + }, + { + "epoch": 0.61, + "learning_rate": 6.923976171040605e-06, + "loss": 0.3045, + "step": 10123 + }, + { + "epoch": 0.61, + "learning_rate": 6.922114450849824e-06, + "loss": 0.2924, + "step": 10124 + }, + { + "epoch": 0.61, + "learning_rate": 6.920252848494908e-06, + "loss": 0.3117, + "step": 10125 + }, + { + "epoch": 0.61, + "learning_rate": 6.918391364047129e-06, + "loss": 0.2974, + "step": 10126 + }, + { + "epoch": 0.61, + "learning_rate": 6.9165299975777485e-06, + "loss": 0.2932, + "step": 10127 + }, + { + "epoch": 0.61, + "learning_rate": 6.9146687491580336e-06, + "loss": 0.2708, + "step": 10128 + }, + { + "epoch": 0.61, + "learning_rate": 6.91280761885924e-06, + "loss": 0.3019, + "step": 10129 + }, + { + "epoch": 0.61, + "learning_rate": 6.910946606752617e-06, + "loss": 0.2893, + "step": 10130 + }, + { + "epoch": 0.61, + "learning_rate": 6.909085712909418e-06, + "loss": 0.3044, + "step": 10131 + }, + { + "epoch": 0.61, + "learning_rate": 6.90722493740088e-06, + "loss": 0.2814, + "step": 10132 + }, + { + "epoch": 0.61, + "learning_rate": 6.905364280298252e-06, + "loss": 0.3058, + "step": 10133 + }, + { + "epoch": 0.61, + "learning_rate": 6.903503741672761e-06, + "loss": 0.2988, + "step": 10134 + }, + { + "epoch": 0.61, + "learning_rate": 6.901643321595639e-06, + "loss": 0.2877, + "step": 10135 + }, + { + "epoch": 0.61, + "learning_rate": 6.899783020138113e-06, + "loss": 0.3234, + "step": 10136 + }, + { + "epoch": 0.61, + "learning_rate": 6.897922837371401e-06, + "loss": 0.2748, + "step": 10137 + }, + { + "epoch": 0.61, + "learning_rate": 6.896062773366724e-06, + "loss": 0.3006, + "step": 10138 + }, + { + "epoch": 0.61, + "learning_rate": 6.894202828195291e-06, + "loss": 0.2999, + "step": 10139 + }, + { + "epoch": 0.61, + "learning_rate": 6.892343001928308e-06, + "loss": 0.3142, + "step": 10140 + }, + { + "epoch": 0.61, + "learning_rate": 6.890483294636985e-06, + "loss": 0.2946, + "step": 10141 + }, + { + "epoch": 0.61, + "learning_rate": 6.888623706392511e-06, + "loss": 0.2754, + "step": 10142 + }, + { + "epoch": 0.61, + "learning_rate": 6.886764237266084e-06, + "loss": 0.2988, + "step": 10143 + }, + { + "epoch": 0.61, + "learning_rate": 6.8849048873288914e-06, + "loss": 0.2744, + "step": 10144 + }, + { + "epoch": 0.61, + "learning_rate": 6.883045656652122e-06, + "loss": 0.3217, + "step": 10145 + }, + { + "epoch": 0.61, + "learning_rate": 6.88118654530695e-06, + "loss": 0.2924, + "step": 10146 + }, + { + "epoch": 0.61, + "learning_rate": 6.879327553364557e-06, + "loss": 0.3145, + "step": 10147 + }, + { + "epoch": 0.61, + "learning_rate": 6.877468680896111e-06, + "loss": 0.3122, + "step": 10148 + }, + { + "epoch": 0.61, + "learning_rate": 6.875609927972775e-06, + "loss": 0.3184, + "step": 10149 + }, + { + "epoch": 0.61, + "learning_rate": 6.873751294665718e-06, + "loss": 0.2688, + "step": 10150 + }, + { + "epoch": 0.61, + "learning_rate": 6.871892781046092e-06, + "loss": 0.3117, + "step": 10151 + }, + { + "epoch": 0.61, + "learning_rate": 6.870034387185049e-06, + "loss": 0.3301, + "step": 10152 + }, + { + "epoch": 0.61, + "learning_rate": 6.868176113153741e-06, + "loss": 0.2857, + "step": 10153 + }, + { + "epoch": 0.61, + "learning_rate": 6.866317959023307e-06, + "loss": 0.3096, + "step": 10154 + }, + { + "epoch": 0.61, + "learning_rate": 6.864459924864891e-06, + "loss": 0.3153, + "step": 10155 + }, + { + "epoch": 0.61, + "learning_rate": 6.862602010749624e-06, + "loss": 0.2804, + "step": 10156 + }, + { + "epoch": 0.61, + "learning_rate": 6.8607442167486345e-06, + "loss": 0.3059, + "step": 10157 + }, + { + "epoch": 0.61, + "learning_rate": 6.858886542933051e-06, + "loss": 0.2919, + "step": 10158 + }, + { + "epoch": 0.61, + "learning_rate": 6.857028989373989e-06, + "loss": 0.2835, + "step": 10159 + }, + { + "epoch": 0.61, + "learning_rate": 6.855171556142571e-06, + "loss": 0.2687, + "step": 10160 + }, + { + "epoch": 0.61, + "learning_rate": 6.853314243309903e-06, + "loss": 0.3304, + "step": 10161 + }, + { + "epoch": 0.61, + "learning_rate": 6.851457050947092e-06, + "loss": 0.3044, + "step": 10162 + }, + { + "epoch": 0.61, + "learning_rate": 6.849599979125244e-06, + "loss": 0.3094, + "step": 10163 + }, + { + "epoch": 0.61, + "learning_rate": 6.847743027915451e-06, + "loss": 0.2983, + "step": 10164 + }, + { + "epoch": 0.61, + "learning_rate": 6.84588619738881e-06, + "loss": 0.2942, + "step": 10165 + }, + { + "epoch": 0.61, + "learning_rate": 6.844029487616408e-06, + "loss": 0.2885, + "step": 10166 + }, + { + "epoch": 0.61, + "learning_rate": 6.8421728986693255e-06, + "loss": 0.2949, + "step": 10167 + }, + { + "epoch": 0.61, + "learning_rate": 6.8403164306186475e-06, + "loss": 0.2792, + "step": 10168 + }, + { + "epoch": 0.61, + "learning_rate": 6.838460083535445e-06, + "loss": 0.2787, + "step": 10169 + }, + { + "epoch": 0.61, + "learning_rate": 6.836603857490784e-06, + "loss": 0.2885, + "step": 10170 + }, + { + "epoch": 0.61, + "learning_rate": 6.834747752555736e-06, + "loss": 0.2838, + "step": 10171 + }, + { + "epoch": 0.61, + "learning_rate": 6.832891768801356e-06, + "loss": 0.2939, + "step": 10172 + }, + { + "epoch": 0.61, + "learning_rate": 6.831035906298704e-06, + "loss": 0.2892, + "step": 10173 + }, + { + "epoch": 0.61, + "learning_rate": 6.829180165118829e-06, + "loss": 0.3015, + "step": 10174 + }, + { + "epoch": 0.61, + "learning_rate": 6.827324545332782e-06, + "loss": 0.2886, + "step": 10175 + }, + { + "epoch": 0.61, + "learning_rate": 6.825469047011594e-06, + "loss": 0.3301, + "step": 10176 + }, + { + "epoch": 0.61, + "learning_rate": 6.823613670226312e-06, + "loss": 0.2803, + "step": 10177 + }, + { + "epoch": 0.61, + "learning_rate": 6.821758415047965e-06, + "loss": 0.3192, + "step": 10178 + }, + { + "epoch": 0.61, + "learning_rate": 6.819903281547579e-06, + "loss": 0.3022, + "step": 10179 + }, + { + "epoch": 0.61, + "learning_rate": 6.81804826979618e-06, + "loss": 0.2856, + "step": 10180 + }, + { + "epoch": 0.62, + "learning_rate": 6.816193379864786e-06, + "loss": 0.2854, + "step": 10181 + }, + { + "epoch": 0.62, + "learning_rate": 6.814338611824412e-06, + "loss": 0.2852, + "step": 10182 + }, + { + "epoch": 0.62, + "learning_rate": 6.812483965746064e-06, + "loss": 0.3089, + "step": 10183 + }, + { + "epoch": 0.62, + "learning_rate": 6.810629441700748e-06, + "loss": 0.2974, + "step": 10184 + }, + { + "epoch": 0.62, + "learning_rate": 6.8087750397594655e-06, + "loss": 0.2928, + "step": 10185 + }, + { + "epoch": 0.62, + "learning_rate": 6.806920759993212e-06, + "loss": 0.2972, + "step": 10186 + }, + { + "epoch": 0.62, + "learning_rate": 6.805066602472973e-06, + "loss": 0.3379, + "step": 10187 + }, + { + "epoch": 0.62, + "learning_rate": 6.80321256726974e-06, + "loss": 0.2981, + "step": 10188 + }, + { + "epoch": 0.62, + "learning_rate": 6.801358654454491e-06, + "loss": 0.2746, + "step": 10189 + }, + { + "epoch": 0.62, + "learning_rate": 6.799504864098203e-06, + "loss": 0.3295, + "step": 10190 + }, + { + "epoch": 0.62, + "learning_rate": 6.797651196271851e-06, + "loss": 0.3158, + "step": 10191 + }, + { + "epoch": 0.62, + "learning_rate": 6.795797651046394e-06, + "loss": 0.3189, + "step": 10192 + }, + { + "epoch": 0.62, + "learning_rate": 6.793944228492804e-06, + "loss": 0.3022, + "step": 10193 + }, + { + "epoch": 0.62, + "learning_rate": 6.792090928682032e-06, + "loss": 0.317, + "step": 10194 + }, + { + "epoch": 0.62, + "learning_rate": 6.7902377516850346e-06, + "loss": 0.2969, + "step": 10195 + }, + { + "epoch": 0.62, + "learning_rate": 6.7883846975727586e-06, + "loss": 0.2907, + "step": 10196 + }, + { + "epoch": 0.62, + "learning_rate": 6.786531766416147e-06, + "loss": 0.3044, + "step": 10197 + }, + { + "epoch": 0.62, + "learning_rate": 6.78467895828614e-06, + "loss": 0.3225, + "step": 10198 + }, + { + "epoch": 0.62, + "learning_rate": 6.782826273253671e-06, + "loss": 0.2905, + "step": 10199 + }, + { + "epoch": 0.62, + "learning_rate": 6.78097371138967e-06, + "loss": 0.2561, + "step": 10200 + }, + { + "epoch": 0.62, + "learning_rate": 6.779121272765062e-06, + "loss": 0.2941, + "step": 10201 + }, + { + "epoch": 0.62, + "learning_rate": 6.777268957450764e-06, + "loss": 0.2786, + "step": 10202 + }, + { + "epoch": 0.62, + "learning_rate": 6.775416765517697e-06, + "loss": 0.2891, + "step": 10203 + }, + { + "epoch": 0.62, + "learning_rate": 6.773564697036767e-06, + "loss": 0.2902, + "step": 10204 + }, + { + "epoch": 0.62, + "learning_rate": 6.7717127520788795e-06, + "loss": 0.2775, + "step": 10205 + }, + { + "epoch": 0.62, + "learning_rate": 6.7698609307149396e-06, + "loss": 0.291, + "step": 10206 + }, + { + "epoch": 0.62, + "learning_rate": 6.76800923301584e-06, + "loss": 0.3045, + "step": 10207 + }, + { + "epoch": 0.62, + "learning_rate": 6.766157659052473e-06, + "loss": 0.3092, + "step": 10208 + }, + { + "epoch": 0.62, + "learning_rate": 6.764306208895731e-06, + "loss": 0.2819, + "step": 10209 + }, + { + "epoch": 0.62, + "learning_rate": 6.762454882616488e-06, + "loss": 0.2773, + "step": 10210 + }, + { + "epoch": 0.62, + "learning_rate": 6.760603680285624e-06, + "loss": 0.2985, + "step": 10211 + }, + { + "epoch": 0.62, + "learning_rate": 6.758752601974013e-06, + "loss": 0.2786, + "step": 10212 + }, + { + "epoch": 0.62, + "learning_rate": 6.756901647752524e-06, + "loss": 0.3008, + "step": 10213 + }, + { + "epoch": 0.62, + "learning_rate": 6.755050817692015e-06, + "loss": 0.2976, + "step": 10214 + }, + { + "epoch": 0.62, + "learning_rate": 6.753200111863352e-06, + "loss": 0.2886, + "step": 10215 + }, + { + "epoch": 0.62, + "learning_rate": 6.751349530337381e-06, + "loss": 0.2861, + "step": 10216 + }, + { + "epoch": 0.62, + "learning_rate": 6.749499073184957e-06, + "loss": 0.2972, + "step": 10217 + }, + { + "epoch": 0.62, + "learning_rate": 6.747648740476923e-06, + "loss": 0.3139, + "step": 10218 + }, + { + "epoch": 0.62, + "learning_rate": 6.745798532284114e-06, + "loss": 0.2884, + "step": 10219 + }, + { + "epoch": 0.62, + "learning_rate": 6.743948448677371e-06, + "loss": 0.2919, + "step": 10220 + }, + { + "epoch": 0.62, + "learning_rate": 6.74209848972752e-06, + "loss": 0.3292, + "step": 10221 + }, + { + "epoch": 0.62, + "learning_rate": 6.740248655505386e-06, + "loss": 0.2709, + "step": 10222 + }, + { + "epoch": 0.62, + "learning_rate": 6.738398946081791e-06, + "loss": 0.3019, + "step": 10223 + }, + { + "epoch": 0.62, + "learning_rate": 6.736549361527548e-06, + "loss": 0.286, + "step": 10224 + }, + { + "epoch": 0.62, + "learning_rate": 6.7346999019134704e-06, + "loss": 0.2958, + "step": 10225 + }, + { + "epoch": 0.62, + "learning_rate": 6.732850567310365e-06, + "loss": 0.3044, + "step": 10226 + }, + { + "epoch": 0.62, + "learning_rate": 6.731001357789029e-06, + "loss": 0.2949, + "step": 10227 + }, + { + "epoch": 0.62, + "learning_rate": 6.729152273420261e-06, + "loss": 0.303, + "step": 10228 + }, + { + "epoch": 0.62, + "learning_rate": 6.727303314274852e-06, + "loss": 0.3054, + "step": 10229 + }, + { + "epoch": 0.62, + "learning_rate": 6.725454480423592e-06, + "loss": 0.2951, + "step": 10230 + }, + { + "epoch": 0.62, + "learning_rate": 6.723605771937259e-06, + "loss": 0.285, + "step": 10231 + }, + { + "epoch": 0.62, + "learning_rate": 6.721757188886631e-06, + "loss": 0.3205, + "step": 10232 + }, + { + "epoch": 0.62, + "learning_rate": 6.719908731342482e-06, + "loss": 0.3143, + "step": 10233 + }, + { + "epoch": 0.62, + "learning_rate": 6.718060399375576e-06, + "loss": 0.2924, + "step": 10234 + }, + { + "epoch": 0.62, + "learning_rate": 6.7162121930566824e-06, + "loss": 0.2882, + "step": 10235 + }, + { + "epoch": 0.62, + "learning_rate": 6.714364112456554e-06, + "loss": 0.3016, + "step": 10236 + }, + { + "epoch": 0.62, + "learning_rate": 6.712516157645943e-06, + "loss": 0.2821, + "step": 10237 + }, + { + "epoch": 0.62, + "learning_rate": 6.710668328695603e-06, + "loss": 0.278, + "step": 10238 + }, + { + "epoch": 0.62, + "learning_rate": 6.7088206256762734e-06, + "loss": 0.2881, + "step": 10239 + }, + { + "epoch": 0.62, + "learning_rate": 6.7069730486586934e-06, + "loss": 0.2899, + "step": 10240 + }, + { + "epoch": 0.62, + "learning_rate": 6.705125597713597e-06, + "loss": 0.2853, + "step": 10241 + }, + { + "epoch": 0.62, + "learning_rate": 6.7032782729117156e-06, + "loss": 0.3334, + "step": 10242 + }, + { + "epoch": 0.62, + "learning_rate": 6.701431074323774e-06, + "loss": 0.2998, + "step": 10243 + }, + { + "epoch": 0.62, + "learning_rate": 6.699584002020485e-06, + "loss": 0.2969, + "step": 10244 + }, + { + "epoch": 0.62, + "learning_rate": 6.6977370560725696e-06, + "loss": 0.3006, + "step": 10245 + }, + { + "epoch": 0.62, + "learning_rate": 6.695890236550733e-06, + "loss": 0.2755, + "step": 10246 + }, + { + "epoch": 0.62, + "learning_rate": 6.694043543525684e-06, + "loss": 0.3061, + "step": 10247 + }, + { + "epoch": 0.62, + "learning_rate": 6.692196977068121e-06, + "loss": 0.3318, + "step": 10248 + }, + { + "epoch": 0.62, + "learning_rate": 6.690350537248738e-06, + "loss": 0.3009, + "step": 10249 + }, + { + "epoch": 0.62, + "learning_rate": 6.688504224138228e-06, + "loss": 0.26, + "step": 10250 + }, + { + "epoch": 0.62, + "learning_rate": 6.686658037807273e-06, + "loss": 0.3055, + "step": 10251 + }, + { + "epoch": 0.62, + "learning_rate": 6.684811978326558e-06, + "loss": 0.3281, + "step": 10252 + }, + { + "epoch": 0.62, + "learning_rate": 6.682966045766758e-06, + "loss": 0.2856, + "step": 10253 + }, + { + "epoch": 0.62, + "learning_rate": 6.681120240198539e-06, + "loss": 0.334, + "step": 10254 + }, + { + "epoch": 0.62, + "learning_rate": 6.679274561692572e-06, + "loss": 0.2966, + "step": 10255 + }, + { + "epoch": 0.62, + "learning_rate": 6.677429010319518e-06, + "loss": 0.2666, + "step": 10256 + }, + { + "epoch": 0.62, + "learning_rate": 6.67558358615003e-06, + "loss": 0.2694, + "step": 10257 + }, + { + "epoch": 0.62, + "learning_rate": 6.673738289254765e-06, + "loss": 0.3125, + "step": 10258 + }, + { + "epoch": 0.62, + "learning_rate": 6.671893119704363e-06, + "loss": 0.2952, + "step": 10259 + }, + { + "epoch": 0.62, + "learning_rate": 6.670048077569471e-06, + "loss": 0.2627, + "step": 10260 + }, + { + "epoch": 0.62, + "learning_rate": 6.668203162920723e-06, + "loss": 0.3129, + "step": 10261 + }, + { + "epoch": 0.62, + "learning_rate": 6.6663583758287494e-06, + "loss": 0.2871, + "step": 10262 + }, + { + "epoch": 0.62, + "learning_rate": 6.664513716364184e-06, + "loss": 0.3068, + "step": 10263 + }, + { + "epoch": 0.62, + "learning_rate": 6.662669184597641e-06, + "loss": 0.3071, + "step": 10264 + }, + { + "epoch": 0.62, + "learning_rate": 6.660824780599744e-06, + "loss": 0.2979, + "step": 10265 + }, + { + "epoch": 0.62, + "learning_rate": 6.658980504441104e-06, + "loss": 0.2921, + "step": 10266 + }, + { + "epoch": 0.62, + "learning_rate": 6.657136356192323e-06, + "loss": 0.284, + "step": 10267 + }, + { + "epoch": 0.62, + "learning_rate": 6.655292335924011e-06, + "loss": 0.2935, + "step": 10268 + }, + { + "epoch": 0.62, + "learning_rate": 6.65344844370676e-06, + "loss": 0.3163, + "step": 10269 + }, + { + "epoch": 0.62, + "learning_rate": 6.651604679611169e-06, + "loss": 0.2902, + "step": 10270 + }, + { + "epoch": 0.62, + "learning_rate": 6.649761043707821e-06, + "loss": 0.308, + "step": 10271 + }, + { + "epoch": 0.62, + "learning_rate": 6.647917536067299e-06, + "loss": 0.2712, + "step": 10272 + }, + { + "epoch": 0.62, + "learning_rate": 6.646074156760185e-06, + "loss": 0.2786, + "step": 10273 + }, + { + "epoch": 0.62, + "learning_rate": 6.6442309058570496e-06, + "loss": 0.2838, + "step": 10274 + }, + { + "epoch": 0.62, + "learning_rate": 6.642387783428459e-06, + "loss": 0.3036, + "step": 10275 + }, + { + "epoch": 0.62, + "learning_rate": 6.64054478954498e-06, + "loss": 0.2914, + "step": 10276 + }, + { + "epoch": 0.62, + "learning_rate": 6.638701924277173e-06, + "loss": 0.304, + "step": 10277 + }, + { + "epoch": 0.62, + "learning_rate": 6.636859187695585e-06, + "loss": 0.2939, + "step": 10278 + }, + { + "epoch": 0.62, + "learning_rate": 6.635016579870767e-06, + "loss": 0.3274, + "step": 10279 + }, + { + "epoch": 0.62, + "learning_rate": 6.6331741008732655e-06, + "loss": 0.2861, + "step": 10280 + }, + { + "epoch": 0.62, + "learning_rate": 6.631331750773614e-06, + "loss": 0.2772, + "step": 10281 + }, + { + "epoch": 0.62, + "learning_rate": 6.629489529642352e-06, + "loss": 0.2765, + "step": 10282 + }, + { + "epoch": 0.62, + "learning_rate": 6.627647437550006e-06, + "loss": 0.2868, + "step": 10283 + }, + { + "epoch": 0.62, + "learning_rate": 6.625805474567098e-06, + "loss": 0.3037, + "step": 10284 + }, + { + "epoch": 0.62, + "learning_rate": 6.623963640764151e-06, + "loss": 0.3019, + "step": 10285 + }, + { + "epoch": 0.62, + "learning_rate": 6.622121936211674e-06, + "loss": 0.2914, + "step": 10286 + }, + { + "epoch": 0.62, + "learning_rate": 6.620280360980182e-06, + "loss": 0.2994, + "step": 10287 + }, + { + "epoch": 0.62, + "learning_rate": 6.618438915140175e-06, + "loss": 0.2935, + "step": 10288 + }, + { + "epoch": 0.62, + "learning_rate": 6.616597598762151e-06, + "loss": 0.2821, + "step": 10289 + }, + { + "epoch": 0.62, + "learning_rate": 6.61475641191661e-06, + "loss": 0.2991, + "step": 10290 + }, + { + "epoch": 0.62, + "learning_rate": 6.612915354674035e-06, + "loss": 0.2778, + "step": 10291 + }, + { + "epoch": 0.62, + "learning_rate": 6.611074427104913e-06, + "loss": 0.334, + "step": 10292 + }, + { + "epoch": 0.62, + "learning_rate": 6.609233629279723e-06, + "loss": 0.3036, + "step": 10293 + }, + { + "epoch": 0.62, + "learning_rate": 6.6073929612689396e-06, + "loss": 0.3281, + "step": 10294 + }, + { + "epoch": 0.62, + "learning_rate": 6.605552423143033e-06, + "loss": 0.3039, + "step": 10295 + }, + { + "epoch": 0.62, + "learning_rate": 6.603712014972467e-06, + "loss": 0.2559, + "step": 10296 + }, + { + "epoch": 0.62, + "learning_rate": 6.6018717368277e-06, + "loss": 0.3058, + "step": 10297 + }, + { + "epoch": 0.62, + "learning_rate": 6.6000315887791875e-06, + "loss": 0.3186, + "step": 10298 + }, + { + "epoch": 0.62, + "learning_rate": 6.5981915708973775e-06, + "loss": 0.3083, + "step": 10299 + }, + { + "epoch": 0.62, + "learning_rate": 6.596351683252718e-06, + "loss": 0.2688, + "step": 10300 + }, + { + "epoch": 0.62, + "learning_rate": 6.594511925915645e-06, + "loss": 0.284, + "step": 10301 + }, + { + "epoch": 0.62, + "learning_rate": 6.592672298956593e-06, + "loss": 0.2921, + "step": 10302 + }, + { + "epoch": 0.62, + "learning_rate": 6.590832802445994e-06, + "loss": 0.3287, + "step": 10303 + }, + { + "epoch": 0.62, + "learning_rate": 6.588993436454272e-06, + "loss": 0.3019, + "step": 10304 + }, + { + "epoch": 0.62, + "learning_rate": 6.587154201051846e-06, + "loss": 0.2785, + "step": 10305 + }, + { + "epoch": 0.62, + "learning_rate": 6.58531509630913e-06, + "loss": 0.3242, + "step": 10306 + }, + { + "epoch": 0.62, + "learning_rate": 6.583476122296533e-06, + "loss": 0.3075, + "step": 10307 + }, + { + "epoch": 0.62, + "learning_rate": 6.581637279084461e-06, + "loss": 0.2762, + "step": 10308 + }, + { + "epoch": 0.62, + "learning_rate": 6.579798566743314e-06, + "loss": 0.2704, + "step": 10309 + }, + { + "epoch": 0.62, + "learning_rate": 6.577959985343482e-06, + "loss": 0.315, + "step": 10310 + }, + { + "epoch": 0.62, + "learning_rate": 6.576121534955364e-06, + "loss": 0.3047, + "step": 10311 + }, + { + "epoch": 0.62, + "learning_rate": 6.574283215649335e-06, + "loss": 0.2934, + "step": 10312 + }, + { + "epoch": 0.62, + "learning_rate": 6.572445027495779e-06, + "loss": 0.3125, + "step": 10313 + }, + { + "epoch": 0.62, + "learning_rate": 6.5706069705650674e-06, + "loss": 0.3331, + "step": 10314 + }, + { + "epoch": 0.62, + "learning_rate": 6.5687690449275735e-06, + "loss": 0.3075, + "step": 10315 + }, + { + "epoch": 0.62, + "learning_rate": 6.566931250653657e-06, + "loss": 0.2344, + "step": 10316 + }, + { + "epoch": 0.62, + "learning_rate": 6.565093587813683e-06, + "loss": 0.3103, + "step": 10317 + }, + { + "epoch": 0.62, + "learning_rate": 6.563256056478002e-06, + "loss": 0.2843, + "step": 10318 + }, + { + "epoch": 0.62, + "learning_rate": 6.561418656716964e-06, + "loss": 0.2868, + "step": 10319 + }, + { + "epoch": 0.62, + "learning_rate": 6.5595813886009145e-06, + "loss": 0.2653, + "step": 10320 + }, + { + "epoch": 0.62, + "learning_rate": 6.5577442522001885e-06, + "loss": 0.2914, + "step": 10321 + }, + { + "epoch": 0.62, + "learning_rate": 6.555907247585127e-06, + "loss": 0.2952, + "step": 10322 + }, + { + "epoch": 0.62, + "learning_rate": 6.5540703748260544e-06, + "loss": 0.279, + "step": 10323 + }, + { + "epoch": 0.62, + "learning_rate": 6.552233633993294e-06, + "loss": 0.2861, + "step": 10324 + }, + { + "epoch": 0.62, + "learning_rate": 6.550397025157169e-06, + "loss": 0.2723, + "step": 10325 + }, + { + "epoch": 0.62, + "learning_rate": 6.54856054838799e-06, + "loss": 0.2999, + "step": 10326 + }, + { + "epoch": 0.62, + "learning_rate": 6.5467242037560686e-06, + "loss": 0.2821, + "step": 10327 + }, + { + "epoch": 0.62, + "learning_rate": 6.5448879913317065e-06, + "loss": 0.2789, + "step": 10328 + }, + { + "epoch": 0.62, + "learning_rate": 6.543051911185201e-06, + "loss": 0.3203, + "step": 10329 + }, + { + "epoch": 0.62, + "learning_rate": 6.541215963386851e-06, + "loss": 0.2902, + "step": 10330 + }, + { + "epoch": 0.62, + "learning_rate": 6.539380148006942e-06, + "loss": 0.2733, + "step": 10331 + }, + { + "epoch": 0.62, + "learning_rate": 6.537544465115755e-06, + "loss": 0.28, + "step": 10332 + }, + { + "epoch": 0.62, + "learning_rate": 6.535708914783574e-06, + "loss": 0.2693, + "step": 10333 + }, + { + "epoch": 0.62, + "learning_rate": 6.533873497080667e-06, + "loss": 0.3041, + "step": 10334 + }, + { + "epoch": 0.62, + "learning_rate": 6.5320382120773074e-06, + "loss": 0.298, + "step": 10335 + }, + { + "epoch": 0.62, + "learning_rate": 6.530203059843756e-06, + "loss": 0.3015, + "step": 10336 + }, + { + "epoch": 0.62, + "learning_rate": 6.528368040450269e-06, + "loss": 0.285, + "step": 10337 + }, + { + "epoch": 0.62, + "learning_rate": 6.526533153967105e-06, + "loss": 0.2693, + "step": 10338 + }, + { + "epoch": 0.62, + "learning_rate": 6.524698400464505e-06, + "loss": 0.2882, + "step": 10339 + }, + { + "epoch": 0.62, + "learning_rate": 6.522863780012719e-06, + "loss": 0.3315, + "step": 10340 + }, + { + "epoch": 0.62, + "learning_rate": 6.521029292681982e-06, + "loss": 0.2985, + "step": 10341 + }, + { + "epoch": 0.62, + "learning_rate": 6.519194938542523e-06, + "loss": 0.3002, + "step": 10342 + }, + { + "epoch": 0.62, + "learning_rate": 6.517360717664576e-06, + "loss": 0.2748, + "step": 10343 + }, + { + "epoch": 0.62, + "learning_rate": 6.515526630118361e-06, + "loss": 0.3073, + "step": 10344 + }, + { + "epoch": 0.62, + "learning_rate": 6.513692675974096e-06, + "loss": 0.3025, + "step": 10345 + }, + { + "epoch": 0.63, + "learning_rate": 6.511858855301989e-06, + "loss": 0.3133, + "step": 10346 + }, + { + "epoch": 0.63, + "learning_rate": 6.510025168172251e-06, + "loss": 0.2952, + "step": 10347 + }, + { + "epoch": 0.63, + "learning_rate": 6.508191614655086e-06, + "loss": 0.2852, + "step": 10348 + }, + { + "epoch": 0.63, + "learning_rate": 6.506358194820686e-06, + "loss": 0.293, + "step": 10349 + }, + { + "epoch": 0.63, + "learning_rate": 6.504524908739247e-06, + "loss": 0.2677, + "step": 10350 + }, + { + "epoch": 0.63, + "learning_rate": 6.502691756480954e-06, + "loss": 0.2478, + "step": 10351 + }, + { + "epoch": 0.63, + "learning_rate": 6.500858738115989e-06, + "loss": 0.3005, + "step": 10352 + }, + { + "epoch": 0.63, + "learning_rate": 6.499025853714531e-06, + "loss": 0.305, + "step": 10353 + }, + { + "epoch": 0.63, + "learning_rate": 6.497193103346745e-06, + "loss": 0.3069, + "step": 10354 + }, + { + "epoch": 0.63, + "learning_rate": 6.495360487082804e-06, + "loss": 0.3105, + "step": 10355 + }, + { + "epoch": 0.63, + "learning_rate": 6.493528004992864e-06, + "loss": 0.2688, + "step": 10356 + }, + { + "epoch": 0.63, + "learning_rate": 6.491695657147085e-06, + "loss": 0.2854, + "step": 10357 + }, + { + "epoch": 0.63, + "learning_rate": 6.489863443615618e-06, + "loss": 0.2764, + "step": 10358 + }, + { + "epoch": 0.63, + "learning_rate": 6.488031364468604e-06, + "loss": 0.3256, + "step": 10359 + }, + { + "epoch": 0.63, + "learning_rate": 6.486199419776189e-06, + "loss": 0.3022, + "step": 10360 + }, + { + "epoch": 0.63, + "learning_rate": 6.484367609608503e-06, + "loss": 0.29, + "step": 10361 + }, + { + "epoch": 0.63, + "learning_rate": 6.482535934035681e-06, + "loss": 0.3052, + "step": 10362 + }, + { + "epoch": 0.63, + "learning_rate": 6.4807043931278476e-06, + "loss": 0.2992, + "step": 10363 + }, + { + "epoch": 0.63, + "learning_rate": 6.478872986955118e-06, + "loss": 0.2857, + "step": 10364 + }, + { + "epoch": 0.63, + "learning_rate": 6.477041715587613e-06, + "loss": 0.2769, + "step": 10365 + }, + { + "epoch": 0.63, + "learning_rate": 6.47521057909544e-06, + "loss": 0.2921, + "step": 10366 + }, + { + "epoch": 0.63, + "learning_rate": 6.473379577548702e-06, + "loss": 0.3119, + "step": 10367 + }, + { + "epoch": 0.63, + "learning_rate": 6.471548711017499e-06, + "loss": 0.2747, + "step": 10368 + }, + { + "epoch": 0.63, + "learning_rate": 6.469717979571925e-06, + "loss": 0.2987, + "step": 10369 + }, + { + "epoch": 0.63, + "learning_rate": 6.467887383282072e-06, + "loss": 0.2691, + "step": 10370 + }, + { + "epoch": 0.63, + "learning_rate": 6.466056922218021e-06, + "loss": 0.2829, + "step": 10371 + }, + { + "epoch": 0.63, + "learning_rate": 6.464226596449849e-06, + "loss": 0.3181, + "step": 10372 + }, + { + "epoch": 0.63, + "learning_rate": 6.462396406047634e-06, + "loss": 0.2945, + "step": 10373 + }, + { + "epoch": 0.63, + "learning_rate": 6.460566351081439e-06, + "loss": 0.2972, + "step": 10374 + }, + { + "epoch": 0.63, + "learning_rate": 6.458736431621332e-06, + "loss": 0.3136, + "step": 10375 + }, + { + "epoch": 0.63, + "learning_rate": 6.4569066477373685e-06, + "loss": 0.3047, + "step": 10376 + }, + { + "epoch": 0.63, + "learning_rate": 6.4550769994996e-06, + "loss": 0.3039, + "step": 10377 + }, + { + "epoch": 0.63, + "learning_rate": 6.453247486978076e-06, + "loss": 0.2838, + "step": 10378 + }, + { + "epoch": 0.63, + "learning_rate": 6.451418110242842e-06, + "loss": 0.3267, + "step": 10379 + }, + { + "epoch": 0.63, + "learning_rate": 6.449588869363927e-06, + "loss": 0.287, + "step": 10380 + }, + { + "epoch": 0.63, + "learning_rate": 6.447759764411366e-06, + "loss": 0.3186, + "step": 10381 + }, + { + "epoch": 0.63, + "learning_rate": 6.44593079545519e-06, + "loss": 0.2999, + "step": 10382 + }, + { + "epoch": 0.63, + "learning_rate": 6.4441019625654165e-06, + "loss": 0.2914, + "step": 10383 + }, + { + "epoch": 0.63, + "learning_rate": 6.44227326581206e-06, + "loss": 0.2898, + "step": 10384 + }, + { + "epoch": 0.63, + "learning_rate": 6.440444705265137e-06, + "loss": 0.2928, + "step": 10385 + }, + { + "epoch": 0.63, + "learning_rate": 6.438616280994648e-06, + "loss": 0.2919, + "step": 10386 + }, + { + "epoch": 0.63, + "learning_rate": 6.436787993070599e-06, + "loss": 0.2903, + "step": 10387 + }, + { + "epoch": 0.63, + "learning_rate": 6.434959841562982e-06, + "loss": 0.3186, + "step": 10388 + }, + { + "epoch": 0.63, + "learning_rate": 6.4331318265417855e-06, + "loss": 0.2633, + "step": 10389 + }, + { + "epoch": 0.63, + "learning_rate": 6.431303948076999e-06, + "loss": 0.2941, + "step": 10390 + }, + { + "epoch": 0.63, + "learning_rate": 6.4294762062385985e-06, + "loss": 0.2987, + "step": 10391 + }, + { + "epoch": 0.63, + "learning_rate": 6.4276486010965614e-06, + "loss": 0.2941, + "step": 10392 + }, + { + "epoch": 0.63, + "learning_rate": 6.425821132720855e-06, + "loss": 0.3026, + "step": 10393 + }, + { + "epoch": 0.63, + "learning_rate": 6.423993801181443e-06, + "loss": 0.2857, + "step": 10394 + }, + { + "epoch": 0.63, + "learning_rate": 6.422166606548288e-06, + "loss": 0.2783, + "step": 10395 + }, + { + "epoch": 0.63, + "learning_rate": 6.420339548891337e-06, + "loss": 0.2916, + "step": 10396 + }, + { + "epoch": 0.63, + "learning_rate": 6.418512628280544e-06, + "loss": 0.2651, + "step": 10397 + }, + { + "epoch": 0.63, + "learning_rate": 6.416685844785851e-06, + "loss": 0.2879, + "step": 10398 + }, + { + "epoch": 0.63, + "learning_rate": 6.414859198477193e-06, + "loss": 0.3013, + "step": 10399 + }, + { + "epoch": 0.63, + "learning_rate": 6.4130326894245055e-06, + "loss": 0.2832, + "step": 10400 + }, + { + "epoch": 0.63, + "learning_rate": 6.411206317697716e-06, + "loss": 0.2828, + "step": 10401 + }, + { + "epoch": 0.63, + "learning_rate": 6.409380083366743e-06, + "loss": 0.2812, + "step": 10402 + }, + { + "epoch": 0.63, + "learning_rate": 6.407553986501507e-06, + "loss": 0.3107, + "step": 10403 + }, + { + "epoch": 0.63, + "learning_rate": 6.405728027171916e-06, + "loss": 0.2991, + "step": 10404 + }, + { + "epoch": 0.63, + "learning_rate": 6.4039022054478814e-06, + "loss": 0.2927, + "step": 10405 + }, + { + "epoch": 0.63, + "learning_rate": 6.402076521399301e-06, + "loss": 0.2624, + "step": 10406 + }, + { + "epoch": 0.63, + "learning_rate": 6.40025097509607e-06, + "loss": 0.3009, + "step": 10407 + }, + { + "epoch": 0.63, + "learning_rate": 6.39842556660808e-06, + "loss": 0.287, + "step": 10408 + }, + { + "epoch": 0.63, + "learning_rate": 6.396600296005213e-06, + "loss": 0.304, + "step": 10409 + }, + { + "epoch": 0.63, + "learning_rate": 6.3947751633573555e-06, + "loss": 0.2859, + "step": 10410 + }, + { + "epoch": 0.63, + "learning_rate": 6.392950168734378e-06, + "loss": 0.3214, + "step": 10411 + }, + { + "epoch": 0.63, + "learning_rate": 6.3911253122061475e-06, + "loss": 0.3108, + "step": 10412 + }, + { + "epoch": 0.63, + "learning_rate": 6.389300593842536e-06, + "loss": 0.2818, + "step": 10413 + }, + { + "epoch": 0.63, + "learning_rate": 6.387476013713394e-06, + "loss": 0.2706, + "step": 10414 + }, + { + "epoch": 0.63, + "learning_rate": 6.385651571888578e-06, + "loss": 0.2902, + "step": 10415 + }, + { + "epoch": 0.63, + "learning_rate": 6.383827268437935e-06, + "loss": 0.3454, + "step": 10416 + }, + { + "epoch": 0.63, + "learning_rate": 6.382003103431309e-06, + "loss": 0.2879, + "step": 10417 + }, + { + "epoch": 0.63, + "learning_rate": 6.38017907693854e-06, + "loss": 0.3192, + "step": 10418 + }, + { + "epoch": 0.63, + "learning_rate": 6.378355189029454e-06, + "loss": 0.2826, + "step": 10419 + }, + { + "epoch": 0.63, + "learning_rate": 6.376531439773884e-06, + "loss": 0.2985, + "step": 10420 + }, + { + "epoch": 0.63, + "learning_rate": 6.374707829241649e-06, + "loss": 0.2552, + "step": 10421 + }, + { + "epoch": 0.63, + "learning_rate": 6.372884357502567e-06, + "loss": 0.3096, + "step": 10422 + }, + { + "epoch": 0.63, + "learning_rate": 6.371061024626446e-06, + "loss": 0.2985, + "step": 10423 + }, + { + "epoch": 0.63, + "learning_rate": 6.369237830683094e-06, + "loss": 0.2651, + "step": 10424 + }, + { + "epoch": 0.63, + "learning_rate": 6.367414775742312e-06, + "loss": 0.3376, + "step": 10425 + }, + { + "epoch": 0.63, + "learning_rate": 6.365591859873892e-06, + "loss": 0.2854, + "step": 10426 + }, + { + "epoch": 0.63, + "learning_rate": 6.363769083147628e-06, + "loss": 0.2634, + "step": 10427 + }, + { + "epoch": 0.63, + "learning_rate": 6.361946445633303e-06, + "loss": 0.285, + "step": 10428 + }, + { + "epoch": 0.63, + "learning_rate": 6.3601239474006915e-06, + "loss": 0.2803, + "step": 10429 + }, + { + "epoch": 0.63, + "learning_rate": 6.358301588519575e-06, + "loss": 0.2916, + "step": 10430 + }, + { + "epoch": 0.63, + "learning_rate": 6.356479369059715e-06, + "loss": 0.3089, + "step": 10431 + }, + { + "epoch": 0.63, + "learning_rate": 6.35465728909088e-06, + "loss": 0.3058, + "step": 10432 + }, + { + "epoch": 0.63, + "learning_rate": 6.352835348682824e-06, + "loss": 0.2868, + "step": 10433 + }, + { + "epoch": 0.63, + "learning_rate": 6.3510135479053e-06, + "loss": 0.3071, + "step": 10434 + }, + { + "epoch": 0.63, + "learning_rate": 6.349191886828059e-06, + "loss": 0.3153, + "step": 10435 + }, + { + "epoch": 0.63, + "learning_rate": 6.3473703655208376e-06, + "loss": 0.3066, + "step": 10436 + }, + { + "epoch": 0.63, + "learning_rate": 6.345548984053373e-06, + "loss": 0.2983, + "step": 10437 + }, + { + "epoch": 0.63, + "learning_rate": 6.343727742495398e-06, + "loss": 0.3008, + "step": 10438 + }, + { + "epoch": 0.63, + "learning_rate": 6.341906640916638e-06, + "loss": 0.2783, + "step": 10439 + }, + { + "epoch": 0.63, + "learning_rate": 6.340085679386812e-06, + "loss": 0.3152, + "step": 10440 + }, + { + "epoch": 0.63, + "learning_rate": 6.338264857975638e-06, + "loss": 0.2916, + "step": 10441 + }, + { + "epoch": 0.63, + "learning_rate": 6.336444176752819e-06, + "loss": 0.2814, + "step": 10442 + }, + { + "epoch": 0.63, + "learning_rate": 6.334623635788067e-06, + "loss": 0.2899, + "step": 10443 + }, + { + "epoch": 0.63, + "learning_rate": 6.332803235151076e-06, + "loss": 0.3078, + "step": 10444 + }, + { + "epoch": 0.63, + "learning_rate": 6.330982974911542e-06, + "loss": 0.3069, + "step": 10445 + }, + { + "epoch": 0.63, + "learning_rate": 6.329162855139153e-06, + "loss": 0.2679, + "step": 10446 + }, + { + "epoch": 0.63, + "learning_rate": 6.32734287590359e-06, + "loss": 0.2938, + "step": 10447 + }, + { + "epoch": 0.63, + "learning_rate": 6.325523037274529e-06, + "loss": 0.2673, + "step": 10448 + }, + { + "epoch": 0.63, + "learning_rate": 6.323703339321645e-06, + "loss": 0.2614, + "step": 10449 + }, + { + "epoch": 0.63, + "learning_rate": 6.321883782114604e-06, + "loss": 0.2839, + "step": 10450 + }, + { + "epoch": 0.63, + "learning_rate": 6.320064365723065e-06, + "loss": 0.2938, + "step": 10451 + }, + { + "epoch": 0.63, + "learning_rate": 6.318245090216686e-06, + "loss": 0.2983, + "step": 10452 + }, + { + "epoch": 0.63, + "learning_rate": 6.316425955665118e-06, + "loss": 0.2793, + "step": 10453 + }, + { + "epoch": 0.63, + "learning_rate": 6.314606962138001e-06, + "loss": 0.2822, + "step": 10454 + }, + { + "epoch": 0.63, + "learning_rate": 6.312788109704981e-06, + "loss": 0.2872, + "step": 10455 + }, + { + "epoch": 0.63, + "learning_rate": 6.310969398435689e-06, + "loss": 0.3009, + "step": 10456 + }, + { + "epoch": 0.63, + "learning_rate": 6.309150828399755e-06, + "loss": 0.2983, + "step": 10457 + }, + { + "epoch": 0.63, + "learning_rate": 6.307332399666802e-06, + "loss": 0.3117, + "step": 10458 + }, + { + "epoch": 0.63, + "learning_rate": 6.305514112306446e-06, + "loss": 0.2881, + "step": 10459 + }, + { + "epoch": 0.63, + "learning_rate": 6.303695966388302e-06, + "loss": 0.3008, + "step": 10460 + }, + { + "epoch": 0.63, + "learning_rate": 6.301877961981974e-06, + "loss": 0.2888, + "step": 10461 + }, + { + "epoch": 0.63, + "learning_rate": 6.300060099157071e-06, + "loss": 0.2536, + "step": 10462 + }, + { + "epoch": 0.63, + "learning_rate": 6.298242377983182e-06, + "loss": 0.2699, + "step": 10463 + }, + { + "epoch": 0.63, + "learning_rate": 6.296424798529898e-06, + "loss": 0.3064, + "step": 10464 + }, + { + "epoch": 0.63, + "learning_rate": 6.294607360866809e-06, + "loss": 0.3181, + "step": 10465 + }, + { + "epoch": 0.63, + "learning_rate": 6.29279006506349e-06, + "loss": 0.2896, + "step": 10466 + }, + { + "epoch": 0.63, + "learning_rate": 6.290972911189521e-06, + "loss": 0.2896, + "step": 10467 + }, + { + "epoch": 0.63, + "learning_rate": 6.289155899314469e-06, + "loss": 0.3061, + "step": 10468 + }, + { + "epoch": 0.63, + "learning_rate": 6.287339029507894e-06, + "loss": 0.258, + "step": 10469 + }, + { + "epoch": 0.63, + "learning_rate": 6.285522301839359e-06, + "loss": 0.2822, + "step": 10470 + }, + { + "epoch": 0.63, + "learning_rate": 6.283705716378417e-06, + "loss": 0.3023, + "step": 10471 + }, + { + "epoch": 0.63, + "learning_rate": 6.28188927319461e-06, + "loss": 0.2997, + "step": 10472 + }, + { + "epoch": 0.63, + "learning_rate": 6.280072972357487e-06, + "loss": 0.254, + "step": 10473 + }, + { + "epoch": 0.63, + "learning_rate": 6.2782568139365775e-06, + "loss": 0.2754, + "step": 10474 + }, + { + "epoch": 0.63, + "learning_rate": 6.276440798001419e-06, + "loss": 0.2755, + "step": 10475 + }, + { + "epoch": 0.63, + "learning_rate": 6.274624924621534e-06, + "loss": 0.2879, + "step": 10476 + }, + { + "epoch": 0.63, + "learning_rate": 6.27280919386644e-06, + "loss": 0.2761, + "step": 10477 + }, + { + "epoch": 0.63, + "learning_rate": 6.270993605805659e-06, + "loss": 0.28, + "step": 10478 + }, + { + "epoch": 0.63, + "learning_rate": 6.269178160508691e-06, + "loss": 0.31, + "step": 10479 + }, + { + "epoch": 0.63, + "learning_rate": 6.267362858045048e-06, + "loss": 0.3207, + "step": 10480 + }, + { + "epoch": 0.63, + "learning_rate": 6.265547698484227e-06, + "loss": 0.3097, + "step": 10481 + }, + { + "epoch": 0.63, + "learning_rate": 6.2637326818957154e-06, + "loss": 0.3199, + "step": 10482 + }, + { + "epoch": 0.63, + "learning_rate": 6.261917808349002e-06, + "loss": 0.2903, + "step": 10483 + }, + { + "epoch": 0.63, + "learning_rate": 6.260103077913573e-06, + "loss": 0.3198, + "step": 10484 + }, + { + "epoch": 0.63, + "learning_rate": 6.258288490658901e-06, + "loss": 0.2924, + "step": 10485 + }, + { + "epoch": 0.63, + "learning_rate": 6.256474046654456e-06, + "loss": 0.2999, + "step": 10486 + }, + { + "epoch": 0.63, + "learning_rate": 6.2546597459697075e-06, + "loss": 0.2934, + "step": 10487 + }, + { + "epoch": 0.63, + "learning_rate": 6.252845588674115e-06, + "loss": 0.2919, + "step": 10488 + }, + { + "epoch": 0.63, + "learning_rate": 6.251031574837127e-06, + "loss": 0.279, + "step": 10489 + }, + { + "epoch": 0.63, + "learning_rate": 6.2492177045282e-06, + "loss": 0.2785, + "step": 10490 + }, + { + "epoch": 0.63, + "learning_rate": 6.24740397781677e-06, + "loss": 0.2762, + "step": 10491 + }, + { + "epoch": 0.63, + "learning_rate": 6.245590394772284e-06, + "loss": 0.2899, + "step": 10492 + }, + { + "epoch": 0.63, + "learning_rate": 6.24377695546417e-06, + "loss": 0.2628, + "step": 10493 + }, + { + "epoch": 0.63, + "learning_rate": 6.241963659961849e-06, + "loss": 0.2987, + "step": 10494 + }, + { + "epoch": 0.63, + "learning_rate": 6.240150508334752e-06, + "loss": 0.29, + "step": 10495 + }, + { + "epoch": 0.63, + "learning_rate": 6.238337500652288e-06, + "loss": 0.3209, + "step": 10496 + }, + { + "epoch": 0.63, + "learning_rate": 6.236524636983874e-06, + "loss": 0.3108, + "step": 10497 + }, + { + "epoch": 0.63, + "learning_rate": 6.234711917398909e-06, + "loss": 0.3041, + "step": 10498 + }, + { + "epoch": 0.63, + "learning_rate": 6.232899341966794e-06, + "loss": 0.2683, + "step": 10499 + }, + { + "epoch": 0.63, + "learning_rate": 6.231086910756925e-06, + "loss": 0.279, + "step": 10500 + }, + { + "epoch": 0.63, + "learning_rate": 6.229274623838686e-06, + "loss": 0.2967, + "step": 10501 + }, + { + "epoch": 0.63, + "learning_rate": 6.2274624812814655e-06, + "loss": 0.2723, + "step": 10502 + }, + { + "epoch": 0.63, + "learning_rate": 6.225650483154637e-06, + "loss": 0.2861, + "step": 10503 + }, + { + "epoch": 0.63, + "learning_rate": 6.223838629527572e-06, + "loss": 0.3051, + "step": 10504 + }, + { + "epoch": 0.63, + "learning_rate": 6.22202692046964e-06, + "loss": 0.2574, + "step": 10505 + }, + { + "epoch": 0.63, + "learning_rate": 6.2202153560502e-06, + "loss": 0.3018, + "step": 10506 + }, + { + "epoch": 0.63, + "learning_rate": 6.218403936338604e-06, + "loss": 0.2907, + "step": 10507 + }, + { + "epoch": 0.63, + "learning_rate": 6.216592661404206e-06, + "loss": 0.286, + "step": 10508 + }, + { + "epoch": 0.63, + "learning_rate": 6.2147815313163475e-06, + "loss": 0.2694, + "step": 10509 + }, + { + "epoch": 0.63, + "learning_rate": 6.21297054614437e-06, + "loss": 0.2965, + "step": 10510 + }, + { + "epoch": 0.63, + "learning_rate": 6.2111597059576035e-06, + "loss": 0.3047, + "step": 10511 + }, + { + "epoch": 0.64, + "learning_rate": 6.209349010825375e-06, + "loss": 0.2926, + "step": 10512 + }, + { + "epoch": 0.64, + "learning_rate": 6.2075384608170095e-06, + "loss": 0.2958, + "step": 10513 + }, + { + "epoch": 0.64, + "learning_rate": 6.205728056001821e-06, + "loss": 0.299, + "step": 10514 + }, + { + "epoch": 0.64, + "learning_rate": 6.203917796449125e-06, + "loss": 0.3206, + "step": 10515 + }, + { + "epoch": 0.64, + "learning_rate": 6.202107682228218e-06, + "loss": 0.2916, + "step": 10516 + }, + { + "epoch": 0.64, + "learning_rate": 6.200297713408406e-06, + "loss": 0.2991, + "step": 10517 + }, + { + "epoch": 0.64, + "learning_rate": 6.198487890058979e-06, + "loss": 0.2825, + "step": 10518 + }, + { + "epoch": 0.64, + "learning_rate": 6.19667821224923e-06, + "loss": 0.298, + "step": 10519 + }, + { + "epoch": 0.64, + "learning_rate": 6.194868680048439e-06, + "loss": 0.2815, + "step": 10520 + }, + { + "epoch": 0.64, + "learning_rate": 6.193059293525883e-06, + "loss": 0.2831, + "step": 10521 + }, + { + "epoch": 0.64, + "learning_rate": 6.191250052750837e-06, + "loss": 0.2733, + "step": 10522 + }, + { + "epoch": 0.64, + "learning_rate": 6.189440957792563e-06, + "loss": 0.3055, + "step": 10523 + }, + { + "epoch": 0.64, + "learning_rate": 6.187632008720326e-06, + "loss": 0.2962, + "step": 10524 + }, + { + "epoch": 0.64, + "learning_rate": 6.185823205603378e-06, + "loss": 0.3061, + "step": 10525 + }, + { + "epoch": 0.64, + "learning_rate": 6.184014548510968e-06, + "loss": 0.285, + "step": 10526 + }, + { + "epoch": 0.64, + "learning_rate": 6.182206037512343e-06, + "loss": 0.2679, + "step": 10527 + }, + { + "epoch": 0.64, + "learning_rate": 6.180397672676739e-06, + "loss": 0.2928, + "step": 10528 + }, + { + "epoch": 0.64, + "learning_rate": 6.1785894540733874e-06, + "loss": 0.2736, + "step": 10529 + }, + { + "epoch": 0.64, + "learning_rate": 6.176781381771519e-06, + "loss": 0.2644, + "step": 10530 + }, + { + "epoch": 0.64, + "learning_rate": 6.174973455840351e-06, + "loss": 0.2891, + "step": 10531 + }, + { + "epoch": 0.64, + "learning_rate": 6.173165676349103e-06, + "loss": 0.2889, + "step": 10532 + }, + { + "epoch": 0.64, + "learning_rate": 6.171358043366984e-06, + "loss": 0.302, + "step": 10533 + }, + { + "epoch": 0.64, + "learning_rate": 6.1695505569631976e-06, + "loss": 0.2985, + "step": 10534 + }, + { + "epoch": 0.64, + "learning_rate": 6.167743217206945e-06, + "loss": 0.2628, + "step": 10535 + }, + { + "epoch": 0.64, + "learning_rate": 6.165936024167415e-06, + "loss": 0.2577, + "step": 10536 + }, + { + "epoch": 0.64, + "learning_rate": 6.1641289779138016e-06, + "loss": 0.3125, + "step": 10537 + }, + { + "epoch": 0.64, + "learning_rate": 6.162322078515284e-06, + "loss": 0.3044, + "step": 10538 + }, + { + "epoch": 0.64, + "learning_rate": 6.160515326041037e-06, + "loss": 0.2896, + "step": 10539 + }, + { + "epoch": 0.64, + "learning_rate": 6.158708720560235e-06, + "loss": 0.2598, + "step": 10540 + }, + { + "epoch": 0.64, + "learning_rate": 6.156902262142042e-06, + "loss": 0.2768, + "step": 10541 + }, + { + "epoch": 0.64, + "learning_rate": 6.155095950855614e-06, + "loss": 0.2786, + "step": 10542 + }, + { + "epoch": 0.64, + "learning_rate": 6.153289786770111e-06, + "loss": 0.296, + "step": 10543 + }, + { + "epoch": 0.64, + "learning_rate": 6.151483769954678e-06, + "loss": 0.3107, + "step": 10544 + }, + { + "epoch": 0.64, + "learning_rate": 6.149677900478459e-06, + "loss": 0.3013, + "step": 10545 + }, + { + "epoch": 0.64, + "learning_rate": 6.147872178410592e-06, + "loss": 0.2741, + "step": 10546 + }, + { + "epoch": 0.64, + "learning_rate": 6.1460666038202044e-06, + "loss": 0.302, + "step": 10547 + }, + { + "epoch": 0.64, + "learning_rate": 6.144261176776427e-06, + "loss": 0.2845, + "step": 10548 + }, + { + "epoch": 0.64, + "learning_rate": 6.142455897348376e-06, + "loss": 0.3052, + "step": 10549 + }, + { + "epoch": 0.64, + "learning_rate": 6.140650765605174e-06, + "loss": 0.2656, + "step": 10550 + }, + { + "epoch": 0.64, + "learning_rate": 6.138845781615918e-06, + "loss": 0.2878, + "step": 10551 + }, + { + "epoch": 0.64, + "learning_rate": 6.137040945449718e-06, + "loss": 0.2958, + "step": 10552 + }, + { + "epoch": 0.64, + "learning_rate": 6.1352362571756686e-06, + "loss": 0.2817, + "step": 10553 + }, + { + "epoch": 0.64, + "learning_rate": 6.1334317168628656e-06, + "loss": 0.2706, + "step": 10554 + }, + { + "epoch": 0.64, + "learning_rate": 6.131627324580394e-06, + "loss": 0.3181, + "step": 10555 + }, + { + "epoch": 0.64, + "learning_rate": 6.1298230803973305e-06, + "loss": 0.2976, + "step": 10556 + }, + { + "epoch": 0.64, + "learning_rate": 6.128018984382755e-06, + "loss": 0.269, + "step": 10557 + }, + { + "epoch": 0.64, + "learning_rate": 6.126215036605735e-06, + "loss": 0.2842, + "step": 10558 + }, + { + "epoch": 0.64, + "learning_rate": 6.124411237135335e-06, + "loss": 0.2994, + "step": 10559 + }, + { + "epoch": 0.64, + "learning_rate": 6.122607586040611e-06, + "loss": 0.2648, + "step": 10560 + }, + { + "epoch": 0.64, + "learning_rate": 6.120804083390615e-06, + "loss": 0.2896, + "step": 10561 + }, + { + "epoch": 0.64, + "learning_rate": 6.119000729254395e-06, + "loss": 0.3114, + "step": 10562 + }, + { + "epoch": 0.64, + "learning_rate": 6.117197523700993e-06, + "loss": 0.2825, + "step": 10563 + }, + { + "epoch": 0.64, + "learning_rate": 6.115394466799441e-06, + "loss": 0.2974, + "step": 10564 + }, + { + "epoch": 0.64, + "learning_rate": 6.11359155861877e-06, + "loss": 0.3004, + "step": 10565 + }, + { + "epoch": 0.64, + "learning_rate": 6.111788799228002e-06, + "loss": 0.2913, + "step": 10566 + }, + { + "epoch": 0.64, + "learning_rate": 6.10998618869616e-06, + "loss": 0.2903, + "step": 10567 + }, + { + "epoch": 0.64, + "learning_rate": 6.108183727092253e-06, + "loss": 0.2677, + "step": 10568 + }, + { + "epoch": 0.64, + "learning_rate": 6.106381414485285e-06, + "loss": 0.2732, + "step": 10569 + }, + { + "epoch": 0.64, + "learning_rate": 6.104579250944263e-06, + "loss": 0.2806, + "step": 10570 + }, + { + "epoch": 0.64, + "learning_rate": 6.102777236538175e-06, + "loss": 0.2867, + "step": 10571 + }, + { + "epoch": 0.64, + "learning_rate": 6.10097537133602e-06, + "loss": 0.2928, + "step": 10572 + }, + { + "epoch": 0.64, + "learning_rate": 6.099173655406774e-06, + "loss": 0.3025, + "step": 10573 + }, + { + "epoch": 0.64, + "learning_rate": 6.097372088819417e-06, + "loss": 0.3202, + "step": 10574 + }, + { + "epoch": 0.64, + "learning_rate": 6.095570671642923e-06, + "loss": 0.2702, + "step": 10575 + }, + { + "epoch": 0.64, + "learning_rate": 6.0937694039462545e-06, + "loss": 0.3336, + "step": 10576 + }, + { + "epoch": 0.64, + "learning_rate": 6.09196828579838e-06, + "loss": 0.2761, + "step": 10577 + }, + { + "epoch": 0.64, + "learning_rate": 6.09016731726825e-06, + "loss": 0.3079, + "step": 10578 + }, + { + "epoch": 0.64, + "learning_rate": 6.088366498424812e-06, + "loss": 0.3281, + "step": 10579 + }, + { + "epoch": 0.64, + "learning_rate": 6.086565829337012e-06, + "loss": 0.2697, + "step": 10580 + }, + { + "epoch": 0.64, + "learning_rate": 6.0847653100737905e-06, + "loss": 0.2659, + "step": 10581 + }, + { + "epoch": 0.64, + "learning_rate": 6.0829649407040744e-06, + "loss": 0.2921, + "step": 10582 + }, + { + "epoch": 0.64, + "learning_rate": 6.0811647212967975e-06, + "loss": 0.3178, + "step": 10583 + }, + { + "epoch": 0.64, + "learning_rate": 6.079364651920876e-06, + "loss": 0.2704, + "step": 10584 + }, + { + "epoch": 0.64, + "learning_rate": 6.077564732645224e-06, + "loss": 0.2825, + "step": 10585 + }, + { + "epoch": 0.64, + "learning_rate": 6.0757649635387515e-06, + "loss": 0.3111, + "step": 10586 + }, + { + "epoch": 0.64, + "learning_rate": 6.073965344670364e-06, + "loss": 0.2899, + "step": 10587 + }, + { + "epoch": 0.64, + "learning_rate": 6.072165876108957e-06, + "loss": 0.2902, + "step": 10588 + }, + { + "epoch": 0.64, + "learning_rate": 6.0703665579234235e-06, + "loss": 0.3172, + "step": 10589 + }, + { + "epoch": 0.64, + "learning_rate": 6.068567390182653e-06, + "loss": 0.3104, + "step": 10590 + }, + { + "epoch": 0.64, + "learning_rate": 6.066768372955519e-06, + "loss": 0.291, + "step": 10591 + }, + { + "epoch": 0.64, + "learning_rate": 6.064969506310902e-06, + "loss": 0.2697, + "step": 10592 + }, + { + "epoch": 0.64, + "learning_rate": 6.0631707903176695e-06, + "loss": 0.275, + "step": 10593 + }, + { + "epoch": 0.64, + "learning_rate": 6.061372225044686e-06, + "loss": 0.2921, + "step": 10594 + }, + { + "epoch": 0.64, + "learning_rate": 6.059573810560808e-06, + "loss": 0.2716, + "step": 10595 + }, + { + "epoch": 0.64, + "learning_rate": 6.057775546934885e-06, + "loss": 0.2811, + "step": 10596 + }, + { + "epoch": 0.64, + "learning_rate": 6.055977434235767e-06, + "loss": 0.3027, + "step": 10597 + }, + { + "epoch": 0.64, + "learning_rate": 6.054179472532291e-06, + "loss": 0.2899, + "step": 10598 + }, + { + "epoch": 0.64, + "learning_rate": 6.052381661893292e-06, + "loss": 0.2854, + "step": 10599 + }, + { + "epoch": 0.64, + "learning_rate": 6.050584002387601e-06, + "loss": 0.2698, + "step": 10600 + }, + { + "epoch": 0.64, + "learning_rate": 6.048786494084036e-06, + "loss": 0.3228, + "step": 10601 + }, + { + "epoch": 0.64, + "learning_rate": 6.046989137051419e-06, + "loss": 0.3005, + "step": 10602 + }, + { + "epoch": 0.64, + "learning_rate": 6.045191931358558e-06, + "loss": 0.296, + "step": 10603 + }, + { + "epoch": 0.64, + "learning_rate": 6.04339487707426e-06, + "loss": 0.2942, + "step": 10604 + }, + { + "epoch": 0.64, + "learning_rate": 6.041597974267326e-06, + "loss": 0.2878, + "step": 10605 + }, + { + "epoch": 0.64, + "learning_rate": 6.039801223006545e-06, + "loss": 0.3083, + "step": 10606 + }, + { + "epoch": 0.64, + "learning_rate": 6.038004623360711e-06, + "loss": 0.3018, + "step": 10607 + }, + { + "epoch": 0.64, + "learning_rate": 6.036208175398602e-06, + "loss": 0.2849, + "step": 10608 + }, + { + "epoch": 0.64, + "learning_rate": 6.034411879188995e-06, + "loss": 0.2895, + "step": 10609 + }, + { + "epoch": 0.64, + "learning_rate": 6.032615734800664e-06, + "loss": 0.3089, + "step": 10610 + }, + { + "epoch": 0.64, + "learning_rate": 6.0308197423023685e-06, + "loss": 0.2952, + "step": 10611 + }, + { + "epoch": 0.64, + "learning_rate": 6.029023901762873e-06, + "loss": 0.2994, + "step": 10612 + }, + { + "epoch": 0.64, + "learning_rate": 6.027228213250926e-06, + "loss": 0.2884, + "step": 10613 + }, + { + "epoch": 0.64, + "learning_rate": 6.0254326768352764e-06, + "loss": 0.3181, + "step": 10614 + }, + { + "epoch": 0.64, + "learning_rate": 6.02363729258467e-06, + "loss": 0.2967, + "step": 10615 + }, + { + "epoch": 0.64, + "learning_rate": 6.021842060567835e-06, + "loss": 0.2949, + "step": 10616 + }, + { + "epoch": 0.64, + "learning_rate": 6.020046980853506e-06, + "loss": 0.2953, + "step": 10617 + }, + { + "epoch": 0.64, + "learning_rate": 6.018252053510409e-06, + "loss": 0.2648, + "step": 10618 + }, + { + "epoch": 0.64, + "learning_rate": 6.016457278607256e-06, + "loss": 0.3147, + "step": 10619 + }, + { + "epoch": 0.64, + "learning_rate": 6.014662656212764e-06, + "loss": 0.2949, + "step": 10620 + }, + { + "epoch": 0.64, + "learning_rate": 6.012868186395635e-06, + "loss": 0.2857, + "step": 10621 + }, + { + "epoch": 0.64, + "learning_rate": 6.011073869224576e-06, + "loss": 0.2898, + "step": 10622 + }, + { + "epoch": 0.64, + "learning_rate": 6.009279704768276e-06, + "loss": 0.2898, + "step": 10623 + }, + { + "epoch": 0.64, + "learning_rate": 6.0074856930954294e-06, + "loss": 0.291, + "step": 10624 + }, + { + "epoch": 0.64, + "learning_rate": 6.005691834274716e-06, + "loss": 0.2939, + "step": 10625 + }, + { + "epoch": 0.64, + "learning_rate": 6.003898128374813e-06, + "loss": 0.2617, + "step": 10626 + }, + { + "epoch": 0.64, + "learning_rate": 6.002104575464393e-06, + "loss": 0.2951, + "step": 10627 + }, + { + "epoch": 0.64, + "learning_rate": 6.0003111756121214e-06, + "loss": 0.2902, + "step": 10628 + }, + { + "epoch": 0.64, + "learning_rate": 5.998517928886658e-06, + "loss": 0.2734, + "step": 10629 + }, + { + "epoch": 0.64, + "learning_rate": 5.996724835356658e-06, + "loss": 0.2906, + "step": 10630 + }, + { + "epoch": 0.64, + "learning_rate": 5.994931895090766e-06, + "loss": 0.2948, + "step": 10631 + }, + { + "epoch": 0.64, + "learning_rate": 5.993139108157626e-06, + "loss": 0.3064, + "step": 10632 + }, + { + "epoch": 0.64, + "learning_rate": 5.9913464746258756e-06, + "loss": 0.2999, + "step": 10633 + }, + { + "epoch": 0.64, + "learning_rate": 5.989553994564141e-06, + "loss": 0.2885, + "step": 10634 + }, + { + "epoch": 0.64, + "learning_rate": 5.987761668041052e-06, + "loss": 0.2951, + "step": 10635 + }, + { + "epoch": 0.64, + "learning_rate": 5.9859694951252244e-06, + "loss": 0.2581, + "step": 10636 + }, + { + "epoch": 0.64, + "learning_rate": 5.984177475885272e-06, + "loss": 0.2888, + "step": 10637 + }, + { + "epoch": 0.64, + "learning_rate": 5.982385610389803e-06, + "loss": 0.2891, + "step": 10638 + }, + { + "epoch": 0.64, + "learning_rate": 5.980593898707412e-06, + "loss": 0.2999, + "step": 10639 + }, + { + "epoch": 0.64, + "learning_rate": 5.978802340906702e-06, + "loss": 0.2891, + "step": 10640 + }, + { + "epoch": 0.64, + "learning_rate": 5.9770109370562556e-06, + "loss": 0.2983, + "step": 10641 + }, + { + "epoch": 0.64, + "learning_rate": 5.975219687224663e-06, + "loss": 0.2789, + "step": 10642 + }, + { + "epoch": 0.64, + "learning_rate": 5.973428591480497e-06, + "loss": 0.2847, + "step": 10643 + }, + { + "epoch": 0.64, + "learning_rate": 5.971637649892329e-06, + "loss": 0.2736, + "step": 10644 + }, + { + "epoch": 0.64, + "learning_rate": 5.969846862528727e-06, + "loss": 0.2766, + "step": 10645 + }, + { + "epoch": 0.64, + "learning_rate": 5.968056229458247e-06, + "loss": 0.3092, + "step": 10646 + }, + { + "epoch": 0.64, + "learning_rate": 5.966265750749449e-06, + "loss": 0.2939, + "step": 10647 + }, + { + "epoch": 0.64, + "learning_rate": 5.964475426470877e-06, + "loss": 0.2882, + "step": 10648 + }, + { + "epoch": 0.64, + "learning_rate": 5.962685256691071e-06, + "loss": 0.3013, + "step": 10649 + }, + { + "epoch": 0.64, + "learning_rate": 5.96089524147857e-06, + "loss": 0.3237, + "step": 10650 + }, + { + "epoch": 0.64, + "learning_rate": 5.959105380901906e-06, + "loss": 0.2972, + "step": 10651 + }, + { + "epoch": 0.64, + "learning_rate": 5.9573156750296e-06, + "loss": 0.2637, + "step": 10652 + }, + { + "epoch": 0.64, + "learning_rate": 5.955526123930168e-06, + "loss": 0.2846, + "step": 10653 + }, + { + "epoch": 0.64, + "learning_rate": 5.9537367276721285e-06, + "loss": 0.3059, + "step": 10654 + }, + { + "epoch": 0.64, + "learning_rate": 5.9519474863239836e-06, + "loss": 0.3064, + "step": 10655 + }, + { + "epoch": 0.64, + "learning_rate": 5.950158399954233e-06, + "loss": 0.3198, + "step": 10656 + }, + { + "epoch": 0.64, + "learning_rate": 5.948369468631376e-06, + "loss": 0.2963, + "step": 10657 + }, + { + "epoch": 0.64, + "learning_rate": 5.946580692423897e-06, + "loss": 0.2972, + "step": 10658 + }, + { + "epoch": 0.64, + "learning_rate": 5.9447920714002815e-06, + "loss": 0.2949, + "step": 10659 + }, + { + "epoch": 0.64, + "learning_rate": 5.943003605629005e-06, + "loss": 0.2818, + "step": 10660 + }, + { + "epoch": 0.64, + "learning_rate": 5.941215295178538e-06, + "loss": 0.2956, + "step": 10661 + }, + { + "epoch": 0.64, + "learning_rate": 5.939427140117346e-06, + "loss": 0.2843, + "step": 10662 + }, + { + "epoch": 0.64, + "learning_rate": 5.937639140513887e-06, + "loss": 0.3033, + "step": 10663 + }, + { + "epoch": 0.64, + "learning_rate": 5.935851296436616e-06, + "loss": 0.3016, + "step": 10664 + }, + { + "epoch": 0.64, + "learning_rate": 5.934063607953981e-06, + "loss": 0.2785, + "step": 10665 + }, + { + "epoch": 0.64, + "learning_rate": 5.932276075134418e-06, + "loss": 0.3111, + "step": 10666 + }, + { + "epoch": 0.64, + "learning_rate": 5.930488698046366e-06, + "loss": 0.3008, + "step": 10667 + }, + { + "epoch": 0.64, + "learning_rate": 5.928701476758254e-06, + "loss": 0.2984, + "step": 10668 + }, + { + "epoch": 0.64, + "learning_rate": 5.926914411338502e-06, + "loss": 0.28, + "step": 10669 + }, + { + "epoch": 0.64, + "learning_rate": 5.925127501855532e-06, + "loss": 0.2905, + "step": 10670 + }, + { + "epoch": 0.64, + "learning_rate": 5.92334074837775e-06, + "loss": 0.2983, + "step": 10671 + }, + { + "epoch": 0.64, + "learning_rate": 5.921554150973567e-06, + "loss": 0.2831, + "step": 10672 + }, + { + "epoch": 0.64, + "learning_rate": 5.919767709711382e-06, + "loss": 0.2719, + "step": 10673 + }, + { + "epoch": 0.64, + "learning_rate": 5.91798142465958e-06, + "loss": 0.3086, + "step": 10674 + }, + { + "epoch": 0.64, + "learning_rate": 5.916195295886558e-06, + "loss": 0.2819, + "step": 10675 + }, + { + "epoch": 0.64, + "learning_rate": 5.914409323460692e-06, + "loss": 0.315, + "step": 10676 + }, + { + "epoch": 0.65, + "learning_rate": 5.912623507450359e-06, + "loss": 0.2997, + "step": 10677 + }, + { + "epoch": 0.65, + "learning_rate": 5.91083784792393e-06, + "loss": 0.285, + "step": 10678 + }, + { + "epoch": 0.65, + "learning_rate": 5.9090523449497635e-06, + "loss": 0.2698, + "step": 10679 + }, + { + "epoch": 0.65, + "learning_rate": 5.907266998596224e-06, + "loss": 0.3078, + "step": 10680 + }, + { + "epoch": 0.65, + "learning_rate": 5.905481808931657e-06, + "loss": 0.291, + "step": 10681 + }, + { + "epoch": 0.65, + "learning_rate": 5.90369677602441e-06, + "loss": 0.3105, + "step": 10682 + }, + { + "epoch": 0.65, + "learning_rate": 5.901911899942825e-06, + "loss": 0.3145, + "step": 10683 + }, + { + "epoch": 0.65, + "learning_rate": 5.9001271807552285e-06, + "loss": 0.2635, + "step": 10684 + }, + { + "epoch": 0.65, + "learning_rate": 5.898342618529956e-06, + "loss": 0.264, + "step": 10685 + }, + { + "epoch": 0.65, + "learning_rate": 5.896558213335326e-06, + "loss": 0.2914, + "step": 10686 + }, + { + "epoch": 0.65, + "learning_rate": 5.89477396523965e-06, + "loss": 0.2753, + "step": 10687 + }, + { + "epoch": 0.65, + "learning_rate": 5.892989874311241e-06, + "loss": 0.3287, + "step": 10688 + }, + { + "epoch": 0.65, + "learning_rate": 5.891205940618402e-06, + "loss": 0.2914, + "step": 10689 + }, + { + "epoch": 0.65, + "learning_rate": 5.88942216422943e-06, + "loss": 0.2995, + "step": 10690 + }, + { + "epoch": 0.65, + "learning_rate": 5.887638545212615e-06, + "loss": 0.3083, + "step": 10691 + }, + { + "epoch": 0.65, + "learning_rate": 5.885855083636246e-06, + "loss": 0.287, + "step": 10692 + }, + { + "epoch": 0.65, + "learning_rate": 5.8840717795685975e-06, + "loss": 0.281, + "step": 10693 + }, + { + "epoch": 0.65, + "learning_rate": 5.8822886330779475e-06, + "loss": 0.2646, + "step": 10694 + }, + { + "epoch": 0.65, + "learning_rate": 5.880505644232559e-06, + "loss": 0.308, + "step": 10695 + }, + { + "epoch": 0.65, + "learning_rate": 5.878722813100694e-06, + "loss": 0.2838, + "step": 10696 + }, + { + "epoch": 0.65, + "learning_rate": 5.876940139750612e-06, + "loss": 0.2729, + "step": 10697 + }, + { + "epoch": 0.65, + "learning_rate": 5.875157624250556e-06, + "loss": 0.308, + "step": 10698 + }, + { + "epoch": 0.65, + "learning_rate": 5.8733752666687726e-06, + "loss": 0.2825, + "step": 10699 + }, + { + "epoch": 0.65, + "learning_rate": 5.8715930670735e-06, + "loss": 0.3029, + "step": 10700 + }, + { + "epoch": 0.65, + "learning_rate": 5.869811025532963e-06, + "loss": 0.2716, + "step": 10701 + }, + { + "epoch": 0.65, + "learning_rate": 5.8680291421153934e-06, + "loss": 0.2669, + "step": 10702 + }, + { + "epoch": 0.65, + "learning_rate": 5.866247416889007e-06, + "loss": 0.2974, + "step": 10703 + }, + { + "epoch": 0.65, + "learning_rate": 5.8644658499220144e-06, + "loss": 0.2712, + "step": 10704 + }, + { + "epoch": 0.65, + "learning_rate": 5.862684441282628e-06, + "loss": 0.287, + "step": 10705 + }, + { + "epoch": 0.65, + "learning_rate": 5.860903191039043e-06, + "loss": 0.2871, + "step": 10706 + }, + { + "epoch": 0.65, + "learning_rate": 5.859122099259456e-06, + "loss": 0.2916, + "step": 10707 + }, + { + "epoch": 0.65, + "learning_rate": 5.857341166012057e-06, + "loss": 0.2815, + "step": 10708 + }, + { + "epoch": 0.65, + "learning_rate": 5.855560391365025e-06, + "loss": 0.2779, + "step": 10709 + }, + { + "epoch": 0.65, + "learning_rate": 5.853779775386541e-06, + "loss": 0.3006, + "step": 10710 + }, + { + "epoch": 0.65, + "learning_rate": 5.851999318144771e-06, + "loss": 0.2699, + "step": 10711 + }, + { + "epoch": 0.65, + "learning_rate": 5.850219019707882e-06, + "loss": 0.2566, + "step": 10712 + }, + { + "epoch": 0.65, + "learning_rate": 5.848438880144031e-06, + "loss": 0.2821, + "step": 10713 + }, + { + "epoch": 0.65, + "learning_rate": 5.846658899521371e-06, + "loss": 0.3158, + "step": 10714 + }, + { + "epoch": 0.65, + "learning_rate": 5.844879077908048e-06, + "loss": 0.2896, + "step": 10715 + }, + { + "epoch": 0.65, + "learning_rate": 5.843099415372197e-06, + "loss": 0.2983, + "step": 10716 + }, + { + "epoch": 0.65, + "learning_rate": 5.84131991198196e-06, + "loss": 0.2489, + "step": 10717 + }, + { + "epoch": 0.65, + "learning_rate": 5.839540567805461e-06, + "loss": 0.2677, + "step": 10718 + }, + { + "epoch": 0.65, + "learning_rate": 5.837761382910821e-06, + "loss": 0.2892, + "step": 10719 + }, + { + "epoch": 0.65, + "learning_rate": 5.8359823573661545e-06, + "loss": 0.2928, + "step": 10720 + }, + { + "epoch": 0.65, + "learning_rate": 5.834203491239574e-06, + "loss": 0.2853, + "step": 10721 + }, + { + "epoch": 0.65, + "learning_rate": 5.83242478459918e-06, + "loss": 0.2935, + "step": 10722 + }, + { + "epoch": 0.65, + "learning_rate": 5.830646237513071e-06, + "loss": 0.2882, + "step": 10723 + }, + { + "epoch": 0.65, + "learning_rate": 5.828867850049335e-06, + "loss": 0.296, + "step": 10724 + }, + { + "epoch": 0.65, + "learning_rate": 5.827089622276063e-06, + "loss": 0.2955, + "step": 10725 + }, + { + "epoch": 0.65, + "learning_rate": 5.82531155426133e-06, + "loss": 0.2938, + "step": 10726 + }, + { + "epoch": 0.65, + "learning_rate": 5.823533646073212e-06, + "loss": 0.3145, + "step": 10727 + }, + { + "epoch": 0.65, + "learning_rate": 5.82175589777977e-06, + "loss": 0.262, + "step": 10728 + }, + { + "epoch": 0.65, + "learning_rate": 5.8199783094490665e-06, + "loss": 0.3002, + "step": 10729 + }, + { + "epoch": 0.65, + "learning_rate": 5.8182008811491606e-06, + "loss": 0.2797, + "step": 10730 + }, + { + "epoch": 0.65, + "learning_rate": 5.816423612948096e-06, + "loss": 0.3092, + "step": 10731 + }, + { + "epoch": 0.65, + "learning_rate": 5.814646504913917e-06, + "loss": 0.2966, + "step": 10732 + }, + { + "epoch": 0.65, + "learning_rate": 5.812869557114659e-06, + "loss": 0.2719, + "step": 10733 + }, + { + "epoch": 0.65, + "learning_rate": 5.81109276961835e-06, + "loss": 0.3061, + "step": 10734 + }, + { + "epoch": 0.65, + "learning_rate": 5.809316142493012e-06, + "loss": 0.3164, + "step": 10735 + }, + { + "epoch": 0.65, + "learning_rate": 5.80753967580667e-06, + "loss": 0.298, + "step": 10736 + }, + { + "epoch": 0.65, + "learning_rate": 5.805763369627331e-06, + "loss": 0.286, + "step": 10737 + }, + { + "epoch": 0.65, + "learning_rate": 5.8039872240229996e-06, + "loss": 0.2985, + "step": 10738 + }, + { + "epoch": 0.65, + "learning_rate": 5.802211239061677e-06, + "loss": 0.2418, + "step": 10739 + }, + { + "epoch": 0.65, + "learning_rate": 5.800435414811352e-06, + "loss": 0.2977, + "step": 10740 + }, + { + "epoch": 0.65, + "learning_rate": 5.798659751340017e-06, + "loss": 0.2903, + "step": 10741 + }, + { + "epoch": 0.65, + "learning_rate": 5.7968842487156505e-06, + "loss": 0.2655, + "step": 10742 + }, + { + "epoch": 0.65, + "learning_rate": 5.795108907006228e-06, + "loss": 0.2563, + "step": 10743 + }, + { + "epoch": 0.65, + "learning_rate": 5.793333726279716e-06, + "loss": 0.305, + "step": 10744 + }, + { + "epoch": 0.65, + "learning_rate": 5.791558706604075e-06, + "loss": 0.2893, + "step": 10745 + }, + { + "epoch": 0.65, + "learning_rate": 5.789783848047266e-06, + "loss": 0.2926, + "step": 10746 + }, + { + "epoch": 0.65, + "learning_rate": 5.7880091506772366e-06, + "loss": 0.2694, + "step": 10747 + }, + { + "epoch": 0.65, + "learning_rate": 5.7862346145619295e-06, + "loss": 0.3184, + "step": 10748 + }, + { + "epoch": 0.65, + "learning_rate": 5.784460239769285e-06, + "loss": 0.2981, + "step": 10749 + }, + { + "epoch": 0.65, + "learning_rate": 5.7826860263672265e-06, + "loss": 0.3192, + "step": 10750 + }, + { + "epoch": 0.65, + "learning_rate": 5.7809119744236906e-06, + "loss": 0.3199, + "step": 10751 + }, + { + "epoch": 0.65, + "learning_rate": 5.7791380840065916e-06, + "loss": 0.2942, + "step": 10752 + }, + { + "epoch": 0.65, + "learning_rate": 5.777364355183841e-06, + "loss": 0.296, + "step": 10753 + }, + { + "epoch": 0.65, + "learning_rate": 5.775590788023346e-06, + "loss": 0.2642, + "step": 10754 + }, + { + "epoch": 0.65, + "learning_rate": 5.773817382593008e-06, + "loss": 0.291, + "step": 10755 + }, + { + "epoch": 0.65, + "learning_rate": 5.77204413896072e-06, + "loss": 0.286, + "step": 10756 + }, + { + "epoch": 0.65, + "learning_rate": 5.7702710571943695e-06, + "loss": 0.2807, + "step": 10757 + }, + { + "epoch": 0.65, + "learning_rate": 5.76849813736184e-06, + "loss": 0.2893, + "step": 10758 + }, + { + "epoch": 0.65, + "learning_rate": 5.766725379531003e-06, + "loss": 0.2826, + "step": 10759 + }, + { + "epoch": 0.65, + "learning_rate": 5.764952783769735e-06, + "loss": 0.293, + "step": 10760 + }, + { + "epoch": 0.65, + "learning_rate": 5.763180350145896e-06, + "loss": 0.3045, + "step": 10761 + }, + { + "epoch": 0.65, + "learning_rate": 5.761408078727344e-06, + "loss": 0.2736, + "step": 10762 + }, + { + "epoch": 0.65, + "learning_rate": 5.759635969581927e-06, + "loss": 0.2852, + "step": 10763 + }, + { + "epoch": 0.65, + "learning_rate": 5.757864022777489e-06, + "loss": 0.2994, + "step": 10764 + }, + { + "epoch": 0.65, + "learning_rate": 5.756092238381874e-06, + "loss": 0.2896, + "step": 10765 + }, + { + "epoch": 0.65, + "learning_rate": 5.754320616462911e-06, + "loss": 0.3242, + "step": 10766 + }, + { + "epoch": 0.65, + "learning_rate": 5.752549157088425e-06, + "loss": 0.2969, + "step": 10767 + }, + { + "epoch": 0.65, + "learning_rate": 5.750777860326238e-06, + "loss": 0.2825, + "step": 10768 + }, + { + "epoch": 0.65, + "learning_rate": 5.749006726244162e-06, + "loss": 0.2941, + "step": 10769 + }, + { + "epoch": 0.65, + "learning_rate": 5.74723575491e-06, + "loss": 0.3276, + "step": 10770 + }, + { + "epoch": 0.65, + "learning_rate": 5.745464946391561e-06, + "loss": 0.2718, + "step": 10771 + }, + { + "epoch": 0.65, + "learning_rate": 5.743694300756638e-06, + "loss": 0.2966, + "step": 10772 + }, + { + "epoch": 0.65, + "learning_rate": 5.7419238180730174e-06, + "loss": 0.2854, + "step": 10773 + }, + { + "epoch": 0.65, + "learning_rate": 5.740153498408483e-06, + "loss": 0.2988, + "step": 10774 + }, + { + "epoch": 0.65, + "learning_rate": 5.738383341830804e-06, + "loss": 0.2863, + "step": 10775 + }, + { + "epoch": 0.65, + "learning_rate": 5.7366133484077625e-06, + "loss": 0.2935, + "step": 10776 + }, + { + "epoch": 0.65, + "learning_rate": 5.734843518207114e-06, + "loss": 0.2921, + "step": 10777 + }, + { + "epoch": 0.65, + "learning_rate": 5.73307385129662e-06, + "loss": 0.2882, + "step": 10778 + }, + { + "epoch": 0.65, + "learning_rate": 5.731304347744029e-06, + "loss": 0.2955, + "step": 10779 + }, + { + "epoch": 0.65, + "learning_rate": 5.729535007617082e-06, + "loss": 0.2846, + "step": 10780 + }, + { + "epoch": 0.65, + "learning_rate": 5.727765830983525e-06, + "loss": 0.2863, + "step": 10781 + }, + { + "epoch": 0.65, + "learning_rate": 5.7259968179110895e-06, + "loss": 0.2914, + "step": 10782 + }, + { + "epoch": 0.65, + "learning_rate": 5.724227968467497e-06, + "loss": 0.2845, + "step": 10783 + }, + { + "epoch": 0.65, + "learning_rate": 5.722459282720472e-06, + "loss": 0.2983, + "step": 10784 + }, + { + "epoch": 0.65, + "learning_rate": 5.720690760737721e-06, + "loss": 0.3181, + "step": 10785 + }, + { + "epoch": 0.65, + "learning_rate": 5.718922402586962e-06, + "loss": 0.2893, + "step": 10786 + }, + { + "epoch": 0.65, + "learning_rate": 5.7171542083358875e-06, + "loss": 0.2422, + "step": 10787 + }, + { + "epoch": 0.65, + "learning_rate": 5.7153861780522e-06, + "loss": 0.2994, + "step": 10788 + }, + { + "epoch": 0.65, + "learning_rate": 5.713618311803576e-06, + "loss": 0.3023, + "step": 10789 + }, + { + "epoch": 0.65, + "learning_rate": 5.711850609657708e-06, + "loss": 0.3019, + "step": 10790 + }, + { + "epoch": 0.65, + "learning_rate": 5.710083071682269e-06, + "loss": 0.2997, + "step": 10791 + }, + { + "epoch": 0.65, + "learning_rate": 5.708315697944926e-06, + "loss": 0.2785, + "step": 10792 + }, + { + "epoch": 0.65, + "learning_rate": 5.706548488513347e-06, + "loss": 0.3117, + "step": 10793 + }, + { + "epoch": 0.65, + "learning_rate": 5.704781443455182e-06, + "loss": 0.2872, + "step": 10794 + }, + { + "epoch": 0.65, + "learning_rate": 5.7030145628380895e-06, + "loss": 0.3211, + "step": 10795 + }, + { + "epoch": 0.65, + "learning_rate": 5.701247846729712e-06, + "loss": 0.3044, + "step": 10796 + }, + { + "epoch": 0.65, + "learning_rate": 5.6994812951976855e-06, + "loss": 0.3309, + "step": 10797 + }, + { + "epoch": 0.65, + "learning_rate": 5.697714908309643e-06, + "loss": 0.2857, + "step": 10798 + }, + { + "epoch": 0.65, + "learning_rate": 5.695948686133206e-06, + "loss": 0.2726, + "step": 10799 + }, + { + "epoch": 0.65, + "learning_rate": 5.694182628736003e-06, + "loss": 0.2872, + "step": 10800 + }, + { + "epoch": 0.65, + "learning_rate": 5.69241673618564e-06, + "loss": 0.2764, + "step": 10801 + }, + { + "epoch": 0.65, + "learning_rate": 5.690651008549725e-06, + "loss": 0.275, + "step": 10802 + }, + { + "epoch": 0.65, + "learning_rate": 5.688885445895858e-06, + "loss": 0.286, + "step": 10803 + }, + { + "epoch": 0.65, + "learning_rate": 5.687120048291634e-06, + "loss": 0.2684, + "step": 10804 + }, + { + "epoch": 0.65, + "learning_rate": 5.685354815804638e-06, + "loss": 0.25, + "step": 10805 + }, + { + "epoch": 0.65, + "learning_rate": 5.683589748502456e-06, + "loss": 0.2847, + "step": 10806 + }, + { + "epoch": 0.65, + "learning_rate": 5.6818248464526604e-06, + "loss": 0.2924, + "step": 10807 + }, + { + "epoch": 0.65, + "learning_rate": 5.68006010972282e-06, + "loss": 0.3016, + "step": 10808 + }, + { + "epoch": 0.65, + "learning_rate": 5.678295538380498e-06, + "loss": 0.2856, + "step": 10809 + }, + { + "epoch": 0.65, + "learning_rate": 5.6765311324932455e-06, + "loss": 0.2944, + "step": 10810 + }, + { + "epoch": 0.65, + "learning_rate": 5.674766892128621e-06, + "loss": 0.3089, + "step": 10811 + }, + { + "epoch": 0.65, + "learning_rate": 5.673002817354163e-06, + "loss": 0.3211, + "step": 10812 + }, + { + "epoch": 0.65, + "learning_rate": 5.671238908237409e-06, + "loss": 0.3165, + "step": 10813 + }, + { + "epoch": 0.65, + "learning_rate": 5.66947516484589e-06, + "loss": 0.2928, + "step": 10814 + }, + { + "epoch": 0.65, + "learning_rate": 5.667711587247124e-06, + "loss": 0.2988, + "step": 10815 + }, + { + "epoch": 0.65, + "learning_rate": 5.665948175508641e-06, + "loss": 0.2828, + "step": 10816 + }, + { + "epoch": 0.65, + "learning_rate": 5.664184929697946e-06, + "loss": 0.2912, + "step": 10817 + }, + { + "epoch": 0.65, + "learning_rate": 5.662421849882544e-06, + "loss": 0.2785, + "step": 10818 + }, + { + "epoch": 0.65, + "learning_rate": 5.660658936129936e-06, + "loss": 0.266, + "step": 10819 + }, + { + "epoch": 0.65, + "learning_rate": 5.6588961885076085e-06, + "loss": 0.3039, + "step": 10820 + }, + { + "epoch": 0.65, + "learning_rate": 5.657133607083058e-06, + "loss": 0.2974, + "step": 10821 + }, + { + "epoch": 0.65, + "learning_rate": 5.655371191923762e-06, + "loss": 0.2985, + "step": 10822 + }, + { + "epoch": 0.65, + "learning_rate": 5.653608943097186e-06, + "loss": 0.2656, + "step": 10823 + }, + { + "epoch": 0.65, + "learning_rate": 5.6518468606708e-06, + "loss": 0.29, + "step": 10824 + }, + { + "epoch": 0.65, + "learning_rate": 5.650084944712071e-06, + "loss": 0.3117, + "step": 10825 + }, + { + "epoch": 0.65, + "learning_rate": 5.648323195288448e-06, + "loss": 0.3107, + "step": 10826 + }, + { + "epoch": 0.65, + "learning_rate": 5.6465616124673804e-06, + "loss": 0.308, + "step": 10827 + }, + { + "epoch": 0.65, + "learning_rate": 5.644800196316311e-06, + "loss": 0.2832, + "step": 10828 + }, + { + "epoch": 0.65, + "learning_rate": 5.643038946902668e-06, + "loss": 0.2892, + "step": 10829 + }, + { + "epoch": 0.65, + "learning_rate": 5.641277864293892e-06, + "loss": 0.2532, + "step": 10830 + }, + { + "epoch": 0.65, + "learning_rate": 5.6395169485573985e-06, + "loss": 0.2787, + "step": 10831 + }, + { + "epoch": 0.65, + "learning_rate": 5.637756199760605e-06, + "loss": 0.2698, + "step": 10832 + }, + { + "epoch": 0.65, + "learning_rate": 5.63599561797092e-06, + "loss": 0.2852, + "step": 10833 + }, + { + "epoch": 0.65, + "learning_rate": 5.634235203255745e-06, + "loss": 0.2992, + "step": 10834 + }, + { + "epoch": 0.65, + "learning_rate": 5.632474955682483e-06, + "loss": 0.293, + "step": 10835 + }, + { + "epoch": 0.65, + "learning_rate": 5.630714875318522e-06, + "loss": 0.2919, + "step": 10836 + }, + { + "epoch": 0.65, + "learning_rate": 5.628954962231243e-06, + "loss": 0.2856, + "step": 10837 + }, + { + "epoch": 0.65, + "learning_rate": 5.627195216488029e-06, + "loss": 0.3022, + "step": 10838 + }, + { + "epoch": 0.65, + "learning_rate": 5.625435638156245e-06, + "loss": 0.3013, + "step": 10839 + }, + { + "epoch": 0.65, + "learning_rate": 5.623676227303257e-06, + "loss": 0.2693, + "step": 10840 + }, + { + "epoch": 0.65, + "learning_rate": 5.621916983996429e-06, + "loss": 0.3013, + "step": 10841 + }, + { + "epoch": 0.65, + "learning_rate": 5.62015790830311e-06, + "loss": 0.2778, + "step": 10842 + }, + { + "epoch": 0.66, + "learning_rate": 5.618399000290645e-06, + "loss": 0.2762, + "step": 10843 + }, + { + "epoch": 0.66, + "learning_rate": 5.616640260026374e-06, + "loss": 0.3178, + "step": 10844 + }, + { + "epoch": 0.66, + "learning_rate": 5.614881687577624e-06, + "loss": 0.2909, + "step": 10845 + }, + { + "epoch": 0.66, + "learning_rate": 5.6131232830117335e-06, + "loss": 0.3125, + "step": 10846 + }, + { + "epoch": 0.66, + "learning_rate": 5.611365046396013e-06, + "loss": 0.3033, + "step": 10847 + }, + { + "epoch": 0.66, + "learning_rate": 5.609606977797781e-06, + "loss": 0.3008, + "step": 10848 + }, + { + "epoch": 0.66, + "learning_rate": 5.6078490772843404e-06, + "loss": 0.2804, + "step": 10849 + }, + { + "epoch": 0.66, + "learning_rate": 5.60609134492299e-06, + "loss": 0.3027, + "step": 10850 + }, + { + "epoch": 0.66, + "learning_rate": 5.604333780781033e-06, + "loss": 0.2932, + "step": 10851 + }, + { + "epoch": 0.66, + "learning_rate": 5.602576384925752e-06, + "loss": 0.2926, + "step": 10852 + }, + { + "epoch": 0.66, + "learning_rate": 5.600819157424428e-06, + "loss": 0.2829, + "step": 10853 + }, + { + "epoch": 0.66, + "learning_rate": 5.599062098344336e-06, + "loss": 0.2906, + "step": 10854 + }, + { + "epoch": 0.66, + "learning_rate": 5.59730520775274e-06, + "loss": 0.2769, + "step": 10855 + }, + { + "epoch": 0.66, + "learning_rate": 5.595548485716915e-06, + "loss": 0.2792, + "step": 10856 + }, + { + "epoch": 0.66, + "learning_rate": 5.593791932304105e-06, + "loss": 0.2666, + "step": 10857 + }, + { + "epoch": 0.66, + "learning_rate": 5.592035547581561e-06, + "loss": 0.2812, + "step": 10858 + }, + { + "epoch": 0.66, + "learning_rate": 5.590279331616524e-06, + "loss": 0.2822, + "step": 10859 + }, + { + "epoch": 0.66, + "learning_rate": 5.588523284476237e-06, + "loss": 0.2818, + "step": 10860 + }, + { + "epoch": 0.66, + "learning_rate": 5.586767406227926e-06, + "loss": 0.2881, + "step": 10861 + }, + { + "epoch": 0.66, + "learning_rate": 5.585011696938813e-06, + "loss": 0.2733, + "step": 10862 + }, + { + "epoch": 0.66, + "learning_rate": 5.583256156676117e-06, + "loss": 0.2886, + "step": 10863 + }, + { + "epoch": 0.66, + "learning_rate": 5.581500785507041e-06, + "loss": 0.258, + "step": 10864 + }, + { + "epoch": 0.66, + "learning_rate": 5.579745583498802e-06, + "loss": 0.2771, + "step": 10865 + }, + { + "epoch": 0.66, + "learning_rate": 5.5779905507185885e-06, + "loss": 0.28, + "step": 10866 + }, + { + "epoch": 0.66, + "learning_rate": 5.5762356872335935e-06, + "loss": 0.2747, + "step": 10867 + }, + { + "epoch": 0.66, + "learning_rate": 5.574480993111e-06, + "loss": 0.2884, + "step": 10868 + }, + { + "epoch": 0.66, + "learning_rate": 5.5727264684179836e-06, + "loss": 0.2684, + "step": 10869 + }, + { + "epoch": 0.66, + "learning_rate": 5.570972113221724e-06, + "loss": 0.2992, + "step": 10870 + }, + { + "epoch": 0.66, + "learning_rate": 5.56921792758938e-06, + "loss": 0.2645, + "step": 10871 + }, + { + "epoch": 0.66, + "learning_rate": 5.567463911588112e-06, + "loss": 0.3005, + "step": 10872 + }, + { + "epoch": 0.66, + "learning_rate": 5.56571006528507e-06, + "loss": 0.2818, + "step": 10873 + }, + { + "epoch": 0.66, + "learning_rate": 5.563956388747401e-06, + "loss": 0.2727, + "step": 10874 + }, + { + "epoch": 0.66, + "learning_rate": 5.562202882042241e-06, + "loss": 0.2952, + "step": 10875 + }, + { + "epoch": 0.66, + "learning_rate": 5.560449545236727e-06, + "loss": 0.2875, + "step": 10876 + }, + { + "epoch": 0.66, + "learning_rate": 5.558696378397983e-06, + "loss": 0.3136, + "step": 10877 + }, + { + "epoch": 0.66, + "learning_rate": 5.5569433815931296e-06, + "loss": 0.2979, + "step": 10878 + }, + { + "epoch": 0.66, + "learning_rate": 5.555190554889278e-06, + "loss": 0.262, + "step": 10879 + }, + { + "epoch": 0.66, + "learning_rate": 5.5534378983535305e-06, + "loss": 0.2836, + "step": 10880 + }, + { + "epoch": 0.66, + "learning_rate": 5.551685412052997e-06, + "loss": 0.2868, + "step": 10881 + }, + { + "epoch": 0.66, + "learning_rate": 5.549933096054765e-06, + "loss": 0.3011, + "step": 10882 + }, + { + "epoch": 0.66, + "learning_rate": 5.548180950425921e-06, + "loss": 0.2845, + "step": 10883 + }, + { + "epoch": 0.66, + "learning_rate": 5.546428975233546e-06, + "loss": 0.3054, + "step": 10884 + }, + { + "epoch": 0.66, + "learning_rate": 5.544677170544711e-06, + "loss": 0.2878, + "step": 10885 + }, + { + "epoch": 0.66, + "learning_rate": 5.54292553642649e-06, + "loss": 0.2881, + "step": 10886 + }, + { + "epoch": 0.66, + "learning_rate": 5.54117407294594e-06, + "loss": 0.2783, + "step": 10887 + }, + { + "epoch": 0.66, + "learning_rate": 5.539422780170114e-06, + "loss": 0.2663, + "step": 10888 + }, + { + "epoch": 0.66, + "learning_rate": 5.537671658166063e-06, + "loss": 0.2679, + "step": 10889 + }, + { + "epoch": 0.66, + "learning_rate": 5.535920707000825e-06, + "loss": 0.305, + "step": 10890 + }, + { + "epoch": 0.66, + "learning_rate": 5.534169926741435e-06, + "loss": 0.2861, + "step": 10891 + }, + { + "epoch": 0.66, + "learning_rate": 5.532419317454922e-06, + "loss": 0.2966, + "step": 10892 + }, + { + "epoch": 0.66, + "learning_rate": 5.530668879208307e-06, + "loss": 0.3234, + "step": 10893 + }, + { + "epoch": 0.66, + "learning_rate": 5.528918612068604e-06, + "loss": 0.2882, + "step": 10894 + }, + { + "epoch": 0.66, + "learning_rate": 5.527168516102823e-06, + "loss": 0.2787, + "step": 10895 + }, + { + "epoch": 0.66, + "learning_rate": 5.525418591377967e-06, + "loss": 0.2902, + "step": 10896 + }, + { + "epoch": 0.66, + "learning_rate": 5.523668837961029e-06, + "loss": 0.2941, + "step": 10897 + }, + { + "epoch": 0.66, + "learning_rate": 5.521919255918999e-06, + "loss": 0.2899, + "step": 10898 + }, + { + "epoch": 0.66, + "learning_rate": 5.520169845318856e-06, + "loss": 0.2838, + "step": 10899 + }, + { + "epoch": 0.66, + "learning_rate": 5.518420606227582e-06, + "loss": 0.2829, + "step": 10900 + }, + { + "epoch": 0.66, + "learning_rate": 5.51667153871214e-06, + "loss": 0.2852, + "step": 10901 + }, + { + "epoch": 0.66, + "learning_rate": 5.514922642839498e-06, + "loss": 0.281, + "step": 10902 + }, + { + "epoch": 0.66, + "learning_rate": 5.513173918676606e-06, + "loss": 0.2899, + "step": 10903 + }, + { + "epoch": 0.66, + "learning_rate": 5.511425366290414e-06, + "loss": 0.2951, + "step": 10904 + }, + { + "epoch": 0.66, + "learning_rate": 5.509676985747871e-06, + "loss": 0.2751, + "step": 10905 + }, + { + "epoch": 0.66, + "learning_rate": 5.507928777115907e-06, + "loss": 0.2868, + "step": 10906 + }, + { + "epoch": 0.66, + "learning_rate": 5.506180740461457e-06, + "loss": 0.291, + "step": 10907 + }, + { + "epoch": 0.66, + "learning_rate": 5.5044328758514385e-06, + "loss": 0.3037, + "step": 10908 + }, + { + "epoch": 0.66, + "learning_rate": 5.5026851833527714e-06, + "loss": 0.2955, + "step": 10909 + }, + { + "epoch": 0.66, + "learning_rate": 5.500937663032359e-06, + "loss": 0.3061, + "step": 10910 + }, + { + "epoch": 0.66, + "learning_rate": 5.499190314957114e-06, + "loss": 0.2874, + "step": 10911 + }, + { + "epoch": 0.66, + "learning_rate": 5.497443139193931e-06, + "loss": 0.2945, + "step": 10912 + }, + { + "epoch": 0.66, + "learning_rate": 5.495696135809696e-06, + "loss": 0.3033, + "step": 10913 + }, + { + "epoch": 0.66, + "learning_rate": 5.4939493048712955e-06, + "loss": 0.2896, + "step": 10914 + }, + { + "epoch": 0.66, + "learning_rate": 5.492202646445601e-06, + "loss": 0.2864, + "step": 10915 + }, + { + "epoch": 0.66, + "learning_rate": 5.490456160599491e-06, + "loss": 0.3009, + "step": 10916 + }, + { + "epoch": 0.66, + "learning_rate": 5.488709847399827e-06, + "loss": 0.2931, + "step": 10917 + }, + { + "epoch": 0.66, + "learning_rate": 5.486963706913462e-06, + "loss": 0.2708, + "step": 10918 + }, + { + "epoch": 0.66, + "learning_rate": 5.485217739207249e-06, + "loss": 0.2574, + "step": 10919 + }, + { + "epoch": 0.66, + "learning_rate": 5.483471944348029e-06, + "loss": 0.2872, + "step": 10920 + }, + { + "epoch": 0.66, + "learning_rate": 5.4817263224026455e-06, + "loss": 0.2799, + "step": 10921 + }, + { + "epoch": 0.66, + "learning_rate": 5.479980873437923e-06, + "loss": 0.2815, + "step": 10922 + }, + { + "epoch": 0.66, + "learning_rate": 5.47823559752069e-06, + "loss": 0.3142, + "step": 10923 + }, + { + "epoch": 0.66, + "learning_rate": 5.476490494717761e-06, + "loss": 0.3041, + "step": 10924 + }, + { + "epoch": 0.66, + "learning_rate": 5.474745565095947e-06, + "loss": 0.293, + "step": 10925 + }, + { + "epoch": 0.66, + "learning_rate": 5.4730008087220536e-06, + "loss": 0.2748, + "step": 10926 + }, + { + "epoch": 0.66, + "learning_rate": 5.471256225662876e-06, + "loss": 0.3041, + "step": 10927 + }, + { + "epoch": 0.66, + "learning_rate": 5.469511815985207e-06, + "loss": 0.3086, + "step": 10928 + }, + { + "epoch": 0.66, + "learning_rate": 5.467767579755825e-06, + "loss": 0.2814, + "step": 10929 + }, + { + "epoch": 0.66, + "learning_rate": 5.466023517041517e-06, + "loss": 0.2733, + "step": 10930 + }, + { + "epoch": 0.66, + "learning_rate": 5.464279627909048e-06, + "loss": 0.2434, + "step": 10931 + }, + { + "epoch": 0.66, + "learning_rate": 5.462535912425184e-06, + "loss": 0.278, + "step": 10932 + }, + { + "epoch": 0.66, + "learning_rate": 5.460792370656682e-06, + "loss": 0.2972, + "step": 10933 + }, + { + "epoch": 0.66, + "learning_rate": 5.459049002670289e-06, + "loss": 0.3002, + "step": 10934 + }, + { + "epoch": 0.66, + "learning_rate": 5.457305808532758e-06, + "loss": 0.2937, + "step": 10935 + }, + { + "epoch": 0.66, + "learning_rate": 5.455562788310822e-06, + "loss": 0.3011, + "step": 10936 + }, + { + "epoch": 0.66, + "learning_rate": 5.453819942071212e-06, + "loss": 0.268, + "step": 10937 + }, + { + "epoch": 0.66, + "learning_rate": 5.452077269880651e-06, + "loss": 0.2888, + "step": 10938 + }, + { + "epoch": 0.66, + "learning_rate": 5.450334771805854e-06, + "loss": 0.286, + "step": 10939 + }, + { + "epoch": 0.66, + "learning_rate": 5.44859244791354e-06, + "loss": 0.2839, + "step": 10940 + }, + { + "epoch": 0.66, + "learning_rate": 5.4468502982704096e-06, + "loss": 0.2867, + "step": 10941 + }, + { + "epoch": 0.66, + "learning_rate": 5.44510832294316e-06, + "loss": 0.2662, + "step": 10942 + }, + { + "epoch": 0.66, + "learning_rate": 5.443366521998481e-06, + "loss": 0.2866, + "step": 10943 + }, + { + "epoch": 0.66, + "learning_rate": 5.4416248955030596e-06, + "loss": 0.2799, + "step": 10944 + }, + { + "epoch": 0.66, + "learning_rate": 5.439883443523566e-06, + "loss": 0.3138, + "step": 10945 + }, + { + "epoch": 0.66, + "learning_rate": 5.4381421661266835e-06, + "loss": 0.2801, + "step": 10946 + }, + { + "epoch": 0.66, + "learning_rate": 5.436401063379068e-06, + "loss": 0.3013, + "step": 10947 + }, + { + "epoch": 0.66, + "learning_rate": 5.434660135347378e-06, + "loss": 0.278, + "step": 10948 + }, + { + "epoch": 0.66, + "learning_rate": 5.432919382098267e-06, + "loss": 0.2699, + "step": 10949 + }, + { + "epoch": 0.66, + "learning_rate": 5.431178803698374e-06, + "loss": 0.2913, + "step": 10950 + }, + { + "epoch": 0.66, + "learning_rate": 5.429438400214342e-06, + "loss": 0.2847, + "step": 10951 + }, + { + "epoch": 0.66, + "learning_rate": 5.427698171712801e-06, + "loss": 0.2977, + "step": 10952 + }, + { + "epoch": 0.66, + "learning_rate": 5.425958118260374e-06, + "loss": 0.2712, + "step": 10953 + }, + { + "epoch": 0.66, + "learning_rate": 5.424218239923678e-06, + "loss": 0.3112, + "step": 10954 + }, + { + "epoch": 0.66, + "learning_rate": 5.42247853676932e-06, + "loss": 0.2981, + "step": 10955 + }, + { + "epoch": 0.66, + "learning_rate": 5.4207390088639114e-06, + "loss": 0.2748, + "step": 10956 + }, + { + "epoch": 0.66, + "learning_rate": 5.418999656274045e-06, + "loss": 0.2852, + "step": 10957 + }, + { + "epoch": 0.66, + "learning_rate": 5.417260479066318e-06, + "loss": 0.2871, + "step": 10958 + }, + { + "epoch": 0.66, + "learning_rate": 5.415521477307299e-06, + "loss": 0.3029, + "step": 10959 + }, + { + "epoch": 0.66, + "learning_rate": 5.4137826510635796e-06, + "loss": 0.2786, + "step": 10960 + }, + { + "epoch": 0.66, + "learning_rate": 5.412044000401726e-06, + "loss": 0.2935, + "step": 10961 + }, + { + "epoch": 0.66, + "learning_rate": 5.4103055253882995e-06, + "loss": 0.2955, + "step": 10962 + }, + { + "epoch": 0.66, + "learning_rate": 5.4085672260898606e-06, + "loss": 0.2766, + "step": 10963 + }, + { + "epoch": 0.66, + "learning_rate": 5.406829102572952e-06, + "loss": 0.2785, + "step": 10964 + }, + { + "epoch": 0.66, + "learning_rate": 5.405091154904127e-06, + "loss": 0.2699, + "step": 10965 + }, + { + "epoch": 0.66, + "learning_rate": 5.4033533831499195e-06, + "loss": 0.291, + "step": 10966 + }, + { + "epoch": 0.66, + "learning_rate": 5.401615787376857e-06, + "loss": 0.2973, + "step": 10967 + }, + { + "epoch": 0.66, + "learning_rate": 5.3998783676514655e-06, + "loss": 0.2698, + "step": 10968 + }, + { + "epoch": 0.66, + "learning_rate": 5.398141124040255e-06, + "loss": 0.2592, + "step": 10969 + }, + { + "epoch": 0.66, + "learning_rate": 5.396404056609745e-06, + "loss": 0.2879, + "step": 10970 + }, + { + "epoch": 0.66, + "learning_rate": 5.394667165426432e-06, + "loss": 0.2877, + "step": 10971 + }, + { + "epoch": 0.66, + "learning_rate": 5.392930450556817e-06, + "loss": 0.2946, + "step": 10972 + }, + { + "epoch": 0.66, + "learning_rate": 5.391193912067386e-06, + "loss": 0.2999, + "step": 10973 + }, + { + "epoch": 0.66, + "learning_rate": 5.389457550024619e-06, + "loss": 0.2921, + "step": 10974 + }, + { + "epoch": 0.66, + "learning_rate": 5.387721364495e-06, + "loss": 0.2699, + "step": 10975 + }, + { + "epoch": 0.66, + "learning_rate": 5.385985355544994e-06, + "loss": 0.2891, + "step": 10976 + }, + { + "epoch": 0.66, + "learning_rate": 5.3842495232410616e-06, + "loss": 0.3075, + "step": 10977 + }, + { + "epoch": 0.66, + "learning_rate": 5.382513867649663e-06, + "loss": 0.287, + "step": 10978 + }, + { + "epoch": 0.66, + "learning_rate": 5.38077838883724e-06, + "loss": 0.258, + "step": 10979 + }, + { + "epoch": 0.66, + "learning_rate": 5.379043086870244e-06, + "loss": 0.2976, + "step": 10980 + }, + { + "epoch": 0.66, + "learning_rate": 5.3773079618151055e-06, + "loss": 0.2825, + "step": 10981 + }, + { + "epoch": 0.66, + "learning_rate": 5.3755730137382545e-06, + "loss": 0.2891, + "step": 10982 + }, + { + "epoch": 0.66, + "learning_rate": 5.373838242706112e-06, + "loss": 0.2921, + "step": 10983 + }, + { + "epoch": 0.66, + "learning_rate": 5.372103648785094e-06, + "loss": 0.2845, + "step": 10984 + }, + { + "epoch": 0.66, + "learning_rate": 5.370369232041603e-06, + "loss": 0.3248, + "step": 10985 + }, + { + "epoch": 0.66, + "learning_rate": 5.368634992542052e-06, + "loss": 0.2737, + "step": 10986 + }, + { + "epoch": 0.66, + "learning_rate": 5.366900930352828e-06, + "loss": 0.2899, + "step": 10987 + }, + { + "epoch": 0.66, + "learning_rate": 5.365167045540323e-06, + "loss": 0.2955, + "step": 10988 + }, + { + "epoch": 0.66, + "learning_rate": 5.363433338170913e-06, + "loss": 0.2746, + "step": 10989 + }, + { + "epoch": 0.66, + "learning_rate": 5.361699808310974e-06, + "loss": 0.2921, + "step": 10990 + }, + { + "epoch": 0.66, + "learning_rate": 5.359966456026877e-06, + "loss": 0.3052, + "step": 10991 + }, + { + "epoch": 0.66, + "learning_rate": 5.3582332813849855e-06, + "loss": 0.2713, + "step": 10992 + }, + { + "epoch": 0.66, + "learning_rate": 5.3565002844516465e-06, + "loss": 0.281, + "step": 10993 + }, + { + "epoch": 0.66, + "learning_rate": 5.354767465293205e-06, + "loss": 0.2962, + "step": 10994 + }, + { + "epoch": 0.66, + "learning_rate": 5.353034823976009e-06, + "loss": 0.3133, + "step": 10995 + }, + { + "epoch": 0.66, + "learning_rate": 5.351302360566391e-06, + "loss": 0.3044, + "step": 10996 + }, + { + "epoch": 0.66, + "learning_rate": 5.349570075130673e-06, + "loss": 0.3262, + "step": 10997 + }, + { + "epoch": 0.66, + "learning_rate": 5.3478379677351814e-06, + "loss": 0.2963, + "step": 10998 + }, + { + "epoch": 0.66, + "learning_rate": 5.346106038446219e-06, + "loss": 0.2891, + "step": 10999 + }, + { + "epoch": 0.66, + "learning_rate": 5.344374287330105e-06, + "loss": 0.2733, + "step": 11000 + }, + { + "epoch": 0.66, + "learning_rate": 5.342642714453132e-06, + "loss": 0.2984, + "step": 11001 + }, + { + "epoch": 0.66, + "learning_rate": 5.340911319881595e-06, + "loss": 0.2974, + "step": 11002 + }, + { + "epoch": 0.66, + "learning_rate": 5.3391801036817785e-06, + "loss": 0.2698, + "step": 11003 + }, + { + "epoch": 0.66, + "learning_rate": 5.337449065919958e-06, + "loss": 0.3016, + "step": 11004 + }, + { + "epoch": 0.66, + "learning_rate": 5.3357182066624126e-06, + "loss": 0.3036, + "step": 11005 + }, + { + "epoch": 0.66, + "learning_rate": 5.333987525975405e-06, + "loss": 0.2931, + "step": 11006 + }, + { + "epoch": 0.66, + "learning_rate": 5.332257023925194e-06, + "loss": 0.2695, + "step": 11007 + }, + { + "epoch": 0.67, + "learning_rate": 5.33052670057803e-06, + "loss": 0.2818, + "step": 11008 + }, + { + "epoch": 0.67, + "learning_rate": 5.328796556000153e-06, + "loss": 0.2755, + "step": 11009 + }, + { + "epoch": 0.67, + "learning_rate": 5.327066590257813e-06, + "loss": 0.3265, + "step": 11010 + }, + { + "epoch": 0.67, + "learning_rate": 5.3253368034172335e-06, + "loss": 0.3172, + "step": 11011 + }, + { + "epoch": 0.67, + "learning_rate": 5.32360719554464e-06, + "loss": 0.3008, + "step": 11012 + }, + { + "epoch": 0.67, + "learning_rate": 5.321877766706252e-06, + "loss": 0.2801, + "step": 11013 + }, + { + "epoch": 0.67, + "learning_rate": 5.320148516968271e-06, + "loss": 0.3051, + "step": 11014 + }, + { + "epoch": 0.67, + "learning_rate": 5.318419446396914e-06, + "loss": 0.2893, + "step": 11015 + }, + { + "epoch": 0.67, + "learning_rate": 5.316690555058372e-06, + "loss": 0.3128, + "step": 11016 + }, + { + "epoch": 0.67, + "learning_rate": 5.3149618430188335e-06, + "loss": 0.2953, + "step": 11017 + }, + { + "epoch": 0.67, + "learning_rate": 5.313233310344484e-06, + "loss": 0.2744, + "step": 11018 + }, + { + "epoch": 0.67, + "learning_rate": 5.311504957101499e-06, + "loss": 0.2719, + "step": 11019 + }, + { + "epoch": 0.67, + "learning_rate": 5.309776783356044e-06, + "loss": 0.2829, + "step": 11020 + }, + { + "epoch": 0.67, + "learning_rate": 5.308048789174289e-06, + "loss": 0.2818, + "step": 11021 + }, + { + "epoch": 0.67, + "learning_rate": 5.3063209746223875e-06, + "loss": 0.2839, + "step": 11022 + }, + { + "epoch": 0.67, + "learning_rate": 5.3045933397664866e-06, + "loss": 0.2769, + "step": 11023 + }, + { + "epoch": 0.67, + "learning_rate": 5.302865884672728e-06, + "loss": 0.2892, + "step": 11024 + }, + { + "epoch": 0.67, + "learning_rate": 5.301138609407242e-06, + "loss": 0.2821, + "step": 11025 + }, + { + "epoch": 0.67, + "learning_rate": 5.299411514036172e-06, + "loss": 0.2931, + "step": 11026 + }, + { + "epoch": 0.67, + "learning_rate": 5.297684598625626e-06, + "loss": 0.2734, + "step": 11027 + }, + { + "epoch": 0.67, + "learning_rate": 5.295957863241721e-06, + "loss": 0.2849, + "step": 11028 + }, + { + "epoch": 0.67, + "learning_rate": 5.294231307950562e-06, + "loss": 0.298, + "step": 11029 + }, + { + "epoch": 0.67, + "learning_rate": 5.292504932818255e-06, + "loss": 0.2782, + "step": 11030 + }, + { + "epoch": 0.67, + "learning_rate": 5.290778737910893e-06, + "loss": 0.2938, + "step": 11031 + }, + { + "epoch": 0.67, + "learning_rate": 5.289052723294563e-06, + "loss": 0.2839, + "step": 11032 + }, + { + "epoch": 0.67, + "learning_rate": 5.287326889035342e-06, + "loss": 0.2866, + "step": 11033 + }, + { + "epoch": 0.67, + "learning_rate": 5.2856012351993e-06, + "loss": 0.293, + "step": 11034 + }, + { + "epoch": 0.67, + "learning_rate": 5.283875761852512e-06, + "loss": 0.2826, + "step": 11035 + }, + { + "epoch": 0.67, + "learning_rate": 5.2821504690610345e-06, + "loss": 0.2903, + "step": 11036 + }, + { + "epoch": 0.67, + "learning_rate": 5.280425356890917e-06, + "loss": 0.2826, + "step": 11037 + }, + { + "epoch": 0.67, + "learning_rate": 5.278700425408206e-06, + "loss": 0.2977, + "step": 11038 + }, + { + "epoch": 0.67, + "learning_rate": 5.276975674678936e-06, + "loss": 0.2598, + "step": 11039 + }, + { + "epoch": 0.67, + "learning_rate": 5.2752511047691455e-06, + "loss": 0.2966, + "step": 11040 + }, + { + "epoch": 0.67, + "learning_rate": 5.273526715744856e-06, + "loss": 0.2889, + "step": 11041 + }, + { + "epoch": 0.67, + "learning_rate": 5.271802507672087e-06, + "loss": 0.2662, + "step": 11042 + }, + { + "epoch": 0.67, + "learning_rate": 5.270078480616846e-06, + "loss": 0.299, + "step": 11043 + }, + { + "epoch": 0.67, + "learning_rate": 5.268354634645134e-06, + "loss": 0.2715, + "step": 11044 + }, + { + "epoch": 0.67, + "learning_rate": 5.266630969822958e-06, + "loss": 0.2832, + "step": 11045 + }, + { + "epoch": 0.67, + "learning_rate": 5.2649074862163005e-06, + "loss": 0.2891, + "step": 11046 + }, + { + "epoch": 0.67, + "learning_rate": 5.263184183891147e-06, + "loss": 0.2917, + "step": 11047 + }, + { + "epoch": 0.67, + "learning_rate": 5.261461062913473e-06, + "loss": 0.2826, + "step": 11048 + }, + { + "epoch": 0.67, + "learning_rate": 5.259738123349243e-06, + "loss": 0.2713, + "step": 11049 + }, + { + "epoch": 0.67, + "learning_rate": 5.258015365264426e-06, + "loss": 0.2729, + "step": 11050 + }, + { + "epoch": 0.67, + "learning_rate": 5.256292788724977e-06, + "loss": 0.2899, + "step": 11051 + }, + { + "epoch": 0.67, + "learning_rate": 5.2545703937968404e-06, + "loss": 0.2581, + "step": 11052 + }, + { + "epoch": 0.67, + "learning_rate": 5.252848180545962e-06, + "loss": 0.2713, + "step": 11053 + }, + { + "epoch": 0.67, + "learning_rate": 5.25112614903827e-06, + "loss": 0.2751, + "step": 11054 + }, + { + "epoch": 0.67, + "learning_rate": 5.249404299339693e-06, + "loss": 0.2669, + "step": 11055 + }, + { + "epoch": 0.67, + "learning_rate": 5.247682631516158e-06, + "loss": 0.309, + "step": 11056 + }, + { + "epoch": 0.67, + "learning_rate": 5.245961145633575e-06, + "loss": 0.3281, + "step": 11057 + }, + { + "epoch": 0.67, + "learning_rate": 5.244239841757848e-06, + "loss": 0.2596, + "step": 11058 + }, + { + "epoch": 0.67, + "learning_rate": 5.242518719954879e-06, + "loss": 0.266, + "step": 11059 + }, + { + "epoch": 0.67, + "learning_rate": 5.240797780290559e-06, + "loss": 0.2616, + "step": 11060 + }, + { + "epoch": 0.67, + "learning_rate": 5.239077022830776e-06, + "loss": 0.302, + "step": 11061 + }, + { + "epoch": 0.67, + "learning_rate": 5.2373564476414065e-06, + "loss": 0.2833, + "step": 11062 + }, + { + "epoch": 0.67, + "learning_rate": 5.2356360547883234e-06, + "loss": 0.2938, + "step": 11063 + }, + { + "epoch": 0.67, + "learning_rate": 5.233915844337387e-06, + "loss": 0.3061, + "step": 11064 + }, + { + "epoch": 0.67, + "learning_rate": 5.232195816354462e-06, + "loss": 0.3147, + "step": 11065 + }, + { + "epoch": 0.67, + "learning_rate": 5.230475970905396e-06, + "loss": 0.2817, + "step": 11066 + }, + { + "epoch": 0.67, + "learning_rate": 5.228756308056033e-06, + "loss": 0.3044, + "step": 11067 + }, + { + "epoch": 0.67, + "learning_rate": 5.227036827872209e-06, + "loss": 0.3139, + "step": 11068 + }, + { + "epoch": 0.67, + "learning_rate": 5.225317530419751e-06, + "loss": 0.2633, + "step": 11069 + }, + { + "epoch": 0.67, + "learning_rate": 5.223598415764489e-06, + "loss": 0.2667, + "step": 11070 + }, + { + "epoch": 0.67, + "learning_rate": 5.221879483972234e-06, + "loss": 0.2863, + "step": 11071 + }, + { + "epoch": 0.67, + "learning_rate": 5.220160735108797e-06, + "loss": 0.2553, + "step": 11072 + }, + { + "epoch": 0.67, + "learning_rate": 5.218442169239977e-06, + "loss": 0.303, + "step": 11073 + }, + { + "epoch": 0.67, + "learning_rate": 5.216723786431566e-06, + "loss": 0.2853, + "step": 11074 + }, + { + "epoch": 0.67, + "learning_rate": 5.21500558674936e-06, + "loss": 0.2956, + "step": 11075 + }, + { + "epoch": 0.67, + "learning_rate": 5.2132875702591355e-06, + "loss": 0.2818, + "step": 11076 + }, + { + "epoch": 0.67, + "learning_rate": 5.211569737026666e-06, + "loss": 0.3133, + "step": 11077 + }, + { + "epoch": 0.67, + "learning_rate": 5.209852087117718e-06, + "loss": 0.2826, + "step": 11078 + }, + { + "epoch": 0.67, + "learning_rate": 5.208134620598048e-06, + "loss": 0.2765, + "step": 11079 + }, + { + "epoch": 0.67, + "learning_rate": 5.206417337533416e-06, + "loss": 0.2938, + "step": 11080 + }, + { + "epoch": 0.67, + "learning_rate": 5.204700237989564e-06, + "loss": 0.2719, + "step": 11081 + }, + { + "epoch": 0.67, + "learning_rate": 5.2029833220322316e-06, + "loss": 0.3078, + "step": 11082 + }, + { + "epoch": 0.67, + "learning_rate": 5.201266589727148e-06, + "loss": 0.2896, + "step": 11083 + }, + { + "epoch": 0.67, + "learning_rate": 5.199550041140034e-06, + "loss": 0.2999, + "step": 11084 + }, + { + "epoch": 0.67, + "learning_rate": 5.197833676336617e-06, + "loss": 0.2909, + "step": 11085 + }, + { + "epoch": 0.67, + "learning_rate": 5.196117495382602e-06, + "loss": 0.3055, + "step": 11086 + }, + { + "epoch": 0.67, + "learning_rate": 5.194401498343693e-06, + "loss": 0.2744, + "step": 11087 + }, + { + "epoch": 0.67, + "learning_rate": 5.192685685285588e-06, + "loss": 0.296, + "step": 11088 + }, + { + "epoch": 0.67, + "learning_rate": 5.1909700562739725e-06, + "loss": 0.3131, + "step": 11089 + }, + { + "epoch": 0.67, + "learning_rate": 5.189254611374528e-06, + "loss": 0.2482, + "step": 11090 + }, + { + "epoch": 0.67, + "learning_rate": 5.187539350652937e-06, + "loss": 0.29, + "step": 11091 + }, + { + "epoch": 0.67, + "learning_rate": 5.185824274174864e-06, + "loss": 0.2845, + "step": 11092 + }, + { + "epoch": 0.67, + "learning_rate": 5.184109382005969e-06, + "loss": 0.2786, + "step": 11093 + }, + { + "epoch": 0.67, + "learning_rate": 5.182394674211906e-06, + "loss": 0.2769, + "step": 11094 + }, + { + "epoch": 0.67, + "learning_rate": 5.1806801508583235e-06, + "loss": 0.3221, + "step": 11095 + }, + { + "epoch": 0.67, + "learning_rate": 5.178965812010861e-06, + "loss": 0.2646, + "step": 11096 + }, + { + "epoch": 0.67, + "learning_rate": 5.177251657735152e-06, + "loss": 0.2835, + "step": 11097 + }, + { + "epoch": 0.67, + "learning_rate": 5.17553768809682e-06, + "loss": 0.2977, + "step": 11098 + }, + { + "epoch": 0.67, + "learning_rate": 5.173823903161481e-06, + "loss": 0.2799, + "step": 11099 + }, + { + "epoch": 0.67, + "learning_rate": 5.1721103029947575e-06, + "loss": 0.3075, + "step": 11100 + }, + { + "epoch": 0.67, + "learning_rate": 5.170396887662246e-06, + "loss": 0.3072, + "step": 11101 + }, + { + "epoch": 0.67, + "learning_rate": 5.168683657229545e-06, + "loss": 0.2917, + "step": 11102 + }, + { + "epoch": 0.67, + "learning_rate": 5.1669706117622475e-06, + "loss": 0.2974, + "step": 11103 + }, + { + "epoch": 0.67, + "learning_rate": 5.165257751325929e-06, + "loss": 0.2972, + "step": 11104 + }, + { + "epoch": 0.67, + "learning_rate": 5.163545075986178e-06, + "loss": 0.2831, + "step": 11105 + }, + { + "epoch": 0.67, + "learning_rate": 5.1618325858085575e-06, + "loss": 0.2874, + "step": 11106 + }, + { + "epoch": 0.67, + "learning_rate": 5.160120280858628e-06, + "loss": 0.2849, + "step": 11107 + }, + { + "epoch": 0.67, + "learning_rate": 5.158408161201948e-06, + "loss": 0.2804, + "step": 11108 + }, + { + "epoch": 0.67, + "learning_rate": 5.156696226904059e-06, + "loss": 0.275, + "step": 11109 + }, + { + "epoch": 0.67, + "learning_rate": 5.1549844780305106e-06, + "loss": 0.2757, + "step": 11110 + }, + { + "epoch": 0.67, + "learning_rate": 5.153272914646832e-06, + "loss": 0.2884, + "step": 11111 + }, + { + "epoch": 0.67, + "learning_rate": 5.1515615368185525e-06, + "loss": 0.272, + "step": 11112 + }, + { + "epoch": 0.67, + "learning_rate": 5.149850344611187e-06, + "loss": 0.3069, + "step": 11113 + }, + { + "epoch": 0.67, + "learning_rate": 5.148139338090249e-06, + "loss": 0.2882, + "step": 11114 + }, + { + "epoch": 0.67, + "learning_rate": 5.146428517321248e-06, + "loss": 0.2748, + "step": 11115 + }, + { + "epoch": 0.67, + "learning_rate": 5.144717882369679e-06, + "loss": 0.2803, + "step": 11116 + }, + { + "epoch": 0.67, + "learning_rate": 5.1430074333010346e-06, + "loss": 0.2765, + "step": 11117 + }, + { + "epoch": 0.67, + "learning_rate": 5.141297170180799e-06, + "loss": 0.323, + "step": 11118 + }, + { + "epoch": 0.67, + "learning_rate": 5.139587093074443e-06, + "loss": 0.2928, + "step": 11119 + }, + { + "epoch": 0.67, + "learning_rate": 5.137877202047445e-06, + "loss": 0.2709, + "step": 11120 + }, + { + "epoch": 0.67, + "learning_rate": 5.136167497165266e-06, + "loss": 0.2613, + "step": 11121 + }, + { + "epoch": 0.67, + "learning_rate": 5.134457978493356e-06, + "loss": 0.2953, + "step": 11122 + }, + { + "epoch": 0.67, + "learning_rate": 5.1327486460971696e-06, + "loss": 0.2811, + "step": 11123 + }, + { + "epoch": 0.67, + "learning_rate": 5.1310395000421435e-06, + "loss": 0.2938, + "step": 11124 + }, + { + "epoch": 0.67, + "learning_rate": 5.129330540393711e-06, + "loss": 0.3147, + "step": 11125 + }, + { + "epoch": 0.67, + "learning_rate": 5.127621767217306e-06, + "loss": 0.3022, + "step": 11126 + }, + { + "epoch": 0.67, + "learning_rate": 5.1259131805783425e-06, + "loss": 0.3032, + "step": 11127 + }, + { + "epoch": 0.67, + "learning_rate": 5.124204780542239e-06, + "loss": 0.2806, + "step": 11128 + }, + { + "epoch": 0.67, + "learning_rate": 5.12249656717439e-06, + "loss": 0.3161, + "step": 11129 + }, + { + "epoch": 0.67, + "learning_rate": 5.120788540540203e-06, + "loss": 0.2729, + "step": 11130 + }, + { + "epoch": 0.67, + "learning_rate": 5.119080700705068e-06, + "loss": 0.256, + "step": 11131 + }, + { + "epoch": 0.67, + "learning_rate": 5.117373047734369e-06, + "loss": 0.3287, + "step": 11132 + }, + { + "epoch": 0.67, + "learning_rate": 5.115665581693481e-06, + "loss": 0.3012, + "step": 11133 + }, + { + "epoch": 0.67, + "learning_rate": 5.11395830264777e-06, + "loss": 0.2842, + "step": 11134 + }, + { + "epoch": 0.67, + "learning_rate": 5.1122512106626085e-06, + "loss": 0.2718, + "step": 11135 + }, + { + "epoch": 0.67, + "learning_rate": 5.110544305803347e-06, + "loss": 0.298, + "step": 11136 + }, + { + "epoch": 0.67, + "learning_rate": 5.108837588135333e-06, + "loss": 0.2923, + "step": 11137 + }, + { + "epoch": 0.67, + "learning_rate": 5.107131057723908e-06, + "loss": 0.2653, + "step": 11138 + }, + { + "epoch": 0.67, + "learning_rate": 5.105424714634404e-06, + "loss": 0.2843, + "step": 11139 + }, + { + "epoch": 0.67, + "learning_rate": 5.103718558932151e-06, + "loss": 0.2969, + "step": 11140 + }, + { + "epoch": 0.67, + "learning_rate": 5.10201259068247e-06, + "loss": 0.2944, + "step": 11141 + }, + { + "epoch": 0.67, + "learning_rate": 5.100306809950671e-06, + "loss": 0.3068, + "step": 11142 + }, + { + "epoch": 0.67, + "learning_rate": 5.09860121680206e-06, + "loss": 0.2659, + "step": 11143 + }, + { + "epoch": 0.67, + "learning_rate": 5.0968958113019296e-06, + "loss": 0.2499, + "step": 11144 + }, + { + "epoch": 0.67, + "learning_rate": 5.09519059351558e-06, + "loss": 0.2799, + "step": 11145 + }, + { + "epoch": 0.67, + "learning_rate": 5.09348556350829e-06, + "loss": 0.2732, + "step": 11146 + }, + { + "epoch": 0.67, + "learning_rate": 5.091780721345336e-06, + "loss": 0.3013, + "step": 11147 + }, + { + "epoch": 0.67, + "learning_rate": 5.090076067091991e-06, + "loss": 0.2796, + "step": 11148 + }, + { + "epoch": 0.67, + "learning_rate": 5.088371600813508e-06, + "loss": 0.278, + "step": 11149 + }, + { + "epoch": 0.67, + "learning_rate": 5.086667322575152e-06, + "loss": 0.2826, + "step": 11150 + }, + { + "epoch": 0.67, + "learning_rate": 5.084963232442167e-06, + "loss": 0.2854, + "step": 11151 + }, + { + "epoch": 0.67, + "learning_rate": 5.083259330479794e-06, + "loss": 0.2871, + "step": 11152 + }, + { + "epoch": 0.67, + "learning_rate": 5.081555616753264e-06, + "loss": 0.2733, + "step": 11153 + }, + { + "epoch": 0.67, + "learning_rate": 5.079852091327804e-06, + "loss": 0.3011, + "step": 11154 + }, + { + "epoch": 0.67, + "learning_rate": 5.078148754268635e-06, + "loss": 0.2679, + "step": 11155 + }, + { + "epoch": 0.67, + "learning_rate": 5.07644560564097e-06, + "loss": 0.2698, + "step": 11156 + }, + { + "epoch": 0.67, + "learning_rate": 5.074742645510009e-06, + "loss": 0.2847, + "step": 11157 + }, + { + "epoch": 0.67, + "learning_rate": 5.073039873940952e-06, + "loss": 0.2956, + "step": 11158 + }, + { + "epoch": 0.67, + "learning_rate": 5.071337290998988e-06, + "loss": 0.3111, + "step": 11159 + }, + { + "epoch": 0.67, + "learning_rate": 5.069634896749296e-06, + "loss": 0.284, + "step": 11160 + }, + { + "epoch": 0.67, + "learning_rate": 5.0679326912570604e-06, + "loss": 0.278, + "step": 11161 + }, + { + "epoch": 0.67, + "learning_rate": 5.066230674587449e-06, + "loss": 0.2697, + "step": 11162 + }, + { + "epoch": 0.67, + "learning_rate": 5.064528846805614e-06, + "loss": 0.2905, + "step": 11163 + }, + { + "epoch": 0.67, + "learning_rate": 5.062827207976713e-06, + "loss": 0.3198, + "step": 11164 + }, + { + "epoch": 0.67, + "learning_rate": 5.061125758165896e-06, + "loss": 0.2708, + "step": 11165 + }, + { + "epoch": 0.67, + "learning_rate": 5.0594244974383e-06, + "loss": 0.2882, + "step": 11166 + }, + { + "epoch": 0.67, + "learning_rate": 5.057723425859058e-06, + "loss": 0.3055, + "step": 11167 + }, + { + "epoch": 0.67, + "learning_rate": 5.0560225434932956e-06, + "loss": 0.2907, + "step": 11168 + }, + { + "epoch": 0.67, + "learning_rate": 5.0543218504061255e-06, + "loss": 0.2794, + "step": 11169 + }, + { + "epoch": 0.67, + "learning_rate": 5.052621346662666e-06, + "loss": 0.2722, + "step": 11170 + }, + { + "epoch": 0.67, + "learning_rate": 5.050921032328016e-06, + "loss": 0.2645, + "step": 11171 + }, + { + "epoch": 0.67, + "learning_rate": 5.049220907467274e-06, + "loss": 0.3085, + "step": 11172 + }, + { + "epoch": 0.67, + "learning_rate": 5.047520972145525e-06, + "loss": 0.2543, + "step": 11173 + }, + { + "epoch": 0.68, + "learning_rate": 5.04582122642785e-06, + "loss": 0.3153, + "step": 11174 + }, + { + "epoch": 0.68, + "learning_rate": 5.044121670379327e-06, + "loss": 0.2697, + "step": 11175 + }, + { + "epoch": 0.68, + "learning_rate": 5.042422304065025e-06, + "loss": 0.2754, + "step": 11176 + }, + { + "epoch": 0.68, + "learning_rate": 5.0407231275499976e-06, + "loss": 0.3039, + "step": 11177 + }, + { + "epoch": 0.68, + "learning_rate": 5.0390241408993e-06, + "loss": 0.2981, + "step": 11178 + }, + { + "epoch": 0.68, + "learning_rate": 5.037325344177975e-06, + "loss": 0.3172, + "step": 11179 + }, + { + "epoch": 0.68, + "learning_rate": 5.035626737451065e-06, + "loss": 0.2764, + "step": 11180 + }, + { + "epoch": 0.68, + "learning_rate": 5.0339283207836e-06, + "loss": 0.2773, + "step": 11181 + }, + { + "epoch": 0.68, + "learning_rate": 5.0322300942406e-06, + "loss": 0.2715, + "step": 11182 + }, + { + "epoch": 0.68, + "learning_rate": 5.030532057887083e-06, + "loss": 0.2817, + "step": 11183 + }, + { + "epoch": 0.68, + "learning_rate": 5.0288342117880555e-06, + "loss": 0.2705, + "step": 11184 + }, + { + "epoch": 0.68, + "learning_rate": 5.027136556008523e-06, + "loss": 0.2893, + "step": 11185 + }, + { + "epoch": 0.68, + "learning_rate": 5.0254390906134775e-06, + "loss": 0.3008, + "step": 11186 + }, + { + "epoch": 0.68, + "learning_rate": 5.023741815667907e-06, + "loss": 0.2998, + "step": 11187 + }, + { + "epoch": 0.68, + "learning_rate": 5.022044731236791e-06, + "loss": 0.2882, + "step": 11188 + }, + { + "epoch": 0.68, + "learning_rate": 5.020347837385096e-06, + "loss": 0.2917, + "step": 11189 + }, + { + "epoch": 0.68, + "learning_rate": 5.018651134177796e-06, + "loss": 0.2578, + "step": 11190 + }, + { + "epoch": 0.68, + "learning_rate": 5.016954621679846e-06, + "loss": 0.2828, + "step": 11191 + }, + { + "epoch": 0.68, + "learning_rate": 5.015258299956194e-06, + "loss": 0.3111, + "step": 11192 + }, + { + "epoch": 0.68, + "learning_rate": 5.013562169071785e-06, + "loss": 0.2875, + "step": 11193 + }, + { + "epoch": 0.68, + "learning_rate": 5.011866229091555e-06, + "loss": 0.2633, + "step": 11194 + }, + { + "epoch": 0.68, + "learning_rate": 5.010170480080427e-06, + "loss": 0.2621, + "step": 11195 + }, + { + "epoch": 0.68, + "learning_rate": 5.0084749221033345e-06, + "loss": 0.2829, + "step": 11196 + }, + { + "epoch": 0.68, + "learning_rate": 5.006779555225181e-06, + "loss": 0.3142, + "step": 11197 + }, + { + "epoch": 0.68, + "learning_rate": 5.005084379510876e-06, + "loss": 0.2759, + "step": 11198 + }, + { + "epoch": 0.68, + "learning_rate": 5.003389395025314e-06, + "loss": 0.3033, + "step": 11199 + }, + { + "epoch": 0.68, + "learning_rate": 5.001694601833396e-06, + "loss": 0.2972, + "step": 11200 + }, + { + "epoch": 0.68, + "learning_rate": 5.000000000000003e-06, + "loss": 0.2829, + "step": 11201 + }, + { + "epoch": 0.68, + "learning_rate": 4.99830558959001e-06, + "loss": 0.2973, + "step": 11202 + }, + { + "epoch": 0.68, + "learning_rate": 4.99661137066829e-06, + "loss": 0.2885, + "step": 11203 + }, + { + "epoch": 0.68, + "learning_rate": 4.9949173432997e-06, + "loss": 0.2981, + "step": 11204 + }, + { + "epoch": 0.68, + "learning_rate": 4.993223507549103e-06, + "loss": 0.2708, + "step": 11205 + }, + { + "epoch": 0.68, + "learning_rate": 4.991529863481343e-06, + "loss": 0.2906, + "step": 11206 + }, + { + "epoch": 0.68, + "learning_rate": 4.989836411161262e-06, + "loss": 0.2722, + "step": 11207 + }, + { + "epoch": 0.68, + "learning_rate": 4.9881431506536915e-06, + "loss": 0.3114, + "step": 11208 + }, + { + "epoch": 0.68, + "learning_rate": 4.986450082023455e-06, + "loss": 0.2924, + "step": 11209 + }, + { + "epoch": 0.68, + "learning_rate": 4.984757205335379e-06, + "loss": 0.2909, + "step": 11210 + }, + { + "epoch": 0.68, + "learning_rate": 4.983064520654269e-06, + "loss": 0.3002, + "step": 11211 + }, + { + "epoch": 0.68, + "learning_rate": 4.981372028044929e-06, + "loss": 0.2812, + "step": 11212 + }, + { + "epoch": 0.68, + "learning_rate": 4.979679727572159e-06, + "loss": 0.296, + "step": 11213 + }, + { + "epoch": 0.68, + "learning_rate": 4.9779876193007404e-06, + "loss": 0.2708, + "step": 11214 + }, + { + "epoch": 0.68, + "learning_rate": 4.976295703295465e-06, + "loss": 0.2765, + "step": 11215 + }, + { + "epoch": 0.68, + "learning_rate": 4.974603979621102e-06, + "loss": 0.2969, + "step": 11216 + }, + { + "epoch": 0.68, + "learning_rate": 4.9729124483424195e-06, + "loss": 0.2753, + "step": 11217 + }, + { + "epoch": 0.68, + "learning_rate": 4.9712211095241755e-06, + "loss": 0.2812, + "step": 11218 + }, + { + "epoch": 0.68, + "learning_rate": 4.969529963231122e-06, + "loss": 0.2899, + "step": 11219 + }, + { + "epoch": 0.68, + "learning_rate": 4.967839009528009e-06, + "loss": 0.3073, + "step": 11220 + }, + { + "epoch": 0.68, + "learning_rate": 4.96614824847957e-06, + "loss": 0.267, + "step": 11221 + }, + { + "epoch": 0.68, + "learning_rate": 4.964457680150536e-06, + "loss": 0.3242, + "step": 11222 + }, + { + "epoch": 0.68, + "learning_rate": 4.96276730460563e-06, + "loss": 0.272, + "step": 11223 + }, + { + "epoch": 0.68, + "learning_rate": 4.9610771219095645e-06, + "loss": 0.2786, + "step": 11224 + }, + { + "epoch": 0.68, + "learning_rate": 4.959387132127054e-06, + "loss": 0.2824, + "step": 11225 + }, + { + "epoch": 0.68, + "learning_rate": 4.957697335322795e-06, + "loss": 0.3209, + "step": 11226 + }, + { + "epoch": 0.68, + "learning_rate": 4.956007731561483e-06, + "loss": 0.2872, + "step": 11227 + }, + { + "epoch": 0.68, + "learning_rate": 4.954318320907801e-06, + "loss": 0.2994, + "step": 11228 + }, + { + "epoch": 0.68, + "learning_rate": 4.952629103426426e-06, + "loss": 0.2916, + "step": 11229 + }, + { + "epoch": 0.68, + "learning_rate": 4.950940079182041e-06, + "loss": 0.2718, + "step": 11230 + }, + { + "epoch": 0.68, + "learning_rate": 4.949251248239295e-06, + "loss": 0.3069, + "step": 11231 + }, + { + "epoch": 0.68, + "learning_rate": 4.947562610662852e-06, + "loss": 0.2779, + "step": 11232 + }, + { + "epoch": 0.68, + "learning_rate": 4.9458741665173584e-06, + "loss": 0.3052, + "step": 11233 + }, + { + "epoch": 0.68, + "learning_rate": 4.944185915867453e-06, + "loss": 0.2757, + "step": 11234 + }, + { + "epoch": 0.68, + "learning_rate": 4.942497858777777e-06, + "loss": 0.3083, + "step": 11235 + }, + { + "epoch": 0.68, + "learning_rate": 4.940809995312954e-06, + "loss": 0.2727, + "step": 11236 + }, + { + "epoch": 0.68, + "learning_rate": 4.9391223255376045e-06, + "loss": 0.2946, + "step": 11237 + }, + { + "epoch": 0.68, + "learning_rate": 4.937434849516337e-06, + "loss": 0.2899, + "step": 11238 + }, + { + "epoch": 0.68, + "learning_rate": 4.935747567313755e-06, + "loss": 0.2793, + "step": 11239 + }, + { + "epoch": 0.68, + "learning_rate": 4.9340604789944615e-06, + "loss": 0.2945, + "step": 11240 + }, + { + "epoch": 0.68, + "learning_rate": 4.932373584623044e-06, + "loss": 0.2796, + "step": 11241 + }, + { + "epoch": 0.68, + "learning_rate": 4.930686884264082e-06, + "loss": 0.299, + "step": 11242 + }, + { + "epoch": 0.68, + "learning_rate": 4.929000377982154e-06, + "loss": 0.2941, + "step": 11243 + }, + { + "epoch": 0.68, + "learning_rate": 4.92731406584182e-06, + "loss": 0.2803, + "step": 11244 + }, + { + "epoch": 0.68, + "learning_rate": 4.925627947907651e-06, + "loss": 0.2861, + "step": 11245 + }, + { + "epoch": 0.68, + "learning_rate": 4.923942024244192e-06, + "loss": 0.2852, + "step": 11246 + }, + { + "epoch": 0.68, + "learning_rate": 4.922256294915991e-06, + "loss": 0.2928, + "step": 11247 + }, + { + "epoch": 0.68, + "learning_rate": 4.920570759987584e-06, + "loss": 0.2988, + "step": 11248 + }, + { + "epoch": 0.68, + "learning_rate": 4.9188854195235e-06, + "loss": 0.2644, + "step": 11249 + }, + { + "epoch": 0.68, + "learning_rate": 4.917200273588264e-06, + "loss": 0.257, + "step": 11250 + }, + { + "epoch": 0.68, + "learning_rate": 4.915515322246393e-06, + "loss": 0.2909, + "step": 11251 + }, + { + "epoch": 0.68, + "learning_rate": 4.913830565562391e-06, + "loss": 0.2818, + "step": 11252 + }, + { + "epoch": 0.68, + "learning_rate": 4.91214600360076e-06, + "loss": 0.2991, + "step": 11253 + }, + { + "epoch": 0.68, + "learning_rate": 4.91046163642599e-06, + "loss": 0.2859, + "step": 11254 + }, + { + "epoch": 0.68, + "learning_rate": 4.908777464102571e-06, + "loss": 0.3139, + "step": 11255 + }, + { + "epoch": 0.68, + "learning_rate": 4.907093486694981e-06, + "loss": 0.3061, + "step": 11256 + }, + { + "epoch": 0.68, + "learning_rate": 4.905409704267689e-06, + "loss": 0.2868, + "step": 11257 + }, + { + "epoch": 0.68, + "learning_rate": 4.903726116885156e-06, + "loss": 0.2941, + "step": 11258 + }, + { + "epoch": 0.68, + "learning_rate": 4.902042724611837e-06, + "loss": 0.2956, + "step": 11259 + }, + { + "epoch": 0.68, + "learning_rate": 4.900359527512186e-06, + "loss": 0.2642, + "step": 11260 + }, + { + "epoch": 0.68, + "learning_rate": 4.8986765256506395e-06, + "loss": 0.2727, + "step": 11261 + }, + { + "epoch": 0.68, + "learning_rate": 4.8969937190916325e-06, + "loss": 0.2935, + "step": 11262 + }, + { + "epoch": 0.68, + "learning_rate": 4.89531110789959e-06, + "loss": 0.2847, + "step": 11263 + }, + { + "epoch": 0.68, + "learning_rate": 4.893628692138929e-06, + "loss": 0.2988, + "step": 11264 + }, + { + "epoch": 0.68, + "learning_rate": 4.891946471874062e-06, + "loss": 0.2656, + "step": 11265 + }, + { + "epoch": 0.68, + "learning_rate": 4.89026444716939e-06, + "loss": 0.2776, + "step": 11266 + }, + { + "epoch": 0.68, + "learning_rate": 4.8885826180893115e-06, + "loss": 0.2697, + "step": 11267 + }, + { + "epoch": 0.68, + "learning_rate": 4.886900984698212e-06, + "loss": 0.2905, + "step": 11268 + }, + { + "epoch": 0.68, + "learning_rate": 4.8852195470604715e-06, + "loss": 0.3019, + "step": 11269 + }, + { + "epoch": 0.68, + "learning_rate": 4.883538305240469e-06, + "loss": 0.2946, + "step": 11270 + }, + { + "epoch": 0.68, + "learning_rate": 4.881857259302566e-06, + "loss": 0.2866, + "step": 11271 + }, + { + "epoch": 0.68, + "learning_rate": 4.880176409311122e-06, + "loss": 0.2769, + "step": 11272 + }, + { + "epoch": 0.68, + "learning_rate": 4.878495755330488e-06, + "loss": 0.2884, + "step": 11273 + }, + { + "epoch": 0.68, + "learning_rate": 4.876815297425002e-06, + "loss": 0.2896, + "step": 11274 + }, + { + "epoch": 0.68, + "learning_rate": 4.875135035659009e-06, + "loss": 0.2782, + "step": 11275 + }, + { + "epoch": 0.68, + "learning_rate": 4.873454970096833e-06, + "loss": 0.285, + "step": 11276 + }, + { + "epoch": 0.68, + "learning_rate": 4.871775100802794e-06, + "loss": 0.2755, + "step": 11277 + }, + { + "epoch": 0.68, + "learning_rate": 4.870095427841206e-06, + "loss": 0.2716, + "step": 11278 + }, + { + "epoch": 0.68, + "learning_rate": 4.8684159512763705e-06, + "loss": 0.2997, + "step": 11279 + }, + { + "epoch": 0.68, + "learning_rate": 4.866736671172595e-06, + "loss": 0.2773, + "step": 11280 + }, + { + "epoch": 0.68, + "learning_rate": 4.865057587594164e-06, + "loss": 0.2988, + "step": 11281 + }, + { + "epoch": 0.68, + "learning_rate": 4.863378700605363e-06, + "loss": 0.2762, + "step": 11282 + }, + { + "epoch": 0.68, + "learning_rate": 4.861700010270464e-06, + "loss": 0.2766, + "step": 11283 + }, + { + "epoch": 0.68, + "learning_rate": 4.860021516653735e-06, + "loss": 0.2899, + "step": 11284 + }, + { + "epoch": 0.68, + "learning_rate": 4.8583432198194424e-06, + "loss": 0.3002, + "step": 11285 + }, + { + "epoch": 0.68, + "learning_rate": 4.856665119831837e-06, + "loss": 0.2593, + "step": 11286 + }, + { + "epoch": 0.68, + "learning_rate": 4.854987216755162e-06, + "loss": 0.2663, + "step": 11287 + }, + { + "epoch": 0.68, + "learning_rate": 4.853309510653658e-06, + "loss": 0.2927, + "step": 11288 + }, + { + "epoch": 0.68, + "learning_rate": 4.85163200159155e-06, + "loss": 0.2719, + "step": 11289 + }, + { + "epoch": 0.68, + "learning_rate": 4.849954689633068e-06, + "loss": 0.2893, + "step": 11290 + }, + { + "epoch": 0.68, + "learning_rate": 4.848277574842425e-06, + "loss": 0.3174, + "step": 11291 + }, + { + "epoch": 0.68, + "learning_rate": 4.846600657283828e-06, + "loss": 0.2866, + "step": 11292 + }, + { + "epoch": 0.68, + "learning_rate": 4.844923937021476e-06, + "loss": 0.3097, + "step": 11293 + }, + { + "epoch": 0.68, + "learning_rate": 4.84324741411956e-06, + "loss": 0.2759, + "step": 11294 + }, + { + "epoch": 0.68, + "learning_rate": 4.841571088642274e-06, + "loss": 0.2907, + "step": 11295 + }, + { + "epoch": 0.68, + "learning_rate": 4.839894960653788e-06, + "loss": 0.2768, + "step": 11296 + }, + { + "epoch": 0.68, + "learning_rate": 4.838219030218274e-06, + "loss": 0.3052, + "step": 11297 + }, + { + "epoch": 0.68, + "learning_rate": 4.836543297399895e-06, + "loss": 0.274, + "step": 11298 + }, + { + "epoch": 0.68, + "learning_rate": 4.8348677622628045e-06, + "loss": 0.2701, + "step": 11299 + }, + { + "epoch": 0.68, + "learning_rate": 4.8331924248711516e-06, + "loss": 0.2804, + "step": 11300 + }, + { + "epoch": 0.68, + "learning_rate": 4.831517285289074e-06, + "loss": 0.3093, + "step": 11301 + }, + { + "epoch": 0.68, + "learning_rate": 4.829842343580704e-06, + "loss": 0.2465, + "step": 11302 + }, + { + "epoch": 0.68, + "learning_rate": 4.8281675998101655e-06, + "loss": 0.274, + "step": 11303 + }, + { + "epoch": 0.68, + "learning_rate": 4.826493054041579e-06, + "loss": 0.3013, + "step": 11304 + }, + { + "epoch": 0.68, + "learning_rate": 4.824818706339053e-06, + "loss": 0.2979, + "step": 11305 + }, + { + "epoch": 0.68, + "learning_rate": 4.8231445567666875e-06, + "loss": 0.2762, + "step": 11306 + }, + { + "epoch": 0.68, + "learning_rate": 4.821470605388578e-06, + "loss": 0.2985, + "step": 11307 + }, + { + "epoch": 0.68, + "learning_rate": 4.819796852268812e-06, + "loss": 0.2801, + "step": 11308 + }, + { + "epoch": 0.68, + "learning_rate": 4.818123297471463e-06, + "loss": 0.2969, + "step": 11309 + }, + { + "epoch": 0.68, + "learning_rate": 4.816449941060611e-06, + "loss": 0.2507, + "step": 11310 + }, + { + "epoch": 0.68, + "learning_rate": 4.814776783100316e-06, + "loss": 0.2846, + "step": 11311 + }, + { + "epoch": 0.68, + "learning_rate": 4.813103823654633e-06, + "loss": 0.3033, + "step": 11312 + }, + { + "epoch": 0.68, + "learning_rate": 4.811431062787614e-06, + "loss": 0.2775, + "step": 11313 + }, + { + "epoch": 0.68, + "learning_rate": 4.809758500563293e-06, + "loss": 0.2746, + "step": 11314 + }, + { + "epoch": 0.68, + "learning_rate": 4.8080861370457124e-06, + "loss": 0.3125, + "step": 11315 + }, + { + "epoch": 0.68, + "learning_rate": 4.806413972298895e-06, + "loss": 0.2697, + "step": 11316 + }, + { + "epoch": 0.68, + "learning_rate": 4.804742006386858e-06, + "loss": 0.3103, + "step": 11317 + }, + { + "epoch": 0.68, + "learning_rate": 4.803070239373612e-06, + "loss": 0.2426, + "step": 11318 + }, + { + "epoch": 0.68, + "learning_rate": 4.801398671323156e-06, + "loss": 0.2759, + "step": 11319 + }, + { + "epoch": 0.68, + "learning_rate": 4.799727302299495e-06, + "loss": 0.2732, + "step": 11320 + }, + { + "epoch": 0.68, + "learning_rate": 4.798056132366612e-06, + "loss": 0.2965, + "step": 11321 + }, + { + "epoch": 0.68, + "learning_rate": 4.796385161588487e-06, + "loss": 0.2984, + "step": 11322 + }, + { + "epoch": 0.68, + "learning_rate": 4.794714390029091e-06, + "loss": 0.2875, + "step": 11323 + }, + { + "epoch": 0.68, + "learning_rate": 4.7930438177523886e-06, + "loss": 0.2736, + "step": 11324 + }, + { + "epoch": 0.68, + "learning_rate": 4.791373444822342e-06, + "loss": 0.2923, + "step": 11325 + }, + { + "epoch": 0.68, + "learning_rate": 4.789703271302899e-06, + "loss": 0.2787, + "step": 11326 + }, + { + "epoch": 0.68, + "learning_rate": 4.788033297257999e-06, + "loss": 0.308, + "step": 11327 + }, + { + "epoch": 0.68, + "learning_rate": 4.786363522751579e-06, + "loss": 0.2941, + "step": 11328 + }, + { + "epoch": 0.68, + "learning_rate": 4.784693947847562e-06, + "loss": 0.2778, + "step": 11329 + }, + { + "epoch": 0.68, + "learning_rate": 4.783024572609873e-06, + "loss": 0.2955, + "step": 11330 + }, + { + "epoch": 0.68, + "learning_rate": 4.781355397102419e-06, + "loss": 0.2923, + "step": 11331 + }, + { + "epoch": 0.68, + "learning_rate": 4.77968642138911e-06, + "loss": 0.2825, + "step": 11332 + }, + { + "epoch": 0.68, + "learning_rate": 4.77801764553383e-06, + "loss": 0.2853, + "step": 11333 + }, + { + "epoch": 0.68, + "learning_rate": 4.77634906960048e-06, + "loss": 0.2666, + "step": 11334 + }, + { + "epoch": 0.68, + "learning_rate": 4.774680693652934e-06, + "loss": 0.2977, + "step": 11335 + }, + { + "epoch": 0.68, + "learning_rate": 4.773012517755069e-06, + "loss": 0.3025, + "step": 11336 + }, + { + "epoch": 0.68, + "learning_rate": 4.771344541970748e-06, + "loss": 0.296, + "step": 11337 + }, + { + "epoch": 0.68, + "learning_rate": 4.769676766363825e-06, + "loss": 0.2799, + "step": 11338 + }, + { + "epoch": 0.69, + "learning_rate": 4.768009190998161e-06, + "loss": 0.3133, + "step": 11339 + }, + { + "epoch": 0.69, + "learning_rate": 4.766341815937591e-06, + "loss": 0.2818, + "step": 11340 + }, + { + "epoch": 0.69, + "learning_rate": 4.764674641245952e-06, + "loss": 0.2988, + "step": 11341 + }, + { + "epoch": 0.69, + "learning_rate": 4.763007666987071e-06, + "loss": 0.291, + "step": 11342 + }, + { + "epoch": 0.69, + "learning_rate": 4.761340893224767e-06, + "loss": 0.2522, + "step": 11343 + }, + { + "epoch": 0.69, + "learning_rate": 4.7596743200228465e-06, + "loss": 0.2991, + "step": 11344 + }, + { + "epoch": 0.69, + "learning_rate": 4.758007947445126e-06, + "loss": 0.269, + "step": 11345 + }, + { + "epoch": 0.69, + "learning_rate": 4.7563417755553945e-06, + "loss": 0.2983, + "step": 11346 + }, + { + "epoch": 0.69, + "learning_rate": 4.754675804417441e-06, + "loss": 0.2734, + "step": 11347 + }, + { + "epoch": 0.69, + "learning_rate": 4.753010034095048e-06, + "loss": 0.28, + "step": 11348 + }, + { + "epoch": 0.69, + "learning_rate": 4.751344464651986e-06, + "loss": 0.2832, + "step": 11349 + }, + { + "epoch": 0.69, + "learning_rate": 4.7496790961520255e-06, + "loss": 0.2944, + "step": 11350 + }, + { + "epoch": 0.69, + "learning_rate": 4.748013928658922e-06, + "loss": 0.2949, + "step": 11351 + }, + { + "epoch": 0.69, + "learning_rate": 4.746348962236428e-06, + "loss": 0.2999, + "step": 11352 + }, + { + "epoch": 0.69, + "learning_rate": 4.744684196948283e-06, + "loss": 0.2769, + "step": 11353 + }, + { + "epoch": 0.69, + "learning_rate": 4.74301963285822e-06, + "loss": 0.2768, + "step": 11354 + }, + { + "epoch": 0.69, + "learning_rate": 4.741355270029975e-06, + "loss": 0.2697, + "step": 11355 + }, + { + "epoch": 0.69, + "learning_rate": 4.739691108527261e-06, + "loss": 0.2879, + "step": 11356 + }, + { + "epoch": 0.69, + "learning_rate": 4.738027148413792e-06, + "loss": 0.3186, + "step": 11357 + }, + { + "epoch": 0.69, + "learning_rate": 4.73636338975327e-06, + "loss": 0.2723, + "step": 11358 + }, + { + "epoch": 0.69, + "learning_rate": 4.734699832609391e-06, + "loss": 0.2719, + "step": 11359 + }, + { + "epoch": 0.69, + "learning_rate": 4.7330364770458485e-06, + "loss": 0.3002, + "step": 11360 + }, + { + "epoch": 0.69, + "learning_rate": 4.731373323126321e-06, + "loss": 0.2656, + "step": 11361 + }, + { + "epoch": 0.69, + "learning_rate": 4.729710370914481e-06, + "loss": 0.3097, + "step": 11362 + }, + { + "epoch": 0.69, + "learning_rate": 4.728047620473995e-06, + "loss": 0.2744, + "step": 11363 + }, + { + "epoch": 0.69, + "learning_rate": 4.726385071868518e-06, + "loss": 0.2836, + "step": 11364 + }, + { + "epoch": 0.69, + "learning_rate": 4.7247227251617045e-06, + "loss": 0.2893, + "step": 11365 + }, + { + "epoch": 0.69, + "learning_rate": 4.723060580417199e-06, + "loss": 0.2847, + "step": 11366 + }, + { + "epoch": 0.69, + "learning_rate": 4.72139863769863e-06, + "loss": 0.285, + "step": 11367 + }, + { + "epoch": 0.69, + "learning_rate": 4.719736897069622e-06, + "loss": 0.3089, + "step": 11368 + }, + { + "epoch": 0.69, + "learning_rate": 4.718075358593802e-06, + "loss": 0.3078, + "step": 11369 + }, + { + "epoch": 0.69, + "learning_rate": 4.71641402233478e-06, + "loss": 0.2626, + "step": 11370 + }, + { + "epoch": 0.69, + "learning_rate": 4.7147528883561565e-06, + "loss": 0.2992, + "step": 11371 + }, + { + "epoch": 0.69, + "learning_rate": 4.71309195672153e-06, + "loss": 0.2912, + "step": 11372 + }, + { + "epoch": 0.69, + "learning_rate": 4.711431227494483e-06, + "loss": 0.279, + "step": 11373 + }, + { + "epoch": 0.69, + "learning_rate": 4.709770700738606e-06, + "loss": 0.2932, + "step": 11374 + }, + { + "epoch": 0.69, + "learning_rate": 4.708110376517465e-06, + "loss": 0.2896, + "step": 11375 + }, + { + "epoch": 0.69, + "learning_rate": 4.706450254894628e-06, + "loss": 0.2866, + "step": 11376 + }, + { + "epoch": 0.69, + "learning_rate": 4.70479033593365e-06, + "loss": 0.2973, + "step": 11377 + }, + { + "epoch": 0.69, + "learning_rate": 4.703130619698082e-06, + "loss": 0.2905, + "step": 11378 + }, + { + "epoch": 0.69, + "learning_rate": 4.701471106251459e-06, + "loss": 0.2674, + "step": 11379 + }, + { + "epoch": 0.69, + "learning_rate": 4.699811795657327e-06, + "loss": 0.2747, + "step": 11380 + }, + { + "epoch": 0.69, + "learning_rate": 4.698152687979205e-06, + "loss": 0.2747, + "step": 11381 + }, + { + "epoch": 0.69, + "learning_rate": 4.6964937832806125e-06, + "loss": 0.2926, + "step": 11382 + }, + { + "epoch": 0.69, + "learning_rate": 4.6948350816250595e-06, + "loss": 0.2677, + "step": 11383 + }, + { + "epoch": 0.69, + "learning_rate": 4.693176583076047e-06, + "loss": 0.3239, + "step": 11384 + }, + { + "epoch": 0.69, + "learning_rate": 4.691518287697076e-06, + "loss": 0.2757, + "step": 11385 + }, + { + "epoch": 0.69, + "learning_rate": 4.68986019555163e-06, + "loss": 0.2872, + "step": 11386 + }, + { + "epoch": 0.69, + "learning_rate": 4.688202306703189e-06, + "loss": 0.2962, + "step": 11387 + }, + { + "epoch": 0.69, + "learning_rate": 4.686544621215224e-06, + "loss": 0.263, + "step": 11388 + }, + { + "epoch": 0.69, + "learning_rate": 4.684887139151196e-06, + "loss": 0.2835, + "step": 11389 + }, + { + "epoch": 0.69, + "learning_rate": 4.6832298605745685e-06, + "loss": 0.2575, + "step": 11390 + }, + { + "epoch": 0.69, + "learning_rate": 4.681572785548787e-06, + "loss": 0.2882, + "step": 11391 + }, + { + "epoch": 0.69, + "learning_rate": 4.679915914137291e-06, + "loss": 0.2642, + "step": 11392 + }, + { + "epoch": 0.69, + "learning_rate": 4.6782592464035124e-06, + "loss": 0.2651, + "step": 11393 + }, + { + "epoch": 0.69, + "learning_rate": 4.676602782410874e-06, + "loss": 0.2988, + "step": 11394 + }, + { + "epoch": 0.69, + "learning_rate": 4.674946522222801e-06, + "loss": 0.2667, + "step": 11395 + }, + { + "epoch": 0.69, + "learning_rate": 4.673290465902697e-06, + "loss": 0.2963, + "step": 11396 + }, + { + "epoch": 0.69, + "learning_rate": 4.671634613513965e-06, + "loss": 0.2642, + "step": 11397 + }, + { + "epoch": 0.69, + "learning_rate": 4.6699789651199975e-06, + "loss": 0.2871, + "step": 11398 + }, + { + "epoch": 0.69, + "learning_rate": 4.668323520784179e-06, + "loss": 0.2945, + "step": 11399 + }, + { + "epoch": 0.69, + "learning_rate": 4.666668280569897e-06, + "loss": 0.2472, + "step": 11400 + }, + { + "epoch": 0.69, + "learning_rate": 4.665013244540512e-06, + "loss": 0.2733, + "step": 11401 + }, + { + "epoch": 0.69, + "learning_rate": 4.663358412759389e-06, + "loss": 0.2853, + "step": 11402 + }, + { + "epoch": 0.69, + "learning_rate": 4.661703785289879e-06, + "loss": 0.2921, + "step": 11403 + }, + { + "epoch": 0.69, + "learning_rate": 4.660049362195337e-06, + "loss": 0.305, + "step": 11404 + }, + { + "epoch": 0.69, + "learning_rate": 4.6583951435390975e-06, + "loss": 0.3192, + "step": 11405 + }, + { + "epoch": 0.69, + "learning_rate": 4.656741129384493e-06, + "loss": 0.2758, + "step": 11406 + }, + { + "epoch": 0.69, + "learning_rate": 4.655087319794846e-06, + "loss": 0.2875, + "step": 11407 + }, + { + "epoch": 0.69, + "learning_rate": 4.653433714833469e-06, + "loss": 0.3079, + "step": 11408 + }, + { + "epoch": 0.69, + "learning_rate": 4.651780314563676e-06, + "loss": 0.2739, + "step": 11409 + }, + { + "epoch": 0.69, + "learning_rate": 4.650127119048764e-06, + "loss": 0.2702, + "step": 11410 + }, + { + "epoch": 0.69, + "learning_rate": 4.648474128352027e-06, + "loss": 0.2914, + "step": 11411 + }, + { + "epoch": 0.69, + "learning_rate": 4.646821342536745e-06, + "loss": 0.2987, + "step": 11412 + }, + { + "epoch": 0.69, + "learning_rate": 4.645168761666197e-06, + "loss": 0.2896, + "step": 11413 + }, + { + "epoch": 0.69, + "learning_rate": 4.643516385803647e-06, + "loss": 0.287, + "step": 11414 + }, + { + "epoch": 0.69, + "learning_rate": 4.641864215012364e-06, + "loss": 0.2932, + "step": 11415 + }, + { + "epoch": 0.69, + "learning_rate": 4.640212249355597e-06, + "loss": 0.272, + "step": 11416 + }, + { + "epoch": 0.69, + "learning_rate": 4.638560488896589e-06, + "loss": 0.2793, + "step": 11417 + }, + { + "epoch": 0.69, + "learning_rate": 4.636908933698581e-06, + "loss": 0.2821, + "step": 11418 + }, + { + "epoch": 0.69, + "learning_rate": 4.635257583824796e-06, + "loss": 0.2599, + "step": 11419 + }, + { + "epoch": 0.69, + "learning_rate": 4.633606439338462e-06, + "loss": 0.2979, + "step": 11420 + }, + { + "epoch": 0.69, + "learning_rate": 4.63195550030279e-06, + "loss": 0.2701, + "step": 11421 + }, + { + "epoch": 0.69, + "learning_rate": 4.6303047667809865e-06, + "loss": 0.2831, + "step": 11422 + }, + { + "epoch": 0.69, + "learning_rate": 4.628654238836247e-06, + "loss": 0.3037, + "step": 11423 + }, + { + "epoch": 0.69, + "learning_rate": 4.627003916531761e-06, + "loss": 0.2704, + "step": 11424 + }, + { + "epoch": 0.69, + "learning_rate": 4.625353799930715e-06, + "loss": 0.2916, + "step": 11425 + }, + { + "epoch": 0.69, + "learning_rate": 4.62370388909628e-06, + "loss": 0.2912, + "step": 11426 + }, + { + "epoch": 0.69, + "learning_rate": 4.622054184091624e-06, + "loss": 0.3025, + "step": 11427 + }, + { + "epoch": 0.69, + "learning_rate": 4.620404684979905e-06, + "loss": 0.2857, + "step": 11428 + }, + { + "epoch": 0.69, + "learning_rate": 4.6187553918242675e-06, + "loss": 0.2987, + "step": 11429 + }, + { + "epoch": 0.69, + "learning_rate": 4.617106304687864e-06, + "loss": 0.2783, + "step": 11430 + }, + { + "epoch": 0.69, + "learning_rate": 4.615457423633825e-06, + "loss": 0.2882, + "step": 11431 + }, + { + "epoch": 0.69, + "learning_rate": 4.613808748725277e-06, + "loss": 0.2785, + "step": 11432 + }, + { + "epoch": 0.69, + "learning_rate": 4.6121602800253386e-06, + "loss": 0.2425, + "step": 11433 + }, + { + "epoch": 0.69, + "learning_rate": 4.610512017597122e-06, + "loss": 0.2757, + "step": 11434 + }, + { + "epoch": 0.69, + "learning_rate": 4.60886396150373e-06, + "loss": 0.2768, + "step": 11435 + }, + { + "epoch": 0.69, + "learning_rate": 4.607216111808258e-06, + "loss": 0.2758, + "step": 11436 + }, + { + "epoch": 0.69, + "learning_rate": 4.605568468573792e-06, + "loss": 0.2854, + "step": 11437 + }, + { + "epoch": 0.69, + "learning_rate": 4.603921031863411e-06, + "loss": 0.3156, + "step": 11438 + }, + { + "epoch": 0.69, + "learning_rate": 4.60227380174019e-06, + "loss": 0.2504, + "step": 11439 + }, + { + "epoch": 0.69, + "learning_rate": 4.600626778267192e-06, + "loss": 0.3055, + "step": 11440 + }, + { + "epoch": 0.69, + "learning_rate": 4.598979961507472e-06, + "loss": 0.2937, + "step": 11441 + }, + { + "epoch": 0.69, + "learning_rate": 4.597333351524077e-06, + "loss": 0.2835, + "step": 11442 + }, + { + "epoch": 0.69, + "learning_rate": 4.595686948380044e-06, + "loss": 0.2916, + "step": 11443 + }, + { + "epoch": 0.69, + "learning_rate": 4.594040752138412e-06, + "loss": 0.2628, + "step": 11444 + }, + { + "epoch": 0.69, + "learning_rate": 4.592394762862201e-06, + "loss": 0.2856, + "step": 11445 + }, + { + "epoch": 0.69, + "learning_rate": 4.590748980614429e-06, + "loss": 0.2815, + "step": 11446 + }, + { + "epoch": 0.69, + "learning_rate": 4.589103405458103e-06, + "loss": 0.2736, + "step": 11447 + }, + { + "epoch": 0.69, + "learning_rate": 4.587458037456223e-06, + "loss": 0.2369, + "step": 11448 + }, + { + "epoch": 0.69, + "learning_rate": 4.585812876671778e-06, + "loss": 0.2539, + "step": 11449 + }, + { + "epoch": 0.69, + "learning_rate": 4.584167923167759e-06, + "loss": 0.28, + "step": 11450 + }, + { + "epoch": 0.69, + "learning_rate": 4.582523177007141e-06, + "loss": 0.299, + "step": 11451 + }, + { + "epoch": 0.69, + "learning_rate": 4.580878638252891e-06, + "loss": 0.3048, + "step": 11452 + }, + { + "epoch": 0.69, + "learning_rate": 4.5792343069679705e-06, + "loss": 0.3103, + "step": 11453 + }, + { + "epoch": 0.69, + "learning_rate": 4.577590183215327e-06, + "loss": 0.2854, + "step": 11454 + }, + { + "epoch": 0.69, + "learning_rate": 4.575946267057913e-06, + "loss": 0.2533, + "step": 11455 + }, + { + "epoch": 0.69, + "learning_rate": 4.5743025585586645e-06, + "loss": 0.2485, + "step": 11456 + }, + { + "epoch": 0.69, + "learning_rate": 4.5726590577805064e-06, + "loss": 0.3034, + "step": 11457 + }, + { + "epoch": 0.69, + "learning_rate": 4.571015764786362e-06, + "loss": 0.2754, + "step": 11458 + }, + { + "epoch": 0.69, + "learning_rate": 4.569372679639139e-06, + "loss": 0.293, + "step": 11459 + }, + { + "epoch": 0.69, + "learning_rate": 4.56772980240175e-06, + "loss": 0.2839, + "step": 11460 + }, + { + "epoch": 0.69, + "learning_rate": 4.566087133137091e-06, + "loss": 0.2637, + "step": 11461 + }, + { + "epoch": 0.69, + "learning_rate": 4.564444671908047e-06, + "loss": 0.298, + "step": 11462 + }, + { + "epoch": 0.69, + "learning_rate": 4.562802418777502e-06, + "loss": 0.2603, + "step": 11463 + }, + { + "epoch": 0.69, + "learning_rate": 4.561160373808324e-06, + "loss": 0.2938, + "step": 11464 + }, + { + "epoch": 0.69, + "learning_rate": 4.559518537063388e-06, + "loss": 0.2627, + "step": 11465 + }, + { + "epoch": 0.69, + "learning_rate": 4.557876908605543e-06, + "loss": 0.2762, + "step": 11466 + }, + { + "epoch": 0.69, + "learning_rate": 4.5562354884976425e-06, + "loss": 0.2733, + "step": 11467 + }, + { + "epoch": 0.69, + "learning_rate": 4.554594276802525e-06, + "loss": 0.29, + "step": 11468 + }, + { + "epoch": 0.69, + "learning_rate": 4.552953273583026e-06, + "loss": 0.2815, + "step": 11469 + }, + { + "epoch": 0.69, + "learning_rate": 4.5513124789019695e-06, + "loss": 0.287, + "step": 11470 + }, + { + "epoch": 0.69, + "learning_rate": 4.549671892822172e-06, + "loss": 0.3006, + "step": 11471 + }, + { + "epoch": 0.69, + "learning_rate": 4.548031515406443e-06, + "loss": 0.2797, + "step": 11472 + }, + { + "epoch": 0.69, + "learning_rate": 4.5463913467175825e-06, + "loss": 0.3058, + "step": 11473 + }, + { + "epoch": 0.69, + "learning_rate": 4.544751386818389e-06, + "loss": 0.3012, + "step": 11474 + }, + { + "epoch": 0.69, + "learning_rate": 4.543111635771644e-06, + "loss": 0.2755, + "step": 11475 + }, + { + "epoch": 0.69, + "learning_rate": 4.5414720936401265e-06, + "loss": 0.308, + "step": 11476 + }, + { + "epoch": 0.69, + "learning_rate": 4.5398327604866056e-06, + "loss": 0.2765, + "step": 11477 + }, + { + "epoch": 0.69, + "learning_rate": 4.5381936363738365e-06, + "loss": 0.2899, + "step": 11478 + }, + { + "epoch": 0.69, + "learning_rate": 4.536554721364584e-06, + "loss": 0.2662, + "step": 11479 + }, + { + "epoch": 0.69, + "learning_rate": 4.534916015521584e-06, + "loss": 0.2821, + "step": 11480 + }, + { + "epoch": 0.69, + "learning_rate": 4.533277518907579e-06, + "loss": 0.2803, + "step": 11481 + }, + { + "epoch": 0.69, + "learning_rate": 4.531639231585298e-06, + "loss": 0.3265, + "step": 11482 + }, + { + "epoch": 0.69, + "learning_rate": 4.530001153617459e-06, + "loss": 0.2715, + "step": 11483 + }, + { + "epoch": 0.69, + "learning_rate": 4.528363285066774e-06, + "loss": 0.2955, + "step": 11484 + }, + { + "epoch": 0.69, + "learning_rate": 4.526725625995956e-06, + "loss": 0.3033, + "step": 11485 + }, + { + "epoch": 0.69, + "learning_rate": 4.525088176467696e-06, + "loss": 0.2513, + "step": 11486 + }, + { + "epoch": 0.69, + "learning_rate": 4.523450936544687e-06, + "loss": 0.3044, + "step": 11487 + }, + { + "epoch": 0.69, + "learning_rate": 4.521813906289607e-06, + "loss": 0.2766, + "step": 11488 + }, + { + "epoch": 0.69, + "learning_rate": 4.5201770857651275e-06, + "loss": 0.2956, + "step": 11489 + }, + { + "epoch": 0.69, + "learning_rate": 4.518540475033919e-06, + "loss": 0.2953, + "step": 11490 + }, + { + "epoch": 0.69, + "learning_rate": 4.516904074158637e-06, + "loss": 0.3209, + "step": 11491 + }, + { + "epoch": 0.69, + "learning_rate": 4.515267883201928e-06, + "loss": 0.291, + "step": 11492 + }, + { + "epoch": 0.69, + "learning_rate": 4.513631902226437e-06, + "loss": 0.2801, + "step": 11493 + }, + { + "epoch": 0.69, + "learning_rate": 4.511996131294789e-06, + "loss": 0.2891, + "step": 11494 + }, + { + "epoch": 0.69, + "learning_rate": 4.51036057046962e-06, + "loss": 0.2704, + "step": 11495 + }, + { + "epoch": 0.69, + "learning_rate": 4.508725219813541e-06, + "loss": 0.2863, + "step": 11496 + }, + { + "epoch": 0.69, + "learning_rate": 4.507090079389161e-06, + "loss": 0.2893, + "step": 11497 + }, + { + "epoch": 0.69, + "learning_rate": 4.505455149259083e-06, + "loss": 0.2849, + "step": 11498 + }, + { + "epoch": 0.69, + "learning_rate": 4.503820429485892e-06, + "loss": 0.3055, + "step": 11499 + }, + { + "epoch": 0.69, + "learning_rate": 4.502185920132184e-06, + "loss": 0.2595, + "step": 11500 + }, + { + "epoch": 0.69, + "learning_rate": 4.500551621260531e-06, + "loss": 0.29, + "step": 11501 + }, + { + "epoch": 0.69, + "learning_rate": 4.498917532933503e-06, + "loss": 0.2824, + "step": 11502 + }, + { + "epoch": 0.69, + "learning_rate": 4.497283655213654e-06, + "loss": 0.2698, + "step": 11503 + }, + { + "epoch": 0.69, + "learning_rate": 4.495649988163542e-06, + "loss": 0.2683, + "step": 11504 + }, + { + "epoch": 0.7, + "learning_rate": 4.494016531845712e-06, + "loss": 0.3089, + "step": 11505 + }, + { + "epoch": 0.7, + "learning_rate": 4.492383286322699e-06, + "loss": 0.2771, + "step": 11506 + }, + { + "epoch": 0.7, + "learning_rate": 4.49075025165703e-06, + "loss": 0.3079, + "step": 11507 + }, + { + "epoch": 0.7, + "learning_rate": 4.489117427911223e-06, + "loss": 0.2612, + "step": 11508 + }, + { + "epoch": 0.7, + "learning_rate": 4.487484815147799e-06, + "loss": 0.2635, + "step": 11509 + }, + { + "epoch": 0.7, + "learning_rate": 4.485852413429256e-06, + "loss": 0.2766, + "step": 11510 + }, + { + "epoch": 0.7, + "learning_rate": 4.48422022281809e-06, + "loss": 0.2589, + "step": 11511 + }, + { + "epoch": 0.7, + "learning_rate": 4.482588243376789e-06, + "loss": 0.2708, + "step": 11512 + }, + { + "epoch": 0.7, + "learning_rate": 4.48095647516783e-06, + "loss": 0.2874, + "step": 11513 + }, + { + "epoch": 0.7, + "learning_rate": 4.479324918253693e-06, + "loss": 0.2874, + "step": 11514 + }, + { + "epoch": 0.7, + "learning_rate": 4.477693572696834e-06, + "loss": 0.2979, + "step": 11515 + }, + { + "epoch": 0.7, + "learning_rate": 4.476062438559713e-06, + "loss": 0.3278, + "step": 11516 + }, + { + "epoch": 0.7, + "learning_rate": 4.474431515904776e-06, + "loss": 0.2888, + "step": 11517 + }, + { + "epoch": 0.7, + "learning_rate": 4.472800804794461e-06, + "loss": 0.3023, + "step": 11518 + }, + { + "epoch": 0.7, + "learning_rate": 4.471170305291197e-06, + "loss": 0.3006, + "step": 11519 + }, + { + "epoch": 0.7, + "learning_rate": 4.469540017457413e-06, + "loss": 0.2824, + "step": 11520 + }, + { + "epoch": 0.7, + "learning_rate": 4.467909941355523e-06, + "loss": 0.2946, + "step": 11521 + }, + { + "epoch": 0.7, + "learning_rate": 4.466280077047932e-06, + "loss": 0.278, + "step": 11522 + }, + { + "epoch": 0.7, + "learning_rate": 4.464650424597039e-06, + "loss": 0.2849, + "step": 11523 + }, + { + "epoch": 0.7, + "learning_rate": 4.463020984065232e-06, + "loss": 0.2732, + "step": 11524 + }, + { + "epoch": 0.7, + "learning_rate": 4.461391755514899e-06, + "loss": 0.2938, + "step": 11525 + }, + { + "epoch": 0.7, + "learning_rate": 4.4597627390084116e-06, + "loss": 0.2746, + "step": 11526 + }, + { + "epoch": 0.7, + "learning_rate": 4.4581339346081385e-06, + "loss": 0.2926, + "step": 11527 + }, + { + "epoch": 0.7, + "learning_rate": 4.456505342376435e-06, + "loss": 0.2821, + "step": 11528 + }, + { + "epoch": 0.7, + "learning_rate": 4.454876962375649e-06, + "loss": 0.269, + "step": 11529 + }, + { + "epoch": 0.7, + "learning_rate": 4.453248794668129e-06, + "loss": 0.2898, + "step": 11530 + }, + { + "epoch": 0.7, + "learning_rate": 4.4516208393162065e-06, + "loss": 0.2693, + "step": 11531 + }, + { + "epoch": 0.7, + "learning_rate": 4.449993096382206e-06, + "loss": 0.2845, + "step": 11532 + }, + { + "epoch": 0.7, + "learning_rate": 4.4483655659284455e-06, + "loss": 0.2769, + "step": 11533 + }, + { + "epoch": 0.7, + "learning_rate": 4.44673824801723e-06, + "loss": 0.273, + "step": 11534 + }, + { + "epoch": 0.7, + "learning_rate": 4.445111142710871e-06, + "loss": 0.2829, + "step": 11535 + }, + { + "epoch": 0.7, + "learning_rate": 4.443484250071659e-06, + "loss": 0.285, + "step": 11536 + }, + { + "epoch": 0.7, + "learning_rate": 4.441857570161871e-06, + "loss": 0.2967, + "step": 11537 + }, + { + "epoch": 0.7, + "learning_rate": 4.440231103043788e-06, + "loss": 0.2891, + "step": 11538 + }, + { + "epoch": 0.7, + "learning_rate": 4.438604848779682e-06, + "loss": 0.298, + "step": 11539 + }, + { + "epoch": 0.7, + "learning_rate": 4.436978807431812e-06, + "loss": 0.2838, + "step": 11540 + }, + { + "epoch": 0.7, + "learning_rate": 4.4353529790624294e-06, + "loss": 0.2743, + "step": 11541 + }, + { + "epoch": 0.7, + "learning_rate": 4.433727363733779e-06, + "loss": 0.2437, + "step": 11542 + }, + { + "epoch": 0.7, + "learning_rate": 4.432101961508094e-06, + "loss": 0.2687, + "step": 11543 + }, + { + "epoch": 0.7, + "learning_rate": 4.43047677244761e-06, + "loss": 0.2856, + "step": 11544 + }, + { + "epoch": 0.7, + "learning_rate": 4.42885179661454e-06, + "loss": 0.2443, + "step": 11545 + }, + { + "epoch": 0.7, + "learning_rate": 4.4272270340711e-06, + "loss": 0.2831, + "step": 11546 + }, + { + "epoch": 0.7, + "learning_rate": 4.425602484879491e-06, + "loss": 0.2817, + "step": 11547 + }, + { + "epoch": 0.7, + "learning_rate": 4.423978149101904e-06, + "loss": 0.2875, + "step": 11548 + }, + { + "epoch": 0.7, + "learning_rate": 4.422354026800536e-06, + "loss": 0.2751, + "step": 11549 + }, + { + "epoch": 0.7, + "learning_rate": 4.420730118037561e-06, + "loss": 0.2871, + "step": 11550 + }, + { + "epoch": 0.7, + "learning_rate": 4.419106422875149e-06, + "loss": 0.3002, + "step": 11551 + }, + { + "epoch": 0.7, + "learning_rate": 4.417482941375464e-06, + "loss": 0.2723, + "step": 11552 + }, + { + "epoch": 0.7, + "learning_rate": 4.4158596736006566e-06, + "loss": 0.2782, + "step": 11553 + }, + { + "epoch": 0.7, + "learning_rate": 4.414236619612879e-06, + "loss": 0.2999, + "step": 11554 + }, + { + "epoch": 0.7, + "learning_rate": 4.412613779474267e-06, + "loss": 0.286, + "step": 11555 + }, + { + "epoch": 0.7, + "learning_rate": 4.410991153246952e-06, + "loss": 0.2501, + "step": 11556 + }, + { + "epoch": 0.7, + "learning_rate": 4.409368740993053e-06, + "loss": 0.2758, + "step": 11557 + }, + { + "epoch": 0.7, + "learning_rate": 4.4077465427746834e-06, + "loss": 0.2914, + "step": 11558 + }, + { + "epoch": 0.7, + "learning_rate": 4.406124558653947e-06, + "loss": 0.2695, + "step": 11559 + }, + { + "epoch": 0.7, + "learning_rate": 4.404502788692947e-06, + "loss": 0.278, + "step": 11560 + }, + { + "epoch": 0.7, + "learning_rate": 4.40288123295377e-06, + "loss": 0.2787, + "step": 11561 + }, + { + "epoch": 0.7, + "learning_rate": 4.401259891498495e-06, + "loss": 0.2931, + "step": 11562 + }, + { + "epoch": 0.7, + "learning_rate": 4.3996387643891945e-06, + "loss": 0.2877, + "step": 11563 + }, + { + "epoch": 0.7, + "learning_rate": 4.39801785168793e-06, + "loss": 0.2832, + "step": 11564 + }, + { + "epoch": 0.7, + "learning_rate": 4.396397153456766e-06, + "loss": 0.2578, + "step": 11565 + }, + { + "epoch": 0.7, + "learning_rate": 4.394776669757745e-06, + "loss": 0.2852, + "step": 11566 + }, + { + "epoch": 0.7, + "learning_rate": 4.393156400652907e-06, + "loss": 0.3058, + "step": 11567 + }, + { + "epoch": 0.7, + "learning_rate": 4.391536346204285e-06, + "loss": 0.261, + "step": 11568 + }, + { + "epoch": 0.7, + "learning_rate": 4.389916506473896e-06, + "loss": 0.2708, + "step": 11569 + }, + { + "epoch": 0.7, + "learning_rate": 4.388296881523768e-06, + "loss": 0.2761, + "step": 11570 + }, + { + "epoch": 0.7, + "learning_rate": 4.3866774714158965e-06, + "loss": 0.2806, + "step": 11571 + }, + { + "epoch": 0.7, + "learning_rate": 4.385058276212283e-06, + "loss": 0.3011, + "step": 11572 + }, + { + "epoch": 0.7, + "learning_rate": 4.383439295974915e-06, + "loss": 0.2815, + "step": 11573 + }, + { + "epoch": 0.7, + "learning_rate": 4.381820530765781e-06, + "loss": 0.315, + "step": 11574 + }, + { + "epoch": 0.7, + "learning_rate": 4.380201980646853e-06, + "loss": 0.2833, + "step": 11575 + }, + { + "epoch": 0.7, + "learning_rate": 4.378583645680096e-06, + "loss": 0.2733, + "step": 11576 + }, + { + "epoch": 0.7, + "learning_rate": 4.3769655259274655e-06, + "loss": 0.255, + "step": 11577 + }, + { + "epoch": 0.7, + "learning_rate": 4.3753476214509105e-06, + "loss": 0.3083, + "step": 11578 + }, + { + "epoch": 0.7, + "learning_rate": 4.373729932312376e-06, + "loss": 0.278, + "step": 11579 + }, + { + "epoch": 0.7, + "learning_rate": 4.372112458573793e-06, + "loss": 0.2847, + "step": 11580 + }, + { + "epoch": 0.7, + "learning_rate": 4.370495200297084e-06, + "loss": 0.2967, + "step": 11581 + }, + { + "epoch": 0.7, + "learning_rate": 4.368878157544169e-06, + "loss": 0.2702, + "step": 11582 + }, + { + "epoch": 0.7, + "learning_rate": 4.367261330376948e-06, + "loss": 0.2709, + "step": 11583 + }, + { + "epoch": 0.7, + "learning_rate": 4.365644718857331e-06, + "loss": 0.2736, + "step": 11584 + }, + { + "epoch": 0.7, + "learning_rate": 4.364028323047205e-06, + "loss": 0.298, + "step": 11585 + }, + { + "epoch": 0.7, + "learning_rate": 4.362412143008452e-06, + "loss": 0.3052, + "step": 11586 + }, + { + "epoch": 0.7, + "learning_rate": 4.360796178802948e-06, + "loss": 0.299, + "step": 11587 + }, + { + "epoch": 0.7, + "learning_rate": 4.359180430492556e-06, + "loss": 0.2695, + "step": 11588 + }, + { + "epoch": 0.7, + "learning_rate": 4.357564898139142e-06, + "loss": 0.2896, + "step": 11589 + }, + { + "epoch": 0.7, + "learning_rate": 4.355949581804552e-06, + "loss": 0.2782, + "step": 11590 + }, + { + "epoch": 0.7, + "learning_rate": 4.354334481550627e-06, + "loss": 0.2761, + "step": 11591 + }, + { + "epoch": 0.7, + "learning_rate": 4.352719597439202e-06, + "loss": 0.2395, + "step": 11592 + }, + { + "epoch": 0.7, + "learning_rate": 4.351104929532103e-06, + "loss": 0.2599, + "step": 11593 + }, + { + "epoch": 0.7, + "learning_rate": 4.349490477891142e-06, + "loss": 0.2874, + "step": 11594 + }, + { + "epoch": 0.7, + "learning_rate": 4.347876242578134e-06, + "loss": 0.2905, + "step": 11595 + }, + { + "epoch": 0.7, + "learning_rate": 4.346262223654878e-06, + "loss": 0.2681, + "step": 11596 + }, + { + "epoch": 0.7, + "learning_rate": 4.344648421183167e-06, + "loss": 0.2985, + "step": 11597 + }, + { + "epoch": 0.7, + "learning_rate": 4.3430348352247805e-06, + "loss": 0.2706, + "step": 11598 + }, + { + "epoch": 0.7, + "learning_rate": 4.341421465841494e-06, + "loss": 0.2699, + "step": 11599 + }, + { + "epoch": 0.7, + "learning_rate": 4.3398083130950815e-06, + "loss": 0.2759, + "step": 11600 + }, + { + "epoch": 0.7, + "learning_rate": 4.3381953770473e-06, + "loss": 0.2789, + "step": 11601 + }, + { + "epoch": 0.7, + "learning_rate": 4.336582657759897e-06, + "loss": 0.2846, + "step": 11602 + }, + { + "epoch": 0.7, + "learning_rate": 4.334970155294617e-06, + "loss": 0.2792, + "step": 11603 + }, + { + "epoch": 0.7, + "learning_rate": 4.333357869713193e-06, + "loss": 0.2944, + "step": 11604 + }, + { + "epoch": 0.7, + "learning_rate": 4.331745801077352e-06, + "loss": 0.2683, + "step": 11605 + }, + { + "epoch": 0.7, + "learning_rate": 4.330133949448809e-06, + "loss": 0.2831, + "step": 11606 + }, + { + "epoch": 0.7, + "learning_rate": 4.3285223148892774e-06, + "loss": 0.2983, + "step": 11607 + }, + { + "epoch": 0.7, + "learning_rate": 4.326910897460452e-06, + "loss": 0.3023, + "step": 11608 + }, + { + "epoch": 0.7, + "learning_rate": 4.3252996972240326e-06, + "loss": 0.2677, + "step": 11609 + }, + { + "epoch": 0.7, + "learning_rate": 4.3236887142417e-06, + "loss": 0.3044, + "step": 11610 + }, + { + "epoch": 0.7, + "learning_rate": 4.322077948575131e-06, + "loss": 0.2812, + "step": 11611 + }, + { + "epoch": 0.7, + "learning_rate": 4.320467400285993e-06, + "loss": 0.2775, + "step": 11612 + }, + { + "epoch": 0.7, + "learning_rate": 4.31885706943594e-06, + "loss": 0.3209, + "step": 11613 + }, + { + "epoch": 0.7, + "learning_rate": 4.317246956086632e-06, + "loss": 0.303, + "step": 11614 + }, + { + "epoch": 0.7, + "learning_rate": 4.315637060299707e-06, + "loss": 0.2545, + "step": 11615 + }, + { + "epoch": 0.7, + "learning_rate": 4.314027382136801e-06, + "loss": 0.2499, + "step": 11616 + }, + { + "epoch": 0.7, + "learning_rate": 4.3124179216595385e-06, + "loss": 0.2854, + "step": 11617 + }, + { + "epoch": 0.7, + "learning_rate": 4.310808678929535e-06, + "loss": 0.2582, + "step": 11618 + }, + { + "epoch": 0.7, + "learning_rate": 4.309199654008405e-06, + "loss": 0.2881, + "step": 11619 + }, + { + "epoch": 0.7, + "learning_rate": 4.307590846957748e-06, + "loss": 0.2935, + "step": 11620 + }, + { + "epoch": 0.7, + "learning_rate": 4.305982257839154e-06, + "loss": 0.3237, + "step": 11621 + }, + { + "epoch": 0.7, + "learning_rate": 4.304373886714211e-06, + "loss": 0.2849, + "step": 11622 + }, + { + "epoch": 0.7, + "learning_rate": 4.302765733644487e-06, + "loss": 0.3058, + "step": 11623 + }, + { + "epoch": 0.7, + "learning_rate": 4.301157798691561e-06, + "loss": 0.2927, + "step": 11624 + }, + { + "epoch": 0.7, + "learning_rate": 4.299550081916987e-06, + "loss": 0.2899, + "step": 11625 + }, + { + "epoch": 0.7, + "learning_rate": 4.297942583382315e-06, + "loss": 0.267, + "step": 11626 + }, + { + "epoch": 0.7, + "learning_rate": 4.2963353031490885e-06, + "loss": 0.2667, + "step": 11627 + }, + { + "epoch": 0.7, + "learning_rate": 4.2947282412788415e-06, + "loss": 0.2571, + "step": 11628 + }, + { + "epoch": 0.7, + "learning_rate": 4.293121397833097e-06, + "loss": 0.2743, + "step": 11629 + }, + { + "epoch": 0.7, + "learning_rate": 4.2915147728733775e-06, + "loss": 0.2655, + "step": 11630 + }, + { + "epoch": 0.7, + "learning_rate": 4.289908366461192e-06, + "loss": 0.3036, + "step": 11631 + }, + { + "epoch": 0.7, + "learning_rate": 4.288302178658038e-06, + "loss": 0.2999, + "step": 11632 + }, + { + "epoch": 0.7, + "learning_rate": 4.2866962095254095e-06, + "loss": 0.2868, + "step": 11633 + }, + { + "epoch": 0.7, + "learning_rate": 4.285090459124787e-06, + "loss": 0.2907, + "step": 11634 + }, + { + "epoch": 0.7, + "learning_rate": 4.283484927517651e-06, + "loss": 0.2786, + "step": 11635 + }, + { + "epoch": 0.7, + "learning_rate": 4.281879614765469e-06, + "loss": 0.2773, + "step": 11636 + }, + { + "epoch": 0.7, + "learning_rate": 4.280274520929698e-06, + "loss": 0.2811, + "step": 11637 + }, + { + "epoch": 0.7, + "learning_rate": 4.278669646071788e-06, + "loss": 0.2536, + "step": 11638 + }, + { + "epoch": 0.7, + "learning_rate": 4.277064990253181e-06, + "loss": 0.2829, + "step": 11639 + }, + { + "epoch": 0.7, + "learning_rate": 4.2754605535353125e-06, + "loss": 0.2801, + "step": 11640 + }, + { + "epoch": 0.7, + "learning_rate": 4.273856335979606e-06, + "loss": 0.3126, + "step": 11641 + }, + { + "epoch": 0.7, + "learning_rate": 4.272252337647479e-06, + "loss": 0.2973, + "step": 11642 + }, + { + "epoch": 0.7, + "learning_rate": 4.270648558600337e-06, + "loss": 0.2686, + "step": 11643 + }, + { + "epoch": 0.7, + "learning_rate": 4.269044998899586e-06, + "loss": 0.2746, + "step": 11644 + }, + { + "epoch": 0.7, + "learning_rate": 4.267441658606617e-06, + "loss": 0.2697, + "step": 11645 + }, + { + "epoch": 0.7, + "learning_rate": 4.26583853778281e-06, + "loss": 0.2676, + "step": 11646 + }, + { + "epoch": 0.7, + "learning_rate": 4.264235636489542e-06, + "loss": 0.2772, + "step": 11647 + }, + { + "epoch": 0.7, + "learning_rate": 4.262632954788175e-06, + "loss": 0.2793, + "step": 11648 + }, + { + "epoch": 0.7, + "learning_rate": 4.261030492740076e-06, + "loss": 0.291, + "step": 11649 + }, + { + "epoch": 0.7, + "learning_rate": 4.2594282504065885e-06, + "loss": 0.3047, + "step": 11650 + }, + { + "epoch": 0.7, + "learning_rate": 4.257826227849057e-06, + "loss": 0.2888, + "step": 11651 + }, + { + "epoch": 0.7, + "learning_rate": 4.256224425128811e-06, + "loss": 0.291, + "step": 11652 + }, + { + "epoch": 0.7, + "learning_rate": 4.2546228423071735e-06, + "loss": 0.2743, + "step": 11653 + }, + { + "epoch": 0.7, + "learning_rate": 4.2530214794454685e-06, + "loss": 0.3008, + "step": 11654 + }, + { + "epoch": 0.7, + "learning_rate": 4.251420336604998e-06, + "loss": 0.322, + "step": 11655 + }, + { + "epoch": 0.7, + "learning_rate": 4.249819413847061e-06, + "loss": 0.2673, + "step": 11656 + }, + { + "epoch": 0.7, + "learning_rate": 4.248218711232952e-06, + "loss": 0.2799, + "step": 11657 + }, + { + "epoch": 0.7, + "learning_rate": 4.246618228823945e-06, + "loss": 0.3011, + "step": 11658 + }, + { + "epoch": 0.7, + "learning_rate": 4.2450179666813255e-06, + "loss": 0.2782, + "step": 11659 + }, + { + "epoch": 0.7, + "learning_rate": 4.243417924866352e-06, + "loss": 0.293, + "step": 11660 + }, + { + "epoch": 0.7, + "learning_rate": 4.241818103440284e-06, + "loss": 0.2649, + "step": 11661 + }, + { + "epoch": 0.7, + "learning_rate": 4.240218502464368e-06, + "loss": 0.279, + "step": 11662 + }, + { + "epoch": 0.7, + "learning_rate": 4.238619121999846e-06, + "loss": 0.2677, + "step": 11663 + }, + { + "epoch": 0.7, + "learning_rate": 4.2370199621079455e-06, + "loss": 0.2783, + "step": 11664 + }, + { + "epoch": 0.7, + "learning_rate": 4.235421022849897e-06, + "loss": 0.3059, + "step": 11665 + }, + { + "epoch": 0.7, + "learning_rate": 4.233822304286912e-06, + "loss": 0.2695, + "step": 11666 + }, + { + "epoch": 0.7, + "learning_rate": 4.232223806480197e-06, + "loss": 0.262, + "step": 11667 + }, + { + "epoch": 0.7, + "learning_rate": 4.230625529490948e-06, + "loss": 0.2826, + "step": 11668 + }, + { + "epoch": 0.7, + "learning_rate": 4.229027473380355e-06, + "loss": 0.2843, + "step": 11669 + }, + { + "epoch": 0.71, + "learning_rate": 4.227429638209602e-06, + "loss": 0.286, + "step": 11670 + }, + { + "epoch": 0.71, + "learning_rate": 4.225832024039861e-06, + "loss": 0.2688, + "step": 11671 + }, + { + "epoch": 0.71, + "learning_rate": 4.224234630932299e-06, + "loss": 0.2694, + "step": 11672 + }, + { + "epoch": 0.71, + "learning_rate": 4.222637458948059e-06, + "loss": 0.2932, + "step": 11673 + }, + { + "epoch": 0.71, + "learning_rate": 4.221040508148302e-06, + "loss": 0.2891, + "step": 11674 + }, + { + "epoch": 0.71, + "learning_rate": 4.219443778594161e-06, + "loss": 0.2941, + "step": 11675 + }, + { + "epoch": 0.71, + "learning_rate": 4.217847270346768e-06, + "loss": 0.2941, + "step": 11676 + }, + { + "epoch": 0.71, + "learning_rate": 4.216250983467242e-06, + "loss": 0.2891, + "step": 11677 + }, + { + "epoch": 0.71, + "learning_rate": 4.2146549180166965e-06, + "loss": 0.3092, + "step": 11678 + }, + { + "epoch": 0.71, + "learning_rate": 4.21305907405624e-06, + "loss": 0.2874, + "step": 11679 + }, + { + "epoch": 0.71, + "learning_rate": 4.211463451646968e-06, + "loss": 0.2768, + "step": 11680 + }, + { + "epoch": 0.71, + "learning_rate": 4.209868050849967e-06, + "loss": 0.3033, + "step": 11681 + }, + { + "epoch": 0.71, + "learning_rate": 4.208272871726317e-06, + "loss": 0.2831, + "step": 11682 + }, + { + "epoch": 0.71, + "learning_rate": 4.2066779143370836e-06, + "loss": 0.2691, + "step": 11683 + }, + { + "epoch": 0.71, + "learning_rate": 4.205083178743339e-06, + "loss": 0.2912, + "step": 11684 + }, + { + "epoch": 0.71, + "learning_rate": 4.203488665006132e-06, + "loss": 0.2753, + "step": 11685 + }, + { + "epoch": 0.71, + "learning_rate": 4.201894373186509e-06, + "loss": 0.2688, + "step": 11686 + }, + { + "epoch": 0.71, + "learning_rate": 4.200300303345506e-06, + "loss": 0.2631, + "step": 11687 + }, + { + "epoch": 0.71, + "learning_rate": 4.198706455544148e-06, + "loss": 0.3001, + "step": 11688 + }, + { + "epoch": 0.71, + "learning_rate": 4.1971128298434626e-06, + "loss": 0.2814, + "step": 11689 + }, + { + "epoch": 0.71, + "learning_rate": 4.195519426304457e-06, + "loss": 0.2485, + "step": 11690 + }, + { + "epoch": 0.71, + "learning_rate": 4.193926244988136e-06, + "loss": 0.2513, + "step": 11691 + }, + { + "epoch": 0.71, + "learning_rate": 4.192333285955491e-06, + "loss": 0.269, + "step": 11692 + }, + { + "epoch": 0.71, + "learning_rate": 4.1907405492675065e-06, + "loss": 0.3054, + "step": 11693 + }, + { + "epoch": 0.71, + "learning_rate": 4.189148034985167e-06, + "loss": 0.2581, + "step": 11694 + }, + { + "epoch": 0.71, + "learning_rate": 4.187555743169437e-06, + "loss": 0.2776, + "step": 11695 + }, + { + "epoch": 0.71, + "learning_rate": 4.185963673881277e-06, + "loss": 0.2559, + "step": 11696 + }, + { + "epoch": 0.71, + "learning_rate": 4.184371827181639e-06, + "loss": 0.3223, + "step": 11697 + }, + { + "epoch": 0.71, + "learning_rate": 4.182780203131467e-06, + "loss": 0.2673, + "step": 11698 + }, + { + "epoch": 0.71, + "learning_rate": 4.1811888017916914e-06, + "loss": 0.2693, + "step": 11699 + }, + { + "epoch": 0.71, + "learning_rate": 4.179597623223246e-06, + "loss": 0.2967, + "step": 11700 + }, + { + "epoch": 0.71, + "learning_rate": 4.178006667487044e-06, + "loss": 0.2753, + "step": 11701 + }, + { + "epoch": 0.71, + "learning_rate": 4.176415934643998e-06, + "loss": 0.2849, + "step": 11702 + }, + { + "epoch": 0.71, + "learning_rate": 4.174825424755005e-06, + "loss": 0.2652, + "step": 11703 + }, + { + "epoch": 0.71, + "learning_rate": 4.173235137880954e-06, + "loss": 0.2665, + "step": 11704 + }, + { + "epoch": 0.71, + "learning_rate": 4.171645074082737e-06, + "loss": 0.2799, + "step": 11705 + }, + { + "epoch": 0.71, + "learning_rate": 4.1700552334212285e-06, + "loss": 0.2598, + "step": 11706 + }, + { + "epoch": 0.71, + "learning_rate": 4.168465615957289e-06, + "loss": 0.2631, + "step": 11707 + }, + { + "epoch": 0.71, + "learning_rate": 4.1668762217517755e-06, + "loss": 0.285, + "step": 11708 + }, + { + "epoch": 0.71, + "learning_rate": 4.165287050865544e-06, + "loss": 0.3058, + "step": 11709 + }, + { + "epoch": 0.71, + "learning_rate": 4.1636981033594326e-06, + "loss": 0.2976, + "step": 11710 + }, + { + "epoch": 0.71, + "learning_rate": 4.162109379294273e-06, + "loss": 0.2468, + "step": 11711 + }, + { + "epoch": 0.71, + "learning_rate": 4.160520878730891e-06, + "loss": 0.2951, + "step": 11712 + }, + { + "epoch": 0.71, + "learning_rate": 4.158932601730097e-06, + "loss": 0.285, + "step": 11713 + }, + { + "epoch": 0.71, + "learning_rate": 4.157344548352703e-06, + "loss": 0.2739, + "step": 11714 + }, + { + "epoch": 0.71, + "learning_rate": 4.155756718659506e-06, + "loss": 0.2765, + "step": 11715 + }, + { + "epoch": 0.71, + "learning_rate": 4.154169112711295e-06, + "loss": 0.2902, + "step": 11716 + }, + { + "epoch": 0.71, + "learning_rate": 4.15258173056885e-06, + "loss": 0.2778, + "step": 11717 + }, + { + "epoch": 0.71, + "learning_rate": 4.150994572292941e-06, + "loss": 0.2967, + "step": 11718 + }, + { + "epoch": 0.71, + "learning_rate": 4.1494076379443385e-06, + "loss": 0.3073, + "step": 11719 + }, + { + "epoch": 0.71, + "learning_rate": 4.147820927583794e-06, + "loss": 0.2799, + "step": 11720 + }, + { + "epoch": 0.71, + "learning_rate": 4.146234441272054e-06, + "loss": 0.2814, + "step": 11721 + }, + { + "epoch": 0.71, + "learning_rate": 4.144648179069859e-06, + "loss": 0.3245, + "step": 11722 + }, + { + "epoch": 0.71, + "learning_rate": 4.143062141037931e-06, + "loss": 0.2753, + "step": 11723 + }, + { + "epoch": 0.71, + "learning_rate": 4.141476327237002e-06, + "loss": 0.3066, + "step": 11724 + }, + { + "epoch": 0.71, + "learning_rate": 4.139890737727777e-06, + "loss": 0.2919, + "step": 11725 + }, + { + "epoch": 0.71, + "learning_rate": 4.138305372570963e-06, + "loss": 0.284, + "step": 11726 + }, + { + "epoch": 0.71, + "learning_rate": 4.136720231827254e-06, + "loss": 0.2884, + "step": 11727 + }, + { + "epoch": 0.71, + "learning_rate": 4.135135315557332e-06, + "loss": 0.2794, + "step": 11728 + }, + { + "epoch": 0.71, + "learning_rate": 4.133550623821884e-06, + "loss": 0.28, + "step": 11729 + }, + { + "epoch": 0.71, + "learning_rate": 4.131966156681575e-06, + "loss": 0.2951, + "step": 11730 + }, + { + "epoch": 0.71, + "learning_rate": 4.130381914197066e-06, + "loss": 0.2794, + "step": 11731 + }, + { + "epoch": 0.71, + "learning_rate": 4.128797896429009e-06, + "loss": 0.2681, + "step": 11732 + }, + { + "epoch": 0.71, + "learning_rate": 4.1272141034380466e-06, + "loss": 0.2919, + "step": 11733 + }, + { + "epoch": 0.71, + "learning_rate": 4.125630535284812e-06, + "loss": 0.2826, + "step": 11734 + }, + { + "epoch": 0.71, + "learning_rate": 4.1240471920299374e-06, + "loss": 0.2792, + "step": 11735 + }, + { + "epoch": 0.71, + "learning_rate": 4.122464073734038e-06, + "loss": 0.2938, + "step": 11736 + }, + { + "epoch": 0.71, + "learning_rate": 4.120881180457723e-06, + "loss": 0.2771, + "step": 11737 + }, + { + "epoch": 0.71, + "learning_rate": 4.119298512261593e-06, + "loss": 0.2737, + "step": 11738 + }, + { + "epoch": 0.71, + "learning_rate": 4.117716069206235e-06, + "loss": 0.2879, + "step": 11739 + }, + { + "epoch": 0.71, + "learning_rate": 4.116133851352243e-06, + "loss": 0.2997, + "step": 11740 + }, + { + "epoch": 0.71, + "learning_rate": 4.114551858760183e-06, + "loss": 0.2799, + "step": 11741 + }, + { + "epoch": 0.71, + "learning_rate": 4.112970091490625e-06, + "loss": 0.2836, + "step": 11742 + }, + { + "epoch": 0.71, + "learning_rate": 4.111388549604119e-06, + "loss": 0.2916, + "step": 11743 + }, + { + "epoch": 0.71, + "learning_rate": 4.1098072331612246e-06, + "loss": 0.3034, + "step": 11744 + }, + { + "epoch": 0.71, + "learning_rate": 4.1082261422224775e-06, + "loss": 0.3153, + "step": 11745 + }, + { + "epoch": 0.71, + "learning_rate": 4.106645276848409e-06, + "loss": 0.2906, + "step": 11746 + }, + { + "epoch": 0.71, + "learning_rate": 4.105064637099543e-06, + "loss": 0.2693, + "step": 11747 + }, + { + "epoch": 0.71, + "learning_rate": 4.103484223036388e-06, + "loss": 0.2624, + "step": 11748 + }, + { + "epoch": 0.71, + "learning_rate": 4.1019040347194584e-06, + "loss": 0.2723, + "step": 11749 + }, + { + "epoch": 0.71, + "learning_rate": 4.100324072209247e-06, + "loss": 0.2893, + "step": 11750 + }, + { + "epoch": 0.71, + "learning_rate": 4.098744335566243e-06, + "loss": 0.2967, + "step": 11751 + }, + { + "epoch": 0.71, + "learning_rate": 4.097164824850926e-06, + "loss": 0.2949, + "step": 11752 + }, + { + "epoch": 0.71, + "learning_rate": 4.095585540123762e-06, + "loss": 0.2799, + "step": 11753 + }, + { + "epoch": 0.71, + "learning_rate": 4.094006481445222e-06, + "loss": 0.293, + "step": 11754 + }, + { + "epoch": 0.71, + "learning_rate": 4.092427648875756e-06, + "loss": 0.2716, + "step": 11755 + }, + { + "epoch": 0.71, + "learning_rate": 4.0908490424758086e-06, + "loss": 0.264, + "step": 11756 + }, + { + "epoch": 0.71, + "learning_rate": 4.0892706623058175e-06, + "loss": 0.2744, + "step": 11757 + }, + { + "epoch": 0.71, + "learning_rate": 4.087692508426205e-06, + "loss": 0.2693, + "step": 11758 + }, + { + "epoch": 0.71, + "learning_rate": 4.086114580897399e-06, + "loss": 0.2972, + "step": 11759 + }, + { + "epoch": 0.71, + "learning_rate": 4.084536879779806e-06, + "loss": 0.2619, + "step": 11760 + }, + { + "epoch": 0.71, + "learning_rate": 4.082959405133826e-06, + "loss": 0.2875, + "step": 11761 + }, + { + "epoch": 0.71, + "learning_rate": 4.0813821570198544e-06, + "loss": 0.2992, + "step": 11762 + }, + { + "epoch": 0.71, + "learning_rate": 4.079805135498271e-06, + "loss": 0.2485, + "step": 11763 + }, + { + "epoch": 0.71, + "learning_rate": 4.07822834062946e-06, + "loss": 0.2762, + "step": 11764 + }, + { + "epoch": 0.71, + "learning_rate": 4.076651772473783e-06, + "loss": 0.2652, + "step": 11765 + }, + { + "epoch": 0.71, + "learning_rate": 4.0750754310915995e-06, + "loss": 0.2937, + "step": 11766 + }, + { + "epoch": 0.71, + "learning_rate": 4.073499316543259e-06, + "loss": 0.3025, + "step": 11767 + }, + { + "epoch": 0.71, + "learning_rate": 4.071923428889103e-06, + "loss": 0.2888, + "step": 11768 + }, + { + "epoch": 0.71, + "learning_rate": 4.0703477681894595e-06, + "loss": 0.2877, + "step": 11769 + }, + { + "epoch": 0.71, + "learning_rate": 4.06877233450466e-06, + "loss": 0.3025, + "step": 11770 + }, + { + "epoch": 0.71, + "learning_rate": 4.067197127895015e-06, + "loss": 0.2952, + "step": 11771 + }, + { + "epoch": 0.71, + "learning_rate": 4.065622148420832e-06, + "loss": 0.2938, + "step": 11772 + }, + { + "epoch": 0.71, + "learning_rate": 4.064047396142407e-06, + "loss": 0.264, + "step": 11773 + }, + { + "epoch": 0.71, + "learning_rate": 4.062472871120031e-06, + "loss": 0.2744, + "step": 11774 + }, + { + "epoch": 0.71, + "learning_rate": 4.0608985734139824e-06, + "loss": 0.2871, + "step": 11775 + }, + { + "epoch": 0.71, + "learning_rate": 4.059324503084533e-06, + "loss": 0.2839, + "step": 11776 + }, + { + "epoch": 0.71, + "learning_rate": 4.057750660191947e-06, + "loss": 0.2828, + "step": 11777 + }, + { + "epoch": 0.71, + "learning_rate": 4.056177044796472e-06, + "loss": 0.2653, + "step": 11778 + }, + { + "epoch": 0.71, + "learning_rate": 4.054603656958363e-06, + "loss": 0.2906, + "step": 11779 + }, + { + "epoch": 0.71, + "learning_rate": 4.0530304967378534e-06, + "loss": 0.2659, + "step": 11780 + }, + { + "epoch": 0.71, + "learning_rate": 4.051457564195169e-06, + "loss": 0.2905, + "step": 11781 + }, + { + "epoch": 0.71, + "learning_rate": 4.0498848593905315e-06, + "loss": 0.2879, + "step": 11782 + }, + { + "epoch": 0.71, + "learning_rate": 4.048312382384146e-06, + "loss": 0.2877, + "step": 11783 + }, + { + "epoch": 0.71, + "learning_rate": 4.04674013323622e-06, + "loss": 0.3061, + "step": 11784 + }, + { + "epoch": 0.71, + "learning_rate": 4.045168112006947e-06, + "loss": 0.2946, + "step": 11785 + }, + { + "epoch": 0.71, + "learning_rate": 4.043596318756507e-06, + "loss": 0.298, + "step": 11786 + }, + { + "epoch": 0.71, + "learning_rate": 4.04202475354508e-06, + "loss": 0.2764, + "step": 11787 + }, + { + "epoch": 0.71, + "learning_rate": 4.0404534164328255e-06, + "loss": 0.2994, + "step": 11788 + }, + { + "epoch": 0.71, + "learning_rate": 4.038882307479912e-06, + "loss": 0.2898, + "step": 11789 + }, + { + "epoch": 0.71, + "learning_rate": 4.037311426746483e-06, + "loss": 0.2545, + "step": 11790 + }, + { + "epoch": 0.71, + "learning_rate": 4.035740774292679e-06, + "loss": 0.269, + "step": 11791 + }, + { + "epoch": 0.71, + "learning_rate": 4.034170350178633e-06, + "loss": 0.2963, + "step": 11792 + }, + { + "epoch": 0.71, + "learning_rate": 4.032600154464466e-06, + "loss": 0.2987, + "step": 11793 + }, + { + "epoch": 0.71, + "learning_rate": 4.031030187210297e-06, + "loss": 0.2926, + "step": 11794 + }, + { + "epoch": 0.71, + "learning_rate": 4.029460448476228e-06, + "loss": 0.2927, + "step": 11795 + }, + { + "epoch": 0.71, + "learning_rate": 4.027890938322359e-06, + "loss": 0.2797, + "step": 11796 + }, + { + "epoch": 0.71, + "learning_rate": 4.026321656808775e-06, + "loss": 0.281, + "step": 11797 + }, + { + "epoch": 0.71, + "learning_rate": 4.0247526039955535e-06, + "loss": 0.2672, + "step": 11798 + }, + { + "epoch": 0.71, + "learning_rate": 4.023183779942772e-06, + "loss": 0.2744, + "step": 11799 + }, + { + "epoch": 0.71, + "learning_rate": 4.021615184710488e-06, + "loss": 0.2765, + "step": 11800 + }, + { + "epoch": 0.71, + "learning_rate": 4.0200468183587556e-06, + "loss": 0.2704, + "step": 11801 + }, + { + "epoch": 0.71, + "learning_rate": 4.018478680947621e-06, + "loss": 0.2775, + "step": 11802 + }, + { + "epoch": 0.71, + "learning_rate": 4.016910772537112e-06, + "loss": 0.2801, + "step": 11803 + }, + { + "epoch": 0.71, + "learning_rate": 4.015343093187266e-06, + "loss": 0.2859, + "step": 11804 + }, + { + "epoch": 0.71, + "learning_rate": 4.013775642958097e-06, + "loss": 0.2843, + "step": 11805 + }, + { + "epoch": 0.71, + "learning_rate": 4.012208421909614e-06, + "loss": 0.2494, + "step": 11806 + }, + { + "epoch": 0.71, + "learning_rate": 4.010641430101816e-06, + "loss": 0.2747, + "step": 11807 + }, + { + "epoch": 0.71, + "learning_rate": 4.009074667594697e-06, + "loss": 0.3013, + "step": 11808 + }, + { + "epoch": 0.71, + "learning_rate": 4.007508134448239e-06, + "loss": 0.2772, + "step": 11809 + }, + { + "epoch": 0.71, + "learning_rate": 4.005941830722416e-06, + "loss": 0.2765, + "step": 11810 + }, + { + "epoch": 0.71, + "learning_rate": 4.0043757564771945e-06, + "loss": 0.2718, + "step": 11811 + }, + { + "epoch": 0.71, + "learning_rate": 4.0028099117725306e-06, + "loss": 0.2962, + "step": 11812 + }, + { + "epoch": 0.71, + "learning_rate": 4.001244296668367e-06, + "loss": 0.2591, + "step": 11813 + }, + { + "epoch": 0.71, + "learning_rate": 3.999678911224653e-06, + "loss": 0.2861, + "step": 11814 + }, + { + "epoch": 0.71, + "learning_rate": 3.998113755501313e-06, + "loss": 0.2785, + "step": 11815 + }, + { + "epoch": 0.71, + "learning_rate": 3.99654882955827e-06, + "loss": 0.2727, + "step": 11816 + }, + { + "epoch": 0.71, + "learning_rate": 3.994984133455435e-06, + "loss": 0.2659, + "step": 11817 + }, + { + "epoch": 0.71, + "learning_rate": 3.993419667252709e-06, + "loss": 0.2443, + "step": 11818 + }, + { + "epoch": 0.71, + "learning_rate": 3.9918554310099955e-06, + "loss": 0.2811, + "step": 11819 + }, + { + "epoch": 0.71, + "learning_rate": 3.990291424787175e-06, + "loss": 0.2783, + "step": 11820 + }, + { + "epoch": 0.71, + "learning_rate": 3.988727648644126e-06, + "loss": 0.2821, + "step": 11821 + }, + { + "epoch": 0.71, + "learning_rate": 3.987164102640718e-06, + "loss": 0.2765, + "step": 11822 + }, + { + "epoch": 0.71, + "learning_rate": 3.985600786836806e-06, + "loss": 0.2593, + "step": 11823 + }, + { + "epoch": 0.71, + "learning_rate": 3.984037701292251e-06, + "loss": 0.3076, + "step": 11824 + }, + { + "epoch": 0.71, + "learning_rate": 3.982474846066886e-06, + "loss": 0.2861, + "step": 11825 + }, + { + "epoch": 0.71, + "learning_rate": 3.98091222122055e-06, + "loss": 0.2793, + "step": 11826 + }, + { + "epoch": 0.71, + "learning_rate": 3.979349826813065e-06, + "loss": 0.267, + "step": 11827 + }, + { + "epoch": 0.71, + "learning_rate": 3.977787662904243e-06, + "loss": 0.2773, + "step": 11828 + }, + { + "epoch": 0.71, + "learning_rate": 3.9762257295539e-06, + "loss": 0.2517, + "step": 11829 + }, + { + "epoch": 0.71, + "learning_rate": 3.974664026821829e-06, + "loss": 0.2845, + "step": 11830 + }, + { + "epoch": 0.71, + "learning_rate": 3.973102554767818e-06, + "loss": 0.2535, + "step": 11831 + }, + { + "epoch": 0.71, + "learning_rate": 3.97154131345165e-06, + "loss": 0.3185, + "step": 11832 + }, + { + "epoch": 0.71, + "learning_rate": 3.9699803029330915e-06, + "loss": 0.2847, + "step": 11833 + }, + { + "epoch": 0.71, + "learning_rate": 3.968419523271915e-06, + "loss": 0.2646, + "step": 11834 + }, + { + "epoch": 0.71, + "learning_rate": 3.966858974527866e-06, + "loss": 0.2885, + "step": 11835 + }, + { + "epoch": 0.72, + "learning_rate": 3.965298656760694e-06, + "loss": 0.2783, + "step": 11836 + }, + { + "epoch": 0.72, + "learning_rate": 3.963738570030135e-06, + "loss": 0.2635, + "step": 11837 + }, + { + "epoch": 0.72, + "learning_rate": 3.962178714395909e-06, + "loss": 0.267, + "step": 11838 + }, + { + "epoch": 0.72, + "learning_rate": 3.9606190899177466e-06, + "loss": 0.2789, + "step": 11839 + }, + { + "epoch": 0.72, + "learning_rate": 3.95905969665535e-06, + "loss": 0.2872, + "step": 11840 + }, + { + "epoch": 0.72, + "learning_rate": 3.957500534668423e-06, + "loss": 0.3004, + "step": 11841 + }, + { + "epoch": 0.72, + "learning_rate": 3.955941604016659e-06, + "loss": 0.2946, + "step": 11842 + }, + { + "epoch": 0.72, + "learning_rate": 3.954382904759731e-06, + "loss": 0.2789, + "step": 11843 + }, + { + "epoch": 0.72, + "learning_rate": 3.952824436957326e-06, + "loss": 0.3103, + "step": 11844 + }, + { + "epoch": 0.72, + "learning_rate": 3.951266200669103e-06, + "loss": 0.2888, + "step": 11845 + }, + { + "epoch": 0.72, + "learning_rate": 3.94970819595472e-06, + "loss": 0.2793, + "step": 11846 + }, + { + "epoch": 0.72, + "learning_rate": 3.948150422873824e-06, + "loss": 0.2881, + "step": 11847 + }, + { + "epoch": 0.72, + "learning_rate": 3.9465928814860524e-06, + "loss": 0.2814, + "step": 11848 + }, + { + "epoch": 0.72, + "learning_rate": 3.94503557185104e-06, + "loss": 0.275, + "step": 11849 + }, + { + "epoch": 0.72, + "learning_rate": 3.943478494028406e-06, + "loss": 0.2819, + "step": 11850 + }, + { + "epoch": 0.72, + "learning_rate": 3.94192164807776e-06, + "loss": 0.3036, + "step": 11851 + }, + { + "epoch": 0.72, + "learning_rate": 3.9403650340587074e-06, + "loss": 0.2514, + "step": 11852 + }, + { + "epoch": 0.72, + "learning_rate": 3.938808652030839e-06, + "loss": 0.2832, + "step": 11853 + }, + { + "epoch": 0.72, + "learning_rate": 3.937252502053748e-06, + "loss": 0.2885, + "step": 11854 + }, + { + "epoch": 0.72, + "learning_rate": 3.935696584187005e-06, + "loss": 0.2698, + "step": 11855 + }, + { + "epoch": 0.72, + "learning_rate": 3.934140898490182e-06, + "loss": 0.2637, + "step": 11856 + }, + { + "epoch": 0.72, + "learning_rate": 3.932585445022834e-06, + "loss": 0.2808, + "step": 11857 + }, + { + "epoch": 0.72, + "learning_rate": 3.93103022384451e-06, + "loss": 0.2773, + "step": 11858 + }, + { + "epoch": 0.72, + "learning_rate": 3.929475235014756e-06, + "loss": 0.3037, + "step": 11859 + }, + { + "epoch": 0.72, + "learning_rate": 3.927920478593103e-06, + "loss": 0.2843, + "step": 11860 + }, + { + "epoch": 0.72, + "learning_rate": 3.926365954639073e-06, + "loss": 0.3085, + "step": 11861 + }, + { + "epoch": 0.72, + "learning_rate": 3.9248116632121815e-06, + "loss": 0.291, + "step": 11862 + }, + { + "epoch": 0.72, + "learning_rate": 3.923257604371928e-06, + "loss": 0.2676, + "step": 11863 + }, + { + "epoch": 0.72, + "learning_rate": 3.92170377817782e-06, + "loss": 0.2789, + "step": 11864 + }, + { + "epoch": 0.72, + "learning_rate": 3.920150184689339e-06, + "loss": 0.2715, + "step": 11865 + }, + { + "epoch": 0.72, + "learning_rate": 3.918596823965964e-06, + "loss": 0.2713, + "step": 11866 + }, + { + "epoch": 0.72, + "learning_rate": 3.917043696067166e-06, + "loss": 0.2977, + "step": 11867 + }, + { + "epoch": 0.72, + "learning_rate": 3.9154908010524026e-06, + "loss": 0.2599, + "step": 11868 + }, + { + "epoch": 0.72, + "learning_rate": 3.913938138981132e-06, + "loss": 0.2985, + "step": 11869 + }, + { + "epoch": 0.72, + "learning_rate": 3.912385709912794e-06, + "loss": 0.2818, + "step": 11870 + }, + { + "epoch": 0.72, + "learning_rate": 3.910833513906822e-06, + "loss": 0.3161, + "step": 11871 + }, + { + "epoch": 0.72, + "learning_rate": 3.909281551022643e-06, + "loss": 0.3278, + "step": 11872 + }, + { + "epoch": 0.72, + "learning_rate": 3.90772982131967e-06, + "loss": 0.2801, + "step": 11873 + }, + { + "epoch": 0.72, + "learning_rate": 3.906178324857316e-06, + "loss": 0.2709, + "step": 11874 + }, + { + "epoch": 0.72, + "learning_rate": 3.904627061694977e-06, + "loss": 0.3071, + "step": 11875 + }, + { + "epoch": 0.72, + "learning_rate": 3.903076031892046e-06, + "loss": 0.2811, + "step": 11876 + }, + { + "epoch": 0.72, + "learning_rate": 3.901525235507893e-06, + "loss": 0.2617, + "step": 11877 + }, + { + "epoch": 0.72, + "learning_rate": 3.8999746726018995e-06, + "loss": 0.2761, + "step": 11878 + }, + { + "epoch": 0.72, + "learning_rate": 3.898424343233427e-06, + "loss": 0.2872, + "step": 11879 + }, + { + "epoch": 0.72, + "learning_rate": 3.896874247461826e-06, + "loss": 0.2963, + "step": 11880 + }, + { + "epoch": 0.72, + "learning_rate": 3.895324385346445e-06, + "loss": 0.2833, + "step": 11881 + }, + { + "epoch": 0.72, + "learning_rate": 3.893774756946618e-06, + "loss": 0.2796, + "step": 11882 + }, + { + "epoch": 0.72, + "learning_rate": 3.8922253623216685e-06, + "loss": 0.3153, + "step": 11883 + }, + { + "epoch": 0.72, + "learning_rate": 3.890676201530923e-06, + "loss": 0.2891, + "step": 11884 + }, + { + "epoch": 0.72, + "learning_rate": 3.889127274633685e-06, + "loss": 0.3273, + "step": 11885 + }, + { + "epoch": 0.72, + "learning_rate": 3.887578581689256e-06, + "loss": 0.2997, + "step": 11886 + }, + { + "epoch": 0.72, + "learning_rate": 3.886030122756928e-06, + "loss": 0.2906, + "step": 11887 + }, + { + "epoch": 0.72, + "learning_rate": 3.884481897895977e-06, + "loss": 0.2921, + "step": 11888 + }, + { + "epoch": 0.72, + "learning_rate": 3.8829339071656856e-06, + "loss": 0.2765, + "step": 11889 + }, + { + "epoch": 0.72, + "learning_rate": 3.881386150625315e-06, + "loss": 0.2792, + "step": 11890 + }, + { + "epoch": 0.72, + "learning_rate": 3.8798386283341185e-06, + "loss": 0.2949, + "step": 11891 + }, + { + "epoch": 0.72, + "learning_rate": 3.8782913403513435e-06, + "loss": 0.2919, + "step": 11892 + }, + { + "epoch": 0.72, + "learning_rate": 3.8767442867362245e-06, + "loss": 0.3054, + "step": 11893 + }, + { + "epoch": 0.72, + "learning_rate": 3.8751974675479964e-06, + "loss": 0.2973, + "step": 11894 + }, + { + "epoch": 0.72, + "learning_rate": 3.873650882845876e-06, + "loss": 0.2914, + "step": 11895 + }, + { + "epoch": 0.72, + "learning_rate": 3.872104532689073e-06, + "loss": 0.2588, + "step": 11896 + }, + { + "epoch": 0.72, + "learning_rate": 3.870558417136788e-06, + "loss": 0.2776, + "step": 11897 + }, + { + "epoch": 0.72, + "learning_rate": 3.8690125362482114e-06, + "loss": 0.2871, + "step": 11898 + }, + { + "epoch": 0.72, + "learning_rate": 3.867466890082534e-06, + "loss": 0.2598, + "step": 11899 + }, + { + "epoch": 0.72, + "learning_rate": 3.865921478698925e-06, + "loss": 0.315, + "step": 11900 + }, + { + "epoch": 0.72, + "learning_rate": 3.8643763021565525e-06, + "loss": 0.2849, + "step": 11901 + }, + { + "epoch": 0.72, + "learning_rate": 3.862831360514572e-06, + "loss": 0.2644, + "step": 11902 + }, + { + "epoch": 0.72, + "learning_rate": 3.8612866538321255e-06, + "loss": 0.2924, + "step": 11903 + }, + { + "epoch": 0.72, + "learning_rate": 3.859742182168363e-06, + "loss": 0.2727, + "step": 11904 + }, + { + "epoch": 0.72, + "learning_rate": 3.858197945582406e-06, + "loss": 0.2829, + "step": 11905 + }, + { + "epoch": 0.72, + "learning_rate": 3.8566539441333785e-06, + "loss": 0.2803, + "step": 11906 + }, + { + "epoch": 0.72, + "learning_rate": 3.8551101778803905e-06, + "loss": 0.2896, + "step": 11907 + }, + { + "epoch": 0.72, + "learning_rate": 3.853566646882542e-06, + "loss": 0.293, + "step": 11908 + }, + { + "epoch": 0.72, + "learning_rate": 3.852023351198933e-06, + "loss": 0.2479, + "step": 11909 + }, + { + "epoch": 0.72, + "learning_rate": 3.850480290888646e-06, + "loss": 0.2812, + "step": 11910 + }, + { + "epoch": 0.72, + "learning_rate": 3.848937466010754e-06, + "loss": 0.2866, + "step": 11911 + }, + { + "epoch": 0.72, + "learning_rate": 3.847394876624321e-06, + "loss": 0.2864, + "step": 11912 + }, + { + "epoch": 0.72, + "learning_rate": 3.845852522788412e-06, + "loss": 0.2955, + "step": 11913 + }, + { + "epoch": 0.72, + "learning_rate": 3.84431040456207e-06, + "loss": 0.2824, + "step": 11914 + }, + { + "epoch": 0.72, + "learning_rate": 3.8427685220043375e-06, + "loss": 0.2997, + "step": 11915 + }, + { + "epoch": 0.72, + "learning_rate": 3.841226875174243e-06, + "loss": 0.2727, + "step": 11916 + }, + { + "epoch": 0.72, + "learning_rate": 3.83968546413081e-06, + "loss": 0.2838, + "step": 11917 + }, + { + "epoch": 0.72, + "learning_rate": 3.838144288933045e-06, + "loss": 0.2659, + "step": 11918 + }, + { + "epoch": 0.72, + "learning_rate": 3.836603349639961e-06, + "loss": 0.2938, + "step": 11919 + }, + { + "epoch": 0.72, + "learning_rate": 3.835062646310547e-06, + "loss": 0.2773, + "step": 11920 + }, + { + "epoch": 0.72, + "learning_rate": 3.833522179003788e-06, + "loss": 0.2847, + "step": 11921 + }, + { + "epoch": 0.72, + "learning_rate": 3.831981947778661e-06, + "loss": 0.2856, + "step": 11922 + }, + { + "epoch": 0.72, + "learning_rate": 3.83044195269413e-06, + "loss": 0.2757, + "step": 11923 + }, + { + "epoch": 0.72, + "learning_rate": 3.828902193809161e-06, + "loss": 0.2945, + "step": 11924 + }, + { + "epoch": 0.72, + "learning_rate": 3.827362671182698e-06, + "loss": 0.2631, + "step": 11925 + }, + { + "epoch": 0.72, + "learning_rate": 3.8258233848736814e-06, + "loss": 0.2913, + "step": 11926 + }, + { + "epoch": 0.72, + "learning_rate": 3.824284334941044e-06, + "loss": 0.2694, + "step": 11927 + }, + { + "epoch": 0.72, + "learning_rate": 3.822745521443702e-06, + "loss": 0.2578, + "step": 11928 + }, + { + "epoch": 0.72, + "learning_rate": 3.821206944440577e-06, + "loss": 0.2956, + "step": 11929 + }, + { + "epoch": 0.72, + "learning_rate": 3.819668603990568e-06, + "loss": 0.2852, + "step": 11930 + }, + { + "epoch": 0.72, + "learning_rate": 3.818130500152573e-06, + "loss": 0.2812, + "step": 11931 + }, + { + "epoch": 0.72, + "learning_rate": 3.816592632985473e-06, + "loss": 0.2681, + "step": 11932 + }, + { + "epoch": 0.72, + "learning_rate": 3.815055002548145e-06, + "loss": 0.2681, + "step": 11933 + }, + { + "epoch": 0.72, + "learning_rate": 3.813517608899463e-06, + "loss": 0.2762, + "step": 11934 + }, + { + "epoch": 0.72, + "learning_rate": 3.8119804520982807e-06, + "loss": 0.2884, + "step": 11935 + }, + { + "epoch": 0.72, + "learning_rate": 3.8104435322034483e-06, + "loss": 0.2764, + "step": 11936 + }, + { + "epoch": 0.72, + "learning_rate": 3.808906849273808e-06, + "loss": 0.2794, + "step": 11937 + }, + { + "epoch": 0.72, + "learning_rate": 3.807370403368186e-06, + "loss": 0.2571, + "step": 11938 + }, + { + "epoch": 0.72, + "learning_rate": 3.8058341945454102e-06, + "loss": 0.2833, + "step": 11939 + }, + { + "epoch": 0.72, + "learning_rate": 3.8042982228642945e-06, + "loss": 0.2903, + "step": 11940 + }, + { + "epoch": 0.72, + "learning_rate": 3.802762488383639e-06, + "loss": 0.2691, + "step": 11941 + }, + { + "epoch": 0.72, + "learning_rate": 3.8012269911622414e-06, + "loss": 0.291, + "step": 11942 + }, + { + "epoch": 0.72, + "learning_rate": 3.799691731258882e-06, + "loss": 0.2695, + "step": 11943 + }, + { + "epoch": 0.72, + "learning_rate": 3.798156708732351e-06, + "loss": 0.2482, + "step": 11944 + }, + { + "epoch": 0.72, + "learning_rate": 3.7966219236414036e-06, + "loss": 0.2676, + "step": 11945 + }, + { + "epoch": 0.72, + "learning_rate": 3.7950873760448036e-06, + "loss": 0.2697, + "step": 11946 + }, + { + "epoch": 0.72, + "learning_rate": 3.7935530660012953e-06, + "loss": 0.292, + "step": 11947 + }, + { + "epoch": 0.72, + "learning_rate": 3.7920189935696284e-06, + "loss": 0.2748, + "step": 11948 + }, + { + "epoch": 0.72, + "learning_rate": 3.790485158808529e-06, + "loss": 0.2896, + "step": 11949 + }, + { + "epoch": 0.72, + "learning_rate": 3.7889515617767203e-06, + "loss": 0.2619, + "step": 11950 + }, + { + "epoch": 0.72, + "learning_rate": 3.7874182025329155e-06, + "loss": 0.2921, + "step": 11951 + }, + { + "epoch": 0.72, + "learning_rate": 3.7858850811358184e-06, + "loss": 0.2878, + "step": 11952 + }, + { + "epoch": 0.72, + "learning_rate": 3.784352197644121e-06, + "loss": 0.2645, + "step": 11953 + }, + { + "epoch": 0.72, + "learning_rate": 3.782819552116517e-06, + "loss": 0.2867, + "step": 11954 + }, + { + "epoch": 0.72, + "learning_rate": 3.781287144611678e-06, + "loss": 0.2879, + "step": 11955 + }, + { + "epoch": 0.72, + "learning_rate": 3.7797549751882734e-06, + "loss": 0.2746, + "step": 11956 + }, + { + "epoch": 0.72, + "learning_rate": 3.7782230439049606e-06, + "loss": 0.2944, + "step": 11957 + }, + { + "epoch": 0.72, + "learning_rate": 3.776691350820386e-06, + "loss": 0.2919, + "step": 11958 + }, + { + "epoch": 0.72, + "learning_rate": 3.775159895993198e-06, + "loss": 0.2878, + "step": 11959 + }, + { + "epoch": 0.72, + "learning_rate": 3.773628679482022e-06, + "loss": 0.291, + "step": 11960 + }, + { + "epoch": 0.72, + "learning_rate": 3.7720977013454817e-06, + "loss": 0.2732, + "step": 11961 + }, + { + "epoch": 0.72, + "learning_rate": 3.7705669616421914e-06, + "loss": 0.2733, + "step": 11962 + }, + { + "epoch": 0.72, + "learning_rate": 3.7690364604307492e-06, + "loss": 0.2803, + "step": 11963 + }, + { + "epoch": 0.72, + "learning_rate": 3.7675061977697567e-06, + "loss": 0.2528, + "step": 11964 + }, + { + "epoch": 0.72, + "learning_rate": 3.7659761737177993e-06, + "loss": 0.308, + "step": 11965 + }, + { + "epoch": 0.72, + "learning_rate": 3.7644463883334494e-06, + "loss": 0.3047, + "step": 11966 + }, + { + "epoch": 0.72, + "learning_rate": 3.762916841675277e-06, + "loss": 0.2966, + "step": 11967 + }, + { + "epoch": 0.72, + "learning_rate": 3.7613875338018345e-06, + "loss": 0.2775, + "step": 11968 + }, + { + "epoch": 0.72, + "learning_rate": 3.7598584647716808e-06, + "loss": 0.2811, + "step": 11969 + }, + { + "epoch": 0.72, + "learning_rate": 3.7583296346433505e-06, + "loss": 0.2882, + "step": 11970 + }, + { + "epoch": 0.72, + "learning_rate": 3.7568010434753755e-06, + "loss": 0.2772, + "step": 11971 + }, + { + "epoch": 0.72, + "learning_rate": 3.755272691326276e-06, + "loss": 0.2867, + "step": 11972 + }, + { + "epoch": 0.72, + "learning_rate": 3.7537445782545612e-06, + "loss": 0.2885, + "step": 11973 + }, + { + "epoch": 0.72, + "learning_rate": 3.752216704318743e-06, + "loss": 0.32, + "step": 11974 + }, + { + "epoch": 0.72, + "learning_rate": 3.75068906957731e-06, + "loss": 0.2921, + "step": 11975 + }, + { + "epoch": 0.72, + "learning_rate": 3.7491616740887493e-06, + "loss": 0.2422, + "step": 11976 + }, + { + "epoch": 0.72, + "learning_rate": 3.7476345179115346e-06, + "loss": 0.2595, + "step": 11977 + }, + { + "epoch": 0.72, + "learning_rate": 3.746107601104134e-06, + "loss": 0.2879, + "step": 11978 + }, + { + "epoch": 0.72, + "learning_rate": 3.744580923725005e-06, + "loss": 0.3001, + "step": 11979 + }, + { + "epoch": 0.72, + "learning_rate": 3.7430544858325946e-06, + "loss": 0.2776, + "step": 11980 + }, + { + "epoch": 0.72, + "learning_rate": 3.7415282874853443e-06, + "loss": 0.2824, + "step": 11981 + }, + { + "epoch": 0.72, + "learning_rate": 3.7400023287416787e-06, + "loss": 0.2776, + "step": 11982 + }, + { + "epoch": 0.72, + "learning_rate": 3.7384766096600255e-06, + "loss": 0.2919, + "step": 11983 + }, + { + "epoch": 0.72, + "learning_rate": 3.736951130298795e-06, + "loss": 0.2623, + "step": 11984 + }, + { + "epoch": 0.72, + "learning_rate": 3.735425890716388e-06, + "loss": 0.2799, + "step": 11985 + }, + { + "epoch": 0.72, + "learning_rate": 3.733900890971197e-06, + "loss": 0.2591, + "step": 11986 + }, + { + "epoch": 0.72, + "learning_rate": 3.7323761311216076e-06, + "loss": 0.2824, + "step": 11987 + }, + { + "epoch": 0.72, + "learning_rate": 3.730851611225992e-06, + "loss": 0.258, + "step": 11988 + }, + { + "epoch": 0.72, + "learning_rate": 3.7293273313427203e-06, + "loss": 0.2923, + "step": 11989 + }, + { + "epoch": 0.72, + "learning_rate": 3.727803291530149e-06, + "loss": 0.2566, + "step": 11990 + }, + { + "epoch": 0.72, + "learning_rate": 3.7262794918466218e-06, + "loss": 0.2508, + "step": 11991 + }, + { + "epoch": 0.72, + "learning_rate": 3.724755932350479e-06, + "loss": 0.2782, + "step": 11992 + }, + { + "epoch": 0.72, + "learning_rate": 3.7232326131000463e-06, + "loss": 0.2743, + "step": 11993 + }, + { + "epoch": 0.72, + "learning_rate": 3.72170953415365e-06, + "loss": 0.2773, + "step": 11994 + }, + { + "epoch": 0.72, + "learning_rate": 3.7201866955695963e-06, + "loss": 0.2966, + "step": 11995 + }, + { + "epoch": 0.72, + "learning_rate": 3.7186640974061885e-06, + "loss": 0.2847, + "step": 11996 + }, + { + "epoch": 0.72, + "learning_rate": 3.717141739721717e-06, + "loss": 0.2864, + "step": 11997 + }, + { + "epoch": 0.72, + "learning_rate": 3.7156196225744623e-06, + "loss": 0.2712, + "step": 11998 + }, + { + "epoch": 0.72, + "learning_rate": 3.7140977460227045e-06, + "loss": 0.2659, + "step": 11999 + }, + { + "epoch": 0.72, + "learning_rate": 3.712576110124705e-06, + "loss": 0.2921, + "step": 12000 + }, + { + "epoch": 0.73, + "learning_rate": 3.71105471493872e-06, + "loss": 0.2969, + "step": 12001 + }, + { + "epoch": 0.73, + "learning_rate": 3.709533560522994e-06, + "loss": 0.2693, + "step": 12002 + }, + { + "epoch": 0.73, + "learning_rate": 3.708012646935761e-06, + "loss": 0.2708, + "step": 12003 + }, + { + "epoch": 0.73, + "learning_rate": 3.706491974235257e-06, + "loss": 0.2977, + "step": 12004 + }, + { + "epoch": 0.73, + "learning_rate": 3.704971542479695e-06, + "loss": 0.2513, + "step": 12005 + }, + { + "epoch": 0.73, + "learning_rate": 3.7034513517272854e-06, + "loss": 0.2804, + "step": 12006 + }, + { + "epoch": 0.73, + "learning_rate": 3.7019314020362275e-06, + "loss": 0.2796, + "step": 12007 + }, + { + "epoch": 0.73, + "learning_rate": 3.7004116934647082e-06, + "loss": 0.2938, + "step": 12008 + }, + { + "epoch": 0.73, + "learning_rate": 3.698892226070918e-06, + "loss": 0.2838, + "step": 12009 + }, + { + "epoch": 0.73, + "learning_rate": 3.697372999913025e-06, + "loss": 0.2792, + "step": 12010 + }, + { + "epoch": 0.73, + "learning_rate": 3.6958540150491917e-06, + "loss": 0.3041, + "step": 12011 + }, + { + "epoch": 0.73, + "learning_rate": 3.6943352715375712e-06, + "loss": 0.2846, + "step": 12012 + }, + { + "epoch": 0.73, + "learning_rate": 3.6928167694363094e-06, + "loss": 0.298, + "step": 12013 + }, + { + "epoch": 0.73, + "learning_rate": 3.6912985088035414e-06, + "loss": 0.2924, + "step": 12014 + }, + { + "epoch": 0.73, + "learning_rate": 3.6897804896973932e-06, + "loss": 0.2801, + "step": 12015 + }, + { + "epoch": 0.73, + "learning_rate": 3.688262712175982e-06, + "loss": 0.2686, + "step": 12016 + }, + { + "epoch": 0.73, + "learning_rate": 3.6867451762974117e-06, + "loss": 0.3002, + "step": 12017 + }, + { + "epoch": 0.73, + "learning_rate": 3.6852278821197874e-06, + "loss": 0.2612, + "step": 12018 + }, + { + "epoch": 0.73, + "learning_rate": 3.683710829701195e-06, + "loss": 0.2681, + "step": 12019 + }, + { + "epoch": 0.73, + "learning_rate": 3.682194019099714e-06, + "loss": 0.2997, + "step": 12020 + }, + { + "epoch": 0.73, + "learning_rate": 3.6806774503734157e-06, + "loss": 0.2691, + "step": 12021 + }, + { + "epoch": 0.73, + "learning_rate": 3.6791611235803603e-06, + "loss": 0.2815, + "step": 12022 + }, + { + "epoch": 0.73, + "learning_rate": 3.6776450387785976e-06, + "loss": 0.2831, + "step": 12023 + }, + { + "epoch": 0.73, + "learning_rate": 3.676129196026177e-06, + "loss": 0.2931, + "step": 12024 + }, + { + "epoch": 0.73, + "learning_rate": 3.6746135953811278e-06, + "loss": 0.2626, + "step": 12025 + }, + { + "epoch": 0.73, + "learning_rate": 3.673098236901477e-06, + "loss": 0.2457, + "step": 12026 + }, + { + "epoch": 0.73, + "learning_rate": 3.6715831206452357e-06, + "loss": 0.2676, + "step": 12027 + }, + { + "epoch": 0.73, + "learning_rate": 3.6700682466704084e-06, + "loss": 0.2609, + "step": 12028 + }, + { + "epoch": 0.73, + "learning_rate": 3.668553615034999e-06, + "loss": 0.2789, + "step": 12029 + }, + { + "epoch": 0.73, + "learning_rate": 3.667039225796989e-06, + "loss": 0.2764, + "step": 12030 + }, + { + "epoch": 0.73, + "learning_rate": 3.6655250790143592e-06, + "loss": 0.2726, + "step": 12031 + }, + { + "epoch": 0.73, + "learning_rate": 3.6640111747450757e-06, + "loss": 0.2663, + "step": 12032 + }, + { + "epoch": 0.73, + "learning_rate": 3.662497513047096e-06, + "loss": 0.2638, + "step": 12033 + }, + { + "epoch": 0.73, + "learning_rate": 3.660984093978376e-06, + "loss": 0.2881, + "step": 12034 + }, + { + "epoch": 0.73, + "learning_rate": 3.6594709175968547e-06, + "loss": 0.3079, + "step": 12035 + }, + { + "epoch": 0.73, + "learning_rate": 3.6579579839604605e-06, + "loss": 0.2712, + "step": 12036 + }, + { + "epoch": 0.73, + "learning_rate": 3.656445293127119e-06, + "loss": 0.2796, + "step": 12037 + }, + { + "epoch": 0.73, + "learning_rate": 3.6549328451547372e-06, + "loss": 0.3164, + "step": 12038 + }, + { + "epoch": 0.73, + "learning_rate": 3.653420640101227e-06, + "loss": 0.2556, + "step": 12039 + }, + { + "epoch": 0.73, + "learning_rate": 3.6519086780244785e-06, + "loss": 0.2794, + "step": 12040 + }, + { + "epoch": 0.73, + "learning_rate": 3.6503969589823774e-06, + "loss": 0.2764, + "step": 12041 + }, + { + "epoch": 0.73, + "learning_rate": 3.648885483032798e-06, + "loss": 0.2977, + "step": 12042 + }, + { + "epoch": 0.73, + "learning_rate": 3.647374250233604e-06, + "loss": 0.2898, + "step": 12043 + }, + { + "epoch": 0.73, + "learning_rate": 3.6458632606426602e-06, + "loss": 0.2799, + "step": 12044 + }, + { + "epoch": 0.73, + "learning_rate": 3.6443525143178103e-06, + "loss": 0.2896, + "step": 12045 + }, + { + "epoch": 0.73, + "learning_rate": 3.642842011316895e-06, + "loss": 0.2924, + "step": 12046 + }, + { + "epoch": 0.73, + "learning_rate": 3.641331751697735e-06, + "loss": 0.2743, + "step": 12047 + }, + { + "epoch": 0.73, + "learning_rate": 3.6398217355181597e-06, + "loss": 0.2786, + "step": 12048 + }, + { + "epoch": 0.73, + "learning_rate": 3.6383119628359753e-06, + "loss": 0.2672, + "step": 12049 + }, + { + "epoch": 0.73, + "learning_rate": 3.6368024337089847e-06, + "loss": 0.2833, + "step": 12050 + }, + { + "epoch": 0.73, + "learning_rate": 3.6352931481949784e-06, + "loss": 0.2789, + "step": 12051 + }, + { + "epoch": 0.73, + "learning_rate": 3.6337841063517355e-06, + "loss": 0.2942, + "step": 12052 + }, + { + "epoch": 0.73, + "learning_rate": 3.6322753082370364e-06, + "loss": 0.2829, + "step": 12053 + }, + { + "epoch": 0.73, + "learning_rate": 3.630766753908641e-06, + "loss": 0.2602, + "step": 12054 + }, + { + "epoch": 0.73, + "learning_rate": 3.629258443424305e-06, + "loss": 0.2955, + "step": 12055 + }, + { + "epoch": 0.73, + "learning_rate": 3.627750376841773e-06, + "loss": 0.2603, + "step": 12056 + }, + { + "epoch": 0.73, + "learning_rate": 3.6262425542187807e-06, + "loss": 0.2786, + "step": 12057 + }, + { + "epoch": 0.73, + "learning_rate": 3.6247349756130512e-06, + "loss": 0.2687, + "step": 12058 + }, + { + "epoch": 0.73, + "learning_rate": 3.6232276410823076e-06, + "loss": 0.2704, + "step": 12059 + }, + { + "epoch": 0.73, + "learning_rate": 3.621720550684256e-06, + "loss": 0.2628, + "step": 12060 + }, + { + "epoch": 0.73, + "learning_rate": 3.620213704476594e-06, + "loss": 0.272, + "step": 12061 + }, + { + "epoch": 0.73, + "learning_rate": 3.6187071025170106e-06, + "loss": 0.2638, + "step": 12062 + }, + { + "epoch": 0.73, + "learning_rate": 3.617200744863183e-06, + "loss": 0.2792, + "step": 12063 + }, + { + "epoch": 0.73, + "learning_rate": 3.6156946315727883e-06, + "loss": 0.2886, + "step": 12064 + }, + { + "epoch": 0.73, + "learning_rate": 3.614188762703482e-06, + "loss": 0.2506, + "step": 12065 + }, + { + "epoch": 0.73, + "learning_rate": 3.612683138312919e-06, + "loss": 0.2584, + "step": 12066 + }, + { + "epoch": 0.73, + "learning_rate": 3.6111777584587393e-06, + "loss": 0.263, + "step": 12067 + }, + { + "epoch": 0.73, + "learning_rate": 3.6096726231985736e-06, + "loss": 0.2652, + "step": 12068 + }, + { + "epoch": 0.73, + "learning_rate": 3.6081677325900523e-06, + "loss": 0.3052, + "step": 12069 + }, + { + "epoch": 0.73, + "learning_rate": 3.6066630866907857e-06, + "loss": 0.2603, + "step": 12070 + }, + { + "epoch": 0.73, + "learning_rate": 3.6051586855583787e-06, + "loss": 0.2779, + "step": 12071 + }, + { + "epoch": 0.73, + "learning_rate": 3.603654529250428e-06, + "loss": 0.2697, + "step": 12072 + }, + { + "epoch": 0.73, + "learning_rate": 3.6021506178245145e-06, + "loss": 0.2881, + "step": 12073 + }, + { + "epoch": 0.73, + "learning_rate": 3.6006469513382224e-06, + "loss": 0.2868, + "step": 12074 + }, + { + "epoch": 0.73, + "learning_rate": 3.5991435298491173e-06, + "loss": 0.2701, + "step": 12075 + }, + { + "epoch": 0.73, + "learning_rate": 3.5976403534147554e-06, + "loss": 0.2822, + "step": 12076 + }, + { + "epoch": 0.73, + "learning_rate": 3.596137422092686e-06, + "loss": 0.3041, + "step": 12077 + }, + { + "epoch": 0.73, + "learning_rate": 3.5946347359404443e-06, + "loss": 0.3156, + "step": 12078 + }, + { + "epoch": 0.73, + "learning_rate": 3.593132295015568e-06, + "loss": 0.2835, + "step": 12079 + }, + { + "epoch": 0.73, + "learning_rate": 3.591630099375577e-06, + "loss": 0.2832, + "step": 12080 + }, + { + "epoch": 0.73, + "learning_rate": 3.590128149077975e-06, + "loss": 0.2644, + "step": 12081 + }, + { + "epoch": 0.73, + "learning_rate": 3.5886264441802656e-06, + "loss": 0.2797, + "step": 12082 + }, + { + "epoch": 0.73, + "learning_rate": 3.587124984739946e-06, + "loss": 0.2861, + "step": 12083 + }, + { + "epoch": 0.73, + "learning_rate": 3.585623770814496e-06, + "loss": 0.2652, + "step": 12084 + }, + { + "epoch": 0.73, + "learning_rate": 3.584122802461388e-06, + "loss": 0.2885, + "step": 12085 + }, + { + "epoch": 0.73, + "learning_rate": 3.5826220797380896e-06, + "loss": 0.2663, + "step": 12086 + }, + { + "epoch": 0.73, + "learning_rate": 3.581121602702049e-06, + "loss": 0.2665, + "step": 12087 + }, + { + "epoch": 0.73, + "learning_rate": 3.5796213714107196e-06, + "loss": 0.2688, + "step": 12088 + }, + { + "epoch": 0.73, + "learning_rate": 3.5781213859215334e-06, + "loss": 0.2875, + "step": 12089 + }, + { + "epoch": 0.73, + "learning_rate": 3.576621646291917e-06, + "loss": 0.2761, + "step": 12090 + }, + { + "epoch": 0.73, + "learning_rate": 3.5751221525792877e-06, + "loss": 0.2775, + "step": 12091 + }, + { + "epoch": 0.73, + "learning_rate": 3.573622904841053e-06, + "loss": 0.2804, + "step": 12092 + }, + { + "epoch": 0.73, + "learning_rate": 3.5721239031346067e-06, + "loss": 0.2628, + "step": 12093 + }, + { + "epoch": 0.73, + "learning_rate": 3.5706251475173457e-06, + "loss": 0.2686, + "step": 12094 + }, + { + "epoch": 0.73, + "learning_rate": 3.5691266380466462e-06, + "loss": 0.2927, + "step": 12095 + }, + { + "epoch": 0.73, + "learning_rate": 3.567628374779877e-06, + "loss": 0.2713, + "step": 12096 + }, + { + "epoch": 0.73, + "learning_rate": 3.566130357774399e-06, + "loss": 0.2765, + "step": 12097 + }, + { + "epoch": 0.73, + "learning_rate": 3.564632587087561e-06, + "loss": 0.2868, + "step": 12098 + }, + { + "epoch": 0.73, + "learning_rate": 3.563135062776709e-06, + "loss": 0.296, + "step": 12099 + }, + { + "epoch": 0.73, + "learning_rate": 3.5616377848991744e-06, + "loss": 0.2825, + "step": 12100 + }, + { + "epoch": 0.73, + "learning_rate": 3.560140753512279e-06, + "loss": 0.2847, + "step": 12101 + }, + { + "epoch": 0.73, + "learning_rate": 3.5586439686733352e-06, + "loss": 0.2754, + "step": 12102 + }, + { + "epoch": 0.73, + "learning_rate": 3.5571474304396458e-06, + "loss": 0.2999, + "step": 12103 + }, + { + "epoch": 0.73, + "learning_rate": 3.55565113886851e-06, + "loss": 0.2972, + "step": 12104 + }, + { + "epoch": 0.73, + "learning_rate": 3.554155094017211e-06, + "loss": 0.2631, + "step": 12105 + }, + { + "epoch": 0.73, + "learning_rate": 3.5526592959430227e-06, + "loss": 0.2874, + "step": 12106 + }, + { + "epoch": 0.73, + "learning_rate": 3.551163744703212e-06, + "loss": 0.2838, + "step": 12107 + }, + { + "epoch": 0.73, + "learning_rate": 3.549668440355033e-06, + "loss": 0.2702, + "step": 12108 + }, + { + "epoch": 0.73, + "learning_rate": 3.54817338295574e-06, + "loss": 0.2522, + "step": 12109 + }, + { + "epoch": 0.73, + "learning_rate": 3.546678572562565e-06, + "loss": 0.2818, + "step": 12110 + }, + { + "epoch": 0.73, + "learning_rate": 3.5451840092327383e-06, + "loss": 0.2879, + "step": 12111 + }, + { + "epoch": 0.73, + "learning_rate": 3.5436896930234775e-06, + "loss": 0.3125, + "step": 12112 + }, + { + "epoch": 0.73, + "learning_rate": 3.542195623991991e-06, + "loss": 0.2899, + "step": 12113 + }, + { + "epoch": 0.73, + "learning_rate": 3.5407018021954865e-06, + "loss": 0.3161, + "step": 12114 + }, + { + "epoch": 0.73, + "learning_rate": 3.5392082276911457e-06, + "loss": 0.3022, + "step": 12115 + }, + { + "epoch": 0.73, + "learning_rate": 3.5377149005361522e-06, + "loss": 0.2994, + "step": 12116 + }, + { + "epoch": 0.73, + "learning_rate": 3.5362218207876742e-06, + "loss": 0.2634, + "step": 12117 + }, + { + "epoch": 0.73, + "learning_rate": 3.5347289885028813e-06, + "loss": 0.3025, + "step": 12118 + }, + { + "epoch": 0.73, + "learning_rate": 3.5332364037389222e-06, + "loss": 0.2609, + "step": 12119 + }, + { + "epoch": 0.73, + "learning_rate": 3.531744066552939e-06, + "loss": 0.284, + "step": 12120 + }, + { + "epoch": 0.73, + "learning_rate": 3.530251977002068e-06, + "loss": 0.2853, + "step": 12121 + }, + { + "epoch": 0.73, + "learning_rate": 3.5287601351434276e-06, + "loss": 0.2638, + "step": 12122 + }, + { + "epoch": 0.73, + "learning_rate": 3.52726854103414e-06, + "loss": 0.2934, + "step": 12123 + }, + { + "epoch": 0.73, + "learning_rate": 3.525777194731308e-06, + "loss": 0.2741, + "step": 12124 + }, + { + "epoch": 0.73, + "learning_rate": 3.5242860962920254e-06, + "loss": 0.2881, + "step": 12125 + }, + { + "epoch": 0.73, + "learning_rate": 3.5227952457733796e-06, + "loss": 0.2709, + "step": 12126 + }, + { + "epoch": 0.73, + "learning_rate": 3.5213046432324437e-06, + "loss": 0.2779, + "step": 12127 + }, + { + "epoch": 0.73, + "learning_rate": 3.519814288726293e-06, + "loss": 0.2801, + "step": 12128 + }, + { + "epoch": 0.73, + "learning_rate": 3.5183241823119797e-06, + "loss": 0.2945, + "step": 12129 + }, + { + "epoch": 0.73, + "learning_rate": 3.516834324046553e-06, + "loss": 0.2765, + "step": 12130 + }, + { + "epoch": 0.73, + "learning_rate": 3.5153447139870512e-06, + "loss": 0.2852, + "step": 12131 + }, + { + "epoch": 0.73, + "learning_rate": 3.513855352190505e-06, + "loss": 0.2832, + "step": 12132 + }, + { + "epoch": 0.73, + "learning_rate": 3.512366238713929e-06, + "loss": 0.2829, + "step": 12133 + }, + { + "epoch": 0.73, + "learning_rate": 3.510877373614342e-06, + "loss": 0.2709, + "step": 12134 + }, + { + "epoch": 0.73, + "learning_rate": 3.5093887569487397e-06, + "loss": 0.2663, + "step": 12135 + }, + { + "epoch": 0.73, + "learning_rate": 3.5079003887741136e-06, + "loss": 0.2793, + "step": 12136 + }, + { + "epoch": 0.73, + "learning_rate": 3.5064122691474455e-06, + "loss": 0.2673, + "step": 12137 + }, + { + "epoch": 0.73, + "learning_rate": 3.504924398125705e-06, + "loss": 0.2718, + "step": 12138 + }, + { + "epoch": 0.73, + "learning_rate": 3.5034367757658604e-06, + "loss": 0.2691, + "step": 12139 + }, + { + "epoch": 0.73, + "learning_rate": 3.5019494021248625e-06, + "loss": 0.2746, + "step": 12140 + }, + { + "epoch": 0.73, + "learning_rate": 3.500462277259654e-06, + "loss": 0.2637, + "step": 12141 + }, + { + "epoch": 0.73, + "learning_rate": 3.4989754012271703e-06, + "loss": 0.2768, + "step": 12142 + }, + { + "epoch": 0.73, + "learning_rate": 3.497488774084331e-06, + "loss": 0.2743, + "step": 12143 + }, + { + "epoch": 0.73, + "learning_rate": 3.4960023958880594e-06, + "loss": 0.273, + "step": 12144 + }, + { + "epoch": 0.73, + "learning_rate": 3.4945162666952557e-06, + "loss": 0.2679, + "step": 12145 + }, + { + "epoch": 0.73, + "learning_rate": 3.493030386562818e-06, + "loss": 0.2687, + "step": 12146 + }, + { + "epoch": 0.73, + "learning_rate": 3.491544755547631e-06, + "loss": 0.2967, + "step": 12147 + }, + { + "epoch": 0.73, + "learning_rate": 3.4900593737065735e-06, + "loss": 0.2797, + "step": 12148 + }, + { + "epoch": 0.73, + "learning_rate": 3.4885742410965105e-06, + "loss": 0.3047, + "step": 12149 + }, + { + "epoch": 0.73, + "learning_rate": 3.4870893577743028e-06, + "loss": 0.275, + "step": 12150 + }, + { + "epoch": 0.73, + "learning_rate": 3.485604723796795e-06, + "loss": 0.2919, + "step": 12151 + }, + { + "epoch": 0.73, + "learning_rate": 3.4841203392208257e-06, + "loss": 0.2854, + "step": 12152 + }, + { + "epoch": 0.73, + "learning_rate": 3.482636204103229e-06, + "loss": 0.2595, + "step": 12153 + }, + { + "epoch": 0.73, + "learning_rate": 3.481152318500822e-06, + "loss": 0.2846, + "step": 12154 + }, + { + "epoch": 0.73, + "learning_rate": 3.479668682470415e-06, + "loss": 0.2902, + "step": 12155 + }, + { + "epoch": 0.73, + "learning_rate": 3.478185296068808e-06, + "loss": 0.2913, + "step": 12156 + }, + { + "epoch": 0.73, + "learning_rate": 3.476702159352787e-06, + "loss": 0.2436, + "step": 12157 + }, + { + "epoch": 0.73, + "learning_rate": 3.475219272379143e-06, + "loss": 0.2667, + "step": 12158 + }, + { + "epoch": 0.73, + "learning_rate": 3.4737366352046443e-06, + "loss": 0.2525, + "step": 12159 + }, + { + "epoch": 0.73, + "learning_rate": 3.472254247886051e-06, + "loss": 0.273, + "step": 12160 + }, + { + "epoch": 0.73, + "learning_rate": 3.4707721104801175e-06, + "loss": 0.2868, + "step": 12161 + }, + { + "epoch": 0.73, + "learning_rate": 3.469290223043583e-06, + "loss": 0.252, + "step": 12162 + }, + { + "epoch": 0.73, + "learning_rate": 3.467808585633188e-06, + "loss": 0.2584, + "step": 12163 + }, + { + "epoch": 0.73, + "learning_rate": 3.466327198305652e-06, + "loss": 0.2826, + "step": 12164 + }, + { + "epoch": 0.73, + "learning_rate": 3.4648460611176926e-06, + "loss": 0.2653, + "step": 12165 + }, + { + "epoch": 0.73, + "learning_rate": 3.4633651741260112e-06, + "loss": 0.3117, + "step": 12166 + }, + { + "epoch": 0.74, + "learning_rate": 3.4618845373873045e-06, + "loss": 0.3039, + "step": 12167 + }, + { + "epoch": 0.74, + "learning_rate": 3.4604041509582555e-06, + "loss": 0.2708, + "step": 12168 + }, + { + "epoch": 0.74, + "learning_rate": 3.458924014895546e-06, + "loss": 0.2524, + "step": 12169 + }, + { + "epoch": 0.74, + "learning_rate": 3.457444129255839e-06, + "loss": 0.2801, + "step": 12170 + }, + { + "epoch": 0.74, + "learning_rate": 3.4559644940957924e-06, + "loss": 0.2916, + "step": 12171 + }, + { + "epoch": 0.74, + "learning_rate": 3.4544851094720534e-06, + "loss": 0.2712, + "step": 12172 + }, + { + "epoch": 0.74, + "learning_rate": 3.453005975441256e-06, + "loss": 0.2847, + "step": 12173 + }, + { + "epoch": 0.74, + "learning_rate": 3.451527092060035e-06, + "loss": 0.2787, + "step": 12174 + }, + { + "epoch": 0.74, + "learning_rate": 3.450048459385006e-06, + "loss": 0.2905, + "step": 12175 + }, + { + "epoch": 0.74, + "learning_rate": 3.4485700774727783e-06, + "loss": 0.2944, + "step": 12176 + }, + { + "epoch": 0.74, + "learning_rate": 3.44709194637995e-06, + "loss": 0.2561, + "step": 12177 + }, + { + "epoch": 0.74, + "learning_rate": 3.445614066163109e-06, + "loss": 0.2734, + "step": 12178 + }, + { + "epoch": 0.74, + "learning_rate": 3.444136436878841e-06, + "loss": 0.3107, + "step": 12179 + }, + { + "epoch": 0.74, + "learning_rate": 3.4426590585837137e-06, + "loss": 0.2832, + "step": 12180 + }, + { + "epoch": 0.74, + "learning_rate": 3.4411819313342887e-06, + "loss": 0.2814, + "step": 12181 + }, + { + "epoch": 0.74, + "learning_rate": 3.439705055187117e-06, + "loss": 0.2776, + "step": 12182 + }, + { + "epoch": 0.74, + "learning_rate": 3.438228430198739e-06, + "loss": 0.2612, + "step": 12183 + }, + { + "epoch": 0.74, + "learning_rate": 3.4367520564256883e-06, + "loss": 0.2895, + "step": 12184 + }, + { + "epoch": 0.74, + "learning_rate": 3.4352759339244867e-06, + "loss": 0.2889, + "step": 12185 + }, + { + "epoch": 0.74, + "learning_rate": 3.433800062751648e-06, + "loss": 0.2575, + "step": 12186 + }, + { + "epoch": 0.74, + "learning_rate": 3.4323244429636716e-06, + "loss": 0.2723, + "step": 12187 + }, + { + "epoch": 0.74, + "learning_rate": 3.4308490746170565e-06, + "loss": 0.2836, + "step": 12188 + }, + { + "epoch": 0.74, + "learning_rate": 3.429373957768285e-06, + "loss": 0.2765, + "step": 12189 + }, + { + "epoch": 0.74, + "learning_rate": 3.427899092473832e-06, + "loss": 0.2779, + "step": 12190 + }, + { + "epoch": 0.74, + "learning_rate": 3.42642447879016e-06, + "loss": 0.2941, + "step": 12191 + }, + { + "epoch": 0.74, + "learning_rate": 3.4249501167737232e-06, + "loss": 0.2573, + "step": 12192 + }, + { + "epoch": 0.74, + "learning_rate": 3.4234760064809715e-06, + "loss": 0.3018, + "step": 12193 + }, + { + "epoch": 0.74, + "learning_rate": 3.4220021479683398e-06, + "loss": 0.2977, + "step": 12194 + }, + { + "epoch": 0.74, + "learning_rate": 3.4205285412922516e-06, + "loss": 0.2585, + "step": 12195 + }, + { + "epoch": 0.74, + "learning_rate": 3.4190551865091247e-06, + "loss": 0.2669, + "step": 12196 + }, + { + "epoch": 0.74, + "learning_rate": 3.417582083675365e-06, + "loss": 0.2748, + "step": 12197 + }, + { + "epoch": 0.74, + "learning_rate": 3.416109232847372e-06, + "loss": 0.2916, + "step": 12198 + }, + { + "epoch": 0.74, + "learning_rate": 3.414636634081533e-06, + "loss": 0.3089, + "step": 12199 + }, + { + "epoch": 0.74, + "learning_rate": 3.4131642874342263e-06, + "loss": 0.2886, + "step": 12200 + }, + { + "epoch": 0.74, + "learning_rate": 3.4116921929618173e-06, + "loss": 0.2859, + "step": 12201 + }, + { + "epoch": 0.74, + "learning_rate": 3.410220350720668e-06, + "loss": 0.2672, + "step": 12202 + }, + { + "epoch": 0.74, + "learning_rate": 3.408748760767122e-06, + "loss": 0.2884, + "step": 12203 + }, + { + "epoch": 0.74, + "learning_rate": 3.407277423157527e-06, + "loss": 0.2938, + "step": 12204 + }, + { + "epoch": 0.74, + "learning_rate": 3.4058063379482065e-06, + "loss": 0.275, + "step": 12205 + }, + { + "epoch": 0.74, + "learning_rate": 3.404335505195483e-06, + "loss": 0.275, + "step": 12206 + }, + { + "epoch": 0.74, + "learning_rate": 3.4028649249556665e-06, + "loss": 0.3343, + "step": 12207 + }, + { + "epoch": 0.74, + "learning_rate": 3.4013945972850536e-06, + "loss": 0.257, + "step": 12208 + }, + { + "epoch": 0.74, + "learning_rate": 3.3999245222399433e-06, + "loss": 0.3122, + "step": 12209 + }, + { + "epoch": 0.74, + "learning_rate": 3.398454699876612e-06, + "loss": 0.296, + "step": 12210 + }, + { + "epoch": 0.74, + "learning_rate": 3.3969851302513324e-06, + "loss": 0.2878, + "step": 12211 + }, + { + "epoch": 0.74, + "learning_rate": 3.395515813420366e-06, + "loss": 0.2849, + "step": 12212 + }, + { + "epoch": 0.74, + "learning_rate": 3.394046749439962e-06, + "loss": 0.3045, + "step": 12213 + }, + { + "epoch": 0.74, + "learning_rate": 3.3925779383663703e-06, + "loss": 0.2659, + "step": 12214 + }, + { + "epoch": 0.74, + "learning_rate": 3.3911093802558194e-06, + "loss": 0.266, + "step": 12215 + }, + { + "epoch": 0.74, + "learning_rate": 3.389641075164537e-06, + "loss": 0.2914, + "step": 12216 + }, + { + "epoch": 0.74, + "learning_rate": 3.388173023148725e-06, + "loss": 0.2683, + "step": 12217 + }, + { + "epoch": 0.74, + "learning_rate": 3.386705224264599e-06, + "loss": 0.2706, + "step": 12218 + }, + { + "epoch": 0.74, + "learning_rate": 3.3852376785683505e-06, + "loss": 0.2959, + "step": 12219 + }, + { + "epoch": 0.74, + "learning_rate": 3.3837703861161617e-06, + "loss": 0.2773, + "step": 12220 + }, + { + "epoch": 0.74, + "learning_rate": 3.382303346964209e-06, + "loss": 0.2828, + "step": 12221 + }, + { + "epoch": 0.74, + "learning_rate": 3.3808365611686534e-06, + "loss": 0.3083, + "step": 12222 + }, + { + "epoch": 0.74, + "learning_rate": 3.3793700287856577e-06, + "loss": 0.2953, + "step": 12223 + }, + { + "epoch": 0.74, + "learning_rate": 3.3779037498713642e-06, + "loss": 0.2722, + "step": 12224 + }, + { + "epoch": 0.74, + "learning_rate": 3.376437724481908e-06, + "loss": 0.2969, + "step": 12225 + }, + { + "epoch": 0.74, + "learning_rate": 3.3749719526734172e-06, + "loss": 0.2775, + "step": 12226 + }, + { + "epoch": 0.74, + "learning_rate": 3.373506434502003e-06, + "loss": 0.2553, + "step": 12227 + }, + { + "epoch": 0.74, + "learning_rate": 3.3720411700237797e-06, + "loss": 0.2757, + "step": 12228 + }, + { + "epoch": 0.74, + "learning_rate": 3.3705761592948416e-06, + "loss": 0.2877, + "step": 12229 + }, + { + "epoch": 0.74, + "learning_rate": 3.3691114023712767e-06, + "loss": 0.2879, + "step": 12230 + }, + { + "epoch": 0.74, + "learning_rate": 3.3676468993091617e-06, + "loss": 0.2701, + "step": 12231 + }, + { + "epoch": 0.74, + "learning_rate": 3.366182650164561e-06, + "loss": 0.2577, + "step": 12232 + }, + { + "epoch": 0.74, + "learning_rate": 3.3647186549935407e-06, + "loss": 0.2778, + "step": 12233 + }, + { + "epoch": 0.74, + "learning_rate": 3.3632549138521455e-06, + "loss": 0.2652, + "step": 12234 + }, + { + "epoch": 0.74, + "learning_rate": 3.3617914267964136e-06, + "loss": 0.2723, + "step": 12235 + }, + { + "epoch": 0.74, + "learning_rate": 3.360328193882375e-06, + "loss": 0.2994, + "step": 12236 + }, + { + "epoch": 0.74, + "learning_rate": 3.3588652151660497e-06, + "loss": 0.2889, + "step": 12237 + }, + { + "epoch": 0.74, + "learning_rate": 3.3574024907034432e-06, + "loss": 0.2691, + "step": 12238 + }, + { + "epoch": 0.74, + "learning_rate": 3.355940020550561e-06, + "loss": 0.2815, + "step": 12239 + }, + { + "epoch": 0.74, + "learning_rate": 3.354477804763392e-06, + "loss": 0.284, + "step": 12240 + }, + { + "epoch": 0.74, + "learning_rate": 3.3530158433979152e-06, + "loss": 0.2779, + "step": 12241 + }, + { + "epoch": 0.74, + "learning_rate": 3.3515541365101035e-06, + "loss": 0.2811, + "step": 12242 + }, + { + "epoch": 0.74, + "learning_rate": 3.3500926841559113e-06, + "loss": 0.268, + "step": 12243 + }, + { + "epoch": 0.74, + "learning_rate": 3.3486314863912984e-06, + "loss": 0.2532, + "step": 12244 + }, + { + "epoch": 0.74, + "learning_rate": 3.3471705432722037e-06, + "loss": 0.2907, + "step": 12245 + }, + { + "epoch": 0.74, + "learning_rate": 3.3457098548545575e-06, + "loss": 0.2807, + "step": 12246 + }, + { + "epoch": 0.74, + "learning_rate": 3.3442494211942824e-06, + "loss": 0.3051, + "step": 12247 + }, + { + "epoch": 0.74, + "learning_rate": 3.342789242347286e-06, + "loss": 0.2585, + "step": 12248 + }, + { + "epoch": 0.74, + "learning_rate": 3.3413293183694795e-06, + "loss": 0.298, + "step": 12249 + }, + { + "epoch": 0.74, + "learning_rate": 3.3398696493167546e-06, + "loss": 0.286, + "step": 12250 + }, + { + "epoch": 0.74, + "learning_rate": 3.338410235244988e-06, + "loss": 0.2797, + "step": 12251 + }, + { + "epoch": 0.74, + "learning_rate": 3.3369510762100523e-06, + "loss": 0.2715, + "step": 12252 + }, + { + "epoch": 0.74, + "learning_rate": 3.335492172267818e-06, + "loss": 0.2683, + "step": 12253 + }, + { + "epoch": 0.74, + "learning_rate": 3.334033523474136e-06, + "loss": 0.2793, + "step": 12254 + }, + { + "epoch": 0.74, + "learning_rate": 3.3325751298848487e-06, + "loss": 0.275, + "step": 12255 + }, + { + "epoch": 0.74, + "learning_rate": 3.3311169915557918e-06, + "loss": 0.2927, + "step": 12256 + }, + { + "epoch": 0.74, + "learning_rate": 3.3296591085427855e-06, + "loss": 0.2938, + "step": 12257 + }, + { + "epoch": 0.74, + "learning_rate": 3.3282014809016505e-06, + "loss": 0.2575, + "step": 12258 + }, + { + "epoch": 0.74, + "learning_rate": 3.326744108688189e-06, + "loss": 0.2645, + "step": 12259 + }, + { + "epoch": 0.74, + "learning_rate": 3.3252869919581964e-06, + "loss": 0.2628, + "step": 12260 + }, + { + "epoch": 0.74, + "learning_rate": 3.323830130767458e-06, + "loss": 0.2659, + "step": 12261 + }, + { + "epoch": 0.74, + "learning_rate": 3.322373525171745e-06, + "loss": 0.2818, + "step": 12262 + }, + { + "epoch": 0.74, + "learning_rate": 3.32091717522683e-06, + "loss": 0.2761, + "step": 12263 + }, + { + "epoch": 0.74, + "learning_rate": 3.3194610809884663e-06, + "loss": 0.2561, + "step": 12264 + }, + { + "epoch": 0.74, + "learning_rate": 3.3180052425123986e-06, + "loss": 0.2666, + "step": 12265 + }, + { + "epoch": 0.74, + "learning_rate": 3.3165496598543655e-06, + "loss": 0.273, + "step": 12266 + }, + { + "epoch": 0.74, + "learning_rate": 3.315094333070088e-06, + "loss": 0.2701, + "step": 12267 + }, + { + "epoch": 0.74, + "learning_rate": 3.31363926221529e-06, + "loss": 0.2849, + "step": 12268 + }, + { + "epoch": 0.74, + "learning_rate": 3.3121844473456757e-06, + "loss": 0.2874, + "step": 12269 + }, + { + "epoch": 0.74, + "learning_rate": 3.3107298885169416e-06, + "loss": 0.2879, + "step": 12270 + }, + { + "epoch": 0.74, + "learning_rate": 3.309275585784775e-06, + "loss": 0.2754, + "step": 12271 + }, + { + "epoch": 0.74, + "learning_rate": 3.307821539204854e-06, + "loss": 0.2944, + "step": 12272 + }, + { + "epoch": 0.74, + "learning_rate": 3.3063677488328427e-06, + "loss": 0.2754, + "step": 12273 + }, + { + "epoch": 0.74, + "learning_rate": 3.3049142147244063e-06, + "loss": 0.2792, + "step": 12274 + }, + { + "epoch": 0.74, + "learning_rate": 3.303460936935188e-06, + "loss": 0.2612, + "step": 12275 + }, + { + "epoch": 0.74, + "learning_rate": 3.3020079155208273e-06, + "loss": 0.2845, + "step": 12276 + }, + { + "epoch": 0.74, + "learning_rate": 3.3005551505369528e-06, + "loss": 0.2733, + "step": 12277 + }, + { + "epoch": 0.74, + "learning_rate": 3.299102642039178e-06, + "loss": 0.2637, + "step": 12278 + }, + { + "epoch": 0.74, + "learning_rate": 3.2976503900831215e-06, + "loss": 0.2503, + "step": 12279 + }, + { + "epoch": 0.74, + "learning_rate": 3.296198394724376e-06, + "loss": 0.2698, + "step": 12280 + }, + { + "epoch": 0.74, + "learning_rate": 3.2947466560185316e-06, + "loss": 0.2928, + "step": 12281 + }, + { + "epoch": 0.74, + "learning_rate": 3.293295174021168e-06, + "loss": 0.2531, + "step": 12282 + }, + { + "epoch": 0.74, + "learning_rate": 3.291843948787853e-06, + "loss": 0.2701, + "step": 12283 + }, + { + "epoch": 0.74, + "learning_rate": 3.290392980374153e-06, + "loss": 0.2775, + "step": 12284 + }, + { + "epoch": 0.74, + "learning_rate": 3.28894226883561e-06, + "loss": 0.2787, + "step": 12285 + }, + { + "epoch": 0.74, + "learning_rate": 3.2874918142277668e-06, + "loss": 0.2274, + "step": 12286 + }, + { + "epoch": 0.74, + "learning_rate": 3.2860416166061505e-06, + "loss": 0.2667, + "step": 12287 + }, + { + "epoch": 0.74, + "learning_rate": 3.2845916760262874e-06, + "loss": 0.3086, + "step": 12288 + }, + { + "epoch": 0.74, + "learning_rate": 3.2831419925436857e-06, + "loss": 0.2807, + "step": 12289 + }, + { + "epoch": 0.74, + "learning_rate": 3.2816925662138455e-06, + "loss": 0.2514, + "step": 12290 + }, + { + "epoch": 0.74, + "learning_rate": 3.2802433970922577e-06, + "loss": 0.2726, + "step": 12291 + }, + { + "epoch": 0.74, + "learning_rate": 3.278794485234399e-06, + "loss": 0.2981, + "step": 12292 + }, + { + "epoch": 0.74, + "learning_rate": 3.2773458306957494e-06, + "loss": 0.2955, + "step": 12293 + }, + { + "epoch": 0.74, + "learning_rate": 3.275897433531765e-06, + "loss": 0.3069, + "step": 12294 + }, + { + "epoch": 0.74, + "learning_rate": 3.2744492937978978e-06, + "loss": 0.266, + "step": 12295 + }, + { + "epoch": 0.74, + "learning_rate": 3.273001411549589e-06, + "loss": 0.2847, + "step": 12296 + }, + { + "epoch": 0.74, + "learning_rate": 3.2715537868422676e-06, + "loss": 0.2764, + "step": 12297 + }, + { + "epoch": 0.74, + "learning_rate": 3.2701064197313625e-06, + "loss": 0.281, + "step": 12298 + }, + { + "epoch": 0.74, + "learning_rate": 3.2686593102722806e-06, + "loss": 0.2969, + "step": 12299 + }, + { + "epoch": 0.74, + "learning_rate": 3.267212458520426e-06, + "loss": 0.2796, + "step": 12300 + }, + { + "epoch": 0.74, + "learning_rate": 3.2657658645311896e-06, + "loss": 0.297, + "step": 12301 + }, + { + "epoch": 0.74, + "learning_rate": 3.2643195283599517e-06, + "loss": 0.2712, + "step": 12302 + }, + { + "epoch": 0.74, + "learning_rate": 3.2628734500620894e-06, + "loss": 0.2849, + "step": 12303 + }, + { + "epoch": 0.74, + "learning_rate": 3.261427629692965e-06, + "loss": 0.296, + "step": 12304 + }, + { + "epoch": 0.74, + "learning_rate": 3.2599820673079286e-06, + "loss": 0.2885, + "step": 12305 + }, + { + "epoch": 0.74, + "learning_rate": 3.2585367629623243e-06, + "loss": 0.2663, + "step": 12306 + }, + { + "epoch": 0.74, + "learning_rate": 3.2570917167114845e-06, + "loss": 0.3027, + "step": 12307 + }, + { + "epoch": 0.74, + "learning_rate": 3.2556469286107296e-06, + "loss": 0.3114, + "step": 12308 + }, + { + "epoch": 0.74, + "learning_rate": 3.254202398715379e-06, + "loss": 0.2672, + "step": 12309 + }, + { + "epoch": 0.74, + "learning_rate": 3.252758127080734e-06, + "loss": 0.2782, + "step": 12310 + }, + { + "epoch": 0.74, + "learning_rate": 3.251314113762086e-06, + "loss": 0.3029, + "step": 12311 + }, + { + "epoch": 0.74, + "learning_rate": 3.2498703588147206e-06, + "loss": 0.2701, + "step": 12312 + }, + { + "epoch": 0.74, + "learning_rate": 3.2484268622939077e-06, + "loss": 0.281, + "step": 12313 + }, + { + "epoch": 0.74, + "learning_rate": 3.246983624254917e-06, + "loss": 0.2599, + "step": 12314 + }, + { + "epoch": 0.74, + "learning_rate": 3.2455406447529993e-06, + "loss": 0.2899, + "step": 12315 + }, + { + "epoch": 0.74, + "learning_rate": 3.2440979238433977e-06, + "loss": 0.285, + "step": 12316 + }, + { + "epoch": 0.74, + "learning_rate": 3.2426554615813488e-06, + "loss": 0.272, + "step": 12317 + }, + { + "epoch": 0.74, + "learning_rate": 3.2412132580220745e-06, + "loss": 0.2705, + "step": 12318 + }, + { + "epoch": 0.74, + "learning_rate": 3.239771313220791e-06, + "loss": 0.2653, + "step": 12319 + }, + { + "epoch": 0.74, + "learning_rate": 3.2383296272327003e-06, + "loss": 0.273, + "step": 12320 + }, + { + "epoch": 0.74, + "learning_rate": 3.2368882001129987e-06, + "loss": 0.274, + "step": 12321 + }, + { + "epoch": 0.74, + "learning_rate": 3.2354470319168674e-06, + "loss": 0.274, + "step": 12322 + }, + { + "epoch": 0.74, + "learning_rate": 3.234006122699487e-06, + "loss": 0.2733, + "step": 12323 + }, + { + "epoch": 0.74, + "learning_rate": 3.2325654725160195e-06, + "loss": 0.2764, + "step": 12324 + }, + { + "epoch": 0.74, + "learning_rate": 3.23112508142162e-06, + "loss": 0.2828, + "step": 12325 + }, + { + "epoch": 0.74, + "learning_rate": 3.229684949471431e-06, + "loss": 0.2792, + "step": 12326 + }, + { + "epoch": 0.74, + "learning_rate": 3.2282450767205873e-06, + "loss": 0.2549, + "step": 12327 + }, + { + "epoch": 0.74, + "learning_rate": 3.2268054632242196e-06, + "loss": 0.2868, + "step": 12328 + }, + { + "epoch": 0.74, + "learning_rate": 3.22536610903744e-06, + "loss": 0.298, + "step": 12329 + }, + { + "epoch": 0.74, + "learning_rate": 3.2239270142153523e-06, + "loss": 0.3076, + "step": 12330 + }, + { + "epoch": 0.74, + "learning_rate": 3.2224881788130523e-06, + "loss": 0.2606, + "step": 12331 + }, + { + "epoch": 0.75, + "learning_rate": 3.2210496028856232e-06, + "loss": 0.2819, + "step": 12332 + }, + { + "epoch": 0.75, + "learning_rate": 3.2196112864881456e-06, + "loss": 0.2789, + "step": 12333 + }, + { + "epoch": 0.75, + "learning_rate": 3.218173229675683e-06, + "loss": 0.2768, + "step": 12334 + }, + { + "epoch": 0.75, + "learning_rate": 3.2167354325032894e-06, + "loss": 0.2641, + "step": 12335 + }, + { + "epoch": 0.75, + "learning_rate": 3.2152978950260116e-06, + "loss": 0.2748, + "step": 12336 + }, + { + "epoch": 0.75, + "learning_rate": 3.213860617298883e-06, + "loss": 0.2979, + "step": 12337 + }, + { + "epoch": 0.75, + "learning_rate": 3.2124235993769327e-06, + "loss": 0.2924, + "step": 12338 + }, + { + "epoch": 0.75, + "learning_rate": 3.2109868413151757e-06, + "loss": 0.2686, + "step": 12339 + }, + { + "epoch": 0.75, + "learning_rate": 3.209550343168617e-06, + "loss": 0.2736, + "step": 12340 + }, + { + "epoch": 0.75, + "learning_rate": 3.2081141049922535e-06, + "loss": 0.2518, + "step": 12341 + }, + { + "epoch": 0.75, + "learning_rate": 3.2066781268410696e-06, + "loss": 0.2854, + "step": 12342 + }, + { + "epoch": 0.75, + "learning_rate": 3.2052424087700384e-06, + "loss": 0.2787, + "step": 12343 + }, + { + "epoch": 0.75, + "learning_rate": 3.2038069508341328e-06, + "loss": 0.2427, + "step": 12344 + }, + { + "epoch": 0.75, + "learning_rate": 3.2023717530883057e-06, + "loss": 0.2533, + "step": 12345 + }, + { + "epoch": 0.75, + "learning_rate": 3.2009368155875033e-06, + "loss": 0.2867, + "step": 12346 + }, + { + "epoch": 0.75, + "learning_rate": 3.1995021383866607e-06, + "loss": 0.2762, + "step": 12347 + }, + { + "epoch": 0.75, + "learning_rate": 3.198067721540702e-06, + "loss": 0.2751, + "step": 12348 + }, + { + "epoch": 0.75, + "learning_rate": 3.196633565104549e-06, + "loss": 0.281, + "step": 12349 + }, + { + "epoch": 0.75, + "learning_rate": 3.1951996691331056e-06, + "loss": 0.278, + "step": 12350 + }, + { + "epoch": 0.75, + "learning_rate": 3.193766033681267e-06, + "loss": 0.2628, + "step": 12351 + }, + { + "epoch": 0.75, + "learning_rate": 3.1923326588039207e-06, + "loss": 0.2766, + "step": 12352 + }, + { + "epoch": 0.75, + "learning_rate": 3.190899544555941e-06, + "loss": 0.3016, + "step": 12353 + }, + { + "epoch": 0.75, + "learning_rate": 3.189466690992197e-06, + "loss": 0.2972, + "step": 12354 + }, + { + "epoch": 0.75, + "learning_rate": 3.188034098167542e-06, + "loss": 0.2768, + "step": 12355 + }, + { + "epoch": 0.75, + "learning_rate": 3.186601766136824e-06, + "loss": 0.2649, + "step": 12356 + }, + { + "epoch": 0.75, + "learning_rate": 3.185169694954877e-06, + "loss": 0.2792, + "step": 12357 + }, + { + "epoch": 0.75, + "learning_rate": 3.183737884676533e-06, + "loss": 0.2792, + "step": 12358 + }, + { + "epoch": 0.75, + "learning_rate": 3.182306335356604e-06, + "loss": 0.2686, + "step": 12359 + }, + { + "epoch": 0.75, + "learning_rate": 3.1808750470498984e-06, + "loss": 0.284, + "step": 12360 + }, + { + "epoch": 0.75, + "learning_rate": 3.179444019811212e-06, + "loss": 0.2446, + "step": 12361 + }, + { + "epoch": 0.75, + "learning_rate": 3.178013253695327e-06, + "loss": 0.2973, + "step": 12362 + }, + { + "epoch": 0.75, + "learning_rate": 3.1765827487570267e-06, + "loss": 0.2564, + "step": 12363 + }, + { + "epoch": 0.75, + "learning_rate": 3.1751525050510745e-06, + "loss": 0.2729, + "step": 12364 + }, + { + "epoch": 0.75, + "learning_rate": 3.173722522632228e-06, + "loss": 0.2839, + "step": 12365 + }, + { + "epoch": 0.75, + "learning_rate": 3.172292801555231e-06, + "loss": 0.2822, + "step": 12366 + }, + { + "epoch": 0.75, + "learning_rate": 3.1708633418748193e-06, + "loss": 0.263, + "step": 12367 + }, + { + "epoch": 0.75, + "learning_rate": 3.1694341436457245e-06, + "loss": 0.2637, + "step": 12368 + }, + { + "epoch": 0.75, + "learning_rate": 3.1680052069226586e-06, + "loss": 0.2891, + "step": 12369 + }, + { + "epoch": 0.75, + "learning_rate": 3.1665765317603302e-06, + "loss": 0.2819, + "step": 12370 + }, + { + "epoch": 0.75, + "learning_rate": 3.1651481182134356e-06, + "loss": 0.2991, + "step": 12371 + }, + { + "epoch": 0.75, + "learning_rate": 3.163719966336656e-06, + "loss": 0.2985, + "step": 12372 + }, + { + "epoch": 0.75, + "learning_rate": 3.1622920761846755e-06, + "loss": 0.267, + "step": 12373 + }, + { + "epoch": 0.75, + "learning_rate": 3.1608644478121576e-06, + "loss": 0.251, + "step": 12374 + }, + { + "epoch": 0.75, + "learning_rate": 3.159437081273757e-06, + "loss": 0.2473, + "step": 12375 + }, + { + "epoch": 0.75, + "learning_rate": 3.1580099766241225e-06, + "loss": 0.2775, + "step": 12376 + }, + { + "epoch": 0.75, + "learning_rate": 3.1565831339178844e-06, + "loss": 0.2665, + "step": 12377 + }, + { + "epoch": 0.75, + "learning_rate": 3.1551565532096764e-06, + "loss": 0.2758, + "step": 12378 + }, + { + "epoch": 0.75, + "learning_rate": 3.153730234554112e-06, + "loss": 0.2613, + "step": 12379 + }, + { + "epoch": 0.75, + "learning_rate": 3.1523041780057962e-06, + "loss": 0.2935, + "step": 12380 + }, + { + "epoch": 0.75, + "learning_rate": 3.150878383619327e-06, + "loss": 0.2517, + "step": 12381 + }, + { + "epoch": 0.75, + "learning_rate": 3.1494528514492886e-06, + "loss": 0.2554, + "step": 12382 + }, + { + "epoch": 0.75, + "learning_rate": 3.148027581550255e-06, + "loss": 0.2931, + "step": 12383 + }, + { + "epoch": 0.75, + "learning_rate": 3.146602573976798e-06, + "loss": 0.2747, + "step": 12384 + }, + { + "epoch": 0.75, + "learning_rate": 3.14517782878347e-06, + "loss": 0.2528, + "step": 12385 + }, + { + "epoch": 0.75, + "learning_rate": 3.1437533460248217e-06, + "loss": 0.286, + "step": 12386 + }, + { + "epoch": 0.75, + "learning_rate": 3.1423291257553777e-06, + "loss": 0.2782, + "step": 12387 + }, + { + "epoch": 0.75, + "learning_rate": 3.140905168029674e-06, + "loss": 0.2723, + "step": 12388 + }, + { + "epoch": 0.75, + "learning_rate": 3.139481472902224e-06, + "loss": 0.2902, + "step": 12389 + }, + { + "epoch": 0.75, + "learning_rate": 3.1380580404275317e-06, + "loss": 0.2757, + "step": 12390 + }, + { + "epoch": 0.75, + "learning_rate": 3.1366348706600947e-06, + "loss": 0.292, + "step": 12391 + }, + { + "epoch": 0.75, + "learning_rate": 3.135211963654393e-06, + "loss": 0.2732, + "step": 12392 + }, + { + "epoch": 0.75, + "learning_rate": 3.1337893194649117e-06, + "loss": 0.2588, + "step": 12393 + }, + { + "epoch": 0.75, + "learning_rate": 3.13236693814611e-06, + "loss": 0.2525, + "step": 12394 + }, + { + "epoch": 0.75, + "learning_rate": 3.1309448197524452e-06, + "loss": 0.2441, + "step": 12395 + }, + { + "epoch": 0.75, + "learning_rate": 3.129522964338362e-06, + "loss": 0.2553, + "step": 12396 + }, + { + "epoch": 0.75, + "learning_rate": 3.128101371958292e-06, + "loss": 0.273, + "step": 12397 + }, + { + "epoch": 0.75, + "learning_rate": 3.1266800426666677e-06, + "loss": 0.2759, + "step": 12398 + }, + { + "epoch": 0.75, + "learning_rate": 3.1252589765179e-06, + "loss": 0.2833, + "step": 12399 + }, + { + "epoch": 0.75, + "learning_rate": 3.123838173566396e-06, + "loss": 0.2772, + "step": 12400 + }, + { + "epoch": 0.75, + "learning_rate": 3.122417633866548e-06, + "loss": 0.2607, + "step": 12401 + }, + { + "epoch": 0.75, + "learning_rate": 3.1209973574727404e-06, + "loss": 0.246, + "step": 12402 + }, + { + "epoch": 0.75, + "learning_rate": 3.119577344439352e-06, + "loss": 0.2704, + "step": 12403 + }, + { + "epoch": 0.75, + "learning_rate": 3.1181575948207467e-06, + "loss": 0.2616, + "step": 12404 + }, + { + "epoch": 0.75, + "learning_rate": 3.1167381086712765e-06, + "loss": 0.2545, + "step": 12405 + }, + { + "epoch": 0.75, + "learning_rate": 3.1153188860452876e-06, + "loss": 0.278, + "step": 12406 + }, + { + "epoch": 0.75, + "learning_rate": 3.1138999269971114e-06, + "loss": 0.2524, + "step": 12407 + }, + { + "epoch": 0.75, + "learning_rate": 3.112481231581078e-06, + "loss": 0.2889, + "step": 12408 + }, + { + "epoch": 0.75, + "learning_rate": 3.1110627998514998e-06, + "loss": 0.2846, + "step": 12409 + }, + { + "epoch": 0.75, + "learning_rate": 3.1096446318626794e-06, + "loss": 0.2532, + "step": 12410 + }, + { + "epoch": 0.75, + "learning_rate": 3.1082267276689115e-06, + "loss": 0.2949, + "step": 12411 + }, + { + "epoch": 0.75, + "learning_rate": 3.1068090873244772e-06, + "loss": 0.2674, + "step": 12412 + }, + { + "epoch": 0.75, + "learning_rate": 3.1053917108836563e-06, + "loss": 0.2875, + "step": 12413 + }, + { + "epoch": 0.75, + "learning_rate": 3.1039745984007095e-06, + "loss": 0.2775, + "step": 12414 + }, + { + "epoch": 0.75, + "learning_rate": 3.102557749929891e-06, + "loss": 0.2716, + "step": 12415 + }, + { + "epoch": 0.75, + "learning_rate": 3.101141165525444e-06, + "loss": 0.2877, + "step": 12416 + }, + { + "epoch": 0.75, + "learning_rate": 3.099724845241603e-06, + "loss": 0.2879, + "step": 12417 + }, + { + "epoch": 0.75, + "learning_rate": 3.0983087891325857e-06, + "loss": 0.2588, + "step": 12418 + }, + { + "epoch": 0.75, + "learning_rate": 3.096892997252614e-06, + "loss": 0.262, + "step": 12419 + }, + { + "epoch": 0.75, + "learning_rate": 3.09547746965589e-06, + "loss": 0.2437, + "step": 12420 + }, + { + "epoch": 0.75, + "learning_rate": 3.094062206396601e-06, + "loss": 0.284, + "step": 12421 + }, + { + "epoch": 0.75, + "learning_rate": 3.0926472075289294e-06, + "loss": 0.2822, + "step": 12422 + }, + { + "epoch": 0.75, + "learning_rate": 3.091232473107054e-06, + "loss": 0.2825, + "step": 12423 + }, + { + "epoch": 0.75, + "learning_rate": 3.0898180031851354e-06, + "loss": 0.2903, + "step": 12424 + }, + { + "epoch": 0.75, + "learning_rate": 3.0884037978173254e-06, + "loss": 0.2939, + "step": 12425 + }, + { + "epoch": 0.75, + "learning_rate": 3.086989857057766e-06, + "loss": 0.2871, + "step": 12426 + }, + { + "epoch": 0.75, + "learning_rate": 3.0855761809605876e-06, + "loss": 0.2896, + "step": 12427 + }, + { + "epoch": 0.75, + "learning_rate": 3.084162769579917e-06, + "loss": 0.2866, + "step": 12428 + }, + { + "epoch": 0.75, + "learning_rate": 3.082749622969865e-06, + "loss": 0.2656, + "step": 12429 + }, + { + "epoch": 0.75, + "learning_rate": 3.081336741184532e-06, + "loss": 0.2987, + "step": 12430 + }, + { + "epoch": 0.75, + "learning_rate": 3.07992412427801e-06, + "loss": 0.2778, + "step": 12431 + }, + { + "epoch": 0.75, + "learning_rate": 3.078511772304377e-06, + "loss": 0.2882, + "step": 12432 + }, + { + "epoch": 0.75, + "learning_rate": 3.0770996853177116e-06, + "loss": 0.2991, + "step": 12433 + }, + { + "epoch": 0.75, + "learning_rate": 3.075687863372071e-06, + "loss": 0.2592, + "step": 12434 + }, + { + "epoch": 0.75, + "learning_rate": 3.0742763065215074e-06, + "loss": 0.2561, + "step": 12435 + }, + { + "epoch": 0.75, + "learning_rate": 3.0728650148200623e-06, + "loss": 0.254, + "step": 12436 + }, + { + "epoch": 0.75, + "learning_rate": 3.071453988321762e-06, + "loss": 0.2758, + "step": 12437 + }, + { + "epoch": 0.75, + "learning_rate": 3.0700432270806333e-06, + "loss": 0.2983, + "step": 12438 + }, + { + "epoch": 0.75, + "learning_rate": 3.0686327311506846e-06, + "loss": 0.2825, + "step": 12439 + }, + { + "epoch": 0.75, + "learning_rate": 3.0672225005859168e-06, + "loss": 0.3034, + "step": 12440 + }, + { + "epoch": 0.75, + "learning_rate": 3.065812535440319e-06, + "loss": 0.293, + "step": 12441 + }, + { + "epoch": 0.75, + "learning_rate": 3.064402835767868e-06, + "loss": 0.2875, + "step": 12442 + }, + { + "epoch": 0.75, + "learning_rate": 3.0629934016225402e-06, + "loss": 0.2803, + "step": 12443 + }, + { + "epoch": 0.75, + "learning_rate": 3.0615842330582923e-06, + "loss": 0.2602, + "step": 12444 + }, + { + "epoch": 0.75, + "learning_rate": 3.0601753301290747e-06, + "loss": 0.2932, + "step": 12445 + }, + { + "epoch": 0.75, + "learning_rate": 3.058766692888825e-06, + "loss": 0.2644, + "step": 12446 + }, + { + "epoch": 0.75, + "learning_rate": 3.0573583213914716e-06, + "loss": 0.2568, + "step": 12447 + }, + { + "epoch": 0.75, + "learning_rate": 3.0559502156909383e-06, + "loss": 0.2635, + "step": 12448 + }, + { + "epoch": 0.75, + "learning_rate": 3.0545423758411298e-06, + "loss": 0.2845, + "step": 12449 + }, + { + "epoch": 0.75, + "learning_rate": 3.0531348018959474e-06, + "loss": 0.2683, + "step": 12450 + }, + { + "epoch": 0.75, + "learning_rate": 3.0517274939092776e-06, + "loss": 0.2815, + "step": 12451 + }, + { + "epoch": 0.75, + "learning_rate": 3.0503204519350003e-06, + "loss": 0.2617, + "step": 12452 + }, + { + "epoch": 0.75, + "learning_rate": 3.048913676026979e-06, + "loss": 0.2546, + "step": 12453 + }, + { + "epoch": 0.75, + "learning_rate": 3.0475071662390822e-06, + "loss": 0.2863, + "step": 12454 + }, + { + "epoch": 0.75, + "learning_rate": 3.046100922625148e-06, + "loss": 0.2824, + "step": 12455 + }, + { + "epoch": 0.75, + "learning_rate": 3.044694945239016e-06, + "loss": 0.2619, + "step": 12456 + }, + { + "epoch": 0.75, + "learning_rate": 3.043289234134512e-06, + "loss": 0.2722, + "step": 12457 + }, + { + "epoch": 0.75, + "learning_rate": 3.0418837893654596e-06, + "loss": 0.2815, + "step": 12458 + }, + { + "epoch": 0.75, + "learning_rate": 3.040478610985661e-06, + "loss": 0.2718, + "step": 12459 + }, + { + "epoch": 0.75, + "learning_rate": 3.0390736990489143e-06, + "loss": 0.2757, + "step": 12460 + }, + { + "epoch": 0.75, + "learning_rate": 3.0376690536090057e-06, + "loss": 0.2782, + "step": 12461 + }, + { + "epoch": 0.75, + "learning_rate": 3.0362646747197087e-06, + "loss": 0.2582, + "step": 12462 + }, + { + "epoch": 0.75, + "learning_rate": 3.034860562434795e-06, + "loss": 0.2843, + "step": 12463 + }, + { + "epoch": 0.75, + "learning_rate": 3.0334567168080186e-06, + "loss": 0.266, + "step": 12464 + }, + { + "epoch": 0.75, + "learning_rate": 3.032053137893124e-06, + "loss": 0.2759, + "step": 12465 + }, + { + "epoch": 0.75, + "learning_rate": 3.0306498257438487e-06, + "loss": 0.2609, + "step": 12466 + }, + { + "epoch": 0.75, + "learning_rate": 3.0292467804139124e-06, + "loss": 0.2747, + "step": 12467 + }, + { + "epoch": 0.75, + "learning_rate": 3.0278440019570387e-06, + "loss": 0.2658, + "step": 12468 + }, + { + "epoch": 0.75, + "learning_rate": 3.026441490426929e-06, + "loss": 0.2747, + "step": 12469 + }, + { + "epoch": 0.75, + "learning_rate": 3.025039245877276e-06, + "loss": 0.2757, + "step": 12470 + }, + { + "epoch": 0.75, + "learning_rate": 3.0236372683617663e-06, + "loss": 0.3082, + "step": 12471 + }, + { + "epoch": 0.75, + "learning_rate": 3.0222355579340713e-06, + "loss": 0.2935, + "step": 12472 + }, + { + "epoch": 0.75, + "learning_rate": 3.02083411464786e-06, + "loss": 0.2679, + "step": 12473 + }, + { + "epoch": 0.75, + "learning_rate": 3.0194329385567832e-06, + "loss": 0.3009, + "step": 12474 + }, + { + "epoch": 0.75, + "learning_rate": 3.0180320297144848e-06, + "loss": 0.2697, + "step": 12475 + }, + { + "epoch": 0.75, + "learning_rate": 3.016631388174598e-06, + "loss": 0.2713, + "step": 12476 + }, + { + "epoch": 0.75, + "learning_rate": 3.0152310139907436e-06, + "loss": 0.2603, + "step": 12477 + }, + { + "epoch": 0.75, + "learning_rate": 3.0138309072165396e-06, + "loss": 0.2589, + "step": 12478 + }, + { + "epoch": 0.75, + "learning_rate": 3.0124310679055857e-06, + "loss": 0.2803, + "step": 12479 + }, + { + "epoch": 0.75, + "learning_rate": 3.011031496111475e-06, + "loss": 0.3022, + "step": 12480 + }, + { + "epoch": 0.75, + "learning_rate": 3.009632191887789e-06, + "loss": 0.2724, + "step": 12481 + }, + { + "epoch": 0.75, + "learning_rate": 3.008233155288097e-06, + "loss": 0.2796, + "step": 12482 + }, + { + "epoch": 0.75, + "learning_rate": 3.0068343863659667e-06, + "loss": 0.2697, + "step": 12483 + }, + { + "epoch": 0.75, + "learning_rate": 3.0054358851749467e-06, + "loss": 0.2988, + "step": 12484 + }, + { + "epoch": 0.75, + "learning_rate": 3.0040376517685764e-06, + "loss": 0.2969, + "step": 12485 + }, + { + "epoch": 0.75, + "learning_rate": 3.0026396862003894e-06, + "loss": 0.2658, + "step": 12486 + }, + { + "epoch": 0.75, + "learning_rate": 3.0012419885239043e-06, + "loss": 0.2744, + "step": 12487 + }, + { + "epoch": 0.75, + "learning_rate": 2.999844558792633e-06, + "loss": 0.2919, + "step": 12488 + }, + { + "epoch": 0.75, + "learning_rate": 2.998447397060075e-06, + "loss": 0.2833, + "step": 12489 + }, + { + "epoch": 0.75, + "learning_rate": 2.997050503379719e-06, + "loss": 0.267, + "step": 12490 + }, + { + "epoch": 0.75, + "learning_rate": 2.995653877805047e-06, + "loss": 0.279, + "step": 12491 + }, + { + "epoch": 0.75, + "learning_rate": 2.994257520389523e-06, + "loss": 0.2811, + "step": 12492 + }, + { + "epoch": 0.75, + "learning_rate": 2.9928614311866146e-06, + "loss": 0.2529, + "step": 12493 + }, + { + "epoch": 0.75, + "learning_rate": 2.991465610249765e-06, + "loss": 0.2962, + "step": 12494 + }, + { + "epoch": 0.75, + "learning_rate": 2.990070057632415e-06, + "loss": 0.2789, + "step": 12495 + }, + { + "epoch": 0.75, + "learning_rate": 2.9886747733879916e-06, + "loss": 0.3256, + "step": 12496 + }, + { + "epoch": 0.75, + "learning_rate": 2.98727975756991e-06, + "loss": 0.2846, + "step": 12497 + }, + { + "epoch": 0.76, + "learning_rate": 2.985885010231584e-06, + "loss": 0.2821, + "step": 12498 + }, + { + "epoch": 0.76, + "learning_rate": 2.9844905314264083e-06, + "loss": 0.2644, + "step": 12499 + }, + { + "epoch": 0.76, + "learning_rate": 2.983096321207769e-06, + "loss": 0.2836, + "step": 12500 + }, + { + "epoch": 0.76, + "learning_rate": 2.9817023796290444e-06, + "loss": 0.2846, + "step": 12501 + }, + { + "epoch": 0.76, + "learning_rate": 2.9803087067435976e-06, + "loss": 0.2879, + "step": 12502 + }, + { + "epoch": 0.76, + "learning_rate": 2.978915302604791e-06, + "loss": 0.286, + "step": 12503 + }, + { + "epoch": 0.76, + "learning_rate": 2.9775221672659683e-06, + "loss": 0.293, + "step": 12504 + }, + { + "epoch": 0.76, + "learning_rate": 2.9761293007804635e-06, + "loss": 0.2719, + "step": 12505 + }, + { + "epoch": 0.76, + "learning_rate": 2.974736703201604e-06, + "loss": 0.2776, + "step": 12506 + }, + { + "epoch": 0.76, + "learning_rate": 2.9733443745826996e-06, + "loss": 0.2732, + "step": 12507 + }, + { + "epoch": 0.76, + "learning_rate": 2.9719523149770636e-06, + "loss": 0.2628, + "step": 12508 + }, + { + "epoch": 0.76, + "learning_rate": 2.9705605244379854e-06, + "loss": 0.2754, + "step": 12509 + }, + { + "epoch": 0.76, + "learning_rate": 2.9691690030187514e-06, + "loss": 0.2817, + "step": 12510 + }, + { + "epoch": 0.76, + "learning_rate": 2.967777750772635e-06, + "loss": 0.2814, + "step": 12511 + }, + { + "epoch": 0.76, + "learning_rate": 2.9663867677528946e-06, + "loss": 0.2944, + "step": 12512 + }, + { + "epoch": 0.76, + "learning_rate": 2.964996054012792e-06, + "loss": 0.2799, + "step": 12513 + }, + { + "epoch": 0.76, + "learning_rate": 2.9636056096055665e-06, + "loss": 0.2593, + "step": 12514 + }, + { + "epoch": 0.76, + "learning_rate": 2.962215434584451e-06, + "loss": 0.2652, + "step": 12515 + }, + { + "epoch": 0.76, + "learning_rate": 2.9608255290026676e-06, + "loss": 0.2832, + "step": 12516 + }, + { + "epoch": 0.76, + "learning_rate": 2.959435892913426e-06, + "loss": 0.3142, + "step": 12517 + }, + { + "epoch": 0.76, + "learning_rate": 2.958046526369933e-06, + "loss": 0.2688, + "step": 12518 + }, + { + "epoch": 0.76, + "learning_rate": 2.956657429425378e-06, + "loss": 0.2872, + "step": 12519 + }, + { + "epoch": 0.76, + "learning_rate": 2.955268602132941e-06, + "loss": 0.2882, + "step": 12520 + }, + { + "epoch": 0.76, + "learning_rate": 2.953880044545795e-06, + "loss": 0.2722, + "step": 12521 + }, + { + "epoch": 0.76, + "learning_rate": 2.9524917567170975e-06, + "loss": 0.2757, + "step": 12522 + }, + { + "epoch": 0.76, + "learning_rate": 2.951103738700001e-06, + "loss": 0.2672, + "step": 12523 + }, + { + "epoch": 0.76, + "learning_rate": 2.9497159905476437e-06, + "loss": 0.2726, + "step": 12524 + }, + { + "epoch": 0.76, + "learning_rate": 2.9483285123131564e-06, + "loss": 0.2723, + "step": 12525 + }, + { + "epoch": 0.76, + "learning_rate": 2.9469413040496573e-06, + "loss": 0.2699, + "step": 12526 + }, + { + "epoch": 0.76, + "learning_rate": 2.9455543658102524e-06, + "loss": 0.2605, + "step": 12527 + }, + { + "epoch": 0.76, + "learning_rate": 2.9441676976480473e-06, + "loss": 0.2593, + "step": 12528 + }, + { + "epoch": 0.76, + "learning_rate": 2.9427812996161263e-06, + "loss": 0.29, + "step": 12529 + }, + { + "epoch": 0.76, + "learning_rate": 2.941395171767566e-06, + "loss": 0.2928, + "step": 12530 + }, + { + "epoch": 0.76, + "learning_rate": 2.9400093141554364e-06, + "loss": 0.2931, + "step": 12531 + }, + { + "epoch": 0.76, + "learning_rate": 2.93862372683279e-06, + "loss": 0.2723, + "step": 12532 + }, + { + "epoch": 0.76, + "learning_rate": 2.9372384098526783e-06, + "loss": 0.2585, + "step": 12533 + }, + { + "epoch": 0.76, + "learning_rate": 2.9358533632681385e-06, + "loss": 0.2698, + "step": 12534 + }, + { + "epoch": 0.76, + "learning_rate": 2.934468587132193e-06, + "loss": 0.2896, + "step": 12535 + }, + { + "epoch": 0.76, + "learning_rate": 2.933084081497859e-06, + "loss": 0.2833, + "step": 12536 + }, + { + "epoch": 0.76, + "learning_rate": 2.931699846418139e-06, + "loss": 0.2787, + "step": 12537 + }, + { + "epoch": 0.76, + "learning_rate": 2.930315881946034e-06, + "loss": 0.3223, + "step": 12538 + }, + { + "epoch": 0.76, + "learning_rate": 2.9289321881345257e-06, + "loss": 0.2578, + "step": 12539 + }, + { + "epoch": 0.76, + "learning_rate": 2.9275487650365873e-06, + "loss": 0.2744, + "step": 12540 + }, + { + "epoch": 0.76, + "learning_rate": 2.926165612705184e-06, + "loss": 0.3145, + "step": 12541 + }, + { + "epoch": 0.76, + "learning_rate": 2.9247827311932663e-06, + "loss": 0.2711, + "step": 12542 + }, + { + "epoch": 0.76, + "learning_rate": 2.923400120553782e-06, + "loss": 0.2977, + "step": 12543 + }, + { + "epoch": 0.76, + "learning_rate": 2.9220177808396623e-06, + "loss": 0.2627, + "step": 12544 + }, + { + "epoch": 0.76, + "learning_rate": 2.920635712103829e-06, + "loss": 0.2885, + "step": 12545 + }, + { + "epoch": 0.76, + "learning_rate": 2.919253914399195e-06, + "loss": 0.2772, + "step": 12546 + }, + { + "epoch": 0.76, + "learning_rate": 2.9178723877786574e-06, + "loss": 0.264, + "step": 12547 + }, + { + "epoch": 0.76, + "learning_rate": 2.916491132295114e-06, + "loss": 0.2567, + "step": 12548 + }, + { + "epoch": 0.76, + "learning_rate": 2.9151101480014444e-06, + "loss": 0.2942, + "step": 12549 + }, + { + "epoch": 0.76, + "learning_rate": 2.9137294349505175e-06, + "loss": 0.2746, + "step": 12550 + }, + { + "epoch": 0.76, + "learning_rate": 2.912348993195193e-06, + "loss": 0.2956, + "step": 12551 + }, + { + "epoch": 0.76, + "learning_rate": 2.910968822788319e-06, + "loss": 0.2896, + "step": 12552 + }, + { + "epoch": 0.76, + "learning_rate": 2.9095889237827412e-06, + "loss": 0.2963, + "step": 12553 + }, + { + "epoch": 0.76, + "learning_rate": 2.9082092962312834e-06, + "loss": 0.2931, + "step": 12554 + }, + { + "epoch": 0.76, + "learning_rate": 2.9068299401867674e-06, + "loss": 0.2794, + "step": 12555 + }, + { + "epoch": 0.76, + "learning_rate": 2.9054508557019977e-06, + "loss": 0.2852, + "step": 12556 + }, + { + "epoch": 0.76, + "learning_rate": 2.9040720428297754e-06, + "loss": 0.2686, + "step": 12557 + }, + { + "epoch": 0.76, + "learning_rate": 2.9026935016228854e-06, + "loss": 0.2849, + "step": 12558 + }, + { + "epoch": 0.76, + "learning_rate": 2.9013152321341063e-06, + "loss": 0.28, + "step": 12559 + }, + { + "epoch": 0.76, + "learning_rate": 2.8999372344162046e-06, + "loss": 0.2646, + "step": 12560 + }, + { + "epoch": 0.76, + "learning_rate": 2.8985595085219355e-06, + "loss": 0.2807, + "step": 12561 + }, + { + "epoch": 0.76, + "learning_rate": 2.897182054504043e-06, + "loss": 0.2787, + "step": 12562 + }, + { + "epoch": 0.76, + "learning_rate": 2.895804872415268e-06, + "loss": 0.2503, + "step": 12563 + }, + { + "epoch": 0.76, + "learning_rate": 2.8944279623083316e-06, + "loss": 0.2705, + "step": 12564 + }, + { + "epoch": 0.76, + "learning_rate": 2.8930513242359494e-06, + "loss": 0.2796, + "step": 12565 + }, + { + "epoch": 0.76, + "learning_rate": 2.8916749582508253e-06, + "loss": 0.2983, + "step": 12566 + }, + { + "epoch": 0.76, + "learning_rate": 2.8902988644056506e-06, + "loss": 0.2619, + "step": 12567 + }, + { + "epoch": 0.76, + "learning_rate": 2.8889230427531134e-06, + "loss": 0.2669, + "step": 12568 + }, + { + "epoch": 0.76, + "learning_rate": 2.8875474933458847e-06, + "loss": 0.2698, + "step": 12569 + }, + { + "epoch": 0.76, + "learning_rate": 2.886172216236626e-06, + "loss": 0.2765, + "step": 12570 + }, + { + "epoch": 0.76, + "learning_rate": 2.88479721147799e-06, + "loss": 0.2684, + "step": 12571 + }, + { + "epoch": 0.76, + "learning_rate": 2.883422479122615e-06, + "loss": 0.3016, + "step": 12572 + }, + { + "epoch": 0.76, + "learning_rate": 2.8820480192231393e-06, + "loss": 0.266, + "step": 12573 + }, + { + "epoch": 0.76, + "learning_rate": 2.8806738318321792e-06, + "loss": 0.2415, + "step": 12574 + }, + { + "epoch": 0.76, + "learning_rate": 2.879299917002345e-06, + "loss": 0.2736, + "step": 12575 + }, + { + "epoch": 0.76, + "learning_rate": 2.877926274786238e-06, + "loss": 0.2621, + "step": 12576 + }, + { + "epoch": 0.76, + "learning_rate": 2.8765529052364428e-06, + "loss": 0.2725, + "step": 12577 + }, + { + "epoch": 0.76, + "learning_rate": 2.875179808405546e-06, + "loss": 0.2704, + "step": 12578 + }, + { + "epoch": 0.76, + "learning_rate": 2.8738069843461123e-06, + "loss": 0.3285, + "step": 12579 + }, + { + "epoch": 0.76, + "learning_rate": 2.8724344331107e-06, + "loss": 0.2658, + "step": 12580 + }, + { + "epoch": 0.76, + "learning_rate": 2.871062154751858e-06, + "loss": 0.2739, + "step": 12581 + }, + { + "epoch": 0.76, + "learning_rate": 2.8696901493221173e-06, + "loss": 0.2706, + "step": 12582 + }, + { + "epoch": 0.76, + "learning_rate": 2.868318416874015e-06, + "loss": 0.2665, + "step": 12583 + }, + { + "epoch": 0.76, + "learning_rate": 2.8669469574600604e-06, + "loss": 0.2699, + "step": 12584 + }, + { + "epoch": 0.76, + "learning_rate": 2.8655757711327627e-06, + "loss": 0.2666, + "step": 12585 + }, + { + "epoch": 0.76, + "learning_rate": 2.864204857944616e-06, + "loss": 0.2729, + "step": 12586 + }, + { + "epoch": 0.76, + "learning_rate": 2.8628342179481008e-06, + "loss": 0.2648, + "step": 12587 + }, + { + "epoch": 0.76, + "learning_rate": 2.8614638511956992e-06, + "loss": 0.3244, + "step": 12588 + }, + { + "epoch": 0.76, + "learning_rate": 2.8600937577398723e-06, + "loss": 0.2621, + "step": 12589 + }, + { + "epoch": 0.76, + "learning_rate": 2.8587239376330743e-06, + "loss": 0.2605, + "step": 12590 + }, + { + "epoch": 0.76, + "learning_rate": 2.8573543909277466e-06, + "loss": 0.262, + "step": 12591 + }, + { + "epoch": 0.76, + "learning_rate": 2.855985117676322e-06, + "loss": 0.2528, + "step": 12592 + }, + { + "epoch": 0.76, + "learning_rate": 2.8546161179312247e-06, + "loss": 0.272, + "step": 12593 + }, + { + "epoch": 0.76, + "learning_rate": 2.8532473917448644e-06, + "loss": 0.2741, + "step": 12594 + }, + { + "epoch": 0.76, + "learning_rate": 2.8518789391696434e-06, + "loss": 0.2687, + "step": 12595 + }, + { + "epoch": 0.76, + "learning_rate": 2.850510760257951e-06, + "loss": 0.2799, + "step": 12596 + }, + { + "epoch": 0.76, + "learning_rate": 2.849142855062167e-06, + "loss": 0.2902, + "step": 12597 + }, + { + "epoch": 0.76, + "learning_rate": 2.8477752236346645e-06, + "loss": 0.2804, + "step": 12598 + }, + { + "epoch": 0.76, + "learning_rate": 2.8464078660278017e-06, + "loss": 0.245, + "step": 12599 + }, + { + "epoch": 0.76, + "learning_rate": 2.8450407822939274e-06, + "loss": 0.2825, + "step": 12600 + }, + { + "epoch": 0.76, + "learning_rate": 2.843673972485379e-06, + "loss": 0.2722, + "step": 12601 + }, + { + "epoch": 0.76, + "learning_rate": 2.8423074366544813e-06, + "loss": 0.2529, + "step": 12602 + }, + { + "epoch": 0.76, + "learning_rate": 2.8409411748535588e-06, + "loss": 0.2702, + "step": 12603 + }, + { + "epoch": 0.76, + "learning_rate": 2.839575187134915e-06, + "loss": 0.2691, + "step": 12604 + }, + { + "epoch": 0.76, + "learning_rate": 2.838209473550846e-06, + "loss": 0.2905, + "step": 12605 + }, + { + "epoch": 0.76, + "learning_rate": 2.8368440341536373e-06, + "loss": 0.2818, + "step": 12606 + }, + { + "epoch": 0.76, + "learning_rate": 2.8354788689955625e-06, + "loss": 0.2596, + "step": 12607 + }, + { + "epoch": 0.76, + "learning_rate": 2.834113978128892e-06, + "loss": 0.2793, + "step": 12608 + }, + { + "epoch": 0.76, + "learning_rate": 2.8327493616058765e-06, + "loss": 0.2727, + "step": 12609 + }, + { + "epoch": 0.76, + "learning_rate": 2.8313850194787605e-06, + "loss": 0.26, + "step": 12610 + }, + { + "epoch": 0.76, + "learning_rate": 2.830020951799777e-06, + "loss": 0.2652, + "step": 12611 + }, + { + "epoch": 0.76, + "learning_rate": 2.8286571586211466e-06, + "loss": 0.2928, + "step": 12612 + }, + { + "epoch": 0.76, + "learning_rate": 2.827293639995088e-06, + "loss": 0.243, + "step": 12613 + }, + { + "epoch": 0.76, + "learning_rate": 2.8259303959737984e-06, + "loss": 0.2568, + "step": 12614 + }, + { + "epoch": 0.76, + "learning_rate": 2.82456742660947e-06, + "loss": 0.2973, + "step": 12615 + }, + { + "epoch": 0.76, + "learning_rate": 2.8232047319542854e-06, + "loss": 0.2924, + "step": 12616 + }, + { + "epoch": 0.76, + "learning_rate": 2.8218423120604087e-06, + "loss": 0.2718, + "step": 12617 + }, + { + "epoch": 0.76, + "learning_rate": 2.8204801669800084e-06, + "loss": 0.2789, + "step": 12618 + }, + { + "epoch": 0.76, + "learning_rate": 2.8191182967652296e-06, + "loss": 0.2739, + "step": 12619 + }, + { + "epoch": 0.76, + "learning_rate": 2.8177567014682105e-06, + "loss": 0.2556, + "step": 12620 + }, + { + "epoch": 0.76, + "learning_rate": 2.8163953811410805e-06, + "loss": 0.2769, + "step": 12621 + }, + { + "epoch": 0.76, + "learning_rate": 2.8150343358359545e-06, + "loss": 0.2729, + "step": 12622 + }, + { + "epoch": 0.76, + "learning_rate": 2.813673565604944e-06, + "loss": 0.2796, + "step": 12623 + }, + { + "epoch": 0.76, + "learning_rate": 2.8123130705001445e-06, + "loss": 0.2857, + "step": 12624 + }, + { + "epoch": 0.76, + "learning_rate": 2.810952850573644e-06, + "loss": 0.2637, + "step": 12625 + }, + { + "epoch": 0.76, + "learning_rate": 2.8095929058775094e-06, + "loss": 0.2893, + "step": 12626 + }, + { + "epoch": 0.76, + "learning_rate": 2.8082332364638153e-06, + "loss": 0.2885, + "step": 12627 + }, + { + "epoch": 0.76, + "learning_rate": 2.806873842384612e-06, + "loss": 0.2754, + "step": 12628 + }, + { + "epoch": 0.76, + "learning_rate": 2.8055147236919444e-06, + "loss": 0.2787, + "step": 12629 + }, + { + "epoch": 0.76, + "learning_rate": 2.8041558804378455e-06, + "loss": 0.2704, + "step": 12630 + }, + { + "epoch": 0.76, + "learning_rate": 2.802797312674338e-06, + "loss": 0.2683, + "step": 12631 + }, + { + "epoch": 0.76, + "learning_rate": 2.8014390204534325e-06, + "loss": 0.2649, + "step": 12632 + }, + { + "epoch": 0.76, + "learning_rate": 2.800081003827135e-06, + "loss": 0.2793, + "step": 12633 + }, + { + "epoch": 0.76, + "learning_rate": 2.7987232628474347e-06, + "loss": 0.2846, + "step": 12634 + }, + { + "epoch": 0.76, + "learning_rate": 2.7973657975663127e-06, + "loss": 0.3002, + "step": 12635 + }, + { + "epoch": 0.76, + "learning_rate": 2.7960086080357383e-06, + "loss": 0.2727, + "step": 12636 + }, + { + "epoch": 0.76, + "learning_rate": 2.7946516943076686e-06, + "loss": 0.2831, + "step": 12637 + }, + { + "epoch": 0.76, + "learning_rate": 2.7932950564340577e-06, + "loss": 0.2835, + "step": 12638 + }, + { + "epoch": 0.76, + "learning_rate": 2.7919386944668425e-06, + "loss": 0.2694, + "step": 12639 + }, + { + "epoch": 0.76, + "learning_rate": 2.79058260845795e-06, + "loss": 0.2593, + "step": 12640 + }, + { + "epoch": 0.76, + "learning_rate": 2.789226798459298e-06, + "loss": 0.2677, + "step": 12641 + }, + { + "epoch": 0.76, + "learning_rate": 2.787871264522789e-06, + "loss": 0.2701, + "step": 12642 + }, + { + "epoch": 0.76, + "learning_rate": 2.786516006700326e-06, + "loss": 0.2864, + "step": 12643 + }, + { + "epoch": 0.76, + "learning_rate": 2.785161025043792e-06, + "loss": 0.2785, + "step": 12644 + }, + { + "epoch": 0.76, + "learning_rate": 2.7838063196050624e-06, + "loss": 0.2627, + "step": 12645 + }, + { + "epoch": 0.76, + "learning_rate": 2.782451890436e-06, + "loss": 0.2655, + "step": 12646 + }, + { + "epoch": 0.76, + "learning_rate": 2.7810977375884563e-06, + "loss": 0.2779, + "step": 12647 + }, + { + "epoch": 0.76, + "learning_rate": 2.7797438611142825e-06, + "loss": 0.2667, + "step": 12648 + }, + { + "epoch": 0.76, + "learning_rate": 2.778390261065306e-06, + "loss": 0.2658, + "step": 12649 + }, + { + "epoch": 0.76, + "learning_rate": 2.777036937493349e-06, + "loss": 0.2856, + "step": 12650 + }, + { + "epoch": 0.76, + "learning_rate": 2.775683890450225e-06, + "loss": 0.3055, + "step": 12651 + }, + { + "epoch": 0.76, + "learning_rate": 2.77433111998773e-06, + "loss": 0.2882, + "step": 12652 + }, + { + "epoch": 0.76, + "learning_rate": 2.7729786261576618e-06, + "loss": 0.3016, + "step": 12653 + }, + { + "epoch": 0.76, + "learning_rate": 2.771626409011796e-06, + "loss": 0.2697, + "step": 12654 + }, + { + "epoch": 0.76, + "learning_rate": 2.770274468601902e-06, + "loss": 0.2595, + "step": 12655 + }, + { + "epoch": 0.76, + "learning_rate": 2.76892280497974e-06, + "loss": 0.2838, + "step": 12656 + }, + { + "epoch": 0.76, + "learning_rate": 2.7675714181970527e-06, + "loss": 0.2722, + "step": 12657 + }, + { + "epoch": 0.76, + "learning_rate": 2.7662203083055848e-06, + "loss": 0.3047, + "step": 12658 + }, + { + "epoch": 0.76, + "learning_rate": 2.7648694753570637e-06, + "loss": 0.2627, + "step": 12659 + }, + { + "epoch": 0.76, + "learning_rate": 2.7635189194031974e-06, + "loss": 0.2598, + "step": 12660 + }, + { + "epoch": 0.76, + "learning_rate": 2.7621686404956936e-06, + "loss": 0.296, + "step": 12661 + }, + { + "epoch": 0.76, + "learning_rate": 2.7608186386862533e-06, + "loss": 0.2706, + "step": 12662 + }, + { + "epoch": 0.76, + "learning_rate": 2.7594689140265563e-06, + "loss": 0.2963, + "step": 12663 + }, + { + "epoch": 0.77, + "learning_rate": 2.7581194665682774e-06, + "loss": 0.2638, + "step": 12664 + }, + { + "epoch": 0.77, + "learning_rate": 2.7567702963630805e-06, + "loss": 0.2799, + "step": 12665 + }, + { + "epoch": 0.77, + "learning_rate": 2.755421403462616e-06, + "loss": 0.3188, + "step": 12666 + }, + { + "epoch": 0.77, + "learning_rate": 2.7540727879185247e-06, + "loss": 0.2518, + "step": 12667 + }, + { + "epoch": 0.77, + "learning_rate": 2.7527244497824433e-06, + "loss": 0.2754, + "step": 12668 + }, + { + "epoch": 0.77, + "learning_rate": 2.75137638910599e-06, + "loss": 0.2748, + "step": 12669 + }, + { + "epoch": 0.77, + "learning_rate": 2.750028605940773e-06, + "loss": 0.2826, + "step": 12670 + }, + { + "epoch": 0.77, + "learning_rate": 2.7486811003383953e-06, + "loss": 0.2948, + "step": 12671 + }, + { + "epoch": 0.77, + "learning_rate": 2.7473338723504383e-06, + "loss": 0.2969, + "step": 12672 + }, + { + "epoch": 0.77, + "learning_rate": 2.7459869220284898e-06, + "loss": 0.2672, + "step": 12673 + }, + { + "epoch": 0.77, + "learning_rate": 2.744640249424113e-06, + "loss": 0.2741, + "step": 12674 + }, + { + "epoch": 0.77, + "learning_rate": 2.7432938545888645e-06, + "loss": 0.2638, + "step": 12675 + }, + { + "epoch": 0.77, + "learning_rate": 2.741947737574291e-06, + "loss": 0.2902, + "step": 12676 + }, + { + "epoch": 0.77, + "learning_rate": 2.740601898431925e-06, + "loss": 0.2792, + "step": 12677 + }, + { + "epoch": 0.77, + "learning_rate": 2.7392563372132974e-06, + "loss": 0.2658, + "step": 12678 + }, + { + "epoch": 0.77, + "learning_rate": 2.7379110539699203e-06, + "loss": 0.2842, + "step": 12679 + }, + { + "epoch": 0.77, + "learning_rate": 2.7365660487532976e-06, + "loss": 0.2783, + "step": 12680 + }, + { + "epoch": 0.77, + "learning_rate": 2.7352213216149203e-06, + "loss": 0.268, + "step": 12681 + }, + { + "epoch": 0.77, + "learning_rate": 2.7338768726062703e-06, + "loss": 0.2789, + "step": 12682 + }, + { + "epoch": 0.77, + "learning_rate": 2.732532701778824e-06, + "loss": 0.2729, + "step": 12683 + }, + { + "epoch": 0.77, + "learning_rate": 2.73118880918404e-06, + "loss": 0.3041, + "step": 12684 + }, + { + "epoch": 0.77, + "learning_rate": 2.7298451948733684e-06, + "loss": 0.2818, + "step": 12685 + }, + { + "epoch": 0.77, + "learning_rate": 2.7285018588982505e-06, + "loss": 0.2605, + "step": 12686 + }, + { + "epoch": 0.77, + "learning_rate": 2.72715880131011e-06, + "loss": 0.2772, + "step": 12687 + }, + { + "epoch": 0.77, + "learning_rate": 2.7258160221603735e-06, + "loss": 0.2653, + "step": 12688 + }, + { + "epoch": 0.77, + "learning_rate": 2.7244735215004448e-06, + "loss": 0.2651, + "step": 12689 + }, + { + "epoch": 0.77, + "learning_rate": 2.7231312993817217e-06, + "loss": 0.2806, + "step": 12690 + }, + { + "epoch": 0.77, + "learning_rate": 2.721789355855591e-06, + "loss": 0.286, + "step": 12691 + }, + { + "epoch": 0.77, + "learning_rate": 2.7204476909734245e-06, + "loss": 0.2557, + "step": 12692 + }, + { + "epoch": 0.77, + "learning_rate": 2.7191063047865975e-06, + "loss": 0.2868, + "step": 12693 + }, + { + "epoch": 0.77, + "learning_rate": 2.7177651973464546e-06, + "loss": 0.2499, + "step": 12694 + }, + { + "epoch": 0.77, + "learning_rate": 2.7164243687043433e-06, + "loss": 0.2924, + "step": 12695 + }, + { + "epoch": 0.77, + "learning_rate": 2.7150838189115934e-06, + "loss": 0.2729, + "step": 12696 + }, + { + "epoch": 0.77, + "learning_rate": 2.713743548019534e-06, + "loss": 0.2437, + "step": 12697 + }, + { + "epoch": 0.77, + "learning_rate": 2.7124035560794736e-06, + "loss": 0.2931, + "step": 12698 + }, + { + "epoch": 0.77, + "learning_rate": 2.7110638431427128e-06, + "loss": 0.2506, + "step": 12699 + }, + { + "epoch": 0.77, + "learning_rate": 2.7097244092605435e-06, + "loss": 0.2497, + "step": 12700 + }, + { + "epoch": 0.77, + "learning_rate": 2.708385254484244e-06, + "loss": 0.2934, + "step": 12701 + }, + { + "epoch": 0.77, + "learning_rate": 2.7070463788650815e-06, + "loss": 0.2719, + "step": 12702 + }, + { + "epoch": 0.77, + "learning_rate": 2.7057077824543187e-06, + "loss": 0.2775, + "step": 12703 + }, + { + "epoch": 0.77, + "learning_rate": 2.7043694653032026e-06, + "loss": 0.2808, + "step": 12704 + }, + { + "epoch": 0.77, + "learning_rate": 2.70303142746297e-06, + "loss": 0.2773, + "step": 12705 + }, + { + "epoch": 0.77, + "learning_rate": 2.701693668984845e-06, + "loss": 0.2857, + "step": 12706 + }, + { + "epoch": 0.77, + "learning_rate": 2.7003561899200428e-06, + "loss": 0.2768, + "step": 12707 + }, + { + "epoch": 0.77, + "learning_rate": 2.6990189903197727e-06, + "loss": 0.2699, + "step": 12708 + }, + { + "epoch": 0.77, + "learning_rate": 2.697682070235227e-06, + "loss": 0.263, + "step": 12709 + }, + { + "epoch": 0.77, + "learning_rate": 2.6963454297175897e-06, + "loss": 0.2985, + "step": 12710 + }, + { + "epoch": 0.77, + "learning_rate": 2.695009068818031e-06, + "loss": 0.2708, + "step": 12711 + }, + { + "epoch": 0.77, + "learning_rate": 2.693672987587712e-06, + "loss": 0.262, + "step": 12712 + }, + { + "epoch": 0.77, + "learning_rate": 2.6923371860777915e-06, + "loss": 0.2932, + "step": 12713 + }, + { + "epoch": 0.77, + "learning_rate": 2.6910016643394042e-06, + "loss": 0.2525, + "step": 12714 + }, + { + "epoch": 0.77, + "learning_rate": 2.6896664224236823e-06, + "loss": 0.2944, + "step": 12715 + }, + { + "epoch": 0.77, + "learning_rate": 2.6883314603817446e-06, + "loss": 0.2649, + "step": 12716 + }, + { + "epoch": 0.77, + "learning_rate": 2.6869967782646968e-06, + "loss": 0.2624, + "step": 12717 + }, + { + "epoch": 0.77, + "learning_rate": 2.6856623761236413e-06, + "loss": 0.2623, + "step": 12718 + }, + { + "epoch": 0.77, + "learning_rate": 2.6843282540096647e-06, + "loss": 0.2688, + "step": 12719 + }, + { + "epoch": 0.77, + "learning_rate": 2.6829944119738415e-06, + "loss": 0.2563, + "step": 12720 + }, + { + "epoch": 0.77, + "learning_rate": 2.681660850067238e-06, + "loss": 0.2817, + "step": 12721 + }, + { + "epoch": 0.77, + "learning_rate": 2.680327568340907e-06, + "loss": 0.2768, + "step": 12722 + }, + { + "epoch": 0.77, + "learning_rate": 2.6789945668458984e-06, + "loss": 0.2903, + "step": 12723 + }, + { + "epoch": 0.77, + "learning_rate": 2.6776618456332416e-06, + "loss": 0.2698, + "step": 12724 + }, + { + "epoch": 0.77, + "learning_rate": 2.67632940475396e-06, + "loss": 0.2759, + "step": 12725 + }, + { + "epoch": 0.77, + "learning_rate": 2.6749972442590667e-06, + "loss": 0.2779, + "step": 12726 + }, + { + "epoch": 0.77, + "learning_rate": 2.6736653641995614e-06, + "loss": 0.2866, + "step": 12727 + }, + { + "epoch": 0.77, + "learning_rate": 2.672333764626436e-06, + "loss": 0.2898, + "step": 12728 + }, + { + "epoch": 0.77, + "learning_rate": 2.67100244559067e-06, + "loss": 0.2775, + "step": 12729 + }, + { + "epoch": 0.77, + "learning_rate": 2.6696714071432326e-06, + "loss": 0.2634, + "step": 12730 + }, + { + "epoch": 0.77, + "learning_rate": 2.6683406493350785e-06, + "loss": 0.2638, + "step": 12731 + }, + { + "epoch": 0.77, + "learning_rate": 2.667010172217163e-06, + "loss": 0.3079, + "step": 12732 + }, + { + "epoch": 0.77, + "learning_rate": 2.6656799758404172e-06, + "loss": 0.2602, + "step": 12733 + }, + { + "epoch": 0.77, + "learning_rate": 2.664350060255769e-06, + "loss": 0.2687, + "step": 12734 + }, + { + "epoch": 0.77, + "learning_rate": 2.6630204255141346e-06, + "loss": 0.2475, + "step": 12735 + }, + { + "epoch": 0.77, + "learning_rate": 2.661691071666415e-06, + "loss": 0.2819, + "step": 12736 + }, + { + "epoch": 0.77, + "learning_rate": 2.6603619987635087e-06, + "loss": 0.2684, + "step": 12737 + }, + { + "epoch": 0.77, + "learning_rate": 2.659033206856297e-06, + "loss": 0.2747, + "step": 12738 + }, + { + "epoch": 0.77, + "learning_rate": 2.657704695995653e-06, + "loss": 0.2557, + "step": 12739 + }, + { + "epoch": 0.77, + "learning_rate": 2.6563764662324378e-06, + "loss": 0.2821, + "step": 12740 + }, + { + "epoch": 0.77, + "learning_rate": 2.655048517617501e-06, + "loss": 0.2854, + "step": 12741 + }, + { + "epoch": 0.77, + "learning_rate": 2.653720850201681e-06, + "loss": 0.2884, + "step": 12742 + }, + { + "epoch": 0.77, + "learning_rate": 2.6523934640358128e-06, + "loss": 0.291, + "step": 12743 + }, + { + "epoch": 0.77, + "learning_rate": 2.651066359170712e-06, + "loss": 0.2543, + "step": 12744 + }, + { + "epoch": 0.77, + "learning_rate": 2.6497395356571864e-06, + "loss": 0.2748, + "step": 12745 + }, + { + "epoch": 0.77, + "learning_rate": 2.6484129935460325e-06, + "loss": 0.2758, + "step": 12746 + }, + { + "epoch": 0.77, + "learning_rate": 2.6470867328880346e-06, + "loss": 0.279, + "step": 12747 + }, + { + "epoch": 0.77, + "learning_rate": 2.6457607537339735e-06, + "loss": 0.2974, + "step": 12748 + }, + { + "epoch": 0.77, + "learning_rate": 2.6444350561346098e-06, + "loss": 0.2906, + "step": 12749 + }, + { + "epoch": 0.77, + "learning_rate": 2.6431096401406998e-06, + "loss": 0.2849, + "step": 12750 + }, + { + "epoch": 0.77, + "learning_rate": 2.641784505802986e-06, + "loss": 0.2662, + "step": 12751 + }, + { + "epoch": 0.77, + "learning_rate": 2.640459653172195e-06, + "loss": 0.2744, + "step": 12752 + }, + { + "epoch": 0.77, + "learning_rate": 2.639135082299058e-06, + "loss": 0.2645, + "step": 12753 + }, + { + "epoch": 0.77, + "learning_rate": 2.6378107932342813e-06, + "loss": 0.2602, + "step": 12754 + }, + { + "epoch": 0.77, + "learning_rate": 2.636486786028564e-06, + "loss": 0.2831, + "step": 12755 + }, + { + "epoch": 0.77, + "learning_rate": 2.6351630607325963e-06, + "loss": 0.2998, + "step": 12756 + }, + { + "epoch": 0.77, + "learning_rate": 2.6338396173970536e-06, + "loss": 0.2734, + "step": 12757 + }, + { + "epoch": 0.77, + "learning_rate": 2.632516456072609e-06, + "loss": 0.2683, + "step": 12758 + }, + { + "epoch": 0.77, + "learning_rate": 2.631193576809917e-06, + "loss": 0.2884, + "step": 12759 + }, + { + "epoch": 0.77, + "learning_rate": 2.6298709796596234e-06, + "loss": 0.2792, + "step": 12760 + }, + { + "epoch": 0.77, + "learning_rate": 2.6285486646723636e-06, + "loss": 0.2793, + "step": 12761 + }, + { + "epoch": 0.77, + "learning_rate": 2.6272266318987606e-06, + "loss": 0.2676, + "step": 12762 + }, + { + "epoch": 0.77, + "learning_rate": 2.6259048813894293e-06, + "loss": 0.275, + "step": 12763 + }, + { + "epoch": 0.77, + "learning_rate": 2.624583413194973e-06, + "loss": 0.2824, + "step": 12764 + }, + { + "epoch": 0.77, + "learning_rate": 2.6232622273659824e-06, + "loss": 0.2857, + "step": 12765 + }, + { + "epoch": 0.77, + "learning_rate": 2.6219413239530365e-06, + "loss": 0.256, + "step": 12766 + }, + { + "epoch": 0.77, + "learning_rate": 2.620620703006711e-06, + "loss": 0.2402, + "step": 12767 + }, + { + "epoch": 0.77, + "learning_rate": 2.6193003645775627e-06, + "loss": 0.2663, + "step": 12768 + }, + { + "epoch": 0.77, + "learning_rate": 2.6179803087161417e-06, + "loss": 0.2697, + "step": 12769 + }, + { + "epoch": 0.77, + "learning_rate": 2.6166605354729836e-06, + "loss": 0.2698, + "step": 12770 + }, + { + "epoch": 0.77, + "learning_rate": 2.615341044898614e-06, + "loss": 0.2492, + "step": 12771 + }, + { + "epoch": 0.77, + "learning_rate": 2.6140218370435542e-06, + "loss": 0.2571, + "step": 12772 + }, + { + "epoch": 0.77, + "learning_rate": 2.612702911958308e-06, + "loss": 0.2713, + "step": 12773 + }, + { + "epoch": 0.77, + "learning_rate": 2.611384269693369e-06, + "loss": 0.2967, + "step": 12774 + }, + { + "epoch": 0.77, + "learning_rate": 2.6100659102992208e-06, + "loss": 0.3097, + "step": 12775 + }, + { + "epoch": 0.77, + "learning_rate": 2.6087478338263373e-06, + "loss": 0.2768, + "step": 12776 + }, + { + "epoch": 0.77, + "learning_rate": 2.6074300403251763e-06, + "loss": 0.2757, + "step": 12777 + }, + { + "epoch": 0.77, + "learning_rate": 2.6061125298461964e-06, + "loss": 0.2536, + "step": 12778 + }, + { + "epoch": 0.77, + "learning_rate": 2.6047953024398356e-06, + "loss": 0.2969, + "step": 12779 + }, + { + "epoch": 0.77, + "learning_rate": 2.6034783581565213e-06, + "loss": 0.2744, + "step": 12780 + }, + { + "epoch": 0.77, + "learning_rate": 2.6021616970466733e-06, + "loss": 0.268, + "step": 12781 + }, + { + "epoch": 0.77, + "learning_rate": 2.6008453191606973e-06, + "loss": 0.2719, + "step": 12782 + }, + { + "epoch": 0.77, + "learning_rate": 2.5995292245489968e-06, + "loss": 0.2944, + "step": 12783 + }, + { + "epoch": 0.77, + "learning_rate": 2.598213413261953e-06, + "loss": 0.2535, + "step": 12784 + }, + { + "epoch": 0.77, + "learning_rate": 2.5968978853499427e-06, + "loss": 0.2969, + "step": 12785 + }, + { + "epoch": 0.77, + "learning_rate": 2.595582640863331e-06, + "loss": 0.2684, + "step": 12786 + }, + { + "epoch": 0.77, + "learning_rate": 2.5942676798524667e-06, + "loss": 0.2852, + "step": 12787 + }, + { + "epoch": 0.77, + "learning_rate": 2.5929530023677008e-06, + "loss": 0.2729, + "step": 12788 + }, + { + "epoch": 0.77, + "learning_rate": 2.59163860845936e-06, + "loss": 0.2676, + "step": 12789 + }, + { + "epoch": 0.77, + "learning_rate": 2.5903244981777674e-06, + "loss": 0.2492, + "step": 12790 + }, + { + "epoch": 0.77, + "learning_rate": 2.5890106715732333e-06, + "loss": 0.2676, + "step": 12791 + }, + { + "epoch": 0.77, + "learning_rate": 2.5876971286960527e-06, + "loss": 0.3008, + "step": 12792 + }, + { + "epoch": 0.77, + "learning_rate": 2.5863838695965214e-06, + "loss": 0.2907, + "step": 12793 + }, + { + "epoch": 0.77, + "learning_rate": 2.5850708943249135e-06, + "loss": 0.273, + "step": 12794 + }, + { + "epoch": 0.77, + "learning_rate": 2.583758202931499e-06, + "loss": 0.2902, + "step": 12795 + }, + { + "epoch": 0.77, + "learning_rate": 2.582445795466525e-06, + "loss": 0.2859, + "step": 12796 + }, + { + "epoch": 0.77, + "learning_rate": 2.581133671980246e-06, + "loss": 0.2662, + "step": 12797 + }, + { + "epoch": 0.77, + "learning_rate": 2.579821832522892e-06, + "loss": 0.2737, + "step": 12798 + }, + { + "epoch": 0.77, + "learning_rate": 2.5785102771446866e-06, + "loss": 0.2695, + "step": 12799 + }, + { + "epoch": 0.77, + "learning_rate": 2.5771990058958442e-06, + "loss": 0.2916, + "step": 12800 + }, + { + "epoch": 0.77, + "learning_rate": 2.57588801882656e-06, + "loss": 0.2891, + "step": 12801 + }, + { + "epoch": 0.77, + "learning_rate": 2.5745773159870334e-06, + "loss": 0.2666, + "step": 12802 + }, + { + "epoch": 0.77, + "learning_rate": 2.573266897427441e-06, + "loss": 0.2719, + "step": 12803 + }, + { + "epoch": 0.77, + "learning_rate": 2.5719567631979512e-06, + "loss": 0.2853, + "step": 12804 + }, + { + "epoch": 0.77, + "learning_rate": 2.570646913348722e-06, + "loss": 0.2674, + "step": 12805 + }, + { + "epoch": 0.77, + "learning_rate": 2.569337347929898e-06, + "loss": 0.2762, + "step": 12806 + }, + { + "epoch": 0.77, + "learning_rate": 2.5680280669916204e-06, + "loss": 0.2786, + "step": 12807 + }, + { + "epoch": 0.77, + "learning_rate": 2.566719070584013e-06, + "loss": 0.3097, + "step": 12808 + }, + { + "epoch": 0.77, + "learning_rate": 2.5654103587571887e-06, + "loss": 0.2832, + "step": 12809 + }, + { + "epoch": 0.77, + "learning_rate": 2.564101931561254e-06, + "loss": 0.2666, + "step": 12810 + }, + { + "epoch": 0.77, + "learning_rate": 2.5627937890462985e-06, + "loss": 0.2667, + "step": 12811 + }, + { + "epoch": 0.77, + "learning_rate": 2.5614859312624018e-06, + "loss": 0.2843, + "step": 12812 + }, + { + "epoch": 0.77, + "learning_rate": 2.560178358259642e-06, + "loss": 0.2867, + "step": 12813 + }, + { + "epoch": 0.77, + "learning_rate": 2.558871070088076e-06, + "loss": 0.2485, + "step": 12814 + }, + { + "epoch": 0.77, + "learning_rate": 2.5575640667977506e-06, + "loss": 0.2436, + "step": 12815 + }, + { + "epoch": 0.77, + "learning_rate": 2.5562573484387066e-06, + "loss": 0.249, + "step": 12816 + }, + { + "epoch": 0.77, + "learning_rate": 2.5549509150609673e-06, + "loss": 0.2768, + "step": 12817 + }, + { + "epoch": 0.77, + "learning_rate": 2.5536447667145557e-06, + "loss": 0.2606, + "step": 12818 + }, + { + "epoch": 0.77, + "learning_rate": 2.5523389034494737e-06, + "loss": 0.2641, + "step": 12819 + }, + { + "epoch": 0.77, + "learning_rate": 2.551033325315716e-06, + "loss": 0.2846, + "step": 12820 + }, + { + "epoch": 0.77, + "learning_rate": 2.5497280323632657e-06, + "loss": 0.2821, + "step": 12821 + }, + { + "epoch": 0.77, + "learning_rate": 2.5484230246420926e-06, + "loss": 0.2877, + "step": 12822 + }, + { + "epoch": 0.77, + "learning_rate": 2.547118302202165e-06, + "loss": 0.2766, + "step": 12823 + }, + { + "epoch": 0.77, + "learning_rate": 2.5458138650934316e-06, + "loss": 0.2775, + "step": 12824 + }, + { + "epoch": 0.77, + "learning_rate": 2.5445097133658305e-06, + "loss": 0.2736, + "step": 12825 + }, + { + "epoch": 0.77, + "learning_rate": 2.543205847069292e-06, + "loss": 0.2529, + "step": 12826 + }, + { + "epoch": 0.77, + "learning_rate": 2.5419022662537306e-06, + "loss": 0.2527, + "step": 12827 + }, + { + "epoch": 0.77, + "learning_rate": 2.540598970969059e-06, + "loss": 0.2732, + "step": 12828 + }, + { + "epoch": 0.78, + "learning_rate": 2.5392959612651756e-06, + "loss": 0.2775, + "step": 12829 + }, + { + "epoch": 0.78, + "learning_rate": 2.537993237191958e-06, + "loss": 0.2673, + "step": 12830 + }, + { + "epoch": 0.78, + "learning_rate": 2.5366907987992807e-06, + "loss": 0.2732, + "step": 12831 + }, + { + "epoch": 0.78, + "learning_rate": 2.5353886461370127e-06, + "loss": 0.2815, + "step": 12832 + }, + { + "epoch": 0.78, + "learning_rate": 2.5340867792550052e-06, + "loss": 0.2701, + "step": 12833 + }, + { + "epoch": 0.78, + "learning_rate": 2.532785198203098e-06, + "loss": 0.2935, + "step": 12834 + }, + { + "epoch": 0.78, + "learning_rate": 2.531483903031122e-06, + "loss": 0.2451, + "step": 12835 + }, + { + "epoch": 0.78, + "learning_rate": 2.530182893788895e-06, + "loss": 0.2669, + "step": 12836 + }, + { + "epoch": 0.78, + "learning_rate": 2.52888217052623e-06, + "loss": 0.2386, + "step": 12837 + }, + { + "epoch": 0.78, + "learning_rate": 2.527581733292923e-06, + "loss": 0.2497, + "step": 12838 + }, + { + "epoch": 0.78, + "learning_rate": 2.526281582138761e-06, + "loss": 0.2846, + "step": 12839 + }, + { + "epoch": 0.78, + "learning_rate": 2.5249817171135194e-06, + "loss": 0.2771, + "step": 12840 + }, + { + "epoch": 0.78, + "learning_rate": 2.523682138266961e-06, + "loss": 0.2764, + "step": 12841 + }, + { + "epoch": 0.78, + "learning_rate": 2.522382845648843e-06, + "loss": 0.2872, + "step": 12842 + }, + { + "epoch": 0.78, + "learning_rate": 2.521083839308909e-06, + "loss": 0.2737, + "step": 12843 + }, + { + "epoch": 0.78, + "learning_rate": 2.519785119296888e-06, + "loss": 0.2633, + "step": 12844 + }, + { + "epoch": 0.78, + "learning_rate": 2.518486685662502e-06, + "loss": 0.2785, + "step": 12845 + }, + { + "epoch": 0.78, + "learning_rate": 2.5171885384554627e-06, + "loss": 0.2506, + "step": 12846 + }, + { + "epoch": 0.78, + "learning_rate": 2.5158906777254654e-06, + "loss": 0.2787, + "step": 12847 + }, + { + "epoch": 0.78, + "learning_rate": 2.514593103522204e-06, + "loss": 0.2762, + "step": 12848 + }, + { + "epoch": 0.78, + "learning_rate": 2.5132958158953523e-06, + "loss": 0.2644, + "step": 12849 + }, + { + "epoch": 0.78, + "learning_rate": 2.511998814894576e-06, + "loss": 0.2946, + "step": 12850 + }, + { + "epoch": 0.78, + "learning_rate": 2.510702100569532e-06, + "loss": 0.2825, + "step": 12851 + }, + { + "epoch": 0.78, + "learning_rate": 2.509405672969861e-06, + "loss": 0.2705, + "step": 12852 + }, + { + "epoch": 0.78, + "learning_rate": 2.508109532145203e-06, + "loss": 0.2527, + "step": 12853 + }, + { + "epoch": 0.78, + "learning_rate": 2.5068136781451757e-06, + "loss": 0.2561, + "step": 12854 + }, + { + "epoch": 0.78, + "learning_rate": 2.5055181110193917e-06, + "loss": 0.2676, + "step": 12855 + }, + { + "epoch": 0.78, + "learning_rate": 2.50422283081745e-06, + "loss": 0.3004, + "step": 12856 + }, + { + "epoch": 0.78, + "learning_rate": 2.502927837588939e-06, + "loss": 0.2898, + "step": 12857 + }, + { + "epoch": 0.78, + "learning_rate": 2.501633131383442e-06, + "loss": 0.2585, + "step": 12858 + }, + { + "epoch": 0.78, + "learning_rate": 2.500338712250523e-06, + "loss": 0.2722, + "step": 12859 + }, + { + "epoch": 0.78, + "learning_rate": 2.4990445802397388e-06, + "loss": 0.2739, + "step": 12860 + }, + { + "epoch": 0.78, + "learning_rate": 2.497750735400635e-06, + "loss": 0.2905, + "step": 12861 + }, + { + "epoch": 0.78, + "learning_rate": 2.496457177782743e-06, + "loss": 0.2907, + "step": 12862 + }, + { + "epoch": 0.78, + "learning_rate": 2.495163907435596e-06, + "loss": 0.3248, + "step": 12863 + }, + { + "epoch": 0.78, + "learning_rate": 2.4938709244086955e-06, + "loss": 0.2754, + "step": 12864 + }, + { + "epoch": 0.78, + "learning_rate": 2.4925782287515476e-06, + "loss": 0.279, + "step": 12865 + }, + { + "epoch": 0.78, + "learning_rate": 2.491285820513639e-06, + "loss": 0.2697, + "step": 12866 + }, + { + "epoch": 0.78, + "learning_rate": 2.4899936997444564e-06, + "loss": 0.2839, + "step": 12867 + }, + { + "epoch": 0.78, + "learning_rate": 2.488701866493464e-06, + "loss": 0.2527, + "step": 12868 + }, + { + "epoch": 0.78, + "learning_rate": 2.4874103208101187e-06, + "loss": 0.2585, + "step": 12869 + }, + { + "epoch": 0.78, + "learning_rate": 2.486119062743867e-06, + "loss": 0.2531, + "step": 12870 + }, + { + "epoch": 0.78, + "learning_rate": 2.4848280923441425e-06, + "loss": 0.2785, + "step": 12871 + }, + { + "epoch": 0.78, + "learning_rate": 2.4835374096603758e-06, + "loss": 0.2712, + "step": 12872 + }, + { + "epoch": 0.78, + "learning_rate": 2.4822470147419765e-06, + "loss": 0.2712, + "step": 12873 + }, + { + "epoch": 0.78, + "learning_rate": 2.4809569076383453e-06, + "loss": 0.2843, + "step": 12874 + }, + { + "epoch": 0.78, + "learning_rate": 2.4796670883988773e-06, + "loss": 0.269, + "step": 12875 + }, + { + "epoch": 0.78, + "learning_rate": 2.478377557072946e-06, + "loss": 0.2681, + "step": 12876 + }, + { + "epoch": 0.78, + "learning_rate": 2.477088313709929e-06, + "loss": 0.2861, + "step": 12877 + }, + { + "epoch": 0.78, + "learning_rate": 2.475799358359181e-06, + "loss": 0.2667, + "step": 12878 + }, + { + "epoch": 0.78, + "learning_rate": 2.47451069107005e-06, + "loss": 0.2783, + "step": 12879 + }, + { + "epoch": 0.78, + "learning_rate": 2.4732223118918707e-06, + "loss": 0.2543, + "step": 12880 + }, + { + "epoch": 0.78, + "learning_rate": 2.4719342208739695e-06, + "loss": 0.2676, + "step": 12881 + }, + { + "epoch": 0.78, + "learning_rate": 2.470646418065656e-06, + "loss": 0.252, + "step": 12882 + }, + { + "epoch": 0.78, + "learning_rate": 2.4693589035162413e-06, + "loss": 0.2779, + "step": 12883 + }, + { + "epoch": 0.78, + "learning_rate": 2.4680716772750136e-06, + "loss": 0.2628, + "step": 12884 + }, + { + "epoch": 0.78, + "learning_rate": 2.4667847393912535e-06, + "loss": 0.2527, + "step": 12885 + }, + { + "epoch": 0.78, + "learning_rate": 2.465498089914231e-06, + "loss": 0.2764, + "step": 12886 + }, + { + "epoch": 0.78, + "learning_rate": 2.4642117288932033e-06, + "loss": 0.2757, + "step": 12887 + }, + { + "epoch": 0.78, + "learning_rate": 2.462925656377424e-06, + "loss": 0.2732, + "step": 12888 + }, + { + "epoch": 0.78, + "learning_rate": 2.461639872416125e-06, + "loss": 0.291, + "step": 12889 + }, + { + "epoch": 0.78, + "learning_rate": 2.460354377058535e-06, + "loss": 0.2711, + "step": 12890 + }, + { + "epoch": 0.78, + "learning_rate": 2.459069170353867e-06, + "loss": 0.2695, + "step": 12891 + }, + { + "epoch": 0.78, + "learning_rate": 2.4577842523513218e-06, + "loss": 0.2849, + "step": 12892 + }, + { + "epoch": 0.78, + "learning_rate": 2.4564996231000982e-06, + "loss": 0.2723, + "step": 12893 + }, + { + "epoch": 0.78, + "learning_rate": 2.4552152826493757e-06, + "loss": 0.2921, + "step": 12894 + }, + { + "epoch": 0.78, + "learning_rate": 2.453931231048323e-06, + "loss": 0.2405, + "step": 12895 + }, + { + "epoch": 0.78, + "learning_rate": 2.452647468346102e-06, + "loss": 0.2884, + "step": 12896 + }, + { + "epoch": 0.78, + "learning_rate": 2.45136399459186e-06, + "loss": 0.2814, + "step": 12897 + }, + { + "epoch": 0.78, + "learning_rate": 2.450080809834734e-06, + "loss": 0.273, + "step": 12898 + }, + { + "epoch": 0.78, + "learning_rate": 2.4487979141238506e-06, + "loss": 0.3013, + "step": 12899 + }, + { + "epoch": 0.78, + "learning_rate": 2.447515307508325e-06, + "loss": 0.2835, + "step": 12900 + }, + { + "epoch": 0.78, + "learning_rate": 2.4462329900372584e-06, + "loss": 0.2549, + "step": 12901 + }, + { + "epoch": 0.78, + "learning_rate": 2.44495096175975e-06, + "loss": 0.2713, + "step": 12902 + }, + { + "epoch": 0.78, + "learning_rate": 2.443669222724878e-06, + "loss": 0.2952, + "step": 12903 + }, + { + "epoch": 0.78, + "learning_rate": 2.4423877729817156e-06, + "loss": 0.2585, + "step": 12904 + }, + { + "epoch": 0.78, + "learning_rate": 2.44110661257932e-06, + "loss": 0.2483, + "step": 12905 + }, + { + "epoch": 0.78, + "learning_rate": 2.439825741566739e-06, + "loss": 0.2821, + "step": 12906 + }, + { + "epoch": 0.78, + "learning_rate": 2.4385451599930166e-06, + "loss": 0.2866, + "step": 12907 + }, + { + "epoch": 0.78, + "learning_rate": 2.437264867907174e-06, + "loss": 0.2566, + "step": 12908 + }, + { + "epoch": 0.78, + "learning_rate": 2.4359848653582285e-06, + "loss": 0.2778, + "step": 12909 + }, + { + "epoch": 0.78, + "learning_rate": 2.4347051523951847e-06, + "loss": 0.2458, + "step": 12910 + }, + { + "epoch": 0.78, + "learning_rate": 2.4334257290670337e-06, + "loss": 0.2753, + "step": 12911 + }, + { + "epoch": 0.78, + "learning_rate": 2.4321465954227606e-06, + "loss": 0.2762, + "step": 12912 + }, + { + "epoch": 0.78, + "learning_rate": 2.4308677515113376e-06, + "loss": 0.2952, + "step": 12913 + }, + { + "epoch": 0.78, + "learning_rate": 2.4295891973817223e-06, + "loss": 0.2507, + "step": 12914 + }, + { + "epoch": 0.78, + "learning_rate": 2.4283109330828647e-06, + "loss": 0.2861, + "step": 12915 + }, + { + "epoch": 0.78, + "learning_rate": 2.427032958663702e-06, + "loss": 0.2853, + "step": 12916 + }, + { + "epoch": 0.78, + "learning_rate": 2.4257552741731593e-06, + "loss": 0.26, + "step": 12917 + }, + { + "epoch": 0.78, + "learning_rate": 2.4244778796601565e-06, + "loss": 0.2787, + "step": 12918 + }, + { + "epoch": 0.78, + "learning_rate": 2.4232007751735976e-06, + "loss": 0.2485, + "step": 12919 + }, + { + "epoch": 0.78, + "learning_rate": 2.4219239607623747e-06, + "loss": 0.2704, + "step": 12920 + }, + { + "epoch": 0.78, + "learning_rate": 2.4206474364753696e-06, + "loss": 0.272, + "step": 12921 + }, + { + "epoch": 0.78, + "learning_rate": 2.4193712023614514e-06, + "loss": 0.298, + "step": 12922 + }, + { + "epoch": 0.78, + "learning_rate": 2.4180952584694872e-06, + "loss": 0.2596, + "step": 12923 + }, + { + "epoch": 0.78, + "learning_rate": 2.4168196048483216e-06, + "loss": 0.2697, + "step": 12924 + }, + { + "epoch": 0.78, + "learning_rate": 2.4155442415467935e-06, + "loss": 0.2457, + "step": 12925 + }, + { + "epoch": 0.78, + "learning_rate": 2.4142691686137286e-06, + "loss": 0.2924, + "step": 12926 + }, + { + "epoch": 0.78, + "learning_rate": 2.412994386097942e-06, + "loss": 0.2659, + "step": 12927 + }, + { + "epoch": 0.78, + "learning_rate": 2.4117198940482424e-06, + "loss": 0.2395, + "step": 12928 + }, + { + "epoch": 0.78, + "learning_rate": 2.41044569251342e-06, + "loss": 0.2673, + "step": 12929 + }, + { + "epoch": 0.78, + "learning_rate": 2.409171781542259e-06, + "loss": 0.2699, + "step": 12930 + }, + { + "epoch": 0.78, + "learning_rate": 2.4078981611835294e-06, + "loss": 0.2605, + "step": 12931 + }, + { + "epoch": 0.78, + "learning_rate": 2.406624831485992e-06, + "loss": 0.2794, + "step": 12932 + }, + { + "epoch": 0.78, + "learning_rate": 2.405351792498395e-06, + "loss": 0.266, + "step": 12933 + }, + { + "epoch": 0.78, + "learning_rate": 2.404079044269476e-06, + "loss": 0.2783, + "step": 12934 + }, + { + "epoch": 0.78, + "learning_rate": 2.4028065868479643e-06, + "loss": 0.2891, + "step": 12935 + }, + { + "epoch": 0.78, + "learning_rate": 2.4015344202825695e-06, + "loss": 0.2348, + "step": 12936 + }, + { + "epoch": 0.78, + "learning_rate": 2.400262544622004e-06, + "loss": 0.2969, + "step": 12937 + }, + { + "epoch": 0.78, + "learning_rate": 2.3989909599149575e-06, + "loss": 0.2771, + "step": 12938 + }, + { + "epoch": 0.78, + "learning_rate": 2.397719666210111e-06, + "loss": 0.2773, + "step": 12939 + }, + { + "epoch": 0.78, + "learning_rate": 2.3964486635561378e-06, + "loss": 0.2578, + "step": 12940 + }, + { + "epoch": 0.78, + "learning_rate": 2.3951779520016937e-06, + "loss": 0.2334, + "step": 12941 + }, + { + "epoch": 0.78, + "learning_rate": 2.393907531595434e-06, + "loss": 0.2514, + "step": 12942 + }, + { + "epoch": 0.78, + "learning_rate": 2.392637402385992e-06, + "loss": 0.2787, + "step": 12943 + }, + { + "epoch": 0.78, + "learning_rate": 2.391367564421996e-06, + "loss": 0.2621, + "step": 12944 + }, + { + "epoch": 0.78, + "learning_rate": 2.3900980177520603e-06, + "loss": 0.2885, + "step": 12945 + }, + { + "epoch": 0.78, + "learning_rate": 2.3888287624247864e-06, + "loss": 0.2969, + "step": 12946 + }, + { + "epoch": 0.78, + "learning_rate": 2.387559798488772e-06, + "loss": 0.2556, + "step": 12947 + }, + { + "epoch": 0.78, + "learning_rate": 2.386291125992598e-06, + "loss": 0.2815, + "step": 12948 + }, + { + "epoch": 0.78, + "learning_rate": 2.385022744984834e-06, + "loss": 0.2967, + "step": 12949 + }, + { + "epoch": 0.78, + "learning_rate": 2.38375465551404e-06, + "loss": 0.2874, + "step": 12950 + }, + { + "epoch": 0.78, + "learning_rate": 2.3824868576287654e-06, + "loss": 0.2836, + "step": 12951 + }, + { + "epoch": 0.78, + "learning_rate": 2.381219351377542e-06, + "loss": 0.2854, + "step": 12952 + }, + { + "epoch": 0.78, + "learning_rate": 2.379952136808903e-06, + "loss": 0.2704, + "step": 12953 + }, + { + "epoch": 0.78, + "learning_rate": 2.378685213971361e-06, + "loss": 0.2644, + "step": 12954 + }, + { + "epoch": 0.78, + "learning_rate": 2.377418582913419e-06, + "loss": 0.2821, + "step": 12955 + }, + { + "epoch": 0.78, + "learning_rate": 2.3761522436835694e-06, + "loss": 0.2637, + "step": 12956 + }, + { + "epoch": 0.78, + "learning_rate": 2.3748861963302914e-06, + "loss": 0.2626, + "step": 12957 + }, + { + "epoch": 0.78, + "learning_rate": 2.3736204409020613e-06, + "loss": 0.2853, + "step": 12958 + }, + { + "epoch": 0.78, + "learning_rate": 2.372354977447334e-06, + "loss": 0.2875, + "step": 12959 + }, + { + "epoch": 0.78, + "learning_rate": 2.371089806014558e-06, + "loss": 0.2609, + "step": 12960 + }, + { + "epoch": 0.78, + "learning_rate": 2.3698249266521702e-06, + "loss": 0.2567, + "step": 12961 + }, + { + "epoch": 0.78, + "learning_rate": 2.368560339408592e-06, + "loss": 0.2891, + "step": 12962 + }, + { + "epoch": 0.78, + "learning_rate": 2.3672960443322455e-06, + "loss": 0.2778, + "step": 12963 + }, + { + "epoch": 0.78, + "learning_rate": 2.36603204147153e-06, + "loss": 0.2803, + "step": 12964 + }, + { + "epoch": 0.78, + "learning_rate": 2.3647683308748392e-06, + "loss": 0.2902, + "step": 12965 + }, + { + "epoch": 0.78, + "learning_rate": 2.3635049125905475e-06, + "loss": 0.2665, + "step": 12966 + }, + { + "epoch": 0.78, + "learning_rate": 2.362241786667031e-06, + "loss": 0.2863, + "step": 12967 + }, + { + "epoch": 0.78, + "learning_rate": 2.360978953152646e-06, + "loss": 0.2746, + "step": 12968 + }, + { + "epoch": 0.78, + "learning_rate": 2.3597164120957404e-06, + "loss": 0.2757, + "step": 12969 + }, + { + "epoch": 0.78, + "learning_rate": 2.3584541635446488e-06, + "loss": 0.3048, + "step": 12970 + }, + { + "epoch": 0.78, + "learning_rate": 2.3571922075476947e-06, + "loss": 0.2592, + "step": 12971 + }, + { + "epoch": 0.78, + "learning_rate": 2.3559305441531952e-06, + "loss": 0.2924, + "step": 12972 + }, + { + "epoch": 0.78, + "learning_rate": 2.3546691734094527e-06, + "loss": 0.2938, + "step": 12973 + }, + { + "epoch": 0.78, + "learning_rate": 2.3534080953647566e-06, + "loss": 0.2672, + "step": 12974 + }, + { + "epoch": 0.78, + "learning_rate": 2.3521473100673863e-06, + "loss": 0.2677, + "step": 12975 + }, + { + "epoch": 0.78, + "learning_rate": 2.3508868175656085e-06, + "loss": 0.3087, + "step": 12976 + }, + { + "epoch": 0.78, + "learning_rate": 2.3496266179076867e-06, + "loss": 0.2473, + "step": 12977 + }, + { + "epoch": 0.78, + "learning_rate": 2.348366711141865e-06, + "loss": 0.2831, + "step": 12978 + }, + { + "epoch": 0.78, + "learning_rate": 2.3471070973163767e-06, + "loss": 0.2649, + "step": 12979 + }, + { + "epoch": 0.78, + "learning_rate": 2.345847776479446e-06, + "loss": 0.2812, + "step": 12980 + }, + { + "epoch": 0.78, + "learning_rate": 2.344588748679285e-06, + "loss": 0.2815, + "step": 12981 + }, + { + "epoch": 0.78, + "learning_rate": 2.3433300139640993e-06, + "loss": 0.2723, + "step": 12982 + }, + { + "epoch": 0.78, + "learning_rate": 2.342071572382075e-06, + "loss": 0.2783, + "step": 12983 + }, + { + "epoch": 0.78, + "learning_rate": 2.340813423981393e-06, + "loss": 0.2708, + "step": 12984 + }, + { + "epoch": 0.78, + "learning_rate": 2.339555568810221e-06, + "loss": 0.2782, + "step": 12985 + }, + { + "epoch": 0.78, + "learning_rate": 2.3382980069167115e-06, + "loss": 0.272, + "step": 12986 + }, + { + "epoch": 0.78, + "learning_rate": 2.337040738349018e-06, + "loss": 0.2593, + "step": 12987 + }, + { + "epoch": 0.78, + "learning_rate": 2.335783763155268e-06, + "loss": 0.2398, + "step": 12988 + }, + { + "epoch": 0.78, + "learning_rate": 2.3345270813835883e-06, + "loss": 0.2672, + "step": 12989 + }, + { + "epoch": 0.78, + "learning_rate": 2.333270693082088e-06, + "loss": 0.2679, + "step": 12990 + }, + { + "epoch": 0.78, + "learning_rate": 2.3320145982988685e-06, + "loss": 0.263, + "step": 12991 + }, + { + "epoch": 0.78, + "learning_rate": 2.330758797082016e-06, + "loss": 0.2591, + "step": 12992 + }, + { + "epoch": 0.78, + "learning_rate": 2.329503289479613e-06, + "loss": 0.2994, + "step": 12993 + }, + { + "epoch": 0.78, + "learning_rate": 2.328248075539725e-06, + "loss": 0.2697, + "step": 12994 + }, + { + "epoch": 0.79, + "learning_rate": 2.3269931553104077e-06, + "loss": 0.2874, + "step": 12995 + }, + { + "epoch": 0.79, + "learning_rate": 2.325738528839704e-06, + "loss": 0.257, + "step": 12996 + }, + { + "epoch": 0.79, + "learning_rate": 2.3244841961756448e-06, + "loss": 0.2571, + "step": 12997 + }, + { + "epoch": 0.79, + "learning_rate": 2.3232301573662575e-06, + "loss": 0.2588, + "step": 12998 + }, + { + "epoch": 0.79, + "learning_rate": 2.321976412459552e-06, + "loss": 0.2641, + "step": 12999 + }, + { + "epoch": 0.79, + "learning_rate": 2.320722961503522e-06, + "loss": 0.2782, + "step": 13000 + }, + { + "epoch": 0.79, + "learning_rate": 2.319469804546156e-06, + "loss": 0.2893, + "step": 13001 + }, + { + "epoch": 0.79, + "learning_rate": 2.318216941635436e-06, + "loss": 0.2726, + "step": 13002 + }, + { + "epoch": 0.79, + "learning_rate": 2.3169643728193247e-06, + "loss": 0.2958, + "step": 13003 + }, + { + "epoch": 0.79, + "learning_rate": 2.315712098145776e-06, + "loss": 0.3019, + "step": 13004 + }, + { + "epoch": 0.79, + "learning_rate": 2.3144601176627335e-06, + "loss": 0.2468, + "step": 13005 + }, + { + "epoch": 0.79, + "learning_rate": 2.3132084314181246e-06, + "loss": 0.2603, + "step": 13006 + }, + { + "epoch": 0.79, + "learning_rate": 2.3119570394598777e-06, + "loss": 0.2612, + "step": 13007 + }, + { + "epoch": 0.79, + "learning_rate": 2.3107059418358966e-06, + "loss": 0.2716, + "step": 13008 + }, + { + "epoch": 0.79, + "learning_rate": 2.309455138594081e-06, + "loss": 0.2913, + "step": 13009 + }, + { + "epoch": 0.79, + "learning_rate": 2.308204629782318e-06, + "loss": 0.2589, + "step": 13010 + }, + { + "epoch": 0.79, + "learning_rate": 2.306954415448478e-06, + "loss": 0.2874, + "step": 13011 + }, + { + "epoch": 0.79, + "learning_rate": 2.3057044956404316e-06, + "loss": 0.2697, + "step": 13012 + }, + { + "epoch": 0.79, + "learning_rate": 2.3044548704060287e-06, + "loss": 0.2776, + "step": 13013 + }, + { + "epoch": 0.79, + "learning_rate": 2.303205539793111e-06, + "loss": 0.2833, + "step": 13014 + }, + { + "epoch": 0.79, + "learning_rate": 2.3019565038495097e-06, + "loss": 0.2792, + "step": 13015 + }, + { + "epoch": 0.79, + "learning_rate": 2.300707762623039e-06, + "loss": 0.2414, + "step": 13016 + }, + { + "epoch": 0.79, + "learning_rate": 2.299459316161513e-06, + "loss": 0.2674, + "step": 13017 + }, + { + "epoch": 0.79, + "learning_rate": 2.2982111645127257e-06, + "loss": 0.3011, + "step": 13018 + }, + { + "epoch": 0.79, + "learning_rate": 2.296963307724461e-06, + "loss": 0.267, + "step": 13019 + }, + { + "epoch": 0.79, + "learning_rate": 2.2957157458444945e-06, + "loss": 0.2838, + "step": 13020 + }, + { + "epoch": 0.79, + "learning_rate": 2.294468478920585e-06, + "loss": 0.2574, + "step": 13021 + }, + { + "epoch": 0.79, + "learning_rate": 2.2932215070004893e-06, + "loss": 0.2888, + "step": 13022 + }, + { + "epoch": 0.79, + "learning_rate": 2.2919748301319442e-06, + "loss": 0.2871, + "step": 13023 + }, + { + "epoch": 0.79, + "learning_rate": 2.290728448362679e-06, + "loss": 0.2839, + "step": 13024 + }, + { + "epoch": 0.79, + "learning_rate": 2.2894823617404107e-06, + "loss": 0.2539, + "step": 13025 + }, + { + "epoch": 0.79, + "learning_rate": 2.288236570312844e-06, + "loss": 0.2586, + "step": 13026 + }, + { + "epoch": 0.79, + "learning_rate": 2.2869910741276736e-06, + "loss": 0.2799, + "step": 13027 + }, + { + "epoch": 0.79, + "learning_rate": 2.2857458732325866e-06, + "loss": 0.2667, + "step": 13028 + }, + { + "epoch": 0.79, + "learning_rate": 2.2845009676752527e-06, + "loss": 0.2697, + "step": 13029 + }, + { + "epoch": 0.79, + "learning_rate": 2.2832563575033317e-06, + "loss": 0.2898, + "step": 13030 + }, + { + "epoch": 0.79, + "learning_rate": 2.2820120427644754e-06, + "loss": 0.2681, + "step": 13031 + }, + { + "epoch": 0.79, + "learning_rate": 2.2807680235063177e-06, + "loss": 0.2669, + "step": 13032 + }, + { + "epoch": 0.79, + "learning_rate": 2.2795242997764943e-06, + "loss": 0.2628, + "step": 13033 + }, + { + "epoch": 0.79, + "learning_rate": 2.2782808716226124e-06, + "loss": 0.2737, + "step": 13034 + }, + { + "epoch": 0.79, + "learning_rate": 2.277037739092278e-06, + "loss": 0.2609, + "step": 13035 + }, + { + "epoch": 0.79, + "learning_rate": 2.275794902233083e-06, + "loss": 0.272, + "step": 13036 + }, + { + "epoch": 0.79, + "learning_rate": 2.2745523610926126e-06, + "loss": 0.2573, + "step": 13037 + }, + { + "epoch": 0.79, + "learning_rate": 2.273310115718436e-06, + "loss": 0.2578, + "step": 13038 + }, + { + "epoch": 0.79, + "learning_rate": 2.272068166158111e-06, + "loss": 0.2549, + "step": 13039 + }, + { + "epoch": 0.79, + "learning_rate": 2.2708265124591867e-06, + "loss": 0.2571, + "step": 13040 + }, + { + "epoch": 0.79, + "learning_rate": 2.269585154669195e-06, + "loss": 0.2616, + "step": 13041 + }, + { + "epoch": 0.79, + "learning_rate": 2.268344092835668e-06, + "loss": 0.2549, + "step": 13042 + }, + { + "epoch": 0.79, + "learning_rate": 2.267103327006115e-06, + "loss": 0.2896, + "step": 13043 + }, + { + "epoch": 0.79, + "learning_rate": 2.2658628572280395e-06, + "loss": 0.2706, + "step": 13044 + }, + { + "epoch": 0.79, + "learning_rate": 2.2646226835489317e-06, + "loss": 0.2843, + "step": 13045 + }, + { + "epoch": 0.79, + "learning_rate": 2.263382806016269e-06, + "loss": 0.2652, + "step": 13046 + }, + { + "epoch": 0.79, + "learning_rate": 2.2621432246775256e-06, + "loss": 0.2736, + "step": 13047 + }, + { + "epoch": 0.79, + "learning_rate": 2.2609039395801547e-06, + "loss": 0.2815, + "step": 13048 + }, + { + "epoch": 0.79, + "learning_rate": 2.2596649507716015e-06, + "loss": 0.2768, + "step": 13049 + }, + { + "epoch": 0.79, + "learning_rate": 2.258426258299302e-06, + "loss": 0.2542, + "step": 13050 + }, + { + "epoch": 0.79, + "learning_rate": 2.2571878622106757e-06, + "loss": 0.2789, + "step": 13051 + }, + { + "epoch": 0.79, + "learning_rate": 2.2559497625531402e-06, + "loss": 0.2582, + "step": 13052 + }, + { + "epoch": 0.79, + "learning_rate": 2.2547119593740917e-06, + "loss": 0.2563, + "step": 13053 + }, + { + "epoch": 0.79, + "learning_rate": 2.2534744527209197e-06, + "loss": 0.2529, + "step": 13054 + }, + { + "epoch": 0.79, + "learning_rate": 2.2522372426410024e-06, + "loss": 0.2758, + "step": 13055 + }, + { + "epoch": 0.79, + "learning_rate": 2.251000329181703e-06, + "loss": 0.2524, + "step": 13056 + }, + { + "epoch": 0.79, + "learning_rate": 2.2497637123903802e-06, + "loss": 0.2522, + "step": 13057 + }, + { + "epoch": 0.79, + "learning_rate": 2.2485273923143767e-06, + "loss": 0.2716, + "step": 13058 + }, + { + "epoch": 0.79, + "learning_rate": 2.2472913690010245e-06, + "loss": 0.261, + "step": 13059 + }, + { + "epoch": 0.79, + "learning_rate": 2.2460556424976444e-06, + "loss": 0.2641, + "step": 13060 + }, + { + "epoch": 0.79, + "learning_rate": 2.244820212851544e-06, + "loss": 0.2586, + "step": 13061 + }, + { + "epoch": 0.79, + "learning_rate": 2.24358508011002e-06, + "loss": 0.2467, + "step": 13062 + }, + { + "epoch": 0.79, + "learning_rate": 2.242350244320365e-06, + "loss": 0.268, + "step": 13063 + }, + { + "epoch": 0.79, + "learning_rate": 2.2411157055298493e-06, + "loss": 0.2542, + "step": 13064 + }, + { + "epoch": 0.79, + "learning_rate": 2.23988146378574e-06, + "loss": 0.2762, + "step": 13065 + }, + { + "epoch": 0.79, + "learning_rate": 2.2386475191352874e-06, + "loss": 0.2921, + "step": 13066 + }, + { + "epoch": 0.79, + "learning_rate": 2.237413871625734e-06, + "loss": 0.2501, + "step": 13067 + }, + { + "epoch": 0.79, + "learning_rate": 2.2361805213043087e-06, + "loss": 0.2956, + "step": 13068 + }, + { + "epoch": 0.79, + "learning_rate": 2.2349474682182302e-06, + "loss": 0.2765, + "step": 13069 + }, + { + "epoch": 0.79, + "learning_rate": 2.233714712414705e-06, + "loss": 0.2653, + "step": 13070 + }, + { + "epoch": 0.79, + "learning_rate": 2.2324822539409275e-06, + "loss": 0.2768, + "step": 13071 + }, + { + "epoch": 0.79, + "learning_rate": 2.231250092844086e-06, + "loss": 0.2676, + "step": 13072 + }, + { + "epoch": 0.79, + "learning_rate": 2.230018229171351e-06, + "loss": 0.297, + "step": 13073 + }, + { + "epoch": 0.79, + "learning_rate": 2.228786662969885e-06, + "loss": 0.2347, + "step": 13074 + }, + { + "epoch": 0.79, + "learning_rate": 2.2275553942868365e-06, + "loss": 0.2819, + "step": 13075 + }, + { + "epoch": 0.79, + "learning_rate": 2.2263244231693426e-06, + "loss": 0.2866, + "step": 13076 + }, + { + "epoch": 0.79, + "learning_rate": 2.2250937496645365e-06, + "loss": 0.2817, + "step": 13077 + }, + { + "epoch": 0.79, + "learning_rate": 2.2238633738195294e-06, + "loss": 0.2825, + "step": 13078 + }, + { + "epoch": 0.79, + "learning_rate": 2.2226332956814277e-06, + "loss": 0.2697, + "step": 13079 + }, + { + "epoch": 0.79, + "learning_rate": 2.2214035152973236e-06, + "loss": 0.2715, + "step": 13080 + }, + { + "epoch": 0.79, + "learning_rate": 2.220174032714297e-06, + "loss": 0.274, + "step": 13081 + }, + { + "epoch": 0.79, + "learning_rate": 2.2189448479794227e-06, + "loss": 0.2758, + "step": 13082 + }, + { + "epoch": 0.79, + "learning_rate": 2.217715961139758e-06, + "loss": 0.2902, + "step": 13083 + }, + { + "epoch": 0.79, + "learning_rate": 2.2164873722423508e-06, + "loss": 0.2835, + "step": 13084 + }, + { + "epoch": 0.79, + "learning_rate": 2.2152590813342344e-06, + "loss": 0.273, + "step": 13085 + }, + { + "epoch": 0.79, + "learning_rate": 2.214031088462434e-06, + "loss": 0.2921, + "step": 13086 + }, + { + "epoch": 0.79, + "learning_rate": 2.212803393673967e-06, + "loss": 0.2683, + "step": 13087 + }, + { + "epoch": 0.79, + "learning_rate": 2.2115759970158325e-06, + "loss": 0.2662, + "step": 13088 + }, + { + "epoch": 0.79, + "learning_rate": 2.2103488985350207e-06, + "loss": 0.2603, + "step": 13089 + }, + { + "epoch": 0.79, + "learning_rate": 2.2091220982785113e-06, + "loss": 0.2773, + "step": 13090 + }, + { + "epoch": 0.79, + "learning_rate": 2.2078955962932693e-06, + "loss": 0.2722, + "step": 13091 + }, + { + "epoch": 0.79, + "learning_rate": 2.206669392626255e-06, + "loss": 0.2895, + "step": 13092 + }, + { + "epoch": 0.79, + "learning_rate": 2.2054434873244136e-06, + "loss": 0.2607, + "step": 13093 + }, + { + "epoch": 0.79, + "learning_rate": 2.204217880434675e-06, + "loss": 0.3011, + "step": 13094 + }, + { + "epoch": 0.79, + "learning_rate": 2.2029925720039633e-06, + "loss": 0.2924, + "step": 13095 + }, + { + "epoch": 0.79, + "learning_rate": 2.201767562079188e-06, + "loss": 0.2646, + "step": 13096 + }, + { + "epoch": 0.79, + "learning_rate": 2.200542850707247e-06, + "loss": 0.2755, + "step": 13097 + }, + { + "epoch": 0.79, + "learning_rate": 2.199318437935031e-06, + "loss": 0.28, + "step": 13098 + }, + { + "epoch": 0.79, + "learning_rate": 2.1980943238094154e-06, + "loss": 0.2718, + "step": 13099 + }, + { + "epoch": 0.79, + "learning_rate": 2.1968705083772645e-06, + "loss": 0.269, + "step": 13100 + }, + { + "epoch": 0.79, + "learning_rate": 2.195646991685432e-06, + "loss": 0.2518, + "step": 13101 + }, + { + "epoch": 0.79, + "learning_rate": 2.1944237737807592e-06, + "loss": 0.2616, + "step": 13102 + }, + { + "epoch": 0.79, + "learning_rate": 2.193200854710077e-06, + "loss": 0.2687, + "step": 13103 + }, + { + "epoch": 0.79, + "learning_rate": 2.1919782345202044e-06, + "loss": 0.2747, + "step": 13104 + }, + { + "epoch": 0.79, + "learning_rate": 2.1907559132579494e-06, + "loss": 0.2796, + "step": 13105 + }, + { + "epoch": 0.79, + "learning_rate": 2.1895338909701046e-06, + "loss": 0.2525, + "step": 13106 + }, + { + "epoch": 0.79, + "learning_rate": 2.18831216770346e-06, + "loss": 0.2447, + "step": 13107 + }, + { + "epoch": 0.79, + "learning_rate": 2.187090743504788e-06, + "loss": 0.2743, + "step": 13108 + }, + { + "epoch": 0.79, + "learning_rate": 2.185869618420848e-06, + "loss": 0.2676, + "step": 13109 + }, + { + "epoch": 0.79, + "learning_rate": 2.1846487924983927e-06, + "loss": 0.2718, + "step": 13110 + }, + { + "epoch": 0.79, + "learning_rate": 2.1834282657841578e-06, + "loss": 0.2651, + "step": 13111 + }, + { + "epoch": 0.79, + "learning_rate": 2.1822080383248746e-06, + "loss": 0.2573, + "step": 13112 + }, + { + "epoch": 0.79, + "learning_rate": 2.180988110167257e-06, + "loss": 0.2634, + "step": 13113 + }, + { + "epoch": 0.79, + "learning_rate": 2.179768481358011e-06, + "loss": 0.2785, + "step": 13114 + }, + { + "epoch": 0.79, + "learning_rate": 2.1785491519438286e-06, + "loss": 0.2626, + "step": 13115 + }, + { + "epoch": 0.79, + "learning_rate": 2.177330121971388e-06, + "loss": 0.2702, + "step": 13116 + }, + { + "epoch": 0.79, + "learning_rate": 2.176111391487367e-06, + "loss": 0.2702, + "step": 13117 + }, + { + "epoch": 0.79, + "learning_rate": 2.1748929605384195e-06, + "loss": 0.2884, + "step": 13118 + }, + { + "epoch": 0.79, + "learning_rate": 2.173674829171193e-06, + "loss": 0.2603, + "step": 13119 + }, + { + "epoch": 0.79, + "learning_rate": 2.172456997432324e-06, + "loss": 0.2666, + "step": 13120 + }, + { + "epoch": 0.79, + "learning_rate": 2.1712394653684343e-06, + "loss": 0.2556, + "step": 13121 + }, + { + "epoch": 0.79, + "learning_rate": 2.170022233026141e-06, + "loss": 0.274, + "step": 13122 + }, + { + "epoch": 0.79, + "learning_rate": 2.1688053004520437e-06, + "loss": 0.2787, + "step": 13123 + }, + { + "epoch": 0.79, + "learning_rate": 2.1675886676927326e-06, + "loss": 0.2679, + "step": 13124 + }, + { + "epoch": 0.79, + "learning_rate": 2.1663723347947853e-06, + "loss": 0.2602, + "step": 13125 + }, + { + "epoch": 0.79, + "learning_rate": 2.165156301804765e-06, + "loss": 0.2566, + "step": 13126 + }, + { + "epoch": 0.79, + "learning_rate": 2.163940568769235e-06, + "loss": 0.2818, + "step": 13127 + }, + { + "epoch": 0.79, + "learning_rate": 2.1627251357347345e-06, + "loss": 0.2646, + "step": 13128 + }, + { + "epoch": 0.79, + "learning_rate": 2.1615100027477976e-06, + "loss": 0.2621, + "step": 13129 + }, + { + "epoch": 0.79, + "learning_rate": 2.1602951698549447e-06, + "loss": 0.2808, + "step": 13130 + }, + { + "epoch": 0.79, + "learning_rate": 2.159080637102684e-06, + "loss": 0.2794, + "step": 13131 + }, + { + "epoch": 0.79, + "learning_rate": 2.157866404537512e-06, + "loss": 0.2747, + "step": 13132 + }, + { + "epoch": 0.79, + "learning_rate": 2.15665247220592e-06, + "loss": 0.2822, + "step": 13133 + }, + { + "epoch": 0.79, + "learning_rate": 2.1554388401543815e-06, + "loss": 0.2769, + "step": 13134 + }, + { + "epoch": 0.79, + "learning_rate": 2.1542255084293617e-06, + "loss": 0.2663, + "step": 13135 + }, + { + "epoch": 0.79, + "learning_rate": 2.153012477077305e-06, + "loss": 0.2679, + "step": 13136 + }, + { + "epoch": 0.79, + "learning_rate": 2.1517997461446593e-06, + "loss": 0.2775, + "step": 13137 + }, + { + "epoch": 0.79, + "learning_rate": 2.150587315677852e-06, + "loss": 0.264, + "step": 13138 + }, + { + "epoch": 0.79, + "learning_rate": 2.149375185723299e-06, + "loss": 0.2651, + "step": 13139 + }, + { + "epoch": 0.79, + "learning_rate": 2.1481633563274084e-06, + "loss": 0.2652, + "step": 13140 + }, + { + "epoch": 0.79, + "learning_rate": 2.146951827536571e-06, + "loss": 0.29, + "step": 13141 + }, + { + "epoch": 0.79, + "learning_rate": 2.1457405993971746e-06, + "loss": 0.2771, + "step": 13142 + }, + { + "epoch": 0.79, + "learning_rate": 2.1445296719555884e-06, + "loss": 0.2898, + "step": 13143 + }, + { + "epoch": 0.79, + "learning_rate": 2.1433190452581733e-06, + "loss": 0.2472, + "step": 13144 + }, + { + "epoch": 0.79, + "learning_rate": 2.1421087193512758e-06, + "loss": 0.2617, + "step": 13145 + }, + { + "epoch": 0.79, + "learning_rate": 2.1408986942812317e-06, + "loss": 0.266, + "step": 13146 + }, + { + "epoch": 0.79, + "learning_rate": 2.1396889700943714e-06, + "loss": 0.292, + "step": 13147 + }, + { + "epoch": 0.79, + "learning_rate": 2.138479546837006e-06, + "loss": 0.2487, + "step": 13148 + }, + { + "epoch": 0.79, + "learning_rate": 2.1372704245554378e-06, + "loss": 0.2677, + "step": 13149 + }, + { + "epoch": 0.79, + "learning_rate": 2.136061603295958e-06, + "loss": 0.2489, + "step": 13150 + }, + { + "epoch": 0.79, + "learning_rate": 2.134853083104843e-06, + "loss": 0.2539, + "step": 13151 + }, + { + "epoch": 0.79, + "learning_rate": 2.133644864028367e-06, + "loss": 0.2808, + "step": 13152 + }, + { + "epoch": 0.79, + "learning_rate": 2.1324369461127813e-06, + "loss": 0.2902, + "step": 13153 + }, + { + "epoch": 0.79, + "learning_rate": 2.131229329404333e-06, + "loss": 0.281, + "step": 13154 + }, + { + "epoch": 0.79, + "learning_rate": 2.1300220139492534e-06, + "loss": 0.2974, + "step": 13155 + }, + { + "epoch": 0.79, + "learning_rate": 2.1288149997937636e-06, + "loss": 0.2593, + "step": 13156 + }, + { + "epoch": 0.79, + "learning_rate": 2.1276082869840765e-06, + "loss": 0.3249, + "step": 13157 + }, + { + "epoch": 0.79, + "learning_rate": 2.126401875566391e-06, + "loss": 0.2725, + "step": 13158 + }, + { + "epoch": 0.79, + "learning_rate": 2.1251957655868915e-06, + "loss": 0.2771, + "step": 13159 + }, + { + "epoch": 0.8, + "learning_rate": 2.1239899570917555e-06, + "loss": 0.2787, + "step": 13160 + }, + { + "epoch": 0.8, + "learning_rate": 2.122784450127142e-06, + "loss": 0.2866, + "step": 13161 + }, + { + "epoch": 0.8, + "learning_rate": 2.1215792447392127e-06, + "loss": 0.2538, + "step": 13162 + }, + { + "epoch": 0.8, + "learning_rate": 2.1203743409741016e-06, + "loss": 0.2589, + "step": 13163 + }, + { + "epoch": 0.8, + "learning_rate": 2.119169738877941e-06, + "loss": 0.3008, + "step": 13164 + }, + { + "epoch": 0.8, + "learning_rate": 2.1179654384968463e-06, + "loss": 0.2418, + "step": 13165 + }, + { + "epoch": 0.8, + "learning_rate": 2.1167614398769266e-06, + "loss": 0.2826, + "step": 13166 + }, + { + "epoch": 0.8, + "learning_rate": 2.1155577430642714e-06, + "loss": 0.2716, + "step": 13167 + }, + { + "epoch": 0.8, + "learning_rate": 2.11435434810497e-06, + "loss": 0.2879, + "step": 13168 + }, + { + "epoch": 0.8, + "learning_rate": 2.113151255045095e-06, + "loss": 0.2614, + "step": 13169 + }, + { + "epoch": 0.8, + "learning_rate": 2.1119484639306997e-06, + "loss": 0.275, + "step": 13170 + }, + { + "epoch": 0.8, + "learning_rate": 2.1107459748078328e-06, + "loss": 0.2733, + "step": 13171 + }, + { + "epoch": 0.8, + "learning_rate": 2.1095437877225377e-06, + "loss": 0.26, + "step": 13172 + }, + { + "epoch": 0.8, + "learning_rate": 2.108341902720835e-06, + "loss": 0.2893, + "step": 13173 + }, + { + "epoch": 0.8, + "learning_rate": 2.1071403198487418e-06, + "loss": 0.2469, + "step": 13174 + }, + { + "epoch": 0.8, + "learning_rate": 2.105939039152256e-06, + "loss": 0.2737, + "step": 13175 + }, + { + "epoch": 0.8, + "learning_rate": 2.1047380606773692e-06, + "loss": 0.2771, + "step": 13176 + }, + { + "epoch": 0.8, + "learning_rate": 2.1035373844700647e-06, + "loss": 0.2557, + "step": 13177 + }, + { + "epoch": 0.8, + "learning_rate": 2.1023370105763064e-06, + "loss": 0.2372, + "step": 13178 + }, + { + "epoch": 0.8, + "learning_rate": 2.101136939042051e-06, + "loss": 0.2877, + "step": 13179 + }, + { + "epoch": 0.8, + "learning_rate": 2.099937169913242e-06, + "loss": 0.2842, + "step": 13180 + }, + { + "epoch": 0.8, + "learning_rate": 2.0987377032358114e-06, + "loss": 0.2489, + "step": 13181 + }, + { + "epoch": 0.8, + "learning_rate": 2.0975385390556848e-06, + "loss": 0.2688, + "step": 13182 + }, + { + "epoch": 0.8, + "learning_rate": 2.0963396774187684e-06, + "loss": 0.2772, + "step": 13183 + }, + { + "epoch": 0.8, + "learning_rate": 2.09514111837096e-06, + "loss": 0.2649, + "step": 13184 + }, + { + "epoch": 0.8, + "learning_rate": 2.093942861958148e-06, + "loss": 0.2754, + "step": 13185 + }, + { + "epoch": 0.8, + "learning_rate": 2.092744908226204e-06, + "loss": 0.2969, + "step": 13186 + }, + { + "epoch": 0.8, + "learning_rate": 2.091547257220995e-06, + "loss": 0.2514, + "step": 13187 + }, + { + "epoch": 0.8, + "learning_rate": 2.0903499089883717e-06, + "loss": 0.2662, + "step": 13188 + }, + { + "epoch": 0.8, + "learning_rate": 2.089152863574173e-06, + "loss": 0.2909, + "step": 13189 + }, + { + "epoch": 0.8, + "learning_rate": 2.0879561210242282e-06, + "loss": 0.2881, + "step": 13190 + }, + { + "epoch": 0.8, + "learning_rate": 2.0867596813843503e-06, + "loss": 0.285, + "step": 13191 + }, + { + "epoch": 0.8, + "learning_rate": 2.085563544700353e-06, + "loss": 0.2533, + "step": 13192 + }, + { + "epoch": 0.8, + "learning_rate": 2.0843677110180238e-06, + "loss": 0.2885, + "step": 13193 + }, + { + "epoch": 0.8, + "learning_rate": 2.0831721803831474e-06, + "loss": 0.2769, + "step": 13194 + }, + { + "epoch": 0.8, + "learning_rate": 2.0819769528414925e-06, + "loss": 0.2715, + "step": 13195 + }, + { + "epoch": 0.8, + "learning_rate": 2.0807820284388157e-06, + "loss": 0.2453, + "step": 13196 + }, + { + "epoch": 0.8, + "learning_rate": 2.07958740722087e-06, + "loss": 0.2476, + "step": 13197 + }, + { + "epoch": 0.8, + "learning_rate": 2.078393089233389e-06, + "loss": 0.3016, + "step": 13198 + }, + { + "epoch": 0.8, + "learning_rate": 2.077199074522096e-06, + "loss": 0.267, + "step": 13199 + }, + { + "epoch": 0.8, + "learning_rate": 2.0760053631327036e-06, + "loss": 0.2726, + "step": 13200 + }, + { + "epoch": 0.8, + "learning_rate": 2.0748119551109125e-06, + "loss": 0.3019, + "step": 13201 + }, + { + "epoch": 0.8, + "learning_rate": 2.0736188505024103e-06, + "loss": 0.2846, + "step": 13202 + }, + { + "epoch": 0.8, + "learning_rate": 2.0724260493528813e-06, + "loss": 0.2575, + "step": 13203 + }, + { + "epoch": 0.8, + "learning_rate": 2.071233551707984e-06, + "loss": 0.2944, + "step": 13204 + }, + { + "epoch": 0.8, + "learning_rate": 2.070041357613376e-06, + "loss": 0.2585, + "step": 13205 + }, + { + "epoch": 0.8, + "learning_rate": 2.0688494671146963e-06, + "loss": 0.2612, + "step": 13206 + }, + { + "epoch": 0.8, + "learning_rate": 2.067657880257582e-06, + "loss": 0.2773, + "step": 13207 + }, + { + "epoch": 0.8, + "learning_rate": 2.0664665970876496e-06, + "loss": 0.2573, + "step": 13208 + }, + { + "epoch": 0.8, + "learning_rate": 2.0652756176505085e-06, + "loss": 0.2896, + "step": 13209 + }, + { + "epoch": 0.8, + "learning_rate": 2.064084941991753e-06, + "loss": 0.2514, + "step": 13210 + }, + { + "epoch": 0.8, + "learning_rate": 2.062894570156966e-06, + "loss": 0.2829, + "step": 13211 + }, + { + "epoch": 0.8, + "learning_rate": 2.061704502191726e-06, + "loss": 0.2926, + "step": 13212 + }, + { + "epoch": 0.8, + "learning_rate": 2.0605147381415903e-06, + "loss": 0.264, + "step": 13213 + }, + { + "epoch": 0.8, + "learning_rate": 2.059325278052111e-06, + "loss": 0.2691, + "step": 13214 + }, + { + "epoch": 0.8, + "learning_rate": 2.0581361219688243e-06, + "loss": 0.3011, + "step": 13215 + }, + { + "epoch": 0.8, + "learning_rate": 2.056947269937255e-06, + "loss": 0.2486, + "step": 13216 + }, + { + "epoch": 0.8, + "learning_rate": 2.055758722002923e-06, + "loss": 0.2599, + "step": 13217 + }, + { + "epoch": 0.8, + "learning_rate": 2.0545704782113284e-06, + "loss": 0.2659, + "step": 13218 + }, + { + "epoch": 0.8, + "learning_rate": 2.053382538607963e-06, + "loss": 0.2713, + "step": 13219 + }, + { + "epoch": 0.8, + "learning_rate": 2.052194903238307e-06, + "loss": 0.263, + "step": 13220 + }, + { + "epoch": 0.8, + "learning_rate": 2.051007572147825e-06, + "loss": 0.2977, + "step": 13221 + }, + { + "epoch": 0.8, + "learning_rate": 2.04982054538198e-06, + "loss": 0.2753, + "step": 13222 + }, + { + "epoch": 0.8, + "learning_rate": 2.0486338229862147e-06, + "loss": 0.2807, + "step": 13223 + }, + { + "epoch": 0.8, + "learning_rate": 2.04744740500596e-06, + "loss": 0.2921, + "step": 13224 + }, + { + "epoch": 0.8, + "learning_rate": 2.046261291486641e-06, + "loss": 0.272, + "step": 13225 + }, + { + "epoch": 0.8, + "learning_rate": 2.045075482473662e-06, + "loss": 0.2613, + "step": 13226 + }, + { + "epoch": 0.8, + "learning_rate": 2.0438899780124276e-06, + "loss": 0.2623, + "step": 13227 + }, + { + "epoch": 0.8, + "learning_rate": 2.0427047781483224e-06, + "loss": 0.2893, + "step": 13228 + }, + { + "epoch": 0.8, + "learning_rate": 2.041519882926721e-06, + "loss": 0.2655, + "step": 13229 + }, + { + "epoch": 0.8, + "learning_rate": 2.0403352923929863e-06, + "loss": 0.2415, + "step": 13230 + }, + { + "epoch": 0.8, + "learning_rate": 2.039151006592468e-06, + "loss": 0.2921, + "step": 13231 + }, + { + "epoch": 0.8, + "learning_rate": 2.0379670255705118e-06, + "loss": 0.2913, + "step": 13232 + }, + { + "epoch": 0.8, + "learning_rate": 2.036783349372442e-06, + "loss": 0.2913, + "step": 13233 + }, + { + "epoch": 0.8, + "learning_rate": 2.0355999780435754e-06, + "loss": 0.2545, + "step": 13234 + }, + { + "epoch": 0.8, + "learning_rate": 2.034416911629219e-06, + "loss": 0.3011, + "step": 13235 + }, + { + "epoch": 0.8, + "learning_rate": 2.033234150174661e-06, + "loss": 0.2674, + "step": 13236 + }, + { + "epoch": 0.8, + "learning_rate": 2.0320516937251923e-06, + "loss": 0.2658, + "step": 13237 + }, + { + "epoch": 0.8, + "learning_rate": 2.030869542326075e-06, + "loss": 0.2694, + "step": 13238 + }, + { + "epoch": 0.8, + "learning_rate": 2.029687696022569e-06, + "loss": 0.2612, + "step": 13239 + }, + { + "epoch": 0.8, + "learning_rate": 2.0285061548599216e-06, + "loss": 0.273, + "step": 13240 + }, + { + "epoch": 0.8, + "learning_rate": 2.0273249188833656e-06, + "loss": 0.2699, + "step": 13241 + }, + { + "epoch": 0.8, + "learning_rate": 2.0261439881381293e-06, + "loss": 0.2725, + "step": 13242 + }, + { + "epoch": 0.8, + "learning_rate": 2.024963362669421e-06, + "loss": 0.3097, + "step": 13243 + }, + { + "epoch": 0.8, + "learning_rate": 2.0237830425224403e-06, + "loss": 0.2609, + "step": 13244 + }, + { + "epoch": 0.8, + "learning_rate": 2.022603027742377e-06, + "loss": 0.2554, + "step": 13245 + }, + { + "epoch": 0.8, + "learning_rate": 2.0214233183744037e-06, + "loss": 0.2874, + "step": 13246 + }, + { + "epoch": 0.8, + "learning_rate": 2.02024391446369e-06, + "loss": 0.2746, + "step": 13247 + }, + { + "epoch": 0.8, + "learning_rate": 2.019064816055387e-06, + "loss": 0.284, + "step": 13248 + }, + { + "epoch": 0.8, + "learning_rate": 2.017886023194637e-06, + "loss": 0.2667, + "step": 13249 + }, + { + "epoch": 0.8, + "learning_rate": 2.0167075359265685e-06, + "loss": 0.2702, + "step": 13250 + }, + { + "epoch": 0.8, + "learning_rate": 2.015529354296297e-06, + "loss": 0.2437, + "step": 13251 + }, + { + "epoch": 0.8, + "learning_rate": 2.014351478348935e-06, + "loss": 0.2619, + "step": 13252 + }, + { + "epoch": 0.8, + "learning_rate": 2.013173908129573e-06, + "loss": 0.256, + "step": 13253 + }, + { + "epoch": 0.8, + "learning_rate": 2.011996643683295e-06, + "loss": 0.2646, + "step": 13254 + }, + { + "epoch": 0.8, + "learning_rate": 2.010819685055172e-06, + "loss": 0.2843, + "step": 13255 + }, + { + "epoch": 0.8, + "learning_rate": 2.0096430322902604e-06, + "loss": 0.2874, + "step": 13256 + }, + { + "epoch": 0.8, + "learning_rate": 2.0084666854336145e-06, + "loss": 0.2609, + "step": 13257 + }, + { + "epoch": 0.8, + "learning_rate": 2.0072906445302665e-06, + "loss": 0.2673, + "step": 13258 + }, + { + "epoch": 0.8, + "learning_rate": 2.00611490962524e-06, + "loss": 0.2591, + "step": 13259 + }, + { + "epoch": 0.8, + "learning_rate": 2.004939480763549e-06, + "loss": 0.2811, + "step": 13260 + }, + { + "epoch": 0.8, + "learning_rate": 2.003764357990192e-06, + "loss": 0.3004, + "step": 13261 + }, + { + "epoch": 0.8, + "learning_rate": 2.0025895413501627e-06, + "loss": 0.2871, + "step": 13262 + }, + { + "epoch": 0.8, + "learning_rate": 2.0014150308884373e-06, + "loss": 0.2536, + "step": 13263 + }, + { + "epoch": 0.8, + "learning_rate": 2.0002408266499796e-06, + "loss": 0.2966, + "step": 13264 + }, + { + "epoch": 0.8, + "learning_rate": 1.999066928679744e-06, + "loss": 0.2712, + "step": 13265 + }, + { + "epoch": 0.8, + "learning_rate": 1.9978933370226707e-06, + "loss": 0.2522, + "step": 13266 + }, + { + "epoch": 0.8, + "learning_rate": 1.9967200517236963e-06, + "loss": 0.2727, + "step": 13267 + }, + { + "epoch": 0.8, + "learning_rate": 1.995547072827735e-06, + "loss": 0.2905, + "step": 13268 + }, + { + "epoch": 0.8, + "learning_rate": 1.9943744003796963e-06, + "loss": 0.2655, + "step": 13269 + }, + { + "epoch": 0.8, + "learning_rate": 1.993202034424474e-06, + "loss": 0.2581, + "step": 13270 + }, + { + "epoch": 0.8, + "learning_rate": 1.9920299750069526e-06, + "loss": 0.304, + "step": 13271 + }, + { + "epoch": 0.8, + "learning_rate": 1.990858222172003e-06, + "loss": 0.2529, + "step": 13272 + }, + { + "epoch": 0.8, + "learning_rate": 1.9896867759644857e-06, + "loss": 0.263, + "step": 13273 + }, + { + "epoch": 0.8, + "learning_rate": 1.98851563642925e-06, + "loss": 0.2708, + "step": 13274 + }, + { + "epoch": 0.8, + "learning_rate": 1.987344803611132e-06, + "loss": 0.2857, + "step": 13275 + }, + { + "epoch": 0.8, + "learning_rate": 1.9861742775549543e-06, + "loss": 0.2666, + "step": 13276 + }, + { + "epoch": 0.8, + "learning_rate": 1.985004058305535e-06, + "loss": 0.2764, + "step": 13277 + }, + { + "epoch": 0.8, + "learning_rate": 1.9838341459076725e-06, + "loss": 0.2542, + "step": 13278 + }, + { + "epoch": 0.8, + "learning_rate": 1.9826645404061584e-06, + "loss": 0.2684, + "step": 13279 + }, + { + "epoch": 0.8, + "learning_rate": 1.9814952418457687e-06, + "loss": 0.2641, + "step": 13280 + }, + { + "epoch": 0.8, + "learning_rate": 1.9803262502712674e-06, + "loss": 0.264, + "step": 13281 + }, + { + "epoch": 0.8, + "learning_rate": 1.9791575657274143e-06, + "loss": 0.2674, + "step": 13282 + }, + { + "epoch": 0.8, + "learning_rate": 1.977989188258951e-06, + "loss": 0.2829, + "step": 13283 + }, + { + "epoch": 0.8, + "learning_rate": 1.9768211179106054e-06, + "loss": 0.2831, + "step": 13284 + }, + { + "epoch": 0.8, + "learning_rate": 1.9756533547270994e-06, + "loss": 0.2546, + "step": 13285 + }, + { + "epoch": 0.8, + "learning_rate": 1.9744858987531366e-06, + "loss": 0.2651, + "step": 13286 + }, + { + "epoch": 0.8, + "learning_rate": 1.9733187500334183e-06, + "loss": 0.2746, + "step": 13287 + }, + { + "epoch": 0.8, + "learning_rate": 1.9721519086126263e-06, + "loss": 0.2921, + "step": 13288 + }, + { + "epoch": 0.8, + "learning_rate": 1.970985374535431e-06, + "loss": 0.2875, + "step": 13289 + }, + { + "epoch": 0.8, + "learning_rate": 1.9698191478464945e-06, + "loss": 0.2574, + "step": 13290 + }, + { + "epoch": 0.8, + "learning_rate": 1.9686532285904625e-06, + "loss": 0.2621, + "step": 13291 + }, + { + "epoch": 0.8, + "learning_rate": 1.967487616811976e-06, + "loss": 0.2631, + "step": 13292 + }, + { + "epoch": 0.8, + "learning_rate": 1.9663223125556584e-06, + "loss": 0.2701, + "step": 13293 + }, + { + "epoch": 0.8, + "learning_rate": 1.965157315866123e-06, + "loss": 0.2748, + "step": 13294 + }, + { + "epoch": 0.8, + "learning_rate": 1.9639926267879718e-06, + "loss": 0.2892, + "step": 13295 + }, + { + "epoch": 0.8, + "learning_rate": 1.962828245365791e-06, + "loss": 0.2335, + "step": 13296 + }, + { + "epoch": 0.8, + "learning_rate": 1.961664171644163e-06, + "loss": 0.2616, + "step": 13297 + }, + { + "epoch": 0.8, + "learning_rate": 1.9605004056676536e-06, + "loss": 0.2688, + "step": 13298 + }, + { + "epoch": 0.8, + "learning_rate": 1.9593369474808165e-06, + "loss": 0.279, + "step": 13299 + }, + { + "epoch": 0.8, + "learning_rate": 1.9581737971281932e-06, + "loss": 0.2631, + "step": 13300 + }, + { + "epoch": 0.8, + "learning_rate": 1.9570109546543124e-06, + "loss": 0.2631, + "step": 13301 + }, + { + "epoch": 0.8, + "learning_rate": 1.9558484201037e-06, + "loss": 0.2878, + "step": 13302 + }, + { + "epoch": 0.8, + "learning_rate": 1.9546861935208574e-06, + "loss": 0.29, + "step": 13303 + }, + { + "epoch": 0.8, + "learning_rate": 1.9535242749502826e-06, + "loss": 0.3104, + "step": 13304 + }, + { + "epoch": 0.8, + "learning_rate": 1.952362664436459e-06, + "loss": 0.2829, + "step": 13305 + }, + { + "epoch": 0.8, + "learning_rate": 1.951201362023858e-06, + "loss": 0.2824, + "step": 13306 + }, + { + "epoch": 0.8, + "learning_rate": 1.9500403677569404e-06, + "loss": 0.2626, + "step": 13307 + }, + { + "epoch": 0.8, + "learning_rate": 1.9488796816801526e-06, + "loss": 0.2905, + "step": 13308 + }, + { + "epoch": 0.8, + "learning_rate": 1.947719303837934e-06, + "loss": 0.2614, + "step": 13309 + }, + { + "epoch": 0.8, + "learning_rate": 1.9465592342747032e-06, + "loss": 0.263, + "step": 13310 + }, + { + "epoch": 0.8, + "learning_rate": 1.9453994730348814e-06, + "loss": 0.2757, + "step": 13311 + }, + { + "epoch": 0.8, + "learning_rate": 1.9442400201628652e-06, + "loss": 0.2785, + "step": 13312 + }, + { + "epoch": 0.8, + "learning_rate": 1.9430808757030452e-06, + "loss": 0.2734, + "step": 13313 + }, + { + "epoch": 0.8, + "learning_rate": 1.9419220396997983e-06, + "loss": 0.2761, + "step": 13314 + }, + { + "epoch": 0.8, + "learning_rate": 1.94076351219749e-06, + "loss": 0.2804, + "step": 13315 + }, + { + "epoch": 0.8, + "learning_rate": 1.939605293240472e-06, + "loss": 0.2644, + "step": 13316 + }, + { + "epoch": 0.8, + "learning_rate": 1.93844738287309e-06, + "loss": 0.2722, + "step": 13317 + }, + { + "epoch": 0.8, + "learning_rate": 1.937289781139674e-06, + "loss": 0.2962, + "step": 13318 + }, + { + "epoch": 0.8, + "learning_rate": 1.93613248808454e-06, + "loss": 0.279, + "step": 13319 + }, + { + "epoch": 0.8, + "learning_rate": 1.934975503751997e-06, + "loss": 0.2836, + "step": 13320 + }, + { + "epoch": 0.8, + "learning_rate": 1.9338188281863355e-06, + "loss": 0.2621, + "step": 13321 + }, + { + "epoch": 0.8, + "learning_rate": 1.9326624614318447e-06, + "loss": 0.2603, + "step": 13322 + }, + { + "epoch": 0.8, + "learning_rate": 1.9315064035327914e-06, + "loss": 0.2683, + "step": 13323 + }, + { + "epoch": 0.8, + "learning_rate": 1.9303506545334373e-06, + "loss": 0.2653, + "step": 13324 + }, + { + "epoch": 0.8, + "learning_rate": 1.929195214478028e-06, + "loss": 0.2815, + "step": 13325 + }, + { + "epoch": 0.81, + "learning_rate": 1.9280400834107972e-06, + "loss": 0.2759, + "step": 13326 + }, + { + "epoch": 0.81, + "learning_rate": 1.9268852613759737e-06, + "loss": 0.2532, + "step": 13327 + }, + { + "epoch": 0.81, + "learning_rate": 1.925730748417768e-06, + "loss": 0.2552, + "step": 13328 + }, + { + "epoch": 0.81, + "learning_rate": 1.9245765445803787e-06, + "loss": 0.2606, + "step": 13329 + }, + { + "epoch": 0.81, + "learning_rate": 1.9234226499079956e-06, + "loss": 0.2744, + "step": 13330 + }, + { + "epoch": 0.81, + "learning_rate": 1.9222690644447916e-06, + "loss": 0.2857, + "step": 13331 + }, + { + "epoch": 0.81, + "learning_rate": 1.9211157882349362e-06, + "loss": 0.2713, + "step": 13332 + }, + { + "epoch": 0.81, + "learning_rate": 1.9199628213225806e-06, + "loss": 0.2527, + "step": 13333 + }, + { + "epoch": 0.81, + "learning_rate": 1.9188101637518653e-06, + "loss": 0.2842, + "step": 13334 + }, + { + "epoch": 0.81, + "learning_rate": 1.917657815566919e-06, + "loss": 0.2641, + "step": 13335 + }, + { + "epoch": 0.81, + "learning_rate": 1.9165057768118578e-06, + "loss": 0.2732, + "step": 13336 + }, + { + "epoch": 0.81, + "learning_rate": 1.915354047530791e-06, + "loss": 0.2772, + "step": 13337 + }, + { + "epoch": 0.81, + "learning_rate": 1.9142026277678092e-06, + "loss": 0.2705, + "step": 13338 + }, + { + "epoch": 0.81, + "learning_rate": 1.9130515175669993e-06, + "loss": 0.2485, + "step": 13339 + }, + { + "epoch": 0.81, + "learning_rate": 1.91190071697242e-06, + "loss": 0.2732, + "step": 13340 + }, + { + "epoch": 0.81, + "learning_rate": 1.9107502260281398e-06, + "loss": 0.2782, + "step": 13341 + }, + { + "epoch": 0.81, + "learning_rate": 1.9096000447782006e-06, + "loss": 0.2785, + "step": 13342 + }, + { + "epoch": 0.81, + "learning_rate": 1.908450173266637e-06, + "loss": 0.266, + "step": 13343 + }, + { + "epoch": 0.81, + "learning_rate": 1.9073006115374726e-06, + "loss": 0.2497, + "step": 13344 + }, + { + "epoch": 0.81, + "learning_rate": 1.9061513596347137e-06, + "loss": 0.2783, + "step": 13345 + }, + { + "epoch": 0.81, + "learning_rate": 1.9050024176023662e-06, + "loss": 0.2637, + "step": 13346 + }, + { + "epoch": 0.81, + "learning_rate": 1.9038537854844131e-06, + "loss": 0.2599, + "step": 13347 + }, + { + "epoch": 0.81, + "learning_rate": 1.9027054633248287e-06, + "loss": 0.2719, + "step": 13348 + }, + { + "epoch": 0.81, + "learning_rate": 1.9015574511675783e-06, + "loss": 0.2487, + "step": 13349 + }, + { + "epoch": 0.81, + "learning_rate": 1.900409749056611e-06, + "loss": 0.2902, + "step": 13350 + }, + { + "epoch": 0.81, + "learning_rate": 1.8992623570358648e-06, + "loss": 0.279, + "step": 13351 + }, + { + "epoch": 0.81, + "learning_rate": 1.8981152751492725e-06, + "loss": 0.2799, + "step": 13352 + }, + { + "epoch": 0.81, + "learning_rate": 1.896968503440747e-06, + "loss": 0.2782, + "step": 13353 + }, + { + "epoch": 0.81, + "learning_rate": 1.8958220419541917e-06, + "loss": 0.2896, + "step": 13354 + }, + { + "epoch": 0.81, + "learning_rate": 1.8946758907334984e-06, + "loss": 0.2881, + "step": 13355 + }, + { + "epoch": 0.81, + "learning_rate": 1.893530049822546e-06, + "loss": 0.264, + "step": 13356 + }, + { + "epoch": 0.81, + "learning_rate": 1.8923845192652057e-06, + "loss": 0.2956, + "step": 13357 + }, + { + "epoch": 0.81, + "learning_rate": 1.891239299105333e-06, + "loss": 0.2926, + "step": 13358 + }, + { + "epoch": 0.81, + "learning_rate": 1.8900943893867707e-06, + "loss": 0.2803, + "step": 13359 + }, + { + "epoch": 0.81, + "learning_rate": 1.8889497901533526e-06, + "loss": 0.2605, + "step": 13360 + }, + { + "epoch": 0.81, + "learning_rate": 1.887805501448896e-06, + "loss": 0.2409, + "step": 13361 + }, + { + "epoch": 0.81, + "learning_rate": 1.886661523317216e-06, + "loss": 0.2549, + "step": 13362 + }, + { + "epoch": 0.81, + "learning_rate": 1.8855178558021047e-06, + "loss": 0.2493, + "step": 13363 + }, + { + "epoch": 0.81, + "learning_rate": 1.884374498947349e-06, + "loss": 0.2646, + "step": 13364 + }, + { + "epoch": 0.81, + "learning_rate": 1.883231452796721e-06, + "loss": 0.304, + "step": 13365 + }, + { + "epoch": 0.81, + "learning_rate": 1.8820887173939794e-06, + "loss": 0.2899, + "step": 13366 + }, + { + "epoch": 0.81, + "learning_rate": 1.880946292782878e-06, + "loss": 0.2734, + "step": 13367 + }, + { + "epoch": 0.81, + "learning_rate": 1.879804179007153e-06, + "loss": 0.2592, + "step": 13368 + }, + { + "epoch": 0.81, + "learning_rate": 1.8786623761105293e-06, + "loss": 0.2877, + "step": 13369 + }, + { + "epoch": 0.81, + "learning_rate": 1.8775208841367198e-06, + "loss": 0.2748, + "step": 13370 + }, + { + "epoch": 0.81, + "learning_rate": 1.876379703129425e-06, + "loss": 0.2898, + "step": 13371 + }, + { + "epoch": 0.81, + "learning_rate": 1.8752388331323379e-06, + "loss": 0.2812, + "step": 13372 + }, + { + "epoch": 0.81, + "learning_rate": 1.8740982741891379e-06, + "loss": 0.2628, + "step": 13373 + }, + { + "epoch": 0.81, + "learning_rate": 1.8729580263434842e-06, + "loss": 0.2799, + "step": 13374 + }, + { + "epoch": 0.81, + "learning_rate": 1.8718180896390325e-06, + "loss": 0.2599, + "step": 13375 + }, + { + "epoch": 0.81, + "learning_rate": 1.8706784641194297e-06, + "loss": 0.2683, + "step": 13376 + }, + { + "epoch": 0.81, + "learning_rate": 1.8695391498283032e-06, + "loss": 0.2531, + "step": 13377 + }, + { + "epoch": 0.81, + "learning_rate": 1.8684001468092706e-06, + "loss": 0.268, + "step": 13378 + }, + { + "epoch": 0.81, + "learning_rate": 1.8672614551059388e-06, + "loss": 0.267, + "step": 13379 + }, + { + "epoch": 0.81, + "learning_rate": 1.8661230747619007e-06, + "loss": 0.2683, + "step": 13380 + }, + { + "epoch": 0.81, + "learning_rate": 1.864985005820742e-06, + "loss": 0.2753, + "step": 13381 + }, + { + "epoch": 0.81, + "learning_rate": 1.8638472483260329e-06, + "loss": 0.2667, + "step": 13382 + }, + { + "epoch": 0.81, + "learning_rate": 1.8627098023213297e-06, + "loss": 0.2493, + "step": 13383 + }, + { + "epoch": 0.81, + "learning_rate": 1.861572667850181e-06, + "loss": 0.2818, + "step": 13384 + }, + { + "epoch": 0.81, + "learning_rate": 1.860435844956121e-06, + "loss": 0.2708, + "step": 13385 + }, + { + "epoch": 0.81, + "learning_rate": 1.8592993336826693e-06, + "loss": 0.2483, + "step": 13386 + }, + { + "epoch": 0.81, + "learning_rate": 1.858163134073343e-06, + "loss": 0.2525, + "step": 13387 + }, + { + "epoch": 0.81, + "learning_rate": 1.8570272461716389e-06, + "loss": 0.2879, + "step": 13388 + }, + { + "epoch": 0.81, + "learning_rate": 1.8558916700210438e-06, + "loss": 0.2644, + "step": 13389 + }, + { + "epoch": 0.81, + "learning_rate": 1.8547564056650314e-06, + "loss": 0.2812, + "step": 13390 + }, + { + "epoch": 0.81, + "learning_rate": 1.8536214531470641e-06, + "loss": 0.2808, + "step": 13391 + }, + { + "epoch": 0.81, + "learning_rate": 1.8524868125105987e-06, + "loss": 0.2836, + "step": 13392 + }, + { + "epoch": 0.81, + "learning_rate": 1.8513524837990704e-06, + "loss": 0.275, + "step": 13393 + }, + { + "epoch": 0.81, + "learning_rate": 1.8502184670559066e-06, + "loss": 0.2783, + "step": 13394 + }, + { + "epoch": 0.81, + "learning_rate": 1.8490847623245246e-06, + "loss": 0.2467, + "step": 13395 + }, + { + "epoch": 0.81, + "learning_rate": 1.847951369648323e-06, + "loss": 0.2916, + "step": 13396 + }, + { + "epoch": 0.81, + "learning_rate": 1.846818289070701e-06, + "loss": 0.2619, + "step": 13397 + }, + { + "epoch": 0.81, + "learning_rate": 1.8456855206350333e-06, + "loss": 0.2864, + "step": 13398 + }, + { + "epoch": 0.81, + "learning_rate": 1.8445530643846898e-06, + "loss": 0.2609, + "step": 13399 + }, + { + "epoch": 0.81, + "learning_rate": 1.8434209203630238e-06, + "loss": 0.2655, + "step": 13400 + }, + { + "epoch": 0.81, + "learning_rate": 1.8422890886133793e-06, + "loss": 0.2846, + "step": 13401 + }, + { + "epoch": 0.81, + "learning_rate": 1.8411575691790907e-06, + "loss": 0.2713, + "step": 13402 + }, + { + "epoch": 0.81, + "learning_rate": 1.8400263621034764e-06, + "loss": 0.2585, + "step": 13403 + }, + { + "epoch": 0.81, + "learning_rate": 1.8388954674298443e-06, + "loss": 0.2951, + "step": 13404 + }, + { + "epoch": 0.81, + "learning_rate": 1.8377648852014907e-06, + "loss": 0.284, + "step": 13405 + }, + { + "epoch": 0.81, + "learning_rate": 1.8366346154616954e-06, + "loss": 0.2799, + "step": 13406 + }, + { + "epoch": 0.81, + "learning_rate": 1.8355046582537406e-06, + "loss": 0.2734, + "step": 13407 + }, + { + "epoch": 0.81, + "learning_rate": 1.8343750136208771e-06, + "loss": 0.2779, + "step": 13408 + }, + { + "epoch": 0.81, + "learning_rate": 1.833245681606356e-06, + "loss": 0.2483, + "step": 13409 + }, + { + "epoch": 0.81, + "learning_rate": 1.8321166622534093e-06, + "loss": 0.297, + "step": 13410 + }, + { + "epoch": 0.81, + "learning_rate": 1.830987955605269e-06, + "loss": 0.2628, + "step": 13411 + }, + { + "epoch": 0.81, + "learning_rate": 1.8298595617051429e-06, + "loss": 0.246, + "step": 13412 + }, + { + "epoch": 0.81, + "learning_rate": 1.828731480596232e-06, + "loss": 0.2638, + "step": 13413 + }, + { + "epoch": 0.81, + "learning_rate": 1.8276037123217239e-06, + "loss": 0.2467, + "step": 13414 + }, + { + "epoch": 0.81, + "learning_rate": 1.826476256924793e-06, + "loss": 0.2666, + "step": 13415 + }, + { + "epoch": 0.81, + "learning_rate": 1.8253491144486068e-06, + "loss": 0.273, + "step": 13416 + }, + { + "epoch": 0.81, + "learning_rate": 1.8242222849363178e-06, + "loss": 0.2486, + "step": 13417 + }, + { + "epoch": 0.81, + "learning_rate": 1.8230957684310647e-06, + "loss": 0.2695, + "step": 13418 + }, + { + "epoch": 0.81, + "learning_rate": 1.8219695649759761e-06, + "loss": 0.2734, + "step": 13419 + }, + { + "epoch": 0.81, + "learning_rate": 1.8208436746141677e-06, + "loss": 0.2487, + "step": 13420 + }, + { + "epoch": 0.81, + "learning_rate": 1.819718097388743e-06, + "loss": 0.2864, + "step": 13421 + }, + { + "epoch": 0.81, + "learning_rate": 1.8185928333427982e-06, + "loss": 0.2713, + "step": 13422 + }, + { + "epoch": 0.81, + "learning_rate": 1.8174678825194103e-06, + "loss": 0.2832, + "step": 13423 + }, + { + "epoch": 0.81, + "learning_rate": 1.8163432449616503e-06, + "loss": 0.2563, + "step": 13424 + }, + { + "epoch": 0.81, + "learning_rate": 1.8152189207125714e-06, + "loss": 0.2374, + "step": 13425 + }, + { + "epoch": 0.81, + "learning_rate": 1.8140949098152184e-06, + "loss": 0.2655, + "step": 13426 + }, + { + "epoch": 0.81, + "learning_rate": 1.8129712123126263e-06, + "loss": 0.2782, + "step": 13427 + }, + { + "epoch": 0.81, + "learning_rate": 1.8118478282478158e-06, + "loss": 0.2708, + "step": 13428 + }, + { + "epoch": 0.81, + "learning_rate": 1.810724757663792e-06, + "loss": 0.266, + "step": 13429 + }, + { + "epoch": 0.81, + "learning_rate": 1.809602000603553e-06, + "loss": 0.2616, + "step": 13430 + }, + { + "epoch": 0.81, + "learning_rate": 1.808479557110081e-06, + "loss": 0.2575, + "step": 13431 + }, + { + "epoch": 0.81, + "learning_rate": 1.807357427226354e-06, + "loss": 0.2628, + "step": 13432 + }, + { + "epoch": 0.81, + "learning_rate": 1.8062356109953272e-06, + "loss": 0.2705, + "step": 13433 + }, + { + "epoch": 0.81, + "learning_rate": 1.805114108459951e-06, + "loss": 0.2655, + "step": 13434 + }, + { + "epoch": 0.81, + "learning_rate": 1.8039929196631623e-06, + "loss": 0.2607, + "step": 13435 + }, + { + "epoch": 0.81, + "learning_rate": 1.8028720446478808e-06, + "loss": 0.2879, + "step": 13436 + }, + { + "epoch": 0.81, + "learning_rate": 1.801751483457026e-06, + "loss": 0.2817, + "step": 13437 + }, + { + "epoch": 0.81, + "learning_rate": 1.8006312361334943e-06, + "loss": 0.2693, + "step": 13438 + }, + { + "epoch": 0.81, + "learning_rate": 1.799511302720175e-06, + "loss": 0.279, + "step": 13439 + }, + { + "epoch": 0.81, + "learning_rate": 1.7983916832599447e-06, + "loss": 0.2586, + "step": 13440 + }, + { + "epoch": 0.81, + "learning_rate": 1.7972723777956657e-06, + "loss": 0.268, + "step": 13441 + }, + { + "epoch": 0.81, + "learning_rate": 1.7961533863701919e-06, + "loss": 0.2581, + "step": 13442 + }, + { + "epoch": 0.81, + "learning_rate": 1.795034709026363e-06, + "loss": 0.2617, + "step": 13443 + }, + { + "epoch": 0.81, + "learning_rate": 1.7939163458070064e-06, + "loss": 0.268, + "step": 13444 + }, + { + "epoch": 0.81, + "learning_rate": 1.7927982967549384e-06, + "loss": 0.2719, + "step": 13445 + }, + { + "epoch": 0.81, + "learning_rate": 1.7916805619129652e-06, + "loss": 0.2513, + "step": 13446 + }, + { + "epoch": 0.81, + "learning_rate": 1.7905631413238777e-06, + "loss": 0.2628, + "step": 13447 + }, + { + "epoch": 0.81, + "learning_rate": 1.7894460350304565e-06, + "loss": 0.2627, + "step": 13448 + }, + { + "epoch": 0.81, + "learning_rate": 1.7883292430754685e-06, + "loss": 0.2899, + "step": 13449 + }, + { + "epoch": 0.81, + "learning_rate": 1.787212765501668e-06, + "loss": 0.2662, + "step": 13450 + }, + { + "epoch": 0.81, + "learning_rate": 1.7860966023518044e-06, + "loss": 0.2754, + "step": 13451 + }, + { + "epoch": 0.81, + "learning_rate": 1.784980753668606e-06, + "loss": 0.2546, + "step": 13452 + }, + { + "epoch": 0.81, + "learning_rate": 1.7838652194947936e-06, + "loss": 0.2733, + "step": 13453 + }, + { + "epoch": 0.81, + "learning_rate": 1.7827499998730747e-06, + "loss": 0.2496, + "step": 13454 + }, + { + "epoch": 0.81, + "learning_rate": 1.7816350948461447e-06, + "loss": 0.2785, + "step": 13455 + }, + { + "epoch": 0.81, + "learning_rate": 1.7805205044566854e-06, + "loss": 0.2585, + "step": 13456 + }, + { + "epoch": 0.81, + "learning_rate": 1.7794062287473734e-06, + "loss": 0.2758, + "step": 13457 + }, + { + "epoch": 0.81, + "learning_rate": 1.778292267760866e-06, + "loss": 0.2832, + "step": 13458 + }, + { + "epoch": 0.81, + "learning_rate": 1.7771786215398113e-06, + "loss": 0.2552, + "step": 13459 + }, + { + "epoch": 0.81, + "learning_rate": 1.7760652901268437e-06, + "loss": 0.2603, + "step": 13460 + }, + { + "epoch": 0.81, + "learning_rate": 1.7749522735645864e-06, + "loss": 0.2704, + "step": 13461 + }, + { + "epoch": 0.81, + "learning_rate": 1.7738395718956526e-06, + "loss": 0.2831, + "step": 13462 + }, + { + "epoch": 0.81, + "learning_rate": 1.7727271851626416e-06, + "loss": 0.2381, + "step": 13463 + }, + { + "epoch": 0.81, + "learning_rate": 1.7716151134081405e-06, + "loss": 0.2644, + "step": 13464 + }, + { + "epoch": 0.81, + "learning_rate": 1.7705033566747243e-06, + "loss": 0.2464, + "step": 13465 + }, + { + "epoch": 0.81, + "learning_rate": 1.7693919150049542e-06, + "loss": 0.2662, + "step": 13466 + }, + { + "epoch": 0.81, + "learning_rate": 1.7682807884413855e-06, + "loss": 0.2825, + "step": 13467 + }, + { + "epoch": 0.81, + "learning_rate": 1.767169977026555e-06, + "loss": 0.2494, + "step": 13468 + }, + { + "epoch": 0.81, + "learning_rate": 1.7660594808029907e-06, + "loss": 0.2797, + "step": 13469 + }, + { + "epoch": 0.81, + "learning_rate": 1.7649492998132067e-06, + "loss": 0.25, + "step": 13470 + }, + { + "epoch": 0.81, + "learning_rate": 1.7638394340997035e-06, + "loss": 0.2723, + "step": 13471 + }, + { + "epoch": 0.81, + "learning_rate": 1.7627298837049766e-06, + "loss": 0.2607, + "step": 13472 + }, + { + "epoch": 0.81, + "learning_rate": 1.7616206486715037e-06, + "loss": 0.2684, + "step": 13473 + }, + { + "epoch": 0.81, + "learning_rate": 1.7605117290417507e-06, + "loss": 0.2581, + "step": 13474 + }, + { + "epoch": 0.81, + "learning_rate": 1.7594031248581712e-06, + "loss": 0.2506, + "step": 13475 + }, + { + "epoch": 0.81, + "learning_rate": 1.7582948361632091e-06, + "loss": 0.251, + "step": 13476 + }, + { + "epoch": 0.81, + "learning_rate": 1.757186862999295e-06, + "loss": 0.2508, + "step": 13477 + }, + { + "epoch": 0.81, + "learning_rate": 1.756079205408846e-06, + "loss": 0.2641, + "step": 13478 + }, + { + "epoch": 0.81, + "learning_rate": 1.7549718634342694e-06, + "loss": 0.2669, + "step": 13479 + }, + { + "epoch": 0.81, + "learning_rate": 1.7538648371179579e-06, + "loss": 0.2547, + "step": 13480 + }, + { + "epoch": 0.81, + "learning_rate": 1.7527581265022964e-06, + "loss": 0.2719, + "step": 13481 + }, + { + "epoch": 0.81, + "learning_rate": 1.7516517316296534e-06, + "loss": 0.329, + "step": 13482 + }, + { + "epoch": 0.81, + "learning_rate": 1.750545652542387e-06, + "loss": 0.2702, + "step": 13483 + }, + { + "epoch": 0.81, + "learning_rate": 1.7494398892828435e-06, + "loss": 0.2999, + "step": 13484 + }, + { + "epoch": 0.81, + "learning_rate": 1.7483344418933535e-06, + "loss": 0.2821, + "step": 13485 + }, + { + "epoch": 0.81, + "learning_rate": 1.747229310416244e-06, + "loss": 0.2656, + "step": 13486 + }, + { + "epoch": 0.81, + "learning_rate": 1.7461244948938226e-06, + "loss": 0.2684, + "step": 13487 + }, + { + "epoch": 0.81, + "learning_rate": 1.7450199953683855e-06, + "loss": 0.2328, + "step": 13488 + }, + { + "epoch": 0.81, + "learning_rate": 1.7439158118822198e-06, + "loss": 0.2665, + "step": 13489 + }, + { + "epoch": 0.81, + "learning_rate": 1.7428119444775982e-06, + "loss": 0.2653, + "step": 13490 + }, + { + "epoch": 0.82, + "learning_rate": 1.7417083931967783e-06, + "loss": 0.2824, + "step": 13491 + }, + { + "epoch": 0.82, + "learning_rate": 1.740605158082015e-06, + "loss": 0.2503, + "step": 13492 + }, + { + "epoch": 0.82, + "learning_rate": 1.7395022391755433e-06, + "loss": 0.2672, + "step": 13493 + }, + { + "epoch": 0.82, + "learning_rate": 1.7383996365195866e-06, + "loss": 0.2574, + "step": 13494 + }, + { + "epoch": 0.82, + "learning_rate": 1.7372973501563595e-06, + "loss": 0.3012, + "step": 13495 + }, + { + "epoch": 0.82, + "learning_rate": 1.7361953801280595e-06, + "loss": 0.269, + "step": 13496 + }, + { + "epoch": 0.82, + "learning_rate": 1.7350937264768808e-06, + "loss": 0.2811, + "step": 13497 + }, + { + "epoch": 0.82, + "learning_rate": 1.733992389244996e-06, + "loss": 0.292, + "step": 13498 + }, + { + "epoch": 0.82, + "learning_rate": 1.7328913684745697e-06, + "loss": 0.2924, + "step": 13499 + }, + { + "epoch": 0.82, + "learning_rate": 1.7317906642077542e-06, + "loss": 0.2494, + "step": 13500 + }, + { + "epoch": 0.82, + "learning_rate": 1.7306902764866885e-06, + "loss": 0.2423, + "step": 13501 + }, + { + "epoch": 0.82, + "learning_rate": 1.729590205353504e-06, + "loss": 0.2691, + "step": 13502 + }, + { + "epoch": 0.82, + "learning_rate": 1.7284904508503153e-06, + "loss": 0.2628, + "step": 13503 + }, + { + "epoch": 0.82, + "learning_rate": 1.727391013019224e-06, + "loss": 0.2741, + "step": 13504 + }, + { + "epoch": 0.82, + "learning_rate": 1.726291891902324e-06, + "loss": 0.272, + "step": 13505 + }, + { + "epoch": 0.82, + "learning_rate": 1.7251930875416922e-06, + "loss": 0.2786, + "step": 13506 + }, + { + "epoch": 0.82, + "learning_rate": 1.724094599979399e-06, + "loss": 0.2677, + "step": 13507 + }, + { + "epoch": 0.82, + "learning_rate": 1.7229964292574996e-06, + "loss": 0.2853, + "step": 13508 + }, + { + "epoch": 0.82, + "learning_rate": 1.7218985754180372e-06, + "loss": 0.278, + "step": 13509 + }, + { + "epoch": 0.82, + "learning_rate": 1.720801038503037e-06, + "loss": 0.2478, + "step": 13510 + }, + { + "epoch": 0.82, + "learning_rate": 1.7197038185545244e-06, + "loss": 0.2628, + "step": 13511 + }, + { + "epoch": 0.82, + "learning_rate": 1.7186069156145046e-06, + "loss": 0.2521, + "step": 13512 + }, + { + "epoch": 0.82, + "learning_rate": 1.7175103297249719e-06, + "loss": 0.2722, + "step": 13513 + }, + { + "epoch": 0.82, + "learning_rate": 1.7164140609279078e-06, + "loss": 0.2765, + "step": 13514 + }, + { + "epoch": 0.82, + "learning_rate": 1.7153181092652815e-06, + "loss": 0.2711, + "step": 13515 + }, + { + "epoch": 0.82, + "learning_rate": 1.7142224747790548e-06, + "loss": 0.2893, + "step": 13516 + }, + { + "epoch": 0.82, + "learning_rate": 1.7131271575111718e-06, + "loss": 0.2584, + "step": 13517 + }, + { + "epoch": 0.82, + "learning_rate": 1.712032157503567e-06, + "loss": 0.2962, + "step": 13518 + }, + { + "epoch": 0.82, + "learning_rate": 1.710937474798161e-06, + "loss": 0.2545, + "step": 13519 + }, + { + "epoch": 0.82, + "learning_rate": 1.7098431094368618e-06, + "loss": 0.2606, + "step": 13520 + }, + { + "epoch": 0.82, + "learning_rate": 1.708749061461572e-06, + "loss": 0.2872, + "step": 13521 + }, + { + "epoch": 0.82, + "learning_rate": 1.707655330914173e-06, + "loss": 0.2616, + "step": 13522 + }, + { + "epoch": 0.82, + "learning_rate": 1.706561917836539e-06, + "loss": 0.2524, + "step": 13523 + }, + { + "epoch": 0.82, + "learning_rate": 1.7054688222705307e-06, + "loss": 0.2508, + "step": 13524 + }, + { + "epoch": 0.82, + "learning_rate": 1.704376044257996e-06, + "loss": 0.2667, + "step": 13525 + }, + { + "epoch": 0.82, + "learning_rate": 1.7032835838407714e-06, + "loss": 0.2535, + "step": 13526 + }, + { + "epoch": 0.82, + "learning_rate": 1.7021914410606844e-06, + "loss": 0.2575, + "step": 13527 + }, + { + "epoch": 0.82, + "learning_rate": 1.701099615959545e-06, + "loss": 0.2645, + "step": 13528 + }, + { + "epoch": 0.82, + "learning_rate": 1.7000081085791541e-06, + "loss": 0.286, + "step": 13529 + }, + { + "epoch": 0.82, + "learning_rate": 1.6989169189612997e-06, + "loss": 0.2679, + "step": 13530 + }, + { + "epoch": 0.82, + "learning_rate": 1.6978260471477547e-06, + "loss": 0.2499, + "step": 13531 + }, + { + "epoch": 0.82, + "learning_rate": 1.696735493180288e-06, + "loss": 0.2446, + "step": 13532 + }, + { + "epoch": 0.82, + "learning_rate": 1.6956452571006488e-06, + "loss": 0.2648, + "step": 13533 + }, + { + "epoch": 0.82, + "learning_rate": 1.6945553389505764e-06, + "loss": 0.2744, + "step": 13534 + }, + { + "epoch": 0.82, + "learning_rate": 1.6934657387717967e-06, + "loss": 0.2568, + "step": 13535 + }, + { + "epoch": 0.82, + "learning_rate": 1.6923764566060241e-06, + "loss": 0.2649, + "step": 13536 + }, + { + "epoch": 0.82, + "learning_rate": 1.6912874924949651e-06, + "loss": 0.2769, + "step": 13537 + }, + { + "epoch": 0.82, + "learning_rate": 1.6901988464803087e-06, + "loss": 0.2812, + "step": 13538 + }, + { + "epoch": 0.82, + "learning_rate": 1.6891105186037326e-06, + "loss": 0.2677, + "step": 13539 + }, + { + "epoch": 0.82, + "learning_rate": 1.6880225089069037e-06, + "loss": 0.2592, + "step": 13540 + }, + { + "epoch": 0.82, + "learning_rate": 1.6869348174314736e-06, + "loss": 0.2577, + "step": 13541 + }, + { + "epoch": 0.82, + "learning_rate": 1.6858474442190886e-06, + "loss": 0.3025, + "step": 13542 + }, + { + "epoch": 0.82, + "learning_rate": 1.6847603893113784e-06, + "loss": 0.2832, + "step": 13543 + }, + { + "epoch": 0.82, + "learning_rate": 1.6836736527499575e-06, + "loss": 0.2379, + "step": 13544 + }, + { + "epoch": 0.82, + "learning_rate": 1.6825872345764282e-06, + "loss": 0.2698, + "step": 13545 + }, + { + "epoch": 0.82, + "learning_rate": 1.6815011348323907e-06, + "loss": 0.2419, + "step": 13546 + }, + { + "epoch": 0.82, + "learning_rate": 1.6804153535594237e-06, + "loss": 0.2522, + "step": 13547 + }, + { + "epoch": 0.82, + "learning_rate": 1.679329890799094e-06, + "loss": 0.2628, + "step": 13548 + }, + { + "epoch": 0.82, + "learning_rate": 1.6782447465929597e-06, + "loss": 0.2711, + "step": 13549 + }, + { + "epoch": 0.82, + "learning_rate": 1.6771599209825628e-06, + "loss": 0.274, + "step": 13550 + }, + { + "epoch": 0.82, + "learning_rate": 1.6760754140094405e-06, + "loss": 0.2878, + "step": 13551 + }, + { + "epoch": 0.82, + "learning_rate": 1.6749912257151102e-06, + "loss": 0.2812, + "step": 13552 + }, + { + "epoch": 0.82, + "learning_rate": 1.673907356141079e-06, + "loss": 0.2688, + "step": 13553 + }, + { + "epoch": 0.82, + "learning_rate": 1.6728238053288425e-06, + "loss": 0.2665, + "step": 13554 + }, + { + "epoch": 0.82, + "learning_rate": 1.6717405733198832e-06, + "loss": 0.2747, + "step": 13555 + }, + { + "epoch": 0.82, + "learning_rate": 1.6706576601556746e-06, + "loss": 0.2747, + "step": 13556 + }, + { + "epoch": 0.82, + "learning_rate": 1.6695750658776766e-06, + "loss": 0.2447, + "step": 13557 + }, + { + "epoch": 0.82, + "learning_rate": 1.668492790527333e-06, + "loss": 0.2467, + "step": 13558 + }, + { + "epoch": 0.82, + "learning_rate": 1.6674108341460793e-06, + "loss": 0.2518, + "step": 13559 + }, + { + "epoch": 0.82, + "learning_rate": 1.666329196775336e-06, + "loss": 0.2884, + "step": 13560 + }, + { + "epoch": 0.82, + "learning_rate": 1.6652478784565173e-06, + "loss": 0.2715, + "step": 13561 + }, + { + "epoch": 0.82, + "learning_rate": 1.6641668792310195e-06, + "loss": 0.2549, + "step": 13562 + }, + { + "epoch": 0.82, + "learning_rate": 1.663086199140228e-06, + "loss": 0.2803, + "step": 13563 + }, + { + "epoch": 0.82, + "learning_rate": 1.6620058382255167e-06, + "loss": 0.2789, + "step": 13564 + }, + { + "epoch": 0.82, + "learning_rate": 1.6609257965282456e-06, + "loss": 0.2985, + "step": 13565 + }, + { + "epoch": 0.82, + "learning_rate": 1.6598460740897627e-06, + "loss": 0.261, + "step": 13566 + }, + { + "epoch": 0.82, + "learning_rate": 1.6587666709514083e-06, + "loss": 0.2619, + "step": 13567 + }, + { + "epoch": 0.82, + "learning_rate": 1.6576875871545062e-06, + "loss": 0.2701, + "step": 13568 + }, + { + "epoch": 0.82, + "learning_rate": 1.6566088227403675e-06, + "loss": 0.2838, + "step": 13569 + }, + { + "epoch": 0.82, + "learning_rate": 1.655530377750293e-06, + "loss": 0.2804, + "step": 13570 + }, + { + "epoch": 0.82, + "learning_rate": 1.6544522522255678e-06, + "loss": 0.3025, + "step": 13571 + }, + { + "epoch": 0.82, + "learning_rate": 1.6533744462074718e-06, + "loss": 0.256, + "step": 13572 + }, + { + "epoch": 0.82, + "learning_rate": 1.6522969597372684e-06, + "loss": 0.2606, + "step": 13573 + }, + { + "epoch": 0.82, + "learning_rate": 1.651219792856207e-06, + "loss": 0.2539, + "step": 13574 + }, + { + "epoch": 0.82, + "learning_rate": 1.6501429456055262e-06, + "loss": 0.242, + "step": 13575 + }, + { + "epoch": 0.82, + "learning_rate": 1.6490664180264515e-06, + "loss": 0.2875, + "step": 13576 + }, + { + "epoch": 0.82, + "learning_rate": 1.6479902101602042e-06, + "loss": 0.2736, + "step": 13577 + }, + { + "epoch": 0.82, + "learning_rate": 1.6469143220479788e-06, + "loss": 0.2711, + "step": 13578 + }, + { + "epoch": 0.82, + "learning_rate": 1.6458387537309683e-06, + "loss": 0.2912, + "step": 13579 + }, + { + "epoch": 0.82, + "learning_rate": 1.6447635052503486e-06, + "loss": 0.2533, + "step": 13580 + }, + { + "epoch": 0.82, + "learning_rate": 1.6436885766472887e-06, + "loss": 0.2836, + "step": 13581 + }, + { + "epoch": 0.82, + "learning_rate": 1.6426139679629417e-06, + "loss": 0.2465, + "step": 13582 + }, + { + "epoch": 0.82, + "learning_rate": 1.6415396792384465e-06, + "loss": 0.2571, + "step": 13583 + }, + { + "epoch": 0.82, + "learning_rate": 1.6404657105149324e-06, + "loss": 0.2751, + "step": 13584 + }, + { + "epoch": 0.82, + "learning_rate": 1.6393920618335135e-06, + "loss": 0.2606, + "step": 13585 + }, + { + "epoch": 0.82, + "learning_rate": 1.6383187332353002e-06, + "loss": 0.2852, + "step": 13586 + }, + { + "epoch": 0.82, + "learning_rate": 1.6372457247613804e-06, + "loss": 0.2748, + "step": 13587 + }, + { + "epoch": 0.82, + "learning_rate": 1.6361730364528339e-06, + "loss": 0.2448, + "step": 13588 + }, + { + "epoch": 0.82, + "learning_rate": 1.6351006683507298e-06, + "loss": 0.2789, + "step": 13589 + }, + { + "epoch": 0.82, + "learning_rate": 1.634028620496121e-06, + "loss": 0.25, + "step": 13590 + }, + { + "epoch": 0.82, + "learning_rate": 1.6329568929300532e-06, + "loss": 0.2635, + "step": 13591 + }, + { + "epoch": 0.82, + "learning_rate": 1.6318854856935563e-06, + "loss": 0.2434, + "step": 13592 + }, + { + "epoch": 0.82, + "learning_rate": 1.6308143988276482e-06, + "loss": 0.2753, + "step": 13593 + }, + { + "epoch": 0.82, + "learning_rate": 1.6297436323733363e-06, + "loss": 0.2619, + "step": 13594 + }, + { + "epoch": 0.82, + "learning_rate": 1.6286731863716098e-06, + "loss": 0.2759, + "step": 13595 + }, + { + "epoch": 0.82, + "learning_rate": 1.6276030608634575e-06, + "loss": 0.2676, + "step": 13596 + }, + { + "epoch": 0.82, + "learning_rate": 1.626533255889845e-06, + "loss": 0.2743, + "step": 13597 + }, + { + "epoch": 0.82, + "learning_rate": 1.625463771491731e-06, + "loss": 0.3022, + "step": 13598 + }, + { + "epoch": 0.82, + "learning_rate": 1.624394607710058e-06, + "loss": 0.2673, + "step": 13599 + }, + { + "epoch": 0.82, + "learning_rate": 1.6233257645857615e-06, + "loss": 0.285, + "step": 13600 + }, + { + "epoch": 0.82, + "learning_rate": 1.622257242159756e-06, + "loss": 0.2699, + "step": 13601 + }, + { + "epoch": 0.82, + "learning_rate": 1.621189040472958e-06, + "loss": 0.2628, + "step": 13602 + }, + { + "epoch": 0.82, + "learning_rate": 1.620121159566258e-06, + "loss": 0.2492, + "step": 13603 + }, + { + "epoch": 0.82, + "learning_rate": 1.61905359948054e-06, + "loss": 0.2379, + "step": 13604 + }, + { + "epoch": 0.82, + "learning_rate": 1.617986360256677e-06, + "loss": 0.2448, + "step": 13605 + }, + { + "epoch": 0.82, + "learning_rate": 1.6169194419355238e-06, + "loss": 0.2718, + "step": 13606 + }, + { + "epoch": 0.82, + "learning_rate": 1.6158528445579314e-06, + "loss": 0.2772, + "step": 13607 + }, + { + "epoch": 0.82, + "learning_rate": 1.6147865681647335e-06, + "loss": 0.2633, + "step": 13608 + }, + { + "epoch": 0.82, + "learning_rate": 1.6137206127967509e-06, + "loss": 0.2765, + "step": 13609 + }, + { + "epoch": 0.82, + "learning_rate": 1.6126549784947943e-06, + "loss": 0.2531, + "step": 13610 + }, + { + "epoch": 0.82, + "learning_rate": 1.61158966529966e-06, + "loss": 0.26, + "step": 13611 + }, + { + "epoch": 0.82, + "learning_rate": 1.610524673252134e-06, + "loss": 0.2577, + "step": 13612 + }, + { + "epoch": 0.82, + "learning_rate": 1.6094600023929885e-06, + "loss": 0.2736, + "step": 13613 + }, + { + "epoch": 0.82, + "learning_rate": 1.608395652762985e-06, + "loss": 0.2912, + "step": 13614 + }, + { + "epoch": 0.82, + "learning_rate": 1.6073316244028681e-06, + "loss": 0.3013, + "step": 13615 + }, + { + "epoch": 0.82, + "learning_rate": 1.6062679173533802e-06, + "loss": 0.2397, + "step": 13616 + }, + { + "epoch": 0.82, + "learning_rate": 1.605204531655241e-06, + "loss": 0.2453, + "step": 13617 + }, + { + "epoch": 0.82, + "learning_rate": 1.604141467349164e-06, + "loss": 0.2624, + "step": 13618 + }, + { + "epoch": 0.82, + "learning_rate": 1.603078724475846e-06, + "loss": 0.2956, + "step": 13619 + }, + { + "epoch": 0.82, + "learning_rate": 1.6020163030759727e-06, + "loss": 0.2561, + "step": 13620 + }, + { + "epoch": 0.82, + "learning_rate": 1.6009542031902226e-06, + "loss": 0.2561, + "step": 13621 + }, + { + "epoch": 0.82, + "learning_rate": 1.5998924248592552e-06, + "loss": 0.291, + "step": 13622 + }, + { + "epoch": 0.82, + "learning_rate": 1.5988309681237224e-06, + "loss": 0.2649, + "step": 13623 + }, + { + "epoch": 0.82, + "learning_rate": 1.5977698330242598e-06, + "loss": 0.2792, + "step": 13624 + }, + { + "epoch": 0.82, + "learning_rate": 1.59670901960149e-06, + "loss": 0.2585, + "step": 13625 + }, + { + "epoch": 0.82, + "learning_rate": 1.595648527896032e-06, + "loss": 0.2941, + "step": 13626 + }, + { + "epoch": 0.82, + "learning_rate": 1.5945883579484833e-06, + "loss": 0.2857, + "step": 13627 + }, + { + "epoch": 0.82, + "learning_rate": 1.5935285097994325e-06, + "loss": 0.2754, + "step": 13628 + }, + { + "epoch": 0.82, + "learning_rate": 1.5924689834894546e-06, + "loss": 0.2478, + "step": 13629 + }, + { + "epoch": 0.82, + "learning_rate": 1.591409779059112e-06, + "loss": 0.2944, + "step": 13630 + }, + { + "epoch": 0.82, + "learning_rate": 1.59035089654896e-06, + "loss": 0.285, + "step": 13631 + }, + { + "epoch": 0.82, + "learning_rate": 1.589292335999536e-06, + "loss": 0.2835, + "step": 13632 + }, + { + "epoch": 0.82, + "learning_rate": 1.5882340974513665e-06, + "loss": 0.2852, + "step": 13633 + }, + { + "epoch": 0.82, + "learning_rate": 1.5871761809449648e-06, + "loss": 0.267, + "step": 13634 + }, + { + "epoch": 0.82, + "learning_rate": 1.586118586520834e-06, + "loss": 0.2732, + "step": 13635 + }, + { + "epoch": 0.82, + "learning_rate": 1.5850613142194604e-06, + "loss": 0.274, + "step": 13636 + }, + { + "epoch": 0.82, + "learning_rate": 1.5840043640813274e-06, + "loss": 0.2787, + "step": 13637 + }, + { + "epoch": 0.82, + "learning_rate": 1.582947736146896e-06, + "loss": 0.2733, + "step": 13638 + }, + { + "epoch": 0.82, + "learning_rate": 1.58189143045662e-06, + "loss": 0.2472, + "step": 13639 + }, + { + "epoch": 0.82, + "learning_rate": 1.5808354470509403e-06, + "loss": 0.261, + "step": 13640 + }, + { + "epoch": 0.82, + "learning_rate": 1.5797797859702813e-06, + "loss": 0.2596, + "step": 13641 + }, + { + "epoch": 0.82, + "learning_rate": 1.5787244472550633e-06, + "loss": 0.2627, + "step": 13642 + }, + { + "epoch": 0.82, + "learning_rate": 1.5776694309456886e-06, + "loss": 0.2383, + "step": 13643 + }, + { + "epoch": 0.82, + "learning_rate": 1.5766147370825468e-06, + "loss": 0.2561, + "step": 13644 + }, + { + "epoch": 0.82, + "learning_rate": 1.575560365706017e-06, + "loss": 0.2546, + "step": 13645 + }, + { + "epoch": 0.82, + "learning_rate": 1.5745063168564657e-06, + "loss": 0.2879, + "step": 13646 + }, + { + "epoch": 0.82, + "learning_rate": 1.573452590574246e-06, + "loss": 0.2775, + "step": 13647 + }, + { + "epoch": 0.82, + "learning_rate": 1.5723991868997002e-06, + "loss": 0.2866, + "step": 13648 + }, + { + "epoch": 0.82, + "learning_rate": 1.571346105873157e-06, + "loss": 0.2532, + "step": 13649 + }, + { + "epoch": 0.82, + "learning_rate": 1.5702933475349324e-06, + "loss": 0.2769, + "step": 13650 + }, + { + "epoch": 0.82, + "learning_rate": 1.5692409119253337e-06, + "loss": 0.2803, + "step": 13651 + }, + { + "epoch": 0.82, + "learning_rate": 1.5681887990846523e-06, + "loss": 0.2719, + "step": 13652 + }, + { + "epoch": 0.82, + "learning_rate": 1.5671370090531669e-06, + "loss": 0.2718, + "step": 13653 + }, + { + "epoch": 0.82, + "learning_rate": 1.566085541871145e-06, + "loss": 0.2393, + "step": 13654 + }, + { + "epoch": 0.82, + "learning_rate": 1.5650343975788397e-06, + "loss": 0.257, + "step": 13655 + }, + { + "epoch": 0.82, + "learning_rate": 1.563983576216499e-06, + "loss": 0.2769, + "step": 13656 + }, + { + "epoch": 0.83, + "learning_rate": 1.562933077824349e-06, + "loss": 0.281, + "step": 13657 + }, + { + "epoch": 0.83, + "learning_rate": 1.5618829024426097e-06, + "loss": 0.2655, + "step": 13658 + }, + { + "epoch": 0.83, + "learning_rate": 1.5608330501114854e-06, + "loss": 0.2471, + "step": 13659 + }, + { + "epoch": 0.83, + "learning_rate": 1.5597835208711687e-06, + "loss": 0.2949, + "step": 13660 + }, + { + "epoch": 0.83, + "learning_rate": 1.5587343147618438e-06, + "loss": 0.2626, + "step": 13661 + }, + { + "epoch": 0.83, + "learning_rate": 1.5576854318236768e-06, + "loss": 0.2779, + "step": 13662 + }, + { + "epoch": 0.83, + "learning_rate": 1.5566368720968239e-06, + "loss": 0.2543, + "step": 13663 + }, + { + "epoch": 0.83, + "learning_rate": 1.5555886356214288e-06, + "loss": 0.2748, + "step": 13664 + }, + { + "epoch": 0.83, + "learning_rate": 1.554540722437622e-06, + "loss": 0.258, + "step": 13665 + }, + { + "epoch": 0.83, + "learning_rate": 1.5534931325855262e-06, + "loss": 0.2799, + "step": 13666 + }, + { + "epoch": 0.83, + "learning_rate": 1.5524458661052444e-06, + "loss": 0.264, + "step": 13667 + }, + { + "epoch": 0.83, + "learning_rate": 1.5513989230368731e-06, + "loss": 0.2626, + "step": 13668 + }, + { + "epoch": 0.83, + "learning_rate": 1.5503523034204927e-06, + "loss": 0.279, + "step": 13669 + }, + { + "epoch": 0.83, + "learning_rate": 1.5493060072961742e-06, + "loss": 0.2854, + "step": 13670 + }, + { + "epoch": 0.83, + "learning_rate": 1.54826003470397e-06, + "loss": 0.2642, + "step": 13671 + }, + { + "epoch": 0.83, + "learning_rate": 1.5472143856839316e-06, + "loss": 0.2994, + "step": 13672 + }, + { + "epoch": 0.83, + "learning_rate": 1.5461690602760882e-06, + "loss": 0.2776, + "step": 13673 + }, + { + "epoch": 0.83, + "learning_rate": 1.5451240585204597e-06, + "loss": 0.2828, + "step": 13674 + }, + { + "epoch": 0.83, + "learning_rate": 1.5440793804570532e-06, + "loss": 0.2849, + "step": 13675 + }, + { + "epoch": 0.83, + "learning_rate": 1.543035026125863e-06, + "loss": 0.2744, + "step": 13676 + }, + { + "epoch": 0.83, + "learning_rate": 1.5419909955668754e-06, + "loss": 0.3062, + "step": 13677 + }, + { + "epoch": 0.83, + "learning_rate": 1.5409472888200583e-06, + "loss": 0.2859, + "step": 13678 + }, + { + "epoch": 0.83, + "learning_rate": 1.5399039059253728e-06, + "loss": 0.2688, + "step": 13679 + }, + { + "epoch": 0.83, + "learning_rate": 1.5388608469227563e-06, + "loss": 0.2715, + "step": 13680 + }, + { + "epoch": 0.83, + "learning_rate": 1.5378181118521506e-06, + "loss": 0.2902, + "step": 13681 + }, + { + "epoch": 0.83, + "learning_rate": 1.5367757007534733e-06, + "loss": 0.2815, + "step": 13682 + }, + { + "epoch": 0.83, + "learning_rate": 1.5357336136666323e-06, + "loss": 0.2503, + "step": 13683 + }, + { + "epoch": 0.83, + "learning_rate": 1.5346918506315244e-06, + "loss": 0.2923, + "step": 13684 + }, + { + "epoch": 0.83, + "learning_rate": 1.5336504116880303e-06, + "loss": 0.2812, + "step": 13685 + }, + { + "epoch": 0.83, + "learning_rate": 1.5326092968760254e-06, + "loss": 0.2475, + "step": 13686 + }, + { + "epoch": 0.83, + "learning_rate": 1.531568506235367e-06, + "loss": 0.2818, + "step": 13687 + }, + { + "epoch": 0.83, + "learning_rate": 1.5305280398059018e-06, + "loss": 0.2641, + "step": 13688 + }, + { + "epoch": 0.83, + "learning_rate": 1.529487897627463e-06, + "loss": 0.2674, + "step": 13689 + }, + { + "epoch": 0.83, + "learning_rate": 1.5284480797398694e-06, + "loss": 0.2702, + "step": 13690 + }, + { + "epoch": 0.83, + "learning_rate": 1.5274085861829345e-06, + "loss": 0.286, + "step": 13691 + }, + { + "epoch": 0.83, + "learning_rate": 1.526369416996455e-06, + "loss": 0.2592, + "step": 13692 + }, + { + "epoch": 0.83, + "learning_rate": 1.5253305722202127e-06, + "loss": 0.2637, + "step": 13693 + }, + { + "epoch": 0.83, + "learning_rate": 1.5242920518939796e-06, + "loss": 0.2613, + "step": 13694 + }, + { + "epoch": 0.83, + "learning_rate": 1.523253856057515e-06, + "loss": 0.2779, + "step": 13695 + }, + { + "epoch": 0.83, + "learning_rate": 1.522215984750568e-06, + "loss": 0.2697, + "step": 13696 + }, + { + "epoch": 0.83, + "learning_rate": 1.5211784380128714e-06, + "loss": 0.2515, + "step": 13697 + }, + { + "epoch": 0.83, + "learning_rate": 1.5201412158841488e-06, + "loss": 0.2658, + "step": 13698 + }, + { + "epoch": 0.83, + "learning_rate": 1.5191043184041078e-06, + "loss": 0.284, + "step": 13699 + }, + { + "epoch": 0.83, + "learning_rate": 1.5180677456124448e-06, + "loss": 0.2732, + "step": 13700 + }, + { + "epoch": 0.83, + "learning_rate": 1.5170314975488499e-06, + "loss": 0.2713, + "step": 13701 + }, + { + "epoch": 0.83, + "learning_rate": 1.515995574252992e-06, + "loss": 0.2659, + "step": 13702 + }, + { + "epoch": 0.83, + "learning_rate": 1.5149599757645305e-06, + "loss": 0.2564, + "step": 13703 + }, + { + "epoch": 0.83, + "learning_rate": 1.513924702123114e-06, + "loss": 0.2699, + "step": 13704 + }, + { + "epoch": 0.83, + "learning_rate": 1.5128897533683784e-06, + "loss": 0.2829, + "step": 13705 + }, + { + "epoch": 0.83, + "learning_rate": 1.511855129539942e-06, + "loss": 0.2739, + "step": 13706 + }, + { + "epoch": 0.83, + "learning_rate": 1.5108208306774207e-06, + "loss": 0.2988, + "step": 13707 + }, + { + "epoch": 0.83, + "learning_rate": 1.509786856820411e-06, + "loss": 0.2626, + "step": 13708 + }, + { + "epoch": 0.83, + "learning_rate": 1.5087532080084976e-06, + "loss": 0.2471, + "step": 13709 + }, + { + "epoch": 0.83, + "learning_rate": 1.5077198842812523e-06, + "loss": 0.2679, + "step": 13710 + }, + { + "epoch": 0.83, + "learning_rate": 1.5066868856782336e-06, + "loss": 0.3011, + "step": 13711 + }, + { + "epoch": 0.83, + "learning_rate": 1.505654212238996e-06, + "loss": 0.2507, + "step": 13712 + }, + { + "epoch": 0.83, + "learning_rate": 1.504621864003073e-06, + "loss": 0.2793, + "step": 13713 + }, + { + "epoch": 0.83, + "learning_rate": 1.5035898410099835e-06, + "loss": 0.2557, + "step": 13714 + }, + { + "epoch": 0.83, + "learning_rate": 1.5025581432992387e-06, + "loss": 0.278, + "step": 13715 + }, + { + "epoch": 0.83, + "learning_rate": 1.501526770910342e-06, + "loss": 0.2556, + "step": 13716 + }, + { + "epoch": 0.83, + "learning_rate": 1.5004957238827755e-06, + "loss": 0.3382, + "step": 13717 + }, + { + "epoch": 0.83, + "learning_rate": 1.499465002256013e-06, + "loss": 0.2602, + "step": 13718 + }, + { + "epoch": 0.83, + "learning_rate": 1.4984346060695154e-06, + "loss": 0.2662, + "step": 13719 + }, + { + "epoch": 0.83, + "learning_rate": 1.4974045353627286e-06, + "loss": 0.2676, + "step": 13720 + }, + { + "epoch": 0.83, + "learning_rate": 1.4963747901750936e-06, + "loss": 0.2842, + "step": 13721 + }, + { + "epoch": 0.83, + "learning_rate": 1.495345370546032e-06, + "loss": 0.2698, + "step": 13722 + }, + { + "epoch": 0.83, + "learning_rate": 1.4943162765149533e-06, + "loss": 0.2799, + "step": 13723 + }, + { + "epoch": 0.83, + "learning_rate": 1.4932875081212561e-06, + "loss": 0.2948, + "step": 13724 + }, + { + "epoch": 0.83, + "learning_rate": 1.4922590654043267e-06, + "loss": 0.278, + "step": 13725 + }, + { + "epoch": 0.83, + "learning_rate": 1.4912309484035403e-06, + "loss": 0.2949, + "step": 13726 + }, + { + "epoch": 0.83, + "learning_rate": 1.4902031571582576e-06, + "loss": 0.2599, + "step": 13727 + }, + { + "epoch": 0.83, + "learning_rate": 1.489175691707826e-06, + "loss": 0.2718, + "step": 13728 + }, + { + "epoch": 0.83, + "learning_rate": 1.4881485520915828e-06, + "loss": 0.254, + "step": 13729 + }, + { + "epoch": 0.83, + "learning_rate": 1.487121738348849e-06, + "loss": 0.2557, + "step": 13730 + }, + { + "epoch": 0.83, + "learning_rate": 1.4860952505189396e-06, + "loss": 0.2578, + "step": 13731 + }, + { + "epoch": 0.83, + "learning_rate": 1.4850690886411523e-06, + "loss": 0.275, + "step": 13732 + }, + { + "epoch": 0.83, + "learning_rate": 1.4840432527547731e-06, + "loss": 0.2472, + "step": 13733 + }, + { + "epoch": 0.83, + "learning_rate": 1.4830177428990745e-06, + "loss": 0.2612, + "step": 13734 + }, + { + "epoch": 0.83, + "learning_rate": 1.4819925591133177e-06, + "loss": 0.2677, + "step": 13735 + }, + { + "epoch": 0.83, + "learning_rate": 1.480967701436755e-06, + "loss": 0.2365, + "step": 13736 + }, + { + "epoch": 0.83, + "learning_rate": 1.4799431699086197e-06, + "loss": 0.2676, + "step": 13737 + }, + { + "epoch": 0.83, + "learning_rate": 1.478918964568138e-06, + "loss": 0.2758, + "step": 13738 + }, + { + "epoch": 0.83, + "learning_rate": 1.4778950854545182e-06, + "loss": 0.2535, + "step": 13739 + }, + { + "epoch": 0.83, + "learning_rate": 1.4768715326069606e-06, + "loss": 0.29, + "step": 13740 + }, + { + "epoch": 0.83, + "learning_rate": 1.47584830606465e-06, + "loss": 0.3005, + "step": 13741 + }, + { + "epoch": 0.83, + "learning_rate": 1.4748254058667644e-06, + "loss": 0.2713, + "step": 13742 + }, + { + "epoch": 0.83, + "learning_rate": 1.4738028320524634e-06, + "loss": 0.2718, + "step": 13743 + }, + { + "epoch": 0.83, + "learning_rate": 1.4727805846608945e-06, + "loss": 0.2644, + "step": 13744 + }, + { + "epoch": 0.83, + "learning_rate": 1.4717586637311943e-06, + "loss": 0.263, + "step": 13745 + }, + { + "epoch": 0.83, + "learning_rate": 1.470737069302486e-06, + "loss": 0.2637, + "step": 13746 + }, + { + "epoch": 0.83, + "learning_rate": 1.469715801413887e-06, + "loss": 0.248, + "step": 13747 + }, + { + "epoch": 0.83, + "learning_rate": 1.4686948601044894e-06, + "loss": 0.2992, + "step": 13748 + }, + { + "epoch": 0.83, + "learning_rate": 1.4676742454133807e-06, + "loss": 0.3039, + "step": 13749 + }, + { + "epoch": 0.83, + "learning_rate": 1.4666539573796345e-06, + "loss": 0.2687, + "step": 13750 + }, + { + "epoch": 0.83, + "learning_rate": 1.4656339960423149e-06, + "loss": 0.2619, + "step": 13751 + }, + { + "epoch": 0.83, + "learning_rate": 1.4646143614404707e-06, + "loss": 0.2606, + "step": 13752 + }, + { + "epoch": 0.83, + "learning_rate": 1.463595053613137e-06, + "loss": 0.2607, + "step": 13753 + }, + { + "epoch": 0.83, + "learning_rate": 1.4625760725993376e-06, + "loss": 0.2528, + "step": 13754 + }, + { + "epoch": 0.83, + "learning_rate": 1.461557418438081e-06, + "loss": 0.262, + "step": 13755 + }, + { + "epoch": 0.83, + "learning_rate": 1.4605390911683714e-06, + "loss": 0.2596, + "step": 13756 + }, + { + "epoch": 0.83, + "learning_rate": 1.4595210908291934e-06, + "loss": 0.2747, + "step": 13757 + }, + { + "epoch": 0.83, + "learning_rate": 1.4585034174595203e-06, + "loss": 0.2713, + "step": 13758 + }, + { + "epoch": 0.83, + "learning_rate": 1.457486071098313e-06, + "loss": 0.2522, + "step": 13759 + }, + { + "epoch": 0.83, + "learning_rate": 1.4564690517845181e-06, + "loss": 0.2698, + "step": 13760 + }, + { + "epoch": 0.83, + "learning_rate": 1.4554523595570768e-06, + "loss": 0.2794, + "step": 13761 + }, + { + "epoch": 0.83, + "learning_rate": 1.4544359944549115e-06, + "loss": 0.2529, + "step": 13762 + }, + { + "epoch": 0.83, + "learning_rate": 1.4534199565169305e-06, + "loss": 0.2708, + "step": 13763 + }, + { + "epoch": 0.83, + "learning_rate": 1.4524042457820353e-06, + "loss": 0.2547, + "step": 13764 + }, + { + "epoch": 0.83, + "learning_rate": 1.451388862289108e-06, + "loss": 0.2783, + "step": 13765 + }, + { + "epoch": 0.83, + "learning_rate": 1.450373806077029e-06, + "loss": 0.2759, + "step": 13766 + }, + { + "epoch": 0.83, + "learning_rate": 1.4493590771846544e-06, + "loss": 0.2871, + "step": 13767 + }, + { + "epoch": 0.83, + "learning_rate": 1.4483446756508346e-06, + "loss": 0.2789, + "step": 13768 + }, + { + "epoch": 0.83, + "learning_rate": 1.447330601514405e-06, + "loss": 0.2659, + "step": 13769 + }, + { + "epoch": 0.83, + "learning_rate": 1.446316854814186e-06, + "loss": 0.2489, + "step": 13770 + }, + { + "epoch": 0.83, + "learning_rate": 1.445303435588995e-06, + "loss": 0.2768, + "step": 13771 + }, + { + "epoch": 0.83, + "learning_rate": 1.4442903438776279e-06, + "loss": 0.2667, + "step": 13772 + }, + { + "epoch": 0.83, + "learning_rate": 1.4432775797188691e-06, + "loss": 0.2866, + "step": 13773 + }, + { + "epoch": 0.83, + "learning_rate": 1.4422651431514923e-06, + "loss": 0.252, + "step": 13774 + }, + { + "epoch": 0.83, + "learning_rate": 1.4412530342142595e-06, + "loss": 0.279, + "step": 13775 + }, + { + "epoch": 0.83, + "learning_rate": 1.4402412529459153e-06, + "loss": 0.2591, + "step": 13776 + }, + { + "epoch": 0.83, + "learning_rate": 1.439229799385201e-06, + "loss": 0.2733, + "step": 13777 + }, + { + "epoch": 0.83, + "learning_rate": 1.438218673570837e-06, + "loss": 0.2427, + "step": 13778 + }, + { + "epoch": 0.83, + "learning_rate": 1.4372078755415343e-06, + "loss": 0.3178, + "step": 13779 + }, + { + "epoch": 0.83, + "learning_rate": 1.4361974053359906e-06, + "loss": 0.2886, + "step": 13780 + }, + { + "epoch": 0.83, + "learning_rate": 1.435187262992891e-06, + "loss": 0.2778, + "step": 13781 + }, + { + "epoch": 0.83, + "learning_rate": 1.4341774485509097e-06, + "loss": 0.2872, + "step": 13782 + }, + { + "epoch": 0.83, + "learning_rate": 1.433167962048706e-06, + "loss": 0.2789, + "step": 13783 + }, + { + "epoch": 0.83, + "learning_rate": 1.432158803524929e-06, + "loss": 0.2734, + "step": 13784 + }, + { + "epoch": 0.83, + "learning_rate": 1.43114997301821e-06, + "loss": 0.2776, + "step": 13785 + }, + { + "epoch": 0.83, + "learning_rate": 1.4301414705671779e-06, + "loss": 0.2468, + "step": 13786 + }, + { + "epoch": 0.83, + "learning_rate": 1.42913329621044e-06, + "loss": 0.2843, + "step": 13787 + }, + { + "epoch": 0.83, + "learning_rate": 1.4281254499865927e-06, + "loss": 0.2659, + "step": 13788 + }, + { + "epoch": 0.83, + "learning_rate": 1.4271179319342231e-06, + "loss": 0.275, + "step": 13789 + }, + { + "epoch": 0.83, + "learning_rate": 1.4261107420919018e-06, + "loss": 0.2575, + "step": 13790 + }, + { + "epoch": 0.83, + "learning_rate": 1.425103880498191e-06, + "loss": 0.2649, + "step": 13791 + }, + { + "epoch": 0.83, + "learning_rate": 1.4240973471916364e-06, + "loss": 0.2785, + "step": 13792 + }, + { + "epoch": 0.83, + "learning_rate": 1.423091142210774e-06, + "loss": 0.2616, + "step": 13793 + }, + { + "epoch": 0.83, + "learning_rate": 1.4220852655941253e-06, + "loss": 0.2589, + "step": 13794 + }, + { + "epoch": 0.83, + "learning_rate": 1.4210797173801971e-06, + "loss": 0.2578, + "step": 13795 + }, + { + "epoch": 0.83, + "learning_rate": 1.4200744976074932e-06, + "loss": 0.2753, + "step": 13796 + }, + { + "epoch": 0.83, + "learning_rate": 1.4190696063144927e-06, + "loss": 0.2493, + "step": 13797 + }, + { + "epoch": 0.83, + "learning_rate": 1.4180650435396703e-06, + "loss": 0.2838, + "step": 13798 + }, + { + "epoch": 0.83, + "learning_rate": 1.4170608093214833e-06, + "loss": 0.2539, + "step": 13799 + }, + { + "epoch": 0.83, + "learning_rate": 1.4160569036983784e-06, + "loss": 0.2686, + "step": 13800 + }, + { + "epoch": 0.83, + "learning_rate": 1.4150533267087918e-06, + "loss": 0.2729, + "step": 13801 + }, + { + "epoch": 0.83, + "learning_rate": 1.4140500783911449e-06, + "loss": 0.2814, + "step": 13802 + }, + { + "epoch": 0.83, + "learning_rate": 1.4130471587838456e-06, + "loss": 0.261, + "step": 13803 + }, + { + "epoch": 0.83, + "learning_rate": 1.4120445679252903e-06, + "loss": 0.2758, + "step": 13804 + }, + { + "epoch": 0.83, + "learning_rate": 1.4110423058538625e-06, + "loss": 0.3105, + "step": 13805 + }, + { + "epoch": 0.83, + "learning_rate": 1.4100403726079348e-06, + "loss": 0.2686, + "step": 13806 + }, + { + "epoch": 0.83, + "learning_rate": 1.4090387682258666e-06, + "loss": 0.2794, + "step": 13807 + }, + { + "epoch": 0.83, + "learning_rate": 1.4080374927460028e-06, + "loss": 0.2766, + "step": 13808 + }, + { + "epoch": 0.83, + "learning_rate": 1.4070365462066759e-06, + "loss": 0.2927, + "step": 13809 + }, + { + "epoch": 0.83, + "learning_rate": 1.4060359286462066e-06, + "loss": 0.2443, + "step": 13810 + }, + { + "epoch": 0.83, + "learning_rate": 1.4050356401029065e-06, + "loss": 0.2768, + "step": 13811 + }, + { + "epoch": 0.83, + "learning_rate": 1.404035680615069e-06, + "loss": 0.2546, + "step": 13812 + }, + { + "epoch": 0.83, + "learning_rate": 1.4030360502209772e-06, + "loss": 0.2824, + "step": 13813 + }, + { + "epoch": 0.83, + "learning_rate": 1.4020367489589027e-06, + "loss": 0.2718, + "step": 13814 + }, + { + "epoch": 0.83, + "learning_rate": 1.4010377768671024e-06, + "loss": 0.2753, + "step": 13815 + }, + { + "epoch": 0.83, + "learning_rate": 1.4000391339838215e-06, + "loss": 0.2942, + "step": 13816 + }, + { + "epoch": 0.83, + "learning_rate": 1.3990408203472938e-06, + "loss": 0.2796, + "step": 13817 + }, + { + "epoch": 0.83, + "learning_rate": 1.3980428359957387e-06, + "loss": 0.2694, + "step": 13818 + }, + { + "epoch": 0.83, + "learning_rate": 1.3970451809673624e-06, + "loss": 0.2824, + "step": 13819 + }, + { + "epoch": 0.83, + "learning_rate": 1.396047855300361e-06, + "loss": 0.2642, + "step": 13820 + }, + { + "epoch": 0.83, + "learning_rate": 1.3950508590329171e-06, + "loss": 0.2538, + "step": 13821 + }, + { + "epoch": 0.84, + "learning_rate": 1.3940541922032013e-06, + "loss": 0.268, + "step": 13822 + }, + { + "epoch": 0.84, + "learning_rate": 1.39305785484937e-06, + "loss": 0.2741, + "step": 13823 + }, + { + "epoch": 0.84, + "learning_rate": 1.392061847009567e-06, + "loss": 0.258, + "step": 13824 + }, + { + "epoch": 0.84, + "learning_rate": 1.3910661687219217e-06, + "loss": 0.2807, + "step": 13825 + }, + { + "epoch": 0.84, + "learning_rate": 1.3900708200245583e-06, + "loss": 0.2916, + "step": 13826 + }, + { + "epoch": 0.84, + "learning_rate": 1.3890758009555815e-06, + "loss": 0.2681, + "step": 13827 + }, + { + "epoch": 0.84, + "learning_rate": 1.3880811115530846e-06, + "loss": 0.2659, + "step": 13828 + }, + { + "epoch": 0.84, + "learning_rate": 1.387086751855149e-06, + "loss": 0.2557, + "step": 13829 + }, + { + "epoch": 0.84, + "learning_rate": 1.3860927218998422e-06, + "loss": 0.2704, + "step": 13830 + }, + { + "epoch": 0.84, + "learning_rate": 1.3850990217252235e-06, + "loss": 0.2909, + "step": 13831 + }, + { + "epoch": 0.84, + "learning_rate": 1.3841056513693352e-06, + "loss": 0.2503, + "step": 13832 + }, + { + "epoch": 0.84, + "learning_rate": 1.3831126108702075e-06, + "loss": 0.2493, + "step": 13833 + }, + { + "epoch": 0.84, + "learning_rate": 1.382119900265858e-06, + "loss": 0.2586, + "step": 13834 + }, + { + "epoch": 0.84, + "learning_rate": 1.3811275195942918e-06, + "loss": 0.2744, + "step": 13835 + }, + { + "epoch": 0.84, + "learning_rate": 1.3801354688935055e-06, + "loss": 0.2797, + "step": 13836 + }, + { + "epoch": 0.84, + "learning_rate": 1.3791437482014769e-06, + "loss": 0.2693, + "step": 13837 + }, + { + "epoch": 0.84, + "learning_rate": 1.378152357556175e-06, + "loss": 0.245, + "step": 13838 + }, + { + "epoch": 0.84, + "learning_rate": 1.3771612969955529e-06, + "loss": 0.273, + "step": 13839 + }, + { + "epoch": 0.84, + "learning_rate": 1.376170566557552e-06, + "loss": 0.2907, + "step": 13840 + }, + { + "epoch": 0.84, + "learning_rate": 1.3751801662801056e-06, + "loss": 0.2836, + "step": 13841 + }, + { + "epoch": 0.84, + "learning_rate": 1.37419009620113e-06, + "loss": 0.2796, + "step": 13842 + }, + { + "epoch": 0.84, + "learning_rate": 1.3732003563585294e-06, + "loss": 0.2567, + "step": 13843 + }, + { + "epoch": 0.84, + "learning_rate": 1.3722109467901946e-06, + "loss": 0.262, + "step": 13844 + }, + { + "epoch": 0.84, + "learning_rate": 1.3712218675340038e-06, + "loss": 0.2656, + "step": 13845 + }, + { + "epoch": 0.84, + "learning_rate": 1.3702331186278273e-06, + "loss": 0.2635, + "step": 13846 + }, + { + "epoch": 0.84, + "learning_rate": 1.3692447001095166e-06, + "loss": 0.2662, + "step": 13847 + }, + { + "epoch": 0.84, + "learning_rate": 1.3682566120169149e-06, + "loss": 0.2559, + "step": 13848 + }, + { + "epoch": 0.84, + "learning_rate": 1.36726885438785e-06, + "loss": 0.2524, + "step": 13849 + }, + { + "epoch": 0.84, + "learning_rate": 1.3662814272601344e-06, + "loss": 0.2676, + "step": 13850 + }, + { + "epoch": 0.84, + "learning_rate": 1.365294330671575e-06, + "loss": 0.264, + "step": 13851 + }, + { + "epoch": 0.84, + "learning_rate": 1.3643075646599625e-06, + "loss": 0.2676, + "step": 13852 + }, + { + "epoch": 0.84, + "learning_rate": 1.3633211292630743e-06, + "loss": 0.2754, + "step": 13853 + }, + { + "epoch": 0.84, + "learning_rate": 1.3623350245186762e-06, + "loss": 0.2669, + "step": 13854 + }, + { + "epoch": 0.84, + "learning_rate": 1.3613492504645177e-06, + "loss": 0.2847, + "step": 13855 + }, + { + "epoch": 0.84, + "learning_rate": 1.360363807138344e-06, + "loss": 0.2792, + "step": 13856 + }, + { + "epoch": 0.84, + "learning_rate": 1.3593786945778798e-06, + "loss": 0.2812, + "step": 13857 + }, + { + "epoch": 0.84, + "learning_rate": 1.358393912820839e-06, + "loss": 0.2699, + "step": 13858 + }, + { + "epoch": 0.84, + "learning_rate": 1.357409461904926e-06, + "loss": 0.2543, + "step": 13859 + }, + { + "epoch": 0.84, + "learning_rate": 1.3564253418678263e-06, + "loss": 0.275, + "step": 13860 + }, + { + "epoch": 0.84, + "learning_rate": 1.355441552747222e-06, + "loss": 0.2739, + "step": 13861 + }, + { + "epoch": 0.84, + "learning_rate": 1.354458094580774e-06, + "loss": 0.2719, + "step": 13862 + }, + { + "epoch": 0.84, + "learning_rate": 1.3534749674061332e-06, + "loss": 0.2581, + "step": 13863 + }, + { + "epoch": 0.84, + "learning_rate": 1.3524921712609396e-06, + "loss": 0.2593, + "step": 13864 + }, + { + "epoch": 0.84, + "learning_rate": 1.3515097061828163e-06, + "loss": 0.2494, + "step": 13865 + }, + { + "epoch": 0.84, + "learning_rate": 1.3505275722093824e-06, + "loss": 0.2658, + "step": 13866 + }, + { + "epoch": 0.84, + "learning_rate": 1.3495457693782344e-06, + "loss": 0.2744, + "step": 13867 + }, + { + "epoch": 0.84, + "learning_rate": 1.3485642977269619e-06, + "loss": 0.2786, + "step": 13868 + }, + { + "epoch": 0.84, + "learning_rate": 1.3475831572931397e-06, + "loss": 0.256, + "step": 13869 + }, + { + "epoch": 0.84, + "learning_rate": 1.3466023481143276e-06, + "loss": 0.2818, + "step": 13870 + }, + { + "epoch": 0.84, + "learning_rate": 1.34562187022808e-06, + "loss": 0.2709, + "step": 13871 + }, + { + "epoch": 0.84, + "learning_rate": 1.3446417236719333e-06, + "loss": 0.2939, + "step": 13872 + }, + { + "epoch": 0.84, + "learning_rate": 1.3436619084834102e-06, + "loss": 0.257, + "step": 13873 + }, + { + "epoch": 0.84, + "learning_rate": 1.3426824247000247e-06, + "loss": 0.2765, + "step": 13874 + }, + { + "epoch": 0.84, + "learning_rate": 1.3417032723592726e-06, + "loss": 0.249, + "step": 13875 + }, + { + "epoch": 0.84, + "learning_rate": 1.3407244514986451e-06, + "loss": 0.2603, + "step": 13876 + }, + { + "epoch": 0.84, + "learning_rate": 1.339745962155613e-06, + "loss": 0.2893, + "step": 13877 + }, + { + "epoch": 0.84, + "learning_rate": 1.3387678043676389e-06, + "loss": 0.2508, + "step": 13878 + }, + { + "epoch": 0.84, + "learning_rate": 1.3377899781721704e-06, + "loss": 0.2741, + "step": 13879 + }, + { + "epoch": 0.84, + "learning_rate": 1.3368124836066422e-06, + "loss": 0.2754, + "step": 13880 + }, + { + "epoch": 0.84, + "learning_rate": 1.3358353207084795e-06, + "loss": 0.3045, + "step": 13881 + }, + { + "epoch": 0.84, + "learning_rate": 1.3348584895150917e-06, + "loss": 0.2861, + "step": 13882 + }, + { + "epoch": 0.84, + "learning_rate": 1.33388199006388e-06, + "loss": 0.293, + "step": 13883 + }, + { + "epoch": 0.84, + "learning_rate": 1.3329058223922209e-06, + "loss": 0.2688, + "step": 13884 + }, + { + "epoch": 0.84, + "learning_rate": 1.3319299865374935e-06, + "loss": 0.2564, + "step": 13885 + }, + { + "epoch": 0.84, + "learning_rate": 1.3309544825370556e-06, + "loss": 0.2517, + "step": 13886 + }, + { + "epoch": 0.84, + "learning_rate": 1.3299793104282544e-06, + "loss": 0.2694, + "step": 13887 + }, + { + "epoch": 0.84, + "learning_rate": 1.329004470248424e-06, + "loss": 0.2586, + "step": 13888 + }, + { + "epoch": 0.84, + "learning_rate": 1.3280299620348847e-06, + "loss": 0.2699, + "step": 13889 + }, + { + "epoch": 0.84, + "learning_rate": 1.3270557858249444e-06, + "loss": 0.248, + "step": 13890 + }, + { + "epoch": 0.84, + "learning_rate": 1.3260819416559022e-06, + "loss": 0.2564, + "step": 13891 + }, + { + "epoch": 0.84, + "learning_rate": 1.3251084295650406e-06, + "loss": 0.2708, + "step": 13892 + }, + { + "epoch": 0.84, + "learning_rate": 1.3241352495896298e-06, + "loss": 0.2586, + "step": 13893 + }, + { + "epoch": 0.84, + "learning_rate": 1.3231624017669265e-06, + "loss": 0.2799, + "step": 13894 + }, + { + "epoch": 0.84, + "learning_rate": 1.3221898861341753e-06, + "loss": 0.2803, + "step": 13895 + }, + { + "epoch": 0.84, + "learning_rate": 1.3212177027286122e-06, + "loss": 0.2652, + "step": 13896 + }, + { + "epoch": 0.84, + "learning_rate": 1.320245851587455e-06, + "loss": 0.28, + "step": 13897 + }, + { + "epoch": 0.84, + "learning_rate": 1.3192743327479107e-06, + "loss": 0.2769, + "step": 13898 + }, + { + "epoch": 0.84, + "learning_rate": 1.318303146247174e-06, + "loss": 0.2559, + "step": 13899 + }, + { + "epoch": 0.84, + "learning_rate": 1.317332292122423e-06, + "loss": 0.2633, + "step": 13900 + }, + { + "epoch": 0.84, + "learning_rate": 1.3163617704108322e-06, + "loss": 0.2937, + "step": 13901 + }, + { + "epoch": 0.84, + "learning_rate": 1.3153915811495542e-06, + "loss": 0.2584, + "step": 13902 + }, + { + "epoch": 0.84, + "learning_rate": 1.314421724375734e-06, + "loss": 0.2628, + "step": 13903 + }, + { + "epoch": 0.84, + "learning_rate": 1.3134522001265014e-06, + "loss": 0.2861, + "step": 13904 + }, + { + "epoch": 0.84, + "learning_rate": 1.3124830084389718e-06, + "loss": 0.257, + "step": 13905 + }, + { + "epoch": 0.84, + "learning_rate": 1.311514149350256e-06, + "loss": 0.2722, + "step": 13906 + }, + { + "epoch": 0.84, + "learning_rate": 1.3105456228974423e-06, + "loss": 0.2626, + "step": 13907 + }, + { + "epoch": 0.84, + "learning_rate": 1.3095774291176121e-06, + "loss": 0.2766, + "step": 13908 + }, + { + "epoch": 0.84, + "learning_rate": 1.3086095680478328e-06, + "loss": 0.2722, + "step": 13909 + }, + { + "epoch": 0.84, + "learning_rate": 1.3076420397251543e-06, + "loss": 0.2797, + "step": 13910 + }, + { + "epoch": 0.84, + "learning_rate": 1.306674844186624e-06, + "loss": 0.2864, + "step": 13911 + }, + { + "epoch": 0.84, + "learning_rate": 1.3057079814692686e-06, + "loss": 0.2487, + "step": 13912 + }, + { + "epoch": 0.84, + "learning_rate": 1.3047414516101031e-06, + "loss": 0.2464, + "step": 13913 + }, + { + "epoch": 0.84, + "learning_rate": 1.3037752546461313e-06, + "loss": 0.2423, + "step": 13914 + }, + { + "epoch": 0.84, + "learning_rate": 1.3028093906143424e-06, + "loss": 0.2732, + "step": 13915 + }, + { + "epoch": 0.84, + "learning_rate": 1.301843859551718e-06, + "loss": 0.2567, + "step": 13916 + }, + { + "epoch": 0.84, + "learning_rate": 1.3008786614952217e-06, + "loss": 0.2578, + "step": 13917 + }, + { + "epoch": 0.84, + "learning_rate": 1.2999137964818043e-06, + "loss": 0.2634, + "step": 13918 + }, + { + "epoch": 0.84, + "learning_rate": 1.2989492645484025e-06, + "loss": 0.2877, + "step": 13919 + }, + { + "epoch": 0.84, + "learning_rate": 1.2979850657319494e-06, + "loss": 0.2616, + "step": 13920 + }, + { + "epoch": 0.84, + "learning_rate": 1.297021200069356e-06, + "loss": 0.2567, + "step": 13921 + }, + { + "epoch": 0.84, + "learning_rate": 1.2960576675975244e-06, + "loss": 0.2547, + "step": 13922 + }, + { + "epoch": 0.84, + "learning_rate": 1.2950944683533416e-06, + "loss": 0.2778, + "step": 13923 + }, + { + "epoch": 0.84, + "learning_rate": 1.294131602373685e-06, + "loss": 0.2807, + "step": 13924 + }, + { + "epoch": 0.84, + "learning_rate": 1.2931690696954135e-06, + "loss": 0.2709, + "step": 13925 + }, + { + "epoch": 0.84, + "learning_rate": 1.2922068703553848e-06, + "loss": 0.2655, + "step": 13926 + }, + { + "epoch": 0.84, + "learning_rate": 1.2912450043904312e-06, + "loss": 0.242, + "step": 13927 + }, + { + "epoch": 0.84, + "learning_rate": 1.290283471837378e-06, + "loss": 0.2518, + "step": 13928 + }, + { + "epoch": 0.84, + "learning_rate": 1.289322272733039e-06, + "loss": 0.2765, + "step": 13929 + }, + { + "epoch": 0.84, + "learning_rate": 1.2883614071142092e-06, + "loss": 0.2287, + "step": 13930 + }, + { + "epoch": 0.84, + "learning_rate": 1.2874008750176804e-06, + "loss": 0.2754, + "step": 13931 + }, + { + "epoch": 0.84, + "learning_rate": 1.2864406764802239e-06, + "loss": 0.2713, + "step": 13932 + }, + { + "epoch": 0.84, + "learning_rate": 1.285480811538601e-06, + "loss": 0.2619, + "step": 13933 + }, + { + "epoch": 0.84, + "learning_rate": 1.284521280229558e-06, + "loss": 0.2501, + "step": 13934 + }, + { + "epoch": 0.84, + "learning_rate": 1.283562082589831e-06, + "loss": 0.2722, + "step": 13935 + }, + { + "epoch": 0.84, + "learning_rate": 1.2826032186561444e-06, + "loss": 0.2627, + "step": 13936 + }, + { + "epoch": 0.84, + "learning_rate": 1.2816446884652066e-06, + "loss": 0.2602, + "step": 13937 + }, + { + "epoch": 0.84, + "learning_rate": 1.2806864920537154e-06, + "loss": 0.2538, + "step": 13938 + }, + { + "epoch": 0.84, + "learning_rate": 1.2797286294583545e-06, + "loss": 0.2701, + "step": 13939 + }, + { + "epoch": 0.84, + "learning_rate": 1.2787711007157932e-06, + "loss": 0.2646, + "step": 13940 + }, + { + "epoch": 0.84, + "learning_rate": 1.2778139058626948e-06, + "loss": 0.2455, + "step": 13941 + }, + { + "epoch": 0.84, + "learning_rate": 1.276857044935702e-06, + "loss": 0.2684, + "step": 13942 + }, + { + "epoch": 0.84, + "learning_rate": 1.275900517971448e-06, + "loss": 0.2624, + "step": 13943 + }, + { + "epoch": 0.84, + "learning_rate": 1.2749443250065552e-06, + "loss": 0.2746, + "step": 13944 + }, + { + "epoch": 0.84, + "learning_rate": 1.2739884660776268e-06, + "loss": 0.238, + "step": 13945 + }, + { + "epoch": 0.84, + "learning_rate": 1.2730329412212627e-06, + "loss": 0.2677, + "step": 13946 + }, + { + "epoch": 0.84, + "learning_rate": 1.2720777504740422e-06, + "loss": 0.2808, + "step": 13947 + }, + { + "epoch": 0.84, + "learning_rate": 1.2711228938725351e-06, + "loss": 0.2637, + "step": 13948 + }, + { + "epoch": 0.84, + "learning_rate": 1.2701683714532976e-06, + "loss": 0.254, + "step": 13949 + }, + { + "epoch": 0.84, + "learning_rate": 1.2692141832528715e-06, + "loss": 0.2709, + "step": 13950 + }, + { + "epoch": 0.84, + "learning_rate": 1.2682603293077933e-06, + "loss": 0.2988, + "step": 13951 + }, + { + "epoch": 0.84, + "learning_rate": 1.267306809654575e-06, + "loss": 0.2688, + "step": 13952 + }, + { + "epoch": 0.84, + "learning_rate": 1.2663536243297236e-06, + "loss": 0.2761, + "step": 13953 + }, + { + "epoch": 0.84, + "learning_rate": 1.2654007733697294e-06, + "loss": 0.2607, + "step": 13954 + }, + { + "epoch": 0.84, + "learning_rate": 1.264448256811076e-06, + "loss": 0.2651, + "step": 13955 + }, + { + "epoch": 0.84, + "learning_rate": 1.2634960746902302e-06, + "loss": 0.2585, + "step": 13956 + }, + { + "epoch": 0.84, + "learning_rate": 1.2625442270436427e-06, + "loss": 0.2617, + "step": 13957 + }, + { + "epoch": 0.84, + "learning_rate": 1.2615927139077566e-06, + "loss": 0.2486, + "step": 13958 + }, + { + "epoch": 0.84, + "learning_rate": 1.2606415353190005e-06, + "loss": 0.2522, + "step": 13959 + }, + { + "epoch": 0.84, + "learning_rate": 1.2596906913137874e-06, + "loss": 0.2825, + "step": 13960 + }, + { + "epoch": 0.84, + "learning_rate": 1.258740181928524e-06, + "loss": 0.2847, + "step": 13961 + }, + { + "epoch": 0.84, + "learning_rate": 1.2577900071995975e-06, + "loss": 0.2903, + "step": 13962 + }, + { + "epoch": 0.84, + "learning_rate": 1.2568401671633868e-06, + "loss": 0.2754, + "step": 13963 + }, + { + "epoch": 0.84, + "learning_rate": 1.255890661856255e-06, + "loss": 0.2694, + "step": 13964 + }, + { + "epoch": 0.84, + "learning_rate": 1.2549414913145509e-06, + "loss": 0.2843, + "step": 13965 + }, + { + "epoch": 0.84, + "learning_rate": 1.2539926555746197e-06, + "loss": 0.2658, + "step": 13966 + }, + { + "epoch": 0.84, + "learning_rate": 1.2530441546727834e-06, + "loss": 0.2944, + "step": 13967 + }, + { + "epoch": 0.84, + "learning_rate": 1.2520959886453543e-06, + "loss": 0.2836, + "step": 13968 + }, + { + "epoch": 0.84, + "learning_rate": 1.2511481575286354e-06, + "loss": 0.2662, + "step": 13969 + }, + { + "epoch": 0.84, + "learning_rate": 1.2502006613589091e-06, + "loss": 0.2772, + "step": 13970 + }, + { + "epoch": 0.84, + "learning_rate": 1.249253500172456e-06, + "loss": 0.2701, + "step": 13971 + }, + { + "epoch": 0.84, + "learning_rate": 1.248306674005535e-06, + "loss": 0.2923, + "step": 13972 + }, + { + "epoch": 0.84, + "learning_rate": 1.2473601828943949e-06, + "loss": 0.2705, + "step": 13973 + }, + { + "epoch": 0.84, + "learning_rate": 1.2464140268752733e-06, + "loss": 0.2688, + "step": 13974 + }, + { + "epoch": 0.84, + "learning_rate": 1.2454682059843892e-06, + "loss": 0.2324, + "step": 13975 + }, + { + "epoch": 0.84, + "learning_rate": 1.244522720257959e-06, + "loss": 0.2801, + "step": 13976 + }, + { + "epoch": 0.84, + "learning_rate": 1.2435775697321762e-06, + "loss": 0.2603, + "step": 13977 + }, + { + "epoch": 0.84, + "learning_rate": 1.2426327544432282e-06, + "loss": 0.2645, + "step": 13978 + }, + { + "epoch": 0.84, + "learning_rate": 1.241688274427285e-06, + "loss": 0.2504, + "step": 13979 + }, + { + "epoch": 0.84, + "learning_rate": 1.2407441297205048e-06, + "loss": 0.2451, + "step": 13980 + }, + { + "epoch": 0.84, + "learning_rate": 1.239800320359037e-06, + "loss": 0.2605, + "step": 13981 + }, + { + "epoch": 0.84, + "learning_rate": 1.238856846379014e-06, + "loss": 0.2893, + "step": 13982 + }, + { + "epoch": 0.84, + "learning_rate": 1.237913707816555e-06, + "loss": 0.2773, + "step": 13983 + }, + { + "epoch": 0.84, + "learning_rate": 1.2369709047077694e-06, + "loss": 0.2447, + "step": 13984 + }, + { + "epoch": 0.84, + "learning_rate": 1.2360284370887509e-06, + "loss": 0.2638, + "step": 13985 + }, + { + "epoch": 0.84, + "learning_rate": 1.235086304995582e-06, + "loss": 0.2681, + "step": 13986 + }, + { + "epoch": 0.84, + "learning_rate": 1.2341445084643323e-06, + "loss": 0.246, + "step": 13987 + }, + { + "epoch": 0.85, + "learning_rate": 1.2332030475310575e-06, + "loss": 0.2935, + "step": 13988 + }, + { + "epoch": 0.85, + "learning_rate": 1.2322619222317999e-06, + "loss": 0.2736, + "step": 13989 + }, + { + "epoch": 0.85, + "learning_rate": 1.2313211326025932e-06, + "loss": 0.2617, + "step": 13990 + }, + { + "epoch": 0.85, + "learning_rate": 1.2303806786794537e-06, + "loss": 0.2694, + "step": 13991 + }, + { + "epoch": 0.85, + "learning_rate": 1.2294405604983872e-06, + "loss": 0.2793, + "step": 13992 + }, + { + "epoch": 0.85, + "learning_rate": 1.228500778095384e-06, + "loss": 0.2658, + "step": 13993 + }, + { + "epoch": 0.85, + "learning_rate": 1.227561331506425e-06, + "loss": 0.2895, + "step": 13994 + }, + { + "epoch": 0.85, + "learning_rate": 1.2266222207674739e-06, + "loss": 0.2554, + "step": 13995 + }, + { + "epoch": 0.85, + "learning_rate": 1.2256834459144885e-06, + "loss": 0.2637, + "step": 13996 + }, + { + "epoch": 0.85, + "learning_rate": 1.2247450069834077e-06, + "loss": 0.2796, + "step": 13997 + }, + { + "epoch": 0.85, + "learning_rate": 1.2238069040101586e-06, + "loss": 0.2672, + "step": 13998 + }, + { + "epoch": 0.85, + "learning_rate": 1.2228691370306566e-06, + "loss": 0.2955, + "step": 13999 + }, + { + "epoch": 0.85, + "learning_rate": 1.2219317060808022e-06, + "loss": 0.2547, + "step": 14000 + }, + { + "epoch": 0.85, + "learning_rate": 1.2209946111964876e-06, + "loss": 0.2737, + "step": 14001 + }, + { + "epoch": 0.85, + "learning_rate": 1.2200578524135886e-06, + "loss": 0.2806, + "step": 14002 + }, + { + "epoch": 0.85, + "learning_rate": 1.2191214297679677e-06, + "loss": 0.2819, + "step": 14003 + }, + { + "epoch": 0.85, + "learning_rate": 1.2181853432954749e-06, + "loss": 0.2627, + "step": 14004 + }, + { + "epoch": 0.85, + "learning_rate": 1.2172495930319484e-06, + "loss": 0.2938, + "step": 14005 + }, + { + "epoch": 0.85, + "learning_rate": 1.216314179013215e-06, + "loss": 0.2588, + "step": 14006 + }, + { + "epoch": 0.85, + "learning_rate": 1.215379101275085e-06, + "loss": 0.2547, + "step": 14007 + }, + { + "epoch": 0.85, + "learning_rate": 1.2144443598533574e-06, + "loss": 0.2845, + "step": 14008 + }, + { + "epoch": 0.85, + "learning_rate": 1.213509954783819e-06, + "loss": 0.2847, + "step": 14009 + }, + { + "epoch": 0.85, + "learning_rate": 1.2125758861022418e-06, + "loss": 0.2529, + "step": 14010 + }, + { + "epoch": 0.85, + "learning_rate": 1.211642153844389e-06, + "loss": 0.26, + "step": 14011 + }, + { + "epoch": 0.85, + "learning_rate": 1.2107087580460076e-06, + "loss": 0.2586, + "step": 14012 + }, + { + "epoch": 0.85, + "learning_rate": 1.2097756987428311e-06, + "loss": 0.3036, + "step": 14013 + }, + { + "epoch": 0.85, + "learning_rate": 1.2088429759705821e-06, + "loss": 0.2709, + "step": 14014 + }, + { + "epoch": 0.85, + "learning_rate": 1.2079105897649668e-06, + "loss": 0.2723, + "step": 14015 + }, + { + "epoch": 0.85, + "learning_rate": 1.2069785401616873e-06, + "loss": 0.2575, + "step": 14016 + }, + { + "epoch": 0.85, + "learning_rate": 1.206046827196422e-06, + "loss": 0.2613, + "step": 14017 + }, + { + "epoch": 0.85, + "learning_rate": 1.2051154509048434e-06, + "loss": 0.2702, + "step": 14018 + }, + { + "epoch": 0.85, + "learning_rate": 1.2041844113226075e-06, + "loss": 0.2656, + "step": 14019 + }, + { + "epoch": 0.85, + "learning_rate": 1.2032537084853602e-06, + "loss": 0.2832, + "step": 14020 + }, + { + "epoch": 0.85, + "learning_rate": 1.2023233424287328e-06, + "loss": 0.2574, + "step": 14021 + }, + { + "epoch": 0.85, + "learning_rate": 1.2013933131883437e-06, + "loss": 0.292, + "step": 14022 + }, + { + "epoch": 0.85, + "learning_rate": 1.2004636207997987e-06, + "loss": 0.2609, + "step": 14023 + }, + { + "epoch": 0.85, + "learning_rate": 1.1995342652986886e-06, + "loss": 0.2585, + "step": 14024 + }, + { + "epoch": 0.85, + "learning_rate": 1.1986052467205978e-06, + "loss": 0.2681, + "step": 14025 + }, + { + "epoch": 0.85, + "learning_rate": 1.1976765651010913e-06, + "loss": 0.2648, + "step": 14026 + }, + { + "epoch": 0.85, + "learning_rate": 1.1967482204757241e-06, + "loss": 0.2599, + "step": 14027 + }, + { + "epoch": 0.85, + "learning_rate": 1.195820212880038e-06, + "loss": 0.2822, + "step": 14028 + }, + { + "epoch": 0.85, + "learning_rate": 1.1948925423495583e-06, + "loss": 0.2564, + "step": 14029 + }, + { + "epoch": 0.85, + "learning_rate": 1.1939652089198027e-06, + "loss": 0.2753, + "step": 14030 + }, + { + "epoch": 0.85, + "learning_rate": 1.1930382126262752e-06, + "loss": 0.2573, + "step": 14031 + }, + { + "epoch": 0.85, + "learning_rate": 1.1921115535044637e-06, + "loss": 0.2433, + "step": 14032 + }, + { + "epoch": 0.85, + "learning_rate": 1.1911852315898465e-06, + "loss": 0.273, + "step": 14033 + }, + { + "epoch": 0.85, + "learning_rate": 1.1902592469178864e-06, + "loss": 0.2623, + "step": 14034 + }, + { + "epoch": 0.85, + "learning_rate": 1.189333599524033e-06, + "loss": 0.2729, + "step": 14035 + }, + { + "epoch": 0.85, + "learning_rate": 1.1884082894437278e-06, + "loss": 0.2469, + "step": 14036 + }, + { + "epoch": 0.85, + "learning_rate": 1.1874833167123944e-06, + "loss": 0.246, + "step": 14037 + }, + { + "epoch": 0.85, + "learning_rate": 1.186558681365445e-06, + "loss": 0.2857, + "step": 14038 + }, + { + "epoch": 0.85, + "learning_rate": 1.1856343834382788e-06, + "loss": 0.2638, + "step": 14039 + }, + { + "epoch": 0.85, + "learning_rate": 1.184710422966281e-06, + "loss": 0.2612, + "step": 14040 + }, + { + "epoch": 0.85, + "learning_rate": 1.1837867999848286e-06, + "loss": 0.2596, + "step": 14041 + }, + { + "epoch": 0.85, + "learning_rate": 1.18286351452928e-06, + "loss": 0.2624, + "step": 14042 + }, + { + "epoch": 0.85, + "learning_rate": 1.1819405666349825e-06, + "loss": 0.2574, + "step": 14043 + }, + { + "epoch": 0.85, + "learning_rate": 1.1810179563372726e-06, + "loss": 0.2655, + "step": 14044 + }, + { + "epoch": 0.85, + "learning_rate": 1.1800956836714683e-06, + "loss": 0.2527, + "step": 14045 + }, + { + "epoch": 0.85, + "learning_rate": 1.1791737486728827e-06, + "loss": 0.2709, + "step": 14046 + }, + { + "epoch": 0.85, + "learning_rate": 1.1782521513768108e-06, + "loss": 0.2906, + "step": 14047 + }, + { + "epoch": 0.85, + "learning_rate": 1.1773308918185354e-06, + "loss": 0.2812, + "step": 14048 + }, + { + "epoch": 0.85, + "learning_rate": 1.1764099700333265e-06, + "loss": 0.2628, + "step": 14049 + }, + { + "epoch": 0.85, + "learning_rate": 1.1754893860564387e-06, + "loss": 0.2725, + "step": 14050 + }, + { + "epoch": 0.85, + "learning_rate": 1.1745691399231208e-06, + "loss": 0.2825, + "step": 14051 + }, + { + "epoch": 0.85, + "learning_rate": 1.173649231668601e-06, + "loss": 0.2725, + "step": 14052 + }, + { + "epoch": 0.85, + "learning_rate": 1.1727296613281026e-06, + "loss": 0.2641, + "step": 14053 + }, + { + "epoch": 0.85, + "learning_rate": 1.1718104289368225e-06, + "loss": 0.2603, + "step": 14054 + }, + { + "epoch": 0.85, + "learning_rate": 1.1708915345299598e-06, + "loss": 0.2765, + "step": 14055 + }, + { + "epoch": 0.85, + "learning_rate": 1.1699729781426916e-06, + "loss": 0.2695, + "step": 14056 + }, + { + "epoch": 0.85, + "learning_rate": 1.1690547598101864e-06, + "loss": 0.2693, + "step": 14057 + }, + { + "epoch": 0.85, + "learning_rate": 1.1681368795675963e-06, + "loss": 0.2755, + "step": 14058 + }, + { + "epoch": 0.85, + "learning_rate": 1.1672193374500595e-06, + "loss": 0.2635, + "step": 14059 + }, + { + "epoch": 0.85, + "learning_rate": 1.1663021334927093e-06, + "loss": 0.2751, + "step": 14060 + }, + { + "epoch": 0.85, + "learning_rate": 1.1653852677306577e-06, + "loss": 0.267, + "step": 14061 + }, + { + "epoch": 0.85, + "learning_rate": 1.1644687401990074e-06, + "loss": 0.2545, + "step": 14062 + }, + { + "epoch": 0.85, + "learning_rate": 1.1635525509328471e-06, + "loss": 0.2684, + "step": 14063 + }, + { + "epoch": 0.85, + "learning_rate": 1.1626366999672512e-06, + "loss": 0.254, + "step": 14064 + }, + { + "epoch": 0.85, + "learning_rate": 1.1617211873372836e-06, + "loss": 0.2662, + "step": 14065 + }, + { + "epoch": 0.85, + "learning_rate": 1.1608060130779963e-06, + "loss": 0.2877, + "step": 14066 + }, + { + "epoch": 0.85, + "learning_rate": 1.1598911772244247e-06, + "loss": 0.2743, + "step": 14067 + }, + { + "epoch": 0.85, + "learning_rate": 1.1589766798115931e-06, + "loss": 0.2483, + "step": 14068 + }, + { + "epoch": 0.85, + "learning_rate": 1.1580625208745145e-06, + "loss": 0.2589, + "step": 14069 + }, + { + "epoch": 0.85, + "learning_rate": 1.1571487004481818e-06, + "loss": 0.2811, + "step": 14070 + }, + { + "epoch": 0.85, + "learning_rate": 1.1562352185675874e-06, + "loss": 0.2574, + "step": 14071 + }, + { + "epoch": 0.85, + "learning_rate": 1.1553220752677009e-06, + "loss": 0.2566, + "step": 14072 + }, + { + "epoch": 0.85, + "learning_rate": 1.1544092705834798e-06, + "loss": 0.2693, + "step": 14073 + }, + { + "epoch": 0.85, + "learning_rate": 1.153496804549873e-06, + "loss": 0.2877, + "step": 14074 + }, + { + "epoch": 0.85, + "learning_rate": 1.1525846772018113e-06, + "loss": 0.2755, + "step": 14075 + }, + { + "epoch": 0.85, + "learning_rate": 1.1516728885742179e-06, + "loss": 0.2785, + "step": 14076 + }, + { + "epoch": 0.85, + "learning_rate": 1.1507614387019995e-06, + "loss": 0.2734, + "step": 14077 + }, + { + "epoch": 0.85, + "learning_rate": 1.1498503276200513e-06, + "loss": 0.2533, + "step": 14078 + }, + { + "epoch": 0.85, + "learning_rate": 1.1489395553632532e-06, + "loss": 0.2898, + "step": 14079 + }, + { + "epoch": 0.85, + "learning_rate": 1.148029121966473e-06, + "loss": 0.2779, + "step": 14080 + }, + { + "epoch": 0.85, + "learning_rate": 1.1471190274645706e-06, + "loss": 0.2617, + "step": 14081 + }, + { + "epoch": 0.85, + "learning_rate": 1.1462092718923856e-06, + "loss": 0.2723, + "step": 14082 + }, + { + "epoch": 0.85, + "learning_rate": 1.1452998552847482e-06, + "loss": 0.264, + "step": 14083 + }, + { + "epoch": 0.85, + "learning_rate": 1.1443907776764751e-06, + "loss": 0.2529, + "step": 14084 + }, + { + "epoch": 0.85, + "learning_rate": 1.1434820391023683e-06, + "loss": 0.2815, + "step": 14085 + }, + { + "epoch": 0.85, + "learning_rate": 1.142573639597223e-06, + "loss": 0.2807, + "step": 14086 + }, + { + "epoch": 0.85, + "learning_rate": 1.1416655791958154e-06, + "loss": 0.2879, + "step": 14087 + }, + { + "epoch": 0.85, + "learning_rate": 1.1407578579329082e-06, + "loss": 0.2726, + "step": 14088 + }, + { + "epoch": 0.85, + "learning_rate": 1.1398504758432517e-06, + "loss": 0.2836, + "step": 14089 + }, + { + "epoch": 0.85, + "learning_rate": 1.1389434329615889e-06, + "loss": 0.2932, + "step": 14090 + }, + { + "epoch": 0.85, + "learning_rate": 1.1380367293226447e-06, + "loss": 0.2595, + "step": 14091 + }, + { + "epoch": 0.85, + "learning_rate": 1.137130364961131e-06, + "loss": 0.2542, + "step": 14092 + }, + { + "epoch": 0.85, + "learning_rate": 1.1362243399117479e-06, + "loss": 0.251, + "step": 14093 + }, + { + "epoch": 0.85, + "learning_rate": 1.13531865420918e-06, + "loss": 0.2782, + "step": 14094 + }, + { + "epoch": 0.85, + "learning_rate": 1.134413307888106e-06, + "loss": 0.2846, + "step": 14095 + }, + { + "epoch": 0.85, + "learning_rate": 1.1335083009831849e-06, + "loss": 0.2691, + "step": 14096 + }, + { + "epoch": 0.85, + "learning_rate": 1.1326036335290635e-06, + "loss": 0.2877, + "step": 14097 + }, + { + "epoch": 0.85, + "learning_rate": 1.1316993055603775e-06, + "loss": 0.2659, + "step": 14098 + }, + { + "epoch": 0.85, + "learning_rate": 1.130795317111748e-06, + "loss": 0.2651, + "step": 14099 + }, + { + "epoch": 0.85, + "learning_rate": 1.129891668217783e-06, + "loss": 0.2513, + "step": 14100 + }, + { + "epoch": 0.85, + "learning_rate": 1.1289883589130823e-06, + "loss": 0.2599, + "step": 14101 + }, + { + "epoch": 0.85, + "learning_rate": 1.1280853892322252e-06, + "loss": 0.2722, + "step": 14102 + }, + { + "epoch": 0.85, + "learning_rate": 1.127182759209784e-06, + "loss": 0.2694, + "step": 14103 + }, + { + "epoch": 0.85, + "learning_rate": 1.126280468880313e-06, + "loss": 0.262, + "step": 14104 + }, + { + "epoch": 0.85, + "learning_rate": 1.1253785182783571e-06, + "loss": 0.2607, + "step": 14105 + }, + { + "epoch": 0.85, + "learning_rate": 1.1244769074384488e-06, + "loss": 0.2605, + "step": 14106 + }, + { + "epoch": 0.85, + "learning_rate": 1.1235756363951045e-06, + "loss": 0.2684, + "step": 14107 + }, + { + "epoch": 0.85, + "learning_rate": 1.122674705182829e-06, + "loss": 0.2677, + "step": 14108 + }, + { + "epoch": 0.85, + "learning_rate": 1.121774113836116e-06, + "loss": 0.2418, + "step": 14109 + }, + { + "epoch": 0.85, + "learning_rate": 1.12087386238944e-06, + "loss": 0.2867, + "step": 14110 + }, + { + "epoch": 0.85, + "learning_rate": 1.1199739508772712e-06, + "loss": 0.2567, + "step": 14111 + }, + { + "epoch": 0.85, + "learning_rate": 1.1190743793340608e-06, + "loss": 0.2976, + "step": 14112 + }, + { + "epoch": 0.85, + "learning_rate": 1.1181751477942493e-06, + "loss": 0.2808, + "step": 14113 + }, + { + "epoch": 0.85, + "learning_rate": 1.1172762562922634e-06, + "loss": 0.2538, + "step": 14114 + }, + { + "epoch": 0.85, + "learning_rate": 1.1163777048625136e-06, + "loss": 0.25, + "step": 14115 + }, + { + "epoch": 0.85, + "learning_rate": 1.1154794935394065e-06, + "loss": 0.2418, + "step": 14116 + }, + { + "epoch": 0.85, + "learning_rate": 1.1145816223573259e-06, + "loss": 0.2434, + "step": 14117 + }, + { + "epoch": 0.85, + "learning_rate": 1.1136840913506475e-06, + "loss": 0.2487, + "step": 14118 + }, + { + "epoch": 0.85, + "learning_rate": 1.1127869005537328e-06, + "loss": 0.2543, + "step": 14119 + }, + { + "epoch": 0.85, + "learning_rate": 1.1118900500009277e-06, + "loss": 0.2836, + "step": 14120 + }, + { + "epoch": 0.85, + "learning_rate": 1.110993539726576e-06, + "loss": 0.2515, + "step": 14121 + }, + { + "epoch": 0.85, + "learning_rate": 1.110097369764993e-06, + "loss": 0.2571, + "step": 14122 + }, + { + "epoch": 0.85, + "learning_rate": 1.1092015401504896e-06, + "loss": 0.2638, + "step": 14123 + }, + { + "epoch": 0.85, + "learning_rate": 1.1083060509173616e-06, + "loss": 0.2733, + "step": 14124 + }, + { + "epoch": 0.85, + "learning_rate": 1.1074109020998958e-06, + "loss": 0.2684, + "step": 14125 + }, + { + "epoch": 0.85, + "learning_rate": 1.1065160937323605e-06, + "loss": 0.2677, + "step": 14126 + }, + { + "epoch": 0.85, + "learning_rate": 1.1056216258490127e-06, + "loss": 0.2245, + "step": 14127 + }, + { + "epoch": 0.85, + "learning_rate": 1.1047274984840994e-06, + "loss": 0.2653, + "step": 14128 + }, + { + "epoch": 0.85, + "learning_rate": 1.1038337116718468e-06, + "loss": 0.2726, + "step": 14129 + }, + { + "epoch": 0.85, + "learning_rate": 1.1029402654464794e-06, + "loss": 0.2599, + "step": 14130 + }, + { + "epoch": 0.85, + "learning_rate": 1.102047159842199e-06, + "loss": 0.2494, + "step": 14131 + }, + { + "epoch": 0.85, + "learning_rate": 1.101154394893199e-06, + "loss": 0.2524, + "step": 14132 + }, + { + "epoch": 0.85, + "learning_rate": 1.1002619706336581e-06, + "loss": 0.303, + "step": 14133 + }, + { + "epoch": 0.85, + "learning_rate": 1.099369887097741e-06, + "loss": 0.2804, + "step": 14134 + }, + { + "epoch": 0.85, + "learning_rate": 1.0984781443196036e-06, + "loss": 0.2705, + "step": 14135 + }, + { + "epoch": 0.85, + "learning_rate": 1.0975867423333852e-06, + "loss": 0.2762, + "step": 14136 + }, + { + "epoch": 0.85, + "learning_rate": 1.0966956811732133e-06, + "loss": 0.2856, + "step": 14137 + }, + { + "epoch": 0.85, + "learning_rate": 1.0958049608732003e-06, + "loss": 0.263, + "step": 14138 + }, + { + "epoch": 0.85, + "learning_rate": 1.0949145814674478e-06, + "loss": 0.2637, + "step": 14139 + }, + { + "epoch": 0.85, + "learning_rate": 1.0940245429900408e-06, + "loss": 0.2829, + "step": 14140 + }, + { + "epoch": 0.85, + "learning_rate": 1.0931348454750602e-06, + "loss": 0.2584, + "step": 14141 + }, + { + "epoch": 0.85, + "learning_rate": 1.0922454889565648e-06, + "loss": 0.2592, + "step": 14142 + }, + { + "epoch": 0.85, + "learning_rate": 1.0913564734686032e-06, + "loss": 0.2871, + "step": 14143 + }, + { + "epoch": 0.85, + "learning_rate": 1.0904677990452105e-06, + "loss": 0.2739, + "step": 14144 + }, + { + "epoch": 0.85, + "learning_rate": 1.0895794657204084e-06, + "loss": 0.2755, + "step": 14145 + }, + { + "epoch": 0.85, + "learning_rate": 1.0886914735282105e-06, + "loss": 0.2676, + "step": 14146 + }, + { + "epoch": 0.85, + "learning_rate": 1.08780382250261e-06, + "loss": 0.3025, + "step": 14147 + }, + { + "epoch": 0.85, + "learning_rate": 1.0869165126775905e-06, + "loss": 0.2609, + "step": 14148 + }, + { + "epoch": 0.85, + "learning_rate": 1.0860295440871238e-06, + "loss": 0.2598, + "step": 14149 + }, + { + "epoch": 0.85, + "learning_rate": 1.0851429167651651e-06, + "loss": 0.2845, + "step": 14150 + }, + { + "epoch": 0.85, + "learning_rate": 1.0842566307456604e-06, + "loss": 0.2799, + "step": 14151 + }, + { + "epoch": 0.85, + "learning_rate": 1.0833706860625416e-06, + "loss": 0.2628, + "step": 14152 + }, + { + "epoch": 0.86, + "learning_rate": 1.0824850827497246e-06, + "loss": 0.2554, + "step": 14153 + }, + { + "epoch": 0.86, + "learning_rate": 1.0815998208411161e-06, + "loss": 0.2655, + "step": 14154 + }, + { + "epoch": 0.86, + "learning_rate": 1.0807149003706075e-06, + "loss": 0.2686, + "step": 14155 + }, + { + "epoch": 0.86, + "learning_rate": 1.0798303213720772e-06, + "loss": 0.2646, + "step": 14156 + }, + { + "epoch": 0.86, + "learning_rate": 1.0789460838793919e-06, + "loss": 0.2473, + "step": 14157 + }, + { + "epoch": 0.86, + "learning_rate": 1.078062187926403e-06, + "loss": 0.2787, + "step": 14158 + }, + { + "epoch": 0.86, + "learning_rate": 1.0771786335469503e-06, + "loss": 0.2891, + "step": 14159 + }, + { + "epoch": 0.86, + "learning_rate": 1.0762954207748622e-06, + "loss": 0.2806, + "step": 14160 + }, + { + "epoch": 0.86, + "learning_rate": 1.0754125496439515e-06, + "loss": 0.2764, + "step": 14161 + }, + { + "epoch": 0.86, + "learning_rate": 1.0745300201880181e-06, + "loss": 0.2557, + "step": 14162 + }, + { + "epoch": 0.86, + "learning_rate": 1.0736478324408494e-06, + "loss": 0.2355, + "step": 14163 + }, + { + "epoch": 0.86, + "learning_rate": 1.0727659864362184e-06, + "loss": 0.2699, + "step": 14164 + }, + { + "epoch": 0.86, + "learning_rate": 1.07188448220789e-06, + "loss": 0.2487, + "step": 14165 + }, + { + "epoch": 0.86, + "learning_rate": 1.0710033197896097e-06, + "loss": 0.2755, + "step": 14166 + }, + { + "epoch": 0.86, + "learning_rate": 1.0701224992151126e-06, + "loss": 0.2706, + "step": 14167 + }, + { + "epoch": 0.86, + "learning_rate": 1.0692420205181208e-06, + "loss": 0.2702, + "step": 14168 + }, + { + "epoch": 0.86, + "learning_rate": 1.0683618837323416e-06, + "loss": 0.2652, + "step": 14169 + }, + { + "epoch": 0.86, + "learning_rate": 1.067482088891475e-06, + "loss": 0.2706, + "step": 14170 + }, + { + "epoch": 0.86, + "learning_rate": 1.0666026360292003e-06, + "loss": 0.2674, + "step": 14171 + }, + { + "epoch": 0.86, + "learning_rate": 1.0657235251791876e-06, + "loss": 0.2626, + "step": 14172 + }, + { + "epoch": 0.86, + "learning_rate": 1.064844756375094e-06, + "loss": 0.2814, + "step": 14173 + }, + { + "epoch": 0.86, + "learning_rate": 1.063966329650563e-06, + "loss": 0.278, + "step": 14174 + }, + { + "epoch": 0.86, + "learning_rate": 1.0630882450392221e-06, + "loss": 0.2451, + "step": 14175 + }, + { + "epoch": 0.86, + "learning_rate": 1.0622105025746921e-06, + "loss": 0.2783, + "step": 14176 + }, + { + "epoch": 0.86, + "learning_rate": 1.061333102290576e-06, + "loss": 0.2461, + "step": 14177 + }, + { + "epoch": 0.86, + "learning_rate": 1.0604560442204647e-06, + "loss": 0.2753, + "step": 14178 + }, + { + "epoch": 0.86, + "learning_rate": 1.059579328397936e-06, + "loss": 0.2581, + "step": 14179 + }, + { + "epoch": 0.86, + "learning_rate": 1.058702954856553e-06, + "loss": 0.2653, + "step": 14180 + }, + { + "epoch": 0.86, + "learning_rate": 1.057826923629871e-06, + "loss": 0.3058, + "step": 14181 + }, + { + "epoch": 0.86, + "learning_rate": 1.0569512347514266e-06, + "loss": 0.2821, + "step": 14182 + }, + { + "epoch": 0.86, + "learning_rate": 1.056075888254745e-06, + "loss": 0.2575, + "step": 14183 + }, + { + "epoch": 0.86, + "learning_rate": 1.0552008841733386e-06, + "loss": 0.2826, + "step": 14184 + }, + { + "epoch": 0.86, + "learning_rate": 1.0543262225407048e-06, + "loss": 0.2527, + "step": 14185 + }, + { + "epoch": 0.86, + "learning_rate": 1.0534519033903334e-06, + "loss": 0.2453, + "step": 14186 + }, + { + "epoch": 0.86, + "learning_rate": 1.0525779267556968e-06, + "loss": 0.2839, + "step": 14187 + }, + { + "epoch": 0.86, + "learning_rate": 1.0517042926702537e-06, + "loss": 0.2575, + "step": 14188 + }, + { + "epoch": 0.86, + "learning_rate": 1.0508310011674517e-06, + "loss": 0.2547, + "step": 14189 + }, + { + "epoch": 0.86, + "learning_rate": 1.0499580522807228e-06, + "loss": 0.2888, + "step": 14190 + }, + { + "epoch": 0.86, + "learning_rate": 1.0490854460434896e-06, + "loss": 0.2808, + "step": 14191 + }, + { + "epoch": 0.86, + "learning_rate": 1.0482131824891583e-06, + "loss": 0.2824, + "step": 14192 + }, + { + "epoch": 0.86, + "learning_rate": 1.0473412616511235e-06, + "loss": 0.2861, + "step": 14193 + }, + { + "epoch": 0.86, + "learning_rate": 1.0464696835627641e-06, + "loss": 0.2797, + "step": 14194 + }, + { + "epoch": 0.86, + "learning_rate": 1.0455984482574532e-06, + "loss": 0.2461, + "step": 14195 + }, + { + "epoch": 0.86, + "learning_rate": 1.0447275557685433e-06, + "loss": 0.25, + "step": 14196 + }, + { + "epoch": 0.86, + "learning_rate": 1.0438570061293762e-06, + "loss": 0.2706, + "step": 14197 + }, + { + "epoch": 0.86, + "learning_rate": 1.04298679937328e-06, + "loss": 0.2744, + "step": 14198 + }, + { + "epoch": 0.86, + "learning_rate": 1.0421169355335692e-06, + "loss": 0.2726, + "step": 14199 + }, + { + "epoch": 0.86, + "learning_rate": 1.041247414643549e-06, + "loss": 0.2769, + "step": 14200 + }, + { + "epoch": 0.86, + "learning_rate": 1.040378236736509e-06, + "loss": 0.2677, + "step": 14201 + }, + { + "epoch": 0.86, + "learning_rate": 1.039509401845723e-06, + "loss": 0.2606, + "step": 14202 + }, + { + "epoch": 0.86, + "learning_rate": 1.0386409100044548e-06, + "loss": 0.2889, + "step": 14203 + }, + { + "epoch": 0.86, + "learning_rate": 1.0377727612459532e-06, + "loss": 0.2681, + "step": 14204 + }, + { + "epoch": 0.86, + "learning_rate": 1.0369049556034583e-06, + "loss": 0.2725, + "step": 14205 + }, + { + "epoch": 0.86, + "learning_rate": 1.0360374931101912e-06, + "loss": 0.2485, + "step": 14206 + }, + { + "epoch": 0.86, + "learning_rate": 1.0351703737993645e-06, + "loss": 0.2652, + "step": 14207 + }, + { + "epoch": 0.86, + "learning_rate": 1.0343035977041726e-06, + "loss": 0.245, + "step": 14208 + }, + { + "epoch": 0.86, + "learning_rate": 1.0334371648578024e-06, + "loss": 0.2854, + "step": 14209 + }, + { + "epoch": 0.86, + "learning_rate": 1.0325710752934214e-06, + "loss": 0.2831, + "step": 14210 + }, + { + "epoch": 0.86, + "learning_rate": 1.0317053290441924e-06, + "loss": 0.2875, + "step": 14211 + }, + { + "epoch": 0.86, + "learning_rate": 1.0308399261432578e-06, + "loss": 0.2571, + "step": 14212 + }, + { + "epoch": 0.86, + "learning_rate": 1.0299748666237485e-06, + "loss": 0.2482, + "step": 14213 + }, + { + "epoch": 0.86, + "learning_rate": 1.029110150518785e-06, + "loss": 0.2624, + "step": 14214 + }, + { + "epoch": 0.86, + "learning_rate": 1.0282457778614697e-06, + "loss": 0.2853, + "step": 14215 + }, + { + "epoch": 0.86, + "learning_rate": 1.027381748684898e-06, + "loss": 0.2679, + "step": 14216 + }, + { + "epoch": 0.86, + "learning_rate": 1.0265180630221472e-06, + "loss": 0.2553, + "step": 14217 + }, + { + "epoch": 0.86, + "learning_rate": 1.025654720906285e-06, + "loss": 0.2775, + "step": 14218 + }, + { + "epoch": 0.86, + "learning_rate": 1.024791722370362e-06, + "loss": 0.2969, + "step": 14219 + }, + { + "epoch": 0.86, + "learning_rate": 1.0239290674474167e-06, + "loss": 0.264, + "step": 14220 + }, + { + "epoch": 0.86, + "learning_rate": 1.0230667561704799e-06, + "loss": 0.2909, + "step": 14221 + }, + { + "epoch": 0.86, + "learning_rate": 1.0222047885725629e-06, + "loss": 0.2733, + "step": 14222 + }, + { + "epoch": 0.86, + "learning_rate": 1.0213431646866667e-06, + "loss": 0.2706, + "step": 14223 + }, + { + "epoch": 0.86, + "learning_rate": 1.0204818845457742e-06, + "loss": 0.2691, + "step": 14224 + }, + { + "epoch": 0.86, + "learning_rate": 1.0196209481828633e-06, + "loss": 0.2582, + "step": 14225 + }, + { + "epoch": 0.86, + "learning_rate": 1.0187603556308934e-06, + "loss": 0.2687, + "step": 14226 + }, + { + "epoch": 0.86, + "learning_rate": 1.0179001069228134e-06, + "loss": 0.251, + "step": 14227 + }, + { + "epoch": 0.86, + "learning_rate": 1.0170402020915548e-06, + "loss": 0.2715, + "step": 14228 + }, + { + "epoch": 0.86, + "learning_rate": 1.0161806411700392e-06, + "loss": 0.2409, + "step": 14229 + }, + { + "epoch": 0.86, + "learning_rate": 1.015321424191178e-06, + "loss": 0.2898, + "step": 14230 + }, + { + "epoch": 0.86, + "learning_rate": 1.0144625511878636e-06, + "loss": 0.2744, + "step": 14231 + }, + { + "epoch": 0.86, + "learning_rate": 1.0136040221929777e-06, + "loss": 0.2595, + "step": 14232 + }, + { + "epoch": 0.86, + "learning_rate": 1.0127458372393895e-06, + "loss": 0.2398, + "step": 14233 + }, + { + "epoch": 0.86, + "learning_rate": 1.0118879963599516e-06, + "loss": 0.2702, + "step": 14234 + }, + { + "epoch": 0.86, + "learning_rate": 1.0110304995875109e-06, + "loss": 0.2531, + "step": 14235 + }, + { + "epoch": 0.86, + "learning_rate": 1.0101733469548924e-06, + "loss": 0.2609, + "step": 14236 + }, + { + "epoch": 0.86, + "learning_rate": 1.0093165384949155e-06, + "loss": 0.2538, + "step": 14237 + }, + { + "epoch": 0.86, + "learning_rate": 1.0084600742403794e-06, + "loss": 0.246, + "step": 14238 + }, + { + "epoch": 0.86, + "learning_rate": 1.0076039542240735e-06, + "loss": 0.2598, + "step": 14239 + }, + { + "epoch": 0.86, + "learning_rate": 1.006748178478777e-06, + "loss": 0.261, + "step": 14240 + }, + { + "epoch": 0.86, + "learning_rate": 1.0058927470372515e-06, + "loss": 0.2651, + "step": 14241 + }, + { + "epoch": 0.86, + "learning_rate": 1.0050376599322475e-06, + "loss": 0.267, + "step": 14242 + }, + { + "epoch": 0.86, + "learning_rate": 1.004182917196501e-06, + "loss": 0.2944, + "step": 14243 + }, + { + "epoch": 0.86, + "learning_rate": 1.0033285188627362e-06, + "loss": 0.3068, + "step": 14244 + }, + { + "epoch": 0.86, + "learning_rate": 1.0024744649636597e-06, + "loss": 0.2633, + "step": 14245 + }, + { + "epoch": 0.86, + "learning_rate": 1.0016207555319757e-06, + "loss": 0.2527, + "step": 14246 + }, + { + "epoch": 0.86, + "learning_rate": 1.0007673906003635e-06, + "loss": 0.2702, + "step": 14247 + }, + { + "epoch": 0.86, + "learning_rate": 9.999143702014947e-07, + "loss": 0.2829, + "step": 14248 + }, + { + "epoch": 0.86, + "learning_rate": 9.990616943680266e-07, + "loss": 0.2603, + "step": 14249 + }, + { + "epoch": 0.86, + "learning_rate": 9.982093631326028e-07, + "loss": 0.2613, + "step": 14250 + }, + { + "epoch": 0.86, + "learning_rate": 9.973573765278577e-07, + "loss": 0.2811, + "step": 14251 + }, + { + "epoch": 0.86, + "learning_rate": 9.965057345864083e-07, + "loss": 0.2977, + "step": 14252 + }, + { + "epoch": 0.86, + "learning_rate": 9.956544373408571e-07, + "loss": 0.2764, + "step": 14253 + }, + { + "epoch": 0.86, + "learning_rate": 9.948034848237985e-07, + "loss": 0.2772, + "step": 14254 + }, + { + "epoch": 0.86, + "learning_rate": 9.939528770678075e-07, + "loss": 0.2879, + "step": 14255 + }, + { + "epoch": 0.86, + "learning_rate": 9.931026141054533e-07, + "loss": 0.2833, + "step": 14256 + }, + { + "epoch": 0.86, + "learning_rate": 9.92252695969288e-07, + "loss": 0.2927, + "step": 14257 + }, + { + "epoch": 0.86, + "learning_rate": 9.914031226918464e-07, + "loss": 0.2595, + "step": 14258 + }, + { + "epoch": 0.86, + "learning_rate": 9.905538943056547e-07, + "loss": 0.2552, + "step": 14259 + }, + { + "epoch": 0.86, + "learning_rate": 9.897050108432293e-07, + "loss": 0.242, + "step": 14260 + }, + { + "epoch": 0.86, + "learning_rate": 9.888564723370664e-07, + "loss": 0.2461, + "step": 14261 + }, + { + "epoch": 0.86, + "learning_rate": 9.880082788196533e-07, + "loss": 0.2527, + "step": 14262 + }, + { + "epoch": 0.86, + "learning_rate": 9.871604303234627e-07, + "loss": 0.2958, + "step": 14263 + }, + { + "epoch": 0.86, + "learning_rate": 9.863129268809512e-07, + "loss": 0.2783, + "step": 14264 + }, + { + "epoch": 0.86, + "learning_rate": 9.854657685245695e-07, + "loss": 0.2651, + "step": 14265 + }, + { + "epoch": 0.86, + "learning_rate": 9.846189552867502e-07, + "loss": 0.258, + "step": 14266 + }, + { + "epoch": 0.86, + "learning_rate": 9.83772487199911e-07, + "loss": 0.2799, + "step": 14267 + }, + { + "epoch": 0.86, + "learning_rate": 9.829263642964603e-07, + "loss": 0.2694, + "step": 14268 + }, + { + "epoch": 0.86, + "learning_rate": 9.82080586608789e-07, + "loss": 0.2944, + "step": 14269 + }, + { + "epoch": 0.86, + "learning_rate": 9.812351541692821e-07, + "loss": 0.2792, + "step": 14270 + }, + { + "epoch": 0.86, + "learning_rate": 9.80390067010304e-07, + "loss": 0.261, + "step": 14271 + }, + { + "epoch": 0.86, + "learning_rate": 9.795453251642084e-07, + "loss": 0.2686, + "step": 14272 + }, + { + "epoch": 0.86, + "learning_rate": 9.787009286633365e-07, + "loss": 0.3044, + "step": 14273 + }, + { + "epoch": 0.86, + "learning_rate": 9.778568775400133e-07, + "loss": 0.2458, + "step": 14274 + }, + { + "epoch": 0.86, + "learning_rate": 9.770131718265586e-07, + "loss": 0.2556, + "step": 14275 + }, + { + "epoch": 0.86, + "learning_rate": 9.761698115552688e-07, + "loss": 0.2759, + "step": 14276 + }, + { + "epoch": 0.86, + "learning_rate": 9.753267967584335e-07, + "loss": 0.231, + "step": 14277 + }, + { + "epoch": 0.86, + "learning_rate": 9.74484127468327e-07, + "loss": 0.2591, + "step": 14278 + }, + { + "epoch": 0.86, + "learning_rate": 9.736418037172101e-07, + "loss": 0.2655, + "step": 14279 + }, + { + "epoch": 0.86, + "learning_rate": 9.727998255373295e-07, + "loss": 0.2854, + "step": 14280 + }, + { + "epoch": 0.86, + "learning_rate": 9.719581929609234e-07, + "loss": 0.2591, + "step": 14281 + }, + { + "epoch": 0.86, + "learning_rate": 9.71116906020212e-07, + "loss": 0.2711, + "step": 14282 + }, + { + "epoch": 0.86, + "learning_rate": 9.702759647474025e-07, + "loss": 0.2666, + "step": 14283 + }, + { + "epoch": 0.86, + "learning_rate": 9.694353691746928e-07, + "loss": 0.2888, + "step": 14284 + }, + { + "epoch": 0.86, + "learning_rate": 9.685951193342603e-07, + "loss": 0.2652, + "step": 14285 + }, + { + "epoch": 0.86, + "learning_rate": 9.67755215258278e-07, + "loss": 0.2596, + "step": 14286 + }, + { + "epoch": 0.86, + "learning_rate": 9.669156569788995e-07, + "loss": 0.2645, + "step": 14287 + }, + { + "epoch": 0.86, + "learning_rate": 9.660764445282689e-07, + "loss": 0.2621, + "step": 14288 + }, + { + "epoch": 0.86, + "learning_rate": 9.652375779385126e-07, + "loss": 0.2822, + "step": 14289 + }, + { + "epoch": 0.86, + "learning_rate": 9.64399057241746e-07, + "loss": 0.2472, + "step": 14290 + }, + { + "epoch": 0.86, + "learning_rate": 9.63560882470076e-07, + "loss": 0.2288, + "step": 14291 + }, + { + "epoch": 0.86, + "learning_rate": 9.627230536555876e-07, + "loss": 0.2818, + "step": 14292 + }, + { + "epoch": 0.86, + "learning_rate": 9.618855708303576e-07, + "loss": 0.2577, + "step": 14293 + }, + { + "epoch": 0.86, + "learning_rate": 9.61048434026448e-07, + "loss": 0.2698, + "step": 14294 + }, + { + "epoch": 0.86, + "learning_rate": 9.602116432759112e-07, + "loss": 0.2701, + "step": 14295 + }, + { + "epoch": 0.86, + "learning_rate": 9.593751986107814e-07, + "loss": 0.2804, + "step": 14296 + }, + { + "epoch": 0.86, + "learning_rate": 9.585391000630827e-07, + "loss": 0.2786, + "step": 14297 + }, + { + "epoch": 0.86, + "learning_rate": 9.577033476648234e-07, + "loss": 0.2698, + "step": 14298 + }, + { + "epoch": 0.86, + "learning_rate": 9.568679414479997e-07, + "loss": 0.2838, + "step": 14299 + }, + { + "epoch": 0.86, + "learning_rate": 9.560328814445974e-07, + "loss": 0.2886, + "step": 14300 + }, + { + "epoch": 0.86, + "learning_rate": 9.551981676865852e-07, + "loss": 0.2356, + "step": 14301 + }, + { + "epoch": 0.86, + "learning_rate": 9.543638002059185e-07, + "loss": 0.2633, + "step": 14302 + }, + { + "epoch": 0.86, + "learning_rate": 9.535297790345433e-07, + "loss": 0.2758, + "step": 14303 + }, + { + "epoch": 0.86, + "learning_rate": 9.52696104204387e-07, + "loss": 0.2747, + "step": 14304 + }, + { + "epoch": 0.86, + "learning_rate": 9.518627757473686e-07, + "loss": 0.267, + "step": 14305 + }, + { + "epoch": 0.86, + "learning_rate": 9.510297936953927e-07, + "loss": 0.2586, + "step": 14306 + }, + { + "epoch": 0.86, + "learning_rate": 9.501971580803471e-07, + "loss": 0.2655, + "step": 14307 + }, + { + "epoch": 0.86, + "learning_rate": 9.493648689341106e-07, + "loss": 0.2464, + "step": 14308 + }, + { + "epoch": 0.86, + "learning_rate": 9.485329262885457e-07, + "loss": 0.2591, + "step": 14309 + }, + { + "epoch": 0.86, + "learning_rate": 9.477013301755055e-07, + "loss": 0.2715, + "step": 14310 + }, + { + "epoch": 0.86, + "learning_rate": 9.468700806268261e-07, + "loss": 0.2644, + "step": 14311 + }, + { + "epoch": 0.86, + "learning_rate": 9.460391776743317e-07, + "loss": 0.2501, + "step": 14312 + }, + { + "epoch": 0.86, + "learning_rate": 9.452086213498323e-07, + "loss": 0.2515, + "step": 14313 + }, + { + "epoch": 0.86, + "learning_rate": 9.443784116851262e-07, + "loss": 0.2676, + "step": 14314 + }, + { + "epoch": 0.86, + "learning_rate": 9.435485487119966e-07, + "loss": 0.2659, + "step": 14315 + }, + { + "epoch": 0.86, + "learning_rate": 9.427190324622171e-07, + "loss": 0.2807, + "step": 14316 + }, + { + "epoch": 0.86, + "learning_rate": 9.418898629675443e-07, + "loss": 0.2482, + "step": 14317 + }, + { + "epoch": 0.86, + "learning_rate": 9.41061040259722e-07, + "loss": 0.2845, + "step": 14318 + }, + { + "epoch": 0.87, + "learning_rate": 9.402325643704823e-07, + "loss": 0.25, + "step": 14319 + }, + { + "epoch": 0.87, + "learning_rate": 9.394044353315401e-07, + "loss": 0.2852, + "step": 14320 + }, + { + "epoch": 0.87, + "learning_rate": 9.385766531746055e-07, + "loss": 0.246, + "step": 14321 + }, + { + "epoch": 0.87, + "learning_rate": 9.377492179313652e-07, + "loss": 0.2706, + "step": 14322 + }, + { + "epoch": 0.87, + "learning_rate": 9.369221296335007e-07, + "loss": 0.2803, + "step": 14323 + }, + { + "epoch": 0.87, + "learning_rate": 9.360953883126744e-07, + "loss": 0.2753, + "step": 14324 + }, + { + "epoch": 0.87, + "learning_rate": 9.352689940005389e-07, + "loss": 0.2619, + "step": 14325 + }, + { + "epoch": 0.87, + "learning_rate": 9.344429467287319e-07, + "loss": 0.2621, + "step": 14326 + }, + { + "epoch": 0.87, + "learning_rate": 9.336172465288796e-07, + "loss": 0.2694, + "step": 14327 + }, + { + "epoch": 0.87, + "learning_rate": 9.327918934325908e-07, + "loss": 0.2543, + "step": 14328 + }, + { + "epoch": 0.87, + "learning_rate": 9.319668874714649e-07, + "loss": 0.2646, + "step": 14329 + }, + { + "epoch": 0.87, + "learning_rate": 9.311422286770899e-07, + "loss": 0.2621, + "step": 14330 + }, + { + "epoch": 0.87, + "learning_rate": 9.303179170810362e-07, + "loss": 0.2515, + "step": 14331 + }, + { + "epoch": 0.87, + "learning_rate": 9.294939527148606e-07, + "loss": 0.2766, + "step": 14332 + }, + { + "epoch": 0.87, + "learning_rate": 9.28670335610109e-07, + "loss": 0.2706, + "step": 14333 + }, + { + "epoch": 0.87, + "learning_rate": 9.27847065798313e-07, + "loss": 0.2634, + "step": 14334 + }, + { + "epoch": 0.87, + "learning_rate": 9.270241433109928e-07, + "loss": 0.2624, + "step": 14335 + }, + { + "epoch": 0.87, + "learning_rate": 9.262015681796532e-07, + "loss": 0.2494, + "step": 14336 + }, + { + "epoch": 0.87, + "learning_rate": 9.253793404357858e-07, + "loss": 0.2473, + "step": 14337 + }, + { + "epoch": 0.87, + "learning_rate": 9.245574601108698e-07, + "loss": 0.2592, + "step": 14338 + }, + { + "epoch": 0.87, + "learning_rate": 9.237359272363678e-07, + "loss": 0.2831, + "step": 14339 + }, + { + "epoch": 0.87, + "learning_rate": 9.229147418437356e-07, + "loss": 0.2826, + "step": 14340 + }, + { + "epoch": 0.87, + "learning_rate": 9.220939039644116e-07, + "loss": 0.2418, + "step": 14341 + }, + { + "epoch": 0.87, + "learning_rate": 9.212734136298206e-07, + "loss": 0.2705, + "step": 14342 + }, + { + "epoch": 0.87, + "learning_rate": 9.204532708713743e-07, + "loss": 0.2765, + "step": 14343 + }, + { + "epoch": 0.87, + "learning_rate": 9.196334757204694e-07, + "loss": 0.284, + "step": 14344 + }, + { + "epoch": 0.87, + "learning_rate": 9.188140282084968e-07, + "loss": 0.2684, + "step": 14345 + }, + { + "epoch": 0.87, + "learning_rate": 9.179949283668265e-07, + "loss": 0.2605, + "step": 14346 + }, + { + "epoch": 0.87, + "learning_rate": 9.171761762268161e-07, + "loss": 0.2656, + "step": 14347 + }, + { + "epoch": 0.87, + "learning_rate": 9.163577718198124e-07, + "loss": 0.2447, + "step": 14348 + }, + { + "epoch": 0.87, + "learning_rate": 9.155397151771472e-07, + "loss": 0.2539, + "step": 14349 + }, + { + "epoch": 0.87, + "learning_rate": 9.147220063301387e-07, + "loss": 0.251, + "step": 14350 + }, + { + "epoch": 0.87, + "learning_rate": 9.139046453100952e-07, + "loss": 0.2585, + "step": 14351 + }, + { + "epoch": 0.87, + "learning_rate": 9.130876321483084e-07, + "loss": 0.2921, + "step": 14352 + }, + { + "epoch": 0.87, + "learning_rate": 9.122709668760565e-07, + "loss": 0.2732, + "step": 14353 + }, + { + "epoch": 0.87, + "learning_rate": 9.114546495246046e-07, + "loss": 0.2736, + "step": 14354 + }, + { + "epoch": 0.87, + "learning_rate": 9.106386801252054e-07, + "loss": 0.2771, + "step": 14355 + }, + { + "epoch": 0.87, + "learning_rate": 9.098230587090995e-07, + "loss": 0.2614, + "step": 14356 + }, + { + "epoch": 0.87, + "learning_rate": 9.090077853075119e-07, + "loss": 0.2778, + "step": 14357 + }, + { + "epoch": 0.87, + "learning_rate": 9.081928599516554e-07, + "loss": 0.2423, + "step": 14358 + }, + { + "epoch": 0.87, + "learning_rate": 9.073782826727285e-07, + "loss": 0.2653, + "step": 14359 + }, + { + "epoch": 0.87, + "learning_rate": 9.065640535019182e-07, + "loss": 0.3027, + "step": 14360 + }, + { + "epoch": 0.87, + "learning_rate": 9.057501724703955e-07, + "loss": 0.2658, + "step": 14361 + }, + { + "epoch": 0.87, + "learning_rate": 9.049366396093195e-07, + "loss": 0.2786, + "step": 14362 + }, + { + "epoch": 0.87, + "learning_rate": 9.041234549498379e-07, + "loss": 0.2545, + "step": 14363 + }, + { + "epoch": 0.87, + "learning_rate": 9.033106185230789e-07, + "loss": 0.2531, + "step": 14364 + }, + { + "epoch": 0.87, + "learning_rate": 9.024981303601677e-07, + "loss": 0.3111, + "step": 14365 + }, + { + "epoch": 0.87, + "learning_rate": 9.01685990492207e-07, + "loss": 0.2229, + "step": 14366 + }, + { + "epoch": 0.87, + "learning_rate": 9.0087419895029e-07, + "loss": 0.2835, + "step": 14367 + }, + { + "epoch": 0.87, + "learning_rate": 9.000627557654951e-07, + "loss": 0.2511, + "step": 14368 + }, + { + "epoch": 0.87, + "learning_rate": 8.992516609688861e-07, + "loss": 0.2359, + "step": 14369 + }, + { + "epoch": 0.87, + "learning_rate": 8.984409145915207e-07, + "loss": 0.2486, + "step": 14370 + }, + { + "epoch": 0.87, + "learning_rate": 8.97630516664435e-07, + "loss": 0.2906, + "step": 14371 + }, + { + "epoch": 0.87, + "learning_rate": 8.968204672186542e-07, + "loss": 0.2628, + "step": 14372 + }, + { + "epoch": 0.87, + "learning_rate": 8.960107662851925e-07, + "loss": 0.2677, + "step": 14373 + }, + { + "epoch": 0.87, + "learning_rate": 8.952014138950471e-07, + "loss": 0.2503, + "step": 14374 + }, + { + "epoch": 0.87, + "learning_rate": 8.943924100792056e-07, + "loss": 0.2811, + "step": 14375 + }, + { + "epoch": 0.87, + "learning_rate": 8.93583754868641e-07, + "loss": 0.275, + "step": 14376 + }, + { + "epoch": 0.87, + "learning_rate": 8.927754482943096e-07, + "loss": 0.2734, + "step": 14377 + }, + { + "epoch": 0.87, + "learning_rate": 8.9196749038716e-07, + "loss": 0.275, + "step": 14378 + }, + { + "epoch": 0.87, + "learning_rate": 8.911598811781209e-07, + "loss": 0.2645, + "step": 14379 + }, + { + "epoch": 0.87, + "learning_rate": 8.903526206981161e-07, + "loss": 0.2613, + "step": 14380 + }, + { + "epoch": 0.87, + "learning_rate": 8.895457089780491e-07, + "loss": 0.2485, + "step": 14381 + }, + { + "epoch": 0.87, + "learning_rate": 8.887391460488115e-07, + "loss": 0.2913, + "step": 14382 + }, + { + "epoch": 0.87, + "learning_rate": 8.879329319412843e-07, + "loss": 0.2599, + "step": 14383 + }, + { + "epoch": 0.87, + "learning_rate": 8.871270666863285e-07, + "loss": 0.278, + "step": 14384 + }, + { + "epoch": 0.87, + "learning_rate": 8.863215503148026e-07, + "loss": 0.2564, + "step": 14385 + }, + { + "epoch": 0.87, + "learning_rate": 8.855163828575419e-07, + "loss": 0.2864, + "step": 14386 + }, + { + "epoch": 0.87, + "learning_rate": 8.847115643453741e-07, + "loss": 0.238, + "step": 14387 + }, + { + "epoch": 0.87, + "learning_rate": 8.839070948091089e-07, + "loss": 0.2831, + "step": 14388 + }, + { + "epoch": 0.87, + "learning_rate": 8.83102974279546e-07, + "loss": 0.2571, + "step": 14389 + }, + { + "epoch": 0.87, + "learning_rate": 8.8229920278747e-07, + "loss": 0.2644, + "step": 14390 + }, + { + "epoch": 0.87, + "learning_rate": 8.814957803636559e-07, + "loss": 0.2676, + "step": 14391 + }, + { + "epoch": 0.87, + "learning_rate": 8.806927070388616e-07, + "loss": 0.2482, + "step": 14392 + }, + { + "epoch": 0.87, + "learning_rate": 8.798899828438334e-07, + "loss": 0.2553, + "step": 14393 + }, + { + "epoch": 0.87, + "learning_rate": 8.790876078092969e-07, + "loss": 0.2621, + "step": 14394 + }, + { + "epoch": 0.87, + "learning_rate": 8.782855819659785e-07, + "loss": 0.2712, + "step": 14395 + }, + { + "epoch": 0.87, + "learning_rate": 8.774839053445805e-07, + "loss": 0.2492, + "step": 14396 + }, + { + "epoch": 0.87, + "learning_rate": 8.766825779757937e-07, + "loss": 0.273, + "step": 14397 + }, + { + "epoch": 0.87, + "learning_rate": 8.758815998902992e-07, + "loss": 0.2577, + "step": 14398 + }, + { + "epoch": 0.87, + "learning_rate": 8.75080971118758e-07, + "loss": 0.2363, + "step": 14399 + }, + { + "epoch": 0.87, + "learning_rate": 8.742806916918278e-07, + "loss": 0.254, + "step": 14400 + }, + { + "epoch": 0.87, + "learning_rate": 8.734807616401431e-07, + "loss": 0.2884, + "step": 14401 + }, + { + "epoch": 0.87, + "learning_rate": 8.726811809943303e-07, + "loss": 0.2669, + "step": 14402 + }, + { + "epoch": 0.87, + "learning_rate": 8.718819497850006e-07, + "loss": 0.2801, + "step": 14403 + }, + { + "epoch": 0.87, + "learning_rate": 8.710830680427507e-07, + "loss": 0.2641, + "step": 14404 + }, + { + "epoch": 0.87, + "learning_rate": 8.702845357981682e-07, + "loss": 0.2646, + "step": 14405 + }, + { + "epoch": 0.87, + "learning_rate": 8.694863530818243e-07, + "loss": 0.2775, + "step": 14406 + }, + { + "epoch": 0.87, + "learning_rate": 8.686885199242767e-07, + "loss": 0.2766, + "step": 14407 + }, + { + "epoch": 0.87, + "learning_rate": 8.678910363560689e-07, + "loss": 0.2666, + "step": 14408 + }, + { + "epoch": 0.87, + "learning_rate": 8.670939024077329e-07, + "loss": 0.2446, + "step": 14409 + }, + { + "epoch": 0.87, + "learning_rate": 8.662971181097879e-07, + "loss": 0.2697, + "step": 14410 + }, + { + "epoch": 0.87, + "learning_rate": 8.655006834927393e-07, + "loss": 0.2432, + "step": 14411 + }, + { + "epoch": 0.87, + "learning_rate": 8.647045985870761e-07, + "loss": 0.2649, + "step": 14412 + }, + { + "epoch": 0.87, + "learning_rate": 8.639088634232762e-07, + "loss": 0.2762, + "step": 14413 + }, + { + "epoch": 0.87, + "learning_rate": 8.63113478031804e-07, + "loss": 0.2586, + "step": 14414 + }, + { + "epoch": 0.87, + "learning_rate": 8.62318442443113e-07, + "loss": 0.2598, + "step": 14415 + }, + { + "epoch": 0.87, + "learning_rate": 8.6152375668764e-07, + "loss": 0.2826, + "step": 14416 + }, + { + "epoch": 0.87, + "learning_rate": 8.607294207958072e-07, + "loss": 0.2768, + "step": 14417 + }, + { + "epoch": 0.87, + "learning_rate": 8.599354347980282e-07, + "loss": 0.2527, + "step": 14418 + }, + { + "epoch": 0.87, + "learning_rate": 8.591417987246964e-07, + "loss": 0.2546, + "step": 14419 + }, + { + "epoch": 0.87, + "learning_rate": 8.583485126062007e-07, + "loss": 0.2712, + "step": 14420 + }, + { + "epoch": 0.87, + "learning_rate": 8.575555764729104e-07, + "loss": 0.2676, + "step": 14421 + }, + { + "epoch": 0.87, + "learning_rate": 8.56762990355181e-07, + "loss": 0.2626, + "step": 14422 + }, + { + "epoch": 0.87, + "learning_rate": 8.559707542833584e-07, + "loss": 0.2446, + "step": 14423 + }, + { + "epoch": 0.87, + "learning_rate": 8.551788682877726e-07, + "loss": 0.2691, + "step": 14424 + }, + { + "epoch": 0.87, + "learning_rate": 8.543873323987384e-07, + "loss": 0.2581, + "step": 14425 + }, + { + "epoch": 0.87, + "learning_rate": 8.535961466465626e-07, + "loss": 0.2582, + "step": 14426 + }, + { + "epoch": 0.87, + "learning_rate": 8.528053110615364e-07, + "loss": 0.2814, + "step": 14427 + }, + { + "epoch": 0.87, + "learning_rate": 8.520148256739336e-07, + "loss": 0.2702, + "step": 14428 + }, + { + "epoch": 0.87, + "learning_rate": 8.512246905140164e-07, + "loss": 0.2674, + "step": 14429 + }, + { + "epoch": 0.87, + "learning_rate": 8.504349056120387e-07, + "loss": 0.2501, + "step": 14430 + }, + { + "epoch": 0.87, + "learning_rate": 8.496454709982371e-07, + "loss": 0.246, + "step": 14431 + }, + { + "epoch": 0.87, + "learning_rate": 8.488563867028321e-07, + "loss": 0.239, + "step": 14432 + }, + { + "epoch": 0.87, + "learning_rate": 8.48067652756035e-07, + "loss": 0.2737, + "step": 14433 + }, + { + "epoch": 0.87, + "learning_rate": 8.472792691880405e-07, + "loss": 0.2415, + "step": 14434 + }, + { + "epoch": 0.87, + "learning_rate": 8.464912360290356e-07, + "loss": 0.2994, + "step": 14435 + }, + { + "epoch": 0.87, + "learning_rate": 8.457035533091873e-07, + "loss": 0.2607, + "step": 14436 + }, + { + "epoch": 0.87, + "learning_rate": 8.449162210586514e-07, + "loss": 0.2719, + "step": 14437 + }, + { + "epoch": 0.87, + "learning_rate": 8.441292393075717e-07, + "loss": 0.273, + "step": 14438 + }, + { + "epoch": 0.87, + "learning_rate": 8.433426080860752e-07, + "loss": 0.279, + "step": 14439 + }, + { + "epoch": 0.87, + "learning_rate": 8.42556327424282e-07, + "loss": 0.25, + "step": 14440 + }, + { + "epoch": 0.87, + "learning_rate": 8.417703973522917e-07, + "loss": 0.244, + "step": 14441 + }, + { + "epoch": 0.87, + "learning_rate": 8.409848179001945e-07, + "loss": 0.2637, + "step": 14442 + }, + { + "epoch": 0.87, + "learning_rate": 8.401995890980653e-07, + "loss": 0.2941, + "step": 14443 + }, + { + "epoch": 0.87, + "learning_rate": 8.394147109759654e-07, + "loss": 0.2856, + "step": 14444 + }, + { + "epoch": 0.87, + "learning_rate": 8.386301835639454e-07, + "loss": 0.26, + "step": 14445 + }, + { + "epoch": 0.87, + "learning_rate": 8.378460068920413e-07, + "loss": 0.2473, + "step": 14446 + }, + { + "epoch": 0.87, + "learning_rate": 8.370621809902734e-07, + "loss": 0.2814, + "step": 14447 + }, + { + "epoch": 0.87, + "learning_rate": 8.362787058886501e-07, + "loss": 0.248, + "step": 14448 + }, + { + "epoch": 0.87, + "learning_rate": 8.354955816171661e-07, + "loss": 0.2896, + "step": 14449 + }, + { + "epoch": 0.87, + "learning_rate": 8.347128082058054e-07, + "loss": 0.2677, + "step": 14450 + }, + { + "epoch": 0.87, + "learning_rate": 8.339303856845348e-07, + "loss": 0.268, + "step": 14451 + }, + { + "epoch": 0.87, + "learning_rate": 8.331483140833097e-07, + "loss": 0.2688, + "step": 14452 + }, + { + "epoch": 0.87, + "learning_rate": 8.323665934320713e-07, + "loss": 0.3064, + "step": 14453 + }, + { + "epoch": 0.87, + "learning_rate": 8.315852237607447e-07, + "loss": 0.2806, + "step": 14454 + }, + { + "epoch": 0.87, + "learning_rate": 8.308042050992482e-07, + "loss": 0.2291, + "step": 14455 + }, + { + "epoch": 0.87, + "learning_rate": 8.300235374774824e-07, + "loss": 0.2761, + "step": 14456 + }, + { + "epoch": 0.87, + "learning_rate": 8.292432209253331e-07, + "loss": 0.2704, + "step": 14457 + }, + { + "epoch": 0.87, + "learning_rate": 8.284632554726757e-07, + "loss": 0.2713, + "step": 14458 + }, + { + "epoch": 0.87, + "learning_rate": 8.276836411493716e-07, + "loss": 0.2589, + "step": 14459 + }, + { + "epoch": 0.87, + "learning_rate": 8.269043779852637e-07, + "loss": 0.2759, + "step": 14460 + }, + { + "epoch": 0.87, + "learning_rate": 8.261254660101936e-07, + "loss": 0.2746, + "step": 14461 + }, + { + "epoch": 0.87, + "learning_rate": 8.253469052539753e-07, + "loss": 0.2793, + "step": 14462 + }, + { + "epoch": 0.87, + "learning_rate": 8.245686957464183e-07, + "loss": 0.268, + "step": 14463 + }, + { + "epoch": 0.87, + "learning_rate": 8.237908375173132e-07, + "loss": 0.2436, + "step": 14464 + }, + { + "epoch": 0.87, + "learning_rate": 8.23013330596445e-07, + "loss": 0.2409, + "step": 14465 + }, + { + "epoch": 0.87, + "learning_rate": 8.222361750135765e-07, + "loss": 0.272, + "step": 14466 + }, + { + "epoch": 0.87, + "learning_rate": 8.21459370798463e-07, + "loss": 0.2807, + "step": 14467 + }, + { + "epoch": 0.87, + "learning_rate": 8.206829179808429e-07, + "loss": 0.2713, + "step": 14468 + }, + { + "epoch": 0.87, + "learning_rate": 8.199068165904412e-07, + "loss": 0.2472, + "step": 14469 + }, + { + "epoch": 0.87, + "learning_rate": 8.191310666569741e-07, + "loss": 0.2588, + "step": 14470 + }, + { + "epoch": 0.87, + "learning_rate": 8.18355668210139e-07, + "loss": 0.2829, + "step": 14471 + }, + { + "epoch": 0.87, + "learning_rate": 8.175806212796222e-07, + "loss": 0.2627, + "step": 14472 + }, + { + "epoch": 0.87, + "learning_rate": 8.168059258950956e-07, + "loss": 0.2613, + "step": 14473 + }, + { + "epoch": 0.87, + "learning_rate": 8.160315820862175e-07, + "loss": 0.2584, + "step": 14474 + }, + { + "epoch": 0.87, + "learning_rate": 8.152575898826354e-07, + "loss": 0.2663, + "step": 14475 + }, + { + "epoch": 0.87, + "learning_rate": 8.1448394931398e-07, + "loss": 0.2506, + "step": 14476 + }, + { + "epoch": 0.87, + "learning_rate": 8.13710660409871e-07, + "loss": 0.2743, + "step": 14477 + }, + { + "epoch": 0.87, + "learning_rate": 8.129377231999113e-07, + "loss": 0.2762, + "step": 14478 + }, + { + "epoch": 0.87, + "learning_rate": 8.121651377136941e-07, + "loss": 0.2766, + "step": 14479 + }, + { + "epoch": 0.87, + "learning_rate": 8.113929039807977e-07, + "loss": 0.2462, + "step": 14480 + }, + { + "epoch": 0.87, + "learning_rate": 8.106210220307875e-07, + "loss": 0.2736, + "step": 14481 + }, + { + "epoch": 0.87, + "learning_rate": 8.098494918932132e-07, + "loss": 0.2833, + "step": 14482 + }, + { + "epoch": 0.87, + "learning_rate": 8.090783135976132e-07, + "loss": 0.2815, + "step": 14483 + }, + { + "epoch": 0.88, + "learning_rate": 8.083074871735097e-07, + "loss": 0.2751, + "step": 14484 + }, + { + "epoch": 0.88, + "learning_rate": 8.075370126504179e-07, + "loss": 0.2747, + "step": 14485 + }, + { + "epoch": 0.88, + "learning_rate": 8.067668900578329e-07, + "loss": 0.281, + "step": 14486 + }, + { + "epoch": 0.88, + "learning_rate": 8.059971194252391e-07, + "loss": 0.2743, + "step": 14487 + }, + { + "epoch": 0.88, + "learning_rate": 8.052277007821053e-07, + "loss": 0.2462, + "step": 14488 + }, + { + "epoch": 0.88, + "learning_rate": 8.044586341578886e-07, + "loss": 0.2676, + "step": 14489 + }, + { + "epoch": 0.88, + "learning_rate": 8.036899195820347e-07, + "loss": 0.2549, + "step": 14490 + }, + { + "epoch": 0.88, + "learning_rate": 8.029215570839733e-07, + "loss": 0.2621, + "step": 14491 + }, + { + "epoch": 0.88, + "learning_rate": 8.021535466931196e-07, + "loss": 0.2592, + "step": 14492 + }, + { + "epoch": 0.88, + "learning_rate": 8.013858884388758e-07, + "loss": 0.2637, + "step": 14493 + }, + { + "epoch": 0.88, + "learning_rate": 8.00618582350634e-07, + "loss": 0.2667, + "step": 14494 + }, + { + "epoch": 0.88, + "learning_rate": 7.998516284577684e-07, + "loss": 0.2776, + "step": 14495 + }, + { + "epoch": 0.88, + "learning_rate": 7.990850267896421e-07, + "loss": 0.2769, + "step": 14496 + }, + { + "epoch": 0.88, + "learning_rate": 7.983187773756052e-07, + "loss": 0.2414, + "step": 14497 + }, + { + "epoch": 0.88, + "learning_rate": 7.975528802449906e-07, + "loss": 0.2691, + "step": 14498 + }, + { + "epoch": 0.88, + "learning_rate": 7.967873354271216e-07, + "loss": 0.2648, + "step": 14499 + }, + { + "epoch": 0.88, + "learning_rate": 7.960221429513081e-07, + "loss": 0.2698, + "step": 14500 + }, + { + "epoch": 0.88, + "learning_rate": 7.952573028468458e-07, + "loss": 0.2684, + "step": 14501 + }, + { + "epoch": 0.88, + "learning_rate": 7.944928151430143e-07, + "loss": 0.2803, + "step": 14502 + }, + { + "epoch": 0.88, + "learning_rate": 7.937286798690825e-07, + "loss": 0.2575, + "step": 14503 + }, + { + "epoch": 0.88, + "learning_rate": 7.929648970543036e-07, + "loss": 0.2718, + "step": 14504 + }, + { + "epoch": 0.88, + "learning_rate": 7.92201466727922e-07, + "loss": 0.2701, + "step": 14505 + }, + { + "epoch": 0.88, + "learning_rate": 7.914383889191635e-07, + "loss": 0.2711, + "step": 14506 + }, + { + "epoch": 0.88, + "learning_rate": 7.906756636572433e-07, + "loss": 0.264, + "step": 14507 + }, + { + "epoch": 0.88, + "learning_rate": 7.899132909713603e-07, + "loss": 0.2893, + "step": 14508 + }, + { + "epoch": 0.88, + "learning_rate": 7.891512708907013e-07, + "loss": 0.2683, + "step": 14509 + }, + { + "epoch": 0.88, + "learning_rate": 7.883896034444427e-07, + "loss": 0.2515, + "step": 14510 + }, + { + "epoch": 0.88, + "learning_rate": 7.876282886617436e-07, + "loss": 0.2557, + "step": 14511 + }, + { + "epoch": 0.88, + "learning_rate": 7.868673265717508e-07, + "loss": 0.2542, + "step": 14512 + }, + { + "epoch": 0.88, + "learning_rate": 7.861067172035963e-07, + "loss": 0.2543, + "step": 14513 + }, + { + "epoch": 0.88, + "learning_rate": 7.853464605864003e-07, + "loss": 0.2709, + "step": 14514 + }, + { + "epoch": 0.88, + "learning_rate": 7.845865567492694e-07, + "loss": 0.2434, + "step": 14515 + }, + { + "epoch": 0.88, + "learning_rate": 7.838270057212971e-07, + "loss": 0.264, + "step": 14516 + }, + { + "epoch": 0.88, + "learning_rate": 7.830678075315623e-07, + "loss": 0.2476, + "step": 14517 + }, + { + "epoch": 0.88, + "learning_rate": 7.823089622091284e-07, + "loss": 0.2762, + "step": 14518 + }, + { + "epoch": 0.88, + "learning_rate": 7.815504697830489e-07, + "loss": 0.2617, + "step": 14519 + }, + { + "epoch": 0.88, + "learning_rate": 7.807923302823639e-07, + "loss": 0.2698, + "step": 14520 + }, + { + "epoch": 0.88, + "learning_rate": 7.800345437360968e-07, + "loss": 0.2857, + "step": 14521 + }, + { + "epoch": 0.88, + "learning_rate": 7.79277110173261e-07, + "loss": 0.2612, + "step": 14522 + }, + { + "epoch": 0.88, + "learning_rate": 7.785200296228523e-07, + "loss": 0.2812, + "step": 14523 + }, + { + "epoch": 0.88, + "learning_rate": 7.777633021138542e-07, + "loss": 0.2613, + "step": 14524 + }, + { + "epoch": 0.88, + "learning_rate": 7.770069276752423e-07, + "loss": 0.2859, + "step": 14525 + }, + { + "epoch": 0.88, + "learning_rate": 7.762509063359724e-07, + "loss": 0.2713, + "step": 14526 + }, + { + "epoch": 0.88, + "learning_rate": 7.754952381249869e-07, + "loss": 0.2575, + "step": 14527 + }, + { + "epoch": 0.88, + "learning_rate": 7.747399230712183e-07, + "loss": 0.263, + "step": 14528 + }, + { + "epoch": 0.88, + "learning_rate": 7.739849612035821e-07, + "loss": 0.2757, + "step": 14529 + }, + { + "epoch": 0.88, + "learning_rate": 7.732303525509821e-07, + "loss": 0.2746, + "step": 14530 + }, + { + "epoch": 0.88, + "learning_rate": 7.724760971423084e-07, + "loss": 0.255, + "step": 14531 + }, + { + "epoch": 0.88, + "learning_rate": 7.717221950064368e-07, + "loss": 0.3002, + "step": 14532 + }, + { + "epoch": 0.88, + "learning_rate": 7.709686461722321e-07, + "loss": 0.2653, + "step": 14533 + }, + { + "epoch": 0.88, + "learning_rate": 7.7021545066854e-07, + "loss": 0.2619, + "step": 14534 + }, + { + "epoch": 0.88, + "learning_rate": 7.694626085242007e-07, + "loss": 0.2746, + "step": 14535 + }, + { + "epoch": 0.88, + "learning_rate": 7.687101197680346e-07, + "loss": 0.2818, + "step": 14536 + }, + { + "epoch": 0.88, + "learning_rate": 7.679579844288509e-07, + "loss": 0.2479, + "step": 14537 + }, + { + "epoch": 0.88, + "learning_rate": 7.672062025354443e-07, + "loss": 0.3103, + "step": 14538 + }, + { + "epoch": 0.88, + "learning_rate": 7.664547741165939e-07, + "loss": 0.2741, + "step": 14539 + }, + { + "epoch": 0.88, + "learning_rate": 7.657036992010736e-07, + "loss": 0.2461, + "step": 14540 + }, + { + "epoch": 0.88, + "learning_rate": 7.649529778176345e-07, + "loss": 0.2314, + "step": 14541 + }, + { + "epoch": 0.88, + "learning_rate": 7.642026099950195e-07, + "loss": 0.2709, + "step": 14542 + }, + { + "epoch": 0.88, + "learning_rate": 7.634525957619543e-07, + "loss": 0.2588, + "step": 14543 + }, + { + "epoch": 0.88, + "learning_rate": 7.627029351471515e-07, + "loss": 0.2776, + "step": 14544 + }, + { + "epoch": 0.88, + "learning_rate": 7.61953628179316e-07, + "loss": 0.2812, + "step": 14545 + }, + { + "epoch": 0.88, + "learning_rate": 7.612046748871327e-07, + "loss": 0.249, + "step": 14546 + }, + { + "epoch": 0.88, + "learning_rate": 7.604560752992751e-07, + "loss": 0.2402, + "step": 14547 + }, + { + "epoch": 0.88, + "learning_rate": 7.597078294444026e-07, + "loss": 0.2529, + "step": 14548 + }, + { + "epoch": 0.88, + "learning_rate": 7.589599373511603e-07, + "loss": 0.292, + "step": 14549 + }, + { + "epoch": 0.88, + "learning_rate": 7.582123990481849e-07, + "loss": 0.2407, + "step": 14550 + }, + { + "epoch": 0.88, + "learning_rate": 7.574652145640926e-07, + "loss": 0.2644, + "step": 14551 + }, + { + "epoch": 0.88, + "learning_rate": 7.567183839274906e-07, + "loss": 0.286, + "step": 14552 + }, + { + "epoch": 0.88, + "learning_rate": 7.559719071669691e-07, + "loss": 0.2557, + "step": 14553 + }, + { + "epoch": 0.88, + "learning_rate": 7.552257843111077e-07, + "loss": 0.2619, + "step": 14554 + }, + { + "epoch": 0.88, + "learning_rate": 7.544800153884724e-07, + "loss": 0.2673, + "step": 14555 + }, + { + "epoch": 0.88, + "learning_rate": 7.537346004276147e-07, + "loss": 0.2547, + "step": 14556 + }, + { + "epoch": 0.88, + "learning_rate": 7.529895394570719e-07, + "loss": 0.3136, + "step": 14557 + }, + { + "epoch": 0.88, + "learning_rate": 7.522448325053688e-07, + "loss": 0.2566, + "step": 14558 + }, + { + "epoch": 0.88, + "learning_rate": 7.515004796010139e-07, + "loss": 0.2457, + "step": 14559 + }, + { + "epoch": 0.88, + "learning_rate": 7.507564807725076e-07, + "loss": 0.3078, + "step": 14560 + }, + { + "epoch": 0.88, + "learning_rate": 7.500128360483338e-07, + "loss": 0.281, + "step": 14561 + }, + { + "epoch": 0.88, + "learning_rate": 7.49269545456961e-07, + "loss": 0.2614, + "step": 14562 + }, + { + "epoch": 0.88, + "learning_rate": 7.485266090268462e-07, + "loss": 0.2492, + "step": 14563 + }, + { + "epoch": 0.88, + "learning_rate": 7.477840267864334e-07, + "loss": 0.2732, + "step": 14564 + }, + { + "epoch": 0.88, + "learning_rate": 7.47041798764151e-07, + "loss": 0.2838, + "step": 14565 + }, + { + "epoch": 0.88, + "learning_rate": 7.462999249884151e-07, + "loss": 0.287, + "step": 14566 + }, + { + "epoch": 0.88, + "learning_rate": 7.455584054876286e-07, + "loss": 0.2637, + "step": 14567 + }, + { + "epoch": 0.88, + "learning_rate": 7.448172402901798e-07, + "loss": 0.2553, + "step": 14568 + }, + { + "epoch": 0.88, + "learning_rate": 7.440764294244418e-07, + "loss": 0.2686, + "step": 14569 + }, + { + "epoch": 0.88, + "learning_rate": 7.433359729187817e-07, + "loss": 0.2518, + "step": 14570 + }, + { + "epoch": 0.88, + "learning_rate": 7.425958708015446e-07, + "loss": 0.2422, + "step": 14571 + }, + { + "epoch": 0.88, + "learning_rate": 7.418561231010646e-07, + "loss": 0.258, + "step": 14572 + }, + { + "epoch": 0.88, + "learning_rate": 7.411167298456634e-07, + "loss": 0.3022, + "step": 14573 + }, + { + "epoch": 0.88, + "learning_rate": 7.403776910636473e-07, + "loss": 0.2607, + "step": 14574 + }, + { + "epoch": 0.88, + "learning_rate": 7.396390067833126e-07, + "loss": 0.2607, + "step": 14575 + }, + { + "epoch": 0.88, + "learning_rate": 7.389006770329377e-07, + "loss": 0.2649, + "step": 14576 + }, + { + "epoch": 0.88, + "learning_rate": 7.3816270184079e-07, + "loss": 0.2362, + "step": 14577 + }, + { + "epoch": 0.88, + "learning_rate": 7.374250812351224e-07, + "loss": 0.2846, + "step": 14578 + }, + { + "epoch": 0.88, + "learning_rate": 7.366878152441725e-07, + "loss": 0.2538, + "step": 14579 + }, + { + "epoch": 0.88, + "learning_rate": 7.359509038961699e-07, + "loss": 0.2624, + "step": 14580 + }, + { + "epoch": 0.88, + "learning_rate": 7.352143472193263e-07, + "loss": 0.2793, + "step": 14581 + }, + { + "epoch": 0.88, + "learning_rate": 7.344781452418393e-07, + "loss": 0.2577, + "step": 14582 + }, + { + "epoch": 0.88, + "learning_rate": 7.33742297991894e-07, + "loss": 0.2462, + "step": 14583 + }, + { + "epoch": 0.88, + "learning_rate": 7.330068054976613e-07, + "loss": 0.2701, + "step": 14584 + }, + { + "epoch": 0.88, + "learning_rate": 7.322716677873021e-07, + "loss": 0.2298, + "step": 14585 + }, + { + "epoch": 0.88, + "learning_rate": 7.315368848889592e-07, + "loss": 0.2697, + "step": 14586 + }, + { + "epoch": 0.88, + "learning_rate": 7.308024568307648e-07, + "loss": 0.2546, + "step": 14587 + }, + { + "epoch": 0.88, + "learning_rate": 7.30068383640834e-07, + "loss": 0.2662, + "step": 14588 + }, + { + "epoch": 0.88, + "learning_rate": 7.2933466534727e-07, + "loss": 0.2619, + "step": 14589 + }, + { + "epoch": 0.88, + "learning_rate": 7.286013019781668e-07, + "loss": 0.2885, + "step": 14590 + }, + { + "epoch": 0.88, + "learning_rate": 7.278682935615989e-07, + "loss": 0.2779, + "step": 14591 + }, + { + "epoch": 0.88, + "learning_rate": 7.271356401256291e-07, + "loss": 0.2494, + "step": 14592 + }, + { + "epoch": 0.88, + "learning_rate": 7.264033416983074e-07, + "loss": 0.2691, + "step": 14593 + }, + { + "epoch": 0.88, + "learning_rate": 7.256713983076668e-07, + "loss": 0.2679, + "step": 14594 + }, + { + "epoch": 0.88, + "learning_rate": 7.24939809981734e-07, + "loss": 0.2517, + "step": 14595 + }, + { + "epoch": 0.88, + "learning_rate": 7.242085767485151e-07, + "loss": 0.2543, + "step": 14596 + }, + { + "epoch": 0.88, + "learning_rate": 7.234776986360059e-07, + "loss": 0.2725, + "step": 14597 + }, + { + "epoch": 0.88, + "learning_rate": 7.227471756721882e-07, + "loss": 0.2938, + "step": 14598 + }, + { + "epoch": 0.88, + "learning_rate": 7.220170078850286e-07, + "loss": 0.2487, + "step": 14599 + }, + { + "epoch": 0.88, + "learning_rate": 7.212871953024814e-07, + "loss": 0.2437, + "step": 14600 + }, + { + "epoch": 0.88, + "learning_rate": 7.205577379524875e-07, + "loss": 0.2776, + "step": 14601 + }, + { + "epoch": 0.88, + "learning_rate": 7.198286358629747e-07, + "loss": 0.2803, + "step": 14602 + }, + { + "epoch": 0.88, + "learning_rate": 7.190998890618561e-07, + "loss": 0.2699, + "step": 14603 + }, + { + "epoch": 0.88, + "learning_rate": 7.183714975770295e-07, + "loss": 0.2582, + "step": 14604 + }, + { + "epoch": 0.88, + "learning_rate": 7.176434614363837e-07, + "loss": 0.2775, + "step": 14605 + }, + { + "epoch": 0.88, + "learning_rate": 7.169157806677918e-07, + "loss": 0.2637, + "step": 14606 + }, + { + "epoch": 0.88, + "learning_rate": 7.161884552991105e-07, + "loss": 0.2662, + "step": 14607 + }, + { + "epoch": 0.88, + "learning_rate": 7.154614853581876e-07, + "loss": 0.252, + "step": 14608 + }, + { + "epoch": 0.88, + "learning_rate": 7.147348708728508e-07, + "loss": 0.2613, + "step": 14609 + }, + { + "epoch": 0.88, + "learning_rate": 7.140086118709233e-07, + "loss": 0.2479, + "step": 14610 + }, + { + "epoch": 0.88, + "learning_rate": 7.132827083802074e-07, + "loss": 0.2666, + "step": 14611 + }, + { + "epoch": 0.88, + "learning_rate": 7.125571604284953e-07, + "loss": 0.281, + "step": 14612 + }, + { + "epoch": 0.88, + "learning_rate": 7.118319680435615e-07, + "loss": 0.2552, + "step": 14613 + }, + { + "epoch": 0.88, + "learning_rate": 7.111071312531715e-07, + "loss": 0.2448, + "step": 14614 + }, + { + "epoch": 0.88, + "learning_rate": 7.103826500850764e-07, + "loss": 0.2395, + "step": 14615 + }, + { + "epoch": 0.88, + "learning_rate": 7.096585245670118e-07, + "loss": 0.2501, + "step": 14616 + }, + { + "epoch": 0.88, + "learning_rate": 7.089347547267012e-07, + "loss": 0.2508, + "step": 14617 + }, + { + "epoch": 0.88, + "learning_rate": 7.082113405918523e-07, + "loss": 0.267, + "step": 14618 + }, + { + "epoch": 0.88, + "learning_rate": 7.074882821901608e-07, + "loss": 0.2755, + "step": 14619 + }, + { + "epoch": 0.88, + "learning_rate": 7.067655795493112e-07, + "loss": 0.248, + "step": 14620 + }, + { + "epoch": 0.88, + "learning_rate": 7.060432326969713e-07, + "loss": 0.2924, + "step": 14621 + }, + { + "epoch": 0.88, + "learning_rate": 7.053212416607947e-07, + "loss": 0.2609, + "step": 14622 + }, + { + "epoch": 0.88, + "learning_rate": 7.045996064684224e-07, + "loss": 0.2546, + "step": 14623 + }, + { + "epoch": 0.88, + "learning_rate": 7.038783271474813e-07, + "loss": 0.2432, + "step": 14624 + }, + { + "epoch": 0.88, + "learning_rate": 7.031574037255884e-07, + "loss": 0.2785, + "step": 14625 + }, + { + "epoch": 0.88, + "learning_rate": 7.024368362303424e-07, + "loss": 0.2599, + "step": 14626 + }, + { + "epoch": 0.88, + "learning_rate": 7.017166246893292e-07, + "loss": 0.2494, + "step": 14627 + }, + { + "epoch": 0.88, + "learning_rate": 7.009967691301234e-07, + "loss": 0.261, + "step": 14628 + }, + { + "epoch": 0.88, + "learning_rate": 7.002772695802817e-07, + "loss": 0.2494, + "step": 14629 + }, + { + "epoch": 0.88, + "learning_rate": 6.995581260673534e-07, + "loss": 0.2404, + "step": 14630 + }, + { + "epoch": 0.88, + "learning_rate": 6.988393386188697e-07, + "loss": 0.2606, + "step": 14631 + }, + { + "epoch": 0.88, + "learning_rate": 6.981209072623485e-07, + "loss": 0.2794, + "step": 14632 + }, + { + "epoch": 0.88, + "learning_rate": 6.974028320252935e-07, + "loss": 0.2563, + "step": 14633 + }, + { + "epoch": 0.88, + "learning_rate": 6.96685112935197e-07, + "loss": 0.3133, + "step": 14634 + }, + { + "epoch": 0.88, + "learning_rate": 6.959677500195383e-07, + "loss": 0.2596, + "step": 14635 + }, + { + "epoch": 0.88, + "learning_rate": 6.952507433057787e-07, + "loss": 0.248, + "step": 14636 + }, + { + "epoch": 0.88, + "learning_rate": 6.945340928213706e-07, + "loss": 0.2553, + "step": 14637 + }, + { + "epoch": 0.88, + "learning_rate": 6.9381779859375e-07, + "loss": 0.2822, + "step": 14638 + }, + { + "epoch": 0.88, + "learning_rate": 6.931018606503381e-07, + "loss": 0.297, + "step": 14639 + }, + { + "epoch": 0.88, + "learning_rate": 6.923862790185476e-07, + "loss": 0.2633, + "step": 14640 + }, + { + "epoch": 0.88, + "learning_rate": 6.916710537257743e-07, + "loss": 0.2871, + "step": 14641 + }, + { + "epoch": 0.88, + "learning_rate": 6.909561847993973e-07, + "loss": 0.2877, + "step": 14642 + }, + { + "epoch": 0.88, + "learning_rate": 6.902416722667882e-07, + "loss": 0.2762, + "step": 14643 + }, + { + "epoch": 0.88, + "learning_rate": 6.895275161552983e-07, + "loss": 0.2919, + "step": 14644 + }, + { + "epoch": 0.88, + "learning_rate": 6.888137164922725e-07, + "loss": 0.2626, + "step": 14645 + }, + { + "epoch": 0.88, + "learning_rate": 6.881002733050369e-07, + "loss": 0.2748, + "step": 14646 + }, + { + "epoch": 0.88, + "learning_rate": 6.873871866209058e-07, + "loss": 0.2646, + "step": 14647 + }, + { + "epoch": 0.88, + "learning_rate": 6.86674456467179e-07, + "loss": 0.254, + "step": 14648 + }, + { + "epoch": 0.88, + "learning_rate": 6.859620828711422e-07, + "loss": 0.2638, + "step": 14649 + }, + { + "epoch": 0.89, + "learning_rate": 6.852500658600702e-07, + "loss": 0.256, + "step": 14650 + }, + { + "epoch": 0.89, + "learning_rate": 6.845384054612225e-07, + "loss": 0.2499, + "step": 14651 + }, + { + "epoch": 0.89, + "learning_rate": 6.838271017018438e-07, + "loss": 0.2556, + "step": 14652 + }, + { + "epoch": 0.89, + "learning_rate": 6.831161546091658e-07, + "loss": 0.24, + "step": 14653 + }, + { + "epoch": 0.89, + "learning_rate": 6.824055642104066e-07, + "loss": 0.28, + "step": 14654 + }, + { + "epoch": 0.89, + "learning_rate": 6.816953305327733e-07, + "loss": 0.2754, + "step": 14655 + }, + { + "epoch": 0.89, + "learning_rate": 6.809854536034543e-07, + "loss": 0.2782, + "step": 14656 + }, + { + "epoch": 0.89, + "learning_rate": 6.802759334496289e-07, + "loss": 0.2478, + "step": 14657 + }, + { + "epoch": 0.89, + "learning_rate": 6.795667700984598e-07, + "loss": 0.2713, + "step": 14658 + }, + { + "epoch": 0.89, + "learning_rate": 6.788579635770964e-07, + "loss": 0.2627, + "step": 14659 + }, + { + "epoch": 0.89, + "learning_rate": 6.78149513912677e-07, + "loss": 0.2543, + "step": 14660 + }, + { + "epoch": 0.89, + "learning_rate": 6.774414211323243e-07, + "loss": 0.2383, + "step": 14661 + }, + { + "epoch": 0.89, + "learning_rate": 6.767336852631456e-07, + "loss": 0.2513, + "step": 14662 + }, + { + "epoch": 0.89, + "learning_rate": 6.76026306332237e-07, + "loss": 0.2603, + "step": 14663 + }, + { + "epoch": 0.89, + "learning_rate": 6.7531928436668e-07, + "loss": 0.2956, + "step": 14664 + }, + { + "epoch": 0.89, + "learning_rate": 6.746126193935443e-07, + "loss": 0.2917, + "step": 14665 + }, + { + "epoch": 0.89, + "learning_rate": 6.739063114398859e-07, + "loss": 0.2686, + "step": 14666 + }, + { + "epoch": 0.89, + "learning_rate": 6.732003605327398e-07, + "loss": 0.2825, + "step": 14667 + }, + { + "epoch": 0.89, + "learning_rate": 6.724947666991355e-07, + "loss": 0.2713, + "step": 14668 + }, + { + "epoch": 0.89, + "learning_rate": 6.717895299660892e-07, + "loss": 0.2747, + "step": 14669 + }, + { + "epoch": 0.89, + "learning_rate": 6.710846503605984e-07, + "loss": 0.2821, + "step": 14670 + }, + { + "epoch": 0.89, + "learning_rate": 6.703801279096489e-07, + "loss": 0.2706, + "step": 14671 + }, + { + "epoch": 0.89, + "learning_rate": 6.696759626402138e-07, + "loss": 0.2676, + "step": 14672 + }, + { + "epoch": 0.89, + "learning_rate": 6.689721545792527e-07, + "loss": 0.285, + "step": 14673 + }, + { + "epoch": 0.89, + "learning_rate": 6.682687037537073e-07, + "loss": 0.261, + "step": 14674 + }, + { + "epoch": 0.89, + "learning_rate": 6.675656101905136e-07, + "loss": 0.2345, + "step": 14675 + }, + { + "epoch": 0.89, + "learning_rate": 6.668628739165861e-07, + "loss": 0.2785, + "step": 14676 + }, + { + "epoch": 0.89, + "learning_rate": 6.661604949588307e-07, + "loss": 0.2497, + "step": 14677 + }, + { + "epoch": 0.89, + "learning_rate": 6.654584733441371e-07, + "loss": 0.2536, + "step": 14678 + }, + { + "epoch": 0.89, + "learning_rate": 6.647568090993794e-07, + "loss": 0.2472, + "step": 14679 + }, + { + "epoch": 0.89, + "learning_rate": 6.64055502251425e-07, + "loss": 0.2681, + "step": 14680 + }, + { + "epoch": 0.89, + "learning_rate": 6.633545528271213e-07, + "loss": 0.258, + "step": 14681 + }, + { + "epoch": 0.89, + "learning_rate": 6.626539608533033e-07, + "loss": 0.2567, + "step": 14682 + }, + { + "epoch": 0.89, + "learning_rate": 6.619537263567932e-07, + "loss": 0.2546, + "step": 14683 + }, + { + "epoch": 0.89, + "learning_rate": 6.612538493643994e-07, + "loss": 0.2638, + "step": 14684 + }, + { + "epoch": 0.89, + "learning_rate": 6.605543299029171e-07, + "loss": 0.2793, + "step": 14685 + }, + { + "epoch": 0.89, + "learning_rate": 6.598551679991261e-07, + "loss": 0.2621, + "step": 14686 + }, + { + "epoch": 0.89, + "learning_rate": 6.591563636797948e-07, + "loss": 0.2667, + "step": 14687 + }, + { + "epoch": 0.89, + "learning_rate": 6.584579169716765e-07, + "loss": 0.2812, + "step": 14688 + }, + { + "epoch": 0.89, + "learning_rate": 6.577598279015085e-07, + "loss": 0.2818, + "step": 14689 + }, + { + "epoch": 0.89, + "learning_rate": 6.570620964960206e-07, + "loss": 0.2554, + "step": 14690 + }, + { + "epoch": 0.89, + "learning_rate": 6.563647227819225e-07, + "loss": 0.2651, + "step": 14691 + }, + { + "epoch": 0.89, + "learning_rate": 6.556677067859152e-07, + "loss": 0.2585, + "step": 14692 + }, + { + "epoch": 0.89, + "learning_rate": 6.549710485346828e-07, + "loss": 0.2723, + "step": 14693 + }, + { + "epoch": 0.89, + "learning_rate": 6.542747480548939e-07, + "loss": 0.2553, + "step": 14694 + }, + { + "epoch": 0.89, + "learning_rate": 6.535788053732106e-07, + "loss": 0.2868, + "step": 14695 + }, + { + "epoch": 0.89, + "learning_rate": 6.52883220516275e-07, + "loss": 0.3015, + "step": 14696 + }, + { + "epoch": 0.89, + "learning_rate": 6.521879935107168e-07, + "loss": 0.2771, + "step": 14697 + }, + { + "epoch": 0.89, + "learning_rate": 6.514931243831535e-07, + "loss": 0.2808, + "step": 14698 + }, + { + "epoch": 0.89, + "learning_rate": 6.507986131601851e-07, + "loss": 0.2719, + "step": 14699 + }, + { + "epoch": 0.89, + "learning_rate": 6.50104459868407e-07, + "loss": 0.2863, + "step": 14700 + }, + { + "epoch": 0.89, + "learning_rate": 6.49410664534389e-07, + "loss": 0.2766, + "step": 14701 + }, + { + "epoch": 0.89, + "learning_rate": 6.487172271846942e-07, + "loss": 0.2414, + "step": 14702 + }, + { + "epoch": 0.89, + "learning_rate": 6.480241478458693e-07, + "loss": 0.2494, + "step": 14703 + }, + { + "epoch": 0.89, + "learning_rate": 6.473314265444519e-07, + "loss": 0.2423, + "step": 14704 + }, + { + "epoch": 0.89, + "learning_rate": 6.466390633069608e-07, + "loss": 0.2598, + "step": 14705 + }, + { + "epoch": 0.89, + "learning_rate": 6.459470581599037e-07, + "loss": 0.2679, + "step": 14706 + }, + { + "epoch": 0.89, + "learning_rate": 6.452554111297727e-07, + "loss": 0.254, + "step": 14707 + }, + { + "epoch": 0.89, + "learning_rate": 6.445641222430477e-07, + "loss": 0.2952, + "step": 14708 + }, + { + "epoch": 0.89, + "learning_rate": 6.438731915261931e-07, + "loss": 0.269, + "step": 14709 + }, + { + "epoch": 0.89, + "learning_rate": 6.431826190056634e-07, + "loss": 0.2637, + "step": 14710 + }, + { + "epoch": 0.89, + "learning_rate": 6.424924047078962e-07, + "loss": 0.2595, + "step": 14711 + }, + { + "epoch": 0.89, + "learning_rate": 6.41802548659316e-07, + "loss": 0.257, + "step": 14712 + }, + { + "epoch": 0.89, + "learning_rate": 6.411130508863339e-07, + "loss": 0.2539, + "step": 14713 + }, + { + "epoch": 0.89, + "learning_rate": 6.404239114153454e-07, + "loss": 0.3107, + "step": 14714 + }, + { + "epoch": 0.89, + "learning_rate": 6.397351302727362e-07, + "loss": 0.2819, + "step": 14715 + }, + { + "epoch": 0.89, + "learning_rate": 6.39046707484876e-07, + "loss": 0.2446, + "step": 14716 + }, + { + "epoch": 0.89, + "learning_rate": 6.383586430781196e-07, + "loss": 0.2515, + "step": 14717 + }, + { + "epoch": 0.89, + "learning_rate": 6.376709370788092e-07, + "loss": 0.26, + "step": 14718 + }, + { + "epoch": 0.89, + "learning_rate": 6.369835895132737e-07, + "loss": 0.2501, + "step": 14719 + }, + { + "epoch": 0.89, + "learning_rate": 6.362966004078297e-07, + "loss": 0.2697, + "step": 14720 + }, + { + "epoch": 0.89, + "learning_rate": 6.356099697887763e-07, + "loss": 0.2441, + "step": 14721 + }, + { + "epoch": 0.89, + "learning_rate": 6.349236976824013e-07, + "loss": 0.256, + "step": 14722 + }, + { + "epoch": 0.89, + "learning_rate": 6.342377841149794e-07, + "loss": 0.2921, + "step": 14723 + }, + { + "epoch": 0.89, + "learning_rate": 6.335522291127682e-07, + "loss": 0.2665, + "step": 14724 + }, + { + "epoch": 0.89, + "learning_rate": 6.328670327020169e-07, + "loss": 0.238, + "step": 14725 + }, + { + "epoch": 0.89, + "learning_rate": 6.321821949089568e-07, + "loss": 0.2839, + "step": 14726 + }, + { + "epoch": 0.89, + "learning_rate": 6.314977157598079e-07, + "loss": 0.2723, + "step": 14727 + }, + { + "epoch": 0.89, + "learning_rate": 6.308135952807726e-07, + "loss": 0.2651, + "step": 14728 + }, + { + "epoch": 0.89, + "learning_rate": 6.301298334980422e-07, + "loss": 0.2952, + "step": 14729 + }, + { + "epoch": 0.89, + "learning_rate": 6.294464304377978e-07, + "loss": 0.2517, + "step": 14730 + }, + { + "epoch": 0.89, + "learning_rate": 6.28763386126201e-07, + "loss": 0.2812, + "step": 14731 + }, + { + "epoch": 0.89, + "learning_rate": 6.280807005894019e-07, + "loss": 0.269, + "step": 14732 + }, + { + "epoch": 0.89, + "learning_rate": 6.273983738535383e-07, + "loss": 0.2824, + "step": 14733 + }, + { + "epoch": 0.89, + "learning_rate": 6.267164059447305e-07, + "loss": 0.2612, + "step": 14734 + }, + { + "epoch": 0.89, + "learning_rate": 6.26034796889089e-07, + "loss": 0.2402, + "step": 14735 + }, + { + "epoch": 0.89, + "learning_rate": 6.25353546712708e-07, + "loss": 0.2642, + "step": 14736 + }, + { + "epoch": 0.89, + "learning_rate": 6.246726554416704e-07, + "loss": 0.2533, + "step": 14737 + }, + { + "epoch": 0.89, + "learning_rate": 6.239921231020418e-07, + "loss": 0.2408, + "step": 14738 + }, + { + "epoch": 0.89, + "learning_rate": 6.233119497198781e-07, + "loss": 0.3105, + "step": 14739 + }, + { + "epoch": 0.89, + "learning_rate": 6.226321353212194e-07, + "loss": 0.2577, + "step": 14740 + }, + { + "epoch": 0.89, + "learning_rate": 6.219526799320919e-07, + "loss": 0.2549, + "step": 14741 + }, + { + "epoch": 0.89, + "learning_rate": 6.212735835785077e-07, + "loss": 0.2811, + "step": 14742 + }, + { + "epoch": 0.89, + "learning_rate": 6.205948462864642e-07, + "loss": 0.2651, + "step": 14743 + }, + { + "epoch": 0.89, + "learning_rate": 6.199164680819502e-07, + "loss": 0.2709, + "step": 14744 + }, + { + "epoch": 0.89, + "learning_rate": 6.192384489909364e-07, + "loss": 0.2697, + "step": 14745 + }, + { + "epoch": 0.89, + "learning_rate": 6.185607890393785e-07, + "loss": 0.2719, + "step": 14746 + }, + { + "epoch": 0.89, + "learning_rate": 6.178834882532225e-07, + "loss": 0.2646, + "step": 14747 + }, + { + "epoch": 0.89, + "learning_rate": 6.172065466583987e-07, + "loss": 0.2612, + "step": 14748 + }, + { + "epoch": 0.89, + "learning_rate": 6.165299642808198e-07, + "loss": 0.2754, + "step": 14749 + }, + { + "epoch": 0.89, + "learning_rate": 6.158537411463939e-07, + "loss": 0.2559, + "step": 14750 + }, + { + "epoch": 0.89, + "learning_rate": 6.151778772810068e-07, + "loss": 0.2621, + "step": 14751 + }, + { + "epoch": 0.89, + "learning_rate": 6.145023727105348e-07, + "loss": 0.2595, + "step": 14752 + }, + { + "epoch": 0.89, + "learning_rate": 6.138272274608404e-07, + "loss": 0.2649, + "step": 14753 + }, + { + "epoch": 0.89, + "learning_rate": 6.131524415577672e-07, + "loss": 0.2525, + "step": 14754 + }, + { + "epoch": 0.89, + "learning_rate": 6.124780150271537e-07, + "loss": 0.2695, + "step": 14755 + }, + { + "epoch": 0.89, + "learning_rate": 6.118039478948179e-07, + "loss": 0.2757, + "step": 14756 + }, + { + "epoch": 0.89, + "learning_rate": 6.111302401865671e-07, + "loss": 0.2759, + "step": 14757 + }, + { + "epoch": 0.89, + "learning_rate": 6.104568919281939e-07, + "loss": 0.2899, + "step": 14758 + }, + { + "epoch": 0.89, + "learning_rate": 6.097839031454755e-07, + "loss": 0.2575, + "step": 14759 + }, + { + "epoch": 0.89, + "learning_rate": 6.091112738641792e-07, + "loss": 0.2716, + "step": 14760 + }, + { + "epoch": 0.89, + "learning_rate": 6.084390041100563e-07, + "loss": 0.2473, + "step": 14761 + }, + { + "epoch": 0.89, + "learning_rate": 6.077670939088442e-07, + "loss": 0.2838, + "step": 14762 + }, + { + "epoch": 0.89, + "learning_rate": 6.070955432862657e-07, + "loss": 0.2831, + "step": 14763 + }, + { + "epoch": 0.89, + "learning_rate": 6.064243522680302e-07, + "loss": 0.2759, + "step": 14764 + }, + { + "epoch": 0.89, + "learning_rate": 6.057535208798371e-07, + "loss": 0.2511, + "step": 14765 + }, + { + "epoch": 0.89, + "learning_rate": 6.05083049147367e-07, + "loss": 0.2602, + "step": 14766 + }, + { + "epoch": 0.89, + "learning_rate": 6.044129370962892e-07, + "loss": 0.2646, + "step": 14767 + }, + { + "epoch": 0.89, + "learning_rate": 6.037431847522591e-07, + "loss": 0.2863, + "step": 14768 + }, + { + "epoch": 0.89, + "learning_rate": 6.030737921409169e-07, + "loss": 0.2605, + "step": 14769 + }, + { + "epoch": 0.89, + "learning_rate": 6.024047592878901e-07, + "loss": 0.2603, + "step": 14770 + }, + { + "epoch": 0.89, + "learning_rate": 6.017360862187937e-07, + "loss": 0.2806, + "step": 14771 + }, + { + "epoch": 0.89, + "learning_rate": 6.010677729592262e-07, + "loss": 0.299, + "step": 14772 + }, + { + "epoch": 0.89, + "learning_rate": 6.003998195347726e-07, + "loss": 0.2598, + "step": 14773 + }, + { + "epoch": 0.89, + "learning_rate": 5.997322259710092e-07, + "loss": 0.2575, + "step": 14774 + }, + { + "epoch": 0.89, + "learning_rate": 5.990649922934921e-07, + "loss": 0.2609, + "step": 14775 + }, + { + "epoch": 0.89, + "learning_rate": 5.983981185277665e-07, + "loss": 0.2754, + "step": 14776 + }, + { + "epoch": 0.89, + "learning_rate": 5.977316046993642e-07, + "loss": 0.2771, + "step": 14777 + }, + { + "epoch": 0.89, + "learning_rate": 5.970654508337992e-07, + "loss": 0.2566, + "step": 14778 + }, + { + "epoch": 0.89, + "learning_rate": 5.963996569565789e-07, + "loss": 0.2525, + "step": 14779 + }, + { + "epoch": 0.89, + "learning_rate": 5.957342230931918e-07, + "loss": 0.2794, + "step": 14780 + }, + { + "epoch": 0.89, + "learning_rate": 5.95069149269114e-07, + "loss": 0.2292, + "step": 14781 + }, + { + "epoch": 0.89, + "learning_rate": 5.944044355098067e-07, + "loss": 0.2588, + "step": 14782 + }, + { + "epoch": 0.89, + "learning_rate": 5.937400818407191e-07, + "loss": 0.274, + "step": 14783 + }, + { + "epoch": 0.89, + "learning_rate": 5.930760882872843e-07, + "loss": 0.2545, + "step": 14784 + }, + { + "epoch": 0.89, + "learning_rate": 5.924124548749255e-07, + "loss": 0.2746, + "step": 14785 + }, + { + "epoch": 0.89, + "learning_rate": 5.917491816290477e-07, + "loss": 0.2606, + "step": 14786 + }, + { + "epoch": 0.89, + "learning_rate": 5.910862685750462e-07, + "loss": 0.2606, + "step": 14787 + }, + { + "epoch": 0.89, + "learning_rate": 5.904237157382975e-07, + "loss": 0.2482, + "step": 14788 + }, + { + "epoch": 0.89, + "learning_rate": 5.89761523144169e-07, + "loss": 0.2552, + "step": 14789 + }, + { + "epoch": 0.89, + "learning_rate": 5.890996908180125e-07, + "loss": 0.2649, + "step": 14790 + }, + { + "epoch": 0.89, + "learning_rate": 5.884382187851667e-07, + "loss": 0.2695, + "step": 14791 + }, + { + "epoch": 0.89, + "learning_rate": 5.877771070709537e-07, + "loss": 0.2527, + "step": 14792 + }, + { + "epoch": 0.89, + "learning_rate": 5.871163557006865e-07, + "loss": 0.245, + "step": 14793 + }, + { + "epoch": 0.89, + "learning_rate": 5.864559646996582e-07, + "loss": 0.2878, + "step": 14794 + }, + { + "epoch": 0.89, + "learning_rate": 5.857959340931563e-07, + "loss": 0.2691, + "step": 14795 + }, + { + "epoch": 0.89, + "learning_rate": 5.851362639064473e-07, + "loss": 0.2736, + "step": 14796 + }, + { + "epoch": 0.89, + "learning_rate": 5.844769541647866e-07, + "loss": 0.2603, + "step": 14797 + }, + { + "epoch": 0.89, + "learning_rate": 5.83818004893415e-07, + "loss": 0.2487, + "step": 14798 + }, + { + "epoch": 0.89, + "learning_rate": 5.831594161175602e-07, + "loss": 0.2659, + "step": 14799 + }, + { + "epoch": 0.89, + "learning_rate": 5.825011878624387e-07, + "loss": 0.2489, + "step": 14800 + }, + { + "epoch": 0.89, + "learning_rate": 5.818433201532481e-07, + "loss": 0.2729, + "step": 14801 + }, + { + "epoch": 0.89, + "learning_rate": 5.81185813015176e-07, + "loss": 0.2573, + "step": 14802 + }, + { + "epoch": 0.89, + "learning_rate": 5.805286664733922e-07, + "loss": 0.2796, + "step": 14803 + }, + { + "epoch": 0.89, + "learning_rate": 5.798718805530579e-07, + "loss": 0.2662, + "step": 14804 + }, + { + "epoch": 0.89, + "learning_rate": 5.792154552793172e-07, + "loss": 0.2751, + "step": 14805 + }, + { + "epoch": 0.89, + "learning_rate": 5.785593906773012e-07, + "loss": 0.2475, + "step": 14806 + }, + { + "epoch": 0.89, + "learning_rate": 5.779036867721277e-07, + "loss": 0.304, + "step": 14807 + }, + { + "epoch": 0.89, + "learning_rate": 5.772483435888976e-07, + "loss": 0.2906, + "step": 14808 + }, + { + "epoch": 0.89, + "learning_rate": 5.765933611527041e-07, + "loss": 0.2577, + "step": 14809 + }, + { + "epoch": 0.89, + "learning_rate": 5.759387394886207e-07, + "loss": 0.266, + "step": 14810 + }, + { + "epoch": 0.89, + "learning_rate": 5.752844786217105e-07, + "loss": 0.286, + "step": 14811 + }, + { + "epoch": 0.89, + "learning_rate": 5.746305785770212e-07, + "loss": 0.2766, + "step": 14812 + }, + { + "epoch": 0.89, + "learning_rate": 5.739770393795852e-07, + "loss": 0.2637, + "step": 14813 + }, + { + "epoch": 0.89, + "learning_rate": 5.733238610544268e-07, + "loss": 0.2751, + "step": 14814 + }, + { + "epoch": 0.9, + "learning_rate": 5.726710436265503e-07, + "loss": 0.2836, + "step": 14815 + }, + { + "epoch": 0.9, + "learning_rate": 5.720185871209494e-07, + "loss": 0.2596, + "step": 14816 + }, + { + "epoch": 0.9, + "learning_rate": 5.713664915626027e-07, + "loss": 0.238, + "step": 14817 + }, + { + "epoch": 0.9, + "learning_rate": 5.707147569764759e-07, + "loss": 0.264, + "step": 14818 + }, + { + "epoch": 0.9, + "learning_rate": 5.700633833875191e-07, + "loss": 0.2699, + "step": 14819 + }, + { + "epoch": 0.9, + "learning_rate": 5.694123708206734e-07, + "loss": 0.2706, + "step": 14820 + }, + { + "epoch": 0.9, + "learning_rate": 5.687617193008588e-07, + "loss": 0.2584, + "step": 14821 + }, + { + "epoch": 0.9, + "learning_rate": 5.681114288529877e-07, + "loss": 0.2561, + "step": 14822 + }, + { + "epoch": 0.9, + "learning_rate": 5.674614995019545e-07, + "loss": 0.2822, + "step": 14823 + }, + { + "epoch": 0.9, + "learning_rate": 5.668119312726417e-07, + "loss": 0.2614, + "step": 14824 + }, + { + "epoch": 0.9, + "learning_rate": 5.661627241899193e-07, + "loss": 0.279, + "step": 14825 + }, + { + "epoch": 0.9, + "learning_rate": 5.655138782786418e-07, + "loss": 0.2631, + "step": 14826 + }, + { + "epoch": 0.9, + "learning_rate": 5.648653935636494e-07, + "loss": 0.2472, + "step": 14827 + }, + { + "epoch": 0.9, + "learning_rate": 5.64217270069769e-07, + "loss": 0.287, + "step": 14828 + }, + { + "epoch": 0.9, + "learning_rate": 5.635695078218117e-07, + "loss": 0.251, + "step": 14829 + }, + { + "epoch": 0.9, + "learning_rate": 5.629221068445822e-07, + "loss": 0.2744, + "step": 14830 + }, + { + "epoch": 0.9, + "learning_rate": 5.622750671628619e-07, + "loss": 0.2298, + "step": 14831 + }, + { + "epoch": 0.9, + "learning_rate": 5.61628388801424e-07, + "loss": 0.2727, + "step": 14832 + }, + { + "epoch": 0.9, + "learning_rate": 5.609820717850257e-07, + "loss": 0.2899, + "step": 14833 + }, + { + "epoch": 0.9, + "learning_rate": 5.603361161384103e-07, + "loss": 0.2747, + "step": 14834 + }, + { + "epoch": 0.9, + "learning_rate": 5.596905218863092e-07, + "loss": 0.26, + "step": 14835 + }, + { + "epoch": 0.9, + "learning_rate": 5.590452890534415e-07, + "loss": 0.2584, + "step": 14836 + }, + { + "epoch": 0.9, + "learning_rate": 5.584004176645052e-07, + "loss": 0.2886, + "step": 14837 + }, + { + "epoch": 0.9, + "learning_rate": 5.577559077441886e-07, + "loss": 0.2765, + "step": 14838 + }, + { + "epoch": 0.9, + "learning_rate": 5.571117593171704e-07, + "loss": 0.2527, + "step": 14839 + }, + { + "epoch": 0.9, + "learning_rate": 5.56467972408109e-07, + "loss": 0.267, + "step": 14840 + }, + { + "epoch": 0.9, + "learning_rate": 5.558245470416534e-07, + "loss": 0.2718, + "step": 14841 + }, + { + "epoch": 0.9, + "learning_rate": 5.551814832424352e-07, + "loss": 0.2684, + "step": 14842 + }, + { + "epoch": 0.9, + "learning_rate": 5.545387810350733e-07, + "loss": 0.2776, + "step": 14843 + }, + { + "epoch": 0.9, + "learning_rate": 5.53896440444176e-07, + "loss": 0.2429, + "step": 14844 + }, + { + "epoch": 0.9, + "learning_rate": 5.532544614943347e-07, + "loss": 0.2839, + "step": 14845 + }, + { + "epoch": 0.9, + "learning_rate": 5.526128442101253e-07, + "loss": 0.2573, + "step": 14846 + }, + { + "epoch": 0.9, + "learning_rate": 5.519715886161137e-07, + "loss": 0.2863, + "step": 14847 + }, + { + "epoch": 0.9, + "learning_rate": 5.513306947368469e-07, + "loss": 0.2513, + "step": 14848 + }, + { + "epoch": 0.9, + "learning_rate": 5.506901625968664e-07, + "loss": 0.267, + "step": 14849 + }, + { + "epoch": 0.9, + "learning_rate": 5.500499922206925e-07, + "loss": 0.2567, + "step": 14850 + }, + { + "epoch": 0.9, + "learning_rate": 5.494101836328325e-07, + "loss": 0.2367, + "step": 14851 + }, + { + "epoch": 0.9, + "learning_rate": 5.487707368577833e-07, + "loss": 0.2845, + "step": 14852 + }, + { + "epoch": 0.9, + "learning_rate": 5.481316519200253e-07, + "loss": 0.2628, + "step": 14853 + }, + { + "epoch": 0.9, + "learning_rate": 5.474929288440234e-07, + "loss": 0.256, + "step": 14854 + }, + { + "epoch": 0.9, + "learning_rate": 5.468545676542347e-07, + "loss": 0.2521, + "step": 14855 + }, + { + "epoch": 0.9, + "learning_rate": 5.462165683750964e-07, + "loss": 0.2554, + "step": 14856 + }, + { + "epoch": 0.9, + "learning_rate": 5.455789310310345e-07, + "loss": 0.268, + "step": 14857 + }, + { + "epoch": 0.9, + "learning_rate": 5.449416556464616e-07, + "loss": 0.256, + "step": 14858 + }, + { + "epoch": 0.9, + "learning_rate": 5.443047422457726e-07, + "loss": 0.2674, + "step": 14859 + }, + { + "epoch": 0.9, + "learning_rate": 5.436681908533558e-07, + "loss": 0.2642, + "step": 14860 + }, + { + "epoch": 0.9, + "learning_rate": 5.430320014935797e-07, + "loss": 0.2734, + "step": 14861 + }, + { + "epoch": 0.9, + "learning_rate": 5.42396174190799e-07, + "loss": 0.2698, + "step": 14862 + }, + { + "epoch": 0.9, + "learning_rate": 5.417607089693578e-07, + "loss": 0.2447, + "step": 14863 + }, + { + "epoch": 0.9, + "learning_rate": 5.41125605853583e-07, + "loss": 0.2433, + "step": 14864 + }, + { + "epoch": 0.9, + "learning_rate": 5.40490864867792e-07, + "loss": 0.2552, + "step": 14865 + }, + { + "epoch": 0.9, + "learning_rate": 5.398564860362843e-07, + "loss": 0.2891, + "step": 14866 + }, + { + "epoch": 0.9, + "learning_rate": 5.39222469383347e-07, + "loss": 0.2741, + "step": 14867 + }, + { + "epoch": 0.9, + "learning_rate": 5.38588814933253e-07, + "loss": 0.262, + "step": 14868 + }, + { + "epoch": 0.9, + "learning_rate": 5.379555227102596e-07, + "loss": 0.2397, + "step": 14869 + }, + { + "epoch": 0.9, + "learning_rate": 5.373225927386172e-07, + "loss": 0.2567, + "step": 14870 + }, + { + "epoch": 0.9, + "learning_rate": 5.366900250425532e-07, + "loss": 0.256, + "step": 14871 + }, + { + "epoch": 0.9, + "learning_rate": 5.360578196462863e-07, + "loss": 0.255, + "step": 14872 + }, + { + "epoch": 0.9, + "learning_rate": 5.354259765740177e-07, + "loss": 0.2476, + "step": 14873 + }, + { + "epoch": 0.9, + "learning_rate": 5.347944958499429e-07, + "loss": 0.251, + "step": 14874 + }, + { + "epoch": 0.9, + "learning_rate": 5.341633774982347e-07, + "loss": 0.2582, + "step": 14875 + }, + { + "epoch": 0.9, + "learning_rate": 5.335326215430548e-07, + "loss": 0.2281, + "step": 14876 + }, + { + "epoch": 0.9, + "learning_rate": 5.329022280085528e-07, + "loss": 0.2593, + "step": 14877 + }, + { + "epoch": 0.9, + "learning_rate": 5.322721969188616e-07, + "loss": 0.2688, + "step": 14878 + }, + { + "epoch": 0.9, + "learning_rate": 5.31642528298103e-07, + "loss": 0.2619, + "step": 14879 + }, + { + "epoch": 0.9, + "learning_rate": 5.310132221703846e-07, + "loss": 0.273, + "step": 14880 + }, + { + "epoch": 0.9, + "learning_rate": 5.303842785597968e-07, + "loss": 0.2709, + "step": 14881 + }, + { + "epoch": 0.9, + "learning_rate": 5.297556974904205e-07, + "loss": 0.2946, + "step": 14882 + }, + { + "epoch": 0.9, + "learning_rate": 5.291274789863177e-07, + "loss": 0.2372, + "step": 14883 + }, + { + "epoch": 0.9, + "learning_rate": 5.284996230715434e-07, + "loss": 0.2454, + "step": 14884 + }, + { + "epoch": 0.9, + "learning_rate": 5.278721297701339e-07, + "loss": 0.2912, + "step": 14885 + }, + { + "epoch": 0.9, + "learning_rate": 5.272449991061113e-07, + "loss": 0.2347, + "step": 14886 + }, + { + "epoch": 0.9, + "learning_rate": 5.266182311034851e-07, + "loss": 0.2688, + "step": 14887 + }, + { + "epoch": 0.9, + "learning_rate": 5.259918257862518e-07, + "loss": 0.2706, + "step": 14888 + }, + { + "epoch": 0.9, + "learning_rate": 5.253657831783921e-07, + "loss": 0.2464, + "step": 14889 + }, + { + "epoch": 0.9, + "learning_rate": 5.247401033038757e-07, + "loss": 0.2655, + "step": 14890 + }, + { + "epoch": 0.9, + "learning_rate": 5.241147861866547e-07, + "loss": 0.2486, + "step": 14891 + }, + { + "epoch": 0.9, + "learning_rate": 5.234898318506698e-07, + "loss": 0.2882, + "step": 14892 + }, + { + "epoch": 0.9, + "learning_rate": 5.228652403198475e-07, + "loss": 0.275, + "step": 14893 + }, + { + "epoch": 0.9, + "learning_rate": 5.222410116180987e-07, + "loss": 0.2667, + "step": 14894 + }, + { + "epoch": 0.9, + "learning_rate": 5.216171457693242e-07, + "loss": 0.2253, + "step": 14895 + }, + { + "epoch": 0.9, + "learning_rate": 5.209936427974072e-07, + "loss": 0.2454, + "step": 14896 + }, + { + "epoch": 0.9, + "learning_rate": 5.203705027262185e-07, + "loss": 0.2832, + "step": 14897 + }, + { + "epoch": 0.9, + "learning_rate": 5.197477255796146e-07, + "loss": 0.2677, + "step": 14898 + }, + { + "epoch": 0.9, + "learning_rate": 5.191253113814366e-07, + "loss": 0.2532, + "step": 14899 + }, + { + "epoch": 0.9, + "learning_rate": 5.185032601555173e-07, + "loss": 0.2987, + "step": 14900 + }, + { + "epoch": 0.9, + "learning_rate": 5.178815719256692e-07, + "loss": 0.2854, + "step": 14901 + }, + { + "epoch": 0.9, + "learning_rate": 5.172602467156939e-07, + "loss": 0.2556, + "step": 14902 + }, + { + "epoch": 0.9, + "learning_rate": 5.166392845493784e-07, + "loss": 0.2684, + "step": 14903 + }, + { + "epoch": 0.9, + "learning_rate": 5.160186854504956e-07, + "loss": 0.2659, + "step": 14904 + }, + { + "epoch": 0.9, + "learning_rate": 5.153984494428065e-07, + "loss": 0.2628, + "step": 14905 + }, + { + "epoch": 0.9, + "learning_rate": 5.147785765500557e-07, + "loss": 0.2595, + "step": 14906 + }, + { + "epoch": 0.9, + "learning_rate": 5.14159066795975e-07, + "loss": 0.2743, + "step": 14907 + }, + { + "epoch": 0.9, + "learning_rate": 5.135399202042801e-07, + "loss": 0.2727, + "step": 14908 + }, + { + "epoch": 0.9, + "learning_rate": 5.129211367986786e-07, + "loss": 0.2635, + "step": 14909 + }, + { + "epoch": 0.9, + "learning_rate": 5.123027166028583e-07, + "loss": 0.2817, + "step": 14910 + }, + { + "epoch": 0.9, + "learning_rate": 5.116846596404957e-07, + "loss": 0.2501, + "step": 14911 + }, + { + "epoch": 0.9, + "learning_rate": 5.110669659352518e-07, + "loss": 0.2624, + "step": 14912 + }, + { + "epoch": 0.9, + "learning_rate": 5.104496355107758e-07, + "loss": 0.2567, + "step": 14913 + }, + { + "epoch": 0.9, + "learning_rate": 5.098326683907018e-07, + "loss": 0.2533, + "step": 14914 + }, + { + "epoch": 0.9, + "learning_rate": 5.09216064598651e-07, + "loss": 0.2504, + "step": 14915 + }, + { + "epoch": 0.9, + "learning_rate": 5.08599824158229e-07, + "loss": 0.287, + "step": 14916 + }, + { + "epoch": 0.9, + "learning_rate": 5.079839470930281e-07, + "loss": 0.2801, + "step": 14917 + }, + { + "epoch": 0.9, + "learning_rate": 5.07368433426626e-07, + "loss": 0.2857, + "step": 14918 + }, + { + "epoch": 0.9, + "learning_rate": 5.067532831825916e-07, + "loss": 0.2503, + "step": 14919 + }, + { + "epoch": 0.9, + "learning_rate": 5.061384963844718e-07, + "loss": 0.2727, + "step": 14920 + }, + { + "epoch": 0.9, + "learning_rate": 5.055240730558042e-07, + "loss": 0.2517, + "step": 14921 + }, + { + "epoch": 0.9, + "learning_rate": 5.049100132201135e-07, + "loss": 0.2753, + "step": 14922 + }, + { + "epoch": 0.9, + "learning_rate": 5.042963169009063e-07, + "loss": 0.2573, + "step": 14923 + }, + { + "epoch": 0.9, + "learning_rate": 5.036829841216783e-07, + "loss": 0.2614, + "step": 14924 + }, + { + "epoch": 0.9, + "learning_rate": 5.030700149059131e-07, + "loss": 0.2527, + "step": 14925 + }, + { + "epoch": 0.9, + "learning_rate": 5.024574092770762e-07, + "loss": 0.2949, + "step": 14926 + }, + { + "epoch": 0.9, + "learning_rate": 5.018451672586211e-07, + "loss": 0.2688, + "step": 14927 + }, + { + "epoch": 0.9, + "learning_rate": 5.012332888739868e-07, + "loss": 0.2573, + "step": 14928 + }, + { + "epoch": 0.9, + "learning_rate": 5.006217741465991e-07, + "loss": 0.2688, + "step": 14929 + }, + { + "epoch": 0.9, + "learning_rate": 5.000106230998702e-07, + "loss": 0.267, + "step": 14930 + }, + { + "epoch": 0.9, + "learning_rate": 4.993998357571983e-07, + "loss": 0.2363, + "step": 14931 + }, + { + "epoch": 0.9, + "learning_rate": 4.987894121419668e-07, + "loss": 0.2665, + "step": 14932 + }, + { + "epoch": 0.9, + "learning_rate": 4.981793522775458e-07, + "loss": 0.2725, + "step": 14933 + }, + { + "epoch": 0.9, + "learning_rate": 4.975696561872878e-07, + "loss": 0.2486, + "step": 14934 + }, + { + "epoch": 0.9, + "learning_rate": 4.969603238945408e-07, + "loss": 0.268, + "step": 14935 + }, + { + "epoch": 0.9, + "learning_rate": 4.963513554226284e-07, + "loss": 0.255, + "step": 14936 + }, + { + "epoch": 0.9, + "learning_rate": 4.957427507948675e-07, + "loss": 0.2585, + "step": 14937 + }, + { + "epoch": 0.9, + "learning_rate": 4.951345100345562e-07, + "loss": 0.2503, + "step": 14938 + }, + { + "epoch": 0.9, + "learning_rate": 4.945266331649823e-07, + "loss": 0.2684, + "step": 14939 + }, + { + "epoch": 0.9, + "learning_rate": 4.939191202094162e-07, + "loss": 0.2535, + "step": 14940 + }, + { + "epoch": 0.9, + "learning_rate": 4.933119711911194e-07, + "loss": 0.2688, + "step": 14941 + }, + { + "epoch": 0.9, + "learning_rate": 4.927051861333332e-07, + "loss": 0.2744, + "step": 14942 + }, + { + "epoch": 0.9, + "learning_rate": 4.92098765059289e-07, + "loss": 0.2521, + "step": 14943 + }, + { + "epoch": 0.9, + "learning_rate": 4.91492707992206e-07, + "loss": 0.262, + "step": 14944 + }, + { + "epoch": 0.9, + "learning_rate": 4.908870149552836e-07, + "loss": 0.2778, + "step": 14945 + }, + { + "epoch": 0.9, + "learning_rate": 4.90281685971713e-07, + "loss": 0.3009, + "step": 14946 + }, + { + "epoch": 0.9, + "learning_rate": 4.89676721064668e-07, + "loss": 0.2641, + "step": 14947 + }, + { + "epoch": 0.9, + "learning_rate": 4.890721202573078e-07, + "loss": 0.2737, + "step": 14948 + }, + { + "epoch": 0.9, + "learning_rate": 4.884678835727818e-07, + "loss": 0.2699, + "step": 14949 + }, + { + "epoch": 0.9, + "learning_rate": 4.878640110342236e-07, + "loss": 0.2446, + "step": 14950 + }, + { + "epoch": 0.9, + "learning_rate": 4.872605026647504e-07, + "loss": 0.262, + "step": 14951 + }, + { + "epoch": 0.9, + "learning_rate": 4.866573584874668e-07, + "loss": 0.2807, + "step": 14952 + }, + { + "epoch": 0.9, + "learning_rate": 4.860545785254645e-07, + "loss": 0.2814, + "step": 14953 + }, + { + "epoch": 0.9, + "learning_rate": 4.854521628018227e-07, + "loss": 0.2563, + "step": 14954 + }, + { + "epoch": 0.9, + "learning_rate": 4.848501113396031e-07, + "loss": 0.2574, + "step": 14955 + }, + { + "epoch": 0.9, + "learning_rate": 4.84248424161855e-07, + "loss": 0.238, + "step": 14956 + }, + { + "epoch": 0.9, + "learning_rate": 4.836471012916144e-07, + "loss": 0.2758, + "step": 14957 + }, + { + "epoch": 0.9, + "learning_rate": 4.830461427519018e-07, + "loss": 0.2464, + "step": 14958 + }, + { + "epoch": 0.9, + "learning_rate": 4.824455485657242e-07, + "loss": 0.2726, + "step": 14959 + }, + { + "epoch": 0.9, + "learning_rate": 4.818453187560768e-07, + "loss": 0.2704, + "step": 14960 + }, + { + "epoch": 0.9, + "learning_rate": 4.812454533459399e-07, + "loss": 0.2596, + "step": 14961 + }, + { + "epoch": 0.9, + "learning_rate": 4.806459523582762e-07, + "loss": 0.2694, + "step": 14962 + }, + { + "epoch": 0.9, + "learning_rate": 4.800468158160409e-07, + "loss": 0.2746, + "step": 14963 + }, + { + "epoch": 0.9, + "learning_rate": 4.794480437421678e-07, + "loss": 0.2755, + "step": 14964 + }, + { + "epoch": 0.9, + "learning_rate": 4.78849636159584e-07, + "loss": 0.2635, + "step": 14965 + }, + { + "epoch": 0.9, + "learning_rate": 4.782515930911979e-07, + "loss": 0.2614, + "step": 14966 + }, + { + "epoch": 0.9, + "learning_rate": 4.776539145599057e-07, + "loss": 0.2741, + "step": 14967 + }, + { + "epoch": 0.9, + "learning_rate": 4.770566005885901e-07, + "loss": 0.2538, + "step": 14968 + }, + { + "epoch": 0.9, + "learning_rate": 4.7645965120011627e-07, + "loss": 0.2465, + "step": 14969 + }, + { + "epoch": 0.9, + "learning_rate": 4.7586306641734137e-07, + "loss": 0.2674, + "step": 14970 + }, + { + "epoch": 0.9, + "learning_rate": 4.752668462631038e-07, + "loss": 0.2436, + "step": 14971 + }, + { + "epoch": 0.9, + "learning_rate": 4.7467099076023313e-07, + "loss": 0.2465, + "step": 14972 + }, + { + "epoch": 0.9, + "learning_rate": 4.7407549993153443e-07, + "loss": 0.2697, + "step": 14973 + }, + { + "epoch": 0.9, + "learning_rate": 4.734803737998117e-07, + "loss": 0.2592, + "step": 14974 + }, + { + "epoch": 0.9, + "learning_rate": 4.728856123878478e-07, + "loss": 0.2751, + "step": 14975 + }, + { + "epoch": 0.9, + "learning_rate": 4.722912157184123e-07, + "loss": 0.25, + "step": 14976 + }, + { + "epoch": 0.9, + "learning_rate": 4.716971838142614e-07, + "loss": 0.2849, + "step": 14977 + }, + { + "epoch": 0.9, + "learning_rate": 4.71103516698137e-07, + "loss": 0.2669, + "step": 14978 + }, + { + "epoch": 0.9, + "learning_rate": 4.705102143927698e-07, + "loss": 0.2736, + "step": 14979 + }, + { + "epoch": 0.9, + "learning_rate": 4.6991727692087154e-07, + "loss": 0.2797, + "step": 14980 + }, + { + "epoch": 0.91, + "learning_rate": 4.693247043051441e-07, + "loss": 0.2472, + "step": 14981 + }, + { + "epoch": 0.91, + "learning_rate": 4.687324965682749e-07, + "loss": 0.264, + "step": 14982 + }, + { + "epoch": 0.91, + "learning_rate": 4.6814065373293247e-07, + "loss": 0.2779, + "step": 14983 + }, + { + "epoch": 0.91, + "learning_rate": 4.675491758217798e-07, + "loss": 0.2651, + "step": 14984 + }, + { + "epoch": 0.91, + "learning_rate": 4.669580628574588e-07, + "loss": 0.2824, + "step": 14985 + }, + { + "epoch": 0.91, + "learning_rate": 4.663673148626013e-07, + "loss": 0.2684, + "step": 14986 + }, + { + "epoch": 0.91, + "learning_rate": 4.6577693185982374e-07, + "loss": 0.2701, + "step": 14987 + }, + { + "epoch": 0.91, + "learning_rate": 4.651869138717269e-07, + "loss": 0.2711, + "step": 14988 + }, + { + "epoch": 0.91, + "learning_rate": 4.645972609209026e-07, + "loss": 0.2667, + "step": 14989 + }, + { + "epoch": 0.91, + "learning_rate": 4.640079730299241e-07, + "loss": 0.2773, + "step": 14990 + }, + { + "epoch": 0.91, + "learning_rate": 4.6341905022135094e-07, + "loss": 0.2468, + "step": 14991 + }, + { + "epoch": 0.91, + "learning_rate": 4.628304925177318e-07, + "loss": 0.2383, + "step": 14992 + }, + { + "epoch": 0.91, + "learning_rate": 4.622422999415965e-07, + "loss": 0.2935, + "step": 14993 + }, + { + "epoch": 0.91, + "learning_rate": 4.61654472515467e-07, + "loss": 0.2687, + "step": 14994 + }, + { + "epoch": 0.91, + "learning_rate": 4.610670102618464e-07, + "loss": 0.262, + "step": 14995 + }, + { + "epoch": 0.91, + "learning_rate": 4.604799132032267e-07, + "loss": 0.257, + "step": 14996 + }, + { + "epoch": 0.91, + "learning_rate": 4.5989318136208326e-07, + "loss": 0.2859, + "step": 14997 + }, + { + "epoch": 0.91, + "learning_rate": 4.5930681476088036e-07, + "loss": 0.269, + "step": 14998 + }, + { + "epoch": 0.91, + "learning_rate": 4.587208134220644e-07, + "loss": 0.2699, + "step": 14999 + }, + { + "epoch": 0.91, + "learning_rate": 4.581351773680731e-07, + "loss": 0.2702, + "step": 15000 + }, + { + "epoch": 0.91, + "learning_rate": 4.5754990662132624e-07, + "loss": 0.2829, + "step": 15001 + }, + { + "epoch": 0.91, + "learning_rate": 4.5696500120423036e-07, + "loss": 0.2759, + "step": 15002 + }, + { + "epoch": 0.91, + "learning_rate": 4.563804611391787e-07, + "loss": 0.2736, + "step": 15003 + }, + { + "epoch": 0.91, + "learning_rate": 4.557962864485499e-07, + "loss": 0.2698, + "step": 15004 + }, + { + "epoch": 0.91, + "learning_rate": 4.552124771547095e-07, + "loss": 0.2861, + "step": 15005 + }, + { + "epoch": 0.91, + "learning_rate": 4.5462903328000964e-07, + "loss": 0.2705, + "step": 15006 + }, + { + "epoch": 0.91, + "learning_rate": 4.540459548467846e-07, + "loss": 0.2726, + "step": 15007 + }, + { + "epoch": 0.91, + "learning_rate": 4.5346324187735656e-07, + "loss": 0.2726, + "step": 15008 + }, + { + "epoch": 0.91, + "learning_rate": 4.528808943940388e-07, + "loss": 0.2687, + "step": 15009 + }, + { + "epoch": 0.91, + "learning_rate": 4.522989124191235e-07, + "loss": 0.2698, + "step": 15010 + }, + { + "epoch": 0.91, + "learning_rate": 4.517172959748917e-07, + "loss": 0.2464, + "step": 15011 + }, + { + "epoch": 0.91, + "learning_rate": 4.511360450836111e-07, + "loss": 0.2747, + "step": 15012 + }, + { + "epoch": 0.91, + "learning_rate": 4.505551597675339e-07, + "loss": 0.2626, + "step": 15013 + }, + { + "epoch": 0.91, + "learning_rate": 4.499746400489002e-07, + "loss": 0.2684, + "step": 15014 + }, + { + "epoch": 0.91, + "learning_rate": 4.493944859499344e-07, + "loss": 0.2905, + "step": 15015 + }, + { + "epoch": 0.91, + "learning_rate": 4.4881469749284755e-07, + "loss": 0.2729, + "step": 15016 + }, + { + "epoch": 0.91, + "learning_rate": 4.482352746998364e-07, + "loss": 0.2718, + "step": 15017 + }, + { + "epoch": 0.91, + "learning_rate": 4.476562175930832e-07, + "loss": 0.2395, + "step": 15018 + }, + { + "epoch": 0.91, + "learning_rate": 4.4707752619476017e-07, + "loss": 0.2708, + "step": 15019 + }, + { + "epoch": 0.91, + "learning_rate": 4.4649920052701966e-07, + "loss": 0.263, + "step": 15020 + }, + { + "epoch": 0.91, + "learning_rate": 4.459212406120028e-07, + "loss": 0.2764, + "step": 15021 + }, + { + "epoch": 0.91, + "learning_rate": 4.4534364647183635e-07, + "loss": 0.2962, + "step": 15022 + }, + { + "epoch": 0.91, + "learning_rate": 4.4476641812863377e-07, + "loss": 0.263, + "step": 15023 + }, + { + "epoch": 0.91, + "learning_rate": 4.4418955560449505e-07, + "loss": 0.2783, + "step": 15024 + }, + { + "epoch": 0.91, + "learning_rate": 4.4361305892150484e-07, + "loss": 0.26, + "step": 15025 + }, + { + "epoch": 0.91, + "learning_rate": 4.4303692810173326e-07, + "loss": 0.2619, + "step": 15026 + }, + { + "epoch": 0.91, + "learning_rate": 4.424611631672382e-07, + "loss": 0.2606, + "step": 15027 + }, + { + "epoch": 0.91, + "learning_rate": 4.418857641400609e-07, + "loss": 0.2635, + "step": 15028 + }, + { + "epoch": 0.91, + "learning_rate": 4.4131073104223267e-07, + "loss": 0.2437, + "step": 15029 + }, + { + "epoch": 0.91, + "learning_rate": 4.4073606389576696e-07, + "loss": 0.2772, + "step": 15030 + }, + { + "epoch": 0.91, + "learning_rate": 4.401617627226651e-07, + "loss": 0.2769, + "step": 15031 + }, + { + "epoch": 0.91, + "learning_rate": 4.395878275449139e-07, + "loss": 0.2693, + "step": 15032 + }, + { + "epoch": 0.91, + "learning_rate": 4.39014258384487e-07, + "loss": 0.2722, + "step": 15033 + }, + { + "epoch": 0.91, + "learning_rate": 4.3844105526334113e-07, + "loss": 0.2754, + "step": 15034 + }, + { + "epoch": 0.91, + "learning_rate": 4.378682182034244e-07, + "loss": 0.2783, + "step": 15035 + }, + { + "epoch": 0.91, + "learning_rate": 4.372957472266659e-07, + "loss": 0.2638, + "step": 15036 + }, + { + "epoch": 0.91, + "learning_rate": 4.3672364235498366e-07, + "loss": 0.2761, + "step": 15037 + }, + { + "epoch": 0.91, + "learning_rate": 4.3615190361027794e-07, + "loss": 0.2635, + "step": 15038 + }, + { + "epoch": 0.91, + "learning_rate": 4.3558053101443895e-07, + "loss": 0.2812, + "step": 15039 + }, + { + "epoch": 0.91, + "learning_rate": 4.3500952458934376e-07, + "loss": 0.3022, + "step": 15040 + }, + { + "epoch": 0.91, + "learning_rate": 4.344388843568503e-07, + "loss": 0.2748, + "step": 15041 + }, + { + "epoch": 0.91, + "learning_rate": 4.3386861033880677e-07, + "loss": 0.2475, + "step": 15042 + }, + { + "epoch": 0.91, + "learning_rate": 4.3329870255704343e-07, + "loss": 0.2609, + "step": 15043 + }, + { + "epoch": 0.91, + "learning_rate": 4.327291610333817e-07, + "loss": 0.2384, + "step": 15044 + }, + { + "epoch": 0.91, + "learning_rate": 4.3215998578962637e-07, + "loss": 0.2799, + "step": 15045 + }, + { + "epoch": 0.91, + "learning_rate": 4.315911768475678e-07, + "loss": 0.2626, + "step": 15046 + }, + { + "epoch": 0.91, + "learning_rate": 4.3102273422898077e-07, + "loss": 0.2532, + "step": 15047 + }, + { + "epoch": 0.91, + "learning_rate": 4.3045465795562903e-07, + "loss": 0.2429, + "step": 15048 + }, + { + "epoch": 0.91, + "learning_rate": 4.298869480492629e-07, + "loss": 0.2719, + "step": 15049 + }, + { + "epoch": 0.91, + "learning_rate": 4.29319604531615e-07, + "loss": 0.2766, + "step": 15050 + }, + { + "epoch": 0.91, + "learning_rate": 4.2875262742440694e-07, + "loss": 0.2753, + "step": 15051 + }, + { + "epoch": 0.91, + "learning_rate": 4.2818601674934456e-07, + "loss": 0.2672, + "step": 15052 + }, + { + "epoch": 0.91, + "learning_rate": 4.2761977252811946e-07, + "loss": 0.221, + "step": 15053 + }, + { + "epoch": 0.91, + "learning_rate": 4.270538947824121e-07, + "loss": 0.2668, + "step": 15054 + }, + { + "epoch": 0.91, + "learning_rate": 4.264883835338862e-07, + "loss": 0.278, + "step": 15055 + }, + { + "epoch": 0.91, + "learning_rate": 4.259232388041923e-07, + "loss": 0.2467, + "step": 15056 + }, + { + "epoch": 0.91, + "learning_rate": 4.2535846061496743e-07, + "loss": 0.2744, + "step": 15057 + }, + { + "epoch": 0.91, + "learning_rate": 4.2479404898782994e-07, + "loss": 0.243, + "step": 15058 + }, + { + "epoch": 0.91, + "learning_rate": 4.242300039443936e-07, + "loss": 0.2397, + "step": 15059 + }, + { + "epoch": 0.91, + "learning_rate": 4.236663255062512e-07, + "loss": 0.2539, + "step": 15060 + }, + { + "epoch": 0.91, + "learning_rate": 4.2310301369498095e-07, + "loss": 0.2716, + "step": 15061 + }, + { + "epoch": 0.91, + "learning_rate": 4.2254006853215015e-07, + "loss": 0.2833, + "step": 15062 + }, + { + "epoch": 0.91, + "learning_rate": 4.2197749003931035e-07, + "loss": 0.2737, + "step": 15063 + }, + { + "epoch": 0.91, + "learning_rate": 4.214152782380021e-07, + "loss": 0.2712, + "step": 15064 + }, + { + "epoch": 0.91, + "learning_rate": 4.2085343314974714e-07, + "loss": 0.2715, + "step": 15065 + }, + { + "epoch": 0.91, + "learning_rate": 4.20291954796056e-07, + "loss": 0.2762, + "step": 15066 + }, + { + "epoch": 0.91, + "learning_rate": 4.1973084319842594e-07, + "loss": 0.2653, + "step": 15067 + }, + { + "epoch": 0.91, + "learning_rate": 4.191700983783364e-07, + "loss": 0.2967, + "step": 15068 + }, + { + "epoch": 0.91, + "learning_rate": 4.18609720357257e-07, + "loss": 0.2748, + "step": 15069 + }, + { + "epoch": 0.91, + "learning_rate": 4.1804970915664267e-07, + "loss": 0.2397, + "step": 15070 + }, + { + "epoch": 0.91, + "learning_rate": 4.1749006479793075e-07, + "loss": 0.2758, + "step": 15071 + }, + { + "epoch": 0.91, + "learning_rate": 4.1693078730254964e-07, + "loss": 0.2619, + "step": 15072 + }, + { + "epoch": 0.91, + "learning_rate": 4.163718766919089e-07, + "loss": 0.2769, + "step": 15073 + }, + { + "epoch": 0.91, + "learning_rate": 4.158133329874081e-07, + "loss": 0.2758, + "step": 15074 + }, + { + "epoch": 0.91, + "learning_rate": 4.15255156210429e-07, + "loss": 0.266, + "step": 15075 + }, + { + "epoch": 0.91, + "learning_rate": 4.146973463823434e-07, + "loss": 0.2596, + "step": 15076 + }, + { + "epoch": 0.91, + "learning_rate": 4.141399035245053e-07, + "loss": 0.2547, + "step": 15077 + }, + { + "epoch": 0.91, + "learning_rate": 4.135828276582543e-07, + "loss": 0.2633, + "step": 15078 + }, + { + "epoch": 0.91, + "learning_rate": 4.1302611880492227e-07, + "loss": 0.2853, + "step": 15079 + }, + { + "epoch": 0.91, + "learning_rate": 4.124697769858199e-07, + "loss": 0.2447, + "step": 15080 + }, + { + "epoch": 0.91, + "learning_rate": 4.1191380222224796e-07, + "loss": 0.2602, + "step": 15081 + }, + { + "epoch": 0.91, + "learning_rate": 4.1135819453549054e-07, + "loss": 0.2539, + "step": 15082 + }, + { + "epoch": 0.91, + "learning_rate": 4.1080295394681837e-07, + "loss": 0.2974, + "step": 15083 + }, + { + "epoch": 0.91, + "learning_rate": 4.1024808047749e-07, + "loss": 0.2733, + "step": 15084 + }, + { + "epoch": 0.91, + "learning_rate": 4.096935741487495e-07, + "loss": 0.273, + "step": 15085 + }, + { + "epoch": 0.91, + "learning_rate": 4.091394349818234e-07, + "loss": 0.2969, + "step": 15086 + }, + { + "epoch": 0.91, + "learning_rate": 4.0858566299792903e-07, + "loss": 0.2503, + "step": 15087 + }, + { + "epoch": 0.91, + "learning_rate": 4.0803225821826495e-07, + "loss": 0.2444, + "step": 15088 + }, + { + "epoch": 0.91, + "learning_rate": 4.07479220664021e-07, + "loss": 0.2785, + "step": 15089 + }, + { + "epoch": 0.91, + "learning_rate": 4.0692655035636797e-07, + "loss": 0.2663, + "step": 15090 + }, + { + "epoch": 0.91, + "learning_rate": 4.063742473164645e-07, + "loss": 0.303, + "step": 15091 + }, + { + "epoch": 0.91, + "learning_rate": 4.0582231156545694e-07, + "loss": 0.2584, + "step": 15092 + }, + { + "epoch": 0.91, + "learning_rate": 4.05270743124474e-07, + "loss": 0.2833, + "step": 15093 + }, + { + "epoch": 0.91, + "learning_rate": 4.0471954201463437e-07, + "loss": 0.2607, + "step": 15094 + }, + { + "epoch": 0.91, + "learning_rate": 4.0416870825704e-07, + "loss": 0.2859, + "step": 15095 + }, + { + "epoch": 0.91, + "learning_rate": 4.0361824187277855e-07, + "loss": 0.2575, + "step": 15096 + }, + { + "epoch": 0.91, + "learning_rate": 4.030681428829253e-07, + "loss": 0.2757, + "step": 15097 + }, + { + "epoch": 0.91, + "learning_rate": 4.02518411308539e-07, + "loss": 0.2624, + "step": 15098 + }, + { + "epoch": 0.91, + "learning_rate": 4.0196904717066833e-07, + "loss": 0.2621, + "step": 15099 + }, + { + "epoch": 0.91, + "learning_rate": 4.014200504903443e-07, + "loss": 0.2843, + "step": 15100 + }, + { + "epoch": 0.91, + "learning_rate": 4.008714212885856e-07, + "loss": 0.2482, + "step": 15101 + }, + { + "epoch": 0.91, + "learning_rate": 4.0032315958639546e-07, + "loss": 0.2751, + "step": 15102 + }, + { + "epoch": 0.91, + "learning_rate": 3.9977526540476483e-07, + "loss": 0.2939, + "step": 15103 + }, + { + "epoch": 0.91, + "learning_rate": 3.9922773876466705e-07, + "loss": 0.2653, + "step": 15104 + }, + { + "epoch": 0.91, + "learning_rate": 3.986805796870674e-07, + "loss": 0.2686, + "step": 15105 + }, + { + "epoch": 0.91, + "learning_rate": 3.981337881929126e-07, + "loss": 0.2493, + "step": 15106 + }, + { + "epoch": 0.91, + "learning_rate": 3.975873643031369e-07, + "loss": 0.2753, + "step": 15107 + }, + { + "epoch": 0.91, + "learning_rate": 3.9704130803865815e-07, + "loss": 0.2825, + "step": 15108 + }, + { + "epoch": 0.91, + "learning_rate": 3.9649561942038396e-07, + "loss": 0.2701, + "step": 15109 + }, + { + "epoch": 0.91, + "learning_rate": 3.959502984692043e-07, + "loss": 0.2835, + "step": 15110 + }, + { + "epoch": 0.91, + "learning_rate": 3.9540534520599806e-07, + "loss": 0.2472, + "step": 15111 + }, + { + "epoch": 0.91, + "learning_rate": 3.948607596516274e-07, + "loss": 0.291, + "step": 15112 + }, + { + "epoch": 0.91, + "learning_rate": 3.943165418269401e-07, + "loss": 0.2694, + "step": 15113 + }, + { + "epoch": 0.91, + "learning_rate": 3.937726917527751e-07, + "loss": 0.2641, + "step": 15114 + }, + { + "epoch": 0.91, + "learning_rate": 3.9322920944995123e-07, + "loss": 0.2786, + "step": 15115 + }, + { + "epoch": 0.91, + "learning_rate": 3.9268609493927634e-07, + "loss": 0.2801, + "step": 15116 + }, + { + "epoch": 0.91, + "learning_rate": 3.9214334824154265e-07, + "loss": 0.257, + "step": 15117 + }, + { + "epoch": 0.91, + "learning_rate": 3.916009693775291e-07, + "loss": 0.263, + "step": 15118 + }, + { + "epoch": 0.91, + "learning_rate": 3.9105895836800134e-07, + "loss": 0.2528, + "step": 15119 + }, + { + "epoch": 0.91, + "learning_rate": 3.905173152337094e-07, + "loss": 0.2395, + "step": 15120 + }, + { + "epoch": 0.91, + "learning_rate": 3.8997603999539e-07, + "loss": 0.266, + "step": 15121 + }, + { + "epoch": 0.91, + "learning_rate": 3.8943513267376664e-07, + "loss": 0.2796, + "step": 15122 + }, + { + "epoch": 0.91, + "learning_rate": 3.8889459328954495e-07, + "loss": 0.2818, + "step": 15123 + }, + { + "epoch": 0.91, + "learning_rate": 3.8835442186342277e-07, + "loss": 0.2711, + "step": 15124 + }, + { + "epoch": 0.91, + "learning_rate": 3.8781461841607804e-07, + "loss": 0.2662, + "step": 15125 + }, + { + "epoch": 0.91, + "learning_rate": 3.872751829681787e-07, + "loss": 0.2684, + "step": 15126 + }, + { + "epoch": 0.91, + "learning_rate": 3.867361155403759e-07, + "loss": 0.2953, + "step": 15127 + }, + { + "epoch": 0.91, + "learning_rate": 3.861974161533066e-07, + "loss": 0.2573, + "step": 15128 + }, + { + "epoch": 0.91, + "learning_rate": 3.856590848275976e-07, + "loss": 0.246, + "step": 15129 + }, + { + "epoch": 0.91, + "learning_rate": 3.851211215838557e-07, + "loss": 0.2514, + "step": 15130 + }, + { + "epoch": 0.91, + "learning_rate": 3.8458352644267895e-07, + "loss": 0.2529, + "step": 15131 + }, + { + "epoch": 0.91, + "learning_rate": 3.8404629942464744e-07, + "loss": 0.2896, + "step": 15132 + }, + { + "epoch": 0.91, + "learning_rate": 3.835094405503281e-07, + "loss": 0.2988, + "step": 15133 + }, + { + "epoch": 0.91, + "learning_rate": 3.829729498402779e-07, + "loss": 0.2427, + "step": 15134 + }, + { + "epoch": 0.91, + "learning_rate": 3.824368273150325e-07, + "loss": 0.2638, + "step": 15135 + }, + { + "epoch": 0.91, + "learning_rate": 3.8190107299512e-07, + "loss": 0.2655, + "step": 15136 + }, + { + "epoch": 0.91, + "learning_rate": 3.8136568690104956e-07, + "loss": 0.2681, + "step": 15137 + }, + { + "epoch": 0.91, + "learning_rate": 3.808306690533181e-07, + "loss": 0.2773, + "step": 15138 + }, + { + "epoch": 0.91, + "learning_rate": 3.802960194724092e-07, + "loss": 0.2441, + "step": 15139 + }, + { + "epoch": 0.91, + "learning_rate": 3.7976173817879326e-07, + "loss": 0.2575, + "step": 15140 + }, + { + "epoch": 0.91, + "learning_rate": 3.7922782519292265e-07, + "loss": 0.2879, + "step": 15141 + }, + { + "epoch": 0.91, + "learning_rate": 3.7869428053524115e-07, + "loss": 0.2557, + "step": 15142 + }, + { + "epoch": 0.91, + "learning_rate": 3.7816110422617127e-07, + "loss": 0.2508, + "step": 15143 + }, + { + "epoch": 0.91, + "learning_rate": 3.7762829628612883e-07, + "loss": 0.2635, + "step": 15144 + }, + { + "epoch": 0.91, + "learning_rate": 3.7709585673550985e-07, + "loss": 0.2575, + "step": 15145 + }, + { + "epoch": 0.92, + "learning_rate": 3.765637855947013e-07, + "loss": 0.2536, + "step": 15146 + }, + { + "epoch": 0.92, + "learning_rate": 3.7603208288407025e-07, + "loss": 0.2903, + "step": 15147 + }, + { + "epoch": 0.92, + "learning_rate": 3.7550074862397366e-07, + "loss": 0.2556, + "step": 15148 + }, + { + "epoch": 0.92, + "learning_rate": 3.749697828347565e-07, + "loss": 0.2706, + "step": 15149 + }, + { + "epoch": 0.92, + "learning_rate": 3.744391855367424e-07, + "loss": 0.2522, + "step": 15150 + }, + { + "epoch": 0.92, + "learning_rate": 3.7390895675024854e-07, + "loss": 0.2511, + "step": 15151 + }, + { + "epoch": 0.92, + "learning_rate": 3.733790964955719e-07, + "loss": 0.272, + "step": 15152 + }, + { + "epoch": 0.92, + "learning_rate": 3.728496047929986e-07, + "loss": 0.2373, + "step": 15153 + }, + { + "epoch": 0.92, + "learning_rate": 3.723204816628012e-07, + "loss": 0.2598, + "step": 15154 + }, + { + "epoch": 0.92, + "learning_rate": 3.7179172712523693e-07, + "loss": 0.2713, + "step": 15155 + }, + { + "epoch": 0.92, + "learning_rate": 3.712633412005484e-07, + "loss": 0.2755, + "step": 15156 + }, + { + "epoch": 0.92, + "learning_rate": 3.7073532390896394e-07, + "loss": 0.275, + "step": 15157 + }, + { + "epoch": 0.92, + "learning_rate": 3.7020767527069957e-07, + "loss": 0.2628, + "step": 15158 + }, + { + "epoch": 0.92, + "learning_rate": 3.696803953059558e-07, + "loss": 0.2467, + "step": 15159 + }, + { + "epoch": 0.92, + "learning_rate": 3.691534840349198e-07, + "loss": 0.2757, + "step": 15160 + }, + { + "epoch": 0.92, + "learning_rate": 3.6862694147776434e-07, + "loss": 0.2246, + "step": 15161 + }, + { + "epoch": 0.92, + "learning_rate": 3.681007676546466e-07, + "loss": 0.2589, + "step": 15162 + }, + { + "epoch": 0.92, + "learning_rate": 3.675749625857106e-07, + "loss": 0.2489, + "step": 15163 + }, + { + "epoch": 0.92, + "learning_rate": 3.6704952629109004e-07, + "loss": 0.2487, + "step": 15164 + }, + { + "epoch": 0.92, + "learning_rate": 3.665244587908989e-07, + "loss": 0.2637, + "step": 15165 + }, + { + "epoch": 0.92, + "learning_rate": 3.6599976010523786e-07, + "loss": 0.2578, + "step": 15166 + }, + { + "epoch": 0.92, + "learning_rate": 3.654754302541974e-07, + "loss": 0.2864, + "step": 15167 + }, + { + "epoch": 0.92, + "learning_rate": 3.649514692578482e-07, + "loss": 0.2732, + "step": 15168 + }, + { + "epoch": 0.92, + "learning_rate": 3.6442787713625416e-07, + "loss": 0.2613, + "step": 15169 + }, + { + "epoch": 0.92, + "learning_rate": 3.6390465390945707e-07, + "loss": 0.2446, + "step": 15170 + }, + { + "epoch": 0.92, + "learning_rate": 3.633817995974909e-07, + "loss": 0.2553, + "step": 15171 + }, + { + "epoch": 0.92, + "learning_rate": 3.628593142203718e-07, + "loss": 0.2503, + "step": 15172 + }, + { + "epoch": 0.92, + "learning_rate": 3.623371977981027e-07, + "loss": 0.264, + "step": 15173 + }, + { + "epoch": 0.92, + "learning_rate": 3.618154503506721e-07, + "loss": 0.2564, + "step": 15174 + }, + { + "epoch": 0.92, + "learning_rate": 3.612940718980562e-07, + "loss": 0.2751, + "step": 15175 + }, + { + "epoch": 0.92, + "learning_rate": 3.6077306246021795e-07, + "loss": 0.2658, + "step": 15176 + }, + { + "epoch": 0.92, + "learning_rate": 3.602524220570991e-07, + "loss": 0.264, + "step": 15177 + }, + { + "epoch": 0.92, + "learning_rate": 3.597321507086349e-07, + "loss": 0.2719, + "step": 15178 + }, + { + "epoch": 0.92, + "learning_rate": 3.592122484347438e-07, + "loss": 0.269, + "step": 15179 + }, + { + "epoch": 0.92, + "learning_rate": 3.5869271525532987e-07, + "loss": 0.2292, + "step": 15180 + }, + { + "epoch": 0.92, + "learning_rate": 3.5817355119028283e-07, + "loss": 0.2839, + "step": 15181 + }, + { + "epoch": 0.92, + "learning_rate": 3.57654756259479e-07, + "loss": 0.2952, + "step": 15182 + }, + { + "epoch": 0.92, + "learning_rate": 3.571363304827802e-07, + "loss": 0.267, + "step": 15183 + }, + { + "epoch": 0.92, + "learning_rate": 3.56618273880035e-07, + "loss": 0.2762, + "step": 15184 + }, + { + "epoch": 0.92, + "learning_rate": 3.561005864710754e-07, + "loss": 0.2522, + "step": 15185 + }, + { + "epoch": 0.92, + "learning_rate": 3.555832682757232e-07, + "loss": 0.2889, + "step": 15186 + }, + { + "epoch": 0.92, + "learning_rate": 3.5506631931378156e-07, + "loss": 0.2653, + "step": 15187 + }, + { + "epoch": 0.92, + "learning_rate": 3.5454973960504237e-07, + "loss": 0.2423, + "step": 15188 + }, + { + "epoch": 0.92, + "learning_rate": 3.540335291692831e-07, + "loss": 0.2531, + "step": 15189 + }, + { + "epoch": 0.92, + "learning_rate": 3.5351768802626695e-07, + "loss": 0.272, + "step": 15190 + }, + { + "epoch": 0.92, + "learning_rate": 3.530022161957425e-07, + "loss": 0.2694, + "step": 15191 + }, + { + "epoch": 0.92, + "learning_rate": 3.524871136974439e-07, + "loss": 0.2706, + "step": 15192 + }, + { + "epoch": 0.92, + "learning_rate": 3.5197238055109107e-07, + "loss": 0.2536, + "step": 15193 + }, + { + "epoch": 0.92, + "learning_rate": 3.5145801677639257e-07, + "loss": 0.2557, + "step": 15194 + }, + { + "epoch": 0.92, + "learning_rate": 3.5094402239303935e-07, + "loss": 0.2405, + "step": 15195 + }, + { + "epoch": 0.92, + "learning_rate": 3.5043039742071015e-07, + "loss": 0.2596, + "step": 15196 + }, + { + "epoch": 0.92, + "learning_rate": 3.499171418790681e-07, + "loss": 0.2812, + "step": 15197 + }, + { + "epoch": 0.92, + "learning_rate": 3.4940425578776194e-07, + "loss": 0.2856, + "step": 15198 + }, + { + "epoch": 0.92, + "learning_rate": 3.488917391664304e-07, + "loss": 0.2603, + "step": 15199 + }, + { + "epoch": 0.92, + "learning_rate": 3.483795920346922e-07, + "loss": 0.2506, + "step": 15200 + }, + { + "epoch": 0.92, + "learning_rate": 3.4786781441215725e-07, + "loss": 0.2709, + "step": 15201 + }, + { + "epoch": 0.92, + "learning_rate": 3.473564063184165e-07, + "loss": 0.2441, + "step": 15202 + }, + { + "epoch": 0.92, + "learning_rate": 3.468453677730488e-07, + "loss": 0.2327, + "step": 15203 + }, + { + "epoch": 0.92, + "learning_rate": 3.4633469879562175e-07, + "loss": 0.2842, + "step": 15204 + }, + { + "epoch": 0.92, + "learning_rate": 3.4582439940568424e-07, + "loss": 0.2719, + "step": 15205 + }, + { + "epoch": 0.92, + "learning_rate": 3.453144696227739e-07, + "loss": 0.2616, + "step": 15206 + }, + { + "epoch": 0.92, + "learning_rate": 3.4480490946641296e-07, + "loss": 0.2786, + "step": 15207 + }, + { + "epoch": 0.92, + "learning_rate": 3.4429571895610915e-07, + "loss": 0.2596, + "step": 15208 + }, + { + "epoch": 0.92, + "learning_rate": 3.4378689811135567e-07, + "loss": 0.2796, + "step": 15209 + }, + { + "epoch": 0.92, + "learning_rate": 3.432784469516359e-07, + "loss": 0.2522, + "step": 15210 + }, + { + "epoch": 0.92, + "learning_rate": 3.427703654964132e-07, + "loss": 0.2888, + "step": 15211 + }, + { + "epoch": 0.92, + "learning_rate": 3.4226265376513856e-07, + "loss": 0.2598, + "step": 15212 + }, + { + "epoch": 0.92, + "learning_rate": 3.41755311777251e-07, + "loss": 0.2726, + "step": 15213 + }, + { + "epoch": 0.92, + "learning_rate": 3.412483395521748e-07, + "loss": 0.2595, + "step": 15214 + }, + { + "epoch": 0.92, + "learning_rate": 3.4074173710931804e-07, + "loss": 0.252, + "step": 15215 + }, + { + "epoch": 0.92, + "learning_rate": 3.4023550446807605e-07, + "loss": 0.2501, + "step": 15216 + }, + { + "epoch": 0.92, + "learning_rate": 3.3972964164782907e-07, + "loss": 0.2854, + "step": 15217 + }, + { + "epoch": 0.92, + "learning_rate": 3.392241486679437e-07, + "loss": 0.26, + "step": 15218 + }, + { + "epoch": 0.92, + "learning_rate": 3.387190255477757e-07, + "loss": 0.2819, + "step": 15219 + }, + { + "epoch": 0.92, + "learning_rate": 3.382142723066595e-07, + "loss": 0.2598, + "step": 15220 + }, + { + "epoch": 0.92, + "learning_rate": 3.37709888963923e-07, + "loss": 0.2826, + "step": 15221 + }, + { + "epoch": 0.92, + "learning_rate": 3.3720587553887294e-07, + "loss": 0.2479, + "step": 15222 + }, + { + "epoch": 0.92, + "learning_rate": 3.367022320508062e-07, + "loss": 0.258, + "step": 15223 + }, + { + "epoch": 0.92, + "learning_rate": 3.361989585190073e-07, + "loss": 0.2633, + "step": 15224 + }, + { + "epoch": 0.92, + "learning_rate": 3.356960549627408e-07, + "loss": 0.2789, + "step": 15225 + }, + { + "epoch": 0.92, + "learning_rate": 3.3519352140126247e-07, + "loss": 0.2407, + "step": 15226 + }, + { + "epoch": 0.92, + "learning_rate": 3.3469135785381026e-07, + "loss": 0.2753, + "step": 15227 + }, + { + "epoch": 0.92, + "learning_rate": 3.341895643396076e-07, + "loss": 0.245, + "step": 15228 + }, + { + "epoch": 0.92, + "learning_rate": 3.336881408778692e-07, + "loss": 0.26, + "step": 15229 + }, + { + "epoch": 0.92, + "learning_rate": 3.331870874877907e-07, + "loss": 0.2753, + "step": 15230 + }, + { + "epoch": 0.92, + "learning_rate": 3.326864041885536e-07, + "loss": 0.272, + "step": 15231 + }, + { + "epoch": 0.92, + "learning_rate": 3.3218609099932684e-07, + "loss": 0.2782, + "step": 15232 + }, + { + "epoch": 0.92, + "learning_rate": 3.3168614793926523e-07, + "loss": 0.2753, + "step": 15233 + }, + { + "epoch": 0.92, + "learning_rate": 3.311865750275078e-07, + "loss": 0.2778, + "step": 15234 + }, + { + "epoch": 0.92, + "learning_rate": 3.306873722831827e-07, + "loss": 0.2614, + "step": 15235 + }, + { + "epoch": 0.92, + "learning_rate": 3.301885397254001e-07, + "loss": 0.2697, + "step": 15236 + }, + { + "epoch": 0.92, + "learning_rate": 3.296900773732581e-07, + "loss": 0.2683, + "step": 15237 + }, + { + "epoch": 0.92, + "learning_rate": 3.291919852458392e-07, + "loss": 0.2591, + "step": 15238 + }, + { + "epoch": 0.92, + "learning_rate": 3.2869426336221476e-07, + "loss": 0.2376, + "step": 15239 + }, + { + "epoch": 0.92, + "learning_rate": 3.2819691174143854e-07, + "loss": 0.2289, + "step": 15240 + }, + { + "epoch": 0.92, + "learning_rate": 3.276999304025508e-07, + "loss": 0.315, + "step": 15241 + }, + { + "epoch": 0.92, + "learning_rate": 3.272033193645807e-07, + "loss": 0.2508, + "step": 15242 + }, + { + "epoch": 0.92, + "learning_rate": 3.267070786465365e-07, + "loss": 0.2547, + "step": 15243 + }, + { + "epoch": 0.92, + "learning_rate": 3.262112082674229e-07, + "loss": 0.2613, + "step": 15244 + }, + { + "epoch": 0.92, + "learning_rate": 3.257157082462192e-07, + "loss": 0.2603, + "step": 15245 + }, + { + "epoch": 0.92, + "learning_rate": 3.2522057860189693e-07, + "loss": 0.2725, + "step": 15246 + }, + { + "epoch": 0.92, + "learning_rate": 3.2472581935341195e-07, + "loss": 0.2508, + "step": 15247 + }, + { + "epoch": 0.92, + "learning_rate": 3.242314305197047e-07, + "loss": 0.2821, + "step": 15248 + }, + { + "epoch": 0.92, + "learning_rate": 3.2373741211970566e-07, + "loss": 0.2854, + "step": 15249 + }, + { + "epoch": 0.92, + "learning_rate": 3.232437641723263e-07, + "loss": 0.2506, + "step": 15250 + }, + { + "epoch": 0.92, + "learning_rate": 3.227504866964659e-07, + "loss": 0.2634, + "step": 15251 + }, + { + "epoch": 0.92, + "learning_rate": 3.2225757971100833e-07, + "loss": 0.2472, + "step": 15252 + }, + { + "epoch": 0.92, + "learning_rate": 3.2176504323482625e-07, + "loss": 0.2598, + "step": 15253 + }, + { + "epoch": 0.92, + "learning_rate": 3.212728772867757e-07, + "loss": 0.2803, + "step": 15254 + }, + { + "epoch": 0.92, + "learning_rate": 3.2078108188569936e-07, + "loss": 0.2667, + "step": 15255 + }, + { + "epoch": 0.92, + "learning_rate": 3.2028965705042434e-07, + "loss": 0.2453, + "step": 15256 + }, + { + "epoch": 0.92, + "learning_rate": 3.197986027997657e-07, + "loss": 0.2863, + "step": 15257 + }, + { + "epoch": 0.92, + "learning_rate": 3.1930791915252166e-07, + "loss": 0.256, + "step": 15258 + }, + { + "epoch": 0.92, + "learning_rate": 3.1881760612748057e-07, + "loss": 0.2257, + "step": 15259 + }, + { + "epoch": 0.92, + "learning_rate": 3.1832766374341293e-07, + "loss": 0.2667, + "step": 15260 + }, + { + "epoch": 0.92, + "learning_rate": 3.1783809201907487e-07, + "loss": 0.2653, + "step": 15261 + }, + { + "epoch": 0.92, + "learning_rate": 3.1734889097321144e-07, + "loss": 0.2846, + "step": 15262 + }, + { + "epoch": 0.92, + "learning_rate": 3.1686006062454757e-07, + "loss": 0.2557, + "step": 15263 + }, + { + "epoch": 0.92, + "learning_rate": 3.1637160099180274e-07, + "loss": 0.2776, + "step": 15264 + }, + { + "epoch": 0.92, + "learning_rate": 3.158835120936765e-07, + "loss": 0.2476, + "step": 15265 + }, + { + "epoch": 0.92, + "learning_rate": 3.153957939488528e-07, + "loss": 0.2759, + "step": 15266 + }, + { + "epoch": 0.92, + "learning_rate": 3.149084465760055e-07, + "loss": 0.2528, + "step": 15267 + }, + { + "epoch": 0.92, + "learning_rate": 3.1442146999379084e-07, + "loss": 0.2944, + "step": 15268 + }, + { + "epoch": 0.92, + "learning_rate": 3.139348642208562e-07, + "loss": 0.2645, + "step": 15269 + }, + { + "epoch": 0.92, + "learning_rate": 3.1344862927582764e-07, + "loss": 0.2699, + "step": 15270 + }, + { + "epoch": 0.92, + "learning_rate": 3.1296276517732147e-07, + "loss": 0.2454, + "step": 15271 + }, + { + "epoch": 0.92, + "learning_rate": 3.124772719439395e-07, + "loss": 0.2786, + "step": 15272 + }, + { + "epoch": 0.92, + "learning_rate": 3.1199214959426793e-07, + "loss": 0.2517, + "step": 15273 + }, + { + "epoch": 0.92, + "learning_rate": 3.115073981468797e-07, + "loss": 0.2683, + "step": 15274 + }, + { + "epoch": 0.92, + "learning_rate": 3.110230176203344e-07, + "loss": 0.2486, + "step": 15275 + }, + { + "epoch": 0.92, + "learning_rate": 3.105390080331749e-07, + "loss": 0.2775, + "step": 15276 + }, + { + "epoch": 0.92, + "learning_rate": 3.100553694039321e-07, + "loss": 0.2854, + "step": 15277 + }, + { + "epoch": 0.92, + "learning_rate": 3.0957210175112217e-07, + "loss": 0.2344, + "step": 15278 + }, + { + "epoch": 0.92, + "learning_rate": 3.0908920509324594e-07, + "loss": 0.2674, + "step": 15279 + }, + { + "epoch": 0.92, + "learning_rate": 3.0860667944879186e-07, + "loss": 0.2595, + "step": 15280 + }, + { + "epoch": 0.92, + "learning_rate": 3.08124524836233e-07, + "loss": 0.2676, + "step": 15281 + }, + { + "epoch": 0.92, + "learning_rate": 3.0764274127402793e-07, + "loss": 0.2528, + "step": 15282 + }, + { + "epoch": 0.92, + "learning_rate": 3.0716132878062186e-07, + "loss": 0.252, + "step": 15283 + }, + { + "epoch": 0.92, + "learning_rate": 3.0668028737444566e-07, + "loss": 0.2711, + "step": 15284 + }, + { + "epoch": 0.92, + "learning_rate": 3.0619961707391677e-07, + "loss": 0.2677, + "step": 15285 + }, + { + "epoch": 0.92, + "learning_rate": 3.0571931789743604e-07, + "loss": 0.2472, + "step": 15286 + }, + { + "epoch": 0.92, + "learning_rate": 3.052393898633932e-07, + "loss": 0.2874, + "step": 15287 + }, + { + "epoch": 0.92, + "learning_rate": 3.047598329901591e-07, + "loss": 0.2868, + "step": 15288 + }, + { + "epoch": 0.92, + "learning_rate": 3.0428064729609576e-07, + "loss": 0.2436, + "step": 15289 + }, + { + "epoch": 0.92, + "learning_rate": 3.0380183279954953e-07, + "loss": 0.2646, + "step": 15290 + }, + { + "epoch": 0.92, + "learning_rate": 3.033233895188503e-07, + "loss": 0.2762, + "step": 15291 + }, + { + "epoch": 0.92, + "learning_rate": 3.028453174723145e-07, + "loss": 0.2514, + "step": 15292 + }, + { + "epoch": 0.92, + "learning_rate": 3.0236761667824523e-07, + "loss": 0.2821, + "step": 15293 + }, + { + "epoch": 0.92, + "learning_rate": 3.0189028715493227e-07, + "loss": 0.2465, + "step": 15294 + }, + { + "epoch": 0.92, + "learning_rate": 3.0141332892065e-07, + "loss": 0.2593, + "step": 15295 + }, + { + "epoch": 0.92, + "learning_rate": 3.009367419936582e-07, + "loss": 0.2607, + "step": 15296 + }, + { + "epoch": 0.92, + "learning_rate": 3.0046052639220223e-07, + "loss": 0.26, + "step": 15297 + }, + { + "epoch": 0.92, + "learning_rate": 2.99984682134512e-07, + "loss": 0.2708, + "step": 15298 + }, + { + "epoch": 0.92, + "learning_rate": 2.9950920923880967e-07, + "loss": 0.2432, + "step": 15299 + }, + { + "epoch": 0.92, + "learning_rate": 2.990341077232961e-07, + "loss": 0.2955, + "step": 15300 + }, + { + "epoch": 0.92, + "learning_rate": 2.985593776061602e-07, + "loss": 0.2679, + "step": 15301 + }, + { + "epoch": 0.92, + "learning_rate": 2.980850189055773e-07, + "loss": 0.2713, + "step": 15302 + }, + { + "epoch": 0.92, + "learning_rate": 2.976110316397074e-07, + "loss": 0.2719, + "step": 15303 + }, + { + "epoch": 0.92, + "learning_rate": 2.971374158266982e-07, + "loss": 0.2748, + "step": 15304 + }, + { + "epoch": 0.92, + "learning_rate": 2.966641714846807e-07, + "loss": 0.266, + "step": 15305 + }, + { + "epoch": 0.92, + "learning_rate": 2.961912986317739e-07, + "loss": 0.2747, + "step": 15306 + }, + { + "epoch": 0.92, + "learning_rate": 2.9571879728608086e-07, + "loss": 0.2607, + "step": 15307 + }, + { + "epoch": 0.92, + "learning_rate": 2.9524666746569065e-07, + "loss": 0.2547, + "step": 15308 + }, + { + "epoch": 0.92, + "learning_rate": 2.9477490918867977e-07, + "loss": 0.2673, + "step": 15309 + }, + { + "epoch": 0.92, + "learning_rate": 2.9430352247310946e-07, + "loss": 0.2656, + "step": 15310 + }, + { + "epoch": 0.92, + "learning_rate": 2.938325073370263e-07, + "loss": 0.2592, + "step": 15311 + }, + { + "epoch": 0.93, + "learning_rate": 2.9336186379846255e-07, + "loss": 0.3029, + "step": 15312 + }, + { + "epoch": 0.93, + "learning_rate": 2.9289159187543605e-07, + "loss": 0.2613, + "step": 15313 + }, + { + "epoch": 0.93, + "learning_rate": 2.9242169158595237e-07, + "loss": 0.2686, + "step": 15314 + }, + { + "epoch": 0.93, + "learning_rate": 2.919521629480016e-07, + "loss": 0.2492, + "step": 15315 + }, + { + "epoch": 0.93, + "learning_rate": 2.9148300597955814e-07, + "loss": 0.2662, + "step": 15316 + }, + { + "epoch": 0.93, + "learning_rate": 2.910142206985833e-07, + "loss": 0.26, + "step": 15317 + }, + { + "epoch": 0.93, + "learning_rate": 2.9054580712302603e-07, + "loss": 0.2455, + "step": 15318 + }, + { + "epoch": 0.93, + "learning_rate": 2.900777652708187e-07, + "loss": 0.2761, + "step": 15319 + }, + { + "epoch": 0.93, + "learning_rate": 2.8961009515988015e-07, + "loss": 0.2546, + "step": 15320 + }, + { + "epoch": 0.93, + "learning_rate": 2.89142796808114e-07, + "loss": 0.2458, + "step": 15321 + }, + { + "epoch": 0.93, + "learning_rate": 2.886758702334125e-07, + "loss": 0.2755, + "step": 15322 + }, + { + "epoch": 0.93, + "learning_rate": 2.8820931545364916e-07, + "loss": 0.2638, + "step": 15323 + }, + { + "epoch": 0.93, + "learning_rate": 2.877431324866897e-07, + "loss": 0.244, + "step": 15324 + }, + { + "epoch": 0.93, + "learning_rate": 2.8727732135037876e-07, + "loss": 0.2829, + "step": 15325 + }, + { + "epoch": 0.93, + "learning_rate": 2.8681188206254985e-07, + "loss": 0.255, + "step": 15326 + }, + { + "epoch": 0.93, + "learning_rate": 2.863468146410242e-07, + "loss": 0.244, + "step": 15327 + }, + { + "epoch": 0.93, + "learning_rate": 2.858821191036043e-07, + "loss": 0.2706, + "step": 15328 + }, + { + "epoch": 0.93, + "learning_rate": 2.8541779546808255e-07, + "loss": 0.2436, + "step": 15329 + }, + { + "epoch": 0.93, + "learning_rate": 2.849538437522348e-07, + "loss": 0.2852, + "step": 15330 + }, + { + "epoch": 0.93, + "learning_rate": 2.844902639738234e-07, + "loss": 0.2665, + "step": 15331 + }, + { + "epoch": 0.93, + "learning_rate": 2.8402705615059753e-07, + "loss": 0.2769, + "step": 15332 + }, + { + "epoch": 0.93, + "learning_rate": 2.835642203002875e-07, + "loss": 0.2504, + "step": 15333 + }, + { + "epoch": 0.93, + "learning_rate": 2.8310175644061686e-07, + "loss": 0.2679, + "step": 15334 + }, + { + "epoch": 0.93, + "learning_rate": 2.826396645892893e-07, + "loss": 0.2688, + "step": 15335 + }, + { + "epoch": 0.93, + "learning_rate": 2.82177944763995e-07, + "loss": 0.2497, + "step": 15336 + }, + { + "epoch": 0.93, + "learning_rate": 2.8171659698241205e-07, + "loss": 0.2846, + "step": 15337 + }, + { + "epoch": 0.93, + "learning_rate": 2.8125562126220086e-07, + "loss": 0.2691, + "step": 15338 + }, + { + "epoch": 0.93, + "learning_rate": 2.807950176210128e-07, + "loss": 0.2674, + "step": 15339 + }, + { + "epoch": 0.93, + "learning_rate": 2.803347860764805e-07, + "loss": 0.2726, + "step": 15340 + }, + { + "epoch": 0.93, + "learning_rate": 2.7987492664622304e-07, + "loss": 0.2646, + "step": 15341 + }, + { + "epoch": 0.93, + "learning_rate": 2.7941543934784765e-07, + "loss": 0.2877, + "step": 15342 + }, + { + "epoch": 0.93, + "learning_rate": 2.789563241989435e-07, + "loss": 0.275, + "step": 15343 + }, + { + "epoch": 0.93, + "learning_rate": 2.784975812170887e-07, + "loss": 0.2779, + "step": 15344 + }, + { + "epoch": 0.93, + "learning_rate": 2.7803921041984703e-07, + "loss": 0.2761, + "step": 15345 + }, + { + "epoch": 0.93, + "learning_rate": 2.775812118247678e-07, + "loss": 0.261, + "step": 15346 + }, + { + "epoch": 0.93, + "learning_rate": 2.7712358544938144e-07, + "loss": 0.2676, + "step": 15347 + }, + { + "epoch": 0.93, + "learning_rate": 2.766663313112117e-07, + "loss": 0.2874, + "step": 15348 + }, + { + "epoch": 0.93, + "learning_rate": 2.762094494277623e-07, + "loss": 0.2493, + "step": 15349 + }, + { + "epoch": 0.93, + "learning_rate": 2.7575293981652593e-07, + "loss": 0.2751, + "step": 15350 + }, + { + "epoch": 0.93, + "learning_rate": 2.752968024949798e-07, + "loss": 0.2589, + "step": 15351 + }, + { + "epoch": 0.93, + "learning_rate": 2.748410374805854e-07, + "loss": 0.255, + "step": 15352 + }, + { + "epoch": 0.93, + "learning_rate": 2.7438564479079444e-07, + "loss": 0.2388, + "step": 15353 + }, + { + "epoch": 0.93, + "learning_rate": 2.7393062444303953e-07, + "loss": 0.2616, + "step": 15354 + }, + { + "epoch": 0.93, + "learning_rate": 2.7347597645474123e-07, + "loss": 0.2779, + "step": 15355 + }, + { + "epoch": 0.93, + "learning_rate": 2.730217008433067e-07, + "loss": 0.2571, + "step": 15356 + }, + { + "epoch": 0.93, + "learning_rate": 2.725677976261265e-07, + "loss": 0.2743, + "step": 15357 + }, + { + "epoch": 0.93, + "learning_rate": 2.7211426682057676e-07, + "loss": 0.2769, + "step": 15358 + }, + { + "epoch": 0.93, + "learning_rate": 2.716611084440246e-07, + "loss": 0.2697, + "step": 15359 + }, + { + "epoch": 0.93, + "learning_rate": 2.7120832251381623e-07, + "loss": 0.2648, + "step": 15360 + }, + { + "epoch": 0.93, + "learning_rate": 2.707559090472878e-07, + "loss": 0.2617, + "step": 15361 + }, + { + "epoch": 0.93, + "learning_rate": 2.703038680617576e-07, + "loss": 0.2513, + "step": 15362 + }, + { + "epoch": 0.93, + "learning_rate": 2.69852199574534e-07, + "loss": 0.2631, + "step": 15363 + }, + { + "epoch": 0.93, + "learning_rate": 2.6940090360290885e-07, + "loss": 0.258, + "step": 15364 + }, + { + "epoch": 0.93, + "learning_rate": 2.689499801641593e-07, + "loss": 0.2515, + "step": 15365 + }, + { + "epoch": 0.93, + "learning_rate": 2.684994292755494e-07, + "loss": 0.2762, + "step": 15366 + }, + { + "epoch": 0.93, + "learning_rate": 2.6804925095432756e-07, + "loss": 0.2734, + "step": 15367 + }, + { + "epoch": 0.93, + "learning_rate": 2.6759944521772774e-07, + "loss": 0.2688, + "step": 15368 + }, + { + "epoch": 0.93, + "learning_rate": 2.6715001208297286e-07, + "loss": 0.2891, + "step": 15369 + }, + { + "epoch": 0.93, + "learning_rate": 2.6670095156726917e-07, + "loss": 0.238, + "step": 15370 + }, + { + "epoch": 0.93, + "learning_rate": 2.662522636878073e-07, + "loss": 0.254, + "step": 15371 + }, + { + "epoch": 0.93, + "learning_rate": 2.658039484617658e-07, + "loss": 0.2652, + "step": 15372 + }, + { + "epoch": 0.93, + "learning_rate": 2.6535600590630763e-07, + "loss": 0.2506, + "step": 15373 + }, + { + "epoch": 0.93, + "learning_rate": 2.649084360385834e-07, + "loss": 0.2662, + "step": 15374 + }, + { + "epoch": 0.93, + "learning_rate": 2.644612388757284e-07, + "loss": 0.2803, + "step": 15375 + }, + { + "epoch": 0.93, + "learning_rate": 2.6401441443486107e-07, + "loss": 0.2612, + "step": 15376 + }, + { + "epoch": 0.93, + "learning_rate": 2.635679627330912e-07, + "loss": 0.2715, + "step": 15377 + }, + { + "epoch": 0.93, + "learning_rate": 2.631218837875071e-07, + "loss": 0.2593, + "step": 15378 + }, + { + "epoch": 0.93, + "learning_rate": 2.6267617761519095e-07, + "loss": 0.2888, + "step": 15379 + }, + { + "epoch": 0.93, + "learning_rate": 2.6223084423320445e-07, + "loss": 0.2659, + "step": 15380 + }, + { + "epoch": 0.93, + "learning_rate": 2.6178588365859735e-07, + "loss": 0.2591, + "step": 15381 + }, + { + "epoch": 0.93, + "learning_rate": 2.6134129590840273e-07, + "loss": 0.2768, + "step": 15382 + }, + { + "epoch": 0.93, + "learning_rate": 2.608970809996436e-07, + "loss": 0.2423, + "step": 15383 + }, + { + "epoch": 0.93, + "learning_rate": 2.604532389493275e-07, + "loss": 0.2585, + "step": 15384 + }, + { + "epoch": 0.93, + "learning_rate": 2.6000976977444526e-07, + "loss": 0.2638, + "step": 15385 + }, + { + "epoch": 0.93, + "learning_rate": 2.595666734919755e-07, + "loss": 0.2614, + "step": 15386 + }, + { + "epoch": 0.93, + "learning_rate": 2.5912395011888027e-07, + "loss": 0.291, + "step": 15387 + }, + { + "epoch": 0.93, + "learning_rate": 2.586815996721126e-07, + "loss": 0.2593, + "step": 15388 + }, + { + "epoch": 0.93, + "learning_rate": 2.5823962216860563e-07, + "loss": 0.2754, + "step": 15389 + }, + { + "epoch": 0.93, + "learning_rate": 2.5779801762527924e-07, + "loss": 0.2651, + "step": 15390 + }, + { + "epoch": 0.93, + "learning_rate": 2.57356786059042e-07, + "loss": 0.2683, + "step": 15391 + }, + { + "epoch": 0.93, + "learning_rate": 2.5691592748678607e-07, + "loss": 0.2547, + "step": 15392 + }, + { + "epoch": 0.93, + "learning_rate": 2.5647544192538785e-07, + "loss": 0.2712, + "step": 15393 + }, + { + "epoch": 0.93, + "learning_rate": 2.5603532939171394e-07, + "loss": 0.2665, + "step": 15394 + }, + { + "epoch": 0.93, + "learning_rate": 2.555955899026119e-07, + "loss": 0.2638, + "step": 15395 + }, + { + "epoch": 0.93, + "learning_rate": 2.551562234749183e-07, + "loss": 0.278, + "step": 15396 + }, + { + "epoch": 0.93, + "learning_rate": 2.54717230125453e-07, + "loss": 0.2577, + "step": 15397 + }, + { + "epoch": 0.93, + "learning_rate": 2.542786098710215e-07, + "loss": 0.2627, + "step": 15398 + }, + { + "epoch": 0.93, + "learning_rate": 2.5384036272842027e-07, + "loss": 0.2485, + "step": 15399 + }, + { + "epoch": 0.93, + "learning_rate": 2.5340248871442375e-07, + "loss": 0.2568, + "step": 15400 + }, + { + "epoch": 0.93, + "learning_rate": 2.529649878457985e-07, + "loss": 0.2683, + "step": 15401 + }, + { + "epoch": 0.93, + "learning_rate": 2.525278601392911e-07, + "loss": 0.2695, + "step": 15402 + }, + { + "epoch": 0.93, + "learning_rate": 2.520911056116393e-07, + "loss": 0.2623, + "step": 15403 + }, + { + "epoch": 0.93, + "learning_rate": 2.5165472427956306e-07, + "loss": 0.2352, + "step": 15404 + }, + { + "epoch": 0.93, + "learning_rate": 2.512187161597701e-07, + "loss": 0.2592, + "step": 15405 + }, + { + "epoch": 0.93, + "learning_rate": 2.507830812689516e-07, + "loss": 0.2688, + "step": 15406 + }, + { + "epoch": 0.93, + "learning_rate": 2.503478196237863e-07, + "loss": 0.2779, + "step": 15407 + }, + { + "epoch": 0.93, + "learning_rate": 2.499129312409365e-07, + "loss": 0.2552, + "step": 15408 + }, + { + "epoch": 0.93, + "learning_rate": 2.494784161370545e-07, + "loss": 0.2593, + "step": 15409 + }, + { + "epoch": 0.93, + "learning_rate": 2.490442743287746e-07, + "loss": 0.2649, + "step": 15410 + }, + { + "epoch": 0.93, + "learning_rate": 2.4861050583271706e-07, + "loss": 0.2927, + "step": 15411 + }, + { + "epoch": 0.93, + "learning_rate": 2.4817711066548953e-07, + "loss": 0.2695, + "step": 15412 + }, + { + "epoch": 0.93, + "learning_rate": 2.4774408884368216e-07, + "loss": 0.2649, + "step": 15413 + }, + { + "epoch": 0.93, + "learning_rate": 2.4731144038387723e-07, + "loss": 0.2451, + "step": 15414 + }, + { + "epoch": 0.93, + "learning_rate": 2.468791653026348e-07, + "loss": 0.2575, + "step": 15415 + }, + { + "epoch": 0.93, + "learning_rate": 2.4644726361650606e-07, + "loss": 0.239, + "step": 15416 + }, + { + "epoch": 0.93, + "learning_rate": 2.460157353420245e-07, + "loss": 0.2755, + "step": 15417 + }, + { + "epoch": 0.93, + "learning_rate": 2.4558458049571354e-07, + "loss": 0.2384, + "step": 15418 + }, + { + "epoch": 0.93, + "learning_rate": 2.451537990940789e-07, + "loss": 0.3131, + "step": 15419 + }, + { + "epoch": 0.93, + "learning_rate": 2.4472339115361176e-07, + "loss": 0.2652, + "step": 15420 + }, + { + "epoch": 0.93, + "learning_rate": 2.4429335669079123e-07, + "loss": 0.2486, + "step": 15421 + }, + { + "epoch": 0.93, + "learning_rate": 2.438636957220808e-07, + "loss": 0.2492, + "step": 15422 + }, + { + "epoch": 0.93, + "learning_rate": 2.434344082639295e-07, + "loss": 0.2446, + "step": 15423 + }, + { + "epoch": 0.93, + "learning_rate": 2.430054943327742e-07, + "loss": 0.2706, + "step": 15424 + }, + { + "epoch": 0.93, + "learning_rate": 2.4257695394503286e-07, + "loss": 0.2648, + "step": 15425 + }, + { + "epoch": 0.93, + "learning_rate": 2.4214878711711467e-07, + "loss": 0.2409, + "step": 15426 + }, + { + "epoch": 0.93, + "learning_rate": 2.4172099386541083e-07, + "loss": 0.2386, + "step": 15427 + }, + { + "epoch": 0.93, + "learning_rate": 2.4129357420629827e-07, + "loss": 0.2819, + "step": 15428 + }, + { + "epoch": 0.93, + "learning_rate": 2.4086652815614175e-07, + "loss": 0.2582, + "step": 15429 + }, + { + "epoch": 0.93, + "learning_rate": 2.4043985573129146e-07, + "loss": 0.2864, + "step": 15430 + }, + { + "epoch": 0.93, + "learning_rate": 2.400135569480799e-07, + "loss": 0.2663, + "step": 15431 + }, + { + "epoch": 0.93, + "learning_rate": 2.3958763182283075e-07, + "loss": 0.2863, + "step": 15432 + }, + { + "epoch": 0.93, + "learning_rate": 2.391620803718475e-07, + "loss": 0.284, + "step": 15433 + }, + { + "epoch": 0.93, + "learning_rate": 2.387369026114239e-07, + "loss": 0.2799, + "step": 15434 + }, + { + "epoch": 0.93, + "learning_rate": 2.38312098557838e-07, + "loss": 0.2612, + "step": 15435 + }, + { + "epoch": 0.93, + "learning_rate": 2.3788766822735344e-07, + "loss": 0.2779, + "step": 15436 + }, + { + "epoch": 0.93, + "learning_rate": 2.3746361163621723e-07, + "loss": 0.2641, + "step": 15437 + }, + { + "epoch": 0.93, + "learning_rate": 2.370399288006664e-07, + "loss": 0.2531, + "step": 15438 + }, + { + "epoch": 0.93, + "learning_rate": 2.3661661973692129e-07, + "loss": 0.2676, + "step": 15439 + }, + { + "epoch": 0.93, + "learning_rate": 2.3619368446118672e-07, + "loss": 0.2652, + "step": 15440 + }, + { + "epoch": 0.93, + "learning_rate": 2.3577112298965643e-07, + "loss": 0.2538, + "step": 15441 + }, + { + "epoch": 0.93, + "learning_rate": 2.3534893533850745e-07, + "loss": 0.258, + "step": 15442 + }, + { + "epoch": 0.93, + "learning_rate": 2.349271215239013e-07, + "loss": 0.2503, + "step": 15443 + }, + { + "epoch": 0.93, + "learning_rate": 2.3450568156199061e-07, + "loss": 0.2553, + "step": 15444 + }, + { + "epoch": 0.93, + "learning_rate": 2.3408461546890693e-07, + "loss": 0.2718, + "step": 15445 + }, + { + "epoch": 0.93, + "learning_rate": 2.3366392326077292e-07, + "loss": 0.2694, + "step": 15446 + }, + { + "epoch": 0.93, + "learning_rate": 2.3324360495369236e-07, + "loss": 0.2637, + "step": 15447 + }, + { + "epoch": 0.93, + "learning_rate": 2.3282366056375904e-07, + "loss": 0.2336, + "step": 15448 + }, + { + "epoch": 0.93, + "learning_rate": 2.32404090107049e-07, + "loss": 0.2708, + "step": 15449 + }, + { + "epoch": 0.93, + "learning_rate": 2.3198489359962606e-07, + "loss": 0.2426, + "step": 15450 + }, + { + "epoch": 0.93, + "learning_rate": 2.315660710575385e-07, + "loss": 0.2666, + "step": 15451 + }, + { + "epoch": 0.93, + "learning_rate": 2.3114762249682122e-07, + "loss": 0.2407, + "step": 15452 + }, + { + "epoch": 0.93, + "learning_rate": 2.3072954793349368e-07, + "loss": 0.2602, + "step": 15453 + }, + { + "epoch": 0.93, + "learning_rate": 2.3031184738356305e-07, + "loss": 0.2433, + "step": 15454 + }, + { + "epoch": 0.93, + "learning_rate": 2.29894520863021e-07, + "loss": 0.2808, + "step": 15455 + }, + { + "epoch": 0.93, + "learning_rate": 2.2947756838784252e-07, + "loss": 0.2815, + "step": 15456 + }, + { + "epoch": 0.93, + "learning_rate": 2.2906098997399151e-07, + "loss": 0.2515, + "step": 15457 + }, + { + "epoch": 0.93, + "learning_rate": 2.2864478563741853e-07, + "loss": 0.2593, + "step": 15458 + }, + { + "epoch": 0.93, + "learning_rate": 2.282289553940553e-07, + "loss": 0.2627, + "step": 15459 + }, + { + "epoch": 0.93, + "learning_rate": 2.2781349925982245e-07, + "loss": 0.2669, + "step": 15460 + }, + { + "epoch": 0.93, + "learning_rate": 2.2739841725062716e-07, + "loss": 0.2725, + "step": 15461 + }, + { + "epoch": 0.93, + "learning_rate": 2.2698370938235792e-07, + "loss": 0.2436, + "step": 15462 + }, + { + "epoch": 0.93, + "learning_rate": 2.265693756708931e-07, + "loss": 0.2588, + "step": 15463 + }, + { + "epoch": 0.93, + "learning_rate": 2.2615541613209558e-07, + "loss": 0.2821, + "step": 15464 + }, + { + "epoch": 0.93, + "learning_rate": 2.2574183078181377e-07, + "loss": 0.2633, + "step": 15465 + }, + { + "epoch": 0.93, + "learning_rate": 2.253286196358817e-07, + "loss": 0.2584, + "step": 15466 + }, + { + "epoch": 0.93, + "learning_rate": 2.2491578271011783e-07, + "loss": 0.2645, + "step": 15467 + }, + { + "epoch": 0.93, + "learning_rate": 2.2450332002032837e-07, + "loss": 0.2655, + "step": 15468 + }, + { + "epoch": 0.93, + "learning_rate": 2.2409123158230406e-07, + "loss": 0.26, + "step": 15469 + }, + { + "epoch": 0.93, + "learning_rate": 2.2367951741182225e-07, + "loss": 0.2561, + "step": 15470 + }, + { + "epoch": 0.93, + "learning_rate": 2.2326817752464593e-07, + "loss": 0.2818, + "step": 15471 + }, + { + "epoch": 0.93, + "learning_rate": 2.2285721193652022e-07, + "loss": 0.2806, + "step": 15472 + }, + { + "epoch": 0.93, + "learning_rate": 2.2244662066318146e-07, + "loss": 0.2718, + "step": 15473 + }, + { + "epoch": 0.93, + "learning_rate": 2.2203640372034707e-07, + "loss": 0.2475, + "step": 15474 + }, + { + "epoch": 0.93, + "learning_rate": 2.216265611237245e-07, + "loss": 0.2496, + "step": 15475 + }, + { + "epoch": 0.93, + "learning_rate": 2.2121709288900338e-07, + "loss": 0.2642, + "step": 15476 + }, + { + "epoch": 0.93, + "learning_rate": 2.20807999031859e-07, + "loss": 0.2747, + "step": 15477 + }, + { + "epoch": 0.94, + "learning_rate": 2.2039927956795327e-07, + "loss": 0.2655, + "step": 15478 + }, + { + "epoch": 0.94, + "learning_rate": 2.1999093451293586e-07, + "loss": 0.2372, + "step": 15479 + }, + { + "epoch": 0.94, + "learning_rate": 2.1958296388243873e-07, + "loss": 0.287, + "step": 15480 + }, + { + "epoch": 0.94, + "learning_rate": 2.191753676920816e-07, + "loss": 0.2656, + "step": 15481 + }, + { + "epoch": 0.94, + "learning_rate": 2.1876814595746864e-07, + "loss": 0.2702, + "step": 15482 + }, + { + "epoch": 0.94, + "learning_rate": 2.1836129869418964e-07, + "loss": 0.2501, + "step": 15483 + }, + { + "epoch": 0.94, + "learning_rate": 2.1795482591782213e-07, + "loss": 0.2776, + "step": 15484 + }, + { + "epoch": 0.94, + "learning_rate": 2.17548727643927e-07, + "loss": 0.2766, + "step": 15485 + }, + { + "epoch": 0.94, + "learning_rate": 2.1714300388805177e-07, + "loss": 0.2817, + "step": 15486 + }, + { + "epoch": 0.94, + "learning_rate": 2.1673765466572738e-07, + "loss": 0.2843, + "step": 15487 + }, + { + "epoch": 0.94, + "learning_rate": 2.1633267999247587e-07, + "loss": 0.2553, + "step": 15488 + }, + { + "epoch": 0.94, + "learning_rate": 2.1592807988380037e-07, + "loss": 0.2958, + "step": 15489 + }, + { + "epoch": 0.94, + "learning_rate": 2.1552385435518963e-07, + "loss": 0.269, + "step": 15490 + }, + { + "epoch": 0.94, + "learning_rate": 2.1512000342212125e-07, + "loss": 0.2427, + "step": 15491 + }, + { + "epoch": 0.94, + "learning_rate": 2.1471652710005397e-07, + "loss": 0.2606, + "step": 15492 + }, + { + "epoch": 0.94, + "learning_rate": 2.1431342540443767e-07, + "loss": 0.2381, + "step": 15493 + }, + { + "epoch": 0.94, + "learning_rate": 2.1391069835070333e-07, + "loss": 0.2418, + "step": 15494 + }, + { + "epoch": 0.94, + "learning_rate": 2.1350834595426972e-07, + "loss": 0.2422, + "step": 15495 + }, + { + "epoch": 0.94, + "learning_rate": 2.131063682305401e-07, + "loss": 0.2651, + "step": 15496 + }, + { + "epoch": 0.94, + "learning_rate": 2.1270476519490434e-07, + "loss": 0.2439, + "step": 15497 + }, + { + "epoch": 0.94, + "learning_rate": 2.1230353686273686e-07, + "loss": 0.2561, + "step": 15498 + }, + { + "epoch": 0.94, + "learning_rate": 2.119026832494009e-07, + "loss": 0.2478, + "step": 15499 + }, + { + "epoch": 0.94, + "learning_rate": 2.1150220437024195e-07, + "loss": 0.2799, + "step": 15500 + }, + { + "epoch": 0.94, + "learning_rate": 2.1110210024059108e-07, + "loss": 0.2535, + "step": 15501 + }, + { + "epoch": 0.94, + "learning_rate": 2.1070237087576827e-07, + "loss": 0.2598, + "step": 15502 + }, + { + "epoch": 0.94, + "learning_rate": 2.1030301629107353e-07, + "loss": 0.2941, + "step": 15503 + }, + { + "epoch": 0.94, + "learning_rate": 2.0990403650180013e-07, + "loss": 0.2638, + "step": 15504 + }, + { + "epoch": 0.94, + "learning_rate": 2.0950543152322035e-07, + "loss": 0.2436, + "step": 15505 + }, + { + "epoch": 0.94, + "learning_rate": 2.0910720137059527e-07, + "loss": 0.2593, + "step": 15506 + }, + { + "epoch": 0.94, + "learning_rate": 2.0870934605917047e-07, + "loss": 0.2596, + "step": 15507 + }, + { + "epoch": 0.94, + "learning_rate": 2.0831186560417826e-07, + "loss": 0.2546, + "step": 15508 + }, + { + "epoch": 0.94, + "learning_rate": 2.0791476002083642e-07, + "loss": 0.2586, + "step": 15509 + }, + { + "epoch": 0.94, + "learning_rate": 2.075180293243484e-07, + "loss": 0.2836, + "step": 15510 + }, + { + "epoch": 0.94, + "learning_rate": 2.0712167352990198e-07, + "loss": 0.2539, + "step": 15511 + }, + { + "epoch": 0.94, + "learning_rate": 2.0672569265267172e-07, + "loss": 0.2422, + "step": 15512 + }, + { + "epoch": 0.94, + "learning_rate": 2.063300867078155e-07, + "loss": 0.2609, + "step": 15513 + }, + { + "epoch": 0.94, + "learning_rate": 2.0593485571048343e-07, + "loss": 0.2614, + "step": 15514 + }, + { + "epoch": 0.94, + "learning_rate": 2.055399996758034e-07, + "loss": 0.2801, + "step": 15515 + }, + { + "epoch": 0.94, + "learning_rate": 2.0514551861889553e-07, + "loss": 0.2553, + "step": 15516 + }, + { + "epoch": 0.94, + "learning_rate": 2.0475141255485776e-07, + "loss": 0.2688, + "step": 15517 + }, + { + "epoch": 0.94, + "learning_rate": 2.0435768149878132e-07, + "loss": 0.2726, + "step": 15518 + }, + { + "epoch": 0.94, + "learning_rate": 2.0396432546573974e-07, + "loss": 0.2619, + "step": 15519 + }, + { + "epoch": 0.94, + "learning_rate": 2.0357134447079208e-07, + "loss": 0.2489, + "step": 15520 + }, + { + "epoch": 0.94, + "learning_rate": 2.0317873852898518e-07, + "loss": 0.2464, + "step": 15521 + }, + { + "epoch": 0.94, + "learning_rate": 2.027865076553459e-07, + "loss": 0.2489, + "step": 15522 + }, + { + "epoch": 0.94, + "learning_rate": 2.0239465186489447e-07, + "loss": 0.2591, + "step": 15523 + }, + { + "epoch": 0.94, + "learning_rate": 2.0200317117263225e-07, + "loss": 0.2771, + "step": 15524 + }, + { + "epoch": 0.94, + "learning_rate": 2.0161206559354607e-07, + "loss": 0.2521, + "step": 15525 + }, + { + "epoch": 0.94, + "learning_rate": 2.0122133514260956e-07, + "loss": 0.248, + "step": 15526 + }, + { + "epoch": 0.94, + "learning_rate": 2.0083097983478073e-07, + "loss": 0.2824, + "step": 15527 + }, + { + "epoch": 0.94, + "learning_rate": 2.0044099968500651e-07, + "loss": 0.2818, + "step": 15528 + }, + { + "epoch": 0.94, + "learning_rate": 2.000513947082161e-07, + "loss": 0.263, + "step": 15529 + }, + { + "epoch": 0.94, + "learning_rate": 1.996621649193242e-07, + "loss": 0.269, + "step": 15530 + }, + { + "epoch": 0.94, + "learning_rate": 1.992733103332345e-07, + "loss": 0.2231, + "step": 15531 + }, + { + "epoch": 0.94, + "learning_rate": 1.988848309648317e-07, + "loss": 0.2553, + "step": 15532 + }, + { + "epoch": 0.94, + "learning_rate": 1.9849672682898947e-07, + "loss": 0.2655, + "step": 15533 + }, + { + "epoch": 0.94, + "learning_rate": 1.9810899794056815e-07, + "loss": 0.2725, + "step": 15534 + }, + { + "epoch": 0.94, + "learning_rate": 1.9772164431441033e-07, + "loss": 0.2458, + "step": 15535 + }, + { + "epoch": 0.94, + "learning_rate": 1.9733466596534523e-07, + "loss": 0.25, + "step": 15536 + }, + { + "epoch": 0.94, + "learning_rate": 1.969480629081888e-07, + "loss": 0.2367, + "step": 15537 + }, + { + "epoch": 0.94, + "learning_rate": 1.965618351577414e-07, + "loss": 0.2592, + "step": 15538 + }, + { + "epoch": 0.94, + "learning_rate": 1.9617598272879123e-07, + "loss": 0.2506, + "step": 15539 + }, + { + "epoch": 0.94, + "learning_rate": 1.9579050563610868e-07, + "loss": 0.2686, + "step": 15540 + }, + { + "epoch": 0.94, + "learning_rate": 1.9540540389445306e-07, + "loss": 0.2584, + "step": 15541 + }, + { + "epoch": 0.94, + "learning_rate": 1.9502067751856702e-07, + "loss": 0.2853, + "step": 15542 + }, + { + "epoch": 0.94, + "learning_rate": 1.9463632652317876e-07, + "loss": 0.264, + "step": 15543 + }, + { + "epoch": 0.94, + "learning_rate": 1.9425235092300543e-07, + "loss": 0.2635, + "step": 15544 + }, + { + "epoch": 0.94, + "learning_rate": 1.9386875073274636e-07, + "loss": 0.2694, + "step": 15545 + }, + { + "epoch": 0.94, + "learning_rate": 1.9348552596708646e-07, + "loss": 0.2729, + "step": 15546 + }, + { + "epoch": 0.94, + "learning_rate": 1.931026766406985e-07, + "loss": 0.2651, + "step": 15547 + }, + { + "epoch": 0.94, + "learning_rate": 1.9272020276823955e-07, + "loss": 0.2681, + "step": 15548 + }, + { + "epoch": 0.94, + "learning_rate": 1.923381043643524e-07, + "loss": 0.2842, + "step": 15549 + }, + { + "epoch": 0.94, + "learning_rate": 1.9195638144366647e-07, + "loss": 0.2614, + "step": 15550 + }, + { + "epoch": 0.94, + "learning_rate": 1.9157503402079446e-07, + "loss": 0.2577, + "step": 15551 + }, + { + "epoch": 0.94, + "learning_rate": 1.9119406211033586e-07, + "loss": 0.2638, + "step": 15552 + }, + { + "epoch": 0.94, + "learning_rate": 1.9081346572687788e-07, + "loss": 0.2641, + "step": 15553 + }, + { + "epoch": 0.94, + "learning_rate": 1.9043324488498993e-07, + "loss": 0.2567, + "step": 15554 + }, + { + "epoch": 0.94, + "learning_rate": 1.900533995992293e-07, + "loss": 0.2898, + "step": 15555 + }, + { + "epoch": 0.94, + "learning_rate": 1.896739298841388e-07, + "loss": 0.2577, + "step": 15556 + }, + { + "epoch": 0.94, + "learning_rate": 1.8929483575424458e-07, + "loss": 0.2776, + "step": 15557 + }, + { + "epoch": 0.94, + "learning_rate": 1.889161172240628e-07, + "loss": 0.2408, + "step": 15558 + }, + { + "epoch": 0.94, + "learning_rate": 1.8853777430809073e-07, + "loss": 0.3101, + "step": 15559 + }, + { + "epoch": 0.94, + "learning_rate": 1.8815980702081237e-07, + "loss": 0.2871, + "step": 15560 + }, + { + "epoch": 0.94, + "learning_rate": 1.877822153767006e-07, + "loss": 0.2726, + "step": 15561 + }, + { + "epoch": 0.94, + "learning_rate": 1.8740499939020828e-07, + "loss": 0.2522, + "step": 15562 + }, + { + "epoch": 0.94, + "learning_rate": 1.8702815907577943e-07, + "loss": 0.2812, + "step": 15563 + }, + { + "epoch": 0.94, + "learning_rate": 1.8665169444784027e-07, + "loss": 0.2609, + "step": 15564 + }, + { + "epoch": 0.94, + "learning_rate": 1.8627560552080481e-07, + "loss": 0.279, + "step": 15565 + }, + { + "epoch": 0.94, + "learning_rate": 1.8589989230906935e-07, + "loss": 0.2525, + "step": 15566 + }, + { + "epoch": 0.94, + "learning_rate": 1.8552455482701903e-07, + "loss": 0.256, + "step": 15567 + }, + { + "epoch": 0.94, + "learning_rate": 1.8514959308902459e-07, + "loss": 0.2475, + "step": 15568 + }, + { + "epoch": 0.94, + "learning_rate": 1.8477500710944008e-07, + "loss": 0.2496, + "step": 15569 + }, + { + "epoch": 0.94, + "learning_rate": 1.8440079690260737e-07, + "loss": 0.2552, + "step": 15570 + }, + { + "epoch": 0.94, + "learning_rate": 1.8402696248285167e-07, + "loss": 0.2762, + "step": 15571 + }, + { + "epoch": 0.94, + "learning_rate": 1.83653503864486e-07, + "loss": 0.2607, + "step": 15572 + }, + { + "epoch": 0.94, + "learning_rate": 1.8328042106180665e-07, + "loss": 0.2465, + "step": 15573 + }, + { + "epoch": 0.94, + "learning_rate": 1.829077140890989e-07, + "loss": 0.2704, + "step": 15574 + }, + { + "epoch": 0.94, + "learning_rate": 1.8253538296063134e-07, + "loss": 0.2566, + "step": 15575 + }, + { + "epoch": 0.94, + "learning_rate": 1.821634276906581e-07, + "loss": 0.2693, + "step": 15576 + }, + { + "epoch": 0.94, + "learning_rate": 1.817918482934189e-07, + "loss": 0.2513, + "step": 15577 + }, + { + "epoch": 0.94, + "learning_rate": 1.8142064478314014e-07, + "loss": 0.2718, + "step": 15578 + }, + { + "epoch": 0.94, + "learning_rate": 1.8104981717403268e-07, + "loss": 0.2476, + "step": 15579 + }, + { + "epoch": 0.94, + "learning_rate": 1.8067936548029518e-07, + "loss": 0.2659, + "step": 15580 + }, + { + "epoch": 0.94, + "learning_rate": 1.8030928971610963e-07, + "loss": 0.2824, + "step": 15581 + }, + { + "epoch": 0.94, + "learning_rate": 1.7993958989564242e-07, + "loss": 0.2492, + "step": 15582 + }, + { + "epoch": 0.94, + "learning_rate": 1.7957026603304895e-07, + "loss": 0.2578, + "step": 15583 + }, + { + "epoch": 0.94, + "learning_rate": 1.79201318142469e-07, + "loss": 0.2656, + "step": 15584 + }, + { + "epoch": 0.94, + "learning_rate": 1.7883274623802793e-07, + "loss": 0.261, + "step": 15585 + }, + { + "epoch": 0.94, + "learning_rate": 1.7846455033383446e-07, + "loss": 0.267, + "step": 15586 + }, + { + "epoch": 0.94, + "learning_rate": 1.7809673044398511e-07, + "loss": 0.2806, + "step": 15587 + }, + { + "epoch": 0.94, + "learning_rate": 1.7772928658256305e-07, + "loss": 0.2946, + "step": 15588 + }, + { + "epoch": 0.94, + "learning_rate": 1.773622187636359e-07, + "loss": 0.2494, + "step": 15589 + }, + { + "epoch": 0.94, + "learning_rate": 1.7699552700125576e-07, + "loss": 0.2531, + "step": 15590 + }, + { + "epoch": 0.94, + "learning_rate": 1.7662921130946143e-07, + "loss": 0.2942, + "step": 15591 + }, + { + "epoch": 0.94, + "learning_rate": 1.7626327170227719e-07, + "loss": 0.2694, + "step": 15592 + }, + { + "epoch": 0.94, + "learning_rate": 1.75897708193713e-07, + "loss": 0.2761, + "step": 15593 + }, + { + "epoch": 0.94, + "learning_rate": 1.755325207977654e-07, + "loss": 0.2667, + "step": 15594 + }, + { + "epoch": 0.94, + "learning_rate": 1.7516770952841433e-07, + "loss": 0.2854, + "step": 15595 + }, + { + "epoch": 0.94, + "learning_rate": 1.748032743996264e-07, + "loss": 0.2595, + "step": 15596 + }, + { + "epoch": 0.94, + "learning_rate": 1.744392154253538e-07, + "loss": 0.2812, + "step": 15597 + }, + { + "epoch": 0.94, + "learning_rate": 1.7407553261953536e-07, + "loss": 0.2744, + "step": 15598 + }, + { + "epoch": 0.94, + "learning_rate": 1.737122259960944e-07, + "loss": 0.266, + "step": 15599 + }, + { + "epoch": 0.94, + "learning_rate": 1.7334929556893865e-07, + "loss": 0.2118, + "step": 15600 + }, + { + "epoch": 0.94, + "learning_rate": 1.729867413519648e-07, + "loss": 0.2886, + "step": 15601 + }, + { + "epoch": 0.94, + "learning_rate": 1.7262456335905065e-07, + "loss": 0.2667, + "step": 15602 + }, + { + "epoch": 0.94, + "learning_rate": 1.7226276160406397e-07, + "loss": 0.2662, + "step": 15603 + }, + { + "epoch": 0.94, + "learning_rate": 1.7190133610085592e-07, + "loss": 0.2757, + "step": 15604 + }, + { + "epoch": 0.94, + "learning_rate": 1.715402868632643e-07, + "loss": 0.2723, + "step": 15605 + }, + { + "epoch": 0.94, + "learning_rate": 1.711796139051103e-07, + "loss": 0.2531, + "step": 15606 + }, + { + "epoch": 0.94, + "learning_rate": 1.7081931724020174e-07, + "loss": 0.2616, + "step": 15607 + }, + { + "epoch": 0.94, + "learning_rate": 1.7045939688233316e-07, + "loss": 0.2712, + "step": 15608 + }, + { + "epoch": 0.94, + "learning_rate": 1.7009985284528574e-07, + "loss": 0.2585, + "step": 15609 + }, + { + "epoch": 0.94, + "learning_rate": 1.6974068514282294e-07, + "loss": 0.244, + "step": 15610 + }, + { + "epoch": 0.94, + "learning_rate": 1.6938189378869486e-07, + "loss": 0.2514, + "step": 15611 + }, + { + "epoch": 0.94, + "learning_rate": 1.6902347879663828e-07, + "loss": 0.2808, + "step": 15612 + }, + { + "epoch": 0.94, + "learning_rate": 1.6866544018037444e-07, + "loss": 0.2266, + "step": 15613 + }, + { + "epoch": 0.94, + "learning_rate": 1.6830777795361242e-07, + "loss": 0.2686, + "step": 15614 + }, + { + "epoch": 0.94, + "learning_rate": 1.6795049213004344e-07, + "loss": 0.266, + "step": 15615 + }, + { + "epoch": 0.94, + "learning_rate": 1.675935827233477e-07, + "loss": 0.2938, + "step": 15616 + }, + { + "epoch": 0.94, + "learning_rate": 1.6723704974718758e-07, + "loss": 0.267, + "step": 15617 + }, + { + "epoch": 0.94, + "learning_rate": 1.668808932152144e-07, + "loss": 0.2538, + "step": 15618 + }, + { + "epoch": 0.94, + "learning_rate": 1.6652511314106167e-07, + "loss": 0.2546, + "step": 15619 + }, + { + "epoch": 0.94, + "learning_rate": 1.6616970953835187e-07, + "loss": 0.2648, + "step": 15620 + }, + { + "epoch": 0.94, + "learning_rate": 1.6581468242069187e-07, + "loss": 0.2815, + "step": 15621 + }, + { + "epoch": 0.94, + "learning_rate": 1.654600318016708e-07, + "loss": 0.2794, + "step": 15622 + }, + { + "epoch": 0.94, + "learning_rate": 1.6510575769487114e-07, + "loss": 0.2402, + "step": 15623 + }, + { + "epoch": 0.94, + "learning_rate": 1.6475186011385201e-07, + "loss": 0.2546, + "step": 15624 + }, + { + "epoch": 0.94, + "learning_rate": 1.6439833907216374e-07, + "loss": 0.2732, + "step": 15625 + }, + { + "epoch": 0.94, + "learning_rate": 1.6404519458334211e-07, + "loss": 0.2651, + "step": 15626 + }, + { + "epoch": 0.94, + "learning_rate": 1.6369242666090412e-07, + "loss": 0.2436, + "step": 15627 + }, + { + "epoch": 0.94, + "learning_rate": 1.6334003531835784e-07, + "loss": 0.2612, + "step": 15628 + }, + { + "epoch": 0.94, + "learning_rate": 1.6298802056919362e-07, + "loss": 0.2467, + "step": 15629 + }, + { + "epoch": 0.94, + "learning_rate": 1.626363824268884e-07, + "loss": 0.2557, + "step": 15630 + }, + { + "epoch": 0.94, + "learning_rate": 1.622851209049048e-07, + "loss": 0.2969, + "step": 15631 + }, + { + "epoch": 0.94, + "learning_rate": 1.619342360166898e-07, + "loss": 0.2655, + "step": 15632 + }, + { + "epoch": 0.94, + "learning_rate": 1.615837277756771e-07, + "loss": 0.2455, + "step": 15633 + }, + { + "epoch": 0.94, + "learning_rate": 1.6123359619528712e-07, + "loss": 0.2821, + "step": 15634 + }, + { + "epoch": 0.94, + "learning_rate": 1.6088384128892353e-07, + "loss": 0.254, + "step": 15635 + }, + { + "epoch": 0.94, + "learning_rate": 1.6053446306997677e-07, + "loss": 0.2556, + "step": 15636 + }, + { + "epoch": 0.94, + "learning_rate": 1.601854615518217e-07, + "loss": 0.2716, + "step": 15637 + }, + { + "epoch": 0.94, + "learning_rate": 1.5983683674782203e-07, + "loss": 0.2599, + "step": 15638 + }, + { + "epoch": 0.94, + "learning_rate": 1.5948858867132265e-07, + "loss": 0.2778, + "step": 15639 + }, + { + "epoch": 0.94, + "learning_rate": 1.591407173356574e-07, + "loss": 0.2499, + "step": 15640 + }, + { + "epoch": 0.94, + "learning_rate": 1.5879322275414334e-07, + "loss": 0.2524, + "step": 15641 + }, + { + "epoch": 0.94, + "learning_rate": 1.584461049400854e-07, + "loss": 0.2825, + "step": 15642 + }, + { + "epoch": 0.95, + "learning_rate": 1.5809936390677073e-07, + "loss": 0.2684, + "step": 15643 + }, + { + "epoch": 0.95, + "learning_rate": 1.577529996674765e-07, + "loss": 0.266, + "step": 15644 + }, + { + "epoch": 0.95, + "learning_rate": 1.5740701223546208e-07, + "loss": 0.2506, + "step": 15645 + }, + { + "epoch": 0.95, + "learning_rate": 1.5706140162397355e-07, + "loss": 0.2797, + "step": 15646 + }, + { + "epoch": 0.95, + "learning_rate": 1.5671616784624365e-07, + "loss": 0.2669, + "step": 15647 + }, + { + "epoch": 0.95, + "learning_rate": 1.563713109154874e-07, + "loss": 0.2486, + "step": 15648 + }, + { + "epoch": 0.95, + "learning_rate": 1.5602683084490866e-07, + "loss": 0.2818, + "step": 15649 + }, + { + "epoch": 0.95, + "learning_rate": 1.5568272764769687e-07, + "loss": 0.2455, + "step": 15650 + }, + { + "epoch": 0.95, + "learning_rate": 1.5533900133702484e-07, + "loss": 0.2354, + "step": 15651 + }, + { + "epoch": 0.95, + "learning_rate": 1.5499565192605093e-07, + "loss": 0.261, + "step": 15652 + }, + { + "epoch": 0.95, + "learning_rate": 1.5465267942792129e-07, + "loss": 0.2609, + "step": 15653 + }, + { + "epoch": 0.95, + "learning_rate": 1.5431008385576762e-07, + "loss": 0.2923, + "step": 15654 + }, + { + "epoch": 0.95, + "learning_rate": 1.539678652227039e-07, + "loss": 0.2746, + "step": 15655 + }, + { + "epoch": 0.95, + "learning_rate": 1.5362602354183298e-07, + "loss": 0.2531, + "step": 15656 + }, + { + "epoch": 0.95, + "learning_rate": 1.5328455882624215e-07, + "loss": 0.2461, + "step": 15657 + }, + { + "epoch": 0.95, + "learning_rate": 1.5294347108900433e-07, + "loss": 0.2471, + "step": 15658 + }, + { + "epoch": 0.95, + "learning_rate": 1.526027603431779e-07, + "loss": 0.2729, + "step": 15659 + }, + { + "epoch": 0.95, + "learning_rate": 1.522624266018069e-07, + "loss": 0.254, + "step": 15660 + }, + { + "epoch": 0.95, + "learning_rate": 1.519224698779198e-07, + "loss": 0.2546, + "step": 15661 + }, + { + "epoch": 0.95, + "learning_rate": 1.5158289018453286e-07, + "loss": 0.2429, + "step": 15662 + }, + { + "epoch": 0.95, + "learning_rate": 1.512436875346468e-07, + "loss": 0.2354, + "step": 15663 + }, + { + "epoch": 0.95, + "learning_rate": 1.50904861941249e-07, + "loss": 0.2787, + "step": 15664 + }, + { + "epoch": 0.95, + "learning_rate": 1.5056641341730903e-07, + "loss": 0.2849, + "step": 15665 + }, + { + "epoch": 0.95, + "learning_rate": 1.502283419757844e-07, + "loss": 0.2514, + "step": 15666 + }, + { + "epoch": 0.95, + "learning_rate": 1.4989064762961914e-07, + "loss": 0.2533, + "step": 15667 + }, + { + "epoch": 0.95, + "learning_rate": 1.495533303917429e-07, + "loss": 0.26, + "step": 15668 + }, + { + "epoch": 0.95, + "learning_rate": 1.492163902750676e-07, + "loss": 0.2616, + "step": 15669 + }, + { + "epoch": 0.95, + "learning_rate": 1.4887982729249406e-07, + "loss": 0.2485, + "step": 15670 + }, + { + "epoch": 0.95, + "learning_rate": 1.4854364145690636e-07, + "loss": 0.2857, + "step": 15671 + }, + { + "epoch": 0.95, + "learning_rate": 1.4820783278117645e-07, + "loss": 0.2568, + "step": 15672 + }, + { + "epoch": 0.95, + "learning_rate": 1.478724012781596e-07, + "loss": 0.2489, + "step": 15673 + }, + { + "epoch": 0.95, + "learning_rate": 1.4753734696069888e-07, + "loss": 0.2852, + "step": 15674 + }, + { + "epoch": 0.95, + "learning_rate": 1.472026698416218e-07, + "loss": 0.2715, + "step": 15675 + }, + { + "epoch": 0.95, + "learning_rate": 1.4686836993374033e-07, + "loss": 0.2662, + "step": 15676 + }, + { + "epoch": 0.95, + "learning_rate": 1.465344472498531e-07, + "loss": 0.291, + "step": 15677 + }, + { + "epoch": 0.95, + "learning_rate": 1.4620090180274439e-07, + "loss": 0.267, + "step": 15678 + }, + { + "epoch": 0.95, + "learning_rate": 1.4586773360518503e-07, + "loss": 0.2566, + "step": 15679 + }, + { + "epoch": 0.95, + "learning_rate": 1.455349426699282e-07, + "loss": 0.28, + "step": 15680 + }, + { + "epoch": 0.95, + "learning_rate": 1.45202529009717e-07, + "loss": 0.2722, + "step": 15681 + }, + { + "epoch": 0.95, + "learning_rate": 1.448704926372757e-07, + "loss": 0.2588, + "step": 15682 + }, + { + "epoch": 0.95, + "learning_rate": 1.4453883356531752e-07, + "loss": 0.255, + "step": 15683 + }, + { + "epoch": 0.95, + "learning_rate": 1.442075518065389e-07, + "loss": 0.2857, + "step": 15684 + }, + { + "epoch": 0.95, + "learning_rate": 1.4387664737362417e-07, + "loss": 0.2999, + "step": 15685 + }, + { + "epoch": 0.95, + "learning_rate": 1.4354612027924209e-07, + "loss": 0.2653, + "step": 15686 + }, + { + "epoch": 0.95, + "learning_rate": 1.4321597053604364e-07, + "loss": 0.2607, + "step": 15687 + }, + { + "epoch": 0.95, + "learning_rate": 1.4288619815667203e-07, + "loss": 0.2684, + "step": 15688 + }, + { + "epoch": 0.95, + "learning_rate": 1.4255680315375164e-07, + "loss": 0.2552, + "step": 15689 + }, + { + "epoch": 0.95, + "learning_rate": 1.4222778553989124e-07, + "loss": 0.2524, + "step": 15690 + }, + { + "epoch": 0.95, + "learning_rate": 1.4189914532769078e-07, + "loss": 0.2624, + "step": 15691 + }, + { + "epoch": 0.95, + "learning_rate": 1.4157088252972796e-07, + "loss": 0.281, + "step": 15692 + }, + { + "epoch": 0.95, + "learning_rate": 1.412429971585738e-07, + "loss": 0.263, + "step": 15693 + }, + { + "epoch": 0.95, + "learning_rate": 1.4091548922677945e-07, + "loss": 0.2785, + "step": 15694 + }, + { + "epoch": 0.95, + "learning_rate": 1.405883587468837e-07, + "loss": 0.2748, + "step": 15695 + }, + { + "epoch": 0.95, + "learning_rate": 1.40261605731411e-07, + "loss": 0.2765, + "step": 15696 + }, + { + "epoch": 0.95, + "learning_rate": 1.3993523019286914e-07, + "loss": 0.2514, + "step": 15697 + }, + { + "epoch": 0.95, + "learning_rate": 1.3960923214375699e-07, + "loss": 0.2713, + "step": 15698 + }, + { + "epoch": 0.95, + "learning_rate": 1.3928361159655234e-07, + "loss": 0.2715, + "step": 15699 + }, + { + "epoch": 0.95, + "learning_rate": 1.3895836856372304e-07, + "loss": 0.2472, + "step": 15700 + }, + { + "epoch": 0.95, + "learning_rate": 1.3863350305772015e-07, + "loss": 0.2711, + "step": 15701 + }, + { + "epoch": 0.95, + "learning_rate": 1.3830901509097938e-07, + "loss": 0.2556, + "step": 15702 + }, + { + "epoch": 0.95, + "learning_rate": 1.379849046759274e-07, + "loss": 0.2621, + "step": 15703 + }, + { + "epoch": 0.95, + "learning_rate": 1.3766117182496985e-07, + "loss": 0.2581, + "step": 15704 + }, + { + "epoch": 0.95, + "learning_rate": 1.3733781655050127e-07, + "loss": 0.2487, + "step": 15705 + }, + { + "epoch": 0.95, + "learning_rate": 1.3701483886490286e-07, + "loss": 0.2758, + "step": 15706 + }, + { + "epoch": 0.95, + "learning_rate": 1.3669223878053584e-07, + "loss": 0.2786, + "step": 15707 + }, + { + "epoch": 0.95, + "learning_rate": 1.3637001630975587e-07, + "loss": 0.2642, + "step": 15708 + }, + { + "epoch": 0.95, + "learning_rate": 1.3604817146489534e-07, + "loss": 0.2575, + "step": 15709 + }, + { + "epoch": 0.95, + "learning_rate": 1.3572670425827773e-07, + "loss": 0.2574, + "step": 15710 + }, + { + "epoch": 0.95, + "learning_rate": 1.354056147022098e-07, + "loss": 0.239, + "step": 15711 + }, + { + "epoch": 0.95, + "learning_rate": 1.35084902808984e-07, + "loss": 0.2454, + "step": 15712 + }, + { + "epoch": 0.95, + "learning_rate": 1.3476456859087827e-07, + "loss": 0.2571, + "step": 15713 + }, + { + "epoch": 0.95, + "learning_rate": 1.3444461206015835e-07, + "loss": 0.2867, + "step": 15714 + }, + { + "epoch": 0.95, + "learning_rate": 1.3412503322907334e-07, + "loss": 0.2506, + "step": 15715 + }, + { + "epoch": 0.95, + "learning_rate": 1.3380583210985676e-07, + "loss": 0.2603, + "step": 15716 + }, + { + "epoch": 0.95, + "learning_rate": 1.3348700871472996e-07, + "loss": 0.2448, + "step": 15717 + }, + { + "epoch": 0.95, + "learning_rate": 1.3316856305589766e-07, + "loss": 0.2568, + "step": 15718 + }, + { + "epoch": 0.95, + "learning_rate": 1.3285049514555448e-07, + "loss": 0.263, + "step": 15719 + }, + { + "epoch": 0.95, + "learning_rate": 1.3253280499587518e-07, + "loss": 0.256, + "step": 15720 + }, + { + "epoch": 0.95, + "learning_rate": 1.3221549261902333e-07, + "loss": 0.2761, + "step": 15721 + }, + { + "epoch": 0.95, + "learning_rate": 1.3189855802714592e-07, + "loss": 0.2522, + "step": 15722 + }, + { + "epoch": 0.95, + "learning_rate": 1.3158200123237764e-07, + "loss": 0.2571, + "step": 15723 + }, + { + "epoch": 0.95, + "learning_rate": 1.3126582224683882e-07, + "loss": 0.2694, + "step": 15724 + }, + { + "epoch": 0.95, + "learning_rate": 1.30950021082632e-07, + "loss": 0.2845, + "step": 15725 + }, + { + "epoch": 0.95, + "learning_rate": 1.3063459775184974e-07, + "loss": 0.2475, + "step": 15726 + }, + { + "epoch": 0.95, + "learning_rate": 1.3031955226656567e-07, + "loss": 0.2626, + "step": 15727 + }, + { + "epoch": 0.95, + "learning_rate": 1.3000488463884242e-07, + "loss": 0.2612, + "step": 15728 + }, + { + "epoch": 0.95, + "learning_rate": 1.296905948807281e-07, + "loss": 0.278, + "step": 15729 + }, + { + "epoch": 0.95, + "learning_rate": 1.293766830042531e-07, + "loss": 0.2514, + "step": 15730 + }, + { + "epoch": 0.95, + "learning_rate": 1.290631490214367e-07, + "loss": 0.2482, + "step": 15731 + }, + { + "epoch": 0.95, + "learning_rate": 1.2874999294428147e-07, + "loss": 0.2701, + "step": 15732 + }, + { + "epoch": 0.95, + "learning_rate": 1.284372147847779e-07, + "loss": 0.2673, + "step": 15733 + }, + { + "epoch": 0.95, + "learning_rate": 1.2812481455489968e-07, + "loss": 0.2641, + "step": 15734 + }, + { + "epoch": 0.95, + "learning_rate": 1.2781279226660616e-07, + "loss": 0.2906, + "step": 15735 + }, + { + "epoch": 0.95, + "learning_rate": 1.2750114793184554e-07, + "loss": 0.2609, + "step": 15736 + }, + { + "epoch": 0.95, + "learning_rate": 1.271898815625461e-07, + "loss": 0.2766, + "step": 15737 + }, + { + "epoch": 0.95, + "learning_rate": 1.26878993170626e-07, + "loss": 0.2395, + "step": 15738 + }, + { + "epoch": 0.95, + "learning_rate": 1.2656848276798804e-07, + "loss": 0.3009, + "step": 15739 + }, + { + "epoch": 0.95, + "learning_rate": 1.2625835036651935e-07, + "loss": 0.2485, + "step": 15740 + }, + { + "epoch": 0.95, + "learning_rate": 1.259485959780926e-07, + "loss": 0.2807, + "step": 15741 + }, + { + "epoch": 0.95, + "learning_rate": 1.2563921961456727e-07, + "loss": 0.2473, + "step": 15742 + }, + { + "epoch": 0.95, + "learning_rate": 1.2533022128778826e-07, + "loss": 0.2595, + "step": 15743 + }, + { + "epoch": 0.95, + "learning_rate": 1.2502160100958505e-07, + "loss": 0.2819, + "step": 15744 + }, + { + "epoch": 0.95, + "learning_rate": 1.2471335879177259e-07, + "loss": 0.2658, + "step": 15745 + }, + { + "epoch": 0.95, + "learning_rate": 1.244054946461537e-07, + "loss": 0.2743, + "step": 15746 + }, + { + "epoch": 0.95, + "learning_rate": 1.2409800858451227e-07, + "loss": 0.2822, + "step": 15747 + }, + { + "epoch": 0.95, + "learning_rate": 1.2379090061861999e-07, + "loss": 0.2503, + "step": 15748 + }, + { + "epoch": 0.95, + "learning_rate": 1.2348417076023743e-07, + "loss": 0.2468, + "step": 15749 + }, + { + "epoch": 0.95, + "learning_rate": 1.2317781902110637e-07, + "loss": 0.2665, + "step": 15750 + }, + { + "epoch": 0.95, + "learning_rate": 1.2287184541295515e-07, + "loss": 0.2635, + "step": 15751 + }, + { + "epoch": 0.95, + "learning_rate": 1.2256624994749666e-07, + "loss": 0.262, + "step": 15752 + }, + { + "epoch": 0.95, + "learning_rate": 1.2226103263643263e-07, + "loss": 0.2532, + "step": 15753 + }, + { + "epoch": 0.95, + "learning_rate": 1.2195619349144706e-07, + "loss": 0.269, + "step": 15754 + }, + { + "epoch": 0.95, + "learning_rate": 1.2165173252421058e-07, + "loss": 0.2591, + "step": 15755 + }, + { + "epoch": 0.95, + "learning_rate": 1.2134764974638058e-07, + "loss": 0.2932, + "step": 15756 + }, + { + "epoch": 0.95, + "learning_rate": 1.210439451695955e-07, + "loss": 0.2905, + "step": 15757 + }, + { + "epoch": 0.95, + "learning_rate": 1.2074061880548715e-07, + "loss": 0.2566, + "step": 15758 + }, + { + "epoch": 0.95, + "learning_rate": 1.2043767066566514e-07, + "loss": 0.2501, + "step": 15759 + }, + { + "epoch": 0.95, + "learning_rate": 1.2013510076172797e-07, + "loss": 0.2612, + "step": 15760 + }, + { + "epoch": 0.95, + "learning_rate": 1.198329091052608e-07, + "loss": 0.2684, + "step": 15761 + }, + { + "epoch": 0.95, + "learning_rate": 1.1953109570783218e-07, + "loss": 0.2626, + "step": 15762 + }, + { + "epoch": 0.95, + "learning_rate": 1.1922966058099618e-07, + "loss": 0.2507, + "step": 15763 + }, + { + "epoch": 0.95, + "learning_rate": 1.1892860373629467e-07, + "loss": 0.268, + "step": 15764 + }, + { + "epoch": 0.95, + "learning_rate": 1.186279251852529e-07, + "loss": 0.2857, + "step": 15765 + }, + { + "epoch": 0.95, + "learning_rate": 1.1832762493938276e-07, + "loss": 0.2829, + "step": 15766 + }, + { + "epoch": 0.95, + "learning_rate": 1.1802770301017841e-07, + "loss": 0.2563, + "step": 15767 + }, + { + "epoch": 0.95, + "learning_rate": 1.1772815940912618e-07, + "loss": 0.2508, + "step": 15768 + }, + { + "epoch": 0.95, + "learning_rate": 1.1742899414769138e-07, + "loss": 0.287, + "step": 15769 + }, + { + "epoch": 0.95, + "learning_rate": 1.1713020723732926e-07, + "loss": 0.2475, + "step": 15770 + }, + { + "epoch": 0.95, + "learning_rate": 1.1683179868947736e-07, + "loss": 0.256, + "step": 15771 + }, + { + "epoch": 0.95, + "learning_rate": 1.1653376851555986e-07, + "loss": 0.2814, + "step": 15772 + }, + { + "epoch": 0.95, + "learning_rate": 1.1623611672698765e-07, + "loss": 0.2369, + "step": 15773 + }, + { + "epoch": 0.95, + "learning_rate": 1.1593884333515604e-07, + "loss": 0.2753, + "step": 15774 + }, + { + "epoch": 0.95, + "learning_rate": 1.1564194835144704e-07, + "loss": 0.2374, + "step": 15775 + }, + { + "epoch": 0.95, + "learning_rate": 1.1534543178722601e-07, + "loss": 0.2543, + "step": 15776 + }, + { + "epoch": 0.95, + "learning_rate": 1.1504929365384387e-07, + "loss": 0.2492, + "step": 15777 + }, + { + "epoch": 0.95, + "learning_rate": 1.1475353396264043e-07, + "loss": 0.2342, + "step": 15778 + }, + { + "epoch": 0.95, + "learning_rate": 1.1445815272493777e-07, + "loss": 0.264, + "step": 15779 + }, + { + "epoch": 0.95, + "learning_rate": 1.1416314995204458e-07, + "loss": 0.2598, + "step": 15780 + }, + { + "epoch": 0.95, + "learning_rate": 1.138685256552552e-07, + "loss": 0.26, + "step": 15781 + }, + { + "epoch": 0.95, + "learning_rate": 1.1357427984584835e-07, + "loss": 0.2655, + "step": 15782 + }, + { + "epoch": 0.95, + "learning_rate": 1.1328041253508948e-07, + "loss": 0.2662, + "step": 15783 + }, + { + "epoch": 0.95, + "learning_rate": 1.1298692373422959e-07, + "loss": 0.273, + "step": 15784 + }, + { + "epoch": 0.95, + "learning_rate": 1.1269381345450525e-07, + "loss": 0.2651, + "step": 15785 + }, + { + "epoch": 0.95, + "learning_rate": 1.1240108170713637e-07, + "loss": 0.2718, + "step": 15786 + }, + { + "epoch": 0.95, + "learning_rate": 1.1210872850333287e-07, + "loss": 0.2634, + "step": 15787 + }, + { + "epoch": 0.95, + "learning_rate": 1.1181675385428469e-07, + "loss": 0.2409, + "step": 15788 + }, + { + "epoch": 0.95, + "learning_rate": 1.1152515777117068e-07, + "loss": 0.2453, + "step": 15789 + }, + { + "epoch": 0.95, + "learning_rate": 1.1123394026515522e-07, + "loss": 0.2457, + "step": 15790 + }, + { + "epoch": 0.95, + "learning_rate": 1.1094310134738718e-07, + "loss": 0.2578, + "step": 15791 + }, + { + "epoch": 0.95, + "learning_rate": 1.1065264102900098e-07, + "loss": 0.2665, + "step": 15792 + }, + { + "epoch": 0.95, + "learning_rate": 1.1036255932111773e-07, + "loss": 0.2667, + "step": 15793 + }, + { + "epoch": 0.95, + "learning_rate": 1.1007285623484187e-07, + "loss": 0.2652, + "step": 15794 + }, + { + "epoch": 0.95, + "learning_rate": 1.0978353178126455e-07, + "loss": 0.2701, + "step": 15795 + }, + { + "epoch": 0.95, + "learning_rate": 1.0949458597146356e-07, + "loss": 0.2398, + "step": 15796 + }, + { + "epoch": 0.95, + "learning_rate": 1.0920601881650006e-07, + "loss": 0.2599, + "step": 15797 + }, + { + "epoch": 0.95, + "learning_rate": 1.0891783032742298e-07, + "loss": 0.2482, + "step": 15798 + }, + { + "epoch": 0.95, + "learning_rate": 1.0863002051526461e-07, + "loss": 0.2471, + "step": 15799 + }, + { + "epoch": 0.95, + "learning_rate": 1.083425893910428e-07, + "loss": 0.2545, + "step": 15800 + }, + { + "epoch": 0.95, + "learning_rate": 1.0805553696576432e-07, + "loss": 0.261, + "step": 15801 + }, + { + "epoch": 0.95, + "learning_rate": 1.0776886325041592e-07, + "loss": 0.2556, + "step": 15802 + }, + { + "epoch": 0.95, + "learning_rate": 1.0748256825597547e-07, + "loss": 0.2931, + "step": 15803 + }, + { + "epoch": 0.95, + "learning_rate": 1.071966519934009e-07, + "loss": 0.2405, + "step": 15804 + }, + { + "epoch": 0.95, + "learning_rate": 1.0691111447364122e-07, + "loss": 0.2494, + "step": 15805 + }, + { + "epoch": 0.95, + "learning_rate": 1.0662595570762768e-07, + "loss": 0.2483, + "step": 15806 + }, + { + "epoch": 0.95, + "learning_rate": 1.0634117570627489e-07, + "loss": 0.2701, + "step": 15807 + }, + { + "epoch": 0.95, + "learning_rate": 1.0605677448048746e-07, + "loss": 0.2553, + "step": 15808 + }, + { + "epoch": 0.96, + "learning_rate": 1.0577275204115445e-07, + "loss": 0.2557, + "step": 15809 + }, + { + "epoch": 0.96, + "learning_rate": 1.0548910839914938e-07, + "loss": 0.2928, + "step": 15810 + }, + { + "epoch": 0.96, + "learning_rate": 1.0520584356532915e-07, + "loss": 0.2693, + "step": 15811 + }, + { + "epoch": 0.96, + "learning_rate": 1.049229575505406e-07, + "loss": 0.2415, + "step": 15812 + }, + { + "epoch": 0.96, + "learning_rate": 1.0464045036561287e-07, + "loss": 0.2575, + "step": 15813 + }, + { + "epoch": 0.96, + "learning_rate": 1.0435832202136286e-07, + "loss": 0.2545, + "step": 15814 + }, + { + "epoch": 0.96, + "learning_rate": 1.0407657252859083e-07, + "loss": 0.2461, + "step": 15815 + }, + { + "epoch": 0.96, + "learning_rate": 1.0379520189808367e-07, + "loss": 0.2684, + "step": 15816 + }, + { + "epoch": 0.96, + "learning_rate": 1.0351421014061281e-07, + "loss": 0.2461, + "step": 15817 + }, + { + "epoch": 0.96, + "learning_rate": 1.0323359726693849e-07, + "loss": 0.2408, + "step": 15818 + }, + { + "epoch": 0.96, + "learning_rate": 1.0295336328780103e-07, + "loss": 0.2528, + "step": 15819 + }, + { + "epoch": 0.96, + "learning_rate": 1.0267350821392963e-07, + "loss": 0.2698, + "step": 15820 + }, + { + "epoch": 0.96, + "learning_rate": 1.0239403205604014e-07, + "loss": 0.2649, + "step": 15821 + }, + { + "epoch": 0.96, + "learning_rate": 1.021149348248307e-07, + "loss": 0.2807, + "step": 15822 + }, + { + "epoch": 0.96, + "learning_rate": 1.0183621653098607e-07, + "loss": 0.2686, + "step": 15823 + }, + { + "epoch": 0.96, + "learning_rate": 1.0155787718517884e-07, + "loss": 0.2422, + "step": 15824 + }, + { + "epoch": 0.96, + "learning_rate": 1.0127991679806271e-07, + "loss": 0.2623, + "step": 15825 + }, + { + "epoch": 0.96, + "learning_rate": 1.0100233538028138e-07, + "loss": 0.2695, + "step": 15826 + }, + { + "epoch": 0.96, + "learning_rate": 1.0072513294246078e-07, + "loss": 0.2584, + "step": 15827 + }, + { + "epoch": 0.96, + "learning_rate": 1.0044830949521356e-07, + "loss": 0.2704, + "step": 15828 + }, + { + "epoch": 0.96, + "learning_rate": 1.0017186504913901e-07, + "loss": 0.2649, + "step": 15829 + }, + { + "epoch": 0.96, + "learning_rate": 9.989579961481866e-08, + "loss": 0.2662, + "step": 15830 + }, + { + "epoch": 0.96, + "learning_rate": 9.962011320282405e-08, + "loss": 0.2885, + "step": 15831 + }, + { + "epoch": 0.96, + "learning_rate": 9.934480582370676e-08, + "loss": 0.2595, + "step": 15832 + }, + { + "epoch": 0.96, + "learning_rate": 9.906987748800945e-08, + "loss": 0.2715, + "step": 15833 + }, + { + "epoch": 0.96, + "learning_rate": 9.879532820625704e-08, + "loss": 0.2776, + "step": 15834 + }, + { + "epoch": 0.96, + "learning_rate": 9.852115798896e-08, + "loss": 0.2619, + "step": 15835 + }, + { + "epoch": 0.96, + "learning_rate": 9.824736684661552e-08, + "loss": 0.2684, + "step": 15836 + }, + { + "epoch": 0.96, + "learning_rate": 9.797395478970406e-08, + "loss": 0.2414, + "step": 15837 + }, + { + "epoch": 0.96, + "learning_rate": 9.770092182869505e-08, + "loss": 0.2755, + "step": 15838 + }, + { + "epoch": 0.96, + "learning_rate": 9.742826797404126e-08, + "loss": 0.2554, + "step": 15839 + }, + { + "epoch": 0.96, + "learning_rate": 9.715599323617985e-08, + "loss": 0.2762, + "step": 15840 + }, + { + "epoch": 0.96, + "learning_rate": 9.688409762553586e-08, + "loss": 0.2736, + "step": 15841 + }, + { + "epoch": 0.96, + "learning_rate": 9.66125811525176e-08, + "loss": 0.2486, + "step": 15842 + }, + { + "epoch": 0.96, + "learning_rate": 9.63414438275212e-08, + "loss": 0.2697, + "step": 15843 + }, + { + "epoch": 0.96, + "learning_rate": 9.607068566092615e-08, + "loss": 0.2412, + "step": 15844 + }, + { + "epoch": 0.96, + "learning_rate": 9.58003066630997e-08, + "loss": 0.2582, + "step": 15845 + }, + { + "epoch": 0.96, + "learning_rate": 9.553030684439135e-08, + "loss": 0.2868, + "step": 15846 + }, + { + "epoch": 0.96, + "learning_rate": 9.526068621513951e-08, + "loss": 0.268, + "step": 15847 + }, + { + "epoch": 0.96, + "learning_rate": 9.499144478566592e-08, + "loss": 0.2606, + "step": 15848 + }, + { + "epoch": 0.96, + "learning_rate": 9.472258256627898e-08, + "loss": 0.2592, + "step": 15849 + }, + { + "epoch": 0.96, + "learning_rate": 9.445409956727048e-08, + "loss": 0.2439, + "step": 15850 + }, + { + "epoch": 0.96, + "learning_rate": 9.418599579892107e-08, + "loss": 0.254, + "step": 15851 + }, + { + "epoch": 0.96, + "learning_rate": 9.391827127149256e-08, + "loss": 0.2408, + "step": 15852 + }, + { + "epoch": 0.96, + "learning_rate": 9.365092599523784e-08, + "loss": 0.2835, + "step": 15853 + }, + { + "epoch": 0.96, + "learning_rate": 9.338395998038985e-08, + "loss": 0.2699, + "step": 15854 + }, + { + "epoch": 0.96, + "learning_rate": 9.31173732371704e-08, + "loss": 0.2613, + "step": 15855 + }, + { + "epoch": 0.96, + "learning_rate": 9.28511657757858e-08, + "loss": 0.2317, + "step": 15856 + }, + { + "epoch": 0.96, + "learning_rate": 9.258533760642562e-08, + "loss": 0.2796, + "step": 15857 + }, + { + "epoch": 0.96, + "learning_rate": 9.231988873926956e-08, + "loss": 0.2506, + "step": 15858 + }, + { + "epoch": 0.96, + "learning_rate": 9.205481918447945e-08, + "loss": 0.2494, + "step": 15859 + }, + { + "epoch": 0.96, + "learning_rate": 9.179012895220273e-08, + "loss": 0.2546, + "step": 15860 + }, + { + "epoch": 0.96, + "learning_rate": 9.152581805257355e-08, + "loss": 0.25, + "step": 15861 + }, + { + "epoch": 0.96, + "learning_rate": 9.126188649571044e-08, + "loss": 0.2404, + "step": 15862 + }, + { + "epoch": 0.96, + "learning_rate": 9.099833429171868e-08, + "loss": 0.2524, + "step": 15863 + }, + { + "epoch": 0.96, + "learning_rate": 9.073516145068905e-08, + "loss": 0.2603, + "step": 15864 + }, + { + "epoch": 0.96, + "learning_rate": 9.047236798269466e-08, + "loss": 0.2546, + "step": 15865 + }, + { + "epoch": 0.96, + "learning_rate": 9.020995389779852e-08, + "loss": 0.2542, + "step": 15866 + }, + { + "epoch": 0.96, + "learning_rate": 8.994791920604484e-08, + "loss": 0.2529, + "step": 15867 + }, + { + "epoch": 0.96, + "learning_rate": 8.968626391746893e-08, + "loss": 0.2513, + "step": 15868 + }, + { + "epoch": 0.96, + "learning_rate": 8.9424988042085e-08, + "loss": 0.2741, + "step": 15869 + }, + { + "epoch": 0.96, + "learning_rate": 8.916409158989836e-08, + "loss": 0.2574, + "step": 15870 + }, + { + "epoch": 0.96, + "learning_rate": 8.890357457089548e-08, + "loss": 0.2804, + "step": 15871 + }, + { + "epoch": 0.96, + "learning_rate": 8.864343699505062e-08, + "loss": 0.2797, + "step": 15872 + }, + { + "epoch": 0.96, + "learning_rate": 8.838367887232358e-08, + "loss": 0.2686, + "step": 15873 + }, + { + "epoch": 0.96, + "learning_rate": 8.812430021265861e-08, + "loss": 0.2743, + "step": 15874 + }, + { + "epoch": 0.96, + "learning_rate": 8.786530102598667e-08, + "loss": 0.2556, + "step": 15875 + }, + { + "epoch": 0.96, + "learning_rate": 8.760668132222206e-08, + "loss": 0.2578, + "step": 15876 + }, + { + "epoch": 0.96, + "learning_rate": 8.734844111126795e-08, + "loss": 0.2467, + "step": 15877 + }, + { + "epoch": 0.96, + "learning_rate": 8.709058040300866e-08, + "loss": 0.2773, + "step": 15878 + }, + { + "epoch": 0.96, + "learning_rate": 8.683309920731852e-08, + "loss": 0.2479, + "step": 15879 + }, + { + "epoch": 0.96, + "learning_rate": 8.65759975340541e-08, + "loss": 0.272, + "step": 15880 + }, + { + "epoch": 0.96, + "learning_rate": 8.631927539305862e-08, + "loss": 0.2612, + "step": 15881 + }, + { + "epoch": 0.96, + "learning_rate": 8.606293279415979e-08, + "loss": 0.2624, + "step": 15882 + }, + { + "epoch": 0.96, + "learning_rate": 8.580696974717195e-08, + "loss": 0.2582, + "step": 15883 + }, + { + "epoch": 0.96, + "learning_rate": 8.555138626189619e-08, + "loss": 0.26, + "step": 15884 + }, + { + "epoch": 0.96, + "learning_rate": 8.529618234811577e-08, + "loss": 0.2257, + "step": 15885 + }, + { + "epoch": 0.96, + "learning_rate": 8.504135801560065e-08, + "loss": 0.2493, + "step": 15886 + }, + { + "epoch": 0.96, + "learning_rate": 8.478691327410749e-08, + "loss": 0.2577, + "step": 15887 + }, + { + "epoch": 0.96, + "learning_rate": 8.453284813337848e-08, + "loss": 0.2669, + "step": 15888 + }, + { + "epoch": 0.96, + "learning_rate": 8.427916260313917e-08, + "loss": 0.2775, + "step": 15889 + }, + { + "epoch": 0.96, + "learning_rate": 8.402585669310293e-08, + "loss": 0.2681, + "step": 15890 + }, + { + "epoch": 0.96, + "learning_rate": 8.377293041296642e-08, + "loss": 0.3026, + "step": 15891 + }, + { + "epoch": 0.96, + "learning_rate": 8.3520383772413e-08, + "loss": 0.2828, + "step": 15892 + }, + { + "epoch": 0.96, + "learning_rate": 8.326821678111163e-08, + "loss": 0.2747, + "step": 15893 + }, + { + "epoch": 0.96, + "learning_rate": 8.301642944871679e-08, + "loss": 0.2468, + "step": 15894 + }, + { + "epoch": 0.96, + "learning_rate": 8.276502178486745e-08, + "loss": 0.2388, + "step": 15895 + }, + { + "epoch": 0.96, + "learning_rate": 8.251399379918923e-08, + "loss": 0.2649, + "step": 15896 + }, + { + "epoch": 0.96, + "learning_rate": 8.226334550129111e-08, + "loss": 0.2408, + "step": 15897 + }, + { + "epoch": 0.96, + "learning_rate": 8.201307690077099e-08, + "loss": 0.2595, + "step": 15898 + }, + { + "epoch": 0.96, + "learning_rate": 8.176318800721006e-08, + "loss": 0.2676, + "step": 15899 + }, + { + "epoch": 0.96, + "learning_rate": 8.151367883017402e-08, + "loss": 0.2478, + "step": 15900 + }, + { + "epoch": 0.96, + "learning_rate": 8.126454937921746e-08, + "loss": 0.2926, + "step": 15901 + }, + { + "epoch": 0.96, + "learning_rate": 8.101579966387496e-08, + "loss": 0.2755, + "step": 15902 + }, + { + "epoch": 0.96, + "learning_rate": 8.076742969367224e-08, + "loss": 0.2515, + "step": 15903 + }, + { + "epoch": 0.96, + "learning_rate": 8.051943947811836e-08, + "loss": 0.2706, + "step": 15904 + }, + { + "epoch": 0.96, + "learning_rate": 8.027182902670571e-08, + "loss": 0.2646, + "step": 15905 + }, + { + "epoch": 0.96, + "learning_rate": 8.002459834891562e-08, + "loss": 0.272, + "step": 15906 + }, + { + "epoch": 0.96, + "learning_rate": 7.977774745421163e-08, + "loss": 0.2613, + "step": 15907 + }, + { + "epoch": 0.96, + "learning_rate": 7.953127635204616e-08, + "loss": 0.2518, + "step": 15908 + }, + { + "epoch": 0.96, + "learning_rate": 7.92851850518539e-08, + "loss": 0.2511, + "step": 15909 + }, + { + "epoch": 0.96, + "learning_rate": 7.90394735630573e-08, + "loss": 0.2384, + "step": 15910 + }, + { + "epoch": 0.96, + "learning_rate": 7.87941418950633e-08, + "loss": 0.2338, + "step": 15911 + }, + { + "epoch": 0.96, + "learning_rate": 7.854919005726325e-08, + "loss": 0.2723, + "step": 15912 + }, + { + "epoch": 0.96, + "learning_rate": 7.830461805903522e-08, + "loss": 0.2585, + "step": 15913 + }, + { + "epoch": 0.96, + "learning_rate": 7.806042590974394e-08, + "loss": 0.2967, + "step": 15914 + }, + { + "epoch": 0.96, + "learning_rate": 7.781661361873749e-08, + "loss": 0.2796, + "step": 15915 + }, + { + "epoch": 0.96, + "learning_rate": 7.757318119534951e-08, + "loss": 0.2704, + "step": 15916 + }, + { + "epoch": 0.96, + "learning_rate": 7.733012864890033e-08, + "loss": 0.2754, + "step": 15917 + }, + { + "epoch": 0.96, + "learning_rate": 7.708745598869582e-08, + "loss": 0.2953, + "step": 15918 + }, + { + "epoch": 0.96, + "learning_rate": 7.684516322402524e-08, + "loss": 0.2786, + "step": 15919 + }, + { + "epoch": 0.96, + "learning_rate": 7.660325036416561e-08, + "loss": 0.2856, + "step": 15920 + }, + { + "epoch": 0.96, + "learning_rate": 7.636171741837838e-08, + "loss": 0.2794, + "step": 15921 + }, + { + "epoch": 0.96, + "learning_rate": 7.612056439590953e-08, + "loss": 0.2906, + "step": 15922 + }, + { + "epoch": 0.96, + "learning_rate": 7.587979130599277e-08, + "loss": 0.2893, + "step": 15923 + }, + { + "epoch": 0.96, + "learning_rate": 7.563939815784738e-08, + "loss": 0.2676, + "step": 15924 + }, + { + "epoch": 0.96, + "learning_rate": 7.539938496067268e-08, + "loss": 0.2658, + "step": 15925 + }, + { + "epoch": 0.96, + "learning_rate": 7.515975172366019e-08, + "loss": 0.2676, + "step": 15926 + }, + { + "epoch": 0.96, + "learning_rate": 7.492049845598481e-08, + "loss": 0.2631, + "step": 15927 + }, + { + "epoch": 0.96, + "learning_rate": 7.468162516680478e-08, + "loss": 0.2513, + "step": 15928 + }, + { + "epoch": 0.96, + "learning_rate": 7.444313186526608e-08, + "loss": 0.269, + "step": 15929 + }, + { + "epoch": 0.96, + "learning_rate": 7.420501856049922e-08, + "loss": 0.2448, + "step": 15930 + }, + { + "epoch": 0.96, + "learning_rate": 7.396728526161911e-08, + "loss": 0.2653, + "step": 15931 + }, + { + "epoch": 0.96, + "learning_rate": 7.372993197772849e-08, + "loss": 0.2659, + "step": 15932 + }, + { + "epoch": 0.96, + "learning_rate": 7.349295871791452e-08, + "loss": 0.245, + "step": 15933 + }, + { + "epoch": 0.96, + "learning_rate": 7.325636549124993e-08, + "loss": 0.2832, + "step": 15934 + }, + { + "epoch": 0.96, + "learning_rate": 7.302015230679194e-08, + "loss": 0.2623, + "step": 15935 + }, + { + "epoch": 0.96, + "learning_rate": 7.278431917358331e-08, + "loss": 0.268, + "step": 15936 + }, + { + "epoch": 0.96, + "learning_rate": 7.254886610065237e-08, + "loss": 0.2582, + "step": 15937 + }, + { + "epoch": 0.96, + "learning_rate": 7.231379309701636e-08, + "loss": 0.2624, + "step": 15938 + }, + { + "epoch": 0.96, + "learning_rate": 7.207910017167252e-08, + "loss": 0.2563, + "step": 15939 + }, + { + "epoch": 0.96, + "learning_rate": 7.1844787333607e-08, + "loss": 0.2518, + "step": 15940 + }, + { + "epoch": 0.96, + "learning_rate": 7.161085459178929e-08, + "loss": 0.2912, + "step": 15941 + }, + { + "epoch": 0.96, + "learning_rate": 7.137730195517556e-08, + "loss": 0.2582, + "step": 15942 + }, + { + "epoch": 0.96, + "learning_rate": 7.114412943270866e-08, + "loss": 0.2602, + "step": 15943 + }, + { + "epoch": 0.96, + "learning_rate": 7.091133703331477e-08, + "loss": 0.2613, + "step": 15944 + }, + { + "epoch": 0.96, + "learning_rate": 7.067892476590676e-08, + "loss": 0.2634, + "step": 15945 + }, + { + "epoch": 0.96, + "learning_rate": 7.044689263938088e-08, + "loss": 0.2665, + "step": 15946 + }, + { + "epoch": 0.96, + "learning_rate": 7.021524066262108e-08, + "loss": 0.2588, + "step": 15947 + }, + { + "epoch": 0.96, + "learning_rate": 6.998396884449699e-08, + "loss": 0.2547, + "step": 15948 + }, + { + "epoch": 0.96, + "learning_rate": 6.975307719386259e-08, + "loss": 0.2722, + "step": 15949 + }, + { + "epoch": 0.96, + "learning_rate": 6.95225657195564e-08, + "loss": 0.2733, + "step": 15950 + }, + { + "epoch": 0.96, + "learning_rate": 6.929243443040467e-08, + "loss": 0.2719, + "step": 15951 + }, + { + "epoch": 0.96, + "learning_rate": 6.906268333521704e-08, + "loss": 0.264, + "step": 15952 + }, + { + "epoch": 0.96, + "learning_rate": 6.88333124427909e-08, + "loss": 0.2633, + "step": 15953 + }, + { + "epoch": 0.96, + "learning_rate": 6.860432176190591e-08, + "loss": 0.2693, + "step": 15954 + }, + { + "epoch": 0.96, + "learning_rate": 6.837571130132947e-08, + "loss": 0.2849, + "step": 15955 + }, + { + "epoch": 0.96, + "learning_rate": 6.814748106981462e-08, + "loss": 0.2726, + "step": 15956 + }, + { + "epoch": 0.96, + "learning_rate": 6.791963107609655e-08, + "loss": 0.2612, + "step": 15957 + }, + { + "epoch": 0.96, + "learning_rate": 6.769216132890277e-08, + "loss": 0.2783, + "step": 15958 + }, + { + "epoch": 0.96, + "learning_rate": 6.746507183693851e-08, + "loss": 0.2605, + "step": 15959 + }, + { + "epoch": 0.96, + "learning_rate": 6.723836260889793e-08, + "loss": 0.2754, + "step": 15960 + }, + { + "epoch": 0.96, + "learning_rate": 6.701203365346188e-08, + "loss": 0.2695, + "step": 15961 + }, + { + "epoch": 0.96, + "learning_rate": 6.678608497929451e-08, + "loss": 0.2635, + "step": 15962 + }, + { + "epoch": 0.96, + "learning_rate": 6.65605165950467e-08, + "loss": 0.2662, + "step": 15963 + }, + { + "epoch": 0.96, + "learning_rate": 6.633532850935486e-08, + "loss": 0.2257, + "step": 15964 + }, + { + "epoch": 0.96, + "learning_rate": 6.611052073083768e-08, + "loss": 0.246, + "step": 15965 + }, + { + "epoch": 0.96, + "learning_rate": 6.588609326810491e-08, + "loss": 0.26, + "step": 15966 + }, + { + "epoch": 0.96, + "learning_rate": 6.566204612974635e-08, + "loss": 0.272, + "step": 15967 + }, + { + "epoch": 0.96, + "learning_rate": 6.543837932434071e-08, + "loss": 0.2815, + "step": 15968 + }, + { + "epoch": 0.96, + "learning_rate": 6.521509286045114e-08, + "loss": 0.2531, + "step": 15969 + }, + { + "epoch": 0.96, + "learning_rate": 6.499218674662633e-08, + "loss": 0.2839, + "step": 15970 + }, + { + "epoch": 0.96, + "learning_rate": 6.476966099139836e-08, + "loss": 0.251, + "step": 15971 + }, + { + "epoch": 0.96, + "learning_rate": 6.45475156032882e-08, + "loss": 0.2613, + "step": 15972 + }, + { + "epoch": 0.96, + "learning_rate": 6.432575059080016e-08, + "loss": 0.2465, + "step": 15973 + }, + { + "epoch": 0.97, + "learning_rate": 6.41043659624252e-08, + "loss": 0.2436, + "step": 15974 + }, + { + "epoch": 0.97, + "learning_rate": 6.388336172663767e-08, + "loss": 0.2379, + "step": 15975 + }, + { + "epoch": 0.97, + "learning_rate": 6.366273789189969e-08, + "loss": 0.2504, + "step": 15976 + }, + { + "epoch": 0.97, + "learning_rate": 6.344249446665673e-08, + "loss": 0.2634, + "step": 15977 + }, + { + "epoch": 0.97, + "learning_rate": 6.322263145934204e-08, + "loss": 0.2818, + "step": 15978 + }, + { + "epoch": 0.97, + "learning_rate": 6.300314887837222e-08, + "loss": 0.2705, + "step": 15979 + }, + { + "epoch": 0.97, + "learning_rate": 6.278404673214944e-08, + "loss": 0.2616, + "step": 15980 + }, + { + "epoch": 0.97, + "learning_rate": 6.256532502906365e-08, + "loss": 0.2718, + "step": 15981 + }, + { + "epoch": 0.97, + "learning_rate": 6.234698377748704e-08, + "loss": 0.2584, + "step": 15982 + }, + { + "epoch": 0.97, + "learning_rate": 6.21290229857796e-08, + "loss": 0.2581, + "step": 15983 + }, + { + "epoch": 0.97, + "learning_rate": 6.191144266228467e-08, + "loss": 0.269, + "step": 15984 + }, + { + "epoch": 0.97, + "learning_rate": 6.169424281533442e-08, + "loss": 0.2609, + "step": 15985 + }, + { + "epoch": 0.97, + "learning_rate": 6.147742345324226e-08, + "loss": 0.2633, + "step": 15986 + }, + { + "epoch": 0.97, + "learning_rate": 6.12609845843093e-08, + "loss": 0.2778, + "step": 15987 + }, + { + "epoch": 0.97, + "learning_rate": 6.104492621682335e-08, + "loss": 0.2461, + "step": 15988 + }, + { + "epoch": 0.97, + "learning_rate": 6.082924835905446e-08, + "loss": 0.2708, + "step": 15989 + }, + { + "epoch": 0.97, + "learning_rate": 6.061395101926048e-08, + "loss": 0.2614, + "step": 15990 + }, + { + "epoch": 0.97, + "learning_rate": 6.039903420568483e-08, + "loss": 0.2659, + "step": 15991 + }, + { + "epoch": 0.97, + "learning_rate": 6.018449792655423e-08, + "loss": 0.2489, + "step": 15992 + }, + { + "epoch": 0.97, + "learning_rate": 5.997034219008213e-08, + "loss": 0.2514, + "step": 15993 + }, + { + "epoch": 0.97, + "learning_rate": 5.975656700446753e-08, + "loss": 0.2443, + "step": 15994 + }, + { + "epoch": 0.97, + "learning_rate": 5.954317237789609e-08, + "loss": 0.2806, + "step": 15995 + }, + { + "epoch": 0.97, + "learning_rate": 5.9330158318535726e-08, + "loss": 0.2468, + "step": 15996 + }, + { + "epoch": 0.97, + "learning_rate": 5.911752483454214e-08, + "loss": 0.2394, + "step": 15997 + }, + { + "epoch": 0.97, + "learning_rate": 5.89052719340566e-08, + "loss": 0.263, + "step": 15998 + }, + { + "epoch": 0.97, + "learning_rate": 5.869339962520482e-08, + "loss": 0.2732, + "step": 15999 + }, + { + "epoch": 0.97, + "learning_rate": 5.84819079160992e-08, + "loss": 0.2716, + "step": 16000 + }, + { + "epoch": 0.97, + "learning_rate": 5.8270796814834387e-08, + "loss": 0.2624, + "step": 16001 + }, + { + "epoch": 0.97, + "learning_rate": 5.8060066329493904e-08, + "loss": 0.2501, + "step": 16002 + }, + { + "epoch": 0.97, + "learning_rate": 5.784971646814575e-08, + "loss": 0.2826, + "step": 16003 + }, + { + "epoch": 0.97, + "learning_rate": 5.7639747238843466e-08, + "loss": 0.2821, + "step": 16004 + }, + { + "epoch": 0.97, + "learning_rate": 5.743015864962398e-08, + "loss": 0.2462, + "step": 16005 + }, + { + "epoch": 0.97, + "learning_rate": 5.722095070851308e-08, + "loss": 0.2649, + "step": 16006 + }, + { + "epoch": 0.97, + "learning_rate": 5.701212342351992e-08, + "loss": 0.2924, + "step": 16007 + }, + { + "epoch": 0.97, + "learning_rate": 5.68036768026381e-08, + "loss": 0.2638, + "step": 16008 + }, + { + "epoch": 0.97, + "learning_rate": 5.6595610853849016e-08, + "loss": 0.2753, + "step": 16009 + }, + { + "epoch": 0.97, + "learning_rate": 5.638792558511963e-08, + "loss": 0.261, + "step": 16010 + }, + { + "epoch": 0.97, + "learning_rate": 5.618062100439803e-08, + "loss": 0.2486, + "step": 16011 + }, + { + "epoch": 0.97, + "learning_rate": 5.5973697119623414e-08, + "loss": 0.2546, + "step": 16012 + }, + { + "epoch": 0.97, + "learning_rate": 5.576715393871612e-08, + "loss": 0.2473, + "step": 16013 + }, + { + "epoch": 0.97, + "learning_rate": 5.556099146958538e-08, + "loss": 0.2641, + "step": 16014 + }, + { + "epoch": 0.97, + "learning_rate": 5.535520972012265e-08, + "loss": 0.2613, + "step": 16015 + }, + { + "epoch": 0.97, + "learning_rate": 5.5149808698206075e-08, + "loss": 0.2467, + "step": 16016 + }, + { + "epoch": 0.97, + "learning_rate": 5.4944788411700475e-08, + "loss": 0.2732, + "step": 16017 + }, + { + "epoch": 0.97, + "learning_rate": 5.4740148868454027e-08, + "loss": 0.2706, + "step": 16018 + }, + { + "epoch": 0.97, + "learning_rate": 5.4535890076302664e-08, + "loss": 0.2517, + "step": 16019 + }, + { + "epoch": 0.97, + "learning_rate": 5.4332012043064595e-08, + "loss": 0.2698, + "step": 16020 + }, + { + "epoch": 0.97, + "learning_rate": 5.4128514776545795e-08, + "loss": 0.2483, + "step": 16021 + }, + { + "epoch": 0.97, + "learning_rate": 5.3925398284536687e-08, + "loss": 0.2768, + "step": 16022 + }, + { + "epoch": 0.97, + "learning_rate": 5.372266257481551e-08, + "loss": 0.2617, + "step": 16023 + }, + { + "epoch": 0.97, + "learning_rate": 5.352030765514271e-08, + "loss": 0.2595, + "step": 16024 + }, + { + "epoch": 0.97, + "learning_rate": 5.331833353326432e-08, + "loss": 0.2619, + "step": 16025 + }, + { + "epoch": 0.97, + "learning_rate": 5.3116740216914154e-08, + "loss": 0.2644, + "step": 16026 + }, + { + "epoch": 0.97, + "learning_rate": 5.291552771380937e-08, + "loss": 0.2508, + "step": 16027 + }, + { + "epoch": 0.97, + "learning_rate": 5.27146960316538e-08, + "loss": 0.2659, + "step": 16028 + }, + { + "epoch": 0.97, + "learning_rate": 5.251424517813686e-08, + "loss": 0.2439, + "step": 16029 + }, + { + "epoch": 0.97, + "learning_rate": 5.231417516093129e-08, + "loss": 0.2987, + "step": 16030 + }, + { + "epoch": 0.97, + "learning_rate": 5.2114485987697636e-08, + "loss": 0.264, + "step": 16031 + }, + { + "epoch": 0.97, + "learning_rate": 5.1915177666080894e-08, + "loss": 0.2659, + "step": 16032 + }, + { + "epoch": 0.97, + "learning_rate": 5.1716250203710516e-08, + "loss": 0.256, + "step": 16033 + }, + { + "epoch": 0.97, + "learning_rate": 5.151770360820485e-08, + "loss": 0.2775, + "step": 16034 + }, + { + "epoch": 0.97, + "learning_rate": 5.1319537887162265e-08, + "loss": 0.2559, + "step": 16035 + }, + { + "epoch": 0.97, + "learning_rate": 5.112175304817113e-08, + "loss": 0.2598, + "step": 16036 + }, + { + "epoch": 0.97, + "learning_rate": 5.092434909880317e-08, + "loss": 0.2515, + "step": 16037 + }, + { + "epoch": 0.97, + "learning_rate": 5.072732604661568e-08, + "loss": 0.2799, + "step": 16038 + }, + { + "epoch": 0.97, + "learning_rate": 5.053068389915261e-08, + "loss": 0.2487, + "step": 16039 + }, + { + "epoch": 0.97, + "learning_rate": 5.0334422663941285e-08, + "loss": 0.2506, + "step": 16040 + }, + { + "epoch": 0.97, + "learning_rate": 5.013854234849569e-08, + "loss": 0.264, + "step": 16041 + }, + { + "epoch": 0.97, + "learning_rate": 4.994304296031538e-08, + "loss": 0.2861, + "step": 16042 + }, + { + "epoch": 0.97, + "learning_rate": 4.9747924506884373e-08, + "loss": 0.2867, + "step": 16043 + }, + { + "epoch": 0.97, + "learning_rate": 4.9553186995674465e-08, + "loss": 0.2757, + "step": 16044 + }, + { + "epoch": 0.97, + "learning_rate": 4.935883043413858e-08, + "loss": 0.2804, + "step": 16045 + }, + { + "epoch": 0.97, + "learning_rate": 4.9164854829719664e-08, + "loss": 0.2457, + "step": 16046 + }, + { + "epoch": 0.97, + "learning_rate": 4.8971260189841775e-08, + "loss": 0.264, + "step": 16047 + }, + { + "epoch": 0.97, + "learning_rate": 4.8778046521920086e-08, + "loss": 0.3009, + "step": 16048 + }, + { + "epoch": 0.97, + "learning_rate": 4.8585213833348686e-08, + "loss": 0.2766, + "step": 16049 + }, + { + "epoch": 0.97, + "learning_rate": 4.839276213151056e-08, + "loss": 0.2872, + "step": 16050 + }, + { + "epoch": 0.97, + "learning_rate": 4.820069142377537e-08, + "loss": 0.2476, + "step": 16051 + }, + { + "epoch": 0.97, + "learning_rate": 4.8009001717493896e-08, + "loss": 0.2737, + "step": 16052 + }, + { + "epoch": 0.97, + "learning_rate": 4.781769302000694e-08, + "loss": 0.267, + "step": 16053 + }, + { + "epoch": 0.97, + "learning_rate": 4.762676533863864e-08, + "loss": 0.2688, + "step": 16054 + }, + { + "epoch": 0.97, + "learning_rate": 4.7436218680697586e-08, + "loss": 0.3033, + "step": 16055 + }, + { + "epoch": 0.97, + "learning_rate": 4.724605305347907e-08, + "loss": 0.2574, + "step": 16056 + }, + { + "epoch": 0.97, + "learning_rate": 4.705626846426392e-08, + "loss": 0.2612, + "step": 16057 + }, + { + "epoch": 0.97, + "learning_rate": 4.6866864920317445e-08, + "loss": 0.2598, + "step": 16058 + }, + { + "epoch": 0.97, + "learning_rate": 4.667784242889162e-08, + "loss": 0.2624, + "step": 16059 + }, + { + "epoch": 0.97, + "learning_rate": 4.648920099722287e-08, + "loss": 0.2927, + "step": 16060 + }, + { + "epoch": 0.97, + "learning_rate": 4.630094063253321e-08, + "loss": 0.25, + "step": 16061 + }, + { + "epoch": 0.97, + "learning_rate": 4.6113061342029085e-08, + "loss": 0.2828, + "step": 16062 + }, + { + "epoch": 0.97, + "learning_rate": 4.592556313290586e-08, + "loss": 0.2641, + "step": 16063 + }, + { + "epoch": 0.97, + "learning_rate": 4.57384460123389e-08, + "loss": 0.2628, + "step": 16064 + }, + { + "epoch": 0.97, + "learning_rate": 4.5551709987494693e-08, + "loss": 0.2669, + "step": 16065 + }, + { + "epoch": 0.97, + "learning_rate": 4.536535506551976e-08, + "loss": 0.2642, + "step": 16066 + }, + { + "epoch": 0.97, + "learning_rate": 4.5179381253550594e-08, + "loss": 0.2539, + "step": 16067 + }, + { + "epoch": 0.97, + "learning_rate": 4.4993788558705954e-08, + "loss": 0.2681, + "step": 16068 + }, + { + "epoch": 0.97, + "learning_rate": 4.480857698809127e-08, + "loss": 0.2439, + "step": 16069 + }, + { + "epoch": 0.97, + "learning_rate": 4.4623746548798643e-08, + "loss": 0.2602, + "step": 16070 + }, + { + "epoch": 0.97, + "learning_rate": 4.4439297247903525e-08, + "loss": 0.257, + "step": 16071 + }, + { + "epoch": 0.97, + "learning_rate": 4.425522909246693e-08, + "loss": 0.2646, + "step": 16072 + }, + { + "epoch": 0.97, + "learning_rate": 4.407154208953546e-08, + "loss": 0.2726, + "step": 16073 + }, + { + "epoch": 0.97, + "learning_rate": 4.3888236246143465e-08, + "loss": 0.2513, + "step": 16074 + }, + { + "epoch": 0.97, + "learning_rate": 4.3705311569307576e-08, + "loss": 0.2775, + "step": 16075 + }, + { + "epoch": 0.97, + "learning_rate": 4.3522768066029955e-08, + "loss": 0.2839, + "step": 16076 + }, + { + "epoch": 0.97, + "learning_rate": 4.334060574330057e-08, + "loss": 0.262, + "step": 16077 + }, + { + "epoch": 0.97, + "learning_rate": 4.315882460809384e-08, + "loss": 0.2595, + "step": 16078 + }, + { + "epoch": 0.97, + "learning_rate": 4.297742466736754e-08, + "loss": 0.2658, + "step": 16079 + }, + { + "epoch": 0.97, + "learning_rate": 4.279640592806833e-08, + "loss": 0.274, + "step": 16080 + }, + { + "epoch": 0.97, + "learning_rate": 4.261576839712511e-08, + "loss": 0.2581, + "step": 16081 + }, + { + "epoch": 0.97, + "learning_rate": 4.243551208145458e-08, + "loss": 0.2517, + "step": 16082 + }, + { + "epoch": 0.97, + "learning_rate": 4.225563698795676e-08, + "loss": 0.2981, + "step": 16083 + }, + { + "epoch": 0.97, + "learning_rate": 4.2076143123518376e-08, + "loss": 0.2593, + "step": 16084 + }, + { + "epoch": 0.97, + "learning_rate": 4.18970304950117e-08, + "loss": 0.2718, + "step": 16085 + }, + { + "epoch": 0.97, + "learning_rate": 4.171829910929459e-08, + "loss": 0.2888, + "step": 16086 + }, + { + "epoch": 0.97, + "learning_rate": 4.1539948973207124e-08, + "loss": 0.2663, + "step": 16087 + }, + { + "epoch": 0.97, + "learning_rate": 4.13619800935805e-08, + "loss": 0.2496, + "step": 16088 + }, + { + "epoch": 0.97, + "learning_rate": 4.118439247722705e-08, + "loss": 0.2794, + "step": 16089 + }, + { + "epoch": 0.97, + "learning_rate": 4.1007186130944674e-08, + "loss": 0.2552, + "step": 16090 + }, + { + "epoch": 0.97, + "learning_rate": 4.083036106151905e-08, + "loss": 0.2386, + "step": 16091 + }, + { + "epoch": 0.97, + "learning_rate": 4.0653917275719215e-08, + "loss": 0.2723, + "step": 16092 + }, + { + "epoch": 0.97, + "learning_rate": 4.0477854780299754e-08, + "loss": 0.2471, + "step": 16093 + }, + { + "epoch": 0.97, + "learning_rate": 4.030217358200306e-08, + "loss": 0.2934, + "step": 16094 + }, + { + "epoch": 0.97, + "learning_rate": 4.012687368755264e-08, + "loss": 0.258, + "step": 16095 + }, + { + "epoch": 0.97, + "learning_rate": 3.99519551036609e-08, + "loss": 0.2479, + "step": 16096 + }, + { + "epoch": 0.97, + "learning_rate": 3.9777417837024714e-08, + "loss": 0.2646, + "step": 16097 + }, + { + "epoch": 0.97, + "learning_rate": 3.9603261894326504e-08, + "loss": 0.2455, + "step": 16098 + }, + { + "epoch": 0.97, + "learning_rate": 3.942948728223317e-08, + "loss": 0.2722, + "step": 16099 + }, + { + "epoch": 0.97, + "learning_rate": 3.925609400739716e-08, + "loss": 0.245, + "step": 16100 + }, + { + "epoch": 0.97, + "learning_rate": 3.9083082076457614e-08, + "loss": 0.2651, + "step": 16101 + }, + { + "epoch": 0.97, + "learning_rate": 3.8910451496037e-08, + "loss": 0.2478, + "step": 16102 + }, + { + "epoch": 0.97, + "learning_rate": 3.873820227274672e-08, + "loss": 0.2755, + "step": 16103 + }, + { + "epoch": 0.97, + "learning_rate": 3.8566334413179254e-08, + "loss": 0.2771, + "step": 16104 + }, + { + "epoch": 0.97, + "learning_rate": 3.839484792391601e-08, + "loss": 0.2747, + "step": 16105 + }, + { + "epoch": 0.97, + "learning_rate": 3.822374281152064e-08, + "loss": 0.2662, + "step": 16106 + }, + { + "epoch": 0.97, + "learning_rate": 3.805301908254455e-08, + "loss": 0.2822, + "step": 16107 + }, + { + "epoch": 0.97, + "learning_rate": 3.788267674352475e-08, + "loss": 0.2669, + "step": 16108 + }, + { + "epoch": 0.97, + "learning_rate": 3.771271580098157e-08, + "loss": 0.2761, + "step": 16109 + }, + { + "epoch": 0.97, + "learning_rate": 3.754313626142203e-08, + "loss": 0.2902, + "step": 16110 + }, + { + "epoch": 0.97, + "learning_rate": 3.7373938131339825e-08, + "loss": 0.246, + "step": 16111 + }, + { + "epoch": 0.97, + "learning_rate": 3.7205121417209774e-08, + "loss": 0.248, + "step": 16112 + }, + { + "epoch": 0.97, + "learning_rate": 3.703668612549782e-08, + "loss": 0.2422, + "step": 16113 + }, + { + "epoch": 0.97, + "learning_rate": 3.6868632262651026e-08, + "loss": 0.2676, + "step": 16114 + }, + { + "epoch": 0.97, + "learning_rate": 3.6700959835104244e-08, + "loss": 0.252, + "step": 16115 + }, + { + "epoch": 0.97, + "learning_rate": 3.6533668849275674e-08, + "loss": 0.2508, + "step": 16116 + }, + { + "epoch": 0.97, + "learning_rate": 3.636675931157019e-08, + "loss": 0.2475, + "step": 16117 + }, + { + "epoch": 0.97, + "learning_rate": 3.620023122837824e-08, + "loss": 0.2525, + "step": 16118 + }, + { + "epoch": 0.97, + "learning_rate": 3.603408460607583e-08, + "loss": 0.2415, + "step": 16119 + }, + { + "epoch": 0.97, + "learning_rate": 3.5868319451023425e-08, + "loss": 0.2655, + "step": 16120 + }, + { + "epoch": 0.97, + "learning_rate": 3.570293576956596e-08, + "loss": 0.2349, + "step": 16121 + }, + { + "epoch": 0.97, + "learning_rate": 3.553793356803614e-08, + "loss": 0.2709, + "step": 16122 + }, + { + "epoch": 0.97, + "learning_rate": 3.537331285275225e-08, + "loss": 0.2934, + "step": 16123 + }, + { + "epoch": 0.97, + "learning_rate": 3.52090736300148e-08, + "loss": 0.2726, + "step": 16124 + }, + { + "epoch": 0.97, + "learning_rate": 3.5045215906113204e-08, + "loss": 0.2554, + "step": 16125 + }, + { + "epoch": 0.97, + "learning_rate": 3.4881739687319115e-08, + "loss": 0.2577, + "step": 16126 + }, + { + "epoch": 0.97, + "learning_rate": 3.471864497989197e-08, + "loss": 0.2573, + "step": 16127 + }, + { + "epoch": 0.97, + "learning_rate": 3.4555931790075656e-08, + "loss": 0.2612, + "step": 16128 + }, + { + "epoch": 0.97, + "learning_rate": 3.439360012410076e-08, + "loss": 0.2607, + "step": 16129 + }, + { + "epoch": 0.97, + "learning_rate": 3.423164998818007e-08, + "loss": 0.2302, + "step": 16130 + }, + { + "epoch": 0.97, + "learning_rate": 3.407008138851531e-08, + "loss": 0.2666, + "step": 16131 + }, + { + "epoch": 0.97, + "learning_rate": 3.390889433129041e-08, + "loss": 0.2963, + "step": 16132 + }, + { + "epoch": 0.97, + "learning_rate": 3.374808882267933e-08, + "loss": 0.2596, + "step": 16133 + }, + { + "epoch": 0.97, + "learning_rate": 3.358766486883602e-08, + "loss": 0.2473, + "step": 16134 + }, + { + "epoch": 0.97, + "learning_rate": 3.342762247590226e-08, + "loss": 0.2511, + "step": 16135 + }, + { + "epoch": 0.97, + "learning_rate": 3.326796165000534e-08, + "loss": 0.2525, + "step": 16136 + }, + { + "epoch": 0.97, + "learning_rate": 3.3108682397259286e-08, + "loss": 0.2584, + "step": 16137 + }, + { + "epoch": 0.97, + "learning_rate": 3.294978472376142e-08, + "loss": 0.273, + "step": 16138 + }, + { + "epoch": 0.97, + "learning_rate": 3.2791268635594675e-08, + "loss": 0.262, + "step": 16139 + }, + { + "epoch": 0.98, + "learning_rate": 3.263313413882752e-08, + "loss": 0.2475, + "step": 16140 + }, + { + "epoch": 0.98, + "learning_rate": 3.2475381239513995e-08, + "loss": 0.2786, + "step": 16141 + }, + { + "epoch": 0.98, + "learning_rate": 3.231800994369483e-08, + "loss": 0.2457, + "step": 16142 + }, + { + "epoch": 0.98, + "learning_rate": 3.2161020257394095e-08, + "loss": 0.2415, + "step": 16143 + }, + { + "epoch": 0.98, + "learning_rate": 3.200441218662253e-08, + "loss": 0.2833, + "step": 16144 + }, + { + "epoch": 0.98, + "learning_rate": 3.1848185737374247e-08, + "loss": 0.3047, + "step": 16145 + }, + { + "epoch": 0.98, + "learning_rate": 3.169234091563334e-08, + "loss": 0.2888, + "step": 16146 + }, + { + "epoch": 0.98, + "learning_rate": 3.15368777273628e-08, + "loss": 0.2386, + "step": 16147 + }, + { + "epoch": 0.98, + "learning_rate": 3.138179617851678e-08, + "loss": 0.2591, + "step": 16148 + }, + { + "epoch": 0.98, + "learning_rate": 3.122709627503162e-08, + "loss": 0.2595, + "step": 16149 + }, + { + "epoch": 0.98, + "learning_rate": 3.1072778022830375e-08, + "loss": 0.2684, + "step": 16150 + }, + { + "epoch": 0.98, + "learning_rate": 3.0918841427821644e-08, + "loss": 0.2584, + "step": 16151 + }, + { + "epoch": 0.98, + "learning_rate": 3.076528649589738e-08, + "loss": 0.2769, + "step": 16152 + }, + { + "epoch": 0.98, + "learning_rate": 3.061211323293733e-08, + "loss": 0.2401, + "step": 16153 + }, + { + "epoch": 0.98, + "learning_rate": 3.045932164480569e-08, + "loss": 0.2819, + "step": 16154 + }, + { + "epoch": 0.98, + "learning_rate": 3.030691173735223e-08, + "loss": 0.2748, + "step": 16155 + }, + { + "epoch": 0.98, + "learning_rate": 3.015488351641116e-08, + "loss": 0.2469, + "step": 16156 + }, + { + "epoch": 0.98, + "learning_rate": 3.0003236987802276e-08, + "loss": 0.2737, + "step": 16157 + }, + { + "epoch": 0.98, + "learning_rate": 2.985197215733315e-08, + "loss": 0.2846, + "step": 16158 + }, + { + "epoch": 0.98, + "learning_rate": 2.9701089030793606e-08, + "loss": 0.2295, + "step": 16159 + }, + { + "epoch": 0.98, + "learning_rate": 2.9550587613961236e-08, + "loss": 0.2806, + "step": 16160 + }, + { + "epoch": 0.98, + "learning_rate": 2.9400467912596985e-08, + "loss": 0.2536, + "step": 16161 + }, + { + "epoch": 0.98, + "learning_rate": 2.9250729932448485e-08, + "loss": 0.2627, + "step": 16162 + }, + { + "epoch": 0.98, + "learning_rate": 2.9101373679247814e-08, + "loss": 0.2589, + "step": 16163 + }, + { + "epoch": 0.98, + "learning_rate": 2.895239915871373e-08, + "loss": 0.255, + "step": 16164 + }, + { + "epoch": 0.98, + "learning_rate": 2.8803806376550557e-08, + "loss": 0.2641, + "step": 16165 + }, + { + "epoch": 0.98, + "learning_rate": 2.8655595338444864e-08, + "loss": 0.2561, + "step": 16166 + }, + { + "epoch": 0.98, + "learning_rate": 2.8507766050072106e-08, + "loss": 0.248, + "step": 16167 + }, + { + "epoch": 0.98, + "learning_rate": 2.83603185170922e-08, + "loss": 0.2546, + "step": 16168 + }, + { + "epoch": 0.98, + "learning_rate": 2.8213252745149522e-08, + "loss": 0.3094, + "step": 16169 + }, + { + "epoch": 0.98, + "learning_rate": 2.806656873987512e-08, + "loss": 0.2596, + "step": 16170 + }, + { + "epoch": 0.98, + "learning_rate": 2.79202665068834e-08, + "loss": 0.2702, + "step": 16171 + }, + { + "epoch": 0.98, + "learning_rate": 2.7774346051776536e-08, + "loss": 0.2698, + "step": 16172 + }, + { + "epoch": 0.98, + "learning_rate": 2.7628807380142286e-08, + "loss": 0.2522, + "step": 16173 + }, + { + "epoch": 0.98, + "learning_rate": 2.748365049754953e-08, + "loss": 0.2888, + "step": 16174 + }, + { + "epoch": 0.98, + "learning_rate": 2.7338875409558265e-08, + "loss": 0.2489, + "step": 16175 + }, + { + "epoch": 0.98, + "learning_rate": 2.7194482121709608e-08, + "loss": 0.2472, + "step": 16176 + }, + { + "epoch": 0.98, + "learning_rate": 2.7050470639531367e-08, + "loss": 0.2951, + "step": 16177 + }, + { + "epoch": 0.98, + "learning_rate": 2.6906840968539128e-08, + "loss": 0.2874, + "step": 16178 + }, + { + "epoch": 0.98, + "learning_rate": 2.6763593114229603e-08, + "loss": 0.2559, + "step": 16179 + }, + { + "epoch": 0.98, + "learning_rate": 2.6620727082087293e-08, + "loss": 0.2732, + "step": 16180 + }, + { + "epoch": 0.98, + "learning_rate": 2.6478242877583383e-08, + "loss": 0.2573, + "step": 16181 + }, + { + "epoch": 0.98, + "learning_rate": 2.633614050617017e-08, + "loss": 0.2432, + "step": 16182 + }, + { + "epoch": 0.98, + "learning_rate": 2.6194419973289975e-08, + "loss": 0.2708, + "step": 16183 + }, + { + "epoch": 0.98, + "learning_rate": 2.6053081284368453e-08, + "loss": 0.2619, + "step": 16184 + }, + { + "epoch": 0.98, + "learning_rate": 2.5912124444815723e-08, + "loss": 0.268, + "step": 16185 + }, + { + "epoch": 0.98, + "learning_rate": 2.577154946002858e-08, + "loss": 0.261, + "step": 16186 + }, + { + "epoch": 0.98, + "learning_rate": 2.563135633538938e-08, + "loss": 0.2912, + "step": 16187 + }, + { + "epoch": 0.98, + "learning_rate": 2.5491545076264945e-08, + "loss": 0.2725, + "step": 16188 + }, + { + "epoch": 0.98, + "learning_rate": 2.5352115688008773e-08, + "loss": 0.2743, + "step": 16189 + }, + { + "epoch": 0.98, + "learning_rate": 2.52130681759577e-08, + "loss": 0.2354, + "step": 16190 + }, + { + "epoch": 0.98, + "learning_rate": 2.507440254543525e-08, + "loss": 0.2769, + "step": 16191 + }, + { + "epoch": 0.98, + "learning_rate": 2.493611880175051e-08, + "loss": 0.2942, + "step": 16192 + }, + { + "epoch": 0.98, + "learning_rate": 2.479821695019813e-08, + "loss": 0.2366, + "step": 16193 + }, + { + "epoch": 0.98, + "learning_rate": 2.466069699605611e-08, + "loss": 0.2613, + "step": 16194 + }, + { + "epoch": 0.98, + "learning_rate": 2.4523558944591353e-08, + "loss": 0.2593, + "step": 16195 + }, + { + "epoch": 0.98, + "learning_rate": 2.438680280105299e-08, + "loss": 0.2793, + "step": 16196 + }, + { + "epoch": 0.98, + "learning_rate": 2.425042857067683e-08, + "loss": 0.2616, + "step": 16197 + }, + { + "epoch": 0.98, + "learning_rate": 2.411443625868426e-08, + "loss": 0.2517, + "step": 16198 + }, + { + "epoch": 0.98, + "learning_rate": 2.397882587028111e-08, + "loss": 0.2697, + "step": 16199 + }, + { + "epoch": 0.98, + "learning_rate": 2.384359741065878e-08, + "loss": 0.233, + "step": 16200 + }, + { + "epoch": 0.98, + "learning_rate": 2.3708750884996468e-08, + "loss": 0.2681, + "step": 16201 + }, + { + "epoch": 0.98, + "learning_rate": 2.3574286298454486e-08, + "loss": 0.2351, + "step": 16202 + }, + { + "epoch": 0.98, + "learning_rate": 2.3440203656182046e-08, + "loss": 0.2637, + "step": 16203 + }, + { + "epoch": 0.98, + "learning_rate": 2.3306502963311716e-08, + "loss": 0.2518, + "step": 16204 + }, + { + "epoch": 0.98, + "learning_rate": 2.317318422496273e-08, + "loss": 0.2621, + "step": 16205 + }, + { + "epoch": 0.98, + "learning_rate": 2.3040247446238783e-08, + "loss": 0.2794, + "step": 16206 + }, + { + "epoch": 0.98, + "learning_rate": 2.290769263223025e-08, + "loss": 0.2528, + "step": 16207 + }, + { + "epoch": 0.98, + "learning_rate": 2.2775519788010848e-08, + "loss": 0.2739, + "step": 16208 + }, + { + "epoch": 0.98, + "learning_rate": 2.2643728918640974e-08, + "loss": 0.2415, + "step": 16209 + }, + { + "epoch": 0.98, + "learning_rate": 2.251232002916659e-08, + "loss": 0.2605, + "step": 16210 + }, + { + "epoch": 0.98, + "learning_rate": 2.238129312461812e-08, + "loss": 0.2405, + "step": 16211 + }, + { + "epoch": 0.98, + "learning_rate": 2.225064821001155e-08, + "loss": 0.2674, + "step": 16212 + }, + { + "epoch": 0.98, + "learning_rate": 2.2120385290350655e-08, + "loss": 0.2748, + "step": 16213 + }, + { + "epoch": 0.98, + "learning_rate": 2.1990504370620337e-08, + "loss": 0.2695, + "step": 16214 + }, + { + "epoch": 0.98, + "learning_rate": 2.1861005455794393e-08, + "loss": 0.279, + "step": 16215 + }, + { + "epoch": 0.98, + "learning_rate": 2.173188855082886e-08, + "loss": 0.2556, + "step": 16216 + }, + { + "epoch": 0.98, + "learning_rate": 2.1603153660668673e-08, + "loss": 0.2423, + "step": 16217 + }, + { + "epoch": 0.98, + "learning_rate": 2.1474800790242112e-08, + "loss": 0.2746, + "step": 16218 + }, + { + "epoch": 0.98, + "learning_rate": 2.1346829944463022e-08, + "loss": 0.2737, + "step": 16219 + }, + { + "epoch": 0.98, + "learning_rate": 2.1219241128229708e-08, + "loss": 0.2513, + "step": 16220 + }, + { + "epoch": 0.98, + "learning_rate": 2.109203434642826e-08, + "loss": 0.2825, + "step": 16221 + }, + { + "epoch": 0.98, + "learning_rate": 2.0965209603928115e-08, + "loss": 0.2616, + "step": 16222 + }, + { + "epoch": 0.98, + "learning_rate": 2.0838766905585394e-08, + "loss": 0.2414, + "step": 16223 + }, + { + "epoch": 0.98, + "learning_rate": 2.0712706256239557e-08, + "loss": 0.2599, + "step": 16224 + }, + { + "epoch": 0.98, + "learning_rate": 2.058702766071785e-08, + "loss": 0.2542, + "step": 16225 + }, + { + "epoch": 0.98, + "learning_rate": 2.0461731123831984e-08, + "loss": 0.2504, + "step": 16226 + }, + { + "epoch": 0.98, + "learning_rate": 2.033681665037812e-08, + "loss": 0.2634, + "step": 16227 + }, + { + "epoch": 0.98, + "learning_rate": 2.0212284245137993e-08, + "loss": 0.2372, + "step": 16228 + }, + { + "epoch": 0.98, + "learning_rate": 2.0088133912881114e-08, + "loss": 0.2702, + "step": 16229 + }, + { + "epoch": 0.98, + "learning_rate": 1.9964365658360352e-08, + "loss": 0.2835, + "step": 16230 + }, + { + "epoch": 0.98, + "learning_rate": 1.9840979486311918e-08, + "loss": 0.2765, + "step": 16231 + }, + { + "epoch": 0.98, + "learning_rate": 1.9717975401462032e-08, + "loss": 0.2528, + "step": 16232 + }, + { + "epoch": 0.98, + "learning_rate": 1.9595353408519148e-08, + "loss": 0.2557, + "step": 16233 + }, + { + "epoch": 0.98, + "learning_rate": 1.947311351217729e-08, + "loss": 0.2734, + "step": 16234 + }, + { + "epoch": 0.98, + "learning_rate": 1.9351255717117156e-08, + "loss": 0.2928, + "step": 16235 + }, + { + "epoch": 0.98, + "learning_rate": 1.9229780028002797e-08, + "loss": 0.2552, + "step": 16236 + }, + { + "epoch": 0.98, + "learning_rate": 1.9108686449486046e-08, + "loss": 0.2455, + "step": 16237 + }, + { + "epoch": 0.98, + "learning_rate": 1.8987974986202084e-08, + "loss": 0.2854, + "step": 16238 + }, + { + "epoch": 0.98, + "learning_rate": 1.886764564277388e-08, + "loss": 0.2489, + "step": 16239 + }, + { + "epoch": 0.98, + "learning_rate": 1.8747698423805527e-08, + "loss": 0.2426, + "step": 16240 + }, + { + "epoch": 0.98, + "learning_rate": 1.862813333389113e-08, + "loss": 0.286, + "step": 16241 + }, + { + "epoch": 0.98, + "learning_rate": 1.8508950377609248e-08, + "loss": 0.2766, + "step": 16242 + }, + { + "epoch": 0.98, + "learning_rate": 1.8390149559519567e-08, + "loss": 0.2626, + "step": 16243 + }, + { + "epoch": 0.98, + "learning_rate": 1.827173088417289e-08, + "loss": 0.3018, + "step": 16244 + }, + { + "epoch": 0.98, + "learning_rate": 1.815369435610226e-08, + "loss": 0.2542, + "step": 16245 + }, + { + "epoch": 0.98, + "learning_rate": 1.8036039979826282e-08, + "loss": 0.2686, + "step": 16246 + }, + { + "epoch": 0.98, + "learning_rate": 1.791876775984913e-08, + "loss": 0.258, + "step": 16247 + }, + { + "epoch": 0.98, + "learning_rate": 1.7801877700660552e-08, + "loss": 0.2838, + "step": 16248 + }, + { + "epoch": 0.98, + "learning_rate": 1.7685369806736964e-08, + "loss": 0.2782, + "step": 16249 + }, + { + "epoch": 0.98, + "learning_rate": 1.756924408253702e-08, + "loss": 0.2748, + "step": 16250 + }, + { + "epoch": 0.98, + "learning_rate": 1.745350053250827e-08, + "loss": 0.2697, + "step": 16251 + }, + { + "epoch": 0.98, + "learning_rate": 1.7338139161079403e-08, + "loss": 0.2552, + "step": 16252 + }, + { + "epoch": 0.98, + "learning_rate": 1.722315997267021e-08, + "loss": 0.2715, + "step": 16253 + }, + { + "epoch": 0.98, + "learning_rate": 1.710856297168051e-08, + "loss": 0.2605, + "step": 16254 + }, + { + "epoch": 0.98, + "learning_rate": 1.69943481624979e-08, + "loss": 0.2394, + "step": 16255 + }, + { + "epoch": 0.98, + "learning_rate": 1.6880515549495547e-08, + "loss": 0.2821, + "step": 16256 + }, + { + "epoch": 0.98, + "learning_rate": 1.676706513703108e-08, + "loss": 0.2586, + "step": 16257 + }, + { + "epoch": 0.98, + "learning_rate": 1.665399692944769e-08, + "loss": 0.2649, + "step": 16258 + }, + { + "epoch": 0.98, + "learning_rate": 1.654131093107525e-08, + "loss": 0.287, + "step": 16259 + }, + { + "epoch": 0.98, + "learning_rate": 1.642900714622697e-08, + "loss": 0.2818, + "step": 16260 + }, + { + "epoch": 0.98, + "learning_rate": 1.631708557920275e-08, + "loss": 0.2653, + "step": 16261 + }, + { + "epoch": 0.98, + "learning_rate": 1.6205546234286938e-08, + "loss": 0.2397, + "step": 16262 + }, + { + "epoch": 0.98, + "learning_rate": 1.6094389115750563e-08, + "loss": 0.264, + "step": 16263 + }, + { + "epoch": 0.98, + "learning_rate": 1.598361422784911e-08, + "loss": 0.2588, + "step": 16264 + }, + { + "epoch": 0.98, + "learning_rate": 1.587322157482252e-08, + "loss": 0.274, + "step": 16265 + }, + { + "epoch": 0.98, + "learning_rate": 1.576321116089852e-08, + "loss": 0.2508, + "step": 16266 + }, + { + "epoch": 0.98, + "learning_rate": 1.5653582990288187e-08, + "loss": 0.2753, + "step": 16267 + }, + { + "epoch": 0.98, + "learning_rate": 1.5544337067188164e-08, + "loss": 0.2783, + "step": 16268 + }, + { + "epoch": 0.98, + "learning_rate": 1.5435473395781774e-08, + "loss": 0.2771, + "step": 16269 + }, + { + "epoch": 0.98, + "learning_rate": 1.5326991980235684e-08, + "loss": 0.2246, + "step": 16270 + }, + { + "epoch": 0.98, + "learning_rate": 1.5218892824704346e-08, + "loss": 0.2655, + "step": 16271 + }, + { + "epoch": 0.98, + "learning_rate": 1.5111175933325562e-08, + "loss": 0.2536, + "step": 16272 + }, + { + "epoch": 0.98, + "learning_rate": 1.500384131022381e-08, + "loss": 0.2348, + "step": 16273 + }, + { + "epoch": 0.98, + "learning_rate": 1.4896888959508027e-08, + "loss": 0.2612, + "step": 16274 + }, + { + "epoch": 0.98, + "learning_rate": 1.4790318885271604e-08, + "loss": 0.2538, + "step": 16275 + }, + { + "epoch": 0.98, + "learning_rate": 1.4684131091596832e-08, + "loss": 0.2761, + "step": 16276 + }, + { + "epoch": 0.98, + "learning_rate": 1.4578325582548236e-08, + "loss": 0.2524, + "step": 16277 + }, + { + "epoch": 0.98, + "learning_rate": 1.4472902362175911e-08, + "loss": 0.2414, + "step": 16278 + }, + { + "epoch": 0.98, + "learning_rate": 1.4367861434516628e-08, + "loss": 0.2605, + "step": 16279 + }, + { + "epoch": 0.98, + "learning_rate": 1.4263202803590503e-08, + "loss": 0.2754, + "step": 16280 + }, + { + "epoch": 0.98, + "learning_rate": 1.4158926473406553e-08, + "loss": 0.2733, + "step": 16281 + }, + { + "epoch": 0.98, + "learning_rate": 1.4055032447954919e-08, + "loss": 0.2725, + "step": 16282 + }, + { + "epoch": 0.98, + "learning_rate": 1.395152073121464e-08, + "loss": 0.2448, + "step": 16283 + }, + { + "epoch": 0.98, + "learning_rate": 1.3848391327148102e-08, + "loss": 0.2589, + "step": 16284 + }, + { + "epoch": 0.98, + "learning_rate": 1.374564423970326e-08, + "loss": 0.2695, + "step": 16285 + }, + { + "epoch": 0.98, + "learning_rate": 1.364327947281474e-08, + "loss": 0.2628, + "step": 16286 + }, + { + "epoch": 0.98, + "learning_rate": 1.3541297030399414e-08, + "loss": 0.2626, + "step": 16287 + }, + { + "epoch": 0.98, + "learning_rate": 1.3439696916364153e-08, + "loss": 0.2786, + "step": 16288 + }, + { + "epoch": 0.98, + "learning_rate": 1.333847913459696e-08, + "loss": 0.2793, + "step": 16289 + }, + { + "epoch": 0.98, + "learning_rate": 1.3237643688974733e-08, + "loss": 0.2852, + "step": 16290 + }, + { + "epoch": 0.98, + "learning_rate": 1.3137190583355497e-08, + "loss": 0.2759, + "step": 16291 + }, + { + "epoch": 0.98, + "learning_rate": 1.3037119821586174e-08, + "loss": 0.2592, + "step": 16292 + }, + { + "epoch": 0.98, + "learning_rate": 1.2937431407499256e-08, + "loss": 0.2458, + "step": 16293 + }, + { + "epoch": 0.98, + "learning_rate": 1.283812534490947e-08, + "loss": 0.231, + "step": 16294 + }, + { + "epoch": 0.98, + "learning_rate": 1.2739201637619324e-08, + "loss": 0.2557, + "step": 16295 + }, + { + "epoch": 0.98, + "learning_rate": 1.2640660289415795e-08, + "loss": 0.2457, + "step": 16296 + }, + { + "epoch": 0.98, + "learning_rate": 1.254250130407142e-08, + "loss": 0.2857, + "step": 16297 + }, + { + "epoch": 0.98, + "learning_rate": 1.244472468534652e-08, + "loss": 0.2591, + "step": 16298 + }, + { + "epoch": 0.98, + "learning_rate": 1.2347330436980331e-08, + "loss": 0.2436, + "step": 16299 + }, + { + "epoch": 0.98, + "learning_rate": 1.2250318562705422e-08, + "loss": 0.2759, + "step": 16300 + }, + { + "epoch": 0.98, + "learning_rate": 1.2153689066233266e-08, + "loss": 0.256, + "step": 16301 + }, + { + "epoch": 0.98, + "learning_rate": 1.2057441951264237e-08, + "loss": 0.2596, + "step": 16302 + }, + { + "epoch": 0.98, + "learning_rate": 1.1961577221483167e-08, + "loss": 0.2476, + "step": 16303 + }, + { + "epoch": 0.98, + "learning_rate": 1.1866094880560452e-08, + "loss": 0.24, + "step": 16304 + }, + { + "epoch": 0.99, + "learning_rate": 1.1770994932150947e-08, + "loss": 0.2935, + "step": 16305 + }, + { + "epoch": 0.99, + "learning_rate": 1.1676277379895073e-08, + "loss": 0.281, + "step": 16306 + }, + { + "epoch": 0.99, + "learning_rate": 1.158194222742104e-08, + "loss": 0.2462, + "step": 16307 + }, + { + "epoch": 0.99, + "learning_rate": 1.148798947833929e-08, + "loss": 0.2496, + "step": 16308 + }, + { + "epoch": 0.99, + "learning_rate": 1.1394419136245837e-08, + "loss": 0.2602, + "step": 16309 + }, + { + "epoch": 0.99, + "learning_rate": 1.1301231204724482e-08, + "loss": 0.2637, + "step": 16310 + }, + { + "epoch": 0.99, + "learning_rate": 1.1208425687342372e-08, + "loss": 0.2826, + "step": 16311 + }, + { + "epoch": 0.99, + "learning_rate": 1.111600258765222e-08, + "loss": 0.2759, + "step": 16312 + }, + { + "epoch": 0.99, + "learning_rate": 1.1023961909192304e-08, + "loss": 0.2528, + "step": 16313 + }, + { + "epoch": 0.99, + "learning_rate": 1.0932303655487586e-08, + "loss": 0.2885, + "step": 16314 + }, + { + "epoch": 0.99, + "learning_rate": 1.084102783004637e-08, + "loss": 0.256, + "step": 16315 + }, + { + "epoch": 0.99, + "learning_rate": 1.0750134436361414e-08, + "loss": 0.242, + "step": 16316 + }, + { + "epoch": 0.99, + "learning_rate": 1.0659623477915494e-08, + "loss": 0.2423, + "step": 16317 + }, + { + "epoch": 0.99, + "learning_rate": 1.05694949581725e-08, + "loss": 0.2588, + "step": 16318 + }, + { + "epoch": 0.99, + "learning_rate": 1.0479748880583007e-08, + "loss": 0.2447, + "step": 16319 + }, + { + "epoch": 0.99, + "learning_rate": 1.0390385248583157e-08, + "loss": 0.2683, + "step": 16320 + }, + { + "epoch": 0.99, + "learning_rate": 1.0301404065592435e-08, + "loss": 0.2532, + "step": 16321 + }, + { + "epoch": 0.99, + "learning_rate": 1.0212805335020337e-08, + "loss": 0.2679, + "step": 16322 + }, + { + "epoch": 0.99, + "learning_rate": 1.0124589060257483e-08, + "loss": 0.278, + "step": 16323 + }, + { + "epoch": 0.99, + "learning_rate": 1.0036755244680064e-08, + "loss": 0.2709, + "step": 16324 + }, + { + "epoch": 0.99, + "learning_rate": 9.949303891653162e-09, + "loss": 0.2614, + "step": 16325 + }, + { + "epoch": 0.99, + "learning_rate": 9.86223500452299e-09, + "loss": 0.2815, + "step": 16326 + }, + { + "epoch": 0.99, + "learning_rate": 9.77554858662244e-09, + "loss": 0.2667, + "step": 16327 + }, + { + "epoch": 0.99, + "learning_rate": 9.689244641272188e-09, + "loss": 0.2605, + "step": 16328 + }, + { + "epoch": 0.99, + "learning_rate": 9.603323171776257e-09, + "loss": 0.2447, + "step": 16329 + }, + { + "epoch": 0.99, + "learning_rate": 9.517784181422018e-09, + "loss": 0.2789, + "step": 16330 + }, + { + "epoch": 0.99, + "learning_rate": 9.43262767348685e-09, + "loss": 0.279, + "step": 16331 + }, + { + "epoch": 0.99, + "learning_rate": 9.347853651229256e-09, + "loss": 0.2427, + "step": 16332 + }, + { + "epoch": 0.99, + "learning_rate": 9.263462117895528e-09, + "loss": 0.2602, + "step": 16333 + }, + { + "epoch": 0.99, + "learning_rate": 9.179453076715305e-09, + "loss": 0.2669, + "step": 16334 + }, + { + "epoch": 0.99, + "learning_rate": 9.095826530907126e-09, + "loss": 0.2441, + "step": 16335 + }, + { + "epoch": 0.99, + "learning_rate": 9.012582483670651e-09, + "loss": 0.2377, + "step": 16336 + }, + { + "epoch": 0.99, + "learning_rate": 8.929720938193331e-09, + "loss": 0.2856, + "step": 16337 + }, + { + "epoch": 0.99, + "learning_rate": 8.847241897648184e-09, + "loss": 0.2557, + "step": 16338 + }, + { + "epoch": 0.99, + "learning_rate": 8.765145365191574e-09, + "loss": 0.2758, + "step": 16339 + }, + { + "epoch": 0.99, + "learning_rate": 8.683431343968762e-09, + "loss": 0.2592, + "step": 16340 + }, + { + "epoch": 0.99, + "learning_rate": 8.602099837105026e-09, + "loss": 0.2789, + "step": 16341 + }, + { + "epoch": 0.99, + "learning_rate": 8.521150847716763e-09, + "loss": 0.2518, + "step": 16342 + }, + { + "epoch": 0.99, + "learning_rate": 8.440584378902605e-09, + "loss": 0.2652, + "step": 16343 + }, + { + "epoch": 0.99, + "learning_rate": 8.36040043374675e-09, + "loss": 0.2355, + "step": 16344 + }, + { + "epoch": 0.99, + "learning_rate": 8.280599015317859e-09, + "loss": 0.2651, + "step": 16345 + }, + { + "epoch": 0.99, + "learning_rate": 8.20118012667237e-09, + "loss": 0.2492, + "step": 16346 + }, + { + "epoch": 0.99, + "learning_rate": 8.122143770850077e-09, + "loss": 0.2815, + "step": 16347 + }, + { + "epoch": 0.99, + "learning_rate": 8.043489950878558e-09, + "loss": 0.2677, + "step": 16348 + }, + { + "epoch": 0.99, + "learning_rate": 7.965218669766516e-09, + "loss": 0.2965, + "step": 16349 + }, + { + "epoch": 0.99, + "learning_rate": 7.887329930512665e-09, + "loss": 0.268, + "step": 16350 + }, + { + "epoch": 0.99, + "learning_rate": 7.809823736096844e-09, + "loss": 0.2769, + "step": 16351 + }, + { + "epoch": 0.99, + "learning_rate": 7.732700089488898e-09, + "loss": 0.2575, + "step": 16352 + }, + { + "epoch": 0.99, + "learning_rate": 7.655958993639801e-09, + "loss": 0.2753, + "step": 16353 + }, + { + "epoch": 0.99, + "learning_rate": 7.579600451488311e-09, + "loss": 0.2708, + "step": 16354 + }, + { + "epoch": 0.99, + "learning_rate": 7.503624465956538e-09, + "loss": 0.2835, + "step": 16355 + }, + { + "epoch": 0.99, + "learning_rate": 7.428031039955486e-09, + "loss": 0.263, + "step": 16356 + }, + { + "epoch": 0.99, + "learning_rate": 7.3528201763761745e-09, + "loss": 0.2688, + "step": 16357 + }, + { + "epoch": 0.99, + "learning_rate": 7.277991878100743e-09, + "loss": 0.2511, + "step": 16358 + }, + { + "epoch": 0.99, + "learning_rate": 7.203546147992457e-09, + "loss": 0.2542, + "step": 16359 + }, + { + "epoch": 0.99, + "learning_rate": 7.12948298890237e-09, + "loss": 0.2679, + "step": 16360 + }, + { + "epoch": 0.99, + "learning_rate": 7.0558024036648795e-09, + "loss": 0.2695, + "step": 16361 + }, + { + "epoch": 0.99, + "learning_rate": 6.982504395102175e-09, + "loss": 0.2951, + "step": 16362 + }, + { + "epoch": 0.99, + "learning_rate": 6.909588966019787e-09, + "loss": 0.2287, + "step": 16363 + }, + { + "epoch": 0.99, + "learning_rate": 6.837056119208818e-09, + "loss": 0.2666, + "step": 16364 + }, + { + "epoch": 0.99, + "learning_rate": 6.764905857445936e-09, + "loss": 0.2737, + "step": 16365 + }, + { + "epoch": 0.99, + "learning_rate": 6.693138183494485e-09, + "loss": 0.2539, + "step": 16366 + }, + { + "epoch": 0.99, + "learning_rate": 6.621753100102268e-09, + "loss": 0.2574, + "step": 16367 + }, + { + "epoch": 0.99, + "learning_rate": 6.5507506100004334e-09, + "loss": 0.2416, + "step": 16368 + }, + { + "epoch": 0.99, + "learning_rate": 6.480130715909028e-09, + "loss": 0.2574, + "step": 16369 + }, + { + "epoch": 0.99, + "learning_rate": 6.409893420530333e-09, + "loss": 0.2754, + "step": 16370 + }, + { + "epoch": 0.99, + "learning_rate": 6.340038726554421e-09, + "loss": 0.2564, + "step": 16371 + }, + { + "epoch": 0.99, + "learning_rate": 6.270566636655817e-09, + "loss": 0.2418, + "step": 16372 + }, + { + "epoch": 0.99, + "learning_rate": 6.201477153493507e-09, + "loss": 0.2552, + "step": 16373 + }, + { + "epoch": 0.99, + "learning_rate": 6.132770279713151e-09, + "loss": 0.2496, + "step": 16374 + }, + { + "epoch": 0.99, + "learning_rate": 6.064446017944869e-09, + "loss": 0.2617, + "step": 16375 + }, + { + "epoch": 0.99, + "learning_rate": 5.996504370803236e-09, + "loss": 0.2804, + "step": 16376 + }, + { + "epoch": 0.99, + "learning_rate": 5.928945340891723e-09, + "loss": 0.2814, + "step": 16377 + }, + { + "epoch": 0.99, + "learning_rate": 5.8617689307949314e-09, + "loss": 0.262, + "step": 16378 + }, + { + "epoch": 0.99, + "learning_rate": 5.794975143086357e-09, + "loss": 0.2563, + "step": 16379 + }, + { + "epoch": 0.99, + "learning_rate": 5.7285639803217326e-09, + "loss": 0.2444, + "step": 16380 + }, + { + "epoch": 0.99, + "learning_rate": 5.662535445043471e-09, + "loss": 0.2365, + "step": 16381 + }, + { + "epoch": 0.99, + "learning_rate": 5.596889539780659e-09, + "loss": 0.2561, + "step": 16382 + }, + { + "epoch": 0.99, + "learning_rate": 5.531626267045731e-09, + "loss": 0.258, + "step": 16383 + }, + { + "epoch": 0.99, + "learning_rate": 5.4667456293377995e-09, + "loss": 0.24, + "step": 16384 + }, + { + "epoch": 0.99, + "learning_rate": 5.4022476291393236e-09, + "loss": 0.2704, + "step": 16385 + }, + { + "epoch": 0.99, + "learning_rate": 5.3381322689216584e-09, + "loss": 0.284, + "step": 16386 + }, + { + "epoch": 0.99, + "learning_rate": 5.274399551138398e-09, + "loss": 0.2365, + "step": 16387 + }, + { + "epoch": 0.99, + "learning_rate": 5.211049478229813e-09, + "loss": 0.2554, + "step": 16388 + }, + { + "epoch": 0.99, + "learning_rate": 5.148082052620629e-09, + "loss": 0.2358, + "step": 16389 + }, + { + "epoch": 0.99, + "learning_rate": 5.085497276722251e-09, + "loss": 0.2529, + "step": 16390 + }, + { + "epoch": 0.99, + "learning_rate": 5.023295152930541e-09, + "loss": 0.2476, + "step": 16391 + }, + { + "epoch": 0.99, + "learning_rate": 4.961475683625816e-09, + "loss": 0.2863, + "step": 16392 + }, + { + "epoch": 0.99, + "learning_rate": 4.900038871176183e-09, + "loss": 0.2706, + "step": 16393 + }, + { + "epoch": 0.99, + "learning_rate": 4.838984717934203e-09, + "loss": 0.2556, + "step": 16394 + }, + { + "epoch": 0.99, + "learning_rate": 4.778313226235786e-09, + "loss": 0.2819, + "step": 16395 + }, + { + "epoch": 0.99, + "learning_rate": 4.71802439840463e-09, + "loss": 0.2554, + "step": 16396 + }, + { + "epoch": 0.99, + "learning_rate": 4.658118236747777e-09, + "loss": 0.2831, + "step": 16397 + }, + { + "epoch": 0.99, + "learning_rate": 4.598594743561169e-09, + "loss": 0.2739, + "step": 16398 + }, + { + "epoch": 0.99, + "learning_rate": 4.5394539211207625e-09, + "loss": 0.2766, + "step": 16399 + }, + { + "epoch": 0.99, + "learning_rate": 4.480695771692522e-09, + "loss": 0.2673, + "step": 16400 + }, + { + "epoch": 0.99, + "learning_rate": 4.4223202975246514e-09, + "loss": 0.2764, + "step": 16401 + }, + { + "epoch": 0.99, + "learning_rate": 4.364327500853138e-09, + "loss": 0.2566, + "step": 16402 + }, + { + "epoch": 0.99, + "learning_rate": 4.306717383898429e-09, + "loss": 0.26, + "step": 16403 + }, + { + "epoch": 0.99, + "learning_rate": 4.249489948865426e-09, + "loss": 0.2595, + "step": 16404 + }, + { + "epoch": 0.99, + "learning_rate": 4.19264519794571e-09, + "loss": 0.2464, + "step": 16405 + }, + { + "epoch": 0.99, + "learning_rate": 4.136183133314209e-09, + "loss": 0.2688, + "step": 16406 + }, + { + "epoch": 0.99, + "learning_rate": 4.080103757133635e-09, + "loss": 0.263, + "step": 16407 + }, + { + "epoch": 0.99, + "learning_rate": 4.024407071551162e-09, + "loss": 0.2571, + "step": 16408 + }, + { + "epoch": 0.99, + "learning_rate": 3.9690930786995266e-09, + "loss": 0.2764, + "step": 16409 + }, + { + "epoch": 0.99, + "learning_rate": 3.914161780694814e-09, + "loss": 0.2489, + "step": 16410 + }, + { + "epoch": 0.99, + "learning_rate": 3.8596131796397874e-09, + "loss": 0.2761, + "step": 16411 + }, + { + "epoch": 0.99, + "learning_rate": 3.805447277624996e-09, + "loss": 0.2677, + "step": 16412 + }, + { + "epoch": 0.99, + "learning_rate": 3.751664076723227e-09, + "loss": 0.2352, + "step": 16413 + }, + { + "epoch": 0.99, + "learning_rate": 3.6982635789928334e-09, + "loss": 0.2672, + "step": 16414 + }, + { + "epoch": 0.99, + "learning_rate": 3.645245786478846e-09, + "loss": 0.2616, + "step": 16415 + }, + { + "epoch": 0.99, + "learning_rate": 3.592610701211863e-09, + "loss": 0.2699, + "step": 16416 + }, + { + "epoch": 0.99, + "learning_rate": 3.540358325204718e-09, + "loss": 0.3027, + "step": 16417 + }, + { + "epoch": 0.99, + "learning_rate": 3.488488660460254e-09, + "loss": 0.2638, + "step": 16418 + }, + { + "epoch": 0.99, + "learning_rate": 3.4370017089635497e-09, + "loss": 0.281, + "step": 16419 + }, + { + "epoch": 0.99, + "learning_rate": 3.385897472685251e-09, + "loss": 0.2794, + "step": 16420 + }, + { + "epoch": 0.99, + "learning_rate": 3.335175953581571e-09, + "loss": 0.264, + "step": 16421 + }, + { + "epoch": 0.99, + "learning_rate": 3.284837153595399e-09, + "loss": 0.2582, + "step": 16422 + }, + { + "epoch": 0.99, + "learning_rate": 3.2348810746529736e-09, + "loss": 0.2496, + "step": 16423 + }, + { + "epoch": 0.99, + "learning_rate": 3.185307718668318e-09, + "loss": 0.2699, + "step": 16424 + }, + { + "epoch": 0.99, + "learning_rate": 3.136117087537693e-09, + "loss": 0.2454, + "step": 16425 + }, + { + "epoch": 0.99, + "learning_rate": 3.0873091831451485e-09, + "loss": 0.2549, + "step": 16426 + }, + { + "epoch": 0.99, + "learning_rate": 3.0388840073591887e-09, + "loss": 0.2972, + "step": 16427 + }, + { + "epoch": 0.99, + "learning_rate": 2.9908415620327756e-09, + "loss": 0.2508, + "step": 16428 + }, + { + "epoch": 0.99, + "learning_rate": 2.9431818490077703e-09, + "loss": 0.2653, + "step": 16429 + }, + { + "epoch": 0.99, + "learning_rate": 2.8959048701060477e-09, + "loss": 0.2716, + "step": 16430 + }, + { + "epoch": 0.99, + "learning_rate": 2.8490106271394924e-09, + "loss": 0.2674, + "step": 16431 + }, + { + "epoch": 0.99, + "learning_rate": 2.8024991219022247e-09, + "loss": 0.2585, + "step": 16432 + }, + { + "epoch": 0.99, + "learning_rate": 2.756370356175042e-09, + "loss": 0.2804, + "step": 16433 + }, + { + "epoch": 0.99, + "learning_rate": 2.71062433172542e-09, + "loss": 0.2623, + "step": 16434 + }, + { + "epoch": 0.99, + "learning_rate": 2.6652610503019593e-09, + "loss": 0.2722, + "step": 16435 + }, + { + "epoch": 0.99, + "learning_rate": 2.62028051364438e-09, + "loss": 0.2807, + "step": 16436 + }, + { + "epoch": 0.99, + "learning_rate": 2.575682723472417e-09, + "loss": 0.2991, + "step": 16437 + }, + { + "epoch": 0.99, + "learning_rate": 2.531467681495814e-09, + "loss": 0.2649, + "step": 16438 + }, + { + "epoch": 0.99, + "learning_rate": 2.4876353894043305e-09, + "loss": 0.2521, + "step": 16439 + }, + { + "epoch": 0.99, + "learning_rate": 2.444185848878844e-09, + "loss": 0.2778, + "step": 16440 + }, + { + "epoch": 0.99, + "learning_rate": 2.4011190615813584e-09, + "loss": 0.28, + "step": 16441 + }, + { + "epoch": 0.99, + "learning_rate": 2.358435029160555e-09, + "loss": 0.2725, + "step": 16442 + }, + { + "epoch": 0.99, + "learning_rate": 2.3161337532517925e-09, + "loss": 0.2693, + "step": 16443 + }, + { + "epoch": 0.99, + "learning_rate": 2.2742152354737756e-09, + "loss": 0.2651, + "step": 16444 + }, + { + "epoch": 0.99, + "learning_rate": 2.232679477430777e-09, + "loss": 0.2501, + "step": 16445 + }, + { + "epoch": 0.99, + "learning_rate": 2.191526480713746e-09, + "loss": 0.2641, + "step": 16446 + }, + { + "epoch": 0.99, + "learning_rate": 2.150756246898089e-09, + "loss": 0.2517, + "step": 16447 + }, + { + "epoch": 0.99, + "learning_rate": 2.1103687775447803e-09, + "loss": 0.2693, + "step": 16448 + }, + { + "epoch": 0.99, + "learning_rate": 2.0703640741992494e-09, + "loss": 0.2612, + "step": 16449 + }, + { + "epoch": 0.99, + "learning_rate": 2.0307421383947146e-09, + "loss": 0.2493, + "step": 16450 + }, + { + "epoch": 0.99, + "learning_rate": 1.99150297164552e-09, + "loss": 0.2778, + "step": 16451 + }, + { + "epoch": 0.99, + "learning_rate": 1.952646575457129e-09, + "loss": 0.2497, + "step": 16452 + }, + { + "epoch": 0.99, + "learning_rate": 1.9141729513139085e-09, + "loss": 0.2633, + "step": 16453 + }, + { + "epoch": 0.99, + "learning_rate": 1.8760821006913454e-09, + "loss": 0.2684, + "step": 16454 + }, + { + "epoch": 0.99, + "learning_rate": 1.8383740250471627e-09, + "loss": 0.2778, + "step": 16455 + }, + { + "epoch": 0.99, + "learning_rate": 1.8010487258235399e-09, + "loss": 0.257, + "step": 16456 + }, + { + "epoch": 0.99, + "learning_rate": 1.7641062044515544e-09, + "loss": 0.2532, + "step": 16457 + }, + { + "epoch": 0.99, + "learning_rate": 1.7275464623434102e-09, + "loss": 0.2588, + "step": 16458 + }, + { + "epoch": 0.99, + "learning_rate": 1.6913695009013187e-09, + "loss": 0.2497, + "step": 16459 + }, + { + "epoch": 0.99, + "learning_rate": 1.6555753215075077e-09, + "loss": 0.2461, + "step": 16460 + }, + { + "epoch": 0.99, + "learning_rate": 1.6201639255342128e-09, + "loss": 0.255, + "step": 16461 + }, + { + "epoch": 0.99, + "learning_rate": 1.5851353143370163e-09, + "loss": 0.2577, + "step": 16462 + }, + { + "epoch": 0.99, + "learning_rate": 1.5504894892559573e-09, + "loss": 0.2701, + "step": 16463 + }, + { + "epoch": 0.99, + "learning_rate": 1.516226451618863e-09, + "loss": 0.2596, + "step": 16464 + }, + { + "epoch": 0.99, + "learning_rate": 1.482346202735796e-09, + "loss": 0.2695, + "step": 16465 + }, + { + "epoch": 0.99, + "learning_rate": 1.4488487439057175e-09, + "loss": 0.2517, + "step": 16466 + }, + { + "epoch": 0.99, + "learning_rate": 1.4157340764098248e-09, + "loss": 0.2493, + "step": 16467 + }, + { + "epoch": 0.99, + "learning_rate": 1.3830022015159927e-09, + "loss": 0.2405, + "step": 16468 + }, + { + "epoch": 0.99, + "learning_rate": 1.350653120477663e-09, + "loss": 0.2806, + "step": 16469 + }, + { + "epoch": 0.99, + "learning_rate": 1.3186868345327341e-09, + "loss": 0.2458, + "step": 16470 + }, + { + "epoch": 1.0, + "learning_rate": 1.2871033449057824e-09, + "loss": 0.2797, + "step": 16471 + }, + { + "epoch": 1.0, + "learning_rate": 1.2559026528058405e-09, + "loss": 0.2522, + "step": 16472 + }, + { + "epoch": 1.0, + "learning_rate": 1.2250847594275083e-09, + "loss": 0.2514, + "step": 16473 + }, + { + "epoch": 1.0, + "learning_rate": 1.194649665949843e-09, + "loss": 0.2447, + "step": 16474 + }, + { + "epoch": 1.0, + "learning_rate": 1.1645973735396888e-09, + "loss": 0.2676, + "step": 16475 + }, + { + "epoch": 1.0, + "learning_rate": 1.1349278833450162e-09, + "loss": 0.2566, + "step": 16476 + }, + { + "epoch": 1.0, + "learning_rate": 1.1056411965038038e-09, + "loss": 0.2746, + "step": 16477 + }, + { + "epoch": 1.0, + "learning_rate": 1.076737314137377e-09, + "loss": 0.2766, + "step": 16478 + }, + { + "epoch": 1.0, + "learning_rate": 1.0482162373504079e-09, + "loss": 0.2806, + "step": 16479 + }, + { + "epoch": 1.0, + "learning_rate": 1.0200779672364658e-09, + "loss": 0.2492, + "step": 16480 + }, + { + "epoch": 1.0, + "learning_rate": 9.923225048724672e-10, + "loss": 0.272, + "step": 16481 + }, + { + "epoch": 1.0, + "learning_rate": 9.649498513208954e-10, + "loss": 0.2559, + "step": 16482 + }, + { + "epoch": 1.0, + "learning_rate": 9.379600076298012e-10, + "loss": 0.2648, + "step": 16483 + }, + { + "epoch": 1.0, + "learning_rate": 9.113529748316918e-10, + "loss": 0.2422, + "step": 16484 + }, + { + "epoch": 1.0, + "learning_rate": 8.851287539468622e-10, + "loss": 0.2681, + "step": 16485 + }, + { + "epoch": 1.0, + "learning_rate": 8.592873459778439e-10, + "loss": 0.278, + "step": 16486 + }, + { + "epoch": 1.0, + "learning_rate": 8.338287519149557e-10, + "loss": 0.2582, + "step": 16487 + }, + { + "epoch": 1.0, + "learning_rate": 8.087529727318633e-10, + "loss": 0.2705, + "step": 16488 + }, + { + "epoch": 1.0, + "learning_rate": 7.840600093889095e-10, + "loss": 0.278, + "step": 16489 + }, + { + "epoch": 1.0, + "learning_rate": 7.597498628320044e-10, + "loss": 0.2405, + "step": 16490 + }, + { + "epoch": 1.0, + "learning_rate": 7.358225339915148e-10, + "loss": 0.2384, + "step": 16491 + }, + { + "epoch": 1.0, + "learning_rate": 7.122780237844851e-10, + "loss": 0.2616, + "step": 16492 + }, + { + "epoch": 1.0, + "learning_rate": 6.891163331101958e-10, + "loss": 0.2299, + "step": 16493 + }, + { + "epoch": 1.0, + "learning_rate": 6.663374628568254e-10, + "loss": 0.2595, + "step": 16494 + }, + { + "epoch": 1.0, + "learning_rate": 6.439414138958988e-10, + "loss": 0.2871, + "step": 16495 + }, + { + "epoch": 1.0, + "learning_rate": 6.219281870845084e-10, + "loss": 0.2674, + "step": 16496 + }, + { + "epoch": 1.0, + "learning_rate": 6.002977832664237e-10, + "loss": 0.2741, + "step": 16497 + }, + { + "epoch": 1.0, + "learning_rate": 5.790502032698709e-10, + "loss": 0.26, + "step": 16498 + }, + { + "epoch": 1.0, + "learning_rate": 5.581854479075332e-10, + "loss": 0.2758, + "step": 16499 + }, + { + "epoch": 1.0, + "learning_rate": 5.377035179776613e-10, + "loss": 0.2776, + "step": 16500 + }, + { + "epoch": 1.0, + "learning_rate": 5.176044142662928e-10, + "loss": 0.2892, + "step": 16501 + }, + { + "epoch": 1.0, + "learning_rate": 4.978881375417022e-10, + "loss": 0.2383, + "step": 16502 + }, + { + "epoch": 1.0, + "learning_rate": 4.785546885577308e-10, + "loss": 0.2571, + "step": 16503 + }, + { + "epoch": 1.0, + "learning_rate": 4.596040680571179e-10, + "loss": 0.2831, + "step": 16504 + }, + { + "epoch": 1.0, + "learning_rate": 4.410362767626186e-10, + "loss": 0.2506, + "step": 16505 + }, + { + "epoch": 1.0, + "learning_rate": 4.228513153869962e-10, + "loss": 0.2405, + "step": 16506 + }, + { + "epoch": 1.0, + "learning_rate": 4.050491846252502e-10, + "loss": 0.2698, + "step": 16507 + }, + { + "epoch": 1.0, + "learning_rate": 3.8762988516016785e-10, + "loss": 0.2634, + "step": 16508 + }, + { + "epoch": 1.0, + "learning_rate": 3.705934176578829e-10, + "loss": 0.2557, + "step": 16509 + }, + { + "epoch": 1.0, + "learning_rate": 3.5393978277120655e-10, + "loss": 0.2614, + "step": 16510 + }, + { + "epoch": 1.0, + "learning_rate": 3.3766898113629655e-10, + "loss": 0.246, + "step": 16511 + }, + { + "epoch": 1.0, + "learning_rate": 3.2178101337709823e-10, + "loss": 0.2595, + "step": 16512 + }, + { + "epoch": 1.0, + "learning_rate": 3.0627588010312404e-10, + "loss": 0.2709, + "step": 16513 + }, + { + "epoch": 1.0, + "learning_rate": 2.9115358190501266e-10, + "loss": 0.2455, + "step": 16514 + }, + { + "epoch": 1.0, + "learning_rate": 2.764141193645209e-10, + "loss": 0.2409, + "step": 16515 + }, + { + "epoch": 1.0, + "learning_rate": 2.620574930445319e-10, + "loss": 0.239, + "step": 16516 + }, + { + "epoch": 1.0, + "learning_rate": 2.48083703494606e-10, + "loss": 0.2684, + "step": 16517 + }, + { + "epoch": 1.0, + "learning_rate": 2.344927512498707e-10, + "loss": 0.237, + "step": 16518 + }, + { + "epoch": 1.0, + "learning_rate": 2.2128463683102064e-10, + "loss": 0.2623, + "step": 16519 + }, + { + "epoch": 1.0, + "learning_rate": 2.084593607443175e-10, + "loss": 0.262, + "step": 16520 + }, + { + "epoch": 1.0, + "learning_rate": 1.9601692347936962e-10, + "loss": 0.2758, + "step": 16521 + }, + { + "epoch": 1.0, + "learning_rate": 1.839573255124627e-10, + "loss": 0.2568, + "step": 16522 + }, + { + "epoch": 1.0, + "learning_rate": 1.7228056730766995e-10, + "loss": 0.2642, + "step": 16523 + }, + { + "epoch": 1.0, + "learning_rate": 1.609866493090806e-10, + "loss": 0.2753, + "step": 16524 + }, + { + "epoch": 1.0, + "learning_rate": 1.5007557195079182e-10, + "loss": 0.2627, + "step": 16525 + }, + { + "epoch": 1.0, + "learning_rate": 1.3954733565024747e-10, + "loss": 0.2734, + "step": 16526 + }, + { + "epoch": 1.0, + "learning_rate": 1.2940194080934833e-10, + "loss": 0.2656, + "step": 16527 + }, + { + "epoch": 1.0, + "learning_rate": 1.1963938781889283e-10, + "loss": 0.266, + "step": 16528 + }, + { + "epoch": 1.0, + "learning_rate": 1.1025967705080576e-10, + "loss": 0.251, + "step": 16529 + }, + { + "epoch": 1.0, + "learning_rate": 1.0126280886479934e-10, + "loss": 0.3153, + "step": 16530 + }, + { + "epoch": 1.0, + "learning_rate": 9.264878360393248e-11, + "loss": 0.2627, + "step": 16531 + }, + { + "epoch": 1.0, + "learning_rate": 8.441760160016188e-11, + "loss": 0.2778, + "step": 16532 + }, + { + "epoch": 1.0, + "learning_rate": 7.656926316879088e-11, + "loss": 0.2667, + "step": 16533 + }, + { + "epoch": 1.0, + "learning_rate": 6.910376860735923e-11, + "loss": 0.2792, + "step": 16534 + }, + { + "epoch": 1.0, + "learning_rate": 6.202111820452494e-11, + "loss": 0.2772, + "step": 16535 + }, + { + "epoch": 1.0, + "learning_rate": 5.53213122300722e-11, + "loss": 0.2614, + "step": 16536 + }, + { + "epoch": 1.0, + "learning_rate": 4.900435094157274e-11, + "loss": 0.2571, + "step": 16537 + }, + { + "epoch": 1.0, + "learning_rate": 4.307023457883475e-11, + "loss": 0.2832, + "step": 16538 + }, + { + "epoch": 1.0, + "learning_rate": 3.7518963371674376e-11, + "loss": 0.2619, + "step": 16539 + }, + { + "epoch": 1.0, + "learning_rate": 3.235053752992379e-11, + "loss": 0.2708, + "step": 16540 + }, + { + "epoch": 1.0, + "learning_rate": 2.756495725342312e-11, + "loss": 0.2716, + "step": 16541 + }, + { + "epoch": 1.0, + "learning_rate": 2.3162222725359174e-11, + "loss": 0.2619, + "step": 16542 + }, + { + "epoch": 1.0, + "learning_rate": 1.9142334113375628e-11, + "loss": 0.3105, + "step": 16543 + }, + { + "epoch": 1.0, + "learning_rate": 1.550529157179348e-11, + "loss": 0.2695, + "step": 16544 + }, + { + "epoch": 1.0, + "learning_rate": 1.2251095239390609e-11, + "loss": 0.2504, + "step": 16545 + }, + { + "epoch": 1.0, + "learning_rate": 9.379745241622219e-12, + "loss": 0.2786, + "step": 16546 + }, + { + "epoch": 1.0, + "learning_rate": 6.891241688400385e-12, + "loss": 0.2592, + "step": 16547 + }, + { + "epoch": 1.0, + "learning_rate": 4.785584672983845e-12, + "loss": 0.2542, + "step": 16548 + }, + { + "epoch": 1.0, + "learning_rate": 3.0627742786393244e-12, + "loss": 0.2653, + "step": 16549 + }, + { + "epoch": 1.0, + "learning_rate": 1.7228105708699816e-12, + "loss": 0.261, + "step": 16550 + }, + { + "epoch": 1.0, + "learning_rate": 7.656935985256297e-13, + "loss": 0.263, + "step": 16551 + }, + { + "epoch": 1.0, + "learning_rate": 1.9142340157429772e-13, + "loss": 0.2624, + "step": 16552 + }, + { + "epoch": 1.0, + "learning_rate": 0.0, + "loss": 0.2653, + "step": 16553 + }, + { + "epoch": 1.0, + "step": 16553, + "total_flos": 750493356687360.0, + "train_loss": 0.3367887742271425, + "train_runtime": 150872.0687, + "train_samples_per_second": 9.216, + "train_steps_per_second": 0.11 + } + ], + "max_steps": 16553, + "num_train_epochs": 1, + "total_flos": 750493356687360.0, + "trial_name": null, + "trial_params": null +}