{ "best_metric": null, "best_model_checkpoint": null, "epoch": 1.0, "eval_steps": 500, "global_step": 11808, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 5.633802816901409e-08, "loss": 1.6758, "step": 1 }, { "epoch": 0.0, "learning_rate": 1.1267605633802818e-07, "loss": 1.7047, "step": 2 }, { "epoch": 0.0, "learning_rate": 1.6901408450704225e-07, "loss": 1.7525, "step": 3 }, { "epoch": 0.0, "learning_rate": 2.2535211267605636e-07, "loss": 1.6836, "step": 4 }, { "epoch": 0.0, "learning_rate": 2.8169014084507043e-07, "loss": 1.7202, "step": 5 }, { "epoch": 0.0, "learning_rate": 3.380281690140845e-07, "loss": 1.6453, "step": 6 }, { "epoch": 0.0, "learning_rate": 3.9436619718309864e-07, "loss": 1.6874, "step": 7 }, { "epoch": 0.0, "learning_rate": 4.507042253521127e-07, "loss": 1.6851, "step": 8 }, { "epoch": 0.0, "learning_rate": 5.070422535211268e-07, "loss": 1.6793, "step": 9 }, { "epoch": 0.0, "learning_rate": 5.633802816901409e-07, "loss": 1.6861, "step": 10 }, { "epoch": 0.0, "learning_rate": 6.197183098591549e-07, "loss": 1.6331, "step": 11 }, { "epoch": 0.0, "learning_rate": 6.76056338028169e-07, "loss": 1.6047, "step": 12 }, { "epoch": 0.0, "learning_rate": 7.323943661971832e-07, "loss": 1.659, "step": 13 }, { "epoch": 0.0, "learning_rate": 7.887323943661973e-07, "loss": 1.5509, "step": 14 }, { "epoch": 0.0, "learning_rate": 8.450704225352114e-07, "loss": 1.5175, "step": 15 }, { "epoch": 0.0, "learning_rate": 9.014084507042254e-07, "loss": 1.4892, "step": 16 }, { "epoch": 0.0, "learning_rate": 9.577464788732395e-07, "loss": 1.5298, "step": 17 }, { "epoch": 0.0, "learning_rate": 1.0140845070422536e-06, "loss": 1.5147, "step": 18 }, { "epoch": 0.0, "learning_rate": 1.0704225352112677e-06, "loss": 1.4906, "step": 19 }, { "epoch": 0.0, "learning_rate": 1.1267605633802817e-06, "loss": 1.3851, "step": 20 }, { "epoch": 0.0, "learning_rate": 1.1830985915492958e-06, "loss": 1.3569, "step": 21 }, { "epoch": 0.0, "learning_rate": 1.2394366197183099e-06, "loss": 1.4156, "step": 22 }, { "epoch": 0.0, "learning_rate": 1.295774647887324e-06, "loss": 1.4451, "step": 23 }, { "epoch": 0.0, "learning_rate": 1.352112676056338e-06, "loss": 1.377, "step": 24 }, { "epoch": 0.0, "learning_rate": 1.4084507042253523e-06, "loss": 1.434, "step": 25 }, { "epoch": 0.0, "learning_rate": 1.4647887323943664e-06, "loss": 1.4416, "step": 26 }, { "epoch": 0.0, "learning_rate": 1.5211267605633803e-06, "loss": 1.4363, "step": 27 }, { "epoch": 0.0, "learning_rate": 1.5774647887323946e-06, "loss": 1.367, "step": 28 }, { "epoch": 0.0, "learning_rate": 1.6338028169014086e-06, "loss": 1.3796, "step": 29 }, { "epoch": 0.0, "learning_rate": 1.6901408450704227e-06, "loss": 1.3944, "step": 30 }, { "epoch": 0.0, "learning_rate": 1.7464788732394366e-06, "loss": 1.3955, "step": 31 }, { "epoch": 0.0, "learning_rate": 1.8028169014084509e-06, "loss": 1.3792, "step": 32 }, { "epoch": 0.0, "learning_rate": 1.859154929577465e-06, "loss": 1.3484, "step": 33 }, { "epoch": 0.0, "learning_rate": 1.915492957746479e-06, "loss": 1.3452, "step": 34 }, { "epoch": 0.0, "learning_rate": 1.971830985915493e-06, "loss": 1.4057, "step": 35 }, { "epoch": 0.0, "learning_rate": 2.028169014084507e-06, "loss": 1.3922, "step": 36 }, { "epoch": 0.0, "learning_rate": 2.0845070422535212e-06, "loss": 1.3642, "step": 37 }, { "epoch": 0.0, "learning_rate": 2.1408450704225353e-06, "loss": 1.3761, "step": 38 }, { "epoch": 0.0, "learning_rate": 2.19718309859155e-06, "loss": 1.315, "step": 39 }, { "epoch": 0.0, "learning_rate": 2.2535211267605635e-06, "loss": 1.3598, "step": 40 }, { "epoch": 0.0, "learning_rate": 2.3098591549295775e-06, "loss": 1.3207, "step": 41 }, { "epoch": 0.0, "learning_rate": 2.3661971830985916e-06, "loss": 1.3736, "step": 42 }, { "epoch": 0.0, "learning_rate": 2.422535211267606e-06, "loss": 1.3514, "step": 43 }, { "epoch": 0.0, "learning_rate": 2.4788732394366198e-06, "loss": 1.2994, "step": 44 }, { "epoch": 0.0, "learning_rate": 2.535211267605634e-06, "loss": 1.3084, "step": 45 }, { "epoch": 0.0, "learning_rate": 2.591549295774648e-06, "loss": 1.3692, "step": 46 }, { "epoch": 0.0, "learning_rate": 2.647887323943662e-06, "loss": 1.3237, "step": 47 }, { "epoch": 0.0, "learning_rate": 2.704225352112676e-06, "loss": 1.3031, "step": 48 }, { "epoch": 0.0, "learning_rate": 2.7605633802816906e-06, "loss": 1.2593, "step": 49 }, { "epoch": 0.0, "learning_rate": 2.8169014084507046e-06, "loss": 1.2897, "step": 50 }, { "epoch": 0.0, "learning_rate": 2.8732394366197187e-06, "loss": 1.2774, "step": 51 }, { "epoch": 0.0, "learning_rate": 2.929577464788733e-06, "loss": 1.3254, "step": 52 }, { "epoch": 0.0, "learning_rate": 2.9859154929577465e-06, "loss": 1.2992, "step": 53 }, { "epoch": 0.0, "learning_rate": 3.0422535211267605e-06, "loss": 1.3385, "step": 54 }, { "epoch": 0.0, "learning_rate": 3.0985915492957746e-06, "loss": 1.2345, "step": 55 }, { "epoch": 0.0, "learning_rate": 3.154929577464789e-06, "loss": 1.3125, "step": 56 }, { "epoch": 0.0, "learning_rate": 3.211267605633803e-06, "loss": 1.246, "step": 57 }, { "epoch": 0.0, "learning_rate": 3.2676056338028173e-06, "loss": 1.2705, "step": 58 }, { "epoch": 0.0, "learning_rate": 3.3239436619718313e-06, "loss": 1.2875, "step": 59 }, { "epoch": 0.01, "learning_rate": 3.3802816901408454e-06, "loss": 1.2807, "step": 60 }, { "epoch": 0.01, "learning_rate": 3.4366197183098595e-06, "loss": 1.3052, "step": 61 }, { "epoch": 0.01, "learning_rate": 3.492957746478873e-06, "loss": 1.2683, "step": 62 }, { "epoch": 0.01, "learning_rate": 3.549295774647888e-06, "loss": 1.3042, "step": 63 }, { "epoch": 0.01, "learning_rate": 3.6056338028169017e-06, "loss": 1.2398, "step": 64 }, { "epoch": 0.01, "learning_rate": 3.6619718309859158e-06, "loss": 1.2773, "step": 65 }, { "epoch": 0.01, "learning_rate": 3.71830985915493e-06, "loss": 1.3149, "step": 66 }, { "epoch": 0.01, "learning_rate": 3.774647887323944e-06, "loss": 1.2483, "step": 67 }, { "epoch": 0.01, "learning_rate": 3.830985915492958e-06, "loss": 1.2334, "step": 68 }, { "epoch": 0.01, "learning_rate": 3.887323943661972e-06, "loss": 1.2872, "step": 69 }, { "epoch": 0.01, "learning_rate": 3.943661971830986e-06, "loss": 1.2578, "step": 70 }, { "epoch": 0.01, "learning_rate": 4.000000000000001e-06, "loss": 1.3085, "step": 71 }, { "epoch": 0.01, "learning_rate": 4.056338028169014e-06, "loss": 1.1872, "step": 72 }, { "epoch": 0.01, "learning_rate": 4.112676056338029e-06, "loss": 1.2736, "step": 73 }, { "epoch": 0.01, "learning_rate": 4.1690140845070425e-06, "loss": 0.9045, "step": 74 }, { "epoch": 0.01, "learning_rate": 4.225352112676057e-06, "loss": 1.2264, "step": 75 }, { "epoch": 0.01, "learning_rate": 4.281690140845071e-06, "loss": 1.2353, "step": 76 }, { "epoch": 0.01, "learning_rate": 4.338028169014084e-06, "loss": 1.2376, "step": 77 }, { "epoch": 0.01, "learning_rate": 4.3943661971831e-06, "loss": 1.2607, "step": 78 }, { "epoch": 0.01, "learning_rate": 4.450704225352113e-06, "loss": 1.2628, "step": 79 }, { "epoch": 0.01, "learning_rate": 4.507042253521127e-06, "loss": 1.2625, "step": 80 }, { "epoch": 0.01, "learning_rate": 4.5633802816901414e-06, "loss": 1.2722, "step": 81 }, { "epoch": 0.01, "learning_rate": 4.619718309859155e-06, "loss": 1.2916, "step": 82 }, { "epoch": 0.01, "learning_rate": 4.67605633802817e-06, "loss": 1.2258, "step": 83 }, { "epoch": 0.01, "learning_rate": 4.732394366197183e-06, "loss": 1.2546, "step": 84 }, { "epoch": 0.01, "learning_rate": 4.788732394366197e-06, "loss": 1.292, "step": 85 }, { "epoch": 0.01, "learning_rate": 4.845070422535212e-06, "loss": 1.2438, "step": 86 }, { "epoch": 0.01, "learning_rate": 4.901408450704226e-06, "loss": 1.2394, "step": 87 }, { "epoch": 0.01, "learning_rate": 4.9577464788732395e-06, "loss": 1.2719, "step": 88 }, { "epoch": 0.01, "learning_rate": 5.014084507042254e-06, "loss": 1.2647, "step": 89 }, { "epoch": 0.01, "learning_rate": 5.070422535211268e-06, "loss": 1.2139, "step": 90 }, { "epoch": 0.01, "learning_rate": 5.126760563380282e-06, "loss": 1.2801, "step": 91 }, { "epoch": 0.01, "learning_rate": 5.183098591549296e-06, "loss": 1.2128, "step": 92 }, { "epoch": 0.01, "learning_rate": 5.23943661971831e-06, "loss": 1.2538, "step": 93 }, { "epoch": 0.01, "learning_rate": 5.295774647887324e-06, "loss": 1.2694, "step": 94 }, { "epoch": 0.01, "learning_rate": 5.352112676056338e-06, "loss": 1.2304, "step": 95 }, { "epoch": 0.01, "learning_rate": 5.408450704225352e-06, "loss": 1.1854, "step": 96 }, { "epoch": 0.01, "learning_rate": 5.4647887323943675e-06, "loss": 1.2255, "step": 97 }, { "epoch": 0.01, "learning_rate": 5.521126760563381e-06, "loss": 1.2205, "step": 98 }, { "epoch": 0.01, "learning_rate": 5.577464788732395e-06, "loss": 1.2677, "step": 99 }, { "epoch": 0.01, "learning_rate": 5.633802816901409e-06, "loss": 1.2443, "step": 100 }, { "epoch": 0.01, "learning_rate": 5.690140845070423e-06, "loss": 1.2587, "step": 101 }, { "epoch": 0.01, "learning_rate": 5.7464788732394374e-06, "loss": 1.2652, "step": 102 }, { "epoch": 0.01, "learning_rate": 5.802816901408451e-06, "loss": 1.243, "step": 103 }, { "epoch": 0.01, "learning_rate": 5.859154929577466e-06, "loss": 1.2186, "step": 104 }, { "epoch": 0.01, "learning_rate": 5.915492957746479e-06, "loss": 1.2016, "step": 105 }, { "epoch": 0.01, "learning_rate": 5.971830985915493e-06, "loss": 1.2045, "step": 106 }, { "epoch": 0.01, "learning_rate": 6.028169014084507e-06, "loss": 1.2107, "step": 107 }, { "epoch": 0.01, "learning_rate": 6.084507042253521e-06, "loss": 1.2462, "step": 108 }, { "epoch": 0.01, "learning_rate": 6.1408450704225356e-06, "loss": 1.2762, "step": 109 }, { "epoch": 0.01, "learning_rate": 6.197183098591549e-06, "loss": 1.2045, "step": 110 }, { "epoch": 0.01, "learning_rate": 6.2535211267605646e-06, "loss": 1.1879, "step": 111 }, { "epoch": 0.01, "learning_rate": 6.309859154929578e-06, "loss": 1.2412, "step": 112 }, { "epoch": 0.01, "learning_rate": 6.366197183098593e-06, "loss": 1.2205, "step": 113 }, { "epoch": 0.01, "learning_rate": 6.422535211267606e-06, "loss": 1.2727, "step": 114 }, { "epoch": 0.01, "learning_rate": 6.478873239436621e-06, "loss": 1.2482, "step": 115 }, { "epoch": 0.01, "learning_rate": 6.5352112676056345e-06, "loss": 1.1716, "step": 116 }, { "epoch": 0.01, "learning_rate": 6.591549295774648e-06, "loss": 1.2014, "step": 117 }, { "epoch": 0.01, "learning_rate": 6.647887323943663e-06, "loss": 1.2357, "step": 118 }, { "epoch": 0.01, "learning_rate": 6.704225352112676e-06, "loss": 1.1992, "step": 119 }, { "epoch": 0.01, "learning_rate": 6.760563380281691e-06, "loss": 1.2865, "step": 120 }, { "epoch": 0.01, "learning_rate": 6.8169014084507045e-06, "loss": 1.1684, "step": 121 }, { "epoch": 0.01, "learning_rate": 6.873239436619719e-06, "loss": 0.8646, "step": 122 }, { "epoch": 0.01, "learning_rate": 6.929577464788733e-06, "loss": 1.2405, "step": 123 }, { "epoch": 0.01, "learning_rate": 6.985915492957746e-06, "loss": 1.2801, "step": 124 }, { "epoch": 0.01, "learning_rate": 7.042253521126761e-06, "loss": 1.2482, "step": 125 }, { "epoch": 0.01, "learning_rate": 7.098591549295776e-06, "loss": 1.2299, "step": 126 }, { "epoch": 0.01, "learning_rate": 7.15492957746479e-06, "loss": 1.2333, "step": 127 }, { "epoch": 0.01, "learning_rate": 7.211267605633803e-06, "loss": 1.2459, "step": 128 }, { "epoch": 0.01, "learning_rate": 7.267605633802818e-06, "loss": 1.1196, "step": 129 }, { "epoch": 0.01, "learning_rate": 7.3239436619718316e-06, "loss": 1.2324, "step": 130 }, { "epoch": 0.01, "learning_rate": 7.380281690140846e-06, "loss": 1.191, "step": 131 }, { "epoch": 0.01, "learning_rate": 7.43661971830986e-06, "loss": 1.2558, "step": 132 }, { "epoch": 0.01, "learning_rate": 7.492957746478874e-06, "loss": 1.1664, "step": 133 }, { "epoch": 0.01, "learning_rate": 7.549295774647888e-06, "loss": 0.9942, "step": 134 }, { "epoch": 0.01, "learning_rate": 7.6056338028169015e-06, "loss": 1.1795, "step": 135 }, { "epoch": 0.01, "learning_rate": 7.661971830985916e-06, "loss": 1.2049, "step": 136 }, { "epoch": 0.01, "learning_rate": 7.71830985915493e-06, "loss": 1.1773, "step": 137 }, { "epoch": 0.01, "learning_rate": 7.774647887323943e-06, "loss": 1.2605, "step": 138 }, { "epoch": 0.01, "learning_rate": 7.830985915492958e-06, "loss": 1.2534, "step": 139 }, { "epoch": 0.01, "learning_rate": 7.887323943661972e-06, "loss": 1.2084, "step": 140 }, { "epoch": 0.01, "learning_rate": 7.943661971830987e-06, "loss": 1.1994, "step": 141 }, { "epoch": 0.01, "learning_rate": 8.000000000000001e-06, "loss": 1.254, "step": 142 }, { "epoch": 0.01, "learning_rate": 8.056338028169016e-06, "loss": 1.206, "step": 143 }, { "epoch": 0.01, "learning_rate": 8.112676056338029e-06, "loss": 1.2211, "step": 144 }, { "epoch": 0.01, "learning_rate": 8.169014084507043e-06, "loss": 1.1465, "step": 145 }, { "epoch": 0.01, "learning_rate": 8.225352112676058e-06, "loss": 1.2084, "step": 146 }, { "epoch": 0.01, "learning_rate": 8.28169014084507e-06, "loss": 1.2381, "step": 147 }, { "epoch": 0.01, "learning_rate": 8.338028169014085e-06, "loss": 1.2778, "step": 148 }, { "epoch": 0.01, "learning_rate": 8.3943661971831e-06, "loss": 1.143, "step": 149 }, { "epoch": 0.01, "learning_rate": 8.450704225352114e-06, "loss": 1.2075, "step": 150 }, { "epoch": 0.01, "learning_rate": 8.507042253521127e-06, "loss": 1.2061, "step": 151 }, { "epoch": 0.01, "learning_rate": 8.563380281690141e-06, "loss": 1.1679, "step": 152 }, { "epoch": 0.01, "learning_rate": 8.619718309859156e-06, "loss": 1.2541, "step": 153 }, { "epoch": 0.01, "learning_rate": 8.676056338028169e-06, "loss": 1.158, "step": 154 }, { "epoch": 0.01, "learning_rate": 8.732394366197183e-06, "loss": 1.2124, "step": 155 }, { "epoch": 0.01, "learning_rate": 8.7887323943662e-06, "loss": 1.1789, "step": 156 }, { "epoch": 0.01, "learning_rate": 8.845070422535212e-06, "loss": 1.2348, "step": 157 }, { "epoch": 0.01, "learning_rate": 8.901408450704227e-06, "loss": 1.1666, "step": 158 }, { "epoch": 0.01, "learning_rate": 8.957746478873241e-06, "loss": 1.2023, "step": 159 }, { "epoch": 0.01, "learning_rate": 9.014084507042254e-06, "loss": 1.207, "step": 160 }, { "epoch": 0.01, "learning_rate": 9.070422535211268e-06, "loss": 1.1985, "step": 161 }, { "epoch": 0.01, "learning_rate": 9.126760563380283e-06, "loss": 1.1738, "step": 162 }, { "epoch": 0.01, "learning_rate": 9.183098591549296e-06, "loss": 1.1161, "step": 163 }, { "epoch": 0.01, "learning_rate": 9.23943661971831e-06, "loss": 1.2031, "step": 164 }, { "epoch": 0.01, "learning_rate": 9.295774647887325e-06, "loss": 1.1912, "step": 165 }, { "epoch": 0.01, "learning_rate": 9.35211267605634e-06, "loss": 1.1356, "step": 166 }, { "epoch": 0.01, "learning_rate": 9.408450704225352e-06, "loss": 1.1926, "step": 167 }, { "epoch": 0.01, "learning_rate": 9.464788732394366e-06, "loss": 1.1677, "step": 168 }, { "epoch": 0.01, "learning_rate": 9.521126760563381e-06, "loss": 1.1826, "step": 169 }, { "epoch": 0.01, "learning_rate": 9.577464788732394e-06, "loss": 1.2039, "step": 170 }, { "epoch": 0.01, "learning_rate": 9.63380281690141e-06, "loss": 1.1575, "step": 171 }, { "epoch": 0.01, "learning_rate": 9.690140845070424e-06, "loss": 1.1824, "step": 172 }, { "epoch": 0.01, "learning_rate": 9.746478873239437e-06, "loss": 1.2443, "step": 173 }, { "epoch": 0.01, "learning_rate": 9.802816901408452e-06, "loss": 1.1692, "step": 174 }, { "epoch": 0.01, "learning_rate": 9.859154929577466e-06, "loss": 1.2121, "step": 175 }, { "epoch": 0.01, "learning_rate": 9.915492957746479e-06, "loss": 1.2333, "step": 176 }, { "epoch": 0.01, "learning_rate": 9.971830985915494e-06, "loss": 1.1548, "step": 177 }, { "epoch": 0.02, "learning_rate": 1.0028169014084508e-05, "loss": 1.2482, "step": 178 }, { "epoch": 0.02, "learning_rate": 1.0084507042253523e-05, "loss": 1.2494, "step": 179 }, { "epoch": 0.02, "learning_rate": 1.0140845070422535e-05, "loss": 0.8335, "step": 180 }, { "epoch": 0.02, "learning_rate": 1.019718309859155e-05, "loss": 1.1706, "step": 181 }, { "epoch": 0.02, "learning_rate": 1.0253521126760564e-05, "loss": 1.1777, "step": 182 }, { "epoch": 0.02, "learning_rate": 1.0309859154929577e-05, "loss": 1.209, "step": 183 }, { "epoch": 0.02, "learning_rate": 1.0366197183098592e-05, "loss": 1.2138, "step": 184 }, { "epoch": 0.02, "learning_rate": 1.0422535211267606e-05, "loss": 1.2227, "step": 185 }, { "epoch": 0.02, "learning_rate": 1.047887323943662e-05, "loss": 1.2412, "step": 186 }, { "epoch": 0.02, "learning_rate": 1.0535211267605633e-05, "loss": 1.1456, "step": 187 }, { "epoch": 0.02, "learning_rate": 1.0591549295774648e-05, "loss": 1.1611, "step": 188 }, { "epoch": 0.02, "learning_rate": 1.0647887323943662e-05, "loss": 1.1808, "step": 189 }, { "epoch": 0.02, "learning_rate": 1.0704225352112675e-05, "loss": 1.2346, "step": 190 }, { "epoch": 0.02, "learning_rate": 1.076056338028169e-05, "loss": 1.1991, "step": 191 }, { "epoch": 0.02, "learning_rate": 1.0816901408450704e-05, "loss": 1.146, "step": 192 }, { "epoch": 0.02, "learning_rate": 1.087323943661972e-05, "loss": 0.8399, "step": 193 }, { "epoch": 0.02, "learning_rate": 1.0929577464788735e-05, "loss": 0.8206, "step": 194 }, { "epoch": 0.02, "learning_rate": 1.0985915492957748e-05, "loss": 1.2049, "step": 195 }, { "epoch": 0.02, "learning_rate": 1.1042253521126762e-05, "loss": 1.1981, "step": 196 }, { "epoch": 0.02, "learning_rate": 1.1098591549295777e-05, "loss": 1.1552, "step": 197 }, { "epoch": 0.02, "learning_rate": 1.115492957746479e-05, "loss": 1.2048, "step": 198 }, { "epoch": 0.02, "learning_rate": 1.1211267605633804e-05, "loss": 1.1684, "step": 199 }, { "epoch": 0.02, "learning_rate": 1.1267605633802819e-05, "loss": 1.1987, "step": 200 }, { "epoch": 0.02, "learning_rate": 1.1323943661971833e-05, "loss": 1.2269, "step": 201 }, { "epoch": 0.02, "learning_rate": 1.1380281690140846e-05, "loss": 1.1212, "step": 202 }, { "epoch": 0.02, "learning_rate": 1.143661971830986e-05, "loss": 1.1972, "step": 203 }, { "epoch": 0.02, "learning_rate": 1.1492957746478875e-05, "loss": 1.1836, "step": 204 }, { "epoch": 0.02, "learning_rate": 1.1549295774647888e-05, "loss": 1.1905, "step": 205 }, { "epoch": 0.02, "learning_rate": 1.1605633802816902e-05, "loss": 1.224, "step": 206 }, { "epoch": 0.02, "learning_rate": 1.1661971830985917e-05, "loss": 1.2214, "step": 207 }, { "epoch": 0.02, "learning_rate": 1.1718309859154931e-05, "loss": 1.1953, "step": 208 }, { "epoch": 0.02, "learning_rate": 1.1774647887323944e-05, "loss": 1.1369, "step": 209 }, { "epoch": 0.02, "learning_rate": 1.1830985915492958e-05, "loss": 1.212, "step": 210 }, { "epoch": 0.02, "learning_rate": 1.1887323943661973e-05, "loss": 1.1668, "step": 211 }, { "epoch": 0.02, "learning_rate": 1.1943661971830986e-05, "loss": 1.161, "step": 212 }, { "epoch": 0.02, "learning_rate": 1.2e-05, "loss": 1.2013, "step": 213 }, { "epoch": 0.02, "learning_rate": 1.2056338028169015e-05, "loss": 1.1892, "step": 214 }, { "epoch": 0.02, "learning_rate": 1.211267605633803e-05, "loss": 0.8918, "step": 215 }, { "epoch": 0.02, "learning_rate": 1.2169014084507042e-05, "loss": 1.1851, "step": 216 }, { "epoch": 0.02, "learning_rate": 1.2225352112676057e-05, "loss": 1.1464, "step": 217 }, { "epoch": 0.02, "learning_rate": 1.2281690140845071e-05, "loss": 1.1905, "step": 218 }, { "epoch": 0.02, "learning_rate": 1.2338028169014084e-05, "loss": 1.2323, "step": 219 }, { "epoch": 0.02, "learning_rate": 1.2394366197183098e-05, "loss": 1.1531, "step": 220 }, { "epoch": 0.02, "learning_rate": 1.2450704225352113e-05, "loss": 1.2402, "step": 221 }, { "epoch": 0.02, "learning_rate": 1.2507042253521129e-05, "loss": 1.1448, "step": 222 }, { "epoch": 0.02, "learning_rate": 1.2563380281690144e-05, "loss": 1.1351, "step": 223 }, { "epoch": 0.02, "learning_rate": 1.2619718309859156e-05, "loss": 1.2281, "step": 224 }, { "epoch": 0.02, "learning_rate": 1.2676056338028171e-05, "loss": 0.9066, "step": 225 }, { "epoch": 0.02, "learning_rate": 1.2732394366197185e-05, "loss": 1.2036, "step": 226 }, { "epoch": 0.02, "learning_rate": 1.2788732394366198e-05, "loss": 0.8498, "step": 227 }, { "epoch": 0.02, "learning_rate": 1.2845070422535213e-05, "loss": 1.1669, "step": 228 }, { "epoch": 0.02, "learning_rate": 1.2901408450704227e-05, "loss": 1.2285, "step": 229 }, { "epoch": 0.02, "learning_rate": 1.2957746478873242e-05, "loss": 1.1688, "step": 230 }, { "epoch": 0.02, "learning_rate": 1.3014084507042255e-05, "loss": 1.1072, "step": 231 }, { "epoch": 0.02, "learning_rate": 1.3070422535211269e-05, "loss": 1.1816, "step": 232 }, { "epoch": 0.02, "learning_rate": 1.3126760563380284e-05, "loss": 1.1939, "step": 233 }, { "epoch": 0.02, "learning_rate": 1.3183098591549296e-05, "loss": 1.2171, "step": 234 }, { "epoch": 0.02, "learning_rate": 1.323943661971831e-05, "loss": 1.1338, "step": 235 }, { "epoch": 0.02, "learning_rate": 1.3295774647887325e-05, "loss": 1.1874, "step": 236 }, { "epoch": 0.02, "learning_rate": 1.335211267605634e-05, "loss": 1.1803, "step": 237 }, { "epoch": 0.02, "learning_rate": 1.3408450704225353e-05, "loss": 1.129, "step": 238 }, { "epoch": 0.02, "learning_rate": 1.3464788732394367e-05, "loss": 1.1781, "step": 239 }, { "epoch": 0.02, "learning_rate": 1.3521126760563382e-05, "loss": 1.2029, "step": 240 }, { "epoch": 0.02, "learning_rate": 1.3577464788732394e-05, "loss": 1.2092, "step": 241 }, { "epoch": 0.02, "learning_rate": 1.3633802816901409e-05, "loss": 1.2089, "step": 242 }, { "epoch": 0.02, "learning_rate": 1.3690140845070423e-05, "loss": 1.1787, "step": 243 }, { "epoch": 0.02, "learning_rate": 1.3746478873239438e-05, "loss": 1.1588, "step": 244 }, { "epoch": 0.02, "learning_rate": 1.380281690140845e-05, "loss": 1.1401, "step": 245 }, { "epoch": 0.02, "learning_rate": 1.3859154929577465e-05, "loss": 1.1614, "step": 246 }, { "epoch": 0.02, "learning_rate": 1.391549295774648e-05, "loss": 1.201, "step": 247 }, { "epoch": 0.02, "learning_rate": 1.3971830985915493e-05, "loss": 1.2737, "step": 248 }, { "epoch": 0.02, "learning_rate": 1.4028169014084507e-05, "loss": 1.2007, "step": 249 }, { "epoch": 0.02, "learning_rate": 1.4084507042253522e-05, "loss": 1.1857, "step": 250 }, { "epoch": 0.02, "learning_rate": 1.4140845070422536e-05, "loss": 0.7196, "step": 251 }, { "epoch": 0.02, "learning_rate": 1.4197183098591552e-05, "loss": 0.8662, "step": 252 }, { "epoch": 0.02, "learning_rate": 1.4253521126760565e-05, "loss": 1.2656, "step": 253 }, { "epoch": 0.02, "learning_rate": 1.430985915492958e-05, "loss": 1.1246, "step": 254 }, { "epoch": 0.02, "learning_rate": 1.4366197183098594e-05, "loss": 1.1544, "step": 255 }, { "epoch": 0.02, "learning_rate": 1.4422535211267607e-05, "loss": 1.1574, "step": 256 }, { "epoch": 0.02, "learning_rate": 1.4478873239436621e-05, "loss": 1.167, "step": 257 }, { "epoch": 0.02, "learning_rate": 1.4535211267605636e-05, "loss": 1.1756, "step": 258 }, { "epoch": 0.02, "learning_rate": 1.459154929577465e-05, "loss": 1.1921, "step": 259 }, { "epoch": 0.02, "learning_rate": 1.4647887323943663e-05, "loss": 1.1922, "step": 260 }, { "epoch": 0.02, "learning_rate": 1.4704225352112678e-05, "loss": 1.2543, "step": 261 }, { "epoch": 0.02, "learning_rate": 1.4760563380281692e-05, "loss": 1.1794, "step": 262 }, { "epoch": 0.02, "learning_rate": 1.4816901408450705e-05, "loss": 1.2149, "step": 263 }, { "epoch": 0.02, "learning_rate": 1.487323943661972e-05, "loss": 1.208, "step": 264 }, { "epoch": 0.02, "learning_rate": 1.4929577464788734e-05, "loss": 1.1603, "step": 265 }, { "epoch": 0.02, "learning_rate": 1.4985915492957748e-05, "loss": 1.1981, "step": 266 }, { "epoch": 0.02, "learning_rate": 1.5042253521126761e-05, "loss": 1.2291, "step": 267 }, { "epoch": 0.02, "learning_rate": 1.5098591549295776e-05, "loss": 1.1986, "step": 268 }, { "epoch": 0.02, "learning_rate": 1.515492957746479e-05, "loss": 1.1793, "step": 269 }, { "epoch": 0.02, "learning_rate": 1.5211267605633803e-05, "loss": 1.1578, "step": 270 }, { "epoch": 0.02, "learning_rate": 1.5267605633802818e-05, "loss": 1.2303, "step": 271 }, { "epoch": 0.02, "learning_rate": 1.5323943661971832e-05, "loss": 1.1921, "step": 272 }, { "epoch": 0.02, "learning_rate": 1.5380281690140847e-05, "loss": 1.2105, "step": 273 }, { "epoch": 0.02, "learning_rate": 1.543661971830986e-05, "loss": 1.217, "step": 274 }, { "epoch": 0.02, "learning_rate": 1.5492957746478872e-05, "loss": 1.1587, "step": 275 }, { "epoch": 0.02, "learning_rate": 1.5549295774647887e-05, "loss": 1.1628, "step": 276 }, { "epoch": 0.02, "learning_rate": 1.56056338028169e-05, "loss": 1.1858, "step": 277 }, { "epoch": 0.02, "learning_rate": 1.5661971830985916e-05, "loss": 1.1539, "step": 278 }, { "epoch": 0.02, "learning_rate": 1.571830985915493e-05, "loss": 1.155, "step": 279 }, { "epoch": 0.02, "learning_rate": 1.5774647887323945e-05, "loss": 1.2411, "step": 280 }, { "epoch": 0.02, "learning_rate": 1.583098591549296e-05, "loss": 1.1943, "step": 281 }, { "epoch": 0.02, "learning_rate": 1.5887323943661974e-05, "loss": 1.2238, "step": 282 }, { "epoch": 0.02, "learning_rate": 1.5943661971830988e-05, "loss": 1.1893, "step": 283 }, { "epoch": 0.02, "learning_rate": 1.6000000000000003e-05, "loss": 0.9031, "step": 284 }, { "epoch": 0.02, "learning_rate": 1.6056338028169017e-05, "loss": 1.2434, "step": 285 }, { "epoch": 0.02, "learning_rate": 1.611267605633803e-05, "loss": 1.1775, "step": 286 }, { "epoch": 0.02, "learning_rate": 1.6169014084507043e-05, "loss": 1.1516, "step": 287 }, { "epoch": 0.02, "learning_rate": 1.6225352112676057e-05, "loss": 1.1763, "step": 288 }, { "epoch": 0.02, "learning_rate": 1.6281690140845072e-05, "loss": 1.2272, "step": 289 }, { "epoch": 0.02, "learning_rate": 1.6338028169014086e-05, "loss": 1.1907, "step": 290 }, { "epoch": 0.02, "learning_rate": 1.63943661971831e-05, "loss": 1.1247, "step": 291 }, { "epoch": 0.02, "learning_rate": 1.6450704225352115e-05, "loss": 1.1241, "step": 292 }, { "epoch": 0.02, "learning_rate": 1.650704225352113e-05, "loss": 1.1795, "step": 293 }, { "epoch": 0.02, "learning_rate": 1.656338028169014e-05, "loss": 1.198, "step": 294 }, { "epoch": 0.02, "learning_rate": 1.6619718309859155e-05, "loss": 1.1811, "step": 295 }, { "epoch": 0.03, "learning_rate": 1.667605633802817e-05, "loss": 1.286, "step": 296 }, { "epoch": 0.03, "learning_rate": 1.6732394366197184e-05, "loss": 1.1791, "step": 297 }, { "epoch": 0.03, "learning_rate": 1.67887323943662e-05, "loss": 1.1969, "step": 298 }, { "epoch": 0.03, "learning_rate": 1.6845070422535213e-05, "loss": 1.2324, "step": 299 }, { "epoch": 0.03, "learning_rate": 1.6901408450704228e-05, "loss": 1.2155, "step": 300 }, { "epoch": 0.03, "learning_rate": 1.695774647887324e-05, "loss": 1.1733, "step": 301 }, { "epoch": 0.03, "learning_rate": 1.7014084507042253e-05, "loss": 1.1409, "step": 302 }, { "epoch": 0.03, "learning_rate": 1.7070422535211268e-05, "loss": 1.2298, "step": 303 }, { "epoch": 0.03, "learning_rate": 1.7126760563380282e-05, "loss": 1.1836, "step": 304 }, { "epoch": 0.03, "learning_rate": 1.7183098591549297e-05, "loss": 1.1849, "step": 305 }, { "epoch": 0.03, "learning_rate": 1.723943661971831e-05, "loss": 1.2186, "step": 306 }, { "epoch": 0.03, "learning_rate": 1.7295774647887326e-05, "loss": 1.2755, "step": 307 }, { "epoch": 0.03, "learning_rate": 1.7352112676056337e-05, "loss": 1.1461, "step": 308 }, { "epoch": 0.03, "learning_rate": 1.740845070422535e-05, "loss": 1.1748, "step": 309 }, { "epoch": 0.03, "learning_rate": 1.7464788732394366e-05, "loss": 1.1958, "step": 310 }, { "epoch": 0.03, "learning_rate": 1.7521126760563384e-05, "loss": 1.1568, "step": 311 }, { "epoch": 0.03, "learning_rate": 1.75774647887324e-05, "loss": 1.2517, "step": 312 }, { "epoch": 0.03, "learning_rate": 1.763380281690141e-05, "loss": 1.1709, "step": 313 }, { "epoch": 0.03, "learning_rate": 1.7690140845070424e-05, "loss": 1.1593, "step": 314 }, { "epoch": 0.03, "learning_rate": 1.774647887323944e-05, "loss": 1.2079, "step": 315 }, { "epoch": 0.03, "learning_rate": 1.7802816901408453e-05, "loss": 1.1576, "step": 316 }, { "epoch": 0.03, "learning_rate": 1.7859154929577468e-05, "loss": 1.1688, "step": 317 }, { "epoch": 0.03, "learning_rate": 1.7915492957746482e-05, "loss": 1.2342, "step": 318 }, { "epoch": 0.03, "learning_rate": 1.7971830985915497e-05, "loss": 1.2488, "step": 319 }, { "epoch": 0.03, "learning_rate": 1.8028169014084508e-05, "loss": 0.8182, "step": 320 }, { "epoch": 0.03, "learning_rate": 1.8084507042253522e-05, "loss": 1.1534, "step": 321 }, { "epoch": 0.03, "learning_rate": 1.8140845070422537e-05, "loss": 1.2095, "step": 322 }, { "epoch": 0.03, "learning_rate": 1.819718309859155e-05, "loss": 1.1522, "step": 323 }, { "epoch": 0.03, "learning_rate": 1.8253521126760566e-05, "loss": 1.1973, "step": 324 }, { "epoch": 0.03, "learning_rate": 1.830985915492958e-05, "loss": 1.2086, "step": 325 }, { "epoch": 0.03, "learning_rate": 1.836619718309859e-05, "loss": 1.2207, "step": 326 }, { "epoch": 0.03, "learning_rate": 1.8422535211267606e-05, "loss": 1.186, "step": 327 }, { "epoch": 0.03, "learning_rate": 1.847887323943662e-05, "loss": 1.2577, "step": 328 }, { "epoch": 0.03, "learning_rate": 1.8535211267605635e-05, "loss": 1.2593, "step": 329 }, { "epoch": 0.03, "learning_rate": 1.859154929577465e-05, "loss": 1.2198, "step": 330 }, { "epoch": 0.03, "learning_rate": 1.8647887323943664e-05, "loss": 1.2265, "step": 331 }, { "epoch": 0.03, "learning_rate": 1.870422535211268e-05, "loss": 1.1833, "step": 332 }, { "epoch": 0.03, "learning_rate": 1.876056338028169e-05, "loss": 1.2177, "step": 333 }, { "epoch": 0.03, "learning_rate": 1.8816901408450704e-05, "loss": 1.1931, "step": 334 }, { "epoch": 0.03, "learning_rate": 1.887323943661972e-05, "loss": 1.2071, "step": 335 }, { "epoch": 0.03, "learning_rate": 1.8929577464788733e-05, "loss": 1.2233, "step": 336 }, { "epoch": 0.03, "learning_rate": 1.8985915492957747e-05, "loss": 1.2438, "step": 337 }, { "epoch": 0.03, "learning_rate": 1.9042253521126762e-05, "loss": 1.2115, "step": 338 }, { "epoch": 0.03, "learning_rate": 1.9098591549295776e-05, "loss": 1.1705, "step": 339 }, { "epoch": 0.03, "learning_rate": 1.9154929577464788e-05, "loss": 1.1472, "step": 340 }, { "epoch": 0.03, "learning_rate": 1.9211267605633805e-05, "loss": 1.2045, "step": 341 }, { "epoch": 0.03, "learning_rate": 1.926760563380282e-05, "loss": 1.1999, "step": 342 }, { "epoch": 0.03, "learning_rate": 1.9323943661971834e-05, "loss": 1.1749, "step": 343 }, { "epoch": 0.03, "learning_rate": 1.938028169014085e-05, "loss": 1.1887, "step": 344 }, { "epoch": 0.03, "learning_rate": 1.943661971830986e-05, "loss": 1.1946, "step": 345 }, { "epoch": 0.03, "learning_rate": 1.9492957746478875e-05, "loss": 1.1714, "step": 346 }, { "epoch": 0.03, "learning_rate": 1.954929577464789e-05, "loss": 1.2002, "step": 347 }, { "epoch": 0.03, "learning_rate": 1.9605633802816904e-05, "loss": 1.2311, "step": 348 }, { "epoch": 0.03, "learning_rate": 1.9661971830985918e-05, "loss": 1.1983, "step": 349 }, { "epoch": 0.03, "learning_rate": 1.9718309859154933e-05, "loss": 1.1966, "step": 350 }, { "epoch": 0.03, "learning_rate": 1.9774647887323947e-05, "loss": 1.2021, "step": 351 }, { "epoch": 0.03, "learning_rate": 1.9830985915492958e-05, "loss": 1.2186, "step": 352 }, { "epoch": 0.03, "learning_rate": 1.9887323943661973e-05, "loss": 1.2158, "step": 353 }, { "epoch": 0.03, "learning_rate": 1.9943661971830987e-05, "loss": 1.1643, "step": 354 }, { "epoch": 0.03, "learning_rate": 2e-05, "loss": 1.2145, "step": 355 }, { "epoch": 0.03, "learning_rate": 1.999999962378923e-05, "loss": 1.188, "step": 356 }, { "epoch": 0.03, "learning_rate": 1.9999998495156944e-05, "loss": 1.2149, "step": 357 }, { "epoch": 0.03, "learning_rate": 1.999999661410323e-05, "loss": 1.1576, "step": 358 }, { "epoch": 0.03, "learning_rate": 1.9999993980628226e-05, "loss": 1.2011, "step": 359 }, { "epoch": 0.03, "learning_rate": 1.9999990594732137e-05, "loss": 1.2205, "step": 360 }, { "epoch": 0.03, "learning_rate": 1.999998645641521e-05, "loss": 1.2138, "step": 361 }, { "epoch": 0.03, "learning_rate": 1.9999981565677763e-05, "loss": 1.2395, "step": 362 }, { "epoch": 0.03, "learning_rate": 1.9999975922520157e-05, "loss": 1.2184, "step": 363 }, { "epoch": 0.03, "learning_rate": 1.9999969526942822e-05, "loss": 1.212, "step": 364 }, { "epoch": 0.03, "learning_rate": 1.9999962378946235e-05, "loss": 1.1969, "step": 365 }, { "epoch": 0.03, "learning_rate": 1.9999954478530937e-05, "loss": 1.1866, "step": 366 }, { "epoch": 0.03, "learning_rate": 1.9999945825697523e-05, "loss": 1.2314, "step": 367 }, { "epoch": 0.03, "learning_rate": 1.9999936420446643e-05, "loss": 1.2251, "step": 368 }, { "epoch": 0.03, "learning_rate": 1.9999926262779005e-05, "loss": 1.1621, "step": 369 }, { "epoch": 0.03, "learning_rate": 1.9999915352695374e-05, "loss": 1.1696, "step": 370 }, { "epoch": 0.03, "learning_rate": 1.9999903690196567e-05, "loss": 1.2116, "step": 371 }, { "epoch": 0.03, "learning_rate": 1.9999891275283465e-05, "loss": 1.2292, "step": 372 }, { "epoch": 0.03, "learning_rate": 1.9999878107957e-05, "loss": 1.2403, "step": 373 }, { "epoch": 0.03, "learning_rate": 1.9999864188218164e-05, "loss": 1.2424, "step": 374 }, { "epoch": 0.03, "learning_rate": 1.9999849516068005e-05, "loss": 1.2007, "step": 375 }, { "epoch": 0.03, "learning_rate": 1.999983409150763e-05, "loss": 1.1757, "step": 376 }, { "epoch": 0.03, "learning_rate": 1.999981791453819e-05, "loss": 1.206, "step": 377 }, { "epoch": 0.03, "learning_rate": 1.9999800985160912e-05, "loss": 1.2113, "step": 378 }, { "epoch": 0.03, "learning_rate": 1.9999783303377066e-05, "loss": 1.1921, "step": 379 }, { "epoch": 0.03, "learning_rate": 1.999976486918798e-05, "loss": 1.1915, "step": 380 }, { "epoch": 0.03, "learning_rate": 1.9999745682595044e-05, "loss": 1.2098, "step": 381 }, { "epoch": 0.03, "learning_rate": 1.99997257435997e-05, "loss": 0.9356, "step": 382 }, { "epoch": 0.03, "learning_rate": 1.999970505220345e-05, "loss": 1.1758, "step": 383 }, { "epoch": 0.03, "learning_rate": 1.9999683608407853e-05, "loss": 1.1967, "step": 384 }, { "epoch": 0.03, "learning_rate": 1.9999661412214517e-05, "loss": 1.1403, "step": 385 }, { "epoch": 0.03, "learning_rate": 1.9999638463625115e-05, "loss": 1.239, "step": 386 }, { "epoch": 0.03, "learning_rate": 1.9999614762641377e-05, "loss": 1.2483, "step": 387 }, { "epoch": 0.03, "learning_rate": 1.999959030926508e-05, "loss": 1.2183, "step": 388 }, { "epoch": 0.03, "learning_rate": 1.9999565103498063e-05, "loss": 1.1948, "step": 389 }, { "epoch": 0.03, "learning_rate": 1.9999539145342228e-05, "loss": 1.2122, "step": 390 }, { "epoch": 0.03, "learning_rate": 1.999951243479953e-05, "loss": 1.2471, "step": 391 }, { "epoch": 0.03, "learning_rate": 1.9999484971871973e-05, "loss": 1.2348, "step": 392 }, { "epoch": 0.03, "learning_rate": 1.9999456756561623e-05, "loss": 1.2749, "step": 393 }, { "epoch": 0.03, "learning_rate": 1.9999427788870607e-05, "loss": 0.8369, "step": 394 }, { "epoch": 0.03, "learning_rate": 1.9999398068801103e-05, "loss": 1.1765, "step": 395 }, { "epoch": 0.03, "learning_rate": 1.999936759635535e-05, "loss": 0.8408, "step": 396 }, { "epoch": 0.03, "learning_rate": 1.9999336371535632e-05, "loss": 1.2479, "step": 397 }, { "epoch": 0.03, "learning_rate": 1.999930439434431e-05, "loss": 1.1938, "step": 398 }, { "epoch": 0.03, "learning_rate": 1.9999271664783784e-05, "loss": 1.1513, "step": 399 }, { "epoch": 0.03, "learning_rate": 1.9999238182856514e-05, "loss": 1.1852, "step": 400 }, { "epoch": 0.03, "learning_rate": 1.9999203948565026e-05, "loss": 1.1643, "step": 401 }, { "epoch": 0.03, "learning_rate": 1.9999168961911888e-05, "loss": 1.1738, "step": 402 }, { "epoch": 0.03, "learning_rate": 1.999913322289974e-05, "loss": 1.1925, "step": 403 }, { "epoch": 0.03, "learning_rate": 1.9999096731531266e-05, "loss": 1.2183, "step": 404 }, { "epoch": 0.03, "learning_rate": 1.9999059487809216e-05, "loss": 1.1347, "step": 405 }, { "epoch": 0.03, "learning_rate": 1.9999021491736385e-05, "loss": 1.1639, "step": 406 }, { "epoch": 0.03, "learning_rate": 1.9998982743315643e-05, "loss": 1.2133, "step": 407 }, { "epoch": 0.03, "learning_rate": 1.9998943242549894e-05, "loss": 1.2067, "step": 408 }, { "epoch": 0.03, "learning_rate": 1.9998902989442118e-05, "loss": 1.1978, "step": 409 }, { "epoch": 0.03, "learning_rate": 1.9998861983995338e-05, "loss": 1.2272, "step": 410 }, { "epoch": 0.03, "learning_rate": 1.9998820226212648e-05, "loss": 1.1799, "step": 411 }, { "epoch": 0.03, "learning_rate": 1.999877771609718e-05, "loss": 1.1651, "step": 412 }, { "epoch": 0.03, "learning_rate": 1.999873445365214e-05, "loss": 1.2148, "step": 413 }, { "epoch": 0.04, "learning_rate": 1.9998690438880775e-05, "loss": 1.1754, "step": 414 }, { "epoch": 0.04, "learning_rate": 1.9998645671786406e-05, "loss": 1.1633, "step": 415 }, { "epoch": 0.04, "learning_rate": 1.9998600152372396e-05, "loss": 1.1922, "step": 416 }, { "epoch": 0.04, "learning_rate": 1.999855388064217e-05, "loss": 0.9027, "step": 417 }, { "epoch": 0.04, "learning_rate": 1.9998506856599213e-05, "loss": 1.2272, "step": 418 }, { "epoch": 0.04, "learning_rate": 1.999845908024706e-05, "loss": 1.2593, "step": 419 }, { "epoch": 0.04, "learning_rate": 1.9998410551589308e-05, "loss": 1.1796, "step": 420 }, { "epoch": 0.04, "learning_rate": 1.9998361270629607e-05, "loss": 1.2075, "step": 421 }, { "epoch": 0.04, "learning_rate": 1.999831123737166e-05, "loss": 1.2436, "step": 422 }, { "epoch": 0.04, "learning_rate": 1.999826045181924e-05, "loss": 1.1952, "step": 423 }, { "epoch": 0.04, "learning_rate": 1.9998208913976166e-05, "loss": 1.1754, "step": 424 }, { "epoch": 0.04, "learning_rate": 1.9998156623846316e-05, "loss": 1.2252, "step": 425 }, { "epoch": 0.04, "learning_rate": 1.9998103581433623e-05, "loss": 1.1929, "step": 426 }, { "epoch": 0.04, "learning_rate": 1.9998049786742075e-05, "loss": 1.2134, "step": 427 }, { "epoch": 0.04, "learning_rate": 1.9997995239775725e-05, "loss": 1.1992, "step": 428 }, { "epoch": 0.04, "learning_rate": 1.9997939940538673e-05, "loss": 1.2128, "step": 429 }, { "epoch": 0.04, "learning_rate": 1.9997883889035086e-05, "loss": 1.2207, "step": 430 }, { "epoch": 0.04, "learning_rate": 1.9997827085269174e-05, "loss": 1.2373, "step": 431 }, { "epoch": 0.04, "learning_rate": 1.9997769529245215e-05, "loss": 1.1953, "step": 432 }, { "epoch": 0.04, "learning_rate": 1.999771122096754e-05, "loss": 1.1751, "step": 433 }, { "epoch": 0.04, "learning_rate": 1.9997652160440534e-05, "loss": 1.1669, "step": 434 }, { "epoch": 0.04, "learning_rate": 1.9997592347668645e-05, "loss": 1.204, "step": 435 }, { "epoch": 0.04, "learning_rate": 1.9997531782656367e-05, "loss": 1.1849, "step": 436 }, { "epoch": 0.04, "learning_rate": 1.999747046540826e-05, "loss": 1.1624, "step": 437 }, { "epoch": 0.04, "learning_rate": 1.9997408395928944e-05, "loss": 1.1832, "step": 438 }, { "epoch": 0.04, "learning_rate": 1.9997345574223078e-05, "loss": 1.2121, "step": 439 }, { "epoch": 0.04, "learning_rate": 1.9997282000295398e-05, "loss": 1.187, "step": 440 }, { "epoch": 0.04, "learning_rate": 1.9997217674150682e-05, "loss": 1.1637, "step": 441 }, { "epoch": 0.04, "learning_rate": 1.999715259579377e-05, "loss": 1.2009, "step": 442 }, { "epoch": 0.04, "learning_rate": 1.9997086765229562e-05, "loss": 1.1717, "step": 443 }, { "epoch": 0.04, "learning_rate": 1.9997020182463013e-05, "loss": 1.2127, "step": 444 }, { "epoch": 0.04, "learning_rate": 1.9996952847499124e-05, "loss": 1.2304, "step": 445 }, { "epoch": 0.04, "learning_rate": 1.999688476034297e-05, "loss": 1.2148, "step": 446 }, { "epoch": 0.04, "learning_rate": 1.999681592099967e-05, "loss": 1.2392, "step": 447 }, { "epoch": 0.04, "learning_rate": 1.9996746329474407e-05, "loss": 1.1912, "step": 448 }, { "epoch": 0.04, "learning_rate": 1.999667598577241e-05, "loss": 1.1999, "step": 449 }, { "epoch": 0.04, "learning_rate": 1.999660488989898e-05, "loss": 1.1942, "step": 450 }, { "epoch": 0.04, "learning_rate": 1.999653304185946e-05, "loss": 1.198, "step": 451 }, { "epoch": 0.04, "learning_rate": 1.9996460441659263e-05, "loss": 1.1643, "step": 452 }, { "epoch": 0.04, "learning_rate": 1.9996387089303847e-05, "loss": 1.219, "step": 453 }, { "epoch": 0.04, "learning_rate": 1.999631298479873e-05, "loss": 1.1611, "step": 454 }, { "epoch": 0.04, "learning_rate": 1.999623812814949e-05, "loss": 1.1906, "step": 455 }, { "epoch": 0.04, "learning_rate": 1.9996162519361758e-05, "loss": 1.1952, "step": 456 }, { "epoch": 0.04, "learning_rate": 1.9996086158441228e-05, "loss": 1.1874, "step": 457 }, { "epoch": 0.04, "learning_rate": 1.9996009045393635e-05, "loss": 1.1983, "step": 458 }, { "epoch": 0.04, "learning_rate": 1.9995931180224794e-05, "loss": 1.232, "step": 459 }, { "epoch": 0.04, "learning_rate": 1.9995852562940555e-05, "loss": 1.1675, "step": 460 }, { "epoch": 0.04, "learning_rate": 1.9995773193546834e-05, "loss": 1.1404, "step": 461 }, { "epoch": 0.04, "learning_rate": 1.9995693072049608e-05, "loss": 1.2031, "step": 462 }, { "epoch": 0.04, "learning_rate": 1.9995612198454898e-05, "loss": 1.1912, "step": 463 }, { "epoch": 0.04, "learning_rate": 1.9995530572768796e-05, "loss": 1.2365, "step": 464 }, { "epoch": 0.04, "learning_rate": 1.999544819499744e-05, "loss": 1.207, "step": 465 }, { "epoch": 0.04, "learning_rate": 1.999536506514703e-05, "loss": 1.1688, "step": 466 }, { "epoch": 0.04, "learning_rate": 1.999528118322382e-05, "loss": 1.1122, "step": 467 }, { "epoch": 0.04, "learning_rate": 1.999519654923412e-05, "loss": 1.1992, "step": 468 }, { "epoch": 0.04, "learning_rate": 1.9995111163184304e-05, "loss": 1.216, "step": 469 }, { "epoch": 0.04, "learning_rate": 1.9995025025080787e-05, "loss": 1.2111, "step": 470 }, { "epoch": 0.04, "learning_rate": 1.9994938134930056e-05, "loss": 1.1766, "step": 471 }, { "epoch": 0.04, "learning_rate": 1.9994850492738653e-05, "loss": 1.1768, "step": 472 }, { "epoch": 0.04, "learning_rate": 1.9994762098513164e-05, "loss": 1.2086, "step": 473 }, { "epoch": 0.04, "learning_rate": 1.9994672952260244e-05, "loss": 1.1788, "step": 474 }, { "epoch": 0.04, "learning_rate": 1.99945830539866e-05, "loss": 0.9225, "step": 475 }, { "epoch": 0.04, "learning_rate": 1.9994492403698994e-05, "loss": 1.1944, "step": 476 }, { "epoch": 0.04, "learning_rate": 1.9994401001404254e-05, "loss": 1.173, "step": 477 }, { "epoch": 0.04, "learning_rate": 1.999430884710925e-05, "loss": 1.1835, "step": 478 }, { "epoch": 0.04, "learning_rate": 1.9994215940820918e-05, "loss": 1.1779, "step": 479 }, { "epoch": 0.04, "learning_rate": 1.999412228254625e-05, "loss": 1.2216, "step": 480 }, { "epoch": 0.04, "learning_rate": 1.999402787229229e-05, "loss": 1.2298, "step": 481 }, { "epoch": 0.04, "learning_rate": 1.999393271006614e-05, "loss": 1.173, "step": 482 }, { "epoch": 0.04, "learning_rate": 1.999383679587497e-05, "loss": 1.2147, "step": 483 }, { "epoch": 0.04, "learning_rate": 1.999374012972599e-05, "loss": 1.143, "step": 484 }, { "epoch": 0.04, "learning_rate": 1.999364271162647e-05, "loss": 1.1983, "step": 485 }, { "epoch": 0.04, "learning_rate": 1.9993544541583744e-05, "loss": 1.1775, "step": 486 }, { "epoch": 0.04, "learning_rate": 1.99934456196052e-05, "loss": 1.1528, "step": 487 }, { "epoch": 0.04, "learning_rate": 1.999334594569828e-05, "loss": 1.1881, "step": 488 }, { "epoch": 0.04, "learning_rate": 1.9993245519870483e-05, "loss": 1.2013, "step": 489 }, { "epoch": 0.04, "learning_rate": 1.9993144342129364e-05, "loss": 1.1518, "step": 490 }, { "epoch": 0.04, "learning_rate": 1.9993042412482536e-05, "loss": 1.1745, "step": 491 }, { "epoch": 0.04, "learning_rate": 1.999293973093767e-05, "loss": 1.1474, "step": 492 }, { "epoch": 0.04, "learning_rate": 1.9992836297502492e-05, "loss": 1.1466, "step": 493 }, { "epoch": 0.04, "learning_rate": 1.9992732112184784e-05, "loss": 1.1565, "step": 494 }, { "epoch": 0.04, "learning_rate": 1.999262717499239e-05, "loss": 1.1966, "step": 495 }, { "epoch": 0.04, "learning_rate": 1.9992521485933193e-05, "loss": 1.1664, "step": 496 }, { "epoch": 0.04, "learning_rate": 1.9992415045015156e-05, "loss": 1.1933, "step": 497 }, { "epoch": 0.04, "learning_rate": 1.9992307852246283e-05, "loss": 1.1647, "step": 498 }, { "epoch": 0.04, "learning_rate": 1.9992199907634643e-05, "loss": 1.2344, "step": 499 }, { "epoch": 0.04, "learning_rate": 1.999209121118836e-05, "loss": 1.2098, "step": 500 }, { "epoch": 0.04, "learning_rate": 1.9991981762915605e-05, "loss": 1.1975, "step": 501 }, { "epoch": 0.04, "learning_rate": 1.999187156282462e-05, "loss": 1.2323, "step": 502 }, { "epoch": 0.04, "learning_rate": 1.999176061092369e-05, "loss": 1.1987, "step": 503 }, { "epoch": 0.04, "learning_rate": 1.9991648907221172e-05, "loss": 1.2452, "step": 504 }, { "epoch": 0.04, "learning_rate": 1.999153645172546e-05, "loss": 1.1576, "step": 505 }, { "epoch": 0.04, "learning_rate": 1.9991423244445024e-05, "loss": 1.2196, "step": 506 }, { "epoch": 0.04, "learning_rate": 1.999130928538838e-05, "loss": 1.2242, "step": 507 }, { "epoch": 0.04, "learning_rate": 1.9991194574564098e-05, "loss": 1.1847, "step": 508 }, { "epoch": 0.04, "learning_rate": 1.9991079111980817e-05, "loss": 1.2107, "step": 509 }, { "epoch": 0.04, "learning_rate": 1.999096289764722e-05, "loss": 1.2522, "step": 510 }, { "epoch": 0.04, "learning_rate": 1.9990845931572048e-05, "loss": 1.2079, "step": 511 }, { "epoch": 0.04, "learning_rate": 1.999072821376411e-05, "loss": 1.2324, "step": 512 }, { "epoch": 0.04, "learning_rate": 1.9990609744232254e-05, "loss": 0.8896, "step": 513 }, { "epoch": 0.04, "learning_rate": 1.9990490522985403e-05, "loss": 1.2021, "step": 514 }, { "epoch": 0.04, "learning_rate": 1.9990370550032522e-05, "loss": 1.1228, "step": 515 }, { "epoch": 0.04, "learning_rate": 1.9990249825382637e-05, "loss": 1.152, "step": 516 }, { "epoch": 0.04, "learning_rate": 1.9990128349044836e-05, "loss": 1.2051, "step": 517 }, { "epoch": 0.04, "learning_rate": 1.999000612102826e-05, "loss": 1.2017, "step": 518 }, { "epoch": 0.04, "learning_rate": 1.9989883141342098e-05, "loss": 1.2153, "step": 519 }, { "epoch": 0.04, "learning_rate": 1.998975940999561e-05, "loss": 1.2556, "step": 520 }, { "epoch": 0.04, "learning_rate": 1.99896349269981e-05, "loss": 1.2086, "step": 521 }, { "epoch": 0.04, "learning_rate": 1.998950969235894e-05, "loss": 1.173, "step": 522 }, { "epoch": 0.04, "learning_rate": 1.998938370608755e-05, "loss": 1.098, "step": 523 }, { "epoch": 0.04, "learning_rate": 1.9989256968193412e-05, "loss": 1.187, "step": 524 }, { "epoch": 0.04, "learning_rate": 1.998912947868606e-05, "loss": 1.1771, "step": 525 }, { "epoch": 0.04, "learning_rate": 1.9989001237575094e-05, "loss": 1.2503, "step": 526 }, { "epoch": 0.04, "learning_rate": 1.998887224487015e-05, "loss": 1.1905, "step": 527 }, { "epoch": 0.04, "learning_rate": 1.9988742500580936e-05, "loss": 1.2188, "step": 528 }, { "epoch": 0.04, "learning_rate": 1.998861200471722e-05, "loss": 1.1705, "step": 529 }, { "epoch": 0.04, "learning_rate": 1.9988480757288823e-05, "loss": 1.2011, "step": 530 }, { "epoch": 0.04, "learning_rate": 1.9988348758305613e-05, "loss": 1.2001, "step": 531 }, { "epoch": 0.05, "learning_rate": 1.9988216007777524e-05, "loss": 1.2172, "step": 532 }, { "epoch": 0.05, "learning_rate": 1.998808250571455e-05, "loss": 1.1977, "step": 533 }, { "epoch": 0.05, "learning_rate": 1.998794825212673e-05, "loss": 1.1539, "step": 534 }, { "epoch": 0.05, "learning_rate": 1.9987813247024166e-05, "loss": 1.1723, "step": 535 }, { "epoch": 0.05, "learning_rate": 1.998767749041702e-05, "loss": 1.1994, "step": 536 }, { "epoch": 0.05, "learning_rate": 1.9987540982315497e-05, "loss": 1.1373, "step": 537 }, { "epoch": 0.05, "learning_rate": 1.9987403722729882e-05, "loss": 1.1626, "step": 538 }, { "epoch": 0.05, "learning_rate": 1.998726571167049e-05, "loss": 1.183, "step": 539 }, { "epoch": 0.05, "learning_rate": 1.9987126949147714e-05, "loss": 1.2028, "step": 540 }, { "epoch": 0.05, "learning_rate": 1.998698743517199e-05, "loss": 1.2013, "step": 541 }, { "epoch": 0.05, "learning_rate": 1.9986847169753824e-05, "loss": 1.2198, "step": 542 }, { "epoch": 0.05, "learning_rate": 1.9986706152903754e-05, "loss": 1.1608, "step": 543 }, { "epoch": 0.05, "learning_rate": 1.9986564384632404e-05, "loss": 1.2059, "step": 544 }, { "epoch": 0.05, "learning_rate": 1.9986421864950435e-05, "loss": 1.1776, "step": 545 }, { "epoch": 0.05, "learning_rate": 1.998627859386857e-05, "loss": 1.2478, "step": 546 }, { "epoch": 0.05, "learning_rate": 1.9986134571397592e-05, "loss": 1.2106, "step": 547 }, { "epoch": 0.05, "learning_rate": 1.998598979754834e-05, "loss": 1.2159, "step": 548 }, { "epoch": 0.05, "learning_rate": 1.99858442723317e-05, "loss": 1.1999, "step": 549 }, { "epoch": 0.05, "learning_rate": 1.9985697995758623e-05, "loss": 1.1479, "step": 550 }, { "epoch": 0.05, "learning_rate": 1.998555096784012e-05, "loss": 1.1496, "step": 551 }, { "epoch": 0.05, "learning_rate": 1.9985403188587252e-05, "loss": 1.1887, "step": 552 }, { "epoch": 0.05, "learning_rate": 1.9985254658011135e-05, "loss": 1.233, "step": 553 }, { "epoch": 0.05, "learning_rate": 1.9985105376122947e-05, "loss": 1.1533, "step": 554 }, { "epoch": 0.05, "learning_rate": 1.998495534293392e-05, "loss": 1.1755, "step": 555 }, { "epoch": 0.05, "learning_rate": 1.9984804558455344e-05, "loss": 1.1891, "step": 556 }, { "epoch": 0.05, "learning_rate": 1.9984653022698562e-05, "loss": 1.1359, "step": 557 }, { "epoch": 0.05, "learning_rate": 1.998450073567498e-05, "loss": 1.2069, "step": 558 }, { "epoch": 0.05, "learning_rate": 1.9984347697396052e-05, "loss": 1.1694, "step": 559 }, { "epoch": 0.05, "learning_rate": 1.9984193907873296e-05, "loss": 1.1687, "step": 560 }, { "epoch": 0.05, "learning_rate": 1.998403936711828e-05, "loss": 1.1846, "step": 561 }, { "epoch": 0.05, "learning_rate": 1.9983884075142634e-05, "loss": 1.1882, "step": 562 }, { "epoch": 0.05, "learning_rate": 1.9983728031958045e-05, "loss": 1.1444, "step": 563 }, { "epoch": 0.05, "learning_rate": 1.998357123757625e-05, "loss": 1.1588, "step": 564 }, { "epoch": 0.05, "learning_rate": 1.9983413692009046e-05, "loss": 1.1911, "step": 565 }, { "epoch": 0.05, "learning_rate": 1.998325539526829e-05, "loss": 1.1593, "step": 566 }, { "epoch": 0.05, "learning_rate": 1.9983096347365895e-05, "loss": 1.1398, "step": 567 }, { "epoch": 0.05, "learning_rate": 1.9982936548313822e-05, "loss": 1.1911, "step": 568 }, { "epoch": 0.05, "learning_rate": 1.99827759981241e-05, "loss": 1.1738, "step": 569 }, { "epoch": 0.05, "learning_rate": 1.99826146968088e-05, "loss": 1.197, "step": 570 }, { "epoch": 0.05, "learning_rate": 1.998245264438007e-05, "loss": 1.2702, "step": 571 }, { "epoch": 0.05, "learning_rate": 1.99822898408501e-05, "loss": 1.1904, "step": 572 }, { "epoch": 0.05, "learning_rate": 1.998212628623114e-05, "loss": 1.2698, "step": 573 }, { "epoch": 0.05, "learning_rate": 1.998196198053549e-05, "loss": 1.186, "step": 574 }, { "epoch": 0.05, "learning_rate": 1.9981796923775516e-05, "loss": 1.2011, "step": 575 }, { "epoch": 0.05, "learning_rate": 1.998163111596364e-05, "loss": 1.144, "step": 576 }, { "epoch": 0.05, "learning_rate": 1.9981464557112336e-05, "loss": 1.1935, "step": 577 }, { "epoch": 0.05, "learning_rate": 1.9981297247234138e-05, "loss": 1.2327, "step": 578 }, { "epoch": 0.05, "learning_rate": 1.998112918634163e-05, "loss": 1.1765, "step": 579 }, { "epoch": 0.05, "learning_rate": 1.9980960374447463e-05, "loss": 1.1803, "step": 580 }, { "epoch": 0.05, "learning_rate": 1.9980790811564333e-05, "loss": 1.137, "step": 581 }, { "epoch": 0.05, "learning_rate": 1.9980620497705004e-05, "loss": 1.1989, "step": 582 }, { "epoch": 0.05, "learning_rate": 1.9980449432882286e-05, "loss": 1.1841, "step": 583 }, { "epoch": 0.05, "learning_rate": 1.9980277617109055e-05, "loss": 1.1529, "step": 584 }, { "epoch": 0.05, "learning_rate": 1.9980105050398236e-05, "loss": 1.186, "step": 585 }, { "epoch": 0.05, "learning_rate": 1.9979931732762812e-05, "loss": 1.2087, "step": 586 }, { "epoch": 0.05, "learning_rate": 1.9979757664215828e-05, "loss": 1.1937, "step": 587 }, { "epoch": 0.05, "learning_rate": 1.9979582844770375e-05, "loss": 1.1856, "step": 588 }, { "epoch": 0.05, "learning_rate": 1.997940727443961e-05, "loss": 0.8899, "step": 589 }, { "epoch": 0.05, "learning_rate": 1.9979230953236746e-05, "loss": 1.1277, "step": 590 }, { "epoch": 0.05, "learning_rate": 1.9979053881175048e-05, "loss": 1.2166, "step": 591 }, { "epoch": 0.05, "learning_rate": 1.9978876058267834e-05, "loss": 1.1867, "step": 592 }, { "epoch": 0.05, "learning_rate": 1.9978697484528495e-05, "loss": 1.1213, "step": 593 }, { "epoch": 0.05, "learning_rate": 1.9978518159970457e-05, "loss": 1.1769, "step": 594 }, { "epoch": 0.05, "learning_rate": 1.9978338084607216e-05, "loss": 1.166, "step": 595 }, { "epoch": 0.05, "learning_rate": 1.9978157258452324e-05, "loss": 1.2302, "step": 596 }, { "epoch": 0.05, "learning_rate": 1.9977975681519385e-05, "loss": 1.1783, "step": 597 }, { "epoch": 0.05, "learning_rate": 1.997779335382206e-05, "loss": 1.172, "step": 598 }, { "epoch": 0.05, "learning_rate": 1.9977610275374068e-05, "loss": 1.1586, "step": 599 }, { "epoch": 0.05, "learning_rate": 1.9977426446189185e-05, "loss": 1.1943, "step": 600 }, { "epoch": 0.05, "learning_rate": 1.997724186628124e-05, "loss": 1.2014, "step": 601 }, { "epoch": 0.05, "learning_rate": 1.9977056535664126e-05, "loss": 1.2212, "step": 602 }, { "epoch": 0.05, "learning_rate": 1.9976870454351783e-05, "loss": 1.1962, "step": 603 }, { "epoch": 0.05, "learning_rate": 1.997668362235822e-05, "loss": 1.1529, "step": 604 }, { "epoch": 0.05, "learning_rate": 1.9976496039697483e-05, "loss": 1.1737, "step": 605 }, { "epoch": 0.05, "learning_rate": 1.9976307706383695e-05, "loss": 1.1647, "step": 606 }, { "epoch": 0.05, "learning_rate": 1.9976118622431024e-05, "loss": 1.1767, "step": 607 }, { "epoch": 0.05, "learning_rate": 1.9975928787853696e-05, "loss": 1.1823, "step": 608 }, { "epoch": 0.05, "learning_rate": 1.9975738202666e-05, "loss": 1.1853, "step": 609 }, { "epoch": 0.05, "learning_rate": 1.9975546866882264e-05, "loss": 1.2208, "step": 610 }, { "epoch": 0.05, "learning_rate": 1.9975354780516896e-05, "loss": 1.1914, "step": 611 }, { "epoch": 0.05, "learning_rate": 1.9975161943584344e-05, "loss": 1.1042, "step": 612 }, { "epoch": 0.05, "learning_rate": 1.997496835609912e-05, "loss": 1.1154, "step": 613 }, { "epoch": 0.05, "learning_rate": 1.9974774018075788e-05, "loss": 1.1772, "step": 614 }, { "epoch": 0.05, "learning_rate": 1.997457892952897e-05, "loss": 1.1757, "step": 615 }, { "epoch": 0.05, "learning_rate": 1.9974383090473344e-05, "loss": 1.2295, "step": 616 }, { "epoch": 0.05, "learning_rate": 1.997418650092365e-05, "loss": 1.1108, "step": 617 }, { "epoch": 0.05, "learning_rate": 1.9973989160894676e-05, "loss": 1.1794, "step": 618 }, { "epoch": 0.05, "learning_rate": 1.9973791070401273e-05, "loss": 1.193, "step": 619 }, { "epoch": 0.05, "learning_rate": 1.997359222945834e-05, "loss": 1.1911, "step": 620 }, { "epoch": 0.05, "learning_rate": 1.9973392638080845e-05, "loss": 1.2561, "step": 621 }, { "epoch": 0.05, "learning_rate": 1.9973192296283802e-05, "loss": 1.1922, "step": 622 }, { "epoch": 0.05, "learning_rate": 1.9972991204082287e-05, "loss": 1.174, "step": 623 }, { "epoch": 0.05, "learning_rate": 1.997278936149143e-05, "loss": 1.2093, "step": 624 }, { "epoch": 0.05, "learning_rate": 1.9972586768526414e-05, "loss": 1.178, "step": 625 }, { "epoch": 0.05, "learning_rate": 1.997238342520249e-05, "loss": 1.1839, "step": 626 }, { "epoch": 0.05, "learning_rate": 1.997217933153495e-05, "loss": 1.0629, "step": 627 }, { "epoch": 0.05, "learning_rate": 1.9971974487539157e-05, "loss": 1.14, "step": 628 }, { "epoch": 0.05, "learning_rate": 1.9971768893230524e-05, "loss": 1.1623, "step": 629 }, { "epoch": 0.05, "learning_rate": 1.9971562548624517e-05, "loss": 1.1807, "step": 630 }, { "epoch": 0.05, "learning_rate": 1.997135545373666e-05, "loss": 1.1597, "step": 631 }, { "epoch": 0.05, "learning_rate": 1.9971147608582542e-05, "loss": 1.1876, "step": 632 }, { "epoch": 0.05, "learning_rate": 1.9970939013177792e-05, "loss": 1.1437, "step": 633 }, { "epoch": 0.05, "learning_rate": 1.9970729667538115e-05, "loss": 1.1462, "step": 634 }, { "epoch": 0.05, "learning_rate": 1.997051957167926e-05, "loss": 1.181, "step": 635 }, { "epoch": 0.05, "learning_rate": 1.997030872561703e-05, "loss": 1.2237, "step": 636 }, { "epoch": 0.05, "learning_rate": 1.9970097129367293e-05, "loss": 0.935, "step": 637 }, { "epoch": 0.05, "learning_rate": 1.9969884782945973e-05, "loss": 1.2022, "step": 638 }, { "epoch": 0.05, "learning_rate": 1.9969671686369043e-05, "loss": 1.111, "step": 639 }, { "epoch": 0.05, "learning_rate": 1.9969457839652537e-05, "loss": 1.1723, "step": 640 }, { "epoch": 0.05, "learning_rate": 1.9969243242812548e-05, "loss": 1.17, "step": 641 }, { "epoch": 0.05, "learning_rate": 1.9969027895865222e-05, "loss": 1.126, "step": 642 }, { "epoch": 0.05, "learning_rate": 1.996881179882676e-05, "loss": 1.1534, "step": 643 }, { "epoch": 0.05, "learning_rate": 1.9968594951713423e-05, "loss": 1.1139, "step": 644 }, { "epoch": 0.05, "learning_rate": 1.9968377354541528e-05, "loss": 1.1522, "step": 645 }, { "epoch": 0.05, "learning_rate": 1.9968159007327442e-05, "loss": 1.1899, "step": 646 }, { "epoch": 0.05, "learning_rate": 1.9967939910087607e-05, "loss": 0.9345, "step": 647 }, { "epoch": 0.05, "learning_rate": 1.9967720062838494e-05, "loss": 1.2403, "step": 648 }, { "epoch": 0.05, "learning_rate": 1.9967499465596652e-05, "loss": 1.2027, "step": 649 }, { "epoch": 0.06, "learning_rate": 1.9967278118378676e-05, "loss": 1.1112, "step": 650 }, { "epoch": 0.06, "learning_rate": 1.9967056021201228e-05, "loss": 0.8931, "step": 651 }, { "epoch": 0.06, "learning_rate": 1.996683317408101e-05, "loss": 1.1541, "step": 652 }, { "epoch": 0.06, "learning_rate": 1.9966609577034793e-05, "loss": 1.1984, "step": 653 }, { "epoch": 0.06, "learning_rate": 1.9966385230079398e-05, "loss": 1.1396, "step": 654 }, { "epoch": 0.06, "learning_rate": 1.9966160133231712e-05, "loss": 1.1995, "step": 655 }, { "epoch": 0.06, "learning_rate": 1.996593428650867e-05, "loss": 1.1858, "step": 656 }, { "epoch": 0.06, "learning_rate": 1.996570768992726e-05, "loss": 1.1602, "step": 657 }, { "epoch": 0.06, "learning_rate": 1.9965480343504534e-05, "loss": 1.121, "step": 658 }, { "epoch": 0.06, "learning_rate": 1.99652522472576e-05, "loss": 1.152, "step": 659 }, { "epoch": 0.06, "learning_rate": 1.996502340120362e-05, "loss": 1.1528, "step": 660 }, { "epoch": 0.06, "learning_rate": 1.9964793805359814e-05, "loss": 1.2321, "step": 661 }, { "epoch": 0.06, "learning_rate": 1.996456345974345e-05, "loss": 1.1874, "step": 662 }, { "epoch": 0.06, "learning_rate": 1.9964332364371872e-05, "loss": 1.1499, "step": 663 }, { "epoch": 0.06, "learning_rate": 1.9964100519262456e-05, "loss": 1.1621, "step": 664 }, { "epoch": 0.06, "learning_rate": 1.9963867924432656e-05, "loss": 1.2025, "step": 665 }, { "epoch": 0.06, "learning_rate": 1.9963634579899965e-05, "loss": 1.1498, "step": 666 }, { "epoch": 0.06, "learning_rate": 1.996340048568195e-05, "loss": 1.1972, "step": 667 }, { "epoch": 0.06, "learning_rate": 1.996316564179621e-05, "loss": 1.1381, "step": 668 }, { "epoch": 0.06, "learning_rate": 1.996293004826043e-05, "loss": 1.0136, "step": 669 }, { "epoch": 0.06, "learning_rate": 1.996269370509233e-05, "loss": 1.1774, "step": 670 }, { "epoch": 0.06, "learning_rate": 1.9962456612309693e-05, "loss": 1.1932, "step": 671 }, { "epoch": 0.06, "learning_rate": 1.996221876993036e-05, "loss": 1.1185, "step": 672 }, { "epoch": 0.06, "learning_rate": 1.9961980177972225e-05, "loss": 1.1226, "step": 673 }, { "epoch": 0.06, "learning_rate": 1.996174083645324e-05, "loss": 1.1787, "step": 674 }, { "epoch": 0.06, "learning_rate": 1.9961500745391418e-05, "loss": 1.1964, "step": 675 }, { "epoch": 0.06, "learning_rate": 1.996125990480482e-05, "loss": 1.2177, "step": 676 }, { "epoch": 0.06, "learning_rate": 1.9961018314711565e-05, "loss": 1.177, "step": 677 }, { "epoch": 0.06, "learning_rate": 1.9960775975129833e-05, "loss": 1.1756, "step": 678 }, { "epoch": 0.06, "learning_rate": 1.996053288607786e-05, "loss": 1.2232, "step": 679 }, { "epoch": 0.06, "learning_rate": 1.9960289047573937e-05, "loss": 1.2583, "step": 680 }, { "epoch": 0.06, "learning_rate": 1.9960044459636404e-05, "loss": 1.1727, "step": 681 }, { "epoch": 0.06, "learning_rate": 1.9959799122283675e-05, "loss": 1.1856, "step": 682 }, { "epoch": 0.06, "learning_rate": 1.9959553035534204e-05, "loss": 1.1341, "step": 683 }, { "epoch": 0.06, "learning_rate": 1.9959306199406504e-05, "loss": 1.1925, "step": 684 }, { "epoch": 0.06, "learning_rate": 1.9959058613919155e-05, "loss": 1.1697, "step": 685 }, { "epoch": 0.06, "learning_rate": 1.9958810279090776e-05, "loss": 1.131, "step": 686 }, { "epoch": 0.06, "learning_rate": 1.995856119494006e-05, "loss": 1.1656, "step": 687 }, { "epoch": 0.06, "learning_rate": 1.9958311361485752e-05, "loss": 1.1511, "step": 688 }, { "epoch": 0.06, "learning_rate": 1.995806077874664e-05, "loss": 1.17, "step": 689 }, { "epoch": 0.06, "learning_rate": 1.9957809446741584e-05, "loss": 1.1564, "step": 690 }, { "epoch": 0.06, "learning_rate": 1.995755736548949e-05, "loss": 1.191, "step": 691 }, { "epoch": 0.06, "learning_rate": 1.9957304535009334e-05, "loss": 1.2054, "step": 692 }, { "epoch": 0.06, "learning_rate": 1.9957050955320132e-05, "loss": 1.1431, "step": 693 }, { "epoch": 0.06, "learning_rate": 1.9956796626440967e-05, "loss": 1.1606, "step": 694 }, { "epoch": 0.06, "learning_rate": 1.9956541548390973e-05, "loss": 1.2005, "step": 695 }, { "epoch": 0.06, "learning_rate": 1.9956285721189347e-05, "loss": 1.1826, "step": 696 }, { "epoch": 0.06, "learning_rate": 1.9956029144855332e-05, "loss": 1.1443, "step": 697 }, { "epoch": 0.06, "learning_rate": 1.995577181940824e-05, "loss": 1.2068, "step": 698 }, { "epoch": 0.06, "learning_rate": 1.9955513744867426e-05, "loss": 1.1967, "step": 699 }, { "epoch": 0.06, "learning_rate": 1.995525492125231e-05, "loss": 1.1551, "step": 700 }, { "epoch": 0.06, "learning_rate": 1.9954995348582373e-05, "loss": 0.928, "step": 701 }, { "epoch": 0.06, "learning_rate": 1.9954735026877138e-05, "loss": 1.1683, "step": 702 }, { "epoch": 0.06, "learning_rate": 1.9954473956156195e-05, "loss": 1.2117, "step": 703 }, { "epoch": 0.06, "learning_rate": 1.9954212136439185e-05, "loss": 1.2192, "step": 704 }, { "epoch": 0.06, "learning_rate": 1.995394956774581e-05, "loss": 1.1865, "step": 705 }, { "epoch": 0.06, "learning_rate": 1.9953686250095832e-05, "loss": 1.1263, "step": 706 }, { "epoch": 0.06, "learning_rate": 1.9953422183509056e-05, "loss": 1.195, "step": 707 }, { "epoch": 0.06, "learning_rate": 1.9953157368005346e-05, "loss": 1.1811, "step": 708 }, { "epoch": 0.06, "learning_rate": 1.9952891803604642e-05, "loss": 1.1969, "step": 709 }, { "epoch": 0.06, "learning_rate": 1.9952625490326913e-05, "loss": 1.1855, "step": 710 }, { "epoch": 0.06, "learning_rate": 1.9952358428192205e-05, "loss": 1.1211, "step": 711 }, { "epoch": 0.06, "learning_rate": 1.9952090617220607e-05, "loss": 1.1819, "step": 712 }, { "epoch": 0.06, "learning_rate": 1.9951822057432274e-05, "loss": 1.1989, "step": 713 }, { "epoch": 0.06, "learning_rate": 1.9951552748847408e-05, "loss": 0.952, "step": 714 }, { "epoch": 0.06, "learning_rate": 1.9951282691486277e-05, "loss": 1.14, "step": 715 }, { "epoch": 0.06, "learning_rate": 1.9951011885369197e-05, "loss": 1.2115, "step": 716 }, { "epoch": 0.06, "learning_rate": 1.995074033051655e-05, "loss": 1.1658, "step": 717 }, { "epoch": 0.06, "learning_rate": 1.995046802694876e-05, "loss": 1.1897, "step": 718 }, { "epoch": 0.06, "learning_rate": 1.9950194974686323e-05, "loss": 1.1575, "step": 719 }, { "epoch": 0.06, "learning_rate": 1.994992117374978e-05, "loss": 1.1824, "step": 720 }, { "epoch": 0.06, "learning_rate": 1.994964662415973e-05, "loss": 1.1161, "step": 721 }, { "epoch": 0.06, "learning_rate": 1.9949371325936838e-05, "loss": 1.1876, "step": 722 }, { "epoch": 0.06, "learning_rate": 1.9949095279101813e-05, "loss": 1.155, "step": 723 }, { "epoch": 0.06, "learning_rate": 1.9948818483675427e-05, "loss": 1.1109, "step": 724 }, { "epoch": 0.06, "learning_rate": 1.9948540939678507e-05, "loss": 1.1511, "step": 725 }, { "epoch": 0.06, "learning_rate": 1.9948262647131934e-05, "loss": 1.1607, "step": 726 }, { "epoch": 0.06, "learning_rate": 1.9947983606056647e-05, "loss": 0.8811, "step": 727 }, { "epoch": 0.06, "learning_rate": 1.9947703816473645e-05, "loss": 1.1411, "step": 728 }, { "epoch": 0.06, "learning_rate": 1.9947423278403976e-05, "loss": 1.1501, "step": 729 }, { "epoch": 0.06, "learning_rate": 1.9947141991868754e-05, "loss": 1.1673, "step": 730 }, { "epoch": 0.06, "learning_rate": 1.9946859956889138e-05, "loss": 1.2101, "step": 731 }, { "epoch": 0.06, "learning_rate": 1.994657717348635e-05, "loss": 1.1464, "step": 732 }, { "epoch": 0.06, "learning_rate": 1.994629364168167e-05, "loss": 1.1547, "step": 733 }, { "epoch": 0.06, "learning_rate": 1.994600936149643e-05, "loss": 1.181, "step": 734 }, { "epoch": 0.06, "learning_rate": 1.994572433295202e-05, "loss": 1.1387, "step": 735 }, { "epoch": 0.06, "learning_rate": 1.9945438556069884e-05, "loss": 1.1697, "step": 736 }, { "epoch": 0.06, "learning_rate": 1.9945152030871524e-05, "loss": 1.2238, "step": 737 }, { "epoch": 0.06, "learning_rate": 1.9944864757378505e-05, "loss": 1.0838, "step": 738 }, { "epoch": 0.06, "learning_rate": 1.9944576735612434e-05, "loss": 1.175, "step": 739 }, { "epoch": 0.06, "learning_rate": 1.994428796559499e-05, "loss": 1.2103, "step": 740 }, { "epoch": 0.06, "learning_rate": 1.9943998447347897e-05, "loss": 1.2016, "step": 741 }, { "epoch": 0.06, "learning_rate": 1.9943708180892938e-05, "loss": 1.2175, "step": 742 }, { "epoch": 0.06, "learning_rate": 1.9943417166251953e-05, "loss": 1.1542, "step": 743 }, { "epoch": 0.06, "learning_rate": 1.994312540344684e-05, "loss": 1.2116, "step": 744 }, { "epoch": 0.06, "learning_rate": 1.9942832892499554e-05, "loss": 1.1453, "step": 745 }, { "epoch": 0.06, "learning_rate": 1.9942539633432103e-05, "loss": 1.1436, "step": 746 }, { "epoch": 0.06, "learning_rate": 1.994224562626655e-05, "loss": 1.0696, "step": 747 }, { "epoch": 0.06, "learning_rate": 1.9941950871025016e-05, "loss": 1.1806, "step": 748 }, { "epoch": 0.06, "learning_rate": 1.994165536772968e-05, "loss": 1.1973, "step": 749 }, { "epoch": 0.06, "learning_rate": 1.9941359116402784e-05, "loss": 1.2294, "step": 750 }, { "epoch": 0.06, "learning_rate": 1.9941062117066608e-05, "loss": 0.8258, "step": 751 }, { "epoch": 0.06, "learning_rate": 1.99407643697435e-05, "loss": 1.2024, "step": 752 }, { "epoch": 0.06, "learning_rate": 1.994046587445587e-05, "loss": 1.2068, "step": 753 }, { "epoch": 0.06, "learning_rate": 1.9940166631226174e-05, "loss": 1.1606, "step": 754 }, { "epoch": 0.06, "learning_rate": 1.9939866640076925e-05, "loss": 1.1747, "step": 755 }, { "epoch": 0.06, "learning_rate": 1.99395659010307e-05, "loss": 0.9006, "step": 756 }, { "epoch": 0.06, "learning_rate": 1.993926441411012e-05, "loss": 0.7913, "step": 757 }, { "epoch": 0.06, "learning_rate": 1.9938962179337877e-05, "loss": 1.2091, "step": 758 }, { "epoch": 0.06, "learning_rate": 1.9938659196736708e-05, "loss": 1.2202, "step": 759 }, { "epoch": 0.06, "learning_rate": 1.993835546632941e-05, "loss": 1.2047, "step": 760 }, { "epoch": 0.06, "learning_rate": 1.993805098813884e-05, "loss": 1.1333, "step": 761 }, { "epoch": 0.06, "learning_rate": 1.99377457621879e-05, "loss": 1.1634, "step": 762 }, { "epoch": 0.06, "learning_rate": 1.9937439788499565e-05, "loss": 1.1105, "step": 763 }, { "epoch": 0.06, "learning_rate": 1.993713306709685e-05, "loss": 1.1572, "step": 764 }, { "epoch": 0.06, "learning_rate": 1.993682559800284e-05, "loss": 1.1577, "step": 765 }, { "epoch": 0.06, "learning_rate": 1.9936517381240657e-05, "loss": 1.2014, "step": 766 }, { "epoch": 0.06, "learning_rate": 1.9936208416833508e-05, "loss": 1.2355, "step": 767 }, { "epoch": 0.07, "learning_rate": 1.9935898704804634e-05, "loss": 1.1726, "step": 768 }, { "epoch": 0.07, "learning_rate": 1.9935588245177332e-05, "loss": 1.1898, "step": 769 }, { "epoch": 0.07, "learning_rate": 1.993527703797497e-05, "loss": 1.1762, "step": 770 }, { "epoch": 0.07, "learning_rate": 1.9934965083220962e-05, "loss": 1.1654, "step": 771 }, { "epoch": 0.07, "learning_rate": 1.9934652380938773e-05, "loss": 1.2092, "step": 772 }, { "epoch": 0.07, "learning_rate": 1.9934338931151942e-05, "loss": 0.9072, "step": 773 }, { "epoch": 0.07, "learning_rate": 1.9934024733884048e-05, "loss": 1.1814, "step": 774 }, { "epoch": 0.07, "learning_rate": 1.993370978915873e-05, "loss": 1.2078, "step": 775 }, { "epoch": 0.07, "learning_rate": 1.9933394096999694e-05, "loss": 1.1461, "step": 776 }, { "epoch": 0.07, "learning_rate": 1.993307765743068e-05, "loss": 1.2009, "step": 777 }, { "epoch": 0.07, "learning_rate": 1.9932760470475513e-05, "loss": 1.2076, "step": 778 }, { "epoch": 0.07, "learning_rate": 1.9932442536158043e-05, "loss": 1.0706, "step": 779 }, { "epoch": 0.07, "learning_rate": 1.9932123854502204e-05, "loss": 1.118, "step": 780 }, { "epoch": 0.07, "learning_rate": 1.993180442553197e-05, "loss": 1.2052, "step": 781 }, { "epoch": 0.07, "learning_rate": 1.9931484249271375e-05, "loss": 1.1333, "step": 782 }, { "epoch": 0.07, "learning_rate": 1.993116332574451e-05, "loss": 1.1642, "step": 783 }, { "epoch": 0.07, "learning_rate": 1.993084165497552e-05, "loss": 1.1522, "step": 784 }, { "epoch": 0.07, "learning_rate": 1.993051923698861e-05, "loss": 1.1938, "step": 785 }, { "epoch": 0.07, "learning_rate": 1.9930196071808045e-05, "loss": 1.1703, "step": 786 }, { "epoch": 0.07, "learning_rate": 1.9929872159458134e-05, "loss": 1.1598, "step": 787 }, { "epoch": 0.07, "learning_rate": 1.992954749996325e-05, "loss": 1.1595, "step": 788 }, { "epoch": 0.07, "learning_rate": 1.9929222093347817e-05, "loss": 1.2331, "step": 789 }, { "epoch": 0.07, "learning_rate": 1.992889593963633e-05, "loss": 1.1902, "step": 790 }, { "epoch": 0.07, "learning_rate": 1.992856903885332e-05, "loss": 1.1492, "step": 791 }, { "epoch": 0.07, "learning_rate": 1.9928241391023387e-05, "loss": 1.178, "step": 792 }, { "epoch": 0.07, "learning_rate": 1.9927912996171186e-05, "loss": 1.1601, "step": 793 }, { "epoch": 0.07, "learning_rate": 1.992758385432142e-05, "loss": 1.1485, "step": 794 }, { "epoch": 0.07, "learning_rate": 1.9927253965498866e-05, "loss": 1.2, "step": 795 }, { "epoch": 0.07, "learning_rate": 1.992692332972833e-05, "loss": 1.182, "step": 796 }, { "epoch": 0.07, "learning_rate": 1.9926591947034703e-05, "loss": 1.2075, "step": 797 }, { "epoch": 0.07, "learning_rate": 1.9926259817442912e-05, "loss": 1.1765, "step": 798 }, { "epoch": 0.07, "learning_rate": 1.992592694097795e-05, "loss": 1.164, "step": 799 }, { "epoch": 0.07, "learning_rate": 1.992559331766486e-05, "loss": 1.1331, "step": 800 }, { "epoch": 0.07, "learning_rate": 1.9925258947528747e-05, "loss": 1.1561, "step": 801 }, { "epoch": 0.07, "learning_rate": 1.9924923830594768e-05, "loss": 1.1562, "step": 802 }, { "epoch": 0.07, "learning_rate": 1.9924587966888143e-05, "loss": 1.167, "step": 803 }, { "epoch": 0.07, "learning_rate": 1.9924251356434135e-05, "loss": 1.1605, "step": 804 }, { "epoch": 0.07, "learning_rate": 1.992391399925808e-05, "loss": 1.1137, "step": 805 }, { "epoch": 0.07, "learning_rate": 1.9923575895385357e-05, "loss": 1.1836, "step": 806 }, { "epoch": 0.07, "learning_rate": 1.9923237044841403e-05, "loss": 1.1977, "step": 807 }, { "epoch": 0.07, "learning_rate": 1.992289744765172e-05, "loss": 1.1438, "step": 808 }, { "epoch": 0.07, "learning_rate": 1.9922557103841852e-05, "loss": 0.9618, "step": 809 }, { "epoch": 0.07, "learning_rate": 1.9922216013437416e-05, "loss": 1.1967, "step": 810 }, { "epoch": 0.07, "learning_rate": 1.992187417646407e-05, "loss": 1.1359, "step": 811 }, { "epoch": 0.07, "learning_rate": 1.992153159294754e-05, "loss": 1.1947, "step": 812 }, { "epoch": 0.07, "learning_rate": 1.9921188262913594e-05, "loss": 1.2046, "step": 813 }, { "epoch": 0.07, "learning_rate": 1.9920844186388077e-05, "loss": 1.174, "step": 814 }, { "epoch": 0.07, "learning_rate": 1.992049936339687e-05, "loss": 1.1277, "step": 815 }, { "epoch": 0.07, "learning_rate": 1.9920153793965917e-05, "loss": 1.1458, "step": 816 }, { "epoch": 0.07, "learning_rate": 1.9919807478121226e-05, "loss": 1.1919, "step": 817 }, { "epoch": 0.07, "learning_rate": 1.9919460415888846e-05, "loss": 1.1534, "step": 818 }, { "epoch": 0.07, "learning_rate": 1.9919112607294898e-05, "loss": 1.2023, "step": 819 }, { "epoch": 0.07, "learning_rate": 1.991876405236555e-05, "loss": 0.9341, "step": 820 }, { "epoch": 0.07, "learning_rate": 1.9918414751127026e-05, "loss": 1.1813, "step": 821 }, { "epoch": 0.07, "learning_rate": 1.9918064703605607e-05, "loss": 1.1434, "step": 822 }, { "epoch": 0.07, "learning_rate": 1.991771390982764e-05, "loss": 1.1923, "step": 823 }, { "epoch": 0.07, "learning_rate": 1.991736236981951e-05, "loss": 1.1543, "step": 824 }, { "epoch": 0.07, "learning_rate": 1.9917010083607668e-05, "loss": 1.1585, "step": 825 }, { "epoch": 0.07, "learning_rate": 1.991665705121863e-05, "loss": 1.1823, "step": 826 }, { "epoch": 0.07, "learning_rate": 1.991630327267895e-05, "loss": 1.1955, "step": 827 }, { "epoch": 0.07, "learning_rate": 1.9915948748015248e-05, "loss": 1.1699, "step": 828 }, { "epoch": 0.07, "learning_rate": 1.9915593477254203e-05, "loss": 1.1743, "step": 829 }, { "epoch": 0.07, "learning_rate": 1.991523746042254e-05, "loss": 1.2038, "step": 830 }, { "epoch": 0.07, "learning_rate": 1.9914880697547055e-05, "loss": 1.1609, "step": 831 }, { "epoch": 0.07, "learning_rate": 1.9914523188654583e-05, "loss": 1.1094, "step": 832 }, { "epoch": 0.07, "learning_rate": 1.9914164933772033e-05, "loss": 1.1631, "step": 833 }, { "epoch": 0.07, "learning_rate": 1.9913805932926356e-05, "loss": 1.1676, "step": 834 }, { "epoch": 0.07, "learning_rate": 1.991344618614456e-05, "loss": 1.1689, "step": 835 }, { "epoch": 0.07, "learning_rate": 1.991308569345372e-05, "loss": 1.0968, "step": 836 }, { "epoch": 0.07, "learning_rate": 1.9912724454880957e-05, "loss": 1.1231, "step": 837 }, { "epoch": 0.07, "learning_rate": 1.991236247045345e-05, "loss": 1.1615, "step": 838 }, { "epoch": 0.07, "learning_rate": 1.991199974019844e-05, "loss": 1.1656, "step": 839 }, { "epoch": 0.07, "learning_rate": 1.9911636264143212e-05, "loss": 1.2231, "step": 840 }, { "epoch": 0.07, "learning_rate": 1.9911272042315124e-05, "loss": 1.1656, "step": 841 }, { "epoch": 0.07, "learning_rate": 1.9910907074741576e-05, "loss": 1.185, "step": 842 }, { "epoch": 0.07, "learning_rate": 1.9910541361450027e-05, "loss": 1.1451, "step": 843 }, { "epoch": 0.07, "learning_rate": 1.9910174902467994e-05, "loss": 1.1385, "step": 844 }, { "epoch": 0.07, "learning_rate": 1.9909807697823055e-05, "loss": 1.1851, "step": 845 }, { "epoch": 0.07, "learning_rate": 1.990943974754284e-05, "loss": 1.1413, "step": 846 }, { "epoch": 0.07, "learning_rate": 1.9909071051655026e-05, "loss": 1.17, "step": 847 }, { "epoch": 0.07, "learning_rate": 1.9908701610187364e-05, "loss": 1.1553, "step": 848 }, { "epoch": 0.07, "learning_rate": 1.9908331423167644e-05, "loss": 1.1284, "step": 849 }, { "epoch": 0.07, "learning_rate": 1.9907960490623727e-05, "loss": 1.1742, "step": 850 }, { "epoch": 0.07, "learning_rate": 1.9907588812583514e-05, "loss": 1.1556, "step": 851 }, { "epoch": 0.07, "learning_rate": 1.9907216389074974e-05, "loss": 1.1791, "step": 852 }, { "epoch": 0.07, "learning_rate": 1.9906843220126134e-05, "loss": 1.1609, "step": 853 }, { "epoch": 0.07, "learning_rate": 1.990646930576507e-05, "loss": 1.1554, "step": 854 }, { "epoch": 0.07, "learning_rate": 1.990609464601991e-05, "loss": 1.1734, "step": 855 }, { "epoch": 0.07, "learning_rate": 1.990571924091885e-05, "loss": 1.1745, "step": 856 }, { "epoch": 0.07, "learning_rate": 1.9905343090490137e-05, "loss": 1.1223, "step": 857 }, { "epoch": 0.07, "learning_rate": 1.990496619476207e-05, "loss": 1.1775, "step": 858 }, { "epoch": 0.07, "learning_rate": 1.990458855376301e-05, "loss": 1.1626, "step": 859 }, { "epoch": 0.07, "learning_rate": 1.990421016752137e-05, "loss": 1.1823, "step": 860 }, { "epoch": 0.07, "learning_rate": 1.990383103606562e-05, "loss": 1.1703, "step": 861 }, { "epoch": 0.07, "learning_rate": 1.990345115942429e-05, "loss": 1.1855, "step": 862 }, { "epoch": 0.07, "learning_rate": 1.9903070537625958e-05, "loss": 1.2064, "step": 863 }, { "epoch": 0.07, "learning_rate": 1.990268917069926e-05, "loss": 1.1638, "step": 864 }, { "epoch": 0.07, "learning_rate": 1.9902307058672905e-05, "loss": 1.1742, "step": 865 }, { "epoch": 0.07, "learning_rate": 1.9901924201575633e-05, "loss": 1.1779, "step": 866 }, { "epoch": 0.07, "learning_rate": 1.9901540599436254e-05, "loss": 1.1587, "step": 867 }, { "epoch": 0.07, "learning_rate": 1.9901156252283625e-05, "loss": 1.2096, "step": 868 }, { "epoch": 0.07, "learning_rate": 1.990077116014667e-05, "loss": 1.222, "step": 869 }, { "epoch": 0.07, "learning_rate": 1.990038532305437e-05, "loss": 1.1526, "step": 870 }, { "epoch": 0.07, "learning_rate": 1.989999874103575e-05, "loss": 1.1738, "step": 871 }, { "epoch": 0.07, "learning_rate": 1.9899611414119895e-05, "loss": 1.1693, "step": 872 }, { "epoch": 0.07, "learning_rate": 1.9899223342335952e-05, "loss": 1.1535, "step": 873 }, { "epoch": 0.07, "learning_rate": 1.989883452571312e-05, "loss": 1.2083, "step": 874 }, { "epoch": 0.07, "learning_rate": 1.989844496428065e-05, "loss": 1.1045, "step": 875 }, { "epoch": 0.07, "learning_rate": 1.989805465806786e-05, "loss": 1.1726, "step": 876 }, { "epoch": 0.07, "learning_rate": 1.9897663607104115e-05, "loss": 1.1793, "step": 877 }, { "epoch": 0.07, "learning_rate": 1.9897271811418838e-05, "loss": 1.1746, "step": 878 }, { "epoch": 0.07, "learning_rate": 1.989687927104151e-05, "loss": 1.1903, "step": 879 }, { "epoch": 0.07, "learning_rate": 1.9896485986001663e-05, "loss": 1.1823, "step": 880 }, { "epoch": 0.07, "learning_rate": 1.9896091956328894e-05, "loss": 1.1827, "step": 881 }, { "epoch": 0.07, "learning_rate": 1.9895697182052844e-05, "loss": 1.1832, "step": 882 }, { "epoch": 0.07, "learning_rate": 1.9895301663203225e-05, "loss": 1.2342, "step": 883 }, { "epoch": 0.07, "learning_rate": 1.989490539980979e-05, "loss": 1.1878, "step": 884 }, { "epoch": 0.07, "learning_rate": 1.9894508391902354e-05, "loss": 1.1925, "step": 885 }, { "epoch": 0.08, "learning_rate": 1.9894110639510794e-05, "loss": 1.1959, "step": 886 }, { "epoch": 0.08, "learning_rate": 1.9893712142665033e-05, "loss": 1.1867, "step": 887 }, { "epoch": 0.08, "learning_rate": 1.989331290139506e-05, "loss": 1.205, "step": 888 }, { "epoch": 0.08, "learning_rate": 1.989291291573091e-05, "loss": 1.2292, "step": 889 }, { "epoch": 0.08, "learning_rate": 1.9892512185702683e-05, "loss": 1.1728, "step": 890 }, { "epoch": 0.08, "learning_rate": 1.9892110711340525e-05, "loss": 1.165, "step": 891 }, { "epoch": 0.08, "learning_rate": 1.9891708492674647e-05, "loss": 1.1577, "step": 892 }, { "epoch": 0.08, "learning_rate": 1.9891305529735314e-05, "loss": 1.2565, "step": 893 }, { "epoch": 0.08, "learning_rate": 1.9890901822552847e-05, "loss": 1.1523, "step": 894 }, { "epoch": 0.08, "learning_rate": 1.9890497371157617e-05, "loss": 1.1499, "step": 895 }, { "epoch": 0.08, "learning_rate": 1.989009217558006e-05, "loss": 1.1212, "step": 896 }, { "epoch": 0.08, "learning_rate": 1.988968623585066e-05, "loss": 1.1495, "step": 897 }, { "epoch": 0.08, "learning_rate": 1.9889279551999966e-05, "loss": 1.1521, "step": 898 }, { "epoch": 0.08, "learning_rate": 1.9888872124058572e-05, "loss": 1.1759, "step": 899 }, { "epoch": 0.08, "learning_rate": 1.9888463952057135e-05, "loss": 1.185, "step": 900 }, { "epoch": 0.08, "learning_rate": 1.988805503602637e-05, "loss": 1.211, "step": 901 }, { "epoch": 0.08, "learning_rate": 1.9887645375997043e-05, "loss": 1.1283, "step": 902 }, { "epoch": 0.08, "learning_rate": 1.9887234971999978e-05, "loss": 1.0821, "step": 903 }, { "epoch": 0.08, "learning_rate": 1.9886823824066056e-05, "loss": 1.1487, "step": 904 }, { "epoch": 0.08, "learning_rate": 1.9886411932226206e-05, "loss": 1.1542, "step": 905 }, { "epoch": 0.08, "learning_rate": 1.9885999296511428e-05, "loss": 0.9365, "step": 906 }, { "epoch": 0.08, "learning_rate": 1.9885585916952766e-05, "loss": 1.1224, "step": 907 }, { "epoch": 0.08, "learning_rate": 1.9885171793581324e-05, "loss": 1.2047, "step": 908 }, { "epoch": 0.08, "learning_rate": 1.988475692642826e-05, "loss": 1.1794, "step": 909 }, { "epoch": 0.08, "learning_rate": 1.988434131552479e-05, "loss": 1.1744, "step": 910 }, { "epoch": 0.08, "learning_rate": 1.9883924960902185e-05, "loss": 1.2017, "step": 911 }, { "epoch": 0.08, "learning_rate": 1.9883507862591776e-05, "loss": 0.9506, "step": 912 }, { "epoch": 0.08, "learning_rate": 1.9883090020624946e-05, "loss": 1.2081, "step": 913 }, { "epoch": 0.08, "learning_rate": 1.988267143503313e-05, "loss": 1.2124, "step": 914 }, { "epoch": 0.08, "learning_rate": 1.9882252105847825e-05, "loss": 1.1698, "step": 915 }, { "epoch": 0.08, "learning_rate": 1.9881832033100585e-05, "loss": 1.1445, "step": 916 }, { "epoch": 0.08, "learning_rate": 1.9881411216823012e-05, "loss": 1.1445, "step": 917 }, { "epoch": 0.08, "learning_rate": 1.9880989657046776e-05, "loss": 0.8994, "step": 918 }, { "epoch": 0.08, "learning_rate": 1.988056735380359e-05, "loss": 1.1471, "step": 919 }, { "epoch": 0.08, "learning_rate": 1.988014430712523e-05, "loss": 1.142, "step": 920 }, { "epoch": 0.08, "learning_rate": 1.9879720517043528e-05, "loss": 1.1624, "step": 921 }, { "epoch": 0.08, "learning_rate": 1.9879295983590373e-05, "loss": 1.1526, "step": 922 }, { "epoch": 0.08, "learning_rate": 1.9878870706797703e-05, "loss": 1.091, "step": 923 }, { "epoch": 0.08, "learning_rate": 1.987844468669752e-05, "loss": 1.1463, "step": 924 }, { "epoch": 0.08, "learning_rate": 1.9878017923321883e-05, "loss": 1.1938, "step": 925 }, { "epoch": 0.08, "learning_rate": 1.9877590416702893e-05, "loss": 1.1212, "step": 926 }, { "epoch": 0.08, "learning_rate": 1.9877162166872727e-05, "loss": 1.1846, "step": 927 }, { "epoch": 0.08, "learning_rate": 1.9876733173863595e-05, "loss": 1.1868, "step": 928 }, { "epoch": 0.08, "learning_rate": 1.9876303437707785e-05, "loss": 1.118, "step": 929 }, { "epoch": 0.08, "learning_rate": 1.9875872958437628e-05, "loss": 1.1823, "step": 930 }, { "epoch": 0.08, "learning_rate": 1.9875441736085514e-05, "loss": 1.1322, "step": 931 }, { "epoch": 0.08, "learning_rate": 1.9875009770683894e-05, "loss": 1.1862, "step": 932 }, { "epoch": 0.08, "learning_rate": 1.987457706226526e-05, "loss": 1.0967, "step": 933 }, { "epoch": 0.08, "learning_rate": 1.987414361086218e-05, "loss": 1.1216, "step": 934 }, { "epoch": 0.08, "learning_rate": 1.987370941650726e-05, "loss": 1.1569, "step": 935 }, { "epoch": 0.08, "learning_rate": 1.9873274479233177e-05, "loss": 1.2262, "step": 936 }, { "epoch": 0.08, "learning_rate": 1.9872838799072652e-05, "loss": 1.1656, "step": 937 }, { "epoch": 0.08, "learning_rate": 1.9872402376058463e-05, "loss": 1.1897, "step": 938 }, { "epoch": 0.08, "learning_rate": 1.9871965210223454e-05, "loss": 1.1509, "step": 939 }, { "epoch": 0.08, "learning_rate": 1.9871527301600516e-05, "loss": 1.1896, "step": 940 }, { "epoch": 0.08, "learning_rate": 1.98710886502226e-05, "loss": 1.1724, "step": 941 }, { "epoch": 0.08, "learning_rate": 1.987064925612271e-05, "loss": 1.1555, "step": 942 }, { "epoch": 0.08, "learning_rate": 1.9870209119333904e-05, "loss": 1.1618, "step": 943 }, { "epoch": 0.08, "learning_rate": 1.98697682398893e-05, "loss": 1.1662, "step": 944 }, { "epoch": 0.08, "learning_rate": 1.9869326617822075e-05, "loss": 1.1731, "step": 945 }, { "epoch": 0.08, "learning_rate": 1.986888425316545e-05, "loss": 1.2264, "step": 946 }, { "epoch": 0.08, "learning_rate": 1.9868441145952718e-05, "loss": 1.1929, "step": 947 }, { "epoch": 0.08, "learning_rate": 1.9867997296217215e-05, "loss": 1.1702, "step": 948 }, { "epoch": 0.08, "learning_rate": 1.9867552703992333e-05, "loss": 1.1445, "step": 949 }, { "epoch": 0.08, "learning_rate": 1.9867107369311536e-05, "loss": 1.1784, "step": 950 }, { "epoch": 0.08, "learning_rate": 1.986666129220832e-05, "loss": 1.1354, "step": 951 }, { "epoch": 0.08, "learning_rate": 1.986621447271625e-05, "loss": 1.192, "step": 952 }, { "epoch": 0.08, "learning_rate": 1.9865766910868954e-05, "loss": 1.1179, "step": 953 }, { "epoch": 0.08, "learning_rate": 1.9865318606700098e-05, "loss": 1.121, "step": 954 }, { "epoch": 0.08, "learning_rate": 1.9864869560243423e-05, "loss": 1.1964, "step": 955 }, { "epoch": 0.08, "learning_rate": 1.9864419771532707e-05, "loss": 1.181, "step": 956 }, { "epoch": 0.08, "learning_rate": 1.9863969240601803e-05, "loss": 1.1781, "step": 957 }, { "epoch": 0.08, "learning_rate": 1.9863517967484597e-05, "loss": 1.1616, "step": 958 }, { "epoch": 0.08, "learning_rate": 1.9863065952215053e-05, "loss": 1.1749, "step": 959 }, { "epoch": 0.08, "learning_rate": 1.986261319482718e-05, "loss": 1.1392, "step": 960 }, { "epoch": 0.08, "learning_rate": 1.9862159695355044e-05, "loss": 1.1026, "step": 961 }, { "epoch": 0.08, "learning_rate": 1.9861705453832766e-05, "loss": 1.1493, "step": 962 }, { "epoch": 0.08, "learning_rate": 1.9861250470294524e-05, "loss": 1.1642, "step": 963 }, { "epoch": 0.08, "learning_rate": 1.9860794744774557e-05, "loss": 1.103, "step": 964 }, { "epoch": 0.08, "learning_rate": 1.9860338277307146e-05, "loss": 1.1721, "step": 965 }, { "epoch": 0.08, "learning_rate": 1.9859881067926646e-05, "loss": 1.1316, "step": 966 }, { "epoch": 0.08, "learning_rate": 1.985942311666745e-05, "loss": 1.1726, "step": 967 }, { "epoch": 0.08, "learning_rate": 1.985896442356402e-05, "loss": 1.113, "step": 968 }, { "epoch": 0.08, "learning_rate": 1.9858504988650873e-05, "loss": 1.1775, "step": 969 }, { "epoch": 0.08, "learning_rate": 1.985804481196257e-05, "loss": 1.196, "step": 970 }, { "epoch": 0.08, "learning_rate": 1.985758389353374e-05, "loss": 1.1703, "step": 971 }, { "epoch": 0.08, "learning_rate": 1.9857122233399058e-05, "loss": 1.1594, "step": 972 }, { "epoch": 0.08, "learning_rate": 1.9856659831593268e-05, "loss": 1.2042, "step": 973 }, { "epoch": 0.08, "learning_rate": 1.9856196688151156e-05, "loss": 1.1487, "step": 974 }, { "epoch": 0.08, "learning_rate": 1.9855732803107575e-05, "loss": 1.2023, "step": 975 }, { "epoch": 0.08, "learning_rate": 1.9855268176497426e-05, "loss": 1.2286, "step": 976 }, { "epoch": 0.08, "learning_rate": 1.985480280835567e-05, "loss": 1.1567, "step": 977 }, { "epoch": 0.08, "learning_rate": 1.985433669871732e-05, "loss": 0.9237, "step": 978 }, { "epoch": 0.08, "learning_rate": 1.9853869847617446e-05, "loss": 1.1723, "step": 979 }, { "epoch": 0.08, "learning_rate": 1.985340225509118e-05, "loss": 1.2031, "step": 980 }, { "epoch": 0.08, "learning_rate": 1.9852933921173702e-05, "loss": 1.2488, "step": 981 }, { "epoch": 0.08, "learning_rate": 1.985246484590025e-05, "loss": 1.1974, "step": 982 }, { "epoch": 0.08, "learning_rate": 1.9851995029306115e-05, "loss": 1.1473, "step": 983 }, { "epoch": 0.08, "learning_rate": 1.985152447142665e-05, "loss": 1.1576, "step": 984 }, { "epoch": 0.08, "learning_rate": 1.9851053172297264e-05, "loss": 1.1657, "step": 985 }, { "epoch": 0.08, "learning_rate": 1.9850581131953416e-05, "loss": 1.2074, "step": 986 }, { "epoch": 0.08, "learning_rate": 1.9850108350430626e-05, "loss": 1.1132, "step": 987 }, { "epoch": 0.08, "learning_rate": 1.9849634827764462e-05, "loss": 1.0701, "step": 988 }, { "epoch": 0.08, "learning_rate": 1.9849160563990552e-05, "loss": 1.157, "step": 989 }, { "epoch": 0.08, "learning_rate": 1.9848685559144586e-05, "loss": 1.1166, "step": 990 }, { "epoch": 0.08, "learning_rate": 1.98482098132623e-05, "loss": 1.2039, "step": 991 }, { "epoch": 0.08, "learning_rate": 1.9847733326379498e-05, "loss": 1.2265, "step": 992 }, { "epoch": 0.08, "learning_rate": 1.984725609853202e-05, "loss": 1.1359, "step": 993 }, { "epoch": 0.08, "learning_rate": 1.9846778129755782e-05, "loss": 1.0944, "step": 994 }, { "epoch": 0.08, "learning_rate": 1.9846299420086746e-05, "loss": 1.1829, "step": 995 }, { "epoch": 0.08, "learning_rate": 1.9845819969560933e-05, "loss": 1.2353, "step": 996 }, { "epoch": 0.08, "learning_rate": 1.984533977821441e-05, "loss": 1.1809, "step": 997 }, { "epoch": 0.08, "learning_rate": 1.9844858846083314e-05, "loss": 1.2415, "step": 998 }, { "epoch": 0.08, "learning_rate": 1.9844377173203834e-05, "loss": 1.1902, "step": 999 }, { "epoch": 0.08, "learning_rate": 1.9843894759612207e-05, "loss": 1.1977, "step": 1000 }, { "epoch": 0.08, "learning_rate": 1.984341160534473e-05, "loss": 1.1936, "step": 1001 }, { "epoch": 0.08, "learning_rate": 1.9842927710437758e-05, "loss": 1.1873, "step": 1002 }, { "epoch": 0.08, "learning_rate": 1.9842443074927704e-05, "loss": 1.1802, "step": 1003 }, { "epoch": 0.09, "learning_rate": 1.9841957698851028e-05, "loss": 1.145, "step": 1004 }, { "epoch": 0.09, "learning_rate": 1.9841471582244255e-05, "loss": 1.1494, "step": 1005 }, { "epoch": 0.09, "learning_rate": 1.9840984725143956e-05, "loss": 1.1068, "step": 1006 }, { "epoch": 0.09, "learning_rate": 1.984049712758677e-05, "loss": 1.1944, "step": 1007 }, { "epoch": 0.09, "learning_rate": 1.984000878960938e-05, "loss": 1.2072, "step": 1008 }, { "epoch": 0.09, "learning_rate": 1.983951971124853e-05, "loss": 1.1043, "step": 1009 }, { "epoch": 0.09, "learning_rate": 1.983902989254102e-05, "loss": 1.1731, "step": 1010 }, { "epoch": 0.09, "learning_rate": 1.9838539333523705e-05, "loss": 0.9526, "step": 1011 }, { "epoch": 0.09, "learning_rate": 1.98380480342335e-05, "loss": 1.1952, "step": 1012 }, { "epoch": 0.09, "learning_rate": 1.9837555994707363e-05, "loss": 1.1345, "step": 1013 }, { "epoch": 0.09, "learning_rate": 1.983706321498232e-05, "loss": 1.1881, "step": 1014 }, { "epoch": 0.09, "learning_rate": 1.983656969509545e-05, "loss": 1.2112, "step": 1015 }, { "epoch": 0.09, "learning_rate": 1.9836075435083887e-05, "loss": 1.2449, "step": 1016 }, { "epoch": 0.09, "learning_rate": 1.983558043498482e-05, "loss": 1.1374, "step": 1017 }, { "epoch": 0.09, "learning_rate": 1.9835084694835494e-05, "loss": 1.2028, "step": 1018 }, { "epoch": 0.09, "learning_rate": 1.9834588214673208e-05, "loss": 1.0858, "step": 1019 }, { "epoch": 0.09, "learning_rate": 1.9834090994535315e-05, "loss": 1.134, "step": 1020 }, { "epoch": 0.09, "learning_rate": 1.983359303445923e-05, "loss": 1.1983, "step": 1021 }, { "epoch": 0.09, "learning_rate": 1.9833094334482427e-05, "loss": 1.138, "step": 1022 }, { "epoch": 0.09, "learning_rate": 1.983259489464242e-05, "loss": 1.1534, "step": 1023 }, { "epoch": 0.09, "learning_rate": 1.9832094714976794e-05, "loss": 1.1911, "step": 1024 }, { "epoch": 0.09, "learning_rate": 1.983159379552318e-05, "loss": 1.1382, "step": 1025 }, { "epoch": 0.09, "learning_rate": 1.9831092136319265e-05, "loss": 1.1863, "step": 1026 }, { "epoch": 0.09, "learning_rate": 1.9830589737402804e-05, "loss": 1.0917, "step": 1027 }, { "epoch": 0.09, "learning_rate": 1.9830086598811592e-05, "loss": 1.1164, "step": 1028 }, { "epoch": 0.09, "learning_rate": 1.9829582720583486e-05, "loss": 1.1851, "step": 1029 }, { "epoch": 0.09, "learning_rate": 1.9829078102756402e-05, "loss": 1.2022, "step": 1030 }, { "epoch": 0.09, "learning_rate": 1.9828572745368308e-05, "loss": 1.1402, "step": 1031 }, { "epoch": 0.09, "learning_rate": 1.9828066648457228e-05, "loss": 1.1437, "step": 1032 }, { "epoch": 0.09, "learning_rate": 1.982755981206124e-05, "loss": 1.1848, "step": 1033 }, { "epoch": 0.09, "learning_rate": 1.9827052236218482e-05, "loss": 1.1415, "step": 1034 }, { "epoch": 0.09, "learning_rate": 1.9826543920967145e-05, "loss": 1.1635, "step": 1035 }, { "epoch": 0.09, "learning_rate": 1.982603486634547e-05, "loss": 1.2124, "step": 1036 }, { "epoch": 0.09, "learning_rate": 1.9825525072391773e-05, "loss": 1.1249, "step": 1037 }, { "epoch": 0.09, "learning_rate": 1.9825014539144395e-05, "loss": 1.1794, "step": 1038 }, { "epoch": 0.09, "learning_rate": 1.9824503266641758e-05, "loss": 1.1814, "step": 1039 }, { "epoch": 0.09, "learning_rate": 1.9823991254922335e-05, "loss": 1.1423, "step": 1040 }, { "epoch": 0.09, "learning_rate": 1.9823478504024645e-05, "loss": 1.177, "step": 1041 }, { "epoch": 0.09, "learning_rate": 1.982296501398727e-05, "loss": 1.2315, "step": 1042 }, { "epoch": 0.09, "learning_rate": 1.9822450784848845e-05, "loss": 1.1682, "step": 1043 }, { "epoch": 0.09, "learning_rate": 1.9821935816648066e-05, "loss": 1.1629, "step": 1044 }, { "epoch": 0.09, "learning_rate": 1.9821420109423675e-05, "loss": 1.1604, "step": 1045 }, { "epoch": 0.09, "learning_rate": 1.982090366321448e-05, "loss": 1.1395, "step": 1046 }, { "epoch": 0.09, "learning_rate": 1.9820386478059333e-05, "loss": 1.1944, "step": 1047 }, { "epoch": 0.09, "learning_rate": 1.981986855399715e-05, "loss": 1.129, "step": 1048 }, { "epoch": 0.09, "learning_rate": 1.981934989106691e-05, "loss": 1.1697, "step": 1049 }, { "epoch": 0.09, "learning_rate": 1.9818830489307626e-05, "loss": 1.1315, "step": 1050 }, { "epoch": 0.09, "learning_rate": 1.9818310348758386e-05, "loss": 1.1265, "step": 1051 }, { "epoch": 0.09, "learning_rate": 1.9817789469458323e-05, "loss": 1.2235, "step": 1052 }, { "epoch": 0.09, "learning_rate": 1.9817267851446633e-05, "loss": 1.1602, "step": 1053 }, { "epoch": 0.09, "learning_rate": 1.981674549476256e-05, "loss": 1.1445, "step": 1054 }, { "epoch": 0.09, "learning_rate": 1.9816222399445405e-05, "loss": 1.1575, "step": 1055 }, { "epoch": 0.09, "learning_rate": 1.9815698565534536e-05, "loss": 1.1677, "step": 1056 }, { "epoch": 0.09, "learning_rate": 1.9815173993069358e-05, "loss": 1.1404, "step": 1057 }, { "epoch": 0.09, "learning_rate": 1.9814648682089344e-05, "loss": 1.1523, "step": 1058 }, { "epoch": 0.09, "learning_rate": 1.981412263263402e-05, "loss": 1.1642, "step": 1059 }, { "epoch": 0.09, "learning_rate": 1.981359584474297e-05, "loss": 1.1601, "step": 1060 }, { "epoch": 0.09, "learning_rate": 1.9813068318455824e-05, "loss": 0.9386, "step": 1061 }, { "epoch": 0.09, "learning_rate": 1.981254005381228e-05, "loss": 1.2126, "step": 1062 }, { "epoch": 0.09, "learning_rate": 1.9812011050852083e-05, "loss": 1.1894, "step": 1063 }, { "epoch": 0.09, "learning_rate": 1.981148130961504e-05, "loss": 1.1446, "step": 1064 }, { "epoch": 0.09, "learning_rate": 1.9810950830141003e-05, "loss": 1.1765, "step": 1065 }, { "epoch": 0.09, "learning_rate": 1.9810419612469895e-05, "loss": 1.1754, "step": 1066 }, { "epoch": 0.09, "learning_rate": 1.9809887656641678e-05, "loss": 1.1973, "step": 1067 }, { "epoch": 0.09, "learning_rate": 1.9809354962696382e-05, "loss": 1.1416, "step": 1068 }, { "epoch": 0.09, "learning_rate": 1.9808821530674087e-05, "loss": 1.1579, "step": 1069 }, { "epoch": 0.09, "learning_rate": 1.980828736061493e-05, "loss": 1.1501, "step": 1070 }, { "epoch": 0.09, "learning_rate": 1.9807752452559098e-05, "loss": 1.2263, "step": 1071 }, { "epoch": 0.09, "learning_rate": 1.980721680654685e-05, "loss": 1.2026, "step": 1072 }, { "epoch": 0.09, "learning_rate": 1.9806680422618476e-05, "loss": 1.1477, "step": 1073 }, { "epoch": 0.09, "learning_rate": 1.9806143300814343e-05, "loss": 1.1629, "step": 1074 }, { "epoch": 0.09, "learning_rate": 1.980560544117487e-05, "loss": 1.1437, "step": 1075 }, { "epoch": 0.09, "learning_rate": 1.980506684374051e-05, "loss": 1.178, "step": 1076 }, { "epoch": 0.09, "learning_rate": 1.9804527508551806e-05, "loss": 1.1523, "step": 1077 }, { "epoch": 0.09, "learning_rate": 1.9803987435649326e-05, "loss": 1.1365, "step": 1078 }, { "epoch": 0.09, "learning_rate": 1.9803446625073713e-05, "loss": 1.1417, "step": 1079 }, { "epoch": 0.09, "learning_rate": 1.9802905076865658e-05, "loss": 1.2588, "step": 1080 }, { "epoch": 0.09, "learning_rate": 1.9802362791065904e-05, "loss": 1.1463, "step": 1081 }, { "epoch": 0.09, "learning_rate": 1.980181976771526e-05, "loss": 1.15, "step": 1082 }, { "epoch": 0.09, "learning_rate": 1.980127600685458e-05, "loss": 1.1993, "step": 1083 }, { "epoch": 0.09, "learning_rate": 1.980073150852478e-05, "loss": 1.13, "step": 1084 }, { "epoch": 0.09, "learning_rate": 1.9800186272766827e-05, "loss": 1.1473, "step": 1085 }, { "epoch": 0.09, "learning_rate": 1.979964029962175e-05, "loss": 1.2337, "step": 1086 }, { "epoch": 0.09, "learning_rate": 1.979909358913062e-05, "loss": 1.1156, "step": 1087 }, { "epoch": 0.09, "learning_rate": 1.9798546141334585e-05, "loss": 1.2058, "step": 1088 }, { "epoch": 0.09, "learning_rate": 1.9797997956274825e-05, "loss": 1.2084, "step": 1089 }, { "epoch": 0.09, "learning_rate": 1.9797449033992596e-05, "loss": 0.9811, "step": 1090 }, { "epoch": 0.09, "learning_rate": 1.979689937452919e-05, "loss": 1.1462, "step": 1091 }, { "epoch": 0.09, "learning_rate": 1.9796348977925975e-05, "loss": 1.1606, "step": 1092 }, { "epoch": 0.09, "learning_rate": 1.979579784422436e-05, "loss": 1.14, "step": 1093 }, { "epoch": 0.09, "learning_rate": 1.9795245973465808e-05, "loss": 1.1769, "step": 1094 }, { "epoch": 0.09, "learning_rate": 1.979469336569185e-05, "loss": 1.2029, "step": 1095 }, { "epoch": 0.09, "learning_rate": 1.9794140020944065e-05, "loss": 1.1136, "step": 1096 }, { "epoch": 0.09, "learning_rate": 1.9793585939264087e-05, "loss": 1.1377, "step": 1097 }, { "epoch": 0.09, "learning_rate": 1.97930311206936e-05, "loss": 1.1681, "step": 1098 }, { "epoch": 0.09, "learning_rate": 1.9792475565274358e-05, "loss": 0.894, "step": 1099 }, { "epoch": 0.09, "learning_rate": 1.979191927304816e-05, "loss": 1.1425, "step": 1100 }, { "epoch": 0.09, "learning_rate": 1.9791362244056858e-05, "loss": 0.9423, "step": 1101 }, { "epoch": 0.09, "learning_rate": 1.979080447834237e-05, "loss": 1.1736, "step": 1102 }, { "epoch": 0.09, "learning_rate": 1.979024597594666e-05, "loss": 1.2008, "step": 1103 }, { "epoch": 0.09, "learning_rate": 1.9789686736911755e-05, "loss": 1.2217, "step": 1104 }, { "epoch": 0.09, "learning_rate": 1.9789126761279733e-05, "loss": 1.101, "step": 1105 }, { "epoch": 0.09, "learning_rate": 1.9788566049092718e-05, "loss": 1.1599, "step": 1106 }, { "epoch": 0.09, "learning_rate": 1.978800460039291e-05, "loss": 1.2067, "step": 1107 }, { "epoch": 0.09, "learning_rate": 1.9787442415222552e-05, "loss": 1.1364, "step": 1108 }, { "epoch": 0.09, "learning_rate": 1.978687949362394e-05, "loss": 1.1533, "step": 1109 }, { "epoch": 0.09, "learning_rate": 1.978631583563943e-05, "loss": 1.1537, "step": 1110 }, { "epoch": 0.09, "learning_rate": 1.978575144131144e-05, "loss": 1.1497, "step": 1111 }, { "epoch": 0.09, "learning_rate": 1.9785186310682425e-05, "loss": 1.1421, "step": 1112 }, { "epoch": 0.09, "learning_rate": 1.9784620443794915e-05, "loss": 1.1667, "step": 1113 }, { "epoch": 0.09, "learning_rate": 1.9784053840691487e-05, "loss": 1.1302, "step": 1114 }, { "epoch": 0.09, "learning_rate": 1.9783486501414766e-05, "loss": 1.2074, "step": 1115 }, { "epoch": 0.09, "learning_rate": 1.9782918426007444e-05, "loss": 1.2014, "step": 1116 }, { "epoch": 0.09, "learning_rate": 1.978234961451227e-05, "loss": 1.1466, "step": 1117 }, { "epoch": 0.09, "learning_rate": 1.9781780066972035e-05, "loss": 1.1144, "step": 1118 }, { "epoch": 0.09, "learning_rate": 1.9781209783429597e-05, "loss": 1.177, "step": 1119 }, { "epoch": 0.09, "learning_rate": 1.978063876392786e-05, "loss": 1.1769, "step": 1120 }, { "epoch": 0.09, "learning_rate": 1.9780067008509795e-05, "loss": 1.1492, "step": 1121 }, { "epoch": 0.1, "learning_rate": 1.977949451721842e-05, "loss": 1.1577, "step": 1122 }, { "epoch": 0.1, "learning_rate": 1.977892129009681e-05, "loss": 1.2183, "step": 1123 }, { "epoch": 0.1, "learning_rate": 1.9778347327188094e-05, "loss": 1.1776, "step": 1124 }, { "epoch": 0.1, "learning_rate": 1.9777772628535465e-05, "loss": 1.1308, "step": 1125 }, { "epoch": 0.1, "learning_rate": 1.9777197194182157e-05, "loss": 1.1811, "step": 1126 }, { "epoch": 0.1, "learning_rate": 1.9776621024171468e-05, "loss": 1.1339, "step": 1127 }, { "epoch": 0.1, "learning_rate": 1.9776044118546752e-05, "loss": 1.1793, "step": 1128 }, { "epoch": 0.1, "learning_rate": 1.977546647735142e-05, "loss": 1.1851, "step": 1129 }, { "epoch": 0.1, "learning_rate": 1.977488810062893e-05, "loss": 0.8926, "step": 1130 }, { "epoch": 0.1, "learning_rate": 1.97743089884228e-05, "loss": 1.1824, "step": 1131 }, { "epoch": 0.1, "learning_rate": 1.9773729140776608e-05, "loss": 1.165, "step": 1132 }, { "epoch": 0.1, "learning_rate": 1.9773148557733977e-05, "loss": 1.188, "step": 1133 }, { "epoch": 0.1, "learning_rate": 1.97725672393386e-05, "loss": 1.1669, "step": 1134 }, { "epoch": 0.1, "learning_rate": 1.9771985185634208e-05, "loss": 1.1449, "step": 1135 }, { "epoch": 0.1, "learning_rate": 1.97714023966646e-05, "loss": 1.1155, "step": 1136 }, { "epoch": 0.1, "learning_rate": 1.9770818872473626e-05, "loss": 1.1938, "step": 1137 }, { "epoch": 0.1, "learning_rate": 1.9770234613105194e-05, "loss": 1.1435, "step": 1138 }, { "epoch": 0.1, "learning_rate": 1.9769649618603257e-05, "loss": 1.1773, "step": 1139 }, { "epoch": 0.1, "learning_rate": 1.9769063889011838e-05, "loss": 1.207, "step": 1140 }, { "epoch": 0.1, "learning_rate": 1.976847742437501e-05, "loss": 1.156, "step": 1141 }, { "epoch": 0.1, "learning_rate": 1.9767890224736893e-05, "loss": 1.1235, "step": 1142 }, { "epoch": 0.1, "learning_rate": 1.9767302290141678e-05, "loss": 1.1493, "step": 1143 }, { "epoch": 0.1, "learning_rate": 1.9766713620633594e-05, "loss": 1.1536, "step": 1144 }, { "epoch": 0.1, "learning_rate": 1.976612421625694e-05, "loss": 1.2024, "step": 1145 }, { "epoch": 0.1, "learning_rate": 1.976553407705606e-05, "loss": 1.189, "step": 1146 }, { "epoch": 0.1, "learning_rate": 1.976494320307536e-05, "loss": 1.1543, "step": 1147 }, { "epoch": 0.1, "learning_rate": 1.9764351594359295e-05, "loss": 1.1659, "step": 1148 }, { "epoch": 0.1, "learning_rate": 1.976375925095238e-05, "loss": 1.1462, "step": 1149 }, { "epoch": 0.1, "learning_rate": 1.9763166172899185e-05, "loss": 1.089, "step": 1150 }, { "epoch": 0.1, "learning_rate": 1.976257236024434e-05, "loss": 1.1052, "step": 1151 }, { "epoch": 0.1, "learning_rate": 1.9761977813032518e-05, "loss": 1.1851, "step": 1152 }, { "epoch": 0.1, "learning_rate": 1.9761382531308454e-05, "loss": 1.1922, "step": 1153 }, { "epoch": 0.1, "learning_rate": 1.976078651511694e-05, "loss": 1.156, "step": 1154 }, { "epoch": 0.1, "learning_rate": 1.9760189764502825e-05, "loss": 1.1474, "step": 1155 }, { "epoch": 0.1, "learning_rate": 1.9759592279511004e-05, "loss": 1.1239, "step": 1156 }, { "epoch": 0.1, "learning_rate": 1.9758994060186433e-05, "loss": 1.2468, "step": 1157 }, { "epoch": 0.1, "learning_rate": 1.975839510657413e-05, "loss": 1.2235, "step": 1158 }, { "epoch": 0.1, "learning_rate": 1.9757795418719156e-05, "loss": 1.0995, "step": 1159 }, { "epoch": 0.1, "learning_rate": 1.975719499666663e-05, "loss": 0.9943, "step": 1160 }, { "epoch": 0.1, "learning_rate": 1.975659384046174e-05, "loss": 1.1828, "step": 1161 }, { "epoch": 0.1, "learning_rate": 1.9755991950149706e-05, "loss": 1.1645, "step": 1162 }, { "epoch": 0.1, "learning_rate": 1.975538932577582e-05, "loss": 1.1917, "step": 1163 }, { "epoch": 0.1, "learning_rate": 1.975478596738543e-05, "loss": 1.1632, "step": 1164 }, { "epoch": 0.1, "learning_rate": 1.9754181875023933e-05, "loss": 1.1744, "step": 1165 }, { "epoch": 0.1, "learning_rate": 1.9753577048736772e-05, "loss": 1.1761, "step": 1166 }, { "epoch": 0.1, "learning_rate": 1.9752971488569464e-05, "loss": 1.1576, "step": 1167 }, { "epoch": 0.1, "learning_rate": 1.975236519456757e-05, "loss": 1.1461, "step": 1168 }, { "epoch": 0.1, "learning_rate": 1.9751758166776713e-05, "loss": 1.1554, "step": 1169 }, { "epoch": 0.1, "learning_rate": 1.9751150405242563e-05, "loss": 1.1522, "step": 1170 }, { "epoch": 0.1, "learning_rate": 1.975054191001085e-05, "loss": 1.3969, "step": 1171 }, { "epoch": 0.1, "learning_rate": 1.9749932681127358e-05, "loss": 1.1689, "step": 1172 }, { "epoch": 0.1, "learning_rate": 1.9749322718637925e-05, "loss": 1.174, "step": 1173 }, { "epoch": 0.1, "learning_rate": 1.9748712022588453e-05, "loss": 1.1483, "step": 1174 }, { "epoch": 0.1, "learning_rate": 1.9748100593024885e-05, "loss": 1.1574, "step": 1175 }, { "epoch": 0.1, "learning_rate": 1.974748842999323e-05, "loss": 1.1066, "step": 1176 }, { "epoch": 0.1, "learning_rate": 1.9746875533539542e-05, "loss": 1.1774, "step": 1177 }, { "epoch": 0.1, "learning_rate": 1.974626190370995e-05, "loss": 1.1574, "step": 1178 }, { "epoch": 0.1, "learning_rate": 1.974564754055061e-05, "loss": 1.6436, "step": 1179 }, { "epoch": 0.1, "learning_rate": 1.9745032444107756e-05, "loss": 1.1637, "step": 1180 }, { "epoch": 0.1, "learning_rate": 1.9744416614427672e-05, "loss": 1.1589, "step": 1181 }, { "epoch": 0.1, "learning_rate": 1.9743800051556685e-05, "loss": 1.2041, "step": 1182 }, { "epoch": 0.1, "learning_rate": 1.9743182755541193e-05, "loss": 1.2534, "step": 1183 }, { "epoch": 0.1, "learning_rate": 1.974256472642764e-05, "loss": 1.1634, "step": 1184 }, { "epoch": 0.1, "learning_rate": 1.9741945964262532e-05, "loss": 1.1882, "step": 1185 }, { "epoch": 0.1, "learning_rate": 1.9741326469092424e-05, "loss": 1.16, "step": 1186 }, { "epoch": 0.1, "learning_rate": 1.974070624096392e-05, "loss": 1.2218, "step": 1187 }, { "epoch": 0.1, "learning_rate": 1.9740085279923704e-05, "loss": 2.8179, "step": 1188 }, { "epoch": 0.1, "learning_rate": 1.9739463586018486e-05, "loss": 1.4673, "step": 1189 }, { "epoch": 0.1, "learning_rate": 1.9738841159295046e-05, "loss": 1.323, "step": 1190 }, { "epoch": 0.1, "learning_rate": 1.9738217999800215e-05, "loss": 1.2891, "step": 1191 }, { "epoch": 0.1, "learning_rate": 1.9737594107580888e-05, "loss": 1.4165, "step": 1192 }, { "epoch": 0.1, "learning_rate": 1.9736969482683998e-05, "loss": 1.3192, "step": 1193 }, { "epoch": 0.1, "learning_rate": 1.973634412515655e-05, "loss": 1.317, "step": 1194 }, { "epoch": 0.1, "learning_rate": 1.9735718035045603e-05, "loss": 1.304, "step": 1195 }, { "epoch": 0.1, "learning_rate": 1.973509121239825e-05, "loss": 1.2442, "step": 1196 }, { "epoch": 0.1, "learning_rate": 1.973446365726166e-05, "loss": 1.2146, "step": 1197 }, { "epoch": 0.1, "learning_rate": 1.973383536968306e-05, "loss": 1.2786, "step": 1198 }, { "epoch": 0.1, "learning_rate": 1.9733206349709718e-05, "loss": 3.0607, "step": 1199 }, { "epoch": 0.1, "learning_rate": 1.973257659738896e-05, "loss": 1.2799, "step": 1200 }, { "epoch": 0.1, "learning_rate": 1.9731946112768172e-05, "loss": 1.293, "step": 1201 }, { "epoch": 0.1, "learning_rate": 1.9731314895894795e-05, "loss": 1.2661, "step": 1202 }, { "epoch": 0.1, "learning_rate": 1.9730682946816323e-05, "loss": 1.2537, "step": 1203 }, { "epoch": 0.1, "learning_rate": 1.9730050265580302e-05, "loss": 1.2223, "step": 1204 }, { "epoch": 0.1, "learning_rate": 1.9729416852234338e-05, "loss": 1.2914, "step": 1205 }, { "epoch": 0.1, "learning_rate": 1.972878270682609e-05, "loss": 1.2437, "step": 1206 }, { "epoch": 0.1, "learning_rate": 1.9728147829403275e-05, "loss": 1.2143, "step": 1207 }, { "epoch": 0.1, "learning_rate": 1.972751222001366e-05, "loss": 1.2327, "step": 1208 }, { "epoch": 0.1, "learning_rate": 1.972687587870507e-05, "loss": 1.3223, "step": 1209 }, { "epoch": 0.1, "learning_rate": 1.9726238805525383e-05, "loss": 1.2003, "step": 1210 }, { "epoch": 0.1, "learning_rate": 1.9725601000522537e-05, "loss": 1.176, "step": 1211 }, { "epoch": 0.1, "learning_rate": 1.9724962463744522e-05, "loss": 1.1907, "step": 1212 }, { "epoch": 0.1, "learning_rate": 1.972432319523938e-05, "loss": 1.2187, "step": 1213 }, { "epoch": 0.1, "learning_rate": 1.9723683195055214e-05, "loss": 1.1771, "step": 1214 }, { "epoch": 0.1, "learning_rate": 1.9723042463240173e-05, "loss": 1.2079, "step": 1215 }, { "epoch": 0.1, "learning_rate": 1.9722400999842474e-05, "loss": 1.1381, "step": 1216 }, { "epoch": 0.1, "learning_rate": 1.972175880491038e-05, "loss": 1.1328, "step": 1217 }, { "epoch": 0.1, "learning_rate": 1.972111587849221e-05, "loss": 1.2154, "step": 1218 }, { "epoch": 0.1, "learning_rate": 1.9720472220636337e-05, "loss": 1.2199, "step": 1219 }, { "epoch": 0.1, "learning_rate": 1.9719827831391197e-05, "loss": 1.2186, "step": 1220 }, { "epoch": 0.1, "learning_rate": 1.9719182710805272e-05, "loss": 1.212, "step": 1221 }, { "epoch": 0.1, "learning_rate": 1.9718536858927102e-05, "loss": 1.1792, "step": 1222 }, { "epoch": 0.1, "learning_rate": 1.9717890275805284e-05, "loss": 1.2434, "step": 1223 }, { "epoch": 0.1, "learning_rate": 1.9717242961488464e-05, "loss": 1.1865, "step": 1224 }, { "epoch": 0.1, "learning_rate": 1.9716594916025353e-05, "loss": 1.2058, "step": 1225 }, { "epoch": 0.1, "learning_rate": 1.9715946139464708e-05, "loss": 1.2332, "step": 1226 }, { "epoch": 0.1, "learning_rate": 1.9715296631855347e-05, "loss": 1.2072, "step": 1227 }, { "epoch": 0.1, "learning_rate": 1.9714646393246134e-05, "loss": 1.2487, "step": 1228 }, { "epoch": 0.1, "learning_rate": 1.9713995423686e-05, "loss": 1.2564, "step": 1229 }, { "epoch": 0.1, "learning_rate": 1.9713343723223924e-05, "loss": 1.1605, "step": 1230 }, { "epoch": 0.1, "learning_rate": 1.9712691291908943e-05, "loss": 2.2183, "step": 1231 }, { "epoch": 0.1, "learning_rate": 1.9712038129790145e-05, "loss": 1.1808, "step": 1232 }, { "epoch": 0.1, "learning_rate": 1.9711384236916676e-05, "loss": 1.2021, "step": 1233 }, { "epoch": 0.1, "learning_rate": 1.9710729613337736e-05, "loss": 1.2675, "step": 1234 }, { "epoch": 0.1, "learning_rate": 1.971007425910258e-05, "loss": 1.2363, "step": 1235 }, { "epoch": 0.1, "learning_rate": 1.9709418174260523e-05, "loss": 1.7601, "step": 1236 }, { "epoch": 0.1, "learning_rate": 1.970876135886092e-05, "loss": 1.1813, "step": 1237 }, { "epoch": 0.1, "learning_rate": 1.97081038129532e-05, "loss": 1.1793, "step": 1238 }, { "epoch": 0.1, "learning_rate": 1.9707445536586838e-05, "loss": 1.2106, "step": 1239 }, { "epoch": 0.11, "learning_rate": 1.970678652981136e-05, "loss": 1.2243, "step": 1240 }, { "epoch": 0.11, "learning_rate": 1.970612679267635e-05, "loss": 1.178, "step": 1241 }, { "epoch": 0.11, "learning_rate": 1.9705466325231456e-05, "loss": 1.1788, "step": 1242 }, { "epoch": 0.11, "learning_rate": 1.9704805127526365e-05, "loss": 1.1728, "step": 1243 }, { "epoch": 0.11, "learning_rate": 1.970414319961083e-05, "loss": 1.144, "step": 1244 }, { "epoch": 0.11, "learning_rate": 1.9703480541534656e-05, "loss": 1.1678, "step": 1245 }, { "epoch": 0.11, "learning_rate": 1.9702817153347702e-05, "loss": 1.243, "step": 1246 }, { "epoch": 0.11, "learning_rate": 1.9702153035099884e-05, "loss": 2.0031, "step": 1247 }, { "epoch": 0.11, "learning_rate": 1.970148818684117e-05, "loss": 1.1797, "step": 1248 }, { "epoch": 0.11, "learning_rate": 1.9700822608621585e-05, "loss": 1.2292, "step": 1249 }, { "epoch": 0.11, "learning_rate": 1.9700156300491207e-05, "loss": 1.2277, "step": 1250 }, { "epoch": 0.11, "learning_rate": 1.969948926250018e-05, "loss": 1.2453, "step": 1251 }, { "epoch": 0.11, "learning_rate": 1.969882149469868e-05, "loss": 1.1802, "step": 1252 }, { "epoch": 0.11, "learning_rate": 1.9698152997136956e-05, "loss": 1.172, "step": 1253 }, { "epoch": 0.11, "learning_rate": 1.969748376986531e-05, "loss": 1.2348, "step": 1254 }, { "epoch": 0.11, "learning_rate": 1.9696813812934094e-05, "loss": 1.2126, "step": 1255 }, { "epoch": 0.11, "learning_rate": 1.9696143126393718e-05, "loss": 1.1863, "step": 1256 }, { "epoch": 0.11, "learning_rate": 1.969547171029465e-05, "loss": 1.1485, "step": 1257 }, { "epoch": 0.11, "learning_rate": 1.9694799564687396e-05, "loss": 1.2497, "step": 1258 }, { "epoch": 0.11, "learning_rate": 1.9694126689622543e-05, "loss": 1.1781, "step": 1259 }, { "epoch": 0.11, "learning_rate": 1.9693453085150716e-05, "loss": 1.2054, "step": 1260 }, { "epoch": 0.11, "learning_rate": 1.9692778751322594e-05, "loss": 1.1792, "step": 1261 }, { "epoch": 0.11, "learning_rate": 1.9692103688188915e-05, "loss": 1.208, "step": 1262 }, { "epoch": 0.11, "learning_rate": 1.969142789580048e-05, "loss": 1.2122, "step": 1263 }, { "epoch": 0.11, "learning_rate": 1.969075137420813e-05, "loss": 1.1539, "step": 1264 }, { "epoch": 0.11, "learning_rate": 1.9690074123462772e-05, "loss": 1.1938, "step": 1265 }, { "epoch": 0.11, "learning_rate": 1.9689396143615362e-05, "loss": 1.1371, "step": 1266 }, { "epoch": 0.11, "learning_rate": 1.9688717434716912e-05, "loss": 1.1754, "step": 1267 }, { "epoch": 0.11, "learning_rate": 1.9688037996818493e-05, "loss": 1.1084, "step": 1268 }, { "epoch": 0.11, "learning_rate": 1.968735782997122e-05, "loss": 1.1742, "step": 1269 }, { "epoch": 0.11, "learning_rate": 1.968667693422628e-05, "loss": 1.1841, "step": 1270 }, { "epoch": 0.11, "learning_rate": 1.9685995309634897e-05, "loss": 1.2132, "step": 1271 }, { "epoch": 0.11, "learning_rate": 1.9685312956248365e-05, "loss": 2.3181, "step": 1272 }, { "epoch": 0.11, "learning_rate": 1.9684629874118016e-05, "loss": 1.1918, "step": 1273 }, { "epoch": 0.11, "learning_rate": 1.9683946063295254e-05, "loss": 1.2108, "step": 1274 }, { "epoch": 0.11, "learning_rate": 1.9683261523831533e-05, "loss": 1.2031, "step": 1275 }, { "epoch": 0.11, "learning_rate": 1.968257625577835e-05, "loss": 1.1533, "step": 1276 }, { "epoch": 0.11, "learning_rate": 1.9681890259187272e-05, "loss": 1.9734, "step": 1277 }, { "epoch": 0.11, "learning_rate": 1.9681203534109915e-05, "loss": 1.2537, "step": 1278 }, { "epoch": 0.11, "learning_rate": 1.9680516080597946e-05, "loss": 1.2656, "step": 1279 }, { "epoch": 0.11, "learning_rate": 1.9679827898703093e-05, "loss": 1.2677, "step": 1280 }, { "epoch": 0.11, "learning_rate": 1.967913898847714e-05, "loss": 1.2856, "step": 1281 }, { "epoch": 0.11, "learning_rate": 1.9678449349971915e-05, "loss": 1.3127, "step": 1282 }, { "epoch": 0.11, "learning_rate": 1.9677758983239313e-05, "loss": 1.274, "step": 1283 }, { "epoch": 0.11, "learning_rate": 1.9677067888331277e-05, "loss": 1.2209, "step": 1284 }, { "epoch": 0.11, "learning_rate": 1.9676376065299804e-05, "loss": 1.2347, "step": 1285 }, { "epoch": 0.11, "learning_rate": 1.9675683514196953e-05, "loss": 1.229, "step": 1286 }, { "epoch": 0.11, "learning_rate": 1.9674990235074828e-05, "loss": 1.1638, "step": 1287 }, { "epoch": 0.11, "learning_rate": 1.96742962279856e-05, "loss": 1.1731, "step": 1288 }, { "epoch": 0.11, "learning_rate": 1.967360149298148e-05, "loss": 1.1824, "step": 1289 }, { "epoch": 0.11, "learning_rate": 1.9672906030114744e-05, "loss": 1.2113, "step": 1290 }, { "epoch": 0.11, "learning_rate": 1.967220983943772e-05, "loss": 1.1888, "step": 1291 }, { "epoch": 0.11, "learning_rate": 1.9671512921002795e-05, "loss": 1.1905, "step": 1292 }, { "epoch": 0.11, "learning_rate": 1.96708152748624e-05, "loss": 1.2279, "step": 1293 }, { "epoch": 0.11, "learning_rate": 1.967011690106903e-05, "loss": 1.1719, "step": 1294 }, { "epoch": 0.11, "learning_rate": 1.9669417799675235e-05, "loss": 1.2232, "step": 1295 }, { "epoch": 0.11, "learning_rate": 1.9668717970733615e-05, "loss": 1.2452, "step": 1296 }, { "epoch": 0.11, "learning_rate": 1.9668017414296823e-05, "loss": 1.2141, "step": 1297 }, { "epoch": 0.11, "learning_rate": 1.9667316130417576e-05, "loss": 1.185, "step": 1298 }, { "epoch": 0.11, "learning_rate": 1.966661411914864e-05, "loss": 1.2004, "step": 1299 }, { "epoch": 0.11, "learning_rate": 1.9665911380542827e-05, "loss": 1.6327, "step": 1300 }, { "epoch": 0.11, "learning_rate": 1.9665207914653022e-05, "loss": 1.2234, "step": 1301 }, { "epoch": 0.11, "learning_rate": 1.9664503721532154e-05, "loss": 1.188, "step": 1302 }, { "epoch": 0.11, "learning_rate": 1.9663798801233206e-05, "loss": 1.1398, "step": 1303 }, { "epoch": 0.11, "learning_rate": 1.966309315380922e-05, "loss": 1.1832, "step": 1304 }, { "epoch": 0.11, "learning_rate": 1.966238677931328e-05, "loss": 1.1541, "step": 1305 }, { "epoch": 0.11, "learning_rate": 1.9661679677798554e-05, "loss": 1.1946, "step": 1306 }, { "epoch": 0.11, "learning_rate": 1.966097184931823e-05, "loss": 1.1914, "step": 1307 }, { "epoch": 0.11, "learning_rate": 1.9660263293925573e-05, "loss": 1.2638, "step": 1308 }, { "epoch": 0.11, "learning_rate": 1.9659554011673894e-05, "loss": 1.1641, "step": 1309 }, { "epoch": 0.11, "learning_rate": 1.9658844002616565e-05, "loss": 1.1509, "step": 1310 }, { "epoch": 0.11, "learning_rate": 1.9658133266807004e-05, "loss": 1.1587, "step": 1311 }, { "epoch": 0.11, "learning_rate": 1.965742180429869e-05, "loss": 1.2171, "step": 1312 }, { "epoch": 0.11, "learning_rate": 1.9656709615145154e-05, "loss": 1.2932, "step": 1313 }, { "epoch": 0.11, "learning_rate": 1.9655996699399985e-05, "loss": 1.188, "step": 1314 }, { "epoch": 0.11, "learning_rate": 1.9655283057116822e-05, "loss": 1.1922, "step": 1315 }, { "epoch": 0.11, "learning_rate": 1.9654568688349364e-05, "loss": 1.2572, "step": 1316 }, { "epoch": 0.11, "learning_rate": 1.965385359315136e-05, "loss": 1.1751, "step": 1317 }, { "epoch": 0.11, "learning_rate": 1.965313777157661e-05, "loss": 1.1308, "step": 1318 }, { "epoch": 0.11, "learning_rate": 1.9652421223678985e-05, "loss": 1.2023, "step": 1319 }, { "epoch": 0.11, "learning_rate": 1.9651703949512392e-05, "loss": 1.1938, "step": 1320 }, { "epoch": 0.11, "learning_rate": 1.96509859491308e-05, "loss": 1.1787, "step": 1321 }, { "epoch": 0.11, "learning_rate": 1.9650267222588234e-05, "loss": 1.1632, "step": 1322 }, { "epoch": 0.11, "learning_rate": 1.9649547769938774e-05, "loss": 1.1784, "step": 1323 }, { "epoch": 0.11, "learning_rate": 1.9648827591236554e-05, "loss": 1.1648, "step": 1324 }, { "epoch": 0.11, "learning_rate": 1.964810668653576e-05, "loss": 1.1549, "step": 1325 }, { "epoch": 0.11, "learning_rate": 1.9647385055890632e-05, "loss": 1.2055, "step": 1326 }, { "epoch": 0.11, "learning_rate": 1.9646662699355472e-05, "loss": 1.138, "step": 1327 }, { "epoch": 0.11, "learning_rate": 1.964593961698463e-05, "loss": 1.1581, "step": 1328 }, { "epoch": 0.11, "learning_rate": 1.964521580883251e-05, "loss": 1.1495, "step": 1329 }, { "epoch": 0.11, "learning_rate": 1.9644491274953573e-05, "loss": 1.1553, "step": 1330 }, { "epoch": 0.11, "learning_rate": 1.964376601540234e-05, "loss": 1.19, "step": 1331 }, { "epoch": 0.11, "learning_rate": 1.9643040030233375e-05, "loss": 1.1128, "step": 1332 }, { "epoch": 0.11, "learning_rate": 1.9642313319501304e-05, "loss": 1.1563, "step": 1333 }, { "epoch": 0.11, "learning_rate": 1.9641585883260808e-05, "loss": 1.1705, "step": 1334 }, { "epoch": 0.11, "learning_rate": 1.9640857721566618e-05, "loss": 1.1514, "step": 1335 }, { "epoch": 0.11, "learning_rate": 1.9640128834473527e-05, "loss": 1.161, "step": 1336 }, { "epoch": 0.11, "learning_rate": 1.963939922203637e-05, "loss": 1.1455, "step": 1337 }, { "epoch": 0.11, "learning_rate": 1.9638668884310055e-05, "loss": 1.1556, "step": 1338 }, { "epoch": 0.11, "learning_rate": 1.9637937821349528e-05, "loss": 1.1607, "step": 1339 }, { "epoch": 0.11, "learning_rate": 1.96372060332098e-05, "loss": 1.1617, "step": 1340 }, { "epoch": 0.11, "learning_rate": 1.9636473519945923e-05, "loss": 1.1564, "step": 1341 }, { "epoch": 0.11, "learning_rate": 1.9635740281613022e-05, "loss": 1.1513, "step": 1342 }, { "epoch": 0.11, "learning_rate": 1.9635006318266264e-05, "loss": 1.1628, "step": 1343 }, { "epoch": 0.11, "learning_rate": 1.9634271629960876e-05, "loss": 1.1644, "step": 1344 }, { "epoch": 0.11, "learning_rate": 1.963353621675214e-05, "loss": 1.1293, "step": 1345 }, { "epoch": 0.11, "learning_rate": 1.9632800078695378e-05, "loss": 1.174, "step": 1346 }, { "epoch": 0.11, "learning_rate": 1.963206321584599e-05, "loss": 1.1453, "step": 1347 }, { "epoch": 0.11, "learning_rate": 1.9631325628259415e-05, "loss": 1.1578, "step": 1348 }, { "epoch": 0.11, "learning_rate": 1.9630587315991153e-05, "loss": 1.1241, "step": 1349 }, { "epoch": 0.11, "learning_rate": 1.9629848279096754e-05, "loss": 1.1599, "step": 1350 }, { "epoch": 0.11, "learning_rate": 1.962910851763183e-05, "loss": 1.2053, "step": 1351 }, { "epoch": 0.11, "learning_rate": 1.962836803165203e-05, "loss": 1.1377, "step": 1352 }, { "epoch": 0.11, "learning_rate": 1.9627626821213082e-05, "loss": 1.1739, "step": 1353 }, { "epoch": 0.11, "learning_rate": 1.962688488637075e-05, "loss": 1.123, "step": 1354 }, { "epoch": 0.11, "learning_rate": 1.9626142227180864e-05, "loss": 1.1765, "step": 1355 }, { "epoch": 0.11, "learning_rate": 1.9625398843699295e-05, "loss": 1.176, "step": 1356 }, { "epoch": 0.11, "learning_rate": 1.9624654735981984e-05, "loss": 1.1998, "step": 1357 }, { "epoch": 0.12, "learning_rate": 1.9623909904084917e-05, "loss": 1.1745, "step": 1358 }, { "epoch": 0.12, "learning_rate": 1.9623164348064138e-05, "loss": 1.1422, "step": 1359 }, { "epoch": 0.12, "learning_rate": 1.962241806797574e-05, "loss": 1.1508, "step": 1360 }, { "epoch": 0.12, "learning_rate": 1.962167106387588e-05, "loss": 1.1538, "step": 1361 }, { "epoch": 0.12, "learning_rate": 1.9620923335820757e-05, "loss": 1.3314, "step": 1362 }, { "epoch": 0.12, "learning_rate": 1.962017488386664e-05, "loss": 1.1677, "step": 1363 }, { "epoch": 0.12, "learning_rate": 1.9619425708069842e-05, "loss": 1.1997, "step": 1364 }, { "epoch": 0.12, "learning_rate": 1.961867580848673e-05, "loss": 1.21, "step": 1365 }, { "epoch": 0.12, "learning_rate": 1.961792518517373e-05, "loss": 1.1622, "step": 1366 }, { "epoch": 0.12, "learning_rate": 1.9617173838187316e-05, "loss": 1.2001, "step": 1367 }, { "epoch": 0.12, "learning_rate": 1.961642176758403e-05, "loss": 1.1601, "step": 1368 }, { "epoch": 0.12, "learning_rate": 1.9615668973420453e-05, "loss": 1.1374, "step": 1369 }, { "epoch": 0.12, "learning_rate": 1.9614915455753225e-05, "loss": 1.1782, "step": 1370 }, { "epoch": 0.12, "learning_rate": 1.9614161214639046e-05, "loss": 1.1338, "step": 1371 }, { "epoch": 0.12, "learning_rate": 1.9613406250134668e-05, "loss": 1.1811, "step": 1372 }, { "epoch": 0.12, "learning_rate": 1.9612650562296892e-05, "loss": 1.0941, "step": 1373 }, { "epoch": 0.12, "learning_rate": 1.9611894151182582e-05, "loss": 1.1288, "step": 1374 }, { "epoch": 0.12, "learning_rate": 1.9611137016848648e-05, "loss": 1.1376, "step": 1375 }, { "epoch": 0.12, "learning_rate": 1.961037915935206e-05, "loss": 1.1615, "step": 1376 }, { "epoch": 0.12, "learning_rate": 1.9609620578749845e-05, "loss": 1.1693, "step": 1377 }, { "epoch": 0.12, "learning_rate": 1.960886127509907e-05, "loss": 1.1721, "step": 1378 }, { "epoch": 0.12, "learning_rate": 1.9608101248456877e-05, "loss": 1.2123, "step": 1379 }, { "epoch": 0.12, "learning_rate": 1.9607340498880448e-05, "loss": 1.1527, "step": 1380 }, { "epoch": 0.12, "learning_rate": 1.960657902642702e-05, "loss": 1.1233, "step": 1381 }, { "epoch": 0.12, "learning_rate": 1.9605816831153897e-05, "loss": 1.1957, "step": 1382 }, { "epoch": 0.12, "learning_rate": 1.9605053913118418e-05, "loss": 1.1604, "step": 1383 }, { "epoch": 0.12, "learning_rate": 1.960429027237799e-05, "loss": 1.1592, "step": 1384 }, { "epoch": 0.12, "learning_rate": 1.9603525908990075e-05, "loss": 1.1663, "step": 1385 }, { "epoch": 0.12, "learning_rate": 1.9602760823012182e-05, "loss": 1.1663, "step": 1386 }, { "epoch": 0.12, "learning_rate": 1.9601995014501874e-05, "loss": 1.1427, "step": 1387 }, { "epoch": 0.12, "learning_rate": 1.9601228483516782e-05, "loss": 1.1056, "step": 1388 }, { "epoch": 0.12, "learning_rate": 1.9600461230114572e-05, "loss": 1.1635, "step": 1389 }, { "epoch": 0.12, "learning_rate": 1.9599693254352978e-05, "loss": 1.197, "step": 1390 }, { "epoch": 0.12, "learning_rate": 1.9598924556289785e-05, "loss": 1.203, "step": 1391 }, { "epoch": 0.12, "learning_rate": 1.9598155135982827e-05, "loss": 1.1453, "step": 1392 }, { "epoch": 0.12, "learning_rate": 1.959738499349e-05, "loss": 1.146, "step": 1393 }, { "epoch": 0.12, "learning_rate": 1.9596614128869256e-05, "loss": 1.133, "step": 1394 }, { "epoch": 0.12, "learning_rate": 1.9595842542178588e-05, "loss": 1.1398, "step": 1395 }, { "epoch": 0.12, "learning_rate": 1.9595070233476056e-05, "loss": 1.1305, "step": 1396 }, { "epoch": 0.12, "learning_rate": 1.959429720281977e-05, "loss": 1.2074, "step": 1397 }, { "epoch": 0.12, "learning_rate": 1.959352345026789e-05, "loss": 1.1871, "step": 1398 }, { "epoch": 0.12, "learning_rate": 1.9592748975878643e-05, "loss": 1.1771, "step": 1399 }, { "epoch": 0.12, "learning_rate": 1.95919737797103e-05, "loss": 1.1041, "step": 1400 }, { "epoch": 0.12, "learning_rate": 1.9591197861821182e-05, "loss": 1.1471, "step": 1401 }, { "epoch": 0.12, "learning_rate": 1.9590421222269674e-05, "loss": 1.1132, "step": 1402 }, { "epoch": 0.12, "learning_rate": 1.9589643861114217e-05, "loss": 1.1909, "step": 1403 }, { "epoch": 0.12, "learning_rate": 1.9588865778413296e-05, "loss": 1.1255, "step": 1404 }, { "epoch": 0.12, "learning_rate": 1.958808697422546e-05, "loss": 1.1404, "step": 1405 }, { "epoch": 0.12, "learning_rate": 1.95873074486093e-05, "loss": 1.1432, "step": 1406 }, { "epoch": 0.12, "learning_rate": 1.9586527201623478e-05, "loss": 1.1597, "step": 1407 }, { "epoch": 0.12, "learning_rate": 1.9585746233326698e-05, "loss": 1.1486, "step": 1408 }, { "epoch": 0.12, "learning_rate": 1.9584964543777718e-05, "loss": 1.1734, "step": 1409 }, { "epoch": 0.12, "learning_rate": 1.958418213303536e-05, "loss": 1.1653, "step": 1410 }, { "epoch": 0.12, "learning_rate": 1.958339900115849e-05, "loss": 1.1404, "step": 1411 }, { "epoch": 0.12, "learning_rate": 1.9582615148206035e-05, "loss": 1.0704, "step": 1412 }, { "epoch": 0.12, "learning_rate": 1.9581830574236976e-05, "loss": 1.2422, "step": 1413 }, { "epoch": 0.12, "learning_rate": 1.958104527931034e-05, "loss": 1.1502, "step": 1414 }, { "epoch": 0.12, "learning_rate": 1.9580259263485218e-05, "loss": 1.193, "step": 1415 }, { "epoch": 0.12, "learning_rate": 1.957947252682075e-05, "loss": 1.149, "step": 1416 }, { "epoch": 0.12, "learning_rate": 1.957868506937613e-05, "loss": 1.1874, "step": 1417 }, { "epoch": 0.12, "learning_rate": 1.9577896891210615e-05, "loss": 1.1498, "step": 1418 }, { "epoch": 0.12, "learning_rate": 1.9577107992383505e-05, "loss": 1.0991, "step": 1419 }, { "epoch": 0.12, "learning_rate": 1.9576318372954157e-05, "loss": 1.1656, "step": 1420 }, { "epoch": 0.12, "learning_rate": 1.957552803298198e-05, "loss": 1.1597, "step": 1421 }, { "epoch": 0.12, "learning_rate": 1.9574736972526455e-05, "loss": 1.134, "step": 1422 }, { "epoch": 0.12, "learning_rate": 1.957394519164709e-05, "loss": 1.1091, "step": 1423 }, { "epoch": 0.12, "learning_rate": 1.957315269040346e-05, "loss": 1.1452, "step": 1424 }, { "epoch": 0.12, "learning_rate": 1.9572359468855203e-05, "loss": 1.1444, "step": 1425 }, { "epoch": 0.12, "learning_rate": 1.9571565527062e-05, "loss": 1.1285, "step": 1426 }, { "epoch": 0.12, "learning_rate": 1.9570770865083582e-05, "loss": 1.1838, "step": 1427 }, { "epoch": 0.12, "learning_rate": 1.956997548297975e-05, "loss": 1.1151, "step": 1428 }, { "epoch": 0.12, "learning_rate": 1.9569179380810348e-05, "loss": 1.1946, "step": 1429 }, { "epoch": 0.12, "learning_rate": 1.9568382558635272e-05, "loss": 1.1606, "step": 1430 }, { "epoch": 0.12, "learning_rate": 1.9567585016514484e-05, "loss": 1.1391, "step": 1431 }, { "epoch": 0.12, "learning_rate": 1.9566786754507983e-05, "loss": 1.155, "step": 1432 }, { "epoch": 0.12, "learning_rate": 1.956598777267584e-05, "loss": 1.1334, "step": 1433 }, { "epoch": 0.12, "learning_rate": 1.956518807107817e-05, "loss": 1.3154, "step": 1434 }, { "epoch": 0.12, "learning_rate": 1.956438764977514e-05, "loss": 1.1684, "step": 1435 }, { "epoch": 0.12, "learning_rate": 1.9563586508826984e-05, "loss": 1.1409, "step": 1436 }, { "epoch": 0.12, "learning_rate": 1.9562784648293978e-05, "loss": 1.1369, "step": 1437 }, { "epoch": 0.12, "learning_rate": 1.956198206823645e-05, "loss": 1.1926, "step": 1438 }, { "epoch": 0.12, "learning_rate": 1.9561178768714795e-05, "loss": 1.1588, "step": 1439 }, { "epoch": 0.12, "learning_rate": 1.9560374749789454e-05, "loss": 1.1734, "step": 1440 }, { "epoch": 0.12, "learning_rate": 1.9559570011520918e-05, "loss": 1.1459, "step": 1441 }, { "epoch": 0.12, "learning_rate": 1.9558764553969744e-05, "loss": 1.1488, "step": 1442 }, { "epoch": 0.12, "learning_rate": 1.955795837719653e-05, "loss": 1.1584, "step": 1443 }, { "epoch": 0.12, "learning_rate": 1.9557151481261938e-05, "loss": 1.1713, "step": 1444 }, { "epoch": 0.12, "learning_rate": 1.9556343866226682e-05, "loss": 1.1251, "step": 1445 }, { "epoch": 0.12, "learning_rate": 1.9555535532151527e-05, "loss": 1.1275, "step": 1446 }, { "epoch": 0.12, "learning_rate": 1.955472647909729e-05, "loss": 1.1608, "step": 1447 }, { "epoch": 0.12, "learning_rate": 1.9553916707124852e-05, "loss": 1.1141, "step": 1448 }, { "epoch": 0.12, "learning_rate": 1.9553106216295142e-05, "loss": 1.1366, "step": 1449 }, { "epoch": 0.12, "learning_rate": 1.9552295006669136e-05, "loss": 1.044, "step": 1450 }, { "epoch": 0.12, "learning_rate": 1.955148307830788e-05, "loss": 1.1687, "step": 1451 }, { "epoch": 0.12, "learning_rate": 1.9550670431272456e-05, "loss": 1.1492, "step": 1452 }, { "epoch": 0.12, "learning_rate": 1.954985706562402e-05, "loss": 1.1645, "step": 1453 }, { "epoch": 0.12, "learning_rate": 1.954904298142376e-05, "loss": 1.1871, "step": 1454 }, { "epoch": 0.12, "learning_rate": 1.9548228178732937e-05, "loss": 1.1593, "step": 1455 }, { "epoch": 0.12, "learning_rate": 1.9547412657612858e-05, "loss": 1.1411, "step": 1456 }, { "epoch": 0.12, "learning_rate": 1.9546596418124885e-05, "loss": 1.1345, "step": 1457 }, { "epoch": 0.12, "learning_rate": 1.954577946033043e-05, "loss": 1.1726, "step": 1458 }, { "epoch": 0.12, "learning_rate": 1.9544961784290964e-05, "loss": 1.117, "step": 1459 }, { "epoch": 0.12, "learning_rate": 1.9544143390068012e-05, "loss": 1.1187, "step": 1460 }, { "epoch": 0.12, "learning_rate": 1.954332427772315e-05, "loss": 1.1782, "step": 1461 }, { "epoch": 0.12, "learning_rate": 1.954250444731801e-05, "loss": 1.1007, "step": 1462 }, { "epoch": 0.12, "learning_rate": 1.954168389891428e-05, "loss": 1.1104, "step": 1463 }, { "epoch": 0.12, "learning_rate": 1.95408626325737e-05, "loss": 1.1316, "step": 1464 }, { "epoch": 0.12, "learning_rate": 1.954004064835806e-05, "loss": 1.1481, "step": 1465 }, { "epoch": 0.12, "learning_rate": 1.953921794632921e-05, "loss": 1.1443, "step": 1466 }, { "epoch": 0.12, "learning_rate": 1.9538394526549053e-05, "loss": 1.1143, "step": 1467 }, { "epoch": 0.12, "learning_rate": 1.9537570389079545e-05, "loss": 1.1706, "step": 1468 }, { "epoch": 0.12, "learning_rate": 1.953674553398269e-05, "loss": 1.1775, "step": 1469 }, { "epoch": 0.12, "learning_rate": 1.953591996132056e-05, "loss": 1.1264, "step": 1470 }, { "epoch": 0.12, "learning_rate": 1.953509367115527e-05, "loss": 1.1085, "step": 1471 }, { "epoch": 0.12, "learning_rate": 1.9534266663548993e-05, "loss": 1.1135, "step": 1472 }, { "epoch": 0.12, "learning_rate": 1.953343893856395e-05, "loss": 1.1384, "step": 1473 }, { "epoch": 0.12, "learning_rate": 1.9532610496262422e-05, "loss": 1.1024, "step": 1474 }, { "epoch": 0.12, "learning_rate": 1.953178133670675e-05, "loss": 1.1489, "step": 1475 }, { "epoch": 0.12, "learning_rate": 1.9530951459959313e-05, "loss": 1.1243, "step": 1476 }, { "epoch": 0.13, "learning_rate": 1.9530120866082555e-05, "loss": 1.0961, "step": 1477 }, { "epoch": 0.13, "learning_rate": 1.9529289555138976e-05, "loss": 1.1166, "step": 1478 }, { "epoch": 0.13, "learning_rate": 1.952845752719112e-05, "loss": 1.1497, "step": 1479 }, { "epoch": 0.13, "learning_rate": 1.9527624782301597e-05, "loss": 1.1183, "step": 1480 }, { "epoch": 0.13, "learning_rate": 1.952679132053306e-05, "loss": 1.09, "step": 1481 }, { "epoch": 0.13, "learning_rate": 1.952595714194822e-05, "loss": 1.1434, "step": 1482 }, { "epoch": 0.13, "learning_rate": 1.9525122246609842e-05, "loss": 1.1895, "step": 1483 }, { "epoch": 0.13, "learning_rate": 1.9524286634580747e-05, "loss": 1.1517, "step": 1484 }, { "epoch": 0.13, "learning_rate": 1.9523450305923808e-05, "loss": 1.2004, "step": 1485 }, { "epoch": 0.13, "learning_rate": 1.9522613260701953e-05, "loss": 1.1563, "step": 1486 }, { "epoch": 0.13, "learning_rate": 1.9521775498978163e-05, "loss": 1.1682, "step": 1487 }, { "epoch": 0.13, "learning_rate": 1.9520937020815468e-05, "loss": 1.1577, "step": 1488 }, { "epoch": 0.13, "learning_rate": 1.9520097826276965e-05, "loss": 1.1432, "step": 1489 }, { "epoch": 0.13, "learning_rate": 1.9519257915425793e-05, "loss": 1.1582, "step": 1490 }, { "epoch": 0.13, "learning_rate": 1.951841728832515e-05, "loss": 1.1826, "step": 1491 }, { "epoch": 0.13, "learning_rate": 1.9517575945038284e-05, "loss": 1.1925, "step": 1492 }, { "epoch": 0.13, "learning_rate": 1.9516733885628497e-05, "loss": 1.1199, "step": 1493 }, { "epoch": 0.13, "learning_rate": 1.9515891110159155e-05, "loss": 1.1619, "step": 1494 }, { "epoch": 0.13, "learning_rate": 1.951504761869367e-05, "loss": 1.1177, "step": 1495 }, { "epoch": 0.13, "learning_rate": 1.9514203411295498e-05, "loss": 1.1413, "step": 1496 }, { "epoch": 0.13, "learning_rate": 1.9513358488028167e-05, "loss": 1.1063, "step": 1497 }, { "epoch": 0.13, "learning_rate": 1.951251284895525e-05, "loss": 1.1879, "step": 1498 }, { "epoch": 0.13, "learning_rate": 1.9511666494140375e-05, "loss": 1.16, "step": 1499 }, { "epoch": 0.13, "learning_rate": 1.951081942364722e-05, "loss": 1.1275, "step": 1500 }, { "epoch": 0.13, "learning_rate": 1.9509971637539523e-05, "loss": 1.0963, "step": 1501 }, { "epoch": 0.13, "learning_rate": 1.950912313588108e-05, "loss": 1.2102, "step": 1502 }, { "epoch": 0.13, "learning_rate": 1.950827391873572e-05, "loss": 1.1752, "step": 1503 }, { "epoch": 0.13, "learning_rate": 1.950742398616735e-05, "loss": 1.1185, "step": 1504 }, { "epoch": 0.13, "learning_rate": 1.9506573338239917e-05, "loss": 1.1134, "step": 1505 }, { "epoch": 0.13, "learning_rate": 1.9505721975017426e-05, "loss": 1.1591, "step": 1506 }, { "epoch": 0.13, "learning_rate": 1.9504869896563936e-05, "loss": 1.1252, "step": 1507 }, { "epoch": 0.13, "learning_rate": 1.950401710294356e-05, "loss": 1.176, "step": 1508 }, { "epoch": 0.13, "learning_rate": 1.9503163594220465e-05, "loss": 1.1146, "step": 1509 }, { "epoch": 0.13, "learning_rate": 1.9502309370458868e-05, "loss": 1.1567, "step": 1510 }, { "epoch": 0.13, "learning_rate": 1.9501454431723044e-05, "loss": 1.161, "step": 1511 }, { "epoch": 0.13, "learning_rate": 1.950059877807732e-05, "loss": 1.0796, "step": 1512 }, { "epoch": 0.13, "learning_rate": 1.9499742409586075e-05, "loss": 1.1089, "step": 1513 }, { "epoch": 0.13, "learning_rate": 1.9498885326313747e-05, "loss": 1.1114, "step": 1514 }, { "epoch": 0.13, "learning_rate": 1.9498027528324826e-05, "loss": 1.1782, "step": 1515 }, { "epoch": 0.13, "learning_rate": 1.9497169015683854e-05, "loss": 1.1133, "step": 1516 }, { "epoch": 0.13, "learning_rate": 1.9496309788455425e-05, "loss": 1.0822, "step": 1517 }, { "epoch": 0.13, "learning_rate": 1.9495449846704187e-05, "loss": 1.1189, "step": 1518 }, { "epoch": 0.13, "learning_rate": 1.949458919049485e-05, "loss": 1.1344, "step": 1519 }, { "epoch": 0.13, "learning_rate": 1.949372781989217e-05, "loss": 1.1614, "step": 1520 }, { "epoch": 0.13, "learning_rate": 1.9492865734960956e-05, "loss": 1.1756, "step": 1521 }, { "epoch": 0.13, "learning_rate": 1.9492002935766074e-05, "loss": 1.1351, "step": 1522 }, { "epoch": 0.13, "learning_rate": 1.949113942237244e-05, "loss": 1.0718, "step": 1523 }, { "epoch": 0.13, "learning_rate": 1.9490275194845036e-05, "loss": 1.0528, "step": 1524 }, { "epoch": 0.13, "learning_rate": 1.9489410253248876e-05, "loss": 1.1676, "step": 1525 }, { "epoch": 0.13, "learning_rate": 1.9488544597649048e-05, "loss": 1.1373, "step": 1526 }, { "epoch": 0.13, "learning_rate": 1.9487678228110685e-05, "loss": 1.1424, "step": 1527 }, { "epoch": 0.13, "learning_rate": 1.9486811144698975e-05, "loss": 1.0941, "step": 1528 }, { "epoch": 0.13, "learning_rate": 1.9485943347479153e-05, "loss": 1.11, "step": 1529 }, { "epoch": 0.13, "learning_rate": 1.948507483651652e-05, "loss": 1.1522, "step": 1530 }, { "epoch": 0.13, "learning_rate": 1.9484205611876427e-05, "loss": 1.1927, "step": 1531 }, { "epoch": 0.13, "learning_rate": 1.948333567362427e-05, "loss": 1.1969, "step": 1532 }, { "epoch": 0.13, "learning_rate": 1.9482465021825504e-05, "loss": 1.156, "step": 1533 }, { "epoch": 0.13, "learning_rate": 1.9481593656545647e-05, "loss": 1.1819, "step": 1534 }, { "epoch": 0.13, "learning_rate": 1.9480721577850256e-05, "loss": 1.1244, "step": 1535 }, { "epoch": 0.13, "learning_rate": 1.947984878580495e-05, "loss": 1.0973, "step": 1536 }, { "epoch": 0.13, "learning_rate": 1.94789752804754e-05, "loss": 1.0874, "step": 1537 }, { "epoch": 0.13, "learning_rate": 1.947810106192733e-05, "loss": 1.1507, "step": 1538 }, { "epoch": 0.13, "learning_rate": 1.9477226130226516e-05, "loss": 1.1418, "step": 1539 }, { "epoch": 0.13, "learning_rate": 1.9476350485438792e-05, "loss": 1.1573, "step": 1540 }, { "epoch": 0.13, "learning_rate": 1.9475474127630046e-05, "loss": 1.1623, "step": 1541 }, { "epoch": 0.13, "learning_rate": 1.9474597056866214e-05, "loss": 1.1695, "step": 1542 }, { "epoch": 0.13, "learning_rate": 1.9473719273213287e-05, "loss": 1.1103, "step": 1543 }, { "epoch": 0.13, "learning_rate": 1.9472840776737314e-05, "loss": 1.0815, "step": 1544 }, { "epoch": 0.13, "learning_rate": 1.9471961567504393e-05, "loss": 1.1847, "step": 1545 }, { "epoch": 0.13, "learning_rate": 1.9471081645580678e-05, "loss": 1.103, "step": 1546 }, { "epoch": 0.13, "learning_rate": 1.9470201011032382e-05, "loss": 1.1728, "step": 1547 }, { "epoch": 0.13, "learning_rate": 1.9469319663925758e-05, "loss": 1.1357, "step": 1548 }, { "epoch": 0.13, "learning_rate": 1.946843760432712e-05, "loss": 1.0755, "step": 1549 }, { "epoch": 0.13, "learning_rate": 1.9467554832302845e-05, "loss": 1.164, "step": 1550 }, { "epoch": 0.13, "learning_rate": 1.9466671347919344e-05, "loss": 1.1765, "step": 1551 }, { "epoch": 0.13, "learning_rate": 1.94657871512431e-05, "loss": 1.1865, "step": 1552 }, { "epoch": 0.13, "learning_rate": 1.946490224234064e-05, "loss": 1.1458, "step": 1553 }, { "epoch": 0.13, "learning_rate": 1.9464016621278544e-05, "loss": 1.1624, "step": 1554 }, { "epoch": 0.13, "learning_rate": 1.946313028812345e-05, "loss": 1.1708, "step": 1555 }, { "epoch": 0.13, "learning_rate": 1.9462243242942045e-05, "loss": 1.1158, "step": 1556 }, { "epoch": 0.13, "learning_rate": 1.9461355485801076e-05, "loss": 1.1617, "step": 1557 }, { "epoch": 0.13, "learning_rate": 1.946046701676734e-05, "loss": 1.1356, "step": 1558 }, { "epoch": 0.13, "learning_rate": 1.9459577835907687e-05, "loss": 1.1267, "step": 1559 }, { "epoch": 0.13, "learning_rate": 1.9458687943289014e-05, "loss": 1.1218, "step": 1560 }, { "epoch": 0.13, "learning_rate": 1.945779733897829e-05, "loss": 1.1594, "step": 1561 }, { "epoch": 0.13, "learning_rate": 1.9456906023042515e-05, "loss": 1.2203, "step": 1562 }, { "epoch": 0.13, "learning_rate": 1.945601399554876e-05, "loss": 1.153, "step": 1563 }, { "epoch": 0.13, "learning_rate": 1.9455121256564142e-05, "loss": 1.1019, "step": 1564 }, { "epoch": 0.13, "learning_rate": 1.9454227806155833e-05, "loss": 1.0148, "step": 1565 }, { "epoch": 0.13, "learning_rate": 1.9453333644391054e-05, "loss": 1.1169, "step": 1566 }, { "epoch": 0.13, "learning_rate": 1.945243877133709e-05, "loss": 0.9975, "step": 1567 }, { "epoch": 0.13, "learning_rate": 1.945154318706127e-05, "loss": 1.1341, "step": 1568 }, { "epoch": 0.13, "learning_rate": 1.945064689163098e-05, "loss": 1.1478, "step": 1569 }, { "epoch": 0.13, "learning_rate": 1.9449749885113656e-05, "loss": 1.0941, "step": 1570 }, { "epoch": 0.13, "learning_rate": 1.9448852167576793e-05, "loss": 1.1618, "step": 1571 }, { "epoch": 0.13, "learning_rate": 1.9447953739087938e-05, "loss": 1.145, "step": 1572 }, { "epoch": 0.13, "learning_rate": 1.9447054599714695e-05, "loss": 1.1637, "step": 1573 }, { "epoch": 0.13, "learning_rate": 1.944615474952471e-05, "loss": 1.117, "step": 1574 }, { "epoch": 0.13, "learning_rate": 1.9445254188585695e-05, "loss": 1.0548, "step": 1575 }, { "epoch": 0.13, "learning_rate": 1.9444352916965402e-05, "loss": 1.1047, "step": 1576 }, { "epoch": 0.13, "learning_rate": 1.9443450934731653e-05, "loss": 1.1523, "step": 1577 }, { "epoch": 0.13, "learning_rate": 1.9442548241952312e-05, "loss": 1.157, "step": 1578 }, { "epoch": 0.13, "learning_rate": 1.94416448386953e-05, "loss": 1.1324, "step": 1579 }, { "epoch": 0.13, "learning_rate": 1.9440740725028593e-05, "loss": 1.184, "step": 1580 }, { "epoch": 0.13, "learning_rate": 1.9439835901020214e-05, "loss": 1.1487, "step": 1581 }, { "epoch": 0.13, "learning_rate": 1.9438930366738246e-05, "loss": 1.1394, "step": 1582 }, { "epoch": 0.13, "learning_rate": 1.9438024122250824e-05, "loss": 1.134, "step": 1583 }, { "epoch": 0.13, "learning_rate": 1.9437117167626137e-05, "loss": 1.1034, "step": 1584 }, { "epoch": 0.13, "learning_rate": 1.9436209502932423e-05, "loss": 1.1307, "step": 1585 }, { "epoch": 0.13, "learning_rate": 1.9435301128237976e-05, "loss": 1.1404, "step": 1586 }, { "epoch": 0.13, "learning_rate": 1.9434392043611148e-05, "loss": 1.2027, "step": 1587 }, { "epoch": 0.13, "learning_rate": 1.943348224912034e-05, "loss": 1.1368, "step": 1588 }, { "epoch": 0.13, "learning_rate": 1.9432571744834002e-05, "loss": 1.1323, "step": 1589 }, { "epoch": 0.13, "learning_rate": 1.943166053082065e-05, "loss": 1.0768, "step": 1590 }, { "epoch": 0.13, "learning_rate": 1.943074860714884e-05, "loss": 1.1399, "step": 1591 }, { "epoch": 0.13, "learning_rate": 1.9429835973887187e-05, "loss": 1.1472, "step": 1592 }, { "epoch": 0.13, "learning_rate": 1.9428922631104364e-05, "loss": 1.2249, "step": 1593 }, { "epoch": 0.13, "learning_rate": 1.942800857886909e-05, "loss": 1.1502, "step": 1594 }, { "epoch": 0.14, "learning_rate": 1.9427093817250137e-05, "loss": 1.0914, "step": 1595 }, { "epoch": 0.14, "learning_rate": 1.942617834631634e-05, "loss": 1.1833, "step": 1596 }, { "epoch": 0.14, "learning_rate": 1.942526216613658e-05, "loss": 1.2183, "step": 1597 }, { "epoch": 0.14, "learning_rate": 1.9424345276779785e-05, "loss": 1.1653, "step": 1598 }, { "epoch": 0.14, "learning_rate": 1.9423427678314953e-05, "loss": 1.1293, "step": 1599 }, { "epoch": 0.14, "learning_rate": 1.942250937081112e-05, "loss": 1.1433, "step": 1600 }, { "epoch": 0.14, "learning_rate": 1.9421590354337385e-05, "loss": 1.1337, "step": 1601 }, { "epoch": 0.14, "learning_rate": 1.9420670628962895e-05, "loss": 1.1317, "step": 1602 }, { "epoch": 0.14, "learning_rate": 1.9419750194756852e-05, "loss": 0.9573, "step": 1603 }, { "epoch": 0.14, "learning_rate": 1.9418829051788513e-05, "loss": 1.1399, "step": 1604 }, { "epoch": 0.14, "learning_rate": 1.9417907200127185e-05, "loss": 1.1576, "step": 1605 }, { "epoch": 0.14, "learning_rate": 1.9416984639842227e-05, "loss": 1.1165, "step": 1606 }, { "epoch": 0.14, "learning_rate": 1.9416061371003065e-05, "loss": 1.1444, "step": 1607 }, { "epoch": 0.14, "learning_rate": 1.941513739367916e-05, "loss": 1.1184, "step": 1608 }, { "epoch": 0.14, "learning_rate": 1.941421270794003e-05, "loss": 1.2291, "step": 1609 }, { "epoch": 0.14, "learning_rate": 1.941328731385526e-05, "loss": 1.152, "step": 1610 }, { "epoch": 0.14, "learning_rate": 1.9412361211494475e-05, "loss": 1.1247, "step": 1611 }, { "epoch": 0.14, "learning_rate": 1.941143440092735e-05, "loss": 1.1057, "step": 1612 }, { "epoch": 0.14, "learning_rate": 1.9410506882223632e-05, "loss": 1.18, "step": 1613 }, { "epoch": 0.14, "learning_rate": 1.9409578655453103e-05, "loss": 1.1329, "step": 1614 }, { "epoch": 0.14, "learning_rate": 1.9408649720685603e-05, "loss": 1.1359, "step": 1615 }, { "epoch": 0.14, "learning_rate": 1.940772007799103e-05, "loss": 1.2097, "step": 1616 }, { "epoch": 0.14, "learning_rate": 1.9406789727439334e-05, "loss": 1.1806, "step": 1617 }, { "epoch": 0.14, "learning_rate": 1.940585866910051e-05, "loss": 1.1151, "step": 1618 }, { "epoch": 0.14, "learning_rate": 1.940492690304462e-05, "loss": 1.1687, "step": 1619 }, { "epoch": 0.14, "learning_rate": 1.9403994429341773e-05, "loss": 1.1273, "step": 1620 }, { "epoch": 0.14, "learning_rate": 1.9403061248062124e-05, "loss": 1.1846, "step": 1621 }, { "epoch": 0.14, "learning_rate": 1.9402127359275888e-05, "loss": 1.1094, "step": 1622 }, { "epoch": 0.14, "learning_rate": 1.9401192763053335e-05, "loss": 1.1362, "step": 1623 }, { "epoch": 0.14, "learning_rate": 1.940025745946479e-05, "loss": 1.1302, "step": 1624 }, { "epoch": 0.14, "learning_rate": 1.939932144858062e-05, "loss": 1.1744, "step": 1625 }, { "epoch": 0.14, "learning_rate": 1.9398384730471256e-05, "loss": 1.1189, "step": 1626 }, { "epoch": 0.14, "learning_rate": 1.939744730520718e-05, "loss": 1.1268, "step": 1627 }, { "epoch": 0.14, "learning_rate": 1.9396509172858924e-05, "loss": 1.193, "step": 1628 }, { "epoch": 0.14, "learning_rate": 1.9395570333497074e-05, "loss": 1.1572, "step": 1629 }, { "epoch": 0.14, "learning_rate": 1.9394630787192276e-05, "loss": 1.1233, "step": 1630 }, { "epoch": 0.14, "learning_rate": 1.9393690534015215e-05, "loss": 1.1464, "step": 1631 }, { "epoch": 0.14, "learning_rate": 1.9392749574036643e-05, "loss": 1.145, "step": 1632 }, { "epoch": 0.14, "learning_rate": 1.9391807907327357e-05, "loss": 1.1687, "step": 1633 }, { "epoch": 0.14, "learning_rate": 1.9390865533958215e-05, "loss": 1.1327, "step": 1634 }, { "epoch": 0.14, "learning_rate": 1.938992245400012e-05, "loss": 1.1858, "step": 1635 }, { "epoch": 0.14, "learning_rate": 1.9388978667524033e-05, "loss": 1.1467, "step": 1636 }, { "epoch": 0.14, "learning_rate": 1.938803417460096e-05, "loss": 1.1501, "step": 1637 }, { "epoch": 0.14, "learning_rate": 1.9387088975301975e-05, "loss": 1.1551, "step": 1638 }, { "epoch": 0.14, "learning_rate": 1.9386143069698192e-05, "loss": 1.1164, "step": 1639 }, { "epoch": 0.14, "learning_rate": 1.9385196457860784e-05, "loss": 1.124, "step": 1640 }, { "epoch": 0.14, "learning_rate": 1.938424913986098e-05, "loss": 1.1792, "step": 1641 }, { "epoch": 0.14, "learning_rate": 1.938330111577005e-05, "loss": 1.116, "step": 1642 }, { "epoch": 0.14, "learning_rate": 1.9382352385659337e-05, "loss": 1.15, "step": 1643 }, { "epoch": 0.14, "learning_rate": 1.9381402949600214e-05, "loss": 1.1123, "step": 1644 }, { "epoch": 0.14, "learning_rate": 1.938045280766412e-05, "loss": 1.1235, "step": 1645 }, { "epoch": 0.14, "learning_rate": 1.9379501959922553e-05, "loss": 1.1029, "step": 1646 }, { "epoch": 0.14, "learning_rate": 1.9378550406447053e-05, "loss": 1.1552, "step": 1647 }, { "epoch": 0.14, "learning_rate": 1.9377598147309218e-05, "loss": 1.1444, "step": 1648 }, { "epoch": 0.14, "learning_rate": 1.9376645182580693e-05, "loss": 1.1824, "step": 1649 }, { "epoch": 0.14, "learning_rate": 1.9375691512333188e-05, "loss": 1.1568, "step": 1650 }, { "epoch": 0.14, "learning_rate": 1.9374737136638453e-05, "loss": 1.1957, "step": 1651 }, { "epoch": 0.14, "learning_rate": 1.9373782055568304e-05, "loss": 1.1327, "step": 1652 }, { "epoch": 0.14, "learning_rate": 1.93728262691946e-05, "loss": 1.0769, "step": 1653 }, { "epoch": 0.14, "learning_rate": 1.937186977758925e-05, "loss": 1.0821, "step": 1654 }, { "epoch": 0.14, "learning_rate": 1.9370912580824238e-05, "loss": 1.1601, "step": 1655 }, { "epoch": 0.14, "learning_rate": 1.9369954678971568e-05, "loss": 1.0923, "step": 1656 }, { "epoch": 0.14, "learning_rate": 1.9368996072103327e-05, "loss": 1.0987, "step": 1657 }, { "epoch": 0.14, "learning_rate": 1.9368036760291637e-05, "loss": 1.152, "step": 1658 }, { "epoch": 0.14, "learning_rate": 1.9367076743608683e-05, "loss": 1.0785, "step": 1659 }, { "epoch": 0.14, "learning_rate": 1.9366116022126694e-05, "loss": 1.1333, "step": 1660 }, { "epoch": 0.14, "learning_rate": 1.9365154595917958e-05, "loss": 1.1816, "step": 1661 }, { "epoch": 0.14, "learning_rate": 1.9364192465054815e-05, "loss": 1.1858, "step": 1662 }, { "epoch": 0.14, "learning_rate": 1.936322962960966e-05, "loss": 1.181, "step": 1663 }, { "epoch": 0.14, "learning_rate": 1.9362266089654937e-05, "loss": 1.0917, "step": 1664 }, { "epoch": 0.14, "learning_rate": 1.9361301845263143e-05, "loss": 1.1627, "step": 1665 }, { "epoch": 0.14, "learning_rate": 1.9360336896506834e-05, "loss": 1.1112, "step": 1666 }, { "epoch": 0.14, "learning_rate": 1.935937124345861e-05, "loss": 1.1481, "step": 1667 }, { "epoch": 0.14, "learning_rate": 1.9358404886191133e-05, "loss": 1.1136, "step": 1668 }, { "epoch": 0.14, "learning_rate": 1.9357437824777112e-05, "loss": 1.1589, "step": 1669 }, { "epoch": 0.14, "learning_rate": 1.9356470059289313e-05, "loss": 1.1064, "step": 1670 }, { "epoch": 0.14, "learning_rate": 1.9355501589800548e-05, "loss": 1.1433, "step": 1671 }, { "epoch": 0.14, "learning_rate": 1.935453241638369e-05, "loss": 1.0868, "step": 1672 }, { "epoch": 0.14, "learning_rate": 1.935356253911166e-05, "loss": 1.1282, "step": 1673 }, { "epoch": 0.14, "learning_rate": 1.9352591958057435e-05, "loss": 1.0571, "step": 1674 }, { "epoch": 0.14, "learning_rate": 1.9351620673294046e-05, "loss": 1.1703, "step": 1675 }, { "epoch": 0.14, "learning_rate": 1.9350648684894567e-05, "loss": 1.1449, "step": 1676 }, { "epoch": 0.14, "learning_rate": 1.9349675992932143e-05, "loss": 1.1281, "step": 1677 }, { "epoch": 0.14, "learning_rate": 1.934870259747995e-05, "loss": 1.1454, "step": 1678 }, { "epoch": 0.14, "learning_rate": 1.9347728498611238e-05, "loss": 1.168, "step": 1679 }, { "epoch": 0.14, "learning_rate": 1.9346753696399295e-05, "loss": 1.1279, "step": 1680 }, { "epoch": 0.14, "learning_rate": 1.934577819091747e-05, "loss": 1.0999, "step": 1681 }, { "epoch": 0.14, "learning_rate": 1.9344801982239163e-05, "loss": 1.1101, "step": 1682 }, { "epoch": 0.14, "learning_rate": 1.934382507043782e-05, "loss": 1.1565, "step": 1683 }, { "epoch": 0.14, "learning_rate": 1.934284745558695e-05, "loss": 1.0722, "step": 1684 }, { "epoch": 0.14, "learning_rate": 1.9341869137760114e-05, "loss": 1.1134, "step": 1685 }, { "epoch": 0.14, "learning_rate": 1.9340890117030916e-05, "loss": 1.1441, "step": 1686 }, { "epoch": 0.14, "learning_rate": 1.9339910393473026e-05, "loss": 1.0704, "step": 1687 }, { "epoch": 0.14, "learning_rate": 1.9338929967160157e-05, "loss": 1.0873, "step": 1688 }, { "epoch": 0.14, "learning_rate": 1.933794883816608e-05, "loss": 1.1154, "step": 1689 }, { "epoch": 0.14, "learning_rate": 1.933696700656461e-05, "loss": 1.1079, "step": 1690 }, { "epoch": 0.14, "learning_rate": 1.9335984472429633e-05, "loss": 1.1079, "step": 1691 }, { "epoch": 0.14, "learning_rate": 1.9335001235835074e-05, "loss": 1.1241, "step": 1692 }, { "epoch": 0.14, "learning_rate": 1.933401729685491e-05, "loss": 1.1346, "step": 1693 }, { "epoch": 0.14, "learning_rate": 1.9333032655563176e-05, "loss": 1.1068, "step": 1694 }, { "epoch": 0.14, "learning_rate": 1.933204731203396e-05, "loss": 1.1527, "step": 1695 }, { "epoch": 0.14, "learning_rate": 1.93310612663414e-05, "loss": 1.1496, "step": 1696 }, { "epoch": 0.14, "learning_rate": 1.9330074518559695e-05, "loss": 0.9714, "step": 1697 }, { "epoch": 0.14, "learning_rate": 1.932908706876308e-05, "loss": 1.1545, "step": 1698 }, { "epoch": 0.14, "learning_rate": 1.9328098917025855e-05, "loss": 1.1307, "step": 1699 }, { "epoch": 0.14, "learning_rate": 1.9327110063422375e-05, "loss": 1.1604, "step": 1700 }, { "epoch": 0.14, "learning_rate": 1.9326120508027037e-05, "loss": 1.1233, "step": 1701 }, { "epoch": 0.14, "learning_rate": 1.9325130250914304e-05, "loss": 1.0951, "step": 1702 }, { "epoch": 0.14, "learning_rate": 1.9324139292158682e-05, "loss": 1.098, "step": 1703 }, { "epoch": 0.14, "learning_rate": 1.9323147631834737e-05, "loss": 1.1267, "step": 1704 }, { "epoch": 0.14, "learning_rate": 1.9322155270017074e-05, "loss": 1.1278, "step": 1705 }, { "epoch": 0.14, "learning_rate": 1.932116220678037e-05, "loss": 1.1621, "step": 1706 }, { "epoch": 0.14, "learning_rate": 1.932016844219934e-05, "loss": 1.0869, "step": 1707 }, { "epoch": 0.14, "learning_rate": 1.931917397634876e-05, "loss": 1.1188, "step": 1708 }, { "epoch": 0.14, "learning_rate": 1.9318178809303454e-05, "loss": 1.1208, "step": 1709 }, { "epoch": 0.14, "learning_rate": 1.93171829411383e-05, "loss": 1.2153, "step": 1710 }, { "epoch": 0.14, "learning_rate": 1.9316186371928228e-05, "loss": 1.1639, "step": 1711 }, { "epoch": 0.14, "learning_rate": 1.9315189101748228e-05, "loss": 1.1552, "step": 1712 }, { "epoch": 0.15, "learning_rate": 1.9314191130673332e-05, "loss": 1.1165, "step": 1713 }, { "epoch": 0.15, "learning_rate": 1.931319245877863e-05, "loss": 1.1141, "step": 1714 }, { "epoch": 0.15, "learning_rate": 1.9312193086139265e-05, "loss": 1.0997, "step": 1715 }, { "epoch": 0.15, "learning_rate": 1.9311193012830428e-05, "loss": 1.1586, "step": 1716 }, { "epoch": 0.15, "learning_rate": 1.9310192238927376e-05, "loss": 1.0914, "step": 1717 }, { "epoch": 0.15, "learning_rate": 1.93091907645054e-05, "loss": 1.179, "step": 1718 }, { "epoch": 0.15, "learning_rate": 1.9308188589639858e-05, "loss": 1.0888, "step": 1719 }, { "epoch": 0.15, "learning_rate": 1.930718571440615e-05, "loss": 1.1123, "step": 1720 }, { "epoch": 0.15, "learning_rate": 1.9306182138879745e-05, "loss": 1.1172, "step": 1721 }, { "epoch": 0.15, "learning_rate": 1.9305177863136145e-05, "loss": 1.1215, "step": 1722 }, { "epoch": 0.15, "learning_rate": 1.930417288725092e-05, "loss": 1.1868, "step": 1723 }, { "epoch": 0.15, "learning_rate": 1.930316721129968e-05, "loss": 1.1198, "step": 1724 }, { "epoch": 0.15, "learning_rate": 1.9302160835358096e-05, "loss": 0.9254, "step": 1725 }, { "epoch": 0.15, "learning_rate": 1.9301153759501896e-05, "loss": 1.1141, "step": 1726 }, { "epoch": 0.15, "learning_rate": 1.9300145983806852e-05, "loss": 1.1499, "step": 1727 }, { "epoch": 0.15, "learning_rate": 1.9299137508348786e-05, "loss": 1.1117, "step": 1728 }, { "epoch": 0.15, "learning_rate": 1.929812833320358e-05, "loss": 1.1607, "step": 1729 }, { "epoch": 0.15, "learning_rate": 1.9297118458447168e-05, "loss": 1.1116, "step": 1730 }, { "epoch": 0.15, "learning_rate": 1.9296107884155537e-05, "loss": 1.1025, "step": 1731 }, { "epoch": 0.15, "learning_rate": 1.9295096610404722e-05, "loss": 1.159, "step": 1732 }, { "epoch": 0.15, "learning_rate": 1.9294084637270816e-05, "loss": 1.1678, "step": 1733 }, { "epoch": 0.15, "learning_rate": 1.929307196482996e-05, "loss": 1.1036, "step": 1734 }, { "epoch": 0.15, "learning_rate": 1.9292058593158348e-05, "loss": 1.0757, "step": 1735 }, { "epoch": 0.15, "learning_rate": 1.929104452233223e-05, "loss": 1.1397, "step": 1736 }, { "epoch": 0.15, "learning_rate": 1.9290029752427908e-05, "loss": 1.1971, "step": 1737 }, { "epoch": 0.15, "learning_rate": 1.928901428352173e-05, "loss": 1.0315, "step": 1738 }, { "epoch": 0.15, "learning_rate": 1.9287998115690115e-05, "loss": 1.1604, "step": 1739 }, { "epoch": 0.15, "learning_rate": 1.9286981249009507e-05, "loss": 1.0979, "step": 1740 }, { "epoch": 0.15, "learning_rate": 1.9285963683556425e-05, "loss": 1.1018, "step": 1741 }, { "epoch": 0.15, "learning_rate": 1.9284945419407434e-05, "loss": 1.1263, "step": 1742 }, { "epoch": 0.15, "learning_rate": 1.9283926456639143e-05, "loss": 1.1133, "step": 1743 }, { "epoch": 0.15, "learning_rate": 1.9282906795328228e-05, "loss": 1.0401, "step": 1744 }, { "epoch": 0.15, "learning_rate": 1.928188643555141e-05, "loss": 1.1396, "step": 1745 }, { "epoch": 0.15, "learning_rate": 1.9280865377385458e-05, "loss": 1.1445, "step": 1746 }, { "epoch": 0.15, "learning_rate": 1.9279843620907204e-05, "loss": 1.1307, "step": 1747 }, { "epoch": 0.15, "learning_rate": 1.9278821166193526e-05, "loss": 1.1232, "step": 1748 }, { "epoch": 0.15, "learning_rate": 1.9277798013321353e-05, "loss": 1.0523, "step": 1749 }, { "epoch": 0.15, "learning_rate": 1.9276774162367665e-05, "loss": 1.1221, "step": 1750 }, { "epoch": 0.15, "learning_rate": 1.9275749613409514e-05, "loss": 1.1357, "step": 1751 }, { "epoch": 0.15, "learning_rate": 1.9274724366523973e-05, "loss": 1.1949, "step": 1752 }, { "epoch": 0.15, "learning_rate": 1.9273698421788197e-05, "loss": 1.1907, "step": 1753 }, { "epoch": 0.15, "learning_rate": 1.927267177927937e-05, "loss": 1.1202, "step": 1754 }, { "epoch": 0.15, "learning_rate": 1.9271644439074745e-05, "loss": 1.0757, "step": 1755 }, { "epoch": 0.15, "learning_rate": 1.9270616401251615e-05, "loss": 1.1512, "step": 1756 }, { "epoch": 0.15, "learning_rate": 1.926958766588734e-05, "loss": 1.1522, "step": 1757 }, { "epoch": 0.15, "learning_rate": 1.926855823305932e-05, "loss": 1.1238, "step": 1758 }, { "epoch": 0.15, "learning_rate": 1.926752810284501e-05, "loss": 1.1763, "step": 1759 }, { "epoch": 0.15, "learning_rate": 1.926649727532192e-05, "loss": 1.1553, "step": 1760 }, { "epoch": 0.15, "learning_rate": 1.9265465750567616e-05, "loss": 1.1295, "step": 1761 }, { "epoch": 0.15, "learning_rate": 1.926443352865971e-05, "loss": 1.1585, "step": 1762 }, { "epoch": 0.15, "learning_rate": 1.926340060967586e-05, "loss": 1.1333, "step": 1763 }, { "epoch": 0.15, "learning_rate": 1.9262366993693796e-05, "loss": 1.1551, "step": 1764 }, { "epoch": 0.15, "learning_rate": 1.926133268079129e-05, "loss": 1.1349, "step": 1765 }, { "epoch": 0.15, "learning_rate": 1.9260297671046156e-05, "loss": 1.0897, "step": 1766 }, { "epoch": 0.15, "learning_rate": 1.9259261964536282e-05, "loss": 1.1297, "step": 1767 }, { "epoch": 0.15, "learning_rate": 1.9258225561339586e-05, "loss": 1.1734, "step": 1768 }, { "epoch": 0.15, "learning_rate": 1.925718846153406e-05, "loss": 1.1311, "step": 1769 }, { "epoch": 0.15, "learning_rate": 1.9256150665197732e-05, "loss": 1.1044, "step": 1770 }, { "epoch": 0.15, "learning_rate": 1.9255112172408682e-05, "loss": 1.164, "step": 1771 }, { "epoch": 0.15, "learning_rate": 1.9254072983245057e-05, "loss": 1.1293, "step": 1772 }, { "epoch": 0.15, "learning_rate": 1.9253033097785048e-05, "loss": 1.1227, "step": 1773 }, { "epoch": 0.15, "learning_rate": 1.925199251610689e-05, "loss": 1.0524, "step": 1774 }, { "epoch": 0.15, "learning_rate": 1.925095123828889e-05, "loss": 1.1172, "step": 1775 }, { "epoch": 0.15, "learning_rate": 1.9249909264409386e-05, "loss": 1.156, "step": 1776 }, { "epoch": 0.15, "learning_rate": 1.9248866594546783e-05, "loss": 1.1634, "step": 1777 }, { "epoch": 0.15, "learning_rate": 1.9247823228779535e-05, "loss": 1.0121, "step": 1778 }, { "epoch": 0.15, "learning_rate": 1.9246779167186146e-05, "loss": 1.1134, "step": 1779 }, { "epoch": 0.15, "learning_rate": 1.9245734409845173e-05, "loss": 1.1755, "step": 1780 }, { "epoch": 0.15, "learning_rate": 1.924468895683522e-05, "loss": 1.1535, "step": 1781 }, { "epoch": 0.15, "learning_rate": 1.924364280823496e-05, "loss": 1.1014, "step": 1782 }, { "epoch": 0.15, "learning_rate": 1.92425959641231e-05, "loss": 1.0669, "step": 1783 }, { "epoch": 0.15, "learning_rate": 1.9241548424578406e-05, "loss": 1.0907, "step": 1784 }, { "epoch": 0.15, "learning_rate": 1.9240500189679706e-05, "loss": 1.0523, "step": 1785 }, { "epoch": 0.15, "learning_rate": 1.9239451259505864e-05, "loss": 1.1389, "step": 1786 }, { "epoch": 0.15, "learning_rate": 1.9238401634135805e-05, "loss": 1.1255, "step": 1787 }, { "epoch": 0.15, "learning_rate": 1.92373513136485e-05, "loss": 1.1677, "step": 1788 }, { "epoch": 0.15, "learning_rate": 1.9236300298122987e-05, "loss": 1.1461, "step": 1789 }, { "epoch": 0.15, "learning_rate": 1.9235248587638343e-05, "loss": 1.1269, "step": 1790 }, { "epoch": 0.15, "learning_rate": 1.92341961822737e-05, "loss": 1.1524, "step": 1791 }, { "epoch": 0.15, "learning_rate": 1.9233143082108245e-05, "loss": 1.1317, "step": 1792 }, { "epoch": 0.15, "learning_rate": 1.923208928722121e-05, "loss": 1.107, "step": 1793 }, { "epoch": 0.15, "learning_rate": 1.923103479769189e-05, "loss": 1.1248, "step": 1794 }, { "epoch": 0.15, "learning_rate": 1.922997961359963e-05, "loss": 1.1941, "step": 1795 }, { "epoch": 0.15, "learning_rate": 1.9228923735023815e-05, "loss": 1.1947, "step": 1796 }, { "epoch": 0.15, "learning_rate": 1.92278671620439e-05, "loss": 0.8719, "step": 1797 }, { "epoch": 0.15, "learning_rate": 1.9226809894739383e-05, "loss": 1.1476, "step": 1798 }, { "epoch": 0.15, "learning_rate": 1.9225751933189806e-05, "loss": 1.1666, "step": 1799 }, { "epoch": 0.15, "learning_rate": 1.9224693277474787e-05, "loss": 1.1138, "step": 1800 }, { "epoch": 0.15, "learning_rate": 1.9223633927673967e-05, "loss": 1.1467, "step": 1801 }, { "epoch": 0.15, "learning_rate": 1.9222573883867063e-05, "loss": 1.1195, "step": 1802 }, { "epoch": 0.15, "learning_rate": 1.9221513146133835e-05, "loss": 1.0964, "step": 1803 }, { "epoch": 0.15, "learning_rate": 1.922045171455409e-05, "loss": 1.1513, "step": 1804 }, { "epoch": 0.15, "learning_rate": 1.9219389589207693e-05, "loss": 1.1705, "step": 1805 }, { "epoch": 0.15, "learning_rate": 1.9218326770174563e-05, "loss": 1.2125, "step": 1806 }, { "epoch": 0.15, "learning_rate": 1.921726325753467e-05, "loss": 1.1789, "step": 1807 }, { "epoch": 0.15, "learning_rate": 1.9216199051368033e-05, "loss": 1.1009, "step": 1808 }, { "epoch": 0.15, "learning_rate": 1.9215134151754726e-05, "loss": 1.1428, "step": 1809 }, { "epoch": 0.15, "learning_rate": 1.9214068558774874e-05, "loss": 1.1225, "step": 1810 }, { "epoch": 0.15, "learning_rate": 1.9213002272508655e-05, "loss": 1.1705, "step": 1811 }, { "epoch": 0.15, "learning_rate": 1.9211935293036293e-05, "loss": 1.1103, "step": 1812 }, { "epoch": 0.15, "learning_rate": 1.921086762043808e-05, "loss": 1.128, "step": 1813 }, { "epoch": 0.15, "learning_rate": 1.9209799254794342e-05, "loss": 1.1902, "step": 1814 }, { "epoch": 0.15, "learning_rate": 1.920873019618547e-05, "loss": 1.1457, "step": 1815 }, { "epoch": 0.15, "learning_rate": 1.92076604446919e-05, "loss": 1.1575, "step": 1816 }, { "epoch": 0.15, "learning_rate": 1.9206590000394124e-05, "loss": 1.1796, "step": 1817 }, { "epoch": 0.15, "learning_rate": 1.9205518863372677e-05, "loss": 1.1218, "step": 1818 }, { "epoch": 0.15, "learning_rate": 1.9204447033708168e-05, "loss": 1.1866, "step": 1819 }, { "epoch": 0.15, "learning_rate": 1.920337451148123e-05, "loss": 1.1517, "step": 1820 }, { "epoch": 0.15, "learning_rate": 1.9202301296772572e-05, "loss": 1.1643, "step": 1821 }, { "epoch": 0.15, "learning_rate": 1.9201227389662937e-05, "loss": 1.0561, "step": 1822 }, { "epoch": 0.15, "learning_rate": 1.9200152790233132e-05, "loss": 1.0946, "step": 1823 }, { "epoch": 0.15, "learning_rate": 1.919907749856401e-05, "loss": 1.1453, "step": 1824 }, { "epoch": 0.15, "learning_rate": 1.9198001514736485e-05, "loss": 1.1588, "step": 1825 }, { "epoch": 0.15, "learning_rate": 1.9196924838831508e-05, "loss": 1.1356, "step": 1826 }, { "epoch": 0.15, "learning_rate": 1.9195847470930094e-05, "loss": 1.0761, "step": 1827 }, { "epoch": 0.15, "learning_rate": 1.919476941111331e-05, "loss": 1.1139, "step": 1828 }, { "epoch": 0.15, "learning_rate": 1.919369065946226e-05, "loss": 1.1534, "step": 1829 }, { "epoch": 0.15, "learning_rate": 1.9192611216058125e-05, "loss": 1.1128, "step": 1830 }, { "epoch": 0.16, "learning_rate": 1.9191531080982115e-05, "loss": 1.0492, "step": 1831 }, { "epoch": 0.16, "learning_rate": 1.919045025431551e-05, "loss": 1.1623, "step": 1832 }, { "epoch": 0.16, "learning_rate": 1.9189368736139626e-05, "loss": 1.1002, "step": 1833 }, { "epoch": 0.16, "learning_rate": 1.9188286526535845e-05, "loss": 1.0885, "step": 1834 }, { "epoch": 0.16, "learning_rate": 1.9187203625585592e-05, "loss": 1.138, "step": 1835 }, { "epoch": 0.16, "learning_rate": 1.9186120033370344e-05, "loss": 1.1184, "step": 1836 }, { "epoch": 0.16, "learning_rate": 1.9185035749971638e-05, "loss": 0.9383, "step": 1837 }, { "epoch": 0.16, "learning_rate": 1.9183950775471055e-05, "loss": 1.1196, "step": 1838 }, { "epoch": 0.16, "learning_rate": 1.9182865109950233e-05, "loss": 1.0942, "step": 1839 }, { "epoch": 0.16, "learning_rate": 1.9181778753490853e-05, "loss": 1.0812, "step": 1840 }, { "epoch": 0.16, "learning_rate": 1.9180691706174664e-05, "loss": 1.1052, "step": 1841 }, { "epoch": 0.16, "learning_rate": 1.9179603968083452e-05, "loss": 1.0968, "step": 1842 }, { "epoch": 0.16, "learning_rate": 1.9178515539299064e-05, "loss": 1.1064, "step": 1843 }, { "epoch": 0.16, "learning_rate": 1.9177426419903392e-05, "loss": 1.1278, "step": 1844 }, { "epoch": 0.16, "learning_rate": 1.9176336609978386e-05, "loss": 1.0909, "step": 1845 }, { "epoch": 0.16, "learning_rate": 1.9175246109606048e-05, "loss": 1.1187, "step": 1846 }, { "epoch": 0.16, "learning_rate": 1.9174154918868427e-05, "loss": 1.1575, "step": 1847 }, { "epoch": 0.16, "learning_rate": 1.9173063037847624e-05, "loss": 1.0801, "step": 1848 }, { "epoch": 0.16, "learning_rate": 1.91719704666258e-05, "loss": 1.132, "step": 1849 }, { "epoch": 0.16, "learning_rate": 1.9170877205285157e-05, "loss": 1.0813, "step": 1850 }, { "epoch": 0.16, "learning_rate": 1.9169783253907956e-05, "loss": 1.1143, "step": 1851 }, { "epoch": 0.16, "learning_rate": 1.916868861257651e-05, "loss": 1.1272, "step": 1852 }, { "epoch": 0.16, "learning_rate": 1.916759328137318e-05, "loss": 1.1015, "step": 1853 }, { "epoch": 0.16, "learning_rate": 1.9166497260380386e-05, "loss": 1.1549, "step": 1854 }, { "epoch": 0.16, "learning_rate": 1.916540054968059e-05, "loss": 1.1618, "step": 1855 }, { "epoch": 0.16, "learning_rate": 1.916430314935631e-05, "loss": 1.1216, "step": 1856 }, { "epoch": 0.16, "learning_rate": 1.916320505949012e-05, "loss": 1.0975, "step": 1857 }, { "epoch": 0.16, "learning_rate": 1.9162106280164642e-05, "loss": 1.1573, "step": 1858 }, { "epoch": 0.16, "learning_rate": 1.9161006811462552e-05, "loss": 1.1437, "step": 1859 }, { "epoch": 0.16, "learning_rate": 1.9159906653466575e-05, "loss": 1.1364, "step": 1860 }, { "epoch": 0.16, "learning_rate": 1.9158805806259486e-05, "loss": 1.1136, "step": 1861 }, { "epoch": 0.16, "learning_rate": 1.9157704269924117e-05, "loss": 1.0867, "step": 1862 }, { "epoch": 0.16, "learning_rate": 1.915660204454335e-05, "loss": 1.1617, "step": 1863 }, { "epoch": 0.16, "learning_rate": 1.915549913020012e-05, "loss": 1.0656, "step": 1864 }, { "epoch": 0.16, "learning_rate": 1.9154395526977415e-05, "loss": 1.1184, "step": 1865 }, { "epoch": 0.16, "learning_rate": 1.915329123495827e-05, "loss": 1.0905, "step": 1866 }, { "epoch": 0.16, "learning_rate": 1.9152186254225773e-05, "loss": 1.1611, "step": 1867 }, { "epoch": 0.16, "learning_rate": 1.9151080584863066e-05, "loss": 1.1109, "step": 1868 }, { "epoch": 0.16, "learning_rate": 1.9149974226953344e-05, "loss": 1.1183, "step": 1869 }, { "epoch": 0.16, "learning_rate": 1.9148867180579847e-05, "loss": 1.1566, "step": 1870 }, { "epoch": 0.16, "learning_rate": 1.9147759445825876e-05, "loss": 1.1417, "step": 1871 }, { "epoch": 0.16, "learning_rate": 1.914665102277478e-05, "loss": 1.1798, "step": 1872 }, { "epoch": 0.16, "learning_rate": 1.9145541911509956e-05, "loss": 1.0941, "step": 1873 }, { "epoch": 0.16, "learning_rate": 1.9144432112114853e-05, "loss": 1.1422, "step": 1874 }, { "epoch": 0.16, "learning_rate": 1.9143321624672983e-05, "loss": 1.1611, "step": 1875 }, { "epoch": 0.16, "learning_rate": 1.9142210449267897e-05, "loss": 1.15, "step": 1876 }, { "epoch": 0.16, "learning_rate": 1.9141098585983197e-05, "loss": 1.1414, "step": 1877 }, { "epoch": 0.16, "learning_rate": 1.913998603490255e-05, "loss": 1.187, "step": 1878 }, { "epoch": 0.16, "learning_rate": 1.9138872796109666e-05, "loss": 1.118, "step": 1879 }, { "epoch": 0.16, "learning_rate": 1.9137758869688304e-05, "loss": 1.1299, "step": 1880 }, { "epoch": 0.16, "learning_rate": 1.9136644255722282e-05, "loss": 1.1462, "step": 1881 }, { "epoch": 0.16, "learning_rate": 1.913552895429546e-05, "loss": 1.1167, "step": 1882 }, { "epoch": 0.16, "learning_rate": 1.9134412965491762e-05, "loss": 1.1204, "step": 1883 }, { "epoch": 0.16, "learning_rate": 1.9133296289395153e-05, "loss": 1.1271, "step": 1884 }, { "epoch": 0.16, "learning_rate": 1.913217892608966e-05, "loss": 1.1364, "step": 1885 }, { "epoch": 0.16, "learning_rate": 1.9131060875659345e-05, "loss": 1.1365, "step": 1886 }, { "epoch": 0.16, "learning_rate": 1.9129942138188346e-05, "loss": 1.1617, "step": 1887 }, { "epoch": 0.16, "learning_rate": 1.912882271376083e-05, "loss": 1.1661, "step": 1888 }, { "epoch": 0.16, "learning_rate": 1.9127702602461025e-05, "loss": 1.166, "step": 1889 }, { "epoch": 0.16, "learning_rate": 1.9126581804373212e-05, "loss": 1.0864, "step": 1890 }, { "epoch": 0.16, "learning_rate": 1.912546031958173e-05, "loss": 1.1151, "step": 1891 }, { "epoch": 0.16, "learning_rate": 1.9124338148170948e-05, "loss": 1.1875, "step": 1892 }, { "epoch": 0.16, "learning_rate": 1.9123215290225314e-05, "loss": 1.0778, "step": 1893 }, { "epoch": 0.16, "learning_rate": 1.91220917458293e-05, "loss": 1.1158, "step": 1894 }, { "epoch": 0.16, "learning_rate": 1.912096751506746e-05, "loss": 1.1679, "step": 1895 }, { "epoch": 0.16, "learning_rate": 1.911984259802437e-05, "loss": 1.0228, "step": 1896 }, { "epoch": 0.16, "learning_rate": 1.911871699478468e-05, "loss": 1.1297, "step": 1897 }, { "epoch": 0.16, "learning_rate": 1.9117590705433078e-05, "loss": 1.1101, "step": 1898 }, { "epoch": 0.16, "learning_rate": 1.911646373005431e-05, "loss": 1.1476, "step": 1899 }, { "epoch": 0.16, "learning_rate": 1.911533606873317e-05, "loss": 1.0674, "step": 1900 }, { "epoch": 0.16, "learning_rate": 1.911420772155451e-05, "loss": 1.1408, "step": 1901 }, { "epoch": 0.16, "learning_rate": 1.9113078688603225e-05, "loss": 1.1276, "step": 1902 }, { "epoch": 0.16, "learning_rate": 1.9111948969964272e-05, "loss": 1.1179, "step": 1903 }, { "epoch": 0.16, "learning_rate": 1.9110818565722646e-05, "loss": 1.1981, "step": 1904 }, { "epoch": 0.16, "learning_rate": 1.91096874759634e-05, "loss": 1.0874, "step": 1905 }, { "epoch": 0.16, "learning_rate": 1.910855570077165e-05, "loss": 1.124, "step": 1906 }, { "epoch": 0.16, "learning_rate": 1.9107423240232546e-05, "loss": 1.1133, "step": 1907 }, { "epoch": 0.16, "learning_rate": 1.91062900944313e-05, "loss": 1.1566, "step": 1908 }, { "epoch": 0.16, "learning_rate": 1.9105156263453163e-05, "loss": 1.1462, "step": 1909 }, { "epoch": 0.16, "learning_rate": 1.910402174738346e-05, "loss": 1.158, "step": 1910 }, { "epoch": 0.16, "learning_rate": 1.9102886546307545e-05, "loss": 1.1646, "step": 1911 }, { "epoch": 0.16, "learning_rate": 1.910175066031084e-05, "loss": 1.0597, "step": 1912 }, { "epoch": 0.16, "learning_rate": 1.9100614089478805e-05, "loss": 1.1147, "step": 1913 }, { "epoch": 0.16, "learning_rate": 1.9099476833896964e-05, "loss": 1.1715, "step": 1914 }, { "epoch": 0.16, "learning_rate": 1.9098338893650883e-05, "loss": 1.1118, "step": 1915 }, { "epoch": 0.16, "learning_rate": 1.909720026882618e-05, "loss": 1.1635, "step": 1916 }, { "epoch": 0.16, "learning_rate": 1.9096060959508534e-05, "loss": 1.1261, "step": 1917 }, { "epoch": 0.16, "learning_rate": 1.9094920965783667e-05, "loss": 1.169, "step": 1918 }, { "epoch": 0.16, "learning_rate": 1.9093780287737355e-05, "loss": 1.0663, "step": 1919 }, { "epoch": 0.16, "learning_rate": 1.909263892545542e-05, "loss": 1.072, "step": 1920 }, { "epoch": 0.16, "learning_rate": 1.9091496879023746e-05, "loss": 1.1515, "step": 1921 }, { "epoch": 0.16, "learning_rate": 1.9090354148528264e-05, "loss": 1.1453, "step": 1922 }, { "epoch": 0.16, "learning_rate": 1.908921073405495e-05, "loss": 1.1126, "step": 1923 }, { "epoch": 0.16, "learning_rate": 1.9088066635689843e-05, "loss": 1.1773, "step": 1924 }, { "epoch": 0.16, "learning_rate": 1.9086921853519023e-05, "loss": 1.0953, "step": 1925 }, { "epoch": 0.16, "learning_rate": 1.9085776387628627e-05, "loss": 1.1376, "step": 1926 }, { "epoch": 0.16, "learning_rate": 1.9084630238104843e-05, "loss": 1.1147, "step": 1927 }, { "epoch": 0.16, "learning_rate": 1.9083483405033908e-05, "loss": 1.0944, "step": 1928 }, { "epoch": 0.16, "learning_rate": 1.9082335888502116e-05, "loss": 1.1454, "step": 1929 }, { "epoch": 0.16, "learning_rate": 1.90811876885958e-05, "loss": 1.172, "step": 1930 }, { "epoch": 0.16, "learning_rate": 1.9080038805401367e-05, "loss": 1.1262, "step": 1931 }, { "epoch": 0.16, "learning_rate": 1.9078889239005248e-05, "loss": 1.108, "step": 1932 }, { "epoch": 0.16, "learning_rate": 1.907773898949395e-05, "loss": 1.0883, "step": 1933 }, { "epoch": 0.16, "learning_rate": 1.9076588056954014e-05, "loss": 1.1227, "step": 1934 }, { "epoch": 0.16, "learning_rate": 1.9075436441472036e-05, "loss": 1.1233, "step": 1935 }, { "epoch": 0.16, "learning_rate": 1.907428414313467e-05, "loss": 1.0688, "step": 1936 }, { "epoch": 0.16, "learning_rate": 1.907313116202862e-05, "loss": 1.1522, "step": 1937 }, { "epoch": 0.16, "learning_rate": 1.9071977498240636e-05, "loss": 1.0737, "step": 1938 }, { "epoch": 0.16, "learning_rate": 1.9070823151857517e-05, "loss": 1.1243, "step": 1939 }, { "epoch": 0.16, "learning_rate": 1.9069668122966128e-05, "loss": 1.1011, "step": 1940 }, { "epoch": 0.16, "learning_rate": 1.906851241165337e-05, "loss": 1.1507, "step": 1941 }, { "epoch": 0.16, "learning_rate": 1.90673560180062e-05, "loss": 1.1392, "step": 1942 }, { "epoch": 0.16, "learning_rate": 1.906619894211163e-05, "loss": 1.1208, "step": 1943 }, { "epoch": 0.16, "learning_rate": 1.9065041184056726e-05, "loss": 1.1323, "step": 1944 }, { "epoch": 0.16, "learning_rate": 1.906388274392859e-05, "loss": 1.1141, "step": 1945 }, { "epoch": 0.16, "learning_rate": 1.906272362181439e-05, "loss": 1.1319, "step": 1946 }, { "epoch": 0.16, "learning_rate": 1.9061563817801345e-05, "loss": 1.1202, "step": 1947 }, { "epoch": 0.16, "learning_rate": 1.9060403331976715e-05, "loss": 1.1141, "step": 1948 }, { "epoch": 0.17, "learning_rate": 1.905924216442782e-05, "loss": 1.1223, "step": 1949 }, { "epoch": 0.17, "learning_rate": 1.9058080315242032e-05, "loss": 1.1526, "step": 1950 }, { "epoch": 0.17, "learning_rate": 1.9056917784506764e-05, "loss": 1.1238, "step": 1951 }, { "epoch": 0.17, "learning_rate": 1.9055754572309495e-05, "loss": 1.107, "step": 1952 }, { "epoch": 0.17, "learning_rate": 1.9054590678737744e-05, "loss": 1.108, "step": 1953 }, { "epoch": 0.17, "learning_rate": 1.905342610387908e-05, "loss": 1.1321, "step": 1954 }, { "epoch": 0.17, "learning_rate": 1.9052260847821133e-05, "loss": 1.1108, "step": 1955 }, { "epoch": 0.17, "learning_rate": 1.9051094910651584e-05, "loss": 0.9968, "step": 1956 }, { "epoch": 0.17, "learning_rate": 1.9049928292458154e-05, "loss": 1.1276, "step": 1957 }, { "epoch": 0.17, "learning_rate": 1.904876099332862e-05, "loss": 1.1402, "step": 1958 }, { "epoch": 0.17, "learning_rate": 1.904759301335082e-05, "loss": 1.1504, "step": 1959 }, { "epoch": 0.17, "learning_rate": 1.9046424352612626e-05, "loss": 1.1301, "step": 1960 }, { "epoch": 0.17, "learning_rate": 1.904525501120198e-05, "loss": 1.1874, "step": 1961 }, { "epoch": 0.17, "learning_rate": 1.904408498920686e-05, "loss": 1.1355, "step": 1962 }, { "epoch": 0.17, "learning_rate": 1.90429142867153e-05, "loss": 1.1112, "step": 1963 }, { "epoch": 0.17, "learning_rate": 1.9041742903815392e-05, "loss": 1.1056, "step": 1964 }, { "epoch": 0.17, "learning_rate": 1.904057084059527e-05, "loss": 1.0899, "step": 1965 }, { "epoch": 0.17, "learning_rate": 1.903939809714312e-05, "loss": 1.1636, "step": 1966 }, { "epoch": 0.17, "learning_rate": 1.9038224673547185e-05, "loss": 1.111, "step": 1967 }, { "epoch": 0.17, "learning_rate": 1.9037050569895755e-05, "loss": 1.1134, "step": 1968 }, { "epoch": 0.17, "learning_rate": 1.903587578627717e-05, "loss": 1.0987, "step": 1969 }, { "epoch": 0.17, "learning_rate": 1.903470032277983e-05, "loss": 1.1269, "step": 1970 }, { "epoch": 0.17, "learning_rate": 1.9033524179492172e-05, "loss": 1.1717, "step": 1971 }, { "epoch": 0.17, "learning_rate": 1.9032347356502696e-05, "loss": 1.1354, "step": 1972 }, { "epoch": 0.17, "learning_rate": 1.9031169853899945e-05, "loss": 1.1026, "step": 1973 }, { "epoch": 0.17, "learning_rate": 1.9029991671772518e-05, "loss": 1.1167, "step": 1974 }, { "epoch": 0.17, "learning_rate": 1.9028812810209067e-05, "loss": 1.1573, "step": 1975 }, { "epoch": 0.17, "learning_rate": 1.902763326929829e-05, "loss": 1.1136, "step": 1976 }, { "epoch": 0.17, "learning_rate": 1.902645304912894e-05, "loss": 1.116, "step": 1977 }, { "epoch": 0.17, "learning_rate": 1.9025272149789814e-05, "loss": 1.1103, "step": 1978 }, { "epoch": 0.17, "learning_rate": 1.902409057136977e-05, "loss": 1.1399, "step": 1979 }, { "epoch": 0.17, "learning_rate": 1.902290831395771e-05, "loss": 1.1084, "step": 1980 }, { "epoch": 0.17, "learning_rate": 1.9021725377642592e-05, "loss": 1.122, "step": 1981 }, { "epoch": 0.17, "learning_rate": 1.9020541762513422e-05, "loss": 1.1225, "step": 1982 }, { "epoch": 0.17, "learning_rate": 1.901935746865926e-05, "loss": 1.1426, "step": 1983 }, { "epoch": 0.17, "learning_rate": 1.901817249616921e-05, "loss": 1.1425, "step": 1984 }, { "epoch": 0.17, "learning_rate": 1.9016986845132432e-05, "loss": 1.1104, "step": 1985 }, { "epoch": 0.17, "learning_rate": 1.9015800515638142e-05, "loss": 1.145, "step": 1986 }, { "epoch": 0.17, "learning_rate": 1.9014613507775596e-05, "loss": 1.0724, "step": 1987 }, { "epoch": 0.17, "learning_rate": 1.9013425821634114e-05, "loss": 1.1067, "step": 1988 }, { "epoch": 0.17, "learning_rate": 1.9012237457303055e-05, "loss": 1.1653, "step": 1989 }, { "epoch": 0.17, "learning_rate": 1.9011048414871834e-05, "loss": 1.1227, "step": 1990 }, { "epoch": 0.17, "learning_rate": 1.9009858694429917e-05, "loss": 1.1024, "step": 1991 }, { "epoch": 0.17, "learning_rate": 1.9008668296066823e-05, "loss": 1.0855, "step": 1992 }, { "epoch": 0.17, "learning_rate": 1.9007477219872122e-05, "loss": 1.1674, "step": 1993 }, { "epoch": 0.17, "learning_rate": 1.9006285465935426e-05, "loss": 1.0919, "step": 1994 }, { "epoch": 0.17, "learning_rate": 1.9005093034346416e-05, "loss": 1.1059, "step": 1995 }, { "epoch": 0.17, "learning_rate": 1.9003899925194803e-05, "loss": 1.1091, "step": 1996 }, { "epoch": 0.17, "learning_rate": 1.9002706138570365e-05, "loss": 1.1007, "step": 1997 }, { "epoch": 0.17, "learning_rate": 1.9001511674562922e-05, "loss": 1.1736, "step": 1998 }, { "epoch": 0.17, "learning_rate": 1.9000316533262347e-05, "loss": 1.1038, "step": 1999 }, { "epoch": 0.17, "learning_rate": 1.899912071475857e-05, "loss": 1.2075, "step": 2000 }, { "epoch": 0.17, "learning_rate": 1.8997924219141564e-05, "loss": 1.0326, "step": 2001 }, { "epoch": 0.17, "learning_rate": 1.899672704650136e-05, "loss": 1.1331, "step": 2002 }, { "epoch": 0.17, "learning_rate": 1.8995529196928025e-05, "loss": 1.1119, "step": 2003 }, { "epoch": 0.17, "learning_rate": 1.89943306705117e-05, "loss": 1.1623, "step": 2004 }, { "epoch": 0.17, "learning_rate": 1.8993131467342556e-05, "loss": 1.1221, "step": 2005 }, { "epoch": 0.17, "learning_rate": 1.899193158751083e-05, "loss": 1.1615, "step": 2006 }, { "epoch": 0.17, "learning_rate": 1.8990731031106796e-05, "loss": 1.1477, "step": 2007 }, { "epoch": 0.17, "learning_rate": 1.8989529798220794e-05, "loss": 1.1262, "step": 2008 }, { "epoch": 0.17, "learning_rate": 1.8988327888943202e-05, "loss": 1.1113, "step": 2009 }, { "epoch": 0.17, "learning_rate": 1.8987125303364462e-05, "loss": 1.1213, "step": 2010 }, { "epoch": 0.17, "learning_rate": 1.8985922041575048e-05, "loss": 1.1624, "step": 2011 }, { "epoch": 0.17, "learning_rate": 1.8984718103665507e-05, "loss": 1.1403, "step": 2012 }, { "epoch": 0.17, "learning_rate": 1.8983513489726418e-05, "loss": 1.11, "step": 2013 }, { "epoch": 0.17, "learning_rate": 1.898230819984842e-05, "loss": 1.1572, "step": 2014 }, { "epoch": 0.17, "learning_rate": 1.8981102234122203e-05, "loss": 1.17, "step": 2015 }, { "epoch": 0.17, "learning_rate": 1.8979895592638508e-05, "loss": 1.1136, "step": 2016 }, { "epoch": 0.17, "learning_rate": 1.8978688275488127e-05, "loss": 1.14, "step": 2017 }, { "epoch": 0.17, "learning_rate": 1.8977480282761893e-05, "loss": 1.1693, "step": 2018 }, { "epoch": 0.17, "learning_rate": 1.89762716145507e-05, "loss": 1.1719, "step": 2019 }, { "epoch": 0.17, "learning_rate": 1.8975062270945504e-05, "loss": 1.1327, "step": 2020 }, { "epoch": 0.17, "learning_rate": 1.8973852252037283e-05, "loss": 1.0962, "step": 2021 }, { "epoch": 0.17, "learning_rate": 1.8972641557917083e-05, "loss": 1.12, "step": 2022 }, { "epoch": 0.17, "learning_rate": 1.8971430188676008e-05, "loss": 1.1584, "step": 2023 }, { "epoch": 0.17, "learning_rate": 1.89702181444052e-05, "loss": 1.121, "step": 2024 }, { "epoch": 0.17, "learning_rate": 1.8969005425195846e-05, "loss": 1.0825, "step": 2025 }, { "epoch": 0.17, "learning_rate": 1.896779203113921e-05, "loss": 1.0886, "step": 2026 }, { "epoch": 0.17, "learning_rate": 1.896657796232658e-05, "loss": 1.1261, "step": 2027 }, { "epoch": 0.17, "learning_rate": 1.896536321884931e-05, "loss": 1.1474, "step": 2028 }, { "epoch": 0.17, "learning_rate": 1.8964147800798798e-05, "loss": 1.1114, "step": 2029 }, { "epoch": 0.17, "learning_rate": 1.8962931708266493e-05, "loss": 1.1436, "step": 2030 }, { "epoch": 0.17, "learning_rate": 1.89617149413439e-05, "loss": 1.104, "step": 2031 }, { "epoch": 0.17, "learning_rate": 1.8960497500122565e-05, "loss": 1.1278, "step": 2032 }, { "epoch": 0.17, "learning_rate": 1.89592793846941e-05, "loss": 1.1954, "step": 2033 }, { "epoch": 0.17, "learning_rate": 1.895806059515015e-05, "loss": 1.1125, "step": 2034 }, { "epoch": 0.17, "learning_rate": 1.8956841131582425e-05, "loss": 1.0609, "step": 2035 }, { "epoch": 0.17, "learning_rate": 1.8955620994082678e-05, "loss": 0.9914, "step": 2036 }, { "epoch": 0.17, "learning_rate": 1.8954400182742713e-05, "loss": 1.1104, "step": 2037 }, { "epoch": 0.17, "learning_rate": 1.8953178697654394e-05, "loss": 1.1343, "step": 2038 }, { "epoch": 0.17, "learning_rate": 1.8951956538909616e-05, "loss": 1.107, "step": 2039 }, { "epoch": 0.17, "learning_rate": 1.8950733706600345e-05, "loss": 1.2076, "step": 2040 }, { "epoch": 0.17, "learning_rate": 1.894951020081859e-05, "loss": 1.1535, "step": 2041 }, { "epoch": 0.17, "learning_rate": 1.8948286021656404e-05, "loss": 1.1506, "step": 2042 }, { "epoch": 0.17, "learning_rate": 1.8947061169205904e-05, "loss": 1.1344, "step": 2043 }, { "epoch": 0.17, "learning_rate": 1.8945835643559248e-05, "loss": 1.118, "step": 2044 }, { "epoch": 0.17, "learning_rate": 1.894460944480865e-05, "loss": 1.161, "step": 2045 }, { "epoch": 0.17, "learning_rate": 1.8943382573046364e-05, "loss": 1.1485, "step": 2046 }, { "epoch": 0.17, "learning_rate": 1.8942155028364704e-05, "loss": 1.1776, "step": 2047 }, { "epoch": 0.17, "learning_rate": 1.8940926810856043e-05, "loss": 1.147, "step": 2048 }, { "epoch": 0.17, "learning_rate": 1.8939697920612784e-05, "loss": 1.1083, "step": 2049 }, { "epoch": 0.17, "learning_rate": 1.893846835772739e-05, "loss": 1.112, "step": 2050 }, { "epoch": 0.17, "learning_rate": 1.893723812229239e-05, "loss": 1.1511, "step": 2051 }, { "epoch": 0.17, "learning_rate": 1.8936007214400338e-05, "loss": 1.0972, "step": 2052 }, { "epoch": 0.17, "learning_rate": 1.8934775634143852e-05, "loss": 1.128, "step": 2053 }, { "epoch": 0.17, "learning_rate": 1.89335433816156e-05, "loss": 1.1238, "step": 2054 }, { "epoch": 0.17, "learning_rate": 1.89323104569083e-05, "loss": 1.126, "step": 2055 }, { "epoch": 0.17, "learning_rate": 1.8931076860114717e-05, "loss": 1.1774, "step": 2056 }, { "epoch": 0.17, "learning_rate": 1.892984259132767e-05, "loss": 1.0688, "step": 2057 }, { "epoch": 0.17, "learning_rate": 1.8928607650640037e-05, "loss": 1.1198, "step": 2058 }, { "epoch": 0.17, "learning_rate": 1.8927372038144724e-05, "loss": 1.131, "step": 2059 }, { "epoch": 0.17, "learning_rate": 1.8926135753934706e-05, "loss": 1.1733, "step": 2060 }, { "epoch": 0.17, "learning_rate": 1.8924898798103005e-05, "loss": 1.1498, "step": 2061 }, { "epoch": 0.17, "learning_rate": 1.8923661170742696e-05, "loss": 1.1611, "step": 2062 }, { "epoch": 0.17, "learning_rate": 1.8922422871946894e-05, "loss": 1.1657, "step": 2063 }, { "epoch": 0.17, "learning_rate": 1.8921183901808773e-05, "loss": 1.1089, "step": 2064 }, { "epoch": 0.17, "learning_rate": 1.8919944260421562e-05, "loss": 1.171, "step": 2065 }, { "epoch": 0.17, "learning_rate": 1.891870394787853e-05, "loss": 1.1036, "step": 2066 }, { "epoch": 0.18, "learning_rate": 1.8917462964272992e-05, "loss": 1.0929, "step": 2067 }, { "epoch": 0.18, "learning_rate": 1.8916221309698333e-05, "loss": 1.1381, "step": 2068 }, { "epoch": 0.18, "learning_rate": 1.8914978984247976e-05, "loss": 1.1698, "step": 2069 }, { "epoch": 0.18, "learning_rate": 1.8913735988015395e-05, "loss": 1.1155, "step": 2070 }, { "epoch": 0.18, "learning_rate": 1.8912492321094117e-05, "loss": 1.0927, "step": 2071 }, { "epoch": 0.18, "learning_rate": 1.8911247983577716e-05, "loss": 1.1304, "step": 2072 }, { "epoch": 0.18, "learning_rate": 1.891000297555982e-05, "loss": 1.1472, "step": 2073 }, { "epoch": 0.18, "learning_rate": 1.8908757297134106e-05, "loss": 1.1138, "step": 2074 }, { "epoch": 0.18, "learning_rate": 1.8907510948394302e-05, "loss": 1.135, "step": 2075 }, { "epoch": 0.18, "learning_rate": 1.8906263929434187e-05, "loss": 1.1113, "step": 2076 }, { "epoch": 0.18, "learning_rate": 1.890501624034758e-05, "loss": 1.1147, "step": 2077 }, { "epoch": 0.18, "learning_rate": 1.8903767881228375e-05, "loss": 1.114, "step": 2078 }, { "epoch": 0.18, "learning_rate": 1.8902518852170493e-05, "loss": 1.1376, "step": 2079 }, { "epoch": 0.18, "learning_rate": 1.8901269153267915e-05, "loss": 1.1527, "step": 2080 }, { "epoch": 0.18, "learning_rate": 1.8900018784614667e-05, "loss": 1.1014, "step": 2081 }, { "epoch": 0.18, "learning_rate": 1.8898767746304835e-05, "loss": 1.1405, "step": 2082 }, { "epoch": 0.18, "learning_rate": 1.8897516038432545e-05, "loss": 1.1043, "step": 2083 }, { "epoch": 0.18, "learning_rate": 1.8896263661091986e-05, "loss": 1.1189, "step": 2084 }, { "epoch": 0.18, "learning_rate": 1.889501061437738e-05, "loss": 1.1077, "step": 2085 }, { "epoch": 0.18, "learning_rate": 1.8893756898383017e-05, "loss": 1.1377, "step": 2086 }, { "epoch": 0.18, "learning_rate": 1.889250251320322e-05, "loss": 1.1068, "step": 2087 }, { "epoch": 0.18, "learning_rate": 1.8891247458932383e-05, "loss": 1.0908, "step": 2088 }, { "epoch": 0.18, "learning_rate": 1.888999173566493e-05, "loss": 1.1126, "step": 2089 }, { "epoch": 0.18, "learning_rate": 1.888873534349535e-05, "loss": 1.1341, "step": 2090 }, { "epoch": 0.18, "learning_rate": 1.8887478282518175e-05, "loss": 1.1133, "step": 2091 }, { "epoch": 0.18, "learning_rate": 1.8886220552827986e-05, "loss": 1.1224, "step": 2092 }, { "epoch": 0.18, "learning_rate": 1.888496215451942e-05, "loss": 1.0578, "step": 2093 }, { "epoch": 0.18, "learning_rate": 1.8883703087687163e-05, "loss": 1.1735, "step": 2094 }, { "epoch": 0.18, "learning_rate": 1.8882443352425945e-05, "loss": 1.1732, "step": 2095 }, { "epoch": 0.18, "learning_rate": 1.8881182948830555e-05, "loss": 1.176, "step": 2096 }, { "epoch": 0.18, "learning_rate": 1.887992187699583e-05, "loss": 1.1703, "step": 2097 }, { "epoch": 0.18, "learning_rate": 1.8878660137016653e-05, "loss": 1.142, "step": 2098 }, { "epoch": 0.18, "learning_rate": 1.8877397728987957e-05, "loss": 1.1029, "step": 2099 }, { "epoch": 0.18, "learning_rate": 1.8876134653004733e-05, "loss": 1.1377, "step": 2100 }, { "epoch": 0.18, "learning_rate": 1.887487090916202e-05, "loss": 1.1209, "step": 2101 }, { "epoch": 0.18, "learning_rate": 1.8873606497554897e-05, "loss": 1.1118, "step": 2102 }, { "epoch": 0.18, "learning_rate": 1.8872341418278508e-05, "loss": 1.1335, "step": 2103 }, { "epoch": 0.18, "learning_rate": 1.8871075671428033e-05, "loss": 1.1387, "step": 2104 }, { "epoch": 0.18, "learning_rate": 1.886980925709872e-05, "loss": 1.123, "step": 2105 }, { "epoch": 0.18, "learning_rate": 1.8868542175385848e-05, "loss": 1.1704, "step": 2106 }, { "epoch": 0.18, "learning_rate": 1.8867274426384758e-05, "loss": 1.1403, "step": 2107 }, { "epoch": 0.18, "learning_rate": 1.886600601019084e-05, "loss": 1.1394, "step": 2108 }, { "epoch": 0.18, "learning_rate": 1.886473692689953e-05, "loss": 1.1177, "step": 2109 }, { "epoch": 0.18, "learning_rate": 1.8863467176606315e-05, "loss": 1.1442, "step": 2110 }, { "epoch": 0.18, "learning_rate": 1.8862196759406734e-05, "loss": 1.1552, "step": 2111 }, { "epoch": 0.18, "learning_rate": 1.8860925675396384e-05, "loss": 1.1434, "step": 2112 }, { "epoch": 0.18, "learning_rate": 1.885965392467089e-05, "loss": 1.1938, "step": 2113 }, { "epoch": 0.18, "learning_rate": 1.8858381507325952e-05, "loss": 1.1099, "step": 2114 }, { "epoch": 0.18, "learning_rate": 1.8857108423457308e-05, "loss": 1.1131, "step": 2115 }, { "epoch": 0.18, "learning_rate": 1.8855834673160742e-05, "loss": 1.0851, "step": 2116 }, { "epoch": 0.18, "learning_rate": 1.8854560256532098e-05, "loss": 1.1422, "step": 2117 }, { "epoch": 0.18, "learning_rate": 1.8853285173667267e-05, "loss": 1.1428, "step": 2118 }, { "epoch": 0.18, "learning_rate": 1.885200942466219e-05, "loss": 1.1006, "step": 2119 }, { "epoch": 0.18, "learning_rate": 1.8850733009612854e-05, "loss": 1.1116, "step": 2120 }, { "epoch": 0.18, "learning_rate": 1.8849455928615298e-05, "loss": 1.1064, "step": 2121 }, { "epoch": 0.18, "learning_rate": 1.8848178181765613e-05, "loss": 1.1306, "step": 2122 }, { "epoch": 0.18, "learning_rate": 1.8846899769159945e-05, "loss": 1.1696, "step": 2123 }, { "epoch": 0.18, "learning_rate": 1.8845620690894474e-05, "loss": 1.1281, "step": 2124 }, { "epoch": 0.18, "learning_rate": 1.8844340947065452e-05, "loss": 1.1112, "step": 2125 }, { "epoch": 0.18, "learning_rate": 1.884306053776916e-05, "loss": 1.1252, "step": 2126 }, { "epoch": 0.18, "learning_rate": 1.8841779463101944e-05, "loss": 1.1313, "step": 2127 }, { "epoch": 0.18, "learning_rate": 1.8840497723160195e-05, "loss": 1.1645, "step": 2128 }, { "epoch": 0.18, "learning_rate": 1.883921531804035e-05, "loss": 1.1583, "step": 2129 }, { "epoch": 0.18, "learning_rate": 1.883793224783891e-05, "loss": 1.039, "step": 2130 }, { "epoch": 0.18, "learning_rate": 1.8836648512652403e-05, "loss": 1.0897, "step": 2131 }, { "epoch": 0.18, "learning_rate": 1.8835364112577425e-05, "loss": 1.1033, "step": 2132 }, { "epoch": 0.18, "learning_rate": 1.8834079047710617e-05, "loss": 1.116, "step": 2133 }, { "epoch": 0.18, "learning_rate": 1.8832793318148676e-05, "loss": 1.1649, "step": 2134 }, { "epoch": 0.18, "learning_rate": 1.8831506923988337e-05, "loss": 1.0853, "step": 2135 }, { "epoch": 0.18, "learning_rate": 1.8830219865326386e-05, "loss": 1.1082, "step": 2136 }, { "epoch": 0.18, "learning_rate": 1.8828932142259674e-05, "loss": 1.1359, "step": 2137 }, { "epoch": 0.18, "learning_rate": 1.8827643754885087e-05, "loss": 1.112, "step": 2138 }, { "epoch": 0.18, "learning_rate": 1.882635470329957e-05, "loss": 1.1237, "step": 2139 }, { "epoch": 0.18, "learning_rate": 1.8825064987600106e-05, "loss": 1.0924, "step": 2140 }, { "epoch": 0.18, "learning_rate": 1.882377460788375e-05, "loss": 1.1351, "step": 2141 }, { "epoch": 0.18, "learning_rate": 1.8822483564247574e-05, "loss": 1.1337, "step": 2142 }, { "epoch": 0.18, "learning_rate": 1.882119185678873e-05, "loss": 1.1181, "step": 2143 }, { "epoch": 0.18, "learning_rate": 1.8819899485604415e-05, "loss": 1.0778, "step": 2144 }, { "epoch": 0.18, "learning_rate": 1.8818606450791856e-05, "loss": 1.111, "step": 2145 }, { "epoch": 0.18, "learning_rate": 1.8817312752448352e-05, "loss": 1.1726, "step": 2146 }, { "epoch": 0.18, "learning_rate": 1.8816018390671243e-05, "loss": 1.1611, "step": 2147 }, { "epoch": 0.18, "learning_rate": 1.8814723365557916e-05, "loss": 1.0805, "step": 2148 }, { "epoch": 0.18, "learning_rate": 1.8813427677205815e-05, "loss": 1.109, "step": 2149 }, { "epoch": 0.18, "learning_rate": 1.881213132571243e-05, "loss": 1.0777, "step": 2150 }, { "epoch": 0.18, "learning_rate": 1.8810834311175303e-05, "loss": 1.139, "step": 2151 }, { "epoch": 0.18, "learning_rate": 1.8809536633692018e-05, "loss": 1.0711, "step": 2152 }, { "epoch": 0.18, "learning_rate": 1.880823829336022e-05, "loss": 1.0785, "step": 2153 }, { "epoch": 0.18, "learning_rate": 1.8806939290277596e-05, "loss": 1.1664, "step": 2154 }, { "epoch": 0.18, "learning_rate": 1.880563962454189e-05, "loss": 1.1876, "step": 2155 }, { "epoch": 0.18, "learning_rate": 1.880433929625089e-05, "loss": 1.1024, "step": 2156 }, { "epoch": 0.18, "learning_rate": 1.8803038305502433e-05, "loss": 1.1235, "step": 2157 }, { "epoch": 0.18, "learning_rate": 1.880173665239441e-05, "loss": 1.112, "step": 2158 }, { "epoch": 0.18, "learning_rate": 1.880043433702476e-05, "loss": 1.086, "step": 2159 }, { "epoch": 0.18, "learning_rate": 1.8799131359491475e-05, "loss": 1.1165, "step": 2160 }, { "epoch": 0.18, "learning_rate": 1.879782771989259e-05, "loss": 1.1083, "step": 2161 }, { "epoch": 0.18, "learning_rate": 1.8796523418326193e-05, "loss": 1.1149, "step": 2162 }, { "epoch": 0.18, "learning_rate": 1.8795218454890423e-05, "loss": 1.1804, "step": 2163 }, { "epoch": 0.18, "learning_rate": 1.8793912829683472e-05, "loss": 1.1097, "step": 2164 }, { "epoch": 0.18, "learning_rate": 1.8792606542803575e-05, "loss": 1.1069, "step": 2165 }, { "epoch": 0.18, "learning_rate": 1.879129959434902e-05, "loss": 1.1767, "step": 2166 }, { "epoch": 0.18, "learning_rate": 1.8789991984418144e-05, "loss": 1.0977, "step": 2167 }, { "epoch": 0.18, "learning_rate": 1.8788683713109336e-05, "loss": 1.0412, "step": 2168 }, { "epoch": 0.18, "learning_rate": 1.8787374780521035e-05, "loss": 1.0793, "step": 2169 }, { "epoch": 0.18, "learning_rate": 1.8786065186751724e-05, "loss": 1.093, "step": 2170 }, { "epoch": 0.18, "learning_rate": 1.8784754931899937e-05, "loss": 1.1716, "step": 2171 }, { "epoch": 0.18, "learning_rate": 1.878344401606427e-05, "loss": 1.1319, "step": 2172 }, { "epoch": 0.18, "learning_rate": 1.8782132439343354e-05, "loss": 1.1139, "step": 2173 }, { "epoch": 0.18, "learning_rate": 1.878082020183587e-05, "loss": 1.13, "step": 2174 }, { "epoch": 0.18, "learning_rate": 1.8779507303640557e-05, "loss": 1.1465, "step": 2175 }, { "epoch": 0.18, "learning_rate": 1.8778193744856207e-05, "loss": 1.0733, "step": 2176 }, { "epoch": 0.18, "learning_rate": 1.877687952558165e-05, "loss": 1.2, "step": 2177 }, { "epoch": 0.18, "learning_rate": 1.8775564645915764e-05, "loss": 0.9625, "step": 2178 }, { "epoch": 0.18, "learning_rate": 1.877424910595749e-05, "loss": 1.0999, "step": 2179 }, { "epoch": 0.18, "learning_rate": 1.8772932905805814e-05, "loss": 1.0865, "step": 2180 }, { "epoch": 0.18, "learning_rate": 1.8771616045559766e-05, "loss": 1.0746, "step": 2181 }, { "epoch": 0.18, "learning_rate": 1.8770298525318433e-05, "loss": 1.1276, "step": 2182 }, { "epoch": 0.18, "learning_rate": 1.8768980345180944e-05, "loss": 1.1296, "step": 2183 }, { "epoch": 0.18, "learning_rate": 1.8767661505246483e-05, "loss": 1.1882, "step": 2184 }, { "epoch": 0.19, "learning_rate": 1.8766342005614282e-05, "loss": 1.0855, "step": 2185 }, { "epoch": 0.19, "learning_rate": 1.8765021846383627e-05, "loss": 1.1466, "step": 2186 }, { "epoch": 0.19, "learning_rate": 1.8763701027653844e-05, "loss": 1.0719, "step": 2187 }, { "epoch": 0.19, "learning_rate": 1.8762379549524318e-05, "loss": 1.1105, "step": 2188 }, { "epoch": 0.19, "learning_rate": 1.8761057412094476e-05, "loss": 1.1232, "step": 2189 }, { "epoch": 0.19, "learning_rate": 1.87597346154638e-05, "loss": 1.1504, "step": 2190 }, { "epoch": 0.19, "learning_rate": 1.8758411159731823e-05, "loss": 1.0913, "step": 2191 }, { "epoch": 0.19, "learning_rate": 1.8757087044998122e-05, "loss": 1.1721, "step": 2192 }, { "epoch": 0.19, "learning_rate": 1.8755762271362326e-05, "loss": 1.1115, "step": 2193 }, { "epoch": 0.19, "learning_rate": 1.875443683892412e-05, "loss": 1.0688, "step": 2194 }, { "epoch": 0.19, "learning_rate": 1.8753110747783223e-05, "loss": 1.1574, "step": 2195 }, { "epoch": 0.19, "learning_rate": 1.8751783998039415e-05, "loss": 1.0872, "step": 2196 }, { "epoch": 0.19, "learning_rate": 1.875045658979253e-05, "loss": 1.104, "step": 2197 }, { "epoch": 0.19, "learning_rate": 1.874912852314244e-05, "loss": 1.1638, "step": 2198 }, { "epoch": 0.19, "learning_rate": 1.8747799798189068e-05, "loss": 1.1177, "step": 2199 }, { "epoch": 0.19, "learning_rate": 1.8746470415032397e-05, "loss": 1.1312, "step": 2200 }, { "epoch": 0.19, "learning_rate": 1.8745140373772452e-05, "loss": 1.1902, "step": 2201 }, { "epoch": 0.19, "learning_rate": 1.8743809674509303e-05, "loss": 1.0991, "step": 2202 }, { "epoch": 0.19, "learning_rate": 1.8742478317343078e-05, "loss": 1.0887, "step": 2203 }, { "epoch": 0.19, "learning_rate": 1.8741146302373955e-05, "loss": 1.0962, "step": 2204 }, { "epoch": 0.19, "learning_rate": 1.873981362970215e-05, "loss": 1.0959, "step": 2205 }, { "epoch": 0.19, "learning_rate": 1.8738480299427938e-05, "loss": 1.1183, "step": 2206 }, { "epoch": 0.19, "learning_rate": 1.8737146311651648e-05, "loss": 1.1412, "step": 2207 }, { "epoch": 0.19, "learning_rate": 1.8735811666473646e-05, "loss": 1.1322, "step": 2208 }, { "epoch": 0.19, "learning_rate": 1.8734476363994353e-05, "loss": 1.115, "step": 2209 }, { "epoch": 0.19, "learning_rate": 1.8733140404314245e-05, "loss": 1.1042, "step": 2210 }, { "epoch": 0.19, "learning_rate": 1.873180378753384e-05, "loss": 1.1246, "step": 2211 }, { "epoch": 0.19, "learning_rate": 1.8730466513753706e-05, "loss": 1.0851, "step": 2212 }, { "epoch": 0.19, "learning_rate": 1.8729128583074465e-05, "loss": 1.0798, "step": 2213 }, { "epoch": 0.19, "learning_rate": 1.8727789995596784e-05, "loss": 1.1583, "step": 2214 }, { "epoch": 0.19, "learning_rate": 1.8726450751421384e-05, "loss": 1.0962, "step": 2215 }, { "epoch": 0.19, "learning_rate": 1.8725110850649028e-05, "loss": 1.0608, "step": 2216 }, { "epoch": 0.19, "learning_rate": 1.8723770293380537e-05, "loss": 1.1451, "step": 2217 }, { "epoch": 0.19, "learning_rate": 1.8722429079716776e-05, "loss": 1.0873, "step": 2218 }, { "epoch": 0.19, "learning_rate": 1.8721087209758657e-05, "loss": 1.1073, "step": 2219 }, { "epoch": 0.19, "learning_rate": 1.8719744683607156e-05, "loss": 1.1314, "step": 2220 }, { "epoch": 0.19, "learning_rate": 1.8718401501363273e-05, "loss": 1.0696, "step": 2221 }, { "epoch": 0.19, "learning_rate": 1.8717057663128084e-05, "loss": 1.1733, "step": 2222 }, { "epoch": 0.19, "learning_rate": 1.87157131690027e-05, "loss": 1.1168, "step": 2223 }, { "epoch": 0.19, "learning_rate": 1.8714368019088278e-05, "loss": 1.1064, "step": 2224 }, { "epoch": 0.19, "learning_rate": 1.8713022213486034e-05, "loss": 1.1218, "step": 2225 }, { "epoch": 0.19, "learning_rate": 1.871167575229723e-05, "loss": 1.0941, "step": 2226 }, { "epoch": 0.19, "learning_rate": 1.8710328635623172e-05, "loss": 1.0801, "step": 2227 }, { "epoch": 0.19, "learning_rate": 1.8708980863565226e-05, "loss": 1.1133, "step": 2228 }, { "epoch": 0.19, "learning_rate": 1.8707632436224798e-05, "loss": 1.1271, "step": 2229 }, { "epoch": 0.19, "learning_rate": 1.8706283353703348e-05, "loss": 1.1353, "step": 2230 }, { "epoch": 0.19, "learning_rate": 1.8704933616102382e-05, "loss": 1.1129, "step": 2231 }, { "epoch": 0.19, "learning_rate": 1.870358322352346e-05, "loss": 1.1711, "step": 2232 }, { "epoch": 0.19, "learning_rate": 1.8702232176068185e-05, "loss": 1.1222, "step": 2233 }, { "epoch": 0.19, "learning_rate": 1.8700880473838214e-05, "loss": 1.1241, "step": 2234 }, { "epoch": 0.19, "learning_rate": 1.8699528116935256e-05, "loss": 1.1752, "step": 2235 }, { "epoch": 0.19, "learning_rate": 1.869817510546106e-05, "loss": 1.0517, "step": 2236 }, { "epoch": 0.19, "learning_rate": 1.869682143951743e-05, "loss": 1.0791, "step": 2237 }, { "epoch": 0.19, "learning_rate": 1.8695467119206217e-05, "loss": 1.141, "step": 2238 }, { "epoch": 0.19, "learning_rate": 1.8694112144629328e-05, "loss": 1.1654, "step": 2239 }, { "epoch": 0.19, "learning_rate": 1.8692756515888714e-05, "loss": 1.1458, "step": 2240 }, { "epoch": 0.19, "learning_rate": 1.8691400233086373e-05, "loss": 1.0804, "step": 2241 }, { "epoch": 0.19, "learning_rate": 1.8690043296324353e-05, "loss": 1.0687, "step": 2242 }, { "epoch": 0.19, "learning_rate": 1.8688685705704756e-05, "loss": 1.1056, "step": 2243 }, { "epoch": 0.19, "learning_rate": 1.868732746132973e-05, "loss": 1.1518, "step": 2244 }, { "epoch": 0.19, "learning_rate": 1.868596856330147e-05, "loss": 1.1099, "step": 2245 }, { "epoch": 0.19, "learning_rate": 1.8684609011722223e-05, "loss": 1.0813, "step": 2246 }, { "epoch": 0.19, "learning_rate": 1.8683248806694286e-05, "loss": 1.1105, "step": 2247 }, { "epoch": 0.19, "learning_rate": 1.8681887948320005e-05, "loss": 1.0823, "step": 2248 }, { "epoch": 0.19, "learning_rate": 1.8680526436701768e-05, "loss": 1.0971, "step": 2249 }, { "epoch": 0.19, "learning_rate": 1.8679164271942025e-05, "loss": 1.162, "step": 2250 }, { "epoch": 0.19, "learning_rate": 1.8677801454143266e-05, "loss": 1.0863, "step": 2251 }, { "epoch": 0.19, "learning_rate": 1.867643798340803e-05, "loss": 1.1521, "step": 2252 }, { "epoch": 0.19, "learning_rate": 1.867507385983891e-05, "loss": 1.1046, "step": 2253 }, { "epoch": 0.19, "learning_rate": 1.867370908353854e-05, "loss": 0.9582, "step": 2254 }, { "epoch": 0.19, "learning_rate": 1.8672343654609618e-05, "loss": 1.128, "step": 2255 }, { "epoch": 0.19, "learning_rate": 1.8670977573154878e-05, "loss": 0.9249, "step": 2256 }, { "epoch": 0.19, "learning_rate": 1.8669610839277103e-05, "loss": 1.0897, "step": 2257 }, { "epoch": 0.19, "learning_rate": 1.8668243453079133e-05, "loss": 1.1094, "step": 2258 }, { "epoch": 0.19, "learning_rate": 1.8666875414663855e-05, "loss": 1.0947, "step": 2259 }, { "epoch": 0.19, "learning_rate": 1.8665506724134195e-05, "loss": 1.0979, "step": 2260 }, { "epoch": 0.19, "learning_rate": 1.8664137381593145e-05, "loss": 1.1811, "step": 2261 }, { "epoch": 0.19, "learning_rate": 1.8662767387143734e-05, "loss": 1.1162, "step": 2262 }, { "epoch": 0.19, "learning_rate": 1.866139674088904e-05, "loss": 1.0937, "step": 2263 }, { "epoch": 0.19, "learning_rate": 1.86600254429322e-05, "loss": 1.1075, "step": 2264 }, { "epoch": 0.19, "learning_rate": 1.865865349337639e-05, "loss": 1.1366, "step": 2265 }, { "epoch": 0.19, "learning_rate": 1.8657280892324833e-05, "loss": 1.0782, "step": 2266 }, { "epoch": 0.19, "learning_rate": 1.8655907639880816e-05, "loss": 1.1229, "step": 2267 }, { "epoch": 0.19, "learning_rate": 1.8654533736147664e-05, "loss": 1.0754, "step": 2268 }, { "epoch": 0.19, "learning_rate": 1.8653159181228748e-05, "loss": 1.1114, "step": 2269 }, { "epoch": 0.19, "learning_rate": 1.8651783975227488e-05, "loss": 1.1175, "step": 2270 }, { "epoch": 0.19, "learning_rate": 1.865040811824737e-05, "loss": 1.1646, "step": 2271 }, { "epoch": 0.19, "learning_rate": 1.8649031610391907e-05, "loss": 1.1168, "step": 2272 }, { "epoch": 0.19, "learning_rate": 1.8647654451764675e-05, "loss": 1.1392, "step": 2273 }, { "epoch": 0.19, "learning_rate": 1.8646276642469295e-05, "loss": 1.1219, "step": 2274 }, { "epoch": 0.19, "learning_rate": 1.8644898182609434e-05, "loss": 1.1183, "step": 2275 }, { "epoch": 0.19, "learning_rate": 1.8643519072288808e-05, "loss": 1.1293, "step": 2276 }, { "epoch": 0.19, "learning_rate": 1.8642139311611186e-05, "loss": 1.0947, "step": 2277 }, { "epoch": 0.19, "learning_rate": 1.8640758900680387e-05, "loss": 1.1549, "step": 2278 }, { "epoch": 0.19, "learning_rate": 1.8639377839600274e-05, "loss": 1.0869, "step": 2279 }, { "epoch": 0.19, "learning_rate": 1.863799612847476e-05, "loss": 1.0656, "step": 2280 }, { "epoch": 0.19, "learning_rate": 1.8636613767407814e-05, "loss": 1.1222, "step": 2281 }, { "epoch": 0.19, "learning_rate": 1.8635230756503437e-05, "loss": 1.1243, "step": 2282 }, { "epoch": 0.19, "learning_rate": 1.8633847095865694e-05, "loss": 1.1745, "step": 2283 }, { "epoch": 0.19, "learning_rate": 1.8632462785598702e-05, "loss": 1.1916, "step": 2284 }, { "epoch": 0.19, "learning_rate": 1.8631077825806612e-05, "loss": 1.1867, "step": 2285 }, { "epoch": 0.19, "learning_rate": 1.862969221659363e-05, "loss": 1.1482, "step": 2286 }, { "epoch": 0.19, "learning_rate": 1.8628305958064018e-05, "loss": 1.1435, "step": 2287 }, { "epoch": 0.19, "learning_rate": 1.862691905032208e-05, "loss": 1.1138, "step": 2288 }, { "epoch": 0.19, "learning_rate": 1.8625531493472164e-05, "loss": 1.1014, "step": 2289 }, { "epoch": 0.19, "learning_rate": 1.8624143287618678e-05, "loss": 1.1147, "step": 2290 }, { "epoch": 0.19, "learning_rate": 1.8622754432866076e-05, "loss": 1.1171, "step": 2291 }, { "epoch": 0.19, "learning_rate": 1.862136492931885e-05, "loss": 1.1107, "step": 2292 }, { "epoch": 0.19, "learning_rate": 1.8619974777081558e-05, "loss": 1.1419, "step": 2293 }, { "epoch": 0.19, "learning_rate": 1.8618583976258793e-05, "loss": 1.1258, "step": 2294 }, { "epoch": 0.19, "learning_rate": 1.8617192526955202e-05, "loss": 1.1693, "step": 2295 }, { "epoch": 0.19, "learning_rate": 1.8615800429275483e-05, "loss": 1.0861, "step": 2296 }, { "epoch": 0.19, "learning_rate": 1.8614407683324377e-05, "loss": 1.1774, "step": 2297 }, { "epoch": 0.19, "learning_rate": 1.8613014289206682e-05, "loss": 1.1365, "step": 2298 }, { "epoch": 0.19, "learning_rate": 1.8611620247027236e-05, "loss": 0.9821, "step": 2299 }, { "epoch": 0.19, "learning_rate": 1.8610225556890934e-05, "loss": 1.1441, "step": 2300 }, { "epoch": 0.19, "learning_rate": 1.860883021890271e-05, "loss": 1.0692, "step": 2301 }, { "epoch": 0.19, "learning_rate": 1.8607434233167557e-05, "loss": 1.1212, "step": 2302 }, { "epoch": 0.2, "learning_rate": 1.8606037599790507e-05, "loss": 1.1378, "step": 2303 }, { "epoch": 0.2, "learning_rate": 1.860464031887665e-05, "loss": 1.1213, "step": 2304 }, { "epoch": 0.2, "learning_rate": 1.860324239053112e-05, "loss": 1.118, "step": 2305 }, { "epoch": 0.2, "learning_rate": 1.86018438148591e-05, "loss": 1.0582, "step": 2306 }, { "epoch": 0.2, "learning_rate": 1.8600444591965817e-05, "loss": 1.1002, "step": 2307 }, { "epoch": 0.2, "learning_rate": 1.8599044721956555e-05, "loss": 1.058, "step": 2308 }, { "epoch": 0.2, "learning_rate": 1.8597644204936647e-05, "loss": 1.1315, "step": 2309 }, { "epoch": 0.2, "learning_rate": 1.859624304101147e-05, "loss": 1.1061, "step": 2310 }, { "epoch": 0.2, "learning_rate": 1.859484123028644e-05, "loss": 1.129, "step": 2311 }, { "epoch": 0.2, "learning_rate": 1.859343877286705e-05, "loss": 1.1374, "step": 2312 }, { "epoch": 0.2, "learning_rate": 1.8592035668858804e-05, "loss": 1.1006, "step": 2313 }, { "epoch": 0.2, "learning_rate": 1.8590631918367293e-05, "loss": 1.1362, "step": 2314 }, { "epoch": 0.2, "learning_rate": 1.858922752149813e-05, "loss": 1.0859, "step": 2315 }, { "epoch": 0.2, "learning_rate": 1.858782247835698e-05, "loss": 1.1131, "step": 2316 }, { "epoch": 0.2, "learning_rate": 1.8586416789049574e-05, "loss": 1.1302, "step": 2317 }, { "epoch": 0.2, "learning_rate": 1.8585010453681665e-05, "loss": 1.0663, "step": 2318 }, { "epoch": 0.2, "learning_rate": 1.8583603472359078e-05, "loss": 1.1347, "step": 2319 }, { "epoch": 0.2, "learning_rate": 1.8582195845187675e-05, "loss": 1.1216, "step": 2320 }, { "epoch": 0.2, "learning_rate": 1.8580787572273363e-05, "loss": 1.0893, "step": 2321 }, { "epoch": 0.2, "learning_rate": 1.8579378653722116e-05, "loss": 1.1266, "step": 2322 }, { "epoch": 0.2, "learning_rate": 1.8577969089639934e-05, "loss": 1.1139, "step": 2323 }, { "epoch": 0.2, "learning_rate": 1.857655888013288e-05, "loss": 1.1402, "step": 2324 }, { "epoch": 0.2, "learning_rate": 1.857514802530706e-05, "loss": 1.0953, "step": 2325 }, { "epoch": 0.2, "learning_rate": 1.8573736525268627e-05, "loss": 1.1362, "step": 2326 }, { "epoch": 0.2, "learning_rate": 1.8572324380123793e-05, "loss": 1.1432, "step": 2327 }, { "epoch": 0.2, "learning_rate": 1.8570911589978802e-05, "loss": 1.177, "step": 2328 }, { "epoch": 0.2, "learning_rate": 1.8569498154939958e-05, "loss": 1.1058, "step": 2329 }, { "epoch": 0.2, "learning_rate": 1.8568084075113613e-05, "loss": 1.1452, "step": 2330 }, { "epoch": 0.2, "learning_rate": 1.856666935060617e-05, "loss": 1.1877, "step": 2331 }, { "epoch": 0.2, "learning_rate": 1.8565253981524064e-05, "loss": 1.096, "step": 2332 }, { "epoch": 0.2, "learning_rate": 1.85638379679738e-05, "loss": 1.0619, "step": 2333 }, { "epoch": 0.2, "learning_rate": 1.8562421310061914e-05, "loss": 1.1947, "step": 2334 }, { "epoch": 0.2, "learning_rate": 1.856100400789501e-05, "loss": 1.107, "step": 2335 }, { "epoch": 0.2, "learning_rate": 1.8559586061579718e-05, "loss": 1.0748, "step": 2336 }, { "epoch": 0.2, "learning_rate": 1.855816747122273e-05, "loss": 1.1037, "step": 2337 }, { "epoch": 0.2, "learning_rate": 1.8556748236930788e-05, "loss": 1.1139, "step": 2338 }, { "epoch": 0.2, "learning_rate": 1.8555328358810673e-05, "loss": 1.0806, "step": 2339 }, { "epoch": 0.2, "learning_rate": 1.8553907836969225e-05, "loss": 1.1081, "step": 2340 }, { "epoch": 0.2, "learning_rate": 1.8552486671513322e-05, "loss": 1.1159, "step": 2341 }, { "epoch": 0.2, "learning_rate": 1.85510648625499e-05, "loss": 1.1161, "step": 2342 }, { "epoch": 0.2, "learning_rate": 1.8549642410185934e-05, "loss": 1.1202, "step": 2343 }, { "epoch": 0.2, "learning_rate": 1.8548219314528455e-05, "loss": 1.1579, "step": 2344 }, { "epoch": 0.2, "learning_rate": 1.8546795575684543e-05, "loss": 1.1431, "step": 2345 }, { "epoch": 0.2, "learning_rate": 1.854537119376132e-05, "loss": 1.1784, "step": 2346 }, { "epoch": 0.2, "learning_rate": 1.8543946168865955e-05, "loss": 1.0706, "step": 2347 }, { "epoch": 0.2, "learning_rate": 1.854252050110568e-05, "loss": 0.9549, "step": 2348 }, { "epoch": 0.2, "learning_rate": 1.8541094190587758e-05, "loss": 1.1859, "step": 2349 }, { "epoch": 0.2, "learning_rate": 1.8539667237419508e-05, "loss": 1.1515, "step": 2350 }, { "epoch": 0.2, "learning_rate": 1.8538239641708297e-05, "loss": 1.0305, "step": 2351 }, { "epoch": 0.2, "learning_rate": 1.8536811403561545e-05, "loss": 1.1218, "step": 2352 }, { "epoch": 0.2, "learning_rate": 1.853538252308671e-05, "loss": 1.1389, "step": 2353 }, { "epoch": 0.2, "learning_rate": 1.853395300039131e-05, "loss": 1.0723, "step": 2354 }, { "epoch": 0.2, "learning_rate": 1.85325228355829e-05, "loss": 1.1173, "step": 2355 }, { "epoch": 0.2, "learning_rate": 1.8531092028769092e-05, "loss": 1.0857, "step": 2356 }, { "epoch": 0.2, "learning_rate": 1.852966058005754e-05, "loss": 1.1235, "step": 2357 }, { "epoch": 0.2, "learning_rate": 1.8528228489555956e-05, "loss": 1.0963, "step": 2358 }, { "epoch": 0.2, "learning_rate": 1.8526795757372084e-05, "loss": 1.103, "step": 2359 }, { "epoch": 0.2, "learning_rate": 1.852536238361373e-05, "loss": 1.1454, "step": 2360 }, { "epoch": 0.2, "learning_rate": 1.8523928368388748e-05, "loss": 1.0944, "step": 2361 }, { "epoch": 0.2, "learning_rate": 1.852249371180503e-05, "loss": 1.1459, "step": 2362 }, { "epoch": 0.2, "learning_rate": 1.8521058413970524e-05, "loss": 1.0994, "step": 2363 }, { "epoch": 0.2, "learning_rate": 1.8519622474993232e-05, "loss": 1.1677, "step": 2364 }, { "epoch": 0.2, "learning_rate": 1.8518185894981186e-05, "loss": 1.1463, "step": 2365 }, { "epoch": 0.2, "learning_rate": 1.8516748674042487e-05, "loss": 1.1299, "step": 2366 }, { "epoch": 0.2, "learning_rate": 1.851531081228527e-05, "loss": 1.1146, "step": 2367 }, { "epoch": 0.2, "learning_rate": 1.8513872309817726e-05, "loss": 1.1276, "step": 2368 }, { "epoch": 0.2, "learning_rate": 1.851243316674808e-05, "loss": 1.0546, "step": 2369 }, { "epoch": 0.2, "learning_rate": 1.851099338318463e-05, "loss": 1.1375, "step": 2370 }, { "epoch": 0.2, "learning_rate": 1.8509552959235707e-05, "loss": 1.1091, "step": 2371 }, { "epoch": 0.2, "learning_rate": 1.8508111895009683e-05, "loss": 1.0986, "step": 2372 }, { "epoch": 0.2, "learning_rate": 1.850667019061499e-05, "loss": 1.1224, "step": 2373 }, { "epoch": 0.2, "learning_rate": 1.850522784616011e-05, "loss": 1.1325, "step": 2374 }, { "epoch": 0.2, "learning_rate": 1.8503784861753567e-05, "loss": 1.1928, "step": 2375 }, { "epoch": 0.2, "learning_rate": 1.8502341237503928e-05, "loss": 1.1085, "step": 2376 }, { "epoch": 0.2, "learning_rate": 1.8500896973519818e-05, "loss": 1.1353, "step": 2377 }, { "epoch": 0.2, "learning_rate": 1.849945206990991e-05, "loss": 1.1375, "step": 2378 }, { "epoch": 0.2, "learning_rate": 1.8498006526782913e-05, "loss": 1.1299, "step": 2379 }, { "epoch": 0.2, "learning_rate": 1.84965603442476e-05, "loss": 1.0937, "step": 2380 }, { "epoch": 0.2, "learning_rate": 1.8495113522412785e-05, "loss": 1.1639, "step": 2381 }, { "epoch": 0.2, "learning_rate": 1.849366606138733e-05, "loss": 1.1175, "step": 2382 }, { "epoch": 0.2, "learning_rate": 1.849221796128014e-05, "loss": 1.1316, "step": 2383 }, { "epoch": 0.2, "learning_rate": 1.8490769222200175e-05, "loss": 1.1039, "step": 2384 }, { "epoch": 0.2, "learning_rate": 1.8489319844256446e-05, "loss": 0.966, "step": 2385 }, { "epoch": 0.2, "learning_rate": 1.8487869827558e-05, "loss": 1.0609, "step": 2386 }, { "epoch": 0.2, "learning_rate": 1.8486419172213946e-05, "loss": 1.1019, "step": 2387 }, { "epoch": 0.2, "learning_rate": 1.848496787833343e-05, "loss": 1.0752, "step": 2388 }, { "epoch": 0.2, "learning_rate": 1.8483515946025655e-05, "loss": 1.1554, "step": 2389 }, { "epoch": 0.2, "learning_rate": 1.8482063375399865e-05, "loss": 1.0794, "step": 2390 }, { "epoch": 0.2, "learning_rate": 1.848061016656535e-05, "loss": 1.1387, "step": 2391 }, { "epoch": 0.2, "learning_rate": 1.847915631963146e-05, "loss": 1.0964, "step": 2392 }, { "epoch": 0.2, "learning_rate": 1.847770183470758e-05, "loss": 1.114, "step": 2393 }, { "epoch": 0.2, "learning_rate": 1.8476246711903156e-05, "loss": 1.1301, "step": 2394 }, { "epoch": 0.2, "learning_rate": 1.8474790951327664e-05, "loss": 1.126, "step": 2395 }, { "epoch": 0.2, "learning_rate": 1.847333455309065e-05, "loss": 1.1337, "step": 2396 }, { "epoch": 0.2, "learning_rate": 1.8471877517301685e-05, "loss": 1.1472, "step": 2397 }, { "epoch": 0.2, "learning_rate": 1.8470419844070403e-05, "loss": 1.1262, "step": 2398 }, { "epoch": 0.2, "learning_rate": 1.8468961533506492e-05, "loss": 1.1051, "step": 2399 }, { "epoch": 0.2, "learning_rate": 1.846750258571967e-05, "loss": 1.1135, "step": 2400 }, { "epoch": 0.2, "learning_rate": 1.8466043000819708e-05, "loss": 1.1498, "step": 2401 }, { "epoch": 0.2, "learning_rate": 1.8464582778916435e-05, "loss": 1.1236, "step": 2402 }, { "epoch": 0.2, "learning_rate": 1.8463121920119717e-05, "loss": 1.0836, "step": 2403 }, { "epoch": 0.2, "learning_rate": 1.8461660424539476e-05, "loss": 1.1374, "step": 2404 }, { "epoch": 0.2, "learning_rate": 1.8460198292285677e-05, "loss": 1.177, "step": 2405 }, { "epoch": 0.2, "learning_rate": 1.8458735523468332e-05, "loss": 1.1411, "step": 2406 }, { "epoch": 0.2, "learning_rate": 1.8457272118197505e-05, "loss": 1.1015, "step": 2407 }, { "epoch": 0.2, "learning_rate": 1.8455808076583304e-05, "loss": 1.0601, "step": 2408 }, { "epoch": 0.2, "learning_rate": 1.845434339873589e-05, "loss": 1.0529, "step": 2409 }, { "epoch": 0.2, "learning_rate": 1.845287808476546e-05, "loss": 1.131, "step": 2410 }, { "epoch": 0.2, "learning_rate": 1.845141213478228e-05, "loss": 1.1239, "step": 2411 }, { "epoch": 0.2, "learning_rate": 1.844994554889664e-05, "loss": 1.0793, "step": 2412 }, { "epoch": 0.2, "learning_rate": 1.8448478327218895e-05, "loss": 1.1746, "step": 2413 }, { "epoch": 0.2, "learning_rate": 1.8447010469859444e-05, "loss": 1.1416, "step": 2414 }, { "epoch": 0.2, "learning_rate": 1.8445541976928726e-05, "loss": 1.0915, "step": 2415 }, { "epoch": 0.2, "learning_rate": 1.8444072848537232e-05, "loss": 1.099, "step": 2416 }, { "epoch": 0.2, "learning_rate": 1.8442603084795513e-05, "loss": 1.1475, "step": 2417 }, { "epoch": 0.2, "learning_rate": 1.8441132685814144e-05, "loss": 1.1632, "step": 2418 }, { "epoch": 0.2, "learning_rate": 1.843966165170377e-05, "loss": 1.1692, "step": 2419 }, { "epoch": 0.2, "learning_rate": 1.8438189982575077e-05, "loss": 1.1413, "step": 2420 }, { "epoch": 0.21, "learning_rate": 1.8436717678538787e-05, "loss": 1.0961, "step": 2421 }, { "epoch": 0.21, "learning_rate": 1.8435244739705683e-05, "loss": 1.1295, "step": 2422 }, { "epoch": 0.21, "learning_rate": 1.8433771166186597e-05, "loss": 1.1105, "step": 2423 }, { "epoch": 0.21, "learning_rate": 1.8432296958092396e-05, "loss": 1.0654, "step": 2424 }, { "epoch": 0.21, "learning_rate": 1.843082211553401e-05, "loss": 1.0884, "step": 2425 }, { "epoch": 0.21, "learning_rate": 1.8429346638622407e-05, "loss": 1.1525, "step": 2426 }, { "epoch": 0.21, "learning_rate": 1.8427870527468604e-05, "loss": 1.1364, "step": 2427 }, { "epoch": 0.21, "learning_rate": 1.8426393782183662e-05, "loss": 1.0997, "step": 2428 }, { "epoch": 0.21, "learning_rate": 1.8424916402878703e-05, "loss": 1.1132, "step": 2429 }, { "epoch": 0.21, "learning_rate": 1.8423438389664884e-05, "loss": 1.1252, "step": 2430 }, { "epoch": 0.21, "learning_rate": 1.8421959742653418e-05, "loss": 1.0869, "step": 2431 }, { "epoch": 0.21, "learning_rate": 1.8420480461955553e-05, "loss": 1.1381, "step": 2432 }, { "epoch": 0.21, "learning_rate": 1.8419000547682597e-05, "loss": 1.1014, "step": 2433 }, { "epoch": 0.21, "learning_rate": 1.8417519999945908e-05, "loss": 1.0599, "step": 2434 }, { "epoch": 0.21, "learning_rate": 1.841603881885688e-05, "loss": 1.1047, "step": 2435 }, { "epoch": 0.21, "learning_rate": 1.8414557004526958e-05, "loss": 1.1425, "step": 2436 }, { "epoch": 0.21, "learning_rate": 1.8413074557067644e-05, "loss": 1.0648, "step": 2437 }, { "epoch": 0.21, "learning_rate": 1.8411591476590474e-05, "loss": 1.1081, "step": 2438 }, { "epoch": 0.21, "learning_rate": 1.841010776320704e-05, "loss": 1.035, "step": 2439 }, { "epoch": 0.21, "learning_rate": 1.8408623417028982e-05, "loss": 1.0879, "step": 2440 }, { "epoch": 0.21, "learning_rate": 1.8407138438167985e-05, "loss": 1.1139, "step": 2441 }, { "epoch": 0.21, "learning_rate": 1.8405652826735782e-05, "loss": 1.1065, "step": 2442 }, { "epoch": 0.21, "learning_rate": 1.840416658284415e-05, "loss": 1.1181, "step": 2443 }, { "epoch": 0.21, "learning_rate": 1.840267970660492e-05, "loss": 1.1279, "step": 2444 }, { "epoch": 0.21, "learning_rate": 1.8401192198129965e-05, "loss": 1.0642, "step": 2445 }, { "epoch": 0.21, "learning_rate": 1.8399704057531217e-05, "loss": 1.0825, "step": 2446 }, { "epoch": 0.21, "learning_rate": 1.8398215284920634e-05, "loss": 1.094, "step": 2447 }, { "epoch": 0.21, "learning_rate": 1.8396725880410248e-05, "loss": 1.0912, "step": 2448 }, { "epoch": 0.21, "learning_rate": 1.8395235844112112e-05, "loss": 1.129, "step": 2449 }, { "epoch": 0.21, "learning_rate": 1.839374517613835e-05, "loss": 1.1132, "step": 2450 }, { "epoch": 0.21, "learning_rate": 1.8392253876601117e-05, "loss": 1.1238, "step": 2451 }, { "epoch": 0.21, "learning_rate": 1.8390761945612624e-05, "loss": 1.1205, "step": 2452 }, { "epoch": 0.21, "learning_rate": 1.8389269383285128e-05, "loss": 1.1167, "step": 2453 }, { "epoch": 0.21, "learning_rate": 1.838777618973093e-05, "loss": 1.1254, "step": 2454 }, { "epoch": 0.21, "learning_rate": 1.838628236506238e-05, "loss": 1.1495, "step": 2455 }, { "epoch": 0.21, "learning_rate": 1.838478790939188e-05, "loss": 1.1282, "step": 2456 }, { "epoch": 0.21, "learning_rate": 1.8383292822831873e-05, "loss": 1.1073, "step": 2457 }, { "epoch": 0.21, "learning_rate": 1.838179710549486e-05, "loss": 1.1924, "step": 2458 }, { "epoch": 0.21, "learning_rate": 1.838030075749337e-05, "loss": 0.9999, "step": 2459 }, { "epoch": 0.21, "learning_rate": 1.8378803778939998e-05, "loss": 1.1591, "step": 2460 }, { "epoch": 0.21, "learning_rate": 1.8377306169947387e-05, "loss": 1.1095, "step": 2461 }, { "epoch": 0.21, "learning_rate": 1.8375807930628208e-05, "loss": 1.0929, "step": 2462 }, { "epoch": 0.21, "learning_rate": 1.8374309061095198e-05, "loss": 1.089, "step": 2463 }, { "epoch": 0.21, "learning_rate": 1.8372809561461135e-05, "loss": 1.1062, "step": 2464 }, { "epoch": 0.21, "learning_rate": 1.8371309431838836e-05, "loss": 1.1149, "step": 2465 }, { "epoch": 0.21, "learning_rate": 1.836980867234119e-05, "loss": 1.1402, "step": 2466 }, { "epoch": 0.21, "learning_rate": 1.836830728308111e-05, "loss": 1.1001, "step": 2467 }, { "epoch": 0.21, "learning_rate": 1.8366805264171557e-05, "loss": 1.1048, "step": 2468 }, { "epoch": 0.21, "learning_rate": 1.836530261572555e-05, "loss": 1.1023, "step": 2469 }, { "epoch": 0.21, "learning_rate": 1.836379933785616e-05, "loss": 1.1211, "step": 2470 }, { "epoch": 0.21, "learning_rate": 1.8362295430676485e-05, "loss": 1.1347, "step": 2471 }, { "epoch": 0.21, "learning_rate": 1.836079089429969e-05, "loss": 1.0853, "step": 2472 }, { "epoch": 0.21, "learning_rate": 1.8359285728838977e-05, "loss": 1.0542, "step": 2473 }, { "epoch": 0.21, "learning_rate": 1.83577799344076e-05, "loss": 1.1286, "step": 2474 }, { "epoch": 0.21, "learning_rate": 1.8356273511118852e-05, "loss": 1.126, "step": 2475 }, { "epoch": 0.21, "learning_rate": 1.8354766459086087e-05, "loss": 1.1184, "step": 2476 }, { "epoch": 0.21, "learning_rate": 1.8353258778422693e-05, "loss": 1.1502, "step": 2477 }, { "epoch": 0.21, "learning_rate": 1.8351750469242117e-05, "loss": 1.1478, "step": 2478 }, { "epoch": 0.21, "learning_rate": 1.835024153165784e-05, "loss": 1.1314, "step": 2479 }, { "epoch": 0.21, "learning_rate": 1.8348731965783403e-05, "loss": 1.1589, "step": 2480 }, { "epoch": 0.21, "learning_rate": 1.834722177173239e-05, "loss": 1.0865, "step": 2481 }, { "epoch": 0.21, "learning_rate": 1.8345710949618426e-05, "loss": 1.0377, "step": 2482 }, { "epoch": 0.21, "learning_rate": 1.8344199499555194e-05, "loss": 1.121, "step": 2483 }, { "epoch": 0.21, "learning_rate": 1.8342687421656414e-05, "loss": 1.0983, "step": 2484 }, { "epoch": 0.21, "learning_rate": 1.8341174716035862e-05, "loss": 1.0729, "step": 2485 }, { "epoch": 0.21, "learning_rate": 1.8339661382807352e-05, "loss": 1.1379, "step": 2486 }, { "epoch": 0.21, "learning_rate": 1.8338147422084756e-05, "loss": 1.0466, "step": 2487 }, { "epoch": 0.21, "learning_rate": 1.833663283398199e-05, "loss": 1.1292, "step": 2488 }, { "epoch": 0.21, "learning_rate": 1.8335117618613006e-05, "loss": 0.9383, "step": 2489 }, { "epoch": 0.21, "learning_rate": 1.8333601776091816e-05, "loss": 1.0923, "step": 2490 }, { "epoch": 0.21, "learning_rate": 1.8332085306532478e-05, "loss": 1.1318, "step": 2491 }, { "epoch": 0.21, "learning_rate": 1.8330568210049086e-05, "loss": 1.0994, "step": 2492 }, { "epoch": 0.21, "learning_rate": 1.8329050486755805e-05, "loss": 1.1254, "step": 2493 }, { "epoch": 0.21, "learning_rate": 1.8327532136766816e-05, "loss": 1.048, "step": 2494 }, { "epoch": 0.21, "learning_rate": 1.8326013160196373e-05, "loss": 1.1524, "step": 2495 }, { "epoch": 0.21, "learning_rate": 1.8324493557158765e-05, "loss": 1.1256, "step": 2496 }, { "epoch": 0.21, "learning_rate": 1.8322973327768327e-05, "loss": 1.0762, "step": 2497 }, { "epoch": 0.21, "learning_rate": 1.8321452472139444e-05, "loss": 1.1568, "step": 2498 }, { "epoch": 0.21, "learning_rate": 1.831993099038655e-05, "loss": 1.0412, "step": 2499 }, { "epoch": 0.21, "learning_rate": 1.831840888262413e-05, "loss": 1.092, "step": 2500 }, { "epoch": 0.21, "learning_rate": 1.83168861489667e-05, "loss": 1.072, "step": 2501 }, { "epoch": 0.21, "learning_rate": 1.8315362789528846e-05, "loss": 1.1114, "step": 2502 }, { "epoch": 0.21, "learning_rate": 1.831383880442518e-05, "loss": 1.1398, "step": 2503 }, { "epoch": 0.21, "learning_rate": 1.831231419377037e-05, "loss": 1.0726, "step": 2504 }, { "epoch": 0.21, "learning_rate": 1.8310788957679135e-05, "loss": 1.1319, "step": 2505 }, { "epoch": 0.21, "learning_rate": 1.8309263096266233e-05, "loss": 1.0985, "step": 2506 }, { "epoch": 0.21, "learning_rate": 1.830773660964648e-05, "loss": 1.1069, "step": 2507 }, { "epoch": 0.21, "learning_rate": 1.8306209497934727e-05, "loss": 1.1398, "step": 2508 }, { "epoch": 0.21, "learning_rate": 1.8304681761245876e-05, "loss": 0.9419, "step": 2509 }, { "epoch": 0.21, "learning_rate": 1.830315339969488e-05, "loss": 1.1099, "step": 2510 }, { "epoch": 0.21, "learning_rate": 1.8301624413396734e-05, "loss": 1.1816, "step": 2511 }, { "epoch": 0.21, "learning_rate": 1.8300094802466483e-05, "loss": 1.1554, "step": 2512 }, { "epoch": 0.21, "learning_rate": 1.8298564567019223e-05, "loss": 1.1269, "step": 2513 }, { "epoch": 0.21, "learning_rate": 1.8297033707170082e-05, "loss": 1.1173, "step": 2514 }, { "epoch": 0.21, "learning_rate": 1.8295502223034255e-05, "loss": 1.1462, "step": 2515 }, { "epoch": 0.21, "learning_rate": 1.829397011472697e-05, "loss": 1.1629, "step": 2516 }, { "epoch": 0.21, "learning_rate": 1.8292437382363502e-05, "loss": 1.0415, "step": 2517 }, { "epoch": 0.21, "learning_rate": 1.8290904026059188e-05, "loss": 1.1201, "step": 2518 }, { "epoch": 0.21, "learning_rate": 1.828937004592939e-05, "loss": 1.082, "step": 2519 }, { "epoch": 0.21, "learning_rate": 1.8287835442089535e-05, "loss": 1.1417, "step": 2520 }, { "epoch": 0.21, "learning_rate": 1.8286300214655088e-05, "loss": 1.0969, "step": 2521 }, { "epoch": 0.21, "learning_rate": 1.8284764363741558e-05, "loss": 1.0997, "step": 2522 }, { "epoch": 0.21, "learning_rate": 1.8283227889464512e-05, "loss": 0.8962, "step": 2523 }, { "epoch": 0.21, "learning_rate": 1.8281690791939557e-05, "loss": 1.1136, "step": 2524 }, { "epoch": 0.21, "learning_rate": 1.8280153071282345e-05, "loss": 0.9684, "step": 2525 }, { "epoch": 0.21, "learning_rate": 1.8278614727608576e-05, "loss": 1.1275, "step": 2526 }, { "epoch": 0.21, "learning_rate": 1.8277075761034004e-05, "loss": 1.1441, "step": 2527 }, { "epoch": 0.21, "learning_rate": 1.8275536171674422e-05, "loss": 1.1377, "step": 2528 }, { "epoch": 0.21, "learning_rate": 1.827399595964567e-05, "loss": 1.195, "step": 2529 }, { "epoch": 0.21, "learning_rate": 1.8272455125063637e-05, "loss": 1.1192, "step": 2530 }, { "epoch": 0.21, "learning_rate": 1.8270913668044257e-05, "loss": 1.0843, "step": 2531 }, { "epoch": 0.21, "learning_rate": 1.8269371588703517e-05, "loss": 1.103, "step": 2532 }, { "epoch": 0.21, "learning_rate": 1.8267828887157445e-05, "loss": 1.0905, "step": 2533 }, { "epoch": 0.21, "learning_rate": 1.826628556352212e-05, "loss": 0.9848, "step": 2534 }, { "epoch": 0.21, "learning_rate": 1.8264741617913656e-05, "loss": 1.0856, "step": 2535 }, { "epoch": 0.21, "learning_rate": 1.8263197050448232e-05, "loss": 1.1398, "step": 2536 }, { "epoch": 0.21, "learning_rate": 1.826165186124206e-05, "loss": 1.0944, "step": 2537 }, { "epoch": 0.21, "learning_rate": 1.8260106050411404e-05, "loss": 1.1353, "step": 2538 }, { "epoch": 0.22, "learning_rate": 1.8258559618072575e-05, "loss": 1.1543, "step": 2539 }, { "epoch": 0.22, "learning_rate": 1.825701256434193e-05, "loss": 1.1298, "step": 2540 }, { "epoch": 0.22, "learning_rate": 1.825546488933587e-05, "loss": 1.0766, "step": 2541 }, { "epoch": 0.22, "learning_rate": 1.825391659317085e-05, "loss": 0.9402, "step": 2542 }, { "epoch": 0.22, "learning_rate": 1.8252367675963362e-05, "loss": 1.1081, "step": 2543 }, { "epoch": 0.22, "learning_rate": 1.8250818137829954e-05, "loss": 1.1708, "step": 2544 }, { "epoch": 0.22, "learning_rate": 1.8249267978887217e-05, "loss": 1.0616, "step": 2545 }, { "epoch": 0.22, "learning_rate": 1.8247717199251783e-05, "loss": 1.0815, "step": 2546 }, { "epoch": 0.22, "learning_rate": 1.824616579904034e-05, "loss": 1.0924, "step": 2547 }, { "epoch": 0.22, "learning_rate": 1.824461377836962e-05, "loss": 1.135, "step": 2548 }, { "epoch": 0.22, "learning_rate": 1.8243061137356396e-05, "loss": 1.1198, "step": 2549 }, { "epoch": 0.22, "learning_rate": 1.8241507876117498e-05, "loss": 1.1134, "step": 2550 }, { "epoch": 0.22, "learning_rate": 1.823995399476979e-05, "loss": 1.1689, "step": 2551 }, { "epoch": 0.22, "learning_rate": 1.8238399493430196e-05, "loss": 1.0888, "step": 2552 }, { "epoch": 0.22, "learning_rate": 1.823684437221567e-05, "loss": 1.137, "step": 2553 }, { "epoch": 0.22, "learning_rate": 1.8235288631243235e-05, "loss": 1.1577, "step": 2554 }, { "epoch": 0.22, "learning_rate": 1.8233732270629943e-05, "loss": 1.0855, "step": 2555 }, { "epoch": 0.22, "learning_rate": 1.8232175290492897e-05, "loss": 1.1018, "step": 2556 }, { "epoch": 0.22, "learning_rate": 1.8230617690949245e-05, "loss": 1.1144, "step": 2557 }, { "epoch": 0.22, "learning_rate": 1.8229059472116188e-05, "loss": 1.051, "step": 2558 }, { "epoch": 0.22, "learning_rate": 1.822750063411097e-05, "loss": 1.1559, "step": 2559 }, { "epoch": 0.22, "learning_rate": 1.8225941177050882e-05, "loss": 1.1388, "step": 2560 }, { "epoch": 0.22, "learning_rate": 1.8224381101053256e-05, "loss": 1.1065, "step": 2561 }, { "epoch": 0.22, "learning_rate": 1.8222820406235483e-05, "loss": 1.1529, "step": 2562 }, { "epoch": 0.22, "learning_rate": 1.8221259092714985e-05, "loss": 1.1266, "step": 2563 }, { "epoch": 0.22, "learning_rate": 1.8219697160609244e-05, "loss": 1.1448, "step": 2564 }, { "epoch": 0.22, "learning_rate": 1.821813461003578e-05, "loss": 1.1507, "step": 2565 }, { "epoch": 0.22, "learning_rate": 1.8216571441112162e-05, "loss": 1.0702, "step": 2566 }, { "epoch": 0.22, "learning_rate": 1.8215007653956012e-05, "loss": 1.0818, "step": 2567 }, { "epoch": 0.22, "learning_rate": 1.821344324868499e-05, "loss": 1.1332, "step": 2568 }, { "epoch": 0.22, "learning_rate": 1.82118782254168e-05, "loss": 1.0941, "step": 2569 }, { "epoch": 0.22, "learning_rate": 1.8210312584269206e-05, "loss": 1.1711, "step": 2570 }, { "epoch": 0.22, "learning_rate": 1.8208746325360005e-05, "loss": 1.1109, "step": 2571 }, { "epoch": 0.22, "learning_rate": 1.8207179448807048e-05, "loss": 1.1074, "step": 2572 }, { "epoch": 0.22, "learning_rate": 1.8205611954728224e-05, "loss": 1.0589, "step": 2573 }, { "epoch": 0.22, "learning_rate": 1.8204043843241484e-05, "loss": 1.0646, "step": 2574 }, { "epoch": 0.22, "learning_rate": 1.820247511446481e-05, "loss": 1.1098, "step": 2575 }, { "epoch": 0.22, "learning_rate": 1.820090576851624e-05, "loss": 1.1656, "step": 2576 }, { "epoch": 0.22, "learning_rate": 1.819933580551385e-05, "loss": 1.1377, "step": 2577 }, { "epoch": 0.22, "learning_rate": 1.8197765225575775e-05, "loss": 1.0557, "step": 2578 }, { "epoch": 0.22, "learning_rate": 1.8196194028820185e-05, "loss": 1.1668, "step": 2579 }, { "epoch": 0.22, "learning_rate": 1.8194622215365297e-05, "loss": 1.1129, "step": 2580 }, { "epoch": 0.22, "learning_rate": 1.819304978532938e-05, "loss": 1.0711, "step": 2581 }, { "epoch": 0.22, "learning_rate": 1.8191476738830747e-05, "loss": 1.1586, "step": 2582 }, { "epoch": 0.22, "learning_rate": 1.8189903075987758e-05, "loss": 1.1258, "step": 2583 }, { "epoch": 0.22, "learning_rate": 1.8188328796918822e-05, "loss": 1.1146, "step": 2584 }, { "epoch": 0.22, "learning_rate": 1.8186753901742383e-05, "loss": 1.0642, "step": 2585 }, { "epoch": 0.22, "learning_rate": 1.8185178390576948e-05, "loss": 1.0846, "step": 2586 }, { "epoch": 0.22, "learning_rate": 1.8183602263541056e-05, "loss": 1.0774, "step": 2587 }, { "epoch": 0.22, "learning_rate": 1.8182025520753303e-05, "loss": 1.1439, "step": 2588 }, { "epoch": 0.22, "learning_rate": 1.8180448162332317e-05, "loss": 1.1411, "step": 2589 }, { "epoch": 0.22, "learning_rate": 1.8178870188396795e-05, "loss": 1.0737, "step": 2590 }, { "epoch": 0.22, "learning_rate": 1.817729159906546e-05, "loss": 1.0988, "step": 2591 }, { "epoch": 0.22, "learning_rate": 1.8175712394457086e-05, "loss": 1.0961, "step": 2592 }, { "epoch": 0.22, "learning_rate": 1.8174132574690503e-05, "loss": 1.0459, "step": 2593 }, { "epoch": 0.22, "learning_rate": 1.8172552139884576e-05, "loss": 1.1649, "step": 2594 }, { "epoch": 0.22, "learning_rate": 1.8170971090158217e-05, "loss": 1.134, "step": 2595 }, { "epoch": 0.22, "learning_rate": 1.8169389425630393e-05, "loss": 1.0773, "step": 2596 }, { "epoch": 0.22, "learning_rate": 1.8167807146420106e-05, "loss": 1.0737, "step": 2597 }, { "epoch": 0.22, "learning_rate": 1.816622425264642e-05, "loss": 1.1127, "step": 2598 }, { "epoch": 0.22, "learning_rate": 1.8164640744428426e-05, "loss": 1.1473, "step": 2599 }, { "epoch": 0.22, "learning_rate": 1.8163056621885276e-05, "loss": 1.0878, "step": 2600 }, { "epoch": 0.22, "learning_rate": 1.8161471885136158e-05, "loss": 1.1059, "step": 2601 }, { "epoch": 0.22, "learning_rate": 1.8159886534300316e-05, "loss": 1.0851, "step": 2602 }, { "epoch": 0.22, "learning_rate": 1.815830056949703e-05, "loss": 1.1177, "step": 2603 }, { "epoch": 0.22, "learning_rate": 1.8156713990845634e-05, "loss": 1.1113, "step": 2604 }, { "epoch": 0.22, "learning_rate": 1.8155126798465507e-05, "loss": 1.1217, "step": 2605 }, { "epoch": 0.22, "learning_rate": 1.8153538992476072e-05, "loss": 1.09, "step": 2606 }, { "epoch": 0.22, "learning_rate": 1.81519505729968e-05, "loss": 1.1044, "step": 2607 }, { "epoch": 0.22, "learning_rate": 1.8150361540147203e-05, "loss": 1.0971, "step": 2608 }, { "epoch": 0.22, "learning_rate": 1.8148771894046846e-05, "loss": 1.1192, "step": 2609 }, { "epoch": 0.22, "learning_rate": 1.814718163481534e-05, "loss": 1.1622, "step": 2610 }, { "epoch": 0.22, "learning_rate": 1.8145590762572335e-05, "loss": 1.1374, "step": 2611 }, { "epoch": 0.22, "learning_rate": 1.8143999277437532e-05, "loss": 0.9488, "step": 2612 }, { "epoch": 0.22, "learning_rate": 1.814240717953068e-05, "loss": 1.1022, "step": 2613 }, { "epoch": 0.22, "learning_rate": 1.8140814468971574e-05, "loss": 1.0894, "step": 2614 }, { "epoch": 0.22, "learning_rate": 1.8139221145880048e-05, "loss": 1.0903, "step": 2615 }, { "epoch": 0.22, "learning_rate": 1.813762721037599e-05, "loss": 1.0774, "step": 2616 }, { "epoch": 0.22, "learning_rate": 1.8136032662579327e-05, "loss": 1.136, "step": 2617 }, { "epoch": 0.22, "learning_rate": 1.8134437502610044e-05, "loss": 1.1096, "step": 2618 }, { "epoch": 0.22, "learning_rate": 1.813284173058816e-05, "loss": 1.1225, "step": 2619 }, { "epoch": 0.22, "learning_rate": 1.813124534663374e-05, "loss": 1.1191, "step": 2620 }, { "epoch": 0.22, "learning_rate": 1.8129648350866906e-05, "loss": 1.1274, "step": 2621 }, { "epoch": 0.22, "learning_rate": 1.8128050743407817e-05, "loss": 1.1156, "step": 2622 }, { "epoch": 0.22, "learning_rate": 1.812645252437668e-05, "loss": 1.1222, "step": 2623 }, { "epoch": 0.22, "learning_rate": 1.8124853693893745e-05, "loss": 1.0912, "step": 2624 }, { "epoch": 0.22, "learning_rate": 1.812325425207932e-05, "loss": 1.1284, "step": 2625 }, { "epoch": 0.22, "learning_rate": 1.8121654199053743e-05, "loss": 1.135, "step": 2626 }, { "epoch": 0.22, "learning_rate": 1.812005353493741e-05, "loss": 1.1186, "step": 2627 }, { "epoch": 0.22, "learning_rate": 1.811845225985076e-05, "loss": 1.0854, "step": 2628 }, { "epoch": 0.22, "learning_rate": 1.8116850373914267e-05, "loss": 1.137, "step": 2629 }, { "epoch": 0.22, "learning_rate": 1.8115247877248467e-05, "loss": 1.1085, "step": 2630 }, { "epoch": 0.22, "learning_rate": 1.8113644769973934e-05, "loss": 1.0856, "step": 2631 }, { "epoch": 0.22, "learning_rate": 1.8112041052211295e-05, "loss": 1.086, "step": 2632 }, { "epoch": 0.22, "learning_rate": 1.8110436724081206e-05, "loss": 1.1104, "step": 2633 }, { "epoch": 0.22, "learning_rate": 1.8108831785704387e-05, "loss": 1.1428, "step": 2634 }, { "epoch": 0.22, "learning_rate": 1.81072262372016e-05, "loss": 1.0972, "step": 2635 }, { "epoch": 0.22, "learning_rate": 1.8105620078693643e-05, "loss": 1.0958, "step": 2636 }, { "epoch": 0.22, "learning_rate": 1.810401331030137e-05, "loss": 1.1299, "step": 2637 }, { "epoch": 0.22, "learning_rate": 1.8102405932145676e-05, "loss": 1.069, "step": 2638 }, { "epoch": 0.22, "learning_rate": 1.810079794434751e-05, "loss": 1.1434, "step": 2639 }, { "epoch": 0.22, "learning_rate": 1.809918934702785e-05, "loss": 1.16, "step": 2640 }, { "epoch": 0.22, "learning_rate": 1.809758014030774e-05, "loss": 1.1509, "step": 2641 }, { "epoch": 0.22, "learning_rate": 1.8095970324308253e-05, "loss": 1.1407, "step": 2642 }, { "epoch": 0.22, "learning_rate": 1.809435989915052e-05, "loss": 1.0803, "step": 2643 }, { "epoch": 0.22, "learning_rate": 1.8092748864955712e-05, "loss": 1.1126, "step": 2644 }, { "epoch": 0.22, "learning_rate": 1.8091137221845042e-05, "loss": 1.0892, "step": 2645 }, { "epoch": 0.22, "learning_rate": 1.8089524969939782e-05, "loss": 1.0884, "step": 2646 }, { "epoch": 0.22, "learning_rate": 1.8087912109361233e-05, "loss": 1.0337, "step": 2647 }, { "epoch": 0.22, "learning_rate": 1.8086298640230756e-05, "loss": 1.1183, "step": 2648 }, { "epoch": 0.22, "learning_rate": 1.808468456266975e-05, "loss": 1.1448, "step": 2649 }, { "epoch": 0.22, "learning_rate": 1.808306987679966e-05, "loss": 1.0726, "step": 2650 }, { "epoch": 0.22, "learning_rate": 1.8081454582741978e-05, "loss": 1.108, "step": 2651 }, { "epoch": 0.22, "learning_rate": 1.8079838680618244e-05, "loss": 0.9599, "step": 2652 }, { "epoch": 0.22, "learning_rate": 1.8078222170550043e-05, "loss": 1.108, "step": 2653 }, { "epoch": 0.22, "learning_rate": 1.8076605052659e-05, "loss": 1.0822, "step": 2654 }, { "epoch": 0.22, "learning_rate": 1.80749873270668e-05, "loss": 1.1352, "step": 2655 }, { "epoch": 0.22, "learning_rate": 1.8073368993895156e-05, "loss": 1.1116, "step": 2656 }, { "epoch": 0.23, "learning_rate": 1.8071750053265835e-05, "loss": 1.1316, "step": 2657 }, { "epoch": 0.23, "learning_rate": 1.8070130505300653e-05, "loss": 1.1086, "step": 2658 }, { "epoch": 0.23, "learning_rate": 1.8068510350121467e-05, "loss": 1.155, "step": 2659 }, { "epoch": 0.23, "learning_rate": 1.8066889587850177e-05, "loss": 1.0982, "step": 2660 }, { "epoch": 0.23, "learning_rate": 1.806526821860874e-05, "loss": 1.0979, "step": 2661 }, { "epoch": 0.23, "learning_rate": 1.806364624251915e-05, "loss": 1.1515, "step": 2662 }, { "epoch": 0.23, "learning_rate": 1.8062023659703443e-05, "loss": 1.0974, "step": 2663 }, { "epoch": 0.23, "learning_rate": 1.8060400470283705e-05, "loss": 1.0619, "step": 2664 }, { "epoch": 0.23, "learning_rate": 1.8058776674382078e-05, "loss": 1.0975, "step": 2665 }, { "epoch": 0.23, "learning_rate": 1.805715227212073e-05, "loss": 1.0925, "step": 2666 }, { "epoch": 0.23, "learning_rate": 1.805552726362189e-05, "loss": 1.1387, "step": 2667 }, { "epoch": 0.23, "learning_rate": 1.805390164900782e-05, "loss": 1.109, "step": 2668 }, { "epoch": 0.23, "learning_rate": 1.8052275428400847e-05, "loss": 1.0428, "step": 2669 }, { "epoch": 0.23, "learning_rate": 1.805064860192332e-05, "loss": 1.1167, "step": 2670 }, { "epoch": 0.23, "learning_rate": 1.804902116969765e-05, "loss": 1.0727, "step": 2671 }, { "epoch": 0.23, "learning_rate": 1.8047393131846286e-05, "loss": 1.1374, "step": 2672 }, { "epoch": 0.23, "learning_rate": 1.8045764488491725e-05, "loss": 1.1518, "step": 2673 }, { "epoch": 0.23, "learning_rate": 1.8044135239756516e-05, "loss": 0.9525, "step": 2674 }, { "epoch": 0.23, "learning_rate": 1.8042505385763242e-05, "loss": 1.0838, "step": 2675 }, { "epoch": 0.23, "learning_rate": 1.8040874926634536e-05, "loss": 0.9522, "step": 2676 }, { "epoch": 0.23, "learning_rate": 1.803924386249308e-05, "loss": 1.1628, "step": 2677 }, { "epoch": 0.23, "learning_rate": 1.8037612193461598e-05, "loss": 1.1515, "step": 2678 }, { "epoch": 0.23, "learning_rate": 1.8035979919662855e-05, "loss": 1.1492, "step": 2679 }, { "epoch": 0.23, "learning_rate": 1.803434704121968e-05, "loss": 1.1537, "step": 2680 }, { "epoch": 0.23, "learning_rate": 1.8032713558254916e-05, "loss": 1.0865, "step": 2681 }, { "epoch": 0.23, "learning_rate": 1.8031079470891484e-05, "loss": 1.0878, "step": 2682 }, { "epoch": 0.23, "learning_rate": 1.8029444779252335e-05, "loss": 1.1239, "step": 2683 }, { "epoch": 0.23, "learning_rate": 1.802780948346046e-05, "loss": 1.0547, "step": 2684 }, { "epoch": 0.23, "learning_rate": 1.8026173583638904e-05, "loss": 1.1685, "step": 2685 }, { "epoch": 0.23, "learning_rate": 1.8024537079910756e-05, "loss": 1.0971, "step": 2686 }, { "epoch": 0.23, "learning_rate": 1.8022899972399157e-05, "loss": 1.0998, "step": 2687 }, { "epoch": 0.23, "learning_rate": 1.8021262261227278e-05, "loss": 1.0969, "step": 2688 }, { "epoch": 0.23, "learning_rate": 1.8019623946518345e-05, "loss": 1.1317, "step": 2689 }, { "epoch": 0.23, "learning_rate": 1.8017985028395633e-05, "loss": 1.1104, "step": 2690 }, { "epoch": 0.23, "learning_rate": 1.8016345506982452e-05, "loss": 1.0412, "step": 2691 }, { "epoch": 0.23, "learning_rate": 1.8014705382402168e-05, "loss": 1.1041, "step": 2692 }, { "epoch": 0.23, "learning_rate": 1.8013064654778186e-05, "loss": 1.1055, "step": 2693 }, { "epoch": 0.23, "learning_rate": 1.8011423324233958e-05, "loss": 1.1396, "step": 2694 }, { "epoch": 0.23, "learning_rate": 1.800978139089298e-05, "loss": 1.1411, "step": 2695 }, { "epoch": 0.23, "learning_rate": 1.8008138854878795e-05, "loss": 1.1306, "step": 2696 }, { "epoch": 0.23, "learning_rate": 1.8006495716314988e-05, "loss": 1.1685, "step": 2697 }, { "epoch": 0.23, "learning_rate": 1.8004851975325198e-05, "loss": 1.0805, "step": 2698 }, { "epoch": 0.23, "learning_rate": 1.8003207632033098e-05, "loss": 1.1286, "step": 2699 }, { "epoch": 0.23, "learning_rate": 1.800156268656242e-05, "loss": 1.0655, "step": 2700 }, { "epoch": 0.23, "learning_rate": 1.7999917139036928e-05, "loss": 1.1069, "step": 2701 }, { "epoch": 0.23, "learning_rate": 1.799827098958043e-05, "loss": 1.0971, "step": 2702 }, { "epoch": 0.23, "learning_rate": 1.7996624238316798e-05, "loss": 1.0868, "step": 2703 }, { "epoch": 0.23, "learning_rate": 1.7994976885369934e-05, "loss": 1.1137, "step": 2704 }, { "epoch": 0.23, "learning_rate": 1.799332893086378e-05, "loss": 1.1358, "step": 2705 }, { "epoch": 0.23, "learning_rate": 1.7991680374922344e-05, "loss": 1.1659, "step": 2706 }, { "epoch": 0.23, "learning_rate": 1.7990031217669653e-05, "loss": 1.1211, "step": 2707 }, { "epoch": 0.23, "learning_rate": 1.7988381459229805e-05, "loss": 1.1393, "step": 2708 }, { "epoch": 0.23, "learning_rate": 1.7986731099726928e-05, "loss": 1.1209, "step": 2709 }, { "epoch": 0.23, "learning_rate": 1.7985080139285197e-05, "loss": 1.1186, "step": 2710 }, { "epoch": 0.23, "learning_rate": 1.7983428578028834e-05, "loss": 1.1179, "step": 2711 }, { "epoch": 0.23, "learning_rate": 1.7981776416082105e-05, "loss": 1.0997, "step": 2712 }, { "epoch": 0.23, "learning_rate": 1.7980123653569328e-05, "loss": 1.1334, "step": 2713 }, { "epoch": 0.23, "learning_rate": 1.7978470290614855e-05, "loss": 1.107, "step": 2714 }, { "epoch": 0.23, "learning_rate": 1.797681632734309e-05, "loss": 1.0475, "step": 2715 }, { "epoch": 0.23, "learning_rate": 1.797516176387848e-05, "loss": 1.1558, "step": 2716 }, { "epoch": 0.23, "learning_rate": 1.7973506600345514e-05, "loss": 1.0674, "step": 2717 }, { "epoch": 0.23, "learning_rate": 1.7971850836868738e-05, "loss": 1.0818, "step": 2718 }, { "epoch": 0.23, "learning_rate": 1.7970194473572732e-05, "loss": 1.1817, "step": 2719 }, { "epoch": 0.23, "learning_rate": 1.7968537510582127e-05, "loss": 1.0774, "step": 2720 }, { "epoch": 0.23, "learning_rate": 1.796687994802159e-05, "loss": 1.0831, "step": 2721 }, { "epoch": 0.23, "learning_rate": 1.7965221786015844e-05, "loss": 1.0665, "step": 2722 }, { "epoch": 0.23, "learning_rate": 1.7963563024689652e-05, "loss": 1.1388, "step": 2723 }, { "epoch": 0.23, "learning_rate": 1.796190366416782e-05, "loss": 1.0455, "step": 2724 }, { "epoch": 0.23, "learning_rate": 1.7960243704575208e-05, "loss": 1.0833, "step": 2725 }, { "epoch": 0.23, "learning_rate": 1.795858314603671e-05, "loss": 1.147, "step": 2726 }, { "epoch": 0.23, "learning_rate": 1.795692198867727e-05, "loss": 1.1277, "step": 2727 }, { "epoch": 0.23, "learning_rate": 1.7955260232621886e-05, "loss": 1.1368, "step": 2728 }, { "epoch": 0.23, "learning_rate": 1.7953597877995578e-05, "loss": 1.0812, "step": 2729 }, { "epoch": 0.23, "learning_rate": 1.795193492492343e-05, "loss": 1.0827, "step": 2730 }, { "epoch": 0.23, "learning_rate": 1.7950271373530573e-05, "loss": 1.1015, "step": 2731 }, { "epoch": 0.23, "learning_rate": 1.7948607223942164e-05, "loss": 1.1266, "step": 2732 }, { "epoch": 0.23, "learning_rate": 1.7946942476283432e-05, "loss": 1.1266, "step": 2733 }, { "epoch": 0.23, "learning_rate": 1.7945277130679626e-05, "loss": 1.0711, "step": 2734 }, { "epoch": 0.23, "learning_rate": 1.794361118725605e-05, "loss": 1.0965, "step": 2735 }, { "epoch": 0.23, "learning_rate": 1.794194464613806e-05, "loss": 1.1354, "step": 2736 }, { "epoch": 0.23, "learning_rate": 1.7940277507451044e-05, "loss": 1.1153, "step": 2737 }, { "epoch": 0.23, "learning_rate": 1.793860977132044e-05, "loss": 1.1553, "step": 2738 }, { "epoch": 0.23, "learning_rate": 1.7936941437871737e-05, "loss": 1.0754, "step": 2739 }, { "epoch": 0.23, "learning_rate": 1.7935272507230465e-05, "loss": 1.0938, "step": 2740 }, { "epoch": 0.23, "learning_rate": 1.7933602979522195e-05, "loss": 1.0865, "step": 2741 }, { "epoch": 0.23, "learning_rate": 1.793193285487254e-05, "loss": 1.0962, "step": 2742 }, { "epoch": 0.23, "learning_rate": 1.7930262133407174e-05, "loss": 1.0572, "step": 2743 }, { "epoch": 0.23, "learning_rate": 1.7928590815251803e-05, "loss": 1.1477, "step": 2744 }, { "epoch": 0.23, "learning_rate": 1.7926918900532174e-05, "loss": 1.1269, "step": 2745 }, { "epoch": 0.23, "learning_rate": 1.7925246389374095e-05, "loss": 1.1384, "step": 2746 }, { "epoch": 0.23, "learning_rate": 1.7923573281903405e-05, "loss": 1.1058, "step": 2747 }, { "epoch": 0.23, "learning_rate": 1.792189957824599e-05, "loss": 1.0839, "step": 2748 }, { "epoch": 0.23, "learning_rate": 1.7920225278527787e-05, "loss": 1.0561, "step": 2749 }, { "epoch": 0.23, "learning_rate": 1.7918550382874767e-05, "loss": 1.1346, "step": 2750 }, { "epoch": 0.23, "learning_rate": 1.7916874891412964e-05, "loss": 1.1338, "step": 2751 }, { "epoch": 0.23, "learning_rate": 1.7915198804268438e-05, "loss": 1.0824, "step": 2752 }, { "epoch": 0.23, "learning_rate": 1.79135221215673e-05, "loss": 1.0542, "step": 2753 }, { "epoch": 0.23, "learning_rate": 1.791184484343571e-05, "loss": 1.0916, "step": 2754 }, { "epoch": 0.23, "learning_rate": 1.791016696999987e-05, "loss": 1.1183, "step": 2755 }, { "epoch": 0.23, "learning_rate": 1.7908488501386032e-05, "loss": 1.0607, "step": 2756 }, { "epoch": 0.23, "learning_rate": 1.7906809437720478e-05, "loss": 1.0959, "step": 2757 }, { "epoch": 0.23, "learning_rate": 1.790512977912955e-05, "loss": 1.1226, "step": 2758 }, { "epoch": 0.23, "learning_rate": 1.7903449525739628e-05, "loss": 1.0974, "step": 2759 }, { "epoch": 0.23, "learning_rate": 1.7901768677677133e-05, "loss": 1.1606, "step": 2760 }, { "epoch": 0.23, "learning_rate": 1.7900087235068546e-05, "loss": 1.0737, "step": 2761 }, { "epoch": 0.23, "learning_rate": 1.7898405198040377e-05, "loss": 1.0886, "step": 2762 }, { "epoch": 0.23, "learning_rate": 1.7896722566719182e-05, "loss": 1.0897, "step": 2763 }, { "epoch": 0.23, "learning_rate": 1.789503934123157e-05, "loss": 1.0839, "step": 2764 }, { "epoch": 0.23, "learning_rate": 1.789335552170419e-05, "loss": 1.1143, "step": 2765 }, { "epoch": 0.23, "learning_rate": 1.789167110826374e-05, "loss": 1.1557, "step": 2766 }, { "epoch": 0.23, "learning_rate": 1.7889986101036954e-05, "loss": 1.1765, "step": 2767 }, { "epoch": 0.23, "learning_rate": 1.7888300500150617e-05, "loss": 1.1593, "step": 2768 }, { "epoch": 0.23, "learning_rate": 1.7886614305731556e-05, "loss": 1.1145, "step": 2769 }, { "epoch": 0.23, "learning_rate": 1.7884927517906647e-05, "loss": 1.1201, "step": 2770 }, { "epoch": 0.23, "learning_rate": 1.7883240136802802e-05, "loss": 1.0754, "step": 2771 }, { "epoch": 0.23, "learning_rate": 1.788155216254699e-05, "loss": 1.1307, "step": 2772 }, { "epoch": 0.23, "learning_rate": 1.787986359526621e-05, "loss": 1.2062, "step": 2773 }, { "epoch": 0.23, "learning_rate": 1.7878174435087523e-05, "loss": 1.1437, "step": 2774 }, { "epoch": 0.24, "learning_rate": 1.787648468213802e-05, "loss": 1.106, "step": 2775 }, { "epoch": 0.24, "learning_rate": 1.787479433654484e-05, "loss": 1.0404, "step": 2776 }, { "epoch": 0.24, "learning_rate": 1.787310339843517e-05, "loss": 1.0902, "step": 2777 }, { "epoch": 0.24, "learning_rate": 1.7871411867936236e-05, "loss": 1.1206, "step": 2778 }, { "epoch": 0.24, "learning_rate": 1.786971974517532e-05, "loss": 1.107, "step": 2779 }, { "epoch": 0.24, "learning_rate": 1.7868027030279737e-05, "loss": 1.0697, "step": 2780 }, { "epoch": 0.24, "learning_rate": 1.7866333723376853e-05, "loss": 0.977, "step": 2781 }, { "epoch": 0.24, "learning_rate": 1.786463982459407e-05, "loss": 1.1218, "step": 2782 }, { "epoch": 0.24, "learning_rate": 1.7862945334058845e-05, "loss": 1.1241, "step": 2783 }, { "epoch": 0.24, "learning_rate": 1.7861250251898675e-05, "loss": 1.109, "step": 2784 }, { "epoch": 0.24, "learning_rate": 1.78595545782411e-05, "loss": 1.1045, "step": 2785 }, { "epoch": 0.24, "learning_rate": 1.7857858313213708e-05, "loss": 1.1137, "step": 2786 }, { "epoch": 0.24, "learning_rate": 1.785616145694413e-05, "loss": 1.174, "step": 2787 }, { "epoch": 0.24, "learning_rate": 1.785446400956004e-05, "loss": 1.1165, "step": 2788 }, { "epoch": 0.24, "learning_rate": 1.7852765971189153e-05, "loss": 1.1446, "step": 2789 }, { "epoch": 0.24, "learning_rate": 1.785106734195924e-05, "loss": 1.1124, "step": 2790 }, { "epoch": 0.24, "learning_rate": 1.7849368121998113e-05, "loss": 1.096, "step": 2791 }, { "epoch": 0.24, "learning_rate": 1.7847668311433612e-05, "loss": 0.9327, "step": 2792 }, { "epoch": 0.24, "learning_rate": 1.7845967910393646e-05, "loss": 1.1129, "step": 2793 }, { "epoch": 0.24, "learning_rate": 1.7844266919006145e-05, "loss": 1.0925, "step": 2794 }, { "epoch": 0.24, "learning_rate": 1.784256533739911e-05, "loss": 1.1531, "step": 2795 }, { "epoch": 0.24, "learning_rate": 1.7840863165700562e-05, "loss": 1.0987, "step": 2796 }, { "epoch": 0.24, "learning_rate": 1.783916040403858e-05, "loss": 1.2105, "step": 2797 }, { "epoch": 0.24, "learning_rate": 1.783745705254128e-05, "loss": 1.1246, "step": 2798 }, { "epoch": 0.24, "learning_rate": 1.7835753111336824e-05, "loss": 1.1223, "step": 2799 }, { "epoch": 0.24, "learning_rate": 1.7834048580553428e-05, "loss": 1.0767, "step": 2800 }, { "epoch": 0.24, "learning_rate": 1.783234346031934e-05, "loss": 1.1274, "step": 2801 }, { "epoch": 0.24, "learning_rate": 1.7830637750762858e-05, "loss": 1.1398, "step": 2802 }, { "epoch": 0.24, "learning_rate": 1.782893145201232e-05, "loss": 1.1223, "step": 2803 }, { "epoch": 0.24, "learning_rate": 1.7827224564196117e-05, "loss": 1.1344, "step": 2804 }, { "epoch": 0.24, "learning_rate": 1.7825517087442678e-05, "loss": 1.1122, "step": 2805 }, { "epoch": 0.24, "learning_rate": 1.782380902188047e-05, "loss": 1.1484, "step": 2806 }, { "epoch": 0.24, "learning_rate": 1.782210036763802e-05, "loss": 1.1585, "step": 2807 }, { "epoch": 0.24, "learning_rate": 1.7820391124843886e-05, "loss": 1.1366, "step": 2808 }, { "epoch": 0.24, "learning_rate": 1.7818681293626676e-05, "loss": 1.0958, "step": 2809 }, { "epoch": 0.24, "learning_rate": 1.7816970874115044e-05, "loss": 1.096, "step": 2810 }, { "epoch": 0.24, "learning_rate": 1.7815259866437682e-05, "loss": 1.0982, "step": 2811 }, { "epoch": 0.24, "learning_rate": 1.7813548270723334e-05, "loss": 1.084, "step": 2812 }, { "epoch": 0.24, "learning_rate": 1.7811836087100784e-05, "loss": 0.956, "step": 2813 }, { "epoch": 0.24, "learning_rate": 1.7810123315698856e-05, "loss": 1.0909, "step": 2814 }, { "epoch": 0.24, "learning_rate": 1.7808409956646422e-05, "loss": 1.1212, "step": 2815 }, { "epoch": 0.24, "learning_rate": 1.7806696010072403e-05, "loss": 0.9426, "step": 2816 }, { "epoch": 0.24, "learning_rate": 1.780498147610576e-05, "loss": 1.1074, "step": 2817 }, { "epoch": 0.24, "learning_rate": 1.78032663548755e-05, "loss": 1.1571, "step": 2818 }, { "epoch": 0.24, "learning_rate": 1.7801550646510662e-05, "loss": 1.1668, "step": 2819 }, { "epoch": 0.24, "learning_rate": 1.7799834351140353e-05, "loss": 1.0841, "step": 2820 }, { "epoch": 0.24, "learning_rate": 1.77981174688937e-05, "loss": 1.1587, "step": 2821 }, { "epoch": 0.24, "learning_rate": 1.7796399999899892e-05, "loss": 1.1366, "step": 2822 }, { "epoch": 0.24, "learning_rate": 1.7794681944288154e-05, "loss": 1.0721, "step": 2823 }, { "epoch": 0.24, "learning_rate": 1.7792963302187755e-05, "loss": 1.1424, "step": 2824 }, { "epoch": 0.24, "learning_rate": 1.7791244073728008e-05, "loss": 1.1243, "step": 2825 }, { "epoch": 0.24, "learning_rate": 1.7789524259038273e-05, "loss": 1.1197, "step": 2826 }, { "epoch": 0.24, "learning_rate": 1.7787803858247952e-05, "loss": 1.0717, "step": 2827 }, { "epoch": 0.24, "learning_rate": 1.7786082871486493e-05, "loss": 1.0997, "step": 2828 }, { "epoch": 0.24, "learning_rate": 1.778436129888338e-05, "loss": 1.1094, "step": 2829 }, { "epoch": 0.24, "learning_rate": 1.778263914056816e-05, "loss": 1.0804, "step": 2830 }, { "epoch": 0.24, "learning_rate": 1.7780916396670403e-05, "loss": 1.121, "step": 2831 }, { "epoch": 0.24, "learning_rate": 1.7779193067319734e-05, "loss": 1.0684, "step": 2832 }, { "epoch": 0.24, "learning_rate": 1.7777469152645822e-05, "loss": 1.0861, "step": 2833 }, { "epoch": 0.24, "learning_rate": 1.777574465277838e-05, "loss": 1.1185, "step": 2834 }, { "epoch": 0.24, "learning_rate": 1.7774019567847152e-05, "loss": 1.0923, "step": 2835 }, { "epoch": 0.24, "learning_rate": 1.777229389798195e-05, "loss": 1.0785, "step": 2836 }, { "epoch": 0.24, "learning_rate": 1.7770567643312608e-05, "loss": 1.0878, "step": 2837 }, { "epoch": 0.24, "learning_rate": 1.7768840803969017e-05, "loss": 1.1068, "step": 2838 }, { "epoch": 0.24, "learning_rate": 1.7767113380081112e-05, "loss": 1.1351, "step": 2839 }, { "epoch": 0.24, "learning_rate": 1.776538537177886e-05, "loss": 1.0578, "step": 2840 }, { "epoch": 0.24, "learning_rate": 1.7763656779192287e-05, "loss": 1.1241, "step": 2841 }, { "epoch": 0.24, "learning_rate": 1.7761927602451453e-05, "loss": 1.1534, "step": 2842 }, { "epoch": 0.24, "learning_rate": 1.7760197841686463e-05, "loss": 1.1048, "step": 2843 }, { "epoch": 0.24, "learning_rate": 1.775846749702747e-05, "loss": 1.1205, "step": 2844 }, { "epoch": 0.24, "learning_rate": 1.775673656860467e-05, "loss": 1.0768, "step": 2845 }, { "epoch": 0.24, "learning_rate": 1.77550050565483e-05, "loss": 1.1044, "step": 2846 }, { "epoch": 0.24, "learning_rate": 1.7753272960988645e-05, "loss": 1.1261, "step": 2847 }, { "epoch": 0.24, "learning_rate": 1.775154028205603e-05, "loss": 1.0974, "step": 2848 }, { "epoch": 0.24, "learning_rate": 1.7749807019880822e-05, "loss": 0.9725, "step": 2849 }, { "epoch": 0.24, "learning_rate": 1.7748073174593443e-05, "loss": 1.1804, "step": 2850 }, { "epoch": 0.24, "learning_rate": 1.774633874632435e-05, "loss": 1.1169, "step": 2851 }, { "epoch": 0.24, "learning_rate": 1.7744603735204033e-05, "loss": 1.107, "step": 2852 }, { "epoch": 0.24, "learning_rate": 1.774286814136305e-05, "loss": 1.0897, "step": 2853 }, { "epoch": 0.24, "learning_rate": 1.7741131964931995e-05, "loss": 1.1379, "step": 2854 }, { "epoch": 0.24, "learning_rate": 1.7739395206041487e-05, "loss": 1.1242, "step": 2855 }, { "epoch": 0.24, "learning_rate": 1.7737657864822215e-05, "loss": 1.0407, "step": 2856 }, { "epoch": 0.24, "learning_rate": 1.7735919941404895e-05, "loss": 1.1089, "step": 2857 }, { "epoch": 0.24, "learning_rate": 1.7734181435920294e-05, "loss": 1.1725, "step": 2858 }, { "epoch": 0.24, "learning_rate": 1.7732442348499222e-05, "loss": 1.0944, "step": 2859 }, { "epoch": 0.24, "learning_rate": 1.7730702679272527e-05, "loss": 1.0279, "step": 2860 }, { "epoch": 0.24, "learning_rate": 1.7728962428371113e-05, "loss": 1.0657, "step": 2861 }, { "epoch": 0.24, "learning_rate": 1.7727221595925913e-05, "loss": 1.1011, "step": 2862 }, { "epoch": 0.24, "learning_rate": 1.7725480182067913e-05, "loss": 1.0964, "step": 2863 }, { "epoch": 0.24, "learning_rate": 1.7723738186928145e-05, "loss": 1.0822, "step": 2864 }, { "epoch": 0.24, "learning_rate": 1.7721995610637673e-05, "loss": 1.1208, "step": 2865 }, { "epoch": 0.24, "learning_rate": 1.772025245332762e-05, "loss": 1.1411, "step": 2866 }, { "epoch": 0.24, "learning_rate": 1.7718508715129136e-05, "loss": 1.07, "step": 2867 }, { "epoch": 0.24, "learning_rate": 1.7716764396173435e-05, "loss": 1.1337, "step": 2868 }, { "epoch": 0.24, "learning_rate": 1.7715019496591752e-05, "loss": 1.0474, "step": 2869 }, { "epoch": 0.24, "learning_rate": 1.7713274016515385e-05, "loss": 1.1041, "step": 2870 }, { "epoch": 0.24, "learning_rate": 1.7711527956075665e-05, "loss": 1.1688, "step": 2871 }, { "epoch": 0.24, "learning_rate": 1.7709781315403966e-05, "loss": 1.0751, "step": 2872 }, { "epoch": 0.24, "learning_rate": 1.7708034094631714e-05, "loss": 1.1383, "step": 2873 }, { "epoch": 0.24, "learning_rate": 1.7706286293890375e-05, "loss": 1.1333, "step": 2874 }, { "epoch": 0.24, "learning_rate": 1.770453791331145e-05, "loss": 1.0867, "step": 2875 }, { "epoch": 0.24, "learning_rate": 1.7702788953026496e-05, "loss": 1.1421, "step": 2876 }, { "epoch": 0.24, "learning_rate": 1.770103941316711e-05, "loss": 1.0969, "step": 2877 }, { "epoch": 0.24, "learning_rate": 1.7699289293864923e-05, "loss": 1.1206, "step": 2878 }, { "epoch": 0.24, "learning_rate": 1.769753859525163e-05, "loss": 1.1104, "step": 2879 }, { "epoch": 0.24, "learning_rate": 1.7695787317458945e-05, "loss": 1.095, "step": 2880 }, { "epoch": 0.24, "learning_rate": 1.7694035460618647e-05, "loss": 1.0788, "step": 2881 }, { "epoch": 0.24, "learning_rate": 1.7692283024862545e-05, "loss": 1.1234, "step": 2882 }, { "epoch": 0.24, "learning_rate": 1.7690530010322494e-05, "loss": 1.0578, "step": 2883 }, { "epoch": 0.24, "learning_rate": 1.7688776417130404e-05, "loss": 1.0706, "step": 2884 }, { "epoch": 0.24, "learning_rate": 1.768702224541821e-05, "loss": 1.1429, "step": 2885 }, { "epoch": 0.24, "learning_rate": 1.7685267495317903e-05, "loss": 1.07, "step": 2886 }, { "epoch": 0.24, "learning_rate": 1.7683512166961514e-05, "loss": 1.1201, "step": 2887 }, { "epoch": 0.24, "learning_rate": 1.7681756260481114e-05, "loss": 1.0817, "step": 2888 }, { "epoch": 0.24, "learning_rate": 1.767999977600883e-05, "loss": 1.099, "step": 2889 }, { "epoch": 0.24, "learning_rate": 1.7678242713676813e-05, "loss": 1.1426, "step": 2890 }, { "epoch": 0.24, "learning_rate": 1.7676485073617278e-05, "loss": 1.0968, "step": 2891 }, { "epoch": 0.24, "learning_rate": 1.7674726855962467e-05, "loss": 1.0943, "step": 2892 }, { "epoch": 0.25, "learning_rate": 1.7672968060844675e-05, "loss": 1.0754, "step": 2893 }, { "epoch": 0.25, "learning_rate": 1.767120868839624e-05, "loss": 1.0721, "step": 2894 }, { "epoch": 0.25, "learning_rate": 1.766944873874953e-05, "loss": 1.0959, "step": 2895 }, { "epoch": 0.25, "learning_rate": 1.766768821203698e-05, "loss": 1.1498, "step": 2896 }, { "epoch": 0.25, "learning_rate": 1.766592710839105e-05, "loss": 1.153, "step": 2897 }, { "epoch": 0.25, "learning_rate": 1.766416542794425e-05, "loss": 1.0713, "step": 2898 }, { "epoch": 0.25, "learning_rate": 1.766240317082913e-05, "loss": 1.0961, "step": 2899 }, { "epoch": 0.25, "learning_rate": 1.7660640337178295e-05, "loss": 1.102, "step": 2900 }, { "epoch": 0.25, "learning_rate": 1.7658876927124374e-05, "loss": 1.138, "step": 2901 }, { "epoch": 0.25, "learning_rate": 1.7657112940800054e-05, "loss": 1.1138, "step": 2902 }, { "epoch": 0.25, "learning_rate": 1.7655348378338062e-05, "loss": 1.0654, "step": 2903 }, { "epoch": 0.25, "learning_rate": 1.765358323987117e-05, "loss": 1.0804, "step": 2904 }, { "epoch": 0.25, "learning_rate": 1.7651817525532182e-05, "loss": 1.1126, "step": 2905 }, { "epoch": 0.25, "learning_rate": 1.765005123545396e-05, "loss": 0.9969, "step": 2906 }, { "epoch": 0.25, "learning_rate": 1.7648284369769405e-05, "loss": 1.1214, "step": 2907 }, { "epoch": 0.25, "learning_rate": 1.7646516928611458e-05, "loss": 1.1166, "step": 2908 }, { "epoch": 0.25, "learning_rate": 1.76447489121131e-05, "loss": 1.0214, "step": 2909 }, { "epoch": 0.25, "learning_rate": 1.764298032040737e-05, "loss": 1.0788, "step": 2910 }, { "epoch": 0.25, "learning_rate": 1.7641211153627334e-05, "loss": 1.1482, "step": 2911 }, { "epoch": 0.25, "learning_rate": 1.7639441411906113e-05, "loss": 1.1213, "step": 2912 }, { "epoch": 0.25, "learning_rate": 1.763767109537686e-05, "loss": 1.109, "step": 2913 }, { "epoch": 0.25, "learning_rate": 1.7635900204172784e-05, "loss": 1.0744, "step": 2914 }, { "epoch": 0.25, "learning_rate": 1.7634128738427122e-05, "loss": 1.1352, "step": 2915 }, { "epoch": 0.25, "learning_rate": 1.7632356698273173e-05, "loss": 1.128, "step": 2916 }, { "epoch": 0.25, "learning_rate": 1.763058408384426e-05, "loss": 1.1228, "step": 2917 }, { "epoch": 0.25, "learning_rate": 1.7628810895273764e-05, "loss": 1.1084, "step": 2918 }, { "epoch": 0.25, "learning_rate": 1.7627037132695104e-05, "loss": 1.1253, "step": 2919 }, { "epoch": 0.25, "learning_rate": 1.7625262796241736e-05, "loss": 1.1515, "step": 2920 }, { "epoch": 0.25, "learning_rate": 1.762348788604717e-05, "loss": 1.1188, "step": 2921 }, { "epoch": 0.25, "learning_rate": 1.7621712402244956e-05, "loss": 1.1021, "step": 2922 }, { "epoch": 0.25, "learning_rate": 1.7619936344968675e-05, "loss": 1.1439, "step": 2923 }, { "epoch": 0.25, "learning_rate": 1.7618159714351973e-05, "loss": 1.1179, "step": 2924 }, { "epoch": 0.25, "learning_rate": 1.761638251052852e-05, "loss": 1.1048, "step": 2925 }, { "epoch": 0.25, "learning_rate": 1.7614604733632044e-05, "loss": 1.0794, "step": 2926 }, { "epoch": 0.25, "learning_rate": 1.76128263837963e-05, "loss": 1.0628, "step": 2927 }, { "epoch": 0.25, "learning_rate": 1.76110474611551e-05, "loss": 1.109, "step": 2928 }, { "epoch": 0.25, "learning_rate": 1.7609267965842295e-05, "loss": 1.1237, "step": 2929 }, { "epoch": 0.25, "learning_rate": 1.7607487897991772e-05, "loss": 1.0624, "step": 2930 }, { "epoch": 0.25, "learning_rate": 1.7605707257737475e-05, "loss": 1.0926, "step": 2931 }, { "epoch": 0.25, "learning_rate": 1.760392604521338e-05, "loss": 1.0954, "step": 2932 }, { "epoch": 0.25, "learning_rate": 1.7602144260553503e-05, "loss": 1.1286, "step": 2933 }, { "epoch": 0.25, "learning_rate": 1.760036190389192e-05, "loss": 1.1277, "step": 2934 }, { "epoch": 0.25, "learning_rate": 1.7598578975362733e-05, "loss": 1.0853, "step": 2935 }, { "epoch": 0.25, "learning_rate": 1.7596795475100096e-05, "loss": 1.1109, "step": 2936 }, { "epoch": 0.25, "learning_rate": 1.75950114032382e-05, "loss": 1.1194, "step": 2937 }, { "epoch": 0.25, "learning_rate": 1.7593226759911286e-05, "loss": 1.0998, "step": 2938 }, { "epoch": 0.25, "learning_rate": 1.7591441545253632e-05, "loss": 1.0455, "step": 2939 }, { "epoch": 0.25, "learning_rate": 1.758965575939956e-05, "loss": 1.0764, "step": 2940 }, { "epoch": 0.25, "learning_rate": 1.7587869402483443e-05, "loss": 1.053, "step": 2941 }, { "epoch": 0.25, "learning_rate": 1.7586082474639688e-05, "loss": 1.1505, "step": 2942 }, { "epoch": 0.25, "learning_rate": 1.758429497600274e-05, "loss": 1.0117, "step": 2943 }, { "epoch": 0.25, "learning_rate": 1.7582506906707106e-05, "loss": 1.0602, "step": 2944 }, { "epoch": 0.25, "learning_rate": 1.758071826688731e-05, "loss": 1.0766, "step": 2945 }, { "epoch": 0.25, "learning_rate": 1.7578929056677944e-05, "loss": 1.1117, "step": 2946 }, { "epoch": 0.25, "learning_rate": 1.7577139276213627e-05, "loss": 1.109, "step": 2947 }, { "epoch": 0.25, "learning_rate": 1.7575348925629035e-05, "loss": 1.1346, "step": 2948 }, { "epoch": 0.25, "learning_rate": 1.7573558005058863e-05, "loss": 1.0933, "step": 2949 }, { "epoch": 0.25, "learning_rate": 1.7571766514637876e-05, "loss": 1.1124, "step": 2950 }, { "epoch": 0.25, "learning_rate": 1.7569974454500866e-05, "loss": 1.1064, "step": 2951 }, { "epoch": 0.25, "learning_rate": 1.7568181824782663e-05, "loss": 1.0802, "step": 2952 }, { "epoch": 0.25, "learning_rate": 1.756638862561816e-05, "loss": 1.1033, "step": 2953 }, { "epoch": 0.25, "learning_rate": 1.756459485714228e-05, "loss": 1.1324, "step": 2954 }, { "epoch": 0.25, "learning_rate": 1.7562800519489986e-05, "loss": 1.0841, "step": 2955 }, { "epoch": 0.25, "learning_rate": 1.7561005612796286e-05, "loss": 1.0434, "step": 2956 }, { "epoch": 0.25, "learning_rate": 1.7559210137196237e-05, "loss": 1.1484, "step": 2957 }, { "epoch": 0.25, "learning_rate": 1.7557414092824935e-05, "loss": 1.1313, "step": 2958 }, { "epoch": 0.25, "learning_rate": 1.7555617479817515e-05, "loss": 1.1316, "step": 2959 }, { "epoch": 0.25, "learning_rate": 1.7553820298309157e-05, "loss": 1.1033, "step": 2960 }, { "epoch": 0.25, "learning_rate": 1.7552022548435088e-05, "loss": 1.14, "step": 2961 }, { "epoch": 0.25, "learning_rate": 1.7550224230330575e-05, "loss": 1.1391, "step": 2962 }, { "epoch": 0.25, "learning_rate": 1.7548425344130928e-05, "loss": 1.0977, "step": 2963 }, { "epoch": 0.25, "learning_rate": 1.754662588997149e-05, "loss": 0.942, "step": 2964 }, { "epoch": 0.25, "learning_rate": 1.754482586798767e-05, "loss": 1.0881, "step": 2965 }, { "epoch": 0.25, "learning_rate": 1.7543025278314894e-05, "loss": 1.138, "step": 2966 }, { "epoch": 0.25, "learning_rate": 1.754122412108865e-05, "loss": 1.1254, "step": 2967 }, { "epoch": 0.25, "learning_rate": 1.753942239644445e-05, "loss": 1.1575, "step": 2968 }, { "epoch": 0.25, "learning_rate": 1.7537620104517876e-05, "loss": 1.064, "step": 2969 }, { "epoch": 0.25, "learning_rate": 1.7535817245444523e-05, "loss": 1.08, "step": 2970 }, { "epoch": 0.25, "learning_rate": 1.753401381936005e-05, "loss": 1.0913, "step": 2971 }, { "epoch": 0.25, "learning_rate": 1.7532209826400142e-05, "loss": 1.0747, "step": 2972 }, { "epoch": 0.25, "learning_rate": 1.7530405266700542e-05, "loss": 1.1076, "step": 2973 }, { "epoch": 0.25, "learning_rate": 1.752860014039703e-05, "loss": 1.0965, "step": 2974 }, { "epoch": 0.25, "learning_rate": 1.7526794447625427e-05, "loss": 1.1064, "step": 2975 }, { "epoch": 0.25, "learning_rate": 1.752498818852159e-05, "loss": 1.1367, "step": 2976 }, { "epoch": 0.25, "learning_rate": 1.7523181363221433e-05, "loss": 1.1271, "step": 2977 }, { "epoch": 0.25, "learning_rate": 1.75213739718609e-05, "loss": 1.1016, "step": 2978 }, { "epoch": 0.25, "learning_rate": 1.751956601457599e-05, "loss": 1.1337, "step": 2979 }, { "epoch": 0.25, "learning_rate": 1.751775749150274e-05, "loss": 1.1201, "step": 2980 }, { "epoch": 0.25, "learning_rate": 1.751594840277721e-05, "loss": 1.0987, "step": 2981 }, { "epoch": 0.25, "learning_rate": 1.751413874853554e-05, "loss": 1.0972, "step": 2982 }, { "epoch": 0.25, "learning_rate": 1.751232852891388e-05, "loss": 1.0892, "step": 2983 }, { "epoch": 0.25, "learning_rate": 1.751051774404844e-05, "loss": 1.1449, "step": 2984 }, { "epoch": 0.25, "learning_rate": 1.7508706394075462e-05, "loss": 1.1214, "step": 2985 }, { "epoch": 0.25, "learning_rate": 1.750689447913124e-05, "loss": 1.1259, "step": 2986 }, { "epoch": 0.25, "learning_rate": 1.7505081999352105e-05, "loss": 1.1146, "step": 2987 }, { "epoch": 0.25, "learning_rate": 1.7503268954874436e-05, "loss": 1.082, "step": 2988 }, { "epoch": 0.25, "learning_rate": 1.7501455345834647e-05, "loss": 1.1264, "step": 2989 }, { "epoch": 0.25, "learning_rate": 1.7499641172369192e-05, "loss": 1.1128, "step": 2990 }, { "epoch": 0.25, "learning_rate": 1.7497826434614585e-05, "loss": 1.072, "step": 2991 }, { "epoch": 0.25, "learning_rate": 1.7496011132707362e-05, "loss": 1.1066, "step": 2992 }, { "epoch": 0.25, "learning_rate": 1.7494195266784115e-05, "loss": 1.1004, "step": 2993 }, { "epoch": 0.25, "learning_rate": 1.749237883698147e-05, "loss": 1.1374, "step": 2994 }, { "epoch": 0.25, "learning_rate": 1.74905618434361e-05, "loss": 1.1155, "step": 2995 }, { "epoch": 0.25, "learning_rate": 1.7488744286284727e-05, "loss": 1.0326, "step": 2996 }, { "epoch": 0.25, "learning_rate": 1.7486926165664098e-05, "loss": 1.1691, "step": 2997 }, { "epoch": 0.25, "learning_rate": 1.7485107481711014e-05, "loss": 1.0903, "step": 2998 }, { "epoch": 0.25, "learning_rate": 1.748328823456232e-05, "loss": 1.0759, "step": 2999 }, { "epoch": 0.25, "learning_rate": 1.7481468424354895e-05, "loss": 1.087, "step": 3000 }, { "epoch": 0.25, "learning_rate": 1.7479648051225672e-05, "loss": 1.0914, "step": 3001 }, { "epoch": 0.25, "learning_rate": 1.7477827115311617e-05, "loss": 1.1149, "step": 3002 }, { "epoch": 0.25, "learning_rate": 1.7476005616749738e-05, "loss": 1.1173, "step": 3003 }, { "epoch": 0.25, "learning_rate": 1.7474183555677093e-05, "loss": 1.0947, "step": 3004 }, { "epoch": 0.25, "learning_rate": 1.7472360932230777e-05, "loss": 1.1121, "step": 3005 }, { "epoch": 0.25, "learning_rate": 1.7470537746547927e-05, "loss": 1.0832, "step": 3006 }, { "epoch": 0.25, "learning_rate": 1.746871399876572e-05, "loss": 1.0885, "step": 3007 }, { "epoch": 0.25, "learning_rate": 1.7466889689021387e-05, "loss": 1.1353, "step": 3008 }, { "epoch": 0.25, "learning_rate": 1.7465064817452186e-05, "loss": 1.0775, "step": 3009 }, { "epoch": 0.25, "learning_rate": 1.7463239384195428e-05, "loss": 1.0915, "step": 3010 }, { "epoch": 0.25, "learning_rate": 1.7461413389388463e-05, "loss": 1.154, "step": 3011 }, { "epoch": 0.26, "learning_rate": 1.7459586833168674e-05, "loss": 1.099, "step": 3012 }, { "epoch": 0.26, "learning_rate": 1.7457759715673508e-05, "loss": 1.079, "step": 3013 }, { "epoch": 0.26, "learning_rate": 1.7455932037040437e-05, "loss": 1.1065, "step": 3014 }, { "epoch": 0.26, "learning_rate": 1.7454103797406972e-05, "loss": 1.139, "step": 3015 }, { "epoch": 0.26, "learning_rate": 1.745227499691068e-05, "loss": 1.1366, "step": 3016 }, { "epoch": 0.26, "learning_rate": 1.7450445635689165e-05, "loss": 1.1175, "step": 3017 }, { "epoch": 0.26, "learning_rate": 1.744861571388007e-05, "loss": 1.0475, "step": 3018 }, { "epoch": 0.26, "learning_rate": 1.7446785231621084e-05, "loss": 1.1171, "step": 3019 }, { "epoch": 0.26, "learning_rate": 1.7444954189049935e-05, "loss": 1.1198, "step": 3020 }, { "epoch": 0.26, "learning_rate": 1.7443122586304394e-05, "loss": 1.0581, "step": 3021 }, { "epoch": 0.26, "learning_rate": 1.7441290423522272e-05, "loss": 1.1267, "step": 3022 }, { "epoch": 0.26, "learning_rate": 1.7439457700841434e-05, "loss": 1.1103, "step": 3023 }, { "epoch": 0.26, "learning_rate": 1.7437624418399767e-05, "loss": 1.1087, "step": 3024 }, { "epoch": 0.26, "learning_rate": 1.743579057633522e-05, "loss": 1.125, "step": 3025 }, { "epoch": 0.26, "learning_rate": 1.743395617478577e-05, "loss": 1.1272, "step": 3026 }, { "epoch": 0.26, "learning_rate": 1.7432121213889442e-05, "loss": 1.1718, "step": 3027 }, { "epoch": 0.26, "learning_rate": 1.7430285693784305e-05, "loss": 1.1041, "step": 3028 }, { "epoch": 0.26, "learning_rate": 1.7428449614608463e-05, "loss": 1.09, "step": 3029 }, { "epoch": 0.26, "learning_rate": 1.7426612976500076e-05, "loss": 1.1155, "step": 3030 }, { "epoch": 0.26, "learning_rate": 1.742477577959732e-05, "loss": 1.0464, "step": 3031 }, { "epoch": 0.26, "learning_rate": 1.742293802403845e-05, "loss": 1.1294, "step": 3032 }, { "epoch": 0.26, "learning_rate": 1.7421099709961728e-05, "loss": 1.1048, "step": 3033 }, { "epoch": 0.26, "learning_rate": 1.7419260837505475e-05, "loss": 1.1308, "step": 3034 }, { "epoch": 0.26, "learning_rate": 1.7417421406808052e-05, "loss": 1.1078, "step": 3035 }, { "epoch": 0.26, "learning_rate": 1.7415581418007867e-05, "loss": 1.1009, "step": 3036 }, { "epoch": 0.26, "learning_rate": 1.7413740871243363e-05, "loss": 1.1128, "step": 3037 }, { "epoch": 0.26, "learning_rate": 1.741189976665302e-05, "loss": 1.1333, "step": 3038 }, { "epoch": 0.26, "learning_rate": 1.7410058104375377e-05, "loss": 1.0727, "step": 3039 }, { "epoch": 0.26, "learning_rate": 1.7408215884548997e-05, "loss": 1.0342, "step": 3040 }, { "epoch": 0.26, "learning_rate": 1.7406373107312494e-05, "loss": 1.1534, "step": 3041 }, { "epoch": 0.26, "learning_rate": 1.740452977280452e-05, "loss": 1.0707, "step": 3042 }, { "epoch": 0.26, "learning_rate": 1.7402685881163778e-05, "loss": 1.0305, "step": 3043 }, { "epoch": 0.26, "learning_rate": 1.7400841432529006e-05, "loss": 1.1134, "step": 3044 }, { "epoch": 0.26, "learning_rate": 1.7398996427038975e-05, "loss": 1.1075, "step": 3045 }, { "epoch": 0.26, "learning_rate": 1.739715086483252e-05, "loss": 1.0355, "step": 3046 }, { "epoch": 0.26, "learning_rate": 1.7395304746048497e-05, "loss": 1.1447, "step": 3047 }, { "epoch": 0.26, "learning_rate": 1.7393458070825812e-05, "loss": 0.9023, "step": 3048 }, { "epoch": 0.26, "learning_rate": 1.7391610839303418e-05, "loss": 1.0647, "step": 3049 }, { "epoch": 0.26, "learning_rate": 1.7389763051620295e-05, "loss": 1.1248, "step": 3050 }, { "epoch": 0.26, "learning_rate": 1.7387914707915485e-05, "loss": 1.131, "step": 3051 }, { "epoch": 0.26, "learning_rate": 1.7386065808328055e-05, "loss": 1.0782, "step": 3052 }, { "epoch": 0.26, "learning_rate": 1.7384216352997124e-05, "loss": 1.0833, "step": 3053 }, { "epoch": 0.26, "learning_rate": 1.7382366342061847e-05, "loss": 1.1347, "step": 3054 }, { "epoch": 0.26, "learning_rate": 1.7380515775661424e-05, "loss": 1.1357, "step": 3055 }, { "epoch": 0.26, "learning_rate": 1.7378664653935092e-05, "loss": 1.1075, "step": 3056 }, { "epoch": 0.26, "learning_rate": 1.737681297702214e-05, "loss": 1.1317, "step": 3057 }, { "epoch": 0.26, "learning_rate": 1.7374960745061882e-05, "loss": 0.9094, "step": 3058 }, { "epoch": 0.26, "learning_rate": 1.7373107958193696e-05, "loss": 1.1231, "step": 3059 }, { "epoch": 0.26, "learning_rate": 1.737125461655698e-05, "loss": 1.104, "step": 3060 }, { "epoch": 0.26, "learning_rate": 1.736940072029119e-05, "loss": 1.1379, "step": 3061 }, { "epoch": 0.26, "learning_rate": 1.736754626953581e-05, "loss": 1.1694, "step": 3062 }, { "epoch": 0.26, "learning_rate": 1.7365691264430382e-05, "loss": 1.0774, "step": 3063 }, { "epoch": 0.26, "learning_rate": 1.7363835705114475e-05, "loss": 1.1491, "step": 3064 }, { "epoch": 0.26, "learning_rate": 1.7361979591727703e-05, "loss": 1.0768, "step": 3065 }, { "epoch": 0.26, "learning_rate": 1.736012292440973e-05, "loss": 1.1446, "step": 3066 }, { "epoch": 0.26, "learning_rate": 1.7358265703300255e-05, "loss": 1.0683, "step": 3067 }, { "epoch": 0.26, "learning_rate": 1.7356407928539016e-05, "loss": 1.0831, "step": 3068 }, { "epoch": 0.26, "learning_rate": 1.735454960026579e-05, "loss": 1.1462, "step": 3069 }, { "epoch": 0.26, "learning_rate": 1.7352690718620415e-05, "loss": 1.0956, "step": 3070 }, { "epoch": 0.26, "learning_rate": 1.7350831283742747e-05, "loss": 1.1037, "step": 3071 }, { "epoch": 0.26, "learning_rate": 1.7348971295772704e-05, "loss": 1.0956, "step": 3072 }, { "epoch": 0.26, "learning_rate": 1.7347110754850222e-05, "loss": 1.1359, "step": 3073 }, { "epoch": 0.26, "learning_rate": 1.7345249661115306e-05, "loss": 1.116, "step": 3074 }, { "epoch": 0.26, "learning_rate": 1.734338801470798e-05, "loss": 1.103, "step": 3075 }, { "epoch": 0.26, "learning_rate": 1.7341525815768316e-05, "loss": 1.053, "step": 3076 }, { "epoch": 0.26, "learning_rate": 1.7339663064436437e-05, "loss": 1.0702, "step": 3077 }, { "epoch": 0.26, "learning_rate": 1.7337799760852497e-05, "loss": 1.1235, "step": 3078 }, { "epoch": 0.26, "learning_rate": 1.73359359051567e-05, "loss": 1.0858, "step": 3079 }, { "epoch": 0.26, "learning_rate": 1.733407149748928e-05, "loss": 1.0728, "step": 3080 }, { "epoch": 0.26, "learning_rate": 1.733220653799052e-05, "loss": 1.1241, "step": 3081 }, { "epoch": 0.26, "learning_rate": 1.7330341026800744e-05, "loss": 1.1151, "step": 3082 }, { "epoch": 0.26, "learning_rate": 1.732847496406032e-05, "loss": 1.1156, "step": 3083 }, { "epoch": 0.26, "learning_rate": 1.732660834990965e-05, "loss": 1.0582, "step": 3084 }, { "epoch": 0.26, "learning_rate": 1.7324741184489186e-05, "loss": 1.0618, "step": 3085 }, { "epoch": 0.26, "learning_rate": 1.7322873467939417e-05, "loss": 1.1083, "step": 3086 }, { "epoch": 0.26, "learning_rate": 1.732100520040087e-05, "loss": 1.1387, "step": 3087 }, { "epoch": 0.26, "learning_rate": 1.7319136382014123e-05, "loss": 1.0924, "step": 3088 }, { "epoch": 0.26, "learning_rate": 1.7317267012919784e-05, "loss": 1.0138, "step": 3089 }, { "epoch": 0.26, "learning_rate": 1.7315397093258516e-05, "loss": 1.1107, "step": 3090 }, { "epoch": 0.26, "learning_rate": 1.731352662317101e-05, "loss": 1.0885, "step": 3091 }, { "epoch": 0.26, "learning_rate": 1.7311655602798006e-05, "loss": 1.0283, "step": 3092 }, { "epoch": 0.26, "learning_rate": 1.7309784032280284e-05, "loss": 1.1206, "step": 3093 }, { "epoch": 0.26, "learning_rate": 1.7307911911758662e-05, "loss": 1.0717, "step": 3094 }, { "epoch": 0.26, "learning_rate": 1.730603924137401e-05, "loss": 1.144, "step": 3095 }, { "epoch": 0.26, "learning_rate": 1.7304166021267225e-05, "loss": 1.1407, "step": 3096 }, { "epoch": 0.26, "learning_rate": 1.7302292251579248e-05, "loss": 1.1514, "step": 3097 }, { "epoch": 0.26, "learning_rate": 1.7300417932451075e-05, "loss": 1.0885, "step": 3098 }, { "epoch": 0.26, "learning_rate": 1.729854306402373e-05, "loss": 1.1219, "step": 3099 }, { "epoch": 0.26, "learning_rate": 1.7296667646438282e-05, "loss": 1.1503, "step": 3100 }, { "epoch": 0.26, "learning_rate": 1.729479167983584e-05, "loss": 1.0893, "step": 3101 }, { "epoch": 0.26, "learning_rate": 1.729291516435756e-05, "loss": 1.0722, "step": 3102 }, { "epoch": 0.26, "learning_rate": 1.729103810014463e-05, "loss": 1.1082, "step": 3103 }, { "epoch": 0.26, "learning_rate": 1.728916048733829e-05, "loss": 0.9097, "step": 3104 }, { "epoch": 0.26, "learning_rate": 1.7287282326079808e-05, "loss": 1.0818, "step": 3105 }, { "epoch": 0.26, "learning_rate": 1.728540361651051e-05, "loss": 1.1307, "step": 3106 }, { "epoch": 0.26, "learning_rate": 1.7283524358771746e-05, "loss": 1.1, "step": 3107 }, { "epoch": 0.26, "learning_rate": 1.728164455300492e-05, "loss": 1.0824, "step": 3108 }, { "epoch": 0.26, "learning_rate": 1.727976419935147e-05, "loss": 0.9903, "step": 3109 }, { "epoch": 0.26, "learning_rate": 1.7277883297952878e-05, "loss": 1.0681, "step": 3110 }, { "epoch": 0.26, "learning_rate": 1.7276001848950672e-05, "loss": 1.1506, "step": 3111 }, { "epoch": 0.26, "learning_rate": 1.727411985248641e-05, "loss": 1.133, "step": 3112 }, { "epoch": 0.26, "learning_rate": 1.72722373087017e-05, "loss": 1.1046, "step": 3113 }, { "epoch": 0.26, "learning_rate": 1.7270354217738192e-05, "loss": 1.13, "step": 3114 }, { "epoch": 0.26, "learning_rate": 1.726847057973757e-05, "loss": 1.1423, "step": 3115 }, { "epoch": 0.26, "learning_rate": 1.726658639484156e-05, "loss": 1.1483, "step": 3116 }, { "epoch": 0.26, "learning_rate": 1.726470166319194e-05, "loss": 1.1153, "step": 3117 }, { "epoch": 0.26, "learning_rate": 1.7262816384930515e-05, "loss": 1.1349, "step": 3118 }, { "epoch": 0.26, "learning_rate": 1.726093056019914e-05, "loss": 1.0854, "step": 3119 }, { "epoch": 0.26, "learning_rate": 1.7259044189139705e-05, "loss": 1.0809, "step": 3120 }, { "epoch": 0.26, "learning_rate": 1.725715727189415e-05, "loss": 1.0727, "step": 3121 }, { "epoch": 0.26, "learning_rate": 1.7255269808604447e-05, "loss": 1.0722, "step": 3122 }, { "epoch": 0.26, "learning_rate": 1.7253381799412617e-05, "loss": 1.0761, "step": 3123 }, { "epoch": 0.26, "learning_rate": 1.7251493244460712e-05, "loss": 1.1152, "step": 3124 }, { "epoch": 0.26, "learning_rate": 1.7249604143890836e-05, "loss": 1.0652, "step": 3125 }, { "epoch": 0.26, "learning_rate": 1.7247714497845125e-05, "loss": 1.0676, "step": 3126 }, { "epoch": 0.26, "learning_rate": 1.7245824306465762e-05, "loss": 1.0794, "step": 3127 }, { "epoch": 0.26, "learning_rate": 1.7243933569894968e-05, "loss": 0.9929, "step": 3128 }, { "epoch": 0.26, "learning_rate": 1.724204228827501e-05, "loss": 1.0445, "step": 3129 }, { "epoch": 0.27, "learning_rate": 1.7240150461748188e-05, "loss": 1.1505, "step": 3130 }, { "epoch": 0.27, "learning_rate": 1.7238258090456846e-05, "loss": 0.9486, "step": 3131 }, { "epoch": 0.27, "learning_rate": 1.7236365174543374e-05, "loss": 1.1359, "step": 3132 }, { "epoch": 0.27, "learning_rate": 1.7234471714150198e-05, "loss": 1.081, "step": 3133 }, { "epoch": 0.27, "learning_rate": 1.7232577709419786e-05, "loss": 1.1589, "step": 3134 }, { "epoch": 0.27, "learning_rate": 1.7230683160494645e-05, "loss": 1.0781, "step": 3135 }, { "epoch": 0.27, "learning_rate": 1.7228788067517326e-05, "loss": 1.1364, "step": 3136 }, { "epoch": 0.27, "learning_rate": 1.7226892430630423e-05, "loss": 1.1245, "step": 3137 }, { "epoch": 0.27, "learning_rate": 1.722499624997656e-05, "loss": 1.0578, "step": 3138 }, { "epoch": 0.27, "learning_rate": 1.722309952569842e-05, "loss": 1.07, "step": 3139 }, { "epoch": 0.27, "learning_rate": 1.7221202257938706e-05, "loss": 1.1054, "step": 3140 }, { "epoch": 0.27, "learning_rate": 1.721930444684018e-05, "loss": 1.117, "step": 3141 }, { "epoch": 0.27, "learning_rate": 1.721740609254564e-05, "loss": 1.1131, "step": 3142 }, { "epoch": 0.27, "learning_rate": 1.7215507195197913e-05, "loss": 1.0904, "step": 3143 }, { "epoch": 0.27, "learning_rate": 1.7213607754939883e-05, "loss": 0.9994, "step": 3144 }, { "epoch": 0.27, "learning_rate": 1.7211707771914463e-05, "loss": 1.04, "step": 3145 }, { "epoch": 0.27, "learning_rate": 1.7209807246264616e-05, "loss": 1.0459, "step": 3146 }, { "epoch": 0.27, "learning_rate": 1.720790617813334e-05, "loss": 1.1203, "step": 3147 }, { "epoch": 0.27, "learning_rate": 1.7206004567663675e-05, "loss": 1.1525, "step": 3148 }, { "epoch": 0.27, "learning_rate": 1.7204102414998706e-05, "loss": 1.1056, "step": 3149 }, { "epoch": 0.27, "learning_rate": 1.7202199720281548e-05, "loss": 0.997, "step": 3150 }, { "epoch": 0.27, "learning_rate": 1.7200296483655373e-05, "loss": 1.1467, "step": 3151 }, { "epoch": 0.27, "learning_rate": 1.7198392705263375e-05, "loss": 1.0531, "step": 3152 }, { "epoch": 0.27, "learning_rate": 1.7196488385248803e-05, "loss": 1.1615, "step": 3153 }, { "epoch": 0.27, "learning_rate": 1.7194583523754942e-05, "loss": 1.172, "step": 3154 }, { "epoch": 0.27, "learning_rate": 1.719267812092512e-05, "loss": 1.0909, "step": 3155 }, { "epoch": 0.27, "learning_rate": 1.71907721769027e-05, "loss": 1.1008, "step": 3156 }, { "epoch": 0.27, "learning_rate": 1.718886569183109e-05, "loss": 1.0872, "step": 3157 }, { "epoch": 0.27, "learning_rate": 1.7186958665853737e-05, "loss": 1.1102, "step": 3158 }, { "epoch": 0.27, "learning_rate": 1.7185051099114132e-05, "loss": 1.0893, "step": 3159 }, { "epoch": 0.27, "learning_rate": 1.7183142991755807e-05, "loss": 1.0536, "step": 3160 }, { "epoch": 0.27, "learning_rate": 1.718123434392233e-05, "loss": 1.0864, "step": 3161 }, { "epoch": 0.27, "learning_rate": 1.717932515575731e-05, "loss": 1.0816, "step": 3162 }, { "epoch": 0.27, "learning_rate": 1.7177415427404396e-05, "loss": 1.0497, "step": 3163 }, { "epoch": 0.27, "learning_rate": 1.7175505159007287e-05, "loss": 1.1028, "step": 3164 }, { "epoch": 0.27, "learning_rate": 1.717359435070971e-05, "loss": 1.0716, "step": 3165 }, { "epoch": 0.27, "learning_rate": 1.7171683002655442e-05, "loss": 1.0849, "step": 3166 }, { "epoch": 0.27, "learning_rate": 1.7169771114988294e-05, "loss": 1.0549, "step": 3167 }, { "epoch": 0.27, "learning_rate": 1.7167858687852123e-05, "loss": 1.1306, "step": 3168 }, { "epoch": 0.27, "learning_rate": 1.7165945721390823e-05, "loss": 1.1611, "step": 3169 }, { "epoch": 0.27, "learning_rate": 1.716403221574833e-05, "loss": 1.1249, "step": 3170 }, { "epoch": 0.27, "learning_rate": 1.716211817106862e-05, "loss": 1.1013, "step": 3171 }, { "epoch": 0.27, "learning_rate": 1.7160203587495707e-05, "loss": 1.1368, "step": 3172 }, { "epoch": 0.27, "learning_rate": 1.7158288465173654e-05, "loss": 1.1058, "step": 3173 }, { "epoch": 0.27, "learning_rate": 1.7156372804246552e-05, "loss": 1.0817, "step": 3174 }, { "epoch": 0.27, "learning_rate": 1.715445660485855e-05, "loss": 1.0988, "step": 3175 }, { "epoch": 0.27, "learning_rate": 1.7152539867153817e-05, "loss": 1.1229, "step": 3176 }, { "epoch": 0.27, "learning_rate": 1.715062259127658e-05, "loss": 1.118, "step": 3177 }, { "epoch": 0.27, "learning_rate": 1.714870477737109e-05, "loss": 1.1167, "step": 3178 }, { "epoch": 0.27, "learning_rate": 1.7146786425581657e-05, "loss": 1.1195, "step": 3179 }, { "epoch": 0.27, "learning_rate": 1.7144867536052613e-05, "loss": 1.1015, "step": 3180 }, { "epoch": 0.27, "learning_rate": 1.7142948108928345e-05, "loss": 1.0598, "step": 3181 }, { "epoch": 0.27, "learning_rate": 1.7141028144353277e-05, "loss": 1.115, "step": 3182 }, { "epoch": 0.27, "learning_rate": 1.7139107642471864e-05, "loss": 1.1593, "step": 3183 }, { "epoch": 0.27, "learning_rate": 1.7137186603428614e-05, "loss": 1.1539, "step": 3184 }, { "epoch": 0.27, "learning_rate": 1.7135265027368068e-05, "loss": 1.1253, "step": 3185 }, { "epoch": 0.27, "learning_rate": 1.713334291443481e-05, "loss": 1.0748, "step": 3186 }, { "epoch": 0.27, "learning_rate": 1.7131420264773468e-05, "loss": 1.0983, "step": 3187 }, { "epoch": 0.27, "learning_rate": 1.71294970785287e-05, "loss": 1.1098, "step": 3188 }, { "epoch": 0.27, "learning_rate": 1.7127573355845214e-05, "loss": 1.0863, "step": 3189 }, { "epoch": 0.27, "learning_rate": 1.7125649096867753e-05, "loss": 1.1673, "step": 3190 }, { "epoch": 0.27, "learning_rate": 1.7123724301741104e-05, "loss": 1.0908, "step": 3191 }, { "epoch": 0.27, "learning_rate": 1.7121798970610095e-05, "loss": 1.095, "step": 3192 }, { "epoch": 0.27, "learning_rate": 1.711987310361958e-05, "loss": 1.1432, "step": 3193 }, { "epoch": 0.27, "learning_rate": 1.7117946700914485e-05, "loss": 1.0818, "step": 3194 }, { "epoch": 0.27, "learning_rate": 1.7116019762639743e-05, "loss": 1.0939, "step": 3195 }, { "epoch": 0.27, "learning_rate": 1.7114092288940344e-05, "loss": 1.0723, "step": 3196 }, { "epoch": 0.27, "learning_rate": 1.711216427996132e-05, "loss": 1.0038, "step": 3197 }, { "epoch": 0.27, "learning_rate": 1.7110235735847726e-05, "loss": 1.1026, "step": 3198 }, { "epoch": 0.27, "learning_rate": 1.7108306656744685e-05, "loss": 1.1004, "step": 3199 }, { "epoch": 0.27, "learning_rate": 1.7106377042797336e-05, "loss": 1.1196, "step": 3200 }, { "epoch": 0.27, "learning_rate": 1.7104446894150867e-05, "loss": 1.1291, "step": 3201 }, { "epoch": 0.27, "learning_rate": 1.7102516210950514e-05, "loss": 1.1167, "step": 3202 }, { "epoch": 0.27, "learning_rate": 1.7100584993341538e-05, "loss": 1.0961, "step": 3203 }, { "epoch": 0.27, "learning_rate": 1.709865324146925e-05, "loss": 1.0882, "step": 3204 }, { "epoch": 0.27, "learning_rate": 1.7096720955479e-05, "loss": 1.1056, "step": 3205 }, { "epoch": 0.27, "learning_rate": 1.7094788135516176e-05, "loss": 1.0306, "step": 3206 }, { "epoch": 0.27, "learning_rate": 1.709285478172621e-05, "loss": 1.1204, "step": 3207 }, { "epoch": 0.27, "learning_rate": 1.7090920894254574e-05, "loss": 1.0851, "step": 3208 }, { "epoch": 0.27, "learning_rate": 1.708898647324677e-05, "loss": 1.1363, "step": 3209 }, { "epoch": 0.27, "learning_rate": 1.7087051518848355e-05, "loss": 1.0565, "step": 3210 }, { "epoch": 0.27, "learning_rate": 1.7085116031204915e-05, "loss": 1.1143, "step": 3211 }, { "epoch": 0.27, "learning_rate": 1.7083180010462085e-05, "loss": 1.0696, "step": 3212 }, { "epoch": 0.27, "learning_rate": 1.7081243456765525e-05, "loss": 1.1013, "step": 3213 }, { "epoch": 0.27, "learning_rate": 1.7079306370260958e-05, "loss": 1.0898, "step": 3214 }, { "epoch": 0.27, "learning_rate": 1.707736875109413e-05, "loss": 1.1367, "step": 3215 }, { "epoch": 0.27, "learning_rate": 1.707543059941083e-05, "loss": 1.1087, "step": 3216 }, { "epoch": 0.27, "learning_rate": 1.707349191535689e-05, "loss": 1.1075, "step": 3217 }, { "epoch": 0.27, "learning_rate": 1.707155269907818e-05, "loss": 1.16, "step": 3218 }, { "epoch": 0.27, "learning_rate": 1.7069612950720608e-05, "loss": 1.0882, "step": 3219 }, { "epoch": 0.27, "learning_rate": 1.7067672670430128e-05, "loss": 1.1094, "step": 3220 }, { "epoch": 0.27, "learning_rate": 1.7065731858352736e-05, "loss": 1.1311, "step": 3221 }, { "epoch": 0.27, "learning_rate": 1.7063790514634454e-05, "loss": 0.8969, "step": 3222 }, { "epoch": 0.27, "learning_rate": 1.7061848639421358e-05, "loss": 1.1068, "step": 3223 }, { "epoch": 0.27, "learning_rate": 1.7059906232859554e-05, "loss": 1.0877, "step": 3224 }, { "epoch": 0.27, "learning_rate": 1.70579632950952e-05, "loss": 1.1163, "step": 3225 }, { "epoch": 0.27, "learning_rate": 1.705601982627448e-05, "loss": 1.1335, "step": 3226 }, { "epoch": 0.27, "learning_rate": 1.7054075826543632e-05, "loss": 1.0908, "step": 3227 }, { "epoch": 0.27, "learning_rate": 1.705213129604892e-05, "loss": 1.0964, "step": 3228 }, { "epoch": 0.27, "learning_rate": 1.7050186234936656e-05, "loss": 0.9411, "step": 3229 }, { "epoch": 0.27, "learning_rate": 1.7048240643353196e-05, "loss": 1.1659, "step": 3230 }, { "epoch": 0.27, "learning_rate": 1.7046294521444924e-05, "loss": 1.0916, "step": 3231 }, { "epoch": 0.27, "learning_rate": 1.7044347869358273e-05, "loss": 1.0578, "step": 3232 }, { "epoch": 0.27, "learning_rate": 1.7042400687239712e-05, "loss": 1.1236, "step": 3233 }, { "epoch": 0.27, "learning_rate": 1.7040452975235754e-05, "loss": 1.1489, "step": 3234 }, { "epoch": 0.27, "learning_rate": 1.7038504733492945e-05, "loss": 1.1321, "step": 3235 }, { "epoch": 0.27, "learning_rate": 1.7036555962157883e-05, "loss": 1.0841, "step": 3236 }, { "epoch": 0.27, "learning_rate": 1.7034606661377187e-05, "loss": 1.1004, "step": 3237 }, { "epoch": 0.27, "learning_rate": 1.7032656831297535e-05, "loss": 1.1332, "step": 3238 }, { "epoch": 0.27, "learning_rate": 1.7030706472065626e-05, "loss": 1.0811, "step": 3239 }, { "epoch": 0.27, "learning_rate": 1.7028755583828222e-05, "loss": 1.1474, "step": 3240 }, { "epoch": 0.27, "learning_rate": 1.702680416673211e-05, "loss": 1.1515, "step": 3241 }, { "epoch": 0.27, "learning_rate": 1.7024852220924105e-05, "loss": 1.0732, "step": 3242 }, { "epoch": 0.27, "learning_rate": 1.7022899746551092e-05, "loss": 1.1065, "step": 3243 }, { "epoch": 0.27, "learning_rate": 1.702094674375997e-05, "loss": 1.0969, "step": 3244 }, { "epoch": 0.27, "learning_rate": 1.7018993212697693e-05, "loss": 1.1352, "step": 3245 }, { "epoch": 0.27, "learning_rate": 1.7017039153511247e-05, "loss": 1.1187, "step": 3246 }, { "epoch": 0.27, "learning_rate": 1.7015084566347657e-05, "loss": 1.1007, "step": 3247 }, { "epoch": 0.28, "learning_rate": 1.7013129451353992e-05, "loss": 1.089, "step": 3248 }, { "epoch": 0.28, "learning_rate": 1.7011173808677358e-05, "loss": 1.0902, "step": 3249 }, { "epoch": 0.28, "learning_rate": 1.7009217638464907e-05, "loss": 1.1228, "step": 3250 }, { "epoch": 0.28, "learning_rate": 1.7007260940863814e-05, "loss": 1.139, "step": 3251 }, { "epoch": 0.28, "learning_rate": 1.700530371602132e-05, "loss": 1.1164, "step": 3252 }, { "epoch": 0.28, "learning_rate": 1.7003345964084678e-05, "loss": 1.0687, "step": 3253 }, { "epoch": 0.28, "learning_rate": 1.70013876852012e-05, "loss": 1.091, "step": 3254 }, { "epoch": 0.28, "learning_rate": 1.6999428879518235e-05, "loss": 1.0885, "step": 3255 }, { "epoch": 0.28, "learning_rate": 1.6997469547183155e-05, "loss": 1.139, "step": 3256 }, { "epoch": 0.28, "learning_rate": 1.6995509688343396e-05, "loss": 1.0821, "step": 3257 }, { "epoch": 0.28, "learning_rate": 1.699354930314642e-05, "loss": 1.153, "step": 3258 }, { "epoch": 0.28, "learning_rate": 1.6991588391739723e-05, "loss": 1.0272, "step": 3259 }, { "epoch": 0.28, "learning_rate": 1.6989626954270858e-05, "loss": 1.1063, "step": 3260 }, { "epoch": 0.28, "learning_rate": 1.6987664990887404e-05, "loss": 1.0969, "step": 3261 }, { "epoch": 0.28, "learning_rate": 1.6985702501736983e-05, "loss": 1.1361, "step": 3262 }, { "epoch": 0.28, "learning_rate": 1.6983739486967256e-05, "loss": 1.0576, "step": 3263 }, { "epoch": 0.28, "learning_rate": 1.6981775946725925e-05, "loss": 1.0686, "step": 3264 }, { "epoch": 0.28, "learning_rate": 1.697981188116073e-05, "loss": 1.1207, "step": 3265 }, { "epoch": 0.28, "learning_rate": 1.6977847290419456e-05, "loss": 1.0674, "step": 3266 }, { "epoch": 0.28, "learning_rate": 1.6975882174649915e-05, "loss": 1.0634, "step": 3267 }, { "epoch": 0.28, "learning_rate": 1.6973916533999976e-05, "loss": 1.0838, "step": 3268 }, { "epoch": 0.28, "learning_rate": 1.6971950368617533e-05, "loss": 1.1477, "step": 3269 }, { "epoch": 0.28, "learning_rate": 1.6969983678650524e-05, "loss": 1.0848, "step": 3270 }, { "epoch": 0.28, "learning_rate": 1.696801646424693e-05, "loss": 1.1269, "step": 3271 }, { "epoch": 0.28, "learning_rate": 1.6966048725554763e-05, "loss": 1.1169, "step": 3272 }, { "epoch": 0.28, "learning_rate": 1.6964080462722086e-05, "loss": 1.1673, "step": 3273 }, { "epoch": 0.28, "learning_rate": 1.696211167589699e-05, "loss": 1.1514, "step": 3274 }, { "epoch": 0.28, "learning_rate": 1.6960142365227615e-05, "loss": 1.0999, "step": 3275 }, { "epoch": 0.28, "learning_rate": 1.6958172530862138e-05, "loss": 1.1108, "step": 3276 }, { "epoch": 0.28, "learning_rate": 1.6956202172948764e-05, "loss": 1.1059, "step": 3277 }, { "epoch": 0.28, "learning_rate": 1.6954231291635757e-05, "loss": 1.0571, "step": 3278 }, { "epoch": 0.28, "learning_rate": 1.6952259887071403e-05, "loss": 1.1354, "step": 3279 }, { "epoch": 0.28, "learning_rate": 1.6950287959404045e-05, "loss": 1.0828, "step": 3280 }, { "epoch": 0.28, "learning_rate": 1.6948315508782042e-05, "loss": 1.1065, "step": 3281 }, { "epoch": 0.28, "learning_rate": 1.6946342535353814e-05, "loss": 0.9321, "step": 3282 }, { "epoch": 0.28, "learning_rate": 1.6944369039267812e-05, "loss": 1.1768, "step": 3283 }, { "epoch": 0.28, "learning_rate": 1.6942395020672522e-05, "loss": 1.1654, "step": 3284 }, { "epoch": 0.28, "learning_rate": 1.6940420479716472e-05, "loss": 1.0862, "step": 3285 }, { "epoch": 0.28, "learning_rate": 1.693844541654824e-05, "loss": 1.0966, "step": 3286 }, { "epoch": 0.28, "learning_rate": 1.6936469831316422e-05, "loss": 1.1328, "step": 3287 }, { "epoch": 0.28, "learning_rate": 1.6934493724169672e-05, "loss": 1.0813, "step": 3288 }, { "epoch": 0.28, "learning_rate": 1.693251709525668e-05, "loss": 1.1147, "step": 3289 }, { "epoch": 0.28, "learning_rate": 1.6930539944726163e-05, "loss": 1.1432, "step": 3290 }, { "epoch": 0.28, "learning_rate": 1.692856227272689e-05, "loss": 1.1143, "step": 3291 }, { "epoch": 0.28, "learning_rate": 1.692658407940767e-05, "loss": 1.0207, "step": 3292 }, { "epoch": 0.28, "learning_rate": 1.692460536491734e-05, "loss": 1.1398, "step": 3293 }, { "epoch": 0.28, "learning_rate": 1.6922626129404785e-05, "loss": 1.0921, "step": 3294 }, { "epoch": 0.28, "learning_rate": 1.6920646373018925e-05, "loss": 1.0633, "step": 3295 }, { "epoch": 0.28, "learning_rate": 1.6918666095908726e-05, "loss": 1.058, "step": 3296 }, { "epoch": 0.28, "learning_rate": 1.6916685298223186e-05, "loss": 1.0036, "step": 3297 }, { "epoch": 0.28, "learning_rate": 1.691470398011134e-05, "loss": 0.89, "step": 3298 }, { "epoch": 0.28, "learning_rate": 1.6912722141722278e-05, "loss": 1.0566, "step": 3299 }, { "epoch": 0.28, "learning_rate": 1.6910739783205102e-05, "loss": 1.1325, "step": 3300 }, { "epoch": 0.28, "learning_rate": 1.6908756904708982e-05, "loss": 1.0764, "step": 3301 }, { "epoch": 0.28, "learning_rate": 1.690677350638311e-05, "loss": 1.1295, "step": 3302 }, { "epoch": 0.28, "learning_rate": 1.6904789588376716e-05, "loss": 1.0871, "step": 3303 }, { "epoch": 0.28, "learning_rate": 1.6902805150839082e-05, "loss": 1.062, "step": 3304 }, { "epoch": 0.28, "learning_rate": 1.690082019391952e-05, "loss": 1.0858, "step": 3305 }, { "epoch": 0.28, "learning_rate": 1.689883471776738e-05, "loss": 1.0889, "step": 3306 }, { "epoch": 0.28, "learning_rate": 1.6896848722532052e-05, "loss": 1.1351, "step": 3307 }, { "epoch": 0.28, "learning_rate": 1.689486220836297e-05, "loss": 1.0991, "step": 3308 }, { "epoch": 0.28, "learning_rate": 1.68928751754096e-05, "loss": 1.1622, "step": 3309 }, { "epoch": 0.28, "learning_rate": 1.6890887623821455e-05, "loss": 1.0891, "step": 3310 }, { "epoch": 0.28, "learning_rate": 1.6888899553748078e-05, "loss": 1.116, "step": 3311 }, { "epoch": 0.28, "learning_rate": 1.6886910965339062e-05, "loss": 1.0632, "step": 3312 }, { "epoch": 0.28, "learning_rate": 1.6884921858744027e-05, "loss": 1.1255, "step": 3313 }, { "epoch": 0.28, "learning_rate": 1.688293223411264e-05, "loss": 1.1412, "step": 3314 }, { "epoch": 0.28, "learning_rate": 1.6880942091594607e-05, "loss": 1.068, "step": 3315 }, { "epoch": 0.28, "learning_rate": 1.6878951431339665e-05, "loss": 1.143, "step": 3316 }, { "epoch": 0.28, "learning_rate": 1.6876960253497596e-05, "loss": 1.0523, "step": 3317 }, { "epoch": 0.28, "learning_rate": 1.6874968558218225e-05, "loss": 1.0982, "step": 3318 }, { "epoch": 0.28, "learning_rate": 1.687297634565141e-05, "loss": 1.1276, "step": 3319 }, { "epoch": 0.28, "learning_rate": 1.6870983615947047e-05, "loss": 1.1153, "step": 3320 }, { "epoch": 0.28, "learning_rate": 1.686899036925508e-05, "loss": 1.1058, "step": 3321 }, { "epoch": 0.28, "learning_rate": 1.6866996605725473e-05, "loss": 1.0877, "step": 3322 }, { "epoch": 0.28, "learning_rate": 1.6865002325508254e-05, "loss": 1.0762, "step": 3323 }, { "epoch": 0.28, "learning_rate": 1.6863007528753467e-05, "loss": 1.0799, "step": 3324 }, { "epoch": 0.28, "learning_rate": 1.6861012215611213e-05, "loss": 1.1019, "step": 3325 }, { "epoch": 0.28, "learning_rate": 1.6859016386231618e-05, "loss": 1.0715, "step": 3326 }, { "epoch": 0.28, "learning_rate": 1.6857020040764853e-05, "loss": 1.1393, "step": 3327 }, { "epoch": 0.28, "learning_rate": 1.6855023179361127e-05, "loss": 1.0592, "step": 3328 }, { "epoch": 0.28, "learning_rate": 1.6853025802170692e-05, "loss": 1.1119, "step": 3329 }, { "epoch": 0.28, "learning_rate": 1.6851027909343834e-05, "loss": 1.091, "step": 3330 }, { "epoch": 0.28, "learning_rate": 1.6849029501030874e-05, "loss": 1.1079, "step": 3331 }, { "epoch": 0.28, "learning_rate": 1.6847030577382182e-05, "loss": 1.0843, "step": 3332 }, { "epoch": 0.28, "learning_rate": 1.684503113854816e-05, "loss": 1.1118, "step": 3333 }, { "epoch": 0.28, "learning_rate": 1.6843031184679247e-05, "loss": 1.1046, "step": 3334 }, { "epoch": 0.28, "learning_rate": 1.6841030715925927e-05, "loss": 1.0967, "step": 3335 }, { "epoch": 0.28, "learning_rate": 1.6839029732438722e-05, "loss": 1.1058, "step": 3336 }, { "epoch": 0.28, "learning_rate": 1.6837028234368184e-05, "loss": 1.0913, "step": 3337 }, { "epoch": 0.28, "learning_rate": 1.683502622186491e-05, "loss": 1.1118, "step": 3338 }, { "epoch": 0.28, "learning_rate": 1.6833023695079545e-05, "loss": 1.0581, "step": 3339 }, { "epoch": 0.28, "learning_rate": 1.6831020654162752e-05, "loss": 1.1275, "step": 3340 }, { "epoch": 0.28, "learning_rate": 1.6829017099265252e-05, "loss": 1.1161, "step": 3341 }, { "epoch": 0.28, "learning_rate": 1.6827013030537793e-05, "loss": 1.1148, "step": 3342 }, { "epoch": 0.28, "learning_rate": 1.6825008448131167e-05, "loss": 1.1225, "step": 3343 }, { "epoch": 0.28, "learning_rate": 1.6823003352196204e-05, "loss": 1.1342, "step": 3344 }, { "epoch": 0.28, "learning_rate": 1.682099774288377e-05, "loss": 1.0747, "step": 3345 }, { "epoch": 0.28, "learning_rate": 1.681899162034477e-05, "loss": 1.0918, "step": 3346 }, { "epoch": 0.28, "learning_rate": 1.6816984984730155e-05, "loss": 1.1332, "step": 3347 }, { "epoch": 0.28, "learning_rate": 1.68149778361909e-05, "loss": 1.1523, "step": 3348 }, { "epoch": 0.28, "learning_rate": 1.681297017487803e-05, "loss": 1.0513, "step": 3349 }, { "epoch": 0.28, "learning_rate": 1.681096200094261e-05, "loss": 1.0973, "step": 3350 }, { "epoch": 0.28, "learning_rate": 1.6808953314535737e-05, "loss": 1.047, "step": 3351 }, { "epoch": 0.28, "learning_rate": 1.6806944115808547e-05, "loss": 1.1022, "step": 3352 }, { "epoch": 0.28, "learning_rate": 1.680493440491222e-05, "loss": 1.1397, "step": 3353 }, { "epoch": 0.28, "learning_rate": 1.680292418199797e-05, "loss": 1.0674, "step": 3354 }, { "epoch": 0.28, "learning_rate": 1.6800913447217048e-05, "loss": 1.1064, "step": 3355 }, { "epoch": 0.28, "learning_rate": 1.6798902200720745e-05, "loss": 1.0396, "step": 3356 }, { "epoch": 0.28, "learning_rate": 1.6796890442660395e-05, "loss": 1.1146, "step": 3357 }, { "epoch": 0.28, "learning_rate": 1.679487817318737e-05, "loss": 1.0955, "step": 3358 }, { "epoch": 0.28, "learning_rate": 1.6792865392453066e-05, "loss": 1.1358, "step": 3359 }, { "epoch": 0.28, "learning_rate": 1.679085210060894e-05, "loss": 1.1154, "step": 3360 }, { "epoch": 0.28, "learning_rate": 1.6788838297806472e-05, "loss": 1.1249, "step": 3361 }, { "epoch": 0.28, "learning_rate": 1.678682398419719e-05, "loss": 1.0479, "step": 3362 }, { "epoch": 0.28, "learning_rate": 1.6784809159932648e-05, "loss": 1.103, "step": 3363 }, { "epoch": 0.28, "learning_rate": 1.678279382516445e-05, "loss": 1.1161, "step": 3364 }, { "epoch": 0.28, "learning_rate": 1.6780777980044233e-05, "loss": 1.0939, "step": 3365 }, { "epoch": 0.29, "learning_rate": 1.677876162472367e-05, "loss": 1.1233, "step": 3366 }, { "epoch": 0.29, "learning_rate": 1.677674475935448e-05, "loss": 1.1047, "step": 3367 }, { "epoch": 0.29, "learning_rate": 1.6774727384088416e-05, "loss": 1.1239, "step": 3368 }, { "epoch": 0.29, "learning_rate": 1.677270949907727e-05, "loss": 1.1443, "step": 3369 }, { "epoch": 0.29, "learning_rate": 1.6770691104472875e-05, "loss": 1.1029, "step": 3370 }, { "epoch": 0.29, "learning_rate": 1.6768672200427092e-05, "loss": 1.0826, "step": 3371 }, { "epoch": 0.29, "learning_rate": 1.676665278709183e-05, "loss": 1.126, "step": 3372 }, { "epoch": 0.29, "learning_rate": 1.6764632864619038e-05, "loss": 1.0599, "step": 3373 }, { "epoch": 0.29, "learning_rate": 1.6762612433160693e-05, "loss": 1.135, "step": 3374 }, { "epoch": 0.29, "learning_rate": 1.6760591492868826e-05, "loss": 1.0411, "step": 3375 }, { "epoch": 0.29, "learning_rate": 1.6758570043895487e-05, "loss": 1.1335, "step": 3376 }, { "epoch": 0.29, "learning_rate": 1.6756548086392776e-05, "loss": 1.058, "step": 3377 }, { "epoch": 0.29, "learning_rate": 1.675452562051284e-05, "loss": 1.0767, "step": 3378 }, { "epoch": 0.29, "learning_rate": 1.675250264640784e-05, "loss": 0.9944, "step": 3379 }, { "epoch": 0.29, "learning_rate": 1.6750479164229993e-05, "loss": 1.1088, "step": 3380 }, { "epoch": 0.29, "learning_rate": 1.674845517413155e-05, "loss": 1.0648, "step": 3381 }, { "epoch": 0.29, "learning_rate": 1.6746430676264808e-05, "loss": 1.1069, "step": 3382 }, { "epoch": 0.29, "learning_rate": 1.6744405670782083e-05, "loss": 1.1222, "step": 3383 }, { "epoch": 0.29, "learning_rate": 1.674238015783575e-05, "loss": 1.0897, "step": 3384 }, { "epoch": 0.29, "learning_rate": 1.674035413757821e-05, "loss": 1.1118, "step": 3385 }, { "epoch": 0.29, "learning_rate": 1.67383276101619e-05, "loss": 1.0799, "step": 3386 }, { "epoch": 0.29, "learning_rate": 1.6736300575739307e-05, "loss": 1.0881, "step": 3387 }, { "epoch": 0.29, "learning_rate": 1.6734273034462948e-05, "loss": 1.0667, "step": 3388 }, { "epoch": 0.29, "learning_rate": 1.6732244986485378e-05, "loss": 1.0518, "step": 3389 }, { "epoch": 0.29, "learning_rate": 1.673021643195919e-05, "loss": 1.035, "step": 3390 }, { "epoch": 0.29, "learning_rate": 1.672818737103702e-05, "loss": 0.9454, "step": 3391 }, { "epoch": 0.29, "learning_rate": 1.672615780387154e-05, "loss": 1.1048, "step": 3392 }, { "epoch": 0.29, "learning_rate": 1.6724127730615457e-05, "loss": 1.0637, "step": 3393 }, { "epoch": 0.29, "learning_rate": 1.672209715142152e-05, "loss": 1.0989, "step": 3394 }, { "epoch": 0.29, "learning_rate": 1.672006606644251e-05, "loss": 1.0676, "step": 3395 }, { "epoch": 0.29, "learning_rate": 1.6718034475831252e-05, "loss": 1.0878, "step": 3396 }, { "epoch": 0.29, "learning_rate": 1.6716002379740607e-05, "loss": 1.1613, "step": 3397 }, { "epoch": 0.29, "learning_rate": 1.671396977832348e-05, "loss": 1.0874, "step": 3398 }, { "epoch": 0.29, "learning_rate": 1.6711936671732798e-05, "loss": 1.1024, "step": 3399 }, { "epoch": 0.29, "learning_rate": 1.6709903060121544e-05, "loss": 1.0558, "step": 3400 }, { "epoch": 0.29, "learning_rate": 1.6707868943642725e-05, "loss": 1.1182, "step": 3401 }, { "epoch": 0.29, "learning_rate": 1.6705834322449402e-05, "loss": 0.9391, "step": 3402 }, { "epoch": 0.29, "learning_rate": 1.6703799196694655e-05, "loss": 1.0974, "step": 3403 }, { "epoch": 0.29, "learning_rate": 1.670176356653162e-05, "loss": 1.1288, "step": 3404 }, { "epoch": 0.29, "learning_rate": 1.6699727432113454e-05, "loss": 1.0954, "step": 3405 }, { "epoch": 0.29, "learning_rate": 1.669769079359336e-05, "loss": 1.0838, "step": 3406 }, { "epoch": 0.29, "learning_rate": 1.6695653651124586e-05, "loss": 1.0826, "step": 3407 }, { "epoch": 0.29, "learning_rate": 1.6693616004860403e-05, "loss": 1.1131, "step": 3408 }, { "epoch": 0.29, "learning_rate": 1.6691577854954137e-05, "loss": 1.1133, "step": 3409 }, { "epoch": 0.29, "learning_rate": 1.6689539201559137e-05, "loss": 1.0789, "step": 3410 }, { "epoch": 0.29, "learning_rate": 1.6687500044828798e-05, "loss": 1.0978, "step": 3411 }, { "epoch": 0.29, "learning_rate": 1.6685460384916544e-05, "loss": 1.0729, "step": 3412 }, { "epoch": 0.29, "learning_rate": 1.668342022197585e-05, "loss": 1.0676, "step": 3413 }, { "epoch": 0.29, "learning_rate": 1.6681379556160226e-05, "loss": 1.1163, "step": 3414 }, { "epoch": 0.29, "learning_rate": 1.6679338387623207e-05, "loss": 1.0848, "step": 3415 }, { "epoch": 0.29, "learning_rate": 1.667729671651838e-05, "loss": 1.0676, "step": 3416 }, { "epoch": 0.29, "learning_rate": 1.667525454299936e-05, "loss": 1.0631, "step": 3417 }, { "epoch": 0.29, "learning_rate": 1.667321186721981e-05, "loss": 1.1112, "step": 3418 }, { "epoch": 0.29, "learning_rate": 1.6671168689333427e-05, "loss": 1.0781, "step": 3419 }, { "epoch": 0.29, "learning_rate": 1.6669125009493936e-05, "loss": 1.0689, "step": 3420 }, { "epoch": 0.29, "learning_rate": 1.6667080827855115e-05, "loss": 1.0921, "step": 3421 }, { "epoch": 0.29, "learning_rate": 1.666503614457077e-05, "loss": 1.1434, "step": 3422 }, { "epoch": 0.29, "learning_rate": 1.6662990959794746e-05, "loss": 1.0917, "step": 3423 }, { "epoch": 0.29, "learning_rate": 1.666094527368093e-05, "loss": 1.1178, "step": 3424 }, { "epoch": 0.29, "learning_rate": 1.6658899086383244e-05, "loss": 1.0732, "step": 3425 }, { "epoch": 0.29, "learning_rate": 1.665685239805564e-05, "loss": 1.067, "step": 3426 }, { "epoch": 0.29, "learning_rate": 1.6654805208852127e-05, "loss": 1.0658, "step": 3427 }, { "epoch": 0.29, "learning_rate": 1.665275751892673e-05, "loss": 1.0895, "step": 3428 }, { "epoch": 0.29, "learning_rate": 1.6650709328433526e-05, "loss": 1.1191, "step": 3429 }, { "epoch": 0.29, "learning_rate": 1.6648660637526628e-05, "loss": 1.1488, "step": 3430 }, { "epoch": 0.29, "learning_rate": 1.6646611446360177e-05, "loss": 1.1139, "step": 3431 }, { "epoch": 0.29, "learning_rate": 1.6644561755088364e-05, "loss": 1.0628, "step": 3432 }, { "epoch": 0.29, "learning_rate": 1.664251156386541e-05, "loss": 1.0993, "step": 3433 }, { "epoch": 0.29, "learning_rate": 1.664046087284558e-05, "loss": 1.1031, "step": 3434 }, { "epoch": 0.29, "learning_rate": 1.6638409682183162e-05, "loss": 1.0692, "step": 3435 }, { "epoch": 0.29, "learning_rate": 1.6636357992032507e-05, "loss": 0.9484, "step": 3436 }, { "epoch": 0.29, "learning_rate": 1.6634305802547974e-05, "loss": 1.0899, "step": 3437 }, { "epoch": 0.29, "learning_rate": 1.663225311388398e-05, "loss": 1.0974, "step": 3438 }, { "epoch": 0.29, "learning_rate": 1.6630199926194982e-05, "loss": 1.1477, "step": 3439 }, { "epoch": 0.29, "learning_rate": 1.662814623963545e-05, "loss": 1.1251, "step": 3440 }, { "epoch": 0.29, "learning_rate": 1.662609205435992e-05, "loss": 1.0628, "step": 3441 }, { "epoch": 0.29, "learning_rate": 1.662403737052295e-05, "loss": 1.1104, "step": 3442 }, { "epoch": 0.29, "learning_rate": 1.662198218827914e-05, "loss": 1.1641, "step": 3443 }, { "epoch": 0.29, "learning_rate": 1.661992650778312e-05, "loss": 1.1042, "step": 3444 }, { "epoch": 0.29, "learning_rate": 1.6617870329189574e-05, "loss": 1.1125, "step": 3445 }, { "epoch": 0.29, "learning_rate": 1.6615813652653204e-05, "loss": 1.1819, "step": 3446 }, { "epoch": 0.29, "learning_rate": 1.6613756478328765e-05, "loss": 1.0923, "step": 3447 }, { "epoch": 0.29, "learning_rate": 1.661169880637104e-05, "loss": 1.1264, "step": 3448 }, { "epoch": 0.29, "learning_rate": 1.6609640636934855e-05, "loss": 1.0399, "step": 3449 }, { "epoch": 0.29, "learning_rate": 1.660758197017507e-05, "loss": 1.0414, "step": 3450 }, { "epoch": 0.29, "learning_rate": 1.660552280624658e-05, "loss": 1.0813, "step": 3451 }, { "epoch": 0.29, "learning_rate": 1.6603463145304325e-05, "loss": 1.0933, "step": 3452 }, { "epoch": 0.29, "learning_rate": 1.660140298750328e-05, "loss": 1.0962, "step": 3453 }, { "epoch": 0.29, "learning_rate": 1.659934233299845e-05, "loss": 1.0884, "step": 3454 }, { "epoch": 0.29, "learning_rate": 1.659728118194489e-05, "loss": 1.0703, "step": 3455 }, { "epoch": 0.29, "learning_rate": 1.659521953449768e-05, "loss": 1.0257, "step": 3456 }, { "epoch": 0.29, "learning_rate": 1.6593157390811943e-05, "loss": 1.1171, "step": 3457 }, { "epoch": 0.29, "learning_rate": 1.659109475104284e-05, "loss": 1.1289, "step": 3458 }, { "epoch": 0.29, "learning_rate": 1.658903161534557e-05, "loss": 1.1248, "step": 3459 }, { "epoch": 0.29, "learning_rate": 1.658696798387537e-05, "loss": 1.1642, "step": 3460 }, { "epoch": 0.29, "learning_rate": 1.658490385678751e-05, "loss": 1.1061, "step": 3461 }, { "epoch": 0.29, "learning_rate": 1.6582839234237292e-05, "loss": 1.1493, "step": 3462 }, { "epoch": 0.29, "learning_rate": 1.6580774116380075e-05, "loss": 1.0662, "step": 3463 }, { "epoch": 0.29, "learning_rate": 1.6578708503371233e-05, "loss": 1.1557, "step": 3464 }, { "epoch": 0.29, "learning_rate": 1.6576642395366192e-05, "loss": 1.0358, "step": 3465 }, { "epoch": 0.29, "learning_rate": 1.657457579252041e-05, "loss": 1.146, "step": 3466 }, { "epoch": 0.29, "learning_rate": 1.657250869498938e-05, "loss": 1.1348, "step": 3467 }, { "epoch": 0.29, "learning_rate": 1.657044110292864e-05, "loss": 0.9108, "step": 3468 }, { "epoch": 0.29, "learning_rate": 1.6568373016493756e-05, "loss": 1.0985, "step": 3469 }, { "epoch": 0.29, "learning_rate": 1.656630443584034e-05, "loss": 1.1262, "step": 3470 }, { "epoch": 0.29, "learning_rate": 1.656423536112403e-05, "loss": 1.1573, "step": 3471 }, { "epoch": 0.29, "learning_rate": 1.656216579250051e-05, "loss": 1.1099, "step": 3472 }, { "epoch": 0.29, "learning_rate": 1.6560095730125498e-05, "loss": 1.1242, "step": 3473 }, { "epoch": 0.29, "learning_rate": 1.655802517415475e-05, "loss": 1.069, "step": 3474 }, { "epoch": 0.29, "learning_rate": 1.6555954124744066e-05, "loss": 1.0897, "step": 3475 }, { "epoch": 0.29, "learning_rate": 1.655388258204927e-05, "loss": 1.1537, "step": 3476 }, { "epoch": 0.29, "learning_rate": 1.6551810546226225e-05, "loss": 1.0685, "step": 3477 }, { "epoch": 0.29, "learning_rate": 1.654973801743084e-05, "loss": 1.0811, "step": 3478 }, { "epoch": 0.29, "learning_rate": 1.654766499581906e-05, "loss": 1.1086, "step": 3479 }, { "epoch": 0.29, "learning_rate": 1.6545591481546862e-05, "loss": 1.1005, "step": 3480 }, { "epoch": 0.29, "learning_rate": 1.654351747477026e-05, "loss": 1.0788, "step": 3481 }, { "epoch": 0.29, "learning_rate": 1.6541442975645304e-05, "loss": 1.1093, "step": 3482 }, { "epoch": 0.29, "learning_rate": 1.6539367984328087e-05, "loss": 1.0679, "step": 3483 }, { "epoch": 0.3, "learning_rate": 1.6537292500974736e-05, "loss": 0.9132, "step": 3484 }, { "epoch": 0.3, "learning_rate": 1.6535216525741415e-05, "loss": 1.0957, "step": 3485 }, { "epoch": 0.3, "learning_rate": 1.6533140058784322e-05, "loss": 1.0746, "step": 3486 }, { "epoch": 0.3, "learning_rate": 1.6531063100259702e-05, "loss": 1.0938, "step": 3487 }, { "epoch": 0.3, "learning_rate": 1.652898565032382e-05, "loss": 1.0648, "step": 3488 }, { "epoch": 0.3, "learning_rate": 1.6526907709132994e-05, "loss": 0.8674, "step": 3489 }, { "epoch": 0.3, "learning_rate": 1.652482927684357e-05, "loss": 1.0439, "step": 3490 }, { "epoch": 0.3, "learning_rate": 1.652275035361194e-05, "loss": 1.0798, "step": 3491 }, { "epoch": 0.3, "learning_rate": 1.652067093959452e-05, "loss": 1.084, "step": 3492 }, { "epoch": 0.3, "learning_rate": 1.651859103494777e-05, "loss": 1.1193, "step": 3493 }, { "epoch": 0.3, "learning_rate": 1.6516510639828188e-05, "loss": 1.0809, "step": 3494 }, { "epoch": 0.3, "learning_rate": 1.651442975439231e-05, "loss": 1.0985, "step": 3495 }, { "epoch": 0.3, "learning_rate": 1.6512348378796697e-05, "loss": 1.1234, "step": 3496 }, { "epoch": 0.3, "learning_rate": 1.6510266513197968e-05, "loss": 1.1041, "step": 3497 }, { "epoch": 0.3, "learning_rate": 1.6508184157752763e-05, "loss": 1.0245, "step": 3498 }, { "epoch": 0.3, "learning_rate": 1.6506101312617757e-05, "loss": 1.1449, "step": 3499 }, { "epoch": 0.3, "learning_rate": 1.6504017977949675e-05, "loss": 1.1028, "step": 3500 }, { "epoch": 0.3, "learning_rate": 1.650193415390527e-05, "loss": 1.0974, "step": 3501 }, { "epoch": 0.3, "learning_rate": 1.6499849840641333e-05, "loss": 1.0976, "step": 3502 }, { "epoch": 0.3, "learning_rate": 1.6497765038314692e-05, "loss": 1.0837, "step": 3503 }, { "epoch": 0.3, "learning_rate": 1.649567974708221e-05, "loss": 1.17, "step": 3504 }, { "epoch": 0.3, "learning_rate": 1.649359396710079e-05, "loss": 1.0974, "step": 3505 }, { "epoch": 0.3, "learning_rate": 1.6491507698527372e-05, "loss": 1.1106, "step": 3506 }, { "epoch": 0.3, "learning_rate": 1.6489420941518934e-05, "loss": 1.0484, "step": 3507 }, { "epoch": 0.3, "learning_rate": 1.648733369623248e-05, "loss": 1.0683, "step": 3508 }, { "epoch": 0.3, "learning_rate": 1.6485245962825064e-05, "loss": 1.1119, "step": 3509 }, { "epoch": 0.3, "learning_rate": 1.6483157741453774e-05, "loss": 1.0768, "step": 3510 }, { "epoch": 0.3, "learning_rate": 1.6481069032275725e-05, "loss": 1.1335, "step": 3511 }, { "epoch": 0.3, "learning_rate": 1.6478979835448083e-05, "loss": 1.1091, "step": 3512 }, { "epoch": 0.3, "learning_rate": 1.647689015112804e-05, "loss": 0.9855, "step": 3513 }, { "epoch": 0.3, "learning_rate": 1.6474799979472828e-05, "loss": 1.0931, "step": 3514 }, { "epoch": 0.3, "learning_rate": 1.647270932063972e-05, "loss": 0.9552, "step": 3515 }, { "epoch": 0.3, "learning_rate": 1.6470618174786018e-05, "loss": 1.0605, "step": 3516 }, { "epoch": 0.3, "learning_rate": 1.6468526542069066e-05, "loss": 1.1071, "step": 3517 }, { "epoch": 0.3, "learning_rate": 1.646643442264624e-05, "loss": 1.1101, "step": 3518 }, { "epoch": 0.3, "learning_rate": 1.646434181667496e-05, "loss": 1.0862, "step": 3519 }, { "epoch": 0.3, "learning_rate": 1.6462248724312674e-05, "loss": 1.0184, "step": 3520 }, { "epoch": 0.3, "learning_rate": 1.6460155145716873e-05, "loss": 1.1001, "step": 3521 }, { "epoch": 0.3, "learning_rate": 1.6458061081045086e-05, "loss": 1.1186, "step": 3522 }, { "epoch": 0.3, "learning_rate": 1.6455966530454864e-05, "loss": 1.0995, "step": 3523 }, { "epoch": 0.3, "learning_rate": 1.6453871494103818e-05, "loss": 1.0558, "step": 3524 }, { "epoch": 0.3, "learning_rate": 1.645177597214958e-05, "loss": 1.0916, "step": 3525 }, { "epoch": 0.3, "learning_rate": 1.644967996474981e-05, "loss": 1.0296, "step": 3526 }, { "epoch": 0.3, "learning_rate": 1.6447583472062234e-05, "loss": 1.0948, "step": 3527 }, { "epoch": 0.3, "learning_rate": 1.6445486494244584e-05, "loss": 1.1254, "step": 3528 }, { "epoch": 0.3, "learning_rate": 1.6443389031454645e-05, "loss": 1.1601, "step": 3529 }, { "epoch": 0.3, "learning_rate": 1.6441291083850238e-05, "loss": 1.1116, "step": 3530 }, { "epoch": 0.3, "learning_rate": 1.643919265158921e-05, "loss": 1.0985, "step": 3531 }, { "epoch": 0.3, "learning_rate": 1.6437093734829454e-05, "loss": 1.0874, "step": 3532 }, { "epoch": 0.3, "learning_rate": 1.6434994333728906e-05, "loss": 1.1062, "step": 3533 }, { "epoch": 0.3, "learning_rate": 1.6432894448445514e-05, "loss": 1.1012, "step": 3534 }, { "epoch": 0.3, "learning_rate": 1.6430794079137288e-05, "loss": 1.1044, "step": 3535 }, { "epoch": 0.3, "learning_rate": 1.6428693225962263e-05, "loss": 1.0004, "step": 3536 }, { "epoch": 0.3, "learning_rate": 1.642659188907851e-05, "loss": 1.1004, "step": 3537 }, { "epoch": 0.3, "learning_rate": 1.6424490068644137e-05, "loss": 1.1472, "step": 3538 }, { "epoch": 0.3, "learning_rate": 1.6422387764817295e-05, "loss": 1.1264, "step": 3539 }, { "epoch": 0.3, "learning_rate": 1.6420284977756158e-05, "loss": 1.1083, "step": 3540 }, { "epoch": 0.3, "learning_rate": 1.6418181707618954e-05, "loss": 1.1232, "step": 3541 }, { "epoch": 0.3, "learning_rate": 1.6416077954563926e-05, "loss": 1.1295, "step": 3542 }, { "epoch": 0.3, "learning_rate": 1.6413973718749376e-05, "loss": 1.1094, "step": 3543 }, { "epoch": 0.3, "learning_rate": 1.6411869000333626e-05, "loss": 1.0699, "step": 3544 }, { "epoch": 0.3, "learning_rate": 1.6409763799475034e-05, "loss": 1.1223, "step": 3545 }, { "epoch": 0.3, "learning_rate": 1.640765811633201e-05, "loss": 1.0842, "step": 3546 }, { "epoch": 0.3, "learning_rate": 1.6405551951062986e-05, "loss": 1.0842, "step": 3547 }, { "epoch": 0.3, "learning_rate": 1.640344530382643e-05, "loss": 1.1604, "step": 3548 }, { "epoch": 0.3, "learning_rate": 1.640133817478086e-05, "loss": 1.0922, "step": 3549 }, { "epoch": 0.3, "learning_rate": 1.6399230564084816e-05, "loss": 1.1271, "step": 3550 }, { "epoch": 0.3, "learning_rate": 1.6397122471896877e-05, "loss": 1.0827, "step": 3551 }, { "epoch": 0.3, "learning_rate": 1.639501389837566e-05, "loss": 1.0427, "step": 3552 }, { "epoch": 0.3, "learning_rate": 1.6392904843679823e-05, "loss": 1.0946, "step": 3553 }, { "epoch": 0.3, "learning_rate": 1.6390795307968056e-05, "loss": 1.1248, "step": 3554 }, { "epoch": 0.3, "learning_rate": 1.6388685291399077e-05, "loss": 1.0455, "step": 3555 }, { "epoch": 0.3, "learning_rate": 1.638657479413166e-05, "loss": 1.0866, "step": 3556 }, { "epoch": 0.3, "learning_rate": 1.6384463816324592e-05, "loss": 1.1059, "step": 3557 }, { "epoch": 0.3, "learning_rate": 1.6382352358136717e-05, "loss": 1.1251, "step": 3558 }, { "epoch": 0.3, "learning_rate": 1.63802404197269e-05, "loss": 1.0614, "step": 3559 }, { "epoch": 0.3, "learning_rate": 1.6378128001254047e-05, "loss": 1.1298, "step": 3560 }, { "epoch": 0.3, "learning_rate": 1.6376015102877106e-05, "loss": 1.0689, "step": 3561 }, { "epoch": 0.3, "learning_rate": 1.6373901724755053e-05, "loss": 1.058, "step": 3562 }, { "epoch": 0.3, "learning_rate": 1.63717878670469e-05, "loss": 1.0552, "step": 3563 }, { "epoch": 0.3, "learning_rate": 1.6369673529911706e-05, "loss": 1.0558, "step": 3564 }, { "epoch": 0.3, "learning_rate": 1.636755871350855e-05, "loss": 1.0931, "step": 3565 }, { "epoch": 0.3, "learning_rate": 1.636544341799656e-05, "loss": 1.0412, "step": 3566 }, { "epoch": 0.3, "learning_rate": 1.6363327643534898e-05, "loss": 1.1683, "step": 3567 }, { "epoch": 0.3, "learning_rate": 1.6361211390282755e-05, "loss": 1.0155, "step": 3568 }, { "epoch": 0.3, "learning_rate": 1.6359094658399365e-05, "loss": 1.0998, "step": 3569 }, { "epoch": 0.3, "learning_rate": 1.6356977448043986e-05, "loss": 1.0813, "step": 3570 }, { "epoch": 0.3, "learning_rate": 1.6354859759375936e-05, "loss": 0.946, "step": 3571 }, { "epoch": 0.3, "learning_rate": 1.6352741592554545e-05, "loss": 1.1116, "step": 3572 }, { "epoch": 0.3, "learning_rate": 1.6350622947739194e-05, "loss": 1.1184, "step": 3573 }, { "epoch": 0.3, "learning_rate": 1.634850382508929e-05, "loss": 1.0739, "step": 3574 }, { "epoch": 0.3, "learning_rate": 1.634638422476428e-05, "loss": 1.1318, "step": 3575 }, { "epoch": 0.3, "learning_rate": 1.6344264146923655e-05, "loss": 1.0818, "step": 3576 }, { "epoch": 0.3, "learning_rate": 1.6342143591726925e-05, "loss": 1.124, "step": 3577 }, { "epoch": 0.3, "learning_rate": 1.634002255933365e-05, "loss": 1.151, "step": 3578 }, { "epoch": 0.3, "learning_rate": 1.6337901049903418e-05, "loss": 1.0697, "step": 3579 }, { "epoch": 0.3, "learning_rate": 1.6335779063595857e-05, "loss": 1.0964, "step": 3580 }, { "epoch": 0.3, "learning_rate": 1.633365660057063e-05, "loss": 1.1051, "step": 3581 }, { "epoch": 0.3, "learning_rate": 1.633153366098744e-05, "loss": 1.0712, "step": 3582 }, { "epoch": 0.3, "learning_rate": 1.6329410245006015e-05, "loss": 1.0924, "step": 3583 }, { "epoch": 0.3, "learning_rate": 1.6327286352786126e-05, "loss": 1.1074, "step": 3584 }, { "epoch": 0.3, "learning_rate": 1.6325161984487588e-05, "loss": 1.1032, "step": 3585 }, { "epoch": 0.3, "learning_rate": 1.632303714027023e-05, "loss": 1.0962, "step": 3586 }, { "epoch": 0.3, "learning_rate": 1.6320911820293938e-05, "loss": 1.0684, "step": 3587 }, { "epoch": 0.3, "learning_rate": 1.6318786024718626e-05, "loss": 1.1271, "step": 3588 }, { "epoch": 0.3, "learning_rate": 1.631665975370424e-05, "loss": 1.0877, "step": 3589 }, { "epoch": 0.3, "learning_rate": 1.6314533007410764e-05, "loss": 1.0988, "step": 3590 }, { "epoch": 0.3, "learning_rate": 1.6312405785998223e-05, "loss": 1.0942, "step": 3591 }, { "epoch": 0.3, "learning_rate": 1.6310278089626672e-05, "loss": 1.1359, "step": 3592 }, { "epoch": 0.3, "learning_rate": 1.6308149918456207e-05, "loss": 0.974, "step": 3593 }, { "epoch": 0.3, "learning_rate": 1.630602127264695e-05, "loss": 1.0956, "step": 3594 }, { "epoch": 0.3, "learning_rate": 1.630389215235907e-05, "loss": 1.09, "step": 3595 }, { "epoch": 0.3, "learning_rate": 1.6301762557752764e-05, "loss": 1.0944, "step": 3596 }, { "epoch": 0.3, "learning_rate": 1.6299632488988265e-05, "loss": 1.12, "step": 3597 }, { "epoch": 0.3, "learning_rate": 1.6297501946225846e-05, "loss": 1.0357, "step": 3598 }, { "epoch": 0.3, "learning_rate": 1.6295370929625817e-05, "loss": 1.0959, "step": 3599 }, { "epoch": 0.3, "learning_rate": 1.6293239439348517e-05, "loss": 1.1066, "step": 3600 }, { "epoch": 0.3, "learning_rate": 1.6291107475554325e-05, "loss": 1.0821, "step": 3601 }, { "epoch": 0.31, "learning_rate": 1.628897503840365e-05, "loss": 1.1003, "step": 3602 }, { "epoch": 0.31, "learning_rate": 1.628684212805695e-05, "loss": 1.0546, "step": 3603 }, { "epoch": 0.31, "learning_rate": 1.6284708744674703e-05, "loss": 1.1207, "step": 3604 }, { "epoch": 0.31, "learning_rate": 1.6282574888417433e-05, "loss": 1.1114, "step": 3605 }, { "epoch": 0.31, "learning_rate": 1.6280440559445692e-05, "loss": 1.1188, "step": 3606 }, { "epoch": 0.31, "learning_rate": 1.6278305757920074e-05, "loss": 1.1262, "step": 3607 }, { "epoch": 0.31, "learning_rate": 1.627617048400121e-05, "loss": 1.0544, "step": 3608 }, { "epoch": 0.31, "learning_rate": 1.627403473784975e-05, "loss": 1.0521, "step": 3609 }, { "epoch": 0.31, "learning_rate": 1.627189851962641e-05, "loss": 1.0806, "step": 3610 }, { "epoch": 0.31, "learning_rate": 1.626976182949191e-05, "loss": 1.0736, "step": 3611 }, { "epoch": 0.31, "learning_rate": 1.626762466760702e-05, "loss": 1.0854, "step": 3612 }, { "epoch": 0.31, "learning_rate": 1.626548703413255e-05, "loss": 1.0951, "step": 3613 }, { "epoch": 0.31, "learning_rate": 1.6263348929229338e-05, "loss": 0.8877, "step": 3614 }, { "epoch": 0.31, "learning_rate": 1.6261210353058264e-05, "loss": 1.0984, "step": 3615 }, { "epoch": 0.31, "learning_rate": 1.625907130578023e-05, "loss": 1.0691, "step": 3616 }, { "epoch": 0.31, "learning_rate": 1.625693178755619e-05, "loss": 1.0621, "step": 3617 }, { "epoch": 0.31, "learning_rate": 1.6254791798547122e-05, "loss": 1.1107, "step": 3618 }, { "epoch": 0.31, "learning_rate": 1.6252651338914044e-05, "loss": 1.0971, "step": 3619 }, { "epoch": 0.31, "learning_rate": 1.6250510408818013e-05, "loss": 1.1286, "step": 3620 }, { "epoch": 0.31, "learning_rate": 1.624836900842011e-05, "loss": 1.0686, "step": 3621 }, { "epoch": 0.31, "learning_rate": 1.6246227137881464e-05, "loss": 1.0982, "step": 3622 }, { "epoch": 0.31, "learning_rate": 1.6244084797363236e-05, "loss": 1.0803, "step": 3623 }, { "epoch": 0.31, "learning_rate": 1.6241941987026612e-05, "loss": 0.9176, "step": 3624 }, { "epoch": 0.31, "learning_rate": 1.623979870703283e-05, "loss": 1.0617, "step": 3625 }, { "epoch": 0.31, "learning_rate": 1.6237654957543147e-05, "loss": 1.1115, "step": 3626 }, { "epoch": 0.31, "learning_rate": 1.6235510738718868e-05, "loss": 1.1177, "step": 3627 }, { "epoch": 0.31, "learning_rate": 1.623336605072133e-05, "loss": 1.1692, "step": 3628 }, { "epoch": 0.31, "learning_rate": 1.6231220893711902e-05, "loss": 1.0231, "step": 3629 }, { "epoch": 0.31, "learning_rate": 1.622907526785199e-05, "loss": 1.0486, "step": 3630 }, { "epoch": 0.31, "learning_rate": 1.6226929173303034e-05, "loss": 1.092, "step": 3631 }, { "epoch": 0.31, "learning_rate": 1.6224782610226515e-05, "loss": 1.0911, "step": 3632 }, { "epoch": 0.31, "learning_rate": 1.6222635578783944e-05, "loss": 1.0774, "step": 3633 }, { "epoch": 0.31, "learning_rate": 1.6220488079136864e-05, "loss": 0.9167, "step": 3634 }, { "epoch": 0.31, "learning_rate": 1.6218340111446862e-05, "loss": 1.063, "step": 3635 }, { "epoch": 0.31, "learning_rate": 1.6216191675875555e-05, "loss": 1.131, "step": 3636 }, { "epoch": 0.31, "learning_rate": 1.6214042772584595e-05, "loss": 1.083, "step": 3637 }, { "epoch": 0.31, "learning_rate": 1.621189340173567e-05, "loss": 1.117, "step": 3638 }, { "epoch": 0.31, "learning_rate": 1.62097435634905e-05, "loss": 1.1004, "step": 3639 }, { "epoch": 0.31, "learning_rate": 1.620759325801085e-05, "loss": 1.0968, "step": 3640 }, { "epoch": 0.31, "learning_rate": 1.620544248545851e-05, "loss": 1.1053, "step": 3641 }, { "epoch": 0.31, "learning_rate": 1.6203291245995312e-05, "loss": 1.1343, "step": 3642 }, { "epoch": 0.31, "learning_rate": 1.6201139539783116e-05, "loss": 1.092, "step": 3643 }, { "epoch": 0.31, "learning_rate": 1.619898736698382e-05, "loss": 1.0782, "step": 3644 }, { "epoch": 0.31, "learning_rate": 1.619683472775936e-05, "loss": 1.1082, "step": 3645 }, { "epoch": 0.31, "learning_rate": 1.619468162227171e-05, "loss": 1.0798, "step": 3646 }, { "epoch": 0.31, "learning_rate": 1.6192528050682868e-05, "loss": 1.1378, "step": 3647 }, { "epoch": 0.31, "learning_rate": 1.6190374013154875e-05, "loss": 1.0904, "step": 3648 }, { "epoch": 0.31, "learning_rate": 1.6188219509849806e-05, "loss": 1.1275, "step": 3649 }, { "epoch": 0.31, "learning_rate": 1.6186064540929767e-05, "loss": 1.1589, "step": 3650 }, { "epoch": 0.31, "learning_rate": 1.618390910655691e-05, "loss": 1.0665, "step": 3651 }, { "epoch": 0.31, "learning_rate": 1.6181753206893408e-05, "loss": 1.0991, "step": 3652 }, { "epoch": 0.31, "learning_rate": 1.617959684210148e-05, "loss": 1.0937, "step": 3653 }, { "epoch": 0.31, "learning_rate": 1.6177440012343372e-05, "loss": 1.0966, "step": 3654 }, { "epoch": 0.31, "learning_rate": 1.6175282717781372e-05, "loss": 1.1508, "step": 3655 }, { "epoch": 0.31, "learning_rate": 1.6173124958577793e-05, "loss": 1.0905, "step": 3656 }, { "epoch": 0.31, "learning_rate": 1.6170966734894996e-05, "loss": 1.1135, "step": 3657 }, { "epoch": 0.31, "learning_rate": 1.6168808046895364e-05, "loss": 1.1181, "step": 3658 }, { "epoch": 0.31, "learning_rate": 1.6166648894741333e-05, "loss": 1.0602, "step": 3659 }, { "epoch": 0.31, "learning_rate": 1.616448927859535e-05, "loss": 1.0168, "step": 3660 }, { "epoch": 0.31, "learning_rate": 1.6162329198619914e-05, "loss": 1.0533, "step": 3661 }, { "epoch": 0.31, "learning_rate": 1.6160168654977555e-05, "loss": 0.9041, "step": 3662 }, { "epoch": 0.31, "learning_rate": 1.6158007647830835e-05, "loss": 1.123, "step": 3663 }, { "epoch": 0.31, "learning_rate": 1.6155846177342354e-05, "loss": 1.0398, "step": 3664 }, { "epoch": 0.31, "learning_rate": 1.6153684243674747e-05, "loss": 0.9513, "step": 3665 }, { "epoch": 0.31, "learning_rate": 1.615152184699068e-05, "loss": 1.1361, "step": 3666 }, { "epoch": 0.31, "learning_rate": 1.6149358987452856e-05, "loss": 1.0435, "step": 3667 }, { "epoch": 0.31, "learning_rate": 1.6147195665224017e-05, "loss": 1.0978, "step": 3668 }, { "epoch": 0.31, "learning_rate": 1.6145031880466934e-05, "loss": 1.0935, "step": 3669 }, { "epoch": 0.31, "learning_rate": 1.614286763334441e-05, "loss": 1.0748, "step": 3670 }, { "epoch": 0.31, "learning_rate": 1.6140702924019296e-05, "loss": 1.1371, "step": 3671 }, { "epoch": 0.31, "learning_rate": 1.6138537752654465e-05, "loss": 1.0978, "step": 3672 }, { "epoch": 0.31, "learning_rate": 1.6136372119412828e-05, "loss": 1.1242, "step": 3673 }, { "epoch": 0.31, "learning_rate": 1.6134206024457336e-05, "loss": 1.0204, "step": 3674 }, { "epoch": 0.31, "learning_rate": 1.6132039467950966e-05, "loss": 1.0894, "step": 3675 }, { "epoch": 0.31, "learning_rate": 1.612987245005674e-05, "loss": 1.0903, "step": 3676 }, { "epoch": 0.31, "learning_rate": 1.61277049709377e-05, "loss": 0.9509, "step": 3677 }, { "epoch": 0.31, "learning_rate": 1.612553703075694e-05, "loss": 1.0539, "step": 3678 }, { "epoch": 0.31, "learning_rate": 1.612336862967758e-05, "loss": 1.0404, "step": 3679 }, { "epoch": 0.31, "learning_rate": 1.612119976786277e-05, "loss": 1.0429, "step": 3680 }, { "epoch": 0.31, "learning_rate": 1.6119030445475706e-05, "loss": 1.1219, "step": 3681 }, { "epoch": 0.31, "learning_rate": 1.6116860662679606e-05, "loss": 1.015, "step": 3682 }, { "epoch": 0.31, "learning_rate": 1.6114690419637736e-05, "loss": 1.0594, "step": 3683 }, { "epoch": 0.31, "learning_rate": 1.6112519716513387e-05, "loss": 1.0484, "step": 3684 }, { "epoch": 0.31, "learning_rate": 1.6110348553469883e-05, "loss": 1.0517, "step": 3685 }, { "epoch": 0.31, "learning_rate": 1.6108176930670594e-05, "loss": 1.1337, "step": 3686 }, { "epoch": 0.31, "learning_rate": 1.610600484827891e-05, "loss": 1.0793, "step": 3687 }, { "epoch": 0.31, "learning_rate": 1.6103832306458273e-05, "loss": 1.0997, "step": 3688 }, { "epoch": 0.31, "learning_rate": 1.610165930537214e-05, "loss": 1.0374, "step": 3689 }, { "epoch": 0.31, "learning_rate": 1.6099485845184016e-05, "loss": 0.9396, "step": 3690 }, { "epoch": 0.31, "learning_rate": 1.609731192605744e-05, "loss": 1.0482, "step": 3691 }, { "epoch": 0.31, "learning_rate": 1.6095137548155978e-05, "loss": 1.0646, "step": 3692 }, { "epoch": 0.31, "learning_rate": 1.6092962711643238e-05, "loss": 1.0863, "step": 3693 }, { "epoch": 0.31, "learning_rate": 1.6090787416682855e-05, "loss": 1.0795, "step": 3694 }, { "epoch": 0.31, "learning_rate": 1.6088611663438506e-05, "loss": 1.0764, "step": 3695 }, { "epoch": 0.31, "learning_rate": 1.60864354520739e-05, "loss": 0.9511, "step": 3696 }, { "epoch": 0.31, "learning_rate": 1.608425878275278e-05, "loss": 1.1161, "step": 3697 }, { "epoch": 0.31, "learning_rate": 1.6082081655638925e-05, "loss": 1.078, "step": 3698 }, { "epoch": 0.31, "learning_rate": 1.607990407089614e-05, "loss": 1.1076, "step": 3699 }, { "epoch": 0.31, "learning_rate": 1.6077726028688275e-05, "loss": 1.0958, "step": 3700 }, { "epoch": 0.31, "learning_rate": 1.607554752917921e-05, "loss": 1.1033, "step": 3701 }, { "epoch": 0.31, "learning_rate": 1.6073368572532863e-05, "loss": 1.0892, "step": 3702 }, { "epoch": 0.31, "learning_rate": 1.6071189158913182e-05, "loss": 1.066, "step": 3703 }, { "epoch": 0.31, "learning_rate": 1.6069009288484148e-05, "loss": 1.0562, "step": 3704 }, { "epoch": 0.31, "learning_rate": 1.606682896140978e-05, "loss": 1.0832, "step": 3705 }, { "epoch": 0.31, "learning_rate": 1.6064648177854134e-05, "loss": 1.0996, "step": 3706 }, { "epoch": 0.31, "learning_rate": 1.6062466937981295e-05, "loss": 1.0663, "step": 3707 }, { "epoch": 0.31, "learning_rate": 1.606028524195538e-05, "loss": 1.0918, "step": 3708 }, { "epoch": 0.31, "learning_rate": 1.605810308994055e-05, "loss": 1.1028, "step": 3709 }, { "epoch": 0.31, "learning_rate": 1.6055920482100994e-05, "loss": 1.0647, "step": 3710 }, { "epoch": 0.31, "learning_rate": 1.6053737418600934e-05, "loss": 1.0037, "step": 3711 }, { "epoch": 0.31, "learning_rate": 1.6051553899604632e-05, "loss": 1.0696, "step": 3712 }, { "epoch": 0.31, "learning_rate": 1.6049369925276373e-05, "loss": 1.1135, "step": 3713 }, { "epoch": 0.31, "learning_rate": 1.604718549578049e-05, "loss": 1.0955, "step": 3714 }, { "epoch": 0.31, "learning_rate": 1.6045000611281347e-05, "loss": 1.0635, "step": 3715 }, { "epoch": 0.31, "learning_rate": 1.6042815271943332e-05, "loss": 1.1417, "step": 3716 }, { "epoch": 0.31, "learning_rate": 1.604062947793088e-05, "loss": 1.1024, "step": 3717 }, { "epoch": 0.31, "learning_rate": 1.6038443229408455e-05, "loss": 0.8677, "step": 3718 }, { "epoch": 0.31, "learning_rate": 1.603625652654055e-05, "loss": 1.0986, "step": 3719 }, { "epoch": 0.32, "learning_rate": 1.60340693694917e-05, "loss": 1.0978, "step": 3720 }, { "epoch": 0.32, "learning_rate": 1.603188175842647e-05, "loss": 1.0801, "step": 3721 }, { "epoch": 0.32, "learning_rate": 1.602969369350947e-05, "loss": 1.109, "step": 3722 }, { "epoch": 0.32, "learning_rate": 1.6027505174905323e-05, "loss": 1.1013, "step": 3723 }, { "epoch": 0.32, "learning_rate": 1.60253162027787e-05, "loss": 1.0985, "step": 3724 }, { "epoch": 0.32, "learning_rate": 1.602312677729431e-05, "loss": 1.1307, "step": 3725 }, { "epoch": 0.32, "learning_rate": 1.602093689861688e-05, "loss": 1.0684, "step": 3726 }, { "epoch": 0.32, "learning_rate": 1.601874656691119e-05, "loss": 1.1392, "step": 3727 }, { "epoch": 0.32, "learning_rate": 1.6016555782342048e-05, "loss": 1.0717, "step": 3728 }, { "epoch": 0.32, "learning_rate": 1.6014364545074282e-05, "loss": 1.0191, "step": 3729 }, { "epoch": 0.32, "learning_rate": 1.601217285527277e-05, "loss": 1.0662, "step": 3730 }, { "epoch": 0.32, "learning_rate": 1.6009980713102425e-05, "loss": 1.0946, "step": 3731 }, { "epoch": 0.32, "learning_rate": 1.600778811872818e-05, "loss": 1.0695, "step": 3732 }, { "epoch": 0.32, "learning_rate": 1.600559507231502e-05, "loss": 1.1016, "step": 3733 }, { "epoch": 0.32, "learning_rate": 1.6003401574027946e-05, "loss": 1.0455, "step": 3734 }, { "epoch": 0.32, "learning_rate": 1.6001207624032006e-05, "loss": 1.0949, "step": 3735 }, { "epoch": 0.32, "learning_rate": 1.5999013222492275e-05, "loss": 1.0815, "step": 3736 }, { "epoch": 0.32, "learning_rate": 1.5996818369573866e-05, "loss": 1.0566, "step": 3737 }, { "epoch": 0.32, "learning_rate": 1.5994623065441925e-05, "loss": 1.1023, "step": 3738 }, { "epoch": 0.32, "learning_rate": 1.599242731026163e-05, "loss": 1.097, "step": 3739 }, { "epoch": 0.32, "learning_rate": 1.5990231104198195e-05, "loss": 1.0436, "step": 3740 }, { "epoch": 0.32, "learning_rate": 1.598803444741687e-05, "loss": 1.066, "step": 3741 }, { "epoch": 0.32, "learning_rate": 1.5985837340082932e-05, "loss": 1.1156, "step": 3742 }, { "epoch": 0.32, "learning_rate": 1.5983639782361694e-05, "loss": 1.1192, "step": 3743 }, { "epoch": 0.32, "learning_rate": 1.598144177441851e-05, "loss": 1.0254, "step": 3744 }, { "epoch": 0.32, "learning_rate": 1.5979243316418765e-05, "loss": 1.0958, "step": 3745 }, { "epoch": 0.32, "learning_rate": 1.597704440852787e-05, "loss": 1.0536, "step": 3746 }, { "epoch": 0.32, "learning_rate": 1.5974845050911276e-05, "loss": 0.9793, "step": 3747 }, { "epoch": 0.32, "learning_rate": 1.5972645243734473e-05, "loss": 1.1346, "step": 3748 }, { "epoch": 0.32, "learning_rate": 1.597044498716297e-05, "loss": 1.0796, "step": 3749 }, { "epoch": 0.32, "learning_rate": 1.596824428136233e-05, "loss": 1.0842, "step": 3750 }, { "epoch": 0.32, "learning_rate": 1.596604312649813e-05, "loss": 1.0834, "step": 3751 }, { "epoch": 0.32, "learning_rate": 1.5963841522735995e-05, "loss": 1.0418, "step": 3752 }, { "epoch": 0.32, "learning_rate": 1.5961639470241576e-05, "loss": 1.056, "step": 3753 }, { "epoch": 0.32, "learning_rate": 1.595943696918056e-05, "loss": 1.1071, "step": 3754 }, { "epoch": 0.32, "learning_rate": 1.5957234019718666e-05, "loss": 1.042, "step": 3755 }, { "epoch": 0.32, "learning_rate": 1.5955030622021654e-05, "loss": 1.0978, "step": 3756 }, { "epoch": 0.32, "learning_rate": 1.595282677625531e-05, "loss": 1.0987, "step": 3757 }, { "epoch": 0.32, "learning_rate": 1.5950622482585454e-05, "loss": 1.0245, "step": 3758 }, { "epoch": 0.32, "learning_rate": 1.5948417741177943e-05, "loss": 1.0565, "step": 3759 }, { "epoch": 0.32, "learning_rate": 1.5946212552198667e-05, "loss": 1.1331, "step": 3760 }, { "epoch": 0.32, "learning_rate": 1.594400691581355e-05, "loss": 1.1224, "step": 3761 }, { "epoch": 0.32, "learning_rate": 1.5941800832188546e-05, "loss": 1.0576, "step": 3762 }, { "epoch": 0.32, "learning_rate": 1.593959430148965e-05, "loss": 1.1646, "step": 3763 }, { "epoch": 0.32, "learning_rate": 1.593738732388288e-05, "loss": 1.1022, "step": 3764 }, { "epoch": 0.32, "learning_rate": 1.5935179899534298e-05, "loss": 1.0972, "step": 3765 }, { "epoch": 0.32, "learning_rate": 1.593297202861e-05, "loss": 1.0305, "step": 3766 }, { "epoch": 0.32, "learning_rate": 1.59307637112761e-05, "loss": 1.0613, "step": 3767 }, { "epoch": 0.32, "learning_rate": 1.5928554947698764e-05, "loss": 1.0793, "step": 3768 }, { "epoch": 0.32, "learning_rate": 1.5926345738044178e-05, "loss": 1.1184, "step": 3769 }, { "epoch": 0.32, "learning_rate": 1.5924136082478577e-05, "loss": 1.1509, "step": 3770 }, { "epoch": 0.32, "learning_rate": 1.5921925981168213e-05, "loss": 1.0401, "step": 3771 }, { "epoch": 0.32, "learning_rate": 1.591971543427938e-05, "loss": 1.0799, "step": 3772 }, { "epoch": 0.32, "learning_rate": 1.5917504441978406e-05, "loss": 1.0869, "step": 3773 }, { "epoch": 0.32, "learning_rate": 1.591529300443165e-05, "loss": 1.1179, "step": 3774 }, { "epoch": 0.32, "learning_rate": 1.591308112180551e-05, "loss": 1.0947, "step": 3775 }, { "epoch": 0.32, "learning_rate": 1.5910868794266403e-05, "loss": 1.12, "step": 3776 }, { "epoch": 0.32, "learning_rate": 1.5908656021980795e-05, "loss": 1.0652, "step": 3777 }, { "epoch": 0.32, "learning_rate": 1.5906442805115183e-05, "loss": 1.1253, "step": 3778 }, { "epoch": 0.32, "learning_rate": 1.5904229143836086e-05, "loss": 1.064, "step": 3779 }, { "epoch": 0.32, "learning_rate": 1.590201503831007e-05, "loss": 1.0375, "step": 3780 }, { "epoch": 0.32, "learning_rate": 1.589980048870373e-05, "loss": 1.0866, "step": 3781 }, { "epoch": 0.32, "learning_rate": 1.5897585495183688e-05, "loss": 1.0881, "step": 3782 }, { "epoch": 0.32, "learning_rate": 1.5895370057916615e-05, "loss": 1.0966, "step": 3783 }, { "epoch": 0.32, "learning_rate": 1.5893154177069195e-05, "loss": 1.1476, "step": 3784 }, { "epoch": 0.32, "learning_rate": 1.589093785280816e-05, "loss": 1.0507, "step": 3785 }, { "epoch": 0.32, "learning_rate": 1.5888721085300267e-05, "loss": 1.0552, "step": 3786 }, { "epoch": 0.32, "learning_rate": 1.5886503874712316e-05, "loss": 1.035, "step": 3787 }, { "epoch": 0.32, "learning_rate": 1.5884286221211133e-05, "loss": 1.1213, "step": 3788 }, { "epoch": 0.32, "learning_rate": 1.5882068124963578e-05, "loss": 1.1093, "step": 3789 }, { "epoch": 0.32, "learning_rate": 1.5879849586136548e-05, "loss": 1.1416, "step": 3790 }, { "epoch": 0.32, "learning_rate": 1.587763060489697e-05, "loss": 1.0608, "step": 3791 }, { "epoch": 0.32, "learning_rate": 1.5875411181411798e-05, "loss": 1.1035, "step": 3792 }, { "epoch": 0.32, "learning_rate": 1.5873191315848037e-05, "loss": 1.1175, "step": 3793 }, { "epoch": 0.32, "learning_rate": 1.5870971008372706e-05, "loss": 1.0785, "step": 3794 }, { "epoch": 0.32, "learning_rate": 1.586875025915287e-05, "loss": 1.1151, "step": 3795 }, { "epoch": 0.32, "learning_rate": 1.5866529068355617e-05, "loss": 1.0545, "step": 3796 }, { "epoch": 0.32, "learning_rate": 1.5864307436148085e-05, "loss": 1.1121, "step": 3797 }, { "epoch": 0.32, "learning_rate": 1.5862085362697425e-05, "loss": 1.0872, "step": 3798 }, { "epoch": 0.32, "learning_rate": 1.5859862848170838e-05, "loss": 1.098, "step": 3799 }, { "epoch": 0.32, "learning_rate": 1.5857639892735544e-05, "loss": 1.0964, "step": 3800 }, { "epoch": 0.32, "learning_rate": 1.5855416496558806e-05, "loss": 1.1133, "step": 3801 }, { "epoch": 0.32, "learning_rate": 1.5853192659807916e-05, "loss": 1.0842, "step": 3802 }, { "epoch": 0.32, "learning_rate": 1.5850968382650204e-05, "loss": 1.0954, "step": 3803 }, { "epoch": 0.32, "learning_rate": 1.5848743665253023e-05, "loss": 1.0592, "step": 3804 }, { "epoch": 0.32, "learning_rate": 1.584651850778377e-05, "loss": 1.0343, "step": 3805 }, { "epoch": 0.32, "learning_rate": 1.584429291040987e-05, "loss": 1.091, "step": 3806 }, { "epoch": 0.32, "learning_rate": 1.584206687329878e-05, "loss": 1.0841, "step": 3807 }, { "epoch": 0.32, "learning_rate": 1.5839840396617997e-05, "loss": 1.1538, "step": 3808 }, { "epoch": 0.32, "learning_rate": 1.583761348053504e-05, "loss": 1.1186, "step": 3809 }, { "epoch": 0.32, "learning_rate": 1.5835386125217473e-05, "loss": 1.0984, "step": 3810 }, { "epoch": 0.32, "learning_rate": 1.5833158330832876e-05, "loss": 1.0821, "step": 3811 }, { "epoch": 0.32, "learning_rate": 1.5830930097548885e-05, "loss": 1.0083, "step": 3812 }, { "epoch": 0.32, "learning_rate": 1.582870142553315e-05, "loss": 1.0705, "step": 3813 }, { "epoch": 0.32, "learning_rate": 1.5826472314953364e-05, "loss": 1.0776, "step": 3814 }, { "epoch": 0.32, "learning_rate": 1.582424276597725e-05, "loss": 1.1026, "step": 3815 }, { "epoch": 0.32, "learning_rate": 1.582201277877256e-05, "loss": 1.0587, "step": 3816 }, { "epoch": 0.32, "learning_rate": 1.581978235350709e-05, "loss": 1.1423, "step": 3817 }, { "epoch": 0.32, "learning_rate": 1.581755149034866e-05, "loss": 1.1777, "step": 3818 }, { "epoch": 0.32, "learning_rate": 1.581532018946512e-05, "loss": 1.0303, "step": 3819 }, { "epoch": 0.32, "learning_rate": 1.5813088451024363e-05, "loss": 1.0411, "step": 3820 }, { "epoch": 0.32, "learning_rate": 1.5810856275194307e-05, "loss": 1.0664, "step": 3821 }, { "epoch": 0.32, "learning_rate": 1.5808623662142907e-05, "loss": 1.06, "step": 3822 }, { "epoch": 0.32, "learning_rate": 1.580639061203815e-05, "loss": 1.0552, "step": 3823 }, { "epoch": 0.32, "learning_rate": 1.5804157125048056e-05, "loss": 1.0826, "step": 3824 }, { "epoch": 0.32, "learning_rate": 1.5801923201340675e-05, "loss": 1.0884, "step": 3825 }, { "epoch": 0.32, "learning_rate": 1.5799688841084097e-05, "loss": 1.116, "step": 3826 }, { "epoch": 0.32, "learning_rate": 1.579745404444643e-05, "loss": 1.0318, "step": 3827 }, { "epoch": 0.32, "learning_rate": 1.5795218811595837e-05, "loss": 0.9184, "step": 3828 }, { "epoch": 0.32, "learning_rate": 1.5792983142700497e-05, "loss": 1.0866, "step": 3829 }, { "epoch": 0.32, "learning_rate": 1.579074703792862e-05, "loss": 1.0261, "step": 3830 }, { "epoch": 0.32, "learning_rate": 1.5788510497448467e-05, "loss": 1.1251, "step": 3831 }, { "epoch": 0.32, "learning_rate": 1.5786273521428313e-05, "loss": 1.0936, "step": 3832 }, { "epoch": 0.32, "learning_rate": 1.5784036110036473e-05, "loss": 1.0984, "step": 3833 }, { "epoch": 0.32, "learning_rate": 1.5781798263441297e-05, "loss": 1.0974, "step": 3834 }, { "epoch": 0.32, "learning_rate": 1.5779559981811165e-05, "loss": 1.117, "step": 3835 }, { "epoch": 0.32, "learning_rate": 1.5777321265314486e-05, "loss": 1.0658, "step": 3836 }, { "epoch": 0.32, "learning_rate": 1.5775082114119714e-05, "loss": 0.8782, "step": 3837 }, { "epoch": 0.33, "learning_rate": 1.5772842528395318e-05, "loss": 1.0819, "step": 3838 }, { "epoch": 0.33, "learning_rate": 1.5770602508309818e-05, "loss": 1.0856, "step": 3839 }, { "epoch": 0.33, "learning_rate": 1.5768362054031753e-05, "loss": 1.0917, "step": 3840 }, { "epoch": 0.33, "learning_rate": 1.57661211657297e-05, "loss": 1.1126, "step": 3841 }, { "epoch": 0.33, "learning_rate": 1.5763879843572268e-05, "loss": 1.083, "step": 3842 }, { "epoch": 0.33, "learning_rate": 1.57616380877281e-05, "loss": 1.0673, "step": 3843 }, { "epoch": 0.33, "learning_rate": 1.575939589836587e-05, "loss": 1.0855, "step": 3844 }, { "epoch": 0.33, "learning_rate": 1.575715327565429e-05, "loss": 1.0726, "step": 3845 }, { "epoch": 0.33, "learning_rate": 1.575491021976209e-05, "loss": 1.0738, "step": 3846 }, { "epoch": 0.33, "learning_rate": 1.575266673085805e-05, "loss": 1.0503, "step": 3847 }, { "epoch": 0.33, "learning_rate": 1.575042280911097e-05, "loss": 1.1355, "step": 3848 }, { "epoch": 0.33, "learning_rate": 1.5748178454689694e-05, "loss": 1.1106, "step": 3849 }, { "epoch": 0.33, "learning_rate": 1.5745933667763084e-05, "loss": 1.1102, "step": 3850 }, { "epoch": 0.33, "learning_rate": 1.574368844850005e-05, "loss": 1.1037, "step": 3851 }, { "epoch": 0.33, "learning_rate": 1.5741442797069523e-05, "loss": 1.0208, "step": 3852 }, { "epoch": 0.33, "learning_rate": 1.5739196713640473e-05, "loss": 1.1115, "step": 3853 }, { "epoch": 0.33, "learning_rate": 1.5736950198381894e-05, "loss": 1.0982, "step": 3854 }, { "epoch": 0.33, "learning_rate": 1.5734703251462827e-05, "loss": 0.8737, "step": 3855 }, { "epoch": 0.33, "learning_rate": 1.5732455873052335e-05, "loss": 1.0993, "step": 3856 }, { "epoch": 0.33, "learning_rate": 1.573020806331951e-05, "loss": 1.1028, "step": 3857 }, { "epoch": 0.33, "learning_rate": 1.5727959822433488e-05, "loss": 1.1037, "step": 3858 }, { "epoch": 0.33, "learning_rate": 1.5725711150563428e-05, "loss": 1.1044, "step": 3859 }, { "epoch": 0.33, "learning_rate": 1.572346204787853e-05, "loss": 1.1063, "step": 3860 }, { "epoch": 0.33, "learning_rate": 1.5721212514548013e-05, "loss": 1.0996, "step": 3861 }, { "epoch": 0.33, "learning_rate": 1.5718962550741147e-05, "loss": 1.0983, "step": 3862 }, { "epoch": 0.33, "learning_rate": 1.5716712156627215e-05, "loss": 1.0716, "step": 3863 }, { "epoch": 0.33, "learning_rate": 1.5714461332375546e-05, "loss": 1.1114, "step": 3864 }, { "epoch": 0.33, "learning_rate": 1.5712210078155495e-05, "loss": 1.0545, "step": 3865 }, { "epoch": 0.33, "learning_rate": 1.5709958394136454e-05, "loss": 1.0707, "step": 3866 }, { "epoch": 0.33, "learning_rate": 1.5707706280487844e-05, "loss": 1.0753, "step": 3867 }, { "epoch": 0.33, "learning_rate": 1.5705453737379113e-05, "loss": 1.0985, "step": 3868 }, { "epoch": 0.33, "learning_rate": 1.5703200764979757e-05, "loss": 1.0632, "step": 3869 }, { "epoch": 0.33, "learning_rate": 1.5700947363459286e-05, "loss": 1.072, "step": 3870 }, { "epoch": 0.33, "learning_rate": 1.5698693532987256e-05, "loss": 1.1087, "step": 3871 }, { "epoch": 0.33, "learning_rate": 1.5696439273733247e-05, "loss": 1.0963, "step": 3872 }, { "epoch": 0.33, "learning_rate": 1.5694184585866875e-05, "loss": 1.0447, "step": 3873 }, { "epoch": 0.33, "learning_rate": 1.569192946955779e-05, "loss": 1.1133, "step": 3874 }, { "epoch": 0.33, "learning_rate": 1.5689673924975668e-05, "loss": 1.1069, "step": 3875 }, { "epoch": 0.33, "learning_rate": 1.5687417952290226e-05, "loss": 1.0777, "step": 3876 }, { "epoch": 0.33, "learning_rate": 1.5685161551671203e-05, "loss": 1.1026, "step": 3877 }, { "epoch": 0.33, "learning_rate": 1.5682904723288375e-05, "loss": 1.0755, "step": 3878 }, { "epoch": 0.33, "learning_rate": 1.568064746731156e-05, "loss": 1.0856, "step": 3879 }, { "epoch": 0.33, "learning_rate": 1.5678389783910587e-05, "loss": 1.1139, "step": 3880 }, { "epoch": 0.33, "learning_rate": 1.5676131673255338e-05, "loss": 1.1127, "step": 3881 }, { "epoch": 0.33, "learning_rate": 1.5673873135515712e-05, "loss": 1.125, "step": 3882 }, { "epoch": 0.33, "learning_rate": 1.5671614170861645e-05, "loss": 1.0859, "step": 3883 }, { "epoch": 0.33, "learning_rate": 1.5669354779463117e-05, "loss": 1.0635, "step": 3884 }, { "epoch": 0.33, "learning_rate": 1.5667094961490115e-05, "loss": 1.0763, "step": 3885 }, { "epoch": 0.33, "learning_rate": 1.566483471711268e-05, "loss": 1.0717, "step": 3886 }, { "epoch": 0.33, "learning_rate": 1.5662574046500886e-05, "loss": 1.0809, "step": 3887 }, { "epoch": 0.33, "learning_rate": 1.5660312949824813e-05, "loss": 1.1037, "step": 3888 }, { "epoch": 0.33, "learning_rate": 1.5658051427254607e-05, "loss": 1.1149, "step": 3889 }, { "epoch": 0.33, "learning_rate": 1.5655789478960418e-05, "loss": 1.1196, "step": 3890 }, { "epoch": 0.33, "learning_rate": 1.5653527105112444e-05, "loss": 1.0939, "step": 3891 }, { "epoch": 0.33, "learning_rate": 1.5651264305880916e-05, "loss": 1.0991, "step": 3892 }, { "epoch": 0.33, "learning_rate": 1.5649001081436083e-05, "loss": 1.0648, "step": 3893 }, { "epoch": 0.33, "learning_rate": 1.564673743194824e-05, "loss": 1.0953, "step": 3894 }, { "epoch": 0.33, "learning_rate": 1.564447335758771e-05, "loss": 1.0374, "step": 3895 }, { "epoch": 0.33, "learning_rate": 1.5642208858524846e-05, "loss": 1.0526, "step": 3896 }, { "epoch": 0.33, "learning_rate": 1.5639943934930034e-05, "loss": 1.1234, "step": 3897 }, { "epoch": 0.33, "learning_rate": 1.5637678586973687e-05, "loss": 1.0697, "step": 3898 }, { "epoch": 0.33, "learning_rate": 1.563541281482626e-05, "loss": 1.0939, "step": 3899 }, { "epoch": 0.33, "learning_rate": 1.5633146618658228e-05, "loss": 1.0826, "step": 3900 }, { "epoch": 0.33, "learning_rate": 1.5630879998640117e-05, "loss": 1.0238, "step": 3901 }, { "epoch": 0.33, "learning_rate": 1.562861295494246e-05, "loss": 1.1249, "step": 3902 }, { "epoch": 0.33, "learning_rate": 1.5626345487735838e-05, "loss": 1.0006, "step": 3903 }, { "epoch": 0.33, "learning_rate": 1.5624077597190864e-05, "loss": 1.0794, "step": 3904 }, { "epoch": 0.33, "learning_rate": 1.5621809283478175e-05, "loss": 1.1169, "step": 3905 }, { "epoch": 0.33, "learning_rate": 1.5619540546768446e-05, "loss": 1.0801, "step": 3906 }, { "epoch": 0.33, "learning_rate": 1.561727138723238e-05, "loss": 1.1098, "step": 3907 }, { "epoch": 0.33, "learning_rate": 1.561500180504071e-05, "loss": 1.1165, "step": 3908 }, { "epoch": 0.33, "learning_rate": 1.5612731800364212e-05, "loss": 1.0517, "step": 3909 }, { "epoch": 0.33, "learning_rate": 1.561046137337368e-05, "loss": 1.1315, "step": 3910 }, { "epoch": 0.33, "learning_rate": 1.560819052423995e-05, "loss": 1.0788, "step": 3911 }, { "epoch": 0.33, "learning_rate": 1.5605919253133883e-05, "loss": 1.1014, "step": 3912 }, { "epoch": 0.33, "learning_rate": 1.5603647560226374e-05, "loss": 1.0544, "step": 3913 }, { "epoch": 0.33, "learning_rate": 1.560137544568835e-05, "loss": 1.0491, "step": 3914 }, { "epoch": 0.33, "learning_rate": 1.5599102909690776e-05, "loss": 1.1219, "step": 3915 }, { "epoch": 0.33, "learning_rate": 1.5596829952404634e-05, "loss": 1.0491, "step": 3916 }, { "epoch": 0.33, "learning_rate": 1.559455657400095e-05, "loss": 1.096, "step": 3917 }, { "epoch": 0.33, "learning_rate": 1.5592282774650774e-05, "loss": 1.0533, "step": 3918 }, { "epoch": 0.33, "learning_rate": 1.5590008554525196e-05, "loss": 1.1226, "step": 3919 }, { "epoch": 0.33, "learning_rate": 1.558773391379534e-05, "loss": 1.1413, "step": 3920 }, { "epoch": 0.33, "learning_rate": 1.5585458852632337e-05, "loss": 1.0794, "step": 3921 }, { "epoch": 0.33, "learning_rate": 1.558318337120738e-05, "loss": 1.0443, "step": 3922 }, { "epoch": 0.33, "learning_rate": 1.5580907469691683e-05, "loss": 1.0861, "step": 3923 }, { "epoch": 0.33, "learning_rate": 1.557863114825648e-05, "loss": 1.1044, "step": 3924 }, { "epoch": 0.33, "learning_rate": 1.5576354407073057e-05, "loss": 1.1265, "step": 3925 }, { "epoch": 0.33, "learning_rate": 1.557407724631271e-05, "loss": 1.1034, "step": 3926 }, { "epoch": 0.33, "learning_rate": 1.5571799666146784e-05, "loss": 1.0887, "step": 3927 }, { "epoch": 0.33, "learning_rate": 1.556952166674665e-05, "loss": 1.0654, "step": 3928 }, { "epoch": 0.33, "learning_rate": 1.556724324828371e-05, "loss": 1.0826, "step": 3929 }, { "epoch": 0.33, "learning_rate": 1.5564964410929394e-05, "loss": 1.1222, "step": 3930 }, { "epoch": 0.33, "learning_rate": 1.5562685154855165e-05, "loss": 1.0804, "step": 3931 }, { "epoch": 0.33, "learning_rate": 1.556040548023252e-05, "loss": 1.107, "step": 3932 }, { "epoch": 0.33, "learning_rate": 1.555812538723299e-05, "loss": 1.0585, "step": 3933 }, { "epoch": 0.33, "learning_rate": 1.5555844876028133e-05, "loss": 1.0493, "step": 3934 }, { "epoch": 0.33, "learning_rate": 1.5553563946789537e-05, "loss": 1.1185, "step": 3935 }, { "epoch": 0.33, "learning_rate": 1.5551282599688832e-05, "loss": 1.1179, "step": 3936 }, { "epoch": 0.33, "learning_rate": 1.5549000834897662e-05, "loss": 1.0968, "step": 3937 }, { "epoch": 0.33, "learning_rate": 1.5546718652587712e-05, "loss": 1.1398, "step": 3938 }, { "epoch": 0.33, "learning_rate": 1.5544436052930704e-05, "loss": 1.1342, "step": 3939 }, { "epoch": 0.33, "learning_rate": 1.5542153036098386e-05, "loss": 1.0358, "step": 3940 }, { "epoch": 0.33, "learning_rate": 1.5539869602262528e-05, "loss": 1.1023, "step": 3941 }, { "epoch": 0.33, "learning_rate": 1.5537585751594954e-05, "loss": 1.0809, "step": 3942 }, { "epoch": 0.33, "learning_rate": 1.5535301484267496e-05, "loss": 1.0599, "step": 3943 }, { "epoch": 0.33, "learning_rate": 1.553301680045203e-05, "loss": 1.0675, "step": 3944 }, { "epoch": 0.33, "learning_rate": 1.553073170032046e-05, "loss": 1.0946, "step": 3945 }, { "epoch": 0.33, "learning_rate": 1.5528446184044723e-05, "loss": 1.1118, "step": 3946 }, { "epoch": 0.33, "learning_rate": 1.5526160251796784e-05, "loss": 1.0874, "step": 3947 }, { "epoch": 0.33, "learning_rate": 1.5523873903748645e-05, "loss": 1.1105, "step": 3948 }, { "epoch": 0.33, "learning_rate": 1.5521587140072332e-05, "loss": 1.1456, "step": 3949 }, { "epoch": 0.33, "learning_rate": 1.5519299960939908e-05, "loss": 1.0871, "step": 3950 }, { "epoch": 0.33, "learning_rate": 1.551701236652347e-05, "loss": 1.0881, "step": 3951 }, { "epoch": 0.33, "learning_rate": 1.5514724356995125e-05, "loss": 1.1147, "step": 3952 }, { "epoch": 0.33, "learning_rate": 1.551243593252705e-05, "loss": 1.103, "step": 3953 }, { "epoch": 0.33, "learning_rate": 1.5510147093291414e-05, "loss": 1.0989, "step": 3954 }, { "epoch": 0.33, "learning_rate": 1.5507857839460446e-05, "loss": 1.0493, "step": 3955 }, { "epoch": 0.34, "learning_rate": 1.5505568171206387e-05, "loss": 1.1311, "step": 3956 }, { "epoch": 0.34, "learning_rate": 1.5503278088701518e-05, "loss": 1.1204, "step": 3957 }, { "epoch": 0.34, "learning_rate": 1.5500987592118145e-05, "loss": 1.1055, "step": 3958 }, { "epoch": 0.34, "learning_rate": 1.5498696681628623e-05, "loss": 1.1168, "step": 3959 }, { "epoch": 0.34, "learning_rate": 1.549640535740531e-05, "loss": 1.1005, "step": 3960 }, { "epoch": 0.34, "learning_rate": 1.5494113619620627e-05, "loss": 1.0731, "step": 3961 }, { "epoch": 0.34, "learning_rate": 1.549182146844699e-05, "loss": 1.1245, "step": 3962 }, { "epoch": 0.34, "learning_rate": 1.5489528904056884e-05, "loss": 1.0833, "step": 3963 }, { "epoch": 0.34, "learning_rate": 1.548723592662279e-05, "loss": 1.0873, "step": 3964 }, { "epoch": 0.34, "learning_rate": 1.548494253631725e-05, "loss": 1.1361, "step": 3965 }, { "epoch": 0.34, "learning_rate": 1.5482648733312812e-05, "loss": 1.0931, "step": 3966 }, { "epoch": 0.34, "learning_rate": 1.5480354517782075e-05, "loss": 1.0719, "step": 3967 }, { "epoch": 0.34, "learning_rate": 1.5478059889897657e-05, "loss": 1.0367, "step": 3968 }, { "epoch": 0.34, "learning_rate": 1.5475764849832215e-05, "loss": 1.1245, "step": 3969 }, { "epoch": 0.34, "learning_rate": 1.5473469397758427e-05, "loss": 1.0364, "step": 3970 }, { "epoch": 0.34, "learning_rate": 1.547117353384901e-05, "loss": 1.0194, "step": 3971 }, { "epoch": 0.34, "learning_rate": 1.5468877258276714e-05, "loss": 1.053, "step": 3972 }, { "epoch": 0.34, "learning_rate": 1.5466580571214307e-05, "loss": 1.035, "step": 3973 }, { "epoch": 0.34, "learning_rate": 1.5464283472834605e-05, "loss": 1.0868, "step": 3974 }, { "epoch": 0.34, "learning_rate": 1.546198596331044e-05, "loss": 1.076, "step": 3975 }, { "epoch": 0.34, "learning_rate": 1.5459688042814684e-05, "loss": 1.1029, "step": 3976 }, { "epoch": 0.34, "learning_rate": 1.5457389711520244e-05, "loss": 1.0953, "step": 3977 }, { "epoch": 0.34, "learning_rate": 1.545509096960004e-05, "loss": 1.0638, "step": 3978 }, { "epoch": 0.34, "learning_rate": 1.545279181722704e-05, "loss": 1.0843, "step": 3979 }, { "epoch": 0.34, "learning_rate": 1.545049225457424e-05, "loss": 1.1222, "step": 3980 }, { "epoch": 0.34, "learning_rate": 1.544819228181466e-05, "loss": 1.0832, "step": 3981 }, { "epoch": 0.34, "learning_rate": 1.5445891899121356e-05, "loss": 1.1457, "step": 3982 }, { "epoch": 0.34, "learning_rate": 1.5443591106667412e-05, "loss": 0.9149, "step": 3983 }, { "epoch": 0.34, "learning_rate": 1.544128990462595e-05, "loss": 1.0724, "step": 3984 }, { "epoch": 0.34, "learning_rate": 1.543898829317011e-05, "loss": 1.1275, "step": 3985 }, { "epoch": 0.34, "learning_rate": 1.5436686272473076e-05, "loss": 1.0438, "step": 3986 }, { "epoch": 0.34, "learning_rate": 1.5434383842708057e-05, "loss": 0.9209, "step": 3987 }, { "epoch": 0.34, "learning_rate": 1.5432081004048288e-05, "loss": 1.1266, "step": 3988 }, { "epoch": 0.34, "learning_rate": 1.5429777756667044e-05, "loss": 1.0859, "step": 3989 }, { "epoch": 0.34, "learning_rate": 1.542747410073762e-05, "loss": 1.0593, "step": 3990 }, { "epoch": 0.34, "learning_rate": 1.5425170036433358e-05, "loss": 1.0752, "step": 3991 }, { "epoch": 0.34, "learning_rate": 1.5422865563927613e-05, "loss": 1.0729, "step": 3992 }, { "epoch": 0.34, "learning_rate": 1.5420560683393778e-05, "loss": 1.05, "step": 3993 }, { "epoch": 0.34, "learning_rate": 1.5418255395005285e-05, "loss": 1.0886, "step": 3994 }, { "epoch": 0.34, "learning_rate": 1.5415949698935582e-05, "loss": 1.0726, "step": 3995 }, { "epoch": 0.34, "learning_rate": 1.5413643595358156e-05, "loss": 1.0143, "step": 3996 }, { "epoch": 0.34, "learning_rate": 1.5411337084446526e-05, "loss": 1.1424, "step": 3997 }, { "epoch": 0.34, "learning_rate": 1.540903016637423e-05, "loss": 1.0628, "step": 3998 }, { "epoch": 0.34, "learning_rate": 1.5406722841314858e-05, "loss": 1.0301, "step": 3999 }, { "epoch": 0.34, "learning_rate": 1.5404415109442012e-05, "loss": 1.0317, "step": 4000 }, { "epoch": 0.34, "learning_rate": 1.540210697092933e-05, "loss": 1.0726, "step": 4001 }, { "epoch": 0.34, "learning_rate": 1.539979842595048e-05, "loss": 1.0491, "step": 4002 }, { "epoch": 0.34, "learning_rate": 1.5397489474679164e-05, "loss": 1.0173, "step": 4003 }, { "epoch": 0.34, "learning_rate": 1.539518011728911e-05, "loss": 1.1203, "step": 4004 }, { "epoch": 0.34, "learning_rate": 1.5392870353954087e-05, "loss": 1.1708, "step": 4005 }, { "epoch": 0.34, "learning_rate": 1.539056018484788e-05, "loss": 1.0923, "step": 4006 }, { "epoch": 0.34, "learning_rate": 1.5388249610144314e-05, "loss": 1.1145, "step": 4007 }, { "epoch": 0.34, "learning_rate": 1.5385938630017233e-05, "loss": 1.0476, "step": 4008 }, { "epoch": 0.34, "learning_rate": 1.538362724464053e-05, "loss": 1.0734, "step": 4009 }, { "epoch": 0.34, "learning_rate": 1.538131545418812e-05, "loss": 1.0499, "step": 4010 }, { "epoch": 0.34, "learning_rate": 1.5379003258833937e-05, "loss": 1.0573, "step": 4011 }, { "epoch": 0.34, "learning_rate": 1.5376690658751964e-05, "loss": 1.1071, "step": 4012 }, { "epoch": 0.34, "learning_rate": 1.5374377654116203e-05, "loss": 1.0882, "step": 4013 }, { "epoch": 0.34, "learning_rate": 1.537206424510069e-05, "loss": 1.0386, "step": 4014 }, { "epoch": 0.34, "learning_rate": 1.5369750431879487e-05, "loss": 1.0682, "step": 4015 }, { "epoch": 0.34, "learning_rate": 1.5367436214626693e-05, "loss": 1.0535, "step": 4016 }, { "epoch": 0.34, "learning_rate": 1.536512159351644e-05, "loss": 1.0442, "step": 4017 }, { "epoch": 0.34, "learning_rate": 1.536280656872288e-05, "loss": 1.0698, "step": 4018 }, { "epoch": 0.34, "learning_rate": 1.5360491140420202e-05, "loss": 1.0846, "step": 4019 }, { "epoch": 0.34, "learning_rate": 1.535817530878262e-05, "loss": 1.09, "step": 4020 }, { "epoch": 0.34, "learning_rate": 1.535585907398439e-05, "loss": 1.0706, "step": 4021 }, { "epoch": 0.34, "learning_rate": 1.535354243619978e-05, "loss": 1.076, "step": 4022 }, { "epoch": 0.34, "learning_rate": 1.5351225395603104e-05, "loss": 1.0986, "step": 4023 }, { "epoch": 0.34, "learning_rate": 1.5348907952368705e-05, "loss": 1.0627, "step": 4024 }, { "epoch": 0.34, "learning_rate": 1.5346590106670944e-05, "loss": 1.089, "step": 4025 }, { "epoch": 0.34, "learning_rate": 1.534427185868423e-05, "loss": 1.0685, "step": 4026 }, { "epoch": 0.34, "learning_rate": 1.5341953208582986e-05, "loss": 1.0479, "step": 4027 }, { "epoch": 0.34, "learning_rate": 1.5339634156541678e-05, "loss": 1.0846, "step": 4028 }, { "epoch": 0.34, "learning_rate": 1.5337314702734787e-05, "loss": 1.0932, "step": 4029 }, { "epoch": 0.34, "learning_rate": 1.5334994847336844e-05, "loss": 1.0904, "step": 4030 }, { "epoch": 0.34, "learning_rate": 1.5332674590522398e-05, "loss": 1.0859, "step": 4031 }, { "epoch": 0.34, "learning_rate": 1.5330353932466025e-05, "loss": 1.0766, "step": 4032 }, { "epoch": 0.34, "learning_rate": 1.532803287334234e-05, "loss": 1.1222, "step": 4033 }, { "epoch": 0.34, "learning_rate": 1.5325711413325984e-05, "loss": 1.014, "step": 4034 }, { "epoch": 0.34, "learning_rate": 1.5323389552591626e-05, "loss": 1.138, "step": 4035 }, { "epoch": 0.34, "learning_rate": 1.5321067291313975e-05, "loss": 1.0775, "step": 4036 }, { "epoch": 0.34, "learning_rate": 1.5318744629667755e-05, "loss": 1.1241, "step": 4037 }, { "epoch": 0.34, "learning_rate": 1.5316421567827732e-05, "loss": 1.0676, "step": 4038 }, { "epoch": 0.34, "learning_rate": 1.5314098105968698e-05, "loss": 0.8604, "step": 4039 }, { "epoch": 0.34, "learning_rate": 1.5311774244265474e-05, "loss": 1.0514, "step": 4040 }, { "epoch": 0.34, "learning_rate": 1.5309449982892912e-05, "loss": 1.1323, "step": 4041 }, { "epoch": 0.34, "learning_rate": 1.5307125322025897e-05, "loss": 1.1014, "step": 4042 }, { "epoch": 0.34, "learning_rate": 1.5304800261839337e-05, "loss": 1.0931, "step": 4043 }, { "epoch": 0.34, "learning_rate": 1.530247480250818e-05, "loss": 1.0128, "step": 4044 }, { "epoch": 0.34, "learning_rate": 1.5300148944207397e-05, "loss": 1.1262, "step": 4045 }, { "epoch": 0.34, "learning_rate": 1.5297822687111986e-05, "loss": 1.0659, "step": 4046 }, { "epoch": 0.34, "learning_rate": 1.529549603139699e-05, "loss": 1.1089, "step": 4047 }, { "epoch": 0.34, "learning_rate": 1.529316897723746e-05, "loss": 1.156, "step": 4048 }, { "epoch": 0.34, "learning_rate": 1.5290841524808497e-05, "loss": 1.0541, "step": 4049 }, { "epoch": 0.34, "learning_rate": 1.5288513674285216e-05, "loss": 1.0936, "step": 4050 }, { "epoch": 0.34, "learning_rate": 1.5286185425842773e-05, "loss": 1.0225, "step": 4051 }, { "epoch": 0.34, "learning_rate": 1.528385677965635e-05, "loss": 1.0997, "step": 4052 }, { "epoch": 0.34, "learning_rate": 1.5281527735901164e-05, "loss": 1.1234, "step": 4053 }, { "epoch": 0.34, "learning_rate": 1.527919829475245e-05, "loss": 1.1054, "step": 4054 }, { "epoch": 0.34, "learning_rate": 1.5276868456385487e-05, "loss": 1.132, "step": 4055 }, { "epoch": 0.34, "learning_rate": 1.527453822097557e-05, "loss": 1.0562, "step": 4056 }, { "epoch": 0.34, "learning_rate": 1.5272207588698037e-05, "loss": 1.0769, "step": 4057 }, { "epoch": 0.34, "learning_rate": 1.5269876559728247e-05, "loss": 1.0674, "step": 4058 }, { "epoch": 0.34, "learning_rate": 1.5267545134241592e-05, "loss": 1.0262, "step": 4059 }, { "epoch": 0.34, "learning_rate": 1.5265213312413496e-05, "loss": 1.0579, "step": 4060 }, { "epoch": 0.34, "learning_rate": 1.5262881094419407e-05, "loss": 1.0631, "step": 4061 }, { "epoch": 0.34, "learning_rate": 1.5260548480434806e-05, "loss": 1.0795, "step": 4062 }, { "epoch": 0.34, "learning_rate": 1.5258215470635206e-05, "loss": 1.1048, "step": 4063 }, { "epoch": 0.34, "learning_rate": 1.5255882065196144e-05, "loss": 1.0617, "step": 4064 }, { "epoch": 0.34, "learning_rate": 1.5253548264293192e-05, "loss": 1.1294, "step": 4065 }, { "epoch": 0.34, "learning_rate": 1.5251214068101954e-05, "loss": 1.13, "step": 4066 }, { "epoch": 0.34, "learning_rate": 1.5248879476798054e-05, "loss": 1.0694, "step": 4067 }, { "epoch": 0.34, "learning_rate": 1.5246544490557159e-05, "loss": 1.1424, "step": 4068 }, { "epoch": 0.34, "learning_rate": 1.5244209109554954e-05, "loss": 1.0823, "step": 4069 }, { "epoch": 0.34, "learning_rate": 1.5241873333967156e-05, "loss": 1.0757, "step": 4070 }, { "epoch": 0.34, "learning_rate": 1.5239537163969517e-05, "loss": 1.0535, "step": 4071 }, { "epoch": 0.34, "learning_rate": 1.5237200599737813e-05, "loss": 0.9211, "step": 4072 }, { "epoch": 0.34, "learning_rate": 1.5234863641447857e-05, "loss": 1.1002, "step": 4073 }, { "epoch": 0.35, "learning_rate": 1.5232526289275482e-05, "loss": 1.0399, "step": 4074 }, { "epoch": 0.35, "learning_rate": 1.5230188543396558e-05, "loss": 1.0998, "step": 4075 }, { "epoch": 0.35, "learning_rate": 1.5227850403986979e-05, "loss": 1.0804, "step": 4076 }, { "epoch": 0.35, "learning_rate": 1.5225511871222674e-05, "loss": 1.1383, "step": 4077 }, { "epoch": 0.35, "learning_rate": 1.5223172945279598e-05, "loss": 1.0633, "step": 4078 }, { "epoch": 0.35, "learning_rate": 1.522083362633374e-05, "loss": 1.0832, "step": 4079 }, { "epoch": 0.35, "learning_rate": 1.5218493914561112e-05, "loss": 1.0545, "step": 4080 }, { "epoch": 0.35, "learning_rate": 1.5216153810137759e-05, "loss": 1.0934, "step": 4081 }, { "epoch": 0.35, "learning_rate": 1.5213813313239754e-05, "loss": 1.0732, "step": 4082 }, { "epoch": 0.35, "learning_rate": 1.5211472424043208e-05, "loss": 1.0685, "step": 4083 }, { "epoch": 0.35, "learning_rate": 1.5209131142724247e-05, "loss": 1.133, "step": 4084 }, { "epoch": 0.35, "learning_rate": 1.5206789469459042e-05, "loss": 1.1022, "step": 4085 }, { "epoch": 0.35, "learning_rate": 1.5204447404423777e-05, "loss": 1.0887, "step": 4086 }, { "epoch": 0.35, "learning_rate": 1.5202104947794678e-05, "loss": 1.0936, "step": 4087 }, { "epoch": 0.35, "learning_rate": 1.5199762099747997e-05, "loss": 1.0927, "step": 4088 }, { "epoch": 0.35, "learning_rate": 1.519741886046001e-05, "loss": 1.0862, "step": 4089 }, { "epoch": 0.35, "learning_rate": 1.5195075230107037e-05, "loss": 1.0521, "step": 4090 }, { "epoch": 0.35, "learning_rate": 1.519273120886541e-05, "loss": 1.0449, "step": 4091 }, { "epoch": 0.35, "learning_rate": 1.51903867969115e-05, "loss": 1.0509, "step": 4092 }, { "epoch": 0.35, "learning_rate": 1.5188041994421708e-05, "loss": 1.0983, "step": 4093 }, { "epoch": 0.35, "learning_rate": 1.5185696801572455e-05, "loss": 1.0902, "step": 4094 }, { "epoch": 0.35, "learning_rate": 1.5183351218540208e-05, "loss": 1.0811, "step": 4095 }, { "epoch": 0.35, "learning_rate": 1.5181005245501447e-05, "loss": 1.0293, "step": 4096 }, { "epoch": 0.35, "learning_rate": 1.5178658882632693e-05, "loss": 1.0781, "step": 4097 }, { "epoch": 0.35, "learning_rate": 1.5176312130110487e-05, "loss": 1.0648, "step": 4098 }, { "epoch": 0.35, "learning_rate": 1.5173964988111406e-05, "loss": 1.1017, "step": 4099 }, { "epoch": 0.35, "learning_rate": 1.5171617456812055e-05, "loss": 1.1589, "step": 4100 }, { "epoch": 0.35, "learning_rate": 1.5169269536389064e-05, "loss": 1.0776, "step": 4101 }, { "epoch": 0.35, "learning_rate": 1.5166921227019095e-05, "loss": 1.1167, "step": 4102 }, { "epoch": 0.35, "learning_rate": 1.5164572528878845e-05, "loss": 1.1181, "step": 4103 }, { "epoch": 0.35, "learning_rate": 1.516222344214503e-05, "loss": 1.0548, "step": 4104 }, { "epoch": 0.35, "learning_rate": 1.5159873966994403e-05, "loss": 1.071, "step": 4105 }, { "epoch": 0.35, "learning_rate": 1.5157524103603743e-05, "loss": 1.0607, "step": 4106 }, { "epoch": 0.35, "learning_rate": 1.5155173852149858e-05, "loss": 1.0914, "step": 4107 }, { "epoch": 0.35, "learning_rate": 1.5152823212809589e-05, "loss": 1.08, "step": 4108 }, { "epoch": 0.35, "learning_rate": 1.5150472185759799e-05, "loss": 0.9345, "step": 4109 }, { "epoch": 0.35, "learning_rate": 1.5148120771177387e-05, "loss": 1.1352, "step": 4110 }, { "epoch": 0.35, "learning_rate": 1.514576896923928e-05, "loss": 1.1032, "step": 4111 }, { "epoch": 0.35, "learning_rate": 1.514341678012243e-05, "loss": 1.1304, "step": 4112 }, { "epoch": 0.35, "learning_rate": 1.5141064204003821e-05, "loss": 1.0803, "step": 4113 }, { "epoch": 0.35, "learning_rate": 1.5138711241060467e-05, "loss": 1.0771, "step": 4114 }, { "epoch": 0.35, "learning_rate": 1.5136357891469407e-05, "loss": 1.1094, "step": 4115 }, { "epoch": 0.35, "learning_rate": 1.5134004155407715e-05, "loss": 1.1247, "step": 4116 }, { "epoch": 0.35, "learning_rate": 1.5131650033052489e-05, "loss": 1.0832, "step": 4117 }, { "epoch": 0.35, "learning_rate": 1.5129295524580863e-05, "loss": 1.0772, "step": 4118 }, { "epoch": 0.35, "learning_rate": 1.512694063016999e-05, "loss": 1.1215, "step": 4119 }, { "epoch": 0.35, "learning_rate": 1.5124585349997061e-05, "loss": 1.0832, "step": 4120 }, { "epoch": 0.35, "learning_rate": 1.512222968423929e-05, "loss": 1.0534, "step": 4121 }, { "epoch": 0.35, "learning_rate": 1.5119873633073923e-05, "loss": 1.0399, "step": 4122 }, { "epoch": 0.35, "learning_rate": 1.5117517196678233e-05, "loss": 1.0522, "step": 4123 }, { "epoch": 0.35, "learning_rate": 1.5115160375229525e-05, "loss": 1.0804, "step": 4124 }, { "epoch": 0.35, "learning_rate": 1.5112803168905133e-05, "loss": 1.0904, "step": 4125 }, { "epoch": 0.35, "learning_rate": 1.5110445577882416e-05, "loss": 1.0961, "step": 4126 }, { "epoch": 0.35, "learning_rate": 1.5108087602338764e-05, "loss": 0.8686, "step": 4127 }, { "epoch": 0.35, "learning_rate": 1.5105729242451597e-05, "loss": 1.0783, "step": 4128 }, { "epoch": 0.35, "learning_rate": 1.5103370498398362e-05, "loss": 1.0751, "step": 4129 }, { "epoch": 0.35, "learning_rate": 1.5101011370356537e-05, "loss": 1.0988, "step": 4130 }, { "epoch": 0.35, "learning_rate": 1.5098651858503628e-05, "loss": 1.059, "step": 4131 }, { "epoch": 0.35, "learning_rate": 1.5096291963017169e-05, "loss": 1.0871, "step": 4132 }, { "epoch": 0.35, "learning_rate": 1.5093931684074726e-05, "loss": 1.0449, "step": 4133 }, { "epoch": 0.35, "learning_rate": 1.5091571021853884e-05, "loss": 1.0846, "step": 4134 }, { "epoch": 0.35, "learning_rate": 1.5089209976532275e-05, "loss": 1.1362, "step": 4135 }, { "epoch": 0.35, "learning_rate": 1.5086848548287543e-05, "loss": 1.1121, "step": 4136 }, { "epoch": 0.35, "learning_rate": 1.5084486737297365e-05, "loss": 1.0585, "step": 4137 }, { "epoch": 0.35, "learning_rate": 1.5082124543739456e-05, "loss": 1.0436, "step": 4138 }, { "epoch": 0.35, "learning_rate": 1.507976196779155e-05, "loss": 1.1408, "step": 4139 }, { "epoch": 0.35, "learning_rate": 1.5077399009631406e-05, "loss": 1.0982, "step": 4140 }, { "epoch": 0.35, "learning_rate": 1.5075035669436826e-05, "loss": 1.0957, "step": 4141 }, { "epoch": 0.35, "learning_rate": 1.5072671947385624e-05, "loss": 0.8864, "step": 4142 }, { "epoch": 0.35, "learning_rate": 1.5070307843655662e-05, "loss": 1.0619, "step": 4143 }, { "epoch": 0.35, "learning_rate": 1.5067943358424812e-05, "loss": 1.0045, "step": 4144 }, { "epoch": 0.35, "learning_rate": 1.5065578491870989e-05, "loss": 1.0604, "step": 4145 }, { "epoch": 0.35, "learning_rate": 1.5063213244172125e-05, "loss": 1.0936, "step": 4146 }, { "epoch": 0.35, "learning_rate": 1.5060847615506191e-05, "loss": 1.0482, "step": 4147 }, { "epoch": 0.35, "learning_rate": 1.505848160605118e-05, "loss": 1.0289, "step": 4148 }, { "epoch": 0.35, "learning_rate": 1.5056115215985117e-05, "loss": 1.0378, "step": 4149 }, { "epoch": 0.35, "learning_rate": 1.505374844548605e-05, "loss": 1.1314, "step": 4150 }, { "epoch": 0.35, "learning_rate": 1.5051381294732064e-05, "loss": 1.0882, "step": 4151 }, { "epoch": 0.35, "learning_rate": 1.5049013763901268e-05, "loss": 1.0854, "step": 4152 }, { "epoch": 0.35, "learning_rate": 1.5046645853171799e-05, "loss": 1.0606, "step": 4153 }, { "epoch": 0.35, "learning_rate": 1.5044277562721827e-05, "loss": 1.1294, "step": 4154 }, { "epoch": 0.35, "learning_rate": 1.5041908892729544e-05, "loss": 1.1181, "step": 4155 }, { "epoch": 0.35, "learning_rate": 1.5039539843373173e-05, "loss": 1.0808, "step": 4156 }, { "epoch": 0.35, "learning_rate": 1.5037170414830966e-05, "loss": 1.086, "step": 4157 }, { "epoch": 0.35, "learning_rate": 1.5034800607281209e-05, "loss": 1.1497, "step": 4158 }, { "epoch": 0.35, "learning_rate": 1.5032430420902207e-05, "loss": 1.0997, "step": 4159 }, { "epoch": 0.35, "learning_rate": 1.5030059855872297e-05, "loss": 1.0452, "step": 4160 }, { "epoch": 0.35, "learning_rate": 1.5027688912369849e-05, "loss": 1.0359, "step": 4161 }, { "epoch": 0.35, "learning_rate": 1.5025317590573255e-05, "loss": 1.1, "step": 4162 }, { "epoch": 0.35, "learning_rate": 1.5022945890660942e-05, "loss": 1.0379, "step": 4163 }, { "epoch": 0.35, "learning_rate": 1.5020573812811359e-05, "loss": 1.0773, "step": 4164 }, { "epoch": 0.35, "learning_rate": 1.5018201357202987e-05, "loss": 1.1274, "step": 4165 }, { "epoch": 0.35, "learning_rate": 1.5015828524014335e-05, "loss": 1.0992, "step": 4166 }, { "epoch": 0.35, "learning_rate": 1.5013455313423939e-05, "loss": 1.0588, "step": 4167 }, { "epoch": 0.35, "learning_rate": 1.5011081725610365e-05, "loss": 1.0393, "step": 4168 }, { "epoch": 0.35, "learning_rate": 1.5008707760752206e-05, "loss": 1.1093, "step": 4169 }, { "epoch": 0.35, "learning_rate": 1.5006333419028086e-05, "loss": 0.9901, "step": 4170 }, { "epoch": 0.35, "learning_rate": 1.5003958700616654e-05, "loss": 1.0752, "step": 4171 }, { "epoch": 0.35, "learning_rate": 1.500158360569659e-05, "loss": 1.0404, "step": 4172 }, { "epoch": 0.35, "learning_rate": 1.4999208134446601e-05, "loss": 1.1413, "step": 4173 }, { "epoch": 0.35, "learning_rate": 1.4996832287045421e-05, "loss": 1.0222, "step": 4174 }, { "epoch": 0.35, "learning_rate": 1.4994456063671819e-05, "loss": 1.006, "step": 4175 }, { "epoch": 0.35, "learning_rate": 1.499207946450458e-05, "loss": 0.9012, "step": 4176 }, { "epoch": 0.35, "learning_rate": 1.498970248972253e-05, "loss": 1.0476, "step": 4177 }, { "epoch": 0.35, "learning_rate": 1.4987325139504517e-05, "loss": 0.9864, "step": 4178 }, { "epoch": 0.35, "learning_rate": 1.4984947414029413e-05, "loss": 1.0697, "step": 4179 }, { "epoch": 0.35, "learning_rate": 1.498256931347613e-05, "loss": 1.0925, "step": 4180 }, { "epoch": 0.35, "learning_rate": 1.4980190838023599e-05, "loss": 1.1206, "step": 4181 }, { "epoch": 0.35, "learning_rate": 1.4977811987850776e-05, "loss": 1.0643, "step": 4182 }, { "epoch": 0.35, "learning_rate": 1.497543276313666e-05, "loss": 1.0792, "step": 4183 }, { "epoch": 0.35, "learning_rate": 1.4973053164060262e-05, "loss": 1.1076, "step": 4184 }, { "epoch": 0.35, "learning_rate": 1.4970673190800633e-05, "loss": 1.0989, "step": 4185 }, { "epoch": 0.35, "learning_rate": 1.4968292843536841e-05, "loss": 1.0816, "step": 4186 }, { "epoch": 0.35, "learning_rate": 1.4965912122447994e-05, "loss": 1.0641, "step": 4187 }, { "epoch": 0.35, "learning_rate": 1.4963531027713222e-05, "loss": 1.1266, "step": 4188 }, { "epoch": 0.35, "learning_rate": 1.4961149559511685e-05, "loss": 1.0352, "step": 4189 }, { "epoch": 0.35, "learning_rate": 1.4958767718022565e-05, "loss": 1.0937, "step": 4190 }, { "epoch": 0.35, "learning_rate": 1.4956385503425081e-05, "loss": 1.0633, "step": 4191 }, { "epoch": 0.36, "learning_rate": 1.4954002915898476e-05, "loss": 1.0673, "step": 4192 }, { "epoch": 0.36, "learning_rate": 1.4951619955622017e-05, "loss": 1.068, "step": 4193 }, { "epoch": 0.36, "learning_rate": 1.4949236622775006e-05, "loss": 0.9724, "step": 4194 }, { "epoch": 0.36, "learning_rate": 1.494685291753677e-05, "loss": 1.1095, "step": 4195 }, { "epoch": 0.36, "learning_rate": 1.4944468840086664e-05, "loss": 1.0964, "step": 4196 }, { "epoch": 0.36, "learning_rate": 1.494208439060407e-05, "loss": 1.0838, "step": 4197 }, { "epoch": 0.36, "learning_rate": 1.49396995692684e-05, "loss": 1.0775, "step": 4198 }, { "epoch": 0.36, "learning_rate": 1.4937314376259093e-05, "loss": 1.1444, "step": 4199 }, { "epoch": 0.36, "learning_rate": 1.4934928811755617e-05, "loss": 1.06, "step": 4200 }, { "epoch": 0.36, "learning_rate": 1.4932542875937465e-05, "loss": 1.0314, "step": 4201 }, { "epoch": 0.36, "learning_rate": 1.4930156568984161e-05, "loss": 1.0924, "step": 4202 }, { "epoch": 0.36, "learning_rate": 1.4927769891075259e-05, "loss": 1.0289, "step": 4203 }, { "epoch": 0.36, "learning_rate": 1.4925382842390333e-05, "loss": 1.1027, "step": 4204 }, { "epoch": 0.36, "learning_rate": 1.492299542310899e-05, "loss": 1.002, "step": 4205 }, { "epoch": 0.36, "learning_rate": 1.4920607633410869e-05, "loss": 1.0734, "step": 4206 }, { "epoch": 0.36, "learning_rate": 1.491821947347563e-05, "loss": 1.1116, "step": 4207 }, { "epoch": 0.36, "learning_rate": 1.4915830943482958e-05, "loss": 1.1222, "step": 4208 }, { "epoch": 0.36, "learning_rate": 1.491344204361258e-05, "loss": 1.0766, "step": 4209 }, { "epoch": 0.36, "learning_rate": 1.4911052774044237e-05, "loss": 1.0171, "step": 4210 }, { "epoch": 0.36, "learning_rate": 1.4908663134957704e-05, "loss": 1.0627, "step": 4211 }, { "epoch": 0.36, "learning_rate": 1.490627312653278e-05, "loss": 1.082, "step": 4212 }, { "epoch": 0.36, "learning_rate": 1.4903882748949297e-05, "loss": 1.0636, "step": 4213 }, { "epoch": 0.36, "learning_rate": 1.4901492002387113e-05, "loss": 1.0539, "step": 4214 }, { "epoch": 0.36, "learning_rate": 1.4899100887026112e-05, "loss": 1.1024, "step": 4215 }, { "epoch": 0.36, "learning_rate": 1.4896709403046205e-05, "loss": 0.9454, "step": 4216 }, { "epoch": 0.36, "learning_rate": 1.4894317550627336e-05, "loss": 1.0855, "step": 4217 }, { "epoch": 0.36, "learning_rate": 1.4891925329949471e-05, "loss": 0.9241, "step": 4218 }, { "epoch": 0.36, "learning_rate": 1.4889532741192607e-05, "loss": 1.0668, "step": 4219 }, { "epoch": 0.36, "learning_rate": 1.4887139784536764e-05, "loss": 1.0937, "step": 4220 }, { "epoch": 0.36, "learning_rate": 1.4884746460161995e-05, "loss": 1.0781, "step": 4221 }, { "epoch": 0.36, "learning_rate": 1.488235276824838e-05, "loss": 1.0309, "step": 4222 }, { "epoch": 0.36, "learning_rate": 1.4879958708976026e-05, "loss": 1.0978, "step": 4223 }, { "epoch": 0.36, "learning_rate": 1.4877564282525066e-05, "loss": 1.0615, "step": 4224 }, { "epoch": 0.36, "learning_rate": 1.4875169489075658e-05, "loss": 1.0731, "step": 4225 }, { "epoch": 0.36, "learning_rate": 1.4872774328808e-05, "loss": 1.0817, "step": 4226 }, { "epoch": 0.36, "learning_rate": 1.4870378801902304e-05, "loss": 1.0937, "step": 4227 }, { "epoch": 0.36, "learning_rate": 1.4867982908538815e-05, "loss": 1.101, "step": 4228 }, { "epoch": 0.36, "learning_rate": 1.4865586648897803e-05, "loss": 1.0614, "step": 4229 }, { "epoch": 0.36, "learning_rate": 1.4863190023159571e-05, "loss": 1.0817, "step": 4230 }, { "epoch": 0.36, "learning_rate": 1.4860793031504443e-05, "loss": 1.0892, "step": 4231 }, { "epoch": 0.36, "learning_rate": 1.485839567411278e-05, "loss": 1.0251, "step": 4232 }, { "epoch": 0.36, "learning_rate": 1.4855997951164958e-05, "loss": 1.0574, "step": 4233 }, { "epoch": 0.36, "learning_rate": 1.4853599862841393e-05, "loss": 1.0607, "step": 4234 }, { "epoch": 0.36, "learning_rate": 1.4851201409322513e-05, "loss": 1.0885, "step": 4235 }, { "epoch": 0.36, "learning_rate": 1.484880259078879e-05, "loss": 1.0359, "step": 4236 }, { "epoch": 0.36, "learning_rate": 1.4846403407420717e-05, "loss": 1.0855, "step": 4237 }, { "epoch": 0.36, "learning_rate": 1.4844003859398805e-05, "loss": 1.0854, "step": 4238 }, { "epoch": 0.36, "learning_rate": 1.4841603946903614e-05, "loss": 1.0184, "step": 4239 }, { "epoch": 0.36, "learning_rate": 1.483920367011571e-05, "loss": 1.0591, "step": 4240 }, { "epoch": 0.36, "learning_rate": 1.4836803029215696e-05, "loss": 1.0843, "step": 4241 }, { "epoch": 0.36, "learning_rate": 1.4834402024384205e-05, "loss": 1.0589, "step": 4242 }, { "epoch": 0.36, "learning_rate": 1.483200065580189e-05, "loss": 1.0834, "step": 4243 }, { "epoch": 0.36, "learning_rate": 1.4829598923649436e-05, "loss": 1.0856, "step": 4244 }, { "epoch": 0.36, "learning_rate": 1.4827196828107555e-05, "loss": 1.0641, "step": 4245 }, { "epoch": 0.36, "learning_rate": 1.4824794369356988e-05, "loss": 0.9814, "step": 4246 }, { "epoch": 0.36, "learning_rate": 1.4822391547578497e-05, "loss": 0.9115, "step": 4247 }, { "epoch": 0.36, "learning_rate": 1.4819988362952875e-05, "loss": 1.1268, "step": 4248 }, { "epoch": 0.36, "learning_rate": 1.4817584815660947e-05, "loss": 1.0649, "step": 4249 }, { "epoch": 0.36, "learning_rate": 1.481518090588356e-05, "loss": 1.1092, "step": 4250 }, { "epoch": 0.36, "learning_rate": 1.4812776633801589e-05, "loss": 1.0616, "step": 4251 }, { "epoch": 0.36, "learning_rate": 1.4810371999595935e-05, "loss": 1.0552, "step": 4252 }, { "epoch": 0.36, "learning_rate": 1.4807967003447528e-05, "loss": 1.0248, "step": 4253 }, { "epoch": 0.36, "learning_rate": 1.4805561645537327e-05, "loss": 1.0781, "step": 4254 }, { "epoch": 0.36, "learning_rate": 1.4803155926046318e-05, "loss": 1.0734, "step": 4255 }, { "epoch": 0.36, "learning_rate": 1.4800749845155507e-05, "loss": 1.0385, "step": 4256 }, { "epoch": 0.36, "learning_rate": 1.4798343403045937e-05, "loss": 1.0397, "step": 4257 }, { "epoch": 0.36, "learning_rate": 1.4795936599898673e-05, "loss": 1.079, "step": 4258 }, { "epoch": 0.36, "learning_rate": 1.4793529435894804e-05, "loss": 1.0471, "step": 4259 }, { "epoch": 0.36, "learning_rate": 1.4791121911215458e-05, "loss": 1.0893, "step": 4260 }, { "epoch": 0.36, "learning_rate": 1.478871402604178e-05, "loss": 1.0959, "step": 4261 }, { "epoch": 0.36, "learning_rate": 1.4786305780554937e-05, "loss": 1.1026, "step": 4262 }, { "epoch": 0.36, "learning_rate": 1.478389717493614e-05, "loss": 1.065, "step": 4263 }, { "epoch": 0.36, "learning_rate": 1.478148820936661e-05, "loss": 1.0257, "step": 4264 }, { "epoch": 0.36, "learning_rate": 1.477907888402761e-05, "loss": 1.0432, "step": 4265 }, { "epoch": 0.36, "learning_rate": 1.4776669199100418e-05, "loss": 1.0618, "step": 4266 }, { "epoch": 0.36, "learning_rate": 1.4774259154766346e-05, "loss": 1.1049, "step": 4267 }, { "epoch": 0.36, "learning_rate": 1.477184875120673e-05, "loss": 1.0388, "step": 4268 }, { "epoch": 0.36, "learning_rate": 1.4769437988602935e-05, "loss": 1.1355, "step": 4269 }, { "epoch": 0.36, "learning_rate": 1.4767026867136352e-05, "loss": 1.0836, "step": 4270 }, { "epoch": 0.36, "learning_rate": 1.4764615386988397e-05, "loss": 0.9495, "step": 4271 }, { "epoch": 0.36, "learning_rate": 1.4762203548340514e-05, "loss": 1.0713, "step": 4272 }, { "epoch": 0.36, "learning_rate": 1.4759791351374182e-05, "loss": 1.1025, "step": 4273 }, { "epoch": 0.36, "learning_rate": 1.4757378796270893e-05, "loss": 1.0912, "step": 4274 }, { "epoch": 0.36, "learning_rate": 1.4754965883212173e-05, "loss": 1.088, "step": 4275 }, { "epoch": 0.36, "learning_rate": 1.4752552612379578e-05, "loss": 1.0699, "step": 4276 }, { "epoch": 0.36, "learning_rate": 1.4750138983954687e-05, "loss": 1.1054, "step": 4277 }, { "epoch": 0.36, "learning_rate": 1.4747724998119104e-05, "loss": 1.0671, "step": 4278 }, { "epoch": 0.36, "learning_rate": 1.4745310655054468e-05, "loss": 1.1027, "step": 4279 }, { "epoch": 0.36, "learning_rate": 1.4742895954942431e-05, "loss": 1.0951, "step": 4280 }, { "epoch": 0.36, "learning_rate": 1.4740480897964689e-05, "loss": 1.1212, "step": 4281 }, { "epoch": 0.36, "learning_rate": 1.473806548430295e-05, "loss": 1.073, "step": 4282 }, { "epoch": 0.36, "learning_rate": 1.4735649714138957e-05, "loss": 1.1089, "step": 4283 }, { "epoch": 0.36, "learning_rate": 1.473323358765448e-05, "loss": 1.0634, "step": 4284 }, { "epoch": 0.36, "learning_rate": 1.4730817105031307e-05, "loss": 1.1089, "step": 4285 }, { "epoch": 0.36, "learning_rate": 1.4728400266451269e-05, "loss": 1.0894, "step": 4286 }, { "epoch": 0.36, "learning_rate": 1.4725983072096203e-05, "loss": 1.0873, "step": 4287 }, { "epoch": 0.36, "learning_rate": 1.4723565522147995e-05, "loss": 1.1187, "step": 4288 }, { "epoch": 0.36, "learning_rate": 1.472114761678854e-05, "loss": 1.1703, "step": 4289 }, { "epoch": 0.36, "learning_rate": 1.4718729356199766e-05, "loss": 1.1296, "step": 4290 }, { "epoch": 0.36, "learning_rate": 1.471631074056363e-05, "loss": 1.0682, "step": 4291 }, { "epoch": 0.36, "learning_rate": 1.4713891770062115e-05, "loss": 1.0595, "step": 4292 }, { "epoch": 0.36, "learning_rate": 1.4711472444877228e-05, "loss": 1.0457, "step": 4293 }, { "epoch": 0.36, "learning_rate": 1.4709052765191006e-05, "loss": 1.0838, "step": 4294 }, { "epoch": 0.36, "learning_rate": 1.4706632731185508e-05, "loss": 1.0712, "step": 4295 }, { "epoch": 0.36, "learning_rate": 1.4704212343042824e-05, "loss": 1.0944, "step": 4296 }, { "epoch": 0.36, "learning_rate": 1.470179160094507e-05, "loss": 1.1146, "step": 4297 }, { "epoch": 0.36, "learning_rate": 1.4699370505074386e-05, "loss": 1.0688, "step": 4298 }, { "epoch": 0.36, "learning_rate": 1.4696949055612943e-05, "loss": 1.0599, "step": 4299 }, { "epoch": 0.36, "learning_rate": 1.4694527252742938e-05, "loss": 1.078, "step": 4300 }, { "epoch": 0.36, "learning_rate": 1.4692105096646584e-05, "loss": 1.1291, "step": 4301 }, { "epoch": 0.36, "learning_rate": 1.4689682587506138e-05, "loss": 1.0922, "step": 4302 }, { "epoch": 0.36, "learning_rate": 1.468725972550387e-05, "loss": 1.0605, "step": 4303 }, { "epoch": 0.36, "learning_rate": 1.4684836510822083e-05, "loss": 1.0768, "step": 4304 }, { "epoch": 0.36, "learning_rate": 1.4682412943643102e-05, "loss": 1.0823, "step": 4305 }, { "epoch": 0.36, "learning_rate": 1.4679989024149289e-05, "loss": 1.0698, "step": 4306 }, { "epoch": 0.36, "learning_rate": 1.4677564752523017e-05, "loss": 1.1001, "step": 4307 }, { "epoch": 0.36, "learning_rate": 1.4675140128946694e-05, "loss": 1.0952, "step": 4308 }, { "epoch": 0.36, "learning_rate": 1.4672715153602761e-05, "loss": 1.0956, "step": 4309 }, { "epoch": 0.37, "learning_rate": 1.467028982667367e-05, "loss": 1.0691, "step": 4310 }, { "epoch": 0.37, "learning_rate": 1.4667864148341916e-05, "loss": 1.0126, "step": 4311 }, { "epoch": 0.37, "learning_rate": 1.4665438118790004e-05, "loss": 1.061, "step": 4312 }, { "epoch": 0.37, "learning_rate": 1.466301173820048e-05, "loss": 1.0122, "step": 4313 }, { "epoch": 0.37, "learning_rate": 1.4660585006755905e-05, "loss": 1.0421, "step": 4314 }, { "epoch": 0.37, "learning_rate": 1.4658157924638875e-05, "loss": 1.0814, "step": 4315 }, { "epoch": 0.37, "learning_rate": 1.4655730492032008e-05, "loss": 1.0713, "step": 4316 }, { "epoch": 0.37, "learning_rate": 1.465330270911795e-05, "loss": 1.0904, "step": 4317 }, { "epoch": 0.37, "learning_rate": 1.465087457607937e-05, "loss": 1.0279, "step": 4318 }, { "epoch": 0.37, "learning_rate": 1.464844609309897e-05, "loss": 1.1038, "step": 4319 }, { "epoch": 0.37, "learning_rate": 1.4646017260359471e-05, "loss": 1.0861, "step": 4320 }, { "epoch": 0.37, "learning_rate": 1.4643588078043624e-05, "loss": 1.1282, "step": 4321 }, { "epoch": 0.37, "learning_rate": 1.464115854633421e-05, "loss": 1.0856, "step": 4322 }, { "epoch": 0.37, "learning_rate": 1.4638728665414026e-05, "loss": 1.0983, "step": 4323 }, { "epoch": 0.37, "learning_rate": 1.4636298435465903e-05, "loss": 1.0693, "step": 4324 }, { "epoch": 0.37, "learning_rate": 1.46338678566727e-05, "loss": 1.0847, "step": 4325 }, { "epoch": 0.37, "learning_rate": 1.46314369292173e-05, "loss": 1.0089, "step": 4326 }, { "epoch": 0.37, "learning_rate": 1.4629005653282602e-05, "loss": 1.1415, "step": 4327 }, { "epoch": 0.37, "learning_rate": 1.462657402905155e-05, "loss": 1.1, "step": 4328 }, { "epoch": 0.37, "learning_rate": 1.4624142056707102e-05, "loss": 1.0762, "step": 4329 }, { "epoch": 0.37, "learning_rate": 1.4621709736432244e-05, "loss": 1.0908, "step": 4330 }, { "epoch": 0.37, "learning_rate": 1.4619277068409987e-05, "loss": 1.0925, "step": 4331 }, { "epoch": 0.37, "learning_rate": 1.4616844052823374e-05, "loss": 1.0587, "step": 4332 }, { "epoch": 0.37, "learning_rate": 1.4614410689855466e-05, "loss": 1.0956, "step": 4333 }, { "epoch": 0.37, "learning_rate": 1.461197697968936e-05, "loss": 1.1047, "step": 4334 }, { "epoch": 0.37, "learning_rate": 1.4609542922508172e-05, "loss": 1.0822, "step": 4335 }, { "epoch": 0.37, "learning_rate": 1.4607108518495042e-05, "loss": 1.0723, "step": 4336 }, { "epoch": 0.37, "learning_rate": 1.4604673767833141e-05, "loss": 1.0478, "step": 4337 }, { "epoch": 0.37, "learning_rate": 1.4602238670705668e-05, "loss": 1.0962, "step": 4338 }, { "epoch": 0.37, "learning_rate": 1.4599803227295844e-05, "loss": 1.1192, "step": 4339 }, { "epoch": 0.37, "learning_rate": 1.4597367437786915e-05, "loss": 1.0669, "step": 4340 }, { "epoch": 0.37, "learning_rate": 1.459493130236216e-05, "loss": 1.0891, "step": 4341 }, { "epoch": 0.37, "learning_rate": 1.4592494821204868e-05, "loss": 1.0591, "step": 4342 }, { "epoch": 0.37, "learning_rate": 1.4590057994498375e-05, "loss": 1.0579, "step": 4343 }, { "epoch": 0.37, "learning_rate": 1.4587620822426032e-05, "loss": 1.0784, "step": 4344 }, { "epoch": 0.37, "learning_rate": 1.4585183305171216e-05, "loss": 0.9881, "step": 4345 }, { "epoch": 0.37, "learning_rate": 1.4582745442917328e-05, "loss": 1.0801, "step": 4346 }, { "epoch": 0.37, "learning_rate": 1.4580307235847802e-05, "loss": 1.0578, "step": 4347 }, { "epoch": 0.37, "learning_rate": 1.4577868684146093e-05, "loss": 1.1319, "step": 4348 }, { "epoch": 0.37, "learning_rate": 1.4575429787995681e-05, "loss": 1.0428, "step": 4349 }, { "epoch": 0.37, "learning_rate": 1.4572990547580075e-05, "loss": 1.0702, "step": 4350 }, { "epoch": 0.37, "learning_rate": 1.457055096308281e-05, "loss": 1.0854, "step": 4351 }, { "epoch": 0.37, "learning_rate": 1.4568111034687445e-05, "loss": 1.0558, "step": 4352 }, { "epoch": 0.37, "learning_rate": 1.4565670762577567e-05, "loss": 1.071, "step": 4353 }, { "epoch": 0.37, "learning_rate": 1.4563230146936782e-05, "loss": 1.0164, "step": 4354 }, { "epoch": 0.37, "learning_rate": 1.456078918794873e-05, "loss": 1.113, "step": 4355 }, { "epoch": 0.37, "learning_rate": 1.4558347885797077e-05, "loss": 1.1129, "step": 4356 }, { "epoch": 0.37, "learning_rate": 1.4555906240665507e-05, "loss": 1.1261, "step": 4357 }, { "epoch": 0.37, "learning_rate": 1.4553464252737737e-05, "loss": 1.0353, "step": 4358 }, { "epoch": 0.37, "learning_rate": 1.4551021922197508e-05, "loss": 1.0806, "step": 4359 }, { "epoch": 0.37, "learning_rate": 1.4548579249228587e-05, "loss": 1.079, "step": 4360 }, { "epoch": 0.37, "learning_rate": 1.4546136234014762e-05, "loss": 1.1037, "step": 4361 }, { "epoch": 0.37, "learning_rate": 1.4543692876739854e-05, "loss": 1.1095, "step": 4362 }, { "epoch": 0.37, "learning_rate": 1.4541249177587706e-05, "loss": 1.0961, "step": 4363 }, { "epoch": 0.37, "learning_rate": 1.453880513674219e-05, "loss": 1.1058, "step": 4364 }, { "epoch": 0.37, "learning_rate": 1.4536360754387193e-05, "loss": 1.0119, "step": 4365 }, { "epoch": 0.37, "learning_rate": 1.4533916030706642e-05, "loss": 1.0818, "step": 4366 }, { "epoch": 0.37, "learning_rate": 1.4531470965884485e-05, "loss": 1.0332, "step": 4367 }, { "epoch": 0.37, "learning_rate": 1.4529025560104686e-05, "loss": 1.0546, "step": 4368 }, { "epoch": 0.37, "learning_rate": 1.452657981355125e-05, "loss": 1.0486, "step": 4369 }, { "epoch": 0.37, "learning_rate": 1.4524133726408195e-05, "loss": 1.0495, "step": 4370 }, { "epoch": 0.37, "learning_rate": 1.4521687298859575e-05, "loss": 1.1112, "step": 4371 }, { "epoch": 0.37, "learning_rate": 1.451924053108946e-05, "loss": 1.1029, "step": 4372 }, { "epoch": 0.37, "learning_rate": 1.4516793423281954e-05, "loss": 1.0968, "step": 4373 }, { "epoch": 0.37, "learning_rate": 1.451434597562118e-05, "loss": 1.0089, "step": 4374 }, { "epoch": 0.37, "learning_rate": 1.451189818829129e-05, "loss": 1.1286, "step": 4375 }, { "epoch": 0.37, "learning_rate": 1.4509450061476457e-05, "loss": 1.0857, "step": 4376 }, { "epoch": 0.37, "learning_rate": 1.450700159536089e-05, "loss": 1.0626, "step": 4377 }, { "epoch": 0.37, "learning_rate": 1.4504552790128813e-05, "loss": 1.0825, "step": 4378 }, { "epoch": 0.37, "learning_rate": 1.4502103645964483e-05, "loss": 1.0486, "step": 4379 }, { "epoch": 0.37, "learning_rate": 1.4499654163052174e-05, "loss": 1.1663, "step": 4380 }, { "epoch": 0.37, "learning_rate": 1.4497204341576192e-05, "loss": 1.067, "step": 4381 }, { "epoch": 0.37, "learning_rate": 1.449475418172087e-05, "loss": 1.0762, "step": 4382 }, { "epoch": 0.37, "learning_rate": 1.4492303683670556e-05, "loss": 1.1485, "step": 4383 }, { "epoch": 0.37, "learning_rate": 1.4489852847609638e-05, "loss": 1.1368, "step": 4384 }, { "epoch": 0.37, "learning_rate": 1.4487401673722519e-05, "loss": 1.0912, "step": 4385 }, { "epoch": 0.37, "learning_rate": 1.4484950162193632e-05, "loss": 1.0712, "step": 4386 }, { "epoch": 0.37, "learning_rate": 1.448249831320743e-05, "loss": 1.0929, "step": 4387 }, { "epoch": 0.37, "learning_rate": 1.4480046126948402e-05, "loss": 1.0886, "step": 4388 }, { "epoch": 0.37, "learning_rate": 1.447759360360105e-05, "loss": 1.0107, "step": 4389 }, { "epoch": 0.37, "learning_rate": 1.4475140743349912e-05, "loss": 0.9967, "step": 4390 }, { "epoch": 0.37, "learning_rate": 1.4472687546379542e-05, "loss": 1.0354, "step": 4391 }, { "epoch": 0.37, "learning_rate": 1.4470234012874527e-05, "loss": 1.0867, "step": 4392 }, { "epoch": 0.37, "learning_rate": 1.4467780143019474e-05, "loss": 1.0719, "step": 4393 }, { "epoch": 0.37, "learning_rate": 1.4465325936999017e-05, "loss": 1.0988, "step": 4394 }, { "epoch": 0.37, "learning_rate": 1.4462871394997819e-05, "loss": 1.0536, "step": 4395 }, { "epoch": 0.37, "learning_rate": 1.4460416517200564e-05, "loss": 1.0469, "step": 4396 }, { "epoch": 0.37, "learning_rate": 1.445796130379196e-05, "loss": 1.0977, "step": 4397 }, { "epoch": 0.37, "learning_rate": 1.4455505754956745e-05, "loss": 1.0765, "step": 4398 }, { "epoch": 0.37, "learning_rate": 1.4453049870879679e-05, "loss": 1.1009, "step": 4399 }, { "epoch": 0.37, "learning_rate": 1.4450593651745547e-05, "loss": 1.0164, "step": 4400 }, { "epoch": 0.37, "learning_rate": 1.4448137097739162e-05, "loss": 1.1181, "step": 4401 }, { "epoch": 0.37, "learning_rate": 1.4445680209045356e-05, "loss": 1.06, "step": 4402 }, { "epoch": 0.37, "learning_rate": 1.4443222985848997e-05, "loss": 1.0905, "step": 4403 }, { "epoch": 0.37, "learning_rate": 1.4440765428334968e-05, "loss": 1.1495, "step": 4404 }, { "epoch": 0.37, "learning_rate": 1.4438307536688181e-05, "loss": 1.0532, "step": 4405 }, { "epoch": 0.37, "learning_rate": 1.4435849311093576e-05, "loss": 1.0502, "step": 4406 }, { "epoch": 0.37, "learning_rate": 1.4433390751736114e-05, "loss": 1.0278, "step": 4407 }, { "epoch": 0.37, "learning_rate": 1.4430931858800779e-05, "loss": 1.0727, "step": 4408 }, { "epoch": 0.37, "learning_rate": 1.4428472632472582e-05, "loss": 1.0473, "step": 4409 }, { "epoch": 0.37, "learning_rate": 1.4426013072936568e-05, "loss": 1.0413, "step": 4410 }, { "epoch": 0.37, "learning_rate": 1.4423553180377796e-05, "loss": 1.0803, "step": 4411 }, { "epoch": 0.37, "learning_rate": 1.4421092954981353e-05, "loss": 1.0747, "step": 4412 }, { "epoch": 0.37, "learning_rate": 1.4418632396932353e-05, "loss": 1.0899, "step": 4413 }, { "epoch": 0.37, "learning_rate": 1.441617150641593e-05, "loss": 1.0569, "step": 4414 }, { "epoch": 0.37, "learning_rate": 1.441371028361725e-05, "loss": 1.0897, "step": 4415 }, { "epoch": 0.37, "learning_rate": 1.4411248728721501e-05, "loss": 1.129, "step": 4416 }, { "epoch": 0.37, "learning_rate": 1.4408786841913895e-05, "loss": 1.1191, "step": 4417 }, { "epoch": 0.37, "learning_rate": 1.4406324623379671e-05, "loss": 1.1149, "step": 4418 }, { "epoch": 0.37, "learning_rate": 1.4403862073304087e-05, "loss": 1.0603, "step": 4419 }, { "epoch": 0.37, "learning_rate": 1.4401399191872434e-05, "loss": 1.0793, "step": 4420 }, { "epoch": 0.37, "learning_rate": 1.4398935979270023e-05, "loss": 0.9565, "step": 4421 }, { "epoch": 0.37, "learning_rate": 1.4396472435682196e-05, "loss": 1.0654, "step": 4422 }, { "epoch": 0.37, "learning_rate": 1.439400856129431e-05, "loss": 1.1078, "step": 4423 }, { "epoch": 0.37, "learning_rate": 1.4391544356291753e-05, "loss": 1.0625, "step": 4424 }, { "epoch": 0.37, "learning_rate": 1.4389079820859941e-05, "loss": 1.102, "step": 4425 }, { "epoch": 0.37, "learning_rate": 1.4386614955184307e-05, "loss": 1.0544, "step": 4426 }, { "epoch": 0.37, "learning_rate": 1.4384149759450314e-05, "loss": 1.0478, "step": 4427 }, { "epoch": 0.38, "learning_rate": 1.4381684233843446e-05, "loss": 1.1259, "step": 4428 }, { "epoch": 0.38, "learning_rate": 1.4379218378549223e-05, "loss": 1.0863, "step": 4429 }, { "epoch": 0.38, "learning_rate": 1.4376752193753169e-05, "loss": 1.0643, "step": 4430 }, { "epoch": 0.38, "learning_rate": 1.4374285679640854e-05, "loss": 1.0719, "step": 4431 }, { "epoch": 0.38, "learning_rate": 1.4371818836397862e-05, "loss": 1.0696, "step": 4432 }, { "epoch": 0.38, "learning_rate": 1.4369351664209802e-05, "loss": 1.0999, "step": 4433 }, { "epoch": 0.38, "learning_rate": 1.4366884163262312e-05, "loss": 1.1095, "step": 4434 }, { "epoch": 0.38, "learning_rate": 1.4364416333741049e-05, "loss": 1.072, "step": 4435 }, { "epoch": 0.38, "learning_rate": 1.4361948175831697e-05, "loss": 1.0757, "step": 4436 }, { "epoch": 0.38, "learning_rate": 1.435947968971997e-05, "loss": 1.0936, "step": 4437 }, { "epoch": 0.38, "learning_rate": 1.4357010875591599e-05, "loss": 1.0505, "step": 4438 }, { "epoch": 0.38, "learning_rate": 1.4354541733632344e-05, "loss": 1.0797, "step": 4439 }, { "epoch": 0.38, "learning_rate": 1.4352072264027987e-05, "loss": 1.0855, "step": 4440 }, { "epoch": 0.38, "learning_rate": 1.4349602466964339e-05, "loss": 1.0516, "step": 4441 }, { "epoch": 0.38, "learning_rate": 1.434713234262723e-05, "loss": 1.0909, "step": 4442 }, { "epoch": 0.38, "learning_rate": 1.434466189120252e-05, "loss": 1.0878, "step": 4443 }, { "epoch": 0.38, "learning_rate": 1.4342191112876087e-05, "loss": 1.1004, "step": 4444 }, { "epoch": 0.38, "learning_rate": 1.4339720007833844e-05, "loss": 0.9209, "step": 4445 }, { "epoch": 0.38, "learning_rate": 1.4337248576261717e-05, "loss": 1.0767, "step": 4446 }, { "epoch": 0.38, "learning_rate": 1.4334776818345666e-05, "loss": 1.0839, "step": 4447 }, { "epoch": 0.38, "learning_rate": 1.4332304734271665e-05, "loss": 1.1015, "step": 4448 }, { "epoch": 0.38, "learning_rate": 1.4329832324225725e-05, "loss": 1.0503, "step": 4449 }, { "epoch": 0.38, "learning_rate": 1.4327359588393873e-05, "loss": 1.0037, "step": 4450 }, { "epoch": 0.38, "learning_rate": 1.4324886526962165e-05, "loss": 1.0518, "step": 4451 }, { "epoch": 0.38, "learning_rate": 1.4322413140116677e-05, "loss": 1.106, "step": 4452 }, { "epoch": 0.38, "learning_rate": 1.4319939428043514e-05, "loss": 1.0711, "step": 4453 }, { "epoch": 0.38, "learning_rate": 1.4317465390928803e-05, "loss": 1.1012, "step": 4454 }, { "epoch": 0.38, "learning_rate": 1.431499102895869e-05, "loss": 1.0951, "step": 4455 }, { "epoch": 0.38, "learning_rate": 1.4312516342319361e-05, "loss": 1.0055, "step": 4456 }, { "epoch": 0.38, "learning_rate": 1.4310041331197012e-05, "loss": 1.0798, "step": 4457 }, { "epoch": 0.38, "learning_rate": 1.4307565995777866e-05, "loss": 1.0827, "step": 4458 }, { "epoch": 0.38, "learning_rate": 1.4305090336248176e-05, "loss": 1.0913, "step": 4459 }, { "epoch": 0.38, "learning_rate": 1.4302614352794218e-05, "loss": 1.0988, "step": 4460 }, { "epoch": 0.38, "learning_rate": 1.4300138045602283e-05, "loss": 1.1023, "step": 4461 }, { "epoch": 0.38, "learning_rate": 1.42976614148587e-05, "loss": 1.1261, "step": 4462 }, { "epoch": 0.38, "learning_rate": 1.4295184460749814e-05, "loss": 1.1101, "step": 4463 }, { "epoch": 0.38, "learning_rate": 1.4292707183461994e-05, "loss": 1.1361, "step": 4464 }, { "epoch": 0.38, "learning_rate": 1.4290229583181639e-05, "loss": 1.0266, "step": 4465 }, { "epoch": 0.38, "learning_rate": 1.4287751660095168e-05, "loss": 1.0797, "step": 4466 }, { "epoch": 0.38, "learning_rate": 1.4285273414389028e-05, "loss": 1.0905, "step": 4467 }, { "epoch": 0.38, "learning_rate": 1.428279484624968e-05, "loss": 1.0391, "step": 4468 }, { "epoch": 0.38, "learning_rate": 1.4280315955863627e-05, "loss": 1.0603, "step": 4469 }, { "epoch": 0.38, "learning_rate": 1.4277836743417376e-05, "loss": 1.0815, "step": 4470 }, { "epoch": 0.38, "learning_rate": 1.4275357209097476e-05, "loss": 1.0726, "step": 4471 }, { "epoch": 0.38, "learning_rate": 1.4272877353090486e-05, "loss": 1.0813, "step": 4472 }, { "epoch": 0.38, "learning_rate": 1.4270397175583001e-05, "loss": 1.0246, "step": 4473 }, { "epoch": 0.38, "learning_rate": 1.4267916676761635e-05, "loss": 1.1226, "step": 4474 }, { "epoch": 0.38, "learning_rate": 1.4265435856813023e-05, "loss": 1.0764, "step": 4475 }, { "epoch": 0.38, "learning_rate": 1.4262954715923827e-05, "loss": 1.083, "step": 4476 }, { "epoch": 0.38, "learning_rate": 1.4260473254280735e-05, "loss": 1.0865, "step": 4477 }, { "epoch": 0.38, "learning_rate": 1.4257991472070459e-05, "loss": 1.0786, "step": 4478 }, { "epoch": 0.38, "learning_rate": 1.4255509369479731e-05, "loss": 1.0449, "step": 4479 }, { "epoch": 0.38, "learning_rate": 1.4253026946695311e-05, "loss": 1.0547, "step": 4480 }, { "epoch": 0.38, "learning_rate": 1.425054420390398e-05, "loss": 1.0802, "step": 4481 }, { "epoch": 0.38, "learning_rate": 1.4248061141292548e-05, "loss": 1.0902, "step": 4482 }, { "epoch": 0.38, "learning_rate": 1.4245577759047844e-05, "loss": 1.12, "step": 4483 }, { "epoch": 0.38, "learning_rate": 1.4243094057356724e-05, "loss": 1.0886, "step": 4484 }, { "epoch": 0.38, "learning_rate": 1.4240610036406067e-05, "loss": 1.0894, "step": 4485 }, { "epoch": 0.38, "learning_rate": 1.4238125696382773e-05, "loss": 1.1013, "step": 4486 }, { "epoch": 0.38, "learning_rate": 1.4235641037473775e-05, "loss": 1.1318, "step": 4487 }, { "epoch": 0.38, "learning_rate": 1.4233156059866018e-05, "loss": 1.0576, "step": 4488 }, { "epoch": 0.38, "learning_rate": 1.4230670763746482e-05, "loss": 1.056, "step": 4489 }, { "epoch": 0.38, "learning_rate": 1.4228185149302164e-05, "loss": 1.0739, "step": 4490 }, { "epoch": 0.38, "learning_rate": 1.4225699216720084e-05, "loss": 1.0485, "step": 4491 }, { "epoch": 0.38, "learning_rate": 1.4223212966187292e-05, "loss": 1.0565, "step": 4492 }, { "epoch": 0.38, "learning_rate": 1.4220726397890858e-05, "loss": 1.0623, "step": 4493 }, { "epoch": 0.38, "learning_rate": 1.421823951201788e-05, "loss": 1.1203, "step": 4494 }, { "epoch": 0.38, "learning_rate": 1.4215752308755475e-05, "loss": 1.1003, "step": 4495 }, { "epoch": 0.38, "learning_rate": 1.4213264788290782e-05, "loss": 1.0686, "step": 4496 }, { "epoch": 0.38, "learning_rate": 1.4210776950810969e-05, "loss": 1.0766, "step": 4497 }, { "epoch": 0.38, "learning_rate": 1.4208288796503227e-05, "loss": 1.1075, "step": 4498 }, { "epoch": 0.38, "learning_rate": 1.4205800325554774e-05, "loss": 1.0884, "step": 4499 }, { "epoch": 0.38, "learning_rate": 1.4203311538152842e-05, "loss": 1.1062, "step": 4500 }, { "epoch": 0.38, "learning_rate": 1.4200822434484696e-05, "loss": 1.0962, "step": 4501 }, { "epoch": 0.38, "learning_rate": 1.4198333014737619e-05, "loss": 1.0282, "step": 4502 }, { "epoch": 0.38, "learning_rate": 1.4195843279098925e-05, "loss": 1.0107, "step": 4503 }, { "epoch": 0.38, "learning_rate": 1.419335322775594e-05, "loss": 1.0726, "step": 4504 }, { "epoch": 0.38, "learning_rate": 1.4190862860896027e-05, "loss": 1.0434, "step": 4505 }, { "epoch": 0.38, "learning_rate": 1.4188372178706562e-05, "loss": 0.8764, "step": 4506 }, { "epoch": 0.38, "learning_rate": 1.4185881181374952e-05, "loss": 1.0743, "step": 4507 }, { "epoch": 0.38, "learning_rate": 1.4183389869088627e-05, "loss": 1.1087, "step": 4508 }, { "epoch": 0.38, "learning_rate": 1.4180898242035033e-05, "loss": 1.1171, "step": 4509 }, { "epoch": 0.38, "learning_rate": 1.417840630040165e-05, "loss": 1.0684, "step": 4510 }, { "epoch": 0.38, "learning_rate": 1.4175914044375976e-05, "loss": 1.119, "step": 4511 }, { "epoch": 0.38, "learning_rate": 1.4173421474145532e-05, "loss": 1.0611, "step": 4512 }, { "epoch": 0.38, "learning_rate": 1.4170928589897868e-05, "loss": 1.1175, "step": 4513 }, { "epoch": 0.38, "learning_rate": 1.4168435391820552e-05, "loss": 1.094, "step": 4514 }, { "epoch": 0.38, "learning_rate": 1.4165941880101178e-05, "loss": 0.9605, "step": 4515 }, { "epoch": 0.38, "learning_rate": 1.4163448054927358e-05, "loss": 1.1094, "step": 4516 }, { "epoch": 0.38, "learning_rate": 1.4160953916486742e-05, "loss": 1.0833, "step": 4517 }, { "epoch": 0.38, "learning_rate": 1.4158459464966988e-05, "loss": 1.0998, "step": 4518 }, { "epoch": 0.38, "learning_rate": 1.4155964700555785e-05, "loss": 1.049, "step": 4519 }, { "epoch": 0.38, "learning_rate": 1.4153469623440843e-05, "loss": 1.0727, "step": 4520 }, { "epoch": 0.38, "learning_rate": 1.4150974233809902e-05, "loss": 1.0521, "step": 4521 }, { "epoch": 0.38, "learning_rate": 1.4148478531850715e-05, "loss": 1.1022, "step": 4522 }, { "epoch": 0.38, "learning_rate": 1.414598251775107e-05, "loss": 1.0666, "step": 4523 }, { "epoch": 0.38, "learning_rate": 1.4143486191698763e-05, "loss": 1.0654, "step": 4524 }, { "epoch": 0.38, "learning_rate": 1.4140989553881633e-05, "loss": 1.0416, "step": 4525 }, { "epoch": 0.38, "learning_rate": 1.4138492604487528e-05, "loss": 1.1037, "step": 4526 }, { "epoch": 0.38, "learning_rate": 1.4135995343704324e-05, "loss": 1.1072, "step": 4527 }, { "epoch": 0.38, "learning_rate": 1.4133497771719922e-05, "loss": 1.1081, "step": 4528 }, { "epoch": 0.38, "learning_rate": 1.4130999888722242e-05, "loss": 1.0694, "step": 4529 }, { "epoch": 0.38, "learning_rate": 1.412850169489923e-05, "loss": 1.0273, "step": 4530 }, { "epoch": 0.38, "learning_rate": 1.4126003190438858e-05, "loss": 1.0312, "step": 4531 }, { "epoch": 0.38, "learning_rate": 1.4123504375529115e-05, "loss": 1.0869, "step": 4532 }, { "epoch": 0.38, "learning_rate": 1.4121005250358022e-05, "loss": 1.0413, "step": 4533 }, { "epoch": 0.38, "learning_rate": 1.4118505815113616e-05, "loss": 1.0849, "step": 4534 }, { "epoch": 0.38, "learning_rate": 1.411600606998396e-05, "loss": 0.8952, "step": 4535 }, { "epoch": 0.38, "learning_rate": 1.4113506015157144e-05, "loss": 0.9209, "step": 4536 }, { "epoch": 0.38, "learning_rate": 1.4111005650821268e-05, "loss": 1.0309, "step": 4537 }, { "epoch": 0.38, "learning_rate": 1.4108504977164474e-05, "loss": 1.0925, "step": 4538 }, { "epoch": 0.38, "learning_rate": 1.4106003994374915e-05, "loss": 1.0871, "step": 4539 }, { "epoch": 0.38, "learning_rate": 1.410350270264077e-05, "loss": 1.0349, "step": 4540 }, { "epoch": 0.38, "learning_rate": 1.4101001102150243e-05, "loss": 1.1121, "step": 4541 }, { "epoch": 0.38, "learning_rate": 1.4098499193091556e-05, "loss": 1.0781, "step": 4542 }, { "epoch": 0.38, "learning_rate": 1.4095996975652962e-05, "loss": 0.8789, "step": 4543 }, { "epoch": 0.38, "learning_rate": 1.4093494450022732e-05, "loss": 1.0781, "step": 4544 }, { "epoch": 0.38, "learning_rate": 1.409099161638916e-05, "loss": 1.0218, "step": 4545 }, { "epoch": 0.38, "learning_rate": 1.4088488474940565e-05, "loss": 0.9604, "step": 4546 }, { "epoch": 0.39, "learning_rate": 1.408598502586529e-05, "loss": 0.9913, "step": 4547 }, { "epoch": 0.39, "learning_rate": 1.4083481269351701e-05, "loss": 1.0668, "step": 4548 }, { "epoch": 0.39, "learning_rate": 1.4080977205588184e-05, "loss": 1.036, "step": 4549 }, { "epoch": 0.39, "learning_rate": 1.4078472834763149e-05, "loss": 1.0811, "step": 4550 }, { "epoch": 0.39, "learning_rate": 1.4075968157065033e-05, "loss": 1.0979, "step": 4551 }, { "epoch": 0.39, "learning_rate": 1.4073463172682293e-05, "loss": 1.0852, "step": 4552 }, { "epoch": 0.39, "learning_rate": 1.407095788180341e-05, "loss": 1.1058, "step": 4553 }, { "epoch": 0.39, "learning_rate": 1.4068452284616883e-05, "loss": 1.1041, "step": 4554 }, { "epoch": 0.39, "learning_rate": 1.4065946381311242e-05, "loss": 1.0824, "step": 4555 }, { "epoch": 0.39, "learning_rate": 1.4063440172075037e-05, "loss": 1.0141, "step": 4556 }, { "epoch": 0.39, "learning_rate": 1.4060933657096842e-05, "loss": 1.0068, "step": 4557 }, { "epoch": 0.39, "learning_rate": 1.4058426836565248e-05, "loss": 1.0768, "step": 4558 }, { "epoch": 0.39, "learning_rate": 1.4055919710668877e-05, "loss": 1.059, "step": 4559 }, { "epoch": 0.39, "learning_rate": 1.4053412279596368e-05, "loss": 1.1083, "step": 4560 }, { "epoch": 0.39, "learning_rate": 1.4050904543536388e-05, "loss": 1.0763, "step": 4561 }, { "epoch": 0.39, "learning_rate": 1.4048396502677623e-05, "loss": 1.1171, "step": 4562 }, { "epoch": 0.39, "learning_rate": 1.4045888157208784e-05, "loss": 1.1063, "step": 4563 }, { "epoch": 0.39, "learning_rate": 1.4043379507318606e-05, "loss": 1.0827, "step": 4564 }, { "epoch": 0.39, "learning_rate": 1.4040870553195842e-05, "loss": 0.8488, "step": 4565 }, { "epoch": 0.39, "learning_rate": 1.4038361295029274e-05, "loss": 1.0804, "step": 4566 }, { "epoch": 0.39, "learning_rate": 1.4035851733007704e-05, "loss": 1.0614, "step": 4567 }, { "epoch": 0.39, "learning_rate": 1.4033341867319951e-05, "loss": 1.1103, "step": 4568 }, { "epoch": 0.39, "learning_rate": 1.403083169815487e-05, "loss": 1.0801, "step": 4569 }, { "epoch": 0.39, "learning_rate": 1.4028321225701329e-05, "loss": 1.0507, "step": 4570 }, { "epoch": 0.39, "learning_rate": 1.4025810450148219e-05, "loss": 1.0757, "step": 4571 }, { "epoch": 0.39, "learning_rate": 1.4023299371684458e-05, "loss": 1.0704, "step": 4572 }, { "epoch": 0.39, "learning_rate": 1.4020787990498987e-05, "loss": 1.0986, "step": 4573 }, { "epoch": 0.39, "learning_rate": 1.4018276306780765e-05, "loss": 1.0964, "step": 4574 }, { "epoch": 0.39, "learning_rate": 1.4015764320718777e-05, "loss": 1.1308, "step": 4575 }, { "epoch": 0.39, "learning_rate": 1.4013252032502029e-05, "loss": 0.9035, "step": 4576 }, { "epoch": 0.39, "learning_rate": 1.4010739442319554e-05, "loss": 1.0651, "step": 4577 }, { "epoch": 0.39, "learning_rate": 1.4008226550360405e-05, "loss": 1.096, "step": 4578 }, { "epoch": 0.39, "learning_rate": 1.4005713356813655e-05, "loss": 1.102, "step": 4579 }, { "epoch": 0.39, "learning_rate": 1.4003199861868401e-05, "loss": 1.0673, "step": 4580 }, { "epoch": 0.39, "learning_rate": 1.4000686065713766e-05, "loss": 1.0784, "step": 4581 }, { "epoch": 0.39, "learning_rate": 1.3998171968538893e-05, "loss": 1.0693, "step": 4582 }, { "epoch": 0.39, "learning_rate": 1.399565757053295e-05, "loss": 1.1283, "step": 4583 }, { "epoch": 0.39, "learning_rate": 1.3993142871885124e-05, "loss": 1.046, "step": 4584 }, { "epoch": 0.39, "learning_rate": 1.3990627872784624e-05, "loss": 0.9958, "step": 4585 }, { "epoch": 0.39, "learning_rate": 1.3988112573420687e-05, "loss": 1.0684, "step": 4586 }, { "epoch": 0.39, "learning_rate": 1.3985596973982567e-05, "loss": 1.0949, "step": 4587 }, { "epoch": 0.39, "learning_rate": 1.3983081074659545e-05, "loss": 1.0809, "step": 4588 }, { "epoch": 0.39, "learning_rate": 1.3980564875640925e-05, "loss": 1.0727, "step": 4589 }, { "epoch": 0.39, "learning_rate": 1.3978048377116028e-05, "loss": 1.0778, "step": 4590 }, { "epoch": 0.39, "learning_rate": 1.39755315792742e-05, "loss": 1.0808, "step": 4591 }, { "epoch": 0.39, "learning_rate": 1.3973014482304812e-05, "loss": 1.1107, "step": 4592 }, { "epoch": 0.39, "learning_rate": 1.3970497086397257e-05, "loss": 1.0427, "step": 4593 }, { "epoch": 0.39, "learning_rate": 1.3967979391740949e-05, "loss": 1.0502, "step": 4594 }, { "epoch": 0.39, "learning_rate": 1.3965461398525319e-05, "loss": 1.0829, "step": 4595 }, { "epoch": 0.39, "learning_rate": 1.3962943106939835e-05, "loss": 1.0064, "step": 4596 }, { "epoch": 0.39, "learning_rate": 1.3960424517173972e-05, "loss": 1.0851, "step": 4597 }, { "epoch": 0.39, "learning_rate": 1.3957905629417236e-05, "loss": 1.0901, "step": 4598 }, { "epoch": 0.39, "learning_rate": 1.3955386443859156e-05, "loss": 1.0596, "step": 4599 }, { "epoch": 0.39, "learning_rate": 1.395286696068928e-05, "loss": 1.0926, "step": 4600 }, { "epoch": 0.39, "learning_rate": 1.3950347180097176e-05, "loss": 1.0512, "step": 4601 }, { "epoch": 0.39, "learning_rate": 1.394782710227244e-05, "loss": 1.0462, "step": 4602 }, { "epoch": 0.39, "learning_rate": 1.3945306727404688e-05, "loss": 1.0977, "step": 4603 }, { "epoch": 0.39, "learning_rate": 1.3942786055683561e-05, "loss": 1.1026, "step": 4604 }, { "epoch": 0.39, "learning_rate": 1.3940265087298717e-05, "loss": 1.1207, "step": 4605 }, { "epoch": 0.39, "learning_rate": 1.3937743822439842e-05, "loss": 1.0464, "step": 4606 }, { "epoch": 0.39, "learning_rate": 1.3935222261296635e-05, "loss": 1.0967, "step": 4607 }, { "epoch": 0.39, "learning_rate": 1.393270040405883e-05, "loss": 1.0898, "step": 4608 }, { "epoch": 0.39, "learning_rate": 1.3930178250916173e-05, "loss": 1.0643, "step": 4609 }, { "epoch": 0.39, "learning_rate": 1.3927655802058438e-05, "loss": 1.1457, "step": 4610 }, { "epoch": 0.39, "learning_rate": 1.3925133057675419e-05, "loss": 1.0952, "step": 4611 }, { "epoch": 0.39, "learning_rate": 1.3922610017956934e-05, "loss": 1.0387, "step": 4612 }, { "epoch": 0.39, "learning_rate": 1.392008668309282e-05, "loss": 1.0539, "step": 4613 }, { "epoch": 0.39, "learning_rate": 1.391756305327294e-05, "loss": 0.954, "step": 4614 }, { "epoch": 0.39, "learning_rate": 1.3915039128687178e-05, "loss": 1.1117, "step": 4615 }, { "epoch": 0.39, "learning_rate": 1.3912514909525437e-05, "loss": 1.0559, "step": 4616 }, { "epoch": 0.39, "learning_rate": 1.3909990395977644e-05, "loss": 1.1526, "step": 4617 }, { "epoch": 0.39, "learning_rate": 1.3907465588233751e-05, "loss": 1.0575, "step": 4618 }, { "epoch": 0.39, "learning_rate": 1.3904940486483731e-05, "loss": 1.0767, "step": 4619 }, { "epoch": 0.39, "learning_rate": 1.3902415090917576e-05, "loss": 0.9989, "step": 4620 }, { "epoch": 0.39, "learning_rate": 1.3899889401725305e-05, "loss": 1.0672, "step": 4621 }, { "epoch": 0.39, "learning_rate": 1.389736341909695e-05, "loss": 1.1158, "step": 4622 }, { "epoch": 0.39, "learning_rate": 1.3894837143222577e-05, "loss": 0.9165, "step": 4623 }, { "epoch": 0.39, "learning_rate": 1.389231057429227e-05, "loss": 1.0803, "step": 4624 }, { "epoch": 0.39, "learning_rate": 1.3889783712496125e-05, "loss": 1.0776, "step": 4625 }, { "epoch": 0.39, "learning_rate": 1.3887256558024279e-05, "loss": 1.0785, "step": 4626 }, { "epoch": 0.39, "learning_rate": 1.3884729111066872e-05, "loss": 1.0741, "step": 4627 }, { "epoch": 0.39, "learning_rate": 1.3882201371814077e-05, "loss": 1.1321, "step": 4628 }, { "epoch": 0.39, "learning_rate": 1.387967334045609e-05, "loss": 1.1124, "step": 4629 }, { "epoch": 0.39, "learning_rate": 1.3877145017183122e-05, "loss": 1.0877, "step": 4630 }, { "epoch": 0.39, "learning_rate": 1.3874616402185412e-05, "loss": 1.0635, "step": 4631 }, { "epoch": 0.39, "learning_rate": 1.3872087495653216e-05, "loss": 1.1058, "step": 4632 }, { "epoch": 0.39, "learning_rate": 1.3869558297776818e-05, "loss": 1.0945, "step": 4633 }, { "epoch": 0.39, "learning_rate": 1.3867028808746517e-05, "loss": 0.9746, "step": 4634 }, { "epoch": 0.39, "learning_rate": 1.3864499028752636e-05, "loss": 1.11, "step": 4635 }, { "epoch": 0.39, "learning_rate": 1.3861968957985523e-05, "loss": 1.0969, "step": 4636 }, { "epoch": 0.39, "learning_rate": 1.3859438596635548e-05, "loss": 1.0489, "step": 4637 }, { "epoch": 0.39, "learning_rate": 1.3856907944893097e-05, "loss": 1.0997, "step": 4638 }, { "epoch": 0.39, "learning_rate": 1.3854377002948582e-05, "loss": 1.0804, "step": 4639 }, { "epoch": 0.39, "learning_rate": 1.3851845770992443e-05, "loss": 1.0677, "step": 4640 }, { "epoch": 0.39, "learning_rate": 1.3849314249215126e-05, "loss": 1.078, "step": 4641 }, { "epoch": 0.39, "learning_rate": 1.3846782437807114e-05, "loss": 1.0329, "step": 4642 }, { "epoch": 0.39, "learning_rate": 1.3844250336958904e-05, "loss": 1.0754, "step": 4643 }, { "epoch": 0.39, "learning_rate": 1.3841717946861018e-05, "loss": 1.0952, "step": 4644 }, { "epoch": 0.39, "learning_rate": 1.3839185267704e-05, "loss": 1.1001, "step": 4645 }, { "epoch": 0.39, "learning_rate": 1.3836652299678406e-05, "loss": 1.0561, "step": 4646 }, { "epoch": 0.39, "learning_rate": 1.3834119042974832e-05, "loss": 1.0632, "step": 4647 }, { "epoch": 0.39, "learning_rate": 1.3831585497783884e-05, "loss": 1.0254, "step": 4648 }, { "epoch": 0.39, "learning_rate": 1.3829051664296184e-05, "loss": 1.0359, "step": 4649 }, { "epoch": 0.39, "learning_rate": 1.3826517542702392e-05, "loss": 1.0387, "step": 4650 }, { "epoch": 0.39, "learning_rate": 1.3823983133193175e-05, "loss": 1.1153, "step": 4651 }, { "epoch": 0.39, "learning_rate": 1.3821448435959228e-05, "loss": 1.0587, "step": 4652 }, { "epoch": 0.39, "learning_rate": 1.3818913451191269e-05, "loss": 1.0743, "step": 4653 }, { "epoch": 0.39, "learning_rate": 1.3816378179080035e-05, "loss": 1.0538, "step": 4654 }, { "epoch": 0.39, "learning_rate": 1.3813842619816284e-05, "loss": 1.0243, "step": 4655 }, { "epoch": 0.39, "learning_rate": 1.3811306773590802e-05, "loss": 1.0571, "step": 4656 }, { "epoch": 0.39, "learning_rate": 1.3808770640594386e-05, "loss": 1.0459, "step": 4657 }, { "epoch": 0.39, "learning_rate": 1.3806234221017862e-05, "loss": 1.105, "step": 4658 }, { "epoch": 0.39, "learning_rate": 1.3803697515052075e-05, "loss": 1.0588, "step": 4659 }, { "epoch": 0.39, "learning_rate": 1.3801160522887895e-05, "loss": 1.0633, "step": 4660 }, { "epoch": 0.39, "learning_rate": 1.3798623244716209e-05, "loss": 1.0686, "step": 4661 }, { "epoch": 0.39, "learning_rate": 1.3796085680727927e-05, "loss": 0.9437, "step": 4662 }, { "epoch": 0.39, "learning_rate": 1.3793547831113977e-05, "loss": 1.0266, "step": 4663 }, { "epoch": 0.39, "learning_rate": 1.3791009696065318e-05, "loss": 1.016, "step": 4664 }, { "epoch": 0.4, "learning_rate": 1.3788471275772924e-05, "loss": 1.0189, "step": 4665 }, { "epoch": 0.4, "learning_rate": 1.3785932570427786e-05, "loss": 1.0989, "step": 4666 }, { "epoch": 0.4, "learning_rate": 1.378339358022093e-05, "loss": 1.1041, "step": 4667 }, { "epoch": 0.4, "learning_rate": 1.378085430534339e-05, "loss": 1.0355, "step": 4668 }, { "epoch": 0.4, "learning_rate": 1.3778314745986226e-05, "loss": 1.0641, "step": 4669 }, { "epoch": 0.4, "learning_rate": 1.3775774902340523e-05, "loss": 1.0865, "step": 4670 }, { "epoch": 0.4, "learning_rate": 1.377323477459738e-05, "loss": 1.0063, "step": 4671 }, { "epoch": 0.4, "learning_rate": 1.3770694362947923e-05, "loss": 0.9674, "step": 4672 }, { "epoch": 0.4, "learning_rate": 1.3768153667583301e-05, "loss": 1.0705, "step": 4673 }, { "epoch": 0.4, "learning_rate": 1.376561268869468e-05, "loss": 1.1656, "step": 4674 }, { "epoch": 0.4, "learning_rate": 1.3763071426473248e-05, "loss": 1.057, "step": 4675 }, { "epoch": 0.4, "learning_rate": 1.3760529881110217e-05, "loss": 1.119, "step": 4676 }, { "epoch": 0.4, "learning_rate": 1.3757988052796815e-05, "loss": 1.0105, "step": 4677 }, { "epoch": 0.4, "learning_rate": 1.3755445941724297e-05, "loss": 1.0569, "step": 4678 }, { "epoch": 0.4, "learning_rate": 1.3752903548083936e-05, "loss": 1.0636, "step": 4679 }, { "epoch": 0.4, "learning_rate": 1.3750360872067023e-05, "loss": 1.0924, "step": 4680 }, { "epoch": 0.4, "learning_rate": 1.3747817913864882e-05, "loss": 1.0455, "step": 4681 }, { "epoch": 0.4, "learning_rate": 1.374527467366885e-05, "loss": 1.0957, "step": 4682 }, { "epoch": 0.4, "learning_rate": 1.3742731151670281e-05, "loss": 1.0611, "step": 4683 }, { "epoch": 0.4, "learning_rate": 1.3740187348060557e-05, "loss": 1.0682, "step": 4684 }, { "epoch": 0.4, "learning_rate": 1.3737643263031082e-05, "loss": 1.0773, "step": 4685 }, { "epoch": 0.4, "learning_rate": 1.3735098896773276e-05, "loss": 1.0796, "step": 4686 }, { "epoch": 0.4, "learning_rate": 1.373255424947858e-05, "loss": 1.0446, "step": 4687 }, { "epoch": 0.4, "learning_rate": 1.3730009321338464e-05, "loss": 1.0828, "step": 4688 }, { "epoch": 0.4, "learning_rate": 1.372746411254441e-05, "loss": 1.0422, "step": 4689 }, { "epoch": 0.4, "learning_rate": 1.3724918623287926e-05, "loss": 1.0488, "step": 4690 }, { "epoch": 0.4, "learning_rate": 1.3722372853760541e-05, "loss": 1.0641, "step": 4691 }, { "epoch": 0.4, "learning_rate": 1.3719826804153805e-05, "loss": 1.0981, "step": 4692 }, { "epoch": 0.4, "learning_rate": 1.3717280474659283e-05, "loss": 1.0465, "step": 4693 }, { "epoch": 0.4, "learning_rate": 1.3714733865468573e-05, "loss": 0.9462, "step": 4694 }, { "epoch": 0.4, "learning_rate": 1.3712186976773286e-05, "loss": 1.0219, "step": 4695 }, { "epoch": 0.4, "learning_rate": 1.3709639808765049e-05, "loss": 0.9815, "step": 4696 }, { "epoch": 0.4, "learning_rate": 1.3707092361635526e-05, "loss": 1.0885, "step": 4697 }, { "epoch": 0.4, "learning_rate": 1.3704544635576387e-05, "loss": 1.1282, "step": 4698 }, { "epoch": 0.4, "learning_rate": 1.3701996630779326e-05, "loss": 1.0782, "step": 4699 }, { "epoch": 0.4, "learning_rate": 1.3699448347436067e-05, "loss": 1.1068, "step": 4700 }, { "epoch": 0.4, "learning_rate": 1.3696899785738345e-05, "loss": 1.1117, "step": 4701 }, { "epoch": 0.4, "learning_rate": 1.3694350945877917e-05, "loss": 1.0667, "step": 4702 }, { "epoch": 0.4, "learning_rate": 1.3691801828046568e-05, "loss": 1.0607, "step": 4703 }, { "epoch": 0.4, "learning_rate": 1.3689252432436093e-05, "loss": 1.0647, "step": 4704 }, { "epoch": 0.4, "learning_rate": 1.3686702759238319e-05, "loss": 1.0669, "step": 4705 }, { "epoch": 0.4, "learning_rate": 1.3684152808645086e-05, "loss": 1.0733, "step": 4706 }, { "epoch": 0.4, "learning_rate": 1.3681602580848261e-05, "loss": 1.0787, "step": 4707 }, { "epoch": 0.4, "learning_rate": 1.3679052076039724e-05, "loss": 1.0857, "step": 4708 }, { "epoch": 0.4, "learning_rate": 1.3676501294411386e-05, "loss": 1.0325, "step": 4709 }, { "epoch": 0.4, "learning_rate": 1.367395023615517e-05, "loss": 1.0876, "step": 4710 }, { "epoch": 0.4, "learning_rate": 1.3671398901463024e-05, "loss": 1.1196, "step": 4711 }, { "epoch": 0.4, "learning_rate": 1.3668847290526914e-05, "loss": 1.0783, "step": 4712 }, { "epoch": 0.4, "learning_rate": 1.3666295403538837e-05, "loss": 1.1161, "step": 4713 }, { "epoch": 0.4, "learning_rate": 1.3663743240690787e-05, "loss": 1.045, "step": 4714 }, { "epoch": 0.4, "learning_rate": 1.3661190802174806e-05, "loss": 1.0978, "step": 4715 }, { "epoch": 0.4, "learning_rate": 1.3658638088182942e-05, "loss": 1.1126, "step": 4716 }, { "epoch": 0.4, "learning_rate": 1.3656085098907268e-05, "loss": 1.0488, "step": 4717 }, { "epoch": 0.4, "learning_rate": 1.3653531834539874e-05, "loss": 1.0554, "step": 4718 }, { "epoch": 0.4, "learning_rate": 1.3650978295272875e-05, "loss": 1.0715, "step": 4719 }, { "epoch": 0.4, "learning_rate": 1.3648424481298404e-05, "loss": 1.0441, "step": 4720 }, { "epoch": 0.4, "learning_rate": 1.3645870392808613e-05, "loss": 1.0296, "step": 4721 }, { "epoch": 0.4, "learning_rate": 1.3643316029995682e-05, "loss": 1.0635, "step": 4722 }, { "epoch": 0.4, "learning_rate": 1.3640761393051802e-05, "loss": 1.1407, "step": 4723 }, { "epoch": 0.4, "learning_rate": 1.3638206482169195e-05, "loss": 1.0525, "step": 4724 }, { "epoch": 0.4, "learning_rate": 1.3635651297540094e-05, "loss": 1.0289, "step": 4725 }, { "epoch": 0.4, "learning_rate": 1.3633095839356757e-05, "loss": 1.0596, "step": 4726 }, { "epoch": 0.4, "learning_rate": 1.3630540107811459e-05, "loss": 1.1017, "step": 4727 }, { "epoch": 0.4, "learning_rate": 1.3627984103096507e-05, "loss": 1.0309, "step": 4728 }, { "epoch": 0.4, "learning_rate": 1.3625427825404214e-05, "loss": 1.1079, "step": 4729 }, { "epoch": 0.4, "learning_rate": 1.362287127492692e-05, "loss": 1.0714, "step": 4730 }, { "epoch": 0.4, "learning_rate": 1.3620314451856989e-05, "loss": 1.0585, "step": 4731 }, { "epoch": 0.4, "learning_rate": 1.3617757356386796e-05, "loss": 1.0327, "step": 4732 }, { "epoch": 0.4, "learning_rate": 1.3615199988708747e-05, "loss": 1.0805, "step": 4733 }, { "epoch": 0.4, "learning_rate": 1.3612642349015266e-05, "loss": 1.0664, "step": 4734 }, { "epoch": 0.4, "learning_rate": 1.361008443749879e-05, "loss": 1.0676, "step": 4735 }, { "epoch": 0.4, "learning_rate": 1.3607526254351784e-05, "loss": 1.083, "step": 4736 }, { "epoch": 0.4, "learning_rate": 1.360496779976673e-05, "loss": 1.0729, "step": 4737 }, { "epoch": 0.4, "learning_rate": 1.3602409073936135e-05, "loss": 1.0772, "step": 4738 }, { "epoch": 0.4, "learning_rate": 1.359985007705252e-05, "loss": 1.0904, "step": 4739 }, { "epoch": 0.4, "learning_rate": 1.3597290809308433e-05, "loss": 1.0414, "step": 4740 }, { "epoch": 0.4, "learning_rate": 1.3594731270896433e-05, "loss": 1.0706, "step": 4741 }, { "epoch": 0.4, "learning_rate": 1.3592171462009107e-05, "loss": 1.0939, "step": 4742 }, { "epoch": 0.4, "learning_rate": 1.3589611382839066e-05, "loss": 1.0354, "step": 4743 }, { "epoch": 0.4, "learning_rate": 1.3587051033578929e-05, "loss": 1.0529, "step": 4744 }, { "epoch": 0.4, "learning_rate": 1.3584490414421345e-05, "loss": 1.0362, "step": 4745 }, { "epoch": 0.4, "learning_rate": 1.3581929525558979e-05, "loss": 1.0287, "step": 4746 }, { "epoch": 0.4, "learning_rate": 1.357936836718452e-05, "loss": 1.0889, "step": 4747 }, { "epoch": 0.4, "learning_rate": 1.3576806939490674e-05, "loss": 1.0672, "step": 4748 }, { "epoch": 0.4, "learning_rate": 1.357424524267017e-05, "loss": 1.0928, "step": 4749 }, { "epoch": 0.4, "learning_rate": 1.3571683276915754e-05, "loss": 1.0653, "step": 4750 }, { "epoch": 0.4, "learning_rate": 1.3569121042420195e-05, "loss": 1.0365, "step": 4751 }, { "epoch": 0.4, "learning_rate": 1.3566558539376278e-05, "loss": 1.0965, "step": 4752 }, { "epoch": 0.4, "learning_rate": 1.3563995767976815e-05, "loss": 1.0934, "step": 4753 }, { "epoch": 0.4, "learning_rate": 1.3561432728414629e-05, "loss": 1.0705, "step": 4754 }, { "epoch": 0.4, "learning_rate": 1.3558869420882577e-05, "loss": 1.0903, "step": 4755 }, { "epoch": 0.4, "learning_rate": 1.3556305845573521e-05, "loss": 1.0892, "step": 4756 }, { "epoch": 0.4, "learning_rate": 1.3553742002680353e-05, "loss": 1.0637, "step": 4757 }, { "epoch": 0.4, "learning_rate": 1.3551177892395977e-05, "loss": 1.0452, "step": 4758 }, { "epoch": 0.4, "learning_rate": 1.3548613514913331e-05, "loss": 1.0658, "step": 4759 }, { "epoch": 0.4, "learning_rate": 1.3546048870425356e-05, "loss": 1.014, "step": 4760 }, { "epoch": 0.4, "learning_rate": 1.3543483959125027e-05, "loss": 1.0737, "step": 4761 }, { "epoch": 0.4, "learning_rate": 1.3540918781205331e-05, "loss": 1.0764, "step": 4762 }, { "epoch": 0.4, "learning_rate": 1.353835333685928e-05, "loss": 1.0785, "step": 4763 }, { "epoch": 0.4, "learning_rate": 1.35357876262799e-05, "loss": 1.1209, "step": 4764 }, { "epoch": 0.4, "learning_rate": 1.3533221649660242e-05, "loss": 1.0771, "step": 4765 }, { "epoch": 0.4, "learning_rate": 1.3530655407193374e-05, "loss": 1.0883, "step": 4766 }, { "epoch": 0.4, "learning_rate": 1.352808889907239e-05, "loss": 1.0522, "step": 4767 }, { "epoch": 0.4, "learning_rate": 1.3525522125490397e-05, "loss": 1.0718, "step": 4768 }, { "epoch": 0.4, "learning_rate": 1.3522955086640523e-05, "loss": 1.0358, "step": 4769 }, { "epoch": 0.4, "learning_rate": 1.352038778271592e-05, "loss": 1.0685, "step": 4770 }, { "epoch": 0.4, "learning_rate": 1.3517820213909756e-05, "loss": 0.919, "step": 4771 }, { "epoch": 0.4, "learning_rate": 1.351525238041522e-05, "loss": 1.0925, "step": 4772 }, { "epoch": 0.4, "learning_rate": 1.3512684282425523e-05, "loss": 1.1108, "step": 4773 }, { "epoch": 0.4, "learning_rate": 1.3510115920133892e-05, "loss": 1.0951, "step": 4774 }, { "epoch": 0.4, "learning_rate": 1.350754729373358e-05, "loss": 1.0413, "step": 4775 }, { "epoch": 0.4, "learning_rate": 1.3504978403417852e-05, "loss": 1.0528, "step": 4776 }, { "epoch": 0.4, "learning_rate": 1.3502409249379997e-05, "loss": 1.0727, "step": 4777 }, { "epoch": 0.4, "learning_rate": 1.3499839831813328e-05, "loss": 1.0426, "step": 4778 }, { "epoch": 0.4, "learning_rate": 1.3497270150911168e-05, "loss": 1.0623, "step": 4779 }, { "epoch": 0.4, "learning_rate": 1.3494700206866869e-05, "loss": 1.0492, "step": 4780 }, { "epoch": 0.4, "learning_rate": 1.34921299998738e-05, "loss": 1.0719, "step": 4781 }, { "epoch": 0.4, "learning_rate": 1.3489559530125346e-05, "loss": 1.0731, "step": 4782 }, { "epoch": 0.41, "learning_rate": 1.3486988797814912e-05, "loss": 1.0925, "step": 4783 }, { "epoch": 0.41, "learning_rate": 1.3484417803135931e-05, "loss": 1.083, "step": 4784 }, { "epoch": 0.41, "learning_rate": 1.3481846546281849e-05, "loss": 1.1205, "step": 4785 }, { "epoch": 0.41, "learning_rate": 1.3479275027446132e-05, "loss": 1.0937, "step": 4786 }, { "epoch": 0.41, "learning_rate": 1.347670324682227e-05, "loss": 1.0739, "step": 4787 }, { "epoch": 0.41, "learning_rate": 1.347413120460376e-05, "loss": 1.1332, "step": 4788 }, { "epoch": 0.41, "learning_rate": 1.3471558900984138e-05, "loss": 1.1009, "step": 4789 }, { "epoch": 0.41, "learning_rate": 1.3468986336156944e-05, "loss": 0.9971, "step": 4790 }, { "epoch": 0.41, "learning_rate": 1.3466413510315746e-05, "loss": 1.0577, "step": 4791 }, { "epoch": 0.41, "learning_rate": 1.346384042365413e-05, "loss": 1.0587, "step": 4792 }, { "epoch": 0.41, "learning_rate": 1.3461267076365694e-05, "loss": 1.0753, "step": 4793 }, { "epoch": 0.41, "learning_rate": 1.3458693468644071e-05, "loss": 0.8425, "step": 4794 }, { "epoch": 0.41, "learning_rate": 1.34561196006829e-05, "loss": 1.0519, "step": 4795 }, { "epoch": 0.41, "learning_rate": 1.3453545472675842e-05, "loss": 1.0755, "step": 4796 }, { "epoch": 0.41, "learning_rate": 1.3450971084816585e-05, "loss": 1.0826, "step": 4797 }, { "epoch": 0.41, "learning_rate": 1.344839643729883e-05, "loss": 1.0514, "step": 4798 }, { "epoch": 0.41, "learning_rate": 1.3445821530316297e-05, "loss": 1.1027, "step": 4799 }, { "epoch": 0.41, "learning_rate": 1.3443246364062727e-05, "loss": 1.0524, "step": 4800 }, { "epoch": 0.41, "learning_rate": 1.3440670938731885e-05, "loss": 0.9111, "step": 4801 }, { "epoch": 0.41, "learning_rate": 1.3438095254517552e-05, "loss": 1.0506, "step": 4802 }, { "epoch": 0.41, "learning_rate": 1.3435519311613524e-05, "loss": 1.0315, "step": 4803 }, { "epoch": 0.41, "learning_rate": 1.3432943110213619e-05, "loss": 1.03, "step": 4804 }, { "epoch": 0.41, "learning_rate": 1.3430366650511681e-05, "loss": 1.0544, "step": 4805 }, { "epoch": 0.41, "learning_rate": 1.3427789932701567e-05, "loss": 0.9884, "step": 4806 }, { "epoch": 0.41, "learning_rate": 1.3425212956977154e-05, "loss": 1.066, "step": 4807 }, { "epoch": 0.41, "learning_rate": 1.342263572353234e-05, "loss": 1.0372, "step": 4808 }, { "epoch": 0.41, "learning_rate": 1.342005823256104e-05, "loss": 1.042, "step": 4809 }, { "epoch": 0.41, "learning_rate": 1.3417480484257191e-05, "loss": 1.0499, "step": 4810 }, { "epoch": 0.41, "learning_rate": 1.3414902478814748e-05, "loss": 1.082, "step": 4811 }, { "epoch": 0.41, "learning_rate": 1.3412324216427688e-05, "loss": 1.0621, "step": 4812 }, { "epoch": 0.41, "learning_rate": 1.3409745697290002e-05, "loss": 1.0649, "step": 4813 }, { "epoch": 0.41, "learning_rate": 1.3407166921595702e-05, "loss": 1.0649, "step": 4814 }, { "epoch": 0.41, "learning_rate": 1.3404587889538826e-05, "loss": 1.0449, "step": 4815 }, { "epoch": 0.41, "learning_rate": 1.340200860131342e-05, "loss": 0.9165, "step": 4816 }, { "epoch": 0.41, "learning_rate": 1.339942905711356e-05, "loss": 1.0349, "step": 4817 }, { "epoch": 0.41, "learning_rate": 1.3396849257133335e-05, "loss": 1.0527, "step": 4818 }, { "epoch": 0.41, "learning_rate": 1.3394269201566853e-05, "loss": 1.0857, "step": 4819 }, { "epoch": 0.41, "learning_rate": 1.3391688890608247e-05, "loss": 1.0647, "step": 4820 }, { "epoch": 0.41, "learning_rate": 1.3389108324451661e-05, "loss": 1.0699, "step": 4821 }, { "epoch": 0.41, "learning_rate": 1.3386527503291262e-05, "loss": 1.0914, "step": 4822 }, { "epoch": 0.41, "learning_rate": 1.3383946427321242e-05, "loss": 1.0114, "step": 4823 }, { "epoch": 0.41, "learning_rate": 1.3381365096735799e-05, "loss": 1.0398, "step": 4824 }, { "epoch": 0.41, "learning_rate": 1.3378783511729163e-05, "loss": 1.027, "step": 4825 }, { "epoch": 0.41, "learning_rate": 1.3376201672495576e-05, "loss": 1.0722, "step": 4826 }, { "epoch": 0.41, "learning_rate": 1.3373619579229304e-05, "loss": 1.0647, "step": 4827 }, { "epoch": 0.41, "learning_rate": 1.3371037232124626e-05, "loss": 1.0519, "step": 4828 }, { "epoch": 0.41, "learning_rate": 1.3368454631375845e-05, "loss": 1.0752, "step": 4829 }, { "epoch": 0.41, "learning_rate": 1.3365871777177279e-05, "loss": 1.0363, "step": 4830 }, { "epoch": 0.41, "learning_rate": 1.3363288669723273e-05, "loss": 1.102, "step": 4831 }, { "epoch": 0.41, "learning_rate": 1.3360705309208179e-05, "loss": 1.0925, "step": 4832 }, { "epoch": 0.41, "learning_rate": 1.3358121695826381e-05, "loss": 1.1181, "step": 4833 }, { "epoch": 0.41, "learning_rate": 1.3355537829772271e-05, "loss": 0.9609, "step": 4834 }, { "epoch": 0.41, "learning_rate": 1.3352953711240266e-05, "loss": 1.0552, "step": 4835 }, { "epoch": 0.41, "learning_rate": 1.33503693404248e-05, "loss": 1.0828, "step": 4836 }, { "epoch": 0.41, "learning_rate": 1.334778471752033e-05, "loss": 1.0716, "step": 4837 }, { "epoch": 0.41, "learning_rate": 1.3345199842721322e-05, "loss": 1.0195, "step": 4838 }, { "epoch": 0.41, "learning_rate": 1.3342614716222273e-05, "loss": 1.0893, "step": 4839 }, { "epoch": 0.41, "learning_rate": 1.3340029338217693e-05, "loss": 1.107, "step": 4840 }, { "epoch": 0.41, "learning_rate": 1.3337443708902109e-05, "loss": 1.0917, "step": 4841 }, { "epoch": 0.41, "learning_rate": 1.3334857828470074e-05, "loss": 1.0938, "step": 4842 }, { "epoch": 0.41, "learning_rate": 1.3332271697116149e-05, "loss": 1.0806, "step": 4843 }, { "epoch": 0.41, "learning_rate": 1.3329685315034922e-05, "loss": 1.0489, "step": 4844 }, { "epoch": 0.41, "learning_rate": 1.3327098682421e-05, "loss": 0.8887, "step": 4845 }, { "epoch": 0.41, "learning_rate": 1.3324511799469006e-05, "loss": 1.1072, "step": 4846 }, { "epoch": 0.41, "learning_rate": 1.3321924666373585e-05, "loss": 1.0586, "step": 4847 }, { "epoch": 0.41, "learning_rate": 1.3319337283329393e-05, "loss": 1.0811, "step": 4848 }, { "epoch": 0.41, "learning_rate": 1.3316749650531114e-05, "loss": 1.0874, "step": 4849 }, { "epoch": 0.41, "learning_rate": 1.3314161768173447e-05, "loss": 1.0305, "step": 4850 }, { "epoch": 0.41, "learning_rate": 1.3311573636451108e-05, "loss": 1.0112, "step": 4851 }, { "epoch": 0.41, "learning_rate": 1.3308985255558832e-05, "loss": 1.0512, "step": 4852 }, { "epoch": 0.41, "learning_rate": 1.330639662569138e-05, "loss": 1.0728, "step": 4853 }, { "epoch": 0.41, "learning_rate": 1.3303807747043522e-05, "loss": 0.9963, "step": 4854 }, { "epoch": 0.41, "learning_rate": 1.3301218619810055e-05, "loss": 1.1068, "step": 4855 }, { "epoch": 0.41, "learning_rate": 1.3298629244185783e-05, "loss": 1.05, "step": 4856 }, { "epoch": 0.41, "learning_rate": 1.3296039620365543e-05, "loss": 1.0646, "step": 4857 }, { "epoch": 0.41, "learning_rate": 1.329344974854418e-05, "loss": 1.0779, "step": 4858 }, { "epoch": 0.41, "learning_rate": 1.3290859628916566e-05, "loss": 1.0384, "step": 4859 }, { "epoch": 0.41, "learning_rate": 1.3288269261677583e-05, "loss": 1.0504, "step": 4860 }, { "epoch": 0.41, "learning_rate": 1.3285678647022136e-05, "loss": 1.0553, "step": 4861 }, { "epoch": 0.41, "learning_rate": 1.328308778514515e-05, "loss": 1.0854, "step": 4862 }, { "epoch": 0.41, "learning_rate": 1.3280496676241566e-05, "loss": 1.046, "step": 4863 }, { "epoch": 0.41, "learning_rate": 1.3277905320506346e-05, "loss": 1.1206, "step": 4864 }, { "epoch": 0.41, "learning_rate": 1.3275313718134465e-05, "loss": 1.0522, "step": 4865 }, { "epoch": 0.41, "learning_rate": 1.3272721869320929e-05, "loss": 1.0578, "step": 4866 }, { "epoch": 0.41, "learning_rate": 1.3270129774260746e-05, "loss": 1.0265, "step": 4867 }, { "epoch": 0.41, "learning_rate": 1.3267537433148955e-05, "loss": 1.1234, "step": 4868 }, { "epoch": 0.41, "learning_rate": 1.326494484618061e-05, "loss": 1.0661, "step": 4869 }, { "epoch": 0.41, "learning_rate": 1.3262352013550781e-05, "loss": 1.0627, "step": 4870 }, { "epoch": 0.41, "learning_rate": 1.3259758935454557e-05, "loss": 1.0591, "step": 4871 }, { "epoch": 0.41, "learning_rate": 1.3257165612087048e-05, "loss": 1.079, "step": 4872 }, { "epoch": 0.41, "learning_rate": 1.3254572043643383e-05, "loss": 1.0714, "step": 4873 }, { "epoch": 0.41, "learning_rate": 1.325197823031871e-05, "loss": 1.0481, "step": 4874 }, { "epoch": 0.41, "learning_rate": 1.3249384172308184e-05, "loss": 1.0662, "step": 4875 }, { "epoch": 0.41, "learning_rate": 1.3246789869806995e-05, "loss": 1.0552, "step": 4876 }, { "epoch": 0.41, "learning_rate": 1.324419532301034e-05, "loss": 1.1042, "step": 4877 }, { "epoch": 0.41, "learning_rate": 1.3241600532113442e-05, "loss": 1.0663, "step": 4878 }, { "epoch": 0.41, "learning_rate": 1.3239005497311536e-05, "loss": 1.1156, "step": 4879 }, { "epoch": 0.41, "learning_rate": 1.323641021879988e-05, "loss": 1.0533, "step": 4880 }, { "epoch": 0.41, "learning_rate": 1.3233814696773743e-05, "loss": 1.0695, "step": 4881 }, { "epoch": 0.41, "learning_rate": 1.3231218931428427e-05, "loss": 1.1012, "step": 4882 }, { "epoch": 0.41, "learning_rate": 1.3228622922959235e-05, "loss": 1.0495, "step": 4883 }, { "epoch": 0.41, "learning_rate": 1.32260266715615e-05, "loss": 1.1214, "step": 4884 }, { "epoch": 0.41, "learning_rate": 1.322343017743057e-05, "loss": 1.0532, "step": 4885 }, { "epoch": 0.41, "learning_rate": 1.3220833440761806e-05, "loss": 1.0913, "step": 4886 }, { "epoch": 0.41, "learning_rate": 1.3218236461750597e-05, "loss": 1.0486, "step": 4887 }, { "epoch": 0.41, "learning_rate": 1.3215639240592344e-05, "loss": 1.0587, "step": 4888 }, { "epoch": 0.41, "learning_rate": 1.3213041777482466e-05, "loss": 1.0948, "step": 4889 }, { "epoch": 0.41, "learning_rate": 1.3210444072616406e-05, "loss": 1.0315, "step": 4890 }, { "epoch": 0.41, "learning_rate": 1.3207846126189617e-05, "loss": 1.0297, "step": 4891 }, { "epoch": 0.41, "learning_rate": 1.3205247938397574e-05, "loss": 1.0532, "step": 4892 }, { "epoch": 0.41, "learning_rate": 1.320264950943577e-05, "loss": 1.0691, "step": 4893 }, { "epoch": 0.41, "learning_rate": 1.320005083949972e-05, "loss": 1.0378, "step": 4894 }, { "epoch": 0.41, "learning_rate": 1.319745192878495e-05, "loss": 1.0305, "step": 4895 }, { "epoch": 0.41, "learning_rate": 1.3194852777487009e-05, "loss": 1.1132, "step": 4896 }, { "epoch": 0.41, "learning_rate": 1.3192253385801461e-05, "loss": 1.0172, "step": 4897 }, { "epoch": 0.41, "learning_rate": 1.3189653753923894e-05, "loss": 1.075, "step": 4898 }, { "epoch": 0.41, "learning_rate": 1.3187053882049907e-05, "loss": 1.0881, "step": 4899 }, { "epoch": 0.41, "learning_rate": 1.3184453770375119e-05, "loss": 1.0829, "step": 4900 }, { "epoch": 0.42, "learning_rate": 1.3181853419095171e-05, "loss": 1.1113, "step": 4901 }, { "epoch": 0.42, "learning_rate": 1.3179252828405718e-05, "loss": 1.0444, "step": 4902 }, { "epoch": 0.42, "learning_rate": 1.317665199850243e-05, "loss": 1.0313, "step": 4903 }, { "epoch": 0.42, "learning_rate": 1.3174050929581004e-05, "loss": 1.0525, "step": 4904 }, { "epoch": 0.42, "learning_rate": 1.3171449621837149e-05, "loss": 1.0397, "step": 4905 }, { "epoch": 0.42, "learning_rate": 1.316884807546659e-05, "loss": 1.1143, "step": 4906 }, { "epoch": 0.42, "learning_rate": 1.3166246290665077e-05, "loss": 1.0843, "step": 4907 }, { "epoch": 0.42, "learning_rate": 1.3163644267628373e-05, "loss": 1.0438, "step": 4908 }, { "epoch": 0.42, "learning_rate": 1.3161042006552259e-05, "loss": 1.0883, "step": 4909 }, { "epoch": 0.42, "learning_rate": 1.3158439507632532e-05, "loss": 1.13, "step": 4910 }, { "epoch": 0.42, "learning_rate": 1.3155836771065014e-05, "loss": 0.8658, "step": 4911 }, { "epoch": 0.42, "learning_rate": 1.315323379704554e-05, "loss": 1.1048, "step": 4912 }, { "epoch": 0.42, "learning_rate": 1.315063058576996e-05, "loss": 1.0282, "step": 4913 }, { "epoch": 0.42, "learning_rate": 1.3148027137434149e-05, "loss": 1.0576, "step": 4914 }, { "epoch": 0.42, "learning_rate": 1.3145423452233997e-05, "loss": 1.038, "step": 4915 }, { "epoch": 0.42, "learning_rate": 1.3142819530365405e-05, "loss": 1.065, "step": 4916 }, { "epoch": 0.42, "learning_rate": 1.3140215372024303e-05, "loss": 1.0311, "step": 4917 }, { "epoch": 0.42, "learning_rate": 1.3137610977406632e-05, "loss": 1.0379, "step": 4918 }, { "epoch": 0.42, "learning_rate": 1.313500634670835e-05, "loss": 1.0701, "step": 4919 }, { "epoch": 0.42, "learning_rate": 1.3132401480125436e-05, "loss": 1.1153, "step": 4920 }, { "epoch": 0.42, "learning_rate": 1.312979637785389e-05, "loss": 0.9018, "step": 4921 }, { "epoch": 0.42, "learning_rate": 1.312719104008972e-05, "loss": 1.0585, "step": 4922 }, { "epoch": 0.42, "learning_rate": 1.3124585467028961e-05, "loss": 1.0745, "step": 4923 }, { "epoch": 0.42, "learning_rate": 1.3121979658867658e-05, "loss": 1.0099, "step": 4924 }, { "epoch": 0.42, "learning_rate": 1.3119373615801882e-05, "loss": 1.0844, "step": 4925 }, { "epoch": 0.42, "learning_rate": 1.3116767338027714e-05, "loss": 1.0499, "step": 4926 }, { "epoch": 0.42, "learning_rate": 1.311416082574126e-05, "loss": 1.0507, "step": 4927 }, { "epoch": 0.42, "learning_rate": 1.3111554079138635e-05, "loss": 1.0803, "step": 4928 }, { "epoch": 0.42, "learning_rate": 1.3108947098415978e-05, "loss": 1.0932, "step": 4929 }, { "epoch": 0.42, "learning_rate": 1.3106339883769444e-05, "loss": 1.0457, "step": 4930 }, { "epoch": 0.42, "learning_rate": 1.3103732435395205e-05, "loss": 1.0787, "step": 4931 }, { "epoch": 0.42, "learning_rate": 1.3101124753489452e-05, "loss": 1.048, "step": 4932 }, { "epoch": 0.42, "learning_rate": 1.309851683824839e-05, "loss": 1.0721, "step": 4933 }, { "epoch": 0.42, "learning_rate": 1.3095908689868248e-05, "loss": 1.0532, "step": 4934 }, { "epoch": 0.42, "learning_rate": 1.3093300308545267e-05, "loss": 1.0218, "step": 4935 }, { "epoch": 0.42, "learning_rate": 1.3090691694475705e-05, "loss": 1.0986, "step": 4936 }, { "epoch": 0.42, "learning_rate": 1.3088082847855845e-05, "loss": 1.0518, "step": 4937 }, { "epoch": 0.42, "learning_rate": 1.3085473768881976e-05, "loss": 1.0002, "step": 4938 }, { "epoch": 0.42, "learning_rate": 1.3082864457750414e-05, "loss": 1.0242, "step": 4939 }, { "epoch": 0.42, "learning_rate": 1.3080254914657491e-05, "loss": 1.0908, "step": 4940 }, { "epoch": 0.42, "learning_rate": 1.3077645139799554e-05, "loss": 1.0988, "step": 4941 }, { "epoch": 0.42, "learning_rate": 1.3075035133372963e-05, "loss": 0.9255, "step": 4942 }, { "epoch": 0.42, "learning_rate": 1.3072424895574104e-05, "loss": 0.9927, "step": 4943 }, { "epoch": 0.42, "learning_rate": 1.306981442659938e-05, "loss": 1.1335, "step": 4944 }, { "epoch": 0.42, "learning_rate": 1.3067203726645204e-05, "loss": 0.942, "step": 4945 }, { "epoch": 0.42, "learning_rate": 1.3064592795908014e-05, "loss": 1.0679, "step": 4946 }, { "epoch": 0.42, "learning_rate": 1.3061981634584258e-05, "loss": 1.0687, "step": 4947 }, { "epoch": 0.42, "learning_rate": 1.3059370242870406e-05, "loss": 1.0671, "step": 4948 }, { "epoch": 0.42, "learning_rate": 1.3056758620962949e-05, "loss": 0.8935, "step": 4949 }, { "epoch": 0.42, "learning_rate": 1.305414676905839e-05, "loss": 1.0361, "step": 4950 }, { "epoch": 0.42, "learning_rate": 1.3051534687353244e-05, "loss": 1.0366, "step": 4951 }, { "epoch": 0.42, "learning_rate": 1.3048922376044058e-05, "loss": 1.1347, "step": 4952 }, { "epoch": 0.42, "learning_rate": 1.3046309835327383e-05, "loss": 1.0298, "step": 4953 }, { "epoch": 0.42, "learning_rate": 1.3043697065399794e-05, "loss": 1.1471, "step": 4954 }, { "epoch": 0.42, "learning_rate": 1.304108406645788e-05, "loss": 1.0396, "step": 4955 }, { "epoch": 0.42, "learning_rate": 1.3038470838698252e-05, "loss": 1.1468, "step": 4956 }, { "epoch": 0.42, "learning_rate": 1.3035857382317532e-05, "loss": 1.0455, "step": 4957 }, { "epoch": 0.42, "learning_rate": 1.3033243697512359e-05, "loss": 1.019, "step": 4958 }, { "epoch": 0.42, "learning_rate": 1.3030629784479398e-05, "loss": 1.0838, "step": 4959 }, { "epoch": 0.42, "learning_rate": 1.3028015643415325e-05, "loss": 1.032, "step": 4960 }, { "epoch": 0.42, "learning_rate": 1.302540127451683e-05, "loss": 1.0914, "step": 4961 }, { "epoch": 0.42, "learning_rate": 1.3022786677980626e-05, "loss": 1.1167, "step": 4962 }, { "epoch": 0.42, "learning_rate": 1.302017185400344e-05, "loss": 1.0663, "step": 4963 }, { "epoch": 0.42, "learning_rate": 1.3017556802782018e-05, "loss": 1.0678, "step": 4964 }, { "epoch": 0.42, "learning_rate": 1.301494152451312e-05, "loss": 1.0562, "step": 4965 }, { "epoch": 0.42, "learning_rate": 1.3012326019393533e-05, "loss": 1.06, "step": 4966 }, { "epoch": 0.42, "learning_rate": 1.3009710287620042e-05, "loss": 0.9889, "step": 4967 }, { "epoch": 0.42, "learning_rate": 1.3007094329389464e-05, "loss": 1.032, "step": 4968 }, { "epoch": 0.42, "learning_rate": 1.3004478144898633e-05, "loss": 1.0637, "step": 4969 }, { "epoch": 0.42, "learning_rate": 1.3001861734344394e-05, "loss": 1.0125, "step": 4970 }, { "epoch": 0.42, "learning_rate": 1.2999245097923611e-05, "loss": 1.0722, "step": 4971 }, { "epoch": 0.42, "learning_rate": 1.2996628235833166e-05, "loss": 1.0745, "step": 4972 }, { "epoch": 0.42, "learning_rate": 1.2994011148269958e-05, "loss": 1.1009, "step": 4973 }, { "epoch": 0.42, "learning_rate": 1.2991393835430901e-05, "loss": 1.0789, "step": 4974 }, { "epoch": 0.42, "learning_rate": 1.2988776297512926e-05, "loss": 1.0574, "step": 4975 }, { "epoch": 0.42, "learning_rate": 1.2986158534712985e-05, "loss": 1.1044, "step": 4976 }, { "epoch": 0.42, "learning_rate": 1.2983540547228043e-05, "loss": 1.0648, "step": 4977 }, { "epoch": 0.42, "learning_rate": 1.2980922335255081e-05, "loss": 1.0695, "step": 4978 }, { "epoch": 0.42, "learning_rate": 1.2978303898991103e-05, "loss": 1.047, "step": 4979 }, { "epoch": 0.42, "learning_rate": 1.2975685238633124e-05, "loss": 1.0464, "step": 4980 }, { "epoch": 0.42, "learning_rate": 1.2973066354378178e-05, "loss": 1.0343, "step": 4981 }, { "epoch": 0.42, "learning_rate": 1.2970447246423312e-05, "loss": 1.0394, "step": 4982 }, { "epoch": 0.42, "learning_rate": 1.2967827914965598e-05, "loss": 1.0697, "step": 4983 }, { "epoch": 0.42, "learning_rate": 1.2965208360202119e-05, "loss": 1.0702, "step": 4984 }, { "epoch": 0.42, "learning_rate": 1.2962588582329973e-05, "loss": 1.0686, "step": 4985 }, { "epoch": 0.42, "learning_rate": 1.2959968581546281e-05, "loss": 1.0317, "step": 4986 }, { "epoch": 0.42, "learning_rate": 1.2957348358048176e-05, "loss": 0.8835, "step": 4987 }, { "epoch": 0.42, "learning_rate": 1.295472791203281e-05, "loss": 1.0204, "step": 4988 }, { "epoch": 0.42, "learning_rate": 1.2952107243697352e-05, "loss": 1.0553, "step": 4989 }, { "epoch": 0.42, "learning_rate": 1.2949486353238982e-05, "loss": 1.05, "step": 4990 }, { "epoch": 0.42, "learning_rate": 1.2946865240854907e-05, "loss": 1.0316, "step": 4991 }, { "epoch": 0.42, "learning_rate": 1.2944243906742343e-05, "loss": 1.0701, "step": 4992 }, { "epoch": 0.42, "learning_rate": 1.2941622351098526e-05, "loss": 1.0678, "step": 4993 }, { "epoch": 0.42, "learning_rate": 1.2939000574120706e-05, "loss": 1.0744, "step": 4994 }, { "epoch": 0.42, "learning_rate": 1.293637857600615e-05, "loss": 1.0185, "step": 4995 }, { "epoch": 0.42, "learning_rate": 1.2933756356952145e-05, "loss": 1.0532, "step": 4996 }, { "epoch": 0.42, "learning_rate": 1.2931133917155994e-05, "loss": 1.0756, "step": 4997 }, { "epoch": 0.42, "learning_rate": 1.2928511256815011e-05, "loss": 1.0613, "step": 4998 }, { "epoch": 0.42, "learning_rate": 1.292588837612653e-05, "loss": 1.0007, "step": 4999 }, { "epoch": 0.42, "learning_rate": 1.2923265275287906e-05, "loss": 1.0634, "step": 5000 }, { "epoch": 0.42, "learning_rate": 1.2920641954496507e-05, "loss": 1.0172, "step": 5001 }, { "epoch": 0.42, "learning_rate": 1.2918018413949714e-05, "loss": 1.0609, "step": 5002 }, { "epoch": 0.42, "learning_rate": 1.2915394653844929e-05, "loss": 1.0504, "step": 5003 }, { "epoch": 0.42, "learning_rate": 1.2912770674379571e-05, "loss": 1.0399, "step": 5004 }, { "epoch": 0.42, "learning_rate": 1.291014647575107e-05, "loss": 1.0423, "step": 5005 }, { "epoch": 0.42, "learning_rate": 1.2907522058156882e-05, "loss": 1.0918, "step": 5006 }, { "epoch": 0.42, "learning_rate": 1.2904897421794472e-05, "loss": 1.0935, "step": 5007 }, { "epoch": 0.42, "learning_rate": 1.290227256686132e-05, "loss": 0.9486, "step": 5008 }, { "epoch": 0.42, "learning_rate": 1.2899647493554927e-05, "loss": 1.0576, "step": 5009 }, { "epoch": 0.42, "learning_rate": 1.2897022202072814e-05, "loss": 1.0714, "step": 5010 }, { "epoch": 0.42, "learning_rate": 1.2894396692612506e-05, "loss": 1.0399, "step": 5011 }, { "epoch": 0.42, "learning_rate": 1.2891770965371555e-05, "loss": 1.0191, "step": 5012 }, { "epoch": 0.42, "learning_rate": 1.2889145020547528e-05, "loss": 1.0517, "step": 5013 }, { "epoch": 0.42, "learning_rate": 1.2886518858338007e-05, "loss": 1.0449, "step": 5014 }, { "epoch": 0.42, "learning_rate": 1.2883892478940587e-05, "loss": 1.1003, "step": 5015 }, { "epoch": 0.42, "learning_rate": 1.2881265882552887e-05, "loss": 1.0567, "step": 5016 }, { "epoch": 0.42, "learning_rate": 1.287863906937253e-05, "loss": 1.0606, "step": 5017 }, { "epoch": 0.42, "learning_rate": 1.2876012039597172e-05, "loss": 1.0386, "step": 5018 }, { "epoch": 0.43, "learning_rate": 1.287338479342447e-05, "loss": 1.0878, "step": 5019 }, { "epoch": 0.43, "learning_rate": 1.287075733105211e-05, "loss": 0.9665, "step": 5020 }, { "epoch": 0.43, "learning_rate": 1.2868129652677784e-05, "loss": 1.0782, "step": 5021 }, { "epoch": 0.43, "learning_rate": 1.2865501758499202e-05, "loss": 1.0811, "step": 5022 }, { "epoch": 0.43, "learning_rate": 1.2862873648714096e-05, "loss": 0.8927, "step": 5023 }, { "epoch": 0.43, "learning_rate": 1.2860245323520208e-05, "loss": 1.0092, "step": 5024 }, { "epoch": 0.43, "learning_rate": 1.2857616783115301e-05, "loss": 1.1084, "step": 5025 }, { "epoch": 0.43, "learning_rate": 1.2854988027697151e-05, "loss": 1.0576, "step": 5026 }, { "epoch": 0.43, "learning_rate": 1.285235905746355e-05, "loss": 1.0907, "step": 5027 }, { "epoch": 0.43, "learning_rate": 1.2849729872612312e-05, "loss": 1.0629, "step": 5028 }, { "epoch": 0.43, "learning_rate": 1.2847100473341257e-05, "loss": 1.0625, "step": 5029 }, { "epoch": 0.43, "learning_rate": 1.284447085984823e-05, "loss": 1.0979, "step": 5030 }, { "epoch": 0.43, "learning_rate": 1.2841841032331088e-05, "loss": 1.062, "step": 5031 }, { "epoch": 0.43, "learning_rate": 1.2839210990987706e-05, "loss": 1.026, "step": 5032 }, { "epoch": 0.43, "learning_rate": 1.283658073601597e-05, "loss": 1.0573, "step": 5033 }, { "epoch": 0.43, "learning_rate": 1.2833950267613788e-05, "loss": 1.0769, "step": 5034 }, { "epoch": 0.43, "learning_rate": 1.2831319585979088e-05, "loss": 1.0188, "step": 5035 }, { "epoch": 0.43, "learning_rate": 1.28286886913098e-05, "loss": 1.013, "step": 5036 }, { "epoch": 0.43, "learning_rate": 1.282605758380388e-05, "loss": 1.108, "step": 5037 }, { "epoch": 0.43, "learning_rate": 1.2823426263659299e-05, "loss": 1.072, "step": 5038 }, { "epoch": 0.43, "learning_rate": 1.2820794731074047e-05, "loss": 1.061, "step": 5039 }, { "epoch": 0.43, "learning_rate": 1.2818162986246119e-05, "loss": 1.0402, "step": 5040 }, { "epoch": 0.43, "learning_rate": 1.281553102937354e-05, "loss": 0.9064, "step": 5041 }, { "epoch": 0.43, "learning_rate": 1.281289886065434e-05, "loss": 0.9966, "step": 5042 }, { "epoch": 0.43, "learning_rate": 1.2810266480286567e-05, "loss": 1.0692, "step": 5043 }, { "epoch": 0.43, "learning_rate": 1.2807633888468292e-05, "loss": 1.08, "step": 5044 }, { "epoch": 0.43, "learning_rate": 1.2805001085397595e-05, "loss": 1.0062, "step": 5045 }, { "epoch": 0.43, "learning_rate": 1.2802368071272578e-05, "loss": 1.0779, "step": 5046 }, { "epoch": 0.43, "learning_rate": 1.2799734846291348e-05, "loss": 1.0779, "step": 5047 }, { "epoch": 0.43, "learning_rate": 1.2797101410652033e-05, "loss": 1.1018, "step": 5048 }, { "epoch": 0.43, "learning_rate": 1.2794467764552784e-05, "loss": 1.0561, "step": 5049 }, { "epoch": 0.43, "learning_rate": 1.2791833908191762e-05, "loss": 1.0117, "step": 5050 }, { "epoch": 0.43, "learning_rate": 1.2789199841767142e-05, "loss": 1.0897, "step": 5051 }, { "epoch": 0.43, "learning_rate": 1.2786565565477118e-05, "loss": 1.0791, "step": 5052 }, { "epoch": 0.43, "learning_rate": 1.2783931079519896e-05, "loss": 1.0479, "step": 5053 }, { "epoch": 0.43, "learning_rate": 1.2781296384093702e-05, "loss": 1.0306, "step": 5054 }, { "epoch": 0.43, "learning_rate": 1.2778661479396779e-05, "loss": 1.0698, "step": 5055 }, { "epoch": 0.43, "learning_rate": 1.2776026365627378e-05, "loss": 1.0875, "step": 5056 }, { "epoch": 0.43, "learning_rate": 1.2773391042983775e-05, "loss": 1.0523, "step": 5057 }, { "epoch": 0.43, "learning_rate": 1.2770755511664255e-05, "loss": 1.0332, "step": 5058 }, { "epoch": 0.43, "learning_rate": 1.276811977186712e-05, "loss": 1.0492, "step": 5059 }, { "epoch": 0.43, "learning_rate": 1.2765483823790692e-05, "loss": 1.0652, "step": 5060 }, { "epoch": 0.43, "learning_rate": 1.2762847667633305e-05, "loss": 1.0718, "step": 5061 }, { "epoch": 0.43, "learning_rate": 1.2760211303593305e-05, "loss": 1.0426, "step": 5062 }, { "epoch": 0.43, "learning_rate": 1.275757473186906e-05, "loss": 1.0397, "step": 5063 }, { "epoch": 0.43, "learning_rate": 1.2754937952658955e-05, "loss": 1.0101, "step": 5064 }, { "epoch": 0.43, "learning_rate": 1.275230096616138e-05, "loss": 1.0848, "step": 5065 }, { "epoch": 0.43, "learning_rate": 1.2749663772574751e-05, "loss": 1.0578, "step": 5066 }, { "epoch": 0.43, "learning_rate": 1.2747026372097499e-05, "loss": 1.0721, "step": 5067 }, { "epoch": 0.43, "learning_rate": 1.2744388764928063e-05, "loss": 1.0721, "step": 5068 }, { "epoch": 0.43, "learning_rate": 1.2741750951264902e-05, "loss": 1.0719, "step": 5069 }, { "epoch": 0.43, "learning_rate": 1.2739112931306497e-05, "loss": 1.0737, "step": 5070 }, { "epoch": 0.43, "learning_rate": 1.273647470525133e-05, "loss": 1.0194, "step": 5071 }, { "epoch": 0.43, "learning_rate": 1.2733836273297916e-05, "loss": 1.0987, "step": 5072 }, { "epoch": 0.43, "learning_rate": 1.2731197635644774e-05, "loss": 1.0774, "step": 5073 }, { "epoch": 0.43, "learning_rate": 1.2728558792490433e-05, "loss": 1.0271, "step": 5074 }, { "epoch": 0.43, "learning_rate": 1.2725919744033453e-05, "loss": 1.1534, "step": 5075 }, { "epoch": 0.43, "learning_rate": 1.2723280490472396e-05, "loss": 1.0598, "step": 5076 }, { "epoch": 0.43, "learning_rate": 1.2720641032005852e-05, "loss": 1.0316, "step": 5077 }, { "epoch": 0.43, "learning_rate": 1.2718001368832413e-05, "loss": 1.0714, "step": 5078 }, { "epoch": 0.43, "learning_rate": 1.2715361501150696e-05, "loss": 1.1077, "step": 5079 }, { "epoch": 0.43, "learning_rate": 1.2712721429159331e-05, "loss": 1.0318, "step": 5080 }, { "epoch": 0.43, "learning_rate": 1.2710081153056961e-05, "loss": 1.0703, "step": 5081 }, { "epoch": 0.43, "learning_rate": 1.2707440673042248e-05, "loss": 1.0643, "step": 5082 }, { "epoch": 0.43, "learning_rate": 1.2704799989313865e-05, "loss": 1.0578, "step": 5083 }, { "epoch": 0.43, "learning_rate": 1.2702159102070506e-05, "loss": 1.0638, "step": 5084 }, { "epoch": 0.43, "learning_rate": 1.2699518011510875e-05, "loss": 1.0968, "step": 5085 }, { "epoch": 0.43, "learning_rate": 1.2696876717833694e-05, "loss": 1.0196, "step": 5086 }, { "epoch": 0.43, "learning_rate": 1.2694235221237695e-05, "loss": 1.0853, "step": 5087 }, { "epoch": 0.43, "learning_rate": 1.2691593521921639e-05, "loss": 0.9983, "step": 5088 }, { "epoch": 0.43, "learning_rate": 1.2688951620084285e-05, "loss": 1.0666, "step": 5089 }, { "epoch": 0.43, "learning_rate": 1.2686309515924418e-05, "loss": 1.0633, "step": 5090 }, { "epoch": 0.43, "learning_rate": 1.2683667209640837e-05, "loss": 1.0349, "step": 5091 }, { "epoch": 0.43, "learning_rate": 1.2681024701432354e-05, "loss": 1.0527, "step": 5092 }, { "epoch": 0.43, "learning_rate": 1.2678381991497796e-05, "loss": 1.0844, "step": 5093 }, { "epoch": 0.43, "learning_rate": 1.2675739080036009e-05, "loss": 1.058, "step": 5094 }, { "epoch": 0.43, "learning_rate": 1.2673095967245846e-05, "loss": 1.0828, "step": 5095 }, { "epoch": 0.43, "learning_rate": 1.2670452653326186e-05, "loss": 1.0631, "step": 5096 }, { "epoch": 0.43, "learning_rate": 1.2667809138475915e-05, "loss": 0.9709, "step": 5097 }, { "epoch": 0.43, "learning_rate": 1.266516542289394e-05, "loss": 1.0474, "step": 5098 }, { "epoch": 0.43, "learning_rate": 1.2662521506779176e-05, "loss": 1.0513, "step": 5099 }, { "epoch": 0.43, "learning_rate": 1.2659877390330557e-05, "loss": 1.0337, "step": 5100 }, { "epoch": 0.43, "learning_rate": 1.2657233073747035e-05, "loss": 1.0447, "step": 5101 }, { "epoch": 0.43, "learning_rate": 1.265458855722757e-05, "loss": 1.0385, "step": 5102 }, { "epoch": 0.43, "learning_rate": 1.2651943840971143e-05, "loss": 1.0396, "step": 5103 }, { "epoch": 0.43, "learning_rate": 1.264929892517675e-05, "loss": 1.0808, "step": 5104 }, { "epoch": 0.43, "learning_rate": 1.2646653810043398e-05, "loss": 1.0202, "step": 5105 }, { "epoch": 0.43, "learning_rate": 1.2644008495770112e-05, "loss": 1.0448, "step": 5106 }, { "epoch": 0.43, "learning_rate": 1.2641362982555932e-05, "loss": 1.0937, "step": 5107 }, { "epoch": 0.43, "learning_rate": 1.263871727059991e-05, "loss": 1.0593, "step": 5108 }, { "epoch": 0.43, "learning_rate": 1.2636071360101116e-05, "loss": 1.1324, "step": 5109 }, { "epoch": 0.43, "learning_rate": 1.2633425251258632e-05, "loss": 1.097, "step": 5110 }, { "epoch": 0.43, "learning_rate": 1.2630778944271561e-05, "loss": 1.083, "step": 5111 }, { "epoch": 0.43, "learning_rate": 1.2628132439339018e-05, "loss": 1.0755, "step": 5112 }, { "epoch": 0.43, "learning_rate": 1.2625485736660126e-05, "loss": 1.1126, "step": 5113 }, { "epoch": 0.43, "learning_rate": 1.262283883643403e-05, "loss": 1.0443, "step": 5114 }, { "epoch": 0.43, "learning_rate": 1.2620191738859893e-05, "loss": 1.1147, "step": 5115 }, { "epoch": 0.43, "learning_rate": 1.2617544444136884e-05, "loss": 1.0583, "step": 5116 }, { "epoch": 0.43, "learning_rate": 1.2614896952464191e-05, "loss": 0.9863, "step": 5117 }, { "epoch": 0.43, "learning_rate": 1.2612249264041019e-05, "loss": 1.078, "step": 5118 }, { "epoch": 0.43, "learning_rate": 1.2609601379066583e-05, "loss": 1.0817, "step": 5119 }, { "epoch": 0.43, "learning_rate": 1.2606953297740119e-05, "loss": 1.0628, "step": 5120 }, { "epoch": 0.43, "learning_rate": 1.260430502026087e-05, "loss": 1.1203, "step": 5121 }, { "epoch": 0.43, "learning_rate": 1.2601656546828106e-05, "loss": 1.0553, "step": 5122 }, { "epoch": 0.43, "learning_rate": 1.2599007877641094e-05, "loss": 1.0273, "step": 5123 }, { "epoch": 0.43, "learning_rate": 1.2596359012899132e-05, "loss": 1.0572, "step": 5124 }, { "epoch": 0.43, "learning_rate": 1.2593709952801527e-05, "loss": 1.0723, "step": 5125 }, { "epoch": 0.43, "learning_rate": 1.2591060697547596e-05, "loss": 1.0684, "step": 5126 }, { "epoch": 0.43, "learning_rate": 1.2588411247336678e-05, "loss": 1.0023, "step": 5127 }, { "epoch": 0.43, "learning_rate": 1.258576160236812e-05, "loss": 0.8959, "step": 5128 }, { "epoch": 0.43, "learning_rate": 1.2583111762841291e-05, "loss": 1.0382, "step": 5129 }, { "epoch": 0.43, "learning_rate": 1.2580461728955566e-05, "loss": 1.0382, "step": 5130 }, { "epoch": 0.43, "learning_rate": 1.2577811500910342e-05, "loss": 1.0312, "step": 5131 }, { "epoch": 0.43, "learning_rate": 1.2575161078905027e-05, "loss": 1.0832, "step": 5132 }, { "epoch": 0.43, "learning_rate": 1.2572510463139045e-05, "loss": 1.0277, "step": 5133 }, { "epoch": 0.43, "learning_rate": 1.2569859653811836e-05, "loss": 1.0312, "step": 5134 }, { "epoch": 0.43, "learning_rate": 1.2567208651122848e-05, "loss": 1.0717, "step": 5135 }, { "epoch": 0.43, "learning_rate": 1.2564557455271551e-05, "loss": 1.1012, "step": 5136 }, { "epoch": 0.44, "learning_rate": 1.2561906066457428e-05, "loss": 1.0093, "step": 5137 }, { "epoch": 0.44, "learning_rate": 1.2559254484879973e-05, "loss": 1.0914, "step": 5138 }, { "epoch": 0.44, "learning_rate": 1.2556602710738695e-05, "loss": 0.8495, "step": 5139 }, { "epoch": 0.44, "learning_rate": 1.2553950744233123e-05, "loss": 1.0269, "step": 5140 }, { "epoch": 0.44, "learning_rate": 1.2551298585562796e-05, "loss": 1.0016, "step": 5141 }, { "epoch": 0.44, "learning_rate": 1.254864623492727e-05, "loss": 1.0978, "step": 5142 }, { "epoch": 0.44, "learning_rate": 1.2545993692526107e-05, "loss": 1.0147, "step": 5143 }, { "epoch": 0.44, "learning_rate": 1.2543340958558896e-05, "loss": 1.0494, "step": 5144 }, { "epoch": 0.44, "learning_rate": 1.254068803322523e-05, "loss": 1.1134, "step": 5145 }, { "epoch": 0.44, "learning_rate": 1.2538034916724723e-05, "loss": 1.0764, "step": 5146 }, { "epoch": 0.44, "learning_rate": 1.2535381609257002e-05, "loss": 1.0835, "step": 5147 }, { "epoch": 0.44, "learning_rate": 1.2532728111021707e-05, "loss": 1.0836, "step": 5148 }, { "epoch": 0.44, "learning_rate": 1.2530074422218492e-05, "loss": 1.0843, "step": 5149 }, { "epoch": 0.44, "learning_rate": 1.2527420543047028e-05, "loss": 1.0586, "step": 5150 }, { "epoch": 0.44, "learning_rate": 1.2524766473707e-05, "loss": 1.0639, "step": 5151 }, { "epoch": 0.44, "learning_rate": 1.25221122143981e-05, "loss": 1.1248, "step": 5152 }, { "epoch": 0.44, "learning_rate": 1.2519457765320046e-05, "loss": 1.0213, "step": 5153 }, { "epoch": 0.44, "learning_rate": 1.2516803126672563e-05, "loss": 1.0909, "step": 5154 }, { "epoch": 0.44, "learning_rate": 1.2514148298655392e-05, "loss": 1.0854, "step": 5155 }, { "epoch": 0.44, "learning_rate": 1.2511493281468284e-05, "loss": 1.0978, "step": 5156 }, { "epoch": 0.44, "learning_rate": 1.2508838075311009e-05, "loss": 1.0572, "step": 5157 }, { "epoch": 0.44, "learning_rate": 1.2506182680383357e-05, "loss": 1.0582, "step": 5158 }, { "epoch": 0.44, "learning_rate": 1.2503527096885118e-05, "loss": 1.0862, "step": 5159 }, { "epoch": 0.44, "learning_rate": 1.2500871325016108e-05, "loss": 1.0832, "step": 5160 }, { "epoch": 0.44, "learning_rate": 1.2498215364976152e-05, "loss": 1.0246, "step": 5161 }, { "epoch": 0.44, "learning_rate": 1.2495559216965091e-05, "loss": 1.0704, "step": 5162 }, { "epoch": 0.44, "learning_rate": 1.2492902881182779e-05, "loss": 1.0951, "step": 5163 }, { "epoch": 0.44, "learning_rate": 1.2490246357829084e-05, "loss": 1.0888, "step": 5164 }, { "epoch": 0.44, "learning_rate": 1.2487589647103887e-05, "loss": 1.0867, "step": 5165 }, { "epoch": 0.44, "learning_rate": 1.2484932749207087e-05, "loss": 1.0981, "step": 5166 }, { "epoch": 0.44, "learning_rate": 1.2482275664338592e-05, "loss": 1.0571, "step": 5167 }, { "epoch": 0.44, "learning_rate": 1.247961839269833e-05, "loss": 1.0583, "step": 5168 }, { "epoch": 0.44, "learning_rate": 1.247696093448624e-05, "loss": 1.0745, "step": 5169 }, { "epoch": 0.44, "learning_rate": 1.2474303289902273e-05, "loss": 1.0882, "step": 5170 }, { "epoch": 0.44, "learning_rate": 1.2471645459146394e-05, "loss": 1.0645, "step": 5171 }, { "epoch": 0.44, "learning_rate": 1.2468987442418588e-05, "loss": 1.0324, "step": 5172 }, { "epoch": 0.44, "learning_rate": 1.2466329239918844e-05, "loss": 1.1021, "step": 5173 }, { "epoch": 0.44, "learning_rate": 1.2463670851847178e-05, "loss": 1.0911, "step": 5174 }, { "epoch": 0.44, "learning_rate": 1.246101227840361e-05, "loss": 0.9981, "step": 5175 }, { "epoch": 0.44, "learning_rate": 1.2458353519788176e-05, "loss": 1.0484, "step": 5176 }, { "epoch": 0.44, "learning_rate": 1.2455694576200927e-05, "loss": 0.8651, "step": 5177 }, { "epoch": 0.44, "learning_rate": 1.245303544784193e-05, "loss": 0.9213, "step": 5178 }, { "epoch": 0.44, "learning_rate": 1.2450376134911259e-05, "loss": 1.0292, "step": 5179 }, { "epoch": 0.44, "learning_rate": 1.2447716637609013e-05, "loss": 1.0281, "step": 5180 }, { "epoch": 0.44, "learning_rate": 1.244505695613529e-05, "loss": 1.0592, "step": 5181 }, { "epoch": 0.44, "learning_rate": 1.2442397090690217e-05, "loss": 1.0846, "step": 5182 }, { "epoch": 0.44, "learning_rate": 1.2439737041473923e-05, "loss": 0.9251, "step": 5183 }, { "epoch": 0.44, "learning_rate": 1.2437076808686559e-05, "loss": 1.027, "step": 5184 }, { "epoch": 0.44, "learning_rate": 1.2434416392528286e-05, "loss": 1.0484, "step": 5185 }, { "epoch": 0.44, "learning_rate": 1.243175579319928e-05, "loss": 1.0437, "step": 5186 }, { "epoch": 0.44, "learning_rate": 1.242909501089973e-05, "loss": 1.0482, "step": 5187 }, { "epoch": 0.44, "learning_rate": 1.2426434045829836e-05, "loss": 1.056, "step": 5188 }, { "epoch": 0.44, "learning_rate": 1.2423772898189819e-05, "loss": 1.0334, "step": 5189 }, { "epoch": 0.44, "learning_rate": 1.2421111568179906e-05, "loss": 1.0673, "step": 5190 }, { "epoch": 0.44, "learning_rate": 1.2418450056000346e-05, "loss": 1.061, "step": 5191 }, { "epoch": 0.44, "learning_rate": 1.2415788361851394e-05, "loss": 0.9808, "step": 5192 }, { "epoch": 0.44, "learning_rate": 1.2413126485933318e-05, "loss": 1.0545, "step": 5193 }, { "epoch": 0.44, "learning_rate": 1.241046442844641e-05, "loss": 1.008, "step": 5194 }, { "epoch": 0.44, "learning_rate": 1.2407802189590965e-05, "loss": 1.028, "step": 5195 }, { "epoch": 0.44, "learning_rate": 1.2405139769567296e-05, "loss": 1.0985, "step": 5196 }, { "epoch": 0.44, "learning_rate": 1.2402477168575729e-05, "loss": 1.0482, "step": 5197 }, { "epoch": 0.44, "learning_rate": 1.2399814386816604e-05, "loss": 1.0222, "step": 5198 }, { "epoch": 0.44, "learning_rate": 1.2397151424490272e-05, "loss": 1.1122, "step": 5199 }, { "epoch": 0.44, "learning_rate": 1.2394488281797106e-05, "loss": 1.0718, "step": 5200 }, { "epoch": 0.44, "learning_rate": 1.2391824958937483e-05, "loss": 1.0779, "step": 5201 }, { "epoch": 0.44, "learning_rate": 1.2389161456111797e-05, "loss": 1.0355, "step": 5202 }, { "epoch": 0.44, "learning_rate": 1.2386497773520458e-05, "loss": 1.0797, "step": 5203 }, { "epoch": 0.44, "learning_rate": 1.2383833911363885e-05, "loss": 0.9782, "step": 5204 }, { "epoch": 0.44, "learning_rate": 1.2381169869842512e-05, "loss": 1.0271, "step": 5205 }, { "epoch": 0.44, "learning_rate": 1.2378505649156792e-05, "loss": 1.0741, "step": 5206 }, { "epoch": 0.44, "learning_rate": 1.2375841249507181e-05, "loss": 1.0619, "step": 5207 }, { "epoch": 0.44, "learning_rate": 1.2373176671094153e-05, "loss": 1.053, "step": 5208 }, { "epoch": 0.44, "learning_rate": 1.2370511914118204e-05, "loss": 1.0398, "step": 5209 }, { "epoch": 0.44, "learning_rate": 1.236784697877983e-05, "loss": 1.0655, "step": 5210 }, { "epoch": 0.44, "learning_rate": 1.2365181865279548e-05, "loss": 1.0858, "step": 5211 }, { "epoch": 0.44, "learning_rate": 1.2362516573817889e-05, "loss": 1.0819, "step": 5212 }, { "epoch": 0.44, "learning_rate": 1.2359851104595394e-05, "loss": 1.0079, "step": 5213 }, { "epoch": 0.44, "learning_rate": 1.2357185457812616e-05, "loss": 1.0441, "step": 5214 }, { "epoch": 0.44, "learning_rate": 1.2354519633670125e-05, "loss": 1.0974, "step": 5215 }, { "epoch": 0.44, "learning_rate": 1.2351853632368507e-05, "loss": 1.0787, "step": 5216 }, { "epoch": 0.44, "learning_rate": 1.2349187454108354e-05, "loss": 1.0512, "step": 5217 }, { "epoch": 0.44, "learning_rate": 1.2346521099090278e-05, "loss": 1.0964, "step": 5218 }, { "epoch": 0.44, "learning_rate": 1.2343854567514897e-05, "loss": 1.0502, "step": 5219 }, { "epoch": 0.44, "learning_rate": 1.234118785958285e-05, "loss": 1.073, "step": 5220 }, { "epoch": 0.44, "learning_rate": 1.2338520975494788e-05, "loss": 1.0636, "step": 5221 }, { "epoch": 0.44, "learning_rate": 1.2335853915451367e-05, "loss": 1.0748, "step": 5222 }, { "epoch": 0.44, "learning_rate": 1.2333186679653268e-05, "loss": 1.0567, "step": 5223 }, { "epoch": 0.44, "learning_rate": 1.2330519268301176e-05, "loss": 1.0235, "step": 5224 }, { "epoch": 0.44, "learning_rate": 1.2327851681595788e-05, "loss": 1.0267, "step": 5225 }, { "epoch": 0.44, "learning_rate": 1.2325183919737831e-05, "loss": 1.0204, "step": 5226 }, { "epoch": 0.44, "learning_rate": 1.2322515982928025e-05, "loss": 1.0753, "step": 5227 }, { "epoch": 0.44, "learning_rate": 1.2319847871367115e-05, "loss": 1.013, "step": 5228 }, { "epoch": 0.44, "learning_rate": 1.231717958525585e-05, "loss": 1.0632, "step": 5229 }, { "epoch": 0.44, "learning_rate": 1.2314511124795003e-05, "loss": 1.0701, "step": 5230 }, { "epoch": 0.44, "learning_rate": 1.2311842490185356e-05, "loss": 1.0366, "step": 5231 }, { "epoch": 0.44, "learning_rate": 1.2309173681627696e-05, "loss": 1.0933, "step": 5232 }, { "epoch": 0.44, "learning_rate": 1.2306504699322835e-05, "loss": 1.0893, "step": 5233 }, { "epoch": 0.44, "learning_rate": 1.2303835543471594e-05, "loss": 1.059, "step": 5234 }, { "epoch": 0.44, "learning_rate": 1.2301166214274801e-05, "loss": 1.0237, "step": 5235 }, { "epoch": 0.44, "learning_rate": 1.2298496711933305e-05, "loss": 1.0682, "step": 5236 }, { "epoch": 0.44, "learning_rate": 1.2295827036647965e-05, "loss": 1.0646, "step": 5237 }, { "epoch": 0.44, "learning_rate": 1.2293157188619653e-05, "loss": 0.9998, "step": 5238 }, { "epoch": 0.44, "learning_rate": 1.2290487168049254e-05, "loss": 1.0431, "step": 5239 }, { "epoch": 0.44, "learning_rate": 1.2287816975137665e-05, "loss": 1.0586, "step": 5240 }, { "epoch": 0.44, "learning_rate": 1.22851466100858e-05, "loss": 1.105, "step": 5241 }, { "epoch": 0.44, "learning_rate": 1.228247607309458e-05, "loss": 1.0453, "step": 5242 }, { "epoch": 0.44, "learning_rate": 1.2279805364364945e-05, "loss": 1.1254, "step": 5243 }, { "epoch": 0.44, "learning_rate": 1.2277134484097842e-05, "loss": 1.1019, "step": 5244 }, { "epoch": 0.44, "learning_rate": 1.2274463432494235e-05, "loss": 1.1149, "step": 5245 }, { "epoch": 0.44, "learning_rate": 1.22717922097551e-05, "loss": 1.0301, "step": 5246 }, { "epoch": 0.44, "learning_rate": 1.2269120816081425e-05, "loss": 1.0169, "step": 5247 }, { "epoch": 0.44, "learning_rate": 1.2266449251674213e-05, "loss": 1.0321, "step": 5248 }, { "epoch": 0.44, "learning_rate": 1.2263777516734477e-05, "loss": 1.0179, "step": 5249 }, { "epoch": 0.44, "learning_rate": 1.2261105611463243e-05, "loss": 1.1378, "step": 5250 }, { "epoch": 0.44, "learning_rate": 1.2258433536061546e-05, "loss": 1.0332, "step": 5251 }, { "epoch": 0.44, "learning_rate": 1.225576129073045e-05, "loss": 1.049, "step": 5252 }, { "epoch": 0.44, "learning_rate": 1.2253088875671016e-05, "loss": 0.9965, "step": 5253 }, { "epoch": 0.44, "learning_rate": 1.225041629108432e-05, "loss": 1.0217, "step": 5254 }, { "epoch": 0.45, "learning_rate": 1.2247743537171455e-05, "loss": 1.0112, "step": 5255 }, { "epoch": 0.45, "learning_rate": 1.2245070614133522e-05, "loss": 1.0495, "step": 5256 }, { "epoch": 0.45, "learning_rate": 1.2242397522171641e-05, "loss": 1.0768, "step": 5257 }, { "epoch": 0.45, "learning_rate": 1.2239724261486941e-05, "loss": 1.0422, "step": 5258 }, { "epoch": 0.45, "learning_rate": 1.223705083228056e-05, "loss": 1.0412, "step": 5259 }, { "epoch": 0.45, "learning_rate": 1.2234377234753656e-05, "loss": 1.0396, "step": 5260 }, { "epoch": 0.45, "learning_rate": 1.2231703469107399e-05, "loss": 1.0507, "step": 5261 }, { "epoch": 0.45, "learning_rate": 1.2229029535542963e-05, "loss": 1.0173, "step": 5262 }, { "epoch": 0.45, "learning_rate": 1.2226355434261543e-05, "loss": 1.0264, "step": 5263 }, { "epoch": 0.45, "learning_rate": 1.2223681165464343e-05, "loss": 1.0893, "step": 5264 }, { "epoch": 0.45, "learning_rate": 1.2221006729352583e-05, "loss": 1.0703, "step": 5265 }, { "epoch": 0.45, "learning_rate": 1.221833212612749e-05, "loss": 1.0725, "step": 5266 }, { "epoch": 0.45, "learning_rate": 1.2215657355990312e-05, "loss": 1.069, "step": 5267 }, { "epoch": 0.45, "learning_rate": 1.2212982419142298e-05, "loss": 1.0803, "step": 5268 }, { "epoch": 0.45, "learning_rate": 1.2210307315784721e-05, "loss": 0.8124, "step": 5269 }, { "epoch": 0.45, "learning_rate": 1.2207632046118861e-05, "loss": 1.0738, "step": 5270 }, { "epoch": 0.45, "learning_rate": 1.2204956610346008e-05, "loss": 1.0862, "step": 5271 }, { "epoch": 0.45, "learning_rate": 1.2202281008667471e-05, "loss": 1.0545, "step": 5272 }, { "epoch": 0.45, "learning_rate": 1.2199605241284567e-05, "loss": 0.9342, "step": 5273 }, { "epoch": 0.45, "learning_rate": 1.2196929308398626e-05, "loss": 1.0301, "step": 5274 }, { "epoch": 0.45, "learning_rate": 1.2194253210210991e-05, "loss": 1.0343, "step": 5275 }, { "epoch": 0.45, "learning_rate": 1.219157694692302e-05, "loss": 1.0686, "step": 5276 }, { "epoch": 0.45, "learning_rate": 1.2188900518736073e-05, "loss": 1.0838, "step": 5277 }, { "epoch": 0.45, "learning_rate": 1.218622392585154e-05, "loss": 1.1112, "step": 5278 }, { "epoch": 0.45, "learning_rate": 1.2183547168470806e-05, "loss": 1.0405, "step": 5279 }, { "epoch": 0.45, "learning_rate": 1.2180870246795282e-05, "loss": 1.0668, "step": 5280 }, { "epoch": 0.45, "learning_rate": 1.2178193161026381e-05, "loss": 1.0361, "step": 5281 }, { "epoch": 0.45, "learning_rate": 1.2175515911365535e-05, "loss": 1.0475, "step": 5282 }, { "epoch": 0.45, "learning_rate": 1.2172838498014184e-05, "loss": 1.0712, "step": 5283 }, { "epoch": 0.45, "learning_rate": 1.2170160921173783e-05, "loss": 1.051, "step": 5284 }, { "epoch": 0.45, "learning_rate": 1.2167483181045803e-05, "loss": 1.0508, "step": 5285 }, { "epoch": 0.45, "learning_rate": 1.2164805277831716e-05, "loss": 1.0479, "step": 5286 }, { "epoch": 0.45, "learning_rate": 1.2162127211733016e-05, "loss": 0.9917, "step": 5287 }, { "epoch": 0.45, "learning_rate": 1.215944898295121e-05, "loss": 1.0449, "step": 5288 }, { "epoch": 0.45, "learning_rate": 1.215677059168781e-05, "loss": 1.0112, "step": 5289 }, { "epoch": 0.45, "learning_rate": 1.2154092038144342e-05, "loss": 1.0742, "step": 5290 }, { "epoch": 0.45, "learning_rate": 1.215141332252235e-05, "loss": 0.9968, "step": 5291 }, { "epoch": 0.45, "learning_rate": 1.2148734445023385e-05, "loss": 1.0003, "step": 5292 }, { "epoch": 0.45, "learning_rate": 1.214605540584901e-05, "loss": 1.028, "step": 5293 }, { "epoch": 0.45, "learning_rate": 1.2143376205200805e-05, "loss": 1.0382, "step": 5294 }, { "epoch": 0.45, "learning_rate": 1.2140696843280355e-05, "loss": 1.0877, "step": 5295 }, { "epoch": 0.45, "learning_rate": 1.2138017320289266e-05, "loss": 1.1221, "step": 5296 }, { "epoch": 0.45, "learning_rate": 1.2135337636429145e-05, "loss": 1.0553, "step": 5297 }, { "epoch": 0.45, "learning_rate": 1.2132657791901621e-05, "loss": 1.053, "step": 5298 }, { "epoch": 0.45, "learning_rate": 1.212997778690833e-05, "loss": 1.0513, "step": 5299 }, { "epoch": 0.45, "learning_rate": 1.2127297621650924e-05, "loss": 1.032, "step": 5300 }, { "epoch": 0.45, "learning_rate": 1.212461729633106e-05, "loss": 1.0488, "step": 5301 }, { "epoch": 0.45, "learning_rate": 1.2121936811150416e-05, "loss": 1.077, "step": 5302 }, { "epoch": 0.45, "learning_rate": 1.2119256166310673e-05, "loss": 1.0556, "step": 5303 }, { "epoch": 0.45, "learning_rate": 1.2116575362013532e-05, "loss": 1.0854, "step": 5304 }, { "epoch": 0.45, "learning_rate": 1.2113894398460702e-05, "loss": 1.0585, "step": 5305 }, { "epoch": 0.45, "learning_rate": 1.2111213275853902e-05, "loss": 0.9956, "step": 5306 }, { "epoch": 0.45, "learning_rate": 1.2108531994394865e-05, "loss": 1.0065, "step": 5307 }, { "epoch": 0.45, "learning_rate": 1.210585055428534e-05, "loss": 1.0411, "step": 5308 }, { "epoch": 0.45, "learning_rate": 1.2103168955727085e-05, "loss": 1.0149, "step": 5309 }, { "epoch": 0.45, "learning_rate": 1.2100487198921866e-05, "loss": 1.0635, "step": 5310 }, { "epoch": 0.45, "learning_rate": 1.2097805284071464e-05, "loss": 1.0699, "step": 5311 }, { "epoch": 0.45, "learning_rate": 1.2095123211377675e-05, "loss": 0.9775, "step": 5312 }, { "epoch": 0.45, "learning_rate": 1.2092440981042302e-05, "loss": 1.0426, "step": 5313 }, { "epoch": 0.45, "learning_rate": 1.2089758593267163e-05, "loss": 1.1258, "step": 5314 }, { "epoch": 0.45, "learning_rate": 1.2087076048254082e-05, "loss": 1.031, "step": 5315 }, { "epoch": 0.45, "learning_rate": 1.2084393346204905e-05, "loss": 1.0447, "step": 5316 }, { "epoch": 0.45, "learning_rate": 1.2081710487321482e-05, "loss": 1.071, "step": 5317 }, { "epoch": 0.45, "learning_rate": 1.207902747180568e-05, "loss": 1.0716, "step": 5318 }, { "epoch": 0.45, "learning_rate": 1.2076344299859369e-05, "loss": 1.0888, "step": 5319 }, { "epoch": 0.45, "learning_rate": 1.207366097168444e-05, "loss": 1.1192, "step": 5320 }, { "epoch": 0.45, "learning_rate": 1.2070977487482793e-05, "loss": 1.0631, "step": 5321 }, { "epoch": 0.45, "learning_rate": 1.206829384745634e-05, "loss": 1.0472, "step": 5322 }, { "epoch": 0.45, "learning_rate": 1.2065610051807002e-05, "loss": 1.0795, "step": 5323 }, { "epoch": 0.45, "learning_rate": 1.2062926100736712e-05, "loss": 0.9775, "step": 5324 }, { "epoch": 0.45, "learning_rate": 1.206024199444742e-05, "loss": 1.0631, "step": 5325 }, { "epoch": 0.45, "learning_rate": 1.205755773314108e-05, "loss": 1.0227, "step": 5326 }, { "epoch": 0.45, "learning_rate": 1.2054873317019665e-05, "loss": 1.0169, "step": 5327 }, { "epoch": 0.45, "learning_rate": 1.2052188746285155e-05, "loss": 1.0582, "step": 5328 }, { "epoch": 0.45, "learning_rate": 1.2049504021139542e-05, "loss": 1.0409, "step": 5329 }, { "epoch": 0.45, "learning_rate": 1.2046819141784831e-05, "loss": 1.0618, "step": 5330 }, { "epoch": 0.45, "learning_rate": 1.204413410842304e-05, "loss": 0.9928, "step": 5331 }, { "epoch": 0.45, "learning_rate": 1.2041448921256196e-05, "loss": 1.0677, "step": 5332 }, { "epoch": 0.45, "learning_rate": 1.2038763580486335e-05, "loss": 1.0153, "step": 5333 }, { "epoch": 0.45, "learning_rate": 1.203607808631551e-05, "loss": 1.081, "step": 5334 }, { "epoch": 0.45, "learning_rate": 1.2033392438945786e-05, "loss": 1.0697, "step": 5335 }, { "epoch": 0.45, "learning_rate": 1.2030706638579233e-05, "loss": 1.0485, "step": 5336 }, { "epoch": 0.45, "learning_rate": 1.2028020685417938e-05, "loss": 1.033, "step": 5337 }, { "epoch": 0.45, "learning_rate": 1.2025334579663999e-05, "loss": 1.1, "step": 5338 }, { "epoch": 0.45, "learning_rate": 1.2022648321519521e-05, "loss": 1.0375, "step": 5339 }, { "epoch": 0.45, "learning_rate": 1.2019961911186626e-05, "loss": 1.0281, "step": 5340 }, { "epoch": 0.45, "learning_rate": 1.2017275348867449e-05, "loss": 1.094, "step": 5341 }, { "epoch": 0.45, "learning_rate": 1.2014588634764126e-05, "loss": 1.1198, "step": 5342 }, { "epoch": 0.45, "learning_rate": 1.2011901769078815e-05, "loss": 1.0798, "step": 5343 }, { "epoch": 0.45, "learning_rate": 1.200921475201368e-05, "loss": 0.9872, "step": 5344 }, { "epoch": 0.45, "learning_rate": 1.2006527583770899e-05, "loss": 1.0418, "step": 5345 }, { "epoch": 0.45, "learning_rate": 1.2003840264552662e-05, "loss": 1.0576, "step": 5346 }, { "epoch": 0.45, "learning_rate": 1.2001152794561164e-05, "loss": 1.0542, "step": 5347 }, { "epoch": 0.45, "learning_rate": 1.1998465173998619e-05, "loss": 1.0539, "step": 5348 }, { "epoch": 0.45, "learning_rate": 1.199577740306725e-05, "loss": 1.0396, "step": 5349 }, { "epoch": 0.45, "learning_rate": 1.1993089481969291e-05, "loss": 1.0512, "step": 5350 }, { "epoch": 0.45, "learning_rate": 1.1990401410906984e-05, "loss": 1.0068, "step": 5351 }, { "epoch": 0.45, "learning_rate": 1.1987713190082588e-05, "loss": 1.0498, "step": 5352 }, { "epoch": 0.45, "learning_rate": 1.1985024819698368e-05, "loss": 1.0574, "step": 5353 }, { "epoch": 0.45, "learning_rate": 1.1982336299956605e-05, "loss": 0.9979, "step": 5354 }, { "epoch": 0.45, "learning_rate": 1.197964763105959e-05, "loss": 1.0574, "step": 5355 }, { "epoch": 0.45, "learning_rate": 1.197695881320962e-05, "loss": 1.0802, "step": 5356 }, { "epoch": 0.45, "learning_rate": 1.1974269846609012e-05, "loss": 1.0545, "step": 5357 }, { "epoch": 0.45, "learning_rate": 1.1971580731460088e-05, "loss": 1.0373, "step": 5358 }, { "epoch": 0.45, "learning_rate": 1.1968891467965182e-05, "loss": 1.0414, "step": 5359 }, { "epoch": 0.45, "learning_rate": 1.1966202056326638e-05, "loss": 1.0851, "step": 5360 }, { "epoch": 0.45, "learning_rate": 1.1963512496746818e-05, "loss": 1.0533, "step": 5361 }, { "epoch": 0.45, "learning_rate": 1.1960822789428087e-05, "loss": 1.0651, "step": 5362 }, { "epoch": 0.45, "learning_rate": 1.1958132934572827e-05, "loss": 1.0242, "step": 5363 }, { "epoch": 0.45, "learning_rate": 1.1955442932383425e-05, "loss": 1.0174, "step": 5364 }, { "epoch": 0.45, "learning_rate": 1.1952752783062287e-05, "loss": 1.0966, "step": 5365 }, { "epoch": 0.45, "learning_rate": 1.1950062486811819e-05, "loss": 1.0577, "step": 5366 }, { "epoch": 0.45, "learning_rate": 1.1947372043834452e-05, "loss": 1.0596, "step": 5367 }, { "epoch": 0.45, "learning_rate": 1.1944681454332615e-05, "loss": 1.143, "step": 5368 }, { "epoch": 0.45, "learning_rate": 1.1941990718508758e-05, "loss": 1.0424, "step": 5369 }, { "epoch": 0.45, "learning_rate": 1.1939299836565334e-05, "loss": 1.0373, "step": 5370 }, { "epoch": 0.45, "learning_rate": 1.1936608808704815e-05, "loss": 1.039, "step": 5371 }, { "epoch": 0.45, "learning_rate": 1.1933917635129676e-05, "loss": 1.0526, "step": 5372 }, { "epoch": 0.46, "learning_rate": 1.1931226316042409e-05, "loss": 1.0248, "step": 5373 }, { "epoch": 0.46, "learning_rate": 1.192853485164551e-05, "loss": 1.0796, "step": 5374 }, { "epoch": 0.46, "learning_rate": 1.1925843242141498e-05, "loss": 1.0146, "step": 5375 }, { "epoch": 0.46, "learning_rate": 1.192315148773289e-05, "loss": 1.0739, "step": 5376 }, { "epoch": 0.46, "learning_rate": 1.192045958862222e-05, "loss": 0.9777, "step": 5377 }, { "epoch": 0.46, "learning_rate": 1.1917767545012033e-05, "loss": 1.0522, "step": 5378 }, { "epoch": 0.46, "learning_rate": 1.1915075357104888e-05, "loss": 1.0759, "step": 5379 }, { "epoch": 0.46, "learning_rate": 1.1912383025103344e-05, "loss": 1.0828, "step": 5380 }, { "epoch": 0.46, "learning_rate": 1.1909690549209985e-05, "loss": 1.0038, "step": 5381 }, { "epoch": 0.46, "learning_rate": 1.1906997929627392e-05, "loss": 1.0472, "step": 5382 }, { "epoch": 0.46, "learning_rate": 1.190430516655817e-05, "loss": 1.0391, "step": 5383 }, { "epoch": 0.46, "learning_rate": 1.1901612260204922e-05, "loss": 1.0722, "step": 5384 }, { "epoch": 0.46, "learning_rate": 1.1898919210770272e-05, "loss": 1.0015, "step": 5385 }, { "epoch": 0.46, "learning_rate": 1.189622601845685e-05, "loss": 1.0806, "step": 5386 }, { "epoch": 0.46, "learning_rate": 1.1893532683467299e-05, "loss": 0.9995, "step": 5387 }, { "epoch": 0.46, "learning_rate": 1.1890839206004267e-05, "loss": 1.0937, "step": 5388 }, { "epoch": 0.46, "learning_rate": 1.1888145586270421e-05, "loss": 0.913, "step": 5389 }, { "epoch": 0.46, "learning_rate": 1.1885451824468434e-05, "loss": 1.0836, "step": 5390 }, { "epoch": 0.46, "learning_rate": 1.1882757920800989e-05, "loss": 1.0898, "step": 5391 }, { "epoch": 0.46, "learning_rate": 1.1880063875470785e-05, "loss": 1.0415, "step": 5392 }, { "epoch": 0.46, "learning_rate": 1.187736968868052e-05, "loss": 1.0038, "step": 5393 }, { "epoch": 0.46, "learning_rate": 1.1874675360632919e-05, "loss": 1.0777, "step": 5394 }, { "epoch": 0.46, "learning_rate": 1.1871980891530706e-05, "loss": 1.0511, "step": 5395 }, { "epoch": 0.46, "learning_rate": 1.1869286281576618e-05, "loss": 1.112, "step": 5396 }, { "epoch": 0.46, "learning_rate": 1.1866591530973399e-05, "loss": 1.0602, "step": 5397 }, { "epoch": 0.46, "learning_rate": 1.1863896639923816e-05, "loss": 1.0861, "step": 5398 }, { "epoch": 0.46, "learning_rate": 1.1861201608630632e-05, "loss": 1.1224, "step": 5399 }, { "epoch": 0.46, "learning_rate": 1.1858506437296632e-05, "loss": 1.1265, "step": 5400 }, { "epoch": 0.46, "learning_rate": 1.1855811126124603e-05, "loss": 1.062, "step": 5401 }, { "epoch": 0.46, "learning_rate": 1.1853115675317345e-05, "loss": 0.886, "step": 5402 }, { "epoch": 0.46, "learning_rate": 1.1850420085077675e-05, "loss": 1.0459, "step": 5403 }, { "epoch": 0.46, "learning_rate": 1.184772435560841e-05, "loss": 1.083, "step": 5404 }, { "epoch": 0.46, "learning_rate": 1.1845028487112384e-05, "loss": 1.0506, "step": 5405 }, { "epoch": 0.46, "learning_rate": 1.184233247979244e-05, "loss": 1.0429, "step": 5406 }, { "epoch": 0.46, "learning_rate": 1.1839636333851432e-05, "loss": 1.0132, "step": 5407 }, { "epoch": 0.46, "learning_rate": 1.1836940049492221e-05, "loss": 1.0483, "step": 5408 }, { "epoch": 0.46, "learning_rate": 1.1834243626917685e-05, "loss": 1.1015, "step": 5409 }, { "epoch": 0.46, "learning_rate": 1.1831547066330708e-05, "loss": 1.0292, "step": 5410 }, { "epoch": 0.46, "learning_rate": 1.1828850367934183e-05, "loss": 1.0978, "step": 5411 }, { "epoch": 0.46, "learning_rate": 1.1826153531931018e-05, "loss": 1.0351, "step": 5412 }, { "epoch": 0.46, "learning_rate": 1.1823456558524124e-05, "loss": 1.0929, "step": 5413 }, { "epoch": 0.46, "learning_rate": 1.1820759447916435e-05, "loss": 1.0521, "step": 5414 }, { "epoch": 0.46, "learning_rate": 1.1818062200310879e-05, "loss": 1.0363, "step": 5415 }, { "epoch": 0.46, "learning_rate": 1.181536481591041e-05, "loss": 0.997, "step": 5416 }, { "epoch": 0.46, "learning_rate": 1.1812667294917976e-05, "loss": 1.0295, "step": 5417 }, { "epoch": 0.46, "learning_rate": 1.1809969637536553e-05, "loss": 1.0426, "step": 5418 }, { "epoch": 0.46, "learning_rate": 1.1807271843969118e-05, "loss": 1.0639, "step": 5419 }, { "epoch": 0.46, "learning_rate": 1.1804573914418652e-05, "loss": 1.0315, "step": 5420 }, { "epoch": 0.46, "learning_rate": 1.180187584908816e-05, "loss": 1.0463, "step": 5421 }, { "epoch": 0.46, "learning_rate": 1.1799177648180646e-05, "loss": 1.0424, "step": 5422 }, { "epoch": 0.46, "learning_rate": 1.1796479311899131e-05, "loss": 1.0587, "step": 5423 }, { "epoch": 0.46, "learning_rate": 1.1793780840446642e-05, "loss": 1.0617, "step": 5424 }, { "epoch": 0.46, "learning_rate": 1.1791082234026218e-05, "loss": 1.0533, "step": 5425 }, { "epoch": 0.46, "learning_rate": 1.1788383492840908e-05, "loss": 1.003, "step": 5426 }, { "epoch": 0.46, "learning_rate": 1.1785684617093773e-05, "loss": 0.9317, "step": 5427 }, { "epoch": 0.46, "learning_rate": 1.1782985606987879e-05, "loss": 1.1082, "step": 5428 }, { "epoch": 0.46, "learning_rate": 1.1780286462726306e-05, "loss": 1.017, "step": 5429 }, { "epoch": 0.46, "learning_rate": 1.1777587184512146e-05, "loss": 1.1168, "step": 5430 }, { "epoch": 0.46, "learning_rate": 1.1774887772548498e-05, "loss": 1.1023, "step": 5431 }, { "epoch": 0.46, "learning_rate": 1.1772188227038468e-05, "loss": 0.9983, "step": 5432 }, { "epoch": 0.46, "learning_rate": 1.176948854818518e-05, "loss": 1.0255, "step": 5433 }, { "epoch": 0.46, "learning_rate": 1.176678873619176e-05, "loss": 1.0214, "step": 5434 }, { "epoch": 0.46, "learning_rate": 1.176408879126135e-05, "loss": 1.0597, "step": 5435 }, { "epoch": 0.46, "learning_rate": 1.1761388713597099e-05, "loss": 1.018, "step": 5436 }, { "epoch": 0.46, "learning_rate": 1.1758688503402167e-05, "loss": 1.028, "step": 5437 }, { "epoch": 0.46, "learning_rate": 1.1755988160879723e-05, "loss": 1.0662, "step": 5438 }, { "epoch": 0.46, "learning_rate": 1.1753287686232949e-05, "loss": 1.0614, "step": 5439 }, { "epoch": 0.46, "learning_rate": 1.175058707966503e-05, "loss": 1.0289, "step": 5440 }, { "epoch": 0.46, "learning_rate": 1.1747886341379168e-05, "loss": 1.0147, "step": 5441 }, { "epoch": 0.46, "learning_rate": 1.1745185471578574e-05, "loss": 1.0711, "step": 5442 }, { "epoch": 0.46, "learning_rate": 1.1742484470466464e-05, "loss": 1.066, "step": 5443 }, { "epoch": 0.46, "learning_rate": 1.173978333824607e-05, "loss": 1.0666, "step": 5444 }, { "epoch": 0.46, "learning_rate": 1.1737082075120629e-05, "loss": 1.083, "step": 5445 }, { "epoch": 0.46, "learning_rate": 1.1734380681293391e-05, "loss": 1.0453, "step": 5446 }, { "epoch": 0.46, "learning_rate": 1.1731679156967615e-05, "loss": 0.9735, "step": 5447 }, { "epoch": 0.46, "learning_rate": 1.172897750234657e-05, "loss": 1.0401, "step": 5448 }, { "epoch": 0.46, "learning_rate": 1.1726275717633531e-05, "loss": 1.0721, "step": 5449 }, { "epoch": 0.46, "learning_rate": 1.1723573803031786e-05, "loss": 1.0712, "step": 5450 }, { "epoch": 0.46, "learning_rate": 1.1720871758744639e-05, "loss": 1.0486, "step": 5451 }, { "epoch": 0.46, "learning_rate": 1.1718169584975391e-05, "loss": 1.056, "step": 5452 }, { "epoch": 0.46, "learning_rate": 1.1715467281927364e-05, "loss": 1.1007, "step": 5453 }, { "epoch": 0.46, "learning_rate": 1.1712764849803882e-05, "loss": 1.105, "step": 5454 }, { "epoch": 0.46, "learning_rate": 1.1710062288808282e-05, "loss": 1.0616, "step": 5455 }, { "epoch": 0.46, "learning_rate": 1.1707359599143912e-05, "loss": 1.0765, "step": 5456 }, { "epoch": 0.46, "learning_rate": 1.170465678101413e-05, "loss": 1.0603, "step": 5457 }, { "epoch": 0.46, "learning_rate": 1.1701953834622297e-05, "loss": 1.058, "step": 5458 }, { "epoch": 0.46, "learning_rate": 1.169925076017179e-05, "loss": 1.0618, "step": 5459 }, { "epoch": 0.46, "learning_rate": 1.1696547557866e-05, "loss": 1.0796, "step": 5460 }, { "epoch": 0.46, "learning_rate": 1.1693844227908313e-05, "loss": 1.0008, "step": 5461 }, { "epoch": 0.46, "learning_rate": 1.1691140770502138e-05, "loss": 1.0461, "step": 5462 }, { "epoch": 0.46, "learning_rate": 1.168843718585089e-05, "loss": 1.077, "step": 5463 }, { "epoch": 0.46, "learning_rate": 1.168573347415799e-05, "loss": 1.0594, "step": 5464 }, { "epoch": 0.46, "learning_rate": 1.168302963562687e-05, "loss": 1.0329, "step": 5465 }, { "epoch": 0.46, "learning_rate": 1.1680325670460976e-05, "loss": 1.0814, "step": 5466 }, { "epoch": 0.46, "learning_rate": 1.167762157886376e-05, "loss": 1.0458, "step": 5467 }, { "epoch": 0.46, "learning_rate": 1.1674917361038682e-05, "loss": 1.0354, "step": 5468 }, { "epoch": 0.46, "learning_rate": 1.1672213017189211e-05, "loss": 1.0603, "step": 5469 }, { "epoch": 0.46, "learning_rate": 1.1669508547518834e-05, "loss": 1.0669, "step": 5470 }, { "epoch": 0.46, "learning_rate": 1.1666803952231034e-05, "loss": 1.0947, "step": 5471 }, { "epoch": 0.46, "learning_rate": 1.1664099231529315e-05, "loss": 1.0486, "step": 5472 }, { "epoch": 0.46, "learning_rate": 1.1661394385617186e-05, "loss": 1.0884, "step": 5473 }, { "epoch": 0.46, "learning_rate": 1.1658689414698163e-05, "loss": 1.1005, "step": 5474 }, { "epoch": 0.46, "learning_rate": 1.1655984318975779e-05, "loss": 1.0497, "step": 5475 }, { "epoch": 0.46, "learning_rate": 1.1653279098653565e-05, "loss": 1.1014, "step": 5476 }, { "epoch": 0.46, "learning_rate": 1.1650573753935067e-05, "loss": 1.0723, "step": 5477 }, { "epoch": 0.46, "learning_rate": 1.1647868285023849e-05, "loss": 1.0624, "step": 5478 }, { "epoch": 0.46, "learning_rate": 1.1645162692123468e-05, "loss": 1.1006, "step": 5479 }, { "epoch": 0.46, "learning_rate": 1.1642456975437505e-05, "loss": 1.0032, "step": 5480 }, { "epoch": 0.46, "learning_rate": 1.1639751135169537e-05, "loss": 1.0523, "step": 5481 }, { "epoch": 0.46, "learning_rate": 1.1637045171523162e-05, "loss": 1.0907, "step": 5482 }, { "epoch": 0.46, "learning_rate": 1.1634339084701985e-05, "loss": 1.038, "step": 5483 }, { "epoch": 0.46, "learning_rate": 1.1631632874909611e-05, "loss": 1.0769, "step": 5484 }, { "epoch": 0.46, "learning_rate": 1.1628926542349662e-05, "loss": 1.0869, "step": 5485 }, { "epoch": 0.46, "learning_rate": 1.1626220087225776e-05, "loss": 1.0441, "step": 5486 }, { "epoch": 0.46, "learning_rate": 1.1623513509741589e-05, "loss": 1.1003, "step": 5487 }, { "epoch": 0.46, "learning_rate": 1.1620806810100744e-05, "loss": 0.9956, "step": 5488 }, { "epoch": 0.46, "learning_rate": 1.1618099988506903e-05, "loss": 1.1238, "step": 5489 }, { "epoch": 0.46, "learning_rate": 1.1615393045163734e-05, "loss": 1.0776, "step": 5490 }, { "epoch": 0.47, "learning_rate": 1.1612685980274913e-05, "loss": 1.0474, "step": 5491 }, { "epoch": 0.47, "learning_rate": 1.1609978794044125e-05, "loss": 1.0275, "step": 5492 }, { "epoch": 0.47, "learning_rate": 1.1607271486675062e-05, "loss": 1.0543, "step": 5493 }, { "epoch": 0.47, "learning_rate": 1.1604564058371433e-05, "loss": 1.0391, "step": 5494 }, { "epoch": 0.47, "learning_rate": 1.1601856509336947e-05, "loss": 1.0588, "step": 5495 }, { "epoch": 0.47, "learning_rate": 1.1599148839775325e-05, "loss": 1.0191, "step": 5496 }, { "epoch": 0.47, "learning_rate": 1.1596441049890298e-05, "loss": 1.0778, "step": 5497 }, { "epoch": 0.47, "learning_rate": 1.159373313988561e-05, "loss": 1.0349, "step": 5498 }, { "epoch": 0.47, "learning_rate": 1.1591025109965007e-05, "loss": 1.0421, "step": 5499 }, { "epoch": 0.47, "learning_rate": 1.1588316960332246e-05, "loss": 1.1022, "step": 5500 }, { "epoch": 0.47, "learning_rate": 1.1585608691191094e-05, "loss": 1.0599, "step": 5501 }, { "epoch": 0.47, "learning_rate": 1.1582900302745334e-05, "loss": 0.9969, "step": 5502 }, { "epoch": 0.47, "learning_rate": 1.158019179519874e-05, "loss": 1.0642, "step": 5503 }, { "epoch": 0.47, "learning_rate": 1.1577483168755115e-05, "loss": 1.0629, "step": 5504 }, { "epoch": 0.47, "learning_rate": 1.1574774423618255e-05, "loss": 1.0696, "step": 5505 }, { "epoch": 0.47, "learning_rate": 1.1572065559991977e-05, "loss": 1.054, "step": 5506 }, { "epoch": 0.47, "learning_rate": 1.1569356578080098e-05, "loss": 1.1118, "step": 5507 }, { "epoch": 0.47, "learning_rate": 1.156664747808645e-05, "loss": 0.9914, "step": 5508 }, { "epoch": 0.47, "learning_rate": 1.156393826021487e-05, "loss": 1.0329, "step": 5509 }, { "epoch": 0.47, "learning_rate": 1.1561228924669206e-05, "loss": 0.9967, "step": 5510 }, { "epoch": 0.47, "learning_rate": 1.1558519471653314e-05, "loss": 1.0743, "step": 5511 }, { "epoch": 0.47, "learning_rate": 1.155580990137106e-05, "loss": 0.9232, "step": 5512 }, { "epoch": 0.47, "learning_rate": 1.1553100214026318e-05, "loss": 0.9819, "step": 5513 }, { "epoch": 0.47, "learning_rate": 1.1550390409822973e-05, "loss": 1.0587, "step": 5514 }, { "epoch": 0.47, "learning_rate": 1.154768048896491e-05, "loss": 1.0883, "step": 5515 }, { "epoch": 0.47, "learning_rate": 1.1544970451656035e-05, "loss": 1.0551, "step": 5516 }, { "epoch": 0.47, "learning_rate": 1.1542260298100254e-05, "loss": 1.0956, "step": 5517 }, { "epoch": 0.47, "learning_rate": 1.1539550028501485e-05, "loss": 1.0569, "step": 5518 }, { "epoch": 0.47, "learning_rate": 1.1536839643063656e-05, "loss": 1.0493, "step": 5519 }, { "epoch": 0.47, "learning_rate": 1.1534129141990702e-05, "loss": 1.038, "step": 5520 }, { "epoch": 0.47, "learning_rate": 1.1531418525486566e-05, "loss": 1.0929, "step": 5521 }, { "epoch": 0.47, "learning_rate": 1.1528707793755201e-05, "loss": 1.0216, "step": 5522 }, { "epoch": 0.47, "learning_rate": 1.1525996947000569e-05, "loss": 1.0525, "step": 5523 }, { "epoch": 0.47, "learning_rate": 1.1523285985426638e-05, "loss": 1.037, "step": 5524 }, { "epoch": 0.47, "learning_rate": 1.1520574909237387e-05, "loss": 1.0567, "step": 5525 }, { "epoch": 0.47, "learning_rate": 1.1517863718636809e-05, "loss": 1.0202, "step": 5526 }, { "epoch": 0.47, "learning_rate": 1.151515241382889e-05, "loss": 1.0556, "step": 5527 }, { "epoch": 0.47, "learning_rate": 1.1512440995017643e-05, "loss": 1.0477, "step": 5528 }, { "epoch": 0.47, "learning_rate": 1.1509729462407075e-05, "loss": 1.0459, "step": 5529 }, { "epoch": 0.47, "learning_rate": 1.1507017816201212e-05, "loss": 1.0246, "step": 5530 }, { "epoch": 0.47, "learning_rate": 1.150430605660408e-05, "loss": 1.0781, "step": 5531 }, { "epoch": 0.47, "learning_rate": 1.150159418381972e-05, "loss": 1.0599, "step": 5532 }, { "epoch": 0.47, "learning_rate": 1.149888219805218e-05, "loss": 1.0073, "step": 5533 }, { "epoch": 0.47, "learning_rate": 1.1496170099505512e-05, "loss": 1.0545, "step": 5534 }, { "epoch": 0.47, "learning_rate": 1.1493457888383786e-05, "loss": 1.1002, "step": 5535 }, { "epoch": 0.47, "learning_rate": 1.1490745564891068e-05, "loss": 1.0945, "step": 5536 }, { "epoch": 0.47, "learning_rate": 1.1488033129231442e-05, "loss": 1.0645, "step": 5537 }, { "epoch": 0.47, "learning_rate": 1.1485320581609001e-05, "loss": 1.1044, "step": 5538 }, { "epoch": 0.47, "learning_rate": 1.1482607922227838e-05, "loss": 1.0322, "step": 5539 }, { "epoch": 0.47, "learning_rate": 1.1479895151292063e-05, "loss": 1.0629, "step": 5540 }, { "epoch": 0.47, "learning_rate": 1.1477182269005787e-05, "loss": 1.0599, "step": 5541 }, { "epoch": 0.47, "learning_rate": 1.1474469275573135e-05, "loss": 1.0013, "step": 5542 }, { "epoch": 0.47, "learning_rate": 1.1471756171198236e-05, "loss": 1.0101, "step": 5543 }, { "epoch": 0.47, "learning_rate": 1.1469042956085234e-05, "loss": 1.0004, "step": 5544 }, { "epoch": 0.47, "learning_rate": 1.1466329630438278e-05, "loss": 1.0398, "step": 5545 }, { "epoch": 0.47, "learning_rate": 1.1463616194461518e-05, "loss": 1.0247, "step": 5546 }, { "epoch": 0.47, "learning_rate": 1.1460902648359124e-05, "loss": 1.0787, "step": 5547 }, { "epoch": 0.47, "learning_rate": 1.1458188992335268e-05, "loss": 1.0432, "step": 5548 }, { "epoch": 0.47, "learning_rate": 1.145547522659413e-05, "loss": 1.0868, "step": 5549 }, { "epoch": 0.47, "learning_rate": 1.14527613513399e-05, "loss": 0.9712, "step": 5550 }, { "epoch": 0.47, "learning_rate": 1.1450047366776778e-05, "loss": 0.9917, "step": 5551 }, { "epoch": 0.47, "learning_rate": 1.1447333273108969e-05, "loss": 0.9992, "step": 5552 }, { "epoch": 0.47, "learning_rate": 1.1444619070540686e-05, "loss": 0.8644, "step": 5553 }, { "epoch": 0.47, "learning_rate": 1.144190475927615e-05, "loss": 0.9635, "step": 5554 }, { "epoch": 0.47, "learning_rate": 1.1439190339519597e-05, "loss": 1.0223, "step": 5555 }, { "epoch": 0.47, "learning_rate": 1.1436475811475261e-05, "loss": 0.8683, "step": 5556 }, { "epoch": 0.47, "learning_rate": 1.1433761175347391e-05, "loss": 1.0602, "step": 5557 }, { "epoch": 0.47, "learning_rate": 1.143104643134024e-05, "loss": 1.0774, "step": 5558 }, { "epoch": 0.47, "learning_rate": 1.1428331579658072e-05, "loss": 0.8675, "step": 5559 }, { "epoch": 0.47, "learning_rate": 1.142561662050516e-05, "loss": 1.078, "step": 5560 }, { "epoch": 0.47, "learning_rate": 1.1422901554085784e-05, "loss": 0.9348, "step": 5561 }, { "epoch": 0.47, "learning_rate": 1.1420186380604227e-05, "loss": 1.0514, "step": 5562 }, { "epoch": 0.47, "learning_rate": 1.1417471100264787e-05, "loss": 1.025, "step": 5563 }, { "epoch": 0.47, "learning_rate": 1.141475571327177e-05, "loss": 1.0732, "step": 5564 }, { "epoch": 0.47, "learning_rate": 1.1412040219829486e-05, "loss": 1.0863, "step": 5565 }, { "epoch": 0.47, "learning_rate": 1.1409324620142252e-05, "loss": 0.9994, "step": 5566 }, { "epoch": 0.47, "learning_rate": 1.1406608914414402e-05, "loss": 1.1067, "step": 5567 }, { "epoch": 0.47, "learning_rate": 1.1403893102850263e-05, "loss": 1.0481, "step": 5568 }, { "epoch": 0.47, "learning_rate": 1.1401177185654181e-05, "loss": 1.0643, "step": 5569 }, { "epoch": 0.47, "learning_rate": 1.1398461163030513e-05, "loss": 0.972, "step": 5570 }, { "epoch": 0.47, "learning_rate": 1.1395745035183612e-05, "loss": 1.0517, "step": 5571 }, { "epoch": 0.47, "learning_rate": 1.139302880231785e-05, "loss": 1.0068, "step": 5572 }, { "epoch": 0.47, "learning_rate": 1.1390312464637598e-05, "loss": 1.0492, "step": 5573 }, { "epoch": 0.47, "learning_rate": 1.1387596022347243e-05, "loss": 1.0315, "step": 5574 }, { "epoch": 0.47, "learning_rate": 1.1384879475651171e-05, "loss": 1.0401, "step": 5575 }, { "epoch": 0.47, "learning_rate": 1.1382162824753786e-05, "loss": 1.0794, "step": 5576 }, { "epoch": 0.47, "learning_rate": 1.1379446069859492e-05, "loss": 0.999, "step": 5577 }, { "epoch": 0.47, "learning_rate": 1.1376729211172703e-05, "loss": 1.0746, "step": 5578 }, { "epoch": 0.47, "learning_rate": 1.1374012248897846e-05, "loss": 1.0857, "step": 5579 }, { "epoch": 0.47, "learning_rate": 1.1371295183239344e-05, "loss": 1.016, "step": 5580 }, { "epoch": 0.47, "learning_rate": 1.1368578014401638e-05, "loss": 1.0525, "step": 5581 }, { "epoch": 0.47, "learning_rate": 1.1365860742589178e-05, "loss": 1.0983, "step": 5582 }, { "epoch": 0.47, "learning_rate": 1.136314336800641e-05, "loss": 1.0305, "step": 5583 }, { "epoch": 0.47, "learning_rate": 1.13604258908578e-05, "loss": 1.0726, "step": 5584 }, { "epoch": 0.47, "learning_rate": 1.1357708311347815e-05, "loss": 1.0314, "step": 5585 }, { "epoch": 0.47, "learning_rate": 1.135499062968093e-05, "loss": 1.0249, "step": 5586 }, { "epoch": 0.47, "learning_rate": 1.1352272846061634e-05, "loss": 1.0183, "step": 5587 }, { "epoch": 0.47, "learning_rate": 1.1349554960694413e-05, "loss": 1.0157, "step": 5588 }, { "epoch": 0.47, "learning_rate": 1.1346836973783769e-05, "loss": 1.1101, "step": 5589 }, { "epoch": 0.47, "learning_rate": 1.1344118885534213e-05, "loss": 0.9752, "step": 5590 }, { "epoch": 0.47, "learning_rate": 1.1341400696150253e-05, "loss": 0.9946, "step": 5591 }, { "epoch": 0.47, "learning_rate": 1.1338682405836417e-05, "loss": 1.0787, "step": 5592 }, { "epoch": 0.47, "learning_rate": 1.1335964014797231e-05, "loss": 1.0321, "step": 5593 }, { "epoch": 0.47, "learning_rate": 1.1333245523237239e-05, "loss": 1.0238, "step": 5594 }, { "epoch": 0.47, "learning_rate": 1.1330526931360981e-05, "loss": 1.0315, "step": 5595 }, { "epoch": 0.47, "learning_rate": 1.1327808239373009e-05, "loss": 1.0797, "step": 5596 }, { "epoch": 0.47, "learning_rate": 1.1325089447477881e-05, "loss": 1.0232, "step": 5597 }, { "epoch": 0.47, "learning_rate": 1.1322370555880173e-05, "loss": 1.0689, "step": 5598 }, { "epoch": 0.47, "learning_rate": 1.1319651564784454e-05, "loss": 1.0583, "step": 5599 }, { "epoch": 0.47, "learning_rate": 1.1316932474395311e-05, "loss": 1.0413, "step": 5600 }, { "epoch": 0.47, "learning_rate": 1.1314213284917328e-05, "loss": 1.0834, "step": 5601 }, { "epoch": 0.47, "learning_rate": 1.1311493996555107e-05, "loss": 1.0227, "step": 5602 }, { "epoch": 0.47, "learning_rate": 1.1308774609513253e-05, "loss": 0.8401, "step": 5603 }, { "epoch": 0.47, "learning_rate": 1.1306055123996378e-05, "loss": 1.1173, "step": 5604 }, { "epoch": 0.47, "learning_rate": 1.1303335540209102e-05, "loss": 1.0619, "step": 5605 }, { "epoch": 0.47, "learning_rate": 1.1300615858356052e-05, "loss": 1.045, "step": 5606 }, { "epoch": 0.47, "learning_rate": 1.1297896078641864e-05, "loss": 1.0517, "step": 5607 }, { "epoch": 0.47, "learning_rate": 1.1295176201271177e-05, "loss": 1.0102, "step": 5608 }, { "epoch": 0.48, "learning_rate": 1.1292456226448645e-05, "loss": 1.0056, "step": 5609 }, { "epoch": 0.48, "learning_rate": 1.128973615437892e-05, "loss": 0.9967, "step": 5610 }, { "epoch": 0.48, "learning_rate": 1.1287015985266671e-05, "loss": 1.0186, "step": 5611 }, { "epoch": 0.48, "learning_rate": 1.1284295719316565e-05, "loss": 1.06, "step": 5612 }, { "epoch": 0.48, "learning_rate": 1.1281575356733283e-05, "loss": 1.0695, "step": 5613 }, { "epoch": 0.48, "learning_rate": 1.127885489772151e-05, "loss": 1.0674, "step": 5614 }, { "epoch": 0.48, "learning_rate": 1.1276134342485936e-05, "loss": 1.0427, "step": 5615 }, { "epoch": 0.48, "learning_rate": 1.1273413691231267e-05, "loss": 1.0872, "step": 5616 }, { "epoch": 0.48, "learning_rate": 1.1270692944162212e-05, "loss": 1.0777, "step": 5617 }, { "epoch": 0.48, "learning_rate": 1.126797210148348e-05, "loss": 1.0482, "step": 5618 }, { "epoch": 0.48, "learning_rate": 1.1265251163399796e-05, "loss": 1.0992, "step": 5619 }, { "epoch": 0.48, "learning_rate": 1.1262530130115887e-05, "loss": 1.0601, "step": 5620 }, { "epoch": 0.48, "learning_rate": 1.1259809001836498e-05, "loss": 1.0423, "step": 5621 }, { "epoch": 0.48, "learning_rate": 1.1257087778766358e-05, "loss": 1.0421, "step": 5622 }, { "epoch": 0.48, "learning_rate": 1.1254366461110229e-05, "loss": 1.0258, "step": 5623 }, { "epoch": 0.48, "learning_rate": 1.1251645049072866e-05, "loss": 1.0491, "step": 5624 }, { "epoch": 0.48, "learning_rate": 1.124892354285903e-05, "loss": 1.0535, "step": 5625 }, { "epoch": 0.48, "learning_rate": 1.1246201942673498e-05, "loss": 1.1, "step": 5626 }, { "epoch": 0.48, "learning_rate": 1.1243480248721046e-05, "loss": 0.8748, "step": 5627 }, { "epoch": 0.48, "learning_rate": 1.1240758461206463e-05, "loss": 1.0919, "step": 5628 }, { "epoch": 0.48, "learning_rate": 1.1238036580334539e-05, "loss": 1.0372, "step": 5629 }, { "epoch": 0.48, "learning_rate": 1.1235314606310077e-05, "loss": 0.9809, "step": 5630 }, { "epoch": 0.48, "learning_rate": 1.1232592539337883e-05, "loss": 1.1341, "step": 5631 }, { "epoch": 0.48, "learning_rate": 1.1229870379622771e-05, "loss": 1.025, "step": 5632 }, { "epoch": 0.48, "learning_rate": 1.122714812736956e-05, "loss": 1.0442, "step": 5633 }, { "epoch": 0.48, "learning_rate": 1.1224425782783083e-05, "loss": 1.0444, "step": 5634 }, { "epoch": 0.48, "learning_rate": 1.1221703346068171e-05, "loss": 1.0457, "step": 5635 }, { "epoch": 0.48, "learning_rate": 1.121898081742967e-05, "loss": 1.0387, "step": 5636 }, { "epoch": 0.48, "learning_rate": 1.1216258197072425e-05, "loss": 0.9759, "step": 5637 }, { "epoch": 0.48, "learning_rate": 1.1213535485201292e-05, "loss": 1.0124, "step": 5638 }, { "epoch": 0.48, "learning_rate": 1.1210812682021135e-05, "loss": 1.0371, "step": 5639 }, { "epoch": 0.48, "learning_rate": 1.1208089787736823e-05, "loss": 1.0783, "step": 5640 }, { "epoch": 0.48, "learning_rate": 1.1205366802553231e-05, "loss": 0.9059, "step": 5641 }, { "epoch": 0.48, "learning_rate": 1.1202643726675247e-05, "loss": 1.0655, "step": 5642 }, { "epoch": 0.48, "learning_rate": 1.1199920560307757e-05, "loss": 1.0478, "step": 5643 }, { "epoch": 0.48, "learning_rate": 1.119719730365566e-05, "loss": 1.0346, "step": 5644 }, { "epoch": 0.48, "learning_rate": 1.119447395692386e-05, "loss": 1.0215, "step": 5645 }, { "epoch": 0.48, "learning_rate": 1.1191750520317264e-05, "loss": 1.04, "step": 5646 }, { "epoch": 0.48, "learning_rate": 1.1189026994040792e-05, "loss": 1.0574, "step": 5647 }, { "epoch": 0.48, "learning_rate": 1.118630337829937e-05, "loss": 1.0034, "step": 5648 }, { "epoch": 0.48, "learning_rate": 1.1183579673297925e-05, "loss": 1.0139, "step": 5649 }, { "epoch": 0.48, "learning_rate": 1.1180855879241394e-05, "loss": 1.0839, "step": 5650 }, { "epoch": 0.48, "learning_rate": 1.1178131996334724e-05, "loss": 1.0862, "step": 5651 }, { "epoch": 0.48, "learning_rate": 1.1175408024782863e-05, "loss": 1.0578, "step": 5652 }, { "epoch": 0.48, "learning_rate": 1.1172683964790773e-05, "loss": 1.0519, "step": 5653 }, { "epoch": 0.48, "learning_rate": 1.1169959816563413e-05, "loss": 1.0828, "step": 5654 }, { "epoch": 0.48, "learning_rate": 1.1167235580305758e-05, "loss": 1.0461, "step": 5655 }, { "epoch": 0.48, "learning_rate": 1.116451125622278e-05, "loss": 1.0503, "step": 5656 }, { "epoch": 0.48, "learning_rate": 1.1161786844519469e-05, "loss": 0.9941, "step": 5657 }, { "epoch": 0.48, "learning_rate": 1.1159062345400812e-05, "loss": 1.0764, "step": 5658 }, { "epoch": 0.48, "learning_rate": 1.115633775907181e-05, "loss": 1.0496, "step": 5659 }, { "epoch": 0.48, "learning_rate": 1.1153613085737461e-05, "loss": 1.1111, "step": 5660 }, { "epoch": 0.48, "learning_rate": 1.115088832560278e-05, "loss": 1.0516, "step": 5661 }, { "epoch": 0.48, "learning_rate": 1.114816347887278e-05, "loss": 1.0356, "step": 5662 }, { "epoch": 0.48, "learning_rate": 1.114543854575249e-05, "loss": 1.0535, "step": 5663 }, { "epoch": 0.48, "learning_rate": 1.1142713526446933e-05, "loss": 1.0209, "step": 5664 }, { "epoch": 0.48, "learning_rate": 1.113998842116115e-05, "loss": 1.0569, "step": 5665 }, { "epoch": 0.48, "learning_rate": 1.113726323010018e-05, "loss": 1.0355, "step": 5666 }, { "epoch": 0.48, "learning_rate": 1.1134537953469077e-05, "loss": 1.0187, "step": 5667 }, { "epoch": 0.48, "learning_rate": 1.1131812591472895e-05, "loss": 1.063, "step": 5668 }, { "epoch": 0.48, "learning_rate": 1.1129087144316692e-05, "loss": 0.9259, "step": 5669 }, { "epoch": 0.48, "learning_rate": 1.1126361612205544e-05, "loss": 1.079, "step": 5670 }, { "epoch": 0.48, "learning_rate": 1.1123635995344519e-05, "loss": 1.0806, "step": 5671 }, { "epoch": 0.48, "learning_rate": 1.1120910293938703e-05, "loss": 1.0478, "step": 5672 }, { "epoch": 0.48, "learning_rate": 1.111818450819318e-05, "loss": 1.0289, "step": 5673 }, { "epoch": 0.48, "learning_rate": 1.1115458638313049e-05, "loss": 1.0384, "step": 5674 }, { "epoch": 0.48, "learning_rate": 1.1112732684503405e-05, "loss": 1.1086, "step": 5675 }, { "epoch": 0.48, "learning_rate": 1.1110006646969356e-05, "loss": 1.0511, "step": 5676 }, { "epoch": 0.48, "learning_rate": 1.1107280525916018e-05, "loss": 0.9599, "step": 5677 }, { "epoch": 0.48, "learning_rate": 1.1104554321548504e-05, "loss": 1.0839, "step": 5678 }, { "epoch": 0.48, "learning_rate": 1.1101828034071947e-05, "loss": 1.0599, "step": 5679 }, { "epoch": 0.48, "learning_rate": 1.1099101663691474e-05, "loss": 1.0457, "step": 5680 }, { "epoch": 0.48, "learning_rate": 1.1096375210612223e-05, "loss": 1.0937, "step": 5681 }, { "epoch": 0.48, "learning_rate": 1.109364867503934e-05, "loss": 1.0878, "step": 5682 }, { "epoch": 0.48, "learning_rate": 1.1090922057177976e-05, "loss": 0.878, "step": 5683 }, { "epoch": 0.48, "learning_rate": 1.1088195357233282e-05, "loss": 1.0528, "step": 5684 }, { "epoch": 0.48, "learning_rate": 1.108546857541043e-05, "loss": 1.0687, "step": 5685 }, { "epoch": 0.48, "learning_rate": 1.1082741711914582e-05, "loss": 1.037, "step": 5686 }, { "epoch": 0.48, "learning_rate": 1.1080014766950916e-05, "loss": 1.0958, "step": 5687 }, { "epoch": 0.48, "learning_rate": 1.107728774072461e-05, "loss": 1.0185, "step": 5688 }, { "epoch": 0.48, "learning_rate": 1.1074560633440858e-05, "loss": 1.055, "step": 5689 }, { "epoch": 0.48, "learning_rate": 1.1071833445304844e-05, "loss": 1.045, "step": 5690 }, { "epoch": 0.48, "learning_rate": 1.1069106176521775e-05, "loss": 1.056, "step": 5691 }, { "epoch": 0.48, "learning_rate": 1.1066378827296854e-05, "loss": 0.9799, "step": 5692 }, { "epoch": 0.48, "learning_rate": 1.1063651397835293e-05, "loss": 1.0404, "step": 5693 }, { "epoch": 0.48, "learning_rate": 1.106092388834231e-05, "loss": 1.1101, "step": 5694 }, { "epoch": 0.48, "learning_rate": 1.1058196299023125e-05, "loss": 1.0341, "step": 5695 }, { "epoch": 0.48, "learning_rate": 1.1055468630082975e-05, "loss": 1.0795, "step": 5696 }, { "epoch": 0.48, "learning_rate": 1.105274088172709e-05, "loss": 1.059, "step": 5697 }, { "epoch": 0.48, "learning_rate": 1.1050013054160713e-05, "loss": 1.1386, "step": 5698 }, { "epoch": 0.48, "learning_rate": 1.104728514758909e-05, "loss": 1.065, "step": 5699 }, { "epoch": 0.48, "learning_rate": 1.104455716221748e-05, "loss": 1.0494, "step": 5700 }, { "epoch": 0.48, "learning_rate": 1.1041829098251137e-05, "loss": 1.0591, "step": 5701 }, { "epoch": 0.48, "learning_rate": 1.103910095589533e-05, "loss": 1.0627, "step": 5702 }, { "epoch": 0.48, "learning_rate": 1.1036372735355328e-05, "loss": 1.0705, "step": 5703 }, { "epoch": 0.48, "learning_rate": 1.1033644436836408e-05, "loss": 1.03, "step": 5704 }, { "epoch": 0.48, "learning_rate": 1.1030916060543852e-05, "loss": 1.03, "step": 5705 }, { "epoch": 0.48, "learning_rate": 1.1028187606682954e-05, "loss": 1.0668, "step": 5706 }, { "epoch": 0.48, "learning_rate": 1.1025459075459003e-05, "loss": 1.0835, "step": 5707 }, { "epoch": 0.48, "learning_rate": 1.1022730467077302e-05, "loss": 1.0378, "step": 5708 }, { "epoch": 0.48, "learning_rate": 1.1020001781743159e-05, "loss": 1.0325, "step": 5709 }, { "epoch": 0.48, "learning_rate": 1.1017273019661884e-05, "loss": 1.0465, "step": 5710 }, { "epoch": 0.48, "learning_rate": 1.1014544181038798e-05, "loss": 1.0044, "step": 5711 }, { "epoch": 0.48, "learning_rate": 1.101181526607922e-05, "loss": 1.0028, "step": 5712 }, { "epoch": 0.48, "learning_rate": 1.100908627498848e-05, "loss": 1.0648, "step": 5713 }, { "epoch": 0.48, "learning_rate": 1.1006357207971919e-05, "loss": 1.0635, "step": 5714 }, { "epoch": 0.48, "learning_rate": 1.100362806523487e-05, "loss": 1.0142, "step": 5715 }, { "epoch": 0.48, "learning_rate": 1.1000898846982686e-05, "loss": 1.0466, "step": 5716 }, { "epoch": 0.48, "learning_rate": 1.0998169553420714e-05, "loss": 1.0322, "step": 5717 }, { "epoch": 0.48, "learning_rate": 1.0995440184754316e-05, "loss": 1.0765, "step": 5718 }, { "epoch": 0.48, "learning_rate": 1.0992710741188852e-05, "loss": 1.0393, "step": 5719 }, { "epoch": 0.48, "learning_rate": 1.0989981222929695e-05, "loss": 1.0056, "step": 5720 }, { "epoch": 0.48, "learning_rate": 1.0987251630182218e-05, "loss": 1.0659, "step": 5721 }, { "epoch": 0.48, "learning_rate": 1.0984521963151802e-05, "loss": 1.0556, "step": 5722 }, { "epoch": 0.48, "learning_rate": 1.0981792222043833e-05, "loss": 1.088, "step": 5723 }, { "epoch": 0.48, "learning_rate": 1.0979062407063701e-05, "loss": 0.9599, "step": 5724 }, { "epoch": 0.48, "learning_rate": 1.0976332518416806e-05, "loss": 1.0547, "step": 5725 }, { "epoch": 0.48, "learning_rate": 1.0973602556308548e-05, "loss": 1.0729, "step": 5726 }, { "epoch": 0.49, "learning_rate": 1.0970872520944336e-05, "loss": 1.0553, "step": 5727 }, { "epoch": 0.49, "learning_rate": 1.0968142412529585e-05, "loss": 1.1406, "step": 5728 }, { "epoch": 0.49, "learning_rate": 1.0965412231269715e-05, "loss": 1.0692, "step": 5729 }, { "epoch": 0.49, "learning_rate": 1.0962681977370144e-05, "loss": 1.0759, "step": 5730 }, { "epoch": 0.49, "learning_rate": 1.095995165103631e-05, "loss": 1.0001, "step": 5731 }, { "epoch": 0.49, "learning_rate": 1.0957221252473647e-05, "loss": 1.0467, "step": 5732 }, { "epoch": 0.49, "learning_rate": 1.0954490781887592e-05, "loss": 1.0363, "step": 5733 }, { "epoch": 0.49, "learning_rate": 1.0951760239483595e-05, "loss": 1.0438, "step": 5734 }, { "epoch": 0.49, "learning_rate": 1.094902962546711e-05, "loss": 1.0957, "step": 5735 }, { "epoch": 0.49, "learning_rate": 1.0946298940043589e-05, "loss": 1.1236, "step": 5736 }, { "epoch": 0.49, "learning_rate": 1.0943568183418499e-05, "loss": 1.0362, "step": 5737 }, { "epoch": 0.49, "learning_rate": 1.0940837355797304e-05, "loss": 1.0419, "step": 5738 }, { "epoch": 0.49, "learning_rate": 1.093810645738548e-05, "loss": 1.0232, "step": 5739 }, { "epoch": 0.49, "learning_rate": 1.0935375488388506e-05, "loss": 1.0138, "step": 5740 }, { "epoch": 0.49, "learning_rate": 1.0932644449011865e-05, "loss": 1.0662, "step": 5741 }, { "epoch": 0.49, "learning_rate": 1.0929913339461046e-05, "loss": 1.0287, "step": 5742 }, { "epoch": 0.49, "learning_rate": 1.0927182159941547e-05, "loss": 1.0939, "step": 5743 }, { "epoch": 0.49, "learning_rate": 1.0924450910658862e-05, "loss": 1.0726, "step": 5744 }, { "epoch": 0.49, "learning_rate": 1.0921719591818498e-05, "loss": 1.0231, "step": 5745 }, { "epoch": 0.49, "learning_rate": 1.0918988203625969e-05, "loss": 1.0043, "step": 5746 }, { "epoch": 0.49, "learning_rate": 1.0916256746286786e-05, "loss": 1.0372, "step": 5747 }, { "epoch": 0.49, "learning_rate": 1.091352522000647e-05, "loss": 1.1068, "step": 5748 }, { "epoch": 0.49, "learning_rate": 1.0910793624990553e-05, "loss": 1.0791, "step": 5749 }, { "epoch": 0.49, "learning_rate": 1.0908061961444559e-05, "loss": 1.0356, "step": 5750 }, { "epoch": 0.49, "learning_rate": 1.0905330229574027e-05, "loss": 1.0071, "step": 5751 }, { "epoch": 0.49, "learning_rate": 1.0902598429584498e-05, "loss": 0.9859, "step": 5752 }, { "epoch": 0.49, "learning_rate": 1.089986656168152e-05, "loss": 1.0645, "step": 5753 }, { "epoch": 0.49, "learning_rate": 1.0897134626070643e-05, "loss": 1.0731, "step": 5754 }, { "epoch": 0.49, "learning_rate": 1.0894402622957426e-05, "loss": 1.04, "step": 5755 }, { "epoch": 0.49, "learning_rate": 1.0891670552547426e-05, "loss": 1.0064, "step": 5756 }, { "epoch": 0.49, "learning_rate": 1.0888938415046213e-05, "loss": 1.0643, "step": 5757 }, { "epoch": 0.49, "learning_rate": 1.0886206210659358e-05, "loss": 1.0531, "step": 5758 }, { "epoch": 0.49, "learning_rate": 1.0883473939592442e-05, "loss": 1.0381, "step": 5759 }, { "epoch": 0.49, "learning_rate": 1.0880741602051039e-05, "loss": 1.0388, "step": 5760 }, { "epoch": 0.49, "learning_rate": 1.0878009198240744e-05, "loss": 1.0526, "step": 5761 }, { "epoch": 0.49, "learning_rate": 1.0875276728367143e-05, "loss": 1.0067, "step": 5762 }, { "epoch": 0.49, "learning_rate": 1.0872544192635836e-05, "loss": 1.0307, "step": 5763 }, { "epoch": 0.49, "learning_rate": 1.0869811591252427e-05, "loss": 1.0698, "step": 5764 }, { "epoch": 0.49, "learning_rate": 1.0867078924422517e-05, "loss": 1.075, "step": 5765 }, { "epoch": 0.49, "learning_rate": 1.0864346192351722e-05, "loss": 1.0215, "step": 5766 }, { "epoch": 0.49, "learning_rate": 1.0861613395245656e-05, "loss": 1.0748, "step": 5767 }, { "epoch": 0.49, "learning_rate": 1.0858880533309944e-05, "loss": 1.0406, "step": 5768 }, { "epoch": 0.49, "learning_rate": 1.085614760675021e-05, "loss": 1.0057, "step": 5769 }, { "epoch": 0.49, "learning_rate": 1.0853414615772084e-05, "loss": 1.0617, "step": 5770 }, { "epoch": 0.49, "learning_rate": 1.0850681560581206e-05, "loss": 1.0187, "step": 5771 }, { "epoch": 0.49, "learning_rate": 1.0847948441383214e-05, "loss": 1.0165, "step": 5772 }, { "epoch": 0.49, "learning_rate": 1.0845215258383755e-05, "loss": 1.1218, "step": 5773 }, { "epoch": 0.49, "learning_rate": 1.084248201178848e-05, "loss": 1.0531, "step": 5774 }, { "epoch": 0.49, "learning_rate": 1.0839748701803042e-05, "loss": 1.0709, "step": 5775 }, { "epoch": 0.49, "learning_rate": 1.0837015328633102e-05, "loss": 1.0436, "step": 5776 }, { "epoch": 0.49, "learning_rate": 1.0834281892484326e-05, "loss": 1.0577, "step": 5777 }, { "epoch": 0.49, "learning_rate": 1.0831548393562383e-05, "loss": 1.0581, "step": 5778 }, { "epoch": 0.49, "learning_rate": 1.082881483207295e-05, "loss": 0.9864, "step": 5779 }, { "epoch": 0.49, "learning_rate": 1.08260812082217e-05, "loss": 0.9995, "step": 5780 }, { "epoch": 0.49, "learning_rate": 1.0823347522214325e-05, "loss": 1.0351, "step": 5781 }, { "epoch": 0.49, "learning_rate": 1.0820613774256508e-05, "loss": 1.0604, "step": 5782 }, { "epoch": 0.49, "learning_rate": 1.0817879964553938e-05, "loss": 1.0467, "step": 5783 }, { "epoch": 0.49, "learning_rate": 1.081514609331232e-05, "loss": 1.03, "step": 5784 }, { "epoch": 0.49, "learning_rate": 1.0812412160737356e-05, "loss": 1.0201, "step": 5785 }, { "epoch": 0.49, "learning_rate": 1.0809678167034749e-05, "loss": 1.0412, "step": 5786 }, { "epoch": 0.49, "learning_rate": 1.0806944112410212e-05, "loss": 1.132, "step": 5787 }, { "epoch": 0.49, "learning_rate": 1.0804209997069462e-05, "loss": 1.015, "step": 5788 }, { "epoch": 0.49, "learning_rate": 1.0801475821218219e-05, "loss": 1.0549, "step": 5789 }, { "epoch": 0.49, "learning_rate": 1.0798741585062209e-05, "loss": 1.0809, "step": 5790 }, { "epoch": 0.49, "learning_rate": 1.0796007288807163e-05, "loss": 1.0148, "step": 5791 }, { "epoch": 0.49, "learning_rate": 1.0793272932658812e-05, "loss": 1.0128, "step": 5792 }, { "epoch": 0.49, "learning_rate": 1.0790538516822897e-05, "loss": 1.0872, "step": 5793 }, { "epoch": 0.49, "learning_rate": 1.0787804041505161e-05, "loss": 0.9999, "step": 5794 }, { "epoch": 0.49, "learning_rate": 1.0785069506911353e-05, "loss": 1.0141, "step": 5795 }, { "epoch": 0.49, "learning_rate": 1.0782334913247222e-05, "loss": 1.0362, "step": 5796 }, { "epoch": 0.49, "learning_rate": 1.0779600260718528e-05, "loss": 1.0661, "step": 5797 }, { "epoch": 0.49, "learning_rate": 1.077686554953103e-05, "loss": 1.0434, "step": 5798 }, { "epoch": 0.49, "learning_rate": 1.0774130779890495e-05, "loss": 1.0668, "step": 5799 }, { "epoch": 0.49, "learning_rate": 1.0771395952002692e-05, "loss": 1.0964, "step": 5800 }, { "epoch": 0.49, "learning_rate": 1.0768661066073399e-05, "loss": 1.0232, "step": 5801 }, { "epoch": 0.49, "learning_rate": 1.0765926122308387e-05, "loss": 1.0216, "step": 5802 }, { "epoch": 0.49, "learning_rate": 1.0763191120913446e-05, "loss": 1.1238, "step": 5803 }, { "epoch": 0.49, "learning_rate": 1.076045606209436e-05, "loss": 1.0421, "step": 5804 }, { "epoch": 0.49, "learning_rate": 1.0757720946056921e-05, "loss": 1.0227, "step": 5805 }, { "epoch": 0.49, "learning_rate": 1.0754985773006928e-05, "loss": 1.1033, "step": 5806 }, { "epoch": 0.49, "learning_rate": 1.0752250543150177e-05, "loss": 1.0758, "step": 5807 }, { "epoch": 0.49, "learning_rate": 1.074951525669248e-05, "loss": 0.8704, "step": 5808 }, { "epoch": 0.49, "learning_rate": 1.0746779913839636e-05, "loss": 1.0131, "step": 5809 }, { "epoch": 0.49, "learning_rate": 1.074404451479746e-05, "loss": 1.0875, "step": 5810 }, { "epoch": 0.49, "learning_rate": 1.0741309059771776e-05, "loss": 1.1006, "step": 5811 }, { "epoch": 0.49, "learning_rate": 1.0738573548968398e-05, "loss": 1.0643, "step": 5812 }, { "epoch": 0.49, "learning_rate": 1.0735837982593158e-05, "loss": 1.0664, "step": 5813 }, { "epoch": 0.49, "learning_rate": 1.0733102360851881e-05, "loss": 1.0375, "step": 5814 }, { "epoch": 0.49, "learning_rate": 1.0730366683950406e-05, "loss": 0.9353, "step": 5815 }, { "epoch": 0.49, "learning_rate": 1.0727630952094566e-05, "loss": 1.0602, "step": 5816 }, { "epoch": 0.49, "learning_rate": 1.0724895165490205e-05, "loss": 1.0599, "step": 5817 }, { "epoch": 0.49, "learning_rate": 1.0722159324343171e-05, "loss": 1.0457, "step": 5818 }, { "epoch": 0.49, "learning_rate": 1.0719423428859314e-05, "loss": 1.0266, "step": 5819 }, { "epoch": 0.49, "learning_rate": 1.0716687479244487e-05, "loss": 1.0879, "step": 5820 }, { "epoch": 0.49, "learning_rate": 1.0713951475704554e-05, "loss": 1.091, "step": 5821 }, { "epoch": 0.49, "learning_rate": 1.0711215418445371e-05, "loss": 1.0016, "step": 5822 }, { "epoch": 0.49, "learning_rate": 1.0708479307672807e-05, "loss": 1.0722, "step": 5823 }, { "epoch": 0.49, "learning_rate": 1.0705743143592734e-05, "loss": 0.9958, "step": 5824 }, { "epoch": 0.49, "learning_rate": 1.0703006926411028e-05, "loss": 1.0586, "step": 5825 }, { "epoch": 0.49, "learning_rate": 1.0700270656333567e-05, "loss": 0.8456, "step": 5826 }, { "epoch": 0.49, "learning_rate": 1.069753433356623e-05, "loss": 1.0608, "step": 5827 }, { "epoch": 0.49, "learning_rate": 1.069479795831491e-05, "loss": 1.033, "step": 5828 }, { "epoch": 0.49, "learning_rate": 1.0692061530785494e-05, "loss": 1.0971, "step": 5829 }, { "epoch": 0.49, "learning_rate": 1.0689325051183878e-05, "loss": 1.0694, "step": 5830 }, { "epoch": 0.49, "learning_rate": 1.068658851971596e-05, "loss": 1.04, "step": 5831 }, { "epoch": 0.49, "learning_rate": 1.068385193658764e-05, "loss": 1.1257, "step": 5832 }, { "epoch": 0.49, "learning_rate": 1.0681115302004831e-05, "loss": 1.0714, "step": 5833 }, { "epoch": 0.49, "learning_rate": 1.0678378616173437e-05, "loss": 1.0698, "step": 5834 }, { "epoch": 0.49, "learning_rate": 1.0675641879299379e-05, "loss": 1.0507, "step": 5835 }, { "epoch": 0.49, "learning_rate": 1.0672905091588566e-05, "loss": 1.019, "step": 5836 }, { "epoch": 0.49, "learning_rate": 1.0670168253246926e-05, "loss": 0.9777, "step": 5837 }, { "epoch": 0.49, "learning_rate": 1.066743136448038e-05, "loss": 1.0374, "step": 5838 }, { "epoch": 0.49, "learning_rate": 1.0664694425494863e-05, "loss": 1.0422, "step": 5839 }, { "epoch": 0.49, "learning_rate": 1.0661957436496306e-05, "loss": 0.9782, "step": 5840 }, { "epoch": 0.49, "learning_rate": 1.0659220397690643e-05, "loss": 1.0738, "step": 5841 }, { "epoch": 0.49, "learning_rate": 1.065648330928382e-05, "loss": 0.9834, "step": 5842 }, { "epoch": 0.49, "learning_rate": 1.0653746171481778e-05, "loss": 1.0629, "step": 5843 }, { "epoch": 0.49, "learning_rate": 1.0651008984490466e-05, "loss": 1.0254, "step": 5844 }, { "epoch": 0.5, "learning_rate": 1.0648271748515833e-05, "loss": 1.0505, "step": 5845 }, { "epoch": 0.5, "learning_rate": 1.0645534463763838e-05, "loss": 0.9468, "step": 5846 }, { "epoch": 0.5, "learning_rate": 1.064279713044044e-05, "loss": 1.0543, "step": 5847 }, { "epoch": 0.5, "learning_rate": 1.06400597487516e-05, "loss": 1.0169, "step": 5848 }, { "epoch": 0.5, "learning_rate": 1.0637322318903287e-05, "loss": 1.0814, "step": 5849 }, { "epoch": 0.5, "learning_rate": 1.063458484110147e-05, "loss": 1.0096, "step": 5850 }, { "epoch": 0.5, "learning_rate": 1.0631847315552122e-05, "loss": 1.0059, "step": 5851 }, { "epoch": 0.5, "learning_rate": 1.0629109742461222e-05, "loss": 1.0663, "step": 5852 }, { "epoch": 0.5, "learning_rate": 1.062637212203475e-05, "loss": 1.0538, "step": 5853 }, { "epoch": 0.5, "learning_rate": 1.0623634454478686e-05, "loss": 1.0653, "step": 5854 }, { "epoch": 0.5, "learning_rate": 1.0620896739999027e-05, "loss": 1.0584, "step": 5855 }, { "epoch": 0.5, "learning_rate": 1.0618158978801756e-05, "loss": 1.0937, "step": 5856 }, { "epoch": 0.5, "learning_rate": 1.0615421171092875e-05, "loss": 1.0184, "step": 5857 }, { "epoch": 0.5, "learning_rate": 1.0612683317078376e-05, "loss": 1.066, "step": 5858 }, { "epoch": 0.5, "learning_rate": 1.0609945416964267e-05, "loss": 1.023, "step": 5859 }, { "epoch": 0.5, "learning_rate": 1.0607207470956551e-05, "loss": 1.0557, "step": 5860 }, { "epoch": 0.5, "learning_rate": 1.0604469479261235e-05, "loss": 1.0843, "step": 5861 }, { "epoch": 0.5, "learning_rate": 1.0601731442084336e-05, "loss": 1.0742, "step": 5862 }, { "epoch": 0.5, "learning_rate": 1.0598993359631867e-05, "loss": 1.0263, "step": 5863 }, { "epoch": 0.5, "learning_rate": 1.0596255232109844e-05, "loss": 1.0132, "step": 5864 }, { "epoch": 0.5, "learning_rate": 1.0593517059724294e-05, "loss": 1.0736, "step": 5865 }, { "epoch": 0.5, "learning_rate": 1.0590778842681241e-05, "loss": 0.9739, "step": 5866 }, { "epoch": 0.5, "learning_rate": 1.0588040581186716e-05, "loss": 1.0906, "step": 5867 }, { "epoch": 0.5, "learning_rate": 1.0585302275446751e-05, "loss": 1.0852, "step": 5868 }, { "epoch": 0.5, "learning_rate": 1.0582563925667384e-05, "loss": 1.023, "step": 5869 }, { "epoch": 0.5, "learning_rate": 1.0579825532054649e-05, "loss": 1.0261, "step": 5870 }, { "epoch": 0.5, "learning_rate": 1.0577087094814593e-05, "loss": 1.0507, "step": 5871 }, { "epoch": 0.5, "learning_rate": 1.057434861415326e-05, "loss": 1.0442, "step": 5872 }, { "epoch": 0.5, "learning_rate": 1.0571610090276703e-05, "loss": 1.017, "step": 5873 }, { "epoch": 0.5, "learning_rate": 1.0568871523390969e-05, "loss": 1.086, "step": 5874 }, { "epoch": 0.5, "learning_rate": 1.0566132913702116e-05, "loss": 1.0604, "step": 5875 }, { "epoch": 0.5, "learning_rate": 1.0563394261416204e-05, "loss": 1.0406, "step": 5876 }, { "epoch": 0.5, "learning_rate": 1.0560655566739293e-05, "loss": 1.0939, "step": 5877 }, { "epoch": 0.5, "learning_rate": 1.055791682987745e-05, "loss": 1.0422, "step": 5878 }, { "epoch": 0.5, "learning_rate": 1.0555178051036744e-05, "loss": 1.0632, "step": 5879 }, { "epoch": 0.5, "learning_rate": 1.0552439230423244e-05, "loss": 1.0208, "step": 5880 }, { "epoch": 0.5, "learning_rate": 1.0549700368243025e-05, "loss": 1.0188, "step": 5881 }, { "epoch": 0.5, "learning_rate": 1.0546961464702167e-05, "loss": 1.0853, "step": 5882 }, { "epoch": 0.5, "learning_rate": 1.054422252000675e-05, "loss": 1.0846, "step": 5883 }, { "epoch": 0.5, "learning_rate": 1.054148353436286e-05, "loss": 0.9795, "step": 5884 }, { "epoch": 0.5, "learning_rate": 1.0538744507976582e-05, "loss": 1.0335, "step": 5885 }, { "epoch": 0.5, "learning_rate": 1.0536005441054005e-05, "loss": 1.0025, "step": 5886 }, { "epoch": 0.5, "learning_rate": 1.0533266333801224e-05, "loss": 1.1262, "step": 5887 }, { "epoch": 0.5, "learning_rate": 1.0530527186424338e-05, "loss": 1.0333, "step": 5888 }, { "epoch": 0.5, "learning_rate": 1.0527787999129445e-05, "loss": 1.0242, "step": 5889 }, { "epoch": 0.5, "learning_rate": 1.052504877212264e-05, "loss": 1.0678, "step": 5890 }, { "epoch": 0.5, "learning_rate": 1.0522309505610038e-05, "loss": 1.0103, "step": 5891 }, { "epoch": 0.5, "learning_rate": 1.0519570199797742e-05, "loss": 1.0541, "step": 5892 }, { "epoch": 0.5, "learning_rate": 1.0516830854891866e-05, "loss": 1.0258, "step": 5893 }, { "epoch": 0.5, "learning_rate": 1.0514091471098523e-05, "loss": 1.0492, "step": 5894 }, { "epoch": 0.5, "learning_rate": 1.0511352048623829e-05, "loss": 1.0666, "step": 5895 }, { "epoch": 0.5, "learning_rate": 1.0508612587673906e-05, "loss": 0.9982, "step": 5896 }, { "epoch": 0.5, "learning_rate": 1.0505873088454876e-05, "loss": 1.0959, "step": 5897 }, { "epoch": 0.5, "learning_rate": 1.0503133551172866e-05, "loss": 0.9923, "step": 5898 }, { "epoch": 0.5, "learning_rate": 1.0500393976034003e-05, "loss": 1.0533, "step": 5899 }, { "epoch": 0.5, "learning_rate": 1.0497654363244418e-05, "loss": 1.0209, "step": 5900 }, { "epoch": 0.5, "learning_rate": 1.049491471301025e-05, "loss": 1.0273, "step": 5901 }, { "epoch": 0.5, "learning_rate": 1.049217502553763e-05, "loss": 1.0629, "step": 5902 }, { "epoch": 0.5, "learning_rate": 1.0489435301032701e-05, "loss": 1.0507, "step": 5903 }, { "epoch": 0.5, "learning_rate": 1.0486695539701608e-05, "loss": 1.0804, "step": 5904 }, { "epoch": 0.5, "learning_rate": 1.0483955741750491e-05, "loss": 1.0386, "step": 5905 }, { "epoch": 0.5, "learning_rate": 1.0481215907385503e-05, "loss": 1.0357, "step": 5906 }, { "epoch": 0.5, "learning_rate": 1.0478476036812795e-05, "loss": 1.0496, "step": 5907 }, { "epoch": 0.5, "learning_rate": 1.0475736130238516e-05, "loss": 1.0189, "step": 5908 }, { "epoch": 0.5, "learning_rate": 1.0472996187868826e-05, "loss": 1.0789, "step": 5909 }, { "epoch": 0.5, "learning_rate": 1.0470256209909886e-05, "loss": 1.0514, "step": 5910 }, { "epoch": 0.5, "learning_rate": 1.0467516196567854e-05, "loss": 1.0689, "step": 5911 }, { "epoch": 0.5, "learning_rate": 1.0464776148048899e-05, "loss": 1.0743, "step": 5912 }, { "epoch": 0.5, "learning_rate": 1.0462036064559182e-05, "loss": 0.9954, "step": 5913 }, { "epoch": 0.5, "learning_rate": 1.045929594630488e-05, "loss": 1.0631, "step": 5914 }, { "epoch": 0.5, "learning_rate": 1.045655579349216e-05, "loss": 1.0613, "step": 5915 }, { "epoch": 0.5, "learning_rate": 1.0453815606327198e-05, "loss": 1.022, "step": 5916 }, { "epoch": 0.5, "learning_rate": 1.0451075385016172e-05, "loss": 0.9956, "step": 5917 }, { "epoch": 0.5, "learning_rate": 1.0448335129765261e-05, "loss": 1.0228, "step": 5918 }, { "epoch": 0.5, "learning_rate": 1.044559484078065e-05, "loss": 1.0404, "step": 5919 }, { "epoch": 0.5, "learning_rate": 1.0442854518268524e-05, "loss": 1.0437, "step": 5920 }, { "epoch": 0.5, "learning_rate": 1.0440114162435071e-05, "loss": 1.0639, "step": 5921 }, { "epoch": 0.5, "learning_rate": 1.043737377348648e-05, "loss": 1.0536, "step": 5922 }, { "epoch": 0.5, "learning_rate": 1.043463335162894e-05, "loss": 1.0385, "step": 5923 }, { "epoch": 0.5, "learning_rate": 1.0431892897068655e-05, "loss": 0.8812, "step": 5924 }, { "epoch": 0.5, "learning_rate": 1.0429152410011816e-05, "loss": 1.0904, "step": 5925 }, { "epoch": 0.5, "learning_rate": 1.0426411890664623e-05, "loss": 1.0563, "step": 5926 }, { "epoch": 0.5, "learning_rate": 1.0423671339233286e-05, "loss": 1.0812, "step": 5927 }, { "epoch": 0.5, "learning_rate": 1.0420930755924e-05, "loss": 1.0576, "step": 5928 }, { "epoch": 0.5, "learning_rate": 1.0418190140942981e-05, "loss": 1.0547, "step": 5929 }, { "epoch": 0.5, "learning_rate": 1.0415449494496433e-05, "loss": 1.0664, "step": 5930 }, { "epoch": 0.5, "learning_rate": 1.0412708816790569e-05, "loss": 1.0509, "step": 5931 }, { "epoch": 0.5, "learning_rate": 1.0409968108031605e-05, "loss": 1.0293, "step": 5932 }, { "epoch": 0.5, "learning_rate": 1.0407227368425759e-05, "loss": 1.0643, "step": 5933 }, { "epoch": 0.5, "learning_rate": 1.0404486598179247e-05, "loss": 1.0421, "step": 5934 }, { "epoch": 0.5, "learning_rate": 1.0401745797498293e-05, "loss": 1.0799, "step": 5935 }, { "epoch": 0.5, "learning_rate": 1.039900496658912e-05, "loss": 1.0698, "step": 5936 }, { "epoch": 0.5, "learning_rate": 1.0396264105657952e-05, "loss": 1.0119, "step": 5937 }, { "epoch": 0.5, "learning_rate": 1.0393523214911021e-05, "loss": 1.0424, "step": 5938 }, { "epoch": 0.5, "learning_rate": 1.0390782294554553e-05, "loss": 1.0495, "step": 5939 }, { "epoch": 0.5, "learning_rate": 1.0388041344794787e-05, "loss": 1.0122, "step": 5940 }, { "epoch": 0.5, "learning_rate": 1.0385300365837952e-05, "loss": 1.0596, "step": 5941 }, { "epoch": 0.5, "learning_rate": 1.038255935789029e-05, "loss": 1.0456, "step": 5942 }, { "epoch": 0.5, "learning_rate": 1.0379818321158035e-05, "loss": 1.0596, "step": 5943 }, { "epoch": 0.5, "learning_rate": 1.0377077255847431e-05, "loss": 1.0197, "step": 5944 }, { "epoch": 0.5, "learning_rate": 1.0374336162164722e-05, "loss": 1.027, "step": 5945 }, { "epoch": 0.5, "learning_rate": 1.0371595040316156e-05, "loss": 0.974, "step": 5946 }, { "epoch": 0.5, "learning_rate": 1.0368853890507977e-05, "loss": 0.9998, "step": 5947 }, { "epoch": 0.5, "learning_rate": 1.0366112712946437e-05, "loss": 1.0606, "step": 5948 }, { "epoch": 0.5, "learning_rate": 1.0363371507837787e-05, "loss": 0.9752, "step": 5949 }, { "epoch": 0.5, "learning_rate": 1.0360630275388281e-05, "loss": 1.0688, "step": 5950 }, { "epoch": 0.5, "learning_rate": 1.0357889015804179e-05, "loss": 1.004, "step": 5951 }, { "epoch": 0.5, "learning_rate": 1.0355147729291733e-05, "loss": 1.0414, "step": 5952 }, { "epoch": 0.5, "learning_rate": 1.0352406416057208e-05, "loss": 1.0006, "step": 5953 }, { "epoch": 0.5, "learning_rate": 1.034966507630687e-05, "loss": 1.0895, "step": 5954 }, { "epoch": 0.5, "learning_rate": 1.0346923710246974e-05, "loss": 1.0179, "step": 5955 }, { "epoch": 0.5, "learning_rate": 1.0344182318083791e-05, "loss": 1.0117, "step": 5956 }, { "epoch": 0.5, "learning_rate": 1.034144090002359e-05, "loss": 1.0534, "step": 5957 }, { "epoch": 0.5, "learning_rate": 1.0338699456272639e-05, "loss": 1.0528, "step": 5958 }, { "epoch": 0.5, "learning_rate": 1.033595798703721e-05, "loss": 0.9964, "step": 5959 }, { "epoch": 0.5, "learning_rate": 1.033321649252358e-05, "loss": 0.8452, "step": 5960 }, { "epoch": 0.5, "learning_rate": 1.0330474972938023e-05, "loss": 1.0478, "step": 5961 }, { "epoch": 0.5, "learning_rate": 1.0327733428486817e-05, "loss": 1.0198, "step": 5962 }, { "epoch": 0.5, "learning_rate": 1.0324991859376244e-05, "loss": 0.993, "step": 5963 }, { "epoch": 0.51, "learning_rate": 1.032225026581258e-05, "loss": 1.0201, "step": 5964 }, { "epoch": 0.51, "learning_rate": 1.0319508648002114e-05, "loss": 1.007, "step": 5965 }, { "epoch": 0.51, "learning_rate": 1.0316767006151127e-05, "loss": 1.0135, "step": 5966 }, { "epoch": 0.51, "learning_rate": 1.031402534046591e-05, "loss": 1.063, "step": 5967 }, { "epoch": 0.51, "learning_rate": 1.031128365115275e-05, "loss": 1.0007, "step": 5968 }, { "epoch": 0.51, "learning_rate": 1.0308541938417938e-05, "loss": 1.0621, "step": 5969 }, { "epoch": 0.51, "learning_rate": 1.0305800202467763e-05, "loss": 1.0359, "step": 5970 }, { "epoch": 0.51, "learning_rate": 1.0303058443508524e-05, "loss": 0.8618, "step": 5971 }, { "epoch": 0.51, "learning_rate": 1.0300316661746517e-05, "loss": 1.0502, "step": 5972 }, { "epoch": 0.51, "learning_rate": 1.0297574857388033e-05, "loss": 1.0289, "step": 5973 }, { "epoch": 0.51, "learning_rate": 1.0294833030639378e-05, "loss": 1.08, "step": 5974 }, { "epoch": 0.51, "learning_rate": 1.0292091181706851e-05, "loss": 1.0555, "step": 5975 }, { "epoch": 0.51, "learning_rate": 1.0289349310796752e-05, "loss": 0.8934, "step": 5976 }, { "epoch": 0.51, "learning_rate": 1.028660741811539e-05, "loss": 1.0903, "step": 5977 }, { "epoch": 0.51, "learning_rate": 1.0283865503869066e-05, "loss": 1.0475, "step": 5978 }, { "epoch": 0.51, "learning_rate": 1.0281123568264091e-05, "loss": 1.0603, "step": 5979 }, { "epoch": 0.51, "learning_rate": 1.0278381611506774e-05, "loss": 1.0573, "step": 5980 }, { "epoch": 0.51, "learning_rate": 1.0275639633803425e-05, "loss": 1.0456, "step": 5981 }, { "epoch": 0.51, "learning_rate": 1.0272897635360357e-05, "loss": 0.9621, "step": 5982 }, { "epoch": 0.51, "learning_rate": 1.027015561638388e-05, "loss": 1.0186, "step": 5983 }, { "epoch": 0.51, "learning_rate": 1.0267413577080315e-05, "loss": 1.0242, "step": 5984 }, { "epoch": 0.51, "learning_rate": 1.0264671517655975e-05, "loss": 0.9858, "step": 5985 }, { "epoch": 0.51, "learning_rate": 1.026192943831718e-05, "loss": 1.0698, "step": 5986 }, { "epoch": 0.51, "learning_rate": 1.025918733927025e-05, "loss": 1.0148, "step": 5987 }, { "epoch": 0.51, "learning_rate": 1.0256445220721508e-05, "loss": 0.9891, "step": 5988 }, { "epoch": 0.51, "learning_rate": 1.0253703082877271e-05, "loss": 1.0362, "step": 5989 }, { "epoch": 0.51, "learning_rate": 1.0250960925943871e-05, "loss": 1.0295, "step": 5990 }, { "epoch": 0.51, "learning_rate": 1.024821875012763e-05, "loss": 1.0342, "step": 5991 }, { "epoch": 0.51, "learning_rate": 1.0245476555634875e-05, "loss": 1.0278, "step": 5992 }, { "epoch": 0.51, "learning_rate": 1.0242734342671933e-05, "loss": 1.0844, "step": 5993 }, { "epoch": 0.51, "learning_rate": 1.0239992111445138e-05, "loss": 1.0443, "step": 5994 }, { "epoch": 0.51, "learning_rate": 1.023724986216082e-05, "loss": 1.0473, "step": 5995 }, { "epoch": 0.51, "learning_rate": 1.0234507595025313e-05, "loss": 1.0196, "step": 5996 }, { "epoch": 0.51, "learning_rate": 1.0231765310244948e-05, "loss": 1.0202, "step": 5997 }, { "epoch": 0.51, "learning_rate": 1.022902300802606e-05, "loss": 0.9975, "step": 5998 }, { "epoch": 0.51, "learning_rate": 1.022628068857499e-05, "loss": 0.9706, "step": 5999 }, { "epoch": 0.51, "learning_rate": 1.0223538352098071e-05, "loss": 1.0982, "step": 6000 }, { "epoch": 0.51, "learning_rate": 1.0220795998801644e-05, "loss": 1.0613, "step": 6001 }, { "epoch": 0.51, "learning_rate": 1.0218053628892056e-05, "loss": 1.0743, "step": 6002 }, { "epoch": 0.51, "learning_rate": 1.0215311242575639e-05, "loss": 0.9859, "step": 6003 }, { "epoch": 0.51, "learning_rate": 1.0212568840058736e-05, "loss": 0.9923, "step": 6004 }, { "epoch": 0.51, "learning_rate": 1.0209826421547701e-05, "loss": 1.0295, "step": 6005 }, { "epoch": 0.51, "learning_rate": 1.0207083987248873e-05, "loss": 0.9072, "step": 6006 }, { "epoch": 0.51, "learning_rate": 1.02043415373686e-05, "loss": 1.0552, "step": 6007 }, { "epoch": 0.51, "learning_rate": 1.0201599072113234e-05, "loss": 1.0177, "step": 6008 }, { "epoch": 0.51, "learning_rate": 1.0198856591689115e-05, "loss": 1.0803, "step": 6009 }, { "epoch": 0.51, "learning_rate": 1.0196114096302597e-05, "loss": 0.9811, "step": 6010 }, { "epoch": 0.51, "learning_rate": 1.0193371586160032e-05, "loss": 1.1033, "step": 6011 }, { "epoch": 0.51, "learning_rate": 1.0190629061467775e-05, "loss": 1.0083, "step": 6012 }, { "epoch": 0.51, "learning_rate": 1.0187886522432173e-05, "loss": 1.0545, "step": 6013 }, { "epoch": 0.51, "learning_rate": 1.0185143969259586e-05, "loss": 1.0327, "step": 6014 }, { "epoch": 0.51, "learning_rate": 1.0182401402156368e-05, "loss": 1.0724, "step": 6015 }, { "epoch": 0.51, "learning_rate": 1.0179658821328876e-05, "loss": 1.0281, "step": 6016 }, { "epoch": 0.51, "learning_rate": 1.0176916226983467e-05, "loss": 1.0496, "step": 6017 }, { "epoch": 0.51, "learning_rate": 1.0174173619326497e-05, "loss": 1.0083, "step": 6018 }, { "epoch": 0.51, "learning_rate": 1.0171430998564334e-05, "loss": 1.0231, "step": 6019 }, { "epoch": 0.51, "learning_rate": 1.016868836490333e-05, "loss": 1.0479, "step": 6020 }, { "epoch": 0.51, "learning_rate": 1.016594571854985e-05, "loss": 1.0338, "step": 6021 }, { "epoch": 0.51, "learning_rate": 1.0163203059710259e-05, "loss": 1.0483, "step": 6022 }, { "epoch": 0.51, "learning_rate": 1.0160460388590918e-05, "loss": 1.0532, "step": 6023 }, { "epoch": 0.51, "learning_rate": 1.0157717705398189e-05, "loss": 1.0524, "step": 6024 }, { "epoch": 0.51, "learning_rate": 1.0154975010338442e-05, "loss": 1.0587, "step": 6025 }, { "epoch": 0.51, "learning_rate": 1.0152232303618042e-05, "loss": 1.0516, "step": 6026 }, { "epoch": 0.51, "learning_rate": 1.0149489585443352e-05, "loss": 0.9842, "step": 6027 }, { "epoch": 0.51, "learning_rate": 1.0146746856020746e-05, "loss": 0.9626, "step": 6028 }, { "epoch": 0.51, "learning_rate": 1.014400411555659e-05, "loss": 0.839, "step": 6029 }, { "epoch": 0.51, "learning_rate": 1.0141261364257253e-05, "loss": 1.0033, "step": 6030 }, { "epoch": 0.51, "learning_rate": 1.0138518602329109e-05, "loss": 1.0461, "step": 6031 }, { "epoch": 0.51, "learning_rate": 1.0135775829978525e-05, "loss": 1.0226, "step": 6032 }, { "epoch": 0.51, "learning_rate": 1.0133033047411876e-05, "loss": 0.997, "step": 6033 }, { "epoch": 0.51, "learning_rate": 1.0130290254835537e-05, "loss": 1.0165, "step": 6034 }, { "epoch": 0.51, "learning_rate": 1.0127547452455875e-05, "loss": 1.0777, "step": 6035 }, { "epoch": 0.51, "learning_rate": 1.0124804640479268e-05, "loss": 1.0947, "step": 6036 }, { "epoch": 0.51, "learning_rate": 1.0122061819112093e-05, "loss": 1.0196, "step": 6037 }, { "epoch": 0.51, "learning_rate": 1.0119318988560722e-05, "loss": 1.0083, "step": 6038 }, { "epoch": 0.51, "learning_rate": 1.0116576149031532e-05, "loss": 1.0205, "step": 6039 }, { "epoch": 0.51, "learning_rate": 1.0113833300730904e-05, "loss": 1.039, "step": 6040 }, { "epoch": 0.51, "learning_rate": 1.0111090443865212e-05, "loss": 1.0317, "step": 6041 }, { "epoch": 0.51, "learning_rate": 1.0108347578640835e-05, "loss": 1.0512, "step": 6042 }, { "epoch": 0.51, "learning_rate": 1.0105604705264153e-05, "loss": 1.0719, "step": 6043 }, { "epoch": 0.51, "learning_rate": 1.0102861823941548e-05, "loss": 1.0355, "step": 6044 }, { "epoch": 0.51, "learning_rate": 1.0100118934879394e-05, "loss": 1.0429, "step": 6045 }, { "epoch": 0.51, "learning_rate": 1.0097376038284079e-05, "loss": 1.0517, "step": 6046 }, { "epoch": 0.51, "learning_rate": 1.009463313436198e-05, "loss": 1.0311, "step": 6047 }, { "epoch": 0.51, "learning_rate": 1.0091890223319478e-05, "loss": 1.0446, "step": 6048 }, { "epoch": 0.51, "learning_rate": 1.0089147305362963e-05, "loss": 0.8766, "step": 6049 }, { "epoch": 0.51, "learning_rate": 1.0086404380698809e-05, "loss": 0.9396, "step": 6050 }, { "epoch": 0.51, "learning_rate": 1.0083661449533402e-05, "loss": 1.0677, "step": 6051 }, { "epoch": 0.51, "learning_rate": 1.0080918512073131e-05, "loss": 1.0454, "step": 6052 }, { "epoch": 0.51, "learning_rate": 1.0078175568524373e-05, "loss": 1.0862, "step": 6053 }, { "epoch": 0.51, "learning_rate": 1.007543261909352e-05, "loss": 0.9939, "step": 6054 }, { "epoch": 0.51, "learning_rate": 1.007268966398695e-05, "loss": 1.0348, "step": 6055 }, { "epoch": 0.51, "learning_rate": 1.0069946703411053e-05, "loss": 1.0484, "step": 6056 }, { "epoch": 0.51, "learning_rate": 1.006720373757222e-05, "loss": 1.0197, "step": 6057 }, { "epoch": 0.51, "learning_rate": 1.006446076667683e-05, "loss": 1.0667, "step": 6058 }, { "epoch": 0.51, "learning_rate": 1.0061717790931274e-05, "loss": 1.0434, "step": 6059 }, { "epoch": 0.51, "learning_rate": 1.005897481054194e-05, "loss": 1.0094, "step": 6060 }, { "epoch": 0.51, "learning_rate": 1.0056231825715212e-05, "loss": 1.0269, "step": 6061 }, { "epoch": 0.51, "learning_rate": 1.0053488836657483e-05, "loss": 1.0743, "step": 6062 }, { "epoch": 0.51, "learning_rate": 1.0050745843575137e-05, "loss": 0.9988, "step": 6063 }, { "epoch": 0.51, "learning_rate": 1.0048002846674562e-05, "loss": 1.0891, "step": 6064 }, { "epoch": 0.51, "learning_rate": 1.0045259846162151e-05, "loss": 1.0296, "step": 6065 }, { "epoch": 0.51, "learning_rate": 1.0042516842244294e-05, "loss": 1.0043, "step": 6066 }, { "epoch": 0.51, "learning_rate": 1.0039773835127375e-05, "loss": 1.0496, "step": 6067 }, { "epoch": 0.51, "learning_rate": 1.0037030825017791e-05, "loss": 0.9923, "step": 6068 }, { "epoch": 0.51, "learning_rate": 1.0034287812121923e-05, "loss": 1.0274, "step": 6069 }, { "epoch": 0.51, "learning_rate": 1.0031544796646168e-05, "loss": 1.0386, "step": 6070 }, { "epoch": 0.51, "learning_rate": 1.0028801778796917e-05, "loss": 1.037, "step": 6071 }, { "epoch": 0.51, "learning_rate": 1.0026058758780555e-05, "loss": 1.0428, "step": 6072 }, { "epoch": 0.51, "learning_rate": 1.0023315736803478e-05, "loss": 1.0957, "step": 6073 }, { "epoch": 0.51, "learning_rate": 1.0020572713072071e-05, "loss": 1.015, "step": 6074 }, { "epoch": 0.51, "learning_rate": 1.0017829687792732e-05, "loss": 1.0562, "step": 6075 }, { "epoch": 0.51, "learning_rate": 1.0015086661171852e-05, "loss": 1.0086, "step": 6076 }, { "epoch": 0.51, "learning_rate": 1.0012343633415815e-05, "loss": 0.9894, "step": 6077 }, { "epoch": 0.51, "learning_rate": 1.0009600604731017e-05, "loss": 1.0269, "step": 6078 }, { "epoch": 0.51, "learning_rate": 1.000685757532385e-05, "loss": 0.9993, "step": 6079 }, { "epoch": 0.51, "learning_rate": 1.0004114545400703e-05, "loss": 1.0267, "step": 6080 }, { "epoch": 0.51, "learning_rate": 1.0001371515167967e-05, "loss": 1.0362, "step": 6081 }, { "epoch": 0.52, "learning_rate": 9.998628484832034e-06, "loss": 0.8014, "step": 6082 }, { "epoch": 0.52, "learning_rate": 9.995885454599304e-06, "loss": 1.0323, "step": 6083 }, { "epoch": 0.52, "learning_rate": 9.993142424676154e-06, "loss": 1.0573, "step": 6084 }, { "epoch": 0.52, "learning_rate": 9.990399395268988e-06, "loss": 1.0181, "step": 6085 }, { "epoch": 0.52, "learning_rate": 9.987656366584188e-06, "loss": 1.0519, "step": 6086 }, { "epoch": 0.52, "learning_rate": 9.984913338828153e-06, "loss": 0.9995, "step": 6087 }, { "epoch": 0.52, "learning_rate": 9.98217031220727e-06, "loss": 1.0345, "step": 6088 }, { "epoch": 0.52, "learning_rate": 9.97942728692793e-06, "loss": 1.0994, "step": 6089 }, { "epoch": 0.52, "learning_rate": 9.976684263196525e-06, "loss": 1.044, "step": 6090 }, { "epoch": 0.52, "learning_rate": 9.97394124121945e-06, "loss": 1.0504, "step": 6091 }, { "epoch": 0.52, "learning_rate": 9.971198221203086e-06, "loss": 1.0068, "step": 6092 }, { "epoch": 0.52, "learning_rate": 9.968455203353837e-06, "loss": 0.9333, "step": 6093 }, { "epoch": 0.52, "learning_rate": 9.965712187878079e-06, "loss": 1.0529, "step": 6094 }, { "epoch": 0.52, "learning_rate": 9.96296917498221e-06, "loss": 1.0739, "step": 6095 }, { "epoch": 0.52, "learning_rate": 9.960226164872626e-06, "loss": 1.0147, "step": 6096 }, { "epoch": 0.52, "learning_rate": 9.957483157755708e-06, "loss": 1.0374, "step": 6097 }, { "epoch": 0.52, "learning_rate": 9.95474015383785e-06, "loss": 1.0278, "step": 6098 }, { "epoch": 0.52, "learning_rate": 9.951997153325438e-06, "loss": 1.0619, "step": 6099 }, { "epoch": 0.52, "learning_rate": 9.949254156424867e-06, "loss": 1.0902, "step": 6100 }, { "epoch": 0.52, "learning_rate": 9.94651116334252e-06, "loss": 0.9877, "step": 6101 }, { "epoch": 0.52, "learning_rate": 9.94376817428479e-06, "loss": 1.0202, "step": 6102 }, { "epoch": 0.52, "learning_rate": 9.941025189458062e-06, "loss": 1.0415, "step": 6103 }, { "epoch": 0.52, "learning_rate": 9.93828220906873e-06, "loss": 0.9929, "step": 6104 }, { "epoch": 0.52, "learning_rate": 9.93553923332317e-06, "loss": 1.0525, "step": 6105 }, { "epoch": 0.52, "learning_rate": 9.932796262427785e-06, "loss": 1.0088, "step": 6106 }, { "epoch": 0.52, "learning_rate": 9.930053296588945e-06, "loss": 0.8284, "step": 6107 }, { "epoch": 0.52, "learning_rate": 9.927310336013054e-06, "loss": 1.0158, "step": 6108 }, { "epoch": 0.52, "learning_rate": 9.924567380906484e-06, "loss": 1.0357, "step": 6109 }, { "epoch": 0.52, "learning_rate": 9.921824431475632e-06, "loss": 0.996, "step": 6110 }, { "epoch": 0.52, "learning_rate": 9.919081487926874e-06, "loss": 1.0868, "step": 6111 }, { "epoch": 0.52, "learning_rate": 9.916338550466601e-06, "loss": 1.0183, "step": 6112 }, { "epoch": 0.52, "learning_rate": 9.913595619301194e-06, "loss": 1.0254, "step": 6113 }, { "epoch": 0.52, "learning_rate": 9.910852694637043e-06, "loss": 1.0581, "step": 6114 }, { "epoch": 0.52, "learning_rate": 9.908109776680524e-06, "loss": 1.0348, "step": 6115 }, { "epoch": 0.52, "learning_rate": 9.905366865638025e-06, "loss": 1.0687, "step": 6116 }, { "epoch": 0.52, "learning_rate": 9.902623961715925e-06, "loss": 1.0109, "step": 6117 }, { "epoch": 0.52, "learning_rate": 9.89988106512061e-06, "loss": 1.0406, "step": 6118 }, { "epoch": 0.52, "learning_rate": 9.897138176058456e-06, "loss": 1.0515, "step": 6119 }, { "epoch": 0.52, "learning_rate": 9.894395294735852e-06, "loss": 1.0788, "step": 6120 }, { "epoch": 0.52, "learning_rate": 9.891652421359166e-06, "loss": 1.0371, "step": 6121 }, { "epoch": 0.52, "learning_rate": 9.88890955613479e-06, "loss": 1.0892, "step": 6122 }, { "epoch": 0.52, "learning_rate": 9.886166699269099e-06, "loss": 1.0019, "step": 6123 }, { "epoch": 0.52, "learning_rate": 9.883423850968468e-06, "loss": 1.0332, "step": 6124 }, { "epoch": 0.52, "learning_rate": 9.880681011439281e-06, "loss": 1.093, "step": 6125 }, { "epoch": 0.52, "learning_rate": 9.87793818088791e-06, "loss": 1.0476, "step": 6126 }, { "epoch": 0.52, "learning_rate": 9.875195359520734e-06, "loss": 1.0345, "step": 6127 }, { "epoch": 0.52, "learning_rate": 9.872452547544128e-06, "loss": 1.0742, "step": 6128 }, { "epoch": 0.52, "learning_rate": 9.869709745164466e-06, "loss": 1.0474, "step": 6129 }, { "epoch": 0.52, "learning_rate": 9.866966952588124e-06, "loss": 1.0573, "step": 6130 }, { "epoch": 0.52, "learning_rate": 9.864224170021476e-06, "loss": 1.0487, "step": 6131 }, { "epoch": 0.52, "learning_rate": 9.861481397670893e-06, "loss": 1.0388, "step": 6132 }, { "epoch": 0.52, "learning_rate": 9.85873863574275e-06, "loss": 1.0386, "step": 6133 }, { "epoch": 0.52, "learning_rate": 9.85599588444341e-06, "loss": 1.0286, "step": 6134 }, { "epoch": 0.52, "learning_rate": 9.853253143979257e-06, "loss": 1.0357, "step": 6135 }, { "epoch": 0.52, "learning_rate": 9.85051041455665e-06, "loss": 1.0311, "step": 6136 }, { "epoch": 0.52, "learning_rate": 9.847767696381965e-06, "loss": 1.0137, "step": 6137 }, { "epoch": 0.52, "learning_rate": 9.845024989661561e-06, "loss": 1.0751, "step": 6138 }, { "epoch": 0.52, "learning_rate": 9.842282294601814e-06, "loss": 1.0584, "step": 6139 }, { "epoch": 0.52, "learning_rate": 9.839539611409086e-06, "loss": 1.014, "step": 6140 }, { "epoch": 0.52, "learning_rate": 9.836796940289746e-06, "loss": 1.0056, "step": 6141 }, { "epoch": 0.52, "learning_rate": 9.834054281450153e-06, "loss": 1.028, "step": 6142 }, { "epoch": 0.52, "learning_rate": 9.831311635096673e-06, "loss": 1.0564, "step": 6143 }, { "epoch": 0.52, "learning_rate": 9.828569001435671e-06, "loss": 0.9718, "step": 6144 }, { "epoch": 0.52, "learning_rate": 9.825826380673506e-06, "loss": 1.051, "step": 6145 }, { "epoch": 0.52, "learning_rate": 9.823083773016538e-06, "loss": 1.0725, "step": 6146 }, { "epoch": 0.52, "learning_rate": 9.820341178671126e-06, "loss": 1.0036, "step": 6147 }, { "epoch": 0.52, "learning_rate": 9.817598597843635e-06, "loss": 1.0826, "step": 6148 }, { "epoch": 0.52, "learning_rate": 9.814856030740416e-06, "loss": 1.0515, "step": 6149 }, { "epoch": 0.52, "learning_rate": 9.81211347756783e-06, "loss": 1.0294, "step": 6150 }, { "epoch": 0.52, "learning_rate": 9.809370938532227e-06, "loss": 0.968, "step": 6151 }, { "epoch": 0.52, "learning_rate": 9.806628413839971e-06, "loss": 1.0098, "step": 6152 }, { "epoch": 0.52, "learning_rate": 9.803885903697407e-06, "loss": 0.9908, "step": 6153 }, { "epoch": 0.52, "learning_rate": 9.80114340831089e-06, "loss": 1.0168, "step": 6154 }, { "epoch": 0.52, "learning_rate": 9.798400927886771e-06, "loss": 0.9886, "step": 6155 }, { "epoch": 0.52, "learning_rate": 9.795658462631402e-06, "loss": 0.9646, "step": 6156 }, { "epoch": 0.52, "learning_rate": 9.792916012751127e-06, "loss": 0.9881, "step": 6157 }, { "epoch": 0.52, "learning_rate": 9.7901735784523e-06, "loss": 1.0581, "step": 6158 }, { "epoch": 0.52, "learning_rate": 9.787431159941263e-06, "loss": 1.0226, "step": 6159 }, { "epoch": 0.52, "learning_rate": 9.784688757424368e-06, "loss": 1.0364, "step": 6160 }, { "epoch": 0.52, "learning_rate": 9.78194637110795e-06, "loss": 1.0383, "step": 6161 }, { "epoch": 0.52, "learning_rate": 9.779204001198359e-06, "loss": 1.0332, "step": 6162 }, { "epoch": 0.52, "learning_rate": 9.77646164790193e-06, "loss": 1.0362, "step": 6163 }, { "epoch": 0.52, "learning_rate": 9.773719311425017e-06, "loss": 1.009, "step": 6164 }, { "epoch": 0.52, "learning_rate": 9.770976991973943e-06, "loss": 1.0199, "step": 6165 }, { "epoch": 0.52, "learning_rate": 9.768234689755057e-06, "loss": 1.0019, "step": 6166 }, { "epoch": 0.52, "learning_rate": 9.765492404974688e-06, "loss": 1.0453, "step": 6167 }, { "epoch": 0.52, "learning_rate": 9.762750137839183e-06, "loss": 1.0424, "step": 6168 }, { "epoch": 0.52, "learning_rate": 9.760007888554864e-06, "loss": 1.0213, "step": 6169 }, { "epoch": 0.52, "learning_rate": 9.757265657328069e-06, "loss": 1.0663, "step": 6170 }, { "epoch": 0.52, "learning_rate": 9.75452344436513e-06, "loss": 1.0358, "step": 6171 }, { "epoch": 0.52, "learning_rate": 9.751781249872377e-06, "loss": 1.0416, "step": 6172 }, { "epoch": 0.52, "learning_rate": 9.749039074056132e-06, "loss": 0.9681, "step": 6173 }, { "epoch": 0.52, "learning_rate": 9.746296917122729e-06, "loss": 1.0636, "step": 6174 }, { "epoch": 0.52, "learning_rate": 9.743554779278497e-06, "loss": 1.0863, "step": 6175 }, { "epoch": 0.52, "learning_rate": 9.74081266072975e-06, "loss": 0.9839, "step": 6176 }, { "epoch": 0.52, "learning_rate": 9.738070561682823e-06, "loss": 1.046, "step": 6177 }, { "epoch": 0.52, "learning_rate": 9.735328482344025e-06, "loss": 1.0059, "step": 6178 }, { "epoch": 0.52, "learning_rate": 9.732586422919689e-06, "loss": 1.0627, "step": 6179 }, { "epoch": 0.52, "learning_rate": 9.729844383616123e-06, "loss": 1.0467, "step": 6180 }, { "epoch": 0.52, "learning_rate": 9.727102364639646e-06, "loss": 1.0028, "step": 6181 }, { "epoch": 0.52, "learning_rate": 9.724360366196577e-06, "loss": 1.098, "step": 6182 }, { "epoch": 0.52, "learning_rate": 9.72161838849323e-06, "loss": 1.0589, "step": 6183 }, { "epoch": 0.52, "learning_rate": 9.718876431735909e-06, "loss": 1.0448, "step": 6184 }, { "epoch": 0.52, "learning_rate": 9.716134496130938e-06, "loss": 1.0437, "step": 6185 }, { "epoch": 0.52, "learning_rate": 9.713392581884613e-06, "loss": 1.0473, "step": 6186 }, { "epoch": 0.52, "learning_rate": 9.710650689203253e-06, "loss": 1.0474, "step": 6187 }, { "epoch": 0.52, "learning_rate": 9.707908818293152e-06, "loss": 0.9889, "step": 6188 }, { "epoch": 0.52, "learning_rate": 9.705166969360626e-06, "loss": 0.9954, "step": 6189 }, { "epoch": 0.52, "learning_rate": 9.702425142611968e-06, "loss": 1.062, "step": 6190 }, { "epoch": 0.52, "learning_rate": 9.69968333825349e-06, "loss": 1.0207, "step": 6191 }, { "epoch": 0.52, "learning_rate": 9.696941556491477e-06, "loss": 1.046, "step": 6192 }, { "epoch": 0.52, "learning_rate": 9.69419979753224e-06, "loss": 1.0187, "step": 6193 }, { "epoch": 0.52, "learning_rate": 9.691458061582064e-06, "loss": 1.1202, "step": 6194 }, { "epoch": 0.52, "learning_rate": 9.688716348847253e-06, "loss": 1.0361, "step": 6195 }, { "epoch": 0.52, "learning_rate": 9.685974659534094e-06, "loss": 1.0329, "step": 6196 }, { "epoch": 0.52, "learning_rate": 9.683232993848875e-06, "loss": 0.9953, "step": 6197 }, { "epoch": 0.52, "learning_rate": 9.680491351997889e-06, "loss": 0.9943, "step": 6198 }, { "epoch": 0.52, "learning_rate": 9.677749734187421e-06, "loss": 1.081, "step": 6199 }, { "epoch": 0.53, "learning_rate": 9.675008140623761e-06, "loss": 1.0178, "step": 6200 }, { "epoch": 0.53, "learning_rate": 9.672266571513183e-06, "loss": 1.0459, "step": 6201 }, { "epoch": 0.53, "learning_rate": 9.669525027061979e-06, "loss": 1.0646, "step": 6202 }, { "epoch": 0.53, "learning_rate": 9.666783507476421e-06, "loss": 1.0396, "step": 6203 }, { "epoch": 0.53, "learning_rate": 9.664042012962792e-06, "loss": 1.054, "step": 6204 }, { "epoch": 0.53, "learning_rate": 9.661300543727363e-06, "loss": 0.8645, "step": 6205 }, { "epoch": 0.53, "learning_rate": 9.658559099976414e-06, "loss": 0.953, "step": 6206 }, { "epoch": 0.53, "learning_rate": 9.655817681916212e-06, "loss": 0.9782, "step": 6207 }, { "epoch": 0.53, "learning_rate": 9.653076289753029e-06, "loss": 1.0255, "step": 6208 }, { "epoch": 0.53, "learning_rate": 9.650334923693133e-06, "loss": 1.0228, "step": 6209 }, { "epoch": 0.53, "learning_rate": 9.647593583942794e-06, "loss": 0.9905, "step": 6210 }, { "epoch": 0.53, "learning_rate": 9.644852270708267e-06, "loss": 1.0217, "step": 6211 }, { "epoch": 0.53, "learning_rate": 9.642110984195826e-06, "loss": 1.0732, "step": 6212 }, { "epoch": 0.53, "learning_rate": 9.63936972461172e-06, "loss": 1.0524, "step": 6213 }, { "epoch": 0.53, "learning_rate": 9.63662849216222e-06, "loss": 1.026, "step": 6214 }, { "epoch": 0.53, "learning_rate": 9.633887287053568e-06, "loss": 1.0283, "step": 6215 }, { "epoch": 0.53, "learning_rate": 9.631146109492028e-06, "loss": 1.0527, "step": 6216 }, { "epoch": 0.53, "learning_rate": 9.628404959683849e-06, "loss": 1.0728, "step": 6217 }, { "epoch": 0.53, "learning_rate": 9.625663837835283e-06, "loss": 1.0321, "step": 6218 }, { "epoch": 0.53, "learning_rate": 9.622922744152572e-06, "loss": 1.0259, "step": 6219 }, { "epoch": 0.53, "learning_rate": 9.620181678841972e-06, "loss": 1.0359, "step": 6220 }, { "epoch": 0.53, "learning_rate": 9.617440642109714e-06, "loss": 0.9418, "step": 6221 }, { "epoch": 0.53, "learning_rate": 9.614699634162053e-06, "loss": 1.0952, "step": 6222 }, { "epoch": 0.53, "learning_rate": 9.611958655205217e-06, "loss": 1.0206, "step": 6223 }, { "epoch": 0.53, "learning_rate": 9.60921770544545e-06, "loss": 0.9844, "step": 6224 }, { "epoch": 0.53, "learning_rate": 9.606476785088984e-06, "loss": 1.0276, "step": 6225 }, { "epoch": 0.53, "learning_rate": 9.60373589434205e-06, "loss": 0.9926, "step": 6226 }, { "epoch": 0.53, "learning_rate": 9.600995033410884e-06, "loss": 1.0319, "step": 6227 }, { "epoch": 0.53, "learning_rate": 9.598254202501708e-06, "loss": 1.0799, "step": 6228 }, { "epoch": 0.53, "learning_rate": 9.595513401820755e-06, "loss": 1.0472, "step": 6229 }, { "epoch": 0.53, "learning_rate": 9.592772631574243e-06, "loss": 1.0957, "step": 6230 }, { "epoch": 0.53, "learning_rate": 9.590031891968397e-06, "loss": 0.8828, "step": 6231 }, { "epoch": 0.53, "learning_rate": 9.587291183209431e-06, "loss": 1.0376, "step": 6232 }, { "epoch": 0.53, "learning_rate": 9.58455050550357e-06, "loss": 1.027, "step": 6233 }, { "epoch": 0.53, "learning_rate": 9.581809859057024e-06, "loss": 1.0351, "step": 6234 }, { "epoch": 0.53, "learning_rate": 9.579069244076001e-06, "loss": 1.0165, "step": 6235 }, { "epoch": 0.53, "learning_rate": 9.576328660766718e-06, "loss": 1.0157, "step": 6236 }, { "epoch": 0.53, "learning_rate": 9.57358810933538e-06, "loss": 1.02, "step": 6237 }, { "epoch": 0.53, "learning_rate": 9.570847589988187e-06, "loss": 1.0541, "step": 6238 }, { "epoch": 0.53, "learning_rate": 9.56810710293135e-06, "loss": 1.0273, "step": 6239 }, { "epoch": 0.53, "learning_rate": 9.565366648371061e-06, "loss": 1.0456, "step": 6240 }, { "epoch": 0.53, "learning_rate": 9.562626226513527e-06, "loss": 1.009, "step": 6241 }, { "epoch": 0.53, "learning_rate": 9.559885837564934e-06, "loss": 1.025, "step": 6242 }, { "epoch": 0.53, "learning_rate": 9.55714548173148e-06, "loss": 0.9934, "step": 6243 }, { "epoch": 0.53, "learning_rate": 9.554405159219353e-06, "loss": 1.0163, "step": 6244 }, { "epoch": 0.53, "learning_rate": 9.551664870234742e-06, "loss": 1.0554, "step": 6245 }, { "epoch": 0.53, "learning_rate": 9.548924614983831e-06, "loss": 1.0658, "step": 6246 }, { "epoch": 0.53, "learning_rate": 9.546184393672807e-06, "loss": 1.0147, "step": 6247 }, { "epoch": 0.53, "learning_rate": 9.543444206507843e-06, "loss": 0.9995, "step": 6248 }, { "epoch": 0.53, "learning_rate": 9.540704053695124e-06, "loss": 1.0631, "step": 6249 }, { "epoch": 0.53, "learning_rate": 9.53796393544082e-06, "loss": 0.9905, "step": 6250 }, { "epoch": 0.53, "learning_rate": 9.535223851951103e-06, "loss": 1.0543, "step": 6251 }, { "epoch": 0.53, "learning_rate": 9.532483803432147e-06, "loss": 1.0568, "step": 6252 }, { "epoch": 0.53, "learning_rate": 9.529743790090113e-06, "loss": 0.992, "step": 6253 }, { "epoch": 0.53, "learning_rate": 9.527003812131175e-06, "loss": 1.046, "step": 6254 }, { "epoch": 0.53, "learning_rate": 9.524263869761484e-06, "loss": 0.9827, "step": 6255 }, { "epoch": 0.53, "learning_rate": 9.521523963187209e-06, "loss": 0.8576, "step": 6256 }, { "epoch": 0.53, "learning_rate": 9.518784092614497e-06, "loss": 1.0465, "step": 6257 }, { "epoch": 0.53, "learning_rate": 9.51604425824951e-06, "loss": 1.0683, "step": 6258 }, { "epoch": 0.53, "learning_rate": 9.513304460298393e-06, "loss": 1.0244, "step": 6259 }, { "epoch": 0.53, "learning_rate": 9.5105646989673e-06, "loss": 0.9523, "step": 6260 }, { "epoch": 0.53, "learning_rate": 9.507824974462372e-06, "loss": 0.9845, "step": 6261 }, { "epoch": 0.53, "learning_rate": 9.505085286989754e-06, "loss": 1.0183, "step": 6262 }, { "epoch": 0.53, "learning_rate": 9.502345636755583e-06, "loss": 1.0384, "step": 6263 }, { "epoch": 0.53, "learning_rate": 9.499606023966002e-06, "loss": 1.0366, "step": 6264 }, { "epoch": 0.53, "learning_rate": 9.496866448827137e-06, "loss": 1.0137, "step": 6265 }, { "epoch": 0.53, "learning_rate": 9.494126911545129e-06, "loss": 1.0656, "step": 6266 }, { "epoch": 0.53, "learning_rate": 9.491387412326096e-06, "loss": 1.0189, "step": 6267 }, { "epoch": 0.53, "learning_rate": 9.488647951376176e-06, "loss": 1.0598, "step": 6268 }, { "epoch": 0.53, "learning_rate": 9.48590852890148e-06, "loss": 1.0823, "step": 6269 }, { "epoch": 0.53, "learning_rate": 9.48316914510814e-06, "loss": 1.0527, "step": 6270 }, { "epoch": 0.53, "learning_rate": 9.48042980020226e-06, "loss": 1.0156, "step": 6271 }, { "epoch": 0.53, "learning_rate": 9.477690494389967e-06, "loss": 1.056, "step": 6272 }, { "epoch": 0.53, "learning_rate": 9.474951227877363e-06, "loss": 1.0253, "step": 6273 }, { "epoch": 0.53, "learning_rate": 9.472212000870562e-06, "loss": 0.8052, "step": 6274 }, { "epoch": 0.53, "learning_rate": 9.469472813575666e-06, "loss": 1.0318, "step": 6275 }, { "epoch": 0.53, "learning_rate": 9.466733666198778e-06, "loss": 0.9868, "step": 6276 }, { "epoch": 0.53, "learning_rate": 9.463994558945996e-06, "loss": 1.0178, "step": 6277 }, { "epoch": 0.53, "learning_rate": 9.461255492023418e-06, "loss": 1.1025, "step": 6278 }, { "epoch": 0.53, "learning_rate": 9.458516465637142e-06, "loss": 1.0762, "step": 6279 }, { "epoch": 0.53, "learning_rate": 9.455777479993248e-06, "loss": 1.1085, "step": 6280 }, { "epoch": 0.53, "learning_rate": 9.453038535297834e-06, "loss": 1.0654, "step": 6281 }, { "epoch": 0.53, "learning_rate": 9.450299631756975e-06, "loss": 1.0421, "step": 6282 }, { "epoch": 0.53, "learning_rate": 9.44756076957676e-06, "loss": 1.062, "step": 6283 }, { "epoch": 0.53, "learning_rate": 9.444821948963258e-06, "loss": 1.0585, "step": 6284 }, { "epoch": 0.53, "learning_rate": 9.442083170122552e-06, "loss": 1.0695, "step": 6285 }, { "epoch": 0.53, "learning_rate": 9.439344433260708e-06, "loss": 1.0636, "step": 6286 }, { "epoch": 0.53, "learning_rate": 9.4366057385838e-06, "loss": 1.0554, "step": 6287 }, { "epoch": 0.53, "learning_rate": 9.433867086297886e-06, "loss": 1.0266, "step": 6288 }, { "epoch": 0.53, "learning_rate": 9.431128476609034e-06, "loss": 1.0475, "step": 6289 }, { "epoch": 0.53, "learning_rate": 9.4283899097233e-06, "loss": 1.0177, "step": 6290 }, { "epoch": 0.53, "learning_rate": 9.425651385846743e-06, "loss": 1.013, "step": 6291 }, { "epoch": 0.53, "learning_rate": 9.42291290518541e-06, "loss": 0.9711, "step": 6292 }, { "epoch": 0.53, "learning_rate": 9.420174467945355e-06, "loss": 1.0569, "step": 6293 }, { "epoch": 0.53, "learning_rate": 9.41743607433262e-06, "loss": 1.0382, "step": 6294 }, { "epoch": 0.53, "learning_rate": 9.414697724553254e-06, "loss": 1.012, "step": 6295 }, { "epoch": 0.53, "learning_rate": 9.411959418813286e-06, "loss": 0.8656, "step": 6296 }, { "epoch": 0.53, "learning_rate": 9.409221157318762e-06, "loss": 0.9688, "step": 6297 }, { "epoch": 0.53, "learning_rate": 9.40648294027571e-06, "loss": 0.9783, "step": 6298 }, { "epoch": 0.53, "learning_rate": 9.403744767890161e-06, "loss": 1.0157, "step": 6299 }, { "epoch": 0.53, "learning_rate": 9.401006640368138e-06, "loss": 1.0289, "step": 6300 }, { "epoch": 0.53, "learning_rate": 9.398268557915668e-06, "loss": 1.0421, "step": 6301 }, { "epoch": 0.53, "learning_rate": 9.395530520738768e-06, "loss": 1.0317, "step": 6302 }, { "epoch": 0.53, "learning_rate": 9.39279252904345e-06, "loss": 1.0287, "step": 6303 }, { "epoch": 0.53, "learning_rate": 9.390054583035737e-06, "loss": 0.8703, "step": 6304 }, { "epoch": 0.53, "learning_rate": 9.387316682921623e-06, "loss": 0.996, "step": 6305 }, { "epoch": 0.53, "learning_rate": 9.384578828907128e-06, "loss": 1.0944, "step": 6306 }, { "epoch": 0.53, "learning_rate": 9.381841021198244e-06, "loss": 1.0561, "step": 6307 }, { "epoch": 0.53, "learning_rate": 9.379103260000978e-06, "loss": 1.0325, "step": 6308 }, { "epoch": 0.53, "learning_rate": 9.376365545521314e-06, "loss": 1.0153, "step": 6309 }, { "epoch": 0.53, "learning_rate": 9.373627877965254e-06, "loss": 1.0008, "step": 6310 }, { "epoch": 0.53, "learning_rate": 9.37089025753878e-06, "loss": 1.0105, "step": 6311 }, { "epoch": 0.53, "learning_rate": 9.36815268444788e-06, "loss": 1.0326, "step": 6312 }, { "epoch": 0.53, "learning_rate": 9.36541515889853e-06, "loss": 1.0885, "step": 6313 }, { "epoch": 0.53, "learning_rate": 9.362677681096714e-06, "loss": 1.008, "step": 6314 }, { "epoch": 0.53, "learning_rate": 9.359940251248401e-06, "loss": 1.0211, "step": 6315 }, { "epoch": 0.53, "learning_rate": 9.357202869559563e-06, "loss": 1.0469, "step": 6316 }, { "epoch": 0.53, "learning_rate": 9.354465536236165e-06, "loss": 0.9739, "step": 6317 }, { "epoch": 0.54, "learning_rate": 9.351728251484172e-06, "loss": 1.0237, "step": 6318 }, { "epoch": 0.54, "learning_rate": 9.34899101550954e-06, "loss": 1.0507, "step": 6319 }, { "epoch": 0.54, "learning_rate": 9.346253828518229e-06, "loss": 1.0461, "step": 6320 }, { "epoch": 0.54, "learning_rate": 9.343516690716183e-06, "loss": 0.9777, "step": 6321 }, { "epoch": 0.54, "learning_rate": 9.34077960230936e-06, "loss": 1.0313, "step": 6322 }, { "epoch": 0.54, "learning_rate": 9.338042563503697e-06, "loss": 0.9975, "step": 6323 }, { "epoch": 0.54, "learning_rate": 9.335305574505142e-06, "loss": 0.9231, "step": 6324 }, { "epoch": 0.54, "learning_rate": 9.332568635519622e-06, "loss": 1.0422, "step": 6325 }, { "epoch": 0.54, "learning_rate": 9.329831746753079e-06, "loss": 1.0058, "step": 6326 }, { "epoch": 0.54, "learning_rate": 9.327094908411436e-06, "loss": 1.038, "step": 6327 }, { "epoch": 0.54, "learning_rate": 9.324358120700626e-06, "loss": 1.0321, "step": 6328 }, { "epoch": 0.54, "learning_rate": 9.321621383826564e-06, "loss": 0.9939, "step": 6329 }, { "epoch": 0.54, "learning_rate": 9.31888469799517e-06, "loss": 1.0111, "step": 6330 }, { "epoch": 0.54, "learning_rate": 9.316148063412361e-06, "loss": 1.0689, "step": 6331 }, { "epoch": 0.54, "learning_rate": 9.313411480284042e-06, "loss": 1.0134, "step": 6332 }, { "epoch": 0.54, "learning_rate": 9.310674948816125e-06, "loss": 1.0292, "step": 6333 }, { "epoch": 0.54, "learning_rate": 9.307938469214508e-06, "loss": 1.0174, "step": 6334 }, { "epoch": 0.54, "learning_rate": 9.305202041685092e-06, "loss": 1.0192, "step": 6335 }, { "epoch": 0.54, "learning_rate": 9.30246566643377e-06, "loss": 1.0865, "step": 6336 }, { "epoch": 0.54, "learning_rate": 9.299729343666437e-06, "loss": 0.9793, "step": 6337 }, { "epoch": 0.54, "learning_rate": 9.296993073588972e-06, "loss": 1.0504, "step": 6338 }, { "epoch": 0.54, "learning_rate": 9.294256856407267e-06, "loss": 1.0196, "step": 6339 }, { "epoch": 0.54, "learning_rate": 9.291520692327194e-06, "loss": 1.013, "step": 6340 }, { "epoch": 0.54, "learning_rate": 9.288784581554634e-06, "loss": 0.9843, "step": 6341 }, { "epoch": 0.54, "learning_rate": 9.28604852429545e-06, "loss": 1.0279, "step": 6342 }, { "epoch": 0.54, "learning_rate": 9.283312520755515e-06, "loss": 1.0517, "step": 6343 }, { "epoch": 0.54, "learning_rate": 9.280576571140687e-06, "loss": 1.0362, "step": 6344 }, { "epoch": 0.54, "learning_rate": 9.277840675656834e-06, "loss": 0.9857, "step": 6345 }, { "epoch": 0.54, "learning_rate": 9.275104834509797e-06, "loss": 1.0557, "step": 6346 }, { "epoch": 0.54, "learning_rate": 9.272369047905439e-06, "loss": 1.09, "step": 6347 }, { "epoch": 0.54, "learning_rate": 9.269633316049597e-06, "loss": 1.0448, "step": 6348 }, { "epoch": 0.54, "learning_rate": 9.266897639148122e-06, "loss": 1.0757, "step": 6349 }, { "epoch": 0.54, "learning_rate": 9.264162017406845e-06, "loss": 0.9843, "step": 6350 }, { "epoch": 0.54, "learning_rate": 9.261426451031606e-06, "loss": 1.0677, "step": 6351 }, { "epoch": 0.54, "learning_rate": 9.258690940228226e-06, "loss": 1.0239, "step": 6352 }, { "epoch": 0.54, "learning_rate": 9.255955485202544e-06, "loss": 1.0759, "step": 6353 }, { "epoch": 0.54, "learning_rate": 9.253220086160369e-06, "loss": 1.0347, "step": 6354 }, { "epoch": 0.54, "learning_rate": 9.250484743307524e-06, "loss": 0.9902, "step": 6355 }, { "epoch": 0.54, "learning_rate": 9.247749456849824e-06, "loss": 1.0494, "step": 6356 }, { "epoch": 0.54, "learning_rate": 9.245014226993072e-06, "loss": 1.0417, "step": 6357 }, { "epoch": 0.54, "learning_rate": 9.24227905394308e-06, "loss": 1.0701, "step": 6358 }, { "epoch": 0.54, "learning_rate": 9.23954393790564e-06, "loss": 1.0347, "step": 6359 }, { "epoch": 0.54, "learning_rate": 9.236808879086556e-06, "loss": 0.9756, "step": 6360 }, { "epoch": 0.54, "learning_rate": 9.234073877691613e-06, "loss": 1.0484, "step": 6361 }, { "epoch": 0.54, "learning_rate": 9.231338933926606e-06, "loss": 0.8315, "step": 6362 }, { "epoch": 0.54, "learning_rate": 9.228604047997308e-06, "loss": 1.0329, "step": 6363 }, { "epoch": 0.54, "learning_rate": 9.225869220109507e-06, "loss": 1.0676, "step": 6364 }, { "epoch": 0.54, "learning_rate": 9.22313445046897e-06, "loss": 1.0393, "step": 6365 }, { "epoch": 0.54, "learning_rate": 9.220399739281474e-06, "loss": 1.064, "step": 6366 }, { "epoch": 0.54, "learning_rate": 9.217665086752778e-06, "loss": 1.0756, "step": 6367 }, { "epoch": 0.54, "learning_rate": 9.21493049308865e-06, "loss": 1.0377, "step": 6368 }, { "epoch": 0.54, "learning_rate": 9.212195958494842e-06, "loss": 1.0419, "step": 6369 }, { "epoch": 0.54, "learning_rate": 9.209461483177106e-06, "loss": 1.0194, "step": 6370 }, { "epoch": 0.54, "learning_rate": 9.20672706734119e-06, "loss": 0.8801, "step": 6371 }, { "epoch": 0.54, "learning_rate": 9.203992711192842e-06, "loss": 1.0206, "step": 6372 }, { "epoch": 0.54, "learning_rate": 9.201258414937793e-06, "loss": 1.0499, "step": 6373 }, { "epoch": 0.54, "learning_rate": 9.198524178781785e-06, "loss": 0.988, "step": 6374 }, { "epoch": 0.54, "learning_rate": 9.19579000293054e-06, "loss": 1.0359, "step": 6375 }, { "epoch": 0.54, "learning_rate": 9.193055887589793e-06, "loss": 1.0298, "step": 6376 }, { "epoch": 0.54, "learning_rate": 9.190321832965255e-06, "loss": 0.9815, "step": 6377 }, { "epoch": 0.54, "learning_rate": 9.18758783926265e-06, "loss": 0.8702, "step": 6378 }, { "epoch": 0.54, "learning_rate": 9.184853906687683e-06, "loss": 1.052, "step": 6379 }, { "epoch": 0.54, "learning_rate": 9.182120035446067e-06, "loss": 0.9989, "step": 6380 }, { "epoch": 0.54, "learning_rate": 9.179386225743497e-06, "loss": 1.012, "step": 6381 }, { "epoch": 0.54, "learning_rate": 9.176652477785676e-06, "loss": 0.8657, "step": 6382 }, { "epoch": 0.54, "learning_rate": 9.173918791778301e-06, "loss": 1.0298, "step": 6383 }, { "epoch": 0.54, "learning_rate": 9.171185167927052e-06, "loss": 0.9541, "step": 6384 }, { "epoch": 0.54, "learning_rate": 9.168451606437619e-06, "loss": 1.0053, "step": 6385 }, { "epoch": 0.54, "learning_rate": 9.165718107515674e-06, "loss": 1.0168, "step": 6386 }, { "epoch": 0.54, "learning_rate": 9.162984671366901e-06, "loss": 1.0369, "step": 6387 }, { "epoch": 0.54, "learning_rate": 9.16025129819696e-06, "loss": 1.0046, "step": 6388 }, { "epoch": 0.54, "learning_rate": 9.157517988211523e-06, "loss": 1.0222, "step": 6389 }, { "epoch": 0.54, "learning_rate": 9.154784741616247e-06, "loss": 1.0327, "step": 6390 }, { "epoch": 0.54, "learning_rate": 9.152051558616788e-06, "loss": 0.9757, "step": 6391 }, { "epoch": 0.54, "learning_rate": 9.149318439418794e-06, "loss": 1.008, "step": 6392 }, { "epoch": 0.54, "learning_rate": 9.146585384227918e-06, "loss": 0.9844, "step": 6393 }, { "epoch": 0.54, "learning_rate": 9.143852393249794e-06, "loss": 0.9942, "step": 6394 }, { "epoch": 0.54, "learning_rate": 9.14111946669006e-06, "loss": 1.0341, "step": 6395 }, { "epoch": 0.54, "learning_rate": 9.138386604754345e-06, "loss": 1.0468, "step": 6396 }, { "epoch": 0.54, "learning_rate": 9.135653807648281e-06, "loss": 1.0388, "step": 6397 }, { "epoch": 0.54, "learning_rate": 9.132921075577485e-06, "loss": 1.0085, "step": 6398 }, { "epoch": 0.54, "learning_rate": 9.130188408747578e-06, "loss": 1.0402, "step": 6399 }, { "epoch": 0.54, "learning_rate": 9.127455807364165e-06, "loss": 1.0033, "step": 6400 }, { "epoch": 0.54, "learning_rate": 9.124723271632862e-06, "loss": 1.0514, "step": 6401 }, { "epoch": 0.54, "learning_rate": 9.12199080175926e-06, "loss": 1.0424, "step": 6402 }, { "epoch": 0.54, "learning_rate": 9.119258397948966e-06, "loss": 1.074, "step": 6403 }, { "epoch": 0.54, "learning_rate": 9.116526060407562e-06, "loss": 1.0018, "step": 6404 }, { "epoch": 0.54, "learning_rate": 9.113793789340645e-06, "loss": 1.044, "step": 6405 }, { "epoch": 0.54, "learning_rate": 9.111061584953789e-06, "loss": 1.0866, "step": 6406 }, { "epoch": 0.54, "learning_rate": 9.108329447452578e-06, "loss": 1.0228, "step": 6407 }, { "epoch": 0.54, "learning_rate": 9.105597377042578e-06, "loss": 0.9695, "step": 6408 }, { "epoch": 0.54, "learning_rate": 9.102865373929359e-06, "loss": 0.988, "step": 6409 }, { "epoch": 0.54, "learning_rate": 9.100133438318483e-06, "loss": 1.0882, "step": 6410 }, { "epoch": 0.54, "learning_rate": 9.097401570415502e-06, "loss": 1.0524, "step": 6411 }, { "epoch": 0.54, "learning_rate": 9.094669770425976e-06, "loss": 1.0025, "step": 6412 }, { "epoch": 0.54, "learning_rate": 9.091938038555441e-06, "loss": 1.048, "step": 6413 }, { "epoch": 0.54, "learning_rate": 9.089206375009449e-06, "loss": 1.0264, "step": 6414 }, { "epoch": 0.54, "learning_rate": 9.086474779993528e-06, "loss": 0.8317, "step": 6415 }, { "epoch": 0.54, "learning_rate": 9.083743253713217e-06, "loss": 1.016, "step": 6416 }, { "epoch": 0.54, "learning_rate": 9.081011796374033e-06, "loss": 1.0179, "step": 6417 }, { "epoch": 0.54, "learning_rate": 9.078280408181504e-06, "loss": 1.0241, "step": 6418 }, { "epoch": 0.54, "learning_rate": 9.07554908934114e-06, "loss": 1.0106, "step": 6419 }, { "epoch": 0.54, "learning_rate": 9.072817840058458e-06, "loss": 1.0431, "step": 6420 }, { "epoch": 0.54, "learning_rate": 9.070086660538955e-06, "loss": 0.9837, "step": 6421 }, { "epoch": 0.54, "learning_rate": 9.067355550988136e-06, "loss": 1.0386, "step": 6422 }, { "epoch": 0.54, "learning_rate": 9.064624511611497e-06, "loss": 1.0023, "step": 6423 }, { "epoch": 0.54, "learning_rate": 9.061893542614525e-06, "loss": 1.0892, "step": 6424 }, { "epoch": 0.54, "learning_rate": 9.0591626442027e-06, "loss": 1.0692, "step": 6425 }, { "epoch": 0.54, "learning_rate": 9.056431816581506e-06, "loss": 1.0034, "step": 6426 }, { "epoch": 0.54, "learning_rate": 9.053701059956413e-06, "loss": 0.9739, "step": 6427 }, { "epoch": 0.54, "learning_rate": 9.050970374532895e-06, "loss": 1.0397, "step": 6428 }, { "epoch": 0.54, "learning_rate": 9.048239760516407e-06, "loss": 1.0169, "step": 6429 }, { "epoch": 0.54, "learning_rate": 9.045509218112413e-06, "loss": 1.0376, "step": 6430 }, { "epoch": 0.54, "learning_rate": 9.042778747526357e-06, "loss": 1.041, "step": 6431 }, { "epoch": 0.54, "learning_rate": 9.040048348963694e-06, "loss": 1.0388, "step": 6432 }, { "epoch": 0.54, "learning_rate": 9.037318022629857e-06, "loss": 1.0497, "step": 6433 }, { "epoch": 0.54, "learning_rate": 9.03458776873029e-06, "loss": 1.0416, "step": 6434 }, { "epoch": 0.54, "learning_rate": 9.031857587470417e-06, "loss": 1.017, "step": 6435 }, { "epoch": 0.55, "learning_rate": 9.029127479055665e-06, "loss": 1.0215, "step": 6436 }, { "epoch": 0.55, "learning_rate": 9.026397443691456e-06, "loss": 1.0325, "step": 6437 }, { "epoch": 0.55, "learning_rate": 9.023667481583196e-06, "loss": 1.0612, "step": 6438 }, { "epoch": 0.55, "learning_rate": 9.020937592936302e-06, "loss": 1.0705, "step": 6439 }, { "epoch": 0.55, "learning_rate": 9.018207777956167e-06, "loss": 0.9993, "step": 6440 }, { "epoch": 0.55, "learning_rate": 9.0154780368482e-06, "loss": 1.0104, "step": 6441 }, { "epoch": 0.55, "learning_rate": 9.012748369817782e-06, "loss": 0.9647, "step": 6442 }, { "epoch": 0.55, "learning_rate": 9.010018777070307e-06, "loss": 1.0362, "step": 6443 }, { "epoch": 0.55, "learning_rate": 9.007289258811148e-06, "loss": 0.9967, "step": 6444 }, { "epoch": 0.55, "learning_rate": 9.004559815245688e-06, "loss": 1.0391, "step": 6445 }, { "epoch": 0.55, "learning_rate": 9.001830446579288e-06, "loss": 1.063, "step": 6446 }, { "epoch": 0.55, "learning_rate": 8.99910115301732e-06, "loss": 0.9641, "step": 6447 }, { "epoch": 0.55, "learning_rate": 8.996371934765133e-06, "loss": 0.9991, "step": 6448 }, { "epoch": 0.55, "learning_rate": 8.993642792028086e-06, "loss": 1.0805, "step": 6449 }, { "epoch": 0.55, "learning_rate": 8.990913725011522e-06, "loss": 1.001, "step": 6450 }, { "epoch": 0.55, "learning_rate": 8.988184733920787e-06, "loss": 1.0841, "step": 6451 }, { "epoch": 0.55, "learning_rate": 8.985455818961206e-06, "loss": 1.0319, "step": 6452 }, { "epoch": 0.55, "learning_rate": 8.98272698033812e-06, "loss": 1.0151, "step": 6453 }, { "epoch": 0.55, "learning_rate": 8.979998218256843e-06, "loss": 1.0449, "step": 6454 }, { "epoch": 0.55, "learning_rate": 8.977269532922701e-06, "loss": 1.034, "step": 6455 }, { "epoch": 0.55, "learning_rate": 8.974540924541e-06, "loss": 0.9942, "step": 6456 }, { "epoch": 0.55, "learning_rate": 8.971812393317053e-06, "loss": 0.9957, "step": 6457 }, { "epoch": 0.55, "learning_rate": 8.96908393945615e-06, "loss": 1.0475, "step": 6458 }, { "epoch": 0.55, "learning_rate": 8.966355563163593e-06, "loss": 1.037, "step": 6459 }, { "epoch": 0.55, "learning_rate": 8.963627264644676e-06, "loss": 1.0145, "step": 6460 }, { "epoch": 0.55, "learning_rate": 8.960899044104673e-06, "loss": 1.0183, "step": 6461 }, { "epoch": 0.55, "learning_rate": 8.958170901748865e-06, "loss": 0.9486, "step": 6462 }, { "epoch": 0.55, "learning_rate": 8.955442837782521e-06, "loss": 1.0392, "step": 6463 }, { "epoch": 0.55, "learning_rate": 8.952714852410912e-06, "loss": 1.0518, "step": 6464 }, { "epoch": 0.55, "learning_rate": 8.949986945839289e-06, "loss": 1.0185, "step": 6465 }, { "epoch": 0.55, "learning_rate": 8.947259118272912e-06, "loss": 1.0792, "step": 6466 }, { "epoch": 0.55, "learning_rate": 8.944531369917027e-06, "loss": 1.0339, "step": 6467 }, { "epoch": 0.55, "learning_rate": 8.941803700976877e-06, "loss": 1.0328, "step": 6468 }, { "epoch": 0.55, "learning_rate": 8.939076111657692e-06, "loss": 1.0123, "step": 6469 }, { "epoch": 0.55, "learning_rate": 8.93634860216471e-06, "loss": 1.0612, "step": 6470 }, { "epoch": 0.55, "learning_rate": 8.933621172703148e-06, "loss": 1.0347, "step": 6471 }, { "epoch": 0.55, "learning_rate": 8.930893823478229e-06, "loss": 0.9919, "step": 6472 }, { "epoch": 0.55, "learning_rate": 8.928166554695157e-06, "loss": 1.0371, "step": 6473 }, { "epoch": 0.55, "learning_rate": 8.925439366559148e-06, "loss": 1.0205, "step": 6474 }, { "epoch": 0.55, "learning_rate": 8.922712259275392e-06, "loss": 0.9982, "step": 6475 }, { "epoch": 0.55, "learning_rate": 8.91998523304909e-06, "loss": 1.0468, "step": 6476 }, { "epoch": 0.55, "learning_rate": 8.917258288085421e-06, "loss": 1.0156, "step": 6477 }, { "epoch": 0.55, "learning_rate": 8.914531424589575e-06, "loss": 1.0327, "step": 6478 }, { "epoch": 0.55, "learning_rate": 8.91180464276672e-06, "loss": 0.8891, "step": 6479 }, { "epoch": 0.55, "learning_rate": 8.909077942822031e-06, "loss": 1.0123, "step": 6480 }, { "epoch": 0.55, "learning_rate": 8.906351324960664e-06, "loss": 1.0399, "step": 6481 }, { "epoch": 0.55, "learning_rate": 8.903624789387782e-06, "loss": 1.0228, "step": 6482 }, { "epoch": 0.55, "learning_rate": 8.90089833630853e-06, "loss": 1.0219, "step": 6483 }, { "epoch": 0.55, "learning_rate": 8.898171965928058e-06, "loss": 1.0223, "step": 6484 }, { "epoch": 0.55, "learning_rate": 8.895445678451498e-06, "loss": 1.0576, "step": 6485 }, { "epoch": 0.55, "learning_rate": 8.892719474083986e-06, "loss": 1.0463, "step": 6486 }, { "epoch": 0.55, "learning_rate": 8.889993353030646e-06, "loss": 1.0018, "step": 6487 }, { "epoch": 0.55, "learning_rate": 8.887267315496598e-06, "loss": 1.0213, "step": 6488 }, { "epoch": 0.55, "learning_rate": 8.884541361686953e-06, "loss": 1.0395, "step": 6489 }, { "epoch": 0.55, "learning_rate": 8.88181549180682e-06, "loss": 0.9882, "step": 6490 }, { "epoch": 0.55, "learning_rate": 8.8790897060613e-06, "loss": 1.0242, "step": 6491 }, { "epoch": 0.55, "learning_rate": 8.876364004655481e-06, "loss": 1.0424, "step": 6492 }, { "epoch": 0.55, "learning_rate": 8.87363838779446e-06, "loss": 1.0254, "step": 6493 }, { "epoch": 0.55, "learning_rate": 8.870912855683308e-06, "loss": 1.038, "step": 6494 }, { "epoch": 0.55, "learning_rate": 8.86818740852711e-06, "loss": 1.029, "step": 6495 }, { "epoch": 0.55, "learning_rate": 8.865462046530924e-06, "loss": 1.046, "step": 6496 }, { "epoch": 0.55, "learning_rate": 8.862736769899822e-06, "loss": 1.0382, "step": 6497 }, { "epoch": 0.55, "learning_rate": 8.860011578838852e-06, "loss": 0.8896, "step": 6498 }, { "epoch": 0.55, "learning_rate": 8.85728647355307e-06, "loss": 1.0699, "step": 6499 }, { "epoch": 0.55, "learning_rate": 8.854561454247512e-06, "loss": 1.0356, "step": 6500 }, { "epoch": 0.55, "learning_rate": 8.851836521127223e-06, "loss": 1.0327, "step": 6501 }, { "epoch": 0.55, "learning_rate": 8.849111674397224e-06, "loss": 0.9242, "step": 6502 }, { "epoch": 0.55, "learning_rate": 8.84638691426254e-06, "loss": 1.0571, "step": 6503 }, { "epoch": 0.55, "learning_rate": 8.843662240928194e-06, "loss": 1.0693, "step": 6504 }, { "epoch": 0.55, "learning_rate": 8.840937654599193e-06, "loss": 0.9771, "step": 6505 }, { "epoch": 0.55, "learning_rate": 8.838213155480533e-06, "loss": 1.0325, "step": 6506 }, { "epoch": 0.55, "learning_rate": 8.835488743777225e-06, "loss": 1.0515, "step": 6507 }, { "epoch": 0.55, "learning_rate": 8.832764419694247e-06, "loss": 0.9956, "step": 6508 }, { "epoch": 0.55, "learning_rate": 8.830040183436592e-06, "loss": 1.0681, "step": 6509 }, { "epoch": 0.55, "learning_rate": 8.82731603520923e-06, "loss": 1.0356, "step": 6510 }, { "epoch": 0.55, "learning_rate": 8.824591975217136e-06, "loss": 1.0737, "step": 6511 }, { "epoch": 0.55, "learning_rate": 8.82186800366528e-06, "loss": 0.9852, "step": 6512 }, { "epoch": 0.55, "learning_rate": 8.819144120758608e-06, "loss": 1.0714, "step": 6513 }, { "epoch": 0.55, "learning_rate": 8.816420326702079e-06, "loss": 1.0841, "step": 6514 }, { "epoch": 0.55, "learning_rate": 8.813696621700632e-06, "loss": 1.0333, "step": 6515 }, { "epoch": 0.55, "learning_rate": 8.81097300595921e-06, "loss": 1.0943, "step": 6516 }, { "epoch": 0.55, "learning_rate": 8.808249479682736e-06, "loss": 1.0041, "step": 6517 }, { "epoch": 0.55, "learning_rate": 8.805526043076145e-06, "loss": 1.0438, "step": 6518 }, { "epoch": 0.55, "learning_rate": 8.80280269634434e-06, "loss": 1.033, "step": 6519 }, { "epoch": 0.55, "learning_rate": 8.800079439692246e-06, "loss": 1.0004, "step": 6520 }, { "epoch": 0.55, "learning_rate": 8.797356273324755e-06, "loss": 0.9759, "step": 6521 }, { "epoch": 0.55, "learning_rate": 8.79463319744677e-06, "loss": 1.0205, "step": 6522 }, { "epoch": 0.55, "learning_rate": 8.791910212263179e-06, "loss": 1.0351, "step": 6523 }, { "epoch": 0.55, "learning_rate": 8.78918731797887e-06, "loss": 1.0594, "step": 6524 }, { "epoch": 0.55, "learning_rate": 8.786464514798712e-06, "loss": 1.0794, "step": 6525 }, { "epoch": 0.55, "learning_rate": 8.783741802927582e-06, "loss": 1.0032, "step": 6526 }, { "epoch": 0.55, "learning_rate": 8.781019182570333e-06, "loss": 1.0039, "step": 6527 }, { "epoch": 0.55, "learning_rate": 8.778296653931832e-06, "loss": 0.9677, "step": 6528 }, { "epoch": 0.55, "learning_rate": 8.77557421721692e-06, "loss": 1.019, "step": 6529 }, { "epoch": 0.55, "learning_rate": 8.772851872630442e-06, "loss": 0.9965, "step": 6530 }, { "epoch": 0.55, "learning_rate": 8.770129620377232e-06, "loss": 0.9728, "step": 6531 }, { "epoch": 0.55, "learning_rate": 8.767407460662122e-06, "loss": 1.0194, "step": 6532 }, { "epoch": 0.55, "learning_rate": 8.764685393689926e-06, "loss": 0.9822, "step": 6533 }, { "epoch": 0.55, "learning_rate": 8.761963419665464e-06, "loss": 1.0393, "step": 6534 }, { "epoch": 0.55, "learning_rate": 8.759241538793539e-06, "loss": 0.9688, "step": 6535 }, { "epoch": 0.55, "learning_rate": 8.756519751278954e-06, "loss": 1.0585, "step": 6536 }, { "epoch": 0.55, "learning_rate": 8.753798057326504e-06, "loss": 0.9749, "step": 6537 }, { "epoch": 0.55, "learning_rate": 8.75107645714097e-06, "loss": 1.0153, "step": 6538 }, { "epoch": 0.55, "learning_rate": 8.748354950927139e-06, "loss": 1.0633, "step": 6539 }, { "epoch": 0.55, "learning_rate": 8.745633538889774e-06, "loss": 0.9731, "step": 6540 }, { "epoch": 0.55, "learning_rate": 8.742912221233644e-06, "loss": 0.9807, "step": 6541 }, { "epoch": 0.55, "learning_rate": 8.740190998163507e-06, "loss": 1.0109, "step": 6542 }, { "epoch": 0.55, "learning_rate": 8.737469869884115e-06, "loss": 1.0099, "step": 6543 }, { "epoch": 0.55, "learning_rate": 8.734748836600206e-06, "loss": 1.0438, "step": 6544 }, { "epoch": 0.55, "learning_rate": 8.732027898516524e-06, "loss": 1.0171, "step": 6545 }, { "epoch": 0.55, "learning_rate": 8.72930705583779e-06, "loss": 1.0481, "step": 6546 }, { "epoch": 0.55, "learning_rate": 8.726586308768736e-06, "loss": 1.0123, "step": 6547 }, { "epoch": 0.55, "learning_rate": 8.723865657514064e-06, "loss": 1.0162, "step": 6548 }, { "epoch": 0.55, "learning_rate": 8.721145102278497e-06, "loss": 0.9941, "step": 6549 }, { "epoch": 0.55, "learning_rate": 8.71842464326672e-06, "loss": 1.0055, "step": 6550 }, { "epoch": 0.55, "learning_rate": 8.71570428068344e-06, "loss": 1.0571, "step": 6551 }, { "epoch": 0.55, "learning_rate": 8.712984014733332e-06, "loss": 1.0486, "step": 6552 }, { "epoch": 0.55, "learning_rate": 8.710263845621083e-06, "loss": 1.0289, "step": 6553 }, { "epoch": 0.56, "learning_rate": 8.707543773551357e-06, "loss": 1.0014, "step": 6554 }, { "epoch": 0.56, "learning_rate": 8.704823798728826e-06, "loss": 1.0461, "step": 6555 }, { "epoch": 0.56, "learning_rate": 8.702103921358141e-06, "loss": 1.008, "step": 6556 }, { "epoch": 0.56, "learning_rate": 8.699384141643951e-06, "loss": 1.0197, "step": 6557 }, { "epoch": 0.56, "learning_rate": 8.696664459790901e-06, "loss": 0.9256, "step": 6558 }, { "epoch": 0.56, "learning_rate": 8.693944876003627e-06, "loss": 1.0274, "step": 6559 }, { "epoch": 0.56, "learning_rate": 8.69122539048675e-06, "loss": 0.9714, "step": 6560 }, { "epoch": 0.56, "learning_rate": 8.688506003444898e-06, "loss": 1.0145, "step": 6561 }, { "epoch": 0.56, "learning_rate": 8.685786715082676e-06, "loss": 1.0123, "step": 6562 }, { "epoch": 0.56, "learning_rate": 8.68306752560469e-06, "loss": 1.0178, "step": 6563 }, { "epoch": 0.56, "learning_rate": 8.680348435215549e-06, "loss": 1.0105, "step": 6564 }, { "epoch": 0.56, "learning_rate": 8.677629444119827e-06, "loss": 0.9823, "step": 6565 }, { "epoch": 0.56, "learning_rate": 8.67491055252212e-06, "loss": 1.0265, "step": 6566 }, { "epoch": 0.56, "learning_rate": 8.672191760626996e-06, "loss": 0.8562, "step": 6567 }, { "epoch": 0.56, "learning_rate": 8.669473068639022e-06, "loss": 1.011, "step": 6568 }, { "epoch": 0.56, "learning_rate": 8.666754476762763e-06, "loss": 0.8877, "step": 6569 }, { "epoch": 0.56, "learning_rate": 8.66403598520277e-06, "loss": 1.0372, "step": 6570 }, { "epoch": 0.56, "learning_rate": 8.661317594163583e-06, "loss": 1.04, "step": 6571 }, { "epoch": 0.56, "learning_rate": 8.65859930384975e-06, "loss": 1.0245, "step": 6572 }, { "epoch": 0.56, "learning_rate": 8.655881114465789e-06, "loss": 1.0223, "step": 6573 }, { "epoch": 0.56, "learning_rate": 8.653163026216233e-06, "loss": 1.0211, "step": 6574 }, { "epoch": 0.56, "learning_rate": 8.650445039305589e-06, "loss": 1.0736, "step": 6575 }, { "epoch": 0.56, "learning_rate": 8.647727153938373e-06, "loss": 0.963, "step": 6576 }, { "epoch": 0.56, "learning_rate": 8.645009370319072e-06, "loss": 1.0259, "step": 6577 }, { "epoch": 0.56, "learning_rate": 8.64229168865219e-06, "loss": 1.0279, "step": 6578 }, { "epoch": 0.56, "learning_rate": 8.639574109142203e-06, "loss": 1.0198, "step": 6579 }, { "epoch": 0.56, "learning_rate": 8.636856631993593e-06, "loss": 1.076, "step": 6580 }, { "epoch": 0.56, "learning_rate": 8.634139257410823e-06, "loss": 1.085, "step": 6581 }, { "epoch": 0.56, "learning_rate": 8.631421985598363e-06, "loss": 1.0204, "step": 6582 }, { "epoch": 0.56, "learning_rate": 8.628704816760659e-06, "loss": 1.0444, "step": 6583 }, { "epoch": 0.56, "learning_rate": 8.62598775110216e-06, "loss": 1.0046, "step": 6584 }, { "epoch": 0.56, "learning_rate": 8.623270788827298e-06, "loss": 1.0163, "step": 6585 }, { "epoch": 0.56, "learning_rate": 8.620553930140513e-06, "loss": 1.0387, "step": 6586 }, { "epoch": 0.56, "learning_rate": 8.617837175246218e-06, "loss": 0.9881, "step": 6587 }, { "epoch": 0.56, "learning_rate": 8.615120524348829e-06, "loss": 1.0454, "step": 6588 }, { "epoch": 0.56, "learning_rate": 8.612403977652762e-06, "loss": 1.0002, "step": 6589 }, { "epoch": 0.56, "learning_rate": 8.609687535362403e-06, "loss": 0.9465, "step": 6590 }, { "epoch": 0.56, "learning_rate": 8.606971197682153e-06, "loss": 1.0303, "step": 6591 }, { "epoch": 0.56, "learning_rate": 8.604254964816388e-06, "loss": 1.0559, "step": 6592 }, { "epoch": 0.56, "learning_rate": 8.601538836969489e-06, "loss": 0.9917, "step": 6593 }, { "epoch": 0.56, "learning_rate": 8.59882281434582e-06, "loss": 1.0286, "step": 6594 }, { "epoch": 0.56, "learning_rate": 8.59610689714974e-06, "loss": 0.9861, "step": 6595 }, { "epoch": 0.56, "learning_rate": 8.593391085585602e-06, "loss": 1.0202, "step": 6596 }, { "epoch": 0.56, "learning_rate": 8.590675379857752e-06, "loss": 0.9985, "step": 6597 }, { "epoch": 0.56, "learning_rate": 8.587959780170515e-06, "loss": 0.9981, "step": 6598 }, { "epoch": 0.56, "learning_rate": 8.585244286728233e-06, "loss": 1.0085, "step": 6599 }, { "epoch": 0.56, "learning_rate": 8.582528899735213e-06, "loss": 1.0034, "step": 6600 }, { "epoch": 0.56, "learning_rate": 8.579813619395778e-06, "loss": 1.0297, "step": 6601 }, { "epoch": 0.56, "learning_rate": 8.57709844591422e-06, "loss": 1.0298, "step": 6602 }, { "epoch": 0.56, "learning_rate": 8.574383379494844e-06, "loss": 1.0077, "step": 6603 }, { "epoch": 0.56, "learning_rate": 8.57166842034193e-06, "loss": 1.0304, "step": 6604 }, { "epoch": 0.56, "learning_rate": 8.568953568659766e-06, "loss": 1.0301, "step": 6605 }, { "epoch": 0.56, "learning_rate": 8.566238824652612e-06, "loss": 1.0187, "step": 6606 }, { "epoch": 0.56, "learning_rate": 8.563524188524744e-06, "loss": 1.019, "step": 6607 }, { "epoch": 0.56, "learning_rate": 8.560809660480404e-06, "loss": 1.0671, "step": 6608 }, { "epoch": 0.56, "learning_rate": 8.558095240723853e-06, "loss": 1.0494, "step": 6609 }, { "epoch": 0.56, "learning_rate": 8.555380929459319e-06, "loss": 0.9859, "step": 6610 }, { "epoch": 0.56, "learning_rate": 8.552666726891035e-06, "loss": 1.0194, "step": 6611 }, { "epoch": 0.56, "learning_rate": 8.549952633223224e-06, "loss": 1.0388, "step": 6612 }, { "epoch": 0.56, "learning_rate": 8.547238648660103e-06, "loss": 1.068, "step": 6613 }, { "epoch": 0.56, "learning_rate": 8.544524773405872e-06, "loss": 0.9992, "step": 6614 }, { "epoch": 0.56, "learning_rate": 8.541811007664734e-06, "loss": 1.0129, "step": 6615 }, { "epoch": 0.56, "learning_rate": 8.539097351640879e-06, "loss": 1.0026, "step": 6616 }, { "epoch": 0.56, "learning_rate": 8.536383805538482e-06, "loss": 1.0693, "step": 6617 }, { "epoch": 0.56, "learning_rate": 8.533670369561726e-06, "loss": 1.0103, "step": 6618 }, { "epoch": 0.56, "learning_rate": 8.530957043914765e-06, "loss": 1.0368, "step": 6619 }, { "epoch": 0.56, "learning_rate": 8.528243828801766e-06, "loss": 1.0084, "step": 6620 }, { "epoch": 0.56, "learning_rate": 8.52553072442687e-06, "loss": 1.0378, "step": 6621 }, { "epoch": 0.56, "learning_rate": 8.522817730994216e-06, "loss": 1.0161, "step": 6622 }, { "epoch": 0.56, "learning_rate": 8.52010484870794e-06, "loss": 1.0001, "step": 6623 }, { "epoch": 0.56, "learning_rate": 8.517392077772167e-06, "loss": 1.0124, "step": 6624 }, { "epoch": 0.56, "learning_rate": 8.514679418391e-06, "loss": 1.0911, "step": 6625 }, { "epoch": 0.56, "learning_rate": 8.511966870768562e-06, "loss": 0.8937, "step": 6626 }, { "epoch": 0.56, "learning_rate": 8.509254435108935e-06, "loss": 1.0272, "step": 6627 }, { "epoch": 0.56, "learning_rate": 8.506542111616219e-06, "loss": 1.0305, "step": 6628 }, { "epoch": 0.56, "learning_rate": 8.50382990049449e-06, "loss": 1.0457, "step": 6629 }, { "epoch": 0.56, "learning_rate": 8.501117801947825e-06, "loss": 1.0147, "step": 6630 }, { "epoch": 0.56, "learning_rate": 8.498405816180282e-06, "loss": 1.0391, "step": 6631 }, { "epoch": 0.56, "learning_rate": 8.495693943395925e-06, "loss": 1.0194, "step": 6632 }, { "epoch": 0.56, "learning_rate": 8.492982183798791e-06, "loss": 0.978, "step": 6633 }, { "epoch": 0.56, "learning_rate": 8.490270537592928e-06, "loss": 1.0158, "step": 6634 }, { "epoch": 0.56, "learning_rate": 8.487559004982362e-06, "loss": 0.9916, "step": 6635 }, { "epoch": 0.56, "learning_rate": 8.484847586171112e-06, "loss": 1.0482, "step": 6636 }, { "epoch": 0.56, "learning_rate": 8.482136281363196e-06, "loss": 1.0561, "step": 6637 }, { "epoch": 0.56, "learning_rate": 8.479425090762615e-06, "loss": 0.986, "step": 6638 }, { "epoch": 0.56, "learning_rate": 8.476714014573366e-06, "loss": 1.0056, "step": 6639 }, { "epoch": 0.56, "learning_rate": 8.474003052999433e-06, "loss": 1.0061, "step": 6640 }, { "epoch": 0.56, "learning_rate": 8.471292206244802e-06, "loss": 1.0448, "step": 6641 }, { "epoch": 0.56, "learning_rate": 8.468581474513436e-06, "loss": 0.9436, "step": 6642 }, { "epoch": 0.56, "learning_rate": 8.465870858009303e-06, "loss": 1.0454, "step": 6643 }, { "epoch": 0.56, "learning_rate": 8.463160356936344e-06, "loss": 0.9929, "step": 6644 }, { "epoch": 0.56, "learning_rate": 8.460449971498518e-06, "loss": 1.0198, "step": 6645 }, { "epoch": 0.56, "learning_rate": 8.457739701899747e-06, "loss": 1.0025, "step": 6646 }, { "epoch": 0.56, "learning_rate": 8.455029548343969e-06, "loss": 0.9457, "step": 6647 }, { "epoch": 0.56, "learning_rate": 8.452319511035091e-06, "loss": 1.0198, "step": 6648 }, { "epoch": 0.56, "learning_rate": 8.44960959017703e-06, "loss": 1.0276, "step": 6649 }, { "epoch": 0.56, "learning_rate": 8.446899785973682e-06, "loss": 1.0597, "step": 6650 }, { "epoch": 0.56, "learning_rate": 8.444190098628942e-06, "loss": 0.9687, "step": 6651 }, { "epoch": 0.56, "learning_rate": 8.441480528346686e-06, "loss": 1.0381, "step": 6652 }, { "epoch": 0.56, "learning_rate": 8.438771075330797e-06, "loss": 1.0627, "step": 6653 }, { "epoch": 0.56, "learning_rate": 8.436061739785132e-06, "loss": 0.9682, "step": 6654 }, { "epoch": 0.56, "learning_rate": 8.433352521913555e-06, "loss": 1.0228, "step": 6655 }, { "epoch": 0.56, "learning_rate": 8.430643421919906e-06, "loss": 1.0188, "step": 6656 }, { "epoch": 0.56, "learning_rate": 8.427934440008028e-06, "loss": 1.0308, "step": 6657 }, { "epoch": 0.56, "learning_rate": 8.425225576381748e-06, "loss": 1.0242, "step": 6658 }, { "epoch": 0.56, "learning_rate": 8.422516831244892e-06, "loss": 1.0451, "step": 6659 }, { "epoch": 0.56, "learning_rate": 8.419808204801261e-06, "loss": 1.0476, "step": 6660 }, { "epoch": 0.56, "learning_rate": 8.417099697254671e-06, "loss": 1.0367, "step": 6661 }, { "epoch": 0.56, "learning_rate": 8.414391308808908e-06, "loss": 1.0502, "step": 6662 }, { "epoch": 0.56, "learning_rate": 8.411683039667757e-06, "loss": 0.9513, "step": 6663 }, { "epoch": 0.56, "learning_rate": 8.408974890034996e-06, "loss": 0.9779, "step": 6664 }, { "epoch": 0.56, "learning_rate": 8.406266860114395e-06, "loss": 1.0087, "step": 6665 }, { "epoch": 0.56, "learning_rate": 8.403558950109705e-06, "loss": 1.0382, "step": 6666 }, { "epoch": 0.56, "learning_rate": 8.400851160224677e-06, "loss": 1.0334, "step": 6667 }, { "epoch": 0.56, "learning_rate": 8.398143490663058e-06, "loss": 1.0083, "step": 6668 }, { "epoch": 0.56, "learning_rate": 8.395435941628569e-06, "loss": 1.0161, "step": 6669 }, { "epoch": 0.56, "learning_rate": 8.39272851332494e-06, "loss": 1.0118, "step": 6670 }, { "epoch": 0.56, "learning_rate": 8.390021205955877e-06, "loss": 1.0707, "step": 6671 }, { "epoch": 0.57, "learning_rate": 8.387314019725088e-06, "loss": 1.0561, "step": 6672 }, { "epoch": 0.57, "learning_rate": 8.384606954836264e-06, "loss": 1.0612, "step": 6673 }, { "epoch": 0.57, "learning_rate": 8.381900011493099e-06, "loss": 1.0343, "step": 6674 }, { "epoch": 0.57, "learning_rate": 8.37919318989926e-06, "loss": 1.0188, "step": 6675 }, { "epoch": 0.57, "learning_rate": 8.376486490258415e-06, "loss": 1.0473, "step": 6676 }, { "epoch": 0.57, "learning_rate": 8.373779912774226e-06, "loss": 1.053, "step": 6677 }, { "epoch": 0.57, "learning_rate": 8.371073457650339e-06, "loss": 1.0234, "step": 6678 }, { "epoch": 0.57, "learning_rate": 8.368367125090392e-06, "loss": 1.009, "step": 6679 }, { "epoch": 0.57, "learning_rate": 8.365660915298022e-06, "loss": 1.0346, "step": 6680 }, { "epoch": 0.57, "learning_rate": 8.362954828476841e-06, "loss": 1.051, "step": 6681 }, { "epoch": 0.57, "learning_rate": 8.360248864830468e-06, "loss": 0.889, "step": 6682 }, { "epoch": 0.57, "learning_rate": 8.3575430245625e-06, "loss": 1.0956, "step": 6683 }, { "epoch": 0.57, "learning_rate": 8.354837307876537e-06, "loss": 0.9642, "step": 6684 }, { "epoch": 0.57, "learning_rate": 8.352131714976154e-06, "loss": 1.0013, "step": 6685 }, { "epoch": 0.57, "learning_rate": 8.349426246064936e-06, "loss": 1.0295, "step": 6686 }, { "epoch": 0.57, "learning_rate": 8.346720901346439e-06, "loss": 1.0512, "step": 6687 }, { "epoch": 0.57, "learning_rate": 8.344015681024225e-06, "loss": 1.0563, "step": 6688 }, { "epoch": 0.57, "learning_rate": 8.341310585301838e-06, "loss": 0.992, "step": 6689 }, { "epoch": 0.57, "learning_rate": 8.338605614382816e-06, "loss": 1.0283, "step": 6690 }, { "epoch": 0.57, "learning_rate": 8.335900768470686e-06, "loss": 1.0055, "step": 6691 }, { "epoch": 0.57, "learning_rate": 8.333196047768966e-06, "loss": 0.9972, "step": 6692 }, { "epoch": 0.57, "learning_rate": 8.33049145248117e-06, "loss": 0.9989, "step": 6693 }, { "epoch": 0.57, "learning_rate": 8.327786982810789e-06, "loss": 1.0098, "step": 6694 }, { "epoch": 0.57, "learning_rate": 8.325082638961321e-06, "loss": 1.0397, "step": 6695 }, { "epoch": 0.57, "learning_rate": 8.322378421136242e-06, "loss": 1.0772, "step": 6696 }, { "epoch": 0.57, "learning_rate": 8.319674329539025e-06, "loss": 0.9821, "step": 6697 }, { "epoch": 0.57, "learning_rate": 8.31697036437313e-06, "loss": 0.994, "step": 6698 }, { "epoch": 0.57, "learning_rate": 8.314266525842014e-06, "loss": 0.985, "step": 6699 }, { "epoch": 0.57, "learning_rate": 8.311562814149112e-06, "loss": 0.9933, "step": 6700 }, { "epoch": 0.57, "learning_rate": 8.308859229497864e-06, "loss": 1.0407, "step": 6701 }, { "epoch": 0.57, "learning_rate": 8.306155772091689e-06, "loss": 1.0019, "step": 6702 }, { "epoch": 0.57, "learning_rate": 8.303452442134003e-06, "loss": 0.883, "step": 6703 }, { "epoch": 0.57, "learning_rate": 8.300749239828211e-06, "loss": 0.9661, "step": 6704 }, { "epoch": 0.57, "learning_rate": 8.298046165377708e-06, "loss": 1.0847, "step": 6705 }, { "epoch": 0.57, "learning_rate": 8.295343218985873e-06, "loss": 0.9019, "step": 6706 }, { "epoch": 0.57, "learning_rate": 8.292640400856093e-06, "loss": 0.9864, "step": 6707 }, { "epoch": 0.57, "learning_rate": 8.28993771119172e-06, "loss": 1.0205, "step": 6708 }, { "epoch": 0.57, "learning_rate": 8.287235150196123e-06, "loss": 1.0513, "step": 6709 }, { "epoch": 0.57, "learning_rate": 8.28453271807264e-06, "loss": 1.0482, "step": 6710 }, { "epoch": 0.57, "learning_rate": 8.281830415024614e-06, "loss": 0.9709, "step": 6711 }, { "epoch": 0.57, "learning_rate": 8.279128241255365e-06, "loss": 1.0308, "step": 6712 }, { "epoch": 0.57, "learning_rate": 8.276426196968219e-06, "loss": 1.0118, "step": 6713 }, { "epoch": 0.57, "learning_rate": 8.273724282366474e-06, "loss": 1.0667, "step": 6714 }, { "epoch": 0.57, "learning_rate": 8.271022497653437e-06, "loss": 1.0139, "step": 6715 }, { "epoch": 0.57, "learning_rate": 8.268320843032388e-06, "loss": 1.0159, "step": 6716 }, { "epoch": 0.57, "learning_rate": 8.265619318706612e-06, "loss": 1.0331, "step": 6717 }, { "epoch": 0.57, "learning_rate": 8.262917924879373e-06, "loss": 1.0587, "step": 6718 }, { "epoch": 0.57, "learning_rate": 8.26021666175393e-06, "loss": 1.047, "step": 6719 }, { "epoch": 0.57, "learning_rate": 8.257515529533538e-06, "loss": 1.0655, "step": 6720 }, { "epoch": 0.57, "learning_rate": 8.254814528421427e-06, "loss": 1.0636, "step": 6721 }, { "epoch": 0.57, "learning_rate": 8.252113658620833e-06, "loss": 1.0481, "step": 6722 }, { "epoch": 0.57, "learning_rate": 8.24941292033497e-06, "loss": 1.0609, "step": 6723 }, { "epoch": 0.57, "learning_rate": 8.246712313767055e-06, "loss": 1.0353, "step": 6724 }, { "epoch": 0.57, "learning_rate": 8.244011839120277e-06, "loss": 1.0046, "step": 6725 }, { "epoch": 0.57, "learning_rate": 8.241311496597835e-06, "loss": 1.0331, "step": 6726 }, { "epoch": 0.57, "learning_rate": 8.238611286402901e-06, "loss": 1.0312, "step": 6727 }, { "epoch": 0.57, "learning_rate": 8.235911208738653e-06, "loss": 0.97, "step": 6728 }, { "epoch": 0.57, "learning_rate": 8.233211263808243e-06, "loss": 1.0402, "step": 6729 }, { "epoch": 0.57, "learning_rate": 8.230511451814823e-06, "loss": 1.0001, "step": 6730 }, { "epoch": 0.57, "learning_rate": 8.227811772961535e-06, "loss": 0.9943, "step": 6731 }, { "epoch": 0.57, "learning_rate": 8.225112227451509e-06, "loss": 1.0096, "step": 6732 }, { "epoch": 0.57, "learning_rate": 8.222412815487857e-06, "loss": 1.0167, "step": 6733 }, { "epoch": 0.57, "learning_rate": 8.219713537273697e-06, "loss": 1.0442, "step": 6734 }, { "epoch": 0.57, "learning_rate": 8.217014393012123e-06, "loss": 1.0198, "step": 6735 }, { "epoch": 0.57, "learning_rate": 8.214315382906232e-06, "loss": 1.0593, "step": 6736 }, { "epoch": 0.57, "learning_rate": 8.211616507159095e-06, "loss": 0.9336, "step": 6737 }, { "epoch": 0.57, "learning_rate": 8.208917765973787e-06, "loss": 0.9996, "step": 6738 }, { "epoch": 0.57, "learning_rate": 8.206219159553362e-06, "loss": 0.9247, "step": 6739 }, { "epoch": 0.57, "learning_rate": 8.203520688100874e-06, "loss": 1.0101, "step": 6740 }, { "epoch": 0.57, "learning_rate": 8.200822351819356e-06, "loss": 0.9925, "step": 6741 }, { "epoch": 0.57, "learning_rate": 8.198124150911843e-06, "loss": 1.04, "step": 6742 }, { "epoch": 0.57, "learning_rate": 8.19542608558135e-06, "loss": 1.0379, "step": 6743 }, { "epoch": 0.57, "learning_rate": 8.192728156030884e-06, "loss": 1.0383, "step": 6744 }, { "epoch": 0.57, "learning_rate": 8.19003036246345e-06, "loss": 1.0375, "step": 6745 }, { "epoch": 0.57, "learning_rate": 8.187332705082024e-06, "loss": 1.0499, "step": 6746 }, { "epoch": 0.57, "learning_rate": 8.184635184089596e-06, "loss": 0.9983, "step": 6747 }, { "epoch": 0.57, "learning_rate": 8.181937799689121e-06, "loss": 1.0592, "step": 6748 }, { "epoch": 0.57, "learning_rate": 8.17924055208357e-06, "loss": 1.104, "step": 6749 }, { "epoch": 0.57, "learning_rate": 8.176543441475876e-06, "loss": 0.9989, "step": 6750 }, { "epoch": 0.57, "learning_rate": 8.173846468068987e-06, "loss": 1.0095, "step": 6751 }, { "epoch": 0.57, "learning_rate": 8.171149632065819e-06, "loss": 0.9737, "step": 6752 }, { "epoch": 0.57, "learning_rate": 8.168452933669295e-06, "loss": 0.9968, "step": 6753 }, { "epoch": 0.57, "learning_rate": 8.165756373082318e-06, "loss": 1.0235, "step": 6754 }, { "epoch": 0.57, "learning_rate": 8.16305995050778e-06, "loss": 1.0898, "step": 6755 }, { "epoch": 0.57, "learning_rate": 8.160363666148571e-06, "loss": 1.0536, "step": 6756 }, { "epoch": 0.57, "learning_rate": 8.157667520207563e-06, "loss": 0.9739, "step": 6757 }, { "epoch": 0.57, "learning_rate": 8.15497151288762e-06, "loss": 1.0091, "step": 6758 }, { "epoch": 0.57, "learning_rate": 8.152275644391595e-06, "loss": 1.005, "step": 6759 }, { "epoch": 0.57, "learning_rate": 8.149579914922328e-06, "loss": 0.9983, "step": 6760 }, { "epoch": 0.57, "learning_rate": 8.146884324682658e-06, "loss": 1.0069, "step": 6761 }, { "epoch": 0.57, "learning_rate": 8.1441888738754e-06, "loss": 1.0335, "step": 6762 }, { "epoch": 0.57, "learning_rate": 8.141493562703373e-06, "loss": 0.9871, "step": 6763 }, { "epoch": 0.57, "learning_rate": 8.13879839136937e-06, "loss": 0.978, "step": 6764 }, { "epoch": 0.57, "learning_rate": 8.13610336007619e-06, "loss": 1.0415, "step": 6765 }, { "epoch": 0.57, "learning_rate": 8.133408469026603e-06, "loss": 0.9887, "step": 6766 }, { "epoch": 0.57, "learning_rate": 8.130713718423388e-06, "loss": 1.0294, "step": 6767 }, { "epoch": 0.57, "learning_rate": 8.128019108469295e-06, "loss": 1.0019, "step": 6768 }, { "epoch": 0.57, "learning_rate": 8.125324639367083e-06, "loss": 1.0266, "step": 6769 }, { "epoch": 0.57, "learning_rate": 8.122630311319482e-06, "loss": 1.0487, "step": 6770 }, { "epoch": 0.57, "learning_rate": 8.119936124529217e-06, "loss": 0.9985, "step": 6771 }, { "epoch": 0.57, "learning_rate": 8.117242079199014e-06, "loss": 1.0072, "step": 6772 }, { "epoch": 0.57, "learning_rate": 8.114548175531567e-06, "loss": 1.0185, "step": 6773 }, { "epoch": 0.57, "learning_rate": 8.111854413729582e-06, "loss": 0.9916, "step": 6774 }, { "epoch": 0.57, "learning_rate": 8.109160793995734e-06, "loss": 1.0443, "step": 6775 }, { "epoch": 0.57, "learning_rate": 8.106467316532705e-06, "loss": 1.0171, "step": 6776 }, { "epoch": 0.57, "learning_rate": 8.103773981543151e-06, "loss": 0.9911, "step": 6777 }, { "epoch": 0.57, "learning_rate": 8.10108078922973e-06, "loss": 1.0131, "step": 6778 }, { "epoch": 0.57, "learning_rate": 8.098387739795078e-06, "loss": 0.9681, "step": 6779 }, { "epoch": 0.57, "learning_rate": 8.095694833441832e-06, "loss": 0.9408, "step": 6780 }, { "epoch": 0.57, "learning_rate": 8.09300207037261e-06, "loss": 0.974, "step": 6781 }, { "epoch": 0.57, "learning_rate": 8.090309450790019e-06, "loss": 1.084, "step": 6782 }, { "epoch": 0.57, "learning_rate": 8.087616974896657e-06, "loss": 0.8814, "step": 6783 }, { "epoch": 0.57, "learning_rate": 8.084924642895117e-06, "loss": 0.9007, "step": 6784 }, { "epoch": 0.57, "learning_rate": 8.082232454987969e-06, "loss": 1.0129, "step": 6785 }, { "epoch": 0.57, "learning_rate": 8.079540411377786e-06, "loss": 0.9839, "step": 6786 }, { "epoch": 0.57, "learning_rate": 8.076848512267115e-06, "loss": 1.0567, "step": 6787 }, { "epoch": 0.57, "learning_rate": 8.074156757858509e-06, "loss": 0.9806, "step": 6788 }, { "epoch": 0.57, "learning_rate": 8.071465148354493e-06, "loss": 1.0405, "step": 6789 }, { "epoch": 0.58, "learning_rate": 8.068773683957598e-06, "loss": 0.9776, "step": 6790 }, { "epoch": 0.58, "learning_rate": 8.066082364870328e-06, "loss": 0.974, "step": 6791 }, { "epoch": 0.58, "learning_rate": 8.06339119129519e-06, "loss": 1.0499, "step": 6792 }, { "epoch": 0.58, "learning_rate": 8.060700163434667e-06, "loss": 1.0452, "step": 6793 }, { "epoch": 0.58, "learning_rate": 8.058009281491245e-06, "loss": 0.9877, "step": 6794 }, { "epoch": 0.58, "learning_rate": 8.055318545667386e-06, "loss": 1.0287, "step": 6795 }, { "epoch": 0.58, "learning_rate": 8.05262795616555e-06, "loss": 1.0154, "step": 6796 }, { "epoch": 0.58, "learning_rate": 8.049937513188183e-06, "loss": 0.8286, "step": 6797 }, { "epoch": 0.58, "learning_rate": 8.047247216937715e-06, "loss": 1.0059, "step": 6798 }, { "epoch": 0.58, "learning_rate": 8.044557067616577e-06, "loss": 1.0218, "step": 6799 }, { "epoch": 0.58, "learning_rate": 8.041867065427174e-06, "loss": 1.0479, "step": 6800 }, { "epoch": 0.58, "learning_rate": 8.039177210571914e-06, "loss": 1.0039, "step": 6801 }, { "epoch": 0.58, "learning_rate": 8.036487503253182e-06, "loss": 1.0619, "step": 6802 }, { "epoch": 0.58, "learning_rate": 8.033797943673363e-06, "loss": 1.0015, "step": 6803 }, { "epoch": 0.58, "learning_rate": 8.03110853203482e-06, "loss": 1.0331, "step": 6804 }, { "epoch": 0.58, "learning_rate": 8.028419268539915e-06, "loss": 1.0427, "step": 6805 }, { "epoch": 0.58, "learning_rate": 8.025730153390988e-06, "loss": 1.0111, "step": 6806 }, { "epoch": 0.58, "learning_rate": 8.023041186790382e-06, "loss": 1.0238, "step": 6807 }, { "epoch": 0.58, "learning_rate": 8.020352368940414e-06, "loss": 1.058, "step": 6808 }, { "epoch": 0.58, "learning_rate": 8.017663700043396e-06, "loss": 1.0621, "step": 6809 }, { "epoch": 0.58, "learning_rate": 8.014975180301634e-06, "loss": 1.0163, "step": 6810 }, { "epoch": 0.58, "learning_rate": 8.012286809917418e-06, "loss": 1.077, "step": 6811 }, { "epoch": 0.58, "learning_rate": 8.009598589093019e-06, "loss": 1.0395, "step": 6812 }, { "epoch": 0.58, "learning_rate": 8.006910518030716e-06, "loss": 0.9662, "step": 6813 }, { "epoch": 0.58, "learning_rate": 8.004222596932753e-06, "loss": 1.0298, "step": 6814 }, { "epoch": 0.58, "learning_rate": 8.001534826001384e-06, "loss": 1.0474, "step": 6815 }, { "epoch": 0.58, "learning_rate": 7.998847205438838e-06, "loss": 1.0204, "step": 6816 }, { "epoch": 0.58, "learning_rate": 7.996159735447343e-06, "loss": 0.9658, "step": 6817 }, { "epoch": 0.58, "learning_rate": 7.993472416229103e-06, "loss": 1.0319, "step": 6818 }, { "epoch": 0.58, "learning_rate": 7.990785247986324e-06, "loss": 1.0023, "step": 6819 }, { "epoch": 0.58, "learning_rate": 7.988098230921187e-06, "loss": 1.038, "step": 6820 }, { "epoch": 0.58, "learning_rate": 7.985411365235879e-06, "loss": 0.9761, "step": 6821 }, { "epoch": 0.58, "learning_rate": 7.982724651132554e-06, "loss": 1.0095, "step": 6822 }, { "epoch": 0.58, "learning_rate": 7.980038088813373e-06, "loss": 0.9764, "step": 6823 }, { "epoch": 0.58, "learning_rate": 7.977351678480482e-06, "loss": 1.0086, "step": 6824 }, { "epoch": 0.58, "learning_rate": 7.974665420336003e-06, "loss": 0.9811, "step": 6825 }, { "epoch": 0.58, "learning_rate": 7.971979314582065e-06, "loss": 0.9789, "step": 6826 }, { "epoch": 0.58, "learning_rate": 7.969293361420769e-06, "loss": 0.9795, "step": 6827 }, { "epoch": 0.58, "learning_rate": 7.966607561054218e-06, "loss": 0.986, "step": 6828 }, { "epoch": 0.58, "learning_rate": 7.963921913684488e-06, "loss": 0.9917, "step": 6829 }, { "epoch": 0.58, "learning_rate": 7.961236419513668e-06, "loss": 1.0199, "step": 6830 }, { "epoch": 0.58, "learning_rate": 7.958551078743806e-06, "loss": 1.01, "step": 6831 }, { "epoch": 0.58, "learning_rate": 7.955865891576962e-06, "loss": 1.0081, "step": 6832 }, { "epoch": 0.58, "learning_rate": 7.953180858215169e-06, "loss": 1.0379, "step": 6833 }, { "epoch": 0.58, "learning_rate": 7.95049597886046e-06, "loss": 1.0569, "step": 6834 }, { "epoch": 0.58, "learning_rate": 7.947811253714847e-06, "loss": 1.0508, "step": 6835 }, { "epoch": 0.58, "learning_rate": 7.945126682980337e-06, "loss": 1.0138, "step": 6836 }, { "epoch": 0.58, "learning_rate": 7.942442266858923e-06, "loss": 1.0202, "step": 6837 }, { "epoch": 0.58, "learning_rate": 7.939758005552585e-06, "loss": 1.0318, "step": 6838 }, { "epoch": 0.58, "learning_rate": 7.937073899263291e-06, "loss": 1.0238, "step": 6839 }, { "epoch": 0.58, "learning_rate": 7.934389948193004e-06, "loss": 1.0611, "step": 6840 }, { "epoch": 0.58, "learning_rate": 7.931706152543663e-06, "loss": 0.9926, "step": 6841 }, { "epoch": 0.58, "learning_rate": 7.92902251251721e-06, "loss": 0.9889, "step": 6842 }, { "epoch": 0.58, "learning_rate": 7.926339028315562e-06, "loss": 0.9796, "step": 6843 }, { "epoch": 0.58, "learning_rate": 7.923655700140636e-06, "loss": 0.9825, "step": 6844 }, { "epoch": 0.58, "learning_rate": 7.920972528194324e-06, "loss": 0.9175, "step": 6845 }, { "epoch": 0.58, "learning_rate": 7.918289512678521e-06, "loss": 1.0334, "step": 6846 }, { "epoch": 0.58, "learning_rate": 7.915606653795097e-06, "loss": 0.9703, "step": 6847 }, { "epoch": 0.58, "learning_rate": 7.91292395174592e-06, "loss": 1.0588, "step": 6848 }, { "epoch": 0.58, "learning_rate": 7.91024140673284e-06, "loss": 0.9928, "step": 6849 }, { "epoch": 0.58, "learning_rate": 7.9075590189577e-06, "loss": 1.0301, "step": 6850 }, { "epoch": 0.58, "learning_rate": 7.904876788622328e-06, "loss": 1.044, "step": 6851 }, { "epoch": 0.58, "learning_rate": 7.902194715928536e-06, "loss": 1.0334, "step": 6852 }, { "epoch": 0.58, "learning_rate": 7.899512801078136e-06, "loss": 1.0107, "step": 6853 }, { "epoch": 0.58, "learning_rate": 7.896831044272915e-06, "loss": 1.0238, "step": 6854 }, { "epoch": 0.58, "learning_rate": 7.894149445714661e-06, "loss": 1.0021, "step": 6855 }, { "epoch": 0.58, "learning_rate": 7.891468005605134e-06, "loss": 0.992, "step": 6856 }, { "epoch": 0.58, "learning_rate": 7.888786724146102e-06, "loss": 1.0648, "step": 6857 }, { "epoch": 0.58, "learning_rate": 7.8861056015393e-06, "loss": 1.0065, "step": 6858 }, { "epoch": 0.58, "learning_rate": 7.883424637986472e-06, "loss": 0.8851, "step": 6859 }, { "epoch": 0.58, "learning_rate": 7.880743833689327e-06, "loss": 1.0591, "step": 6860 }, { "epoch": 0.58, "learning_rate": 7.878063188849587e-06, "loss": 1.0182, "step": 6861 }, { "epoch": 0.58, "learning_rate": 7.875382703668941e-06, "loss": 1.0533, "step": 6862 }, { "epoch": 0.58, "learning_rate": 7.87270237834908e-06, "loss": 0.9656, "step": 6863 }, { "epoch": 0.58, "learning_rate": 7.870022213091671e-06, "loss": 1.014, "step": 6864 }, { "epoch": 0.58, "learning_rate": 7.867342208098384e-06, "loss": 1.0092, "step": 6865 }, { "epoch": 0.58, "learning_rate": 7.864662363570858e-06, "loss": 0.9809, "step": 6866 }, { "epoch": 0.58, "learning_rate": 7.861982679710741e-06, "loss": 1.0565, "step": 6867 }, { "epoch": 0.58, "learning_rate": 7.859303156719646e-06, "loss": 1.0398, "step": 6868 }, { "epoch": 0.58, "learning_rate": 7.8566237947992e-06, "loss": 0.9794, "step": 6869 }, { "epoch": 0.58, "learning_rate": 7.853944594150991e-06, "loss": 1.0261, "step": 6870 }, { "epoch": 0.58, "learning_rate": 7.85126555497662e-06, "loss": 0.9872, "step": 6871 }, { "epoch": 0.58, "learning_rate": 7.848586677477652e-06, "loss": 1.0296, "step": 6872 }, { "epoch": 0.58, "learning_rate": 7.845907961855663e-06, "loss": 0.9973, "step": 6873 }, { "epoch": 0.58, "learning_rate": 7.843229408312194e-06, "loss": 1.0032, "step": 6874 }, { "epoch": 0.58, "learning_rate": 7.840551017048791e-06, "loss": 1.0132, "step": 6875 }, { "epoch": 0.58, "learning_rate": 7.837872788266986e-06, "loss": 1.002, "step": 6876 }, { "epoch": 0.58, "learning_rate": 7.835194722168284e-06, "loss": 1.007, "step": 6877 }, { "epoch": 0.58, "learning_rate": 7.832516818954202e-06, "loss": 0.9547, "step": 6878 }, { "epoch": 0.58, "learning_rate": 7.829839078826215e-06, "loss": 1.0083, "step": 6879 }, { "epoch": 0.58, "learning_rate": 7.827161501985818e-06, "loss": 1.0389, "step": 6880 }, { "epoch": 0.58, "learning_rate": 7.824484088634467e-06, "loss": 1.0115, "step": 6881 }, { "epoch": 0.58, "learning_rate": 7.821806838973622e-06, "loss": 1.025, "step": 6882 }, { "epoch": 0.58, "learning_rate": 7.81912975320472e-06, "loss": 0.9863, "step": 6883 }, { "epoch": 0.58, "learning_rate": 7.816452831529196e-06, "loss": 1.0221, "step": 6884 }, { "epoch": 0.58, "learning_rate": 7.813776074148462e-06, "loss": 1.0617, "step": 6885 }, { "epoch": 0.58, "learning_rate": 7.811099481263928e-06, "loss": 0.9664, "step": 6886 }, { "epoch": 0.58, "learning_rate": 7.808423053076983e-06, "loss": 1.0531, "step": 6887 }, { "epoch": 0.58, "learning_rate": 7.805746789789012e-06, "loss": 0.9813, "step": 6888 }, { "epoch": 0.58, "learning_rate": 7.803070691601376e-06, "loss": 0.9564, "step": 6889 }, { "epoch": 0.58, "learning_rate": 7.800394758715435e-06, "loss": 1.0291, "step": 6890 }, { "epoch": 0.58, "learning_rate": 7.79771899133253e-06, "loss": 0.8708, "step": 6891 }, { "epoch": 0.58, "learning_rate": 7.795043389653995e-06, "loss": 1.0175, "step": 6892 }, { "epoch": 0.58, "learning_rate": 7.79236795388114e-06, "loss": 0.9375, "step": 6893 }, { "epoch": 0.58, "learning_rate": 7.789692684215284e-06, "loss": 0.9985, "step": 6894 }, { "epoch": 0.58, "learning_rate": 7.787017580857706e-06, "loss": 1.0034, "step": 6895 }, { "epoch": 0.58, "learning_rate": 7.784342644009695e-06, "loss": 1.0381, "step": 6896 }, { "epoch": 0.58, "learning_rate": 7.781667873872513e-06, "loss": 0.9989, "step": 6897 }, { "epoch": 0.58, "learning_rate": 7.778993270647424e-06, "loss": 1.0448, "step": 6898 }, { "epoch": 0.58, "learning_rate": 7.77631883453566e-06, "loss": 1.0236, "step": 6899 }, { "epoch": 0.58, "learning_rate": 7.77364456573846e-06, "loss": 1.0232, "step": 6900 }, { "epoch": 0.58, "learning_rate": 7.770970464457039e-06, "loss": 1.0495, "step": 6901 }, { "epoch": 0.58, "learning_rate": 7.768296530892603e-06, "loss": 1.0496, "step": 6902 }, { "epoch": 0.58, "learning_rate": 7.765622765246345e-06, "loss": 0.983, "step": 6903 }, { "epoch": 0.58, "learning_rate": 7.762949167719441e-06, "loss": 1.0118, "step": 6904 }, { "epoch": 0.58, "learning_rate": 7.760275738513064e-06, "loss": 0.9651, "step": 6905 }, { "epoch": 0.58, "learning_rate": 7.757602477828359e-06, "loss": 1.0344, "step": 6906 }, { "epoch": 0.58, "learning_rate": 7.754929385866481e-06, "loss": 0.9809, "step": 6907 }, { "epoch": 0.59, "learning_rate": 7.752256462828547e-06, "loss": 0.9211, "step": 6908 }, { "epoch": 0.59, "learning_rate": 7.749583708915683e-06, "loss": 0.9725, "step": 6909 }, { "epoch": 0.59, "learning_rate": 7.746911124328985e-06, "loss": 1.0128, "step": 6910 }, { "epoch": 0.59, "learning_rate": 7.744238709269552e-06, "loss": 1.0415, "step": 6911 }, { "epoch": 0.59, "learning_rate": 7.741566463938454e-06, "loss": 1.0203, "step": 6912 }, { "epoch": 0.59, "learning_rate": 7.738894388536764e-06, "loss": 1.0563, "step": 6913 }, { "epoch": 0.59, "learning_rate": 7.736222483265526e-06, "loss": 1.007, "step": 6914 }, { "epoch": 0.59, "learning_rate": 7.73355074832579e-06, "loss": 1.0521, "step": 6915 }, { "epoch": 0.59, "learning_rate": 7.730879183918578e-06, "loss": 0.9989, "step": 6916 }, { "epoch": 0.59, "learning_rate": 7.728207790244903e-06, "loss": 0.977, "step": 6917 }, { "epoch": 0.59, "learning_rate": 7.725536567505767e-06, "loss": 1.0413, "step": 6918 }, { "epoch": 0.59, "learning_rate": 7.722865515902162e-06, "loss": 1.0538, "step": 6919 }, { "epoch": 0.59, "learning_rate": 7.720194635635058e-06, "loss": 1.0294, "step": 6920 }, { "epoch": 0.59, "learning_rate": 7.717523926905423e-06, "loss": 1.0522, "step": 6921 }, { "epoch": 0.59, "learning_rate": 7.714853389914203e-06, "loss": 1.0741, "step": 6922 }, { "epoch": 0.59, "learning_rate": 7.71218302486234e-06, "loss": 0.8295, "step": 6923 }, { "epoch": 0.59, "learning_rate": 7.709512831950749e-06, "loss": 1.0366, "step": 6924 }, { "epoch": 0.59, "learning_rate": 7.706842811380353e-06, "loss": 0.9925, "step": 6925 }, { "epoch": 0.59, "learning_rate": 7.70417296335204e-06, "loss": 0.9782, "step": 6926 }, { "epoch": 0.59, "learning_rate": 7.701503288066699e-06, "loss": 1.0365, "step": 6927 }, { "epoch": 0.59, "learning_rate": 7.6988337857252e-06, "loss": 1.0073, "step": 6928 }, { "epoch": 0.59, "learning_rate": 7.69616445652841e-06, "loss": 1.0446, "step": 6929 }, { "epoch": 0.59, "learning_rate": 7.693495300677169e-06, "loss": 0.9174, "step": 6930 }, { "epoch": 0.59, "learning_rate": 7.690826318372306e-06, "loss": 1.0377, "step": 6931 }, { "epoch": 0.59, "learning_rate": 7.68815750981465e-06, "loss": 0.9747, "step": 6932 }, { "epoch": 0.59, "learning_rate": 7.685488875204995e-06, "loss": 1.0227, "step": 6933 }, { "epoch": 0.59, "learning_rate": 7.682820414744152e-06, "loss": 1.0302, "step": 6934 }, { "epoch": 0.59, "learning_rate": 7.680152128632887e-06, "loss": 0.9848, "step": 6935 }, { "epoch": 0.59, "learning_rate": 7.677484017071978e-06, "loss": 1.0369, "step": 6936 }, { "epoch": 0.59, "learning_rate": 7.67481608026217e-06, "loss": 0.9927, "step": 6937 }, { "epoch": 0.59, "learning_rate": 7.672148318404213e-06, "loss": 1.0052, "step": 6938 }, { "epoch": 0.59, "learning_rate": 7.669480731698829e-06, "loss": 1.009, "step": 6939 }, { "epoch": 0.59, "learning_rate": 7.666813320346737e-06, "loss": 1.0133, "step": 6940 }, { "epoch": 0.59, "learning_rate": 7.664146084548634e-06, "loss": 0.9831, "step": 6941 }, { "epoch": 0.59, "learning_rate": 7.661479024505216e-06, "loss": 1.0082, "step": 6942 }, { "epoch": 0.59, "learning_rate": 7.658812140417151e-06, "loss": 1.0275, "step": 6943 }, { "epoch": 0.59, "learning_rate": 7.656145432485105e-06, "loss": 0.9716, "step": 6944 }, { "epoch": 0.59, "learning_rate": 7.653478900909725e-06, "loss": 0.989, "step": 6945 }, { "epoch": 0.59, "learning_rate": 7.650812545891651e-06, "loss": 1.0807, "step": 6946 }, { "epoch": 0.59, "learning_rate": 7.648146367631496e-06, "loss": 1.0395, "step": 6947 }, { "epoch": 0.59, "learning_rate": 7.64548036632988e-06, "loss": 1.0079, "step": 6948 }, { "epoch": 0.59, "learning_rate": 7.64281454218739e-06, "loss": 0.9981, "step": 6949 }, { "epoch": 0.59, "learning_rate": 7.640148895404613e-06, "loss": 0.9852, "step": 6950 }, { "epoch": 0.59, "learning_rate": 7.637483426182115e-06, "loss": 1.0238, "step": 6951 }, { "epoch": 0.59, "learning_rate": 7.634818134720452e-06, "loss": 1.0421, "step": 6952 }, { "epoch": 0.59, "learning_rate": 7.632153021220172e-06, "loss": 0.9988, "step": 6953 }, { "epoch": 0.59, "learning_rate": 7.629488085881798e-06, "loss": 1.0475, "step": 6954 }, { "epoch": 0.59, "learning_rate": 7.626823328905848e-06, "loss": 0.9857, "step": 6955 }, { "epoch": 0.59, "learning_rate": 7.624158750492823e-06, "loss": 1.01, "step": 6956 }, { "epoch": 0.59, "learning_rate": 7.621494350843213e-06, "loss": 0.9817, "step": 6957 }, { "epoch": 0.59, "learning_rate": 7.618830130157488e-06, "loss": 0.9608, "step": 6958 }, { "epoch": 0.59, "learning_rate": 7.616166088636117e-06, "loss": 1.0097, "step": 6959 }, { "epoch": 0.59, "learning_rate": 7.613502226479541e-06, "loss": 1.0547, "step": 6960 }, { "epoch": 0.59, "learning_rate": 7.610838543888204e-06, "loss": 1.0207, "step": 6961 }, { "epoch": 0.59, "learning_rate": 7.608175041062517e-06, "loss": 1.042, "step": 6962 }, { "epoch": 0.59, "learning_rate": 7.6055117182028955e-06, "loss": 0.9968, "step": 6963 }, { "epoch": 0.59, "learning_rate": 7.602848575509727e-06, "loss": 0.9696, "step": 6964 }, { "epoch": 0.59, "learning_rate": 7.6001856131834e-06, "loss": 0.9578, "step": 6965 }, { "epoch": 0.59, "learning_rate": 7.597522831424275e-06, "loss": 0.9828, "step": 6966 }, { "epoch": 0.59, "learning_rate": 7.594860230432709e-06, "loss": 1.0007, "step": 6967 }, { "epoch": 0.59, "learning_rate": 7.5921978104090375e-06, "loss": 1.0406, "step": 6968 }, { "epoch": 0.59, "learning_rate": 7.589535571553595e-06, "loss": 1.0123, "step": 6969 }, { "epoch": 0.59, "learning_rate": 7.586873514066685e-06, "loss": 1.001, "step": 6970 }, { "epoch": 0.59, "learning_rate": 7.5842116381486105e-06, "loss": 0.9699, "step": 6971 }, { "epoch": 0.59, "learning_rate": 7.581549943999656e-06, "loss": 0.8999, "step": 6972 }, { "epoch": 0.59, "learning_rate": 7.578888431820097e-06, "loss": 0.9885, "step": 6973 }, { "epoch": 0.59, "learning_rate": 7.576227101810183e-06, "loss": 1.0103, "step": 6974 }, { "epoch": 0.59, "learning_rate": 7.573565954170169e-06, "loss": 0.9759, "step": 6975 }, { "epoch": 0.59, "learning_rate": 7.570904989100274e-06, "loss": 1.0049, "step": 6976 }, { "epoch": 0.59, "learning_rate": 7.568244206800726e-06, "loss": 0.9999, "step": 6977 }, { "epoch": 0.59, "learning_rate": 7.565583607471716e-06, "loss": 1.0001, "step": 6978 }, { "epoch": 0.59, "learning_rate": 7.562923191313441e-06, "loss": 1.0422, "step": 6979 }, { "epoch": 0.59, "learning_rate": 7.5602629585260786e-06, "loss": 1.0102, "step": 6980 }, { "epoch": 0.59, "learning_rate": 7.557602909309787e-06, "loss": 1.028, "step": 6981 }, { "epoch": 0.59, "learning_rate": 7.554943043864712e-06, "loss": 0.9998, "step": 6982 }, { "epoch": 0.59, "learning_rate": 7.5522833623909905e-06, "loss": 1.0059, "step": 6983 }, { "epoch": 0.59, "learning_rate": 7.549623865088743e-06, "loss": 1.031, "step": 6984 }, { "epoch": 0.59, "learning_rate": 7.546964552158072e-06, "loss": 1.0328, "step": 6985 }, { "epoch": 0.59, "learning_rate": 7.5443054237990754e-06, "loss": 1.0568, "step": 6986 }, { "epoch": 0.59, "learning_rate": 7.541646480211825e-06, "loss": 1.0194, "step": 6987 }, { "epoch": 0.59, "learning_rate": 7.538987721596393e-06, "loss": 1.0373, "step": 6988 }, { "epoch": 0.59, "learning_rate": 7.536329148152823e-06, "loss": 1.0204, "step": 6989 }, { "epoch": 0.59, "learning_rate": 7.533670760081159e-06, "loss": 0.989, "step": 6990 }, { "epoch": 0.59, "learning_rate": 7.531012557581416e-06, "loss": 0.9373, "step": 6991 }, { "epoch": 0.59, "learning_rate": 7.528354540853611e-06, "loss": 1.0483, "step": 6992 }, { "epoch": 0.59, "learning_rate": 7.525696710097731e-06, "loss": 1.0187, "step": 6993 }, { "epoch": 0.59, "learning_rate": 7.523039065513764e-06, "loss": 1.0442, "step": 6994 }, { "epoch": 0.59, "learning_rate": 7.520381607301673e-06, "loss": 1.0446, "step": 6995 }, { "epoch": 0.59, "learning_rate": 7.5177243356614115e-06, "loss": 1.0443, "step": 6996 }, { "epoch": 0.59, "learning_rate": 7.5150672507929175e-06, "loss": 1.0092, "step": 6997 }, { "epoch": 0.59, "learning_rate": 7.5124103528961165e-06, "loss": 1.0521, "step": 6998 }, { "epoch": 0.59, "learning_rate": 7.509753642170921e-06, "loss": 0.9925, "step": 6999 }, { "epoch": 0.59, "learning_rate": 7.507097118817226e-06, "loss": 1.0437, "step": 7000 }, { "epoch": 0.59, "learning_rate": 7.504440783034912e-06, "loss": 1.0335, "step": 7001 }, { "epoch": 0.59, "learning_rate": 7.501784635023852e-06, "loss": 1.0371, "step": 7002 }, { "epoch": 0.59, "learning_rate": 7.499128674983894e-06, "loss": 0.9872, "step": 7003 }, { "epoch": 0.59, "learning_rate": 7.496472903114882e-06, "loss": 0.8492, "step": 7004 }, { "epoch": 0.59, "learning_rate": 7.493817319616648e-06, "loss": 0.9925, "step": 7005 }, { "epoch": 0.59, "learning_rate": 7.491161924688991e-06, "loss": 1.0573, "step": 7006 }, { "epoch": 0.59, "learning_rate": 7.48850671853172e-06, "loss": 0.9612, "step": 7007 }, { "epoch": 0.59, "learning_rate": 7.4858517013446135e-06, "loss": 1.0547, "step": 7008 }, { "epoch": 0.59, "learning_rate": 7.48319687332744e-06, "loss": 0.9953, "step": 7009 }, { "epoch": 0.59, "learning_rate": 7.480542234679955e-06, "loss": 1.0439, "step": 7010 }, { "epoch": 0.59, "learning_rate": 7.4778877856019025e-06, "loss": 1.0344, "step": 7011 }, { "epoch": 0.59, "learning_rate": 7.475233526293002e-06, "loss": 1.0097, "step": 7012 }, { "epoch": 0.59, "learning_rate": 7.472579456952973e-06, "loss": 1.0295, "step": 7013 }, { "epoch": 0.59, "learning_rate": 7.4699255777815075e-06, "loss": 1.0466, "step": 7014 }, { "epoch": 0.59, "learning_rate": 7.467271888978296e-06, "loss": 1.0584, "step": 7015 }, { "epoch": 0.59, "learning_rate": 7.464618390742999e-06, "loss": 1.0032, "step": 7016 }, { "epoch": 0.59, "learning_rate": 7.461965083275281e-06, "loss": 1.0437, "step": 7017 }, { "epoch": 0.59, "learning_rate": 7.459311966774773e-06, "loss": 1.0437, "step": 7018 }, { "epoch": 0.59, "learning_rate": 7.45665904144111e-06, "loss": 1.0688, "step": 7019 }, { "epoch": 0.59, "learning_rate": 7.454006307473896e-06, "loss": 1.0524, "step": 7020 }, { "epoch": 0.59, "learning_rate": 7.451353765072735e-06, "loss": 1.0174, "step": 7021 }, { "epoch": 0.59, "learning_rate": 7.448701414437206e-06, "loss": 1.0137, "step": 7022 }, { "epoch": 0.59, "learning_rate": 7.446049255766878e-06, "loss": 1.0618, "step": 7023 }, { "epoch": 0.59, "learning_rate": 7.4433972892613075e-06, "loss": 1.0223, "step": 7024 }, { "epoch": 0.59, "learning_rate": 7.440745515120034e-06, "loss": 0.9853, "step": 7025 }, { "epoch": 0.6, "learning_rate": 7.438093933542576e-06, "loss": 1.0469, "step": 7026 }, { "epoch": 0.6, "learning_rate": 7.435442544728453e-06, "loss": 0.9953, "step": 7027 }, { "epoch": 0.6, "learning_rate": 7.432791348877155e-06, "loss": 0.9899, "step": 7028 }, { "epoch": 0.6, "learning_rate": 7.430140346188171e-06, "loss": 1.0415, "step": 7029 }, { "epoch": 0.6, "learning_rate": 7.427489536860957e-06, "loss": 1.035, "step": 7030 }, { "epoch": 0.6, "learning_rate": 7.424838921094973e-06, "loss": 0.993, "step": 7031 }, { "epoch": 0.6, "learning_rate": 7.422188499089661e-06, "loss": 0.9742, "step": 7032 }, { "epoch": 0.6, "learning_rate": 7.419538271044436e-06, "loss": 1.014, "step": 7033 }, { "epoch": 0.6, "learning_rate": 7.4168882371587125e-06, "loss": 1.0791, "step": 7034 }, { "epoch": 0.6, "learning_rate": 7.414238397631881e-06, "loss": 0.9914, "step": 7035 }, { "epoch": 0.6, "learning_rate": 7.4115887526633235e-06, "loss": 0.9923, "step": 7036 }, { "epoch": 0.6, "learning_rate": 7.4089393024524045e-06, "loss": 1.0697, "step": 7037 }, { "epoch": 0.6, "learning_rate": 7.406290047198476e-06, "loss": 0.9793, "step": 7038 }, { "epoch": 0.6, "learning_rate": 7.403640987100867e-06, "loss": 0.9569, "step": 7039 }, { "epoch": 0.6, "learning_rate": 7.400992122358907e-06, "loss": 1.0215, "step": 7040 }, { "epoch": 0.6, "learning_rate": 7.398343453171897e-06, "loss": 1.023, "step": 7041 }, { "epoch": 0.6, "learning_rate": 7.395694979739132e-06, "loss": 1.0374, "step": 7042 }, { "epoch": 0.6, "learning_rate": 7.393046702259884e-06, "loss": 1.0188, "step": 7043 }, { "epoch": 0.6, "learning_rate": 7.390398620933421e-06, "loss": 0.9612, "step": 7044 }, { "epoch": 0.6, "learning_rate": 7.387750735958985e-06, "loss": 0.9582, "step": 7045 }, { "epoch": 0.6, "learning_rate": 7.385103047535814e-06, "loss": 1.0355, "step": 7046 }, { "epoch": 0.6, "learning_rate": 7.38245555586312e-06, "loss": 1.0187, "step": 7047 }, { "epoch": 0.6, "learning_rate": 7.379808261140111e-06, "loss": 0.911, "step": 7048 }, { "epoch": 0.6, "learning_rate": 7.377161163565972e-06, "loss": 1.0563, "step": 7049 }, { "epoch": 0.6, "learning_rate": 7.374514263339878e-06, "loss": 1.0292, "step": 7050 }, { "epoch": 0.6, "learning_rate": 7.371867560660986e-06, "loss": 0.9979, "step": 7051 }, { "epoch": 0.6, "learning_rate": 7.369221055728442e-06, "loss": 0.9827, "step": 7052 }, { "epoch": 0.6, "learning_rate": 7.366574748741368e-06, "loss": 0.9888, "step": 7053 }, { "epoch": 0.6, "learning_rate": 7.363928639898889e-06, "loss": 1.0573, "step": 7054 }, { "epoch": 0.6, "learning_rate": 7.361282729400094e-06, "loss": 0.9902, "step": 7055 }, { "epoch": 0.6, "learning_rate": 7.35863701744407e-06, "loss": 1.0062, "step": 7056 }, { "epoch": 0.6, "learning_rate": 7.3559915042298916e-06, "loss": 1.0453, "step": 7057 }, { "epoch": 0.6, "learning_rate": 7.353346189956603e-06, "loss": 1.0087, "step": 7058 }, { "epoch": 0.6, "learning_rate": 7.350701074823252e-06, "loss": 1.0285, "step": 7059 }, { "epoch": 0.6, "learning_rate": 7.348056159028858e-06, "loss": 0.9722, "step": 7060 }, { "epoch": 0.6, "learning_rate": 7.345411442772433e-06, "loss": 1.032, "step": 7061 }, { "epoch": 0.6, "learning_rate": 7.342766926252969e-06, "loss": 1.0346, "step": 7062 }, { "epoch": 0.6, "learning_rate": 7.340122609669445e-06, "loss": 1.042, "step": 7063 }, { "epoch": 0.6, "learning_rate": 7.337478493220827e-06, "loss": 1.0627, "step": 7064 }, { "epoch": 0.6, "learning_rate": 7.334834577106065e-06, "loss": 1.0058, "step": 7065 }, { "epoch": 0.6, "learning_rate": 7.332190861524084e-06, "loss": 0.9841, "step": 7066 }, { "epoch": 0.6, "learning_rate": 7.3295473466738176e-06, "loss": 0.9966, "step": 7067 }, { "epoch": 0.6, "learning_rate": 7.326904032754155e-06, "loss": 0.9871, "step": 7068 }, { "epoch": 0.6, "learning_rate": 7.324260919963997e-06, "loss": 1.0228, "step": 7069 }, { "epoch": 0.6, "learning_rate": 7.321618008502206e-06, "loss": 1.0257, "step": 7070 }, { "epoch": 0.6, "learning_rate": 7.318975298567651e-06, "loss": 0.9807, "step": 7071 }, { "epoch": 0.6, "learning_rate": 7.316332790359165e-06, "loss": 0.9854, "step": 7072 }, { "epoch": 0.6, "learning_rate": 7.313690484075586e-06, "loss": 1.0277, "step": 7073 }, { "epoch": 0.6, "learning_rate": 7.311048379915718e-06, "loss": 0.9988, "step": 7074 }, { "epoch": 0.6, "learning_rate": 7.308406478078367e-06, "loss": 1.0124, "step": 7075 }, { "epoch": 0.6, "learning_rate": 7.305764778762308e-06, "loss": 1.0223, "step": 7076 }, { "epoch": 0.6, "learning_rate": 7.3031232821663115e-06, "loss": 1.0317, "step": 7077 }, { "epoch": 0.6, "learning_rate": 7.300481988489129e-06, "loss": 0.9794, "step": 7078 }, { "epoch": 0.6, "learning_rate": 7.297840897929498e-06, "loss": 1.0108, "step": 7079 }, { "epoch": 0.6, "learning_rate": 7.295200010686136e-06, "loss": 1.0046, "step": 7080 }, { "epoch": 0.6, "learning_rate": 7.2925593269577566e-06, "loss": 1.048, "step": 7081 }, { "epoch": 0.6, "learning_rate": 7.28991884694304e-06, "loss": 0.8727, "step": 7082 }, { "epoch": 0.6, "learning_rate": 7.2872785708406705e-06, "loss": 0.9606, "step": 7083 }, { "epoch": 0.6, "learning_rate": 7.284638498849306e-06, "loss": 1.0441, "step": 7084 }, { "epoch": 0.6, "learning_rate": 7.2819986311675886e-06, "loss": 1.0047, "step": 7085 }, { "epoch": 0.6, "learning_rate": 7.2793589679941525e-06, "loss": 0.9701, "step": 7086 }, { "epoch": 0.6, "learning_rate": 7.276719509527603e-06, "loss": 1.0485, "step": 7087 }, { "epoch": 0.6, "learning_rate": 7.274080255966551e-06, "loss": 1.0475, "step": 7088 }, { "epoch": 0.6, "learning_rate": 7.2714412075095696e-06, "loss": 0.9874, "step": 7089 }, { "epoch": 0.6, "learning_rate": 7.26880236435523e-06, "loss": 0.9497, "step": 7090 }, { "epoch": 0.6, "learning_rate": 7.2661637267020835e-06, "loss": 1.0379, "step": 7091 }, { "epoch": 0.6, "learning_rate": 7.26352529474867e-06, "loss": 1.0022, "step": 7092 }, { "epoch": 0.6, "learning_rate": 7.260887068693504e-06, "loss": 1.0067, "step": 7093 }, { "epoch": 0.6, "learning_rate": 7.2582490487350995e-06, "loss": 1.086, "step": 7094 }, { "epoch": 0.6, "learning_rate": 7.25561123507194e-06, "loss": 0.9989, "step": 7095 }, { "epoch": 0.6, "learning_rate": 7.252973627902507e-06, "loss": 1.0356, "step": 7096 }, { "epoch": 0.6, "learning_rate": 7.250336227425251e-06, "loss": 1.0052, "step": 7097 }, { "epoch": 0.6, "learning_rate": 7.247699033838626e-06, "loss": 0.9466, "step": 7098 }, { "epoch": 0.6, "learning_rate": 7.24506204734105e-06, "loss": 1.0336, "step": 7099 }, { "epoch": 0.6, "learning_rate": 7.242425268130944e-06, "loss": 0.9975, "step": 7100 }, { "epoch": 0.6, "learning_rate": 7.239788696406698e-06, "loss": 1.0137, "step": 7101 }, { "epoch": 0.6, "learning_rate": 7.2371523323667e-06, "loss": 1.0429, "step": 7102 }, { "epoch": 0.6, "learning_rate": 7.2345161762093105e-06, "loss": 1.0599, "step": 7103 }, { "epoch": 0.6, "learning_rate": 7.231880228132882e-06, "loss": 0.9553, "step": 7104 }, { "epoch": 0.6, "learning_rate": 7.229244488335748e-06, "loss": 1.0346, "step": 7105 }, { "epoch": 0.6, "learning_rate": 7.22660895701623e-06, "loss": 1.0061, "step": 7106 }, { "epoch": 0.6, "learning_rate": 7.223973634372623e-06, "loss": 0.989, "step": 7107 }, { "epoch": 0.6, "learning_rate": 7.221338520603222e-06, "loss": 1.0181, "step": 7108 }, { "epoch": 0.6, "learning_rate": 7.2187036159063e-06, "loss": 0.9839, "step": 7109 }, { "epoch": 0.6, "learning_rate": 7.2160689204801035e-06, "loss": 0.869, "step": 7110 }, { "epoch": 0.6, "learning_rate": 7.213434434522886e-06, "loss": 0.8307, "step": 7111 }, { "epoch": 0.6, "learning_rate": 7.210800158232858e-06, "loss": 0.9926, "step": 7112 }, { "epoch": 0.6, "learning_rate": 7.2081660918082395e-06, "loss": 1.0064, "step": 7113 }, { "epoch": 0.6, "learning_rate": 7.2055322354472165e-06, "loss": 1.0604, "step": 7114 }, { "epoch": 0.6, "learning_rate": 7.202898589347969e-06, "loss": 0.9797, "step": 7115 }, { "epoch": 0.6, "learning_rate": 7.200265153708656e-06, "loss": 1.0367, "step": 7116 }, { "epoch": 0.6, "learning_rate": 7.197631928727425e-06, "loss": 1.0197, "step": 7117 }, { "epoch": 0.6, "learning_rate": 7.1949989146024046e-06, "loss": 0.9992, "step": 7118 }, { "epoch": 0.6, "learning_rate": 7.19236611153171e-06, "loss": 0.9405, "step": 7119 }, { "epoch": 0.6, "learning_rate": 7.189733519713433e-06, "loss": 1.0704, "step": 7120 }, { "epoch": 0.6, "learning_rate": 7.187101139345666e-06, "loss": 0.981, "step": 7121 }, { "epoch": 0.6, "learning_rate": 7.184468970626463e-06, "loss": 0.953, "step": 7122 }, { "epoch": 0.6, "learning_rate": 7.1818370137538855e-06, "loss": 1.0103, "step": 7123 }, { "epoch": 0.6, "learning_rate": 7.179205268925958e-06, "loss": 0.9613, "step": 7124 }, { "epoch": 0.6, "learning_rate": 7.176573736340704e-06, "loss": 1.0517, "step": 7125 }, { "epoch": 0.6, "learning_rate": 7.173942416196123e-06, "loss": 0.988, "step": 7126 }, { "epoch": 0.6, "learning_rate": 7.171311308690205e-06, "loss": 0.9749, "step": 7127 }, { "epoch": 0.6, "learning_rate": 7.168680414020916e-06, "loss": 1.0313, "step": 7128 }, { "epoch": 0.6, "learning_rate": 7.166049732386213e-06, "loss": 0.9877, "step": 7129 }, { "epoch": 0.6, "learning_rate": 7.163419263984034e-06, "loss": 1.0395, "step": 7130 }, { "epoch": 0.6, "learning_rate": 7.160789009012299e-06, "loss": 0.9477, "step": 7131 }, { "epoch": 0.6, "learning_rate": 7.158158967668916e-06, "loss": 1.0291, "step": 7132 }, { "epoch": 0.6, "learning_rate": 7.155529140151775e-06, "loss": 0.9367, "step": 7133 }, { "epoch": 0.6, "learning_rate": 7.152899526658745e-06, "loss": 1.049, "step": 7134 }, { "epoch": 0.6, "learning_rate": 7.150270127387688e-06, "loss": 0.9789, "step": 7135 }, { "epoch": 0.6, "learning_rate": 7.147640942536451e-06, "loss": 1.0141, "step": 7136 }, { "epoch": 0.6, "learning_rate": 7.14501197230285e-06, "loss": 0.9764, "step": 7137 }, { "epoch": 0.6, "learning_rate": 7.142383216884703e-06, "loss": 0.9977, "step": 7138 }, { "epoch": 0.6, "learning_rate": 7.1397546764797935e-06, "loss": 0.9575, "step": 7139 }, { "epoch": 0.6, "learning_rate": 7.137126351285908e-06, "loss": 0.9937, "step": 7140 }, { "epoch": 0.6, "learning_rate": 7.1344982415008005e-06, "loss": 1.0159, "step": 7141 }, { "epoch": 0.6, "learning_rate": 7.131870347322219e-06, "loss": 1.0002, "step": 7142 }, { "epoch": 0.6, "learning_rate": 7.129242668947891e-06, "loss": 1.0123, "step": 7143 }, { "epoch": 0.61, "learning_rate": 7.126615206575531e-06, "loss": 1.0147, "step": 7144 }, { "epoch": 0.61, "learning_rate": 7.12398796040283e-06, "loss": 1.027, "step": 7145 }, { "epoch": 0.61, "learning_rate": 7.121360930627472e-06, "loss": 0.9478, "step": 7146 }, { "epoch": 0.61, "learning_rate": 7.118734117447117e-06, "loss": 1.0517, "step": 7147 }, { "epoch": 0.61, "learning_rate": 7.116107521059417e-06, "loss": 1.0016, "step": 7148 }, { "epoch": 0.61, "learning_rate": 7.113481141661995e-06, "loss": 1.0296, "step": 7149 }, { "epoch": 0.61, "learning_rate": 7.110854979452476e-06, "loss": 1.0474, "step": 7150 }, { "epoch": 0.61, "learning_rate": 7.108229034628447e-06, "loss": 1.01, "step": 7151 }, { "epoch": 0.61, "learning_rate": 7.105603307387499e-06, "loss": 1.0078, "step": 7152 }, { "epoch": 0.61, "learning_rate": 7.1029777979271905e-06, "loss": 1.0196, "step": 7153 }, { "epoch": 0.61, "learning_rate": 7.100352506445076e-06, "loss": 1.0166, "step": 7154 }, { "epoch": 0.61, "learning_rate": 7.097727433138683e-06, "loss": 1.0497, "step": 7155 }, { "epoch": 0.61, "learning_rate": 7.095102578205532e-06, "loss": 1.0077, "step": 7156 }, { "epoch": 0.61, "learning_rate": 7.092477941843121e-06, "loss": 0.994, "step": 7157 }, { "epoch": 0.61, "learning_rate": 7.089853524248931e-06, "loss": 0.8581, "step": 7158 }, { "epoch": 0.61, "learning_rate": 7.087229325620433e-06, "loss": 1.0082, "step": 7159 }, { "epoch": 0.61, "learning_rate": 7.084605346155071e-06, "loss": 0.9592, "step": 7160 }, { "epoch": 0.61, "learning_rate": 7.081981586050289e-06, "loss": 1.0357, "step": 7161 }, { "epoch": 0.61, "learning_rate": 7.079358045503495e-06, "loss": 1.0127, "step": 7162 }, { "epoch": 0.61, "learning_rate": 7.076734724712097e-06, "loss": 1.0441, "step": 7163 }, { "epoch": 0.61, "learning_rate": 7.074111623873471e-06, "loss": 1.052, "step": 7164 }, { "epoch": 0.61, "learning_rate": 7.071488743184993e-06, "loss": 1.0438, "step": 7165 }, { "epoch": 0.61, "learning_rate": 7.068866082844008e-06, "loss": 1.0223, "step": 7166 }, { "epoch": 0.61, "learning_rate": 7.066243643047857e-06, "loss": 1.0518, "step": 7167 }, { "epoch": 0.61, "learning_rate": 7.063621423993852e-06, "loss": 1.0048, "step": 7168 }, { "epoch": 0.61, "learning_rate": 7.060999425879297e-06, "loss": 1.0657, "step": 7169 }, { "epoch": 0.61, "learning_rate": 7.0583776489014754e-06, "loss": 1.0632, "step": 7170 }, { "epoch": 0.61, "learning_rate": 7.05575609325766e-06, "loss": 1.0261, "step": 7171 }, { "epoch": 0.61, "learning_rate": 7.053134759145094e-06, "loss": 0.9829, "step": 7172 }, { "epoch": 0.61, "learning_rate": 7.050513646761021e-06, "loss": 1.0548, "step": 7173 }, { "epoch": 0.61, "learning_rate": 7.047892756302651e-06, "loss": 0.9755, "step": 7174 }, { "epoch": 0.61, "learning_rate": 7.045272087967194e-06, "loss": 1.0427, "step": 7175 }, { "epoch": 0.61, "learning_rate": 7.042651641951825e-06, "loss": 1.0121, "step": 7176 }, { "epoch": 0.61, "learning_rate": 7.040031418453723e-06, "loss": 1.0412, "step": 7177 }, { "epoch": 0.61, "learning_rate": 7.037411417670029e-06, "loss": 0.9647, "step": 7178 }, { "epoch": 0.61, "learning_rate": 7.034791639797886e-06, "loss": 1.0024, "step": 7179 }, { "epoch": 0.61, "learning_rate": 7.0321720850344036e-06, "loss": 0.9693, "step": 7180 }, { "epoch": 0.61, "learning_rate": 7.029552753576692e-06, "loss": 1.0212, "step": 7181 }, { "epoch": 0.61, "learning_rate": 7.026933645621825e-06, "loss": 1.0459, "step": 7182 }, { "epoch": 0.61, "learning_rate": 7.024314761366879e-06, "loss": 0.9995, "step": 7183 }, { "epoch": 0.61, "learning_rate": 7.021696101008899e-06, "loss": 0.8402, "step": 7184 }, { "epoch": 0.61, "learning_rate": 7.019077664744921e-06, "loss": 1.0236, "step": 7185 }, { "epoch": 0.61, "learning_rate": 7.01645945277196e-06, "loss": 1.0778, "step": 7186 }, { "epoch": 0.61, "learning_rate": 7.013841465287017e-06, "loss": 1.0277, "step": 7187 }, { "epoch": 0.61, "learning_rate": 7.0112237024870775e-06, "loss": 1.0616, "step": 7188 }, { "epoch": 0.61, "learning_rate": 7.0086061645691e-06, "loss": 0.9981, "step": 7189 }, { "epoch": 0.61, "learning_rate": 7.005988851730046e-06, "loss": 0.9453, "step": 7190 }, { "epoch": 0.61, "learning_rate": 7.003371764166834e-06, "loss": 1.0453, "step": 7191 }, { "epoch": 0.61, "learning_rate": 7.0007549020763906e-06, "loss": 0.9355, "step": 7192 }, { "epoch": 0.61, "learning_rate": 6.998138265655605e-06, "loss": 0.9977, "step": 7193 }, { "epoch": 0.61, "learning_rate": 6.995521855101368e-06, "loss": 0.9671, "step": 7194 }, { "epoch": 0.61, "learning_rate": 6.992905670610537e-06, "loss": 0.9435, "step": 7195 }, { "epoch": 0.61, "learning_rate": 6.990289712379961e-06, "loss": 1.0271, "step": 7196 }, { "epoch": 0.61, "learning_rate": 6.987673980606471e-06, "loss": 1.0186, "step": 7197 }, { "epoch": 0.61, "learning_rate": 6.985058475486881e-06, "loss": 1.0074, "step": 7198 }, { "epoch": 0.61, "learning_rate": 6.982443197217983e-06, "loss": 1.0289, "step": 7199 }, { "epoch": 0.61, "learning_rate": 6.9798281459965635e-06, "loss": 1.0142, "step": 7200 }, { "epoch": 0.61, "learning_rate": 6.9772133220193764e-06, "loss": 1.0551, "step": 7201 }, { "epoch": 0.61, "learning_rate": 6.974598725483175e-06, "loss": 0.9769, "step": 7202 }, { "epoch": 0.61, "learning_rate": 6.971984356584678e-06, "loss": 0.9988, "step": 7203 }, { "epoch": 0.61, "learning_rate": 6.969370215520606e-06, "loss": 1.0355, "step": 7204 }, { "epoch": 0.61, "learning_rate": 6.966756302487643e-06, "loss": 1.0281, "step": 7205 }, { "epoch": 0.61, "learning_rate": 6.964142617682475e-06, "loss": 1.0, "step": 7206 }, { "epoch": 0.61, "learning_rate": 6.96152916130175e-06, "loss": 0.9782, "step": 7207 }, { "epoch": 0.61, "learning_rate": 6.958915933542123e-06, "loss": 0.9976, "step": 7208 }, { "epoch": 0.61, "learning_rate": 6.956302934600207e-06, "loss": 0.9818, "step": 7209 }, { "epoch": 0.61, "learning_rate": 6.95369016467262e-06, "loss": 1.0046, "step": 7210 }, { "epoch": 0.61, "learning_rate": 6.951077623955946e-06, "loss": 1.0519, "step": 7211 }, { "epoch": 0.61, "learning_rate": 6.948465312646756e-06, "loss": 1.0116, "step": 7212 }, { "epoch": 0.61, "learning_rate": 6.945853230941615e-06, "loss": 1.0462, "step": 7213 }, { "epoch": 0.61, "learning_rate": 6.943241379037051e-06, "loss": 0.9913, "step": 7214 }, { "epoch": 0.61, "learning_rate": 6.9406297571295955e-06, "loss": 0.9859, "step": 7215 }, { "epoch": 0.61, "learning_rate": 6.938018365415744e-06, "loss": 1.0072, "step": 7216 }, { "epoch": 0.61, "learning_rate": 6.93540720409199e-06, "loss": 1.0329, "step": 7217 }, { "epoch": 0.61, "learning_rate": 6.932796273354796e-06, "loss": 0.9586, "step": 7218 }, { "epoch": 0.61, "learning_rate": 6.930185573400623e-06, "loss": 1.0248, "step": 7219 }, { "epoch": 0.61, "learning_rate": 6.927575104425896e-06, "loss": 1.0181, "step": 7220 }, { "epoch": 0.61, "learning_rate": 6.92496486662704e-06, "loss": 0.9938, "step": 7221 }, { "epoch": 0.61, "learning_rate": 6.9223548602004505e-06, "loss": 0.9542, "step": 7222 }, { "epoch": 0.61, "learning_rate": 6.9197450853425104e-06, "loss": 0.9966, "step": 7223 }, { "epoch": 0.61, "learning_rate": 6.9171355422495865e-06, "loss": 0.9845, "step": 7224 }, { "epoch": 0.61, "learning_rate": 6.9145262311180285e-06, "loss": 1.004, "step": 7225 }, { "epoch": 0.61, "learning_rate": 6.9119171521441585e-06, "loss": 1.0739, "step": 7226 }, { "epoch": 0.61, "learning_rate": 6.909308305524299e-06, "loss": 1.0182, "step": 7227 }, { "epoch": 0.61, "learning_rate": 6.9066996914547365e-06, "loss": 1.0224, "step": 7228 }, { "epoch": 0.61, "learning_rate": 6.904091310131757e-06, "loss": 0.9951, "step": 7229 }, { "epoch": 0.61, "learning_rate": 6.901483161751612e-06, "loss": 1.0102, "step": 7230 }, { "epoch": 0.61, "learning_rate": 6.8988752465105545e-06, "loss": 1.0197, "step": 7231 }, { "epoch": 0.61, "learning_rate": 6.8962675646047985e-06, "loss": 1.019, "step": 7232 }, { "epoch": 0.61, "learning_rate": 6.893660116230562e-06, "loss": 1.0121, "step": 7233 }, { "epoch": 0.61, "learning_rate": 6.891052901584025e-06, "loss": 1.0198, "step": 7234 }, { "epoch": 0.61, "learning_rate": 6.8884459208613695e-06, "loss": 0.9716, "step": 7235 }, { "epoch": 0.61, "learning_rate": 6.885839174258742e-06, "loss": 1.0138, "step": 7236 }, { "epoch": 0.61, "learning_rate": 6.883232661972288e-06, "loss": 1.0225, "step": 7237 }, { "epoch": 0.61, "learning_rate": 6.880626384198122e-06, "loss": 0.9929, "step": 7238 }, { "epoch": 0.61, "learning_rate": 6.878020341132342e-06, "loss": 1.0453, "step": 7239 }, { "epoch": 0.61, "learning_rate": 6.875414532971043e-06, "loss": 1.0345, "step": 7240 }, { "epoch": 0.61, "learning_rate": 6.8728089599102805e-06, "loss": 1.0019, "step": 7241 }, { "epoch": 0.61, "learning_rate": 6.870203622146113e-06, "loss": 1.0084, "step": 7242 }, { "epoch": 0.61, "learning_rate": 6.8675985198745636e-06, "loss": 1.0418, "step": 7243 }, { "epoch": 0.61, "learning_rate": 6.864993653291654e-06, "loss": 0.9817, "step": 7244 }, { "epoch": 0.61, "learning_rate": 6.86238902259337e-06, "loss": 0.9691, "step": 7245 }, { "epoch": 0.61, "learning_rate": 6.859784627975701e-06, "loss": 0.9851, "step": 7246 }, { "epoch": 0.61, "learning_rate": 6.857180469634596e-06, "loss": 1.0244, "step": 7247 }, { "epoch": 0.61, "learning_rate": 6.854576547766007e-06, "loss": 0.8414, "step": 7248 }, { "epoch": 0.61, "learning_rate": 6.8519728625658525e-06, "loss": 1.0363, "step": 7249 }, { "epoch": 0.61, "learning_rate": 6.849369414230042e-06, "loss": 0.9735, "step": 7250 }, { "epoch": 0.61, "learning_rate": 6.846766202954462e-06, "loss": 0.9897, "step": 7251 }, { "epoch": 0.61, "learning_rate": 6.84416322893499e-06, "loss": 1.017, "step": 7252 }, { "epoch": 0.61, "learning_rate": 6.841560492367471e-06, "loss": 1.0295, "step": 7253 }, { "epoch": 0.61, "learning_rate": 6.838957993447748e-06, "loss": 0.9778, "step": 7254 }, { "epoch": 0.61, "learning_rate": 6.836355732371631e-06, "loss": 1.0482, "step": 7255 }, { "epoch": 0.61, "learning_rate": 6.833753709334927e-06, "loss": 0.9737, "step": 7256 }, { "epoch": 0.61, "learning_rate": 6.831151924533413e-06, "loss": 1.0333, "step": 7257 }, { "epoch": 0.61, "learning_rate": 6.828550378162857e-06, "loss": 0.9273, "step": 7258 }, { "epoch": 0.61, "learning_rate": 6.825949070418999e-06, "loss": 1.0392, "step": 7259 }, { "epoch": 0.61, "learning_rate": 6.823348001497574e-06, "loss": 0.9866, "step": 7260 }, { "epoch": 0.61, "learning_rate": 6.820747171594287e-06, "loss": 1.0474, "step": 7261 }, { "epoch": 0.62, "learning_rate": 6.818146580904833e-06, "loss": 1.0528, "step": 7262 }, { "epoch": 0.62, "learning_rate": 6.815546229624883e-06, "loss": 0.9846, "step": 7263 }, { "epoch": 0.62, "learning_rate": 6.812946117950095e-06, "loss": 1.0175, "step": 7264 }, { "epoch": 0.62, "learning_rate": 6.8103462460761085e-06, "loss": 0.9832, "step": 7265 }, { "epoch": 0.62, "learning_rate": 6.8077466141985385e-06, "loss": 1.0347, "step": 7266 }, { "epoch": 0.62, "learning_rate": 6.8051472225129935e-06, "loss": 0.9544, "step": 7267 }, { "epoch": 0.62, "learning_rate": 6.802548071215052e-06, "loss": 0.8203, "step": 7268 }, { "epoch": 0.62, "learning_rate": 6.799949160500283e-06, "loss": 0.9984, "step": 7269 }, { "epoch": 0.62, "learning_rate": 6.79735049056423e-06, "loss": 0.9328, "step": 7270 }, { "epoch": 0.62, "learning_rate": 6.7947520616024295e-06, "loss": 1.0549, "step": 7271 }, { "epoch": 0.62, "learning_rate": 6.792153873810385e-06, "loss": 0.9742, "step": 7272 }, { "epoch": 0.62, "learning_rate": 6.789555927383598e-06, "loss": 0.9953, "step": 7273 }, { "epoch": 0.62, "learning_rate": 6.786958222517533e-06, "loss": 0.9826, "step": 7274 }, { "epoch": 0.62, "learning_rate": 6.784360759407658e-06, "loss": 0.9705, "step": 7275 }, { "epoch": 0.62, "learning_rate": 6.781763538249406e-06, "loss": 1.0051, "step": 7276 }, { "epoch": 0.62, "learning_rate": 6.779166559238197e-06, "loss": 0.9063, "step": 7277 }, { "epoch": 0.62, "learning_rate": 6.7765698225694345e-06, "loss": 1.0449, "step": 7278 }, { "epoch": 0.62, "learning_rate": 6.773973328438504e-06, "loss": 0.9411, "step": 7279 }, { "epoch": 0.62, "learning_rate": 6.7713770770407675e-06, "loss": 0.9961, "step": 7280 }, { "epoch": 0.62, "learning_rate": 6.7687810685715785e-06, "loss": 0.9948, "step": 7281 }, { "epoch": 0.62, "learning_rate": 6.766185303226258e-06, "loss": 1.0193, "step": 7282 }, { "epoch": 0.62, "learning_rate": 6.763589781200126e-06, "loss": 1.006, "step": 7283 }, { "epoch": 0.62, "learning_rate": 6.7609945026884674e-06, "loss": 0.942, "step": 7284 }, { "epoch": 0.62, "learning_rate": 6.7583994678865625e-06, "loss": 0.977, "step": 7285 }, { "epoch": 0.62, "learning_rate": 6.7558046769896625e-06, "loss": 0.9752, "step": 7286 }, { "epoch": 0.62, "learning_rate": 6.753210130193011e-06, "loss": 0.9637, "step": 7287 }, { "epoch": 0.62, "learning_rate": 6.7506158276918185e-06, "loss": 0.9891, "step": 7288 }, { "epoch": 0.62, "learning_rate": 6.748021769681297e-06, "loss": 0.9659, "step": 7289 }, { "epoch": 0.62, "learning_rate": 6.7454279563566195e-06, "loss": 1.0018, "step": 7290 }, { "epoch": 0.62, "learning_rate": 6.742834387912952e-06, "loss": 0.9978, "step": 7291 }, { "epoch": 0.62, "learning_rate": 6.740241064545446e-06, "loss": 0.9962, "step": 7292 }, { "epoch": 0.62, "learning_rate": 6.73764798644922e-06, "loss": 1.0208, "step": 7293 }, { "epoch": 0.62, "learning_rate": 6.735055153819393e-06, "loss": 1.0282, "step": 7294 }, { "epoch": 0.62, "learning_rate": 6.732462566851045e-06, "loss": 0.9377, "step": 7295 }, { "epoch": 0.62, "learning_rate": 6.729870225739257e-06, "loss": 0.9709, "step": 7296 }, { "epoch": 0.62, "learning_rate": 6.727278130679072e-06, "loss": 0.983, "step": 7297 }, { "epoch": 0.62, "learning_rate": 6.724686281865535e-06, "loss": 1.0303, "step": 7298 }, { "epoch": 0.62, "learning_rate": 6.722094679493656e-06, "loss": 1.0015, "step": 7299 }, { "epoch": 0.62, "learning_rate": 6.719503323758438e-06, "loss": 1.0485, "step": 7300 }, { "epoch": 0.62, "learning_rate": 6.7169122148548515e-06, "loss": 0.9843, "step": 7301 }, { "epoch": 0.62, "learning_rate": 6.714321352977868e-06, "loss": 1.0309, "step": 7302 }, { "epoch": 0.62, "learning_rate": 6.711730738322421e-06, "loss": 0.9831, "step": 7303 }, { "epoch": 0.62, "learning_rate": 6.709140371083438e-06, "loss": 1.0164, "step": 7304 }, { "epoch": 0.62, "learning_rate": 6.706550251455821e-06, "loss": 0.9817, "step": 7305 }, { "epoch": 0.62, "learning_rate": 6.7039603796344625e-06, "loss": 0.977, "step": 7306 }, { "epoch": 0.62, "learning_rate": 6.7013707558142195e-06, "loss": 1.0242, "step": 7307 }, { "epoch": 0.62, "learning_rate": 6.698781380189951e-06, "loss": 1.0068, "step": 7308 }, { "epoch": 0.62, "learning_rate": 6.6961922529564796e-06, "loss": 0.9589, "step": 7309 }, { "epoch": 0.62, "learning_rate": 6.693603374308624e-06, "loss": 1.0325, "step": 7310 }, { "epoch": 0.62, "learning_rate": 6.69101474444117e-06, "loss": 1.0194, "step": 7311 }, { "epoch": 0.62, "learning_rate": 6.688426363548899e-06, "loss": 1.0035, "step": 7312 }, { "epoch": 0.62, "learning_rate": 6.685838231826558e-06, "loss": 0.9698, "step": 7313 }, { "epoch": 0.62, "learning_rate": 6.6832503494688906e-06, "loss": 1.0167, "step": 7314 }, { "epoch": 0.62, "learning_rate": 6.68066271667061e-06, "loss": 0.9879, "step": 7315 }, { "epoch": 0.62, "learning_rate": 6.678075333626418e-06, "loss": 0.9675, "step": 7316 }, { "epoch": 0.62, "learning_rate": 6.675488200530998e-06, "loss": 0.9783, "step": 7317 }, { "epoch": 0.62, "learning_rate": 6.672901317579e-06, "loss": 0.9909, "step": 7318 }, { "epoch": 0.62, "learning_rate": 6.67031468496508e-06, "loss": 0.9309, "step": 7319 }, { "epoch": 0.62, "learning_rate": 6.6677283028838536e-06, "loss": 1.0168, "step": 7320 }, { "epoch": 0.62, "learning_rate": 6.665142171529931e-06, "loss": 1.0506, "step": 7321 }, { "epoch": 0.62, "learning_rate": 6.66255629109789e-06, "loss": 1.0576, "step": 7322 }, { "epoch": 0.62, "learning_rate": 6.659970661782309e-06, "loss": 0.9896, "step": 7323 }, { "epoch": 0.62, "learning_rate": 6.657385283777726e-06, "loss": 1.1062, "step": 7324 }, { "epoch": 0.62, "learning_rate": 6.6548001572786805e-06, "loss": 0.9959, "step": 7325 }, { "epoch": 0.62, "learning_rate": 6.652215282479673e-06, "loss": 0.9932, "step": 7326 }, { "epoch": 0.62, "learning_rate": 6.649630659575202e-06, "loss": 1.0483, "step": 7327 }, { "epoch": 0.62, "learning_rate": 6.6470462887597354e-06, "loss": 1.0008, "step": 7328 }, { "epoch": 0.62, "learning_rate": 6.644462170227733e-06, "loss": 0.9941, "step": 7329 }, { "epoch": 0.62, "learning_rate": 6.641878304173621e-06, "loss": 0.9992, "step": 7330 }, { "epoch": 0.62, "learning_rate": 6.639294690791822e-06, "loss": 0.9979, "step": 7331 }, { "epoch": 0.62, "learning_rate": 6.6367113302767305e-06, "loss": 0.871, "step": 7332 }, { "epoch": 0.62, "learning_rate": 6.634128222822725e-06, "loss": 1.0442, "step": 7333 }, { "epoch": 0.62, "learning_rate": 6.6315453686241584e-06, "loss": 1.0065, "step": 7334 }, { "epoch": 0.62, "learning_rate": 6.6289627678753794e-06, "loss": 1.0182, "step": 7335 }, { "epoch": 0.62, "learning_rate": 6.6263804207706995e-06, "loss": 0.9772, "step": 7336 }, { "epoch": 0.62, "learning_rate": 6.623798327504428e-06, "loss": 0.9893, "step": 7337 }, { "epoch": 0.62, "learning_rate": 6.62121648827084e-06, "loss": 0.8212, "step": 7338 }, { "epoch": 0.62, "learning_rate": 6.618634903264206e-06, "loss": 1.0309, "step": 7339 }, { "epoch": 0.62, "learning_rate": 6.616053572678762e-06, "loss": 1.0396, "step": 7340 }, { "epoch": 0.62, "learning_rate": 6.613472496708742e-06, "loss": 1.0204, "step": 7341 }, { "epoch": 0.62, "learning_rate": 6.610891675548341e-06, "loss": 0.919, "step": 7342 }, { "epoch": 0.62, "learning_rate": 6.608311109391754e-06, "loss": 0.9784, "step": 7343 }, { "epoch": 0.62, "learning_rate": 6.605730798433148e-06, "loss": 1.0119, "step": 7344 }, { "epoch": 0.62, "learning_rate": 6.6031507428666655e-06, "loss": 1.0159, "step": 7345 }, { "epoch": 0.62, "learning_rate": 6.600570942886442e-06, "loss": 0.9565, "step": 7346 }, { "epoch": 0.62, "learning_rate": 6.597991398686579e-06, "loss": 0.9378, "step": 7347 }, { "epoch": 0.62, "learning_rate": 6.595412110461177e-06, "loss": 1.0148, "step": 7348 }, { "epoch": 0.62, "learning_rate": 6.592833078404298e-06, "loss": 1.0641, "step": 7349 }, { "epoch": 0.62, "learning_rate": 6.5902543027100015e-06, "loss": 0.9953, "step": 7350 }, { "epoch": 0.62, "learning_rate": 6.5876757835723135e-06, "loss": 0.9748, "step": 7351 }, { "epoch": 0.62, "learning_rate": 6.585097521185254e-06, "loss": 0.9979, "step": 7352 }, { "epoch": 0.62, "learning_rate": 6.582519515742809e-06, "loss": 1.0185, "step": 7353 }, { "epoch": 0.62, "learning_rate": 6.579941767438962e-06, "loss": 1.0079, "step": 7354 }, { "epoch": 0.62, "learning_rate": 6.577364276467663e-06, "loss": 0.9666, "step": 7355 }, { "epoch": 0.62, "learning_rate": 6.574787043022848e-06, "loss": 1.0575, "step": 7356 }, { "epoch": 0.62, "learning_rate": 6.572210067298435e-06, "loss": 1.0298, "step": 7357 }, { "epoch": 0.62, "learning_rate": 6.56963334948832e-06, "loss": 1.0464, "step": 7358 }, { "epoch": 0.62, "learning_rate": 6.567056889786383e-06, "loss": 0.9772, "step": 7359 }, { "epoch": 0.62, "learning_rate": 6.564480688386483e-06, "loss": 1.0493, "step": 7360 }, { "epoch": 0.62, "learning_rate": 6.5619047454824526e-06, "loss": 1.0575, "step": 7361 }, { "epoch": 0.62, "learning_rate": 6.5593290612681184e-06, "loss": 1.0014, "step": 7362 }, { "epoch": 0.62, "learning_rate": 6.556753635937274e-06, "loss": 1.0249, "step": 7363 }, { "epoch": 0.62, "learning_rate": 6.554178469683709e-06, "loss": 0.9617, "step": 7364 }, { "epoch": 0.62, "learning_rate": 6.551603562701174e-06, "loss": 0.9467, "step": 7365 }, { "epoch": 0.62, "learning_rate": 6.549028915183419e-06, "loss": 0.9525, "step": 7366 }, { "epoch": 0.62, "learning_rate": 6.54645452732416e-06, "loss": 0.9637, "step": 7367 }, { "epoch": 0.62, "learning_rate": 6.5438803993171044e-06, "loss": 1.0342, "step": 7368 }, { "epoch": 0.62, "learning_rate": 6.541306531355931e-06, "loss": 1.0423, "step": 7369 }, { "epoch": 0.62, "learning_rate": 6.538732923634306e-06, "loss": 0.9692, "step": 7370 }, { "epoch": 0.62, "learning_rate": 6.536159576345875e-06, "loss": 0.9432, "step": 7371 }, { "epoch": 0.62, "learning_rate": 6.5335864896842536e-06, "loss": 0.9788, "step": 7372 }, { "epoch": 0.62, "learning_rate": 6.531013663843058e-06, "loss": 1.0401, "step": 7373 }, { "epoch": 0.62, "learning_rate": 6.528441099015863e-06, "loss": 1.0041, "step": 7374 }, { "epoch": 0.62, "learning_rate": 6.525868795396242e-06, "loss": 0.9909, "step": 7375 }, { "epoch": 0.62, "learning_rate": 6.523296753177732e-06, "loss": 0.9647, "step": 7376 }, { "epoch": 0.62, "learning_rate": 6.520724972553869e-06, "loss": 0.884, "step": 7377 }, { "epoch": 0.62, "learning_rate": 6.51815345371815e-06, "loss": 0.9723, "step": 7378 }, { "epoch": 0.62, "learning_rate": 6.5155821968640695e-06, "loss": 0.9748, "step": 7379 }, { "epoch": 0.62, "learning_rate": 6.513011202185088e-06, "loss": 1.0677, "step": 7380 }, { "epoch": 0.63, "learning_rate": 6.5104404698746595e-06, "loss": 0.9959, "step": 7381 }, { "epoch": 0.63, "learning_rate": 6.507870000126204e-06, "loss": 1.0399, "step": 7382 }, { "epoch": 0.63, "learning_rate": 6.505299793133133e-06, "loss": 1.0259, "step": 7383 }, { "epoch": 0.63, "learning_rate": 6.502729849088833e-06, "loss": 0.9705, "step": 7384 }, { "epoch": 0.63, "learning_rate": 6.500160168186678e-06, "loss": 0.9964, "step": 7385 }, { "epoch": 0.63, "learning_rate": 6.497590750620004e-06, "loss": 1.0076, "step": 7386 }, { "epoch": 0.63, "learning_rate": 6.4950215965821535e-06, "loss": 1.0262, "step": 7387 }, { "epoch": 0.63, "learning_rate": 6.4924527062664235e-06, "loss": 1.0413, "step": 7388 }, { "epoch": 0.63, "learning_rate": 6.489884079866112e-06, "loss": 0.8464, "step": 7389 }, { "epoch": 0.63, "learning_rate": 6.4873157175744805e-06, "loss": 0.9675, "step": 7390 }, { "epoch": 0.63, "learning_rate": 6.484747619584786e-06, "loss": 1.039, "step": 7391 }, { "epoch": 0.63, "learning_rate": 6.482179786090248e-06, "loss": 0.9724, "step": 7392 }, { "epoch": 0.63, "learning_rate": 6.479612217284086e-06, "loss": 1.0424, "step": 7393 }, { "epoch": 0.63, "learning_rate": 6.47704491335948e-06, "loss": 1.0258, "step": 7394 }, { "epoch": 0.63, "learning_rate": 6.474477874509606e-06, "loss": 0.9989, "step": 7395 }, { "epoch": 0.63, "learning_rate": 6.471911100927611e-06, "loss": 1.007, "step": 7396 }, { "epoch": 0.63, "learning_rate": 6.469344592806625e-06, "loss": 1.0298, "step": 7397 }, { "epoch": 0.63, "learning_rate": 6.466778350339762e-06, "loss": 1.0085, "step": 7398 }, { "epoch": 0.63, "learning_rate": 6.4642123737201014e-06, "loss": 1.0067, "step": 7399 }, { "epoch": 0.63, "learning_rate": 6.461646663140724e-06, "loss": 1.0533, "step": 7400 }, { "epoch": 0.63, "learning_rate": 6.4590812187946685e-06, "loss": 0.964, "step": 7401 }, { "epoch": 0.63, "learning_rate": 6.456516040874975e-06, "loss": 1.0043, "step": 7402 }, { "epoch": 0.63, "learning_rate": 6.453951129574644e-06, "loss": 1.026, "step": 7403 }, { "epoch": 0.63, "learning_rate": 6.451386485086672e-06, "loss": 0.9472, "step": 7404 }, { "epoch": 0.63, "learning_rate": 6.448822107604022e-06, "loss": 1.0009, "step": 7405 }, { "epoch": 0.63, "learning_rate": 6.446257997319651e-06, "loss": 1.0011, "step": 7406 }, { "epoch": 0.63, "learning_rate": 6.443694154426481e-06, "loss": 1.0157, "step": 7407 }, { "epoch": 0.63, "learning_rate": 6.441130579117427e-06, "loss": 0.8646, "step": 7408 }, { "epoch": 0.63, "learning_rate": 6.4385672715853735e-06, "loss": 1.0145, "step": 7409 }, { "epoch": 0.63, "learning_rate": 6.43600423202319e-06, "loss": 0.9257, "step": 7410 }, { "epoch": 0.63, "learning_rate": 6.433441460623726e-06, "loss": 0.989, "step": 7411 }, { "epoch": 0.63, "learning_rate": 6.430878957579811e-06, "loss": 1.0042, "step": 7412 }, { "epoch": 0.63, "learning_rate": 6.428316723084249e-06, "loss": 1.0259, "step": 7413 }, { "epoch": 0.63, "learning_rate": 6.425754757329834e-06, "loss": 1.0385, "step": 7414 }, { "epoch": 0.63, "learning_rate": 6.423193060509327e-06, "loss": 1.0123, "step": 7415 }, { "epoch": 0.63, "learning_rate": 6.420631632815484e-06, "loss": 1.0203, "step": 7416 }, { "epoch": 0.63, "learning_rate": 6.418070474441024e-06, "loss": 1.01, "step": 7417 }, { "epoch": 0.63, "learning_rate": 6.415509585578661e-06, "loss": 1.0303, "step": 7418 }, { "epoch": 0.63, "learning_rate": 6.412948966421075e-06, "loss": 1.0104, "step": 7419 }, { "epoch": 0.63, "learning_rate": 6.4103886171609364e-06, "loss": 1.0444, "step": 7420 }, { "epoch": 0.63, "learning_rate": 6.407828537990894e-06, "loss": 1.0277, "step": 7421 }, { "epoch": 0.63, "learning_rate": 6.40526872910357e-06, "loss": 0.9626, "step": 7422 }, { "epoch": 0.63, "learning_rate": 6.4027091906915704e-06, "loss": 1.0047, "step": 7423 }, { "epoch": 0.63, "learning_rate": 6.4001499229474804e-06, "loss": 1.0256, "step": 7424 }, { "epoch": 0.63, "learning_rate": 6.397590926063868e-06, "loss": 0.9866, "step": 7425 }, { "epoch": 0.63, "learning_rate": 6.39503220023327e-06, "loss": 0.9532, "step": 7426 }, { "epoch": 0.63, "learning_rate": 6.39247374564822e-06, "loss": 1.041, "step": 7427 }, { "epoch": 0.63, "learning_rate": 6.389915562501212e-06, "loss": 1.0451, "step": 7428 }, { "epoch": 0.63, "learning_rate": 6.387357650984738e-06, "loss": 0.9994, "step": 7429 }, { "epoch": 0.63, "learning_rate": 6.384800011291253e-06, "loss": 0.9606, "step": 7430 }, { "epoch": 0.63, "learning_rate": 6.382242643613208e-06, "loss": 1.0039, "step": 7431 }, { "epoch": 0.63, "learning_rate": 6.3796855481430145e-06, "loss": 0.9752, "step": 7432 }, { "epoch": 0.63, "learning_rate": 6.377128725073084e-06, "loss": 0.9943, "step": 7433 }, { "epoch": 0.63, "learning_rate": 6.3745721745957886e-06, "loss": 0.824, "step": 7434 }, { "epoch": 0.63, "learning_rate": 6.372015896903497e-06, "loss": 1.0135, "step": 7435 }, { "epoch": 0.63, "learning_rate": 6.369459892188543e-06, "loss": 0.9598, "step": 7436 }, { "epoch": 0.63, "learning_rate": 6.366904160643248e-06, "loss": 1.0517, "step": 7437 }, { "epoch": 0.63, "learning_rate": 6.36434870245991e-06, "loss": 0.9931, "step": 7438 }, { "epoch": 0.63, "learning_rate": 6.36179351783081e-06, "loss": 1.0369, "step": 7439 }, { "epoch": 0.63, "learning_rate": 6.359238606948199e-06, "loss": 1.0398, "step": 7440 }, { "epoch": 0.63, "learning_rate": 6.356683970004323e-06, "loss": 0.9819, "step": 7441 }, { "epoch": 0.63, "learning_rate": 6.354129607191389e-06, "loss": 1.0251, "step": 7442 }, { "epoch": 0.63, "learning_rate": 6.351575518701602e-06, "loss": 0.8338, "step": 7443 }, { "epoch": 0.63, "learning_rate": 6.3490217047271275e-06, "loss": 1.0414, "step": 7444 }, { "epoch": 0.63, "learning_rate": 6.346468165460127e-06, "loss": 1.0209, "step": 7445 }, { "epoch": 0.63, "learning_rate": 6.343914901092733e-06, "loss": 1.0139, "step": 7446 }, { "epoch": 0.63, "learning_rate": 6.341361911817057e-06, "loss": 1.0311, "step": 7447 }, { "epoch": 0.63, "learning_rate": 6.338809197825195e-06, "loss": 1.0088, "step": 7448 }, { "epoch": 0.63, "learning_rate": 6.336256759309214e-06, "loss": 0.9802, "step": 7449 }, { "epoch": 0.63, "learning_rate": 6.333704596461167e-06, "loss": 1.0307, "step": 7450 }, { "epoch": 0.63, "learning_rate": 6.331152709473085e-06, "loss": 0.9852, "step": 7451 }, { "epoch": 0.63, "learning_rate": 6.328601098536979e-06, "loss": 0.9998, "step": 7452 }, { "epoch": 0.63, "learning_rate": 6.32604976384483e-06, "loss": 1.0129, "step": 7453 }, { "epoch": 0.63, "learning_rate": 6.3234987055886154e-06, "loss": 1.0074, "step": 7454 }, { "epoch": 0.63, "learning_rate": 6.3209479239602755e-06, "loss": 0.9453, "step": 7455 }, { "epoch": 0.63, "learning_rate": 6.318397419151743e-06, "loss": 0.9541, "step": 7456 }, { "epoch": 0.63, "learning_rate": 6.315847191354916e-06, "loss": 1.022, "step": 7457 }, { "epoch": 0.63, "learning_rate": 6.313297240761685e-06, "loss": 1.0088, "step": 7458 }, { "epoch": 0.63, "learning_rate": 6.310747567563909e-06, "loss": 1.0325, "step": 7459 }, { "epoch": 0.63, "learning_rate": 6.308198171953438e-06, "loss": 0.9796, "step": 7460 }, { "epoch": 0.63, "learning_rate": 6.305649054122085e-06, "loss": 1.0053, "step": 7461 }, { "epoch": 0.63, "learning_rate": 6.3031002142616595e-06, "loss": 0.7935, "step": 7462 }, { "epoch": 0.63, "learning_rate": 6.300551652563936e-06, "loss": 1.0165, "step": 7463 }, { "epoch": 0.63, "learning_rate": 6.298003369220675e-06, "loss": 1.0114, "step": 7464 }, { "epoch": 0.63, "learning_rate": 6.295455364423617e-06, "loss": 1.0353, "step": 7465 }, { "epoch": 0.63, "learning_rate": 6.2929076383644794e-06, "loss": 0.9706, "step": 7466 }, { "epoch": 0.63, "learning_rate": 6.290360191234952e-06, "loss": 0.9461, "step": 7467 }, { "epoch": 0.63, "learning_rate": 6.287813023226721e-06, "loss": 0.9993, "step": 7468 }, { "epoch": 0.63, "learning_rate": 6.285266134531429e-06, "loss": 1.0147, "step": 7469 }, { "epoch": 0.63, "learning_rate": 6.282719525340721e-06, "loss": 0.9767, "step": 7470 }, { "epoch": 0.63, "learning_rate": 6.2801731958462e-06, "loss": 1.0384, "step": 7471 }, { "epoch": 0.63, "learning_rate": 6.27762714623946e-06, "loss": 0.9892, "step": 7472 }, { "epoch": 0.63, "learning_rate": 6.2750813767120765e-06, "loss": 0.9645, "step": 7473 }, { "epoch": 0.63, "learning_rate": 6.272535887455593e-06, "loss": 1.009, "step": 7474 }, { "epoch": 0.63, "learning_rate": 6.269990678661539e-06, "loss": 0.8468, "step": 7475 }, { "epoch": 0.63, "learning_rate": 6.267445750521421e-06, "loss": 0.9984, "step": 7476 }, { "epoch": 0.63, "learning_rate": 6.264901103226727e-06, "loss": 0.9921, "step": 7477 }, { "epoch": 0.63, "learning_rate": 6.2623567369689194e-06, "loss": 0.9862, "step": 7478 }, { "epoch": 0.63, "learning_rate": 6.259812651939445e-06, "loss": 1.0442, "step": 7479 }, { "epoch": 0.63, "learning_rate": 6.257268848329721e-06, "loss": 1.0189, "step": 7480 }, { "epoch": 0.63, "learning_rate": 6.254725326331154e-06, "loss": 0.9931, "step": 7481 }, { "epoch": 0.63, "learning_rate": 6.2521820861351175e-06, "loss": 1.0191, "step": 7482 }, { "epoch": 0.63, "learning_rate": 6.249639127932979e-06, "loss": 0.8188, "step": 7483 }, { "epoch": 0.63, "learning_rate": 6.2470964519160685e-06, "loss": 1.0524, "step": 7484 }, { "epoch": 0.63, "learning_rate": 6.2445540582757095e-06, "loss": 1.0482, "step": 7485 }, { "epoch": 0.63, "learning_rate": 6.242011947203188e-06, "loss": 1.0267, "step": 7486 }, { "epoch": 0.63, "learning_rate": 6.239470118889787e-06, "loss": 0.9725, "step": 7487 }, { "epoch": 0.63, "learning_rate": 6.236928573526753e-06, "loss": 0.9934, "step": 7488 }, { "epoch": 0.63, "learning_rate": 6.234387311305323e-06, "loss": 0.9828, "step": 7489 }, { "epoch": 0.63, "learning_rate": 6.231846332416701e-06, "loss": 1.0301, "step": 7490 }, { "epoch": 0.63, "learning_rate": 6.229305637052079e-06, "loss": 0.9984, "step": 7491 }, { "epoch": 0.63, "learning_rate": 6.226765225402624e-06, "loss": 1.0289, "step": 7492 }, { "epoch": 0.63, "learning_rate": 6.224225097659484e-06, "loss": 0.9977, "step": 7493 }, { "epoch": 0.63, "learning_rate": 6.221685254013777e-06, "loss": 1.0835, "step": 7494 }, { "epoch": 0.63, "learning_rate": 6.219145694656615e-06, "loss": 0.9903, "step": 7495 }, { "epoch": 0.63, "learning_rate": 6.216606419779072e-06, "loss": 0.9529, "step": 7496 }, { "epoch": 0.63, "learning_rate": 6.214067429572213e-06, "loss": 0.8481, "step": 7497 }, { "epoch": 0.63, "learning_rate": 6.2115287242270795e-06, "loss": 1.0207, "step": 7498 }, { "epoch": 0.64, "learning_rate": 6.2089903039346825e-06, "loss": 0.9723, "step": 7499 }, { "epoch": 0.64, "learning_rate": 6.206452168886026e-06, "loss": 1.002, "step": 7500 }, { "epoch": 0.64, "learning_rate": 6.203914319272077e-06, "loss": 1.0394, "step": 7501 }, { "epoch": 0.64, "learning_rate": 6.201376755283794e-06, "loss": 0.983, "step": 7502 }, { "epoch": 0.64, "learning_rate": 6.198839477112106e-06, "loss": 0.99, "step": 7503 }, { "epoch": 0.64, "learning_rate": 6.196302484947927e-06, "loss": 0.958, "step": 7504 }, { "epoch": 0.64, "learning_rate": 6.193765778982139e-06, "loss": 0.8998, "step": 7505 }, { "epoch": 0.64, "learning_rate": 6.191229359405617e-06, "loss": 1.0164, "step": 7506 }, { "epoch": 0.64, "learning_rate": 6.188693226409199e-06, "loss": 1.0041, "step": 7507 }, { "epoch": 0.64, "learning_rate": 6.186157380183717e-06, "loss": 0.9709, "step": 7508 }, { "epoch": 0.64, "learning_rate": 6.183621820919967e-06, "loss": 0.9371, "step": 7509 }, { "epoch": 0.64, "learning_rate": 6.181086548808734e-06, "loss": 1.0141, "step": 7510 }, { "epoch": 0.64, "learning_rate": 6.178551564040774e-06, "loss": 0.9899, "step": 7511 }, { "epoch": 0.64, "learning_rate": 6.176016866806831e-06, "loss": 1.0313, "step": 7512 }, { "epoch": 0.64, "learning_rate": 6.1734824572976114e-06, "loss": 0.9777, "step": 7513 }, { "epoch": 0.64, "learning_rate": 6.170948335703819e-06, "loss": 0.9616, "step": 7514 }, { "epoch": 0.64, "learning_rate": 6.16841450221612e-06, "loss": 0.7654, "step": 7515 }, { "epoch": 0.64, "learning_rate": 6.1658809570251696e-06, "loss": 1.0327, "step": 7516 }, { "epoch": 0.64, "learning_rate": 6.163347700321596e-06, "loss": 0.9733, "step": 7517 }, { "epoch": 0.64, "learning_rate": 6.160814732296006e-06, "loss": 0.9622, "step": 7518 }, { "epoch": 0.64, "learning_rate": 6.1582820531389845e-06, "loss": 1.0488, "step": 7519 }, { "epoch": 0.64, "learning_rate": 6.155749663041099e-06, "loss": 0.9715, "step": 7520 }, { "epoch": 0.64, "learning_rate": 6.153217562192888e-06, "loss": 1.0444, "step": 7521 }, { "epoch": 0.64, "learning_rate": 6.15068575078488e-06, "loss": 0.9978, "step": 7522 }, { "epoch": 0.64, "learning_rate": 6.148154229007561e-06, "loss": 0.869, "step": 7523 }, { "epoch": 0.64, "learning_rate": 6.145622997051417e-06, "loss": 0.9995, "step": 7524 }, { "epoch": 0.64, "learning_rate": 6.143092055106906e-06, "loss": 0.9619, "step": 7525 }, { "epoch": 0.64, "learning_rate": 6.140561403364454e-06, "loss": 0.9759, "step": 7526 }, { "epoch": 0.64, "learning_rate": 6.138031042014479e-06, "loss": 0.9977, "step": 7527 }, { "epoch": 0.64, "learning_rate": 6.135500971247367e-06, "loss": 0.9648, "step": 7528 }, { "epoch": 0.64, "learning_rate": 6.132971191253487e-06, "loss": 1.0042, "step": 7529 }, { "epoch": 0.64, "learning_rate": 6.130441702223184e-06, "loss": 1.0613, "step": 7530 }, { "epoch": 0.64, "learning_rate": 6.127912504346787e-06, "loss": 1.0158, "step": 7531 }, { "epoch": 0.64, "learning_rate": 6.125383597814589e-06, "loss": 1.0313, "step": 7532 }, { "epoch": 0.64, "learning_rate": 6.12285498281688e-06, "loss": 1.0061, "step": 7533 }, { "epoch": 0.64, "learning_rate": 6.120326659543911e-06, "loss": 1.0086, "step": 7534 }, { "epoch": 0.64, "learning_rate": 6.117798628185926e-06, "loss": 0.8707, "step": 7535 }, { "epoch": 0.64, "learning_rate": 6.115270888933131e-06, "loss": 0.9721, "step": 7536 }, { "epoch": 0.64, "learning_rate": 6.112743441975727e-06, "loss": 1.0317, "step": 7537 }, { "epoch": 0.64, "learning_rate": 6.110216287503876e-06, "loss": 1.0026, "step": 7538 }, { "epoch": 0.64, "learning_rate": 6.107689425707737e-06, "loss": 1.0334, "step": 7539 }, { "epoch": 0.64, "learning_rate": 6.105162856777425e-06, "loss": 1.0012, "step": 7540 }, { "epoch": 0.64, "learning_rate": 6.102636580903054e-06, "loss": 0.9708, "step": 7541 }, { "epoch": 0.64, "learning_rate": 6.100110598274699e-06, "loss": 1.0229, "step": 7542 }, { "epoch": 0.64, "learning_rate": 6.097584909082428e-06, "loss": 1.0042, "step": 7543 }, { "epoch": 0.64, "learning_rate": 6.095059513516273e-06, "loss": 1.0085, "step": 7544 }, { "epoch": 0.64, "learning_rate": 6.092534411766251e-06, "loss": 1.0234, "step": 7545 }, { "epoch": 0.64, "learning_rate": 6.09000960402236e-06, "loss": 0.9746, "step": 7546 }, { "epoch": 0.64, "learning_rate": 6.087485090474569e-06, "loss": 1.0322, "step": 7547 }, { "epoch": 0.64, "learning_rate": 6.0849608713128264e-06, "loss": 0.991, "step": 7548 }, { "epoch": 0.64, "learning_rate": 6.08243694672706e-06, "loss": 0.9838, "step": 7549 }, { "epoch": 0.64, "learning_rate": 6.079913316907182e-06, "loss": 1.0186, "step": 7550 }, { "epoch": 0.64, "learning_rate": 6.077389982043067e-06, "loss": 1.0288, "step": 7551 }, { "epoch": 0.64, "learning_rate": 6.0748669423245845e-06, "loss": 0.994, "step": 7552 }, { "epoch": 0.64, "learning_rate": 6.072344197941563e-06, "loss": 1.05, "step": 7553 }, { "epoch": 0.64, "learning_rate": 6.069821749083831e-06, "loss": 0.9719, "step": 7554 }, { "epoch": 0.64, "learning_rate": 6.067299595941174e-06, "loss": 1.0505, "step": 7555 }, { "epoch": 0.64, "learning_rate": 6.064777738703368e-06, "loss": 1.0125, "step": 7556 }, { "epoch": 0.64, "learning_rate": 6.062256177560162e-06, "loss": 1.0386, "step": 7557 }, { "epoch": 0.64, "learning_rate": 6.059734912701286e-06, "loss": 1.0359, "step": 7558 }, { "epoch": 0.64, "learning_rate": 6.0572139443164405e-06, "loss": 1.0087, "step": 7559 }, { "epoch": 0.64, "learning_rate": 6.054693272595313e-06, "loss": 0.9975, "step": 7560 }, { "epoch": 0.64, "learning_rate": 6.052172897727562e-06, "loss": 1.0216, "step": 7561 }, { "epoch": 0.64, "learning_rate": 6.0496528199028295e-06, "loss": 0.9998, "step": 7562 }, { "epoch": 0.64, "learning_rate": 6.047133039310724e-06, "loss": 0.9893, "step": 7563 }, { "epoch": 0.64, "learning_rate": 6.044613556140848e-06, "loss": 1.0218, "step": 7564 }, { "epoch": 0.64, "learning_rate": 6.042094370582765e-06, "loss": 0.9957, "step": 7565 }, { "epoch": 0.64, "learning_rate": 6.039575482826032e-06, "loss": 0.9823, "step": 7566 }, { "epoch": 0.64, "learning_rate": 6.037056893060169e-06, "loss": 1.0104, "step": 7567 }, { "epoch": 0.64, "learning_rate": 6.034538601474684e-06, "loss": 0.8917, "step": 7568 }, { "epoch": 0.64, "learning_rate": 6.032020608259055e-06, "loss": 1.05, "step": 7569 }, { "epoch": 0.64, "learning_rate": 6.029502913602747e-06, "loss": 0.9752, "step": 7570 }, { "epoch": 0.64, "learning_rate": 6.0269855176951895e-06, "loss": 0.9899, "step": 7571 }, { "epoch": 0.64, "learning_rate": 6.024468420725803e-06, "loss": 0.9803, "step": 7572 }, { "epoch": 0.64, "learning_rate": 6.021951622883975e-06, "loss": 1.0328, "step": 7573 }, { "epoch": 0.64, "learning_rate": 6.0194351243590796e-06, "loss": 1.0127, "step": 7574 }, { "epoch": 0.64, "learning_rate": 6.0169189253404554e-06, "loss": 0.9715, "step": 7575 }, { "epoch": 0.64, "learning_rate": 6.014403026017432e-06, "loss": 1.0192, "step": 7576 }, { "epoch": 0.64, "learning_rate": 6.011887426579317e-06, "loss": 1.0191, "step": 7577 }, { "epoch": 0.64, "learning_rate": 6.009372127215377e-06, "loss": 1.0216, "step": 7578 }, { "epoch": 0.64, "learning_rate": 6.00685712811488e-06, "loss": 0.993, "step": 7579 }, { "epoch": 0.64, "learning_rate": 6.0043424294670505e-06, "loss": 1.0017, "step": 7580 }, { "epoch": 0.64, "learning_rate": 6.001828031461107e-06, "loss": 1.0423, "step": 7581 }, { "epoch": 0.64, "learning_rate": 5.999313934286235e-06, "loss": 1.0083, "step": 7582 }, { "epoch": 0.64, "learning_rate": 5.996800138131601e-06, "loss": 0.9354, "step": 7583 }, { "epoch": 0.64, "learning_rate": 5.994286643186349e-06, "loss": 1.0256, "step": 7584 }, { "epoch": 0.64, "learning_rate": 5.991773449639599e-06, "loss": 0.802, "step": 7585 }, { "epoch": 0.64, "learning_rate": 5.989260557680447e-06, "loss": 0.9904, "step": 7586 }, { "epoch": 0.64, "learning_rate": 5.986747967497975e-06, "loss": 1.0221, "step": 7587 }, { "epoch": 0.64, "learning_rate": 5.9842356792812265e-06, "loss": 1.0142, "step": 7588 }, { "epoch": 0.64, "learning_rate": 5.981723693219241e-06, "loss": 1.0017, "step": 7589 }, { "epoch": 0.64, "learning_rate": 5.979212009501017e-06, "loss": 0.8741, "step": 7590 }, { "epoch": 0.64, "learning_rate": 5.976700628315546e-06, "loss": 0.9939, "step": 7591 }, { "epoch": 0.64, "learning_rate": 5.974189549851784e-06, "loss": 0.9704, "step": 7592 }, { "epoch": 0.64, "learning_rate": 5.971678774298677e-06, "loss": 1.0049, "step": 7593 }, { "epoch": 0.64, "learning_rate": 5.969168301845132e-06, "loss": 1.032, "step": 7594 }, { "epoch": 0.64, "learning_rate": 5.966658132680052e-06, "loss": 0.9959, "step": 7595 }, { "epoch": 0.64, "learning_rate": 5.9641482669923e-06, "loss": 1.0234, "step": 7596 }, { "epoch": 0.64, "learning_rate": 5.961638704970729e-06, "loss": 1.029, "step": 7597 }, { "epoch": 0.64, "learning_rate": 5.959129446804159e-06, "loss": 1.0315, "step": 7598 }, { "epoch": 0.64, "learning_rate": 5.956620492681395e-06, "loss": 0.9886, "step": 7599 }, { "epoch": 0.64, "learning_rate": 5.954111842791217e-06, "loss": 0.9913, "step": 7600 }, { "epoch": 0.64, "learning_rate": 5.951603497322376e-06, "loss": 0.9838, "step": 7601 }, { "epoch": 0.64, "learning_rate": 5.949095456463614e-06, "loss": 0.9718, "step": 7602 }, { "epoch": 0.64, "learning_rate": 5.946587720403632e-06, "loss": 0.9916, "step": 7603 }, { "epoch": 0.64, "learning_rate": 5.944080289331127e-06, "loss": 0.9707, "step": 7604 }, { "epoch": 0.64, "learning_rate": 5.9415731634347525e-06, "loss": 1.0392, "step": 7605 }, { "epoch": 0.64, "learning_rate": 5.939066342903161e-06, "loss": 0.9646, "step": 7606 }, { "epoch": 0.64, "learning_rate": 5.936559827924963e-06, "loss": 0.9958, "step": 7607 }, { "epoch": 0.64, "learning_rate": 5.9340536186887595e-06, "loss": 1.0103, "step": 7608 }, { "epoch": 0.64, "learning_rate": 5.931547715383119e-06, "loss": 1.0143, "step": 7609 }, { "epoch": 0.64, "learning_rate": 5.929042118196594e-06, "loss": 1.0125, "step": 7610 }, { "epoch": 0.64, "learning_rate": 5.926536827317709e-06, "loss": 0.9853, "step": 7611 }, { "epoch": 0.64, "learning_rate": 5.924031842934971e-06, "loss": 0.9081, "step": 7612 }, { "epoch": 0.64, "learning_rate": 5.9215271652368534e-06, "loss": 0.9676, "step": 7613 }, { "epoch": 0.64, "learning_rate": 5.919022794411821e-06, "loss": 0.9919, "step": 7614 }, { "epoch": 0.64, "learning_rate": 5.916518730648302e-06, "loss": 1.032, "step": 7615 }, { "epoch": 0.64, "learning_rate": 5.914014974134714e-06, "loss": 1.0231, "step": 7616 }, { "epoch": 0.65, "learning_rate": 5.9115115250594366e-06, "loss": 0.9626, "step": 7617 }, { "epoch": 0.65, "learning_rate": 5.9090083836108455e-06, "loss": 1.0534, "step": 7618 }, { "epoch": 0.65, "learning_rate": 5.9065055499772725e-06, "loss": 1.045, "step": 7619 }, { "epoch": 0.65, "learning_rate": 5.904003024347043e-06, "loss": 1.0546, "step": 7620 }, { "epoch": 0.65, "learning_rate": 5.901500806908447e-06, "loss": 0.9848, "step": 7621 }, { "epoch": 0.65, "learning_rate": 5.8989988978497616e-06, "loss": 0.9846, "step": 7622 }, { "epoch": 0.65, "learning_rate": 5.8964972973592335e-06, "loss": 0.9878, "step": 7623 }, { "epoch": 0.65, "learning_rate": 5.893996005625088e-06, "loss": 0.9519, "step": 7624 }, { "epoch": 0.65, "learning_rate": 5.891495022835528e-06, "loss": 1.0399, "step": 7625 }, { "epoch": 0.65, "learning_rate": 5.888994349178734e-06, "loss": 0.9611, "step": 7626 }, { "epoch": 0.65, "learning_rate": 5.88649398484286e-06, "loss": 1.0375, "step": 7627 }, { "epoch": 0.65, "learning_rate": 5.883993930016038e-06, "loss": 0.9797, "step": 7628 }, { "epoch": 0.65, "learning_rate": 5.881494184886385e-06, "loss": 0.955, "step": 7629 }, { "epoch": 0.65, "learning_rate": 5.8789947496419776e-06, "loss": 1.0474, "step": 7630 }, { "epoch": 0.65, "learning_rate": 5.876495624470886e-06, "loss": 1.0174, "step": 7631 }, { "epoch": 0.65, "learning_rate": 5.873996809561143e-06, "loss": 1.0351, "step": 7632 }, { "epoch": 0.65, "learning_rate": 5.871498305100774e-06, "loss": 0.9829, "step": 7633 }, { "epoch": 0.65, "learning_rate": 5.869000111277759e-06, "loss": 0.9922, "step": 7634 }, { "epoch": 0.65, "learning_rate": 5.866502228280081e-06, "loss": 0.9887, "step": 7635 }, { "epoch": 0.65, "learning_rate": 5.864004656295677e-06, "loss": 1.0142, "step": 7636 }, { "epoch": 0.65, "learning_rate": 5.861507395512473e-06, "loss": 0.9329, "step": 7637 }, { "epoch": 0.65, "learning_rate": 5.859010446118369e-06, "loss": 1.0077, "step": 7638 }, { "epoch": 0.65, "learning_rate": 5.85651380830124e-06, "loss": 0.9674, "step": 7639 }, { "epoch": 0.65, "learning_rate": 5.854017482248935e-06, "loss": 1.0109, "step": 7640 }, { "epoch": 0.65, "learning_rate": 5.85152146814929e-06, "loss": 1.0017, "step": 7641 }, { "epoch": 0.65, "learning_rate": 5.849025766190102e-06, "loss": 0.9565, "step": 7642 }, { "epoch": 0.65, "learning_rate": 5.846530376559162e-06, "loss": 1.0299, "step": 7643 }, { "epoch": 0.65, "learning_rate": 5.844035299444219e-06, "loss": 1.04, "step": 7644 }, { "epoch": 0.65, "learning_rate": 5.841540535033019e-06, "loss": 1.0169, "step": 7645 }, { "epoch": 0.65, "learning_rate": 5.839046083513262e-06, "loss": 1.0334, "step": 7646 }, { "epoch": 0.65, "learning_rate": 5.836551945072646e-06, "loss": 0.9947, "step": 7647 }, { "epoch": 0.65, "learning_rate": 5.8340581198988265e-06, "loss": 0.9828, "step": 7648 }, { "epoch": 0.65, "learning_rate": 5.831564608179452e-06, "loss": 0.9711, "step": 7649 }, { "epoch": 0.65, "learning_rate": 5.829071410102132e-06, "loss": 0.9546, "step": 7650 }, { "epoch": 0.65, "learning_rate": 5.826578525854469e-06, "loss": 0.9894, "step": 7651 }, { "epoch": 0.65, "learning_rate": 5.824085955624025e-06, "loss": 1.026, "step": 7652 }, { "epoch": 0.65, "learning_rate": 5.821593699598349e-06, "loss": 1.0201, "step": 7653 }, { "epoch": 0.65, "learning_rate": 5.8191017579649664e-06, "loss": 0.9815, "step": 7654 }, { "epoch": 0.65, "learning_rate": 5.816610130911377e-06, "loss": 0.9825, "step": 7655 }, { "epoch": 0.65, "learning_rate": 5.814118818625048e-06, "loss": 1.0441, "step": 7656 }, { "epoch": 0.65, "learning_rate": 5.811627821293441e-06, "loss": 0.8634, "step": 7657 }, { "epoch": 0.65, "learning_rate": 5.809137139103979e-06, "loss": 0.9929, "step": 7658 }, { "epoch": 0.65, "learning_rate": 5.806646772244063e-06, "loss": 0.9963, "step": 7659 }, { "epoch": 0.65, "learning_rate": 5.804156720901082e-06, "loss": 0.9888, "step": 7660 }, { "epoch": 0.65, "learning_rate": 5.801666985262382e-06, "loss": 0.9694, "step": 7661 }, { "epoch": 0.65, "learning_rate": 5.799177565515308e-06, "loss": 0.9926, "step": 7662 }, { "epoch": 0.65, "learning_rate": 5.7966884618471595e-06, "loss": 0.9932, "step": 7663 }, { "epoch": 0.65, "learning_rate": 5.79419967444523e-06, "loss": 0.9852, "step": 7664 }, { "epoch": 0.65, "learning_rate": 5.791711203496773e-06, "loss": 0.9884, "step": 7665 }, { "epoch": 0.65, "learning_rate": 5.7892230491890345e-06, "loss": 0.9947, "step": 7666 }, { "epoch": 0.65, "learning_rate": 5.7867352117092215e-06, "loss": 1.0426, "step": 7667 }, { "epoch": 0.65, "learning_rate": 5.784247691244531e-06, "loss": 1.0578, "step": 7668 }, { "epoch": 0.65, "learning_rate": 5.781760487982122e-06, "loss": 0.9874, "step": 7669 }, { "epoch": 0.65, "learning_rate": 5.779273602109146e-06, "loss": 0.9799, "step": 7670 }, { "epoch": 0.65, "learning_rate": 5.7767870338127105e-06, "loss": 1.0047, "step": 7671 }, { "epoch": 0.65, "learning_rate": 5.774300783279922e-06, "loss": 1.0173, "step": 7672 }, { "epoch": 0.65, "learning_rate": 5.7718148506978415e-06, "loss": 0.9799, "step": 7673 }, { "epoch": 0.65, "learning_rate": 5.7693292362535235e-06, "loss": 1.0409, "step": 7674 }, { "epoch": 0.65, "learning_rate": 5.766843940133984e-06, "loss": 0.9802, "step": 7675 }, { "epoch": 0.65, "learning_rate": 5.7643589625262305e-06, "loss": 1.0183, "step": 7676 }, { "epoch": 0.65, "learning_rate": 5.761874303617227e-06, "loss": 1.0124, "step": 7677 }, { "epoch": 0.65, "learning_rate": 5.759389963593937e-06, "loss": 1.0984, "step": 7678 }, { "epoch": 0.65, "learning_rate": 5.756905942643277e-06, "loss": 1.0545, "step": 7679 }, { "epoch": 0.65, "learning_rate": 5.7544222409521545e-06, "loss": 0.9802, "step": 7680 }, { "epoch": 0.65, "learning_rate": 5.751938858707452e-06, "loss": 0.9725, "step": 7681 }, { "epoch": 0.65, "learning_rate": 5.749455796096022e-06, "loss": 1.0177, "step": 7682 }, { "epoch": 0.65, "learning_rate": 5.746973053304689e-06, "loss": 0.9926, "step": 7683 }, { "epoch": 0.65, "learning_rate": 5.744490630520272e-06, "loss": 0.8294, "step": 7684 }, { "epoch": 0.65, "learning_rate": 5.742008527929545e-06, "loss": 1.0393, "step": 7685 }, { "epoch": 0.65, "learning_rate": 5.739526745719266e-06, "loss": 1.0107, "step": 7686 }, { "epoch": 0.65, "learning_rate": 5.737045284076177e-06, "loss": 0.9871, "step": 7687 }, { "epoch": 0.65, "learning_rate": 5.73456414318698e-06, "loss": 1.0709, "step": 7688 }, { "epoch": 0.65, "learning_rate": 5.732083323238369e-06, "loss": 0.852, "step": 7689 }, { "epoch": 0.65, "learning_rate": 5.729602824417e-06, "loss": 1.0163, "step": 7690 }, { "epoch": 0.65, "learning_rate": 5.727122646909518e-06, "loss": 1.0279, "step": 7691 }, { "epoch": 0.65, "learning_rate": 5.724642790902528e-06, "loss": 1.0116, "step": 7692 }, { "epoch": 0.65, "learning_rate": 5.722163256582629e-06, "loss": 1.0273, "step": 7693 }, { "epoch": 0.65, "learning_rate": 5.719684044136378e-06, "loss": 0.935, "step": 7694 }, { "epoch": 0.65, "learning_rate": 5.7172051537503226e-06, "loss": 1.0124, "step": 7695 }, { "epoch": 0.65, "learning_rate": 5.714726585610976e-06, "loss": 0.9479, "step": 7696 }, { "epoch": 0.65, "learning_rate": 5.7122483399048356e-06, "loss": 1.0566, "step": 7697 }, { "epoch": 0.65, "learning_rate": 5.709770416818363e-06, "loss": 0.9981, "step": 7698 }, { "epoch": 0.65, "learning_rate": 5.7072928165380105e-06, "loss": 1.0212, "step": 7699 }, { "epoch": 0.65, "learning_rate": 5.70481553925019e-06, "loss": 1.012, "step": 7700 }, { "epoch": 0.65, "learning_rate": 5.702338585141305e-06, "loss": 0.9523, "step": 7701 }, { "epoch": 0.65, "learning_rate": 5.699861954397719e-06, "loss": 0.9957, "step": 7702 }, { "epoch": 0.65, "learning_rate": 5.697385647205788e-06, "loss": 0.9261, "step": 7703 }, { "epoch": 0.65, "learning_rate": 5.694909663751825e-06, "loss": 1.0339, "step": 7704 }, { "epoch": 0.65, "learning_rate": 5.6924340042221336e-06, "loss": 0.9763, "step": 7705 }, { "epoch": 0.65, "learning_rate": 5.689958668802991e-06, "loss": 0.9926, "step": 7706 }, { "epoch": 0.65, "learning_rate": 5.687483657680639e-06, "loss": 0.9878, "step": 7707 }, { "epoch": 0.65, "learning_rate": 5.68500897104131e-06, "loss": 1.0009, "step": 7708 }, { "epoch": 0.65, "learning_rate": 5.682534609071202e-06, "loss": 0.9802, "step": 7709 }, { "epoch": 0.65, "learning_rate": 5.680060571956487e-06, "loss": 1.0364, "step": 7710 }, { "epoch": 0.65, "learning_rate": 5.677586859883325e-06, "loss": 0.9834, "step": 7711 }, { "epoch": 0.65, "learning_rate": 5.675113473037839e-06, "loss": 1.0683, "step": 7712 }, { "epoch": 0.65, "learning_rate": 5.672640411606127e-06, "loss": 1.0609, "step": 7713 }, { "epoch": 0.65, "learning_rate": 5.670167675774279e-06, "loss": 0.9729, "step": 7714 }, { "epoch": 0.65, "learning_rate": 5.6676952657283366e-06, "loss": 1.0139, "step": 7715 }, { "epoch": 0.65, "learning_rate": 5.66522318165434e-06, "loss": 0.9598, "step": 7716 }, { "epoch": 0.65, "learning_rate": 5.662751423738286e-06, "loss": 1.0172, "step": 7717 }, { "epoch": 0.65, "learning_rate": 5.66027999216616e-06, "loss": 1.0113, "step": 7718 }, { "epoch": 0.65, "learning_rate": 5.657808887123914e-06, "loss": 1.0499, "step": 7719 }, { "epoch": 0.65, "learning_rate": 5.655338108797485e-06, "loss": 0.8811, "step": 7720 }, { "epoch": 0.65, "learning_rate": 5.652867657372774e-06, "loss": 0.9955, "step": 7721 }, { "epoch": 0.65, "learning_rate": 5.6503975330356664e-06, "loss": 0.9592, "step": 7722 }, { "epoch": 0.65, "learning_rate": 5.647927735972016e-06, "loss": 1.0456, "step": 7723 }, { "epoch": 0.65, "learning_rate": 5.645458266367662e-06, "loss": 1.0216, "step": 7724 }, { "epoch": 0.65, "learning_rate": 5.6429891244084045e-06, "loss": 0.9834, "step": 7725 }, { "epoch": 0.65, "learning_rate": 5.640520310280036e-06, "loss": 1.0005, "step": 7726 }, { "epoch": 0.65, "learning_rate": 5.638051824168305e-06, "loss": 0.9745, "step": 7727 }, { "epoch": 0.65, "learning_rate": 5.635583666258957e-06, "loss": 0.9679, "step": 7728 }, { "epoch": 0.65, "learning_rate": 5.63311583673769e-06, "loss": 0.9948, "step": 7729 }, { "epoch": 0.65, "learning_rate": 5.630648335790201e-06, "loss": 1.0341, "step": 7730 }, { "epoch": 0.65, "learning_rate": 5.628181163602139e-06, "loss": 0.9857, "step": 7731 }, { "epoch": 0.65, "learning_rate": 5.625714320359143e-06, "loss": 0.9871, "step": 7732 }, { "epoch": 0.65, "learning_rate": 5.623247806246831e-06, "loss": 0.9454, "step": 7733 }, { "epoch": 0.65, "learning_rate": 5.620781621450777e-06, "loss": 1.0123, "step": 7734 }, { "epoch": 0.66, "learning_rate": 5.618315766156553e-06, "loss": 1.0358, "step": 7735 }, { "epoch": 0.66, "learning_rate": 5.6158502405496895e-06, "loss": 1.0067, "step": 7736 }, { "epoch": 0.66, "learning_rate": 5.6133850448156934e-06, "loss": 1.0041, "step": 7737 }, { "epoch": 0.66, "learning_rate": 5.610920179140062e-06, "loss": 1.0101, "step": 7738 }, { "epoch": 0.66, "learning_rate": 5.608455643708251e-06, "loss": 1.0252, "step": 7739 }, { "epoch": 0.66, "learning_rate": 5.605991438705693e-06, "loss": 1.0007, "step": 7740 }, { "epoch": 0.66, "learning_rate": 5.603527564317809e-06, "loss": 0.9004, "step": 7741 }, { "epoch": 0.66, "learning_rate": 5.601064020729978e-06, "loss": 1.0312, "step": 7742 }, { "epoch": 0.66, "learning_rate": 5.598600808127571e-06, "loss": 0.9756, "step": 7743 }, { "epoch": 0.66, "learning_rate": 5.596137926695917e-06, "loss": 0.965, "step": 7744 }, { "epoch": 0.66, "learning_rate": 5.5936753766203355e-06, "loss": 0.9825, "step": 7745 }, { "epoch": 0.66, "learning_rate": 5.591213158086107e-06, "loss": 1.0265, "step": 7746 }, { "epoch": 0.66, "learning_rate": 5.588751271278503e-06, "loss": 1.0087, "step": 7747 }, { "epoch": 0.66, "learning_rate": 5.586289716382751e-06, "loss": 0.9925, "step": 7748 }, { "epoch": 0.66, "learning_rate": 5.583828493584074e-06, "loss": 1.0084, "step": 7749 }, { "epoch": 0.66, "learning_rate": 5.581367603067651e-06, "loss": 0.9959, "step": 7750 }, { "epoch": 0.66, "learning_rate": 5.578907045018651e-06, "loss": 1.0031, "step": 7751 }, { "epoch": 0.66, "learning_rate": 5.576446819622205e-06, "loss": 0.9994, "step": 7752 }, { "epoch": 0.66, "learning_rate": 5.573986927063435e-06, "loss": 0.9472, "step": 7753 }, { "epoch": 0.66, "learning_rate": 5.571527367527418e-06, "loss": 0.9885, "step": 7754 }, { "epoch": 0.66, "learning_rate": 5.569068141199227e-06, "loss": 0.9892, "step": 7755 }, { "epoch": 0.66, "learning_rate": 5.5666092482638904e-06, "loss": 1.0147, "step": 7756 }, { "epoch": 0.66, "learning_rate": 5.564150688906424e-06, "loss": 0.9364, "step": 7757 }, { "epoch": 0.66, "learning_rate": 5.561692463311818e-06, "loss": 1.0197, "step": 7758 }, { "epoch": 0.66, "learning_rate": 5.559234571665031e-06, "loss": 0.9546, "step": 7759 }, { "epoch": 0.66, "learning_rate": 5.5567770141510045e-06, "loss": 0.9718, "step": 7760 }, { "epoch": 0.66, "learning_rate": 5.554319790954643e-06, "loss": 1.0642, "step": 7761 }, { "epoch": 0.66, "learning_rate": 5.551862902260841e-06, "loss": 0.9793, "step": 7762 }, { "epoch": 0.66, "learning_rate": 5.549406348254457e-06, "loss": 1.0359, "step": 7763 }, { "epoch": 0.66, "learning_rate": 5.546950129120323e-06, "loss": 1.0298, "step": 7764 }, { "epoch": 0.66, "learning_rate": 5.544494245043257e-06, "loss": 0.9665, "step": 7765 }, { "epoch": 0.66, "learning_rate": 5.542038696208044e-06, "loss": 1.0234, "step": 7766 }, { "epoch": 0.66, "learning_rate": 5.539583482799438e-06, "loss": 0.9784, "step": 7767 }, { "epoch": 0.66, "learning_rate": 5.537128605002185e-06, "loss": 0.9684, "step": 7768 }, { "epoch": 0.66, "learning_rate": 5.5346740630009844e-06, "loss": 1.0139, "step": 7769 }, { "epoch": 0.66, "learning_rate": 5.532219856980531e-06, "loss": 1.0015, "step": 7770 }, { "epoch": 0.66, "learning_rate": 5.5297659871254764e-06, "loss": 0.9962, "step": 7771 }, { "epoch": 0.66, "learning_rate": 5.527312453620462e-06, "loss": 0.9826, "step": 7772 }, { "epoch": 0.66, "learning_rate": 5.52485925665009e-06, "loss": 0.9914, "step": 7773 }, { "epoch": 0.66, "learning_rate": 5.522406396398954e-06, "loss": 1.0345, "step": 7774 }, { "epoch": 0.66, "learning_rate": 5.5199538730516e-06, "loss": 0.9559, "step": 7775 }, { "epoch": 0.66, "learning_rate": 5.5175016867925735e-06, "loss": 0.9119, "step": 7776 }, { "epoch": 0.66, "learning_rate": 5.515049837806372e-06, "loss": 0.9592, "step": 7777 }, { "epoch": 0.66, "learning_rate": 5.512598326277485e-06, "loss": 0.9412, "step": 7778 }, { "epoch": 0.66, "learning_rate": 5.510147152390365e-06, "loss": 0.9764, "step": 7779 }, { "epoch": 0.66, "learning_rate": 5.507696316329448e-06, "loss": 1.0052, "step": 7780 }, { "epoch": 0.66, "learning_rate": 5.505245818279135e-06, "loss": 0.963, "step": 7781 }, { "epoch": 0.66, "learning_rate": 5.5027956584238114e-06, "loss": 1.0232, "step": 7782 }, { "epoch": 0.66, "learning_rate": 5.500345836947828e-06, "loss": 1.0453, "step": 7783 }, { "epoch": 0.66, "learning_rate": 5.4978963540355165e-06, "loss": 1.0181, "step": 7784 }, { "epoch": 0.66, "learning_rate": 5.4954472098711865e-06, "loss": 0.9336, "step": 7785 }, { "epoch": 0.66, "learning_rate": 5.492998404639108e-06, "loss": 1.0367, "step": 7786 }, { "epoch": 0.66, "learning_rate": 5.4905499385235425e-06, "loss": 1.0237, "step": 7787 }, { "epoch": 0.66, "learning_rate": 5.488101811708715e-06, "loss": 1.0002, "step": 7788 }, { "epoch": 0.66, "learning_rate": 5.485654024378822e-06, "loss": 1.0403, "step": 7789 }, { "epoch": 0.66, "learning_rate": 5.48320657671805e-06, "loss": 1.0097, "step": 7790 }, { "epoch": 0.66, "learning_rate": 5.480759468910539e-06, "loss": 0.9916, "step": 7791 }, { "epoch": 0.66, "learning_rate": 5.478312701140427e-06, "loss": 0.9685, "step": 7792 }, { "epoch": 0.66, "learning_rate": 5.475866273591809e-06, "loss": 0.9401, "step": 7793 }, { "epoch": 0.66, "learning_rate": 5.473420186448753e-06, "loss": 1.009, "step": 7794 }, { "epoch": 0.66, "learning_rate": 5.470974439895318e-06, "loss": 0.9755, "step": 7795 }, { "epoch": 0.66, "learning_rate": 5.4685290341155195e-06, "loss": 0.9655, "step": 7796 }, { "epoch": 0.66, "learning_rate": 5.466083969293361e-06, "loss": 1.0058, "step": 7797 }, { "epoch": 0.66, "learning_rate": 5.4636392456128085e-06, "loss": 1.0135, "step": 7798 }, { "epoch": 0.66, "learning_rate": 5.461194863257816e-06, "loss": 1.0559, "step": 7799 }, { "epoch": 0.66, "learning_rate": 5.458750822412296e-06, "loss": 0.9737, "step": 7800 }, { "epoch": 0.66, "learning_rate": 5.4563071232601495e-06, "loss": 0.9305, "step": 7801 }, { "epoch": 0.66, "learning_rate": 5.45386376598524e-06, "loss": 1.0187, "step": 7802 }, { "epoch": 0.66, "learning_rate": 5.451420750771419e-06, "loss": 0.9837, "step": 7803 }, { "epoch": 0.66, "learning_rate": 5.448978077802493e-06, "loss": 0.9925, "step": 7804 }, { "epoch": 0.66, "learning_rate": 5.446535747262267e-06, "loss": 0.9193, "step": 7805 }, { "epoch": 0.66, "learning_rate": 5.444093759334496e-06, "loss": 1.0076, "step": 7806 }, { "epoch": 0.66, "learning_rate": 5.441652114202928e-06, "loss": 1.0079, "step": 7807 }, { "epoch": 0.66, "learning_rate": 5.439210812051272e-06, "loss": 0.9958, "step": 7808 }, { "epoch": 0.66, "learning_rate": 5.43676985306322e-06, "loss": 1.0328, "step": 7809 }, { "epoch": 0.66, "learning_rate": 5.434329237422436e-06, "loss": 1.0135, "step": 7810 }, { "epoch": 0.66, "learning_rate": 5.431888965312554e-06, "loss": 1.0508, "step": 7811 }, { "epoch": 0.66, "learning_rate": 5.429449036917189e-06, "loss": 1.0439, "step": 7812 }, { "epoch": 0.66, "learning_rate": 5.427009452419922e-06, "loss": 1.0057, "step": 7813 }, { "epoch": 0.66, "learning_rate": 5.4245702120043195e-06, "loss": 1.001, "step": 7814 }, { "epoch": 0.66, "learning_rate": 5.42213131585391e-06, "loss": 1.0139, "step": 7815 }, { "epoch": 0.66, "learning_rate": 5.419692764152199e-06, "loss": 0.995, "step": 7816 }, { "epoch": 0.66, "learning_rate": 5.417254557082674e-06, "loss": 0.9932, "step": 7817 }, { "epoch": 0.66, "learning_rate": 5.41481669482879e-06, "loss": 1.0195, "step": 7818 }, { "epoch": 0.66, "learning_rate": 5.412379177573971e-06, "loss": 1.0309, "step": 7819 }, { "epoch": 0.66, "learning_rate": 5.409942005501629e-06, "loss": 0.9745, "step": 7820 }, { "epoch": 0.66, "learning_rate": 5.407505178795135e-06, "loss": 1.0411, "step": 7821 }, { "epoch": 0.66, "learning_rate": 5.4050686976378476e-06, "loss": 0.9853, "step": 7822 }, { "epoch": 0.66, "learning_rate": 5.402632562213088e-06, "loss": 0.9349, "step": 7823 }, { "epoch": 0.66, "learning_rate": 5.400196772704161e-06, "loss": 0.9126, "step": 7824 }, { "epoch": 0.66, "learning_rate": 5.397761329294333e-06, "loss": 0.9453, "step": 7825 }, { "epoch": 0.66, "learning_rate": 5.395326232166862e-06, "loss": 1.0035, "step": 7826 }, { "epoch": 0.66, "learning_rate": 5.392891481504962e-06, "loss": 0.9671, "step": 7827 }, { "epoch": 0.66, "learning_rate": 5.390457077491834e-06, "loss": 1.0024, "step": 7828 }, { "epoch": 0.66, "learning_rate": 5.388023020310642e-06, "loss": 0.8418, "step": 7829 }, { "epoch": 0.66, "learning_rate": 5.385589310144537e-06, "loss": 0.8957, "step": 7830 }, { "epoch": 0.66, "learning_rate": 5.38315594717663e-06, "loss": 1.0261, "step": 7831 }, { "epoch": 0.66, "learning_rate": 5.380722931590018e-06, "loss": 0.9888, "step": 7832 }, { "epoch": 0.66, "learning_rate": 5.37829026356776e-06, "loss": 1.0596, "step": 7833 }, { "epoch": 0.66, "learning_rate": 5.375857943292904e-06, "loss": 1.0126, "step": 7834 }, { "epoch": 0.66, "learning_rate": 5.373425970948452e-06, "loss": 1.0285, "step": 7835 }, { "epoch": 0.66, "learning_rate": 5.3709943467173975e-06, "loss": 1.0193, "step": 7836 }, { "epoch": 0.66, "learning_rate": 5.368563070782704e-06, "loss": 1.0028, "step": 7837 }, { "epoch": 0.66, "learning_rate": 5.366132143327298e-06, "loss": 1.0237, "step": 7838 }, { "epoch": 0.66, "learning_rate": 5.363701564534098e-06, "loss": 1.0278, "step": 7839 }, { "epoch": 0.66, "learning_rate": 5.361271334585973e-06, "loss": 0.9887, "step": 7840 }, { "epoch": 0.66, "learning_rate": 5.358841453665793e-06, "loss": 0.9926, "step": 7841 }, { "epoch": 0.66, "learning_rate": 5.356411921956377e-06, "loss": 0.9329, "step": 7842 }, { "epoch": 0.66, "learning_rate": 5.35398273964053e-06, "loss": 1.0406, "step": 7843 }, { "epoch": 0.66, "learning_rate": 5.351553906901032e-06, "loss": 0.9904, "step": 7844 }, { "epoch": 0.66, "learning_rate": 5.349125423920634e-06, "loss": 0.9484, "step": 7845 }, { "epoch": 0.66, "learning_rate": 5.346697290882053e-06, "loss": 0.9984, "step": 7846 }, { "epoch": 0.66, "learning_rate": 5.344269507967996e-06, "loss": 0.9951, "step": 7847 }, { "epoch": 0.66, "learning_rate": 5.341842075361128e-06, "loss": 0.9619, "step": 7848 }, { "epoch": 0.66, "learning_rate": 5.339414993244101e-06, "loss": 0.9641, "step": 7849 }, { "epoch": 0.66, "learning_rate": 5.336988261799525e-06, "loss": 1.014, "step": 7850 }, { "epoch": 0.66, "learning_rate": 5.3345618812100005e-06, "loss": 1.0464, "step": 7851 }, { "epoch": 0.66, "learning_rate": 5.332135851658088e-06, "loss": 1.0158, "step": 7852 }, { "epoch": 0.67, "learning_rate": 5.329710173326333e-06, "loss": 0.9634, "step": 7853 }, { "epoch": 0.67, "learning_rate": 5.3272848463972425e-06, "loss": 1.0398, "step": 7854 }, { "epoch": 0.67, "learning_rate": 5.324859871053309e-06, "loss": 0.9999, "step": 7855 }, { "epoch": 0.67, "learning_rate": 5.322435247476987e-06, "loss": 0.9957, "step": 7856 }, { "epoch": 0.67, "learning_rate": 5.320010975850718e-06, "loss": 1.0137, "step": 7857 }, { "epoch": 0.67, "learning_rate": 5.317587056356898e-06, "loss": 0.9593, "step": 7858 }, { "epoch": 0.67, "learning_rate": 5.315163489177922e-06, "loss": 0.9913, "step": 7859 }, { "epoch": 0.67, "learning_rate": 5.312740274496133e-06, "loss": 0.9704, "step": 7860 }, { "epoch": 0.67, "learning_rate": 5.310317412493864e-06, "loss": 0.9992, "step": 7861 }, { "epoch": 0.67, "learning_rate": 5.307894903353416e-06, "loss": 1.021, "step": 7862 }, { "epoch": 0.67, "learning_rate": 5.305472747257063e-06, "loss": 1.0005, "step": 7863 }, { "epoch": 0.67, "learning_rate": 5.303050944387056e-06, "loss": 1.0042, "step": 7864 }, { "epoch": 0.67, "learning_rate": 5.300629494925611e-06, "loss": 0.9877, "step": 7865 }, { "epoch": 0.67, "learning_rate": 5.29820839905493e-06, "loss": 1.0382, "step": 7866 }, { "epoch": 0.67, "learning_rate": 5.295787656957175e-06, "loss": 0.9796, "step": 7867 }, { "epoch": 0.67, "learning_rate": 5.293367268814494e-06, "loss": 0.9703, "step": 7868 }, { "epoch": 0.67, "learning_rate": 5.2909472348089966e-06, "loss": 0.9785, "step": 7869 }, { "epoch": 0.67, "learning_rate": 5.288527555122772e-06, "loss": 0.9822, "step": 7870 }, { "epoch": 0.67, "learning_rate": 5.286108229937888e-06, "loss": 0.971, "step": 7871 }, { "epoch": 0.67, "learning_rate": 5.2836892594363745e-06, "loss": 0.8757, "step": 7872 }, { "epoch": 0.67, "learning_rate": 5.281270643800237e-06, "loss": 0.9861, "step": 7873 }, { "epoch": 0.67, "learning_rate": 5.278852383211466e-06, "loss": 0.9874, "step": 7874 }, { "epoch": 0.67, "learning_rate": 5.276434477852008e-06, "loss": 1.0088, "step": 7875 }, { "epoch": 0.67, "learning_rate": 5.2740169279038e-06, "loss": 0.9793, "step": 7876 }, { "epoch": 0.67, "learning_rate": 5.271599733548735e-06, "loss": 1.0264, "step": 7877 }, { "epoch": 0.67, "learning_rate": 5.269182894968696e-06, "loss": 1.0011, "step": 7878 }, { "epoch": 0.67, "learning_rate": 5.266766412345526e-06, "loss": 0.9999, "step": 7879 }, { "epoch": 0.67, "learning_rate": 5.264350285861049e-06, "loss": 0.9753, "step": 7880 }, { "epoch": 0.67, "learning_rate": 5.261934515697053e-06, "loss": 0.9799, "step": 7881 }, { "epoch": 0.67, "learning_rate": 5.2595191020353165e-06, "loss": 0.9894, "step": 7882 }, { "epoch": 0.67, "learning_rate": 5.2571040450575705e-06, "loss": 0.8269, "step": 7883 }, { "epoch": 0.67, "learning_rate": 5.254689344945538e-06, "loss": 0.9006, "step": 7884 }, { "epoch": 0.67, "learning_rate": 5.252275001880897e-06, "loss": 1.0534, "step": 7885 }, { "epoch": 0.67, "learning_rate": 5.249861016045317e-06, "loss": 0.9688, "step": 7886 }, { "epoch": 0.67, "learning_rate": 5.247447387620424e-06, "loss": 0.9955, "step": 7887 }, { "epoch": 0.67, "learning_rate": 5.245034116787826e-06, "loss": 0.9941, "step": 7888 }, { "epoch": 0.67, "learning_rate": 5.24262120372911e-06, "loss": 1.0214, "step": 7889 }, { "epoch": 0.67, "learning_rate": 5.240208648625818e-06, "loss": 0.9999, "step": 7890 }, { "epoch": 0.67, "learning_rate": 5.237796451659485e-06, "loss": 1.0379, "step": 7891 }, { "epoch": 0.67, "learning_rate": 5.2353846130116025e-06, "loss": 1.0256, "step": 7892 }, { "epoch": 0.67, "learning_rate": 5.23297313286365e-06, "loss": 0.8854, "step": 7893 }, { "epoch": 0.67, "learning_rate": 5.2305620113970635e-06, "loss": 1.0295, "step": 7894 }, { "epoch": 0.67, "learning_rate": 5.22815124879327e-06, "loss": 1.0217, "step": 7895 }, { "epoch": 0.67, "learning_rate": 5.225740845233657e-06, "loss": 1.0246, "step": 7896 }, { "epoch": 0.67, "learning_rate": 5.223330800899582e-06, "loss": 1.0137, "step": 7897 }, { "epoch": 0.67, "learning_rate": 5.220921115972393e-06, "loss": 0.9951, "step": 7898 }, { "epoch": 0.67, "learning_rate": 5.218511790633393e-06, "loss": 1.0457, "step": 7899 }, { "epoch": 0.67, "learning_rate": 5.216102825063865e-06, "loss": 0.9851, "step": 7900 }, { "epoch": 0.67, "learning_rate": 5.213694219445069e-06, "loss": 1.0225, "step": 7901 }, { "epoch": 0.67, "learning_rate": 5.211285973958227e-06, "loss": 0.9535, "step": 7902 }, { "epoch": 0.67, "learning_rate": 5.208878088784547e-06, "loss": 0.9686, "step": 7903 }, { "epoch": 0.67, "learning_rate": 5.206470564105197e-06, "loss": 0.9781, "step": 7904 }, { "epoch": 0.67, "learning_rate": 5.204063400101333e-06, "loss": 1.0358, "step": 7905 }, { "epoch": 0.67, "learning_rate": 5.201656596954067e-06, "loss": 1.0732, "step": 7906 }, { "epoch": 0.67, "learning_rate": 5.1992501548444975e-06, "loss": 1.0102, "step": 7907 }, { "epoch": 0.67, "learning_rate": 5.196844073953686e-06, "loss": 0.9809, "step": 7908 }, { "epoch": 0.67, "learning_rate": 5.194438354462676e-06, "loss": 0.9972, "step": 7909 }, { "epoch": 0.67, "learning_rate": 5.192032996552474e-06, "loss": 1.0093, "step": 7910 }, { "epoch": 0.67, "learning_rate": 5.18962800040407e-06, "loss": 1.0098, "step": 7911 }, { "epoch": 0.67, "learning_rate": 5.187223366198415e-06, "loss": 0.9701, "step": 7912 }, { "epoch": 0.67, "learning_rate": 5.18481909411644e-06, "loss": 1.0132, "step": 7913 }, { "epoch": 0.67, "learning_rate": 5.182415184339056e-06, "loss": 0.9942, "step": 7914 }, { "epoch": 0.67, "learning_rate": 5.180011637047126e-06, "loss": 0.9649, "step": 7915 }, { "epoch": 0.67, "learning_rate": 5.177608452421507e-06, "loss": 1.0107, "step": 7916 }, { "epoch": 0.67, "learning_rate": 5.175205630643014e-06, "loss": 1.0087, "step": 7917 }, { "epoch": 0.67, "learning_rate": 5.172803171892446e-06, "loss": 0.9765, "step": 7918 }, { "epoch": 0.67, "learning_rate": 5.170401076350564e-06, "loss": 1.0298, "step": 7919 }, { "epoch": 0.67, "learning_rate": 5.167999344198111e-06, "loss": 0.9963, "step": 7920 }, { "epoch": 0.67, "learning_rate": 5.165597975615795e-06, "loss": 0.9742, "step": 7921 }, { "epoch": 0.67, "learning_rate": 5.163196970784303e-06, "loss": 1.0204, "step": 7922 }, { "epoch": 0.67, "learning_rate": 5.1607963298842926e-06, "loss": 0.9959, "step": 7923 }, { "epoch": 0.67, "learning_rate": 5.158396053096386e-06, "loss": 0.9711, "step": 7924 }, { "epoch": 0.67, "learning_rate": 5.155996140601196e-06, "loss": 0.9759, "step": 7925 }, { "epoch": 0.67, "learning_rate": 5.1535965925792905e-06, "loss": 1.002, "step": 7926 }, { "epoch": 0.67, "learning_rate": 5.151197409211213e-06, "loss": 0.9963, "step": 7927 }, { "epoch": 0.67, "learning_rate": 5.148798590677493e-06, "loss": 0.9779, "step": 7928 }, { "epoch": 0.67, "learning_rate": 5.146400137158614e-06, "loss": 0.9882, "step": 7929 }, { "epoch": 0.67, "learning_rate": 5.144002048835046e-06, "loss": 0.9898, "step": 7930 }, { "epoch": 0.67, "learning_rate": 5.141604325887224e-06, "loss": 0.9757, "step": 7931 }, { "epoch": 0.67, "learning_rate": 5.139206968495561e-06, "loss": 0.9995, "step": 7932 }, { "epoch": 0.67, "learning_rate": 5.136809976840433e-06, "loss": 0.995, "step": 7933 }, { "epoch": 0.67, "learning_rate": 5.134413351102203e-06, "loss": 0.9645, "step": 7934 }, { "epoch": 0.67, "learning_rate": 5.13201709146119e-06, "loss": 0.9788, "step": 7935 }, { "epoch": 0.67, "learning_rate": 5.129621198097702e-06, "loss": 1.0594, "step": 7936 }, { "epoch": 0.67, "learning_rate": 5.127225671192002e-06, "loss": 1.0405, "step": 7937 }, { "epoch": 0.67, "learning_rate": 5.1248305109243454e-06, "loss": 0.9752, "step": 7938 }, { "epoch": 0.67, "learning_rate": 5.122435717474938e-06, "loss": 0.9991, "step": 7939 }, { "epoch": 0.67, "learning_rate": 5.120041291023976e-06, "loss": 0.997, "step": 7940 }, { "epoch": 0.67, "learning_rate": 5.117647231751624e-06, "loss": 0.9699, "step": 7941 }, { "epoch": 0.67, "learning_rate": 5.115253539838005e-06, "loss": 0.9817, "step": 7942 }, { "epoch": 0.67, "learning_rate": 5.11286021546324e-06, "loss": 0.9907, "step": 7943 }, { "epoch": 0.67, "learning_rate": 5.110467258807394e-06, "loss": 1.0021, "step": 7944 }, { "epoch": 0.67, "learning_rate": 5.1080746700505315e-06, "loss": 1.0068, "step": 7945 }, { "epoch": 0.67, "learning_rate": 5.105682449372662e-06, "loss": 0.9993, "step": 7946 }, { "epoch": 0.67, "learning_rate": 5.103290596953795e-06, "loss": 0.9511, "step": 7947 }, { "epoch": 0.67, "learning_rate": 5.100899112973887e-06, "loss": 0.9782, "step": 7948 }, { "epoch": 0.67, "learning_rate": 5.098507997612887e-06, "loss": 0.9467, "step": 7949 }, { "epoch": 0.67, "learning_rate": 5.096117251050705e-06, "loss": 0.9431, "step": 7950 }, { "epoch": 0.67, "learning_rate": 5.093726873467222e-06, "loss": 0.9855, "step": 7951 }, { "epoch": 0.67, "learning_rate": 5.091336865042301e-06, "loss": 0.9761, "step": 7952 }, { "epoch": 0.67, "learning_rate": 5.088947225955769e-06, "loss": 0.9321, "step": 7953 }, { "epoch": 0.67, "learning_rate": 5.086557956387424e-06, "loss": 0.9893, "step": 7954 }, { "epoch": 0.67, "learning_rate": 5.084169056517047e-06, "loss": 1.024, "step": 7955 }, { "epoch": 0.67, "learning_rate": 5.081780526524376e-06, "loss": 0.908, "step": 7956 }, { "epoch": 0.67, "learning_rate": 5.0793923665891365e-06, "loss": 0.9786, "step": 7957 }, { "epoch": 0.67, "learning_rate": 5.077004576891011e-06, "loss": 0.9834, "step": 7958 }, { "epoch": 0.67, "learning_rate": 5.074617157609671e-06, "loss": 0.9807, "step": 7959 }, { "epoch": 0.67, "learning_rate": 5.072230108924744e-06, "loss": 1.0007, "step": 7960 }, { "epoch": 0.67, "learning_rate": 5.069843431015842e-06, "loss": 0.9746, "step": 7961 }, { "epoch": 0.67, "learning_rate": 5.067457124062537e-06, "loss": 0.9944, "step": 7962 }, { "epoch": 0.67, "learning_rate": 5.065071188244388e-06, "loss": 1.0419, "step": 7963 }, { "epoch": 0.67, "learning_rate": 5.06268562374091e-06, "loss": 1.0052, "step": 7964 }, { "epoch": 0.67, "learning_rate": 5.0603004307316016e-06, "loss": 1.0196, "step": 7965 }, { "epoch": 0.67, "learning_rate": 5.057915609395933e-06, "loss": 1.0039, "step": 7966 }, { "epoch": 0.67, "learning_rate": 5.055531159913338e-06, "loss": 1.0122, "step": 7967 }, { "epoch": 0.67, "learning_rate": 5.0531470824632324e-06, "loss": 0.9872, "step": 7968 }, { "epoch": 0.67, "learning_rate": 5.050763377224994e-06, "loss": 1.012, "step": 7969 }, { "epoch": 0.67, "learning_rate": 5.048380044377985e-06, "loss": 0.9986, "step": 7970 }, { "epoch": 0.68, "learning_rate": 5.045997084101525e-06, "loss": 0.9731, "step": 7971 }, { "epoch": 0.68, "learning_rate": 5.04361449657492e-06, "loss": 0.9776, "step": 7972 }, { "epoch": 0.68, "learning_rate": 5.041232281977433e-06, "loss": 0.9852, "step": 7973 }, { "epoch": 0.68, "learning_rate": 5.038850440488316e-06, "loss": 1.0083, "step": 7974 }, { "epoch": 0.68, "learning_rate": 5.036468972286775e-06, "loss": 0.9823, "step": 7975 }, { "epoch": 0.68, "learning_rate": 5.034087877552005e-06, "loss": 0.9609, "step": 7976 }, { "epoch": 0.68, "learning_rate": 5.031707156463162e-06, "loss": 1.0192, "step": 7977 }, { "epoch": 0.68, "learning_rate": 5.029326809199371e-06, "loss": 0.9618, "step": 7978 }, { "epoch": 0.68, "learning_rate": 5.026946835939742e-06, "loss": 1.0046, "step": 7979 }, { "epoch": 0.68, "learning_rate": 5.024567236863347e-06, "loss": 0.9807, "step": 7980 }, { "epoch": 0.68, "learning_rate": 5.022188012149227e-06, "loss": 0.9838, "step": 7981 }, { "epoch": 0.68, "learning_rate": 5.019809161976409e-06, "loss": 0.9634, "step": 7982 }, { "epoch": 0.68, "learning_rate": 5.017430686523874e-06, "loss": 1.0249, "step": 7983 }, { "epoch": 0.68, "learning_rate": 5.015052585970591e-06, "loss": 0.9499, "step": 7984 }, { "epoch": 0.68, "learning_rate": 5.0126748604954865e-06, "loss": 1.0138, "step": 7985 }, { "epoch": 0.68, "learning_rate": 5.010297510277473e-06, "loss": 1.0277, "step": 7986 }, { "epoch": 0.68, "learning_rate": 5.007920535495421e-06, "loss": 0.9428, "step": 7987 }, { "epoch": 0.68, "learning_rate": 5.005543936328185e-06, "loss": 0.9861, "step": 7988 }, { "epoch": 0.68, "learning_rate": 5.0031677129545794e-06, "loss": 1.0443, "step": 7989 }, { "epoch": 0.68, "learning_rate": 5.000791865553402e-06, "loss": 1.0246, "step": 7990 }, { "epoch": 0.68, "learning_rate": 4.9984163943034125e-06, "loss": 1.0226, "step": 7991 }, { "epoch": 0.68, "learning_rate": 4.996041299383346e-06, "loss": 0.9765, "step": 7992 }, { "epoch": 0.68, "learning_rate": 4.9936665809719166e-06, "loss": 1.0201, "step": 7993 }, { "epoch": 0.68, "learning_rate": 4.991292239247795e-06, "loss": 0.962, "step": 7994 }, { "epoch": 0.68, "learning_rate": 4.988918274389638e-06, "loss": 0.9518, "step": 7995 }, { "epoch": 0.68, "learning_rate": 4.986544686576062e-06, "loss": 1.0139, "step": 7996 }, { "epoch": 0.68, "learning_rate": 4.984171475985667e-06, "loss": 1.0124, "step": 7997 }, { "epoch": 0.68, "learning_rate": 4.981798642797013e-06, "loss": 1.0483, "step": 7998 }, { "epoch": 0.68, "learning_rate": 4.979426187188643e-06, "loss": 1.0089, "step": 7999 }, { "epoch": 0.68, "learning_rate": 4.977054109339058e-06, "loss": 1.031, "step": 8000 }, { "epoch": 0.68, "learning_rate": 4.974682409426745e-06, "loss": 1.0142, "step": 8001 }, { "epoch": 0.68, "learning_rate": 4.972311087630152e-06, "loss": 1.0003, "step": 8002 }, { "epoch": 0.68, "learning_rate": 4.969940144127705e-06, "loss": 0.8894, "step": 8003 }, { "epoch": 0.68, "learning_rate": 4.9675695790978e-06, "loss": 0.9808, "step": 8004 }, { "epoch": 0.68, "learning_rate": 4.9651993927187945e-06, "loss": 0.982, "step": 8005 }, { "epoch": 0.68, "learning_rate": 4.9628295851690375e-06, "loss": 1.0122, "step": 8006 }, { "epoch": 0.68, "learning_rate": 4.960460156626835e-06, "loss": 0.9965, "step": 8007 }, { "epoch": 0.68, "learning_rate": 4.958091107270462e-06, "loss": 0.9991, "step": 8008 }, { "epoch": 0.68, "learning_rate": 4.955722437278179e-06, "loss": 0.9818, "step": 8009 }, { "epoch": 0.68, "learning_rate": 4.953354146828203e-06, "loss": 0.9532, "step": 8010 }, { "epoch": 0.68, "learning_rate": 4.950986236098737e-06, "loss": 0.9587, "step": 8011 }, { "epoch": 0.68, "learning_rate": 4.948618705267939e-06, "loss": 0.9858, "step": 8012 }, { "epoch": 0.68, "learning_rate": 4.946251554513955e-06, "loss": 1.0058, "step": 8013 }, { "epoch": 0.68, "learning_rate": 4.9438847840148875e-06, "loss": 1.0049, "step": 8014 }, { "epoch": 0.68, "learning_rate": 4.9415183939488244e-06, "loss": 1.0208, "step": 8015 }, { "epoch": 0.68, "learning_rate": 4.939152384493812e-06, "loss": 1.0047, "step": 8016 }, { "epoch": 0.68, "learning_rate": 4.936786755827875e-06, "loss": 1.0212, "step": 8017 }, { "epoch": 0.68, "learning_rate": 4.9344215081290135e-06, "loss": 0.9304, "step": 8018 }, { "epoch": 0.68, "learning_rate": 4.932056641575188e-06, "loss": 1.0338, "step": 8019 }, { "epoch": 0.68, "learning_rate": 4.929692156344341e-06, "loss": 0.9599, "step": 8020 }, { "epoch": 0.68, "learning_rate": 4.927328052614375e-06, "loss": 0.9579, "step": 8021 }, { "epoch": 0.68, "learning_rate": 4.924964330563178e-06, "loss": 0.9976, "step": 8022 }, { "epoch": 0.68, "learning_rate": 4.9226009903685945e-06, "loss": 0.9623, "step": 8023 }, { "epoch": 0.68, "learning_rate": 4.920238032208453e-06, "loss": 0.9478, "step": 8024 }, { "epoch": 0.68, "learning_rate": 4.917875456260542e-06, "loss": 1.0049, "step": 8025 }, { "epoch": 0.68, "learning_rate": 4.915513262702634e-06, "loss": 1.0093, "step": 8026 }, { "epoch": 0.68, "learning_rate": 4.913151451712457e-06, "loss": 1.0119, "step": 8027 }, { "epoch": 0.68, "learning_rate": 4.910790023467726e-06, "loss": 0.9805, "step": 8028 }, { "epoch": 0.68, "learning_rate": 4.9084289781461184e-06, "loss": 1.0095, "step": 8029 }, { "epoch": 0.68, "learning_rate": 4.9060683159252775e-06, "loss": 1.0022, "step": 8030 }, { "epoch": 0.68, "learning_rate": 4.903708036982835e-06, "loss": 0.998, "step": 8031 }, { "epoch": 0.68, "learning_rate": 4.901348141496375e-06, "loss": 1.002, "step": 8032 }, { "epoch": 0.68, "learning_rate": 4.898988629643467e-06, "loss": 0.9736, "step": 8033 }, { "epoch": 0.68, "learning_rate": 4.896629501601644e-06, "loss": 0.9609, "step": 8034 }, { "epoch": 0.68, "learning_rate": 4.8942707575484085e-06, "loss": 0.9779, "step": 8035 }, { "epoch": 0.68, "learning_rate": 4.891912397661243e-06, "loss": 0.9901, "step": 8036 }, { "epoch": 0.68, "learning_rate": 4.889554422117588e-06, "loss": 1.0264, "step": 8037 }, { "epoch": 0.68, "learning_rate": 4.887196831094872e-06, "loss": 0.9929, "step": 8038 }, { "epoch": 0.68, "learning_rate": 4.884839624770478e-06, "loss": 1.0208, "step": 8039 }, { "epoch": 0.68, "learning_rate": 4.8824828033217715e-06, "loss": 1.0158, "step": 8040 }, { "epoch": 0.68, "learning_rate": 4.880126366926081e-06, "loss": 0.9509, "step": 8041 }, { "epoch": 0.68, "learning_rate": 4.877770315760716e-06, "loss": 0.9831, "step": 8042 }, { "epoch": 0.68, "learning_rate": 4.875414650002942e-06, "loss": 0.9297, "step": 8043 }, { "epoch": 0.68, "learning_rate": 4.8730593698300096e-06, "loss": 1.0296, "step": 8044 }, { "epoch": 0.68, "learning_rate": 4.870704475419139e-06, "loss": 0.8538, "step": 8045 }, { "epoch": 0.68, "learning_rate": 4.8683499669475095e-06, "loss": 0.9564, "step": 8046 }, { "epoch": 0.68, "learning_rate": 4.865995844592287e-06, "loss": 0.9853, "step": 8047 }, { "epoch": 0.68, "learning_rate": 4.863642108530594e-06, "loss": 1.0251, "step": 8048 }, { "epoch": 0.68, "learning_rate": 4.861288758939536e-06, "loss": 1.0087, "step": 8049 }, { "epoch": 0.68, "learning_rate": 4.858935795996179e-06, "loss": 1.0286, "step": 8050 }, { "epoch": 0.68, "learning_rate": 4.856583219877572e-06, "loss": 0.9687, "step": 8051 }, { "epoch": 0.68, "learning_rate": 4.854231030760719e-06, "loss": 0.9821, "step": 8052 }, { "epoch": 0.68, "learning_rate": 4.851879228822613e-06, "loss": 0.9793, "step": 8053 }, { "epoch": 0.68, "learning_rate": 4.849527814240199e-06, "loss": 0.9686, "step": 8054 }, { "epoch": 0.68, "learning_rate": 4.847176787190413e-06, "loss": 0.9826, "step": 8055 }, { "epoch": 0.68, "learning_rate": 4.844826147850145e-06, "loss": 1.0461, "step": 8056 }, { "epoch": 0.68, "learning_rate": 4.842475896396259e-06, "loss": 1.0308, "step": 8057 }, { "epoch": 0.68, "learning_rate": 4.840126033005602e-06, "loss": 0.9587, "step": 8058 }, { "epoch": 0.68, "learning_rate": 4.837776557854977e-06, "loss": 1.0079, "step": 8059 }, { "epoch": 0.68, "learning_rate": 4.835427471121161e-06, "loss": 0.9916, "step": 8060 }, { "epoch": 0.68, "learning_rate": 4.833078772980912e-06, "loss": 1.0258, "step": 8061 }, { "epoch": 0.68, "learning_rate": 4.830730463610942e-06, "loss": 1.0464, "step": 8062 }, { "epoch": 0.68, "learning_rate": 4.828382543187952e-06, "loss": 0.999, "step": 8063 }, { "epoch": 0.68, "learning_rate": 4.826035011888597e-06, "loss": 0.9942, "step": 8064 }, { "epoch": 0.68, "learning_rate": 4.823687869889518e-06, "loss": 0.9984, "step": 8065 }, { "epoch": 0.68, "learning_rate": 4.821341117367309e-06, "loss": 1.0311, "step": 8066 }, { "epoch": 0.68, "learning_rate": 4.818994754498555e-06, "loss": 1.0132, "step": 8067 }, { "epoch": 0.68, "learning_rate": 4.816648781459794e-06, "loss": 0.9878, "step": 8068 }, { "epoch": 0.68, "learning_rate": 4.814303198427544e-06, "loss": 0.9503, "step": 8069 }, { "epoch": 0.68, "learning_rate": 4.811958005578296e-06, "loss": 0.9675, "step": 8070 }, { "epoch": 0.68, "learning_rate": 4.809613203088502e-06, "loss": 1.0453, "step": 8071 }, { "epoch": 0.68, "learning_rate": 4.807268791134592e-06, "loss": 1.0662, "step": 8072 }, { "epoch": 0.68, "learning_rate": 4.804924769892965e-06, "loss": 1.0, "step": 8073 }, { "epoch": 0.68, "learning_rate": 4.802581139539991e-06, "loss": 0.9825, "step": 8074 }, { "epoch": 0.68, "learning_rate": 4.800237900252005e-06, "loss": 1.0123, "step": 8075 }, { "epoch": 0.68, "learning_rate": 4.797895052205325e-06, "loss": 1.0148, "step": 8076 }, { "epoch": 0.68, "learning_rate": 4.795552595576224e-06, "loss": 1.0375, "step": 8077 }, { "epoch": 0.68, "learning_rate": 4.793210530540962e-06, "loss": 1.0439, "step": 8078 }, { "epoch": 0.68, "learning_rate": 4.790868857275751e-06, "loss": 0.9572, "step": 8079 }, { "epoch": 0.68, "learning_rate": 4.788527575956794e-06, "loss": 1.0124, "step": 8080 }, { "epoch": 0.68, "learning_rate": 4.7861866867602445e-06, "loss": 0.9821, "step": 8081 }, { "epoch": 0.68, "learning_rate": 4.783846189862244e-06, "loss": 0.9867, "step": 8082 }, { "epoch": 0.68, "learning_rate": 4.781506085438894e-06, "loss": 0.9297, "step": 8083 }, { "epoch": 0.68, "learning_rate": 4.779166373666263e-06, "loss": 1.0112, "step": 8084 }, { "epoch": 0.68, "learning_rate": 4.776827054720406e-06, "loss": 0.9847, "step": 8085 }, { "epoch": 0.68, "learning_rate": 4.774488128777332e-06, "loss": 0.9662, "step": 8086 }, { "epoch": 0.68, "learning_rate": 4.772149596013026e-06, "loss": 0.978, "step": 8087 }, { "epoch": 0.68, "learning_rate": 4.769811456603449e-06, "loss": 0.9543, "step": 8088 }, { "epoch": 0.69, "learning_rate": 4.767473710724522e-06, "loss": 1.0171, "step": 8089 }, { "epoch": 0.69, "learning_rate": 4.765136358552148e-06, "loss": 1.0372, "step": 8090 }, { "epoch": 0.69, "learning_rate": 4.762799400262189e-06, "loss": 0.9806, "step": 8091 }, { "epoch": 0.69, "learning_rate": 4.760462836030488e-06, "loss": 0.9631, "step": 8092 }, { "epoch": 0.69, "learning_rate": 4.758126666032847e-06, "loss": 0.8136, "step": 8093 }, { "epoch": 0.69, "learning_rate": 4.755790890445052e-06, "loss": 1.013, "step": 8094 }, { "epoch": 0.69, "learning_rate": 4.753455509442844e-06, "loss": 0.9142, "step": 8095 }, { "epoch": 0.69, "learning_rate": 4.751120523201944e-06, "loss": 1.0017, "step": 8096 }, { "epoch": 0.69, "learning_rate": 4.748785931898049e-06, "loss": 0.9781, "step": 8097 }, { "epoch": 0.69, "learning_rate": 4.7464517357068085e-06, "loss": 0.9067, "step": 8098 }, { "epoch": 0.69, "learning_rate": 4.744117934803861e-06, "loss": 1.0378, "step": 8099 }, { "epoch": 0.69, "learning_rate": 4.741784529364798e-06, "loss": 1.0225, "step": 8100 }, { "epoch": 0.69, "learning_rate": 4.739451519565198e-06, "loss": 0.971, "step": 8101 }, { "epoch": 0.69, "learning_rate": 4.737118905580595e-06, "loss": 1.0544, "step": 8102 }, { "epoch": 0.69, "learning_rate": 4.7347866875865064e-06, "loss": 0.9937, "step": 8103 }, { "epoch": 0.69, "learning_rate": 4.732454865758407e-06, "loss": 0.9961, "step": 8104 }, { "epoch": 0.69, "learning_rate": 4.730123440271753e-06, "loss": 1.0199, "step": 8105 }, { "epoch": 0.69, "learning_rate": 4.7277924113019626e-06, "loss": 1.0058, "step": 8106 }, { "epoch": 0.69, "learning_rate": 4.72546177902443e-06, "loss": 0.9536, "step": 8107 }, { "epoch": 0.69, "learning_rate": 4.723131543614514e-06, "loss": 0.9869, "step": 8108 }, { "epoch": 0.69, "learning_rate": 4.720801705247552e-06, "loss": 0.9751, "step": 8109 }, { "epoch": 0.69, "learning_rate": 4.718472264098841e-06, "loss": 1.0222, "step": 8110 }, { "epoch": 0.69, "learning_rate": 4.7161432203436516e-06, "loss": 0.9801, "step": 8111 }, { "epoch": 0.69, "learning_rate": 4.713814574157232e-06, "loss": 0.987, "step": 8112 }, { "epoch": 0.69, "learning_rate": 4.7114863257147924e-06, "loss": 0.9815, "step": 8113 }, { "epoch": 0.69, "learning_rate": 4.709158475191511e-06, "loss": 0.9805, "step": 8114 }, { "epoch": 0.69, "learning_rate": 4.706831022762547e-06, "loss": 0.9868, "step": 8115 }, { "epoch": 0.69, "learning_rate": 4.704503968603016e-06, "loss": 0.9595, "step": 8116 }, { "epoch": 0.69, "learning_rate": 4.702177312888017e-06, "loss": 0.9719, "step": 8117 }, { "epoch": 0.69, "learning_rate": 4.6998510557926065e-06, "loss": 1.0485, "step": 8118 }, { "epoch": 0.69, "learning_rate": 4.697525197491825e-06, "loss": 0.9474, "step": 8119 }, { "epoch": 0.69, "learning_rate": 4.695199738160665e-06, "loss": 1.0386, "step": 8120 }, { "epoch": 0.69, "learning_rate": 4.692874677974106e-06, "loss": 0.9453, "step": 8121 }, { "epoch": 0.69, "learning_rate": 4.690550017107092e-06, "loss": 0.9787, "step": 8122 }, { "epoch": 0.69, "learning_rate": 4.688225755734528e-06, "loss": 0.9925, "step": 8123 }, { "epoch": 0.69, "learning_rate": 4.685901894031306e-06, "loss": 0.9802, "step": 8124 }, { "epoch": 0.69, "learning_rate": 4.683578432172269e-06, "loss": 1.0361, "step": 8125 }, { "epoch": 0.69, "learning_rate": 4.681255370332247e-06, "loss": 1.0294, "step": 8126 }, { "epoch": 0.69, "learning_rate": 4.678932708686026e-06, "loss": 0.9977, "step": 8127 }, { "epoch": 0.69, "learning_rate": 4.676610447408374e-06, "loss": 0.9898, "step": 8128 }, { "epoch": 0.69, "learning_rate": 4.6742885866740165e-06, "loss": 0.9856, "step": 8129 }, { "epoch": 0.69, "learning_rate": 4.6719671266576625e-06, "loss": 0.9324, "step": 8130 }, { "epoch": 0.69, "learning_rate": 4.669646067533975e-06, "loss": 0.9607, "step": 8131 }, { "epoch": 0.69, "learning_rate": 4.667325409477604e-06, "loss": 0.9609, "step": 8132 }, { "epoch": 0.69, "learning_rate": 4.665005152663154e-06, "loss": 0.9654, "step": 8133 }, { "epoch": 0.69, "learning_rate": 4.662685297265213e-06, "loss": 0.9894, "step": 8134 }, { "epoch": 0.69, "learning_rate": 4.660365843458325e-06, "loss": 1.0271, "step": 8135 }, { "epoch": 0.69, "learning_rate": 4.658046791417016e-06, "loss": 0.9789, "step": 8136 }, { "epoch": 0.69, "learning_rate": 4.655728141315775e-06, "loss": 1.0079, "step": 8137 }, { "epoch": 0.69, "learning_rate": 4.6534098933290575e-06, "loss": 0.8529, "step": 8138 }, { "epoch": 0.69, "learning_rate": 4.651092047631301e-06, "loss": 1.0281, "step": 8139 }, { "epoch": 0.69, "learning_rate": 4.648774604396903e-06, "loss": 0.9548, "step": 8140 }, { "epoch": 0.69, "learning_rate": 4.6464575638002266e-06, "loss": 1.0023, "step": 8141 }, { "epoch": 0.69, "learning_rate": 4.644140926015619e-06, "loss": 0.9842, "step": 8142 }, { "epoch": 0.69, "learning_rate": 4.6418246912173834e-06, "loss": 0.9974, "step": 8143 }, { "epoch": 0.69, "learning_rate": 4.6395088595798045e-06, "loss": 1.0157, "step": 8144 }, { "epoch": 0.69, "learning_rate": 4.637193431277123e-06, "loss": 1.0162, "step": 8145 }, { "epoch": 0.69, "learning_rate": 4.634878406483564e-06, "loss": 0.8743, "step": 8146 }, { "epoch": 0.69, "learning_rate": 4.6325637853733075e-06, "loss": 0.9895, "step": 8147 }, { "epoch": 0.69, "learning_rate": 4.630249568120514e-06, "loss": 0.9979, "step": 8148 }, { "epoch": 0.69, "learning_rate": 4.627935754899315e-06, "loss": 0.9676, "step": 8149 }, { "epoch": 0.69, "learning_rate": 4.625622345883799e-06, "loss": 0.9512, "step": 8150 }, { "epoch": 0.69, "learning_rate": 4.623309341248039e-06, "loss": 0.9273, "step": 8151 }, { "epoch": 0.69, "learning_rate": 4.620996741166064e-06, "loss": 1.0343, "step": 8152 }, { "epoch": 0.69, "learning_rate": 4.618684545811884e-06, "loss": 1.0296, "step": 8153 }, { "epoch": 0.69, "learning_rate": 4.616372755359468e-06, "loss": 0.9453, "step": 8154 }, { "epoch": 0.69, "learning_rate": 4.614061369982769e-06, "loss": 1.0074, "step": 8155 }, { "epoch": 0.69, "learning_rate": 4.61175038985569e-06, "loss": 1.033, "step": 8156 }, { "epoch": 0.69, "learning_rate": 4.6094398151521224e-06, "loss": 0.905, "step": 8157 }, { "epoch": 0.69, "learning_rate": 4.607129646045912e-06, "loss": 0.9762, "step": 8158 }, { "epoch": 0.69, "learning_rate": 4.60481988271089e-06, "loss": 0.9236, "step": 8159 }, { "epoch": 0.69, "learning_rate": 4.602510525320837e-06, "loss": 1.0035, "step": 8160 }, { "epoch": 0.69, "learning_rate": 4.600201574049523e-06, "loss": 0.9879, "step": 8161 }, { "epoch": 0.69, "learning_rate": 4.597893029070673e-06, "loss": 0.9654, "step": 8162 }, { "epoch": 0.69, "learning_rate": 4.595584890557991e-06, "loss": 0.9222, "step": 8163 }, { "epoch": 0.69, "learning_rate": 4.5932771586851455e-06, "loss": 1.0347, "step": 8164 }, { "epoch": 0.69, "learning_rate": 4.59096983362577e-06, "loss": 0.9729, "step": 8165 }, { "epoch": 0.69, "learning_rate": 4.58866291555348e-06, "loss": 0.9962, "step": 8166 }, { "epoch": 0.69, "learning_rate": 4.58635640464185e-06, "loss": 1.0242, "step": 8167 }, { "epoch": 0.69, "learning_rate": 4.584050301064422e-06, "loss": 1.0148, "step": 8168 }, { "epoch": 0.69, "learning_rate": 4.58174460499472e-06, "loss": 0.9688, "step": 8169 }, { "epoch": 0.69, "learning_rate": 4.579439316606224e-06, "loss": 0.9983, "step": 8170 }, { "epoch": 0.69, "learning_rate": 4.577134436072393e-06, "loss": 0.9811, "step": 8171 }, { "epoch": 0.69, "learning_rate": 4.574829963566646e-06, "loss": 0.9866, "step": 8172 }, { "epoch": 0.69, "learning_rate": 4.57252589926238e-06, "loss": 1.0568, "step": 8173 }, { "epoch": 0.69, "learning_rate": 4.570222243332961e-06, "loss": 1.0297, "step": 8174 }, { "epoch": 0.69, "learning_rate": 4.567918995951715e-06, "loss": 0.9982, "step": 8175 }, { "epoch": 0.69, "learning_rate": 4.565616157291948e-06, "loss": 0.9495, "step": 8176 }, { "epoch": 0.69, "learning_rate": 4.5633137275269235e-06, "loss": 1.0033, "step": 8177 }, { "epoch": 0.69, "learning_rate": 4.561011706829891e-06, "loss": 0.9533, "step": 8178 }, { "epoch": 0.69, "learning_rate": 4.558710095374051e-06, "loss": 0.9878, "step": 8179 }, { "epoch": 0.69, "learning_rate": 4.556408893332589e-06, "loss": 1.0097, "step": 8180 }, { "epoch": 0.69, "learning_rate": 4.5541081008786445e-06, "loss": 1.0122, "step": 8181 }, { "epoch": 0.69, "learning_rate": 4.551807718185342e-06, "loss": 0.9923, "step": 8182 }, { "epoch": 0.69, "learning_rate": 4.549507745425761e-06, "loss": 1.0069, "step": 8183 }, { "epoch": 0.69, "learning_rate": 4.5472081827729604e-06, "loss": 1.0634, "step": 8184 }, { "epoch": 0.69, "learning_rate": 4.544909030399961e-06, "loss": 1.0071, "step": 8185 }, { "epoch": 0.69, "learning_rate": 4.542610288479761e-06, "loss": 0.9969, "step": 8186 }, { "epoch": 0.69, "learning_rate": 4.540311957185315e-06, "loss": 0.9543, "step": 8187 }, { "epoch": 0.69, "learning_rate": 4.538014036689563e-06, "loss": 0.964, "step": 8188 }, { "epoch": 0.69, "learning_rate": 4.535716527165397e-06, "loss": 0.9761, "step": 8189 }, { "epoch": 0.69, "learning_rate": 4.5334194287856945e-06, "loss": 0.9661, "step": 8190 }, { "epoch": 0.69, "learning_rate": 4.5311227417232915e-06, "loss": 1.0383, "step": 8191 }, { "epoch": 0.69, "learning_rate": 4.528826466150991e-06, "loss": 0.9953, "step": 8192 }, { "epoch": 0.69, "learning_rate": 4.526530602241577e-06, "loss": 1.0242, "step": 8193 }, { "epoch": 0.69, "learning_rate": 4.524235150167791e-06, "loss": 1.0114, "step": 8194 }, { "epoch": 0.69, "learning_rate": 4.521940110102344e-06, "loss": 0.9213, "step": 8195 }, { "epoch": 0.69, "learning_rate": 4.519645482217929e-06, "loss": 0.9692, "step": 8196 }, { "epoch": 0.69, "learning_rate": 4.5173512666871895e-06, "loss": 0.9201, "step": 8197 }, { "epoch": 0.69, "learning_rate": 4.515057463682757e-06, "loss": 0.9668, "step": 8198 }, { "epoch": 0.69, "learning_rate": 4.5127640733772115e-06, "loss": 0.9816, "step": 8199 }, { "epoch": 0.69, "learning_rate": 4.510471095943119e-06, "loss": 0.958, "step": 8200 }, { "epoch": 0.69, "learning_rate": 4.5081785315530105e-06, "loss": 0.9918, "step": 8201 }, { "epoch": 0.69, "learning_rate": 4.505886380379375e-06, "loss": 0.9525, "step": 8202 }, { "epoch": 0.69, "learning_rate": 4.5035946425946885e-06, "loss": 0.9596, "step": 8203 }, { "epoch": 0.69, "learning_rate": 4.501303318371378e-06, "loss": 0.9683, "step": 8204 }, { "epoch": 0.69, "learning_rate": 4.499012407881856e-06, "loss": 1.0262, "step": 8205 }, { "epoch": 0.69, "learning_rate": 4.496721911298485e-06, "loss": 0.9677, "step": 8206 }, { "epoch": 0.7, "learning_rate": 4.494431828793617e-06, "loss": 0.9701, "step": 8207 }, { "epoch": 0.7, "learning_rate": 4.492142160539555e-06, "loss": 0.9881, "step": 8208 }, { "epoch": 0.7, "learning_rate": 4.489852906708587e-06, "loss": 0.9305, "step": 8209 }, { "epoch": 0.7, "learning_rate": 4.487564067472951e-06, "loss": 0.9734, "step": 8210 }, { "epoch": 0.7, "learning_rate": 4.485275643004875e-06, "loss": 0.9898, "step": 8211 }, { "epoch": 0.7, "learning_rate": 4.482987633476535e-06, "loss": 0.9897, "step": 8212 }, { "epoch": 0.7, "learning_rate": 4.480700039060095e-06, "loss": 1.0055, "step": 8213 }, { "epoch": 0.7, "learning_rate": 4.478412859927669e-06, "loss": 0.9991, "step": 8214 }, { "epoch": 0.7, "learning_rate": 4.476126096251359e-06, "loss": 1.0001, "step": 8215 }, { "epoch": 0.7, "learning_rate": 4.473839748203217e-06, "loss": 0.9652, "step": 8216 }, { "epoch": 0.7, "learning_rate": 4.471553815955281e-06, "loss": 0.9806, "step": 8217 }, { "epoch": 0.7, "learning_rate": 4.469268299679544e-06, "loss": 0.9448, "step": 8218 }, { "epoch": 0.7, "learning_rate": 4.466983199547973e-06, "loss": 0.869, "step": 8219 }, { "epoch": 0.7, "learning_rate": 4.464698515732509e-06, "loss": 0.9858, "step": 8220 }, { "epoch": 0.7, "learning_rate": 4.462414248405053e-06, "loss": 0.8411, "step": 8221 }, { "epoch": 0.7, "learning_rate": 4.460130397737473e-06, "loss": 1.035, "step": 8222 }, { "epoch": 0.7, "learning_rate": 4.457846963901621e-06, "loss": 0.999, "step": 8223 }, { "epoch": 0.7, "learning_rate": 4.455563947069298e-06, "loss": 0.8667, "step": 8224 }, { "epoch": 0.7, "learning_rate": 4.453281347412289e-06, "loss": 0.996, "step": 8225 }, { "epoch": 0.7, "learning_rate": 4.450999165102343e-06, "loss": 0.9729, "step": 8226 }, { "epoch": 0.7, "learning_rate": 4.44871740031117e-06, "loss": 1.0076, "step": 8227 }, { "epoch": 0.7, "learning_rate": 4.446436053210464e-06, "loss": 1.0036, "step": 8228 }, { "epoch": 0.7, "learning_rate": 4.4441551239718675e-06, "loss": 0.988, "step": 8229 }, { "epoch": 0.7, "learning_rate": 4.441874612767012e-06, "loss": 0.9302, "step": 8230 }, { "epoch": 0.7, "learning_rate": 4.43959451976748e-06, "loss": 0.9944, "step": 8231 }, { "epoch": 0.7, "learning_rate": 4.43731484514484e-06, "loss": 1.0223, "step": 8232 }, { "epoch": 0.7, "learning_rate": 4.43503558907061e-06, "loss": 1.0128, "step": 8233 }, { "epoch": 0.7, "learning_rate": 4.432756751716294e-06, "loss": 0.8956, "step": 8234 }, { "epoch": 0.7, "learning_rate": 4.4304783332533495e-06, "loss": 0.9681, "step": 8235 }, { "epoch": 0.7, "learning_rate": 4.428200333853217e-06, "loss": 1.0026, "step": 8236 }, { "epoch": 0.7, "learning_rate": 4.42592275368729e-06, "loss": 0.9824, "step": 8237 }, { "epoch": 0.7, "learning_rate": 4.4236455929269485e-06, "loss": 0.9545, "step": 8238 }, { "epoch": 0.7, "learning_rate": 4.421368851743521e-06, "loss": 0.9723, "step": 8239 }, { "epoch": 0.7, "learning_rate": 4.419092530308321e-06, "loss": 0.8466, "step": 8240 }, { "epoch": 0.7, "learning_rate": 4.416816628792619e-06, "loss": 1.0249, "step": 8241 }, { "epoch": 0.7, "learning_rate": 4.414541147367665e-06, "loss": 0.9876, "step": 8242 }, { "epoch": 0.7, "learning_rate": 4.4122660862046644e-06, "loss": 0.9841, "step": 8243 }, { "epoch": 0.7, "learning_rate": 4.409991445474804e-06, "loss": 0.9725, "step": 8244 }, { "epoch": 0.7, "learning_rate": 4.407717225349229e-06, "loss": 0.9527, "step": 8245 }, { "epoch": 0.7, "learning_rate": 4.405443425999053e-06, "loss": 0.9517, "step": 8246 }, { "epoch": 0.7, "learning_rate": 4.403170047595371e-06, "loss": 0.8103, "step": 8247 }, { "epoch": 0.7, "learning_rate": 4.40089709030923e-06, "loss": 0.989, "step": 8248 }, { "epoch": 0.7, "learning_rate": 4.39862455431165e-06, "loss": 0.9991, "step": 8249 }, { "epoch": 0.7, "learning_rate": 4.39635243977363e-06, "loss": 0.9872, "step": 8250 }, { "epoch": 0.7, "learning_rate": 4.394080746866119e-06, "loss": 0.9831, "step": 8251 }, { "epoch": 0.7, "learning_rate": 4.39180947576005e-06, "loss": 1.0304, "step": 8252 }, { "epoch": 0.7, "learning_rate": 4.389538626626322e-06, "loss": 0.9913, "step": 8253 }, { "epoch": 0.7, "learning_rate": 4.387268199635789e-06, "loss": 0.9851, "step": 8254 }, { "epoch": 0.7, "learning_rate": 4.384998194959292e-06, "loss": 0.9988, "step": 8255 }, { "epoch": 0.7, "learning_rate": 4.382728612767624e-06, "loss": 0.9919, "step": 8256 }, { "epoch": 0.7, "learning_rate": 4.3804594532315585e-06, "loss": 0.9753, "step": 8257 }, { "epoch": 0.7, "learning_rate": 4.378190716521826e-06, "loss": 1.0342, "step": 8258 }, { "epoch": 0.7, "learning_rate": 4.375922402809139e-06, "loss": 0.9347, "step": 8259 }, { "epoch": 0.7, "learning_rate": 4.373654512264162e-06, "loss": 0.9753, "step": 8260 }, { "epoch": 0.7, "learning_rate": 4.371387045057544e-06, "loss": 0.9799, "step": 8261 }, { "epoch": 0.7, "learning_rate": 4.369120001359886e-06, "loss": 0.9651, "step": 8262 }, { "epoch": 0.7, "learning_rate": 4.3668533813417725e-06, "loss": 0.936, "step": 8263 }, { "epoch": 0.7, "learning_rate": 4.364587185173742e-06, "loss": 0.9364, "step": 8264 }, { "epoch": 0.7, "learning_rate": 4.362321413026317e-06, "loss": 1.0136, "step": 8265 }, { "epoch": 0.7, "learning_rate": 4.3600560650699685e-06, "loss": 0.9882, "step": 8266 }, { "epoch": 0.7, "learning_rate": 4.357791141475156e-06, "loss": 0.9974, "step": 8267 }, { "epoch": 0.7, "learning_rate": 4.355526642412289e-06, "loss": 1.0286, "step": 8268 }, { "epoch": 0.7, "learning_rate": 4.3532625680517605e-06, "loss": 1.0276, "step": 8269 }, { "epoch": 0.7, "learning_rate": 4.35099891856392e-06, "loss": 0.9886, "step": 8270 }, { "epoch": 0.7, "learning_rate": 4.348735694119087e-06, "loss": 0.9757, "step": 8271 }, { "epoch": 0.7, "learning_rate": 4.346472894887559e-06, "loss": 1.003, "step": 8272 }, { "epoch": 0.7, "learning_rate": 4.3442105210395854e-06, "loss": 1.0083, "step": 8273 }, { "epoch": 0.7, "learning_rate": 4.3419485727453995e-06, "loss": 1.0266, "step": 8274 }, { "epoch": 0.7, "learning_rate": 4.339687050175192e-06, "loss": 0.9683, "step": 8275 }, { "epoch": 0.7, "learning_rate": 4.337425953499121e-06, "loss": 0.9788, "step": 8276 }, { "epoch": 0.7, "learning_rate": 4.335165282887319e-06, "loss": 0.9412, "step": 8277 }, { "epoch": 0.7, "learning_rate": 4.332905038509888e-06, "loss": 0.8344, "step": 8278 }, { "epoch": 0.7, "learning_rate": 4.330645220536887e-06, "loss": 0.9807, "step": 8279 }, { "epoch": 0.7, "learning_rate": 4.328385829138357e-06, "loss": 0.9979, "step": 8280 }, { "epoch": 0.7, "learning_rate": 4.326126864484291e-06, "loss": 0.9814, "step": 8281 }, { "epoch": 0.7, "learning_rate": 4.323868326744667e-06, "loss": 1.0372, "step": 8282 }, { "epoch": 0.7, "learning_rate": 4.321610216089414e-06, "loss": 1.0077, "step": 8283 }, { "epoch": 0.7, "learning_rate": 4.319352532688444e-06, "loss": 1.0267, "step": 8284 }, { "epoch": 0.7, "learning_rate": 4.317095276711623e-06, "loss": 0.9744, "step": 8285 }, { "epoch": 0.7, "learning_rate": 4.3148384483288e-06, "loss": 0.8392, "step": 8286 }, { "epoch": 0.7, "learning_rate": 4.312582047709776e-06, "loss": 1.0154, "step": 8287 }, { "epoch": 0.7, "learning_rate": 4.3103260750243334e-06, "loss": 0.9051, "step": 8288 }, { "epoch": 0.7, "learning_rate": 4.308070530442211e-06, "loss": 0.9319, "step": 8289 }, { "epoch": 0.7, "learning_rate": 4.305815414133126e-06, "loss": 0.9889, "step": 8290 }, { "epoch": 0.7, "learning_rate": 4.303560726266753e-06, "loss": 1.0039, "step": 8291 }, { "epoch": 0.7, "learning_rate": 4.301306467012747e-06, "loss": 0.9967, "step": 8292 }, { "epoch": 0.7, "learning_rate": 4.299052636540715e-06, "loss": 0.9604, "step": 8293 }, { "epoch": 0.7, "learning_rate": 4.296799235020247e-06, "loss": 1.0328, "step": 8294 }, { "epoch": 0.7, "learning_rate": 4.294546262620887e-06, "loss": 1.039, "step": 8295 }, { "epoch": 0.7, "learning_rate": 4.2922937195121595e-06, "loss": 1.0313, "step": 8296 }, { "epoch": 0.7, "learning_rate": 4.29004160586355e-06, "loss": 0.995, "step": 8297 }, { "epoch": 0.7, "learning_rate": 4.287789921844506e-06, "loss": 0.9753, "step": 8298 }, { "epoch": 0.7, "learning_rate": 4.285538667624458e-06, "loss": 0.9559, "step": 8299 }, { "epoch": 0.7, "learning_rate": 4.283287843372791e-06, "loss": 1.0393, "step": 8300 }, { "epoch": 0.7, "learning_rate": 4.281037449258858e-06, "loss": 0.8615, "step": 8301 }, { "epoch": 0.7, "learning_rate": 4.278787485451988e-06, "loss": 0.9397, "step": 8302 }, { "epoch": 0.7, "learning_rate": 4.276537952121475e-06, "loss": 0.9831, "step": 8303 }, { "epoch": 0.7, "learning_rate": 4.274288849436573e-06, "loss": 0.9632, "step": 8304 }, { "epoch": 0.7, "learning_rate": 4.272040177566515e-06, "loss": 1.0271, "step": 8305 }, { "epoch": 0.7, "learning_rate": 4.269791936680492e-06, "loss": 0.9815, "step": 8306 }, { "epoch": 0.7, "learning_rate": 4.26754412694767e-06, "loss": 0.8388, "step": 8307 }, { "epoch": 0.7, "learning_rate": 4.265296748537174e-06, "loss": 1.0409, "step": 8308 }, { "epoch": 0.7, "learning_rate": 4.263049801618107e-06, "loss": 0.941, "step": 8309 }, { "epoch": 0.7, "learning_rate": 4.260803286359531e-06, "loss": 0.9516, "step": 8310 }, { "epoch": 0.7, "learning_rate": 4.25855720293048e-06, "loss": 0.9569, "step": 8311 }, { "epoch": 0.7, "learning_rate": 4.2563115514999505e-06, "loss": 0.9356, "step": 8312 }, { "epoch": 0.7, "learning_rate": 4.254066332236918e-06, "loss": 0.9828, "step": 8313 }, { "epoch": 0.7, "learning_rate": 4.251821545310308e-06, "loss": 0.9658, "step": 8314 }, { "epoch": 0.7, "learning_rate": 4.249577190889032e-06, "loss": 0.9744, "step": 8315 }, { "epoch": 0.7, "learning_rate": 4.247333269141952e-06, "loss": 1.0213, "step": 8316 }, { "epoch": 0.7, "learning_rate": 4.245089780237913e-06, "loss": 0.9917, "step": 8317 }, { "epoch": 0.7, "learning_rate": 4.242846724345713e-06, "loss": 0.9812, "step": 8318 }, { "epoch": 0.7, "learning_rate": 4.240604101634131e-06, "loss": 1.0021, "step": 8319 }, { "epoch": 0.7, "learning_rate": 4.2383619122719e-06, "loss": 0.9986, "step": 8320 }, { "epoch": 0.7, "learning_rate": 4.2361201564277345e-06, "loss": 0.9921, "step": 8321 }, { "epoch": 0.7, "learning_rate": 4.233878834270302e-06, "loss": 0.9569, "step": 8322 }, { "epoch": 0.7, "learning_rate": 4.231637945968251e-06, "loss": 0.8036, "step": 8323 }, { "epoch": 0.7, "learning_rate": 4.229397491690187e-06, "loss": 0.9788, "step": 8324 }, { "epoch": 0.71, "learning_rate": 4.227157471604684e-06, "loss": 0.943, "step": 8325 }, { "epoch": 0.71, "learning_rate": 4.224917885880291e-06, "loss": 0.9895, "step": 8326 }, { "epoch": 0.71, "learning_rate": 4.222678734685518e-06, "loss": 0.9496, "step": 8327 }, { "epoch": 0.71, "learning_rate": 4.22044001818884e-06, "loss": 1.0446, "step": 8328 }, { "epoch": 0.71, "learning_rate": 4.218201736558705e-06, "loss": 0.9053, "step": 8329 }, { "epoch": 0.71, "learning_rate": 4.21596388996353e-06, "loss": 0.7919, "step": 8330 }, { "epoch": 0.71, "learning_rate": 4.213726478571688e-06, "loss": 0.9833, "step": 8331 }, { "epoch": 0.71, "learning_rate": 4.211489502551536e-06, "loss": 0.98, "step": 8332 }, { "epoch": 0.71, "learning_rate": 4.2092529620713785e-06, "loss": 0.9417, "step": 8333 }, { "epoch": 0.71, "learning_rate": 4.207016857299508e-06, "loss": 0.9708, "step": 8334 }, { "epoch": 0.71, "learning_rate": 4.204781188404163e-06, "loss": 0.9903, "step": 8335 }, { "epoch": 0.71, "learning_rate": 4.202545955553571e-06, "loss": 0.9656, "step": 8336 }, { "epoch": 0.71, "learning_rate": 4.200311158915906e-06, "loss": 0.954, "step": 8337 }, { "epoch": 0.71, "learning_rate": 4.198076798659327e-06, "loss": 1.0187, "step": 8338 }, { "epoch": 0.71, "learning_rate": 4.195842874951944e-06, "loss": 0.9764, "step": 8339 }, { "epoch": 0.71, "learning_rate": 4.193609387961851e-06, "loss": 0.9973, "step": 8340 }, { "epoch": 0.71, "learning_rate": 4.1913763378570926e-06, "loss": 0.9567, "step": 8341 }, { "epoch": 0.71, "learning_rate": 4.189143724805695e-06, "loss": 1.0439, "step": 8342 }, { "epoch": 0.71, "learning_rate": 4.1869115489756384e-06, "loss": 1.0072, "step": 8343 }, { "epoch": 0.71, "learning_rate": 4.184679810534883e-06, "loss": 0.9436, "step": 8344 }, { "epoch": 0.71, "learning_rate": 4.1824485096513426e-06, "loss": 1.0416, "step": 8345 }, { "epoch": 0.71, "learning_rate": 4.180217646492912e-06, "loss": 0.9905, "step": 8346 }, { "epoch": 0.71, "learning_rate": 4.17798722122744e-06, "loss": 0.9654, "step": 8347 }, { "epoch": 0.71, "learning_rate": 4.175757234022755e-06, "loss": 0.9355, "step": 8348 }, { "epoch": 0.71, "learning_rate": 4.173527685046638e-06, "loss": 0.9821, "step": 8349 }, { "epoch": 0.71, "learning_rate": 4.171298574466854e-06, "loss": 1.021, "step": 8350 }, { "epoch": 0.71, "learning_rate": 4.169069902451122e-06, "loss": 0.9451, "step": 8351 }, { "epoch": 0.71, "learning_rate": 4.166841669167126e-06, "loss": 0.9422, "step": 8352 }, { "epoch": 0.71, "learning_rate": 4.164613874782535e-06, "loss": 1.017, "step": 8353 }, { "epoch": 0.71, "learning_rate": 4.162386519464962e-06, "loss": 0.9932, "step": 8354 }, { "epoch": 0.71, "learning_rate": 4.1601596033820065e-06, "loss": 0.9432, "step": 8355 }, { "epoch": 0.71, "learning_rate": 4.15793312670122e-06, "loss": 1.0203, "step": 8356 }, { "epoch": 0.71, "learning_rate": 4.1557070895901334e-06, "loss": 0.8563, "step": 8357 }, { "epoch": 0.71, "learning_rate": 4.153481492216232e-06, "loss": 0.9881, "step": 8358 }, { "epoch": 0.71, "learning_rate": 4.15125633474698e-06, "loss": 0.9784, "step": 8359 }, { "epoch": 0.71, "learning_rate": 4.1490316173497994e-06, "loss": 0.9638, "step": 8360 }, { "epoch": 0.71, "learning_rate": 4.146807340192087e-06, "loss": 0.9252, "step": 8361 }, { "epoch": 0.71, "learning_rate": 4.144583503441197e-06, "loss": 0.9564, "step": 8362 }, { "epoch": 0.71, "learning_rate": 4.1423601072644595e-06, "loss": 0.9654, "step": 8363 }, { "epoch": 0.71, "learning_rate": 4.140137151829164e-06, "loss": 0.9975, "step": 8364 }, { "epoch": 0.71, "learning_rate": 4.137914637302576e-06, "loss": 1.0114, "step": 8365 }, { "epoch": 0.71, "learning_rate": 4.135692563851916e-06, "loss": 0.9399, "step": 8366 }, { "epoch": 0.71, "learning_rate": 4.1334709316443846e-06, "loss": 0.9262, "step": 8367 }, { "epoch": 0.71, "learning_rate": 4.131249740847133e-06, "loss": 1.0039, "step": 8368 }, { "epoch": 0.71, "learning_rate": 4.129028991627299e-06, "loss": 0.9725, "step": 8369 }, { "epoch": 0.71, "learning_rate": 4.126808684151966e-06, "loss": 1.0131, "step": 8370 }, { "epoch": 0.71, "learning_rate": 4.124588818588204e-06, "loss": 1.0221, "step": 8371 }, { "epoch": 0.71, "learning_rate": 4.122369395103033e-06, "loss": 0.9815, "step": 8372 }, { "epoch": 0.71, "learning_rate": 4.120150413863454e-06, "loss": 0.9645, "step": 8373 }, { "epoch": 0.71, "learning_rate": 4.11793187503642e-06, "loss": 0.9006, "step": 8374 }, { "epoch": 0.71, "learning_rate": 4.115713778788868e-06, "loss": 0.9652, "step": 8375 }, { "epoch": 0.71, "learning_rate": 4.113496125287684e-06, "loss": 0.9439, "step": 8376 }, { "epoch": 0.71, "learning_rate": 4.111278914699734e-06, "loss": 0.9718, "step": 8377 }, { "epoch": 0.71, "learning_rate": 4.109062147191847e-06, "loss": 0.9708, "step": 8378 }, { "epoch": 0.71, "learning_rate": 4.1068458229308085e-06, "loss": 1.0355, "step": 8379 }, { "epoch": 0.71, "learning_rate": 4.1046299420833904e-06, "loss": 1.0217, "step": 8380 }, { "epoch": 0.71, "learning_rate": 4.102414504816312e-06, "loss": 0.9874, "step": 8381 }, { "epoch": 0.71, "learning_rate": 4.100199511296274e-06, "loss": 0.9756, "step": 8382 }, { "epoch": 0.71, "learning_rate": 4.097984961689932e-06, "loss": 0.9796, "step": 8383 }, { "epoch": 0.71, "learning_rate": 4.095770856163918e-06, "loss": 0.9673, "step": 8384 }, { "epoch": 0.71, "learning_rate": 4.09355719488482e-06, "loss": 0.9533, "step": 8385 }, { "epoch": 0.71, "learning_rate": 4.0913439780192075e-06, "loss": 1.0403, "step": 8386 }, { "epoch": 0.71, "learning_rate": 4.089131205733598e-06, "loss": 0.9351, "step": 8387 }, { "epoch": 0.71, "learning_rate": 4.086918878194495e-06, "loss": 0.9852, "step": 8388 }, { "epoch": 0.71, "learning_rate": 4.084706995568348e-06, "loss": 1.0084, "step": 8389 }, { "epoch": 0.71, "learning_rate": 4.082495558021596e-06, "loss": 1.0101, "step": 8390 }, { "epoch": 0.71, "learning_rate": 4.08028456572062e-06, "loss": 0.8341, "step": 8391 }, { "epoch": 0.71, "learning_rate": 4.07807401883179e-06, "loss": 0.9984, "step": 8392 }, { "epoch": 0.71, "learning_rate": 4.0758639175214255e-06, "loss": 1.0132, "step": 8393 }, { "epoch": 0.71, "learning_rate": 4.073654261955824e-06, "loss": 1.0345, "step": 8394 }, { "epoch": 0.71, "learning_rate": 4.07144505230124e-06, "loss": 1.0061, "step": 8395 }, { "epoch": 0.71, "learning_rate": 4.069236288723904e-06, "loss": 0.9984, "step": 8396 }, { "epoch": 0.71, "learning_rate": 4.067027971390004e-06, "loss": 1.0209, "step": 8397 }, { "epoch": 0.71, "learning_rate": 4.064820100465704e-06, "loss": 0.9867, "step": 8398 }, { "epoch": 0.71, "learning_rate": 4.062612676117121e-06, "loss": 0.9275, "step": 8399 }, { "epoch": 0.71, "learning_rate": 4.0604056985103545e-06, "loss": 0.8567, "step": 8400 }, { "epoch": 0.71, "learning_rate": 4.0581991678114555e-06, "loss": 0.9555, "step": 8401 }, { "epoch": 0.71, "learning_rate": 4.055993084186454e-06, "loss": 1.0088, "step": 8402 }, { "epoch": 0.71, "learning_rate": 4.053787447801334e-06, "loss": 1.0399, "step": 8403 }, { "epoch": 0.71, "learning_rate": 4.05158225882206e-06, "loss": 1.0122, "step": 8404 }, { "epoch": 0.71, "learning_rate": 4.049377517414551e-06, "loss": 0.9378, "step": 8405 }, { "epoch": 0.71, "learning_rate": 4.0471732237446935e-06, "loss": 0.9795, "step": 8406 }, { "epoch": 0.71, "learning_rate": 4.04496937797835e-06, "loss": 0.9681, "step": 8407 }, { "epoch": 0.71, "learning_rate": 4.042765980281335e-06, "loss": 0.9634, "step": 8408 }, { "epoch": 0.71, "learning_rate": 4.040563030819444e-06, "loss": 0.7328, "step": 8409 }, { "epoch": 0.71, "learning_rate": 4.038360529758426e-06, "loss": 0.9937, "step": 8410 }, { "epoch": 0.71, "learning_rate": 4.036158477264008e-06, "loss": 1.0164, "step": 8411 }, { "epoch": 0.71, "learning_rate": 4.033956873501871e-06, "loss": 0.9789, "step": 8412 }, { "epoch": 0.71, "learning_rate": 4.031755718637673e-06, "loss": 1.0182, "step": 8413 }, { "epoch": 0.71, "learning_rate": 4.029555012837029e-06, "loss": 0.9803, "step": 8414 }, { "epoch": 0.71, "learning_rate": 4.027354756265532e-06, "loss": 1.0046, "step": 8415 }, { "epoch": 0.71, "learning_rate": 4.025154949088724e-06, "loss": 0.9759, "step": 8416 }, { "epoch": 0.71, "learning_rate": 4.022955591472134e-06, "loss": 0.7645, "step": 8417 }, { "epoch": 0.71, "learning_rate": 4.020756683581238e-06, "loss": 1.0229, "step": 8418 }, { "epoch": 0.71, "learning_rate": 4.018558225581492e-06, "loss": 0.9632, "step": 8419 }, { "epoch": 0.71, "learning_rate": 4.016360217638307e-06, "loss": 0.9383, "step": 8420 }, { "epoch": 0.71, "learning_rate": 4.014162659917074e-06, "loss": 0.9285, "step": 8421 }, { "epoch": 0.71, "learning_rate": 4.011965552583133e-06, "loss": 0.9714, "step": 8422 }, { "epoch": 0.71, "learning_rate": 4.009768895801808e-06, "loss": 0.9706, "step": 8423 }, { "epoch": 0.71, "learning_rate": 4.007572689738371e-06, "loss": 0.9728, "step": 8424 }, { "epoch": 0.71, "learning_rate": 4.005376934558078e-06, "loss": 0.9371, "step": 8425 }, { "epoch": 0.71, "learning_rate": 4.003181630426135e-06, "loss": 1.0117, "step": 8426 }, { "epoch": 0.71, "learning_rate": 4.000986777507729e-06, "loss": 0.976, "step": 8427 }, { "epoch": 0.71, "learning_rate": 3.998792375967996e-06, "loss": 1.0132, "step": 8428 }, { "epoch": 0.71, "learning_rate": 3.996598425972057e-06, "loss": 1.006, "step": 8429 }, { "epoch": 0.71, "learning_rate": 3.994404927684982e-06, "loss": 0.8068, "step": 8430 }, { "epoch": 0.71, "learning_rate": 3.99221188127182e-06, "loss": 0.801, "step": 8431 }, { "epoch": 0.71, "learning_rate": 3.990019286897579e-06, "loss": 1.0007, "step": 8432 }, { "epoch": 0.71, "learning_rate": 3.987827144727231e-06, "loss": 0.9719, "step": 8433 }, { "epoch": 0.71, "learning_rate": 3.985635454925723e-06, "loss": 0.9832, "step": 8434 }, { "epoch": 0.71, "learning_rate": 3.983444217657957e-06, "loss": 0.9055, "step": 8435 }, { "epoch": 0.71, "learning_rate": 3.981253433088812e-06, "loss": 1.0578, "step": 8436 }, { "epoch": 0.71, "learning_rate": 3.97906310138312e-06, "loss": 0.9981, "step": 8437 }, { "epoch": 0.71, "learning_rate": 3.976873222705696e-06, "loss": 0.9604, "step": 8438 }, { "epoch": 0.71, "learning_rate": 3.9746837972213016e-06, "loss": 0.9988, "step": 8439 }, { "epoch": 0.71, "learning_rate": 3.972494825094681e-06, "loss": 0.7839, "step": 8440 }, { "epoch": 0.71, "learning_rate": 3.9703063064905334e-06, "loss": 0.9815, "step": 8441 }, { "epoch": 0.71, "learning_rate": 3.96811824157353e-06, "loss": 0.9719, "step": 8442 }, { "epoch": 0.72, "learning_rate": 3.965930630508301e-06, "loss": 0.9966, "step": 8443 }, { "epoch": 0.72, "learning_rate": 3.963743473459455e-06, "loss": 0.9651, "step": 8444 }, { "epoch": 0.72, "learning_rate": 3.961556770591549e-06, "loss": 1.0339, "step": 8445 }, { "epoch": 0.72, "learning_rate": 3.959370522069124e-06, "loss": 1.013, "step": 8446 }, { "epoch": 0.72, "learning_rate": 3.957184728056669e-06, "loss": 0.9542, "step": 8447 }, { "epoch": 0.72, "learning_rate": 3.954999388718658e-06, "loss": 0.9893, "step": 8448 }, { "epoch": 0.72, "learning_rate": 3.952814504219511e-06, "loss": 0.9964, "step": 8449 }, { "epoch": 0.72, "learning_rate": 3.95063007472363e-06, "loss": 0.9348, "step": 8450 }, { "epoch": 0.72, "learning_rate": 3.948446100395372e-06, "loss": 0.9978, "step": 8451 }, { "epoch": 0.72, "learning_rate": 3.94626258139907e-06, "loss": 0.794, "step": 8452 }, { "epoch": 0.72, "learning_rate": 3.944079517899008e-06, "loss": 1.0159, "step": 8453 }, { "epoch": 0.72, "learning_rate": 3.941896910059453e-06, "loss": 1.0571, "step": 8454 }, { "epoch": 0.72, "learning_rate": 3.93971475804462e-06, "loss": 0.9951, "step": 8455 }, { "epoch": 0.72, "learning_rate": 3.937533062018709e-06, "loss": 0.9901, "step": 8456 }, { "epoch": 0.72, "learning_rate": 3.935351822145868e-06, "loss": 0.9763, "step": 8457 }, { "epoch": 0.72, "learning_rate": 3.933171038590222e-06, "loss": 0.972, "step": 8458 }, { "epoch": 0.72, "learning_rate": 3.930990711515857e-06, "loss": 0.9797, "step": 8459 }, { "epoch": 0.72, "learning_rate": 3.928810841086821e-06, "loss": 1.0147, "step": 8460 }, { "epoch": 0.72, "learning_rate": 3.92663142746714e-06, "loss": 0.9901, "step": 8461 }, { "epoch": 0.72, "learning_rate": 3.924452470820791e-06, "loss": 0.9994, "step": 8462 }, { "epoch": 0.72, "learning_rate": 3.9222739713117296e-06, "loss": 1.0059, "step": 8463 }, { "epoch": 0.72, "learning_rate": 3.920095929103863e-06, "loss": 1.0216, "step": 8464 }, { "epoch": 0.72, "learning_rate": 3.9179183443610815e-06, "loss": 0.9655, "step": 8465 }, { "epoch": 0.72, "learning_rate": 3.9157412172472205e-06, "loss": 0.9679, "step": 8466 }, { "epoch": 0.72, "learning_rate": 3.913564547926102e-06, "loss": 1.0505, "step": 8467 }, { "epoch": 0.72, "learning_rate": 3.911388336561494e-06, "loss": 1.0204, "step": 8468 }, { "epoch": 0.72, "learning_rate": 3.909212583317149e-06, "loss": 0.9921, "step": 8469 }, { "epoch": 0.72, "learning_rate": 3.907037288356767e-06, "loss": 1.0309, "step": 8470 }, { "epoch": 0.72, "learning_rate": 3.904862451844027e-06, "loss": 0.9913, "step": 8471 }, { "epoch": 0.72, "learning_rate": 3.902688073942563e-06, "loss": 0.9725, "step": 8472 }, { "epoch": 0.72, "learning_rate": 3.900514154815987e-06, "loss": 1.0208, "step": 8473 }, { "epoch": 0.72, "learning_rate": 3.898340694627863e-06, "loss": 0.9808, "step": 8474 }, { "epoch": 0.72, "learning_rate": 3.896167693541732e-06, "loss": 1.0317, "step": 8475 }, { "epoch": 0.72, "learning_rate": 3.89399515172109e-06, "loss": 1.0476, "step": 8476 }, { "epoch": 0.72, "learning_rate": 3.891823069329411e-06, "loss": 0.9872, "step": 8477 }, { "epoch": 0.72, "learning_rate": 3.889651446530119e-06, "loss": 0.9225, "step": 8478 }, { "epoch": 0.72, "learning_rate": 3.887480283486617e-06, "loss": 0.9583, "step": 8479 }, { "epoch": 0.72, "learning_rate": 3.8853095803622645e-06, "loss": 1.0134, "step": 8480 }, { "epoch": 0.72, "learning_rate": 3.883139337320395e-06, "loss": 0.9484, "step": 8481 }, { "epoch": 0.72, "learning_rate": 3.880969554524296e-06, "loss": 0.8002, "step": 8482 }, { "epoch": 0.72, "learning_rate": 3.878800232137231e-06, "loss": 1.0034, "step": 8483 }, { "epoch": 0.72, "learning_rate": 3.876631370322421e-06, "loss": 0.989, "step": 8484 }, { "epoch": 0.72, "learning_rate": 3.874462969243061e-06, "loss": 0.9504, "step": 8485 }, { "epoch": 0.72, "learning_rate": 3.872295029062303e-06, "loss": 0.9631, "step": 8486 }, { "epoch": 0.72, "learning_rate": 3.870127549943265e-06, "loss": 0.9653, "step": 8487 }, { "epoch": 0.72, "learning_rate": 3.867960532049038e-06, "loss": 0.9842, "step": 8488 }, { "epoch": 0.72, "learning_rate": 3.865793975542666e-06, "loss": 0.9861, "step": 8489 }, { "epoch": 0.72, "learning_rate": 3.863627880587174e-06, "loss": 0.9692, "step": 8490 }, { "epoch": 0.72, "learning_rate": 3.861462247345536e-06, "loss": 0.9967, "step": 8491 }, { "epoch": 0.72, "learning_rate": 3.859297075980707e-06, "loss": 1.027, "step": 8492 }, { "epoch": 0.72, "learning_rate": 3.8571323666555906e-06, "loss": 0.9924, "step": 8493 }, { "epoch": 0.72, "learning_rate": 3.854968119533071e-06, "loss": 0.8303, "step": 8494 }, { "epoch": 0.72, "learning_rate": 3.852804334775984e-06, "loss": 0.9627, "step": 8495 }, { "epoch": 0.72, "learning_rate": 3.850641012547146e-06, "loss": 0.9849, "step": 8496 }, { "epoch": 0.72, "learning_rate": 3.848478153009322e-06, "loss": 0.9994, "step": 8497 }, { "epoch": 0.72, "learning_rate": 3.846315756325256e-06, "loss": 0.9085, "step": 8498 }, { "epoch": 0.72, "learning_rate": 3.8441538226576465e-06, "loss": 1.0461, "step": 8499 }, { "epoch": 0.72, "learning_rate": 3.841992352169167e-06, "loss": 1.0147, "step": 8500 }, { "epoch": 0.72, "learning_rate": 3.839831345022447e-06, "loss": 0.9972, "step": 8501 }, { "epoch": 0.72, "learning_rate": 3.837670801380089e-06, "loss": 1.0186, "step": 8502 }, { "epoch": 0.72, "learning_rate": 3.8355107214046516e-06, "loss": 0.9752, "step": 8503 }, { "epoch": 0.72, "learning_rate": 3.8333511052586716e-06, "loss": 0.8006, "step": 8504 }, { "epoch": 0.72, "learning_rate": 3.831191953104635e-06, "loss": 0.9707, "step": 8505 }, { "epoch": 0.72, "learning_rate": 3.8290332651050075e-06, "loss": 0.9679, "step": 8506 }, { "epoch": 0.72, "learning_rate": 3.826875041422209e-06, "loss": 0.9079, "step": 8507 }, { "epoch": 0.72, "learning_rate": 3.8247172822186325e-06, "loss": 1.0021, "step": 8508 }, { "epoch": 0.72, "learning_rate": 3.822559987656629e-06, "loss": 0.9864, "step": 8509 }, { "epoch": 0.72, "learning_rate": 3.820403157898523e-06, "loss": 1.0341, "step": 8510 }, { "epoch": 0.72, "learning_rate": 3.818246793106591e-06, "loss": 1.0025, "step": 8511 }, { "epoch": 0.72, "learning_rate": 3.816090893443092e-06, "loss": 1.0008, "step": 8512 }, { "epoch": 0.72, "learning_rate": 3.8139354590702347e-06, "loss": 0.9948, "step": 8513 }, { "epoch": 0.72, "learning_rate": 3.8117804901501966e-06, "loss": 0.9937, "step": 8514 }, { "epoch": 0.72, "learning_rate": 3.809625986845129e-06, "loss": 1.0303, "step": 8515 }, { "epoch": 0.72, "learning_rate": 3.807471949317134e-06, "loss": 0.9485, "step": 8516 }, { "epoch": 0.72, "learning_rate": 3.805318377728293e-06, "loss": 0.9796, "step": 8517 }, { "epoch": 0.72, "learning_rate": 3.8031652722406396e-06, "loss": 0.987, "step": 8518 }, { "epoch": 0.72, "learning_rate": 3.8010126330161835e-06, "loss": 0.9889, "step": 8519 }, { "epoch": 0.72, "learning_rate": 3.7988604602168878e-06, "loss": 1.0527, "step": 8520 }, { "epoch": 0.72, "learning_rate": 3.796708754004692e-06, "loss": 0.964, "step": 8521 }, { "epoch": 0.72, "learning_rate": 3.7945575145414903e-06, "loss": 1.0063, "step": 8522 }, { "epoch": 0.72, "learning_rate": 3.792406741989153e-06, "loss": 0.9832, "step": 8523 }, { "epoch": 0.72, "learning_rate": 3.7902564365095016e-06, "loss": 0.9743, "step": 8524 }, { "epoch": 0.72, "learning_rate": 3.7881065982643363e-06, "loss": 0.9707, "step": 8525 }, { "epoch": 0.72, "learning_rate": 3.785957227415409e-06, "loss": 1.0033, "step": 8526 }, { "epoch": 0.72, "learning_rate": 3.7838083241244493e-06, "loss": 0.9843, "step": 8527 }, { "epoch": 0.72, "learning_rate": 3.7816598885531395e-06, "loss": 0.834, "step": 8528 }, { "epoch": 0.72, "learning_rate": 3.7795119208631393e-06, "loss": 0.9901, "step": 8529 }, { "epoch": 0.72, "learning_rate": 3.7773644212160587e-06, "loss": 0.9423, "step": 8530 }, { "epoch": 0.72, "learning_rate": 3.7752173897734878e-06, "loss": 1.0054, "step": 8531 }, { "epoch": 0.72, "learning_rate": 3.7730708266969673e-06, "loss": 1.005, "step": 8532 }, { "epoch": 0.72, "learning_rate": 3.770924732148015e-06, "loss": 0.953, "step": 8533 }, { "epoch": 0.72, "learning_rate": 3.768779106288101e-06, "loss": 1.0061, "step": 8534 }, { "epoch": 0.72, "learning_rate": 3.766633949278674e-06, "loss": 1.0076, "step": 8535 }, { "epoch": 0.72, "learning_rate": 3.7644892612811337e-06, "loss": 0.9604, "step": 8536 }, { "epoch": 0.72, "learning_rate": 3.7623450424568575e-06, "loss": 1.0152, "step": 8537 }, { "epoch": 0.72, "learning_rate": 3.760201292967177e-06, "loss": 0.9905, "step": 8538 }, { "epoch": 0.72, "learning_rate": 3.7580580129733914e-06, "loss": 0.9784, "step": 8539 }, { "epoch": 0.72, "learning_rate": 3.7559152026367695e-06, "loss": 0.9659, "step": 8540 }, { "epoch": 0.72, "learning_rate": 3.7537728621185367e-06, "loss": 0.9985, "step": 8541 }, { "epoch": 0.72, "learning_rate": 3.751630991579892e-06, "loss": 0.9655, "step": 8542 }, { "epoch": 0.72, "learning_rate": 3.7494895911819896e-06, "loss": 0.9578, "step": 8543 }, { "epoch": 0.72, "learning_rate": 3.7473486610859577e-06, "loss": 0.9387, "step": 8544 }, { "epoch": 0.72, "learning_rate": 3.745208201452879e-06, "loss": 1.0081, "step": 8545 }, { "epoch": 0.72, "learning_rate": 3.743068212443813e-06, "loss": 1.0059, "step": 8546 }, { "epoch": 0.72, "learning_rate": 3.740928694219771e-06, "loss": 0.9856, "step": 8547 }, { "epoch": 0.72, "learning_rate": 3.7387896469417406e-06, "loss": 0.9802, "step": 8548 }, { "epoch": 0.72, "learning_rate": 3.7366510707706614e-06, "loss": 0.983, "step": 8549 }, { "epoch": 0.72, "learning_rate": 3.7345129658674527e-06, "loss": 0.9392, "step": 8550 }, { "epoch": 0.72, "learning_rate": 3.732375332392982e-06, "loss": 0.9684, "step": 8551 }, { "epoch": 0.72, "learning_rate": 3.730238170508097e-06, "loss": 0.9481, "step": 8552 }, { "epoch": 0.72, "learning_rate": 3.7281014803735947e-06, "loss": 0.8468, "step": 8553 }, { "epoch": 0.72, "learning_rate": 3.7259652621502518e-06, "loss": 0.909, "step": 8554 }, { "epoch": 0.72, "learning_rate": 3.723829515998795e-06, "loss": 0.9396, "step": 8555 }, { "epoch": 0.72, "learning_rate": 3.7216942420799285e-06, "loss": 0.9233, "step": 8556 }, { "epoch": 0.72, "learning_rate": 3.719559440554309e-06, "loss": 0.962, "step": 8557 }, { "epoch": 0.72, "learning_rate": 3.717425111582571e-06, "loss": 1.0019, "step": 8558 }, { "epoch": 0.72, "learning_rate": 3.7152912553252975e-06, "loss": 1.0036, "step": 8559 }, { "epoch": 0.72, "learning_rate": 3.7131578719430526e-06, "loss": 1.0054, "step": 8560 }, { "epoch": 0.73, "learning_rate": 3.711024961596349e-06, "loss": 1.026, "step": 8561 }, { "epoch": 0.73, "learning_rate": 3.7088925244456784e-06, "loss": 0.9784, "step": 8562 }, { "epoch": 0.73, "learning_rate": 3.7067605606514843e-06, "loss": 0.8929, "step": 8563 }, { "epoch": 0.73, "learning_rate": 3.704629070374186e-06, "loss": 0.9442, "step": 8564 }, { "epoch": 0.73, "learning_rate": 3.7024980537741574e-06, "loss": 0.9399, "step": 8565 }, { "epoch": 0.73, "learning_rate": 3.7003675110117388e-06, "loss": 1.0098, "step": 8566 }, { "epoch": 0.73, "learning_rate": 3.6982374422472422e-06, "loss": 0.9293, "step": 8567 }, { "epoch": 0.73, "learning_rate": 3.696107847640934e-06, "loss": 0.9688, "step": 8568 }, { "epoch": 0.73, "learning_rate": 3.6939787273530536e-06, "loss": 0.9923, "step": 8569 }, { "epoch": 0.73, "learning_rate": 3.6918500815437954e-06, "loss": 0.9883, "step": 8570 }, { "epoch": 0.73, "learning_rate": 3.6897219103733294e-06, "loss": 0.9841, "step": 8571 }, { "epoch": 0.73, "learning_rate": 3.6875942140017784e-06, "loss": 0.9972, "step": 8572 }, { "epoch": 0.73, "learning_rate": 3.6854669925892396e-06, "loss": 1.0107, "step": 8573 }, { "epoch": 0.73, "learning_rate": 3.6833402462957636e-06, "loss": 0.9961, "step": 8574 }, { "epoch": 0.73, "learning_rate": 3.6812139752813792e-06, "loss": 1.0197, "step": 8575 }, { "epoch": 0.73, "learning_rate": 3.679088179706064e-06, "loss": 0.9965, "step": 8576 }, { "epoch": 0.73, "learning_rate": 3.6769628597297737e-06, "loss": 0.8545, "step": 8577 }, { "epoch": 0.73, "learning_rate": 3.6748380155124163e-06, "loss": 0.9671, "step": 8578 }, { "epoch": 0.73, "learning_rate": 3.672713647213876e-06, "loss": 0.9892, "step": 8579 }, { "epoch": 0.73, "learning_rate": 3.670589754993987e-06, "loss": 0.9648, "step": 8580 }, { "epoch": 0.73, "learning_rate": 3.6684663390125654e-06, "loss": 0.9483, "step": 8581 }, { "epoch": 0.73, "learning_rate": 3.6663433994293707e-06, "loss": 0.9099, "step": 8582 }, { "epoch": 0.73, "learning_rate": 3.664220936404147e-06, "loss": 0.9549, "step": 8583 }, { "epoch": 0.73, "learning_rate": 3.6620989500965853e-06, "loss": 0.9709, "step": 8584 }, { "epoch": 0.73, "learning_rate": 3.6599774406663557e-06, "loss": 0.9616, "step": 8585 }, { "epoch": 0.73, "learning_rate": 3.657856408273077e-06, "loss": 0.9604, "step": 8586 }, { "epoch": 0.73, "learning_rate": 3.6557358530763485e-06, "loss": 0.9893, "step": 8587 }, { "epoch": 0.73, "learning_rate": 3.653615775235718e-06, "loss": 0.959, "step": 8588 }, { "epoch": 0.73, "learning_rate": 3.6514961749107125e-06, "loss": 0.9497, "step": 8589 }, { "epoch": 0.73, "learning_rate": 3.6493770522608065e-06, "loss": 1.017, "step": 8590 }, { "epoch": 0.73, "learning_rate": 3.6472584074454565e-06, "loss": 0.9889, "step": 8591 }, { "epoch": 0.73, "learning_rate": 3.6451402406240675e-06, "loss": 0.9718, "step": 8592 }, { "epoch": 0.73, "learning_rate": 3.6430225519560146e-06, "loss": 1.0282, "step": 8593 }, { "epoch": 0.73, "learning_rate": 3.6409053416006424e-06, "loss": 0.9512, "step": 8594 }, { "epoch": 0.73, "learning_rate": 3.6387886097172475e-06, "loss": 0.9927, "step": 8595 }, { "epoch": 0.73, "learning_rate": 3.6366723564651053e-06, "loss": 0.9969, "step": 8596 }, { "epoch": 0.73, "learning_rate": 3.634556582003439e-06, "loss": 0.9627, "step": 8597 }, { "epoch": 0.73, "learning_rate": 3.6324412864914517e-06, "loss": 0.9815, "step": 8598 }, { "epoch": 0.73, "learning_rate": 3.6303264700882954e-06, "loss": 0.8358, "step": 8599 }, { "epoch": 0.73, "learning_rate": 3.6282121329531016e-06, "loss": 1.0075, "step": 8600 }, { "epoch": 0.73, "learning_rate": 3.62609827524495e-06, "loss": 0.971, "step": 8601 }, { "epoch": 0.73, "learning_rate": 3.6239848971228987e-06, "loss": 1.0106, "step": 8602 }, { "epoch": 0.73, "learning_rate": 3.621871998745955e-06, "loss": 0.9754, "step": 8603 }, { "epoch": 0.73, "learning_rate": 3.6197595802731054e-06, "loss": 0.9824, "step": 8604 }, { "epoch": 0.73, "learning_rate": 3.6176476418632865e-06, "loss": 0.9567, "step": 8605 }, { "epoch": 0.73, "learning_rate": 3.615536183675411e-06, "loss": 0.9761, "step": 8606 }, { "epoch": 0.73, "learning_rate": 3.6134252058683437e-06, "loss": 0.9376, "step": 8607 }, { "epoch": 0.73, "learning_rate": 3.611314708600926e-06, "loss": 1.0231, "step": 8608 }, { "epoch": 0.73, "learning_rate": 3.6092046920319477e-06, "loss": 1.0205, "step": 8609 }, { "epoch": 0.73, "learning_rate": 3.6070951563201805e-06, "loss": 0.9943, "step": 8610 }, { "epoch": 0.73, "learning_rate": 3.6049861016243413e-06, "loss": 1.0214, "step": 8611 }, { "epoch": 0.73, "learning_rate": 3.602877528103128e-06, "loss": 0.9947, "step": 8612 }, { "epoch": 0.73, "learning_rate": 3.600769435915187e-06, "loss": 0.9787, "step": 8613 }, { "epoch": 0.73, "learning_rate": 3.5986618252191387e-06, "loss": 0.9519, "step": 8614 }, { "epoch": 0.73, "learning_rate": 3.5965546961735685e-06, "loss": 0.9351, "step": 8615 }, { "epoch": 0.73, "learning_rate": 3.594448048937017e-06, "loss": 0.8973, "step": 8616 }, { "epoch": 0.73, "learning_rate": 3.5923418836679904e-06, "loss": 0.9473, "step": 8617 }, { "epoch": 0.73, "learning_rate": 3.590236200524968e-06, "loss": 0.9994, "step": 8618 }, { "epoch": 0.73, "learning_rate": 3.5881309996663815e-06, "loss": 1.0168, "step": 8619 }, { "epoch": 0.73, "learning_rate": 3.5860262812506273e-06, "loss": 0.9485, "step": 8620 }, { "epoch": 0.73, "learning_rate": 3.5839220454360766e-06, "loss": 0.9959, "step": 8621 }, { "epoch": 0.73, "learning_rate": 3.581818292381051e-06, "loss": 0.9661, "step": 8622 }, { "epoch": 0.73, "learning_rate": 3.579715022243845e-06, "loss": 0.9622, "step": 8623 }, { "epoch": 0.73, "learning_rate": 3.577612235182709e-06, "loss": 0.9712, "step": 8624 }, { "epoch": 0.73, "learning_rate": 3.575509931355866e-06, "loss": 0.9801, "step": 8625 }, { "epoch": 0.73, "learning_rate": 3.5734081109214934e-06, "loss": 1.0009, "step": 8626 }, { "epoch": 0.73, "learning_rate": 3.5713067740377427e-06, "loss": 0.8143, "step": 8627 }, { "epoch": 0.73, "learning_rate": 3.569205920862715e-06, "loss": 0.9824, "step": 8628 }, { "epoch": 0.73, "learning_rate": 3.5671055515544915e-06, "loss": 0.9816, "step": 8629 }, { "epoch": 0.73, "learning_rate": 3.5650056662711e-06, "loss": 0.9683, "step": 8630 }, { "epoch": 0.73, "learning_rate": 3.5629062651705484e-06, "loss": 0.9326, "step": 8631 }, { "epoch": 0.73, "learning_rate": 3.560807348410793e-06, "loss": 0.9436, "step": 8632 }, { "epoch": 0.73, "learning_rate": 3.5587089161497678e-06, "loss": 0.9523, "step": 8633 }, { "epoch": 0.73, "learning_rate": 3.5566109685453566e-06, "loss": 1.0033, "step": 8634 }, { "epoch": 0.73, "learning_rate": 3.5545135057554193e-06, "loss": 0.9779, "step": 8635 }, { "epoch": 0.73, "learning_rate": 3.552416527937769e-06, "loss": 0.9798, "step": 8636 }, { "epoch": 0.73, "learning_rate": 3.550320035250191e-06, "loss": 0.9529, "step": 8637 }, { "epoch": 0.73, "learning_rate": 3.5482240278504242e-06, "loss": 0.9819, "step": 8638 }, { "epoch": 0.73, "learning_rate": 3.5461285058961848e-06, "loss": 0.966, "step": 8639 }, { "epoch": 0.73, "learning_rate": 3.544033469545135e-06, "loss": 0.9579, "step": 8640 }, { "epoch": 0.73, "learning_rate": 3.541938918954916e-06, "loss": 0.9601, "step": 8641 }, { "epoch": 0.73, "learning_rate": 3.5398448542831265e-06, "loss": 0.9967, "step": 8642 }, { "epoch": 0.73, "learning_rate": 3.537751275687329e-06, "loss": 1.0162, "step": 8643 }, { "epoch": 0.73, "learning_rate": 3.5356581833250416e-06, "loss": 1.0183, "step": 8644 }, { "epoch": 0.73, "learning_rate": 3.5335655773537624e-06, "loss": 0.9861, "step": 8645 }, { "epoch": 0.73, "learning_rate": 3.5314734579309393e-06, "loss": 0.9958, "step": 8646 }, { "epoch": 0.73, "learning_rate": 3.5293818252139843e-06, "loss": 0.997, "step": 8647 }, { "epoch": 0.73, "learning_rate": 3.5272906793602845e-06, "loss": 0.9616, "step": 8648 }, { "epoch": 0.73, "learning_rate": 3.525200020527173e-06, "loss": 0.9553, "step": 8649 }, { "epoch": 0.73, "learning_rate": 3.523109848871964e-06, "loss": 0.9294, "step": 8650 }, { "epoch": 0.73, "learning_rate": 3.521020164551919e-06, "loss": 0.987, "step": 8651 }, { "epoch": 0.73, "learning_rate": 3.518930967724279e-06, "loss": 0.9595, "step": 8652 }, { "epoch": 0.73, "learning_rate": 3.5168422585462305e-06, "loss": 0.9789, "step": 8653 }, { "epoch": 0.73, "learning_rate": 3.51475403717494e-06, "loss": 0.9676, "step": 8654 }, { "epoch": 0.73, "learning_rate": 3.5126663037675225e-06, "loss": 1.027, "step": 8655 }, { "epoch": 0.73, "learning_rate": 3.510579058481072e-06, "loss": 0.9838, "step": 8656 }, { "epoch": 0.73, "learning_rate": 3.508492301472629e-06, "loss": 0.825, "step": 8657 }, { "epoch": 0.73, "learning_rate": 3.5064060328992134e-06, "loss": 0.9357, "step": 8658 }, { "epoch": 0.73, "learning_rate": 3.5043202529177922e-06, "loss": 0.9551, "step": 8659 }, { "epoch": 0.73, "learning_rate": 3.5022349616853125e-06, "loss": 1.0084, "step": 8660 }, { "epoch": 0.73, "learning_rate": 3.5001501593586686e-06, "loss": 0.9835, "step": 8661 }, { "epoch": 0.73, "learning_rate": 3.4980658460947326e-06, "loss": 1.0063, "step": 8662 }, { "epoch": 0.73, "learning_rate": 3.495982022050325e-06, "loss": 1.003, "step": 8663 }, { "epoch": 0.73, "learning_rate": 3.4938986873822445e-06, "loss": 0.9598, "step": 8664 }, { "epoch": 0.73, "learning_rate": 3.4918158422472403e-06, "loss": 0.9699, "step": 8665 }, { "epoch": 0.73, "learning_rate": 3.4897334868020317e-06, "loss": 0.9744, "step": 8666 }, { "epoch": 0.73, "learning_rate": 3.487651621203303e-06, "loss": 0.9538, "step": 8667 }, { "epoch": 0.73, "learning_rate": 3.4855702456076924e-06, "loss": 0.9415, "step": 8668 }, { "epoch": 0.73, "learning_rate": 3.4834893601718136e-06, "loss": 0.9914, "step": 8669 }, { "epoch": 0.73, "learning_rate": 3.481408965052233e-06, "loss": 1.0269, "step": 8670 }, { "epoch": 0.73, "learning_rate": 3.4793290604054817e-06, "loss": 0.9581, "step": 8671 }, { "epoch": 0.73, "learning_rate": 3.4772496463880625e-06, "loss": 0.9968, "step": 8672 }, { "epoch": 0.73, "learning_rate": 3.4751707231564313e-06, "loss": 0.997, "step": 8673 }, { "epoch": 0.73, "learning_rate": 3.4730922908670073e-06, "loss": 0.9371, "step": 8674 }, { "epoch": 0.73, "learning_rate": 3.471014349676184e-06, "loss": 0.9625, "step": 8675 }, { "epoch": 0.73, "learning_rate": 3.4689368997403018e-06, "loss": 0.9877, "step": 8676 }, { "epoch": 0.73, "learning_rate": 3.46685994121568e-06, "loss": 0.9609, "step": 8677 }, { "epoch": 0.73, "learning_rate": 3.464783474258587e-06, "loss": 1.0187, "step": 8678 }, { "epoch": 0.74, "learning_rate": 3.4627074990252684e-06, "loss": 0.9558, "step": 8679 }, { "epoch": 0.74, "learning_rate": 3.4606320156719154e-06, "loss": 0.9545, "step": 8680 }, { "epoch": 0.74, "learning_rate": 3.4585570243547007e-06, "loss": 0.9857, "step": 8681 }, { "epoch": 0.74, "learning_rate": 3.456482525229744e-06, "loss": 0.9726, "step": 8682 }, { "epoch": 0.74, "learning_rate": 3.454408518453143e-06, "loss": 0.9675, "step": 8683 }, { "epoch": 0.74, "learning_rate": 3.452335004180941e-06, "loss": 0.981, "step": 8684 }, { "epoch": 0.74, "learning_rate": 3.450261982569162e-06, "loss": 1.0336, "step": 8685 }, { "epoch": 0.74, "learning_rate": 3.448189453773778e-06, "loss": 1.0061, "step": 8686 }, { "epoch": 0.74, "learning_rate": 3.446117417950737e-06, "loss": 0.9843, "step": 8687 }, { "epoch": 0.74, "learning_rate": 3.444045875255937e-06, "loss": 0.978, "step": 8688 }, { "epoch": 0.74, "learning_rate": 3.441974825845251e-06, "loss": 0.985, "step": 8689 }, { "epoch": 0.74, "learning_rate": 3.4399042698745044e-06, "loss": 0.8537, "step": 8690 }, { "epoch": 0.74, "learning_rate": 3.4378342074994954e-06, "loss": 0.9901, "step": 8691 }, { "epoch": 0.74, "learning_rate": 3.4357646388759736e-06, "loss": 0.9421, "step": 8692 }, { "epoch": 0.74, "learning_rate": 3.4336955641596613e-06, "loss": 0.9952, "step": 8693 }, { "epoch": 0.74, "learning_rate": 3.4316269835062445e-06, "loss": 1.0326, "step": 8694 }, { "epoch": 0.74, "learning_rate": 3.429558897071358e-06, "loss": 0.9467, "step": 8695 }, { "epoch": 0.74, "learning_rate": 3.4274913050106196e-06, "loss": 0.9844, "step": 8696 }, { "epoch": 0.74, "learning_rate": 3.425424207479593e-06, "loss": 1.0004, "step": 8697 }, { "epoch": 0.74, "learning_rate": 3.423357604633809e-06, "loss": 0.9891, "step": 8698 }, { "epoch": 0.74, "learning_rate": 3.42129149662877e-06, "loss": 0.927, "step": 8699 }, { "epoch": 0.74, "learning_rate": 3.419225883619931e-06, "loss": 1.0255, "step": 8700 }, { "epoch": 0.74, "learning_rate": 3.417160765762709e-06, "loss": 0.9544, "step": 8701 }, { "epoch": 0.74, "learning_rate": 3.4150961432124962e-06, "loss": 0.9728, "step": 8702 }, { "epoch": 0.74, "learning_rate": 3.4130320161246323e-06, "loss": 0.983, "step": 8703 }, { "epoch": 0.74, "learning_rate": 3.4109683846544318e-06, "loss": 1.0652, "step": 8704 }, { "epoch": 0.74, "learning_rate": 3.4089052489571617e-06, "loss": 1.011, "step": 8705 }, { "epoch": 0.74, "learning_rate": 3.4068426091880612e-06, "loss": 1.0098, "step": 8706 }, { "epoch": 0.74, "learning_rate": 3.404780465502324e-06, "loss": 1.0026, "step": 8707 }, { "epoch": 0.74, "learning_rate": 3.402718818055115e-06, "loss": 0.9836, "step": 8708 }, { "epoch": 0.74, "learning_rate": 3.4006576670015524e-06, "loss": 0.9442, "step": 8709 }, { "epoch": 0.74, "learning_rate": 3.398597012496725e-06, "loss": 0.9566, "step": 8710 }, { "epoch": 0.74, "learning_rate": 3.3965368546956767e-06, "loss": 0.9868, "step": 8711 }, { "epoch": 0.74, "learning_rate": 3.394477193753425e-06, "loss": 0.9888, "step": 8712 }, { "epoch": 0.74, "learning_rate": 3.3924180298249343e-06, "loss": 0.9791, "step": 8713 }, { "epoch": 0.74, "learning_rate": 3.39035936306515e-06, "loss": 0.9825, "step": 8714 }, { "epoch": 0.74, "learning_rate": 3.3883011936289624e-06, "loss": 0.9934, "step": 8715 }, { "epoch": 0.74, "learning_rate": 3.3862435216712385e-06, "loss": 0.9622, "step": 8716 }, { "epoch": 0.74, "learning_rate": 3.384186347346797e-06, "loss": 0.8458, "step": 8717 }, { "epoch": 0.74, "learning_rate": 3.3821296708104268e-06, "loss": 1.0046, "step": 8718 }, { "epoch": 0.74, "learning_rate": 3.38007349221688e-06, "loss": 0.8923, "step": 8719 }, { "epoch": 0.74, "learning_rate": 3.3780178117208606e-06, "loss": 1.0028, "step": 8720 }, { "epoch": 0.74, "learning_rate": 3.375962629477051e-06, "loss": 0.9499, "step": 8721 }, { "epoch": 0.74, "learning_rate": 3.373907945640078e-06, "loss": 1.025, "step": 8722 }, { "epoch": 0.74, "learning_rate": 3.3718537603645497e-06, "loss": 0.9982, "step": 8723 }, { "epoch": 0.74, "learning_rate": 3.369800073805023e-06, "loss": 0.9882, "step": 8724 }, { "epoch": 0.74, "learning_rate": 3.3677468861160177e-06, "loss": 0.9768, "step": 8725 }, { "epoch": 0.74, "learning_rate": 3.3656941974520287e-06, "loss": 0.9747, "step": 8726 }, { "epoch": 0.74, "learning_rate": 3.363642007967499e-06, "loss": 0.9669, "step": 8727 }, { "epoch": 0.74, "learning_rate": 3.3615903178168384e-06, "loss": 0.9767, "step": 8728 }, { "epoch": 0.74, "learning_rate": 3.3595391271544263e-06, "loss": 0.9744, "step": 8729 }, { "epoch": 0.74, "learning_rate": 3.3574884361345914e-06, "loss": 0.9862, "step": 8730 }, { "epoch": 0.74, "learning_rate": 3.3554382449116395e-06, "loss": 0.9676, "step": 8731 }, { "epoch": 0.74, "learning_rate": 3.3533885536398256e-06, "loss": 0.9824, "step": 8732 }, { "epoch": 0.74, "learning_rate": 3.351339362473378e-06, "loss": 0.9666, "step": 8733 }, { "epoch": 0.74, "learning_rate": 3.349290671566476e-06, "loss": 1.0221, "step": 8734 }, { "epoch": 0.74, "learning_rate": 3.3472424810732753e-06, "loss": 0.9634, "step": 8735 }, { "epoch": 0.74, "learning_rate": 3.3451947911478776e-06, "loss": 0.9929, "step": 8736 }, { "epoch": 0.74, "learning_rate": 3.3431476019443642e-06, "loss": 0.9544, "step": 8737 }, { "epoch": 0.74, "learning_rate": 3.3411009136167614e-06, "loss": 0.9533, "step": 8738 }, { "epoch": 0.74, "learning_rate": 3.339054726319074e-06, "loss": 0.9955, "step": 8739 }, { "epoch": 0.74, "learning_rate": 3.3370090402052557e-06, "loss": 0.9645, "step": 8740 }, { "epoch": 0.74, "learning_rate": 3.334963855429234e-06, "loss": 0.9424, "step": 8741 }, { "epoch": 0.74, "learning_rate": 3.3329191721448862e-06, "loss": 0.9798, "step": 8742 }, { "epoch": 0.74, "learning_rate": 3.3308749905060665e-06, "loss": 0.9595, "step": 8743 }, { "epoch": 0.74, "learning_rate": 3.328831310666576e-06, "loss": 1.0582, "step": 8744 }, { "epoch": 0.74, "learning_rate": 3.3267881327801874e-06, "loss": 0.9983, "step": 8745 }, { "epoch": 0.74, "learning_rate": 3.32474545700064e-06, "loss": 0.9888, "step": 8746 }, { "epoch": 0.74, "learning_rate": 3.322703283481621e-06, "loss": 0.98, "step": 8747 }, { "epoch": 0.74, "learning_rate": 3.3206616123767953e-06, "loss": 0.9611, "step": 8748 }, { "epoch": 0.74, "learning_rate": 3.3186204438397786e-06, "loss": 0.9912, "step": 8749 }, { "epoch": 0.74, "learning_rate": 3.3165797780241483e-06, "loss": 0.9813, "step": 8750 }, { "epoch": 0.74, "learning_rate": 3.3145396150834573e-06, "loss": 0.9287, "step": 8751 }, { "epoch": 0.74, "learning_rate": 3.3124999551712057e-06, "loss": 1.0154, "step": 8752 }, { "epoch": 0.74, "learning_rate": 3.3104607984408656e-06, "loss": 0.9309, "step": 8753 }, { "epoch": 0.74, "learning_rate": 3.3084221450458665e-06, "loss": 0.975, "step": 8754 }, { "epoch": 0.74, "learning_rate": 3.3063839951395973e-06, "loss": 1.003, "step": 8755 }, { "epoch": 0.74, "learning_rate": 3.304346348875419e-06, "loss": 1.0052, "step": 8756 }, { "epoch": 0.74, "learning_rate": 3.302309206406642e-06, "loss": 0.9695, "step": 8757 }, { "epoch": 0.74, "learning_rate": 3.3002725678865522e-06, "loss": 0.9634, "step": 8758 }, { "epoch": 0.74, "learning_rate": 3.2982364334683838e-06, "loss": 0.9303, "step": 8759 }, { "epoch": 0.74, "learning_rate": 3.2962008033053473e-06, "loss": 0.9885, "step": 8760 }, { "epoch": 0.74, "learning_rate": 3.2941656775505994e-06, "loss": 1.0405, "step": 8761 }, { "epoch": 0.74, "learning_rate": 3.292131056357276e-06, "loss": 0.9736, "step": 8762 }, { "epoch": 0.74, "learning_rate": 3.290096939878459e-06, "loss": 1.0095, "step": 8763 }, { "epoch": 0.74, "learning_rate": 3.2880633282672057e-06, "loss": 0.9659, "step": 8764 }, { "epoch": 0.74, "learning_rate": 3.2860302216765237e-06, "loss": 0.9795, "step": 8765 }, { "epoch": 0.74, "learning_rate": 3.283997620259395e-06, "loss": 1.0042, "step": 8766 }, { "epoch": 0.74, "learning_rate": 3.2819655241687497e-06, "loss": 1.0279, "step": 8767 }, { "epoch": 0.74, "learning_rate": 3.279933933557494e-06, "loss": 0.9757, "step": 8768 }, { "epoch": 0.74, "learning_rate": 3.2779028485784826e-06, "loss": 0.996, "step": 8769 }, { "epoch": 0.74, "learning_rate": 3.2758722693845415e-06, "loss": 1.0051, "step": 8770 }, { "epoch": 0.74, "learning_rate": 3.2738421961284594e-06, "loss": 0.9642, "step": 8771 }, { "epoch": 0.74, "learning_rate": 3.271812628962977e-06, "loss": 0.9923, "step": 8772 }, { "epoch": 0.74, "learning_rate": 3.2697835680408106e-06, "loss": 0.9434, "step": 8773 }, { "epoch": 0.74, "learning_rate": 3.267755013514623e-06, "loss": 0.9877, "step": 8774 }, { "epoch": 0.74, "learning_rate": 3.265726965537054e-06, "loss": 0.9265, "step": 8775 }, { "epoch": 0.74, "learning_rate": 3.2636994242606958e-06, "loss": 0.9143, "step": 8776 }, { "epoch": 0.74, "learning_rate": 3.2616723898381007e-06, "loss": 1.0117, "step": 8777 }, { "epoch": 0.74, "learning_rate": 3.259645862421795e-06, "loss": 0.9714, "step": 8778 }, { "epoch": 0.74, "learning_rate": 3.2576198421642546e-06, "loss": 0.9967, "step": 8779 }, { "epoch": 0.74, "learning_rate": 3.2555943292179194e-06, "loss": 0.9475, "step": 8780 }, { "epoch": 0.74, "learning_rate": 3.253569323735198e-06, "loss": 0.9386, "step": 8781 }, { "epoch": 0.74, "learning_rate": 3.251544825868451e-06, "loss": 0.9694, "step": 8782 }, { "epoch": 0.74, "learning_rate": 3.2495208357700138e-06, "loss": 0.9774, "step": 8783 }, { "epoch": 0.74, "learning_rate": 3.247497353592166e-06, "loss": 0.8944, "step": 8784 }, { "epoch": 0.74, "learning_rate": 3.245474379487168e-06, "loss": 0.977, "step": 8785 }, { "epoch": 0.74, "learning_rate": 3.243451913607225e-06, "loss": 1.0205, "step": 8786 }, { "epoch": 0.74, "learning_rate": 3.2414299561045195e-06, "loss": 1.0118, "step": 8787 }, { "epoch": 0.74, "learning_rate": 3.2394085071311786e-06, "loss": 0.9681, "step": 8788 }, { "epoch": 0.74, "learning_rate": 3.2373875668393094e-06, "loss": 0.9275, "step": 8789 }, { "epoch": 0.74, "learning_rate": 3.2353671353809656e-06, "loss": 0.9826, "step": 8790 }, { "epoch": 0.74, "learning_rate": 3.2333472129081732e-06, "loss": 0.9597, "step": 8791 }, { "epoch": 0.74, "learning_rate": 3.231327799572912e-06, "loss": 0.9685, "step": 8792 }, { "epoch": 0.74, "learning_rate": 3.22930889552713e-06, "loss": 0.8642, "step": 8793 }, { "epoch": 0.74, "learning_rate": 3.2272905009227296e-06, "loss": 1.0186, "step": 8794 }, { "epoch": 0.74, "learning_rate": 3.2252726159115855e-06, "loss": 0.8862, "step": 8795 }, { "epoch": 0.74, "learning_rate": 3.223255240645521e-06, "loss": 0.9998, "step": 8796 }, { "epoch": 0.75, "learning_rate": 3.22123837527633e-06, "loss": 0.9295, "step": 8797 }, { "epoch": 0.75, "learning_rate": 3.219222019955771e-06, "loss": 0.9874, "step": 8798 }, { "epoch": 0.75, "learning_rate": 3.217206174835551e-06, "loss": 0.9934, "step": 8799 }, { "epoch": 0.75, "learning_rate": 3.215190840067354e-06, "loss": 0.9932, "step": 8800 }, { "epoch": 0.75, "learning_rate": 3.21317601580281e-06, "loss": 1.0108, "step": 8801 }, { "epoch": 0.75, "learning_rate": 3.211161702193527e-06, "loss": 0.991, "step": 8802 }, { "epoch": 0.75, "learning_rate": 3.2091478993910617e-06, "loss": 0.9948, "step": 8803 }, { "epoch": 0.75, "learning_rate": 3.2071346075469347e-06, "loss": 1.0144, "step": 8804 }, { "epoch": 0.75, "learning_rate": 3.205121826812636e-06, "loss": 0.9828, "step": 8805 }, { "epoch": 0.75, "learning_rate": 3.20310955733961e-06, "loss": 0.9703, "step": 8806 }, { "epoch": 0.75, "learning_rate": 3.201097799279258e-06, "loss": 0.9767, "step": 8807 }, { "epoch": 0.75, "learning_rate": 3.1990865527829586e-06, "loss": 0.9178, "step": 8808 }, { "epoch": 0.75, "learning_rate": 3.1970758180020346e-06, "loss": 0.9984, "step": 8809 }, { "epoch": 0.75, "learning_rate": 3.195065595087784e-06, "loss": 0.9773, "step": 8810 }, { "epoch": 0.75, "learning_rate": 3.193055884191454e-06, "loss": 0.9763, "step": 8811 }, { "epoch": 0.75, "learning_rate": 3.1910466854642675e-06, "loss": 0.9788, "step": 8812 }, { "epoch": 0.75, "learning_rate": 3.1890379990573917e-06, "loss": 0.9808, "step": 8813 }, { "epoch": 0.75, "learning_rate": 3.1870298251219734e-06, "loss": 0.9976, "step": 8814 }, { "epoch": 0.75, "learning_rate": 3.1850221638091037e-06, "loss": 0.9626, "step": 8815 }, { "epoch": 0.75, "learning_rate": 3.1830150152698526e-06, "loss": 0.9384, "step": 8816 }, { "epoch": 0.75, "learning_rate": 3.1810083796552314e-06, "loss": 0.9781, "step": 8817 }, { "epoch": 0.75, "learning_rate": 3.179002257116235e-06, "loss": 0.9413, "step": 8818 }, { "epoch": 0.75, "learning_rate": 3.1769966478037983e-06, "loss": 0.9891, "step": 8819 }, { "epoch": 0.75, "learning_rate": 3.1749915518688356e-06, "loss": 0.9867, "step": 8820 }, { "epoch": 0.75, "learning_rate": 3.1729869694622085e-06, "loss": 0.9407, "step": 8821 }, { "epoch": 0.75, "learning_rate": 3.170982900734748e-06, "loss": 0.7822, "step": 8822 }, { "epoch": 0.75, "learning_rate": 3.1689793458372484e-06, "loss": 0.9208, "step": 8823 }, { "epoch": 0.75, "learning_rate": 3.1669763049204562e-06, "loss": 0.9933, "step": 8824 }, { "epoch": 0.75, "learning_rate": 3.1649737781350896e-06, "loss": 0.9477, "step": 8825 }, { "epoch": 0.75, "learning_rate": 3.162971765631817e-06, "loss": 1.0286, "step": 8826 }, { "epoch": 0.75, "learning_rate": 3.1609702675612806e-06, "loss": 1.0037, "step": 8827 }, { "epoch": 0.75, "learning_rate": 3.1589692840740715e-06, "loss": 1.0045, "step": 8828 }, { "epoch": 0.75, "learning_rate": 3.1569688153207532e-06, "loss": 0.9861, "step": 8829 }, { "epoch": 0.75, "learning_rate": 3.1549688614518437e-06, "loss": 1.0078, "step": 8830 }, { "epoch": 0.75, "learning_rate": 3.1529694226178186e-06, "loss": 0.9375, "step": 8831 }, { "epoch": 0.75, "learning_rate": 3.150970498969128e-06, "loss": 0.9601, "step": 8832 }, { "epoch": 0.75, "learning_rate": 3.1489720906561706e-06, "loss": 0.9715, "step": 8833 }, { "epoch": 0.75, "learning_rate": 3.14697419782931e-06, "loss": 0.9602, "step": 8834 }, { "epoch": 0.75, "learning_rate": 3.1449768206388764e-06, "loss": 0.9841, "step": 8835 }, { "epoch": 0.75, "learning_rate": 3.142979959235151e-06, "loss": 0.9481, "step": 8836 }, { "epoch": 0.75, "learning_rate": 3.140983613768388e-06, "loss": 0.9288, "step": 8837 }, { "epoch": 0.75, "learning_rate": 3.1389877843887908e-06, "loss": 0.964, "step": 8838 }, { "epoch": 0.75, "learning_rate": 3.1369924712465362e-06, "loss": 0.9448, "step": 8839 }, { "epoch": 0.75, "learning_rate": 3.13499767449175e-06, "loss": 0.8905, "step": 8840 }, { "epoch": 0.75, "learning_rate": 3.13300339427453e-06, "loss": 0.8641, "step": 8841 }, { "epoch": 0.75, "learning_rate": 3.1310096307449246e-06, "loss": 0.968, "step": 8842 }, { "epoch": 0.75, "learning_rate": 3.1290163840529554e-06, "loss": 0.9937, "step": 8843 }, { "epoch": 0.75, "learning_rate": 3.127023654348592e-06, "loss": 0.9605, "step": 8844 }, { "epoch": 0.75, "learning_rate": 3.1250314417817785e-06, "loss": 0.9078, "step": 8845 }, { "epoch": 0.75, "learning_rate": 3.1230397465024066e-06, "loss": 1.0109, "step": 8846 }, { "epoch": 0.75, "learning_rate": 3.121048568660341e-06, "loss": 0.9749, "step": 8847 }, { "epoch": 0.75, "learning_rate": 3.1190579084053973e-06, "loss": 0.9788, "step": 8848 }, { "epoch": 0.75, "learning_rate": 3.1170677658873594e-06, "loss": 0.9331, "step": 8849 }, { "epoch": 0.75, "learning_rate": 3.115078141255974e-06, "loss": 0.9457, "step": 8850 }, { "epoch": 0.75, "learning_rate": 3.113089034660938e-06, "loss": 1.0312, "step": 8851 }, { "epoch": 0.75, "learning_rate": 3.111100446251921e-06, "loss": 0.9393, "step": 8852 }, { "epoch": 0.75, "learning_rate": 3.109112376178545e-06, "loss": 0.9717, "step": 8853 }, { "epoch": 0.75, "learning_rate": 3.1071248245904007e-06, "loss": 0.9998, "step": 8854 }, { "epoch": 0.75, "learning_rate": 3.1051377916370307e-06, "loss": 0.9533, "step": 8855 }, { "epoch": 0.75, "learning_rate": 3.1031512774679497e-06, "loss": 0.9494, "step": 8856 }, { "epoch": 0.75, "learning_rate": 3.101165282232624e-06, "loss": 0.9857, "step": 8857 }, { "epoch": 0.75, "learning_rate": 3.0991798060804814e-06, "loss": 0.9597, "step": 8858 }, { "epoch": 0.75, "learning_rate": 3.097194849160918e-06, "loss": 0.9974, "step": 8859 }, { "epoch": 0.75, "learning_rate": 3.095210411623286e-06, "loss": 0.9832, "step": 8860 }, { "epoch": 0.75, "learning_rate": 3.0932264936168934e-06, "loss": 0.9176, "step": 8861 }, { "epoch": 0.75, "learning_rate": 3.091243095291022e-06, "loss": 1.0352, "step": 8862 }, { "epoch": 0.75, "learning_rate": 3.0892602167948992e-06, "loss": 1.0308, "step": 8863 }, { "epoch": 0.75, "learning_rate": 3.0872778582777286e-06, "loss": 0.9649, "step": 8864 }, { "epoch": 0.75, "learning_rate": 3.0852960198886604e-06, "loss": 0.9798, "step": 8865 }, { "epoch": 0.75, "learning_rate": 3.083314701776818e-06, "loss": 0.9886, "step": 8866 }, { "epoch": 0.75, "learning_rate": 3.081333904091276e-06, "loss": 0.9442, "step": 8867 }, { "epoch": 0.75, "learning_rate": 3.0793536269810774e-06, "loss": 0.9415, "step": 8868 }, { "epoch": 0.75, "learning_rate": 3.0773738705952184e-06, "loss": 1.0018, "step": 8869 }, { "epoch": 0.75, "learning_rate": 3.075394635082665e-06, "loss": 0.9519, "step": 8870 }, { "epoch": 0.75, "learning_rate": 3.073415920592333e-06, "loss": 1.0271, "step": 8871 }, { "epoch": 0.75, "learning_rate": 3.0714377272731123e-06, "loss": 1.0327, "step": 8872 }, { "epoch": 0.75, "learning_rate": 3.0694600552738395e-06, "loss": 0.9767, "step": 8873 }, { "epoch": 0.75, "learning_rate": 3.067482904743322e-06, "loss": 0.9951, "step": 8874 }, { "epoch": 0.75, "learning_rate": 3.0655062758303287e-06, "loss": 1.0026, "step": 8875 }, { "epoch": 0.75, "learning_rate": 3.063530168683578e-06, "loss": 0.9736, "step": 8876 }, { "epoch": 0.75, "learning_rate": 3.0615545834517635e-06, "loss": 1.0052, "step": 8877 }, { "epoch": 0.75, "learning_rate": 3.0595795202835266e-06, "loss": 1.0022, "step": 8878 }, { "epoch": 0.75, "learning_rate": 3.05760497932748e-06, "loss": 0.9452, "step": 8879 }, { "epoch": 0.75, "learning_rate": 3.0556309607321877e-06, "loss": 0.9858, "step": 8880 }, { "epoch": 0.75, "learning_rate": 3.0536574646461857e-06, "loss": 1.0039, "step": 8881 }, { "epoch": 0.75, "learning_rate": 3.0516844912179565e-06, "loss": 0.9672, "step": 8882 }, { "epoch": 0.75, "learning_rate": 3.049712040595957e-06, "loss": 0.9672, "step": 8883 }, { "epoch": 0.75, "learning_rate": 3.0477401129285966e-06, "loss": 1.0538, "step": 8884 }, { "epoch": 0.75, "learning_rate": 3.045768708364244e-06, "loss": 0.9911, "step": 8885 }, { "epoch": 0.75, "learning_rate": 3.043797827051238e-06, "loss": 0.9585, "step": 8886 }, { "epoch": 0.75, "learning_rate": 3.041827469137868e-06, "loss": 0.9887, "step": 8887 }, { "epoch": 0.75, "learning_rate": 3.0398576347723862e-06, "loss": 0.9679, "step": 8888 }, { "epoch": 0.75, "learning_rate": 3.0378883241030133e-06, "loss": 0.977, "step": 8889 }, { "epoch": 0.75, "learning_rate": 3.035919537277917e-06, "loss": 1.0069, "step": 8890 }, { "epoch": 0.75, "learning_rate": 3.0339512744452413e-06, "loss": 1.047, "step": 8891 }, { "epoch": 0.75, "learning_rate": 3.0319835357530746e-06, "loss": 0.96, "step": 8892 }, { "epoch": 0.75, "learning_rate": 3.0300163213494803e-06, "loss": 1.0126, "step": 8893 }, { "epoch": 0.75, "learning_rate": 3.028049631382469e-06, "loss": 1.0148, "step": 8894 }, { "epoch": 0.75, "learning_rate": 3.026083466000027e-06, "loss": 0.986, "step": 8895 }, { "epoch": 0.75, "learning_rate": 3.0241178253500857e-06, "loss": 0.9759, "step": 8896 }, { "epoch": 0.75, "learning_rate": 3.022152709580549e-06, "loss": 1.008, "step": 8897 }, { "epoch": 0.75, "learning_rate": 3.0201881188392722e-06, "loss": 0.9936, "step": 8898 }, { "epoch": 0.75, "learning_rate": 3.0182240532740805e-06, "loss": 0.9691, "step": 8899 }, { "epoch": 0.75, "learning_rate": 3.016260513032747e-06, "loss": 1.0385, "step": 8900 }, { "epoch": 0.75, "learning_rate": 3.014297498263018e-06, "loss": 0.9807, "step": 8901 }, { "epoch": 0.75, "learning_rate": 3.012335009112598e-06, "loss": 1.0058, "step": 8902 }, { "epoch": 0.75, "learning_rate": 3.010373045729141e-06, "loss": 0.9035, "step": 8903 }, { "epoch": 0.75, "learning_rate": 3.0084116082602767e-06, "loss": 0.9492, "step": 8904 }, { "epoch": 0.75, "learning_rate": 3.006450696853581e-06, "loss": 0.9846, "step": 8905 }, { "epoch": 0.75, "learning_rate": 3.004490311656605e-06, "loss": 0.9905, "step": 8906 }, { "epoch": 0.75, "learning_rate": 3.002530452816843e-06, "loss": 0.9847, "step": 8907 }, { "epoch": 0.75, "learning_rate": 3.000571120481769e-06, "loss": 1.0012, "step": 8908 }, { "epoch": 0.75, "learning_rate": 2.9986123147987978e-06, "loss": 0.9356, "step": 8909 }, { "epoch": 0.75, "learning_rate": 2.9966540359153227e-06, "loss": 0.9871, "step": 8910 }, { "epoch": 0.75, "learning_rate": 2.9946962839786844e-06, "loss": 0.9683, "step": 8911 }, { "epoch": 0.75, "learning_rate": 2.992739059136186e-06, "loss": 1.0069, "step": 8912 }, { "epoch": 0.75, "learning_rate": 2.9907823615350983e-06, "loss": 0.9662, "step": 8913 }, { "epoch": 0.75, "learning_rate": 2.9888261913226458e-06, "loss": 0.9429, "step": 8914 }, { "epoch": 0.75, "learning_rate": 2.9868705486460113e-06, "loss": 0.9745, "step": 8915 }, { "epoch": 0.76, "learning_rate": 2.984915433652348e-06, "loss": 0.9763, "step": 8916 }, { "epoch": 0.76, "learning_rate": 2.982960846488756e-06, "loss": 0.9922, "step": 8917 }, { "epoch": 0.76, "learning_rate": 2.9810067873023095e-06, "loss": 0.8789, "step": 8918 }, { "epoch": 0.76, "learning_rate": 2.9790532562400308e-06, "loss": 0.9773, "step": 8919 }, { "epoch": 0.76, "learning_rate": 2.9771002534489124e-06, "loss": 0.9504, "step": 8920 }, { "epoch": 0.76, "learning_rate": 2.9751477790758966e-06, "loss": 0.9188, "step": 8921 }, { "epoch": 0.76, "learning_rate": 2.9731958332678977e-06, "loss": 0.9934, "step": 8922 }, { "epoch": 0.76, "learning_rate": 2.971244416171779e-06, "loss": 0.9725, "step": 8923 }, { "epoch": 0.76, "learning_rate": 2.969293527934376e-06, "loss": 0.9466, "step": 8924 }, { "epoch": 0.76, "learning_rate": 2.96734316870247e-06, "loss": 1.0161, "step": 8925 }, { "epoch": 0.76, "learning_rate": 2.965393338622814e-06, "loss": 0.8256, "step": 8926 }, { "epoch": 0.76, "learning_rate": 2.9634440378421204e-06, "loss": 0.9482, "step": 8927 }, { "epoch": 0.76, "learning_rate": 2.961495266507053e-06, "loss": 0.9702, "step": 8928 }, { "epoch": 0.76, "learning_rate": 2.959547024764248e-06, "loss": 0.9426, "step": 8929 }, { "epoch": 0.76, "learning_rate": 2.9575993127602874e-06, "loss": 0.9494, "step": 8930 }, { "epoch": 0.76, "learning_rate": 2.9556521306417284e-06, "loss": 0.9765, "step": 8931 }, { "epoch": 0.76, "learning_rate": 2.9537054785550767e-06, "loss": 0.934, "step": 8932 }, { "epoch": 0.76, "learning_rate": 2.951759356646806e-06, "loss": 0.9476, "step": 8933 }, { "epoch": 0.76, "learning_rate": 2.949813765063342e-06, "loss": 0.9752, "step": 8934 }, { "epoch": 0.76, "learning_rate": 2.9478687039510823e-06, "loss": 0.9567, "step": 8935 }, { "epoch": 0.76, "learning_rate": 2.9459241734563694e-06, "loss": 0.9635, "step": 8936 }, { "epoch": 0.76, "learning_rate": 2.9439801737255202e-06, "loss": 0.9545, "step": 8937 }, { "epoch": 0.76, "learning_rate": 2.9420367049048037e-06, "loss": 1.0196, "step": 8938 }, { "epoch": 0.76, "learning_rate": 2.9400937671404473e-06, "loss": 0.9592, "step": 8939 }, { "epoch": 0.76, "learning_rate": 2.938151360578647e-06, "loss": 0.8227, "step": 8940 }, { "epoch": 0.76, "learning_rate": 2.9362094853655523e-06, "loss": 1.0115, "step": 8941 }, { "epoch": 0.76, "learning_rate": 2.9342681416472695e-06, "loss": 0.9689, "step": 8942 }, { "epoch": 0.76, "learning_rate": 2.932327329569875e-06, "loss": 1.0097, "step": 8943 }, { "epoch": 0.76, "learning_rate": 2.9303870492793962e-06, "loss": 0.9823, "step": 8944 }, { "epoch": 0.76, "learning_rate": 2.9284473009218275e-06, "loss": 0.9383, "step": 8945 }, { "epoch": 0.76, "learning_rate": 2.926508084643115e-06, "loss": 0.9832, "step": 8946 }, { "epoch": 0.76, "learning_rate": 2.9245694005891757e-06, "loss": 0.9822, "step": 8947 }, { "epoch": 0.76, "learning_rate": 2.922631248905873e-06, "loss": 0.9374, "step": 8948 }, { "epoch": 0.76, "learning_rate": 2.920693629739044e-06, "loss": 0.992, "step": 8949 }, { "epoch": 0.76, "learning_rate": 2.9187565432344756e-06, "loss": 0.9374, "step": 8950 }, { "epoch": 0.76, "learning_rate": 2.916819989537921e-06, "loss": 0.9636, "step": 8951 }, { "epoch": 0.76, "learning_rate": 2.914883968795087e-06, "loss": 0.9765, "step": 8952 }, { "epoch": 0.76, "learning_rate": 2.912948481151646e-06, "loss": 1.0078, "step": 8953 }, { "epoch": 0.76, "learning_rate": 2.911013526753231e-06, "loss": 1.0082, "step": 8954 }, { "epoch": 0.76, "learning_rate": 2.9090791057454273e-06, "loss": 0.9335, "step": 8955 }, { "epoch": 0.76, "learning_rate": 2.9071452182737903e-06, "loss": 0.9849, "step": 8956 }, { "epoch": 0.76, "learning_rate": 2.9052118644838235e-06, "loss": 0.982, "step": 8957 }, { "epoch": 0.76, "learning_rate": 2.9032790445210024e-06, "loss": 0.9437, "step": 8958 }, { "epoch": 0.76, "learning_rate": 2.901346758530752e-06, "loss": 0.9948, "step": 8959 }, { "epoch": 0.76, "learning_rate": 2.8994150066584658e-06, "loss": 0.9662, "step": 8960 }, { "epoch": 0.76, "learning_rate": 2.8974837890494876e-06, "loss": 0.981, "step": 8961 }, { "epoch": 0.76, "learning_rate": 2.895553105849134e-06, "loss": 1.0139, "step": 8962 }, { "epoch": 0.76, "learning_rate": 2.893622957202665e-06, "loss": 0.9328, "step": 8963 }, { "epoch": 0.76, "learning_rate": 2.891693343255317e-06, "loss": 0.9862, "step": 8964 }, { "epoch": 0.76, "learning_rate": 2.889764264152276e-06, "loss": 1.0007, "step": 8965 }, { "epoch": 0.76, "learning_rate": 2.887835720038684e-06, "loss": 0.9884, "step": 8966 }, { "epoch": 0.76, "learning_rate": 2.8859077110596577e-06, "loss": 0.9708, "step": 8967 }, { "epoch": 0.76, "learning_rate": 2.883980237360261e-06, "loss": 0.9809, "step": 8968 }, { "epoch": 0.76, "learning_rate": 2.8820532990855177e-06, "loss": 0.9877, "step": 8969 }, { "epoch": 0.76, "learning_rate": 2.8801268963804206e-06, "loss": 0.9366, "step": 8970 }, { "epoch": 0.76, "learning_rate": 2.878201029389911e-06, "loss": 0.9664, "step": 8971 }, { "epoch": 0.76, "learning_rate": 2.8762756982589003e-06, "loss": 0.9915, "step": 8972 }, { "epoch": 0.76, "learning_rate": 2.8743509031322505e-06, "loss": 0.9336, "step": 8973 }, { "epoch": 0.76, "learning_rate": 2.872426644154791e-06, "loss": 0.8576, "step": 8974 }, { "epoch": 0.76, "learning_rate": 2.870502921471302e-06, "loss": 0.9934, "step": 8975 }, { "epoch": 0.76, "learning_rate": 2.8685797352265367e-06, "loss": 0.9511, "step": 8976 }, { "epoch": 0.76, "learning_rate": 2.86665708556519e-06, "loss": 0.9453, "step": 8977 }, { "epoch": 0.76, "learning_rate": 2.864734972631932e-06, "loss": 0.9719, "step": 8978 }, { "epoch": 0.76, "learning_rate": 2.862813396571388e-06, "loss": 0.8435, "step": 8979 }, { "epoch": 0.76, "learning_rate": 2.8608923575281366e-06, "loss": 0.9834, "step": 8980 }, { "epoch": 0.76, "learning_rate": 2.858971855646727e-06, "loss": 0.9296, "step": 8981 }, { "epoch": 0.76, "learning_rate": 2.857051891071655e-06, "loss": 1.0383, "step": 8982 }, { "epoch": 0.76, "learning_rate": 2.85513246394739e-06, "loss": 0.9334, "step": 8983 }, { "epoch": 0.76, "learning_rate": 2.8532135744183466e-06, "loss": 1.003, "step": 8984 }, { "epoch": 0.76, "learning_rate": 2.8512952226289126e-06, "loss": 0.9469, "step": 8985 }, { "epoch": 0.76, "learning_rate": 2.849377408723423e-06, "loss": 1.0102, "step": 8986 }, { "epoch": 0.76, "learning_rate": 2.8474601328461847e-06, "loss": 0.9642, "step": 8987 }, { "epoch": 0.76, "learning_rate": 2.845543395141451e-06, "loss": 1.0057, "step": 8988 }, { "epoch": 0.76, "learning_rate": 2.843627195753448e-06, "loss": 0.9884, "step": 8989 }, { "epoch": 0.76, "learning_rate": 2.841711534826348e-06, "loss": 1.0217, "step": 8990 }, { "epoch": 0.76, "learning_rate": 2.8397964125042955e-06, "loss": 0.9587, "step": 8991 }, { "epoch": 0.76, "learning_rate": 2.837881828931387e-06, "loss": 0.9854, "step": 8992 }, { "epoch": 0.76, "learning_rate": 2.8359677842516743e-06, "loss": 0.9774, "step": 8993 }, { "epoch": 0.76, "learning_rate": 2.834054278609182e-06, "loss": 0.9278, "step": 8994 }, { "epoch": 0.76, "learning_rate": 2.8321413121478825e-06, "loss": 0.9792, "step": 8995 }, { "epoch": 0.76, "learning_rate": 2.83022888501171e-06, "loss": 0.9858, "step": 8996 }, { "epoch": 0.76, "learning_rate": 2.8283169973445635e-06, "loss": 0.9999, "step": 8997 }, { "epoch": 0.76, "learning_rate": 2.8264056492902937e-06, "loss": 1.0043, "step": 8998 }, { "epoch": 0.76, "learning_rate": 2.8244948409927185e-06, "loss": 0.9626, "step": 8999 }, { "epoch": 0.76, "learning_rate": 2.822584572595606e-06, "loss": 0.9586, "step": 9000 }, { "epoch": 0.76, "learning_rate": 2.820674844242696e-06, "loss": 1.0523, "step": 9001 }, { "epoch": 0.76, "learning_rate": 2.8187656560776733e-06, "loss": 0.9108, "step": 9002 }, { "epoch": 0.76, "learning_rate": 2.8168570082441958e-06, "loss": 0.9748, "step": 9003 }, { "epoch": 0.76, "learning_rate": 2.8149489008858676e-06, "loss": 0.9867, "step": 9004 }, { "epoch": 0.76, "learning_rate": 2.8130413341462638e-06, "loss": 0.9522, "step": 9005 }, { "epoch": 0.76, "learning_rate": 2.8111343081689145e-06, "loss": 0.9918, "step": 9006 }, { "epoch": 0.76, "learning_rate": 2.8092278230973037e-06, "loss": 0.8295, "step": 9007 }, { "epoch": 0.76, "learning_rate": 2.8073218790748845e-06, "loss": 0.9473, "step": 9008 }, { "epoch": 0.76, "learning_rate": 2.8054164762450597e-06, "loss": 0.9664, "step": 9009 }, { "epoch": 0.76, "learning_rate": 2.803511614751201e-06, "loss": 1.0069, "step": 9010 }, { "epoch": 0.76, "learning_rate": 2.801607294736628e-06, "loss": 1.0223, "step": 9011 }, { "epoch": 0.76, "learning_rate": 2.799703516344632e-06, "loss": 1.0, "step": 9012 }, { "epoch": 0.76, "learning_rate": 2.797800279718451e-06, "loss": 0.9409, "step": 9013 }, { "epoch": 0.76, "learning_rate": 2.795897585001297e-06, "loss": 0.994, "step": 9014 }, { "epoch": 0.76, "learning_rate": 2.793995432336324e-06, "loss": 1.0239, "step": 9015 }, { "epoch": 0.76, "learning_rate": 2.7920938218666602e-06, "loss": 0.9721, "step": 9016 }, { "epoch": 0.76, "learning_rate": 2.7901927537353867e-06, "loss": 1.0272, "step": 9017 }, { "epoch": 0.76, "learning_rate": 2.7882922280855383e-06, "loss": 0.9498, "step": 9018 }, { "epoch": 0.76, "learning_rate": 2.786392245060121e-06, "loss": 0.9528, "step": 9019 }, { "epoch": 0.76, "learning_rate": 2.784492804802088e-06, "loss": 0.9818, "step": 9020 }, { "epoch": 0.76, "learning_rate": 2.782593907454364e-06, "loss": 0.9583, "step": 9021 }, { "epoch": 0.76, "learning_rate": 2.7806955531598213e-06, "loss": 0.9776, "step": 9022 }, { "epoch": 0.76, "learning_rate": 2.7787977420612956e-06, "loss": 0.9075, "step": 9023 }, { "epoch": 0.76, "learning_rate": 2.776900474301586e-06, "loss": 0.9928, "step": 9024 }, { "epoch": 0.76, "learning_rate": 2.775003750023443e-06, "loss": 0.93, "step": 9025 }, { "epoch": 0.76, "learning_rate": 2.773107569369584e-06, "loss": 0.9553, "step": 9026 }, { "epoch": 0.76, "learning_rate": 2.771211932482677e-06, "loss": 0.9938, "step": 9027 }, { "epoch": 0.76, "learning_rate": 2.7693168395053603e-06, "loss": 0.972, "step": 9028 }, { "epoch": 0.76, "learning_rate": 2.7674222905802186e-06, "loss": 0.9798, "step": 9029 }, { "epoch": 0.76, "learning_rate": 2.765528285849803e-06, "loss": 0.983, "step": 9030 }, { "epoch": 0.76, "learning_rate": 2.763634825456628e-06, "loss": 0.9461, "step": 9031 }, { "epoch": 0.76, "learning_rate": 2.7617419095431553e-06, "loss": 0.9833, "step": 9032 }, { "epoch": 0.76, "learning_rate": 2.7598495382518163e-06, "loss": 0.951, "step": 9033 }, { "epoch": 0.77, "learning_rate": 2.757957711724992e-06, "loss": 0.9593, "step": 9034 }, { "epoch": 0.77, "learning_rate": 2.7560664301050344e-06, "loss": 0.975, "step": 9035 }, { "epoch": 0.77, "learning_rate": 2.7541756935342403e-06, "loss": 0.9539, "step": 9036 }, { "epoch": 0.77, "learning_rate": 2.7522855021548787e-06, "loss": 0.9542, "step": 9037 }, { "epoch": 0.77, "learning_rate": 2.7503958561091658e-06, "loss": 0.9962, "step": 9038 }, { "epoch": 0.77, "learning_rate": 2.7485067555392907e-06, "loss": 0.9942, "step": 9039 }, { "epoch": 0.77, "learning_rate": 2.7466182005873844e-06, "loss": 0.8449, "step": 9040 }, { "epoch": 0.77, "learning_rate": 2.744730191395554e-06, "loss": 0.9618, "step": 9041 }, { "epoch": 0.77, "learning_rate": 2.74284272810585e-06, "loss": 0.9906, "step": 9042 }, { "epoch": 0.77, "learning_rate": 2.7409558108602963e-06, "loss": 0.9762, "step": 9043 }, { "epoch": 0.77, "learning_rate": 2.7390694398008645e-06, "loss": 0.9817, "step": 9044 }, { "epoch": 0.77, "learning_rate": 2.7371836150694874e-06, "loss": 0.9534, "step": 9045 }, { "epoch": 0.77, "learning_rate": 2.735298336808064e-06, "loss": 0.9688, "step": 9046 }, { "epoch": 0.77, "learning_rate": 2.7334136051584435e-06, "loss": 0.9443, "step": 9047 }, { "epoch": 0.77, "learning_rate": 2.731529420262434e-06, "loss": 0.9977, "step": 9048 }, { "epoch": 0.77, "learning_rate": 2.7296457822618115e-06, "loss": 1.0007, "step": 9049 }, { "epoch": 0.77, "learning_rate": 2.7277626912983003e-06, "loss": 0.9894, "step": 9050 }, { "epoch": 0.77, "learning_rate": 2.725880147513593e-06, "loss": 0.9643, "step": 9051 }, { "epoch": 0.77, "learning_rate": 2.7239981510493306e-06, "loss": 0.9689, "step": 9052 }, { "epoch": 0.77, "learning_rate": 2.722116702047125e-06, "loss": 1.0311, "step": 9053 }, { "epoch": 0.77, "learning_rate": 2.7202358006485342e-06, "loss": 0.9547, "step": 9054 }, { "epoch": 0.77, "learning_rate": 2.718355446995086e-06, "loss": 0.9244, "step": 9055 }, { "epoch": 0.77, "learning_rate": 2.716475641228258e-06, "loss": 0.9972, "step": 9056 }, { "epoch": 0.77, "learning_rate": 2.7145963834894926e-06, "loss": 0.9215, "step": 9057 }, { "epoch": 0.77, "learning_rate": 2.7127176739201945e-06, "loss": 0.9845, "step": 9058 }, { "epoch": 0.77, "learning_rate": 2.7108395126617124e-06, "loss": 0.9679, "step": 9059 }, { "epoch": 0.77, "learning_rate": 2.7089618998553713e-06, "loss": 0.9843, "step": 9060 }, { "epoch": 0.77, "learning_rate": 2.7070848356424406e-06, "loss": 0.9651, "step": 9061 }, { "epoch": 0.77, "learning_rate": 2.705208320164161e-06, "loss": 0.9838, "step": 9062 }, { "epoch": 0.77, "learning_rate": 2.7033323535617183e-06, "loss": 1.0071, "step": 9063 }, { "epoch": 0.77, "learning_rate": 2.701456935976272e-06, "loss": 1.0084, "step": 9064 }, { "epoch": 0.77, "learning_rate": 2.699582067548925e-06, "loss": 1.0269, "step": 9065 }, { "epoch": 0.77, "learning_rate": 2.6977077484207536e-06, "loss": 0.9718, "step": 9066 }, { "epoch": 0.77, "learning_rate": 2.695833978732779e-06, "loss": 0.917, "step": 9067 }, { "epoch": 0.77, "learning_rate": 2.693960758625993e-06, "loss": 0.9687, "step": 9068 }, { "epoch": 0.77, "learning_rate": 2.692088088241336e-06, "loss": 0.9516, "step": 9069 }, { "epoch": 0.77, "learning_rate": 2.6902159677197172e-06, "loss": 0.9834, "step": 9070 }, { "epoch": 0.77, "learning_rate": 2.6883443972019964e-06, "loss": 0.9687, "step": 9071 }, { "epoch": 0.77, "learning_rate": 2.6864733768289907e-06, "loss": 0.9838, "step": 9072 }, { "epoch": 0.77, "learning_rate": 2.6846029067414857e-06, "loss": 0.9825, "step": 9073 }, { "epoch": 0.77, "learning_rate": 2.6827329870802188e-06, "loss": 0.9414, "step": 9074 }, { "epoch": 0.77, "learning_rate": 2.6808636179858804e-06, "loss": 0.9713, "step": 9075 }, { "epoch": 0.77, "learning_rate": 2.678994799599134e-06, "loss": 0.9946, "step": 9076 }, { "epoch": 0.77, "learning_rate": 2.6771265320605877e-06, "loss": 0.9778, "step": 9077 }, { "epoch": 0.77, "learning_rate": 2.6752588155108183e-06, "loss": 0.9842, "step": 9078 }, { "epoch": 0.77, "learning_rate": 2.6733916500903535e-06, "loss": 1.0245, "step": 9079 }, { "epoch": 0.77, "learning_rate": 2.671525035939686e-06, "loss": 1.0131, "step": 9080 }, { "epoch": 0.77, "learning_rate": 2.6696589731992583e-06, "loss": 0.9847, "step": 9081 }, { "epoch": 0.77, "learning_rate": 2.6677934620094824e-06, "loss": 0.9257, "step": 9082 }, { "epoch": 0.77, "learning_rate": 2.665928502510724e-06, "loss": 0.9722, "step": 9083 }, { "epoch": 0.77, "learning_rate": 2.664064094843302e-06, "loss": 1.0115, "step": 9084 }, { "epoch": 0.77, "learning_rate": 2.662200239147502e-06, "loss": 0.9948, "step": 9085 }, { "epoch": 0.77, "learning_rate": 2.6603369355635624e-06, "loss": 0.9835, "step": 9086 }, { "epoch": 0.77, "learning_rate": 2.658474184231685e-06, "loss": 0.9837, "step": 9087 }, { "epoch": 0.77, "learning_rate": 2.6566119852920238e-06, "loss": 0.9578, "step": 9088 }, { "epoch": 0.77, "learning_rate": 2.654750338884697e-06, "loss": 0.9265, "step": 9089 }, { "epoch": 0.77, "learning_rate": 2.652889245149777e-06, "loss": 0.9803, "step": 9090 }, { "epoch": 0.77, "learning_rate": 2.6510287042272996e-06, "loss": 0.9396, "step": 9091 }, { "epoch": 0.77, "learning_rate": 2.649168716257251e-06, "loss": 0.8975, "step": 9092 }, { "epoch": 0.77, "learning_rate": 2.6473092813795874e-06, "loss": 1.0366, "step": 9093 }, { "epoch": 0.77, "learning_rate": 2.6454503997342095e-06, "loss": 0.9564, "step": 9094 }, { "epoch": 0.77, "learning_rate": 2.6435920714609898e-06, "loss": 1.0135, "step": 9095 }, { "epoch": 0.77, "learning_rate": 2.6417342966997474e-06, "loss": 0.8601, "step": 9096 }, { "epoch": 0.77, "learning_rate": 2.6398770755902704e-06, "loss": 0.8731, "step": 9097 }, { "epoch": 0.77, "learning_rate": 2.638020408272298e-06, "loss": 1.0414, "step": 9098 }, { "epoch": 0.77, "learning_rate": 2.636164294885527e-06, "loss": 0.9753, "step": 9099 }, { "epoch": 0.77, "learning_rate": 2.6343087355696207e-06, "loss": 0.9844, "step": 9100 }, { "epoch": 0.77, "learning_rate": 2.632453730464193e-06, "loss": 1.0018, "step": 9101 }, { "epoch": 0.77, "learning_rate": 2.6305992797088144e-06, "loss": 0.9641, "step": 9102 }, { "epoch": 0.77, "learning_rate": 2.6287453834430244e-06, "loss": 0.9733, "step": 9103 }, { "epoch": 0.77, "learning_rate": 2.626892041806308e-06, "loss": 0.9921, "step": 9104 }, { "epoch": 0.77, "learning_rate": 2.625039254938121e-06, "loss": 0.9959, "step": 9105 }, { "epoch": 0.77, "learning_rate": 2.6231870229778655e-06, "loss": 0.959, "step": 9106 }, { "epoch": 0.77, "learning_rate": 2.6213353460649114e-06, "loss": 0.9304, "step": 9107 }, { "epoch": 0.77, "learning_rate": 2.61948422433858e-06, "loss": 0.9344, "step": 9108 }, { "epoch": 0.77, "learning_rate": 2.6176336579381535e-06, "loss": 0.96, "step": 9109 }, { "epoch": 0.77, "learning_rate": 2.6157836470028775e-06, "loss": 0.9627, "step": 9110 }, { "epoch": 0.77, "learning_rate": 2.613934191671945e-06, "loss": 0.9939, "step": 9111 }, { "epoch": 0.77, "learning_rate": 2.6120852920845173e-06, "loss": 0.9438, "step": 9112 }, { "epoch": 0.77, "learning_rate": 2.610236948379705e-06, "loss": 0.973, "step": 9113 }, { "epoch": 0.77, "learning_rate": 2.608389160696587e-06, "loss": 0.8352, "step": 9114 }, { "epoch": 0.77, "learning_rate": 2.606541929174189e-06, "loss": 0.9912, "step": 9115 }, { "epoch": 0.77, "learning_rate": 2.604695253951506e-06, "loss": 0.9615, "step": 9116 }, { "epoch": 0.77, "learning_rate": 2.60284913516748e-06, "loss": 0.9499, "step": 9117 }, { "epoch": 0.77, "learning_rate": 2.601003572961024e-06, "loss": 1.0341, "step": 9118 }, { "epoch": 0.77, "learning_rate": 2.599158567470995e-06, "loss": 0.9578, "step": 9119 }, { "epoch": 0.77, "learning_rate": 2.5973141188362225e-06, "loss": 0.9659, "step": 9120 }, { "epoch": 0.77, "learning_rate": 2.595470227195479e-06, "loss": 0.9823, "step": 9121 }, { "epoch": 0.77, "learning_rate": 2.5936268926875097e-06, "loss": 0.9785, "step": 9122 }, { "epoch": 0.77, "learning_rate": 2.591784115451006e-06, "loss": 1.0025, "step": 9123 }, { "epoch": 0.77, "learning_rate": 2.589941895624626e-06, "loss": 0.9604, "step": 9124 }, { "epoch": 0.77, "learning_rate": 2.5881002333469817e-06, "loss": 1.0087, "step": 9125 }, { "epoch": 0.77, "learning_rate": 2.58625912875664e-06, "loss": 0.9719, "step": 9126 }, { "epoch": 0.77, "learning_rate": 2.5844185819921353e-06, "loss": 0.9707, "step": 9127 }, { "epoch": 0.77, "learning_rate": 2.582578593191951e-06, "loss": 0.8088, "step": 9128 }, { "epoch": 0.77, "learning_rate": 2.58073916249453e-06, "loss": 0.9813, "step": 9129 }, { "epoch": 0.77, "learning_rate": 2.57890029003828e-06, "loss": 0.982, "step": 9130 }, { "epoch": 0.77, "learning_rate": 2.5770619759615557e-06, "loss": 0.9017, "step": 9131 }, { "epoch": 0.77, "learning_rate": 2.575224220402682e-06, "loss": 0.8302, "step": 9132 }, { "epoch": 0.77, "learning_rate": 2.5733870234999293e-06, "loss": 0.9389, "step": 9133 }, { "epoch": 0.77, "learning_rate": 2.5715503853915367e-06, "loss": 0.9444, "step": 9134 }, { "epoch": 0.77, "learning_rate": 2.5697143062156984e-06, "loss": 0.9685, "step": 9135 }, { "epoch": 0.77, "learning_rate": 2.5678787861105593e-06, "loss": 0.9981, "step": 9136 }, { "epoch": 0.77, "learning_rate": 2.566043825214234e-06, "loss": 0.993, "step": 9137 }, { "epoch": 0.77, "learning_rate": 2.564209423664782e-06, "loss": 0.9559, "step": 9138 }, { "epoch": 0.77, "learning_rate": 2.5623755816002347e-06, "loss": 0.9475, "step": 9139 }, { "epoch": 0.77, "learning_rate": 2.560542299158568e-06, "loss": 0.986, "step": 9140 }, { "epoch": 0.77, "learning_rate": 2.5587095764777293e-06, "loss": 0.9368, "step": 9141 }, { "epoch": 0.77, "learning_rate": 2.5568774136956086e-06, "loss": 0.9495, "step": 9142 }, { "epoch": 0.77, "learning_rate": 2.5550458109500686e-06, "loss": 1.048, "step": 9143 }, { "epoch": 0.77, "learning_rate": 2.5532147683789166e-06, "loss": 0.9602, "step": 9144 }, { "epoch": 0.77, "learning_rate": 2.551384286119931e-06, "loss": 1.0217, "step": 9145 }, { "epoch": 0.77, "learning_rate": 2.5495543643108346e-06, "loss": 0.9417, "step": 9146 }, { "epoch": 0.77, "learning_rate": 2.547725003089321e-06, "loss": 1.023, "step": 9147 }, { "epoch": 0.77, "learning_rate": 2.545896202593029e-06, "loss": 0.9641, "step": 9148 }, { "epoch": 0.77, "learning_rate": 2.5440679629595676e-06, "loss": 0.9894, "step": 9149 }, { "epoch": 0.77, "learning_rate": 2.5422402843264914e-06, "loss": 0.9722, "step": 9150 }, { "epoch": 0.77, "learning_rate": 2.540413166831326e-06, "loss": 0.9749, "step": 9151 }, { "epoch": 0.78, "learning_rate": 2.538586610611542e-06, "loss": 1.0018, "step": 9152 }, { "epoch": 0.78, "learning_rate": 2.5367606158045733e-06, "loss": 0.9937, "step": 9153 }, { "epoch": 0.78, "learning_rate": 2.5349351825478164e-06, "loss": 0.9711, "step": 9154 }, { "epoch": 0.78, "learning_rate": 2.533110310978617e-06, "loss": 0.9486, "step": 9155 }, { "epoch": 0.78, "learning_rate": 2.531286001234281e-06, "loss": 0.9409, "step": 9156 }, { "epoch": 0.78, "learning_rate": 2.5294622534520787e-06, "loss": 0.9567, "step": 9157 }, { "epoch": 0.78, "learning_rate": 2.5276390677692263e-06, "loss": 0.9848, "step": 9158 }, { "epoch": 0.78, "learning_rate": 2.525816444322912e-06, "loss": 0.9207, "step": 9159 }, { "epoch": 0.78, "learning_rate": 2.5239943832502643e-06, "loss": 0.9894, "step": 9160 }, { "epoch": 0.78, "learning_rate": 2.5221728846883863e-06, "loss": 0.9259, "step": 9161 }, { "epoch": 0.78, "learning_rate": 2.520351948774331e-06, "loss": 1.0076, "step": 9162 }, { "epoch": 0.78, "learning_rate": 2.5185315756451066e-06, "loss": 0.9422, "step": 9163 }, { "epoch": 0.78, "learning_rate": 2.516711765437685e-06, "loss": 1.0192, "step": 9164 }, { "epoch": 0.78, "learning_rate": 2.514892518288988e-06, "loss": 1.0105, "step": 9165 }, { "epoch": 0.78, "learning_rate": 2.513073834335906e-06, "loss": 0.9791, "step": 9166 }, { "epoch": 0.78, "learning_rate": 2.511255713715275e-06, "loss": 0.9709, "step": 9167 }, { "epoch": 0.78, "learning_rate": 2.509438156563899e-06, "loss": 0.9572, "step": 9168 }, { "epoch": 0.78, "learning_rate": 2.50762116301853e-06, "loss": 0.9654, "step": 9169 }, { "epoch": 0.78, "learning_rate": 2.5058047332158874e-06, "loss": 0.9441, "step": 9170 }, { "epoch": 0.78, "learning_rate": 2.503988867292638e-06, "loss": 0.9842, "step": 9171 }, { "epoch": 0.78, "learning_rate": 2.5021735653854183e-06, "loss": 0.9814, "step": 9172 }, { "epoch": 0.78, "learning_rate": 2.5003588276308075e-06, "loss": 0.9697, "step": 9173 }, { "epoch": 0.78, "learning_rate": 2.498544654165358e-06, "loss": 0.9234, "step": 9174 }, { "epoch": 0.78, "learning_rate": 2.4967310451255655e-06, "loss": 1.0354, "step": 9175 }, { "epoch": 0.78, "learning_rate": 2.494918000647897e-06, "loss": 0.9939, "step": 9176 }, { "epoch": 0.78, "learning_rate": 2.4931055208687615e-06, "loss": 0.9536, "step": 9177 }, { "epoch": 0.78, "learning_rate": 2.4912936059245417e-06, "loss": 1.0311, "step": 9178 }, { "epoch": 0.78, "learning_rate": 2.4894822559515663e-06, "loss": 0.9544, "step": 9179 }, { "epoch": 0.78, "learning_rate": 2.487671471086123e-06, "loss": 0.9669, "step": 9180 }, { "epoch": 0.78, "learning_rate": 2.485861251464464e-06, "loss": 1.0085, "step": 9181 }, { "epoch": 0.78, "learning_rate": 2.484051597222792e-06, "loss": 0.955, "step": 9182 }, { "epoch": 0.78, "learning_rate": 2.482242508497267e-06, "loss": 0.9596, "step": 9183 }, { "epoch": 0.78, "learning_rate": 2.4804339854240123e-06, "loss": 0.9491, "step": 9184 }, { "epoch": 0.78, "learning_rate": 2.4786260281391004e-06, "loss": 0.9112, "step": 9185 }, { "epoch": 0.78, "learning_rate": 2.4768186367785694e-06, "loss": 0.9501, "step": 9186 }, { "epoch": 0.78, "learning_rate": 2.4750118114784137e-06, "loss": 0.9547, "step": 9187 }, { "epoch": 0.78, "learning_rate": 2.4732055523745767e-06, "loss": 0.9769, "step": 9188 }, { "epoch": 0.78, "learning_rate": 2.471399859602972e-06, "loss": 0.9741, "step": 9189 }, { "epoch": 0.78, "learning_rate": 2.4695947332994574e-06, "loss": 0.9549, "step": 9190 }, { "epoch": 0.78, "learning_rate": 2.46779017359986e-06, "loss": 0.9316, "step": 9191 }, { "epoch": 0.78, "learning_rate": 2.4659861806399533e-06, "loss": 0.9729, "step": 9192 }, { "epoch": 0.78, "learning_rate": 2.46418275455548e-06, "loss": 0.9807, "step": 9193 }, { "epoch": 0.78, "learning_rate": 2.4623798954821256e-06, "loss": 0.9748, "step": 9194 }, { "epoch": 0.78, "learning_rate": 2.4605776035555494e-06, "loss": 0.9831, "step": 9195 }, { "epoch": 0.78, "learning_rate": 2.4587758789113537e-06, "loss": 1.0088, "step": 9196 }, { "epoch": 0.78, "learning_rate": 2.456974721685109e-06, "loss": 0.9759, "step": 9197 }, { "epoch": 0.78, "learning_rate": 2.4551741320123324e-06, "loss": 0.9358, "step": 9198 }, { "epoch": 0.78, "learning_rate": 2.4533741100285115e-06, "loss": 0.9373, "step": 9199 }, { "epoch": 0.78, "learning_rate": 2.4515746558690757e-06, "loss": 0.9967, "step": 9200 }, { "epoch": 0.78, "learning_rate": 2.449775769669427e-06, "loss": 0.9465, "step": 9201 }, { "epoch": 0.78, "learning_rate": 2.4479774515649114e-06, "loss": 0.9762, "step": 9202 }, { "epoch": 0.78, "learning_rate": 2.446179701690845e-06, "loss": 0.9488, "step": 9203 }, { "epoch": 0.78, "learning_rate": 2.444382520182487e-06, "loss": 1.0229, "step": 9204 }, { "epoch": 0.78, "learning_rate": 2.442585907175068e-06, "loss": 0.9614, "step": 9205 }, { "epoch": 0.78, "learning_rate": 2.440789862803765e-06, "loss": 0.987, "step": 9206 }, { "epoch": 0.78, "learning_rate": 2.4389943872037147e-06, "loss": 0.9183, "step": 9207 }, { "epoch": 0.78, "learning_rate": 2.4371994805100175e-06, "loss": 1.0016, "step": 9208 }, { "epoch": 0.78, "learning_rate": 2.435405142857724e-06, "loss": 0.9735, "step": 9209 }, { "epoch": 0.78, "learning_rate": 2.43361137438184e-06, "loss": 0.9642, "step": 9210 }, { "epoch": 0.78, "learning_rate": 2.4318181752173365e-06, "loss": 0.9044, "step": 9211 }, { "epoch": 0.78, "learning_rate": 2.4300255454991407e-06, "loss": 0.9925, "step": 9212 }, { "epoch": 0.78, "learning_rate": 2.4282334853621258e-06, "loss": 0.961, "step": 9213 }, { "epoch": 0.78, "learning_rate": 2.426441994941139e-06, "loss": 0.9991, "step": 9214 }, { "epoch": 0.78, "learning_rate": 2.4246510743709683e-06, "loss": 1.0203, "step": 9215 }, { "epoch": 0.78, "learning_rate": 2.422860723786373e-06, "loss": 0.9958, "step": 9216 }, { "epoch": 0.78, "learning_rate": 2.421070943322057e-06, "loss": 1.0322, "step": 9217 }, { "epoch": 0.78, "learning_rate": 2.4192817331126926e-06, "loss": 0.9467, "step": 9218 }, { "epoch": 0.78, "learning_rate": 2.4174930932928985e-06, "loss": 0.9307, "step": 9219 }, { "epoch": 0.78, "learning_rate": 2.4157050239972624e-06, "loss": 0.9493, "step": 9220 }, { "epoch": 0.78, "learning_rate": 2.413917525360314e-06, "loss": 1.0139, "step": 9221 }, { "epoch": 0.78, "learning_rate": 2.4121305975165576e-06, "loss": 1.0049, "step": 9222 }, { "epoch": 0.78, "learning_rate": 2.4103442406004375e-06, "loss": 0.9724, "step": 9223 }, { "epoch": 0.78, "learning_rate": 2.4085584547463702e-06, "loss": 0.9563, "step": 9224 }, { "epoch": 0.78, "learning_rate": 2.4067732400887156e-06, "loss": 0.9823, "step": 9225 }, { "epoch": 0.78, "learning_rate": 2.4049885967618024e-06, "loss": 0.9861, "step": 9226 }, { "epoch": 0.78, "learning_rate": 2.403204524899906e-06, "loss": 0.9974, "step": 9227 }, { "epoch": 0.78, "learning_rate": 2.4014210246372695e-06, "loss": 0.9323, "step": 9228 }, { "epoch": 0.78, "learning_rate": 2.39963809610808e-06, "loss": 0.92, "step": 9229 }, { "epoch": 0.78, "learning_rate": 2.3978557394464973e-06, "loss": 0.9681, "step": 9230 }, { "epoch": 0.78, "learning_rate": 2.396073954786623e-06, "loss": 0.9708, "step": 9231 }, { "epoch": 0.78, "learning_rate": 2.394292742262527e-06, "loss": 0.9621, "step": 9232 }, { "epoch": 0.78, "learning_rate": 2.39251210200823e-06, "loss": 0.9614, "step": 9233 }, { "epoch": 0.78, "learning_rate": 2.390732034157708e-06, "loss": 1.0114, "step": 9234 }, { "epoch": 0.78, "learning_rate": 2.388952538844903e-06, "loss": 0.9295, "step": 9235 }, { "epoch": 0.78, "learning_rate": 2.3871736162037053e-06, "loss": 0.9447, "step": 9236 }, { "epoch": 0.78, "learning_rate": 2.38539526636796e-06, "loss": 0.9227, "step": 9237 }, { "epoch": 0.78, "learning_rate": 2.3836174894714803e-06, "loss": 1.012, "step": 9238 }, { "epoch": 0.78, "learning_rate": 2.3818402856480303e-06, "loss": 1.0092, "step": 9239 }, { "epoch": 0.78, "learning_rate": 2.380063655031326e-06, "loss": 0.9951, "step": 9240 }, { "epoch": 0.78, "learning_rate": 2.3782875977550502e-06, "loss": 0.9794, "step": 9241 }, { "epoch": 0.78, "learning_rate": 2.3765121139528314e-06, "loss": 1.0041, "step": 9242 }, { "epoch": 0.78, "learning_rate": 2.374737203758267e-06, "loss": 0.9404, "step": 9243 }, { "epoch": 0.78, "learning_rate": 2.372962867304899e-06, "loss": 0.9365, "step": 9244 }, { "epoch": 0.78, "learning_rate": 2.371189104726238e-06, "loss": 0.999, "step": 9245 }, { "epoch": 0.78, "learning_rate": 2.369415916155741e-06, "loss": 1.0061, "step": 9246 }, { "epoch": 0.78, "learning_rate": 2.3676433017268308e-06, "loss": 0.9796, "step": 9247 }, { "epoch": 0.78, "learning_rate": 2.3658712615728775e-06, "loss": 1.0564, "step": 9248 }, { "epoch": 0.78, "learning_rate": 2.3640997958272193e-06, "loss": 0.9527, "step": 9249 }, { "epoch": 0.78, "learning_rate": 2.362328904623139e-06, "loss": 0.9923, "step": 9250 }, { "epoch": 0.78, "learning_rate": 2.3605585880938885e-06, "loss": 0.9906, "step": 9251 }, { "epoch": 0.78, "learning_rate": 2.3587888463726637e-06, "loss": 0.9259, "step": 9252 }, { "epoch": 0.78, "learning_rate": 2.3570196795926304e-06, "loss": 0.9077, "step": 9253 }, { "epoch": 0.78, "learning_rate": 2.3552510878868984e-06, "loss": 0.9443, "step": 9254 }, { "epoch": 0.78, "learning_rate": 2.353483071388546e-06, "loss": 0.9788, "step": 9255 }, { "epoch": 0.78, "learning_rate": 2.3517156302305967e-06, "loss": 0.9936, "step": 9256 }, { "epoch": 0.78, "learning_rate": 2.3499487645460415e-06, "loss": 0.9413, "step": 9257 }, { "epoch": 0.78, "learning_rate": 2.3481824744678226e-06, "loss": 1.0005, "step": 9258 }, { "epoch": 0.78, "learning_rate": 2.3464167601288357e-06, "loss": 0.9349, "step": 9259 }, { "epoch": 0.78, "learning_rate": 2.344651621661941e-06, "loss": 0.9176, "step": 9260 }, { "epoch": 0.78, "learning_rate": 2.3428870591999474e-06, "loss": 0.9307, "step": 9261 }, { "epoch": 0.78, "learning_rate": 2.3411230728756294e-06, "loss": 1.0044, "step": 9262 }, { "epoch": 0.78, "learning_rate": 2.3393596628217075e-06, "loss": 1.0301, "step": 9263 }, { "epoch": 0.78, "learning_rate": 2.337596829170871e-06, "loss": 0.9338, "step": 9264 }, { "epoch": 0.78, "learning_rate": 2.335834572055752e-06, "loss": 0.9677, "step": 9265 }, { "epoch": 0.78, "learning_rate": 2.334072891608954e-06, "loss": 0.9961, "step": 9266 }, { "epoch": 0.78, "learning_rate": 2.332311787963022e-06, "loss": 0.9755, "step": 9267 }, { "epoch": 0.78, "learning_rate": 2.330551261250472e-06, "loss": 0.9088, "step": 9268 }, { "epoch": 0.78, "learning_rate": 2.328791311603765e-06, "loss": 0.9415, "step": 9269 }, { "epoch": 0.79, "learning_rate": 2.3270319391553275e-06, "loss": 0.9967, "step": 9270 }, { "epoch": 0.79, "learning_rate": 2.3252731440375333e-06, "loss": 0.9717, "step": 9271 }, { "epoch": 0.79, "learning_rate": 2.323514926382724e-06, "loss": 0.9974, "step": 9272 }, { "epoch": 0.79, "learning_rate": 2.3217572863231865e-06, "loss": 0.9433, "step": 9273 }, { "epoch": 0.79, "learning_rate": 2.3200002239911733e-06, "loss": 0.977, "step": 9274 }, { "epoch": 0.79, "learning_rate": 2.3182437395188852e-06, "loss": 0.9822, "step": 9275 }, { "epoch": 0.79, "learning_rate": 2.31648783303849e-06, "loss": 1.0195, "step": 9276 }, { "epoch": 0.79, "learning_rate": 2.314732504682099e-06, "loss": 0.9537, "step": 9277 }, { "epoch": 0.79, "learning_rate": 2.312977754581793e-06, "loss": 0.9416, "step": 9278 }, { "epoch": 0.79, "learning_rate": 2.3112235828695973e-06, "loss": 1.0108, "step": 9279 }, { "epoch": 0.79, "learning_rate": 2.3094699896775065e-06, "loss": 0.9944, "step": 9280 }, { "epoch": 0.79, "learning_rate": 2.307716975137457e-06, "loss": 0.9463, "step": 9281 }, { "epoch": 0.79, "learning_rate": 2.305964539381357e-06, "loss": 0.9428, "step": 9282 }, { "epoch": 0.79, "learning_rate": 2.3042126825410562e-06, "loss": 0.9118, "step": 9283 }, { "epoch": 0.79, "learning_rate": 2.302461404748375e-06, "loss": 1.0198, "step": 9284 }, { "epoch": 0.79, "learning_rate": 2.300710706135081e-06, "loss": 0.9946, "step": 9285 }, { "epoch": 0.79, "learning_rate": 2.2989605868328946e-06, "loss": 0.983, "step": 9286 }, { "epoch": 0.79, "learning_rate": 2.2972110469735075e-06, "loss": 0.9892, "step": 9287 }, { "epoch": 0.79, "learning_rate": 2.295462086688556e-06, "loss": 0.9989, "step": 9288 }, { "epoch": 0.79, "learning_rate": 2.2937137061096303e-06, "loss": 0.9553, "step": 9289 }, { "epoch": 0.79, "learning_rate": 2.2919659053682862e-06, "loss": 0.9727, "step": 9290 }, { "epoch": 0.79, "learning_rate": 2.290218684596036e-06, "loss": 0.9856, "step": 9291 }, { "epoch": 0.79, "learning_rate": 2.288472043924338e-06, "loss": 1.0105, "step": 9292 }, { "epoch": 0.79, "learning_rate": 2.286725983484618e-06, "loss": 0.9562, "step": 9293 }, { "epoch": 0.79, "learning_rate": 2.2849805034082485e-06, "loss": 0.9207, "step": 9294 }, { "epoch": 0.79, "learning_rate": 2.283235603826569e-06, "loss": 0.9831, "step": 9295 }, { "epoch": 0.79, "learning_rate": 2.2814912848708636e-06, "loss": 0.9897, "step": 9296 }, { "epoch": 0.79, "learning_rate": 2.279747546672384e-06, "loss": 0.9603, "step": 9297 }, { "epoch": 0.79, "learning_rate": 2.2780043893623283e-06, "loss": 0.9816, "step": 9298 }, { "epoch": 0.79, "learning_rate": 2.2762618130718593e-06, "loss": 1.027, "step": 9299 }, { "epoch": 0.79, "learning_rate": 2.274519817932087e-06, "loss": 1.0176, "step": 9300 }, { "epoch": 0.79, "learning_rate": 2.2727784040740896e-06, "loss": 0.998, "step": 9301 }, { "epoch": 0.79, "learning_rate": 2.271037571628889e-06, "loss": 0.9912, "step": 9302 }, { "epoch": 0.79, "learning_rate": 2.269297320727474e-06, "loss": 0.9868, "step": 9303 }, { "epoch": 0.79, "learning_rate": 2.267557651500779e-06, "loss": 0.9713, "step": 9304 }, { "epoch": 0.79, "learning_rate": 2.265818564079707e-06, "loss": 0.9974, "step": 9305 }, { "epoch": 0.79, "learning_rate": 2.264080058595105e-06, "loss": 0.9578, "step": 9306 }, { "epoch": 0.79, "learning_rate": 2.2623421351777875e-06, "loss": 0.9739, "step": 9307 }, { "epoch": 0.79, "learning_rate": 2.260604793958513e-06, "loss": 0.9342, "step": 9308 }, { "epoch": 0.79, "learning_rate": 2.2588680350680094e-06, "loss": 0.9771, "step": 9309 }, { "epoch": 0.79, "learning_rate": 2.2571318586369486e-06, "loss": 0.9872, "step": 9310 }, { "epoch": 0.79, "learning_rate": 2.2553962647959683e-06, "loss": 0.8249, "step": 9311 }, { "epoch": 0.79, "learning_rate": 2.2536612536756575e-06, "loss": 0.9773, "step": 9312 }, { "epoch": 0.79, "learning_rate": 2.2519268254065595e-06, "loss": 1.0192, "step": 9313 }, { "epoch": 0.79, "learning_rate": 2.25019298011918e-06, "loss": 0.9851, "step": 9314 }, { "epoch": 0.79, "learning_rate": 2.2484597179439738e-06, "loss": 0.9292, "step": 9315 }, { "epoch": 0.79, "learning_rate": 2.2467270390113582e-06, "loss": 0.9562, "step": 9316 }, { "epoch": 0.79, "learning_rate": 2.2449949434517016e-06, "loss": 0.9883, "step": 9317 }, { "epoch": 0.79, "learning_rate": 2.2432634313953338e-06, "loss": 0.9929, "step": 9318 }, { "epoch": 0.79, "learning_rate": 2.2415325029725323e-06, "loss": 0.9905, "step": 9319 }, { "epoch": 0.79, "learning_rate": 2.239802158313541e-06, "loss": 0.9541, "step": 9320 }, { "epoch": 0.79, "learning_rate": 2.2380723975485508e-06, "loss": 0.9997, "step": 9321 }, { "epoch": 0.79, "learning_rate": 2.2363432208077172e-06, "loss": 0.9856, "step": 9322 }, { "epoch": 0.79, "learning_rate": 2.234614628221141e-06, "loss": 0.9871, "step": 9323 }, { "epoch": 0.79, "learning_rate": 2.2328866199188924e-06, "loss": 0.9646, "step": 9324 }, { "epoch": 0.79, "learning_rate": 2.231159196030983e-06, "loss": 1.0041, "step": 9325 }, { "epoch": 0.79, "learning_rate": 2.229432356687394e-06, "loss": 0.975, "step": 9326 }, { "epoch": 0.79, "learning_rate": 2.2277061020180534e-06, "loss": 0.9675, "step": 9327 }, { "epoch": 0.79, "learning_rate": 2.22598043215285e-06, "loss": 0.998, "step": 9328 }, { "epoch": 0.79, "learning_rate": 2.2242553472216244e-06, "loss": 0.8732, "step": 9329 }, { "epoch": 0.79, "learning_rate": 2.2225308473541784e-06, "loss": 0.9892, "step": 9330 }, { "epoch": 0.79, "learning_rate": 2.220806932680265e-06, "loss": 1.0323, "step": 9331 }, { "epoch": 0.79, "learning_rate": 2.2190836033295983e-06, "loss": 1.0096, "step": 9332 }, { "epoch": 0.79, "learning_rate": 2.2173608594318396e-06, "loss": 0.9441, "step": 9333 }, { "epoch": 0.79, "learning_rate": 2.215638701116619e-06, "loss": 0.9575, "step": 9334 }, { "epoch": 0.79, "learning_rate": 2.21391712851351e-06, "loss": 0.9709, "step": 9335 }, { "epoch": 0.79, "learning_rate": 2.212196141752051e-06, "loss": 0.9718, "step": 9336 }, { "epoch": 0.79, "learning_rate": 2.2104757409617283e-06, "loss": 0.9563, "step": 9337 }, { "epoch": 0.79, "learning_rate": 2.208755926271995e-06, "loss": 0.9735, "step": 9338 }, { "epoch": 0.79, "learning_rate": 2.2070366978122494e-06, "loss": 0.9421, "step": 9339 }, { "epoch": 0.79, "learning_rate": 2.205318055711848e-06, "loss": 0.9426, "step": 9340 }, { "epoch": 0.79, "learning_rate": 2.20360000010011e-06, "loss": 0.9917, "step": 9341 }, { "epoch": 0.79, "learning_rate": 2.201882531106301e-06, "loss": 0.9456, "step": 9342 }, { "epoch": 0.79, "learning_rate": 2.2001656488596512e-06, "loss": 0.8041, "step": 9343 }, { "epoch": 0.79, "learning_rate": 2.198449353489339e-06, "loss": 0.9471, "step": 9344 }, { "epoch": 0.79, "learning_rate": 2.196733645124506e-06, "loss": 1.0062, "step": 9345 }, { "epoch": 0.79, "learning_rate": 2.195018523894241e-06, "loss": 0.9845, "step": 9346 }, { "epoch": 0.79, "learning_rate": 2.1933039899276e-06, "loss": 0.8884, "step": 9347 }, { "epoch": 0.79, "learning_rate": 2.19159004335358e-06, "loss": 0.9687, "step": 9348 }, { "epoch": 0.79, "learning_rate": 2.18987668430115e-06, "loss": 0.9954, "step": 9349 }, { "epoch": 0.79, "learning_rate": 2.1881639128992195e-06, "loss": 0.9842, "step": 9350 }, { "epoch": 0.79, "learning_rate": 2.1864517292766675e-06, "loss": 0.9831, "step": 9351 }, { "epoch": 0.79, "learning_rate": 2.1847401335623175e-06, "loss": 0.9319, "step": 9352 }, { "epoch": 0.79, "learning_rate": 2.1830291258849578e-06, "loss": 0.9724, "step": 9353 }, { "epoch": 0.79, "learning_rate": 2.1813187063733243e-06, "loss": 0.9667, "step": 9354 }, { "epoch": 0.79, "learning_rate": 2.1796088751561175e-06, "loss": 0.9551, "step": 9355 }, { "epoch": 0.79, "learning_rate": 2.1778996323619828e-06, "loss": 0.9517, "step": 9356 }, { "epoch": 0.79, "learning_rate": 2.1761909781195334e-06, "loss": 0.9209, "step": 9357 }, { "epoch": 0.79, "learning_rate": 2.1744829125573265e-06, "loss": 1.0269, "step": 9358 }, { "epoch": 0.79, "learning_rate": 2.1727754358038856e-06, "loss": 0.9976, "step": 9359 }, { "epoch": 0.79, "learning_rate": 2.171068547987679e-06, "loss": 1.0098, "step": 9360 }, { "epoch": 0.79, "learning_rate": 2.1693622492371448e-06, "loss": 0.9896, "step": 9361 }, { "epoch": 0.79, "learning_rate": 2.1676565396806604e-06, "loss": 0.9892, "step": 9362 }, { "epoch": 0.79, "learning_rate": 2.1659514194465737e-06, "loss": 1.001, "step": 9363 }, { "epoch": 0.79, "learning_rate": 2.164246888663175e-06, "loss": 0.9859, "step": 9364 }, { "epoch": 0.79, "learning_rate": 2.1625429474587243e-06, "loss": 0.9703, "step": 9365 }, { "epoch": 0.79, "learning_rate": 2.160839595961426e-06, "loss": 1.0112, "step": 9366 }, { "epoch": 0.79, "learning_rate": 2.15913683429944e-06, "loss": 0.9875, "step": 9367 }, { "epoch": 0.79, "learning_rate": 2.1574346626008937e-06, "loss": 0.9516, "step": 9368 }, { "epoch": 0.79, "learning_rate": 2.1557330809938548e-06, "loss": 0.9237, "step": 9369 }, { "epoch": 0.79, "learning_rate": 2.1540320896063603e-06, "loss": 0.919, "step": 9370 }, { "epoch": 0.79, "learning_rate": 2.1523316885663905e-06, "loss": 0.9716, "step": 9371 }, { "epoch": 0.79, "learning_rate": 2.150631878001893e-06, "loss": 0.983, "step": 9372 }, { "epoch": 0.79, "learning_rate": 2.1489326580407588e-06, "loss": 0.9072, "step": 9373 }, { "epoch": 0.79, "learning_rate": 2.1472340288108483e-06, "loss": 0.9994, "step": 9374 }, { "epoch": 0.79, "learning_rate": 2.1455359904399633e-06, "loss": 0.9835, "step": 9375 }, { "epoch": 0.79, "learning_rate": 2.143838543055874e-06, "loss": 1.0045, "step": 9376 }, { "epoch": 0.79, "learning_rate": 2.142141686786293e-06, "loss": 0.9934, "step": 9377 }, { "epoch": 0.79, "learning_rate": 2.1404454217589023e-06, "loss": 0.8013, "step": 9378 }, { "epoch": 0.79, "learning_rate": 2.1387497481013276e-06, "loss": 0.9805, "step": 9379 }, { "epoch": 0.79, "learning_rate": 2.1370546659411582e-06, "loss": 1.0051, "step": 9380 }, { "epoch": 0.79, "learning_rate": 2.1353601754059316e-06, "loss": 0.8208, "step": 9381 }, { "epoch": 0.79, "learning_rate": 2.133666276623152e-06, "loss": 0.9823, "step": 9382 }, { "epoch": 0.79, "learning_rate": 2.131972969720264e-06, "loss": 0.9886, "step": 9383 }, { "epoch": 0.79, "learning_rate": 2.1302802548246813e-06, "loss": 0.9434, "step": 9384 }, { "epoch": 0.79, "learning_rate": 2.1285881320637638e-06, "loss": 0.9978, "step": 9385 }, { "epoch": 0.79, "learning_rate": 2.126896601564835e-06, "loss": 0.9592, "step": 9386 }, { "epoch": 0.79, "learning_rate": 2.1252056634551632e-06, "loss": 0.9707, "step": 9387 }, { "epoch": 0.8, "learning_rate": 2.123515317861985e-06, "loss": 0.9736, "step": 9388 }, { "epoch": 0.8, "learning_rate": 2.1218255649124787e-06, "loss": 1.0171, "step": 9389 }, { "epoch": 0.8, "learning_rate": 2.1201364047337903e-06, "loss": 0.9735, "step": 9390 }, { "epoch": 0.8, "learning_rate": 2.118447837453013e-06, "loss": 0.956, "step": 9391 }, { "epoch": 0.8, "learning_rate": 2.1167598631972e-06, "loss": 0.9667, "step": 9392 }, { "epoch": 0.8, "learning_rate": 2.115072482093359e-06, "loss": 0.9504, "step": 9393 }, { "epoch": 0.8, "learning_rate": 2.1133856942684463e-06, "loss": 0.976, "step": 9394 }, { "epoch": 0.8, "learning_rate": 2.111699499849387e-06, "loss": 0.9894, "step": 9395 }, { "epoch": 0.8, "learning_rate": 2.110013898963048e-06, "loss": 0.9485, "step": 9396 }, { "epoch": 0.8, "learning_rate": 2.1083288917362623e-06, "loss": 0.9785, "step": 9397 }, { "epoch": 0.8, "learning_rate": 2.1066444782958094e-06, "loss": 0.9442, "step": 9398 }, { "epoch": 0.8, "learning_rate": 2.1049606587684325e-06, "loss": 0.974, "step": 9399 }, { "epoch": 0.8, "learning_rate": 2.1032774332808203e-06, "loss": 0.9662, "step": 9400 }, { "epoch": 0.8, "learning_rate": 2.101594801959629e-06, "loss": 0.955, "step": 9401 }, { "epoch": 0.8, "learning_rate": 2.0999127649314564e-06, "loss": 1.0036, "step": 9402 }, { "epoch": 0.8, "learning_rate": 2.098231322322869e-06, "loss": 0.9258, "step": 9403 }, { "epoch": 0.8, "learning_rate": 2.096550474260376e-06, "loss": 0.9711, "step": 9404 }, { "epoch": 0.8, "learning_rate": 2.094870220870455e-06, "loss": 0.9307, "step": 9405 }, { "epoch": 0.8, "learning_rate": 2.0931905622795247e-06, "loss": 0.9881, "step": 9406 }, { "epoch": 0.8, "learning_rate": 2.091511498613973e-06, "loss": 0.975, "step": 9407 }, { "epoch": 0.8, "learning_rate": 2.0898330300001292e-06, "loss": 0.8025, "step": 9408 }, { "epoch": 0.8, "learning_rate": 2.0881551565642922e-06, "loss": 0.9499, "step": 9409 }, { "epoch": 0.8, "learning_rate": 2.086477878432701e-06, "loss": 0.993, "step": 9410 }, { "epoch": 0.8, "learning_rate": 2.084801195731566e-06, "loss": 1.0111, "step": 9411 }, { "epoch": 0.8, "learning_rate": 2.0831251085870375e-06, "loss": 0.9797, "step": 9412 }, { "epoch": 0.8, "learning_rate": 2.081449617125233e-06, "loss": 0.9824, "step": 9413 }, { "epoch": 0.8, "learning_rate": 2.0797747214722154e-06, "loss": 0.9673, "step": 9414 }, { "epoch": 0.8, "learning_rate": 2.0781004217540124e-06, "loss": 0.9832, "step": 9415 }, { "epoch": 0.8, "learning_rate": 2.076426718096596e-06, "loss": 0.9312, "step": 9416 }, { "epoch": 0.8, "learning_rate": 2.0747536106259058e-06, "loss": 0.9814, "step": 9417 }, { "epoch": 0.8, "learning_rate": 2.073081099467824e-06, "loss": 0.989, "step": 9418 }, { "epoch": 0.8, "learning_rate": 2.0714091847482e-06, "loss": 1.0156, "step": 9419 }, { "epoch": 0.8, "learning_rate": 2.069737866592828e-06, "loss": 0.9762, "step": 9420 }, { "epoch": 0.8, "learning_rate": 2.06806714512746e-06, "loss": 0.975, "step": 9421 }, { "epoch": 0.8, "learning_rate": 2.0663970204778105e-06, "loss": 0.9126, "step": 9422 }, { "epoch": 0.8, "learning_rate": 2.064727492769538e-06, "loss": 0.987, "step": 9423 }, { "epoch": 0.8, "learning_rate": 2.063058562128265e-06, "loss": 0.9854, "step": 9424 }, { "epoch": 0.8, "learning_rate": 2.061390228679562e-06, "loss": 0.943, "step": 9425 }, { "epoch": 0.8, "learning_rate": 2.059722492548961e-06, "loss": 0.9421, "step": 9426 }, { "epoch": 0.8, "learning_rate": 2.058055353861944e-06, "loss": 0.9379, "step": 9427 }, { "epoch": 0.8, "learning_rate": 2.0563888127439524e-06, "loss": 0.9769, "step": 9428 }, { "epoch": 0.8, "learning_rate": 2.054722869320377e-06, "loss": 1.0335, "step": 9429 }, { "epoch": 0.8, "learning_rate": 2.0530575237165717e-06, "loss": 0.9329, "step": 9430 }, { "epoch": 0.8, "learning_rate": 2.0513927760578354e-06, "loss": 1.0182, "step": 9431 }, { "epoch": 0.8, "learning_rate": 2.0497286264694315e-06, "loss": 0.9873, "step": 9432 }, { "epoch": 0.8, "learning_rate": 2.048065075076571e-06, "loss": 0.9789, "step": 9433 }, { "epoch": 0.8, "learning_rate": 2.046402122004426e-06, "loss": 0.935, "step": 9434 }, { "epoch": 0.8, "learning_rate": 2.0447397673781177e-06, "loss": 0.9965, "step": 9435 }, { "epoch": 0.8, "learning_rate": 2.04307801132273e-06, "loss": 0.9613, "step": 9436 }, { "epoch": 0.8, "learning_rate": 2.0414168539632896e-06, "loss": 0.9857, "step": 9437 }, { "epoch": 0.8, "learning_rate": 2.039756295424793e-06, "loss": 1.0218, "step": 9438 }, { "epoch": 0.8, "learning_rate": 2.0380963358321793e-06, "loss": 0.9503, "step": 9439 }, { "epoch": 0.8, "learning_rate": 2.036436975310351e-06, "loss": 0.9525, "step": 9440 }, { "epoch": 0.8, "learning_rate": 2.0347782139841575e-06, "loss": 0.8333, "step": 9441 }, { "epoch": 0.8, "learning_rate": 2.033120051978412e-06, "loss": 0.9877, "step": 9442 }, { "epoch": 0.8, "learning_rate": 2.0314624894178745e-06, "loss": 0.9488, "step": 9443 }, { "epoch": 0.8, "learning_rate": 2.029805526427269e-06, "loss": 0.9618, "step": 9444 }, { "epoch": 0.8, "learning_rate": 2.028149163131261e-06, "loss": 0.9534, "step": 9445 }, { "epoch": 0.8, "learning_rate": 2.0264933996544866e-06, "loss": 0.9372, "step": 9446 }, { "epoch": 0.8, "learning_rate": 2.024838236121526e-06, "loss": 0.8119, "step": 9447 }, { "epoch": 0.8, "learning_rate": 2.0231836726569143e-06, "loss": 0.9495, "step": 9448 }, { "epoch": 0.8, "learning_rate": 2.0215297093851494e-06, "loss": 0.975, "step": 9449 }, { "epoch": 0.8, "learning_rate": 2.0198763464306738e-06, "loss": 0.9671, "step": 9450 }, { "epoch": 0.8, "learning_rate": 2.0182235839178954e-06, "loss": 0.9576, "step": 9451 }, { "epoch": 0.8, "learning_rate": 2.016571421971167e-06, "loss": 1.0353, "step": 9452 }, { "epoch": 0.8, "learning_rate": 2.0149198607148056e-06, "loss": 0.9768, "step": 9453 }, { "epoch": 0.8, "learning_rate": 2.0132689002730734e-06, "loss": 0.966, "step": 9454 }, { "epoch": 0.8, "learning_rate": 2.011618540770197e-06, "loss": 0.9282, "step": 9455 }, { "epoch": 0.8, "learning_rate": 2.009968782330347e-06, "loss": 0.9755, "step": 9456 }, { "epoch": 0.8, "learning_rate": 2.0083196250776617e-06, "loss": 0.9763, "step": 9457 }, { "epoch": 0.8, "learning_rate": 2.0066710691362213e-06, "loss": 1.0017, "step": 9458 }, { "epoch": 0.8, "learning_rate": 2.0050231146300715e-06, "loss": 0.9501, "step": 9459 }, { "epoch": 0.8, "learning_rate": 2.0033757616832026e-06, "loss": 0.9574, "step": 9460 }, { "epoch": 0.8, "learning_rate": 2.001729010419571e-06, "loss": 0.9949, "step": 9461 }, { "epoch": 0.8, "learning_rate": 2.0000828609630762e-06, "loss": 0.9945, "step": 9462 }, { "epoch": 0.8, "learning_rate": 1.9984373134375835e-06, "loss": 0.9612, "step": 9463 }, { "epoch": 0.8, "learning_rate": 1.996792367966901e-06, "loss": 0.969, "step": 9464 }, { "epoch": 0.8, "learning_rate": 1.9951480246748056e-06, "loss": 0.9514, "step": 9465 }, { "epoch": 0.8, "learning_rate": 1.9935042836850138e-06, "loss": 0.9476, "step": 9466 }, { "epoch": 0.8, "learning_rate": 1.99186114512121e-06, "loss": 0.9776, "step": 9467 }, { "epoch": 0.8, "learning_rate": 1.990218609107023e-06, "loss": 1.0031, "step": 9468 }, { "epoch": 0.8, "learning_rate": 1.9885766757660453e-06, "loss": 0.989, "step": 9469 }, { "epoch": 0.8, "learning_rate": 1.9869353452218144e-06, "loss": 0.9582, "step": 9470 }, { "epoch": 0.8, "learning_rate": 1.9852946175978326e-06, "loss": 0.9342, "step": 9471 }, { "epoch": 0.8, "learning_rate": 1.9836544930175473e-06, "loss": 0.9583, "step": 9472 }, { "epoch": 0.8, "learning_rate": 1.982014971604369e-06, "loss": 0.9184, "step": 9473 }, { "epoch": 0.8, "learning_rate": 1.9803760534816572e-06, "loss": 0.9434, "step": 9474 }, { "epoch": 0.8, "learning_rate": 1.9787377387727245e-06, "loss": 0.9579, "step": 9475 }, { "epoch": 0.8, "learning_rate": 1.9771000276008478e-06, "loss": 0.9632, "step": 9476 }, { "epoch": 0.8, "learning_rate": 1.975462920089244e-06, "loss": 0.9925, "step": 9477 }, { "epoch": 0.8, "learning_rate": 1.9738264163610996e-06, "loss": 0.9795, "step": 9478 }, { "epoch": 0.8, "learning_rate": 1.972190516539545e-06, "loss": 0.9231, "step": 9479 }, { "epoch": 0.8, "learning_rate": 1.970555220747671e-06, "loss": 0.9752, "step": 9480 }, { "epoch": 0.8, "learning_rate": 1.968920529108517e-06, "loss": 0.9831, "step": 9481 }, { "epoch": 0.8, "learning_rate": 1.9672864417450866e-06, "loss": 0.9606, "step": 9482 }, { "epoch": 0.8, "learning_rate": 1.965652958780325e-06, "loss": 0.9957, "step": 9483 }, { "epoch": 0.8, "learning_rate": 1.9640200803371455e-06, "loss": 0.9651, "step": 9484 }, { "epoch": 0.8, "learning_rate": 1.962387806538405e-06, "loss": 0.9964, "step": 9485 }, { "epoch": 0.8, "learning_rate": 1.9607561375069216e-06, "loss": 0.9857, "step": 9486 }, { "epoch": 0.8, "learning_rate": 1.9591250733654645e-06, "loss": 1.0183, "step": 9487 }, { "epoch": 0.8, "learning_rate": 1.95749461423676e-06, "loss": 0.9862, "step": 9488 }, { "epoch": 0.8, "learning_rate": 1.9558647602434833e-06, "loss": 0.9703, "step": 9489 }, { "epoch": 0.8, "learning_rate": 1.9542355115082744e-06, "loss": 0.9526, "step": 9490 }, { "epoch": 0.8, "learning_rate": 1.952606868153716e-06, "loss": 0.9563, "step": 9491 }, { "epoch": 0.8, "learning_rate": 1.950978830302355e-06, "loss": 1.0106, "step": 9492 }, { "epoch": 0.8, "learning_rate": 1.949351398076683e-06, "loss": 0.9771, "step": 9493 }, { "epoch": 0.8, "learning_rate": 1.9477245715991575e-06, "loss": 0.9542, "step": 9494 }, { "epoch": 0.8, "learning_rate": 1.9460983509921806e-06, "loss": 0.9804, "step": 9495 }, { "epoch": 0.8, "learning_rate": 1.9444727363781146e-06, "loss": 1.0264, "step": 9496 }, { "epoch": 0.8, "learning_rate": 1.942847727879272e-06, "loss": 0.976, "step": 9497 }, { "epoch": 0.8, "learning_rate": 1.9412233256179257e-06, "loss": 0.9622, "step": 9498 }, { "epoch": 0.8, "learning_rate": 1.9395995297162963e-06, "loss": 0.9711, "step": 9499 }, { "epoch": 0.8, "learning_rate": 1.9379763402965603e-06, "loss": 0.9376, "step": 9500 }, { "epoch": 0.8, "learning_rate": 1.9363537574808543e-06, "loss": 0.9448, "step": 9501 }, { "epoch": 0.8, "learning_rate": 1.9347317813912605e-06, "loss": 0.9889, "step": 9502 }, { "epoch": 0.8, "learning_rate": 1.9331104121498247e-06, "loss": 0.9668, "step": 9503 }, { "epoch": 0.8, "learning_rate": 1.9314896498785363e-06, "loss": 0.9425, "step": 9504 }, { "epoch": 0.8, "learning_rate": 1.929869494699351e-06, "loss": 0.9491, "step": 9505 }, { "epoch": 0.81, "learning_rate": 1.9282499467341676e-06, "loss": 0.9411, "step": 9506 }, { "epoch": 0.81, "learning_rate": 1.9266310061048486e-06, "loss": 0.9906, "step": 9507 }, { "epoch": 0.81, "learning_rate": 1.925012672933202e-06, "loss": 0.9664, "step": 9508 }, { "epoch": 0.81, "learning_rate": 1.9233949473410007e-06, "loss": 0.9899, "step": 9509 }, { "epoch": 0.81, "learning_rate": 1.921777829449959e-06, "loss": 0.9297, "step": 9510 }, { "epoch": 0.81, "learning_rate": 1.9201613193817593e-06, "loss": 0.9646, "step": 9511 }, { "epoch": 0.81, "learning_rate": 1.9185454172580243e-06, "loss": 0.9531, "step": 9512 }, { "epoch": 0.81, "learning_rate": 1.9169301232003444e-06, "loss": 0.9934, "step": 9513 }, { "epoch": 0.81, "learning_rate": 1.9153154373302527e-06, "loss": 0.9688, "step": 9514 }, { "epoch": 0.81, "learning_rate": 1.9137013597692465e-06, "loss": 0.983, "step": 9515 }, { "epoch": 0.81, "learning_rate": 1.9120878906387672e-06, "loss": 0.7532, "step": 9516 }, { "epoch": 0.81, "learning_rate": 1.9104750300602202e-06, "loss": 0.9288, "step": 9517 }, { "epoch": 0.81, "learning_rate": 1.9088627781549574e-06, "loss": 1.0112, "step": 9518 }, { "epoch": 0.81, "learning_rate": 1.90725113504429e-06, "loss": 0.9611, "step": 9519 }, { "epoch": 0.81, "learning_rate": 1.9056401008494807e-06, "loss": 0.9605, "step": 9520 }, { "epoch": 0.81, "learning_rate": 1.9040296756917486e-06, "loss": 0.9716, "step": 9521 }, { "epoch": 0.81, "learning_rate": 1.9024198596922615e-06, "loss": 0.9592, "step": 9522 }, { "epoch": 0.81, "learning_rate": 1.9008106529721515e-06, "loss": 0.8773, "step": 9523 }, { "epoch": 0.81, "learning_rate": 1.8992020556524927e-06, "loss": 0.9343, "step": 9524 }, { "epoch": 0.81, "learning_rate": 1.8975940678543249e-06, "loss": 0.9863, "step": 9525 }, { "epoch": 0.81, "learning_rate": 1.8959866896986334e-06, "loss": 0.9704, "step": 9526 }, { "epoch": 0.81, "learning_rate": 1.8943799213063597e-06, "loss": 1.001, "step": 9527 }, { "epoch": 0.81, "learning_rate": 1.8927737627984044e-06, "loss": 0.9829, "step": 9528 }, { "epoch": 0.81, "learning_rate": 1.8911682142956133e-06, "loss": 0.9387, "step": 9529 }, { "epoch": 0.81, "learning_rate": 1.8895632759187976e-06, "loss": 0.9882, "step": 9530 }, { "epoch": 0.81, "learning_rate": 1.8879589477887095e-06, "loss": 0.9372, "step": 9531 }, { "epoch": 0.81, "learning_rate": 1.886355230026068e-06, "loss": 0.9687, "step": 9532 }, { "epoch": 0.81, "learning_rate": 1.8847521227515354e-06, "loss": 0.9492, "step": 9533 }, { "epoch": 0.81, "learning_rate": 1.8831496260857374e-06, "loss": 0.952, "step": 9534 }, { "epoch": 0.81, "learning_rate": 1.8815477401492443e-06, "loss": 1.0217, "step": 9535 }, { "epoch": 0.81, "learning_rate": 1.879946465062592e-06, "loss": 0.914, "step": 9536 }, { "epoch": 0.81, "learning_rate": 1.8783458009462563e-06, "loss": 0.9827, "step": 9537 }, { "epoch": 0.81, "learning_rate": 1.8767457479206818e-06, "loss": 1.0023, "step": 9538 }, { "epoch": 0.81, "learning_rate": 1.8751463061062547e-06, "loss": 0.9416, "step": 9539 }, { "epoch": 0.81, "learning_rate": 1.873547475623324e-06, "loss": 1.0186, "step": 9540 }, { "epoch": 0.81, "learning_rate": 1.8719492565921849e-06, "loss": 1.0158, "step": 9541 }, { "epoch": 0.81, "learning_rate": 1.8703516491330964e-06, "loss": 0.933, "step": 9542 }, { "epoch": 0.81, "learning_rate": 1.8687546533662614e-06, "loss": 0.9747, "step": 9543 }, { "epoch": 0.81, "learning_rate": 1.867158269411844e-06, "loss": 0.9254, "step": 9544 }, { "epoch": 0.81, "learning_rate": 1.8655624973899566e-06, "loss": 0.9938, "step": 9545 }, { "epoch": 0.81, "learning_rate": 1.8639673374206735e-06, "loss": 1.0017, "step": 9546 }, { "epoch": 0.81, "learning_rate": 1.8623727896240117e-06, "loss": 1.008, "step": 9547 }, { "epoch": 0.81, "learning_rate": 1.8607788541199545e-06, "loss": 0.9913, "step": 9548 }, { "epoch": 0.81, "learning_rate": 1.859185531028427e-06, "loss": 0.7849, "step": 9549 }, { "epoch": 0.81, "learning_rate": 1.8575928204693206e-06, "loss": 0.8812, "step": 9550 }, { "epoch": 0.81, "learning_rate": 1.8560007225624678e-06, "loss": 1.0272, "step": 9551 }, { "epoch": 0.81, "learning_rate": 1.854409237427668e-06, "loss": 0.9382, "step": 9552 }, { "epoch": 0.81, "learning_rate": 1.8528183651846643e-06, "loss": 0.8089, "step": 9553 }, { "epoch": 0.81, "learning_rate": 1.8512281059531545e-06, "loss": 0.9576, "step": 9554 }, { "epoch": 0.81, "learning_rate": 1.8496384598528005e-06, "loss": 0.9593, "step": 9555 }, { "epoch": 0.81, "learning_rate": 1.8480494270032034e-06, "loss": 0.97, "step": 9556 }, { "epoch": 0.81, "learning_rate": 1.8464610075239309e-06, "loss": 1.0025, "step": 9557 }, { "epoch": 0.81, "learning_rate": 1.8448732015344939e-06, "loss": 0.9851, "step": 9558 }, { "epoch": 0.81, "learning_rate": 1.8432860091543692e-06, "loss": 0.9604, "step": 9559 }, { "epoch": 0.81, "learning_rate": 1.8416994305029734e-06, "loss": 0.9537, "step": 9560 }, { "epoch": 0.81, "learning_rate": 1.8401134656996899e-06, "loss": 0.9778, "step": 9561 }, { "epoch": 0.81, "learning_rate": 1.8385281148638446e-06, "loss": 0.9651, "step": 9562 }, { "epoch": 0.81, "learning_rate": 1.8369433781147295e-06, "loss": 0.9921, "step": 9563 }, { "epoch": 0.81, "learning_rate": 1.835359255571576e-06, "loss": 0.9436, "step": 9564 }, { "epoch": 0.81, "learning_rate": 1.8337757473535832e-06, "loss": 0.9376, "step": 9565 }, { "epoch": 0.81, "learning_rate": 1.8321928535798938e-06, "loss": 0.9545, "step": 9566 }, { "epoch": 0.81, "learning_rate": 1.8306105743696112e-06, "loss": 0.9886, "step": 9567 }, { "epoch": 0.81, "learning_rate": 1.8290289098417857e-06, "loss": 0.9908, "step": 9568 }, { "epoch": 0.81, "learning_rate": 1.8274478601154288e-06, "loss": 0.9805, "step": 9569 }, { "epoch": 0.81, "learning_rate": 1.8258674253094987e-06, "loss": 0.9998, "step": 9570 }, { "epoch": 0.81, "learning_rate": 1.824287605542916e-06, "loss": 0.9894, "step": 9571 }, { "epoch": 0.81, "learning_rate": 1.822708400934542e-06, "loss": 0.9872, "step": 9572 }, { "epoch": 0.81, "learning_rate": 1.8211298116032072e-06, "loss": 0.9455, "step": 9573 }, { "epoch": 0.81, "learning_rate": 1.8195518376676824e-06, "loss": 0.9109, "step": 9574 }, { "epoch": 0.81, "learning_rate": 1.817974479246699e-06, "loss": 0.9777, "step": 9575 }, { "epoch": 0.81, "learning_rate": 1.8163977364589459e-06, "loss": 0.9568, "step": 9576 }, { "epoch": 0.81, "learning_rate": 1.8148216094230554e-06, "loss": 0.977, "step": 9577 }, { "epoch": 0.81, "learning_rate": 1.8132460982576172e-06, "loss": 0.9504, "step": 9578 }, { "epoch": 0.81, "learning_rate": 1.8116712030811811e-06, "loss": 0.9376, "step": 9579 }, { "epoch": 0.81, "learning_rate": 1.8100969240122446e-06, "loss": 0.9686, "step": 9580 }, { "epoch": 0.81, "learning_rate": 1.8085232611692549e-06, "loss": 0.9726, "step": 9581 }, { "epoch": 0.81, "learning_rate": 1.8069502146706242e-06, "loss": 0.9175, "step": 9582 }, { "epoch": 0.81, "learning_rate": 1.8053777846347066e-06, "loss": 0.9728, "step": 9583 }, { "epoch": 0.81, "learning_rate": 1.8038059711798206e-06, "loss": 0.8248, "step": 9584 }, { "epoch": 0.81, "learning_rate": 1.8022347744242263e-06, "loss": 0.9766, "step": 9585 }, { "epoch": 0.81, "learning_rate": 1.8006641944861513e-06, "loss": 0.9512, "step": 9586 }, { "epoch": 0.81, "learning_rate": 1.7990942314837623e-06, "loss": 0.9914, "step": 9587 }, { "epoch": 0.81, "learning_rate": 1.7975248855351924e-06, "loss": 0.9147, "step": 9588 }, { "epoch": 0.81, "learning_rate": 1.7959561567585182e-06, "loss": 0.9961, "step": 9589 }, { "epoch": 0.81, "learning_rate": 1.7943880452717789e-06, "loss": 0.9415, "step": 9590 }, { "epoch": 0.81, "learning_rate": 1.7928205511929564e-06, "loss": 0.9802, "step": 9591 }, { "epoch": 0.81, "learning_rate": 1.79125367464e-06, "loss": 0.9089, "step": 9592 }, { "epoch": 0.81, "learning_rate": 1.7896874157307965e-06, "loss": 0.9745, "step": 9593 }, { "epoch": 0.81, "learning_rate": 1.7881217745832024e-06, "loss": 0.9532, "step": 9594 }, { "epoch": 0.81, "learning_rate": 1.786556751315014e-06, "loss": 0.9751, "step": 9595 }, { "epoch": 0.81, "learning_rate": 1.7849923460439911e-06, "loss": 0.9758, "step": 9596 }, { "epoch": 0.81, "learning_rate": 1.7834285588878385e-06, "loss": 0.9986, "step": 9597 }, { "epoch": 0.81, "learning_rate": 1.7818653899642246e-06, "loss": 0.9579, "step": 9598 }, { "epoch": 0.81, "learning_rate": 1.7803028393907595e-06, "loss": 0.9341, "step": 9599 }, { "epoch": 0.81, "learning_rate": 1.7787409072850192e-06, "loss": 0.9536, "step": 9600 }, { "epoch": 0.81, "learning_rate": 1.7771795937645208e-06, "loss": 0.9511, "step": 9601 }, { "epoch": 0.81, "learning_rate": 1.7756188989467437e-06, "loss": 0.9952, "step": 9602 }, { "epoch": 0.81, "learning_rate": 1.7740588229491208e-06, "loss": 0.9649, "step": 9603 }, { "epoch": 0.81, "learning_rate": 1.7724993658890321e-06, "loss": 0.9598, "step": 9604 }, { "epoch": 0.81, "learning_rate": 1.7709405278838132e-06, "loss": 0.9961, "step": 9605 }, { "epoch": 0.81, "learning_rate": 1.7693823090507579e-06, "loss": 0.9642, "step": 9606 }, { "epoch": 0.81, "learning_rate": 1.7678247095071089e-06, "loss": 0.9518, "step": 9607 }, { "epoch": 0.81, "learning_rate": 1.7662677293700603e-06, "loss": 0.9429, "step": 9608 }, { "epoch": 0.81, "learning_rate": 1.7647113687567684e-06, "loss": 0.9217, "step": 9609 }, { "epoch": 0.81, "learning_rate": 1.7631556277843298e-06, "loss": 1.0422, "step": 9610 }, { "epoch": 0.81, "learning_rate": 1.7616005065698094e-06, "loss": 0.9698, "step": 9611 }, { "epoch": 0.81, "learning_rate": 1.7600460052302125e-06, "loss": 0.9101, "step": 9612 }, { "epoch": 0.81, "learning_rate": 1.758492123882507e-06, "loss": 0.9506, "step": 9613 }, { "epoch": 0.81, "learning_rate": 1.7569388626436047e-06, "loss": 1.0364, "step": 9614 }, { "epoch": 0.81, "learning_rate": 1.755386221630383e-06, "loss": 0.9219, "step": 9615 }, { "epoch": 0.81, "learning_rate": 1.753834200959661e-06, "loss": 0.9635, "step": 9616 }, { "epoch": 0.81, "learning_rate": 1.7522828007482196e-06, "loss": 0.9711, "step": 9617 }, { "epoch": 0.81, "learning_rate": 1.7507320211127865e-06, "loss": 0.9447, "step": 9618 }, { "epoch": 0.81, "learning_rate": 1.7491818621700484e-06, "loss": 0.9077, "step": 9619 }, { "epoch": 0.81, "learning_rate": 1.7476323240366388e-06, "loss": 0.9146, "step": 9620 }, { "epoch": 0.81, "learning_rate": 1.7460834068291532e-06, "loss": 0.9626, "step": 9621 }, { "epoch": 0.81, "learning_rate": 1.7445351106641317e-06, "loss": 0.942, "step": 9622 }, { "epoch": 0.81, "learning_rate": 1.7429874356580734e-06, "loss": 0.9285, "step": 9623 }, { "epoch": 0.82, "learning_rate": 1.7414403819274272e-06, "loss": 0.9453, "step": 9624 }, { "epoch": 0.82, "learning_rate": 1.7398939495885992e-06, "loss": 0.9482, "step": 9625 }, { "epoch": 0.82, "learning_rate": 1.7383481387579415e-06, "loss": 0.9826, "step": 9626 }, { "epoch": 0.82, "learning_rate": 1.736802949551768e-06, "loss": 0.9431, "step": 9627 }, { "epoch": 0.82, "learning_rate": 1.7352583820863444e-06, "loss": 0.9656, "step": 9628 }, { "epoch": 0.82, "learning_rate": 1.7337144364778812e-06, "loss": 0.9585, "step": 9629 }, { "epoch": 0.82, "learning_rate": 1.7321711128425544e-06, "loss": 0.9773, "step": 9630 }, { "epoch": 0.82, "learning_rate": 1.7306284112964844e-06, "loss": 1.0058, "step": 9631 }, { "epoch": 0.82, "learning_rate": 1.7290863319557439e-06, "loss": 0.9484, "step": 9632 }, { "epoch": 0.82, "learning_rate": 1.7275448749363667e-06, "loss": 1.0066, "step": 9633 }, { "epoch": 0.82, "learning_rate": 1.7260040403543354e-06, "loss": 0.9837, "step": 9634 }, { "epoch": 0.82, "learning_rate": 1.724463828325581e-06, "loss": 0.9639, "step": 9635 }, { "epoch": 0.82, "learning_rate": 1.7229242389659984e-06, "loss": 0.9672, "step": 9636 }, { "epoch": 0.82, "learning_rate": 1.7213852723914248e-06, "loss": 0.9743, "step": 9637 }, { "epoch": 0.82, "learning_rate": 1.7198469287176588e-06, "loss": 0.9431, "step": 9638 }, { "epoch": 0.82, "learning_rate": 1.7183092080604458e-06, "loss": 1.0282, "step": 9639 }, { "epoch": 0.82, "learning_rate": 1.7167721105354907e-06, "loss": 0.9784, "step": 9640 }, { "epoch": 0.82, "learning_rate": 1.7152356362584444e-06, "loss": 1.0457, "step": 9641 }, { "epoch": 0.82, "learning_rate": 1.7136997853449177e-06, "loss": 0.9804, "step": 9642 }, { "epoch": 0.82, "learning_rate": 1.7121645579104674e-06, "loss": 0.9709, "step": 9643 }, { "epoch": 0.82, "learning_rate": 1.710629954070613e-06, "loss": 0.9586, "step": 9644 }, { "epoch": 0.82, "learning_rate": 1.7090959739408154e-06, "loss": 0.9652, "step": 9645 }, { "epoch": 0.82, "learning_rate": 1.707562617636499e-06, "loss": 1.0048, "step": 9646 }, { "epoch": 0.82, "learning_rate": 1.7060298852730327e-06, "loss": 1.0117, "step": 9647 }, { "epoch": 0.82, "learning_rate": 1.7044977769657477e-06, "loss": 0.9759, "step": 9648 }, { "epoch": 0.82, "learning_rate": 1.7029662928299184e-06, "loss": 0.9441, "step": 9649 }, { "epoch": 0.82, "learning_rate": 1.7014354329807815e-06, "loss": 0.8871, "step": 9650 }, { "epoch": 0.82, "learning_rate": 1.6999051975335178e-06, "loss": 0.9529, "step": 9651 }, { "epoch": 0.82, "learning_rate": 1.6983755866032692e-06, "loss": 0.7801, "step": 9652 }, { "epoch": 0.82, "learning_rate": 1.696846600305122e-06, "loss": 0.9362, "step": 9653 }, { "epoch": 0.82, "learning_rate": 1.6953182387541244e-06, "loss": 0.9995, "step": 9654 }, { "epoch": 0.82, "learning_rate": 1.6937905020652745e-06, "loss": 0.9483, "step": 9655 }, { "epoch": 0.82, "learning_rate": 1.692263390353519e-06, "loss": 0.9829, "step": 9656 }, { "epoch": 0.82, "learning_rate": 1.690736903733765e-06, "loss": 0.9638, "step": 9657 }, { "epoch": 0.82, "learning_rate": 1.6892110423208674e-06, "loss": 0.9541, "step": 9658 }, { "epoch": 0.82, "learning_rate": 1.6876858062296308e-06, "loss": 0.9384, "step": 9659 }, { "epoch": 0.82, "learning_rate": 1.6861611955748237e-06, "loss": 0.9869, "step": 9660 }, { "epoch": 0.82, "learning_rate": 1.6846372104711583e-06, "loss": 0.9485, "step": 9661 }, { "epoch": 0.82, "learning_rate": 1.6831138510333e-06, "loss": 0.9943, "step": 9662 }, { "epoch": 0.82, "learning_rate": 1.6815911173758747e-06, "loss": 0.9851, "step": 9663 }, { "epoch": 0.82, "learning_rate": 1.6800690096134508e-06, "loss": 1.0101, "step": 9664 }, { "epoch": 0.82, "learning_rate": 1.67854752786056e-06, "loss": 0.9858, "step": 9665 }, { "epoch": 0.82, "learning_rate": 1.6770266722316775e-06, "loss": 0.9293, "step": 9666 }, { "epoch": 0.82, "learning_rate": 1.6755064428412405e-06, "loss": 0.9736, "step": 9667 }, { "epoch": 0.82, "learning_rate": 1.6739868398036284e-06, "loss": 0.8034, "step": 9668 }, { "epoch": 0.82, "learning_rate": 1.672467863233186e-06, "loss": 0.9336, "step": 9669 }, { "epoch": 0.82, "learning_rate": 1.6709495132441989e-06, "loss": 0.9798, "step": 9670 }, { "epoch": 0.82, "learning_rate": 1.6694317899509148e-06, "loss": 0.9658, "step": 9671 }, { "epoch": 0.82, "learning_rate": 1.6679146934675272e-06, "loss": 0.8265, "step": 9672 }, { "epoch": 0.82, "learning_rate": 1.6663982239081888e-06, "loss": 0.9794, "step": 9673 }, { "epoch": 0.82, "learning_rate": 1.6648823813869986e-06, "loss": 0.9538, "step": 9674 }, { "epoch": 0.82, "learning_rate": 1.6633671660180161e-06, "loss": 0.9541, "step": 9675 }, { "epoch": 0.82, "learning_rate": 1.661852577915245e-06, "loss": 0.9058, "step": 9676 }, { "epoch": 0.82, "learning_rate": 1.6603386171926505e-06, "loss": 1.0206, "step": 9677 }, { "epoch": 0.82, "learning_rate": 1.6588252839641416e-06, "loss": 0.9252, "step": 9678 }, { "epoch": 0.82, "learning_rate": 1.6573125783435874e-06, "loss": 0.976, "step": 9679 }, { "epoch": 0.82, "learning_rate": 1.6558005004448096e-06, "loss": 0.9213, "step": 9680 }, { "epoch": 0.82, "learning_rate": 1.6542890503815744e-06, "loss": 0.9789, "step": 9681 }, { "epoch": 0.82, "learning_rate": 1.652778228267613e-06, "loss": 0.9912, "step": 9682 }, { "epoch": 0.82, "learning_rate": 1.6512680342165965e-06, "loss": 0.9507, "step": 9683 }, { "epoch": 0.82, "learning_rate": 1.649758468342162e-06, "loss": 0.978, "step": 9684 }, { "epoch": 0.82, "learning_rate": 1.6482495307578872e-06, "loss": 0.9528, "step": 9685 }, { "epoch": 0.82, "learning_rate": 1.6467412215773071e-06, "loss": 0.9668, "step": 9686 }, { "epoch": 0.82, "learning_rate": 1.6452335409139154e-06, "loss": 1.0059, "step": 9687 }, { "epoch": 0.82, "learning_rate": 1.64372648888115e-06, "loss": 0.9643, "step": 9688 }, { "epoch": 0.82, "learning_rate": 1.6422200655924025e-06, "loss": 0.9529, "step": 9689 }, { "epoch": 0.82, "learning_rate": 1.6407142711610246e-06, "loss": 0.9942, "step": 9690 }, { "epoch": 0.82, "learning_rate": 1.6392091057003101e-06, "loss": 0.9602, "step": 9691 }, { "epoch": 0.82, "learning_rate": 1.6377045693235161e-06, "loss": 0.9696, "step": 9692 }, { "epoch": 0.82, "learning_rate": 1.636200662143842e-06, "loss": 0.947, "step": 9693 }, { "epoch": 0.82, "learning_rate": 1.6346973842744508e-06, "loss": 0.9601, "step": 9694 }, { "epoch": 0.82, "learning_rate": 1.633194735828446e-06, "loss": 1.0329, "step": 9695 }, { "epoch": 0.82, "learning_rate": 1.6316927169188967e-06, "loss": 0.995, "step": 9696 }, { "epoch": 0.82, "learning_rate": 1.6301913276588121e-06, "loss": 0.9869, "step": 9697 }, { "epoch": 0.82, "learning_rate": 1.6286905681611642e-06, "loss": 1.0164, "step": 9698 }, { "epoch": 0.82, "learning_rate": 1.6271904385388704e-06, "loss": 0.9121, "step": 9699 }, { "epoch": 0.82, "learning_rate": 1.625690938904807e-06, "loss": 0.9402, "step": 9700 }, { "epoch": 0.82, "learning_rate": 1.624192069371795e-06, "loss": 0.9748, "step": 9701 }, { "epoch": 0.82, "learning_rate": 1.6226938300526175e-06, "loss": 0.956, "step": 9702 }, { "epoch": 0.82, "learning_rate": 1.6211962210600008e-06, "loss": 0.9487, "step": 9703 }, { "epoch": 0.82, "learning_rate": 1.6196992425066328e-06, "loss": 0.9868, "step": 9704 }, { "epoch": 0.82, "learning_rate": 1.6182028945051441e-06, "loss": 0.9459, "step": 9705 }, { "epoch": 0.82, "learning_rate": 1.6167071771681276e-06, "loss": 0.9404, "step": 9706 }, { "epoch": 0.82, "learning_rate": 1.6152120906081237e-06, "loss": 0.9442, "step": 9707 }, { "epoch": 0.82, "learning_rate": 1.6137176349376226e-06, "loss": 0.9735, "step": 9708 }, { "epoch": 0.82, "learning_rate": 1.6122238102690746e-06, "loss": 0.9522, "step": 9709 }, { "epoch": 0.82, "learning_rate": 1.610730616714875e-06, "loss": 1.0297, "step": 9710 }, { "epoch": 0.82, "learning_rate": 1.6092380543873786e-06, "loss": 0.9447, "step": 9711 }, { "epoch": 0.82, "learning_rate": 1.6077461233988856e-06, "loss": 0.9631, "step": 9712 }, { "epoch": 0.82, "learning_rate": 1.6062548238616516e-06, "loss": 0.9969, "step": 9713 }, { "epoch": 0.82, "learning_rate": 1.6047641558878902e-06, "loss": 0.9871, "step": 9714 }, { "epoch": 0.82, "learning_rate": 1.6032741195897573e-06, "loss": 0.9853, "step": 9715 }, { "epoch": 0.82, "learning_rate": 1.6017847150793663e-06, "loss": 0.9319, "step": 9716 }, { "epoch": 0.82, "learning_rate": 1.6002959424687881e-06, "loss": 0.9745, "step": 9717 }, { "epoch": 0.82, "learning_rate": 1.5988078018700348e-06, "loss": 0.9942, "step": 9718 }, { "epoch": 0.82, "learning_rate": 1.5973202933950839e-06, "loss": 0.904, "step": 9719 }, { "epoch": 0.82, "learning_rate": 1.5958334171558532e-06, "loss": 0.9401, "step": 9720 }, { "epoch": 0.82, "learning_rate": 1.5943471732642224e-06, "loss": 1.0065, "step": 9721 }, { "epoch": 0.82, "learning_rate": 1.5928615618320164e-06, "loss": 0.9332, "step": 9722 }, { "epoch": 0.82, "learning_rate": 1.5913765829710192e-06, "loss": 0.9708, "step": 9723 }, { "epoch": 0.82, "learning_rate": 1.5898922367929603e-06, "loss": 0.9693, "step": 9724 }, { "epoch": 0.82, "learning_rate": 1.5884085234095292e-06, "loss": 1.0045, "step": 9725 }, { "epoch": 0.82, "learning_rate": 1.5869254429323577e-06, "loss": 0.9544, "step": 9726 }, { "epoch": 0.82, "learning_rate": 1.5854429954730432e-06, "loss": 1.0088, "step": 9727 }, { "epoch": 0.82, "learning_rate": 1.5839611811431222e-06, "loss": 0.9428, "step": 9728 }, { "epoch": 0.82, "learning_rate": 1.582480000054095e-06, "loss": 0.97, "step": 9729 }, { "epoch": 0.82, "learning_rate": 1.5809994523174033e-06, "loss": 0.9591, "step": 9730 }, { "epoch": 0.82, "learning_rate": 1.5795195380444483e-06, "loss": 0.9585, "step": 9731 }, { "epoch": 0.82, "learning_rate": 1.5780402573465858e-06, "loss": 0.9526, "step": 9732 }, { "epoch": 0.82, "learning_rate": 1.5765616103351156e-06, "loss": 0.8813, "step": 9733 }, { "epoch": 0.82, "learning_rate": 1.5750835971212975e-06, "loss": 0.9317, "step": 9734 }, { "epoch": 0.82, "learning_rate": 1.573606217816337e-06, "loss": 0.9657, "step": 9735 }, { "epoch": 0.82, "learning_rate": 1.5721294725313996e-06, "loss": 0.9648, "step": 9736 }, { "epoch": 0.82, "learning_rate": 1.5706533613775965e-06, "loss": 0.9909, "step": 9737 }, { "epoch": 0.82, "learning_rate": 1.5691778844659899e-06, "loss": 0.937, "step": 9738 }, { "epoch": 0.82, "learning_rate": 1.5677030419076044e-06, "loss": 1.0172, "step": 9739 }, { "epoch": 0.82, "learning_rate": 1.5662288338134047e-06, "loss": 0.9995, "step": 9740 }, { "epoch": 0.82, "learning_rate": 1.5647552602943183e-06, "loss": 0.9189, "step": 9741 }, { "epoch": 0.83, "learning_rate": 1.563282321461217e-06, "loss": 0.9392, "step": 9742 }, { "epoch": 0.83, "learning_rate": 1.5618100174249272e-06, "loss": 0.9284, "step": 9743 }, { "epoch": 0.83, "learning_rate": 1.560338348296232e-06, "loss": 0.9934, "step": 9744 }, { "epoch": 0.83, "learning_rate": 1.5588673141858568e-06, "loss": 1.0025, "step": 9745 }, { "epoch": 0.83, "learning_rate": 1.557396915204492e-06, "loss": 0.9232, "step": 9746 }, { "epoch": 0.83, "learning_rate": 1.5559271514627694e-06, "loss": 0.9146, "step": 9747 }, { "epoch": 0.83, "learning_rate": 1.5544580230712801e-06, "loss": 1.0471, "step": 9748 }, { "epoch": 0.83, "learning_rate": 1.5529895301405606e-06, "loss": 0.951, "step": 9749 }, { "epoch": 0.83, "learning_rate": 1.5515216727811077e-06, "loss": 0.8968, "step": 9750 }, { "epoch": 0.83, "learning_rate": 1.5500544511033622e-06, "loss": 0.9369, "step": 9751 }, { "epoch": 0.83, "learning_rate": 1.5485878652177243e-06, "loss": 0.9838, "step": 9752 }, { "epoch": 0.83, "learning_rate": 1.5471219152345406e-06, "loss": 0.9536, "step": 9753 }, { "epoch": 0.83, "learning_rate": 1.5456566012641162e-06, "loss": 0.9791, "step": 9754 }, { "epoch": 0.83, "learning_rate": 1.544191923416698e-06, "loss": 0.9171, "step": 9755 }, { "epoch": 0.83, "learning_rate": 1.5427278818024992e-06, "loss": 0.9755, "step": 9756 }, { "epoch": 0.83, "learning_rate": 1.5412644765316697e-06, "loss": 0.9561, "step": 9757 }, { "epoch": 0.83, "learning_rate": 1.5398017077143245e-06, "loss": 0.9924, "step": 9758 }, { "epoch": 0.83, "learning_rate": 1.5383395754605257e-06, "loss": 0.9563, "step": 9759 }, { "epoch": 0.83, "learning_rate": 1.5368780798802829e-06, "loss": 0.975, "step": 9760 }, { "epoch": 0.83, "learning_rate": 1.5354172210835683e-06, "loss": 0.9465, "step": 9761 }, { "epoch": 0.83, "learning_rate": 1.5339569991802937e-06, "loss": 1.0022, "step": 9762 }, { "epoch": 0.83, "learning_rate": 1.532497414280335e-06, "loss": 0.9876, "step": 9763 }, { "epoch": 0.83, "learning_rate": 1.531038466493512e-06, "loss": 0.9923, "step": 9764 }, { "epoch": 0.83, "learning_rate": 1.5295801559295965e-06, "loss": 0.962, "step": 9765 }, { "epoch": 0.83, "learning_rate": 1.5281224826983199e-06, "loss": 0.9649, "step": 9766 }, { "epoch": 0.83, "learning_rate": 1.5266654469093578e-06, "loss": 0.9816, "step": 9767 }, { "epoch": 0.83, "learning_rate": 1.5252090486723381e-06, "loss": 0.9562, "step": 9768 }, { "epoch": 0.83, "learning_rate": 1.5237532880968498e-06, "loss": 0.8853, "step": 9769 }, { "epoch": 0.83, "learning_rate": 1.5222981652924208e-06, "loss": 0.9773, "step": 9770 }, { "epoch": 0.83, "learning_rate": 1.5208436803685434e-06, "loss": 0.9942, "step": 9771 }, { "epoch": 0.83, "learning_rate": 1.5193898334346512e-06, "loss": 0.9866, "step": 9772 }, { "epoch": 0.83, "learning_rate": 1.5179366246001404e-06, "loss": 0.9952, "step": 9773 }, { "epoch": 0.83, "learning_rate": 1.5164840539743463e-06, "loss": 0.9744, "step": 9774 }, { "epoch": 0.83, "learning_rate": 1.515032121666572e-06, "loss": 0.9983, "step": 9775 }, { "epoch": 0.83, "learning_rate": 1.5135808277860553e-06, "loss": 0.9682, "step": 9776 }, { "epoch": 0.83, "learning_rate": 1.5121301724420024e-06, "loss": 0.9189, "step": 9777 }, { "epoch": 0.83, "learning_rate": 1.5106801557435568e-06, "loss": 0.9276, "step": 9778 }, { "epoch": 0.83, "learning_rate": 1.5092307777998272e-06, "loss": 0.9422, "step": 9779 }, { "epoch": 0.83, "learning_rate": 1.5077820387198627e-06, "loss": 0.89, "step": 9780 }, { "epoch": 0.83, "learning_rate": 1.5063339386126752e-06, "loss": 0.9682, "step": 9781 }, { "epoch": 0.83, "learning_rate": 1.5048864775872153e-06, "loss": 0.9836, "step": 9782 }, { "epoch": 0.83, "learning_rate": 1.5034396557523989e-06, "loss": 0.9583, "step": 9783 }, { "epoch": 0.83, "learning_rate": 1.5019934732170883e-06, "loss": 0.9803, "step": 9784 }, { "epoch": 0.83, "learning_rate": 1.5005479300900926e-06, "loss": 0.9941, "step": 9785 }, { "epoch": 0.83, "learning_rate": 1.4991030264801831e-06, "loss": 0.9408, "step": 9786 }, { "epoch": 0.83, "learning_rate": 1.4976587624960725e-06, "loss": 0.9517, "step": 9787 }, { "epoch": 0.83, "learning_rate": 1.4962151382464352e-06, "loss": 0.9143, "step": 9788 }, { "epoch": 0.83, "learning_rate": 1.4947721538398873e-06, "loss": 0.9351, "step": 9789 }, { "epoch": 0.83, "learning_rate": 1.493329809385008e-06, "loss": 0.9701, "step": 9790 }, { "epoch": 0.83, "learning_rate": 1.4918881049903188e-06, "loss": 0.9599, "step": 9791 }, { "epoch": 0.83, "learning_rate": 1.4904470407642947e-06, "loss": 0.9817, "step": 9792 }, { "epoch": 0.83, "learning_rate": 1.4890066168153695e-06, "loss": 1.0342, "step": 9793 }, { "epoch": 0.83, "learning_rate": 1.4875668332519212e-06, "loss": 0.9932, "step": 9794 }, { "epoch": 0.83, "learning_rate": 1.4861276901822784e-06, "loss": 0.9724, "step": 9795 }, { "epoch": 0.83, "learning_rate": 1.484689187714733e-06, "loss": 0.9717, "step": 9796 }, { "epoch": 0.83, "learning_rate": 1.4832513259575144e-06, "loss": 0.9135, "step": 9797 }, { "epoch": 0.83, "learning_rate": 1.4818141050188151e-06, "loss": 0.982, "step": 9798 }, { "epoch": 0.83, "learning_rate": 1.480377525006771e-06, "loss": 0.9377, "step": 9799 }, { "epoch": 0.83, "learning_rate": 1.4789415860294776e-06, "loss": 1.0063, "step": 9800 }, { "epoch": 0.83, "learning_rate": 1.4775062881949732e-06, "loss": 0.9296, "step": 9801 }, { "epoch": 0.83, "learning_rate": 1.4760716316112567e-06, "loss": 0.9604, "step": 9802 }, { "epoch": 0.83, "learning_rate": 1.474637616386272e-06, "loss": 0.9414, "step": 9803 }, { "epoch": 0.83, "learning_rate": 1.4732042426279202e-06, "loss": 0.9984, "step": 9804 }, { "epoch": 0.83, "learning_rate": 1.4717715104440477e-06, "loss": 0.9672, "step": 9805 }, { "epoch": 0.83, "learning_rate": 1.4703394199424615e-06, "loss": 0.928, "step": 9806 }, { "epoch": 0.83, "learning_rate": 1.4689079712309095e-06, "loss": 0.9725, "step": 9807 }, { "epoch": 0.83, "learning_rate": 1.4674771644171026e-06, "loss": 1.0165, "step": 9808 }, { "epoch": 0.83, "learning_rate": 1.4660469996086912e-06, "loss": 0.9305, "step": 9809 }, { "epoch": 0.83, "learning_rate": 1.4646174769132893e-06, "loss": 0.9517, "step": 9810 }, { "epoch": 0.83, "learning_rate": 1.4631885964384572e-06, "loss": 0.9482, "step": 9811 }, { "epoch": 0.83, "learning_rate": 1.4617603582917039e-06, "loss": 0.9825, "step": 9812 }, { "epoch": 0.83, "learning_rate": 1.4603327625804963e-06, "loss": 0.9442, "step": 9813 }, { "epoch": 0.83, "learning_rate": 1.4589058094122456e-06, "loss": 0.9894, "step": 9814 }, { "epoch": 0.83, "learning_rate": 1.4574794988943243e-06, "loss": 0.9032, "step": 9815 }, { "epoch": 0.83, "learning_rate": 1.4560538311340445e-06, "loss": 0.9634, "step": 9816 }, { "epoch": 0.83, "learning_rate": 1.4546288062386838e-06, "loss": 1.0341, "step": 9817 }, { "epoch": 0.83, "learning_rate": 1.4532044243154597e-06, "loss": 0.9097, "step": 9818 }, { "epoch": 0.83, "learning_rate": 1.4517806854715455e-06, "loss": 0.9635, "step": 9819 }, { "epoch": 0.83, "learning_rate": 1.4503575898140687e-06, "loss": 0.9586, "step": 9820 }, { "epoch": 0.83, "learning_rate": 1.448935137450105e-06, "loss": 1.0019, "step": 9821 }, { "epoch": 0.83, "learning_rate": 1.4475133284866805e-06, "loss": 0.9577, "step": 9822 }, { "epoch": 0.83, "learning_rate": 1.446092163030779e-06, "loss": 0.95, "step": 9823 }, { "epoch": 0.83, "learning_rate": 1.444671641189328e-06, "loss": 0.9884, "step": 9824 }, { "epoch": 0.83, "learning_rate": 1.4432517630692156e-06, "loss": 0.968, "step": 9825 }, { "epoch": 0.83, "learning_rate": 1.4418325287772717e-06, "loss": 0.9425, "step": 9826 }, { "epoch": 0.83, "learning_rate": 1.4404139384202865e-06, "loss": 0.9412, "step": 9827 }, { "epoch": 0.83, "learning_rate": 1.4389959921049946e-06, "loss": 1.0195, "step": 9828 }, { "epoch": 0.83, "learning_rate": 1.4375786899380872e-06, "loss": 1.009, "step": 9829 }, { "epoch": 0.83, "learning_rate": 1.4361620320262038e-06, "loss": 0.9783, "step": 9830 }, { "epoch": 0.83, "learning_rate": 1.4347460184759398e-06, "loss": 0.9328, "step": 9831 }, { "epoch": 0.83, "learning_rate": 1.4333306493938348e-06, "loss": 0.9604, "step": 9832 }, { "epoch": 0.83, "learning_rate": 1.431915924886388e-06, "loss": 0.9582, "step": 9833 }, { "epoch": 0.83, "learning_rate": 1.4305018450600427e-06, "loss": 0.9979, "step": 9834 }, { "epoch": 0.83, "learning_rate": 1.4290884100211999e-06, "loss": 0.9717, "step": 9835 }, { "epoch": 0.83, "learning_rate": 1.427675619876211e-06, "loss": 0.9531, "step": 9836 }, { "epoch": 0.83, "learning_rate": 1.426263474731372e-06, "loss": 0.9478, "step": 9837 }, { "epoch": 0.83, "learning_rate": 1.4248519746929424e-06, "loss": 0.9504, "step": 9838 }, { "epoch": 0.83, "learning_rate": 1.4234411198671205e-06, "loss": 0.9766, "step": 9839 }, { "epoch": 0.83, "learning_rate": 1.4220309103600671e-06, "loss": 0.8947, "step": 9840 }, { "epoch": 0.83, "learning_rate": 1.420621346277884e-06, "loss": 0.9615, "step": 9841 }, { "epoch": 0.83, "learning_rate": 1.419212427726636e-06, "loss": 0.9514, "step": 9842 }, { "epoch": 0.83, "learning_rate": 1.4178041548123277e-06, "loss": 0.9267, "step": 9843 }, { "epoch": 0.83, "learning_rate": 1.4163965276409241e-06, "loss": 0.9052, "step": 9844 }, { "epoch": 0.83, "learning_rate": 1.4149895463183383e-06, "loss": 0.9419, "step": 9845 }, { "epoch": 0.83, "learning_rate": 1.41358321095043e-06, "loss": 0.9592, "step": 9846 }, { "epoch": 0.83, "learning_rate": 1.4121775216430211e-06, "loss": 0.9649, "step": 9847 }, { "epoch": 0.83, "learning_rate": 1.410772478501875e-06, "loss": 0.9475, "step": 9848 }, { "epoch": 0.83, "learning_rate": 1.4093680816327094e-06, "loss": 0.9389, "step": 9849 }, { "epoch": 0.83, "learning_rate": 1.4079643311411972e-06, "loss": 0.9757, "step": 9850 }, { "epoch": 0.83, "learning_rate": 1.4065612271329554e-06, "loss": 0.916, "step": 9851 }, { "epoch": 0.83, "learning_rate": 1.4051587697135617e-06, "loss": 0.9355, "step": 9852 }, { "epoch": 0.83, "learning_rate": 1.4037569589885359e-06, "loss": 0.9991, "step": 9853 }, { "epoch": 0.83, "learning_rate": 1.402355795063356e-06, "loss": 0.9442, "step": 9854 }, { "epoch": 0.83, "learning_rate": 1.4009552780434466e-06, "loss": 0.9865, "step": 9855 }, { "epoch": 0.83, "learning_rate": 1.3995554080341877e-06, "loss": 0.9946, "step": 9856 }, { "epoch": 0.83, "learning_rate": 1.3981561851409054e-06, "loss": 0.9323, "step": 9857 }, { "epoch": 0.83, "learning_rate": 1.396757609468884e-06, "loss": 0.9647, "step": 9858 }, { "epoch": 0.83, "learning_rate": 1.3953596811233517e-06, "loss": 0.9772, "step": 9859 }, { "epoch": 0.84, "learning_rate": 1.3939624002094953e-06, "loss": 1.0009, "step": 9860 }, { "epoch": 0.84, "learning_rate": 1.3925657668324455e-06, "loss": 0.9393, "step": 9861 }, { "epoch": 0.84, "learning_rate": 1.39116978109729e-06, "loss": 0.9408, "step": 9862 }, { "epoch": 0.84, "learning_rate": 1.3897744431090675e-06, "loss": 1.0063, "step": 9863 }, { "epoch": 0.84, "learning_rate": 1.3883797529727626e-06, "loss": 0.9312, "step": 9864 }, { "epoch": 0.84, "learning_rate": 1.3869857107933182e-06, "loss": 1.0733, "step": 9865 }, { "epoch": 0.84, "learning_rate": 1.3855923166756214e-06, "loss": 0.9333, "step": 9866 }, { "epoch": 0.84, "learning_rate": 1.3841995707245193e-06, "loss": 0.9873, "step": 9867 }, { "epoch": 0.84, "learning_rate": 1.3828074730447983e-06, "loss": 0.909, "step": 9868 }, { "epoch": 0.84, "learning_rate": 1.38141602374121e-06, "loss": 1.0049, "step": 9869 }, { "epoch": 0.84, "learning_rate": 1.380025222918443e-06, "loss": 0.949, "step": 9870 }, { "epoch": 0.84, "learning_rate": 1.3786350706811513e-06, "loss": 0.8506, "step": 9871 }, { "epoch": 0.84, "learning_rate": 1.377245567133928e-06, "loss": 0.9565, "step": 9872 }, { "epoch": 0.84, "learning_rate": 1.3758567123813227e-06, "loss": 0.9842, "step": 9873 }, { "epoch": 0.84, "learning_rate": 1.3744685065278386e-06, "loss": 0.8405, "step": 9874 }, { "epoch": 0.84, "learning_rate": 1.3730809496779252e-06, "loss": 0.9678, "step": 9875 }, { "epoch": 0.84, "learning_rate": 1.3716940419359837e-06, "loss": 1.0132, "step": 9876 }, { "epoch": 0.84, "learning_rate": 1.3703077834063717e-06, "loss": 0.9919, "step": 9877 }, { "epoch": 0.84, "learning_rate": 1.368922174193391e-06, "loss": 1.0029, "step": 9878 }, { "epoch": 0.84, "learning_rate": 1.367537214401301e-06, "loss": 0.8348, "step": 9879 }, { "epoch": 0.84, "learning_rate": 1.3661529041343057e-06, "loss": 0.9838, "step": 9880 }, { "epoch": 0.84, "learning_rate": 1.3647692434965665e-06, "loss": 0.9307, "step": 9881 }, { "epoch": 0.84, "learning_rate": 1.363386232592191e-06, "loss": 0.9684, "step": 9882 }, { "epoch": 0.84, "learning_rate": 1.3620038715252415e-06, "loss": 0.9225, "step": 9883 }, { "epoch": 0.84, "learning_rate": 1.3606221603997272e-06, "loss": 0.9868, "step": 9884 }, { "epoch": 0.84, "learning_rate": 1.3592410993196147e-06, "loss": 0.9386, "step": 9885 }, { "epoch": 0.84, "learning_rate": 1.3578606883888147e-06, "loss": 0.9933, "step": 9886 }, { "epoch": 0.84, "learning_rate": 1.3564809277111934e-06, "loss": 1.0036, "step": 9887 }, { "epoch": 0.84, "learning_rate": 1.3551018173905695e-06, "loss": 0.9678, "step": 9888 }, { "epoch": 0.84, "learning_rate": 1.3537233575307064e-06, "loss": 0.9657, "step": 9889 }, { "epoch": 0.84, "learning_rate": 1.3523455482353255e-06, "loss": 0.944, "step": 9890 }, { "epoch": 0.84, "learning_rate": 1.3509683896080926e-06, "loss": 0.9542, "step": 9891 }, { "epoch": 0.84, "learning_rate": 1.349591881752632e-06, "loss": 0.9391, "step": 9892 }, { "epoch": 0.84, "learning_rate": 1.3482160247725118e-06, "loss": 0.938, "step": 9893 }, { "epoch": 0.84, "learning_rate": 1.3468408187712578e-06, "loss": 0.9691, "step": 9894 }, { "epoch": 0.84, "learning_rate": 1.3454662638523396e-06, "loss": 0.9467, "step": 9895 }, { "epoch": 0.84, "learning_rate": 1.3440923601191847e-06, "loss": 0.9712, "step": 9896 }, { "epoch": 0.84, "learning_rate": 1.3427191076751666e-06, "loss": 0.9308, "step": 9897 }, { "epoch": 0.84, "learning_rate": 1.3413465066236141e-06, "loss": 0.9392, "step": 9898 }, { "epoch": 0.84, "learning_rate": 1.3399745570678035e-06, "loss": 0.9542, "step": 9899 }, { "epoch": 0.84, "learning_rate": 1.338603259110961e-06, "loss": 0.7927, "step": 9900 }, { "epoch": 0.84, "learning_rate": 1.3372326128562707e-06, "loss": 0.9551, "step": 9901 }, { "epoch": 0.84, "learning_rate": 1.335862618406859e-06, "loss": 0.917, "step": 9902 }, { "epoch": 0.84, "learning_rate": 1.334493275865807e-06, "loss": 0.9489, "step": 9903 }, { "epoch": 0.84, "learning_rate": 1.3331245853361507e-06, "loss": 0.9333, "step": 9904 }, { "epoch": 0.84, "learning_rate": 1.3317565469208693e-06, "loss": 1.0022, "step": 9905 }, { "epoch": 0.84, "learning_rate": 1.3303891607229003e-06, "loss": 0.9432, "step": 9906 }, { "epoch": 0.84, "learning_rate": 1.3290224268451245e-06, "loss": 0.9462, "step": 9907 }, { "epoch": 0.84, "learning_rate": 1.3276563453903845e-06, "loss": 0.9447, "step": 9908 }, { "epoch": 0.84, "learning_rate": 1.3262909164614601e-06, "loss": 0.9526, "step": 9909 }, { "epoch": 0.84, "learning_rate": 1.3249261401610946e-06, "loss": 0.9499, "step": 9910 }, { "epoch": 0.84, "learning_rate": 1.3235620165919727e-06, "loss": 1.0035, "step": 9911 }, { "epoch": 0.84, "learning_rate": 1.3221985458567378e-06, "loss": 0.9412, "step": 9912 }, { "epoch": 0.84, "learning_rate": 1.3208357280579764e-06, "loss": 0.9461, "step": 9913 }, { "epoch": 0.84, "learning_rate": 1.3194735632982314e-06, "loss": 0.9519, "step": 9914 }, { "epoch": 0.84, "learning_rate": 1.3181120516799973e-06, "loss": 0.9662, "step": 9915 }, { "epoch": 0.84, "learning_rate": 1.3167511933057132e-06, "loss": 0.9997, "step": 9916 }, { "epoch": 0.84, "learning_rate": 1.3153909882777782e-06, "loss": 0.9317, "step": 9917 }, { "epoch": 0.84, "learning_rate": 1.3140314366985306e-06, "loss": 1.013, "step": 9918 }, { "epoch": 0.84, "learning_rate": 1.312672538670272e-06, "loss": 1.0364, "step": 9919 }, { "epoch": 0.84, "learning_rate": 1.311314294295245e-06, "loss": 0.9774, "step": 9920 }, { "epoch": 0.84, "learning_rate": 1.3099567036756489e-06, "loss": 1.0087, "step": 9921 }, { "epoch": 0.84, "learning_rate": 1.3085997669136297e-06, "loss": 0.9729, "step": 9922 }, { "epoch": 0.84, "learning_rate": 1.307243484111288e-06, "loss": 0.9947, "step": 9923 }, { "epoch": 0.84, "learning_rate": 1.3058878553706722e-06, "loss": 1.0102, "step": 9924 }, { "epoch": 0.84, "learning_rate": 1.304532880793785e-06, "loss": 0.9142, "step": 9925 }, { "epoch": 0.84, "learning_rate": 1.3031785604825752e-06, "loss": 1.003, "step": 9926 }, { "epoch": 0.84, "learning_rate": 1.3018248945389445e-06, "loss": 0.9333, "step": 9927 }, { "epoch": 0.84, "learning_rate": 1.3004718830647479e-06, "loss": 0.9978, "step": 9928 }, { "epoch": 0.84, "learning_rate": 1.2991195261617885e-06, "loss": 0.9473, "step": 9929 }, { "epoch": 0.84, "learning_rate": 1.297767823931817e-06, "loss": 0.9556, "step": 9930 }, { "epoch": 0.84, "learning_rate": 1.2964167764765433e-06, "loss": 0.9683, "step": 9931 }, { "epoch": 0.84, "learning_rate": 1.2950663838976197e-06, "loss": 0.9969, "step": 9932 }, { "epoch": 0.84, "learning_rate": 1.2937166462966545e-06, "loss": 0.886, "step": 9933 }, { "epoch": 0.84, "learning_rate": 1.2923675637752031e-06, "loss": 0.9496, "step": 9934 }, { "epoch": 0.84, "learning_rate": 1.291019136434777e-06, "loss": 0.9453, "step": 9935 }, { "epoch": 0.84, "learning_rate": 1.2896713643768289e-06, "loss": 0.9297, "step": 9936 }, { "epoch": 0.84, "learning_rate": 1.2883242477027745e-06, "loss": 0.9651, "step": 9937 }, { "epoch": 0.84, "learning_rate": 1.2869777865139688e-06, "loss": 0.9827, "step": 9938 }, { "epoch": 0.84, "learning_rate": 1.2856319809117234e-06, "loss": 0.8498, "step": 9939 }, { "epoch": 0.84, "learning_rate": 1.2842868309973034e-06, "loss": 0.9747, "step": 9940 }, { "epoch": 0.84, "learning_rate": 1.2829423368719153e-06, "loss": 0.9808, "step": 9941 }, { "epoch": 0.84, "learning_rate": 1.281598498636727e-06, "loss": 0.9902, "step": 9942 }, { "epoch": 0.84, "learning_rate": 1.2802553163928466e-06, "loss": 0.9355, "step": 9943 }, { "epoch": 0.84, "learning_rate": 1.2789127902413434e-06, "loss": 0.9717, "step": 9944 }, { "epoch": 0.84, "learning_rate": 1.2775709202832265e-06, "loss": 0.9106, "step": 9945 }, { "epoch": 0.84, "learning_rate": 1.2762297066194662e-06, "loss": 0.9975, "step": 9946 }, { "epoch": 0.84, "learning_rate": 1.274889149350974e-06, "loss": 0.9902, "step": 9947 }, { "epoch": 0.84, "learning_rate": 1.2735492485786206e-06, "loss": 0.9757, "step": 9948 }, { "epoch": 0.84, "learning_rate": 1.272210004403217e-06, "loss": 0.9922, "step": 9949 }, { "epoch": 0.84, "learning_rate": 1.2708714169255377e-06, "loss": 0.9271, "step": 9950 }, { "epoch": 0.84, "learning_rate": 1.269533486246295e-06, "loss": 0.9876, "step": 9951 }, { "epoch": 0.84, "learning_rate": 1.268196212466163e-06, "loss": 0.9768, "step": 9952 }, { "epoch": 0.84, "learning_rate": 1.2668595956857576e-06, "loss": 0.9977, "step": 9953 }, { "epoch": 0.84, "learning_rate": 1.2655236360056478e-06, "loss": 0.9602, "step": 9954 }, { "epoch": 0.84, "learning_rate": 1.2641883335263571e-06, "loss": 0.9496, "step": 9955 }, { "epoch": 0.84, "learning_rate": 1.2628536883483566e-06, "loss": 0.9633, "step": 9956 }, { "epoch": 0.84, "learning_rate": 1.2615197005720636e-06, "loss": 0.9087, "step": 9957 }, { "epoch": 0.84, "learning_rate": 1.2601863702978546e-06, "loss": 0.9361, "step": 9958 }, { "epoch": 0.84, "learning_rate": 1.258853697626049e-06, "loss": 0.9723, "step": 9959 }, { "epoch": 0.84, "learning_rate": 1.2575216826569248e-06, "loss": 0.9792, "step": 9960 }, { "epoch": 0.84, "learning_rate": 1.2561903254906992e-06, "loss": 0.9603, "step": 9961 }, { "epoch": 0.84, "learning_rate": 1.2548596262275526e-06, "loss": 0.9842, "step": 9962 }, { "epoch": 0.84, "learning_rate": 1.2535295849676044e-06, "loss": 0.9259, "step": 9963 }, { "epoch": 0.84, "learning_rate": 1.2522002018109348e-06, "loss": 0.9651, "step": 9964 }, { "epoch": 0.84, "learning_rate": 1.2508714768575648e-06, "loss": 0.9646, "step": 9965 }, { "epoch": 0.84, "learning_rate": 1.2495434102074722e-06, "loss": 1.0017, "step": 9966 }, { "epoch": 0.84, "learning_rate": 1.248216001960586e-06, "loss": 0.9575, "step": 9967 }, { "epoch": 0.84, "learning_rate": 1.2468892522167787e-06, "loss": 1.0202, "step": 9968 }, { "epoch": 0.84, "learning_rate": 1.2455631610758834e-06, "loss": 0.9675, "step": 9969 }, { "epoch": 0.84, "learning_rate": 1.2442377286376718e-06, "loss": 1.0364, "step": 9970 }, { "epoch": 0.84, "learning_rate": 1.2429129550018792e-06, "loss": 0.9877, "step": 9971 }, { "epoch": 0.84, "learning_rate": 1.2415888402681775e-06, "loss": 0.8989, "step": 9972 }, { "epoch": 0.84, "learning_rate": 1.240265384536201e-06, "loss": 0.9094, "step": 9973 }, { "epoch": 0.84, "learning_rate": 1.2389425879055262e-06, "loss": 0.9374, "step": 9974 }, { "epoch": 0.84, "learning_rate": 1.2376204504756862e-06, "loss": 0.9544, "step": 9975 }, { "epoch": 0.84, "learning_rate": 1.2362989723461582e-06, "loss": 0.9459, "step": 9976 }, { "epoch": 0.84, "learning_rate": 1.2349781536163762e-06, "loss": 0.9609, "step": 9977 }, { "epoch": 0.85, "learning_rate": 1.2336579943857196e-06, "loss": 0.98, "step": 9978 }, { "epoch": 0.85, "learning_rate": 1.232338494753519e-06, "loss": 0.9994, "step": 9979 }, { "epoch": 0.85, "learning_rate": 1.231019654819059e-06, "loss": 0.9329, "step": 9980 }, { "epoch": 0.85, "learning_rate": 1.2297014746815683e-06, "loss": 1.0358, "step": 9981 }, { "epoch": 0.85, "learning_rate": 1.2283839544402354e-06, "loss": 0.9472, "step": 9982 }, { "epoch": 0.85, "learning_rate": 1.2270670941941886e-06, "loss": 0.953, "step": 9983 }, { "epoch": 0.85, "learning_rate": 1.225750894042511e-06, "loss": 0.9863, "step": 9984 }, { "epoch": 0.85, "learning_rate": 1.2244353540842402e-06, "loss": 0.9857, "step": 9985 }, { "epoch": 0.85, "learning_rate": 1.2231204744183556e-06, "loss": 0.9456, "step": 9986 }, { "epoch": 0.85, "learning_rate": 1.221806255143797e-06, "loss": 0.9515, "step": 9987 }, { "epoch": 0.85, "learning_rate": 1.2204926963594433e-06, "loss": 0.9796, "step": 9988 }, { "epoch": 0.85, "learning_rate": 1.219179798164134e-06, "loss": 1.0003, "step": 9989 }, { "epoch": 0.85, "learning_rate": 1.2178675606566514e-06, "loss": 0.9783, "step": 9990 }, { "epoch": 0.85, "learning_rate": 1.216555983935731e-06, "loss": 0.9476, "step": 9991 }, { "epoch": 0.85, "learning_rate": 1.2152450681000638e-06, "loss": 0.9708, "step": 9992 }, { "epoch": 0.85, "learning_rate": 1.2139348132482786e-06, "loss": 0.996, "step": 9993 }, { "epoch": 0.85, "learning_rate": 1.2126252194789679e-06, "loss": 0.9375, "step": 9994 }, { "epoch": 0.85, "learning_rate": 1.2113162868906646e-06, "loss": 0.9309, "step": 9995 }, { "epoch": 0.85, "learning_rate": 1.2100080155818583e-06, "loss": 0.9681, "step": 9996 }, { "epoch": 0.85, "learning_rate": 1.2087004056509821e-06, "loss": 0.952, "step": 9997 }, { "epoch": 0.85, "learning_rate": 1.2073934571964286e-06, "loss": 0.9875, "step": 9998 }, { "epoch": 0.85, "learning_rate": 1.20608717031653e-06, "loss": 0.9679, "step": 9999 }, { "epoch": 0.85, "learning_rate": 1.204781545109579e-06, "loss": 1.0154, "step": 10000 }, { "epoch": 0.85, "learning_rate": 1.2034765816738103e-06, "loss": 0.9748, "step": 10001 }, { "epoch": 0.85, "learning_rate": 1.2021722801074143e-06, "loss": 0.9448, "step": 10002 }, { "epoch": 0.85, "learning_rate": 1.2008686405085278e-06, "loss": 0.9913, "step": 10003 }, { "epoch": 0.85, "learning_rate": 1.1995656629752416e-06, "loss": 0.9382, "step": 10004 }, { "epoch": 0.85, "learning_rate": 1.1982633476055927e-06, "loss": 0.9685, "step": 10005 }, { "epoch": 0.85, "learning_rate": 1.1969616944975693e-06, "loss": 0.9285, "step": 10006 }, { "epoch": 0.85, "learning_rate": 1.1956607037491141e-06, "loss": 0.9719, "step": 10007 }, { "epoch": 0.85, "learning_rate": 1.1943603754581134e-06, "loss": 0.9784, "step": 10008 }, { "epoch": 0.85, "learning_rate": 1.1930607097224055e-06, "loss": 0.9261, "step": 10009 }, { "epoch": 0.85, "learning_rate": 1.1917617066397847e-06, "loss": 0.9434, "step": 10010 }, { "epoch": 0.85, "learning_rate": 1.1904633663079856e-06, "loss": 0.9686, "step": 10011 }, { "epoch": 0.85, "learning_rate": 1.1891656888247026e-06, "loss": 0.9748, "step": 10012 }, { "epoch": 0.85, "learning_rate": 1.1878686742875723e-06, "loss": 0.9766, "step": 10013 }, { "epoch": 0.85, "learning_rate": 1.1865723227941882e-06, "loss": 0.9525, "step": 10014 }, { "epoch": 0.85, "learning_rate": 1.1852766344420862e-06, "loss": 0.9471, "step": 10015 }, { "epoch": 0.85, "learning_rate": 1.1839816093287615e-06, "loss": 0.9769, "step": 10016 }, { "epoch": 0.85, "learning_rate": 1.1826872475516515e-06, "loss": 0.933, "step": 10017 }, { "epoch": 0.85, "learning_rate": 1.1813935492081463e-06, "loss": 0.9989, "step": 10018 }, { "epoch": 0.85, "learning_rate": 1.1801005143955901e-06, "loss": 0.9498, "step": 10019 }, { "epoch": 0.85, "learning_rate": 1.1788081432112698e-06, "loss": 0.9855, "step": 10020 }, { "epoch": 0.85, "learning_rate": 1.1775164357524293e-06, "loss": 0.9614, "step": 10021 }, { "epoch": 0.85, "learning_rate": 1.176225392116256e-06, "loss": 0.9277, "step": 10022 }, { "epoch": 0.85, "learning_rate": 1.1749350123998937e-06, "loss": 0.9759, "step": 10023 }, { "epoch": 0.85, "learning_rate": 1.173645296700432e-06, "loss": 0.993, "step": 10024 }, { "epoch": 0.85, "learning_rate": 1.172356245114914e-06, "loss": 0.985, "step": 10025 }, { "epoch": 0.85, "learning_rate": 1.1710678577403255e-06, "loss": 0.891, "step": 10026 }, { "epoch": 0.85, "learning_rate": 1.1697801346736147e-06, "loss": 1.0007, "step": 10027 }, { "epoch": 0.85, "learning_rate": 1.1684930760116664e-06, "loss": 0.9954, "step": 10028 }, { "epoch": 0.85, "learning_rate": 1.1672066818513261e-06, "loss": 0.9325, "step": 10029 }, { "epoch": 0.85, "learning_rate": 1.165920952289381e-06, "loss": 0.9748, "step": 10030 }, { "epoch": 0.85, "learning_rate": 1.1646358874225761e-06, "loss": 0.938, "step": 10031 }, { "epoch": 0.85, "learning_rate": 1.1633514873476003e-06, "loss": 0.9553, "step": 10032 }, { "epoch": 0.85, "learning_rate": 1.1620677521610936e-06, "loss": 0.9921, "step": 10033 }, { "epoch": 0.85, "learning_rate": 1.1607846819596502e-06, "loss": 0.9807, "step": 10034 }, { "epoch": 0.85, "learning_rate": 1.1595022768398078e-06, "loss": 0.9898, "step": 10035 }, { "epoch": 0.85, "learning_rate": 1.1582205368980581e-06, "loss": 0.9261, "step": 10036 }, { "epoch": 0.85, "learning_rate": 1.1569394622308439e-06, "loss": 0.9554, "step": 10037 }, { "epoch": 0.85, "learning_rate": 1.1556590529345523e-06, "loss": 0.9022, "step": 10038 }, { "epoch": 0.85, "learning_rate": 1.1543793091055288e-06, "loss": 0.9371, "step": 10039 }, { "epoch": 0.85, "learning_rate": 1.1531002308400595e-06, "loss": 0.9451, "step": 10040 }, { "epoch": 0.85, "learning_rate": 1.15182181823439e-06, "loss": 1.001, "step": 10041 }, { "epoch": 0.85, "learning_rate": 1.150544071384705e-06, "loss": 0.9443, "step": 10042 }, { "epoch": 0.85, "learning_rate": 1.149266990387148e-06, "loss": 0.9784, "step": 10043 }, { "epoch": 0.85, "learning_rate": 1.1479905753378118e-06, "loss": 0.9295, "step": 10044 }, { "epoch": 0.85, "learning_rate": 1.146714826332731e-06, "loss": 0.927, "step": 10045 }, { "epoch": 0.85, "learning_rate": 1.1454397434679022e-06, "loss": 0.989, "step": 10046 }, { "epoch": 0.85, "learning_rate": 1.1441653268392582e-06, "loss": 0.9554, "step": 10047 }, { "epoch": 0.85, "learning_rate": 1.1428915765426962e-06, "loss": 0.7545, "step": 10048 }, { "epoch": 0.85, "learning_rate": 1.1416184926740493e-06, "loss": 0.9016, "step": 10049 }, { "epoch": 0.85, "learning_rate": 1.1403460753291128e-06, "loss": 0.9927, "step": 10050 }, { "epoch": 0.85, "learning_rate": 1.1390743246036207e-06, "loss": 0.9409, "step": 10051 }, { "epoch": 0.85, "learning_rate": 1.137803240593267e-06, "loss": 0.9815, "step": 10052 }, { "epoch": 0.85, "learning_rate": 1.1365328233936867e-06, "loss": 0.9419, "step": 10053 }, { "epoch": 0.85, "learning_rate": 1.1352630731004733e-06, "loss": 0.9888, "step": 10054 }, { "epoch": 0.85, "learning_rate": 1.1339939898091612e-06, "loss": 0.9893, "step": 10055 }, { "epoch": 0.85, "learning_rate": 1.132725573615242e-06, "loss": 0.9201, "step": 10056 }, { "epoch": 0.85, "learning_rate": 1.1314578246141516e-06, "loss": 0.9614, "step": 10057 }, { "epoch": 0.85, "learning_rate": 1.1301907429012815e-06, "loss": 0.9107, "step": 10058 }, { "epoch": 0.85, "learning_rate": 1.1289243285719675e-06, "loss": 1.007, "step": 10059 }, { "epoch": 0.85, "learning_rate": 1.1276585817214946e-06, "loss": 0.9081, "step": 10060 }, { "epoch": 0.85, "learning_rate": 1.1263935024451055e-06, "loss": 1.0133, "step": 10061 }, { "epoch": 0.85, "learning_rate": 1.1251290908379853e-06, "loss": 0.9608, "step": 10062 }, { "epoch": 0.85, "learning_rate": 1.1238653469952688e-06, "loss": 0.926, "step": 10063 }, { "epoch": 0.85, "learning_rate": 1.1226022710120477e-06, "loss": 0.9752, "step": 10064 }, { "epoch": 0.85, "learning_rate": 1.1213398629833527e-06, "loss": 0.9234, "step": 10065 }, { "epoch": 0.85, "learning_rate": 1.1200781230041757e-06, "loss": 0.9527, "step": 10066 }, { "epoch": 0.85, "learning_rate": 1.1188170511694484e-06, "loss": 0.9945, "step": 10067 }, { "epoch": 0.85, "learning_rate": 1.1175566475740574e-06, "loss": 0.8941, "step": 10068 }, { "epoch": 0.85, "learning_rate": 1.116296912312842e-06, "loss": 0.9732, "step": 10069 }, { "epoch": 0.85, "learning_rate": 1.1150378454805822e-06, "loss": 0.9447, "step": 10070 }, { "epoch": 0.85, "learning_rate": 1.1137794471720175e-06, "loss": 0.911, "step": 10071 }, { "epoch": 0.85, "learning_rate": 1.1125217174818282e-06, "loss": 0.9593, "step": 10072 }, { "epoch": 0.85, "learning_rate": 1.1112646565046525e-06, "loss": 0.9324, "step": 10073 }, { "epoch": 0.85, "learning_rate": 1.1100082643350696e-06, "loss": 0.9841, "step": 10074 }, { "epoch": 0.85, "learning_rate": 1.1087525410676193e-06, "loss": 0.9295, "step": 10075 }, { "epoch": 0.85, "learning_rate": 1.107497486796779e-06, "loss": 0.9797, "step": 10076 }, { "epoch": 0.85, "learning_rate": 1.1062431016169871e-06, "loss": 0.9935, "step": 10077 }, { "epoch": 0.85, "learning_rate": 1.1049893856226212e-06, "loss": 0.9276, "step": 10078 }, { "epoch": 0.85, "learning_rate": 1.103736338908017e-06, "loss": 0.9395, "step": 10079 }, { "epoch": 0.85, "learning_rate": 1.1024839615674542e-06, "loss": 0.9176, "step": 10080 }, { "epoch": 0.85, "learning_rate": 1.1012322536951681e-06, "loss": 0.9664, "step": 10081 }, { "epoch": 0.85, "learning_rate": 1.099981215385334e-06, "loss": 0.9453, "step": 10082 }, { "epoch": 0.85, "learning_rate": 1.098730846732089e-06, "loss": 0.9554, "step": 10083 }, { "epoch": 0.85, "learning_rate": 1.0974811478295078e-06, "loss": 0.9522, "step": 10084 }, { "epoch": 0.85, "learning_rate": 1.0962321187716252e-06, "loss": 0.9975, "step": 10085 }, { "epoch": 0.85, "learning_rate": 1.0949837596524192e-06, "loss": 0.8251, "step": 10086 }, { "epoch": 0.85, "learning_rate": 1.0937360705658174e-06, "loss": 0.9263, "step": 10087 }, { "epoch": 0.85, "learning_rate": 1.0924890516057007e-06, "loss": 0.9903, "step": 10088 }, { "epoch": 0.85, "learning_rate": 1.0912427028658967e-06, "loss": 0.9413, "step": 10089 }, { "epoch": 0.85, "learning_rate": 1.0899970244401814e-06, "loss": 1.0046, "step": 10090 }, { "epoch": 0.85, "learning_rate": 1.0887520164222864e-06, "loss": 0.9674, "step": 10091 }, { "epoch": 0.85, "learning_rate": 1.0875076789058857e-06, "loss": 0.9525, "step": 10092 }, { "epoch": 0.85, "learning_rate": 1.086264011984608e-06, "loss": 0.9856, "step": 10093 }, { "epoch": 0.85, "learning_rate": 1.0850210157520258e-06, "loss": 0.9269, "step": 10094 }, { "epoch": 0.85, "learning_rate": 1.083778690301668e-06, "loss": 0.9808, "step": 10095 }, { "epoch": 0.86, "learning_rate": 1.0825370357270105e-06, "loss": 0.8788, "step": 10096 }, { "epoch": 0.86, "learning_rate": 1.0812960521214755e-06, "loss": 0.939, "step": 10097 }, { "epoch": 0.86, "learning_rate": 1.0800557395784406e-06, "loss": 0.9871, "step": 10098 }, { "epoch": 0.86, "learning_rate": 1.0788160981912254e-06, "loss": 0.9778, "step": 10099 }, { "epoch": 0.86, "learning_rate": 1.0775771280531067e-06, "loss": 0.9418, "step": 10100 }, { "epoch": 0.86, "learning_rate": 1.0763388292573052e-06, "loss": 0.9665, "step": 10101 }, { "epoch": 0.86, "learning_rate": 1.0751012018969942e-06, "loss": 0.9675, "step": 10102 }, { "epoch": 0.86, "learning_rate": 1.073864246065295e-06, "loss": 0.9681, "step": 10103 }, { "epoch": 0.86, "learning_rate": 1.0726279618552792e-06, "loss": 0.979, "step": 10104 }, { "epoch": 0.86, "learning_rate": 1.0713923493599665e-06, "loss": 0.9458, "step": 10105 }, { "epoch": 0.86, "learning_rate": 1.0701574086723288e-06, "loss": 0.9596, "step": 10106 }, { "epoch": 0.86, "learning_rate": 1.0689231398852829e-06, "loss": 0.9302, "step": 10107 }, { "epoch": 0.86, "learning_rate": 1.0676895430917022e-06, "loss": 0.9534, "step": 10108 }, { "epoch": 0.86, "learning_rate": 1.0664566183844005e-06, "loss": 1.005, "step": 10109 }, { "epoch": 0.86, "learning_rate": 1.0652243658561502e-06, "loss": 1.0118, "step": 10110 }, { "epoch": 0.86, "learning_rate": 1.0639927855996634e-06, "loss": 0.9734, "step": 10111 }, { "epoch": 0.86, "learning_rate": 1.062761877707612e-06, "loss": 0.9442, "step": 10112 }, { "epoch": 0.86, "learning_rate": 1.0615316422726096e-06, "loss": 0.9435, "step": 10113 }, { "epoch": 0.86, "learning_rate": 1.06030207938722e-06, "loss": 1.0108, "step": 10114 }, { "epoch": 0.86, "learning_rate": 1.0590731891439632e-06, "loss": 0.963, "step": 10115 }, { "epoch": 0.86, "learning_rate": 1.0578449716352989e-06, "loss": 0.9706, "step": 10116 }, { "epoch": 0.86, "learning_rate": 1.0566174269536423e-06, "loss": 0.9377, "step": 10117 }, { "epoch": 0.86, "learning_rate": 1.0553905551913568e-06, "loss": 0.9107, "step": 10118 }, { "epoch": 0.86, "learning_rate": 1.0541643564407545e-06, "loss": 0.9845, "step": 10119 }, { "epoch": 0.86, "learning_rate": 1.0529388307940957e-06, "loss": 0.9766, "step": 10120 }, { "epoch": 0.86, "learning_rate": 1.051713978343597e-06, "loss": 0.9717, "step": 10121 }, { "epoch": 0.86, "learning_rate": 1.0504897991814123e-06, "loss": 0.9784, "step": 10122 }, { "epoch": 0.86, "learning_rate": 1.0492662933996567e-06, "loss": 0.9546, "step": 10123 }, { "epoch": 0.86, "learning_rate": 1.0480434610903856e-06, "loss": 0.9782, "step": 10124 }, { "epoch": 0.86, "learning_rate": 1.0468213023456109e-06, "loss": 0.9856, "step": 10125 }, { "epoch": 0.86, "learning_rate": 1.0455998172572868e-06, "loss": 0.9441, "step": 10126 }, { "epoch": 0.86, "learning_rate": 1.0443790059173243e-06, "loss": 0.9262, "step": 10127 }, { "epoch": 0.86, "learning_rate": 1.0431588684175753e-06, "loss": 0.9707, "step": 10128 }, { "epoch": 0.86, "learning_rate": 1.041939404849851e-06, "loss": 1.0017, "step": 10129 }, { "epoch": 0.86, "learning_rate": 1.0407206153059013e-06, "loss": 0.9659, "step": 10130 }, { "epoch": 0.86, "learning_rate": 1.039502499877435e-06, "loss": 0.9495, "step": 10131 }, { "epoch": 0.86, "learning_rate": 1.0382850586561023e-06, "loss": 0.9809, "step": 10132 }, { "epoch": 0.86, "learning_rate": 1.0370682917335084e-06, "loss": 0.9423, "step": 10133 }, { "epoch": 0.86, "learning_rate": 1.0358521992012028e-06, "loss": 0.9536, "step": 10134 }, { "epoch": 0.86, "learning_rate": 1.0346367811506907e-06, "loss": 0.9725, "step": 10135 }, { "epoch": 0.86, "learning_rate": 1.0334220376734195e-06, "loss": 0.9928, "step": 10136 }, { "epoch": 0.86, "learning_rate": 1.032207968860791e-06, "loss": 0.955, "step": 10137 }, { "epoch": 0.86, "learning_rate": 1.0309945748041527e-06, "loss": 0.8367, "step": 10138 }, { "epoch": 0.86, "learning_rate": 1.0297818555948058e-06, "loss": 0.9033, "step": 10139 }, { "epoch": 0.86, "learning_rate": 1.028569811323996e-06, "loss": 1.0147, "step": 10140 }, { "epoch": 0.86, "learning_rate": 1.0273584420829185e-06, "loss": 0.9891, "step": 10141 }, { "epoch": 0.86, "learning_rate": 1.0261477479627225e-06, "loss": 0.937, "step": 10142 }, { "epoch": 0.86, "learning_rate": 1.0249377290545016e-06, "loss": 0.9738, "step": 10143 }, { "epoch": 0.86, "learning_rate": 1.0237283854492984e-06, "loss": 0.9608, "step": 10144 }, { "epoch": 0.86, "learning_rate": 1.0225197172381107e-06, "loss": 1.0012, "step": 10145 }, { "epoch": 0.86, "learning_rate": 1.0213117245118775e-06, "loss": 0.94, "step": 10146 }, { "epoch": 0.86, "learning_rate": 1.0201044073614918e-06, "loss": 0.9901, "step": 10147 }, { "epoch": 0.86, "learning_rate": 1.0188977658777978e-06, "loss": 0.9538, "step": 10148 }, { "epoch": 0.86, "learning_rate": 1.0176918001515811e-06, "loss": 0.8401, "step": 10149 }, { "epoch": 0.86, "learning_rate": 1.0164865102735854e-06, "loss": 0.9923, "step": 10150 }, { "epoch": 0.86, "learning_rate": 1.0152818963344957e-06, "loss": 0.9941, "step": 10151 }, { "epoch": 0.86, "learning_rate": 1.0140779584249527e-06, "loss": 0.9774, "step": 10152 }, { "epoch": 0.86, "learning_rate": 1.0128746966355407e-06, "loss": 1.0122, "step": 10153 }, { "epoch": 0.86, "learning_rate": 1.0116721110567984e-06, "loss": 0.9533, "step": 10154 }, { "epoch": 0.86, "learning_rate": 1.0104702017792067e-06, "loss": 0.9155, "step": 10155 }, { "epoch": 0.86, "learning_rate": 1.0092689688932057e-06, "loss": 1.026, "step": 10156 }, { "epoch": 0.86, "learning_rate": 1.0080684124891737e-06, "loss": 0.9335, "step": 10157 }, { "epoch": 0.86, "learning_rate": 1.0068685326574467e-06, "loss": 0.9926, "step": 10158 }, { "epoch": 0.86, "learning_rate": 1.0056693294883024e-06, "loss": 0.9328, "step": 10159 }, { "epoch": 0.86, "learning_rate": 1.0044708030719764e-06, "loss": 0.8872, "step": 10160 }, { "epoch": 0.86, "learning_rate": 1.003272953498643e-06, "loss": 0.9237, "step": 10161 }, { "epoch": 0.86, "learning_rate": 1.0020757808584358e-06, "loss": 0.9427, "step": 10162 }, { "epoch": 0.86, "learning_rate": 1.0008792852414285e-06, "loss": 0.9179, "step": 10163 }, { "epoch": 0.86, "learning_rate": 9.99683466737652e-07, "loss": 0.9836, "step": 10164 }, { "epoch": 0.86, "learning_rate": 9.984883254370791e-07, "loss": 0.997, "step": 10165 }, { "epoch": 0.86, "learning_rate": 9.97293861429638e-07, "loss": 1.0125, "step": 10166 }, { "epoch": 0.86, "learning_rate": 9.961000748051996e-07, "loss": 0.9595, "step": 10167 }, { "epoch": 0.86, "learning_rate": 9.94906965653587e-07, "loss": 0.9848, "step": 10168 }, { "epoch": 0.86, "learning_rate": 9.93714534064575e-07, "loss": 0.9359, "step": 10169 }, { "epoch": 0.86, "learning_rate": 9.925227801278826e-07, "loss": 0.9757, "step": 10170 }, { "epoch": 0.86, "learning_rate": 9.913317039331794e-07, "loss": 0.9844, "step": 10171 }, { "epoch": 0.86, "learning_rate": 9.901413055700848e-07, "loss": 0.972, "step": 10172 }, { "epoch": 0.86, "learning_rate": 9.889515851281705e-07, "loss": 0.8804, "step": 10173 }, { "epoch": 0.86, "learning_rate": 9.877625426969484e-07, "loss": 0.9961, "step": 10174 }, { "epoch": 0.86, "learning_rate": 9.8657417836589e-07, "loss": 0.9654, "step": 10175 }, { "epoch": 0.86, "learning_rate": 9.85386492224405e-07, "loss": 0.9325, "step": 10176 }, { "epoch": 0.86, "learning_rate": 9.841994843618619e-07, "loss": 0.9868, "step": 10177 }, { "epoch": 0.86, "learning_rate": 9.830131548675693e-07, "loss": 0.9634, "step": 10178 }, { "epoch": 0.86, "learning_rate": 9.818275038307934e-07, "loss": 0.9785, "step": 10179 }, { "epoch": 0.86, "learning_rate": 9.806425313407431e-07, "loss": 0.9786, "step": 10180 }, { "epoch": 0.86, "learning_rate": 9.79458237486579e-07, "loss": 1.0005, "step": 10181 }, { "epoch": 0.86, "learning_rate": 9.78274622357408e-07, "loss": 0.9743, "step": 10182 }, { "epoch": 0.86, "learning_rate": 9.770916860422907e-07, "loss": 0.9586, "step": 10183 }, { "epoch": 0.86, "learning_rate": 9.759094286302318e-07, "loss": 0.9462, "step": 10184 }, { "epoch": 0.86, "learning_rate": 9.747278502101887e-07, "loss": 0.9153, "step": 10185 }, { "epoch": 0.86, "learning_rate": 9.735469508710626e-07, "loss": 1.0281, "step": 10186 }, { "epoch": 0.86, "learning_rate": 9.723667307017115e-07, "loss": 0.9333, "step": 10187 }, { "epoch": 0.86, "learning_rate": 9.71187189790933e-07, "loss": 0.9008, "step": 10188 }, { "epoch": 0.86, "learning_rate": 9.70008328227483e-07, "loss": 0.9464, "step": 10189 }, { "epoch": 0.86, "learning_rate": 9.68830146100057e-07, "loss": 0.9502, "step": 10190 }, { "epoch": 0.86, "learning_rate": 9.676526434973078e-07, "loss": 0.9641, "step": 10191 }, { "epoch": 0.86, "learning_rate": 9.664758205078294e-07, "loss": 0.9576, "step": 10192 }, { "epoch": 0.86, "learning_rate": 9.652996772201728e-07, "loss": 0.9811, "step": 10193 }, { "epoch": 0.86, "learning_rate": 9.641242137228313e-07, "loss": 0.9815, "step": 10194 }, { "epoch": 0.86, "learning_rate": 9.629494301042474e-07, "loss": 0.9356, "step": 10195 }, { "epoch": 0.86, "learning_rate": 9.617753264528184e-07, "loss": 0.9566, "step": 10196 }, { "epoch": 0.86, "learning_rate": 9.606019028568847e-07, "loss": 0.9683, "step": 10197 }, { "epoch": 0.86, "learning_rate": 9.594291594047344e-07, "loss": 0.9822, "step": 10198 }, { "epoch": 0.86, "learning_rate": 9.582570961846104e-07, "loss": 0.948, "step": 10199 }, { "epoch": 0.86, "learning_rate": 9.570857132847012e-07, "loss": 0.9968, "step": 10200 }, { "epoch": 0.86, "learning_rate": 9.559150107931424e-07, "loss": 0.9317, "step": 10201 }, { "epoch": 0.86, "learning_rate": 9.54744988798023e-07, "loss": 0.9719, "step": 10202 }, { "epoch": 0.86, "learning_rate": 9.535756473873748e-07, "loss": 0.99, "step": 10203 }, { "epoch": 0.86, "learning_rate": 9.524069866491847e-07, "loss": 0.9879, "step": 10204 }, { "epoch": 0.86, "learning_rate": 9.512390066713816e-07, "loss": 0.9698, "step": 10205 }, { "epoch": 0.86, "learning_rate": 9.500717075418509e-07, "loss": 0.9665, "step": 10206 }, { "epoch": 0.86, "learning_rate": 9.489050893484186e-07, "loss": 1.0041, "step": 10207 }, { "epoch": 0.86, "learning_rate": 9.47739152178867e-07, "loss": 0.9309, "step": 10208 }, { "epoch": 0.86, "learning_rate": 9.465738961209214e-07, "loss": 0.9236, "step": 10209 }, { "epoch": 0.86, "learning_rate": 9.454093212622606e-07, "loss": 1.0415, "step": 10210 }, { "epoch": 0.86, "learning_rate": 9.442454276905067e-07, "loss": 0.9751, "step": 10211 }, { "epoch": 0.86, "learning_rate": 9.430822154932362e-07, "loss": 0.9854, "step": 10212 }, { "epoch": 0.86, "learning_rate": 9.419196847579692e-07, "loss": 0.9766, "step": 10213 }, { "epoch": 0.87, "learning_rate": 9.407578355721803e-07, "loss": 1.019, "step": 10214 }, { "epoch": 0.87, "learning_rate": 9.395966680232849e-07, "loss": 0.9518, "step": 10215 }, { "epoch": 0.87, "learning_rate": 9.384361821986565e-07, "loss": 0.9704, "step": 10216 }, { "epoch": 0.87, "learning_rate": 9.372763781856098e-07, "loss": 0.9559, "step": 10217 }, { "epoch": 0.87, "learning_rate": 9.361172560714127e-07, "loss": 0.9547, "step": 10218 }, { "epoch": 0.87, "learning_rate": 9.349588159432776e-07, "loss": 0.941, "step": 10219 }, { "epoch": 0.87, "learning_rate": 9.338010578883705e-07, "loss": 0.9564, "step": 10220 }, { "epoch": 0.87, "learning_rate": 9.326439819938026e-07, "loss": 0.9731, "step": 10221 }, { "epoch": 0.87, "learning_rate": 9.314875883466335e-07, "loss": 0.9711, "step": 10222 }, { "epoch": 0.87, "learning_rate": 9.303318770338754e-07, "loss": 1.0109, "step": 10223 }, { "epoch": 0.87, "learning_rate": 9.291768481424845e-07, "loss": 0.9613, "step": 10224 }, { "epoch": 0.87, "learning_rate": 9.280225017593691e-07, "loss": 0.9705, "step": 10225 }, { "epoch": 0.87, "learning_rate": 9.268688379713819e-07, "loss": 0.934, "step": 10226 }, { "epoch": 0.87, "learning_rate": 9.257158568653313e-07, "loss": 0.9625, "step": 10227 }, { "epoch": 0.87, "learning_rate": 9.245635585279655e-07, "loss": 0.8998, "step": 10228 }, { "epoch": 0.87, "learning_rate": 9.234119430459898e-07, "loss": 0.986, "step": 10229 }, { "epoch": 0.87, "learning_rate": 9.222610105060514e-07, "loss": 0.9508, "step": 10230 }, { "epoch": 0.87, "learning_rate": 9.211107609947512e-07, "loss": 0.9524, "step": 10231 }, { "epoch": 0.87, "learning_rate": 9.199611945986342e-07, "loss": 0.9416, "step": 10232 }, { "epoch": 0.87, "learning_rate": 9.188123114042003e-07, "loss": 0.7254, "step": 10233 }, { "epoch": 0.87, "learning_rate": 9.176641114978879e-07, "loss": 0.9903, "step": 10234 }, { "epoch": 0.87, "learning_rate": 9.165165949660959e-07, "loss": 0.923, "step": 10235 }, { "epoch": 0.87, "learning_rate": 9.153697618951607e-07, "loss": 0.8962, "step": 10236 }, { "epoch": 0.87, "learning_rate": 9.142236123713776e-07, "loss": 0.8984, "step": 10237 }, { "epoch": 0.87, "learning_rate": 9.130781464809802e-07, "loss": 0.9513, "step": 10238 }, { "epoch": 0.87, "learning_rate": 9.119333643101614e-07, "loss": 0.9797, "step": 10239 }, { "epoch": 0.87, "learning_rate": 9.107892659450512e-07, "loss": 0.9727, "step": 10240 }, { "epoch": 0.87, "learning_rate": 9.096458514717399e-07, "loss": 0.9554, "step": 10241 }, { "epoch": 0.87, "learning_rate": 9.085031209762552e-07, "loss": 0.9809, "step": 10242 }, { "epoch": 0.87, "learning_rate": 9.073610745445815e-07, "loss": 0.9917, "step": 10243 }, { "epoch": 0.87, "learning_rate": 9.062197122626481e-07, "loss": 0.9759, "step": 10244 }, { "epoch": 0.87, "learning_rate": 9.050790342163352e-07, "loss": 1.0044, "step": 10245 }, { "epoch": 0.87, "learning_rate": 9.039390404914683e-07, "loss": 0.9427, "step": 10246 }, { "epoch": 0.87, "learning_rate": 9.027997311738213e-07, "loss": 1.0129, "step": 10247 }, { "epoch": 0.87, "learning_rate": 9.01661106349121e-07, "loss": 0.9393, "step": 10248 }, { "epoch": 0.87, "learning_rate": 9.005231661030378e-07, "loss": 0.9487, "step": 10249 }, { "epoch": 0.87, "learning_rate": 8.993859105211966e-07, "loss": 0.9476, "step": 10250 }, { "epoch": 0.87, "learning_rate": 8.982493396891623e-07, "loss": 0.9595, "step": 10251 }, { "epoch": 0.87, "learning_rate": 8.971134536924564e-07, "loss": 0.9543, "step": 10252 }, { "epoch": 0.87, "learning_rate": 8.959782526165417e-07, "loss": 0.8909, "step": 10253 }, { "epoch": 0.87, "learning_rate": 8.948437365468388e-07, "loss": 0.9821, "step": 10254 }, { "epoch": 0.87, "learning_rate": 8.937099055687049e-07, "loss": 0.8158, "step": 10255 }, { "epoch": 0.87, "learning_rate": 8.925767597674573e-07, "loss": 0.93, "step": 10256 }, { "epoch": 0.87, "learning_rate": 8.914442992283512e-07, "loss": 0.9199, "step": 10257 }, { "epoch": 0.87, "learning_rate": 8.903125240366006e-07, "loss": 0.991, "step": 10258 }, { "epoch": 0.87, "learning_rate": 8.891814342773574e-07, "loss": 0.9768, "step": 10259 }, { "epoch": 0.87, "learning_rate": 8.880510300357326e-07, "loss": 0.936, "step": 10260 }, { "epoch": 0.87, "learning_rate": 8.869213113967756e-07, "loss": 0.981, "step": 10261 }, { "epoch": 0.87, "learning_rate": 8.857922784454931e-07, "loss": 0.9691, "step": 10262 }, { "epoch": 0.87, "learning_rate": 8.846639312668315e-07, "loss": 0.9425, "step": 10263 }, { "epoch": 0.87, "learning_rate": 8.83536269945694e-07, "loss": 0.9895, "step": 10264 }, { "epoch": 0.87, "learning_rate": 8.824092945669249e-07, "loss": 0.9892, "step": 10265 }, { "epoch": 0.87, "learning_rate": 8.812830052153232e-07, "loss": 1.0204, "step": 10266 }, { "epoch": 0.87, "learning_rate": 8.80157401975631e-07, "loss": 0.949, "step": 10267 }, { "epoch": 0.87, "learning_rate": 8.790324849325437e-07, "loss": 0.8154, "step": 10268 }, { "epoch": 0.87, "learning_rate": 8.779082541706996e-07, "loss": 0.9963, "step": 10269 }, { "epoch": 0.87, "learning_rate": 8.767847097746906e-07, "loss": 0.9205, "step": 10270 }, { "epoch": 0.87, "learning_rate": 8.756618518290527e-07, "loss": 0.9195, "step": 10271 }, { "epoch": 0.87, "learning_rate": 8.745396804182748e-07, "loss": 0.944, "step": 10272 }, { "epoch": 0.87, "learning_rate": 8.734181956267896e-07, "loss": 0.973, "step": 10273 }, { "epoch": 0.87, "learning_rate": 8.722973975389782e-07, "loss": 0.9553, "step": 10274 }, { "epoch": 0.87, "learning_rate": 8.711772862391754e-07, "loss": 0.9613, "step": 10275 }, { "epoch": 0.87, "learning_rate": 8.700578618116573e-07, "loss": 0.82, "step": 10276 }, { "epoch": 0.87, "learning_rate": 8.689391243406564e-07, "loss": 1.0006, "step": 10277 }, { "epoch": 0.87, "learning_rate": 8.678210739103443e-07, "loss": 0.966, "step": 10278 }, { "epoch": 0.87, "learning_rate": 8.667037106048493e-07, "loss": 0.9663, "step": 10279 }, { "epoch": 0.87, "learning_rate": 8.655870345082396e-07, "loss": 0.8375, "step": 10280 }, { "epoch": 0.87, "learning_rate": 8.644710457045425e-07, "loss": 0.9659, "step": 10281 }, { "epoch": 0.87, "learning_rate": 8.633557442777208e-07, "loss": 0.9934, "step": 10282 }, { "epoch": 0.87, "learning_rate": 8.622411303116984e-07, "loss": 0.9044, "step": 10283 }, { "epoch": 0.87, "learning_rate": 8.611272038903351e-07, "loss": 0.9436, "step": 10284 }, { "epoch": 0.87, "learning_rate": 8.600139650974515e-07, "loss": 0.9464, "step": 10285 }, { "epoch": 0.87, "learning_rate": 8.58901414016804e-07, "loss": 0.9496, "step": 10286 }, { "epoch": 0.87, "learning_rate": 8.577895507321088e-07, "loss": 0.9516, "step": 10287 }, { "epoch": 0.87, "learning_rate": 8.566783753270203e-07, "loss": 0.9968, "step": 10288 }, { "epoch": 0.87, "learning_rate": 8.555678878851492e-07, "loss": 0.9815, "step": 10289 }, { "epoch": 0.87, "learning_rate": 8.544580884900478e-07, "loss": 0.9881, "step": 10290 }, { "epoch": 0.87, "learning_rate": 8.533489772252235e-07, "loss": 0.9257, "step": 10291 }, { "epoch": 0.87, "learning_rate": 8.522405541741242e-07, "loss": 0.9691, "step": 10292 }, { "epoch": 0.87, "learning_rate": 8.51132819420154e-07, "loss": 0.9766, "step": 10293 }, { "epoch": 0.87, "learning_rate": 8.500257730466577e-07, "loss": 0.9488, "step": 10294 }, { "epoch": 0.87, "learning_rate": 8.489194151369351e-07, "loss": 0.9439, "step": 10295 }, { "epoch": 0.87, "learning_rate": 8.478137457742274e-07, "loss": 0.9113, "step": 10296 }, { "epoch": 0.87, "learning_rate": 8.467087650417316e-07, "loss": 0.9251, "step": 10297 }, { "epoch": 0.87, "learning_rate": 8.456044730225843e-07, "loss": 0.9729, "step": 10298 }, { "epoch": 0.87, "learning_rate": 8.445008697998792e-07, "loss": 0.9323, "step": 10299 }, { "epoch": 0.87, "learning_rate": 8.433979554566518e-07, "loss": 0.9412, "step": 10300 }, { "epoch": 0.87, "learning_rate": 8.42295730075886e-07, "loss": 0.9365, "step": 10301 }, { "epoch": 0.87, "learning_rate": 8.411941937405188e-07, "loss": 0.9477, "step": 10302 }, { "epoch": 0.87, "learning_rate": 8.400933465334293e-07, "loss": 0.9451, "step": 10303 }, { "epoch": 0.87, "learning_rate": 8.389931885374514e-07, "loss": 0.9654, "step": 10304 }, { "epoch": 0.87, "learning_rate": 8.378937198353588e-07, "loss": 0.9063, "step": 10305 }, { "epoch": 0.87, "learning_rate": 8.367949405098819e-07, "loss": 0.9877, "step": 10306 }, { "epoch": 0.87, "learning_rate": 8.356968506436914e-07, "loss": 0.9365, "step": 10307 }, { "epoch": 0.87, "learning_rate": 8.345994503194132e-07, "loss": 0.945, "step": 10308 }, { "epoch": 0.87, "learning_rate": 8.335027396196161e-07, "loss": 0.9957, "step": 10309 }, { "epoch": 0.87, "learning_rate": 8.324067186268203e-07, "loss": 0.9972, "step": 10310 }, { "epoch": 0.87, "learning_rate": 8.313113874234913e-07, "loss": 1.0053, "step": 10311 }, { "epoch": 0.87, "learning_rate": 8.302167460920463e-07, "loss": 0.9944, "step": 10312 }, { "epoch": 0.87, "learning_rate": 8.291227947148462e-07, "loss": 0.9908, "step": 10313 }, { "epoch": 0.87, "learning_rate": 8.280295333742038e-07, "loss": 0.9588, "step": 10314 }, { "epoch": 0.87, "learning_rate": 8.269369621523771e-07, "loss": 0.9734, "step": 10315 }, { "epoch": 0.87, "learning_rate": 8.258450811315766e-07, "loss": 0.9794, "step": 10316 }, { "epoch": 0.87, "learning_rate": 8.247538903939522e-07, "loss": 1.0094, "step": 10317 }, { "epoch": 0.87, "learning_rate": 8.23663390021614e-07, "loss": 1.0061, "step": 10318 }, { "epoch": 0.87, "learning_rate": 8.225735800966072e-07, "loss": 0.9946, "step": 10319 }, { "epoch": 0.87, "learning_rate": 8.214844607009376e-07, "loss": 0.9102, "step": 10320 }, { "epoch": 0.87, "learning_rate": 8.203960319165472e-07, "loss": 0.9554, "step": 10321 }, { "epoch": 0.87, "learning_rate": 8.193082938253372e-07, "loss": 0.9259, "step": 10322 }, { "epoch": 0.87, "learning_rate": 8.182212465091466e-07, "loss": 0.9796, "step": 10323 }, { "epoch": 0.87, "learning_rate": 8.1713489004977e-07, "loss": 0.9527, "step": 10324 }, { "epoch": 0.87, "learning_rate": 8.160492245289453e-07, "loss": 0.9671, "step": 10325 }, { "epoch": 0.87, "learning_rate": 8.149642500283627e-07, "loss": 0.9367, "step": 10326 }, { "epoch": 0.87, "learning_rate": 8.138799666296571e-07, "loss": 0.9698, "step": 10327 }, { "epoch": 0.87, "learning_rate": 8.127963744144097e-07, "loss": 0.9505, "step": 10328 }, { "epoch": 0.87, "learning_rate": 8.117134734641563e-07, "loss": 0.931, "step": 10329 }, { "epoch": 0.87, "learning_rate": 8.106312638603742e-07, "loss": 0.9287, "step": 10330 }, { "epoch": 0.87, "learning_rate": 8.095497456844925e-07, "loss": 0.8996, "step": 10331 }, { "epoch": 0.88, "learning_rate": 8.084689190178852e-07, "loss": 0.9265, "step": 10332 }, { "epoch": 0.88, "learning_rate": 8.073887839418781e-07, "loss": 0.9839, "step": 10333 }, { "epoch": 0.88, "learning_rate": 8.063093405377409e-07, "loss": 0.9684, "step": 10334 }, { "epoch": 0.88, "learning_rate": 8.052305888866963e-07, "loss": 0.937, "step": 10335 }, { "epoch": 0.88, "learning_rate": 8.041525290699082e-07, "loss": 0.8228, "step": 10336 }, { "epoch": 0.88, "learning_rate": 8.030751611684951e-07, "loss": 0.9693, "step": 10337 }, { "epoch": 0.88, "learning_rate": 8.019984852635177e-07, "loss": 0.9557, "step": 10338 }, { "epoch": 0.88, "learning_rate": 8.0092250143599e-07, "loss": 0.9292, "step": 10339 }, { "epoch": 0.88, "learning_rate": 7.998472097668697e-07, "loss": 1.016, "step": 10340 }, { "epoch": 0.88, "learning_rate": 7.987726103370663e-07, "loss": 0.9706, "step": 10341 }, { "epoch": 0.88, "learning_rate": 7.976987032274308e-07, "loss": 0.9945, "step": 10342 }, { "epoch": 0.88, "learning_rate": 7.966254885187708e-07, "loss": 0.963, "step": 10343 }, { "epoch": 0.88, "learning_rate": 7.955529662918338e-07, "loss": 0.9487, "step": 10344 }, { "epoch": 0.88, "learning_rate": 7.94481136627322e-07, "loss": 0.9247, "step": 10345 }, { "epoch": 0.88, "learning_rate": 7.934099996058786e-07, "loss": 1.0076, "step": 10346 }, { "epoch": 0.88, "learning_rate": 7.923395553081015e-07, "loss": 0.9962, "step": 10347 }, { "epoch": 0.88, "learning_rate": 7.912698038145306e-07, "loss": 0.9645, "step": 10348 }, { "epoch": 0.88, "learning_rate": 7.902007452056581e-07, "loss": 0.9661, "step": 10349 }, { "epoch": 0.88, "learning_rate": 7.89132379561921e-07, "loss": 0.9783, "step": 10350 }, { "epoch": 0.88, "learning_rate": 7.880647069637071e-07, "loss": 0.9317, "step": 10351 }, { "epoch": 0.88, "learning_rate": 7.869977274913476e-07, "loss": 0.9834, "step": 10352 }, { "epoch": 0.88, "learning_rate": 7.859314412251284e-07, "loss": 0.9726, "step": 10353 }, { "epoch": 0.88, "learning_rate": 7.848658482452765e-07, "loss": 0.9626, "step": 10354 }, { "epoch": 0.88, "learning_rate": 7.838009486319686e-07, "loss": 0.9502, "step": 10355 }, { "epoch": 0.88, "learning_rate": 7.827367424653321e-07, "loss": 0.9695, "step": 10356 }, { "epoch": 0.88, "learning_rate": 7.816732298254381e-07, "loss": 0.9372, "step": 10357 }, { "epoch": 0.88, "learning_rate": 7.806104107923096e-07, "loss": 0.9456, "step": 10358 }, { "epoch": 0.88, "learning_rate": 7.795482854459135e-07, "loss": 0.9, "step": 10359 }, { "epoch": 0.88, "learning_rate": 7.784868538661694e-07, "loss": 0.9469, "step": 10360 }, { "epoch": 0.88, "learning_rate": 7.774261161329377e-07, "loss": 0.9589, "step": 10361 }, { "epoch": 0.88, "learning_rate": 7.763660723260347e-07, "loss": 0.9814, "step": 10362 }, { "epoch": 0.88, "learning_rate": 7.753067225252165e-07, "loss": 0.9084, "step": 10363 }, { "epoch": 0.88, "learning_rate": 7.742480668101949e-07, "loss": 0.9608, "step": 10364 }, { "epoch": 0.88, "learning_rate": 7.731901052606206e-07, "loss": 0.9341, "step": 10365 }, { "epoch": 0.88, "learning_rate": 7.72132837956101e-07, "loss": 0.9435, "step": 10366 }, { "epoch": 0.88, "learning_rate": 7.710762649761849e-07, "loss": 0.9799, "step": 10367 }, { "epoch": 0.88, "learning_rate": 7.70020386400373e-07, "loss": 0.9682, "step": 10368 }, { "epoch": 0.88, "learning_rate": 7.689652023081096e-07, "loss": 0.9949, "step": 10369 }, { "epoch": 0.88, "learning_rate": 7.679107127787921e-07, "loss": 0.9624, "step": 10370 }, { "epoch": 0.88, "learning_rate": 7.668569178917584e-07, "loss": 0.9945, "step": 10371 }, { "epoch": 0.88, "learning_rate": 7.658038177263027e-07, "loss": 0.9976, "step": 10372 }, { "epoch": 0.88, "learning_rate": 7.647514123616584e-07, "loss": 1.0031, "step": 10373 }, { "epoch": 0.88, "learning_rate": 7.636997018770143e-07, "loss": 0.9703, "step": 10374 }, { "epoch": 0.88, "learning_rate": 7.626486863515004e-07, "loss": 0.9502, "step": 10375 }, { "epoch": 0.88, "learning_rate": 7.615983658642001e-07, "loss": 1.0015, "step": 10376 }, { "epoch": 0.88, "learning_rate": 7.605487404941392e-07, "loss": 0.9302, "step": 10377 }, { "epoch": 0.88, "learning_rate": 7.594998103202967e-07, "loss": 0.9793, "step": 10378 }, { "epoch": 0.88, "learning_rate": 7.584515754215927e-07, "loss": 1.0164, "step": 10379 }, { "epoch": 0.88, "learning_rate": 7.57404035876903e-07, "loss": 0.9855, "step": 10380 }, { "epoch": 0.88, "learning_rate": 7.563571917650437e-07, "loss": 1.0094, "step": 10381 }, { "epoch": 0.88, "learning_rate": 7.553110431647814e-07, "loss": 0.929, "step": 10382 }, { "epoch": 0.88, "learning_rate": 7.542655901548324e-07, "loss": 1.0285, "step": 10383 }, { "epoch": 0.88, "learning_rate": 7.532208328138568e-07, "loss": 0.9681, "step": 10384 }, { "epoch": 0.88, "learning_rate": 7.521767712204664e-07, "loss": 0.9492, "step": 10385 }, { "epoch": 0.88, "learning_rate": 7.511334054532171e-07, "loss": 0.9109, "step": 10386 }, { "epoch": 0.88, "learning_rate": 7.500907355906162e-07, "loss": 0.846, "step": 10387 }, { "epoch": 0.88, "learning_rate": 7.49048761711112e-07, "loss": 0.9723, "step": 10388 }, { "epoch": 0.88, "learning_rate": 7.480074838931106e-07, "loss": 1.0273, "step": 10389 }, { "epoch": 0.88, "learning_rate": 7.469669022149551e-07, "loss": 0.97, "step": 10390 }, { "epoch": 0.88, "learning_rate": 7.459270167549448e-07, "loss": 0.944, "step": 10391 }, { "epoch": 0.88, "learning_rate": 7.448878275913196e-07, "loss": 0.9403, "step": 10392 }, { "epoch": 0.88, "learning_rate": 7.438493348022735e-07, "loss": 0.9947, "step": 10393 }, { "epoch": 0.88, "learning_rate": 7.428115384659417e-07, "loss": 0.9437, "step": 10394 }, { "epoch": 0.88, "learning_rate": 7.417744386604131e-07, "loss": 0.9629, "step": 10395 }, { "epoch": 0.88, "learning_rate": 7.407380354637195e-07, "loss": 1.014, "step": 10396 }, { "epoch": 0.88, "learning_rate": 7.397023289538441e-07, "loss": 0.8981, "step": 10397 }, { "epoch": 0.88, "learning_rate": 7.386673192087124e-07, "loss": 0.9733, "step": 10398 }, { "epoch": 0.88, "learning_rate": 7.376330063062042e-07, "loss": 0.9514, "step": 10399 }, { "epoch": 0.88, "learning_rate": 7.365993903241409e-07, "loss": 0.9518, "step": 10400 }, { "epoch": 0.88, "learning_rate": 7.355664713402966e-07, "loss": 0.9595, "step": 10401 }, { "epoch": 0.88, "learning_rate": 7.345342494323871e-07, "loss": 0.9377, "step": 10402 }, { "epoch": 0.88, "learning_rate": 7.335027246780824e-07, "loss": 0.9765, "step": 10403 }, { "epoch": 0.88, "learning_rate": 7.324718971549926e-07, "loss": 0.9391, "step": 10404 }, { "epoch": 0.88, "learning_rate": 7.314417669406837e-07, "loss": 0.9307, "step": 10405 }, { "epoch": 0.88, "learning_rate": 7.304123341126624e-07, "loss": 1.0052, "step": 10406 }, { "epoch": 0.88, "learning_rate": 7.293835987483866e-07, "loss": 0.9438, "step": 10407 }, { "epoch": 0.88, "learning_rate": 7.283555609252602e-07, "loss": 0.9662, "step": 10408 }, { "epoch": 0.88, "learning_rate": 7.273282207206322e-07, "loss": 0.9527, "step": 10409 }, { "epoch": 0.88, "learning_rate": 7.263015782118066e-07, "loss": 0.9985, "step": 10410 }, { "epoch": 0.88, "learning_rate": 7.252756334760269e-07, "loss": 0.9678, "step": 10411 }, { "epoch": 0.88, "learning_rate": 7.242503865904904e-07, "loss": 0.9661, "step": 10412 }, { "epoch": 0.88, "learning_rate": 7.232258376323343e-07, "loss": 0.9376, "step": 10413 }, { "epoch": 0.88, "learning_rate": 7.222019866786523e-07, "loss": 0.9395, "step": 10414 }, { "epoch": 0.88, "learning_rate": 7.211788338064784e-07, "loss": 0.9758, "step": 10415 }, { "epoch": 0.88, "learning_rate": 7.20156379092799e-07, "loss": 0.9309, "step": 10416 }, { "epoch": 0.88, "learning_rate": 7.191346226145435e-07, "loss": 0.9722, "step": 10417 }, { "epoch": 0.88, "learning_rate": 7.181135644485937e-07, "loss": 0.9504, "step": 10418 }, { "epoch": 0.88, "learning_rate": 7.170932046717727e-07, "loss": 0.9582, "step": 10419 }, { "epoch": 0.88, "learning_rate": 7.16073543360859e-07, "loss": 0.9145, "step": 10420 }, { "epoch": 0.88, "learning_rate": 7.15054580592569e-07, "loss": 0.981, "step": 10421 }, { "epoch": 0.88, "learning_rate": 7.14036316443576e-07, "loss": 0.9913, "step": 10422 }, { "epoch": 0.88, "learning_rate": 7.13018750990494e-07, "loss": 0.893, "step": 10423 }, { "epoch": 0.88, "learning_rate": 7.120018843098886e-07, "loss": 0.8545, "step": 10424 }, { "epoch": 0.88, "learning_rate": 7.109857164782685e-07, "loss": 0.9214, "step": 10425 }, { "epoch": 0.88, "learning_rate": 7.099702475720959e-07, "loss": 0.9341, "step": 10426 }, { "epoch": 0.88, "learning_rate": 7.08955477667772e-07, "loss": 0.9789, "step": 10427 }, { "epoch": 0.88, "learning_rate": 7.079414068416557e-07, "loss": 0.8911, "step": 10428 }, { "epoch": 0.88, "learning_rate": 7.069280351700436e-07, "loss": 0.9062, "step": 10429 }, { "epoch": 0.88, "learning_rate": 7.059153627291871e-07, "loss": 0.9522, "step": 10430 }, { "epoch": 0.88, "learning_rate": 7.049033895952784e-07, "loss": 0.9924, "step": 10431 }, { "epoch": 0.88, "learning_rate": 7.038921158444645e-07, "loss": 0.9132, "step": 10432 }, { "epoch": 0.88, "learning_rate": 7.028815415528323e-07, "loss": 0.745, "step": 10433 }, { "epoch": 0.88, "learning_rate": 7.01871666796422e-07, "loss": 0.9398, "step": 10434 }, { "epoch": 0.88, "learning_rate": 7.008624916512185e-07, "loss": 0.9476, "step": 10435 }, { "epoch": 0.88, "learning_rate": 6.998540161931522e-07, "loss": 0.9141, "step": 10436 }, { "epoch": 0.88, "learning_rate": 6.988462404981066e-07, "loss": 0.9625, "step": 10437 }, { "epoch": 0.88, "learning_rate": 6.978391646419036e-07, "loss": 0.9891, "step": 10438 }, { "epoch": 0.88, "learning_rate": 6.968327887003234e-07, "loss": 0.9419, "step": 10439 }, { "epoch": 0.88, "learning_rate": 6.958271127490846e-07, "loss": 0.9798, "step": 10440 }, { "epoch": 0.88, "learning_rate": 6.948221368638575e-07, "loss": 0.9878, "step": 10441 }, { "epoch": 0.88, "learning_rate": 6.938178611202573e-07, "loss": 0.9526, "step": 10442 }, { "epoch": 0.88, "learning_rate": 6.928142855938513e-07, "loss": 0.9502, "step": 10443 }, { "epoch": 0.88, "learning_rate": 6.918114103601459e-07, "loss": 0.9263, "step": 10444 }, { "epoch": 0.88, "learning_rate": 6.908092354946039e-07, "loss": 0.9596, "step": 10445 }, { "epoch": 0.88, "learning_rate": 6.898077610726273e-07, "loss": 0.9341, "step": 10446 }, { "epoch": 0.88, "learning_rate": 6.888069871695724e-07, "loss": 0.8881, "step": 10447 }, { "epoch": 0.88, "learning_rate": 6.878069138607379e-07, "loss": 0.9831, "step": 10448 }, { "epoch": 0.88, "learning_rate": 6.868075412213726e-07, "loss": 0.9979, "step": 10449 }, { "epoch": 0.88, "learning_rate": 6.858088693266695e-07, "loss": 0.9621, "step": 10450 }, { "epoch": 0.89, "learning_rate": 6.848108982517732e-07, "loss": 0.9334, "step": 10451 }, { "epoch": 0.89, "learning_rate": 6.838136280717711e-07, "loss": 0.993, "step": 10452 }, { "epoch": 0.89, "learning_rate": 6.828170588617023e-07, "loss": 0.9578, "step": 10453 }, { "epoch": 0.89, "learning_rate": 6.818211906965478e-07, "loss": 0.9541, "step": 10454 }, { "epoch": 0.89, "learning_rate": 6.808260236512421e-07, "loss": 0.967, "step": 10455 }, { "epoch": 0.89, "learning_rate": 6.798315578006597e-07, "loss": 0.9923, "step": 10456 }, { "epoch": 0.89, "learning_rate": 6.78837793219631e-07, "loss": 0.9827, "step": 10457 }, { "epoch": 0.89, "learning_rate": 6.778447299829261e-07, "loss": 0.9507, "step": 10458 }, { "epoch": 0.89, "learning_rate": 6.768523681652661e-07, "loss": 0.9532, "step": 10459 }, { "epoch": 0.89, "learning_rate": 6.758607078413171e-07, "loss": 0.9447, "step": 10460 }, { "epoch": 0.89, "learning_rate": 6.74869749085697e-07, "loss": 0.8576, "step": 10461 }, { "epoch": 0.89, "learning_rate": 6.738794919729641e-07, "loss": 0.9875, "step": 10462 }, { "epoch": 0.89, "learning_rate": 6.728899365776287e-07, "loss": 0.7815, "step": 10463 }, { "epoch": 0.89, "learning_rate": 6.719010829741479e-07, "loss": 0.9468, "step": 10464 }, { "epoch": 0.89, "learning_rate": 6.709129312369234e-07, "loss": 0.9694, "step": 10465 }, { "epoch": 0.89, "learning_rate": 6.699254814403089e-07, "loss": 0.9137, "step": 10466 }, { "epoch": 0.89, "learning_rate": 6.689387336585984e-07, "loss": 0.9276, "step": 10467 }, { "epoch": 0.89, "learning_rate": 6.679526879660414e-07, "loss": 0.9782, "step": 10468 }, { "epoch": 0.89, "learning_rate": 6.669673444368252e-07, "loss": 0.9824, "step": 10469 }, { "epoch": 0.89, "learning_rate": 6.659827031450927e-07, "loss": 0.9427, "step": 10470 }, { "epoch": 0.89, "learning_rate": 6.649987641649281e-07, "loss": 0.945, "step": 10471 }, { "epoch": 0.89, "learning_rate": 6.640155275703686e-07, "loss": 0.9103, "step": 10472 }, { "epoch": 0.89, "learning_rate": 6.630329934353907e-07, "loss": 0.9538, "step": 10473 }, { "epoch": 0.89, "learning_rate": 6.620511618339254e-07, "loss": 0.9591, "step": 10474 }, { "epoch": 0.89, "learning_rate": 6.610700328398456e-07, "loss": 0.974, "step": 10475 }, { "epoch": 0.89, "learning_rate": 6.600896065269768e-07, "loss": 0.9482, "step": 10476 }, { "epoch": 0.89, "learning_rate": 6.591098829690845e-07, "loss": 0.9717, "step": 10477 }, { "epoch": 0.89, "learning_rate": 6.581308622398896e-07, "loss": 0.9036, "step": 10478 }, { "epoch": 0.89, "learning_rate": 6.5715254441305e-07, "loss": 0.988, "step": 10479 }, { "epoch": 0.89, "learning_rate": 6.561749295621822e-07, "loss": 0.9271, "step": 10480 }, { "epoch": 0.89, "learning_rate": 6.551980177608397e-07, "loss": 0.944, "step": 10481 }, { "epoch": 0.89, "learning_rate": 6.542218090825314e-07, "loss": 0.9863, "step": 10482 }, { "epoch": 0.89, "learning_rate": 6.532463036007053e-07, "loss": 0.9548, "step": 10483 }, { "epoch": 0.89, "learning_rate": 6.522715013887637e-07, "loss": 0.9072, "step": 10484 }, { "epoch": 0.89, "learning_rate": 6.512974025200503e-07, "loss": 0.9411, "step": 10485 }, { "epoch": 0.89, "learning_rate": 6.503240070678618e-07, "loss": 0.8681, "step": 10486 }, { "epoch": 0.89, "learning_rate": 6.493513151054354e-07, "loss": 0.9546, "step": 10487 }, { "epoch": 0.89, "learning_rate": 6.483793267059591e-07, "loss": 0.9515, "step": 10488 }, { "epoch": 0.89, "learning_rate": 6.474080419425688e-07, "loss": 0.9371, "step": 10489 }, { "epoch": 0.89, "learning_rate": 6.464374608883429e-07, "loss": 0.9122, "step": 10490 }, { "epoch": 0.89, "learning_rate": 6.454675836163149e-07, "loss": 0.9717, "step": 10491 }, { "epoch": 0.89, "learning_rate": 6.444984101994556e-07, "loss": 0.9234, "step": 10492 }, { "epoch": 0.89, "learning_rate": 6.435299407106921e-07, "loss": 0.9428, "step": 10493 }, { "epoch": 0.89, "learning_rate": 6.425621752228894e-07, "loss": 0.9392, "step": 10494 }, { "epoch": 0.89, "learning_rate": 6.415951138088694e-07, "loss": 0.9716, "step": 10495 }, { "epoch": 0.89, "learning_rate": 6.406287565413904e-07, "loss": 0.9997, "step": 10496 }, { "epoch": 0.89, "learning_rate": 6.396631034931688e-07, "loss": 1.0214, "step": 10497 }, { "epoch": 0.89, "learning_rate": 6.386981547368576e-07, "loss": 0.9156, "step": 10498 }, { "epoch": 0.89, "learning_rate": 6.377339103450653e-07, "loss": 1.0624, "step": 10499 }, { "epoch": 0.89, "learning_rate": 6.367703703903417e-07, "loss": 1.0027, "step": 10500 }, { "epoch": 0.89, "learning_rate": 6.358075349451864e-07, "loss": 0.9167, "step": 10501 }, { "epoch": 0.89, "learning_rate": 6.348454040820429e-07, "loss": 1.0215, "step": 10502 }, { "epoch": 0.89, "learning_rate": 6.338839778733085e-07, "loss": 0.9531, "step": 10503 }, { "epoch": 0.89, "learning_rate": 6.32923256391319e-07, "loss": 0.9889, "step": 10504 }, { "epoch": 0.89, "learning_rate": 6.319632397083631e-07, "loss": 0.975, "step": 10505 }, { "epoch": 0.89, "learning_rate": 6.310039278966728e-07, "loss": 0.9781, "step": 10506 }, { "epoch": 0.89, "learning_rate": 6.30045321028433e-07, "loss": 1.0, "step": 10507 }, { "epoch": 0.89, "learning_rate": 6.290874191757656e-07, "loss": 0.9281, "step": 10508 }, { "epoch": 0.89, "learning_rate": 6.281302224107488e-07, "loss": 0.8847, "step": 10509 }, { "epoch": 0.89, "learning_rate": 6.271737308054027e-07, "loss": 0.9915, "step": 10510 }, { "epoch": 0.89, "learning_rate": 6.262179444316985e-07, "loss": 0.9273, "step": 10511 }, { "epoch": 0.89, "learning_rate": 6.252628633615465e-07, "loss": 0.9653, "step": 10512 }, { "epoch": 0.89, "learning_rate": 6.243084876668148e-07, "loss": 0.9897, "step": 10513 }, { "epoch": 0.89, "learning_rate": 6.233548174193094e-07, "loss": 0.9543, "step": 10514 }, { "epoch": 0.89, "learning_rate": 6.22401852690786e-07, "loss": 0.9652, "step": 10515 }, { "epoch": 0.89, "learning_rate": 6.214495935529497e-07, "loss": 0.911, "step": 10516 }, { "epoch": 0.89, "learning_rate": 6.204980400774485e-07, "loss": 0.953, "step": 10517 }, { "epoch": 0.89, "learning_rate": 6.19547192335882e-07, "loss": 0.9367, "step": 10518 }, { "epoch": 0.89, "learning_rate": 6.185970503997906e-07, "loss": 0.9852, "step": 10519 }, { "epoch": 0.89, "learning_rate": 6.176476143406684e-07, "loss": 0.9656, "step": 10520 }, { "epoch": 0.89, "learning_rate": 6.166988842299504e-07, "loss": 0.9542, "step": 10521 }, { "epoch": 0.89, "learning_rate": 6.157508601390239e-07, "loss": 0.8778, "step": 10522 }, { "epoch": 0.89, "learning_rate": 6.148035421392162e-07, "loss": 0.9932, "step": 10523 }, { "epoch": 0.89, "learning_rate": 6.138569303018104e-07, "loss": 0.9639, "step": 10524 }, { "epoch": 0.89, "learning_rate": 6.129110246980274e-07, "loss": 0.9441, "step": 10525 }, { "epoch": 0.89, "learning_rate": 6.119658253990423e-07, "loss": 0.9516, "step": 10526 }, { "epoch": 0.89, "learning_rate": 6.110213324759706e-07, "loss": 0.9419, "step": 10527 }, { "epoch": 0.89, "learning_rate": 6.10077545999882e-07, "loss": 0.9311, "step": 10528 }, { "epoch": 0.89, "learning_rate": 6.091344660417853e-07, "loss": 0.9158, "step": 10529 }, { "epoch": 0.89, "learning_rate": 6.081920926726436e-07, "loss": 0.9242, "step": 10530 }, { "epoch": 0.89, "learning_rate": 6.072504259633594e-07, "loss": 0.9584, "step": 10531 }, { "epoch": 0.89, "learning_rate": 6.063094659847879e-07, "loss": 0.9929, "step": 10532 }, { "epoch": 0.89, "learning_rate": 6.053692128077271e-07, "loss": 1.0052, "step": 10533 }, { "epoch": 0.89, "learning_rate": 6.04429666502927e-07, "loss": 0.9595, "step": 10534 }, { "epoch": 0.89, "learning_rate": 6.034908271410778e-07, "loss": 0.9895, "step": 10535 }, { "epoch": 0.89, "learning_rate": 6.025526947928206e-07, "loss": 0.8106, "step": 10536 }, { "epoch": 0.89, "learning_rate": 6.016152695287448e-07, "loss": 0.9931, "step": 10537 }, { "epoch": 0.89, "learning_rate": 6.006785514193825e-07, "loss": 0.8509, "step": 10538 }, { "epoch": 0.89, "learning_rate": 5.997425405352131e-07, "loss": 0.9433, "step": 10539 }, { "epoch": 0.89, "learning_rate": 5.988072369466669e-07, "loss": 0.9381, "step": 10540 }, { "epoch": 0.89, "learning_rate": 5.978726407241153e-07, "loss": 0.9902, "step": 10541 }, { "epoch": 0.89, "learning_rate": 5.9693875193788e-07, "loss": 0.919, "step": 10542 }, { "epoch": 0.89, "learning_rate": 5.960055706582313e-07, "loss": 0.9222, "step": 10543 }, { "epoch": 0.89, "learning_rate": 5.950730969553797e-07, "loss": 0.9492, "step": 10544 }, { "epoch": 0.89, "learning_rate": 5.941413308994915e-07, "loss": 0.967, "step": 10545 }, { "epoch": 0.89, "learning_rate": 5.932102725606692e-07, "loss": 0.9508, "step": 10546 }, { "epoch": 0.89, "learning_rate": 5.922799220089726e-07, "loss": 0.96, "step": 10547 }, { "epoch": 0.89, "learning_rate": 5.913502793143988e-07, "loss": 0.9685, "step": 10548 }, { "epoch": 0.89, "learning_rate": 5.904213445469009e-07, "loss": 0.9672, "step": 10549 }, { "epoch": 0.89, "learning_rate": 5.894931177763696e-07, "loss": 0.9894, "step": 10550 }, { "epoch": 0.89, "learning_rate": 5.885655990726502e-07, "loss": 0.9629, "step": 10551 }, { "epoch": 0.89, "learning_rate": 5.876387885055279e-07, "loss": 0.985, "step": 10552 }, { "epoch": 0.89, "learning_rate": 5.867126861447414e-07, "loss": 0.9904, "step": 10553 }, { "epoch": 0.89, "learning_rate": 5.857872920599694e-07, "loss": 0.9053, "step": 10554 }, { "epoch": 0.89, "learning_rate": 5.84862606320844e-07, "loss": 0.8977, "step": 10555 }, { "epoch": 0.89, "learning_rate": 5.839386289969363e-07, "loss": 0.9849, "step": 10556 }, { "epoch": 0.89, "learning_rate": 5.830153601577726e-07, "loss": 0.9445, "step": 10557 }, { "epoch": 0.89, "learning_rate": 5.820927998728176e-07, "loss": 0.976, "step": 10558 }, { "epoch": 0.89, "learning_rate": 5.811709482114903e-07, "loss": 0.9142, "step": 10559 }, { "epoch": 0.89, "learning_rate": 5.802498052431493e-07, "loss": 0.924, "step": 10560 }, { "epoch": 0.89, "learning_rate": 5.793293710371073e-07, "loss": 0.9289, "step": 10561 }, { "epoch": 0.89, "learning_rate": 5.784096456626165e-07, "loss": 0.9792, "step": 10562 }, { "epoch": 0.89, "learning_rate": 5.774906291888805e-07, "loss": 0.9721, "step": 10563 }, { "epoch": 0.89, "learning_rate": 5.765723216850494e-07, "loss": 0.9361, "step": 10564 }, { "epoch": 0.89, "learning_rate": 5.75654723220217e-07, "loss": 0.9748, "step": 10565 }, { "epoch": 0.89, "learning_rate": 5.747378338634236e-07, "loss": 0.9835, "step": 10566 }, { "epoch": 0.89, "learning_rate": 5.738216536836616e-07, "loss": 0.9806, "step": 10567 }, { "epoch": 0.89, "learning_rate": 5.72906182749865e-07, "loss": 0.7553, "step": 10568 }, { "epoch": 0.9, "learning_rate": 5.71991421130913e-07, "loss": 0.9768, "step": 10569 }, { "epoch": 0.9, "learning_rate": 5.710773688956383e-07, "loss": 0.9814, "step": 10570 }, { "epoch": 0.9, "learning_rate": 5.701640261128139e-07, "loss": 0.9142, "step": 10571 }, { "epoch": 0.9, "learning_rate": 5.692513928511634e-07, "loss": 1.0105, "step": 10572 }, { "epoch": 0.9, "learning_rate": 5.68339469179352e-07, "loss": 0.913, "step": 10573 }, { "epoch": 0.9, "learning_rate": 5.674282551659993e-07, "loss": 0.9626, "step": 10574 }, { "epoch": 0.9, "learning_rate": 5.665177508796627e-07, "loss": 0.989, "step": 10575 }, { "epoch": 0.9, "learning_rate": 5.65607956388855e-07, "loss": 0.9747, "step": 10576 }, { "epoch": 0.9, "learning_rate": 5.646988717620261e-07, "loss": 1.0169, "step": 10577 }, { "epoch": 0.9, "learning_rate": 5.637904970675811e-07, "loss": 0.9767, "step": 10578 }, { "epoch": 0.9, "learning_rate": 5.628828323738666e-07, "loss": 0.9471, "step": 10579 }, { "epoch": 0.9, "learning_rate": 5.619758777491779e-07, "loss": 0.9476, "step": 10580 }, { "epoch": 0.9, "learning_rate": 5.610696332617549e-07, "loss": 0.8842, "step": 10581 }, { "epoch": 0.9, "learning_rate": 5.601640989797885e-07, "loss": 0.9923, "step": 10582 }, { "epoch": 0.9, "learning_rate": 5.592592749714088e-07, "loss": 0.9625, "step": 10583 }, { "epoch": 0.9, "learning_rate": 5.583551613047012e-07, "loss": 0.9315, "step": 10584 }, { "epoch": 0.9, "learning_rate": 5.574517580476879e-07, "loss": 0.9427, "step": 10585 }, { "epoch": 0.9, "learning_rate": 5.565490652683481e-07, "loss": 0.901, "step": 10586 }, { "epoch": 0.9, "learning_rate": 5.556470830345984e-07, "loss": 0.9586, "step": 10587 }, { "epoch": 0.9, "learning_rate": 5.547458114143079e-07, "loss": 0.912, "step": 10588 }, { "epoch": 0.9, "learning_rate": 5.538452504752912e-07, "loss": 0.9371, "step": 10589 }, { "epoch": 0.9, "learning_rate": 5.529454002853063e-07, "loss": 0.9473, "step": 10590 }, { "epoch": 0.9, "learning_rate": 5.520462609120614e-07, "loss": 0.9305, "step": 10591 }, { "epoch": 0.9, "learning_rate": 5.511478324232089e-07, "loss": 0.9231, "step": 10592 }, { "epoch": 0.9, "learning_rate": 5.502501148863471e-07, "loss": 0.9068, "step": 10593 }, { "epoch": 0.9, "learning_rate": 5.493531083690251e-07, "loss": 1.016, "step": 10594 }, { "epoch": 0.9, "learning_rate": 5.484568129387347e-07, "loss": 0.9388, "step": 10595 }, { "epoch": 0.9, "learning_rate": 5.475612286629129e-07, "loss": 0.912, "step": 10596 }, { "epoch": 0.9, "learning_rate": 5.46666355608948e-07, "loss": 0.9632, "step": 10597 }, { "epoch": 0.9, "learning_rate": 5.457721938441707e-07, "loss": 0.978, "step": 10598 }, { "epoch": 0.9, "learning_rate": 5.448787434358604e-07, "loss": 1.0066, "step": 10599 }, { "epoch": 0.9, "learning_rate": 5.439860044512413e-07, "loss": 0.9929, "step": 10600 }, { "epoch": 0.9, "learning_rate": 5.430939769574872e-07, "loss": 0.9631, "step": 10601 }, { "epoch": 0.9, "learning_rate": 5.422026610217135e-07, "loss": 0.9472, "step": 10602 }, { "epoch": 0.9, "learning_rate": 5.413120567109875e-07, "loss": 0.9475, "step": 10603 }, { "epoch": 0.9, "learning_rate": 5.404221640923168e-07, "loss": 0.9259, "step": 10604 }, { "epoch": 0.9, "learning_rate": 5.395329832326612e-07, "loss": 0.9784, "step": 10605 }, { "epoch": 0.9, "learning_rate": 5.386445141989238e-07, "loss": 0.9844, "step": 10606 }, { "epoch": 0.9, "learning_rate": 5.377567570579568e-07, "loss": 0.9575, "step": 10607 }, { "epoch": 0.9, "learning_rate": 5.368697118765531e-07, "loss": 0.9325, "step": 10608 }, { "epoch": 0.9, "learning_rate": 5.359833787214597e-07, "loss": 0.996, "step": 10609 }, { "epoch": 0.9, "learning_rate": 5.35097757659363e-07, "loss": 1.018, "step": 10610 }, { "epoch": 0.9, "learning_rate": 5.342128487569032e-07, "loss": 0.928, "step": 10611 }, { "epoch": 0.9, "learning_rate": 5.33328652080658e-07, "loss": 0.9337, "step": 10612 }, { "epoch": 0.9, "learning_rate": 5.324451676971598e-07, "loss": 0.9437, "step": 10613 }, { "epoch": 0.9, "learning_rate": 5.31562395672881e-07, "loss": 0.98, "step": 10614 }, { "epoch": 0.9, "learning_rate": 5.306803360742452e-07, "loss": 0.9613, "step": 10615 }, { "epoch": 0.9, "learning_rate": 5.297989889676213e-07, "loss": 0.9727, "step": 10616 }, { "epoch": 0.9, "learning_rate": 5.28918354419321e-07, "loss": 0.942, "step": 10617 }, { "epoch": 0.9, "learning_rate": 5.28038432495609e-07, "loss": 0.9165, "step": 10618 }, { "epoch": 0.9, "learning_rate": 5.271592232626887e-07, "loss": 0.9991, "step": 10619 }, { "epoch": 0.9, "learning_rate": 5.262807267867154e-07, "loss": 0.9509, "step": 10620 }, { "epoch": 0.9, "learning_rate": 5.254029431337892e-07, "loss": 0.9467, "step": 10621 }, { "epoch": 0.9, "learning_rate": 5.245258723699564e-07, "loss": 0.9713, "step": 10622 }, { "epoch": 0.9, "learning_rate": 5.236495145612075e-07, "loss": 0.9978, "step": 10623 }, { "epoch": 0.9, "learning_rate": 5.227738697734863e-07, "loss": 0.9924, "step": 10624 }, { "epoch": 0.9, "learning_rate": 5.218989380726725e-07, "loss": 0.9443, "step": 10625 }, { "epoch": 0.9, "learning_rate": 5.210247195246021e-07, "loss": 0.9498, "step": 10626 }, { "epoch": 0.9, "learning_rate": 5.201512141950515e-07, "loss": 0.9406, "step": 10627 }, { "epoch": 0.9, "learning_rate": 5.19278422149746e-07, "loss": 1.0052, "step": 10628 }, { "epoch": 0.9, "learning_rate": 5.18406343454354e-07, "loss": 1.0371, "step": 10629 }, { "epoch": 0.9, "learning_rate": 5.175349781744965e-07, "loss": 0.9911, "step": 10630 }, { "epoch": 0.9, "learning_rate": 5.166643263757332e-07, "loss": 0.9309, "step": 10631 }, { "epoch": 0.9, "learning_rate": 5.15794388123576e-07, "loss": 0.9095, "step": 10632 }, { "epoch": 0.9, "learning_rate": 5.149251634834796e-07, "loss": 0.974, "step": 10633 }, { "epoch": 0.9, "learning_rate": 5.14056652520849e-07, "loss": 0.9676, "step": 10634 }, { "epoch": 0.9, "learning_rate": 5.131888553010278e-07, "loss": 0.9726, "step": 10635 }, { "epoch": 0.9, "learning_rate": 5.12321771889317e-07, "loss": 0.9906, "step": 10636 }, { "epoch": 0.9, "learning_rate": 5.114554023509521e-07, "loss": 0.9577, "step": 10637 }, { "epoch": 0.9, "learning_rate": 5.105897467511256e-07, "loss": 0.9441, "step": 10638 }, { "epoch": 0.9, "learning_rate": 5.097248051549675e-07, "loss": 0.9555, "step": 10639 }, { "epoch": 0.9, "learning_rate": 5.08860577627559e-07, "loss": 0.9394, "step": 10640 }, { "epoch": 0.9, "learning_rate": 5.079970642339283e-07, "loss": 0.9993, "step": 10641 }, { "epoch": 0.9, "learning_rate": 5.071342650390454e-07, "loss": 0.9389, "step": 10642 }, { "epoch": 0.9, "learning_rate": 5.062721801078319e-07, "loss": 0.9495, "step": 10643 }, { "epoch": 0.9, "learning_rate": 5.054108095051491e-07, "loss": 1.0121, "step": 10644 }, { "epoch": 0.9, "learning_rate": 5.045501532958119e-07, "loss": 0.9804, "step": 10645 }, { "epoch": 0.9, "learning_rate": 5.036902115445774e-07, "loss": 0.9815, "step": 10646 }, { "epoch": 0.9, "learning_rate": 5.028309843161471e-07, "loss": 0.9834, "step": 10647 }, { "epoch": 0.9, "learning_rate": 5.019724716751739e-07, "loss": 0.9718, "step": 10648 }, { "epoch": 0.9, "learning_rate": 5.011146736862537e-07, "loss": 0.9412, "step": 10649 }, { "epoch": 0.9, "learning_rate": 5.002575904139262e-07, "loss": 0.9355, "step": 10650 }, { "epoch": 0.9, "learning_rate": 4.99401221922684e-07, "loss": 0.9827, "step": 10651 }, { "epoch": 0.9, "learning_rate": 4.985455682769591e-07, "loss": 0.9391, "step": 10652 }, { "epoch": 0.9, "learning_rate": 4.976906295411343e-07, "loss": 0.9972, "step": 10653 }, { "epoch": 0.9, "learning_rate": 4.968364057795372e-07, "loss": 0.9818, "step": 10654 }, { "epoch": 0.9, "learning_rate": 4.959828970564407e-07, "loss": 0.9892, "step": 10655 }, { "epoch": 0.9, "learning_rate": 4.951301034360645e-07, "loss": 0.9568, "step": 10656 }, { "epoch": 0.9, "learning_rate": 4.942780249825763e-07, "loss": 0.9429, "step": 10657 }, { "epoch": 0.9, "learning_rate": 4.934266617600858e-07, "loss": 1.0188, "step": 10658 }, { "epoch": 0.9, "learning_rate": 4.925760138326541e-07, "loss": 0.9293, "step": 10659 }, { "epoch": 0.9, "learning_rate": 4.917260812642832e-07, "loss": 0.9679, "step": 10660 }, { "epoch": 0.9, "learning_rate": 4.908768641189254e-07, "loss": 0.9892, "step": 10661 }, { "epoch": 0.9, "learning_rate": 4.900283624604763e-07, "loss": 0.9949, "step": 10662 }, { "epoch": 0.9, "learning_rate": 4.891805763527824e-07, "loss": 0.956, "step": 10663 }, { "epoch": 0.9, "learning_rate": 4.883335058596284e-07, "loss": 0.9427, "step": 10664 }, { "epoch": 0.9, "learning_rate": 4.874871510447532e-07, "loss": 0.9571, "step": 10665 }, { "epoch": 0.9, "learning_rate": 4.866415119718348e-07, "loss": 0.9587, "step": 10666 }, { "epoch": 0.9, "learning_rate": 4.857965887045035e-07, "loss": 0.9179, "step": 10667 }, { "epoch": 0.9, "learning_rate": 4.84952381306335e-07, "loss": 0.9805, "step": 10668 }, { "epoch": 0.9, "learning_rate": 4.841088898408453e-07, "loss": 0.9608, "step": 10669 }, { "epoch": 0.9, "learning_rate": 4.832661143715034e-07, "loss": 0.9581, "step": 10670 }, { "epoch": 0.9, "learning_rate": 4.824240549617187e-07, "loss": 0.9541, "step": 10671 }, { "epoch": 0.9, "learning_rate": 4.815827116748528e-07, "loss": 0.9687, "step": 10672 }, { "epoch": 0.9, "learning_rate": 4.807420845742084e-07, "loss": 0.801, "step": 10673 }, { "epoch": 0.9, "learning_rate": 4.799021737230347e-07, "loss": 0.8633, "step": 10674 }, { "epoch": 0.9, "learning_rate": 4.790629791845324e-07, "loss": 0.9535, "step": 10675 }, { "epoch": 0.9, "learning_rate": 4.782245010218411e-07, "loss": 0.9028, "step": 10676 }, { "epoch": 0.9, "learning_rate": 4.773867392980491e-07, "loss": 0.9254, "step": 10677 }, { "epoch": 0.9, "learning_rate": 4.7654969407619487e-07, "loss": 0.8849, "step": 10678 }, { "epoch": 0.9, "learning_rate": 4.7571336541925584e-07, "loss": 0.963, "step": 10679 }, { "epoch": 0.9, "learning_rate": 4.748777533901616e-07, "loss": 0.9315, "step": 10680 }, { "epoch": 0.9, "learning_rate": 4.7404285805178417e-07, "loss": 1.0032, "step": 10681 }, { "epoch": 0.9, "learning_rate": 4.732086794669444e-07, "loss": 0.944, "step": 10682 }, { "epoch": 0.9, "learning_rate": 4.723752176984042e-07, "loss": 0.9398, "step": 10683 }, { "epoch": 0.9, "learning_rate": 4.715424728088802e-07, "loss": 0.9767, "step": 10684 }, { "epoch": 0.9, "learning_rate": 4.707104448610245e-07, "loss": 0.9717, "step": 10685 }, { "epoch": 0.9, "learning_rate": 4.698791339174458e-07, "loss": 0.9692, "step": 10686 }, { "epoch": 0.91, "learning_rate": 4.690485400406886e-07, "loss": 0.9715, "step": 10687 }, { "epoch": 0.91, "learning_rate": 4.682186632932528e-07, "loss": 0.8754, "step": 10688 }, { "epoch": 0.91, "learning_rate": 4.673895037375775e-07, "loss": 0.945, "step": 10689 }, { "epoch": 0.91, "learning_rate": 4.665610614360527e-07, "loss": 1.0137, "step": 10690 }, { "epoch": 0.91, "learning_rate": 4.657333364510097e-07, "loss": 0.9964, "step": 10691 }, { "epoch": 0.91, "learning_rate": 4.649063288447297e-07, "loss": 1.0022, "step": 10692 }, { "epoch": 0.91, "learning_rate": 4.6408003867943974e-07, "loss": 0.9538, "step": 10693 }, { "epoch": 0.91, "learning_rate": 4.632544660173088e-07, "loss": 0.924, "step": 10694 }, { "epoch": 0.91, "learning_rate": 4.6242961092045737e-07, "loss": 0.9977, "step": 10695 }, { "epoch": 0.91, "learning_rate": 4.6160547345094785e-07, "loss": 0.955, "step": 10696 }, { "epoch": 0.91, "learning_rate": 4.6078205367079076e-07, "loss": 0.9725, "step": 10697 }, { "epoch": 0.91, "learning_rate": 4.5995935164194205e-07, "loss": 0.9922, "step": 10698 }, { "epoch": 0.91, "learning_rate": 4.591373674263022e-07, "loss": 0.9867, "step": 10699 }, { "epoch": 0.91, "learning_rate": 4.5831610108572067e-07, "loss": 0.9811, "step": 10700 }, { "epoch": 0.91, "learning_rate": 4.574955526819902e-07, "loss": 0.9824, "step": 10701 }, { "epoch": 0.91, "learning_rate": 4.5667572227685254e-07, "loss": 0.9593, "step": 10702 }, { "epoch": 0.91, "learning_rate": 4.5585660993199166e-07, "loss": 1.0095, "step": 10703 }, { "epoch": 0.91, "learning_rate": 4.5503821570903827e-07, "loss": 0.9606, "step": 10704 }, { "epoch": 0.91, "learning_rate": 4.5422053966957314e-07, "loss": 0.8438, "step": 10705 }, { "epoch": 0.91, "learning_rate": 4.53403581875117e-07, "loss": 0.9725, "step": 10706 }, { "epoch": 0.91, "learning_rate": 4.5258734238714297e-07, "loss": 0.988, "step": 10707 }, { "epoch": 0.91, "learning_rate": 4.51771821267063e-07, "loss": 0.9774, "step": 10708 }, { "epoch": 0.91, "learning_rate": 4.5095701857624244e-07, "loss": 0.9611, "step": 10709 }, { "epoch": 0.91, "learning_rate": 4.5014293437598446e-07, "loss": 0.9429, "step": 10710 }, { "epoch": 0.91, "learning_rate": 4.4932956872754564e-07, "loss": 0.9468, "step": 10711 }, { "epoch": 0.91, "learning_rate": 4.485169216921248e-07, "loss": 0.9438, "step": 10712 }, { "epoch": 0.91, "learning_rate": 4.4770499333086634e-07, "loss": 0.9865, "step": 10713 }, { "epoch": 0.91, "learning_rate": 4.468937837048626e-07, "loss": 0.9544, "step": 10714 }, { "epoch": 0.91, "learning_rate": 4.4608329287515016e-07, "loss": 0.976, "step": 10715 }, { "epoch": 0.91, "learning_rate": 4.4527352090271147e-07, "loss": 0.9403, "step": 10716 }, { "epoch": 0.91, "learning_rate": 4.444644678484777e-07, "loss": 0.9315, "step": 10717 }, { "epoch": 0.91, "learning_rate": 4.4365613377332027e-07, "loss": 0.9608, "step": 10718 }, { "epoch": 0.91, "learning_rate": 4.428485187380627e-07, "loss": 0.9702, "step": 10719 }, { "epoch": 0.91, "learning_rate": 4.4204162280347186e-07, "loss": 0.9616, "step": 10720 }, { "epoch": 0.91, "learning_rate": 4.4123544603025924e-07, "loss": 0.9039, "step": 10721 }, { "epoch": 0.91, "learning_rate": 4.4042998847908304e-07, "loss": 0.9687, "step": 10722 }, { "epoch": 0.91, "learning_rate": 4.3962525021054804e-07, "loss": 0.9158, "step": 10723 }, { "epoch": 0.91, "learning_rate": 4.388212312852058e-07, "loss": 0.9425, "step": 10724 }, { "epoch": 0.91, "learning_rate": 4.3801793176355024e-07, "loss": 1.0043, "step": 10725 }, { "epoch": 0.91, "learning_rate": 4.37215351706024e-07, "loss": 0.9744, "step": 10726 }, { "epoch": 0.91, "learning_rate": 4.3641349117301557e-07, "loss": 0.9987, "step": 10727 }, { "epoch": 0.91, "learning_rate": 4.3561235022485884e-07, "loss": 0.9577, "step": 10728 }, { "epoch": 0.91, "learning_rate": 4.348119289218322e-07, "loss": 0.9271, "step": 10729 }, { "epoch": 0.91, "learning_rate": 4.3401222732416204e-07, "loss": 0.9589, "step": 10730 }, { "epoch": 0.91, "learning_rate": 4.3321324549201795e-07, "loss": 0.9529, "step": 10731 }, { "epoch": 0.91, "learning_rate": 4.324149834855207e-07, "loss": 0.8977, "step": 10732 }, { "epoch": 0.91, "learning_rate": 4.316174413647289e-07, "loss": 0.9634, "step": 10733 }, { "epoch": 0.91, "learning_rate": 4.3082061918965466e-07, "loss": 0.9729, "step": 10734 }, { "epoch": 0.91, "learning_rate": 4.300245170202499e-07, "loss": 0.9221, "step": 10735 }, { "epoch": 0.91, "learning_rate": 4.2922913491641794e-07, "loss": 0.9472, "step": 10736 }, { "epoch": 0.91, "learning_rate": 4.28434472938003e-07, "loss": 0.9556, "step": 10737 }, { "epoch": 0.91, "learning_rate": 4.2764053114479843e-07, "loss": 0.9801, "step": 10738 }, { "epoch": 0.91, "learning_rate": 4.2684730959653975e-07, "loss": 0.8487, "step": 10739 }, { "epoch": 0.91, "learning_rate": 4.260548083529148e-07, "loss": 0.9509, "step": 10740 }, { "epoch": 0.91, "learning_rate": 4.2526302747354807e-07, "loss": 0.979, "step": 10741 }, { "epoch": 0.91, "learning_rate": 4.2447196701801976e-07, "loss": 0.9393, "step": 10742 }, { "epoch": 0.91, "learning_rate": 4.236816270458466e-07, "loss": 0.9578, "step": 10743 }, { "epoch": 0.91, "learning_rate": 4.228920076164977e-07, "loss": 0.7711, "step": 10744 }, { "epoch": 0.91, "learning_rate": 4.2210310878938674e-07, "loss": 0.9512, "step": 10745 }, { "epoch": 0.91, "learning_rate": 4.213149306238695e-07, "loss": 0.8254, "step": 10746 }, { "epoch": 0.91, "learning_rate": 4.2052747317925304e-07, "loss": 0.9776, "step": 10747 }, { "epoch": 0.91, "learning_rate": 4.1974073651478543e-07, "loss": 0.9446, "step": 10748 }, { "epoch": 0.91, "learning_rate": 4.189547206896638e-07, "loss": 0.9715, "step": 10749 }, { "epoch": 0.91, "learning_rate": 4.181694257630275e-07, "loss": 1.0021, "step": 10750 }, { "epoch": 0.91, "learning_rate": 4.1738485179396603e-07, "loss": 0.9822, "step": 10751 }, { "epoch": 0.91, "learning_rate": 4.1660099884151207e-07, "loss": 0.9681, "step": 10752 }, { "epoch": 0.91, "learning_rate": 4.158178669646418e-07, "loss": 0.9488, "step": 10753 }, { "epoch": 0.91, "learning_rate": 4.150354562222836e-07, "loss": 0.9309, "step": 10754 }, { "epoch": 0.91, "learning_rate": 4.1425376667330596e-07, "loss": 0.9321, "step": 10755 }, { "epoch": 0.91, "learning_rate": 4.134727983765241e-07, "loss": 0.9706, "step": 10756 }, { "epoch": 0.91, "learning_rate": 4.1269255139070095e-07, "loss": 0.9561, "step": 10757 }, { "epoch": 0.91, "learning_rate": 4.1191302577454297e-07, "loss": 0.9241, "step": 10758 }, { "epoch": 0.91, "learning_rate": 4.1113422158670533e-07, "loss": 0.9299, "step": 10759 }, { "epoch": 0.91, "learning_rate": 4.1035613888578465e-07, "loss": 0.8953, "step": 10760 }, { "epoch": 0.91, "learning_rate": 4.0957877773032616e-07, "loss": 0.9801, "step": 10761 }, { "epoch": 0.91, "learning_rate": 4.0880213817882097e-07, "loss": 0.9748, "step": 10762 }, { "epoch": 0.91, "learning_rate": 4.080262202897045e-07, "loss": 1.0186, "step": 10763 }, { "epoch": 0.91, "learning_rate": 4.0725102412135786e-07, "loss": 0.9344, "step": 10764 }, { "epoch": 0.91, "learning_rate": 4.06476549732111e-07, "loss": 0.8898, "step": 10765 }, { "epoch": 0.91, "learning_rate": 4.0570279718023296e-07, "loss": 0.9762, "step": 10766 }, { "epoch": 0.91, "learning_rate": 4.049297665239471e-07, "loss": 0.902, "step": 10767 }, { "epoch": 0.91, "learning_rate": 4.041574578214147e-07, "loss": 0.9108, "step": 10768 }, { "epoch": 0.91, "learning_rate": 4.03385871130747e-07, "loss": 0.9272, "step": 10769 }, { "epoch": 0.91, "learning_rate": 4.0261500650999873e-07, "loss": 0.9185, "step": 10770 }, { "epoch": 0.91, "learning_rate": 4.018448640171735e-07, "loss": 1.0159, "step": 10771 }, { "epoch": 0.91, "learning_rate": 4.010754437102171e-07, "loss": 1.019, "step": 10772 }, { "epoch": 0.91, "learning_rate": 4.003067456470222e-07, "loss": 0.9393, "step": 10773 }, { "epoch": 0.91, "learning_rate": 3.995387698854292e-07, "loss": 0.947, "step": 10774 }, { "epoch": 0.91, "learning_rate": 3.9877151648321955e-07, "loss": 0.9851, "step": 10775 }, { "epoch": 0.91, "learning_rate": 3.9800498549812607e-07, "loss": 0.9765, "step": 10776 }, { "epoch": 0.91, "learning_rate": 3.972391769878203e-07, "loss": 0.9768, "step": 10777 }, { "epoch": 0.91, "learning_rate": 3.964740910099274e-07, "loss": 0.9632, "step": 10778 }, { "epoch": 0.91, "learning_rate": 3.957097276220112e-07, "loss": 0.9255, "step": 10779 }, { "epoch": 0.91, "learning_rate": 3.9494608688158465e-07, "loss": 1.0248, "step": 10780 }, { "epoch": 0.91, "learning_rate": 3.941831688461073e-07, "loss": 0.9592, "step": 10781 }, { "epoch": 0.91, "learning_rate": 3.93420973572981e-07, "loss": 0.9749, "step": 10782 }, { "epoch": 0.91, "learning_rate": 3.926595011195544e-07, "loss": 0.9351, "step": 10783 }, { "epoch": 0.91, "learning_rate": 3.9189875154312497e-07, "loss": 0.962, "step": 10784 }, { "epoch": 0.91, "learning_rate": 3.911387249009302e-07, "loss": 0.9315, "step": 10785 }, { "epoch": 0.91, "learning_rate": 3.9037942125015884e-07, "loss": 1.0158, "step": 10786 }, { "epoch": 0.91, "learning_rate": 3.896208406479407e-07, "loss": 0.9816, "step": 10787 }, { "epoch": 0.91, "learning_rate": 3.8886298315135354e-07, "loss": 0.9728, "step": 10788 }, { "epoch": 0.91, "learning_rate": 3.8810584881741943e-07, "loss": 0.9217, "step": 10789 }, { "epoch": 0.91, "learning_rate": 3.873494377031095e-07, "loss": 1.0374, "step": 10790 }, { "epoch": 0.91, "learning_rate": 3.865937498653338e-07, "loss": 0.9167, "step": 10791 }, { "epoch": 0.91, "learning_rate": 3.8583878536095574e-07, "loss": 0.9565, "step": 10792 }, { "epoch": 0.91, "learning_rate": 3.8508454424677656e-07, "loss": 0.9822, "step": 10793 }, { "epoch": 0.91, "learning_rate": 3.843310265795508e-07, "loss": 0.9782, "step": 10794 }, { "epoch": 0.91, "learning_rate": 3.8357823241597205e-07, "loss": 0.9581, "step": 10795 }, { "epoch": 0.91, "learning_rate": 3.8282616181268274e-07, "loss": 0.972, "step": 10796 }, { "epoch": 0.91, "learning_rate": 3.820748148262721e-07, "loss": 0.9811, "step": 10797 }, { "epoch": 0.91, "learning_rate": 3.813241915132704e-07, "loss": 0.9859, "step": 10798 }, { "epoch": 0.91, "learning_rate": 3.8057429193015915e-07, "loss": 0.7492, "step": 10799 }, { "epoch": 0.91, "learning_rate": 3.798251161333588e-07, "loss": 0.9301, "step": 10800 }, { "epoch": 0.91, "learning_rate": 3.790766641792432e-07, "loss": 0.955, "step": 10801 }, { "epoch": 0.91, "learning_rate": 3.783289361241227e-07, "loss": 0.958, "step": 10802 }, { "epoch": 0.91, "learning_rate": 3.775819320242624e-07, "loss": 0.9382, "step": 10803 }, { "epoch": 0.91, "learning_rate": 3.7683565193586514e-07, "loss": 0.9867, "step": 10804 }, { "epoch": 0.92, "learning_rate": 3.7609009591508485e-07, "loss": 1.0122, "step": 10805 }, { "epoch": 0.92, "learning_rate": 3.753452640180177e-07, "loss": 1.0427, "step": 10806 }, { "epoch": 0.92, "learning_rate": 3.746011563007057e-07, "loss": 0.9861, "step": 10807 }, { "epoch": 0.92, "learning_rate": 3.738577728191406e-07, "loss": 0.994, "step": 10808 }, { "epoch": 0.92, "learning_rate": 3.7311511362925215e-07, "loss": 1.0044, "step": 10809 }, { "epoch": 0.92, "learning_rate": 3.7237317878692006e-07, "loss": 0.9658, "step": 10810 }, { "epoch": 0.92, "learning_rate": 3.7163196834797187e-07, "loss": 0.9444, "step": 10811 }, { "epoch": 0.92, "learning_rate": 3.708914823681753e-07, "loss": 0.9342, "step": 10812 }, { "epoch": 0.92, "learning_rate": 3.7015172090324794e-07, "loss": 0.9598, "step": 10813 }, { "epoch": 0.92, "learning_rate": 3.6941268400884855e-07, "loss": 0.9781, "step": 10814 }, { "epoch": 0.92, "learning_rate": 3.6867437174058606e-07, "loss": 0.9715, "step": 10815 }, { "epoch": 0.92, "learning_rate": 3.679367841540116e-07, "loss": 0.988, "step": 10816 }, { "epoch": 0.92, "learning_rate": 3.671999213046251e-07, "loss": 0.9413, "step": 10817 }, { "epoch": 0.92, "learning_rate": 3.664637832478657e-07, "loss": 0.9878, "step": 10818 }, { "epoch": 0.92, "learning_rate": 3.6572837003912567e-07, "loss": 0.9465, "step": 10819 }, { "epoch": 0.92, "learning_rate": 3.649936817337363e-07, "loss": 0.9629, "step": 10820 }, { "epoch": 0.92, "learning_rate": 3.6425971838698007e-07, "loss": 0.9409, "step": 10821 }, { "epoch": 0.92, "learning_rate": 3.635264800540783e-07, "loss": 1.0201, "step": 10822 }, { "epoch": 0.92, "learning_rate": 3.627939667902036e-07, "loss": 0.9714, "step": 10823 }, { "epoch": 0.92, "learning_rate": 3.62062178650473e-07, "loss": 0.9495, "step": 10824 }, { "epoch": 0.92, "learning_rate": 3.613311156899457e-07, "loss": 0.9175, "step": 10825 }, { "epoch": 0.92, "learning_rate": 3.6060077796363e-07, "loss": 0.9934, "step": 10826 }, { "epoch": 0.92, "learning_rate": 3.598711655264764e-07, "loss": 0.8421, "step": 10827 }, { "epoch": 0.92, "learning_rate": 3.591422784333842e-07, "loss": 0.9358, "step": 10828 }, { "epoch": 0.92, "learning_rate": 3.584141167391952e-07, "loss": 0.9407, "step": 10829 }, { "epoch": 0.92, "learning_rate": 3.576866804986989e-07, "loss": 0.9979, "step": 10830 }, { "epoch": 0.92, "learning_rate": 3.5695996976662817e-07, "loss": 1.0266, "step": 10831 }, { "epoch": 0.92, "learning_rate": 3.5623398459766254e-07, "loss": 0.9801, "step": 10832 }, { "epoch": 0.92, "learning_rate": 3.5550872504642844e-07, "loss": 0.9509, "step": 10833 }, { "epoch": 0.92, "learning_rate": 3.547841911674921e-07, "loss": 0.9864, "step": 10834 }, { "epoch": 0.92, "learning_rate": 3.5406038301537327e-07, "loss": 0.9597, "step": 10835 }, { "epoch": 0.92, "learning_rate": 3.533373006445306e-07, "loss": 0.9445, "step": 10836 }, { "epoch": 0.92, "learning_rate": 3.526149441093685e-07, "loss": 0.9645, "step": 10837 }, { "epoch": 0.92, "learning_rate": 3.5189331346424326e-07, "loss": 0.9716, "step": 10838 }, { "epoch": 0.92, "learning_rate": 3.5117240876344714e-07, "loss": 0.9039, "step": 10839 }, { "epoch": 0.92, "learning_rate": 3.504522300612267e-07, "loss": 1.0019, "step": 10840 }, { "epoch": 0.92, "learning_rate": 3.497327774117676e-07, "loss": 0.9513, "step": 10841 }, { "epoch": 0.92, "learning_rate": 3.4901405086920303e-07, "loss": 0.7548, "step": 10842 }, { "epoch": 0.92, "learning_rate": 3.4829605048761096e-07, "loss": 0.9503, "step": 10843 }, { "epoch": 0.92, "learning_rate": 3.4757877632101813e-07, "loss": 0.8824, "step": 10844 }, { "epoch": 0.92, "learning_rate": 3.468622284233902e-07, "loss": 0.9662, "step": 10845 }, { "epoch": 0.92, "learning_rate": 3.461464068486442e-07, "loss": 0.9705, "step": 10846 }, { "epoch": 0.92, "learning_rate": 3.45431311650638e-07, "loss": 0.9278, "step": 10847 }, { "epoch": 0.92, "learning_rate": 3.447169428831787e-07, "loss": 0.9418, "step": 10848 }, { "epoch": 0.92, "learning_rate": 3.440033006000165e-07, "loss": 0.9624, "step": 10849 }, { "epoch": 0.92, "learning_rate": 3.432903848548474e-07, "loss": 0.9745, "step": 10850 }, { "epoch": 0.92, "learning_rate": 3.42578195701313e-07, "loss": 0.9357, "step": 10851 }, { "epoch": 0.92, "learning_rate": 3.4186673319299814e-07, "loss": 0.9446, "step": 10852 }, { "epoch": 0.92, "learning_rate": 3.4115599738343774e-07, "loss": 0.9278, "step": 10853 }, { "epoch": 0.92, "learning_rate": 3.4044598832610685e-07, "loss": 0.9802, "step": 10854 }, { "epoch": 0.92, "learning_rate": 3.3973670607442943e-07, "loss": 0.9506, "step": 10855 }, { "epoch": 0.92, "learning_rate": 3.3902815068177163e-07, "loss": 0.989, "step": 10856 }, { "epoch": 0.92, "learning_rate": 3.383203222014497e-07, "loss": 0.9339, "step": 10857 }, { "epoch": 0.92, "learning_rate": 3.3761322068671777e-07, "loss": 0.9499, "step": 10858 }, { "epoch": 0.92, "learning_rate": 3.369068461907843e-07, "loss": 0.9402, "step": 10859 }, { "epoch": 0.92, "learning_rate": 3.362011987667957e-07, "loss": 0.9031, "step": 10860 }, { "epoch": 0.92, "learning_rate": 3.3549627846784616e-07, "loss": 1.0095, "step": 10861 }, { "epoch": 0.92, "learning_rate": 3.347920853469777e-07, "loss": 0.9688, "step": 10862 }, { "epoch": 0.92, "learning_rate": 3.340886194571735e-07, "loss": 0.9368, "step": 10863 }, { "epoch": 0.92, "learning_rate": 3.333858808513635e-07, "loss": 0.9737, "step": 10864 }, { "epoch": 0.92, "learning_rate": 3.3268386958242525e-07, "loss": 0.9603, "step": 10865 }, { "epoch": 0.92, "learning_rate": 3.319825857031767e-07, "loss": 0.9906, "step": 10866 }, { "epoch": 0.92, "learning_rate": 3.312820292663876e-07, "loss": 0.955, "step": 10867 }, { "epoch": 0.92, "learning_rate": 3.3058220032476495e-07, "loss": 0.9727, "step": 10868 }, { "epoch": 0.92, "learning_rate": 3.2988309893096984e-07, "loss": 0.9471, "step": 10869 }, { "epoch": 0.92, "learning_rate": 3.2918472513760127e-07, "loss": 1.0238, "step": 10870 }, { "epoch": 0.92, "learning_rate": 3.284870789972072e-07, "loss": 0.9839, "step": 10871 }, { "epoch": 0.92, "learning_rate": 3.277901605622802e-07, "loss": 0.9506, "step": 10872 }, { "epoch": 0.92, "learning_rate": 3.270939698852582e-07, "loss": 0.8436, "step": 10873 }, { "epoch": 0.92, "learning_rate": 3.263985070185216e-07, "loss": 0.8171, "step": 10874 }, { "epoch": 0.92, "learning_rate": 3.257037720144018e-07, "loss": 0.9203, "step": 10875 }, { "epoch": 0.92, "learning_rate": 3.2500976492517154e-07, "loss": 0.8999, "step": 10876 }, { "epoch": 0.92, "learning_rate": 3.2431648580304784e-07, "loss": 0.9164, "step": 10877 }, { "epoch": 0.92, "learning_rate": 3.2362393470019683e-07, "loss": 0.939, "step": 10878 }, { "epoch": 0.92, "learning_rate": 3.229321116687256e-07, "loss": 0.9327, "step": 10879 }, { "epoch": 0.92, "learning_rate": 3.222410167606893e-07, "loss": 0.9156, "step": 10880 }, { "epoch": 0.92, "learning_rate": 3.2155065002808517e-07, "loss": 0.9347, "step": 10881 }, { "epoch": 0.92, "learning_rate": 3.2086101152286274e-07, "loss": 0.9896, "step": 10882 }, { "epoch": 0.92, "learning_rate": 3.201721012969061e-07, "loss": 0.9617, "step": 10883 }, { "epoch": 0.92, "learning_rate": 3.19483919402056e-07, "loss": 0.9491, "step": 10884 }, { "epoch": 0.92, "learning_rate": 3.1879646589008775e-07, "loss": 0.9635, "step": 10885 }, { "epoch": 0.92, "learning_rate": 3.1810974081273096e-07, "loss": 0.9879, "step": 10886 }, { "epoch": 0.92, "learning_rate": 3.1742374422165324e-07, "loss": 0.9734, "step": 10887 }, { "epoch": 0.92, "learning_rate": 3.167384761684711e-07, "loss": 0.966, "step": 10888 }, { "epoch": 0.92, "learning_rate": 3.160539367047466e-07, "loss": 0.9368, "step": 10889 }, { "epoch": 0.92, "learning_rate": 3.153701258819863e-07, "loss": 0.9425, "step": 10890 }, { "epoch": 0.92, "learning_rate": 3.146870437516392e-07, "loss": 0.9706, "step": 10891 }, { "epoch": 0.92, "learning_rate": 3.140046903651051e-07, "loss": 0.7513, "step": 10892 }, { "epoch": 0.92, "learning_rate": 3.133230657737218e-07, "loss": 0.9479, "step": 10893 }, { "epoch": 0.92, "learning_rate": 3.126421700287807e-07, "loss": 1.0021, "step": 10894 }, { "epoch": 0.92, "learning_rate": 3.119620031815096e-07, "loss": 0.9561, "step": 10895 }, { "epoch": 0.92, "learning_rate": 3.1128256528308975e-07, "loss": 0.9685, "step": 10896 }, { "epoch": 0.92, "learning_rate": 3.106038563846392e-07, "loss": 0.9629, "step": 10897 }, { "epoch": 0.92, "learning_rate": 3.099258765372304e-07, "loss": 0.9336, "step": 10898 }, { "epoch": 0.92, "learning_rate": 3.0924862579187144e-07, "loss": 0.9362, "step": 10899 }, { "epoch": 0.92, "learning_rate": 3.0857210419952155e-07, "loss": 0.9717, "step": 10900 }, { "epoch": 0.92, "learning_rate": 3.0789631181108557e-07, "loss": 0.9835, "step": 10901 }, { "epoch": 0.92, "learning_rate": 3.0722124867740957e-07, "loss": 0.8738, "step": 10902 }, { "epoch": 0.92, "learning_rate": 3.065469148492883e-07, "loss": 0.949, "step": 10903 }, { "epoch": 0.92, "learning_rate": 3.0587331037745807e-07, "loss": 1.0062, "step": 10904 }, { "epoch": 0.92, "learning_rate": 3.052004353126037e-07, "loss": 0.8968, "step": 10905 }, { "epoch": 0.92, "learning_rate": 3.0452828970535364e-07, "loss": 1.0146, "step": 10906 }, { "epoch": 0.92, "learning_rate": 3.038568736062819e-07, "loss": 0.9609, "step": 10907 }, { "epoch": 0.92, "learning_rate": 3.031861870659059e-07, "loss": 0.9368, "step": 10908 }, { "epoch": 0.92, "learning_rate": 3.0251623013469065e-07, "loss": 0.9487, "step": 10909 }, { "epoch": 0.92, "learning_rate": 3.0184700286304494e-07, "loss": 0.9782, "step": 10910 }, { "epoch": 0.92, "learning_rate": 3.0117850530132385e-07, "loss": 0.9629, "step": 10911 }, { "epoch": 0.92, "learning_rate": 3.0051073749982393e-07, "loss": 0.9512, "step": 10912 }, { "epoch": 0.92, "learning_rate": 2.998436995087928e-07, "loss": 0.9697, "step": 10913 }, { "epoch": 0.92, "learning_rate": 2.9917739137841797e-07, "loss": 0.946, "step": 10914 }, { "epoch": 0.92, "learning_rate": 2.985118131588327e-07, "loss": 0.9815, "step": 10915 }, { "epoch": 0.92, "learning_rate": 2.9784696490011924e-07, "loss": 0.9605, "step": 10916 }, { "epoch": 0.92, "learning_rate": 2.971828466523008e-07, "loss": 0.9268, "step": 10917 }, { "epoch": 0.92, "learning_rate": 2.9651945846534635e-07, "loss": 0.9414, "step": 10918 }, { "epoch": 0.92, "learning_rate": 2.958568003891726e-07, "loss": 0.9886, "step": 10919 }, { "epoch": 0.92, "learning_rate": 2.9519487247363645e-07, "loss": 0.94, "step": 10920 }, { "epoch": 0.92, "learning_rate": 2.945336747685468e-07, "loss": 0.9538, "step": 10921 }, { "epoch": 0.92, "learning_rate": 2.9387320732364945e-07, "loss": 0.9092, "step": 10922 }, { "epoch": 0.93, "learning_rate": 2.9321347018864245e-07, "loss": 0.9508, "step": 10923 }, { "epoch": 0.93, "learning_rate": 2.925544634131638e-07, "loss": 0.9269, "step": 10924 }, { "epoch": 0.93, "learning_rate": 2.918961870467996e-07, "loss": 0.9587, "step": 10925 }, { "epoch": 0.93, "learning_rate": 2.9123864113907995e-07, "loss": 0.9136, "step": 10926 }, { "epoch": 0.93, "learning_rate": 2.905818257394799e-07, "loss": 0.9321, "step": 10927 }, { "epoch": 0.93, "learning_rate": 2.899257408974199e-07, "loss": 1.0028, "step": 10928 }, { "epoch": 0.93, "learning_rate": 2.892703866622648e-07, "loss": 0.942, "step": 10929 }, { "epoch": 0.93, "learning_rate": 2.8861576308332526e-07, "loss": 0.9387, "step": 10930 }, { "epoch": 0.93, "learning_rate": 2.8796187020985633e-07, "loss": 0.9975, "step": 10931 }, { "epoch": 0.93, "learning_rate": 2.873087080910586e-07, "loss": 0.9521, "step": 10932 }, { "epoch": 0.93, "learning_rate": 2.866562767760761e-07, "loss": 0.939, "step": 10933 }, { "epoch": 0.93, "learning_rate": 2.860045763140018e-07, "loss": 0.9382, "step": 10934 }, { "epoch": 0.93, "learning_rate": 2.853536067538676e-07, "loss": 0.9188, "step": 10935 }, { "epoch": 0.93, "learning_rate": 2.847033681446576e-07, "loss": 0.9736, "step": 10936 }, { "epoch": 0.93, "learning_rate": 2.840538605352927e-07, "loss": 0.9654, "step": 10937 }, { "epoch": 0.93, "learning_rate": 2.834050839746483e-07, "loss": 0.9572, "step": 10938 }, { "epoch": 0.93, "learning_rate": 2.8275703851153635e-07, "loss": 0.9852, "step": 10939 }, { "epoch": 0.93, "learning_rate": 2.8210972419471793e-07, "loss": 0.9617, "step": 10940 }, { "epoch": 0.93, "learning_rate": 2.814631410728996e-07, "loss": 0.9501, "step": 10941 }, { "epoch": 0.93, "learning_rate": 2.808172891947292e-07, "loss": 0.9669, "step": 10942 }, { "epoch": 0.93, "learning_rate": 2.801721686088044e-07, "loss": 0.9824, "step": 10943 }, { "epoch": 0.93, "learning_rate": 2.795277793636642e-07, "loss": 0.9562, "step": 10944 }, { "epoch": 0.93, "learning_rate": 2.788841215077931e-07, "loss": 0.9454, "step": 10945 }, { "epoch": 0.93, "learning_rate": 2.782411950896235e-07, "loss": 0.9449, "step": 10946 }, { "epoch": 0.93, "learning_rate": 2.7759900015752793e-07, "loss": 0.959, "step": 10947 }, { "epoch": 0.93, "learning_rate": 2.769575367598298e-07, "loss": 0.8828, "step": 10948 }, { "epoch": 0.93, "learning_rate": 2.7631680494479064e-07, "loss": 0.9373, "step": 10949 }, { "epoch": 0.93, "learning_rate": 2.7567680476062287e-07, "loss": 0.8943, "step": 10950 }, { "epoch": 0.93, "learning_rate": 2.7503753625548135e-07, "loss": 0.984, "step": 10951 }, { "epoch": 0.93, "learning_rate": 2.743989994774643e-07, "loss": 0.962, "step": 10952 }, { "epoch": 0.93, "learning_rate": 2.737611944746188e-07, "loss": 0.984, "step": 10953 }, { "epoch": 0.93, "learning_rate": 2.7312412129493313e-07, "loss": 0.9379, "step": 10954 }, { "epoch": 0.93, "learning_rate": 2.7248777998634347e-07, "loss": 0.9465, "step": 10955 }, { "epoch": 0.93, "learning_rate": 2.7185217059672695e-07, "loss": 0.9555, "step": 10956 }, { "epoch": 0.93, "learning_rate": 2.712172931739121e-07, "loss": 0.9904, "step": 10957 }, { "epoch": 0.93, "learning_rate": 2.7058314776566396e-07, "loss": 0.9353, "step": 10958 }, { "epoch": 0.93, "learning_rate": 2.6994973441970106e-07, "loss": 0.9935, "step": 10959 }, { "epoch": 0.93, "learning_rate": 2.693170531836797e-07, "loss": 0.9508, "step": 10960 }, { "epoch": 0.93, "learning_rate": 2.6868510410520634e-07, "loss": 0.9237, "step": 10961 }, { "epoch": 0.93, "learning_rate": 2.6805388723182835e-07, "loss": 0.9572, "step": 10962 }, { "epoch": 0.93, "learning_rate": 2.6742340261104226e-07, "loss": 0.9697, "step": 10963 }, { "epoch": 0.93, "learning_rate": 2.6679365029028457e-07, "loss": 0.9454, "step": 10964 }, { "epoch": 0.93, "learning_rate": 2.6616463031694073e-07, "loss": 0.9172, "step": 10965 }, { "epoch": 0.93, "learning_rate": 2.655363427383395e-07, "loss": 0.9762, "step": 10966 }, { "epoch": 0.93, "learning_rate": 2.649087876017531e-07, "loss": 0.9941, "step": 10967 }, { "epoch": 0.93, "learning_rate": 2.642819649544026e-07, "loss": 0.9581, "step": 10968 }, { "epoch": 0.93, "learning_rate": 2.6365587484344924e-07, "loss": 0.9537, "step": 10969 }, { "epoch": 0.93, "learning_rate": 2.630305173160019e-07, "loss": 0.9178, "step": 10970 }, { "epoch": 0.93, "learning_rate": 2.624058924191153e-07, "loss": 0.9199, "step": 10971 }, { "epoch": 0.93, "learning_rate": 2.6178200019978506e-07, "loss": 0.963, "step": 10972 }, { "epoch": 0.93, "learning_rate": 2.61158840704957e-07, "loss": 0.9903, "step": 10973 }, { "epoch": 0.93, "learning_rate": 2.60536413981517e-07, "loss": 0.9834, "step": 10974 }, { "epoch": 0.93, "learning_rate": 2.599147200762986e-07, "loss": 0.976, "step": 10975 }, { "epoch": 0.93, "learning_rate": 2.592937590360778e-07, "loss": 0.9599, "step": 10976 }, { "epoch": 0.93, "learning_rate": 2.586735309075783e-07, "loss": 0.9005, "step": 10977 }, { "epoch": 0.93, "learning_rate": 2.5805403573746943e-07, "loss": 0.9368, "step": 10978 }, { "epoch": 0.93, "learning_rate": 2.5743527357235955e-07, "loss": 0.9423, "step": 10979 }, { "epoch": 0.93, "learning_rate": 2.5681724445880906e-07, "loss": 0.9628, "step": 10980 }, { "epoch": 0.93, "learning_rate": 2.561999484433175e-07, "loss": 0.9238, "step": 10981 }, { "epoch": 0.93, "learning_rate": 2.555833855723322e-07, "loss": 0.9413, "step": 10982 }, { "epoch": 0.93, "learning_rate": 2.5496755589224485e-07, "loss": 0.9876, "step": 10983 }, { "epoch": 0.93, "learning_rate": 2.543524594493918e-07, "loss": 0.9594, "step": 10984 }, { "epoch": 0.93, "learning_rate": 2.5373809629005375e-07, "loss": 0.9699, "step": 10985 }, { "epoch": 0.93, "learning_rate": 2.531244664604582e-07, "loss": 0.9625, "step": 10986 }, { "epoch": 0.93, "learning_rate": 2.5251157000677373e-07, "loss": 0.9599, "step": 10987 }, { "epoch": 0.93, "learning_rate": 2.518994069751179e-07, "loss": 0.9664, "step": 10988 }, { "epoch": 0.93, "learning_rate": 2.5128797741154953e-07, "loss": 0.9273, "step": 10989 }, { "epoch": 0.93, "learning_rate": 2.5067728136207504e-07, "loss": 0.9565, "step": 10990 }, { "epoch": 0.93, "learning_rate": 2.5006731887264434e-07, "loss": 0.9383, "step": 10991 }, { "epoch": 0.93, "learning_rate": 2.49458089989153e-07, "loss": 0.9549, "step": 10992 }, { "epoch": 0.93, "learning_rate": 2.4884959475743985e-07, "loss": 0.9476, "step": 10993 }, { "epoch": 0.93, "learning_rate": 2.482418332232883e-07, "loss": 0.9402, "step": 10994 }, { "epoch": 0.93, "learning_rate": 2.4763480543243066e-07, "loss": 1.022, "step": 10995 }, { "epoch": 0.93, "learning_rate": 2.4702851143053817e-07, "loss": 0.9694, "step": 10996 }, { "epoch": 0.93, "learning_rate": 2.464229512632299e-07, "loss": 0.9289, "step": 10997 }, { "epoch": 0.93, "learning_rate": 2.458181249760716e-07, "loss": 0.9279, "step": 10998 }, { "epoch": 0.93, "learning_rate": 2.4521403261457023e-07, "loss": 0.9507, "step": 10999 }, { "epoch": 0.93, "learning_rate": 2.4461067422417826e-07, "loss": 0.9477, "step": 11000 }, { "epoch": 0.93, "learning_rate": 2.44008049850295e-07, "loss": 0.9212, "step": 11001 }, { "epoch": 0.93, "learning_rate": 2.434061595382631e-07, "loss": 0.9493, "step": 11002 }, { "epoch": 0.93, "learning_rate": 2.4280500333336953e-07, "loss": 0.9507, "step": 11003 }, { "epoch": 0.93, "learning_rate": 2.42204581280846e-07, "loss": 1.0024, "step": 11004 }, { "epoch": 0.93, "learning_rate": 2.416048934258719e-07, "loss": 0.9778, "step": 11005 }, { "epoch": 0.93, "learning_rate": 2.410059398135667e-07, "loss": 0.9741, "step": 11006 }, { "epoch": 0.93, "learning_rate": 2.4040772048899875e-07, "loss": 0.9442, "step": 11007 }, { "epoch": 0.93, "learning_rate": 2.398102354971776e-07, "loss": 0.9089, "step": 11008 }, { "epoch": 0.93, "learning_rate": 2.392134848830607e-07, "loss": 0.9356, "step": 11009 }, { "epoch": 0.93, "learning_rate": 2.386174686915466e-07, "loss": 0.967, "step": 11010 }, { "epoch": 0.93, "learning_rate": 2.3802218696748482e-07, "loss": 0.9098, "step": 11011 }, { "epoch": 0.93, "learning_rate": 2.3742763975566074e-07, "loss": 0.9532, "step": 11012 }, { "epoch": 0.93, "learning_rate": 2.3683382710081415e-07, "loss": 0.9442, "step": 11013 }, { "epoch": 0.93, "learning_rate": 2.362407490476215e-07, "loss": 0.9929, "step": 11014 }, { "epoch": 0.93, "learning_rate": 2.3564840564070933e-07, "loss": 0.9616, "step": 11015 }, { "epoch": 0.93, "learning_rate": 2.3505679692464423e-07, "loss": 0.9589, "step": 11016 }, { "epoch": 0.93, "learning_rate": 2.3446592294394277e-07, "loss": 0.9106, "step": 11017 }, { "epoch": 0.93, "learning_rate": 2.3387578374306275e-07, "loss": 0.9264, "step": 11018 }, { "epoch": 0.93, "learning_rate": 2.3328637936640751e-07, "loss": 0.9843, "step": 11019 }, { "epoch": 0.93, "learning_rate": 2.3269770985832497e-07, "loss": 0.9786, "step": 11020 }, { "epoch": 0.93, "learning_rate": 2.3210977526310628e-07, "loss": 0.8985, "step": 11021 }, { "epoch": 0.93, "learning_rate": 2.315225756249928e-07, "loss": 0.9901, "step": 11022 }, { "epoch": 0.93, "learning_rate": 2.3093611098816248e-07, "loss": 1.015, "step": 11023 }, { "epoch": 0.93, "learning_rate": 2.303503813967445e-07, "loss": 0.9341, "step": 11024 }, { "epoch": 0.93, "learning_rate": 2.2976538689481131e-07, "loss": 0.9157, "step": 11025 }, { "epoch": 0.93, "learning_rate": 2.2918112752637555e-07, "loss": 0.9666, "step": 11026 }, { "epoch": 0.93, "learning_rate": 2.2859760333540205e-07, "loss": 0.9888, "step": 11027 }, { "epoch": 0.93, "learning_rate": 2.2801481436579344e-07, "loss": 0.926, "step": 11028 }, { "epoch": 0.93, "learning_rate": 2.2743276066140242e-07, "loss": 0.9482, "step": 11029 }, { "epoch": 0.93, "learning_rate": 2.2685144226602286e-07, "loss": 0.9761, "step": 11030 }, { "epoch": 0.93, "learning_rate": 2.262708592233942e-07, "loss": 1.0191, "step": 11031 }, { "epoch": 0.93, "learning_rate": 2.2569101157720152e-07, "loss": 1.0133, "step": 11032 }, { "epoch": 0.93, "learning_rate": 2.2511189937107215e-07, "loss": 0.9323, "step": 11033 }, { "epoch": 0.93, "learning_rate": 2.245335226485823e-07, "loss": 0.8871, "step": 11034 }, { "epoch": 0.93, "learning_rate": 2.2395588145324832e-07, "loss": 0.9547, "step": 11035 }, { "epoch": 0.93, "learning_rate": 2.233789758285343e-07, "loss": 0.9699, "step": 11036 }, { "epoch": 0.93, "learning_rate": 2.2280280581784554e-07, "loss": 0.9705, "step": 11037 }, { "epoch": 0.93, "learning_rate": 2.2222737146453843e-07, "loss": 0.9493, "step": 11038 }, { "epoch": 0.93, "learning_rate": 2.2165267281190507e-07, "loss": 0.9268, "step": 11039 }, { "epoch": 0.93, "learning_rate": 2.2107870990319191e-07, "loss": 0.9304, "step": 11040 }, { "epoch": 0.94, "learning_rate": 2.2050548278158113e-07, "loss": 0.8458, "step": 11041 }, { "epoch": 0.94, "learning_rate": 2.1993299149020598e-07, "loss": 0.9126, "step": 11042 }, { "epoch": 0.94, "learning_rate": 2.1936123607214088e-07, "loss": 0.9992, "step": 11043 }, { "epoch": 0.94, "learning_rate": 2.187902165704059e-07, "loss": 0.9537, "step": 11044 }, { "epoch": 0.94, "learning_rate": 2.1821993302796662e-07, "loss": 0.9709, "step": 11045 }, { "epoch": 0.94, "learning_rate": 2.176503854877321e-07, "loss": 0.9242, "step": 11046 }, { "epoch": 0.94, "learning_rate": 2.170815739925558e-07, "loss": 1.0054, "step": 11047 }, { "epoch": 0.94, "learning_rate": 2.1651349858523574e-07, "loss": 0.9376, "step": 11048 }, { "epoch": 0.94, "learning_rate": 2.159461593085177e-07, "loss": 0.9803, "step": 11049 }, { "epoch": 0.94, "learning_rate": 2.1537955620508755e-07, "loss": 0.9236, "step": 11050 }, { "epoch": 0.94, "learning_rate": 2.1481368931757673e-07, "loss": 0.9138, "step": 11051 }, { "epoch": 0.94, "learning_rate": 2.1424855868856453e-07, "loss": 1.0388, "step": 11052 }, { "epoch": 0.94, "learning_rate": 2.1368416436057026e-07, "loss": 0.9631, "step": 11053 }, { "epoch": 0.94, "learning_rate": 2.131205063760633e-07, "loss": 0.9352, "step": 11054 }, { "epoch": 0.94, "learning_rate": 2.1255758477745082e-07, "loss": 0.95, "step": 11055 }, { "epoch": 0.94, "learning_rate": 2.1199539960709115e-07, "loss": 0.9768, "step": 11056 }, { "epoch": 0.94, "learning_rate": 2.114339509072838e-07, "loss": 0.9657, "step": 11057 }, { "epoch": 0.94, "learning_rate": 2.1087323872027166e-07, "loss": 0.9806, "step": 11058 }, { "epoch": 0.94, "learning_rate": 2.103132630882465e-07, "loss": 0.9051, "step": 11059 }, { "epoch": 0.94, "learning_rate": 2.0975402405334023e-07, "loss": 0.9146, "step": 11060 }, { "epoch": 0.94, "learning_rate": 2.0919552165763135e-07, "loss": 0.951, "step": 11061 }, { "epoch": 0.94, "learning_rate": 2.0863775594314295e-07, "loss": 0.9068, "step": 11062 }, { "epoch": 0.94, "learning_rate": 2.0808072695184367e-07, "loss": 0.9198, "step": 11063 }, { "epoch": 0.94, "learning_rate": 2.0752443472564443e-07, "loss": 0.964, "step": 11064 }, { "epoch": 0.94, "learning_rate": 2.0696887930640174e-07, "loss": 0.9304, "step": 11065 }, { "epoch": 0.94, "learning_rate": 2.0641406073591664e-07, "loss": 0.9523, "step": 11066 }, { "epoch": 0.94, "learning_rate": 2.058599790559368e-07, "loss": 0.9252, "step": 11067 }, { "epoch": 0.94, "learning_rate": 2.053066343081489e-07, "loss": 0.9558, "step": 11068 }, { "epoch": 0.94, "learning_rate": 2.0475402653419186e-07, "loss": 0.9327, "step": 11069 }, { "epoch": 0.94, "learning_rate": 2.042021557756424e-07, "loss": 0.886, "step": 11070 }, { "epoch": 0.94, "learning_rate": 2.036510220740251e-07, "loss": 1.0024, "step": 11071 }, { "epoch": 0.94, "learning_rate": 2.031006254708079e-07, "loss": 0.9654, "step": 11072 }, { "epoch": 0.94, "learning_rate": 2.0255096600740654e-07, "loss": 0.943, "step": 11073 }, { "epoch": 0.94, "learning_rate": 2.0200204372517573e-07, "loss": 0.9492, "step": 11074 }, { "epoch": 0.94, "learning_rate": 2.014538586654169e-07, "loss": 0.9574, "step": 11075 }, { "epoch": 0.94, "learning_rate": 2.0090641086938034e-07, "loss": 0.9232, "step": 11076 }, { "epoch": 0.94, "learning_rate": 2.0035970037825426e-07, "loss": 0.957, "step": 11077 }, { "epoch": 0.94, "learning_rate": 1.998137272331735e-07, "loss": 0.9318, "step": 11078 }, { "epoch": 0.94, "learning_rate": 1.9926849147522187e-07, "loss": 0.9488, "step": 11079 }, { "epoch": 0.94, "learning_rate": 1.98723993145421e-07, "loss": 0.9699, "step": 11080 }, { "epoch": 0.94, "learning_rate": 1.9818023228474038e-07, "loss": 0.9764, "step": 11081 }, { "epoch": 0.94, "learning_rate": 1.976372089340961e-07, "loss": 0.9271, "step": 11082 }, { "epoch": 0.94, "learning_rate": 1.9709492313434442e-07, "loss": 1.0027, "step": 11083 }, { "epoch": 0.94, "learning_rate": 1.9655337492628824e-07, "loss": 0.985, "step": 11084 }, { "epoch": 0.94, "learning_rate": 1.9601256435067495e-07, "loss": 0.9539, "step": 11085 }, { "epoch": 0.94, "learning_rate": 1.9547249144819647e-07, "loss": 0.9519, "step": 11086 }, { "epoch": 0.94, "learning_rate": 1.9493315625948915e-07, "loss": 0.9662, "step": 11087 }, { "epoch": 0.94, "learning_rate": 1.9439455882513392e-07, "loss": 0.8835, "step": 11088 }, { "epoch": 0.94, "learning_rate": 1.9385669918565496e-07, "loss": 0.9678, "step": 11089 }, { "epoch": 0.94, "learning_rate": 1.933195773815244e-07, "loss": 0.9129, "step": 11090 }, { "epoch": 0.94, "learning_rate": 1.9278319345315323e-07, "loss": 0.9954, "step": 11091 }, { "epoch": 0.94, "learning_rate": 1.9224754744090247e-07, "loss": 0.9529, "step": 11092 }, { "epoch": 0.94, "learning_rate": 1.9171263938507323e-07, "loss": 0.956, "step": 11093 }, { "epoch": 0.94, "learning_rate": 1.9117846932591554e-07, "loss": 0.9518, "step": 11094 }, { "epoch": 0.94, "learning_rate": 1.906450373036195e-07, "loss": 0.9662, "step": 11095 }, { "epoch": 0.94, "learning_rate": 1.9011234335832408e-07, "loss": 0.9473, "step": 11096 }, { "epoch": 0.94, "learning_rate": 1.8958038753010721e-07, "loss": 0.9388, "step": 11097 }, { "epoch": 0.94, "learning_rate": 1.890491698589969e-07, "loss": 0.8667, "step": 11098 }, { "epoch": 0.94, "learning_rate": 1.8851869038496118e-07, "loss": 0.9982, "step": 11099 }, { "epoch": 0.94, "learning_rate": 1.87988949147917e-07, "loss": 0.9312, "step": 11100 }, { "epoch": 0.94, "learning_rate": 1.8745994618772135e-07, "loss": 0.9686, "step": 11101 }, { "epoch": 0.94, "learning_rate": 1.8693168154417685e-07, "loss": 0.8209, "step": 11102 }, { "epoch": 0.94, "learning_rate": 1.8640415525703282e-07, "loss": 0.9808, "step": 11103 }, { "epoch": 0.94, "learning_rate": 1.8587736736598194e-07, "loss": 0.92, "step": 11104 }, { "epoch": 0.94, "learning_rate": 1.8535131791065808e-07, "loss": 0.9623, "step": 11105 }, { "epoch": 0.94, "learning_rate": 1.8482600693064512e-07, "loss": 0.9747, "step": 11106 }, { "epoch": 0.94, "learning_rate": 1.84301434465467e-07, "loss": 0.949, "step": 11107 }, { "epoch": 0.94, "learning_rate": 1.8377760055459437e-07, "loss": 0.941, "step": 11108 }, { "epoch": 0.94, "learning_rate": 1.8325450523744238e-07, "loss": 0.9963, "step": 11109 }, { "epoch": 0.94, "learning_rate": 1.827321485533684e-07, "loss": 0.9641, "step": 11110 }, { "epoch": 0.94, "learning_rate": 1.822105305416777e-07, "loss": 0.983, "step": 11111 }, { "epoch": 0.94, "learning_rate": 1.8168965124161441e-07, "loss": 0.9468, "step": 11112 }, { "epoch": 0.94, "learning_rate": 1.8116951069237389e-07, "loss": 0.8895, "step": 11113 }, { "epoch": 0.94, "learning_rate": 1.8065010893309143e-07, "loss": 0.9784, "step": 11114 }, { "epoch": 0.94, "learning_rate": 1.8013144600284803e-07, "loss": 0.9291, "step": 11115 }, { "epoch": 0.94, "learning_rate": 1.7961352194066806e-07, "loss": 0.9921, "step": 11116 }, { "epoch": 0.94, "learning_rate": 1.7909633678552362e-07, "loss": 0.9107, "step": 11117 }, { "epoch": 0.94, "learning_rate": 1.7857989057632585e-07, "loss": 1.0019, "step": 11118 }, { "epoch": 0.94, "learning_rate": 1.7806418335193698e-07, "loss": 0.9664, "step": 11119 }, { "epoch": 0.94, "learning_rate": 1.7754921515115487e-07, "loss": 0.9866, "step": 11120 }, { "epoch": 0.94, "learning_rate": 1.7703498601273183e-07, "loss": 0.9482, "step": 11121 }, { "epoch": 0.94, "learning_rate": 1.7652149597535695e-07, "loss": 0.9851, "step": 11122 }, { "epoch": 0.94, "learning_rate": 1.7600874507766707e-07, "loss": 0.9517, "step": 11123 }, { "epoch": 0.94, "learning_rate": 1.7549673335824135e-07, "loss": 0.9608, "step": 11124 }, { "epoch": 0.94, "learning_rate": 1.7498546085560675e-07, "loss": 0.9408, "step": 11125 }, { "epoch": 0.94, "learning_rate": 1.7447492760823136e-07, "loss": 0.9368, "step": 11126 }, { "epoch": 0.94, "learning_rate": 1.7396513365452895e-07, "loss": 0.9379, "step": 11127 }, { "epoch": 0.94, "learning_rate": 1.7345607903285766e-07, "loss": 0.9503, "step": 11128 }, { "epoch": 0.94, "learning_rate": 1.729477637815191e-07, "loss": 0.9467, "step": 11129 }, { "epoch": 0.94, "learning_rate": 1.724401879387605e-07, "loss": 0.899, "step": 11130 }, { "epoch": 0.94, "learning_rate": 1.7193335154277457e-07, "loss": 0.8966, "step": 11131 }, { "epoch": 0.94, "learning_rate": 1.7142725463169307e-07, "loss": 0.9567, "step": 11132 }, { "epoch": 0.94, "learning_rate": 1.7092189724359886e-07, "loss": 1.0278, "step": 11133 }, { "epoch": 0.94, "learning_rate": 1.7041727941651597e-07, "loss": 0.943, "step": 11134 }, { "epoch": 0.94, "learning_rate": 1.6991340118841183e-07, "loss": 0.9899, "step": 11135 }, { "epoch": 0.94, "learning_rate": 1.694102625971994e-07, "loss": 0.9868, "step": 11136 }, { "epoch": 0.94, "learning_rate": 1.6890786368073619e-07, "loss": 0.9507, "step": 11137 }, { "epoch": 0.94, "learning_rate": 1.6840620447682532e-07, "loss": 0.9607, "step": 11138 }, { "epoch": 0.94, "learning_rate": 1.679052850232088e-07, "loss": 0.968, "step": 11139 }, { "epoch": 0.94, "learning_rate": 1.6740510535758093e-07, "loss": 0.9729, "step": 11140 }, { "epoch": 0.94, "learning_rate": 1.6690566551757382e-07, "loss": 0.9507, "step": 11141 }, { "epoch": 0.94, "learning_rate": 1.664069655407685e-07, "loss": 1.0079, "step": 11142 }, { "epoch": 0.94, "learning_rate": 1.6590900546468614e-07, "loss": 0.9659, "step": 11143 }, { "epoch": 0.94, "learning_rate": 1.654117853267956e-07, "loss": 0.9627, "step": 11144 }, { "epoch": 0.94, "learning_rate": 1.6491530516450805e-07, "loss": 0.9348, "step": 11145 }, { "epoch": 0.94, "learning_rate": 1.644195650151803e-07, "loss": 0.9837, "step": 11146 }, { "epoch": 0.94, "learning_rate": 1.639245649161114e-07, "loss": 0.9491, "step": 11147 }, { "epoch": 0.94, "learning_rate": 1.6343030490454936e-07, "loss": 0.9493, "step": 11148 }, { "epoch": 0.94, "learning_rate": 1.6293678501767995e-07, "loss": 1.0121, "step": 11149 }, { "epoch": 0.94, "learning_rate": 1.6244400529264016e-07, "loss": 0.9562, "step": 11150 }, { "epoch": 0.94, "learning_rate": 1.6195196576650362e-07, "loss": 0.9936, "step": 11151 }, { "epoch": 0.94, "learning_rate": 1.614606664762963e-07, "loss": 0.9019, "step": 11152 }, { "epoch": 0.94, "learning_rate": 1.6097010745898202e-07, "loss": 0.9678, "step": 11153 }, { "epoch": 0.94, "learning_rate": 1.6048028875147337e-07, "loss": 0.9223, "step": 11154 }, { "epoch": 0.94, "learning_rate": 1.5999121039062427e-07, "loss": 0.9242, "step": 11155 }, { "epoch": 0.94, "learning_rate": 1.5950287241323302e-07, "loss": 1.0366, "step": 11156 }, { "epoch": 0.94, "learning_rate": 1.5901527485604584e-07, "loss": 0.9734, "step": 11157 }, { "epoch": 0.94, "learning_rate": 1.585284177557489e-07, "loss": 0.9652, "step": 11158 }, { "epoch": 0.95, "learning_rate": 1.58042301148974e-07, "loss": 0.7785, "step": 11159 }, { "epoch": 0.95, "learning_rate": 1.5755692507229858e-07, "loss": 0.9993, "step": 11160 }, { "epoch": 0.95, "learning_rate": 1.570722895622434e-07, "loss": 0.9686, "step": 11161 }, { "epoch": 0.95, "learning_rate": 1.5658839465527265e-07, "loss": 0.973, "step": 11162 }, { "epoch": 0.95, "learning_rate": 1.5610524038779718e-07, "loss": 0.9146, "step": 11163 }, { "epoch": 0.95, "learning_rate": 1.5562282679616792e-07, "loss": 0.9233, "step": 11164 }, { "epoch": 0.95, "learning_rate": 1.5514115391668581e-07, "loss": 0.9718, "step": 11165 }, { "epoch": 0.95, "learning_rate": 1.546602217855908e-07, "loss": 0.975, "step": 11166 }, { "epoch": 0.95, "learning_rate": 1.541800304390706e-07, "loss": 0.9769, "step": 11167 }, { "epoch": 0.95, "learning_rate": 1.5370057991325405e-07, "loss": 0.9283, "step": 11168 }, { "epoch": 0.95, "learning_rate": 1.5322187024421786e-07, "loss": 0.9347, "step": 11169 }, { "epoch": 0.95, "learning_rate": 1.527439014679799e-07, "loss": 0.9657, "step": 11170 }, { "epoch": 0.95, "learning_rate": 1.5226667362050473e-07, "loss": 0.9626, "step": 11171 }, { "epoch": 0.95, "learning_rate": 1.5179018673769806e-07, "loss": 0.9671, "step": 11172 }, { "epoch": 0.95, "learning_rate": 1.5131444085541459e-07, "loss": 0.997, "step": 11173 }, { "epoch": 0.95, "learning_rate": 1.5083943600944896e-07, "loss": 0.9856, "step": 11174 }, { "epoch": 0.95, "learning_rate": 1.503651722355415e-07, "loss": 1.0148, "step": 11175 }, { "epoch": 0.95, "learning_rate": 1.4989164956937587e-07, "loss": 0.9285, "step": 11176 }, { "epoch": 0.95, "learning_rate": 1.4941886804658358e-07, "loss": 0.9392, "step": 11177 }, { "epoch": 0.95, "learning_rate": 1.4894682770273506e-07, "loss": 0.9697, "step": 11178 }, { "epoch": 0.95, "learning_rate": 1.484755285733497e-07, "loss": 0.8904, "step": 11179 }, { "epoch": 0.95, "learning_rate": 1.4800497069388685e-07, "loss": 0.9632, "step": 11180 }, { "epoch": 0.95, "learning_rate": 1.4753515409975494e-07, "loss": 0.9932, "step": 11181 }, { "epoch": 0.95, "learning_rate": 1.470660788263012e-07, "loss": 0.9751, "step": 11182 }, { "epoch": 0.95, "learning_rate": 1.4659774490882183e-07, "loss": 0.9023, "step": 11183 }, { "epoch": 0.95, "learning_rate": 1.4613015238255536e-07, "loss": 0.9799, "step": 11184 }, { "epoch": 0.95, "learning_rate": 1.4566330128268245e-07, "loss": 0.9638, "step": 11185 }, { "epoch": 0.95, "learning_rate": 1.4519719164433289e-07, "loss": 0.964, "step": 11186 }, { "epoch": 0.95, "learning_rate": 1.447318235025752e-07, "loss": 0.9388, "step": 11187 }, { "epoch": 0.95, "learning_rate": 1.4426719689242587e-07, "loss": 1.0253, "step": 11188 }, { "epoch": 0.95, "learning_rate": 1.4380331184884467e-07, "loss": 0.9598, "step": 11189 }, { "epoch": 0.95, "learning_rate": 1.433401684067348e-07, "loss": 0.971, "step": 11190 }, { "epoch": 0.95, "learning_rate": 1.428777666009429e-07, "loss": 0.9167, "step": 11191 }, { "epoch": 0.95, "learning_rate": 1.4241610646626436e-07, "loss": 0.92, "step": 11192 }, { "epoch": 0.95, "learning_rate": 1.4195518803743258e-07, "loss": 0.9288, "step": 11193 }, { "epoch": 0.95, "learning_rate": 1.414950113491287e-07, "loss": 0.9502, "step": 11194 }, { "epoch": 0.95, "learning_rate": 1.410355764359783e-07, "loss": 0.9779, "step": 11195 }, { "epoch": 0.95, "learning_rate": 1.4057688333255048e-07, "loss": 0.9718, "step": 11196 }, { "epoch": 0.95, "learning_rate": 1.4011893207335536e-07, "loss": 0.9439, "step": 11197 }, { "epoch": 0.95, "learning_rate": 1.3966172269285428e-07, "loss": 0.9738, "step": 11198 }, { "epoch": 0.95, "learning_rate": 1.3920525522544526e-07, "loss": 0.963, "step": 11199 }, { "epoch": 0.95, "learning_rate": 1.3874952970547638e-07, "loss": 0.9573, "step": 11200 }, { "epoch": 0.95, "learning_rate": 1.382945461672358e-07, "loss": 0.9593, "step": 11201 }, { "epoch": 0.95, "learning_rate": 1.3784030464495946e-07, "loss": 0.9732, "step": 11202 }, { "epoch": 0.95, "learning_rate": 1.3738680517282222e-07, "loss": 0.9726, "step": 11203 }, { "epoch": 0.95, "learning_rate": 1.3693404778494902e-07, "loss": 0.9709, "step": 11204 }, { "epoch": 0.95, "learning_rate": 1.364820325154048e-07, "loss": 0.9676, "step": 11205 }, { "epoch": 0.95, "learning_rate": 1.3603075939820131e-07, "loss": 0.966, "step": 11206 }, { "epoch": 0.95, "learning_rate": 1.3558022846729358e-07, "loss": 0.986, "step": 11207 }, { "epoch": 0.95, "learning_rate": 1.351304397565789e-07, "loss": 0.9914, "step": 11208 }, { "epoch": 0.95, "learning_rate": 1.3468139329990137e-07, "loss": 0.9662, "step": 11209 }, { "epoch": 0.95, "learning_rate": 1.3423308913104726e-07, "loss": 0.9442, "step": 11210 }, { "epoch": 0.95, "learning_rate": 1.3378552728375073e-07, "loss": 0.9294, "step": 11211 }, { "epoch": 0.95, "learning_rate": 1.3333870779168368e-07, "loss": 0.8955, "step": 11212 }, { "epoch": 0.95, "learning_rate": 1.3289263068846815e-07, "loss": 0.9484, "step": 11213 }, { "epoch": 0.95, "learning_rate": 1.3244729600766616e-07, "loss": 0.9642, "step": 11214 }, { "epoch": 0.95, "learning_rate": 1.3200270378278756e-07, "loss": 0.9673, "step": 11215 }, { "epoch": 0.95, "learning_rate": 1.315588540472834e-07, "loss": 0.9943, "step": 11216 }, { "epoch": 0.95, "learning_rate": 1.311157468345503e-07, "loss": 0.9492, "step": 11217 }, { "epoch": 0.95, "learning_rate": 1.3067338217792714e-07, "loss": 0.9319, "step": 11218 }, { "epoch": 0.95, "learning_rate": 1.302317601107006e-07, "loss": 0.9576, "step": 11219 }, { "epoch": 0.95, "learning_rate": 1.2979088066609746e-07, "loss": 0.9268, "step": 11220 }, { "epoch": 0.95, "learning_rate": 1.293507438772923e-07, "loss": 0.9958, "step": 11221 }, { "epoch": 0.95, "learning_rate": 1.2891134977740083e-07, "loss": 0.878, "step": 11222 }, { "epoch": 0.95, "learning_rate": 1.284726983994833e-07, "loss": 0.9878, "step": 11223 }, { "epoch": 0.95, "learning_rate": 1.280347897765466e-07, "loss": 0.9676, "step": 11224 }, { "epoch": 0.95, "learning_rate": 1.2759762394153775e-07, "loss": 1.0054, "step": 11225 }, { "epoch": 0.95, "learning_rate": 1.2716120092735152e-07, "loss": 0.974, "step": 11226 }, { "epoch": 0.95, "learning_rate": 1.267255207668261e-07, "loss": 0.9749, "step": 11227 }, { "epoch": 0.95, "learning_rate": 1.262905834927397e-07, "loss": 0.9685, "step": 11228 }, { "epoch": 0.95, "learning_rate": 1.2585638913782173e-07, "loss": 0.9347, "step": 11229 }, { "epoch": 0.95, "learning_rate": 1.2542293773474045e-07, "loss": 0.916, "step": 11230 }, { "epoch": 0.95, "learning_rate": 1.2499022931610872e-07, "loss": 0.9373, "step": 11231 }, { "epoch": 0.95, "learning_rate": 1.2455826391448488e-07, "loss": 0.9803, "step": 11232 }, { "epoch": 0.95, "learning_rate": 1.24127041562373e-07, "loss": 0.9979, "step": 11233 }, { "epoch": 0.95, "learning_rate": 1.2369656229221705e-07, "loss": 1.0229, "step": 11234 }, { "epoch": 0.95, "learning_rate": 1.2326682613640673e-07, "loss": 0.9882, "step": 11235 }, { "epoch": 0.95, "learning_rate": 1.2283783312727727e-07, "loss": 0.9536, "step": 11236 }, { "epoch": 0.95, "learning_rate": 1.2240958329710728e-07, "loss": 0.9482, "step": 11237 }, { "epoch": 0.95, "learning_rate": 1.219820766781199e-07, "loss": 0.9622, "step": 11238 }, { "epoch": 0.95, "learning_rate": 1.2155531330247939e-07, "loss": 1.0172, "step": 11239 }, { "epoch": 0.95, "learning_rate": 1.2112929320229783e-07, "loss": 0.9215, "step": 11240 }, { "epoch": 0.95, "learning_rate": 1.2070401640962958e-07, "loss": 0.9509, "step": 11241 }, { "epoch": 0.95, "learning_rate": 1.2027948295647462e-07, "loss": 0.9533, "step": 11242 }, { "epoch": 0.95, "learning_rate": 1.1985569287477294e-07, "loss": 0.9781, "step": 11243 }, { "epoch": 0.95, "learning_rate": 1.194326461964146e-07, "loss": 0.7593, "step": 11244 }, { "epoch": 0.95, "learning_rate": 1.1901034295322855e-07, "loss": 0.9485, "step": 11245 }, { "epoch": 0.95, "learning_rate": 1.1858878317699052e-07, "loss": 1.0092, "step": 11246 }, { "epoch": 0.95, "learning_rate": 1.1816796689941845e-07, "loss": 0.9099, "step": 11247 }, { "epoch": 0.95, "learning_rate": 1.1774789415217703e-07, "loss": 0.9718, "step": 11248 }, { "epoch": 0.95, "learning_rate": 1.1732856496687207e-07, "loss": 0.9101, "step": 11249 }, { "epoch": 0.95, "learning_rate": 1.1690997937505721e-07, "loss": 0.9203, "step": 11250 }, { "epoch": 0.95, "learning_rate": 1.1649213740822396e-07, "loss": 0.9538, "step": 11251 }, { "epoch": 0.95, "learning_rate": 1.1607503909781492e-07, "loss": 0.9379, "step": 11252 }, { "epoch": 0.95, "learning_rate": 1.1565868447521167e-07, "loss": 0.9263, "step": 11253 }, { "epoch": 0.95, "learning_rate": 1.1524307357174358e-07, "loss": 0.9326, "step": 11254 }, { "epoch": 0.95, "learning_rate": 1.1482820641867898e-07, "loss": 0.9802, "step": 11255 }, { "epoch": 0.95, "learning_rate": 1.1441408304723623e-07, "loss": 0.963, "step": 11256 }, { "epoch": 0.95, "learning_rate": 1.1400070348857372e-07, "loss": 0.9542, "step": 11257 }, { "epoch": 0.95, "learning_rate": 1.1358806777379549e-07, "loss": 0.9366, "step": 11258 }, { "epoch": 0.95, "learning_rate": 1.1317617593394781e-07, "loss": 0.9298, "step": 11259 }, { "epoch": 0.95, "learning_rate": 1.1276502800002365e-07, "loss": 1.0098, "step": 11260 }, { "epoch": 0.95, "learning_rate": 1.1235462400295826e-07, "loss": 0.9403, "step": 11261 }, { "epoch": 0.95, "learning_rate": 1.1194496397363142e-07, "loss": 0.9587, "step": 11262 }, { "epoch": 0.95, "learning_rate": 1.1153604794286621e-07, "loss": 0.9191, "step": 11263 }, { "epoch": 0.95, "learning_rate": 1.1112787594143138e-07, "loss": 0.9402, "step": 11264 }, { "epoch": 0.95, "learning_rate": 1.1072044800003789e-07, "loss": 0.9577, "step": 11265 }, { "epoch": 0.95, "learning_rate": 1.1031376414934125e-07, "loss": 0.9398, "step": 11266 }, { "epoch": 0.95, "learning_rate": 1.0990782441994253e-07, "loss": 1.0071, "step": 11267 }, { "epoch": 0.95, "learning_rate": 1.0950262884238394e-07, "loss": 0.9491, "step": 11268 }, { "epoch": 0.95, "learning_rate": 1.0909817744715556e-07, "loss": 0.952, "step": 11269 }, { "epoch": 0.95, "learning_rate": 1.0869447026468638e-07, "loss": 0.9506, "step": 11270 }, { "epoch": 0.95, "learning_rate": 1.082915073253532e-07, "loss": 0.9839, "step": 11271 }, { "epoch": 0.95, "learning_rate": 1.0788928865947624e-07, "loss": 0.9543, "step": 11272 }, { "epoch": 0.95, "learning_rate": 1.0748781429732013e-07, "loss": 0.9587, "step": 11273 }, { "epoch": 0.95, "learning_rate": 1.0708708426909076e-07, "loss": 0.8143, "step": 11274 }, { "epoch": 0.95, "learning_rate": 1.0668709860494175e-07, "loss": 0.9872, "step": 11275 }, { "epoch": 0.95, "learning_rate": 1.0628785733496794e-07, "loss": 0.9592, "step": 11276 }, { "epoch": 0.96, "learning_rate": 1.0588936048920862e-07, "loss": 0.9923, "step": 11277 }, { "epoch": 0.96, "learning_rate": 1.0549160809764758e-07, "loss": 0.9086, "step": 11278 }, { "epoch": 0.96, "learning_rate": 1.0509460019021423e-07, "loss": 0.8794, "step": 11279 }, { "epoch": 0.96, "learning_rate": 1.0469833679677798e-07, "loss": 1.0305, "step": 11280 }, { "epoch": 0.96, "learning_rate": 1.0430281794715725e-07, "loss": 0.9586, "step": 11281 }, { "epoch": 0.96, "learning_rate": 1.0390804367110818e-07, "loss": 0.9412, "step": 11282 }, { "epoch": 0.96, "learning_rate": 1.0351401399833816e-07, "loss": 0.8541, "step": 11283 }, { "epoch": 0.96, "learning_rate": 1.0312072895849234e-07, "loss": 0.9776, "step": 11284 }, { "epoch": 0.96, "learning_rate": 1.0272818858116373e-07, "loss": 1.0124, "step": 11285 }, { "epoch": 0.96, "learning_rate": 1.0233639289588649e-07, "loss": 0.9715, "step": 11286 }, { "epoch": 0.96, "learning_rate": 1.0194534193214145e-07, "loss": 0.9157, "step": 11287 }, { "epoch": 0.96, "learning_rate": 1.0155503571935176e-07, "loss": 0.968, "step": 11288 }, { "epoch": 0.96, "learning_rate": 1.0116547428688395e-07, "loss": 0.9586, "step": 11289 }, { "epoch": 0.96, "learning_rate": 1.007766576640512e-07, "loss": 0.9029, "step": 11290 }, { "epoch": 0.96, "learning_rate": 1.0038858588010792e-07, "loss": 1.0119, "step": 11291 }, { "epoch": 0.96, "learning_rate": 1.0000125896425405e-07, "loss": 0.995, "step": 11292 }, { "epoch": 0.96, "learning_rate": 9.961467694563076e-08, "loss": 0.9746, "step": 11293 }, { "epoch": 0.96, "learning_rate": 9.92288398533281e-08, "loss": 0.9288, "step": 11294 }, { "epoch": 0.96, "learning_rate": 9.884374771637617e-08, "loss": 0.9247, "step": 11295 }, { "epoch": 0.96, "learning_rate": 9.845940056374958e-08, "loss": 0.96, "step": 11296 }, { "epoch": 0.96, "learning_rate": 9.807579842436854e-08, "loss": 0.9502, "step": 11297 }, { "epoch": 0.96, "learning_rate": 9.769294132709551e-08, "loss": 0.9596, "step": 11298 }, { "epoch": 0.96, "learning_rate": 9.731082930073743e-08, "loss": 0.9357, "step": 11299 }, { "epoch": 0.96, "learning_rate": 9.692946237404466e-08, "loss": 0.9019, "step": 11300 }, { "epoch": 0.96, "learning_rate": 9.654884057571312e-08, "loss": 0.9444, "step": 11301 }, { "epoch": 0.96, "learning_rate": 9.616896393438213e-08, "loss": 0.9778, "step": 11302 }, { "epoch": 0.96, "learning_rate": 9.578983247863217e-08, "loss": 0.9229, "step": 11303 }, { "epoch": 0.96, "learning_rate": 9.54114462369915e-08, "loss": 1.0023, "step": 11304 }, { "epoch": 0.96, "learning_rate": 9.50338052379307e-08, "loss": 0.9953, "step": 11305 }, { "epoch": 0.96, "learning_rate": 9.465690950986484e-08, "loss": 0.9625, "step": 11306 }, { "epoch": 0.96, "learning_rate": 9.428075908115009e-08, "loss": 0.8058, "step": 11307 }, { "epoch": 0.96, "learning_rate": 9.39053539800916e-08, "loss": 0.9096, "step": 11308 }, { "epoch": 0.96, "learning_rate": 9.353069423493344e-08, "loss": 0.9637, "step": 11309 }, { "epoch": 0.96, "learning_rate": 9.31567798738675e-08, "loss": 0.9689, "step": 11310 }, { "epoch": 0.96, "learning_rate": 9.278361092502575e-08, "loss": 0.9325, "step": 11311 }, { "epoch": 0.96, "learning_rate": 9.241118741648902e-08, "loss": 0.9881, "step": 11312 }, { "epoch": 0.96, "learning_rate": 9.203950937627714e-08, "loss": 0.9766, "step": 11313 }, { "epoch": 0.96, "learning_rate": 9.166857683235775e-08, "loss": 0.9662, "step": 11314 }, { "epoch": 0.96, "learning_rate": 9.12983898126385e-08, "loss": 0.9687, "step": 11315 }, { "epoch": 0.96, "learning_rate": 9.09289483449749e-08, "loss": 0.8776, "step": 11316 }, { "epoch": 0.96, "learning_rate": 9.05602524571636e-08, "loss": 0.9054, "step": 11317 }, { "epoch": 0.96, "learning_rate": 9.019230217694463e-08, "loss": 0.9216, "step": 11318 }, { "epoch": 0.96, "learning_rate": 8.982509753200696e-08, "loss": 0.9263, "step": 11319 }, { "epoch": 0.96, "learning_rate": 8.945863854997627e-08, "loss": 0.9261, "step": 11320 }, { "epoch": 0.96, "learning_rate": 8.909292525842828e-08, "loss": 0.9198, "step": 11321 }, { "epoch": 0.96, "learning_rate": 8.872795768487873e-08, "loss": 0.9846, "step": 11322 }, { "epoch": 0.96, "learning_rate": 8.836373585678903e-08, "loss": 0.9121, "step": 11323 }, { "epoch": 0.96, "learning_rate": 8.800025980156391e-08, "loss": 0.9532, "step": 11324 }, { "epoch": 0.96, "learning_rate": 8.763752954655147e-08, "loss": 0.9092, "step": 11325 }, { "epoch": 0.96, "learning_rate": 8.727554511904546e-08, "loss": 0.9876, "step": 11326 }, { "epoch": 0.96, "learning_rate": 8.691430654628186e-08, "loss": 1.0104, "step": 11327 }, { "epoch": 0.96, "learning_rate": 8.655381385544114e-08, "loss": 0.9658, "step": 11328 }, { "epoch": 0.96, "learning_rate": 8.619406707364719e-08, "loss": 0.8414, "step": 11329 }, { "epoch": 0.96, "learning_rate": 8.583506622796722e-08, "loss": 0.9989, "step": 11330 }, { "epoch": 0.96, "learning_rate": 8.54768113454163e-08, "loss": 0.9568, "step": 11331 }, { "epoch": 0.96, "learning_rate": 8.51193024529473e-08, "loss": 0.9557, "step": 11332 }, { "epoch": 0.96, "learning_rate": 8.476253957746095e-08, "loss": 0.8873, "step": 11333 }, { "epoch": 0.96, "learning_rate": 8.440652274580018e-08, "loss": 0.9488, "step": 11334 }, { "epoch": 0.96, "learning_rate": 8.405125198475472e-08, "loss": 0.9555, "step": 11335 }, { "epoch": 0.96, "learning_rate": 8.369672732105316e-08, "loss": 0.9697, "step": 11336 }, { "epoch": 0.96, "learning_rate": 8.334294878137306e-08, "loss": 0.9634, "step": 11337 }, { "epoch": 0.96, "learning_rate": 8.29899163923309e-08, "loss": 0.9598, "step": 11338 }, { "epoch": 0.96, "learning_rate": 8.26376301804932e-08, "loss": 1.0105, "step": 11339 }, { "epoch": 0.96, "learning_rate": 8.228609017236211e-08, "loss": 0.9751, "step": 11340 }, { "epoch": 0.96, "learning_rate": 8.193529639439201e-08, "loss": 0.9567, "step": 11341 }, { "epoch": 0.96, "learning_rate": 8.158524887297625e-08, "loss": 0.9625, "step": 11342 }, { "epoch": 0.96, "learning_rate": 8.123594763445265e-08, "loss": 0.964, "step": 11343 }, { "epoch": 0.96, "learning_rate": 8.088739270510459e-08, "loss": 0.9072, "step": 11344 }, { "epoch": 0.96, "learning_rate": 8.053958411115558e-08, "loss": 0.9564, "step": 11345 }, { "epoch": 0.96, "learning_rate": 8.019252187877801e-08, "loss": 1.01, "step": 11346 }, { "epoch": 0.96, "learning_rate": 7.984620603408433e-08, "loss": 0.9492, "step": 11347 }, { "epoch": 0.96, "learning_rate": 7.950063660313367e-08, "loss": 0.9904, "step": 11348 }, { "epoch": 0.96, "learning_rate": 7.915581361192526e-08, "loss": 0.9452, "step": 11349 }, { "epoch": 0.96, "learning_rate": 7.881173708640499e-08, "loss": 1.0017, "step": 11350 }, { "epoch": 0.96, "learning_rate": 7.846840705246217e-08, "loss": 0.9613, "step": 11351 }, { "epoch": 0.96, "learning_rate": 7.812582353593057e-08, "loss": 0.9377, "step": 11352 }, { "epoch": 0.96, "learning_rate": 7.778398656258624e-08, "loss": 0.9791, "step": 11353 }, { "epoch": 0.96, "learning_rate": 7.744289615814859e-08, "loss": 0.9432, "step": 11354 }, { "epoch": 0.96, "learning_rate": 7.710255234828379e-08, "loss": 0.9633, "step": 11355 }, { "epoch": 0.96, "learning_rate": 7.67629551585991e-08, "loss": 0.9452, "step": 11356 }, { "epoch": 0.96, "learning_rate": 7.642410461464634e-08, "loss": 0.9471, "step": 11357 }, { "epoch": 0.96, "learning_rate": 7.608600074192174e-08, "loss": 0.9366, "step": 11358 }, { "epoch": 0.96, "learning_rate": 7.574864356586498e-08, "loss": 0.9381, "step": 11359 }, { "epoch": 0.96, "learning_rate": 7.541203311185908e-08, "loss": 0.9664, "step": 11360 }, { "epoch": 0.96, "learning_rate": 7.507616940523265e-08, "loss": 0.9995, "step": 11361 }, { "epoch": 0.96, "learning_rate": 7.474105247125551e-08, "loss": 0.9818, "step": 11362 }, { "epoch": 0.96, "learning_rate": 7.440668233514304e-08, "loss": 0.9436, "step": 11363 }, { "epoch": 0.96, "learning_rate": 7.407305902205398e-08, "loss": 1.0132, "step": 11364 }, { "epoch": 0.96, "learning_rate": 7.374018255708937e-08, "loss": 0.9446, "step": 11365 }, { "epoch": 0.96, "learning_rate": 7.340805296529918e-08, "loss": 0.9524, "step": 11366 }, { "epoch": 0.96, "learning_rate": 7.307667027166898e-08, "loss": 0.9997, "step": 11367 }, { "epoch": 0.96, "learning_rate": 7.27460345011366e-08, "loss": 0.9234, "step": 11368 }, { "epoch": 0.96, "learning_rate": 7.24161456785788e-08, "loss": 0.9149, "step": 11369 }, { "epoch": 0.96, "learning_rate": 7.208700382881573e-08, "loss": 0.9396, "step": 11370 }, { "epoch": 0.96, "learning_rate": 7.175860897661313e-08, "loss": 0.9391, "step": 11371 }, { "epoch": 0.96, "learning_rate": 7.143096114668124e-08, "loss": 0.9648, "step": 11372 }, { "epoch": 0.96, "learning_rate": 7.110406036367257e-08, "loss": 0.971, "step": 11373 }, { "epoch": 0.96, "learning_rate": 7.077790665218299e-08, "loss": 0.9403, "step": 11374 }, { "epoch": 0.96, "learning_rate": 7.045250003675397e-08, "loss": 0.9617, "step": 11375 }, { "epoch": 0.96, "learning_rate": 7.012784054186927e-08, "loss": 0.965, "step": 11376 }, { "epoch": 0.96, "learning_rate": 6.980392819195713e-08, "loss": 0.9376, "step": 11377 }, { "epoch": 0.96, "learning_rate": 6.948076301138918e-08, "loss": 0.931, "step": 11378 }, { "epoch": 0.96, "learning_rate": 6.91583450244826e-08, "loss": 0.9773, "step": 11379 }, { "epoch": 0.96, "learning_rate": 6.883667425549356e-08, "loss": 0.9799, "step": 11380 }, { "epoch": 0.96, "learning_rate": 6.851575072862937e-08, "loss": 0.9587, "step": 11381 }, { "epoch": 0.96, "learning_rate": 6.819557446803293e-08, "loss": 0.993, "step": 11382 }, { "epoch": 0.96, "learning_rate": 6.787614549779831e-08, "loss": 0.9813, "step": 11383 }, { "epoch": 0.96, "learning_rate": 6.755746384195738e-08, "loss": 1.0131, "step": 11384 }, { "epoch": 0.96, "learning_rate": 6.723952952449098e-08, "loss": 0.995, "step": 11385 }, { "epoch": 0.96, "learning_rate": 6.692234256931995e-08, "loss": 0.9901, "step": 11386 }, { "epoch": 0.96, "learning_rate": 6.660590300030967e-08, "loss": 1.0319, "step": 11387 }, { "epoch": 0.96, "learning_rate": 6.629021084126996e-08, "loss": 0.9978, "step": 11388 }, { "epoch": 0.96, "learning_rate": 6.597526611595406e-08, "loss": 0.9978, "step": 11389 }, { "epoch": 0.96, "learning_rate": 6.566106884805967e-08, "loss": 0.9336, "step": 11390 }, { "epoch": 0.96, "learning_rate": 6.534761906122789e-08, "loss": 0.9989, "step": 11391 }, { "epoch": 0.96, "learning_rate": 6.503491677904206e-08, "loss": 0.9364, "step": 11392 }, { "epoch": 0.96, "learning_rate": 6.472296202503226e-08, "loss": 0.9587, "step": 11393 }, { "epoch": 0.96, "learning_rate": 6.441175482266859e-08, "loss": 0.9262, "step": 11394 }, { "epoch": 0.97, "learning_rate": 6.410129519536901e-08, "loss": 1.022, "step": 11395 }, { "epoch": 0.97, "learning_rate": 6.379158316649259e-08, "loss": 0.9717, "step": 11396 }, { "epoch": 0.97, "learning_rate": 6.348261875934181e-08, "loss": 0.8248, "step": 11397 }, { "epoch": 0.97, "learning_rate": 6.317440199716474e-08, "loss": 0.9856, "step": 11398 }, { "epoch": 0.97, "learning_rate": 6.286693290315171e-08, "loss": 0.8988, "step": 11399 }, { "epoch": 0.97, "learning_rate": 6.256021150043756e-08, "loss": 0.9028, "step": 11400 }, { "epoch": 0.97, "learning_rate": 6.22542378121005e-08, "loss": 0.9964, "step": 11401 }, { "epoch": 0.97, "learning_rate": 6.19490118611632e-08, "loss": 0.9055, "step": 11402 }, { "epoch": 0.97, "learning_rate": 6.164453367059064e-08, "loss": 0.923, "step": 11403 }, { "epoch": 0.97, "learning_rate": 6.134080326329451e-08, "loss": 0.9619, "step": 11404 }, { "epoch": 0.97, "learning_rate": 6.103782066212427e-08, "loss": 0.9392, "step": 11405 }, { "epoch": 0.97, "learning_rate": 6.073558588988171e-08, "loss": 0.9477, "step": 11406 }, { "epoch": 0.97, "learning_rate": 6.043409896930307e-08, "loss": 0.951, "step": 11407 }, { "epoch": 0.97, "learning_rate": 6.01333599230769e-08, "loss": 0.9464, "step": 11408 }, { "epoch": 0.97, "learning_rate": 5.983336877382839e-08, "loss": 0.9691, "step": 11409 }, { "epoch": 0.97, "learning_rate": 5.953412554413063e-08, "loss": 0.9765, "step": 11410 }, { "epoch": 0.97, "learning_rate": 5.923563025650003e-08, "loss": 0.9611, "step": 11411 }, { "epoch": 0.97, "learning_rate": 5.8937882933395305e-08, "loss": 0.9364, "step": 11412 }, { "epoch": 0.97, "learning_rate": 5.8640883597218534e-08, "loss": 0.9385, "step": 11413 }, { "epoch": 0.97, "learning_rate": 5.834463227031961e-08, "loss": 0.9562, "step": 11414 }, { "epoch": 0.97, "learning_rate": 5.8049128974985156e-08, "loss": 0.9384, "step": 11415 }, { "epoch": 0.97, "learning_rate": 5.775437373345294e-08, "loss": 0.9584, "step": 11416 }, { "epoch": 0.97, "learning_rate": 5.746036656789966e-08, "loss": 0.9786, "step": 11417 }, { "epoch": 0.97, "learning_rate": 5.7167107500446517e-08, "loss": 0.9983, "step": 11418 }, { "epoch": 0.97, "learning_rate": 5.68745965531603e-08, "loss": 0.9189, "step": 11419 }, { "epoch": 0.97, "learning_rate": 5.658283374804785e-08, "loss": 0.9444, "step": 11420 }, { "epoch": 0.97, "learning_rate": 5.6291819107064935e-08, "loss": 0.9418, "step": 11421 }, { "epoch": 0.97, "learning_rate": 5.600155265210627e-08, "loss": 0.85, "step": 11422 }, { "epoch": 0.97, "learning_rate": 5.571203440501216e-08, "loss": 0.8387, "step": 11423 }, { "epoch": 0.97, "learning_rate": 5.542326438756629e-08, "loss": 0.9506, "step": 11424 }, { "epoch": 0.97, "learning_rate": 5.513524262149794e-08, "loss": 0.9342, "step": 11425 }, { "epoch": 0.97, "learning_rate": 5.484796912847756e-08, "loss": 0.9245, "step": 11426 }, { "epoch": 0.97, "learning_rate": 5.4561443930120084e-08, "loss": 0.9578, "step": 11427 }, { "epoch": 0.97, "learning_rate": 5.427566704798381e-08, "loss": 0.9653, "step": 11428 }, { "epoch": 0.97, "learning_rate": 5.399063850357267e-08, "loss": 0.9307, "step": 11429 }, { "epoch": 0.97, "learning_rate": 5.370635831833171e-08, "loss": 0.9704, "step": 11430 }, { "epoch": 0.97, "learning_rate": 5.3422826513651604e-08, "loss": 0.9685, "step": 11431 }, { "epoch": 0.97, "learning_rate": 5.314004311086418e-08, "loss": 0.9822, "step": 11432 }, { "epoch": 0.97, "learning_rate": 5.285800813124797e-08, "loss": 0.961, "step": 11433 }, { "epoch": 0.97, "learning_rate": 5.257672159602378e-08, "loss": 0.9466, "step": 11434 }, { "epoch": 0.97, "learning_rate": 5.2296183526356904e-08, "loss": 0.9305, "step": 11435 }, { "epoch": 0.97, "learning_rate": 5.201639394335489e-08, "loss": 0.9917, "step": 11436 }, { "epoch": 0.97, "learning_rate": 5.173735286806869e-08, "loss": 1.0035, "step": 11437 }, { "epoch": 0.97, "learning_rate": 5.1459060321495945e-08, "loss": 0.9612, "step": 11438 }, { "epoch": 0.97, "learning_rate": 5.1181516324574356e-08, "loss": 0.934, "step": 11439 }, { "epoch": 0.97, "learning_rate": 5.090472089818832e-08, "loss": 0.9523, "step": 11440 }, { "epoch": 0.97, "learning_rate": 5.062867406316341e-08, "loss": 0.964, "step": 11441 }, { "epoch": 0.97, "learning_rate": 5.035337584027078e-08, "loss": 0.9367, "step": 11442 }, { "epoch": 0.97, "learning_rate": 5.007882625022387e-08, "loss": 0.9915, "step": 11443 }, { "epoch": 0.97, "learning_rate": 4.980502531367948e-08, "loss": 0.9905, "step": 11444 }, { "epoch": 0.97, "learning_rate": 4.953197305124113e-08, "loss": 0.9804, "step": 11445 }, { "epoch": 0.97, "learning_rate": 4.92596694834524e-08, "loss": 0.9544, "step": 11446 }, { "epoch": 0.97, "learning_rate": 4.898811463080355e-08, "loss": 0.9942, "step": 11447 }, { "epoch": 0.97, "learning_rate": 4.871730851372492e-08, "loss": 0.9233, "step": 11448 }, { "epoch": 0.97, "learning_rate": 4.844725115259352e-08, "loss": 0.9692, "step": 11449 }, { "epoch": 0.97, "learning_rate": 4.8177942567728675e-08, "loss": 0.9632, "step": 11450 }, { "epoch": 0.97, "learning_rate": 4.790938277939416e-08, "loss": 1.0146, "step": 11451 }, { "epoch": 0.97, "learning_rate": 4.764157180779716e-08, "loss": 0.971, "step": 11452 }, { "epoch": 0.97, "learning_rate": 4.73745096730871e-08, "loss": 0.956, "step": 11453 }, { "epoch": 0.97, "learning_rate": 4.710819639536013e-08, "loss": 0.8977, "step": 11454 }, { "epoch": 0.97, "learning_rate": 4.6842631994653554e-08, "loss": 0.9944, "step": 11455 }, { "epoch": 0.97, "learning_rate": 4.6577816490949166e-08, "loss": 1.0196, "step": 11456 }, { "epoch": 0.97, "learning_rate": 4.6313749904171033e-08, "loss": 0.947, "step": 11457 }, { "epoch": 0.97, "learning_rate": 4.6050432254188815e-08, "loss": 0.9591, "step": 11458 }, { "epoch": 0.97, "learning_rate": 4.578786356081555e-08, "loss": 0.9556, "step": 11459 }, { "epoch": 0.97, "learning_rate": 4.5526043843807655e-08, "loss": 0.9784, "step": 11460 }, { "epoch": 0.97, "learning_rate": 4.526497312286382e-08, "loss": 0.9854, "step": 11461 }, { "epoch": 0.97, "learning_rate": 4.5004651417628334e-08, "loss": 0.9657, "step": 11462 }, { "epoch": 0.97, "learning_rate": 4.474507874768885e-08, "loss": 0.9189, "step": 11463 }, { "epoch": 0.97, "learning_rate": 4.448625513257532e-08, "loss": 0.9425, "step": 11464 }, { "epoch": 0.97, "learning_rate": 4.422818059176215e-08, "loss": 0.9505, "step": 11465 }, { "epoch": 0.97, "learning_rate": 4.397085514466826e-08, "loss": 0.8178, "step": 11466 }, { "epoch": 0.97, "learning_rate": 4.371427881065482e-08, "loss": 0.9646, "step": 11467 }, { "epoch": 0.97, "learning_rate": 4.345845160902751e-08, "loss": 1.0221, "step": 11468 }, { "epoch": 0.97, "learning_rate": 4.320337355903426e-08, "loss": 0.9691, "step": 11469 }, { "epoch": 0.97, "learning_rate": 4.2949044679868603e-08, "loss": 0.9369, "step": 11470 }, { "epoch": 0.97, "learning_rate": 4.2695464990667456e-08, "loss": 0.9156, "step": 11471 }, { "epoch": 0.97, "learning_rate": 4.244263451051001e-08, "loss": 0.9991, "step": 11472 }, { "epoch": 0.97, "learning_rate": 4.219055325841881e-08, "loss": 0.8454, "step": 11473 }, { "epoch": 0.97, "learning_rate": 4.193922125336203e-08, "loss": 0.9856, "step": 11474 }, { "epoch": 0.97, "learning_rate": 4.16886385142512e-08, "loss": 1.0131, "step": 11475 }, { "epoch": 0.97, "learning_rate": 4.143880505993902e-08, "loss": 0.9167, "step": 11476 }, { "epoch": 0.97, "learning_rate": 4.11897209092238e-08, "loss": 0.9518, "step": 11477 }, { "epoch": 0.97, "learning_rate": 4.09413860808483e-08, "loss": 0.9561, "step": 11478 }, { "epoch": 0.97, "learning_rate": 4.069380059349759e-08, "loss": 0.9683, "step": 11479 }, { "epoch": 0.97, "learning_rate": 4.044696446579899e-08, "loss": 0.9611, "step": 11480 }, { "epoch": 0.97, "learning_rate": 4.0200877716326526e-08, "loss": 0.9124, "step": 11481 }, { "epoch": 0.97, "learning_rate": 3.9955540363595394e-08, "loss": 0.9243, "step": 11482 }, { "epoch": 0.97, "learning_rate": 3.9710952426066376e-08, "loss": 0.9038, "step": 11483 }, { "epoch": 0.97, "learning_rate": 3.946711392214142e-08, "loss": 0.9124, "step": 11484 }, { "epoch": 0.97, "learning_rate": 3.9224024870169184e-08, "loss": 1.0045, "step": 11485 }, { "epoch": 0.97, "learning_rate": 3.8981685288438376e-08, "loss": 0.9295, "step": 11486 }, { "epoch": 0.97, "learning_rate": 3.8740095195184405e-08, "loss": 0.9977, "step": 11487 }, { "epoch": 0.97, "learning_rate": 3.8499254608583836e-08, "loss": 0.8654, "step": 11488 }, { "epoch": 0.97, "learning_rate": 3.825916354675996e-08, "loss": 0.8854, "step": 11489 }, { "epoch": 0.97, "learning_rate": 3.80198220277761e-08, "loss": 0.9601, "step": 11490 }, { "epoch": 0.97, "learning_rate": 3.778123006964118e-08, "loss": 1.0105, "step": 11491 }, { "epoch": 0.97, "learning_rate": 3.754338769030752e-08, "loss": 0.9448, "step": 11492 }, { "epoch": 0.97, "learning_rate": 3.730629490767079e-08, "loss": 0.9707, "step": 11493 }, { "epoch": 0.97, "learning_rate": 3.7069951739570064e-08, "loss": 0.9564, "step": 11494 }, { "epoch": 0.97, "learning_rate": 3.683435820379e-08, "loss": 0.9354, "step": 11495 }, { "epoch": 0.97, "learning_rate": 3.6599514318054195e-08, "loss": 0.978, "step": 11496 }, { "epoch": 0.97, "learning_rate": 3.636542010003519e-08, "loss": 0.9442, "step": 11497 }, { "epoch": 0.97, "learning_rate": 3.613207556734666e-08, "loss": 0.9378, "step": 11498 }, { "epoch": 0.97, "learning_rate": 3.5899480737544565e-08, "loss": 0.9616, "step": 11499 }, { "epoch": 0.97, "learning_rate": 3.566763562813047e-08, "loss": 0.9283, "step": 11500 }, { "epoch": 0.97, "learning_rate": 3.543654025655041e-08, "loss": 0.9807, "step": 11501 }, { "epoch": 0.97, "learning_rate": 3.5206194640189374e-08, "loss": 1.0004, "step": 11502 }, { "epoch": 0.97, "learning_rate": 3.4976598796381264e-08, "loss": 0.9552, "step": 11503 }, { "epoch": 0.97, "learning_rate": 3.474775274240228e-08, "loss": 0.942, "step": 11504 }, { "epoch": 0.97, "learning_rate": 3.451965649546751e-08, "loss": 0.953, "step": 11505 }, { "epoch": 0.97, "learning_rate": 3.429231007274325e-08, "loss": 0.9602, "step": 11506 }, { "epoch": 0.97, "learning_rate": 3.4065713491334696e-08, "loss": 0.9409, "step": 11507 }, { "epoch": 0.97, "learning_rate": 3.383986676828932e-08, "loss": 0.8973, "step": 11508 }, { "epoch": 0.97, "learning_rate": 3.361476992060242e-08, "loss": 0.787, "step": 11509 }, { "epoch": 0.97, "learning_rate": 3.339042296521045e-08, "loss": 0.9407, "step": 11510 }, { "epoch": 0.97, "learning_rate": 3.316682591899323e-08, "loss": 0.9334, "step": 11511 }, { "epoch": 0.97, "learning_rate": 3.2943978798776196e-08, "loss": 0.9675, "step": 11512 }, { "epoch": 0.98, "learning_rate": 3.2721881621323726e-08, "loss": 0.8732, "step": 11513 }, { "epoch": 0.98, "learning_rate": 3.250053440335021e-08, "loss": 0.9162, "step": 11514 }, { "epoch": 0.98, "learning_rate": 3.22799371615079e-08, "loss": 0.7486, "step": 11515 }, { "epoch": 0.98, "learning_rate": 3.206008991239684e-08, "loss": 0.9239, "step": 11516 }, { "epoch": 0.98, "learning_rate": 3.1840992672557135e-08, "loss": 0.9509, "step": 11517 }, { "epoch": 0.98, "learning_rate": 3.162264545847449e-08, "loss": 0.7791, "step": 11518 }, { "epoch": 0.98, "learning_rate": 3.140504828657909e-08, "loss": 0.9505, "step": 11519 }, { "epoch": 0.98, "learning_rate": 3.118820117324228e-08, "loss": 0.9495, "step": 11520 }, { "epoch": 0.98, "learning_rate": 3.097210413478102e-08, "loss": 0.9646, "step": 11521 }, { "epoch": 0.98, "learning_rate": 3.0756757187453415e-08, "loss": 0.9473, "step": 11522 }, { "epoch": 0.98, "learning_rate": 3.0542160347464265e-08, "loss": 0.9876, "step": 11523 }, { "epoch": 0.98, "learning_rate": 3.032831363095956e-08, "loss": 0.935, "step": 11524 }, { "epoch": 0.98, "learning_rate": 3.011521705402865e-08, "loss": 0.968, "step": 11525 }, { "epoch": 0.98, "learning_rate": 2.990287063270758e-08, "loss": 0.9171, "step": 11526 }, { "epoch": 0.98, "learning_rate": 2.9691274382971368e-08, "loss": 0.9523, "step": 11527 }, { "epoch": 0.98, "learning_rate": 2.9480428320742827e-08, "loss": 0.9657, "step": 11528 }, { "epoch": 0.98, "learning_rate": 2.927033246188593e-08, "loss": 0.9646, "step": 11529 }, { "epoch": 0.98, "learning_rate": 2.9060986822208036e-08, "loss": 0.9664, "step": 11530 }, { "epoch": 0.98, "learning_rate": 2.885239141746099e-08, "loss": 0.8646, "step": 11531 }, { "epoch": 0.98, "learning_rate": 2.864454626334112e-08, "loss": 0.8062, "step": 11532 }, { "epoch": 0.98, "learning_rate": 2.8437451375485924e-08, "loss": 0.9695, "step": 11533 }, { "epoch": 0.98, "learning_rate": 2.823110676947738e-08, "loss": 0.9407, "step": 11534 }, { "epoch": 0.98, "learning_rate": 2.8025512460843064e-08, "loss": 1.0137, "step": 11535 }, { "epoch": 0.98, "learning_rate": 2.7820668465049495e-08, "loss": 0.9852, "step": 11536 }, { "epoch": 0.98, "learning_rate": 2.761657479751323e-08, "loss": 0.9145, "step": 11537 }, { "epoch": 0.98, "learning_rate": 2.7413231473587542e-08, "loss": 0.987, "step": 11538 }, { "epoch": 0.98, "learning_rate": 2.721063850857353e-08, "loss": 0.9906, "step": 11539 }, { "epoch": 0.98, "learning_rate": 2.7008795917715658e-08, "loss": 0.9808, "step": 11540 }, { "epoch": 0.98, "learning_rate": 2.6807703716199562e-08, "loss": 0.998, "step": 11541 }, { "epoch": 0.98, "learning_rate": 2.660736191915647e-08, "loss": 0.9768, "step": 11542 }, { "epoch": 0.98, "learning_rate": 2.6407770541660993e-08, "loss": 0.9405, "step": 11543 }, { "epoch": 0.98, "learning_rate": 2.6208929598730004e-08, "loss": 0.9557, "step": 11544 }, { "epoch": 0.98, "learning_rate": 2.601083910532598e-08, "loss": 0.9042, "step": 11545 }, { "epoch": 0.98, "learning_rate": 2.5813499076351445e-08, "loss": 0.9994, "step": 11546 }, { "epoch": 0.98, "learning_rate": 2.5616909526656742e-08, "loss": 0.9274, "step": 11547 }, { "epoch": 0.98, "learning_rate": 2.542107047103226e-08, "loss": 0.9557, "step": 11548 }, { "epoch": 0.98, "learning_rate": 2.5225981924215103e-08, "loss": 1.015, "step": 11549 }, { "epoch": 0.98, "learning_rate": 2.503164390088242e-08, "loss": 0.924, "step": 11550 }, { "epoch": 0.98, "learning_rate": 2.4838056415656953e-08, "loss": 0.9245, "step": 11551 }, { "epoch": 0.98, "learning_rate": 2.4645219483105943e-08, "loss": 1.0097, "step": 11552 }, { "epoch": 0.98, "learning_rate": 2.445313311773778e-08, "loss": 0.9655, "step": 11553 }, { "epoch": 0.98, "learning_rate": 2.4261797334005354e-08, "loss": 0.9664, "step": 11554 }, { "epoch": 0.98, "learning_rate": 2.4071212146304923e-08, "loss": 0.9554, "step": 11555 }, { "epoch": 0.98, "learning_rate": 2.3881377568977238e-08, "loss": 0.9783, "step": 11556 }, { "epoch": 0.98, "learning_rate": 2.369229361630643e-08, "loss": 0.9176, "step": 11557 }, { "epoch": 0.98, "learning_rate": 2.350396030251778e-08, "loss": 1.0156, "step": 11558 }, { "epoch": 0.98, "learning_rate": 2.3316377641783295e-08, "loss": 0.9998, "step": 11559 }, { "epoch": 0.98, "learning_rate": 2.3129545648217235e-08, "loss": 0.9702, "step": 11560 }, { "epoch": 0.98, "learning_rate": 2.294346433587613e-08, "loss": 0.9769, "step": 11561 }, { "epoch": 0.98, "learning_rate": 2.275813371876212e-08, "loss": 0.9696, "step": 11562 }, { "epoch": 0.98, "learning_rate": 2.2573553810818494e-08, "loss": 0.9395, "step": 11563 }, { "epoch": 0.98, "learning_rate": 2.238972462593525e-08, "loss": 0.9828, "step": 11564 }, { "epoch": 0.98, "learning_rate": 2.2206646177943548e-08, "loss": 0.9333, "step": 11565 }, { "epoch": 0.98, "learning_rate": 2.2024318480617924e-08, "loss": 0.9532, "step": 11566 }, { "epoch": 0.98, "learning_rate": 2.1842741547677405e-08, "loss": 1.0021, "step": 11567 }, { "epoch": 0.98, "learning_rate": 2.1661915392784393e-08, "loss": 0.961, "step": 11568 }, { "epoch": 0.98, "learning_rate": 2.1481840029544677e-08, "loss": 1.0153, "step": 11569 }, { "epoch": 0.98, "learning_rate": 2.130251547150741e-08, "loss": 0.9981, "step": 11570 }, { "epoch": 0.98, "learning_rate": 2.1123941732165144e-08, "loss": 0.9885, "step": 11571 }, { "epoch": 0.98, "learning_rate": 2.0946118824954897e-08, "loss": 0.9464, "step": 11572 }, { "epoch": 0.98, "learning_rate": 2.0769046763255973e-08, "loss": 0.9209, "step": 11573 }, { "epoch": 0.98, "learning_rate": 2.059272556039105e-08, "loss": 0.9924, "step": 11574 }, { "epoch": 0.98, "learning_rate": 2.0417155229627283e-08, "loss": 0.9613, "step": 11575 }, { "epoch": 0.98, "learning_rate": 2.0242335784175226e-08, "loss": 0.8169, "step": 11576 }, { "epoch": 0.98, "learning_rate": 2.006826723718991e-08, "loss": 0.9123, "step": 11577 }, { "epoch": 0.98, "learning_rate": 1.9894949601766412e-08, "loss": 0.982, "step": 11578 }, { "epoch": 0.98, "learning_rate": 1.9722382890946522e-08, "loss": 0.9813, "step": 11579 }, { "epoch": 0.98, "learning_rate": 1.9550567117714305e-08, "loss": 0.9673, "step": 11580 }, { "epoch": 0.98, "learning_rate": 1.937950229499719e-08, "loss": 0.9417, "step": 11581 }, { "epoch": 0.98, "learning_rate": 1.9209188435667102e-08, "loss": 0.9604, "step": 11582 }, { "epoch": 0.98, "learning_rate": 1.9039625552539354e-08, "loss": 0.9539, "step": 11583 }, { "epoch": 0.98, "learning_rate": 1.8870813658371513e-08, "loss": 0.9652, "step": 11584 }, { "epoch": 0.98, "learning_rate": 1.870275276586564e-08, "loss": 0.9666, "step": 11585 }, { "epoch": 0.98, "learning_rate": 1.8535442887664958e-08, "loss": 0.9674, "step": 11586 }, { "epoch": 0.98, "learning_rate": 1.8368884036361612e-08, "loss": 1.0071, "step": 11587 }, { "epoch": 0.98, "learning_rate": 1.8203076224485584e-08, "loss": 0.9353, "step": 11588 }, { "epoch": 0.98, "learning_rate": 1.8038019464514667e-08, "loss": 0.9769, "step": 11589 }, { "epoch": 0.98, "learning_rate": 1.7873713768864488e-08, "loss": 0.9991, "step": 11590 }, { "epoch": 0.98, "learning_rate": 1.7710159149901818e-08, "loss": 0.9481, "step": 11591 }, { "epoch": 0.98, "learning_rate": 1.754735561993015e-08, "loss": 1.0091, "step": 11592 }, { "epoch": 0.98, "learning_rate": 1.738530319119969e-08, "loss": 0.9735, "step": 11593 }, { "epoch": 0.98, "learning_rate": 1.7224001875905117e-08, "loss": 0.9152, "step": 11594 }, { "epoch": 0.98, "learning_rate": 1.7063451686181176e-08, "loss": 0.9706, "step": 11595 }, { "epoch": 0.98, "learning_rate": 1.6903652634108202e-08, "loss": 0.9769, "step": 11596 }, { "epoch": 0.98, "learning_rate": 1.6744604731711024e-08, "loss": 0.9561, "step": 11597 }, { "epoch": 0.98, "learning_rate": 1.658630799095562e-08, "loss": 0.9479, "step": 11598 }, { "epoch": 0.98, "learning_rate": 1.6428762423753575e-08, "loss": 0.9662, "step": 11599 }, { "epoch": 0.98, "learning_rate": 1.627196804195874e-08, "loss": 0.9959, "step": 11600 }, { "epoch": 0.98, "learning_rate": 1.6115924857367238e-08, "loss": 0.9523, "step": 11601 }, { "epoch": 0.98, "learning_rate": 1.5960632881721893e-08, "loss": 0.983, "step": 11602 }, { "epoch": 0.98, "learning_rate": 1.5806092126706695e-08, "loss": 0.9156, "step": 11603 }, { "epoch": 0.98, "learning_rate": 1.565230260395012e-08, "loss": 0.9776, "step": 11604 }, { "epoch": 0.98, "learning_rate": 1.5499264325021802e-08, "loss": 0.9663, "step": 11605 }, { "epoch": 0.98, "learning_rate": 1.5346977301438083e-08, "loss": 0.9458, "step": 11606 }, { "epoch": 0.98, "learning_rate": 1.5195441544657574e-08, "loss": 0.7868, "step": 11607 }, { "epoch": 0.98, "learning_rate": 1.504465706608116e-08, "loss": 0.9605, "step": 11608 }, { "epoch": 0.98, "learning_rate": 1.4894623877054205e-08, "loss": 0.9662, "step": 11609 }, { "epoch": 0.98, "learning_rate": 1.4745341988867678e-08, "loss": 1.0337, "step": 11610 }, { "epoch": 0.98, "learning_rate": 1.4596811412750378e-08, "loss": 0.8717, "step": 11611 }, { "epoch": 0.98, "learning_rate": 1.4449032159881138e-08, "loss": 0.9417, "step": 11612 }, { "epoch": 0.98, "learning_rate": 1.4302004241377731e-08, "loss": 0.9669, "step": 11613 }, { "epoch": 0.98, "learning_rate": 1.4155727668303531e-08, "loss": 1.0094, "step": 11614 }, { "epoch": 0.98, "learning_rate": 1.4010202451663068e-08, "loss": 1.0119, "step": 11615 }, { "epoch": 0.98, "learning_rate": 1.3865428602408693e-08, "loss": 0.9111, "step": 11616 }, { "epoch": 0.98, "learning_rate": 1.3721406131430582e-08, "loss": 0.8295, "step": 11617 }, { "epoch": 0.98, "learning_rate": 1.3578135049567843e-08, "loss": 0.9416, "step": 11618 }, { "epoch": 0.98, "learning_rate": 1.343561536759852e-08, "loss": 0.927, "step": 11619 }, { "epoch": 0.98, "learning_rate": 1.329384709624737e-08, "loss": 0.9442, "step": 11620 }, { "epoch": 0.98, "learning_rate": 1.3152830246180303e-08, "loss": 0.9611, "step": 11621 }, { "epoch": 0.98, "learning_rate": 1.3012564828008834e-08, "loss": 0.9481, "step": 11622 }, { "epoch": 0.98, "learning_rate": 1.2873050852286739e-08, "loss": 0.9704, "step": 11623 }, { "epoch": 0.98, "learning_rate": 1.273428832951007e-08, "loss": 0.9818, "step": 11624 }, { "epoch": 0.98, "learning_rate": 1.2596277270120471e-08, "loss": 0.9997, "step": 11625 }, { "epoch": 0.98, "learning_rate": 1.2459017684502971e-08, "loss": 0.9442, "step": 11626 }, { "epoch": 0.98, "learning_rate": 1.2322509582984865e-08, "loss": 0.9075, "step": 11627 }, { "epoch": 0.98, "learning_rate": 1.2186752975835714e-08, "loss": 0.9628, "step": 11628 }, { "epoch": 0.98, "learning_rate": 1.2051747873271791e-08, "loss": 0.9234, "step": 11629 }, { "epoch": 0.98, "learning_rate": 1.1917494285451636e-08, "loss": 0.9625, "step": 11630 }, { "epoch": 0.99, "learning_rate": 1.178399222247495e-08, "loss": 1.0031, "step": 11631 }, { "epoch": 0.99, "learning_rate": 1.1651241694388138e-08, "loss": 0.9409, "step": 11632 }, { "epoch": 0.99, "learning_rate": 1.1519242711178769e-08, "loss": 0.9708, "step": 11633 }, { "epoch": 0.99, "learning_rate": 1.1387995282780006e-08, "loss": 0.973, "step": 11634 }, { "epoch": 0.99, "learning_rate": 1.1257499419065065e-08, "loss": 0.9622, "step": 11635 }, { "epoch": 0.99, "learning_rate": 1.1127755129854978e-08, "loss": 1.0019, "step": 11636 }, { "epoch": 0.99, "learning_rate": 1.0998762424910826e-08, "loss": 0.9297, "step": 11637 }, { "epoch": 0.99, "learning_rate": 1.0870521313938177e-08, "loss": 0.955, "step": 11638 }, { "epoch": 0.99, "learning_rate": 1.074303180658709e-08, "loss": 0.9048, "step": 11639 }, { "epoch": 0.99, "learning_rate": 1.0616293912448784e-08, "loss": 0.9895, "step": 11640 }, { "epoch": 0.99, "learning_rate": 1.0490307641061181e-08, "loss": 1.0096, "step": 11641 }, { "epoch": 0.99, "learning_rate": 1.036507300190115e-08, "loss": 0.9694, "step": 11642 }, { "epoch": 0.99, "learning_rate": 1.024059000439448e-08, "loss": 0.935, "step": 11643 }, { "epoch": 0.99, "learning_rate": 1.0116858657904794e-08, "loss": 0.9636, "step": 11644 }, { "epoch": 0.99, "learning_rate": 9.993878971744641e-09, "loss": 0.9584, "step": 11645 }, { "epoch": 0.99, "learning_rate": 9.871650955164402e-09, "loss": 0.992, "step": 11646 }, { "epoch": 0.99, "learning_rate": 9.750174617363384e-09, "loss": 0.9227, "step": 11647 }, { "epoch": 0.99, "learning_rate": 9.629449967479832e-09, "loss": 0.9916, "step": 11648 }, { "epoch": 0.99, "learning_rate": 9.5094770145987e-09, "loss": 0.9829, "step": 11649 }, { "epoch": 0.99, "learning_rate": 9.390255767746104e-09, "loss": 1.0274, "step": 11650 }, { "epoch": 0.99, "learning_rate": 9.271786235892644e-09, "loss": 1.0229, "step": 11651 }, { "epoch": 0.99, "learning_rate": 9.154068427952301e-09, "loss": 0.9708, "step": 11652 }, { "epoch": 0.99, "learning_rate": 9.037102352783544e-09, "loss": 0.8203, "step": 11653 }, { "epoch": 0.99, "learning_rate": 8.920888019184893e-09, "loss": 0.9708, "step": 11654 }, { "epoch": 0.99, "learning_rate": 8.805425435902681e-09, "loss": 0.9531, "step": 11655 }, { "epoch": 0.99, "learning_rate": 8.690714611623297e-09, "loss": 0.9865, "step": 11656 }, { "epoch": 0.99, "learning_rate": 8.576755554978721e-09, "loss": 0.9829, "step": 11657 }, { "epoch": 0.99, "learning_rate": 8.4635482745421e-09, "loss": 0.9701, "step": 11658 }, { "epoch": 0.99, "learning_rate": 8.35109277883217e-09, "loss": 0.9538, "step": 11659 }, { "epoch": 0.99, "learning_rate": 8.239389076311055e-09, "loss": 1.0112, "step": 11660 }, { "epoch": 0.99, "learning_rate": 8.128437175383142e-09, "loss": 0.9711, "step": 11661 }, { "epoch": 0.99, "learning_rate": 8.018237084396196e-09, "loss": 0.9757, "step": 11662 }, { "epoch": 0.99, "learning_rate": 7.908788811642477e-09, "loss": 0.9072, "step": 11663 }, { "epoch": 0.99, "learning_rate": 7.800092365356505e-09, "loss": 0.944, "step": 11664 }, { "epoch": 0.99, "learning_rate": 7.692147753717294e-09, "loss": 0.9807, "step": 11665 }, { "epoch": 0.99, "learning_rate": 7.584954984846128e-09, "loss": 0.9608, "step": 11666 }, { "epoch": 0.99, "learning_rate": 7.478514066808773e-09, "loss": 0.9549, "step": 11667 }, { "epoch": 0.99, "learning_rate": 7.372825007615492e-09, "loss": 0.9492, "step": 11668 }, { "epoch": 0.99, "learning_rate": 7.26788781521659e-09, "loss": 0.9395, "step": 11669 }, { "epoch": 0.99, "learning_rate": 7.163702497509084e-09, "loss": 0.9675, "step": 11670 }, { "epoch": 0.99, "learning_rate": 7.0602690623311484e-09, "loss": 0.926, "step": 11671 }, { "epoch": 0.99, "learning_rate": 6.957587517465447e-09, "loss": 0.7457, "step": 11672 }, { "epoch": 0.99, "learning_rate": 6.855657870639132e-09, "loss": 0.9699, "step": 11673 }, { "epoch": 0.99, "learning_rate": 6.754480129520513e-09, "loss": 0.9616, "step": 11674 }, { "epoch": 0.99, "learning_rate": 6.65405430172128e-09, "loss": 0.9432, "step": 11675 }, { "epoch": 0.99, "learning_rate": 6.554380394800941e-09, "loss": 0.9094, "step": 11676 }, { "epoch": 0.99, "learning_rate": 6.455458416255722e-09, "loss": 0.96, "step": 11677 }, { "epoch": 0.99, "learning_rate": 6.357288373531889e-09, "loss": 0.9872, "step": 11678 }, { "epoch": 0.99, "learning_rate": 6.259870274012425e-09, "loss": 0.9708, "step": 11679 }, { "epoch": 0.99, "learning_rate": 6.163204125030353e-09, "loss": 0.9427, "step": 11680 }, { "epoch": 0.99, "learning_rate": 6.0672899338587446e-09, "loss": 0.9519, "step": 11681 }, { "epoch": 0.99, "learning_rate": 5.972127707711828e-09, "loss": 0.9828, "step": 11682 }, { "epoch": 0.99, "learning_rate": 5.877717453752763e-09, "loss": 0.9865, "step": 11683 }, { "epoch": 0.99, "learning_rate": 5.784059179083645e-09, "loss": 0.9667, "step": 11684 }, { "epoch": 0.99, "learning_rate": 5.6911528907521714e-09, "loss": 0.9568, "step": 11685 }, { "epoch": 0.99, "learning_rate": 5.598998595747196e-09, "loss": 0.9845, "step": 11686 }, { "epoch": 0.99, "learning_rate": 5.507596301005391e-09, "loss": 0.9494, "step": 11687 }, { "epoch": 0.99, "learning_rate": 5.416946013401259e-09, "loss": 0.9534, "step": 11688 }, { "epoch": 0.99, "learning_rate": 5.327047739758229e-09, "loss": 0.9874, "step": 11689 }, { "epoch": 0.99, "learning_rate": 5.23790148683756e-09, "loss": 0.985, "step": 11690 }, { "epoch": 0.99, "learning_rate": 5.14950726134944e-09, "loss": 0.9745, "step": 11691 }, { "epoch": 0.99, "learning_rate": 5.061865069942995e-09, "loss": 0.9665, "step": 11692 }, { "epoch": 0.99, "learning_rate": 4.974974919214059e-09, "loss": 0.9621, "step": 11693 }, { "epoch": 0.99, "learning_rate": 4.888836815698517e-09, "loss": 0.9844, "step": 11694 }, { "epoch": 0.99, "learning_rate": 4.8034507658800685e-09, "loss": 0.9454, "step": 11695 }, { "epoch": 0.99, "learning_rate": 4.718816776181356e-09, "loss": 0.9261, "step": 11696 }, { "epoch": 0.99, "learning_rate": 4.634934852970619e-09, "loss": 0.7831, "step": 11697 }, { "epoch": 0.99, "learning_rate": 4.551805002560583e-09, "loss": 0.9652, "step": 11698 }, { "epoch": 0.99, "learning_rate": 4.469427231205137e-09, "loss": 0.9244, "step": 11699 }, { "epoch": 0.99, "learning_rate": 4.387801545102655e-09, "loss": 0.9717, "step": 11700 }, { "epoch": 0.99, "learning_rate": 4.3069279503948905e-09, "loss": 1.0085, "step": 11701 }, { "epoch": 0.99, "learning_rate": 4.226806453166976e-09, "loss": 0.9482, "step": 11702 }, { "epoch": 0.99, "learning_rate": 4.1474370594474235e-09, "loss": 0.9827, "step": 11703 }, { "epoch": 0.99, "learning_rate": 4.068819775208122e-09, "loss": 0.9592, "step": 11704 }, { "epoch": 0.99, "learning_rate": 3.990954606364339e-09, "loss": 0.9629, "step": 11705 }, { "epoch": 0.99, "learning_rate": 3.913841558774722e-09, "loss": 0.9779, "step": 11706 }, { "epoch": 0.99, "learning_rate": 3.837480638242408e-09, "loss": 0.9553, "step": 11707 }, { "epoch": 0.99, "learning_rate": 3.7618718505116895e-09, "loss": 0.9623, "step": 11708 }, { "epoch": 0.99, "learning_rate": 3.687015201271349e-09, "loss": 0.9647, "step": 11709 }, { "epoch": 0.99, "learning_rate": 3.612910696155769e-09, "loss": 0.924, "step": 11710 }, { "epoch": 0.99, "learning_rate": 3.5395583407382696e-09, "loss": 0.8248, "step": 11711 }, { "epoch": 0.99, "learning_rate": 3.4669581405388786e-09, "loss": 0.9287, "step": 11712 }, { "epoch": 0.99, "learning_rate": 3.3951101010221143e-09, "loss": 0.9777, "step": 11713 }, { "epoch": 0.99, "learning_rate": 3.324014227591432e-09, "loss": 0.9286, "step": 11714 }, { "epoch": 0.99, "learning_rate": 3.253670525596997e-09, "loss": 0.9833, "step": 11715 }, { "epoch": 0.99, "learning_rate": 3.1840790003312417e-09, "loss": 0.9931, "step": 11716 }, { "epoch": 0.99, "learning_rate": 3.1152396570321986e-09, "loss": 0.9962, "step": 11717 }, { "epoch": 0.99, "learning_rate": 3.0471525008768376e-09, "loss": 0.9641, "step": 11718 }, { "epoch": 0.99, "learning_rate": 2.9798175369910585e-09, "loss": 0.9446, "step": 11719 }, { "epoch": 0.99, "learning_rate": 2.9132347704385888e-09, "loss": 0.9354, "step": 11720 }, { "epoch": 0.99, "learning_rate": 2.847404206230975e-09, "loss": 0.9002, "step": 11721 }, { "epoch": 0.99, "learning_rate": 2.7823258493209217e-09, "loss": 0.9794, "step": 11722 }, { "epoch": 0.99, "learning_rate": 2.7179997046045127e-09, "loss": 0.9823, "step": 11723 }, { "epoch": 0.99, "learning_rate": 2.6544257769223204e-09, "loss": 0.9326, "step": 11724 }, { "epoch": 0.99, "learning_rate": 2.591604071058296e-09, "loss": 0.9409, "step": 11725 }, { "epoch": 0.99, "learning_rate": 2.5295345917386584e-09, "loss": 0.964, "step": 11726 }, { "epoch": 0.99, "learning_rate": 2.468217343634116e-09, "loss": 0.9948, "step": 11727 }, { "epoch": 0.99, "learning_rate": 2.407652331357646e-09, "loss": 0.9613, "step": 11728 }, { "epoch": 0.99, "learning_rate": 2.3478395594667135e-09, "loss": 0.9732, "step": 11729 }, { "epoch": 0.99, "learning_rate": 2.2887790324610526e-09, "loss": 0.9846, "step": 11730 }, { "epoch": 0.99, "learning_rate": 2.230470754785996e-09, "loss": 0.92, "step": 11731 }, { "epoch": 0.99, "learning_rate": 2.172914730826925e-09, "loss": 0.9164, "step": 11732 }, { "epoch": 0.99, "learning_rate": 2.1161109649159296e-09, "loss": 0.9443, "step": 11733 }, { "epoch": 0.99, "learning_rate": 2.060059461327368e-09, "loss": 0.9815, "step": 11734 }, { "epoch": 0.99, "learning_rate": 2.0047602242767584e-09, "loss": 0.9759, "step": 11735 }, { "epoch": 0.99, "learning_rate": 1.9502132579263257e-09, "loss": 0.9363, "step": 11736 }, { "epoch": 0.99, "learning_rate": 1.8964185663805645e-09, "loss": 0.8847, "step": 11737 }, { "epoch": 0.99, "learning_rate": 1.8433761536862383e-09, "loss": 0.9544, "step": 11738 }, { "epoch": 0.99, "learning_rate": 1.7910860238345983e-09, "loss": 0.9905, "step": 11739 }, { "epoch": 0.99, "learning_rate": 1.7395481807602754e-09, "loss": 0.9324, "step": 11740 }, { "epoch": 0.99, "learning_rate": 1.6887626283401682e-09, "loss": 0.9023, "step": 11741 }, { "epoch": 0.99, "learning_rate": 1.6387293703978847e-09, "loss": 0.9941, "step": 11742 }, { "epoch": 0.99, "learning_rate": 1.5894484106959706e-09, "loss": 0.9685, "step": 11743 }, { "epoch": 0.99, "learning_rate": 1.540919752941461e-09, "loss": 0.9686, "step": 11744 }, { "epoch": 0.99, "learning_rate": 1.4931434007892098e-09, "loss": 0.9804, "step": 11745 }, { "epoch": 0.99, "learning_rate": 1.4461193578307885e-09, "loss": 0.9352, "step": 11746 }, { "epoch": 0.99, "learning_rate": 1.3998476276055883e-09, "loss": 0.9706, "step": 11747 }, { "epoch": 0.99, "learning_rate": 1.3543282135963788e-09, "loss": 0.9663, "step": 11748 }, { "epoch": 1.0, "learning_rate": 1.3095611192259771e-09, "loss": 0.9021, "step": 11749 }, { "epoch": 1.0, "learning_rate": 1.2655463478639108e-09, "loss": 0.9542, "step": 11750 }, { "epoch": 1.0, "learning_rate": 1.2222839028219747e-09, "loss": 0.982, "step": 11751 }, { "epoch": 1.0, "learning_rate": 1.1797737873542325e-09, "loss": 1.0457, "step": 11752 }, { "epoch": 1.0, "learning_rate": 1.1380160046614574e-09, "loss": 0.96, "step": 11753 }, { "epoch": 1.0, "learning_rate": 1.0970105578833601e-09, "loss": 0.9984, "step": 11754 }, { "epoch": 1.0, "learning_rate": 1.0567574501074706e-09, "loss": 1.0162, "step": 11755 }, { "epoch": 1.0, "learning_rate": 1.0172566843602571e-09, "loss": 0.9899, "step": 11756 }, { "epoch": 1.0, "learning_rate": 9.785082636148968e-10, "loss": 0.9372, "step": 11757 }, { "epoch": 1.0, "learning_rate": 9.405121907868352e-10, "loss": 0.9498, "step": 11758 }, { "epoch": 1.0, "learning_rate": 9.03268468734897e-10, "loss": 0.9756, "step": 11759 }, { "epoch": 1.0, "learning_rate": 8.667771002623948e-10, "loss": 0.9731, "step": 11760 }, { "epoch": 1.0, "learning_rate": 8.310380881126901e-10, "loss": 0.9931, "step": 11761 }, { "epoch": 1.0, "learning_rate": 7.960514349769633e-10, "loss": 0.9156, "step": 11762 }, { "epoch": 1.0, "learning_rate": 7.618171434875531e-10, "loss": 0.9825, "step": 11763 }, { "epoch": 1.0, "learning_rate": 7.283352162190671e-10, "loss": 0.9665, "step": 11764 }, { "epoch": 1.0, "learning_rate": 6.956056556917112e-10, "loss": 0.933, "step": 11765 }, { "epoch": 1.0, "learning_rate": 6.636284643679603e-10, "loss": 0.9414, "step": 11766 }, { "epoch": 1.0, "learning_rate": 6.324036446536675e-10, "loss": 0.9063, "step": 11767 }, { "epoch": 1.0, "learning_rate": 6.019311988980647e-10, "loss": 0.9351, "step": 11768 }, { "epoch": 1.0, "learning_rate": 5.722111293937626e-10, "loss": 0.9597, "step": 11769 }, { "epoch": 1.0, "learning_rate": 5.432434383789708e-10, "loss": 0.9392, "step": 11770 }, { "epoch": 1.0, "learning_rate": 5.150281280308367e-10, "loss": 0.9568, "step": 11771 }, { "epoch": 1.0, "learning_rate": 4.875652004732167e-10, "loss": 0.9832, "step": 11772 }, { "epoch": 1.0, "learning_rate": 4.6085465777223613e-10, "loss": 1.0297, "step": 11773 }, { "epoch": 1.0, "learning_rate": 4.3489650193850877e-10, "loss": 0.941, "step": 11774 }, { "epoch": 1.0, "learning_rate": 4.096907349238066e-10, "loss": 1.0269, "step": 11775 }, { "epoch": 1.0, "learning_rate": 3.852373586266112e-10, "loss": 0.9998, "step": 11776 }, { "epoch": 1.0, "learning_rate": 3.6153637488545167e-10, "loss": 0.9881, "step": 11777 }, { "epoch": 1.0, "learning_rate": 3.3858778548334637e-10, "loss": 0.9472, "step": 11778 }, { "epoch": 1.0, "learning_rate": 3.1639159214780225e-10, "loss": 0.9694, "step": 11779 }, { "epoch": 1.0, "learning_rate": 2.9494779654859476e-10, "loss": 0.9294, "step": 11780 }, { "epoch": 1.0, "learning_rate": 2.7425640029998815e-10, "loss": 0.9605, "step": 11781 }, { "epoch": 1.0, "learning_rate": 2.543174049574049e-10, "loss": 0.9145, "step": 11782 }, { "epoch": 1.0, "learning_rate": 2.351308120218665e-10, "loss": 0.9772, "step": 11783 }, { "epoch": 1.0, "learning_rate": 2.1669662293666294e-10, "loss": 0.9741, "step": 11784 }, { "epoch": 1.0, "learning_rate": 1.9901483908957299e-10, "loss": 0.9689, "step": 11785 }, { "epoch": 1.0, "learning_rate": 1.820854618106438e-10, "loss": 0.9306, "step": 11786 }, { "epoch": 1.0, "learning_rate": 1.659084923733012e-10, "loss": 0.9291, "step": 11787 }, { "epoch": 1.0, "learning_rate": 1.5048393199545987e-10, "loss": 0.9452, "step": 11788 }, { "epoch": 1.0, "learning_rate": 1.358117818361926e-10, "loss": 0.9562, "step": 11789 }, { "epoch": 1.0, "learning_rate": 1.2189204300128154e-10, "loss": 0.9484, "step": 11790 }, { "epoch": 1.0, "learning_rate": 1.0872471653766703e-10, "loss": 0.9629, "step": 11791 }, { "epoch": 1.0, "learning_rate": 9.630980343566799e-11, "loss": 0.97, "step": 11792 }, { "epoch": 1.0, "learning_rate": 8.464730462898197e-11, "loss": 0.9313, "step": 11793 }, { "epoch": 1.0, "learning_rate": 7.373722099579539e-11, "loss": 0.9189, "step": 11794 }, { "epoch": 1.0, "learning_rate": 6.357955335767329e-11, "loss": 0.9812, "step": 11795 }, { "epoch": 1.0, "learning_rate": 5.41743024773389e-11, "loss": 0.9495, "step": 11796 }, { "epoch": 1.0, "learning_rate": 4.55214690631145e-11, "loss": 0.9708, "step": 11797 }, { "epoch": 1.0, "learning_rate": 3.7621053765590775e-11, "loss": 0.9349, "step": 11798 }, { "epoch": 1.0, "learning_rate": 3.04730571809575e-11, "loss": 0.9431, "step": 11799 }, { "epoch": 1.0, "learning_rate": 2.407747984545239e-11, "loss": 0.9341, "step": 11800 }, { "epoch": 1.0, "learning_rate": 1.8434322239802017e-11, "loss": 0.9409, "step": 11801 }, { "epoch": 1.0, "learning_rate": 1.354358479033202e-11, "loss": 0.9448, "step": 11802 }, { "epoch": 1.0, "learning_rate": 9.405267864526224e-12, "loss": 0.9722, "step": 11803 }, { "epoch": 1.0, "learning_rate": 6.019371773247073e-12, "loss": 0.962, "step": 11804 }, { "epoch": 1.0, "learning_rate": 3.3858967707356417e-12, "loss": 0.9813, "step": 11805 }, { "epoch": 1.0, "learning_rate": 1.5048430568320726e-12, "loss": 0.9109, "step": 11806 }, { "epoch": 1.0, "learning_rate": 3.7621077142446784e-13, "loss": 0.7963, "step": 11807 }, { "epoch": 1.0, "learning_rate": 0.0, "loss": 0.9768, "step": 11808 }, { "epoch": 1.0, "step": 11808, "total_flos": 3.828985861407716e+19, "train_loss": 1.0456744015196635, "train_runtime": 76041.8071, "train_samples_per_second": 19.875, "train_steps_per_second": 0.155 } ], "logging_steps": 1.0, "max_steps": 11808, "num_input_tokens_seen": 0, "num_train_epochs": 1, "save_steps": 1000, "total_flos": 3.828985861407716e+19, "train_batch_size": 8, "trial_name": null, "trial_params": null }