{ "best_metric": 1.4374229907989502, "best_model_checkpoint": "bird_sounds_classification_2/checkpoint-52500", "epoch": 10.0, "global_step": 52500, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 6.000000000000001e-08, "loss": 3.8409, "step": 1 }, { "epoch": 0.0, "learning_rate": 1.2000000000000002e-07, "loss": 3.8115, "step": 2 }, { "epoch": 0.0, "learning_rate": 1.8e-07, "loss": 3.8496, "step": 3 }, { "epoch": 0.0, "learning_rate": 2.4000000000000003e-07, "loss": 3.8188, "step": 4 }, { "epoch": 0.0, "learning_rate": 3.0000000000000004e-07, "loss": 3.8076, "step": 5 }, { "epoch": 0.0, "learning_rate": 3.6e-07, "loss": 3.8327, "step": 6 }, { "epoch": 0.0, "learning_rate": 4.2000000000000006e-07, "loss": 3.8284, "step": 7 }, { "epoch": 0.0, "learning_rate": 4.800000000000001e-07, "loss": 3.8288, "step": 8 }, { "epoch": 0.0, "learning_rate": 5.4e-07, "loss": 3.8159, "step": 9 }, { "epoch": 0.0, "learning_rate": 6.000000000000001e-07, "loss": 3.8522, "step": 10 }, { "epoch": 0.0, "learning_rate": 6.6e-07, "loss": 3.8133, "step": 11 }, { "epoch": 0.0, "learning_rate": 7.2e-07, "loss": 3.812, "step": 12 }, { "epoch": 0.0, "learning_rate": 7.8e-07, "loss": 3.8424, "step": 13 }, { "epoch": 0.0, "learning_rate": 8.400000000000001e-07, "loss": 3.9233, "step": 14 }, { "epoch": 0.0, "learning_rate": 9e-07, "loss": 3.8258, "step": 15 }, { "epoch": 0.0, "learning_rate": 9.600000000000001e-07, "loss": 3.8377, "step": 16 }, { "epoch": 0.0, "learning_rate": 1.0200000000000002e-06, "loss": 3.8402, "step": 17 }, { "epoch": 0.0, "learning_rate": 1.08e-06, "loss": 3.8522, "step": 18 }, { "epoch": 0.0, "learning_rate": 1.14e-06, "loss": 3.8257, "step": 19 }, { "epoch": 0.0, "learning_rate": 1.2000000000000002e-06, "loss": 3.8196, "step": 20 }, { "epoch": 0.0, "learning_rate": 1.26e-06, "loss": 3.8401, "step": 21 }, { "epoch": 0.0, "learning_rate": 1.32e-06, "loss": 3.8201, "step": 22 }, { "epoch": 0.0, "learning_rate": 1.3800000000000001e-06, "loss": 3.8195, "step": 23 }, { "epoch": 0.0, "learning_rate": 1.44e-06, "loss": 3.8198, "step": 24 }, { "epoch": 0.0, "learning_rate": 1.5e-06, "loss": 3.8285, "step": 25 }, { "epoch": 0.0, "learning_rate": 1.56e-06, "loss": 3.8289, "step": 26 }, { "epoch": 0.01, "learning_rate": 1.6200000000000002e-06, "loss": 3.8279, "step": 27 }, { "epoch": 0.01, "learning_rate": 1.6800000000000002e-06, "loss": 3.817, "step": 28 }, { "epoch": 0.01, "learning_rate": 1.7399999999999999e-06, "loss": 3.8149, "step": 29 }, { "epoch": 0.01, "learning_rate": 1.8e-06, "loss": 3.8429, "step": 30 }, { "epoch": 0.01, "learning_rate": 1.86e-06, "loss": 3.8467, "step": 31 }, { "epoch": 0.01, "learning_rate": 1.9200000000000003e-06, "loss": 3.8313, "step": 32 }, { "epoch": 0.01, "learning_rate": 1.98e-06, "loss": 3.83, "step": 33 }, { "epoch": 0.01, "learning_rate": 2.0400000000000004e-06, "loss": 3.833, "step": 34 }, { "epoch": 0.01, "learning_rate": 2.1e-06, "loss": 3.8461, "step": 35 }, { "epoch": 0.01, "learning_rate": 2.16e-06, "loss": 3.8324, "step": 36 }, { "epoch": 0.01, "learning_rate": 2.22e-06, "loss": 3.8152, "step": 37 }, { "epoch": 0.01, "learning_rate": 2.28e-06, "loss": 3.8281, "step": 38 }, { "epoch": 0.01, "learning_rate": 2.34e-06, "loss": 3.8273, "step": 39 }, { "epoch": 0.01, "learning_rate": 2.4000000000000003e-06, "loss": 3.8327, "step": 40 }, { "epoch": 0.01, "learning_rate": 2.4599999999999997e-06, "loss": 3.8305, "step": 41 }, { "epoch": 0.01, "learning_rate": 2.52e-06, "loss": 3.8263, "step": 42 }, { "epoch": 0.01, "learning_rate": 2.58e-06, "loss": 3.8335, "step": 43 }, { "epoch": 0.01, "learning_rate": 2.64e-06, "loss": 3.8231, "step": 44 }, { "epoch": 0.01, "learning_rate": 2.7e-06, "loss": 3.8181, "step": 45 }, { "epoch": 0.01, "learning_rate": 2.7600000000000003e-06, "loss": 3.8383, "step": 46 }, { "epoch": 0.01, "learning_rate": 2.82e-06, "loss": 3.8484, "step": 47 }, { "epoch": 0.01, "learning_rate": 2.88e-06, "loss": 3.8058, "step": 48 }, { "epoch": 0.01, "learning_rate": 2.9400000000000002e-06, "loss": 3.7957, "step": 49 }, { "epoch": 0.01, "learning_rate": 3e-06, "loss": 3.8399, "step": 50 }, { "epoch": 0.01, "learning_rate": 2.999942802669209e-06, "loss": 3.8484, "step": 51 }, { "epoch": 0.01, "learning_rate": 2.9998856053384177e-06, "loss": 3.8435, "step": 52 }, { "epoch": 0.01, "learning_rate": 2.9998284080076265e-06, "loss": 3.7893, "step": 53 }, { "epoch": 0.01, "learning_rate": 2.999771210676835e-06, "loss": 3.8257, "step": 54 }, { "epoch": 0.01, "learning_rate": 2.999714013346044e-06, "loss": 3.8424, "step": 55 }, { "epoch": 0.01, "learning_rate": 2.999656816015253e-06, "loss": 3.825, "step": 56 }, { "epoch": 0.01, "learning_rate": 2.999599618684461e-06, "loss": 3.8331, "step": 57 }, { "epoch": 0.01, "learning_rate": 2.9995424213536704e-06, "loss": 3.8465, "step": 58 }, { "epoch": 0.01, "learning_rate": 2.999485224022879e-06, "loss": 3.8263, "step": 59 }, { "epoch": 0.01, "learning_rate": 2.999428026692088e-06, "loss": 3.8102, "step": 60 }, { "epoch": 0.01, "learning_rate": 2.9993708293612964e-06, "loss": 3.8239, "step": 61 }, { "epoch": 0.01, "learning_rate": 2.999313632030505e-06, "loss": 3.8213, "step": 62 }, { "epoch": 0.01, "learning_rate": 2.9992564346997144e-06, "loss": 3.853, "step": 63 }, { "epoch": 0.01, "learning_rate": 2.9991992373689228e-06, "loss": 3.804, "step": 64 }, { "epoch": 0.01, "learning_rate": 2.9991420400381316e-06, "loss": 3.8134, "step": 65 }, { "epoch": 0.01, "learning_rate": 2.9990848427073404e-06, "loss": 3.8243, "step": 66 }, { "epoch": 0.01, "learning_rate": 2.999027645376549e-06, "loss": 3.8519, "step": 67 }, { "epoch": 0.01, "learning_rate": 2.998970448045758e-06, "loss": 3.801, "step": 68 }, { "epoch": 0.01, "learning_rate": 2.9989132507149667e-06, "loss": 3.8296, "step": 69 }, { "epoch": 0.01, "learning_rate": 2.9988560533841755e-06, "loss": 3.8851, "step": 70 }, { "epoch": 0.01, "learning_rate": 2.998798856053384e-06, "loss": 3.8509, "step": 71 }, { "epoch": 0.01, "learning_rate": 2.998741658722593e-06, "loss": 3.8196, "step": 72 }, { "epoch": 0.01, "learning_rate": 2.998684461391802e-06, "loss": 3.8084, "step": 73 }, { "epoch": 0.01, "learning_rate": 2.9986272640610103e-06, "loss": 3.8442, "step": 74 }, { "epoch": 0.01, "learning_rate": 2.9985700667302195e-06, "loss": 3.8233, "step": 75 }, { "epoch": 0.01, "learning_rate": 2.998512869399428e-06, "loss": 3.8138, "step": 76 }, { "epoch": 0.01, "learning_rate": 2.998455672068637e-06, "loss": 3.8093, "step": 77 }, { "epoch": 0.01, "learning_rate": 2.998398474737846e-06, "loss": 3.838, "step": 78 }, { "epoch": 0.02, "learning_rate": 2.9983412774070542e-06, "loss": 3.8349, "step": 79 }, { "epoch": 0.02, "learning_rate": 2.9982840800762635e-06, "loss": 3.8033, "step": 80 }, { "epoch": 0.02, "learning_rate": 2.998226882745472e-06, "loss": 3.8154, "step": 81 }, { "epoch": 0.02, "learning_rate": 2.9981696854146806e-06, "loss": 3.8341, "step": 82 }, { "epoch": 0.02, "learning_rate": 2.9981124880838894e-06, "loss": 3.8384, "step": 83 }, { "epoch": 0.02, "learning_rate": 2.9980552907530982e-06, "loss": 3.8432, "step": 84 }, { "epoch": 0.02, "learning_rate": 2.997998093422307e-06, "loss": 3.8288, "step": 85 }, { "epoch": 0.02, "learning_rate": 2.997940896091516e-06, "loss": 3.8656, "step": 86 }, { "epoch": 0.02, "learning_rate": 2.9978836987607246e-06, "loss": 3.8461, "step": 87 }, { "epoch": 0.02, "learning_rate": 2.9978265014299334e-06, "loss": 3.8157, "step": 88 }, { "epoch": 0.02, "learning_rate": 2.997769304099142e-06, "loss": 3.8248, "step": 89 }, { "epoch": 0.02, "learning_rate": 2.997712106768351e-06, "loss": 3.8439, "step": 90 }, { "epoch": 0.02, "learning_rate": 2.9976549094375598e-06, "loss": 3.813, "step": 91 }, { "epoch": 0.02, "learning_rate": 2.9975977121067686e-06, "loss": 3.8364, "step": 92 }, { "epoch": 0.02, "learning_rate": 2.997540514775977e-06, "loss": 3.8344, "step": 93 }, { "epoch": 0.02, "learning_rate": 2.997483317445186e-06, "loss": 3.8238, "step": 94 }, { "epoch": 0.02, "learning_rate": 2.997426120114395e-06, "loss": 3.802, "step": 95 }, { "epoch": 0.02, "learning_rate": 2.9973689227836033e-06, "loss": 3.8273, "step": 96 }, { "epoch": 0.02, "learning_rate": 2.9973117254528125e-06, "loss": 3.8337, "step": 97 }, { "epoch": 0.02, "learning_rate": 2.997254528122021e-06, "loss": 3.807, "step": 98 }, { "epoch": 0.02, "learning_rate": 2.9971973307912297e-06, "loss": 3.8401, "step": 99 }, { "epoch": 0.02, "learning_rate": 2.9971401334604385e-06, "loss": 3.8277, "step": 100 }, { "epoch": 0.02, "learning_rate": 2.9970829361296473e-06, "loss": 3.8279, "step": 101 }, { "epoch": 0.02, "learning_rate": 2.997025738798856e-06, "loss": 3.8313, "step": 102 }, { "epoch": 0.02, "learning_rate": 2.996968541468065e-06, "loss": 3.8243, "step": 103 }, { "epoch": 0.02, "learning_rate": 2.9969113441372737e-06, "loss": 3.8475, "step": 104 }, { "epoch": 0.02, "learning_rate": 2.9968541468064825e-06, "loss": 3.8342, "step": 105 }, { "epoch": 0.02, "learning_rate": 2.9967969494756913e-06, "loss": 3.849, "step": 106 }, { "epoch": 0.02, "learning_rate": 2.9967397521449e-06, "loss": 3.8109, "step": 107 }, { "epoch": 0.02, "learning_rate": 2.996682554814109e-06, "loss": 3.8192, "step": 108 }, { "epoch": 0.02, "learning_rate": 2.9966253574833176e-06, "loss": 3.829, "step": 109 }, { "epoch": 0.02, "learning_rate": 2.996568160152526e-06, "loss": 3.8549, "step": 110 }, { "epoch": 0.02, "learning_rate": 2.9965109628217352e-06, "loss": 3.8104, "step": 111 }, { "epoch": 0.02, "learning_rate": 2.996453765490944e-06, "loss": 3.824, "step": 112 }, { "epoch": 0.02, "learning_rate": 2.9963965681601524e-06, "loss": 3.8297, "step": 113 }, { "epoch": 0.02, "learning_rate": 2.9963393708293616e-06, "loss": 3.8307, "step": 114 }, { "epoch": 0.02, "learning_rate": 2.99628217349857e-06, "loss": 3.8413, "step": 115 }, { "epoch": 0.02, "learning_rate": 2.9962249761677788e-06, "loss": 3.84, "step": 116 }, { "epoch": 0.02, "learning_rate": 2.996167778836988e-06, "loss": 3.8472, "step": 117 }, { "epoch": 0.02, "learning_rate": 2.9961105815061964e-06, "loss": 3.8197, "step": 118 }, { "epoch": 0.02, "learning_rate": 2.996053384175405e-06, "loss": 3.8287, "step": 119 }, { "epoch": 0.02, "learning_rate": 2.995996186844614e-06, "loss": 3.8222, "step": 120 }, { "epoch": 0.02, "learning_rate": 2.9959389895138227e-06, "loss": 3.8349, "step": 121 }, { "epoch": 0.02, "learning_rate": 2.9958817921830315e-06, "loss": 3.8329, "step": 122 }, { "epoch": 0.02, "learning_rate": 2.9958245948522403e-06, "loss": 3.8177, "step": 123 }, { "epoch": 0.02, "learning_rate": 2.995767397521449e-06, "loss": 3.863, "step": 124 }, { "epoch": 0.02, "learning_rate": 2.995710200190658e-06, "loss": 3.8544, "step": 125 }, { "epoch": 0.02, "learning_rate": 2.9956530028598667e-06, "loss": 3.8147, "step": 126 }, { "epoch": 0.02, "learning_rate": 2.9955958055290755e-06, "loss": 3.8437, "step": 127 }, { "epoch": 0.02, "learning_rate": 2.9955386081982843e-06, "loss": 3.8404, "step": 128 }, { "epoch": 0.02, "learning_rate": 2.995481410867493e-06, "loss": 3.8217, "step": 129 }, { "epoch": 0.02, "learning_rate": 2.9954242135367015e-06, "loss": 3.8341, "step": 130 }, { "epoch": 0.02, "learning_rate": 2.9953670162059107e-06, "loss": 3.8344, "step": 131 }, { "epoch": 0.03, "learning_rate": 2.995309818875119e-06, "loss": 3.8384, "step": 132 }, { "epoch": 0.03, "learning_rate": 2.995252621544328e-06, "loss": 3.8294, "step": 133 }, { "epoch": 0.03, "learning_rate": 2.995195424213537e-06, "loss": 3.8095, "step": 134 }, { "epoch": 0.03, "learning_rate": 2.9951382268827454e-06, "loss": 3.816, "step": 135 }, { "epoch": 0.03, "learning_rate": 2.9950810295519542e-06, "loss": 3.8062, "step": 136 }, { "epoch": 0.03, "learning_rate": 2.995023832221163e-06, "loss": 3.8259, "step": 137 }, { "epoch": 0.03, "learning_rate": 2.994966634890372e-06, "loss": 3.807, "step": 138 }, { "epoch": 0.03, "learning_rate": 2.9949094375595806e-06, "loss": 3.8418, "step": 139 }, { "epoch": 0.03, "learning_rate": 2.9948522402287894e-06, "loss": 3.8135, "step": 140 }, { "epoch": 0.03, "learning_rate": 2.994795042897998e-06, "loss": 3.8458, "step": 141 }, { "epoch": 0.03, "learning_rate": 2.994737845567207e-06, "loss": 3.805, "step": 142 }, { "epoch": 0.03, "learning_rate": 2.9946806482364158e-06, "loss": 3.8247, "step": 143 }, { "epoch": 0.03, "learning_rate": 2.9946234509056246e-06, "loss": 3.8272, "step": 144 }, { "epoch": 0.03, "learning_rate": 2.9945662535748334e-06, "loss": 3.8677, "step": 145 }, { "epoch": 0.03, "learning_rate": 2.994509056244042e-06, "loss": 3.8162, "step": 146 }, { "epoch": 0.03, "learning_rate": 2.9944518589132505e-06, "loss": 3.8138, "step": 147 }, { "epoch": 0.03, "learning_rate": 2.9943946615824598e-06, "loss": 3.8308, "step": 148 }, { "epoch": 0.03, "learning_rate": 2.994337464251668e-06, "loss": 3.8477, "step": 149 }, { "epoch": 0.03, "learning_rate": 2.994280266920877e-06, "loss": 3.8287, "step": 150 }, { "epoch": 0.03, "learning_rate": 2.994223069590086e-06, "loss": 3.828, "step": 151 }, { "epoch": 0.03, "learning_rate": 2.9941658722592945e-06, "loss": 3.8517, "step": 152 }, { "epoch": 0.03, "learning_rate": 2.9941086749285033e-06, "loss": 3.8424, "step": 153 }, { "epoch": 0.03, "learning_rate": 2.994051477597712e-06, "loss": 3.8134, "step": 154 }, { "epoch": 0.03, "learning_rate": 2.993994280266921e-06, "loss": 3.807, "step": 155 }, { "epoch": 0.03, "learning_rate": 2.9939370829361297e-06, "loss": 3.8409, "step": 156 }, { "epoch": 0.03, "learning_rate": 2.9938798856053385e-06, "loss": 3.8226, "step": 157 }, { "epoch": 0.03, "learning_rate": 2.9938226882745473e-06, "loss": 3.8405, "step": 158 }, { "epoch": 0.03, "learning_rate": 2.993765490943756e-06, "loss": 3.8264, "step": 159 }, { "epoch": 0.03, "learning_rate": 2.993708293612965e-06, "loss": 3.8199, "step": 160 }, { "epoch": 0.03, "learning_rate": 2.9936510962821736e-06, "loss": 3.8209, "step": 161 }, { "epoch": 0.03, "learning_rate": 2.9935938989513824e-06, "loss": 3.846, "step": 162 }, { "epoch": 0.03, "learning_rate": 2.9935367016205912e-06, "loss": 3.8205, "step": 163 }, { "epoch": 0.03, "learning_rate": 2.9934795042897996e-06, "loss": 3.8368, "step": 164 }, { "epoch": 0.03, "learning_rate": 2.993422306959009e-06, "loss": 3.8449, "step": 165 }, { "epoch": 0.03, "learning_rate": 2.9933651096282176e-06, "loss": 3.8094, "step": 166 }, { "epoch": 0.03, "learning_rate": 2.993307912297426e-06, "loss": 3.8091, "step": 167 }, { "epoch": 0.03, "learning_rate": 2.993250714966635e-06, "loss": 3.8239, "step": 168 }, { "epoch": 0.03, "learning_rate": 2.9931935176358436e-06, "loss": 3.8115, "step": 169 }, { "epoch": 0.03, "learning_rate": 2.9931363203050524e-06, "loss": 3.817, "step": 170 }, { "epoch": 0.03, "learning_rate": 2.993079122974261e-06, "loss": 3.8211, "step": 171 }, { "epoch": 0.03, "learning_rate": 2.99302192564347e-06, "loss": 3.8296, "step": 172 }, { "epoch": 0.03, "learning_rate": 2.992964728312679e-06, "loss": 3.8305, "step": 173 }, { "epoch": 0.03, "learning_rate": 2.9929075309818875e-06, "loss": 3.8168, "step": 174 }, { "epoch": 0.03, "learning_rate": 2.9928503336510963e-06, "loss": 3.8158, "step": 175 }, { "epoch": 0.03, "learning_rate": 2.992793136320305e-06, "loss": 3.8196, "step": 176 }, { "epoch": 0.03, "learning_rate": 2.992735938989514e-06, "loss": 3.7929, "step": 177 }, { "epoch": 0.03, "learning_rate": 2.9926787416587227e-06, "loss": 3.7885, "step": 178 }, { "epoch": 0.03, "learning_rate": 2.9926215443279315e-06, "loss": 3.838, "step": 179 }, { "epoch": 0.03, "learning_rate": 2.9925643469971403e-06, "loss": 3.8383, "step": 180 }, { "epoch": 0.03, "learning_rate": 2.9925071496663487e-06, "loss": 3.81, "step": 181 }, { "epoch": 0.03, "learning_rate": 2.992449952335558e-06, "loss": 3.8062, "step": 182 }, { "epoch": 0.03, "learning_rate": 2.9923927550047667e-06, "loss": 3.8355, "step": 183 }, { "epoch": 0.04, "learning_rate": 2.992335557673975e-06, "loss": 3.8156, "step": 184 }, { "epoch": 0.04, "learning_rate": 2.9922783603431843e-06, "loss": 3.8205, "step": 185 }, { "epoch": 0.04, "learning_rate": 2.9922211630123926e-06, "loss": 3.8264, "step": 186 }, { "epoch": 0.04, "learning_rate": 2.9921639656816014e-06, "loss": 3.7861, "step": 187 }, { "epoch": 0.04, "learning_rate": 2.9921067683508102e-06, "loss": 3.8161, "step": 188 }, { "epoch": 0.04, "learning_rate": 2.992049571020019e-06, "loss": 3.8159, "step": 189 }, { "epoch": 0.04, "learning_rate": 2.9919923736892282e-06, "loss": 3.8217, "step": 190 }, { "epoch": 0.04, "learning_rate": 2.9919351763584366e-06, "loss": 3.847, "step": 191 }, { "epoch": 0.04, "learning_rate": 2.9918779790276454e-06, "loss": 3.8452, "step": 192 }, { "epoch": 0.04, "learning_rate": 2.991820781696854e-06, "loss": 3.8212, "step": 193 }, { "epoch": 0.04, "learning_rate": 2.991763584366063e-06, "loss": 3.8184, "step": 194 }, { "epoch": 0.04, "learning_rate": 2.991706387035272e-06, "loss": 3.8193, "step": 195 }, { "epoch": 0.04, "learning_rate": 2.9916491897044806e-06, "loss": 3.8144, "step": 196 }, { "epoch": 0.04, "learning_rate": 2.9915919923736894e-06, "loss": 3.801, "step": 197 }, { "epoch": 0.04, "learning_rate": 2.9915347950428977e-06, "loss": 3.8461, "step": 198 }, { "epoch": 0.04, "learning_rate": 2.991477597712107e-06, "loss": 3.8666, "step": 199 }, { "epoch": 0.04, "learning_rate": 2.9914204003813158e-06, "loss": 3.7996, "step": 200 }, { "epoch": 0.04, "learning_rate": 2.991363203050524e-06, "loss": 3.8339, "step": 201 }, { "epoch": 0.04, "learning_rate": 2.9913060057197333e-06, "loss": 3.8294, "step": 202 }, { "epoch": 0.04, "learning_rate": 2.9912488083889417e-06, "loss": 3.8391, "step": 203 }, { "epoch": 0.04, "learning_rate": 2.9911916110581505e-06, "loss": 3.783, "step": 204 }, { "epoch": 0.04, "learning_rate": 2.9911344137273597e-06, "loss": 3.8161, "step": 205 }, { "epoch": 0.04, "learning_rate": 2.991077216396568e-06, "loss": 3.8235, "step": 206 }, { "epoch": 0.04, "learning_rate": 2.9910200190657773e-06, "loss": 3.8116, "step": 207 }, { "epoch": 0.04, "learning_rate": 2.9909628217349857e-06, "loss": 3.8205, "step": 208 }, { "epoch": 0.04, "learning_rate": 2.9909056244041945e-06, "loss": 3.804, "step": 209 }, { "epoch": 0.04, "learning_rate": 2.9908484270734033e-06, "loss": 3.8327, "step": 210 }, { "epoch": 0.04, "learning_rate": 2.990791229742612e-06, "loss": 3.8331, "step": 211 }, { "epoch": 0.04, "learning_rate": 2.990734032411821e-06, "loss": 3.7979, "step": 212 }, { "epoch": 0.04, "learning_rate": 2.9906768350810297e-06, "loss": 3.8335, "step": 213 }, { "epoch": 0.04, "learning_rate": 2.9906196377502385e-06, "loss": 3.824, "step": 214 }, { "epoch": 0.04, "learning_rate": 2.9905624404194472e-06, "loss": 3.8136, "step": 215 }, { "epoch": 0.04, "learning_rate": 2.990505243088656e-06, "loss": 3.8144, "step": 216 }, { "epoch": 0.04, "learning_rate": 2.990448045757865e-06, "loss": 3.8108, "step": 217 }, { "epoch": 0.04, "learning_rate": 2.990390848427073e-06, "loss": 3.8156, "step": 218 }, { "epoch": 0.04, "learning_rate": 2.9903336510962824e-06, "loss": 3.8362, "step": 219 }, { "epoch": 0.04, "learning_rate": 2.990276453765491e-06, "loss": 3.8459, "step": 220 }, { "epoch": 0.04, "learning_rate": 2.9902192564347e-06, "loss": 3.8199, "step": 221 }, { "epoch": 0.04, "learning_rate": 2.990162059103909e-06, "loss": 3.8433, "step": 222 }, { "epoch": 0.04, "learning_rate": 2.990104861773117e-06, "loss": 3.8097, "step": 223 }, { "epoch": 0.04, "learning_rate": 2.9900476644423264e-06, "loss": 3.7971, "step": 224 }, { "epoch": 0.04, "learning_rate": 2.9899904671115348e-06, "loss": 3.8316, "step": 225 }, { "epoch": 0.04, "learning_rate": 2.9899332697807436e-06, "loss": 3.8192, "step": 226 }, { "epoch": 0.04, "learning_rate": 2.9898760724499523e-06, "loss": 3.845, "step": 227 }, { "epoch": 0.04, "learning_rate": 2.989818875119161e-06, "loss": 3.851, "step": 228 }, { "epoch": 0.04, "learning_rate": 2.98976167778837e-06, "loss": 3.8106, "step": 229 }, { "epoch": 0.04, "learning_rate": 2.9897044804575787e-06, "loss": 3.8424, "step": 230 }, { "epoch": 0.04, "learning_rate": 2.9896472831267875e-06, "loss": 3.8163, "step": 231 }, { "epoch": 0.04, "learning_rate": 2.9895900857959963e-06, "loss": 3.8055, "step": 232 }, { "epoch": 0.04, "learning_rate": 2.989532888465205e-06, "loss": 3.8341, "step": 233 }, { "epoch": 0.04, "learning_rate": 2.989475691134414e-06, "loss": 3.8238, "step": 234 }, { "epoch": 0.04, "learning_rate": 2.9894184938036223e-06, "loss": 3.8431, "step": 235 }, { "epoch": 0.04, "learning_rate": 2.9893612964728315e-06, "loss": 3.7812, "step": 236 }, { "epoch": 0.05, "learning_rate": 2.98930409914204e-06, "loss": 3.8085, "step": 237 }, { "epoch": 0.05, "learning_rate": 2.989246901811249e-06, "loss": 3.7934, "step": 238 }, { "epoch": 0.05, "learning_rate": 2.989189704480458e-06, "loss": 3.8471, "step": 239 }, { "epoch": 0.05, "learning_rate": 2.9891325071496662e-06, "loss": 3.7891, "step": 240 }, { "epoch": 0.05, "learning_rate": 2.9890753098188755e-06, "loss": 3.8452, "step": 241 }, { "epoch": 0.05, "learning_rate": 2.989018112488084e-06, "loss": 3.8113, "step": 242 }, { "epoch": 0.05, "learning_rate": 2.9889609151572926e-06, "loss": 3.801, "step": 243 }, { "epoch": 0.05, "learning_rate": 2.9889037178265014e-06, "loss": 3.7982, "step": 244 }, { "epoch": 0.05, "learning_rate": 2.9888465204957102e-06, "loss": 3.8331, "step": 245 }, { "epoch": 0.05, "learning_rate": 2.988789323164919e-06, "loss": 3.813, "step": 246 }, { "epoch": 0.05, "learning_rate": 2.988732125834128e-06, "loss": 3.787, "step": 247 }, { "epoch": 0.05, "learning_rate": 2.9886749285033366e-06, "loss": 3.8407, "step": 248 }, { "epoch": 0.05, "learning_rate": 2.9886177311725454e-06, "loss": 3.8453, "step": 249 }, { "epoch": 0.05, "learning_rate": 2.988560533841754e-06, "loss": 3.8411, "step": 250 }, { "epoch": 0.05, "learning_rate": 2.988503336510963e-06, "loss": 3.8271, "step": 251 }, { "epoch": 0.05, "learning_rate": 2.9884461391801713e-06, "loss": 3.8197, "step": 252 }, { "epoch": 0.05, "learning_rate": 2.9883889418493806e-06, "loss": 3.8268, "step": 253 }, { "epoch": 0.05, "learning_rate": 2.9883317445185894e-06, "loss": 3.8008, "step": 254 }, { "epoch": 0.05, "learning_rate": 2.988274547187798e-06, "loss": 3.8288, "step": 255 }, { "epoch": 0.05, "learning_rate": 2.988217349857007e-06, "loss": 3.8458, "step": 256 }, { "epoch": 0.05, "learning_rate": 2.9881601525262153e-06, "loss": 3.8206, "step": 257 }, { "epoch": 0.05, "learning_rate": 2.9881029551954245e-06, "loss": 3.7998, "step": 258 }, { "epoch": 0.05, "learning_rate": 2.988045757864633e-06, "loss": 3.8822, "step": 259 }, { "epoch": 0.05, "learning_rate": 2.9879885605338417e-06, "loss": 3.7824, "step": 260 }, { "epoch": 0.05, "learning_rate": 2.987931363203051e-06, "loss": 3.8664, "step": 261 }, { "epoch": 0.05, "learning_rate": 2.9878741658722593e-06, "loss": 3.7952, "step": 262 }, { "epoch": 0.05, "learning_rate": 2.987816968541468e-06, "loss": 3.8203, "step": 263 }, { "epoch": 0.05, "learning_rate": 2.987759771210677e-06, "loss": 3.8299, "step": 264 }, { "epoch": 0.05, "learning_rate": 2.9877025738798857e-06, "loss": 3.8395, "step": 265 }, { "epoch": 0.05, "learning_rate": 2.9876453765490945e-06, "loss": 3.8174, "step": 266 }, { "epoch": 0.05, "learning_rate": 2.9875881792183033e-06, "loss": 3.8275, "step": 267 }, { "epoch": 0.05, "learning_rate": 2.987530981887512e-06, "loss": 3.8391, "step": 268 }, { "epoch": 0.05, "learning_rate": 2.9874737845567204e-06, "loss": 3.7951, "step": 269 }, { "epoch": 0.05, "learning_rate": 2.9874165872259296e-06, "loss": 3.8002, "step": 270 }, { "epoch": 0.05, "learning_rate": 2.9873593898951384e-06, "loss": 3.818, "step": 271 }, { "epoch": 0.05, "learning_rate": 2.9873021925643472e-06, "loss": 3.8266, "step": 272 }, { "epoch": 0.05, "learning_rate": 2.987244995233556e-06, "loss": 3.8049, "step": 273 }, { "epoch": 0.05, "learning_rate": 2.9871877979027644e-06, "loss": 3.854, "step": 274 }, { "epoch": 0.05, "learning_rate": 2.9871306005719736e-06, "loss": 3.8051, "step": 275 }, { "epoch": 0.05, "learning_rate": 2.987073403241182e-06, "loss": 3.8384, "step": 276 }, { "epoch": 0.05, "learning_rate": 2.9870162059103908e-06, "loss": 3.7955, "step": 277 }, { "epoch": 0.05, "learning_rate": 2.9869590085796e-06, "loss": 3.868, "step": 278 }, { "epoch": 0.05, "learning_rate": 2.9869018112488084e-06, "loss": 3.7929, "step": 279 }, { "epoch": 0.05, "learning_rate": 2.986844613918017e-06, "loss": 3.821, "step": 280 }, { "epoch": 0.05, "learning_rate": 2.986787416587226e-06, "loss": 3.8441, "step": 281 }, { "epoch": 0.05, "learning_rate": 2.9867302192564347e-06, "loss": 3.8417, "step": 282 }, { "epoch": 0.05, "learning_rate": 2.9866730219256435e-06, "loss": 3.8391, "step": 283 }, { "epoch": 0.05, "learning_rate": 2.9866158245948523e-06, "loss": 3.8155, "step": 284 }, { "epoch": 0.05, "learning_rate": 2.986558627264061e-06, "loss": 3.7889, "step": 285 }, { "epoch": 0.05, "learning_rate": 2.98650142993327e-06, "loss": 3.813, "step": 286 }, { "epoch": 0.05, "learning_rate": 2.9864442326024787e-06, "loss": 3.8251, "step": 287 }, { "epoch": 0.05, "learning_rate": 2.9863870352716875e-06, "loss": 3.8609, "step": 288 }, { "epoch": 0.06, "learning_rate": 2.9863298379408963e-06, "loss": 3.8243, "step": 289 }, { "epoch": 0.06, "learning_rate": 2.986272640610105e-06, "loss": 3.809, "step": 290 }, { "epoch": 0.06, "learning_rate": 2.9862154432793135e-06, "loss": 3.8364, "step": 291 }, { "epoch": 0.06, "learning_rate": 2.9861582459485227e-06, "loss": 3.7969, "step": 292 }, { "epoch": 0.06, "learning_rate": 2.986101048617731e-06, "loss": 3.7262, "step": 293 }, { "epoch": 0.06, "learning_rate": 2.98604385128694e-06, "loss": 3.7967, "step": 294 }, { "epoch": 0.06, "learning_rate": 2.985986653956149e-06, "loss": 3.8294, "step": 295 }, { "epoch": 0.06, "learning_rate": 2.9859294566253574e-06, "loss": 3.8678, "step": 296 }, { "epoch": 0.06, "learning_rate": 2.9858722592945662e-06, "loss": 3.841, "step": 297 }, { "epoch": 0.06, "learning_rate": 2.985815061963775e-06, "loss": 3.8234, "step": 298 }, { "epoch": 0.06, "learning_rate": 2.985757864632984e-06, "loss": 3.8346, "step": 299 }, { "epoch": 0.06, "learning_rate": 2.9857006673021926e-06, "loss": 3.8395, "step": 300 }, { "epoch": 0.06, "learning_rate": 2.9856434699714014e-06, "loss": 3.8413, "step": 301 }, { "epoch": 0.06, "learning_rate": 2.98558627264061e-06, "loss": 3.8426, "step": 302 }, { "epoch": 0.06, "learning_rate": 2.985529075309819e-06, "loss": 3.8278, "step": 303 }, { "epoch": 0.06, "learning_rate": 2.9854718779790278e-06, "loss": 3.7619, "step": 304 }, { "epoch": 0.06, "learning_rate": 2.9854146806482366e-06, "loss": 3.8382, "step": 305 }, { "epoch": 0.06, "learning_rate": 2.9853574833174454e-06, "loss": 3.8284, "step": 306 }, { "epoch": 0.06, "learning_rate": 2.985300285986654e-06, "loss": 3.8257, "step": 307 }, { "epoch": 0.06, "learning_rate": 2.9852430886558625e-06, "loss": 3.8108, "step": 308 }, { "epoch": 0.06, "learning_rate": 2.9851858913250717e-06, "loss": 3.7963, "step": 309 }, { "epoch": 0.06, "learning_rate": 2.9851286939942805e-06, "loss": 3.8264, "step": 310 }, { "epoch": 0.06, "learning_rate": 2.985071496663489e-06, "loss": 3.8106, "step": 311 }, { "epoch": 0.06, "learning_rate": 2.985014299332698e-06, "loss": 3.8233, "step": 312 }, { "epoch": 0.06, "learning_rate": 2.9849571020019065e-06, "loss": 3.7968, "step": 313 }, { "epoch": 0.06, "learning_rate": 2.9848999046711153e-06, "loss": 3.8258, "step": 314 }, { "epoch": 0.06, "learning_rate": 2.984842707340324e-06, "loss": 3.829, "step": 315 }, { "epoch": 0.06, "learning_rate": 2.984785510009533e-06, "loss": 3.8259, "step": 316 }, { "epoch": 0.06, "learning_rate": 2.9847283126787417e-06, "loss": 3.8004, "step": 317 }, { "epoch": 0.06, "learning_rate": 2.9846711153479505e-06, "loss": 3.8296, "step": 318 }, { "epoch": 0.06, "learning_rate": 2.9846139180171593e-06, "loss": 3.8209, "step": 319 }, { "epoch": 0.06, "learning_rate": 2.984556720686368e-06, "loss": 3.8334, "step": 320 }, { "epoch": 0.06, "learning_rate": 2.984499523355577e-06, "loss": 3.8093, "step": 321 }, { "epoch": 0.06, "learning_rate": 2.9844423260247856e-06, "loss": 3.7859, "step": 322 }, { "epoch": 0.06, "learning_rate": 2.9843851286939944e-06, "loss": 3.8071, "step": 323 }, { "epoch": 0.06, "learning_rate": 2.9843279313632032e-06, "loss": 3.7858, "step": 324 }, { "epoch": 0.06, "learning_rate": 2.9842707340324116e-06, "loss": 3.8317, "step": 325 }, { "epoch": 0.06, "learning_rate": 2.984213536701621e-06, "loss": 3.8335, "step": 326 }, { "epoch": 0.06, "learning_rate": 2.9841563393708296e-06, "loss": 3.7504, "step": 327 }, { "epoch": 0.06, "learning_rate": 2.984099142040038e-06, "loss": 3.7998, "step": 328 }, { "epoch": 0.06, "learning_rate": 2.984041944709247e-06, "loss": 3.8277, "step": 329 }, { "epoch": 0.06, "learning_rate": 2.9839847473784556e-06, "loss": 3.8285, "step": 330 }, { "epoch": 0.06, "learning_rate": 2.9839275500476644e-06, "loss": 3.8264, "step": 331 }, { "epoch": 0.06, "learning_rate": 2.983870352716873e-06, "loss": 3.7415, "step": 332 }, { "epoch": 0.06, "learning_rate": 2.983813155386082e-06, "loss": 3.8066, "step": 333 }, { "epoch": 0.06, "learning_rate": 2.9837559580552907e-06, "loss": 3.8446, "step": 334 }, { "epoch": 0.06, "learning_rate": 2.9836987607244995e-06, "loss": 3.8218, "step": 335 }, { "epoch": 0.06, "learning_rate": 2.9836415633937083e-06, "loss": 3.8025, "step": 336 }, { "epoch": 0.06, "learning_rate": 2.983584366062917e-06, "loss": 3.7533, "step": 337 }, { "epoch": 0.06, "learning_rate": 2.983527168732126e-06, "loss": 3.7726, "step": 338 }, { "epoch": 0.06, "learning_rate": 2.9834699714013347e-06, "loss": 3.8211, "step": 339 }, { "epoch": 0.06, "learning_rate": 2.9834127740705435e-06, "loss": 3.8139, "step": 340 }, { "epoch": 0.06, "learning_rate": 2.9833555767397523e-06, "loss": 3.8443, "step": 341 }, { "epoch": 0.07, "learning_rate": 2.983298379408961e-06, "loss": 3.8326, "step": 342 }, { "epoch": 0.07, "learning_rate": 2.98324118207817e-06, "loss": 3.8021, "step": 343 }, { "epoch": 0.07, "learning_rate": 2.9831839847473787e-06, "loss": 3.778, "step": 344 }, { "epoch": 0.07, "learning_rate": 2.983126787416587e-06, "loss": 3.8353, "step": 345 }, { "epoch": 0.07, "learning_rate": 2.9830695900857963e-06, "loss": 3.8169, "step": 346 }, { "epoch": 0.07, "learning_rate": 2.9830123927550046e-06, "loss": 3.8451, "step": 347 }, { "epoch": 0.07, "learning_rate": 2.9829551954242134e-06, "loss": 3.8064, "step": 348 }, { "epoch": 0.07, "learning_rate": 2.9828979980934227e-06, "loss": 3.7952, "step": 349 }, { "epoch": 0.07, "learning_rate": 2.982840800762631e-06, "loss": 3.8167, "step": 350 }, { "epoch": 0.07, "learning_rate": 2.9827836034318402e-06, "loss": 3.7935, "step": 351 }, { "epoch": 0.07, "learning_rate": 2.9827264061010486e-06, "loss": 3.8113, "step": 352 }, { "epoch": 0.07, "learning_rate": 2.9826692087702574e-06, "loss": 3.7714, "step": 353 }, { "epoch": 0.07, "learning_rate": 2.982612011439466e-06, "loss": 3.7706, "step": 354 }, { "epoch": 0.07, "learning_rate": 2.982554814108675e-06, "loss": 3.8509, "step": 355 }, { "epoch": 0.07, "learning_rate": 2.9824976167778838e-06, "loss": 3.838, "step": 356 }, { "epoch": 0.07, "learning_rate": 2.9824404194470926e-06, "loss": 3.8194, "step": 357 }, { "epoch": 0.07, "learning_rate": 2.9823832221163014e-06, "loss": 3.8219, "step": 358 }, { "epoch": 0.07, "learning_rate": 2.98232602478551e-06, "loss": 3.8855, "step": 359 }, { "epoch": 0.07, "learning_rate": 2.982268827454719e-06, "loss": 3.7391, "step": 360 }, { "epoch": 0.07, "learning_rate": 2.9822116301239278e-06, "loss": 3.7752, "step": 361 }, { "epoch": 0.07, "learning_rate": 2.982154432793136e-06, "loss": 3.8104, "step": 362 }, { "epoch": 0.07, "learning_rate": 2.9820972354623453e-06, "loss": 3.8196, "step": 363 }, { "epoch": 0.07, "learning_rate": 2.9820400381315537e-06, "loss": 3.7002, "step": 364 }, { "epoch": 0.07, "learning_rate": 2.9819828408007625e-06, "loss": 3.8023, "step": 365 }, { "epoch": 0.07, "learning_rate": 2.9819256434699717e-06, "loss": 3.8249, "step": 366 }, { "epoch": 0.07, "learning_rate": 2.98186844613918e-06, "loss": 3.8116, "step": 367 }, { "epoch": 0.07, "learning_rate": 2.9818112488083893e-06, "loss": 3.711, "step": 368 }, { "epoch": 0.07, "learning_rate": 2.9817540514775977e-06, "loss": 3.9098, "step": 369 }, { "epoch": 0.07, "learning_rate": 2.9816968541468065e-06, "loss": 3.7811, "step": 370 }, { "epoch": 0.07, "learning_rate": 2.9816396568160153e-06, "loss": 3.7752, "step": 371 }, { "epoch": 0.07, "learning_rate": 2.981582459485224e-06, "loss": 3.7572, "step": 372 }, { "epoch": 0.07, "learning_rate": 2.981525262154433e-06, "loss": 3.8082, "step": 373 }, { "epoch": 0.07, "learning_rate": 2.9814680648236416e-06, "loss": 3.8409, "step": 374 }, { "epoch": 0.07, "learning_rate": 2.9814108674928504e-06, "loss": 3.8035, "step": 375 }, { "epoch": 0.07, "learning_rate": 2.9813536701620592e-06, "loss": 3.8306, "step": 376 }, { "epoch": 0.07, "learning_rate": 2.981296472831268e-06, "loss": 3.7698, "step": 377 }, { "epoch": 0.07, "learning_rate": 2.981239275500477e-06, "loss": 3.8113, "step": 378 }, { "epoch": 0.07, "learning_rate": 2.981182078169685e-06, "loss": 3.768, "step": 379 }, { "epoch": 0.07, "learning_rate": 2.9811248808388944e-06, "loss": 3.8559, "step": 380 }, { "epoch": 0.07, "learning_rate": 2.9810676835081028e-06, "loss": 3.8404, "step": 381 }, { "epoch": 0.07, "learning_rate": 2.9810104861773116e-06, "loss": 3.7388, "step": 382 }, { "epoch": 0.07, "learning_rate": 2.980953288846521e-06, "loss": 3.8, "step": 383 }, { "epoch": 0.07, "learning_rate": 2.980896091515729e-06, "loss": 3.8353, "step": 384 }, { "epoch": 0.07, "learning_rate": 2.9808388941849384e-06, "loss": 3.8419, "step": 385 }, { "epoch": 0.07, "learning_rate": 2.9807816968541468e-06, "loss": 3.8242, "step": 386 }, { "epoch": 0.07, "learning_rate": 2.9807244995233555e-06, "loss": 3.8036, "step": 387 }, { "epoch": 0.07, "learning_rate": 2.9806673021925648e-06, "loss": 3.8058, "step": 388 }, { "epoch": 0.07, "learning_rate": 2.980610104861773e-06, "loss": 3.8166, "step": 389 }, { "epoch": 0.07, "learning_rate": 2.980552907530982e-06, "loss": 3.8854, "step": 390 }, { "epoch": 0.07, "learning_rate": 2.9804957102001907e-06, "loss": 3.789, "step": 391 }, { "epoch": 0.07, "learning_rate": 2.9804385128693995e-06, "loss": 3.7909, "step": 392 }, { "epoch": 0.07, "learning_rate": 2.9803813155386083e-06, "loss": 3.7891, "step": 393 }, { "epoch": 0.08, "learning_rate": 2.980324118207817e-06, "loss": 3.7047, "step": 394 }, { "epoch": 0.08, "learning_rate": 2.980266920877026e-06, "loss": 3.7908, "step": 395 }, { "epoch": 0.08, "learning_rate": 2.9802097235462343e-06, "loss": 3.827, "step": 396 }, { "epoch": 0.08, "learning_rate": 2.9801525262154435e-06, "loss": 3.8092, "step": 397 }, { "epoch": 0.08, "learning_rate": 2.9800953288846523e-06, "loss": 3.8749, "step": 398 }, { "epoch": 0.08, "learning_rate": 2.9800381315538606e-06, "loss": 3.816, "step": 399 }, { "epoch": 0.08, "learning_rate": 2.97998093422307e-06, "loss": 3.8213, "step": 400 }, { "epoch": 0.08, "learning_rate": 2.9799237368922782e-06, "loss": 3.8512, "step": 401 }, { "epoch": 0.08, "learning_rate": 2.9798665395614875e-06, "loss": 3.7538, "step": 402 }, { "epoch": 0.08, "learning_rate": 2.979809342230696e-06, "loss": 3.7768, "step": 403 }, { "epoch": 0.08, "learning_rate": 2.9797521448999046e-06, "loss": 3.8113, "step": 404 }, { "epoch": 0.08, "learning_rate": 2.979694947569114e-06, "loss": 3.8123, "step": 405 }, { "epoch": 0.08, "learning_rate": 2.979637750238322e-06, "loss": 3.8102, "step": 406 }, { "epoch": 0.08, "learning_rate": 2.979580552907531e-06, "loss": 3.8072, "step": 407 }, { "epoch": 0.08, "learning_rate": 2.97952335557674e-06, "loss": 3.782, "step": 408 }, { "epoch": 0.08, "learning_rate": 2.9794661582459486e-06, "loss": 3.8109, "step": 409 }, { "epoch": 0.08, "learning_rate": 2.9794089609151574e-06, "loss": 3.7942, "step": 410 }, { "epoch": 0.08, "learning_rate": 2.979351763584366e-06, "loss": 3.8421, "step": 411 }, { "epoch": 0.08, "learning_rate": 2.979294566253575e-06, "loss": 3.7145, "step": 412 }, { "epoch": 0.08, "learning_rate": 2.9792373689227833e-06, "loss": 3.8248, "step": 413 }, { "epoch": 0.08, "learning_rate": 2.9791801715919926e-06, "loss": 3.8178, "step": 414 }, { "epoch": 0.08, "learning_rate": 2.9791229742612013e-06, "loss": 3.7323, "step": 415 }, { "epoch": 0.08, "learning_rate": 2.9790657769304097e-06, "loss": 3.789, "step": 416 }, { "epoch": 0.08, "learning_rate": 2.979008579599619e-06, "loss": 3.7916, "step": 417 }, { "epoch": 0.08, "learning_rate": 2.9789513822688273e-06, "loss": 3.7489, "step": 418 }, { "epoch": 0.08, "learning_rate": 2.9788941849380365e-06, "loss": 3.8172, "step": 419 }, { "epoch": 0.08, "learning_rate": 2.978836987607245e-06, "loss": 3.8386, "step": 420 }, { "epoch": 0.08, "learning_rate": 2.9787797902764537e-06, "loss": 3.8295, "step": 421 }, { "epoch": 0.08, "learning_rate": 2.978722592945663e-06, "loss": 3.7134, "step": 422 }, { "epoch": 0.08, "learning_rate": 2.9786653956148713e-06, "loss": 3.8003, "step": 423 }, { "epoch": 0.08, "learning_rate": 2.97860819828408e-06, "loss": 3.8219, "step": 424 }, { "epoch": 0.08, "learning_rate": 2.978551000953289e-06, "loss": 3.7614, "step": 425 }, { "epoch": 0.08, "learning_rate": 2.9784938036224977e-06, "loss": 3.8152, "step": 426 }, { "epoch": 0.08, "learning_rate": 2.9784366062917065e-06, "loss": 3.7526, "step": 427 }, { "epoch": 0.08, "learning_rate": 2.9783794089609152e-06, "loss": 3.7393, "step": 428 }, { "epoch": 0.08, "learning_rate": 2.978322211630124e-06, "loss": 3.7256, "step": 429 }, { "epoch": 0.08, "learning_rate": 2.978265014299333e-06, "loss": 3.8231, "step": 430 }, { "epoch": 0.08, "learning_rate": 2.9782078169685416e-06, "loss": 3.8526, "step": 431 }, { "epoch": 0.08, "learning_rate": 2.9781506196377504e-06, "loss": 3.8589, "step": 432 }, { "epoch": 0.08, "learning_rate": 2.9780934223069592e-06, "loss": 3.8204, "step": 433 }, { "epoch": 0.08, "learning_rate": 2.978036224976168e-06, "loss": 3.7986, "step": 434 }, { "epoch": 0.08, "learning_rate": 2.9779790276453764e-06, "loss": 3.8301, "step": 435 }, { "epoch": 0.08, "learning_rate": 2.9779218303145856e-06, "loss": 3.7885, "step": 436 }, { "epoch": 0.08, "learning_rate": 2.9778646329837944e-06, "loss": 3.8088, "step": 437 }, { "epoch": 0.08, "learning_rate": 2.9778074356530028e-06, "loss": 3.7659, "step": 438 }, { "epoch": 0.08, "learning_rate": 2.977750238322212e-06, "loss": 3.8339, "step": 439 }, { "epoch": 0.08, "learning_rate": 2.9776930409914203e-06, "loss": 3.8151, "step": 440 }, { "epoch": 0.08, "learning_rate": 2.977635843660629e-06, "loss": 3.7853, "step": 441 }, { "epoch": 0.08, "learning_rate": 2.977578646329838e-06, "loss": 3.7845, "step": 442 }, { "epoch": 0.08, "learning_rate": 2.9775214489990467e-06, "loss": 3.7883, "step": 443 }, { "epoch": 0.08, "learning_rate": 2.9774642516682555e-06, "loss": 3.775, "step": 444 }, { "epoch": 0.08, "learning_rate": 2.9774070543374643e-06, "loss": 3.6974, "step": 445 }, { "epoch": 0.08, "learning_rate": 2.977349857006673e-06, "loss": 3.7953, "step": 446 }, { "epoch": 0.09, "learning_rate": 2.977292659675882e-06, "loss": 3.7982, "step": 447 }, { "epoch": 0.09, "learning_rate": 2.9772354623450907e-06, "loss": 3.768, "step": 448 }, { "epoch": 0.09, "learning_rate": 2.9771782650142995e-06, "loss": 3.7813, "step": 449 }, { "epoch": 0.09, "learning_rate": 2.9771210676835083e-06, "loss": 3.7367, "step": 450 }, { "epoch": 0.09, "learning_rate": 2.977063870352717e-06, "loss": 3.7667, "step": 451 }, { "epoch": 0.09, "learning_rate": 2.9770066730219254e-06, "loss": 3.7436, "step": 452 }, { "epoch": 0.09, "learning_rate": 2.9769494756911347e-06, "loss": 3.7524, "step": 453 }, { "epoch": 0.09, "learning_rate": 2.9768922783603435e-06, "loss": 3.789, "step": 454 }, { "epoch": 0.09, "learning_rate": 2.976835081029552e-06, "loss": 3.8241, "step": 455 }, { "epoch": 0.09, "learning_rate": 2.976777883698761e-06, "loss": 3.7449, "step": 456 }, { "epoch": 0.09, "learning_rate": 2.9767206863679694e-06, "loss": 3.8099, "step": 457 }, { "epoch": 0.09, "learning_rate": 2.9766634890371782e-06, "loss": 3.777, "step": 458 }, { "epoch": 0.09, "learning_rate": 2.976606291706387e-06, "loss": 3.8336, "step": 459 }, { "epoch": 0.09, "learning_rate": 2.976549094375596e-06, "loss": 3.8339, "step": 460 }, { "epoch": 0.09, "learning_rate": 2.9764918970448046e-06, "loss": 3.856, "step": 461 }, { "epoch": 0.09, "learning_rate": 2.9764346997140134e-06, "loss": 3.8801, "step": 462 }, { "epoch": 0.09, "learning_rate": 2.976377502383222e-06, "loss": 3.753, "step": 463 }, { "epoch": 0.09, "learning_rate": 2.976320305052431e-06, "loss": 3.8696, "step": 464 }, { "epoch": 0.09, "learning_rate": 2.9762631077216398e-06, "loss": 3.7927, "step": 465 }, { "epoch": 0.09, "learning_rate": 2.9762059103908486e-06, "loss": 3.8808, "step": 466 }, { "epoch": 0.09, "learning_rate": 2.9761487130600574e-06, "loss": 3.7821, "step": 467 }, { "epoch": 0.09, "learning_rate": 2.976091515729266e-06, "loss": 3.8038, "step": 468 }, { "epoch": 0.09, "learning_rate": 2.9760343183984745e-06, "loss": 3.8325, "step": 469 }, { "epoch": 0.09, "learning_rate": 2.9759771210676837e-06, "loss": 3.7883, "step": 470 }, { "epoch": 0.09, "learning_rate": 2.9759199237368925e-06, "loss": 3.8107, "step": 471 }, { "epoch": 0.09, "learning_rate": 2.975862726406101e-06, "loss": 3.7987, "step": 472 }, { "epoch": 0.09, "learning_rate": 2.97580552907531e-06, "loss": 3.826, "step": 473 }, { "epoch": 0.09, "learning_rate": 2.9757483317445185e-06, "loss": 3.8253, "step": 474 }, { "epoch": 0.09, "learning_rate": 2.9756911344137273e-06, "loss": 3.7895, "step": 475 }, { "epoch": 0.09, "learning_rate": 2.9756339370829365e-06, "loss": 3.8068, "step": 476 }, { "epoch": 0.09, "learning_rate": 2.975576739752145e-06, "loss": 3.737, "step": 477 }, { "epoch": 0.09, "learning_rate": 2.9755195424213537e-06, "loss": 3.7553, "step": 478 }, { "epoch": 0.09, "learning_rate": 2.9754623450905625e-06, "loss": 3.8415, "step": 479 }, { "epoch": 0.09, "learning_rate": 2.9754051477597713e-06, "loss": 3.7706, "step": 480 }, { "epoch": 0.09, "learning_rate": 2.97534795042898e-06, "loss": 3.8166, "step": 481 }, { "epoch": 0.09, "learning_rate": 2.975290753098189e-06, "loss": 3.8182, "step": 482 }, { "epoch": 0.09, "learning_rate": 2.9752335557673976e-06, "loss": 3.8532, "step": 483 }, { "epoch": 0.09, "learning_rate": 2.9751763584366064e-06, "loss": 3.8385, "step": 484 }, { "epoch": 0.09, "learning_rate": 2.9751191611058152e-06, "loss": 3.7619, "step": 485 }, { "epoch": 0.09, "learning_rate": 2.975061963775024e-06, "loss": 3.7618, "step": 486 }, { "epoch": 0.09, "learning_rate": 2.975004766444233e-06, "loss": 3.8767, "step": 487 }, { "epoch": 0.09, "learning_rate": 2.9749475691134416e-06, "loss": 3.8022, "step": 488 }, { "epoch": 0.09, "learning_rate": 2.97489037178265e-06, "loss": 3.8129, "step": 489 }, { "epoch": 0.09, "learning_rate": 2.974833174451859e-06, "loss": 3.7882, "step": 490 }, { "epoch": 0.09, "learning_rate": 2.9747759771210676e-06, "loss": 3.7689, "step": 491 }, { "epoch": 0.09, "learning_rate": 2.9747187797902764e-06, "loss": 3.8623, "step": 492 }, { "epoch": 0.09, "learning_rate": 2.9746615824594856e-06, "loss": 3.6873, "step": 493 }, { "epoch": 0.09, "learning_rate": 2.974604385128694e-06, "loss": 3.8334, "step": 494 }, { "epoch": 0.09, "learning_rate": 2.9745471877979027e-06, "loss": 3.7793, "step": 495 }, { "epoch": 0.09, "learning_rate": 2.9744899904671115e-06, "loss": 3.7766, "step": 496 }, { "epoch": 0.09, "learning_rate": 2.9744327931363203e-06, "loss": 3.7554, "step": 497 }, { "epoch": 0.09, "learning_rate": 2.974375595805529e-06, "loss": 3.7761, "step": 498 }, { "epoch": 0.1, "learning_rate": 2.974318398474738e-06, "loss": 3.7186, "step": 499 }, { "epoch": 0.1, "learning_rate": 2.9742612011439467e-06, "loss": 3.8113, "step": 500 }, { "epoch": 0.1, "learning_rate": 2.9742040038131555e-06, "loss": 3.8538, "step": 501 }, { "epoch": 0.1, "learning_rate": 2.9741468064823643e-06, "loss": 3.8685, "step": 502 }, { "epoch": 0.1, "learning_rate": 2.974089609151573e-06, "loss": 3.7203, "step": 503 }, { "epoch": 0.1, "learning_rate": 2.974032411820782e-06, "loss": 3.7665, "step": 504 }, { "epoch": 0.1, "learning_rate": 2.9739752144899907e-06, "loss": 3.831, "step": 505 }, { "epoch": 0.1, "learning_rate": 2.973918017159199e-06, "loss": 3.8067, "step": 506 }, { "epoch": 0.1, "learning_rate": 2.9738608198284083e-06, "loss": 3.7714, "step": 507 }, { "epoch": 0.1, "learning_rate": 2.9738036224976166e-06, "loss": 3.8242, "step": 508 }, { "epoch": 0.1, "learning_rate": 2.9737464251668254e-06, "loss": 3.8214, "step": 509 }, { "epoch": 0.1, "learning_rate": 2.9736892278360346e-06, "loss": 3.7472, "step": 510 }, { "epoch": 0.1, "learning_rate": 2.973632030505243e-06, "loss": 3.818, "step": 511 }, { "epoch": 0.1, "learning_rate": 2.973574833174452e-06, "loss": 3.7173, "step": 512 }, { "epoch": 0.1, "learning_rate": 2.9735176358436606e-06, "loss": 3.7795, "step": 513 }, { "epoch": 0.1, "learning_rate": 2.9734604385128694e-06, "loss": 3.7547, "step": 514 }, { "epoch": 0.1, "learning_rate": 2.9734032411820786e-06, "loss": 3.7299, "step": 515 }, { "epoch": 0.1, "learning_rate": 2.973346043851287e-06, "loss": 3.7941, "step": 516 }, { "epoch": 0.1, "learning_rate": 2.9732888465204958e-06, "loss": 3.8286, "step": 517 }, { "epoch": 0.1, "learning_rate": 2.9732316491897046e-06, "loss": 3.811, "step": 518 }, { "epoch": 0.1, "learning_rate": 2.9731744518589134e-06, "loss": 3.8168, "step": 519 }, { "epoch": 0.1, "learning_rate": 2.973117254528122e-06, "loss": 3.8118, "step": 520 }, { "epoch": 0.1, "learning_rate": 2.973060057197331e-06, "loss": 3.7819, "step": 521 }, { "epoch": 0.1, "learning_rate": 2.9730028598665397e-06, "loss": 3.8216, "step": 522 }, { "epoch": 0.1, "learning_rate": 2.972945662535748e-06, "loss": 3.8142, "step": 523 }, { "epoch": 0.1, "learning_rate": 2.9728884652049573e-06, "loss": 3.8461, "step": 524 }, { "epoch": 0.1, "learning_rate": 2.972831267874166e-06, "loss": 3.7969, "step": 525 }, { "epoch": 0.1, "learning_rate": 2.9727740705433745e-06, "loss": 3.6417, "step": 526 }, { "epoch": 0.1, "learning_rate": 2.9727168732125837e-06, "loss": 3.7798, "step": 527 }, { "epoch": 0.1, "learning_rate": 2.972659675881792e-06, "loss": 3.81, "step": 528 }, { "epoch": 0.1, "learning_rate": 2.972602478551001e-06, "loss": 3.8437, "step": 529 }, { "epoch": 0.1, "learning_rate": 2.9725452812202097e-06, "loss": 3.7962, "step": 530 }, { "epoch": 0.1, "learning_rate": 2.9724880838894185e-06, "loss": 3.8301, "step": 531 }, { "epoch": 0.1, "learning_rate": 2.9724308865586277e-06, "loss": 3.8964, "step": 532 }, { "epoch": 0.1, "learning_rate": 2.972373689227836e-06, "loss": 3.7596, "step": 533 }, { "epoch": 0.1, "learning_rate": 2.972316491897045e-06, "loss": 3.7809, "step": 534 }, { "epoch": 0.1, "learning_rate": 2.9722592945662536e-06, "loss": 3.8611, "step": 535 }, { "epoch": 0.1, "learning_rate": 2.9722020972354624e-06, "loss": 3.8144, "step": 536 }, { "epoch": 0.1, "learning_rate": 2.9721448999046712e-06, "loss": 3.8206, "step": 537 }, { "epoch": 0.1, "learning_rate": 2.97208770257388e-06, "loss": 3.8105, "step": 538 }, { "epoch": 0.1, "learning_rate": 2.972030505243089e-06, "loss": 3.8347, "step": 539 }, { "epoch": 0.1, "learning_rate": 2.971973307912297e-06, "loss": 3.7494, "step": 540 }, { "epoch": 0.1, "learning_rate": 2.9719161105815064e-06, "loss": 3.6991, "step": 541 }, { "epoch": 0.1, "learning_rate": 2.971858913250715e-06, "loss": 3.7583, "step": 542 }, { "epoch": 0.1, "learning_rate": 2.9718017159199236e-06, "loss": 3.8131, "step": 543 }, { "epoch": 0.1, "learning_rate": 2.9717445185891328e-06, "loss": 3.7776, "step": 544 }, { "epoch": 0.1, "learning_rate": 2.971687321258341e-06, "loss": 3.7663, "step": 545 }, { "epoch": 0.1, "learning_rate": 2.97163012392755e-06, "loss": 3.7979, "step": 546 }, { "epoch": 0.1, "learning_rate": 2.9715729265967587e-06, "loss": 3.7725, "step": 547 }, { "epoch": 0.1, "learning_rate": 2.9715157292659675e-06, "loss": 3.753, "step": 548 }, { "epoch": 0.1, "learning_rate": 2.9714585319351768e-06, "loss": 3.7781, "step": 549 }, { "epoch": 0.1, "learning_rate": 2.971401334604385e-06, "loss": 3.7322, "step": 550 }, { "epoch": 0.1, "learning_rate": 2.971344137273594e-06, "loss": 3.834, "step": 551 }, { "epoch": 0.11, "learning_rate": 2.9712869399428027e-06, "loss": 3.8309, "step": 552 }, { "epoch": 0.11, "learning_rate": 2.9712297426120115e-06, "loss": 3.8182, "step": 553 }, { "epoch": 0.11, "learning_rate": 2.9711725452812203e-06, "loss": 3.8555, "step": 554 }, { "epoch": 0.11, "learning_rate": 2.971115347950429e-06, "loss": 3.7921, "step": 555 }, { "epoch": 0.11, "learning_rate": 2.971058150619638e-06, "loss": 3.8682, "step": 556 }, { "epoch": 0.11, "learning_rate": 2.9710009532888463e-06, "loss": 3.7719, "step": 557 }, { "epoch": 0.11, "learning_rate": 2.9709437559580555e-06, "loss": 3.7348, "step": 558 }, { "epoch": 0.11, "learning_rate": 2.9708865586272643e-06, "loss": 3.7962, "step": 559 }, { "epoch": 0.11, "learning_rate": 2.9708293612964726e-06, "loss": 3.7676, "step": 560 }, { "epoch": 0.11, "learning_rate": 2.970772163965682e-06, "loss": 3.7745, "step": 561 }, { "epoch": 0.11, "learning_rate": 2.9707149666348902e-06, "loss": 3.8301, "step": 562 }, { "epoch": 0.11, "learning_rate": 2.9706577693040994e-06, "loss": 3.811, "step": 563 }, { "epoch": 0.11, "learning_rate": 2.9706005719733082e-06, "loss": 3.7194, "step": 564 }, { "epoch": 0.11, "learning_rate": 2.9705433746425166e-06, "loss": 3.7975, "step": 565 }, { "epoch": 0.11, "learning_rate": 2.970486177311726e-06, "loss": 3.764, "step": 566 }, { "epoch": 0.11, "learning_rate": 2.970428979980934e-06, "loss": 3.7875, "step": 567 }, { "epoch": 0.11, "learning_rate": 2.970371782650143e-06, "loss": 3.7047, "step": 568 }, { "epoch": 0.11, "learning_rate": 2.9703145853193518e-06, "loss": 3.7292, "step": 569 }, { "epoch": 0.11, "learning_rate": 2.9702573879885606e-06, "loss": 3.7504, "step": 570 }, { "epoch": 0.11, "learning_rate": 2.9702001906577694e-06, "loss": 3.7045, "step": 571 }, { "epoch": 0.11, "learning_rate": 2.970142993326978e-06, "loss": 3.7907, "step": 572 }, { "epoch": 0.11, "learning_rate": 2.970085795996187e-06, "loss": 3.797, "step": 573 }, { "epoch": 0.11, "learning_rate": 2.9700285986653958e-06, "loss": 3.8288, "step": 574 }, { "epoch": 0.11, "learning_rate": 2.9699714013346045e-06, "loss": 3.6926, "step": 575 }, { "epoch": 0.11, "learning_rate": 2.9699142040038133e-06, "loss": 3.8394, "step": 576 }, { "epoch": 0.11, "learning_rate": 2.9698570066730217e-06, "loss": 3.6418, "step": 577 }, { "epoch": 0.11, "learning_rate": 2.969799809342231e-06, "loss": 3.8276, "step": 578 }, { "epoch": 0.11, "learning_rate": 2.9697426120114393e-06, "loss": 3.6839, "step": 579 }, { "epoch": 0.11, "learning_rate": 2.9696854146806485e-06, "loss": 3.7218, "step": 580 }, { "epoch": 0.11, "learning_rate": 2.9696282173498573e-06, "loss": 3.738, "step": 581 }, { "epoch": 0.11, "learning_rate": 2.9695710200190657e-06, "loss": 3.7356, "step": 582 }, { "epoch": 0.11, "learning_rate": 2.969513822688275e-06, "loss": 3.9355, "step": 583 }, { "epoch": 0.11, "learning_rate": 2.9694566253574833e-06, "loss": 3.7842, "step": 584 }, { "epoch": 0.11, "learning_rate": 2.969399428026692e-06, "loss": 3.7995, "step": 585 }, { "epoch": 0.11, "learning_rate": 2.969342230695901e-06, "loss": 3.8899, "step": 586 }, { "epoch": 0.11, "learning_rate": 2.9692850333651097e-06, "loss": 3.7705, "step": 587 }, { "epoch": 0.11, "learning_rate": 2.9692278360343184e-06, "loss": 3.7538, "step": 588 }, { "epoch": 0.11, "learning_rate": 2.9691706387035272e-06, "loss": 3.7872, "step": 589 }, { "epoch": 0.11, "learning_rate": 2.969113441372736e-06, "loss": 3.8535, "step": 590 }, { "epoch": 0.11, "learning_rate": 2.969056244041945e-06, "loss": 3.8514, "step": 591 }, { "epoch": 0.11, "learning_rate": 2.9689990467111536e-06, "loss": 3.8475, "step": 592 }, { "epoch": 0.11, "learning_rate": 2.9689418493803624e-06, "loss": 3.7453, "step": 593 }, { "epoch": 0.11, "learning_rate": 2.9688846520495708e-06, "loss": 3.8397, "step": 594 }, { "epoch": 0.11, "learning_rate": 2.96882745471878e-06, "loss": 3.7883, "step": 595 }, { "epoch": 0.11, "learning_rate": 2.9687702573879884e-06, "loss": 3.7095, "step": 596 }, { "epoch": 0.11, "learning_rate": 2.9687130600571976e-06, "loss": 3.6727, "step": 597 }, { "epoch": 0.11, "learning_rate": 2.9686558627264064e-06, "loss": 3.7675, "step": 598 }, { "epoch": 0.11, "learning_rate": 2.9685986653956148e-06, "loss": 3.7711, "step": 599 }, { "epoch": 0.11, "learning_rate": 2.968541468064824e-06, "loss": 3.7889, "step": 600 }, { "epoch": 0.11, "learning_rate": 2.9684842707340323e-06, "loss": 3.9116, "step": 601 }, { "epoch": 0.11, "learning_rate": 2.968427073403241e-06, "loss": 3.7964, "step": 602 }, { "epoch": 0.11, "learning_rate": 2.96836987607245e-06, "loss": 3.7681, "step": 603 }, { "epoch": 0.12, "learning_rate": 2.9683126787416587e-06, "loss": 3.738, "step": 604 }, { "epoch": 0.12, "learning_rate": 2.9682554814108675e-06, "loss": 3.7654, "step": 605 }, { "epoch": 0.12, "learning_rate": 2.9681982840800763e-06, "loss": 3.7623, "step": 606 }, { "epoch": 0.12, "learning_rate": 2.968141086749285e-06, "loss": 3.6783, "step": 607 }, { "epoch": 0.12, "learning_rate": 2.968083889418494e-06, "loss": 3.7238, "step": 608 }, { "epoch": 0.12, "learning_rate": 2.9680266920877027e-06, "loss": 3.887, "step": 609 }, { "epoch": 0.12, "learning_rate": 2.9679694947569115e-06, "loss": 3.7598, "step": 610 }, { "epoch": 0.12, "learning_rate": 2.96791229742612e-06, "loss": 3.9036, "step": 611 }, { "epoch": 0.12, "learning_rate": 2.967855100095329e-06, "loss": 3.7107, "step": 612 }, { "epoch": 0.12, "learning_rate": 2.967797902764538e-06, "loss": 3.8131, "step": 613 }, { "epoch": 0.12, "learning_rate": 2.9677407054337467e-06, "loss": 3.8267, "step": 614 }, { "epoch": 0.12, "learning_rate": 2.9676835081029555e-06, "loss": 3.7773, "step": 615 }, { "epoch": 0.12, "learning_rate": 2.967626310772164e-06, "loss": 3.7739, "step": 616 }, { "epoch": 0.12, "learning_rate": 2.967569113441373e-06, "loss": 3.7783, "step": 617 }, { "epoch": 0.12, "learning_rate": 2.9675119161105814e-06, "loss": 3.8361, "step": 618 }, { "epoch": 0.12, "learning_rate": 2.96745471877979e-06, "loss": 3.7932, "step": 619 }, { "epoch": 0.12, "learning_rate": 2.9673975214489994e-06, "loss": 3.6646, "step": 620 }, { "epoch": 0.12, "learning_rate": 2.967340324118208e-06, "loss": 3.8212, "step": 621 }, { "epoch": 0.12, "learning_rate": 2.9672831267874166e-06, "loss": 3.7223, "step": 622 }, { "epoch": 0.12, "learning_rate": 2.9672259294566254e-06, "loss": 3.8491, "step": 623 }, { "epoch": 0.12, "learning_rate": 2.967168732125834e-06, "loss": 3.818, "step": 624 }, { "epoch": 0.12, "learning_rate": 2.967111534795043e-06, "loss": 3.8019, "step": 625 }, { "epoch": 0.12, "learning_rate": 2.9670543374642518e-06, "loss": 3.7437, "step": 626 }, { "epoch": 0.12, "learning_rate": 2.9669971401334606e-06, "loss": 3.7832, "step": 627 }, { "epoch": 0.12, "learning_rate": 2.9669399428026694e-06, "loss": 3.7744, "step": 628 }, { "epoch": 0.12, "learning_rate": 2.966882745471878e-06, "loss": 3.7861, "step": 629 }, { "epoch": 0.12, "learning_rate": 2.966825548141087e-06, "loss": 3.8399, "step": 630 }, { "epoch": 0.12, "learning_rate": 2.9667683508102957e-06, "loss": 3.8565, "step": 631 }, { "epoch": 0.12, "learning_rate": 2.9667111534795045e-06, "loss": 3.8148, "step": 632 }, { "epoch": 0.12, "learning_rate": 2.966653956148713e-06, "loss": 3.8197, "step": 633 }, { "epoch": 0.12, "learning_rate": 2.966596758817922e-06, "loss": 3.7718, "step": 634 }, { "epoch": 0.12, "learning_rate": 2.9665395614871305e-06, "loss": 3.7643, "step": 635 }, { "epoch": 0.12, "learning_rate": 2.9664823641563393e-06, "loss": 3.7683, "step": 636 }, { "epoch": 0.12, "learning_rate": 2.9664251668255485e-06, "loss": 3.8166, "step": 637 }, { "epoch": 0.12, "learning_rate": 2.966367969494757e-06, "loss": 3.8133, "step": 638 }, { "epoch": 0.12, "learning_rate": 2.9663107721639657e-06, "loss": 3.8286, "step": 639 }, { "epoch": 0.12, "learning_rate": 2.9662535748331745e-06, "loss": 3.7951, "step": 640 }, { "epoch": 0.12, "learning_rate": 2.9661963775023832e-06, "loss": 3.6838, "step": 641 }, { "epoch": 0.12, "learning_rate": 2.966139180171592e-06, "loss": 3.7587, "step": 642 }, { "epoch": 0.12, "learning_rate": 2.966081982840801e-06, "loss": 3.7554, "step": 643 }, { "epoch": 0.12, "learning_rate": 2.9660247855100096e-06, "loss": 3.8009, "step": 644 }, { "epoch": 0.12, "learning_rate": 2.9659675881792184e-06, "loss": 3.8026, "step": 645 }, { "epoch": 0.12, "learning_rate": 2.9659103908484272e-06, "loss": 3.8435, "step": 646 }, { "epoch": 0.12, "learning_rate": 2.965853193517636e-06, "loss": 3.7406, "step": 647 }, { "epoch": 0.12, "learning_rate": 2.965795996186845e-06, "loss": 3.8649, "step": 648 }, { "epoch": 0.12, "learning_rate": 2.9657387988560536e-06, "loss": 3.8154, "step": 649 }, { "epoch": 0.12, "learning_rate": 2.965681601525262e-06, "loss": 3.7525, "step": 650 }, { "epoch": 0.12, "learning_rate": 2.965624404194471e-06, "loss": 3.8024, "step": 651 }, { "epoch": 0.12, "learning_rate": 2.96556720686368e-06, "loss": 3.7923, "step": 652 }, { "epoch": 0.12, "learning_rate": 2.9655100095328883e-06, "loss": 3.7582, "step": 653 }, { "epoch": 0.12, "learning_rate": 2.9654528122020976e-06, "loss": 3.7668, "step": 654 }, { "epoch": 0.12, "learning_rate": 2.965395614871306e-06, "loss": 3.8226, "step": 655 }, { "epoch": 0.12, "learning_rate": 2.9653384175405147e-06, "loss": 3.8029, "step": 656 }, { "epoch": 0.13, "learning_rate": 2.9652812202097235e-06, "loss": 3.6569, "step": 657 }, { "epoch": 0.13, "learning_rate": 2.9652240228789323e-06, "loss": 3.8341, "step": 658 }, { "epoch": 0.13, "learning_rate": 2.965166825548141e-06, "loss": 3.7634, "step": 659 }, { "epoch": 0.13, "learning_rate": 2.96510962821735e-06, "loss": 3.7493, "step": 660 }, { "epoch": 0.13, "learning_rate": 2.9650524308865587e-06, "loss": 3.8005, "step": 661 }, { "epoch": 0.13, "learning_rate": 2.9649952335557675e-06, "loss": 3.7232, "step": 662 }, { "epoch": 0.13, "learning_rate": 2.9649380362249763e-06, "loss": 3.7434, "step": 663 }, { "epoch": 0.13, "learning_rate": 2.964880838894185e-06, "loss": 3.8015, "step": 664 }, { "epoch": 0.13, "learning_rate": 2.964823641563394e-06, "loss": 3.7734, "step": 665 }, { "epoch": 0.13, "learning_rate": 2.9647664442326027e-06, "loss": 3.8279, "step": 666 }, { "epoch": 0.13, "learning_rate": 2.964709246901811e-06, "loss": 3.8376, "step": 667 }, { "epoch": 0.13, "learning_rate": 2.9646520495710203e-06, "loss": 3.7868, "step": 668 }, { "epoch": 0.13, "learning_rate": 2.964594852240229e-06, "loss": 3.7768, "step": 669 }, { "epoch": 0.13, "learning_rate": 2.9645376549094374e-06, "loss": 3.7916, "step": 670 }, { "epoch": 0.13, "learning_rate": 2.9644804575786466e-06, "loss": 3.8051, "step": 671 }, { "epoch": 0.13, "learning_rate": 2.964423260247855e-06, "loss": 3.8049, "step": 672 }, { "epoch": 0.13, "learning_rate": 2.964366062917064e-06, "loss": 3.7574, "step": 673 }, { "epoch": 0.13, "learning_rate": 2.9643088655862726e-06, "loss": 3.7372, "step": 674 }, { "epoch": 0.13, "learning_rate": 2.9642516682554814e-06, "loss": 3.8444, "step": 675 }, { "epoch": 0.13, "learning_rate": 2.96419447092469e-06, "loss": 3.8264, "step": 676 }, { "epoch": 0.13, "learning_rate": 2.964137273593899e-06, "loss": 3.8216, "step": 677 }, { "epoch": 0.13, "learning_rate": 2.9640800762631078e-06, "loss": 3.8461, "step": 678 }, { "epoch": 0.13, "learning_rate": 2.9640228789323166e-06, "loss": 3.7291, "step": 679 }, { "epoch": 0.13, "learning_rate": 2.9639656816015254e-06, "loss": 3.6285, "step": 680 }, { "epoch": 0.13, "learning_rate": 2.963908484270734e-06, "loss": 3.7683, "step": 681 }, { "epoch": 0.13, "learning_rate": 2.963851286939943e-06, "loss": 3.739, "step": 682 }, { "epoch": 0.13, "learning_rate": 2.9637940896091517e-06, "loss": 3.8196, "step": 683 }, { "epoch": 0.13, "learning_rate": 2.96373689227836e-06, "loss": 3.8834, "step": 684 }, { "epoch": 0.13, "learning_rate": 2.9636796949475693e-06, "loss": 3.6835, "step": 685 }, { "epoch": 0.13, "learning_rate": 2.963622497616778e-06, "loss": 3.6934, "step": 686 }, { "epoch": 0.13, "learning_rate": 2.9635653002859865e-06, "loss": 3.7732, "step": 687 }, { "epoch": 0.13, "learning_rate": 2.9635081029551957e-06, "loss": 3.7905, "step": 688 }, { "epoch": 0.13, "learning_rate": 2.963450905624404e-06, "loss": 3.7626, "step": 689 }, { "epoch": 0.13, "learning_rate": 2.963393708293613e-06, "loss": 3.7268, "step": 690 }, { "epoch": 0.13, "learning_rate": 2.9633365109628217e-06, "loss": 3.7389, "step": 691 }, { "epoch": 0.13, "learning_rate": 2.9632793136320305e-06, "loss": 3.7151, "step": 692 }, { "epoch": 0.13, "learning_rate": 2.9632221163012397e-06, "loss": 3.69, "step": 693 }, { "epoch": 0.13, "learning_rate": 2.963164918970448e-06, "loss": 3.8207, "step": 694 }, { "epoch": 0.13, "learning_rate": 2.963107721639657e-06, "loss": 3.7701, "step": 695 }, { "epoch": 0.13, "learning_rate": 2.9630505243088656e-06, "loss": 3.7353, "step": 696 }, { "epoch": 0.13, "learning_rate": 2.9629933269780744e-06, "loss": 3.8707, "step": 697 }, { "epoch": 0.13, "learning_rate": 2.9629361296472832e-06, "loss": 3.7649, "step": 698 }, { "epoch": 0.13, "learning_rate": 2.962878932316492e-06, "loss": 3.6928, "step": 699 }, { "epoch": 0.13, "learning_rate": 2.962821734985701e-06, "loss": 3.8112, "step": 700 }, { "epoch": 0.13, "learning_rate": 2.9627645376549096e-06, "loss": 3.8669, "step": 701 }, { "epoch": 0.13, "learning_rate": 2.9627073403241184e-06, "loss": 3.8286, "step": 702 }, { "epoch": 0.13, "learning_rate": 2.962650142993327e-06, "loss": 3.706, "step": 703 }, { "epoch": 0.13, "learning_rate": 2.9625929456625356e-06, "loss": 3.8456, "step": 704 }, { "epoch": 0.13, "learning_rate": 2.9625357483317448e-06, "loss": 3.6205, "step": 705 }, { "epoch": 0.13, "learning_rate": 2.962478551000953e-06, "loss": 3.794, "step": 706 }, { "epoch": 0.13, "learning_rate": 2.962421353670162e-06, "loss": 3.8968, "step": 707 }, { "epoch": 0.13, "learning_rate": 2.962364156339371e-06, "loss": 3.8457, "step": 708 }, { "epoch": 0.14, "learning_rate": 2.9623069590085795e-06, "loss": 3.805, "step": 709 }, { "epoch": 0.14, "learning_rate": 2.9622497616777887e-06, "loss": 3.7633, "step": 710 }, { "epoch": 0.14, "learning_rate": 2.962192564346997e-06, "loss": 3.8084, "step": 711 }, { "epoch": 0.14, "learning_rate": 2.962135367016206e-06, "loss": 3.834, "step": 712 }, { "epoch": 0.14, "learning_rate": 2.9620781696854147e-06, "loss": 3.811, "step": 713 }, { "epoch": 0.14, "learning_rate": 2.9620209723546235e-06, "loss": 3.8673, "step": 714 }, { "epoch": 0.14, "learning_rate": 2.9619637750238323e-06, "loss": 3.8371, "step": 715 }, { "epoch": 0.14, "learning_rate": 2.961906577693041e-06, "loss": 3.8179, "step": 716 }, { "epoch": 0.14, "learning_rate": 2.96184938036225e-06, "loss": 3.7165, "step": 717 }, { "epoch": 0.14, "learning_rate": 2.9617921830314587e-06, "loss": 3.7595, "step": 718 }, { "epoch": 0.14, "learning_rate": 2.9617349857006675e-06, "loss": 3.8025, "step": 719 }, { "epoch": 0.14, "learning_rate": 2.9616777883698763e-06, "loss": 3.761, "step": 720 }, { "epoch": 0.14, "learning_rate": 2.9616205910390846e-06, "loss": 3.7388, "step": 721 }, { "epoch": 0.14, "learning_rate": 2.961563393708294e-06, "loss": 3.7548, "step": 722 }, { "epoch": 0.14, "learning_rate": 2.9615061963775022e-06, "loss": 3.7196, "step": 723 }, { "epoch": 0.14, "learning_rate": 2.961448999046711e-06, "loss": 3.742, "step": 724 }, { "epoch": 0.14, "learning_rate": 2.9613918017159202e-06, "loss": 3.8959, "step": 725 }, { "epoch": 0.14, "learning_rate": 2.9613346043851286e-06, "loss": 3.7904, "step": 726 }, { "epoch": 0.14, "learning_rate": 2.961277407054338e-06, "loss": 3.759, "step": 727 }, { "epoch": 0.14, "learning_rate": 2.961220209723546e-06, "loss": 3.8192, "step": 728 }, { "epoch": 0.14, "learning_rate": 2.961163012392755e-06, "loss": 3.795, "step": 729 }, { "epoch": 0.14, "learning_rate": 2.9611058150619638e-06, "loss": 3.8202, "step": 730 }, { "epoch": 0.14, "learning_rate": 2.9610486177311726e-06, "loss": 3.7694, "step": 731 }, { "epoch": 0.14, "learning_rate": 2.9609914204003814e-06, "loss": 3.7429, "step": 732 }, { "epoch": 0.14, "learning_rate": 2.96093422306959e-06, "loss": 3.7281, "step": 733 }, { "epoch": 0.14, "learning_rate": 2.960877025738799e-06, "loss": 3.7857, "step": 734 }, { "epoch": 0.14, "learning_rate": 2.9608198284080077e-06, "loss": 3.8398, "step": 735 }, { "epoch": 0.14, "learning_rate": 2.9607626310772165e-06, "loss": 3.7175, "step": 736 }, { "epoch": 0.14, "learning_rate": 2.9607054337464253e-06, "loss": 3.6008, "step": 737 }, { "epoch": 0.14, "learning_rate": 2.9606482364156337e-06, "loss": 3.8269, "step": 738 }, { "epoch": 0.14, "learning_rate": 2.960591039084843e-06, "loss": 3.7225, "step": 739 }, { "epoch": 0.14, "learning_rate": 2.9605338417540517e-06, "loss": 3.7704, "step": 740 }, { "epoch": 0.14, "learning_rate": 2.96047664442326e-06, "loss": 3.6994, "step": 741 }, { "epoch": 0.14, "learning_rate": 2.9604194470924693e-06, "loss": 3.6442, "step": 742 }, { "epoch": 0.14, "learning_rate": 2.9603622497616777e-06, "loss": 3.7496, "step": 743 }, { "epoch": 0.14, "learning_rate": 2.960305052430887e-06, "loss": 3.8498, "step": 744 }, { "epoch": 0.14, "learning_rate": 2.9602478551000953e-06, "loss": 3.8499, "step": 745 }, { "epoch": 0.14, "learning_rate": 2.960190657769304e-06, "loss": 3.778, "step": 746 }, { "epoch": 0.14, "learning_rate": 2.9601334604385133e-06, "loss": 3.7088, "step": 747 }, { "epoch": 0.14, "learning_rate": 2.9600762631077216e-06, "loss": 3.6838, "step": 748 }, { "epoch": 0.14, "learning_rate": 2.9600190657769304e-06, "loss": 3.7266, "step": 749 }, { "epoch": 0.14, "learning_rate": 2.9599618684461392e-06, "loss": 3.8478, "step": 750 }, { "epoch": 0.14, "learning_rate": 2.959904671115348e-06, "loss": 3.6952, "step": 751 }, { "epoch": 0.14, "learning_rate": 2.959847473784557e-06, "loss": 3.774, "step": 752 }, { "epoch": 0.14, "learning_rate": 2.9597902764537656e-06, "loss": 3.7769, "step": 753 }, { "epoch": 0.14, "learning_rate": 2.9597330791229744e-06, "loss": 3.8023, "step": 754 }, { "epoch": 0.14, "learning_rate": 2.9596758817921828e-06, "loss": 3.8856, "step": 755 }, { "epoch": 0.14, "learning_rate": 2.959618684461392e-06, "loss": 3.7855, "step": 756 }, { "epoch": 0.14, "learning_rate": 2.9595614871306008e-06, "loss": 3.7806, "step": 757 }, { "epoch": 0.14, "learning_rate": 2.9595042897998096e-06, "loss": 3.8527, "step": 758 }, { "epoch": 0.14, "learning_rate": 2.9594470924690184e-06, "loss": 3.8279, "step": 759 }, { "epoch": 0.14, "learning_rate": 2.9593898951382267e-06, "loss": 3.8039, "step": 760 }, { "epoch": 0.14, "learning_rate": 2.959332697807436e-06, "loss": 3.8077, "step": 761 }, { "epoch": 0.15, "learning_rate": 2.9592755004766443e-06, "loss": 3.7325, "step": 762 }, { "epoch": 0.15, "learning_rate": 2.959218303145853e-06, "loss": 3.766, "step": 763 }, { "epoch": 0.15, "learning_rate": 2.9591611058150623e-06, "loss": 3.824, "step": 764 }, { "epoch": 0.15, "learning_rate": 2.9591039084842707e-06, "loss": 3.8572, "step": 765 }, { "epoch": 0.15, "learning_rate": 2.9590467111534795e-06, "loss": 3.7764, "step": 766 }, { "epoch": 0.15, "learning_rate": 2.9589895138226883e-06, "loss": 3.8657, "step": 767 }, { "epoch": 0.15, "learning_rate": 2.958932316491897e-06, "loss": 3.8743, "step": 768 }, { "epoch": 0.15, "learning_rate": 2.958875119161106e-06, "loss": 3.7203, "step": 769 }, { "epoch": 0.15, "learning_rate": 2.9588179218303147e-06, "loss": 3.7326, "step": 770 }, { "epoch": 0.15, "learning_rate": 2.9587607244995235e-06, "loss": 3.771, "step": 771 }, { "epoch": 0.15, "learning_rate": 2.958703527168732e-06, "loss": 3.8178, "step": 772 }, { "epoch": 0.15, "learning_rate": 2.958646329837941e-06, "loss": 3.7351, "step": 773 }, { "epoch": 0.15, "learning_rate": 2.95858913250715e-06, "loss": 3.8605, "step": 774 }, { "epoch": 0.15, "learning_rate": 2.9585319351763587e-06, "loss": 3.8193, "step": 775 }, { "epoch": 0.15, "learning_rate": 2.9584747378455674e-06, "loss": 3.7385, "step": 776 }, { "epoch": 0.15, "learning_rate": 2.958417540514776e-06, "loss": 3.7229, "step": 777 }, { "epoch": 0.15, "learning_rate": 2.958360343183985e-06, "loss": 3.8221, "step": 778 }, { "epoch": 0.15, "learning_rate": 2.9583031458531934e-06, "loss": 3.6814, "step": 779 }, { "epoch": 0.15, "learning_rate": 2.958245948522402e-06, "loss": 3.7404, "step": 780 }, { "epoch": 0.15, "learning_rate": 2.9581887511916114e-06, "loss": 3.694, "step": 781 }, { "epoch": 0.15, "learning_rate": 2.9581315538608198e-06, "loss": 3.6972, "step": 782 }, { "epoch": 0.15, "learning_rate": 2.9580743565300286e-06, "loss": 3.7869, "step": 783 }, { "epoch": 0.15, "learning_rate": 2.9580171591992374e-06, "loss": 3.8516, "step": 784 }, { "epoch": 0.15, "learning_rate": 2.957959961868446e-06, "loss": 3.8887, "step": 785 }, { "epoch": 0.15, "learning_rate": 2.957902764537655e-06, "loss": 3.8573, "step": 786 }, { "epoch": 0.15, "learning_rate": 2.9578455672068638e-06, "loss": 3.6596, "step": 787 }, { "epoch": 0.15, "learning_rate": 2.9577883698760725e-06, "loss": 3.8217, "step": 788 }, { "epoch": 0.15, "learning_rate": 2.9577311725452813e-06, "loss": 3.7668, "step": 789 }, { "epoch": 0.15, "learning_rate": 2.95767397521449e-06, "loss": 3.8416, "step": 790 }, { "epoch": 0.15, "learning_rate": 2.957616777883699e-06, "loss": 3.7782, "step": 791 }, { "epoch": 0.15, "learning_rate": 2.9575595805529077e-06, "loss": 3.7388, "step": 792 }, { "epoch": 0.15, "learning_rate": 2.9575023832221165e-06, "loss": 3.6341, "step": 793 }, { "epoch": 0.15, "learning_rate": 2.957445185891325e-06, "loss": 3.9046, "step": 794 }, { "epoch": 0.15, "learning_rate": 2.957387988560534e-06, "loss": 3.6879, "step": 795 }, { "epoch": 0.15, "learning_rate": 2.957330791229743e-06, "loss": 3.6916, "step": 796 }, { "epoch": 0.15, "learning_rate": 2.9572735938989513e-06, "loss": 3.7096, "step": 797 }, { "epoch": 0.15, "learning_rate": 2.9572163965681605e-06, "loss": 3.7918, "step": 798 }, { "epoch": 0.15, "learning_rate": 2.957159199237369e-06, "loss": 3.7739, "step": 799 }, { "epoch": 0.15, "learning_rate": 2.9571020019065777e-06, "loss": 3.7878, "step": 800 }, { "epoch": 0.15, "learning_rate": 2.9570448045757864e-06, "loss": 3.825, "step": 801 }, { "epoch": 0.15, "learning_rate": 2.9569876072449952e-06, "loss": 3.6129, "step": 802 }, { "epoch": 0.15, "learning_rate": 2.956930409914204e-06, "loss": 3.8148, "step": 803 }, { "epoch": 0.15, "learning_rate": 2.956873212583413e-06, "loss": 3.8528, "step": 804 }, { "epoch": 0.15, "learning_rate": 2.9568160152526216e-06, "loss": 3.7436, "step": 805 }, { "epoch": 0.15, "learning_rate": 2.9567588179218304e-06, "loss": 3.7575, "step": 806 }, { "epoch": 0.15, "learning_rate": 2.956701620591039e-06, "loss": 3.7202, "step": 807 }, { "epoch": 0.15, "learning_rate": 2.956644423260248e-06, "loss": 3.6457, "step": 808 }, { "epoch": 0.15, "learning_rate": 2.956587225929457e-06, "loss": 3.8583, "step": 809 }, { "epoch": 0.15, "learning_rate": 2.9565300285986656e-06, "loss": 3.7082, "step": 810 }, { "epoch": 0.15, "learning_rate": 2.956472831267874e-06, "loss": 3.798, "step": 811 }, { "epoch": 0.15, "learning_rate": 2.956415633937083e-06, "loss": 3.7761, "step": 812 }, { "epoch": 0.15, "learning_rate": 2.956358436606292e-06, "loss": 3.769, "step": 813 }, { "epoch": 0.16, "learning_rate": 2.9563012392755003e-06, "loss": 3.8694, "step": 814 }, { "epoch": 0.16, "learning_rate": 2.9562440419447096e-06, "loss": 3.7726, "step": 815 }, { "epoch": 0.16, "learning_rate": 2.956186844613918e-06, "loss": 3.7541, "step": 816 }, { "epoch": 0.16, "learning_rate": 2.9561296472831267e-06, "loss": 3.7306, "step": 817 }, { "epoch": 0.16, "learning_rate": 2.9560724499523355e-06, "loss": 3.7755, "step": 818 }, { "epoch": 0.16, "learning_rate": 2.9560152526215443e-06, "loss": 3.694, "step": 819 }, { "epoch": 0.16, "learning_rate": 2.955958055290753e-06, "loss": 3.6665, "step": 820 }, { "epoch": 0.16, "learning_rate": 2.955900857959962e-06, "loss": 3.6162, "step": 821 }, { "epoch": 0.16, "learning_rate": 2.9558436606291707e-06, "loss": 3.6875, "step": 822 }, { "epoch": 0.16, "learning_rate": 2.9557864632983795e-06, "loss": 3.7569, "step": 823 }, { "epoch": 0.16, "learning_rate": 2.9557292659675883e-06, "loss": 3.7746, "step": 824 }, { "epoch": 0.16, "learning_rate": 2.955672068636797e-06, "loss": 3.7981, "step": 825 }, { "epoch": 0.16, "learning_rate": 2.955614871306006e-06, "loss": 3.8265, "step": 826 }, { "epoch": 0.16, "learning_rate": 2.9555576739752147e-06, "loss": 3.7768, "step": 827 }, { "epoch": 0.16, "learning_rate": 2.9555004766444235e-06, "loss": 3.7672, "step": 828 }, { "epoch": 0.16, "learning_rate": 2.9554432793136322e-06, "loss": 3.7061, "step": 829 }, { "epoch": 0.16, "learning_rate": 2.955386081982841e-06, "loss": 3.8155, "step": 830 }, { "epoch": 0.16, "learning_rate": 2.9553288846520494e-06, "loss": 3.7065, "step": 831 }, { "epoch": 0.16, "learning_rate": 2.9552716873212586e-06, "loss": 3.7874, "step": 832 }, { "epoch": 0.16, "learning_rate": 2.955214489990467e-06, "loss": 3.85, "step": 833 }, { "epoch": 0.16, "learning_rate": 2.955157292659676e-06, "loss": 3.8118, "step": 834 }, { "epoch": 0.16, "learning_rate": 2.955100095328885e-06, "loss": 3.7337, "step": 835 }, { "epoch": 0.16, "learning_rate": 2.9550428979980934e-06, "loss": 3.8026, "step": 836 }, { "epoch": 0.16, "learning_rate": 2.954985700667302e-06, "loss": 3.7663, "step": 837 }, { "epoch": 0.16, "learning_rate": 2.954928503336511e-06, "loss": 3.7693, "step": 838 }, { "epoch": 0.16, "learning_rate": 2.9548713060057198e-06, "loss": 3.7177, "step": 839 }, { "epoch": 0.16, "learning_rate": 2.9548141086749286e-06, "loss": 3.6557, "step": 840 }, { "epoch": 0.16, "learning_rate": 2.9547569113441374e-06, "loss": 3.8407, "step": 841 }, { "epoch": 0.16, "learning_rate": 2.954699714013346e-06, "loss": 3.7616, "step": 842 }, { "epoch": 0.16, "learning_rate": 2.954642516682555e-06, "loss": 3.6564, "step": 843 }, { "epoch": 0.16, "learning_rate": 2.9545853193517637e-06, "loss": 3.7412, "step": 844 }, { "epoch": 0.16, "learning_rate": 2.9545281220209725e-06, "loss": 3.6949, "step": 845 }, { "epoch": 0.16, "learning_rate": 2.9544709246901813e-06, "loss": 3.7214, "step": 846 }, { "epoch": 0.16, "learning_rate": 2.95441372735939e-06, "loss": 3.7583, "step": 847 }, { "epoch": 0.16, "learning_rate": 2.9543565300285985e-06, "loss": 3.7469, "step": 848 }, { "epoch": 0.16, "learning_rate": 2.9542993326978077e-06, "loss": 3.7665, "step": 849 }, { "epoch": 0.16, "learning_rate": 2.954242135367016e-06, "loss": 3.8291, "step": 850 }, { "epoch": 0.16, "learning_rate": 2.954184938036225e-06, "loss": 3.8768, "step": 851 }, { "epoch": 0.16, "learning_rate": 2.954127740705434e-06, "loss": 3.7376, "step": 852 }, { "epoch": 0.16, "learning_rate": 2.9540705433746425e-06, "loss": 3.7389, "step": 853 }, { "epoch": 0.16, "learning_rate": 2.9540133460438512e-06, "loss": 3.6861, "step": 854 }, { "epoch": 0.16, "learning_rate": 2.95395614871306e-06, "loss": 3.7655, "step": 855 }, { "epoch": 0.16, "learning_rate": 2.953898951382269e-06, "loss": 3.8453, "step": 856 }, { "epoch": 0.16, "learning_rate": 2.9538417540514776e-06, "loss": 3.7669, "step": 857 }, { "epoch": 0.16, "learning_rate": 2.9537845567206864e-06, "loss": 3.7838, "step": 858 }, { "epoch": 0.16, "learning_rate": 2.9537273593898952e-06, "loss": 3.8357, "step": 859 }, { "epoch": 0.16, "learning_rate": 2.953670162059104e-06, "loss": 3.6964, "step": 860 }, { "epoch": 0.16, "learning_rate": 2.953612964728313e-06, "loss": 3.8145, "step": 861 }, { "epoch": 0.16, "learning_rate": 2.9535557673975216e-06, "loss": 3.5886, "step": 862 }, { "epoch": 0.16, "learning_rate": 2.9534985700667304e-06, "loss": 3.7692, "step": 863 }, { "epoch": 0.16, "learning_rate": 2.953441372735939e-06, "loss": 3.7698, "step": 864 }, { "epoch": 0.16, "learning_rate": 2.9533841754051476e-06, "loss": 3.8602, "step": 865 }, { "epoch": 0.16, "learning_rate": 2.9533269780743568e-06, "loss": 3.7941, "step": 866 }, { "epoch": 0.17, "learning_rate": 2.953269780743565e-06, "loss": 3.6565, "step": 867 }, { "epoch": 0.17, "learning_rate": 2.953212583412774e-06, "loss": 3.6141, "step": 868 }, { "epoch": 0.17, "learning_rate": 2.953155386081983e-06, "loss": 3.7684, "step": 869 }, { "epoch": 0.17, "learning_rate": 2.9530981887511915e-06, "loss": 3.7926, "step": 870 }, { "epoch": 0.17, "learning_rate": 2.9530409914204003e-06, "loss": 3.8377, "step": 871 }, { "epoch": 0.17, "learning_rate": 2.952983794089609e-06, "loss": 3.6911, "step": 872 }, { "epoch": 0.17, "learning_rate": 2.952926596758818e-06, "loss": 3.7173, "step": 873 }, { "epoch": 0.17, "learning_rate": 2.952869399428027e-06, "loss": 3.801, "step": 874 }, { "epoch": 0.17, "learning_rate": 2.9528122020972355e-06, "loss": 3.7748, "step": 875 }, { "epoch": 0.17, "learning_rate": 2.9527550047664443e-06, "loss": 3.7654, "step": 876 }, { "epoch": 0.17, "learning_rate": 2.952697807435653e-06, "loss": 3.7718, "step": 877 }, { "epoch": 0.17, "learning_rate": 2.952640610104862e-06, "loss": 3.7756, "step": 878 }, { "epoch": 0.17, "learning_rate": 2.9525834127740707e-06, "loss": 3.8811, "step": 879 }, { "epoch": 0.17, "learning_rate": 2.9525262154432795e-06, "loss": 3.7243, "step": 880 }, { "epoch": 0.17, "learning_rate": 2.9524690181124883e-06, "loss": 3.7688, "step": 881 }, { "epoch": 0.17, "learning_rate": 2.9524118207816966e-06, "loss": 3.73, "step": 882 }, { "epoch": 0.17, "learning_rate": 2.952354623450906e-06, "loss": 3.7472, "step": 883 }, { "epoch": 0.17, "learning_rate": 2.9522974261201146e-06, "loss": 3.6866, "step": 884 }, { "epoch": 0.17, "learning_rate": 2.952240228789323e-06, "loss": 3.679, "step": 885 }, { "epoch": 0.17, "learning_rate": 2.9521830314585322e-06, "loss": 3.7839, "step": 886 }, { "epoch": 0.17, "learning_rate": 2.9521258341277406e-06, "loss": 3.823, "step": 887 }, { "epoch": 0.17, "learning_rate": 2.95206863679695e-06, "loss": 3.7425, "step": 888 }, { "epoch": 0.17, "learning_rate": 2.952011439466158e-06, "loss": 3.8503, "step": 889 }, { "epoch": 0.17, "learning_rate": 2.951954242135367e-06, "loss": 3.7177, "step": 890 }, { "epoch": 0.17, "learning_rate": 2.951897044804576e-06, "loss": 3.7011, "step": 891 }, { "epoch": 0.17, "learning_rate": 2.9518398474737846e-06, "loss": 3.7293, "step": 892 }, { "epoch": 0.17, "learning_rate": 2.9517826501429934e-06, "loss": 3.757, "step": 893 }, { "epoch": 0.17, "learning_rate": 2.951725452812202e-06, "loss": 3.6787, "step": 894 }, { "epoch": 0.17, "learning_rate": 2.951668255481411e-06, "loss": 3.7074, "step": 895 }, { "epoch": 0.17, "learning_rate": 2.9516110581506197e-06, "loss": 3.8252, "step": 896 }, { "epoch": 0.17, "learning_rate": 2.9515538608198285e-06, "loss": 3.5581, "step": 897 }, { "epoch": 0.17, "learning_rate": 2.9514966634890373e-06, "loss": 3.8109, "step": 898 }, { "epoch": 0.17, "learning_rate": 2.9514394661582457e-06, "loss": 3.7842, "step": 899 }, { "epoch": 0.17, "learning_rate": 2.951382268827455e-06, "loss": 3.804, "step": 900 }, { "epoch": 0.17, "learning_rate": 2.9513250714966637e-06, "loss": 3.6304, "step": 901 }, { "epoch": 0.17, "learning_rate": 2.951267874165872e-06, "loss": 3.7867, "step": 902 }, { "epoch": 0.17, "learning_rate": 2.9512106768350813e-06, "loss": 3.748, "step": 903 }, { "epoch": 0.17, "learning_rate": 2.9511534795042897e-06, "loss": 3.7636, "step": 904 }, { "epoch": 0.17, "learning_rate": 2.951096282173499e-06, "loss": 3.5956, "step": 905 }, { "epoch": 0.17, "learning_rate": 2.9510390848427073e-06, "loss": 3.8803, "step": 906 }, { "epoch": 0.17, "learning_rate": 2.950981887511916e-06, "loss": 3.8514, "step": 907 }, { "epoch": 0.17, "learning_rate": 2.9509246901811253e-06, "loss": 3.8179, "step": 908 }, { "epoch": 0.17, "learning_rate": 2.9508674928503336e-06, "loss": 3.6552, "step": 909 }, { "epoch": 0.17, "learning_rate": 2.9508102955195424e-06, "loss": 3.7512, "step": 910 }, { "epoch": 0.17, "learning_rate": 2.9507530981887512e-06, "loss": 3.7011, "step": 911 }, { "epoch": 0.17, "learning_rate": 2.95069590085796e-06, "loss": 3.7382, "step": 912 }, { "epoch": 0.17, "learning_rate": 2.950638703527169e-06, "loss": 3.8666, "step": 913 }, { "epoch": 0.17, "learning_rate": 2.9505815061963776e-06, "loss": 3.558, "step": 914 }, { "epoch": 0.17, "learning_rate": 2.9505243088655864e-06, "loss": 3.8116, "step": 915 }, { "epoch": 0.17, "learning_rate": 2.9504671115347948e-06, "loss": 3.8128, "step": 916 }, { "epoch": 0.17, "learning_rate": 2.950409914204004e-06, "loss": 3.8017, "step": 917 }, { "epoch": 0.17, "learning_rate": 2.9503527168732128e-06, "loss": 3.6966, "step": 918 }, { "epoch": 0.18, "learning_rate": 2.950295519542421e-06, "loss": 3.8309, "step": 919 }, { "epoch": 0.18, "learning_rate": 2.9502383222116304e-06, "loss": 3.7244, "step": 920 }, { "epoch": 0.18, "learning_rate": 2.9501811248808387e-06, "loss": 3.9055, "step": 921 }, { "epoch": 0.18, "learning_rate": 2.950123927550048e-06, "loss": 3.751, "step": 922 }, { "epoch": 0.18, "learning_rate": 2.9500667302192568e-06, "loss": 3.7481, "step": 923 }, { "epoch": 0.18, "learning_rate": 2.950009532888465e-06, "loss": 3.9002, "step": 924 }, { "epoch": 0.18, "learning_rate": 2.9499523355576743e-06, "loss": 3.8451, "step": 925 }, { "epoch": 0.18, "learning_rate": 2.9498951382268827e-06, "loss": 3.497, "step": 926 }, { "epoch": 0.18, "learning_rate": 2.9498379408960915e-06, "loss": 3.7618, "step": 927 }, { "epoch": 0.18, "learning_rate": 2.9497807435653003e-06, "loss": 3.8327, "step": 928 }, { "epoch": 0.18, "learning_rate": 2.949723546234509e-06, "loss": 3.8199, "step": 929 }, { "epoch": 0.18, "learning_rate": 2.949666348903718e-06, "loss": 3.7316, "step": 930 }, { "epoch": 0.18, "learning_rate": 2.9496091515729267e-06, "loss": 3.7577, "step": 931 }, { "epoch": 0.18, "learning_rate": 2.9495519542421355e-06, "loss": 3.7225, "step": 932 }, { "epoch": 0.18, "learning_rate": 2.9494947569113443e-06, "loss": 3.6464, "step": 933 }, { "epoch": 0.18, "learning_rate": 2.949437559580553e-06, "loss": 3.8716, "step": 934 }, { "epoch": 0.18, "learning_rate": 2.949380362249762e-06, "loss": 3.774, "step": 935 }, { "epoch": 0.18, "learning_rate": 2.9493231649189702e-06, "loss": 3.7376, "step": 936 }, { "epoch": 0.18, "learning_rate": 2.9492659675881794e-06, "loss": 3.8261, "step": 937 }, { "epoch": 0.18, "learning_rate": 2.949208770257388e-06, "loss": 3.7253, "step": 938 }, { "epoch": 0.18, "learning_rate": 2.949151572926597e-06, "loss": 3.8319, "step": 939 }, { "epoch": 0.18, "learning_rate": 2.949094375595806e-06, "loss": 3.6853, "step": 940 }, { "epoch": 0.18, "learning_rate": 2.949037178265014e-06, "loss": 3.7129, "step": 941 }, { "epoch": 0.18, "learning_rate": 2.9489799809342234e-06, "loss": 3.6622, "step": 942 }, { "epoch": 0.18, "learning_rate": 2.9489227836034318e-06, "loss": 3.5679, "step": 943 }, { "epoch": 0.18, "learning_rate": 2.9488655862726406e-06, "loss": 3.7869, "step": 944 }, { "epoch": 0.18, "learning_rate": 2.9488083889418494e-06, "loss": 3.8251, "step": 945 }, { "epoch": 0.18, "learning_rate": 2.948751191611058e-06, "loss": 3.8031, "step": 946 }, { "epoch": 0.18, "learning_rate": 2.948693994280267e-06, "loss": 3.5222, "step": 947 }, { "epoch": 0.18, "learning_rate": 2.9486367969494757e-06, "loss": 3.6658, "step": 948 }, { "epoch": 0.18, "learning_rate": 2.9485795996186845e-06, "loss": 3.6412, "step": 949 }, { "epoch": 0.18, "learning_rate": 2.9485224022878933e-06, "loss": 3.7113, "step": 950 }, { "epoch": 0.18, "learning_rate": 2.948465204957102e-06, "loss": 3.7706, "step": 951 }, { "epoch": 0.18, "learning_rate": 2.948408007626311e-06, "loss": 3.6081, "step": 952 }, { "epoch": 0.18, "learning_rate": 2.9483508102955197e-06, "loss": 3.7093, "step": 953 }, { "epoch": 0.18, "learning_rate": 2.9482936129647285e-06, "loss": 3.7024, "step": 954 }, { "epoch": 0.18, "learning_rate": 2.948236415633937e-06, "loss": 3.7691, "step": 955 }, { "epoch": 0.18, "learning_rate": 2.948179218303146e-06, "loss": 3.7762, "step": 956 }, { "epoch": 0.18, "learning_rate": 2.948122020972355e-06, "loss": 3.7673, "step": 957 }, { "epoch": 0.18, "learning_rate": 2.9480648236415633e-06, "loss": 3.8829, "step": 958 }, { "epoch": 0.18, "learning_rate": 2.9480076263107725e-06, "loss": 3.8306, "step": 959 }, { "epoch": 0.18, "learning_rate": 2.947950428979981e-06, "loss": 3.6834, "step": 960 }, { "epoch": 0.18, "learning_rate": 2.9478932316491896e-06, "loss": 3.7841, "step": 961 }, { "epoch": 0.18, "learning_rate": 2.947836034318399e-06, "loss": 3.8242, "step": 962 }, { "epoch": 0.18, "learning_rate": 2.9477788369876072e-06, "loss": 3.8282, "step": 963 }, { "epoch": 0.18, "learning_rate": 2.947721639656816e-06, "loss": 3.7686, "step": 964 }, { "epoch": 0.18, "learning_rate": 2.947664442326025e-06, "loss": 3.7518, "step": 965 }, { "epoch": 0.18, "learning_rate": 2.9476072449952336e-06, "loss": 3.835, "step": 966 }, { "epoch": 0.18, "learning_rate": 2.9475500476644424e-06, "loss": 3.7985, "step": 967 }, { "epoch": 0.18, "learning_rate": 2.947492850333651e-06, "loss": 3.6384, "step": 968 }, { "epoch": 0.18, "learning_rate": 2.94743565300286e-06, "loss": 3.7448, "step": 969 }, { "epoch": 0.18, "learning_rate": 2.947378455672069e-06, "loss": 3.8579, "step": 970 }, { "epoch": 0.18, "learning_rate": 2.9473212583412776e-06, "loss": 3.6909, "step": 971 }, { "epoch": 0.19, "learning_rate": 2.9472640610104864e-06, "loss": 3.7093, "step": 972 }, { "epoch": 0.19, "learning_rate": 2.947206863679695e-06, "loss": 3.7959, "step": 973 }, { "epoch": 0.19, "learning_rate": 2.947149666348904e-06, "loss": 3.7117, "step": 974 }, { "epoch": 0.19, "learning_rate": 2.9470924690181123e-06, "loss": 3.6753, "step": 975 }, { "epoch": 0.19, "learning_rate": 2.9470352716873216e-06, "loss": 3.8085, "step": 976 }, { "epoch": 0.19, "learning_rate": 2.94697807435653e-06, "loss": 3.8017, "step": 977 }, { "epoch": 0.19, "learning_rate": 2.9469208770257387e-06, "loss": 3.8313, "step": 978 }, { "epoch": 0.19, "learning_rate": 2.946863679694948e-06, "loss": 3.7524, "step": 979 }, { "epoch": 0.19, "learning_rate": 2.9468064823641563e-06, "loss": 3.7989, "step": 980 }, { "epoch": 0.19, "learning_rate": 2.946749285033365e-06, "loss": 3.865, "step": 981 }, { "epoch": 0.19, "learning_rate": 2.946692087702574e-06, "loss": 3.6332, "step": 982 }, { "epoch": 0.19, "learning_rate": 2.9466348903717827e-06, "loss": 3.6148, "step": 983 }, { "epoch": 0.19, "learning_rate": 2.9465776930409915e-06, "loss": 3.6614, "step": 984 }, { "epoch": 0.19, "learning_rate": 2.9465204957102003e-06, "loss": 3.9075, "step": 985 }, { "epoch": 0.19, "learning_rate": 2.946463298379409e-06, "loss": 3.7643, "step": 986 }, { "epoch": 0.19, "learning_rate": 2.946406101048618e-06, "loss": 3.7553, "step": 987 }, { "epoch": 0.19, "learning_rate": 2.9463489037178267e-06, "loss": 3.7456, "step": 988 }, { "epoch": 0.19, "learning_rate": 2.9462917063870354e-06, "loss": 3.7496, "step": 989 }, { "epoch": 0.19, "learning_rate": 2.9462345090562442e-06, "loss": 3.789, "step": 990 }, { "epoch": 0.19, "learning_rate": 2.946177311725453e-06, "loss": 3.7779, "step": 991 }, { "epoch": 0.19, "learning_rate": 2.9461201143946614e-06, "loss": 3.7452, "step": 992 }, { "epoch": 0.19, "learning_rate": 2.9460629170638706e-06, "loss": 3.7031, "step": 993 }, { "epoch": 0.19, "learning_rate": 2.946005719733079e-06, "loss": 3.6622, "step": 994 }, { "epoch": 0.19, "learning_rate": 2.9459485224022878e-06, "loss": 3.687, "step": 995 }, { "epoch": 0.19, "learning_rate": 2.945891325071497e-06, "loss": 3.7071, "step": 996 }, { "epoch": 0.19, "learning_rate": 2.9458341277407054e-06, "loss": 3.8103, "step": 997 }, { "epoch": 0.19, "learning_rate": 2.945776930409914e-06, "loss": 3.7924, "step": 998 }, { "epoch": 0.19, "learning_rate": 2.945719733079123e-06, "loss": 3.7409, "step": 999 }, { "epoch": 0.19, "learning_rate": 2.9456625357483318e-06, "loss": 3.7323, "step": 1000 }, { "epoch": 0.19, "learning_rate": 2.9456053384175406e-06, "loss": 3.8082, "step": 1001 }, { "epoch": 0.19, "learning_rate": 2.9455481410867493e-06, "loss": 3.7123, "step": 1002 }, { "epoch": 0.19, "learning_rate": 2.945490943755958e-06, "loss": 3.8131, "step": 1003 }, { "epoch": 0.19, "learning_rate": 2.945433746425167e-06, "loss": 3.8459, "step": 1004 }, { "epoch": 0.19, "learning_rate": 2.9453765490943757e-06, "loss": 3.8184, "step": 1005 }, { "epoch": 0.19, "learning_rate": 2.9453193517635845e-06, "loss": 3.809, "step": 1006 }, { "epoch": 0.19, "learning_rate": 2.9452621544327933e-06, "loss": 3.7278, "step": 1007 }, { "epoch": 0.19, "learning_rate": 2.945204957102002e-06, "loss": 3.6534, "step": 1008 }, { "epoch": 0.19, "learning_rate": 2.9451477597712105e-06, "loss": 3.8108, "step": 1009 }, { "epoch": 0.19, "learning_rate": 2.9450905624404197e-06, "loss": 3.7502, "step": 1010 }, { "epoch": 0.19, "learning_rate": 2.9450333651096285e-06, "loss": 3.6357, "step": 1011 }, { "epoch": 0.19, "learning_rate": 2.944976167778837e-06, "loss": 3.8693, "step": 1012 }, { "epoch": 0.19, "learning_rate": 2.944918970448046e-06, "loss": 3.884, "step": 1013 }, { "epoch": 0.19, "learning_rate": 2.9448617731172544e-06, "loss": 3.6011, "step": 1014 }, { "epoch": 0.19, "learning_rate": 2.9448045757864632e-06, "loss": 3.7498, "step": 1015 }, { "epoch": 0.19, "learning_rate": 2.944747378455672e-06, "loss": 3.7998, "step": 1016 }, { "epoch": 0.19, "learning_rate": 2.944690181124881e-06, "loss": 3.8635, "step": 1017 }, { "epoch": 0.19, "learning_rate": 2.94463298379409e-06, "loss": 3.8444, "step": 1018 }, { "epoch": 0.19, "learning_rate": 2.9445757864632984e-06, "loss": 3.7255, "step": 1019 }, { "epoch": 0.19, "learning_rate": 2.944518589132507e-06, "loss": 3.7256, "step": 1020 }, { "epoch": 0.19, "learning_rate": 2.944461391801716e-06, "loss": 3.7161, "step": 1021 }, { "epoch": 0.19, "learning_rate": 2.944404194470925e-06, "loss": 3.7133, "step": 1022 }, { "epoch": 0.19, "learning_rate": 2.9443469971401336e-06, "loss": 3.7726, "step": 1023 }, { "epoch": 0.2, "learning_rate": 2.9442897998093424e-06, "loss": 3.8322, "step": 1024 }, { "epoch": 0.2, "learning_rate": 2.944232602478551e-06, "loss": 3.6348, "step": 1025 }, { "epoch": 0.2, "learning_rate": 2.9441754051477595e-06, "loss": 3.5791, "step": 1026 }, { "epoch": 0.2, "learning_rate": 2.9441182078169688e-06, "loss": 3.6905, "step": 1027 }, { "epoch": 0.2, "learning_rate": 2.9440610104861776e-06, "loss": 3.7736, "step": 1028 }, { "epoch": 0.2, "learning_rate": 2.944003813155386e-06, "loss": 3.6831, "step": 1029 }, { "epoch": 0.2, "learning_rate": 2.943946615824595e-06, "loss": 3.6782, "step": 1030 }, { "epoch": 0.2, "learning_rate": 2.9438894184938035e-06, "loss": 3.7014, "step": 1031 }, { "epoch": 0.2, "learning_rate": 2.9438322211630123e-06, "loss": 3.8058, "step": 1032 }, { "epoch": 0.2, "learning_rate": 2.943775023832221e-06, "loss": 3.7518, "step": 1033 }, { "epoch": 0.2, "learning_rate": 2.94371782650143e-06, "loss": 3.8159, "step": 1034 }, { "epoch": 0.2, "learning_rate": 2.943660629170639e-06, "loss": 3.8273, "step": 1035 }, { "epoch": 0.2, "learning_rate": 2.9436034318398475e-06, "loss": 3.6392, "step": 1036 }, { "epoch": 0.2, "learning_rate": 2.9435462345090563e-06, "loss": 3.7144, "step": 1037 }, { "epoch": 0.2, "learning_rate": 2.943489037178265e-06, "loss": 3.638, "step": 1038 }, { "epoch": 0.2, "learning_rate": 2.943431839847474e-06, "loss": 3.7843, "step": 1039 }, { "epoch": 0.2, "learning_rate": 2.9433746425166827e-06, "loss": 3.7377, "step": 1040 }, { "epoch": 0.2, "learning_rate": 2.9433174451858915e-06, "loss": 3.6549, "step": 1041 }, { "epoch": 0.2, "learning_rate": 2.9432602478551002e-06, "loss": 3.8017, "step": 1042 }, { "epoch": 0.2, "learning_rate": 2.9432030505243086e-06, "loss": 3.8256, "step": 1043 }, { "epoch": 0.2, "learning_rate": 2.943145853193518e-06, "loss": 3.8052, "step": 1044 }, { "epoch": 0.2, "learning_rate": 2.9430886558627266e-06, "loss": 3.7041, "step": 1045 }, { "epoch": 0.2, "learning_rate": 2.943031458531935e-06, "loss": 3.7383, "step": 1046 }, { "epoch": 0.2, "learning_rate": 2.9429742612011442e-06, "loss": 3.7672, "step": 1047 }, { "epoch": 0.2, "learning_rate": 2.9429170638703526e-06, "loss": 3.6536, "step": 1048 }, { "epoch": 0.2, "learning_rate": 2.9428598665395614e-06, "loss": 3.6752, "step": 1049 }, { "epoch": 0.2, "learning_rate": 2.9428026692087706e-06, "loss": 3.6907, "step": 1050 }, { "epoch": 0.2, "learning_rate": 2.942745471877979e-06, "loss": 3.8216, "step": 1051 }, { "epoch": 0.2, "learning_rate": 2.942688274547188e-06, "loss": 3.665, "step": 1052 }, { "epoch": 0.2, "learning_rate": 2.9426310772163966e-06, "loss": 3.6184, "step": 1053 }, { "epoch": 0.2, "learning_rate": 2.9425738798856054e-06, "loss": 3.7034, "step": 1054 }, { "epoch": 0.2, "learning_rate": 2.942516682554814e-06, "loss": 3.8285, "step": 1055 }, { "epoch": 0.2, "learning_rate": 2.942459485224023e-06, "loss": 3.6946, "step": 1056 }, { "epoch": 0.2, "learning_rate": 2.9424022878932317e-06, "loss": 3.8011, "step": 1057 }, { "epoch": 0.2, "learning_rate": 2.9423450905624405e-06, "loss": 3.6894, "step": 1058 }, { "epoch": 0.2, "learning_rate": 2.9422878932316493e-06, "loss": 3.7426, "step": 1059 }, { "epoch": 0.2, "learning_rate": 2.942230695900858e-06, "loss": 3.6749, "step": 1060 }, { "epoch": 0.2, "learning_rate": 2.942173498570067e-06, "loss": 3.8042, "step": 1061 }, { "epoch": 0.2, "learning_rate": 2.9421163012392757e-06, "loss": 3.546, "step": 1062 }, { "epoch": 0.2, "learning_rate": 2.942059103908484e-06, "loss": 3.6621, "step": 1063 }, { "epoch": 0.2, "learning_rate": 2.9420019065776933e-06, "loss": 3.8505, "step": 1064 }, { "epoch": 0.2, "learning_rate": 2.9419447092469017e-06, "loss": 3.7286, "step": 1065 }, { "epoch": 0.2, "learning_rate": 2.9418875119161105e-06, "loss": 3.7385, "step": 1066 }, { "epoch": 0.2, "learning_rate": 2.9418303145853197e-06, "loss": 3.8225, "step": 1067 }, { "epoch": 0.2, "learning_rate": 2.941773117254528e-06, "loss": 3.5729, "step": 1068 }, { "epoch": 0.2, "learning_rate": 2.9417159199237373e-06, "loss": 3.792, "step": 1069 }, { "epoch": 0.2, "learning_rate": 2.9416587225929456e-06, "loss": 3.7921, "step": 1070 }, { "epoch": 0.2, "learning_rate": 2.9416015252621544e-06, "loss": 3.7349, "step": 1071 }, { "epoch": 0.2, "learning_rate": 2.9415443279313632e-06, "loss": 3.7081, "step": 1072 }, { "epoch": 0.2, "learning_rate": 2.941487130600572e-06, "loss": 3.7234, "step": 1073 }, { "epoch": 0.2, "learning_rate": 2.941429933269781e-06, "loss": 3.8493, "step": 1074 }, { "epoch": 0.2, "learning_rate": 2.9413727359389896e-06, "loss": 3.759, "step": 1075 }, { "epoch": 0.2, "learning_rate": 2.9413155386081984e-06, "loss": 3.8742, "step": 1076 }, { "epoch": 0.21, "learning_rate": 2.941258341277407e-06, "loss": 3.8614, "step": 1077 }, { "epoch": 0.21, "learning_rate": 2.941201143946616e-06, "loss": 3.6198, "step": 1078 }, { "epoch": 0.21, "learning_rate": 2.9411439466158248e-06, "loss": 3.5939, "step": 1079 }, { "epoch": 0.21, "learning_rate": 2.941086749285033e-06, "loss": 3.5632, "step": 1080 }, { "epoch": 0.21, "learning_rate": 2.9410295519542424e-06, "loss": 3.8129, "step": 1081 }, { "epoch": 0.21, "learning_rate": 2.9409723546234507e-06, "loss": 3.8213, "step": 1082 }, { "epoch": 0.21, "learning_rate": 2.9409151572926595e-06, "loss": 3.8112, "step": 1083 }, { "epoch": 0.21, "learning_rate": 2.9408579599618687e-06, "loss": 3.6913, "step": 1084 }, { "epoch": 0.21, "learning_rate": 2.940800762631077e-06, "loss": 3.8023, "step": 1085 }, { "epoch": 0.21, "learning_rate": 2.9407435653002863e-06, "loss": 3.725, "step": 1086 }, { "epoch": 0.21, "learning_rate": 2.9406863679694947e-06, "loss": 3.7774, "step": 1087 }, { "epoch": 0.21, "learning_rate": 2.9406291706387035e-06, "loss": 3.8057, "step": 1088 }, { "epoch": 0.21, "learning_rate": 2.9405719733079123e-06, "loss": 3.8703, "step": 1089 }, { "epoch": 0.21, "learning_rate": 2.940514775977121e-06, "loss": 3.8513, "step": 1090 }, { "epoch": 0.21, "learning_rate": 2.94045757864633e-06, "loss": 3.7566, "step": 1091 }, { "epoch": 0.21, "learning_rate": 2.9404003813155387e-06, "loss": 3.7589, "step": 1092 }, { "epoch": 0.21, "learning_rate": 2.9403431839847475e-06, "loss": 3.7468, "step": 1093 }, { "epoch": 0.21, "learning_rate": 2.9402859866539563e-06, "loss": 3.8279, "step": 1094 }, { "epoch": 0.21, "learning_rate": 2.940228789323165e-06, "loss": 3.708, "step": 1095 }, { "epoch": 0.21, "learning_rate": 2.940171591992374e-06, "loss": 3.8873, "step": 1096 }, { "epoch": 0.21, "learning_rate": 2.9401143946615822e-06, "loss": 3.6955, "step": 1097 }, { "epoch": 0.21, "learning_rate": 2.9400571973307914e-06, "loss": 3.8154, "step": 1098 }, { "epoch": 0.21, "learning_rate": 2.9400000000000002e-06, "loss": 3.5807, "step": 1099 }, { "epoch": 0.21, "learning_rate": 2.939942802669209e-06, "loss": 3.7173, "step": 1100 }, { "epoch": 0.21, "learning_rate": 2.939885605338418e-06, "loss": 3.8754, "step": 1101 }, { "epoch": 0.21, "learning_rate": 2.939828408007626e-06, "loss": 3.6276, "step": 1102 }, { "epoch": 0.21, "learning_rate": 2.9397712106768354e-06, "loss": 3.8846, "step": 1103 }, { "epoch": 0.21, "learning_rate": 2.9397140133460438e-06, "loss": 3.6609, "step": 1104 }, { "epoch": 0.21, "learning_rate": 2.9396568160152526e-06, "loss": 3.796, "step": 1105 }, { "epoch": 0.21, "learning_rate": 2.9395996186844618e-06, "loss": 3.7685, "step": 1106 }, { "epoch": 0.21, "learning_rate": 2.93954242135367e-06, "loss": 3.826, "step": 1107 }, { "epoch": 0.21, "learning_rate": 2.939485224022879e-06, "loss": 3.7843, "step": 1108 }, { "epoch": 0.21, "learning_rate": 2.9394280266920877e-06, "loss": 3.8149, "step": 1109 }, { "epoch": 0.21, "learning_rate": 2.9393708293612965e-06, "loss": 3.6336, "step": 1110 }, { "epoch": 0.21, "learning_rate": 2.9393136320305053e-06, "loss": 3.68, "step": 1111 }, { "epoch": 0.21, "learning_rate": 2.939256434699714e-06, "loss": 3.7425, "step": 1112 }, { "epoch": 0.21, "learning_rate": 2.939199237368923e-06, "loss": 3.7142, "step": 1113 }, { "epoch": 0.21, "learning_rate": 2.9391420400381313e-06, "loss": 3.8068, "step": 1114 }, { "epoch": 0.21, "learning_rate": 2.9390848427073405e-06, "loss": 3.6859, "step": 1115 }, { "epoch": 0.21, "learning_rate": 2.9390276453765493e-06, "loss": 3.7353, "step": 1116 }, { "epoch": 0.21, "learning_rate": 2.938970448045758e-06, "loss": 3.6914, "step": 1117 }, { "epoch": 0.21, "learning_rate": 2.938913250714967e-06, "loss": 3.8092, "step": 1118 }, { "epoch": 0.21, "learning_rate": 2.9388560533841753e-06, "loss": 3.6351, "step": 1119 }, { "epoch": 0.21, "learning_rate": 2.9387988560533845e-06, "loss": 3.9271, "step": 1120 }, { "epoch": 0.21, "learning_rate": 2.938741658722593e-06, "loss": 3.7783, "step": 1121 }, { "epoch": 0.21, "learning_rate": 2.9386844613918016e-06, "loss": 3.8333, "step": 1122 }, { "epoch": 0.21, "learning_rate": 2.938627264061011e-06, "loss": 3.7641, "step": 1123 }, { "epoch": 0.21, "learning_rate": 2.9385700667302192e-06, "loss": 3.7065, "step": 1124 }, { "epoch": 0.21, "learning_rate": 2.938512869399428e-06, "loss": 3.777, "step": 1125 }, { "epoch": 0.21, "learning_rate": 2.938455672068637e-06, "loss": 3.6736, "step": 1126 }, { "epoch": 0.21, "learning_rate": 2.9383984747378456e-06, "loss": 3.7801, "step": 1127 }, { "epoch": 0.21, "learning_rate": 2.9383412774070544e-06, "loss": 3.8032, "step": 1128 }, { "epoch": 0.22, "learning_rate": 2.938284080076263e-06, "loss": 3.7912, "step": 1129 }, { "epoch": 0.22, "learning_rate": 2.938226882745472e-06, "loss": 3.8048, "step": 1130 }, { "epoch": 0.22, "learning_rate": 2.9381696854146804e-06, "loss": 3.7218, "step": 1131 }, { "epoch": 0.22, "learning_rate": 2.9381124880838896e-06, "loss": 3.6339, "step": 1132 }, { "epoch": 0.22, "learning_rate": 2.9380552907530984e-06, "loss": 3.8735, "step": 1133 }, { "epoch": 0.22, "learning_rate": 2.937998093422307e-06, "loss": 3.7548, "step": 1134 }, { "epoch": 0.22, "learning_rate": 2.937940896091516e-06, "loss": 3.7679, "step": 1135 }, { "epoch": 0.22, "learning_rate": 2.9378836987607243e-06, "loss": 3.7538, "step": 1136 }, { "epoch": 0.22, "learning_rate": 2.9378265014299335e-06, "loss": 3.6114, "step": 1137 }, { "epoch": 0.22, "learning_rate": 2.9377693040991423e-06, "loss": 3.8263, "step": 1138 }, { "epoch": 0.22, "learning_rate": 2.9377121067683507e-06, "loss": 3.9246, "step": 1139 }, { "epoch": 0.22, "learning_rate": 2.93765490943756e-06, "loss": 3.8039, "step": 1140 }, { "epoch": 0.22, "learning_rate": 2.9375977121067683e-06, "loss": 3.638, "step": 1141 }, { "epoch": 0.22, "learning_rate": 2.937540514775977e-06, "loss": 3.7583, "step": 1142 }, { "epoch": 0.22, "learning_rate": 2.937483317445186e-06, "loss": 3.7543, "step": 1143 }, { "epoch": 0.22, "learning_rate": 2.9374261201143947e-06, "loss": 3.6958, "step": 1144 }, { "epoch": 0.22, "learning_rate": 2.9373689227836035e-06, "loss": 3.6902, "step": 1145 }, { "epoch": 0.22, "learning_rate": 2.9373117254528123e-06, "loss": 3.7497, "step": 1146 }, { "epoch": 0.22, "learning_rate": 2.937254528122021e-06, "loss": 3.7407, "step": 1147 }, { "epoch": 0.22, "learning_rate": 2.93719733079123e-06, "loss": 3.7161, "step": 1148 }, { "epoch": 0.22, "learning_rate": 2.9371401334604386e-06, "loss": 3.8481, "step": 1149 }, { "epoch": 0.22, "learning_rate": 2.9370829361296474e-06, "loss": 3.7246, "step": 1150 }, { "epoch": 0.22, "learning_rate": 2.9370257387988562e-06, "loss": 3.6209, "step": 1151 }, { "epoch": 0.22, "learning_rate": 2.936968541468065e-06, "loss": 3.8377, "step": 1152 }, { "epoch": 0.22, "learning_rate": 2.9369113441372734e-06, "loss": 3.5785, "step": 1153 }, { "epoch": 0.22, "learning_rate": 2.9368541468064826e-06, "loss": 3.7373, "step": 1154 }, { "epoch": 0.22, "learning_rate": 2.9367969494756914e-06, "loss": 3.6119, "step": 1155 }, { "epoch": 0.22, "learning_rate": 2.9367397521448998e-06, "loss": 3.6856, "step": 1156 }, { "epoch": 0.22, "learning_rate": 2.936682554814109e-06, "loss": 3.7281, "step": 1157 }, { "epoch": 0.22, "learning_rate": 2.9366253574833174e-06, "loss": 3.7524, "step": 1158 }, { "epoch": 0.22, "learning_rate": 2.936568160152526e-06, "loss": 3.7508, "step": 1159 }, { "epoch": 0.22, "learning_rate": 2.936510962821735e-06, "loss": 3.8325, "step": 1160 }, { "epoch": 0.22, "learning_rate": 2.9364537654909437e-06, "loss": 3.799, "step": 1161 }, { "epoch": 0.22, "learning_rate": 2.9363965681601525e-06, "loss": 3.7572, "step": 1162 }, { "epoch": 0.22, "learning_rate": 2.9363393708293613e-06, "loss": 3.8641, "step": 1163 }, { "epoch": 0.22, "learning_rate": 2.93628217349857e-06, "loss": 3.7232, "step": 1164 }, { "epoch": 0.22, "learning_rate": 2.936224976167779e-06, "loss": 3.7115, "step": 1165 }, { "epoch": 0.22, "learning_rate": 2.9361677788369877e-06, "loss": 3.8027, "step": 1166 }, { "epoch": 0.22, "learning_rate": 2.9361105815061965e-06, "loss": 3.6568, "step": 1167 }, { "epoch": 0.22, "learning_rate": 2.9360533841754053e-06, "loss": 3.9403, "step": 1168 }, { "epoch": 0.22, "learning_rate": 2.935996186844614e-06, "loss": 3.7653, "step": 1169 }, { "epoch": 0.22, "learning_rate": 2.9359389895138225e-06, "loss": 3.7368, "step": 1170 }, { "epoch": 0.22, "learning_rate": 2.9358817921830317e-06, "loss": 3.678, "step": 1171 }, { "epoch": 0.22, "learning_rate": 2.9358245948522405e-06, "loss": 3.8502, "step": 1172 }, { "epoch": 0.22, "learning_rate": 2.935767397521449e-06, "loss": 3.7163, "step": 1173 }, { "epoch": 0.22, "learning_rate": 2.935710200190658e-06, "loss": 3.6871, "step": 1174 }, { "epoch": 0.22, "learning_rate": 2.9356530028598664e-06, "loss": 3.6241, "step": 1175 }, { "epoch": 0.22, "learning_rate": 2.9355958055290752e-06, "loss": 3.7288, "step": 1176 }, { "epoch": 0.22, "learning_rate": 2.935538608198284e-06, "loss": 3.7098, "step": 1177 }, { "epoch": 0.22, "learning_rate": 2.935481410867493e-06, "loss": 3.7554, "step": 1178 }, { "epoch": 0.22, "learning_rate": 2.9354242135367016e-06, "loss": 3.65, "step": 1179 }, { "epoch": 0.22, "learning_rate": 2.9353670162059104e-06, "loss": 3.8242, "step": 1180 }, { "epoch": 0.22, "learning_rate": 2.935309818875119e-06, "loss": 3.6401, "step": 1181 }, { "epoch": 0.23, "learning_rate": 2.935252621544328e-06, "loss": 3.6608, "step": 1182 }, { "epoch": 0.23, "learning_rate": 2.935195424213537e-06, "loss": 3.8808, "step": 1183 }, { "epoch": 0.23, "learning_rate": 2.9351382268827456e-06, "loss": 3.5679, "step": 1184 }, { "epoch": 0.23, "learning_rate": 2.9350810295519544e-06, "loss": 3.8602, "step": 1185 }, { "epoch": 0.23, "learning_rate": 2.935023832221163e-06, "loss": 3.9045, "step": 1186 }, { "epoch": 0.23, "learning_rate": 2.934966634890372e-06, "loss": 3.8076, "step": 1187 }, { "epoch": 0.23, "learning_rate": 2.9349094375595808e-06, "loss": 3.6852, "step": 1188 }, { "epoch": 0.23, "learning_rate": 2.9348522402287896e-06, "loss": 3.9048, "step": 1189 }, { "epoch": 0.23, "learning_rate": 2.934795042897998e-06, "loss": 3.7568, "step": 1190 }, { "epoch": 0.23, "learning_rate": 2.934737845567207e-06, "loss": 3.7841, "step": 1191 }, { "epoch": 0.23, "learning_rate": 2.9346806482364155e-06, "loss": 3.6301, "step": 1192 }, { "epoch": 0.23, "learning_rate": 2.9346234509056243e-06, "loss": 3.571, "step": 1193 }, { "epoch": 0.23, "learning_rate": 2.9345662535748335e-06, "loss": 3.5489, "step": 1194 }, { "epoch": 0.23, "learning_rate": 2.934509056244042e-06, "loss": 3.5248, "step": 1195 }, { "epoch": 0.23, "learning_rate": 2.9344518589132507e-06, "loss": 3.6352, "step": 1196 }, { "epoch": 0.23, "learning_rate": 2.9343946615824595e-06, "loss": 3.706, "step": 1197 }, { "epoch": 0.23, "learning_rate": 2.9343374642516683e-06, "loss": 3.7481, "step": 1198 }, { "epoch": 0.23, "learning_rate": 2.934280266920877e-06, "loss": 3.768, "step": 1199 }, { "epoch": 0.23, "learning_rate": 2.934223069590086e-06, "loss": 3.6508, "step": 1200 }, { "epoch": 0.23, "learning_rate": 2.9341658722592947e-06, "loss": 3.7806, "step": 1201 }, { "epoch": 0.23, "learning_rate": 2.9341086749285034e-06, "loss": 3.6853, "step": 1202 }, { "epoch": 0.23, "learning_rate": 2.9340514775977122e-06, "loss": 3.676, "step": 1203 }, { "epoch": 0.23, "learning_rate": 2.933994280266921e-06, "loss": 3.7716, "step": 1204 }, { "epoch": 0.23, "learning_rate": 2.93393708293613e-06, "loss": 3.6892, "step": 1205 }, { "epoch": 0.23, "learning_rate": 2.9338798856053386e-06, "loss": 3.759, "step": 1206 }, { "epoch": 0.23, "learning_rate": 2.933822688274547e-06, "loss": 3.6969, "step": 1207 }, { "epoch": 0.23, "learning_rate": 2.9337654909437562e-06, "loss": 3.661, "step": 1208 }, { "epoch": 0.23, "learning_rate": 2.9337082936129646e-06, "loss": 3.9355, "step": 1209 }, { "epoch": 0.23, "learning_rate": 2.9336510962821734e-06, "loss": 3.6332, "step": 1210 }, { "epoch": 0.23, "learning_rate": 2.9335938989513826e-06, "loss": 3.6325, "step": 1211 }, { "epoch": 0.23, "learning_rate": 2.933536701620591e-06, "loss": 3.7361, "step": 1212 }, { "epoch": 0.23, "learning_rate": 2.9334795042897998e-06, "loss": 3.7205, "step": 1213 }, { "epoch": 0.23, "learning_rate": 2.9334223069590086e-06, "loss": 3.5486, "step": 1214 }, { "epoch": 0.23, "learning_rate": 2.9333651096282173e-06, "loss": 3.8003, "step": 1215 }, { "epoch": 0.23, "learning_rate": 2.933307912297426e-06, "loss": 3.7278, "step": 1216 }, { "epoch": 0.23, "learning_rate": 2.933250714966635e-06, "loss": 3.7184, "step": 1217 }, { "epoch": 0.23, "learning_rate": 2.9331935176358437e-06, "loss": 3.7275, "step": 1218 }, { "epoch": 0.23, "learning_rate": 2.9331363203050525e-06, "loss": 3.7396, "step": 1219 }, { "epoch": 0.23, "learning_rate": 2.9330791229742613e-06, "loss": 3.6293, "step": 1220 }, { "epoch": 0.23, "learning_rate": 2.93302192564347e-06, "loss": 3.6758, "step": 1221 }, { "epoch": 0.23, "learning_rate": 2.932964728312679e-06, "loss": 3.5327, "step": 1222 }, { "epoch": 0.23, "learning_rate": 2.9329075309818877e-06, "loss": 3.5382, "step": 1223 }, { "epoch": 0.23, "learning_rate": 2.932850333651096e-06, "loss": 3.6764, "step": 1224 }, { "epoch": 0.23, "learning_rate": 2.9327931363203053e-06, "loss": 3.7801, "step": 1225 }, { "epoch": 0.23, "learning_rate": 2.9327359389895137e-06, "loss": 3.7113, "step": 1226 }, { "epoch": 0.23, "learning_rate": 2.9326787416587224e-06, "loss": 3.5893, "step": 1227 }, { "epoch": 0.23, "learning_rate": 2.9326215443279317e-06, "loss": 3.7748, "step": 1228 }, { "epoch": 0.23, "learning_rate": 2.93256434699714e-06, "loss": 3.7891, "step": 1229 }, { "epoch": 0.23, "learning_rate": 2.9325071496663493e-06, "loss": 3.6677, "step": 1230 }, { "epoch": 0.23, "learning_rate": 2.9324499523355576e-06, "loss": 3.687, "step": 1231 }, { "epoch": 0.23, "learning_rate": 2.9323927550047664e-06, "loss": 3.4937, "step": 1232 }, { "epoch": 0.23, "learning_rate": 2.9323355576739756e-06, "loss": 3.6022, "step": 1233 }, { "epoch": 0.24, "learning_rate": 2.932278360343184e-06, "loss": 3.713, "step": 1234 }, { "epoch": 0.24, "learning_rate": 2.932221163012393e-06, "loss": 3.7745, "step": 1235 }, { "epoch": 0.24, "learning_rate": 2.9321639656816016e-06, "loss": 3.8276, "step": 1236 }, { "epoch": 0.24, "learning_rate": 2.9321067683508104e-06, "loss": 3.7672, "step": 1237 }, { "epoch": 0.24, "learning_rate": 2.932049571020019e-06, "loss": 3.7394, "step": 1238 }, { "epoch": 0.24, "learning_rate": 2.931992373689228e-06, "loss": 3.7036, "step": 1239 }, { "epoch": 0.24, "learning_rate": 2.9319351763584368e-06, "loss": 3.8025, "step": 1240 }, { "epoch": 0.24, "learning_rate": 2.931877979027645e-06, "loss": 3.8589, "step": 1241 }, { "epoch": 0.24, "learning_rate": 2.9318207816968544e-06, "loss": 3.7339, "step": 1242 }, { "epoch": 0.24, "learning_rate": 2.931763584366063e-06, "loss": 3.6284, "step": 1243 }, { "epoch": 0.24, "learning_rate": 2.9317063870352715e-06, "loss": 3.7644, "step": 1244 }, { "epoch": 0.24, "learning_rate": 2.9316491897044807e-06, "loss": 3.6622, "step": 1245 }, { "epoch": 0.24, "learning_rate": 2.931591992373689e-06, "loss": 3.8255, "step": 1246 }, { "epoch": 0.24, "learning_rate": 2.9315347950428983e-06, "loss": 3.5823, "step": 1247 }, { "epoch": 0.24, "learning_rate": 2.9314775977121067e-06, "loss": 3.6811, "step": 1248 }, { "epoch": 0.24, "learning_rate": 2.9314204003813155e-06, "loss": 3.5744, "step": 1249 }, { "epoch": 0.24, "learning_rate": 2.9313632030505247e-06, "loss": 3.7048, "step": 1250 }, { "epoch": 0.24, "learning_rate": 2.931306005719733e-06, "loss": 3.6476, "step": 1251 }, { "epoch": 0.24, "learning_rate": 2.931248808388942e-06, "loss": 3.7452, "step": 1252 }, { "epoch": 0.24, "learning_rate": 2.9311916110581507e-06, "loss": 3.6947, "step": 1253 }, { "epoch": 0.24, "learning_rate": 2.9311344137273595e-06, "loss": 3.6459, "step": 1254 }, { "epoch": 0.24, "learning_rate": 2.9310772163965683e-06, "loss": 3.7014, "step": 1255 }, { "epoch": 0.24, "learning_rate": 2.931020019065777e-06, "loss": 3.6983, "step": 1256 }, { "epoch": 0.24, "learning_rate": 2.930962821734986e-06, "loss": 3.7183, "step": 1257 }, { "epoch": 0.24, "learning_rate": 2.930905624404194e-06, "loss": 3.6926, "step": 1258 }, { "epoch": 0.24, "learning_rate": 2.9308484270734034e-06, "loss": 3.593, "step": 1259 }, { "epoch": 0.24, "learning_rate": 2.9307912297426122e-06, "loss": 3.9084, "step": 1260 }, { "epoch": 0.24, "learning_rate": 2.9307340324118206e-06, "loss": 3.749, "step": 1261 }, { "epoch": 0.24, "learning_rate": 2.93067683508103e-06, "loss": 3.7194, "step": 1262 }, { "epoch": 0.24, "learning_rate": 2.930619637750238e-06, "loss": 3.6873, "step": 1263 }, { "epoch": 0.24, "learning_rate": 2.9305624404194474e-06, "loss": 3.8415, "step": 1264 }, { "epoch": 0.24, "learning_rate": 2.9305052430886558e-06, "loss": 3.7418, "step": 1265 }, { "epoch": 0.24, "learning_rate": 2.9304480457578646e-06, "loss": 3.9032, "step": 1266 }, { "epoch": 0.24, "learning_rate": 2.9303908484270738e-06, "loss": 3.5395, "step": 1267 }, { "epoch": 0.24, "learning_rate": 2.930333651096282e-06, "loss": 3.7132, "step": 1268 }, { "epoch": 0.24, "learning_rate": 2.930276453765491e-06, "loss": 3.6486, "step": 1269 }, { "epoch": 0.24, "learning_rate": 2.9302192564346997e-06, "loss": 3.5801, "step": 1270 }, { "epoch": 0.24, "learning_rate": 2.9301620591039085e-06, "loss": 3.686, "step": 1271 }, { "epoch": 0.24, "learning_rate": 2.9301048617731173e-06, "loss": 3.736, "step": 1272 }, { "epoch": 0.24, "learning_rate": 2.930047664442326e-06, "loss": 3.52, "step": 1273 }, { "epoch": 0.24, "learning_rate": 2.929990467111535e-06, "loss": 3.6372, "step": 1274 }, { "epoch": 0.24, "learning_rate": 2.9299332697807437e-06, "loss": 3.7571, "step": 1275 }, { "epoch": 0.24, "learning_rate": 2.9298760724499525e-06, "loss": 3.6757, "step": 1276 }, { "epoch": 0.24, "learning_rate": 2.9298188751191613e-06, "loss": 3.9493, "step": 1277 }, { "epoch": 0.24, "learning_rate": 2.9297616777883697e-06, "loss": 3.8351, "step": 1278 }, { "epoch": 0.24, "learning_rate": 2.929704480457579e-06, "loss": 3.7148, "step": 1279 }, { "epoch": 0.24, "learning_rate": 2.9296472831267872e-06, "loss": 3.6932, "step": 1280 }, { "epoch": 0.24, "learning_rate": 2.9295900857959965e-06, "loss": 3.871, "step": 1281 }, { "epoch": 0.24, "learning_rate": 2.9295328884652053e-06, "loss": 3.7024, "step": 1282 }, { "epoch": 0.24, "learning_rate": 2.9294756911344136e-06, "loss": 3.8207, "step": 1283 }, { "epoch": 0.24, "learning_rate": 2.929418493803623e-06, "loss": 3.8006, "step": 1284 }, { "epoch": 0.24, "learning_rate": 2.9293612964728312e-06, "loss": 3.7549, "step": 1285 }, { "epoch": 0.24, "learning_rate": 2.92930409914204e-06, "loss": 3.7522, "step": 1286 }, { "epoch": 0.25, "learning_rate": 2.929246901811249e-06, "loss": 3.6595, "step": 1287 }, { "epoch": 0.25, "learning_rate": 2.9291897044804576e-06, "loss": 3.6617, "step": 1288 }, { "epoch": 0.25, "learning_rate": 2.9291325071496664e-06, "loss": 3.5947, "step": 1289 }, { "epoch": 0.25, "learning_rate": 2.929075309818875e-06, "loss": 3.7959, "step": 1290 }, { "epoch": 0.25, "learning_rate": 2.929018112488084e-06, "loss": 3.7015, "step": 1291 }, { "epoch": 0.25, "learning_rate": 2.9289609151572928e-06, "loss": 3.6761, "step": 1292 }, { "epoch": 0.25, "learning_rate": 2.9289037178265016e-06, "loss": 3.7757, "step": 1293 }, { "epoch": 0.25, "learning_rate": 2.9288465204957104e-06, "loss": 3.653, "step": 1294 }, { "epoch": 0.25, "learning_rate": 2.928789323164919e-06, "loss": 3.5027, "step": 1295 }, { "epoch": 0.25, "learning_rate": 2.928732125834128e-06, "loss": 3.5197, "step": 1296 }, { "epoch": 0.25, "learning_rate": 2.9286749285033363e-06, "loss": 3.5929, "step": 1297 }, { "epoch": 0.25, "learning_rate": 2.9286177311725455e-06, "loss": 3.7205, "step": 1298 }, { "epoch": 0.25, "learning_rate": 2.9285605338417543e-06, "loss": 3.731, "step": 1299 }, { "epoch": 0.25, "learning_rate": 2.9285033365109627e-06, "loss": 3.8429, "step": 1300 }, { "epoch": 0.25, "learning_rate": 2.928446139180172e-06, "loss": 3.7223, "step": 1301 }, { "epoch": 0.25, "learning_rate": 2.9283889418493803e-06, "loss": 3.8462, "step": 1302 }, { "epoch": 0.25, "learning_rate": 2.928331744518589e-06, "loss": 3.6811, "step": 1303 }, { "epoch": 0.25, "learning_rate": 2.928274547187798e-06, "loss": 3.7481, "step": 1304 }, { "epoch": 0.25, "learning_rate": 2.9282173498570067e-06, "loss": 3.7083, "step": 1305 }, { "epoch": 0.25, "learning_rate": 2.9281601525262155e-06, "loss": 3.9309, "step": 1306 }, { "epoch": 0.25, "learning_rate": 2.9281029551954243e-06, "loss": 3.785, "step": 1307 }, { "epoch": 0.25, "learning_rate": 2.928045757864633e-06, "loss": 3.5218, "step": 1308 }, { "epoch": 0.25, "learning_rate": 2.927988560533842e-06, "loss": 3.595, "step": 1309 }, { "epoch": 0.25, "learning_rate": 2.9279313632030506e-06, "loss": 3.6773, "step": 1310 }, { "epoch": 0.25, "learning_rate": 2.9278741658722594e-06, "loss": 3.8424, "step": 1311 }, { "epoch": 0.25, "learning_rate": 2.9278169685414682e-06, "loss": 3.7584, "step": 1312 }, { "epoch": 0.25, "learning_rate": 2.927759771210677e-06, "loss": 3.5371, "step": 1313 }, { "epoch": 0.25, "learning_rate": 2.9277025738798854e-06, "loss": 3.6412, "step": 1314 }, { "epoch": 0.25, "learning_rate": 2.9276453765490946e-06, "loss": 3.6332, "step": 1315 }, { "epoch": 0.25, "learning_rate": 2.9275881792183034e-06, "loss": 3.7632, "step": 1316 }, { "epoch": 0.25, "learning_rate": 2.9275309818875118e-06, "loss": 3.5702, "step": 1317 }, { "epoch": 0.25, "learning_rate": 2.927473784556721e-06, "loss": 3.9506, "step": 1318 }, { "epoch": 0.25, "learning_rate": 2.9274165872259294e-06, "loss": 3.6307, "step": 1319 }, { "epoch": 0.25, "learning_rate": 2.927359389895138e-06, "loss": 3.8739, "step": 1320 }, { "epoch": 0.25, "learning_rate": 2.9273021925643474e-06, "loss": 3.7933, "step": 1321 }, { "epoch": 0.25, "learning_rate": 2.9272449952335557e-06, "loss": 3.7929, "step": 1322 }, { "epoch": 0.25, "learning_rate": 2.9271877979027645e-06, "loss": 3.836, "step": 1323 }, { "epoch": 0.25, "learning_rate": 2.9271306005719733e-06, "loss": 3.6461, "step": 1324 }, { "epoch": 0.25, "learning_rate": 2.927073403241182e-06, "loss": 3.7759, "step": 1325 }, { "epoch": 0.25, "learning_rate": 2.927016205910391e-06, "loss": 3.7727, "step": 1326 }, { "epoch": 0.25, "learning_rate": 2.9269590085795997e-06, "loss": 3.6738, "step": 1327 }, { "epoch": 0.25, "learning_rate": 2.9269018112488085e-06, "loss": 3.4934, "step": 1328 }, { "epoch": 0.25, "learning_rate": 2.9268446139180173e-06, "loss": 3.8151, "step": 1329 }, { "epoch": 0.25, "learning_rate": 2.926787416587226e-06, "loss": 3.6308, "step": 1330 }, { "epoch": 0.25, "learning_rate": 2.926730219256435e-06, "loss": 3.8073, "step": 1331 }, { "epoch": 0.25, "learning_rate": 2.9266730219256437e-06, "loss": 3.7166, "step": 1332 }, { "epoch": 0.25, "learning_rate": 2.9266158245948525e-06, "loss": 3.7323, "step": 1333 }, { "epoch": 0.25, "learning_rate": 2.926558627264061e-06, "loss": 3.6483, "step": 1334 }, { "epoch": 0.25, "learning_rate": 2.92650142993327e-06, "loss": 3.7817, "step": 1335 }, { "epoch": 0.25, "learning_rate": 2.9264442326024784e-06, "loss": 3.7123, "step": 1336 }, { "epoch": 0.25, "learning_rate": 2.9263870352716872e-06, "loss": 3.7978, "step": 1337 }, { "epoch": 0.25, "learning_rate": 2.9263298379408964e-06, "loss": 3.8255, "step": 1338 }, { "epoch": 0.26, "learning_rate": 2.926272640610105e-06, "loss": 3.7831, "step": 1339 }, { "epoch": 0.26, "learning_rate": 2.9262154432793136e-06, "loss": 3.7211, "step": 1340 }, { "epoch": 0.26, "learning_rate": 2.9261582459485224e-06, "loss": 3.7592, "step": 1341 }, { "epoch": 0.26, "learning_rate": 2.926101048617731e-06, "loss": 3.5993, "step": 1342 }, { "epoch": 0.26, "learning_rate": 2.92604385128694e-06, "loss": 3.6561, "step": 1343 }, { "epoch": 0.26, "learning_rate": 2.9259866539561488e-06, "loss": 3.8288, "step": 1344 }, { "epoch": 0.26, "learning_rate": 2.9259294566253576e-06, "loss": 3.7197, "step": 1345 }, { "epoch": 0.26, "learning_rate": 2.9258722592945664e-06, "loss": 3.8492, "step": 1346 }, { "epoch": 0.26, "learning_rate": 2.925815061963775e-06, "loss": 3.7301, "step": 1347 }, { "epoch": 0.26, "learning_rate": 2.925757864632984e-06, "loss": 3.8336, "step": 1348 }, { "epoch": 0.26, "learning_rate": 2.9257006673021928e-06, "loss": 3.9418, "step": 1349 }, { "epoch": 0.26, "learning_rate": 2.9256434699714015e-06, "loss": 3.6956, "step": 1350 }, { "epoch": 0.26, "learning_rate": 2.92558627264061e-06, "loss": 3.9166, "step": 1351 }, { "epoch": 0.26, "learning_rate": 2.925529075309819e-06, "loss": 3.51, "step": 1352 }, { "epoch": 0.26, "learning_rate": 2.9254718779790275e-06, "loss": 3.6657, "step": 1353 }, { "epoch": 0.26, "learning_rate": 2.9254146806482363e-06, "loss": 3.8024, "step": 1354 }, { "epoch": 0.26, "learning_rate": 2.9253574833174455e-06, "loss": 3.7425, "step": 1355 }, { "epoch": 0.26, "learning_rate": 2.925300285986654e-06, "loss": 3.7795, "step": 1356 }, { "epoch": 0.26, "learning_rate": 2.9252430886558627e-06, "loss": 3.7096, "step": 1357 }, { "epoch": 0.26, "learning_rate": 2.9251858913250715e-06, "loss": 3.7701, "step": 1358 }, { "epoch": 0.26, "learning_rate": 2.9251286939942803e-06, "loss": 3.6596, "step": 1359 }, { "epoch": 0.26, "learning_rate": 2.9250714966634895e-06, "loss": 3.8888, "step": 1360 }, { "epoch": 0.26, "learning_rate": 2.925014299332698e-06, "loss": 3.7094, "step": 1361 }, { "epoch": 0.26, "learning_rate": 2.9249571020019066e-06, "loss": 3.7988, "step": 1362 }, { "epoch": 0.26, "learning_rate": 2.9248999046711154e-06, "loss": 3.7847, "step": 1363 }, { "epoch": 0.26, "learning_rate": 2.9248427073403242e-06, "loss": 3.6135, "step": 1364 }, { "epoch": 0.26, "learning_rate": 2.924785510009533e-06, "loss": 3.6193, "step": 1365 }, { "epoch": 0.26, "learning_rate": 2.924728312678742e-06, "loss": 3.7949, "step": 1366 }, { "epoch": 0.26, "learning_rate": 2.9246711153479506e-06, "loss": 3.7283, "step": 1367 }, { "epoch": 0.26, "learning_rate": 2.924613918017159e-06, "loss": 3.5621, "step": 1368 }, { "epoch": 0.26, "learning_rate": 2.924556720686368e-06, "loss": 3.5832, "step": 1369 }, { "epoch": 0.26, "learning_rate": 2.924499523355577e-06, "loss": 3.7494, "step": 1370 }, { "epoch": 0.26, "learning_rate": 2.9244423260247854e-06, "loss": 3.6979, "step": 1371 }, { "epoch": 0.26, "learning_rate": 2.9243851286939946e-06, "loss": 3.7926, "step": 1372 }, { "epoch": 0.26, "learning_rate": 2.924327931363203e-06, "loss": 3.6946, "step": 1373 }, { "epoch": 0.26, "learning_rate": 2.9242707340324118e-06, "loss": 3.7952, "step": 1374 }, { "epoch": 0.26, "learning_rate": 2.9242135367016205e-06, "loss": 3.5093, "step": 1375 }, { "epoch": 0.26, "learning_rate": 2.9241563393708293e-06, "loss": 3.5596, "step": 1376 }, { "epoch": 0.26, "learning_rate": 2.9240991420400386e-06, "loss": 3.6597, "step": 1377 }, { "epoch": 0.26, "learning_rate": 2.924041944709247e-06, "loss": 3.8317, "step": 1378 }, { "epoch": 0.26, "learning_rate": 2.9239847473784557e-06, "loss": 3.7826, "step": 1379 }, { "epoch": 0.26, "learning_rate": 2.9239275500476645e-06, "loss": 3.5875, "step": 1380 }, { "epoch": 0.26, "learning_rate": 2.9238703527168733e-06, "loss": 3.5967, "step": 1381 }, { "epoch": 0.26, "learning_rate": 2.923813155386082e-06, "loss": 3.6503, "step": 1382 }, { "epoch": 0.26, "learning_rate": 2.923755958055291e-06, "loss": 3.7773, "step": 1383 }, { "epoch": 0.26, "learning_rate": 2.9236987607244997e-06, "loss": 3.5775, "step": 1384 }, { "epoch": 0.26, "learning_rate": 2.923641563393708e-06, "loss": 3.8204, "step": 1385 }, { "epoch": 0.26, "learning_rate": 2.9235843660629173e-06, "loss": 3.6942, "step": 1386 }, { "epoch": 0.26, "learning_rate": 2.923527168732126e-06, "loss": 3.932, "step": 1387 }, { "epoch": 0.26, "learning_rate": 2.9234699714013344e-06, "loss": 3.8606, "step": 1388 }, { "epoch": 0.26, "learning_rate": 2.9234127740705437e-06, "loss": 3.6845, "step": 1389 }, { "epoch": 0.26, "learning_rate": 2.923355576739752e-06, "loss": 3.7903, "step": 1390 }, { "epoch": 0.26, "learning_rate": 2.923298379408961e-06, "loss": 3.7636, "step": 1391 }, { "epoch": 0.27, "learning_rate": 2.9232411820781696e-06, "loss": 3.762, "step": 1392 }, { "epoch": 0.27, "learning_rate": 2.9231839847473784e-06, "loss": 3.6442, "step": 1393 }, { "epoch": 0.27, "learning_rate": 2.9231267874165876e-06, "loss": 3.8075, "step": 1394 }, { "epoch": 0.27, "learning_rate": 2.923069590085796e-06, "loss": 3.8106, "step": 1395 }, { "epoch": 0.27, "learning_rate": 2.923012392755005e-06, "loss": 3.773, "step": 1396 }, { "epoch": 0.27, "learning_rate": 2.9229551954242136e-06, "loss": 3.6467, "step": 1397 }, { "epoch": 0.27, "learning_rate": 2.9228979980934224e-06, "loss": 3.702, "step": 1398 }, { "epoch": 0.27, "learning_rate": 2.922840800762631e-06, "loss": 3.7184, "step": 1399 }, { "epoch": 0.27, "learning_rate": 2.92278360343184e-06, "loss": 3.6153, "step": 1400 }, { "epoch": 0.27, "learning_rate": 2.9227264061010488e-06, "loss": 3.8256, "step": 1401 }, { "epoch": 0.27, "learning_rate": 2.922669208770257e-06, "loss": 3.8733, "step": 1402 }, { "epoch": 0.27, "learning_rate": 2.9226120114394663e-06, "loss": 3.9106, "step": 1403 }, { "epoch": 0.27, "learning_rate": 2.922554814108675e-06, "loss": 3.6822, "step": 1404 }, { "epoch": 0.27, "learning_rate": 2.9224976167778835e-06, "loss": 3.6165, "step": 1405 }, { "epoch": 0.27, "learning_rate": 2.9224404194470927e-06, "loss": 3.7604, "step": 1406 }, { "epoch": 0.27, "learning_rate": 2.922383222116301e-06, "loss": 3.7768, "step": 1407 }, { "epoch": 0.27, "learning_rate": 2.92232602478551e-06, "loss": 3.6988, "step": 1408 }, { "epoch": 0.27, "learning_rate": 2.922268827454719e-06, "loss": 3.6877, "step": 1409 }, { "epoch": 0.27, "learning_rate": 2.9222116301239275e-06, "loss": 3.7182, "step": 1410 }, { "epoch": 0.27, "learning_rate": 2.9221544327931367e-06, "loss": 3.5958, "step": 1411 }, { "epoch": 0.27, "learning_rate": 2.922097235462345e-06, "loss": 3.7274, "step": 1412 }, { "epoch": 0.27, "learning_rate": 2.922040038131554e-06, "loss": 3.5959, "step": 1413 }, { "epoch": 0.27, "learning_rate": 2.9219828408007627e-06, "loss": 3.6809, "step": 1414 }, { "epoch": 0.27, "learning_rate": 2.9219256434699715e-06, "loss": 3.7644, "step": 1415 }, { "epoch": 0.27, "learning_rate": 2.9218684461391802e-06, "loss": 3.7859, "step": 1416 }, { "epoch": 0.27, "learning_rate": 2.921811248808389e-06, "loss": 3.724, "step": 1417 }, { "epoch": 0.27, "learning_rate": 2.921754051477598e-06, "loss": 3.6615, "step": 1418 }, { "epoch": 0.27, "learning_rate": 2.9216968541468066e-06, "loss": 3.6764, "step": 1419 }, { "epoch": 0.27, "learning_rate": 2.9216396568160154e-06, "loss": 3.8069, "step": 1420 }, { "epoch": 0.27, "learning_rate": 2.9215824594852242e-06, "loss": 3.7812, "step": 1421 }, { "epoch": 0.27, "learning_rate": 2.9215252621544326e-06, "loss": 3.8033, "step": 1422 }, { "epoch": 0.27, "learning_rate": 2.921468064823642e-06, "loss": 3.7905, "step": 1423 }, { "epoch": 0.27, "learning_rate": 2.92141086749285e-06, "loss": 3.7641, "step": 1424 }, { "epoch": 0.27, "learning_rate": 2.9213536701620594e-06, "loss": 3.7004, "step": 1425 }, { "epoch": 0.27, "learning_rate": 2.921296472831268e-06, "loss": 3.8431, "step": 1426 }, { "epoch": 0.27, "learning_rate": 2.9212392755004766e-06, "loss": 3.8115, "step": 1427 }, { "epoch": 0.27, "learning_rate": 2.9211820781696858e-06, "loss": 3.7144, "step": 1428 }, { "epoch": 0.27, "learning_rate": 2.921124880838894e-06, "loss": 3.4638, "step": 1429 }, { "epoch": 0.27, "learning_rate": 2.921067683508103e-06, "loss": 3.7483, "step": 1430 }, { "epoch": 0.27, "learning_rate": 2.9210104861773117e-06, "loss": 3.784, "step": 1431 }, { "epoch": 0.27, "learning_rate": 2.9209532888465205e-06, "loss": 3.6809, "step": 1432 }, { "epoch": 0.27, "learning_rate": 2.9208960915157293e-06, "loss": 3.6426, "step": 1433 }, { "epoch": 0.27, "learning_rate": 2.920838894184938e-06, "loss": 3.6137, "step": 1434 }, { "epoch": 0.27, "learning_rate": 2.920781696854147e-06, "loss": 3.7294, "step": 1435 }, { "epoch": 0.27, "learning_rate": 2.9207244995233557e-06, "loss": 3.5946, "step": 1436 }, { "epoch": 0.27, "learning_rate": 2.9206673021925645e-06, "loss": 3.7825, "step": 1437 }, { "epoch": 0.27, "learning_rate": 2.9206101048617733e-06, "loss": 3.84, "step": 1438 }, { "epoch": 0.27, "learning_rate": 2.9205529075309817e-06, "loss": 3.5326, "step": 1439 }, { "epoch": 0.27, "learning_rate": 2.920495710200191e-06, "loss": 3.7578, "step": 1440 }, { "epoch": 0.27, "learning_rate": 2.9204385128693992e-06, "loss": 3.6227, "step": 1441 }, { "epoch": 0.27, "learning_rate": 2.9203813155386085e-06, "loss": 3.6781, "step": 1442 }, { "epoch": 0.27, "learning_rate": 2.9203241182078173e-06, "loss": 3.8899, "step": 1443 }, { "epoch": 0.28, "learning_rate": 2.9202669208770256e-06, "loss": 3.7369, "step": 1444 }, { "epoch": 0.28, "learning_rate": 2.920209723546235e-06, "loss": 3.7049, "step": 1445 }, { "epoch": 0.28, "learning_rate": 2.920152526215443e-06, "loss": 3.6975, "step": 1446 }, { "epoch": 0.28, "learning_rate": 2.920095328884652e-06, "loss": 3.7335, "step": 1447 }, { "epoch": 0.28, "learning_rate": 2.9200381315538612e-06, "loss": 3.5065, "step": 1448 }, { "epoch": 0.28, "learning_rate": 2.9199809342230696e-06, "loss": 3.6253, "step": 1449 }, { "epoch": 0.28, "learning_rate": 2.9199237368922784e-06, "loss": 3.8763, "step": 1450 }, { "epoch": 0.28, "learning_rate": 2.919866539561487e-06, "loss": 3.7791, "step": 1451 }, { "epoch": 0.28, "learning_rate": 2.919809342230696e-06, "loss": 3.8321, "step": 1452 }, { "epoch": 0.28, "learning_rate": 2.9197521448999048e-06, "loss": 3.807, "step": 1453 }, { "epoch": 0.28, "learning_rate": 2.9196949475691136e-06, "loss": 3.6991, "step": 1454 }, { "epoch": 0.28, "learning_rate": 2.9196377502383224e-06, "loss": 3.5902, "step": 1455 }, { "epoch": 0.28, "learning_rate": 2.9195805529075307e-06, "loss": 3.7902, "step": 1456 }, { "epoch": 0.28, "learning_rate": 2.91952335557674e-06, "loss": 3.7007, "step": 1457 }, { "epoch": 0.28, "learning_rate": 2.9194661582459487e-06, "loss": 3.6807, "step": 1458 }, { "epoch": 0.28, "learning_rate": 2.9194089609151575e-06, "loss": 3.758, "step": 1459 }, { "epoch": 0.28, "learning_rate": 2.9193517635843663e-06, "loss": 3.5342, "step": 1460 }, { "epoch": 0.28, "learning_rate": 2.9192945662535747e-06, "loss": 3.6798, "step": 1461 }, { "epoch": 0.28, "learning_rate": 2.919237368922784e-06, "loss": 3.7546, "step": 1462 }, { "epoch": 0.28, "learning_rate": 2.9191801715919923e-06, "loss": 3.8648, "step": 1463 }, { "epoch": 0.28, "learning_rate": 2.919122974261201e-06, "loss": 3.6288, "step": 1464 }, { "epoch": 0.28, "learning_rate": 2.9190657769304103e-06, "loss": 3.4447, "step": 1465 }, { "epoch": 0.28, "learning_rate": 2.9190085795996187e-06, "loss": 3.9023, "step": 1466 }, { "epoch": 0.28, "learning_rate": 2.9189513822688275e-06, "loss": 3.8022, "step": 1467 }, { "epoch": 0.28, "learning_rate": 2.9188941849380363e-06, "loss": 3.7675, "step": 1468 }, { "epoch": 0.28, "learning_rate": 2.918836987607245e-06, "loss": 3.8959, "step": 1469 }, { "epoch": 0.28, "learning_rate": 2.918779790276454e-06, "loss": 3.8781, "step": 1470 }, { "epoch": 0.28, "learning_rate": 2.9187225929456626e-06, "loss": 3.8099, "step": 1471 }, { "epoch": 0.28, "learning_rate": 2.9186653956148714e-06, "loss": 3.7182, "step": 1472 }, { "epoch": 0.28, "learning_rate": 2.91860819828408e-06, "loss": 3.6069, "step": 1473 }, { "epoch": 0.28, "learning_rate": 2.918551000953289e-06, "loss": 3.6428, "step": 1474 }, { "epoch": 0.28, "learning_rate": 2.918493803622498e-06, "loss": 3.583, "step": 1475 }, { "epoch": 0.28, "learning_rate": 2.9184366062917066e-06, "loss": 3.67, "step": 1476 }, { "epoch": 0.28, "learning_rate": 2.9183794089609154e-06, "loss": 3.6736, "step": 1477 }, { "epoch": 0.28, "learning_rate": 2.9183222116301238e-06, "loss": 3.6136, "step": 1478 }, { "epoch": 0.28, "learning_rate": 2.918265014299333e-06, "loss": 3.7599, "step": 1479 }, { "epoch": 0.28, "learning_rate": 2.9182078169685414e-06, "loss": 3.6548, "step": 1480 }, { "epoch": 0.28, "learning_rate": 2.91815061963775e-06, "loss": 3.8469, "step": 1481 }, { "epoch": 0.28, "learning_rate": 2.9180934223069594e-06, "loss": 3.6148, "step": 1482 }, { "epoch": 0.28, "learning_rate": 2.9180362249761677e-06, "loss": 3.8065, "step": 1483 }, { "epoch": 0.28, "learning_rate": 2.9179790276453765e-06, "loss": 3.9144, "step": 1484 }, { "epoch": 0.28, "learning_rate": 2.9179218303145853e-06, "loss": 3.6857, "step": 1485 }, { "epoch": 0.28, "learning_rate": 2.917864632983794e-06, "loss": 3.7945, "step": 1486 }, { "epoch": 0.28, "learning_rate": 2.917807435653003e-06, "loss": 3.7373, "step": 1487 }, { "epoch": 0.28, "learning_rate": 2.9177502383222117e-06, "loss": 3.624, "step": 1488 }, { "epoch": 0.28, "learning_rate": 2.9176930409914205e-06, "loss": 3.7095, "step": 1489 }, { "epoch": 0.28, "learning_rate": 2.9176358436606293e-06, "loss": 3.7103, "step": 1490 }, { "epoch": 0.28, "learning_rate": 2.917578646329838e-06, "loss": 3.7413, "step": 1491 }, { "epoch": 0.28, "learning_rate": 2.917521448999047e-06, "loss": 3.8167, "step": 1492 }, { "epoch": 0.28, "learning_rate": 2.9174642516682557e-06, "loss": 3.5791, "step": 1493 }, { "epoch": 0.28, "learning_rate": 2.9174070543374645e-06, "loss": 3.7544, "step": 1494 }, { "epoch": 0.28, "learning_rate": 2.917349857006673e-06, "loss": 3.7003, "step": 1495 }, { "epoch": 0.28, "learning_rate": 2.917292659675882e-06, "loss": 3.8134, "step": 1496 }, { "epoch": 0.29, "learning_rate": 2.917235462345091e-06, "loss": 3.7953, "step": 1497 }, { "epoch": 0.29, "learning_rate": 2.9171782650142992e-06, "loss": 3.7387, "step": 1498 }, { "epoch": 0.29, "learning_rate": 2.9171210676835084e-06, "loss": 3.7747, "step": 1499 }, { "epoch": 0.29, "learning_rate": 2.917063870352717e-06, "loss": 3.8693, "step": 1500 }, { "epoch": 0.29, "learning_rate": 2.9170066730219256e-06, "loss": 3.6712, "step": 1501 }, { "epoch": 0.29, "learning_rate": 2.9169494756911344e-06, "loss": 3.6466, "step": 1502 }, { "epoch": 0.29, "learning_rate": 2.916892278360343e-06, "loss": 3.7733, "step": 1503 }, { "epoch": 0.29, "learning_rate": 2.916835081029552e-06, "loss": 3.7726, "step": 1504 }, { "epoch": 0.29, "learning_rate": 2.9167778836987608e-06, "loss": 3.6488, "step": 1505 }, { "epoch": 0.29, "learning_rate": 2.9167206863679696e-06, "loss": 3.7015, "step": 1506 }, { "epoch": 0.29, "learning_rate": 2.9166634890371784e-06, "loss": 3.8556, "step": 1507 }, { "epoch": 0.29, "learning_rate": 2.916606291706387e-06, "loss": 3.7167, "step": 1508 }, { "epoch": 0.29, "learning_rate": 2.916549094375596e-06, "loss": 3.6333, "step": 1509 }, { "epoch": 0.29, "learning_rate": 2.9164918970448047e-06, "loss": 3.7842, "step": 1510 }, { "epoch": 0.29, "learning_rate": 2.9164346997140135e-06, "loss": 3.679, "step": 1511 }, { "epoch": 0.29, "learning_rate": 2.916377502383222e-06, "loss": 3.6027, "step": 1512 }, { "epoch": 0.29, "learning_rate": 2.916320305052431e-06, "loss": 3.4445, "step": 1513 }, { "epoch": 0.29, "learning_rate": 2.91626310772164e-06, "loss": 3.7916, "step": 1514 }, { "epoch": 0.29, "learning_rate": 2.9162059103908483e-06, "loss": 3.7468, "step": 1515 }, { "epoch": 0.29, "learning_rate": 2.9161487130600575e-06, "loss": 3.8471, "step": 1516 }, { "epoch": 0.29, "learning_rate": 2.916091515729266e-06, "loss": 3.7102, "step": 1517 }, { "epoch": 0.29, "learning_rate": 2.9160343183984747e-06, "loss": 3.8374, "step": 1518 }, { "epoch": 0.29, "learning_rate": 2.9159771210676835e-06, "loss": 3.8047, "step": 1519 }, { "epoch": 0.29, "learning_rate": 2.9159199237368923e-06, "loss": 3.7461, "step": 1520 }, { "epoch": 0.29, "learning_rate": 2.915862726406101e-06, "loss": 3.5912, "step": 1521 }, { "epoch": 0.29, "learning_rate": 2.91580552907531e-06, "loss": 3.644, "step": 1522 }, { "epoch": 0.29, "learning_rate": 2.9157483317445186e-06, "loss": 3.7423, "step": 1523 }, { "epoch": 0.29, "learning_rate": 2.9156911344137274e-06, "loss": 3.7316, "step": 1524 }, { "epoch": 0.29, "learning_rate": 2.9156339370829362e-06, "loss": 3.7307, "step": 1525 }, { "epoch": 0.29, "learning_rate": 2.915576739752145e-06, "loss": 3.7942, "step": 1526 }, { "epoch": 0.29, "learning_rate": 2.915519542421354e-06, "loss": 3.8229, "step": 1527 }, { "epoch": 0.29, "learning_rate": 2.9154623450905626e-06, "loss": 3.6278, "step": 1528 }, { "epoch": 0.29, "learning_rate": 2.915405147759771e-06, "loss": 3.8046, "step": 1529 }, { "epoch": 0.29, "learning_rate": 2.91534795042898e-06, "loss": 3.6412, "step": 1530 }, { "epoch": 0.29, "learning_rate": 2.915290753098189e-06, "loss": 3.8251, "step": 1531 }, { "epoch": 0.29, "learning_rate": 2.9152335557673974e-06, "loss": 3.7636, "step": 1532 }, { "epoch": 0.29, "learning_rate": 2.9151763584366066e-06, "loss": 3.7284, "step": 1533 }, { "epoch": 0.29, "learning_rate": 2.915119161105815e-06, "loss": 3.8176, "step": 1534 }, { "epoch": 0.29, "learning_rate": 2.9150619637750237e-06, "loss": 3.5079, "step": 1535 }, { "epoch": 0.29, "learning_rate": 2.915004766444233e-06, "loss": 3.5886, "step": 1536 }, { "epoch": 0.29, "learning_rate": 2.9149475691134413e-06, "loss": 3.7893, "step": 1537 }, { "epoch": 0.29, "learning_rate": 2.91489037178265e-06, "loss": 3.6751, "step": 1538 }, { "epoch": 0.29, "learning_rate": 2.914833174451859e-06, "loss": 3.818, "step": 1539 }, { "epoch": 0.29, "learning_rate": 2.9147759771210677e-06, "loss": 3.7607, "step": 1540 }, { "epoch": 0.29, "learning_rate": 2.9147187797902765e-06, "loss": 3.6005, "step": 1541 }, { "epoch": 0.29, "learning_rate": 2.9146615824594853e-06, "loss": 3.6684, "step": 1542 }, { "epoch": 0.29, "learning_rate": 2.914604385128694e-06, "loss": 3.7033, "step": 1543 }, { "epoch": 0.29, "learning_rate": 2.914547187797903e-06, "loss": 3.4945, "step": 1544 }, { "epoch": 0.29, "learning_rate": 2.9144899904671117e-06, "loss": 3.9457, "step": 1545 }, { "epoch": 0.29, "learning_rate": 2.9144327931363205e-06, "loss": 3.6748, "step": 1546 }, { "epoch": 0.29, "learning_rate": 2.9143755958055293e-06, "loss": 3.757, "step": 1547 }, { "epoch": 0.29, "learning_rate": 2.914318398474738e-06, "loss": 3.3821, "step": 1548 }, { "epoch": 0.3, "learning_rate": 2.9142612011439464e-06, "loss": 3.9807, "step": 1549 }, { "epoch": 0.3, "learning_rate": 2.9142040038131557e-06, "loss": 3.7661, "step": 1550 }, { "epoch": 0.3, "learning_rate": 2.914146806482364e-06, "loss": 3.4506, "step": 1551 }, { "epoch": 0.3, "learning_rate": 2.914089609151573e-06, "loss": 3.6884, "step": 1552 }, { "epoch": 0.3, "learning_rate": 2.914032411820782e-06, "loss": 3.594, "step": 1553 }, { "epoch": 0.3, "learning_rate": 2.9139752144899904e-06, "loss": 3.7826, "step": 1554 }, { "epoch": 0.3, "learning_rate": 2.9139180171591996e-06, "loss": 3.705, "step": 1555 }, { "epoch": 0.3, "learning_rate": 2.913860819828408e-06, "loss": 3.4923, "step": 1556 }, { "epoch": 0.3, "learning_rate": 2.9138036224976168e-06, "loss": 3.5843, "step": 1557 }, { "epoch": 0.3, "learning_rate": 2.9137464251668256e-06, "loss": 3.7534, "step": 1558 }, { "epoch": 0.3, "learning_rate": 2.9136892278360344e-06, "loss": 3.7237, "step": 1559 }, { "epoch": 0.3, "learning_rate": 2.913632030505243e-06, "loss": 3.3877, "step": 1560 }, { "epoch": 0.3, "learning_rate": 2.913574833174452e-06, "loss": 3.9438, "step": 1561 }, { "epoch": 0.3, "learning_rate": 2.9135176358436608e-06, "loss": 3.653, "step": 1562 }, { "epoch": 0.3, "learning_rate": 2.9134604385128695e-06, "loss": 3.6958, "step": 1563 }, { "epoch": 0.3, "learning_rate": 2.9134032411820783e-06, "loss": 3.6065, "step": 1564 }, { "epoch": 0.3, "learning_rate": 2.913346043851287e-06, "loss": 3.8591, "step": 1565 }, { "epoch": 0.3, "learning_rate": 2.9132888465204955e-06, "loss": 3.6273, "step": 1566 }, { "epoch": 0.3, "learning_rate": 2.9132316491897047e-06, "loss": 3.6819, "step": 1567 }, { "epoch": 0.3, "learning_rate": 2.913174451858913e-06, "loss": 3.682, "step": 1568 }, { "epoch": 0.3, "learning_rate": 2.913117254528122e-06, "loss": 3.4695, "step": 1569 }, { "epoch": 0.3, "learning_rate": 2.913060057197331e-06, "loss": 3.545, "step": 1570 }, { "epoch": 0.3, "learning_rate": 2.9130028598665395e-06, "loss": 3.9946, "step": 1571 }, { "epoch": 0.3, "learning_rate": 2.9129456625357487e-06, "loss": 3.6669, "step": 1572 }, { "epoch": 0.3, "learning_rate": 2.912888465204957e-06, "loss": 3.6756, "step": 1573 }, { "epoch": 0.3, "learning_rate": 2.912831267874166e-06, "loss": 3.9031, "step": 1574 }, { "epoch": 0.3, "learning_rate": 2.9127740705433746e-06, "loss": 3.5856, "step": 1575 }, { "epoch": 0.3, "learning_rate": 2.9127168732125834e-06, "loss": 3.6006, "step": 1576 }, { "epoch": 0.3, "learning_rate": 2.9126596758817922e-06, "loss": 3.6434, "step": 1577 }, { "epoch": 0.3, "learning_rate": 2.912602478551001e-06, "loss": 3.5609, "step": 1578 }, { "epoch": 0.3, "learning_rate": 2.91254528122021e-06, "loss": 3.9452, "step": 1579 }, { "epoch": 0.3, "learning_rate": 2.9124880838894186e-06, "loss": 3.4853, "step": 1580 }, { "epoch": 0.3, "learning_rate": 2.9124308865586274e-06, "loss": 3.7177, "step": 1581 }, { "epoch": 0.3, "learning_rate": 2.912373689227836e-06, "loss": 3.6808, "step": 1582 }, { "epoch": 0.3, "learning_rate": 2.9123164918970446e-06, "loss": 3.7915, "step": 1583 }, { "epoch": 0.3, "learning_rate": 2.912259294566254e-06, "loss": 3.5489, "step": 1584 }, { "epoch": 0.3, "learning_rate": 2.9122020972354626e-06, "loss": 3.5412, "step": 1585 }, { "epoch": 0.3, "learning_rate": 2.912144899904671e-06, "loss": 3.7801, "step": 1586 }, { "epoch": 0.3, "learning_rate": 2.91208770257388e-06, "loss": 3.7503, "step": 1587 }, { "epoch": 0.3, "learning_rate": 2.9120305052430885e-06, "loss": 3.6028, "step": 1588 }, { "epoch": 0.3, "learning_rate": 2.9119733079122978e-06, "loss": 3.8361, "step": 1589 }, { "epoch": 0.3, "learning_rate": 2.911916110581506e-06, "loss": 3.734, "step": 1590 }, { "epoch": 0.3, "learning_rate": 2.911858913250715e-06, "loss": 3.8842, "step": 1591 }, { "epoch": 0.3, "learning_rate": 2.911801715919924e-06, "loss": 3.7609, "step": 1592 }, { "epoch": 0.3, "learning_rate": 2.9117445185891325e-06, "loss": 3.8059, "step": 1593 }, { "epoch": 0.3, "learning_rate": 2.9116873212583413e-06, "loss": 3.636, "step": 1594 }, { "epoch": 0.3, "learning_rate": 2.91163012392755e-06, "loss": 3.8503, "step": 1595 }, { "epoch": 0.3, "learning_rate": 2.911572926596759e-06, "loss": 3.6416, "step": 1596 }, { "epoch": 0.3, "learning_rate": 2.9115157292659677e-06, "loss": 3.7955, "step": 1597 }, { "epoch": 0.3, "learning_rate": 2.9114585319351765e-06, "loss": 3.6905, "step": 1598 }, { "epoch": 0.3, "learning_rate": 2.9114013346043853e-06, "loss": 3.8596, "step": 1599 }, { "epoch": 0.3, "learning_rate": 2.9113441372735936e-06, "loss": 3.5767, "step": 1600 }, { "epoch": 0.3, "learning_rate": 2.911286939942803e-06, "loss": 3.7491, "step": 1601 }, { "epoch": 0.31, "learning_rate": 2.9112297426120117e-06, "loss": 3.5664, "step": 1602 }, { "epoch": 0.31, "learning_rate": 2.91117254528122e-06, "loss": 3.5569, "step": 1603 }, { "epoch": 0.31, "learning_rate": 2.9111153479504292e-06, "loss": 3.6911, "step": 1604 }, { "epoch": 0.31, "learning_rate": 2.9110581506196376e-06, "loss": 3.5815, "step": 1605 }, { "epoch": 0.31, "learning_rate": 2.911000953288847e-06, "loss": 3.6308, "step": 1606 }, { "epoch": 0.31, "learning_rate": 2.910943755958055e-06, "loss": 3.5254, "step": 1607 }, { "epoch": 0.31, "learning_rate": 2.910886558627264e-06, "loss": 3.6669, "step": 1608 }, { "epoch": 0.31, "learning_rate": 2.9108293612964732e-06, "loss": 3.5842, "step": 1609 }, { "epoch": 0.31, "learning_rate": 2.9107721639656816e-06, "loss": 3.5627, "step": 1610 }, { "epoch": 0.31, "learning_rate": 2.9107149666348904e-06, "loss": 3.6541, "step": 1611 }, { "epoch": 0.31, "learning_rate": 2.910657769304099e-06, "loss": 3.6432, "step": 1612 }, { "epoch": 0.31, "learning_rate": 2.910600571973308e-06, "loss": 3.778, "step": 1613 }, { "epoch": 0.31, "learning_rate": 2.9105433746425168e-06, "loss": 3.7795, "step": 1614 }, { "epoch": 0.31, "learning_rate": 2.9104861773117256e-06, "loss": 3.681, "step": 1615 }, { "epoch": 0.31, "learning_rate": 2.9104289799809343e-06, "loss": 3.5557, "step": 1616 }, { "epoch": 0.31, "learning_rate": 2.9103717826501427e-06, "loss": 3.5844, "step": 1617 }, { "epoch": 0.31, "learning_rate": 2.910314585319352e-06, "loss": 3.7371, "step": 1618 }, { "epoch": 0.31, "learning_rate": 2.9102573879885607e-06, "loss": 3.7429, "step": 1619 }, { "epoch": 0.31, "learning_rate": 2.9102001906577695e-06, "loss": 3.604, "step": 1620 }, { "epoch": 0.31, "learning_rate": 2.9101429933269783e-06, "loss": 3.4514, "step": 1621 }, { "epoch": 0.31, "learning_rate": 2.9100857959961867e-06, "loss": 3.7025, "step": 1622 }, { "epoch": 0.31, "learning_rate": 2.910028598665396e-06, "loss": 3.9684, "step": 1623 }, { "epoch": 0.31, "learning_rate": 2.9099714013346043e-06, "loss": 3.7603, "step": 1624 }, { "epoch": 0.31, "learning_rate": 2.909914204003813e-06, "loss": 3.6058, "step": 1625 }, { "epoch": 0.31, "learning_rate": 2.9098570066730223e-06, "loss": 3.711, "step": 1626 }, { "epoch": 0.31, "learning_rate": 2.9097998093422307e-06, "loss": 3.4008, "step": 1627 }, { "epoch": 0.31, "learning_rate": 2.9097426120114395e-06, "loss": 3.3929, "step": 1628 }, { "epoch": 0.31, "learning_rate": 2.9096854146806482e-06, "loss": 3.5446, "step": 1629 }, { "epoch": 0.31, "learning_rate": 2.909628217349857e-06, "loss": 3.7502, "step": 1630 }, { "epoch": 0.31, "learning_rate": 2.909571020019066e-06, "loss": 3.8111, "step": 1631 }, { "epoch": 0.31, "learning_rate": 2.9095138226882746e-06, "loss": 3.8261, "step": 1632 }, { "epoch": 0.31, "learning_rate": 2.9094566253574834e-06, "loss": 3.6415, "step": 1633 }, { "epoch": 0.31, "learning_rate": 2.9093994280266922e-06, "loss": 3.6948, "step": 1634 }, { "epoch": 0.31, "learning_rate": 2.909342230695901e-06, "loss": 3.8957, "step": 1635 }, { "epoch": 0.31, "learning_rate": 2.90928503336511e-06, "loss": 3.6724, "step": 1636 }, { "epoch": 0.31, "learning_rate": 2.9092278360343186e-06, "loss": 3.7513, "step": 1637 }, { "epoch": 0.31, "learning_rate": 2.9091706387035274e-06, "loss": 3.5644, "step": 1638 }, { "epoch": 0.31, "learning_rate": 2.9091134413727358e-06, "loss": 3.8026, "step": 1639 }, { "epoch": 0.31, "learning_rate": 2.909056244041945e-06, "loss": 3.6345, "step": 1640 }, { "epoch": 0.31, "learning_rate": 2.9089990467111538e-06, "loss": 3.6602, "step": 1641 }, { "epoch": 0.31, "learning_rate": 2.908941849380362e-06, "loss": 3.631, "step": 1642 }, { "epoch": 0.31, "learning_rate": 2.9088846520495714e-06, "loss": 3.7489, "step": 1643 }, { "epoch": 0.31, "learning_rate": 2.9088274547187797e-06, "loss": 3.7391, "step": 1644 }, { "epoch": 0.31, "learning_rate": 2.9087702573879885e-06, "loss": 3.6739, "step": 1645 }, { "epoch": 0.31, "learning_rate": 2.9087130600571973e-06, "loss": 3.6541, "step": 1646 }, { "epoch": 0.31, "learning_rate": 2.908655862726406e-06, "loss": 3.5124, "step": 1647 }, { "epoch": 0.31, "learning_rate": 2.908598665395615e-06, "loss": 3.5406, "step": 1648 }, { "epoch": 0.31, "learning_rate": 2.9085414680648237e-06, "loss": 3.822, "step": 1649 }, { "epoch": 0.31, "learning_rate": 2.9084842707340325e-06, "loss": 3.5479, "step": 1650 }, { "epoch": 0.31, "learning_rate": 2.9084270734032413e-06, "loss": 3.6229, "step": 1651 }, { "epoch": 0.31, "learning_rate": 2.90836987607245e-06, "loss": 3.8284, "step": 1652 }, { "epoch": 0.31, "learning_rate": 2.908312678741659e-06, "loss": 3.8971, "step": 1653 }, { "epoch": 0.32, "learning_rate": 2.9082554814108677e-06, "loss": 3.8263, "step": 1654 }, { "epoch": 0.32, "learning_rate": 2.9081982840800765e-06, "loss": 3.5802, "step": 1655 }, { "epoch": 0.32, "learning_rate": 2.908141086749285e-06, "loss": 3.782, "step": 1656 }, { "epoch": 0.32, "learning_rate": 2.908083889418494e-06, "loss": 3.7266, "step": 1657 }, { "epoch": 0.32, "learning_rate": 2.908026692087703e-06, "loss": 3.8606, "step": 1658 }, { "epoch": 0.32, "learning_rate": 2.9079694947569112e-06, "loss": 3.6313, "step": 1659 }, { "epoch": 0.32, "learning_rate": 2.9079122974261204e-06, "loss": 3.7638, "step": 1660 }, { "epoch": 0.32, "learning_rate": 2.907855100095329e-06, "loss": 3.7774, "step": 1661 }, { "epoch": 0.32, "learning_rate": 2.9077979027645376e-06, "loss": 3.663, "step": 1662 }, { "epoch": 0.32, "learning_rate": 2.9077407054337464e-06, "loss": 3.8608, "step": 1663 }, { "epoch": 0.32, "learning_rate": 2.907683508102955e-06, "loss": 3.9575, "step": 1664 }, { "epoch": 0.32, "learning_rate": 2.907626310772164e-06, "loss": 3.6513, "step": 1665 }, { "epoch": 0.32, "learning_rate": 2.9075691134413728e-06, "loss": 3.8051, "step": 1666 }, { "epoch": 0.32, "learning_rate": 2.9075119161105816e-06, "loss": 3.6867, "step": 1667 }, { "epoch": 0.32, "learning_rate": 2.9074547187797904e-06, "loss": 3.8667, "step": 1668 }, { "epoch": 0.32, "learning_rate": 2.907397521448999e-06, "loss": 3.5457, "step": 1669 }, { "epoch": 0.32, "learning_rate": 2.907340324118208e-06, "loss": 3.7045, "step": 1670 }, { "epoch": 0.32, "learning_rate": 2.9072831267874167e-06, "loss": 3.5241, "step": 1671 }, { "epoch": 0.32, "learning_rate": 2.9072259294566255e-06, "loss": 3.6544, "step": 1672 }, { "epoch": 0.32, "learning_rate": 2.9071687321258343e-06, "loss": 3.6403, "step": 1673 }, { "epoch": 0.32, "learning_rate": 2.907111534795043e-06, "loss": 3.5417, "step": 1674 }, { "epoch": 0.32, "learning_rate": 2.907054337464252e-06, "loss": 3.6569, "step": 1675 }, { "epoch": 0.32, "learning_rate": 2.9069971401334603e-06, "loss": 3.655, "step": 1676 }, { "epoch": 0.32, "learning_rate": 2.9069399428026695e-06, "loss": 3.4555, "step": 1677 }, { "epoch": 0.32, "learning_rate": 2.906882745471878e-06, "loss": 3.8695, "step": 1678 }, { "epoch": 0.32, "learning_rate": 2.9068255481410867e-06, "loss": 3.5734, "step": 1679 }, { "epoch": 0.32, "learning_rate": 2.906768350810296e-06, "loss": 3.4498, "step": 1680 }, { "epoch": 0.32, "learning_rate": 2.9067111534795043e-06, "loss": 3.7129, "step": 1681 }, { "epoch": 0.32, "learning_rate": 2.906653956148713e-06, "loss": 3.5006, "step": 1682 }, { "epoch": 0.32, "learning_rate": 2.906596758817922e-06, "loss": 3.8508, "step": 1683 }, { "epoch": 0.32, "learning_rate": 2.9065395614871306e-06, "loss": 3.7572, "step": 1684 }, { "epoch": 0.32, "learning_rate": 2.9064823641563394e-06, "loss": 3.8304, "step": 1685 }, { "epoch": 0.32, "learning_rate": 2.9064251668255482e-06, "loss": 3.6946, "step": 1686 }, { "epoch": 0.32, "learning_rate": 2.906367969494757e-06, "loss": 3.583, "step": 1687 }, { "epoch": 0.32, "learning_rate": 2.906310772163966e-06, "loss": 3.7533, "step": 1688 }, { "epoch": 0.32, "learning_rate": 2.9062535748331746e-06, "loss": 3.687, "step": 1689 }, { "epoch": 0.32, "learning_rate": 2.9061963775023834e-06, "loss": 3.6362, "step": 1690 }, { "epoch": 0.32, "learning_rate": 2.906139180171592e-06, "loss": 3.8031, "step": 1691 }, { "epoch": 0.32, "learning_rate": 2.906081982840801e-06, "loss": 3.9055, "step": 1692 }, { "epoch": 0.32, "learning_rate": 2.9060247855100094e-06, "loss": 3.5266, "step": 1693 }, { "epoch": 0.32, "learning_rate": 2.9059675881792186e-06, "loss": 3.3423, "step": 1694 }, { "epoch": 0.32, "learning_rate": 2.905910390848427e-06, "loss": 3.5307, "step": 1695 }, { "epoch": 0.32, "learning_rate": 2.9058531935176357e-06, "loss": 3.5875, "step": 1696 }, { "epoch": 0.32, "learning_rate": 2.905795996186845e-06, "loss": 3.8823, "step": 1697 }, { "epoch": 0.32, "learning_rate": 2.9057387988560533e-06, "loss": 3.5589, "step": 1698 }, { "epoch": 0.32, "learning_rate": 2.905681601525262e-06, "loss": 3.6386, "step": 1699 }, { "epoch": 0.32, "learning_rate": 2.905624404194471e-06, "loss": 3.6026, "step": 1700 }, { "epoch": 0.32, "learning_rate": 2.9055672068636797e-06, "loss": 3.5854, "step": 1701 }, { "epoch": 0.32, "learning_rate": 2.9055100095328885e-06, "loss": 3.8006, "step": 1702 }, { "epoch": 0.32, "learning_rate": 2.9054528122020973e-06, "loss": 3.6364, "step": 1703 }, { "epoch": 0.32, "learning_rate": 2.905395614871306e-06, "loss": 3.6924, "step": 1704 }, { "epoch": 0.32, "learning_rate": 2.905338417540515e-06, "loss": 3.6517, "step": 1705 }, { "epoch": 0.32, "learning_rate": 2.9052812202097237e-06, "loss": 3.6997, "step": 1706 }, { "epoch": 0.33, "learning_rate": 2.9052240228789325e-06, "loss": 3.8875, "step": 1707 }, { "epoch": 0.33, "learning_rate": 2.9051668255481413e-06, "loss": 3.6845, "step": 1708 }, { "epoch": 0.33, "learning_rate": 2.90510962821735e-06, "loss": 3.8868, "step": 1709 }, { "epoch": 0.33, "learning_rate": 2.9050524308865584e-06, "loss": 3.6627, "step": 1710 }, { "epoch": 0.33, "learning_rate": 2.9049952335557676e-06, "loss": 3.9377, "step": 1711 }, { "epoch": 0.33, "learning_rate": 2.904938036224976e-06, "loss": 3.542, "step": 1712 }, { "epoch": 0.33, "learning_rate": 2.904880838894185e-06, "loss": 3.756, "step": 1713 }, { "epoch": 0.33, "learning_rate": 2.904823641563394e-06, "loss": 3.7534, "step": 1714 }, { "epoch": 0.33, "learning_rate": 2.9047664442326024e-06, "loss": 3.6594, "step": 1715 }, { "epoch": 0.33, "learning_rate": 2.904709246901811e-06, "loss": 3.531, "step": 1716 }, { "epoch": 0.33, "learning_rate": 2.90465204957102e-06, "loss": 3.6877, "step": 1717 }, { "epoch": 0.33, "learning_rate": 2.9045948522402288e-06, "loss": 3.7618, "step": 1718 }, { "epoch": 0.33, "learning_rate": 2.904537654909438e-06, "loss": 3.7831, "step": 1719 }, { "epoch": 0.33, "learning_rate": 2.9044804575786464e-06, "loss": 3.4731, "step": 1720 }, { "epoch": 0.33, "learning_rate": 2.904423260247855e-06, "loss": 3.6525, "step": 1721 }, { "epoch": 0.33, "learning_rate": 2.904366062917064e-06, "loss": 3.5715, "step": 1722 }, { "epoch": 0.33, "learning_rate": 2.9043088655862727e-06, "loss": 3.5023, "step": 1723 }, { "epoch": 0.33, "learning_rate": 2.9042516682554815e-06, "loss": 3.8134, "step": 1724 }, { "epoch": 0.33, "learning_rate": 2.9041944709246903e-06, "loss": 3.6691, "step": 1725 }, { "epoch": 0.33, "learning_rate": 2.904137273593899e-06, "loss": 3.6874, "step": 1726 }, { "epoch": 0.33, "learning_rate": 2.9040800762631075e-06, "loss": 3.6566, "step": 1727 }, { "epoch": 0.33, "learning_rate": 2.9040228789323167e-06, "loss": 3.6122, "step": 1728 }, { "epoch": 0.33, "learning_rate": 2.9039656816015255e-06, "loss": 3.7017, "step": 1729 }, { "epoch": 0.33, "learning_rate": 2.903908484270734e-06, "loss": 3.8048, "step": 1730 }, { "epoch": 0.33, "learning_rate": 2.903851286939943e-06, "loss": 3.6478, "step": 1731 }, { "epoch": 0.33, "learning_rate": 2.9037940896091515e-06, "loss": 3.6689, "step": 1732 }, { "epoch": 0.33, "learning_rate": 2.9037368922783603e-06, "loss": 3.8088, "step": 1733 }, { "epoch": 0.33, "learning_rate": 2.903679694947569e-06, "loss": 3.6878, "step": 1734 }, { "epoch": 0.33, "learning_rate": 2.903622497616778e-06, "loss": 3.8059, "step": 1735 }, { "epoch": 0.33, "learning_rate": 2.903565300285987e-06, "loss": 3.8024, "step": 1736 }, { "epoch": 0.33, "learning_rate": 2.9035081029551954e-06, "loss": 3.8329, "step": 1737 }, { "epoch": 0.33, "learning_rate": 2.9034509056244042e-06, "loss": 3.4417, "step": 1738 }, { "epoch": 0.33, "learning_rate": 2.903393708293613e-06, "loss": 3.6786, "step": 1739 }, { "epoch": 0.33, "learning_rate": 2.903336510962822e-06, "loss": 3.5724, "step": 1740 }, { "epoch": 0.33, "learning_rate": 2.9032793136320306e-06, "loss": 3.4928, "step": 1741 }, { "epoch": 0.33, "learning_rate": 2.9032221163012394e-06, "loss": 3.8494, "step": 1742 }, { "epoch": 0.33, "learning_rate": 2.903164918970448e-06, "loss": 3.6382, "step": 1743 }, { "epoch": 0.33, "learning_rate": 2.9031077216396566e-06, "loss": 3.565, "step": 1744 }, { "epoch": 0.33, "learning_rate": 2.9030505243088658e-06, "loss": 3.6343, "step": 1745 }, { "epoch": 0.33, "learning_rate": 2.9029933269780746e-06, "loss": 3.782, "step": 1746 }, { "epoch": 0.33, "learning_rate": 2.902936129647283e-06, "loss": 3.8442, "step": 1747 }, { "epoch": 0.33, "learning_rate": 2.902878932316492e-06, "loss": 3.8339, "step": 1748 }, { "epoch": 0.33, "learning_rate": 2.9028217349857005e-06, "loss": 3.8124, "step": 1749 }, { "epoch": 0.33, "learning_rate": 2.9027645376549093e-06, "loss": 3.6022, "step": 1750 }, { "epoch": 0.33, "learning_rate": 2.902707340324118e-06, "loss": 3.753, "step": 1751 }, { "epoch": 0.33, "learning_rate": 2.902650142993327e-06, "loss": 3.6099, "step": 1752 }, { "epoch": 0.33, "learning_rate": 2.902592945662536e-06, "loss": 3.6277, "step": 1753 }, { "epoch": 0.33, "learning_rate": 2.9025357483317445e-06, "loss": 3.7147, "step": 1754 }, { "epoch": 0.33, "learning_rate": 2.9024785510009533e-06, "loss": 3.6093, "step": 1755 }, { "epoch": 0.33, "learning_rate": 2.902421353670162e-06, "loss": 3.808, "step": 1756 }, { "epoch": 0.33, "learning_rate": 2.902364156339371e-06, "loss": 3.8408, "step": 1757 }, { "epoch": 0.33, "learning_rate": 2.9023069590085797e-06, "loss": 3.6985, "step": 1758 }, { "epoch": 0.34, "learning_rate": 2.9022497616777885e-06, "loss": 3.7768, "step": 1759 }, { "epoch": 0.34, "learning_rate": 2.9021925643469973e-06, "loss": 3.525, "step": 1760 }, { "epoch": 0.34, "learning_rate": 2.902135367016206e-06, "loss": 3.3757, "step": 1761 }, { "epoch": 0.34, "learning_rate": 2.902078169685415e-06, "loss": 3.8263, "step": 1762 }, { "epoch": 0.34, "learning_rate": 2.9020209723546237e-06, "loss": 3.682, "step": 1763 }, { "epoch": 0.34, "learning_rate": 2.901963775023832e-06, "loss": 3.8401, "step": 1764 }, { "epoch": 0.34, "learning_rate": 2.9019065776930412e-06, "loss": 3.6072, "step": 1765 }, { "epoch": 0.34, "learning_rate": 2.9018493803622496e-06, "loss": 3.8392, "step": 1766 }, { "epoch": 0.34, "learning_rate": 2.901792183031459e-06, "loss": 3.7286, "step": 1767 }, { "epoch": 0.34, "learning_rate": 2.9017349857006676e-06, "loss": 3.7656, "step": 1768 }, { "epoch": 0.34, "learning_rate": 2.901677788369876e-06, "loss": 3.6438, "step": 1769 }, { "epoch": 0.34, "learning_rate": 2.901620591039085e-06, "loss": 3.8569, "step": 1770 }, { "epoch": 0.34, "learning_rate": 2.9015633937082936e-06, "loss": 3.5421, "step": 1771 }, { "epoch": 0.34, "learning_rate": 2.9015061963775024e-06, "loss": 3.7103, "step": 1772 }, { "epoch": 0.34, "learning_rate": 2.901448999046711e-06, "loss": 3.7911, "step": 1773 }, { "epoch": 0.34, "learning_rate": 2.90139180171592e-06, "loss": 3.7943, "step": 1774 }, { "epoch": 0.34, "learning_rate": 2.9013346043851288e-06, "loss": 3.7583, "step": 1775 }, { "epoch": 0.34, "learning_rate": 2.9012774070543375e-06, "loss": 3.6676, "step": 1776 }, { "epoch": 0.34, "learning_rate": 2.9012202097235463e-06, "loss": 3.6142, "step": 1777 }, { "epoch": 0.34, "learning_rate": 2.901163012392755e-06, "loss": 3.8487, "step": 1778 }, { "epoch": 0.34, "learning_rate": 2.901105815061964e-06, "loss": 3.7301, "step": 1779 }, { "epoch": 0.34, "learning_rate": 2.9010486177311727e-06, "loss": 3.6695, "step": 1780 }, { "epoch": 0.34, "learning_rate": 2.900991420400381e-06, "loss": 3.6547, "step": 1781 }, { "epoch": 0.34, "learning_rate": 2.9009342230695903e-06, "loss": 3.5978, "step": 1782 }, { "epoch": 0.34, "learning_rate": 2.9008770257387987e-06, "loss": 3.7006, "step": 1783 }, { "epoch": 0.34, "learning_rate": 2.900819828408008e-06, "loss": 3.7212, "step": 1784 }, { "epoch": 0.34, "learning_rate": 2.9007626310772167e-06, "loss": 3.7164, "step": 1785 }, { "epoch": 0.34, "learning_rate": 2.900705433746425e-06, "loss": 3.531, "step": 1786 }, { "epoch": 0.34, "learning_rate": 2.9006482364156343e-06, "loss": 3.605, "step": 1787 }, { "epoch": 0.34, "learning_rate": 2.9005910390848427e-06, "loss": 3.6571, "step": 1788 }, { "epoch": 0.34, "learning_rate": 2.9005338417540514e-06, "loss": 3.699, "step": 1789 }, { "epoch": 0.34, "learning_rate": 2.9004766444232602e-06, "loss": 3.4122, "step": 1790 }, { "epoch": 0.34, "learning_rate": 2.900419447092469e-06, "loss": 3.7802, "step": 1791 }, { "epoch": 0.34, "learning_rate": 2.900362249761678e-06, "loss": 3.5652, "step": 1792 }, { "epoch": 0.34, "learning_rate": 2.9003050524308866e-06, "loss": 3.5089, "step": 1793 }, { "epoch": 0.34, "learning_rate": 2.9002478551000954e-06, "loss": 3.6948, "step": 1794 }, { "epoch": 0.34, "learning_rate": 2.900190657769304e-06, "loss": 3.6731, "step": 1795 }, { "epoch": 0.34, "learning_rate": 2.900133460438513e-06, "loss": 3.808, "step": 1796 }, { "epoch": 0.34, "learning_rate": 2.900076263107722e-06, "loss": 3.8377, "step": 1797 }, { "epoch": 0.34, "learning_rate": 2.90001906577693e-06, "loss": 3.6908, "step": 1798 }, { "epoch": 0.34, "learning_rate": 2.8999618684461394e-06, "loss": 3.4685, "step": 1799 }, { "epoch": 0.34, "learning_rate": 2.8999046711153478e-06, "loss": 3.6929, "step": 1800 }, { "epoch": 0.34, "learning_rate": 2.899847473784557e-06, "loss": 3.7314, "step": 1801 }, { "epoch": 0.34, "learning_rate": 2.8997902764537658e-06, "loss": 3.5732, "step": 1802 }, { "epoch": 0.34, "learning_rate": 2.899733079122974e-06, "loss": 3.7971, "step": 1803 }, { "epoch": 0.34, "learning_rate": 2.8996758817921834e-06, "loss": 3.6199, "step": 1804 }, { "epoch": 0.34, "learning_rate": 2.8996186844613917e-06, "loss": 3.4654, "step": 1805 }, { "epoch": 0.34, "learning_rate": 2.8995614871306005e-06, "loss": 3.4784, "step": 1806 }, { "epoch": 0.34, "learning_rate": 2.8995042897998097e-06, "loss": 3.4492, "step": 1807 }, { "epoch": 0.34, "learning_rate": 2.899447092469018e-06, "loss": 3.6931, "step": 1808 }, { "epoch": 0.34, "learning_rate": 2.899389895138227e-06, "loss": 3.779, "step": 1809 }, { "epoch": 0.34, "learning_rate": 2.8993326978074357e-06, "loss": 3.7492, "step": 1810 }, { "epoch": 0.34, "learning_rate": 2.8992755004766445e-06, "loss": 3.7771, "step": 1811 }, { "epoch": 0.35, "learning_rate": 2.8992183031458533e-06, "loss": 3.6202, "step": 1812 }, { "epoch": 0.35, "learning_rate": 2.899161105815062e-06, "loss": 3.8263, "step": 1813 }, { "epoch": 0.35, "learning_rate": 2.899103908484271e-06, "loss": 3.7279, "step": 1814 }, { "epoch": 0.35, "learning_rate": 2.8990467111534792e-06, "loss": 3.7347, "step": 1815 }, { "epoch": 0.35, "learning_rate": 2.8989895138226885e-06, "loss": 3.5622, "step": 1816 }, { "epoch": 0.35, "learning_rate": 2.8989323164918972e-06, "loss": 3.7297, "step": 1817 }, { "epoch": 0.35, "learning_rate": 2.898875119161106e-06, "loss": 3.5643, "step": 1818 }, { "epoch": 0.35, "learning_rate": 2.898817921830315e-06, "loss": 3.6682, "step": 1819 }, { "epoch": 0.35, "learning_rate": 2.898760724499523e-06, "loss": 3.5794, "step": 1820 }, { "epoch": 0.35, "learning_rate": 2.8987035271687324e-06, "loss": 3.65, "step": 1821 }, { "epoch": 0.35, "learning_rate": 2.898646329837941e-06, "loss": 3.7417, "step": 1822 }, { "epoch": 0.35, "learning_rate": 2.8985891325071496e-06, "loss": 3.516, "step": 1823 }, { "epoch": 0.35, "learning_rate": 2.898531935176359e-06, "loss": 3.7516, "step": 1824 }, { "epoch": 0.35, "learning_rate": 2.898474737845567e-06, "loss": 3.5545, "step": 1825 }, { "epoch": 0.35, "learning_rate": 2.898417540514776e-06, "loss": 3.6409, "step": 1826 }, { "epoch": 0.35, "learning_rate": 2.8983603431839848e-06, "loss": 3.7852, "step": 1827 }, { "epoch": 0.35, "learning_rate": 2.8983031458531936e-06, "loss": 3.3654, "step": 1828 }, { "epoch": 0.35, "learning_rate": 2.8982459485224023e-06, "loss": 3.8013, "step": 1829 }, { "epoch": 0.35, "learning_rate": 2.898188751191611e-06, "loss": 3.5901, "step": 1830 }, { "epoch": 0.35, "learning_rate": 2.89813155386082e-06, "loss": 3.5425, "step": 1831 }, { "epoch": 0.35, "learning_rate": 2.8980743565300287e-06, "loss": 3.6804, "step": 1832 }, { "epoch": 0.35, "learning_rate": 2.8980171591992375e-06, "loss": 3.631, "step": 1833 }, { "epoch": 0.35, "learning_rate": 2.8979599618684463e-06, "loss": 3.6038, "step": 1834 }, { "epoch": 0.35, "learning_rate": 2.897902764537655e-06, "loss": 3.7056, "step": 1835 }, { "epoch": 0.35, "learning_rate": 2.897845567206864e-06, "loss": 3.8091, "step": 1836 }, { "epoch": 0.35, "learning_rate": 2.8977883698760723e-06, "loss": 3.8018, "step": 1837 }, { "epoch": 0.35, "learning_rate": 2.8977311725452815e-06, "loss": 3.8144, "step": 1838 }, { "epoch": 0.35, "learning_rate": 2.89767397521449e-06, "loss": 3.7603, "step": 1839 }, { "epoch": 0.35, "learning_rate": 2.8976167778836987e-06, "loss": 3.7198, "step": 1840 }, { "epoch": 0.35, "learning_rate": 2.897559580552908e-06, "loss": 3.7327, "step": 1841 }, { "epoch": 0.35, "learning_rate": 2.8975023832221162e-06, "loss": 3.8716, "step": 1842 }, { "epoch": 0.35, "learning_rate": 2.897445185891325e-06, "loss": 3.6245, "step": 1843 }, { "epoch": 0.35, "learning_rate": 2.897387988560534e-06, "loss": 3.8065, "step": 1844 }, { "epoch": 0.35, "learning_rate": 2.8973307912297426e-06, "loss": 3.5802, "step": 1845 }, { "epoch": 0.35, "learning_rate": 2.8972735938989514e-06, "loss": 3.554, "step": 1846 }, { "epoch": 0.35, "learning_rate": 2.8972163965681602e-06, "loss": 3.8768, "step": 1847 }, { "epoch": 0.35, "learning_rate": 2.897159199237369e-06, "loss": 3.3602, "step": 1848 }, { "epoch": 0.35, "learning_rate": 2.897102001906578e-06, "loss": 3.4708, "step": 1849 }, { "epoch": 0.35, "learning_rate": 2.8970448045757866e-06, "loss": 3.5403, "step": 1850 }, { "epoch": 0.35, "learning_rate": 2.8969876072449954e-06, "loss": 3.7281, "step": 1851 }, { "epoch": 0.35, "learning_rate": 2.896930409914204e-06, "loss": 3.4687, "step": 1852 }, { "epoch": 0.35, "learning_rate": 2.896873212583413e-06, "loss": 3.7881, "step": 1853 }, { "epoch": 0.35, "learning_rate": 2.8968160152526213e-06, "loss": 3.7092, "step": 1854 }, { "epoch": 0.35, "learning_rate": 2.8967588179218306e-06, "loss": 3.6096, "step": 1855 }, { "epoch": 0.35, "learning_rate": 2.8967016205910394e-06, "loss": 3.4786, "step": 1856 }, { "epoch": 0.35, "learning_rate": 2.8966444232602477e-06, "loss": 3.4464, "step": 1857 }, { "epoch": 0.35, "learning_rate": 2.896587225929457e-06, "loss": 3.8097, "step": 1858 }, { "epoch": 0.35, "learning_rate": 2.8965300285986653e-06, "loss": 3.3704, "step": 1859 }, { "epoch": 0.35, "learning_rate": 2.896472831267874e-06, "loss": 3.7062, "step": 1860 }, { "epoch": 0.35, "learning_rate": 2.896415633937083e-06, "loss": 3.5933, "step": 1861 }, { "epoch": 0.35, "learning_rate": 2.8963584366062917e-06, "loss": 3.7777, "step": 1862 }, { "epoch": 0.35, "learning_rate": 2.8963012392755005e-06, "loss": 3.6775, "step": 1863 }, { "epoch": 0.36, "learning_rate": 2.8962440419447093e-06, "loss": 3.4978, "step": 1864 }, { "epoch": 0.36, "learning_rate": 2.896186844613918e-06, "loss": 3.7223, "step": 1865 }, { "epoch": 0.36, "learning_rate": 2.896129647283127e-06, "loss": 3.5849, "step": 1866 }, { "epoch": 0.36, "learning_rate": 2.8960724499523357e-06, "loss": 3.7679, "step": 1867 }, { "epoch": 0.36, "learning_rate": 2.8960152526215445e-06, "loss": 3.6956, "step": 1868 }, { "epoch": 0.36, "learning_rate": 2.8959580552907533e-06, "loss": 3.6309, "step": 1869 }, { "epoch": 0.36, "learning_rate": 2.895900857959962e-06, "loss": 3.8282, "step": 1870 }, { "epoch": 0.36, "learning_rate": 2.8958436606291704e-06, "loss": 3.6566, "step": 1871 }, { "epoch": 0.36, "learning_rate": 2.8957864632983796e-06, "loss": 3.6071, "step": 1872 }, { "epoch": 0.36, "learning_rate": 2.8957292659675884e-06, "loss": 3.7993, "step": 1873 }, { "epoch": 0.36, "learning_rate": 2.895672068636797e-06, "loss": 3.7466, "step": 1874 }, { "epoch": 0.36, "learning_rate": 2.895614871306006e-06, "loss": 3.7354, "step": 1875 }, { "epoch": 0.36, "learning_rate": 2.8955576739752144e-06, "loss": 3.7948, "step": 1876 }, { "epoch": 0.36, "learning_rate": 2.895500476644423e-06, "loss": 3.5737, "step": 1877 }, { "epoch": 0.36, "learning_rate": 2.895443279313632e-06, "loss": 3.8084, "step": 1878 }, { "epoch": 0.36, "learning_rate": 2.8953860819828408e-06, "loss": 3.5561, "step": 1879 }, { "epoch": 0.36, "learning_rate": 2.8953288846520496e-06, "loss": 3.5782, "step": 1880 }, { "epoch": 0.36, "learning_rate": 2.8952716873212584e-06, "loss": 3.5369, "step": 1881 }, { "epoch": 0.36, "learning_rate": 2.895214489990467e-06, "loss": 3.5675, "step": 1882 }, { "epoch": 0.36, "learning_rate": 2.895157292659676e-06, "loss": 3.7013, "step": 1883 }, { "epoch": 0.36, "learning_rate": 2.8951000953288847e-06, "loss": 3.3676, "step": 1884 }, { "epoch": 0.36, "learning_rate": 2.8950428979980935e-06, "loss": 3.424, "step": 1885 }, { "epoch": 0.36, "learning_rate": 2.8949857006673023e-06, "loss": 3.6566, "step": 1886 }, { "epoch": 0.36, "learning_rate": 2.894928503336511e-06, "loss": 3.7005, "step": 1887 }, { "epoch": 0.36, "learning_rate": 2.8948713060057195e-06, "loss": 3.7544, "step": 1888 }, { "epoch": 0.36, "learning_rate": 2.8948141086749287e-06, "loss": 3.8797, "step": 1889 }, { "epoch": 0.36, "learning_rate": 2.8947569113441375e-06, "loss": 3.5663, "step": 1890 }, { "epoch": 0.36, "learning_rate": 2.894699714013346e-06, "loss": 3.8504, "step": 1891 }, { "epoch": 0.36, "learning_rate": 2.894642516682555e-06, "loss": 3.8478, "step": 1892 }, { "epoch": 0.36, "learning_rate": 2.8945853193517635e-06, "loss": 3.531, "step": 1893 }, { "epoch": 0.36, "learning_rate": 2.8945281220209723e-06, "loss": 3.7317, "step": 1894 }, { "epoch": 0.36, "learning_rate": 2.8944709246901815e-06, "loss": 3.8383, "step": 1895 }, { "epoch": 0.36, "learning_rate": 2.89441372735939e-06, "loss": 3.4908, "step": 1896 }, { "epoch": 0.36, "learning_rate": 2.894356530028599e-06, "loss": 3.6974, "step": 1897 }, { "epoch": 0.36, "learning_rate": 2.8942993326978074e-06, "loss": 3.8691, "step": 1898 }, { "epoch": 0.36, "learning_rate": 2.8942421353670162e-06, "loss": 3.6705, "step": 1899 }, { "epoch": 0.36, "learning_rate": 2.894184938036225e-06, "loss": 3.5357, "step": 1900 }, { "epoch": 0.36, "learning_rate": 2.894127740705434e-06, "loss": 3.6398, "step": 1901 }, { "epoch": 0.36, "learning_rate": 2.8940705433746426e-06, "loss": 3.3967, "step": 1902 }, { "epoch": 0.36, "learning_rate": 2.8940133460438514e-06, "loss": 3.6185, "step": 1903 }, { "epoch": 0.36, "learning_rate": 2.89395614871306e-06, "loss": 3.5953, "step": 1904 }, { "epoch": 0.36, "learning_rate": 2.893898951382269e-06, "loss": 3.8783, "step": 1905 }, { "epoch": 0.36, "learning_rate": 2.8938417540514778e-06, "loss": 3.5636, "step": 1906 }, { "epoch": 0.36, "learning_rate": 2.8937845567206866e-06, "loss": 3.8018, "step": 1907 }, { "epoch": 0.36, "learning_rate": 2.893727359389895e-06, "loss": 3.5603, "step": 1908 }, { "epoch": 0.36, "learning_rate": 2.893670162059104e-06, "loss": 3.6682, "step": 1909 }, { "epoch": 0.36, "learning_rate": 2.8936129647283125e-06, "loss": 3.4299, "step": 1910 }, { "epoch": 0.36, "learning_rate": 2.8935557673975213e-06, "loss": 3.7495, "step": 1911 }, { "epoch": 0.36, "learning_rate": 2.8934985700667305e-06, "loss": 3.8658, "step": 1912 }, { "epoch": 0.36, "learning_rate": 2.893441372735939e-06, "loss": 3.5526, "step": 1913 }, { "epoch": 0.36, "learning_rate": 2.893384175405148e-06, "loss": 3.4994, "step": 1914 }, { "epoch": 0.36, "learning_rate": 2.8933269780743565e-06, "loss": 3.6609, "step": 1915 }, { "epoch": 0.36, "learning_rate": 2.8932697807435653e-06, "loss": 3.805, "step": 1916 }, { "epoch": 0.37, "learning_rate": 2.893212583412774e-06, "loss": 3.6264, "step": 1917 }, { "epoch": 0.37, "learning_rate": 2.893155386081983e-06, "loss": 3.4856, "step": 1918 }, { "epoch": 0.37, "learning_rate": 2.8930981887511917e-06, "loss": 3.7567, "step": 1919 }, { "epoch": 0.37, "learning_rate": 2.8930409914204005e-06, "loss": 3.689, "step": 1920 }, { "epoch": 0.37, "learning_rate": 2.8929837940896093e-06, "loss": 3.9835, "step": 1921 }, { "epoch": 0.37, "learning_rate": 2.892926596758818e-06, "loss": 3.6755, "step": 1922 }, { "epoch": 0.37, "learning_rate": 2.892869399428027e-06, "loss": 3.5629, "step": 1923 }, { "epoch": 0.37, "learning_rate": 2.8928122020972356e-06, "loss": 3.7442, "step": 1924 }, { "epoch": 0.37, "learning_rate": 2.892755004766444e-06, "loss": 3.5507, "step": 1925 }, { "epoch": 0.37, "learning_rate": 2.8926978074356532e-06, "loss": 3.406, "step": 1926 }, { "epoch": 0.37, "learning_rate": 2.8926406101048616e-06, "loss": 3.5372, "step": 1927 }, { "epoch": 0.37, "learning_rate": 2.8925834127740704e-06, "loss": 3.8035, "step": 1928 }, { "epoch": 0.37, "learning_rate": 2.8925262154432796e-06, "loss": 3.7798, "step": 1929 }, { "epoch": 0.37, "learning_rate": 2.892469018112488e-06, "loss": 3.7062, "step": 1930 }, { "epoch": 0.37, "learning_rate": 2.892411820781697e-06, "loss": 3.5973, "step": 1931 }, { "epoch": 0.37, "learning_rate": 2.8923546234509056e-06, "loss": 3.4706, "step": 1932 }, { "epoch": 0.37, "learning_rate": 2.8922974261201144e-06, "loss": 3.7098, "step": 1933 }, { "epoch": 0.37, "learning_rate": 2.892240228789323e-06, "loss": 3.8563, "step": 1934 }, { "epoch": 0.37, "learning_rate": 2.892183031458532e-06, "loss": 3.5863, "step": 1935 }, { "epoch": 0.37, "learning_rate": 2.8921258341277407e-06, "loss": 3.5114, "step": 1936 }, { "epoch": 0.37, "learning_rate": 2.8920686367969495e-06, "loss": 3.3894, "step": 1937 }, { "epoch": 0.37, "learning_rate": 2.8920114394661583e-06, "loss": 3.6737, "step": 1938 }, { "epoch": 0.37, "learning_rate": 2.891954242135367e-06, "loss": 3.8734, "step": 1939 }, { "epoch": 0.37, "learning_rate": 2.891897044804576e-06, "loss": 3.6524, "step": 1940 }, { "epoch": 0.37, "learning_rate": 2.8918398474737847e-06, "loss": 3.6627, "step": 1941 }, { "epoch": 0.37, "learning_rate": 2.891782650142993e-06, "loss": 3.7534, "step": 1942 }, { "epoch": 0.37, "learning_rate": 2.8917254528122023e-06, "loss": 3.7708, "step": 1943 }, { "epoch": 0.37, "learning_rate": 2.891668255481411e-06, "loss": 3.4609, "step": 1944 }, { "epoch": 0.37, "learning_rate": 2.8916110581506195e-06, "loss": 3.6413, "step": 1945 }, { "epoch": 0.37, "learning_rate": 2.8915538608198287e-06, "loss": 3.6435, "step": 1946 }, { "epoch": 0.37, "learning_rate": 2.891496663489037e-06, "loss": 3.6637, "step": 1947 }, { "epoch": 0.37, "learning_rate": 2.8914394661582463e-06, "loss": 3.742, "step": 1948 }, { "epoch": 0.37, "learning_rate": 2.8913822688274546e-06, "loss": 3.6946, "step": 1949 }, { "epoch": 0.37, "learning_rate": 2.8913250714966634e-06, "loss": 3.7863, "step": 1950 }, { "epoch": 0.37, "learning_rate": 2.8912678741658727e-06, "loss": 3.612, "step": 1951 }, { "epoch": 0.37, "learning_rate": 2.891210676835081e-06, "loss": 3.6243, "step": 1952 }, { "epoch": 0.37, "learning_rate": 2.89115347950429e-06, "loss": 3.5626, "step": 1953 }, { "epoch": 0.37, "learning_rate": 2.8910962821734986e-06, "loss": 3.9437, "step": 1954 }, { "epoch": 0.37, "learning_rate": 2.8910390848427074e-06, "loss": 3.5323, "step": 1955 }, { "epoch": 0.37, "learning_rate": 2.890981887511916e-06, "loss": 3.8842, "step": 1956 }, { "epoch": 0.37, "learning_rate": 2.890924690181125e-06, "loss": 3.5441, "step": 1957 }, { "epoch": 0.37, "learning_rate": 2.8908674928503338e-06, "loss": 3.7242, "step": 1958 }, { "epoch": 0.37, "learning_rate": 2.890810295519542e-06, "loss": 3.8507, "step": 1959 }, { "epoch": 0.37, "learning_rate": 2.8907530981887514e-06, "loss": 3.8529, "step": 1960 }, { "epoch": 0.37, "learning_rate": 2.89069590085796e-06, "loss": 3.8667, "step": 1961 }, { "epoch": 0.37, "learning_rate": 2.890638703527169e-06, "loss": 3.6017, "step": 1962 }, { "epoch": 0.37, "learning_rate": 2.8905815061963778e-06, "loss": 3.5487, "step": 1963 }, { "epoch": 0.37, "learning_rate": 2.890524308865586e-06, "loss": 3.6777, "step": 1964 }, { "epoch": 0.37, "learning_rate": 2.8904671115347953e-06, "loss": 3.7026, "step": 1965 }, { "epoch": 0.37, "learning_rate": 2.8904099142040037e-06, "loss": 3.719, "step": 1966 }, { "epoch": 0.37, "learning_rate": 2.8903527168732125e-06, "loss": 3.6713, "step": 1967 }, { "epoch": 0.37, "learning_rate": 2.8902955195424217e-06, "loss": 3.5733, "step": 1968 }, { "epoch": 0.38, "learning_rate": 2.89023832221163e-06, "loss": 3.5482, "step": 1969 }, { "epoch": 0.38, "learning_rate": 2.890181124880839e-06, "loss": 3.5427, "step": 1970 }, { "epoch": 0.38, "learning_rate": 2.8901239275500477e-06, "loss": 3.7208, "step": 1971 }, { "epoch": 0.38, "learning_rate": 2.8900667302192565e-06, "loss": 3.5409, "step": 1972 }, { "epoch": 0.38, "learning_rate": 2.8900095328884653e-06, "loss": 3.7035, "step": 1973 }, { "epoch": 0.38, "learning_rate": 2.889952335557674e-06, "loss": 3.4665, "step": 1974 }, { "epoch": 0.38, "learning_rate": 2.889895138226883e-06, "loss": 3.6825, "step": 1975 }, { "epoch": 0.38, "learning_rate": 2.8898379408960912e-06, "loss": 3.6892, "step": 1976 }, { "epoch": 0.38, "learning_rate": 2.8897807435653004e-06, "loss": 3.4264, "step": 1977 }, { "epoch": 0.38, "learning_rate": 2.8897235462345092e-06, "loss": 3.4025, "step": 1978 }, { "epoch": 0.38, "learning_rate": 2.889666348903718e-06, "loss": 3.6964, "step": 1979 }, { "epoch": 0.38, "learning_rate": 2.889609151572927e-06, "loss": 3.5035, "step": 1980 }, { "epoch": 0.38, "learning_rate": 2.889551954242135e-06, "loss": 3.5575, "step": 1981 }, { "epoch": 0.38, "learning_rate": 2.8894947569113444e-06, "loss": 3.596, "step": 1982 }, { "epoch": 0.38, "learning_rate": 2.889437559580553e-06, "loss": 3.6735, "step": 1983 }, { "epoch": 0.38, "learning_rate": 2.8893803622497616e-06, "loss": 3.5184, "step": 1984 }, { "epoch": 0.38, "learning_rate": 2.889323164918971e-06, "loss": 3.7117, "step": 1985 }, { "epoch": 0.38, "learning_rate": 2.889265967588179e-06, "loss": 3.6623, "step": 1986 }, { "epoch": 0.38, "learning_rate": 2.889208770257388e-06, "loss": 3.6337, "step": 1987 }, { "epoch": 0.38, "learning_rate": 2.8891515729265968e-06, "loss": 3.6173, "step": 1988 }, { "epoch": 0.38, "learning_rate": 2.8890943755958055e-06, "loss": 3.6312, "step": 1989 }, { "epoch": 0.38, "learning_rate": 2.8890371782650143e-06, "loss": 3.7536, "step": 1990 }, { "epoch": 0.38, "learning_rate": 2.888979980934223e-06, "loss": 3.6136, "step": 1991 }, { "epoch": 0.38, "learning_rate": 2.888922783603432e-06, "loss": 3.8712, "step": 1992 }, { "epoch": 0.38, "learning_rate": 2.8888655862726407e-06, "loss": 3.8299, "step": 1993 }, { "epoch": 0.38, "learning_rate": 2.8888083889418495e-06, "loss": 3.5087, "step": 1994 }, { "epoch": 0.38, "learning_rate": 2.8887511916110583e-06, "loss": 3.8572, "step": 1995 }, { "epoch": 0.38, "learning_rate": 2.888693994280267e-06, "loss": 3.6368, "step": 1996 }, { "epoch": 0.38, "learning_rate": 2.888636796949476e-06, "loss": 3.4645, "step": 1997 }, { "epoch": 0.38, "learning_rate": 2.8885795996186843e-06, "loss": 3.638, "step": 1998 }, { "epoch": 0.38, "learning_rate": 2.8885224022878935e-06, "loss": 3.6348, "step": 1999 }, { "epoch": 0.38, "learning_rate": 2.8884652049571023e-06, "loss": 3.4796, "step": 2000 }, { "epoch": 0.38, "learning_rate": 2.8884080076263107e-06, "loss": 3.7602, "step": 2001 }, { "epoch": 0.38, "learning_rate": 2.88835081029552e-06, "loss": 3.5738, "step": 2002 }, { "epoch": 0.38, "learning_rate": 2.8882936129647282e-06, "loss": 3.6263, "step": 2003 }, { "epoch": 0.38, "learning_rate": 2.888236415633937e-06, "loss": 3.613, "step": 2004 }, { "epoch": 0.38, "learning_rate": 2.888179218303146e-06, "loss": 3.7656, "step": 2005 }, { "epoch": 0.38, "learning_rate": 2.8881220209723546e-06, "loss": 3.5265, "step": 2006 }, { "epoch": 0.38, "learning_rate": 2.8880648236415634e-06, "loss": 3.3811, "step": 2007 }, { "epoch": 0.38, "learning_rate": 2.888007626310772e-06, "loss": 3.6138, "step": 2008 }, { "epoch": 0.38, "learning_rate": 2.887950428979981e-06, "loss": 3.5536, "step": 2009 }, { "epoch": 0.38, "learning_rate": 2.88789323164919e-06, "loss": 3.5274, "step": 2010 }, { "epoch": 0.38, "learning_rate": 2.8878360343183986e-06, "loss": 3.7341, "step": 2011 }, { "epoch": 0.38, "learning_rate": 2.8877788369876074e-06, "loss": 3.9158, "step": 2012 }, { "epoch": 0.38, "learning_rate": 2.887721639656816e-06, "loss": 3.9364, "step": 2013 }, { "epoch": 0.38, "learning_rate": 2.887664442326025e-06, "loss": 3.6148, "step": 2014 }, { "epoch": 0.38, "learning_rate": 2.8876072449952333e-06, "loss": 3.681, "step": 2015 }, { "epoch": 0.38, "learning_rate": 2.8875500476644426e-06, "loss": 3.4648, "step": 2016 }, { "epoch": 0.38, "learning_rate": 2.8874928503336514e-06, "loss": 3.6126, "step": 2017 }, { "epoch": 0.38, "learning_rate": 2.8874356530028597e-06, "loss": 3.7202, "step": 2018 }, { "epoch": 0.38, "learning_rate": 2.887378455672069e-06, "loss": 3.7329, "step": 2019 }, { "epoch": 0.38, "learning_rate": 2.8873212583412773e-06, "loss": 3.6927, "step": 2020 }, { "epoch": 0.38, "learning_rate": 2.887264061010486e-06, "loss": 3.5882, "step": 2021 }, { "epoch": 0.39, "learning_rate": 2.887206863679695e-06, "loss": 3.6504, "step": 2022 }, { "epoch": 0.39, "learning_rate": 2.8871496663489037e-06, "loss": 3.6778, "step": 2023 }, { "epoch": 0.39, "learning_rate": 2.8870924690181125e-06, "loss": 3.5556, "step": 2024 }, { "epoch": 0.39, "learning_rate": 2.8870352716873213e-06, "loss": 3.8747, "step": 2025 }, { "epoch": 0.39, "learning_rate": 2.88697807435653e-06, "loss": 3.8674, "step": 2026 }, { "epoch": 0.39, "learning_rate": 2.886920877025739e-06, "loss": 3.673, "step": 2027 }, { "epoch": 0.39, "learning_rate": 2.8868636796949477e-06, "loss": 3.8378, "step": 2028 }, { "epoch": 0.39, "learning_rate": 2.8868064823641565e-06, "loss": 3.4374, "step": 2029 }, { "epoch": 0.39, "learning_rate": 2.8867492850333652e-06, "loss": 3.72, "step": 2030 }, { "epoch": 0.39, "learning_rate": 2.886692087702574e-06, "loss": 3.8239, "step": 2031 }, { "epoch": 0.39, "learning_rate": 2.886634890371783e-06, "loss": 3.59, "step": 2032 }, { "epoch": 0.39, "learning_rate": 2.8865776930409916e-06, "loss": 3.6, "step": 2033 }, { "epoch": 0.39, "learning_rate": 2.8865204957102004e-06, "loss": 3.5494, "step": 2034 }, { "epoch": 0.39, "learning_rate": 2.886463298379409e-06, "loss": 3.6576, "step": 2035 }, { "epoch": 0.39, "learning_rate": 2.886406101048618e-06, "loss": 3.7511, "step": 2036 }, { "epoch": 0.39, "learning_rate": 2.8863489037178264e-06, "loss": 3.6837, "step": 2037 }, { "epoch": 0.39, "learning_rate": 2.886291706387035e-06, "loss": 3.8152, "step": 2038 }, { "epoch": 0.39, "learning_rate": 2.8862345090562444e-06, "loss": 3.598, "step": 2039 }, { "epoch": 0.39, "learning_rate": 2.8861773117254528e-06, "loss": 3.8489, "step": 2040 }, { "epoch": 0.39, "learning_rate": 2.8861201143946616e-06, "loss": 3.8959, "step": 2041 }, { "epoch": 0.39, "learning_rate": 2.8860629170638704e-06, "loss": 3.8946, "step": 2042 }, { "epoch": 0.39, "learning_rate": 2.886005719733079e-06, "loss": 3.8356, "step": 2043 }, { "epoch": 0.39, "learning_rate": 2.885948522402288e-06, "loss": 3.5322, "step": 2044 }, { "epoch": 0.39, "learning_rate": 2.8858913250714967e-06, "loss": 3.5591, "step": 2045 }, { "epoch": 0.39, "learning_rate": 2.8858341277407055e-06, "loss": 3.5088, "step": 2046 }, { "epoch": 0.39, "learning_rate": 2.8857769304099143e-06, "loss": 3.5022, "step": 2047 }, { "epoch": 0.39, "learning_rate": 2.885719733079123e-06, "loss": 3.4525, "step": 2048 }, { "epoch": 0.39, "learning_rate": 2.885662535748332e-06, "loss": 3.8187, "step": 2049 }, { "epoch": 0.39, "learning_rate": 2.8856053384175407e-06, "loss": 3.5681, "step": 2050 }, { "epoch": 0.39, "learning_rate": 2.8855481410867495e-06, "loss": 3.6394, "step": 2051 }, { "epoch": 0.39, "learning_rate": 2.885490943755958e-06, "loss": 3.6289, "step": 2052 }, { "epoch": 0.39, "learning_rate": 2.885433746425167e-06, "loss": 3.5723, "step": 2053 }, { "epoch": 0.39, "learning_rate": 2.8853765490943755e-06, "loss": 3.6548, "step": 2054 }, { "epoch": 0.39, "learning_rate": 2.8853193517635842e-06, "loss": 3.7763, "step": 2055 }, { "epoch": 0.39, "learning_rate": 2.8852621544327935e-06, "loss": 3.6171, "step": 2056 }, { "epoch": 0.39, "learning_rate": 2.885204957102002e-06, "loss": 3.4147, "step": 2057 }, { "epoch": 0.39, "learning_rate": 2.8851477597712106e-06, "loss": 3.5095, "step": 2058 }, { "epoch": 0.39, "learning_rate": 2.8850905624404194e-06, "loss": 3.5884, "step": 2059 }, { "epoch": 0.39, "learning_rate": 2.8850333651096282e-06, "loss": 3.5694, "step": 2060 }, { "epoch": 0.39, "learning_rate": 2.884976167778837e-06, "loss": 3.7273, "step": 2061 }, { "epoch": 0.39, "learning_rate": 2.884918970448046e-06, "loss": 3.6739, "step": 2062 }, { "epoch": 0.39, "learning_rate": 2.8848617731172546e-06, "loss": 3.5047, "step": 2063 }, { "epoch": 0.39, "learning_rate": 2.8848045757864634e-06, "loss": 3.7826, "step": 2064 }, { "epoch": 0.39, "learning_rate": 2.884747378455672e-06, "loss": 3.8685, "step": 2065 }, { "epoch": 0.39, "learning_rate": 2.884690181124881e-06, "loss": 3.4542, "step": 2066 }, { "epoch": 0.39, "learning_rate": 2.8846329837940898e-06, "loss": 3.3244, "step": 2067 }, { "epoch": 0.39, "learning_rate": 2.8845757864632986e-06, "loss": 3.5487, "step": 2068 }, { "epoch": 0.39, "learning_rate": 2.884518589132507e-06, "loss": 3.6896, "step": 2069 }, { "epoch": 0.39, "learning_rate": 2.884461391801716e-06, "loss": 3.6591, "step": 2070 }, { "epoch": 0.39, "learning_rate": 2.884404194470925e-06, "loss": 3.7846, "step": 2071 }, { "epoch": 0.39, "learning_rate": 2.8843469971401333e-06, "loss": 3.8066, "step": 2072 }, { "epoch": 0.39, "learning_rate": 2.8842897998093425e-06, "loss": 3.801, "step": 2073 }, { "epoch": 0.4, "learning_rate": 2.884232602478551e-06, "loss": 3.3937, "step": 2074 }, { "epoch": 0.4, "learning_rate": 2.8841754051477597e-06, "loss": 3.8593, "step": 2075 }, { "epoch": 0.4, "learning_rate": 2.8841182078169685e-06, "loss": 3.7228, "step": 2076 }, { "epoch": 0.4, "learning_rate": 2.8840610104861773e-06, "loss": 3.7993, "step": 2077 }, { "epoch": 0.4, "learning_rate": 2.8840038131553865e-06, "loss": 3.8793, "step": 2078 }, { "epoch": 0.4, "learning_rate": 2.883946615824595e-06, "loss": 3.4701, "step": 2079 }, { "epoch": 0.4, "learning_rate": 2.8838894184938037e-06, "loss": 3.6072, "step": 2080 }, { "epoch": 0.4, "learning_rate": 2.8838322211630125e-06, "loss": 3.677, "step": 2081 }, { "epoch": 0.4, "learning_rate": 2.8837750238322213e-06, "loss": 3.8042, "step": 2082 }, { "epoch": 0.4, "learning_rate": 2.88371782650143e-06, "loss": 3.5724, "step": 2083 }, { "epoch": 0.4, "learning_rate": 2.883660629170639e-06, "loss": 3.8013, "step": 2084 }, { "epoch": 0.4, "learning_rate": 2.8836034318398476e-06, "loss": 3.7646, "step": 2085 }, { "epoch": 0.4, "learning_rate": 2.883546234509056e-06, "loss": 3.7488, "step": 2086 }, { "epoch": 0.4, "learning_rate": 2.8834890371782652e-06, "loss": 3.9018, "step": 2087 }, { "epoch": 0.4, "learning_rate": 2.883431839847474e-06, "loss": 3.4192, "step": 2088 }, { "epoch": 0.4, "learning_rate": 2.8833746425166824e-06, "loss": 3.8099, "step": 2089 }, { "epoch": 0.4, "learning_rate": 2.8833174451858916e-06, "loss": 3.7439, "step": 2090 }, { "epoch": 0.4, "learning_rate": 2.8832602478551e-06, "loss": 3.8363, "step": 2091 }, { "epoch": 0.4, "learning_rate": 2.883203050524309e-06, "loss": 3.6336, "step": 2092 }, { "epoch": 0.4, "learning_rate": 2.8831458531935176e-06, "loss": 3.5786, "step": 2093 }, { "epoch": 0.4, "learning_rate": 2.8830886558627264e-06, "loss": 3.8483, "step": 2094 }, { "epoch": 0.4, "learning_rate": 2.8830314585319356e-06, "loss": 3.6774, "step": 2095 }, { "epoch": 0.4, "learning_rate": 2.882974261201144e-06, "loss": 3.6232, "step": 2096 }, { "epoch": 0.4, "learning_rate": 2.8829170638703527e-06, "loss": 3.7414, "step": 2097 }, { "epoch": 0.4, "learning_rate": 2.8828598665395615e-06, "loss": 3.5881, "step": 2098 }, { "epoch": 0.4, "learning_rate": 2.8828026692087703e-06, "loss": 3.733, "step": 2099 }, { "epoch": 0.4, "learning_rate": 2.882745471877979e-06, "loss": 3.5455, "step": 2100 }, { "epoch": 0.4, "learning_rate": 2.882688274547188e-06, "loss": 3.5271, "step": 2101 }, { "epoch": 0.4, "learning_rate": 2.8826310772163967e-06, "loss": 3.5684, "step": 2102 }, { "epoch": 0.4, "learning_rate": 2.882573879885605e-06, "loss": 3.4596, "step": 2103 }, { "epoch": 0.4, "learning_rate": 2.8825166825548143e-06, "loss": 3.3039, "step": 2104 }, { "epoch": 0.4, "learning_rate": 2.882459485224023e-06, "loss": 3.8156, "step": 2105 }, { "epoch": 0.4, "learning_rate": 2.8824022878932315e-06, "loss": 3.5622, "step": 2106 }, { "epoch": 0.4, "learning_rate": 2.8823450905624407e-06, "loss": 3.6609, "step": 2107 }, { "epoch": 0.4, "learning_rate": 2.882287893231649e-06, "loss": 3.5663, "step": 2108 }, { "epoch": 0.4, "learning_rate": 2.8822306959008583e-06, "loss": 3.8672, "step": 2109 }, { "epoch": 0.4, "learning_rate": 2.8821734985700666e-06, "loss": 3.7863, "step": 2110 }, { "epoch": 0.4, "learning_rate": 2.8821163012392754e-06, "loss": 3.6533, "step": 2111 }, { "epoch": 0.4, "learning_rate": 2.8820591039084846e-06, "loss": 3.5966, "step": 2112 }, { "epoch": 0.4, "learning_rate": 2.882001906577693e-06, "loss": 3.6601, "step": 2113 }, { "epoch": 0.4, "learning_rate": 2.881944709246902e-06, "loss": 3.4654, "step": 2114 }, { "epoch": 0.4, "learning_rate": 2.8818875119161106e-06, "loss": 3.7132, "step": 2115 }, { "epoch": 0.4, "learning_rate": 2.8818303145853194e-06, "loss": 3.4879, "step": 2116 }, { "epoch": 0.4, "learning_rate": 2.881773117254528e-06, "loss": 3.6403, "step": 2117 }, { "epoch": 0.4, "learning_rate": 2.881715919923737e-06, "loss": 3.8112, "step": 2118 }, { "epoch": 0.4, "learning_rate": 2.8816587225929458e-06, "loss": 3.6598, "step": 2119 }, { "epoch": 0.4, "learning_rate": 2.8816015252621546e-06, "loss": 3.7806, "step": 2120 }, { "epoch": 0.4, "learning_rate": 2.8815443279313634e-06, "loss": 3.6395, "step": 2121 }, { "epoch": 0.4, "learning_rate": 2.881487130600572e-06, "loss": 3.5622, "step": 2122 }, { "epoch": 0.4, "learning_rate": 2.8814299332697805e-06, "loss": 3.7984, "step": 2123 }, { "epoch": 0.4, "learning_rate": 2.8813727359389897e-06, "loss": 3.5669, "step": 2124 }, { "epoch": 0.4, "learning_rate": 2.881315538608198e-06, "loss": 3.7309, "step": 2125 }, { "epoch": 0.4, "learning_rate": 2.8812583412774073e-06, "loss": 3.7061, "step": 2126 }, { "epoch": 0.41, "learning_rate": 2.881201143946616e-06, "loss": 3.7994, "step": 2127 }, { "epoch": 0.41, "learning_rate": 2.8811439466158245e-06, "loss": 3.5686, "step": 2128 }, { "epoch": 0.41, "learning_rate": 2.8810867492850337e-06, "loss": 3.6987, "step": 2129 }, { "epoch": 0.41, "learning_rate": 2.881029551954242e-06, "loss": 3.4377, "step": 2130 }, { "epoch": 0.41, "learning_rate": 2.880972354623451e-06, "loss": 3.4313, "step": 2131 }, { "epoch": 0.41, "learning_rate": 2.8809151572926597e-06, "loss": 3.8487, "step": 2132 }, { "epoch": 0.41, "learning_rate": 2.8808579599618685e-06, "loss": 3.5483, "step": 2133 }, { "epoch": 0.41, "learning_rate": 2.8808007626310773e-06, "loss": 3.8548, "step": 2134 }, { "epoch": 0.41, "learning_rate": 2.880743565300286e-06, "loss": 3.6566, "step": 2135 }, { "epoch": 0.41, "learning_rate": 2.880686367969495e-06, "loss": 3.8274, "step": 2136 }, { "epoch": 0.41, "learning_rate": 2.8806291706387036e-06, "loss": 3.6567, "step": 2137 }, { "epoch": 0.41, "learning_rate": 2.8805719733079124e-06, "loss": 3.7994, "step": 2138 }, { "epoch": 0.41, "learning_rate": 2.8805147759771212e-06, "loss": 3.4836, "step": 2139 }, { "epoch": 0.41, "learning_rate": 2.8804575786463296e-06, "loss": 3.6761, "step": 2140 }, { "epoch": 0.41, "learning_rate": 2.880400381315539e-06, "loss": 3.7026, "step": 2141 }, { "epoch": 0.41, "learning_rate": 2.880343183984747e-06, "loss": 3.9155, "step": 2142 }, { "epoch": 0.41, "learning_rate": 2.8802859866539564e-06, "loss": 3.6415, "step": 2143 }, { "epoch": 0.41, "learning_rate": 2.880228789323165e-06, "loss": 3.7287, "step": 2144 }, { "epoch": 0.41, "learning_rate": 2.8801715919923736e-06, "loss": 3.6884, "step": 2145 }, { "epoch": 0.41, "learning_rate": 2.880114394661583e-06, "loss": 3.5522, "step": 2146 }, { "epoch": 0.41, "learning_rate": 2.880057197330791e-06, "loss": 3.5987, "step": 2147 }, { "epoch": 0.41, "learning_rate": 2.88e-06, "loss": 3.6424, "step": 2148 }, { "epoch": 0.41, "learning_rate": 2.8799428026692087e-06, "loss": 3.614, "step": 2149 }, { "epoch": 0.41, "learning_rate": 2.8798856053384175e-06, "loss": 3.6125, "step": 2150 }, { "epoch": 0.41, "learning_rate": 2.8798284080076263e-06, "loss": 3.8118, "step": 2151 }, { "epoch": 0.41, "learning_rate": 2.879771210676835e-06, "loss": 3.8459, "step": 2152 }, { "epoch": 0.41, "learning_rate": 2.879714013346044e-06, "loss": 3.7421, "step": 2153 }, { "epoch": 0.41, "learning_rate": 2.8796568160152527e-06, "loss": 3.4138, "step": 2154 }, { "epoch": 0.41, "learning_rate": 2.8795996186844615e-06, "loss": 3.6109, "step": 2155 }, { "epoch": 0.41, "learning_rate": 2.8795424213536703e-06, "loss": 3.6509, "step": 2156 }, { "epoch": 0.41, "learning_rate": 2.879485224022879e-06, "loss": 3.7807, "step": 2157 }, { "epoch": 0.41, "learning_rate": 2.879428026692088e-06, "loss": 3.7005, "step": 2158 }, { "epoch": 0.41, "learning_rate": 2.8793708293612967e-06, "loss": 3.5923, "step": 2159 }, { "epoch": 0.41, "learning_rate": 2.8793136320305055e-06, "loss": 3.6631, "step": 2160 }, { "epoch": 0.41, "learning_rate": 2.8792564346997143e-06, "loss": 3.5979, "step": 2161 }, { "epoch": 0.41, "learning_rate": 2.8791992373689226e-06, "loss": 3.6967, "step": 2162 }, { "epoch": 0.41, "learning_rate": 2.879142040038132e-06, "loss": 3.7571, "step": 2163 }, { "epoch": 0.41, "learning_rate": 2.8790848427073402e-06, "loss": 3.5256, "step": 2164 }, { "epoch": 0.41, "learning_rate": 2.879027645376549e-06, "loss": 3.5904, "step": 2165 }, { "epoch": 0.41, "learning_rate": 2.8789704480457582e-06, "loss": 3.5926, "step": 2166 }, { "epoch": 0.41, "learning_rate": 2.8789132507149666e-06, "loss": 3.4667, "step": 2167 }, { "epoch": 0.41, "learning_rate": 2.8788560533841754e-06, "loss": 3.4246, "step": 2168 }, { "epoch": 0.41, "learning_rate": 2.878798856053384e-06, "loss": 3.5369, "step": 2169 }, { "epoch": 0.41, "learning_rate": 2.878741658722593e-06, "loss": 3.5988, "step": 2170 }, { "epoch": 0.41, "learning_rate": 2.878684461391802e-06, "loss": 3.6622, "step": 2171 }, { "epoch": 0.41, "learning_rate": 2.8786272640610106e-06, "loss": 3.7293, "step": 2172 }, { "epoch": 0.41, "learning_rate": 2.8785700667302194e-06, "loss": 3.6247, "step": 2173 }, { "epoch": 0.41, "learning_rate": 2.878512869399428e-06, "loss": 3.6136, "step": 2174 }, { "epoch": 0.41, "learning_rate": 2.878455672068637e-06, "loss": 3.531, "step": 2175 }, { "epoch": 0.41, "learning_rate": 2.8783984747378458e-06, "loss": 3.8194, "step": 2176 }, { "epoch": 0.41, "learning_rate": 2.8783412774070546e-06, "loss": 3.7001, "step": 2177 }, { "epoch": 0.41, "learning_rate": 2.8782840800762633e-06, "loss": 3.6885, "step": 2178 }, { "epoch": 0.42, "learning_rate": 2.8782268827454717e-06, "loss": 3.8184, "step": 2179 }, { "epoch": 0.42, "learning_rate": 2.878169685414681e-06, "loss": 3.6188, "step": 2180 }, { "epoch": 0.42, "learning_rate": 2.8781124880838893e-06, "loss": 3.762, "step": 2181 }, { "epoch": 0.42, "learning_rate": 2.878055290753098e-06, "loss": 3.7305, "step": 2182 }, { "epoch": 0.42, "learning_rate": 2.8779980934223073e-06, "loss": 3.6533, "step": 2183 }, { "epoch": 0.42, "learning_rate": 2.8779408960915157e-06, "loss": 3.7882, "step": 2184 }, { "epoch": 0.42, "learning_rate": 2.8778836987607245e-06, "loss": 3.556, "step": 2185 }, { "epoch": 0.42, "learning_rate": 2.8778265014299333e-06, "loss": 3.7359, "step": 2186 }, { "epoch": 0.42, "learning_rate": 2.877769304099142e-06, "loss": 3.4671, "step": 2187 }, { "epoch": 0.42, "learning_rate": 2.877712106768351e-06, "loss": 3.4021, "step": 2188 }, { "epoch": 0.42, "learning_rate": 2.8776549094375597e-06, "loss": 3.7737, "step": 2189 }, { "epoch": 0.42, "learning_rate": 2.8775977121067684e-06, "loss": 3.5733, "step": 2190 }, { "epoch": 0.42, "learning_rate": 2.8775405147759772e-06, "loss": 3.7167, "step": 2191 }, { "epoch": 0.42, "learning_rate": 2.877483317445186e-06, "loss": 3.6069, "step": 2192 }, { "epoch": 0.42, "learning_rate": 2.877426120114395e-06, "loss": 3.6946, "step": 2193 }, { "epoch": 0.42, "learning_rate": 2.8773689227836036e-06, "loss": 3.5828, "step": 2194 }, { "epoch": 0.42, "learning_rate": 2.8773117254528124e-06, "loss": 3.6048, "step": 2195 }, { "epoch": 0.42, "learning_rate": 2.8772545281220208e-06, "loss": 3.5847, "step": 2196 }, { "epoch": 0.42, "learning_rate": 2.87719733079123e-06, "loss": 3.6042, "step": 2197 }, { "epoch": 0.42, "learning_rate": 2.8771401334604384e-06, "loss": 3.5981, "step": 2198 }, { "epoch": 0.42, "learning_rate": 2.877082936129647e-06, "loss": 3.6728, "step": 2199 }, { "epoch": 0.42, "learning_rate": 2.8770257387988564e-06, "loss": 3.55, "step": 2200 }, { "epoch": 0.42, "learning_rate": 2.8769685414680648e-06, "loss": 3.6648, "step": 2201 }, { "epoch": 0.42, "learning_rate": 2.8769113441372735e-06, "loss": 3.5161, "step": 2202 }, { "epoch": 0.42, "learning_rate": 2.8768541468064823e-06, "loss": 3.6488, "step": 2203 }, { "epoch": 0.42, "learning_rate": 2.876796949475691e-06, "loss": 3.7857, "step": 2204 }, { "epoch": 0.42, "learning_rate": 2.8767397521449e-06, "loss": 3.7435, "step": 2205 }, { "epoch": 0.42, "learning_rate": 2.8766825548141087e-06, "loss": 3.4295, "step": 2206 }, { "epoch": 0.42, "learning_rate": 2.8766253574833175e-06, "loss": 3.6693, "step": 2207 }, { "epoch": 0.42, "learning_rate": 2.8765681601525263e-06, "loss": 3.6826, "step": 2208 }, { "epoch": 0.42, "learning_rate": 2.876510962821735e-06, "loss": 3.7975, "step": 2209 }, { "epoch": 0.42, "learning_rate": 2.876453765490944e-06, "loss": 3.5033, "step": 2210 }, { "epoch": 0.42, "learning_rate": 2.8763965681601527e-06, "loss": 3.6052, "step": 2211 }, { "epoch": 0.42, "learning_rate": 2.8763393708293615e-06, "loss": 3.573, "step": 2212 }, { "epoch": 0.42, "learning_rate": 2.87628217349857e-06, "loss": 3.4397, "step": 2213 }, { "epoch": 0.42, "learning_rate": 2.876224976167779e-06, "loss": 3.845, "step": 2214 }, { "epoch": 0.42, "learning_rate": 2.876167778836988e-06, "loss": 3.7974, "step": 2215 }, { "epoch": 0.42, "learning_rate": 2.8761105815061962e-06, "loss": 3.4479, "step": 2216 }, { "epoch": 0.42, "learning_rate": 2.8760533841754055e-06, "loss": 3.5476, "step": 2217 }, { "epoch": 0.42, "learning_rate": 2.875996186844614e-06, "loss": 3.6692, "step": 2218 }, { "epoch": 0.42, "learning_rate": 2.8759389895138226e-06, "loss": 3.8091, "step": 2219 }, { "epoch": 0.42, "learning_rate": 2.8758817921830314e-06, "loss": 3.8294, "step": 2220 }, { "epoch": 0.42, "learning_rate": 2.87582459485224e-06, "loss": 3.4463, "step": 2221 }, { "epoch": 0.42, "learning_rate": 2.8757673975214494e-06, "loss": 3.2312, "step": 2222 }, { "epoch": 0.42, "learning_rate": 2.875710200190658e-06, "loss": 3.7774, "step": 2223 }, { "epoch": 0.42, "learning_rate": 2.8756530028598666e-06, "loss": 3.7565, "step": 2224 }, { "epoch": 0.42, "learning_rate": 2.8755958055290754e-06, "loss": 3.7594, "step": 2225 }, { "epoch": 0.42, "learning_rate": 2.875538608198284e-06, "loss": 3.6622, "step": 2226 }, { "epoch": 0.42, "learning_rate": 2.875481410867493e-06, "loss": 3.6846, "step": 2227 }, { "epoch": 0.42, "learning_rate": 2.8754242135367018e-06, "loss": 3.7453, "step": 2228 }, { "epoch": 0.42, "learning_rate": 2.8753670162059106e-06, "loss": 3.591, "step": 2229 }, { "epoch": 0.42, "learning_rate": 2.875309818875119e-06, "loss": 3.3149, "step": 2230 }, { "epoch": 0.42, "learning_rate": 2.875252621544328e-06, "loss": 3.4518, "step": 2231 }, { "epoch": 0.43, "learning_rate": 2.875195424213537e-06, "loss": 3.6754, "step": 2232 }, { "epoch": 0.43, "learning_rate": 2.8751382268827453e-06, "loss": 3.7874, "step": 2233 }, { "epoch": 0.43, "learning_rate": 2.8750810295519545e-06, "loss": 3.7145, "step": 2234 }, { "epoch": 0.43, "learning_rate": 2.875023832221163e-06, "loss": 3.6613, "step": 2235 }, { "epoch": 0.43, "learning_rate": 2.8749666348903717e-06, "loss": 3.5942, "step": 2236 }, { "epoch": 0.43, "learning_rate": 2.8749094375595805e-06, "loss": 3.6415, "step": 2237 }, { "epoch": 0.43, "learning_rate": 2.8748522402287893e-06, "loss": 3.8677, "step": 2238 }, { "epoch": 0.43, "learning_rate": 2.8747950428979985e-06, "loss": 3.6008, "step": 2239 }, { "epoch": 0.43, "learning_rate": 2.874737845567207e-06, "loss": 3.7428, "step": 2240 }, { "epoch": 0.43, "learning_rate": 2.8746806482364157e-06, "loss": 3.6361, "step": 2241 }, { "epoch": 0.43, "learning_rate": 2.8746234509056245e-06, "loss": 3.8185, "step": 2242 }, { "epoch": 0.43, "learning_rate": 2.8745662535748332e-06, "loss": 3.607, "step": 2243 }, { "epoch": 0.43, "learning_rate": 2.874509056244042e-06, "loss": 3.6174, "step": 2244 }, { "epoch": 0.43, "learning_rate": 2.874451858913251e-06, "loss": 3.6402, "step": 2245 }, { "epoch": 0.43, "learning_rate": 2.8743946615824596e-06, "loss": 3.6535, "step": 2246 }, { "epoch": 0.43, "learning_rate": 2.874337464251668e-06, "loss": 3.7537, "step": 2247 }, { "epoch": 0.43, "learning_rate": 2.8742802669208772e-06, "loss": 3.7826, "step": 2248 }, { "epoch": 0.43, "learning_rate": 2.874223069590086e-06, "loss": 3.87, "step": 2249 }, { "epoch": 0.43, "learning_rate": 2.8741658722592944e-06, "loss": 3.5346, "step": 2250 }, { "epoch": 0.43, "learning_rate": 2.8741086749285036e-06, "loss": 3.4874, "step": 2251 }, { "epoch": 0.43, "learning_rate": 2.874051477597712e-06, "loss": 3.3564, "step": 2252 }, { "epoch": 0.43, "learning_rate": 2.8739942802669208e-06, "loss": 3.8498, "step": 2253 }, { "epoch": 0.43, "learning_rate": 2.87393708293613e-06, "loss": 3.5865, "step": 2254 }, { "epoch": 0.43, "learning_rate": 2.8738798856053384e-06, "loss": 3.6447, "step": 2255 }, { "epoch": 0.43, "learning_rate": 2.8738226882745476e-06, "loss": 3.7683, "step": 2256 }, { "epoch": 0.43, "learning_rate": 2.873765490943756e-06, "loss": 3.6407, "step": 2257 }, { "epoch": 0.43, "learning_rate": 2.8737082936129647e-06, "loss": 3.5877, "step": 2258 }, { "epoch": 0.43, "learning_rate": 2.8736510962821735e-06, "loss": 3.5868, "step": 2259 }, { "epoch": 0.43, "learning_rate": 2.8735938989513823e-06, "loss": 3.4128, "step": 2260 }, { "epoch": 0.43, "learning_rate": 2.873536701620591e-06, "loss": 3.531, "step": 2261 }, { "epoch": 0.43, "learning_rate": 2.8734795042898e-06, "loss": 3.7454, "step": 2262 }, { "epoch": 0.43, "learning_rate": 2.8734223069590087e-06, "loss": 3.6351, "step": 2263 }, { "epoch": 0.43, "learning_rate": 2.8733651096282175e-06, "loss": 3.5476, "step": 2264 }, { "epoch": 0.43, "learning_rate": 2.8733079122974263e-06, "loss": 3.3842, "step": 2265 }, { "epoch": 0.43, "learning_rate": 2.873250714966635e-06, "loss": 3.4836, "step": 2266 }, { "epoch": 0.43, "learning_rate": 2.8731935176358435e-06, "loss": 3.5365, "step": 2267 }, { "epoch": 0.43, "learning_rate": 2.8731363203050527e-06, "loss": 3.752, "step": 2268 }, { "epoch": 0.43, "learning_rate": 2.873079122974261e-06, "loss": 3.7848, "step": 2269 }, { "epoch": 0.43, "learning_rate": 2.87302192564347e-06, "loss": 3.6407, "step": 2270 }, { "epoch": 0.43, "learning_rate": 2.872964728312679e-06, "loss": 3.8988, "step": 2271 }, { "epoch": 0.43, "learning_rate": 2.8729075309818874e-06, "loss": 3.5343, "step": 2272 }, { "epoch": 0.43, "learning_rate": 2.8728503336510966e-06, "loss": 3.5246, "step": 2273 }, { "epoch": 0.43, "learning_rate": 2.872793136320305e-06, "loss": 3.8259, "step": 2274 }, { "epoch": 0.43, "learning_rate": 2.872735938989514e-06, "loss": 3.5404, "step": 2275 }, { "epoch": 0.43, "learning_rate": 2.8726787416587226e-06, "loss": 3.2128, "step": 2276 }, { "epoch": 0.43, "learning_rate": 2.8726215443279314e-06, "loss": 3.3831, "step": 2277 }, { "epoch": 0.43, "learning_rate": 2.87256434699714e-06, "loss": 3.6477, "step": 2278 }, { "epoch": 0.43, "learning_rate": 2.872507149666349e-06, "loss": 3.4843, "step": 2279 }, { "epoch": 0.43, "learning_rate": 2.8724499523355578e-06, "loss": 3.6616, "step": 2280 }, { "epoch": 0.43, "learning_rate": 2.8723927550047666e-06, "loss": 3.7367, "step": 2281 }, { "epoch": 0.43, "learning_rate": 2.8723355576739754e-06, "loss": 3.708, "step": 2282 }, { "epoch": 0.43, "learning_rate": 2.872278360343184e-06, "loss": 3.7684, "step": 2283 }, { "epoch": 0.44, "learning_rate": 2.8722211630123925e-06, "loss": 3.6559, "step": 2284 }, { "epoch": 0.44, "learning_rate": 2.8721639656816017e-06, "loss": 3.8259, "step": 2285 }, { "epoch": 0.44, "learning_rate": 2.87210676835081e-06, "loss": 3.665, "step": 2286 }, { "epoch": 0.44, "learning_rate": 2.8720495710200193e-06, "loss": 3.6101, "step": 2287 }, { "epoch": 0.44, "learning_rate": 2.871992373689228e-06, "loss": 3.7437, "step": 2288 }, { "epoch": 0.44, "learning_rate": 2.8719351763584365e-06, "loss": 3.6063, "step": 2289 }, { "epoch": 0.44, "learning_rate": 2.8718779790276457e-06, "loss": 3.8339, "step": 2290 }, { "epoch": 0.44, "learning_rate": 2.871820781696854e-06, "loss": 3.5774, "step": 2291 }, { "epoch": 0.44, "learning_rate": 2.871763584366063e-06, "loss": 3.9609, "step": 2292 }, { "epoch": 0.44, "learning_rate": 2.871706387035272e-06, "loss": 3.581, "step": 2293 }, { "epoch": 0.44, "learning_rate": 2.8716491897044805e-06, "loss": 3.6018, "step": 2294 }, { "epoch": 0.44, "learning_rate": 2.8715919923736893e-06, "loss": 3.718, "step": 2295 }, { "epoch": 0.44, "learning_rate": 2.871534795042898e-06, "loss": 3.3312, "step": 2296 }, { "epoch": 0.44, "learning_rate": 2.871477597712107e-06, "loss": 3.474, "step": 2297 }, { "epoch": 0.44, "learning_rate": 2.8714204003813156e-06, "loss": 3.6219, "step": 2298 }, { "epoch": 0.44, "learning_rate": 2.8713632030505244e-06, "loss": 3.5269, "step": 2299 }, { "epoch": 0.44, "learning_rate": 2.8713060057197332e-06, "loss": 3.6969, "step": 2300 }, { "epoch": 0.44, "learning_rate": 2.8712488083889416e-06, "loss": 3.52, "step": 2301 }, { "epoch": 0.44, "learning_rate": 2.871191611058151e-06, "loss": 3.4255, "step": 2302 }, { "epoch": 0.44, "learning_rate": 2.8711344137273596e-06, "loss": 3.6832, "step": 2303 }, { "epoch": 0.44, "learning_rate": 2.8710772163965684e-06, "loss": 3.2545, "step": 2304 }, { "epoch": 0.44, "learning_rate": 2.871020019065777e-06, "loss": 3.7807, "step": 2305 }, { "epoch": 0.44, "learning_rate": 2.8709628217349856e-06, "loss": 3.6503, "step": 2306 }, { "epoch": 0.44, "learning_rate": 2.8709056244041948e-06, "loss": 3.5325, "step": 2307 }, { "epoch": 0.44, "learning_rate": 2.870848427073403e-06, "loss": 3.4926, "step": 2308 }, { "epoch": 0.44, "learning_rate": 2.870791229742612e-06, "loss": 3.6988, "step": 2309 }, { "epoch": 0.44, "learning_rate": 2.870734032411821e-06, "loss": 3.7468, "step": 2310 }, { "epoch": 0.44, "learning_rate": 2.8706768350810295e-06, "loss": 3.5923, "step": 2311 }, { "epoch": 0.44, "learning_rate": 2.8706196377502383e-06, "loss": 3.7203, "step": 2312 }, { "epoch": 0.44, "learning_rate": 2.870562440419447e-06, "loss": 3.1616, "step": 2313 }, { "epoch": 0.44, "learning_rate": 2.870505243088656e-06, "loss": 3.8622, "step": 2314 }, { "epoch": 0.44, "learning_rate": 2.8704480457578647e-06, "loss": 3.4778, "step": 2315 }, { "epoch": 0.44, "learning_rate": 2.8703908484270735e-06, "loss": 3.7629, "step": 2316 }, { "epoch": 0.44, "learning_rate": 2.8703336510962823e-06, "loss": 3.7596, "step": 2317 }, { "epoch": 0.44, "learning_rate": 2.8702764537654907e-06, "loss": 3.5249, "step": 2318 }, { "epoch": 0.44, "learning_rate": 2.8702192564347e-06, "loss": 3.5326, "step": 2319 }, { "epoch": 0.44, "learning_rate": 2.8701620591039087e-06, "loss": 3.4989, "step": 2320 }, { "epoch": 0.44, "learning_rate": 2.8701048617731175e-06, "loss": 3.6707, "step": 2321 }, { "epoch": 0.44, "learning_rate": 2.8700476644423263e-06, "loss": 3.854, "step": 2322 }, { "epoch": 0.44, "learning_rate": 2.8699904671115346e-06, "loss": 3.597, "step": 2323 }, { "epoch": 0.44, "learning_rate": 2.869933269780744e-06, "loss": 3.7374, "step": 2324 }, { "epoch": 0.44, "learning_rate": 2.8698760724499522e-06, "loss": 3.488, "step": 2325 }, { "epoch": 0.44, "learning_rate": 2.869818875119161e-06, "loss": 3.3027, "step": 2326 }, { "epoch": 0.44, "learning_rate": 2.8697616777883702e-06, "loss": 3.5812, "step": 2327 }, { "epoch": 0.44, "learning_rate": 2.8697044804575786e-06, "loss": 3.7154, "step": 2328 }, { "epoch": 0.44, "learning_rate": 2.8696472831267874e-06, "loss": 3.6374, "step": 2329 }, { "epoch": 0.44, "learning_rate": 2.869590085795996e-06, "loss": 3.6072, "step": 2330 }, { "epoch": 0.44, "learning_rate": 2.869532888465205e-06, "loss": 3.6104, "step": 2331 }, { "epoch": 0.44, "learning_rate": 2.8694756911344138e-06, "loss": 3.4992, "step": 2332 }, { "epoch": 0.44, "learning_rate": 2.8694184938036226e-06, "loss": 3.4842, "step": 2333 }, { "epoch": 0.44, "learning_rate": 2.8693612964728314e-06, "loss": 3.6982, "step": 2334 }, { "epoch": 0.44, "learning_rate": 2.8693040991420397e-06, "loss": 3.3922, "step": 2335 }, { "epoch": 0.44, "learning_rate": 2.869246901811249e-06, "loss": 3.7011, "step": 2336 }, { "epoch": 0.45, "learning_rate": 2.8691897044804578e-06, "loss": 3.3107, "step": 2337 }, { "epoch": 0.45, "learning_rate": 2.8691325071496665e-06, "loss": 3.5828, "step": 2338 }, { "epoch": 0.45, "learning_rate": 2.8690753098188753e-06, "loss": 3.6288, "step": 2339 }, { "epoch": 0.45, "learning_rate": 2.8690181124880837e-06, "loss": 3.6471, "step": 2340 }, { "epoch": 0.45, "learning_rate": 2.868960915157293e-06, "loss": 3.8695, "step": 2341 }, { "epoch": 0.45, "learning_rate": 2.8689037178265017e-06, "loss": 3.6891, "step": 2342 }, { "epoch": 0.45, "learning_rate": 2.86884652049571e-06, "loss": 3.6723, "step": 2343 }, { "epoch": 0.45, "learning_rate": 2.8687893231649193e-06, "loss": 3.506, "step": 2344 }, { "epoch": 0.45, "learning_rate": 2.8687321258341277e-06, "loss": 3.7907, "step": 2345 }, { "epoch": 0.45, "learning_rate": 2.8686749285033365e-06, "loss": 3.6595, "step": 2346 }, { "epoch": 0.45, "learning_rate": 2.8686177311725453e-06, "loss": 3.4527, "step": 2347 }, { "epoch": 0.45, "learning_rate": 2.868560533841754e-06, "loss": 3.6856, "step": 2348 }, { "epoch": 0.45, "learning_rate": 2.868503336510963e-06, "loss": 3.5024, "step": 2349 }, { "epoch": 0.45, "learning_rate": 2.8684461391801716e-06, "loss": 3.7184, "step": 2350 }, { "epoch": 0.45, "learning_rate": 2.8683889418493804e-06, "loss": 3.6564, "step": 2351 }, { "epoch": 0.45, "learning_rate": 2.8683317445185892e-06, "loss": 3.5545, "step": 2352 }, { "epoch": 0.45, "learning_rate": 2.868274547187798e-06, "loss": 3.6991, "step": 2353 }, { "epoch": 0.45, "learning_rate": 2.868217349857007e-06, "loss": 3.7228, "step": 2354 }, { "epoch": 0.45, "learning_rate": 2.8681601525262156e-06, "loss": 3.5077, "step": 2355 }, { "epoch": 0.45, "learning_rate": 2.8681029551954244e-06, "loss": 3.5128, "step": 2356 }, { "epoch": 0.45, "learning_rate": 2.8680457578646328e-06, "loss": 3.6538, "step": 2357 }, { "epoch": 0.45, "learning_rate": 2.867988560533842e-06, "loss": 3.7346, "step": 2358 }, { "epoch": 0.45, "learning_rate": 2.867931363203051e-06, "loss": 3.6439, "step": 2359 }, { "epoch": 0.45, "learning_rate": 2.867874165872259e-06, "loss": 3.716, "step": 2360 }, { "epoch": 0.45, "learning_rate": 2.8678169685414684e-06, "loss": 3.7491, "step": 2361 }, { "epoch": 0.45, "learning_rate": 2.8677597712106767e-06, "loss": 3.4642, "step": 2362 }, { "epoch": 0.45, "learning_rate": 2.8677025738798855e-06, "loss": 3.8868, "step": 2363 }, { "epoch": 0.45, "learning_rate": 2.8676453765490943e-06, "loss": 3.5042, "step": 2364 }, { "epoch": 0.45, "learning_rate": 2.867588179218303e-06, "loss": 3.5119, "step": 2365 }, { "epoch": 0.45, "learning_rate": 2.867530981887512e-06, "loss": 3.7538, "step": 2366 }, { "epoch": 0.45, "learning_rate": 2.8674737845567207e-06, "loss": 3.6814, "step": 2367 }, { "epoch": 0.45, "learning_rate": 2.8674165872259295e-06, "loss": 3.5223, "step": 2368 }, { "epoch": 0.45, "learning_rate": 2.8673593898951383e-06, "loss": 3.587, "step": 2369 }, { "epoch": 0.45, "learning_rate": 2.867302192564347e-06, "loss": 3.8768, "step": 2370 }, { "epoch": 0.45, "learning_rate": 2.867244995233556e-06, "loss": 3.4547, "step": 2371 }, { "epoch": 0.45, "learning_rate": 2.8671877979027647e-06, "loss": 3.7789, "step": 2372 }, { "epoch": 0.45, "learning_rate": 2.8671306005719735e-06, "loss": 3.6952, "step": 2373 }, { "epoch": 0.45, "learning_rate": 2.867073403241182e-06, "loss": 3.5339, "step": 2374 }, { "epoch": 0.45, "learning_rate": 2.867016205910391e-06, "loss": 3.6186, "step": 2375 }, { "epoch": 0.45, "learning_rate": 2.8669590085796e-06, "loss": 3.7065, "step": 2376 }, { "epoch": 0.45, "learning_rate": 2.8669018112488082e-06, "loss": 3.8467, "step": 2377 }, { "epoch": 0.45, "learning_rate": 2.8668446139180175e-06, "loss": 3.7318, "step": 2378 }, { "epoch": 0.45, "learning_rate": 2.866787416587226e-06, "loss": 3.6244, "step": 2379 }, { "epoch": 0.45, "learning_rate": 2.8667302192564346e-06, "loss": 3.4856, "step": 2380 }, { "epoch": 0.45, "learning_rate": 2.866673021925644e-06, "loss": 3.6146, "step": 2381 }, { "epoch": 0.45, "learning_rate": 2.866615824594852e-06, "loss": 3.6228, "step": 2382 }, { "epoch": 0.45, "learning_rate": 2.866558627264061e-06, "loss": 3.7846, "step": 2383 }, { "epoch": 0.45, "learning_rate": 2.86650142993327e-06, "loss": 3.8022, "step": 2384 }, { "epoch": 0.45, "learning_rate": 2.8664442326024786e-06, "loss": 3.493, "step": 2385 }, { "epoch": 0.45, "learning_rate": 2.8663870352716874e-06, "loss": 3.5311, "step": 2386 }, { "epoch": 0.45, "learning_rate": 2.866329837940896e-06, "loss": 3.5756, "step": 2387 }, { "epoch": 0.45, "learning_rate": 2.866272640610105e-06, "loss": 3.5804, "step": 2388 }, { "epoch": 0.46, "learning_rate": 2.8662154432793138e-06, "loss": 3.5837, "step": 2389 }, { "epoch": 0.46, "learning_rate": 2.8661582459485226e-06, "loss": 3.6396, "step": 2390 }, { "epoch": 0.46, "learning_rate": 2.8661010486177313e-06, "loss": 3.5101, "step": 2391 }, { "epoch": 0.46, "learning_rate": 2.86604385128694e-06, "loss": 3.8042, "step": 2392 }, { "epoch": 0.46, "learning_rate": 2.865986653956149e-06, "loss": 3.4294, "step": 2393 }, { "epoch": 0.46, "learning_rate": 2.8659294566253573e-06, "loss": 3.7639, "step": 2394 }, { "epoch": 0.46, "learning_rate": 2.8658722592945665e-06, "loss": 3.7165, "step": 2395 }, { "epoch": 0.46, "learning_rate": 2.865815061963775e-06, "loss": 3.5068, "step": 2396 }, { "epoch": 0.46, "learning_rate": 2.8657578646329837e-06, "loss": 3.7052, "step": 2397 }, { "epoch": 0.46, "learning_rate": 2.865700667302193e-06, "loss": 3.4406, "step": 2398 }, { "epoch": 0.46, "learning_rate": 2.8656434699714013e-06, "loss": 3.6406, "step": 2399 }, { "epoch": 0.46, "learning_rate": 2.86558627264061e-06, "loss": 3.4815, "step": 2400 }, { "epoch": 0.46, "learning_rate": 2.865529075309819e-06, "loss": 3.491, "step": 2401 }, { "epoch": 0.46, "learning_rate": 2.8654718779790277e-06, "loss": 3.5255, "step": 2402 }, { "epoch": 0.46, "learning_rate": 2.8654146806482364e-06, "loss": 3.8146, "step": 2403 }, { "epoch": 0.46, "learning_rate": 2.8653574833174452e-06, "loss": 3.7201, "step": 2404 }, { "epoch": 0.46, "learning_rate": 2.865300285986654e-06, "loss": 3.3016, "step": 2405 }, { "epoch": 0.46, "learning_rate": 2.865243088655863e-06, "loss": 3.4741, "step": 2406 }, { "epoch": 0.46, "learning_rate": 2.8651858913250716e-06, "loss": 3.3996, "step": 2407 }, { "epoch": 0.46, "learning_rate": 2.8651286939942804e-06, "loss": 3.7531, "step": 2408 }, { "epoch": 0.46, "learning_rate": 2.865071496663489e-06, "loss": 3.632, "step": 2409 }, { "epoch": 0.46, "learning_rate": 2.865014299332698e-06, "loss": 3.7019, "step": 2410 }, { "epoch": 0.46, "learning_rate": 2.8649571020019064e-06, "loss": 3.8363, "step": 2411 }, { "epoch": 0.46, "learning_rate": 2.8648999046711156e-06, "loss": 3.7411, "step": 2412 }, { "epoch": 0.46, "learning_rate": 2.864842707340324e-06, "loss": 3.7305, "step": 2413 }, { "epoch": 0.46, "learning_rate": 2.8647855100095328e-06, "loss": 3.4563, "step": 2414 }, { "epoch": 0.46, "learning_rate": 2.864728312678742e-06, "loss": 3.7906, "step": 2415 }, { "epoch": 0.46, "learning_rate": 2.8646711153479503e-06, "loss": 3.839, "step": 2416 }, { "epoch": 0.46, "learning_rate": 2.864613918017159e-06, "loss": 3.497, "step": 2417 }, { "epoch": 0.46, "learning_rate": 2.864556720686368e-06, "loss": 3.7068, "step": 2418 }, { "epoch": 0.46, "learning_rate": 2.8644995233555767e-06, "loss": 3.6569, "step": 2419 }, { "epoch": 0.46, "learning_rate": 2.8644423260247855e-06, "loss": 3.612, "step": 2420 }, { "epoch": 0.46, "learning_rate": 2.8643851286939943e-06, "loss": 3.7278, "step": 2421 }, { "epoch": 0.46, "learning_rate": 2.864327931363203e-06, "loss": 3.6442, "step": 2422 }, { "epoch": 0.46, "learning_rate": 2.864270734032412e-06, "loss": 3.6131, "step": 2423 }, { "epoch": 0.46, "learning_rate": 2.8642135367016207e-06, "loss": 3.7057, "step": 2424 }, { "epoch": 0.46, "learning_rate": 2.8641563393708295e-06, "loss": 3.4781, "step": 2425 }, { "epoch": 0.46, "learning_rate": 2.8640991420400383e-06, "loss": 3.3039, "step": 2426 }, { "epoch": 0.46, "learning_rate": 2.864041944709247e-06, "loss": 3.433, "step": 2427 }, { "epoch": 0.46, "learning_rate": 2.8639847473784554e-06, "loss": 3.3447, "step": 2428 }, { "epoch": 0.46, "learning_rate": 2.8639275500476647e-06, "loss": 3.4573, "step": 2429 }, { "epoch": 0.46, "learning_rate": 2.8638703527168735e-06, "loss": 3.3145, "step": 2430 }, { "epoch": 0.46, "learning_rate": 2.863813155386082e-06, "loss": 3.5298, "step": 2431 }, { "epoch": 0.46, "learning_rate": 2.863755958055291e-06, "loss": 3.6108, "step": 2432 }, { "epoch": 0.46, "learning_rate": 2.8636987607244994e-06, "loss": 3.8749, "step": 2433 }, { "epoch": 0.46, "learning_rate": 2.8636415633937086e-06, "loss": 3.7341, "step": 2434 }, { "epoch": 0.46, "learning_rate": 2.863584366062917e-06, "loss": 3.631, "step": 2435 }, { "epoch": 0.46, "learning_rate": 2.863527168732126e-06, "loss": 3.7323, "step": 2436 }, { "epoch": 0.46, "learning_rate": 2.863469971401335e-06, "loss": 3.6197, "step": 2437 }, { "epoch": 0.46, "learning_rate": 2.8634127740705434e-06, "loss": 3.4425, "step": 2438 }, { "epoch": 0.46, "learning_rate": 2.863355576739752e-06, "loss": 3.8402, "step": 2439 }, { "epoch": 0.46, "learning_rate": 2.863298379408961e-06, "loss": 3.6801, "step": 2440 }, { "epoch": 0.46, "learning_rate": 2.8632411820781698e-06, "loss": 3.4542, "step": 2441 }, { "epoch": 0.47, "learning_rate": 2.8631839847473786e-06, "loss": 3.9109, "step": 2442 }, { "epoch": 0.47, "learning_rate": 2.8631267874165874e-06, "loss": 3.5974, "step": 2443 }, { "epoch": 0.47, "learning_rate": 2.863069590085796e-06, "loss": 3.432, "step": 2444 }, { "epoch": 0.47, "learning_rate": 2.8630123927550045e-06, "loss": 3.2619, "step": 2445 }, { "epoch": 0.47, "learning_rate": 2.8629551954242137e-06, "loss": 3.611, "step": 2446 }, { "epoch": 0.47, "learning_rate": 2.8628979980934225e-06, "loss": 3.4268, "step": 2447 }, { "epoch": 0.47, "learning_rate": 2.862840800762631e-06, "loss": 3.4197, "step": 2448 }, { "epoch": 0.47, "learning_rate": 2.86278360343184e-06, "loss": 3.7213, "step": 2449 }, { "epoch": 0.47, "learning_rate": 2.8627264061010485e-06, "loss": 3.8904, "step": 2450 }, { "epoch": 0.47, "learning_rate": 2.8626692087702577e-06, "loss": 3.4991, "step": 2451 }, { "epoch": 0.47, "learning_rate": 2.862612011439466e-06, "loss": 3.3972, "step": 2452 }, { "epoch": 0.47, "learning_rate": 2.862554814108675e-06, "loss": 3.5382, "step": 2453 }, { "epoch": 0.47, "learning_rate": 2.862497616777884e-06, "loss": 3.8599, "step": 2454 }, { "epoch": 0.47, "learning_rate": 2.8624404194470925e-06, "loss": 3.4166, "step": 2455 }, { "epoch": 0.47, "learning_rate": 2.8623832221163013e-06, "loss": 3.4865, "step": 2456 }, { "epoch": 0.47, "learning_rate": 2.86232602478551e-06, "loss": 3.7633, "step": 2457 }, { "epoch": 0.47, "learning_rate": 2.862268827454719e-06, "loss": 3.6172, "step": 2458 }, { "epoch": 0.47, "learning_rate": 2.8622116301239276e-06, "loss": 3.6702, "step": 2459 }, { "epoch": 0.47, "learning_rate": 2.8621544327931364e-06, "loss": 3.5904, "step": 2460 }, { "epoch": 0.47, "learning_rate": 2.8620972354623452e-06, "loss": 3.8477, "step": 2461 }, { "epoch": 0.47, "learning_rate": 2.8620400381315536e-06, "loss": 3.304, "step": 2462 }, { "epoch": 0.47, "learning_rate": 2.861982840800763e-06, "loss": 3.619, "step": 2463 }, { "epoch": 0.47, "learning_rate": 2.8619256434699716e-06, "loss": 3.5887, "step": 2464 }, { "epoch": 0.47, "learning_rate": 2.86186844613918e-06, "loss": 3.8444, "step": 2465 }, { "epoch": 0.47, "learning_rate": 2.861811248808389e-06, "loss": 3.4135, "step": 2466 }, { "epoch": 0.47, "learning_rate": 2.8617540514775976e-06, "loss": 3.7144, "step": 2467 }, { "epoch": 0.47, "learning_rate": 2.8616968541468068e-06, "loss": 3.4523, "step": 2468 }, { "epoch": 0.47, "learning_rate": 2.8616396568160156e-06, "loss": 3.5247, "step": 2469 }, { "epoch": 0.47, "learning_rate": 2.861582459485224e-06, "loss": 3.6393, "step": 2470 }, { "epoch": 0.47, "learning_rate": 2.861525262154433e-06, "loss": 3.4195, "step": 2471 }, { "epoch": 0.47, "learning_rate": 2.8614680648236415e-06, "loss": 3.6709, "step": 2472 }, { "epoch": 0.47, "learning_rate": 2.8614108674928503e-06, "loss": 3.4222, "step": 2473 }, { "epoch": 0.47, "learning_rate": 2.861353670162059e-06, "loss": 3.7503, "step": 2474 }, { "epoch": 0.47, "learning_rate": 2.861296472831268e-06, "loss": 3.6224, "step": 2475 }, { "epoch": 0.47, "learning_rate": 2.8612392755004767e-06, "loss": 3.5707, "step": 2476 }, { "epoch": 0.47, "learning_rate": 2.8611820781696855e-06, "loss": 3.4342, "step": 2477 }, { "epoch": 0.47, "learning_rate": 2.8611248808388943e-06, "loss": 3.4554, "step": 2478 }, { "epoch": 0.47, "learning_rate": 2.861067683508103e-06, "loss": 3.3307, "step": 2479 }, { "epoch": 0.47, "learning_rate": 2.861010486177312e-06, "loss": 3.4897, "step": 2480 }, { "epoch": 0.47, "learning_rate": 2.8609532888465207e-06, "loss": 3.7969, "step": 2481 }, { "epoch": 0.47, "learning_rate": 2.860896091515729e-06, "loss": 3.4177, "step": 2482 }, { "epoch": 0.47, "learning_rate": 2.8608388941849383e-06, "loss": 3.672, "step": 2483 }, { "epoch": 0.47, "learning_rate": 2.8607816968541466e-06, "loss": 3.529, "step": 2484 }, { "epoch": 0.47, "learning_rate": 2.860724499523356e-06, "loss": 3.7093, "step": 2485 }, { "epoch": 0.47, "learning_rate": 2.8606673021925646e-06, "loss": 3.7206, "step": 2486 }, { "epoch": 0.47, "learning_rate": 2.860610104861773e-06, "loss": 3.359, "step": 2487 }, { "epoch": 0.47, "learning_rate": 2.8605529075309822e-06, "loss": 3.3056, "step": 2488 }, { "epoch": 0.47, "learning_rate": 2.8604957102001906e-06, "loss": 3.2484, "step": 2489 }, { "epoch": 0.47, "learning_rate": 2.8604385128693994e-06, "loss": 3.3672, "step": 2490 }, { "epoch": 0.47, "learning_rate": 2.860381315538608e-06, "loss": 3.4744, "step": 2491 }, { "epoch": 0.47, "learning_rate": 2.860324118207817e-06, "loss": 3.7972, "step": 2492 }, { "epoch": 0.47, "learning_rate": 2.8602669208770258e-06, "loss": 3.8072, "step": 2493 }, { "epoch": 0.48, "learning_rate": 2.8602097235462346e-06, "loss": 3.2584, "step": 2494 }, { "epoch": 0.48, "learning_rate": 2.8601525262154434e-06, "loss": 3.7531, "step": 2495 }, { "epoch": 0.48, "learning_rate": 2.860095328884652e-06, "loss": 3.5517, "step": 2496 }, { "epoch": 0.48, "learning_rate": 2.860038131553861e-06, "loss": 3.6663, "step": 2497 }, { "epoch": 0.48, "learning_rate": 2.8599809342230697e-06, "loss": 3.5163, "step": 2498 }, { "epoch": 0.48, "learning_rate": 2.8599237368922785e-06, "loss": 3.4635, "step": 2499 }, { "epoch": 0.48, "learning_rate": 2.8598665395614873e-06, "loss": 3.5319, "step": 2500 }, { "epoch": 0.48, "learning_rate": 2.8598093422306957e-06, "loss": 3.812, "step": 2501 }, { "epoch": 0.48, "learning_rate": 2.859752144899905e-06, "loss": 3.8581, "step": 2502 }, { "epoch": 0.48, "learning_rate": 2.8596949475691137e-06, "loss": 3.7103, "step": 2503 }, { "epoch": 0.48, "learning_rate": 2.859637750238322e-06, "loss": 3.6532, "step": 2504 }, { "epoch": 0.48, "learning_rate": 2.8595805529075313e-06, "loss": 3.8703, "step": 2505 }, { "epoch": 0.48, "learning_rate": 2.8595233555767397e-06, "loss": 3.743, "step": 2506 }, { "epoch": 0.48, "learning_rate": 2.8594661582459485e-06, "loss": 3.5075, "step": 2507 }, { "epoch": 0.48, "learning_rate": 2.8594089609151573e-06, "loss": 3.7256, "step": 2508 }, { "epoch": 0.48, "learning_rate": 2.859351763584366e-06, "loss": 3.7304, "step": 2509 }, { "epoch": 0.48, "learning_rate": 2.859294566253575e-06, "loss": 3.5407, "step": 2510 }, { "epoch": 0.48, "learning_rate": 2.8592373689227836e-06, "loss": 3.5611, "step": 2511 }, { "epoch": 0.48, "learning_rate": 2.8591801715919924e-06, "loss": 3.5375, "step": 2512 }, { "epoch": 0.48, "learning_rate": 2.8591229742612012e-06, "loss": 3.7696, "step": 2513 }, { "epoch": 0.48, "learning_rate": 2.85906577693041e-06, "loss": 3.7507, "step": 2514 }, { "epoch": 0.48, "learning_rate": 2.859008579599619e-06, "loss": 3.6996, "step": 2515 }, { "epoch": 0.48, "learning_rate": 2.8589513822688276e-06, "loss": 3.7334, "step": 2516 }, { "epoch": 0.48, "learning_rate": 2.8588941849380364e-06, "loss": 3.6568, "step": 2517 }, { "epoch": 0.48, "learning_rate": 2.858836987607245e-06, "loss": 3.3617, "step": 2518 }, { "epoch": 0.48, "learning_rate": 2.858779790276454e-06, "loss": 3.4302, "step": 2519 }, { "epoch": 0.48, "learning_rate": 2.8587225929456628e-06, "loss": 3.8173, "step": 2520 }, { "epoch": 0.48, "learning_rate": 2.858665395614871e-06, "loss": 3.5464, "step": 2521 }, { "epoch": 0.48, "learning_rate": 2.8586081982840804e-06, "loss": 3.8891, "step": 2522 }, { "epoch": 0.48, "learning_rate": 2.8585510009532887e-06, "loss": 3.4289, "step": 2523 }, { "epoch": 0.48, "learning_rate": 2.8584938036224975e-06, "loss": 3.4223, "step": 2524 }, { "epoch": 0.48, "learning_rate": 2.8584366062917068e-06, "loss": 3.292, "step": 2525 }, { "epoch": 0.48, "learning_rate": 2.858379408960915e-06, "loss": 3.6016, "step": 2526 }, { "epoch": 0.48, "learning_rate": 2.858322211630124e-06, "loss": 3.9097, "step": 2527 }, { "epoch": 0.48, "learning_rate": 2.8582650142993327e-06, "loss": 3.2267, "step": 2528 }, { "epoch": 0.48, "learning_rate": 2.8582078169685415e-06, "loss": 3.7459, "step": 2529 }, { "epoch": 0.48, "learning_rate": 2.8581506196377503e-06, "loss": 3.6365, "step": 2530 }, { "epoch": 0.48, "learning_rate": 2.858093422306959e-06, "loss": 3.5222, "step": 2531 }, { "epoch": 0.48, "learning_rate": 2.858036224976168e-06, "loss": 3.9396, "step": 2532 }, { "epoch": 0.48, "learning_rate": 2.8579790276453767e-06, "loss": 3.2512, "step": 2533 }, { "epoch": 0.48, "learning_rate": 2.8579218303145855e-06, "loss": 3.5355, "step": 2534 }, { "epoch": 0.48, "learning_rate": 2.8578646329837943e-06, "loss": 3.7215, "step": 2535 }, { "epoch": 0.48, "learning_rate": 2.857807435653003e-06, "loss": 3.5517, "step": 2536 }, { "epoch": 0.48, "learning_rate": 2.857750238322212e-06, "loss": 3.5821, "step": 2537 }, { "epoch": 0.48, "learning_rate": 2.8576930409914202e-06, "loss": 3.5251, "step": 2538 }, { "epoch": 0.48, "learning_rate": 2.8576358436606294e-06, "loss": 3.4237, "step": 2539 }, { "epoch": 0.48, "learning_rate": 2.857578646329838e-06, "loss": 3.387, "step": 2540 }, { "epoch": 0.48, "learning_rate": 2.8575214489990466e-06, "loss": 3.5593, "step": 2541 }, { "epoch": 0.48, "learning_rate": 2.857464251668256e-06, "loss": 3.3293, "step": 2542 }, { "epoch": 0.48, "learning_rate": 2.857407054337464e-06, "loss": 3.4815, "step": 2543 }, { "epoch": 0.48, "learning_rate": 2.857349857006673e-06, "loss": 3.4008, "step": 2544 }, { "epoch": 0.48, "learning_rate": 2.8572926596758818e-06, "loss": 3.6496, "step": 2545 }, { "epoch": 0.48, "learning_rate": 2.8572354623450906e-06, "loss": 3.8026, "step": 2546 }, { "epoch": 0.49, "learning_rate": 2.8571782650142994e-06, "loss": 3.5326, "step": 2547 }, { "epoch": 0.49, "learning_rate": 2.857121067683508e-06, "loss": 3.3114, "step": 2548 }, { "epoch": 0.49, "learning_rate": 2.857063870352717e-06, "loss": 3.811, "step": 2549 }, { "epoch": 0.49, "learning_rate": 2.8570066730219258e-06, "loss": 3.4468, "step": 2550 }, { "epoch": 0.49, "learning_rate": 2.8569494756911345e-06, "loss": 3.1408, "step": 2551 }, { "epoch": 0.49, "learning_rate": 2.8568922783603433e-06, "loss": 3.6016, "step": 2552 }, { "epoch": 0.49, "learning_rate": 2.856835081029552e-06, "loss": 3.985, "step": 2553 }, { "epoch": 0.49, "learning_rate": 2.856777883698761e-06, "loss": 3.5025, "step": 2554 }, { "epoch": 0.49, "learning_rate": 2.8567206863679693e-06, "loss": 3.6643, "step": 2555 }, { "epoch": 0.49, "learning_rate": 2.8566634890371785e-06, "loss": 3.3595, "step": 2556 }, { "epoch": 0.49, "learning_rate": 2.856606291706387e-06, "loss": 3.3587, "step": 2557 }, { "epoch": 0.49, "learning_rate": 2.8565490943755957e-06, "loss": 3.7746, "step": 2558 }, { "epoch": 0.49, "learning_rate": 2.856491897044805e-06, "loss": 3.8237, "step": 2559 }, { "epoch": 0.49, "learning_rate": 2.8564346997140133e-06, "loss": 3.3104, "step": 2560 }, { "epoch": 0.49, "learning_rate": 2.856377502383222e-06, "loss": 3.2161, "step": 2561 }, { "epoch": 0.49, "learning_rate": 2.856320305052431e-06, "loss": 3.4118, "step": 2562 }, { "epoch": 0.49, "learning_rate": 2.8562631077216396e-06, "loss": 3.2735, "step": 2563 }, { "epoch": 0.49, "learning_rate": 2.856205910390849e-06, "loss": 3.7978, "step": 2564 }, { "epoch": 0.49, "learning_rate": 2.8561487130600572e-06, "loss": 3.5647, "step": 2565 }, { "epoch": 0.49, "learning_rate": 2.856091515729266e-06, "loss": 3.7498, "step": 2566 }, { "epoch": 0.49, "learning_rate": 2.856034318398475e-06, "loss": 3.3381, "step": 2567 }, { "epoch": 0.49, "learning_rate": 2.8559771210676836e-06, "loss": 3.6959, "step": 2568 }, { "epoch": 0.49, "learning_rate": 2.8559199237368924e-06, "loss": 3.6774, "step": 2569 }, { "epoch": 0.49, "learning_rate": 2.855862726406101e-06, "loss": 3.4953, "step": 2570 }, { "epoch": 0.49, "learning_rate": 2.85580552907531e-06, "loss": 3.4812, "step": 2571 }, { "epoch": 0.49, "learning_rate": 2.8557483317445184e-06, "loss": 3.6662, "step": 2572 }, { "epoch": 0.49, "learning_rate": 2.8556911344137276e-06, "loss": 3.4039, "step": 2573 }, { "epoch": 0.49, "learning_rate": 2.8556339370829364e-06, "loss": 3.7146, "step": 2574 }, { "epoch": 0.49, "learning_rate": 2.8555767397521448e-06, "loss": 3.8067, "step": 2575 }, { "epoch": 0.49, "learning_rate": 2.855519542421354e-06, "loss": 3.5664, "step": 2576 }, { "epoch": 0.49, "learning_rate": 2.8554623450905623e-06, "loss": 3.7504, "step": 2577 }, { "epoch": 0.49, "learning_rate": 2.855405147759771e-06, "loss": 3.6076, "step": 2578 }, { "epoch": 0.49, "learning_rate": 2.85534795042898e-06, "loss": 3.7342, "step": 2579 }, { "epoch": 0.49, "learning_rate": 2.8552907530981887e-06, "loss": 3.3483, "step": 2580 }, { "epoch": 0.49, "learning_rate": 2.855233555767398e-06, "loss": 3.8883, "step": 2581 }, { "epoch": 0.49, "learning_rate": 2.8551763584366063e-06, "loss": 3.7771, "step": 2582 }, { "epoch": 0.49, "learning_rate": 2.855119161105815e-06, "loss": 3.7908, "step": 2583 }, { "epoch": 0.49, "learning_rate": 2.855061963775024e-06, "loss": 3.5278, "step": 2584 }, { "epoch": 0.49, "learning_rate": 2.8550047664442327e-06, "loss": 3.5173, "step": 2585 }, { "epoch": 0.49, "learning_rate": 2.8549475691134415e-06, "loss": 3.5037, "step": 2586 }, { "epoch": 0.49, "learning_rate": 2.8548903717826503e-06, "loss": 3.7589, "step": 2587 }, { "epoch": 0.49, "learning_rate": 2.854833174451859e-06, "loss": 3.3933, "step": 2588 }, { "epoch": 0.49, "learning_rate": 2.8547759771210674e-06, "loss": 3.6859, "step": 2589 }, { "epoch": 0.49, "learning_rate": 2.8547187797902767e-06, "loss": 3.5407, "step": 2590 }, { "epoch": 0.49, "learning_rate": 2.8546615824594855e-06, "loss": 4.0373, "step": 2591 }, { "epoch": 0.49, "learning_rate": 2.854604385128694e-06, "loss": 3.5904, "step": 2592 }, { "epoch": 0.49, "learning_rate": 2.854547187797903e-06, "loss": 3.5668, "step": 2593 }, { "epoch": 0.49, "learning_rate": 2.8544899904671114e-06, "loss": 3.534, "step": 2594 }, { "epoch": 0.49, "learning_rate": 2.85443279313632e-06, "loss": 3.5138, "step": 2595 }, { "epoch": 0.49, "learning_rate": 2.854375595805529e-06, "loss": 3.4385, "step": 2596 }, { "epoch": 0.49, "learning_rate": 2.854318398474738e-06, "loss": 3.8642, "step": 2597 }, { "epoch": 0.49, "learning_rate": 2.854261201143947e-06, "loss": 3.584, "step": 2598 }, { "epoch": 0.5, "learning_rate": 2.8542040038131554e-06, "loss": 3.5931, "step": 2599 }, { "epoch": 0.5, "learning_rate": 2.854146806482364e-06, "loss": 3.775, "step": 2600 }, { "epoch": 0.5, "learning_rate": 2.854089609151573e-06, "loss": 3.7712, "step": 2601 }, { "epoch": 0.5, "learning_rate": 2.8540324118207818e-06, "loss": 3.7693, "step": 2602 }, { "epoch": 0.5, "learning_rate": 2.8539752144899906e-06, "loss": 3.6084, "step": 2603 }, { "epoch": 0.5, "learning_rate": 2.8539180171591993e-06, "loss": 3.8676, "step": 2604 }, { "epoch": 0.5, "learning_rate": 2.853860819828408e-06, "loss": 3.963, "step": 2605 }, { "epoch": 0.5, "learning_rate": 2.853803622497617e-06, "loss": 3.5938, "step": 2606 }, { "epoch": 0.5, "learning_rate": 2.8537464251668257e-06, "loss": 3.7974, "step": 2607 }, { "epoch": 0.5, "learning_rate": 2.8536892278360345e-06, "loss": 3.6252, "step": 2608 }, { "epoch": 0.5, "learning_rate": 2.853632030505243e-06, "loss": 3.867, "step": 2609 }, { "epoch": 0.5, "learning_rate": 2.853574833174452e-06, "loss": 3.6502, "step": 2610 }, { "epoch": 0.5, "learning_rate": 2.8535176358436605e-06, "loss": 3.6869, "step": 2611 }, { "epoch": 0.5, "learning_rate": 2.8534604385128693e-06, "loss": 3.7872, "step": 2612 }, { "epoch": 0.5, "learning_rate": 2.8534032411820785e-06, "loss": 3.7194, "step": 2613 }, { "epoch": 0.5, "learning_rate": 2.853346043851287e-06, "loss": 3.83, "step": 2614 }, { "epoch": 0.5, "learning_rate": 2.853288846520496e-06, "loss": 3.3007, "step": 2615 }, { "epoch": 0.5, "learning_rate": 2.8532316491897044e-06, "loss": 3.6001, "step": 2616 }, { "epoch": 0.5, "learning_rate": 2.8531744518589132e-06, "loss": 3.7915, "step": 2617 }, { "epoch": 0.5, "learning_rate": 2.853117254528122e-06, "loss": 3.6908, "step": 2618 }, { "epoch": 0.5, "learning_rate": 2.853060057197331e-06, "loss": 3.7926, "step": 2619 }, { "epoch": 0.5, "learning_rate": 2.8530028598665396e-06, "loss": 3.5979, "step": 2620 }, { "epoch": 0.5, "learning_rate": 2.8529456625357484e-06, "loss": 3.6502, "step": 2621 }, { "epoch": 0.5, "learning_rate": 2.8528884652049572e-06, "loss": 3.2683, "step": 2622 }, { "epoch": 0.5, "learning_rate": 2.852831267874166e-06, "loss": 3.5066, "step": 2623 }, { "epoch": 0.5, "learning_rate": 2.852774070543375e-06, "loss": 3.8866, "step": 2624 }, { "epoch": 0.5, "learning_rate": 2.8527168732125836e-06, "loss": 3.6633, "step": 2625 }, { "epoch": 0.5, "learning_rate": 2.852659675881792e-06, "loss": 3.2519, "step": 2626 }, { "epoch": 0.5, "learning_rate": 2.852602478551001e-06, "loss": 3.7592, "step": 2627 }, { "epoch": 0.5, "learning_rate": 2.8525452812202096e-06, "loss": 3.6655, "step": 2628 }, { "epoch": 0.5, "learning_rate": 2.8524880838894188e-06, "loss": 3.9322, "step": 2629 }, { "epoch": 0.5, "learning_rate": 2.8524308865586276e-06, "loss": 3.748, "step": 2630 }, { "epoch": 0.5, "learning_rate": 2.852373689227836e-06, "loss": 3.582, "step": 2631 }, { "epoch": 0.5, "learning_rate": 2.852316491897045e-06, "loss": 3.4884, "step": 2632 }, { "epoch": 0.5, "learning_rate": 2.8522592945662535e-06, "loss": 3.5305, "step": 2633 }, { "epoch": 0.5, "learning_rate": 2.8522020972354623e-06, "loss": 3.881, "step": 2634 }, { "epoch": 0.5, "learning_rate": 2.852144899904671e-06, "loss": 3.4599, "step": 2635 }, { "epoch": 0.5, "learning_rate": 2.85208770257388e-06, "loss": 3.7208, "step": 2636 }, { "epoch": 0.5, "learning_rate": 2.8520305052430887e-06, "loss": 3.56, "step": 2637 }, { "epoch": 0.5, "learning_rate": 2.8519733079122975e-06, "loss": 3.5158, "step": 2638 }, { "epoch": 0.5, "learning_rate": 2.8519161105815063e-06, "loss": 3.5651, "step": 2639 }, { "epoch": 0.5, "learning_rate": 2.851858913250715e-06, "loss": 3.7722, "step": 2640 }, { "epoch": 0.5, "learning_rate": 2.851801715919924e-06, "loss": 3.6545, "step": 2641 }, { "epoch": 0.5, "learning_rate": 2.8517445185891327e-06, "loss": 3.4668, "step": 2642 }, { "epoch": 0.5, "learning_rate": 2.851687321258341e-06, "loss": 3.7832, "step": 2643 }, { "epoch": 0.5, "learning_rate": 2.8516301239275503e-06, "loss": 3.6732, "step": 2644 }, { "epoch": 0.5, "learning_rate": 2.8515729265967586e-06, "loss": 3.6444, "step": 2645 }, { "epoch": 0.5, "learning_rate": 2.851515729265968e-06, "loss": 3.3736, "step": 2646 }, { "epoch": 0.5, "learning_rate": 2.8514585319351766e-06, "loss": 3.3023, "step": 2647 }, { "epoch": 0.5, "learning_rate": 2.851401334604385e-06, "loss": 3.6518, "step": 2648 }, { "epoch": 0.5, "learning_rate": 2.8513441372735942e-06, "loss": 3.3022, "step": 2649 }, { "epoch": 0.5, "learning_rate": 2.8512869399428026e-06, "loss": 3.7615, "step": 2650 }, { "epoch": 0.5, "learning_rate": 2.8512297426120114e-06, "loss": 3.4942, "step": 2651 }, { "epoch": 0.51, "learning_rate": 2.8511725452812206e-06, "loss": 3.3201, "step": 2652 }, { "epoch": 0.51, "learning_rate": 2.851115347950429e-06, "loss": 3.4666, "step": 2653 }, { "epoch": 0.51, "learning_rate": 2.8510581506196378e-06, "loss": 3.7371, "step": 2654 }, { "epoch": 0.51, "learning_rate": 2.8510009532888466e-06, "loss": 3.8061, "step": 2655 }, { "epoch": 0.51, "learning_rate": 2.8509437559580554e-06, "loss": 3.5048, "step": 2656 }, { "epoch": 0.51, "learning_rate": 2.850886558627264e-06, "loss": 3.4612, "step": 2657 }, { "epoch": 0.51, "learning_rate": 2.850829361296473e-06, "loss": 3.8031, "step": 2658 }, { "epoch": 0.51, "learning_rate": 2.8507721639656817e-06, "loss": 3.5059, "step": 2659 }, { "epoch": 0.51, "learning_rate": 2.85071496663489e-06, "loss": 3.7101, "step": 2660 }, { "epoch": 0.51, "learning_rate": 2.8506577693040993e-06, "loss": 3.6309, "step": 2661 }, { "epoch": 0.51, "learning_rate": 2.850600571973308e-06, "loss": 3.8176, "step": 2662 }, { "epoch": 0.51, "learning_rate": 2.850543374642517e-06, "loss": 3.6013, "step": 2663 }, { "epoch": 0.51, "learning_rate": 2.8504861773117257e-06, "loss": 3.4916, "step": 2664 }, { "epoch": 0.51, "learning_rate": 2.850428979980934e-06, "loss": 3.792, "step": 2665 }, { "epoch": 0.51, "learning_rate": 2.8503717826501433e-06, "loss": 3.8238, "step": 2666 }, { "epoch": 0.51, "learning_rate": 2.8503145853193517e-06, "loss": 3.6253, "step": 2667 }, { "epoch": 0.51, "learning_rate": 2.8502573879885605e-06, "loss": 3.3588, "step": 2668 }, { "epoch": 0.51, "learning_rate": 2.8502001906577697e-06, "loss": 3.5141, "step": 2669 }, { "epoch": 0.51, "learning_rate": 2.850142993326978e-06, "loss": 3.5724, "step": 2670 }, { "epoch": 0.51, "learning_rate": 2.850085795996187e-06, "loss": 3.5649, "step": 2671 }, { "epoch": 0.51, "learning_rate": 2.8500285986653956e-06, "loss": 3.7767, "step": 2672 }, { "epoch": 0.51, "learning_rate": 2.8499714013346044e-06, "loss": 3.5265, "step": 2673 }, { "epoch": 0.51, "learning_rate": 2.8499142040038132e-06, "loss": 3.799, "step": 2674 }, { "epoch": 0.51, "learning_rate": 2.849857006673022e-06, "loss": 3.6187, "step": 2675 }, { "epoch": 0.51, "learning_rate": 2.849799809342231e-06, "loss": 3.6525, "step": 2676 }, { "epoch": 0.51, "learning_rate": 2.849742612011439e-06, "loss": 3.7031, "step": 2677 }, { "epoch": 0.51, "learning_rate": 2.8496854146806484e-06, "loss": 3.7205, "step": 2678 }, { "epoch": 0.51, "learning_rate": 2.849628217349857e-06, "loss": 3.902, "step": 2679 }, { "epoch": 0.51, "learning_rate": 2.849571020019066e-06, "loss": 3.8707, "step": 2680 }, { "epoch": 0.51, "learning_rate": 2.8495138226882748e-06, "loss": 3.5857, "step": 2681 }, { "epoch": 0.51, "learning_rate": 2.849456625357483e-06, "loss": 3.8498, "step": 2682 }, { "epoch": 0.51, "learning_rate": 2.8493994280266924e-06, "loss": 3.6212, "step": 2683 }, { "epoch": 0.51, "learning_rate": 2.8493422306959007e-06, "loss": 3.2423, "step": 2684 }, { "epoch": 0.51, "learning_rate": 2.8492850333651095e-06, "loss": 3.7307, "step": 2685 }, { "epoch": 0.51, "learning_rate": 2.8492278360343187e-06, "loss": 3.863, "step": 2686 }, { "epoch": 0.51, "learning_rate": 2.849170638703527e-06, "loss": 3.6935, "step": 2687 }, { "epoch": 0.51, "learning_rate": 2.849113441372736e-06, "loss": 3.7132, "step": 2688 }, { "epoch": 0.51, "learning_rate": 2.8490562440419447e-06, "loss": 3.6249, "step": 2689 }, { "epoch": 0.51, "learning_rate": 2.8489990467111535e-06, "loss": 3.3422, "step": 2690 }, { "epoch": 0.51, "learning_rate": 2.8489418493803623e-06, "loss": 3.6228, "step": 2691 }, { "epoch": 0.51, "learning_rate": 2.848884652049571e-06, "loss": 3.2327, "step": 2692 }, { "epoch": 0.51, "learning_rate": 2.84882745471878e-06, "loss": 3.2899, "step": 2693 }, { "epoch": 0.51, "learning_rate": 2.8487702573879887e-06, "loss": 3.6568, "step": 2694 }, { "epoch": 0.51, "learning_rate": 2.8487130600571975e-06, "loss": 3.69, "step": 2695 }, { "epoch": 0.51, "learning_rate": 2.8486558627264063e-06, "loss": 3.8075, "step": 2696 }, { "epoch": 0.51, "learning_rate": 2.848598665395615e-06, "loss": 3.7091, "step": 2697 }, { "epoch": 0.51, "learning_rate": 2.848541468064824e-06, "loss": 3.4992, "step": 2698 }, { "epoch": 0.51, "learning_rate": 2.8484842707340322e-06, "loss": 3.5869, "step": 2699 }, { "epoch": 0.51, "learning_rate": 2.8484270734032414e-06, "loss": 3.8632, "step": 2700 }, { "epoch": 0.51, "learning_rate": 2.8483698760724502e-06, "loss": 3.526, "step": 2701 }, { "epoch": 0.51, "learning_rate": 2.8483126787416586e-06, "loss": 3.4131, "step": 2702 }, { "epoch": 0.51, "learning_rate": 2.848255481410868e-06, "loss": 3.7593, "step": 2703 }, { "epoch": 0.52, "learning_rate": 2.848198284080076e-06, "loss": 3.3208, "step": 2704 }, { "epoch": 0.52, "learning_rate": 2.848141086749285e-06, "loss": 3.5831, "step": 2705 }, { "epoch": 0.52, "learning_rate": 2.8480838894184938e-06, "loss": 3.701, "step": 2706 }, { "epoch": 0.52, "learning_rate": 2.8480266920877026e-06, "loss": 3.4853, "step": 2707 }, { "epoch": 0.52, "learning_rate": 2.8479694947569114e-06, "loss": 3.5504, "step": 2708 }, { "epoch": 0.52, "learning_rate": 2.84791229742612e-06, "loss": 3.8565, "step": 2709 }, { "epoch": 0.52, "learning_rate": 2.847855100095329e-06, "loss": 3.7163, "step": 2710 }, { "epoch": 0.52, "learning_rate": 2.8477979027645377e-06, "loss": 3.4553, "step": 2711 }, { "epoch": 0.52, "learning_rate": 2.8477407054337465e-06, "loss": 3.2885, "step": 2712 }, { "epoch": 0.52, "learning_rate": 2.8476835081029553e-06, "loss": 3.8232, "step": 2713 }, { "epoch": 0.52, "learning_rate": 2.847626310772164e-06, "loss": 3.695, "step": 2714 }, { "epoch": 0.52, "learning_rate": 2.847569113441373e-06, "loss": 3.4279, "step": 2715 }, { "epoch": 0.52, "learning_rate": 2.8475119161105813e-06, "loss": 3.1698, "step": 2716 }, { "epoch": 0.52, "learning_rate": 2.8474547187797905e-06, "loss": 3.4334, "step": 2717 }, { "epoch": 0.52, "learning_rate": 2.8473975214489993e-06, "loss": 3.2539, "step": 2718 }, { "epoch": 0.52, "learning_rate": 2.8473403241182077e-06, "loss": 3.5049, "step": 2719 }, { "epoch": 0.52, "learning_rate": 2.847283126787417e-06, "loss": 3.5139, "step": 2720 }, { "epoch": 0.52, "learning_rate": 2.8472259294566253e-06, "loss": 3.4054, "step": 2721 }, { "epoch": 0.52, "learning_rate": 2.847168732125834e-06, "loss": 3.7252, "step": 2722 }, { "epoch": 0.52, "learning_rate": 2.847111534795043e-06, "loss": 3.4214, "step": 2723 }, { "epoch": 0.52, "learning_rate": 2.8470543374642516e-06, "loss": 3.8235, "step": 2724 }, { "epoch": 0.52, "learning_rate": 2.8469971401334604e-06, "loss": 3.599, "step": 2725 }, { "epoch": 0.52, "learning_rate": 2.8469399428026692e-06, "loss": 3.9867, "step": 2726 }, { "epoch": 0.52, "learning_rate": 2.846882745471878e-06, "loss": 3.5689, "step": 2727 }, { "epoch": 0.52, "learning_rate": 2.846825548141087e-06, "loss": 3.883, "step": 2728 }, { "epoch": 0.52, "learning_rate": 2.8467683508102956e-06, "loss": 3.6344, "step": 2729 }, { "epoch": 0.52, "learning_rate": 2.8467111534795044e-06, "loss": 3.8367, "step": 2730 }, { "epoch": 0.52, "learning_rate": 2.846653956148713e-06, "loss": 3.63, "step": 2731 }, { "epoch": 0.52, "learning_rate": 2.846596758817922e-06, "loss": 3.348, "step": 2732 }, { "epoch": 0.52, "learning_rate": 2.8465395614871304e-06, "loss": 3.8496, "step": 2733 }, { "epoch": 0.52, "learning_rate": 2.8464823641563396e-06, "loss": 3.8288, "step": 2734 }, { "epoch": 0.52, "learning_rate": 2.8464251668255484e-06, "loss": 3.6593, "step": 2735 }, { "epoch": 0.52, "learning_rate": 2.8463679694947567e-06, "loss": 3.4016, "step": 2736 }, { "epoch": 0.52, "learning_rate": 2.846310772163966e-06, "loss": 3.4317, "step": 2737 }, { "epoch": 0.52, "learning_rate": 2.8462535748331743e-06, "loss": 3.5553, "step": 2738 }, { "epoch": 0.52, "learning_rate": 2.846196377502383e-06, "loss": 3.8038, "step": 2739 }, { "epoch": 0.52, "learning_rate": 2.8461391801715923e-06, "loss": 3.5759, "step": 2740 }, { "epoch": 0.52, "learning_rate": 2.8460819828408007e-06, "loss": 3.5374, "step": 2741 }, { "epoch": 0.52, "learning_rate": 2.8460247855100095e-06, "loss": 3.6881, "step": 2742 }, { "epoch": 0.52, "learning_rate": 2.8459675881792183e-06, "loss": 3.7235, "step": 2743 }, { "epoch": 0.52, "learning_rate": 2.845910390848427e-06, "loss": 3.441, "step": 2744 }, { "epoch": 0.52, "learning_rate": 2.845853193517636e-06, "loss": 3.6958, "step": 2745 }, { "epoch": 0.52, "learning_rate": 2.8457959961868447e-06, "loss": 3.4551, "step": 2746 }, { "epoch": 0.52, "learning_rate": 2.8457387988560535e-06, "loss": 3.6458, "step": 2747 }, { "epoch": 0.52, "learning_rate": 2.8456816015252623e-06, "loss": 3.8215, "step": 2748 }, { "epoch": 0.52, "learning_rate": 2.845624404194471e-06, "loss": 3.5076, "step": 2749 }, { "epoch": 0.52, "learning_rate": 2.84556720686368e-06, "loss": 3.5864, "step": 2750 }, { "epoch": 0.52, "learning_rate": 2.8455100095328887e-06, "loss": 3.8394, "step": 2751 }, { "epoch": 0.52, "learning_rate": 2.8454528122020974e-06, "loss": 3.5974, "step": 2752 }, { "epoch": 0.52, "learning_rate": 2.845395614871306e-06, "loss": 3.6124, "step": 2753 }, { "epoch": 0.52, "learning_rate": 2.845338417540515e-06, "loss": 3.658, "step": 2754 }, { "epoch": 0.52, "learning_rate": 2.8452812202097234e-06, "loss": 3.7134, "step": 2755 }, { "epoch": 0.52, "learning_rate": 2.845224022878932e-06, "loss": 3.3912, "step": 2756 }, { "epoch": 0.53, "learning_rate": 2.8451668255481414e-06, "loss": 3.7547, "step": 2757 }, { "epoch": 0.53, "learning_rate": 2.8451096282173498e-06, "loss": 3.1705, "step": 2758 }, { "epoch": 0.53, "learning_rate": 2.845052430886559e-06, "loss": 3.7059, "step": 2759 }, { "epoch": 0.53, "learning_rate": 2.8449952335557674e-06, "loss": 3.7365, "step": 2760 }, { "epoch": 0.53, "learning_rate": 2.844938036224976e-06, "loss": 3.4075, "step": 2761 }, { "epoch": 0.53, "learning_rate": 2.844880838894185e-06, "loss": 3.454, "step": 2762 }, { "epoch": 0.53, "learning_rate": 2.8448236415633938e-06, "loss": 3.6505, "step": 2763 }, { "epoch": 0.53, "learning_rate": 2.8447664442326025e-06, "loss": 3.4226, "step": 2764 }, { "epoch": 0.53, "learning_rate": 2.8447092469018113e-06, "loss": 3.7409, "step": 2765 }, { "epoch": 0.53, "learning_rate": 2.84465204957102e-06, "loss": 3.2663, "step": 2766 }, { "epoch": 0.53, "learning_rate": 2.844594852240229e-06, "loss": 3.8236, "step": 2767 }, { "epoch": 0.53, "learning_rate": 2.8445376549094377e-06, "loss": 3.4042, "step": 2768 }, { "epoch": 0.53, "learning_rate": 2.8444804575786465e-06, "loss": 3.5432, "step": 2769 }, { "epoch": 0.53, "learning_rate": 2.844423260247855e-06, "loss": 3.8497, "step": 2770 }, { "epoch": 0.53, "learning_rate": 2.844366062917064e-06, "loss": 3.5439, "step": 2771 }, { "epoch": 0.53, "learning_rate": 2.8443088655862725e-06, "loss": 3.5269, "step": 2772 }, { "epoch": 0.53, "learning_rate": 2.8442516682554813e-06, "loss": 3.5898, "step": 2773 }, { "epoch": 0.53, "learning_rate": 2.8441944709246905e-06, "loss": 3.675, "step": 2774 }, { "epoch": 0.53, "learning_rate": 2.844137273593899e-06, "loss": 3.8919, "step": 2775 }, { "epoch": 0.53, "learning_rate": 2.844080076263108e-06, "loss": 3.5371, "step": 2776 }, { "epoch": 0.53, "learning_rate": 2.8440228789323164e-06, "loss": 3.4604, "step": 2777 }, { "epoch": 0.53, "learning_rate": 2.8439656816015252e-06, "loss": 3.3871, "step": 2778 }, { "epoch": 0.53, "learning_rate": 2.8439084842707345e-06, "loss": 3.6565, "step": 2779 }, { "epoch": 0.53, "learning_rate": 2.843851286939943e-06, "loss": 3.5738, "step": 2780 }, { "epoch": 0.53, "learning_rate": 2.8437940896091516e-06, "loss": 3.7877, "step": 2781 }, { "epoch": 0.53, "learning_rate": 2.8437368922783604e-06, "loss": 3.7742, "step": 2782 }, { "epoch": 0.53, "learning_rate": 2.843679694947569e-06, "loss": 3.4785, "step": 2783 }, { "epoch": 0.53, "learning_rate": 2.843622497616778e-06, "loss": 3.5344, "step": 2784 }, { "epoch": 0.53, "learning_rate": 2.843565300285987e-06, "loss": 3.46, "step": 2785 }, { "epoch": 0.53, "learning_rate": 2.8435081029551956e-06, "loss": 3.4002, "step": 2786 }, { "epoch": 0.53, "learning_rate": 2.843450905624404e-06, "loss": 3.6988, "step": 2787 }, { "epoch": 0.53, "learning_rate": 2.843393708293613e-06, "loss": 3.5057, "step": 2788 }, { "epoch": 0.53, "learning_rate": 2.843336510962822e-06, "loss": 3.5595, "step": 2789 }, { "epoch": 0.53, "learning_rate": 2.8432793136320303e-06, "loss": 3.7108, "step": 2790 }, { "epoch": 0.53, "learning_rate": 2.8432221163012396e-06, "loss": 3.7893, "step": 2791 }, { "epoch": 0.53, "learning_rate": 2.843164918970448e-06, "loss": 3.7232, "step": 2792 }, { "epoch": 0.53, "learning_rate": 2.843107721639657e-06, "loss": 3.3746, "step": 2793 }, { "epoch": 0.53, "learning_rate": 2.8430505243088655e-06, "loss": 3.849, "step": 2794 }, { "epoch": 0.53, "learning_rate": 2.8429933269780743e-06, "loss": 3.6491, "step": 2795 }, { "epoch": 0.53, "learning_rate": 2.8429361296472835e-06, "loss": 3.8646, "step": 2796 }, { "epoch": 0.53, "learning_rate": 2.842878932316492e-06, "loss": 3.5526, "step": 2797 }, { "epoch": 0.53, "learning_rate": 2.8428217349857007e-06, "loss": 3.5385, "step": 2798 }, { "epoch": 0.53, "learning_rate": 2.8427645376549095e-06, "loss": 3.4364, "step": 2799 }, { "epoch": 0.53, "learning_rate": 2.8427073403241183e-06, "loss": 3.5889, "step": 2800 }, { "epoch": 0.53, "learning_rate": 2.842650142993327e-06, "loss": 3.5321, "step": 2801 }, { "epoch": 0.53, "learning_rate": 2.842592945662536e-06, "loss": 3.4461, "step": 2802 }, { "epoch": 0.53, "learning_rate": 2.8425357483317447e-06, "loss": 3.4689, "step": 2803 }, { "epoch": 0.53, "learning_rate": 2.842478551000953e-06, "loss": 3.5071, "step": 2804 }, { "epoch": 0.53, "learning_rate": 2.8424213536701622e-06, "loss": 3.4182, "step": 2805 }, { "epoch": 0.53, "learning_rate": 2.842364156339371e-06, "loss": 3.8527, "step": 2806 }, { "epoch": 0.53, "learning_rate": 2.8423069590085794e-06, "loss": 3.7866, "step": 2807 }, { "epoch": 0.53, "learning_rate": 2.8422497616777886e-06, "loss": 3.5741, "step": 2808 }, { "epoch": 0.54, "learning_rate": 2.842192564346997e-06, "loss": 3.5747, "step": 2809 }, { "epoch": 0.54, "learning_rate": 2.8421353670162062e-06, "loss": 3.3355, "step": 2810 }, { "epoch": 0.54, "learning_rate": 2.8420781696854146e-06, "loss": 3.6997, "step": 2811 }, { "epoch": 0.54, "learning_rate": 2.8420209723546234e-06, "loss": 3.4346, "step": 2812 }, { "epoch": 0.54, "learning_rate": 2.8419637750238326e-06, "loss": 3.8296, "step": 2813 }, { "epoch": 0.54, "learning_rate": 2.841906577693041e-06, "loss": 3.5488, "step": 2814 }, { "epoch": 0.54, "learning_rate": 2.8418493803622498e-06, "loss": 3.5482, "step": 2815 }, { "epoch": 0.54, "learning_rate": 2.8417921830314586e-06, "loss": 3.756, "step": 2816 }, { "epoch": 0.54, "learning_rate": 2.8417349857006673e-06, "loss": 3.8172, "step": 2817 }, { "epoch": 0.54, "learning_rate": 2.841677788369876e-06, "loss": 3.4585, "step": 2818 }, { "epoch": 0.54, "learning_rate": 2.841620591039085e-06, "loss": 3.3139, "step": 2819 }, { "epoch": 0.54, "learning_rate": 2.8415633937082937e-06, "loss": 3.8918, "step": 2820 }, { "epoch": 0.54, "learning_rate": 2.841506196377502e-06, "loss": 3.474, "step": 2821 }, { "epoch": 0.54, "learning_rate": 2.8414489990467113e-06, "loss": 3.3687, "step": 2822 }, { "epoch": 0.54, "learning_rate": 2.84139180171592e-06, "loss": 3.7587, "step": 2823 }, { "epoch": 0.54, "learning_rate": 2.841334604385129e-06, "loss": 3.6883, "step": 2824 }, { "epoch": 0.54, "learning_rate": 2.8412774070543377e-06, "loss": 3.6103, "step": 2825 }, { "epoch": 0.54, "learning_rate": 2.841220209723546e-06, "loss": 3.545, "step": 2826 }, { "epoch": 0.54, "learning_rate": 2.8411630123927553e-06, "loss": 3.9564, "step": 2827 }, { "epoch": 0.54, "learning_rate": 2.841105815061964e-06, "loss": 3.8264, "step": 2828 }, { "epoch": 0.54, "learning_rate": 2.8410486177311725e-06, "loss": 3.4753, "step": 2829 }, { "epoch": 0.54, "learning_rate": 2.8409914204003817e-06, "loss": 3.5659, "step": 2830 }, { "epoch": 0.54, "learning_rate": 2.84093422306959e-06, "loss": 3.6268, "step": 2831 }, { "epoch": 0.54, "learning_rate": 2.840877025738799e-06, "loss": 3.685, "step": 2832 }, { "epoch": 0.54, "learning_rate": 2.8408198284080076e-06, "loss": 3.0628, "step": 2833 }, { "epoch": 0.54, "learning_rate": 2.8407626310772164e-06, "loss": 3.4919, "step": 2834 }, { "epoch": 0.54, "learning_rate": 2.8407054337464252e-06, "loss": 3.4637, "step": 2835 }, { "epoch": 0.54, "learning_rate": 2.840648236415634e-06, "loss": 3.6003, "step": 2836 }, { "epoch": 0.54, "learning_rate": 2.840591039084843e-06, "loss": 3.7439, "step": 2837 }, { "epoch": 0.54, "learning_rate": 2.8405338417540516e-06, "loss": 3.8542, "step": 2838 }, { "epoch": 0.54, "learning_rate": 2.8404766444232604e-06, "loss": 3.5785, "step": 2839 }, { "epoch": 0.54, "learning_rate": 2.840419447092469e-06, "loss": 3.4866, "step": 2840 }, { "epoch": 0.54, "learning_rate": 2.840362249761678e-06, "loss": 3.2939, "step": 2841 }, { "epoch": 0.54, "learning_rate": 2.8403050524308868e-06, "loss": 3.3144, "step": 2842 }, { "epoch": 0.54, "learning_rate": 2.840247855100095e-06, "loss": 3.5509, "step": 2843 }, { "epoch": 0.54, "learning_rate": 2.8401906577693044e-06, "loss": 3.6591, "step": 2844 }, { "epoch": 0.54, "learning_rate": 2.840133460438513e-06, "loss": 3.4796, "step": 2845 }, { "epoch": 0.54, "learning_rate": 2.8400762631077215e-06, "loss": 3.7689, "step": 2846 }, { "epoch": 0.54, "learning_rate": 2.8400190657769307e-06, "loss": 3.6985, "step": 2847 }, { "epoch": 0.54, "learning_rate": 2.839961868446139e-06, "loss": 3.67, "step": 2848 }, { "epoch": 0.54, "learning_rate": 2.839904671115348e-06, "loss": 3.4578, "step": 2849 }, { "epoch": 0.54, "learning_rate": 2.8398474737845567e-06, "loss": 3.4234, "step": 2850 }, { "epoch": 0.54, "learning_rate": 2.8397902764537655e-06, "loss": 3.5468, "step": 2851 }, { "epoch": 0.54, "learning_rate": 2.8397330791229743e-06, "loss": 3.5415, "step": 2852 }, { "epoch": 0.54, "learning_rate": 2.839675881792183e-06, "loss": 3.7315, "step": 2853 }, { "epoch": 0.54, "learning_rate": 2.839618684461392e-06, "loss": 3.5216, "step": 2854 }, { "epoch": 0.54, "learning_rate": 2.8395614871306007e-06, "loss": 3.5415, "step": 2855 }, { "epoch": 0.54, "learning_rate": 2.8395042897998095e-06, "loss": 3.2839, "step": 2856 }, { "epoch": 0.54, "learning_rate": 2.8394470924690183e-06, "loss": 3.8014, "step": 2857 }, { "epoch": 0.54, "learning_rate": 2.839389895138227e-06, "loss": 3.5279, "step": 2858 }, { "epoch": 0.54, "learning_rate": 2.839332697807436e-06, "loss": 3.3168, "step": 2859 }, { "epoch": 0.54, "learning_rate": 2.8392755004766442e-06, "loss": 3.4257, "step": 2860 }, { "epoch": 0.54, "learning_rate": 2.8392183031458534e-06, "loss": 3.7269, "step": 2861 }, { "epoch": 0.55, "learning_rate": 2.8391611058150622e-06, "loss": 3.7271, "step": 2862 }, { "epoch": 0.55, "learning_rate": 2.8391039084842706e-06, "loss": 3.1115, "step": 2863 }, { "epoch": 0.55, "learning_rate": 2.83904671115348e-06, "loss": 3.7779, "step": 2864 }, { "epoch": 0.55, "learning_rate": 2.838989513822688e-06, "loss": 3.6583, "step": 2865 }, { "epoch": 0.55, "learning_rate": 2.838932316491897e-06, "loss": 3.7692, "step": 2866 }, { "epoch": 0.55, "learning_rate": 2.838875119161106e-06, "loss": 3.693, "step": 2867 }, { "epoch": 0.55, "learning_rate": 2.8388179218303146e-06, "loss": 3.7322, "step": 2868 }, { "epoch": 0.55, "learning_rate": 2.8387607244995234e-06, "loss": 3.389, "step": 2869 }, { "epoch": 0.55, "learning_rate": 2.838703527168732e-06, "loss": 3.3984, "step": 2870 }, { "epoch": 0.55, "learning_rate": 2.838646329837941e-06, "loss": 3.7464, "step": 2871 }, { "epoch": 0.55, "learning_rate": 2.8385891325071497e-06, "loss": 3.5438, "step": 2872 }, { "epoch": 0.55, "learning_rate": 2.8385319351763585e-06, "loss": 3.4825, "step": 2873 }, { "epoch": 0.55, "learning_rate": 2.8384747378455673e-06, "loss": 3.7681, "step": 2874 }, { "epoch": 0.55, "learning_rate": 2.838417540514776e-06, "loss": 3.439, "step": 2875 }, { "epoch": 0.55, "learning_rate": 2.838360343183985e-06, "loss": 3.8569, "step": 2876 }, { "epoch": 0.55, "learning_rate": 2.8383031458531937e-06, "loss": 3.3732, "step": 2877 }, { "epoch": 0.55, "learning_rate": 2.8382459485224025e-06, "loss": 3.4861, "step": 2878 }, { "epoch": 0.55, "learning_rate": 2.8381887511916113e-06, "loss": 3.9449, "step": 2879 }, { "epoch": 0.55, "learning_rate": 2.8381315538608197e-06, "loss": 3.3462, "step": 2880 }, { "epoch": 0.55, "learning_rate": 2.838074356530029e-06, "loss": 3.1744, "step": 2881 }, { "epoch": 0.55, "learning_rate": 2.8380171591992373e-06, "loss": 3.2146, "step": 2882 }, { "epoch": 0.55, "learning_rate": 2.837959961868446e-06, "loss": 3.2924, "step": 2883 }, { "epoch": 0.55, "learning_rate": 2.8379027645376553e-06, "loss": 3.6151, "step": 2884 }, { "epoch": 0.55, "learning_rate": 2.8378455672068636e-06, "loss": 3.9217, "step": 2885 }, { "epoch": 0.55, "learning_rate": 2.8377883698760724e-06, "loss": 3.6167, "step": 2886 }, { "epoch": 0.55, "learning_rate": 2.8377311725452812e-06, "loss": 3.4649, "step": 2887 }, { "epoch": 0.55, "learning_rate": 2.83767397521449e-06, "loss": 3.707, "step": 2888 }, { "epoch": 0.55, "learning_rate": 2.837616777883699e-06, "loss": 3.6989, "step": 2889 }, { "epoch": 0.55, "learning_rate": 2.8375595805529076e-06, "loss": 3.5394, "step": 2890 }, { "epoch": 0.55, "learning_rate": 2.8375023832221164e-06, "loss": 3.3651, "step": 2891 }, { "epoch": 0.55, "learning_rate": 2.837445185891325e-06, "loss": 3.5979, "step": 2892 }, { "epoch": 0.55, "learning_rate": 2.837387988560534e-06, "loss": 3.651, "step": 2893 }, { "epoch": 0.55, "learning_rate": 2.8373307912297428e-06, "loss": 3.698, "step": 2894 }, { "epoch": 0.55, "learning_rate": 2.8372735938989516e-06, "loss": 3.6152, "step": 2895 }, { "epoch": 0.55, "learning_rate": 2.8372163965681604e-06, "loss": 3.2891, "step": 2896 }, { "epoch": 0.55, "learning_rate": 2.8371591992373687e-06, "loss": 3.898, "step": 2897 }, { "epoch": 0.55, "learning_rate": 2.837102001906578e-06, "loss": 3.3835, "step": 2898 }, { "epoch": 0.55, "learning_rate": 2.8370448045757863e-06, "loss": 3.9082, "step": 2899 }, { "epoch": 0.55, "learning_rate": 2.836987607244995e-06, "loss": 3.4611, "step": 2900 }, { "epoch": 0.55, "learning_rate": 2.8369304099142043e-06, "loss": 3.3281, "step": 2901 }, { "epoch": 0.55, "learning_rate": 2.8368732125834127e-06, "loss": 3.9273, "step": 2902 }, { "epoch": 0.55, "learning_rate": 2.8368160152526215e-06, "loss": 3.8496, "step": 2903 }, { "epoch": 0.55, "learning_rate": 2.8367588179218303e-06, "loss": 3.5318, "step": 2904 }, { "epoch": 0.55, "learning_rate": 2.836701620591039e-06, "loss": 3.7693, "step": 2905 }, { "epoch": 0.55, "learning_rate": 2.836644423260248e-06, "loss": 3.5533, "step": 2906 }, { "epoch": 0.55, "learning_rate": 2.8365872259294567e-06, "loss": 3.5773, "step": 2907 }, { "epoch": 0.55, "learning_rate": 2.8365300285986655e-06, "loss": 3.7932, "step": 2908 }, { "epoch": 0.55, "learning_rate": 2.8364728312678743e-06, "loss": 3.6279, "step": 2909 }, { "epoch": 0.55, "learning_rate": 2.836415633937083e-06, "loss": 3.6394, "step": 2910 }, { "epoch": 0.55, "learning_rate": 2.836358436606292e-06, "loss": 3.6212, "step": 2911 }, { "epoch": 0.55, "learning_rate": 2.8363012392755006e-06, "loss": 3.665, "step": 2912 }, { "epoch": 0.55, "learning_rate": 2.8362440419447094e-06, "loss": 3.696, "step": 2913 }, { "epoch": 0.56, "learning_rate": 2.836186844613918e-06, "loss": 3.8784, "step": 2914 }, { "epoch": 0.56, "learning_rate": 2.836129647283127e-06, "loss": 3.7447, "step": 2915 }, { "epoch": 0.56, "learning_rate": 2.836072449952336e-06, "loss": 3.5347, "step": 2916 }, { "epoch": 0.56, "learning_rate": 2.836015252621544e-06, "loss": 3.4656, "step": 2917 }, { "epoch": 0.56, "learning_rate": 2.8359580552907534e-06, "loss": 3.3983, "step": 2918 }, { "epoch": 0.56, "learning_rate": 2.8359008579599618e-06, "loss": 3.5057, "step": 2919 }, { "epoch": 0.56, "learning_rate": 2.8358436606291706e-06, "loss": 3.724, "step": 2920 }, { "epoch": 0.56, "learning_rate": 2.8357864632983794e-06, "loss": 3.763, "step": 2921 }, { "epoch": 0.56, "learning_rate": 2.835729265967588e-06, "loss": 3.3855, "step": 2922 }, { "epoch": 0.56, "learning_rate": 2.8356720686367974e-06, "loss": 3.3194, "step": 2923 }, { "epoch": 0.56, "learning_rate": 2.8356148713060057e-06, "loss": 3.6948, "step": 2924 }, { "epoch": 0.56, "learning_rate": 2.8355576739752145e-06, "loss": 3.8185, "step": 2925 }, { "epoch": 0.56, "learning_rate": 2.8355004766444233e-06, "loss": 3.5332, "step": 2926 }, { "epoch": 0.56, "learning_rate": 2.835443279313632e-06, "loss": 3.692, "step": 2927 }, { "epoch": 0.56, "learning_rate": 2.835386081982841e-06, "loss": 3.8277, "step": 2928 }, { "epoch": 0.56, "learning_rate": 2.8353288846520497e-06, "loss": 3.6599, "step": 2929 }, { "epoch": 0.56, "learning_rate": 2.8352716873212585e-06, "loss": 3.927, "step": 2930 }, { "epoch": 0.56, "learning_rate": 2.835214489990467e-06, "loss": 3.3226, "step": 2931 }, { "epoch": 0.56, "learning_rate": 2.835157292659676e-06, "loss": 3.3724, "step": 2932 }, { "epoch": 0.56, "learning_rate": 2.835100095328885e-06, "loss": 3.4672, "step": 2933 }, { "epoch": 0.56, "learning_rate": 2.8350428979980933e-06, "loss": 3.5254, "step": 2934 }, { "epoch": 0.56, "learning_rate": 2.8349857006673025e-06, "loss": 3.4333, "step": 2935 }, { "epoch": 0.56, "learning_rate": 2.834928503336511e-06, "loss": 3.316, "step": 2936 }, { "epoch": 0.56, "learning_rate": 2.8348713060057196e-06, "loss": 3.6612, "step": 2937 }, { "epoch": 0.56, "learning_rate": 2.8348141086749284e-06, "loss": 3.5936, "step": 2938 }, { "epoch": 0.56, "learning_rate": 2.8347569113441372e-06, "loss": 3.5335, "step": 2939 }, { "epoch": 0.56, "learning_rate": 2.8346997140133464e-06, "loss": 3.5489, "step": 2940 }, { "epoch": 0.56, "learning_rate": 2.834642516682555e-06, "loss": 3.643, "step": 2941 }, { "epoch": 0.56, "learning_rate": 2.8345853193517636e-06, "loss": 3.5646, "step": 2942 }, { "epoch": 0.56, "learning_rate": 2.8345281220209724e-06, "loss": 3.2888, "step": 2943 }, { "epoch": 0.56, "learning_rate": 2.834470924690181e-06, "loss": 3.5748, "step": 2944 }, { "epoch": 0.56, "learning_rate": 2.83441372735939e-06, "loss": 3.8149, "step": 2945 }, { "epoch": 0.56, "learning_rate": 2.8343565300285988e-06, "loss": 3.4143, "step": 2946 }, { "epoch": 0.56, "learning_rate": 2.8342993326978076e-06, "loss": 3.6362, "step": 2947 }, { "epoch": 0.56, "learning_rate": 2.834242135367016e-06, "loss": 3.6651, "step": 2948 }, { "epoch": 0.56, "learning_rate": 2.834184938036225e-06, "loss": 3.784, "step": 2949 }, { "epoch": 0.56, "learning_rate": 2.834127740705434e-06, "loss": 3.6232, "step": 2950 }, { "epoch": 0.56, "learning_rate": 2.8340705433746423e-06, "loss": 3.4973, "step": 2951 }, { "epoch": 0.56, "learning_rate": 2.8340133460438515e-06, "loss": 3.5675, "step": 2952 }, { "epoch": 0.56, "learning_rate": 2.83395614871306e-06, "loss": 3.5532, "step": 2953 }, { "epoch": 0.56, "learning_rate": 2.833898951382269e-06, "loss": 3.3191, "step": 2954 }, { "epoch": 0.56, "learning_rate": 2.8338417540514775e-06, "loss": 3.8609, "step": 2955 }, { "epoch": 0.56, "learning_rate": 2.8337845567206863e-06, "loss": 3.5909, "step": 2956 }, { "epoch": 0.56, "learning_rate": 2.8337273593898955e-06, "loss": 3.6869, "step": 2957 }, { "epoch": 0.56, "learning_rate": 2.833670162059104e-06, "loss": 3.9267, "step": 2958 }, { "epoch": 0.56, "learning_rate": 2.8336129647283127e-06, "loss": 3.6812, "step": 2959 }, { "epoch": 0.56, "learning_rate": 2.8335557673975215e-06, "loss": 3.7773, "step": 2960 }, { "epoch": 0.56, "learning_rate": 2.8334985700667303e-06, "loss": 3.7944, "step": 2961 }, { "epoch": 0.56, "learning_rate": 2.833441372735939e-06, "loss": 3.4238, "step": 2962 }, { "epoch": 0.56, "learning_rate": 2.833384175405148e-06, "loss": 3.4723, "step": 2963 }, { "epoch": 0.56, "learning_rate": 2.8333269780743567e-06, "loss": 3.7515, "step": 2964 }, { "epoch": 0.56, "learning_rate": 2.8332697807435654e-06, "loss": 3.5236, "step": 2965 }, { "epoch": 0.56, "learning_rate": 2.8332125834127742e-06, "loss": 3.6572, "step": 2966 }, { "epoch": 0.57, "learning_rate": 2.833155386081983e-06, "loss": 3.5394, "step": 2967 }, { "epoch": 0.57, "learning_rate": 2.8330981887511914e-06, "loss": 3.36, "step": 2968 }, { "epoch": 0.57, "learning_rate": 2.8330409914204006e-06, "loss": 3.7814, "step": 2969 }, { "epoch": 0.57, "learning_rate": 2.832983794089609e-06, "loss": 3.6221, "step": 2970 }, { "epoch": 0.57, "learning_rate": 2.832926596758818e-06, "loss": 3.5557, "step": 2971 }, { "epoch": 0.57, "learning_rate": 2.832869399428027e-06, "loss": 3.2702, "step": 2972 }, { "epoch": 0.57, "learning_rate": 2.8328122020972354e-06, "loss": 3.3491, "step": 2973 }, { "epoch": 0.57, "learning_rate": 2.8327550047664446e-06, "loss": 3.7503, "step": 2974 }, { "epoch": 0.57, "learning_rate": 2.832697807435653e-06, "loss": 3.7768, "step": 2975 }, { "epoch": 0.57, "learning_rate": 2.8326406101048618e-06, "loss": 3.6231, "step": 2976 }, { "epoch": 0.57, "learning_rate": 2.8325834127740705e-06, "loss": 3.3231, "step": 2977 }, { "epoch": 0.57, "learning_rate": 2.8325262154432793e-06, "loss": 3.6279, "step": 2978 }, { "epoch": 0.57, "learning_rate": 2.832469018112488e-06, "loss": 3.4192, "step": 2979 }, { "epoch": 0.57, "learning_rate": 2.832411820781697e-06, "loss": 3.8247, "step": 2980 }, { "epoch": 0.57, "learning_rate": 2.8323546234509057e-06, "loss": 3.6169, "step": 2981 }, { "epoch": 0.57, "learning_rate": 2.8322974261201145e-06, "loss": 3.8147, "step": 2982 }, { "epoch": 0.57, "learning_rate": 2.8322402287893233e-06, "loss": 3.6488, "step": 2983 }, { "epoch": 0.57, "learning_rate": 2.832183031458532e-06, "loss": 3.3817, "step": 2984 }, { "epoch": 0.57, "learning_rate": 2.8321258341277405e-06, "loss": 3.4315, "step": 2985 }, { "epoch": 0.57, "learning_rate": 2.8320686367969497e-06, "loss": 3.6154, "step": 2986 }, { "epoch": 0.57, "learning_rate": 2.832011439466158e-06, "loss": 3.5969, "step": 2987 }, { "epoch": 0.57, "learning_rate": 2.8319542421353673e-06, "loss": 3.7776, "step": 2988 }, { "epoch": 0.57, "learning_rate": 2.831897044804576e-06, "loss": 3.2038, "step": 2989 }, { "epoch": 0.57, "learning_rate": 2.8318398474737844e-06, "loss": 3.841, "step": 2990 }, { "epoch": 0.57, "learning_rate": 2.8317826501429937e-06, "loss": 3.8723, "step": 2991 }, { "epoch": 0.57, "learning_rate": 2.831725452812202e-06, "loss": 3.365, "step": 2992 }, { "epoch": 0.57, "learning_rate": 2.831668255481411e-06, "loss": 3.5233, "step": 2993 }, { "epoch": 0.57, "learning_rate": 2.8316110581506196e-06, "loss": 3.6568, "step": 2994 }, { "epoch": 0.57, "learning_rate": 2.8315538608198284e-06, "loss": 3.5645, "step": 2995 }, { "epoch": 0.57, "learning_rate": 2.831496663489037e-06, "loss": 3.2386, "step": 2996 }, { "epoch": 0.57, "learning_rate": 2.831439466158246e-06, "loss": 3.1911, "step": 2997 }, { "epoch": 0.57, "learning_rate": 2.831382268827455e-06, "loss": 3.1639, "step": 2998 }, { "epoch": 0.57, "learning_rate": 2.8313250714966636e-06, "loss": 3.7942, "step": 2999 }, { "epoch": 0.57, "learning_rate": 2.8312678741658724e-06, "loss": 3.6122, "step": 3000 }, { "epoch": 0.57, "learning_rate": 2.831210676835081e-06, "loss": 3.4923, "step": 3001 }, { "epoch": 0.57, "learning_rate": 2.8311534795042895e-06, "loss": 3.381, "step": 3002 }, { "epoch": 0.57, "learning_rate": 2.8310962821734988e-06, "loss": 3.476, "step": 3003 }, { "epoch": 0.57, "learning_rate": 2.8310390848427076e-06, "loss": 3.5392, "step": 3004 }, { "epoch": 0.57, "learning_rate": 2.8309818875119164e-06, "loss": 3.7882, "step": 3005 }, { "epoch": 0.57, "learning_rate": 2.830924690181125e-06, "loss": 3.3809, "step": 3006 }, { "epoch": 0.57, "learning_rate": 2.8308674928503335e-06, "loss": 3.5058, "step": 3007 }, { "epoch": 0.57, "learning_rate": 2.8308102955195427e-06, "loss": 3.549, "step": 3008 }, { "epoch": 0.57, "learning_rate": 2.830753098188751e-06, "loss": 3.7693, "step": 3009 }, { "epoch": 0.57, "learning_rate": 2.83069590085796e-06, "loss": 3.9393, "step": 3010 }, { "epoch": 0.57, "learning_rate": 2.830638703527169e-06, "loss": 3.6634, "step": 3011 }, { "epoch": 0.57, "learning_rate": 2.8305815061963775e-06, "loss": 3.703, "step": 3012 }, { "epoch": 0.57, "learning_rate": 2.8305243088655863e-06, "loss": 3.65, "step": 3013 }, { "epoch": 0.57, "learning_rate": 2.830467111534795e-06, "loss": 3.6325, "step": 3014 }, { "epoch": 0.57, "learning_rate": 2.830409914204004e-06, "loss": 3.3794, "step": 3015 }, { "epoch": 0.57, "learning_rate": 2.8303527168732127e-06, "loss": 3.6409, "step": 3016 }, { "epoch": 0.57, "learning_rate": 2.8302955195424215e-06, "loss": 3.9662, "step": 3017 }, { "epoch": 0.57, "learning_rate": 2.8302383222116302e-06, "loss": 3.9319, "step": 3018 }, { "epoch": 0.58, "learning_rate": 2.8301811248808386e-06, "loss": 3.2572, "step": 3019 }, { "epoch": 0.58, "learning_rate": 2.830123927550048e-06, "loss": 3.8014, "step": 3020 }, { "epoch": 0.58, "learning_rate": 2.8300667302192566e-06, "loss": 3.1477, "step": 3021 }, { "epoch": 0.58, "learning_rate": 2.8300095328884654e-06, "loss": 3.5386, "step": 3022 }, { "epoch": 0.58, "learning_rate": 2.8299523355576742e-06, "loss": 3.9239, "step": 3023 }, { "epoch": 0.58, "learning_rate": 2.8298951382268826e-06, "loss": 3.5338, "step": 3024 }, { "epoch": 0.58, "learning_rate": 2.829837940896092e-06, "loss": 3.507, "step": 3025 }, { "epoch": 0.58, "learning_rate": 2.8297807435653e-06, "loss": 3.5203, "step": 3026 }, { "epoch": 0.58, "learning_rate": 2.829723546234509e-06, "loss": 3.6354, "step": 3027 }, { "epoch": 0.58, "learning_rate": 2.829666348903718e-06, "loss": 3.7162, "step": 3028 }, { "epoch": 0.58, "learning_rate": 2.8296091515729266e-06, "loss": 3.6755, "step": 3029 }, { "epoch": 0.58, "learning_rate": 2.8295519542421353e-06, "loss": 3.385, "step": 3030 }, { "epoch": 0.58, "learning_rate": 2.829494756911344e-06, "loss": 3.6654, "step": 3031 }, { "epoch": 0.58, "learning_rate": 2.829437559580553e-06, "loss": 3.4499, "step": 3032 }, { "epoch": 0.58, "learning_rate": 2.8293803622497617e-06, "loss": 3.777, "step": 3033 }, { "epoch": 0.58, "learning_rate": 2.8293231649189705e-06, "loss": 3.4252, "step": 3034 }, { "epoch": 0.58, "learning_rate": 2.8292659675881793e-06, "loss": 3.5276, "step": 3035 }, { "epoch": 0.58, "learning_rate": 2.829208770257388e-06, "loss": 3.6718, "step": 3036 }, { "epoch": 0.58, "learning_rate": 2.829151572926597e-06, "loss": 3.4365, "step": 3037 }, { "epoch": 0.58, "learning_rate": 2.8290943755958057e-06, "loss": 3.5333, "step": 3038 }, { "epoch": 0.58, "learning_rate": 2.8290371782650145e-06, "loss": 3.928, "step": 3039 }, { "epoch": 0.58, "learning_rate": 2.8289799809342233e-06, "loss": 3.4147, "step": 3040 }, { "epoch": 0.58, "learning_rate": 2.8289227836034317e-06, "loss": 3.3952, "step": 3041 }, { "epoch": 0.58, "learning_rate": 2.828865586272641e-06, "loss": 3.7066, "step": 3042 }, { "epoch": 0.58, "learning_rate": 2.8288083889418492e-06, "loss": 3.9107, "step": 3043 }, { "epoch": 0.58, "learning_rate": 2.828751191611058e-06, "loss": 3.5608, "step": 3044 }, { "epoch": 0.58, "learning_rate": 2.8286939942802673e-06, "loss": 3.5246, "step": 3045 }, { "epoch": 0.58, "learning_rate": 2.8286367969494756e-06, "loss": 3.6647, "step": 3046 }, { "epoch": 0.58, "learning_rate": 2.8285795996186844e-06, "loss": 3.6969, "step": 3047 }, { "epoch": 0.58, "learning_rate": 2.8285224022878932e-06, "loss": 3.7539, "step": 3048 }, { "epoch": 0.58, "learning_rate": 2.828465204957102e-06, "loss": 3.7107, "step": 3049 }, { "epoch": 0.58, "learning_rate": 2.828408007626311e-06, "loss": 3.1785, "step": 3050 }, { "epoch": 0.58, "learning_rate": 2.8283508102955196e-06, "loss": 3.7272, "step": 3051 }, { "epoch": 0.58, "learning_rate": 2.8282936129647284e-06, "loss": 3.4888, "step": 3052 }, { "epoch": 0.58, "learning_rate": 2.828236415633937e-06, "loss": 3.6203, "step": 3053 }, { "epoch": 0.58, "learning_rate": 2.828179218303146e-06, "loss": 3.5566, "step": 3054 }, { "epoch": 0.58, "learning_rate": 2.8281220209723548e-06, "loss": 3.6077, "step": 3055 }, { "epoch": 0.58, "learning_rate": 2.8280648236415636e-06, "loss": 3.6272, "step": 3056 }, { "epoch": 0.58, "learning_rate": 2.8280076263107724e-06, "loss": 3.5873, "step": 3057 }, { "epoch": 0.58, "learning_rate": 2.8279504289799807e-06, "loss": 3.5737, "step": 3058 }, { "epoch": 0.58, "learning_rate": 2.82789323164919e-06, "loss": 3.3506, "step": 3059 }, { "epoch": 0.58, "learning_rate": 2.8278360343183987e-06, "loss": 3.6968, "step": 3060 }, { "epoch": 0.58, "learning_rate": 2.827778836987607e-06, "loss": 3.3755, "step": 3061 }, { "epoch": 0.58, "learning_rate": 2.8277216396568163e-06, "loss": 3.4715, "step": 3062 }, { "epoch": 0.58, "learning_rate": 2.8276644423260247e-06, "loss": 3.644, "step": 3063 }, { "epoch": 0.58, "learning_rate": 2.8276072449952335e-06, "loss": 3.4339, "step": 3064 }, { "epoch": 0.58, "learning_rate": 2.8275500476644423e-06, "loss": 3.6316, "step": 3065 }, { "epoch": 0.58, "learning_rate": 2.827492850333651e-06, "loss": 3.3192, "step": 3066 }, { "epoch": 0.58, "learning_rate": 2.82743565300286e-06, "loss": 3.6967, "step": 3067 }, { "epoch": 0.58, "learning_rate": 2.8273784556720687e-06, "loss": 3.7168, "step": 3068 }, { "epoch": 0.58, "learning_rate": 2.8273212583412775e-06, "loss": 3.2331, "step": 3069 }, { "epoch": 0.58, "learning_rate": 2.8272640610104863e-06, "loss": 3.7517, "step": 3070 }, { "epoch": 0.58, "learning_rate": 2.827206863679695e-06, "loss": 3.3646, "step": 3071 }, { "epoch": 0.59, "learning_rate": 2.827149666348904e-06, "loss": 3.3396, "step": 3072 }, { "epoch": 0.59, "learning_rate": 2.8270924690181126e-06, "loss": 3.5826, "step": 3073 }, { "epoch": 0.59, "learning_rate": 2.8270352716873214e-06, "loss": 3.7277, "step": 3074 }, { "epoch": 0.59, "learning_rate": 2.82697807435653e-06, "loss": 3.7303, "step": 3075 }, { "epoch": 0.59, "learning_rate": 2.826920877025739e-06, "loss": 3.2022, "step": 3076 }, { "epoch": 0.59, "learning_rate": 2.826863679694948e-06, "loss": 3.4586, "step": 3077 }, { "epoch": 0.59, "learning_rate": 2.826806482364156e-06, "loss": 3.4953, "step": 3078 }, { "epoch": 0.59, "learning_rate": 2.8267492850333654e-06, "loss": 3.3473, "step": 3079 }, { "epoch": 0.59, "learning_rate": 2.8266920877025738e-06, "loss": 3.6802, "step": 3080 }, { "epoch": 0.59, "learning_rate": 2.8266348903717826e-06, "loss": 3.6228, "step": 3081 }, { "epoch": 0.59, "learning_rate": 2.8265776930409914e-06, "loss": 3.6488, "step": 3082 }, { "epoch": 0.59, "learning_rate": 2.8265204957102e-06, "loss": 3.5168, "step": 3083 }, { "epoch": 0.59, "learning_rate": 2.826463298379409e-06, "loss": 3.4243, "step": 3084 }, { "epoch": 0.59, "learning_rate": 2.8264061010486177e-06, "loss": 3.4551, "step": 3085 }, { "epoch": 0.59, "learning_rate": 2.8263489037178265e-06, "loss": 3.7956, "step": 3086 }, { "epoch": 0.59, "learning_rate": 2.8262917063870353e-06, "loss": 3.6464, "step": 3087 }, { "epoch": 0.59, "learning_rate": 2.826234509056244e-06, "loss": 3.6832, "step": 3088 }, { "epoch": 0.59, "learning_rate": 2.826177311725453e-06, "loss": 3.2675, "step": 3089 }, { "epoch": 0.59, "learning_rate": 2.8261201143946617e-06, "loss": 3.925, "step": 3090 }, { "epoch": 0.59, "learning_rate": 2.8260629170638705e-06, "loss": 3.5291, "step": 3091 }, { "epoch": 0.59, "learning_rate": 2.8260057197330793e-06, "loss": 3.5841, "step": 3092 }, { "epoch": 0.59, "learning_rate": 2.825948522402288e-06, "loss": 3.3329, "step": 3093 }, { "epoch": 0.59, "learning_rate": 2.825891325071497e-06, "loss": 3.7111, "step": 3094 }, { "epoch": 0.59, "learning_rate": 2.8258341277407053e-06, "loss": 3.7657, "step": 3095 }, { "epoch": 0.59, "learning_rate": 2.8257769304099145e-06, "loss": 3.7579, "step": 3096 }, { "epoch": 0.59, "learning_rate": 2.825719733079123e-06, "loss": 3.4734, "step": 3097 }, { "epoch": 0.59, "learning_rate": 2.8256625357483316e-06, "loss": 3.0457, "step": 3098 }, { "epoch": 0.59, "learning_rate": 2.825605338417541e-06, "loss": 3.5783, "step": 3099 }, { "epoch": 0.59, "learning_rate": 2.8255481410867492e-06, "loss": 3.4137, "step": 3100 }, { "epoch": 0.59, "learning_rate": 2.8254909437559584e-06, "loss": 3.2948, "step": 3101 }, { "epoch": 0.59, "learning_rate": 2.825433746425167e-06, "loss": 3.7131, "step": 3102 }, { "epoch": 0.59, "learning_rate": 2.8253765490943756e-06, "loss": 3.7752, "step": 3103 }, { "epoch": 0.59, "learning_rate": 2.8253193517635844e-06, "loss": 3.6805, "step": 3104 }, { "epoch": 0.59, "learning_rate": 2.825262154432793e-06, "loss": 3.676, "step": 3105 }, { "epoch": 0.59, "learning_rate": 2.825204957102002e-06, "loss": 3.8787, "step": 3106 }, { "epoch": 0.59, "learning_rate": 2.8251477597712108e-06, "loss": 3.1673, "step": 3107 }, { "epoch": 0.59, "learning_rate": 2.8250905624404196e-06, "loss": 3.4513, "step": 3108 }, { "epoch": 0.59, "learning_rate": 2.8250333651096284e-06, "loss": 3.3078, "step": 3109 }, { "epoch": 0.59, "learning_rate": 2.824976167778837e-06, "loss": 3.5412, "step": 3110 }, { "epoch": 0.59, "learning_rate": 2.824918970448046e-06, "loss": 3.3619, "step": 3111 }, { "epoch": 0.59, "learning_rate": 2.8248617731172543e-06, "loss": 3.4663, "step": 3112 }, { "epoch": 0.59, "learning_rate": 2.8248045757864635e-06, "loss": 3.5769, "step": 3113 }, { "epoch": 0.59, "learning_rate": 2.824747378455672e-06, "loss": 3.214, "step": 3114 }, { "epoch": 0.59, "learning_rate": 2.8246901811248807e-06, "loss": 3.3807, "step": 3115 }, { "epoch": 0.59, "learning_rate": 2.82463298379409e-06, "loss": 3.4173, "step": 3116 }, { "epoch": 0.59, "learning_rate": 2.8245757864632983e-06, "loss": 3.5969, "step": 3117 }, { "epoch": 0.59, "learning_rate": 2.8245185891325075e-06, "loss": 3.474, "step": 3118 }, { "epoch": 0.59, "learning_rate": 2.824461391801716e-06, "loss": 3.5747, "step": 3119 }, { "epoch": 0.59, "learning_rate": 2.8244041944709247e-06, "loss": 3.5924, "step": 3120 }, { "epoch": 0.59, "learning_rate": 2.8243469971401335e-06, "loss": 3.5221, "step": 3121 }, { "epoch": 0.59, "learning_rate": 2.8242897998093423e-06, "loss": 3.5609, "step": 3122 }, { "epoch": 0.59, "learning_rate": 2.824232602478551e-06, "loss": 3.3797, "step": 3123 }, { "epoch": 0.6, "learning_rate": 2.82417540514776e-06, "loss": 3.7907, "step": 3124 }, { "epoch": 0.6, "learning_rate": 2.8241182078169686e-06, "loss": 3.3711, "step": 3125 }, { "epoch": 0.6, "learning_rate": 2.8240610104861774e-06, "loss": 3.5589, "step": 3126 }, { "epoch": 0.6, "learning_rate": 2.8240038131553862e-06, "loss": 3.497, "step": 3127 }, { "epoch": 0.6, "learning_rate": 2.823946615824595e-06, "loss": 3.6044, "step": 3128 }, { "epoch": 0.6, "learning_rate": 2.8238894184938034e-06, "loss": 3.6765, "step": 3129 }, { "epoch": 0.6, "learning_rate": 2.8238322211630126e-06, "loss": 3.7233, "step": 3130 }, { "epoch": 0.6, "learning_rate": 2.823775023832221e-06, "loss": 3.7709, "step": 3131 }, { "epoch": 0.6, "learning_rate": 2.8237178265014298e-06, "loss": 3.4206, "step": 3132 }, { "epoch": 0.6, "learning_rate": 2.823660629170639e-06, "loss": 3.5249, "step": 3133 }, { "epoch": 0.6, "learning_rate": 2.8236034318398474e-06, "loss": 3.7292, "step": 3134 }, { "epoch": 0.6, "learning_rate": 2.8235462345090566e-06, "loss": 3.5111, "step": 3135 }, { "epoch": 0.6, "learning_rate": 2.823489037178265e-06, "loss": 3.3763, "step": 3136 }, { "epoch": 0.6, "learning_rate": 2.8234318398474737e-06, "loss": 3.3677, "step": 3137 }, { "epoch": 0.6, "learning_rate": 2.823374642516683e-06, "loss": 3.1279, "step": 3138 }, { "epoch": 0.6, "learning_rate": 2.8233174451858913e-06, "loss": 3.3013, "step": 3139 }, { "epoch": 0.6, "learning_rate": 2.8232602478551e-06, "loss": 3.5773, "step": 3140 }, { "epoch": 0.6, "learning_rate": 2.823203050524309e-06, "loss": 3.8041, "step": 3141 }, { "epoch": 0.6, "learning_rate": 2.8231458531935177e-06, "loss": 3.6483, "step": 3142 }, { "epoch": 0.6, "learning_rate": 2.8230886558627265e-06, "loss": 3.6274, "step": 3143 }, { "epoch": 0.6, "learning_rate": 2.8230314585319353e-06, "loss": 3.7088, "step": 3144 }, { "epoch": 0.6, "learning_rate": 2.822974261201144e-06, "loss": 3.7018, "step": 3145 }, { "epoch": 0.6, "learning_rate": 2.8229170638703525e-06, "loss": 3.5856, "step": 3146 }, { "epoch": 0.6, "learning_rate": 2.8228598665395617e-06, "loss": 3.5619, "step": 3147 }, { "epoch": 0.6, "learning_rate": 2.8228026692087705e-06, "loss": 3.7433, "step": 3148 }, { "epoch": 0.6, "learning_rate": 2.822745471877979e-06, "loss": 3.5917, "step": 3149 }, { "epoch": 0.6, "learning_rate": 2.822688274547188e-06, "loss": 3.2322, "step": 3150 }, { "epoch": 0.6, "learning_rate": 2.8226310772163964e-06, "loss": 3.4428, "step": 3151 }, { "epoch": 0.6, "learning_rate": 2.8225738798856057e-06, "loss": 3.7253, "step": 3152 }, { "epoch": 0.6, "learning_rate": 2.822516682554814e-06, "loss": 3.6234, "step": 3153 }, { "epoch": 0.6, "learning_rate": 2.822459485224023e-06, "loss": 3.5478, "step": 3154 }, { "epoch": 0.6, "learning_rate": 2.822402287893232e-06, "loss": 3.3437, "step": 3155 }, { "epoch": 0.6, "learning_rate": 2.8223450905624404e-06, "loss": 3.6252, "step": 3156 }, { "epoch": 0.6, "learning_rate": 2.822287893231649e-06, "loss": 3.5389, "step": 3157 }, { "epoch": 0.6, "learning_rate": 2.822230695900858e-06, "loss": 3.8513, "step": 3158 }, { "epoch": 0.6, "learning_rate": 2.8221734985700668e-06, "loss": 3.2073, "step": 3159 }, { "epoch": 0.6, "learning_rate": 2.8221163012392756e-06, "loss": 3.6297, "step": 3160 }, { "epoch": 0.6, "learning_rate": 2.8220591039084844e-06, "loss": 3.5698, "step": 3161 }, { "epoch": 0.6, "learning_rate": 2.822001906577693e-06, "loss": 3.0629, "step": 3162 }, { "epoch": 0.6, "learning_rate": 2.8219447092469015e-06, "loss": 3.6503, "step": 3163 }, { "epoch": 0.6, "learning_rate": 2.8218875119161108e-06, "loss": 3.6425, "step": 3164 }, { "epoch": 0.6, "learning_rate": 2.8218303145853196e-06, "loss": 3.5069, "step": 3165 }, { "epoch": 0.6, "learning_rate": 2.8217731172545283e-06, "loss": 3.9209, "step": 3166 }, { "epoch": 0.6, "learning_rate": 2.821715919923737e-06, "loss": 3.4103, "step": 3167 }, { "epoch": 0.6, "learning_rate": 2.8216587225929455e-06, "loss": 3.4747, "step": 3168 }, { "epoch": 0.6, "learning_rate": 2.8216015252621547e-06, "loss": 3.5986, "step": 3169 }, { "epoch": 0.6, "learning_rate": 2.821544327931363e-06, "loss": 3.7283, "step": 3170 }, { "epoch": 0.6, "learning_rate": 2.821487130600572e-06, "loss": 3.635, "step": 3171 }, { "epoch": 0.6, "learning_rate": 2.821429933269781e-06, "loss": 3.8549, "step": 3172 }, { "epoch": 0.6, "learning_rate": 2.8213727359389895e-06, "loss": 3.6474, "step": 3173 }, { "epoch": 0.6, "learning_rate": 2.8213155386081983e-06, "loss": 3.2948, "step": 3174 }, { "epoch": 0.6, "learning_rate": 2.821258341277407e-06, "loss": 3.7592, "step": 3175 }, { "epoch": 0.6, "learning_rate": 2.821201143946616e-06, "loss": 3.4953, "step": 3176 }, { "epoch": 0.61, "learning_rate": 2.8211439466158247e-06, "loss": 3.337, "step": 3177 }, { "epoch": 0.61, "learning_rate": 2.8210867492850334e-06, "loss": 3.2857, "step": 3178 }, { "epoch": 0.61, "learning_rate": 2.8210295519542422e-06, "loss": 3.584, "step": 3179 }, { "epoch": 0.61, "learning_rate": 2.820972354623451e-06, "loss": 3.4053, "step": 3180 }, { "epoch": 0.61, "learning_rate": 2.82091515729266e-06, "loss": 3.7187, "step": 3181 }, { "epoch": 0.61, "learning_rate": 2.8208579599618686e-06, "loss": 3.5828, "step": 3182 }, { "epoch": 0.61, "learning_rate": 2.8208007626310774e-06, "loss": 3.4962, "step": 3183 }, { "epoch": 0.61, "learning_rate": 2.820743565300286e-06, "loss": 3.6114, "step": 3184 }, { "epoch": 0.61, "learning_rate": 2.8206863679694946e-06, "loss": 3.5869, "step": 3185 }, { "epoch": 0.61, "learning_rate": 2.820629170638704e-06, "loss": 3.1753, "step": 3186 }, { "epoch": 0.61, "learning_rate": 2.8205719733079126e-06, "loss": 3.4727, "step": 3187 }, { "epoch": 0.61, "learning_rate": 2.820514775977121e-06, "loss": 3.6035, "step": 3188 }, { "epoch": 0.61, "learning_rate": 2.82045757864633e-06, "loss": 3.418, "step": 3189 }, { "epoch": 0.61, "learning_rate": 2.8204003813155385e-06, "loss": 3.6016, "step": 3190 }, { "epoch": 0.61, "learning_rate": 2.8203431839847473e-06, "loss": 3.5529, "step": 3191 }, { "epoch": 0.61, "learning_rate": 2.820285986653956e-06, "loss": 3.7614, "step": 3192 }, { "epoch": 0.61, "learning_rate": 2.820228789323165e-06, "loss": 3.6728, "step": 3193 }, { "epoch": 0.61, "learning_rate": 2.8201715919923737e-06, "loss": 3.676, "step": 3194 }, { "epoch": 0.61, "learning_rate": 2.8201143946615825e-06, "loss": 3.7665, "step": 3195 }, { "epoch": 0.61, "learning_rate": 2.8200571973307913e-06, "loss": 3.1945, "step": 3196 }, { "epoch": 0.61, "learning_rate": 2.82e-06, "loss": 3.3664, "step": 3197 }, { "epoch": 0.61, "learning_rate": 2.819942802669209e-06, "loss": 3.6506, "step": 3198 }, { "epoch": 0.61, "learning_rate": 2.8198856053384177e-06, "loss": 3.4186, "step": 3199 }, { "epoch": 0.61, "learning_rate": 2.8198284080076265e-06, "loss": 3.6666, "step": 3200 }, { "epoch": 0.61, "learning_rate": 2.8197712106768353e-06, "loss": 3.7415, "step": 3201 }, { "epoch": 0.61, "learning_rate": 2.8197140133460437e-06, "loss": 3.4095, "step": 3202 }, { "epoch": 0.61, "learning_rate": 2.819656816015253e-06, "loss": 3.0663, "step": 3203 }, { "epoch": 0.61, "learning_rate": 2.8195996186844617e-06, "loss": 3.6291, "step": 3204 }, { "epoch": 0.61, "learning_rate": 2.81954242135367e-06, "loss": 3.4977, "step": 3205 }, { "epoch": 0.61, "learning_rate": 2.8194852240228792e-06, "loss": 3.7615, "step": 3206 }, { "epoch": 0.61, "learning_rate": 2.8194280266920876e-06, "loss": 3.5224, "step": 3207 }, { "epoch": 0.61, "learning_rate": 2.8193708293612964e-06, "loss": 3.1673, "step": 3208 }, { "epoch": 0.61, "learning_rate": 2.819313632030505e-06, "loss": 3.7427, "step": 3209 }, { "epoch": 0.61, "learning_rate": 2.819256434699714e-06, "loss": 3.7021, "step": 3210 }, { "epoch": 0.61, "learning_rate": 2.819199237368923e-06, "loss": 3.7488, "step": 3211 }, { "epoch": 0.61, "learning_rate": 2.8191420400381316e-06, "loss": 3.6655, "step": 3212 }, { "epoch": 0.61, "learning_rate": 2.8190848427073404e-06, "loss": 3.9085, "step": 3213 }, { "epoch": 0.61, "learning_rate": 2.819027645376549e-06, "loss": 3.1637, "step": 3214 }, { "epoch": 0.61, "learning_rate": 2.818970448045758e-06, "loss": 3.3688, "step": 3215 }, { "epoch": 0.61, "learning_rate": 2.8189132507149668e-06, "loss": 3.7361, "step": 3216 }, { "epoch": 0.61, "learning_rate": 2.8188560533841756e-06, "loss": 3.1492, "step": 3217 }, { "epoch": 0.61, "learning_rate": 2.8187988560533844e-06, "loss": 3.4295, "step": 3218 }, { "epoch": 0.61, "learning_rate": 2.8187416587225927e-06, "loss": 3.495, "step": 3219 }, { "epoch": 0.61, "learning_rate": 2.818684461391802e-06, "loss": 3.4508, "step": 3220 }, { "epoch": 0.61, "learning_rate": 2.8186272640610107e-06, "loss": 3.9157, "step": 3221 }, { "epoch": 0.61, "learning_rate": 2.818570066730219e-06, "loss": 3.3539, "step": 3222 }, { "epoch": 0.61, "learning_rate": 2.8185128693994283e-06, "loss": 3.67, "step": 3223 }, { "epoch": 0.61, "learning_rate": 2.8184556720686367e-06, "loss": 3.6893, "step": 3224 }, { "epoch": 0.61, "learning_rate": 2.8183984747378455e-06, "loss": 3.6558, "step": 3225 }, { "epoch": 0.61, "learning_rate": 2.8183412774070547e-06, "loss": 3.4423, "step": 3226 }, { "epoch": 0.61, "learning_rate": 2.818284080076263e-06, "loss": 3.4248, "step": 3227 }, { "epoch": 0.61, "learning_rate": 2.818226882745472e-06, "loss": 3.2241, "step": 3228 }, { "epoch": 0.62, "learning_rate": 2.8181696854146807e-06, "loss": 3.7471, "step": 3229 }, { "epoch": 0.62, "learning_rate": 2.8181124880838895e-06, "loss": 3.8975, "step": 3230 }, { "epoch": 0.62, "learning_rate": 2.8180552907530982e-06, "loss": 3.6956, "step": 3231 }, { "epoch": 0.62, "learning_rate": 2.817998093422307e-06, "loss": 3.4827, "step": 3232 }, { "epoch": 0.62, "learning_rate": 2.817940896091516e-06, "loss": 3.7168, "step": 3233 }, { "epoch": 0.62, "learning_rate": 2.8178836987607246e-06, "loss": 3.7893, "step": 3234 }, { "epoch": 0.62, "learning_rate": 2.8178265014299334e-06, "loss": 3.1791, "step": 3235 }, { "epoch": 0.62, "learning_rate": 2.8177693040991422e-06, "loss": 3.1175, "step": 3236 }, { "epoch": 0.62, "learning_rate": 2.817712106768351e-06, "loss": 3.7126, "step": 3237 }, { "epoch": 0.62, "learning_rate": 2.81765490943756e-06, "loss": 3.2893, "step": 3238 }, { "epoch": 0.62, "learning_rate": 2.817597712106768e-06, "loss": 3.5864, "step": 3239 }, { "epoch": 0.62, "learning_rate": 2.8175405147759774e-06, "loss": 3.5437, "step": 3240 }, { "epoch": 0.62, "learning_rate": 2.8174833174451858e-06, "loss": 3.5572, "step": 3241 }, { "epoch": 0.62, "learning_rate": 2.8174261201143946e-06, "loss": 3.2807, "step": 3242 }, { "epoch": 0.62, "learning_rate": 2.8173689227836038e-06, "loss": 3.627, "step": 3243 }, { "epoch": 0.62, "learning_rate": 2.817311725452812e-06, "loss": 3.5203, "step": 3244 }, { "epoch": 0.62, "learning_rate": 2.817254528122021e-06, "loss": 3.8152, "step": 3245 }, { "epoch": 0.62, "learning_rate": 2.8171973307912297e-06, "loss": 3.2629, "step": 3246 }, { "epoch": 0.62, "learning_rate": 2.8171401334604385e-06, "loss": 3.5616, "step": 3247 }, { "epoch": 0.62, "learning_rate": 2.8170829361296473e-06, "loss": 3.63, "step": 3248 }, { "epoch": 0.62, "learning_rate": 2.817025738798856e-06, "loss": 3.1964, "step": 3249 }, { "epoch": 0.62, "learning_rate": 2.816968541468065e-06, "loss": 3.6973, "step": 3250 }, { "epoch": 0.62, "learning_rate": 2.8169113441372737e-06, "loss": 3.1751, "step": 3251 }, { "epoch": 0.62, "learning_rate": 2.8168541468064825e-06, "loss": 3.713, "step": 3252 }, { "epoch": 0.62, "learning_rate": 2.8167969494756913e-06, "loss": 3.457, "step": 3253 }, { "epoch": 0.62, "learning_rate": 2.8167397521449e-06, "loss": 3.3806, "step": 3254 }, { "epoch": 0.62, "learning_rate": 2.816682554814109e-06, "loss": 3.4691, "step": 3255 }, { "epoch": 0.62, "learning_rate": 2.8166253574833172e-06, "loss": 3.3589, "step": 3256 }, { "epoch": 0.62, "learning_rate": 2.8165681601525265e-06, "loss": 3.7699, "step": 3257 }, { "epoch": 0.62, "learning_rate": 2.816510962821735e-06, "loss": 3.4709, "step": 3258 }, { "epoch": 0.62, "learning_rate": 2.8164537654909436e-06, "loss": 3.7613, "step": 3259 }, { "epoch": 0.62, "learning_rate": 2.816396568160153e-06, "loss": 3.3552, "step": 3260 }, { "epoch": 0.62, "learning_rate": 2.8163393708293612e-06, "loss": 3.8315, "step": 3261 }, { "epoch": 0.62, "learning_rate": 2.81628217349857e-06, "loss": 3.5094, "step": 3262 }, { "epoch": 0.62, "learning_rate": 2.816224976167779e-06, "loss": 3.6152, "step": 3263 }, { "epoch": 0.62, "learning_rate": 2.8161677788369876e-06, "loss": 3.2326, "step": 3264 }, { "epoch": 0.62, "learning_rate": 2.8161105815061964e-06, "loss": 3.4253, "step": 3265 }, { "epoch": 0.62, "learning_rate": 2.816053384175405e-06, "loss": 3.7688, "step": 3266 }, { "epoch": 0.62, "learning_rate": 2.815996186844614e-06, "loss": 3.5819, "step": 3267 }, { "epoch": 0.62, "learning_rate": 2.8159389895138228e-06, "loss": 3.5756, "step": 3268 }, { "epoch": 0.62, "learning_rate": 2.8158817921830316e-06, "loss": 3.8257, "step": 3269 }, { "epoch": 0.62, "learning_rate": 2.8158245948522404e-06, "loss": 3.2148, "step": 3270 }, { "epoch": 0.62, "learning_rate": 2.815767397521449e-06, "loss": 3.217, "step": 3271 }, { "epoch": 0.62, "learning_rate": 2.815710200190658e-06, "loss": 3.4652, "step": 3272 }, { "epoch": 0.62, "learning_rate": 2.8156530028598663e-06, "loss": 3.7407, "step": 3273 }, { "epoch": 0.62, "learning_rate": 2.8155958055290755e-06, "loss": 3.0951, "step": 3274 }, { "epoch": 0.62, "learning_rate": 2.8155386081982843e-06, "loss": 3.271, "step": 3275 }, { "epoch": 0.62, "learning_rate": 2.8154814108674927e-06, "loss": 3.6296, "step": 3276 }, { "epoch": 0.62, "learning_rate": 2.815424213536702e-06, "loss": 3.5759, "step": 3277 }, { "epoch": 0.62, "learning_rate": 2.8153670162059103e-06, "loss": 3.5387, "step": 3278 }, { "epoch": 0.62, "learning_rate": 2.815309818875119e-06, "loss": 3.6707, "step": 3279 }, { "epoch": 0.62, "learning_rate": 2.815252621544328e-06, "loss": 3.78, "step": 3280 }, { "epoch": 0.62, "learning_rate": 2.8151954242135367e-06, "loss": 3.5547, "step": 3281 }, { "epoch": 0.63, "learning_rate": 2.815138226882746e-06, "loss": 3.7688, "step": 3282 }, { "epoch": 0.63, "learning_rate": 2.8150810295519543e-06, "loss": 3.2325, "step": 3283 }, { "epoch": 0.63, "learning_rate": 2.815023832221163e-06, "loss": 3.438, "step": 3284 }, { "epoch": 0.63, "learning_rate": 2.814966634890372e-06, "loss": 3.5558, "step": 3285 }, { "epoch": 0.63, "learning_rate": 2.8149094375595806e-06, "loss": 3.7635, "step": 3286 }, { "epoch": 0.63, "learning_rate": 2.8148522402287894e-06, "loss": 3.5607, "step": 3287 }, { "epoch": 0.63, "learning_rate": 2.8147950428979982e-06, "loss": 3.6053, "step": 3288 }, { "epoch": 0.63, "learning_rate": 2.814737845567207e-06, "loss": 3.4986, "step": 3289 }, { "epoch": 0.63, "learning_rate": 2.8146806482364154e-06, "loss": 3.7129, "step": 3290 }, { "epoch": 0.63, "learning_rate": 2.8146234509056246e-06, "loss": 3.5718, "step": 3291 }, { "epoch": 0.63, "learning_rate": 2.8145662535748334e-06, "loss": 3.5275, "step": 3292 }, { "epoch": 0.63, "learning_rate": 2.8145090562440418e-06, "loss": 3.6437, "step": 3293 }, { "epoch": 0.63, "learning_rate": 2.814451858913251e-06, "loss": 3.7374, "step": 3294 }, { "epoch": 0.63, "learning_rate": 2.8143946615824594e-06, "loss": 3.6503, "step": 3295 }, { "epoch": 0.63, "learning_rate": 2.8143374642516686e-06, "loss": 3.6126, "step": 3296 }, { "epoch": 0.63, "learning_rate": 2.814280266920877e-06, "loss": 3.6501, "step": 3297 }, { "epoch": 0.63, "learning_rate": 2.8142230695900857e-06, "loss": 3.6057, "step": 3298 }, { "epoch": 0.63, "learning_rate": 2.814165872259295e-06, "loss": 3.5644, "step": 3299 }, { "epoch": 0.63, "learning_rate": 2.8141086749285033e-06, "loss": 3.5708, "step": 3300 }, { "epoch": 0.63, "learning_rate": 2.814051477597712e-06, "loss": 3.7359, "step": 3301 }, { "epoch": 0.63, "learning_rate": 2.813994280266921e-06, "loss": 3.6839, "step": 3302 }, { "epoch": 0.63, "learning_rate": 2.8139370829361297e-06, "loss": 3.558, "step": 3303 }, { "epoch": 0.63, "learning_rate": 2.8138798856053385e-06, "loss": 3.4269, "step": 3304 }, { "epoch": 0.63, "learning_rate": 2.8138226882745473e-06, "loss": 3.7127, "step": 3305 }, { "epoch": 0.63, "learning_rate": 2.813765490943756e-06, "loss": 3.2645, "step": 3306 }, { "epoch": 0.63, "learning_rate": 2.8137082936129645e-06, "loss": 3.2927, "step": 3307 }, { "epoch": 0.63, "learning_rate": 2.8136510962821737e-06, "loss": 3.4058, "step": 3308 }, { "epoch": 0.63, "learning_rate": 2.8135938989513825e-06, "loss": 3.4537, "step": 3309 }, { "epoch": 0.63, "learning_rate": 2.813536701620591e-06, "loss": 3.4628, "step": 3310 }, { "epoch": 0.63, "learning_rate": 2.8134795042898e-06, "loss": 3.3123, "step": 3311 }, { "epoch": 0.63, "learning_rate": 2.8134223069590084e-06, "loss": 3.6699, "step": 3312 }, { "epoch": 0.63, "learning_rate": 2.8133651096282176e-06, "loss": 3.234, "step": 3313 }, { "epoch": 0.63, "learning_rate": 2.8133079122974264e-06, "loss": 3.2517, "step": 3314 }, { "epoch": 0.63, "learning_rate": 2.813250714966635e-06, "loss": 3.4346, "step": 3315 }, { "epoch": 0.63, "learning_rate": 2.813193517635844e-06, "loss": 3.2137, "step": 3316 }, { "epoch": 0.63, "learning_rate": 2.8131363203050524e-06, "loss": 3.4464, "step": 3317 }, { "epoch": 0.63, "learning_rate": 2.813079122974261e-06, "loss": 3.481, "step": 3318 }, { "epoch": 0.63, "learning_rate": 2.81302192564347e-06, "loss": 3.5787, "step": 3319 }, { "epoch": 0.63, "learning_rate": 2.8129647283126788e-06, "loss": 3.2337, "step": 3320 }, { "epoch": 0.63, "learning_rate": 2.8129075309818876e-06, "loss": 3.5924, "step": 3321 }, { "epoch": 0.63, "learning_rate": 2.8128503336510964e-06, "loss": 3.6202, "step": 3322 }, { "epoch": 0.63, "learning_rate": 2.812793136320305e-06, "loss": 3.4406, "step": 3323 }, { "epoch": 0.63, "learning_rate": 2.812735938989514e-06, "loss": 3.3418, "step": 3324 }, { "epoch": 0.63, "learning_rate": 2.8126787416587227e-06, "loss": 3.4426, "step": 3325 }, { "epoch": 0.63, "learning_rate": 2.8126215443279315e-06, "loss": 3.6916, "step": 3326 }, { "epoch": 0.63, "learning_rate": 2.81256434699714e-06, "loss": 3.4891, "step": 3327 }, { "epoch": 0.63, "learning_rate": 2.812507149666349e-06, "loss": 3.2594, "step": 3328 }, { "epoch": 0.63, "learning_rate": 2.8124499523355575e-06, "loss": 3.8301, "step": 3329 }, { "epoch": 0.63, "learning_rate": 2.8123927550047667e-06, "loss": 3.9479, "step": 3330 }, { "epoch": 0.63, "learning_rate": 2.8123355576739755e-06, "loss": 3.4838, "step": 3331 }, { "epoch": 0.63, "learning_rate": 2.812278360343184e-06, "loss": 3.6246, "step": 3332 }, { "epoch": 0.63, "learning_rate": 2.812221163012393e-06, "loss": 3.1943, "step": 3333 }, { "epoch": 0.64, "learning_rate": 2.8121639656816015e-06, "loss": 3.622, "step": 3334 }, { "epoch": 0.64, "learning_rate": 2.8121067683508103e-06, "loss": 3.2987, "step": 3335 }, { "epoch": 0.64, "learning_rate": 2.812049571020019e-06, "loss": 3.8626, "step": 3336 }, { "epoch": 0.64, "learning_rate": 2.811992373689228e-06, "loss": 3.6721, "step": 3337 }, { "epoch": 0.64, "learning_rate": 2.8119351763584366e-06, "loss": 3.1842, "step": 3338 }, { "epoch": 0.64, "learning_rate": 2.8118779790276454e-06, "loss": 3.7552, "step": 3339 }, { "epoch": 0.64, "learning_rate": 2.8118207816968542e-06, "loss": 3.4804, "step": 3340 }, { "epoch": 0.64, "learning_rate": 2.811763584366063e-06, "loss": 3.4696, "step": 3341 }, { "epoch": 0.64, "learning_rate": 2.811706387035272e-06, "loss": 3.4938, "step": 3342 }, { "epoch": 0.64, "learning_rate": 2.8116491897044806e-06, "loss": 3.5536, "step": 3343 }, { "epoch": 0.64, "learning_rate": 2.811591992373689e-06, "loss": 3.5208, "step": 3344 }, { "epoch": 0.64, "learning_rate": 2.811534795042898e-06, "loss": 3.477, "step": 3345 }, { "epoch": 0.64, "learning_rate": 2.8114775977121066e-06, "loss": 3.6402, "step": 3346 }, { "epoch": 0.64, "learning_rate": 2.811420400381316e-06, "loss": 3.805, "step": 3347 }, { "epoch": 0.64, "learning_rate": 2.8113632030505246e-06, "loss": 3.268, "step": 3348 }, { "epoch": 0.64, "learning_rate": 2.811306005719733e-06, "loss": 3.7154, "step": 3349 }, { "epoch": 0.64, "learning_rate": 2.811248808388942e-06, "loss": 3.2931, "step": 3350 }, { "epoch": 0.64, "learning_rate": 2.8111916110581505e-06, "loss": 3.4888, "step": 3351 }, { "epoch": 0.64, "learning_rate": 2.8111344137273593e-06, "loss": 3.452, "step": 3352 }, { "epoch": 0.64, "learning_rate": 2.811077216396568e-06, "loss": 3.9928, "step": 3353 }, { "epoch": 0.64, "learning_rate": 2.811020019065777e-06, "loss": 3.3554, "step": 3354 }, { "epoch": 0.64, "learning_rate": 2.8109628217349857e-06, "loss": 3.4513, "step": 3355 }, { "epoch": 0.64, "learning_rate": 2.8109056244041945e-06, "loss": 3.4703, "step": 3356 }, { "epoch": 0.64, "learning_rate": 2.8108484270734033e-06, "loss": 3.5455, "step": 3357 }, { "epoch": 0.64, "learning_rate": 2.810791229742612e-06, "loss": 3.2716, "step": 3358 }, { "epoch": 0.64, "learning_rate": 2.810734032411821e-06, "loss": 3.4314, "step": 3359 }, { "epoch": 0.64, "learning_rate": 2.8106768350810297e-06, "loss": 3.3014, "step": 3360 }, { "epoch": 0.64, "learning_rate": 2.8106196377502385e-06, "loss": 3.2944, "step": 3361 }, { "epoch": 0.64, "learning_rate": 2.8105624404194473e-06, "loss": 3.4797, "step": 3362 }, { "epoch": 0.64, "learning_rate": 2.810505243088656e-06, "loss": 3.8435, "step": 3363 }, { "epoch": 0.64, "learning_rate": 2.810448045757865e-06, "loss": 3.6444, "step": 3364 }, { "epoch": 0.64, "learning_rate": 2.8103908484270737e-06, "loss": 3.6809, "step": 3365 }, { "epoch": 0.64, "learning_rate": 2.810333651096282e-06, "loss": 3.2519, "step": 3366 }, { "epoch": 0.64, "learning_rate": 2.8102764537654912e-06, "loss": 3.6857, "step": 3367 }, { "epoch": 0.64, "learning_rate": 2.8102192564346996e-06, "loss": 3.6412, "step": 3368 }, { "epoch": 0.64, "learning_rate": 2.8101620591039084e-06, "loss": 3.807, "step": 3369 }, { "epoch": 0.64, "learning_rate": 2.8101048617731176e-06, "loss": 3.5361, "step": 3370 }, { "epoch": 0.64, "learning_rate": 2.810047664442326e-06, "loss": 3.6725, "step": 3371 }, { "epoch": 0.64, "learning_rate": 2.809990467111535e-06, "loss": 3.672, "step": 3372 }, { "epoch": 0.64, "learning_rate": 2.8099332697807436e-06, "loss": 3.6242, "step": 3373 }, { "epoch": 0.64, "learning_rate": 2.8098760724499524e-06, "loss": 3.8602, "step": 3374 }, { "epoch": 0.64, "learning_rate": 2.809818875119161e-06, "loss": 3.792, "step": 3375 }, { "epoch": 0.64, "learning_rate": 2.80976167778837e-06, "loss": 3.4527, "step": 3376 }, { "epoch": 0.64, "learning_rate": 2.8097044804575788e-06, "loss": 3.3171, "step": 3377 }, { "epoch": 0.64, "learning_rate": 2.8096472831267876e-06, "loss": 3.8406, "step": 3378 }, { "epoch": 0.64, "learning_rate": 2.8095900857959963e-06, "loss": 3.461, "step": 3379 }, { "epoch": 0.64, "learning_rate": 2.809532888465205e-06, "loss": 3.5986, "step": 3380 }, { "epoch": 0.64, "learning_rate": 2.809475691134414e-06, "loss": 3.6553, "step": 3381 }, { "epoch": 0.64, "learning_rate": 2.8094184938036227e-06, "loss": 3.1607, "step": 3382 }, { "epoch": 0.64, "learning_rate": 2.809361296472831e-06, "loss": 3.2103, "step": 3383 }, { "epoch": 0.64, "learning_rate": 2.8093040991420403e-06, "loss": 3.8222, "step": 3384 }, { "epoch": 0.64, "learning_rate": 2.8092469018112487e-06, "loss": 3.7509, "step": 3385 }, { "epoch": 0.64, "learning_rate": 2.8091897044804575e-06, "loss": 3.3287, "step": 3386 }, { "epoch": 0.65, "learning_rate": 2.8091325071496667e-06, "loss": 3.2556, "step": 3387 }, { "epoch": 0.65, "learning_rate": 2.809075309818875e-06, "loss": 3.5614, "step": 3388 }, { "epoch": 0.65, "learning_rate": 2.809018112488084e-06, "loss": 3.7591, "step": 3389 }, { "epoch": 0.65, "learning_rate": 2.8089609151572927e-06, "loss": 3.6092, "step": 3390 }, { "epoch": 0.65, "learning_rate": 2.8089037178265014e-06, "loss": 3.4002, "step": 3391 }, { "epoch": 0.65, "learning_rate": 2.8088465204957102e-06, "loss": 3.6292, "step": 3392 }, { "epoch": 0.65, "learning_rate": 2.808789323164919e-06, "loss": 3.3511, "step": 3393 }, { "epoch": 0.65, "learning_rate": 2.808732125834128e-06, "loss": 3.318, "step": 3394 }, { "epoch": 0.65, "learning_rate": 2.8086749285033366e-06, "loss": 3.6488, "step": 3395 }, { "epoch": 0.65, "learning_rate": 2.8086177311725454e-06, "loss": 3.346, "step": 3396 }, { "epoch": 0.65, "learning_rate": 2.808560533841754e-06, "loss": 3.3479, "step": 3397 }, { "epoch": 0.65, "learning_rate": 2.808503336510963e-06, "loss": 3.1526, "step": 3398 }, { "epoch": 0.65, "learning_rate": 2.808446139180172e-06, "loss": 3.6955, "step": 3399 }, { "epoch": 0.65, "learning_rate": 2.80838894184938e-06, "loss": 3.4013, "step": 3400 }, { "epoch": 0.65, "learning_rate": 2.8083317445185894e-06, "loss": 3.3796, "step": 3401 }, { "epoch": 0.65, "learning_rate": 2.808274547187798e-06, "loss": 3.6244, "step": 3402 }, { "epoch": 0.65, "learning_rate": 2.8082173498570065e-06, "loss": 3.6263, "step": 3403 }, { "epoch": 0.65, "learning_rate": 2.8081601525262158e-06, "loss": 3.6199, "step": 3404 }, { "epoch": 0.65, "learning_rate": 2.808102955195424e-06, "loss": 3.3462, "step": 3405 }, { "epoch": 0.65, "learning_rate": 2.808045757864633e-06, "loss": 3.241, "step": 3406 }, { "epoch": 0.65, "learning_rate": 2.8079885605338417e-06, "loss": 3.2194, "step": 3407 }, { "epoch": 0.65, "learning_rate": 2.8079313632030505e-06, "loss": 3.425, "step": 3408 }, { "epoch": 0.65, "learning_rate": 2.8078741658722593e-06, "loss": 3.5806, "step": 3409 }, { "epoch": 0.65, "learning_rate": 2.807816968541468e-06, "loss": 3.3326, "step": 3410 }, { "epoch": 0.65, "learning_rate": 2.807759771210677e-06, "loss": 3.6809, "step": 3411 }, { "epoch": 0.65, "learning_rate": 2.8077025738798857e-06, "loss": 3.8057, "step": 3412 }, { "epoch": 0.65, "learning_rate": 2.8076453765490945e-06, "loss": 3.2615, "step": 3413 }, { "epoch": 0.65, "learning_rate": 2.8075881792183033e-06, "loss": 3.5893, "step": 3414 }, { "epoch": 0.65, "learning_rate": 2.807530981887512e-06, "loss": 3.4566, "step": 3415 }, { "epoch": 0.65, "learning_rate": 2.807473784556721e-06, "loss": 3.4042, "step": 3416 }, { "epoch": 0.65, "learning_rate": 2.8074165872259292e-06, "loss": 3.39, "step": 3417 }, { "epoch": 0.65, "learning_rate": 2.8073593898951385e-06, "loss": 3.7978, "step": 3418 }, { "epoch": 0.65, "learning_rate": 2.8073021925643473e-06, "loss": 3.6099, "step": 3419 }, { "epoch": 0.65, "learning_rate": 2.8072449952335556e-06, "loss": 3.8881, "step": 3420 }, { "epoch": 0.65, "learning_rate": 2.807187797902765e-06, "loss": 3.6931, "step": 3421 }, { "epoch": 0.65, "learning_rate": 2.807130600571973e-06, "loss": 3.5985, "step": 3422 }, { "epoch": 0.65, "learning_rate": 2.807073403241182e-06, "loss": 3.5852, "step": 3423 }, { "epoch": 0.65, "learning_rate": 2.807016205910391e-06, "loss": 3.5561, "step": 3424 }, { "epoch": 0.65, "learning_rate": 2.8069590085795996e-06, "loss": 3.3554, "step": 3425 }, { "epoch": 0.65, "learning_rate": 2.806901811248809e-06, "loss": 3.2565, "step": 3426 }, { "epoch": 0.65, "learning_rate": 2.806844613918017e-06, "loss": 3.5233, "step": 3427 }, { "epoch": 0.65, "learning_rate": 2.806787416587226e-06, "loss": 3.6403, "step": 3428 }, { "epoch": 0.65, "learning_rate": 2.8067302192564348e-06, "loss": 3.3638, "step": 3429 }, { "epoch": 0.65, "learning_rate": 2.8066730219256436e-06, "loss": 3.0755, "step": 3430 }, { "epoch": 0.65, "learning_rate": 2.8066158245948524e-06, "loss": 3.7088, "step": 3431 }, { "epoch": 0.65, "learning_rate": 2.806558627264061e-06, "loss": 3.9491, "step": 3432 }, { "epoch": 0.65, "learning_rate": 2.80650142993327e-06, "loss": 3.6239, "step": 3433 }, { "epoch": 0.65, "learning_rate": 2.8064442326024783e-06, "loss": 3.7998, "step": 3434 }, { "epoch": 0.65, "learning_rate": 2.8063870352716875e-06, "loss": 3.6245, "step": 3435 }, { "epoch": 0.65, "learning_rate": 2.8063298379408963e-06, "loss": 3.3681, "step": 3436 }, { "epoch": 0.65, "learning_rate": 2.8062726406101047e-06, "loss": 3.4616, "step": 3437 }, { "epoch": 0.65, "learning_rate": 2.806215443279314e-06, "loss": 3.3594, "step": 3438 }, { "epoch": 0.66, "learning_rate": 2.8061582459485223e-06, "loss": 3.5168, "step": 3439 }, { "epoch": 0.66, "learning_rate": 2.806101048617731e-06, "loss": 3.5166, "step": 3440 }, { "epoch": 0.66, "learning_rate": 2.80604385128694e-06, "loss": 3.3384, "step": 3441 }, { "epoch": 0.66, "learning_rate": 2.8059866539561487e-06, "loss": 2.9793, "step": 3442 }, { "epoch": 0.66, "learning_rate": 2.805929456625358e-06, "loss": 3.637, "step": 3443 }, { "epoch": 0.66, "learning_rate": 2.8058722592945662e-06, "loss": 3.374, "step": 3444 }, { "epoch": 0.66, "learning_rate": 2.805815061963775e-06, "loss": 3.2841, "step": 3445 }, { "epoch": 0.66, "learning_rate": 2.805757864632984e-06, "loss": 3.6137, "step": 3446 }, { "epoch": 0.66, "learning_rate": 2.8057006673021926e-06, "loss": 3.2459, "step": 3447 }, { "epoch": 0.66, "learning_rate": 2.8056434699714014e-06, "loss": 3.7537, "step": 3448 }, { "epoch": 0.66, "learning_rate": 2.8055862726406102e-06, "loss": 3.5301, "step": 3449 }, { "epoch": 0.66, "learning_rate": 2.805529075309819e-06, "loss": 3.5851, "step": 3450 }, { "epoch": 0.66, "learning_rate": 2.805471877979028e-06, "loss": 3.5735, "step": 3451 }, { "epoch": 0.66, "learning_rate": 2.8054146806482366e-06, "loss": 3.4324, "step": 3452 }, { "epoch": 0.66, "learning_rate": 2.8053574833174454e-06, "loss": 3.4413, "step": 3453 }, { "epoch": 0.66, "learning_rate": 2.8053002859866538e-06, "loss": 3.8126, "step": 3454 }, { "epoch": 0.66, "learning_rate": 2.805243088655863e-06, "loss": 3.2936, "step": 3455 }, { "epoch": 0.66, "learning_rate": 2.8051858913250714e-06, "loss": 3.8616, "step": 3456 }, { "epoch": 0.66, "learning_rate": 2.80512869399428e-06, "loss": 3.5164, "step": 3457 }, { "epoch": 0.66, "learning_rate": 2.8050714966634894e-06, "loss": 3.6905, "step": 3458 }, { "epoch": 0.66, "learning_rate": 2.8050142993326977e-06, "loss": 3.5444, "step": 3459 }, { "epoch": 0.66, "learning_rate": 2.804957102001907e-06, "loss": 3.2426, "step": 3460 }, { "epoch": 0.66, "learning_rate": 2.8048999046711153e-06, "loss": 3.5028, "step": 3461 }, { "epoch": 0.66, "learning_rate": 2.804842707340324e-06, "loss": 3.3174, "step": 3462 }, { "epoch": 0.66, "learning_rate": 2.804785510009533e-06, "loss": 3.4907, "step": 3463 }, { "epoch": 0.66, "learning_rate": 2.8047283126787417e-06, "loss": 3.6333, "step": 3464 }, { "epoch": 0.66, "learning_rate": 2.8046711153479505e-06, "loss": 3.5763, "step": 3465 }, { "epoch": 0.66, "learning_rate": 2.8046139180171593e-06, "loss": 3.3259, "step": 3466 }, { "epoch": 0.66, "learning_rate": 2.804556720686368e-06, "loss": 3.4623, "step": 3467 }, { "epoch": 0.66, "learning_rate": 2.804499523355577e-06, "loss": 3.9071, "step": 3468 }, { "epoch": 0.66, "learning_rate": 2.8044423260247857e-06, "loss": 3.2737, "step": 3469 }, { "epoch": 0.66, "learning_rate": 2.8043851286939945e-06, "loss": 3.9108, "step": 3470 }, { "epoch": 0.66, "learning_rate": 2.804327931363203e-06, "loss": 3.686, "step": 3471 }, { "epoch": 0.66, "learning_rate": 2.804270734032412e-06, "loss": 3.3882, "step": 3472 }, { "epoch": 0.66, "learning_rate": 2.8042135367016204e-06, "loss": 3.7877, "step": 3473 }, { "epoch": 0.66, "learning_rate": 2.8041563393708292e-06, "loss": 3.7326, "step": 3474 }, { "epoch": 0.66, "learning_rate": 2.8040991420400384e-06, "loss": 3.4908, "step": 3475 }, { "epoch": 0.66, "learning_rate": 2.804041944709247e-06, "loss": 3.6131, "step": 3476 }, { "epoch": 0.66, "learning_rate": 2.803984747378456e-06, "loss": 3.7011, "step": 3477 }, { "epoch": 0.66, "learning_rate": 2.8039275500476644e-06, "loss": 3.6674, "step": 3478 }, { "epoch": 0.66, "learning_rate": 2.803870352716873e-06, "loss": 3.8927, "step": 3479 }, { "epoch": 0.66, "learning_rate": 2.803813155386082e-06, "loss": 3.1763, "step": 3480 }, { "epoch": 0.66, "learning_rate": 2.8037559580552908e-06, "loss": 3.3606, "step": 3481 }, { "epoch": 0.66, "learning_rate": 2.8036987607244996e-06, "loss": 3.496, "step": 3482 }, { "epoch": 0.66, "learning_rate": 2.8036415633937084e-06, "loss": 3.4091, "step": 3483 }, { "epoch": 0.66, "learning_rate": 2.803584366062917e-06, "loss": 3.213, "step": 3484 }, { "epoch": 0.66, "learning_rate": 2.803527168732126e-06, "loss": 3.2023, "step": 3485 }, { "epoch": 0.66, "learning_rate": 2.8034699714013347e-06, "loss": 3.6754, "step": 3486 }, { "epoch": 0.66, "learning_rate": 2.8034127740705435e-06, "loss": 3.6333, "step": 3487 }, { "epoch": 0.66, "learning_rate": 2.803355576739752e-06, "loss": 3.6366, "step": 3488 }, { "epoch": 0.66, "learning_rate": 2.803298379408961e-06, "loss": 3.6601, "step": 3489 }, { "epoch": 0.66, "learning_rate": 2.80324118207817e-06, "loss": 3.7532, "step": 3490 }, { "epoch": 0.66, "learning_rate": 2.8031839847473787e-06, "loss": 3.9213, "step": 3491 }, { "epoch": 0.67, "learning_rate": 2.8031267874165875e-06, "loss": 3.4921, "step": 3492 }, { "epoch": 0.67, "learning_rate": 2.803069590085796e-06, "loss": 3.4628, "step": 3493 }, { "epoch": 0.67, "learning_rate": 2.803012392755005e-06, "loss": 3.3497, "step": 3494 }, { "epoch": 0.67, "learning_rate": 2.8029551954242135e-06, "loss": 3.5803, "step": 3495 }, { "epoch": 0.67, "learning_rate": 2.8028979980934223e-06, "loss": 3.5004, "step": 3496 }, { "epoch": 0.67, "learning_rate": 2.8028408007626315e-06, "loss": 3.2845, "step": 3497 }, { "epoch": 0.67, "learning_rate": 2.80278360343184e-06, "loss": 3.8195, "step": 3498 }, { "epoch": 0.67, "learning_rate": 2.8027264061010486e-06, "loss": 3.6353, "step": 3499 }, { "epoch": 0.67, "learning_rate": 2.8026692087702574e-06, "loss": 3.5928, "step": 3500 }, { "epoch": 0.67, "learning_rate": 2.8026120114394662e-06, "loss": 3.8601, "step": 3501 }, { "epoch": 0.67, "learning_rate": 2.802554814108675e-06, "loss": 3.8213, "step": 3502 }, { "epoch": 0.67, "learning_rate": 2.802497616777884e-06, "loss": 3.1783, "step": 3503 }, { "epoch": 0.67, "learning_rate": 2.8024404194470926e-06, "loss": 3.6829, "step": 3504 }, { "epoch": 0.67, "learning_rate": 2.802383222116301e-06, "loss": 3.0841, "step": 3505 }, { "epoch": 0.67, "learning_rate": 2.80232602478551e-06, "loss": 4.0469, "step": 3506 }, { "epoch": 0.67, "learning_rate": 2.802268827454719e-06, "loss": 3.8644, "step": 3507 }, { "epoch": 0.67, "learning_rate": 2.8022116301239278e-06, "loss": 3.5715, "step": 3508 }, { "epoch": 0.67, "learning_rate": 2.8021544327931366e-06, "loss": 2.9631, "step": 3509 }, { "epoch": 0.67, "learning_rate": 2.802097235462345e-06, "loss": 3.1809, "step": 3510 }, { "epoch": 0.67, "learning_rate": 2.802040038131554e-06, "loss": 3.1472, "step": 3511 }, { "epoch": 0.67, "learning_rate": 2.8019828408007625e-06, "loss": 3.3672, "step": 3512 }, { "epoch": 0.67, "learning_rate": 2.8019256434699713e-06, "loss": 3.7447, "step": 3513 }, { "epoch": 0.67, "learning_rate": 2.8018684461391805e-06, "loss": 3.3936, "step": 3514 }, { "epoch": 0.67, "learning_rate": 2.801811248808389e-06, "loss": 3.4611, "step": 3515 }, { "epoch": 0.67, "learning_rate": 2.8017540514775977e-06, "loss": 3.4497, "step": 3516 }, { "epoch": 0.67, "learning_rate": 2.8016968541468065e-06, "loss": 3.3317, "step": 3517 }, { "epoch": 0.67, "learning_rate": 2.8016396568160153e-06, "loss": 3.9405, "step": 3518 }, { "epoch": 0.67, "learning_rate": 2.801582459485224e-06, "loss": 3.7021, "step": 3519 }, { "epoch": 0.67, "learning_rate": 2.801525262154433e-06, "loss": 3.7092, "step": 3520 }, { "epoch": 0.67, "learning_rate": 2.8014680648236417e-06, "loss": 3.3994, "step": 3521 }, { "epoch": 0.67, "learning_rate": 2.80141086749285e-06, "loss": 3.6871, "step": 3522 }, { "epoch": 0.67, "learning_rate": 2.8013536701620593e-06, "loss": 3.3851, "step": 3523 }, { "epoch": 0.67, "learning_rate": 2.801296472831268e-06, "loss": 3.6829, "step": 3524 }, { "epoch": 0.67, "learning_rate": 2.801239275500477e-06, "loss": 3.5984, "step": 3525 }, { "epoch": 0.67, "learning_rate": 2.8011820781696856e-06, "loss": 3.5954, "step": 3526 }, { "epoch": 0.67, "learning_rate": 2.801124880838894e-06, "loss": 3.3921, "step": 3527 }, { "epoch": 0.67, "learning_rate": 2.8010676835081032e-06, "loss": 3.5304, "step": 3528 }, { "epoch": 0.67, "learning_rate": 2.8010104861773116e-06, "loss": 3.4191, "step": 3529 }, { "epoch": 0.67, "learning_rate": 2.8009532888465204e-06, "loss": 3.3768, "step": 3530 }, { "epoch": 0.67, "learning_rate": 2.8008960915157296e-06, "loss": 3.6752, "step": 3531 }, { "epoch": 0.67, "learning_rate": 2.800838894184938e-06, "loss": 3.157, "step": 3532 }, { "epoch": 0.67, "learning_rate": 2.8007816968541468e-06, "loss": 3.6025, "step": 3533 }, { "epoch": 0.67, "learning_rate": 2.8007244995233556e-06, "loss": 3.1072, "step": 3534 }, { "epoch": 0.67, "learning_rate": 2.8006673021925644e-06, "loss": 3.491, "step": 3535 }, { "epoch": 0.67, "learning_rate": 2.800610104861773e-06, "loss": 3.9296, "step": 3536 }, { "epoch": 0.67, "learning_rate": 2.800552907530982e-06, "loss": 3.6399, "step": 3537 }, { "epoch": 0.67, "learning_rate": 2.8004957102001908e-06, "loss": 3.5823, "step": 3538 }, { "epoch": 0.67, "learning_rate": 2.8004385128693995e-06, "loss": 3.3105, "step": 3539 }, { "epoch": 0.67, "learning_rate": 2.8003813155386083e-06, "loss": 3.4954, "step": 3540 }, { "epoch": 0.67, "learning_rate": 2.800324118207817e-06, "loss": 3.1912, "step": 3541 }, { "epoch": 0.67, "learning_rate": 2.800266920877026e-06, "loss": 3.6829, "step": 3542 }, { "epoch": 0.67, "learning_rate": 2.8002097235462347e-06, "loss": 3.6281, "step": 3543 }, { "epoch": 0.68, "learning_rate": 2.800152526215443e-06, "loss": 3.5842, "step": 3544 }, { "epoch": 0.68, "learning_rate": 2.8000953288846523e-06, "loss": 3.6731, "step": 3545 }, { "epoch": 0.68, "learning_rate": 2.800038131553861e-06, "loss": 3.4791, "step": 3546 }, { "epoch": 0.68, "learning_rate": 2.7999809342230695e-06, "loss": 3.3192, "step": 3547 }, { "epoch": 0.68, "learning_rate": 2.7999237368922787e-06, "loss": 3.3894, "step": 3548 }, { "epoch": 0.68, "learning_rate": 2.799866539561487e-06, "loss": 2.9896, "step": 3549 }, { "epoch": 0.68, "learning_rate": 2.799809342230696e-06, "loss": 3.6098, "step": 3550 }, { "epoch": 0.68, "learning_rate": 2.7997521448999046e-06, "loss": 3.5835, "step": 3551 }, { "epoch": 0.68, "learning_rate": 2.7996949475691134e-06, "loss": 3.5694, "step": 3552 }, { "epoch": 0.68, "learning_rate": 2.7996377502383222e-06, "loss": 3.4163, "step": 3553 }, { "epoch": 0.68, "learning_rate": 2.799580552907531e-06, "loss": 3.5561, "step": 3554 }, { "epoch": 0.68, "learning_rate": 2.79952335557674e-06, "loss": 3.6344, "step": 3555 }, { "epoch": 0.68, "learning_rate": 2.7994661582459486e-06, "loss": 3.5212, "step": 3556 }, { "epoch": 0.68, "learning_rate": 2.7994089609151574e-06, "loss": 3.5602, "step": 3557 }, { "epoch": 0.68, "learning_rate": 2.799351763584366e-06, "loss": 3.8107, "step": 3558 }, { "epoch": 0.68, "learning_rate": 2.799294566253575e-06, "loss": 3.7388, "step": 3559 }, { "epoch": 0.68, "learning_rate": 2.799237368922784e-06, "loss": 3.5198, "step": 3560 }, { "epoch": 0.68, "learning_rate": 2.799180171591992e-06, "loss": 3.6959, "step": 3561 }, { "epoch": 0.68, "learning_rate": 2.7991229742612014e-06, "loss": 3.791, "step": 3562 }, { "epoch": 0.68, "learning_rate": 2.79906577693041e-06, "loss": 3.7461, "step": 3563 }, { "epoch": 0.68, "learning_rate": 2.7990085795996185e-06, "loss": 3.8297, "step": 3564 }, { "epoch": 0.68, "learning_rate": 2.7989513822688278e-06, "loss": 3.8113, "step": 3565 }, { "epoch": 0.68, "learning_rate": 2.798894184938036e-06, "loss": 3.4966, "step": 3566 }, { "epoch": 0.68, "learning_rate": 2.798836987607245e-06, "loss": 3.8698, "step": 3567 }, { "epoch": 0.68, "learning_rate": 2.7987797902764537e-06, "loss": 3.5936, "step": 3568 }, { "epoch": 0.68, "learning_rate": 2.7987225929456625e-06, "loss": 3.4146, "step": 3569 }, { "epoch": 0.68, "learning_rate": 2.7986653956148713e-06, "loss": 3.5837, "step": 3570 }, { "epoch": 0.68, "learning_rate": 2.79860819828408e-06, "loss": 3.5, "step": 3571 }, { "epoch": 0.68, "learning_rate": 2.798551000953289e-06, "loss": 3.3965, "step": 3572 }, { "epoch": 0.68, "learning_rate": 2.7984938036224977e-06, "loss": 3.5885, "step": 3573 }, { "epoch": 0.68, "learning_rate": 2.7984366062917065e-06, "loss": 3.4769, "step": 3574 }, { "epoch": 0.68, "learning_rate": 2.7983794089609153e-06, "loss": 3.8061, "step": 3575 }, { "epoch": 0.68, "learning_rate": 2.798322211630124e-06, "loss": 3.3403, "step": 3576 }, { "epoch": 0.68, "learning_rate": 2.798265014299333e-06, "loss": 3.5439, "step": 3577 }, { "epoch": 0.68, "learning_rate": 2.7982078169685412e-06, "loss": 3.4294, "step": 3578 }, { "epoch": 0.68, "learning_rate": 2.7981506196377504e-06, "loss": 3.5948, "step": 3579 }, { "epoch": 0.68, "learning_rate": 2.7980934223069592e-06, "loss": 3.5023, "step": 3580 }, { "epoch": 0.68, "learning_rate": 2.7980362249761676e-06, "loss": 3.3569, "step": 3581 }, { "epoch": 0.68, "learning_rate": 2.797979027645377e-06, "loss": 3.4222, "step": 3582 }, { "epoch": 0.68, "learning_rate": 2.797921830314585e-06, "loss": 3.9999, "step": 3583 }, { "epoch": 0.68, "learning_rate": 2.797864632983794e-06, "loss": 3.5853, "step": 3584 }, { "epoch": 0.68, "learning_rate": 2.7978074356530032e-06, "loss": 3.778, "step": 3585 }, { "epoch": 0.68, "learning_rate": 2.7977502383222116e-06, "loss": 3.4459, "step": 3586 }, { "epoch": 0.68, "learning_rate": 2.7976930409914204e-06, "loss": 3.3282, "step": 3587 }, { "epoch": 0.68, "learning_rate": 2.797635843660629e-06, "loss": 3.4921, "step": 3588 }, { "epoch": 0.68, "learning_rate": 2.797578646329838e-06, "loss": 3.5259, "step": 3589 }, { "epoch": 0.68, "learning_rate": 2.7975214489990468e-06, "loss": 3.6396, "step": 3590 }, { "epoch": 0.68, "learning_rate": 2.7974642516682556e-06, "loss": 3.6469, "step": 3591 }, { "epoch": 0.68, "learning_rate": 2.7974070543374643e-06, "loss": 3.462, "step": 3592 }, { "epoch": 0.68, "learning_rate": 2.797349857006673e-06, "loss": 3.3884, "step": 3593 }, { "epoch": 0.68, "learning_rate": 2.797292659675882e-06, "loss": 3.6032, "step": 3594 }, { "epoch": 0.68, "learning_rate": 2.7972354623450907e-06, "loss": 3.4527, "step": 3595 }, { "epoch": 0.68, "learning_rate": 2.7971782650142995e-06, "loss": 3.9802, "step": 3596 }, { "epoch": 0.69, "learning_rate": 2.7971210676835083e-06, "loss": 3.5127, "step": 3597 }, { "epoch": 0.69, "learning_rate": 2.7970638703527167e-06, "loss": 3.511, "step": 3598 }, { "epoch": 0.69, "learning_rate": 2.797006673021926e-06, "loss": 3.7208, "step": 3599 }, { "epoch": 0.69, "learning_rate": 2.7969494756911343e-06, "loss": 3.389, "step": 3600 }, { "epoch": 0.69, "learning_rate": 2.796892278360343e-06, "loss": 3.7796, "step": 3601 }, { "epoch": 0.69, "learning_rate": 2.7968350810295523e-06, "loss": 3.4229, "step": 3602 }, { "epoch": 0.69, "learning_rate": 2.7967778836987607e-06, "loss": 3.5324, "step": 3603 }, { "epoch": 0.69, "learning_rate": 2.7967206863679694e-06, "loss": 3.2526, "step": 3604 }, { "epoch": 0.69, "learning_rate": 2.7966634890371782e-06, "loss": 3.5923, "step": 3605 }, { "epoch": 0.69, "learning_rate": 2.796606291706387e-06, "loss": 3.7205, "step": 3606 }, { "epoch": 0.69, "learning_rate": 2.796549094375596e-06, "loss": 3.9009, "step": 3607 }, { "epoch": 0.69, "learning_rate": 2.7964918970448046e-06, "loss": 3.2715, "step": 3608 }, { "epoch": 0.69, "learning_rate": 2.7964346997140134e-06, "loss": 3.2143, "step": 3609 }, { "epoch": 0.69, "learning_rate": 2.796377502383222e-06, "loss": 3.4973, "step": 3610 }, { "epoch": 0.69, "learning_rate": 2.796320305052431e-06, "loss": 3.4271, "step": 3611 }, { "epoch": 0.69, "learning_rate": 2.79626310772164e-06, "loss": 3.687, "step": 3612 }, { "epoch": 0.69, "learning_rate": 2.7962059103908486e-06, "loss": 3.4526, "step": 3613 }, { "epoch": 0.69, "learning_rate": 2.7961487130600574e-06, "loss": 3.166, "step": 3614 }, { "epoch": 0.69, "learning_rate": 2.7960915157292658e-06, "loss": 3.4081, "step": 3615 }, { "epoch": 0.69, "learning_rate": 2.796034318398475e-06, "loss": 3.4401, "step": 3616 }, { "epoch": 0.69, "learning_rate": 2.7959771210676833e-06, "loss": 3.6029, "step": 3617 }, { "epoch": 0.69, "learning_rate": 2.795919923736892e-06, "loss": 3.7242, "step": 3618 }, { "epoch": 0.69, "learning_rate": 2.7958627264061014e-06, "loss": 3.2707, "step": 3619 }, { "epoch": 0.69, "learning_rate": 2.7958055290753097e-06, "loss": 3.4738, "step": 3620 }, { "epoch": 0.69, "learning_rate": 2.795748331744519e-06, "loss": 3.352, "step": 3621 }, { "epoch": 0.69, "learning_rate": 2.7956911344137273e-06, "loss": 3.3508, "step": 3622 }, { "epoch": 0.69, "learning_rate": 2.795633937082936e-06, "loss": 3.7358, "step": 3623 }, { "epoch": 0.69, "learning_rate": 2.7955767397521453e-06, "loss": 3.0918, "step": 3624 }, { "epoch": 0.69, "learning_rate": 2.7955195424213537e-06, "loss": 3.4008, "step": 3625 }, { "epoch": 0.69, "learning_rate": 2.7954623450905625e-06, "loss": 3.0459, "step": 3626 }, { "epoch": 0.69, "learning_rate": 2.7954051477597713e-06, "loss": 3.5782, "step": 3627 }, { "epoch": 0.69, "learning_rate": 2.79534795042898e-06, "loss": 3.4871, "step": 3628 }, { "epoch": 0.69, "learning_rate": 2.795290753098189e-06, "loss": 3.3895, "step": 3629 }, { "epoch": 0.69, "learning_rate": 2.7952335557673977e-06, "loss": 3.7419, "step": 3630 }, { "epoch": 0.69, "learning_rate": 2.7951763584366065e-06, "loss": 3.704, "step": 3631 }, { "epoch": 0.69, "learning_rate": 2.795119161105815e-06, "loss": 3.5664, "step": 3632 }, { "epoch": 0.69, "learning_rate": 2.795061963775024e-06, "loss": 3.537, "step": 3633 }, { "epoch": 0.69, "learning_rate": 2.795004766444233e-06, "loss": 2.8308, "step": 3634 }, { "epoch": 0.69, "learning_rate": 2.794947569113441e-06, "loss": 3.4149, "step": 3635 }, { "epoch": 0.69, "learning_rate": 2.7948903717826504e-06, "loss": 3.0002, "step": 3636 }, { "epoch": 0.69, "learning_rate": 2.794833174451859e-06, "loss": 3.5877, "step": 3637 }, { "epoch": 0.69, "learning_rate": 2.794775977121068e-06, "loss": 3.1835, "step": 3638 }, { "epoch": 0.69, "learning_rate": 2.7947187797902764e-06, "loss": 3.5784, "step": 3639 }, { "epoch": 0.69, "learning_rate": 2.794661582459485e-06, "loss": 3.3704, "step": 3640 }, { "epoch": 0.69, "learning_rate": 2.7946043851286944e-06, "loss": 3.3537, "step": 3641 }, { "epoch": 0.69, "learning_rate": 2.7945471877979028e-06, "loss": 3.6042, "step": 3642 }, { "epoch": 0.69, "learning_rate": 2.7944899904671116e-06, "loss": 3.5932, "step": 3643 }, { "epoch": 0.69, "learning_rate": 2.7944327931363204e-06, "loss": 3.6949, "step": 3644 }, { "epoch": 0.69, "learning_rate": 2.794375595805529e-06, "loss": 3.4433, "step": 3645 }, { "epoch": 0.69, "learning_rate": 2.794318398474738e-06, "loss": 3.2788, "step": 3646 }, { "epoch": 0.69, "learning_rate": 2.7942612011439467e-06, "loss": 3.785, "step": 3647 }, { "epoch": 0.69, "learning_rate": 2.7942040038131555e-06, "loss": 3.402, "step": 3648 }, { "epoch": 0.7, "learning_rate": 2.794146806482364e-06, "loss": 3.4377, "step": 3649 }, { "epoch": 0.7, "learning_rate": 2.794089609151573e-06, "loss": 3.0721, "step": 3650 }, { "epoch": 0.7, "learning_rate": 2.794032411820782e-06, "loss": 3.7126, "step": 3651 }, { "epoch": 0.7, "learning_rate": 2.7939752144899903e-06, "loss": 3.3803, "step": 3652 }, { "epoch": 0.7, "learning_rate": 2.7939180171591995e-06, "loss": 3.5536, "step": 3653 }, { "epoch": 0.7, "learning_rate": 2.793860819828408e-06, "loss": 3.313, "step": 3654 }, { "epoch": 0.7, "learning_rate": 2.793803622497617e-06, "loss": 3.7445, "step": 3655 }, { "epoch": 0.7, "learning_rate": 2.7937464251668255e-06, "loss": 3.8751, "step": 3656 }, { "epoch": 0.7, "learning_rate": 2.7936892278360343e-06, "loss": 3.3498, "step": 3657 }, { "epoch": 0.7, "learning_rate": 2.7936320305052435e-06, "loss": 3.3796, "step": 3658 }, { "epoch": 0.7, "learning_rate": 2.793574833174452e-06, "loss": 3.3946, "step": 3659 }, { "epoch": 0.7, "learning_rate": 2.7935176358436606e-06, "loss": 3.6998, "step": 3660 }, { "epoch": 0.7, "learning_rate": 2.7934604385128694e-06, "loss": 3.5396, "step": 3661 }, { "epoch": 0.7, "learning_rate": 2.7934032411820782e-06, "loss": 3.534, "step": 3662 }, { "epoch": 0.7, "learning_rate": 2.793346043851287e-06, "loss": 3.4426, "step": 3663 }, { "epoch": 0.7, "learning_rate": 2.793288846520496e-06, "loss": 3.7483, "step": 3664 }, { "epoch": 0.7, "learning_rate": 2.7932316491897046e-06, "loss": 3.5594, "step": 3665 }, { "epoch": 0.7, "learning_rate": 2.793174451858913e-06, "loss": 3.4816, "step": 3666 }, { "epoch": 0.7, "learning_rate": 2.793117254528122e-06, "loss": 3.4899, "step": 3667 }, { "epoch": 0.7, "learning_rate": 2.793060057197331e-06, "loss": 3.5462, "step": 3668 }, { "epoch": 0.7, "learning_rate": 2.7930028598665394e-06, "loss": 3.1682, "step": 3669 }, { "epoch": 0.7, "learning_rate": 2.7929456625357486e-06, "loss": 3.6779, "step": 3670 }, { "epoch": 0.7, "learning_rate": 2.792888465204957e-06, "loss": 3.6838, "step": 3671 }, { "epoch": 0.7, "learning_rate": 2.792831267874166e-06, "loss": 3.3863, "step": 3672 }, { "epoch": 0.7, "learning_rate": 2.792774070543375e-06, "loss": 3.3854, "step": 3673 }, { "epoch": 0.7, "learning_rate": 2.7927168732125833e-06, "loss": 3.559, "step": 3674 }, { "epoch": 0.7, "learning_rate": 2.7926596758817925e-06, "loss": 3.7433, "step": 3675 }, { "epoch": 0.7, "learning_rate": 2.792602478551001e-06, "loss": 3.3645, "step": 3676 }, { "epoch": 0.7, "learning_rate": 2.7925452812202097e-06, "loss": 3.5441, "step": 3677 }, { "epoch": 0.7, "learning_rate": 2.7924880838894185e-06, "loss": 3.5439, "step": 3678 }, { "epoch": 0.7, "learning_rate": 2.7924308865586273e-06, "loss": 3.712, "step": 3679 }, { "epoch": 0.7, "learning_rate": 2.792373689227836e-06, "loss": 3.5625, "step": 3680 }, { "epoch": 0.7, "learning_rate": 2.792316491897045e-06, "loss": 3.5673, "step": 3681 }, { "epoch": 0.7, "learning_rate": 2.7922592945662537e-06, "loss": 3.5375, "step": 3682 }, { "epoch": 0.7, "learning_rate": 2.7922020972354625e-06, "loss": 3.1952, "step": 3683 }, { "epoch": 0.7, "learning_rate": 2.7921448999046713e-06, "loss": 3.3652, "step": 3684 }, { "epoch": 0.7, "learning_rate": 2.79208770257388e-06, "loss": 3.3634, "step": 3685 }, { "epoch": 0.7, "learning_rate": 2.7920305052430884e-06, "loss": 3.5532, "step": 3686 }, { "epoch": 0.7, "learning_rate": 2.7919733079122976e-06, "loss": 3.4951, "step": 3687 }, { "epoch": 0.7, "learning_rate": 2.791916110581506e-06, "loss": 3.0523, "step": 3688 }, { "epoch": 0.7, "learning_rate": 2.7918589132507152e-06, "loss": 3.8381, "step": 3689 }, { "epoch": 0.7, "learning_rate": 2.791801715919924e-06, "loss": 3.551, "step": 3690 }, { "epoch": 0.7, "learning_rate": 2.7917445185891324e-06, "loss": 3.7148, "step": 3691 }, { "epoch": 0.7, "learning_rate": 2.7916873212583416e-06, "loss": 3.2464, "step": 3692 }, { "epoch": 0.7, "learning_rate": 2.79163012392755e-06, "loss": 3.6906, "step": 3693 }, { "epoch": 0.7, "learning_rate": 2.7915729265967588e-06, "loss": 3.7112, "step": 3694 }, { "epoch": 0.7, "learning_rate": 2.7915157292659676e-06, "loss": 3.5308, "step": 3695 }, { "epoch": 0.7, "learning_rate": 2.7914585319351764e-06, "loss": 3.3734, "step": 3696 }, { "epoch": 0.7, "learning_rate": 2.791401334604385e-06, "loss": 3.706, "step": 3697 }, { "epoch": 0.7, "learning_rate": 2.791344137273594e-06, "loss": 3.2804, "step": 3698 }, { "epoch": 0.7, "learning_rate": 2.7912869399428027e-06, "loss": 3.4009, "step": 3699 }, { "epoch": 0.7, "learning_rate": 2.7912297426120115e-06, "loss": 3.6278, "step": 3700 }, { "epoch": 0.7, "learning_rate": 2.7911725452812203e-06, "loss": 3.276, "step": 3701 }, { "epoch": 0.71, "learning_rate": 2.791115347950429e-06, "loss": 3.4755, "step": 3702 }, { "epoch": 0.71, "learning_rate": 2.791058150619638e-06, "loss": 3.7302, "step": 3703 }, { "epoch": 0.71, "learning_rate": 2.7910009532888467e-06, "loss": 3.6031, "step": 3704 }, { "epoch": 0.71, "learning_rate": 2.790943755958055e-06, "loss": 3.7277, "step": 3705 }, { "epoch": 0.71, "learning_rate": 2.7908865586272643e-06, "loss": 3.4903, "step": 3706 }, { "epoch": 0.71, "learning_rate": 2.790829361296473e-06, "loss": 3.4334, "step": 3707 }, { "epoch": 0.71, "learning_rate": 2.7907721639656815e-06, "loss": 3.3699, "step": 3708 }, { "epoch": 0.71, "learning_rate": 2.7907149666348907e-06, "loss": 3.2789, "step": 3709 }, { "epoch": 0.71, "learning_rate": 2.790657769304099e-06, "loss": 3.6724, "step": 3710 }, { "epoch": 0.71, "learning_rate": 2.790600571973308e-06, "loss": 3.3869, "step": 3711 }, { "epoch": 0.71, "learning_rate": 2.790543374642517e-06, "loss": 3.7845, "step": 3712 }, { "epoch": 0.71, "learning_rate": 2.7904861773117254e-06, "loss": 3.266, "step": 3713 }, { "epoch": 0.71, "learning_rate": 2.7904289799809342e-06, "loss": 3.3196, "step": 3714 }, { "epoch": 0.71, "learning_rate": 2.790371782650143e-06, "loss": 3.3768, "step": 3715 }, { "epoch": 0.71, "learning_rate": 2.790314585319352e-06, "loss": 3.3891, "step": 3716 }, { "epoch": 0.71, "learning_rate": 2.7902573879885606e-06, "loss": 3.7615, "step": 3717 }, { "epoch": 0.71, "learning_rate": 2.7902001906577694e-06, "loss": 3.6393, "step": 3718 }, { "epoch": 0.71, "learning_rate": 2.790142993326978e-06, "loss": 3.6784, "step": 3719 }, { "epoch": 0.71, "learning_rate": 2.790085795996187e-06, "loss": 3.4935, "step": 3720 }, { "epoch": 0.71, "learning_rate": 2.7900285986653958e-06, "loss": 3.5599, "step": 3721 }, { "epoch": 0.71, "learning_rate": 2.7899714013346046e-06, "loss": 3.6096, "step": 3722 }, { "epoch": 0.71, "learning_rate": 2.7899142040038134e-06, "loss": 3.4831, "step": 3723 }, { "epoch": 0.71, "learning_rate": 2.789857006673022e-06, "loss": 3.4481, "step": 3724 }, { "epoch": 0.71, "learning_rate": 2.7897998093422305e-06, "loss": 3.8846, "step": 3725 }, { "epoch": 0.71, "learning_rate": 2.7897426120114398e-06, "loss": 3.5498, "step": 3726 }, { "epoch": 0.71, "learning_rate": 2.789685414680648e-06, "loss": 3.0998, "step": 3727 }, { "epoch": 0.71, "learning_rate": 2.789628217349857e-06, "loss": 3.263, "step": 3728 }, { "epoch": 0.71, "learning_rate": 2.789571020019066e-06, "loss": 3.1444, "step": 3729 }, { "epoch": 0.71, "learning_rate": 2.7895138226882745e-06, "loss": 3.5965, "step": 3730 }, { "epoch": 0.71, "learning_rate": 2.7894566253574833e-06, "loss": 3.0458, "step": 3731 }, { "epoch": 0.71, "learning_rate": 2.789399428026692e-06, "loss": 3.3492, "step": 3732 }, { "epoch": 0.71, "learning_rate": 2.789342230695901e-06, "loss": 3.7311, "step": 3733 }, { "epoch": 0.71, "learning_rate": 2.7892850333651097e-06, "loss": 3.2909, "step": 3734 }, { "epoch": 0.71, "learning_rate": 2.7892278360343185e-06, "loss": 3.4838, "step": 3735 }, { "epoch": 0.71, "learning_rate": 2.7891706387035273e-06, "loss": 3.4536, "step": 3736 }, { "epoch": 0.71, "learning_rate": 2.789113441372736e-06, "loss": 3.007, "step": 3737 }, { "epoch": 0.71, "learning_rate": 2.789056244041945e-06, "loss": 3.6493, "step": 3738 }, { "epoch": 0.71, "learning_rate": 2.7889990467111536e-06, "loss": 3.7372, "step": 3739 }, { "epoch": 0.71, "learning_rate": 2.7889418493803624e-06, "loss": 3.2687, "step": 3740 }, { "epoch": 0.71, "learning_rate": 2.7888846520495712e-06, "loss": 3.5188, "step": 3741 }, { "epoch": 0.71, "learning_rate": 2.7888274547187796e-06, "loss": 3.2772, "step": 3742 }, { "epoch": 0.71, "learning_rate": 2.788770257387989e-06, "loss": 3.8071, "step": 3743 }, { "epoch": 0.71, "learning_rate": 2.788713060057197e-06, "loss": 3.4704, "step": 3744 }, { "epoch": 0.71, "learning_rate": 2.788655862726406e-06, "loss": 3.3606, "step": 3745 }, { "epoch": 0.71, "learning_rate": 2.788598665395615e-06, "loss": 3.4773, "step": 3746 }, { "epoch": 0.71, "learning_rate": 2.7885414680648236e-06, "loss": 3.3606, "step": 3747 }, { "epoch": 0.71, "learning_rate": 2.7884842707340324e-06, "loss": 3.4076, "step": 3748 }, { "epoch": 0.71, "learning_rate": 2.788427073403241e-06, "loss": 3.6509, "step": 3749 }, { "epoch": 0.71, "learning_rate": 2.78836987607245e-06, "loss": 3.538, "step": 3750 }, { "epoch": 0.71, "learning_rate": 2.7883126787416588e-06, "loss": 3.5428, "step": 3751 }, { "epoch": 0.71, "learning_rate": 2.7882554814108675e-06, "loss": 3.6864, "step": 3752 }, { "epoch": 0.71, "learning_rate": 2.7881982840800763e-06, "loss": 3.5396, "step": 3753 }, { "epoch": 0.72, "learning_rate": 2.788141086749285e-06, "loss": 3.4844, "step": 3754 }, { "epoch": 0.72, "learning_rate": 2.788083889418494e-06, "loss": 3.1347, "step": 3755 }, { "epoch": 0.72, "learning_rate": 2.7880266920877027e-06, "loss": 3.4677, "step": 3756 }, { "epoch": 0.72, "learning_rate": 2.7879694947569115e-06, "loss": 3.6269, "step": 3757 }, { "epoch": 0.72, "learning_rate": 2.7879122974261203e-06, "loss": 3.5295, "step": 3758 }, { "epoch": 0.72, "learning_rate": 2.7878551000953287e-06, "loss": 3.6984, "step": 3759 }, { "epoch": 0.72, "learning_rate": 2.787797902764538e-06, "loss": 3.2275, "step": 3760 }, { "epoch": 0.72, "learning_rate": 2.7877407054337467e-06, "loss": 3.3041, "step": 3761 }, { "epoch": 0.72, "learning_rate": 2.787683508102955e-06, "loss": 3.2511, "step": 3762 }, { "epoch": 0.72, "learning_rate": 2.7876263107721643e-06, "loss": 3.3563, "step": 3763 }, { "epoch": 0.72, "learning_rate": 2.7875691134413726e-06, "loss": 3.5001, "step": 3764 }, { "epoch": 0.72, "learning_rate": 2.7875119161105814e-06, "loss": 3.4457, "step": 3765 }, { "epoch": 0.72, "learning_rate": 2.7874547187797902e-06, "loss": 3.1735, "step": 3766 }, { "epoch": 0.72, "learning_rate": 2.787397521448999e-06, "loss": 3.5428, "step": 3767 }, { "epoch": 0.72, "learning_rate": 2.7873403241182082e-06, "loss": 3.629, "step": 3768 }, { "epoch": 0.72, "learning_rate": 2.7872831267874166e-06, "loss": 3.5402, "step": 3769 }, { "epoch": 0.72, "learning_rate": 2.7872259294566254e-06, "loss": 3.1965, "step": 3770 }, { "epoch": 0.72, "learning_rate": 2.787168732125834e-06, "loss": 3.8736, "step": 3771 }, { "epoch": 0.72, "learning_rate": 2.787111534795043e-06, "loss": 3.602, "step": 3772 }, { "epoch": 0.72, "learning_rate": 2.787054337464252e-06, "loss": 3.1527, "step": 3773 }, { "epoch": 0.72, "learning_rate": 2.7869971401334606e-06, "loss": 3.6022, "step": 3774 }, { "epoch": 0.72, "learning_rate": 2.7869399428026694e-06, "loss": 3.4477, "step": 3775 }, { "epoch": 0.72, "learning_rate": 2.7868827454718777e-06, "loss": 3.6772, "step": 3776 }, { "epoch": 0.72, "learning_rate": 2.786825548141087e-06, "loss": 3.2591, "step": 3777 }, { "epoch": 0.72, "learning_rate": 2.7867683508102958e-06, "loss": 3.1967, "step": 3778 }, { "epoch": 0.72, "learning_rate": 2.786711153479504e-06, "loss": 3.6941, "step": 3779 }, { "epoch": 0.72, "learning_rate": 2.7866539561487133e-06, "loss": 3.514, "step": 3780 }, { "epoch": 0.72, "learning_rate": 2.7865967588179217e-06, "loss": 3.3179, "step": 3781 }, { "epoch": 0.72, "learning_rate": 2.7865395614871305e-06, "loss": 3.5342, "step": 3782 }, { "epoch": 0.72, "learning_rate": 2.7864823641563393e-06, "loss": 3.5433, "step": 3783 }, { "epoch": 0.72, "learning_rate": 2.786425166825548e-06, "loss": 3.6336, "step": 3784 }, { "epoch": 0.72, "learning_rate": 2.7863679694947573e-06, "loss": 3.9006, "step": 3785 }, { "epoch": 0.72, "learning_rate": 2.7863107721639657e-06, "loss": 3.59, "step": 3786 }, { "epoch": 0.72, "learning_rate": 2.7862535748331745e-06, "loss": 3.2729, "step": 3787 }, { "epoch": 0.72, "learning_rate": 2.7861963775023833e-06, "loss": 3.6534, "step": 3788 }, { "epoch": 0.72, "learning_rate": 2.786139180171592e-06, "loss": 3.4776, "step": 3789 }, { "epoch": 0.72, "learning_rate": 2.786081982840801e-06, "loss": 3.6729, "step": 3790 }, { "epoch": 0.72, "learning_rate": 2.7860247855100097e-06, "loss": 3.5526, "step": 3791 }, { "epoch": 0.72, "learning_rate": 2.7859675881792185e-06, "loss": 3.6378, "step": 3792 }, { "epoch": 0.72, "learning_rate": 2.785910390848427e-06, "loss": 3.2351, "step": 3793 }, { "epoch": 0.72, "learning_rate": 2.785853193517636e-06, "loss": 3.1931, "step": 3794 }, { "epoch": 0.72, "learning_rate": 2.785795996186845e-06, "loss": 3.3486, "step": 3795 }, { "epoch": 0.72, "learning_rate": 2.785738798856053e-06, "loss": 3.2893, "step": 3796 }, { "epoch": 0.72, "learning_rate": 2.7856816015252624e-06, "loss": 3.5454, "step": 3797 }, { "epoch": 0.72, "learning_rate": 2.785624404194471e-06, "loss": 3.5128, "step": 3798 }, { "epoch": 0.72, "learning_rate": 2.7855672068636796e-06, "loss": 3.414, "step": 3799 }, { "epoch": 0.72, "learning_rate": 2.785510009532889e-06, "loss": 3.4889, "step": 3800 }, { "epoch": 0.72, "learning_rate": 2.785452812202097e-06, "loss": 3.5037, "step": 3801 }, { "epoch": 0.72, "learning_rate": 2.7853956148713064e-06, "loss": 3.5691, "step": 3802 }, { "epoch": 0.72, "learning_rate": 2.7853384175405148e-06, "loss": 3.6018, "step": 3803 }, { "epoch": 0.72, "learning_rate": 2.7852812202097236e-06, "loss": 3.5748, "step": 3804 }, { "epoch": 0.72, "learning_rate": 2.7852240228789323e-06, "loss": 3.244, "step": 3805 }, { "epoch": 0.72, "learning_rate": 2.785166825548141e-06, "loss": 3.3661, "step": 3806 }, { "epoch": 0.73, "learning_rate": 2.78510962821735e-06, "loss": 3.4138, "step": 3807 }, { "epoch": 0.73, "learning_rate": 2.7850524308865587e-06, "loss": 3.8293, "step": 3808 }, { "epoch": 0.73, "learning_rate": 2.7849952335557675e-06, "loss": 3.5723, "step": 3809 }, { "epoch": 0.73, "learning_rate": 2.7849380362249763e-06, "loss": 3.8666, "step": 3810 }, { "epoch": 0.73, "learning_rate": 2.784880838894185e-06, "loss": 3.2505, "step": 3811 }, { "epoch": 0.73, "learning_rate": 2.784823641563394e-06, "loss": 3.6316, "step": 3812 }, { "epoch": 0.73, "learning_rate": 2.7847664442326023e-06, "loss": 3.8174, "step": 3813 }, { "epoch": 0.73, "learning_rate": 2.7847092469018115e-06, "loss": 3.556, "step": 3814 }, { "epoch": 0.73, "learning_rate": 2.78465204957102e-06, "loss": 3.7226, "step": 3815 }, { "epoch": 0.73, "learning_rate": 2.7845948522402287e-06, "loss": 3.2725, "step": 3816 }, { "epoch": 0.73, "learning_rate": 2.784537654909438e-06, "loss": 3.4141, "step": 3817 }, { "epoch": 0.73, "learning_rate": 2.7844804575786462e-06, "loss": 3.6948, "step": 3818 }, { "epoch": 0.73, "learning_rate": 2.7844232602478555e-06, "loss": 3.5084, "step": 3819 }, { "epoch": 0.73, "learning_rate": 2.784366062917064e-06, "loss": 3.3128, "step": 3820 }, { "epoch": 0.73, "learning_rate": 2.7843088655862726e-06, "loss": 3.4629, "step": 3821 }, { "epoch": 0.73, "learning_rate": 2.7842516682554814e-06, "loss": 3.4803, "step": 3822 }, { "epoch": 0.73, "learning_rate": 2.7841944709246902e-06, "loss": 3.5589, "step": 3823 }, { "epoch": 0.73, "learning_rate": 2.784137273593899e-06, "loss": 3.7716, "step": 3824 }, { "epoch": 0.73, "learning_rate": 2.784080076263108e-06, "loss": 3.5055, "step": 3825 }, { "epoch": 0.73, "learning_rate": 2.7840228789323166e-06, "loss": 3.4364, "step": 3826 }, { "epoch": 0.73, "learning_rate": 2.7839656816015254e-06, "loss": 3.4902, "step": 3827 }, { "epoch": 0.73, "learning_rate": 2.783908484270734e-06, "loss": 3.4783, "step": 3828 }, { "epoch": 0.73, "learning_rate": 2.783851286939943e-06, "loss": 3.1724, "step": 3829 }, { "epoch": 0.73, "learning_rate": 2.7837940896091513e-06, "loss": 3.5358, "step": 3830 }, { "epoch": 0.73, "learning_rate": 2.7837368922783606e-06, "loss": 3.4191, "step": 3831 }, { "epoch": 0.73, "learning_rate": 2.783679694947569e-06, "loss": 3.1436, "step": 3832 }, { "epoch": 0.73, "learning_rate": 2.783622497616778e-06, "loss": 3.4514, "step": 3833 }, { "epoch": 0.73, "learning_rate": 2.783565300285987e-06, "loss": 3.1788, "step": 3834 }, { "epoch": 0.73, "learning_rate": 2.7835081029551953e-06, "loss": 3.7969, "step": 3835 }, { "epoch": 0.73, "learning_rate": 2.7834509056244045e-06, "loss": 3.5846, "step": 3836 }, { "epoch": 0.73, "learning_rate": 2.783393708293613e-06, "loss": 3.3894, "step": 3837 }, { "epoch": 0.73, "learning_rate": 2.7833365109628217e-06, "loss": 3.4256, "step": 3838 }, { "epoch": 0.73, "learning_rate": 2.7832793136320305e-06, "loss": 3.813, "step": 3839 }, { "epoch": 0.73, "learning_rate": 2.7832221163012393e-06, "loss": 3.5246, "step": 3840 }, { "epoch": 0.73, "learning_rate": 2.783164918970448e-06, "loss": 3.481, "step": 3841 }, { "epoch": 0.73, "learning_rate": 2.783107721639657e-06, "loss": 3.7657, "step": 3842 }, { "epoch": 0.73, "learning_rate": 2.7830505243088657e-06, "loss": 3.6849, "step": 3843 }, { "epoch": 0.73, "learning_rate": 2.7829933269780745e-06, "loss": 3.8237, "step": 3844 }, { "epoch": 0.73, "learning_rate": 2.7829361296472833e-06, "loss": 3.3774, "step": 3845 }, { "epoch": 0.73, "learning_rate": 2.782878932316492e-06, "loss": 3.3919, "step": 3846 }, { "epoch": 0.73, "learning_rate": 2.7828217349857004e-06, "loss": 3.4198, "step": 3847 }, { "epoch": 0.73, "learning_rate": 2.7827645376549096e-06, "loss": 3.8529, "step": 3848 }, { "epoch": 0.73, "learning_rate": 2.7827073403241184e-06, "loss": 3.5247, "step": 3849 }, { "epoch": 0.73, "learning_rate": 2.7826501429933272e-06, "loss": 3.8097, "step": 3850 }, { "epoch": 0.73, "learning_rate": 2.782592945662536e-06, "loss": 3.5941, "step": 3851 }, { "epoch": 0.73, "learning_rate": 2.7825357483317444e-06, "loss": 3.6959, "step": 3852 }, { "epoch": 0.73, "learning_rate": 2.7824785510009536e-06, "loss": 2.9355, "step": 3853 }, { "epoch": 0.73, "learning_rate": 2.782421353670162e-06, "loss": 3.7359, "step": 3854 }, { "epoch": 0.73, "learning_rate": 2.7823641563393708e-06, "loss": 3.4861, "step": 3855 }, { "epoch": 0.73, "learning_rate": 2.78230695900858e-06, "loss": 3.5417, "step": 3856 }, { "epoch": 0.73, "learning_rate": 2.7822497616777884e-06, "loss": 3.1292, "step": 3857 }, { "epoch": 0.73, "learning_rate": 2.782192564346997e-06, "loss": 3.7881, "step": 3858 }, { "epoch": 0.74, "learning_rate": 2.782135367016206e-06, "loss": 3.6778, "step": 3859 }, { "epoch": 0.74, "learning_rate": 2.7820781696854147e-06, "loss": 3.5951, "step": 3860 }, { "epoch": 0.74, "learning_rate": 2.7820209723546235e-06, "loss": 3.5369, "step": 3861 }, { "epoch": 0.74, "learning_rate": 2.7819637750238323e-06, "loss": 3.5283, "step": 3862 }, { "epoch": 0.74, "learning_rate": 2.781906577693041e-06, "loss": 3.1636, "step": 3863 }, { "epoch": 0.74, "learning_rate": 2.7818493803622495e-06, "loss": 3.4, "step": 3864 }, { "epoch": 0.74, "learning_rate": 2.7817921830314587e-06, "loss": 3.4141, "step": 3865 }, { "epoch": 0.74, "learning_rate": 2.7817349857006675e-06, "loss": 3.6257, "step": 3866 }, { "epoch": 0.74, "learning_rate": 2.7816777883698763e-06, "loss": 3.7393, "step": 3867 }, { "epoch": 0.74, "learning_rate": 2.781620591039085e-06, "loss": 3.4802, "step": 3868 }, { "epoch": 0.74, "learning_rate": 2.7815633937082935e-06, "loss": 3.7251, "step": 3869 }, { "epoch": 0.74, "learning_rate": 2.7815061963775027e-06, "loss": 3.2584, "step": 3870 }, { "epoch": 0.74, "learning_rate": 2.781448999046711e-06, "loss": 3.5825, "step": 3871 }, { "epoch": 0.74, "learning_rate": 2.78139180171592e-06, "loss": 3.7623, "step": 3872 }, { "epoch": 0.74, "learning_rate": 2.781334604385129e-06, "loss": 3.6394, "step": 3873 }, { "epoch": 0.74, "learning_rate": 2.7812774070543374e-06, "loss": 3.3612, "step": 3874 }, { "epoch": 0.74, "learning_rate": 2.7812202097235462e-06, "loss": 3.145, "step": 3875 }, { "epoch": 0.74, "learning_rate": 2.781163012392755e-06, "loss": 3.4741, "step": 3876 }, { "epoch": 0.74, "learning_rate": 2.781105815061964e-06, "loss": 2.9927, "step": 3877 }, { "epoch": 0.74, "learning_rate": 2.7810486177311726e-06, "loss": 3.3965, "step": 3878 }, { "epoch": 0.74, "learning_rate": 2.7809914204003814e-06, "loss": 3.5269, "step": 3879 }, { "epoch": 0.74, "learning_rate": 2.78093422306959e-06, "loss": 3.0664, "step": 3880 }, { "epoch": 0.74, "learning_rate": 2.7808770257387986e-06, "loss": 3.3133, "step": 3881 }, { "epoch": 0.74, "learning_rate": 2.7808198284080078e-06, "loss": 3.2536, "step": 3882 }, { "epoch": 0.74, "learning_rate": 2.7807626310772166e-06, "loss": 3.2805, "step": 3883 }, { "epoch": 0.74, "learning_rate": 2.7807054337464254e-06, "loss": 3.5351, "step": 3884 }, { "epoch": 0.74, "learning_rate": 2.780648236415634e-06, "loss": 3.2929, "step": 3885 }, { "epoch": 0.74, "learning_rate": 2.7805910390848425e-06, "loss": 3.215, "step": 3886 }, { "epoch": 0.74, "learning_rate": 2.7805338417540517e-06, "loss": 3.1973, "step": 3887 }, { "epoch": 0.74, "learning_rate": 2.78047664442326e-06, "loss": 3.5383, "step": 3888 }, { "epoch": 0.74, "learning_rate": 2.780419447092469e-06, "loss": 2.9192, "step": 3889 }, { "epoch": 0.74, "learning_rate": 2.780362249761678e-06, "loss": 3.4867, "step": 3890 }, { "epoch": 0.74, "learning_rate": 2.7803050524308865e-06, "loss": 3.3856, "step": 3891 }, { "epoch": 0.74, "learning_rate": 2.7802478551000953e-06, "loss": 3.3214, "step": 3892 }, { "epoch": 0.74, "learning_rate": 2.780190657769304e-06, "loss": 3.5112, "step": 3893 }, { "epoch": 0.74, "learning_rate": 2.780133460438513e-06, "loss": 3.5674, "step": 3894 }, { "epoch": 0.74, "learning_rate": 2.7800762631077217e-06, "loss": 3.1629, "step": 3895 }, { "epoch": 0.74, "learning_rate": 2.7800190657769305e-06, "loss": 3.3288, "step": 3896 }, { "epoch": 0.74, "learning_rate": 2.7799618684461393e-06, "loss": 3.6524, "step": 3897 }, { "epoch": 0.74, "learning_rate": 2.779904671115348e-06, "loss": 3.6595, "step": 3898 }, { "epoch": 0.74, "learning_rate": 2.779847473784557e-06, "loss": 3.4558, "step": 3899 }, { "epoch": 0.74, "learning_rate": 2.7797902764537656e-06, "loss": 3.7364, "step": 3900 }, { "epoch": 0.74, "learning_rate": 2.7797330791229744e-06, "loss": 3.3115, "step": 3901 }, { "epoch": 0.74, "learning_rate": 2.7796758817921832e-06, "loss": 3.3887, "step": 3902 }, { "epoch": 0.74, "learning_rate": 2.7796186844613916e-06, "loss": 3.4852, "step": 3903 }, { "epoch": 0.74, "learning_rate": 2.779561487130601e-06, "loss": 3.2458, "step": 3904 }, { "epoch": 0.74, "learning_rate": 2.7795042897998096e-06, "loss": 3.6113, "step": 3905 }, { "epoch": 0.74, "learning_rate": 2.779447092469018e-06, "loss": 3.1745, "step": 3906 }, { "epoch": 0.74, "learning_rate": 2.779389895138227e-06, "loss": 3.8336, "step": 3907 }, { "epoch": 0.74, "learning_rate": 2.7793326978074356e-06, "loss": 3.0599, "step": 3908 }, { "epoch": 0.74, "learning_rate": 2.7792755004766444e-06, "loss": 3.7273, "step": 3909 }, { "epoch": 0.74, "learning_rate": 2.779218303145853e-06, "loss": 3.3196, "step": 3910 }, { "epoch": 0.74, "learning_rate": 2.779161105815062e-06, "loss": 3.9969, "step": 3911 }, { "epoch": 0.75, "learning_rate": 2.7791039084842707e-06, "loss": 3.1901, "step": 3912 }, { "epoch": 0.75, "learning_rate": 2.7790467111534795e-06, "loss": 3.3609, "step": 3913 }, { "epoch": 0.75, "learning_rate": 2.7789895138226883e-06, "loss": 3.5538, "step": 3914 }, { "epoch": 0.75, "learning_rate": 2.778932316491897e-06, "loss": 3.5634, "step": 3915 }, { "epoch": 0.75, "learning_rate": 2.778875119161106e-06, "loss": 3.1946, "step": 3916 }, { "epoch": 0.75, "learning_rate": 2.7788179218303147e-06, "loss": 3.2497, "step": 3917 }, { "epoch": 0.75, "learning_rate": 2.7787607244995235e-06, "loss": 3.3312, "step": 3918 }, { "epoch": 0.75, "learning_rate": 2.7787035271687323e-06, "loss": 3.6564, "step": 3919 }, { "epoch": 0.75, "learning_rate": 2.7786463298379407e-06, "loss": 3.0425, "step": 3920 }, { "epoch": 0.75, "learning_rate": 2.77858913250715e-06, "loss": 3.383, "step": 3921 }, { "epoch": 0.75, "learning_rate": 2.7785319351763587e-06, "loss": 3.5748, "step": 3922 }, { "epoch": 0.75, "learning_rate": 2.778474737845567e-06, "loss": 3.983, "step": 3923 }, { "epoch": 0.75, "learning_rate": 2.7784175405147763e-06, "loss": 3.8304, "step": 3924 }, { "epoch": 0.75, "learning_rate": 2.7783603431839846e-06, "loss": 3.2677, "step": 3925 }, { "epoch": 0.75, "learning_rate": 2.7783031458531934e-06, "loss": 3.5976, "step": 3926 }, { "epoch": 0.75, "learning_rate": 2.7782459485224022e-06, "loss": 3.1295, "step": 3927 }, { "epoch": 0.75, "learning_rate": 2.778188751191611e-06, "loss": 3.4508, "step": 3928 }, { "epoch": 0.75, "learning_rate": 2.77813155386082e-06, "loss": 3.9436, "step": 3929 }, { "epoch": 0.75, "learning_rate": 2.7780743565300286e-06, "loss": 3.7816, "step": 3930 }, { "epoch": 0.75, "learning_rate": 2.7780171591992374e-06, "loss": 3.3499, "step": 3931 }, { "epoch": 0.75, "learning_rate": 2.777959961868446e-06, "loss": 3.7748, "step": 3932 }, { "epoch": 0.75, "learning_rate": 2.777902764537655e-06, "loss": 3.4918, "step": 3933 }, { "epoch": 0.75, "learning_rate": 2.7778455672068638e-06, "loss": 3.8379, "step": 3934 }, { "epoch": 0.75, "learning_rate": 2.7777883698760726e-06, "loss": 3.1551, "step": 3935 }, { "epoch": 0.75, "learning_rate": 2.7777311725452814e-06, "loss": 3.7442, "step": 3936 }, { "epoch": 0.75, "learning_rate": 2.77767397521449e-06, "loss": 3.2919, "step": 3937 }, { "epoch": 0.75, "learning_rate": 2.777616777883699e-06, "loss": 3.6796, "step": 3938 }, { "epoch": 0.75, "learning_rate": 2.7775595805529078e-06, "loss": 3.7861, "step": 3939 }, { "epoch": 0.75, "learning_rate": 2.777502383222116e-06, "loss": 3.7862, "step": 3940 }, { "epoch": 0.75, "learning_rate": 2.7774451858913253e-06, "loss": 3.5468, "step": 3941 }, { "epoch": 0.75, "learning_rate": 2.7773879885605337e-06, "loss": 3.8352, "step": 3942 }, { "epoch": 0.75, "learning_rate": 2.7773307912297425e-06, "loss": 3.43, "step": 3943 }, { "epoch": 0.75, "learning_rate": 2.7772735938989517e-06, "loss": 3.4455, "step": 3944 }, { "epoch": 0.75, "learning_rate": 2.77721639656816e-06, "loss": 3.7748, "step": 3945 }, { "epoch": 0.75, "learning_rate": 2.777159199237369e-06, "loss": 3.3993, "step": 3946 }, { "epoch": 0.75, "learning_rate": 2.7771020019065777e-06, "loss": 3.4596, "step": 3947 }, { "epoch": 0.75, "learning_rate": 2.7770448045757865e-06, "loss": 3.5673, "step": 3948 }, { "epoch": 0.75, "learning_rate": 2.7769876072449953e-06, "loss": 3.3252, "step": 3949 }, { "epoch": 0.75, "learning_rate": 2.776930409914204e-06, "loss": 3.0835, "step": 3950 }, { "epoch": 0.75, "learning_rate": 2.776873212583413e-06, "loss": 3.4047, "step": 3951 }, { "epoch": 0.75, "learning_rate": 2.7768160152526217e-06, "loss": 3.6773, "step": 3952 }, { "epoch": 0.75, "learning_rate": 2.7767588179218304e-06, "loss": 3.1461, "step": 3953 }, { "epoch": 0.75, "learning_rate": 2.7767016205910392e-06, "loss": 3.1892, "step": 3954 }, { "epoch": 0.75, "learning_rate": 2.776644423260248e-06, "loss": 3.564, "step": 3955 }, { "epoch": 0.75, "learning_rate": 2.776587225929457e-06, "loss": 3.6625, "step": 3956 }, { "epoch": 0.75, "learning_rate": 2.776530028598665e-06, "loss": 3.4989, "step": 3957 }, { "epoch": 0.75, "learning_rate": 2.7764728312678744e-06, "loss": 3.3154, "step": 3958 }, { "epoch": 0.75, "learning_rate": 2.7764156339370828e-06, "loss": 3.2794, "step": 3959 }, { "epoch": 0.75, "learning_rate": 2.7763584366062916e-06, "loss": 3.3191, "step": 3960 }, { "epoch": 0.75, "learning_rate": 2.776301239275501e-06, "loss": 3.5103, "step": 3961 }, { "epoch": 0.75, "learning_rate": 2.776244041944709e-06, "loss": 3.6917, "step": 3962 }, { "epoch": 0.75, "learning_rate": 2.7761868446139184e-06, "loss": 3.483, "step": 3963 }, { "epoch": 0.76, "learning_rate": 2.7761296472831268e-06, "loss": 3.3245, "step": 3964 }, { "epoch": 0.76, "learning_rate": 2.7760724499523355e-06, "loss": 3.505, "step": 3965 }, { "epoch": 0.76, "learning_rate": 2.7760152526215443e-06, "loss": 3.4867, "step": 3966 }, { "epoch": 0.76, "learning_rate": 2.775958055290753e-06, "loss": 3.3784, "step": 3967 }, { "epoch": 0.76, "learning_rate": 2.775900857959962e-06, "loss": 3.3943, "step": 3968 }, { "epoch": 0.76, "learning_rate": 2.7758436606291707e-06, "loss": 4.0271, "step": 3969 }, { "epoch": 0.76, "learning_rate": 2.7757864632983795e-06, "loss": 3.7632, "step": 3970 }, { "epoch": 0.76, "learning_rate": 2.7757292659675883e-06, "loss": 3.2843, "step": 3971 }, { "epoch": 0.76, "learning_rate": 2.775672068636797e-06, "loss": 3.1137, "step": 3972 }, { "epoch": 0.76, "learning_rate": 2.775614871306006e-06, "loss": 3.4501, "step": 3973 }, { "epoch": 0.76, "learning_rate": 2.7755576739752143e-06, "loss": 3.554, "step": 3974 }, { "epoch": 0.76, "learning_rate": 2.7755004766444235e-06, "loss": 3.8198, "step": 3975 }, { "epoch": 0.76, "learning_rate": 2.775443279313632e-06, "loss": 3.6693, "step": 3976 }, { "epoch": 0.76, "learning_rate": 2.7753860819828406e-06, "loss": 3.1179, "step": 3977 }, { "epoch": 0.76, "learning_rate": 2.77532888465205e-06, "loss": 3.3964, "step": 3978 }, { "epoch": 0.76, "learning_rate": 2.7752716873212582e-06, "loss": 3.5003, "step": 3979 }, { "epoch": 0.76, "learning_rate": 2.7752144899904675e-06, "loss": 3.5329, "step": 3980 }, { "epoch": 0.76, "learning_rate": 2.775157292659676e-06, "loss": 3.6465, "step": 3981 }, { "epoch": 0.76, "learning_rate": 2.7751000953288846e-06, "loss": 3.2723, "step": 3982 }, { "epoch": 0.76, "learning_rate": 2.775042897998094e-06, "loss": 3.5877, "step": 3983 }, { "epoch": 0.76, "learning_rate": 2.774985700667302e-06, "loss": 3.6338, "step": 3984 }, { "epoch": 0.76, "learning_rate": 2.774928503336511e-06, "loss": 3.1016, "step": 3985 }, { "epoch": 0.76, "learning_rate": 2.77487130600572e-06, "loss": 3.384, "step": 3986 }, { "epoch": 0.76, "learning_rate": 2.7748141086749286e-06, "loss": 3.7006, "step": 3987 }, { "epoch": 0.76, "learning_rate": 2.7747569113441374e-06, "loss": 3.7635, "step": 3988 }, { "epoch": 0.76, "learning_rate": 2.774699714013346e-06, "loss": 3.2555, "step": 3989 }, { "epoch": 0.76, "learning_rate": 2.774642516682555e-06, "loss": 3.0686, "step": 3990 }, { "epoch": 0.76, "learning_rate": 2.7745853193517633e-06, "loss": 3.2253, "step": 3991 }, { "epoch": 0.76, "learning_rate": 2.7745281220209726e-06, "loss": 3.3238, "step": 3992 }, { "epoch": 0.76, "learning_rate": 2.7744709246901813e-06, "loss": 3.5413, "step": 3993 }, { "epoch": 0.76, "learning_rate": 2.7744137273593897e-06, "loss": 3.6303, "step": 3994 }, { "epoch": 0.76, "learning_rate": 2.774356530028599e-06, "loss": 3.6386, "step": 3995 }, { "epoch": 0.76, "learning_rate": 2.7742993326978073e-06, "loss": 3.0864, "step": 3996 }, { "epoch": 0.76, "learning_rate": 2.7742421353670165e-06, "loss": 3.4857, "step": 3997 }, { "epoch": 0.76, "learning_rate": 2.774184938036225e-06, "loss": 3.2275, "step": 3998 }, { "epoch": 0.76, "learning_rate": 2.7741277407054337e-06, "loss": 3.6803, "step": 3999 }, { "epoch": 0.76, "learning_rate": 2.774070543374643e-06, "loss": 3.5955, "step": 4000 }, { "epoch": 0.76, "learning_rate": 2.7740133460438513e-06, "loss": 3.3056, "step": 4001 }, { "epoch": 0.76, "learning_rate": 2.77395614871306e-06, "loss": 2.8361, "step": 4002 }, { "epoch": 0.76, "learning_rate": 2.773898951382269e-06, "loss": 3.5287, "step": 4003 }, { "epoch": 0.76, "learning_rate": 2.7738417540514777e-06, "loss": 3.7093, "step": 4004 }, { "epoch": 0.76, "learning_rate": 2.7737845567206865e-06, "loss": 3.3148, "step": 4005 }, { "epoch": 0.76, "learning_rate": 2.7737273593898952e-06, "loss": 3.0089, "step": 4006 }, { "epoch": 0.76, "learning_rate": 2.773670162059104e-06, "loss": 3.4477, "step": 4007 }, { "epoch": 0.76, "learning_rate": 2.7736129647283124e-06, "loss": 3.37, "step": 4008 }, { "epoch": 0.76, "learning_rate": 2.7735557673975216e-06, "loss": 3.4499, "step": 4009 }, { "epoch": 0.76, "learning_rate": 2.7734985700667304e-06, "loss": 3.4387, "step": 4010 }, { "epoch": 0.76, "learning_rate": 2.773441372735939e-06, "loss": 3.6823, "step": 4011 }, { "epoch": 0.76, "learning_rate": 2.773384175405148e-06, "loss": 3.374, "step": 4012 }, { "epoch": 0.76, "learning_rate": 2.7733269780743564e-06, "loss": 3.8265, "step": 4013 }, { "epoch": 0.76, "learning_rate": 2.7732697807435656e-06, "loss": 3.526, "step": 4014 }, { "epoch": 0.76, "learning_rate": 2.773212583412774e-06, "loss": 3.4608, "step": 4015 }, { "epoch": 0.76, "learning_rate": 2.7731553860819828e-06, "loss": 3.2354, "step": 4016 }, { "epoch": 0.77, "learning_rate": 2.773098188751192e-06, "loss": 4.1785, "step": 4017 }, { "epoch": 0.77, "learning_rate": 2.7730409914204003e-06, "loss": 3.526, "step": 4018 }, { "epoch": 0.77, "learning_rate": 2.772983794089609e-06, "loss": 3.6234, "step": 4019 }, { "epoch": 0.77, "learning_rate": 2.772926596758818e-06, "loss": 3.1019, "step": 4020 }, { "epoch": 0.77, "learning_rate": 2.7728693994280267e-06, "loss": 3.0158, "step": 4021 }, { "epoch": 0.77, "learning_rate": 2.7728122020972355e-06, "loss": 3.4387, "step": 4022 }, { "epoch": 0.77, "learning_rate": 2.7727550047664443e-06, "loss": 3.8676, "step": 4023 }, { "epoch": 0.77, "learning_rate": 2.772697807435653e-06, "loss": 3.2298, "step": 4024 }, { "epoch": 0.77, "learning_rate": 2.772640610104862e-06, "loss": 3.6352, "step": 4025 }, { "epoch": 0.77, "learning_rate": 2.7725834127740707e-06, "loss": 3.9075, "step": 4026 }, { "epoch": 0.77, "learning_rate": 2.7725262154432795e-06, "loss": 3.5236, "step": 4027 }, { "epoch": 0.77, "learning_rate": 2.7724690181124883e-06, "loss": 3.6184, "step": 4028 }, { "epoch": 0.77, "learning_rate": 2.772411820781697e-06, "loss": 3.4121, "step": 4029 }, { "epoch": 0.77, "learning_rate": 2.7723546234509055e-06, "loss": 3.2358, "step": 4030 }, { "epoch": 0.77, "learning_rate": 2.7722974261201147e-06, "loss": 3.2556, "step": 4031 }, { "epoch": 0.77, "learning_rate": 2.7722402287893235e-06, "loss": 3.3885, "step": 4032 }, { "epoch": 0.77, "learning_rate": 2.772183031458532e-06, "loss": 3.6092, "step": 4033 }, { "epoch": 0.77, "learning_rate": 2.772125834127741e-06, "loss": 3.4379, "step": 4034 }, { "epoch": 0.77, "learning_rate": 2.7720686367969494e-06, "loss": 3.8692, "step": 4035 }, { "epoch": 0.77, "learning_rate": 2.7720114394661582e-06, "loss": 3.3895, "step": 4036 }, { "epoch": 0.77, "learning_rate": 2.771954242135367e-06, "loss": 3.8078, "step": 4037 }, { "epoch": 0.77, "learning_rate": 2.771897044804576e-06, "loss": 3.3477, "step": 4038 }, { "epoch": 0.77, "learning_rate": 2.7718398474737846e-06, "loss": 3.3648, "step": 4039 }, { "epoch": 0.77, "learning_rate": 2.7717826501429934e-06, "loss": 3.5527, "step": 4040 }, { "epoch": 0.77, "learning_rate": 2.771725452812202e-06, "loss": 3.3028, "step": 4041 }, { "epoch": 0.77, "learning_rate": 2.771668255481411e-06, "loss": 3.9784, "step": 4042 }, { "epoch": 0.77, "learning_rate": 2.7716110581506198e-06, "loss": 4.0217, "step": 4043 }, { "epoch": 0.77, "learning_rate": 2.7715538608198286e-06, "loss": 3.6417, "step": 4044 }, { "epoch": 0.77, "learning_rate": 2.7714966634890374e-06, "loss": 3.1361, "step": 4045 }, { "epoch": 0.77, "learning_rate": 2.771439466158246e-06, "loss": 3.659, "step": 4046 }, { "epoch": 0.77, "learning_rate": 2.7713822688274545e-06, "loss": 3.2353, "step": 4047 }, { "epoch": 0.77, "learning_rate": 2.7713250714966637e-06, "loss": 3.0901, "step": 4048 }, { "epoch": 0.77, "learning_rate": 2.7712678741658725e-06, "loss": 3.5174, "step": 4049 }, { "epoch": 0.77, "learning_rate": 2.771210676835081e-06, "loss": 3.778, "step": 4050 }, { "epoch": 0.77, "learning_rate": 2.77115347950429e-06, "loss": 3.835, "step": 4051 }, { "epoch": 0.77, "learning_rate": 2.7710962821734985e-06, "loss": 3.7602, "step": 4052 }, { "epoch": 0.77, "learning_rate": 2.7710390848427073e-06, "loss": 3.5294, "step": 4053 }, { "epoch": 0.77, "learning_rate": 2.770981887511916e-06, "loss": 3.3751, "step": 4054 }, { "epoch": 0.77, "learning_rate": 2.770924690181125e-06, "loss": 3.5143, "step": 4055 }, { "epoch": 0.77, "learning_rate": 2.7708674928503337e-06, "loss": 3.49, "step": 4056 }, { "epoch": 0.77, "learning_rate": 2.7708102955195425e-06, "loss": 2.861, "step": 4057 }, { "epoch": 0.77, "learning_rate": 2.7707530981887513e-06, "loss": 3.1976, "step": 4058 }, { "epoch": 0.77, "learning_rate": 2.77069590085796e-06, "loss": 3.583, "step": 4059 }, { "epoch": 0.77, "learning_rate": 2.770638703527169e-06, "loss": 3.5496, "step": 4060 }, { "epoch": 0.77, "learning_rate": 2.7705815061963776e-06, "loss": 2.8712, "step": 4061 }, { "epoch": 0.77, "learning_rate": 2.7705243088655864e-06, "loss": 3.3937, "step": 4062 }, { "epoch": 0.77, "learning_rate": 2.7704671115347952e-06, "loss": 3.2758, "step": 4063 }, { "epoch": 0.77, "learning_rate": 2.7704099142040036e-06, "loss": 3.4064, "step": 4064 }, { "epoch": 0.77, "learning_rate": 2.770352716873213e-06, "loss": 3.3698, "step": 4065 }, { "epoch": 0.77, "learning_rate": 2.7702955195424216e-06, "loss": 3.535, "step": 4066 }, { "epoch": 0.77, "learning_rate": 2.77023832221163e-06, "loss": 3.5066, "step": 4067 }, { "epoch": 0.77, "learning_rate": 2.770181124880839e-06, "loss": 3.2826, "step": 4068 }, { "epoch": 0.78, "learning_rate": 2.7701239275500476e-06, "loss": 3.2778, "step": 4069 }, { "epoch": 0.78, "learning_rate": 2.7700667302192564e-06, "loss": 3.2871, "step": 4070 }, { "epoch": 0.78, "learning_rate": 2.7700095328884656e-06, "loss": 3.5483, "step": 4071 }, { "epoch": 0.78, "learning_rate": 2.769952335557674e-06, "loss": 3.2762, "step": 4072 }, { "epoch": 0.78, "learning_rate": 2.7698951382268827e-06, "loss": 3.3753, "step": 4073 }, { "epoch": 0.78, "learning_rate": 2.7698379408960915e-06, "loss": 3.5518, "step": 4074 }, { "epoch": 0.78, "learning_rate": 2.7697807435653003e-06, "loss": 3.1954, "step": 4075 }, { "epoch": 0.78, "learning_rate": 2.769723546234509e-06, "loss": 3.5523, "step": 4076 }, { "epoch": 0.78, "learning_rate": 2.769666348903718e-06, "loss": 2.8926, "step": 4077 }, { "epoch": 0.78, "learning_rate": 2.7696091515729267e-06, "loss": 3.3913, "step": 4078 }, { "epoch": 0.78, "learning_rate": 2.7695519542421355e-06, "loss": 3.5757, "step": 4079 }, { "epoch": 0.78, "learning_rate": 2.7694947569113443e-06, "loss": 3.3019, "step": 4080 }, { "epoch": 0.78, "learning_rate": 2.769437559580553e-06, "loss": 3.5524, "step": 4081 }, { "epoch": 0.78, "learning_rate": 2.769380362249762e-06, "loss": 3.6058, "step": 4082 }, { "epoch": 0.78, "learning_rate": 2.7693231649189707e-06, "loss": 3.5868, "step": 4083 }, { "epoch": 0.78, "learning_rate": 2.769265967588179e-06, "loss": 3.6992, "step": 4084 }, { "epoch": 0.78, "learning_rate": 2.7692087702573883e-06, "loss": 3.1939, "step": 4085 }, { "epoch": 0.78, "learning_rate": 2.7691515729265966e-06, "loss": 3.6817, "step": 4086 }, { "epoch": 0.78, "learning_rate": 2.7690943755958054e-06, "loss": 3.3575, "step": 4087 }, { "epoch": 0.78, "learning_rate": 2.7690371782650146e-06, "loss": 3.5365, "step": 4088 }, { "epoch": 0.78, "learning_rate": 2.768979980934223e-06, "loss": 3.4154, "step": 4089 }, { "epoch": 0.78, "learning_rate": 2.768922783603432e-06, "loss": 3.7301, "step": 4090 }, { "epoch": 0.78, "learning_rate": 2.7688655862726406e-06, "loss": 3.2377, "step": 4091 }, { "epoch": 0.78, "learning_rate": 2.7688083889418494e-06, "loss": 3.8465, "step": 4092 }, { "epoch": 0.78, "learning_rate": 2.768751191611058e-06, "loss": 3.4938, "step": 4093 }, { "epoch": 0.78, "learning_rate": 2.768693994280267e-06, "loss": 3.6254, "step": 4094 }, { "epoch": 0.78, "learning_rate": 2.7686367969494758e-06, "loss": 3.2451, "step": 4095 }, { "epoch": 0.78, "learning_rate": 2.7685795996186846e-06, "loss": 2.9375, "step": 4096 }, { "epoch": 0.78, "learning_rate": 2.7685224022878934e-06, "loss": 2.9685, "step": 4097 }, { "epoch": 0.78, "learning_rate": 2.768465204957102e-06, "loss": 3.4235, "step": 4098 }, { "epoch": 0.78, "learning_rate": 2.768408007626311e-06, "loss": 3.6586, "step": 4099 }, { "epoch": 0.78, "learning_rate": 2.7683508102955197e-06, "loss": 3.1057, "step": 4100 }, { "epoch": 0.78, "learning_rate": 2.768293612964728e-06, "loss": 3.1846, "step": 4101 }, { "epoch": 0.78, "learning_rate": 2.7682364156339373e-06, "loss": 3.5796, "step": 4102 }, { "epoch": 0.78, "learning_rate": 2.7681792183031457e-06, "loss": 3.7215, "step": 4103 }, { "epoch": 0.78, "learning_rate": 2.7681220209723545e-06, "loss": 3.2693, "step": 4104 }, { "epoch": 0.78, "learning_rate": 2.7680648236415637e-06, "loss": 3.5266, "step": 4105 }, { "epoch": 0.78, "learning_rate": 2.768007626310772e-06, "loss": 3.379, "step": 4106 }, { "epoch": 0.78, "learning_rate": 2.767950428979981e-06, "loss": 2.9531, "step": 4107 }, { "epoch": 0.78, "learning_rate": 2.7678932316491897e-06, "loss": 3.2692, "step": 4108 }, { "epoch": 0.78, "learning_rate": 2.7678360343183985e-06, "loss": 3.2637, "step": 4109 }, { "epoch": 0.78, "learning_rate": 2.7677788369876077e-06, "loss": 2.8452, "step": 4110 }, { "epoch": 0.78, "learning_rate": 2.767721639656816e-06, "loss": 3.6189, "step": 4111 }, { "epoch": 0.78, "learning_rate": 2.767664442326025e-06, "loss": 3.2832, "step": 4112 }, { "epoch": 0.78, "learning_rate": 2.7676072449952336e-06, "loss": 3.1614, "step": 4113 }, { "epoch": 0.78, "learning_rate": 2.7675500476644424e-06, "loss": 3.5454, "step": 4114 }, { "epoch": 0.78, "learning_rate": 2.7674928503336512e-06, "loss": 3.4823, "step": 4115 }, { "epoch": 0.78, "learning_rate": 2.76743565300286e-06, "loss": 3.425, "step": 4116 }, { "epoch": 0.78, "learning_rate": 2.767378455672069e-06, "loss": 3.1514, "step": 4117 }, { "epoch": 0.78, "learning_rate": 2.767321258341277e-06, "loss": 3.0766, "step": 4118 }, { "epoch": 0.78, "learning_rate": 2.7672640610104864e-06, "loss": 3.77, "step": 4119 }, { "epoch": 0.78, "learning_rate": 2.767206863679695e-06, "loss": 3.3291, "step": 4120 }, { "epoch": 0.78, "learning_rate": 2.7671496663489036e-06, "loss": 3.3697, "step": 4121 }, { "epoch": 0.79, "learning_rate": 2.7670924690181128e-06, "loss": 3.3228, "step": 4122 }, { "epoch": 0.79, "learning_rate": 2.767035271687321e-06, "loss": 3.8205, "step": 4123 }, { "epoch": 0.79, "learning_rate": 2.76697807435653e-06, "loss": 3.6689, "step": 4124 }, { "epoch": 0.79, "learning_rate": 2.7669208770257387e-06, "loss": 3.1785, "step": 4125 }, { "epoch": 0.79, "learning_rate": 2.7668636796949475e-06, "loss": 4.0197, "step": 4126 }, { "epoch": 0.79, "learning_rate": 2.7668064823641568e-06, "loss": 3.471, "step": 4127 }, { "epoch": 0.79, "learning_rate": 2.766749285033365e-06, "loss": 3.6265, "step": 4128 }, { "epoch": 0.79, "learning_rate": 2.766692087702574e-06, "loss": 3.438, "step": 4129 }, { "epoch": 0.79, "learning_rate": 2.7666348903717827e-06, "loss": 3.325, "step": 4130 }, { "epoch": 0.79, "learning_rate": 2.7665776930409915e-06, "loss": 3.7171, "step": 4131 }, { "epoch": 0.79, "learning_rate": 2.7665204957102003e-06, "loss": 3.501, "step": 4132 }, { "epoch": 0.79, "learning_rate": 2.766463298379409e-06, "loss": 3.6383, "step": 4133 }, { "epoch": 0.79, "learning_rate": 2.766406101048618e-06, "loss": 3.4837, "step": 4134 }, { "epoch": 0.79, "learning_rate": 2.7663489037178263e-06, "loss": 3.4872, "step": 4135 }, { "epoch": 0.79, "learning_rate": 2.7662917063870355e-06, "loss": 3.6869, "step": 4136 }, { "epoch": 0.79, "learning_rate": 2.7662345090562443e-06, "loss": 3.56, "step": 4137 }, { "epoch": 0.79, "learning_rate": 2.7661773117254526e-06, "loss": 3.8088, "step": 4138 }, { "epoch": 0.79, "learning_rate": 2.766120114394662e-06, "loss": 3.1325, "step": 4139 }, { "epoch": 0.79, "learning_rate": 2.7660629170638702e-06, "loss": 3.3583, "step": 4140 }, { "epoch": 0.79, "learning_rate": 2.766005719733079e-06, "loss": 3.7646, "step": 4141 }, { "epoch": 0.79, "learning_rate": 2.765948522402288e-06, "loss": 3.7172, "step": 4142 }, { "epoch": 0.79, "learning_rate": 2.7658913250714966e-06, "loss": 3.7099, "step": 4143 }, { "epoch": 0.79, "learning_rate": 2.765834127740706e-06, "loss": 3.7197, "step": 4144 }, { "epoch": 0.79, "learning_rate": 2.765776930409914e-06, "loss": 3.6737, "step": 4145 }, { "epoch": 0.79, "learning_rate": 2.765719733079123e-06, "loss": 3.4195, "step": 4146 }, { "epoch": 0.79, "learning_rate": 2.7656625357483318e-06, "loss": 3.6515, "step": 4147 }, { "epoch": 0.79, "learning_rate": 2.7656053384175406e-06, "loss": 3.0047, "step": 4148 }, { "epoch": 0.79, "learning_rate": 2.7655481410867494e-06, "loss": 3.8151, "step": 4149 }, { "epoch": 0.79, "learning_rate": 2.765490943755958e-06, "loss": 3.1083, "step": 4150 }, { "epoch": 0.79, "learning_rate": 2.765433746425167e-06, "loss": 3.3185, "step": 4151 }, { "epoch": 0.79, "learning_rate": 2.7653765490943753e-06, "loss": 3.264, "step": 4152 }, { "epoch": 0.79, "learning_rate": 2.7653193517635845e-06, "loss": 3.2129, "step": 4153 }, { "epoch": 0.79, "learning_rate": 2.7652621544327933e-06, "loss": 3.3634, "step": 4154 }, { "epoch": 0.79, "learning_rate": 2.7652049571020017e-06, "loss": 3.8121, "step": 4155 }, { "epoch": 0.79, "learning_rate": 2.765147759771211e-06, "loss": 3.4929, "step": 4156 }, { "epoch": 0.79, "learning_rate": 2.7650905624404193e-06, "loss": 3.2573, "step": 4157 }, { "epoch": 0.79, "learning_rate": 2.7650333651096285e-06, "loss": 3.8044, "step": 4158 }, { "epoch": 0.79, "learning_rate": 2.7649761677788373e-06, "loss": 3.5783, "step": 4159 }, { "epoch": 0.79, "learning_rate": 2.7649189704480457e-06, "loss": 3.6052, "step": 4160 }, { "epoch": 0.79, "learning_rate": 2.764861773117255e-06, "loss": 3.4877, "step": 4161 }, { "epoch": 0.79, "learning_rate": 2.7648045757864633e-06, "loss": 3.6607, "step": 4162 }, { "epoch": 0.79, "learning_rate": 2.764747378455672e-06, "loss": 3.5381, "step": 4163 }, { "epoch": 0.79, "learning_rate": 2.764690181124881e-06, "loss": 3.4766, "step": 4164 }, { "epoch": 0.79, "learning_rate": 2.7646329837940897e-06, "loss": 3.9019, "step": 4165 }, { "epoch": 0.79, "learning_rate": 2.7645757864632984e-06, "loss": 3.4145, "step": 4166 }, { "epoch": 0.79, "learning_rate": 2.7645185891325072e-06, "loss": 3.5761, "step": 4167 }, { "epoch": 0.79, "learning_rate": 2.764461391801716e-06, "loss": 3.3813, "step": 4168 }, { "epoch": 0.79, "learning_rate": 2.764404194470925e-06, "loss": 3.1899, "step": 4169 }, { "epoch": 0.79, "learning_rate": 2.7643469971401336e-06, "loss": 3.4051, "step": 4170 }, { "epoch": 0.79, "learning_rate": 2.7642897998093424e-06, "loss": 3.6379, "step": 4171 }, { "epoch": 0.79, "learning_rate": 2.7642326024785508e-06, "loss": 3.4455, "step": 4172 }, { "epoch": 0.79, "learning_rate": 2.76417540514776e-06, "loss": 3.127, "step": 4173 }, { "epoch": 0.8, "learning_rate": 2.7641182078169684e-06, "loss": 2.9141, "step": 4174 }, { "epoch": 0.8, "learning_rate": 2.7640610104861776e-06, "loss": 3.2648, "step": 4175 }, { "epoch": 0.8, "learning_rate": 2.7640038131553864e-06, "loss": 3.9555, "step": 4176 }, { "epoch": 0.8, "learning_rate": 2.7639466158245948e-06, "loss": 3.2206, "step": 4177 }, { "epoch": 0.8, "learning_rate": 2.763889418493804e-06, "loss": 3.4227, "step": 4178 }, { "epoch": 0.8, "learning_rate": 2.7638322211630123e-06, "loss": 3.357, "step": 4179 }, { "epoch": 0.8, "learning_rate": 2.763775023832221e-06, "loss": 3.5502, "step": 4180 }, { "epoch": 0.8, "learning_rate": 2.76371782650143e-06, "loss": 3.7104, "step": 4181 }, { "epoch": 0.8, "learning_rate": 2.7636606291706387e-06, "loss": 3.5783, "step": 4182 }, { "epoch": 0.8, "learning_rate": 2.7636034318398475e-06, "loss": 2.9993, "step": 4183 }, { "epoch": 0.8, "learning_rate": 2.7635462345090563e-06, "loss": 3.7297, "step": 4184 }, { "epoch": 0.8, "learning_rate": 2.763489037178265e-06, "loss": 3.0281, "step": 4185 }, { "epoch": 0.8, "learning_rate": 2.763431839847474e-06, "loss": 3.0769, "step": 4186 }, { "epoch": 0.8, "learning_rate": 2.7633746425166827e-06, "loss": 2.9251, "step": 4187 }, { "epoch": 0.8, "learning_rate": 2.7633174451858915e-06, "loss": 3.5449, "step": 4188 }, { "epoch": 0.8, "learning_rate": 2.7632602478551e-06, "loss": 3.4857, "step": 4189 }, { "epoch": 0.8, "learning_rate": 2.763203050524309e-06, "loss": 3.4232, "step": 4190 }, { "epoch": 0.8, "learning_rate": 2.7631458531935174e-06, "loss": 2.9501, "step": 4191 }, { "epoch": 0.8, "learning_rate": 2.7630886558627267e-06, "loss": 3.4893, "step": 4192 }, { "epoch": 0.8, "learning_rate": 2.7630314585319355e-06, "loss": 3.4176, "step": 4193 }, { "epoch": 0.8, "learning_rate": 2.762974261201144e-06, "loss": 3.4411, "step": 4194 }, { "epoch": 0.8, "learning_rate": 2.762917063870353e-06, "loss": 3.2333, "step": 4195 }, { "epoch": 0.8, "learning_rate": 2.7628598665395614e-06, "loss": 3.3645, "step": 4196 }, { "epoch": 0.8, "learning_rate": 2.76280266920877e-06, "loss": 3.4224, "step": 4197 }, { "epoch": 0.8, "learning_rate": 2.7627454718779794e-06, "loss": 2.9357, "step": 4198 }, { "epoch": 0.8, "learning_rate": 2.762688274547188e-06, "loss": 3.3433, "step": 4199 }, { "epoch": 0.8, "learning_rate": 2.7626310772163966e-06, "loss": 3.8322, "step": 4200 }, { "epoch": 0.8, "learning_rate": 2.7625738798856054e-06, "loss": 3.137, "step": 4201 }, { "epoch": 0.8, "learning_rate": 2.762516682554814e-06, "loss": 3.3852, "step": 4202 }, { "epoch": 0.8, "learning_rate": 2.762459485224023e-06, "loss": 3.5969, "step": 4203 }, { "epoch": 0.8, "learning_rate": 2.7624022878932318e-06, "loss": 3.6649, "step": 4204 }, { "epoch": 0.8, "learning_rate": 2.7623450905624406e-06, "loss": 3.755, "step": 4205 }, { "epoch": 0.8, "learning_rate": 2.762287893231649e-06, "loss": 2.7869, "step": 4206 }, { "epoch": 0.8, "learning_rate": 2.762230695900858e-06, "loss": 3.3179, "step": 4207 }, { "epoch": 0.8, "learning_rate": 2.762173498570067e-06, "loss": 3.3532, "step": 4208 }, { "epoch": 0.8, "learning_rate": 2.7621163012392757e-06, "loss": 3.5767, "step": 4209 }, { "epoch": 0.8, "learning_rate": 2.7620591039084845e-06, "loss": 3.1201, "step": 4210 }, { "epoch": 0.8, "learning_rate": 2.762001906577693e-06, "loss": 4.2101, "step": 4211 }, { "epoch": 0.8, "learning_rate": 2.761944709246902e-06, "loss": 3.5168, "step": 4212 }, { "epoch": 0.8, "learning_rate": 2.7618875119161105e-06, "loss": 3.5244, "step": 4213 }, { "epoch": 0.8, "learning_rate": 2.7618303145853193e-06, "loss": 3.4213, "step": 4214 }, { "epoch": 0.8, "learning_rate": 2.7617731172545285e-06, "loss": 3.2107, "step": 4215 }, { "epoch": 0.8, "learning_rate": 2.761715919923737e-06, "loss": 3.346, "step": 4216 }, { "epoch": 0.8, "learning_rate": 2.7616587225929457e-06, "loss": 3.4584, "step": 4217 }, { "epoch": 0.8, "learning_rate": 2.7616015252621545e-06, "loss": 3.5461, "step": 4218 }, { "epoch": 0.8, "learning_rate": 2.7615443279313632e-06, "loss": 3.0725, "step": 4219 }, { "epoch": 0.8, "learning_rate": 2.761487130600572e-06, "loss": 3.25, "step": 4220 }, { "epoch": 0.8, "learning_rate": 2.761429933269781e-06, "loss": 3.6098, "step": 4221 }, { "epoch": 0.8, "learning_rate": 2.7613727359389896e-06, "loss": 3.2964, "step": 4222 }, { "epoch": 0.8, "learning_rate": 2.7613155386081984e-06, "loss": 3.1887, "step": 4223 }, { "epoch": 0.8, "learning_rate": 2.7612583412774072e-06, "loss": 3.2915, "step": 4224 }, { "epoch": 0.8, "learning_rate": 2.761201143946616e-06, "loss": 3.6929, "step": 4225 }, { "epoch": 0.8, "learning_rate": 2.761143946615825e-06, "loss": 3.4372, "step": 4226 }, { "epoch": 0.81, "learning_rate": 2.7610867492850336e-06, "loss": 3.0906, "step": 4227 }, { "epoch": 0.81, "learning_rate": 2.761029551954242e-06, "loss": 3.6145, "step": 4228 }, { "epoch": 0.81, "learning_rate": 2.760972354623451e-06, "loss": 3.7514, "step": 4229 }, { "epoch": 0.81, "learning_rate": 2.7609151572926596e-06, "loss": 3.8346, "step": 4230 }, { "epoch": 0.81, "learning_rate": 2.7608579599618683e-06, "loss": 3.8863, "step": 4231 }, { "epoch": 0.81, "learning_rate": 2.7608007626310776e-06, "loss": 3.0568, "step": 4232 }, { "epoch": 0.81, "learning_rate": 2.760743565300286e-06, "loss": 3.6728, "step": 4233 }, { "epoch": 0.81, "learning_rate": 2.7606863679694947e-06, "loss": 2.9146, "step": 4234 }, { "epoch": 0.81, "learning_rate": 2.7606291706387035e-06, "loss": 3.3562, "step": 4235 }, { "epoch": 0.81, "learning_rate": 2.7605719733079123e-06, "loss": 3.6087, "step": 4236 }, { "epoch": 0.81, "learning_rate": 2.760514775977121e-06, "loss": 3.6157, "step": 4237 }, { "epoch": 0.81, "learning_rate": 2.76045757864633e-06, "loss": 3.5317, "step": 4238 }, { "epoch": 0.81, "learning_rate": 2.7604003813155387e-06, "loss": 3.4426, "step": 4239 }, { "epoch": 0.81, "learning_rate": 2.7603431839847475e-06, "loss": 3.5049, "step": 4240 }, { "epoch": 0.81, "learning_rate": 2.7602859866539563e-06, "loss": 3.6397, "step": 4241 }, { "epoch": 0.81, "learning_rate": 2.760228789323165e-06, "loss": 3.5679, "step": 4242 }, { "epoch": 0.81, "learning_rate": 2.760171591992374e-06, "loss": 3.5333, "step": 4243 }, { "epoch": 0.81, "learning_rate": 2.7601143946615827e-06, "loss": 3.0622, "step": 4244 }, { "epoch": 0.81, "learning_rate": 2.760057197330791e-06, "loss": 3.8129, "step": 4245 }, { "epoch": 0.81, "learning_rate": 2.7600000000000003e-06, "loss": 3.3999, "step": 4246 }, { "epoch": 0.81, "learning_rate": 2.759942802669209e-06, "loss": 3.5328, "step": 4247 }, { "epoch": 0.81, "learning_rate": 2.7598856053384174e-06, "loss": 3.2588, "step": 4248 }, { "epoch": 0.81, "learning_rate": 2.7598284080076266e-06, "loss": 3.6378, "step": 4249 }, { "epoch": 0.81, "learning_rate": 2.759771210676835e-06, "loss": 3.4173, "step": 4250 }, { "epoch": 0.81, "learning_rate": 2.759714013346044e-06, "loss": 2.9907, "step": 4251 }, { "epoch": 0.81, "learning_rate": 2.7596568160152526e-06, "loss": 3.4232, "step": 4252 }, { "epoch": 0.81, "learning_rate": 2.7595996186844614e-06, "loss": 3.4002, "step": 4253 }, { "epoch": 0.81, "learning_rate": 2.75954242135367e-06, "loss": 3.2301, "step": 4254 }, { "epoch": 0.81, "learning_rate": 2.759485224022879e-06, "loss": 3.2497, "step": 4255 }, { "epoch": 0.81, "learning_rate": 2.7594280266920878e-06, "loss": 3.7132, "step": 4256 }, { "epoch": 0.81, "learning_rate": 2.7593708293612966e-06, "loss": 3.5742, "step": 4257 }, { "epoch": 0.81, "learning_rate": 2.7593136320305054e-06, "loss": 3.4677, "step": 4258 }, { "epoch": 0.81, "learning_rate": 2.759256434699714e-06, "loss": 3.8077, "step": 4259 }, { "epoch": 0.81, "learning_rate": 2.759199237368923e-06, "loss": 3.4892, "step": 4260 }, { "epoch": 0.81, "learning_rate": 2.7591420400381317e-06, "loss": 3.3228, "step": 4261 }, { "epoch": 0.81, "learning_rate": 2.75908484270734e-06, "loss": 3.3798, "step": 4262 }, { "epoch": 0.81, "learning_rate": 2.7590276453765493e-06, "loss": 3.2296, "step": 4263 }, { "epoch": 0.81, "learning_rate": 2.758970448045758e-06, "loss": 3.5029, "step": 4264 }, { "epoch": 0.81, "learning_rate": 2.7589132507149665e-06, "loss": 3.4133, "step": 4265 }, { "epoch": 0.81, "learning_rate": 2.7588560533841757e-06, "loss": 3.7192, "step": 4266 }, { "epoch": 0.81, "learning_rate": 2.758798856053384e-06, "loss": 3.4648, "step": 4267 }, { "epoch": 0.81, "learning_rate": 2.758741658722593e-06, "loss": 3.507, "step": 4268 }, { "epoch": 0.81, "learning_rate": 2.7586844613918017e-06, "loss": 3.4615, "step": 4269 }, { "epoch": 0.81, "learning_rate": 2.7586272640610105e-06, "loss": 3.3286, "step": 4270 }, { "epoch": 0.81, "learning_rate": 2.7585700667302193e-06, "loss": 3.3278, "step": 4271 }, { "epoch": 0.81, "learning_rate": 2.758512869399428e-06, "loss": 3.1645, "step": 4272 }, { "epoch": 0.81, "learning_rate": 2.758455672068637e-06, "loss": 3.5053, "step": 4273 }, { "epoch": 0.81, "learning_rate": 2.7583984747378456e-06, "loss": 3.4779, "step": 4274 }, { "epoch": 0.81, "learning_rate": 2.7583412774070544e-06, "loss": 3.5436, "step": 4275 }, { "epoch": 0.81, "learning_rate": 2.7582840800762632e-06, "loss": 3.0795, "step": 4276 }, { "epoch": 0.81, "learning_rate": 2.758226882745472e-06, "loss": 3.9237, "step": 4277 }, { "epoch": 0.81, "learning_rate": 2.758169685414681e-06, "loss": 3.4143, "step": 4278 }, { "epoch": 0.82, "learning_rate": 2.758112488083889e-06, "loss": 3.7272, "step": 4279 }, { "epoch": 0.82, "learning_rate": 2.7580552907530984e-06, "loss": 3.3545, "step": 4280 }, { "epoch": 0.82, "learning_rate": 2.757998093422307e-06, "loss": 3.4567, "step": 4281 }, { "epoch": 0.82, "learning_rate": 2.7579408960915156e-06, "loss": 3.6838, "step": 4282 }, { "epoch": 0.82, "learning_rate": 2.7578836987607248e-06, "loss": 3.2283, "step": 4283 }, { "epoch": 0.82, "learning_rate": 2.757826501429933e-06, "loss": 3.3973, "step": 4284 }, { "epoch": 0.82, "learning_rate": 2.757769304099142e-06, "loss": 3.7292, "step": 4285 }, { "epoch": 0.82, "learning_rate": 2.7577121067683507e-06, "loss": 3.7274, "step": 4286 }, { "epoch": 0.82, "learning_rate": 2.7576549094375595e-06, "loss": 3.0821, "step": 4287 }, { "epoch": 0.82, "learning_rate": 2.7575977121067687e-06, "loss": 3.6392, "step": 4288 }, { "epoch": 0.82, "learning_rate": 2.757540514775977e-06, "loss": 3.4833, "step": 4289 }, { "epoch": 0.82, "learning_rate": 2.757483317445186e-06, "loss": 2.9462, "step": 4290 }, { "epoch": 0.82, "learning_rate": 2.7574261201143947e-06, "loss": 3.6552, "step": 4291 }, { "epoch": 0.82, "learning_rate": 2.7573689227836035e-06, "loss": 3.1485, "step": 4292 }, { "epoch": 0.82, "learning_rate": 2.7573117254528123e-06, "loss": 3.0425, "step": 4293 }, { "epoch": 0.82, "learning_rate": 2.757254528122021e-06, "loss": 3.3375, "step": 4294 }, { "epoch": 0.82, "learning_rate": 2.75719733079123e-06, "loss": 3.5115, "step": 4295 }, { "epoch": 0.82, "learning_rate": 2.7571401334604387e-06, "loss": 3.5564, "step": 4296 }, { "epoch": 0.82, "learning_rate": 2.7570829361296475e-06, "loss": 3.525, "step": 4297 }, { "epoch": 0.82, "learning_rate": 2.7570257387988563e-06, "loss": 3.1616, "step": 4298 }, { "epoch": 0.82, "learning_rate": 2.7569685414680646e-06, "loss": 3.4416, "step": 4299 }, { "epoch": 0.82, "learning_rate": 2.756911344137274e-06, "loss": 3.7589, "step": 4300 }, { "epoch": 0.82, "learning_rate": 2.7568541468064822e-06, "loss": 3.0537, "step": 4301 }, { "epoch": 0.82, "learning_rate": 2.756796949475691e-06, "loss": 3.4533, "step": 4302 }, { "epoch": 0.82, "learning_rate": 2.7567397521449002e-06, "loss": 3.0527, "step": 4303 }, { "epoch": 0.82, "learning_rate": 2.7566825548141086e-06, "loss": 3.2519, "step": 4304 }, { "epoch": 0.82, "learning_rate": 2.756625357483318e-06, "loss": 3.3713, "step": 4305 }, { "epoch": 0.82, "learning_rate": 2.756568160152526e-06, "loss": 3.092, "step": 4306 }, { "epoch": 0.82, "learning_rate": 2.756510962821735e-06, "loss": 3.3326, "step": 4307 }, { "epoch": 0.82, "learning_rate": 2.7564537654909438e-06, "loss": 2.9642, "step": 4308 }, { "epoch": 0.82, "learning_rate": 2.7563965681601526e-06, "loss": 3.7239, "step": 4309 }, { "epoch": 0.82, "learning_rate": 2.7563393708293614e-06, "loss": 3.7854, "step": 4310 }, { "epoch": 0.82, "learning_rate": 2.75628217349857e-06, "loss": 3.7147, "step": 4311 }, { "epoch": 0.82, "learning_rate": 2.756224976167779e-06, "loss": 3.5986, "step": 4312 }, { "epoch": 0.82, "learning_rate": 2.7561677788369877e-06, "loss": 3.6192, "step": 4313 }, { "epoch": 0.82, "learning_rate": 2.7561105815061965e-06, "loss": 3.1897, "step": 4314 }, { "epoch": 0.82, "learning_rate": 2.7560533841754053e-06, "loss": 3.2469, "step": 4315 }, { "epoch": 0.82, "learning_rate": 2.7559961868446137e-06, "loss": 3.1312, "step": 4316 }, { "epoch": 0.82, "learning_rate": 2.755938989513823e-06, "loss": 3.6148, "step": 4317 }, { "epoch": 0.82, "learning_rate": 2.7558817921830313e-06, "loss": 3.9111, "step": 4318 }, { "epoch": 0.82, "learning_rate": 2.75582459485224e-06, "loss": 3.6259, "step": 4319 }, { "epoch": 0.82, "learning_rate": 2.7557673975214493e-06, "loss": 3.7566, "step": 4320 }, { "epoch": 0.82, "learning_rate": 2.7557102001906577e-06, "loss": 3.7707, "step": 4321 }, { "epoch": 0.82, "learning_rate": 2.755653002859867e-06, "loss": 3.776, "step": 4322 }, { "epoch": 0.82, "learning_rate": 2.7555958055290753e-06, "loss": 2.8857, "step": 4323 }, { "epoch": 0.82, "learning_rate": 2.755538608198284e-06, "loss": 3.1465, "step": 4324 }, { "epoch": 0.82, "learning_rate": 2.755481410867493e-06, "loss": 3.1478, "step": 4325 }, { "epoch": 0.82, "learning_rate": 2.7554242135367016e-06, "loss": 3.8802, "step": 4326 }, { "epoch": 0.82, "learning_rate": 2.7553670162059104e-06, "loss": 3.8768, "step": 4327 }, { "epoch": 0.82, "learning_rate": 2.7553098188751192e-06, "loss": 3.6698, "step": 4328 }, { "epoch": 0.82, "learning_rate": 2.755252621544328e-06, "loss": 3.7518, "step": 4329 }, { "epoch": 0.82, "learning_rate": 2.755195424213537e-06, "loss": 3.1211, "step": 4330 }, { "epoch": 0.82, "learning_rate": 2.7551382268827456e-06, "loss": 3.7619, "step": 4331 }, { "epoch": 0.83, "learning_rate": 2.7550810295519544e-06, "loss": 3.3408, "step": 4332 }, { "epoch": 0.83, "learning_rate": 2.7550238322211628e-06, "loss": 3.3648, "step": 4333 }, { "epoch": 0.83, "learning_rate": 2.754966634890372e-06, "loss": 3.2114, "step": 4334 }, { "epoch": 0.83, "learning_rate": 2.754909437559581e-06, "loss": 3.5607, "step": 4335 }, { "epoch": 0.83, "learning_rate": 2.754852240228789e-06, "loss": 3.0431, "step": 4336 }, { "epoch": 0.83, "learning_rate": 2.7547950428979984e-06, "loss": 3.4751, "step": 4337 }, { "epoch": 0.83, "learning_rate": 2.7547378455672067e-06, "loss": 3.5445, "step": 4338 }, { "epoch": 0.83, "learning_rate": 2.754680648236416e-06, "loss": 3.5538, "step": 4339 }, { "epoch": 0.83, "learning_rate": 2.7546234509056243e-06, "loss": 3.1637, "step": 4340 }, { "epoch": 0.83, "learning_rate": 2.754566253574833e-06, "loss": 3.2786, "step": 4341 }, { "epoch": 0.83, "learning_rate": 2.7545090562440423e-06, "loss": 3.4499, "step": 4342 }, { "epoch": 0.83, "learning_rate": 2.7544518589132507e-06, "loss": 4.0193, "step": 4343 }, { "epoch": 0.83, "learning_rate": 2.7543946615824595e-06, "loss": 3.3999, "step": 4344 }, { "epoch": 0.83, "learning_rate": 2.7543374642516683e-06, "loss": 3.2146, "step": 4345 }, { "epoch": 0.83, "learning_rate": 2.754280266920877e-06, "loss": 3.2637, "step": 4346 }, { "epoch": 0.83, "learning_rate": 2.754223069590086e-06, "loss": 3.9073, "step": 4347 }, { "epoch": 0.83, "learning_rate": 2.7541658722592947e-06, "loss": 3.7391, "step": 4348 }, { "epoch": 0.83, "learning_rate": 2.7541086749285035e-06, "loss": 3.1046, "step": 4349 }, { "epoch": 0.83, "learning_rate": 2.754051477597712e-06, "loss": 3.387, "step": 4350 }, { "epoch": 0.83, "learning_rate": 2.753994280266921e-06, "loss": 3.2411, "step": 4351 }, { "epoch": 0.83, "learning_rate": 2.75393708293613e-06, "loss": 3.4859, "step": 4352 }, { "epoch": 0.83, "learning_rate": 2.7538798856053382e-06, "loss": 3.6165, "step": 4353 }, { "epoch": 0.83, "learning_rate": 2.7538226882745474e-06, "loss": 3.4397, "step": 4354 }, { "epoch": 0.83, "learning_rate": 2.753765490943756e-06, "loss": 3.3354, "step": 4355 }, { "epoch": 0.83, "learning_rate": 2.753708293612965e-06, "loss": 3.7959, "step": 4356 }, { "epoch": 0.83, "learning_rate": 2.7536510962821734e-06, "loss": 3.2041, "step": 4357 }, { "epoch": 0.83, "learning_rate": 2.753593898951382e-06, "loss": 3.433, "step": 4358 }, { "epoch": 0.83, "learning_rate": 2.7535367016205914e-06, "loss": 3.6691, "step": 4359 }, { "epoch": 0.83, "learning_rate": 2.7534795042897998e-06, "loss": 3.7409, "step": 4360 }, { "epoch": 0.83, "learning_rate": 2.7534223069590086e-06, "loss": 3.7873, "step": 4361 }, { "epoch": 0.83, "learning_rate": 2.7533651096282174e-06, "loss": 3.3677, "step": 4362 }, { "epoch": 0.83, "learning_rate": 2.753307912297426e-06, "loss": 3.7124, "step": 4363 }, { "epoch": 0.83, "learning_rate": 2.753250714966635e-06, "loss": 3.3231, "step": 4364 }, { "epoch": 0.83, "learning_rate": 2.7531935176358438e-06, "loss": 3.3566, "step": 4365 }, { "epoch": 0.83, "learning_rate": 2.7531363203050525e-06, "loss": 3.404, "step": 4366 }, { "epoch": 0.83, "learning_rate": 2.753079122974261e-06, "loss": 3.0258, "step": 4367 }, { "epoch": 0.83, "learning_rate": 2.75302192564347e-06, "loss": 3.6604, "step": 4368 }, { "epoch": 0.83, "learning_rate": 2.752964728312679e-06, "loss": 3.4991, "step": 4369 }, { "epoch": 0.83, "learning_rate": 2.7529075309818877e-06, "loss": 3.3027, "step": 4370 }, { "epoch": 0.83, "learning_rate": 2.7528503336510965e-06, "loss": 3.4711, "step": 4371 }, { "epoch": 0.83, "learning_rate": 2.752793136320305e-06, "loss": 3.4063, "step": 4372 }, { "epoch": 0.83, "learning_rate": 2.752735938989514e-06, "loss": 3.857, "step": 4373 }, { "epoch": 0.83, "learning_rate": 2.7526787416587225e-06, "loss": 3.4006, "step": 4374 }, { "epoch": 0.83, "learning_rate": 2.7526215443279313e-06, "loss": 3.3565, "step": 4375 }, { "epoch": 0.83, "learning_rate": 2.7525643469971405e-06, "loss": 3.3747, "step": 4376 }, { "epoch": 0.83, "learning_rate": 2.752507149666349e-06, "loss": 3.3806, "step": 4377 }, { "epoch": 0.83, "learning_rate": 2.7524499523355577e-06, "loss": 3.2293, "step": 4378 }, { "epoch": 0.83, "learning_rate": 2.7523927550047664e-06, "loss": 3.2342, "step": 4379 }, { "epoch": 0.83, "learning_rate": 2.7523355576739752e-06, "loss": 3.7345, "step": 4380 }, { "epoch": 0.83, "learning_rate": 2.752278360343184e-06, "loss": 3.4986, "step": 4381 }, { "epoch": 0.83, "learning_rate": 2.752221163012393e-06, "loss": 3.5307, "step": 4382 }, { "epoch": 0.83, "learning_rate": 2.7521639656816016e-06, "loss": 3.7809, "step": 4383 }, { "epoch": 0.84, "learning_rate": 2.7521067683508104e-06, "loss": 3.3446, "step": 4384 }, { "epoch": 0.84, "learning_rate": 2.752049571020019e-06, "loss": 3.2917, "step": 4385 }, { "epoch": 0.84, "learning_rate": 2.751992373689228e-06, "loss": 3.3079, "step": 4386 }, { "epoch": 0.84, "learning_rate": 2.751935176358437e-06, "loss": 3.3712, "step": 4387 }, { "epoch": 0.84, "learning_rate": 2.7518779790276456e-06, "loss": 3.1678, "step": 4388 }, { "epoch": 0.84, "learning_rate": 2.751820781696854e-06, "loss": 3.8564, "step": 4389 }, { "epoch": 0.84, "learning_rate": 2.751763584366063e-06, "loss": 3.8824, "step": 4390 }, { "epoch": 0.84, "learning_rate": 2.751706387035272e-06, "loss": 3.694, "step": 4391 }, { "epoch": 0.84, "learning_rate": 2.7516491897044803e-06, "loss": 3.7544, "step": 4392 }, { "epoch": 0.84, "learning_rate": 2.7515919923736896e-06, "loss": 3.4262, "step": 4393 }, { "epoch": 0.84, "learning_rate": 2.751534795042898e-06, "loss": 3.5673, "step": 4394 }, { "epoch": 0.84, "learning_rate": 2.7514775977121067e-06, "loss": 3.7054, "step": 4395 }, { "epoch": 0.84, "learning_rate": 2.7514204003813155e-06, "loss": 3.4596, "step": 4396 }, { "epoch": 0.84, "learning_rate": 2.7513632030505243e-06, "loss": 3.7273, "step": 4397 }, { "epoch": 0.84, "learning_rate": 2.751306005719733e-06, "loss": 3.5468, "step": 4398 }, { "epoch": 0.84, "learning_rate": 2.751248808388942e-06, "loss": 3.3162, "step": 4399 }, { "epoch": 0.84, "learning_rate": 2.7511916110581507e-06, "loss": 3.5066, "step": 4400 }, { "epoch": 0.84, "learning_rate": 2.7511344137273595e-06, "loss": 3.1368, "step": 4401 }, { "epoch": 0.84, "learning_rate": 2.7510772163965683e-06, "loss": 3.7677, "step": 4402 }, { "epoch": 0.84, "learning_rate": 2.751020019065777e-06, "loss": 3.6365, "step": 4403 }, { "epoch": 0.84, "learning_rate": 2.750962821734986e-06, "loss": 3.6314, "step": 4404 }, { "epoch": 0.84, "learning_rate": 2.7509056244041947e-06, "loss": 3.6653, "step": 4405 }, { "epoch": 0.84, "learning_rate": 2.750848427073403e-06, "loss": 3.2317, "step": 4406 }, { "epoch": 0.84, "learning_rate": 2.7507912297426122e-06, "loss": 3.7439, "step": 4407 }, { "epoch": 0.84, "learning_rate": 2.750734032411821e-06, "loss": 3.625, "step": 4408 }, { "epoch": 0.84, "learning_rate": 2.7506768350810294e-06, "loss": 3.135, "step": 4409 }, { "epoch": 0.84, "learning_rate": 2.7506196377502386e-06, "loss": 3.9487, "step": 4410 }, { "epoch": 0.84, "learning_rate": 2.750562440419447e-06, "loss": 3.6201, "step": 4411 }, { "epoch": 0.84, "learning_rate": 2.750505243088656e-06, "loss": 3.185, "step": 4412 }, { "epoch": 0.84, "learning_rate": 2.7504480457578646e-06, "loss": 3.4273, "step": 4413 }, { "epoch": 0.84, "learning_rate": 2.7503908484270734e-06, "loss": 3.4015, "step": 4414 }, { "epoch": 0.84, "learning_rate": 2.750333651096282e-06, "loss": 3.2363, "step": 4415 }, { "epoch": 0.84, "learning_rate": 2.750276453765491e-06, "loss": 3.2493, "step": 4416 }, { "epoch": 0.84, "learning_rate": 2.7502192564346998e-06, "loss": 3.4618, "step": 4417 }, { "epoch": 0.84, "learning_rate": 2.7501620591039086e-06, "loss": 3.1325, "step": 4418 }, { "epoch": 0.84, "learning_rate": 2.7501048617731174e-06, "loss": 3.3542, "step": 4419 }, { "epoch": 0.84, "learning_rate": 2.750047664442326e-06, "loss": 3.6489, "step": 4420 }, { "epoch": 0.84, "learning_rate": 2.749990467111535e-06, "loss": 3.5049, "step": 4421 }, { "epoch": 0.84, "learning_rate": 2.7499332697807437e-06, "loss": 2.9829, "step": 4422 }, { "epoch": 0.84, "learning_rate": 2.7498760724499525e-06, "loss": 3.5103, "step": 4423 }, { "epoch": 0.84, "learning_rate": 2.7498188751191613e-06, "loss": 3.2505, "step": 4424 }, { "epoch": 0.84, "learning_rate": 2.74976167778837e-06, "loss": 3.4896, "step": 4425 }, { "epoch": 0.84, "learning_rate": 2.7497044804575785e-06, "loss": 3.2183, "step": 4426 }, { "epoch": 0.84, "learning_rate": 2.7496472831267877e-06, "loss": 2.9933, "step": 4427 }, { "epoch": 0.84, "learning_rate": 2.749590085795996e-06, "loss": 3.4754, "step": 4428 }, { "epoch": 0.84, "learning_rate": 2.749532888465205e-06, "loss": 3.6572, "step": 4429 }, { "epoch": 0.84, "learning_rate": 2.749475691134414e-06, "loss": 3.3569, "step": 4430 }, { "epoch": 0.84, "learning_rate": 2.7494184938036225e-06, "loss": 3.3525, "step": 4431 }, { "epoch": 0.84, "learning_rate": 2.7493612964728312e-06, "loss": 2.8952, "step": 4432 }, { "epoch": 0.84, "learning_rate": 2.74930409914204e-06, "loss": 3.8664, "step": 4433 }, { "epoch": 0.84, "learning_rate": 2.749246901811249e-06, "loss": 3.5891, "step": 4434 }, { "epoch": 0.84, "learning_rate": 2.7491897044804576e-06, "loss": 3.3164, "step": 4435 }, { "epoch": 0.84, "learning_rate": 2.7491325071496664e-06, "loss": 3.3764, "step": 4436 }, { "epoch": 0.85, "learning_rate": 2.7490753098188752e-06, "loss": 3.5376, "step": 4437 }, { "epoch": 0.85, "learning_rate": 2.749018112488084e-06, "loss": 3.2606, "step": 4438 }, { "epoch": 0.85, "learning_rate": 2.748960915157293e-06, "loss": 3.8992, "step": 4439 }, { "epoch": 0.85, "learning_rate": 2.7489037178265016e-06, "loss": 3.212, "step": 4440 }, { "epoch": 0.85, "learning_rate": 2.7488465204957104e-06, "loss": 3.7278, "step": 4441 }, { "epoch": 0.85, "learning_rate": 2.748789323164919e-06, "loss": 3.8059, "step": 4442 }, { "epoch": 0.85, "learning_rate": 2.7487321258341276e-06, "loss": 3.1259, "step": 4443 }, { "epoch": 0.85, "learning_rate": 2.7486749285033368e-06, "loss": 3.9218, "step": 4444 }, { "epoch": 0.85, "learning_rate": 2.748617731172545e-06, "loss": 2.9341, "step": 4445 }, { "epoch": 0.85, "learning_rate": 2.748560533841754e-06, "loss": 3.6688, "step": 4446 }, { "epoch": 0.85, "learning_rate": 2.748503336510963e-06, "loss": 3.0383, "step": 4447 }, { "epoch": 0.85, "learning_rate": 2.7484461391801715e-06, "loss": 3.9605, "step": 4448 }, { "epoch": 0.85, "learning_rate": 2.7483889418493803e-06, "loss": 3.3753, "step": 4449 }, { "epoch": 0.85, "learning_rate": 2.748331744518589e-06, "loss": 3.6385, "step": 4450 }, { "epoch": 0.85, "learning_rate": 2.748274547187798e-06, "loss": 3.1884, "step": 4451 }, { "epoch": 0.85, "learning_rate": 2.7482173498570067e-06, "loss": 3.2074, "step": 4452 }, { "epoch": 0.85, "learning_rate": 2.7481601525262155e-06, "loss": 3.1487, "step": 4453 }, { "epoch": 0.85, "learning_rate": 2.7481029551954243e-06, "loss": 3.6414, "step": 4454 }, { "epoch": 0.85, "learning_rate": 2.748045757864633e-06, "loss": 3.3144, "step": 4455 }, { "epoch": 0.85, "learning_rate": 2.747988560533842e-06, "loss": 3.5282, "step": 4456 }, { "epoch": 0.85, "learning_rate": 2.7479313632030507e-06, "loss": 3.0759, "step": 4457 }, { "epoch": 0.85, "learning_rate": 2.7478741658722595e-06, "loss": 3.2775, "step": 4458 }, { "epoch": 0.85, "learning_rate": 2.7478169685414683e-06, "loss": 3.415, "step": 4459 }, { "epoch": 0.85, "learning_rate": 2.7477597712106766e-06, "loss": 3.7836, "step": 4460 }, { "epoch": 0.85, "learning_rate": 2.747702573879886e-06, "loss": 3.7725, "step": 4461 }, { "epoch": 0.85, "learning_rate": 2.7476453765490942e-06, "loss": 3.6639, "step": 4462 }, { "epoch": 0.85, "learning_rate": 2.747588179218303e-06, "loss": 3.3643, "step": 4463 }, { "epoch": 0.85, "learning_rate": 2.7475309818875122e-06, "loss": 3.3899, "step": 4464 }, { "epoch": 0.85, "learning_rate": 2.7474737845567206e-06, "loss": 3.5383, "step": 4465 }, { "epoch": 0.85, "learning_rate": 2.7474165872259294e-06, "loss": 3.7083, "step": 4466 }, { "epoch": 0.85, "learning_rate": 2.747359389895138e-06, "loss": 2.955, "step": 4467 }, { "epoch": 0.85, "learning_rate": 2.747302192564347e-06, "loss": 3.3983, "step": 4468 }, { "epoch": 0.85, "learning_rate": 2.747244995233556e-06, "loss": 3.1639, "step": 4469 }, { "epoch": 0.85, "learning_rate": 2.7471877979027646e-06, "loss": 3.128, "step": 4470 }, { "epoch": 0.85, "learning_rate": 2.7471306005719734e-06, "loss": 3.281, "step": 4471 }, { "epoch": 0.85, "learning_rate": 2.747073403241182e-06, "loss": 3.316, "step": 4472 }, { "epoch": 0.85, "learning_rate": 2.747016205910391e-06, "loss": 3.7317, "step": 4473 }, { "epoch": 0.85, "learning_rate": 2.7469590085795997e-06, "loss": 3.2272, "step": 4474 }, { "epoch": 0.85, "learning_rate": 2.7469018112488085e-06, "loss": 3.5924, "step": 4475 }, { "epoch": 0.85, "learning_rate": 2.7468446139180173e-06, "loss": 3.2961, "step": 4476 }, { "epoch": 0.85, "learning_rate": 2.7467874165872257e-06, "loss": 3.2612, "step": 4477 }, { "epoch": 0.85, "learning_rate": 2.746730219256435e-06, "loss": 3.2014, "step": 4478 }, { "epoch": 0.85, "learning_rate": 2.7466730219256437e-06, "loss": 3.6879, "step": 4479 }, { "epoch": 0.85, "learning_rate": 2.746615824594852e-06, "loss": 3.3433, "step": 4480 }, { "epoch": 0.85, "learning_rate": 2.7465586272640613e-06, "loss": 3.3038, "step": 4481 }, { "epoch": 0.85, "learning_rate": 2.7465014299332697e-06, "loss": 3.4673, "step": 4482 }, { "epoch": 0.85, "learning_rate": 2.7464442326024785e-06, "loss": 3.3233, "step": 4483 }, { "epoch": 0.85, "learning_rate": 2.7463870352716873e-06, "loss": 3.7073, "step": 4484 }, { "epoch": 0.85, "learning_rate": 2.746329837940896e-06, "loss": 2.8957, "step": 4485 }, { "epoch": 0.85, "learning_rate": 2.7462726406101053e-06, "loss": 3.6592, "step": 4486 }, { "epoch": 0.85, "learning_rate": 2.7462154432793136e-06, "loss": 3.6176, "step": 4487 }, { "epoch": 0.85, "learning_rate": 2.7461582459485224e-06, "loss": 3.1352, "step": 4488 }, { "epoch": 0.86, "learning_rate": 2.7461010486177312e-06, "loss": 3.2305, "step": 4489 }, { "epoch": 0.86, "learning_rate": 2.74604385128694e-06, "loss": 3.3959, "step": 4490 }, { "epoch": 0.86, "learning_rate": 2.745986653956149e-06, "loss": 3.44, "step": 4491 }, { "epoch": 0.86, "learning_rate": 2.7459294566253576e-06, "loss": 3.4589, "step": 4492 }, { "epoch": 0.86, "learning_rate": 2.7458722592945664e-06, "loss": 3.7642, "step": 4493 }, { "epoch": 0.86, "learning_rate": 2.7458150619637748e-06, "loss": 3.5434, "step": 4494 }, { "epoch": 0.86, "learning_rate": 2.745757864632984e-06, "loss": 3.4461, "step": 4495 }, { "epoch": 0.86, "learning_rate": 2.7457006673021928e-06, "loss": 3.4701, "step": 4496 }, { "epoch": 0.86, "learning_rate": 2.745643469971401e-06, "loss": 3.3228, "step": 4497 }, { "epoch": 0.86, "learning_rate": 2.7455862726406104e-06, "loss": 3.3321, "step": 4498 }, { "epoch": 0.86, "learning_rate": 2.7455290753098187e-06, "loss": 3.4687, "step": 4499 }, { "epoch": 0.86, "learning_rate": 2.745471877979028e-06, "loss": 3.0563, "step": 4500 }, { "epoch": 0.86, "learning_rate": 2.7454146806482363e-06, "loss": 3.5904, "step": 4501 }, { "epoch": 0.86, "learning_rate": 2.745357483317445e-06, "loss": 3.5204, "step": 4502 }, { "epoch": 0.86, "learning_rate": 2.7453002859866543e-06, "loss": 3.2438, "step": 4503 }, { "epoch": 0.86, "learning_rate": 2.7452430886558627e-06, "loss": 3.4796, "step": 4504 }, { "epoch": 0.86, "learning_rate": 2.7451858913250715e-06, "loss": 3.5209, "step": 4505 }, { "epoch": 0.86, "learning_rate": 2.7451286939942803e-06, "loss": 3.4725, "step": 4506 }, { "epoch": 0.86, "learning_rate": 2.745071496663489e-06, "loss": 3.7865, "step": 4507 }, { "epoch": 0.86, "learning_rate": 2.745014299332698e-06, "loss": 3.6789, "step": 4508 }, { "epoch": 0.86, "learning_rate": 2.7449571020019067e-06, "loss": 3.6719, "step": 4509 }, { "epoch": 0.86, "learning_rate": 2.7448999046711155e-06, "loss": 3.3716, "step": 4510 }, { "epoch": 0.86, "learning_rate": 2.7448427073403243e-06, "loss": 3.2725, "step": 4511 }, { "epoch": 0.86, "learning_rate": 2.744785510009533e-06, "loss": 3.1178, "step": 4512 }, { "epoch": 0.86, "learning_rate": 2.744728312678742e-06, "loss": 3.1557, "step": 4513 }, { "epoch": 0.86, "learning_rate": 2.7446711153479502e-06, "loss": 3.4528, "step": 4514 }, { "epoch": 0.86, "learning_rate": 2.7446139180171594e-06, "loss": 3.5615, "step": 4515 }, { "epoch": 0.86, "learning_rate": 2.744556720686368e-06, "loss": 3.1623, "step": 4516 }, { "epoch": 0.86, "learning_rate": 2.744499523355577e-06, "loss": 3.6759, "step": 4517 }, { "epoch": 0.86, "learning_rate": 2.744442326024786e-06, "loss": 3.5085, "step": 4518 }, { "epoch": 0.86, "learning_rate": 2.744385128693994e-06, "loss": 3.4607, "step": 4519 }, { "epoch": 0.86, "learning_rate": 2.7443279313632034e-06, "loss": 3.6871, "step": 4520 }, { "epoch": 0.86, "learning_rate": 2.7442707340324118e-06, "loss": 3.2028, "step": 4521 }, { "epoch": 0.86, "learning_rate": 2.7442135367016206e-06, "loss": 3.2709, "step": 4522 }, { "epoch": 0.86, "learning_rate": 2.7441563393708294e-06, "loss": 3.826, "step": 4523 }, { "epoch": 0.86, "learning_rate": 2.744099142040038e-06, "loss": 3.942, "step": 4524 }, { "epoch": 0.86, "learning_rate": 2.744041944709247e-06, "loss": 3.5156, "step": 4525 }, { "epoch": 0.86, "learning_rate": 2.7439847473784557e-06, "loss": 3.3517, "step": 4526 }, { "epoch": 0.86, "learning_rate": 2.7439275500476645e-06, "loss": 3.499, "step": 4527 }, { "epoch": 0.86, "learning_rate": 2.7438703527168733e-06, "loss": 3.7872, "step": 4528 }, { "epoch": 0.86, "learning_rate": 2.743813155386082e-06, "loss": 3.6295, "step": 4529 }, { "epoch": 0.86, "learning_rate": 2.743755958055291e-06, "loss": 2.9614, "step": 4530 }, { "epoch": 0.86, "learning_rate": 2.7436987607244993e-06, "loss": 3.239, "step": 4531 }, { "epoch": 0.86, "learning_rate": 2.7436415633937085e-06, "loss": 3.7861, "step": 4532 }, { "epoch": 0.86, "learning_rate": 2.743584366062917e-06, "loss": 3.4412, "step": 4533 }, { "epoch": 0.86, "learning_rate": 2.743527168732126e-06, "loss": 3.8384, "step": 4534 }, { "epoch": 0.86, "learning_rate": 2.743469971401335e-06, "loss": 2.8894, "step": 4535 }, { "epoch": 0.86, "learning_rate": 2.7434127740705433e-06, "loss": 3.8329, "step": 4536 }, { "epoch": 0.86, "learning_rate": 2.7433555767397525e-06, "loss": 2.9086, "step": 4537 }, { "epoch": 0.86, "learning_rate": 2.743298379408961e-06, "loss": 3.5513, "step": 4538 }, { "epoch": 0.86, "learning_rate": 2.7432411820781696e-06, "loss": 3.6351, "step": 4539 }, { "epoch": 0.86, "learning_rate": 2.7431839847473784e-06, "loss": 2.9157, "step": 4540 }, { "epoch": 0.86, "learning_rate": 2.7431267874165872e-06, "loss": 3.1611, "step": 4541 }, { "epoch": 0.87, "learning_rate": 2.743069590085796e-06, "loss": 3.2711, "step": 4542 }, { "epoch": 0.87, "learning_rate": 2.743012392755005e-06, "loss": 3.4288, "step": 4543 }, { "epoch": 0.87, "learning_rate": 2.7429551954242136e-06, "loss": 3.6086, "step": 4544 }, { "epoch": 0.87, "learning_rate": 2.7428979980934224e-06, "loss": 3.7902, "step": 4545 }, { "epoch": 0.87, "learning_rate": 2.742840800762631e-06, "loss": 3.3502, "step": 4546 }, { "epoch": 0.87, "learning_rate": 2.74278360343184e-06, "loss": 3.0359, "step": 4547 }, { "epoch": 0.87, "learning_rate": 2.7427264061010484e-06, "loss": 3.4184, "step": 4548 }, { "epoch": 0.87, "learning_rate": 2.7426692087702576e-06, "loss": 3.5896, "step": 4549 }, { "epoch": 0.87, "learning_rate": 2.742612011439466e-06, "loss": 2.8701, "step": 4550 }, { "epoch": 0.87, "learning_rate": 2.742554814108675e-06, "loss": 3.6829, "step": 4551 }, { "epoch": 0.87, "learning_rate": 2.742497616777884e-06, "loss": 3.1233, "step": 4552 }, { "epoch": 0.87, "learning_rate": 2.7424404194470923e-06, "loss": 3.9171, "step": 4553 }, { "epoch": 0.87, "learning_rate": 2.7423832221163016e-06, "loss": 3.0668, "step": 4554 }, { "epoch": 0.87, "learning_rate": 2.74232602478551e-06, "loss": 3.7304, "step": 4555 }, { "epoch": 0.87, "learning_rate": 2.7422688274547187e-06, "loss": 3.7456, "step": 4556 }, { "epoch": 0.87, "learning_rate": 2.742211630123928e-06, "loss": 3.1784, "step": 4557 }, { "epoch": 0.87, "learning_rate": 2.7421544327931363e-06, "loss": 3.6612, "step": 4558 }, { "epoch": 0.87, "learning_rate": 2.742097235462345e-06, "loss": 3.2536, "step": 4559 }, { "epoch": 0.87, "learning_rate": 2.742040038131554e-06, "loss": 3.3406, "step": 4560 }, { "epoch": 0.87, "learning_rate": 2.7419828408007627e-06, "loss": 3.4665, "step": 4561 }, { "epoch": 0.87, "learning_rate": 2.7419256434699715e-06, "loss": 3.7872, "step": 4562 }, { "epoch": 0.87, "learning_rate": 2.7418684461391803e-06, "loss": 3.7128, "step": 4563 }, { "epoch": 0.87, "learning_rate": 2.741811248808389e-06, "loss": 3.1212, "step": 4564 }, { "epoch": 0.87, "learning_rate": 2.741754051477598e-06, "loss": 3.6923, "step": 4565 }, { "epoch": 0.87, "learning_rate": 2.7416968541468067e-06, "loss": 3.4814, "step": 4566 }, { "epoch": 0.87, "learning_rate": 2.7416396568160154e-06, "loss": 3.232, "step": 4567 }, { "epoch": 0.87, "learning_rate": 2.7415824594852242e-06, "loss": 3.6381, "step": 4568 }, { "epoch": 0.87, "learning_rate": 2.741525262154433e-06, "loss": 3.1837, "step": 4569 }, { "epoch": 0.87, "learning_rate": 2.7414680648236414e-06, "loss": 2.9162, "step": 4570 }, { "epoch": 0.87, "learning_rate": 2.7414108674928506e-06, "loss": 3.1061, "step": 4571 }, { "epoch": 0.87, "learning_rate": 2.741353670162059e-06, "loss": 3.4674, "step": 4572 }, { "epoch": 0.87, "learning_rate": 2.7412964728312678e-06, "loss": 3.3577, "step": 4573 }, { "epoch": 0.87, "learning_rate": 2.741239275500477e-06, "loss": 3.646, "step": 4574 }, { "epoch": 0.87, "learning_rate": 2.7411820781696854e-06, "loss": 3.7967, "step": 4575 }, { "epoch": 0.87, "learning_rate": 2.741124880838894e-06, "loss": 3.4059, "step": 4576 }, { "epoch": 0.87, "learning_rate": 2.741067683508103e-06, "loss": 3.6006, "step": 4577 }, { "epoch": 0.87, "learning_rate": 2.7410104861773118e-06, "loss": 3.3825, "step": 4578 }, { "epoch": 0.87, "learning_rate": 2.7409532888465206e-06, "loss": 3.2782, "step": 4579 }, { "epoch": 0.87, "learning_rate": 2.7408960915157293e-06, "loss": 2.9467, "step": 4580 }, { "epoch": 0.87, "learning_rate": 2.740838894184938e-06, "loss": 3.4391, "step": 4581 }, { "epoch": 0.87, "learning_rate": 2.740781696854147e-06, "loss": 3.2938, "step": 4582 }, { "epoch": 0.87, "learning_rate": 2.7407244995233557e-06, "loss": 3.5357, "step": 4583 }, { "epoch": 0.87, "learning_rate": 2.7406673021925645e-06, "loss": 2.8164, "step": 4584 }, { "epoch": 0.87, "learning_rate": 2.7406101048617733e-06, "loss": 3.2016, "step": 4585 }, { "epoch": 0.87, "learning_rate": 2.740552907530982e-06, "loss": 3.5369, "step": 4586 }, { "epoch": 0.87, "learning_rate": 2.7404957102001905e-06, "loss": 3.603, "step": 4587 }, { "epoch": 0.87, "learning_rate": 2.7404385128693997e-06, "loss": 3.7368, "step": 4588 }, { "epoch": 0.87, "learning_rate": 2.740381315538608e-06, "loss": 3.8073, "step": 4589 }, { "epoch": 0.87, "learning_rate": 2.740324118207817e-06, "loss": 3.6241, "step": 4590 }, { "epoch": 0.87, "learning_rate": 2.740266920877026e-06, "loss": 3.0292, "step": 4591 }, { "epoch": 0.87, "learning_rate": 2.7402097235462344e-06, "loss": 3.0905, "step": 4592 }, { "epoch": 0.87, "learning_rate": 2.7401525262154432e-06, "loss": 3.4182, "step": 4593 }, { "epoch": 0.88, "learning_rate": 2.740095328884652e-06, "loss": 3.8454, "step": 4594 }, { "epoch": 0.88, "learning_rate": 2.740038131553861e-06, "loss": 3.1045, "step": 4595 }, { "epoch": 0.88, "learning_rate": 2.7399809342230696e-06, "loss": 3.3629, "step": 4596 }, { "epoch": 0.88, "learning_rate": 2.7399237368922784e-06, "loss": 3.2224, "step": 4597 }, { "epoch": 0.88, "learning_rate": 2.739866539561487e-06, "loss": 3.6538, "step": 4598 }, { "epoch": 0.88, "learning_rate": 2.739809342230696e-06, "loss": 3.4369, "step": 4599 }, { "epoch": 0.88, "learning_rate": 2.739752144899905e-06, "loss": 3.3932, "step": 4600 }, { "epoch": 0.88, "learning_rate": 2.7396949475691136e-06, "loss": 3.5926, "step": 4601 }, { "epoch": 0.88, "learning_rate": 2.7396377502383224e-06, "loss": 3.6537, "step": 4602 }, { "epoch": 0.88, "learning_rate": 2.739580552907531e-06, "loss": 3.4393, "step": 4603 }, { "epoch": 0.88, "learning_rate": 2.7395233555767395e-06, "loss": 3.8121, "step": 4604 }, { "epoch": 0.88, "learning_rate": 2.7394661582459488e-06, "loss": 2.6779, "step": 4605 }, { "epoch": 0.88, "learning_rate": 2.7394089609151576e-06, "loss": 3.6331, "step": 4606 }, { "epoch": 0.88, "learning_rate": 2.739351763584366e-06, "loss": 2.8709, "step": 4607 }, { "epoch": 0.88, "learning_rate": 2.739294566253575e-06, "loss": 3.1987, "step": 4608 }, { "epoch": 0.88, "learning_rate": 2.7392373689227835e-06, "loss": 3.2737, "step": 4609 }, { "epoch": 0.88, "learning_rate": 2.7391801715919923e-06, "loss": 3.2944, "step": 4610 }, { "epoch": 0.88, "learning_rate": 2.739122974261201e-06, "loss": 3.2596, "step": 4611 }, { "epoch": 0.88, "learning_rate": 2.73906577693041e-06, "loss": 3.6099, "step": 4612 }, { "epoch": 0.88, "learning_rate": 2.7390085795996187e-06, "loss": 3.5149, "step": 4613 }, { "epoch": 0.88, "learning_rate": 2.7389513822688275e-06, "loss": 2.8575, "step": 4614 }, { "epoch": 0.88, "learning_rate": 2.7388941849380363e-06, "loss": 3.5916, "step": 4615 }, { "epoch": 0.88, "learning_rate": 2.738836987607245e-06, "loss": 3.4374, "step": 4616 }, { "epoch": 0.88, "learning_rate": 2.738779790276454e-06, "loss": 3.4185, "step": 4617 }, { "epoch": 0.88, "learning_rate": 2.7387225929456627e-06, "loss": 2.8587, "step": 4618 }, { "epoch": 0.88, "learning_rate": 2.7386653956148715e-06, "loss": 4.1387, "step": 4619 }, { "epoch": 0.88, "learning_rate": 2.7386081982840803e-06, "loss": 3.537, "step": 4620 }, { "epoch": 0.88, "learning_rate": 2.7385510009532886e-06, "loss": 3.7254, "step": 4621 }, { "epoch": 0.88, "learning_rate": 2.738493803622498e-06, "loss": 3.1678, "step": 4622 }, { "epoch": 0.88, "learning_rate": 2.7384366062917066e-06, "loss": 3.3912, "step": 4623 }, { "epoch": 0.88, "learning_rate": 2.738379408960915e-06, "loss": 2.9731, "step": 4624 }, { "epoch": 0.88, "learning_rate": 2.7383222116301242e-06, "loss": 3.4567, "step": 4625 }, { "epoch": 0.88, "learning_rate": 2.7382650142993326e-06, "loss": 3.196, "step": 4626 }, { "epoch": 0.88, "learning_rate": 2.7382078169685414e-06, "loss": 3.3131, "step": 4627 }, { "epoch": 0.88, "learning_rate": 2.73815061963775e-06, "loss": 3.6756, "step": 4628 }, { "epoch": 0.88, "learning_rate": 2.738093422306959e-06, "loss": 3.0776, "step": 4629 }, { "epoch": 0.88, "learning_rate": 2.738036224976168e-06, "loss": 3.53, "step": 4630 }, { "epoch": 0.88, "learning_rate": 2.7379790276453766e-06, "loss": 3.6167, "step": 4631 }, { "epoch": 0.88, "learning_rate": 2.7379218303145854e-06, "loss": 3.3478, "step": 4632 }, { "epoch": 0.88, "learning_rate": 2.737864632983794e-06, "loss": 3.4693, "step": 4633 }, { "epoch": 0.88, "learning_rate": 2.737807435653003e-06, "loss": 3.4226, "step": 4634 }, { "epoch": 0.88, "learning_rate": 2.7377502383222117e-06, "loss": 3.3009, "step": 4635 }, { "epoch": 0.88, "learning_rate": 2.7376930409914205e-06, "loss": 3.4379, "step": 4636 }, { "epoch": 0.88, "learning_rate": 2.7376358436606293e-06, "loss": 3.2386, "step": 4637 }, { "epoch": 0.88, "learning_rate": 2.7375786463298377e-06, "loss": 3.2236, "step": 4638 }, { "epoch": 0.88, "learning_rate": 2.737521448999047e-06, "loss": 3.538, "step": 4639 }, { "epoch": 0.88, "learning_rate": 2.7374642516682557e-06, "loss": 3.7376, "step": 4640 }, { "epoch": 0.88, "learning_rate": 2.737407054337464e-06, "loss": 3.3063, "step": 4641 }, { "epoch": 0.88, "learning_rate": 2.7373498570066733e-06, "loss": 3.488, "step": 4642 }, { "epoch": 0.88, "learning_rate": 2.7372926596758817e-06, "loss": 3.6451, "step": 4643 }, { "epoch": 0.88, "learning_rate": 2.7372354623450905e-06, "loss": 3.5554, "step": 4644 }, { "epoch": 0.88, "learning_rate": 2.7371782650142997e-06, "loss": 3.39, "step": 4645 }, { "epoch": 0.88, "learning_rate": 2.737121067683508e-06, "loss": 3.7765, "step": 4646 }, { "epoch": 0.89, "learning_rate": 2.7370638703527173e-06, "loss": 3.2697, "step": 4647 }, { "epoch": 0.89, "learning_rate": 2.7370066730219256e-06, "loss": 3.2199, "step": 4648 }, { "epoch": 0.89, "learning_rate": 2.7369494756911344e-06, "loss": 4.064, "step": 4649 }, { "epoch": 0.89, "learning_rate": 2.7368922783603432e-06, "loss": 3.4239, "step": 4650 }, { "epoch": 0.89, "learning_rate": 2.736835081029552e-06, "loss": 2.9356, "step": 4651 }, { "epoch": 0.89, "learning_rate": 2.736777883698761e-06, "loss": 3.1131, "step": 4652 }, { "epoch": 0.89, "learning_rate": 2.7367206863679696e-06, "loss": 3.3003, "step": 4653 }, { "epoch": 0.89, "learning_rate": 2.7366634890371784e-06, "loss": 3.6911, "step": 4654 }, { "epoch": 0.89, "learning_rate": 2.736606291706387e-06, "loss": 3.1917, "step": 4655 }, { "epoch": 0.89, "learning_rate": 2.736549094375596e-06, "loss": 3.325, "step": 4656 }, { "epoch": 0.89, "learning_rate": 2.7364918970448048e-06, "loss": 3.6125, "step": 4657 }, { "epoch": 0.89, "learning_rate": 2.736434699714013e-06, "loss": 3.6002, "step": 4658 }, { "epoch": 0.89, "learning_rate": 2.7363775023832224e-06, "loss": 3.2099, "step": 4659 }, { "epoch": 0.89, "learning_rate": 2.7363203050524307e-06, "loss": 3.6195, "step": 4660 }, { "epoch": 0.89, "learning_rate": 2.7362631077216395e-06, "loss": 3.5492, "step": 4661 }, { "epoch": 0.89, "learning_rate": 2.7362059103908487e-06, "loss": 3.4451, "step": 4662 }, { "epoch": 0.89, "learning_rate": 2.736148713060057e-06, "loss": 3.1487, "step": 4663 }, { "epoch": 0.89, "learning_rate": 2.7360915157292663e-06, "loss": 3.2984, "step": 4664 }, { "epoch": 0.89, "learning_rate": 2.7360343183984747e-06, "loss": 3.7718, "step": 4665 }, { "epoch": 0.89, "learning_rate": 2.7359771210676835e-06, "loss": 3.1356, "step": 4666 }, { "epoch": 0.89, "learning_rate": 2.7359199237368923e-06, "loss": 3.764, "step": 4667 }, { "epoch": 0.89, "learning_rate": 2.735862726406101e-06, "loss": 3.4604, "step": 4668 }, { "epoch": 0.89, "learning_rate": 2.73580552907531e-06, "loss": 3.2307, "step": 4669 }, { "epoch": 0.89, "learning_rate": 2.7357483317445187e-06, "loss": 3.5432, "step": 4670 }, { "epoch": 0.89, "learning_rate": 2.7356911344137275e-06, "loss": 3.2912, "step": 4671 }, { "epoch": 0.89, "learning_rate": 2.7356339370829363e-06, "loss": 3.3021, "step": 4672 }, { "epoch": 0.89, "learning_rate": 2.735576739752145e-06, "loss": 3.2838, "step": 4673 }, { "epoch": 0.89, "learning_rate": 2.735519542421354e-06, "loss": 3.2196, "step": 4674 }, { "epoch": 0.89, "learning_rate": 2.7354623450905622e-06, "loss": 3.4031, "step": 4675 }, { "epoch": 0.89, "learning_rate": 2.7354051477597714e-06, "loss": 3.5422, "step": 4676 }, { "epoch": 0.89, "learning_rate": 2.73534795042898e-06, "loss": 3.4344, "step": 4677 }, { "epoch": 0.89, "learning_rate": 2.7352907530981886e-06, "loss": 3.3761, "step": 4678 }, { "epoch": 0.89, "learning_rate": 2.735233555767398e-06, "loss": 3.5436, "step": 4679 }, { "epoch": 0.89, "learning_rate": 2.735176358436606e-06, "loss": 2.8825, "step": 4680 }, { "epoch": 0.89, "learning_rate": 2.7351191611058154e-06, "loss": 3.8002, "step": 4681 }, { "epoch": 0.89, "learning_rate": 2.7350619637750238e-06, "loss": 3.4182, "step": 4682 }, { "epoch": 0.89, "learning_rate": 2.7350047664442326e-06, "loss": 3.5543, "step": 4683 }, { "epoch": 0.89, "learning_rate": 2.7349475691134414e-06, "loss": 3.4634, "step": 4684 }, { "epoch": 0.89, "learning_rate": 2.73489037178265e-06, "loss": 3.751, "step": 4685 }, { "epoch": 0.89, "learning_rate": 2.734833174451859e-06, "loss": 3.1142, "step": 4686 }, { "epoch": 0.89, "learning_rate": 2.7347759771210677e-06, "loss": 3.8126, "step": 4687 }, { "epoch": 0.89, "learning_rate": 2.7347187797902765e-06, "loss": 3.3868, "step": 4688 }, { "epoch": 0.89, "learning_rate": 2.7346615824594853e-06, "loss": 2.9033, "step": 4689 }, { "epoch": 0.89, "learning_rate": 2.734604385128694e-06, "loss": 3.5904, "step": 4690 }, { "epoch": 0.89, "learning_rate": 2.734547187797903e-06, "loss": 3.4045, "step": 4691 }, { "epoch": 0.89, "learning_rate": 2.7344899904671113e-06, "loss": 3.6066, "step": 4692 }, { "epoch": 0.89, "learning_rate": 2.7344327931363205e-06, "loss": 3.585, "step": 4693 }, { "epoch": 0.89, "learning_rate": 2.7343755958055293e-06, "loss": 3.4209, "step": 4694 }, { "epoch": 0.89, "learning_rate": 2.734318398474738e-06, "loss": 3.8496, "step": 4695 }, { "epoch": 0.89, "learning_rate": 2.734261201143947e-06, "loss": 3.2878, "step": 4696 }, { "epoch": 0.89, "learning_rate": 2.7342040038131553e-06, "loss": 3.2194, "step": 4697 }, { "epoch": 0.89, "learning_rate": 2.7341468064823645e-06, "loss": 3.431, "step": 4698 }, { "epoch": 0.9, "learning_rate": 2.734089609151573e-06, "loss": 3.2457, "step": 4699 }, { "epoch": 0.9, "learning_rate": 2.7340324118207816e-06, "loss": 2.9408, "step": 4700 }, { "epoch": 0.9, "learning_rate": 2.733975214489991e-06, "loss": 3.4187, "step": 4701 }, { "epoch": 0.9, "learning_rate": 2.7339180171591992e-06, "loss": 3.4337, "step": 4702 }, { "epoch": 0.9, "learning_rate": 2.733860819828408e-06, "loss": 3.615, "step": 4703 }, { "epoch": 0.9, "learning_rate": 2.733803622497617e-06, "loss": 3.5986, "step": 4704 }, { "epoch": 0.9, "learning_rate": 2.7337464251668256e-06, "loss": 3.363, "step": 4705 }, { "epoch": 0.9, "learning_rate": 2.7336892278360344e-06, "loss": 2.8382, "step": 4706 }, { "epoch": 0.9, "learning_rate": 2.733632030505243e-06, "loss": 3.4416, "step": 4707 }, { "epoch": 0.9, "learning_rate": 2.733574833174452e-06, "loss": 3.2745, "step": 4708 }, { "epoch": 0.9, "learning_rate": 2.7335176358436604e-06, "loss": 3.0138, "step": 4709 }, { "epoch": 0.9, "learning_rate": 2.7334604385128696e-06, "loss": 3.2052, "step": 4710 }, { "epoch": 0.9, "learning_rate": 2.7334032411820784e-06, "loss": 3.3766, "step": 4711 }, { "epoch": 0.9, "learning_rate": 2.733346043851287e-06, "loss": 3.6155, "step": 4712 }, { "epoch": 0.9, "learning_rate": 2.733288846520496e-06, "loss": 2.7969, "step": 4713 }, { "epoch": 0.9, "learning_rate": 2.7332316491897043e-06, "loss": 3.7012, "step": 4714 }, { "epoch": 0.9, "learning_rate": 2.7331744518589135e-06, "loss": 3.3701, "step": 4715 }, { "epoch": 0.9, "learning_rate": 2.733117254528122e-06, "loss": 3.0114, "step": 4716 }, { "epoch": 0.9, "learning_rate": 2.7330600571973307e-06, "loss": 3.413, "step": 4717 }, { "epoch": 0.9, "learning_rate": 2.73300285986654e-06, "loss": 3.4737, "step": 4718 }, { "epoch": 0.9, "learning_rate": 2.7329456625357483e-06, "loss": 3.9742, "step": 4719 }, { "epoch": 0.9, "learning_rate": 2.732888465204957e-06, "loss": 3.165, "step": 4720 }, { "epoch": 0.9, "learning_rate": 2.732831267874166e-06, "loss": 3.625, "step": 4721 }, { "epoch": 0.9, "learning_rate": 2.7327740705433747e-06, "loss": 3.3123, "step": 4722 }, { "epoch": 0.9, "learning_rate": 2.7327168732125835e-06, "loss": 3.4742, "step": 4723 }, { "epoch": 0.9, "learning_rate": 2.7326596758817923e-06, "loss": 3.7143, "step": 4724 }, { "epoch": 0.9, "learning_rate": 2.732602478551001e-06, "loss": 3.4791, "step": 4725 }, { "epoch": 0.9, "learning_rate": 2.7325452812202094e-06, "loss": 3.7669, "step": 4726 }, { "epoch": 0.9, "learning_rate": 2.7324880838894186e-06, "loss": 2.9763, "step": 4727 }, { "epoch": 0.9, "learning_rate": 2.7324308865586274e-06, "loss": 3.4744, "step": 4728 }, { "epoch": 0.9, "learning_rate": 2.7323736892278362e-06, "loss": 3.0395, "step": 4729 }, { "epoch": 0.9, "learning_rate": 2.732316491897045e-06, "loss": 3.2703, "step": 4730 }, { "epoch": 0.9, "learning_rate": 2.7322592945662534e-06, "loss": 2.9113, "step": 4731 }, { "epoch": 0.9, "learning_rate": 2.7322020972354626e-06, "loss": 2.9852, "step": 4732 }, { "epoch": 0.9, "learning_rate": 2.7321448999046714e-06, "loss": 3.6387, "step": 4733 }, { "epoch": 0.9, "learning_rate": 2.7320877025738798e-06, "loss": 3.09, "step": 4734 }, { "epoch": 0.9, "learning_rate": 2.732030505243089e-06, "loss": 3.4654, "step": 4735 }, { "epoch": 0.9, "learning_rate": 2.7319733079122974e-06, "loss": 3.8028, "step": 4736 }, { "epoch": 0.9, "learning_rate": 2.731916110581506e-06, "loss": 3.2851, "step": 4737 }, { "epoch": 0.9, "learning_rate": 2.731858913250715e-06, "loss": 3.1256, "step": 4738 }, { "epoch": 0.9, "learning_rate": 2.7318017159199237e-06, "loss": 2.8726, "step": 4739 }, { "epoch": 0.9, "learning_rate": 2.7317445185891325e-06, "loss": 3.2761, "step": 4740 }, { "epoch": 0.9, "learning_rate": 2.7316873212583413e-06, "loss": 3.6808, "step": 4741 }, { "epoch": 0.9, "learning_rate": 2.73163012392755e-06, "loss": 2.8432, "step": 4742 }, { "epoch": 0.9, "learning_rate": 2.731572926596759e-06, "loss": 3.2722, "step": 4743 }, { "epoch": 0.9, "learning_rate": 2.7315157292659677e-06, "loss": 3.7988, "step": 4744 }, { "epoch": 0.9, "learning_rate": 2.7314585319351765e-06, "loss": 3.509, "step": 4745 }, { "epoch": 0.9, "learning_rate": 2.7314013346043853e-06, "loss": 3.2893, "step": 4746 }, { "epoch": 0.9, "learning_rate": 2.731344137273594e-06, "loss": 3.6261, "step": 4747 }, { "epoch": 0.9, "learning_rate": 2.7312869399428025e-06, "loss": 3.7276, "step": 4748 }, { "epoch": 0.9, "learning_rate": 2.7312297426120117e-06, "loss": 3.6212, "step": 4749 }, { "epoch": 0.9, "learning_rate": 2.7311725452812205e-06, "loss": 3.2306, "step": 4750 }, { "epoch": 0.9, "learning_rate": 2.731115347950429e-06, "loss": 3.5379, "step": 4751 }, { "epoch": 0.91, "learning_rate": 2.731058150619638e-06, "loss": 3.0792, "step": 4752 }, { "epoch": 0.91, "learning_rate": 2.7310009532888464e-06, "loss": 3.8098, "step": 4753 }, { "epoch": 0.91, "learning_rate": 2.7309437559580552e-06, "loss": 3.3024, "step": 4754 }, { "epoch": 0.91, "learning_rate": 2.730886558627264e-06, "loss": 3.2685, "step": 4755 }, { "epoch": 0.91, "learning_rate": 2.730829361296473e-06, "loss": 3.4101, "step": 4756 }, { "epoch": 0.91, "learning_rate": 2.7307721639656816e-06, "loss": 3.5229, "step": 4757 }, { "epoch": 0.91, "learning_rate": 2.7307149666348904e-06, "loss": 3.6851, "step": 4758 }, { "epoch": 0.91, "learning_rate": 2.730657769304099e-06, "loss": 3.0251, "step": 4759 }, { "epoch": 0.91, "learning_rate": 2.730600571973308e-06, "loss": 3.6728, "step": 4760 }, { "epoch": 0.91, "learning_rate": 2.730543374642517e-06, "loss": 2.8, "step": 4761 }, { "epoch": 0.91, "learning_rate": 2.7304861773117256e-06, "loss": 3.8421, "step": 4762 }, { "epoch": 0.91, "learning_rate": 2.7304289799809344e-06, "loss": 3.9274, "step": 4763 }, { "epoch": 0.91, "learning_rate": 2.730371782650143e-06, "loss": 3.2846, "step": 4764 }, { "epoch": 0.91, "learning_rate": 2.7303145853193515e-06, "loss": 4.0054, "step": 4765 }, { "epoch": 0.91, "learning_rate": 2.7302573879885608e-06, "loss": 2.8411, "step": 4766 }, { "epoch": 0.91, "learning_rate": 2.7302001906577696e-06, "loss": 3.363, "step": 4767 }, { "epoch": 0.91, "learning_rate": 2.730142993326978e-06, "loss": 3.7931, "step": 4768 }, { "epoch": 0.91, "learning_rate": 2.730085795996187e-06, "loss": 3.3751, "step": 4769 }, { "epoch": 0.91, "learning_rate": 2.7300285986653955e-06, "loss": 3.2788, "step": 4770 }, { "epoch": 0.91, "learning_rate": 2.7299714013346043e-06, "loss": 3.4575, "step": 4771 }, { "epoch": 0.91, "learning_rate": 2.729914204003813e-06, "loss": 3.1607, "step": 4772 }, { "epoch": 0.91, "learning_rate": 2.729857006673022e-06, "loss": 3.0376, "step": 4773 }, { "epoch": 0.91, "learning_rate": 2.7297998093422307e-06, "loss": 3.5553, "step": 4774 }, { "epoch": 0.91, "learning_rate": 2.7297426120114395e-06, "loss": 2.7105, "step": 4775 }, { "epoch": 0.91, "learning_rate": 2.7296854146806483e-06, "loss": 3.3005, "step": 4776 }, { "epoch": 0.91, "learning_rate": 2.729628217349857e-06, "loss": 3.0634, "step": 4777 }, { "epoch": 0.91, "learning_rate": 2.729571020019066e-06, "loss": 3.1801, "step": 4778 }, { "epoch": 0.91, "learning_rate": 2.7295138226882747e-06, "loss": 3.1512, "step": 4779 }, { "epoch": 0.91, "learning_rate": 2.7294566253574834e-06, "loss": 3.5538, "step": 4780 }, { "epoch": 0.91, "learning_rate": 2.7293994280266922e-06, "loss": 2.8523, "step": 4781 }, { "epoch": 0.91, "learning_rate": 2.729342230695901e-06, "loss": 3.276, "step": 4782 }, { "epoch": 0.91, "learning_rate": 2.72928503336511e-06, "loss": 2.9062, "step": 4783 }, { "epoch": 0.91, "learning_rate": 2.7292278360343186e-06, "loss": 3.0393, "step": 4784 }, { "epoch": 0.91, "learning_rate": 2.729170638703527e-06, "loss": 3.5937, "step": 4785 }, { "epoch": 0.91, "learning_rate": 2.7291134413727362e-06, "loss": 3.7709, "step": 4786 }, { "epoch": 0.91, "learning_rate": 2.7290562440419446e-06, "loss": 3.7766, "step": 4787 }, { "epoch": 0.91, "learning_rate": 2.7289990467111534e-06, "loss": 3.4166, "step": 4788 }, { "epoch": 0.91, "learning_rate": 2.7289418493803626e-06, "loss": 3.5136, "step": 4789 }, { "epoch": 0.91, "learning_rate": 2.728884652049571e-06, "loss": 3.5743, "step": 4790 }, { "epoch": 0.91, "learning_rate": 2.7288274547187798e-06, "loss": 3.6606, "step": 4791 }, { "epoch": 0.91, "learning_rate": 2.7287702573879886e-06, "loss": 3.4705, "step": 4792 }, { "epoch": 0.91, "learning_rate": 2.7287130600571973e-06, "loss": 3.2287, "step": 4793 }, { "epoch": 0.91, "learning_rate": 2.728655862726406e-06, "loss": 3.818, "step": 4794 }, { "epoch": 0.91, "learning_rate": 2.728598665395615e-06, "loss": 3.7523, "step": 4795 }, { "epoch": 0.91, "learning_rate": 2.7285414680648237e-06, "loss": 3.3646, "step": 4796 }, { "epoch": 0.91, "learning_rate": 2.7284842707340325e-06, "loss": 3.3031, "step": 4797 }, { "epoch": 0.91, "learning_rate": 2.7284270734032413e-06, "loss": 2.8608, "step": 4798 }, { "epoch": 0.91, "learning_rate": 2.72836987607245e-06, "loss": 3.1712, "step": 4799 }, { "epoch": 0.91, "learning_rate": 2.728312678741659e-06, "loss": 3.5168, "step": 4800 }, { "epoch": 0.91, "learning_rate": 2.7282554814108677e-06, "loss": 3.4211, "step": 4801 }, { "epoch": 0.91, "learning_rate": 2.728198284080076e-06, "loss": 3.0631, "step": 4802 }, { "epoch": 0.91, "learning_rate": 2.7281410867492853e-06, "loss": 3.7294, "step": 4803 }, { "epoch": 0.92, "learning_rate": 2.7280838894184937e-06, "loss": 3.1035, "step": 4804 }, { "epoch": 0.92, "learning_rate": 2.7280266920877024e-06, "loss": 3.6289, "step": 4805 }, { "epoch": 0.92, "learning_rate": 2.7279694947569117e-06, "loss": 3.8186, "step": 4806 }, { "epoch": 0.92, "learning_rate": 2.72791229742612e-06, "loss": 3.723, "step": 4807 }, { "epoch": 0.92, "learning_rate": 2.727855100095329e-06, "loss": 4.0047, "step": 4808 }, { "epoch": 0.92, "learning_rate": 2.7277979027645376e-06, "loss": 3.4744, "step": 4809 }, { "epoch": 0.92, "learning_rate": 2.7277407054337464e-06, "loss": 2.9943, "step": 4810 }, { "epoch": 0.92, "learning_rate": 2.727683508102955e-06, "loss": 3.5152, "step": 4811 }, { "epoch": 0.92, "learning_rate": 2.727626310772164e-06, "loss": 3.5136, "step": 4812 }, { "epoch": 0.92, "learning_rate": 2.727569113441373e-06, "loss": 3.0293, "step": 4813 }, { "epoch": 0.92, "learning_rate": 2.7275119161105816e-06, "loss": 2.6836, "step": 4814 }, { "epoch": 0.92, "learning_rate": 2.7274547187797904e-06, "loss": 3.0462, "step": 4815 }, { "epoch": 0.92, "learning_rate": 2.727397521448999e-06, "loss": 3.4595, "step": 4816 }, { "epoch": 0.92, "learning_rate": 2.727340324118208e-06, "loss": 3.5165, "step": 4817 }, { "epoch": 0.92, "learning_rate": 2.7272831267874168e-06, "loss": 2.6689, "step": 4818 }, { "epoch": 0.92, "learning_rate": 2.727225929456625e-06, "loss": 3.5003, "step": 4819 }, { "epoch": 0.92, "learning_rate": 2.7271687321258344e-06, "loss": 3.2884, "step": 4820 }, { "epoch": 0.92, "learning_rate": 2.727111534795043e-06, "loss": 3.3616, "step": 4821 }, { "epoch": 0.92, "learning_rate": 2.7270543374642515e-06, "loss": 3.3442, "step": 4822 }, { "epoch": 0.92, "learning_rate": 2.7269971401334607e-06, "loss": 3.2255, "step": 4823 }, { "epoch": 0.92, "learning_rate": 2.726939942802669e-06, "loss": 3.2423, "step": 4824 }, { "epoch": 0.92, "learning_rate": 2.7268827454718783e-06, "loss": 3.3376, "step": 4825 }, { "epoch": 0.92, "learning_rate": 2.7268255481410867e-06, "loss": 3.3491, "step": 4826 }, { "epoch": 0.92, "learning_rate": 2.7267683508102955e-06, "loss": 3.5879, "step": 4827 }, { "epoch": 0.92, "learning_rate": 2.7267111534795047e-06, "loss": 3.8654, "step": 4828 }, { "epoch": 0.92, "learning_rate": 2.726653956148713e-06, "loss": 3.3757, "step": 4829 }, { "epoch": 0.92, "learning_rate": 2.726596758817922e-06, "loss": 3.9112, "step": 4830 }, { "epoch": 0.92, "learning_rate": 2.7265395614871307e-06, "loss": 3.4925, "step": 4831 }, { "epoch": 0.92, "learning_rate": 2.7264823641563395e-06, "loss": 3.0236, "step": 4832 }, { "epoch": 0.92, "learning_rate": 2.7264251668255483e-06, "loss": 3.6047, "step": 4833 }, { "epoch": 0.92, "learning_rate": 2.726367969494757e-06, "loss": 3.4408, "step": 4834 }, { "epoch": 0.92, "learning_rate": 2.726310772163966e-06, "loss": 3.7281, "step": 4835 }, { "epoch": 0.92, "learning_rate": 2.726253574833174e-06, "loss": 3.5046, "step": 4836 }, { "epoch": 0.92, "learning_rate": 2.7261963775023834e-06, "loss": 3.6467, "step": 4837 }, { "epoch": 0.92, "learning_rate": 2.7261391801715922e-06, "loss": 3.4333, "step": 4838 }, { "epoch": 0.92, "learning_rate": 2.7260819828408006e-06, "loss": 3.4273, "step": 4839 }, { "epoch": 0.92, "learning_rate": 2.72602478551001e-06, "loss": 3.2508, "step": 4840 }, { "epoch": 0.92, "learning_rate": 2.725967588179218e-06, "loss": 3.7306, "step": 4841 }, { "epoch": 0.92, "learning_rate": 2.7259103908484274e-06, "loss": 3.7413, "step": 4842 }, { "epoch": 0.92, "learning_rate": 2.7258531935176358e-06, "loss": 3.4405, "step": 4843 }, { "epoch": 0.92, "learning_rate": 2.7257959961868446e-06, "loss": 3.0625, "step": 4844 }, { "epoch": 0.92, "learning_rate": 2.7257387988560538e-06, "loss": 2.9361, "step": 4845 }, { "epoch": 0.92, "learning_rate": 2.725681601525262e-06, "loss": 3.3475, "step": 4846 }, { "epoch": 0.92, "learning_rate": 2.725624404194471e-06, "loss": 3.0867, "step": 4847 }, { "epoch": 0.92, "learning_rate": 2.7255672068636797e-06, "loss": 3.6782, "step": 4848 }, { "epoch": 0.92, "learning_rate": 2.7255100095328885e-06, "loss": 3.236, "step": 4849 }, { "epoch": 0.92, "learning_rate": 2.7254528122020973e-06, "loss": 3.5056, "step": 4850 }, { "epoch": 0.92, "learning_rate": 2.725395614871306e-06, "loss": 3.2425, "step": 4851 }, { "epoch": 0.92, "learning_rate": 2.725338417540515e-06, "loss": 3.1411, "step": 4852 }, { "epoch": 0.92, "learning_rate": 2.7252812202097233e-06, "loss": 3.3431, "step": 4853 }, { "epoch": 0.92, "learning_rate": 2.7252240228789325e-06, "loss": 3.4983, "step": 4854 }, { "epoch": 0.92, "learning_rate": 2.7251668255481413e-06, "loss": 3.1332, "step": 4855 }, { "epoch": 0.92, "learning_rate": 2.7251096282173497e-06, "loss": 3.3273, "step": 4856 }, { "epoch": 0.93, "learning_rate": 2.725052430886559e-06, "loss": 3.2858, "step": 4857 }, { "epoch": 0.93, "learning_rate": 2.7249952335557672e-06, "loss": 2.8601, "step": 4858 }, { "epoch": 0.93, "learning_rate": 2.7249380362249765e-06, "loss": 3.395, "step": 4859 }, { "epoch": 0.93, "learning_rate": 2.724880838894185e-06, "loss": 3.2686, "step": 4860 }, { "epoch": 0.93, "learning_rate": 2.7248236415633936e-06, "loss": 3.4927, "step": 4861 }, { "epoch": 0.93, "learning_rate": 2.724766444232603e-06, "loss": 3.4364, "step": 4862 }, { "epoch": 0.93, "learning_rate": 2.7247092469018112e-06, "loss": 3.2601, "step": 4863 }, { "epoch": 0.93, "learning_rate": 2.72465204957102e-06, "loss": 3.374, "step": 4864 }, { "epoch": 0.93, "learning_rate": 2.724594852240229e-06, "loss": 3.5288, "step": 4865 }, { "epoch": 0.93, "learning_rate": 2.7245376549094376e-06, "loss": 3.3817, "step": 4866 }, { "epoch": 0.93, "learning_rate": 2.7244804575786464e-06, "loss": 3.2057, "step": 4867 }, { "epoch": 0.93, "learning_rate": 2.724423260247855e-06, "loss": 3.3596, "step": 4868 }, { "epoch": 0.93, "learning_rate": 2.724366062917064e-06, "loss": 4.1569, "step": 4869 }, { "epoch": 0.93, "learning_rate": 2.7243088655862728e-06, "loss": 3.5192, "step": 4870 }, { "epoch": 0.93, "learning_rate": 2.7242516682554816e-06, "loss": 3.1665, "step": 4871 }, { "epoch": 0.93, "learning_rate": 2.7241944709246904e-06, "loss": 2.8103, "step": 4872 }, { "epoch": 0.93, "learning_rate": 2.7241372735938987e-06, "loss": 3.4283, "step": 4873 }, { "epoch": 0.93, "learning_rate": 2.724080076263108e-06, "loss": 3.3999, "step": 4874 }, { "epoch": 0.93, "learning_rate": 2.7240228789323163e-06, "loss": 2.9094, "step": 4875 }, { "epoch": 0.93, "learning_rate": 2.7239656816015255e-06, "loss": 3.2086, "step": 4876 }, { "epoch": 0.93, "learning_rate": 2.7239084842707343e-06, "loss": 3.3183, "step": 4877 }, { "epoch": 0.93, "learning_rate": 2.7238512869399427e-06, "loss": 3.7662, "step": 4878 }, { "epoch": 0.93, "learning_rate": 2.723794089609152e-06, "loss": 3.6205, "step": 4879 }, { "epoch": 0.93, "learning_rate": 2.7237368922783603e-06, "loss": 3.0694, "step": 4880 }, { "epoch": 0.93, "learning_rate": 2.723679694947569e-06, "loss": 3.2579, "step": 4881 }, { "epoch": 0.93, "learning_rate": 2.723622497616778e-06, "loss": 3.631, "step": 4882 }, { "epoch": 0.93, "learning_rate": 2.7235653002859867e-06, "loss": 3.6208, "step": 4883 }, { "epoch": 0.93, "learning_rate": 2.7235081029551955e-06, "loss": 3.2364, "step": 4884 }, { "epoch": 0.93, "learning_rate": 2.7234509056244043e-06, "loss": 3.4962, "step": 4885 }, { "epoch": 0.93, "learning_rate": 2.723393708293613e-06, "loss": 3.137, "step": 4886 }, { "epoch": 0.93, "learning_rate": 2.723336510962822e-06, "loss": 3.1823, "step": 4887 }, { "epoch": 0.93, "learning_rate": 2.7232793136320306e-06, "loss": 3.0286, "step": 4888 }, { "epoch": 0.93, "learning_rate": 2.7232221163012394e-06, "loss": 3.5445, "step": 4889 }, { "epoch": 0.93, "learning_rate": 2.7231649189704482e-06, "loss": 3.4694, "step": 4890 }, { "epoch": 0.93, "learning_rate": 2.723107721639657e-06, "loss": 2.9363, "step": 4891 }, { "epoch": 0.93, "learning_rate": 2.7230505243088654e-06, "loss": 3.1213, "step": 4892 }, { "epoch": 0.93, "learning_rate": 2.7229933269780746e-06, "loss": 3.6062, "step": 4893 }, { "epoch": 0.93, "learning_rate": 2.7229361296472834e-06, "loss": 3.5434, "step": 4894 }, { "epoch": 0.93, "learning_rate": 2.7228789323164918e-06, "loss": 3.3909, "step": 4895 }, { "epoch": 0.93, "learning_rate": 2.722821734985701e-06, "loss": 3.2182, "step": 4896 }, { "epoch": 0.93, "learning_rate": 2.7227645376549094e-06, "loss": 3.5436, "step": 4897 }, { "epoch": 0.93, "learning_rate": 2.722707340324118e-06, "loss": 3.7542, "step": 4898 }, { "epoch": 0.93, "learning_rate": 2.722650142993327e-06, "loss": 3.3544, "step": 4899 }, { "epoch": 0.93, "learning_rate": 2.7225929456625357e-06, "loss": 3.8498, "step": 4900 }, { "epoch": 0.93, "learning_rate": 2.7225357483317445e-06, "loss": 3.5226, "step": 4901 }, { "epoch": 0.93, "learning_rate": 2.7224785510009533e-06, "loss": 3.7573, "step": 4902 }, { "epoch": 0.93, "learning_rate": 2.722421353670162e-06, "loss": 3.8045, "step": 4903 }, { "epoch": 0.93, "learning_rate": 2.722364156339371e-06, "loss": 3.4707, "step": 4904 }, { "epoch": 0.93, "learning_rate": 2.7223069590085797e-06, "loss": 2.8163, "step": 4905 }, { "epoch": 0.93, "learning_rate": 2.7222497616777885e-06, "loss": 3.6956, "step": 4906 }, { "epoch": 0.93, "learning_rate": 2.7221925643469973e-06, "loss": 2.9911, "step": 4907 }, { "epoch": 0.93, "learning_rate": 2.722135367016206e-06, "loss": 3.3798, "step": 4908 }, { "epoch": 0.94, "learning_rate": 2.7220781696854145e-06, "loss": 3.6889, "step": 4909 }, { "epoch": 0.94, "learning_rate": 2.7220209723546237e-06, "loss": 3.6946, "step": 4910 }, { "epoch": 0.94, "learning_rate": 2.7219637750238325e-06, "loss": 3.12, "step": 4911 }, { "epoch": 0.94, "learning_rate": 2.721906577693041e-06, "loss": 3.4502, "step": 4912 }, { "epoch": 0.94, "learning_rate": 2.72184938036225e-06, "loss": 3.3379, "step": 4913 }, { "epoch": 0.94, "learning_rate": 2.7217921830314584e-06, "loss": 3.4313, "step": 4914 }, { "epoch": 0.94, "learning_rate": 2.7217349857006672e-06, "loss": 3.4441, "step": 4915 }, { "epoch": 0.94, "learning_rate": 2.7216777883698764e-06, "loss": 3.0863, "step": 4916 }, { "epoch": 0.94, "learning_rate": 2.721620591039085e-06, "loss": 2.826, "step": 4917 }, { "epoch": 0.94, "learning_rate": 2.7215633937082936e-06, "loss": 3.3588, "step": 4918 }, { "epoch": 0.94, "learning_rate": 2.7215061963775024e-06, "loss": 3.3532, "step": 4919 }, { "epoch": 0.94, "learning_rate": 2.721448999046711e-06, "loss": 3.1315, "step": 4920 }, { "epoch": 0.94, "learning_rate": 2.72139180171592e-06, "loss": 3.0113, "step": 4921 }, { "epoch": 0.94, "learning_rate": 2.7213346043851288e-06, "loss": 3.2509, "step": 4922 }, { "epoch": 0.94, "learning_rate": 2.7212774070543376e-06, "loss": 3.3524, "step": 4923 }, { "epoch": 0.94, "learning_rate": 2.7212202097235464e-06, "loss": 3.3757, "step": 4924 }, { "epoch": 0.94, "learning_rate": 2.721163012392755e-06, "loss": 3.4266, "step": 4925 }, { "epoch": 0.94, "learning_rate": 2.721105815061964e-06, "loss": 2.93, "step": 4926 }, { "epoch": 0.94, "learning_rate": 2.7210486177311728e-06, "loss": 3.6831, "step": 4927 }, { "epoch": 0.94, "learning_rate": 2.7209914204003815e-06, "loss": 3.4814, "step": 4928 }, { "epoch": 0.94, "learning_rate": 2.72093422306959e-06, "loss": 3.0121, "step": 4929 }, { "epoch": 0.94, "learning_rate": 2.720877025738799e-06, "loss": 3.6235, "step": 4930 }, { "epoch": 0.94, "learning_rate": 2.7208198284080075e-06, "loss": 2.8884, "step": 4931 }, { "epoch": 0.94, "learning_rate": 2.7207626310772163e-06, "loss": 3.4778, "step": 4932 }, { "epoch": 0.94, "learning_rate": 2.7207054337464255e-06, "loss": 3.0721, "step": 4933 }, { "epoch": 0.94, "learning_rate": 2.720648236415634e-06, "loss": 3.4554, "step": 4934 }, { "epoch": 0.94, "learning_rate": 2.7205910390848427e-06, "loss": 3.5, "step": 4935 }, { "epoch": 0.94, "learning_rate": 2.7205338417540515e-06, "loss": 3.0495, "step": 4936 }, { "epoch": 0.94, "learning_rate": 2.7204766444232603e-06, "loss": 3.253, "step": 4937 }, { "epoch": 0.94, "learning_rate": 2.720419447092469e-06, "loss": 2.866, "step": 4938 }, { "epoch": 0.94, "learning_rate": 2.720362249761678e-06, "loss": 3.1515, "step": 4939 }, { "epoch": 0.94, "learning_rate": 2.7203050524308866e-06, "loss": 3.0224, "step": 4940 }, { "epoch": 0.94, "learning_rate": 2.7202478551000954e-06, "loss": 3.2663, "step": 4941 }, { "epoch": 0.94, "learning_rate": 2.7201906577693042e-06, "loss": 3.4562, "step": 4942 }, { "epoch": 0.94, "learning_rate": 2.720133460438513e-06, "loss": 2.9833, "step": 4943 }, { "epoch": 0.94, "learning_rate": 2.720076263107722e-06, "loss": 3.2546, "step": 4944 }, { "epoch": 0.94, "learning_rate": 2.7200190657769306e-06, "loss": 3.7443, "step": 4945 }, { "epoch": 0.94, "learning_rate": 2.719961868446139e-06, "loss": 3.1981, "step": 4946 }, { "epoch": 0.94, "learning_rate": 2.719904671115348e-06, "loss": 3.6245, "step": 4947 }, { "epoch": 0.94, "learning_rate": 2.7198474737845566e-06, "loss": 3.4484, "step": 4948 }, { "epoch": 0.94, "learning_rate": 2.7197902764537654e-06, "loss": 3.5946, "step": 4949 }, { "epoch": 0.94, "learning_rate": 2.7197330791229746e-06, "loss": 2.9641, "step": 4950 }, { "epoch": 0.94, "learning_rate": 2.719675881792183e-06, "loss": 3.8317, "step": 4951 }, { "epoch": 0.94, "learning_rate": 2.7196186844613918e-06, "loss": 3.3512, "step": 4952 }, { "epoch": 0.94, "learning_rate": 2.7195614871306005e-06, "loss": 3.7576, "step": 4953 }, { "epoch": 0.94, "learning_rate": 2.7195042897998093e-06, "loss": 3.3508, "step": 4954 }, { "epoch": 0.94, "learning_rate": 2.7194470924690186e-06, "loss": 3.1386, "step": 4955 }, { "epoch": 0.94, "learning_rate": 2.719389895138227e-06, "loss": 2.8742, "step": 4956 }, { "epoch": 0.94, "learning_rate": 2.7193326978074357e-06, "loss": 3.3463, "step": 4957 }, { "epoch": 0.94, "learning_rate": 2.7192755004766445e-06, "loss": 3.231, "step": 4958 }, { "epoch": 0.94, "learning_rate": 2.7192183031458533e-06, "loss": 3.6116, "step": 4959 }, { "epoch": 0.94, "learning_rate": 2.719161105815062e-06, "loss": 3.5708, "step": 4960 }, { "epoch": 0.94, "learning_rate": 2.719103908484271e-06, "loss": 3.34, "step": 4961 }, { "epoch": 0.95, "learning_rate": 2.7190467111534797e-06, "loss": 3.6051, "step": 4962 }, { "epoch": 0.95, "learning_rate": 2.718989513822688e-06, "loss": 3.1126, "step": 4963 }, { "epoch": 0.95, "learning_rate": 2.7189323164918973e-06, "loss": 3.4024, "step": 4964 }, { "epoch": 0.95, "learning_rate": 2.718875119161106e-06, "loss": 3.2737, "step": 4965 }, { "epoch": 0.95, "learning_rate": 2.7188179218303144e-06, "loss": 3.6025, "step": 4966 }, { "epoch": 0.95, "learning_rate": 2.7187607244995237e-06, "loss": 3.6101, "step": 4967 }, { "epoch": 0.95, "learning_rate": 2.718703527168732e-06, "loss": 3.1893, "step": 4968 }, { "epoch": 0.95, "learning_rate": 2.718646329837941e-06, "loss": 3.1769, "step": 4969 }, { "epoch": 0.95, "learning_rate": 2.7185891325071496e-06, "loss": 3.527, "step": 4970 }, { "epoch": 0.95, "learning_rate": 2.7185319351763584e-06, "loss": 3.5969, "step": 4971 }, { "epoch": 0.95, "learning_rate": 2.7184747378455676e-06, "loss": 2.7174, "step": 4972 }, { "epoch": 0.95, "learning_rate": 2.718417540514776e-06, "loss": 3.0213, "step": 4973 }, { "epoch": 0.95, "learning_rate": 2.718360343183985e-06, "loss": 3.4047, "step": 4974 }, { "epoch": 0.95, "learning_rate": 2.7183031458531936e-06, "loss": 3.2902, "step": 4975 }, { "epoch": 0.95, "learning_rate": 2.7182459485224024e-06, "loss": 2.8294, "step": 4976 }, { "epoch": 0.95, "learning_rate": 2.718188751191611e-06, "loss": 3.4689, "step": 4977 }, { "epoch": 0.95, "learning_rate": 2.71813155386082e-06, "loss": 3.3413, "step": 4978 }, { "epoch": 0.95, "learning_rate": 2.7180743565300288e-06, "loss": 3.837, "step": 4979 }, { "epoch": 0.95, "learning_rate": 2.718017159199237e-06, "loss": 3.5949, "step": 4980 }, { "epoch": 0.95, "learning_rate": 2.7179599618684463e-06, "loss": 3.1421, "step": 4981 }, { "epoch": 0.95, "learning_rate": 2.717902764537655e-06, "loss": 3.5832, "step": 4982 }, { "epoch": 0.95, "learning_rate": 2.7178455672068635e-06, "loss": 3.4363, "step": 4983 }, { "epoch": 0.95, "learning_rate": 2.7177883698760727e-06, "loss": 3.4402, "step": 4984 }, { "epoch": 0.95, "learning_rate": 2.717731172545281e-06, "loss": 3.4893, "step": 4985 }, { "epoch": 0.95, "learning_rate": 2.71767397521449e-06, "loss": 3.6543, "step": 4986 }, { "epoch": 0.95, "learning_rate": 2.7176167778836987e-06, "loss": 3.5676, "step": 4987 }, { "epoch": 0.95, "learning_rate": 2.7175595805529075e-06, "loss": 3.3494, "step": 4988 }, { "epoch": 0.95, "learning_rate": 2.7175023832221167e-06, "loss": 3.5399, "step": 4989 }, { "epoch": 0.95, "learning_rate": 2.717445185891325e-06, "loss": 3.56, "step": 4990 }, { "epoch": 0.95, "learning_rate": 2.717387988560534e-06, "loss": 2.9276, "step": 4991 }, { "epoch": 0.95, "learning_rate": 2.7173307912297427e-06, "loss": 3.5599, "step": 4992 }, { "epoch": 0.95, "learning_rate": 2.7172735938989515e-06, "loss": 3.2658, "step": 4993 }, { "epoch": 0.95, "learning_rate": 2.7172163965681602e-06, "loss": 3.5055, "step": 4994 }, { "epoch": 0.95, "learning_rate": 2.717159199237369e-06, "loss": 2.8624, "step": 4995 }, { "epoch": 0.95, "learning_rate": 2.717102001906578e-06, "loss": 2.7427, "step": 4996 }, { "epoch": 0.95, "learning_rate": 2.717044804575786e-06, "loss": 3.4177, "step": 4997 }, { "epoch": 0.95, "learning_rate": 2.7169876072449954e-06, "loss": 3.398, "step": 4998 }, { "epoch": 0.95, "learning_rate": 2.7169304099142042e-06, "loss": 3.0902, "step": 4999 }, { "epoch": 0.95, "learning_rate": 2.7168732125834126e-06, "loss": 3.1023, "step": 5000 }, { "epoch": 0.95, "learning_rate": 2.716816015252622e-06, "loss": 3.4219, "step": 5001 }, { "epoch": 0.95, "learning_rate": 2.71675881792183e-06, "loss": 3.6195, "step": 5002 }, { "epoch": 0.95, "learning_rate": 2.716701620591039e-06, "loss": 3.4665, "step": 5003 }, { "epoch": 0.95, "learning_rate": 2.716644423260248e-06, "loss": 3.5153, "step": 5004 }, { "epoch": 0.95, "learning_rate": 2.7165872259294566e-06, "loss": 3.0179, "step": 5005 }, { "epoch": 0.95, "learning_rate": 2.7165300285986658e-06, "loss": 2.9154, "step": 5006 }, { "epoch": 0.95, "learning_rate": 2.716472831267874e-06, "loss": 3.3705, "step": 5007 }, { "epoch": 0.95, "learning_rate": 2.716415633937083e-06, "loss": 3.0337, "step": 5008 }, { "epoch": 0.95, "learning_rate": 2.7163584366062917e-06, "loss": 3.6892, "step": 5009 }, { "epoch": 0.95, "learning_rate": 2.7163012392755005e-06, "loss": 3.6282, "step": 5010 }, { "epoch": 0.95, "learning_rate": 2.7162440419447093e-06, "loss": 3.2451, "step": 5011 }, { "epoch": 0.95, "learning_rate": 2.716186844613918e-06, "loss": 3.5149, "step": 5012 }, { "epoch": 0.95, "learning_rate": 2.716129647283127e-06, "loss": 3.8153, "step": 5013 }, { "epoch": 0.96, "learning_rate": 2.7160724499523357e-06, "loss": 3.5084, "step": 5014 }, { "epoch": 0.96, "learning_rate": 2.7160152526215445e-06, "loss": 3.2177, "step": 5015 }, { "epoch": 0.96, "learning_rate": 2.7159580552907533e-06, "loss": 3.2224, "step": 5016 }, { "epoch": 0.96, "learning_rate": 2.7159008579599617e-06, "loss": 3.2368, "step": 5017 }, { "epoch": 0.96, "learning_rate": 2.715843660629171e-06, "loss": 3.4676, "step": 5018 }, { "epoch": 0.96, "learning_rate": 2.7157864632983792e-06, "loss": 3.7427, "step": 5019 }, { "epoch": 0.96, "learning_rate": 2.715729265967588e-06, "loss": 3.7118, "step": 5020 }, { "epoch": 0.96, "learning_rate": 2.7156720686367973e-06, "loss": 3.5459, "step": 5021 }, { "epoch": 0.96, "learning_rate": 2.7156148713060056e-06, "loss": 3.3854, "step": 5022 }, { "epoch": 0.96, "learning_rate": 2.715557673975215e-06, "loss": 3.709, "step": 5023 }, { "epoch": 0.96, "learning_rate": 2.7155004766444232e-06, "loss": 3.3535, "step": 5024 }, { "epoch": 0.96, "learning_rate": 2.715443279313632e-06, "loss": 3.4538, "step": 5025 }, { "epoch": 0.96, "learning_rate": 2.715386081982841e-06, "loss": 3.1633, "step": 5026 }, { "epoch": 0.96, "learning_rate": 2.7153288846520496e-06, "loss": 3.7065, "step": 5027 }, { "epoch": 0.96, "learning_rate": 2.7152716873212584e-06, "loss": 3.117, "step": 5028 }, { "epoch": 0.96, "learning_rate": 2.715214489990467e-06, "loss": 2.8979, "step": 5029 }, { "epoch": 0.96, "learning_rate": 2.715157292659676e-06, "loss": 3.313, "step": 5030 }, { "epoch": 0.96, "learning_rate": 2.7151000953288848e-06, "loss": 3.1367, "step": 5031 }, { "epoch": 0.96, "learning_rate": 2.7150428979980936e-06, "loss": 3.414, "step": 5032 }, { "epoch": 0.96, "learning_rate": 2.7149857006673024e-06, "loss": 3.0942, "step": 5033 }, { "epoch": 0.96, "learning_rate": 2.7149285033365107e-06, "loss": 3.3825, "step": 5034 }, { "epoch": 0.96, "learning_rate": 2.71487130600572e-06, "loss": 3.2608, "step": 5035 }, { "epoch": 0.96, "learning_rate": 2.7148141086749283e-06, "loss": 3.5516, "step": 5036 }, { "epoch": 0.96, "learning_rate": 2.7147569113441375e-06, "loss": 2.6953, "step": 5037 }, { "epoch": 0.96, "learning_rate": 2.7146997140133463e-06, "loss": 3.1322, "step": 5038 }, { "epoch": 0.96, "learning_rate": 2.7146425166825547e-06, "loss": 3.4135, "step": 5039 }, { "epoch": 0.96, "learning_rate": 2.714585319351764e-06, "loss": 3.1685, "step": 5040 }, { "epoch": 0.96, "learning_rate": 2.7145281220209723e-06, "loss": 3.5773, "step": 5041 }, { "epoch": 0.96, "learning_rate": 2.714470924690181e-06, "loss": 3.6607, "step": 5042 }, { "epoch": 0.96, "learning_rate": 2.7144137273593903e-06, "loss": 3.7861, "step": 5043 }, { "epoch": 0.96, "learning_rate": 2.7143565300285987e-06, "loss": 3.3228, "step": 5044 }, { "epoch": 0.96, "learning_rate": 2.7142993326978075e-06, "loss": 3.7076, "step": 5045 }, { "epoch": 0.96, "learning_rate": 2.7142421353670163e-06, "loss": 3.6735, "step": 5046 }, { "epoch": 0.96, "learning_rate": 2.714184938036225e-06, "loss": 3.4182, "step": 5047 }, { "epoch": 0.96, "learning_rate": 2.714127740705434e-06, "loss": 3.7538, "step": 5048 }, { "epoch": 0.96, "learning_rate": 2.7140705433746426e-06, "loss": 3.075, "step": 5049 }, { "epoch": 0.96, "learning_rate": 2.7140133460438514e-06, "loss": 3.2557, "step": 5050 }, { "epoch": 0.96, "learning_rate": 2.71395614871306e-06, "loss": 2.991, "step": 5051 }, { "epoch": 0.96, "learning_rate": 2.713898951382269e-06, "loss": 3.3771, "step": 5052 }, { "epoch": 0.96, "learning_rate": 2.713841754051478e-06, "loss": 3.3246, "step": 5053 }, { "epoch": 0.96, "learning_rate": 2.7137845567206866e-06, "loss": 3.0543, "step": 5054 }, { "epoch": 0.96, "learning_rate": 2.7137273593898954e-06, "loss": 3.5107, "step": 5055 }, { "epoch": 0.96, "learning_rate": 2.7136701620591038e-06, "loss": 3.6016, "step": 5056 }, { "epoch": 0.96, "learning_rate": 2.713612964728313e-06, "loss": 3.4946, "step": 5057 }, { "epoch": 0.96, "learning_rate": 2.7135557673975214e-06, "loss": 3.6441, "step": 5058 }, { "epoch": 0.96, "learning_rate": 2.71349857006673e-06, "loss": 2.8634, "step": 5059 }, { "epoch": 0.96, "learning_rate": 2.7134413727359394e-06, "loss": 3.494, "step": 5060 }, { "epoch": 0.96, "learning_rate": 2.7133841754051477e-06, "loss": 3.8397, "step": 5061 }, { "epoch": 0.96, "learning_rate": 2.7133269780743565e-06, "loss": 2.9226, "step": 5062 }, { "epoch": 0.96, "learning_rate": 2.7132697807435653e-06, "loss": 3.1519, "step": 5063 }, { "epoch": 0.96, "learning_rate": 2.713212583412774e-06, "loss": 3.0761, "step": 5064 }, { "epoch": 0.96, "learning_rate": 2.713155386081983e-06, "loss": 3.3465, "step": 5065 }, { "epoch": 0.96, "learning_rate": 2.7130981887511917e-06, "loss": 3.5305, "step": 5066 }, { "epoch": 0.97, "learning_rate": 2.7130409914204005e-06, "loss": 3.4168, "step": 5067 }, { "epoch": 0.97, "learning_rate": 2.712983794089609e-06, "loss": 3.3744, "step": 5068 }, { "epoch": 0.97, "learning_rate": 2.712926596758818e-06, "loss": 3.5047, "step": 5069 }, { "epoch": 0.97, "learning_rate": 2.712869399428027e-06, "loss": 3.4491, "step": 5070 }, { "epoch": 0.97, "learning_rate": 2.7128122020972357e-06, "loss": 3.5014, "step": 5071 }, { "epoch": 0.97, "learning_rate": 2.7127550047664445e-06, "loss": 3.4875, "step": 5072 }, { "epoch": 0.97, "learning_rate": 2.712697807435653e-06, "loss": 3.4188, "step": 5073 }, { "epoch": 0.97, "learning_rate": 2.712640610104862e-06, "loss": 3.1121, "step": 5074 }, { "epoch": 0.97, "learning_rate": 2.7125834127740704e-06, "loss": 3.5769, "step": 5075 }, { "epoch": 0.97, "learning_rate": 2.7125262154432792e-06, "loss": 3.5806, "step": 5076 }, { "epoch": 0.97, "learning_rate": 2.7124690181124884e-06, "loss": 3.1169, "step": 5077 }, { "epoch": 0.97, "learning_rate": 2.712411820781697e-06, "loss": 3.6331, "step": 5078 }, { "epoch": 0.97, "learning_rate": 2.7123546234509056e-06, "loss": 3.3525, "step": 5079 }, { "epoch": 0.97, "learning_rate": 2.7122974261201144e-06, "loss": 3.5686, "step": 5080 }, { "epoch": 0.97, "learning_rate": 2.712240228789323e-06, "loss": 3.3168, "step": 5081 }, { "epoch": 0.97, "learning_rate": 2.712183031458532e-06, "loss": 3.7196, "step": 5082 }, { "epoch": 0.97, "learning_rate": 2.7121258341277408e-06, "loss": 3.8169, "step": 5083 }, { "epoch": 0.97, "learning_rate": 2.7120686367969496e-06, "loss": 3.6506, "step": 5084 }, { "epoch": 0.97, "learning_rate": 2.712011439466158e-06, "loss": 3.459, "step": 5085 }, { "epoch": 0.97, "learning_rate": 2.711954242135367e-06, "loss": 3.3488, "step": 5086 }, { "epoch": 0.97, "learning_rate": 2.711897044804576e-06, "loss": 3.0908, "step": 5087 }, { "epoch": 0.97, "learning_rate": 2.7118398474737847e-06, "loss": 3.6128, "step": 5088 }, { "epoch": 0.97, "learning_rate": 2.7117826501429935e-06, "loss": 3.4136, "step": 5089 }, { "epoch": 0.97, "learning_rate": 2.711725452812202e-06, "loss": 3.9031, "step": 5090 }, { "epoch": 0.97, "learning_rate": 2.711668255481411e-06, "loss": 3.2955, "step": 5091 }, { "epoch": 0.97, "learning_rate": 2.71161105815062e-06, "loss": 2.9833, "step": 5092 }, { "epoch": 0.97, "learning_rate": 2.7115538608198283e-06, "loss": 3.1801, "step": 5093 }, { "epoch": 0.97, "learning_rate": 2.7114966634890375e-06, "loss": 3.5554, "step": 5094 }, { "epoch": 0.97, "learning_rate": 2.711439466158246e-06, "loss": 3.6834, "step": 5095 }, { "epoch": 0.97, "learning_rate": 2.7113822688274547e-06, "loss": 3.2964, "step": 5096 }, { "epoch": 0.97, "learning_rate": 2.7113250714966635e-06, "loss": 3.329, "step": 5097 }, { "epoch": 0.97, "learning_rate": 2.7112678741658723e-06, "loss": 3.5102, "step": 5098 }, { "epoch": 0.97, "learning_rate": 2.711210676835081e-06, "loss": 3.8312, "step": 5099 }, { "epoch": 0.97, "learning_rate": 2.71115347950429e-06, "loss": 3.0749, "step": 5100 }, { "epoch": 0.97, "learning_rate": 2.7110962821734986e-06, "loss": 3.2172, "step": 5101 }, { "epoch": 0.97, "learning_rate": 2.7110390848427074e-06, "loss": 2.791, "step": 5102 }, { "epoch": 0.97, "learning_rate": 2.7109818875119162e-06, "loss": 3.2419, "step": 5103 }, { "epoch": 0.97, "learning_rate": 2.710924690181125e-06, "loss": 3.2075, "step": 5104 }, { "epoch": 0.97, "learning_rate": 2.710867492850334e-06, "loss": 3.8831, "step": 5105 }, { "epoch": 0.97, "learning_rate": 2.7108102955195426e-06, "loss": 3.3977, "step": 5106 }, { "epoch": 0.97, "learning_rate": 2.710753098188751e-06, "loss": 3.6595, "step": 5107 }, { "epoch": 0.97, "learning_rate": 2.71069590085796e-06, "loss": 3.8465, "step": 5108 }, { "epoch": 0.97, "learning_rate": 2.710638703527169e-06, "loss": 3.3044, "step": 5109 }, { "epoch": 0.97, "learning_rate": 2.7105815061963774e-06, "loss": 3.0915, "step": 5110 }, { "epoch": 0.97, "learning_rate": 2.7105243088655866e-06, "loss": 3.3014, "step": 5111 }, { "epoch": 0.97, "learning_rate": 2.710467111534795e-06, "loss": 3.6421, "step": 5112 }, { "epoch": 0.97, "learning_rate": 2.7104099142040037e-06, "loss": 3.6134, "step": 5113 }, { "epoch": 0.97, "learning_rate": 2.7103527168732125e-06, "loss": 3.8186, "step": 5114 }, { "epoch": 0.97, "learning_rate": 2.7102955195424213e-06, "loss": 3.5876, "step": 5115 }, { "epoch": 0.97, "learning_rate": 2.71023832221163e-06, "loss": 3.3671, "step": 5116 }, { "epoch": 0.97, "learning_rate": 2.710181124880839e-06, "loss": 3.7489, "step": 5117 }, { "epoch": 0.97, "learning_rate": 2.7101239275500477e-06, "loss": 3.2698, "step": 5118 }, { "epoch": 0.98, "learning_rate": 2.7100667302192565e-06, "loss": 3.6582, "step": 5119 }, { "epoch": 0.98, "learning_rate": 2.7100095328884653e-06, "loss": 3.3987, "step": 5120 }, { "epoch": 0.98, "learning_rate": 2.709952335557674e-06, "loss": 3.1515, "step": 5121 }, { "epoch": 0.98, "learning_rate": 2.709895138226883e-06, "loss": 3.6024, "step": 5122 }, { "epoch": 0.98, "learning_rate": 2.7098379408960917e-06, "loss": 3.0022, "step": 5123 }, { "epoch": 0.98, "learning_rate": 2.7097807435653e-06, "loss": 3.6257, "step": 5124 }, { "epoch": 0.98, "learning_rate": 2.7097235462345093e-06, "loss": 3.6029, "step": 5125 }, { "epoch": 0.98, "learning_rate": 2.709666348903718e-06, "loss": 3.4175, "step": 5126 }, { "epoch": 0.98, "learning_rate": 2.7096091515729264e-06, "loss": 3.2936, "step": 5127 }, { "epoch": 0.98, "learning_rate": 2.7095519542421357e-06, "loss": 3.0039, "step": 5128 }, { "epoch": 0.98, "learning_rate": 2.709494756911344e-06, "loss": 3.721, "step": 5129 }, { "epoch": 0.98, "learning_rate": 2.709437559580553e-06, "loss": 3.0732, "step": 5130 }, { "epoch": 0.98, "learning_rate": 2.709380362249762e-06, "loss": 3.0468, "step": 5131 }, { "epoch": 0.98, "learning_rate": 2.7093231649189704e-06, "loss": 3.2156, "step": 5132 }, { "epoch": 0.98, "learning_rate": 2.709265967588179e-06, "loss": 3.5169, "step": 5133 }, { "epoch": 0.98, "learning_rate": 2.709208770257388e-06, "loss": 3.3617, "step": 5134 }, { "epoch": 0.98, "learning_rate": 2.7091515729265968e-06, "loss": 3.582, "step": 5135 }, { "epoch": 0.98, "learning_rate": 2.7090943755958056e-06, "loss": 3.3999, "step": 5136 }, { "epoch": 0.98, "learning_rate": 2.7090371782650144e-06, "loss": 3.5896, "step": 5137 }, { "epoch": 0.98, "learning_rate": 2.708979980934223e-06, "loss": 2.843, "step": 5138 }, { "epoch": 0.98, "learning_rate": 2.708922783603432e-06, "loss": 3.0411, "step": 5139 }, { "epoch": 0.98, "learning_rate": 2.7088655862726408e-06, "loss": 3.4322, "step": 5140 }, { "epoch": 0.98, "learning_rate": 2.7088083889418495e-06, "loss": 3.1664, "step": 5141 }, { "epoch": 0.98, "learning_rate": 2.7087511916110583e-06, "loss": 3.1636, "step": 5142 }, { "epoch": 0.98, "learning_rate": 2.708693994280267e-06, "loss": 3.134, "step": 5143 }, { "epoch": 0.98, "learning_rate": 2.7086367969494755e-06, "loss": 3.2899, "step": 5144 }, { "epoch": 0.98, "learning_rate": 2.7085795996186847e-06, "loss": 3.6862, "step": 5145 }, { "epoch": 0.98, "learning_rate": 2.708522402287893e-06, "loss": 3.9465, "step": 5146 }, { "epoch": 0.98, "learning_rate": 2.708465204957102e-06, "loss": 3.2824, "step": 5147 }, { "epoch": 0.98, "learning_rate": 2.708408007626311e-06, "loss": 2.9241, "step": 5148 }, { "epoch": 0.98, "learning_rate": 2.7083508102955195e-06, "loss": 3.4135, "step": 5149 }, { "epoch": 0.98, "learning_rate": 2.7082936129647283e-06, "loss": 3.4272, "step": 5150 }, { "epoch": 0.98, "learning_rate": 2.708236415633937e-06, "loss": 3.2595, "step": 5151 }, { "epoch": 0.98, "learning_rate": 2.708179218303146e-06, "loss": 3.057, "step": 5152 }, { "epoch": 0.98, "learning_rate": 2.7081220209723546e-06, "loss": 3.3539, "step": 5153 }, { "epoch": 0.98, "learning_rate": 2.7080648236415634e-06, "loss": 3.4939, "step": 5154 }, { "epoch": 0.98, "learning_rate": 2.7080076263107722e-06, "loss": 3.4276, "step": 5155 }, { "epoch": 0.98, "learning_rate": 2.707950428979981e-06, "loss": 3.1016, "step": 5156 }, { "epoch": 0.98, "learning_rate": 2.70789323164919e-06, "loss": 3.3818, "step": 5157 }, { "epoch": 0.98, "learning_rate": 2.7078360343183986e-06, "loss": 3.1197, "step": 5158 }, { "epoch": 0.98, "learning_rate": 2.7077788369876074e-06, "loss": 3.1848, "step": 5159 }, { "epoch": 0.98, "learning_rate": 2.707721639656816e-06, "loss": 3.6198, "step": 5160 }, { "epoch": 0.98, "learning_rate": 2.7076644423260246e-06, "loss": 3.3109, "step": 5161 }, { "epoch": 0.98, "learning_rate": 2.707607244995234e-06, "loss": 3.3002, "step": 5162 }, { "epoch": 0.98, "learning_rate": 2.707550047664442e-06, "loss": 3.3068, "step": 5163 }, { "epoch": 0.98, "learning_rate": 2.707492850333651e-06, "loss": 3.196, "step": 5164 }, { "epoch": 0.98, "learning_rate": 2.70743565300286e-06, "loss": 3.4441, "step": 5165 }, { "epoch": 0.98, "learning_rate": 2.7073784556720685e-06, "loss": 3.4265, "step": 5166 }, { "epoch": 0.98, "learning_rate": 2.7073212583412778e-06, "loss": 3.5225, "step": 5167 }, { "epoch": 0.98, "learning_rate": 2.707264061010486e-06, "loss": 3.5163, "step": 5168 }, { "epoch": 0.98, "learning_rate": 2.707206863679695e-06, "loss": 3.4, "step": 5169 }, { "epoch": 0.98, "learning_rate": 2.7071496663489037e-06, "loss": 3.3524, "step": 5170 }, { "epoch": 0.98, "learning_rate": 2.7070924690181125e-06, "loss": 3.6248, "step": 5171 }, { "epoch": 0.99, "learning_rate": 2.7070352716873213e-06, "loss": 3.0059, "step": 5172 }, { "epoch": 0.99, "learning_rate": 2.70697807435653e-06, "loss": 3.1127, "step": 5173 }, { "epoch": 0.99, "learning_rate": 2.706920877025739e-06, "loss": 3.1426, "step": 5174 }, { "epoch": 0.99, "learning_rate": 2.7068636796949477e-06, "loss": 3.7792, "step": 5175 }, { "epoch": 0.99, "learning_rate": 2.7068064823641565e-06, "loss": 3.8628, "step": 5176 }, { "epoch": 0.99, "learning_rate": 2.7067492850333653e-06, "loss": 3.5803, "step": 5177 }, { "epoch": 0.99, "learning_rate": 2.7066920877025736e-06, "loss": 3.1325, "step": 5178 }, { "epoch": 0.99, "learning_rate": 2.706634890371783e-06, "loss": 3.4466, "step": 5179 }, { "epoch": 0.99, "learning_rate": 2.7065776930409917e-06, "loss": 3.2078, "step": 5180 }, { "epoch": 0.99, "learning_rate": 2.7065204957102e-06, "loss": 3.2561, "step": 5181 }, { "epoch": 0.99, "learning_rate": 2.7064632983794092e-06, "loss": 3.4421, "step": 5182 }, { "epoch": 0.99, "learning_rate": 2.7064061010486176e-06, "loss": 2.8286, "step": 5183 }, { "epoch": 0.99, "learning_rate": 2.706348903717827e-06, "loss": 3.0544, "step": 5184 }, { "epoch": 0.99, "learning_rate": 2.706291706387035e-06, "loss": 3.0225, "step": 5185 }, { "epoch": 0.99, "learning_rate": 2.706234509056244e-06, "loss": 3.553, "step": 5186 }, { "epoch": 0.99, "learning_rate": 2.7061773117254532e-06, "loss": 3.6105, "step": 5187 }, { "epoch": 0.99, "learning_rate": 2.7061201143946616e-06, "loss": 3.8107, "step": 5188 }, { "epoch": 0.99, "learning_rate": 2.7060629170638704e-06, "loss": 3.157, "step": 5189 }, { "epoch": 0.99, "learning_rate": 2.706005719733079e-06, "loss": 3.4053, "step": 5190 }, { "epoch": 0.99, "learning_rate": 2.705948522402288e-06, "loss": 2.9442, "step": 5191 }, { "epoch": 0.99, "learning_rate": 2.7058913250714968e-06, "loss": 2.9847, "step": 5192 }, { "epoch": 0.99, "learning_rate": 2.7058341277407056e-06, "loss": 3.339, "step": 5193 }, { "epoch": 0.99, "learning_rate": 2.7057769304099143e-06, "loss": 3.2831, "step": 5194 }, { "epoch": 0.99, "learning_rate": 2.7057197330791227e-06, "loss": 3.7389, "step": 5195 }, { "epoch": 0.99, "learning_rate": 2.705662535748332e-06, "loss": 3.1041, "step": 5196 }, { "epoch": 0.99, "learning_rate": 2.7056053384175407e-06, "loss": 3.4854, "step": 5197 }, { "epoch": 0.99, "learning_rate": 2.705548141086749e-06, "loss": 3.0507, "step": 5198 }, { "epoch": 0.99, "learning_rate": 2.7054909437559583e-06, "loss": 3.0467, "step": 5199 }, { "epoch": 0.99, "learning_rate": 2.7054337464251667e-06, "loss": 3.1357, "step": 5200 }, { "epoch": 0.99, "learning_rate": 2.705376549094376e-06, "loss": 2.7067, "step": 5201 }, { "epoch": 0.99, "learning_rate": 2.7053193517635843e-06, "loss": 3.4798, "step": 5202 }, { "epoch": 0.99, "learning_rate": 2.705262154432793e-06, "loss": 3.3311, "step": 5203 }, { "epoch": 0.99, "learning_rate": 2.7052049571020023e-06, "loss": 2.8941, "step": 5204 }, { "epoch": 0.99, "learning_rate": 2.7051477597712107e-06, "loss": 3.1515, "step": 5205 }, { "epoch": 0.99, "learning_rate": 2.7050905624404195e-06, "loss": 3.425, "step": 5206 }, { "epoch": 0.99, "learning_rate": 2.7050333651096282e-06, "loss": 3.5479, "step": 5207 }, { "epoch": 0.99, "learning_rate": 2.704976167778837e-06, "loss": 3.5506, "step": 5208 }, { "epoch": 0.99, "learning_rate": 2.704918970448046e-06, "loss": 2.6534, "step": 5209 }, { "epoch": 0.99, "learning_rate": 2.7048617731172546e-06, "loss": 3.2913, "step": 5210 }, { "epoch": 0.99, "learning_rate": 2.7048045757864634e-06, "loss": 3.7013, "step": 5211 }, { "epoch": 0.99, "learning_rate": 2.704747378455672e-06, "loss": 3.08, "step": 5212 }, { "epoch": 0.99, "learning_rate": 2.704690181124881e-06, "loss": 2.9056, "step": 5213 }, { "epoch": 0.99, "learning_rate": 2.70463298379409e-06, "loss": 3.5519, "step": 5214 }, { "epoch": 0.99, "learning_rate": 2.704575786463298e-06, "loss": 3.6027, "step": 5215 }, { "epoch": 0.99, "learning_rate": 2.7045185891325074e-06, "loss": 3.0138, "step": 5216 }, { "epoch": 0.99, "learning_rate": 2.7044613918017158e-06, "loss": 3.111, "step": 5217 }, { "epoch": 0.99, "learning_rate": 2.704404194470925e-06, "loss": 3.3238, "step": 5218 }, { "epoch": 0.99, "learning_rate": 2.7043469971401333e-06, "loss": 3.4408, "step": 5219 }, { "epoch": 0.99, "learning_rate": 2.704289799809342e-06, "loss": 3.6509, "step": 5220 }, { "epoch": 0.99, "learning_rate": 2.7042326024785514e-06, "loss": 3.7569, "step": 5221 }, { "epoch": 0.99, "learning_rate": 2.7041754051477597e-06, "loss": 3.6438, "step": 5222 }, { "epoch": 0.99, "learning_rate": 2.7041182078169685e-06, "loss": 3.5088, "step": 5223 }, { "epoch": 1.0, "learning_rate": 2.7040610104861773e-06, "loss": 3.4651, "step": 5224 }, { "epoch": 1.0, "learning_rate": 2.704003813155386e-06, "loss": 3.978, "step": 5225 }, { "epoch": 1.0, "learning_rate": 2.703946615824595e-06, "loss": 2.9467, "step": 5226 }, { "epoch": 1.0, "learning_rate": 2.7038894184938037e-06, "loss": 3.4972, "step": 5227 }, { "epoch": 1.0, "learning_rate": 2.7038322211630125e-06, "loss": 3.0687, "step": 5228 }, { "epoch": 1.0, "learning_rate": 2.7037750238322213e-06, "loss": 3.3231, "step": 5229 }, { "epoch": 1.0, "learning_rate": 2.70371782650143e-06, "loss": 3.0782, "step": 5230 }, { "epoch": 1.0, "learning_rate": 2.703660629170639e-06, "loss": 3.2373, "step": 5231 }, { "epoch": 1.0, "learning_rate": 2.7036034318398477e-06, "loss": 3.2768, "step": 5232 }, { "epoch": 1.0, "learning_rate": 2.7035462345090565e-06, "loss": 3.0904, "step": 5233 }, { "epoch": 1.0, "learning_rate": 2.703489037178265e-06, "loss": 3.749, "step": 5234 }, { "epoch": 1.0, "learning_rate": 2.703431839847474e-06, "loss": 3.3516, "step": 5235 }, { "epoch": 1.0, "learning_rate": 2.703374642516683e-06, "loss": 3.4932, "step": 5236 }, { "epoch": 1.0, "learning_rate": 2.7033174451858912e-06, "loss": 3.5927, "step": 5237 }, { "epoch": 1.0, "learning_rate": 2.7032602478551004e-06, "loss": 2.9911, "step": 5238 }, { "epoch": 1.0, "learning_rate": 2.703203050524309e-06, "loss": 3.0366, "step": 5239 }, { "epoch": 1.0, "learning_rate": 2.7031458531935176e-06, "loss": 3.1601, "step": 5240 }, { "epoch": 1.0, "learning_rate": 2.7030886558627264e-06, "loss": 3.6641, "step": 5241 }, { "epoch": 1.0, "learning_rate": 2.703031458531935e-06, "loss": 3.598, "step": 5242 }, { "epoch": 1.0, "learning_rate": 2.702974261201144e-06, "loss": 2.8205, "step": 5243 }, { "epoch": 1.0, "learning_rate": 2.7029170638703528e-06, "loss": 3.6618, "step": 5244 }, { "epoch": 1.0, "learning_rate": 2.7028598665395616e-06, "loss": 2.9833, "step": 5245 }, { "epoch": 1.0, "learning_rate": 2.7028026692087704e-06, "loss": 2.7753, "step": 5246 }, { "epoch": 1.0, "learning_rate": 2.702745471877979e-06, "loss": 3.2475, "step": 5247 }, { "epoch": 1.0, "learning_rate": 2.702688274547188e-06, "loss": 3.2182, "step": 5248 }, { "epoch": 1.0, "learning_rate": 2.7026310772163967e-06, "loss": 2.8317, "step": 5249 }, { "epoch": 1.0, "learning_rate": 2.7025738798856055e-06, "loss": 3.5673, "step": 5250 }, { "epoch": 1.0, "eval_accuracy": 0.21131590227175312, "eval_loss": 3.271235227584839, "eval_runtime": 123.1829, "eval_samples_per_second": 18.939, "eval_steps_per_second": 4.741, "step": 5250 }, { "epoch": 1.0, "learning_rate": 2.702516682554814e-06, "loss": 3.4992, "step": 5251 }, { "epoch": 1.0, "learning_rate": 2.702459485224023e-06, "loss": 2.9149, "step": 5252 }, { "epoch": 1.0, "learning_rate": 2.702402287893232e-06, "loss": 3.3112, "step": 5253 }, { "epoch": 1.0, "learning_rate": 2.7023450905624403e-06, "loss": 3.7219, "step": 5254 }, { "epoch": 1.0, "learning_rate": 2.7022878932316495e-06, "loss": 3.5139, "step": 5255 }, { "epoch": 1.0, "learning_rate": 2.702230695900858e-06, "loss": 3.782, "step": 5256 }, { "epoch": 1.0, "learning_rate": 2.7021734985700667e-06, "loss": 3.5357, "step": 5257 }, { "epoch": 1.0, "learning_rate": 2.7021163012392755e-06, "loss": 3.2598, "step": 5258 }, { "epoch": 1.0, "learning_rate": 2.7020591039084843e-06, "loss": 3.3222, "step": 5259 }, { "epoch": 1.0, "learning_rate": 2.702001906577693e-06, "loss": 3.2277, "step": 5260 }, { "epoch": 1.0, "learning_rate": 2.701944709246902e-06, "loss": 3.1745, "step": 5261 }, { "epoch": 1.0, "learning_rate": 2.7018875119161106e-06, "loss": 3.6287, "step": 5262 }, { "epoch": 1.0, "learning_rate": 2.7018303145853194e-06, "loss": 3.5939, "step": 5263 }, { "epoch": 1.0, "learning_rate": 2.7017731172545282e-06, "loss": 3.6787, "step": 5264 }, { "epoch": 1.0, "learning_rate": 2.701715919923737e-06, "loss": 3.2387, "step": 5265 }, { "epoch": 1.0, "learning_rate": 2.701658722592946e-06, "loss": 3.6075, "step": 5266 }, { "epoch": 1.0, "learning_rate": 2.7016015252621546e-06, "loss": 3.1366, "step": 5267 }, { "epoch": 1.0, "learning_rate": 2.7015443279313634e-06, "loss": 2.7251, "step": 5268 }, { "epoch": 1.0, "learning_rate": 2.701487130600572e-06, "loss": 2.8941, "step": 5269 }, { "epoch": 1.0, "learning_rate": 2.701429933269781e-06, "loss": 3.0749, "step": 5270 }, { "epoch": 1.0, "learning_rate": 2.7013727359389894e-06, "loss": 3.1518, "step": 5271 }, { "epoch": 1.0, "learning_rate": 2.7013155386081986e-06, "loss": 2.9339, "step": 5272 }, { "epoch": 1.0, "learning_rate": 2.701258341277407e-06, "loss": 3.318, "step": 5273 }, { "epoch": 1.0, "learning_rate": 2.7012011439466157e-06, "loss": 3.6303, "step": 5274 }, { "epoch": 1.0, "learning_rate": 2.701143946615825e-06, "loss": 2.6944, "step": 5275 }, { "epoch": 1.0, "learning_rate": 2.7010867492850333e-06, "loss": 2.84, "step": 5276 }, { "epoch": 1.01, "learning_rate": 2.701029551954242e-06, "loss": 3.3036, "step": 5277 }, { "epoch": 1.01, "learning_rate": 2.700972354623451e-06, "loss": 3.1797, "step": 5278 }, { "epoch": 1.01, "learning_rate": 2.7009151572926597e-06, "loss": 2.6171, "step": 5279 }, { "epoch": 1.01, "learning_rate": 2.7008579599618685e-06, "loss": 3.0037, "step": 5280 }, { "epoch": 1.01, "learning_rate": 2.7008007626310773e-06, "loss": 3.1499, "step": 5281 }, { "epoch": 1.01, "learning_rate": 2.700743565300286e-06, "loss": 3.312, "step": 5282 }, { "epoch": 1.01, "learning_rate": 2.700686367969495e-06, "loss": 3.2193, "step": 5283 }, { "epoch": 1.01, "learning_rate": 2.7006291706387037e-06, "loss": 2.9149, "step": 5284 }, { "epoch": 1.01, "learning_rate": 2.7005719733079125e-06, "loss": 3.2954, "step": 5285 }, { "epoch": 1.01, "learning_rate": 2.7005147759771213e-06, "loss": 3.1761, "step": 5286 }, { "epoch": 1.01, "learning_rate": 2.70045757864633e-06, "loss": 3.5346, "step": 5287 }, { "epoch": 1.01, "learning_rate": 2.7004003813155384e-06, "loss": 3.2865, "step": 5288 }, { "epoch": 1.01, "learning_rate": 2.7003431839847476e-06, "loss": 3.6132, "step": 5289 }, { "epoch": 1.01, "learning_rate": 2.700285986653956e-06, "loss": 3.1811, "step": 5290 }, { "epoch": 1.01, "learning_rate": 2.700228789323165e-06, "loss": 2.8616, "step": 5291 }, { "epoch": 1.01, "learning_rate": 2.700171591992374e-06, "loss": 3.4741, "step": 5292 }, { "epoch": 1.01, "learning_rate": 2.7001143946615824e-06, "loss": 2.7278, "step": 5293 }, { "epoch": 1.01, "learning_rate": 2.700057197330791e-06, "loss": 3.7858, "step": 5294 }, { "epoch": 1.01, "learning_rate": 2.7e-06, "loss": 3.1697, "step": 5295 }, { "epoch": 1.01, "learning_rate": 2.6999428026692088e-06, "loss": 3.4729, "step": 5296 }, { "epoch": 1.01, "learning_rate": 2.6998856053384176e-06, "loss": 2.7757, "step": 5297 }, { "epoch": 1.01, "learning_rate": 2.6998284080076264e-06, "loss": 2.9571, "step": 5298 }, { "epoch": 1.01, "learning_rate": 2.699771210676835e-06, "loss": 3.3744, "step": 5299 }, { "epoch": 1.01, "learning_rate": 2.699714013346044e-06, "loss": 2.9137, "step": 5300 }, { "epoch": 1.01, "learning_rate": 2.6996568160152527e-06, "loss": 3.2412, "step": 5301 }, { "epoch": 1.01, "learning_rate": 2.6995996186844615e-06, "loss": 3.6303, "step": 5302 }, { "epoch": 1.01, "learning_rate": 2.6995424213536703e-06, "loss": 3.4394, "step": 5303 }, { "epoch": 1.01, "learning_rate": 2.699485224022879e-06, "loss": 3.6428, "step": 5304 }, { "epoch": 1.01, "learning_rate": 2.6994280266920875e-06, "loss": 3.2869, "step": 5305 }, { "epoch": 1.01, "learning_rate": 2.6993708293612967e-06, "loss": 2.8489, "step": 5306 }, { "epoch": 1.01, "learning_rate": 2.699313632030505e-06, "loss": 3.4008, "step": 5307 }, { "epoch": 1.01, "learning_rate": 2.699256434699714e-06, "loss": 3.3978, "step": 5308 }, { "epoch": 1.01, "learning_rate": 2.699199237368923e-06, "loss": 3.2124, "step": 5309 }, { "epoch": 1.01, "learning_rate": 2.6991420400381315e-06, "loss": 3.2705, "step": 5310 }, { "epoch": 1.01, "learning_rate": 2.6990848427073403e-06, "loss": 2.9738, "step": 5311 }, { "epoch": 1.01, "learning_rate": 2.699027645376549e-06, "loss": 3.2528, "step": 5312 }, { "epoch": 1.01, "learning_rate": 2.698970448045758e-06, "loss": 3.5323, "step": 5313 }, { "epoch": 1.01, "learning_rate": 2.698913250714967e-06, "loss": 3.5524, "step": 5314 }, { "epoch": 1.01, "learning_rate": 2.6988560533841754e-06, "loss": 2.4808, "step": 5315 }, { "epoch": 1.01, "learning_rate": 2.6987988560533842e-06, "loss": 3.99, "step": 5316 }, { "epoch": 1.01, "learning_rate": 2.698741658722593e-06, "loss": 3.2257, "step": 5317 }, { "epoch": 1.01, "learning_rate": 2.698684461391802e-06, "loss": 3.7284, "step": 5318 }, { "epoch": 1.01, "learning_rate": 2.6986272640610106e-06, "loss": 3.7527, "step": 5319 }, { "epoch": 1.01, "learning_rate": 2.6985700667302194e-06, "loss": 4.0089, "step": 5320 }, { "epoch": 1.01, "learning_rate": 2.698512869399428e-06, "loss": 3.3987, "step": 5321 }, { "epoch": 1.01, "learning_rate": 2.6984556720686366e-06, "loss": 3.0766, "step": 5322 }, { "epoch": 1.01, "learning_rate": 2.6983984747378458e-06, "loss": 3.8404, "step": 5323 }, { "epoch": 1.01, "learning_rate": 2.6983412774070546e-06, "loss": 2.6405, "step": 5324 }, { "epoch": 1.01, "learning_rate": 2.698284080076263e-06, "loss": 2.7876, "step": 5325 }, { "epoch": 1.01, "learning_rate": 2.698226882745472e-06, "loss": 3.2909, "step": 5326 }, { "epoch": 1.01, "learning_rate": 2.6981696854146805e-06, "loss": 2.7662, "step": 5327 }, { "epoch": 1.01, "learning_rate": 2.6981124880838893e-06, "loss": 3.2666, "step": 5328 }, { "epoch": 1.02, "learning_rate": 2.698055290753098e-06, "loss": 3.7871, "step": 5329 }, { "epoch": 1.02, "learning_rate": 2.697998093422307e-06, "loss": 3.2403, "step": 5330 }, { "epoch": 1.02, "learning_rate": 2.697940896091516e-06, "loss": 3.0757, "step": 5331 }, { "epoch": 1.02, "learning_rate": 2.6978836987607245e-06, "loss": 3.4772, "step": 5332 }, { "epoch": 1.02, "learning_rate": 2.6978265014299333e-06, "loss": 3.1245, "step": 5333 }, { "epoch": 1.02, "learning_rate": 2.697769304099142e-06, "loss": 3.5327, "step": 5334 }, { "epoch": 1.02, "learning_rate": 2.697712106768351e-06, "loss": 3.4771, "step": 5335 }, { "epoch": 1.02, "learning_rate": 2.6976549094375597e-06, "loss": 3.5169, "step": 5336 }, { "epoch": 1.02, "learning_rate": 2.6975977121067685e-06, "loss": 3.1522, "step": 5337 }, { "epoch": 1.02, "learning_rate": 2.6975405147759773e-06, "loss": 3.6353, "step": 5338 }, { "epoch": 1.02, "learning_rate": 2.6974833174451856e-06, "loss": 3.7856, "step": 5339 }, { "epoch": 1.02, "learning_rate": 2.697426120114395e-06, "loss": 3.5198, "step": 5340 }, { "epoch": 1.02, "learning_rate": 2.6973689227836037e-06, "loss": 3.312, "step": 5341 }, { "epoch": 1.02, "learning_rate": 2.697311725452812e-06, "loss": 3.4071, "step": 5342 }, { "epoch": 1.02, "learning_rate": 2.6972545281220212e-06, "loss": 3.8587, "step": 5343 }, { "epoch": 1.02, "learning_rate": 2.6971973307912296e-06, "loss": 3.6706, "step": 5344 }, { "epoch": 1.02, "learning_rate": 2.6971401334604384e-06, "loss": 2.9743, "step": 5345 }, { "epoch": 1.02, "learning_rate": 2.697082936129647e-06, "loss": 3.486, "step": 5346 }, { "epoch": 1.02, "learning_rate": 2.697025738798856e-06, "loss": 3.1903, "step": 5347 }, { "epoch": 1.02, "learning_rate": 2.696968541468065e-06, "loss": 2.888, "step": 5348 }, { "epoch": 1.02, "learning_rate": 2.6969113441372736e-06, "loss": 2.9369, "step": 5349 }, { "epoch": 1.02, "learning_rate": 2.6968541468064824e-06, "loss": 3.5129, "step": 5350 }, { "epoch": 1.02, "learning_rate": 2.696796949475691e-06, "loss": 3.3333, "step": 5351 }, { "epoch": 1.02, "learning_rate": 2.6967397521449e-06, "loss": 3.0826, "step": 5352 }, { "epoch": 1.02, "learning_rate": 2.6966825548141088e-06, "loss": 3.3594, "step": 5353 }, { "epoch": 1.02, "learning_rate": 2.6966253574833175e-06, "loss": 3.0967, "step": 5354 }, { "epoch": 1.02, "learning_rate": 2.6965681601525263e-06, "loss": 3.2821, "step": 5355 }, { "epoch": 1.02, "learning_rate": 2.696510962821735e-06, "loss": 3.1511, "step": 5356 }, { "epoch": 1.02, "learning_rate": 2.696453765490944e-06, "loss": 2.9444, "step": 5357 }, { "epoch": 1.02, "learning_rate": 2.6963965681601527e-06, "loss": 2.8049, "step": 5358 }, { "epoch": 1.02, "learning_rate": 2.696339370829361e-06, "loss": 3.5104, "step": 5359 }, { "epoch": 1.02, "learning_rate": 2.6962821734985703e-06, "loss": 3.2047, "step": 5360 }, { "epoch": 1.02, "learning_rate": 2.6962249761677787e-06, "loss": 3.4058, "step": 5361 }, { "epoch": 1.02, "learning_rate": 2.696167778836988e-06, "loss": 3.2616, "step": 5362 }, { "epoch": 1.02, "learning_rate": 2.6961105815061967e-06, "loss": 3.7038, "step": 5363 }, { "epoch": 1.02, "learning_rate": 2.696053384175405e-06, "loss": 3.2642, "step": 5364 }, { "epoch": 1.02, "learning_rate": 2.6959961868446143e-06, "loss": 3.3705, "step": 5365 }, { "epoch": 1.02, "learning_rate": 2.6959389895138227e-06, "loss": 2.8897, "step": 5366 }, { "epoch": 1.02, "learning_rate": 2.6958817921830314e-06, "loss": 3.1845, "step": 5367 }, { "epoch": 1.02, "learning_rate": 2.6958245948522402e-06, "loss": 3.3978, "step": 5368 }, { "epoch": 1.02, "learning_rate": 2.695767397521449e-06, "loss": 3.4976, "step": 5369 }, { "epoch": 1.02, "learning_rate": 2.695710200190658e-06, "loss": 3.1515, "step": 5370 }, { "epoch": 1.02, "learning_rate": 2.6956530028598666e-06, "loss": 3.5457, "step": 5371 }, { "epoch": 1.02, "learning_rate": 2.6955958055290754e-06, "loss": 2.9252, "step": 5372 }, { "epoch": 1.02, "learning_rate": 2.695538608198284e-06, "loss": 3.4005, "step": 5373 }, { "epoch": 1.02, "learning_rate": 2.695481410867493e-06, "loss": 3.3667, "step": 5374 }, { "epoch": 1.02, "learning_rate": 2.695424213536702e-06, "loss": 3.477, "step": 5375 }, { "epoch": 1.02, "learning_rate": 2.69536701620591e-06, "loss": 3.6415, "step": 5376 }, { "epoch": 1.02, "learning_rate": 2.6953098188751194e-06, "loss": 3.1857, "step": 5377 }, { "epoch": 1.02, "learning_rate": 2.6952526215443278e-06, "loss": 3.5441, "step": 5378 }, { "epoch": 1.02, "learning_rate": 2.695195424213537e-06, "loss": 3.3273, "step": 5379 }, { "epoch": 1.02, "learning_rate": 2.6951382268827458e-06, "loss": 3.2262, "step": 5380 }, { "epoch": 1.02, "learning_rate": 2.695081029551954e-06, "loss": 3.4324, "step": 5381 }, { "epoch": 1.03, "learning_rate": 2.6950238322211634e-06, "loss": 3.6662, "step": 5382 }, { "epoch": 1.03, "learning_rate": 2.6949666348903717e-06, "loss": 3.2266, "step": 5383 }, { "epoch": 1.03, "learning_rate": 2.6949094375595805e-06, "loss": 3.4885, "step": 5384 }, { "epoch": 1.03, "learning_rate": 2.6948522402287893e-06, "loss": 3.7451, "step": 5385 }, { "epoch": 1.03, "learning_rate": 2.694795042897998e-06, "loss": 2.9909, "step": 5386 }, { "epoch": 1.03, "learning_rate": 2.694737845567207e-06, "loss": 3.9255, "step": 5387 }, { "epoch": 1.03, "learning_rate": 2.6946806482364157e-06, "loss": 3.5898, "step": 5388 }, { "epoch": 1.03, "learning_rate": 2.6946234509056245e-06, "loss": 3.593, "step": 5389 }, { "epoch": 1.03, "learning_rate": 2.6945662535748333e-06, "loss": 3.3219, "step": 5390 }, { "epoch": 1.03, "learning_rate": 2.694509056244042e-06, "loss": 3.2313, "step": 5391 }, { "epoch": 1.03, "learning_rate": 2.694451858913251e-06, "loss": 3.0929, "step": 5392 }, { "epoch": 1.03, "learning_rate": 2.6943946615824592e-06, "loss": 3.3528, "step": 5393 }, { "epoch": 1.03, "learning_rate": 2.6943374642516685e-06, "loss": 3.1901, "step": 5394 }, { "epoch": 1.03, "learning_rate": 2.694280266920877e-06, "loss": 3.3798, "step": 5395 }, { "epoch": 1.03, "learning_rate": 2.694223069590086e-06, "loss": 2.9918, "step": 5396 }, { "epoch": 1.03, "learning_rate": 2.694165872259295e-06, "loss": 3.6261, "step": 5397 }, { "epoch": 1.03, "learning_rate": 2.694108674928503e-06, "loss": 3.1921, "step": 5398 }, { "epoch": 1.03, "learning_rate": 2.6940514775977124e-06, "loss": 3.6462, "step": 5399 }, { "epoch": 1.03, "learning_rate": 2.693994280266921e-06, "loss": 3.2007, "step": 5400 }, { "epoch": 1.03, "learning_rate": 2.6939370829361296e-06, "loss": 3.1053, "step": 5401 }, { "epoch": 1.03, "learning_rate": 2.693879885605339e-06, "loss": 3.2882, "step": 5402 }, { "epoch": 1.03, "learning_rate": 2.693822688274547e-06, "loss": 3.74, "step": 5403 }, { "epoch": 1.03, "learning_rate": 2.693765490943756e-06, "loss": 3.4042, "step": 5404 }, { "epoch": 1.03, "learning_rate": 2.6937082936129648e-06, "loss": 3.2231, "step": 5405 }, { "epoch": 1.03, "learning_rate": 2.6936510962821736e-06, "loss": 3.7617, "step": 5406 }, { "epoch": 1.03, "learning_rate": 2.6935938989513824e-06, "loss": 3.2856, "step": 5407 }, { "epoch": 1.03, "learning_rate": 2.693536701620591e-06, "loss": 3.38, "step": 5408 }, { "epoch": 1.03, "learning_rate": 2.6934795042898e-06, "loss": 3.3874, "step": 5409 }, { "epoch": 1.03, "learning_rate": 2.6934223069590083e-06, "loss": 2.8684, "step": 5410 }, { "epoch": 1.03, "learning_rate": 2.6933651096282175e-06, "loss": 2.9881, "step": 5411 }, { "epoch": 1.03, "learning_rate": 2.6933079122974263e-06, "loss": 3.6524, "step": 5412 }, { "epoch": 1.03, "learning_rate": 2.693250714966635e-06, "loss": 3.1604, "step": 5413 }, { "epoch": 1.03, "learning_rate": 2.693193517635844e-06, "loss": 3.0037, "step": 5414 }, { "epoch": 1.03, "learning_rate": 2.6931363203050523e-06, "loss": 3.8238, "step": 5415 }, { "epoch": 1.03, "learning_rate": 2.6930791229742615e-06, "loss": 3.9248, "step": 5416 }, { "epoch": 1.03, "learning_rate": 2.69302192564347e-06, "loss": 3.7664, "step": 5417 }, { "epoch": 1.03, "learning_rate": 2.6929647283126787e-06, "loss": 3.6367, "step": 5418 }, { "epoch": 1.03, "learning_rate": 2.692907530981888e-06, "loss": 3.1284, "step": 5419 }, { "epoch": 1.03, "learning_rate": 2.6928503336510962e-06, "loss": 2.866, "step": 5420 }, { "epoch": 1.03, "learning_rate": 2.692793136320305e-06, "loss": 2.801, "step": 5421 }, { "epoch": 1.03, "learning_rate": 2.692735938989514e-06, "loss": 3.1025, "step": 5422 }, { "epoch": 1.03, "learning_rate": 2.6926787416587226e-06, "loss": 3.2196, "step": 5423 }, { "epoch": 1.03, "learning_rate": 2.6926215443279314e-06, "loss": 3.3724, "step": 5424 }, { "epoch": 1.03, "learning_rate": 2.6925643469971402e-06, "loss": 2.8341, "step": 5425 }, { "epoch": 1.03, "learning_rate": 2.692507149666349e-06, "loss": 3.3854, "step": 5426 }, { "epoch": 1.03, "learning_rate": 2.692449952335558e-06, "loss": 3.3245, "step": 5427 }, { "epoch": 1.03, "learning_rate": 2.6923927550047666e-06, "loss": 2.9734, "step": 5428 }, { "epoch": 1.03, "learning_rate": 2.6923355576739754e-06, "loss": 3.3316, "step": 5429 }, { "epoch": 1.03, "learning_rate": 2.692278360343184e-06, "loss": 3.0924, "step": 5430 }, { "epoch": 1.03, "learning_rate": 2.692221163012393e-06, "loss": 3.6508, "step": 5431 }, { "epoch": 1.03, "learning_rate": 2.6921639656816013e-06, "loss": 3.7384, "step": 5432 }, { "epoch": 1.03, "learning_rate": 2.6921067683508106e-06, "loss": 2.5226, "step": 5433 }, { "epoch": 1.04, "learning_rate": 2.692049571020019e-06, "loss": 3.174, "step": 5434 }, { "epoch": 1.04, "learning_rate": 2.6919923736892277e-06, "loss": 3.3193, "step": 5435 }, { "epoch": 1.04, "learning_rate": 2.691935176358437e-06, "loss": 3.3683, "step": 5436 }, { "epoch": 1.04, "learning_rate": 2.6918779790276453e-06, "loss": 3.3456, "step": 5437 }, { "epoch": 1.04, "learning_rate": 2.691820781696854e-06, "loss": 3.4307, "step": 5438 }, { "epoch": 1.04, "learning_rate": 2.691763584366063e-06, "loss": 3.336, "step": 5439 }, { "epoch": 1.04, "learning_rate": 2.6917063870352717e-06, "loss": 3.2888, "step": 5440 }, { "epoch": 1.04, "learning_rate": 2.6916491897044805e-06, "loss": 2.9865, "step": 5441 }, { "epoch": 1.04, "learning_rate": 2.6915919923736893e-06, "loss": 3.6914, "step": 5442 }, { "epoch": 1.04, "learning_rate": 2.691534795042898e-06, "loss": 3.6047, "step": 5443 }, { "epoch": 1.04, "learning_rate": 2.691477597712107e-06, "loss": 3.5105, "step": 5444 }, { "epoch": 1.04, "learning_rate": 2.6914204003813157e-06, "loss": 3.2246, "step": 5445 }, { "epoch": 1.04, "learning_rate": 2.6913632030505245e-06, "loss": 2.8599, "step": 5446 }, { "epoch": 1.04, "learning_rate": 2.6913060057197333e-06, "loss": 3.3796, "step": 5447 }, { "epoch": 1.04, "learning_rate": 2.691248808388942e-06, "loss": 3.5126, "step": 5448 }, { "epoch": 1.04, "learning_rate": 2.6911916110581504e-06, "loss": 3.3228, "step": 5449 }, { "epoch": 1.04, "learning_rate": 2.6911344137273596e-06, "loss": 3.1769, "step": 5450 }, { "epoch": 1.04, "learning_rate": 2.6910772163965684e-06, "loss": 3.1121, "step": 5451 }, { "epoch": 1.04, "learning_rate": 2.691020019065777e-06, "loss": 2.9339, "step": 5452 }, { "epoch": 1.04, "learning_rate": 2.690962821734986e-06, "loss": 3.9896, "step": 5453 }, { "epoch": 1.04, "learning_rate": 2.6909056244041944e-06, "loss": 3.2049, "step": 5454 }, { "epoch": 1.04, "learning_rate": 2.690848427073403e-06, "loss": 3.5966, "step": 5455 }, { "epoch": 1.04, "learning_rate": 2.690791229742612e-06, "loss": 3.3519, "step": 5456 }, { "epoch": 1.04, "learning_rate": 2.6907340324118208e-06, "loss": 3.3325, "step": 5457 }, { "epoch": 1.04, "learning_rate": 2.6906768350810296e-06, "loss": 3.5719, "step": 5458 }, { "epoch": 1.04, "learning_rate": 2.6906196377502384e-06, "loss": 3.4764, "step": 5459 }, { "epoch": 1.04, "learning_rate": 2.690562440419447e-06, "loss": 3.1667, "step": 5460 }, { "epoch": 1.04, "learning_rate": 2.690505243088656e-06, "loss": 3.7686, "step": 5461 }, { "epoch": 1.04, "learning_rate": 2.6904480457578647e-06, "loss": 3.4288, "step": 5462 }, { "epoch": 1.04, "learning_rate": 2.6903908484270735e-06, "loss": 2.7245, "step": 5463 }, { "epoch": 1.04, "learning_rate": 2.6903336510962823e-06, "loss": 3.2099, "step": 5464 }, { "epoch": 1.04, "learning_rate": 2.690276453765491e-06, "loss": 3.1826, "step": 5465 }, { "epoch": 1.04, "learning_rate": 2.6902192564346995e-06, "loss": 3.4833, "step": 5466 }, { "epoch": 1.04, "learning_rate": 2.6901620591039087e-06, "loss": 3.1663, "step": 5467 }, { "epoch": 1.04, "learning_rate": 2.6901048617731175e-06, "loss": 3.3267, "step": 5468 }, { "epoch": 1.04, "learning_rate": 2.690047664442326e-06, "loss": 3.4498, "step": 5469 }, { "epoch": 1.04, "learning_rate": 2.689990467111535e-06, "loss": 3.5419, "step": 5470 }, { "epoch": 1.04, "learning_rate": 2.6899332697807435e-06, "loss": 3.5891, "step": 5471 }, { "epoch": 1.04, "learning_rate": 2.6898760724499523e-06, "loss": 3.9812, "step": 5472 }, { "epoch": 1.04, "learning_rate": 2.689818875119161e-06, "loss": 3.2521, "step": 5473 }, { "epoch": 1.04, "learning_rate": 2.68976167778837e-06, "loss": 3.103, "step": 5474 }, { "epoch": 1.04, "learning_rate": 2.6897044804575786e-06, "loss": 3.7128, "step": 5475 }, { "epoch": 1.04, "learning_rate": 2.6896472831267874e-06, "loss": 3.5788, "step": 5476 }, { "epoch": 1.04, "learning_rate": 2.6895900857959962e-06, "loss": 3.8816, "step": 5477 }, { "epoch": 1.04, "learning_rate": 2.689532888465205e-06, "loss": 3.8435, "step": 5478 }, { "epoch": 1.04, "learning_rate": 2.689475691134414e-06, "loss": 3.0194, "step": 5479 }, { "epoch": 1.04, "learning_rate": 2.6894184938036226e-06, "loss": 3.2104, "step": 5480 }, { "epoch": 1.04, "learning_rate": 2.6893612964728314e-06, "loss": 3.2802, "step": 5481 }, { "epoch": 1.04, "learning_rate": 2.68930409914204e-06, "loss": 3.9534, "step": 5482 }, { "epoch": 1.04, "learning_rate": 2.6892469018112486e-06, "loss": 3.5534, "step": 5483 }, { "epoch": 1.04, "learning_rate": 2.6891897044804578e-06, "loss": 3.3306, "step": 5484 }, { "epoch": 1.04, "learning_rate": 2.6891325071496666e-06, "loss": 3.2418, "step": 5485 }, { "epoch": 1.04, "learning_rate": 2.689075309818875e-06, "loss": 3.521, "step": 5486 }, { "epoch": 1.05, "learning_rate": 2.689018112488084e-06, "loss": 2.8251, "step": 5487 }, { "epoch": 1.05, "learning_rate": 2.6889609151572925e-06, "loss": 3.5593, "step": 5488 }, { "epoch": 1.05, "learning_rate": 2.6889037178265013e-06, "loss": 3.3928, "step": 5489 }, { "epoch": 1.05, "learning_rate": 2.6888465204957105e-06, "loss": 3.286, "step": 5490 }, { "epoch": 1.05, "learning_rate": 2.688789323164919e-06, "loss": 3.191, "step": 5491 }, { "epoch": 1.05, "learning_rate": 2.688732125834128e-06, "loss": 3.2283, "step": 5492 }, { "epoch": 1.05, "learning_rate": 2.6886749285033365e-06, "loss": 3.3806, "step": 5493 }, { "epoch": 1.05, "learning_rate": 2.6886177311725453e-06, "loss": 3.379, "step": 5494 }, { "epoch": 1.05, "learning_rate": 2.688560533841754e-06, "loss": 3.1544, "step": 5495 }, { "epoch": 1.05, "learning_rate": 2.688503336510963e-06, "loss": 3.3555, "step": 5496 }, { "epoch": 1.05, "learning_rate": 2.6884461391801717e-06, "loss": 3.281, "step": 5497 }, { "epoch": 1.05, "learning_rate": 2.6883889418493805e-06, "loss": 3.0022, "step": 5498 }, { "epoch": 1.05, "learning_rate": 2.6883317445185893e-06, "loss": 3.3965, "step": 5499 }, { "epoch": 1.05, "learning_rate": 2.688274547187798e-06, "loss": 3.5586, "step": 5500 }, { "epoch": 1.05, "learning_rate": 2.688217349857007e-06, "loss": 3.2464, "step": 5501 }, { "epoch": 1.05, "learning_rate": 2.6881601525262156e-06, "loss": 3.554, "step": 5502 }, { "epoch": 1.05, "learning_rate": 2.688102955195424e-06, "loss": 3.0744, "step": 5503 }, { "epoch": 1.05, "learning_rate": 2.6880457578646332e-06, "loss": 3.3187, "step": 5504 }, { "epoch": 1.05, "learning_rate": 2.6879885605338416e-06, "loss": 3.0097, "step": 5505 }, { "epoch": 1.05, "learning_rate": 2.6879313632030504e-06, "loss": 3.8157, "step": 5506 }, { "epoch": 1.05, "learning_rate": 2.6878741658722596e-06, "loss": 3.2417, "step": 5507 }, { "epoch": 1.05, "learning_rate": 2.687816968541468e-06, "loss": 3.265, "step": 5508 }, { "epoch": 1.05, "learning_rate": 2.687759771210677e-06, "loss": 3.3986, "step": 5509 }, { "epoch": 1.05, "learning_rate": 2.6877025738798856e-06, "loss": 3.6651, "step": 5510 }, { "epoch": 1.05, "learning_rate": 2.6876453765490944e-06, "loss": 2.9609, "step": 5511 }, { "epoch": 1.05, "learning_rate": 2.687588179218303e-06, "loss": 3.728, "step": 5512 }, { "epoch": 1.05, "learning_rate": 2.687530981887512e-06, "loss": 3.3254, "step": 5513 }, { "epoch": 1.05, "learning_rate": 2.6874737845567207e-06, "loss": 3.491, "step": 5514 }, { "epoch": 1.05, "learning_rate": 2.6874165872259295e-06, "loss": 3.1531, "step": 5515 }, { "epoch": 1.05, "learning_rate": 2.6873593898951383e-06, "loss": 3.1522, "step": 5516 }, { "epoch": 1.05, "learning_rate": 2.687302192564347e-06, "loss": 3.2767, "step": 5517 }, { "epoch": 1.05, "learning_rate": 2.687244995233556e-06, "loss": 3.1866, "step": 5518 }, { "epoch": 1.05, "learning_rate": 2.6871877979027647e-06, "loss": 3.238, "step": 5519 }, { "epoch": 1.05, "learning_rate": 2.687130600571973e-06, "loss": 3.464, "step": 5520 }, { "epoch": 1.05, "learning_rate": 2.6870734032411823e-06, "loss": 3.1506, "step": 5521 }, { "epoch": 1.05, "learning_rate": 2.6870162059103907e-06, "loss": 3.34, "step": 5522 }, { "epoch": 1.05, "learning_rate": 2.6869590085795995e-06, "loss": 3.4069, "step": 5523 }, { "epoch": 1.05, "learning_rate": 2.6869018112488087e-06, "loss": 3.5464, "step": 5524 }, { "epoch": 1.05, "learning_rate": 2.686844613918017e-06, "loss": 3.5293, "step": 5525 }, { "epoch": 1.05, "learning_rate": 2.6867874165872263e-06, "loss": 2.9488, "step": 5526 }, { "epoch": 1.05, "learning_rate": 2.6867302192564346e-06, "loss": 2.7095, "step": 5527 }, { "epoch": 1.05, "learning_rate": 2.6866730219256434e-06, "loss": 3.2257, "step": 5528 }, { "epoch": 1.05, "learning_rate": 2.6866158245948527e-06, "loss": 3.3705, "step": 5529 }, { "epoch": 1.05, "learning_rate": 2.686558627264061e-06, "loss": 3.5346, "step": 5530 }, { "epoch": 1.05, "learning_rate": 2.68650142993327e-06, "loss": 3.1553, "step": 5531 }, { "epoch": 1.05, "learning_rate": 2.6864442326024786e-06, "loss": 3.3764, "step": 5532 }, { "epoch": 1.05, "learning_rate": 2.6863870352716874e-06, "loss": 3.3327, "step": 5533 }, { "epoch": 1.05, "learning_rate": 2.686329837940896e-06, "loss": 3.4549, "step": 5534 }, { "epoch": 1.05, "learning_rate": 2.686272640610105e-06, "loss": 3.4285, "step": 5535 }, { "epoch": 1.05, "learning_rate": 2.6862154432793138e-06, "loss": 3.1489, "step": 5536 }, { "epoch": 1.05, "learning_rate": 2.686158245948522e-06, "loss": 3.2483, "step": 5537 }, { "epoch": 1.05, "learning_rate": 2.6861010486177314e-06, "loss": 3.7292, "step": 5538 }, { "epoch": 1.06, "learning_rate": 2.68604385128694e-06, "loss": 3.5113, "step": 5539 }, { "epoch": 1.06, "learning_rate": 2.6859866539561485e-06, "loss": 3.0187, "step": 5540 }, { "epoch": 1.06, "learning_rate": 2.6859294566253578e-06, "loss": 3.8028, "step": 5541 }, { "epoch": 1.06, "learning_rate": 2.685872259294566e-06, "loss": 3.2334, "step": 5542 }, { "epoch": 1.06, "learning_rate": 2.6858150619637753e-06, "loss": 3.5087, "step": 5543 }, { "epoch": 1.06, "learning_rate": 2.6857578646329837e-06, "loss": 3.1738, "step": 5544 }, { "epoch": 1.06, "learning_rate": 2.6857006673021925e-06, "loss": 3.6362, "step": 5545 }, { "epoch": 1.06, "learning_rate": 2.6856434699714017e-06, "loss": 3.2649, "step": 5546 }, { "epoch": 1.06, "learning_rate": 2.68558627264061e-06, "loss": 3.0966, "step": 5547 }, { "epoch": 1.06, "learning_rate": 2.685529075309819e-06, "loss": 3.1692, "step": 5548 }, { "epoch": 1.06, "learning_rate": 2.6854718779790277e-06, "loss": 3.5611, "step": 5549 }, { "epoch": 1.06, "learning_rate": 2.6854146806482365e-06, "loss": 3.3455, "step": 5550 }, { "epoch": 1.06, "learning_rate": 2.6853574833174453e-06, "loss": 3.152, "step": 5551 }, { "epoch": 1.06, "learning_rate": 2.685300285986654e-06, "loss": 3.2561, "step": 5552 }, { "epoch": 1.06, "learning_rate": 2.685243088655863e-06, "loss": 3.184, "step": 5553 }, { "epoch": 1.06, "learning_rate": 2.6851858913250712e-06, "loss": 3.0704, "step": 5554 }, { "epoch": 1.06, "learning_rate": 2.6851286939942804e-06, "loss": 3.7996, "step": 5555 }, { "epoch": 1.06, "learning_rate": 2.6850714966634892e-06, "loss": 3.0882, "step": 5556 }, { "epoch": 1.06, "learning_rate": 2.685014299332698e-06, "loss": 2.7718, "step": 5557 }, { "epoch": 1.06, "learning_rate": 2.684957102001907e-06, "loss": 3.6113, "step": 5558 }, { "epoch": 1.06, "learning_rate": 2.684899904671115e-06, "loss": 2.7747, "step": 5559 }, { "epoch": 1.06, "learning_rate": 2.6848427073403244e-06, "loss": 2.8641, "step": 5560 }, { "epoch": 1.06, "learning_rate": 2.6847855100095328e-06, "loss": 3.5428, "step": 5561 }, { "epoch": 1.06, "learning_rate": 2.6847283126787416e-06, "loss": 3.2692, "step": 5562 }, { "epoch": 1.06, "learning_rate": 2.684671115347951e-06, "loss": 3.2061, "step": 5563 }, { "epoch": 1.06, "learning_rate": 2.684613918017159e-06, "loss": 3.3066, "step": 5564 }, { "epoch": 1.06, "learning_rate": 2.684556720686368e-06, "loss": 3.2619, "step": 5565 }, { "epoch": 1.06, "learning_rate": 2.6844995233555768e-06, "loss": 3.5352, "step": 5566 }, { "epoch": 1.06, "learning_rate": 2.6844423260247855e-06, "loss": 3.3295, "step": 5567 }, { "epoch": 1.06, "learning_rate": 2.6843851286939943e-06, "loss": 3.0428, "step": 5568 }, { "epoch": 1.06, "learning_rate": 2.684327931363203e-06, "loss": 3.4644, "step": 5569 }, { "epoch": 1.06, "learning_rate": 2.684270734032412e-06, "loss": 3.0902, "step": 5570 }, { "epoch": 1.06, "learning_rate": 2.6842135367016203e-06, "loss": 3.3825, "step": 5571 }, { "epoch": 1.06, "learning_rate": 2.6841563393708295e-06, "loss": 3.4212, "step": 5572 }, { "epoch": 1.06, "learning_rate": 2.6840991420400383e-06, "loss": 3.0714, "step": 5573 }, { "epoch": 1.06, "learning_rate": 2.684041944709247e-06, "loss": 3.5878, "step": 5574 }, { "epoch": 1.06, "learning_rate": 2.683984747378456e-06, "loss": 3.6865, "step": 5575 }, { "epoch": 1.06, "learning_rate": 2.6839275500476643e-06, "loss": 3.8593, "step": 5576 }, { "epoch": 1.06, "learning_rate": 2.6838703527168735e-06, "loss": 3.3054, "step": 5577 }, { "epoch": 1.06, "learning_rate": 2.6838131553860823e-06, "loss": 3.1158, "step": 5578 }, { "epoch": 1.06, "learning_rate": 2.6837559580552907e-06, "loss": 3.0799, "step": 5579 }, { "epoch": 1.06, "learning_rate": 2.6836987607245e-06, "loss": 3.171, "step": 5580 }, { "epoch": 1.06, "learning_rate": 2.6836415633937082e-06, "loss": 3.6461, "step": 5581 }, { "epoch": 1.06, "learning_rate": 2.683584366062917e-06, "loss": 3.037, "step": 5582 }, { "epoch": 1.06, "learning_rate": 2.683527168732126e-06, "loss": 3.6496, "step": 5583 }, { "epoch": 1.06, "learning_rate": 2.6834699714013346e-06, "loss": 3.2029, "step": 5584 }, { "epoch": 1.06, "learning_rate": 2.6834127740705434e-06, "loss": 3.055, "step": 5585 }, { "epoch": 1.06, "learning_rate": 2.683355576739752e-06, "loss": 3.6713, "step": 5586 }, { "epoch": 1.06, "learning_rate": 2.683298379408961e-06, "loss": 3.2253, "step": 5587 }, { "epoch": 1.06, "learning_rate": 2.68324118207817e-06, "loss": 3.5128, "step": 5588 }, { "epoch": 1.06, "learning_rate": 2.6831839847473786e-06, "loss": 3.9453, "step": 5589 }, { "epoch": 1.06, "learning_rate": 2.6831267874165874e-06, "loss": 3.2687, "step": 5590 }, { "epoch": 1.06, "learning_rate": 2.683069590085796e-06, "loss": 3.3305, "step": 5591 }, { "epoch": 1.07, "learning_rate": 2.683012392755005e-06, "loss": 3.538, "step": 5592 }, { "epoch": 1.07, "learning_rate": 2.6829551954242133e-06, "loss": 2.8464, "step": 5593 }, { "epoch": 1.07, "learning_rate": 2.6828979980934226e-06, "loss": 3.3362, "step": 5594 }, { "epoch": 1.07, "learning_rate": 2.6828408007626314e-06, "loss": 3.6292, "step": 5595 }, { "epoch": 1.07, "learning_rate": 2.6827836034318397e-06, "loss": 3.1457, "step": 5596 }, { "epoch": 1.07, "learning_rate": 2.682726406101049e-06, "loss": 3.2371, "step": 5597 }, { "epoch": 1.07, "learning_rate": 2.6826692087702573e-06, "loss": 3.2672, "step": 5598 }, { "epoch": 1.07, "learning_rate": 2.682612011439466e-06, "loss": 3.3896, "step": 5599 }, { "epoch": 1.07, "learning_rate": 2.682554814108675e-06, "loss": 3.0265, "step": 5600 }, { "epoch": 1.07, "learning_rate": 2.6824976167778837e-06, "loss": 3.588, "step": 5601 }, { "epoch": 1.07, "learning_rate": 2.6824404194470925e-06, "loss": 2.7045, "step": 5602 }, { "epoch": 1.07, "learning_rate": 2.6823832221163013e-06, "loss": 3.1288, "step": 5603 }, { "epoch": 1.07, "learning_rate": 2.68232602478551e-06, "loss": 3.4675, "step": 5604 }, { "epoch": 1.07, "learning_rate": 2.682268827454719e-06, "loss": 2.8427, "step": 5605 }, { "epoch": 1.07, "learning_rate": 2.6822116301239277e-06, "loss": 3.6719, "step": 5606 }, { "epoch": 1.07, "learning_rate": 2.6821544327931365e-06, "loss": 2.9957, "step": 5607 }, { "epoch": 1.07, "learning_rate": 2.6820972354623452e-06, "loss": 3.0943, "step": 5608 }, { "epoch": 1.07, "learning_rate": 2.682040038131554e-06, "loss": 3.5072, "step": 5609 }, { "epoch": 1.07, "learning_rate": 2.6819828408007624e-06, "loss": 3.7126, "step": 5610 }, { "epoch": 1.07, "learning_rate": 2.6819256434699716e-06, "loss": 3.3887, "step": 5611 }, { "epoch": 1.07, "learning_rate": 2.6818684461391804e-06, "loss": 3.0195, "step": 5612 }, { "epoch": 1.07, "learning_rate": 2.681811248808389e-06, "loss": 3.6174, "step": 5613 }, { "epoch": 1.07, "learning_rate": 2.681754051477598e-06, "loss": 3.4781, "step": 5614 }, { "epoch": 1.07, "learning_rate": 2.6816968541468064e-06, "loss": 3.4002, "step": 5615 }, { "epoch": 1.07, "learning_rate": 2.681639656816015e-06, "loss": 3.0642, "step": 5616 }, { "epoch": 1.07, "learning_rate": 2.681582459485224e-06, "loss": 2.9284, "step": 5617 }, { "epoch": 1.07, "learning_rate": 2.6815252621544328e-06, "loss": 3.5053, "step": 5618 }, { "epoch": 1.07, "learning_rate": 2.6814680648236416e-06, "loss": 2.6542, "step": 5619 }, { "epoch": 1.07, "learning_rate": 2.6814108674928504e-06, "loss": 3.4925, "step": 5620 }, { "epoch": 1.07, "learning_rate": 2.681353670162059e-06, "loss": 3.0399, "step": 5621 }, { "epoch": 1.07, "learning_rate": 2.681296472831268e-06, "loss": 3.7712, "step": 5622 }, { "epoch": 1.07, "learning_rate": 2.6812392755004767e-06, "loss": 3.2911, "step": 5623 }, { "epoch": 1.07, "learning_rate": 2.6811820781696855e-06, "loss": 3.0355, "step": 5624 }, { "epoch": 1.07, "learning_rate": 2.6811248808388943e-06, "loss": 2.6089, "step": 5625 }, { "epoch": 1.07, "learning_rate": 2.681067683508103e-06, "loss": 3.7575, "step": 5626 }, { "epoch": 1.07, "learning_rate": 2.681010486177312e-06, "loss": 2.8796, "step": 5627 }, { "epoch": 1.07, "learning_rate": 2.6809532888465207e-06, "loss": 3.4535, "step": 5628 }, { "epoch": 1.07, "learning_rate": 2.6808960915157295e-06, "loss": 3.6651, "step": 5629 }, { "epoch": 1.07, "learning_rate": 2.680838894184938e-06, "loss": 3.3602, "step": 5630 }, { "epoch": 1.07, "learning_rate": 2.680781696854147e-06, "loss": 3.2069, "step": 5631 }, { "epoch": 1.07, "learning_rate": 2.6807244995233555e-06, "loss": 3.1884, "step": 5632 }, { "epoch": 1.07, "learning_rate": 2.6806673021925642e-06, "loss": 3.6517, "step": 5633 }, { "epoch": 1.07, "learning_rate": 2.6806101048617735e-06, "loss": 3.7005, "step": 5634 }, { "epoch": 1.07, "learning_rate": 2.680552907530982e-06, "loss": 3.49, "step": 5635 }, { "epoch": 1.07, "learning_rate": 2.6804957102001906e-06, "loss": 3.5337, "step": 5636 }, { "epoch": 1.07, "learning_rate": 2.6804385128693994e-06, "loss": 3.6679, "step": 5637 }, { "epoch": 1.07, "learning_rate": 2.6803813155386082e-06, "loss": 3.064, "step": 5638 }, { "epoch": 1.07, "learning_rate": 2.680324118207817e-06, "loss": 3.4313, "step": 5639 }, { "epoch": 1.07, "learning_rate": 2.680266920877026e-06, "loss": 3.5834, "step": 5640 }, { "epoch": 1.07, "learning_rate": 2.6802097235462346e-06, "loss": 3.2216, "step": 5641 }, { "epoch": 1.07, "learning_rate": 2.6801525262154434e-06, "loss": 2.8868, "step": 5642 }, { "epoch": 1.07, "learning_rate": 2.680095328884652e-06, "loss": 3.4719, "step": 5643 }, { "epoch": 1.08, "learning_rate": 2.680038131553861e-06, "loss": 2.6433, "step": 5644 }, { "epoch": 1.08, "learning_rate": 2.6799809342230698e-06, "loss": 3.631, "step": 5645 }, { "epoch": 1.08, "learning_rate": 2.6799237368922786e-06, "loss": 3.1527, "step": 5646 }, { "epoch": 1.08, "learning_rate": 2.679866539561487e-06, "loss": 3.4518, "step": 5647 }, { "epoch": 1.08, "learning_rate": 2.679809342230696e-06, "loss": 3.367, "step": 5648 }, { "epoch": 1.08, "learning_rate": 2.6797521448999045e-06, "loss": 3.4886, "step": 5649 }, { "epoch": 1.08, "learning_rate": 2.6796949475691133e-06, "loss": 2.9778, "step": 5650 }, { "epoch": 1.08, "learning_rate": 2.6796377502383225e-06, "loss": 3.4304, "step": 5651 }, { "epoch": 1.08, "learning_rate": 2.679580552907531e-06, "loss": 3.2402, "step": 5652 }, { "epoch": 1.08, "learning_rate": 2.6795233555767397e-06, "loss": 3.4119, "step": 5653 }, { "epoch": 1.08, "learning_rate": 2.6794661582459485e-06, "loss": 3.4006, "step": 5654 }, { "epoch": 1.08, "learning_rate": 2.6794089609151573e-06, "loss": 3.3537, "step": 5655 }, { "epoch": 1.08, "learning_rate": 2.679351763584366e-06, "loss": 3.0741, "step": 5656 }, { "epoch": 1.08, "learning_rate": 2.679294566253575e-06, "loss": 3.0584, "step": 5657 }, { "epoch": 1.08, "learning_rate": 2.6792373689227837e-06, "loss": 3.7136, "step": 5658 }, { "epoch": 1.08, "learning_rate": 2.6791801715919925e-06, "loss": 3.1484, "step": 5659 }, { "epoch": 1.08, "learning_rate": 2.6791229742612013e-06, "loss": 3.3378, "step": 5660 }, { "epoch": 1.08, "learning_rate": 2.67906577693041e-06, "loss": 3.722, "step": 5661 }, { "epoch": 1.08, "learning_rate": 2.679008579599619e-06, "loss": 3.7579, "step": 5662 }, { "epoch": 1.08, "learning_rate": 2.6789513822688276e-06, "loss": 3.2408, "step": 5663 }, { "epoch": 1.08, "learning_rate": 2.678894184938036e-06, "loss": 3.5581, "step": 5664 }, { "epoch": 1.08, "learning_rate": 2.6788369876072452e-06, "loss": 3.2321, "step": 5665 }, { "epoch": 1.08, "learning_rate": 2.678779790276454e-06, "loss": 2.9711, "step": 5666 }, { "epoch": 1.08, "learning_rate": 2.6787225929456624e-06, "loss": 3.6808, "step": 5667 }, { "epoch": 1.08, "learning_rate": 2.6786653956148716e-06, "loss": 2.9023, "step": 5668 }, { "epoch": 1.08, "learning_rate": 2.67860819828408e-06, "loss": 3.1895, "step": 5669 }, { "epoch": 1.08, "learning_rate": 2.6785510009532888e-06, "loss": 3.4021, "step": 5670 }, { "epoch": 1.08, "learning_rate": 2.6784938036224976e-06, "loss": 3.1385, "step": 5671 }, { "epoch": 1.08, "learning_rate": 2.6784366062917064e-06, "loss": 2.984, "step": 5672 }, { "epoch": 1.08, "learning_rate": 2.6783794089609156e-06, "loss": 3.5632, "step": 5673 }, { "epoch": 1.08, "learning_rate": 2.678322211630124e-06, "loss": 3.018, "step": 5674 }, { "epoch": 1.08, "learning_rate": 2.6782650142993327e-06, "loss": 3.3027, "step": 5675 }, { "epoch": 1.08, "learning_rate": 2.6782078169685415e-06, "loss": 3.3304, "step": 5676 }, { "epoch": 1.08, "learning_rate": 2.6781506196377503e-06, "loss": 3.5758, "step": 5677 }, { "epoch": 1.08, "learning_rate": 2.678093422306959e-06, "loss": 3.0644, "step": 5678 }, { "epoch": 1.08, "learning_rate": 2.678036224976168e-06, "loss": 3.534, "step": 5679 }, { "epoch": 1.08, "learning_rate": 2.6779790276453767e-06, "loss": 3.4163, "step": 5680 }, { "epoch": 1.08, "learning_rate": 2.677921830314585e-06, "loss": 3.4564, "step": 5681 }, { "epoch": 1.08, "learning_rate": 2.6778646329837943e-06, "loss": 3.3837, "step": 5682 }, { "epoch": 1.08, "learning_rate": 2.677807435653003e-06, "loss": 3.4, "step": 5683 }, { "epoch": 1.08, "learning_rate": 2.6777502383222115e-06, "loss": 3.42, "step": 5684 }, { "epoch": 1.08, "learning_rate": 2.6776930409914207e-06, "loss": 2.7884, "step": 5685 }, { "epoch": 1.08, "learning_rate": 2.677635843660629e-06, "loss": 3.45, "step": 5686 }, { "epoch": 1.08, "learning_rate": 2.677578646329838e-06, "loss": 3.537, "step": 5687 }, { "epoch": 1.08, "learning_rate": 2.6775214489990466e-06, "loss": 3.1802, "step": 5688 }, { "epoch": 1.08, "learning_rate": 2.6774642516682554e-06, "loss": 2.9515, "step": 5689 }, { "epoch": 1.08, "learning_rate": 2.6774070543374646e-06, "loss": 3.4262, "step": 5690 }, { "epoch": 1.08, "learning_rate": 2.677349857006673e-06, "loss": 3.2679, "step": 5691 }, { "epoch": 1.08, "learning_rate": 2.677292659675882e-06, "loss": 3.5835, "step": 5692 }, { "epoch": 1.08, "learning_rate": 2.6772354623450906e-06, "loss": 3.3189, "step": 5693 }, { "epoch": 1.08, "learning_rate": 2.6771782650142994e-06, "loss": 3.1451, "step": 5694 }, { "epoch": 1.08, "learning_rate": 2.677121067683508e-06, "loss": 3.2517, "step": 5695 }, { "epoch": 1.08, "learning_rate": 2.677063870352717e-06, "loss": 3.0847, "step": 5696 }, { "epoch": 1.09, "learning_rate": 2.6770066730219258e-06, "loss": 3.6031, "step": 5697 }, { "epoch": 1.09, "learning_rate": 2.676949475691134e-06, "loss": 3.0099, "step": 5698 }, { "epoch": 1.09, "learning_rate": 2.6768922783603434e-06, "loss": 3.6288, "step": 5699 }, { "epoch": 1.09, "learning_rate": 2.676835081029552e-06, "loss": 3.1769, "step": 5700 }, { "epoch": 1.09, "learning_rate": 2.6767778836987605e-06, "loss": 3.114, "step": 5701 }, { "epoch": 1.09, "learning_rate": 2.6767206863679698e-06, "loss": 3.4449, "step": 5702 }, { "epoch": 1.09, "learning_rate": 2.676663489037178e-06, "loss": 3.3814, "step": 5703 }, { "epoch": 1.09, "learning_rate": 2.6766062917063873e-06, "loss": 3.4701, "step": 5704 }, { "epoch": 1.09, "learning_rate": 2.6765490943755957e-06, "loss": 3.6224, "step": 5705 }, { "epoch": 1.09, "learning_rate": 2.6764918970448045e-06, "loss": 3.0519, "step": 5706 }, { "epoch": 1.09, "learning_rate": 2.6764346997140137e-06, "loss": 3.2419, "step": 5707 }, { "epoch": 1.09, "learning_rate": 2.676377502383222e-06, "loss": 3.3932, "step": 5708 }, { "epoch": 1.09, "learning_rate": 2.676320305052431e-06, "loss": 3.2506, "step": 5709 }, { "epoch": 1.09, "learning_rate": 2.6762631077216397e-06, "loss": 3.6702, "step": 5710 }, { "epoch": 1.09, "learning_rate": 2.6762059103908485e-06, "loss": 3.3154, "step": 5711 }, { "epoch": 1.09, "learning_rate": 2.6761487130600573e-06, "loss": 2.9801, "step": 5712 }, { "epoch": 1.09, "learning_rate": 2.676091515729266e-06, "loss": 2.9644, "step": 5713 }, { "epoch": 1.09, "learning_rate": 2.676034318398475e-06, "loss": 3.0396, "step": 5714 }, { "epoch": 1.09, "learning_rate": 2.6759771210676836e-06, "loss": 3.7056, "step": 5715 }, { "epoch": 1.09, "learning_rate": 2.6759199237368924e-06, "loss": 3.6451, "step": 5716 }, { "epoch": 1.09, "learning_rate": 2.6758627264061012e-06, "loss": 3.3861, "step": 5717 }, { "epoch": 1.09, "learning_rate": 2.6758055290753096e-06, "loss": 3.3077, "step": 5718 }, { "epoch": 1.09, "learning_rate": 2.675748331744519e-06, "loss": 3.2689, "step": 5719 }, { "epoch": 1.09, "learning_rate": 2.675691134413727e-06, "loss": 3.4142, "step": 5720 }, { "epoch": 1.09, "learning_rate": 2.6756339370829364e-06, "loss": 2.7894, "step": 5721 }, { "epoch": 1.09, "learning_rate": 2.675576739752145e-06, "loss": 3.003, "step": 5722 }, { "epoch": 1.09, "learning_rate": 2.6755195424213536e-06, "loss": 2.6492, "step": 5723 }, { "epoch": 1.09, "learning_rate": 2.675462345090563e-06, "loss": 2.9608, "step": 5724 }, { "epoch": 1.09, "learning_rate": 2.675405147759771e-06, "loss": 3.0743, "step": 5725 }, { "epoch": 1.09, "learning_rate": 2.67534795042898e-06, "loss": 3.201, "step": 5726 }, { "epoch": 1.09, "learning_rate": 2.6752907530981887e-06, "loss": 3.8143, "step": 5727 }, { "epoch": 1.09, "learning_rate": 2.6752335557673975e-06, "loss": 2.6184, "step": 5728 }, { "epoch": 1.09, "learning_rate": 2.6751763584366063e-06, "loss": 2.6682, "step": 5729 }, { "epoch": 1.09, "learning_rate": 2.675119161105815e-06, "loss": 2.9385, "step": 5730 }, { "epoch": 1.09, "learning_rate": 2.675061963775024e-06, "loss": 2.771, "step": 5731 }, { "epoch": 1.09, "learning_rate": 2.6750047664442327e-06, "loss": 3.583, "step": 5732 }, { "epoch": 1.09, "learning_rate": 2.6749475691134415e-06, "loss": 2.5496, "step": 5733 }, { "epoch": 1.09, "learning_rate": 2.6748903717826503e-06, "loss": 2.9419, "step": 5734 }, { "epoch": 1.09, "learning_rate": 2.6748331744518587e-06, "loss": 3.519, "step": 5735 }, { "epoch": 1.09, "learning_rate": 2.674775977121068e-06, "loss": 3.2274, "step": 5736 }, { "epoch": 1.09, "learning_rate": 2.6747187797902763e-06, "loss": 3.3607, "step": 5737 }, { "epoch": 1.09, "learning_rate": 2.6746615824594855e-06, "loss": 3.0811, "step": 5738 }, { "epoch": 1.09, "learning_rate": 2.6746043851286943e-06, "loss": 2.6604, "step": 5739 }, { "epoch": 1.09, "learning_rate": 2.6745471877979026e-06, "loss": 2.6063, "step": 5740 }, { "epoch": 1.09, "learning_rate": 2.674489990467112e-06, "loss": 3.6239, "step": 5741 }, { "epoch": 1.09, "learning_rate": 2.6744327931363202e-06, "loss": 2.4674, "step": 5742 }, { "epoch": 1.09, "learning_rate": 2.674375595805529e-06, "loss": 3.2804, "step": 5743 }, { "epoch": 1.09, "learning_rate": 2.674318398474738e-06, "loss": 3.0589, "step": 5744 }, { "epoch": 1.09, "learning_rate": 2.6742612011439466e-06, "loss": 2.8108, "step": 5745 }, { "epoch": 1.09, "learning_rate": 2.6742040038131554e-06, "loss": 3.4943, "step": 5746 }, { "epoch": 1.09, "learning_rate": 2.674146806482364e-06, "loss": 3.5848, "step": 5747 }, { "epoch": 1.09, "learning_rate": 2.674089609151573e-06, "loss": 2.828, "step": 5748 }, { "epoch": 1.1, "learning_rate": 2.674032411820782e-06, "loss": 3.6929, "step": 5749 }, { "epoch": 1.1, "learning_rate": 2.6739752144899906e-06, "loss": 3.5223, "step": 5750 }, { "epoch": 1.1, "learning_rate": 2.6739180171591994e-06, "loss": 3.0526, "step": 5751 }, { "epoch": 1.1, "learning_rate": 2.6738608198284077e-06, "loss": 3.458, "step": 5752 }, { "epoch": 1.1, "learning_rate": 2.673803622497617e-06, "loss": 3.4714, "step": 5753 }, { "epoch": 1.1, "learning_rate": 2.6737464251668258e-06, "loss": 3.5524, "step": 5754 }, { "epoch": 1.1, "learning_rate": 2.6736892278360346e-06, "loss": 3.5992, "step": 5755 }, { "epoch": 1.1, "learning_rate": 2.6736320305052433e-06, "loss": 3.6666, "step": 5756 }, { "epoch": 1.1, "learning_rate": 2.6735748331744517e-06, "loss": 3.1699, "step": 5757 }, { "epoch": 1.1, "learning_rate": 2.673517635843661e-06, "loss": 3.0568, "step": 5758 }, { "epoch": 1.1, "learning_rate": 2.6734604385128693e-06, "loss": 3.3465, "step": 5759 }, { "epoch": 1.1, "learning_rate": 2.673403241182078e-06, "loss": 3.4334, "step": 5760 }, { "epoch": 1.1, "learning_rate": 2.6733460438512873e-06, "loss": 2.8946, "step": 5761 }, { "epoch": 1.1, "learning_rate": 2.6732888465204957e-06, "loss": 3.8221, "step": 5762 }, { "epoch": 1.1, "learning_rate": 2.6732316491897045e-06, "loss": 3.0401, "step": 5763 }, { "epoch": 1.1, "learning_rate": 2.6731744518589133e-06, "loss": 3.0741, "step": 5764 }, { "epoch": 1.1, "learning_rate": 2.673117254528122e-06, "loss": 3.5868, "step": 5765 }, { "epoch": 1.1, "learning_rate": 2.673060057197331e-06, "loss": 3.6266, "step": 5766 }, { "epoch": 1.1, "learning_rate": 2.6730028598665397e-06, "loss": 2.6765, "step": 5767 }, { "epoch": 1.1, "learning_rate": 2.6729456625357484e-06, "loss": 2.8878, "step": 5768 }, { "epoch": 1.1, "learning_rate": 2.6728884652049572e-06, "loss": 3.5531, "step": 5769 }, { "epoch": 1.1, "learning_rate": 2.672831267874166e-06, "loss": 2.8992, "step": 5770 }, { "epoch": 1.1, "learning_rate": 2.672774070543375e-06, "loss": 3.1915, "step": 5771 }, { "epoch": 1.1, "learning_rate": 2.6727168732125836e-06, "loss": 3.34, "step": 5772 }, { "epoch": 1.1, "learning_rate": 2.6726596758817924e-06, "loss": 3.486, "step": 5773 }, { "epoch": 1.1, "learning_rate": 2.6726024785510008e-06, "loss": 3.7246, "step": 5774 }, { "epoch": 1.1, "learning_rate": 2.67254528122021e-06, "loss": 3.0939, "step": 5775 }, { "epoch": 1.1, "learning_rate": 2.6724880838894184e-06, "loss": 3.7979, "step": 5776 }, { "epoch": 1.1, "learning_rate": 2.672430886558627e-06, "loss": 3.7977, "step": 5777 }, { "epoch": 1.1, "learning_rate": 2.6723736892278364e-06, "loss": 3.6493, "step": 5778 }, { "epoch": 1.1, "learning_rate": 2.6723164918970448e-06, "loss": 3.0339, "step": 5779 }, { "epoch": 1.1, "learning_rate": 2.6722592945662536e-06, "loss": 3.5043, "step": 5780 }, { "epoch": 1.1, "learning_rate": 2.6722020972354623e-06, "loss": 3.164, "step": 5781 }, { "epoch": 1.1, "learning_rate": 2.672144899904671e-06, "loss": 3.4829, "step": 5782 }, { "epoch": 1.1, "learning_rate": 2.67208770257388e-06, "loss": 3.3556, "step": 5783 }, { "epoch": 1.1, "learning_rate": 2.6720305052430887e-06, "loss": 3.1566, "step": 5784 }, { "epoch": 1.1, "learning_rate": 2.6719733079122975e-06, "loss": 3.4791, "step": 5785 }, { "epoch": 1.1, "learning_rate": 2.6719161105815063e-06, "loss": 3.5612, "step": 5786 }, { "epoch": 1.1, "learning_rate": 2.671858913250715e-06, "loss": 3.1811, "step": 5787 }, { "epoch": 1.1, "learning_rate": 2.671801715919924e-06, "loss": 3.4301, "step": 5788 }, { "epoch": 1.1, "learning_rate": 2.6717445185891327e-06, "loss": 3.3079, "step": 5789 }, { "epoch": 1.1, "learning_rate": 2.6716873212583415e-06, "loss": 3.568, "step": 5790 }, { "epoch": 1.1, "learning_rate": 2.67163012392755e-06, "loss": 3.2365, "step": 5791 }, { "epoch": 1.1, "learning_rate": 2.671572926596759e-06, "loss": 3.2732, "step": 5792 }, { "epoch": 1.1, "learning_rate": 2.6715157292659674e-06, "loss": 3.5931, "step": 5793 }, { "epoch": 1.1, "learning_rate": 2.6714585319351762e-06, "loss": 2.7772, "step": 5794 }, { "epoch": 1.1, "learning_rate": 2.6714013346043855e-06, "loss": 3.6718, "step": 5795 }, { "epoch": 1.1, "learning_rate": 2.671344137273594e-06, "loss": 3.4594, "step": 5796 }, { "epoch": 1.1, "learning_rate": 2.6712869399428026e-06, "loss": 2.5166, "step": 5797 }, { "epoch": 1.1, "learning_rate": 2.6712297426120114e-06, "loss": 3.6683, "step": 5798 }, { "epoch": 1.1, "learning_rate": 2.67117254528122e-06, "loss": 3.3991, "step": 5799 }, { "epoch": 1.1, "learning_rate": 2.671115347950429e-06, "loss": 3.195, "step": 5800 }, { "epoch": 1.1, "learning_rate": 2.671058150619638e-06, "loss": 3.2128, "step": 5801 }, { "epoch": 1.11, "learning_rate": 2.6710009532888466e-06, "loss": 3.2887, "step": 5802 }, { "epoch": 1.11, "learning_rate": 2.6709437559580554e-06, "loss": 3.7638, "step": 5803 }, { "epoch": 1.11, "learning_rate": 2.670886558627264e-06, "loss": 3.3531, "step": 5804 }, { "epoch": 1.11, "learning_rate": 2.670829361296473e-06, "loss": 3.3537, "step": 5805 }, { "epoch": 1.11, "learning_rate": 2.6707721639656818e-06, "loss": 3.7449, "step": 5806 }, { "epoch": 1.11, "learning_rate": 2.6707149666348906e-06, "loss": 3.4262, "step": 5807 }, { "epoch": 1.11, "learning_rate": 2.670657769304099e-06, "loss": 3.8799, "step": 5808 }, { "epoch": 1.11, "learning_rate": 2.670600571973308e-06, "loss": 3.2944, "step": 5809 }, { "epoch": 1.11, "learning_rate": 2.670543374642517e-06, "loss": 2.8272, "step": 5810 }, { "epoch": 1.11, "learning_rate": 2.6704861773117253e-06, "loss": 3.3424, "step": 5811 }, { "epoch": 1.11, "learning_rate": 2.6704289799809345e-06, "loss": 3.1289, "step": 5812 }, { "epoch": 1.11, "learning_rate": 2.670371782650143e-06, "loss": 3.522, "step": 5813 }, { "epoch": 1.11, "learning_rate": 2.6703145853193517e-06, "loss": 3.382, "step": 5814 }, { "epoch": 1.11, "learning_rate": 2.6702573879885605e-06, "loss": 3.6993, "step": 5815 }, { "epoch": 1.11, "learning_rate": 2.6702001906577693e-06, "loss": 3.0512, "step": 5816 }, { "epoch": 1.11, "learning_rate": 2.670142993326978e-06, "loss": 3.4073, "step": 5817 }, { "epoch": 1.11, "learning_rate": 2.670085795996187e-06, "loss": 3.145, "step": 5818 }, { "epoch": 1.11, "learning_rate": 2.6700285986653957e-06, "loss": 2.9569, "step": 5819 }, { "epoch": 1.11, "learning_rate": 2.6699714013346045e-06, "loss": 3.1093, "step": 5820 }, { "epoch": 1.11, "learning_rate": 2.6699142040038132e-06, "loss": 3.0248, "step": 5821 }, { "epoch": 1.11, "learning_rate": 2.669857006673022e-06, "loss": 3.3306, "step": 5822 }, { "epoch": 1.11, "learning_rate": 2.669799809342231e-06, "loss": 3.4733, "step": 5823 }, { "epoch": 1.11, "learning_rate": 2.6697426120114396e-06, "loss": 3.1081, "step": 5824 }, { "epoch": 1.11, "learning_rate": 2.669685414680648e-06, "loss": 3.0564, "step": 5825 }, { "epoch": 1.11, "learning_rate": 2.6696282173498572e-06, "loss": 3.5551, "step": 5826 }, { "epoch": 1.11, "learning_rate": 2.669571020019066e-06, "loss": 3.3852, "step": 5827 }, { "epoch": 1.11, "learning_rate": 2.6695138226882744e-06, "loss": 2.9794, "step": 5828 }, { "epoch": 1.11, "learning_rate": 2.6694566253574836e-06, "loss": 3.9837, "step": 5829 }, { "epoch": 1.11, "learning_rate": 2.669399428026692e-06, "loss": 3.668, "step": 5830 }, { "epoch": 1.11, "learning_rate": 2.6693422306959008e-06, "loss": 2.817, "step": 5831 }, { "epoch": 1.11, "learning_rate": 2.6692850333651096e-06, "loss": 3.0559, "step": 5832 }, { "epoch": 1.11, "learning_rate": 2.6692278360343184e-06, "loss": 3.4919, "step": 5833 }, { "epoch": 1.11, "learning_rate": 2.6691706387035276e-06, "loss": 3.3875, "step": 5834 }, { "epoch": 1.11, "learning_rate": 2.669113441372736e-06, "loss": 3.3255, "step": 5835 }, { "epoch": 1.11, "learning_rate": 2.6690562440419447e-06, "loss": 3.62, "step": 5836 }, { "epoch": 1.11, "learning_rate": 2.6689990467111535e-06, "loss": 3.8705, "step": 5837 }, { "epoch": 1.11, "learning_rate": 2.6689418493803623e-06, "loss": 3.1561, "step": 5838 }, { "epoch": 1.11, "learning_rate": 2.668884652049571e-06, "loss": 3.2865, "step": 5839 }, { "epoch": 1.11, "learning_rate": 2.66882745471878e-06, "loss": 3.5209, "step": 5840 }, { "epoch": 1.11, "learning_rate": 2.6687702573879887e-06, "loss": 2.9703, "step": 5841 }, { "epoch": 1.11, "learning_rate": 2.6687130600571975e-06, "loss": 3.7096, "step": 5842 }, { "epoch": 1.11, "learning_rate": 2.6686558627264063e-06, "loss": 3.424, "step": 5843 }, { "epoch": 1.11, "learning_rate": 2.668598665395615e-06, "loss": 3.0591, "step": 5844 }, { "epoch": 1.11, "learning_rate": 2.6685414680648235e-06, "loss": 3.721, "step": 5845 }, { "epoch": 1.11, "learning_rate": 2.6684842707340327e-06, "loss": 3.5154, "step": 5846 }, { "epoch": 1.11, "learning_rate": 2.668427073403241e-06, "loss": 3.2111, "step": 5847 }, { "epoch": 1.11, "learning_rate": 2.66836987607245e-06, "loss": 3.5785, "step": 5848 }, { "epoch": 1.11, "learning_rate": 2.668312678741659e-06, "loss": 3.3075, "step": 5849 }, { "epoch": 1.11, "learning_rate": 2.6682554814108674e-06, "loss": 2.8221, "step": 5850 }, { "epoch": 1.11, "learning_rate": 2.6681982840800766e-06, "loss": 3.2682, "step": 5851 }, { "epoch": 1.11, "learning_rate": 2.668141086749285e-06, "loss": 3.1047, "step": 5852 }, { "epoch": 1.11, "learning_rate": 2.668083889418494e-06, "loss": 3.5779, "step": 5853 }, { "epoch": 1.12, "learning_rate": 2.6680266920877026e-06, "loss": 3.247, "step": 5854 }, { "epoch": 1.12, "learning_rate": 2.6679694947569114e-06, "loss": 3.5824, "step": 5855 }, { "epoch": 1.12, "learning_rate": 2.66791229742612e-06, "loss": 3.3724, "step": 5856 }, { "epoch": 1.12, "learning_rate": 2.667855100095329e-06, "loss": 3.482, "step": 5857 }, { "epoch": 1.12, "learning_rate": 2.6677979027645378e-06, "loss": 3.1947, "step": 5858 }, { "epoch": 1.12, "learning_rate": 2.6677407054337466e-06, "loss": 3.0499, "step": 5859 }, { "epoch": 1.12, "learning_rate": 2.6676835081029554e-06, "loss": 3.3046, "step": 5860 }, { "epoch": 1.12, "learning_rate": 2.667626310772164e-06, "loss": 3.2942, "step": 5861 }, { "epoch": 1.12, "learning_rate": 2.6675691134413725e-06, "loss": 3.349, "step": 5862 }, { "epoch": 1.12, "learning_rate": 2.6675119161105817e-06, "loss": 2.956, "step": 5863 }, { "epoch": 1.12, "learning_rate": 2.66745471877979e-06, "loss": 3.1776, "step": 5864 }, { "epoch": 1.12, "learning_rate": 2.667397521448999e-06, "loss": 3.2615, "step": 5865 }, { "epoch": 1.12, "learning_rate": 2.667340324118208e-06, "loss": 2.7871, "step": 5866 }, { "epoch": 1.12, "learning_rate": 2.6672831267874165e-06, "loss": 3.3477, "step": 5867 }, { "epoch": 1.12, "learning_rate": 2.6672259294566257e-06, "loss": 3.6687, "step": 5868 }, { "epoch": 1.12, "learning_rate": 2.667168732125834e-06, "loss": 3.2228, "step": 5869 }, { "epoch": 1.12, "learning_rate": 2.667111534795043e-06, "loss": 2.6788, "step": 5870 }, { "epoch": 1.12, "learning_rate": 2.6670543374642517e-06, "loss": 3.3627, "step": 5871 }, { "epoch": 1.12, "learning_rate": 2.6669971401334605e-06, "loss": 3.5112, "step": 5872 }, { "epoch": 1.12, "learning_rate": 2.6669399428026693e-06, "loss": 3.1877, "step": 5873 }, { "epoch": 1.12, "learning_rate": 2.666882745471878e-06, "loss": 3.1925, "step": 5874 }, { "epoch": 1.12, "learning_rate": 2.666825548141087e-06, "loss": 2.7856, "step": 5875 }, { "epoch": 1.12, "learning_rate": 2.6667683508102956e-06, "loss": 3.2041, "step": 5876 }, { "epoch": 1.12, "learning_rate": 2.6667111534795044e-06, "loss": 3.4822, "step": 5877 }, { "epoch": 1.12, "learning_rate": 2.6666539561487132e-06, "loss": 3.5772, "step": 5878 }, { "epoch": 1.12, "learning_rate": 2.6665967588179216e-06, "loss": 3.3664, "step": 5879 }, { "epoch": 1.12, "learning_rate": 2.666539561487131e-06, "loss": 3.3355, "step": 5880 }, { "epoch": 1.12, "learning_rate": 2.666482364156339e-06, "loss": 3.4803, "step": 5881 }, { "epoch": 1.12, "learning_rate": 2.666425166825548e-06, "loss": 3.3681, "step": 5882 }, { "epoch": 1.12, "learning_rate": 2.666367969494757e-06, "loss": 3.3188, "step": 5883 }, { "epoch": 1.12, "learning_rate": 2.6663107721639656e-06, "loss": 2.9047, "step": 5884 }, { "epoch": 1.12, "learning_rate": 2.6662535748331748e-06, "loss": 2.8866, "step": 5885 }, { "epoch": 1.12, "learning_rate": 2.666196377502383e-06, "loss": 3.3166, "step": 5886 }, { "epoch": 1.12, "learning_rate": 2.666139180171592e-06, "loss": 2.8733, "step": 5887 }, { "epoch": 1.12, "learning_rate": 2.666081982840801e-06, "loss": 3.5365, "step": 5888 }, { "epoch": 1.12, "learning_rate": 2.6660247855100095e-06, "loss": 3.3956, "step": 5889 }, { "epoch": 1.12, "learning_rate": 2.6659675881792183e-06, "loss": 3.6976, "step": 5890 }, { "epoch": 1.12, "learning_rate": 2.665910390848427e-06, "loss": 3.2383, "step": 5891 }, { "epoch": 1.12, "learning_rate": 2.665853193517636e-06, "loss": 3.4543, "step": 5892 }, { "epoch": 1.12, "learning_rate": 2.6657959961868447e-06, "loss": 3.4203, "step": 5893 }, { "epoch": 1.12, "learning_rate": 2.6657387988560535e-06, "loss": 3.7679, "step": 5894 }, { "epoch": 1.12, "learning_rate": 2.6656816015252623e-06, "loss": 3.0993, "step": 5895 }, { "epoch": 1.12, "learning_rate": 2.6656244041944707e-06, "loss": 2.8998, "step": 5896 }, { "epoch": 1.12, "learning_rate": 2.66556720686368e-06, "loss": 2.6228, "step": 5897 }, { "epoch": 1.12, "learning_rate": 2.6655100095328887e-06, "loss": 4.0099, "step": 5898 }, { "epoch": 1.12, "learning_rate": 2.6654528122020975e-06, "loss": 3.49, "step": 5899 }, { "epoch": 1.12, "learning_rate": 2.6653956148713063e-06, "loss": 3.1639, "step": 5900 }, { "epoch": 1.12, "learning_rate": 2.6653384175405146e-06, "loss": 2.7156, "step": 5901 }, { "epoch": 1.12, "learning_rate": 2.665281220209724e-06, "loss": 3.0397, "step": 5902 }, { "epoch": 1.12, "learning_rate": 2.6652240228789322e-06, "loss": 3.9002, "step": 5903 }, { "epoch": 1.12, "learning_rate": 2.665166825548141e-06, "loss": 3.0164, "step": 5904 }, { "epoch": 1.12, "learning_rate": 2.6651096282173502e-06, "loss": 3.2443, "step": 5905 }, { "epoch": 1.12, "learning_rate": 2.6650524308865586e-06, "loss": 2.8223, "step": 5906 }, { "epoch": 1.13, "learning_rate": 2.6649952335557674e-06, "loss": 3.7485, "step": 5907 }, { "epoch": 1.13, "learning_rate": 2.664938036224976e-06, "loss": 3.4981, "step": 5908 }, { "epoch": 1.13, "learning_rate": 2.664880838894185e-06, "loss": 3.5303, "step": 5909 }, { "epoch": 1.13, "learning_rate": 2.6648236415633938e-06, "loss": 3.4605, "step": 5910 }, { "epoch": 1.13, "learning_rate": 2.6647664442326026e-06, "loss": 3.4669, "step": 5911 }, { "epoch": 1.13, "learning_rate": 2.6647092469018114e-06, "loss": 3.526, "step": 5912 }, { "epoch": 1.13, "learning_rate": 2.6646520495710197e-06, "loss": 3.5634, "step": 5913 }, { "epoch": 1.13, "learning_rate": 2.664594852240229e-06, "loss": 2.9805, "step": 5914 }, { "epoch": 1.13, "learning_rate": 2.6645376549094378e-06, "loss": 3.2144, "step": 5915 }, { "epoch": 1.13, "learning_rate": 2.6644804575786465e-06, "loss": 3.1949, "step": 5916 }, { "epoch": 1.13, "learning_rate": 2.6644232602478553e-06, "loss": 3.2479, "step": 5917 }, { "epoch": 1.13, "learning_rate": 2.6643660629170637e-06, "loss": 3.0069, "step": 5918 }, { "epoch": 1.13, "learning_rate": 2.664308865586273e-06, "loss": 3.1119, "step": 5919 }, { "epoch": 1.13, "learning_rate": 2.6642516682554813e-06, "loss": 2.9284, "step": 5920 }, { "epoch": 1.13, "learning_rate": 2.66419447092469e-06, "loss": 3.5043, "step": 5921 }, { "epoch": 1.13, "learning_rate": 2.6641372735938993e-06, "loss": 3.5078, "step": 5922 }, { "epoch": 1.13, "learning_rate": 2.6640800762631077e-06, "loss": 3.2245, "step": 5923 }, { "epoch": 1.13, "learning_rate": 2.6640228789323165e-06, "loss": 3.9785, "step": 5924 }, { "epoch": 1.13, "learning_rate": 2.6639656816015253e-06, "loss": 3.6664, "step": 5925 }, { "epoch": 1.13, "learning_rate": 2.663908484270734e-06, "loss": 3.4421, "step": 5926 }, { "epoch": 1.13, "learning_rate": 2.663851286939943e-06, "loss": 3.9055, "step": 5927 }, { "epoch": 1.13, "learning_rate": 2.6637940896091516e-06, "loss": 3.3299, "step": 5928 }, { "epoch": 1.13, "learning_rate": 2.6637368922783604e-06, "loss": 3.1697, "step": 5929 }, { "epoch": 1.13, "learning_rate": 2.663679694947569e-06, "loss": 3.1852, "step": 5930 }, { "epoch": 1.13, "learning_rate": 2.663622497616778e-06, "loss": 3.0819, "step": 5931 }, { "epoch": 1.13, "learning_rate": 2.663565300285987e-06, "loss": 3.3519, "step": 5932 }, { "epoch": 1.13, "learning_rate": 2.6635081029551956e-06, "loss": 2.7783, "step": 5933 }, { "epoch": 1.13, "learning_rate": 2.6634509056244044e-06, "loss": 3.6897, "step": 5934 }, { "epoch": 1.13, "learning_rate": 2.6633937082936128e-06, "loss": 3.6629, "step": 5935 }, { "epoch": 1.13, "learning_rate": 2.663336510962822e-06, "loss": 3.0914, "step": 5936 }, { "epoch": 1.13, "learning_rate": 2.663279313632031e-06, "loss": 3.3047, "step": 5937 }, { "epoch": 1.13, "learning_rate": 2.663222116301239e-06, "loss": 3.4217, "step": 5938 }, { "epoch": 1.13, "learning_rate": 2.6631649189704484e-06, "loss": 3.6383, "step": 5939 }, { "epoch": 1.13, "learning_rate": 2.6631077216396567e-06, "loss": 3.3135, "step": 5940 }, { "epoch": 1.13, "learning_rate": 2.6630505243088655e-06, "loss": 3.103, "step": 5941 }, { "epoch": 1.13, "learning_rate": 2.6629933269780743e-06, "loss": 3.6687, "step": 5942 }, { "epoch": 1.13, "learning_rate": 2.662936129647283e-06, "loss": 2.692, "step": 5943 }, { "epoch": 1.13, "learning_rate": 2.662878932316492e-06, "loss": 3.1381, "step": 5944 }, { "epoch": 1.13, "learning_rate": 2.6628217349857007e-06, "loss": 2.9455, "step": 5945 }, { "epoch": 1.13, "learning_rate": 2.6627645376549095e-06, "loss": 3.2017, "step": 5946 }, { "epoch": 1.13, "learning_rate": 2.6627073403241183e-06, "loss": 3.4091, "step": 5947 }, { "epoch": 1.13, "learning_rate": 2.662650142993327e-06, "loss": 3.3257, "step": 5948 }, { "epoch": 1.13, "learning_rate": 2.662592945662536e-06, "loss": 3.2004, "step": 5949 }, { "epoch": 1.13, "learning_rate": 2.6625357483317447e-06, "loss": 3.1845, "step": 5950 }, { "epoch": 1.13, "learning_rate": 2.6624785510009535e-06, "loss": 2.7815, "step": 5951 }, { "epoch": 1.13, "learning_rate": 2.662421353670162e-06, "loss": 3.3657, "step": 5952 }, { "epoch": 1.13, "learning_rate": 2.662364156339371e-06, "loss": 2.5914, "step": 5953 }, { "epoch": 1.13, "learning_rate": 2.66230695900858e-06, "loss": 2.7727, "step": 5954 }, { "epoch": 1.13, "learning_rate": 2.6622497616777882e-06, "loss": 3.4562, "step": 5955 }, { "epoch": 1.13, "learning_rate": 2.6621925643469975e-06, "loss": 3.3731, "step": 5956 }, { "epoch": 1.13, "learning_rate": 2.662135367016206e-06, "loss": 3.2338, "step": 5957 }, { "epoch": 1.13, "learning_rate": 2.6620781696854146e-06, "loss": 3.1232, "step": 5958 }, { "epoch": 1.14, "learning_rate": 2.6620209723546234e-06, "loss": 3.6934, "step": 5959 }, { "epoch": 1.14, "learning_rate": 2.661963775023832e-06, "loss": 3.5286, "step": 5960 }, { "epoch": 1.14, "learning_rate": 2.661906577693041e-06, "loss": 4.0104, "step": 5961 }, { "epoch": 1.14, "learning_rate": 2.66184938036225e-06, "loss": 3.0847, "step": 5962 }, { "epoch": 1.14, "learning_rate": 2.6617921830314586e-06, "loss": 2.8916, "step": 5963 }, { "epoch": 1.14, "learning_rate": 2.6617349857006674e-06, "loss": 3.4605, "step": 5964 }, { "epoch": 1.14, "learning_rate": 2.661677788369876e-06, "loss": 2.875, "step": 5965 }, { "epoch": 1.14, "learning_rate": 2.661620591039085e-06, "loss": 2.8974, "step": 5966 }, { "epoch": 1.14, "learning_rate": 2.6615633937082938e-06, "loss": 2.8213, "step": 5967 }, { "epoch": 1.14, "learning_rate": 2.6615061963775026e-06, "loss": 3.3149, "step": 5968 }, { "epoch": 1.14, "learning_rate": 2.661448999046711e-06, "loss": 3.1793, "step": 5969 }, { "epoch": 1.14, "learning_rate": 2.66139180171592e-06, "loss": 2.8879, "step": 5970 }, { "epoch": 1.14, "learning_rate": 2.661334604385129e-06, "loss": 3.2752, "step": 5971 }, { "epoch": 1.14, "learning_rate": 2.6612774070543373e-06, "loss": 2.9614, "step": 5972 }, { "epoch": 1.14, "learning_rate": 2.6612202097235465e-06, "loss": 2.8635, "step": 5973 }, { "epoch": 1.14, "learning_rate": 2.661163012392755e-06, "loss": 2.8987, "step": 5974 }, { "epoch": 1.14, "learning_rate": 2.6611058150619637e-06, "loss": 3.7582, "step": 5975 }, { "epoch": 1.14, "learning_rate": 2.661048617731173e-06, "loss": 3.2271, "step": 5976 }, { "epoch": 1.14, "learning_rate": 2.6609914204003813e-06, "loss": 3.246, "step": 5977 }, { "epoch": 1.14, "learning_rate": 2.66093422306959e-06, "loss": 2.8383, "step": 5978 }, { "epoch": 1.14, "learning_rate": 2.660877025738799e-06, "loss": 3.2675, "step": 5979 }, { "epoch": 1.14, "learning_rate": 2.6608198284080077e-06, "loss": 3.1785, "step": 5980 }, { "epoch": 1.14, "learning_rate": 2.6607626310772164e-06, "loss": 3.3127, "step": 5981 }, { "epoch": 1.14, "learning_rate": 2.6607054337464252e-06, "loss": 2.7868, "step": 5982 }, { "epoch": 1.14, "learning_rate": 2.660648236415634e-06, "loss": 3.558, "step": 5983 }, { "epoch": 1.14, "learning_rate": 2.660591039084843e-06, "loss": 2.9047, "step": 5984 }, { "epoch": 1.14, "learning_rate": 2.6605338417540516e-06, "loss": 3.707, "step": 5985 }, { "epoch": 1.14, "learning_rate": 2.6604766444232604e-06, "loss": 3.163, "step": 5986 }, { "epoch": 1.14, "learning_rate": 2.6604194470924692e-06, "loss": 2.8377, "step": 5987 }, { "epoch": 1.14, "learning_rate": 2.660362249761678e-06, "loss": 3.3155, "step": 5988 }, { "epoch": 1.14, "learning_rate": 2.6603050524308864e-06, "loss": 3.2728, "step": 5989 }, { "epoch": 1.14, "learning_rate": 2.6602478551000956e-06, "loss": 3.1859, "step": 5990 }, { "epoch": 1.14, "learning_rate": 2.660190657769304e-06, "loss": 3.1541, "step": 5991 }, { "epoch": 1.14, "learning_rate": 2.6601334604385128e-06, "loss": 3.3403, "step": 5992 }, { "epoch": 1.14, "learning_rate": 2.660076263107722e-06, "loss": 3.2078, "step": 5993 }, { "epoch": 1.14, "learning_rate": 2.6600190657769303e-06, "loss": 2.6497, "step": 5994 }, { "epoch": 1.14, "learning_rate": 2.659961868446139e-06, "loss": 3.0712, "step": 5995 }, { "epoch": 1.14, "learning_rate": 2.659904671115348e-06, "loss": 3.6374, "step": 5996 }, { "epoch": 1.14, "learning_rate": 2.6598474737845567e-06, "loss": 3.5541, "step": 5997 }, { "epoch": 1.14, "learning_rate": 2.6597902764537655e-06, "loss": 3.67, "step": 5998 }, { "epoch": 1.14, "learning_rate": 2.6597330791229743e-06, "loss": 2.7384, "step": 5999 }, { "epoch": 1.14, "learning_rate": 2.659675881792183e-06, "loss": 3.2154, "step": 6000 }, { "epoch": 1.14, "learning_rate": 2.659618684461392e-06, "loss": 3.6317, "step": 6001 }, { "epoch": 1.14, "learning_rate": 2.6595614871306007e-06, "loss": 3.1118, "step": 6002 }, { "epoch": 1.14, "learning_rate": 2.6595042897998095e-06, "loss": 3.2344, "step": 6003 }, { "epoch": 1.14, "learning_rate": 2.6594470924690183e-06, "loss": 3.0848, "step": 6004 }, { "epoch": 1.14, "learning_rate": 2.659389895138227e-06, "loss": 3.1451, "step": 6005 }, { "epoch": 1.14, "learning_rate": 2.6593326978074354e-06, "loss": 3.441, "step": 6006 }, { "epoch": 1.14, "learning_rate": 2.6592755004766447e-06, "loss": 2.4361, "step": 6007 }, { "epoch": 1.14, "learning_rate": 2.659218303145853e-06, "loss": 3.0155, "step": 6008 }, { "epoch": 1.14, "learning_rate": 2.659161105815062e-06, "loss": 3.1747, "step": 6009 }, { "epoch": 1.14, "learning_rate": 2.659103908484271e-06, "loss": 3.4098, "step": 6010 }, { "epoch": 1.14, "learning_rate": 2.6590467111534794e-06, "loss": 3.0229, "step": 6011 }, { "epoch": 1.15, "learning_rate": 2.658989513822688e-06, "loss": 3.3068, "step": 6012 }, { "epoch": 1.15, "learning_rate": 2.658932316491897e-06, "loss": 3.5532, "step": 6013 }, { "epoch": 1.15, "learning_rate": 2.658875119161106e-06, "loss": 3.3517, "step": 6014 }, { "epoch": 1.15, "learning_rate": 2.6588179218303146e-06, "loss": 3.0731, "step": 6015 }, { "epoch": 1.15, "learning_rate": 2.6587607244995234e-06, "loss": 3.8607, "step": 6016 }, { "epoch": 1.15, "learning_rate": 2.658703527168732e-06, "loss": 3.2866, "step": 6017 }, { "epoch": 1.15, "learning_rate": 2.658646329837941e-06, "loss": 2.8784, "step": 6018 }, { "epoch": 1.15, "learning_rate": 2.6585891325071498e-06, "loss": 2.7954, "step": 6019 }, { "epoch": 1.15, "learning_rate": 2.6585319351763586e-06, "loss": 3.031, "step": 6020 }, { "epoch": 1.15, "learning_rate": 2.6584747378455674e-06, "loss": 2.7102, "step": 6021 }, { "epoch": 1.15, "learning_rate": 2.658417540514776e-06, "loss": 3.8015, "step": 6022 }, { "epoch": 1.15, "learning_rate": 2.6583603431839845e-06, "loss": 3.7307, "step": 6023 }, { "epoch": 1.15, "learning_rate": 2.6583031458531937e-06, "loss": 3.3871, "step": 6024 }, { "epoch": 1.15, "learning_rate": 2.6582459485224025e-06, "loss": 3.2809, "step": 6025 }, { "epoch": 1.15, "learning_rate": 2.658188751191611e-06, "loss": 3.3201, "step": 6026 }, { "epoch": 1.15, "learning_rate": 2.65813155386082e-06, "loss": 2.815, "step": 6027 }, { "epoch": 1.15, "learning_rate": 2.6580743565300285e-06, "loss": 3.4005, "step": 6028 }, { "epoch": 1.15, "learning_rate": 2.6580171591992377e-06, "loss": 3.1464, "step": 6029 }, { "epoch": 1.15, "learning_rate": 2.657959961868446e-06, "loss": 3.2367, "step": 6030 }, { "epoch": 1.15, "learning_rate": 2.657902764537655e-06, "loss": 3.2005, "step": 6031 }, { "epoch": 1.15, "learning_rate": 2.657845567206864e-06, "loss": 3.5139, "step": 6032 }, { "epoch": 1.15, "learning_rate": 2.6577883698760725e-06, "loss": 3.6775, "step": 6033 }, { "epoch": 1.15, "learning_rate": 2.6577311725452813e-06, "loss": 3.4156, "step": 6034 }, { "epoch": 1.15, "learning_rate": 2.65767397521449e-06, "loss": 2.5232, "step": 6035 }, { "epoch": 1.15, "learning_rate": 2.657616777883699e-06, "loss": 3.271, "step": 6036 }, { "epoch": 1.15, "learning_rate": 2.6575595805529076e-06, "loss": 3.1228, "step": 6037 }, { "epoch": 1.15, "learning_rate": 2.6575023832221164e-06, "loss": 3.0997, "step": 6038 }, { "epoch": 1.15, "learning_rate": 2.6574451858913252e-06, "loss": 3.6606, "step": 6039 }, { "epoch": 1.15, "learning_rate": 2.6573879885605336e-06, "loss": 3.7033, "step": 6040 }, { "epoch": 1.15, "learning_rate": 2.657330791229743e-06, "loss": 3.6583, "step": 6041 }, { "epoch": 1.15, "learning_rate": 2.6572735938989516e-06, "loss": 3.3933, "step": 6042 }, { "epoch": 1.15, "learning_rate": 2.65721639656816e-06, "loss": 3.6031, "step": 6043 }, { "epoch": 1.15, "learning_rate": 2.657159199237369e-06, "loss": 3.8238, "step": 6044 }, { "epoch": 1.15, "learning_rate": 2.6571020019065776e-06, "loss": 3.3331, "step": 6045 }, { "epoch": 1.15, "learning_rate": 2.6570448045757868e-06, "loss": 3.1895, "step": 6046 }, { "epoch": 1.15, "learning_rate": 2.656987607244995e-06, "loss": 3.2692, "step": 6047 }, { "epoch": 1.15, "learning_rate": 2.656930409914204e-06, "loss": 2.6499, "step": 6048 }, { "epoch": 1.15, "learning_rate": 2.656873212583413e-06, "loss": 3.1917, "step": 6049 }, { "epoch": 1.15, "learning_rate": 2.6568160152526215e-06, "loss": 3.6908, "step": 6050 }, { "epoch": 1.15, "learning_rate": 2.6567588179218303e-06, "loss": 3.0344, "step": 6051 }, { "epoch": 1.15, "learning_rate": 2.656701620591039e-06, "loss": 3.205, "step": 6052 }, { "epoch": 1.15, "learning_rate": 2.656644423260248e-06, "loss": 3.0031, "step": 6053 }, { "epoch": 1.15, "learning_rate": 2.6565872259294567e-06, "loss": 3.6553, "step": 6054 }, { "epoch": 1.15, "learning_rate": 2.6565300285986655e-06, "loss": 3.5925, "step": 6055 }, { "epoch": 1.15, "learning_rate": 2.6564728312678743e-06, "loss": 2.8737, "step": 6056 }, { "epoch": 1.15, "learning_rate": 2.6564156339370827e-06, "loss": 3.2809, "step": 6057 }, { "epoch": 1.15, "learning_rate": 2.656358436606292e-06, "loss": 3.1346, "step": 6058 }, { "epoch": 1.15, "learning_rate": 2.6563012392755007e-06, "loss": 3.8685, "step": 6059 }, { "epoch": 1.15, "learning_rate": 2.656244041944709e-06, "loss": 2.8737, "step": 6060 }, { "epoch": 1.15, "learning_rate": 2.6561868446139183e-06, "loss": 2.9239, "step": 6061 }, { "epoch": 1.15, "learning_rate": 2.6561296472831266e-06, "loss": 3.5412, "step": 6062 }, { "epoch": 1.15, "learning_rate": 2.656072449952336e-06, "loss": 2.493, "step": 6063 }, { "epoch": 1.16, "learning_rate": 2.6560152526215446e-06, "loss": 2.8621, "step": 6064 }, { "epoch": 1.16, "learning_rate": 2.655958055290753e-06, "loss": 3.1782, "step": 6065 }, { "epoch": 1.16, "learning_rate": 2.6559008579599622e-06, "loss": 2.9553, "step": 6066 }, { "epoch": 1.16, "learning_rate": 2.6558436606291706e-06, "loss": 3.7765, "step": 6067 }, { "epoch": 1.16, "learning_rate": 2.6557864632983794e-06, "loss": 3.0451, "step": 6068 }, { "epoch": 1.16, "learning_rate": 2.655729265967588e-06, "loss": 3.8272, "step": 6069 }, { "epoch": 1.16, "learning_rate": 2.655672068636797e-06, "loss": 3.4144, "step": 6070 }, { "epoch": 1.16, "learning_rate": 2.6556148713060058e-06, "loss": 2.7085, "step": 6071 }, { "epoch": 1.16, "learning_rate": 2.6555576739752146e-06, "loss": 2.5033, "step": 6072 }, { "epoch": 1.16, "learning_rate": 2.6555004766444234e-06, "loss": 2.6511, "step": 6073 }, { "epoch": 1.16, "learning_rate": 2.655443279313632e-06, "loss": 3.6683, "step": 6074 }, { "epoch": 1.16, "learning_rate": 2.655386081982841e-06, "loss": 3.6244, "step": 6075 }, { "epoch": 1.16, "learning_rate": 2.6553288846520497e-06, "loss": 2.839, "step": 6076 }, { "epoch": 1.16, "learning_rate": 2.655271687321258e-06, "loss": 2.4326, "step": 6077 }, { "epoch": 1.16, "learning_rate": 2.6552144899904673e-06, "loss": 3.7658, "step": 6078 }, { "epoch": 1.16, "learning_rate": 2.6551572926596757e-06, "loss": 2.9191, "step": 6079 }, { "epoch": 1.16, "learning_rate": 2.655100095328885e-06, "loss": 3.2174, "step": 6080 }, { "epoch": 1.16, "learning_rate": 2.6550428979980937e-06, "loss": 3.0363, "step": 6081 }, { "epoch": 1.16, "learning_rate": 2.654985700667302e-06, "loss": 3.5411, "step": 6082 }, { "epoch": 1.16, "learning_rate": 2.6549285033365113e-06, "loss": 2.9811, "step": 6083 }, { "epoch": 1.16, "learning_rate": 2.6548713060057197e-06, "loss": 2.9986, "step": 6084 }, { "epoch": 1.16, "learning_rate": 2.6548141086749285e-06, "loss": 3.6379, "step": 6085 }, { "epoch": 1.16, "learning_rate": 2.6547569113441373e-06, "loss": 3.4463, "step": 6086 }, { "epoch": 1.16, "learning_rate": 2.654699714013346e-06, "loss": 3.0238, "step": 6087 }, { "epoch": 1.16, "learning_rate": 2.654642516682555e-06, "loss": 3.1662, "step": 6088 }, { "epoch": 1.16, "learning_rate": 2.6545853193517636e-06, "loss": 3.9141, "step": 6089 }, { "epoch": 1.16, "learning_rate": 2.6545281220209724e-06, "loss": 2.7297, "step": 6090 }, { "epoch": 1.16, "learning_rate": 2.6544709246901812e-06, "loss": 3.5331, "step": 6091 }, { "epoch": 1.16, "learning_rate": 2.65441372735939e-06, "loss": 3.4562, "step": 6092 }, { "epoch": 1.16, "learning_rate": 2.654356530028599e-06, "loss": 3.5614, "step": 6093 }, { "epoch": 1.16, "learning_rate": 2.6542993326978076e-06, "loss": 3.3436, "step": 6094 }, { "epoch": 1.16, "learning_rate": 2.6542421353670164e-06, "loss": 3.2471, "step": 6095 }, { "epoch": 1.16, "learning_rate": 2.6541849380362248e-06, "loss": 3.4656, "step": 6096 }, { "epoch": 1.16, "learning_rate": 2.654127740705434e-06, "loss": 2.5851, "step": 6097 }, { "epoch": 1.16, "learning_rate": 2.6540705433746428e-06, "loss": 3.3055, "step": 6098 }, { "epoch": 1.16, "learning_rate": 2.654013346043851e-06, "loss": 3.5825, "step": 6099 }, { "epoch": 1.16, "learning_rate": 2.6539561487130604e-06, "loss": 3.2701, "step": 6100 }, { "epoch": 1.16, "learning_rate": 2.6538989513822687e-06, "loss": 2.819, "step": 6101 }, { "epoch": 1.16, "learning_rate": 2.6538417540514775e-06, "loss": 3.3451, "step": 6102 }, { "epoch": 1.16, "learning_rate": 2.6537845567206863e-06, "loss": 3.4415, "step": 6103 }, { "epoch": 1.16, "learning_rate": 2.653727359389895e-06, "loss": 3.5731, "step": 6104 }, { "epoch": 1.16, "learning_rate": 2.653670162059104e-06, "loss": 3.355, "step": 6105 }, { "epoch": 1.16, "learning_rate": 2.6536129647283127e-06, "loss": 3.8967, "step": 6106 }, { "epoch": 1.16, "learning_rate": 2.6535557673975215e-06, "loss": 3.376, "step": 6107 }, { "epoch": 1.16, "learning_rate": 2.6534985700667303e-06, "loss": 3.3111, "step": 6108 }, { "epoch": 1.16, "learning_rate": 2.653441372735939e-06, "loss": 3.0006, "step": 6109 }, { "epoch": 1.16, "learning_rate": 2.653384175405148e-06, "loss": 3.0349, "step": 6110 }, { "epoch": 1.16, "learning_rate": 2.6533269780743567e-06, "loss": 3.292, "step": 6111 }, { "epoch": 1.16, "learning_rate": 2.6532697807435655e-06, "loss": 3.8034, "step": 6112 }, { "epoch": 1.16, "learning_rate": 2.6532125834127743e-06, "loss": 3.3458, "step": 6113 }, { "epoch": 1.16, "learning_rate": 2.653155386081983e-06, "loss": 3.139, "step": 6114 }, { "epoch": 1.16, "learning_rate": 2.653098188751192e-06, "loss": 2.8197, "step": 6115 }, { "epoch": 1.16, "learning_rate": 2.6530409914204002e-06, "loss": 3.3877, "step": 6116 }, { "epoch": 1.17, "learning_rate": 2.6529837940896094e-06, "loss": 3.3206, "step": 6117 }, { "epoch": 1.17, "learning_rate": 2.652926596758818e-06, "loss": 3.1214, "step": 6118 }, { "epoch": 1.17, "learning_rate": 2.6528693994280266e-06, "loss": 3.19, "step": 6119 }, { "epoch": 1.17, "learning_rate": 2.652812202097236e-06, "loss": 2.8085, "step": 6120 }, { "epoch": 1.17, "learning_rate": 2.652755004766444e-06, "loss": 3.5947, "step": 6121 }, { "epoch": 1.17, "learning_rate": 2.652697807435653e-06, "loss": 3.4836, "step": 6122 }, { "epoch": 1.17, "learning_rate": 2.6526406101048618e-06, "loss": 3.2474, "step": 6123 }, { "epoch": 1.17, "learning_rate": 2.6525834127740706e-06, "loss": 3.1019, "step": 6124 }, { "epoch": 1.17, "learning_rate": 2.6525262154432794e-06, "loss": 3.4294, "step": 6125 }, { "epoch": 1.17, "learning_rate": 2.652469018112488e-06, "loss": 3.18, "step": 6126 }, { "epoch": 1.17, "learning_rate": 2.652411820781697e-06, "loss": 3.3859, "step": 6127 }, { "epoch": 1.17, "learning_rate": 2.6523546234509058e-06, "loss": 3.3057, "step": 6128 }, { "epoch": 1.17, "learning_rate": 2.6522974261201145e-06, "loss": 3.5753, "step": 6129 }, { "epoch": 1.17, "learning_rate": 2.6522402287893233e-06, "loss": 3.6291, "step": 6130 }, { "epoch": 1.17, "learning_rate": 2.652183031458532e-06, "loss": 2.885, "step": 6131 }, { "epoch": 1.17, "learning_rate": 2.652125834127741e-06, "loss": 2.9027, "step": 6132 }, { "epoch": 1.17, "learning_rate": 2.6520686367969493e-06, "loss": 3.5295, "step": 6133 }, { "epoch": 1.17, "learning_rate": 2.6520114394661585e-06, "loss": 3.359, "step": 6134 }, { "epoch": 1.17, "learning_rate": 2.651954242135367e-06, "loss": 3.4262, "step": 6135 }, { "epoch": 1.17, "learning_rate": 2.6518970448045757e-06, "loss": 3.0955, "step": 6136 }, { "epoch": 1.17, "learning_rate": 2.651839847473785e-06, "loss": 3.412, "step": 6137 }, { "epoch": 1.17, "learning_rate": 2.6517826501429933e-06, "loss": 3.1615, "step": 6138 }, { "epoch": 1.17, "learning_rate": 2.651725452812202e-06, "loss": 3.226, "step": 6139 }, { "epoch": 1.17, "learning_rate": 2.651668255481411e-06, "loss": 3.1197, "step": 6140 }, { "epoch": 1.17, "learning_rate": 2.6516110581506196e-06, "loss": 2.7089, "step": 6141 }, { "epoch": 1.17, "learning_rate": 2.6515538608198284e-06, "loss": 3.3038, "step": 6142 }, { "epoch": 1.17, "learning_rate": 2.6514966634890372e-06, "loss": 3.5336, "step": 6143 }, { "epoch": 1.17, "learning_rate": 2.651439466158246e-06, "loss": 3.2285, "step": 6144 }, { "epoch": 1.17, "learning_rate": 2.651382268827455e-06, "loss": 2.9084, "step": 6145 }, { "epoch": 1.17, "learning_rate": 2.6513250714966636e-06, "loss": 3.1285, "step": 6146 }, { "epoch": 1.17, "learning_rate": 2.6512678741658724e-06, "loss": 3.4613, "step": 6147 }, { "epoch": 1.17, "learning_rate": 2.651210676835081e-06, "loss": 2.9986, "step": 6148 }, { "epoch": 1.17, "learning_rate": 2.65115347950429e-06, "loss": 3.5219, "step": 6149 }, { "epoch": 1.17, "learning_rate": 2.6510962821734984e-06, "loss": 2.892, "step": 6150 }, { "epoch": 1.17, "learning_rate": 2.6510390848427076e-06, "loss": 2.9889, "step": 6151 }, { "epoch": 1.17, "learning_rate": 2.6509818875119164e-06, "loss": 3.2236, "step": 6152 }, { "epoch": 1.17, "learning_rate": 2.6509246901811248e-06, "loss": 3.1883, "step": 6153 }, { "epoch": 1.17, "learning_rate": 2.650867492850334e-06, "loss": 2.6539, "step": 6154 }, { "epoch": 1.17, "learning_rate": 2.6508102955195423e-06, "loss": 3.3561, "step": 6155 }, { "epoch": 1.17, "learning_rate": 2.650753098188751e-06, "loss": 3.5791, "step": 6156 }, { "epoch": 1.17, "learning_rate": 2.65069590085796e-06, "loss": 2.91, "step": 6157 }, { "epoch": 1.17, "learning_rate": 2.6506387035271687e-06, "loss": 3.2653, "step": 6158 }, { "epoch": 1.17, "learning_rate": 2.650581506196378e-06, "loss": 3.4066, "step": 6159 }, { "epoch": 1.17, "learning_rate": 2.6505243088655863e-06, "loss": 2.8227, "step": 6160 }, { "epoch": 1.17, "learning_rate": 2.650467111534795e-06, "loss": 3.5526, "step": 6161 }, { "epoch": 1.17, "learning_rate": 2.650409914204004e-06, "loss": 2.8678, "step": 6162 }, { "epoch": 1.17, "learning_rate": 2.6503527168732127e-06, "loss": 3.531, "step": 6163 }, { "epoch": 1.17, "learning_rate": 2.6502955195424215e-06, "loss": 3.4746, "step": 6164 }, { "epoch": 1.17, "learning_rate": 2.6502383222116303e-06, "loss": 3.6094, "step": 6165 }, { "epoch": 1.17, "learning_rate": 2.650181124880839e-06, "loss": 3.4509, "step": 6166 }, { "epoch": 1.17, "learning_rate": 2.6501239275500474e-06, "loss": 3.3202, "step": 6167 }, { "epoch": 1.17, "learning_rate": 2.6500667302192567e-06, "loss": 3.3465, "step": 6168 }, { "epoch": 1.18, "learning_rate": 2.6500095328884655e-06, "loss": 3.2496, "step": 6169 }, { "epoch": 1.18, "learning_rate": 2.649952335557674e-06, "loss": 3.1925, "step": 6170 }, { "epoch": 1.18, "learning_rate": 2.649895138226883e-06, "loss": 2.6525, "step": 6171 }, { "epoch": 1.18, "learning_rate": 2.6498379408960914e-06, "loss": 3.0896, "step": 6172 }, { "epoch": 1.18, "learning_rate": 2.6497807435653e-06, "loss": 3.012, "step": 6173 }, { "epoch": 1.18, "learning_rate": 2.649723546234509e-06, "loss": 2.9153, "step": 6174 }, { "epoch": 1.18, "learning_rate": 2.649666348903718e-06, "loss": 3.2764, "step": 6175 }, { "epoch": 1.18, "learning_rate": 2.649609151572927e-06, "loss": 3.304, "step": 6176 }, { "epoch": 1.18, "learning_rate": 2.6495519542421354e-06, "loss": 2.8952, "step": 6177 }, { "epoch": 1.18, "learning_rate": 2.649494756911344e-06, "loss": 3.4965, "step": 6178 }, { "epoch": 1.18, "learning_rate": 2.649437559580553e-06, "loss": 3.1298, "step": 6179 }, { "epoch": 1.18, "learning_rate": 2.6493803622497618e-06, "loss": 3.0381, "step": 6180 }, { "epoch": 1.18, "learning_rate": 2.6493231649189706e-06, "loss": 3.7185, "step": 6181 }, { "epoch": 1.18, "learning_rate": 2.6492659675881793e-06, "loss": 3.3141, "step": 6182 }, { "epoch": 1.18, "learning_rate": 2.649208770257388e-06, "loss": 3.0802, "step": 6183 }, { "epoch": 1.18, "learning_rate": 2.6491515729265965e-06, "loss": 2.5961, "step": 6184 }, { "epoch": 1.18, "learning_rate": 2.6490943755958057e-06, "loss": 2.9295, "step": 6185 }, { "epoch": 1.18, "learning_rate": 2.6490371782650145e-06, "loss": 2.6195, "step": 6186 }, { "epoch": 1.18, "learning_rate": 2.648979980934223e-06, "loss": 3.5244, "step": 6187 }, { "epoch": 1.18, "learning_rate": 2.648922783603432e-06, "loss": 3.1793, "step": 6188 }, { "epoch": 1.18, "learning_rate": 2.6488655862726405e-06, "loss": 3.4327, "step": 6189 }, { "epoch": 1.18, "learning_rate": 2.6488083889418493e-06, "loss": 2.934, "step": 6190 }, { "epoch": 1.18, "learning_rate": 2.648751191611058e-06, "loss": 2.9836, "step": 6191 }, { "epoch": 1.18, "learning_rate": 2.648693994280267e-06, "loss": 3.4697, "step": 6192 }, { "epoch": 1.18, "learning_rate": 2.648636796949476e-06, "loss": 2.7444, "step": 6193 }, { "epoch": 1.18, "learning_rate": 2.6485795996186845e-06, "loss": 2.9561, "step": 6194 }, { "epoch": 1.18, "learning_rate": 2.6485224022878932e-06, "loss": 3.1495, "step": 6195 }, { "epoch": 1.18, "learning_rate": 2.648465204957102e-06, "loss": 3.365, "step": 6196 }, { "epoch": 1.18, "learning_rate": 2.648408007626311e-06, "loss": 3.0128, "step": 6197 }, { "epoch": 1.18, "learning_rate": 2.6483508102955196e-06, "loss": 3.1216, "step": 6198 }, { "epoch": 1.18, "learning_rate": 2.6482936129647284e-06, "loss": 3.1706, "step": 6199 }, { "epoch": 1.18, "learning_rate": 2.6482364156339372e-06, "loss": 3.3476, "step": 6200 }, { "epoch": 1.18, "learning_rate": 2.648179218303146e-06, "loss": 2.8375, "step": 6201 }, { "epoch": 1.18, "learning_rate": 2.648122020972355e-06, "loss": 2.8534, "step": 6202 }, { "epoch": 1.18, "learning_rate": 2.6480648236415636e-06, "loss": 3.3482, "step": 6203 }, { "epoch": 1.18, "learning_rate": 2.648007626310772e-06, "loss": 3.1707, "step": 6204 }, { "epoch": 1.18, "learning_rate": 2.647950428979981e-06, "loss": 3.652, "step": 6205 }, { "epoch": 1.18, "learning_rate": 2.6478932316491896e-06, "loss": 3.55, "step": 6206 }, { "epoch": 1.18, "learning_rate": 2.6478360343183983e-06, "loss": 3.9119, "step": 6207 }, { "epoch": 1.18, "learning_rate": 2.6477788369876076e-06, "loss": 2.7558, "step": 6208 }, { "epoch": 1.18, "learning_rate": 2.647721639656816e-06, "loss": 3.2093, "step": 6209 }, { "epoch": 1.18, "learning_rate": 2.647664442326025e-06, "loss": 3.5133, "step": 6210 }, { "epoch": 1.18, "learning_rate": 2.6476072449952335e-06, "loss": 3.1398, "step": 6211 }, { "epoch": 1.18, "learning_rate": 2.6475500476644423e-06, "loss": 3.264, "step": 6212 }, { "epoch": 1.18, "learning_rate": 2.647492850333651e-06, "loss": 3.2597, "step": 6213 }, { "epoch": 1.18, "learning_rate": 2.64743565300286e-06, "loss": 3.4195, "step": 6214 }, { "epoch": 1.18, "learning_rate": 2.6473784556720687e-06, "loss": 3.769, "step": 6215 }, { "epoch": 1.18, "learning_rate": 2.6473212583412775e-06, "loss": 3.7476, "step": 6216 }, { "epoch": 1.18, "learning_rate": 2.6472640610104863e-06, "loss": 3.1873, "step": 6217 }, { "epoch": 1.18, "learning_rate": 2.647206863679695e-06, "loss": 3.2726, "step": 6218 }, { "epoch": 1.18, "learning_rate": 2.647149666348904e-06, "loss": 3.1597, "step": 6219 }, { "epoch": 1.18, "learning_rate": 2.6470924690181127e-06, "loss": 3.5135, "step": 6220 }, { "epoch": 1.18, "learning_rate": 2.647035271687321e-06, "loss": 3.4064, "step": 6221 }, { "epoch": 1.19, "learning_rate": 2.6469780743565303e-06, "loss": 3.0558, "step": 6222 }, { "epoch": 1.19, "learning_rate": 2.6469208770257386e-06, "loss": 2.8908, "step": 6223 }, { "epoch": 1.19, "learning_rate": 2.646863679694948e-06, "loss": 3.2729, "step": 6224 }, { "epoch": 1.19, "learning_rate": 2.6468064823641566e-06, "loss": 3.2116, "step": 6225 }, { "epoch": 1.19, "learning_rate": 2.646749285033365e-06, "loss": 3.1659, "step": 6226 }, { "epoch": 1.19, "learning_rate": 2.6466920877025742e-06, "loss": 3.3142, "step": 6227 }, { "epoch": 1.19, "learning_rate": 2.6466348903717826e-06, "loss": 3.4424, "step": 6228 }, { "epoch": 1.19, "learning_rate": 2.6465776930409914e-06, "loss": 3.0197, "step": 6229 }, { "epoch": 1.19, "learning_rate": 2.6465204957102e-06, "loss": 3.4653, "step": 6230 }, { "epoch": 1.19, "learning_rate": 2.646463298379409e-06, "loss": 3.5868, "step": 6231 }, { "epoch": 1.19, "learning_rate": 2.6464061010486178e-06, "loss": 3.6373, "step": 6232 }, { "epoch": 1.19, "learning_rate": 2.6463489037178266e-06, "loss": 2.8733, "step": 6233 }, { "epoch": 1.19, "learning_rate": 2.6462917063870354e-06, "loss": 2.911, "step": 6234 }, { "epoch": 1.19, "learning_rate": 2.646234509056244e-06, "loss": 3.5549, "step": 6235 }, { "epoch": 1.19, "learning_rate": 2.646177311725453e-06, "loss": 2.6809, "step": 6236 }, { "epoch": 1.19, "learning_rate": 2.6461201143946617e-06, "loss": 3.2143, "step": 6237 }, { "epoch": 1.19, "learning_rate": 2.64606291706387e-06, "loss": 3.4763, "step": 6238 }, { "epoch": 1.19, "learning_rate": 2.6460057197330793e-06, "loss": 3.1956, "step": 6239 }, { "epoch": 1.19, "learning_rate": 2.6459485224022877e-06, "loss": 2.9322, "step": 6240 }, { "epoch": 1.19, "learning_rate": 2.645891325071497e-06, "loss": 3.8655, "step": 6241 }, { "epoch": 1.19, "learning_rate": 2.6458341277407057e-06, "loss": 3.7891, "step": 6242 }, { "epoch": 1.19, "learning_rate": 2.645776930409914e-06, "loss": 3.1811, "step": 6243 }, { "epoch": 1.19, "learning_rate": 2.6457197330791233e-06, "loss": 3.0822, "step": 6244 }, { "epoch": 1.19, "learning_rate": 2.6456625357483317e-06, "loss": 3.2385, "step": 6245 }, { "epoch": 1.19, "learning_rate": 2.6456053384175405e-06, "loss": 3.7493, "step": 6246 }, { "epoch": 1.19, "learning_rate": 2.6455481410867497e-06, "loss": 3.1247, "step": 6247 }, { "epoch": 1.19, "learning_rate": 2.645490943755958e-06, "loss": 3.0253, "step": 6248 }, { "epoch": 1.19, "learning_rate": 2.645433746425167e-06, "loss": 3.5613, "step": 6249 }, { "epoch": 1.19, "learning_rate": 2.6453765490943756e-06, "loss": 3.1538, "step": 6250 }, { "epoch": 1.19, "learning_rate": 2.6453193517635844e-06, "loss": 2.9442, "step": 6251 }, { "epoch": 1.19, "learning_rate": 2.6452621544327932e-06, "loss": 3.4913, "step": 6252 }, { "epoch": 1.19, "learning_rate": 2.645204957102002e-06, "loss": 3.3333, "step": 6253 }, { "epoch": 1.19, "learning_rate": 2.645147759771211e-06, "loss": 2.9827, "step": 6254 }, { "epoch": 1.19, "learning_rate": 2.645090562440419e-06, "loss": 2.9262, "step": 6255 }, { "epoch": 1.19, "learning_rate": 2.6450333651096284e-06, "loss": 3.8184, "step": 6256 }, { "epoch": 1.19, "learning_rate": 2.644976167778837e-06, "loss": 3.0502, "step": 6257 }, { "epoch": 1.19, "learning_rate": 2.644918970448046e-06, "loss": 2.6236, "step": 6258 }, { "epoch": 1.19, "learning_rate": 2.6448617731172548e-06, "loss": 3.3798, "step": 6259 }, { "epoch": 1.19, "learning_rate": 2.644804575786463e-06, "loss": 3.7895, "step": 6260 }, { "epoch": 1.19, "learning_rate": 2.6447473784556724e-06, "loss": 3.1179, "step": 6261 }, { "epoch": 1.19, "learning_rate": 2.6446901811248807e-06, "loss": 3.6466, "step": 6262 }, { "epoch": 1.19, "learning_rate": 2.6446329837940895e-06, "loss": 3.3285, "step": 6263 }, { "epoch": 1.19, "learning_rate": 2.6445757864632987e-06, "loss": 3.1593, "step": 6264 }, { "epoch": 1.19, "learning_rate": 2.644518589132507e-06, "loss": 3.2706, "step": 6265 }, { "epoch": 1.19, "learning_rate": 2.644461391801716e-06, "loss": 2.7897, "step": 6266 }, { "epoch": 1.19, "learning_rate": 2.6444041944709247e-06, "loss": 2.6984, "step": 6267 }, { "epoch": 1.19, "learning_rate": 2.6443469971401335e-06, "loss": 3.1095, "step": 6268 }, { "epoch": 1.19, "learning_rate": 2.6442897998093423e-06, "loss": 2.7653, "step": 6269 }, { "epoch": 1.19, "learning_rate": 2.644232602478551e-06, "loss": 3.3578, "step": 6270 }, { "epoch": 1.19, "learning_rate": 2.64417540514776e-06, "loss": 3.3464, "step": 6271 }, { "epoch": 1.19, "learning_rate": 2.6441182078169683e-06, "loss": 3.4997, "step": 6272 }, { "epoch": 1.19, "learning_rate": 2.6440610104861775e-06, "loss": 2.9652, "step": 6273 }, { "epoch": 1.2, "learning_rate": 2.6440038131553863e-06, "loss": 3.4834, "step": 6274 }, { "epoch": 1.2, "learning_rate": 2.643946615824595e-06, "loss": 3.1397, "step": 6275 }, { "epoch": 1.2, "learning_rate": 2.643889418493804e-06, "loss": 3.4737, "step": 6276 }, { "epoch": 1.2, "learning_rate": 2.6438322211630122e-06, "loss": 3.581, "step": 6277 }, { "epoch": 1.2, "learning_rate": 2.6437750238322214e-06, "loss": 3.3131, "step": 6278 }, { "epoch": 1.2, "learning_rate": 2.64371782650143e-06, "loss": 3.0783, "step": 6279 }, { "epoch": 1.2, "learning_rate": 2.6436606291706386e-06, "loss": 3.448, "step": 6280 }, { "epoch": 1.2, "learning_rate": 2.643603431839848e-06, "loss": 3.3057, "step": 6281 }, { "epoch": 1.2, "learning_rate": 2.643546234509056e-06, "loss": 3.2946, "step": 6282 }, { "epoch": 1.2, "learning_rate": 2.643489037178265e-06, "loss": 3.3822, "step": 6283 }, { "epoch": 1.2, "learning_rate": 2.6434318398474738e-06, "loss": 3.4866, "step": 6284 }, { "epoch": 1.2, "learning_rate": 2.6433746425166826e-06, "loss": 3.6046, "step": 6285 }, { "epoch": 1.2, "learning_rate": 2.6433174451858914e-06, "loss": 3.0212, "step": 6286 }, { "epoch": 1.2, "learning_rate": 2.6432602478551e-06, "loss": 2.5044, "step": 6287 }, { "epoch": 1.2, "learning_rate": 2.643203050524309e-06, "loss": 3.1887, "step": 6288 }, { "epoch": 1.2, "learning_rate": 2.6431458531935177e-06, "loss": 3.204, "step": 6289 }, { "epoch": 1.2, "learning_rate": 2.6430886558627265e-06, "loss": 3.3963, "step": 6290 }, { "epoch": 1.2, "learning_rate": 2.6430314585319353e-06, "loss": 2.9471, "step": 6291 }, { "epoch": 1.2, "learning_rate": 2.642974261201144e-06, "loss": 3.8919, "step": 6292 }, { "epoch": 1.2, "learning_rate": 2.642917063870353e-06, "loss": 3.2097, "step": 6293 }, { "epoch": 1.2, "learning_rate": 2.6428598665395613e-06, "loss": 3.3786, "step": 6294 }, { "epoch": 1.2, "learning_rate": 2.6428026692087705e-06, "loss": 2.6278, "step": 6295 }, { "epoch": 1.2, "learning_rate": 2.6427454718779793e-06, "loss": 3.1135, "step": 6296 }, { "epoch": 1.2, "learning_rate": 2.6426882745471877e-06, "loss": 2.9705, "step": 6297 }, { "epoch": 1.2, "learning_rate": 2.642631077216397e-06, "loss": 3.0393, "step": 6298 }, { "epoch": 1.2, "learning_rate": 2.6425738798856053e-06, "loss": 3.615, "step": 6299 }, { "epoch": 1.2, "learning_rate": 2.642516682554814e-06, "loss": 3.6359, "step": 6300 }, { "epoch": 1.2, "learning_rate": 2.642459485224023e-06, "loss": 2.9447, "step": 6301 }, { "epoch": 1.2, "learning_rate": 2.6424022878932316e-06, "loss": 2.7371, "step": 6302 }, { "epoch": 1.2, "learning_rate": 2.6423450905624404e-06, "loss": 3.5112, "step": 6303 }, { "epoch": 1.2, "learning_rate": 2.6422878932316492e-06, "loss": 2.9618, "step": 6304 }, { "epoch": 1.2, "learning_rate": 2.642230695900858e-06, "loss": 3.0722, "step": 6305 }, { "epoch": 1.2, "learning_rate": 2.642173498570067e-06, "loss": 3.4854, "step": 6306 }, { "epoch": 1.2, "learning_rate": 2.6421163012392756e-06, "loss": 2.9585, "step": 6307 }, { "epoch": 1.2, "learning_rate": 2.6420591039084844e-06, "loss": 3.6579, "step": 6308 }, { "epoch": 1.2, "learning_rate": 2.642001906577693e-06, "loss": 3.1877, "step": 6309 }, { "epoch": 1.2, "learning_rate": 2.641944709246902e-06, "loss": 3.3575, "step": 6310 }, { "epoch": 1.2, "learning_rate": 2.6418875119161104e-06, "loss": 2.522, "step": 6311 }, { "epoch": 1.2, "learning_rate": 2.6418303145853196e-06, "loss": 3.1276, "step": 6312 }, { "epoch": 1.2, "learning_rate": 2.6417731172545284e-06, "loss": 3.275, "step": 6313 }, { "epoch": 1.2, "learning_rate": 2.6417159199237367e-06, "loss": 3.7215, "step": 6314 }, { "epoch": 1.2, "learning_rate": 2.641658722592946e-06, "loss": 2.7131, "step": 6315 }, { "epoch": 1.2, "learning_rate": 2.6416015252621543e-06, "loss": 3.0773, "step": 6316 }, { "epoch": 1.2, "learning_rate": 2.641544327931363e-06, "loss": 3.4253, "step": 6317 }, { "epoch": 1.2, "learning_rate": 2.641487130600572e-06, "loss": 2.7894, "step": 6318 }, { "epoch": 1.2, "learning_rate": 2.6414299332697807e-06, "loss": 3.4646, "step": 6319 }, { "epoch": 1.2, "learning_rate": 2.6413727359389895e-06, "loss": 3.3352, "step": 6320 }, { "epoch": 1.2, "learning_rate": 2.6413155386081983e-06, "loss": 3.0966, "step": 6321 }, { "epoch": 1.2, "learning_rate": 2.641258341277407e-06, "loss": 3.5127, "step": 6322 }, { "epoch": 1.2, "learning_rate": 2.641201143946616e-06, "loss": 2.8171, "step": 6323 }, { "epoch": 1.2, "learning_rate": 2.6411439466158247e-06, "loss": 3.2036, "step": 6324 }, { "epoch": 1.2, "learning_rate": 2.6410867492850335e-06, "loss": 2.9701, "step": 6325 }, { "epoch": 1.2, "learning_rate": 2.6410295519542423e-06, "loss": 2.9443, "step": 6326 }, { "epoch": 1.21, "learning_rate": 2.640972354623451e-06, "loss": 3.4796, "step": 6327 }, { "epoch": 1.21, "learning_rate": 2.6409151572926594e-06, "loss": 3.023, "step": 6328 }, { "epoch": 1.21, "learning_rate": 2.6408579599618687e-06, "loss": 3.067, "step": 6329 }, { "epoch": 1.21, "learning_rate": 2.6408007626310774e-06, "loss": 3.1606, "step": 6330 }, { "epoch": 1.21, "learning_rate": 2.640743565300286e-06, "loss": 3.5047, "step": 6331 }, { "epoch": 1.21, "learning_rate": 2.640686367969495e-06, "loss": 2.7489, "step": 6332 }, { "epoch": 1.21, "learning_rate": 2.6406291706387034e-06, "loss": 3.5656, "step": 6333 }, { "epoch": 1.21, "learning_rate": 2.640571973307912e-06, "loss": 3.7316, "step": 6334 }, { "epoch": 1.21, "learning_rate": 2.6405147759771214e-06, "loss": 3.3607, "step": 6335 }, { "epoch": 1.21, "learning_rate": 2.6404575786463298e-06, "loss": 3.0261, "step": 6336 }, { "epoch": 1.21, "learning_rate": 2.6404003813155386e-06, "loss": 2.3028, "step": 6337 }, { "epoch": 1.21, "learning_rate": 2.6403431839847474e-06, "loss": 3.2036, "step": 6338 }, { "epoch": 1.21, "learning_rate": 2.640285986653956e-06, "loss": 3.05, "step": 6339 }, { "epoch": 1.21, "learning_rate": 2.640228789323165e-06, "loss": 3.3864, "step": 6340 }, { "epoch": 1.21, "learning_rate": 2.6401715919923738e-06, "loss": 3.7105, "step": 6341 }, { "epoch": 1.21, "learning_rate": 2.6401143946615825e-06, "loss": 2.734, "step": 6342 }, { "epoch": 1.21, "learning_rate": 2.6400571973307913e-06, "loss": 3.4215, "step": 6343 }, { "epoch": 1.21, "learning_rate": 2.64e-06, "loss": 3.4471, "step": 6344 }, { "epoch": 1.21, "learning_rate": 2.639942802669209e-06, "loss": 3.0967, "step": 6345 }, { "epoch": 1.21, "learning_rate": 2.6398856053384177e-06, "loss": 3.717, "step": 6346 }, { "epoch": 1.21, "learning_rate": 2.6398284080076265e-06, "loss": 3.3626, "step": 6347 }, { "epoch": 1.21, "learning_rate": 2.639771210676835e-06, "loss": 3.6286, "step": 6348 }, { "epoch": 1.21, "learning_rate": 2.639714013346044e-06, "loss": 3.1885, "step": 6349 }, { "epoch": 1.21, "learning_rate": 2.6396568160152525e-06, "loss": 2.8721, "step": 6350 }, { "epoch": 1.21, "learning_rate": 2.6395996186844613e-06, "loss": 2.9485, "step": 6351 }, { "epoch": 1.21, "learning_rate": 2.6395424213536705e-06, "loss": 3.2569, "step": 6352 }, { "epoch": 1.21, "learning_rate": 2.639485224022879e-06, "loss": 2.5602, "step": 6353 }, { "epoch": 1.21, "learning_rate": 2.6394280266920876e-06, "loss": 3.5389, "step": 6354 }, { "epoch": 1.21, "learning_rate": 2.6393708293612964e-06, "loss": 3.5712, "step": 6355 }, { "epoch": 1.21, "learning_rate": 2.6393136320305052e-06, "loss": 2.8549, "step": 6356 }, { "epoch": 1.21, "learning_rate": 2.639256434699714e-06, "loss": 3.6229, "step": 6357 }, { "epoch": 1.21, "learning_rate": 2.639199237368923e-06, "loss": 3.0889, "step": 6358 }, { "epoch": 1.21, "learning_rate": 2.6391420400381316e-06, "loss": 2.8662, "step": 6359 }, { "epoch": 1.21, "learning_rate": 2.6390848427073404e-06, "loss": 2.983, "step": 6360 }, { "epoch": 1.21, "learning_rate": 2.639027645376549e-06, "loss": 2.6465, "step": 6361 }, { "epoch": 1.21, "learning_rate": 2.638970448045758e-06, "loss": 3.1332, "step": 6362 }, { "epoch": 1.21, "learning_rate": 2.638913250714967e-06, "loss": 3.1976, "step": 6363 }, { "epoch": 1.21, "learning_rate": 2.6388560533841756e-06, "loss": 3.0279, "step": 6364 }, { "epoch": 1.21, "learning_rate": 2.638798856053384e-06, "loss": 3.3235, "step": 6365 }, { "epoch": 1.21, "learning_rate": 2.638741658722593e-06, "loss": 3.2954, "step": 6366 }, { "epoch": 1.21, "learning_rate": 2.6386844613918015e-06, "loss": 3.8885, "step": 6367 }, { "epoch": 1.21, "learning_rate": 2.6386272640610103e-06, "loss": 3.0936, "step": 6368 }, { "epoch": 1.21, "learning_rate": 2.6385700667302196e-06, "loss": 2.7968, "step": 6369 }, { "epoch": 1.21, "learning_rate": 2.638512869399428e-06, "loss": 3.0098, "step": 6370 }, { "epoch": 1.21, "learning_rate": 2.638455672068637e-06, "loss": 3.4671, "step": 6371 }, { "epoch": 1.21, "learning_rate": 2.6383984747378455e-06, "loss": 3.8657, "step": 6372 }, { "epoch": 1.21, "learning_rate": 2.6383412774070543e-06, "loss": 3.3517, "step": 6373 }, { "epoch": 1.21, "learning_rate": 2.6382840800762635e-06, "loss": 3.5385, "step": 6374 }, { "epoch": 1.21, "learning_rate": 2.638226882745472e-06, "loss": 3.4079, "step": 6375 }, { "epoch": 1.21, "learning_rate": 2.6381696854146807e-06, "loss": 3.4989, "step": 6376 }, { "epoch": 1.21, "learning_rate": 2.6381124880838895e-06, "loss": 2.8852, "step": 6377 }, { "epoch": 1.21, "learning_rate": 2.6380552907530983e-06, "loss": 3.1013, "step": 6378 }, { "epoch": 1.22, "learning_rate": 2.637998093422307e-06, "loss": 3.3053, "step": 6379 }, { "epoch": 1.22, "learning_rate": 2.637940896091516e-06, "loss": 3.2756, "step": 6380 }, { "epoch": 1.22, "learning_rate": 2.6378836987607247e-06, "loss": 2.8774, "step": 6381 }, { "epoch": 1.22, "learning_rate": 2.637826501429933e-06, "loss": 3.3679, "step": 6382 }, { "epoch": 1.22, "learning_rate": 2.6377693040991422e-06, "loss": 3.2681, "step": 6383 }, { "epoch": 1.22, "learning_rate": 2.637712106768351e-06, "loss": 3.7313, "step": 6384 }, { "epoch": 1.22, "learning_rate": 2.6376549094375594e-06, "loss": 3.4879, "step": 6385 }, { "epoch": 1.22, "learning_rate": 2.6375977121067686e-06, "loss": 3.3408, "step": 6386 }, { "epoch": 1.22, "learning_rate": 2.637540514775977e-06, "loss": 3.6926, "step": 6387 }, { "epoch": 1.22, "learning_rate": 2.6374833174451862e-06, "loss": 3.6399, "step": 6388 }, { "epoch": 1.22, "learning_rate": 2.6374261201143946e-06, "loss": 3.3363, "step": 6389 }, { "epoch": 1.22, "learning_rate": 2.6373689227836034e-06, "loss": 3.2296, "step": 6390 }, { "epoch": 1.22, "learning_rate": 2.6373117254528126e-06, "loss": 3.0821, "step": 6391 }, { "epoch": 1.22, "learning_rate": 2.637254528122021e-06, "loss": 3.5548, "step": 6392 }, { "epoch": 1.22, "learning_rate": 2.6371973307912298e-06, "loss": 3.2741, "step": 6393 }, { "epoch": 1.22, "learning_rate": 2.6371401334604386e-06, "loss": 3.8336, "step": 6394 }, { "epoch": 1.22, "learning_rate": 2.6370829361296473e-06, "loss": 3.194, "step": 6395 }, { "epoch": 1.22, "learning_rate": 2.637025738798856e-06, "loss": 3.0289, "step": 6396 }, { "epoch": 1.22, "learning_rate": 2.636968541468065e-06, "loss": 3.4023, "step": 6397 }, { "epoch": 1.22, "learning_rate": 2.6369113441372737e-06, "loss": 3.1532, "step": 6398 }, { "epoch": 1.22, "learning_rate": 2.636854146806482e-06, "loss": 3.5897, "step": 6399 }, { "epoch": 1.22, "learning_rate": 2.6367969494756913e-06, "loss": 3.0935, "step": 6400 }, { "epoch": 1.22, "learning_rate": 2.6367397521449e-06, "loss": 3.3785, "step": 6401 }, { "epoch": 1.22, "learning_rate": 2.6366825548141085e-06, "loss": 2.9653, "step": 6402 }, { "epoch": 1.22, "learning_rate": 2.6366253574833177e-06, "loss": 3.3779, "step": 6403 }, { "epoch": 1.22, "learning_rate": 2.636568160152526e-06, "loss": 3.4182, "step": 6404 }, { "epoch": 1.22, "learning_rate": 2.6365109628217353e-06, "loss": 3.3388, "step": 6405 }, { "epoch": 1.22, "learning_rate": 2.6364537654909437e-06, "loss": 3.6709, "step": 6406 }, { "epoch": 1.22, "learning_rate": 2.6363965681601525e-06, "loss": 3.7247, "step": 6407 }, { "epoch": 1.22, "learning_rate": 2.6363393708293617e-06, "loss": 2.3174, "step": 6408 }, { "epoch": 1.22, "learning_rate": 2.63628217349857e-06, "loss": 3.4931, "step": 6409 }, { "epoch": 1.22, "learning_rate": 2.636224976167779e-06, "loss": 3.3748, "step": 6410 }, { "epoch": 1.22, "learning_rate": 2.6361677788369876e-06, "loss": 3.1519, "step": 6411 }, { "epoch": 1.22, "learning_rate": 2.6361105815061964e-06, "loss": 3.6557, "step": 6412 }, { "epoch": 1.22, "learning_rate": 2.6360533841754052e-06, "loss": 3.6552, "step": 6413 }, { "epoch": 1.22, "learning_rate": 2.635996186844614e-06, "loss": 3.073, "step": 6414 }, { "epoch": 1.22, "learning_rate": 2.635938989513823e-06, "loss": 3.2212, "step": 6415 }, { "epoch": 1.22, "learning_rate": 2.635881792183031e-06, "loss": 2.5642, "step": 6416 }, { "epoch": 1.22, "learning_rate": 2.6358245948522404e-06, "loss": 3.2304, "step": 6417 }, { "epoch": 1.22, "learning_rate": 2.635767397521449e-06, "loss": 2.6253, "step": 6418 }, { "epoch": 1.22, "learning_rate": 2.6357102001906576e-06, "loss": 3.0675, "step": 6419 }, { "epoch": 1.22, "learning_rate": 2.6356530028598668e-06, "loss": 3.0818, "step": 6420 }, { "epoch": 1.22, "learning_rate": 2.635595805529075e-06, "loss": 3.2567, "step": 6421 }, { "epoch": 1.22, "learning_rate": 2.6355386081982844e-06, "loss": 3.4933, "step": 6422 }, { "epoch": 1.22, "learning_rate": 2.635481410867493e-06, "loss": 2.9464, "step": 6423 }, { "epoch": 1.22, "learning_rate": 2.6354242135367015e-06, "loss": 2.7964, "step": 6424 }, { "epoch": 1.22, "learning_rate": 2.6353670162059107e-06, "loss": 3.5833, "step": 6425 }, { "epoch": 1.22, "learning_rate": 2.635309818875119e-06, "loss": 2.9703, "step": 6426 }, { "epoch": 1.22, "learning_rate": 2.635252621544328e-06, "loss": 3.6746, "step": 6427 }, { "epoch": 1.22, "learning_rate": 2.6351954242135367e-06, "loss": 2.5666, "step": 6428 }, { "epoch": 1.22, "learning_rate": 2.6351382268827455e-06, "loss": 2.9801, "step": 6429 }, { "epoch": 1.22, "learning_rate": 2.6350810295519543e-06, "loss": 3.0257, "step": 6430 }, { "epoch": 1.22, "learning_rate": 2.635023832221163e-06, "loss": 3.185, "step": 6431 }, { "epoch": 1.23, "learning_rate": 2.634966634890372e-06, "loss": 2.9991, "step": 6432 }, { "epoch": 1.23, "learning_rate": 2.6349094375595807e-06, "loss": 2.8881, "step": 6433 }, { "epoch": 1.23, "learning_rate": 2.6348522402287895e-06, "loss": 3.7185, "step": 6434 }, { "epoch": 1.23, "learning_rate": 2.6347950428979983e-06, "loss": 3.0811, "step": 6435 }, { "epoch": 1.23, "learning_rate": 2.634737845567207e-06, "loss": 3.5315, "step": 6436 }, { "epoch": 1.23, "learning_rate": 2.634680648236416e-06, "loss": 3.5131, "step": 6437 }, { "epoch": 1.23, "learning_rate": 2.6346234509056242e-06, "loss": 3.4204, "step": 6438 }, { "epoch": 1.23, "learning_rate": 2.6345662535748334e-06, "loss": 3.076, "step": 6439 }, { "epoch": 1.23, "learning_rate": 2.6345090562440422e-06, "loss": 3.3853, "step": 6440 }, { "epoch": 1.23, "learning_rate": 2.6344518589132506e-06, "loss": 3.5682, "step": 6441 }, { "epoch": 1.23, "learning_rate": 2.63439466158246e-06, "loss": 3.3677, "step": 6442 }, { "epoch": 1.23, "learning_rate": 2.634337464251668e-06, "loss": 3.5422, "step": 6443 }, { "epoch": 1.23, "learning_rate": 2.634280266920877e-06, "loss": 3.4269, "step": 6444 }, { "epoch": 1.23, "learning_rate": 2.6342230695900858e-06, "loss": 3.1008, "step": 6445 }, { "epoch": 1.23, "learning_rate": 2.6341658722592946e-06, "loss": 3.9554, "step": 6446 }, { "epoch": 1.23, "learning_rate": 2.6341086749285034e-06, "loss": 3.1477, "step": 6447 }, { "epoch": 1.23, "learning_rate": 2.634051477597712e-06, "loss": 2.8928, "step": 6448 }, { "epoch": 1.23, "learning_rate": 2.633994280266921e-06, "loss": 3.5809, "step": 6449 }, { "epoch": 1.23, "learning_rate": 2.6339370829361297e-06, "loss": 2.9267, "step": 6450 }, { "epoch": 1.23, "learning_rate": 2.6338798856053385e-06, "loss": 3.0559, "step": 6451 }, { "epoch": 1.23, "learning_rate": 2.6338226882745473e-06, "loss": 3.3967, "step": 6452 }, { "epoch": 1.23, "learning_rate": 2.633765490943756e-06, "loss": 2.872, "step": 6453 }, { "epoch": 1.23, "learning_rate": 2.633708293612965e-06, "loss": 3.8312, "step": 6454 }, { "epoch": 1.23, "learning_rate": 2.6336510962821733e-06, "loss": 2.8387, "step": 6455 }, { "epoch": 1.23, "learning_rate": 2.6335938989513825e-06, "loss": 3.6552, "step": 6456 }, { "epoch": 1.23, "learning_rate": 2.6335367016205913e-06, "loss": 3.0207, "step": 6457 }, { "epoch": 1.23, "learning_rate": 2.6334795042897997e-06, "loss": 3.4794, "step": 6458 }, { "epoch": 1.23, "learning_rate": 2.633422306959009e-06, "loss": 3.3455, "step": 6459 }, { "epoch": 1.23, "learning_rate": 2.6333651096282173e-06, "loss": 3.1249, "step": 6460 }, { "epoch": 1.23, "learning_rate": 2.633307912297426e-06, "loss": 3.6893, "step": 6461 }, { "epoch": 1.23, "learning_rate": 2.6332507149666353e-06, "loss": 3.486, "step": 6462 }, { "epoch": 1.23, "learning_rate": 2.6331935176358436e-06, "loss": 3.171, "step": 6463 }, { "epoch": 1.23, "learning_rate": 2.6331363203050524e-06, "loss": 3.2592, "step": 6464 }, { "epoch": 1.23, "learning_rate": 2.6330791229742612e-06, "loss": 3.3862, "step": 6465 }, { "epoch": 1.23, "learning_rate": 2.63302192564347e-06, "loss": 3.2913, "step": 6466 }, { "epoch": 1.23, "learning_rate": 2.632964728312679e-06, "loss": 3.1208, "step": 6467 }, { "epoch": 1.23, "learning_rate": 2.6329075309818876e-06, "loss": 3.0915, "step": 6468 }, { "epoch": 1.23, "learning_rate": 2.6328503336510964e-06, "loss": 3.6914, "step": 6469 }, { "epoch": 1.23, "learning_rate": 2.632793136320305e-06, "loss": 3.1014, "step": 6470 }, { "epoch": 1.23, "learning_rate": 2.632735938989514e-06, "loss": 3.2144, "step": 6471 }, { "epoch": 1.23, "learning_rate": 2.6326787416587228e-06, "loss": 3.1289, "step": 6472 }, { "epoch": 1.23, "learning_rate": 2.6326215443279316e-06, "loss": 3.3777, "step": 6473 }, { "epoch": 1.23, "learning_rate": 2.6325643469971404e-06, "loss": 2.8842, "step": 6474 }, { "epoch": 1.23, "learning_rate": 2.6325071496663487e-06, "loss": 2.8499, "step": 6475 }, { "epoch": 1.23, "learning_rate": 2.632449952335558e-06, "loss": 3.0766, "step": 6476 }, { "epoch": 1.23, "learning_rate": 2.6323927550047663e-06, "loss": 3.4568, "step": 6477 }, { "epoch": 1.23, "learning_rate": 2.632335557673975e-06, "loss": 3.2636, "step": 6478 }, { "epoch": 1.23, "learning_rate": 2.6322783603431843e-06, "loss": 3.5714, "step": 6479 }, { "epoch": 1.23, "learning_rate": 2.6322211630123927e-06, "loss": 3.0996, "step": 6480 }, { "epoch": 1.23, "learning_rate": 2.6321639656816015e-06, "loss": 3.5913, "step": 6481 }, { "epoch": 1.23, "learning_rate": 2.6321067683508103e-06, "loss": 3.0133, "step": 6482 }, { "epoch": 1.23, "learning_rate": 2.632049571020019e-06, "loss": 3.1855, "step": 6483 }, { "epoch": 1.24, "learning_rate": 2.631992373689228e-06, "loss": 3.1537, "step": 6484 }, { "epoch": 1.24, "learning_rate": 2.6319351763584367e-06, "loss": 3.1098, "step": 6485 }, { "epoch": 1.24, "learning_rate": 2.6318779790276455e-06, "loss": 3.2025, "step": 6486 }, { "epoch": 1.24, "learning_rate": 2.6318207816968543e-06, "loss": 3.3119, "step": 6487 }, { "epoch": 1.24, "learning_rate": 2.631763584366063e-06, "loss": 3.2039, "step": 6488 }, { "epoch": 1.24, "learning_rate": 2.631706387035272e-06, "loss": 3.0244, "step": 6489 }, { "epoch": 1.24, "learning_rate": 2.6316491897044806e-06, "loss": 2.7738, "step": 6490 }, { "epoch": 1.24, "learning_rate": 2.6315919923736894e-06, "loss": 3.0822, "step": 6491 }, { "epoch": 1.24, "learning_rate": 2.631534795042898e-06, "loss": 3.8098, "step": 6492 }, { "epoch": 1.24, "learning_rate": 2.631477597712107e-06, "loss": 2.9893, "step": 6493 }, { "epoch": 1.24, "learning_rate": 2.6314204003813154e-06, "loss": 3.6697, "step": 6494 }, { "epoch": 1.24, "learning_rate": 2.631363203050524e-06, "loss": 3.7281, "step": 6495 }, { "epoch": 1.24, "learning_rate": 2.6313060057197334e-06, "loss": 3.5216, "step": 6496 }, { "epoch": 1.24, "learning_rate": 2.6312488083889418e-06, "loss": 3.5827, "step": 6497 }, { "epoch": 1.24, "learning_rate": 2.6311916110581506e-06, "loss": 2.8757, "step": 6498 }, { "epoch": 1.24, "learning_rate": 2.6311344137273594e-06, "loss": 3.1664, "step": 6499 }, { "epoch": 1.24, "learning_rate": 2.631077216396568e-06, "loss": 3.2073, "step": 6500 }, { "epoch": 1.24, "learning_rate": 2.631020019065777e-06, "loss": 2.6538, "step": 6501 }, { "epoch": 1.24, "learning_rate": 2.6309628217349857e-06, "loss": 3.2145, "step": 6502 }, { "epoch": 1.24, "learning_rate": 2.6309056244041945e-06, "loss": 3.4608, "step": 6503 }, { "epoch": 1.24, "learning_rate": 2.6308484270734033e-06, "loss": 3.579, "step": 6504 }, { "epoch": 1.24, "learning_rate": 2.630791229742612e-06, "loss": 3.4169, "step": 6505 }, { "epoch": 1.24, "learning_rate": 2.630734032411821e-06, "loss": 2.6244, "step": 6506 }, { "epoch": 1.24, "learning_rate": 2.6306768350810297e-06, "loss": 3.7028, "step": 6507 }, { "epoch": 1.24, "learning_rate": 2.6306196377502385e-06, "loss": 3.4088, "step": 6508 }, { "epoch": 1.24, "learning_rate": 2.630562440419447e-06, "loss": 3.2098, "step": 6509 }, { "epoch": 1.24, "learning_rate": 2.630505243088656e-06, "loss": 3.2997, "step": 6510 }, { "epoch": 1.24, "learning_rate": 2.630448045757865e-06, "loss": 3.7635, "step": 6511 }, { "epoch": 1.24, "learning_rate": 2.6303908484270733e-06, "loss": 3.2743, "step": 6512 }, { "epoch": 1.24, "learning_rate": 2.6303336510962825e-06, "loss": 2.7562, "step": 6513 }, { "epoch": 1.24, "learning_rate": 2.630276453765491e-06, "loss": 3.9405, "step": 6514 }, { "epoch": 1.24, "learning_rate": 2.6302192564346996e-06, "loss": 3.2077, "step": 6515 }, { "epoch": 1.24, "learning_rate": 2.6301620591039084e-06, "loss": 3.1393, "step": 6516 }, { "epoch": 1.24, "learning_rate": 2.6301048617731172e-06, "loss": 2.7232, "step": 6517 }, { "epoch": 1.24, "learning_rate": 2.6300476644423264e-06, "loss": 3.4841, "step": 6518 }, { "epoch": 1.24, "learning_rate": 2.629990467111535e-06, "loss": 2.6649, "step": 6519 }, { "epoch": 1.24, "learning_rate": 2.6299332697807436e-06, "loss": 3.1904, "step": 6520 }, { "epoch": 1.24, "learning_rate": 2.6298760724499524e-06, "loss": 3.496, "step": 6521 }, { "epoch": 1.24, "learning_rate": 2.629818875119161e-06, "loss": 2.8312, "step": 6522 }, { "epoch": 1.24, "learning_rate": 2.62976167778837e-06, "loss": 3.4989, "step": 6523 }, { "epoch": 1.24, "learning_rate": 2.6297044804575788e-06, "loss": 3.4281, "step": 6524 }, { "epoch": 1.24, "learning_rate": 2.6296472831267876e-06, "loss": 3.0662, "step": 6525 }, { "epoch": 1.24, "learning_rate": 2.629590085795996e-06, "loss": 2.9092, "step": 6526 }, { "epoch": 1.24, "learning_rate": 2.629532888465205e-06, "loss": 3.2826, "step": 6527 }, { "epoch": 1.24, "learning_rate": 2.629475691134414e-06, "loss": 3.4677, "step": 6528 }, { "epoch": 1.24, "learning_rate": 2.6294184938036223e-06, "loss": 3.4913, "step": 6529 }, { "epoch": 1.24, "learning_rate": 2.6293612964728315e-06, "loss": 3.6222, "step": 6530 }, { "epoch": 1.24, "learning_rate": 2.62930409914204e-06, "loss": 2.6695, "step": 6531 }, { "epoch": 1.24, "learning_rate": 2.6292469018112487e-06, "loss": 3.3017, "step": 6532 }, { "epoch": 1.24, "learning_rate": 2.6291897044804575e-06, "loss": 2.8306, "step": 6533 }, { "epoch": 1.24, "learning_rate": 2.6291325071496663e-06, "loss": 2.9426, "step": 6534 }, { "epoch": 1.24, "learning_rate": 2.6290753098188755e-06, "loss": 3.3713, "step": 6535 }, { "epoch": 1.24, "learning_rate": 2.629018112488084e-06, "loss": 3.5675, "step": 6536 }, { "epoch": 1.25, "learning_rate": 2.6289609151572927e-06, "loss": 3.3403, "step": 6537 }, { "epoch": 1.25, "learning_rate": 2.6289037178265015e-06, "loss": 3.2125, "step": 6538 }, { "epoch": 1.25, "learning_rate": 2.6288465204957103e-06, "loss": 3.3804, "step": 6539 }, { "epoch": 1.25, "learning_rate": 2.628789323164919e-06, "loss": 2.3533, "step": 6540 }, { "epoch": 1.25, "learning_rate": 2.628732125834128e-06, "loss": 3.5254, "step": 6541 }, { "epoch": 1.25, "learning_rate": 2.6286749285033367e-06, "loss": 3.4514, "step": 6542 }, { "epoch": 1.25, "learning_rate": 2.628617731172545e-06, "loss": 3.0718, "step": 6543 }, { "epoch": 1.25, "learning_rate": 2.6285605338417542e-06, "loss": 3.742, "step": 6544 }, { "epoch": 1.25, "learning_rate": 2.628503336510963e-06, "loss": 2.6708, "step": 6545 }, { "epoch": 1.25, "learning_rate": 2.6284461391801714e-06, "loss": 2.9965, "step": 6546 }, { "epoch": 1.25, "learning_rate": 2.6283889418493806e-06, "loss": 3.462, "step": 6547 }, { "epoch": 1.25, "learning_rate": 2.628331744518589e-06, "loss": 2.4927, "step": 6548 }, { "epoch": 1.25, "learning_rate": 2.6282745471877978e-06, "loss": 2.4298, "step": 6549 }, { "epoch": 1.25, "learning_rate": 2.6282173498570066e-06, "loss": 2.6856, "step": 6550 }, { "epoch": 1.25, "learning_rate": 2.6281601525262154e-06, "loss": 3.0822, "step": 6551 }, { "epoch": 1.25, "learning_rate": 2.6281029551954246e-06, "loss": 3.4741, "step": 6552 }, { "epoch": 1.25, "learning_rate": 2.628045757864633e-06, "loss": 2.9767, "step": 6553 }, { "epoch": 1.25, "learning_rate": 2.6279885605338418e-06, "loss": 2.9682, "step": 6554 }, { "epoch": 1.25, "learning_rate": 2.6279313632030505e-06, "loss": 3.0448, "step": 6555 }, { "epoch": 1.25, "learning_rate": 2.6278741658722593e-06, "loss": 3.004, "step": 6556 }, { "epoch": 1.25, "learning_rate": 2.627816968541468e-06, "loss": 3.0898, "step": 6557 }, { "epoch": 1.25, "learning_rate": 2.627759771210677e-06, "loss": 3.1491, "step": 6558 }, { "epoch": 1.25, "learning_rate": 2.6277025738798857e-06, "loss": 3.0483, "step": 6559 }, { "epoch": 1.25, "learning_rate": 2.6276453765490945e-06, "loss": 3.4785, "step": 6560 }, { "epoch": 1.25, "learning_rate": 2.6275881792183033e-06, "loss": 3.0579, "step": 6561 }, { "epoch": 1.25, "learning_rate": 2.627530981887512e-06, "loss": 3.6151, "step": 6562 }, { "epoch": 1.25, "learning_rate": 2.6274737845567205e-06, "loss": 3.0074, "step": 6563 }, { "epoch": 1.25, "learning_rate": 2.6274165872259297e-06, "loss": 3.6188, "step": 6564 }, { "epoch": 1.25, "learning_rate": 2.627359389895138e-06, "loss": 3.2781, "step": 6565 }, { "epoch": 1.25, "learning_rate": 2.6273021925643473e-06, "loss": 3.6645, "step": 6566 }, { "epoch": 1.25, "learning_rate": 2.627244995233556e-06, "loss": 3.3473, "step": 6567 }, { "epoch": 1.25, "learning_rate": 2.6271877979027644e-06, "loss": 3.1593, "step": 6568 }, { "epoch": 1.25, "learning_rate": 2.6271306005719737e-06, "loss": 3.0085, "step": 6569 }, { "epoch": 1.25, "learning_rate": 2.627073403241182e-06, "loss": 3.4109, "step": 6570 }, { "epoch": 1.25, "learning_rate": 2.627016205910391e-06, "loss": 3.6637, "step": 6571 }, { "epoch": 1.25, "learning_rate": 2.6269590085795996e-06, "loss": 2.9846, "step": 6572 }, { "epoch": 1.25, "learning_rate": 2.6269018112488084e-06, "loss": 2.7225, "step": 6573 }, { "epoch": 1.25, "learning_rate": 2.626844613918017e-06, "loss": 3.0471, "step": 6574 }, { "epoch": 1.25, "learning_rate": 2.626787416587226e-06, "loss": 2.4147, "step": 6575 }, { "epoch": 1.25, "learning_rate": 2.626730219256435e-06, "loss": 3.5028, "step": 6576 }, { "epoch": 1.25, "learning_rate": 2.6266730219256436e-06, "loss": 3.3152, "step": 6577 }, { "epoch": 1.25, "learning_rate": 2.6266158245948524e-06, "loss": 2.7757, "step": 6578 }, { "epoch": 1.25, "learning_rate": 2.626558627264061e-06, "loss": 3.1896, "step": 6579 }, { "epoch": 1.25, "learning_rate": 2.6265014299332695e-06, "loss": 3.399, "step": 6580 }, { "epoch": 1.25, "learning_rate": 2.6264442326024788e-06, "loss": 3.575, "step": 6581 }, { "epoch": 1.25, "learning_rate": 2.626387035271687e-06, "loss": 3.5213, "step": 6582 }, { "epoch": 1.25, "learning_rate": 2.6263298379408964e-06, "loss": 3.1198, "step": 6583 }, { "epoch": 1.25, "learning_rate": 2.626272640610105e-06, "loss": 3.1913, "step": 6584 }, { "epoch": 1.25, "learning_rate": 2.6262154432793135e-06, "loss": 3.3841, "step": 6585 }, { "epoch": 1.25, "learning_rate": 2.6261582459485227e-06, "loss": 3.6325, "step": 6586 }, { "epoch": 1.25, "learning_rate": 2.626101048617731e-06, "loss": 3.1714, "step": 6587 }, { "epoch": 1.25, "learning_rate": 2.62604385128694e-06, "loss": 3.243, "step": 6588 }, { "epoch": 1.26, "learning_rate": 2.6259866539561487e-06, "loss": 2.9559, "step": 6589 }, { "epoch": 1.26, "learning_rate": 2.6259294566253575e-06, "loss": 3.5074, "step": 6590 }, { "epoch": 1.26, "learning_rate": 2.6258722592945663e-06, "loss": 3.7025, "step": 6591 }, { "epoch": 1.26, "learning_rate": 2.625815061963775e-06, "loss": 3.2186, "step": 6592 }, { "epoch": 1.26, "learning_rate": 2.625757864632984e-06, "loss": 3.3947, "step": 6593 }, { "epoch": 1.26, "learning_rate": 2.6257006673021927e-06, "loss": 3.0098, "step": 6594 }, { "epoch": 1.26, "learning_rate": 2.6256434699714015e-06, "loss": 3.8181, "step": 6595 }, { "epoch": 1.26, "learning_rate": 2.6255862726406102e-06, "loss": 3.7219, "step": 6596 }, { "epoch": 1.26, "learning_rate": 2.6255290753098186e-06, "loss": 2.7623, "step": 6597 }, { "epoch": 1.26, "learning_rate": 2.625471877979028e-06, "loss": 3.1965, "step": 6598 }, { "epoch": 1.26, "learning_rate": 2.6254146806482366e-06, "loss": 2.678, "step": 6599 }, { "epoch": 1.26, "learning_rate": 2.6253574833174454e-06, "loss": 3.1745, "step": 6600 }, { "epoch": 1.26, "learning_rate": 2.6253002859866542e-06, "loss": 3.11, "step": 6601 }, { "epoch": 1.26, "learning_rate": 2.6252430886558626e-06, "loss": 2.8851, "step": 6602 }, { "epoch": 1.26, "learning_rate": 2.625185891325072e-06, "loss": 3.5837, "step": 6603 }, { "epoch": 1.26, "learning_rate": 2.62512869399428e-06, "loss": 2.9939, "step": 6604 }, { "epoch": 1.26, "learning_rate": 2.625071496663489e-06, "loss": 2.8475, "step": 6605 }, { "epoch": 1.26, "learning_rate": 2.625014299332698e-06, "loss": 3.4677, "step": 6606 }, { "epoch": 1.26, "learning_rate": 2.6249571020019066e-06, "loss": 3.0257, "step": 6607 }, { "epoch": 1.26, "learning_rate": 2.6248999046711153e-06, "loss": 3.3538, "step": 6608 }, { "epoch": 1.26, "learning_rate": 2.624842707340324e-06, "loss": 2.8567, "step": 6609 }, { "epoch": 1.26, "learning_rate": 2.624785510009533e-06, "loss": 3.8608, "step": 6610 }, { "epoch": 1.26, "learning_rate": 2.6247283126787417e-06, "loss": 2.7604, "step": 6611 }, { "epoch": 1.26, "learning_rate": 2.6246711153479505e-06, "loss": 2.9234, "step": 6612 }, { "epoch": 1.26, "learning_rate": 2.6246139180171593e-06, "loss": 2.8908, "step": 6613 }, { "epoch": 1.26, "learning_rate": 2.6245567206863677e-06, "loss": 3.6792, "step": 6614 }, { "epoch": 1.26, "learning_rate": 2.624499523355577e-06, "loss": 3.3326, "step": 6615 }, { "epoch": 1.26, "learning_rate": 2.6244423260247857e-06, "loss": 2.8095, "step": 6616 }, { "epoch": 1.26, "learning_rate": 2.6243851286939945e-06, "loss": 3.0393, "step": 6617 }, { "epoch": 1.26, "learning_rate": 2.6243279313632033e-06, "loss": 3.2675, "step": 6618 }, { "epoch": 1.26, "learning_rate": 2.6242707340324117e-06, "loss": 2.8645, "step": 6619 }, { "epoch": 1.26, "learning_rate": 2.624213536701621e-06, "loss": 3.6058, "step": 6620 }, { "epoch": 1.26, "learning_rate": 2.6241563393708292e-06, "loss": 3.1194, "step": 6621 }, { "epoch": 1.26, "learning_rate": 2.624099142040038e-06, "loss": 2.8964, "step": 6622 }, { "epoch": 1.26, "learning_rate": 2.6240419447092473e-06, "loss": 4.036, "step": 6623 }, { "epoch": 1.26, "learning_rate": 2.6239847473784556e-06, "loss": 3.7044, "step": 6624 }, { "epoch": 1.26, "learning_rate": 2.6239275500476644e-06, "loss": 3.1486, "step": 6625 }, { "epoch": 1.26, "learning_rate": 2.6238703527168732e-06, "loss": 3.4756, "step": 6626 }, { "epoch": 1.26, "learning_rate": 2.623813155386082e-06, "loss": 2.6895, "step": 6627 }, { "epoch": 1.26, "learning_rate": 2.623755958055291e-06, "loss": 2.9727, "step": 6628 }, { "epoch": 1.26, "learning_rate": 2.6236987607244996e-06, "loss": 2.2349, "step": 6629 }, { "epoch": 1.26, "learning_rate": 2.6236415633937084e-06, "loss": 3.3007, "step": 6630 }, { "epoch": 1.26, "learning_rate": 2.623584366062917e-06, "loss": 3.4829, "step": 6631 }, { "epoch": 1.26, "learning_rate": 2.623527168732126e-06, "loss": 2.8367, "step": 6632 }, { "epoch": 1.26, "learning_rate": 2.6234699714013348e-06, "loss": 3.6427, "step": 6633 }, { "epoch": 1.26, "learning_rate": 2.6234127740705436e-06, "loss": 3.0102, "step": 6634 }, { "epoch": 1.26, "learning_rate": 2.6233555767397524e-06, "loss": 3.5673, "step": 6635 }, { "epoch": 1.26, "learning_rate": 2.6232983794089607e-06, "loss": 3.4749, "step": 6636 }, { "epoch": 1.26, "learning_rate": 2.62324118207817e-06, "loss": 3.6426, "step": 6637 }, { "epoch": 1.26, "learning_rate": 2.6231839847473783e-06, "loss": 3.7236, "step": 6638 }, { "epoch": 1.26, "learning_rate": 2.623126787416587e-06, "loss": 2.9555, "step": 6639 }, { "epoch": 1.26, "learning_rate": 2.6230695900857963e-06, "loss": 2.7197, "step": 6640 }, { "epoch": 1.26, "learning_rate": 2.6230123927550047e-06, "loss": 3.4459, "step": 6641 }, { "epoch": 1.27, "learning_rate": 2.6229551954242135e-06, "loss": 3.2737, "step": 6642 }, { "epoch": 1.27, "learning_rate": 2.6228979980934223e-06, "loss": 2.8609, "step": 6643 }, { "epoch": 1.27, "learning_rate": 2.622840800762631e-06, "loss": 2.8441, "step": 6644 }, { "epoch": 1.27, "learning_rate": 2.62278360343184e-06, "loss": 3.6182, "step": 6645 }, { "epoch": 1.27, "learning_rate": 2.6227264061010487e-06, "loss": 2.6332, "step": 6646 }, { "epoch": 1.27, "learning_rate": 2.6226692087702575e-06, "loss": 3.2877, "step": 6647 }, { "epoch": 1.27, "learning_rate": 2.6226120114394663e-06, "loss": 3.418, "step": 6648 }, { "epoch": 1.27, "learning_rate": 2.622554814108675e-06, "loss": 3.4631, "step": 6649 }, { "epoch": 1.27, "learning_rate": 2.622497616777884e-06, "loss": 2.86, "step": 6650 }, { "epoch": 1.27, "learning_rate": 2.6224404194470926e-06, "loss": 3.1192, "step": 6651 }, { "epoch": 1.27, "learning_rate": 2.6223832221163014e-06, "loss": 3.1801, "step": 6652 }, { "epoch": 1.27, "learning_rate": 2.62232602478551e-06, "loss": 3.1587, "step": 6653 }, { "epoch": 1.27, "learning_rate": 2.622268827454719e-06, "loss": 2.9249, "step": 6654 }, { "epoch": 1.27, "learning_rate": 2.622211630123928e-06, "loss": 3.3784, "step": 6655 }, { "epoch": 1.27, "learning_rate": 2.622154432793136e-06, "loss": 3.3987, "step": 6656 }, { "epoch": 1.27, "learning_rate": 2.6220972354623454e-06, "loss": 3.216, "step": 6657 }, { "epoch": 1.27, "learning_rate": 2.6220400381315538e-06, "loss": 2.5515, "step": 6658 }, { "epoch": 1.27, "learning_rate": 2.6219828408007626e-06, "loss": 3.3224, "step": 6659 }, { "epoch": 1.27, "learning_rate": 2.6219256434699714e-06, "loss": 3.0713, "step": 6660 }, { "epoch": 1.27, "learning_rate": 2.62186844613918e-06, "loss": 3.2441, "step": 6661 }, { "epoch": 1.27, "learning_rate": 2.621811248808389e-06, "loss": 2.7417, "step": 6662 }, { "epoch": 1.27, "learning_rate": 2.6217540514775977e-06, "loss": 3.154, "step": 6663 }, { "epoch": 1.27, "learning_rate": 2.6216968541468065e-06, "loss": 3.6294, "step": 6664 }, { "epoch": 1.27, "learning_rate": 2.6216396568160153e-06, "loss": 3.1916, "step": 6665 }, { "epoch": 1.27, "learning_rate": 2.621582459485224e-06, "loss": 3.0152, "step": 6666 }, { "epoch": 1.27, "learning_rate": 2.621525262154433e-06, "loss": 3.1842, "step": 6667 }, { "epoch": 1.27, "learning_rate": 2.6214680648236417e-06, "loss": 3.0151, "step": 6668 }, { "epoch": 1.27, "learning_rate": 2.6214108674928505e-06, "loss": 3.3626, "step": 6669 }, { "epoch": 1.27, "learning_rate": 2.621353670162059e-06, "loss": 2.5922, "step": 6670 }, { "epoch": 1.27, "learning_rate": 2.621296472831268e-06, "loss": 2.5383, "step": 6671 }, { "epoch": 1.27, "learning_rate": 2.621239275500477e-06, "loss": 3.5106, "step": 6672 }, { "epoch": 1.27, "learning_rate": 2.6211820781696853e-06, "loss": 3.3256, "step": 6673 }, { "epoch": 1.27, "learning_rate": 2.6211248808388945e-06, "loss": 3.3689, "step": 6674 }, { "epoch": 1.27, "learning_rate": 2.621067683508103e-06, "loss": 3.4668, "step": 6675 }, { "epoch": 1.27, "learning_rate": 2.6210104861773116e-06, "loss": 2.797, "step": 6676 }, { "epoch": 1.27, "learning_rate": 2.6209532888465204e-06, "loss": 3.4865, "step": 6677 }, { "epoch": 1.27, "learning_rate": 2.6208960915157292e-06, "loss": 2.9898, "step": 6678 }, { "epoch": 1.27, "learning_rate": 2.620838894184938e-06, "loss": 2.869, "step": 6679 }, { "epoch": 1.27, "learning_rate": 2.620781696854147e-06, "loss": 3.1953, "step": 6680 }, { "epoch": 1.27, "learning_rate": 2.6207244995233556e-06, "loss": 3.5074, "step": 6681 }, { "epoch": 1.27, "learning_rate": 2.6206673021925644e-06, "loss": 3.7846, "step": 6682 }, { "epoch": 1.27, "learning_rate": 2.620610104861773e-06, "loss": 2.6301, "step": 6683 }, { "epoch": 1.27, "learning_rate": 2.620552907530982e-06, "loss": 2.9668, "step": 6684 }, { "epoch": 1.27, "learning_rate": 2.6204957102001908e-06, "loss": 3.2934, "step": 6685 }, { "epoch": 1.27, "learning_rate": 2.6204385128693996e-06, "loss": 2.8819, "step": 6686 }, { "epoch": 1.27, "learning_rate": 2.6203813155386084e-06, "loss": 3.2448, "step": 6687 }, { "epoch": 1.27, "learning_rate": 2.620324118207817e-06, "loss": 3.8508, "step": 6688 }, { "epoch": 1.27, "learning_rate": 2.620266920877026e-06, "loss": 3.0558, "step": 6689 }, { "epoch": 1.27, "learning_rate": 2.6202097235462343e-06, "loss": 3.2802, "step": 6690 }, { "epoch": 1.27, "learning_rate": 2.6201525262154435e-06, "loss": 2.6769, "step": 6691 }, { "epoch": 1.27, "learning_rate": 2.620095328884652e-06, "loss": 3.591, "step": 6692 }, { "epoch": 1.27, "learning_rate": 2.6200381315538607e-06, "loss": 3.7406, "step": 6693 }, { "epoch": 1.28, "learning_rate": 2.61998093422307e-06, "loss": 3.4626, "step": 6694 }, { "epoch": 1.28, "learning_rate": 2.6199237368922783e-06, "loss": 3.4121, "step": 6695 }, { "epoch": 1.28, "learning_rate": 2.6198665395614875e-06, "loss": 2.5143, "step": 6696 }, { "epoch": 1.28, "learning_rate": 2.619809342230696e-06, "loss": 3.029, "step": 6697 }, { "epoch": 1.28, "learning_rate": 2.6197521448999047e-06, "loss": 3.3179, "step": 6698 }, { "epoch": 1.28, "learning_rate": 2.6196949475691135e-06, "loss": 2.9153, "step": 6699 }, { "epoch": 1.28, "learning_rate": 2.6196377502383223e-06, "loss": 3.5831, "step": 6700 }, { "epoch": 1.28, "learning_rate": 2.619580552907531e-06, "loss": 2.95, "step": 6701 }, { "epoch": 1.28, "learning_rate": 2.61952335557674e-06, "loss": 3.2925, "step": 6702 }, { "epoch": 1.28, "learning_rate": 2.6194661582459486e-06, "loss": 3.1158, "step": 6703 }, { "epoch": 1.28, "learning_rate": 2.6194089609151574e-06, "loss": 2.5749, "step": 6704 }, { "epoch": 1.28, "learning_rate": 2.6193517635843662e-06, "loss": 2.9599, "step": 6705 }, { "epoch": 1.28, "learning_rate": 2.619294566253575e-06, "loss": 3.3725, "step": 6706 }, { "epoch": 1.28, "learning_rate": 2.6192373689227834e-06, "loss": 3.1375, "step": 6707 }, { "epoch": 1.28, "learning_rate": 2.6191801715919926e-06, "loss": 2.5738, "step": 6708 }, { "epoch": 1.28, "learning_rate": 2.619122974261201e-06, "loss": 2.8949, "step": 6709 }, { "epoch": 1.28, "learning_rate": 2.6190657769304098e-06, "loss": 3.4751, "step": 6710 }, { "epoch": 1.28, "learning_rate": 2.619008579599619e-06, "loss": 3.2302, "step": 6711 }, { "epoch": 1.28, "learning_rate": 2.6189513822688274e-06, "loss": 3.0225, "step": 6712 }, { "epoch": 1.28, "learning_rate": 2.6188941849380366e-06, "loss": 2.8801, "step": 6713 }, { "epoch": 1.28, "learning_rate": 2.618836987607245e-06, "loss": 3.3696, "step": 6714 }, { "epoch": 1.28, "learning_rate": 2.6187797902764537e-06, "loss": 3.9028, "step": 6715 }, { "epoch": 1.28, "learning_rate": 2.6187225929456625e-06, "loss": 3.6037, "step": 6716 }, { "epoch": 1.28, "learning_rate": 2.6186653956148713e-06, "loss": 3.4379, "step": 6717 }, { "epoch": 1.28, "learning_rate": 2.61860819828408e-06, "loss": 2.9501, "step": 6718 }, { "epoch": 1.28, "learning_rate": 2.618551000953289e-06, "loss": 2.8932, "step": 6719 }, { "epoch": 1.28, "learning_rate": 2.6184938036224977e-06, "loss": 3.6219, "step": 6720 }, { "epoch": 1.28, "learning_rate": 2.6184366062917065e-06, "loss": 2.8351, "step": 6721 }, { "epoch": 1.28, "learning_rate": 2.6183794089609153e-06, "loss": 3.7113, "step": 6722 }, { "epoch": 1.28, "learning_rate": 2.618322211630124e-06, "loss": 2.5875, "step": 6723 }, { "epoch": 1.28, "learning_rate": 2.6182650142993325e-06, "loss": 3.2413, "step": 6724 }, { "epoch": 1.28, "learning_rate": 2.6182078169685417e-06, "loss": 3.2836, "step": 6725 }, { "epoch": 1.28, "learning_rate": 2.61815061963775e-06, "loss": 3.9837, "step": 6726 }, { "epoch": 1.28, "learning_rate": 2.618093422306959e-06, "loss": 3.0691, "step": 6727 }, { "epoch": 1.28, "learning_rate": 2.618036224976168e-06, "loss": 2.9581, "step": 6728 }, { "epoch": 1.28, "learning_rate": 2.6179790276453764e-06, "loss": 2.7275, "step": 6729 }, { "epoch": 1.28, "learning_rate": 2.6179218303145857e-06, "loss": 3.3432, "step": 6730 }, { "epoch": 1.28, "learning_rate": 2.617864632983794e-06, "loss": 3.0971, "step": 6731 }, { "epoch": 1.28, "learning_rate": 2.617807435653003e-06, "loss": 3.5334, "step": 6732 }, { "epoch": 1.28, "learning_rate": 2.617750238322212e-06, "loss": 3.3718, "step": 6733 }, { "epoch": 1.28, "learning_rate": 2.6176930409914204e-06, "loss": 2.9884, "step": 6734 }, { "epoch": 1.28, "learning_rate": 2.617635843660629e-06, "loss": 3.5156, "step": 6735 }, { "epoch": 1.28, "learning_rate": 2.617578646329838e-06, "loss": 3.4813, "step": 6736 }, { "epoch": 1.28, "learning_rate": 2.6175214489990468e-06, "loss": 3.9743, "step": 6737 }, { "epoch": 1.28, "learning_rate": 2.6174642516682556e-06, "loss": 3.3481, "step": 6738 }, { "epoch": 1.28, "learning_rate": 2.6174070543374644e-06, "loss": 3.3369, "step": 6739 }, { "epoch": 1.28, "learning_rate": 2.617349857006673e-06, "loss": 3.1974, "step": 6740 }, { "epoch": 1.28, "learning_rate": 2.6172926596758815e-06, "loss": 3.5908, "step": 6741 }, { "epoch": 1.28, "learning_rate": 2.6172354623450908e-06, "loss": 3.3037, "step": 6742 }, { "epoch": 1.28, "learning_rate": 2.6171782650142996e-06, "loss": 3.7347, "step": 6743 }, { "epoch": 1.28, "learning_rate": 2.617121067683508e-06, "loss": 2.7781, "step": 6744 }, { "epoch": 1.28, "learning_rate": 2.617063870352717e-06, "loss": 3.3224, "step": 6745 }, { "epoch": 1.28, "learning_rate": 2.6170066730219255e-06, "loss": 3.4245, "step": 6746 }, { "epoch": 1.29, "learning_rate": 2.6169494756911347e-06, "loss": 3.1533, "step": 6747 }, { "epoch": 1.29, "learning_rate": 2.616892278360343e-06, "loss": 3.4835, "step": 6748 }, { "epoch": 1.29, "learning_rate": 2.616835081029552e-06, "loss": 2.8328, "step": 6749 }, { "epoch": 1.29, "learning_rate": 2.616777883698761e-06, "loss": 2.934, "step": 6750 }, { "epoch": 1.29, "learning_rate": 2.6167206863679695e-06, "loss": 3.0868, "step": 6751 }, { "epoch": 1.29, "learning_rate": 2.6166634890371783e-06, "loss": 3.2473, "step": 6752 }, { "epoch": 1.29, "learning_rate": 2.616606291706387e-06, "loss": 3.8874, "step": 6753 }, { "epoch": 1.29, "learning_rate": 2.616549094375596e-06, "loss": 3.4656, "step": 6754 }, { "epoch": 1.29, "learning_rate": 2.6164918970448047e-06, "loss": 2.7207, "step": 6755 }, { "epoch": 1.29, "learning_rate": 2.6164346997140134e-06, "loss": 2.6134, "step": 6756 }, { "epoch": 1.29, "learning_rate": 2.6163775023832222e-06, "loss": 3.9123, "step": 6757 }, { "epoch": 1.29, "learning_rate": 2.6163203050524306e-06, "loss": 3.7687, "step": 6758 }, { "epoch": 1.29, "learning_rate": 2.61626310772164e-06, "loss": 2.7494, "step": 6759 }, { "epoch": 1.29, "learning_rate": 2.6162059103908486e-06, "loss": 2.8344, "step": 6760 }, { "epoch": 1.29, "learning_rate": 2.6161487130600574e-06, "loss": 2.8038, "step": 6761 }, { "epoch": 1.29, "learning_rate": 2.616091515729266e-06, "loss": 3.6493, "step": 6762 }, { "epoch": 1.29, "learning_rate": 2.6160343183984746e-06, "loss": 3.1356, "step": 6763 }, { "epoch": 1.29, "learning_rate": 2.615977121067684e-06, "loss": 3.2366, "step": 6764 }, { "epoch": 1.29, "learning_rate": 2.615919923736892e-06, "loss": 3.9906, "step": 6765 }, { "epoch": 1.29, "learning_rate": 2.615862726406101e-06, "loss": 3.5249, "step": 6766 }, { "epoch": 1.29, "learning_rate": 2.61580552907531e-06, "loss": 3.1811, "step": 6767 }, { "epoch": 1.29, "learning_rate": 2.6157483317445185e-06, "loss": 3.2265, "step": 6768 }, { "epoch": 1.29, "learning_rate": 2.6156911344137273e-06, "loss": 2.6569, "step": 6769 }, { "epoch": 1.29, "learning_rate": 2.615633937082936e-06, "loss": 3.0011, "step": 6770 }, { "epoch": 1.29, "learning_rate": 2.615576739752145e-06, "loss": 2.6649, "step": 6771 }, { "epoch": 1.29, "learning_rate": 2.6155195424213537e-06, "loss": 2.9935, "step": 6772 }, { "epoch": 1.29, "learning_rate": 2.6154623450905625e-06, "loss": 2.6813, "step": 6773 }, { "epoch": 1.29, "learning_rate": 2.6154051477597713e-06, "loss": 3.1127, "step": 6774 }, { "epoch": 1.29, "learning_rate": 2.61534795042898e-06, "loss": 2.8423, "step": 6775 }, { "epoch": 1.29, "learning_rate": 2.615290753098189e-06, "loss": 3.0692, "step": 6776 }, { "epoch": 1.29, "learning_rate": 2.6152335557673977e-06, "loss": 3.0361, "step": 6777 }, { "epoch": 1.29, "learning_rate": 2.6151763584366065e-06, "loss": 3.2587, "step": 6778 }, { "epoch": 1.29, "learning_rate": 2.6151191611058153e-06, "loss": 2.8959, "step": 6779 }, { "epoch": 1.29, "learning_rate": 2.6150619637750237e-06, "loss": 3.0678, "step": 6780 }, { "epoch": 1.29, "learning_rate": 2.615004766444233e-06, "loss": 3.9007, "step": 6781 }, { "epoch": 1.29, "learning_rate": 2.6149475691134417e-06, "loss": 3.8407, "step": 6782 }, { "epoch": 1.29, "learning_rate": 2.61489037178265e-06, "loss": 3.0144, "step": 6783 }, { "epoch": 1.29, "learning_rate": 2.6148331744518593e-06, "loss": 2.72, "step": 6784 }, { "epoch": 1.29, "learning_rate": 2.6147759771210676e-06, "loss": 2.9789, "step": 6785 }, { "epoch": 1.29, "learning_rate": 2.6147187797902764e-06, "loss": 3.5349, "step": 6786 }, { "epoch": 1.29, "learning_rate": 2.614661582459485e-06, "loss": 3.6475, "step": 6787 }, { "epoch": 1.29, "learning_rate": 2.614604385128694e-06, "loss": 3.2224, "step": 6788 }, { "epoch": 1.29, "learning_rate": 2.614547187797903e-06, "loss": 2.9364, "step": 6789 }, { "epoch": 1.29, "learning_rate": 2.6144899904671116e-06, "loss": 3.1732, "step": 6790 }, { "epoch": 1.29, "learning_rate": 2.6144327931363204e-06, "loss": 3.533, "step": 6791 }, { "epoch": 1.29, "learning_rate": 2.614375595805529e-06, "loss": 2.7155, "step": 6792 }, { "epoch": 1.29, "learning_rate": 2.614318398474738e-06, "loss": 3.6979, "step": 6793 }, { "epoch": 1.29, "learning_rate": 2.6142612011439468e-06, "loss": 3.1536, "step": 6794 }, { "epoch": 1.29, "learning_rate": 2.6142040038131556e-06, "loss": 2.6373, "step": 6795 }, { "epoch": 1.29, "learning_rate": 2.6141468064823644e-06, "loss": 3.5991, "step": 6796 }, { "epoch": 1.29, "learning_rate": 2.6140896091515727e-06, "loss": 3.0573, "step": 6797 }, { "epoch": 1.29, "learning_rate": 2.614032411820782e-06, "loss": 3.2726, "step": 6798 }, { "epoch": 1.3, "learning_rate": 2.6139752144899907e-06, "loss": 2.8134, "step": 6799 }, { "epoch": 1.3, "learning_rate": 2.613918017159199e-06, "loss": 2.7991, "step": 6800 }, { "epoch": 1.3, "learning_rate": 2.6138608198284083e-06, "loss": 3.6646, "step": 6801 }, { "epoch": 1.3, "learning_rate": 2.6138036224976167e-06, "loss": 3.2972, "step": 6802 }, { "epoch": 1.3, "learning_rate": 2.6137464251668255e-06, "loss": 3.8029, "step": 6803 }, { "epoch": 1.3, "learning_rate": 2.6136892278360343e-06, "loss": 2.8965, "step": 6804 }, { "epoch": 1.3, "learning_rate": 2.613632030505243e-06, "loss": 3.4875, "step": 6805 }, { "epoch": 1.3, "learning_rate": 2.613574833174452e-06, "loss": 3.703, "step": 6806 }, { "epoch": 1.3, "learning_rate": 2.6135176358436607e-06, "loss": 2.6931, "step": 6807 }, { "epoch": 1.3, "learning_rate": 2.6134604385128695e-06, "loss": 2.7707, "step": 6808 }, { "epoch": 1.3, "learning_rate": 2.6134032411820782e-06, "loss": 3.2962, "step": 6809 }, { "epoch": 1.3, "learning_rate": 2.613346043851287e-06, "loss": 3.9632, "step": 6810 }, { "epoch": 1.3, "learning_rate": 2.613288846520496e-06, "loss": 2.8471, "step": 6811 }, { "epoch": 1.3, "learning_rate": 2.6132316491897046e-06, "loss": 3.1178, "step": 6812 }, { "epoch": 1.3, "learning_rate": 2.6131744518589134e-06, "loss": 3.3792, "step": 6813 }, { "epoch": 1.3, "learning_rate": 2.613117254528122e-06, "loss": 2.9457, "step": 6814 }, { "epoch": 1.3, "learning_rate": 2.613060057197331e-06, "loss": 3.36, "step": 6815 }, { "epoch": 1.3, "learning_rate": 2.61300285986654e-06, "loss": 3.6599, "step": 6816 }, { "epoch": 1.3, "learning_rate": 2.612945662535748e-06, "loss": 2.8872, "step": 6817 }, { "epoch": 1.3, "learning_rate": 2.6128884652049574e-06, "loss": 3.4049, "step": 6818 }, { "epoch": 1.3, "learning_rate": 2.6128312678741658e-06, "loss": 3.1152, "step": 6819 }, { "epoch": 1.3, "learning_rate": 2.6127740705433746e-06, "loss": 3.7707, "step": 6820 }, { "epoch": 1.3, "learning_rate": 2.6127168732125838e-06, "loss": 3.5066, "step": 6821 }, { "epoch": 1.3, "learning_rate": 2.612659675881792e-06, "loss": 3.0116, "step": 6822 }, { "epoch": 1.3, "learning_rate": 2.612602478551001e-06, "loss": 3.7041, "step": 6823 }, { "epoch": 1.3, "learning_rate": 2.6125452812202097e-06, "loss": 3.5858, "step": 6824 }, { "epoch": 1.3, "learning_rate": 2.6124880838894185e-06, "loss": 3.4561, "step": 6825 }, { "epoch": 1.3, "learning_rate": 2.6124308865586273e-06, "loss": 2.649, "step": 6826 }, { "epoch": 1.3, "learning_rate": 2.612373689227836e-06, "loss": 3.4877, "step": 6827 }, { "epoch": 1.3, "learning_rate": 2.612316491897045e-06, "loss": 3.3147, "step": 6828 }, { "epoch": 1.3, "learning_rate": 2.6122592945662537e-06, "loss": 2.9671, "step": 6829 }, { "epoch": 1.3, "learning_rate": 2.6122020972354625e-06, "loss": 2.6339, "step": 6830 }, { "epoch": 1.3, "learning_rate": 2.6121448999046713e-06, "loss": 3.2333, "step": 6831 }, { "epoch": 1.3, "learning_rate": 2.61208770257388e-06, "loss": 3.1494, "step": 6832 }, { "epoch": 1.3, "learning_rate": 2.612030505243089e-06, "loss": 3.3175, "step": 6833 }, { "epoch": 1.3, "learning_rate": 2.6119733079122972e-06, "loss": 2.8076, "step": 6834 }, { "epoch": 1.3, "learning_rate": 2.6119161105815065e-06, "loss": 2.4389, "step": 6835 }, { "epoch": 1.3, "learning_rate": 2.611858913250715e-06, "loss": 3.2483, "step": 6836 }, { "epoch": 1.3, "learning_rate": 2.6118017159199236e-06, "loss": 3.8618, "step": 6837 }, { "epoch": 1.3, "learning_rate": 2.611744518589133e-06, "loss": 2.9172, "step": 6838 }, { "epoch": 1.3, "learning_rate": 2.6116873212583412e-06, "loss": 3.4224, "step": 6839 }, { "epoch": 1.3, "learning_rate": 2.61163012392755e-06, "loss": 3.2539, "step": 6840 }, { "epoch": 1.3, "learning_rate": 2.611572926596759e-06, "loss": 3.0657, "step": 6841 }, { "epoch": 1.3, "learning_rate": 2.6115157292659676e-06, "loss": 3.2474, "step": 6842 }, { "epoch": 1.3, "learning_rate": 2.6114585319351764e-06, "loss": 3.6276, "step": 6843 }, { "epoch": 1.3, "learning_rate": 2.611401334604385e-06, "loss": 2.3692, "step": 6844 }, { "epoch": 1.3, "learning_rate": 2.611344137273594e-06, "loss": 2.482, "step": 6845 }, { "epoch": 1.3, "learning_rate": 2.6112869399428028e-06, "loss": 2.9431, "step": 6846 }, { "epoch": 1.3, "learning_rate": 2.6112297426120116e-06, "loss": 3.3511, "step": 6847 }, { "epoch": 1.3, "learning_rate": 2.6111725452812204e-06, "loss": 3.4315, "step": 6848 }, { "epoch": 1.3, "learning_rate": 2.611115347950429e-06, "loss": 2.5345, "step": 6849 }, { "epoch": 1.3, "learning_rate": 2.611058150619638e-06, "loss": 3.0378, "step": 6850 }, { "epoch": 1.3, "learning_rate": 2.6110009532888463e-06, "loss": 3.2582, "step": 6851 }, { "epoch": 1.31, "learning_rate": 2.6109437559580555e-06, "loss": 3.0153, "step": 6852 }, { "epoch": 1.31, "learning_rate": 2.610886558627264e-06, "loss": 3.6638, "step": 6853 }, { "epoch": 1.31, "learning_rate": 2.6108293612964727e-06, "loss": 3.7821, "step": 6854 }, { "epoch": 1.31, "learning_rate": 2.610772163965682e-06, "loss": 3.3861, "step": 6855 }, { "epoch": 1.31, "learning_rate": 2.6107149666348903e-06, "loss": 3.4729, "step": 6856 }, { "epoch": 1.31, "learning_rate": 2.610657769304099e-06, "loss": 3.3118, "step": 6857 }, { "epoch": 1.31, "learning_rate": 2.610600571973308e-06, "loss": 3.5517, "step": 6858 }, { "epoch": 1.31, "learning_rate": 2.6105433746425167e-06, "loss": 2.6367, "step": 6859 }, { "epoch": 1.31, "learning_rate": 2.610486177311726e-06, "loss": 3.808, "step": 6860 }, { "epoch": 1.31, "learning_rate": 2.6104289799809343e-06, "loss": 3.6277, "step": 6861 }, { "epoch": 1.31, "learning_rate": 2.610371782650143e-06, "loss": 3.2625, "step": 6862 }, { "epoch": 1.31, "learning_rate": 2.610314585319352e-06, "loss": 2.6368, "step": 6863 }, { "epoch": 1.31, "learning_rate": 2.6102573879885606e-06, "loss": 2.9191, "step": 6864 }, { "epoch": 1.31, "learning_rate": 2.6102001906577694e-06, "loss": 3.4777, "step": 6865 }, { "epoch": 1.31, "learning_rate": 2.6101429933269782e-06, "loss": 3.0292, "step": 6866 }, { "epoch": 1.31, "learning_rate": 2.610085795996187e-06, "loss": 3.6149, "step": 6867 }, { "epoch": 1.31, "learning_rate": 2.6100285986653954e-06, "loss": 3.1073, "step": 6868 }, { "epoch": 1.31, "learning_rate": 2.6099714013346046e-06, "loss": 3.2133, "step": 6869 }, { "epoch": 1.31, "learning_rate": 2.6099142040038134e-06, "loss": 3.6527, "step": 6870 }, { "epoch": 1.31, "learning_rate": 2.6098570066730218e-06, "loss": 2.8425, "step": 6871 }, { "epoch": 1.31, "learning_rate": 2.609799809342231e-06, "loss": 3.2031, "step": 6872 }, { "epoch": 1.31, "learning_rate": 2.6097426120114394e-06, "loss": 3.0175, "step": 6873 }, { "epoch": 1.31, "learning_rate": 2.609685414680648e-06, "loss": 3.258, "step": 6874 }, { "epoch": 1.31, "learning_rate": 2.609628217349857e-06, "loss": 2.8231, "step": 6875 }, { "epoch": 1.31, "learning_rate": 2.6095710200190657e-06, "loss": 3.0074, "step": 6876 }, { "epoch": 1.31, "learning_rate": 2.609513822688275e-06, "loss": 2.8049, "step": 6877 }, { "epoch": 1.31, "learning_rate": 2.6094566253574833e-06, "loss": 3.5533, "step": 6878 }, { "epoch": 1.31, "learning_rate": 2.609399428026692e-06, "loss": 3.1986, "step": 6879 }, { "epoch": 1.31, "learning_rate": 2.609342230695901e-06, "loss": 2.7454, "step": 6880 }, { "epoch": 1.31, "learning_rate": 2.6092850333651097e-06, "loss": 3.704, "step": 6881 }, { "epoch": 1.31, "learning_rate": 2.6092278360343185e-06, "loss": 3.2456, "step": 6882 }, { "epoch": 1.31, "learning_rate": 2.6091706387035273e-06, "loss": 3.1326, "step": 6883 }, { "epoch": 1.31, "learning_rate": 2.609113441372736e-06, "loss": 2.8022, "step": 6884 }, { "epoch": 1.31, "learning_rate": 2.6090562440419445e-06, "loss": 3.3783, "step": 6885 }, { "epoch": 1.31, "learning_rate": 2.6089990467111537e-06, "loss": 3.7098, "step": 6886 }, { "epoch": 1.31, "learning_rate": 2.6089418493803625e-06, "loss": 2.9324, "step": 6887 }, { "epoch": 1.31, "learning_rate": 2.608884652049571e-06, "loss": 3.0006, "step": 6888 }, { "epoch": 1.31, "learning_rate": 2.60882745471878e-06, "loss": 3.1494, "step": 6889 }, { "epoch": 1.31, "learning_rate": 2.6087702573879884e-06, "loss": 2.8995, "step": 6890 }, { "epoch": 1.31, "learning_rate": 2.6087130600571976e-06, "loss": 2.9429, "step": 6891 }, { "epoch": 1.31, "learning_rate": 2.608655862726406e-06, "loss": 3.0958, "step": 6892 }, { "epoch": 1.31, "learning_rate": 2.608598665395615e-06, "loss": 3.2067, "step": 6893 }, { "epoch": 1.31, "learning_rate": 2.608541468064824e-06, "loss": 2.6093, "step": 6894 }, { "epoch": 1.31, "learning_rate": 2.6084842707340324e-06, "loss": 3.3677, "step": 6895 }, { "epoch": 1.31, "learning_rate": 2.608427073403241e-06, "loss": 3.3825, "step": 6896 }, { "epoch": 1.31, "learning_rate": 2.60836987607245e-06, "loss": 2.7364, "step": 6897 }, { "epoch": 1.31, "learning_rate": 2.6083126787416588e-06, "loss": 2.7328, "step": 6898 }, { "epoch": 1.31, "learning_rate": 2.6082554814108676e-06, "loss": 3.3397, "step": 6899 }, { "epoch": 1.31, "learning_rate": 2.6081982840800764e-06, "loss": 2.9547, "step": 6900 }, { "epoch": 1.31, "learning_rate": 2.608141086749285e-06, "loss": 2.5341, "step": 6901 }, { "epoch": 1.31, "learning_rate": 2.6080838894184935e-06, "loss": 2.582, "step": 6902 }, { "epoch": 1.31, "learning_rate": 2.6080266920877027e-06, "loss": 3.1672, "step": 6903 }, { "epoch": 1.32, "learning_rate": 2.6079694947569115e-06, "loss": 3.5275, "step": 6904 }, { "epoch": 1.32, "learning_rate": 2.60791229742612e-06, "loss": 3.8139, "step": 6905 }, { "epoch": 1.32, "learning_rate": 2.607855100095329e-06, "loss": 3.0995, "step": 6906 }, { "epoch": 1.32, "learning_rate": 2.6077979027645375e-06, "loss": 3.5652, "step": 6907 }, { "epoch": 1.32, "learning_rate": 2.6077407054337467e-06, "loss": 3.6621, "step": 6908 }, { "epoch": 1.32, "learning_rate": 2.6076835081029555e-06, "loss": 3.2746, "step": 6909 }, { "epoch": 1.32, "learning_rate": 2.607626310772164e-06, "loss": 3.5131, "step": 6910 }, { "epoch": 1.32, "learning_rate": 2.607569113441373e-06, "loss": 3.3864, "step": 6911 }, { "epoch": 1.32, "learning_rate": 2.6075119161105815e-06, "loss": 3.3452, "step": 6912 }, { "epoch": 1.32, "learning_rate": 2.6074547187797903e-06, "loss": 3.1994, "step": 6913 }, { "epoch": 1.32, "learning_rate": 2.607397521448999e-06, "loss": 3.0604, "step": 6914 }, { "epoch": 1.32, "learning_rate": 2.607340324118208e-06, "loss": 2.7257, "step": 6915 }, { "epoch": 1.32, "learning_rate": 2.6072831267874166e-06, "loss": 3.5489, "step": 6916 }, { "epoch": 1.32, "learning_rate": 2.6072259294566254e-06, "loss": 2.8256, "step": 6917 }, { "epoch": 1.32, "learning_rate": 2.6071687321258342e-06, "loss": 3.3437, "step": 6918 }, { "epoch": 1.32, "learning_rate": 2.607111534795043e-06, "loss": 3.4239, "step": 6919 }, { "epoch": 1.32, "learning_rate": 2.607054337464252e-06, "loss": 3.4562, "step": 6920 }, { "epoch": 1.32, "learning_rate": 2.6069971401334606e-06, "loss": 3.9626, "step": 6921 }, { "epoch": 1.32, "learning_rate": 2.606939942802669e-06, "loss": 3.3292, "step": 6922 }, { "epoch": 1.32, "learning_rate": 2.606882745471878e-06, "loss": 3.2386, "step": 6923 }, { "epoch": 1.32, "learning_rate": 2.6068255481410866e-06, "loss": 2.9545, "step": 6924 }, { "epoch": 1.32, "learning_rate": 2.606768350810296e-06, "loss": 3.2594, "step": 6925 }, { "epoch": 1.32, "learning_rate": 2.6067111534795046e-06, "loss": 3.7302, "step": 6926 }, { "epoch": 1.32, "learning_rate": 2.606653956148713e-06, "loss": 2.9802, "step": 6927 }, { "epoch": 1.32, "learning_rate": 2.606596758817922e-06, "loss": 3.3041, "step": 6928 }, { "epoch": 1.32, "learning_rate": 2.6065395614871305e-06, "loss": 2.2388, "step": 6929 }, { "epoch": 1.32, "learning_rate": 2.6064823641563393e-06, "loss": 3.619, "step": 6930 }, { "epoch": 1.32, "learning_rate": 2.606425166825548e-06, "loss": 2.5373, "step": 6931 }, { "epoch": 1.32, "learning_rate": 2.606367969494757e-06, "loss": 3.9866, "step": 6932 }, { "epoch": 1.32, "learning_rate": 2.6063107721639657e-06, "loss": 2.7384, "step": 6933 }, { "epoch": 1.32, "learning_rate": 2.6062535748331745e-06, "loss": 3.4767, "step": 6934 }, { "epoch": 1.32, "learning_rate": 2.6061963775023833e-06, "loss": 2.9158, "step": 6935 }, { "epoch": 1.32, "learning_rate": 2.606139180171592e-06, "loss": 3.3914, "step": 6936 }, { "epoch": 1.32, "learning_rate": 2.606081982840801e-06, "loss": 2.9205, "step": 6937 }, { "epoch": 1.32, "learning_rate": 2.6060247855100097e-06, "loss": 3.3647, "step": 6938 }, { "epoch": 1.32, "learning_rate": 2.605967588179218e-06, "loss": 3.416, "step": 6939 }, { "epoch": 1.32, "learning_rate": 2.6059103908484273e-06, "loss": 2.7669, "step": 6940 }, { "epoch": 1.32, "learning_rate": 2.6058531935176356e-06, "loss": 2.9093, "step": 6941 }, { "epoch": 1.32, "learning_rate": 2.605795996186845e-06, "loss": 3.0623, "step": 6942 }, { "epoch": 1.32, "learning_rate": 2.6057387988560537e-06, "loss": 3.5882, "step": 6943 }, { "epoch": 1.32, "learning_rate": 2.605681601525262e-06, "loss": 3.0135, "step": 6944 }, { "epoch": 1.32, "learning_rate": 2.6056244041944712e-06, "loss": 2.523, "step": 6945 }, { "epoch": 1.32, "learning_rate": 2.6055672068636796e-06, "loss": 3.8233, "step": 6946 }, { "epoch": 1.32, "learning_rate": 2.6055100095328884e-06, "loss": 2.9382, "step": 6947 }, { "epoch": 1.32, "learning_rate": 2.605452812202097e-06, "loss": 3.2266, "step": 6948 }, { "epoch": 1.32, "learning_rate": 2.605395614871306e-06, "loss": 2.2396, "step": 6949 }, { "epoch": 1.32, "learning_rate": 2.605338417540515e-06, "loss": 2.9509, "step": 6950 }, { "epoch": 1.32, "learning_rate": 2.6052812202097236e-06, "loss": 3.001, "step": 6951 }, { "epoch": 1.32, "learning_rate": 2.6052240228789324e-06, "loss": 3.6513, "step": 6952 }, { "epoch": 1.32, "learning_rate": 2.605166825548141e-06, "loss": 2.9111, "step": 6953 }, { "epoch": 1.32, "learning_rate": 2.60510962821735e-06, "loss": 2.698, "step": 6954 }, { "epoch": 1.32, "learning_rate": 2.6050524308865588e-06, "loss": 3.4999, "step": 6955 }, { "epoch": 1.32, "learning_rate": 2.604995233555767e-06, "loss": 3.3477, "step": 6956 }, { "epoch": 1.33, "learning_rate": 2.6049380362249763e-06, "loss": 3.6408, "step": 6957 }, { "epoch": 1.33, "learning_rate": 2.604880838894185e-06, "loss": 3.038, "step": 6958 }, { "epoch": 1.33, "learning_rate": 2.604823641563394e-06, "loss": 3.44, "step": 6959 }, { "epoch": 1.33, "learning_rate": 2.6047664442326027e-06, "loss": 2.6408, "step": 6960 }, { "epoch": 1.33, "learning_rate": 2.604709246901811e-06, "loss": 2.9511, "step": 6961 }, { "epoch": 1.33, "learning_rate": 2.6046520495710203e-06, "loss": 2.8109, "step": 6962 }, { "epoch": 1.33, "learning_rate": 2.6045948522402287e-06, "loss": 3.0591, "step": 6963 }, { "epoch": 1.33, "learning_rate": 2.6045376549094375e-06, "loss": 3.0912, "step": 6964 }, { "epoch": 1.33, "learning_rate": 2.6044804575786467e-06, "loss": 2.9073, "step": 6965 }, { "epoch": 1.33, "learning_rate": 2.604423260247855e-06, "loss": 3.2583, "step": 6966 }, { "epoch": 1.33, "learning_rate": 2.604366062917064e-06, "loss": 3.2257, "step": 6967 }, { "epoch": 1.33, "learning_rate": 2.6043088655862727e-06, "loss": 3.1342, "step": 6968 }, { "epoch": 1.33, "learning_rate": 2.6042516682554814e-06, "loss": 3.4224, "step": 6969 }, { "epoch": 1.33, "learning_rate": 2.6041944709246902e-06, "loss": 3.4619, "step": 6970 }, { "epoch": 1.33, "learning_rate": 2.604137273593899e-06, "loss": 3.1215, "step": 6971 }, { "epoch": 1.33, "learning_rate": 2.604080076263108e-06, "loss": 2.8159, "step": 6972 }, { "epoch": 1.33, "learning_rate": 2.6040228789323166e-06, "loss": 2.7789, "step": 6973 }, { "epoch": 1.33, "learning_rate": 2.6039656816015254e-06, "loss": 2.8956, "step": 6974 }, { "epoch": 1.33, "learning_rate": 2.603908484270734e-06, "loss": 3.3263, "step": 6975 }, { "epoch": 1.33, "learning_rate": 2.603851286939943e-06, "loss": 2.8512, "step": 6976 }, { "epoch": 1.33, "learning_rate": 2.603794089609152e-06, "loss": 3.5235, "step": 6977 }, { "epoch": 1.33, "learning_rate": 2.60373689227836e-06, "loss": 3.3818, "step": 6978 }, { "epoch": 1.33, "learning_rate": 2.6036796949475694e-06, "loss": 2.9905, "step": 6979 }, { "epoch": 1.33, "learning_rate": 2.6036224976167778e-06, "loss": 3.5185, "step": 6980 }, { "epoch": 1.33, "learning_rate": 2.6035653002859866e-06, "loss": 3.4266, "step": 6981 }, { "epoch": 1.33, "learning_rate": 2.6035081029551958e-06, "loss": 3.5189, "step": 6982 }, { "epoch": 1.33, "learning_rate": 2.603450905624404e-06, "loss": 2.5273, "step": 6983 }, { "epoch": 1.33, "learning_rate": 2.603393708293613e-06, "loss": 2.999, "step": 6984 }, { "epoch": 1.33, "learning_rate": 2.6033365109628217e-06, "loss": 2.7693, "step": 6985 }, { "epoch": 1.33, "learning_rate": 2.6032793136320305e-06, "loss": 3.5417, "step": 6986 }, { "epoch": 1.33, "learning_rate": 2.6032221163012393e-06, "loss": 2.9465, "step": 6987 }, { "epoch": 1.33, "learning_rate": 2.603164918970448e-06, "loss": 3.6824, "step": 6988 }, { "epoch": 1.33, "learning_rate": 2.603107721639657e-06, "loss": 2.6333, "step": 6989 }, { "epoch": 1.33, "learning_rate": 2.6030505243088657e-06, "loss": 3.3114, "step": 6990 }, { "epoch": 1.33, "learning_rate": 2.6029933269780745e-06, "loss": 2.9916, "step": 6991 }, { "epoch": 1.33, "learning_rate": 2.6029361296472833e-06, "loss": 3.0607, "step": 6992 }, { "epoch": 1.33, "learning_rate": 2.602878932316492e-06, "loss": 3.0234, "step": 6993 }, { "epoch": 1.33, "learning_rate": 2.602821734985701e-06, "loss": 3.0299, "step": 6994 }, { "epoch": 1.33, "learning_rate": 2.6027645376549092e-06, "loss": 3.5485, "step": 6995 }, { "epoch": 1.33, "learning_rate": 2.6027073403241185e-06, "loss": 2.7704, "step": 6996 }, { "epoch": 1.33, "learning_rate": 2.6026501429933273e-06, "loss": 2.9947, "step": 6997 }, { "epoch": 1.33, "learning_rate": 2.6025929456625356e-06, "loss": 3.2924, "step": 6998 }, { "epoch": 1.33, "learning_rate": 2.602535748331745e-06, "loss": 2.5449, "step": 6999 }, { "epoch": 1.33, "learning_rate": 2.602478551000953e-06, "loss": 3.7029, "step": 7000 }, { "epoch": 1.33, "learning_rate": 2.602421353670162e-06, "loss": 3.1036, "step": 7001 }, { "epoch": 1.33, "learning_rate": 2.602364156339371e-06, "loss": 2.6061, "step": 7002 }, { "epoch": 1.33, "learning_rate": 2.6023069590085796e-06, "loss": 2.5341, "step": 7003 }, { "epoch": 1.33, "learning_rate": 2.6022497616777884e-06, "loss": 2.7662, "step": 7004 }, { "epoch": 1.33, "learning_rate": 2.602192564346997e-06, "loss": 2.8013, "step": 7005 }, { "epoch": 1.33, "learning_rate": 2.602135367016206e-06, "loss": 3.4329, "step": 7006 }, { "epoch": 1.33, "learning_rate": 2.6020781696854148e-06, "loss": 3.3071, "step": 7007 }, { "epoch": 1.33, "learning_rate": 2.6020209723546236e-06, "loss": 2.9666, "step": 7008 }, { "epoch": 1.34, "learning_rate": 2.6019637750238324e-06, "loss": 2.8516, "step": 7009 }, { "epoch": 1.34, "learning_rate": 2.601906577693041e-06, "loss": 3.188, "step": 7010 }, { "epoch": 1.34, "learning_rate": 2.60184938036225e-06, "loss": 2.8129, "step": 7011 }, { "epoch": 1.34, "learning_rate": 2.6017921830314583e-06, "loss": 3.1896, "step": 7012 }, { "epoch": 1.34, "learning_rate": 2.6017349857006675e-06, "loss": 3.4022, "step": 7013 }, { "epoch": 1.34, "learning_rate": 2.6016777883698763e-06, "loss": 3.1718, "step": 7014 }, { "epoch": 1.34, "learning_rate": 2.6016205910390847e-06, "loss": 3.3169, "step": 7015 }, { "epoch": 1.34, "learning_rate": 2.601563393708294e-06, "loss": 3.4173, "step": 7016 }, { "epoch": 1.34, "learning_rate": 2.6015061963775023e-06, "loss": 2.8922, "step": 7017 }, { "epoch": 1.34, "learning_rate": 2.601448999046711e-06, "loss": 3.4392, "step": 7018 }, { "epoch": 1.34, "learning_rate": 2.60139180171592e-06, "loss": 2.6519, "step": 7019 }, { "epoch": 1.34, "learning_rate": 2.6013346043851287e-06, "loss": 3.045, "step": 7020 }, { "epoch": 1.34, "learning_rate": 2.6012774070543375e-06, "loss": 2.4047, "step": 7021 }, { "epoch": 1.34, "learning_rate": 2.6012202097235462e-06, "loss": 3.1304, "step": 7022 }, { "epoch": 1.34, "learning_rate": 2.601163012392755e-06, "loss": 3.6141, "step": 7023 }, { "epoch": 1.34, "learning_rate": 2.601105815061964e-06, "loss": 3.2901, "step": 7024 }, { "epoch": 1.34, "learning_rate": 2.6010486177311726e-06, "loss": 3.0314, "step": 7025 }, { "epoch": 1.34, "learning_rate": 2.6009914204003814e-06, "loss": 4.0759, "step": 7026 }, { "epoch": 1.34, "learning_rate": 2.6009342230695902e-06, "loss": 3.5136, "step": 7027 }, { "epoch": 1.34, "learning_rate": 2.600877025738799e-06, "loss": 3.6843, "step": 7028 }, { "epoch": 1.34, "learning_rate": 2.6008198284080074e-06, "loss": 3.0667, "step": 7029 }, { "epoch": 1.34, "learning_rate": 2.6007626310772166e-06, "loss": 3.0324, "step": 7030 }, { "epoch": 1.34, "learning_rate": 2.6007054337464254e-06, "loss": 3.9721, "step": 7031 }, { "epoch": 1.34, "learning_rate": 2.6006482364156338e-06, "loss": 2.6228, "step": 7032 }, { "epoch": 1.34, "learning_rate": 2.600591039084843e-06, "loss": 2.6126, "step": 7033 }, { "epoch": 1.34, "learning_rate": 2.6005338417540514e-06, "loss": 3.0116, "step": 7034 }, { "epoch": 1.34, "learning_rate": 2.60047664442326e-06, "loss": 2.7204, "step": 7035 }, { "epoch": 1.34, "learning_rate": 2.600419447092469e-06, "loss": 2.83, "step": 7036 }, { "epoch": 1.34, "learning_rate": 2.6003622497616777e-06, "loss": 3.5521, "step": 7037 }, { "epoch": 1.34, "learning_rate": 2.600305052430887e-06, "loss": 3.4322, "step": 7038 }, { "epoch": 1.34, "learning_rate": 2.6002478551000953e-06, "loss": 2.9617, "step": 7039 }, { "epoch": 1.34, "learning_rate": 2.600190657769304e-06, "loss": 3.9572, "step": 7040 }, { "epoch": 1.34, "learning_rate": 2.600133460438513e-06, "loss": 2.9912, "step": 7041 }, { "epoch": 1.34, "learning_rate": 2.6000762631077217e-06, "loss": 3.3356, "step": 7042 }, { "epoch": 1.34, "learning_rate": 2.6000190657769305e-06, "loss": 3.3987, "step": 7043 }, { "epoch": 1.34, "learning_rate": 2.5999618684461393e-06, "loss": 2.4542, "step": 7044 }, { "epoch": 1.34, "learning_rate": 2.599904671115348e-06, "loss": 3.1328, "step": 7045 }, { "epoch": 1.34, "learning_rate": 2.599847473784557e-06, "loss": 3.6505, "step": 7046 }, { "epoch": 1.34, "learning_rate": 2.5997902764537657e-06, "loss": 3.5286, "step": 7047 }, { "epoch": 1.34, "learning_rate": 2.5997330791229745e-06, "loss": 3.0541, "step": 7048 }, { "epoch": 1.34, "learning_rate": 2.599675881792183e-06, "loss": 2.9301, "step": 7049 }, { "epoch": 1.34, "learning_rate": 2.599618684461392e-06, "loss": 2.9145, "step": 7050 }, { "epoch": 1.34, "learning_rate": 2.5995614871306004e-06, "loss": 2.7646, "step": 7051 }, { "epoch": 1.34, "learning_rate": 2.5995042897998092e-06, "loss": 3.7121, "step": 7052 }, { "epoch": 1.34, "learning_rate": 2.5994470924690184e-06, "loss": 2.7571, "step": 7053 }, { "epoch": 1.34, "learning_rate": 2.599389895138227e-06, "loss": 3.7606, "step": 7054 }, { "epoch": 1.34, "learning_rate": 2.599332697807436e-06, "loss": 3.0814, "step": 7055 }, { "epoch": 1.34, "learning_rate": 2.5992755004766444e-06, "loss": 2.7909, "step": 7056 }, { "epoch": 1.34, "learning_rate": 2.599218303145853e-06, "loss": 3.0945, "step": 7057 }, { "epoch": 1.34, "learning_rate": 2.599161105815062e-06, "loss": 2.2955, "step": 7058 }, { "epoch": 1.34, "learning_rate": 2.5991039084842708e-06, "loss": 2.8358, "step": 7059 }, { "epoch": 1.34, "learning_rate": 2.5990467111534796e-06, "loss": 3.1656, "step": 7060 }, { "epoch": 1.34, "learning_rate": 2.5989895138226884e-06, "loss": 2.637, "step": 7061 }, { "epoch": 1.35, "learning_rate": 2.598932316491897e-06, "loss": 2.8395, "step": 7062 }, { "epoch": 1.35, "learning_rate": 2.598875119161106e-06, "loss": 2.832, "step": 7063 }, { "epoch": 1.35, "learning_rate": 2.5988179218303147e-06, "loss": 3.7733, "step": 7064 }, { "epoch": 1.35, "learning_rate": 2.5987607244995235e-06, "loss": 3.1374, "step": 7065 }, { "epoch": 1.35, "learning_rate": 2.598703527168732e-06, "loss": 2.6841, "step": 7066 }, { "epoch": 1.35, "learning_rate": 2.598646329837941e-06, "loss": 3.662, "step": 7067 }, { "epoch": 1.35, "learning_rate": 2.5985891325071495e-06, "loss": 3.4987, "step": 7068 }, { "epoch": 1.35, "learning_rate": 2.5985319351763583e-06, "loss": 2.9834, "step": 7069 }, { "epoch": 1.35, "learning_rate": 2.5984747378455675e-06, "loss": 2.9314, "step": 7070 }, { "epoch": 1.35, "learning_rate": 2.598417540514776e-06, "loss": 3.0766, "step": 7071 }, { "epoch": 1.35, "learning_rate": 2.598360343183985e-06, "loss": 2.8465, "step": 7072 }, { "epoch": 1.35, "learning_rate": 2.5983031458531935e-06, "loss": 3.0823, "step": 7073 }, { "epoch": 1.35, "learning_rate": 2.5982459485224023e-06, "loss": 3.3037, "step": 7074 }, { "epoch": 1.35, "learning_rate": 2.598188751191611e-06, "loss": 3.3063, "step": 7075 }, { "epoch": 1.35, "learning_rate": 2.59813155386082e-06, "loss": 3.0053, "step": 7076 }, { "epoch": 1.35, "learning_rate": 2.5980743565300286e-06, "loss": 2.7372, "step": 7077 }, { "epoch": 1.35, "learning_rate": 2.5980171591992374e-06, "loss": 2.9914, "step": 7078 }, { "epoch": 1.35, "learning_rate": 2.5979599618684462e-06, "loss": 3.1787, "step": 7079 }, { "epoch": 1.35, "learning_rate": 2.597902764537655e-06, "loss": 3.477, "step": 7080 }, { "epoch": 1.35, "learning_rate": 2.597845567206864e-06, "loss": 2.6007, "step": 7081 }, { "epoch": 1.35, "learning_rate": 2.5977883698760726e-06, "loss": 2.7299, "step": 7082 }, { "epoch": 1.35, "learning_rate": 2.597731172545281e-06, "loss": 3.4502, "step": 7083 }, { "epoch": 1.35, "learning_rate": 2.59767397521449e-06, "loss": 3.0781, "step": 7084 }, { "epoch": 1.35, "learning_rate": 2.597616777883699e-06, "loss": 2.8543, "step": 7085 }, { "epoch": 1.35, "learning_rate": 2.5975595805529074e-06, "loss": 3.0341, "step": 7086 }, { "epoch": 1.35, "learning_rate": 2.5975023832221166e-06, "loss": 3.2332, "step": 7087 }, { "epoch": 1.35, "learning_rate": 2.597445185891325e-06, "loss": 3.3828, "step": 7088 }, { "epoch": 1.35, "learning_rate": 2.597387988560534e-06, "loss": 3.23, "step": 7089 }, { "epoch": 1.35, "learning_rate": 2.5973307912297425e-06, "loss": 2.5636, "step": 7090 }, { "epoch": 1.35, "learning_rate": 2.5972735938989513e-06, "loss": 3.1133, "step": 7091 }, { "epoch": 1.35, "learning_rate": 2.5972163965681605e-06, "loss": 3.3135, "step": 7092 }, { "epoch": 1.35, "learning_rate": 2.597159199237369e-06, "loss": 3.2126, "step": 7093 }, { "epoch": 1.35, "learning_rate": 2.5971020019065777e-06, "loss": 3.2806, "step": 7094 }, { "epoch": 1.35, "learning_rate": 2.5970448045757865e-06, "loss": 2.7709, "step": 7095 }, { "epoch": 1.35, "learning_rate": 2.5969876072449953e-06, "loss": 2.3832, "step": 7096 }, { "epoch": 1.35, "learning_rate": 2.596930409914204e-06, "loss": 3.4113, "step": 7097 }, { "epoch": 1.35, "learning_rate": 2.596873212583413e-06, "loss": 3.0666, "step": 7098 }, { "epoch": 1.35, "learning_rate": 2.5968160152526217e-06, "loss": 3.1313, "step": 7099 }, { "epoch": 1.35, "learning_rate": 2.59675881792183e-06, "loss": 3.7405, "step": 7100 }, { "epoch": 1.35, "learning_rate": 2.5967016205910393e-06, "loss": 2.4314, "step": 7101 }, { "epoch": 1.35, "learning_rate": 2.596644423260248e-06, "loss": 3.4183, "step": 7102 }, { "epoch": 1.35, "learning_rate": 2.596587225929457e-06, "loss": 2.8523, "step": 7103 }, { "epoch": 1.35, "learning_rate": 2.5965300285986656e-06, "loss": 3.9382, "step": 7104 }, { "epoch": 1.35, "learning_rate": 2.596472831267874e-06, "loss": 3.6548, "step": 7105 }, { "epoch": 1.35, "learning_rate": 2.5964156339370832e-06, "loss": 3.3699, "step": 7106 }, { "epoch": 1.35, "learning_rate": 2.5963584366062916e-06, "loss": 3.5032, "step": 7107 }, { "epoch": 1.35, "learning_rate": 2.5963012392755004e-06, "loss": 3.3032, "step": 7108 }, { "epoch": 1.35, "learning_rate": 2.5962440419447096e-06, "loss": 3.734, "step": 7109 }, { "epoch": 1.35, "learning_rate": 2.596186844613918e-06, "loss": 3.0154, "step": 7110 }, { "epoch": 1.35, "learning_rate": 2.5961296472831268e-06, "loss": 3.774, "step": 7111 }, { "epoch": 1.35, "learning_rate": 2.5960724499523356e-06, "loss": 2.9805, "step": 7112 }, { "epoch": 1.35, "learning_rate": 2.5960152526215444e-06, "loss": 2.7221, "step": 7113 }, { "epoch": 1.36, "learning_rate": 2.595958055290753e-06, "loss": 3.1102, "step": 7114 }, { "epoch": 1.36, "learning_rate": 2.595900857959962e-06, "loss": 3.2088, "step": 7115 }, { "epoch": 1.36, "learning_rate": 2.5958436606291708e-06, "loss": 3.1224, "step": 7116 }, { "epoch": 1.36, "learning_rate": 2.595786463298379e-06, "loss": 3.971, "step": 7117 }, { "epoch": 1.36, "learning_rate": 2.5957292659675883e-06, "loss": 3.5353, "step": 7118 }, { "epoch": 1.36, "learning_rate": 2.595672068636797e-06, "loss": 3.2057, "step": 7119 }, { "epoch": 1.36, "learning_rate": 2.595614871306006e-06, "loss": 3.5862, "step": 7120 }, { "epoch": 1.36, "learning_rate": 2.5955576739752147e-06, "loss": 3.2224, "step": 7121 }, { "epoch": 1.36, "learning_rate": 2.595500476644423e-06, "loss": 3.8772, "step": 7122 }, { "epoch": 1.36, "learning_rate": 2.5954432793136323e-06, "loss": 3.1175, "step": 7123 }, { "epoch": 1.36, "learning_rate": 2.5953860819828407e-06, "loss": 2.8301, "step": 7124 }, { "epoch": 1.36, "learning_rate": 2.5953288846520495e-06, "loss": 3.3657, "step": 7125 }, { "epoch": 1.36, "learning_rate": 2.5952716873212587e-06, "loss": 3.0131, "step": 7126 }, { "epoch": 1.36, "learning_rate": 2.595214489990467e-06, "loss": 2.8222, "step": 7127 }, { "epoch": 1.36, "learning_rate": 2.595157292659676e-06, "loss": 3.446, "step": 7128 }, { "epoch": 1.36, "learning_rate": 2.5951000953288846e-06, "loss": 3.5518, "step": 7129 }, { "epoch": 1.36, "learning_rate": 2.5950428979980934e-06, "loss": 3.2448, "step": 7130 }, { "epoch": 1.36, "learning_rate": 2.5949857006673022e-06, "loss": 2.6927, "step": 7131 }, { "epoch": 1.36, "learning_rate": 2.594928503336511e-06, "loss": 3.139, "step": 7132 }, { "epoch": 1.36, "learning_rate": 2.59487130600572e-06, "loss": 3.3547, "step": 7133 }, { "epoch": 1.36, "learning_rate": 2.5948141086749286e-06, "loss": 2.8014, "step": 7134 }, { "epoch": 1.36, "learning_rate": 2.5947569113441374e-06, "loss": 2.6153, "step": 7135 }, { "epoch": 1.36, "learning_rate": 2.594699714013346e-06, "loss": 3.0606, "step": 7136 }, { "epoch": 1.36, "learning_rate": 2.594642516682555e-06, "loss": 3.3429, "step": 7137 }, { "epoch": 1.36, "learning_rate": 2.594585319351764e-06, "loss": 3.2086, "step": 7138 }, { "epoch": 1.36, "learning_rate": 2.594528122020972e-06, "loss": 3.5391, "step": 7139 }, { "epoch": 1.36, "learning_rate": 2.5944709246901814e-06, "loss": 2.4587, "step": 7140 }, { "epoch": 1.36, "learning_rate": 2.59441372735939e-06, "loss": 3.5362, "step": 7141 }, { "epoch": 1.36, "learning_rate": 2.5943565300285985e-06, "loss": 3.0893, "step": 7142 }, { "epoch": 1.36, "learning_rate": 2.5942993326978078e-06, "loss": 2.6265, "step": 7143 }, { "epoch": 1.36, "learning_rate": 2.594242135367016e-06, "loss": 2.3928, "step": 7144 }, { "epoch": 1.36, "learning_rate": 2.594184938036225e-06, "loss": 2.8817, "step": 7145 }, { "epoch": 1.36, "learning_rate": 2.5941277407054337e-06, "loss": 3.7133, "step": 7146 }, { "epoch": 1.36, "learning_rate": 2.5940705433746425e-06, "loss": 3.8297, "step": 7147 }, { "epoch": 1.36, "learning_rate": 2.5940133460438513e-06, "loss": 3.3885, "step": 7148 }, { "epoch": 1.36, "learning_rate": 2.59395614871306e-06, "loss": 3.664, "step": 7149 }, { "epoch": 1.36, "learning_rate": 2.593898951382269e-06, "loss": 2.2437, "step": 7150 }, { "epoch": 1.36, "learning_rate": 2.5938417540514777e-06, "loss": 2.7572, "step": 7151 }, { "epoch": 1.36, "learning_rate": 2.5937845567206865e-06, "loss": 3.1949, "step": 7152 }, { "epoch": 1.36, "learning_rate": 2.5937273593898953e-06, "loss": 3.3002, "step": 7153 }, { "epoch": 1.36, "learning_rate": 2.593670162059104e-06, "loss": 2.8988, "step": 7154 }, { "epoch": 1.36, "learning_rate": 2.593612964728313e-06, "loss": 3.0382, "step": 7155 }, { "epoch": 1.36, "learning_rate": 2.5935557673975212e-06, "loss": 3.278, "step": 7156 }, { "epoch": 1.36, "learning_rate": 2.5934985700667305e-06, "loss": 2.8848, "step": 7157 }, { "epoch": 1.36, "learning_rate": 2.5934413727359392e-06, "loss": 2.9692, "step": 7158 }, { "epoch": 1.36, "learning_rate": 2.5933841754051476e-06, "loss": 3.1151, "step": 7159 }, { "epoch": 1.36, "learning_rate": 2.593326978074357e-06, "loss": 3.4966, "step": 7160 }, { "epoch": 1.36, "learning_rate": 2.593269780743565e-06, "loss": 3.4209, "step": 7161 }, { "epoch": 1.36, "learning_rate": 2.593212583412774e-06, "loss": 3.1738, "step": 7162 }, { "epoch": 1.36, "learning_rate": 2.593155386081983e-06, "loss": 2.6273, "step": 7163 }, { "epoch": 1.36, "learning_rate": 2.5930981887511916e-06, "loss": 3.3267, "step": 7164 }, { "epoch": 1.36, "learning_rate": 2.5930409914204004e-06, "loss": 2.6934, "step": 7165 }, { "epoch": 1.36, "learning_rate": 2.592983794089609e-06, "loss": 3.286, "step": 7166 }, { "epoch": 1.37, "learning_rate": 2.592926596758818e-06, "loss": 2.5337, "step": 7167 }, { "epoch": 1.37, "learning_rate": 2.5928693994280268e-06, "loss": 2.8351, "step": 7168 }, { "epoch": 1.37, "learning_rate": 2.5928122020972356e-06, "loss": 2.8477, "step": 7169 }, { "epoch": 1.37, "learning_rate": 2.5927550047664443e-06, "loss": 3.4954, "step": 7170 }, { "epoch": 1.37, "learning_rate": 2.592697807435653e-06, "loss": 3.2163, "step": 7171 }, { "epoch": 1.37, "learning_rate": 2.592640610104862e-06, "loss": 3.252, "step": 7172 }, { "epoch": 1.37, "learning_rate": 2.5925834127740707e-06, "loss": 3.6276, "step": 7173 }, { "epoch": 1.37, "learning_rate": 2.5925262154432795e-06, "loss": 2.9796, "step": 7174 }, { "epoch": 1.37, "learning_rate": 2.5924690181124883e-06, "loss": 2.9227, "step": 7175 }, { "epoch": 1.37, "learning_rate": 2.5924118207816967e-06, "loss": 3.575, "step": 7176 }, { "epoch": 1.37, "learning_rate": 2.592354623450906e-06, "loss": 3.163, "step": 7177 }, { "epoch": 1.37, "learning_rate": 2.5922974261201143e-06, "loss": 3.0242, "step": 7178 }, { "epoch": 1.37, "learning_rate": 2.592240228789323e-06, "loss": 2.7819, "step": 7179 }, { "epoch": 1.37, "learning_rate": 2.5921830314585323e-06, "loss": 3.0901, "step": 7180 }, { "epoch": 1.37, "learning_rate": 2.5921258341277407e-06, "loss": 3.4206, "step": 7181 }, { "epoch": 1.37, "learning_rate": 2.5920686367969494e-06, "loss": 3.1386, "step": 7182 }, { "epoch": 1.37, "learning_rate": 2.5920114394661582e-06, "loss": 3.5875, "step": 7183 }, { "epoch": 1.37, "learning_rate": 2.591954242135367e-06, "loss": 3.1472, "step": 7184 }, { "epoch": 1.37, "learning_rate": 2.591897044804576e-06, "loss": 3.0875, "step": 7185 }, { "epoch": 1.37, "learning_rate": 2.5918398474737846e-06, "loss": 3.0442, "step": 7186 }, { "epoch": 1.37, "learning_rate": 2.5917826501429934e-06, "loss": 3.1771, "step": 7187 }, { "epoch": 1.37, "learning_rate": 2.591725452812202e-06, "loss": 3.4574, "step": 7188 }, { "epoch": 1.37, "learning_rate": 2.591668255481411e-06, "loss": 3.7292, "step": 7189 }, { "epoch": 1.37, "learning_rate": 2.59161105815062e-06, "loss": 3.2235, "step": 7190 }, { "epoch": 1.37, "learning_rate": 2.5915538608198286e-06, "loss": 2.8587, "step": 7191 }, { "epoch": 1.37, "learning_rate": 2.5914966634890374e-06, "loss": 3.2356, "step": 7192 }, { "epoch": 1.37, "learning_rate": 2.5914394661582458e-06, "loss": 3.18, "step": 7193 }, { "epoch": 1.37, "learning_rate": 2.591382268827455e-06, "loss": 3.0173, "step": 7194 }, { "epoch": 1.37, "learning_rate": 2.5913250714966633e-06, "loss": 3.6335, "step": 7195 }, { "epoch": 1.37, "learning_rate": 2.591267874165872e-06, "loss": 3.5749, "step": 7196 }, { "epoch": 1.37, "learning_rate": 2.5912106768350814e-06, "loss": 2.7684, "step": 7197 }, { "epoch": 1.37, "learning_rate": 2.5911534795042897e-06, "loss": 2.903, "step": 7198 }, { "epoch": 1.37, "learning_rate": 2.5910962821734985e-06, "loss": 3.2716, "step": 7199 }, { "epoch": 1.37, "learning_rate": 2.5910390848427073e-06, "loss": 3.0086, "step": 7200 }, { "epoch": 1.37, "learning_rate": 2.590981887511916e-06, "loss": 3.1959, "step": 7201 }, { "epoch": 1.37, "learning_rate": 2.590924690181125e-06, "loss": 3.4035, "step": 7202 }, { "epoch": 1.37, "learning_rate": 2.5908674928503337e-06, "loss": 2.92, "step": 7203 }, { "epoch": 1.37, "learning_rate": 2.5908102955195425e-06, "loss": 2.8912, "step": 7204 }, { "epoch": 1.37, "learning_rate": 2.5907530981887513e-06, "loss": 3.2935, "step": 7205 }, { "epoch": 1.37, "learning_rate": 2.59069590085796e-06, "loss": 3.1345, "step": 7206 }, { "epoch": 1.37, "learning_rate": 2.590638703527169e-06, "loss": 3.2172, "step": 7207 }, { "epoch": 1.37, "learning_rate": 2.5905815061963777e-06, "loss": 2.9634, "step": 7208 }, { "epoch": 1.37, "learning_rate": 2.5905243088655865e-06, "loss": 3.1292, "step": 7209 }, { "epoch": 1.37, "learning_rate": 2.590467111534795e-06, "loss": 3.2568, "step": 7210 }, { "epoch": 1.37, "learning_rate": 2.590409914204004e-06, "loss": 3.0811, "step": 7211 }, { "epoch": 1.37, "learning_rate": 2.5903527168732124e-06, "loss": 3.0579, "step": 7212 }, { "epoch": 1.37, "learning_rate": 2.590295519542421e-06, "loss": 2.7182, "step": 7213 }, { "epoch": 1.37, "learning_rate": 2.5902383222116304e-06, "loss": 3.4646, "step": 7214 }, { "epoch": 1.37, "learning_rate": 2.590181124880839e-06, "loss": 3.0565, "step": 7215 }, { "epoch": 1.37, "learning_rate": 2.5901239275500476e-06, "loss": 3.0261, "step": 7216 }, { "epoch": 1.37, "learning_rate": 2.5900667302192564e-06, "loss": 3.0972, "step": 7217 }, { "epoch": 1.37, "learning_rate": 2.590009532888465e-06, "loss": 3.1385, "step": 7218 }, { "epoch": 1.38, "learning_rate": 2.5899523355576744e-06, "loss": 2.7784, "step": 7219 }, { "epoch": 1.38, "learning_rate": 2.5898951382268828e-06, "loss": 3.3278, "step": 7220 }, { "epoch": 1.38, "learning_rate": 2.5898379408960916e-06, "loss": 3.024, "step": 7221 }, { "epoch": 1.38, "learning_rate": 2.5897807435653004e-06, "loss": 3.2648, "step": 7222 }, { "epoch": 1.38, "learning_rate": 2.589723546234509e-06, "loss": 2.9586, "step": 7223 }, { "epoch": 1.38, "learning_rate": 2.589666348903718e-06, "loss": 3.239, "step": 7224 }, { "epoch": 1.38, "learning_rate": 2.5896091515729267e-06, "loss": 3.1116, "step": 7225 }, { "epoch": 1.38, "learning_rate": 2.5895519542421355e-06, "loss": 3.3756, "step": 7226 }, { "epoch": 1.38, "learning_rate": 2.589494756911344e-06, "loss": 3.6327, "step": 7227 }, { "epoch": 1.38, "learning_rate": 2.589437559580553e-06, "loss": 3.2688, "step": 7228 }, { "epoch": 1.38, "learning_rate": 2.589380362249762e-06, "loss": 3.6617, "step": 7229 }, { "epoch": 1.38, "learning_rate": 2.5893231649189703e-06, "loss": 2.7728, "step": 7230 }, { "epoch": 1.38, "learning_rate": 2.5892659675881795e-06, "loss": 2.9953, "step": 7231 }, { "epoch": 1.38, "learning_rate": 2.589208770257388e-06, "loss": 3.1298, "step": 7232 }, { "epoch": 1.38, "learning_rate": 2.589151572926597e-06, "loss": 3.2199, "step": 7233 }, { "epoch": 1.38, "learning_rate": 2.5890943755958055e-06, "loss": 3.8845, "step": 7234 }, { "epoch": 1.38, "learning_rate": 2.5890371782650143e-06, "loss": 2.7491, "step": 7235 }, { "epoch": 1.38, "learning_rate": 2.5889799809342235e-06, "loss": 3.6725, "step": 7236 }, { "epoch": 1.38, "learning_rate": 2.588922783603432e-06, "loss": 3.3297, "step": 7237 }, { "epoch": 1.38, "learning_rate": 2.5888655862726406e-06, "loss": 3.1883, "step": 7238 }, { "epoch": 1.38, "learning_rate": 2.5888083889418494e-06, "loss": 3.2485, "step": 7239 }, { "epoch": 1.38, "learning_rate": 2.5887511916110582e-06, "loss": 2.8362, "step": 7240 }, { "epoch": 1.38, "learning_rate": 2.588693994280267e-06, "loss": 3.2239, "step": 7241 }, { "epoch": 1.38, "learning_rate": 2.588636796949476e-06, "loss": 3.4303, "step": 7242 }, { "epoch": 1.38, "learning_rate": 2.5885795996186846e-06, "loss": 3.1584, "step": 7243 }, { "epoch": 1.38, "learning_rate": 2.588522402287893e-06, "loss": 3.8175, "step": 7244 }, { "epoch": 1.38, "learning_rate": 2.588465204957102e-06, "loss": 3.8747, "step": 7245 }, { "epoch": 1.38, "learning_rate": 2.588408007626311e-06, "loss": 3.7323, "step": 7246 }, { "epoch": 1.38, "learning_rate": 2.5883508102955194e-06, "loss": 3.3047, "step": 7247 }, { "epoch": 1.38, "learning_rate": 2.5882936129647286e-06, "loss": 3.1653, "step": 7248 }, { "epoch": 1.38, "learning_rate": 2.588236415633937e-06, "loss": 2.9722, "step": 7249 }, { "epoch": 1.38, "learning_rate": 2.588179218303146e-06, "loss": 3.6161, "step": 7250 }, { "epoch": 1.38, "learning_rate": 2.5881220209723545e-06, "loss": 2.6851, "step": 7251 }, { "epoch": 1.38, "learning_rate": 2.5880648236415633e-06, "loss": 3.2792, "step": 7252 }, { "epoch": 1.38, "learning_rate": 2.5880076263107725e-06, "loss": 3.5096, "step": 7253 }, { "epoch": 1.38, "learning_rate": 2.587950428979981e-06, "loss": 3.6952, "step": 7254 }, { "epoch": 1.38, "learning_rate": 2.5878932316491897e-06, "loss": 3.6159, "step": 7255 }, { "epoch": 1.38, "learning_rate": 2.5878360343183985e-06, "loss": 3.4266, "step": 7256 }, { "epoch": 1.38, "learning_rate": 2.5877788369876073e-06, "loss": 2.965, "step": 7257 }, { "epoch": 1.38, "learning_rate": 2.587721639656816e-06, "loss": 3.5085, "step": 7258 }, { "epoch": 1.38, "learning_rate": 2.587664442326025e-06, "loss": 3.4236, "step": 7259 }, { "epoch": 1.38, "learning_rate": 2.5876072449952337e-06, "loss": 3.4613, "step": 7260 }, { "epoch": 1.38, "learning_rate": 2.587550047664442e-06, "loss": 2.9198, "step": 7261 }, { "epoch": 1.38, "learning_rate": 2.5874928503336513e-06, "loss": 3.5363, "step": 7262 }, { "epoch": 1.38, "learning_rate": 2.58743565300286e-06, "loss": 3.5589, "step": 7263 }, { "epoch": 1.38, "learning_rate": 2.5873784556720684e-06, "loss": 2.8595, "step": 7264 }, { "epoch": 1.38, "learning_rate": 2.5873212583412776e-06, "loss": 2.7204, "step": 7265 }, { "epoch": 1.38, "learning_rate": 2.587264061010486e-06, "loss": 3.2489, "step": 7266 }, { "epoch": 1.38, "learning_rate": 2.5872068636796952e-06, "loss": 3.2316, "step": 7267 }, { "epoch": 1.38, "learning_rate": 2.587149666348904e-06, "loss": 3.5293, "step": 7268 }, { "epoch": 1.38, "learning_rate": 2.5870924690181124e-06, "loss": 3.159, "step": 7269 }, { "epoch": 1.38, "learning_rate": 2.5870352716873216e-06, "loss": 3.5527, "step": 7270 }, { "epoch": 1.38, "learning_rate": 2.58697807435653e-06, "loss": 2.9073, "step": 7271 }, { "epoch": 1.39, "learning_rate": 2.5869208770257388e-06, "loss": 3.023, "step": 7272 }, { "epoch": 1.39, "learning_rate": 2.5868636796949476e-06, "loss": 3.09, "step": 7273 }, { "epoch": 1.39, "learning_rate": 2.5868064823641564e-06, "loss": 2.8714, "step": 7274 }, { "epoch": 1.39, "learning_rate": 2.586749285033365e-06, "loss": 4.0123, "step": 7275 }, { "epoch": 1.39, "learning_rate": 2.586692087702574e-06, "loss": 2.6017, "step": 7276 }, { "epoch": 1.39, "learning_rate": 2.5866348903717827e-06, "loss": 3.5216, "step": 7277 }, { "epoch": 1.39, "learning_rate": 2.5865776930409915e-06, "loss": 2.6987, "step": 7278 }, { "epoch": 1.39, "learning_rate": 2.5865204957102003e-06, "loss": 2.957, "step": 7279 }, { "epoch": 1.39, "learning_rate": 2.586463298379409e-06, "loss": 3.507, "step": 7280 }, { "epoch": 1.39, "learning_rate": 2.5864061010486175e-06, "loss": 2.8364, "step": 7281 }, { "epoch": 1.39, "learning_rate": 2.5863489037178267e-06, "loss": 2.8414, "step": 7282 }, { "epoch": 1.39, "learning_rate": 2.586291706387035e-06, "loss": 3.3776, "step": 7283 }, { "epoch": 1.39, "learning_rate": 2.5862345090562443e-06, "loss": 3.1201, "step": 7284 }, { "epoch": 1.39, "learning_rate": 2.586177311725453e-06, "loss": 2.9617, "step": 7285 }, { "epoch": 1.39, "learning_rate": 2.5861201143946615e-06, "loss": 2.9345, "step": 7286 }, { "epoch": 1.39, "learning_rate": 2.5860629170638707e-06, "loss": 3.6053, "step": 7287 }, { "epoch": 1.39, "learning_rate": 2.586005719733079e-06, "loss": 3.185, "step": 7288 }, { "epoch": 1.39, "learning_rate": 2.585948522402288e-06, "loss": 3.0395, "step": 7289 }, { "epoch": 1.39, "learning_rate": 2.5858913250714966e-06, "loss": 3.1591, "step": 7290 }, { "epoch": 1.39, "learning_rate": 2.5858341277407054e-06, "loss": 3.0692, "step": 7291 }, { "epoch": 1.39, "learning_rate": 2.5857769304099142e-06, "loss": 3.4537, "step": 7292 }, { "epoch": 1.39, "learning_rate": 2.585719733079123e-06, "loss": 3.4646, "step": 7293 }, { "epoch": 1.39, "learning_rate": 2.585662535748332e-06, "loss": 2.2271, "step": 7294 }, { "epoch": 1.39, "learning_rate": 2.5856053384175406e-06, "loss": 3.8674, "step": 7295 }, { "epoch": 1.39, "learning_rate": 2.5855481410867494e-06, "loss": 2.7402, "step": 7296 }, { "epoch": 1.39, "learning_rate": 2.585490943755958e-06, "loss": 2.8405, "step": 7297 }, { "epoch": 1.39, "learning_rate": 2.585433746425167e-06, "loss": 3.2104, "step": 7298 }, { "epoch": 1.39, "learning_rate": 2.5853765490943758e-06, "loss": 3.6791, "step": 7299 }, { "epoch": 1.39, "learning_rate": 2.585319351763584e-06, "loss": 2.9774, "step": 7300 }, { "epoch": 1.39, "learning_rate": 2.5852621544327934e-06, "loss": 3.3206, "step": 7301 }, { "epoch": 1.39, "learning_rate": 2.585204957102002e-06, "loss": 3.7124, "step": 7302 }, { "epoch": 1.39, "learning_rate": 2.5851477597712105e-06, "loss": 3.0519, "step": 7303 }, { "epoch": 1.39, "learning_rate": 2.5850905624404198e-06, "loss": 3.5292, "step": 7304 }, { "epoch": 1.39, "learning_rate": 2.585033365109628e-06, "loss": 2.9324, "step": 7305 }, { "epoch": 1.39, "learning_rate": 2.584976167778837e-06, "loss": 3.1354, "step": 7306 }, { "epoch": 1.39, "learning_rate": 2.584918970448046e-06, "loss": 2.8976, "step": 7307 }, { "epoch": 1.39, "learning_rate": 2.5848617731172545e-06, "loss": 2.4485, "step": 7308 }, { "epoch": 1.39, "learning_rate": 2.5848045757864633e-06, "loss": 3.5119, "step": 7309 }, { "epoch": 1.39, "learning_rate": 2.584747378455672e-06, "loss": 3.4816, "step": 7310 }, { "epoch": 1.39, "learning_rate": 2.584690181124881e-06, "loss": 3.0964, "step": 7311 }, { "epoch": 1.39, "learning_rate": 2.5846329837940897e-06, "loss": 2.8619, "step": 7312 }, { "epoch": 1.39, "learning_rate": 2.5845757864632985e-06, "loss": 3.5481, "step": 7313 }, { "epoch": 1.39, "learning_rate": 2.5845185891325073e-06, "loss": 3.0597, "step": 7314 }, { "epoch": 1.39, "learning_rate": 2.584461391801716e-06, "loss": 2.7512, "step": 7315 }, { "epoch": 1.39, "learning_rate": 2.584404194470925e-06, "loss": 3.1478, "step": 7316 }, { "epoch": 1.39, "learning_rate": 2.5843469971401336e-06, "loss": 2.883, "step": 7317 }, { "epoch": 1.39, "learning_rate": 2.5842897998093424e-06, "loss": 3.503, "step": 7318 }, { "epoch": 1.39, "learning_rate": 2.5842326024785512e-06, "loss": 2.9722, "step": 7319 }, { "epoch": 1.39, "learning_rate": 2.5841754051477596e-06, "loss": 3.4298, "step": 7320 }, { "epoch": 1.39, "learning_rate": 2.584118207816969e-06, "loss": 3.0912, "step": 7321 }, { "epoch": 1.39, "learning_rate": 2.584061010486177e-06, "loss": 3.3458, "step": 7322 }, { "epoch": 1.39, "learning_rate": 2.584003813155386e-06, "loss": 3.455, "step": 7323 }, { "epoch": 1.4, "learning_rate": 2.583946615824595e-06, "loss": 2.4693, "step": 7324 }, { "epoch": 1.4, "learning_rate": 2.5838894184938036e-06, "loss": 3.2269, "step": 7325 }, { "epoch": 1.4, "learning_rate": 2.5838322211630124e-06, "loss": 3.1862, "step": 7326 }, { "epoch": 1.4, "learning_rate": 2.583775023832221e-06, "loss": 3.4101, "step": 7327 }, { "epoch": 1.4, "learning_rate": 2.58371782650143e-06, "loss": 2.6558, "step": 7328 }, { "epoch": 1.4, "learning_rate": 2.5836606291706388e-06, "loss": 3.1039, "step": 7329 }, { "epoch": 1.4, "learning_rate": 2.5836034318398475e-06, "loss": 2.332, "step": 7330 }, { "epoch": 1.4, "learning_rate": 2.5835462345090563e-06, "loss": 2.8779, "step": 7331 }, { "epoch": 1.4, "learning_rate": 2.583489037178265e-06, "loss": 3.0337, "step": 7332 }, { "epoch": 1.4, "learning_rate": 2.583431839847474e-06, "loss": 2.4614, "step": 7333 }, { "epoch": 1.4, "learning_rate": 2.5833746425166827e-06, "loss": 2.8351, "step": 7334 }, { "epoch": 1.4, "learning_rate": 2.5833174451858915e-06, "loss": 2.7963, "step": 7335 }, { "epoch": 1.4, "learning_rate": 2.5832602478551003e-06, "loss": 3.0726, "step": 7336 }, { "epoch": 1.4, "learning_rate": 2.5832030505243087e-06, "loss": 2.7195, "step": 7337 }, { "epoch": 1.4, "learning_rate": 2.583145853193518e-06, "loss": 2.7433, "step": 7338 }, { "epoch": 1.4, "learning_rate": 2.5830886558627263e-06, "loss": 3.4657, "step": 7339 }, { "epoch": 1.4, "learning_rate": 2.583031458531935e-06, "loss": 2.33, "step": 7340 }, { "epoch": 1.4, "learning_rate": 2.5829742612011443e-06, "loss": 3.5018, "step": 7341 }, { "epoch": 1.4, "learning_rate": 2.5829170638703526e-06, "loss": 2.88, "step": 7342 }, { "epoch": 1.4, "learning_rate": 2.5828598665395614e-06, "loss": 2.9719, "step": 7343 }, { "epoch": 1.4, "learning_rate": 2.5828026692087702e-06, "loss": 3.7033, "step": 7344 }, { "epoch": 1.4, "learning_rate": 2.582745471877979e-06, "loss": 3.2451, "step": 7345 }, { "epoch": 1.4, "learning_rate": 2.582688274547188e-06, "loss": 3.4484, "step": 7346 }, { "epoch": 1.4, "learning_rate": 2.5826310772163966e-06, "loss": 3.714, "step": 7347 }, { "epoch": 1.4, "learning_rate": 2.5825738798856054e-06, "loss": 3.359, "step": 7348 }, { "epoch": 1.4, "learning_rate": 2.582516682554814e-06, "loss": 2.634, "step": 7349 }, { "epoch": 1.4, "learning_rate": 2.582459485224023e-06, "loss": 3.025, "step": 7350 }, { "epoch": 1.4, "learning_rate": 2.582402287893232e-06, "loss": 3.3598, "step": 7351 }, { "epoch": 1.4, "learning_rate": 2.5823450905624406e-06, "loss": 2.4903, "step": 7352 }, { "epoch": 1.4, "learning_rate": 2.5822878932316494e-06, "loss": 2.6817, "step": 7353 }, { "epoch": 1.4, "learning_rate": 2.5822306959008578e-06, "loss": 3.1499, "step": 7354 }, { "epoch": 1.4, "learning_rate": 2.582173498570067e-06, "loss": 3.4639, "step": 7355 }, { "epoch": 1.4, "learning_rate": 2.5821163012392758e-06, "loss": 3.2479, "step": 7356 }, { "epoch": 1.4, "learning_rate": 2.582059103908484e-06, "loss": 2.7757, "step": 7357 }, { "epoch": 1.4, "learning_rate": 2.5820019065776933e-06, "loss": 2.7192, "step": 7358 }, { "epoch": 1.4, "learning_rate": 2.5819447092469017e-06, "loss": 3.0105, "step": 7359 }, { "epoch": 1.4, "learning_rate": 2.5818875119161105e-06, "loss": 2.6806, "step": 7360 }, { "epoch": 1.4, "learning_rate": 2.5818303145853193e-06, "loss": 3.7662, "step": 7361 }, { "epoch": 1.4, "learning_rate": 2.581773117254528e-06, "loss": 3.8207, "step": 7362 }, { "epoch": 1.4, "learning_rate": 2.5817159199237373e-06, "loss": 3.2691, "step": 7363 }, { "epoch": 1.4, "learning_rate": 2.5816587225929457e-06, "loss": 3.0477, "step": 7364 }, { "epoch": 1.4, "learning_rate": 2.5816015252621545e-06, "loss": 2.7736, "step": 7365 }, { "epoch": 1.4, "learning_rate": 2.5815443279313633e-06, "loss": 2.7398, "step": 7366 }, { "epoch": 1.4, "learning_rate": 2.581487130600572e-06, "loss": 3.4971, "step": 7367 }, { "epoch": 1.4, "learning_rate": 2.581429933269781e-06, "loss": 3.099, "step": 7368 }, { "epoch": 1.4, "learning_rate": 2.5813727359389897e-06, "loss": 3.0976, "step": 7369 }, { "epoch": 1.4, "learning_rate": 2.5813155386081985e-06, "loss": 3.5906, "step": 7370 }, { "epoch": 1.4, "learning_rate": 2.581258341277407e-06, "loss": 2.6432, "step": 7371 }, { "epoch": 1.4, "learning_rate": 2.581201143946616e-06, "loss": 3.6409, "step": 7372 }, { "epoch": 1.4, "learning_rate": 2.581143946615825e-06, "loss": 3.5149, "step": 7373 }, { "epoch": 1.4, "learning_rate": 2.581086749285033e-06, "loss": 2.5164, "step": 7374 }, { "epoch": 1.4, "learning_rate": 2.5810295519542424e-06, "loss": 3.687, "step": 7375 }, { "epoch": 1.4, "learning_rate": 2.580972354623451e-06, "loss": 2.8348, "step": 7376 }, { "epoch": 1.41, "learning_rate": 2.5809151572926596e-06, "loss": 3.498, "step": 7377 }, { "epoch": 1.41, "learning_rate": 2.5808579599618684e-06, "loss": 2.6399, "step": 7378 }, { "epoch": 1.41, "learning_rate": 2.580800762631077e-06, "loss": 2.489, "step": 7379 }, { "epoch": 1.41, "learning_rate": 2.5807435653002864e-06, "loss": 2.8316, "step": 7380 }, { "epoch": 1.41, "learning_rate": 2.5806863679694948e-06, "loss": 4.0737, "step": 7381 }, { "epoch": 1.41, "learning_rate": 2.5806291706387036e-06, "loss": 2.6964, "step": 7382 }, { "epoch": 1.41, "learning_rate": 2.5805719733079123e-06, "loss": 2.9681, "step": 7383 }, { "epoch": 1.41, "learning_rate": 2.580514775977121e-06, "loss": 3.0327, "step": 7384 }, { "epoch": 1.41, "learning_rate": 2.58045757864633e-06, "loss": 3.3983, "step": 7385 }, { "epoch": 1.41, "learning_rate": 2.5804003813155387e-06, "loss": 3.721, "step": 7386 }, { "epoch": 1.41, "learning_rate": 2.5803431839847475e-06, "loss": 2.8619, "step": 7387 }, { "epoch": 1.41, "learning_rate": 2.580285986653956e-06, "loss": 3.1833, "step": 7388 }, { "epoch": 1.41, "learning_rate": 2.580228789323165e-06, "loss": 3.5436, "step": 7389 }, { "epoch": 1.41, "learning_rate": 2.580171591992374e-06, "loss": 2.9394, "step": 7390 }, { "epoch": 1.41, "learning_rate": 2.5801143946615823e-06, "loss": 3.2729, "step": 7391 }, { "epoch": 1.41, "learning_rate": 2.5800571973307915e-06, "loss": 3.2135, "step": 7392 }, { "epoch": 1.41, "learning_rate": 2.58e-06, "loss": 3.7199, "step": 7393 }, { "epoch": 1.41, "learning_rate": 2.5799428026692087e-06, "loss": 2.9037, "step": 7394 }, { "epoch": 1.41, "learning_rate": 2.579885605338418e-06, "loss": 2.657, "step": 7395 }, { "epoch": 1.41, "learning_rate": 2.5798284080076262e-06, "loss": 2.9183, "step": 7396 }, { "epoch": 1.41, "learning_rate": 2.5797712106768355e-06, "loss": 3.1805, "step": 7397 }, { "epoch": 1.41, "learning_rate": 2.579714013346044e-06, "loss": 2.71, "step": 7398 }, { "epoch": 1.41, "learning_rate": 2.5796568160152526e-06, "loss": 2.5013, "step": 7399 }, { "epoch": 1.41, "learning_rate": 2.5795996186844614e-06, "loss": 3.2757, "step": 7400 }, { "epoch": 1.41, "learning_rate": 2.5795424213536702e-06, "loss": 3.0158, "step": 7401 }, { "epoch": 1.41, "learning_rate": 2.579485224022879e-06, "loss": 2.8799, "step": 7402 }, { "epoch": 1.41, "learning_rate": 2.579428026692088e-06, "loss": 3.6431, "step": 7403 }, { "epoch": 1.41, "learning_rate": 2.5793708293612966e-06, "loss": 2.5995, "step": 7404 }, { "epoch": 1.41, "learning_rate": 2.5793136320305054e-06, "loss": 3.1523, "step": 7405 }, { "epoch": 1.41, "learning_rate": 2.579256434699714e-06, "loss": 3.3672, "step": 7406 }, { "epoch": 1.41, "learning_rate": 2.579199237368923e-06, "loss": 2.9448, "step": 7407 }, { "epoch": 1.41, "learning_rate": 2.5791420400381313e-06, "loss": 2.9821, "step": 7408 }, { "epoch": 1.41, "learning_rate": 2.5790848427073406e-06, "loss": 3.0561, "step": 7409 }, { "epoch": 1.41, "learning_rate": 2.579027645376549e-06, "loss": 2.4893, "step": 7410 }, { "epoch": 1.41, "learning_rate": 2.5789704480457577e-06, "loss": 2.3212, "step": 7411 }, { "epoch": 1.41, "learning_rate": 2.578913250714967e-06, "loss": 2.7334, "step": 7412 }, { "epoch": 1.41, "learning_rate": 2.5788560533841753e-06, "loss": 2.3584, "step": 7413 }, { "epoch": 1.41, "learning_rate": 2.5787988560533845e-06, "loss": 2.6458, "step": 7414 }, { "epoch": 1.41, "learning_rate": 2.578741658722593e-06, "loss": 3.046, "step": 7415 }, { "epoch": 1.41, "learning_rate": 2.5786844613918017e-06, "loss": 3.0717, "step": 7416 }, { "epoch": 1.41, "learning_rate": 2.5786272640610105e-06, "loss": 3.0716, "step": 7417 }, { "epoch": 1.41, "learning_rate": 2.5785700667302193e-06, "loss": 2.8246, "step": 7418 }, { "epoch": 1.41, "learning_rate": 2.578512869399428e-06, "loss": 3.7351, "step": 7419 }, { "epoch": 1.41, "learning_rate": 2.578455672068637e-06, "loss": 3.0485, "step": 7420 }, { "epoch": 1.41, "learning_rate": 2.5783984747378457e-06, "loss": 3.188, "step": 7421 }, { "epoch": 1.41, "learning_rate": 2.5783412774070545e-06, "loss": 3.1403, "step": 7422 }, { "epoch": 1.41, "learning_rate": 2.5782840800762633e-06, "loss": 3.3981, "step": 7423 }, { "epoch": 1.41, "learning_rate": 2.578226882745472e-06, "loss": 2.6434, "step": 7424 }, { "epoch": 1.41, "learning_rate": 2.5781696854146804e-06, "loss": 3.5132, "step": 7425 }, { "epoch": 1.41, "learning_rate": 2.5781124880838896e-06, "loss": 2.9734, "step": 7426 }, { "epoch": 1.41, "learning_rate": 2.578055290753098e-06, "loss": 3.546, "step": 7427 }, { "epoch": 1.41, "learning_rate": 2.5779980934223072e-06, "loss": 2.9884, "step": 7428 }, { "epoch": 1.42, "learning_rate": 2.577940896091516e-06, "loss": 3.1473, "step": 7429 }, { "epoch": 1.42, "learning_rate": 2.5778836987607244e-06, "loss": 3.3441, "step": 7430 }, { "epoch": 1.42, "learning_rate": 2.5778265014299336e-06, "loss": 3.3711, "step": 7431 }, { "epoch": 1.42, "learning_rate": 2.577769304099142e-06, "loss": 3.7319, "step": 7432 }, { "epoch": 1.42, "learning_rate": 2.5777121067683508e-06, "loss": 2.9074, "step": 7433 }, { "epoch": 1.42, "learning_rate": 2.5776549094375596e-06, "loss": 3.2243, "step": 7434 }, { "epoch": 1.42, "learning_rate": 2.5775977121067684e-06, "loss": 2.4584, "step": 7435 }, { "epoch": 1.42, "learning_rate": 2.577540514775977e-06, "loss": 3.5965, "step": 7436 }, { "epoch": 1.42, "learning_rate": 2.577483317445186e-06, "loss": 2.704, "step": 7437 }, { "epoch": 1.42, "learning_rate": 2.5774261201143947e-06, "loss": 2.7994, "step": 7438 }, { "epoch": 1.42, "learning_rate": 2.5773689227836035e-06, "loss": 2.6687, "step": 7439 }, { "epoch": 1.42, "learning_rate": 2.5773117254528123e-06, "loss": 3.5613, "step": 7440 }, { "epoch": 1.42, "learning_rate": 2.577254528122021e-06, "loss": 3.3039, "step": 7441 }, { "epoch": 1.42, "learning_rate": 2.5771973307912295e-06, "loss": 3.2119, "step": 7442 }, { "epoch": 1.42, "learning_rate": 2.5771401334604387e-06, "loss": 2.4333, "step": 7443 }, { "epoch": 1.42, "learning_rate": 2.5770829361296475e-06, "loss": 3.5784, "step": 7444 }, { "epoch": 1.42, "learning_rate": 2.5770257387988563e-06, "loss": 3.5891, "step": 7445 }, { "epoch": 1.42, "learning_rate": 2.576968541468065e-06, "loss": 3.0929, "step": 7446 }, { "epoch": 1.42, "learning_rate": 2.5769113441372735e-06, "loss": 3.2633, "step": 7447 }, { "epoch": 1.42, "learning_rate": 2.5768541468064827e-06, "loss": 3.64, "step": 7448 }, { "epoch": 1.42, "learning_rate": 2.576796949475691e-06, "loss": 3.628, "step": 7449 }, { "epoch": 1.42, "learning_rate": 2.5767397521449e-06, "loss": 3.1755, "step": 7450 }, { "epoch": 1.42, "learning_rate": 2.576682554814109e-06, "loss": 2.856, "step": 7451 }, { "epoch": 1.42, "learning_rate": 2.5766253574833174e-06, "loss": 3.199, "step": 7452 }, { "epoch": 1.42, "learning_rate": 2.5765681601525262e-06, "loss": 2.856, "step": 7453 }, { "epoch": 1.42, "learning_rate": 2.576510962821735e-06, "loss": 3.205, "step": 7454 }, { "epoch": 1.42, "learning_rate": 2.576453765490944e-06, "loss": 2.9228, "step": 7455 }, { "epoch": 1.42, "learning_rate": 2.5763965681601526e-06, "loss": 2.5174, "step": 7456 }, { "epoch": 1.42, "learning_rate": 2.5763393708293614e-06, "loss": 3.1619, "step": 7457 }, { "epoch": 1.42, "learning_rate": 2.57628217349857e-06, "loss": 3.3563, "step": 7458 }, { "epoch": 1.42, "learning_rate": 2.5762249761677786e-06, "loss": 2.6196, "step": 7459 }, { "epoch": 1.42, "learning_rate": 2.5761677788369878e-06, "loss": 2.5191, "step": 7460 }, { "epoch": 1.42, "learning_rate": 2.5761105815061966e-06, "loss": 3.4203, "step": 7461 }, { "epoch": 1.42, "learning_rate": 2.5760533841754054e-06, "loss": 2.8512, "step": 7462 }, { "epoch": 1.42, "learning_rate": 2.575996186844614e-06, "loss": 3.4109, "step": 7463 }, { "epoch": 1.42, "learning_rate": 2.5759389895138225e-06, "loss": 3.4384, "step": 7464 }, { "epoch": 1.42, "learning_rate": 2.5758817921830317e-06, "loss": 3.49, "step": 7465 }, { "epoch": 1.42, "learning_rate": 2.57582459485224e-06, "loss": 2.4179, "step": 7466 }, { "epoch": 1.42, "learning_rate": 2.575767397521449e-06, "loss": 3.3025, "step": 7467 }, { "epoch": 1.42, "learning_rate": 2.575710200190658e-06, "loss": 2.3424, "step": 7468 }, { "epoch": 1.42, "learning_rate": 2.5756530028598665e-06, "loss": 3.6248, "step": 7469 }, { "epoch": 1.42, "learning_rate": 2.5755958055290753e-06, "loss": 3.6374, "step": 7470 }, { "epoch": 1.42, "learning_rate": 2.575538608198284e-06, "loss": 3.4132, "step": 7471 }, { "epoch": 1.42, "learning_rate": 2.575481410867493e-06, "loss": 3.1239, "step": 7472 }, { "epoch": 1.42, "learning_rate": 2.5754242135367017e-06, "loss": 2.925, "step": 7473 }, { "epoch": 1.42, "learning_rate": 2.5753670162059105e-06, "loss": 3.3346, "step": 7474 }, { "epoch": 1.42, "learning_rate": 2.5753098188751193e-06, "loss": 3.2924, "step": 7475 }, { "epoch": 1.42, "learning_rate": 2.5752526215443276e-06, "loss": 3.0116, "step": 7476 }, { "epoch": 1.42, "learning_rate": 2.575195424213537e-06, "loss": 3.3432, "step": 7477 }, { "epoch": 1.42, "learning_rate": 2.5751382268827456e-06, "loss": 3.0271, "step": 7478 }, { "epoch": 1.42, "learning_rate": 2.5750810295519544e-06, "loss": 2.8579, "step": 7479 }, { "epoch": 1.42, "learning_rate": 2.5750238322211632e-06, "loss": 2.1926, "step": 7480 }, { "epoch": 1.42, "learning_rate": 2.5749666348903716e-06, "loss": 2.8079, "step": 7481 }, { "epoch": 1.43, "learning_rate": 2.574909437559581e-06, "loss": 3.3569, "step": 7482 }, { "epoch": 1.43, "learning_rate": 2.5748522402287896e-06, "loss": 2.9565, "step": 7483 }, { "epoch": 1.43, "learning_rate": 2.574795042897998e-06, "loss": 4.0395, "step": 7484 }, { "epoch": 1.43, "learning_rate": 2.574737845567207e-06, "loss": 3.4723, "step": 7485 }, { "epoch": 1.43, "learning_rate": 2.5746806482364156e-06, "loss": 3.5686, "step": 7486 }, { "epoch": 1.43, "learning_rate": 2.5746234509056244e-06, "loss": 2.8443, "step": 7487 }, { "epoch": 1.43, "learning_rate": 2.574566253574833e-06, "loss": 2.602, "step": 7488 }, { "epoch": 1.43, "learning_rate": 2.574509056244042e-06, "loss": 2.9544, "step": 7489 }, { "epoch": 1.43, "learning_rate": 2.5744518589132507e-06, "loss": 2.8277, "step": 7490 }, { "epoch": 1.43, "learning_rate": 2.5743946615824595e-06, "loss": 2.4724, "step": 7491 }, { "epoch": 1.43, "learning_rate": 2.5743374642516683e-06, "loss": 2.9276, "step": 7492 }, { "epoch": 1.43, "learning_rate": 2.574280266920877e-06, "loss": 2.9367, "step": 7493 }, { "epoch": 1.43, "learning_rate": 2.574223069590086e-06, "loss": 3.2364, "step": 7494 }, { "epoch": 1.43, "learning_rate": 2.5741658722592947e-06, "loss": 2.9758, "step": 7495 }, { "epoch": 1.43, "learning_rate": 2.5741086749285035e-06, "loss": 2.9138, "step": 7496 }, { "epoch": 1.43, "learning_rate": 2.5740514775977123e-06, "loss": 3.0123, "step": 7497 }, { "epoch": 1.43, "learning_rate": 2.5739942802669207e-06, "loss": 3.2572, "step": 7498 }, { "epoch": 1.43, "learning_rate": 2.57393708293613e-06, "loss": 3.224, "step": 7499 }, { "epoch": 1.43, "learning_rate": 2.5738798856053387e-06, "loss": 2.395, "step": 7500 }, { "epoch": 1.43, "learning_rate": 2.573822688274547e-06, "loss": 3.3833, "step": 7501 }, { "epoch": 1.43, "learning_rate": 2.5737654909437563e-06, "loss": 3.2235, "step": 7502 }, { "epoch": 1.43, "learning_rate": 2.5737082936129646e-06, "loss": 3.6835, "step": 7503 }, { "epoch": 1.43, "learning_rate": 2.5736510962821734e-06, "loss": 3.6294, "step": 7504 }, { "epoch": 1.43, "learning_rate": 2.5735938989513822e-06, "loss": 3.1113, "step": 7505 }, { "epoch": 1.43, "learning_rate": 2.573536701620591e-06, "loss": 3.4795, "step": 7506 }, { "epoch": 1.43, "learning_rate": 2.5734795042898e-06, "loss": 3.2493, "step": 7507 }, { "epoch": 1.43, "learning_rate": 2.5734223069590086e-06, "loss": 3.4708, "step": 7508 }, { "epoch": 1.43, "learning_rate": 2.5733651096282174e-06, "loss": 3.1843, "step": 7509 }, { "epoch": 1.43, "learning_rate": 2.573307912297426e-06, "loss": 3.1581, "step": 7510 }, { "epoch": 1.43, "learning_rate": 2.573250714966635e-06, "loss": 2.8702, "step": 7511 }, { "epoch": 1.43, "learning_rate": 2.5731935176358438e-06, "loss": 2.7924, "step": 7512 }, { "epoch": 1.43, "learning_rate": 2.5731363203050526e-06, "loss": 2.1688, "step": 7513 }, { "epoch": 1.43, "learning_rate": 2.5730791229742614e-06, "loss": 3.2157, "step": 7514 }, { "epoch": 1.43, "learning_rate": 2.5730219256434697e-06, "loss": 3.2299, "step": 7515 }, { "epoch": 1.43, "learning_rate": 2.572964728312679e-06, "loss": 2.4286, "step": 7516 }, { "epoch": 1.43, "learning_rate": 2.5729075309818878e-06, "loss": 3.1482, "step": 7517 }, { "epoch": 1.43, "learning_rate": 2.572850333651096e-06, "loss": 2.9492, "step": 7518 }, { "epoch": 1.43, "learning_rate": 2.5727931363203053e-06, "loss": 2.835, "step": 7519 }, { "epoch": 1.43, "learning_rate": 2.5727359389895137e-06, "loss": 3.8084, "step": 7520 }, { "epoch": 1.43, "learning_rate": 2.5726787416587225e-06, "loss": 2.4089, "step": 7521 }, { "epoch": 1.43, "learning_rate": 2.5726215443279313e-06, "loss": 2.7282, "step": 7522 }, { "epoch": 1.43, "learning_rate": 2.57256434699714e-06, "loss": 3.0663, "step": 7523 }, { "epoch": 1.43, "learning_rate": 2.572507149666349e-06, "loss": 3.3627, "step": 7524 }, { "epoch": 1.43, "learning_rate": 2.5724499523355577e-06, "loss": 2.5698, "step": 7525 }, { "epoch": 1.43, "learning_rate": 2.5723927550047665e-06, "loss": 2.6706, "step": 7526 }, { "epoch": 1.43, "learning_rate": 2.5723355576739753e-06, "loss": 2.9455, "step": 7527 }, { "epoch": 1.43, "learning_rate": 2.572278360343184e-06, "loss": 3.2625, "step": 7528 }, { "epoch": 1.43, "learning_rate": 2.572221163012393e-06, "loss": 2.9309, "step": 7529 }, { "epoch": 1.43, "learning_rate": 2.5721639656816017e-06, "loss": 2.9264, "step": 7530 }, { "epoch": 1.43, "learning_rate": 2.5721067683508104e-06, "loss": 3.0948, "step": 7531 }, { "epoch": 1.43, "learning_rate": 2.5720495710200192e-06, "loss": 2.7991, "step": 7532 }, { "epoch": 1.43, "learning_rate": 2.571992373689228e-06, "loss": 2.8051, "step": 7533 }, { "epoch": 1.44, "learning_rate": 2.571935176358437e-06, "loss": 3.7767, "step": 7534 }, { "epoch": 1.44, "learning_rate": 2.571877979027645e-06, "loss": 2.1962, "step": 7535 }, { "epoch": 1.44, "learning_rate": 2.5718207816968544e-06, "loss": 3.244, "step": 7536 }, { "epoch": 1.44, "learning_rate": 2.5717635843660628e-06, "loss": 2.9555, "step": 7537 }, { "epoch": 1.44, "learning_rate": 2.5717063870352716e-06, "loss": 3.5459, "step": 7538 }, { "epoch": 1.44, "learning_rate": 2.571649189704481e-06, "loss": 3.3714, "step": 7539 }, { "epoch": 1.44, "learning_rate": 2.571591992373689e-06, "loss": 2.9466, "step": 7540 }, { "epoch": 1.44, "learning_rate": 2.571534795042898e-06, "loss": 3.1983, "step": 7541 }, { "epoch": 1.44, "learning_rate": 2.5714775977121068e-06, "loss": 3.4973, "step": 7542 }, { "epoch": 1.44, "learning_rate": 2.5714204003813155e-06, "loss": 3.678, "step": 7543 }, { "epoch": 1.44, "learning_rate": 2.5713632030505243e-06, "loss": 2.7919, "step": 7544 }, { "epoch": 1.44, "learning_rate": 2.571306005719733e-06, "loss": 2.7117, "step": 7545 }, { "epoch": 1.44, "learning_rate": 2.571248808388942e-06, "loss": 3.2272, "step": 7546 }, { "epoch": 1.44, "learning_rate": 2.5711916110581507e-06, "loss": 3.3198, "step": 7547 }, { "epoch": 1.44, "learning_rate": 2.5711344137273595e-06, "loss": 3.6764, "step": 7548 }, { "epoch": 1.44, "learning_rate": 2.5710772163965683e-06, "loss": 2.8401, "step": 7549 }, { "epoch": 1.44, "learning_rate": 2.571020019065777e-06, "loss": 3.1494, "step": 7550 }, { "epoch": 1.44, "learning_rate": 2.570962821734986e-06, "loss": 2.7577, "step": 7551 }, { "epoch": 1.44, "learning_rate": 2.5709056244041943e-06, "loss": 3.5922, "step": 7552 }, { "epoch": 1.44, "learning_rate": 2.5708484270734035e-06, "loss": 3.2306, "step": 7553 }, { "epoch": 1.44, "learning_rate": 2.570791229742612e-06, "loss": 2.5566, "step": 7554 }, { "epoch": 1.44, "learning_rate": 2.5707340324118206e-06, "loss": 2.7979, "step": 7555 }, { "epoch": 1.44, "learning_rate": 2.57067683508103e-06, "loss": 3.5806, "step": 7556 }, { "epoch": 1.44, "learning_rate": 2.5706196377502382e-06, "loss": 3.2879, "step": 7557 }, { "epoch": 1.44, "learning_rate": 2.5705624404194475e-06, "loss": 2.9301, "step": 7558 }, { "epoch": 1.44, "learning_rate": 2.570505243088656e-06, "loss": 2.7678, "step": 7559 }, { "epoch": 1.44, "learning_rate": 2.5704480457578646e-06, "loss": 2.6883, "step": 7560 }, { "epoch": 1.44, "learning_rate": 2.5703908484270734e-06, "loss": 2.9203, "step": 7561 }, { "epoch": 1.44, "learning_rate": 2.570333651096282e-06, "loss": 3.6308, "step": 7562 }, { "epoch": 1.44, "learning_rate": 2.570276453765491e-06, "loss": 3.0202, "step": 7563 }, { "epoch": 1.44, "learning_rate": 2.5702192564347e-06, "loss": 3.3496, "step": 7564 }, { "epoch": 1.44, "learning_rate": 2.5701620591039086e-06, "loss": 3.3831, "step": 7565 }, { "epoch": 1.44, "learning_rate": 2.5701048617731174e-06, "loss": 3.4612, "step": 7566 }, { "epoch": 1.44, "learning_rate": 2.570047664442326e-06, "loss": 3.5084, "step": 7567 }, { "epoch": 1.44, "learning_rate": 2.569990467111535e-06, "loss": 2.9862, "step": 7568 }, { "epoch": 1.44, "learning_rate": 2.5699332697807433e-06, "loss": 2.6415, "step": 7569 }, { "epoch": 1.44, "learning_rate": 2.5698760724499526e-06, "loss": 3.6664, "step": 7570 }, { "epoch": 1.44, "learning_rate": 2.569818875119161e-06, "loss": 2.3453, "step": 7571 }, { "epoch": 1.44, "learning_rate": 2.5697616777883697e-06, "loss": 3.3785, "step": 7572 }, { "epoch": 1.44, "learning_rate": 2.569704480457579e-06, "loss": 2.9049, "step": 7573 }, { "epoch": 1.44, "learning_rate": 2.5696472831267873e-06, "loss": 3.2109, "step": 7574 }, { "epoch": 1.44, "learning_rate": 2.5695900857959965e-06, "loss": 3.1238, "step": 7575 }, { "epoch": 1.44, "learning_rate": 2.569532888465205e-06, "loss": 3.3038, "step": 7576 }, { "epoch": 1.44, "learning_rate": 2.5694756911344137e-06, "loss": 3.6597, "step": 7577 }, { "epoch": 1.44, "learning_rate": 2.569418493803623e-06, "loss": 2.7508, "step": 7578 }, { "epoch": 1.44, "learning_rate": 2.5693612964728313e-06, "loss": 3.019, "step": 7579 }, { "epoch": 1.44, "learning_rate": 2.56930409914204e-06, "loss": 3.2652, "step": 7580 }, { "epoch": 1.44, "learning_rate": 2.569246901811249e-06, "loss": 3.4278, "step": 7581 }, { "epoch": 1.44, "learning_rate": 2.5691897044804577e-06, "loss": 3.0453, "step": 7582 }, { "epoch": 1.44, "learning_rate": 2.5691325071496665e-06, "loss": 2.9572, "step": 7583 }, { "epoch": 1.44, "learning_rate": 2.5690753098188752e-06, "loss": 3.1901, "step": 7584 }, { "epoch": 1.44, "learning_rate": 2.569018112488084e-06, "loss": 3.2254, "step": 7585 }, { "epoch": 1.44, "learning_rate": 2.5689609151572924e-06, "loss": 3.0508, "step": 7586 }, { "epoch": 1.45, "learning_rate": 2.5689037178265016e-06, "loss": 2.9004, "step": 7587 }, { "epoch": 1.45, "learning_rate": 2.5688465204957104e-06, "loss": 3.5637, "step": 7588 }, { "epoch": 1.45, "learning_rate": 2.568789323164919e-06, "loss": 3.7483, "step": 7589 }, { "epoch": 1.45, "learning_rate": 2.568732125834128e-06, "loss": 2.9143, "step": 7590 }, { "epoch": 1.45, "learning_rate": 2.5686749285033364e-06, "loss": 3.0457, "step": 7591 }, { "epoch": 1.45, "learning_rate": 2.5686177311725456e-06, "loss": 3.3851, "step": 7592 }, { "epoch": 1.45, "learning_rate": 2.568560533841754e-06, "loss": 3.0406, "step": 7593 }, { "epoch": 1.45, "learning_rate": 2.5685033365109628e-06, "loss": 3.2823, "step": 7594 }, { "epoch": 1.45, "learning_rate": 2.568446139180172e-06, "loss": 3.1517, "step": 7595 }, { "epoch": 1.45, "learning_rate": 2.5683889418493803e-06, "loss": 3.1035, "step": 7596 }, { "epoch": 1.45, "learning_rate": 2.568331744518589e-06, "loss": 2.7547, "step": 7597 }, { "epoch": 1.45, "learning_rate": 2.568274547187798e-06, "loss": 3.1748, "step": 7598 }, { "epoch": 1.45, "learning_rate": 2.5682173498570067e-06, "loss": 3.2595, "step": 7599 }, { "epoch": 1.45, "learning_rate": 2.5681601525262155e-06, "loss": 3.6187, "step": 7600 }, { "epoch": 1.45, "learning_rate": 2.5681029551954243e-06, "loss": 3.2355, "step": 7601 }, { "epoch": 1.45, "learning_rate": 2.568045757864633e-06, "loss": 3.1436, "step": 7602 }, { "epoch": 1.45, "learning_rate": 2.5679885605338415e-06, "loss": 2.7685, "step": 7603 }, { "epoch": 1.45, "learning_rate": 2.5679313632030507e-06, "loss": 3.9572, "step": 7604 }, { "epoch": 1.45, "learning_rate": 2.5678741658722595e-06, "loss": 3.0539, "step": 7605 }, { "epoch": 1.45, "learning_rate": 2.567816968541468e-06, "loss": 3.0517, "step": 7606 }, { "epoch": 1.45, "learning_rate": 2.567759771210677e-06, "loss": 3.0344, "step": 7607 }, { "epoch": 1.45, "learning_rate": 2.5677025738798855e-06, "loss": 3.5971, "step": 7608 }, { "epoch": 1.45, "learning_rate": 2.5676453765490947e-06, "loss": 2.8852, "step": 7609 }, { "epoch": 1.45, "learning_rate": 2.567588179218303e-06, "loss": 2.7883, "step": 7610 }, { "epoch": 1.45, "learning_rate": 2.567530981887512e-06, "loss": 3.44, "step": 7611 }, { "epoch": 1.45, "learning_rate": 2.567473784556721e-06, "loss": 4.1542, "step": 7612 }, { "epoch": 1.45, "learning_rate": 2.5674165872259294e-06, "loss": 2.5773, "step": 7613 }, { "epoch": 1.45, "learning_rate": 2.5673593898951382e-06, "loss": 3.4112, "step": 7614 }, { "epoch": 1.45, "learning_rate": 2.567302192564347e-06, "loss": 2.7196, "step": 7615 }, { "epoch": 1.45, "learning_rate": 2.567244995233556e-06, "loss": 2.9408, "step": 7616 }, { "epoch": 1.45, "learning_rate": 2.5671877979027646e-06, "loss": 3.6802, "step": 7617 }, { "epoch": 1.45, "learning_rate": 2.5671306005719734e-06, "loss": 2.9184, "step": 7618 }, { "epoch": 1.45, "learning_rate": 2.567073403241182e-06, "loss": 2.608, "step": 7619 }, { "epoch": 1.45, "learning_rate": 2.567016205910391e-06, "loss": 3.0614, "step": 7620 }, { "epoch": 1.45, "learning_rate": 2.5669590085795998e-06, "loss": 2.7865, "step": 7621 }, { "epoch": 1.45, "learning_rate": 2.5669018112488086e-06, "loss": 3.7595, "step": 7622 }, { "epoch": 1.45, "learning_rate": 2.566844613918017e-06, "loss": 3.0757, "step": 7623 }, { "epoch": 1.45, "learning_rate": 2.566787416587226e-06, "loss": 2.8658, "step": 7624 }, { "epoch": 1.45, "learning_rate": 2.5667302192564345e-06, "loss": 3.1604, "step": 7625 }, { "epoch": 1.45, "learning_rate": 2.5666730219256437e-06, "loss": 3.238, "step": 7626 }, { "epoch": 1.45, "learning_rate": 2.5666158245948525e-06, "loss": 3.6965, "step": 7627 }, { "epoch": 1.45, "learning_rate": 2.566558627264061e-06, "loss": 3.406, "step": 7628 }, { "epoch": 1.45, "learning_rate": 2.56650142993327e-06, "loss": 2.78, "step": 7629 }, { "epoch": 1.45, "learning_rate": 2.5664442326024785e-06, "loss": 3.4955, "step": 7630 }, { "epoch": 1.45, "learning_rate": 2.5663870352716873e-06, "loss": 3.1389, "step": 7631 }, { "epoch": 1.45, "learning_rate": 2.566329837940896e-06, "loss": 2.8413, "step": 7632 }, { "epoch": 1.45, "learning_rate": 2.566272640610105e-06, "loss": 3.2122, "step": 7633 }, { "epoch": 1.45, "learning_rate": 2.5662154432793137e-06, "loss": 2.8416, "step": 7634 }, { "epoch": 1.45, "learning_rate": 2.5661582459485225e-06, "loss": 3.2982, "step": 7635 }, { "epoch": 1.45, "learning_rate": 2.5661010486177313e-06, "loss": 3.0342, "step": 7636 }, { "epoch": 1.45, "learning_rate": 2.56604385128694e-06, "loss": 3.0418, "step": 7637 }, { "epoch": 1.45, "learning_rate": 2.565986653956149e-06, "loss": 3.0719, "step": 7638 }, { "epoch": 1.46, "learning_rate": 2.5659294566253576e-06, "loss": 3.2407, "step": 7639 }, { "epoch": 1.46, "learning_rate": 2.5658722592945664e-06, "loss": 3.4342, "step": 7640 }, { "epoch": 1.46, "learning_rate": 2.5658150619637752e-06, "loss": 2.9117, "step": 7641 }, { "epoch": 1.46, "learning_rate": 2.5657578646329836e-06, "loss": 3.5076, "step": 7642 }, { "epoch": 1.46, "learning_rate": 2.565700667302193e-06, "loss": 4.117, "step": 7643 }, { "epoch": 1.46, "learning_rate": 2.5656434699714016e-06, "loss": 3.1439, "step": 7644 }, { "epoch": 1.46, "learning_rate": 2.56558627264061e-06, "loss": 3.1596, "step": 7645 }, { "epoch": 1.46, "learning_rate": 2.565529075309819e-06, "loss": 2.8078, "step": 7646 }, { "epoch": 1.46, "learning_rate": 2.5654718779790276e-06, "loss": 3.5296, "step": 7647 }, { "epoch": 1.46, "learning_rate": 2.5654146806482364e-06, "loss": 3.4238, "step": 7648 }, { "epoch": 1.46, "learning_rate": 2.565357483317445e-06, "loss": 3.1906, "step": 7649 }, { "epoch": 1.46, "learning_rate": 2.565300285986654e-06, "loss": 2.5858, "step": 7650 }, { "epoch": 1.46, "learning_rate": 2.5652430886558627e-06, "loss": 3.0057, "step": 7651 }, { "epoch": 1.46, "learning_rate": 2.5651858913250715e-06, "loss": 2.5723, "step": 7652 }, { "epoch": 1.46, "learning_rate": 2.5651286939942803e-06, "loss": 2.6452, "step": 7653 }, { "epoch": 1.46, "learning_rate": 2.565071496663489e-06, "loss": 3.4732, "step": 7654 }, { "epoch": 1.46, "learning_rate": 2.565014299332698e-06, "loss": 3.7653, "step": 7655 }, { "epoch": 1.46, "learning_rate": 2.5649571020019067e-06, "loss": 3.417, "step": 7656 }, { "epoch": 1.46, "learning_rate": 2.5648999046711155e-06, "loss": 3.4008, "step": 7657 }, { "epoch": 1.46, "learning_rate": 2.5648427073403243e-06, "loss": 3.2998, "step": 7658 }, { "epoch": 1.46, "learning_rate": 2.5647855100095327e-06, "loss": 3.0517, "step": 7659 }, { "epoch": 1.46, "learning_rate": 2.564728312678742e-06, "loss": 3.1227, "step": 7660 }, { "epoch": 1.46, "learning_rate": 2.5646711153479507e-06, "loss": 3.5136, "step": 7661 }, { "epoch": 1.46, "learning_rate": 2.564613918017159e-06, "loss": 2.622, "step": 7662 }, { "epoch": 1.46, "learning_rate": 2.5645567206863683e-06, "loss": 2.4872, "step": 7663 }, { "epoch": 1.46, "learning_rate": 2.5644995233555766e-06, "loss": 2.7506, "step": 7664 }, { "epoch": 1.46, "learning_rate": 2.5644423260247854e-06, "loss": 3.5971, "step": 7665 }, { "epoch": 1.46, "learning_rate": 2.5643851286939946e-06, "loss": 2.8088, "step": 7666 }, { "epoch": 1.46, "learning_rate": 2.564327931363203e-06, "loss": 2.5592, "step": 7667 }, { "epoch": 1.46, "learning_rate": 2.564270734032412e-06, "loss": 3.0116, "step": 7668 }, { "epoch": 1.46, "learning_rate": 2.5642135367016206e-06, "loss": 3.8801, "step": 7669 }, { "epoch": 1.46, "learning_rate": 2.5641563393708294e-06, "loss": 3.057, "step": 7670 }, { "epoch": 1.46, "learning_rate": 2.564099142040038e-06, "loss": 3.3736, "step": 7671 }, { "epoch": 1.46, "learning_rate": 2.564041944709247e-06, "loss": 2.51, "step": 7672 }, { "epoch": 1.46, "learning_rate": 2.5639847473784558e-06, "loss": 2.7392, "step": 7673 }, { "epoch": 1.46, "learning_rate": 2.5639275500476646e-06, "loss": 3.2634, "step": 7674 }, { "epoch": 1.46, "learning_rate": 2.5638703527168734e-06, "loss": 3.4122, "step": 7675 }, { "epoch": 1.46, "learning_rate": 2.563813155386082e-06, "loss": 3.1238, "step": 7676 }, { "epoch": 1.46, "learning_rate": 2.563755958055291e-06, "loss": 2.9648, "step": 7677 }, { "epoch": 1.46, "learning_rate": 2.5636987607244997e-06, "loss": 2.8876, "step": 7678 }, { "epoch": 1.46, "learning_rate": 2.563641563393708e-06, "loss": 3.2015, "step": 7679 }, { "epoch": 1.46, "learning_rate": 2.5635843660629173e-06, "loss": 3.1751, "step": 7680 }, { "epoch": 1.46, "learning_rate": 2.5635271687321257e-06, "loss": 2.3765, "step": 7681 }, { "epoch": 1.46, "learning_rate": 2.5634699714013345e-06, "loss": 3.3132, "step": 7682 }, { "epoch": 1.46, "learning_rate": 2.5634127740705437e-06, "loss": 3.0883, "step": 7683 }, { "epoch": 1.46, "learning_rate": 2.563355576739752e-06, "loss": 3.4615, "step": 7684 }, { "epoch": 1.46, "learning_rate": 2.563298379408961e-06, "loss": 3.5434, "step": 7685 }, { "epoch": 1.46, "learning_rate": 2.5632411820781697e-06, "loss": 2.8932, "step": 7686 }, { "epoch": 1.46, "learning_rate": 2.5631839847473785e-06, "loss": 2.4991, "step": 7687 }, { "epoch": 1.46, "learning_rate": 2.5631267874165873e-06, "loss": 3.5397, "step": 7688 }, { "epoch": 1.46, "learning_rate": 2.563069590085796e-06, "loss": 3.4579, "step": 7689 }, { "epoch": 1.46, "learning_rate": 2.563012392755005e-06, "loss": 2.7874, "step": 7690 }, { "epoch": 1.46, "learning_rate": 2.5629551954242136e-06, "loss": 3.2027, "step": 7691 }, { "epoch": 1.47, "learning_rate": 2.5628979980934224e-06, "loss": 2.7609, "step": 7692 }, { "epoch": 1.47, "learning_rate": 2.5628408007626312e-06, "loss": 2.941, "step": 7693 }, { "epoch": 1.47, "learning_rate": 2.56278360343184e-06, "loss": 3.2543, "step": 7694 }, { "epoch": 1.47, "learning_rate": 2.562726406101049e-06, "loss": 3.5411, "step": 7695 }, { "epoch": 1.47, "learning_rate": 2.562669208770257e-06, "loss": 2.6207, "step": 7696 }, { "epoch": 1.47, "learning_rate": 2.5626120114394664e-06, "loss": 3.7649, "step": 7697 }, { "epoch": 1.47, "learning_rate": 2.5625548141086748e-06, "loss": 3.2675, "step": 7698 }, { "epoch": 1.47, "learning_rate": 2.5624976167778836e-06, "loss": 3.1644, "step": 7699 }, { "epoch": 1.47, "learning_rate": 2.5624404194470928e-06, "loss": 3.0919, "step": 7700 }, { "epoch": 1.47, "learning_rate": 2.562383222116301e-06, "loss": 3.1469, "step": 7701 }, { "epoch": 1.47, "learning_rate": 2.56232602478551e-06, "loss": 2.4058, "step": 7702 }, { "epoch": 1.47, "learning_rate": 2.5622688274547187e-06, "loss": 3.5106, "step": 7703 }, { "epoch": 1.47, "learning_rate": 2.5622116301239275e-06, "loss": 2.5706, "step": 7704 }, { "epoch": 1.47, "learning_rate": 2.5621544327931368e-06, "loss": 2.4408, "step": 7705 }, { "epoch": 1.47, "learning_rate": 2.562097235462345e-06, "loss": 2.9603, "step": 7706 }, { "epoch": 1.47, "learning_rate": 2.562040038131554e-06, "loss": 3.7724, "step": 7707 }, { "epoch": 1.47, "learning_rate": 2.5619828408007627e-06, "loss": 3.5705, "step": 7708 }, { "epoch": 1.47, "learning_rate": 2.5619256434699715e-06, "loss": 3.15, "step": 7709 }, { "epoch": 1.47, "learning_rate": 2.5618684461391803e-06, "loss": 2.9921, "step": 7710 }, { "epoch": 1.47, "learning_rate": 2.561811248808389e-06, "loss": 3.182, "step": 7711 }, { "epoch": 1.47, "learning_rate": 2.561754051477598e-06, "loss": 3.47, "step": 7712 }, { "epoch": 1.47, "learning_rate": 2.5616968541468063e-06, "loss": 3.4283, "step": 7713 }, { "epoch": 1.47, "learning_rate": 2.5616396568160155e-06, "loss": 3.3174, "step": 7714 }, { "epoch": 1.47, "learning_rate": 2.5615824594852243e-06, "loss": 2.7841, "step": 7715 }, { "epoch": 1.47, "learning_rate": 2.5615252621544326e-06, "loss": 3.0759, "step": 7716 }, { "epoch": 1.47, "learning_rate": 2.561468064823642e-06, "loss": 3.4661, "step": 7717 }, { "epoch": 1.47, "learning_rate": 2.5614108674928502e-06, "loss": 3.1232, "step": 7718 }, { "epoch": 1.47, "learning_rate": 2.561353670162059e-06, "loss": 2.9955, "step": 7719 }, { "epoch": 1.47, "learning_rate": 2.561296472831268e-06, "loss": 3.2339, "step": 7720 }, { "epoch": 1.47, "learning_rate": 2.5612392755004766e-06, "loss": 3.2299, "step": 7721 }, { "epoch": 1.47, "learning_rate": 2.561182078169686e-06, "loss": 3.1508, "step": 7722 }, { "epoch": 1.47, "learning_rate": 2.561124880838894e-06, "loss": 2.6844, "step": 7723 }, { "epoch": 1.47, "learning_rate": 2.561067683508103e-06, "loss": 2.6686, "step": 7724 }, { "epoch": 1.47, "learning_rate": 2.5610104861773118e-06, "loss": 3.1143, "step": 7725 }, { "epoch": 1.47, "learning_rate": 2.5609532888465206e-06, "loss": 3.2924, "step": 7726 }, { "epoch": 1.47, "learning_rate": 2.5608960915157294e-06, "loss": 3.187, "step": 7727 }, { "epoch": 1.47, "learning_rate": 2.560838894184938e-06, "loss": 2.6133, "step": 7728 }, { "epoch": 1.47, "learning_rate": 2.560781696854147e-06, "loss": 2.7115, "step": 7729 }, { "epoch": 1.47, "learning_rate": 2.5607244995233553e-06, "loss": 3.0142, "step": 7730 }, { "epoch": 1.47, "learning_rate": 2.5606673021925645e-06, "loss": 3.3531, "step": 7731 }, { "epoch": 1.47, "learning_rate": 2.5606101048617733e-06, "loss": 3.1372, "step": 7732 }, { "epoch": 1.47, "learning_rate": 2.5605529075309817e-06, "loss": 3.1809, "step": 7733 }, { "epoch": 1.47, "learning_rate": 2.560495710200191e-06, "loss": 3.0864, "step": 7734 }, { "epoch": 1.47, "learning_rate": 2.5604385128693993e-06, "loss": 2.9757, "step": 7735 }, { "epoch": 1.47, "learning_rate": 2.560381315538608e-06, "loss": 2.1764, "step": 7736 }, { "epoch": 1.47, "learning_rate": 2.560324118207817e-06, "loss": 3.2652, "step": 7737 }, { "epoch": 1.47, "learning_rate": 2.5602669208770257e-06, "loss": 3.2641, "step": 7738 }, { "epoch": 1.47, "learning_rate": 2.560209723546235e-06, "loss": 3.3138, "step": 7739 }, { "epoch": 1.47, "learning_rate": 2.5601525262154433e-06, "loss": 2.9662, "step": 7740 }, { "epoch": 1.47, "learning_rate": 2.560095328884652e-06, "loss": 2.254, "step": 7741 }, { "epoch": 1.47, "learning_rate": 2.560038131553861e-06, "loss": 3.0057, "step": 7742 }, { "epoch": 1.47, "learning_rate": 2.5599809342230697e-06, "loss": 3.2214, "step": 7743 }, { "epoch": 1.48, "learning_rate": 2.5599237368922784e-06, "loss": 3.042, "step": 7744 }, { "epoch": 1.48, "learning_rate": 2.5598665395614872e-06, "loss": 2.9115, "step": 7745 }, { "epoch": 1.48, "learning_rate": 2.559809342230696e-06, "loss": 4.1142, "step": 7746 }, { "epoch": 1.48, "learning_rate": 2.5597521448999044e-06, "loss": 2.8289, "step": 7747 }, { "epoch": 1.48, "learning_rate": 2.5596949475691136e-06, "loss": 2.2362, "step": 7748 }, { "epoch": 1.48, "learning_rate": 2.5596377502383224e-06, "loss": 3.0915, "step": 7749 }, { "epoch": 1.48, "learning_rate": 2.5595805529075308e-06, "loss": 2.7293, "step": 7750 }, { "epoch": 1.48, "learning_rate": 2.55952335557674e-06, "loss": 3.0297, "step": 7751 }, { "epoch": 1.48, "learning_rate": 2.5594661582459484e-06, "loss": 3.8655, "step": 7752 }, { "epoch": 1.48, "learning_rate": 2.559408960915157e-06, "loss": 3.0344, "step": 7753 }, { "epoch": 1.48, "learning_rate": 2.5593517635843664e-06, "loss": 3.4707, "step": 7754 }, { "epoch": 1.48, "learning_rate": 2.5592945662535748e-06, "loss": 3.1645, "step": 7755 }, { "epoch": 1.48, "learning_rate": 2.559237368922784e-06, "loss": 3.2801, "step": 7756 }, { "epoch": 1.48, "learning_rate": 2.5591801715919923e-06, "loss": 3.3541, "step": 7757 }, { "epoch": 1.48, "learning_rate": 2.559122974261201e-06, "loss": 2.7378, "step": 7758 }, { "epoch": 1.48, "learning_rate": 2.55906577693041e-06, "loss": 3.279, "step": 7759 }, { "epoch": 1.48, "learning_rate": 2.5590085795996187e-06, "loss": 2.5532, "step": 7760 }, { "epoch": 1.48, "learning_rate": 2.5589513822688275e-06, "loss": 3.3661, "step": 7761 }, { "epoch": 1.48, "learning_rate": 2.5588941849380363e-06, "loss": 2.7118, "step": 7762 }, { "epoch": 1.48, "learning_rate": 2.558836987607245e-06, "loss": 2.8018, "step": 7763 }, { "epoch": 1.48, "learning_rate": 2.558779790276454e-06, "loss": 3.4534, "step": 7764 }, { "epoch": 1.48, "learning_rate": 2.5587225929456627e-06, "loss": 2.9892, "step": 7765 }, { "epoch": 1.48, "learning_rate": 2.5586653956148715e-06, "loss": 2.694, "step": 7766 }, { "epoch": 1.48, "learning_rate": 2.55860819828408e-06, "loss": 2.5515, "step": 7767 }, { "epoch": 1.48, "learning_rate": 2.558551000953289e-06, "loss": 2.6553, "step": 7768 }, { "epoch": 1.48, "learning_rate": 2.5584938036224974e-06, "loss": 3.1586, "step": 7769 }, { "epoch": 1.48, "learning_rate": 2.5584366062917067e-06, "loss": 3.0909, "step": 7770 }, { "epoch": 1.48, "learning_rate": 2.5583794089609155e-06, "loss": 3.2027, "step": 7771 }, { "epoch": 1.48, "learning_rate": 2.558322211630124e-06, "loss": 2.9071, "step": 7772 }, { "epoch": 1.48, "learning_rate": 2.558265014299333e-06, "loss": 3.0087, "step": 7773 }, { "epoch": 1.48, "learning_rate": 2.5582078169685414e-06, "loss": 3.2281, "step": 7774 }, { "epoch": 1.48, "learning_rate": 2.55815061963775e-06, "loss": 3.46, "step": 7775 }, { "epoch": 1.48, "learning_rate": 2.558093422306959e-06, "loss": 2.9684, "step": 7776 }, { "epoch": 1.48, "learning_rate": 2.558036224976168e-06, "loss": 3.511, "step": 7777 }, { "epoch": 1.48, "learning_rate": 2.5579790276453766e-06, "loss": 2.4885, "step": 7778 }, { "epoch": 1.48, "learning_rate": 2.5579218303145854e-06, "loss": 3.0011, "step": 7779 }, { "epoch": 1.48, "learning_rate": 2.557864632983794e-06, "loss": 2.4262, "step": 7780 }, { "epoch": 1.48, "learning_rate": 2.557807435653003e-06, "loss": 3.6504, "step": 7781 }, { "epoch": 1.48, "learning_rate": 2.5577502383222118e-06, "loss": 2.6817, "step": 7782 }, { "epoch": 1.48, "learning_rate": 2.5576930409914206e-06, "loss": 3.3611, "step": 7783 }, { "epoch": 1.48, "learning_rate": 2.557635843660629e-06, "loss": 3.5115, "step": 7784 }, { "epoch": 1.48, "learning_rate": 2.557578646329838e-06, "loss": 2.6709, "step": 7785 }, { "epoch": 1.48, "learning_rate": 2.5575214489990465e-06, "loss": 2.7997, "step": 7786 }, { "epoch": 1.48, "learning_rate": 2.5574642516682557e-06, "loss": 3.1953, "step": 7787 }, { "epoch": 1.48, "learning_rate": 2.5574070543374645e-06, "loss": 2.5349, "step": 7788 }, { "epoch": 1.48, "learning_rate": 2.557349857006673e-06, "loss": 3.0669, "step": 7789 }, { "epoch": 1.48, "learning_rate": 2.557292659675882e-06, "loss": 3.182, "step": 7790 }, { "epoch": 1.48, "learning_rate": 2.5572354623450905e-06, "loss": 2.9085, "step": 7791 }, { "epoch": 1.48, "learning_rate": 2.5571782650142993e-06, "loss": 3.399, "step": 7792 }, { "epoch": 1.48, "learning_rate": 2.5571210676835085e-06, "loss": 2.9828, "step": 7793 }, { "epoch": 1.48, "learning_rate": 2.557063870352717e-06, "loss": 2.0978, "step": 7794 }, { "epoch": 1.48, "learning_rate": 2.5570066730219257e-06, "loss": 3.1253, "step": 7795 }, { "epoch": 1.48, "learning_rate": 2.5569494756911345e-06, "loss": 2.5289, "step": 7796 }, { "epoch": 1.49, "learning_rate": 2.5568922783603432e-06, "loss": 2.7784, "step": 7797 }, { "epoch": 1.49, "learning_rate": 2.556835081029552e-06, "loss": 1.9714, "step": 7798 }, { "epoch": 1.49, "learning_rate": 2.556777883698761e-06, "loss": 3.3484, "step": 7799 }, { "epoch": 1.49, "learning_rate": 2.5567206863679696e-06, "loss": 2.8098, "step": 7800 }, { "epoch": 1.49, "learning_rate": 2.556663489037178e-06, "loss": 3.4456, "step": 7801 }, { "epoch": 1.49, "learning_rate": 2.5566062917063872e-06, "loss": 2.9539, "step": 7802 }, { "epoch": 1.49, "learning_rate": 2.556549094375596e-06, "loss": 3.3433, "step": 7803 }, { "epoch": 1.49, "learning_rate": 2.556491897044805e-06, "loss": 3.811, "step": 7804 }, { "epoch": 1.49, "learning_rate": 2.5564346997140136e-06, "loss": 3.5728, "step": 7805 }, { "epoch": 1.49, "learning_rate": 2.556377502383222e-06, "loss": 2.8356, "step": 7806 }, { "epoch": 1.49, "learning_rate": 2.556320305052431e-06, "loss": 3.3022, "step": 7807 }, { "epoch": 1.49, "learning_rate": 2.5562631077216396e-06, "loss": 2.9141, "step": 7808 }, { "epoch": 1.49, "learning_rate": 2.5562059103908483e-06, "loss": 3.6051, "step": 7809 }, { "epoch": 1.49, "learning_rate": 2.5561487130600576e-06, "loss": 2.8832, "step": 7810 }, { "epoch": 1.49, "learning_rate": 2.556091515729266e-06, "loss": 2.7579, "step": 7811 }, { "epoch": 1.49, "learning_rate": 2.5560343183984747e-06, "loss": 3.1324, "step": 7812 }, { "epoch": 1.49, "learning_rate": 2.5559771210676835e-06, "loss": 3.6679, "step": 7813 }, { "epoch": 1.49, "learning_rate": 2.5559199237368923e-06, "loss": 3.0162, "step": 7814 }, { "epoch": 1.49, "learning_rate": 2.555862726406101e-06, "loss": 2.7356, "step": 7815 }, { "epoch": 1.49, "learning_rate": 2.55580552907531e-06, "loss": 3.1192, "step": 7816 }, { "epoch": 1.49, "learning_rate": 2.5557483317445187e-06, "loss": 3.7217, "step": 7817 }, { "epoch": 1.49, "learning_rate": 2.555691134413727e-06, "loss": 2.7349, "step": 7818 }, { "epoch": 1.49, "learning_rate": 2.5556339370829363e-06, "loss": 2.4381, "step": 7819 }, { "epoch": 1.49, "learning_rate": 2.555576739752145e-06, "loss": 3.6089, "step": 7820 }, { "epoch": 1.49, "learning_rate": 2.555519542421354e-06, "loss": 2.6935, "step": 7821 }, { "epoch": 1.49, "learning_rate": 2.5554623450905627e-06, "loss": 3.0578, "step": 7822 }, { "epoch": 1.49, "learning_rate": 2.555405147759771e-06, "loss": 3.6226, "step": 7823 }, { "epoch": 1.49, "learning_rate": 2.5553479504289803e-06, "loss": 3.9039, "step": 7824 }, { "epoch": 1.49, "learning_rate": 2.5552907530981886e-06, "loss": 2.8326, "step": 7825 }, { "epoch": 1.49, "learning_rate": 2.5552335557673974e-06, "loss": 3.1392, "step": 7826 }, { "epoch": 1.49, "learning_rate": 2.5551763584366066e-06, "loss": 2.2706, "step": 7827 }, { "epoch": 1.49, "learning_rate": 2.555119161105815e-06, "loss": 3.4703, "step": 7828 }, { "epoch": 1.49, "learning_rate": 2.555061963775024e-06, "loss": 2.9943, "step": 7829 }, { "epoch": 1.49, "learning_rate": 2.5550047664442326e-06, "loss": 3.0543, "step": 7830 }, { "epoch": 1.49, "learning_rate": 2.5549475691134414e-06, "loss": 2.9293, "step": 7831 }, { "epoch": 1.49, "learning_rate": 2.55489037178265e-06, "loss": 2.508, "step": 7832 }, { "epoch": 1.49, "learning_rate": 2.554833174451859e-06, "loss": 3.3519, "step": 7833 }, { "epoch": 1.49, "learning_rate": 2.5547759771210678e-06, "loss": 3.0702, "step": 7834 }, { "epoch": 1.49, "learning_rate": 2.5547187797902766e-06, "loss": 2.5625, "step": 7835 }, { "epoch": 1.49, "learning_rate": 2.5546615824594854e-06, "loss": 2.4394, "step": 7836 }, { "epoch": 1.49, "learning_rate": 2.554604385128694e-06, "loss": 3.341, "step": 7837 }, { "epoch": 1.49, "learning_rate": 2.554547187797903e-06, "loss": 3.194, "step": 7838 }, { "epoch": 1.49, "learning_rate": 2.5544899904671117e-06, "loss": 3.5562, "step": 7839 }, { "epoch": 1.49, "learning_rate": 2.55443279313632e-06, "loss": 2.9606, "step": 7840 }, { "epoch": 1.49, "learning_rate": 2.5543755958055293e-06, "loss": 2.6335, "step": 7841 }, { "epoch": 1.49, "learning_rate": 2.554318398474738e-06, "loss": 3.5302, "step": 7842 }, { "epoch": 1.49, "learning_rate": 2.5542612011439465e-06, "loss": 3.8019, "step": 7843 }, { "epoch": 1.49, "learning_rate": 2.5542040038131557e-06, "loss": 3.1188, "step": 7844 }, { "epoch": 1.49, "learning_rate": 2.554146806482364e-06, "loss": 2.9281, "step": 7845 }, { "epoch": 1.49, "learning_rate": 2.554089609151573e-06, "loss": 2.544, "step": 7846 }, { "epoch": 1.49, "learning_rate": 2.5540324118207817e-06, "loss": 3.2388, "step": 7847 }, { "epoch": 1.49, "learning_rate": 2.5539752144899905e-06, "loss": 2.6165, "step": 7848 }, { "epoch": 1.5, "learning_rate": 2.5539180171591993e-06, "loss": 3.2981, "step": 7849 }, { "epoch": 1.5, "learning_rate": 2.553860819828408e-06, "loss": 3.0739, "step": 7850 }, { "epoch": 1.5, "learning_rate": 2.553803622497617e-06, "loss": 3.6837, "step": 7851 }, { "epoch": 1.5, "learning_rate": 2.5537464251668256e-06, "loss": 2.8237, "step": 7852 }, { "epoch": 1.5, "learning_rate": 2.5536892278360344e-06, "loss": 3.0362, "step": 7853 }, { "epoch": 1.5, "learning_rate": 2.5536320305052432e-06, "loss": 3.5276, "step": 7854 }, { "epoch": 1.5, "learning_rate": 2.553574833174452e-06, "loss": 2.7116, "step": 7855 }, { "epoch": 1.5, "learning_rate": 2.553517635843661e-06, "loss": 2.877, "step": 7856 }, { "epoch": 1.5, "learning_rate": 2.553460438512869e-06, "loss": 2.9104, "step": 7857 }, { "epoch": 1.5, "learning_rate": 2.5534032411820784e-06, "loss": 3.2864, "step": 7858 }, { "epoch": 1.5, "learning_rate": 2.553346043851287e-06, "loss": 3.1881, "step": 7859 }, { "epoch": 1.5, "learning_rate": 2.5532888465204956e-06, "loss": 3.2336, "step": 7860 }, { "epoch": 1.5, "learning_rate": 2.5532316491897048e-06, "loss": 3.0174, "step": 7861 }, { "epoch": 1.5, "learning_rate": 2.553174451858913e-06, "loss": 2.5802, "step": 7862 }, { "epoch": 1.5, "learning_rate": 2.553117254528122e-06, "loss": 3.2759, "step": 7863 }, { "epoch": 1.5, "learning_rate": 2.5530600571973307e-06, "loss": 2.8274, "step": 7864 }, { "epoch": 1.5, "learning_rate": 2.5530028598665395e-06, "loss": 3.5194, "step": 7865 }, { "epoch": 1.5, "learning_rate": 2.5529456625357483e-06, "loss": 2.732, "step": 7866 }, { "epoch": 1.5, "learning_rate": 2.552888465204957e-06, "loss": 3.5231, "step": 7867 }, { "epoch": 1.5, "learning_rate": 2.552831267874166e-06, "loss": 3.1539, "step": 7868 }, { "epoch": 1.5, "learning_rate": 2.5527740705433747e-06, "loss": 2.9122, "step": 7869 }, { "epoch": 1.5, "learning_rate": 2.5527168732125835e-06, "loss": 2.6538, "step": 7870 }, { "epoch": 1.5, "learning_rate": 2.5526596758817923e-06, "loss": 3.8286, "step": 7871 }, { "epoch": 1.5, "learning_rate": 2.552602478551001e-06, "loss": 2.8554, "step": 7872 }, { "epoch": 1.5, "learning_rate": 2.55254528122021e-06, "loss": 3.043, "step": 7873 }, { "epoch": 1.5, "learning_rate": 2.5524880838894183e-06, "loss": 3.2482, "step": 7874 }, { "epoch": 1.5, "learning_rate": 2.5524308865586275e-06, "loss": 3.7572, "step": 7875 }, { "epoch": 1.5, "learning_rate": 2.5523736892278363e-06, "loss": 3.2684, "step": 7876 }, { "epoch": 1.5, "learning_rate": 2.5523164918970446e-06, "loss": 2.6495, "step": 7877 }, { "epoch": 1.5, "learning_rate": 2.552259294566254e-06, "loss": 2.5072, "step": 7878 }, { "epoch": 1.5, "learning_rate": 2.5522020972354622e-06, "loss": 2.7473, "step": 7879 }, { "epoch": 1.5, "learning_rate": 2.552144899904671e-06, "loss": 2.8619, "step": 7880 }, { "epoch": 1.5, "learning_rate": 2.55208770257388e-06, "loss": 3.4833, "step": 7881 }, { "epoch": 1.5, "learning_rate": 2.5520305052430886e-06, "loss": 2.995, "step": 7882 }, { "epoch": 1.5, "learning_rate": 2.5519733079122974e-06, "loss": 3.0482, "step": 7883 }, { "epoch": 1.5, "learning_rate": 2.551916110581506e-06, "loss": 3.111, "step": 7884 }, { "epoch": 1.5, "learning_rate": 2.551858913250715e-06, "loss": 3.3866, "step": 7885 }, { "epoch": 1.5, "learning_rate": 2.5518017159199238e-06, "loss": 2.8415, "step": 7886 }, { "epoch": 1.5, "learning_rate": 2.5517445185891326e-06, "loss": 3.5453, "step": 7887 }, { "epoch": 1.5, "learning_rate": 2.5516873212583414e-06, "loss": 2.8376, "step": 7888 }, { "epoch": 1.5, "learning_rate": 2.55163012392755e-06, "loss": 3.3751, "step": 7889 }, { "epoch": 1.5, "learning_rate": 2.551572926596759e-06, "loss": 3.2919, "step": 7890 }, { "epoch": 1.5, "learning_rate": 2.5515157292659677e-06, "loss": 3.5714, "step": 7891 }, { "epoch": 1.5, "learning_rate": 2.5514585319351765e-06, "loss": 3.6022, "step": 7892 }, { "epoch": 1.5, "learning_rate": 2.5514013346043853e-06, "loss": 2.5428, "step": 7893 }, { "epoch": 1.5, "learning_rate": 2.5513441372735937e-06, "loss": 2.891, "step": 7894 }, { "epoch": 1.5, "learning_rate": 2.551286939942803e-06, "loss": 3.4677, "step": 7895 }, { "epoch": 1.5, "learning_rate": 2.5512297426120113e-06, "loss": 2.9934, "step": 7896 }, { "epoch": 1.5, "learning_rate": 2.55117254528122e-06, "loss": 3.4745, "step": 7897 }, { "epoch": 1.5, "learning_rate": 2.5511153479504293e-06, "loss": 3.7495, "step": 7898 }, { "epoch": 1.5, "learning_rate": 2.5510581506196377e-06, "loss": 2.868, "step": 7899 }, { "epoch": 1.5, "learning_rate": 2.551000953288847e-06, "loss": 3.1544, "step": 7900 }, { "epoch": 1.5, "learning_rate": 2.5509437559580553e-06, "loss": 2.7711, "step": 7901 }, { "epoch": 1.51, "learning_rate": 2.550886558627264e-06, "loss": 2.1286, "step": 7902 }, { "epoch": 1.51, "learning_rate": 2.550829361296473e-06, "loss": 2.7143, "step": 7903 }, { "epoch": 1.51, "learning_rate": 2.5507721639656816e-06, "loss": 3.6783, "step": 7904 }, { "epoch": 1.51, "learning_rate": 2.5507149666348904e-06, "loss": 3.8934, "step": 7905 }, { "epoch": 1.51, "learning_rate": 2.5506577693040992e-06, "loss": 2.9477, "step": 7906 }, { "epoch": 1.51, "learning_rate": 2.550600571973308e-06, "loss": 3.329, "step": 7907 }, { "epoch": 1.51, "learning_rate": 2.550543374642517e-06, "loss": 2.9628, "step": 7908 }, { "epoch": 1.51, "learning_rate": 2.5504861773117256e-06, "loss": 3.6804, "step": 7909 }, { "epoch": 1.51, "learning_rate": 2.5504289799809344e-06, "loss": 2.6609, "step": 7910 }, { "epoch": 1.51, "learning_rate": 2.5503717826501428e-06, "loss": 2.5267, "step": 7911 }, { "epoch": 1.51, "learning_rate": 2.550314585319352e-06, "loss": 3.6338, "step": 7912 }, { "epoch": 1.51, "learning_rate": 2.5502573879885604e-06, "loss": 2.5953, "step": 7913 }, { "epoch": 1.51, "learning_rate": 2.550200190657769e-06, "loss": 3.2964, "step": 7914 }, { "epoch": 1.51, "learning_rate": 2.5501429933269784e-06, "loss": 2.4888, "step": 7915 }, { "epoch": 1.51, "learning_rate": 2.5500857959961867e-06, "loss": 3.3517, "step": 7916 }, { "epoch": 1.51, "learning_rate": 2.550028598665396e-06, "loss": 2.4022, "step": 7917 }, { "epoch": 1.51, "learning_rate": 2.5499714013346043e-06, "loss": 3.3967, "step": 7918 }, { "epoch": 1.51, "learning_rate": 2.549914204003813e-06, "loss": 3.8568, "step": 7919 }, { "epoch": 1.51, "learning_rate": 2.549857006673022e-06, "loss": 3.0516, "step": 7920 }, { "epoch": 1.51, "learning_rate": 2.5497998093422307e-06, "loss": 2.2841, "step": 7921 }, { "epoch": 1.51, "learning_rate": 2.5497426120114395e-06, "loss": 3.2482, "step": 7922 }, { "epoch": 1.51, "learning_rate": 2.5496854146806483e-06, "loss": 2.9792, "step": 7923 }, { "epoch": 1.51, "learning_rate": 2.549628217349857e-06, "loss": 2.9833, "step": 7924 }, { "epoch": 1.51, "learning_rate": 2.549571020019066e-06, "loss": 3.2149, "step": 7925 }, { "epoch": 1.51, "learning_rate": 2.5495138226882747e-06, "loss": 2.426, "step": 7926 }, { "epoch": 1.51, "learning_rate": 2.5494566253574835e-06, "loss": 4.037, "step": 7927 }, { "epoch": 1.51, "learning_rate": 2.549399428026692e-06, "loss": 3.6428, "step": 7928 }, { "epoch": 1.51, "learning_rate": 2.549342230695901e-06, "loss": 2.9807, "step": 7929 }, { "epoch": 1.51, "learning_rate": 2.54928503336511e-06, "loss": 2.8968, "step": 7930 }, { "epoch": 1.51, "learning_rate": 2.5492278360343182e-06, "loss": 2.5398, "step": 7931 }, { "epoch": 1.51, "learning_rate": 2.5491706387035274e-06, "loss": 3.0222, "step": 7932 }, { "epoch": 1.51, "learning_rate": 2.549113441372736e-06, "loss": 3.1468, "step": 7933 }, { "epoch": 1.51, "learning_rate": 2.549056244041945e-06, "loss": 3.6979, "step": 7934 }, { "epoch": 1.51, "learning_rate": 2.5489990467111534e-06, "loss": 2.7914, "step": 7935 }, { "epoch": 1.51, "learning_rate": 2.548941849380362e-06, "loss": 3.4028, "step": 7936 }, { "epoch": 1.51, "learning_rate": 2.5488846520495714e-06, "loss": 2.8603, "step": 7937 }, { "epoch": 1.51, "learning_rate": 2.5488274547187798e-06, "loss": 3.5418, "step": 7938 }, { "epoch": 1.51, "learning_rate": 2.5487702573879886e-06, "loss": 3.5182, "step": 7939 }, { "epoch": 1.51, "learning_rate": 2.5487130600571974e-06, "loss": 2.7551, "step": 7940 }, { "epoch": 1.51, "learning_rate": 2.548655862726406e-06, "loss": 3.3429, "step": 7941 }, { "epoch": 1.51, "learning_rate": 2.548598665395615e-06, "loss": 2.4416, "step": 7942 }, { "epoch": 1.51, "learning_rate": 2.5485414680648238e-06, "loss": 3.1138, "step": 7943 }, { "epoch": 1.51, "learning_rate": 2.5484842707340326e-06, "loss": 2.896, "step": 7944 }, { "epoch": 1.51, "learning_rate": 2.548427073403241e-06, "loss": 2.9457, "step": 7945 }, { "epoch": 1.51, "learning_rate": 2.54836987607245e-06, "loss": 3.3548, "step": 7946 }, { "epoch": 1.51, "learning_rate": 2.548312678741659e-06, "loss": 3.2745, "step": 7947 }, { "epoch": 1.51, "learning_rate": 2.5482554814108673e-06, "loss": 2.7613, "step": 7948 }, { "epoch": 1.51, "learning_rate": 2.5481982840800765e-06, "loss": 2.9108, "step": 7949 }, { "epoch": 1.51, "learning_rate": 2.548141086749285e-06, "loss": 2.6106, "step": 7950 }, { "epoch": 1.51, "learning_rate": 2.548083889418494e-06, "loss": 3.3638, "step": 7951 }, { "epoch": 1.51, "learning_rate": 2.5480266920877025e-06, "loss": 2.8149, "step": 7952 }, { "epoch": 1.51, "learning_rate": 2.5479694947569113e-06, "loss": 2.4824, "step": 7953 }, { "epoch": 1.52, "learning_rate": 2.5479122974261205e-06, "loss": 3.5347, "step": 7954 }, { "epoch": 1.52, "learning_rate": 2.547855100095329e-06, "loss": 3.114, "step": 7955 }, { "epoch": 1.52, "learning_rate": 2.5477979027645377e-06, "loss": 3.0046, "step": 7956 }, { "epoch": 1.52, "learning_rate": 2.5477407054337464e-06, "loss": 2.4198, "step": 7957 }, { "epoch": 1.52, "learning_rate": 2.5476835081029552e-06, "loss": 2.9495, "step": 7958 }, { "epoch": 1.52, "learning_rate": 2.547626310772164e-06, "loss": 2.85, "step": 7959 }, { "epoch": 1.52, "learning_rate": 2.547569113441373e-06, "loss": 3.1346, "step": 7960 }, { "epoch": 1.52, "learning_rate": 2.5475119161105816e-06, "loss": 3.4765, "step": 7961 }, { "epoch": 1.52, "learning_rate": 2.54745471877979e-06, "loss": 3.2395, "step": 7962 }, { "epoch": 1.52, "learning_rate": 2.547397521448999e-06, "loss": 3.2679, "step": 7963 }, { "epoch": 1.52, "learning_rate": 2.547340324118208e-06, "loss": 2.4907, "step": 7964 }, { "epoch": 1.52, "learning_rate": 2.547283126787417e-06, "loss": 2.8157, "step": 7965 }, { "epoch": 1.52, "learning_rate": 2.5472259294566256e-06, "loss": 2.962, "step": 7966 }, { "epoch": 1.52, "learning_rate": 2.547168732125834e-06, "loss": 3.0372, "step": 7967 }, { "epoch": 1.52, "learning_rate": 2.547111534795043e-06, "loss": 3.0936, "step": 7968 }, { "epoch": 1.52, "learning_rate": 2.5470543374642515e-06, "loss": 3.444, "step": 7969 }, { "epoch": 1.52, "learning_rate": 2.5469971401334603e-06, "loss": 2.301, "step": 7970 }, { "epoch": 1.52, "learning_rate": 2.5469399428026696e-06, "loss": 2.6224, "step": 7971 }, { "epoch": 1.52, "learning_rate": 2.546882745471878e-06, "loss": 2.5427, "step": 7972 }, { "epoch": 1.52, "learning_rate": 2.5468255481410867e-06, "loss": 2.7973, "step": 7973 }, { "epoch": 1.52, "learning_rate": 2.5467683508102955e-06, "loss": 3.0823, "step": 7974 }, { "epoch": 1.52, "learning_rate": 2.5467111534795043e-06, "loss": 2.8047, "step": 7975 }, { "epoch": 1.52, "learning_rate": 2.546653956148713e-06, "loss": 2.6803, "step": 7976 }, { "epoch": 1.52, "learning_rate": 2.546596758817922e-06, "loss": 2.9877, "step": 7977 }, { "epoch": 1.52, "learning_rate": 2.5465395614871307e-06, "loss": 2.5912, "step": 7978 }, { "epoch": 1.52, "learning_rate": 2.5464823641563395e-06, "loss": 2.359, "step": 7979 }, { "epoch": 1.52, "learning_rate": 2.5464251668255483e-06, "loss": 3.2466, "step": 7980 }, { "epoch": 1.52, "learning_rate": 2.546367969494757e-06, "loss": 2.6504, "step": 7981 }, { "epoch": 1.52, "learning_rate": 2.546310772163966e-06, "loss": 3.163, "step": 7982 }, { "epoch": 1.52, "learning_rate": 2.5462535748331747e-06, "loss": 3.1024, "step": 7983 }, { "epoch": 1.52, "learning_rate": 2.546196377502383e-06, "loss": 3.6868, "step": 7984 }, { "epoch": 1.52, "learning_rate": 2.5461391801715922e-06, "loss": 3.5159, "step": 7985 }, { "epoch": 1.52, "learning_rate": 2.546081982840801e-06, "loss": 2.9295, "step": 7986 }, { "epoch": 1.52, "learning_rate": 2.5460247855100094e-06, "loss": 3.0774, "step": 7987 }, { "epoch": 1.52, "learning_rate": 2.5459675881792186e-06, "loss": 3.5463, "step": 7988 }, { "epoch": 1.52, "learning_rate": 2.545910390848427e-06, "loss": 2.6573, "step": 7989 }, { "epoch": 1.52, "learning_rate": 2.545853193517636e-06, "loss": 3.4763, "step": 7990 }, { "epoch": 1.52, "learning_rate": 2.5457959961868446e-06, "loss": 2.9712, "step": 7991 }, { "epoch": 1.52, "learning_rate": 2.5457387988560534e-06, "loss": 2.9974, "step": 7992 }, { "epoch": 1.52, "learning_rate": 2.545681601525262e-06, "loss": 2.6303, "step": 7993 }, { "epoch": 1.52, "learning_rate": 2.545624404194471e-06, "loss": 3.1411, "step": 7994 }, { "epoch": 1.52, "learning_rate": 2.5455672068636798e-06, "loss": 2.9813, "step": 7995 }, { "epoch": 1.52, "learning_rate": 2.5455100095328886e-06, "loss": 3.5386, "step": 7996 }, { "epoch": 1.52, "learning_rate": 2.5454528122020974e-06, "loss": 3.2597, "step": 7997 }, { "epoch": 1.52, "learning_rate": 2.545395614871306e-06, "loss": 3.1776, "step": 7998 }, { "epoch": 1.52, "learning_rate": 2.545338417540515e-06, "loss": 4.1876, "step": 7999 }, { "epoch": 1.52, "learning_rate": 2.5452812202097237e-06, "loss": 2.9612, "step": 8000 }, { "epoch": 1.52, "learning_rate": 2.545224022878932e-06, "loss": 3.0318, "step": 8001 }, { "epoch": 1.52, "learning_rate": 2.5451668255481413e-06, "loss": 2.9848, "step": 8002 }, { "epoch": 1.52, "learning_rate": 2.54510962821735e-06, "loss": 3.4673, "step": 8003 }, { "epoch": 1.52, "learning_rate": 2.5450524308865585e-06, "loss": 3.0507, "step": 8004 }, { "epoch": 1.52, "learning_rate": 2.5449952335557677e-06, "loss": 2.7683, "step": 8005 }, { "epoch": 1.52, "learning_rate": 2.544938036224976e-06, "loss": 2.6994, "step": 8006 }, { "epoch": 1.53, "learning_rate": 2.544880838894185e-06, "loss": 3.8342, "step": 8007 }, { "epoch": 1.53, "learning_rate": 2.5448236415633937e-06, "loss": 3.1241, "step": 8008 }, { "epoch": 1.53, "learning_rate": 2.5447664442326025e-06, "loss": 3.2058, "step": 8009 }, { "epoch": 1.53, "learning_rate": 2.5447092469018112e-06, "loss": 2.7263, "step": 8010 }, { "epoch": 1.53, "learning_rate": 2.54465204957102e-06, "loss": 2.2726, "step": 8011 }, { "epoch": 1.53, "learning_rate": 2.544594852240229e-06, "loss": 3.3292, "step": 8012 }, { "epoch": 1.53, "learning_rate": 2.5445376549094376e-06, "loss": 2.933, "step": 8013 }, { "epoch": 1.53, "learning_rate": 2.5444804575786464e-06, "loss": 3.0206, "step": 8014 }, { "epoch": 1.53, "learning_rate": 2.5444232602478552e-06, "loss": 3.7064, "step": 8015 }, { "epoch": 1.53, "learning_rate": 2.544366062917064e-06, "loss": 2.6902, "step": 8016 }, { "epoch": 1.53, "learning_rate": 2.544308865586273e-06, "loss": 2.9221, "step": 8017 }, { "epoch": 1.53, "learning_rate": 2.5442516682554816e-06, "loss": 3.1667, "step": 8018 }, { "epoch": 1.53, "learning_rate": 2.5441944709246904e-06, "loss": 3.1786, "step": 8019 }, { "epoch": 1.53, "learning_rate": 2.544137273593899e-06, "loss": 3.5308, "step": 8020 }, { "epoch": 1.53, "learning_rate": 2.5440800762631076e-06, "loss": 2.7016, "step": 8021 }, { "epoch": 1.53, "learning_rate": 2.5440228789323168e-06, "loss": 2.6678, "step": 8022 }, { "epoch": 1.53, "learning_rate": 2.543965681601525e-06, "loss": 2.6059, "step": 8023 }, { "epoch": 1.53, "learning_rate": 2.543908484270734e-06, "loss": 3.2222, "step": 8024 }, { "epoch": 1.53, "learning_rate": 2.543851286939943e-06, "loss": 2.5781, "step": 8025 }, { "epoch": 1.53, "learning_rate": 2.5437940896091515e-06, "loss": 3.7256, "step": 8026 }, { "epoch": 1.53, "learning_rate": 2.5437368922783603e-06, "loss": 2.525, "step": 8027 }, { "epoch": 1.53, "learning_rate": 2.543679694947569e-06, "loss": 3.342, "step": 8028 }, { "epoch": 1.53, "learning_rate": 2.543622497616778e-06, "loss": 2.9522, "step": 8029 }, { "epoch": 1.53, "learning_rate": 2.5435653002859867e-06, "loss": 2.5115, "step": 8030 }, { "epoch": 1.53, "learning_rate": 2.5435081029551955e-06, "loss": 2.8258, "step": 8031 }, { "epoch": 1.53, "learning_rate": 2.5434509056244043e-06, "loss": 3.4939, "step": 8032 }, { "epoch": 1.53, "learning_rate": 2.543393708293613e-06, "loss": 2.9746, "step": 8033 }, { "epoch": 1.53, "learning_rate": 2.543336510962822e-06, "loss": 2.9319, "step": 8034 }, { "epoch": 1.53, "learning_rate": 2.5432793136320307e-06, "loss": 2.6091, "step": 8035 }, { "epoch": 1.53, "learning_rate": 2.5432221163012395e-06, "loss": 2.6918, "step": 8036 }, { "epoch": 1.53, "learning_rate": 2.5431649189704483e-06, "loss": 3.2771, "step": 8037 }, { "epoch": 1.53, "learning_rate": 2.5431077216396566e-06, "loss": 2.2096, "step": 8038 }, { "epoch": 1.53, "learning_rate": 2.543050524308866e-06, "loss": 3.3246, "step": 8039 }, { "epoch": 1.53, "learning_rate": 2.5429933269780742e-06, "loss": 3.6885, "step": 8040 }, { "epoch": 1.53, "learning_rate": 2.542936129647283e-06, "loss": 3.7699, "step": 8041 }, { "epoch": 1.53, "learning_rate": 2.5428789323164922e-06, "loss": 3.2508, "step": 8042 }, { "epoch": 1.53, "learning_rate": 2.5428217349857006e-06, "loss": 3.8187, "step": 8043 }, { "epoch": 1.53, "learning_rate": 2.5427645376549094e-06, "loss": 2.2839, "step": 8044 }, { "epoch": 1.53, "learning_rate": 2.542707340324118e-06, "loss": 2.8934, "step": 8045 }, { "epoch": 1.53, "learning_rate": 2.542650142993327e-06, "loss": 2.46, "step": 8046 }, { "epoch": 1.53, "learning_rate": 2.5425929456625358e-06, "loss": 3.0714, "step": 8047 }, { "epoch": 1.53, "learning_rate": 2.5425357483317446e-06, "loss": 3.1529, "step": 8048 }, { "epoch": 1.53, "learning_rate": 2.5424785510009534e-06, "loss": 3.6912, "step": 8049 }, { "epoch": 1.53, "learning_rate": 2.542421353670162e-06, "loss": 3.061, "step": 8050 }, { "epoch": 1.53, "learning_rate": 2.542364156339371e-06, "loss": 3.8268, "step": 8051 }, { "epoch": 1.53, "learning_rate": 2.5423069590085797e-06, "loss": 3.5449, "step": 8052 }, { "epoch": 1.53, "learning_rate": 2.5422497616777885e-06, "loss": 2.7532, "step": 8053 }, { "epoch": 1.53, "learning_rate": 2.5421925643469973e-06, "loss": 2.7652, "step": 8054 }, { "epoch": 1.53, "learning_rate": 2.5421353670162057e-06, "loss": 3.2404, "step": 8055 }, { "epoch": 1.53, "learning_rate": 2.542078169685415e-06, "loss": 3.2079, "step": 8056 }, { "epoch": 1.53, "learning_rate": 2.5420209723546233e-06, "loss": 3.2833, "step": 8057 }, { "epoch": 1.53, "learning_rate": 2.541963775023832e-06, "loss": 3.0143, "step": 8058 }, { "epoch": 1.54, "learning_rate": 2.5419065776930413e-06, "loss": 3.726, "step": 8059 }, { "epoch": 1.54, "learning_rate": 2.5418493803622497e-06, "loss": 2.9778, "step": 8060 }, { "epoch": 1.54, "learning_rate": 2.5417921830314585e-06, "loss": 2.9441, "step": 8061 }, { "epoch": 1.54, "learning_rate": 2.5417349857006673e-06, "loss": 3.3362, "step": 8062 }, { "epoch": 1.54, "learning_rate": 2.541677788369876e-06, "loss": 3.7073, "step": 8063 }, { "epoch": 1.54, "learning_rate": 2.5416205910390853e-06, "loss": 3.081, "step": 8064 }, { "epoch": 1.54, "learning_rate": 2.5415633937082936e-06, "loss": 3.0333, "step": 8065 }, { "epoch": 1.54, "learning_rate": 2.5415061963775024e-06, "loss": 3.1804, "step": 8066 }, { "epoch": 1.54, "learning_rate": 2.5414489990467112e-06, "loss": 2.9683, "step": 8067 }, { "epoch": 1.54, "learning_rate": 2.54139180171592e-06, "loss": 2.9016, "step": 8068 }, { "epoch": 1.54, "learning_rate": 2.541334604385129e-06, "loss": 2.8037, "step": 8069 }, { "epoch": 1.54, "learning_rate": 2.5412774070543376e-06, "loss": 3.3991, "step": 8070 }, { "epoch": 1.54, "learning_rate": 2.5412202097235464e-06, "loss": 2.5447, "step": 8071 }, { "epoch": 1.54, "learning_rate": 2.5411630123927548e-06, "loss": 2.9081, "step": 8072 }, { "epoch": 1.54, "learning_rate": 2.541105815061964e-06, "loss": 3.1437, "step": 8073 }, { "epoch": 1.54, "learning_rate": 2.5410486177311728e-06, "loss": 3.2616, "step": 8074 }, { "epoch": 1.54, "learning_rate": 2.540991420400381e-06, "loss": 3.7236, "step": 8075 }, { "epoch": 1.54, "learning_rate": 2.5409342230695904e-06, "loss": 2.6011, "step": 8076 }, { "epoch": 1.54, "learning_rate": 2.5408770257387987e-06, "loss": 3.2777, "step": 8077 }, { "epoch": 1.54, "learning_rate": 2.5408198284080075e-06, "loss": 2.6239, "step": 8078 }, { "epoch": 1.54, "learning_rate": 2.5407626310772163e-06, "loss": 2.8096, "step": 8079 }, { "epoch": 1.54, "learning_rate": 2.540705433746425e-06, "loss": 3.377, "step": 8080 }, { "epoch": 1.54, "learning_rate": 2.5406482364156343e-06, "loss": 2.9934, "step": 8081 }, { "epoch": 1.54, "learning_rate": 2.5405910390848427e-06, "loss": 3.0519, "step": 8082 }, { "epoch": 1.54, "learning_rate": 2.5405338417540515e-06, "loss": 2.4546, "step": 8083 }, { "epoch": 1.54, "learning_rate": 2.5404766444232603e-06, "loss": 3.1288, "step": 8084 }, { "epoch": 1.54, "learning_rate": 2.540419447092469e-06, "loss": 3.41, "step": 8085 }, { "epoch": 1.54, "learning_rate": 2.540362249761678e-06, "loss": 3.6174, "step": 8086 }, { "epoch": 1.54, "learning_rate": 2.5403050524308867e-06, "loss": 2.427, "step": 8087 }, { "epoch": 1.54, "learning_rate": 2.5402478551000955e-06, "loss": 2.258, "step": 8088 }, { "epoch": 1.54, "learning_rate": 2.540190657769304e-06, "loss": 3.8436, "step": 8089 }, { "epoch": 1.54, "learning_rate": 2.540133460438513e-06, "loss": 3.4192, "step": 8090 }, { "epoch": 1.54, "learning_rate": 2.540076263107722e-06, "loss": 3.1712, "step": 8091 }, { "epoch": 1.54, "learning_rate": 2.5400190657769302e-06, "loss": 3.2498, "step": 8092 }, { "epoch": 1.54, "learning_rate": 2.5399618684461394e-06, "loss": 3.0146, "step": 8093 }, { "epoch": 1.54, "learning_rate": 2.539904671115348e-06, "loss": 3.3485, "step": 8094 }, { "epoch": 1.54, "learning_rate": 2.539847473784557e-06, "loss": 3.6969, "step": 8095 }, { "epoch": 1.54, "learning_rate": 2.5397902764537654e-06, "loss": 3.1307, "step": 8096 }, { "epoch": 1.54, "learning_rate": 2.539733079122974e-06, "loss": 3.0709, "step": 8097 }, { "epoch": 1.54, "learning_rate": 2.5396758817921834e-06, "loss": 3.0897, "step": 8098 }, { "epoch": 1.54, "learning_rate": 2.5396186844613918e-06, "loss": 2.8584, "step": 8099 }, { "epoch": 1.54, "learning_rate": 2.5395614871306006e-06, "loss": 3.2842, "step": 8100 }, { "epoch": 1.54, "learning_rate": 2.5395042897998094e-06, "loss": 3.3147, "step": 8101 }, { "epoch": 1.54, "learning_rate": 2.539447092469018e-06, "loss": 3.1562, "step": 8102 }, { "epoch": 1.54, "learning_rate": 2.539389895138227e-06, "loss": 3.5585, "step": 8103 }, { "epoch": 1.54, "learning_rate": 2.5393326978074357e-06, "loss": 2.7617, "step": 8104 }, { "epoch": 1.54, "learning_rate": 2.5392755004766445e-06, "loss": 2.9086, "step": 8105 }, { "epoch": 1.54, "learning_rate": 2.5392183031458533e-06, "loss": 3.0559, "step": 8106 }, { "epoch": 1.54, "learning_rate": 2.539161105815062e-06, "loss": 2.6743, "step": 8107 }, { "epoch": 1.54, "learning_rate": 2.539103908484271e-06, "loss": 2.5229, "step": 8108 }, { "epoch": 1.54, "learning_rate": 2.5390467111534793e-06, "loss": 2.5047, "step": 8109 }, { "epoch": 1.54, "learning_rate": 2.5389895138226885e-06, "loss": 3.4426, "step": 8110 }, { "epoch": 1.54, "learning_rate": 2.538932316491897e-06, "loss": 3.6412, "step": 8111 }, { "epoch": 1.55, "learning_rate": 2.538875119161106e-06, "loss": 3.3704, "step": 8112 }, { "epoch": 1.55, "learning_rate": 2.538817921830315e-06, "loss": 3.0387, "step": 8113 }, { "epoch": 1.55, "learning_rate": 2.5387607244995233e-06, "loss": 3.365, "step": 8114 }, { "epoch": 1.55, "learning_rate": 2.5387035271687325e-06, "loss": 2.7755, "step": 8115 }, { "epoch": 1.55, "learning_rate": 2.538646329837941e-06, "loss": 3.8933, "step": 8116 }, { "epoch": 1.55, "learning_rate": 2.5385891325071496e-06, "loss": 3.2132, "step": 8117 }, { "epoch": 1.55, "learning_rate": 2.5385319351763584e-06, "loss": 2.8432, "step": 8118 }, { "epoch": 1.55, "learning_rate": 2.5384747378455672e-06, "loss": 3.6857, "step": 8119 }, { "epoch": 1.55, "learning_rate": 2.538417540514776e-06, "loss": 3.4729, "step": 8120 }, { "epoch": 1.55, "learning_rate": 2.538360343183985e-06, "loss": 3.7159, "step": 8121 }, { "epoch": 1.55, "learning_rate": 2.5383031458531936e-06, "loss": 2.6752, "step": 8122 }, { "epoch": 1.55, "learning_rate": 2.5382459485224024e-06, "loss": 2.8174, "step": 8123 }, { "epoch": 1.55, "learning_rate": 2.538188751191611e-06, "loss": 2.8994, "step": 8124 }, { "epoch": 1.55, "learning_rate": 2.53813155386082e-06, "loss": 2.8015, "step": 8125 }, { "epoch": 1.55, "learning_rate": 2.5380743565300284e-06, "loss": 3.9002, "step": 8126 }, { "epoch": 1.55, "learning_rate": 2.5380171591992376e-06, "loss": 2.9258, "step": 8127 }, { "epoch": 1.55, "learning_rate": 2.537959961868446e-06, "loss": 3.5859, "step": 8128 }, { "epoch": 1.55, "learning_rate": 2.537902764537655e-06, "loss": 3.0669, "step": 8129 }, { "epoch": 1.55, "learning_rate": 2.537845567206864e-06, "loss": 2.6377, "step": 8130 }, { "epoch": 1.55, "learning_rate": 2.5377883698760723e-06, "loss": 2.5197, "step": 8131 }, { "epoch": 1.55, "learning_rate": 2.5377311725452816e-06, "loss": 3.0428, "step": 8132 }, { "epoch": 1.55, "learning_rate": 2.53767397521449e-06, "loss": 3.1685, "step": 8133 }, { "epoch": 1.55, "learning_rate": 2.5376167778836987e-06, "loss": 2.667, "step": 8134 }, { "epoch": 1.55, "learning_rate": 2.5375595805529075e-06, "loss": 2.5914, "step": 8135 }, { "epoch": 1.55, "learning_rate": 2.5375023832221163e-06, "loss": 3.463, "step": 8136 }, { "epoch": 1.55, "learning_rate": 2.537445185891325e-06, "loss": 3.8702, "step": 8137 }, { "epoch": 1.55, "learning_rate": 2.537387988560534e-06, "loss": 3.6367, "step": 8138 }, { "epoch": 1.55, "learning_rate": 2.5373307912297427e-06, "loss": 2.9084, "step": 8139 }, { "epoch": 1.55, "learning_rate": 2.5372735938989515e-06, "loss": 3.5449, "step": 8140 }, { "epoch": 1.55, "learning_rate": 2.5372163965681603e-06, "loss": 2.8179, "step": 8141 }, { "epoch": 1.55, "learning_rate": 2.537159199237369e-06, "loss": 2.92, "step": 8142 }, { "epoch": 1.55, "learning_rate": 2.5371020019065774e-06, "loss": 2.4827, "step": 8143 }, { "epoch": 1.55, "learning_rate": 2.5370448045757867e-06, "loss": 3.2012, "step": 8144 }, { "epoch": 1.55, "learning_rate": 2.536987607244995e-06, "loss": 3.8168, "step": 8145 }, { "epoch": 1.55, "learning_rate": 2.5369304099142042e-06, "loss": 2.8545, "step": 8146 }, { "epoch": 1.55, "learning_rate": 2.536873212583413e-06, "loss": 2.5371, "step": 8147 }, { "epoch": 1.55, "learning_rate": 2.5368160152526214e-06, "loss": 3.384, "step": 8148 }, { "epoch": 1.55, "learning_rate": 2.5367588179218306e-06, "loss": 2.8248, "step": 8149 }, { "epoch": 1.55, "learning_rate": 2.536701620591039e-06, "loss": 2.9513, "step": 8150 }, { "epoch": 1.55, "learning_rate": 2.536644423260248e-06, "loss": 2.8996, "step": 8151 }, { "epoch": 1.55, "learning_rate": 2.536587225929457e-06, "loss": 3.2569, "step": 8152 }, { "epoch": 1.55, "learning_rate": 2.5365300285986654e-06, "loss": 3.6654, "step": 8153 }, { "epoch": 1.55, "learning_rate": 2.536472831267874e-06, "loss": 3.2397, "step": 8154 }, { "epoch": 1.55, "learning_rate": 2.536415633937083e-06, "loss": 2.73, "step": 8155 }, { "epoch": 1.55, "learning_rate": 2.5363584366062918e-06, "loss": 2.5496, "step": 8156 }, { "epoch": 1.55, "learning_rate": 2.5363012392755006e-06, "loss": 3.3608, "step": 8157 }, { "epoch": 1.55, "learning_rate": 2.5362440419447093e-06, "loss": 2.3557, "step": 8158 }, { "epoch": 1.55, "learning_rate": 2.536186844613918e-06, "loss": 2.1736, "step": 8159 }, { "epoch": 1.55, "learning_rate": 2.536129647283127e-06, "loss": 4.1234, "step": 8160 }, { "epoch": 1.55, "learning_rate": 2.5360724499523357e-06, "loss": 3.089, "step": 8161 }, { "epoch": 1.55, "learning_rate": 2.5360152526215445e-06, "loss": 3.5236, "step": 8162 }, { "epoch": 1.55, "learning_rate": 2.5359580552907533e-06, "loss": 2.9251, "step": 8163 }, { "epoch": 1.56, "learning_rate": 2.535900857959962e-06, "loss": 3.6957, "step": 8164 }, { "epoch": 1.56, "learning_rate": 2.5358436606291705e-06, "loss": 3.0175, "step": 8165 }, { "epoch": 1.56, "learning_rate": 2.5357864632983797e-06, "loss": 2.5189, "step": 8166 }, { "epoch": 1.56, "learning_rate": 2.535729265967588e-06, "loss": 3.3507, "step": 8167 }, { "epoch": 1.56, "learning_rate": 2.535672068636797e-06, "loss": 2.5282, "step": 8168 }, { "epoch": 1.56, "learning_rate": 2.535614871306006e-06, "loss": 2.5789, "step": 8169 }, { "epoch": 1.56, "learning_rate": 2.5355576739752144e-06, "loss": 2.8134, "step": 8170 }, { "epoch": 1.56, "learning_rate": 2.5355004766444232e-06, "loss": 3.0731, "step": 8171 }, { "epoch": 1.56, "learning_rate": 2.535443279313632e-06, "loss": 3.1078, "step": 8172 }, { "epoch": 1.56, "learning_rate": 2.535386081982841e-06, "loss": 3.1933, "step": 8173 }, { "epoch": 1.56, "learning_rate": 2.5353288846520496e-06, "loss": 3.4773, "step": 8174 }, { "epoch": 1.56, "learning_rate": 2.5352716873212584e-06, "loss": 3.1758, "step": 8175 }, { "epoch": 1.56, "learning_rate": 2.535214489990467e-06, "loss": 3.8676, "step": 8176 }, { "epoch": 1.56, "learning_rate": 2.535157292659676e-06, "loss": 3.5919, "step": 8177 }, { "epoch": 1.56, "learning_rate": 2.535100095328885e-06, "loss": 3.278, "step": 8178 }, { "epoch": 1.56, "learning_rate": 2.5350428979980936e-06, "loss": 3.875, "step": 8179 }, { "epoch": 1.56, "learning_rate": 2.5349857006673024e-06, "loss": 3.0222, "step": 8180 }, { "epoch": 1.56, "learning_rate": 2.534928503336511e-06, "loss": 2.5472, "step": 8181 }, { "epoch": 1.56, "learning_rate": 2.5348713060057195e-06, "loss": 2.974, "step": 8182 }, { "epoch": 1.56, "learning_rate": 2.5348141086749288e-06, "loss": 3.3468, "step": 8183 }, { "epoch": 1.56, "learning_rate": 2.534756911344137e-06, "loss": 3.3314, "step": 8184 }, { "epoch": 1.56, "learning_rate": 2.534699714013346e-06, "loss": 2.8719, "step": 8185 }, { "epoch": 1.56, "learning_rate": 2.534642516682555e-06, "loss": 2.9326, "step": 8186 }, { "epoch": 1.56, "learning_rate": 2.5345853193517635e-06, "loss": 3.0118, "step": 8187 }, { "epoch": 1.56, "learning_rate": 2.5345281220209723e-06, "loss": 2.413, "step": 8188 }, { "epoch": 1.56, "learning_rate": 2.534470924690181e-06, "loss": 3.6622, "step": 8189 }, { "epoch": 1.56, "learning_rate": 2.53441372735939e-06, "loss": 3.5697, "step": 8190 }, { "epoch": 1.56, "learning_rate": 2.5343565300285987e-06, "loss": 3.4854, "step": 8191 }, { "epoch": 1.56, "learning_rate": 2.5342993326978075e-06, "loss": 2.3623, "step": 8192 }, { "epoch": 1.56, "learning_rate": 2.5342421353670163e-06, "loss": 3.3403, "step": 8193 }, { "epoch": 1.56, "learning_rate": 2.534184938036225e-06, "loss": 3.41, "step": 8194 }, { "epoch": 1.56, "learning_rate": 2.534127740705434e-06, "loss": 3.2078, "step": 8195 }, { "epoch": 1.56, "learning_rate": 2.5340705433746427e-06, "loss": 3.3161, "step": 8196 }, { "epoch": 1.56, "learning_rate": 2.5340133460438515e-06, "loss": 2.5267, "step": 8197 }, { "epoch": 1.56, "learning_rate": 2.5339561487130603e-06, "loss": 3.4387, "step": 8198 }, { "epoch": 1.56, "learning_rate": 2.5338989513822686e-06, "loss": 3.4515, "step": 8199 }, { "epoch": 1.56, "learning_rate": 2.533841754051478e-06, "loss": 2.7483, "step": 8200 }, { "epoch": 1.56, "learning_rate": 2.5337845567206866e-06, "loss": 3.8267, "step": 8201 }, { "epoch": 1.56, "learning_rate": 2.533727359389895e-06, "loss": 3.5844, "step": 8202 }, { "epoch": 1.56, "learning_rate": 2.5336701620591042e-06, "loss": 3.01, "step": 8203 }, { "epoch": 1.56, "learning_rate": 2.5336129647283126e-06, "loss": 2.888, "step": 8204 }, { "epoch": 1.56, "learning_rate": 2.5335557673975214e-06, "loss": 3.8079, "step": 8205 }, { "epoch": 1.56, "learning_rate": 2.53349857006673e-06, "loss": 2.959, "step": 8206 }, { "epoch": 1.56, "learning_rate": 2.533441372735939e-06, "loss": 3.1174, "step": 8207 }, { "epoch": 1.56, "learning_rate": 2.5333841754051478e-06, "loss": 2.3299, "step": 8208 }, { "epoch": 1.56, "learning_rate": 2.5333269780743566e-06, "loss": 2.6391, "step": 8209 }, { "epoch": 1.56, "learning_rate": 2.5332697807435654e-06, "loss": 3.558, "step": 8210 }, { "epoch": 1.56, "learning_rate": 2.533212583412774e-06, "loss": 3.0474, "step": 8211 }, { "epoch": 1.56, "learning_rate": 2.533155386081983e-06, "loss": 2.6255, "step": 8212 }, { "epoch": 1.56, "learning_rate": 2.5330981887511917e-06, "loss": 2.6392, "step": 8213 }, { "epoch": 1.56, "learning_rate": 2.5330409914204005e-06, "loss": 2.8397, "step": 8214 }, { "epoch": 1.56, "learning_rate": 2.5329837940896093e-06, "loss": 3.3999, "step": 8215 }, { "epoch": 1.56, "learning_rate": 2.5329265967588177e-06, "loss": 2.8793, "step": 8216 }, { "epoch": 1.57, "learning_rate": 2.532869399428027e-06, "loss": 2.3848, "step": 8217 }, { "epoch": 1.57, "learning_rate": 2.5328122020972357e-06, "loss": 2.6116, "step": 8218 }, { "epoch": 1.57, "learning_rate": 2.532755004766444e-06, "loss": 2.6865, "step": 8219 }, { "epoch": 1.57, "learning_rate": 2.5326978074356533e-06, "loss": 2.3298, "step": 8220 }, { "epoch": 1.57, "learning_rate": 2.5326406101048617e-06, "loss": 2.3906, "step": 8221 }, { "epoch": 1.57, "learning_rate": 2.5325834127740705e-06, "loss": 3.2244, "step": 8222 }, { "epoch": 1.57, "learning_rate": 2.5325262154432792e-06, "loss": 3.3855, "step": 8223 }, { "epoch": 1.57, "learning_rate": 2.532469018112488e-06, "loss": 3.574, "step": 8224 }, { "epoch": 1.57, "learning_rate": 2.5324118207816973e-06, "loss": 3.8182, "step": 8225 }, { "epoch": 1.57, "learning_rate": 2.5323546234509056e-06, "loss": 3.1109, "step": 8226 }, { "epoch": 1.57, "learning_rate": 2.5322974261201144e-06, "loss": 3.266, "step": 8227 }, { "epoch": 1.57, "learning_rate": 2.5322402287893232e-06, "loss": 3.375, "step": 8228 }, { "epoch": 1.57, "learning_rate": 2.532183031458532e-06, "loss": 3.029, "step": 8229 }, { "epoch": 1.57, "learning_rate": 2.532125834127741e-06, "loss": 2.8137, "step": 8230 }, { "epoch": 1.57, "learning_rate": 2.5320686367969496e-06, "loss": 3.4451, "step": 8231 }, { "epoch": 1.57, "learning_rate": 2.5320114394661584e-06, "loss": 3.0868, "step": 8232 }, { "epoch": 1.57, "learning_rate": 2.5319542421353668e-06, "loss": 3.6338, "step": 8233 }, { "epoch": 1.57, "learning_rate": 2.531897044804576e-06, "loss": 2.3981, "step": 8234 }, { "epoch": 1.57, "learning_rate": 2.5318398474737848e-06, "loss": 2.7208, "step": 8235 }, { "epoch": 1.57, "learning_rate": 2.531782650142993e-06, "loss": 2.8743, "step": 8236 }, { "epoch": 1.57, "learning_rate": 2.5317254528122024e-06, "loss": 2.3419, "step": 8237 }, { "epoch": 1.57, "learning_rate": 2.5316682554814107e-06, "loss": 3.4184, "step": 8238 }, { "epoch": 1.57, "learning_rate": 2.5316110581506195e-06, "loss": 3.026, "step": 8239 }, { "epoch": 1.57, "learning_rate": 2.5315538608198287e-06, "loss": 2.8771, "step": 8240 }, { "epoch": 1.57, "learning_rate": 2.531496663489037e-06, "loss": 2.9187, "step": 8241 }, { "epoch": 1.57, "learning_rate": 2.5314394661582463e-06, "loss": 3.7618, "step": 8242 }, { "epoch": 1.57, "learning_rate": 2.5313822688274547e-06, "loss": 2.8379, "step": 8243 }, { "epoch": 1.57, "learning_rate": 2.5313250714966635e-06, "loss": 2.9894, "step": 8244 }, { "epoch": 1.57, "learning_rate": 2.5312678741658723e-06, "loss": 2.5345, "step": 8245 }, { "epoch": 1.57, "learning_rate": 2.531210676835081e-06, "loss": 2.8168, "step": 8246 }, { "epoch": 1.57, "learning_rate": 2.53115347950429e-06, "loss": 2.9012, "step": 8247 }, { "epoch": 1.57, "learning_rate": 2.5310962821734987e-06, "loss": 3.3756, "step": 8248 }, { "epoch": 1.57, "learning_rate": 2.5310390848427075e-06, "loss": 2.8987, "step": 8249 }, { "epoch": 1.57, "learning_rate": 2.5309818875119163e-06, "loss": 3.1955, "step": 8250 }, { "epoch": 1.57, "learning_rate": 2.530924690181125e-06, "loss": 2.5664, "step": 8251 }, { "epoch": 1.57, "learning_rate": 2.530867492850334e-06, "loss": 3.2257, "step": 8252 }, { "epoch": 1.57, "learning_rate": 2.5308102955195422e-06, "loss": 3.008, "step": 8253 }, { "epoch": 1.57, "learning_rate": 2.5307530981887514e-06, "loss": 3.2442, "step": 8254 }, { "epoch": 1.57, "learning_rate": 2.53069590085796e-06, "loss": 3.092, "step": 8255 }, { "epoch": 1.57, "learning_rate": 2.5306387035271686e-06, "loss": 2.5471, "step": 8256 }, { "epoch": 1.57, "learning_rate": 2.530581506196378e-06, "loss": 2.6362, "step": 8257 }, { "epoch": 1.57, "learning_rate": 2.530524308865586e-06, "loss": 2.5582, "step": 8258 }, { "epoch": 1.57, "learning_rate": 2.5304671115347954e-06, "loss": 2.6623, "step": 8259 }, { "epoch": 1.57, "learning_rate": 2.5304099142040038e-06, "loss": 3.3719, "step": 8260 }, { "epoch": 1.57, "learning_rate": 2.5303527168732126e-06, "loss": 2.7499, "step": 8261 }, { "epoch": 1.57, "learning_rate": 2.5302955195424214e-06, "loss": 3.7132, "step": 8262 }, { "epoch": 1.57, "learning_rate": 2.53023832221163e-06, "loss": 3.3691, "step": 8263 }, { "epoch": 1.57, "learning_rate": 2.530181124880839e-06, "loss": 2.6728, "step": 8264 }, { "epoch": 1.57, "learning_rate": 2.5301239275500477e-06, "loss": 3.0924, "step": 8265 }, { "epoch": 1.57, "learning_rate": 2.5300667302192565e-06, "loss": 3.0218, "step": 8266 }, { "epoch": 1.57, "learning_rate": 2.5300095328884653e-06, "loss": 2.889, "step": 8267 }, { "epoch": 1.57, "learning_rate": 2.529952335557674e-06, "loss": 2.6341, "step": 8268 }, { "epoch": 1.58, "learning_rate": 2.529895138226883e-06, "loss": 3.3862, "step": 8269 }, { "epoch": 1.58, "learning_rate": 2.5298379408960913e-06, "loss": 3.4941, "step": 8270 }, { "epoch": 1.58, "learning_rate": 2.5297807435653005e-06, "loss": 3.5947, "step": 8271 }, { "epoch": 1.58, "learning_rate": 2.529723546234509e-06, "loss": 3.0656, "step": 8272 }, { "epoch": 1.58, "learning_rate": 2.5296663489037177e-06, "loss": 3.4006, "step": 8273 }, { "epoch": 1.58, "learning_rate": 2.529609151572927e-06, "loss": 3.5107, "step": 8274 }, { "epoch": 1.58, "learning_rate": 2.5295519542421353e-06, "loss": 2.6348, "step": 8275 }, { "epoch": 1.58, "learning_rate": 2.5294947569113445e-06, "loss": 2.5075, "step": 8276 }, { "epoch": 1.58, "learning_rate": 2.529437559580553e-06, "loss": 3.4716, "step": 8277 }, { "epoch": 1.58, "learning_rate": 2.5293803622497616e-06, "loss": 3.0663, "step": 8278 }, { "epoch": 1.58, "learning_rate": 2.5293231649189704e-06, "loss": 3.0155, "step": 8279 }, { "epoch": 1.58, "learning_rate": 2.5292659675881792e-06, "loss": 2.6161, "step": 8280 }, { "epoch": 1.58, "learning_rate": 2.529208770257388e-06, "loss": 3.1762, "step": 8281 }, { "epoch": 1.58, "learning_rate": 2.529151572926597e-06, "loss": 3.3783, "step": 8282 }, { "epoch": 1.58, "learning_rate": 2.5290943755958056e-06, "loss": 2.8913, "step": 8283 }, { "epoch": 1.58, "learning_rate": 2.5290371782650144e-06, "loss": 2.9389, "step": 8284 }, { "epoch": 1.58, "learning_rate": 2.528979980934223e-06, "loss": 2.8146, "step": 8285 }, { "epoch": 1.58, "learning_rate": 2.528922783603432e-06, "loss": 3.4012, "step": 8286 }, { "epoch": 1.58, "learning_rate": 2.5288655862726404e-06, "loss": 3.5249, "step": 8287 }, { "epoch": 1.58, "learning_rate": 2.5288083889418496e-06, "loss": 3.1567, "step": 8288 }, { "epoch": 1.58, "learning_rate": 2.5287511916110584e-06, "loss": 2.9774, "step": 8289 }, { "epoch": 1.58, "learning_rate": 2.5286939942802667e-06, "loss": 3.2839, "step": 8290 }, { "epoch": 1.58, "learning_rate": 2.528636796949476e-06, "loss": 3.1383, "step": 8291 }, { "epoch": 1.58, "learning_rate": 2.5285795996186843e-06, "loss": 3.5166, "step": 8292 }, { "epoch": 1.58, "learning_rate": 2.5285224022878935e-06, "loss": 3.4772, "step": 8293 }, { "epoch": 1.58, "learning_rate": 2.528465204957102e-06, "loss": 2.9041, "step": 8294 }, { "epoch": 1.58, "learning_rate": 2.5284080076263107e-06, "loss": 2.8253, "step": 8295 }, { "epoch": 1.58, "learning_rate": 2.52835081029552e-06, "loss": 2.7225, "step": 8296 }, { "epoch": 1.58, "learning_rate": 2.5282936129647283e-06, "loss": 3.2823, "step": 8297 }, { "epoch": 1.58, "learning_rate": 2.528236415633937e-06, "loss": 3.0937, "step": 8298 }, { "epoch": 1.58, "learning_rate": 2.528179218303146e-06, "loss": 3.1774, "step": 8299 }, { "epoch": 1.58, "learning_rate": 2.5281220209723547e-06, "loss": 2.6835, "step": 8300 }, { "epoch": 1.58, "learning_rate": 2.5280648236415635e-06, "loss": 3.3677, "step": 8301 }, { "epoch": 1.58, "learning_rate": 2.5280076263107723e-06, "loss": 3.3536, "step": 8302 }, { "epoch": 1.58, "learning_rate": 2.527950428979981e-06, "loss": 2.6953, "step": 8303 }, { "epoch": 1.58, "learning_rate": 2.5278932316491894e-06, "loss": 3.4699, "step": 8304 }, { "epoch": 1.58, "learning_rate": 2.5278360343183986e-06, "loss": 3.0367, "step": 8305 }, { "epoch": 1.58, "learning_rate": 2.5277788369876074e-06, "loss": 3.9206, "step": 8306 }, { "epoch": 1.58, "learning_rate": 2.5277216396568162e-06, "loss": 2.6633, "step": 8307 }, { "epoch": 1.58, "learning_rate": 2.527664442326025e-06, "loss": 3.4318, "step": 8308 }, { "epoch": 1.58, "learning_rate": 2.5276072449952334e-06, "loss": 2.7198, "step": 8309 }, { "epoch": 1.58, "learning_rate": 2.5275500476644426e-06, "loss": 2.8058, "step": 8310 }, { "epoch": 1.58, "learning_rate": 2.527492850333651e-06, "loss": 2.9155, "step": 8311 }, { "epoch": 1.58, "learning_rate": 2.5274356530028598e-06, "loss": 3.1296, "step": 8312 }, { "epoch": 1.58, "learning_rate": 2.527378455672069e-06, "loss": 3.2311, "step": 8313 }, { "epoch": 1.58, "learning_rate": 2.5273212583412774e-06, "loss": 3.2828, "step": 8314 }, { "epoch": 1.58, "learning_rate": 2.527264061010486e-06, "loss": 2.7483, "step": 8315 }, { "epoch": 1.58, "learning_rate": 2.527206863679695e-06, "loss": 3.0121, "step": 8316 }, { "epoch": 1.58, "learning_rate": 2.5271496663489038e-06, "loss": 3.4963, "step": 8317 }, { "epoch": 1.58, "learning_rate": 2.5270924690181125e-06, "loss": 3.0508, "step": 8318 }, { "epoch": 1.58, "learning_rate": 2.5270352716873213e-06, "loss": 2.5289, "step": 8319 }, { "epoch": 1.58, "learning_rate": 2.52697807435653e-06, "loss": 2.925, "step": 8320 }, { "epoch": 1.58, "learning_rate": 2.5269208770257385e-06, "loss": 2.9136, "step": 8321 }, { "epoch": 1.59, "learning_rate": 2.5268636796949477e-06, "loss": 3.2654, "step": 8322 }, { "epoch": 1.59, "learning_rate": 2.5268064823641565e-06, "loss": 3.3518, "step": 8323 }, { "epoch": 1.59, "learning_rate": 2.5267492850333653e-06, "loss": 3.012, "step": 8324 }, { "epoch": 1.59, "learning_rate": 2.526692087702574e-06, "loss": 3.9475, "step": 8325 }, { "epoch": 1.59, "learning_rate": 2.5266348903717825e-06, "loss": 2.6625, "step": 8326 }, { "epoch": 1.59, "learning_rate": 2.5265776930409917e-06, "loss": 3.3744, "step": 8327 }, { "epoch": 1.59, "learning_rate": 2.5265204957102005e-06, "loss": 3.0472, "step": 8328 }, { "epoch": 1.59, "learning_rate": 2.526463298379409e-06, "loss": 2.9545, "step": 8329 }, { "epoch": 1.59, "learning_rate": 2.526406101048618e-06, "loss": 3.7165, "step": 8330 }, { "epoch": 1.59, "learning_rate": 2.5263489037178264e-06, "loss": 2.9094, "step": 8331 }, { "epoch": 1.59, "learning_rate": 2.5262917063870352e-06, "loss": 2.7147, "step": 8332 }, { "epoch": 1.59, "learning_rate": 2.526234509056244e-06, "loss": 3.4024, "step": 8333 }, { "epoch": 1.59, "learning_rate": 2.526177311725453e-06, "loss": 3.5374, "step": 8334 }, { "epoch": 1.59, "learning_rate": 2.5261201143946616e-06, "loss": 3.6539, "step": 8335 }, { "epoch": 1.59, "learning_rate": 2.5260629170638704e-06, "loss": 3.4613, "step": 8336 }, { "epoch": 1.59, "learning_rate": 2.526005719733079e-06, "loss": 2.9779, "step": 8337 }, { "epoch": 1.59, "learning_rate": 2.525948522402288e-06, "loss": 2.6035, "step": 8338 }, { "epoch": 1.59, "learning_rate": 2.525891325071497e-06, "loss": 2.9257, "step": 8339 }, { "epoch": 1.59, "learning_rate": 2.5258341277407056e-06, "loss": 3.0671, "step": 8340 }, { "epoch": 1.59, "learning_rate": 2.5257769304099144e-06, "loss": 2.8341, "step": 8341 }, { "epoch": 1.59, "learning_rate": 2.525719733079123e-06, "loss": 3.5808, "step": 8342 }, { "epoch": 1.59, "learning_rate": 2.5256625357483315e-06, "loss": 3.2655, "step": 8343 }, { "epoch": 1.59, "learning_rate": 2.5256053384175408e-06, "loss": 3.1576, "step": 8344 }, { "epoch": 1.59, "learning_rate": 2.5255481410867496e-06, "loss": 2.3976, "step": 8345 }, { "epoch": 1.59, "learning_rate": 2.525490943755958e-06, "loss": 2.9363, "step": 8346 }, { "epoch": 1.59, "learning_rate": 2.525433746425167e-06, "loss": 3.119, "step": 8347 }, { "epoch": 1.59, "learning_rate": 2.5253765490943755e-06, "loss": 3.1837, "step": 8348 }, { "epoch": 1.59, "learning_rate": 2.5253193517635843e-06, "loss": 2.7714, "step": 8349 }, { "epoch": 1.59, "learning_rate": 2.525262154432793e-06, "loss": 2.7682, "step": 8350 }, { "epoch": 1.59, "learning_rate": 2.525204957102002e-06, "loss": 3.2791, "step": 8351 }, { "epoch": 1.59, "learning_rate": 2.5251477597712107e-06, "loss": 3.6052, "step": 8352 }, { "epoch": 1.59, "learning_rate": 2.5250905624404195e-06, "loss": 3.1075, "step": 8353 }, { "epoch": 1.59, "learning_rate": 2.5250333651096283e-06, "loss": 3.2116, "step": 8354 }, { "epoch": 1.59, "learning_rate": 2.524976167778837e-06, "loss": 2.9482, "step": 8355 }, { "epoch": 1.59, "learning_rate": 2.524918970448046e-06, "loss": 3.579, "step": 8356 }, { "epoch": 1.59, "learning_rate": 2.5248617731172547e-06, "loss": 2.1088, "step": 8357 }, { "epoch": 1.59, "learning_rate": 2.5248045757864634e-06, "loss": 2.2632, "step": 8358 }, { "epoch": 1.59, "learning_rate": 2.5247473784556722e-06, "loss": 3.5357, "step": 8359 }, { "epoch": 1.59, "learning_rate": 2.5246901811248806e-06, "loss": 2.2159, "step": 8360 }, { "epoch": 1.59, "learning_rate": 2.52463298379409e-06, "loss": 3.5552, "step": 8361 }, { "epoch": 1.59, "learning_rate": 2.5245757864632986e-06, "loss": 3.3389, "step": 8362 }, { "epoch": 1.59, "learning_rate": 2.524518589132507e-06, "loss": 3.4569, "step": 8363 }, { "epoch": 1.59, "learning_rate": 2.5244613918017162e-06, "loss": 2.8754, "step": 8364 }, { "epoch": 1.59, "learning_rate": 2.5244041944709246e-06, "loss": 3.2587, "step": 8365 }, { "epoch": 1.59, "learning_rate": 2.5243469971401334e-06, "loss": 3.7482, "step": 8366 }, { "epoch": 1.59, "learning_rate": 2.524289799809342e-06, "loss": 3.1043, "step": 8367 }, { "epoch": 1.59, "learning_rate": 2.524232602478551e-06, "loss": 2.5847, "step": 8368 }, { "epoch": 1.59, "learning_rate": 2.5241754051477598e-06, "loss": 3.3281, "step": 8369 }, { "epoch": 1.59, "learning_rate": 2.5241182078169686e-06, "loss": 3.7441, "step": 8370 }, { "epoch": 1.59, "learning_rate": 2.5240610104861773e-06, "loss": 3.056, "step": 8371 }, { "epoch": 1.59, "learning_rate": 2.524003813155386e-06, "loss": 3.1502, "step": 8372 }, { "epoch": 1.59, "learning_rate": 2.523946615824595e-06, "loss": 3.2558, "step": 8373 }, { "epoch": 1.6, "learning_rate": 2.5238894184938037e-06, "loss": 3.364, "step": 8374 }, { "epoch": 1.6, "learning_rate": 2.5238322211630125e-06, "loss": 3.385, "step": 8375 }, { "epoch": 1.6, "learning_rate": 2.5237750238322213e-06, "loss": 3.2905, "step": 8376 }, { "epoch": 1.6, "learning_rate": 2.52371782650143e-06, "loss": 3.0032, "step": 8377 }, { "epoch": 1.6, "learning_rate": 2.523660629170639e-06, "loss": 3.621, "step": 8378 }, { "epoch": 1.6, "learning_rate": 2.5236034318398477e-06, "loss": 2.2281, "step": 8379 }, { "epoch": 1.6, "learning_rate": 2.523546234509056e-06, "loss": 2.7904, "step": 8380 }, { "epoch": 1.6, "learning_rate": 2.5234890371782653e-06, "loss": 3.1781, "step": 8381 }, { "epoch": 1.6, "learning_rate": 2.5234318398474737e-06, "loss": 2.6649, "step": 8382 }, { "epoch": 1.6, "learning_rate": 2.5233746425166824e-06, "loss": 2.8264, "step": 8383 }, { "epoch": 1.6, "learning_rate": 2.5233174451858917e-06, "loss": 3.0991, "step": 8384 }, { "epoch": 1.6, "learning_rate": 2.5232602478551e-06, "loss": 2.8187, "step": 8385 }, { "epoch": 1.6, "learning_rate": 2.523203050524309e-06, "loss": 3.1469, "step": 8386 }, { "epoch": 1.6, "learning_rate": 2.5231458531935176e-06, "loss": 3.6056, "step": 8387 }, { "epoch": 1.6, "learning_rate": 2.5230886558627264e-06, "loss": 1.9762, "step": 8388 }, { "epoch": 1.6, "learning_rate": 2.523031458531935e-06, "loss": 3.2378, "step": 8389 }, { "epoch": 1.6, "learning_rate": 2.522974261201144e-06, "loss": 3.2088, "step": 8390 }, { "epoch": 1.6, "learning_rate": 2.522917063870353e-06, "loss": 3.2396, "step": 8391 }, { "epoch": 1.6, "learning_rate": 2.5228598665395616e-06, "loss": 2.7771, "step": 8392 }, { "epoch": 1.6, "learning_rate": 2.5228026692087704e-06, "loss": 2.7768, "step": 8393 }, { "epoch": 1.6, "learning_rate": 2.522745471877979e-06, "loss": 2.8714, "step": 8394 }, { "epoch": 1.6, "learning_rate": 2.522688274547188e-06, "loss": 3.5336, "step": 8395 }, { "epoch": 1.6, "learning_rate": 2.5226310772163968e-06, "loss": 3.7142, "step": 8396 }, { "epoch": 1.6, "learning_rate": 2.522573879885605e-06, "loss": 3.2978, "step": 8397 }, { "epoch": 1.6, "learning_rate": 2.5225166825548144e-06, "loss": 3.3046, "step": 8398 }, { "epoch": 1.6, "learning_rate": 2.5224594852240227e-06, "loss": 2.6845, "step": 8399 }, { "epoch": 1.6, "learning_rate": 2.5224022878932315e-06, "loss": 3.4452, "step": 8400 }, { "epoch": 1.6, "learning_rate": 2.5223450905624407e-06, "loss": 2.7265, "step": 8401 }, { "epoch": 1.6, "learning_rate": 2.522287893231649e-06, "loss": 3.8131, "step": 8402 }, { "epoch": 1.6, "learning_rate": 2.522230695900858e-06, "loss": 2.7529, "step": 8403 }, { "epoch": 1.6, "learning_rate": 2.5221734985700667e-06, "loss": 3.1, "step": 8404 }, { "epoch": 1.6, "learning_rate": 2.5221163012392755e-06, "loss": 2.8366, "step": 8405 }, { "epoch": 1.6, "learning_rate": 2.5220591039084843e-06, "loss": 2.9272, "step": 8406 }, { "epoch": 1.6, "learning_rate": 2.522001906577693e-06, "loss": 2.6154, "step": 8407 }, { "epoch": 1.6, "learning_rate": 2.521944709246902e-06, "loss": 3.6864, "step": 8408 }, { "epoch": 1.6, "learning_rate": 2.5218875119161107e-06, "loss": 3.4943, "step": 8409 }, { "epoch": 1.6, "learning_rate": 2.5218303145853195e-06, "loss": 2.9668, "step": 8410 }, { "epoch": 1.6, "learning_rate": 2.5217731172545283e-06, "loss": 3.1188, "step": 8411 }, { "epoch": 1.6, "learning_rate": 2.521715919923737e-06, "loss": 2.9887, "step": 8412 }, { "epoch": 1.6, "learning_rate": 2.521658722592946e-06, "loss": 3.53, "step": 8413 }, { "epoch": 1.6, "learning_rate": 2.521601525262154e-06, "loss": 2.5133, "step": 8414 }, { "epoch": 1.6, "learning_rate": 2.5215443279313634e-06, "loss": 2.9435, "step": 8415 }, { "epoch": 1.6, "learning_rate": 2.5214871306005722e-06, "loss": 2.6072, "step": 8416 }, { "epoch": 1.6, "learning_rate": 2.5214299332697806e-06, "loss": 2.9669, "step": 8417 }, { "epoch": 1.6, "learning_rate": 2.52137273593899e-06, "loss": 3.3072, "step": 8418 }, { "epoch": 1.6, "learning_rate": 2.521315538608198e-06, "loss": 3.0508, "step": 8419 }, { "epoch": 1.6, "learning_rate": 2.521258341277407e-06, "loss": 3.6592, "step": 8420 }, { "epoch": 1.6, "learning_rate": 2.5212011439466158e-06, "loss": 2.4054, "step": 8421 }, { "epoch": 1.6, "learning_rate": 2.5211439466158246e-06, "loss": 2.7334, "step": 8422 }, { "epoch": 1.6, "learning_rate": 2.5210867492850338e-06, "loss": 3.0663, "step": 8423 }, { "epoch": 1.6, "learning_rate": 2.521029551954242e-06, "loss": 2.9971, "step": 8424 }, { "epoch": 1.6, "learning_rate": 2.520972354623451e-06, "loss": 3.4098, "step": 8425 }, { "epoch": 1.6, "learning_rate": 2.5209151572926597e-06, "loss": 3.2645, "step": 8426 }, { "epoch": 1.61, "learning_rate": 2.5208579599618685e-06, "loss": 2.575, "step": 8427 }, { "epoch": 1.61, "learning_rate": 2.5208007626310773e-06, "loss": 2.925, "step": 8428 }, { "epoch": 1.61, "learning_rate": 2.520743565300286e-06, "loss": 2.5369, "step": 8429 }, { "epoch": 1.61, "learning_rate": 2.520686367969495e-06, "loss": 2.9991, "step": 8430 }, { "epoch": 1.61, "learning_rate": 2.5206291706387033e-06, "loss": 2.774, "step": 8431 }, { "epoch": 1.61, "learning_rate": 2.5205719733079125e-06, "loss": 2.6204, "step": 8432 }, { "epoch": 1.61, "learning_rate": 2.5205147759771213e-06, "loss": 3.6251, "step": 8433 }, { "epoch": 1.61, "learning_rate": 2.5204575786463297e-06, "loss": 3.0704, "step": 8434 }, { "epoch": 1.61, "learning_rate": 2.520400381315539e-06, "loss": 3.1501, "step": 8435 }, { "epoch": 1.61, "learning_rate": 2.5203431839847473e-06, "loss": 2.6366, "step": 8436 }, { "epoch": 1.61, "learning_rate": 2.5202859866539565e-06, "loss": 3.2527, "step": 8437 }, { "epoch": 1.61, "learning_rate": 2.520228789323165e-06, "loss": 2.4669, "step": 8438 }, { "epoch": 1.61, "learning_rate": 2.5201715919923736e-06, "loss": 2.0844, "step": 8439 }, { "epoch": 1.61, "learning_rate": 2.520114394661583e-06, "loss": 2.9793, "step": 8440 }, { "epoch": 1.61, "learning_rate": 2.5200571973307912e-06, "loss": 2.2483, "step": 8441 }, { "epoch": 1.61, "learning_rate": 2.52e-06, "loss": 3.6501, "step": 8442 }, { "epoch": 1.61, "learning_rate": 2.519942802669209e-06, "loss": 3.5857, "step": 8443 }, { "epoch": 1.61, "learning_rate": 2.5198856053384176e-06, "loss": 3.8338, "step": 8444 }, { "epoch": 1.61, "learning_rate": 2.5198284080076264e-06, "loss": 3.8648, "step": 8445 }, { "epoch": 1.61, "learning_rate": 2.519771210676835e-06, "loss": 3.7324, "step": 8446 }, { "epoch": 1.61, "learning_rate": 2.519714013346044e-06, "loss": 3.557, "step": 8447 }, { "epoch": 1.61, "learning_rate": 2.5196568160152524e-06, "loss": 4.1013, "step": 8448 }, { "epoch": 1.61, "learning_rate": 2.5195996186844616e-06, "loss": 2.7196, "step": 8449 }, { "epoch": 1.61, "learning_rate": 2.5195424213536704e-06, "loss": 3.2237, "step": 8450 }, { "epoch": 1.61, "learning_rate": 2.5194852240228787e-06, "loss": 2.9501, "step": 8451 }, { "epoch": 1.61, "learning_rate": 2.519428026692088e-06, "loss": 3.0815, "step": 8452 }, { "epoch": 1.61, "learning_rate": 2.5193708293612963e-06, "loss": 2.9623, "step": 8453 }, { "epoch": 1.61, "learning_rate": 2.5193136320305055e-06, "loss": 3.4753, "step": 8454 }, { "epoch": 1.61, "learning_rate": 2.519256434699714e-06, "loss": 2.8035, "step": 8455 }, { "epoch": 1.61, "learning_rate": 2.5191992373689227e-06, "loss": 2.8206, "step": 8456 }, { "epoch": 1.61, "learning_rate": 2.519142040038132e-06, "loss": 3.1769, "step": 8457 }, { "epoch": 1.61, "learning_rate": 2.5190848427073403e-06, "loss": 3.4938, "step": 8458 }, { "epoch": 1.61, "learning_rate": 2.519027645376549e-06, "loss": 2.7086, "step": 8459 }, { "epoch": 1.61, "learning_rate": 2.518970448045758e-06, "loss": 3.8669, "step": 8460 }, { "epoch": 1.61, "learning_rate": 2.5189132507149667e-06, "loss": 3.0561, "step": 8461 }, { "epoch": 1.61, "learning_rate": 2.5188560533841755e-06, "loss": 3.3599, "step": 8462 }, { "epoch": 1.61, "learning_rate": 2.5187988560533843e-06, "loss": 2.8626, "step": 8463 }, { "epoch": 1.61, "learning_rate": 2.518741658722593e-06, "loss": 2.9861, "step": 8464 }, { "epoch": 1.61, "learning_rate": 2.518684461391802e-06, "loss": 3.5302, "step": 8465 }, { "epoch": 1.61, "learning_rate": 2.5186272640610106e-06, "loss": 3.7999, "step": 8466 }, { "epoch": 1.61, "learning_rate": 2.5185700667302194e-06, "loss": 3.7257, "step": 8467 }, { "epoch": 1.61, "learning_rate": 2.518512869399428e-06, "loss": 3.1429, "step": 8468 }, { "epoch": 1.61, "learning_rate": 2.518455672068637e-06, "loss": 2.8093, "step": 8469 }, { "epoch": 1.61, "learning_rate": 2.5183984747378454e-06, "loss": 3.5283, "step": 8470 }, { "epoch": 1.61, "learning_rate": 2.5183412774070546e-06, "loss": 2.5605, "step": 8471 }, { "epoch": 1.61, "learning_rate": 2.5182840800762634e-06, "loss": 2.7997, "step": 8472 }, { "epoch": 1.61, "learning_rate": 2.5182268827454718e-06, "loss": 2.4761, "step": 8473 }, { "epoch": 1.61, "learning_rate": 2.518169685414681e-06, "loss": 2.4745, "step": 8474 }, { "epoch": 1.61, "learning_rate": 2.5181124880838894e-06, "loss": 2.9776, "step": 8475 }, { "epoch": 1.61, "learning_rate": 2.518055290753098e-06, "loss": 3.2793, "step": 8476 }, { "epoch": 1.61, "learning_rate": 2.517998093422307e-06, "loss": 3.1856, "step": 8477 }, { "epoch": 1.61, "learning_rate": 2.5179408960915157e-06, "loss": 3.6249, "step": 8478 }, { "epoch": 1.62, "learning_rate": 2.5178836987607245e-06, "loss": 3.3875, "step": 8479 }, { "epoch": 1.62, "learning_rate": 2.5178265014299333e-06, "loss": 2.7964, "step": 8480 }, { "epoch": 1.62, "learning_rate": 2.517769304099142e-06, "loss": 2.5359, "step": 8481 }, { "epoch": 1.62, "learning_rate": 2.517712106768351e-06, "loss": 3.4516, "step": 8482 }, { "epoch": 1.62, "learning_rate": 2.5176549094375597e-06, "loss": 2.4881, "step": 8483 }, { "epoch": 1.62, "learning_rate": 2.5175977121067685e-06, "loss": 2.9221, "step": 8484 }, { "epoch": 1.62, "learning_rate": 2.517540514775977e-06, "loss": 2.9798, "step": 8485 }, { "epoch": 1.62, "learning_rate": 2.517483317445186e-06, "loss": 3.0003, "step": 8486 }, { "epoch": 1.62, "learning_rate": 2.5174261201143945e-06, "loss": 2.6792, "step": 8487 }, { "epoch": 1.62, "learning_rate": 2.5173689227836037e-06, "loss": 3.3914, "step": 8488 }, { "epoch": 1.62, "learning_rate": 2.5173117254528125e-06, "loss": 2.8018, "step": 8489 }, { "epoch": 1.62, "learning_rate": 2.517254528122021e-06, "loss": 2.6128, "step": 8490 }, { "epoch": 1.62, "learning_rate": 2.51719733079123e-06, "loss": 3.793, "step": 8491 }, { "epoch": 1.62, "learning_rate": 2.5171401334604384e-06, "loss": 3.4162, "step": 8492 }, { "epoch": 1.62, "learning_rate": 2.5170829361296472e-06, "loss": 2.7324, "step": 8493 }, { "epoch": 1.62, "learning_rate": 2.517025738798856e-06, "loss": 2.9034, "step": 8494 }, { "epoch": 1.62, "learning_rate": 2.516968541468065e-06, "loss": 3.5625, "step": 8495 }, { "epoch": 1.62, "learning_rate": 2.5169113441372736e-06, "loss": 3.0329, "step": 8496 }, { "epoch": 1.62, "learning_rate": 2.5168541468064824e-06, "loss": 3.4718, "step": 8497 }, { "epoch": 1.62, "learning_rate": 2.516796949475691e-06, "loss": 3.0671, "step": 8498 }, { "epoch": 1.62, "learning_rate": 2.5167397521449e-06, "loss": 2.9179, "step": 8499 }, { "epoch": 1.62, "learning_rate": 2.5166825548141088e-06, "loss": 3.1249, "step": 8500 }, { "epoch": 1.62, "learning_rate": 2.5166253574833176e-06, "loss": 3.1597, "step": 8501 }, { "epoch": 1.62, "learning_rate": 2.5165681601525264e-06, "loss": 2.8168, "step": 8502 }, { "epoch": 1.62, "learning_rate": 2.516510962821735e-06, "loss": 2.7005, "step": 8503 }, { "epoch": 1.62, "learning_rate": 2.516453765490944e-06, "loss": 2.8002, "step": 8504 }, { "epoch": 1.62, "learning_rate": 2.5163965681601528e-06, "loss": 2.302, "step": 8505 }, { "epoch": 1.62, "learning_rate": 2.5163393708293615e-06, "loss": 3.5101, "step": 8506 }, { "epoch": 1.62, "learning_rate": 2.51628217349857e-06, "loss": 2.3733, "step": 8507 }, { "epoch": 1.62, "learning_rate": 2.516224976167779e-06, "loss": 3.4228, "step": 8508 }, { "epoch": 1.62, "learning_rate": 2.5161677788369875e-06, "loss": 2.288, "step": 8509 }, { "epoch": 1.62, "learning_rate": 2.5161105815061963e-06, "loss": 3.4214, "step": 8510 }, { "epoch": 1.62, "learning_rate": 2.5160533841754055e-06, "loss": 2.9976, "step": 8511 }, { "epoch": 1.62, "learning_rate": 2.515996186844614e-06, "loss": 3.23, "step": 8512 }, { "epoch": 1.62, "learning_rate": 2.5159389895138227e-06, "loss": 3.2634, "step": 8513 }, { "epoch": 1.62, "learning_rate": 2.5158817921830315e-06, "loss": 3.0936, "step": 8514 }, { "epoch": 1.62, "learning_rate": 2.5158245948522403e-06, "loss": 3.4376, "step": 8515 }, { "epoch": 1.62, "learning_rate": 2.515767397521449e-06, "loss": 2.7975, "step": 8516 }, { "epoch": 1.62, "learning_rate": 2.515710200190658e-06, "loss": 3.6433, "step": 8517 }, { "epoch": 1.62, "learning_rate": 2.5156530028598666e-06, "loss": 3.6235, "step": 8518 }, { "epoch": 1.62, "learning_rate": 2.5155958055290754e-06, "loss": 2.8798, "step": 8519 }, { "epoch": 1.62, "learning_rate": 2.5155386081982842e-06, "loss": 2.8917, "step": 8520 }, { "epoch": 1.62, "learning_rate": 2.515481410867493e-06, "loss": 3.5943, "step": 8521 }, { "epoch": 1.62, "learning_rate": 2.515424213536702e-06, "loss": 3.2946, "step": 8522 }, { "epoch": 1.62, "learning_rate": 2.5153670162059106e-06, "loss": 2.9814, "step": 8523 }, { "epoch": 1.62, "learning_rate": 2.515309818875119e-06, "loss": 3.0112, "step": 8524 }, { "epoch": 1.62, "learning_rate": 2.515252621544328e-06, "loss": 2.9283, "step": 8525 }, { "epoch": 1.62, "learning_rate": 2.5151954242135366e-06, "loss": 2.7636, "step": 8526 }, { "epoch": 1.62, "learning_rate": 2.5151382268827454e-06, "loss": 2.4573, "step": 8527 }, { "epoch": 1.62, "learning_rate": 2.5150810295519546e-06, "loss": 3.3841, "step": 8528 }, { "epoch": 1.62, "learning_rate": 2.515023832221163e-06, "loss": 2.6501, "step": 8529 }, { "epoch": 1.62, "learning_rate": 2.5149666348903718e-06, "loss": 2.2976, "step": 8530 }, { "epoch": 1.62, "learning_rate": 2.5149094375595805e-06, "loss": 2.7191, "step": 8531 }, { "epoch": 1.63, "learning_rate": 2.5148522402287893e-06, "loss": 2.3747, "step": 8532 }, { "epoch": 1.63, "learning_rate": 2.514795042897998e-06, "loss": 2.3137, "step": 8533 }, { "epoch": 1.63, "learning_rate": 2.514737845567207e-06, "loss": 2.7154, "step": 8534 }, { "epoch": 1.63, "learning_rate": 2.5146806482364157e-06, "loss": 3.0274, "step": 8535 }, { "epoch": 1.63, "learning_rate": 2.5146234509056245e-06, "loss": 2.6313, "step": 8536 }, { "epoch": 1.63, "learning_rate": 2.5145662535748333e-06, "loss": 3.7517, "step": 8537 }, { "epoch": 1.63, "learning_rate": 2.514509056244042e-06, "loss": 3.685, "step": 8538 }, { "epoch": 1.63, "learning_rate": 2.514451858913251e-06, "loss": 2.6166, "step": 8539 }, { "epoch": 1.63, "learning_rate": 2.5143946615824597e-06, "loss": 3.0368, "step": 8540 }, { "epoch": 1.63, "learning_rate": 2.514337464251668e-06, "loss": 2.5269, "step": 8541 }, { "epoch": 1.63, "learning_rate": 2.5142802669208773e-06, "loss": 3.0688, "step": 8542 }, { "epoch": 1.63, "learning_rate": 2.5142230695900856e-06, "loss": 2.9613, "step": 8543 }, { "epoch": 1.63, "learning_rate": 2.5141658722592944e-06, "loss": 2.4853, "step": 8544 }, { "epoch": 1.63, "learning_rate": 2.5141086749285037e-06, "loss": 2.7025, "step": 8545 }, { "epoch": 1.63, "learning_rate": 2.514051477597712e-06, "loss": 3.1953, "step": 8546 }, { "epoch": 1.63, "learning_rate": 2.513994280266921e-06, "loss": 2.7268, "step": 8547 }, { "epoch": 1.63, "learning_rate": 2.5139370829361296e-06, "loss": 3.2553, "step": 8548 }, { "epoch": 1.63, "learning_rate": 2.5138798856053384e-06, "loss": 2.8932, "step": 8549 }, { "epoch": 1.63, "learning_rate": 2.513822688274547e-06, "loss": 2.8078, "step": 8550 }, { "epoch": 1.63, "learning_rate": 2.513765490943756e-06, "loss": 3.6094, "step": 8551 }, { "epoch": 1.63, "learning_rate": 2.513708293612965e-06, "loss": 3.4757, "step": 8552 }, { "epoch": 1.63, "learning_rate": 2.5136510962821736e-06, "loss": 3.1147, "step": 8553 }, { "epoch": 1.63, "learning_rate": 2.5135938989513824e-06, "loss": 3.0017, "step": 8554 }, { "epoch": 1.63, "learning_rate": 2.513536701620591e-06, "loss": 2.3376, "step": 8555 }, { "epoch": 1.63, "learning_rate": 2.5134795042898e-06, "loss": 2.8403, "step": 8556 }, { "epoch": 1.63, "learning_rate": 2.5134223069590088e-06, "loss": 2.8232, "step": 8557 }, { "epoch": 1.63, "learning_rate": 2.513365109628217e-06, "loss": 2.6921, "step": 8558 }, { "epoch": 1.63, "learning_rate": 2.5133079122974263e-06, "loss": 3.1965, "step": 8559 }, { "epoch": 1.63, "learning_rate": 2.513250714966635e-06, "loss": 2.6, "step": 8560 }, { "epoch": 1.63, "learning_rate": 2.5131935176358435e-06, "loss": 3.0344, "step": 8561 }, { "epoch": 1.63, "learning_rate": 2.5131363203050527e-06, "loss": 2.9936, "step": 8562 }, { "epoch": 1.63, "learning_rate": 2.513079122974261e-06, "loss": 3.0323, "step": 8563 }, { "epoch": 1.63, "learning_rate": 2.51302192564347e-06, "loss": 2.8642, "step": 8564 }, { "epoch": 1.63, "learning_rate": 2.5129647283126787e-06, "loss": 2.96, "step": 8565 }, { "epoch": 1.63, "learning_rate": 2.5129075309818875e-06, "loss": 3.1685, "step": 8566 }, { "epoch": 1.63, "learning_rate": 2.5128503336510967e-06, "loss": 2.7881, "step": 8567 }, { "epoch": 1.63, "learning_rate": 2.512793136320305e-06, "loss": 3.7947, "step": 8568 }, { "epoch": 1.63, "learning_rate": 2.512735938989514e-06, "loss": 3.6067, "step": 8569 }, { "epoch": 1.63, "learning_rate": 2.5126787416587227e-06, "loss": 3.1984, "step": 8570 }, { "epoch": 1.63, "learning_rate": 2.5126215443279315e-06, "loss": 2.4749, "step": 8571 }, { "epoch": 1.63, "learning_rate": 2.5125643469971402e-06, "loss": 2.6958, "step": 8572 }, { "epoch": 1.63, "learning_rate": 2.512507149666349e-06, "loss": 3.6208, "step": 8573 }, { "epoch": 1.63, "learning_rate": 2.512449952335558e-06, "loss": 2.2138, "step": 8574 }, { "epoch": 1.63, "learning_rate": 2.512392755004766e-06, "loss": 2.5835, "step": 8575 }, { "epoch": 1.63, "learning_rate": 2.5123355576739754e-06, "loss": 3.3173, "step": 8576 }, { "epoch": 1.63, "learning_rate": 2.5122783603431842e-06, "loss": 3.4716, "step": 8577 }, { "epoch": 1.63, "learning_rate": 2.5122211630123926e-06, "loss": 2.698, "step": 8578 }, { "epoch": 1.63, "learning_rate": 2.512163965681602e-06, "loss": 2.7957, "step": 8579 }, { "epoch": 1.63, "learning_rate": 2.51210676835081e-06, "loss": 3.1866, "step": 8580 }, { "epoch": 1.63, "learning_rate": 2.512049571020019e-06, "loss": 2.5815, "step": 8581 }, { "epoch": 1.63, "learning_rate": 2.5119923736892278e-06, "loss": 2.6964, "step": 8582 }, { "epoch": 1.63, "learning_rate": 2.5119351763584366e-06, "loss": 3.0864, "step": 8583 }, { "epoch": 1.64, "learning_rate": 2.5118779790276458e-06, "loss": 3.1837, "step": 8584 }, { "epoch": 1.64, "learning_rate": 2.511820781696854e-06, "loss": 3.7817, "step": 8585 }, { "epoch": 1.64, "learning_rate": 2.511763584366063e-06, "loss": 3.2478, "step": 8586 }, { "epoch": 1.64, "learning_rate": 2.5117063870352717e-06, "loss": 3.0796, "step": 8587 }, { "epoch": 1.64, "learning_rate": 2.5116491897044805e-06, "loss": 3.9869, "step": 8588 }, { "epoch": 1.64, "learning_rate": 2.5115919923736893e-06, "loss": 2.5432, "step": 8589 }, { "epoch": 1.64, "learning_rate": 2.511534795042898e-06, "loss": 3.5588, "step": 8590 }, { "epoch": 1.64, "learning_rate": 2.511477597712107e-06, "loss": 2.9794, "step": 8591 }, { "epoch": 1.64, "learning_rate": 2.5114204003813153e-06, "loss": 2.8234, "step": 8592 }, { "epoch": 1.64, "learning_rate": 2.5113632030505245e-06, "loss": 3.8002, "step": 8593 }, { "epoch": 1.64, "learning_rate": 2.5113060057197333e-06, "loss": 3.5518, "step": 8594 }, { "epoch": 1.64, "learning_rate": 2.5112488083889417e-06, "loss": 3.1853, "step": 8595 }, { "epoch": 1.64, "learning_rate": 2.511191611058151e-06, "loss": 3.4608, "step": 8596 }, { "epoch": 1.64, "learning_rate": 2.5111344137273592e-06, "loss": 2.429, "step": 8597 }, { "epoch": 1.64, "learning_rate": 2.511077216396568e-06, "loss": 3.2023, "step": 8598 }, { "epoch": 1.64, "learning_rate": 2.5110200190657773e-06, "loss": 2.423, "step": 8599 }, { "epoch": 1.64, "learning_rate": 2.5109628217349856e-06, "loss": 2.5731, "step": 8600 }, { "epoch": 1.64, "learning_rate": 2.510905624404195e-06, "loss": 2.594, "step": 8601 }, { "epoch": 1.64, "learning_rate": 2.5108484270734032e-06, "loss": 2.9776, "step": 8602 }, { "epoch": 1.64, "learning_rate": 2.510791229742612e-06, "loss": 2.3751, "step": 8603 }, { "epoch": 1.64, "learning_rate": 2.510734032411821e-06, "loss": 2.5203, "step": 8604 }, { "epoch": 1.64, "learning_rate": 2.5106768350810296e-06, "loss": 2.8732, "step": 8605 }, { "epoch": 1.64, "learning_rate": 2.5106196377502384e-06, "loss": 3.0652, "step": 8606 }, { "epoch": 1.64, "learning_rate": 2.510562440419447e-06, "loss": 2.8612, "step": 8607 }, { "epoch": 1.64, "learning_rate": 2.510505243088656e-06, "loss": 2.6411, "step": 8608 }, { "epoch": 1.64, "learning_rate": 2.5104480457578648e-06, "loss": 2.291, "step": 8609 }, { "epoch": 1.64, "learning_rate": 2.5103908484270736e-06, "loss": 3.6008, "step": 8610 }, { "epoch": 1.64, "learning_rate": 2.5103336510962824e-06, "loss": 3.3664, "step": 8611 }, { "epoch": 1.64, "learning_rate": 2.5102764537654907e-06, "loss": 3.2246, "step": 8612 }, { "epoch": 1.64, "learning_rate": 2.5102192564347e-06, "loss": 3.4347, "step": 8613 }, { "epoch": 1.64, "learning_rate": 2.5101620591039083e-06, "loss": 3.1933, "step": 8614 }, { "epoch": 1.64, "learning_rate": 2.510104861773117e-06, "loss": 3.0545, "step": 8615 }, { "epoch": 1.64, "learning_rate": 2.5100476644423263e-06, "loss": 3.2053, "step": 8616 }, { "epoch": 1.64, "learning_rate": 2.5099904671115347e-06, "loss": 3.2655, "step": 8617 }, { "epoch": 1.64, "learning_rate": 2.509933269780744e-06, "loss": 3.11, "step": 8618 }, { "epoch": 1.64, "learning_rate": 2.5098760724499523e-06, "loss": 2.9974, "step": 8619 }, { "epoch": 1.64, "learning_rate": 2.509818875119161e-06, "loss": 3.4832, "step": 8620 }, { "epoch": 1.64, "learning_rate": 2.50976167778837e-06, "loss": 3.3226, "step": 8621 }, { "epoch": 1.64, "learning_rate": 2.5097044804575787e-06, "loss": 2.8445, "step": 8622 }, { "epoch": 1.64, "learning_rate": 2.5096472831267875e-06, "loss": 2.7981, "step": 8623 }, { "epoch": 1.64, "learning_rate": 2.5095900857959963e-06, "loss": 3.1827, "step": 8624 }, { "epoch": 1.64, "learning_rate": 2.509532888465205e-06, "loss": 2.465, "step": 8625 }, { "epoch": 1.64, "learning_rate": 2.509475691134414e-06, "loss": 2.9478, "step": 8626 }, { "epoch": 1.64, "learning_rate": 2.5094184938036226e-06, "loss": 3.0707, "step": 8627 }, { "epoch": 1.64, "learning_rate": 2.5093612964728314e-06, "loss": 3.6073, "step": 8628 }, { "epoch": 1.64, "learning_rate": 2.50930409914204e-06, "loss": 2.7636, "step": 8629 }, { "epoch": 1.64, "learning_rate": 2.509246901811249e-06, "loss": 2.7628, "step": 8630 }, { "epoch": 1.64, "learning_rate": 2.5091897044804574e-06, "loss": 2.8437, "step": 8631 }, { "epoch": 1.64, "learning_rate": 2.5091325071496666e-06, "loss": 3.4842, "step": 8632 }, { "epoch": 1.64, "learning_rate": 2.5090753098188754e-06, "loss": 3.0401, "step": 8633 }, { "epoch": 1.64, "learning_rate": 2.5090181124880838e-06, "loss": 3.6219, "step": 8634 }, { "epoch": 1.64, "learning_rate": 2.508960915157293e-06, "loss": 2.6399, "step": 8635 }, { "epoch": 1.64, "learning_rate": 2.5089037178265014e-06, "loss": 2.7062, "step": 8636 }, { "epoch": 1.65, "learning_rate": 2.50884652049571e-06, "loss": 3.3692, "step": 8637 }, { "epoch": 1.65, "learning_rate": 2.5087893231649194e-06, "loss": 2.5643, "step": 8638 }, { "epoch": 1.65, "learning_rate": 2.5087321258341277e-06, "loss": 2.3383, "step": 8639 }, { "epoch": 1.65, "learning_rate": 2.5086749285033365e-06, "loss": 2.5982, "step": 8640 }, { "epoch": 1.65, "learning_rate": 2.5086177311725453e-06, "loss": 3.1736, "step": 8641 }, { "epoch": 1.65, "learning_rate": 2.508560533841754e-06, "loss": 2.5102, "step": 8642 }, { "epoch": 1.65, "learning_rate": 2.508503336510963e-06, "loss": 2.538, "step": 8643 }, { "epoch": 1.65, "learning_rate": 2.5084461391801717e-06, "loss": 3.0693, "step": 8644 }, { "epoch": 1.65, "learning_rate": 2.5083889418493805e-06, "loss": 2.9483, "step": 8645 }, { "epoch": 1.65, "learning_rate": 2.508331744518589e-06, "loss": 3.7451, "step": 8646 }, { "epoch": 1.65, "learning_rate": 2.508274547187798e-06, "loss": 2.8555, "step": 8647 }, { "epoch": 1.65, "learning_rate": 2.508217349857007e-06, "loss": 3.752, "step": 8648 }, { "epoch": 1.65, "learning_rate": 2.5081601525262157e-06, "loss": 2.8067, "step": 8649 }, { "epoch": 1.65, "learning_rate": 2.5081029551954245e-06, "loss": 2.9265, "step": 8650 }, { "epoch": 1.65, "learning_rate": 2.508045757864633e-06, "loss": 3.4499, "step": 8651 }, { "epoch": 1.65, "learning_rate": 2.507988560533842e-06, "loss": 2.3475, "step": 8652 }, { "epoch": 1.65, "learning_rate": 2.5079313632030504e-06, "loss": 2.9991, "step": 8653 }, { "epoch": 1.65, "learning_rate": 2.5078741658722592e-06, "loss": 3.1932, "step": 8654 }, { "epoch": 1.65, "learning_rate": 2.5078169685414684e-06, "loss": 3.2331, "step": 8655 }, { "epoch": 1.65, "learning_rate": 2.507759771210677e-06, "loss": 3.4939, "step": 8656 }, { "epoch": 1.65, "learning_rate": 2.5077025738798856e-06, "loss": 3.5613, "step": 8657 }, { "epoch": 1.65, "learning_rate": 2.5076453765490944e-06, "loss": 2.6448, "step": 8658 }, { "epoch": 1.65, "learning_rate": 2.507588179218303e-06, "loss": 2.574, "step": 8659 }, { "epoch": 1.65, "learning_rate": 2.507530981887512e-06, "loss": 2.7108, "step": 8660 }, { "epoch": 1.65, "learning_rate": 2.5074737845567208e-06, "loss": 2.6341, "step": 8661 }, { "epoch": 1.65, "learning_rate": 2.5074165872259296e-06, "loss": 3.304, "step": 8662 }, { "epoch": 1.65, "learning_rate": 2.507359389895138e-06, "loss": 3.4024, "step": 8663 }, { "epoch": 1.65, "learning_rate": 2.507302192564347e-06, "loss": 3.714, "step": 8664 }, { "epoch": 1.65, "learning_rate": 2.507244995233556e-06, "loss": 3.2927, "step": 8665 }, { "epoch": 1.65, "learning_rate": 2.5071877979027647e-06, "loss": 2.8924, "step": 8666 }, { "epoch": 1.65, "learning_rate": 2.5071306005719735e-06, "loss": 3.0851, "step": 8667 }, { "epoch": 1.65, "learning_rate": 2.507073403241182e-06, "loss": 2.8913, "step": 8668 }, { "epoch": 1.65, "learning_rate": 2.507016205910391e-06, "loss": 3.3502, "step": 8669 }, { "epoch": 1.65, "learning_rate": 2.5069590085795995e-06, "loss": 2.5342, "step": 8670 }, { "epoch": 1.65, "learning_rate": 2.5069018112488083e-06, "loss": 3.2809, "step": 8671 }, { "epoch": 1.65, "learning_rate": 2.5068446139180175e-06, "loss": 3.2162, "step": 8672 }, { "epoch": 1.65, "learning_rate": 2.506787416587226e-06, "loss": 3.5777, "step": 8673 }, { "epoch": 1.65, "learning_rate": 2.5067302192564347e-06, "loss": 2.7353, "step": 8674 }, { "epoch": 1.65, "learning_rate": 2.5066730219256435e-06, "loss": 3.2131, "step": 8675 }, { "epoch": 1.65, "learning_rate": 2.5066158245948523e-06, "loss": 3.5376, "step": 8676 }, { "epoch": 1.65, "learning_rate": 2.506558627264061e-06, "loss": 3.1853, "step": 8677 }, { "epoch": 1.65, "learning_rate": 2.50650142993327e-06, "loss": 3.0839, "step": 8678 }, { "epoch": 1.65, "learning_rate": 2.5064442326024786e-06, "loss": 2.7048, "step": 8679 }, { "epoch": 1.65, "learning_rate": 2.506387035271687e-06, "loss": 2.7792, "step": 8680 }, { "epoch": 1.65, "learning_rate": 2.5063298379408962e-06, "loss": 2.2486, "step": 8681 }, { "epoch": 1.65, "learning_rate": 2.506272640610105e-06, "loss": 3.319, "step": 8682 }, { "epoch": 1.65, "learning_rate": 2.506215443279314e-06, "loss": 2.811, "step": 8683 }, { "epoch": 1.65, "learning_rate": 2.5061582459485226e-06, "loss": 3.1434, "step": 8684 }, { "epoch": 1.65, "learning_rate": 2.506101048617731e-06, "loss": 2.9224, "step": 8685 }, { "epoch": 1.65, "learning_rate": 2.50604385128694e-06, "loss": 3.3514, "step": 8686 }, { "epoch": 1.65, "learning_rate": 2.505986653956149e-06, "loss": 2.4278, "step": 8687 }, { "epoch": 1.65, "learning_rate": 2.5059294566253574e-06, "loss": 3.0641, "step": 8688 }, { "epoch": 1.66, "learning_rate": 2.5058722592945666e-06, "loss": 2.4281, "step": 8689 }, { "epoch": 1.66, "learning_rate": 2.505815061963775e-06, "loss": 3.1999, "step": 8690 }, { "epoch": 1.66, "learning_rate": 2.5057578646329837e-06, "loss": 1.8779, "step": 8691 }, { "epoch": 1.66, "learning_rate": 2.5057006673021925e-06, "loss": 3.2015, "step": 8692 }, { "epoch": 1.66, "learning_rate": 2.5056434699714013e-06, "loss": 2.6934, "step": 8693 }, { "epoch": 1.66, "learning_rate": 2.50558627264061e-06, "loss": 3.0358, "step": 8694 }, { "epoch": 1.66, "learning_rate": 2.505529075309819e-06, "loss": 2.7696, "step": 8695 }, { "epoch": 1.66, "learning_rate": 2.5054718779790277e-06, "loss": 3.3383, "step": 8696 }, { "epoch": 1.66, "learning_rate": 2.5054146806482365e-06, "loss": 3.4143, "step": 8697 }, { "epoch": 1.66, "learning_rate": 2.5053574833174453e-06, "loss": 3.1023, "step": 8698 }, { "epoch": 1.66, "learning_rate": 2.505300285986654e-06, "loss": 3.1434, "step": 8699 }, { "epoch": 1.66, "learning_rate": 2.505243088655863e-06, "loss": 3.822, "step": 8700 }, { "epoch": 1.66, "learning_rate": 2.5051858913250717e-06, "loss": 2.8821, "step": 8701 }, { "epoch": 1.66, "learning_rate": 2.50512869399428e-06, "loss": 3.0658, "step": 8702 }, { "epoch": 1.66, "learning_rate": 2.5050714966634893e-06, "loss": 3.2793, "step": 8703 }, { "epoch": 1.66, "learning_rate": 2.505014299332698e-06, "loss": 2.8764, "step": 8704 }, { "epoch": 1.66, "learning_rate": 2.5049571020019064e-06, "loss": 2.9517, "step": 8705 }, { "epoch": 1.66, "learning_rate": 2.5048999046711157e-06, "loss": 2.9098, "step": 8706 }, { "epoch": 1.66, "learning_rate": 2.504842707340324e-06, "loss": 3.8137, "step": 8707 }, { "epoch": 1.66, "learning_rate": 2.504785510009533e-06, "loss": 3.3632, "step": 8708 }, { "epoch": 1.66, "learning_rate": 2.5047283126787416e-06, "loss": 3.735, "step": 8709 }, { "epoch": 1.66, "learning_rate": 2.5046711153479504e-06, "loss": 3.3331, "step": 8710 }, { "epoch": 1.66, "learning_rate": 2.504613918017159e-06, "loss": 2.4763, "step": 8711 }, { "epoch": 1.66, "learning_rate": 2.504556720686368e-06, "loss": 3.075, "step": 8712 }, { "epoch": 1.66, "learning_rate": 2.5044995233555768e-06, "loss": 3.0004, "step": 8713 }, { "epoch": 1.66, "learning_rate": 2.5044423260247856e-06, "loss": 2.5823, "step": 8714 }, { "epoch": 1.66, "learning_rate": 2.5043851286939944e-06, "loss": 2.9676, "step": 8715 }, { "epoch": 1.66, "learning_rate": 2.504327931363203e-06, "loss": 3.3754, "step": 8716 }, { "epoch": 1.66, "learning_rate": 2.504270734032412e-06, "loss": 3.3286, "step": 8717 }, { "epoch": 1.66, "learning_rate": 2.5042135367016208e-06, "loss": 2.9853, "step": 8718 }, { "epoch": 1.66, "learning_rate": 2.504156339370829e-06, "loss": 3.7716, "step": 8719 }, { "epoch": 1.66, "learning_rate": 2.5040991420400383e-06, "loss": 3.0037, "step": 8720 }, { "epoch": 1.66, "learning_rate": 2.504041944709247e-06, "loss": 2.4857, "step": 8721 }, { "epoch": 1.66, "learning_rate": 2.5039847473784555e-06, "loss": 3.6784, "step": 8722 }, { "epoch": 1.66, "learning_rate": 2.5039275500476647e-06, "loss": 2.8567, "step": 8723 }, { "epoch": 1.66, "learning_rate": 2.503870352716873e-06, "loss": 2.6515, "step": 8724 }, { "epoch": 1.66, "learning_rate": 2.503813155386082e-06, "loss": 3.6687, "step": 8725 }, { "epoch": 1.66, "learning_rate": 2.503755958055291e-06, "loss": 2.5032, "step": 8726 }, { "epoch": 1.66, "learning_rate": 2.5036987607244995e-06, "loss": 3.3914, "step": 8727 }, { "epoch": 1.66, "learning_rate": 2.5036415633937083e-06, "loss": 2.3431, "step": 8728 }, { "epoch": 1.66, "learning_rate": 2.503584366062917e-06, "loss": 3.1033, "step": 8729 }, { "epoch": 1.66, "learning_rate": 2.503527168732126e-06, "loss": 3.533, "step": 8730 }, { "epoch": 1.66, "learning_rate": 2.5034699714013347e-06, "loss": 3.0759, "step": 8731 }, { "epoch": 1.66, "learning_rate": 2.5034127740705434e-06, "loss": 3.0042, "step": 8732 }, { "epoch": 1.66, "learning_rate": 2.5033555767397522e-06, "loss": 2.845, "step": 8733 }, { "epoch": 1.66, "learning_rate": 2.503298379408961e-06, "loss": 2.5516, "step": 8734 }, { "epoch": 1.66, "learning_rate": 2.50324118207817e-06, "loss": 3.0448, "step": 8735 }, { "epoch": 1.66, "learning_rate": 2.5031839847473786e-06, "loss": 2.5554, "step": 8736 }, { "epoch": 1.66, "learning_rate": 2.5031267874165874e-06, "loss": 3.3459, "step": 8737 }, { "epoch": 1.66, "learning_rate": 2.503069590085796e-06, "loss": 2.6622, "step": 8738 }, { "epoch": 1.66, "learning_rate": 2.5030123927550046e-06, "loss": 3.4349, "step": 8739 }, { "epoch": 1.66, "learning_rate": 2.502955195424214e-06, "loss": 2.3889, "step": 8740 }, { "epoch": 1.66, "learning_rate": 2.502897998093422e-06, "loss": 3.3423, "step": 8741 }, { "epoch": 1.67, "learning_rate": 2.502840800762631e-06, "loss": 3.5159, "step": 8742 }, { "epoch": 1.67, "learning_rate": 2.50278360343184e-06, "loss": 3.4519, "step": 8743 }, { "epoch": 1.67, "learning_rate": 2.5027264061010485e-06, "loss": 4.166, "step": 8744 }, { "epoch": 1.67, "learning_rate": 2.5026692087702573e-06, "loss": 3.1214, "step": 8745 }, { "epoch": 1.67, "learning_rate": 2.502612011439466e-06, "loss": 3.1722, "step": 8746 }, { "epoch": 1.67, "learning_rate": 2.502554814108675e-06, "loss": 2.7509, "step": 8747 }, { "epoch": 1.67, "learning_rate": 2.5024976167778837e-06, "loss": 2.7619, "step": 8748 }, { "epoch": 1.67, "learning_rate": 2.5024404194470925e-06, "loss": 3.0217, "step": 8749 }, { "epoch": 1.67, "learning_rate": 2.5023832221163013e-06, "loss": 3.596, "step": 8750 }, { "epoch": 1.67, "learning_rate": 2.50232602478551e-06, "loss": 3.0513, "step": 8751 }, { "epoch": 1.67, "learning_rate": 2.502268827454719e-06, "loss": 3.3931, "step": 8752 }, { "epoch": 1.67, "learning_rate": 2.5022116301239277e-06, "loss": 3.0357, "step": 8753 }, { "epoch": 1.67, "learning_rate": 2.5021544327931365e-06, "loss": 2.5828, "step": 8754 }, { "epoch": 1.67, "learning_rate": 2.5020972354623453e-06, "loss": 3.4991, "step": 8755 }, { "epoch": 1.67, "learning_rate": 2.5020400381315536e-06, "loss": 2.9549, "step": 8756 }, { "epoch": 1.67, "learning_rate": 2.501982840800763e-06, "loss": 2.4943, "step": 8757 }, { "epoch": 1.67, "learning_rate": 2.5019256434699712e-06, "loss": 3.1362, "step": 8758 }, { "epoch": 1.67, "learning_rate": 2.50186844613918e-06, "loss": 2.4258, "step": 8759 }, { "epoch": 1.67, "learning_rate": 2.5018112488083892e-06, "loss": 3.7563, "step": 8760 }, { "epoch": 1.67, "learning_rate": 2.5017540514775976e-06, "loss": 2.4339, "step": 8761 }, { "epoch": 1.67, "learning_rate": 2.501696854146807e-06, "loss": 2.8593, "step": 8762 }, { "epoch": 1.67, "learning_rate": 2.501639656816015e-06, "loss": 3.3649, "step": 8763 }, { "epoch": 1.67, "learning_rate": 2.501582459485224e-06, "loss": 3.0635, "step": 8764 }, { "epoch": 1.67, "learning_rate": 2.501525262154433e-06, "loss": 3.1604, "step": 8765 }, { "epoch": 1.67, "learning_rate": 2.5014680648236416e-06, "loss": 3.2999, "step": 8766 }, { "epoch": 1.67, "learning_rate": 2.5014108674928504e-06, "loss": 2.9947, "step": 8767 }, { "epoch": 1.67, "learning_rate": 2.501353670162059e-06, "loss": 2.6036, "step": 8768 }, { "epoch": 1.67, "learning_rate": 2.501296472831268e-06, "loss": 3.1414, "step": 8769 }, { "epoch": 1.67, "learning_rate": 2.5012392755004768e-06, "loss": 2.6016, "step": 8770 }, { "epoch": 1.67, "learning_rate": 2.5011820781696856e-06, "loss": 3.153, "step": 8771 }, { "epoch": 1.67, "learning_rate": 2.5011248808388943e-06, "loss": 2.4097, "step": 8772 }, { "epoch": 1.67, "learning_rate": 2.5010676835081027e-06, "loss": 2.7409, "step": 8773 }, { "epoch": 1.67, "learning_rate": 2.501010486177312e-06, "loss": 2.782, "step": 8774 }, { "epoch": 1.67, "learning_rate": 2.5009532888465207e-06, "loss": 3.1973, "step": 8775 }, { "epoch": 1.67, "learning_rate": 2.500896091515729e-06, "loss": 2.2714, "step": 8776 }, { "epoch": 1.67, "learning_rate": 2.5008388941849383e-06, "loss": 3.5499, "step": 8777 }, { "epoch": 1.67, "learning_rate": 2.5007816968541467e-06, "loss": 3.7, "step": 8778 }, { "epoch": 1.67, "learning_rate": 2.500724499523356e-06, "loss": 2.6248, "step": 8779 }, { "epoch": 1.67, "learning_rate": 2.5006673021925643e-06, "loss": 3.6604, "step": 8780 }, { "epoch": 1.67, "learning_rate": 2.500610104861773e-06, "loss": 2.656, "step": 8781 }, { "epoch": 1.67, "learning_rate": 2.5005529075309823e-06, "loss": 2.8358, "step": 8782 }, { "epoch": 1.67, "learning_rate": 2.5004957102001907e-06, "loss": 3.0575, "step": 8783 }, { "epoch": 1.67, "learning_rate": 2.5004385128693995e-06, "loss": 3.1965, "step": 8784 }, { "epoch": 1.67, "learning_rate": 2.5003813155386082e-06, "loss": 3.2677, "step": 8785 }, { "epoch": 1.67, "learning_rate": 2.500324118207817e-06, "loss": 3.0218, "step": 8786 }, { "epoch": 1.67, "learning_rate": 2.500266920877026e-06, "loss": 3.2048, "step": 8787 }, { "epoch": 1.67, "learning_rate": 2.5002097235462346e-06, "loss": 3.0838, "step": 8788 }, { "epoch": 1.67, "learning_rate": 2.5001525262154434e-06, "loss": 2.7882, "step": 8789 }, { "epoch": 1.67, "learning_rate": 2.500095328884652e-06, "loss": 2.9918, "step": 8790 }, { "epoch": 1.67, "learning_rate": 2.500038131553861e-06, "loss": 2.9068, "step": 8791 }, { "epoch": 1.67, "learning_rate": 2.49998093422307e-06, "loss": 3.177, "step": 8792 }, { "epoch": 1.67, "learning_rate": 2.499923736892278e-06, "loss": 3.6219, "step": 8793 }, { "epoch": 1.68, "learning_rate": 2.4998665395614874e-06, "loss": 2.5542, "step": 8794 }, { "epoch": 1.68, "learning_rate": 2.4998093422306958e-06, "loss": 3.3, "step": 8795 }, { "epoch": 1.68, "learning_rate": 2.499752144899905e-06, "loss": 2.2803, "step": 8796 }, { "epoch": 1.68, "learning_rate": 2.4996949475691133e-06, "loss": 2.8357, "step": 8797 }, { "epoch": 1.68, "learning_rate": 2.499637750238322e-06, "loss": 2.8439, "step": 8798 }, { "epoch": 1.68, "learning_rate": 2.4995805529075314e-06, "loss": 3.1861, "step": 8799 }, { "epoch": 1.68, "learning_rate": 2.4995233555767397e-06, "loss": 2.8274, "step": 8800 }, { "epoch": 1.68, "learning_rate": 2.4994661582459485e-06, "loss": 2.4241, "step": 8801 }, { "epoch": 1.68, "learning_rate": 2.4994089609151573e-06, "loss": 2.1441, "step": 8802 }, { "epoch": 1.68, "learning_rate": 2.499351763584366e-06, "loss": 2.8659, "step": 8803 }, { "epoch": 1.68, "learning_rate": 2.499294566253575e-06, "loss": 3.426, "step": 8804 }, { "epoch": 1.68, "learning_rate": 2.4992373689227837e-06, "loss": 2.4897, "step": 8805 }, { "epoch": 1.68, "learning_rate": 2.4991801715919925e-06, "loss": 3.1824, "step": 8806 }, { "epoch": 1.68, "learning_rate": 2.499122974261201e-06, "loss": 2.7475, "step": 8807 }, { "epoch": 1.68, "learning_rate": 2.49906577693041e-06, "loss": 3.214, "step": 8808 }, { "epoch": 1.68, "learning_rate": 2.499008579599619e-06, "loss": 3.1129, "step": 8809 }, { "epoch": 1.68, "learning_rate": 2.4989513822688272e-06, "loss": 2.944, "step": 8810 }, { "epoch": 1.68, "learning_rate": 2.4988941849380365e-06, "loss": 3.0826, "step": 8811 }, { "epoch": 1.68, "learning_rate": 2.498836987607245e-06, "loss": 2.3883, "step": 8812 }, { "epoch": 1.68, "learning_rate": 2.498779790276454e-06, "loss": 3.6647, "step": 8813 }, { "epoch": 1.68, "learning_rate": 2.498722592945663e-06, "loss": 2.4315, "step": 8814 }, { "epoch": 1.68, "learning_rate": 2.4986653956148712e-06, "loss": 2.8636, "step": 8815 }, { "epoch": 1.68, "learning_rate": 2.4986081982840804e-06, "loss": 2.5619, "step": 8816 }, { "epoch": 1.68, "learning_rate": 2.498551000953289e-06, "loss": 3.0598, "step": 8817 }, { "epoch": 1.68, "learning_rate": 2.4984938036224976e-06, "loss": 2.091, "step": 8818 }, { "epoch": 1.68, "learning_rate": 2.4984366062917064e-06, "loss": 3.1043, "step": 8819 }, { "epoch": 1.68, "learning_rate": 2.498379408960915e-06, "loss": 3.2171, "step": 8820 }, { "epoch": 1.68, "learning_rate": 2.498322211630124e-06, "loss": 3.4563, "step": 8821 }, { "epoch": 1.68, "learning_rate": 2.4982650142993328e-06, "loss": 3.4713, "step": 8822 }, { "epoch": 1.68, "learning_rate": 2.4982078169685416e-06, "loss": 3.2888, "step": 8823 }, { "epoch": 1.68, "learning_rate": 2.4981506196377504e-06, "loss": 2.3878, "step": 8824 }, { "epoch": 1.68, "learning_rate": 2.498093422306959e-06, "loss": 3.13, "step": 8825 }, { "epoch": 1.68, "learning_rate": 2.498036224976168e-06, "loss": 2.9673, "step": 8826 }, { "epoch": 1.68, "learning_rate": 2.4979790276453767e-06, "loss": 2.5715, "step": 8827 }, { "epoch": 1.68, "learning_rate": 2.4979218303145855e-06, "loss": 2.588, "step": 8828 }, { "epoch": 1.68, "learning_rate": 2.497864632983794e-06, "loss": 2.2875, "step": 8829 }, { "epoch": 1.68, "learning_rate": 2.497807435653003e-06, "loss": 2.8429, "step": 8830 }, { "epoch": 1.68, "learning_rate": 2.497750238322212e-06, "loss": 3.0267, "step": 8831 }, { "epoch": 1.68, "learning_rate": 2.4976930409914203e-06, "loss": 3.1139, "step": 8832 }, { "epoch": 1.68, "learning_rate": 2.4976358436606295e-06, "loss": 2.3029, "step": 8833 }, { "epoch": 1.68, "learning_rate": 2.497578646329838e-06, "loss": 3.1061, "step": 8834 }, { "epoch": 1.68, "learning_rate": 2.4975214489990467e-06, "loss": 2.4791, "step": 8835 }, { "epoch": 1.68, "learning_rate": 2.4974642516682555e-06, "loss": 3.0981, "step": 8836 }, { "epoch": 1.68, "learning_rate": 2.4974070543374643e-06, "loss": 2.7764, "step": 8837 }, { "epoch": 1.68, "learning_rate": 2.497349857006673e-06, "loss": 3.0251, "step": 8838 }, { "epoch": 1.68, "learning_rate": 2.497292659675882e-06, "loss": 2.7336, "step": 8839 }, { "epoch": 1.68, "learning_rate": 2.4972354623450906e-06, "loss": 3.8951, "step": 8840 }, { "epoch": 1.68, "learning_rate": 2.4971782650142994e-06, "loss": 2.8273, "step": 8841 }, { "epoch": 1.68, "learning_rate": 2.4971210676835082e-06, "loss": 2.8439, "step": 8842 }, { "epoch": 1.68, "learning_rate": 2.497063870352717e-06, "loss": 2.7445, "step": 8843 }, { "epoch": 1.68, "learning_rate": 2.497006673021926e-06, "loss": 3.7804, "step": 8844 }, { "epoch": 1.68, "learning_rate": 2.4969494756911346e-06, "loss": 3.1759, "step": 8845 }, { "epoch": 1.68, "learning_rate": 2.496892278360343e-06, "loss": 3.2121, "step": 8846 }, { "epoch": 1.69, "learning_rate": 2.496835081029552e-06, "loss": 2.3722, "step": 8847 }, { "epoch": 1.69, "learning_rate": 2.496777883698761e-06, "loss": 3.5344, "step": 8848 }, { "epoch": 1.69, "learning_rate": 2.4967206863679694e-06, "loss": 3.5331, "step": 8849 }, { "epoch": 1.69, "learning_rate": 2.4966634890371786e-06, "loss": 3.5828, "step": 8850 }, { "epoch": 1.69, "learning_rate": 2.496606291706387e-06, "loss": 3.3237, "step": 8851 }, { "epoch": 1.69, "learning_rate": 2.4965490943755957e-06, "loss": 2.6158, "step": 8852 }, { "epoch": 1.69, "learning_rate": 2.4964918970448045e-06, "loss": 3.1194, "step": 8853 }, { "epoch": 1.69, "learning_rate": 2.4964346997140133e-06, "loss": 3.2709, "step": 8854 }, { "epoch": 1.69, "learning_rate": 2.496377502383222e-06, "loss": 2.9017, "step": 8855 }, { "epoch": 1.69, "learning_rate": 2.496320305052431e-06, "loss": 3.6234, "step": 8856 }, { "epoch": 1.69, "learning_rate": 2.4962631077216397e-06, "loss": 2.3542, "step": 8857 }, { "epoch": 1.69, "learning_rate": 2.4962059103908485e-06, "loss": 2.6693, "step": 8858 }, { "epoch": 1.69, "learning_rate": 2.4961487130600573e-06, "loss": 3.6616, "step": 8859 }, { "epoch": 1.69, "learning_rate": 2.496091515729266e-06, "loss": 3.2009, "step": 8860 }, { "epoch": 1.69, "learning_rate": 2.496034318398475e-06, "loss": 3.2148, "step": 8861 }, { "epoch": 1.69, "learning_rate": 2.4959771210676837e-06, "loss": 2.7314, "step": 8862 }, { "epoch": 1.69, "learning_rate": 2.4959199237368925e-06, "loss": 2.75, "step": 8863 }, { "epoch": 1.69, "learning_rate": 2.4958627264061013e-06, "loss": 2.5664, "step": 8864 }, { "epoch": 1.69, "learning_rate": 2.49580552907531e-06, "loss": 2.3655, "step": 8865 }, { "epoch": 1.69, "learning_rate": 2.4957483317445184e-06, "loss": 3.8173, "step": 8866 }, { "epoch": 1.69, "learning_rate": 2.4956911344137276e-06, "loss": 3.218, "step": 8867 }, { "epoch": 1.69, "learning_rate": 2.495633937082936e-06, "loss": 2.8156, "step": 8868 }, { "epoch": 1.69, "learning_rate": 2.495576739752145e-06, "loss": 3.7201, "step": 8869 }, { "epoch": 1.69, "learning_rate": 2.495519542421354e-06, "loss": 3.099, "step": 8870 }, { "epoch": 1.69, "learning_rate": 2.4954623450905624e-06, "loss": 2.6198, "step": 8871 }, { "epoch": 1.69, "learning_rate": 2.495405147759771e-06, "loss": 2.9921, "step": 8872 }, { "epoch": 1.69, "learning_rate": 2.49534795042898e-06, "loss": 2.6862, "step": 8873 }, { "epoch": 1.69, "learning_rate": 2.4952907530981888e-06, "loss": 2.5223, "step": 8874 }, { "epoch": 1.69, "learning_rate": 2.4952335557673976e-06, "loss": 2.8181, "step": 8875 }, { "epoch": 1.69, "learning_rate": 2.4951763584366064e-06, "loss": 2.6922, "step": 8876 }, { "epoch": 1.69, "learning_rate": 2.495119161105815e-06, "loss": 2.7317, "step": 8877 }, { "epoch": 1.69, "learning_rate": 2.495061963775024e-06, "loss": 3.0545, "step": 8878 }, { "epoch": 1.69, "learning_rate": 2.4950047664442327e-06, "loss": 2.7373, "step": 8879 }, { "epoch": 1.69, "learning_rate": 2.4949475691134415e-06, "loss": 2.8442, "step": 8880 }, { "epoch": 1.69, "learning_rate": 2.4948903717826503e-06, "loss": 2.8185, "step": 8881 }, { "epoch": 1.69, "learning_rate": 2.494833174451859e-06, "loss": 2.7024, "step": 8882 }, { "epoch": 1.69, "learning_rate": 2.4947759771210675e-06, "loss": 3.1154, "step": 8883 }, { "epoch": 1.69, "learning_rate": 2.4947187797902767e-06, "loss": 2.8817, "step": 8884 }, { "epoch": 1.69, "learning_rate": 2.494661582459485e-06, "loss": 3.2724, "step": 8885 }, { "epoch": 1.69, "learning_rate": 2.494604385128694e-06, "loss": 3.8494, "step": 8886 }, { "epoch": 1.69, "learning_rate": 2.494547187797903e-06, "loss": 3.037, "step": 8887 }, { "epoch": 1.69, "learning_rate": 2.4944899904671115e-06, "loss": 3.0159, "step": 8888 }, { "epoch": 1.69, "learning_rate": 2.4944327931363203e-06, "loss": 2.4553, "step": 8889 }, { "epoch": 1.69, "learning_rate": 2.494375595805529e-06, "loss": 2.8619, "step": 8890 }, { "epoch": 1.69, "learning_rate": 2.494318398474738e-06, "loss": 3.408, "step": 8891 }, { "epoch": 1.69, "learning_rate": 2.4942612011439466e-06, "loss": 2.9906, "step": 8892 }, { "epoch": 1.69, "learning_rate": 2.4942040038131554e-06, "loss": 3.2057, "step": 8893 }, { "epoch": 1.69, "learning_rate": 2.4941468064823642e-06, "loss": 3.2415, "step": 8894 }, { "epoch": 1.69, "learning_rate": 2.494089609151573e-06, "loss": 3.4346, "step": 8895 }, { "epoch": 1.69, "learning_rate": 2.494032411820782e-06, "loss": 3.3652, "step": 8896 }, { "epoch": 1.69, "learning_rate": 2.4939752144899906e-06, "loss": 3.4166, "step": 8897 }, { "epoch": 1.69, "learning_rate": 2.4939180171591994e-06, "loss": 2.7159, "step": 8898 }, { "epoch": 1.7, "learning_rate": 2.493860819828408e-06, "loss": 2.9682, "step": 8899 }, { "epoch": 1.7, "learning_rate": 2.4938036224976166e-06, "loss": 2.4534, "step": 8900 }, { "epoch": 1.7, "learning_rate": 2.4937464251668258e-06, "loss": 3.1679, "step": 8901 }, { "epoch": 1.7, "learning_rate": 2.493689227836034e-06, "loss": 2.9189, "step": 8902 }, { "epoch": 1.7, "learning_rate": 2.493632030505243e-06, "loss": 2.8454, "step": 8903 }, { "epoch": 1.7, "learning_rate": 2.493574833174452e-06, "loss": 3.4251, "step": 8904 }, { "epoch": 1.7, "learning_rate": 2.4935176358436605e-06, "loss": 2.6405, "step": 8905 }, { "epoch": 1.7, "learning_rate": 2.4934604385128693e-06, "loss": 2.2447, "step": 8906 }, { "epoch": 1.7, "learning_rate": 2.493403241182078e-06, "loss": 2.943, "step": 8907 }, { "epoch": 1.7, "learning_rate": 2.493346043851287e-06, "loss": 3.5252, "step": 8908 }, { "epoch": 1.7, "learning_rate": 2.493288846520496e-06, "loss": 3.5713, "step": 8909 }, { "epoch": 1.7, "learning_rate": 2.4932316491897045e-06, "loss": 3.0762, "step": 8910 }, { "epoch": 1.7, "learning_rate": 2.4931744518589133e-06, "loss": 3.6449, "step": 8911 }, { "epoch": 1.7, "learning_rate": 2.493117254528122e-06, "loss": 3.4092, "step": 8912 }, { "epoch": 1.7, "learning_rate": 2.493060057197331e-06, "loss": 2.358, "step": 8913 }, { "epoch": 1.7, "learning_rate": 2.4930028598665397e-06, "loss": 2.6931, "step": 8914 }, { "epoch": 1.7, "learning_rate": 2.4929456625357485e-06, "loss": 2.9091, "step": 8915 }, { "epoch": 1.7, "learning_rate": 2.4928884652049573e-06, "loss": 3.1697, "step": 8916 }, { "epoch": 1.7, "learning_rate": 2.4928312678741656e-06, "loss": 3.8834, "step": 8917 }, { "epoch": 1.7, "learning_rate": 2.492774070543375e-06, "loss": 3.6971, "step": 8918 }, { "epoch": 1.7, "learning_rate": 2.4927168732125837e-06, "loss": 3.3276, "step": 8919 }, { "epoch": 1.7, "learning_rate": 2.492659675881792e-06, "loss": 3.3966, "step": 8920 }, { "epoch": 1.7, "learning_rate": 2.4926024785510012e-06, "loss": 3.333, "step": 8921 }, { "epoch": 1.7, "learning_rate": 2.4925452812202096e-06, "loss": 2.8263, "step": 8922 }, { "epoch": 1.7, "learning_rate": 2.4924880838894184e-06, "loss": 2.8355, "step": 8923 }, { "epoch": 1.7, "learning_rate": 2.492430886558627e-06, "loss": 3.0227, "step": 8924 }, { "epoch": 1.7, "learning_rate": 2.492373689227836e-06, "loss": 2.8656, "step": 8925 }, { "epoch": 1.7, "learning_rate": 2.492316491897045e-06, "loss": 3.7611, "step": 8926 }, { "epoch": 1.7, "learning_rate": 2.4922592945662536e-06, "loss": 3.2775, "step": 8927 }, { "epoch": 1.7, "learning_rate": 2.4922020972354624e-06, "loss": 2.6396, "step": 8928 }, { "epoch": 1.7, "learning_rate": 2.492144899904671e-06, "loss": 2.5628, "step": 8929 }, { "epoch": 1.7, "learning_rate": 2.49208770257388e-06, "loss": 3.18, "step": 8930 }, { "epoch": 1.7, "learning_rate": 2.4920305052430888e-06, "loss": 3.0785, "step": 8931 }, { "epoch": 1.7, "learning_rate": 2.4919733079122975e-06, "loss": 3.1502, "step": 8932 }, { "epoch": 1.7, "learning_rate": 2.4919161105815063e-06, "loss": 3.5815, "step": 8933 }, { "epoch": 1.7, "learning_rate": 2.4918589132507147e-06, "loss": 3.5731, "step": 8934 }, { "epoch": 1.7, "learning_rate": 2.491801715919924e-06, "loss": 2.4729, "step": 8935 }, { "epoch": 1.7, "learning_rate": 2.4917445185891327e-06, "loss": 2.8783, "step": 8936 }, { "epoch": 1.7, "learning_rate": 2.491687321258341e-06, "loss": 3.0022, "step": 8937 }, { "epoch": 1.7, "learning_rate": 2.4916301239275503e-06, "loss": 2.4751, "step": 8938 }, { "epoch": 1.7, "learning_rate": 2.4915729265967587e-06, "loss": 3.2138, "step": 8939 }, { "epoch": 1.7, "learning_rate": 2.4915157292659675e-06, "loss": 3.5166, "step": 8940 }, { "epoch": 1.7, "learning_rate": 2.4914585319351763e-06, "loss": 3.4284, "step": 8941 }, { "epoch": 1.7, "learning_rate": 2.491401334604385e-06, "loss": 3.4538, "step": 8942 }, { "epoch": 1.7, "learning_rate": 2.4913441372735943e-06, "loss": 3.9503, "step": 8943 }, { "epoch": 1.7, "learning_rate": 2.4912869399428027e-06, "loss": 3.6242, "step": 8944 }, { "epoch": 1.7, "learning_rate": 2.4912297426120114e-06, "loss": 3.779, "step": 8945 }, { "epoch": 1.7, "learning_rate": 2.4911725452812202e-06, "loss": 3.681, "step": 8946 }, { "epoch": 1.7, "learning_rate": 2.491115347950429e-06, "loss": 4.0109, "step": 8947 }, { "epoch": 1.7, "learning_rate": 2.491058150619638e-06, "loss": 2.7314, "step": 8948 }, { "epoch": 1.7, "learning_rate": 2.4910009532888466e-06, "loss": 3.3825, "step": 8949 }, { "epoch": 1.7, "learning_rate": 2.4909437559580554e-06, "loss": 3.1864, "step": 8950 }, { "epoch": 1.7, "learning_rate": 2.490886558627264e-06, "loss": 2.7699, "step": 8951 }, { "epoch": 1.71, "learning_rate": 2.490829361296473e-06, "loss": 3.5058, "step": 8952 }, { "epoch": 1.71, "learning_rate": 2.490772163965682e-06, "loss": 2.5919, "step": 8953 }, { "epoch": 1.71, "learning_rate": 2.49071496663489e-06, "loss": 3.6658, "step": 8954 }, { "epoch": 1.71, "learning_rate": 2.4906577693040994e-06, "loss": 2.9746, "step": 8955 }, { "epoch": 1.71, "learning_rate": 2.4906005719733078e-06, "loss": 3.4496, "step": 8956 }, { "epoch": 1.71, "learning_rate": 2.4905433746425165e-06, "loss": 2.734, "step": 8957 }, { "epoch": 1.71, "learning_rate": 2.4904861773117258e-06, "loss": 3.2261, "step": 8958 }, { "epoch": 1.71, "learning_rate": 2.490428979980934e-06, "loss": 2.9083, "step": 8959 }, { "epoch": 1.71, "learning_rate": 2.4903717826501434e-06, "loss": 3.1053, "step": 8960 }, { "epoch": 1.71, "learning_rate": 2.4903145853193517e-06, "loss": 2.5222, "step": 8961 }, { "epoch": 1.71, "learning_rate": 2.4902573879885605e-06, "loss": 3.5559, "step": 8962 }, { "epoch": 1.71, "learning_rate": 2.4902001906577693e-06, "loss": 2.6105, "step": 8963 }, { "epoch": 1.71, "learning_rate": 2.490142993326978e-06, "loss": 3.3562, "step": 8964 }, { "epoch": 1.71, "learning_rate": 2.490085795996187e-06, "loss": 2.8353, "step": 8965 }, { "epoch": 1.71, "learning_rate": 2.4900285986653957e-06, "loss": 2.9461, "step": 8966 }, { "epoch": 1.71, "learning_rate": 2.4899714013346045e-06, "loss": 3.0814, "step": 8967 }, { "epoch": 1.71, "learning_rate": 2.4899142040038133e-06, "loss": 3.1585, "step": 8968 }, { "epoch": 1.71, "learning_rate": 2.489857006673022e-06, "loss": 2.9634, "step": 8969 }, { "epoch": 1.71, "learning_rate": 2.489799809342231e-06, "loss": 2.9109, "step": 8970 }, { "epoch": 1.71, "learning_rate": 2.4897426120114392e-06, "loss": 2.2336, "step": 8971 }, { "epoch": 1.71, "learning_rate": 2.4896854146806485e-06, "loss": 3.8825, "step": 8972 }, { "epoch": 1.71, "learning_rate": 2.489628217349857e-06, "loss": 3.4218, "step": 8973 }, { "epoch": 1.71, "learning_rate": 2.489571020019066e-06, "loss": 2.5489, "step": 8974 }, { "epoch": 1.71, "learning_rate": 2.489513822688275e-06, "loss": 3.0325, "step": 8975 }, { "epoch": 1.71, "learning_rate": 2.489456625357483e-06, "loss": 2.3608, "step": 8976 }, { "epoch": 1.71, "learning_rate": 2.4893994280266924e-06, "loss": 2.6746, "step": 8977 }, { "epoch": 1.71, "learning_rate": 2.489342230695901e-06, "loss": 2.3437, "step": 8978 }, { "epoch": 1.71, "learning_rate": 2.4892850333651096e-06, "loss": 3.1551, "step": 8979 }, { "epoch": 1.71, "learning_rate": 2.4892278360343184e-06, "loss": 3.2311, "step": 8980 }, { "epoch": 1.71, "learning_rate": 2.489170638703527e-06, "loss": 3.2535, "step": 8981 }, { "epoch": 1.71, "learning_rate": 2.489113441372736e-06, "loss": 2.9469, "step": 8982 }, { "epoch": 1.71, "learning_rate": 2.4890562440419448e-06, "loss": 3.3926, "step": 8983 }, { "epoch": 1.71, "learning_rate": 2.4889990467111536e-06, "loss": 2.8846, "step": 8984 }, { "epoch": 1.71, "learning_rate": 2.4889418493803624e-06, "loss": 2.6346, "step": 8985 }, { "epoch": 1.71, "learning_rate": 2.488884652049571e-06, "loss": 2.9517, "step": 8986 }, { "epoch": 1.71, "learning_rate": 2.48882745471878e-06, "loss": 3.0808, "step": 8987 }, { "epoch": 1.71, "learning_rate": 2.4887702573879883e-06, "loss": 2.7091, "step": 8988 }, { "epoch": 1.71, "learning_rate": 2.4887130600571975e-06, "loss": 3.1677, "step": 8989 }, { "epoch": 1.71, "learning_rate": 2.488655862726406e-06, "loss": 2.7711, "step": 8990 }, { "epoch": 1.71, "learning_rate": 2.488598665395615e-06, "loss": 2.9089, "step": 8991 }, { "epoch": 1.71, "learning_rate": 2.488541468064824e-06, "loss": 2.4178, "step": 8992 }, { "epoch": 1.71, "learning_rate": 2.4884842707340323e-06, "loss": 3.753, "step": 8993 }, { "epoch": 1.71, "learning_rate": 2.4884270734032415e-06, "loss": 2.8865, "step": 8994 }, { "epoch": 1.71, "learning_rate": 2.48836987607245e-06, "loss": 2.5192, "step": 8995 }, { "epoch": 1.71, "learning_rate": 2.4883126787416587e-06, "loss": 2.6409, "step": 8996 }, { "epoch": 1.71, "learning_rate": 2.488255481410868e-06, "loss": 3.1819, "step": 8997 }, { "epoch": 1.71, "learning_rate": 2.4881982840800762e-06, "loss": 2.8763, "step": 8998 }, { "epoch": 1.71, "learning_rate": 2.488141086749285e-06, "loss": 2.5313, "step": 8999 }, { "epoch": 1.71, "learning_rate": 2.488083889418494e-06, "loss": 2.1167, "step": 9000 }, { "epoch": 1.71, "learning_rate": 2.4880266920877026e-06, "loss": 3.0833, "step": 9001 }, { "epoch": 1.71, "learning_rate": 2.4879694947569114e-06, "loss": 3.044, "step": 9002 }, { "epoch": 1.71, "learning_rate": 2.4879122974261202e-06, "loss": 2.5333, "step": 9003 }, { "epoch": 1.72, "learning_rate": 2.487855100095329e-06, "loss": 2.4578, "step": 9004 }, { "epoch": 1.72, "learning_rate": 2.4877979027645374e-06, "loss": 3.4804, "step": 9005 }, { "epoch": 1.72, "learning_rate": 2.4877407054337466e-06, "loss": 3.109, "step": 9006 }, { "epoch": 1.72, "learning_rate": 2.4876835081029554e-06, "loss": 2.8039, "step": 9007 }, { "epoch": 1.72, "learning_rate": 2.487626310772164e-06, "loss": 2.8497, "step": 9008 }, { "epoch": 1.72, "learning_rate": 2.487569113441373e-06, "loss": 3.2195, "step": 9009 }, { "epoch": 1.72, "learning_rate": 2.4875119161105813e-06, "loss": 3.4969, "step": 9010 }, { "epoch": 1.72, "learning_rate": 2.4874547187797906e-06, "loss": 2.768, "step": 9011 }, { "epoch": 1.72, "learning_rate": 2.487397521448999e-06, "loss": 3.1224, "step": 9012 }, { "epoch": 1.72, "learning_rate": 2.4873403241182077e-06, "loss": 3.9398, "step": 9013 }, { "epoch": 1.72, "learning_rate": 2.487283126787417e-06, "loss": 3.3697, "step": 9014 }, { "epoch": 1.72, "learning_rate": 2.4872259294566253e-06, "loss": 3.3427, "step": 9015 }, { "epoch": 1.72, "learning_rate": 2.487168732125834e-06, "loss": 2.7579, "step": 9016 }, { "epoch": 1.72, "learning_rate": 2.487111534795043e-06, "loss": 2.8572, "step": 9017 }, { "epoch": 1.72, "learning_rate": 2.4870543374642517e-06, "loss": 2.7229, "step": 9018 }, { "epoch": 1.72, "learning_rate": 2.4869971401334605e-06, "loss": 2.8574, "step": 9019 }, { "epoch": 1.72, "learning_rate": 2.4869399428026693e-06, "loss": 3.3309, "step": 9020 }, { "epoch": 1.72, "learning_rate": 2.486882745471878e-06, "loss": 3.444, "step": 9021 }, { "epoch": 1.72, "learning_rate": 2.4868255481410865e-06, "loss": 2.824, "step": 9022 }, { "epoch": 1.72, "learning_rate": 2.4867683508102957e-06, "loss": 2.8039, "step": 9023 }, { "epoch": 1.72, "learning_rate": 2.4867111534795045e-06, "loss": 3.2371, "step": 9024 }, { "epoch": 1.72, "learning_rate": 2.4866539561487133e-06, "loss": 2.6387, "step": 9025 }, { "epoch": 1.72, "learning_rate": 2.486596758817922e-06, "loss": 3.664, "step": 9026 }, { "epoch": 1.72, "learning_rate": 2.4865395614871304e-06, "loss": 2.406, "step": 9027 }, { "epoch": 1.72, "learning_rate": 2.4864823641563396e-06, "loss": 2.8136, "step": 9028 }, { "epoch": 1.72, "learning_rate": 2.486425166825548e-06, "loss": 3.2917, "step": 9029 }, { "epoch": 1.72, "learning_rate": 2.486367969494757e-06, "loss": 2.3527, "step": 9030 }, { "epoch": 1.72, "learning_rate": 2.486310772163966e-06, "loss": 2.8486, "step": 9031 }, { "epoch": 1.72, "learning_rate": 2.4862535748331744e-06, "loss": 2.266, "step": 9032 }, { "epoch": 1.72, "learning_rate": 2.486196377502383e-06, "loss": 2.9302, "step": 9033 }, { "epoch": 1.72, "learning_rate": 2.486139180171592e-06, "loss": 3.7425, "step": 9034 }, { "epoch": 1.72, "learning_rate": 2.4860819828408008e-06, "loss": 2.6623, "step": 9035 }, { "epoch": 1.72, "learning_rate": 2.4860247855100096e-06, "loss": 2.3936, "step": 9036 }, { "epoch": 1.72, "learning_rate": 2.4859675881792184e-06, "loss": 3.6729, "step": 9037 }, { "epoch": 1.72, "learning_rate": 2.485910390848427e-06, "loss": 2.8713, "step": 9038 }, { "epoch": 1.72, "learning_rate": 2.485853193517636e-06, "loss": 2.5309, "step": 9039 }, { "epoch": 1.72, "learning_rate": 2.4857959961868447e-06, "loss": 2.587, "step": 9040 }, { "epoch": 1.72, "learning_rate": 2.4857387988560535e-06, "loss": 3.2389, "step": 9041 }, { "epoch": 1.72, "learning_rate": 2.4856816015252623e-06, "loss": 3.5794, "step": 9042 }, { "epoch": 1.72, "learning_rate": 2.485624404194471e-06, "loss": 2.7761, "step": 9043 }, { "epoch": 1.72, "learning_rate": 2.4855672068636795e-06, "loss": 2.4651, "step": 9044 }, { "epoch": 1.72, "learning_rate": 2.4855100095328887e-06, "loss": 3.0703, "step": 9045 }, { "epoch": 1.72, "learning_rate": 2.4854528122020975e-06, "loss": 2.3476, "step": 9046 }, { "epoch": 1.72, "learning_rate": 2.485395614871306e-06, "loss": 2.8381, "step": 9047 }, { "epoch": 1.72, "learning_rate": 2.485338417540515e-06, "loss": 3.2524, "step": 9048 }, { "epoch": 1.72, "learning_rate": 2.4852812202097235e-06, "loss": 3.2514, "step": 9049 }, { "epoch": 1.72, "learning_rate": 2.4852240228789323e-06, "loss": 2.5701, "step": 9050 }, { "epoch": 1.72, "learning_rate": 2.485166825548141e-06, "loss": 2.3975, "step": 9051 }, { "epoch": 1.72, "learning_rate": 2.48510962821735e-06, "loss": 3.2427, "step": 9052 }, { "epoch": 1.72, "learning_rate": 2.4850524308865586e-06, "loss": 3.2415, "step": 9053 }, { "epoch": 1.72, "learning_rate": 2.4849952335557674e-06, "loss": 3.3064, "step": 9054 }, { "epoch": 1.72, "learning_rate": 2.4849380362249762e-06, "loss": 3.2381, "step": 9055 }, { "epoch": 1.72, "learning_rate": 2.484880838894185e-06, "loss": 2.8396, "step": 9056 }, { "epoch": 1.73, "learning_rate": 2.484823641563394e-06, "loss": 3.251, "step": 9057 }, { "epoch": 1.73, "learning_rate": 2.4847664442326026e-06, "loss": 2.4861, "step": 9058 }, { "epoch": 1.73, "learning_rate": 2.4847092469018114e-06, "loss": 3.0609, "step": 9059 }, { "epoch": 1.73, "learning_rate": 2.48465204957102e-06, "loss": 3.1038, "step": 9060 }, { "epoch": 1.73, "learning_rate": 2.4845948522402286e-06, "loss": 3.5448, "step": 9061 }, { "epoch": 1.73, "learning_rate": 2.4845376549094378e-06, "loss": 3.1341, "step": 9062 }, { "epoch": 1.73, "learning_rate": 2.4844804575786466e-06, "loss": 3.508, "step": 9063 }, { "epoch": 1.73, "learning_rate": 2.484423260247855e-06, "loss": 2.6598, "step": 9064 }, { "epoch": 1.73, "learning_rate": 2.484366062917064e-06, "loss": 3.0861, "step": 9065 }, { "epoch": 1.73, "learning_rate": 2.4843088655862725e-06, "loss": 2.9505, "step": 9066 }, { "epoch": 1.73, "learning_rate": 2.4842516682554813e-06, "loss": 3.0103, "step": 9067 }, { "epoch": 1.73, "learning_rate": 2.48419447092469e-06, "loss": 2.5154, "step": 9068 }, { "epoch": 1.73, "learning_rate": 2.484137273593899e-06, "loss": 2.9146, "step": 9069 }, { "epoch": 1.73, "learning_rate": 2.4840800762631077e-06, "loss": 3.0142, "step": 9070 }, { "epoch": 1.73, "learning_rate": 2.4840228789323165e-06, "loss": 3.3697, "step": 9071 }, { "epoch": 1.73, "learning_rate": 2.4839656816015253e-06, "loss": 2.9831, "step": 9072 }, { "epoch": 1.73, "learning_rate": 2.483908484270734e-06, "loss": 2.9007, "step": 9073 }, { "epoch": 1.73, "learning_rate": 2.483851286939943e-06, "loss": 2.8905, "step": 9074 }, { "epoch": 1.73, "learning_rate": 2.4837940896091517e-06, "loss": 2.3074, "step": 9075 }, { "epoch": 1.73, "learning_rate": 2.4837368922783605e-06, "loss": 3.3599, "step": 9076 }, { "epoch": 1.73, "learning_rate": 2.4836796949475693e-06, "loss": 2.8735, "step": 9077 }, { "epoch": 1.73, "learning_rate": 2.4836224976167776e-06, "loss": 2.1912, "step": 9078 }, { "epoch": 1.73, "learning_rate": 2.483565300285987e-06, "loss": 2.7683, "step": 9079 }, { "epoch": 1.73, "learning_rate": 2.4835081029551956e-06, "loss": 3.5033, "step": 9080 }, { "epoch": 1.73, "learning_rate": 2.483450905624404e-06, "loss": 2.5464, "step": 9081 }, { "epoch": 1.73, "learning_rate": 2.4833937082936132e-06, "loss": 3.251, "step": 9082 }, { "epoch": 1.73, "learning_rate": 2.4833365109628216e-06, "loss": 3.1821, "step": 9083 }, { "epoch": 1.73, "learning_rate": 2.4832793136320304e-06, "loss": 2.8687, "step": 9084 }, { "epoch": 1.73, "learning_rate": 2.4832221163012396e-06, "loss": 2.552, "step": 9085 }, { "epoch": 1.73, "learning_rate": 2.483164918970448e-06, "loss": 3.2548, "step": 9086 }, { "epoch": 1.73, "learning_rate": 2.4831077216396568e-06, "loss": 3.3573, "step": 9087 }, { "epoch": 1.73, "learning_rate": 2.4830505243088656e-06, "loss": 2.6275, "step": 9088 }, { "epoch": 1.73, "learning_rate": 2.4829933269780744e-06, "loss": 2.5096, "step": 9089 }, { "epoch": 1.73, "learning_rate": 2.482936129647283e-06, "loss": 3.1319, "step": 9090 }, { "epoch": 1.73, "learning_rate": 2.482878932316492e-06, "loss": 3.742, "step": 9091 }, { "epoch": 1.73, "learning_rate": 2.4828217349857007e-06, "loss": 3.4192, "step": 9092 }, { "epoch": 1.73, "learning_rate": 2.4827645376549095e-06, "loss": 3.2653, "step": 9093 }, { "epoch": 1.73, "learning_rate": 2.4827073403241183e-06, "loss": 2.8047, "step": 9094 }, { "epoch": 1.73, "learning_rate": 2.482650142993327e-06, "loss": 2.7866, "step": 9095 }, { "epoch": 1.73, "learning_rate": 2.482592945662536e-06, "loss": 2.8033, "step": 9096 }, { "epoch": 1.73, "learning_rate": 2.4825357483317447e-06, "loss": 2.7591, "step": 9097 }, { "epoch": 1.73, "learning_rate": 2.482478551000953e-06, "loss": 2.9483, "step": 9098 }, { "epoch": 1.73, "learning_rate": 2.4824213536701623e-06, "loss": 2.7434, "step": 9099 }, { "epoch": 1.73, "learning_rate": 2.4823641563393707e-06, "loss": 3.1434, "step": 9100 }, { "epoch": 1.73, "learning_rate": 2.4823069590085795e-06, "loss": 3.2281, "step": 9101 }, { "epoch": 1.73, "learning_rate": 2.4822497616777887e-06, "loss": 2.9641, "step": 9102 }, { "epoch": 1.73, "learning_rate": 2.482192564346997e-06, "loss": 3.1006, "step": 9103 }, { "epoch": 1.73, "learning_rate": 2.4821353670162063e-06, "loss": 3.0383, "step": 9104 }, { "epoch": 1.73, "learning_rate": 2.4820781696854146e-06, "loss": 3.8392, "step": 9105 }, { "epoch": 1.73, "learning_rate": 2.4820209723546234e-06, "loss": 2.2823, "step": 9106 }, { "epoch": 1.73, "learning_rate": 2.4819637750238322e-06, "loss": 2.8573, "step": 9107 }, { "epoch": 1.73, "learning_rate": 2.481906577693041e-06, "loss": 3.7131, "step": 9108 }, { "epoch": 1.74, "learning_rate": 2.48184938036225e-06, "loss": 3.0367, "step": 9109 }, { "epoch": 1.74, "learning_rate": 2.4817921830314586e-06, "loss": 2.8594, "step": 9110 }, { "epoch": 1.74, "learning_rate": 2.4817349857006674e-06, "loss": 2.4506, "step": 9111 }, { "epoch": 1.74, "learning_rate": 2.481677788369876e-06, "loss": 3.214, "step": 9112 }, { "epoch": 1.74, "learning_rate": 2.481620591039085e-06, "loss": 3.3594, "step": 9113 }, { "epoch": 1.74, "learning_rate": 2.481563393708294e-06, "loss": 3.4788, "step": 9114 }, { "epoch": 1.74, "learning_rate": 2.481506196377502e-06, "loss": 3.0647, "step": 9115 }, { "epoch": 1.74, "learning_rate": 2.4814489990467114e-06, "loss": 3.7892, "step": 9116 }, { "epoch": 1.74, "learning_rate": 2.4813918017159197e-06, "loss": 3.1517, "step": 9117 }, { "epoch": 1.74, "learning_rate": 2.4813346043851285e-06, "loss": 2.9953, "step": 9118 }, { "epoch": 1.74, "learning_rate": 2.4812774070543378e-06, "loss": 2.8204, "step": 9119 }, { "epoch": 1.74, "learning_rate": 2.481220209723546e-06, "loss": 2.8205, "step": 9120 }, { "epoch": 1.74, "learning_rate": 2.4811630123927553e-06, "loss": 2.8543, "step": 9121 }, { "epoch": 1.74, "learning_rate": 2.4811058150619637e-06, "loss": 3.1136, "step": 9122 }, { "epoch": 1.74, "learning_rate": 2.4810486177311725e-06, "loss": 2.3062, "step": 9123 }, { "epoch": 1.74, "learning_rate": 2.4809914204003817e-06, "loss": 2.805, "step": 9124 }, { "epoch": 1.74, "learning_rate": 2.48093422306959e-06, "loss": 3.6709, "step": 9125 }, { "epoch": 1.74, "learning_rate": 2.480877025738799e-06, "loss": 3.0822, "step": 9126 }, { "epoch": 1.74, "learning_rate": 2.4808198284080077e-06, "loss": 2.2603, "step": 9127 }, { "epoch": 1.74, "learning_rate": 2.4807626310772165e-06, "loss": 3.2189, "step": 9128 }, { "epoch": 1.74, "learning_rate": 2.4807054337464253e-06, "loss": 3.5596, "step": 9129 }, { "epoch": 1.74, "learning_rate": 2.480648236415634e-06, "loss": 2.232, "step": 9130 }, { "epoch": 1.74, "learning_rate": 2.480591039084843e-06, "loss": 2.9486, "step": 9131 }, { "epoch": 1.74, "learning_rate": 2.4805338417540512e-06, "loss": 2.9564, "step": 9132 }, { "epoch": 1.74, "learning_rate": 2.4804766444232604e-06, "loss": 2.1587, "step": 9133 }, { "epoch": 1.74, "learning_rate": 2.4804194470924692e-06, "loss": 3.1985, "step": 9134 }, { "epoch": 1.74, "learning_rate": 2.4803622497616776e-06, "loss": 3.0455, "step": 9135 }, { "epoch": 1.74, "learning_rate": 2.480305052430887e-06, "loss": 3.0195, "step": 9136 }, { "epoch": 1.74, "learning_rate": 2.480247855100095e-06, "loss": 2.5143, "step": 9137 }, { "epoch": 1.74, "learning_rate": 2.4801906577693044e-06, "loss": 3.4008, "step": 9138 }, { "epoch": 1.74, "learning_rate": 2.4801334604385128e-06, "loss": 2.6428, "step": 9139 }, { "epoch": 1.74, "learning_rate": 2.4800762631077216e-06, "loss": 3.9309, "step": 9140 }, { "epoch": 1.74, "learning_rate": 2.480019065776931e-06, "loss": 3.3389, "step": 9141 }, { "epoch": 1.74, "learning_rate": 2.479961868446139e-06, "loss": 2.9976, "step": 9142 }, { "epoch": 1.74, "learning_rate": 2.479904671115348e-06, "loss": 2.6474, "step": 9143 }, { "epoch": 1.74, "learning_rate": 2.4798474737845568e-06, "loss": 2.0817, "step": 9144 }, { "epoch": 1.74, "learning_rate": 2.4797902764537655e-06, "loss": 3.2559, "step": 9145 }, { "epoch": 1.74, "learning_rate": 2.4797330791229743e-06, "loss": 2.7368, "step": 9146 }, { "epoch": 1.74, "learning_rate": 2.479675881792183e-06, "loss": 2.5377, "step": 9147 }, { "epoch": 1.74, "learning_rate": 2.479618684461392e-06, "loss": 3.5495, "step": 9148 }, { "epoch": 1.74, "learning_rate": 2.4795614871306003e-06, "loss": 2.8523, "step": 9149 }, { "epoch": 1.74, "learning_rate": 2.4795042897998095e-06, "loss": 2.9187, "step": 9150 }, { "epoch": 1.74, "learning_rate": 2.4794470924690183e-06, "loss": 2.2262, "step": 9151 }, { "epoch": 1.74, "learning_rate": 2.4793898951382267e-06, "loss": 3.1153, "step": 9152 }, { "epoch": 1.74, "learning_rate": 2.479332697807436e-06, "loss": 2.1726, "step": 9153 }, { "epoch": 1.74, "learning_rate": 2.4792755004766443e-06, "loss": 3.5206, "step": 9154 }, { "epoch": 1.74, "learning_rate": 2.4792183031458535e-06, "loss": 2.7264, "step": 9155 }, { "epoch": 1.74, "learning_rate": 2.479161105815062e-06, "loss": 3.2601, "step": 9156 }, { "epoch": 1.74, "learning_rate": 2.4791039084842707e-06, "loss": 2.9212, "step": 9157 }, { "epoch": 1.74, "learning_rate": 2.47904671115348e-06, "loss": 2.2223, "step": 9158 }, { "epoch": 1.74, "learning_rate": 2.4789895138226882e-06, "loss": 2.9834, "step": 9159 }, { "epoch": 1.74, "learning_rate": 2.478932316491897e-06, "loss": 2.7416, "step": 9160 }, { "epoch": 1.74, "learning_rate": 2.478875119161106e-06, "loss": 3.4962, "step": 9161 }, { "epoch": 1.75, "learning_rate": 2.4788179218303146e-06, "loss": 3.7155, "step": 9162 }, { "epoch": 1.75, "learning_rate": 2.4787607244995234e-06, "loss": 2.5086, "step": 9163 }, { "epoch": 1.75, "learning_rate": 2.478703527168732e-06, "loss": 3.0468, "step": 9164 }, { "epoch": 1.75, "learning_rate": 2.478646329837941e-06, "loss": 3.0076, "step": 9165 }, { "epoch": 1.75, "learning_rate": 2.4785891325071494e-06, "loss": 2.7158, "step": 9166 }, { "epoch": 1.75, "learning_rate": 2.4785319351763586e-06, "loss": 3.1019, "step": 9167 }, { "epoch": 1.75, "learning_rate": 2.4784747378455674e-06, "loss": 2.7518, "step": 9168 }, { "epoch": 1.75, "learning_rate": 2.478417540514776e-06, "loss": 2.6459, "step": 9169 }, { "epoch": 1.75, "learning_rate": 2.478360343183985e-06, "loss": 3.9145, "step": 9170 }, { "epoch": 1.75, "learning_rate": 2.4783031458531933e-06, "loss": 3.0796, "step": 9171 }, { "epoch": 1.75, "learning_rate": 2.4782459485224026e-06, "loss": 3.2649, "step": 9172 }, { "epoch": 1.75, "learning_rate": 2.4781887511916114e-06, "loss": 3.3083, "step": 9173 }, { "epoch": 1.75, "learning_rate": 2.4781315538608197e-06, "loss": 2.5279, "step": 9174 }, { "epoch": 1.75, "learning_rate": 2.478074356530029e-06, "loss": 3.0813, "step": 9175 }, { "epoch": 1.75, "learning_rate": 2.4780171591992373e-06, "loss": 3.1142, "step": 9176 }, { "epoch": 1.75, "learning_rate": 2.477959961868446e-06, "loss": 3.1022, "step": 9177 }, { "epoch": 1.75, "learning_rate": 2.477902764537655e-06, "loss": 2.9508, "step": 9178 }, { "epoch": 1.75, "learning_rate": 2.4778455672068637e-06, "loss": 3.1451, "step": 9179 }, { "epoch": 1.75, "learning_rate": 2.4777883698760725e-06, "loss": 3.4255, "step": 9180 }, { "epoch": 1.75, "learning_rate": 2.4777311725452813e-06, "loss": 2.9364, "step": 9181 }, { "epoch": 1.75, "learning_rate": 2.47767397521449e-06, "loss": 2.6738, "step": 9182 }, { "epoch": 1.75, "learning_rate": 2.477616777883699e-06, "loss": 3.1616, "step": 9183 }, { "epoch": 1.75, "learning_rate": 2.4775595805529077e-06, "loss": 2.8563, "step": 9184 }, { "epoch": 1.75, "learning_rate": 2.4775023832221165e-06, "loss": 2.7638, "step": 9185 }, { "epoch": 1.75, "learning_rate": 2.4774451858913252e-06, "loss": 2.9885, "step": 9186 }, { "epoch": 1.75, "learning_rate": 2.477387988560534e-06, "loss": 3.5912, "step": 9187 }, { "epoch": 1.75, "learning_rate": 2.4773307912297424e-06, "loss": 2.4731, "step": 9188 }, { "epoch": 1.75, "learning_rate": 2.4772735938989516e-06, "loss": 3.0768, "step": 9189 }, { "epoch": 1.75, "learning_rate": 2.4772163965681604e-06, "loss": 2.8986, "step": 9190 }, { "epoch": 1.75, "learning_rate": 2.477159199237369e-06, "loss": 2.4409, "step": 9191 }, { "epoch": 1.75, "learning_rate": 2.477102001906578e-06, "loss": 3.7644, "step": 9192 }, { "epoch": 1.75, "learning_rate": 2.4770448045757864e-06, "loss": 2.7992, "step": 9193 }, { "epoch": 1.75, "learning_rate": 2.476987607244995e-06, "loss": 2.791, "step": 9194 }, { "epoch": 1.75, "learning_rate": 2.476930409914204e-06, "loss": 3.272, "step": 9195 }, { "epoch": 1.75, "learning_rate": 2.4768732125834128e-06, "loss": 3.9528, "step": 9196 }, { "epoch": 1.75, "learning_rate": 2.4768160152526216e-06, "loss": 3.066, "step": 9197 }, { "epoch": 1.75, "learning_rate": 2.4767588179218304e-06, "loss": 2.9826, "step": 9198 }, { "epoch": 1.75, "learning_rate": 2.476701620591039e-06, "loss": 2.6829, "step": 9199 }, { "epoch": 1.75, "learning_rate": 2.476644423260248e-06, "loss": 2.8858, "step": 9200 }, { "epoch": 1.75, "learning_rate": 2.4765872259294567e-06, "loss": 2.4023, "step": 9201 }, { "epoch": 1.75, "learning_rate": 2.4765300285986655e-06, "loss": 3.2381, "step": 9202 }, { "epoch": 1.75, "learning_rate": 2.4764728312678743e-06, "loss": 2.934, "step": 9203 }, { "epoch": 1.75, "learning_rate": 2.476415633937083e-06, "loss": 2.8172, "step": 9204 }, { "epoch": 1.75, "learning_rate": 2.4763584366062915e-06, "loss": 2.9274, "step": 9205 }, { "epoch": 1.75, "learning_rate": 2.4763012392755007e-06, "loss": 3.15, "step": 9206 }, { "epoch": 1.75, "learning_rate": 2.4762440419447095e-06, "loss": 2.8176, "step": 9207 }, { "epoch": 1.75, "learning_rate": 2.476186844613918e-06, "loss": 3.0904, "step": 9208 }, { "epoch": 1.75, "learning_rate": 2.476129647283127e-06, "loss": 3.1659, "step": 9209 }, { "epoch": 1.75, "learning_rate": 2.4760724499523355e-06, "loss": 3.3239, "step": 9210 }, { "epoch": 1.75, "learning_rate": 2.4760152526215442e-06, "loss": 3.0697, "step": 9211 }, { "epoch": 1.75, "learning_rate": 2.475958055290753e-06, "loss": 3.6049, "step": 9212 }, { "epoch": 1.75, "learning_rate": 2.475900857959962e-06, "loss": 2.7562, "step": 9213 }, { "epoch": 1.76, "learning_rate": 2.4758436606291706e-06, "loss": 3.3099, "step": 9214 }, { "epoch": 1.76, "learning_rate": 2.4757864632983794e-06, "loss": 3.309, "step": 9215 }, { "epoch": 1.76, "learning_rate": 2.4757292659675882e-06, "loss": 2.5261, "step": 9216 }, { "epoch": 1.76, "learning_rate": 2.475672068636797e-06, "loss": 2.5128, "step": 9217 }, { "epoch": 1.76, "learning_rate": 2.475614871306006e-06, "loss": 3.1524, "step": 9218 }, { "epoch": 1.76, "learning_rate": 2.4755576739752146e-06, "loss": 3.287, "step": 9219 }, { "epoch": 1.76, "learning_rate": 2.4755004766444234e-06, "loss": 3.2851, "step": 9220 }, { "epoch": 1.76, "learning_rate": 2.475443279313632e-06, "loss": 2.9824, "step": 9221 }, { "epoch": 1.76, "learning_rate": 2.475386081982841e-06, "loss": 3.3747, "step": 9222 }, { "epoch": 1.76, "learning_rate": 2.4753288846520498e-06, "loss": 3.5782, "step": 9223 }, { "epoch": 1.76, "learning_rate": 2.4752716873212586e-06, "loss": 3.4743, "step": 9224 }, { "epoch": 1.76, "learning_rate": 2.475214489990467e-06, "loss": 2.7031, "step": 9225 }, { "epoch": 1.76, "learning_rate": 2.475157292659676e-06, "loss": 3.0704, "step": 9226 }, { "epoch": 1.76, "learning_rate": 2.4751000953288845e-06, "loss": 2.9911, "step": 9227 }, { "epoch": 1.76, "learning_rate": 2.4750428979980933e-06, "loss": 2.8554, "step": 9228 }, { "epoch": 1.76, "learning_rate": 2.4749857006673025e-06, "loss": 3.2465, "step": 9229 }, { "epoch": 1.76, "learning_rate": 2.474928503336511e-06, "loss": 3.1529, "step": 9230 }, { "epoch": 1.76, "learning_rate": 2.4748713060057197e-06, "loss": 3.2235, "step": 9231 }, { "epoch": 1.76, "learning_rate": 2.4748141086749285e-06, "loss": 2.4175, "step": 9232 }, { "epoch": 1.76, "learning_rate": 2.4747569113441373e-06, "loss": 3.4763, "step": 9233 }, { "epoch": 1.76, "learning_rate": 2.474699714013346e-06, "loss": 2.3729, "step": 9234 }, { "epoch": 1.76, "learning_rate": 2.474642516682555e-06, "loss": 3.1355, "step": 9235 }, { "epoch": 1.76, "learning_rate": 2.4745853193517637e-06, "loss": 3.8191, "step": 9236 }, { "epoch": 1.76, "learning_rate": 2.4745281220209725e-06, "loss": 3.2645, "step": 9237 }, { "epoch": 1.76, "learning_rate": 2.4744709246901813e-06, "loss": 3.1089, "step": 9238 }, { "epoch": 1.76, "learning_rate": 2.47441372735939e-06, "loss": 2.6994, "step": 9239 }, { "epoch": 1.76, "learning_rate": 2.474356530028599e-06, "loss": 3.0607, "step": 9240 }, { "epoch": 1.76, "learning_rate": 2.4742993326978076e-06, "loss": 3.2434, "step": 9241 }, { "epoch": 1.76, "learning_rate": 2.474242135367016e-06, "loss": 2.5083, "step": 9242 }, { "epoch": 1.76, "learning_rate": 2.4741849380362252e-06, "loss": 3.2485, "step": 9243 }, { "epoch": 1.76, "learning_rate": 2.4741277407054336e-06, "loss": 2.5122, "step": 9244 }, { "epoch": 1.76, "learning_rate": 2.4740705433746424e-06, "loss": 3.1476, "step": 9245 }, { "epoch": 1.76, "learning_rate": 2.4740133460438516e-06, "loss": 3.0744, "step": 9246 }, { "epoch": 1.76, "learning_rate": 2.47395614871306e-06, "loss": 2.5142, "step": 9247 }, { "epoch": 1.76, "learning_rate": 2.4738989513822688e-06, "loss": 2.9263, "step": 9248 }, { "epoch": 1.76, "learning_rate": 2.4738417540514776e-06, "loss": 3.2677, "step": 9249 }, { "epoch": 1.76, "learning_rate": 2.4737845567206864e-06, "loss": 3.0423, "step": 9250 }, { "epoch": 1.76, "learning_rate": 2.473727359389895e-06, "loss": 3.348, "step": 9251 }, { "epoch": 1.76, "learning_rate": 2.473670162059104e-06, "loss": 2.4986, "step": 9252 }, { "epoch": 1.76, "learning_rate": 2.4736129647283127e-06, "loss": 3.1803, "step": 9253 }, { "epoch": 1.76, "learning_rate": 2.4735557673975215e-06, "loss": 2.5548, "step": 9254 }, { "epoch": 1.76, "learning_rate": 2.4734985700667303e-06, "loss": 2.8571, "step": 9255 }, { "epoch": 1.76, "learning_rate": 2.473441372735939e-06, "loss": 2.674, "step": 9256 }, { "epoch": 1.76, "learning_rate": 2.473384175405148e-06, "loss": 3.4426, "step": 9257 }, { "epoch": 1.76, "learning_rate": 2.4733269780743567e-06, "loss": 3.1022, "step": 9258 }, { "epoch": 1.76, "learning_rate": 2.473269780743565e-06, "loss": 3.1267, "step": 9259 }, { "epoch": 1.76, "learning_rate": 2.4732125834127743e-06, "loss": 2.7132, "step": 9260 }, { "epoch": 1.76, "learning_rate": 2.473155386081983e-06, "loss": 2.3814, "step": 9261 }, { "epoch": 1.76, "learning_rate": 2.4730981887511915e-06, "loss": 2.7777, "step": 9262 }, { "epoch": 1.76, "learning_rate": 2.4730409914204007e-06, "loss": 3.8984, "step": 9263 }, { "epoch": 1.76, "learning_rate": 2.472983794089609e-06, "loss": 3.1187, "step": 9264 }, { "epoch": 1.76, "learning_rate": 2.472926596758818e-06, "loss": 3.4937, "step": 9265 }, { "epoch": 1.76, "learning_rate": 2.4728693994280266e-06, "loss": 2.8909, "step": 9266 }, { "epoch": 1.77, "learning_rate": 2.4728122020972354e-06, "loss": 3.1259, "step": 9267 }, { "epoch": 1.77, "learning_rate": 2.4727550047664446e-06, "loss": 3.2633, "step": 9268 }, { "epoch": 1.77, "learning_rate": 2.472697807435653e-06, "loss": 2.8868, "step": 9269 }, { "epoch": 1.77, "learning_rate": 2.472640610104862e-06, "loss": 3.0628, "step": 9270 }, { "epoch": 1.77, "learning_rate": 2.4725834127740706e-06, "loss": 3.0149, "step": 9271 }, { "epoch": 1.77, "learning_rate": 2.4725262154432794e-06, "loss": 2.7964, "step": 9272 }, { "epoch": 1.77, "learning_rate": 2.472469018112488e-06, "loss": 3.0059, "step": 9273 }, { "epoch": 1.77, "learning_rate": 2.472411820781697e-06, "loss": 2.0852, "step": 9274 }, { "epoch": 1.77, "learning_rate": 2.4723546234509058e-06, "loss": 2.5261, "step": 9275 }, { "epoch": 1.77, "learning_rate": 2.472297426120114e-06, "loss": 2.6774, "step": 9276 }, { "epoch": 1.77, "learning_rate": 2.4722402287893234e-06, "loss": 3.8669, "step": 9277 }, { "epoch": 1.77, "learning_rate": 2.472183031458532e-06, "loss": 2.8364, "step": 9278 }, { "epoch": 1.77, "learning_rate": 2.4721258341277405e-06, "loss": 2.9314, "step": 9279 }, { "epoch": 1.77, "learning_rate": 2.4720686367969498e-06, "loss": 3.5749, "step": 9280 }, { "epoch": 1.77, "learning_rate": 2.472011439466158e-06, "loss": 3.0578, "step": 9281 }, { "epoch": 1.77, "learning_rate": 2.471954242135367e-06, "loss": 3.1966, "step": 9282 }, { "epoch": 1.77, "learning_rate": 2.4718970448045757e-06, "loss": 3.0669, "step": 9283 }, { "epoch": 1.77, "learning_rate": 2.4718398474737845e-06, "loss": 2.5324, "step": 9284 }, { "epoch": 1.77, "learning_rate": 2.4717826501429937e-06, "loss": 2.2505, "step": 9285 }, { "epoch": 1.77, "learning_rate": 2.471725452812202e-06, "loss": 3.1719, "step": 9286 }, { "epoch": 1.77, "learning_rate": 2.471668255481411e-06, "loss": 2.9235, "step": 9287 }, { "epoch": 1.77, "learning_rate": 2.4716110581506197e-06, "loss": 3.0642, "step": 9288 }, { "epoch": 1.77, "learning_rate": 2.4715538608198285e-06, "loss": 3.2976, "step": 9289 }, { "epoch": 1.77, "learning_rate": 2.4714966634890373e-06, "loss": 3.1483, "step": 9290 }, { "epoch": 1.77, "learning_rate": 2.471439466158246e-06, "loss": 2.5291, "step": 9291 }, { "epoch": 1.77, "learning_rate": 2.471382268827455e-06, "loss": 3.4521, "step": 9292 }, { "epoch": 1.77, "learning_rate": 2.4713250714966632e-06, "loss": 2.6874, "step": 9293 }, { "epoch": 1.77, "learning_rate": 2.4712678741658724e-06, "loss": 2.6448, "step": 9294 }, { "epoch": 1.77, "learning_rate": 2.4712106768350812e-06, "loss": 3.2996, "step": 9295 }, { "epoch": 1.77, "learning_rate": 2.4711534795042896e-06, "loss": 2.6509, "step": 9296 }, { "epoch": 1.77, "learning_rate": 2.471096282173499e-06, "loss": 2.6249, "step": 9297 }, { "epoch": 1.77, "learning_rate": 2.471039084842707e-06, "loss": 2.5359, "step": 9298 }, { "epoch": 1.77, "learning_rate": 2.4709818875119164e-06, "loss": 3.2027, "step": 9299 }, { "epoch": 1.77, "learning_rate": 2.4709246901811248e-06, "loss": 3.4478, "step": 9300 }, { "epoch": 1.77, "learning_rate": 2.4708674928503336e-06, "loss": 3.17, "step": 9301 }, { "epoch": 1.77, "learning_rate": 2.470810295519543e-06, "loss": 3.2635, "step": 9302 }, { "epoch": 1.77, "learning_rate": 2.470753098188751e-06, "loss": 2.4226, "step": 9303 }, { "epoch": 1.77, "learning_rate": 2.47069590085796e-06, "loss": 3.2848, "step": 9304 }, { "epoch": 1.77, "learning_rate": 2.4706387035271687e-06, "loss": 2.6885, "step": 9305 }, { "epoch": 1.77, "learning_rate": 2.4705815061963775e-06, "loss": 2.6958, "step": 9306 }, { "epoch": 1.77, "learning_rate": 2.4705243088655863e-06, "loss": 3.0217, "step": 9307 }, { "epoch": 1.77, "learning_rate": 2.470467111534795e-06, "loss": 2.6441, "step": 9308 }, { "epoch": 1.77, "learning_rate": 2.470409914204004e-06, "loss": 2.3119, "step": 9309 }, { "epoch": 1.77, "learning_rate": 2.4703527168732127e-06, "loss": 2.7129, "step": 9310 }, { "epoch": 1.77, "learning_rate": 2.4702955195424215e-06, "loss": 2.6195, "step": 9311 }, { "epoch": 1.77, "learning_rate": 2.4702383222116303e-06, "loss": 2.563, "step": 9312 }, { "epoch": 1.77, "learning_rate": 2.4701811248808387e-06, "loss": 2.5271, "step": 9313 }, { "epoch": 1.77, "learning_rate": 2.470123927550048e-06, "loss": 2.3602, "step": 9314 }, { "epoch": 1.77, "learning_rate": 2.4700667302192563e-06, "loss": 3.0649, "step": 9315 }, { "epoch": 1.77, "learning_rate": 2.4700095328884655e-06, "loss": 2.5313, "step": 9316 }, { "epoch": 1.77, "learning_rate": 2.4699523355576743e-06, "loss": 3.305, "step": 9317 }, { "epoch": 1.77, "learning_rate": 2.4698951382268826e-06, "loss": 3.1375, "step": 9318 }, { "epoch": 1.78, "learning_rate": 2.469837940896092e-06, "loss": 3.5316, "step": 9319 }, { "epoch": 1.78, "learning_rate": 2.4697807435653002e-06, "loss": 2.4138, "step": 9320 }, { "epoch": 1.78, "learning_rate": 2.469723546234509e-06, "loss": 3.2045, "step": 9321 }, { "epoch": 1.78, "learning_rate": 2.469666348903718e-06, "loss": 3.3761, "step": 9322 }, { "epoch": 1.78, "learning_rate": 2.4696091515729266e-06, "loss": 2.4529, "step": 9323 }, { "epoch": 1.78, "learning_rate": 2.4695519542421354e-06, "loss": 3.2253, "step": 9324 }, { "epoch": 1.78, "learning_rate": 2.469494756911344e-06, "loss": 3.0438, "step": 9325 }, { "epoch": 1.78, "learning_rate": 2.469437559580553e-06, "loss": 3.1831, "step": 9326 }, { "epoch": 1.78, "learning_rate": 2.469380362249762e-06, "loss": 3.2323, "step": 9327 }, { "epoch": 1.78, "learning_rate": 2.4693231649189706e-06, "loss": 2.2631, "step": 9328 }, { "epoch": 1.78, "learning_rate": 2.4692659675881794e-06, "loss": 2.3237, "step": 9329 }, { "epoch": 1.78, "learning_rate": 2.4692087702573877e-06, "loss": 2.7801, "step": 9330 }, { "epoch": 1.78, "learning_rate": 2.469151572926597e-06, "loss": 3.0656, "step": 9331 }, { "epoch": 1.78, "learning_rate": 2.4690943755958053e-06, "loss": 3.2291, "step": 9332 }, { "epoch": 1.78, "learning_rate": 2.4690371782650146e-06, "loss": 3.1405, "step": 9333 }, { "epoch": 1.78, "learning_rate": 2.4689799809342233e-06, "loss": 3.1284, "step": 9334 }, { "epoch": 1.78, "learning_rate": 2.4689227836034317e-06, "loss": 2.1809, "step": 9335 }, { "epoch": 1.78, "learning_rate": 2.468865586272641e-06, "loss": 3.4502, "step": 9336 }, { "epoch": 1.78, "learning_rate": 2.4688083889418493e-06, "loss": 3.5722, "step": 9337 }, { "epoch": 1.78, "learning_rate": 2.468751191611058e-06, "loss": 2.7138, "step": 9338 }, { "epoch": 1.78, "learning_rate": 2.468693994280267e-06, "loss": 2.8082, "step": 9339 }, { "epoch": 1.78, "learning_rate": 2.4686367969494757e-06, "loss": 3.0213, "step": 9340 }, { "epoch": 1.78, "learning_rate": 2.4685795996186845e-06, "loss": 3.1344, "step": 9341 }, { "epoch": 1.78, "learning_rate": 2.4685224022878933e-06, "loss": 2.7907, "step": 9342 }, { "epoch": 1.78, "learning_rate": 2.468465204957102e-06, "loss": 3.3929, "step": 9343 }, { "epoch": 1.78, "learning_rate": 2.468408007626311e-06, "loss": 3.6195, "step": 9344 }, { "epoch": 1.78, "learning_rate": 2.4683508102955197e-06, "loss": 3.1878, "step": 9345 }, { "epoch": 1.78, "learning_rate": 2.4682936129647284e-06, "loss": 2.8934, "step": 9346 }, { "epoch": 1.78, "learning_rate": 2.468236415633937e-06, "loss": 3.0315, "step": 9347 }, { "epoch": 1.78, "learning_rate": 2.468179218303146e-06, "loss": 3.423, "step": 9348 }, { "epoch": 1.78, "learning_rate": 2.468122020972355e-06, "loss": 3.7264, "step": 9349 }, { "epoch": 1.78, "learning_rate": 2.4680648236415636e-06, "loss": 2.7708, "step": 9350 }, { "epoch": 1.78, "learning_rate": 2.4680076263107724e-06, "loss": 2.6492, "step": 9351 }, { "epoch": 1.78, "learning_rate": 2.4679504289799808e-06, "loss": 3.2979, "step": 9352 }, { "epoch": 1.78, "learning_rate": 2.46789323164919e-06, "loss": 2.0005, "step": 9353 }, { "epoch": 1.78, "learning_rate": 2.4678360343183984e-06, "loss": 3.7812, "step": 9354 }, { "epoch": 1.78, "learning_rate": 2.467778836987607e-06, "loss": 3.1832, "step": 9355 }, { "epoch": 1.78, "learning_rate": 2.4677216396568164e-06, "loss": 3.6659, "step": 9356 }, { "epoch": 1.78, "learning_rate": 2.4676644423260248e-06, "loss": 2.8209, "step": 9357 }, { "epoch": 1.78, "learning_rate": 2.4676072449952336e-06, "loss": 3.195, "step": 9358 }, { "epoch": 1.78, "learning_rate": 2.4675500476644423e-06, "loss": 2.5987, "step": 9359 }, { "epoch": 1.78, "learning_rate": 2.467492850333651e-06, "loss": 3.2962, "step": 9360 }, { "epoch": 1.78, "learning_rate": 2.46743565300286e-06, "loss": 2.7273, "step": 9361 }, { "epoch": 1.78, "learning_rate": 2.4673784556720687e-06, "loss": 3.0751, "step": 9362 }, { "epoch": 1.78, "learning_rate": 2.4673212583412775e-06, "loss": 2.4804, "step": 9363 }, { "epoch": 1.78, "learning_rate": 2.4672640610104863e-06, "loss": 3.247, "step": 9364 }, { "epoch": 1.78, "learning_rate": 2.467206863679695e-06, "loss": 3.2505, "step": 9365 }, { "epoch": 1.78, "learning_rate": 2.467149666348904e-06, "loss": 3.3586, "step": 9366 }, { "epoch": 1.78, "learning_rate": 2.4670924690181127e-06, "loss": 3.1189, "step": 9367 }, { "epoch": 1.78, "learning_rate": 2.4670352716873215e-06, "loss": 3.5884, "step": 9368 }, { "epoch": 1.78, "learning_rate": 2.46697807435653e-06, "loss": 2.992, "step": 9369 }, { "epoch": 1.78, "learning_rate": 2.466920877025739e-06, "loss": 2.7016, "step": 9370 }, { "epoch": 1.78, "learning_rate": 2.4668636796949474e-06, "loss": 3.3355, "step": 9371 }, { "epoch": 1.79, "learning_rate": 2.4668064823641562e-06, "loss": 3.2623, "step": 9372 }, { "epoch": 1.79, "learning_rate": 2.4667492850333655e-06, "loss": 3.5921, "step": 9373 }, { "epoch": 1.79, "learning_rate": 2.466692087702574e-06, "loss": 3.1546, "step": 9374 }, { "epoch": 1.79, "learning_rate": 2.4666348903717826e-06, "loss": 3.1033, "step": 9375 }, { "epoch": 1.79, "learning_rate": 2.4665776930409914e-06, "loss": 2.7589, "step": 9376 }, { "epoch": 1.79, "learning_rate": 2.4665204957102e-06, "loss": 2.5783, "step": 9377 }, { "epoch": 1.79, "learning_rate": 2.466463298379409e-06, "loss": 2.8665, "step": 9378 }, { "epoch": 1.79, "learning_rate": 2.466406101048618e-06, "loss": 3.6673, "step": 9379 }, { "epoch": 1.79, "learning_rate": 2.4663489037178266e-06, "loss": 2.7852, "step": 9380 }, { "epoch": 1.79, "learning_rate": 2.4662917063870354e-06, "loss": 3.7688, "step": 9381 }, { "epoch": 1.79, "learning_rate": 2.466234509056244e-06, "loss": 2.9578, "step": 9382 }, { "epoch": 1.79, "learning_rate": 2.466177311725453e-06, "loss": 2.4275, "step": 9383 }, { "epoch": 1.79, "learning_rate": 2.4661201143946618e-06, "loss": 3.3808, "step": 9384 }, { "epoch": 1.79, "learning_rate": 2.4660629170638706e-06, "loss": 3.2164, "step": 9385 }, { "epoch": 1.79, "learning_rate": 2.466005719733079e-06, "loss": 3.7125, "step": 9386 }, { "epoch": 1.79, "learning_rate": 2.465948522402288e-06, "loss": 3.2167, "step": 9387 }, { "epoch": 1.79, "learning_rate": 2.4658913250714965e-06, "loss": 2.9941, "step": 9388 }, { "epoch": 1.79, "learning_rate": 2.4658341277407053e-06, "loss": 3.1843, "step": 9389 }, { "epoch": 1.79, "learning_rate": 2.4657769304099145e-06, "loss": 2.9293, "step": 9390 }, { "epoch": 1.79, "learning_rate": 2.465719733079123e-06, "loss": 4.0621, "step": 9391 }, { "epoch": 1.79, "learning_rate": 2.4656625357483317e-06, "loss": 3.2393, "step": 9392 }, { "epoch": 1.79, "learning_rate": 2.4656053384175405e-06, "loss": 2.4296, "step": 9393 }, { "epoch": 1.79, "learning_rate": 2.4655481410867493e-06, "loss": 3.2326, "step": 9394 }, { "epoch": 1.79, "learning_rate": 2.465490943755958e-06, "loss": 3.0228, "step": 9395 }, { "epoch": 1.79, "learning_rate": 2.465433746425167e-06, "loss": 2.0565, "step": 9396 }, { "epoch": 1.79, "learning_rate": 2.4653765490943757e-06, "loss": 4.1125, "step": 9397 }, { "epoch": 1.79, "learning_rate": 2.4653193517635845e-06, "loss": 2.9784, "step": 9398 }, { "epoch": 1.79, "learning_rate": 2.4652621544327933e-06, "loss": 3.4629, "step": 9399 }, { "epoch": 1.79, "learning_rate": 2.465204957102002e-06, "loss": 2.2483, "step": 9400 }, { "epoch": 1.79, "learning_rate": 2.465147759771211e-06, "loss": 3.1356, "step": 9401 }, { "epoch": 1.79, "learning_rate": 2.4650905624404196e-06, "loss": 2.7453, "step": 9402 }, { "epoch": 1.79, "learning_rate": 2.465033365109628e-06, "loss": 2.5309, "step": 9403 }, { "epoch": 1.79, "learning_rate": 2.4649761677788372e-06, "loss": 2.5225, "step": 9404 }, { "epoch": 1.79, "learning_rate": 2.464918970448046e-06, "loss": 3.4978, "step": 9405 }, { "epoch": 1.79, "learning_rate": 2.4648617731172544e-06, "loss": 3.0323, "step": 9406 }, { "epoch": 1.79, "learning_rate": 2.4648045757864636e-06, "loss": 2.6084, "step": 9407 }, { "epoch": 1.79, "learning_rate": 2.464747378455672e-06, "loss": 3.2281, "step": 9408 }, { "epoch": 1.79, "learning_rate": 2.4646901811248808e-06, "loss": 2.912, "step": 9409 }, { "epoch": 1.79, "learning_rate": 2.4646329837940896e-06, "loss": 2.9232, "step": 9410 }, { "epoch": 1.79, "learning_rate": 2.4645757864632984e-06, "loss": 2.9506, "step": 9411 }, { "epoch": 1.79, "learning_rate": 2.464518589132507e-06, "loss": 2.5722, "step": 9412 }, { "epoch": 1.79, "learning_rate": 2.464461391801716e-06, "loss": 3.159, "step": 9413 }, { "epoch": 1.79, "learning_rate": 2.4644041944709247e-06, "loss": 2.6754, "step": 9414 }, { "epoch": 1.79, "learning_rate": 2.4643469971401335e-06, "loss": 2.7081, "step": 9415 }, { "epoch": 1.79, "learning_rate": 2.4642897998093423e-06, "loss": 2.6313, "step": 9416 }, { "epoch": 1.79, "learning_rate": 2.464232602478551e-06, "loss": 2.4185, "step": 9417 }, { "epoch": 1.79, "learning_rate": 2.46417540514776e-06, "loss": 2.8434, "step": 9418 }, { "epoch": 1.79, "learning_rate": 2.4641182078169687e-06, "loss": 2.4626, "step": 9419 }, { "epoch": 1.79, "learning_rate": 2.464061010486177e-06, "loss": 2.9396, "step": 9420 }, { "epoch": 1.79, "learning_rate": 2.4640038131553863e-06, "loss": 2.8476, "step": 9421 }, { "epoch": 1.79, "learning_rate": 2.463946615824595e-06, "loss": 3.0569, "step": 9422 }, { "epoch": 1.79, "learning_rate": 2.4638894184938035e-06, "loss": 2.6441, "step": 9423 }, { "epoch": 1.8, "learning_rate": 2.4638322211630127e-06, "loss": 2.4674, "step": 9424 }, { "epoch": 1.8, "learning_rate": 2.463775023832221e-06, "loss": 3.2473, "step": 9425 }, { "epoch": 1.8, "learning_rate": 2.46371782650143e-06, "loss": 1.7588, "step": 9426 }, { "epoch": 1.8, "learning_rate": 2.4636606291706386e-06, "loss": 2.9834, "step": 9427 }, { "epoch": 1.8, "learning_rate": 2.4636034318398474e-06, "loss": 3.186, "step": 9428 }, { "epoch": 1.8, "learning_rate": 2.4635462345090566e-06, "loss": 3.5346, "step": 9429 }, { "epoch": 1.8, "learning_rate": 2.463489037178265e-06, "loss": 2.8878, "step": 9430 }, { "epoch": 1.8, "learning_rate": 2.463431839847474e-06, "loss": 2.6108, "step": 9431 }, { "epoch": 1.8, "learning_rate": 2.4633746425166826e-06, "loss": 2.4618, "step": 9432 }, { "epoch": 1.8, "learning_rate": 2.4633174451858914e-06, "loss": 3.1157, "step": 9433 }, { "epoch": 1.8, "learning_rate": 2.4632602478551e-06, "loss": 2.4476, "step": 9434 }, { "epoch": 1.8, "learning_rate": 2.463203050524309e-06, "loss": 2.821, "step": 9435 }, { "epoch": 1.8, "learning_rate": 2.4631458531935178e-06, "loss": 3.2131, "step": 9436 }, { "epoch": 1.8, "learning_rate": 2.4630886558627266e-06, "loss": 2.6921, "step": 9437 }, { "epoch": 1.8, "learning_rate": 2.4630314585319354e-06, "loss": 3.0435, "step": 9438 }, { "epoch": 1.8, "learning_rate": 2.462974261201144e-06, "loss": 3.2492, "step": 9439 }, { "epoch": 1.8, "learning_rate": 2.4629170638703525e-06, "loss": 3.5626, "step": 9440 }, { "epoch": 1.8, "learning_rate": 2.4628598665395617e-06, "loss": 2.6265, "step": 9441 }, { "epoch": 1.8, "learning_rate": 2.46280266920877e-06, "loss": 3.0958, "step": 9442 }, { "epoch": 1.8, "learning_rate": 2.462745471877979e-06, "loss": 3.0497, "step": 9443 }, { "epoch": 1.8, "learning_rate": 2.462688274547188e-06, "loss": 3.6074, "step": 9444 }, { "epoch": 1.8, "learning_rate": 2.4626310772163965e-06, "loss": 3.7551, "step": 9445 }, { "epoch": 1.8, "learning_rate": 2.4625738798856057e-06, "loss": 2.3195, "step": 9446 }, { "epoch": 1.8, "learning_rate": 2.462516682554814e-06, "loss": 2.071, "step": 9447 }, { "epoch": 1.8, "learning_rate": 2.462459485224023e-06, "loss": 2.3481, "step": 9448 }, { "epoch": 1.8, "learning_rate": 2.4624022878932317e-06, "loss": 3.2517, "step": 9449 }, { "epoch": 1.8, "learning_rate": 2.4623450905624405e-06, "loss": 1.9771, "step": 9450 }, { "epoch": 1.8, "learning_rate": 2.4622878932316493e-06, "loss": 3.6078, "step": 9451 }, { "epoch": 1.8, "learning_rate": 2.462230695900858e-06, "loss": 2.9915, "step": 9452 }, { "epoch": 1.8, "learning_rate": 2.462173498570067e-06, "loss": 3.2969, "step": 9453 }, { "epoch": 1.8, "learning_rate": 2.4621163012392756e-06, "loss": 3.7298, "step": 9454 }, { "epoch": 1.8, "learning_rate": 2.4620591039084844e-06, "loss": 2.9568, "step": 9455 }, { "epoch": 1.8, "learning_rate": 2.4620019065776932e-06, "loss": 2.7561, "step": 9456 }, { "epoch": 1.8, "learning_rate": 2.4619447092469016e-06, "loss": 3.7711, "step": 9457 }, { "epoch": 1.8, "learning_rate": 2.461887511916111e-06, "loss": 3.7077, "step": 9458 }, { "epoch": 1.8, "learning_rate": 2.461830314585319e-06, "loss": 4.0321, "step": 9459 }, { "epoch": 1.8, "learning_rate": 2.461773117254528e-06, "loss": 2.8891, "step": 9460 }, { "epoch": 1.8, "learning_rate": 2.461715919923737e-06, "loss": 3.3974, "step": 9461 }, { "epoch": 1.8, "learning_rate": 2.4616587225929456e-06, "loss": 2.6847, "step": 9462 }, { "epoch": 1.8, "learning_rate": 2.4616015252621548e-06, "loss": 2.8913, "step": 9463 }, { "epoch": 1.8, "learning_rate": 2.461544327931363e-06, "loss": 3.1466, "step": 9464 }, { "epoch": 1.8, "learning_rate": 2.461487130600572e-06, "loss": 3.0236, "step": 9465 }, { "epoch": 1.8, "learning_rate": 2.4614299332697807e-06, "loss": 2.5332, "step": 9466 }, { "epoch": 1.8, "learning_rate": 2.4613727359389895e-06, "loss": 2.7827, "step": 9467 }, { "epoch": 1.8, "learning_rate": 2.4613155386081983e-06, "loss": 2.8821, "step": 9468 }, { "epoch": 1.8, "learning_rate": 2.461258341277407e-06, "loss": 3.0947, "step": 9469 }, { "epoch": 1.8, "learning_rate": 2.461201143946616e-06, "loss": 3.0414, "step": 9470 }, { "epoch": 1.8, "learning_rate": 2.4611439466158247e-06, "loss": 3.6235, "step": 9471 }, { "epoch": 1.8, "learning_rate": 2.4610867492850335e-06, "loss": 2.7536, "step": 9472 }, { "epoch": 1.8, "learning_rate": 2.4610295519542423e-06, "loss": 3.5113, "step": 9473 }, { "epoch": 1.8, "learning_rate": 2.4609723546234507e-06, "loss": 2.9914, "step": 9474 }, { "epoch": 1.8, "learning_rate": 2.46091515729266e-06, "loss": 3.3843, "step": 9475 }, { "epoch": 1.8, "learning_rate": 2.4608579599618683e-06, "loss": 3.033, "step": 9476 }, { "epoch": 1.81, "learning_rate": 2.460800762631077e-06, "loss": 2.2643, "step": 9477 }, { "epoch": 1.81, "learning_rate": 2.4607435653002863e-06, "loss": 2.6025, "step": 9478 }, { "epoch": 1.81, "learning_rate": 2.4606863679694946e-06, "loss": 3.2378, "step": 9479 }, { "epoch": 1.81, "learning_rate": 2.460629170638704e-06, "loss": 2.8907, "step": 9480 }, { "epoch": 1.81, "learning_rate": 2.4605719733079122e-06, "loss": 3.1261, "step": 9481 }, { "epoch": 1.81, "learning_rate": 2.460514775977121e-06, "loss": 3.6448, "step": 9482 }, { "epoch": 1.81, "learning_rate": 2.4604575786463302e-06, "loss": 2.8001, "step": 9483 }, { "epoch": 1.81, "learning_rate": 2.4604003813155386e-06, "loss": 3.2434, "step": 9484 }, { "epoch": 1.81, "learning_rate": 2.4603431839847474e-06, "loss": 3.8741, "step": 9485 }, { "epoch": 1.81, "learning_rate": 2.460285986653956e-06, "loss": 2.8937, "step": 9486 }, { "epoch": 1.81, "learning_rate": 2.460228789323165e-06, "loss": 3.2481, "step": 9487 }, { "epoch": 1.81, "learning_rate": 2.4601715919923738e-06, "loss": 2.5237, "step": 9488 }, { "epoch": 1.81, "learning_rate": 2.4601143946615826e-06, "loss": 3.189, "step": 9489 }, { "epoch": 1.81, "learning_rate": 2.4600571973307914e-06, "loss": 2.4056, "step": 9490 }, { "epoch": 1.81, "learning_rate": 2.4599999999999997e-06, "loss": 1.8649, "step": 9491 }, { "epoch": 1.81, "learning_rate": 2.459942802669209e-06, "loss": 2.8856, "step": 9492 }, { "epoch": 1.81, "learning_rate": 2.4598856053384178e-06, "loss": 3.1672, "step": 9493 }, { "epoch": 1.81, "learning_rate": 2.4598284080076265e-06, "loss": 3.3856, "step": 9494 }, { "epoch": 1.81, "learning_rate": 2.4597712106768353e-06, "loss": 3.4536, "step": 9495 }, { "epoch": 1.81, "learning_rate": 2.4597140133460437e-06, "loss": 3.636, "step": 9496 }, { "epoch": 1.81, "learning_rate": 2.459656816015253e-06, "loss": 3.0239, "step": 9497 }, { "epoch": 1.81, "learning_rate": 2.4595996186844613e-06, "loss": 2.5091, "step": 9498 }, { "epoch": 1.81, "learning_rate": 2.45954242135367e-06, "loss": 3.2136, "step": 9499 }, { "epoch": 1.81, "learning_rate": 2.4594852240228793e-06, "loss": 2.8853, "step": 9500 }, { "epoch": 1.81, "learning_rate": 2.4594280266920877e-06, "loss": 2.7712, "step": 9501 }, { "epoch": 1.81, "learning_rate": 2.4593708293612965e-06, "loss": 2.4113, "step": 9502 }, { "epoch": 1.81, "learning_rate": 2.4593136320305053e-06, "loss": 2.7622, "step": 9503 }, { "epoch": 1.81, "learning_rate": 2.459256434699714e-06, "loss": 3.2123, "step": 9504 }, { "epoch": 1.81, "learning_rate": 2.459199237368923e-06, "loss": 3.2818, "step": 9505 }, { "epoch": 1.81, "learning_rate": 2.4591420400381316e-06, "loss": 2.8486, "step": 9506 }, { "epoch": 1.81, "learning_rate": 2.4590848427073404e-06, "loss": 2.2149, "step": 9507 }, { "epoch": 1.81, "learning_rate": 2.459027645376549e-06, "loss": 3.0834, "step": 9508 }, { "epoch": 1.81, "learning_rate": 2.458970448045758e-06, "loss": 2.7239, "step": 9509 }, { "epoch": 1.81, "learning_rate": 2.458913250714967e-06, "loss": 2.4405, "step": 9510 }, { "epoch": 1.81, "learning_rate": 2.4588560533841756e-06, "loss": 2.5901, "step": 9511 }, { "epoch": 1.81, "learning_rate": 2.4587988560533844e-06, "loss": 2.7589, "step": 9512 }, { "epoch": 1.81, "learning_rate": 2.4587416587225928e-06, "loss": 3.156, "step": 9513 }, { "epoch": 1.81, "learning_rate": 2.458684461391802e-06, "loss": 3.3862, "step": 9514 }, { "epoch": 1.81, "learning_rate": 2.4586272640610104e-06, "loss": 2.6064, "step": 9515 }, { "epoch": 1.81, "learning_rate": 2.458570066730219e-06, "loss": 2.7898, "step": 9516 }, { "epoch": 1.81, "learning_rate": 2.4585128693994284e-06, "loss": 3.3525, "step": 9517 }, { "epoch": 1.81, "learning_rate": 2.4584556720686368e-06, "loss": 3.6855, "step": 9518 }, { "epoch": 1.81, "learning_rate": 2.4583984747378455e-06, "loss": 2.7125, "step": 9519 }, { "epoch": 1.81, "learning_rate": 2.4583412774070543e-06, "loss": 2.0241, "step": 9520 }, { "epoch": 1.81, "learning_rate": 2.458284080076263e-06, "loss": 2.7385, "step": 9521 }, { "epoch": 1.81, "learning_rate": 2.458226882745472e-06, "loss": 2.806, "step": 9522 }, { "epoch": 1.81, "learning_rate": 2.4581696854146807e-06, "loss": 2.7467, "step": 9523 }, { "epoch": 1.81, "learning_rate": 2.4581124880838895e-06, "loss": 3.0557, "step": 9524 }, { "epoch": 1.81, "learning_rate": 2.458055290753098e-06, "loss": 2.3212, "step": 9525 }, { "epoch": 1.81, "learning_rate": 2.457998093422307e-06, "loss": 2.207, "step": 9526 }, { "epoch": 1.81, "learning_rate": 2.457940896091516e-06, "loss": 2.959, "step": 9527 }, { "epoch": 1.81, "learning_rate": 2.4578836987607247e-06, "loss": 2.6609, "step": 9528 }, { "epoch": 1.82, "learning_rate": 2.4578265014299335e-06, "loss": 2.7556, "step": 9529 }, { "epoch": 1.82, "learning_rate": 2.457769304099142e-06, "loss": 3.253, "step": 9530 }, { "epoch": 1.82, "learning_rate": 2.457712106768351e-06, "loss": 3.8656, "step": 9531 }, { "epoch": 1.82, "learning_rate": 2.45765490943756e-06, "loss": 2.6369, "step": 9532 }, { "epoch": 1.82, "learning_rate": 2.4575977121067682e-06, "loss": 3.4536, "step": 9533 }, { "epoch": 1.82, "learning_rate": 2.4575405147759775e-06, "loss": 2.1048, "step": 9534 }, { "epoch": 1.82, "learning_rate": 2.457483317445186e-06, "loss": 3.1818, "step": 9535 }, { "epoch": 1.82, "learning_rate": 2.4574261201143946e-06, "loss": 3.6272, "step": 9536 }, { "epoch": 1.82, "learning_rate": 2.4573689227836034e-06, "loss": 2.7902, "step": 9537 }, { "epoch": 1.82, "learning_rate": 2.457311725452812e-06, "loss": 3.0922, "step": 9538 }, { "epoch": 1.82, "learning_rate": 2.457254528122021e-06, "loss": 3.1364, "step": 9539 }, { "epoch": 1.82, "learning_rate": 2.45719733079123e-06, "loss": 3.1189, "step": 9540 }, { "epoch": 1.82, "learning_rate": 2.4571401334604386e-06, "loss": 3.2252, "step": 9541 }, { "epoch": 1.82, "learning_rate": 2.4570829361296474e-06, "loss": 2.9621, "step": 9542 }, { "epoch": 1.82, "learning_rate": 2.457025738798856e-06, "loss": 3.0974, "step": 9543 }, { "epoch": 1.82, "learning_rate": 2.456968541468065e-06, "loss": 3.4513, "step": 9544 }, { "epoch": 1.82, "learning_rate": 2.4569113441372738e-06, "loss": 2.768, "step": 9545 }, { "epoch": 1.82, "learning_rate": 2.4568541468064826e-06, "loss": 2.424, "step": 9546 }, { "epoch": 1.82, "learning_rate": 2.456796949475691e-06, "loss": 2.625, "step": 9547 }, { "epoch": 1.82, "learning_rate": 2.4567397521449e-06, "loss": 3.0354, "step": 9548 }, { "epoch": 1.82, "learning_rate": 2.456682554814109e-06, "loss": 3.2128, "step": 9549 }, { "epoch": 1.82, "learning_rate": 2.4566253574833173e-06, "loss": 2.3801, "step": 9550 }, { "epoch": 1.82, "learning_rate": 2.4565681601525265e-06, "loss": 2.7088, "step": 9551 }, { "epoch": 1.82, "learning_rate": 2.456510962821735e-06, "loss": 2.4732, "step": 9552 }, { "epoch": 1.82, "learning_rate": 2.4564537654909437e-06, "loss": 2.6266, "step": 9553 }, { "epoch": 1.82, "learning_rate": 2.4563965681601525e-06, "loss": 2.988, "step": 9554 }, { "epoch": 1.82, "learning_rate": 2.4563393708293613e-06, "loss": 3.1227, "step": 9555 }, { "epoch": 1.82, "learning_rate": 2.45628217349857e-06, "loss": 2.6441, "step": 9556 }, { "epoch": 1.82, "learning_rate": 2.456224976167779e-06, "loss": 3.3618, "step": 9557 }, { "epoch": 1.82, "learning_rate": 2.4561677788369877e-06, "loss": 4.1199, "step": 9558 }, { "epoch": 1.82, "learning_rate": 2.4561105815061964e-06, "loss": 3.047, "step": 9559 }, { "epoch": 1.82, "learning_rate": 2.4560533841754052e-06, "loss": 3.2481, "step": 9560 }, { "epoch": 1.82, "learning_rate": 2.455996186844614e-06, "loss": 3.1125, "step": 9561 }, { "epoch": 1.82, "learning_rate": 2.455938989513823e-06, "loss": 3.323, "step": 9562 }, { "epoch": 1.82, "learning_rate": 2.4558817921830316e-06, "loss": 3.2372, "step": 9563 }, { "epoch": 1.82, "learning_rate": 2.45582459485224e-06, "loss": 3.3104, "step": 9564 }, { "epoch": 1.82, "learning_rate": 2.4557673975214492e-06, "loss": 3.0199, "step": 9565 }, { "epoch": 1.82, "learning_rate": 2.455710200190658e-06, "loss": 3.3084, "step": 9566 }, { "epoch": 1.82, "learning_rate": 2.4556530028598664e-06, "loss": 2.7872, "step": 9567 }, { "epoch": 1.82, "learning_rate": 2.4555958055290756e-06, "loss": 3.2131, "step": 9568 }, { "epoch": 1.82, "learning_rate": 2.455538608198284e-06, "loss": 3.633, "step": 9569 }, { "epoch": 1.82, "learning_rate": 2.4554814108674928e-06, "loss": 2.1712, "step": 9570 }, { "epoch": 1.82, "learning_rate": 2.455424213536702e-06, "loss": 2.6648, "step": 9571 }, { "epoch": 1.82, "learning_rate": 2.4553670162059103e-06, "loss": 3.2525, "step": 9572 }, { "epoch": 1.82, "learning_rate": 2.455309818875119e-06, "loss": 3.2587, "step": 9573 }, { "epoch": 1.82, "learning_rate": 2.455252621544328e-06, "loss": 3.2165, "step": 9574 }, { "epoch": 1.82, "learning_rate": 2.4551954242135367e-06, "loss": 2.3677, "step": 9575 }, { "epoch": 1.82, "learning_rate": 2.4551382268827455e-06, "loss": 2.6919, "step": 9576 }, { "epoch": 1.82, "learning_rate": 2.4550810295519543e-06, "loss": 2.9619, "step": 9577 }, { "epoch": 1.82, "learning_rate": 2.455023832221163e-06, "loss": 3.2499, "step": 9578 }, { "epoch": 1.82, "learning_rate": 2.454966634890372e-06, "loss": 3.4897, "step": 9579 }, { "epoch": 1.82, "learning_rate": 2.4549094375595807e-06, "loss": 3.1457, "step": 9580 }, { "epoch": 1.82, "learning_rate": 2.4548522402287895e-06, "loss": 2.3451, "step": 9581 }, { "epoch": 1.83, "learning_rate": 2.4547950428979983e-06, "loss": 2.6248, "step": 9582 }, { "epoch": 1.83, "learning_rate": 2.454737845567207e-06, "loss": 2.2308, "step": 9583 }, { "epoch": 1.83, "learning_rate": 2.4546806482364154e-06, "loss": 2.5854, "step": 9584 }, { "epoch": 1.83, "learning_rate": 2.4546234509056247e-06, "loss": 3.3719, "step": 9585 }, { "epoch": 1.83, "learning_rate": 2.454566253574833e-06, "loss": 4.0341, "step": 9586 }, { "epoch": 1.83, "learning_rate": 2.454509056244042e-06, "loss": 3.7253, "step": 9587 }, { "epoch": 1.83, "learning_rate": 2.454451858913251e-06, "loss": 3.1917, "step": 9588 }, { "epoch": 1.83, "learning_rate": 2.4543946615824594e-06, "loss": 2.9356, "step": 9589 }, { "epoch": 1.83, "learning_rate": 2.454337464251668e-06, "loss": 2.6142, "step": 9590 }, { "epoch": 1.83, "learning_rate": 2.454280266920877e-06, "loss": 3.1443, "step": 9591 }, { "epoch": 1.83, "learning_rate": 2.454223069590086e-06, "loss": 3.3884, "step": 9592 }, { "epoch": 1.83, "learning_rate": 2.4541658722592946e-06, "loss": 3.0805, "step": 9593 }, { "epoch": 1.83, "learning_rate": 2.4541086749285034e-06, "loss": 2.4329, "step": 9594 }, { "epoch": 1.83, "learning_rate": 2.454051477597712e-06, "loss": 3.6926, "step": 9595 }, { "epoch": 1.83, "learning_rate": 2.453994280266921e-06, "loss": 3.165, "step": 9596 }, { "epoch": 1.83, "learning_rate": 2.4539370829361298e-06, "loss": 2.8784, "step": 9597 }, { "epoch": 1.83, "learning_rate": 2.4538798856053386e-06, "loss": 3.4514, "step": 9598 }, { "epoch": 1.83, "learning_rate": 2.4538226882745474e-06, "loss": 2.7036, "step": 9599 }, { "epoch": 1.83, "learning_rate": 2.453765490943756e-06, "loss": 2.8253, "step": 9600 }, { "epoch": 1.83, "learning_rate": 2.4537082936129645e-06, "loss": 2.7911, "step": 9601 }, { "epoch": 1.83, "learning_rate": 2.4536510962821737e-06, "loss": 3.4664, "step": 9602 }, { "epoch": 1.83, "learning_rate": 2.453593898951382e-06, "loss": 2.8751, "step": 9603 }, { "epoch": 1.83, "learning_rate": 2.453536701620591e-06, "loss": 3.0342, "step": 9604 }, { "epoch": 1.83, "learning_rate": 2.4534795042898e-06, "loss": 3.3113, "step": 9605 }, { "epoch": 1.83, "learning_rate": 2.4534223069590085e-06, "loss": 3.4238, "step": 9606 }, { "epoch": 1.83, "learning_rate": 2.4533651096282173e-06, "loss": 2.7555, "step": 9607 }, { "epoch": 1.83, "learning_rate": 2.453307912297426e-06, "loss": 2.5338, "step": 9608 }, { "epoch": 1.83, "learning_rate": 2.453250714966635e-06, "loss": 2.9162, "step": 9609 }, { "epoch": 1.83, "learning_rate": 2.4531935176358437e-06, "loss": 2.5165, "step": 9610 }, { "epoch": 1.83, "learning_rate": 2.4531363203050525e-06, "loss": 2.6471, "step": 9611 }, { "epoch": 1.83, "learning_rate": 2.4530791229742613e-06, "loss": 2.6457, "step": 9612 }, { "epoch": 1.83, "learning_rate": 2.45302192564347e-06, "loss": 2.7685, "step": 9613 }, { "epoch": 1.83, "learning_rate": 2.452964728312679e-06, "loss": 3.2374, "step": 9614 }, { "epoch": 1.83, "learning_rate": 2.4529075309818876e-06, "loss": 3.0598, "step": 9615 }, { "epoch": 1.83, "learning_rate": 2.4528503336510964e-06, "loss": 2.0801, "step": 9616 }, { "epoch": 1.83, "learning_rate": 2.4527931363203052e-06, "loss": 2.7013, "step": 9617 }, { "epoch": 1.83, "learning_rate": 2.4527359389895136e-06, "loss": 2.5066, "step": 9618 }, { "epoch": 1.83, "learning_rate": 2.452678741658723e-06, "loss": 2.7126, "step": 9619 }, { "epoch": 1.83, "learning_rate": 2.4526215443279316e-06, "loss": 2.4341, "step": 9620 }, { "epoch": 1.83, "learning_rate": 2.45256434699714e-06, "loss": 1.9949, "step": 9621 }, { "epoch": 1.83, "learning_rate": 2.452507149666349e-06, "loss": 3.0719, "step": 9622 }, { "epoch": 1.83, "learning_rate": 2.4524499523355576e-06, "loss": 2.4396, "step": 9623 }, { "epoch": 1.83, "learning_rate": 2.4523927550047664e-06, "loss": 2.2635, "step": 9624 }, { "epoch": 1.83, "learning_rate": 2.452335557673975e-06, "loss": 2.8795, "step": 9625 }, { "epoch": 1.83, "learning_rate": 2.452278360343184e-06, "loss": 3.3301, "step": 9626 }, { "epoch": 1.83, "learning_rate": 2.452221163012393e-06, "loss": 2.361, "step": 9627 }, { "epoch": 1.83, "learning_rate": 2.4521639656816015e-06, "loss": 2.516, "step": 9628 }, { "epoch": 1.83, "learning_rate": 2.4521067683508103e-06, "loss": 3.8173, "step": 9629 }, { "epoch": 1.83, "learning_rate": 2.452049571020019e-06, "loss": 1.8899, "step": 9630 }, { "epoch": 1.83, "learning_rate": 2.451992373689228e-06, "loss": 2.6238, "step": 9631 }, { "epoch": 1.83, "learning_rate": 2.4519351763584367e-06, "loss": 2.8792, "step": 9632 }, { "epoch": 1.83, "learning_rate": 2.4518779790276455e-06, "loss": 2.957, "step": 9633 }, { "epoch": 1.84, "learning_rate": 2.4518207816968543e-06, "loss": 2.8828, "step": 9634 }, { "epoch": 1.84, "learning_rate": 2.4517635843660627e-06, "loss": 2.959, "step": 9635 }, { "epoch": 1.84, "learning_rate": 2.451706387035272e-06, "loss": 2.7258, "step": 9636 }, { "epoch": 1.84, "learning_rate": 2.4516491897044807e-06, "loss": 2.6937, "step": 9637 }, { "epoch": 1.84, "learning_rate": 2.451591992373689e-06, "loss": 3.5995, "step": 9638 }, { "epoch": 1.84, "learning_rate": 2.4515347950428983e-06, "loss": 3.2995, "step": 9639 }, { "epoch": 1.84, "learning_rate": 2.4514775977121066e-06, "loss": 3.1444, "step": 9640 }, { "epoch": 1.84, "learning_rate": 2.451420400381316e-06, "loss": 3.1813, "step": 9641 }, { "epoch": 1.84, "learning_rate": 2.4513632030505242e-06, "loss": 2.9973, "step": 9642 }, { "epoch": 1.84, "learning_rate": 2.451306005719733e-06, "loss": 3.1758, "step": 9643 }, { "epoch": 1.84, "learning_rate": 2.4512488083889422e-06, "loss": 3.7406, "step": 9644 }, { "epoch": 1.84, "learning_rate": 2.4511916110581506e-06, "loss": 3.3055, "step": 9645 }, { "epoch": 1.84, "learning_rate": 2.4511344137273594e-06, "loss": 2.716, "step": 9646 }, { "epoch": 1.84, "learning_rate": 2.451077216396568e-06, "loss": 3.23, "step": 9647 }, { "epoch": 1.84, "learning_rate": 2.451020019065777e-06, "loss": 3.2513, "step": 9648 }, { "epoch": 1.84, "learning_rate": 2.4509628217349858e-06, "loss": 3.2515, "step": 9649 }, { "epoch": 1.84, "learning_rate": 2.4509056244041946e-06, "loss": 3.6455, "step": 9650 }, { "epoch": 1.84, "learning_rate": 2.4508484270734034e-06, "loss": 3.107, "step": 9651 }, { "epoch": 1.84, "learning_rate": 2.4507912297426117e-06, "loss": 2.1053, "step": 9652 }, { "epoch": 1.84, "learning_rate": 2.450734032411821e-06, "loss": 3.6911, "step": 9653 }, { "epoch": 1.84, "learning_rate": 2.4506768350810297e-06, "loss": 2.9504, "step": 9654 }, { "epoch": 1.84, "learning_rate": 2.450619637750238e-06, "loss": 2.4782, "step": 9655 }, { "epoch": 1.84, "learning_rate": 2.4505624404194473e-06, "loss": 2.9052, "step": 9656 }, { "epoch": 1.84, "learning_rate": 2.4505052430886557e-06, "loss": 2.68, "step": 9657 }, { "epoch": 1.84, "learning_rate": 2.450448045757865e-06, "loss": 2.3119, "step": 9658 }, { "epoch": 1.84, "learning_rate": 2.4503908484270737e-06, "loss": 3.411, "step": 9659 }, { "epoch": 1.84, "learning_rate": 2.450333651096282e-06, "loss": 2.8922, "step": 9660 }, { "epoch": 1.84, "learning_rate": 2.4502764537654913e-06, "loss": 2.1157, "step": 9661 }, { "epoch": 1.84, "learning_rate": 2.4502192564346997e-06, "loss": 2.957, "step": 9662 }, { "epoch": 1.84, "learning_rate": 2.4501620591039085e-06, "loss": 3.1023, "step": 9663 }, { "epoch": 1.84, "learning_rate": 2.4501048617731173e-06, "loss": 3.3031, "step": 9664 }, { "epoch": 1.84, "learning_rate": 2.450047664442326e-06, "loss": 2.4963, "step": 9665 }, { "epoch": 1.84, "learning_rate": 2.449990467111535e-06, "loss": 2.1678, "step": 9666 }, { "epoch": 1.84, "learning_rate": 2.4499332697807436e-06, "loss": 2.805, "step": 9667 }, { "epoch": 1.84, "learning_rate": 2.4498760724499524e-06, "loss": 3.3035, "step": 9668 }, { "epoch": 1.84, "learning_rate": 2.4498188751191612e-06, "loss": 3.5418, "step": 9669 }, { "epoch": 1.84, "learning_rate": 2.44976167778837e-06, "loss": 2.1473, "step": 9670 }, { "epoch": 1.84, "learning_rate": 2.449704480457579e-06, "loss": 3.5732, "step": 9671 }, { "epoch": 1.84, "learning_rate": 2.449647283126787e-06, "loss": 2.2418, "step": 9672 }, { "epoch": 1.84, "learning_rate": 2.4495900857959964e-06, "loss": 2.7745, "step": 9673 }, { "epoch": 1.84, "learning_rate": 2.4495328884652048e-06, "loss": 2.8346, "step": 9674 }, { "epoch": 1.84, "learning_rate": 2.449475691134414e-06, "loss": 3.3718, "step": 9675 }, { "epoch": 1.84, "learning_rate": 2.4494184938036228e-06, "loss": 2.8155, "step": 9676 }, { "epoch": 1.84, "learning_rate": 2.449361296472831e-06, "loss": 2.9332, "step": 9677 }, { "epoch": 1.84, "learning_rate": 2.4493040991420404e-06, "loss": 2.0805, "step": 9678 }, { "epoch": 1.84, "learning_rate": 2.4492469018112487e-06, "loss": 2.0431, "step": 9679 }, { "epoch": 1.84, "learning_rate": 2.4491897044804575e-06, "loss": 2.4135, "step": 9680 }, { "epoch": 1.84, "learning_rate": 2.4491325071496663e-06, "loss": 3.3201, "step": 9681 }, { "epoch": 1.84, "learning_rate": 2.449075309818875e-06, "loss": 3.2192, "step": 9682 }, { "epoch": 1.84, "learning_rate": 2.449018112488084e-06, "loss": 2.8699, "step": 9683 }, { "epoch": 1.84, "learning_rate": 2.4489609151572927e-06, "loss": 3.212, "step": 9684 }, { "epoch": 1.84, "learning_rate": 2.4489037178265015e-06, "loss": 3.1497, "step": 9685 }, { "epoch": 1.84, "learning_rate": 2.4488465204957103e-06, "loss": 2.3504, "step": 9686 }, { "epoch": 1.85, "learning_rate": 2.448789323164919e-06, "loss": 1.7925, "step": 9687 }, { "epoch": 1.85, "learning_rate": 2.448732125834128e-06, "loss": 3.0529, "step": 9688 }, { "epoch": 1.85, "learning_rate": 2.4486749285033363e-06, "loss": 2.8609, "step": 9689 }, { "epoch": 1.85, "learning_rate": 2.4486177311725455e-06, "loss": 2.5977, "step": 9690 }, { "epoch": 1.85, "learning_rate": 2.448560533841754e-06, "loss": 2.5778, "step": 9691 }, { "epoch": 1.85, "learning_rate": 2.448503336510963e-06, "loss": 2.704, "step": 9692 }, { "epoch": 1.85, "learning_rate": 2.448446139180172e-06, "loss": 3.395, "step": 9693 }, { "epoch": 1.85, "learning_rate": 2.4483889418493802e-06, "loss": 2.4743, "step": 9694 }, { "epoch": 1.85, "learning_rate": 2.4483317445185894e-06, "loss": 2.7523, "step": 9695 }, { "epoch": 1.85, "learning_rate": 2.448274547187798e-06, "loss": 2.9508, "step": 9696 }, { "epoch": 1.85, "learning_rate": 2.4482173498570066e-06, "loss": 2.8045, "step": 9697 }, { "epoch": 1.85, "learning_rate": 2.4481601525262154e-06, "loss": 2.896, "step": 9698 }, { "epoch": 1.85, "learning_rate": 2.448102955195424e-06, "loss": 3.7174, "step": 9699 }, { "epoch": 1.85, "learning_rate": 2.448045757864633e-06, "loss": 3.1605, "step": 9700 }, { "epoch": 1.85, "learning_rate": 2.4479885605338418e-06, "loss": 3.192, "step": 9701 }, { "epoch": 1.85, "learning_rate": 2.4479313632030506e-06, "loss": 2.8529, "step": 9702 }, { "epoch": 1.85, "learning_rate": 2.4478741658722594e-06, "loss": 3.0794, "step": 9703 }, { "epoch": 1.85, "learning_rate": 2.447816968541468e-06, "loss": 3.2976, "step": 9704 }, { "epoch": 1.85, "learning_rate": 2.447759771210677e-06, "loss": 2.3204, "step": 9705 }, { "epoch": 1.85, "learning_rate": 2.4477025738798858e-06, "loss": 3.5588, "step": 9706 }, { "epoch": 1.85, "learning_rate": 2.4476453765490945e-06, "loss": 3.2179, "step": 9707 }, { "epoch": 1.85, "learning_rate": 2.4475881792183033e-06, "loss": 3.6352, "step": 9708 }, { "epoch": 1.85, "learning_rate": 2.447530981887512e-06, "loss": 3.8136, "step": 9709 }, { "epoch": 1.85, "learning_rate": 2.447473784556721e-06, "loss": 2.8047, "step": 9710 }, { "epoch": 1.85, "learning_rate": 2.4474165872259293e-06, "loss": 2.6018, "step": 9711 }, { "epoch": 1.85, "learning_rate": 2.4473593898951385e-06, "loss": 2.702, "step": 9712 }, { "epoch": 1.85, "learning_rate": 2.447302192564347e-06, "loss": 3.0732, "step": 9713 }, { "epoch": 1.85, "learning_rate": 2.4472449952335557e-06, "loss": 2.4526, "step": 9714 }, { "epoch": 1.85, "learning_rate": 2.447187797902765e-06, "loss": 2.8043, "step": 9715 }, { "epoch": 1.85, "learning_rate": 2.4471306005719733e-06, "loss": 3.4195, "step": 9716 }, { "epoch": 1.85, "learning_rate": 2.447073403241182e-06, "loss": 3.0805, "step": 9717 }, { "epoch": 1.85, "learning_rate": 2.447016205910391e-06, "loss": 3.0405, "step": 9718 }, { "epoch": 1.85, "learning_rate": 2.4469590085795996e-06, "loss": 3.0893, "step": 9719 }, { "epoch": 1.85, "learning_rate": 2.4469018112488084e-06, "loss": 2.7831, "step": 9720 }, { "epoch": 1.85, "learning_rate": 2.4468446139180172e-06, "loss": 2.8408, "step": 9721 }, { "epoch": 1.85, "learning_rate": 2.446787416587226e-06, "loss": 3.1724, "step": 9722 }, { "epoch": 1.85, "learning_rate": 2.446730219256435e-06, "loss": 2.6747, "step": 9723 }, { "epoch": 1.85, "learning_rate": 2.4466730219256436e-06, "loss": 3.0266, "step": 9724 }, { "epoch": 1.85, "learning_rate": 2.4466158245948524e-06, "loss": 2.1606, "step": 9725 }, { "epoch": 1.85, "learning_rate": 2.446558627264061e-06, "loss": 2.1218, "step": 9726 }, { "epoch": 1.85, "learning_rate": 2.44650142993327e-06, "loss": 2.6972, "step": 9727 }, { "epoch": 1.85, "learning_rate": 2.4464442326024784e-06, "loss": 3.4866, "step": 9728 }, { "epoch": 1.85, "learning_rate": 2.4463870352716876e-06, "loss": 2.4751, "step": 9729 }, { "epoch": 1.85, "learning_rate": 2.446329837940896e-06, "loss": 3.514, "step": 9730 }, { "epoch": 1.85, "learning_rate": 2.4462726406101048e-06, "loss": 2.6648, "step": 9731 }, { "epoch": 1.85, "learning_rate": 2.446215443279314e-06, "loss": 2.6053, "step": 9732 }, { "epoch": 1.85, "learning_rate": 2.4461582459485223e-06, "loss": 3.7057, "step": 9733 }, { "epoch": 1.85, "learning_rate": 2.446101048617731e-06, "loss": 3.7875, "step": 9734 }, { "epoch": 1.85, "learning_rate": 2.44604385128694e-06, "loss": 3.3781, "step": 9735 }, { "epoch": 1.85, "learning_rate": 2.4459866539561487e-06, "loss": 2.7223, "step": 9736 }, { "epoch": 1.85, "learning_rate": 2.4459294566253575e-06, "loss": 2.9782, "step": 9737 }, { "epoch": 1.85, "learning_rate": 2.4458722592945663e-06, "loss": 3.2055, "step": 9738 }, { "epoch": 1.86, "learning_rate": 2.445815061963775e-06, "loss": 2.5285, "step": 9739 }, { "epoch": 1.86, "learning_rate": 2.445757864632984e-06, "loss": 2.8234, "step": 9740 }, { "epoch": 1.86, "learning_rate": 2.4457006673021927e-06, "loss": 3.0177, "step": 9741 }, { "epoch": 1.86, "learning_rate": 2.4456434699714015e-06, "loss": 3.4936, "step": 9742 }, { "epoch": 1.86, "learning_rate": 2.4455862726406103e-06, "loss": 2.4897, "step": 9743 }, { "epoch": 1.86, "learning_rate": 2.445529075309819e-06, "loss": 3.425, "step": 9744 }, { "epoch": 1.86, "learning_rate": 2.4454718779790274e-06, "loss": 2.6314, "step": 9745 }, { "epoch": 1.86, "learning_rate": 2.4454146806482367e-06, "loss": 3.5126, "step": 9746 }, { "epoch": 1.86, "learning_rate": 2.4453574833174455e-06, "loss": 3.239, "step": 9747 }, { "epoch": 1.86, "learning_rate": 2.445300285986654e-06, "loss": 2.832, "step": 9748 }, { "epoch": 1.86, "learning_rate": 2.445243088655863e-06, "loss": 3.5001, "step": 9749 }, { "epoch": 1.86, "learning_rate": 2.4451858913250714e-06, "loss": 3.1647, "step": 9750 }, { "epoch": 1.86, "learning_rate": 2.44512869399428e-06, "loss": 2.1917, "step": 9751 }, { "epoch": 1.86, "learning_rate": 2.445071496663489e-06, "loss": 3.194, "step": 9752 }, { "epoch": 1.86, "learning_rate": 2.445014299332698e-06, "loss": 3.3081, "step": 9753 }, { "epoch": 1.86, "learning_rate": 2.4449571020019066e-06, "loss": 2.8837, "step": 9754 }, { "epoch": 1.86, "learning_rate": 2.4448999046711154e-06, "loss": 2.2898, "step": 9755 }, { "epoch": 1.86, "learning_rate": 2.444842707340324e-06, "loss": 3.3115, "step": 9756 }, { "epoch": 1.86, "learning_rate": 2.444785510009533e-06, "loss": 2.8664, "step": 9757 }, { "epoch": 1.86, "learning_rate": 2.4447283126787418e-06, "loss": 2.9193, "step": 9758 }, { "epoch": 1.86, "learning_rate": 2.4446711153479506e-06, "loss": 2.8017, "step": 9759 }, { "epoch": 1.86, "learning_rate": 2.4446139180171593e-06, "loss": 3.0174, "step": 9760 }, { "epoch": 1.86, "learning_rate": 2.444556720686368e-06, "loss": 3.2989, "step": 9761 }, { "epoch": 1.86, "learning_rate": 2.4444995233555765e-06, "loss": 3.442, "step": 9762 }, { "epoch": 1.86, "learning_rate": 2.4444423260247857e-06, "loss": 2.1584, "step": 9763 }, { "epoch": 1.86, "learning_rate": 2.4443851286939945e-06, "loss": 3.6891, "step": 9764 }, { "epoch": 1.86, "learning_rate": 2.444327931363203e-06, "loss": 2.6847, "step": 9765 }, { "epoch": 1.86, "learning_rate": 2.444270734032412e-06, "loss": 2.7901, "step": 9766 }, { "epoch": 1.86, "learning_rate": 2.4442135367016205e-06, "loss": 3.199, "step": 9767 }, { "epoch": 1.86, "learning_rate": 2.4441563393708293e-06, "loss": 2.1019, "step": 9768 }, { "epoch": 1.86, "learning_rate": 2.444099142040038e-06, "loss": 2.9776, "step": 9769 }, { "epoch": 1.86, "learning_rate": 2.444041944709247e-06, "loss": 2.5153, "step": 9770 }, { "epoch": 1.86, "learning_rate": 2.443984747378456e-06, "loss": 2.7744, "step": 9771 }, { "epoch": 1.86, "learning_rate": 2.4439275500476645e-06, "loss": 2.4309, "step": 9772 }, { "epoch": 1.86, "learning_rate": 2.4438703527168732e-06, "loss": 2.5449, "step": 9773 }, { "epoch": 1.86, "learning_rate": 2.443813155386082e-06, "loss": 2.5995, "step": 9774 }, { "epoch": 1.86, "learning_rate": 2.443755958055291e-06, "loss": 2.8563, "step": 9775 }, { "epoch": 1.86, "learning_rate": 2.4436987607244996e-06, "loss": 2.7524, "step": 9776 }, { "epoch": 1.86, "learning_rate": 2.4436415633937084e-06, "loss": 2.9013, "step": 9777 }, { "epoch": 1.86, "learning_rate": 2.4435843660629172e-06, "loss": 2.7182, "step": 9778 }, { "epoch": 1.86, "learning_rate": 2.4435271687321256e-06, "loss": 2.5367, "step": 9779 }, { "epoch": 1.86, "learning_rate": 2.443469971401335e-06, "loss": 2.6095, "step": 9780 }, { "epoch": 1.86, "learning_rate": 2.4434127740705436e-06, "loss": 3.1483, "step": 9781 }, { "epoch": 1.86, "learning_rate": 2.443355576739752e-06, "loss": 3.282, "step": 9782 }, { "epoch": 1.86, "learning_rate": 2.443298379408961e-06, "loss": 2.8145, "step": 9783 }, { "epoch": 1.86, "learning_rate": 2.4432411820781696e-06, "loss": 2.5684, "step": 9784 }, { "epoch": 1.86, "learning_rate": 2.4431839847473783e-06, "loss": 3.3686, "step": 9785 }, { "epoch": 1.86, "learning_rate": 2.443126787416587e-06, "loss": 2.5154, "step": 9786 }, { "epoch": 1.86, "learning_rate": 2.443069590085796e-06, "loss": 3.0999, "step": 9787 }, { "epoch": 1.86, "learning_rate": 2.443012392755005e-06, "loss": 2.5543, "step": 9788 }, { "epoch": 1.86, "learning_rate": 2.4429551954242135e-06, "loss": 3.2858, "step": 9789 }, { "epoch": 1.86, "learning_rate": 2.4428979980934223e-06, "loss": 2.6054, "step": 9790 }, { "epoch": 1.86, "learning_rate": 2.442840800762631e-06, "loss": 2.3648, "step": 9791 }, { "epoch": 1.87, "learning_rate": 2.44278360343184e-06, "loss": 2.6321, "step": 9792 }, { "epoch": 1.87, "learning_rate": 2.4427264061010487e-06, "loss": 3.1488, "step": 9793 }, { "epoch": 1.87, "learning_rate": 2.4426692087702575e-06, "loss": 2.8959, "step": 9794 }, { "epoch": 1.87, "learning_rate": 2.4426120114394663e-06, "loss": 3.3843, "step": 9795 }, { "epoch": 1.87, "learning_rate": 2.442554814108675e-06, "loss": 2.2552, "step": 9796 }, { "epoch": 1.87, "learning_rate": 2.442497616777884e-06, "loss": 2.234, "step": 9797 }, { "epoch": 1.87, "learning_rate": 2.4424404194470927e-06, "loss": 3.2619, "step": 9798 }, { "epoch": 1.87, "learning_rate": 2.442383222116301e-06, "loss": 3.5313, "step": 9799 }, { "epoch": 1.87, "learning_rate": 2.4423260247855103e-06, "loss": 2.467, "step": 9800 }, { "epoch": 1.87, "learning_rate": 2.4422688274547186e-06, "loss": 2.8722, "step": 9801 }, { "epoch": 1.87, "learning_rate": 2.4422116301239274e-06, "loss": 2.2144, "step": 9802 }, { "epoch": 1.87, "learning_rate": 2.4421544327931366e-06, "loss": 2.514, "step": 9803 }, { "epoch": 1.87, "learning_rate": 2.442097235462345e-06, "loss": 3.0031, "step": 9804 }, { "epoch": 1.87, "learning_rate": 2.4420400381315542e-06, "loss": 2.7856, "step": 9805 }, { "epoch": 1.87, "learning_rate": 2.4419828408007626e-06, "loss": 2.5601, "step": 9806 }, { "epoch": 1.87, "learning_rate": 2.4419256434699714e-06, "loss": 2.3388, "step": 9807 }, { "epoch": 1.87, "learning_rate": 2.44186844613918e-06, "loss": 4.0246, "step": 9808 }, { "epoch": 1.87, "learning_rate": 2.441811248808389e-06, "loss": 3.1262, "step": 9809 }, { "epoch": 1.87, "learning_rate": 2.4417540514775978e-06, "loss": 2.6332, "step": 9810 }, { "epoch": 1.87, "learning_rate": 2.4416968541468066e-06, "loss": 2.9352, "step": 9811 }, { "epoch": 1.87, "learning_rate": 2.4416396568160154e-06, "loss": 2.4744, "step": 9812 }, { "epoch": 1.87, "learning_rate": 2.441582459485224e-06, "loss": 2.5453, "step": 9813 }, { "epoch": 1.87, "learning_rate": 2.441525262154433e-06, "loss": 3.4158, "step": 9814 }, { "epoch": 1.87, "learning_rate": 2.4414680648236417e-06, "loss": 2.9042, "step": 9815 }, { "epoch": 1.87, "learning_rate": 2.44141086749285e-06, "loss": 3.4522, "step": 9816 }, { "epoch": 1.87, "learning_rate": 2.4413536701620593e-06, "loss": 3.1861, "step": 9817 }, { "epoch": 1.87, "learning_rate": 2.4412964728312677e-06, "loss": 3.4567, "step": 9818 }, { "epoch": 1.87, "learning_rate": 2.4412392755004765e-06, "loss": 2.7316, "step": 9819 }, { "epoch": 1.87, "learning_rate": 2.4411820781696857e-06, "loss": 2.3652, "step": 9820 }, { "epoch": 1.87, "learning_rate": 2.441124880838894e-06, "loss": 2.636, "step": 9821 }, { "epoch": 1.87, "learning_rate": 2.4410676835081033e-06, "loss": 2.7017, "step": 9822 }, { "epoch": 1.87, "learning_rate": 2.4410104861773117e-06, "loss": 2.4442, "step": 9823 }, { "epoch": 1.87, "learning_rate": 2.4409532888465205e-06, "loss": 2.5068, "step": 9824 }, { "epoch": 1.87, "learning_rate": 2.4408960915157293e-06, "loss": 2.764, "step": 9825 }, { "epoch": 1.87, "learning_rate": 2.440838894184938e-06, "loss": 3.1371, "step": 9826 }, { "epoch": 1.87, "learning_rate": 2.440781696854147e-06, "loss": 3.1988, "step": 9827 }, { "epoch": 1.87, "learning_rate": 2.4407244995233556e-06, "loss": 2.8525, "step": 9828 }, { "epoch": 1.87, "learning_rate": 2.4406673021925644e-06, "loss": 2.2023, "step": 9829 }, { "epoch": 1.87, "learning_rate": 2.4406101048617732e-06, "loss": 3.0669, "step": 9830 }, { "epoch": 1.87, "learning_rate": 2.440552907530982e-06, "loss": 2.597, "step": 9831 }, { "epoch": 1.87, "learning_rate": 2.440495710200191e-06, "loss": 2.6268, "step": 9832 }, { "epoch": 1.87, "learning_rate": 2.440438512869399e-06, "loss": 3.7164, "step": 9833 }, { "epoch": 1.87, "learning_rate": 2.4403813155386084e-06, "loss": 3.9002, "step": 9834 }, { "epoch": 1.87, "learning_rate": 2.440324118207817e-06, "loss": 2.9182, "step": 9835 }, { "epoch": 1.87, "learning_rate": 2.440266920877026e-06, "loss": 3.4682, "step": 9836 }, { "epoch": 1.87, "learning_rate": 2.4402097235462348e-06, "loss": 3.1472, "step": 9837 }, { "epoch": 1.87, "learning_rate": 2.440152526215443e-06, "loss": 2.4571, "step": 9838 }, { "epoch": 1.87, "learning_rate": 2.4400953288846524e-06, "loss": 3.8045, "step": 9839 }, { "epoch": 1.87, "learning_rate": 2.4400381315538607e-06, "loss": 2.9027, "step": 9840 }, { "epoch": 1.87, "learning_rate": 2.4399809342230695e-06, "loss": 2.867, "step": 9841 }, { "epoch": 1.87, "learning_rate": 2.4399237368922787e-06, "loss": 3.6811, "step": 9842 }, { "epoch": 1.87, "learning_rate": 2.439866539561487e-06, "loss": 3.2564, "step": 9843 }, { "epoch": 1.88, "learning_rate": 2.439809342230696e-06, "loss": 2.6374, "step": 9844 }, { "epoch": 1.88, "learning_rate": 2.4397521448999047e-06, "loss": 3.2475, "step": 9845 }, { "epoch": 1.88, "learning_rate": 2.4396949475691135e-06, "loss": 2.7542, "step": 9846 }, { "epoch": 1.88, "learning_rate": 2.4396377502383223e-06, "loss": 2.797, "step": 9847 }, { "epoch": 1.88, "learning_rate": 2.439580552907531e-06, "loss": 3.1451, "step": 9848 }, { "epoch": 1.88, "learning_rate": 2.43952335557674e-06, "loss": 2.79, "step": 9849 }, { "epoch": 1.88, "learning_rate": 2.4394661582459483e-06, "loss": 2.9993, "step": 9850 }, { "epoch": 1.88, "learning_rate": 2.4394089609151575e-06, "loss": 2.8758, "step": 9851 }, { "epoch": 1.88, "learning_rate": 2.4393517635843663e-06, "loss": 2.4757, "step": 9852 }, { "epoch": 1.88, "learning_rate": 2.439294566253575e-06, "loss": 2.9452, "step": 9853 }, { "epoch": 1.88, "learning_rate": 2.439237368922784e-06, "loss": 3.1792, "step": 9854 }, { "epoch": 1.88, "learning_rate": 2.4391801715919922e-06, "loss": 3.3053, "step": 9855 }, { "epoch": 1.88, "learning_rate": 2.4391229742612014e-06, "loss": 3.1601, "step": 9856 }, { "epoch": 1.88, "learning_rate": 2.43906577693041e-06, "loss": 2.0799, "step": 9857 }, { "epoch": 1.88, "learning_rate": 2.4390085795996186e-06, "loss": 3.1936, "step": 9858 }, { "epoch": 1.88, "learning_rate": 2.438951382268828e-06, "loss": 3.9497, "step": 9859 }, { "epoch": 1.88, "learning_rate": 2.438894184938036e-06, "loss": 2.8287, "step": 9860 }, { "epoch": 1.88, "learning_rate": 2.438836987607245e-06, "loss": 2.568, "step": 9861 }, { "epoch": 1.88, "learning_rate": 2.4387797902764538e-06, "loss": 3.4176, "step": 9862 }, { "epoch": 1.88, "learning_rate": 2.4387225929456626e-06, "loss": 3.3453, "step": 9863 }, { "epoch": 1.88, "learning_rate": 2.4386653956148714e-06, "loss": 2.9806, "step": 9864 }, { "epoch": 1.88, "learning_rate": 2.43860819828408e-06, "loss": 3.0703, "step": 9865 }, { "epoch": 1.88, "learning_rate": 2.438551000953289e-06, "loss": 2.066, "step": 9866 }, { "epoch": 1.88, "learning_rate": 2.4384938036224973e-06, "loss": 2.6391, "step": 9867 }, { "epoch": 1.88, "learning_rate": 2.4384366062917065e-06, "loss": 3.4897, "step": 9868 }, { "epoch": 1.88, "learning_rate": 2.4383794089609153e-06, "loss": 3.4387, "step": 9869 }, { "epoch": 1.88, "learning_rate": 2.438322211630124e-06, "loss": 2.3717, "step": 9870 }, { "epoch": 1.88, "learning_rate": 2.438265014299333e-06, "loss": 2.4897, "step": 9871 }, { "epoch": 1.88, "learning_rate": 2.4382078169685413e-06, "loss": 2.3788, "step": 9872 }, { "epoch": 1.88, "learning_rate": 2.4381506196377505e-06, "loss": 3.2174, "step": 9873 }, { "epoch": 1.88, "learning_rate": 2.438093422306959e-06, "loss": 3.1816, "step": 9874 }, { "epoch": 1.88, "learning_rate": 2.4380362249761677e-06, "loss": 3.2881, "step": 9875 }, { "epoch": 1.88, "learning_rate": 2.437979027645377e-06, "loss": 3.7005, "step": 9876 }, { "epoch": 1.88, "learning_rate": 2.4379218303145853e-06, "loss": 2.9752, "step": 9877 }, { "epoch": 1.88, "learning_rate": 2.437864632983794e-06, "loss": 2.7175, "step": 9878 }, { "epoch": 1.88, "learning_rate": 2.437807435653003e-06, "loss": 2.9431, "step": 9879 }, { "epoch": 1.88, "learning_rate": 2.4377502383222116e-06, "loss": 3.3783, "step": 9880 }, { "epoch": 1.88, "learning_rate": 2.4376930409914204e-06, "loss": 2.7146, "step": 9881 }, { "epoch": 1.88, "learning_rate": 2.4376358436606292e-06, "loss": 3.0937, "step": 9882 }, { "epoch": 1.88, "learning_rate": 2.437578646329838e-06, "loss": 3.0979, "step": 9883 }, { "epoch": 1.88, "learning_rate": 2.437521448999047e-06, "loss": 3.0094, "step": 9884 }, { "epoch": 1.88, "learning_rate": 2.4374642516682556e-06, "loss": 2.6054, "step": 9885 }, { "epoch": 1.88, "learning_rate": 2.4374070543374644e-06, "loss": 2.4652, "step": 9886 }, { "epoch": 1.88, "learning_rate": 2.437349857006673e-06, "loss": 2.2772, "step": 9887 }, { "epoch": 1.88, "learning_rate": 2.437292659675882e-06, "loss": 2.9456, "step": 9888 }, { "epoch": 1.88, "learning_rate": 2.4372354623450904e-06, "loss": 3.414, "step": 9889 }, { "epoch": 1.88, "learning_rate": 2.4371782650142996e-06, "loss": 2.9461, "step": 9890 }, { "epoch": 1.88, "learning_rate": 2.4371210676835084e-06, "loss": 3.1151, "step": 9891 }, { "epoch": 1.88, "learning_rate": 2.4370638703527167e-06, "loss": 2.4936, "step": 9892 }, { "epoch": 1.88, "learning_rate": 2.437006673021926e-06, "loss": 3.4738, "step": 9893 }, { "epoch": 1.88, "learning_rate": 2.4369494756911343e-06, "loss": 3.2131, "step": 9894 }, { "epoch": 1.88, "learning_rate": 2.436892278360343e-06, "loss": 3.0354, "step": 9895 }, { "epoch": 1.88, "learning_rate": 2.436835081029552e-06, "loss": 3.1373, "step": 9896 }, { "epoch": 1.89, "learning_rate": 2.4367778836987607e-06, "loss": 3.1665, "step": 9897 }, { "epoch": 1.89, "learning_rate": 2.4367206863679695e-06, "loss": 2.4514, "step": 9898 }, { "epoch": 1.89, "learning_rate": 2.4366634890371783e-06, "loss": 2.7511, "step": 9899 }, { "epoch": 1.89, "learning_rate": 2.436606291706387e-06, "loss": 2.9549, "step": 9900 }, { "epoch": 1.89, "learning_rate": 2.436549094375596e-06, "loss": 2.6052, "step": 9901 }, { "epoch": 1.89, "learning_rate": 2.4364918970448047e-06, "loss": 2.8094, "step": 9902 }, { "epoch": 1.89, "learning_rate": 2.4364346997140135e-06, "loss": 3.3756, "step": 9903 }, { "epoch": 1.89, "learning_rate": 2.4363775023832223e-06, "loss": 2.5369, "step": 9904 }, { "epoch": 1.89, "learning_rate": 2.436320305052431e-06, "loss": 3.0452, "step": 9905 }, { "epoch": 1.89, "learning_rate": 2.4362631077216394e-06, "loss": 2.4616, "step": 9906 }, { "epoch": 1.89, "learning_rate": 2.4362059103908487e-06, "loss": 3.1584, "step": 9907 }, { "epoch": 1.89, "learning_rate": 2.4361487130600574e-06, "loss": 2.8249, "step": 9908 }, { "epoch": 1.89, "learning_rate": 2.436091515729266e-06, "loss": 2.9395, "step": 9909 }, { "epoch": 1.89, "learning_rate": 2.436034318398475e-06, "loss": 3.4529, "step": 9910 }, { "epoch": 1.89, "learning_rate": 2.4359771210676834e-06, "loss": 3.1607, "step": 9911 }, { "epoch": 1.89, "learning_rate": 2.435919923736892e-06, "loss": 3.7493, "step": 9912 }, { "epoch": 1.89, "learning_rate": 2.435862726406101e-06, "loss": 3.273, "step": 9913 }, { "epoch": 1.89, "learning_rate": 2.4358055290753098e-06, "loss": 3.2132, "step": 9914 }, { "epoch": 1.89, "learning_rate": 2.4357483317445186e-06, "loss": 2.756, "step": 9915 }, { "epoch": 1.89, "learning_rate": 2.4356911344137274e-06, "loss": 2.7532, "step": 9916 }, { "epoch": 1.89, "learning_rate": 2.435633937082936e-06, "loss": 3.3161, "step": 9917 }, { "epoch": 1.89, "learning_rate": 2.435576739752145e-06, "loss": 2.8099, "step": 9918 }, { "epoch": 1.89, "learning_rate": 2.4355195424213538e-06, "loss": 2.3406, "step": 9919 }, { "epoch": 1.89, "learning_rate": 2.4354623450905625e-06, "loss": 3.4107, "step": 9920 }, { "epoch": 1.89, "learning_rate": 2.4354051477597713e-06, "loss": 2.3449, "step": 9921 }, { "epoch": 1.89, "learning_rate": 2.43534795042898e-06, "loss": 2.9724, "step": 9922 }, { "epoch": 1.89, "learning_rate": 2.4352907530981885e-06, "loss": 2.6677, "step": 9923 }, { "epoch": 1.89, "learning_rate": 2.4352335557673977e-06, "loss": 2.6941, "step": 9924 }, { "epoch": 1.89, "learning_rate": 2.4351763584366065e-06, "loss": 3.3463, "step": 9925 }, { "epoch": 1.89, "learning_rate": 2.435119161105815e-06, "loss": 2.5781, "step": 9926 }, { "epoch": 1.89, "learning_rate": 2.435061963775024e-06, "loss": 2.9117, "step": 9927 }, { "epoch": 1.89, "learning_rate": 2.4350047664442325e-06, "loss": 2.4186, "step": 9928 }, { "epoch": 1.89, "learning_rate": 2.4349475691134413e-06, "loss": 3.4491, "step": 9929 }, { "epoch": 1.89, "learning_rate": 2.4348903717826505e-06, "loss": 2.7351, "step": 9930 }, { "epoch": 1.89, "learning_rate": 2.434833174451859e-06, "loss": 2.9265, "step": 9931 }, { "epoch": 1.89, "learning_rate": 2.4347759771210676e-06, "loss": 3.441, "step": 9932 }, { "epoch": 1.89, "learning_rate": 2.4347187797902764e-06, "loss": 2.3454, "step": 9933 }, { "epoch": 1.89, "learning_rate": 2.4346615824594852e-06, "loss": 3.0174, "step": 9934 }, { "epoch": 1.89, "learning_rate": 2.434604385128694e-06, "loss": 2.3432, "step": 9935 }, { "epoch": 1.89, "learning_rate": 2.434547187797903e-06, "loss": 2.9179, "step": 9936 }, { "epoch": 1.89, "learning_rate": 2.4344899904671116e-06, "loss": 3.6072, "step": 9937 }, { "epoch": 1.89, "learning_rate": 2.4344327931363204e-06, "loss": 2.8441, "step": 9938 }, { "epoch": 1.89, "learning_rate": 2.434375595805529e-06, "loss": 2.1888, "step": 9939 }, { "epoch": 1.89, "learning_rate": 2.434318398474738e-06, "loss": 2.2139, "step": 9940 }, { "epoch": 1.89, "learning_rate": 2.434261201143947e-06, "loss": 2.6687, "step": 9941 }, { "epoch": 1.89, "learning_rate": 2.4342040038131556e-06, "loss": 2.7016, "step": 9942 }, { "epoch": 1.89, "learning_rate": 2.434146806482364e-06, "loss": 2.6641, "step": 9943 }, { "epoch": 1.89, "learning_rate": 2.434089609151573e-06, "loss": 2.6726, "step": 9944 }, { "epoch": 1.89, "learning_rate": 2.4340324118207815e-06, "loss": 2.7172, "step": 9945 }, { "epoch": 1.89, "learning_rate": 2.4339752144899903e-06, "loss": 2.4842, "step": 9946 }, { "epoch": 1.89, "learning_rate": 2.4339180171591996e-06, "loss": 3.0032, "step": 9947 }, { "epoch": 1.89, "learning_rate": 2.433860819828408e-06, "loss": 2.4491, "step": 9948 }, { "epoch": 1.9, "learning_rate": 2.4338036224976167e-06, "loss": 2.9599, "step": 9949 }, { "epoch": 1.9, "learning_rate": 2.4337464251668255e-06, "loss": 3.8718, "step": 9950 }, { "epoch": 1.9, "learning_rate": 2.4336892278360343e-06, "loss": 3.3035, "step": 9951 }, { "epoch": 1.9, "learning_rate": 2.433632030505243e-06, "loss": 2.4055, "step": 9952 }, { "epoch": 1.9, "learning_rate": 2.433574833174452e-06, "loss": 3.5119, "step": 9953 }, { "epoch": 1.9, "learning_rate": 2.4335176358436607e-06, "loss": 3.2661, "step": 9954 }, { "epoch": 1.9, "learning_rate": 2.4334604385128695e-06, "loss": 3.128, "step": 9955 }, { "epoch": 1.9, "learning_rate": 2.4334032411820783e-06, "loss": 3.5152, "step": 9956 }, { "epoch": 1.9, "learning_rate": 2.433346043851287e-06, "loss": 3.0943, "step": 9957 }, { "epoch": 1.9, "learning_rate": 2.433288846520496e-06, "loss": 3.052, "step": 9958 }, { "epoch": 1.9, "learning_rate": 2.4332316491897047e-06, "loss": 2.5469, "step": 9959 }, { "epoch": 1.9, "learning_rate": 2.433174451858913e-06, "loss": 2.4651, "step": 9960 }, { "epoch": 1.9, "learning_rate": 2.4331172545281222e-06, "loss": 3.4995, "step": 9961 }, { "epoch": 1.9, "learning_rate": 2.4330600571973306e-06, "loss": 3.2773, "step": 9962 }, { "epoch": 1.9, "learning_rate": 2.4330028598665394e-06, "loss": 2.9807, "step": 9963 }, { "epoch": 1.9, "learning_rate": 2.4329456625357486e-06, "loss": 2.6148, "step": 9964 }, { "epoch": 1.9, "learning_rate": 2.432888465204957e-06, "loss": 2.6873, "step": 9965 }, { "epoch": 1.9, "learning_rate": 2.4328312678741662e-06, "loss": 3.2161, "step": 9966 }, { "epoch": 1.9, "learning_rate": 2.4327740705433746e-06, "loss": 3.573, "step": 9967 }, { "epoch": 1.9, "learning_rate": 2.4327168732125834e-06, "loss": 2.9391, "step": 9968 }, { "epoch": 1.9, "learning_rate": 2.4326596758817926e-06, "loss": 3.0746, "step": 9969 }, { "epoch": 1.9, "learning_rate": 2.432602478551001e-06, "loss": 2.977, "step": 9970 }, { "epoch": 1.9, "learning_rate": 2.4325452812202098e-06, "loss": 3.3339, "step": 9971 }, { "epoch": 1.9, "learning_rate": 2.4324880838894186e-06, "loss": 3.0325, "step": 9972 }, { "epoch": 1.9, "learning_rate": 2.4324308865586273e-06, "loss": 3.5029, "step": 9973 }, { "epoch": 1.9, "learning_rate": 2.432373689227836e-06, "loss": 2.8683, "step": 9974 }, { "epoch": 1.9, "learning_rate": 2.432316491897045e-06, "loss": 3.0233, "step": 9975 }, { "epoch": 1.9, "learning_rate": 2.4322592945662537e-06, "loss": 2.5384, "step": 9976 }, { "epoch": 1.9, "learning_rate": 2.432202097235462e-06, "loss": 3.1493, "step": 9977 }, { "epoch": 1.9, "learning_rate": 2.4321448999046713e-06, "loss": 2.8473, "step": 9978 }, { "epoch": 1.9, "learning_rate": 2.43208770257388e-06, "loss": 2.8284, "step": 9979 }, { "epoch": 1.9, "learning_rate": 2.4320305052430885e-06, "loss": 2.6816, "step": 9980 }, { "epoch": 1.9, "learning_rate": 2.4319733079122977e-06, "loss": 2.9751, "step": 9981 }, { "epoch": 1.9, "learning_rate": 2.431916110581506e-06, "loss": 3.157, "step": 9982 }, { "epoch": 1.9, "learning_rate": 2.4318589132507153e-06, "loss": 2.8562, "step": 9983 }, { "epoch": 1.9, "learning_rate": 2.4318017159199237e-06, "loss": 2.051, "step": 9984 }, { "epoch": 1.9, "learning_rate": 2.4317445185891325e-06, "loss": 2.5971, "step": 9985 }, { "epoch": 1.9, "learning_rate": 2.4316873212583417e-06, "loss": 2.8199, "step": 9986 }, { "epoch": 1.9, "learning_rate": 2.43163012392755e-06, "loss": 2.1672, "step": 9987 }, { "epoch": 1.9, "learning_rate": 2.431572926596759e-06, "loss": 3.2066, "step": 9988 }, { "epoch": 1.9, "learning_rate": 2.4315157292659676e-06, "loss": 2.2597, "step": 9989 }, { "epoch": 1.9, "learning_rate": 2.4314585319351764e-06, "loss": 1.9886, "step": 9990 }, { "epoch": 1.9, "learning_rate": 2.4314013346043852e-06, "loss": 3.3274, "step": 9991 }, { "epoch": 1.9, "learning_rate": 2.431344137273594e-06, "loss": 2.5425, "step": 9992 }, { "epoch": 1.9, "learning_rate": 2.431286939942803e-06, "loss": 3.2249, "step": 9993 }, { "epoch": 1.9, "learning_rate": 2.431229742612011e-06, "loss": 3.1635, "step": 9994 }, { "epoch": 1.9, "learning_rate": 2.4311725452812204e-06, "loss": 2.8029, "step": 9995 }, { "epoch": 1.9, "learning_rate": 2.431115347950429e-06, "loss": 2.3847, "step": 9996 }, { "epoch": 1.9, "learning_rate": 2.4310581506196376e-06, "loss": 2.0651, "step": 9997 }, { "epoch": 1.9, "learning_rate": 2.4310009532888468e-06, "loss": 3.189, "step": 9998 }, { "epoch": 1.9, "learning_rate": 2.430943755958055e-06, "loss": 2.7196, "step": 9999 }, { "epoch": 1.9, "learning_rate": 2.4308865586272644e-06, "loss": 3.1787, "step": 10000 }, { "epoch": 1.9, "learning_rate": 2.4308293612964727e-06, "loss": 2.6789, "step": 10001 }, { "epoch": 1.91, "learning_rate": 2.4307721639656815e-06, "loss": 4.3133, "step": 10002 }, { "epoch": 1.91, "learning_rate": 2.4307149666348907e-06, "loss": 3.0725, "step": 10003 }, { "epoch": 1.91, "learning_rate": 2.430657769304099e-06, "loss": 2.4258, "step": 10004 }, { "epoch": 1.91, "learning_rate": 2.430600571973308e-06, "loss": 2.9334, "step": 10005 }, { "epoch": 1.91, "learning_rate": 2.4305433746425167e-06, "loss": 2.5129, "step": 10006 }, { "epoch": 1.91, "learning_rate": 2.4304861773117255e-06, "loss": 2.4504, "step": 10007 }, { "epoch": 1.91, "learning_rate": 2.4304289799809343e-06, "loss": 2.2435, "step": 10008 }, { "epoch": 1.91, "learning_rate": 2.430371782650143e-06, "loss": 2.8096, "step": 10009 }, { "epoch": 1.91, "learning_rate": 2.430314585319352e-06, "loss": 2.0572, "step": 10010 }, { "epoch": 1.91, "learning_rate": 2.4302573879885602e-06, "loss": 3.4343, "step": 10011 }, { "epoch": 1.91, "learning_rate": 2.4302001906577695e-06, "loss": 2.7871, "step": 10012 }, { "epoch": 1.91, "learning_rate": 2.4301429933269783e-06, "loss": 3.1374, "step": 10013 }, { "epoch": 1.91, "learning_rate": 2.4300857959961866e-06, "loss": 3.7289, "step": 10014 }, { "epoch": 1.91, "learning_rate": 2.430028598665396e-06, "loss": 3.0988, "step": 10015 }, { "epoch": 1.91, "learning_rate": 2.4299714013346042e-06, "loss": 2.6934, "step": 10016 }, { "epoch": 1.91, "learning_rate": 2.4299142040038134e-06, "loss": 3.4314, "step": 10017 }, { "epoch": 1.91, "learning_rate": 2.4298570066730222e-06, "loss": 1.7749, "step": 10018 }, { "epoch": 1.91, "learning_rate": 2.4297998093422306e-06, "loss": 3.1345, "step": 10019 }, { "epoch": 1.91, "learning_rate": 2.42974261201144e-06, "loss": 2.5034, "step": 10020 }, { "epoch": 1.91, "learning_rate": 2.429685414680648e-06, "loss": 3.4104, "step": 10021 }, { "epoch": 1.91, "learning_rate": 2.429628217349857e-06, "loss": 2.3726, "step": 10022 }, { "epoch": 1.91, "learning_rate": 2.4295710200190658e-06, "loss": 2.4077, "step": 10023 }, { "epoch": 1.91, "learning_rate": 2.4295138226882746e-06, "loss": 3.1346, "step": 10024 }, { "epoch": 1.91, "learning_rate": 2.4294566253574834e-06, "loss": 3.1272, "step": 10025 }, { "epoch": 1.91, "learning_rate": 2.429399428026692e-06, "loss": 2.3421, "step": 10026 }, { "epoch": 1.91, "learning_rate": 2.429342230695901e-06, "loss": 2.6559, "step": 10027 }, { "epoch": 1.91, "learning_rate": 2.4292850333651097e-06, "loss": 2.2247, "step": 10028 }, { "epoch": 1.91, "learning_rate": 2.4292278360343185e-06, "loss": 3.0973, "step": 10029 }, { "epoch": 1.91, "learning_rate": 2.4291706387035273e-06, "loss": 2.5544, "step": 10030 }, { "epoch": 1.91, "learning_rate": 2.429113441372736e-06, "loss": 2.768, "step": 10031 }, { "epoch": 1.91, "learning_rate": 2.429056244041945e-06, "loss": 2.921, "step": 10032 }, { "epoch": 1.91, "learning_rate": 2.4289990467111533e-06, "loss": 2.6947, "step": 10033 }, { "epoch": 1.91, "learning_rate": 2.4289418493803625e-06, "loss": 3.5666, "step": 10034 }, { "epoch": 1.91, "learning_rate": 2.4288846520495713e-06, "loss": 2.9234, "step": 10035 }, { "epoch": 1.91, "learning_rate": 2.4288274547187797e-06, "loss": 2.2555, "step": 10036 }, { "epoch": 1.91, "learning_rate": 2.428770257387989e-06, "loss": 2.5104, "step": 10037 }, { "epoch": 1.91, "learning_rate": 2.4287130600571973e-06, "loss": 2.6728, "step": 10038 }, { "epoch": 1.91, "learning_rate": 2.428655862726406e-06, "loss": 3.0413, "step": 10039 }, { "epoch": 1.91, "learning_rate": 2.428598665395615e-06, "loss": 3.0472, "step": 10040 }, { "epoch": 1.91, "learning_rate": 2.4285414680648236e-06, "loss": 2.8927, "step": 10041 }, { "epoch": 1.91, "learning_rate": 2.4284842707340324e-06, "loss": 3.1264, "step": 10042 }, { "epoch": 1.91, "learning_rate": 2.4284270734032412e-06, "loss": 3.5648, "step": 10043 }, { "epoch": 1.91, "learning_rate": 2.42836987607245e-06, "loss": 2.7512, "step": 10044 }, { "epoch": 1.91, "learning_rate": 2.428312678741659e-06, "loss": 2.9817, "step": 10045 }, { "epoch": 1.91, "learning_rate": 2.4282554814108676e-06, "loss": 2.5432, "step": 10046 }, { "epoch": 1.91, "learning_rate": 2.4281982840800764e-06, "loss": 3.7774, "step": 10047 }, { "epoch": 1.91, "learning_rate": 2.428141086749285e-06, "loss": 2.8433, "step": 10048 }, { "epoch": 1.91, "learning_rate": 2.428083889418494e-06, "loss": 2.5269, "step": 10049 }, { "epoch": 1.91, "learning_rate": 2.4280266920877024e-06, "loss": 2.6331, "step": 10050 }, { "epoch": 1.91, "learning_rate": 2.4279694947569116e-06, "loss": 2.4713, "step": 10051 }, { "epoch": 1.91, "learning_rate": 2.4279122974261204e-06, "loss": 3.0229, "step": 10052 }, { "epoch": 1.91, "learning_rate": 2.4278551000953287e-06, "loss": 3.2444, "step": 10053 }, { "epoch": 1.92, "learning_rate": 2.427797902764538e-06, "loss": 2.4737, "step": 10054 }, { "epoch": 1.92, "learning_rate": 2.4277407054337463e-06, "loss": 2.3424, "step": 10055 }, { "epoch": 1.92, "learning_rate": 2.427683508102955e-06, "loss": 2.8749, "step": 10056 }, { "epoch": 1.92, "learning_rate": 2.4276263107721643e-06, "loss": 3.2781, "step": 10057 }, { "epoch": 1.92, "learning_rate": 2.4275691134413727e-06, "loss": 2.9906, "step": 10058 }, { "epoch": 1.92, "learning_rate": 2.4275119161105815e-06, "loss": 3.0484, "step": 10059 }, { "epoch": 1.92, "learning_rate": 2.4274547187797903e-06, "loss": 3.0464, "step": 10060 }, { "epoch": 1.92, "learning_rate": 2.427397521448999e-06, "loss": 2.3393, "step": 10061 }, { "epoch": 1.92, "learning_rate": 2.427340324118208e-06, "loss": 2.8325, "step": 10062 }, { "epoch": 1.92, "learning_rate": 2.4272831267874167e-06, "loss": 2.5589, "step": 10063 }, { "epoch": 1.92, "learning_rate": 2.4272259294566255e-06, "loss": 2.9705, "step": 10064 }, { "epoch": 1.92, "learning_rate": 2.4271687321258343e-06, "loss": 4.067, "step": 10065 }, { "epoch": 1.92, "learning_rate": 2.427111534795043e-06, "loss": 3.0269, "step": 10066 }, { "epoch": 1.92, "learning_rate": 2.427054337464252e-06, "loss": 3.0807, "step": 10067 }, { "epoch": 1.92, "learning_rate": 2.4269971401334606e-06, "loss": 2.2823, "step": 10068 }, { "epoch": 1.92, "learning_rate": 2.4269399428026694e-06, "loss": 3.0638, "step": 10069 }, { "epoch": 1.92, "learning_rate": 2.426882745471878e-06, "loss": 3.5415, "step": 10070 }, { "epoch": 1.92, "learning_rate": 2.426825548141087e-06, "loss": 3.326, "step": 10071 }, { "epoch": 1.92, "learning_rate": 2.4267683508102954e-06, "loss": 3.4203, "step": 10072 }, { "epoch": 1.92, "learning_rate": 2.426711153479504e-06, "loss": 3.3299, "step": 10073 }, { "epoch": 1.92, "learning_rate": 2.4266539561487134e-06, "loss": 3.4228, "step": 10074 }, { "epoch": 1.92, "learning_rate": 2.4265967588179218e-06, "loss": 2.753, "step": 10075 }, { "epoch": 1.92, "learning_rate": 2.4265395614871306e-06, "loss": 3.125, "step": 10076 }, { "epoch": 1.92, "learning_rate": 2.4264823641563394e-06, "loss": 2.462, "step": 10077 }, { "epoch": 1.92, "learning_rate": 2.426425166825548e-06, "loss": 3.0459, "step": 10078 }, { "epoch": 1.92, "learning_rate": 2.426367969494757e-06, "loss": 2.8715, "step": 10079 }, { "epoch": 1.92, "learning_rate": 2.4263107721639657e-06, "loss": 3.7287, "step": 10080 }, { "epoch": 1.92, "learning_rate": 2.4262535748331745e-06, "loss": 2.2642, "step": 10081 }, { "epoch": 1.92, "learning_rate": 2.4261963775023833e-06, "loss": 3.281, "step": 10082 }, { "epoch": 1.92, "learning_rate": 2.426139180171592e-06, "loss": 2.8086, "step": 10083 }, { "epoch": 1.92, "learning_rate": 2.426081982840801e-06, "loss": 2.3762, "step": 10084 }, { "epoch": 1.92, "learning_rate": 2.4260247855100097e-06, "loss": 2.772, "step": 10085 }, { "epoch": 1.92, "learning_rate": 2.4259675881792185e-06, "loss": 3.0901, "step": 10086 }, { "epoch": 1.92, "learning_rate": 2.425910390848427e-06, "loss": 2.8508, "step": 10087 }, { "epoch": 1.92, "learning_rate": 2.425853193517636e-06, "loss": 3.5269, "step": 10088 }, { "epoch": 1.92, "learning_rate": 2.4257959961868445e-06, "loss": 3.192, "step": 10089 }, { "epoch": 1.92, "learning_rate": 2.4257387988560533e-06, "loss": 2.0876, "step": 10090 }, { "epoch": 1.92, "learning_rate": 2.4256816015252625e-06, "loss": 3.5383, "step": 10091 }, { "epoch": 1.92, "learning_rate": 2.425624404194471e-06, "loss": 3.6399, "step": 10092 }, { "epoch": 1.92, "learning_rate": 2.4255672068636796e-06, "loss": 2.398, "step": 10093 }, { "epoch": 1.92, "learning_rate": 2.4255100095328884e-06, "loss": 3.2333, "step": 10094 }, { "epoch": 1.92, "learning_rate": 2.4254528122020972e-06, "loss": 3.7726, "step": 10095 }, { "epoch": 1.92, "learning_rate": 2.425395614871306e-06, "loss": 2.8736, "step": 10096 }, { "epoch": 1.92, "learning_rate": 2.425338417540515e-06, "loss": 3.5163, "step": 10097 }, { "epoch": 1.92, "learning_rate": 2.4252812202097236e-06, "loss": 2.8126, "step": 10098 }, { "epoch": 1.92, "learning_rate": 2.4252240228789324e-06, "loss": 2.6785, "step": 10099 }, { "epoch": 1.92, "learning_rate": 2.425166825548141e-06, "loss": 3.0499, "step": 10100 }, { "epoch": 1.92, "learning_rate": 2.42510962821735e-06, "loss": 3.1373, "step": 10101 }, { "epoch": 1.92, "learning_rate": 2.4250524308865588e-06, "loss": 2.1024, "step": 10102 }, { "epoch": 1.92, "learning_rate": 2.4249952335557676e-06, "loss": 3.327, "step": 10103 }, { "epoch": 1.92, "learning_rate": 2.424938036224976e-06, "loss": 3.5485, "step": 10104 }, { "epoch": 1.92, "learning_rate": 2.424880838894185e-06, "loss": 3.2106, "step": 10105 }, { "epoch": 1.92, "learning_rate": 2.424823641563394e-06, "loss": 2.7907, "step": 10106 }, { "epoch": 1.93, "learning_rate": 2.4247664442326023e-06, "loss": 3.9038, "step": 10107 }, { "epoch": 1.93, "learning_rate": 2.4247092469018116e-06, "loss": 2.8971, "step": 10108 }, { "epoch": 1.93, "learning_rate": 2.42465204957102e-06, "loss": 2.5349, "step": 10109 }, { "epoch": 1.93, "learning_rate": 2.4245948522402287e-06, "loss": 2.3304, "step": 10110 }, { "epoch": 1.93, "learning_rate": 2.4245376549094375e-06, "loss": 2.7387, "step": 10111 }, { "epoch": 1.93, "learning_rate": 2.4244804575786463e-06, "loss": 3.7107, "step": 10112 }, { "epoch": 1.93, "learning_rate": 2.4244232602478555e-06, "loss": 2.6655, "step": 10113 }, { "epoch": 1.93, "learning_rate": 2.424366062917064e-06, "loss": 2.3966, "step": 10114 }, { "epoch": 1.93, "learning_rate": 2.4243088655862727e-06, "loss": 3.0917, "step": 10115 }, { "epoch": 1.93, "learning_rate": 2.4242516682554815e-06, "loss": 2.2795, "step": 10116 }, { "epoch": 1.93, "learning_rate": 2.4241944709246903e-06, "loss": 3.1209, "step": 10117 }, { "epoch": 1.93, "learning_rate": 2.424137273593899e-06, "loss": 2.5517, "step": 10118 }, { "epoch": 1.93, "learning_rate": 2.424080076263108e-06, "loss": 2.79, "step": 10119 }, { "epoch": 1.93, "learning_rate": 2.4240228789323167e-06, "loss": 2.5761, "step": 10120 }, { "epoch": 1.93, "learning_rate": 2.423965681601525e-06, "loss": 3.287, "step": 10121 }, { "epoch": 1.93, "learning_rate": 2.4239084842707342e-06, "loss": 3.6923, "step": 10122 }, { "epoch": 1.93, "learning_rate": 2.423851286939943e-06, "loss": 2.6635, "step": 10123 }, { "epoch": 1.93, "learning_rate": 2.4237940896091514e-06, "loss": 2.3175, "step": 10124 }, { "epoch": 1.93, "learning_rate": 2.4237368922783606e-06, "loss": 2.2305, "step": 10125 }, { "epoch": 1.93, "learning_rate": 2.423679694947569e-06, "loss": 1.8562, "step": 10126 }, { "epoch": 1.93, "learning_rate": 2.4236224976167778e-06, "loss": 2.5496, "step": 10127 }, { "epoch": 1.93, "learning_rate": 2.4235653002859866e-06, "loss": 3.2563, "step": 10128 }, { "epoch": 1.93, "learning_rate": 2.4235081029551954e-06, "loss": 1.984, "step": 10129 }, { "epoch": 1.93, "learning_rate": 2.4234509056244046e-06, "loss": 3.5328, "step": 10130 }, { "epoch": 1.93, "learning_rate": 2.423393708293613e-06, "loss": 3.6786, "step": 10131 }, { "epoch": 1.93, "learning_rate": 2.4233365109628218e-06, "loss": 3.4301, "step": 10132 }, { "epoch": 1.93, "learning_rate": 2.4232793136320305e-06, "loss": 3.4161, "step": 10133 }, { "epoch": 1.93, "learning_rate": 2.4232221163012393e-06, "loss": 3.1724, "step": 10134 }, { "epoch": 1.93, "learning_rate": 2.423164918970448e-06, "loss": 2.7283, "step": 10135 }, { "epoch": 1.93, "learning_rate": 2.423107721639657e-06, "loss": 2.4461, "step": 10136 }, { "epoch": 1.93, "learning_rate": 2.4230505243088657e-06, "loss": 3.1752, "step": 10137 }, { "epoch": 1.93, "learning_rate": 2.422993326978074e-06, "loss": 4.0379, "step": 10138 }, { "epoch": 1.93, "learning_rate": 2.4229361296472833e-06, "loss": 2.8582, "step": 10139 }, { "epoch": 1.93, "learning_rate": 2.422878932316492e-06, "loss": 2.5873, "step": 10140 }, { "epoch": 1.93, "learning_rate": 2.4228217349857005e-06, "loss": 2.8672, "step": 10141 }, { "epoch": 1.93, "learning_rate": 2.4227645376549097e-06, "loss": 2.4807, "step": 10142 }, { "epoch": 1.93, "learning_rate": 2.422707340324118e-06, "loss": 3.3578, "step": 10143 }, { "epoch": 1.93, "learning_rate": 2.422650142993327e-06, "loss": 3.6927, "step": 10144 }, { "epoch": 1.93, "learning_rate": 2.422592945662536e-06, "loss": 2.9664, "step": 10145 }, { "epoch": 1.93, "learning_rate": 2.4225357483317444e-06, "loss": 3.33, "step": 10146 }, { "epoch": 1.93, "learning_rate": 2.4224785510009537e-06, "loss": 2.5004, "step": 10147 }, { "epoch": 1.93, "learning_rate": 2.422421353670162e-06, "loss": 2.7989, "step": 10148 }, { "epoch": 1.93, "learning_rate": 2.422364156339371e-06, "loss": 2.9381, "step": 10149 }, { "epoch": 1.93, "learning_rate": 2.4223069590085796e-06, "loss": 4.0502, "step": 10150 }, { "epoch": 1.93, "learning_rate": 2.4222497616777884e-06, "loss": 2.8752, "step": 10151 }, { "epoch": 1.93, "learning_rate": 2.422192564346997e-06, "loss": 2.1617, "step": 10152 }, { "epoch": 1.93, "learning_rate": 2.422135367016206e-06, "loss": 3.4027, "step": 10153 }, { "epoch": 1.93, "learning_rate": 2.422078169685415e-06, "loss": 2.9836, "step": 10154 }, { "epoch": 1.93, "learning_rate": 2.4220209723546236e-06, "loss": 3.1849, "step": 10155 }, { "epoch": 1.93, "learning_rate": 2.4219637750238324e-06, "loss": 2.5566, "step": 10156 }, { "epoch": 1.93, "learning_rate": 2.421906577693041e-06, "loss": 2.8288, "step": 10157 }, { "epoch": 1.93, "learning_rate": 2.4218493803622495e-06, "loss": 2.8384, "step": 10158 }, { "epoch": 1.94, "learning_rate": 2.4217921830314588e-06, "loss": 1.9823, "step": 10159 }, { "epoch": 1.94, "learning_rate": 2.421734985700667e-06, "loss": 2.9491, "step": 10160 }, { "epoch": 1.94, "learning_rate": 2.4216777883698764e-06, "loss": 2.8195, "step": 10161 }, { "epoch": 1.94, "learning_rate": 2.421620591039085e-06, "loss": 3.3827, "step": 10162 }, { "epoch": 1.94, "learning_rate": 2.4215633937082935e-06, "loss": 2.7482, "step": 10163 }, { "epoch": 1.94, "learning_rate": 2.4215061963775027e-06, "loss": 3.5164, "step": 10164 }, { "epoch": 1.94, "learning_rate": 2.421448999046711e-06, "loss": 3.295, "step": 10165 }, { "epoch": 1.94, "learning_rate": 2.42139180171592e-06, "loss": 3.1679, "step": 10166 }, { "epoch": 1.94, "learning_rate": 2.4213346043851287e-06, "loss": 2.8607, "step": 10167 }, { "epoch": 1.94, "learning_rate": 2.4212774070543375e-06, "loss": 2.0292, "step": 10168 }, { "epoch": 1.94, "learning_rate": 2.4212202097235463e-06, "loss": 3.0242, "step": 10169 }, { "epoch": 1.94, "learning_rate": 2.421163012392755e-06, "loss": 3.3809, "step": 10170 }, { "epoch": 1.94, "learning_rate": 2.421105815061964e-06, "loss": 2.5127, "step": 10171 }, { "epoch": 1.94, "learning_rate": 2.4210486177311727e-06, "loss": 2.5506, "step": 10172 }, { "epoch": 1.94, "learning_rate": 2.4209914204003815e-06, "loss": 2.9641, "step": 10173 }, { "epoch": 1.94, "learning_rate": 2.4209342230695902e-06, "loss": 3.1346, "step": 10174 }, { "epoch": 1.94, "learning_rate": 2.4208770257387986e-06, "loss": 2.5999, "step": 10175 }, { "epoch": 1.94, "learning_rate": 2.420819828408008e-06, "loss": 3.4578, "step": 10176 }, { "epoch": 1.94, "learning_rate": 2.420762631077216e-06, "loss": 3.2468, "step": 10177 }, { "epoch": 1.94, "learning_rate": 2.4207054337464254e-06, "loss": 3.322, "step": 10178 }, { "epoch": 1.94, "learning_rate": 2.4206482364156342e-06, "loss": 2.5659, "step": 10179 }, { "epoch": 1.94, "learning_rate": 2.4205910390848426e-06, "loss": 3.19, "step": 10180 }, { "epoch": 1.94, "learning_rate": 2.420533841754052e-06, "loss": 2.283, "step": 10181 }, { "epoch": 1.94, "learning_rate": 2.42047664442326e-06, "loss": 2.646, "step": 10182 }, { "epoch": 1.94, "learning_rate": 2.420419447092469e-06, "loss": 2.5755, "step": 10183 }, { "epoch": 1.94, "learning_rate": 2.4203622497616778e-06, "loss": 3.384, "step": 10184 }, { "epoch": 1.94, "learning_rate": 2.4203050524308866e-06, "loss": 3.0614, "step": 10185 }, { "epoch": 1.94, "learning_rate": 2.4202478551000954e-06, "loss": 2.4232, "step": 10186 }, { "epoch": 1.94, "learning_rate": 2.420190657769304e-06, "loss": 2.5604, "step": 10187 }, { "epoch": 1.94, "learning_rate": 2.420133460438513e-06, "loss": 3.009, "step": 10188 }, { "epoch": 1.94, "learning_rate": 2.4200762631077217e-06, "loss": 3.5908, "step": 10189 }, { "epoch": 1.94, "learning_rate": 2.4200190657769305e-06, "loss": 3.2162, "step": 10190 }, { "epoch": 1.94, "learning_rate": 2.4199618684461393e-06, "loss": 2.3491, "step": 10191 }, { "epoch": 1.94, "learning_rate": 2.4199046711153477e-06, "loss": 3.0319, "step": 10192 }, { "epoch": 1.94, "learning_rate": 2.419847473784557e-06, "loss": 3.2143, "step": 10193 }, { "epoch": 1.94, "learning_rate": 2.4197902764537657e-06, "loss": 3.9125, "step": 10194 }, { "epoch": 1.94, "learning_rate": 2.4197330791229745e-06, "loss": 3.244, "step": 10195 }, { "epoch": 1.94, "learning_rate": 2.4196758817921833e-06, "loss": 3.4611, "step": 10196 }, { "epoch": 1.94, "learning_rate": 2.4196186844613917e-06, "loss": 2.7503, "step": 10197 }, { "epoch": 1.94, "learning_rate": 2.419561487130601e-06, "loss": 2.4304, "step": 10198 }, { "epoch": 1.94, "learning_rate": 2.4195042897998092e-06, "loss": 3.5756, "step": 10199 }, { "epoch": 1.94, "learning_rate": 2.419447092469018e-06, "loss": 2.7316, "step": 10200 }, { "epoch": 1.94, "learning_rate": 2.4193898951382273e-06, "loss": 3.1566, "step": 10201 }, { "epoch": 1.94, "learning_rate": 2.4193326978074356e-06, "loss": 3.0532, "step": 10202 }, { "epoch": 1.94, "learning_rate": 2.4192755004766444e-06, "loss": 3.0342, "step": 10203 }, { "epoch": 1.94, "learning_rate": 2.4192183031458532e-06, "loss": 2.3521, "step": 10204 }, { "epoch": 1.94, "learning_rate": 2.419161105815062e-06, "loss": 2.5837, "step": 10205 }, { "epoch": 1.94, "learning_rate": 2.419103908484271e-06, "loss": 3.3845, "step": 10206 }, { "epoch": 1.94, "learning_rate": 2.4190467111534796e-06, "loss": 2.9996, "step": 10207 }, { "epoch": 1.94, "learning_rate": 2.4189895138226884e-06, "loss": 2.8713, "step": 10208 }, { "epoch": 1.94, "learning_rate": 2.4189323164918968e-06, "loss": 2.5474, "step": 10209 }, { "epoch": 1.94, "learning_rate": 2.418875119161106e-06, "loss": 3.4123, "step": 10210 }, { "epoch": 1.94, "learning_rate": 2.4188179218303148e-06, "loss": 3.4344, "step": 10211 }, { "epoch": 1.95, "learning_rate": 2.4187607244995236e-06, "loss": 2.5177, "step": 10212 }, { "epoch": 1.95, "learning_rate": 2.4187035271687324e-06, "loss": 2.9801, "step": 10213 }, { "epoch": 1.95, "learning_rate": 2.4186463298379407e-06, "loss": 3.6068, "step": 10214 }, { "epoch": 1.95, "learning_rate": 2.41858913250715e-06, "loss": 2.8731, "step": 10215 }, { "epoch": 1.95, "learning_rate": 2.4185319351763583e-06, "loss": 2.6402, "step": 10216 }, { "epoch": 1.95, "learning_rate": 2.418474737845567e-06, "loss": 3.1385, "step": 10217 }, { "epoch": 1.95, "learning_rate": 2.4184175405147763e-06, "loss": 2.5234, "step": 10218 }, { "epoch": 1.95, "learning_rate": 2.4183603431839847e-06, "loss": 2.8224, "step": 10219 }, { "epoch": 1.95, "learning_rate": 2.4183031458531935e-06, "loss": 2.8898, "step": 10220 }, { "epoch": 1.95, "learning_rate": 2.4182459485224023e-06, "loss": 3.3045, "step": 10221 }, { "epoch": 1.95, "learning_rate": 2.418188751191611e-06, "loss": 2.8848, "step": 10222 }, { "epoch": 1.95, "learning_rate": 2.41813155386082e-06, "loss": 3.0271, "step": 10223 }, { "epoch": 1.95, "learning_rate": 2.4180743565300287e-06, "loss": 3.1153, "step": 10224 }, { "epoch": 1.95, "learning_rate": 2.4180171591992375e-06, "loss": 2.5325, "step": 10225 }, { "epoch": 1.95, "learning_rate": 2.417959961868446e-06, "loss": 2.9516, "step": 10226 }, { "epoch": 1.95, "learning_rate": 2.417902764537655e-06, "loss": 3.1117, "step": 10227 }, { "epoch": 1.95, "learning_rate": 2.417845567206864e-06, "loss": 3.2072, "step": 10228 }, { "epoch": 1.95, "learning_rate": 2.4177883698760726e-06, "loss": 3.2427, "step": 10229 }, { "epoch": 1.95, "learning_rate": 2.4177311725452814e-06, "loss": 2.8146, "step": 10230 }, { "epoch": 1.95, "learning_rate": 2.41767397521449e-06, "loss": 2.5891, "step": 10231 }, { "epoch": 1.95, "learning_rate": 2.417616777883699e-06, "loss": 2.8738, "step": 10232 }, { "epoch": 1.95, "learning_rate": 2.4175595805529074e-06, "loss": 3.6541, "step": 10233 }, { "epoch": 1.95, "learning_rate": 2.417502383222116e-06, "loss": 2.9631, "step": 10234 }, { "epoch": 1.95, "learning_rate": 2.4174451858913254e-06, "loss": 2.9389, "step": 10235 }, { "epoch": 1.95, "learning_rate": 2.4173879885605338e-06, "loss": 4.1011, "step": 10236 }, { "epoch": 1.95, "learning_rate": 2.4173307912297426e-06, "loss": 2.7787, "step": 10237 }, { "epoch": 1.95, "learning_rate": 2.4172735938989514e-06, "loss": 2.7682, "step": 10238 }, { "epoch": 1.95, "learning_rate": 2.41721639656816e-06, "loss": 2.7043, "step": 10239 }, { "epoch": 1.95, "learning_rate": 2.417159199237369e-06, "loss": 2.7524, "step": 10240 }, { "epoch": 1.95, "learning_rate": 2.4171020019065777e-06, "loss": 3.4995, "step": 10241 }, { "epoch": 1.95, "learning_rate": 2.4170448045757865e-06, "loss": 2.7973, "step": 10242 }, { "epoch": 1.95, "learning_rate": 2.4169876072449953e-06, "loss": 3.063, "step": 10243 }, { "epoch": 1.95, "learning_rate": 2.416930409914204e-06, "loss": 3.3902, "step": 10244 }, { "epoch": 1.95, "learning_rate": 2.416873212583413e-06, "loss": 2.918, "step": 10245 }, { "epoch": 1.95, "learning_rate": 2.4168160152526217e-06, "loss": 2.5493, "step": 10246 }, { "epoch": 1.95, "learning_rate": 2.4167588179218305e-06, "loss": 2.1819, "step": 10247 }, { "epoch": 1.95, "learning_rate": 2.416701620591039e-06, "loss": 3.2412, "step": 10248 }, { "epoch": 1.95, "learning_rate": 2.416644423260248e-06, "loss": 2.5437, "step": 10249 }, { "epoch": 1.95, "learning_rate": 2.416587225929457e-06, "loss": 3.0085, "step": 10250 }, { "epoch": 1.95, "learning_rate": 2.4165300285986653e-06, "loss": 2.6877, "step": 10251 }, { "epoch": 1.95, "learning_rate": 2.4164728312678745e-06, "loss": 3.2455, "step": 10252 }, { "epoch": 1.95, "learning_rate": 2.416415633937083e-06, "loss": 3.0313, "step": 10253 }, { "epoch": 1.95, "learning_rate": 2.4163584366062916e-06, "loss": 3.2836, "step": 10254 }, { "epoch": 1.95, "learning_rate": 2.4163012392755004e-06, "loss": 3.4167, "step": 10255 }, { "epoch": 1.95, "learning_rate": 2.4162440419447092e-06, "loss": 3.0017, "step": 10256 }, { "epoch": 1.95, "learning_rate": 2.416186844613918e-06, "loss": 2.9996, "step": 10257 }, { "epoch": 1.95, "learning_rate": 2.416129647283127e-06, "loss": 3.259, "step": 10258 }, { "epoch": 1.95, "learning_rate": 2.4160724499523356e-06, "loss": 2.8068, "step": 10259 }, { "epoch": 1.95, "learning_rate": 2.4160152526215444e-06, "loss": 3.6617, "step": 10260 }, { "epoch": 1.95, "learning_rate": 2.415958055290753e-06, "loss": 2.236, "step": 10261 }, { "epoch": 1.95, "learning_rate": 2.415900857959962e-06, "loss": 2.6046, "step": 10262 }, { "epoch": 1.95, "learning_rate": 2.4158436606291708e-06, "loss": 2.5364, "step": 10263 }, { "epoch": 1.96, "learning_rate": 2.4157864632983796e-06, "loss": 2.0918, "step": 10264 }, { "epoch": 1.96, "learning_rate": 2.415729265967588e-06, "loss": 2.5133, "step": 10265 }, { "epoch": 1.96, "learning_rate": 2.415672068636797e-06, "loss": 2.6689, "step": 10266 }, { "epoch": 1.96, "learning_rate": 2.415614871306006e-06, "loss": 3.6349, "step": 10267 }, { "epoch": 1.96, "learning_rate": 2.4155576739752143e-06, "loss": 2.7682, "step": 10268 }, { "epoch": 1.96, "learning_rate": 2.4155004766444235e-06, "loss": 3.7779, "step": 10269 }, { "epoch": 1.96, "learning_rate": 2.415443279313632e-06, "loss": 2.9519, "step": 10270 }, { "epoch": 1.96, "learning_rate": 2.4153860819828407e-06, "loss": 2.9891, "step": 10271 }, { "epoch": 1.96, "learning_rate": 2.4153288846520495e-06, "loss": 2.6528, "step": 10272 }, { "epoch": 1.96, "learning_rate": 2.4152716873212583e-06, "loss": 2.2723, "step": 10273 }, { "epoch": 1.96, "learning_rate": 2.415214489990467e-06, "loss": 2.0588, "step": 10274 }, { "epoch": 1.96, "learning_rate": 2.415157292659676e-06, "loss": 2.8168, "step": 10275 }, { "epoch": 1.96, "learning_rate": 2.4151000953288847e-06, "loss": 3.396, "step": 10276 }, { "epoch": 1.96, "learning_rate": 2.4150428979980935e-06, "loss": 3.7617, "step": 10277 }, { "epoch": 1.96, "learning_rate": 2.4149857006673023e-06, "loss": 3.7065, "step": 10278 }, { "epoch": 1.96, "learning_rate": 2.414928503336511e-06, "loss": 2.6081, "step": 10279 }, { "epoch": 1.96, "learning_rate": 2.41487130600572e-06, "loss": 2.838, "step": 10280 }, { "epoch": 1.96, "learning_rate": 2.4148141086749286e-06, "loss": 2.676, "step": 10281 }, { "epoch": 1.96, "learning_rate": 2.4147569113441374e-06, "loss": 3.6431, "step": 10282 }, { "epoch": 1.96, "learning_rate": 2.4146997140133462e-06, "loss": 3.4558, "step": 10283 }, { "epoch": 1.96, "learning_rate": 2.414642516682555e-06, "loss": 3.0537, "step": 10284 }, { "epoch": 1.96, "learning_rate": 2.4145853193517634e-06, "loss": 2.7893, "step": 10285 }, { "epoch": 1.96, "learning_rate": 2.4145281220209726e-06, "loss": 2.4293, "step": 10286 }, { "epoch": 1.96, "learning_rate": 2.414470924690181e-06, "loss": 3.587, "step": 10287 }, { "epoch": 1.96, "learning_rate": 2.4144137273593898e-06, "loss": 3.3093, "step": 10288 }, { "epoch": 1.96, "learning_rate": 2.414356530028599e-06, "loss": 3.1239, "step": 10289 }, { "epoch": 1.96, "learning_rate": 2.4142993326978074e-06, "loss": 3.1198, "step": 10290 }, { "epoch": 1.96, "learning_rate": 2.414242135367016e-06, "loss": 2.7628, "step": 10291 }, { "epoch": 1.96, "learning_rate": 2.414184938036225e-06, "loss": 2.496, "step": 10292 }, { "epoch": 1.96, "learning_rate": 2.4141277407054337e-06, "loss": 2.8213, "step": 10293 }, { "epoch": 1.96, "learning_rate": 2.4140705433746425e-06, "loss": 3.6011, "step": 10294 }, { "epoch": 1.96, "learning_rate": 2.4140133460438513e-06, "loss": 2.6054, "step": 10295 }, { "epoch": 1.96, "learning_rate": 2.41395614871306e-06, "loss": 2.7818, "step": 10296 }, { "epoch": 1.96, "learning_rate": 2.413898951382269e-06, "loss": 2.6798, "step": 10297 }, { "epoch": 1.96, "learning_rate": 2.4138417540514777e-06, "loss": 2.2832, "step": 10298 }, { "epoch": 1.96, "learning_rate": 2.4137845567206865e-06, "loss": 3.2526, "step": 10299 }, { "epoch": 1.96, "learning_rate": 2.4137273593898953e-06, "loss": 2.4748, "step": 10300 }, { "epoch": 1.96, "learning_rate": 2.413670162059104e-06, "loss": 3.3746, "step": 10301 }, { "epoch": 1.96, "learning_rate": 2.4136129647283125e-06, "loss": 1.9151, "step": 10302 }, { "epoch": 1.96, "learning_rate": 2.4135557673975217e-06, "loss": 2.583, "step": 10303 }, { "epoch": 1.96, "learning_rate": 2.41349857006673e-06, "loss": 3.4593, "step": 10304 }, { "epoch": 1.96, "learning_rate": 2.413441372735939e-06, "loss": 3.3773, "step": 10305 }, { "epoch": 1.96, "learning_rate": 2.413384175405148e-06, "loss": 2.606, "step": 10306 }, { "epoch": 1.96, "learning_rate": 2.4133269780743564e-06, "loss": 2.7587, "step": 10307 }, { "epoch": 1.96, "learning_rate": 2.4132697807435657e-06, "loss": 2.8402, "step": 10308 }, { "epoch": 1.96, "learning_rate": 2.413212583412774e-06, "loss": 2.0247, "step": 10309 }, { "epoch": 1.96, "learning_rate": 2.413155386081983e-06, "loss": 3.5052, "step": 10310 }, { "epoch": 1.96, "learning_rate": 2.4130981887511916e-06, "loss": 2.9331, "step": 10311 }, { "epoch": 1.96, "learning_rate": 2.4130409914204004e-06, "loss": 2.2956, "step": 10312 }, { "epoch": 1.96, "learning_rate": 2.412983794089609e-06, "loss": 3.1599, "step": 10313 }, { "epoch": 1.96, "learning_rate": 2.412926596758818e-06, "loss": 2.5343, "step": 10314 }, { "epoch": 1.96, "learning_rate": 2.4128693994280268e-06, "loss": 2.5814, "step": 10315 }, { "epoch": 1.96, "learning_rate": 2.4128122020972356e-06, "loss": 3.2442, "step": 10316 }, { "epoch": 1.97, "learning_rate": 2.4127550047664444e-06, "loss": 2.6856, "step": 10317 }, { "epoch": 1.97, "learning_rate": 2.412697807435653e-06, "loss": 2.3234, "step": 10318 }, { "epoch": 1.97, "learning_rate": 2.4126406101048615e-06, "loss": 2.9473, "step": 10319 }, { "epoch": 1.97, "learning_rate": 2.4125834127740708e-06, "loss": 2.9139, "step": 10320 }, { "epoch": 1.97, "learning_rate": 2.412526215443279e-06, "loss": 2.6895, "step": 10321 }, { "epoch": 1.97, "learning_rate": 2.412469018112488e-06, "loss": 2.7153, "step": 10322 }, { "epoch": 1.97, "learning_rate": 2.412411820781697e-06, "loss": 3.208, "step": 10323 }, { "epoch": 1.97, "learning_rate": 2.4123546234509055e-06, "loss": 2.745, "step": 10324 }, { "epoch": 1.97, "learning_rate": 2.4122974261201147e-06, "loss": 3.3919, "step": 10325 }, { "epoch": 1.97, "learning_rate": 2.412240228789323e-06, "loss": 2.8775, "step": 10326 }, { "epoch": 1.97, "learning_rate": 2.412183031458532e-06, "loss": 3.2812, "step": 10327 }, { "epoch": 1.97, "learning_rate": 2.412125834127741e-06, "loss": 3.7929, "step": 10328 }, { "epoch": 1.97, "learning_rate": 2.4120686367969495e-06, "loss": 2.5245, "step": 10329 }, { "epoch": 1.97, "learning_rate": 2.4120114394661583e-06, "loss": 2.4742, "step": 10330 }, { "epoch": 1.97, "learning_rate": 2.411954242135367e-06, "loss": 3.1795, "step": 10331 }, { "epoch": 1.97, "learning_rate": 2.411897044804576e-06, "loss": 2.8101, "step": 10332 }, { "epoch": 1.97, "learning_rate": 2.4118398474737847e-06, "loss": 2.9474, "step": 10333 }, { "epoch": 1.97, "learning_rate": 2.4117826501429934e-06, "loss": 3.7136, "step": 10334 }, { "epoch": 1.97, "learning_rate": 2.4117254528122022e-06, "loss": 3.3616, "step": 10335 }, { "epoch": 1.97, "learning_rate": 2.4116682554814106e-06, "loss": 3.3711, "step": 10336 }, { "epoch": 1.97, "learning_rate": 2.41161105815062e-06, "loss": 2.6369, "step": 10337 }, { "epoch": 1.97, "learning_rate": 2.4115538608198286e-06, "loss": 2.5256, "step": 10338 }, { "epoch": 1.97, "learning_rate": 2.411496663489037e-06, "loss": 2.233, "step": 10339 }, { "epoch": 1.97, "learning_rate": 2.411439466158246e-06, "loss": 3.4936, "step": 10340 }, { "epoch": 1.97, "learning_rate": 2.4113822688274546e-06, "loss": 3.3596, "step": 10341 }, { "epoch": 1.97, "learning_rate": 2.411325071496664e-06, "loss": 1.8368, "step": 10342 }, { "epoch": 1.97, "learning_rate": 2.411267874165872e-06, "loss": 2.2576, "step": 10343 }, { "epoch": 1.97, "learning_rate": 2.411210676835081e-06, "loss": 3.0326, "step": 10344 }, { "epoch": 1.97, "learning_rate": 2.41115347950429e-06, "loss": 2.9636, "step": 10345 }, { "epoch": 1.97, "learning_rate": 2.4110962821734985e-06, "loss": 2.8568, "step": 10346 }, { "epoch": 1.97, "learning_rate": 2.4110390848427073e-06, "loss": 3.7622, "step": 10347 }, { "epoch": 1.97, "learning_rate": 2.410981887511916e-06, "loss": 1.8842, "step": 10348 }, { "epoch": 1.97, "learning_rate": 2.410924690181125e-06, "loss": 2.4472, "step": 10349 }, { "epoch": 1.97, "learning_rate": 2.4108674928503337e-06, "loss": 3.0801, "step": 10350 }, { "epoch": 1.97, "learning_rate": 2.4108102955195425e-06, "loss": 3.2238, "step": 10351 }, { "epoch": 1.97, "learning_rate": 2.4107530981887513e-06, "loss": 3.6012, "step": 10352 }, { "epoch": 1.97, "learning_rate": 2.4106959008579597e-06, "loss": 3.2898, "step": 10353 }, { "epoch": 1.97, "learning_rate": 2.410638703527169e-06, "loss": 2.4685, "step": 10354 }, { "epoch": 1.97, "learning_rate": 2.4105815061963777e-06, "loss": 3.4417, "step": 10355 }, { "epoch": 1.97, "learning_rate": 2.410524308865586e-06, "loss": 3.0966, "step": 10356 }, { "epoch": 1.97, "learning_rate": 2.4104671115347953e-06, "loss": 3.6255, "step": 10357 }, { "epoch": 1.97, "learning_rate": 2.4104099142040037e-06, "loss": 3.7082, "step": 10358 }, { "epoch": 1.97, "learning_rate": 2.410352716873213e-06, "loss": 2.4308, "step": 10359 }, { "epoch": 1.97, "learning_rate": 2.4102955195424212e-06, "loss": 3.4029, "step": 10360 }, { "epoch": 1.97, "learning_rate": 2.41023832221163e-06, "loss": 2.7135, "step": 10361 }, { "epoch": 1.97, "learning_rate": 2.4101811248808393e-06, "loss": 2.9597, "step": 10362 }, { "epoch": 1.97, "learning_rate": 2.4101239275500476e-06, "loss": 3.234, "step": 10363 }, { "epoch": 1.97, "learning_rate": 2.4100667302192564e-06, "loss": 2.9063, "step": 10364 }, { "epoch": 1.97, "learning_rate": 2.410009532888465e-06, "loss": 3.1457, "step": 10365 }, { "epoch": 1.97, "learning_rate": 2.409952335557674e-06, "loss": 3.2429, "step": 10366 }, { "epoch": 1.97, "learning_rate": 2.409895138226883e-06, "loss": 2.6307, "step": 10367 }, { "epoch": 1.97, "learning_rate": 2.4098379408960916e-06, "loss": 2.9832, "step": 10368 }, { "epoch": 1.98, "learning_rate": 2.4097807435653004e-06, "loss": 3.2223, "step": 10369 }, { "epoch": 1.98, "learning_rate": 2.409723546234509e-06, "loss": 2.04, "step": 10370 }, { "epoch": 1.98, "learning_rate": 2.409666348903718e-06, "loss": 2.3951, "step": 10371 }, { "epoch": 1.98, "learning_rate": 2.4096091515729268e-06, "loss": 1.9835, "step": 10372 }, { "epoch": 1.98, "learning_rate": 2.4095519542421356e-06, "loss": 2.3333, "step": 10373 }, { "epoch": 1.98, "learning_rate": 2.4094947569113444e-06, "loss": 3.4165, "step": 10374 }, { "epoch": 1.98, "learning_rate": 2.4094375595805527e-06, "loss": 3.011, "step": 10375 }, { "epoch": 1.98, "learning_rate": 2.409380362249762e-06, "loss": 2.6408, "step": 10376 }, { "epoch": 1.98, "learning_rate": 2.4093231649189707e-06, "loss": 3.6291, "step": 10377 }, { "epoch": 1.98, "learning_rate": 2.409265967588179e-06, "loss": 2.8577, "step": 10378 }, { "epoch": 1.98, "learning_rate": 2.4092087702573883e-06, "loss": 2.0764, "step": 10379 }, { "epoch": 1.98, "learning_rate": 2.4091515729265967e-06, "loss": 3.5699, "step": 10380 }, { "epoch": 1.98, "learning_rate": 2.4090943755958055e-06, "loss": 2.1515, "step": 10381 }, { "epoch": 1.98, "learning_rate": 2.4090371782650143e-06, "loss": 2.8147, "step": 10382 }, { "epoch": 1.98, "learning_rate": 2.408979980934223e-06, "loss": 3.6834, "step": 10383 }, { "epoch": 1.98, "learning_rate": 2.408922783603432e-06, "loss": 2.8629, "step": 10384 }, { "epoch": 1.98, "learning_rate": 2.4088655862726407e-06, "loss": 2.5427, "step": 10385 }, { "epoch": 1.98, "learning_rate": 2.4088083889418495e-06, "loss": 1.8204, "step": 10386 }, { "epoch": 1.98, "learning_rate": 2.4087511916110582e-06, "loss": 2.4915, "step": 10387 }, { "epoch": 1.98, "learning_rate": 2.408693994280267e-06, "loss": 2.5023, "step": 10388 }, { "epoch": 1.98, "learning_rate": 2.408636796949476e-06, "loss": 3.401, "step": 10389 }, { "epoch": 1.98, "learning_rate": 2.4085795996186846e-06, "loss": 3.736, "step": 10390 }, { "epoch": 1.98, "learning_rate": 2.4085224022878934e-06, "loss": 3.213, "step": 10391 }, { "epoch": 1.98, "learning_rate": 2.408465204957102e-06, "loss": 3.4785, "step": 10392 }, { "epoch": 1.98, "learning_rate": 2.408408007626311e-06, "loss": 2.5031, "step": 10393 }, { "epoch": 1.98, "learning_rate": 2.40835081029552e-06, "loss": 3.1831, "step": 10394 }, { "epoch": 1.98, "learning_rate": 2.408293612964728e-06, "loss": 2.7267, "step": 10395 }, { "epoch": 1.98, "learning_rate": 2.4082364156339374e-06, "loss": 2.672, "step": 10396 }, { "epoch": 1.98, "learning_rate": 2.4081792183031458e-06, "loss": 2.6727, "step": 10397 }, { "epoch": 1.98, "learning_rate": 2.4081220209723546e-06, "loss": 2.1254, "step": 10398 }, { "epoch": 1.98, "learning_rate": 2.4080648236415634e-06, "loss": 2.7506, "step": 10399 }, { "epoch": 1.98, "learning_rate": 2.408007626310772e-06, "loss": 2.7835, "step": 10400 }, { "epoch": 1.98, "learning_rate": 2.407950428979981e-06, "loss": 2.1297, "step": 10401 }, { "epoch": 1.98, "learning_rate": 2.4078932316491897e-06, "loss": 3.0263, "step": 10402 }, { "epoch": 1.98, "learning_rate": 2.4078360343183985e-06, "loss": 1.8654, "step": 10403 }, { "epoch": 1.98, "learning_rate": 2.4077788369876073e-06, "loss": 3.3459, "step": 10404 }, { "epoch": 1.98, "learning_rate": 2.407721639656816e-06, "loss": 3.4605, "step": 10405 }, { "epoch": 1.98, "learning_rate": 2.407664442326025e-06, "loss": 2.8954, "step": 10406 }, { "epoch": 1.98, "learning_rate": 2.4076072449952337e-06, "loss": 3.0792, "step": 10407 }, { "epoch": 1.98, "learning_rate": 2.4075500476644425e-06, "loss": 3.3501, "step": 10408 }, { "epoch": 1.98, "learning_rate": 2.407492850333651e-06, "loss": 2.8209, "step": 10409 }, { "epoch": 1.98, "learning_rate": 2.40743565300286e-06, "loss": 2.8108, "step": 10410 }, { "epoch": 1.98, "learning_rate": 2.407378455672069e-06, "loss": 4.0329, "step": 10411 }, { "epoch": 1.98, "learning_rate": 2.4073212583412772e-06, "loss": 2.5012, "step": 10412 }, { "epoch": 1.98, "learning_rate": 2.4072640610104865e-06, "loss": 2.975, "step": 10413 }, { "epoch": 1.98, "learning_rate": 2.407206863679695e-06, "loss": 2.7706, "step": 10414 }, { "epoch": 1.98, "learning_rate": 2.4071496663489036e-06, "loss": 2.2088, "step": 10415 }, { "epoch": 1.98, "learning_rate": 2.407092469018113e-06, "loss": 2.4564, "step": 10416 }, { "epoch": 1.98, "learning_rate": 2.4070352716873212e-06, "loss": 3.1393, "step": 10417 }, { "epoch": 1.98, "learning_rate": 2.40697807435653e-06, "loss": 2.5596, "step": 10418 }, { "epoch": 1.98, "learning_rate": 2.406920877025739e-06, "loss": 2.5384, "step": 10419 }, { "epoch": 1.98, "learning_rate": 2.4068636796949476e-06, "loss": 3.0951, "step": 10420 }, { "epoch": 1.98, "learning_rate": 2.4068064823641564e-06, "loss": 2.955, "step": 10421 }, { "epoch": 1.99, "learning_rate": 2.406749285033365e-06, "loss": 3.1482, "step": 10422 }, { "epoch": 1.99, "learning_rate": 2.406692087702574e-06, "loss": 3.1156, "step": 10423 }, { "epoch": 1.99, "learning_rate": 2.4066348903717828e-06, "loss": 2.4065, "step": 10424 }, { "epoch": 1.99, "learning_rate": 2.4065776930409916e-06, "loss": 2.2313, "step": 10425 }, { "epoch": 1.99, "learning_rate": 2.4065204957102004e-06, "loss": 2.1419, "step": 10426 }, { "epoch": 1.99, "learning_rate": 2.406463298379409e-06, "loss": 2.8667, "step": 10427 }, { "epoch": 1.99, "learning_rate": 2.406406101048618e-06, "loss": 2.3293, "step": 10428 }, { "epoch": 1.99, "learning_rate": 2.4063489037178263e-06, "loss": 3.6001, "step": 10429 }, { "epoch": 1.99, "learning_rate": 2.4062917063870355e-06, "loss": 3.1066, "step": 10430 }, { "epoch": 1.99, "learning_rate": 2.406234509056244e-06, "loss": 3.1755, "step": 10431 }, { "epoch": 1.99, "learning_rate": 2.4061773117254527e-06, "loss": 2.9374, "step": 10432 }, { "epoch": 1.99, "learning_rate": 2.406120114394662e-06, "loss": 3.0584, "step": 10433 }, { "epoch": 1.99, "learning_rate": 2.4060629170638703e-06, "loss": 3.2212, "step": 10434 }, { "epoch": 1.99, "learning_rate": 2.406005719733079e-06, "loss": 2.8258, "step": 10435 }, { "epoch": 1.99, "learning_rate": 2.405948522402288e-06, "loss": 2.325, "step": 10436 }, { "epoch": 1.99, "learning_rate": 2.4058913250714967e-06, "loss": 4.0317, "step": 10437 }, { "epoch": 1.99, "learning_rate": 2.4058341277407055e-06, "loss": 3.5556, "step": 10438 }, { "epoch": 1.99, "learning_rate": 2.4057769304099143e-06, "loss": 2.8728, "step": 10439 }, { "epoch": 1.99, "learning_rate": 2.405719733079123e-06, "loss": 3.7857, "step": 10440 }, { "epoch": 1.99, "learning_rate": 2.405662535748332e-06, "loss": 2.9007, "step": 10441 }, { "epoch": 1.99, "learning_rate": 2.4056053384175406e-06, "loss": 2.8872, "step": 10442 }, { "epoch": 1.99, "learning_rate": 2.4055481410867494e-06, "loss": 3.1921, "step": 10443 }, { "epoch": 1.99, "learning_rate": 2.4054909437559582e-06, "loss": 2.0391, "step": 10444 }, { "epoch": 1.99, "learning_rate": 2.405433746425167e-06, "loss": 2.8493, "step": 10445 }, { "epoch": 1.99, "learning_rate": 2.4053765490943754e-06, "loss": 2.7632, "step": 10446 }, { "epoch": 1.99, "learning_rate": 2.4053193517635846e-06, "loss": 3.7175, "step": 10447 }, { "epoch": 1.99, "learning_rate": 2.405262154432793e-06, "loss": 2.8158, "step": 10448 }, { "epoch": 1.99, "learning_rate": 2.4052049571020018e-06, "loss": 3.4315, "step": 10449 }, { "epoch": 1.99, "learning_rate": 2.405147759771211e-06, "loss": 3.3129, "step": 10450 }, { "epoch": 1.99, "learning_rate": 2.4050905624404194e-06, "loss": 2.2953, "step": 10451 }, { "epoch": 1.99, "learning_rate": 2.405033365109628e-06, "loss": 2.7659, "step": 10452 }, { "epoch": 1.99, "learning_rate": 2.404976167778837e-06, "loss": 3.0417, "step": 10453 }, { "epoch": 1.99, "learning_rate": 2.4049189704480457e-06, "loss": 3.3338, "step": 10454 }, { "epoch": 1.99, "learning_rate": 2.404861773117255e-06, "loss": 3.3992, "step": 10455 }, { "epoch": 1.99, "learning_rate": 2.4048045757864633e-06, "loss": 2.6637, "step": 10456 }, { "epoch": 1.99, "learning_rate": 2.404747378455672e-06, "loss": 3.0896, "step": 10457 }, { "epoch": 1.99, "learning_rate": 2.404690181124881e-06, "loss": 2.155, "step": 10458 }, { "epoch": 1.99, "learning_rate": 2.4046329837940897e-06, "loss": 2.6094, "step": 10459 }, { "epoch": 1.99, "learning_rate": 2.4045757864632985e-06, "loss": 3.2536, "step": 10460 }, { "epoch": 1.99, "learning_rate": 2.4045185891325073e-06, "loss": 2.4072, "step": 10461 }, { "epoch": 1.99, "learning_rate": 2.404461391801716e-06, "loss": 3.0383, "step": 10462 }, { "epoch": 1.99, "learning_rate": 2.4044041944709245e-06, "loss": 2.2625, "step": 10463 }, { "epoch": 1.99, "learning_rate": 2.4043469971401337e-06, "loss": 2.599, "step": 10464 }, { "epoch": 1.99, "learning_rate": 2.4042897998093425e-06, "loss": 3.7464, "step": 10465 }, { "epoch": 1.99, "learning_rate": 2.404232602478551e-06, "loss": 3.3822, "step": 10466 }, { "epoch": 1.99, "learning_rate": 2.40417540514776e-06, "loss": 2.9673, "step": 10467 }, { "epoch": 1.99, "learning_rate": 2.4041182078169684e-06, "loss": 3.1998, "step": 10468 }, { "epoch": 1.99, "learning_rate": 2.4040610104861772e-06, "loss": 2.6692, "step": 10469 }, { "epoch": 1.99, "learning_rate": 2.404003813155386e-06, "loss": 2.6637, "step": 10470 }, { "epoch": 1.99, "learning_rate": 2.403946615824595e-06, "loss": 2.2605, "step": 10471 }, { "epoch": 1.99, "learning_rate": 2.403889418493804e-06, "loss": 3.045, "step": 10472 }, { "epoch": 1.99, "learning_rate": 2.4038322211630124e-06, "loss": 2.566, "step": 10473 }, { "epoch": 2.0, "learning_rate": 2.403775023832221e-06, "loss": 2.8649, "step": 10474 }, { "epoch": 2.0, "learning_rate": 2.40371782650143e-06, "loss": 2.6556, "step": 10475 }, { "epoch": 2.0, "learning_rate": 2.4036606291706388e-06, "loss": 3.2265, "step": 10476 }, { "epoch": 2.0, "learning_rate": 2.4036034318398476e-06, "loss": 3.3872, "step": 10477 }, { "epoch": 2.0, "learning_rate": 2.4035462345090564e-06, "loss": 2.4623, "step": 10478 }, { "epoch": 2.0, "learning_rate": 2.403489037178265e-06, "loss": 2.9766, "step": 10479 }, { "epoch": 2.0, "learning_rate": 2.4034318398474735e-06, "loss": 2.8207, "step": 10480 }, { "epoch": 2.0, "learning_rate": 2.4033746425166828e-06, "loss": 1.9185, "step": 10481 }, { "epoch": 2.0, "learning_rate": 2.4033174451858915e-06, "loss": 2.7095, "step": 10482 }, { "epoch": 2.0, "learning_rate": 2.4032602478551e-06, "loss": 2.8768, "step": 10483 }, { "epoch": 2.0, "learning_rate": 2.403203050524309e-06, "loss": 3.1553, "step": 10484 }, { "epoch": 2.0, "learning_rate": 2.4031458531935175e-06, "loss": 2.8076, "step": 10485 }, { "epoch": 2.0, "learning_rate": 2.4030886558627263e-06, "loss": 3.3604, "step": 10486 }, { "epoch": 2.0, "learning_rate": 2.403031458531935e-06, "loss": 2.63, "step": 10487 }, { "epoch": 2.0, "learning_rate": 2.402974261201144e-06, "loss": 2.6733, "step": 10488 }, { "epoch": 2.0, "learning_rate": 2.402917063870353e-06, "loss": 2.4482, "step": 10489 }, { "epoch": 2.0, "learning_rate": 2.4028598665395615e-06, "loss": 2.4523, "step": 10490 }, { "epoch": 2.0, "learning_rate": 2.4028026692087703e-06, "loss": 2.7376, "step": 10491 }, { "epoch": 2.0, "learning_rate": 2.402745471877979e-06, "loss": 3.105, "step": 10492 }, { "epoch": 2.0, "learning_rate": 2.402688274547188e-06, "loss": 2.2366, "step": 10493 }, { "epoch": 2.0, "learning_rate": 2.4026310772163966e-06, "loss": 3.023, "step": 10494 }, { "epoch": 2.0, "learning_rate": 2.4025738798856054e-06, "loss": 3.3286, "step": 10495 }, { "epoch": 2.0, "learning_rate": 2.4025166825548142e-06, "loss": 3.4931, "step": 10496 }, { "epoch": 2.0, "learning_rate": 2.4024594852240226e-06, "loss": 3.2534, "step": 10497 }, { "epoch": 2.0, "learning_rate": 2.402402287893232e-06, "loss": 2.5585, "step": 10498 }, { "epoch": 2.0, "learning_rate": 2.4023450905624406e-06, "loss": 2.4096, "step": 10499 }, { "epoch": 2.0, "learning_rate": 2.402287893231649e-06, "loss": 3.9969, "step": 10500 }, { "epoch": 2.0, "eval_accuracy": 0.2944706386626661, "eval_loss": 2.795963764190674, "eval_runtime": 126.4489, "eval_samples_per_second": 18.45, "eval_steps_per_second": 4.618, "step": 10500 }, { "epoch": 2.0, "learning_rate": 2.402230695900858e-06, "loss": 2.843, "step": 10501 }, { "epoch": 2.0, "learning_rate": 2.4021734985700666e-06, "loss": 2.8955, "step": 10502 }, { "epoch": 2.0, "learning_rate": 2.402116301239276e-06, "loss": 3.0997, "step": 10503 }, { "epoch": 2.0, "learning_rate": 2.4020591039084846e-06, "loss": 2.9831, "step": 10504 }, { "epoch": 2.0, "learning_rate": 2.402001906577693e-06, "loss": 3.1094, "step": 10505 }, { "epoch": 2.0, "learning_rate": 2.401944709246902e-06, "loss": 2.805, "step": 10506 }, { "epoch": 2.0, "learning_rate": 2.4018875119161105e-06, "loss": 3.1908, "step": 10507 }, { "epoch": 2.0, "learning_rate": 2.4018303145853193e-06, "loss": 1.9827, "step": 10508 }, { "epoch": 2.0, "learning_rate": 2.401773117254528e-06, "loss": 2.6243, "step": 10509 }, { "epoch": 2.0, "learning_rate": 2.401715919923737e-06, "loss": 2.4575, "step": 10510 }, { "epoch": 2.0, "learning_rate": 2.4016587225929457e-06, "loss": 2.8174, "step": 10511 }, { "epoch": 2.0, "learning_rate": 2.4016015252621545e-06, "loss": 3.7872, "step": 10512 }, { "epoch": 2.0, "learning_rate": 2.4015443279313633e-06, "loss": 2.9226, "step": 10513 }, { "epoch": 2.0, "learning_rate": 2.401487130600572e-06, "loss": 3.6303, "step": 10514 }, { "epoch": 2.0, "learning_rate": 2.401429933269781e-06, "loss": 2.4059, "step": 10515 }, { "epoch": 2.0, "learning_rate": 2.4013727359389897e-06, "loss": 2.2161, "step": 10516 }, { "epoch": 2.0, "learning_rate": 2.401315538608198e-06, "loss": 2.8251, "step": 10517 }, { "epoch": 2.0, "learning_rate": 2.4012583412774073e-06, "loss": 2.1327, "step": 10518 }, { "epoch": 2.0, "learning_rate": 2.4012011439466156e-06, "loss": 1.5835, "step": 10519 }, { "epoch": 2.0, "learning_rate": 2.401143946615825e-06, "loss": 2.6931, "step": 10520 }, { "epoch": 2.0, "learning_rate": 2.4010867492850337e-06, "loss": 3.6382, "step": 10521 }, { "epoch": 2.0, "learning_rate": 2.401029551954242e-06, "loss": 2.8016, "step": 10522 }, { "epoch": 2.0, "learning_rate": 2.4009723546234512e-06, "loss": 2.6109, "step": 10523 }, { "epoch": 2.0, "learning_rate": 2.4009151572926596e-06, "loss": 3.8099, "step": 10524 }, { "epoch": 2.0, "learning_rate": 2.4008579599618684e-06, "loss": 2.8866, "step": 10525 }, { "epoch": 2.0, "learning_rate": 2.400800762631077e-06, "loss": 2.6021, "step": 10526 }, { "epoch": 2.01, "learning_rate": 2.400743565300286e-06, "loss": 2.3466, "step": 10527 }, { "epoch": 2.01, "learning_rate": 2.400686367969495e-06, "loss": 2.573, "step": 10528 }, { "epoch": 2.01, "learning_rate": 2.4006291706387036e-06, "loss": 2.1045, "step": 10529 }, { "epoch": 2.01, "learning_rate": 2.4005719733079124e-06, "loss": 3.1737, "step": 10530 }, { "epoch": 2.01, "learning_rate": 2.400514775977121e-06, "loss": 2.9688, "step": 10531 }, { "epoch": 2.01, "learning_rate": 2.40045757864633e-06, "loss": 3.1548, "step": 10532 }, { "epoch": 2.01, "learning_rate": 2.4004003813155388e-06, "loss": 3.1215, "step": 10533 }, { "epoch": 2.01, "learning_rate": 2.400343183984747e-06, "loss": 3.0923, "step": 10534 }, { "epoch": 2.01, "learning_rate": 2.4002859866539563e-06, "loss": 3.2422, "step": 10535 }, { "epoch": 2.01, "learning_rate": 2.4002287893231647e-06, "loss": 2.4817, "step": 10536 }, { "epoch": 2.01, "learning_rate": 2.400171591992374e-06, "loss": 3.3299, "step": 10537 }, { "epoch": 2.01, "learning_rate": 2.4001143946615827e-06, "loss": 2.7129, "step": 10538 }, { "epoch": 2.01, "learning_rate": 2.400057197330791e-06, "loss": 1.9881, "step": 10539 }, { "epoch": 2.01, "learning_rate": 2.4000000000000003e-06, "loss": 3.1342, "step": 10540 }, { "epoch": 2.01, "learning_rate": 2.3999428026692087e-06, "loss": 2.8941, "step": 10541 }, { "epoch": 2.01, "learning_rate": 2.3998856053384175e-06, "loss": 3.2016, "step": 10542 }, { "epoch": 2.01, "learning_rate": 2.3998284080076263e-06, "loss": 3.2521, "step": 10543 }, { "epoch": 2.01, "learning_rate": 2.399771210676835e-06, "loss": 2.9401, "step": 10544 }, { "epoch": 2.01, "learning_rate": 2.399714013346044e-06, "loss": 2.3983, "step": 10545 }, { "epoch": 2.01, "learning_rate": 2.3996568160152527e-06, "loss": 3.4846, "step": 10546 }, { "epoch": 2.01, "learning_rate": 2.3995996186844614e-06, "loss": 2.9054, "step": 10547 }, { "epoch": 2.01, "learning_rate": 2.3995424213536702e-06, "loss": 2.7527, "step": 10548 }, { "epoch": 2.01, "learning_rate": 2.399485224022879e-06, "loss": 2.7973, "step": 10549 }, { "epoch": 2.01, "learning_rate": 2.399428026692088e-06, "loss": 3.8201, "step": 10550 }, { "epoch": 2.01, "learning_rate": 2.399370829361296e-06, "loss": 2.5124, "step": 10551 }, { "epoch": 2.01, "learning_rate": 2.3993136320305054e-06, "loss": 2.5751, "step": 10552 }, { "epoch": 2.01, "learning_rate": 2.399256434699714e-06, "loss": 2.6721, "step": 10553 }, { "epoch": 2.01, "learning_rate": 2.399199237368923e-06, "loss": 3.5927, "step": 10554 }, { "epoch": 2.01, "learning_rate": 2.399142040038132e-06, "loss": 3.4426, "step": 10555 }, { "epoch": 2.01, "learning_rate": 2.39908484270734e-06, "loss": 3.7424, "step": 10556 }, { "epoch": 2.01, "learning_rate": 2.3990276453765494e-06, "loss": 2.8306, "step": 10557 }, { "epoch": 2.01, "learning_rate": 2.3989704480457578e-06, "loss": 2.1999, "step": 10558 }, { "epoch": 2.01, "learning_rate": 2.3989132507149666e-06, "loss": 2.7805, "step": 10559 }, { "epoch": 2.01, "learning_rate": 2.3988560533841758e-06, "loss": 3.505, "step": 10560 }, { "epoch": 2.01, "learning_rate": 2.398798856053384e-06, "loss": 3.7002, "step": 10561 }, { "epoch": 2.01, "learning_rate": 2.398741658722593e-06, "loss": 3.1706, "step": 10562 }, { "epoch": 2.01, "learning_rate": 2.3986844613918017e-06, "loss": 3.2759, "step": 10563 }, { "epoch": 2.01, "learning_rate": 2.3986272640610105e-06, "loss": 3.1532, "step": 10564 }, { "epoch": 2.01, "learning_rate": 2.3985700667302193e-06, "loss": 2.7755, "step": 10565 }, { "epoch": 2.01, "learning_rate": 2.398512869399428e-06, "loss": 3.0276, "step": 10566 }, { "epoch": 2.01, "learning_rate": 2.398455672068637e-06, "loss": 2.1768, "step": 10567 }, { "epoch": 2.01, "learning_rate": 2.3983984747378457e-06, "loss": 2.0455, "step": 10568 }, { "epoch": 2.01, "learning_rate": 2.3983412774070545e-06, "loss": 3.055, "step": 10569 }, { "epoch": 2.01, "learning_rate": 2.3982840800762633e-06, "loss": 3.297, "step": 10570 }, { "epoch": 2.01, "learning_rate": 2.398226882745472e-06, "loss": 2.1477, "step": 10571 }, { "epoch": 2.01, "learning_rate": 2.398169685414681e-06, "loss": 2.4235, "step": 10572 }, { "epoch": 2.01, "learning_rate": 2.3981124880838892e-06, "loss": 3.5754, "step": 10573 }, { "epoch": 2.01, "learning_rate": 2.3980552907530985e-06, "loss": 3.1278, "step": 10574 }, { "epoch": 2.01, "learning_rate": 2.397998093422307e-06, "loss": 3.1446, "step": 10575 }, { "epoch": 2.01, "learning_rate": 2.3979408960915156e-06, "loss": 3.5617, "step": 10576 }, { "epoch": 2.01, "learning_rate": 2.397883698760725e-06, "loss": 3.4749, "step": 10577 }, { "epoch": 2.01, "learning_rate": 2.397826501429933e-06, "loss": 3.429, "step": 10578 }, { "epoch": 2.02, "learning_rate": 2.397769304099142e-06, "loss": 1.6499, "step": 10579 }, { "epoch": 2.02, "learning_rate": 2.397712106768351e-06, "loss": 3.8371, "step": 10580 }, { "epoch": 2.02, "learning_rate": 2.3976549094375596e-06, "loss": 2.439, "step": 10581 }, { "epoch": 2.02, "learning_rate": 2.3975977121067684e-06, "loss": 3.1679, "step": 10582 }, { "epoch": 2.02, "learning_rate": 2.397540514775977e-06, "loss": 2.9923, "step": 10583 }, { "epoch": 2.02, "learning_rate": 2.397483317445186e-06, "loss": 2.9816, "step": 10584 }, { "epoch": 2.02, "learning_rate": 2.3974261201143948e-06, "loss": 2.6316, "step": 10585 }, { "epoch": 2.02, "learning_rate": 2.3973689227836036e-06, "loss": 2.2871, "step": 10586 }, { "epoch": 2.02, "learning_rate": 2.3973117254528124e-06, "loss": 3.5458, "step": 10587 }, { "epoch": 2.02, "learning_rate": 2.397254528122021e-06, "loss": 3.514, "step": 10588 }, { "epoch": 2.02, "learning_rate": 2.39719733079123e-06, "loss": 4.0567, "step": 10589 }, { "epoch": 2.02, "learning_rate": 2.3971401334604383e-06, "loss": 2.2347, "step": 10590 }, { "epoch": 2.02, "learning_rate": 2.3970829361296475e-06, "loss": 2.2847, "step": 10591 }, { "epoch": 2.02, "learning_rate": 2.3970257387988563e-06, "loss": 2.0206, "step": 10592 }, { "epoch": 2.02, "learning_rate": 2.3969685414680647e-06, "loss": 3.4714, "step": 10593 }, { "epoch": 2.02, "learning_rate": 2.396911344137274e-06, "loss": 3.0443, "step": 10594 }, { "epoch": 2.02, "learning_rate": 2.3968541468064823e-06, "loss": 2.7463, "step": 10595 }, { "epoch": 2.02, "learning_rate": 2.396796949475691e-06, "loss": 2.657, "step": 10596 }, { "epoch": 2.02, "learning_rate": 2.3967397521449e-06, "loss": 2.6136, "step": 10597 }, { "epoch": 2.02, "learning_rate": 2.3966825548141087e-06, "loss": 2.243, "step": 10598 }, { "epoch": 2.02, "learning_rate": 2.3966253574833175e-06, "loss": 2.6466, "step": 10599 }, { "epoch": 2.02, "learning_rate": 2.3965681601525262e-06, "loss": 2.5835, "step": 10600 }, { "epoch": 2.02, "learning_rate": 2.396510962821735e-06, "loss": 3.1967, "step": 10601 }, { "epoch": 2.02, "learning_rate": 2.396453765490944e-06, "loss": 3.1308, "step": 10602 }, { "epoch": 2.02, "learning_rate": 2.3963965681601526e-06, "loss": 2.3753, "step": 10603 }, { "epoch": 2.02, "learning_rate": 2.3963393708293614e-06, "loss": 1.7241, "step": 10604 }, { "epoch": 2.02, "learning_rate": 2.3962821734985702e-06, "loss": 3.7085, "step": 10605 }, { "epoch": 2.02, "learning_rate": 2.396224976167779e-06, "loss": 3.0158, "step": 10606 }, { "epoch": 2.02, "learning_rate": 2.3961677788369874e-06, "loss": 2.2452, "step": 10607 }, { "epoch": 2.02, "learning_rate": 2.3961105815061966e-06, "loss": 3.4156, "step": 10608 }, { "epoch": 2.02, "learning_rate": 2.3960533841754054e-06, "loss": 2.4233, "step": 10609 }, { "epoch": 2.02, "learning_rate": 2.3959961868446138e-06, "loss": 1.9774, "step": 10610 }, { "epoch": 2.02, "learning_rate": 2.395938989513823e-06, "loss": 2.9935, "step": 10611 }, { "epoch": 2.02, "learning_rate": 2.3958817921830314e-06, "loss": 2.4899, "step": 10612 }, { "epoch": 2.02, "learning_rate": 2.39582459485224e-06, "loss": 2.2232, "step": 10613 }, { "epoch": 2.02, "learning_rate": 2.395767397521449e-06, "loss": 3.4381, "step": 10614 }, { "epoch": 2.02, "learning_rate": 2.3957102001906577e-06, "loss": 2.822, "step": 10615 }, { "epoch": 2.02, "learning_rate": 2.3956530028598665e-06, "loss": 2.1876, "step": 10616 }, { "epoch": 2.02, "learning_rate": 2.3955958055290753e-06, "loss": 2.8929, "step": 10617 }, { "epoch": 2.02, "learning_rate": 2.395538608198284e-06, "loss": 2.5412, "step": 10618 }, { "epoch": 2.02, "learning_rate": 2.395481410867493e-06, "loss": 3.1309, "step": 10619 }, { "epoch": 2.02, "learning_rate": 2.3954242135367017e-06, "loss": 3.5145, "step": 10620 }, { "epoch": 2.02, "learning_rate": 2.3953670162059105e-06, "loss": 2.6927, "step": 10621 }, { "epoch": 2.02, "learning_rate": 2.3953098188751193e-06, "loss": 3.0207, "step": 10622 }, { "epoch": 2.02, "learning_rate": 2.395252621544328e-06, "loss": 2.5811, "step": 10623 }, { "epoch": 2.02, "learning_rate": 2.3951954242135365e-06, "loss": 2.706, "step": 10624 }, { "epoch": 2.02, "learning_rate": 2.3951382268827457e-06, "loss": 2.7427, "step": 10625 }, { "epoch": 2.02, "learning_rate": 2.3950810295519545e-06, "loss": 2.3986, "step": 10626 }, { "epoch": 2.02, "learning_rate": 2.395023832221163e-06, "loss": 3.6295, "step": 10627 }, { "epoch": 2.02, "learning_rate": 2.394966634890372e-06, "loss": 3.2355, "step": 10628 }, { "epoch": 2.02, "learning_rate": 2.3949094375595804e-06, "loss": 3.1893, "step": 10629 }, { "epoch": 2.02, "learning_rate": 2.3948522402287892e-06, "loss": 3.1732, "step": 10630 }, { "epoch": 2.02, "learning_rate": 2.394795042897998e-06, "loss": 2.7289, "step": 10631 }, { "epoch": 2.03, "learning_rate": 2.394737845567207e-06, "loss": 2.5982, "step": 10632 }, { "epoch": 2.03, "learning_rate": 2.394680648236416e-06, "loss": 2.8505, "step": 10633 }, { "epoch": 2.03, "learning_rate": 2.3946234509056244e-06, "loss": 2.6729, "step": 10634 }, { "epoch": 2.03, "learning_rate": 2.394566253574833e-06, "loss": 2.9385, "step": 10635 }, { "epoch": 2.03, "learning_rate": 2.394509056244042e-06, "loss": 2.2007, "step": 10636 }, { "epoch": 2.03, "learning_rate": 2.3944518589132508e-06, "loss": 2.5363, "step": 10637 }, { "epoch": 2.03, "learning_rate": 2.3943946615824596e-06, "loss": 2.3513, "step": 10638 }, { "epoch": 2.03, "learning_rate": 2.3943374642516684e-06, "loss": 3.6345, "step": 10639 }, { "epoch": 2.03, "learning_rate": 2.394280266920877e-06, "loss": 3.4506, "step": 10640 }, { "epoch": 2.03, "learning_rate": 2.394223069590086e-06, "loss": 2.8266, "step": 10641 }, { "epoch": 2.03, "learning_rate": 2.3941658722592947e-06, "loss": 3.2206, "step": 10642 }, { "epoch": 2.03, "learning_rate": 2.3941086749285035e-06, "loss": 2.6655, "step": 10643 }, { "epoch": 2.03, "learning_rate": 2.394051477597712e-06, "loss": 3.3201, "step": 10644 }, { "epoch": 2.03, "learning_rate": 2.393994280266921e-06, "loss": 2.6789, "step": 10645 }, { "epoch": 2.03, "learning_rate": 2.3939370829361295e-06, "loss": 2.4269, "step": 10646 }, { "epoch": 2.03, "learning_rate": 2.3938798856053383e-06, "loss": 2.7938, "step": 10647 }, { "epoch": 2.03, "learning_rate": 2.3938226882745475e-06, "loss": 2.6933, "step": 10648 }, { "epoch": 2.03, "learning_rate": 2.393765490943756e-06, "loss": 2.6778, "step": 10649 }, { "epoch": 2.03, "learning_rate": 2.393708293612965e-06, "loss": 3.2187, "step": 10650 }, { "epoch": 2.03, "learning_rate": 2.3936510962821735e-06, "loss": 2.3998, "step": 10651 }, { "epoch": 2.03, "learning_rate": 2.3935938989513823e-06, "loss": 3.1744, "step": 10652 }, { "epoch": 2.03, "learning_rate": 2.393536701620591e-06, "loss": 2.8941, "step": 10653 }, { "epoch": 2.03, "learning_rate": 2.3934795042898e-06, "loss": 2.3418, "step": 10654 }, { "epoch": 2.03, "learning_rate": 2.3934223069590086e-06, "loss": 2.3115, "step": 10655 }, { "epoch": 2.03, "learning_rate": 2.3933651096282174e-06, "loss": 3.0474, "step": 10656 }, { "epoch": 2.03, "learning_rate": 2.3933079122974262e-06, "loss": 2.3907, "step": 10657 }, { "epoch": 2.03, "learning_rate": 2.393250714966635e-06, "loss": 3.2832, "step": 10658 }, { "epoch": 2.03, "learning_rate": 2.393193517635844e-06, "loss": 2.9998, "step": 10659 }, { "epoch": 2.03, "learning_rate": 2.3931363203050526e-06, "loss": 2.683, "step": 10660 }, { "epoch": 2.03, "learning_rate": 2.393079122974261e-06, "loss": 3.0037, "step": 10661 }, { "epoch": 2.03, "learning_rate": 2.39302192564347e-06, "loss": 2.3668, "step": 10662 }, { "epoch": 2.03, "learning_rate": 2.3929647283126786e-06, "loss": 3.0926, "step": 10663 }, { "epoch": 2.03, "learning_rate": 2.3929075309818874e-06, "loss": 2.4861, "step": 10664 }, { "epoch": 2.03, "learning_rate": 2.3928503336510966e-06, "loss": 2.7403, "step": 10665 }, { "epoch": 2.03, "learning_rate": 2.392793136320305e-06, "loss": 3.2692, "step": 10666 }, { "epoch": 2.03, "learning_rate": 2.392735938989514e-06, "loss": 2.0355, "step": 10667 }, { "epoch": 2.03, "learning_rate": 2.3926787416587225e-06, "loss": 2.1678, "step": 10668 }, { "epoch": 2.03, "learning_rate": 2.3926215443279313e-06, "loss": 2.7433, "step": 10669 }, { "epoch": 2.03, "learning_rate": 2.39256434699714e-06, "loss": 2.7214, "step": 10670 }, { "epoch": 2.03, "learning_rate": 2.392507149666349e-06, "loss": 2.7379, "step": 10671 }, { "epoch": 2.03, "learning_rate": 2.3924499523355577e-06, "loss": 2.3979, "step": 10672 }, { "epoch": 2.03, "learning_rate": 2.3923927550047665e-06, "loss": 2.0494, "step": 10673 }, { "epoch": 2.03, "learning_rate": 2.3923355576739753e-06, "loss": 1.8635, "step": 10674 }, { "epoch": 2.03, "learning_rate": 2.392278360343184e-06, "loss": 3.2022, "step": 10675 }, { "epoch": 2.03, "learning_rate": 2.392221163012393e-06, "loss": 3.0851, "step": 10676 }, { "epoch": 2.03, "learning_rate": 2.3921639656816017e-06, "loss": 2.8615, "step": 10677 }, { "epoch": 2.03, "learning_rate": 2.39210676835081e-06, "loss": 2.7454, "step": 10678 }, { "epoch": 2.03, "learning_rate": 2.3920495710200193e-06, "loss": 2.4957, "step": 10679 }, { "epoch": 2.03, "learning_rate": 2.391992373689228e-06, "loss": 2.643, "step": 10680 }, { "epoch": 2.03, "learning_rate": 2.3919351763584364e-06, "loss": 2.8194, "step": 10681 }, { "epoch": 2.03, "learning_rate": 2.3918779790276456e-06, "loss": 2.8612, "step": 10682 }, { "epoch": 2.03, "learning_rate": 2.391820781696854e-06, "loss": 2.7553, "step": 10683 }, { "epoch": 2.04, "learning_rate": 2.3917635843660632e-06, "loss": 3.8329, "step": 10684 }, { "epoch": 2.04, "learning_rate": 2.3917063870352716e-06, "loss": 3.0134, "step": 10685 }, { "epoch": 2.04, "learning_rate": 2.3916491897044804e-06, "loss": 3.2415, "step": 10686 }, { "epoch": 2.04, "learning_rate": 2.3915919923736896e-06, "loss": 2.7257, "step": 10687 }, { "epoch": 2.04, "learning_rate": 2.391534795042898e-06, "loss": 2.8145, "step": 10688 }, { "epoch": 2.04, "learning_rate": 2.3914775977121068e-06, "loss": 2.1389, "step": 10689 }, { "epoch": 2.04, "learning_rate": 2.3914204003813156e-06, "loss": 2.708, "step": 10690 }, { "epoch": 2.04, "learning_rate": 2.3913632030505244e-06, "loss": 2.4294, "step": 10691 }, { "epoch": 2.04, "learning_rate": 2.391306005719733e-06, "loss": 2.4746, "step": 10692 }, { "epoch": 2.04, "learning_rate": 2.391248808388942e-06, "loss": 2.3125, "step": 10693 }, { "epoch": 2.04, "learning_rate": 2.3911916110581508e-06, "loss": 3.6337, "step": 10694 }, { "epoch": 2.04, "learning_rate": 2.391134413727359e-06, "loss": 3.0048, "step": 10695 }, { "epoch": 2.04, "learning_rate": 2.3910772163965683e-06, "loss": 2.9031, "step": 10696 }, { "epoch": 2.04, "learning_rate": 2.391020019065777e-06, "loss": 2.4267, "step": 10697 }, { "epoch": 2.04, "learning_rate": 2.390962821734986e-06, "loss": 2.7513, "step": 10698 }, { "epoch": 2.04, "learning_rate": 2.3909056244041947e-06, "loss": 2.7291, "step": 10699 }, { "epoch": 2.04, "learning_rate": 2.390848427073403e-06, "loss": 2.9614, "step": 10700 }, { "epoch": 2.04, "learning_rate": 2.3907912297426123e-06, "loss": 3.1767, "step": 10701 }, { "epoch": 2.04, "learning_rate": 2.3907340324118207e-06, "loss": 2.9249, "step": 10702 }, { "epoch": 2.04, "learning_rate": 2.3906768350810295e-06, "loss": 2.7947, "step": 10703 }, { "epoch": 2.04, "learning_rate": 2.3906196377502387e-06, "loss": 2.5289, "step": 10704 }, { "epoch": 2.04, "learning_rate": 2.390562440419447e-06, "loss": 2.7369, "step": 10705 }, { "epoch": 2.04, "learning_rate": 2.390505243088656e-06, "loss": 2.4022, "step": 10706 }, { "epoch": 2.04, "learning_rate": 2.3904480457578646e-06, "loss": 3.9411, "step": 10707 }, { "epoch": 2.04, "learning_rate": 2.3903908484270734e-06, "loss": 2.5176, "step": 10708 }, { "epoch": 2.04, "learning_rate": 2.3903336510962822e-06, "loss": 2.3908, "step": 10709 }, { "epoch": 2.04, "learning_rate": 2.390276453765491e-06, "loss": 3.1004, "step": 10710 }, { "epoch": 2.04, "learning_rate": 2.3902192564347e-06, "loss": 3.3469, "step": 10711 }, { "epoch": 2.04, "learning_rate": 2.390162059103908e-06, "loss": 2.7821, "step": 10712 }, { "epoch": 2.04, "learning_rate": 2.3901048617731174e-06, "loss": 3.2919, "step": 10713 }, { "epoch": 2.04, "learning_rate": 2.390047664442326e-06, "loss": 2.6541, "step": 10714 }, { "epoch": 2.04, "learning_rate": 2.389990467111535e-06, "loss": 2.1098, "step": 10715 }, { "epoch": 2.04, "learning_rate": 2.389933269780744e-06, "loss": 2.2862, "step": 10716 }, { "epoch": 2.04, "learning_rate": 2.389876072449952e-06, "loss": 2.8318, "step": 10717 }, { "epoch": 2.04, "learning_rate": 2.3898188751191614e-06, "loss": 2.2971, "step": 10718 }, { "epoch": 2.04, "learning_rate": 2.3897616777883697e-06, "loss": 2.6601, "step": 10719 }, { "epoch": 2.04, "learning_rate": 2.3897044804575785e-06, "loss": 2.8184, "step": 10720 }, { "epoch": 2.04, "learning_rate": 2.3896472831267878e-06, "loss": 2.5728, "step": 10721 }, { "epoch": 2.04, "learning_rate": 2.389590085795996e-06, "loss": 3.476, "step": 10722 }, { "epoch": 2.04, "learning_rate": 2.389532888465205e-06, "loss": 2.8788, "step": 10723 }, { "epoch": 2.04, "learning_rate": 2.3894756911344137e-06, "loss": 3.1636, "step": 10724 }, { "epoch": 2.04, "learning_rate": 2.3894184938036225e-06, "loss": 3.573, "step": 10725 }, { "epoch": 2.04, "learning_rate": 2.3893612964728313e-06, "loss": 2.0829, "step": 10726 }, { "epoch": 2.04, "learning_rate": 2.38930409914204e-06, "loss": 2.2214, "step": 10727 }, { "epoch": 2.04, "learning_rate": 2.389246901811249e-06, "loss": 3.613, "step": 10728 }, { "epoch": 2.04, "learning_rate": 2.3891897044804577e-06, "loss": 3.3174, "step": 10729 }, { "epoch": 2.04, "learning_rate": 2.3891325071496665e-06, "loss": 2.3553, "step": 10730 }, { "epoch": 2.04, "learning_rate": 2.3890753098188753e-06, "loss": 2.6029, "step": 10731 }, { "epoch": 2.04, "learning_rate": 2.389018112488084e-06, "loss": 2.8857, "step": 10732 }, { "epoch": 2.04, "learning_rate": 2.388960915157293e-06, "loss": 3.4889, "step": 10733 }, { "epoch": 2.04, "learning_rate": 2.3889037178265012e-06, "loss": 2.2503, "step": 10734 }, { "epoch": 2.04, "learning_rate": 2.3888465204957105e-06, "loss": 2.6831, "step": 10735 }, { "epoch": 2.04, "learning_rate": 2.3887893231649192e-06, "loss": 3.4435, "step": 10736 }, { "epoch": 2.05, "learning_rate": 2.3887321258341276e-06, "loss": 2.3229, "step": 10737 }, { "epoch": 2.05, "learning_rate": 2.388674928503337e-06, "loss": 2.4637, "step": 10738 }, { "epoch": 2.05, "learning_rate": 2.388617731172545e-06, "loss": 2.7666, "step": 10739 }, { "epoch": 2.05, "learning_rate": 2.388560533841754e-06, "loss": 3.001, "step": 10740 }, { "epoch": 2.05, "learning_rate": 2.388503336510963e-06, "loss": 2.3654, "step": 10741 }, { "epoch": 2.05, "learning_rate": 2.3884461391801716e-06, "loss": 2.736, "step": 10742 }, { "epoch": 2.05, "learning_rate": 2.3883889418493804e-06, "loss": 2.9203, "step": 10743 }, { "epoch": 2.05, "learning_rate": 2.388331744518589e-06, "loss": 3.4, "step": 10744 }, { "epoch": 2.05, "learning_rate": 2.388274547187798e-06, "loss": 2.7282, "step": 10745 }, { "epoch": 2.05, "learning_rate": 2.3882173498570068e-06, "loss": 2.6815, "step": 10746 }, { "epoch": 2.05, "learning_rate": 2.3881601525262156e-06, "loss": 2.8717, "step": 10747 }, { "epoch": 2.05, "learning_rate": 2.3881029551954243e-06, "loss": 3.42, "step": 10748 }, { "epoch": 2.05, "learning_rate": 2.388045757864633e-06, "loss": 3.2756, "step": 10749 }, { "epoch": 2.05, "learning_rate": 2.387988560533842e-06, "loss": 2.7774, "step": 10750 }, { "epoch": 2.05, "learning_rate": 2.3879313632030503e-06, "loss": 3.105, "step": 10751 }, { "epoch": 2.05, "learning_rate": 2.3878741658722595e-06, "loss": 3.1533, "step": 10752 }, { "epoch": 2.05, "learning_rate": 2.3878169685414683e-06, "loss": 2.4973, "step": 10753 }, { "epoch": 2.05, "learning_rate": 2.3877597712106767e-06, "loss": 3.1964, "step": 10754 }, { "epoch": 2.05, "learning_rate": 2.387702573879886e-06, "loss": 3.2802, "step": 10755 }, { "epoch": 2.05, "learning_rate": 2.3876453765490943e-06, "loss": 3.0128, "step": 10756 }, { "epoch": 2.05, "learning_rate": 2.387588179218303e-06, "loss": 3.5653, "step": 10757 }, { "epoch": 2.05, "learning_rate": 2.387530981887512e-06, "loss": 2.7198, "step": 10758 }, { "epoch": 2.05, "learning_rate": 2.3874737845567207e-06, "loss": 2.9322, "step": 10759 }, { "epoch": 2.05, "learning_rate": 2.3874165872259294e-06, "loss": 3.3351, "step": 10760 }, { "epoch": 2.05, "learning_rate": 2.3873593898951382e-06, "loss": 3.6314, "step": 10761 }, { "epoch": 2.05, "learning_rate": 2.387302192564347e-06, "loss": 3.5216, "step": 10762 }, { "epoch": 2.05, "learning_rate": 2.387244995233556e-06, "loss": 2.553, "step": 10763 }, { "epoch": 2.05, "learning_rate": 2.3871877979027646e-06, "loss": 3.1959, "step": 10764 }, { "epoch": 2.05, "learning_rate": 2.3871306005719734e-06, "loss": 2.9552, "step": 10765 }, { "epoch": 2.05, "learning_rate": 2.3870734032411822e-06, "loss": 2.8174, "step": 10766 }, { "epoch": 2.05, "learning_rate": 2.387016205910391e-06, "loss": 3.2892, "step": 10767 }, { "epoch": 2.05, "learning_rate": 2.3869590085796e-06, "loss": 3.1469, "step": 10768 }, { "epoch": 2.05, "learning_rate": 2.3869018112488086e-06, "loss": 2.5348, "step": 10769 }, { "epoch": 2.05, "learning_rate": 2.3868446139180174e-06, "loss": 3.1598, "step": 10770 }, { "epoch": 2.05, "learning_rate": 2.3867874165872258e-06, "loss": 3.2561, "step": 10771 }, { "epoch": 2.05, "learning_rate": 2.386730219256435e-06, "loss": 2.95, "step": 10772 }, { "epoch": 2.05, "learning_rate": 2.3866730219256433e-06, "loss": 4.2683, "step": 10773 }, { "epoch": 2.05, "learning_rate": 2.386615824594852e-06, "loss": 3.0604, "step": 10774 }, { "epoch": 2.05, "learning_rate": 2.3865586272640614e-06, "loss": 2.649, "step": 10775 }, { "epoch": 2.05, "learning_rate": 2.3865014299332697e-06, "loss": 2.9407, "step": 10776 }, { "epoch": 2.05, "learning_rate": 2.3864442326024785e-06, "loss": 2.7106, "step": 10777 }, { "epoch": 2.05, "learning_rate": 2.3863870352716873e-06, "loss": 2.0102, "step": 10778 }, { "epoch": 2.05, "learning_rate": 2.386329837940896e-06, "loss": 3.1344, "step": 10779 }, { "epoch": 2.05, "learning_rate": 2.386272640610105e-06, "loss": 3.308, "step": 10780 }, { "epoch": 2.05, "learning_rate": 2.3862154432793137e-06, "loss": 2.8888, "step": 10781 }, { "epoch": 2.05, "learning_rate": 2.3861582459485225e-06, "loss": 3.7211, "step": 10782 }, { "epoch": 2.05, "learning_rate": 2.3861010486177313e-06, "loss": 3.0624, "step": 10783 }, { "epoch": 2.05, "learning_rate": 2.38604385128694e-06, "loss": 2.6506, "step": 10784 }, { "epoch": 2.05, "learning_rate": 2.385986653956149e-06, "loss": 3.03, "step": 10785 }, { "epoch": 2.05, "learning_rate": 2.3859294566253577e-06, "loss": 2.8628, "step": 10786 }, { "epoch": 2.05, "learning_rate": 2.3858722592945665e-06, "loss": 2.1826, "step": 10787 }, { "epoch": 2.05, "learning_rate": 2.385815061963775e-06, "loss": 2.1813, "step": 10788 }, { "epoch": 2.06, "learning_rate": 2.385757864632984e-06, "loss": 3.2795, "step": 10789 }, { "epoch": 2.06, "learning_rate": 2.3857006673021924e-06, "loss": 3.2297, "step": 10790 }, { "epoch": 2.06, "learning_rate": 2.385643469971401e-06, "loss": 3.6614, "step": 10791 }, { "epoch": 2.06, "learning_rate": 2.3855862726406104e-06, "loss": 2.5912, "step": 10792 }, { "epoch": 2.06, "learning_rate": 2.385529075309819e-06, "loss": 2.7261, "step": 10793 }, { "epoch": 2.06, "learning_rate": 2.3854718779790276e-06, "loss": 3.3021, "step": 10794 }, { "epoch": 2.06, "learning_rate": 2.3854146806482364e-06, "loss": 3.0356, "step": 10795 }, { "epoch": 2.06, "learning_rate": 2.385357483317445e-06, "loss": 2.491, "step": 10796 }, { "epoch": 2.06, "learning_rate": 2.385300285986654e-06, "loss": 2.0315, "step": 10797 }, { "epoch": 2.06, "learning_rate": 2.3852430886558628e-06, "loss": 2.5953, "step": 10798 }, { "epoch": 2.06, "learning_rate": 2.3851858913250716e-06, "loss": 3.7149, "step": 10799 }, { "epoch": 2.06, "learning_rate": 2.3851286939942804e-06, "loss": 2.4959, "step": 10800 }, { "epoch": 2.06, "learning_rate": 2.385071496663489e-06, "loss": 2.1808, "step": 10801 }, { "epoch": 2.06, "learning_rate": 2.385014299332698e-06, "loss": 2.236, "step": 10802 }, { "epoch": 2.06, "learning_rate": 2.3849571020019067e-06, "loss": 2.4781, "step": 10803 }, { "epoch": 2.06, "learning_rate": 2.3848999046711155e-06, "loss": 3.7287, "step": 10804 }, { "epoch": 2.06, "learning_rate": 2.384842707340324e-06, "loss": 2.366, "step": 10805 }, { "epoch": 2.06, "learning_rate": 2.384785510009533e-06, "loss": 2.5576, "step": 10806 }, { "epoch": 2.06, "learning_rate": 2.3847283126787415e-06, "loss": 2.8413, "step": 10807 }, { "epoch": 2.06, "learning_rate": 2.3846711153479503e-06, "loss": 3.1781, "step": 10808 }, { "epoch": 2.06, "learning_rate": 2.3846139180171595e-06, "loss": 2.9746, "step": 10809 }, { "epoch": 2.06, "learning_rate": 2.384556720686368e-06, "loss": 2.6101, "step": 10810 }, { "epoch": 2.06, "learning_rate": 2.3844995233555767e-06, "loss": 3.2634, "step": 10811 }, { "epoch": 2.06, "learning_rate": 2.3844423260247855e-06, "loss": 2.5734, "step": 10812 }, { "epoch": 2.06, "learning_rate": 2.3843851286939943e-06, "loss": 3.3548, "step": 10813 }, { "epoch": 2.06, "learning_rate": 2.3843279313632035e-06, "loss": 2.5064, "step": 10814 }, { "epoch": 2.06, "learning_rate": 2.384270734032412e-06, "loss": 2.3202, "step": 10815 }, { "epoch": 2.06, "learning_rate": 2.3842135367016206e-06, "loss": 3.7517, "step": 10816 }, { "epoch": 2.06, "learning_rate": 2.3841563393708294e-06, "loss": 3.1178, "step": 10817 }, { "epoch": 2.06, "learning_rate": 2.3840991420400382e-06, "loss": 3.0002, "step": 10818 }, { "epoch": 2.06, "learning_rate": 2.384041944709247e-06, "loss": 3.4286, "step": 10819 }, { "epoch": 2.06, "learning_rate": 2.383984747378456e-06, "loss": 2.1326, "step": 10820 }, { "epoch": 2.06, "learning_rate": 2.3839275500476646e-06, "loss": 3.1829, "step": 10821 }, { "epoch": 2.06, "learning_rate": 2.383870352716873e-06, "loss": 3.2566, "step": 10822 }, { "epoch": 2.06, "learning_rate": 2.383813155386082e-06, "loss": 2.8965, "step": 10823 }, { "epoch": 2.06, "learning_rate": 2.383755958055291e-06, "loss": 2.2825, "step": 10824 }, { "epoch": 2.06, "learning_rate": 2.3836987607244994e-06, "loss": 2.4298, "step": 10825 }, { "epoch": 2.06, "learning_rate": 2.3836415633937086e-06, "loss": 2.0761, "step": 10826 }, { "epoch": 2.06, "learning_rate": 2.383584366062917e-06, "loss": 3.0591, "step": 10827 }, { "epoch": 2.06, "learning_rate": 2.383527168732126e-06, "loss": 3.0448, "step": 10828 }, { "epoch": 2.06, "learning_rate": 2.3834699714013345e-06, "loss": 2.2003, "step": 10829 }, { "epoch": 2.06, "learning_rate": 2.3834127740705433e-06, "loss": 2.8577, "step": 10830 }, { "epoch": 2.06, "learning_rate": 2.3833555767397525e-06, "loss": 3.3724, "step": 10831 }, { "epoch": 2.06, "learning_rate": 2.383298379408961e-06, "loss": 2.812, "step": 10832 }, { "epoch": 2.06, "learning_rate": 2.3832411820781697e-06, "loss": 2.0702, "step": 10833 }, { "epoch": 2.06, "learning_rate": 2.3831839847473785e-06, "loss": 3.0502, "step": 10834 }, { "epoch": 2.06, "learning_rate": 2.3831267874165873e-06, "loss": 2.6222, "step": 10835 }, { "epoch": 2.06, "learning_rate": 2.383069590085796e-06, "loss": 2.9021, "step": 10836 }, { "epoch": 2.06, "learning_rate": 2.383012392755005e-06, "loss": 2.6141, "step": 10837 }, { "epoch": 2.06, "learning_rate": 2.3829551954242137e-06, "loss": 3.2133, "step": 10838 }, { "epoch": 2.06, "learning_rate": 2.382897998093422e-06, "loss": 2.9757, "step": 10839 }, { "epoch": 2.06, "learning_rate": 2.3828408007626313e-06, "loss": 2.5613, "step": 10840 }, { "epoch": 2.06, "learning_rate": 2.38278360343184e-06, "loss": 3.1898, "step": 10841 }, { "epoch": 2.07, "learning_rate": 2.3827264061010484e-06, "loss": 2.5186, "step": 10842 }, { "epoch": 2.07, "learning_rate": 2.3826692087702576e-06, "loss": 2.3162, "step": 10843 }, { "epoch": 2.07, "learning_rate": 2.382612011439466e-06, "loss": 3.3982, "step": 10844 }, { "epoch": 2.07, "learning_rate": 2.3825548141086752e-06, "loss": 3.5675, "step": 10845 }, { "epoch": 2.07, "learning_rate": 2.3824976167778836e-06, "loss": 3.3794, "step": 10846 }, { "epoch": 2.07, "learning_rate": 2.3824404194470924e-06, "loss": 2.3429, "step": 10847 }, { "epoch": 2.07, "learning_rate": 2.3823832221163016e-06, "loss": 2.5319, "step": 10848 }, { "epoch": 2.07, "learning_rate": 2.38232602478551e-06, "loss": 2.8816, "step": 10849 }, { "epoch": 2.07, "learning_rate": 2.3822688274547188e-06, "loss": 2.7393, "step": 10850 }, { "epoch": 2.07, "learning_rate": 2.3822116301239276e-06, "loss": 2.7818, "step": 10851 }, { "epoch": 2.07, "learning_rate": 2.3821544327931364e-06, "loss": 2.4308, "step": 10852 }, { "epoch": 2.07, "learning_rate": 2.382097235462345e-06, "loss": 3.3561, "step": 10853 }, { "epoch": 2.07, "learning_rate": 2.382040038131554e-06, "loss": 3.0642, "step": 10854 }, { "epoch": 2.07, "learning_rate": 2.3819828408007627e-06, "loss": 3.2495, "step": 10855 }, { "epoch": 2.07, "learning_rate": 2.381925643469971e-06, "loss": 3.6448, "step": 10856 }, { "epoch": 2.07, "learning_rate": 2.3818684461391803e-06, "loss": 2.9688, "step": 10857 }, { "epoch": 2.07, "learning_rate": 2.381811248808389e-06, "loss": 3.1947, "step": 10858 }, { "epoch": 2.07, "learning_rate": 2.3817540514775975e-06, "loss": 3.2958, "step": 10859 }, { "epoch": 2.07, "learning_rate": 2.3816968541468067e-06, "loss": 3.7384, "step": 10860 }, { "epoch": 2.07, "learning_rate": 2.381639656816015e-06, "loss": 2.3602, "step": 10861 }, { "epoch": 2.07, "learning_rate": 2.3815824594852243e-06, "loss": 3.0006, "step": 10862 }, { "epoch": 2.07, "learning_rate": 2.381525262154433e-06, "loss": 1.9331, "step": 10863 }, { "epoch": 2.07, "learning_rate": 2.3814680648236415e-06, "loss": 1.7014, "step": 10864 }, { "epoch": 2.07, "learning_rate": 2.3814108674928507e-06, "loss": 3.0523, "step": 10865 }, { "epoch": 2.07, "learning_rate": 2.381353670162059e-06, "loss": 3.3129, "step": 10866 }, { "epoch": 2.07, "learning_rate": 2.381296472831268e-06, "loss": 2.3769, "step": 10867 }, { "epoch": 2.07, "learning_rate": 2.3812392755004766e-06, "loss": 2.8137, "step": 10868 }, { "epoch": 2.07, "learning_rate": 2.3811820781696854e-06, "loss": 3.0318, "step": 10869 }, { "epoch": 2.07, "learning_rate": 2.3811248808388942e-06, "loss": 2.8908, "step": 10870 }, { "epoch": 2.07, "learning_rate": 2.381067683508103e-06, "loss": 2.2659, "step": 10871 }, { "epoch": 2.07, "learning_rate": 2.381010486177312e-06, "loss": 3.2679, "step": 10872 }, { "epoch": 2.07, "learning_rate": 2.3809532888465206e-06, "loss": 2.6736, "step": 10873 }, { "epoch": 2.07, "learning_rate": 2.3808960915157294e-06, "loss": 1.9341, "step": 10874 }, { "epoch": 2.07, "learning_rate": 2.380838894184938e-06, "loss": 3.7373, "step": 10875 }, { "epoch": 2.07, "learning_rate": 2.3807816968541466e-06, "loss": 2.6613, "step": 10876 }, { "epoch": 2.07, "learning_rate": 2.3807244995233558e-06, "loss": 3.3905, "step": 10877 }, { "epoch": 2.07, "learning_rate": 2.380667302192564e-06, "loss": 3.1793, "step": 10878 }, { "epoch": 2.07, "learning_rate": 2.3806101048617734e-06, "loss": 2.0003, "step": 10879 }, { "epoch": 2.07, "learning_rate": 2.380552907530982e-06, "loss": 2.6393, "step": 10880 }, { "epoch": 2.07, "learning_rate": 2.3804957102001905e-06, "loss": 3.4115, "step": 10881 }, { "epoch": 2.07, "learning_rate": 2.3804385128693998e-06, "loss": 4.1699, "step": 10882 }, { "epoch": 2.07, "learning_rate": 2.380381315538608e-06, "loss": 3.7139, "step": 10883 }, { "epoch": 2.07, "learning_rate": 2.380324118207817e-06, "loss": 2.1083, "step": 10884 }, { "epoch": 2.07, "learning_rate": 2.3802669208770257e-06, "loss": 2.2383, "step": 10885 }, { "epoch": 2.07, "learning_rate": 2.3802097235462345e-06, "loss": 2.6624, "step": 10886 }, { "epoch": 2.07, "learning_rate": 2.3801525262154433e-06, "loss": 3.1387, "step": 10887 }, { "epoch": 2.07, "learning_rate": 2.380095328884652e-06, "loss": 2.3552, "step": 10888 }, { "epoch": 2.07, "learning_rate": 2.380038131553861e-06, "loss": 2.4733, "step": 10889 }, { "epoch": 2.07, "learning_rate": 2.3799809342230697e-06, "loss": 2.64, "step": 10890 }, { "epoch": 2.07, "learning_rate": 2.3799237368922785e-06, "loss": 3.0165, "step": 10891 }, { "epoch": 2.07, "learning_rate": 2.3798665395614873e-06, "loss": 2.5077, "step": 10892 }, { "epoch": 2.07, "learning_rate": 2.3798093422306956e-06, "loss": 2.9898, "step": 10893 }, { "epoch": 2.08, "learning_rate": 2.379752144899905e-06, "loss": 2.2232, "step": 10894 }, { "epoch": 2.08, "learning_rate": 2.3796949475691132e-06, "loss": 3.4869, "step": 10895 }, { "epoch": 2.08, "learning_rate": 2.3796377502383224e-06, "loss": 3.5996, "step": 10896 }, { "epoch": 2.08, "learning_rate": 2.3795805529075312e-06, "loss": 3.3331, "step": 10897 }, { "epoch": 2.08, "learning_rate": 2.3795233555767396e-06, "loss": 3.0348, "step": 10898 }, { "epoch": 2.08, "learning_rate": 2.379466158245949e-06, "loss": 3.7171, "step": 10899 }, { "epoch": 2.08, "learning_rate": 2.379408960915157e-06, "loss": 2.1461, "step": 10900 }, { "epoch": 2.08, "learning_rate": 2.379351763584366e-06, "loss": 2.6165, "step": 10901 }, { "epoch": 2.08, "learning_rate": 2.379294566253575e-06, "loss": 3.1752, "step": 10902 }, { "epoch": 2.08, "learning_rate": 2.3792373689227836e-06, "loss": 2.4714, "step": 10903 }, { "epoch": 2.08, "learning_rate": 2.3791801715919924e-06, "loss": 2.3237, "step": 10904 }, { "epoch": 2.08, "learning_rate": 2.379122974261201e-06, "loss": 2.605, "step": 10905 }, { "epoch": 2.08, "learning_rate": 2.37906577693041e-06, "loss": 2.0611, "step": 10906 }, { "epoch": 2.08, "learning_rate": 2.3790085795996188e-06, "loss": 3.3182, "step": 10907 }, { "epoch": 2.08, "learning_rate": 2.3789513822688275e-06, "loss": 3.3986, "step": 10908 }, { "epoch": 2.08, "learning_rate": 2.3788941849380363e-06, "loss": 2.7556, "step": 10909 }, { "epoch": 2.08, "learning_rate": 2.378836987607245e-06, "loss": 2.6275, "step": 10910 }, { "epoch": 2.08, "learning_rate": 2.378779790276454e-06, "loss": 3.2972, "step": 10911 }, { "epoch": 2.08, "learning_rate": 2.3787225929456627e-06, "loss": 2.8026, "step": 10912 }, { "epoch": 2.08, "learning_rate": 2.3786653956148715e-06, "loss": 3.4405, "step": 10913 }, { "epoch": 2.08, "learning_rate": 2.3786081982840803e-06, "loss": 2.3366, "step": 10914 }, { "epoch": 2.08, "learning_rate": 2.3785510009532887e-06, "loss": 3.6075, "step": 10915 }, { "epoch": 2.08, "learning_rate": 2.378493803622498e-06, "loss": 2.9189, "step": 10916 }, { "epoch": 2.08, "learning_rate": 2.3784366062917063e-06, "loss": 3.4053, "step": 10917 }, { "epoch": 2.08, "learning_rate": 2.378379408960915e-06, "loss": 2.3906, "step": 10918 }, { "epoch": 2.08, "learning_rate": 2.3783222116301243e-06, "loss": 2.9349, "step": 10919 }, { "epoch": 2.08, "learning_rate": 2.3782650142993326e-06, "loss": 2.7295, "step": 10920 }, { "epoch": 2.08, "learning_rate": 2.3782078169685414e-06, "loss": 2.6434, "step": 10921 }, { "epoch": 2.08, "learning_rate": 2.3781506196377502e-06, "loss": 2.5503, "step": 10922 }, { "epoch": 2.08, "learning_rate": 2.378093422306959e-06, "loss": 2.4452, "step": 10923 }, { "epoch": 2.08, "learning_rate": 2.378036224976168e-06, "loss": 2.5046, "step": 10924 }, { "epoch": 2.08, "learning_rate": 2.3779790276453766e-06, "loss": 2.8685, "step": 10925 }, { "epoch": 2.08, "learning_rate": 2.3779218303145854e-06, "loss": 2.8975, "step": 10926 }, { "epoch": 2.08, "learning_rate": 2.377864632983794e-06, "loss": 3.0908, "step": 10927 }, { "epoch": 2.08, "learning_rate": 2.377807435653003e-06, "loss": 3.2925, "step": 10928 }, { "epoch": 2.08, "learning_rate": 2.377750238322212e-06, "loss": 3.3533, "step": 10929 }, { "epoch": 2.08, "learning_rate": 2.3776930409914206e-06, "loss": 2.9249, "step": 10930 }, { "epoch": 2.08, "learning_rate": 2.3776358436606294e-06, "loss": 2.5781, "step": 10931 }, { "epoch": 2.08, "learning_rate": 2.3775786463298378e-06, "loss": 3.8732, "step": 10932 }, { "epoch": 2.08, "learning_rate": 2.377521448999047e-06, "loss": 2.5032, "step": 10933 }, { "epoch": 2.08, "learning_rate": 2.3774642516682553e-06, "loss": 2.7805, "step": 10934 }, { "epoch": 2.08, "learning_rate": 2.377407054337464e-06, "loss": 2.935, "step": 10935 }, { "epoch": 2.08, "learning_rate": 2.3773498570066733e-06, "loss": 3.3955, "step": 10936 }, { "epoch": 2.08, "learning_rate": 2.3772926596758817e-06, "loss": 2.9106, "step": 10937 }, { "epoch": 2.08, "learning_rate": 2.3772354623450905e-06, "loss": 3.6315, "step": 10938 }, { "epoch": 2.08, "learning_rate": 2.3771782650142993e-06, "loss": 3.1881, "step": 10939 }, { "epoch": 2.08, "learning_rate": 2.377121067683508e-06, "loss": 2.4587, "step": 10940 }, { "epoch": 2.08, "learning_rate": 2.377063870352717e-06, "loss": 2.284, "step": 10941 }, { "epoch": 2.08, "learning_rate": 2.3770066730219257e-06, "loss": 3.0693, "step": 10942 }, { "epoch": 2.08, "learning_rate": 2.3769494756911345e-06, "loss": 3.705, "step": 10943 }, { "epoch": 2.08, "learning_rate": 2.3768922783603433e-06, "loss": 2.5548, "step": 10944 }, { "epoch": 2.08, "learning_rate": 2.376835081029552e-06, "loss": 2.2989, "step": 10945 }, { "epoch": 2.08, "learning_rate": 2.376777883698761e-06, "loss": 2.8086, "step": 10946 }, { "epoch": 2.09, "learning_rate": 2.3767206863679697e-06, "loss": 2.8825, "step": 10947 }, { "epoch": 2.09, "learning_rate": 2.3766634890371785e-06, "loss": 2.13, "step": 10948 }, { "epoch": 2.09, "learning_rate": 2.376606291706387e-06, "loss": 2.6541, "step": 10949 }, { "epoch": 2.09, "learning_rate": 2.376549094375596e-06, "loss": 2.9381, "step": 10950 }, { "epoch": 2.09, "learning_rate": 2.376491897044805e-06, "loss": 2.5737, "step": 10951 }, { "epoch": 2.09, "learning_rate": 2.376434699714013e-06, "loss": 2.7128, "step": 10952 }, { "epoch": 2.09, "learning_rate": 2.3763775023832224e-06, "loss": 4.0203, "step": 10953 }, { "epoch": 2.09, "learning_rate": 2.376320305052431e-06, "loss": 3.3049, "step": 10954 }, { "epoch": 2.09, "learning_rate": 2.3762631077216396e-06, "loss": 1.9495, "step": 10955 }, { "epoch": 2.09, "learning_rate": 2.3762059103908484e-06, "loss": 2.3067, "step": 10956 }, { "epoch": 2.09, "learning_rate": 2.376148713060057e-06, "loss": 2.6647, "step": 10957 }, { "epoch": 2.09, "learning_rate": 2.376091515729266e-06, "loss": 2.4491, "step": 10958 }, { "epoch": 2.09, "learning_rate": 2.3760343183984748e-06, "loss": 2.8764, "step": 10959 }, { "epoch": 2.09, "learning_rate": 2.3759771210676836e-06, "loss": 2.9635, "step": 10960 }, { "epoch": 2.09, "learning_rate": 2.3759199237368923e-06, "loss": 2.4286, "step": 10961 }, { "epoch": 2.09, "learning_rate": 2.375862726406101e-06, "loss": 3.0491, "step": 10962 }, { "epoch": 2.09, "learning_rate": 2.37580552907531e-06, "loss": 3.5278, "step": 10963 }, { "epoch": 2.09, "learning_rate": 2.3757483317445187e-06, "loss": 2.9227, "step": 10964 }, { "epoch": 2.09, "learning_rate": 2.3756911344137275e-06, "loss": 3.0096, "step": 10965 }, { "epoch": 2.09, "learning_rate": 2.375633937082936e-06, "loss": 2.5496, "step": 10966 }, { "epoch": 2.09, "learning_rate": 2.375576739752145e-06, "loss": 3.3264, "step": 10967 }, { "epoch": 2.09, "learning_rate": 2.375519542421354e-06, "loss": 3.3878, "step": 10968 }, { "epoch": 2.09, "learning_rate": 2.3754623450905623e-06, "loss": 2.5358, "step": 10969 }, { "epoch": 2.09, "learning_rate": 2.3754051477597715e-06, "loss": 2.7448, "step": 10970 }, { "epoch": 2.09, "learning_rate": 2.37534795042898e-06, "loss": 3.3417, "step": 10971 }, { "epoch": 2.09, "learning_rate": 2.3752907530981887e-06, "loss": 3.7127, "step": 10972 }, { "epoch": 2.09, "learning_rate": 2.3752335557673975e-06, "loss": 3.0041, "step": 10973 }, { "epoch": 2.09, "learning_rate": 2.3751763584366062e-06, "loss": 3.0057, "step": 10974 }, { "epoch": 2.09, "learning_rate": 2.3751191611058155e-06, "loss": 2.5569, "step": 10975 }, { "epoch": 2.09, "learning_rate": 2.375061963775024e-06, "loss": 2.6534, "step": 10976 }, { "epoch": 2.09, "learning_rate": 2.3750047664442326e-06, "loss": 2.1391, "step": 10977 }, { "epoch": 2.09, "learning_rate": 2.3749475691134414e-06, "loss": 3.2545, "step": 10978 }, { "epoch": 2.09, "learning_rate": 2.3748903717826502e-06, "loss": 3.4417, "step": 10979 }, { "epoch": 2.09, "learning_rate": 2.374833174451859e-06, "loss": 1.4907, "step": 10980 }, { "epoch": 2.09, "learning_rate": 2.374775977121068e-06, "loss": 2.9886, "step": 10981 }, { "epoch": 2.09, "learning_rate": 2.3747187797902766e-06, "loss": 1.5975, "step": 10982 }, { "epoch": 2.09, "learning_rate": 2.374661582459485e-06, "loss": 2.7749, "step": 10983 }, { "epoch": 2.09, "learning_rate": 2.374604385128694e-06, "loss": 3.2233, "step": 10984 }, { "epoch": 2.09, "learning_rate": 2.374547187797903e-06, "loss": 2.2405, "step": 10985 }, { "epoch": 2.09, "learning_rate": 2.3744899904671113e-06, "loss": 1.7707, "step": 10986 }, { "epoch": 2.09, "learning_rate": 2.3744327931363206e-06, "loss": 3.1021, "step": 10987 }, { "epoch": 2.09, "learning_rate": 2.374375595805529e-06, "loss": 2.8876, "step": 10988 }, { "epoch": 2.09, "learning_rate": 2.3743183984747377e-06, "loss": 2.6613, "step": 10989 }, { "epoch": 2.09, "learning_rate": 2.374261201143947e-06, "loss": 3.0306, "step": 10990 }, { "epoch": 2.09, "learning_rate": 2.3742040038131553e-06, "loss": 3.2796, "step": 10991 }, { "epoch": 2.09, "learning_rate": 2.3741468064823645e-06, "loss": 3.5132, "step": 10992 }, { "epoch": 2.09, "learning_rate": 2.374089609151573e-06, "loss": 2.1202, "step": 10993 }, { "epoch": 2.09, "learning_rate": 2.3740324118207817e-06, "loss": 2.7722, "step": 10994 }, { "epoch": 2.09, "learning_rate": 2.3739752144899905e-06, "loss": 3.6406, "step": 10995 }, { "epoch": 2.09, "learning_rate": 2.3739180171591993e-06, "loss": 3.1463, "step": 10996 }, { "epoch": 2.09, "learning_rate": 2.373860819828408e-06, "loss": 2.9155, "step": 10997 }, { "epoch": 2.09, "learning_rate": 2.373803622497617e-06, "loss": 2.3483, "step": 10998 }, { "epoch": 2.1, "learning_rate": 2.3737464251668257e-06, "loss": 2.4816, "step": 10999 }, { "epoch": 2.1, "learning_rate": 2.3736892278360345e-06, "loss": 3.5469, "step": 11000 }, { "epoch": 2.1, "learning_rate": 2.3736320305052433e-06, "loss": 2.5983, "step": 11001 }, { "epoch": 2.1, "learning_rate": 2.373574833174452e-06, "loss": 3.071, "step": 11002 }, { "epoch": 2.1, "learning_rate": 2.3735176358436604e-06, "loss": 2.3348, "step": 11003 }, { "epoch": 2.1, "learning_rate": 2.3734604385128696e-06, "loss": 3.083, "step": 11004 }, { "epoch": 2.1, "learning_rate": 2.373403241182078e-06, "loss": 2.854, "step": 11005 }, { "epoch": 2.1, "learning_rate": 2.373346043851287e-06, "loss": 3.1553, "step": 11006 }, { "epoch": 2.1, "learning_rate": 2.373288846520496e-06, "loss": 3.0056, "step": 11007 }, { "epoch": 2.1, "learning_rate": 2.3732316491897044e-06, "loss": 3.7863, "step": 11008 }, { "epoch": 2.1, "learning_rate": 2.3731744518589136e-06, "loss": 2.8879, "step": 11009 }, { "epoch": 2.1, "learning_rate": 2.373117254528122e-06, "loss": 3.3667, "step": 11010 }, { "epoch": 2.1, "learning_rate": 2.3730600571973308e-06, "loss": 3.0581, "step": 11011 }, { "epoch": 2.1, "learning_rate": 2.3730028598665396e-06, "loss": 3.5108, "step": 11012 }, { "epoch": 2.1, "learning_rate": 2.3729456625357484e-06, "loss": 2.4064, "step": 11013 }, { "epoch": 2.1, "learning_rate": 2.372888465204957e-06, "loss": 2.3864, "step": 11014 }, { "epoch": 2.1, "learning_rate": 2.372831267874166e-06, "loss": 2.1544, "step": 11015 }, { "epoch": 2.1, "learning_rate": 2.3727740705433747e-06, "loss": 2.8384, "step": 11016 }, { "epoch": 2.1, "learning_rate": 2.3727168732125835e-06, "loss": 2.5359, "step": 11017 }, { "epoch": 2.1, "learning_rate": 2.3726596758817923e-06, "loss": 3.2862, "step": 11018 }, { "epoch": 2.1, "learning_rate": 2.372602478551001e-06, "loss": 1.9702, "step": 11019 }, { "epoch": 2.1, "learning_rate": 2.3725452812202095e-06, "loss": 3.8761, "step": 11020 }, { "epoch": 2.1, "learning_rate": 2.3724880838894187e-06, "loss": 2.4677, "step": 11021 }, { "epoch": 2.1, "learning_rate": 2.372430886558627e-06, "loss": 2.5144, "step": 11022 }, { "epoch": 2.1, "learning_rate": 2.372373689227836e-06, "loss": 2.7787, "step": 11023 }, { "epoch": 2.1, "learning_rate": 2.372316491897045e-06, "loss": 2.9846, "step": 11024 }, { "epoch": 2.1, "learning_rate": 2.3722592945662535e-06, "loss": 2.3687, "step": 11025 }, { "epoch": 2.1, "learning_rate": 2.3722020972354627e-06, "loss": 2.3951, "step": 11026 }, { "epoch": 2.1, "learning_rate": 2.372144899904671e-06, "loss": 2.7642, "step": 11027 }, { "epoch": 2.1, "learning_rate": 2.37208770257388e-06, "loss": 3.0357, "step": 11028 }, { "epoch": 2.1, "learning_rate": 2.3720305052430886e-06, "loss": 2.7977, "step": 11029 }, { "epoch": 2.1, "learning_rate": 2.3719733079122974e-06, "loss": 1.9603, "step": 11030 }, { "epoch": 2.1, "learning_rate": 2.3719161105815062e-06, "loss": 3.7448, "step": 11031 }, { "epoch": 2.1, "learning_rate": 2.371858913250715e-06, "loss": 3.0107, "step": 11032 }, { "epoch": 2.1, "learning_rate": 2.371801715919924e-06, "loss": 2.4466, "step": 11033 }, { "epoch": 2.1, "learning_rate": 2.3717445185891326e-06, "loss": 3.0222, "step": 11034 }, { "epoch": 2.1, "learning_rate": 2.3716873212583414e-06, "loss": 3.2431, "step": 11035 }, { "epoch": 2.1, "learning_rate": 2.37163012392755e-06, "loss": 2.8347, "step": 11036 }, { "epoch": 2.1, "learning_rate": 2.3715729265967586e-06, "loss": 3.2959, "step": 11037 }, { "epoch": 2.1, "learning_rate": 2.3715157292659678e-06, "loss": 2.7467, "step": 11038 }, { "epoch": 2.1, "learning_rate": 2.3714585319351766e-06, "loss": 2.6028, "step": 11039 }, { "epoch": 2.1, "learning_rate": 2.3714013346043854e-06, "loss": 3.2433, "step": 11040 }, { "epoch": 2.1, "learning_rate": 2.371344137273594e-06, "loss": 2.8859, "step": 11041 }, { "epoch": 2.1, "learning_rate": 2.3712869399428025e-06, "loss": 3.3182, "step": 11042 }, { "epoch": 2.1, "learning_rate": 2.3712297426120117e-06, "loss": 3.2053, "step": 11043 }, { "epoch": 2.1, "learning_rate": 2.37117254528122e-06, "loss": 2.5, "step": 11044 }, { "epoch": 2.1, "learning_rate": 2.371115347950429e-06, "loss": 4.0787, "step": 11045 }, { "epoch": 2.1, "learning_rate": 2.371058150619638e-06, "loss": 3.0911, "step": 11046 }, { "epoch": 2.1, "learning_rate": 2.3710009532888465e-06, "loss": 1.9911, "step": 11047 }, { "epoch": 2.1, "learning_rate": 2.3709437559580553e-06, "loss": 2.659, "step": 11048 }, { "epoch": 2.1, "learning_rate": 2.370886558627264e-06, "loss": 2.7787, "step": 11049 }, { "epoch": 2.1, "learning_rate": 2.370829361296473e-06, "loss": 2.7203, "step": 11050 }, { "epoch": 2.1, "learning_rate": 2.3707721639656817e-06, "loss": 3.2452, "step": 11051 }, { "epoch": 2.11, "learning_rate": 2.3707149666348905e-06, "loss": 3.2093, "step": 11052 }, { "epoch": 2.11, "learning_rate": 2.3706577693040993e-06, "loss": 3.6205, "step": 11053 }, { "epoch": 2.11, "learning_rate": 2.3706005719733076e-06, "loss": 2.3354, "step": 11054 }, { "epoch": 2.11, "learning_rate": 2.370543374642517e-06, "loss": 2.7287, "step": 11055 }, { "epoch": 2.11, "learning_rate": 2.3704861773117256e-06, "loss": 3.6307, "step": 11056 }, { "epoch": 2.11, "learning_rate": 2.3704289799809344e-06, "loss": 2.8454, "step": 11057 }, { "epoch": 2.11, "learning_rate": 2.3703717826501432e-06, "loss": 3.6618, "step": 11058 }, { "epoch": 2.11, "learning_rate": 2.3703145853193516e-06, "loss": 1.7331, "step": 11059 }, { "epoch": 2.11, "learning_rate": 2.370257387988561e-06, "loss": 3.5102, "step": 11060 }, { "epoch": 2.11, "learning_rate": 2.370200190657769e-06, "loss": 2.9395, "step": 11061 }, { "epoch": 2.11, "learning_rate": 2.370142993326978e-06, "loss": 3.1709, "step": 11062 }, { "epoch": 2.11, "learning_rate": 2.370085795996187e-06, "loss": 2.4673, "step": 11063 }, { "epoch": 2.11, "learning_rate": 2.3700285986653956e-06, "loss": 2.3942, "step": 11064 }, { "epoch": 2.11, "learning_rate": 2.3699714013346044e-06, "loss": 1.9971, "step": 11065 }, { "epoch": 2.11, "learning_rate": 2.369914204003813e-06, "loss": 2.2486, "step": 11066 }, { "epoch": 2.11, "learning_rate": 2.369857006673022e-06, "loss": 3.6978, "step": 11067 }, { "epoch": 2.11, "learning_rate": 2.3697998093422307e-06, "loss": 1.7496, "step": 11068 }, { "epoch": 2.11, "learning_rate": 2.3697426120114395e-06, "loss": 2.774, "step": 11069 }, { "epoch": 2.11, "learning_rate": 2.3696854146806483e-06, "loss": 3.0202, "step": 11070 }, { "epoch": 2.11, "learning_rate": 2.3696282173498567e-06, "loss": 2.7668, "step": 11071 }, { "epoch": 2.11, "learning_rate": 2.369571020019066e-06, "loss": 2.1168, "step": 11072 }, { "epoch": 2.11, "learning_rate": 2.3695138226882747e-06, "loss": 2.2442, "step": 11073 }, { "epoch": 2.11, "learning_rate": 2.3694566253574835e-06, "loss": 3.2564, "step": 11074 }, { "epoch": 2.11, "learning_rate": 2.3693994280266923e-06, "loss": 2.8696, "step": 11075 }, { "epoch": 2.11, "learning_rate": 2.3693422306959007e-06, "loss": 2.8129, "step": 11076 }, { "epoch": 2.11, "learning_rate": 2.36928503336511e-06, "loss": 2.6057, "step": 11077 }, { "epoch": 2.11, "learning_rate": 2.3692278360343187e-06, "loss": 2.783, "step": 11078 }, { "epoch": 2.11, "learning_rate": 2.369170638703527e-06, "loss": 2.7831, "step": 11079 }, { "epoch": 2.11, "learning_rate": 2.3691134413727363e-06, "loss": 2.7368, "step": 11080 }, { "epoch": 2.11, "learning_rate": 2.3690562440419446e-06, "loss": 2.3239, "step": 11081 }, { "epoch": 2.11, "learning_rate": 2.3689990467111534e-06, "loss": 2.5195, "step": 11082 }, { "epoch": 2.11, "learning_rate": 2.3689418493803622e-06, "loss": 2.9941, "step": 11083 }, { "epoch": 2.11, "learning_rate": 2.368884652049571e-06, "loss": 2.9831, "step": 11084 }, { "epoch": 2.11, "learning_rate": 2.36882745471878e-06, "loss": 2.6764, "step": 11085 }, { "epoch": 2.11, "learning_rate": 2.3687702573879886e-06, "loss": 2.8368, "step": 11086 }, { "epoch": 2.11, "learning_rate": 2.3687130600571974e-06, "loss": 2.6946, "step": 11087 }, { "epoch": 2.11, "learning_rate": 2.368655862726406e-06, "loss": 2.4622, "step": 11088 }, { "epoch": 2.11, "learning_rate": 2.368598665395615e-06, "loss": 3.7214, "step": 11089 }, { "epoch": 2.11, "learning_rate": 2.3685414680648238e-06, "loss": 2.5075, "step": 11090 }, { "epoch": 2.11, "learning_rate": 2.3684842707340326e-06, "loss": 3.2907, "step": 11091 }, { "epoch": 2.11, "learning_rate": 2.3684270734032414e-06, "loss": 3.2577, "step": 11092 }, { "epoch": 2.11, "learning_rate": 2.3683698760724497e-06, "loss": 2.8652, "step": 11093 }, { "epoch": 2.11, "learning_rate": 2.368312678741659e-06, "loss": 2.6481, "step": 11094 }, { "epoch": 2.11, "learning_rate": 2.3682554814108678e-06, "loss": 2.1841, "step": 11095 }, { "epoch": 2.11, "learning_rate": 2.368198284080076e-06, "loss": 3.2114, "step": 11096 }, { "epoch": 2.11, "learning_rate": 2.3681410867492853e-06, "loss": 2.9633, "step": 11097 }, { "epoch": 2.11, "learning_rate": 2.3680838894184937e-06, "loss": 2.7582, "step": 11098 }, { "epoch": 2.11, "learning_rate": 2.3680266920877025e-06, "loss": 3.3613, "step": 11099 }, { "epoch": 2.11, "learning_rate": 2.3679694947569113e-06, "loss": 2.4396, "step": 11100 }, { "epoch": 2.11, "learning_rate": 2.36791229742612e-06, "loss": 3.2967, "step": 11101 }, { "epoch": 2.11, "learning_rate": 2.367855100095329e-06, "loss": 2.4905, "step": 11102 }, { "epoch": 2.11, "learning_rate": 2.3677979027645377e-06, "loss": 3.1502, "step": 11103 }, { "epoch": 2.12, "learning_rate": 2.3677407054337465e-06, "loss": 3.5186, "step": 11104 }, { "epoch": 2.12, "learning_rate": 2.3676835081029553e-06, "loss": 3.0577, "step": 11105 }, { "epoch": 2.12, "learning_rate": 2.367626310772164e-06, "loss": 2.3043, "step": 11106 }, { "epoch": 2.12, "learning_rate": 2.367569113441373e-06, "loss": 2.5705, "step": 11107 }, { "epoch": 2.12, "learning_rate": 2.3675119161105817e-06, "loss": 2.6152, "step": 11108 }, { "epoch": 2.12, "learning_rate": 2.3674547187797904e-06, "loss": 3.0805, "step": 11109 }, { "epoch": 2.12, "learning_rate": 2.367397521448999e-06, "loss": 1.5298, "step": 11110 }, { "epoch": 2.12, "learning_rate": 2.367340324118208e-06, "loss": 2.7323, "step": 11111 }, { "epoch": 2.12, "learning_rate": 2.367283126787417e-06, "loss": 2.7726, "step": 11112 }, { "epoch": 2.12, "learning_rate": 2.367225929456625e-06, "loss": 3.1565, "step": 11113 }, { "epoch": 2.12, "learning_rate": 2.3671687321258344e-06, "loss": 3.5009, "step": 11114 }, { "epoch": 2.12, "learning_rate": 2.3671115347950428e-06, "loss": 3.3601, "step": 11115 }, { "epoch": 2.12, "learning_rate": 2.3670543374642516e-06, "loss": 2.9937, "step": 11116 }, { "epoch": 2.12, "learning_rate": 2.3669971401334604e-06, "loss": 3.6224, "step": 11117 }, { "epoch": 2.12, "learning_rate": 2.366939942802669e-06, "loss": 2.994, "step": 11118 }, { "epoch": 2.12, "learning_rate": 2.366882745471878e-06, "loss": 3.2333, "step": 11119 }, { "epoch": 2.12, "learning_rate": 2.3668255481410868e-06, "loss": 2.9358, "step": 11120 }, { "epoch": 2.12, "learning_rate": 2.3667683508102955e-06, "loss": 2.0961, "step": 11121 }, { "epoch": 2.12, "learning_rate": 2.3667111534795043e-06, "loss": 2.2869, "step": 11122 }, { "epoch": 2.12, "learning_rate": 2.366653956148713e-06, "loss": 2.3982, "step": 11123 }, { "epoch": 2.12, "learning_rate": 2.366596758817922e-06, "loss": 2.8924, "step": 11124 }, { "epoch": 2.12, "learning_rate": 2.3665395614871307e-06, "loss": 2.6419, "step": 11125 }, { "epoch": 2.12, "learning_rate": 2.3664823641563395e-06, "loss": 2.6375, "step": 11126 }, { "epoch": 2.12, "learning_rate": 2.3664251668255483e-06, "loss": 3.3665, "step": 11127 }, { "epoch": 2.12, "learning_rate": 2.366367969494757e-06, "loss": 3.6927, "step": 11128 }, { "epoch": 2.12, "learning_rate": 2.366310772163966e-06, "loss": 3.2223, "step": 11129 }, { "epoch": 2.12, "learning_rate": 2.3662535748331743e-06, "loss": 2.9793, "step": 11130 }, { "epoch": 2.12, "learning_rate": 2.3661963775023835e-06, "loss": 3.2672, "step": 11131 }, { "epoch": 2.12, "learning_rate": 2.366139180171592e-06, "loss": 2.2959, "step": 11132 }, { "epoch": 2.12, "learning_rate": 2.3660819828408006e-06, "loss": 3.0719, "step": 11133 }, { "epoch": 2.12, "learning_rate": 2.36602478551001e-06, "loss": 2.6545, "step": 11134 }, { "epoch": 2.12, "learning_rate": 2.3659675881792182e-06, "loss": 3.0246, "step": 11135 }, { "epoch": 2.12, "learning_rate": 2.365910390848427e-06, "loss": 2.6682, "step": 11136 }, { "epoch": 2.12, "learning_rate": 2.365853193517636e-06, "loss": 1.8833, "step": 11137 }, { "epoch": 2.12, "learning_rate": 2.3657959961868446e-06, "loss": 2.8938, "step": 11138 }, { "epoch": 2.12, "learning_rate": 2.3657387988560534e-06, "loss": 2.3437, "step": 11139 }, { "epoch": 2.12, "learning_rate": 2.365681601525262e-06, "loss": 3.3025, "step": 11140 }, { "epoch": 2.12, "learning_rate": 2.365624404194471e-06, "loss": 2.2338, "step": 11141 }, { "epoch": 2.12, "learning_rate": 2.36556720686368e-06, "loss": 2.6275, "step": 11142 }, { "epoch": 2.12, "learning_rate": 2.3655100095328886e-06, "loss": 2.3797, "step": 11143 }, { "epoch": 2.12, "learning_rate": 2.3654528122020974e-06, "loss": 3.3491, "step": 11144 }, { "epoch": 2.12, "learning_rate": 2.365395614871306e-06, "loss": 3.1277, "step": 11145 }, { "epoch": 2.12, "learning_rate": 2.365338417540515e-06, "loss": 4.0066, "step": 11146 }, { "epoch": 2.12, "learning_rate": 2.3652812202097233e-06, "loss": 1.8231, "step": 11147 }, { "epoch": 2.12, "learning_rate": 2.3652240228789326e-06, "loss": 2.5092, "step": 11148 }, { "epoch": 2.12, "learning_rate": 2.365166825548141e-06, "loss": 2.1563, "step": 11149 }, { "epoch": 2.12, "learning_rate": 2.3651096282173497e-06, "loss": 2.626, "step": 11150 }, { "epoch": 2.12, "learning_rate": 2.365052430886559e-06, "loss": 3.162, "step": 11151 }, { "epoch": 2.12, "learning_rate": 2.3649952335557673e-06, "loss": 3.0748, "step": 11152 }, { "epoch": 2.12, "learning_rate": 2.364938036224976e-06, "loss": 3.2305, "step": 11153 }, { "epoch": 2.12, "learning_rate": 2.364880838894185e-06, "loss": 2.6951, "step": 11154 }, { "epoch": 2.12, "learning_rate": 2.3648236415633937e-06, "loss": 2.2644, "step": 11155 }, { "epoch": 2.12, "learning_rate": 2.3647664442326025e-06, "loss": 2.9569, "step": 11156 }, { "epoch": 2.13, "learning_rate": 2.3647092469018113e-06, "loss": 2.6582, "step": 11157 }, { "epoch": 2.13, "learning_rate": 2.36465204957102e-06, "loss": 2.8666, "step": 11158 }, { "epoch": 2.13, "learning_rate": 2.364594852240229e-06, "loss": 3.3027, "step": 11159 }, { "epoch": 2.13, "learning_rate": 2.3645376549094377e-06, "loss": 2.3015, "step": 11160 }, { "epoch": 2.13, "learning_rate": 2.3644804575786465e-06, "loss": 3.2565, "step": 11161 }, { "epoch": 2.13, "learning_rate": 2.3644232602478552e-06, "loss": 2.5763, "step": 11162 }, { "epoch": 2.13, "learning_rate": 2.364366062917064e-06, "loss": 2.2245, "step": 11163 }, { "epoch": 2.13, "learning_rate": 2.3643088655862724e-06, "loss": 2.7709, "step": 11164 }, { "epoch": 2.13, "learning_rate": 2.3642516682554816e-06, "loss": 2.2848, "step": 11165 }, { "epoch": 2.13, "learning_rate": 2.3641944709246904e-06, "loss": 2.5117, "step": 11166 }, { "epoch": 2.13, "learning_rate": 2.364137273593899e-06, "loss": 3.7101, "step": 11167 }, { "epoch": 2.13, "learning_rate": 2.364080076263108e-06, "loss": 2.8364, "step": 11168 }, { "epoch": 2.13, "learning_rate": 2.3640228789323164e-06, "loss": 2.743, "step": 11169 }, { "epoch": 2.13, "learning_rate": 2.3639656816015256e-06, "loss": 3.0706, "step": 11170 }, { "epoch": 2.13, "learning_rate": 2.363908484270734e-06, "loss": 2.5769, "step": 11171 }, { "epoch": 2.13, "learning_rate": 2.3638512869399428e-06, "loss": 2.3626, "step": 11172 }, { "epoch": 2.13, "learning_rate": 2.363794089609152e-06, "loss": 3.4453, "step": 11173 }, { "epoch": 2.13, "learning_rate": 2.3637368922783603e-06, "loss": 2.1814, "step": 11174 }, { "epoch": 2.13, "learning_rate": 2.363679694947569e-06, "loss": 2.6635, "step": 11175 }, { "epoch": 2.13, "learning_rate": 2.363622497616778e-06, "loss": 2.6005, "step": 11176 }, { "epoch": 2.13, "learning_rate": 2.3635653002859867e-06, "loss": 2.6459, "step": 11177 }, { "epoch": 2.13, "learning_rate": 2.3635081029551955e-06, "loss": 2.628, "step": 11178 }, { "epoch": 2.13, "learning_rate": 2.3634509056244043e-06, "loss": 3.3462, "step": 11179 }, { "epoch": 2.13, "learning_rate": 2.363393708293613e-06, "loss": 2.9267, "step": 11180 }, { "epoch": 2.13, "learning_rate": 2.3633365109628215e-06, "loss": 2.3992, "step": 11181 }, { "epoch": 2.13, "learning_rate": 2.3632793136320307e-06, "loss": 2.5949, "step": 11182 }, { "epoch": 2.13, "learning_rate": 2.3632221163012395e-06, "loss": 2.6174, "step": 11183 }, { "epoch": 2.13, "learning_rate": 2.363164918970448e-06, "loss": 2.4621, "step": 11184 }, { "epoch": 2.13, "learning_rate": 2.363107721639657e-06, "loss": 2.847, "step": 11185 }, { "epoch": 2.13, "learning_rate": 2.3630505243088655e-06, "loss": 3.7161, "step": 11186 }, { "epoch": 2.13, "learning_rate": 2.3629933269780747e-06, "loss": 2.3727, "step": 11187 }, { "epoch": 2.13, "learning_rate": 2.362936129647283e-06, "loss": 2.358, "step": 11188 }, { "epoch": 2.13, "learning_rate": 2.362878932316492e-06, "loss": 3.1128, "step": 11189 }, { "epoch": 2.13, "learning_rate": 2.362821734985701e-06, "loss": 2.4029, "step": 11190 }, { "epoch": 2.13, "learning_rate": 2.3627645376549094e-06, "loss": 2.4228, "step": 11191 }, { "epoch": 2.13, "learning_rate": 2.3627073403241182e-06, "loss": 2.4433, "step": 11192 }, { "epoch": 2.13, "learning_rate": 2.362650142993327e-06, "loss": 3.0573, "step": 11193 }, { "epoch": 2.13, "learning_rate": 2.362592945662536e-06, "loss": 2.4726, "step": 11194 }, { "epoch": 2.13, "learning_rate": 2.3625357483317446e-06, "loss": 2.4553, "step": 11195 }, { "epoch": 2.13, "learning_rate": 2.3624785510009534e-06, "loss": 2.0155, "step": 11196 }, { "epoch": 2.13, "learning_rate": 2.362421353670162e-06, "loss": 3.0701, "step": 11197 }, { "epoch": 2.13, "learning_rate": 2.3623641563393706e-06, "loss": 2.6936, "step": 11198 }, { "epoch": 2.13, "learning_rate": 2.3623069590085798e-06, "loss": 2.6345, "step": 11199 }, { "epoch": 2.13, "learning_rate": 2.3622497616777886e-06, "loss": 2.5501, "step": 11200 }, { "epoch": 2.13, "learning_rate": 2.362192564346997e-06, "loss": 2.7043, "step": 11201 }, { "epoch": 2.13, "learning_rate": 2.362135367016206e-06, "loss": 2.4468, "step": 11202 }, { "epoch": 2.13, "learning_rate": 2.3620781696854145e-06, "loss": 1.9832, "step": 11203 }, { "epoch": 2.13, "learning_rate": 2.3620209723546237e-06, "loss": 2.5629, "step": 11204 }, { "epoch": 2.13, "learning_rate": 2.361963775023832e-06, "loss": 2.8979, "step": 11205 }, { "epoch": 2.13, "learning_rate": 2.361906577693041e-06, "loss": 2.913, "step": 11206 }, { "epoch": 2.13, "learning_rate": 2.36184938036225e-06, "loss": 2.6976, "step": 11207 }, { "epoch": 2.13, "learning_rate": 2.3617921830314585e-06, "loss": 2.8859, "step": 11208 }, { "epoch": 2.14, "learning_rate": 2.3617349857006673e-06, "loss": 3.1773, "step": 11209 }, { "epoch": 2.14, "learning_rate": 2.361677788369876e-06, "loss": 2.9217, "step": 11210 }, { "epoch": 2.14, "learning_rate": 2.361620591039085e-06, "loss": 2.7309, "step": 11211 }, { "epoch": 2.14, "learning_rate": 2.3615633937082937e-06, "loss": 2.896, "step": 11212 }, { "epoch": 2.14, "learning_rate": 2.3615061963775025e-06, "loss": 2.2035, "step": 11213 }, { "epoch": 2.14, "learning_rate": 2.3614489990467113e-06, "loss": 2.4713, "step": 11214 }, { "epoch": 2.14, "learning_rate": 2.36139180171592e-06, "loss": 2.4015, "step": 11215 }, { "epoch": 2.14, "learning_rate": 2.361334604385129e-06, "loss": 3.2034, "step": 11216 }, { "epoch": 2.14, "learning_rate": 2.3612774070543376e-06, "loss": 2.9467, "step": 11217 }, { "epoch": 2.14, "learning_rate": 2.361220209723546e-06, "loss": 2.5295, "step": 11218 }, { "epoch": 2.14, "learning_rate": 2.3611630123927552e-06, "loss": 2.7846, "step": 11219 }, { "epoch": 2.14, "learning_rate": 2.3611058150619636e-06, "loss": 2.1803, "step": 11220 }, { "epoch": 2.14, "learning_rate": 2.361048617731173e-06, "loss": 2.4801, "step": 11221 }, { "epoch": 2.14, "learning_rate": 2.3609914204003816e-06, "loss": 2.9901, "step": 11222 }, { "epoch": 2.14, "learning_rate": 2.36093422306959e-06, "loss": 2.5147, "step": 11223 }, { "epoch": 2.14, "learning_rate": 2.360877025738799e-06, "loss": 2.8151, "step": 11224 }, { "epoch": 2.14, "learning_rate": 2.3608198284080076e-06, "loss": 3.1815, "step": 11225 }, { "epoch": 2.14, "learning_rate": 2.3607626310772164e-06, "loss": 3.7078, "step": 11226 }, { "epoch": 2.14, "learning_rate": 2.360705433746425e-06, "loss": 2.3268, "step": 11227 }, { "epoch": 2.14, "learning_rate": 2.360648236415634e-06, "loss": 3.3826, "step": 11228 }, { "epoch": 2.14, "learning_rate": 2.3605910390848427e-06, "loss": 2.5964, "step": 11229 }, { "epoch": 2.14, "learning_rate": 2.3605338417540515e-06, "loss": 2.8906, "step": 11230 }, { "epoch": 2.14, "learning_rate": 2.3604766444232603e-06, "loss": 2.7115, "step": 11231 }, { "epoch": 2.14, "learning_rate": 2.360419447092469e-06, "loss": 3.1563, "step": 11232 }, { "epoch": 2.14, "learning_rate": 2.360362249761678e-06, "loss": 2.8325, "step": 11233 }, { "epoch": 2.14, "learning_rate": 2.3603050524308867e-06, "loss": 2.943, "step": 11234 }, { "epoch": 2.14, "learning_rate": 2.3602478551000955e-06, "loss": 3.6613, "step": 11235 }, { "epoch": 2.14, "learning_rate": 2.3601906577693043e-06, "loss": 2.8868, "step": 11236 }, { "epoch": 2.14, "learning_rate": 2.3601334604385127e-06, "loss": 2.486, "step": 11237 }, { "epoch": 2.14, "learning_rate": 2.360076263107722e-06, "loss": 2.2858, "step": 11238 }, { "epoch": 2.14, "learning_rate": 2.3600190657769307e-06, "loss": 2.5035, "step": 11239 }, { "epoch": 2.14, "learning_rate": 2.359961868446139e-06, "loss": 3.2875, "step": 11240 }, { "epoch": 2.14, "learning_rate": 2.3599046711153483e-06, "loss": 2.6582, "step": 11241 }, { "epoch": 2.14, "learning_rate": 2.3598474737845566e-06, "loss": 3.0344, "step": 11242 }, { "epoch": 2.14, "learning_rate": 2.3597902764537654e-06, "loss": 3.1199, "step": 11243 }, { "epoch": 2.14, "learning_rate": 2.3597330791229742e-06, "loss": 2.7181, "step": 11244 }, { "epoch": 2.14, "learning_rate": 2.359675881792183e-06, "loss": 2.1486, "step": 11245 }, { "epoch": 2.14, "learning_rate": 2.359618684461392e-06, "loss": 3.1315, "step": 11246 }, { "epoch": 2.14, "learning_rate": 2.3595614871306006e-06, "loss": 2.7244, "step": 11247 }, { "epoch": 2.14, "learning_rate": 2.3595042897998094e-06, "loss": 3.3511, "step": 11248 }, { "epoch": 2.14, "learning_rate": 2.359447092469018e-06, "loss": 3.307, "step": 11249 }, { "epoch": 2.14, "learning_rate": 2.359389895138227e-06, "loss": 3.2743, "step": 11250 }, { "epoch": 2.14, "learning_rate": 2.3593326978074358e-06, "loss": 3.3933, "step": 11251 }, { "epoch": 2.14, "learning_rate": 2.3592755004766446e-06, "loss": 3.3525, "step": 11252 }, { "epoch": 2.14, "learning_rate": 2.3592183031458534e-06, "loss": 2.7336, "step": 11253 }, { "epoch": 2.14, "learning_rate": 2.3591611058150617e-06, "loss": 2.2425, "step": 11254 }, { "epoch": 2.14, "learning_rate": 2.359103908484271e-06, "loss": 2.5961, "step": 11255 }, { "epoch": 2.14, "learning_rate": 2.3590467111534797e-06, "loss": 2.6929, "step": 11256 }, { "epoch": 2.14, "learning_rate": 2.358989513822688e-06, "loss": 2.6174, "step": 11257 }, { "epoch": 2.14, "learning_rate": 2.3589323164918973e-06, "loss": 3.5545, "step": 11258 }, { "epoch": 2.14, "learning_rate": 2.3588751191611057e-06, "loss": 2.7902, "step": 11259 }, { "epoch": 2.14, "learning_rate": 2.3588179218303145e-06, "loss": 2.3178, "step": 11260 }, { "epoch": 2.14, "learning_rate": 2.3587607244995237e-06, "loss": 2.5313, "step": 11261 }, { "epoch": 2.15, "learning_rate": 2.358703527168732e-06, "loss": 3.0589, "step": 11262 }, { "epoch": 2.15, "learning_rate": 2.358646329837941e-06, "loss": 3.109, "step": 11263 }, { "epoch": 2.15, "learning_rate": 2.3585891325071497e-06, "loss": 3.528, "step": 11264 }, { "epoch": 2.15, "learning_rate": 2.3585319351763585e-06, "loss": 2.6407, "step": 11265 }, { "epoch": 2.15, "learning_rate": 2.3584747378455673e-06, "loss": 2.6496, "step": 11266 }, { "epoch": 2.15, "learning_rate": 2.358417540514776e-06, "loss": 3.3361, "step": 11267 }, { "epoch": 2.15, "learning_rate": 2.358360343183985e-06, "loss": 3.1701, "step": 11268 }, { "epoch": 2.15, "learning_rate": 2.3583031458531936e-06, "loss": 2.2027, "step": 11269 }, { "epoch": 2.15, "learning_rate": 2.3582459485224024e-06, "loss": 3.4833, "step": 11270 }, { "epoch": 2.15, "learning_rate": 2.3581887511916112e-06, "loss": 2.7293, "step": 11271 }, { "epoch": 2.15, "learning_rate": 2.35813155386082e-06, "loss": 3.0911, "step": 11272 }, { "epoch": 2.15, "learning_rate": 2.358074356530029e-06, "loss": 2.4221, "step": 11273 }, { "epoch": 2.15, "learning_rate": 2.358017159199237e-06, "loss": 2.7001, "step": 11274 }, { "epoch": 2.15, "learning_rate": 2.3579599618684464e-06, "loss": 3.1046, "step": 11275 }, { "epoch": 2.15, "learning_rate": 2.3579027645376548e-06, "loss": 3.672, "step": 11276 }, { "epoch": 2.15, "learning_rate": 2.3578455672068636e-06, "loss": 2.3711, "step": 11277 }, { "epoch": 2.15, "learning_rate": 2.357788369876073e-06, "loss": 2.5334, "step": 11278 }, { "epoch": 2.15, "learning_rate": 2.357731172545281e-06, "loss": 2.4548, "step": 11279 }, { "epoch": 2.15, "learning_rate": 2.35767397521449e-06, "loss": 2.179, "step": 11280 }, { "epoch": 2.15, "learning_rate": 2.3576167778836987e-06, "loss": 3.6334, "step": 11281 }, { "epoch": 2.15, "learning_rate": 2.3575595805529075e-06, "loss": 2.8666, "step": 11282 }, { "epoch": 2.15, "learning_rate": 2.3575023832221163e-06, "loss": 3.4479, "step": 11283 }, { "epoch": 2.15, "learning_rate": 2.357445185891325e-06, "loss": 3.0509, "step": 11284 }, { "epoch": 2.15, "learning_rate": 2.357387988560534e-06, "loss": 2.1158, "step": 11285 }, { "epoch": 2.15, "learning_rate": 2.3573307912297427e-06, "loss": 3.463, "step": 11286 }, { "epoch": 2.15, "learning_rate": 2.3572735938989515e-06, "loss": 2.7979, "step": 11287 }, { "epoch": 2.15, "learning_rate": 2.3572163965681603e-06, "loss": 3.1621, "step": 11288 }, { "epoch": 2.15, "learning_rate": 2.357159199237369e-06, "loss": 3.977, "step": 11289 }, { "epoch": 2.15, "learning_rate": 2.357102001906578e-06, "loss": 2.3813, "step": 11290 }, { "epoch": 2.15, "learning_rate": 2.3570448045757863e-06, "loss": 2.7911, "step": 11291 }, { "epoch": 2.15, "learning_rate": 2.3569876072449955e-06, "loss": 2.6386, "step": 11292 }, { "epoch": 2.15, "learning_rate": 2.356930409914204e-06, "loss": 2.9071, "step": 11293 }, { "epoch": 2.15, "learning_rate": 2.3568732125834126e-06, "loss": 3.4467, "step": 11294 }, { "epoch": 2.15, "learning_rate": 2.356816015252622e-06, "loss": 2.3512, "step": 11295 }, { "epoch": 2.15, "learning_rate": 2.3567588179218302e-06, "loss": 3.3367, "step": 11296 }, { "epoch": 2.15, "learning_rate": 2.356701620591039e-06, "loss": 3.7182, "step": 11297 }, { "epoch": 2.15, "learning_rate": 2.356644423260248e-06, "loss": 2.5031, "step": 11298 }, { "epoch": 2.15, "learning_rate": 2.3565872259294566e-06, "loss": 2.9838, "step": 11299 }, { "epoch": 2.15, "learning_rate": 2.356530028598666e-06, "loss": 3.4527, "step": 11300 }, { "epoch": 2.15, "learning_rate": 2.356472831267874e-06, "loss": 2.3465, "step": 11301 }, { "epoch": 2.15, "learning_rate": 2.356415633937083e-06, "loss": 4.1643, "step": 11302 }, { "epoch": 2.15, "learning_rate": 2.3563584366062918e-06, "loss": 3.6266, "step": 11303 }, { "epoch": 2.15, "learning_rate": 2.3563012392755006e-06, "loss": 3.2543, "step": 11304 }, { "epoch": 2.15, "learning_rate": 2.3562440419447094e-06, "loss": 2.4022, "step": 11305 }, { "epoch": 2.15, "learning_rate": 2.356186844613918e-06, "loss": 2.6509, "step": 11306 }, { "epoch": 2.15, "learning_rate": 2.356129647283127e-06, "loss": 2.6316, "step": 11307 }, { "epoch": 2.15, "learning_rate": 2.3560724499523353e-06, "loss": 2.0734, "step": 11308 }, { "epoch": 2.15, "learning_rate": 2.3560152526215445e-06, "loss": 2.7705, "step": 11309 }, { "epoch": 2.15, "learning_rate": 2.3559580552907533e-06, "loss": 3.1043, "step": 11310 }, { "epoch": 2.15, "learning_rate": 2.3559008579599617e-06, "loss": 1.8783, "step": 11311 }, { "epoch": 2.15, "learning_rate": 2.355843660629171e-06, "loss": 2.44, "step": 11312 }, { "epoch": 2.15, "learning_rate": 2.3557864632983793e-06, "loss": 2.8944, "step": 11313 }, { "epoch": 2.16, "learning_rate": 2.355729265967588e-06, "loss": 3.5513, "step": 11314 }, { "epoch": 2.16, "learning_rate": 2.355672068636797e-06, "loss": 3.1334, "step": 11315 }, { "epoch": 2.16, "learning_rate": 2.3556148713060057e-06, "loss": 3.5957, "step": 11316 }, { "epoch": 2.16, "learning_rate": 2.355557673975215e-06, "loss": 3.5761, "step": 11317 }, { "epoch": 2.16, "learning_rate": 2.3555004766444233e-06, "loss": 2.922, "step": 11318 }, { "epoch": 2.16, "learning_rate": 2.355443279313632e-06, "loss": 3.2261, "step": 11319 }, { "epoch": 2.16, "learning_rate": 2.355386081982841e-06, "loss": 2.2613, "step": 11320 }, { "epoch": 2.16, "learning_rate": 2.3553288846520497e-06, "loss": 3.2688, "step": 11321 }, { "epoch": 2.16, "learning_rate": 2.3552716873212584e-06, "loss": 2.694, "step": 11322 }, { "epoch": 2.16, "learning_rate": 2.3552144899904672e-06, "loss": 1.9347, "step": 11323 }, { "epoch": 2.16, "learning_rate": 2.355157292659676e-06, "loss": 2.6836, "step": 11324 }, { "epoch": 2.16, "learning_rate": 2.3551000953288844e-06, "loss": 3.7734, "step": 11325 }, { "epoch": 2.16, "learning_rate": 2.3550428979980936e-06, "loss": 2.7479, "step": 11326 }, { "epoch": 2.16, "learning_rate": 2.3549857006673024e-06, "loss": 2.3892, "step": 11327 }, { "epoch": 2.16, "learning_rate": 2.3549285033365108e-06, "loss": 2.7277, "step": 11328 }, { "epoch": 2.16, "learning_rate": 2.35487130600572e-06, "loss": 2.1087, "step": 11329 }, { "epoch": 2.16, "learning_rate": 2.3548141086749284e-06, "loss": 4.0226, "step": 11330 }, { "epoch": 2.16, "learning_rate": 2.354756911344137e-06, "loss": 2.829, "step": 11331 }, { "epoch": 2.16, "learning_rate": 2.354699714013346e-06, "loss": 3.0657, "step": 11332 }, { "epoch": 2.16, "learning_rate": 2.3546425166825548e-06, "loss": 1.9802, "step": 11333 }, { "epoch": 2.16, "learning_rate": 2.354585319351764e-06, "loss": 3.0652, "step": 11334 }, { "epoch": 2.16, "learning_rate": 2.3545281220209723e-06, "loss": 2.6695, "step": 11335 }, { "epoch": 2.16, "learning_rate": 2.354470924690181e-06, "loss": 2.1843, "step": 11336 }, { "epoch": 2.16, "learning_rate": 2.35441372735939e-06, "loss": 2.7436, "step": 11337 }, { "epoch": 2.16, "learning_rate": 2.3543565300285987e-06, "loss": 3.4567, "step": 11338 }, { "epoch": 2.16, "learning_rate": 2.3542993326978075e-06, "loss": 2.1045, "step": 11339 }, { "epoch": 2.16, "learning_rate": 2.3542421353670163e-06, "loss": 3.1733, "step": 11340 }, { "epoch": 2.16, "learning_rate": 2.354184938036225e-06, "loss": 3.3557, "step": 11341 }, { "epoch": 2.16, "learning_rate": 2.3541277407054335e-06, "loss": 2.7029, "step": 11342 }, { "epoch": 2.16, "learning_rate": 2.3540705433746427e-06, "loss": 2.5609, "step": 11343 }, { "epoch": 2.16, "learning_rate": 2.3540133460438515e-06, "loss": 2.6675, "step": 11344 }, { "epoch": 2.16, "learning_rate": 2.35395614871306e-06, "loss": 2.4419, "step": 11345 }, { "epoch": 2.16, "learning_rate": 2.353898951382269e-06, "loss": 2.3358, "step": 11346 }, { "epoch": 2.16, "learning_rate": 2.3538417540514774e-06, "loss": 2.3347, "step": 11347 }, { "epoch": 2.16, "learning_rate": 2.3537845567206862e-06, "loss": 2.2092, "step": 11348 }, { "epoch": 2.16, "learning_rate": 2.3537273593898955e-06, "loss": 2.3289, "step": 11349 }, { "epoch": 2.16, "learning_rate": 2.353670162059104e-06, "loss": 3.2828, "step": 11350 }, { "epoch": 2.16, "learning_rate": 2.353612964728313e-06, "loss": 3.0021, "step": 11351 }, { "epoch": 2.16, "learning_rate": 2.3535557673975214e-06, "loss": 2.0288, "step": 11352 }, { "epoch": 2.16, "learning_rate": 2.35349857006673e-06, "loss": 2.4695, "step": 11353 }, { "epoch": 2.16, "learning_rate": 2.353441372735939e-06, "loss": 2.7494, "step": 11354 }, { "epoch": 2.16, "learning_rate": 2.353384175405148e-06, "loss": 2.0129, "step": 11355 }, { "epoch": 2.16, "learning_rate": 2.3533269780743566e-06, "loss": 2.4431, "step": 11356 }, { "epoch": 2.16, "learning_rate": 2.3532697807435654e-06, "loss": 2.8217, "step": 11357 }, { "epoch": 2.16, "learning_rate": 2.353212583412774e-06, "loss": 2.251, "step": 11358 }, { "epoch": 2.16, "learning_rate": 2.353155386081983e-06, "loss": 2.942, "step": 11359 }, { "epoch": 2.16, "learning_rate": 2.3530981887511918e-06, "loss": 3.6523, "step": 11360 }, { "epoch": 2.16, "learning_rate": 2.3530409914204006e-06, "loss": 2.2011, "step": 11361 }, { "epoch": 2.16, "learning_rate": 2.352983794089609e-06, "loss": 2.6329, "step": 11362 }, { "epoch": 2.16, "learning_rate": 2.352926596758818e-06, "loss": 3.6284, "step": 11363 }, { "epoch": 2.16, "learning_rate": 2.3528693994280265e-06, "loss": 2.6893, "step": 11364 }, { "epoch": 2.16, "learning_rate": 2.3528122020972357e-06, "loss": 2.7341, "step": 11365 }, { "epoch": 2.16, "learning_rate": 2.3527550047664445e-06, "loss": 3.0561, "step": 11366 }, { "epoch": 2.17, "learning_rate": 2.352697807435653e-06, "loss": 2.9746, "step": 11367 }, { "epoch": 2.17, "learning_rate": 2.352640610104862e-06, "loss": 3.2947, "step": 11368 }, { "epoch": 2.17, "learning_rate": 2.3525834127740705e-06, "loss": 2.6188, "step": 11369 }, { "epoch": 2.17, "learning_rate": 2.3525262154432793e-06, "loss": 2.9853, "step": 11370 }, { "epoch": 2.17, "learning_rate": 2.352469018112488e-06, "loss": 2.7882, "step": 11371 }, { "epoch": 2.17, "learning_rate": 2.352411820781697e-06, "loss": 3.807, "step": 11372 }, { "epoch": 2.17, "learning_rate": 2.3523546234509057e-06, "loss": 3.3157, "step": 11373 }, { "epoch": 2.17, "learning_rate": 2.3522974261201145e-06, "loss": 2.9935, "step": 11374 }, { "epoch": 2.17, "learning_rate": 2.3522402287893232e-06, "loss": 2.7755, "step": 11375 }, { "epoch": 2.17, "learning_rate": 2.352183031458532e-06, "loss": 2.6451, "step": 11376 }, { "epoch": 2.17, "learning_rate": 2.352125834127741e-06, "loss": 2.9176, "step": 11377 }, { "epoch": 2.17, "learning_rate": 2.3520686367969496e-06, "loss": 2.1379, "step": 11378 }, { "epoch": 2.17, "learning_rate": 2.352011439466158e-06, "loss": 2.8928, "step": 11379 }, { "epoch": 2.17, "learning_rate": 2.3519542421353672e-06, "loss": 2.9418, "step": 11380 }, { "epoch": 2.17, "learning_rate": 2.3518970448045756e-06, "loss": 2.7559, "step": 11381 }, { "epoch": 2.17, "learning_rate": 2.351839847473785e-06, "loss": 3.0042, "step": 11382 }, { "epoch": 2.17, "learning_rate": 2.3517826501429936e-06, "loss": 2.9418, "step": 11383 }, { "epoch": 2.17, "learning_rate": 2.351725452812202e-06, "loss": 3.2871, "step": 11384 }, { "epoch": 2.17, "learning_rate": 2.351668255481411e-06, "loss": 2.7881, "step": 11385 }, { "epoch": 2.17, "learning_rate": 2.3516110581506196e-06, "loss": 3.0799, "step": 11386 }, { "epoch": 2.17, "learning_rate": 2.3515538608198283e-06, "loss": 2.1632, "step": 11387 }, { "epoch": 2.17, "learning_rate": 2.3514966634890376e-06, "loss": 2.6744, "step": 11388 }, { "epoch": 2.17, "learning_rate": 2.351439466158246e-06, "loss": 3.3618, "step": 11389 }, { "epoch": 2.17, "learning_rate": 2.3513822688274547e-06, "loss": 2.2718, "step": 11390 }, { "epoch": 2.17, "learning_rate": 2.3513250714966635e-06, "loss": 3.1994, "step": 11391 }, { "epoch": 2.17, "learning_rate": 2.3512678741658723e-06, "loss": 1.7556, "step": 11392 }, { "epoch": 2.17, "learning_rate": 2.351210676835081e-06, "loss": 3.5527, "step": 11393 }, { "epoch": 2.17, "learning_rate": 2.35115347950429e-06, "loss": 1.6583, "step": 11394 }, { "epoch": 2.17, "learning_rate": 2.3510962821734987e-06, "loss": 3.0803, "step": 11395 }, { "epoch": 2.17, "learning_rate": 2.351039084842707e-06, "loss": 3.0688, "step": 11396 }, { "epoch": 2.17, "learning_rate": 2.3509818875119163e-06, "loss": 2.6605, "step": 11397 }, { "epoch": 2.17, "learning_rate": 2.350924690181125e-06, "loss": 3.7018, "step": 11398 }, { "epoch": 2.17, "learning_rate": 2.350867492850334e-06, "loss": 3.5147, "step": 11399 }, { "epoch": 2.17, "learning_rate": 2.3508102955195427e-06, "loss": 3.1531, "step": 11400 }, { "epoch": 2.17, "learning_rate": 2.350753098188751e-06, "loss": 2.2366, "step": 11401 }, { "epoch": 2.17, "learning_rate": 2.3506959008579603e-06, "loss": 3.2422, "step": 11402 }, { "epoch": 2.17, "learning_rate": 2.3506387035271686e-06, "loss": 2.1533, "step": 11403 }, { "epoch": 2.17, "learning_rate": 2.3505815061963774e-06, "loss": 2.4406, "step": 11404 }, { "epoch": 2.17, "learning_rate": 2.3505243088655866e-06, "loss": 2.6982, "step": 11405 }, { "epoch": 2.17, "learning_rate": 2.350467111534795e-06, "loss": 2.7324, "step": 11406 }, { "epoch": 2.17, "learning_rate": 2.350409914204004e-06, "loss": 1.7521, "step": 11407 }, { "epoch": 2.17, "learning_rate": 2.3503527168732126e-06, "loss": 3.6217, "step": 11408 }, { "epoch": 2.17, "learning_rate": 2.3502955195424214e-06, "loss": 2.4901, "step": 11409 }, { "epoch": 2.17, "learning_rate": 2.35023832221163e-06, "loss": 2.6393, "step": 11410 }, { "epoch": 2.17, "learning_rate": 2.350181124880839e-06, "loss": 3.5278, "step": 11411 }, { "epoch": 2.17, "learning_rate": 2.3501239275500478e-06, "loss": 2.9656, "step": 11412 }, { "epoch": 2.17, "learning_rate": 2.350066730219256e-06, "loss": 2.7999, "step": 11413 }, { "epoch": 2.17, "learning_rate": 2.3500095328884654e-06, "loss": 3.1997, "step": 11414 }, { "epoch": 2.17, "learning_rate": 2.349952335557674e-06, "loss": 2.1007, "step": 11415 }, { "epoch": 2.17, "learning_rate": 2.349895138226883e-06, "loss": 2.616, "step": 11416 }, { "epoch": 2.17, "learning_rate": 2.3498379408960917e-06, "loss": 3.1242, "step": 11417 }, { "epoch": 2.17, "learning_rate": 2.3497807435653e-06, "loss": 3.6877, "step": 11418 }, { "epoch": 2.18, "learning_rate": 2.3497235462345093e-06, "loss": 2.3961, "step": 11419 }, { "epoch": 2.18, "learning_rate": 2.3496663489037177e-06, "loss": 2.5246, "step": 11420 }, { "epoch": 2.18, "learning_rate": 2.3496091515729265e-06, "loss": 2.9397, "step": 11421 }, { "epoch": 2.18, "learning_rate": 2.3495519542421357e-06, "loss": 2.3942, "step": 11422 }, { "epoch": 2.18, "learning_rate": 2.349494756911344e-06, "loss": 2.8333, "step": 11423 }, { "epoch": 2.18, "learning_rate": 2.349437559580553e-06, "loss": 2.2136, "step": 11424 }, { "epoch": 2.18, "learning_rate": 2.3493803622497617e-06, "loss": 3.568, "step": 11425 }, { "epoch": 2.18, "learning_rate": 2.3493231649189705e-06, "loss": 3.3227, "step": 11426 }, { "epoch": 2.18, "learning_rate": 2.3492659675881793e-06, "loss": 3.1014, "step": 11427 }, { "epoch": 2.18, "learning_rate": 2.349208770257388e-06, "loss": 3.2708, "step": 11428 }, { "epoch": 2.18, "learning_rate": 2.349151572926597e-06, "loss": 1.7558, "step": 11429 }, { "epoch": 2.18, "learning_rate": 2.3490943755958056e-06, "loss": 2.928, "step": 11430 }, { "epoch": 2.18, "learning_rate": 2.3490371782650144e-06, "loss": 2.3828, "step": 11431 }, { "epoch": 2.18, "learning_rate": 2.3489799809342232e-06, "loss": 3.179, "step": 11432 }, { "epoch": 2.18, "learning_rate": 2.348922783603432e-06, "loss": 3.7049, "step": 11433 }, { "epoch": 2.18, "learning_rate": 2.348865586272641e-06, "loss": 2.5371, "step": 11434 }, { "epoch": 2.18, "learning_rate": 2.348808388941849e-06, "loss": 2.35, "step": 11435 }, { "epoch": 2.18, "learning_rate": 2.3487511916110584e-06, "loss": 3.0729, "step": 11436 }, { "epoch": 2.18, "learning_rate": 2.348693994280267e-06, "loss": 2.84, "step": 11437 }, { "epoch": 2.18, "learning_rate": 2.3486367969494756e-06, "loss": 2.7299, "step": 11438 }, { "epoch": 2.18, "learning_rate": 2.3485795996186848e-06, "loss": 2.2394, "step": 11439 }, { "epoch": 2.18, "learning_rate": 2.348522402287893e-06, "loss": 2.5974, "step": 11440 }, { "epoch": 2.18, "learning_rate": 2.348465204957102e-06, "loss": 3.2927, "step": 11441 }, { "epoch": 2.18, "learning_rate": 2.3484080076263107e-06, "loss": 3.2713, "step": 11442 }, { "epoch": 2.18, "learning_rate": 2.3483508102955195e-06, "loss": 2.788, "step": 11443 }, { "epoch": 2.18, "learning_rate": 2.3482936129647283e-06, "loss": 2.6067, "step": 11444 }, { "epoch": 2.18, "learning_rate": 2.348236415633937e-06, "loss": 3.0347, "step": 11445 }, { "epoch": 2.18, "learning_rate": 2.348179218303146e-06, "loss": 2.9469, "step": 11446 }, { "epoch": 2.18, "learning_rate": 2.3481220209723547e-06, "loss": 3.9501, "step": 11447 }, { "epoch": 2.18, "learning_rate": 2.3480648236415635e-06, "loss": 3.9564, "step": 11448 }, { "epoch": 2.18, "learning_rate": 2.3480076263107723e-06, "loss": 3.4422, "step": 11449 }, { "epoch": 2.18, "learning_rate": 2.347950428979981e-06, "loss": 1.8624, "step": 11450 }, { "epoch": 2.18, "learning_rate": 2.34789323164919e-06, "loss": 2.7627, "step": 11451 }, { "epoch": 2.18, "learning_rate": 2.3478360343183983e-06, "loss": 3.8414, "step": 11452 }, { "epoch": 2.18, "learning_rate": 2.3477788369876075e-06, "loss": 3.1463, "step": 11453 }, { "epoch": 2.18, "learning_rate": 2.3477216396568163e-06, "loss": 3.302, "step": 11454 }, { "epoch": 2.18, "learning_rate": 2.3476644423260246e-06, "loss": 3.232, "step": 11455 }, { "epoch": 2.18, "learning_rate": 2.347607244995234e-06, "loss": 2.4809, "step": 11456 }, { "epoch": 2.18, "learning_rate": 2.3475500476644422e-06, "loss": 2.4427, "step": 11457 }, { "epoch": 2.18, "learning_rate": 2.347492850333651e-06, "loss": 3.4924, "step": 11458 }, { "epoch": 2.18, "learning_rate": 2.34743565300286e-06, "loss": 2.54, "step": 11459 }, { "epoch": 2.18, "learning_rate": 2.3473784556720686e-06, "loss": 2.5536, "step": 11460 }, { "epoch": 2.18, "learning_rate": 2.3473212583412774e-06, "loss": 3.0346, "step": 11461 }, { "epoch": 2.18, "learning_rate": 2.347264061010486e-06, "loss": 2.4424, "step": 11462 }, { "epoch": 2.18, "learning_rate": 2.347206863679695e-06, "loss": 2.3754, "step": 11463 }, { "epoch": 2.18, "learning_rate": 2.3471496663489038e-06, "loss": 2.2647, "step": 11464 }, { "epoch": 2.18, "learning_rate": 2.3470924690181126e-06, "loss": 2.6317, "step": 11465 }, { "epoch": 2.18, "learning_rate": 2.3470352716873214e-06, "loss": 2.523, "step": 11466 }, { "epoch": 2.18, "learning_rate": 2.34697807435653e-06, "loss": 2.0908, "step": 11467 }, { "epoch": 2.18, "learning_rate": 2.346920877025739e-06, "loss": 2.7439, "step": 11468 }, { "epoch": 2.18, "learning_rate": 2.3468636796949473e-06, "loss": 2.4754, "step": 11469 }, { "epoch": 2.18, "learning_rate": 2.3468064823641565e-06, "loss": 2.7799, "step": 11470 }, { "epoch": 2.18, "learning_rate": 2.3467492850333653e-06, "loss": 2.9082, "step": 11471 }, { "epoch": 2.19, "learning_rate": 2.3466920877025737e-06, "loss": 2.1875, "step": 11472 }, { "epoch": 2.19, "learning_rate": 2.346634890371783e-06, "loss": 1.6217, "step": 11473 }, { "epoch": 2.19, "learning_rate": 2.3465776930409913e-06, "loss": 3.0065, "step": 11474 }, { "epoch": 2.19, "learning_rate": 2.3465204957102e-06, "loss": 1.7057, "step": 11475 }, { "epoch": 2.19, "learning_rate": 2.3464632983794093e-06, "loss": 2.5161, "step": 11476 }, { "epoch": 2.19, "learning_rate": 2.3464061010486177e-06, "loss": 2.977, "step": 11477 }, { "epoch": 2.19, "learning_rate": 2.3463489037178265e-06, "loss": 2.7951, "step": 11478 }, { "epoch": 2.19, "learning_rate": 2.3462917063870353e-06, "loss": 2.7627, "step": 11479 }, { "epoch": 2.19, "learning_rate": 2.346234509056244e-06, "loss": 2.7074, "step": 11480 }, { "epoch": 2.19, "learning_rate": 2.346177311725453e-06, "loss": 3.3345, "step": 11481 }, { "epoch": 2.19, "learning_rate": 2.3461201143946616e-06, "loss": 2.7087, "step": 11482 }, { "epoch": 2.19, "learning_rate": 2.3460629170638704e-06, "loss": 2.4744, "step": 11483 }, { "epoch": 2.19, "learning_rate": 2.3460057197330792e-06, "loss": 2.8791, "step": 11484 }, { "epoch": 2.19, "learning_rate": 2.345948522402288e-06, "loss": 2.5456, "step": 11485 }, { "epoch": 2.19, "learning_rate": 2.345891325071497e-06, "loss": 2.4131, "step": 11486 }, { "epoch": 2.19, "learning_rate": 2.3458341277407056e-06, "loss": 1.6497, "step": 11487 }, { "epoch": 2.19, "learning_rate": 2.3457769304099144e-06, "loss": 2.539, "step": 11488 }, { "epoch": 2.19, "learning_rate": 2.3457197330791228e-06, "loss": 2.7737, "step": 11489 }, { "epoch": 2.19, "learning_rate": 2.345662535748332e-06, "loss": 2.353, "step": 11490 }, { "epoch": 2.19, "learning_rate": 2.3456053384175404e-06, "loss": 3.3593, "step": 11491 }, { "epoch": 2.19, "learning_rate": 2.345548141086749e-06, "loss": 2.9458, "step": 11492 }, { "epoch": 2.19, "learning_rate": 2.3454909437559584e-06, "loss": 3.5624, "step": 11493 }, { "epoch": 2.19, "learning_rate": 2.3454337464251667e-06, "loss": 2.5847, "step": 11494 }, { "epoch": 2.19, "learning_rate": 2.345376549094376e-06, "loss": 2.6277, "step": 11495 }, { "epoch": 2.19, "learning_rate": 2.3453193517635843e-06, "loss": 2.2805, "step": 11496 }, { "epoch": 2.19, "learning_rate": 2.345262154432793e-06, "loss": 2.566, "step": 11497 }, { "epoch": 2.19, "learning_rate": 2.345204957102002e-06, "loss": 2.273, "step": 11498 }, { "epoch": 2.19, "learning_rate": 2.3451477597712107e-06, "loss": 1.7206, "step": 11499 }, { "epoch": 2.19, "learning_rate": 2.3450905624404195e-06, "loss": 2.5036, "step": 11500 }, { "epoch": 2.19, "learning_rate": 2.3450333651096283e-06, "loss": 3.1173, "step": 11501 }, { "epoch": 2.19, "learning_rate": 2.344976167778837e-06, "loss": 2.471, "step": 11502 }, { "epoch": 2.19, "learning_rate": 2.344918970448046e-06, "loss": 2.9961, "step": 11503 }, { "epoch": 2.19, "learning_rate": 2.3448617731172547e-06, "loss": 3.0842, "step": 11504 }, { "epoch": 2.19, "learning_rate": 2.3448045757864635e-06, "loss": 1.9142, "step": 11505 }, { "epoch": 2.19, "learning_rate": 2.344747378455672e-06, "loss": 3.75, "step": 11506 }, { "epoch": 2.19, "learning_rate": 2.344690181124881e-06, "loss": 2.5425, "step": 11507 }, { "epoch": 2.19, "learning_rate": 2.3446329837940894e-06, "loss": 2.815, "step": 11508 }, { "epoch": 2.19, "learning_rate": 2.3445757864632982e-06, "loss": 2.8403, "step": 11509 }, { "epoch": 2.19, "learning_rate": 2.3445185891325074e-06, "loss": 3.4852, "step": 11510 }, { "epoch": 2.19, "learning_rate": 2.344461391801716e-06, "loss": 2.2034, "step": 11511 }, { "epoch": 2.19, "learning_rate": 2.344404194470925e-06, "loss": 2.9831, "step": 11512 }, { "epoch": 2.19, "learning_rate": 2.3443469971401334e-06, "loss": 3.1592, "step": 11513 }, { "epoch": 2.19, "learning_rate": 2.344289799809342e-06, "loss": 2.6678, "step": 11514 }, { "epoch": 2.19, "learning_rate": 2.344232602478551e-06, "loss": 2.7461, "step": 11515 }, { "epoch": 2.19, "learning_rate": 2.3441754051477598e-06, "loss": 2.3895, "step": 11516 }, { "epoch": 2.19, "learning_rate": 2.3441182078169686e-06, "loss": 2.8231, "step": 11517 }, { "epoch": 2.19, "learning_rate": 2.3440610104861774e-06, "loss": 2.2292, "step": 11518 }, { "epoch": 2.19, "learning_rate": 2.344003813155386e-06, "loss": 2.8493, "step": 11519 }, { "epoch": 2.19, "learning_rate": 2.343946615824595e-06, "loss": 3.9743, "step": 11520 }, { "epoch": 2.19, "learning_rate": 2.3438894184938038e-06, "loss": 2.7313, "step": 11521 }, { "epoch": 2.19, "learning_rate": 2.3438322211630126e-06, "loss": 2.0775, "step": 11522 }, { "epoch": 2.19, "learning_rate": 2.343775023832221e-06, "loss": 2.8313, "step": 11523 }, { "epoch": 2.2, "learning_rate": 2.34371782650143e-06, "loss": 3.4847, "step": 11524 }, { "epoch": 2.2, "learning_rate": 2.343660629170639e-06, "loss": 3.1187, "step": 11525 }, { "epoch": 2.2, "learning_rate": 2.3436034318398473e-06, "loss": 3.4291, "step": 11526 }, { "epoch": 2.2, "learning_rate": 2.3435462345090565e-06, "loss": 3.7982, "step": 11527 }, { "epoch": 2.2, "learning_rate": 2.343489037178265e-06, "loss": 2.9961, "step": 11528 }, { "epoch": 2.2, "learning_rate": 2.343431839847474e-06, "loss": 3.033, "step": 11529 }, { "epoch": 2.2, "learning_rate": 2.3433746425166825e-06, "loss": 2.5955, "step": 11530 }, { "epoch": 2.2, "learning_rate": 2.3433174451858913e-06, "loss": 3.0718, "step": 11531 }, { "epoch": 2.2, "learning_rate": 2.3432602478551005e-06, "loss": 2.2668, "step": 11532 }, { "epoch": 2.2, "learning_rate": 2.343203050524309e-06, "loss": 2.1537, "step": 11533 }, { "epoch": 2.2, "learning_rate": 2.3431458531935177e-06, "loss": 3.1157, "step": 11534 }, { "epoch": 2.2, "learning_rate": 2.3430886558627264e-06, "loss": 2.7425, "step": 11535 }, { "epoch": 2.2, "learning_rate": 2.3430314585319352e-06, "loss": 3.376, "step": 11536 }, { "epoch": 2.2, "learning_rate": 2.342974261201144e-06, "loss": 3.6426, "step": 11537 }, { "epoch": 2.2, "learning_rate": 2.342917063870353e-06, "loss": 2.3303, "step": 11538 }, { "epoch": 2.2, "learning_rate": 2.3428598665395616e-06, "loss": 2.5096, "step": 11539 }, { "epoch": 2.2, "learning_rate": 2.34280266920877e-06, "loss": 3.4815, "step": 11540 }, { "epoch": 2.2, "learning_rate": 2.342745471877979e-06, "loss": 2.8773, "step": 11541 }, { "epoch": 2.2, "learning_rate": 2.342688274547188e-06, "loss": 2.234, "step": 11542 }, { "epoch": 2.2, "learning_rate": 2.3426310772163964e-06, "loss": 2.2721, "step": 11543 }, { "epoch": 2.2, "learning_rate": 2.3425738798856056e-06, "loss": 3.0004, "step": 11544 }, { "epoch": 2.2, "learning_rate": 2.342516682554814e-06, "loss": 2.7954, "step": 11545 }, { "epoch": 2.2, "learning_rate": 2.342459485224023e-06, "loss": 2.8592, "step": 11546 }, { "epoch": 2.2, "learning_rate": 2.3424022878932315e-06, "loss": 2.0905, "step": 11547 }, { "epoch": 2.2, "learning_rate": 2.3423450905624403e-06, "loss": 3.0952, "step": 11548 }, { "epoch": 2.2, "learning_rate": 2.3422878932316496e-06, "loss": 2.3336, "step": 11549 }, { "epoch": 2.2, "learning_rate": 2.342230695900858e-06, "loss": 2.9498, "step": 11550 }, { "epoch": 2.2, "learning_rate": 2.3421734985700667e-06, "loss": 2.1858, "step": 11551 }, { "epoch": 2.2, "learning_rate": 2.3421163012392755e-06, "loss": 2.7176, "step": 11552 }, { "epoch": 2.2, "learning_rate": 2.3420591039084843e-06, "loss": 3.2856, "step": 11553 }, { "epoch": 2.2, "learning_rate": 2.342001906577693e-06, "loss": 2.4749, "step": 11554 }, { "epoch": 2.2, "learning_rate": 2.341944709246902e-06, "loss": 2.3102, "step": 11555 }, { "epoch": 2.2, "learning_rate": 2.3418875119161107e-06, "loss": 3.3826, "step": 11556 }, { "epoch": 2.2, "learning_rate": 2.341830314585319e-06, "loss": 2.9675, "step": 11557 }, { "epoch": 2.2, "learning_rate": 2.3417731172545283e-06, "loss": 2.7953, "step": 11558 }, { "epoch": 2.2, "learning_rate": 2.341715919923737e-06, "loss": 2.1761, "step": 11559 }, { "epoch": 2.2, "learning_rate": 2.3416587225929454e-06, "loss": 2.4945, "step": 11560 }, { "epoch": 2.2, "learning_rate": 2.3416015252621547e-06, "loss": 2.7126, "step": 11561 }, { "epoch": 2.2, "learning_rate": 2.341544327931363e-06, "loss": 2.848, "step": 11562 }, { "epoch": 2.2, "learning_rate": 2.3414871306005723e-06, "loss": 3.0954, "step": 11563 }, { "epoch": 2.2, "learning_rate": 2.3414299332697806e-06, "loss": 2.8346, "step": 11564 }, { "epoch": 2.2, "learning_rate": 2.3413727359389894e-06, "loss": 3.2496, "step": 11565 }, { "epoch": 2.2, "learning_rate": 2.3413155386081986e-06, "loss": 3.3795, "step": 11566 }, { "epoch": 2.2, "learning_rate": 2.341258341277407e-06, "loss": 3.2185, "step": 11567 }, { "epoch": 2.2, "learning_rate": 2.341201143946616e-06, "loss": 2.5777, "step": 11568 }, { "epoch": 2.2, "learning_rate": 2.3411439466158246e-06, "loss": 3.2321, "step": 11569 }, { "epoch": 2.2, "learning_rate": 2.3410867492850334e-06, "loss": 3.1182, "step": 11570 }, { "epoch": 2.2, "learning_rate": 2.341029551954242e-06, "loss": 2.8155, "step": 11571 }, { "epoch": 2.2, "learning_rate": 2.340972354623451e-06, "loss": 2.7007, "step": 11572 }, { "epoch": 2.2, "learning_rate": 2.3409151572926598e-06, "loss": 2.9453, "step": 11573 }, { "epoch": 2.2, "learning_rate": 2.3408579599618686e-06, "loss": 3.0818, "step": 11574 }, { "epoch": 2.2, "learning_rate": 2.3408007626310774e-06, "loss": 3.0675, "step": 11575 }, { "epoch": 2.2, "learning_rate": 2.340743565300286e-06, "loss": 1.9746, "step": 11576 }, { "epoch": 2.21, "learning_rate": 2.340686367969495e-06, "loss": 3.1991, "step": 11577 }, { "epoch": 2.21, "learning_rate": 2.3406291706387037e-06, "loss": 2.4253, "step": 11578 }, { "epoch": 2.21, "learning_rate": 2.340571973307912e-06, "loss": 3.5725, "step": 11579 }, { "epoch": 2.21, "learning_rate": 2.3405147759771213e-06, "loss": 2.7839, "step": 11580 }, { "epoch": 2.21, "learning_rate": 2.34045757864633e-06, "loss": 3.2066, "step": 11581 }, { "epoch": 2.21, "learning_rate": 2.3404003813155385e-06, "loss": 3.8727, "step": 11582 }, { "epoch": 2.21, "learning_rate": 2.3403431839847477e-06, "loss": 3.1787, "step": 11583 }, { "epoch": 2.21, "learning_rate": 2.340285986653956e-06, "loss": 2.5781, "step": 11584 }, { "epoch": 2.21, "learning_rate": 2.340228789323165e-06, "loss": 2.471, "step": 11585 }, { "epoch": 2.21, "learning_rate": 2.3401715919923737e-06, "loss": 2.5159, "step": 11586 }, { "epoch": 2.21, "learning_rate": 2.3401143946615825e-06, "loss": 2.9008, "step": 11587 }, { "epoch": 2.21, "learning_rate": 2.3400571973307912e-06, "loss": 2.3738, "step": 11588 }, { "epoch": 2.21, "learning_rate": 2.34e-06, "loss": 3.4157, "step": 11589 }, { "epoch": 2.21, "learning_rate": 2.339942802669209e-06, "loss": 1.9436, "step": 11590 }, { "epoch": 2.21, "learning_rate": 2.3398856053384176e-06, "loss": 2.9506, "step": 11591 }, { "epoch": 2.21, "learning_rate": 2.3398284080076264e-06, "loss": 3.2925, "step": 11592 }, { "epoch": 2.21, "learning_rate": 2.3397712106768352e-06, "loss": 2.2457, "step": 11593 }, { "epoch": 2.21, "learning_rate": 2.339714013346044e-06, "loss": 3.7635, "step": 11594 }, { "epoch": 2.21, "learning_rate": 2.339656816015253e-06, "loss": 3.8338, "step": 11595 }, { "epoch": 2.21, "learning_rate": 2.339599618684461e-06, "loss": 3.2842, "step": 11596 }, { "epoch": 2.21, "learning_rate": 2.3395424213536704e-06, "loss": 3.0865, "step": 11597 }, { "epoch": 2.21, "learning_rate": 2.339485224022879e-06, "loss": 2.4123, "step": 11598 }, { "epoch": 2.21, "learning_rate": 2.3394280266920876e-06, "loss": 3.4972, "step": 11599 }, { "epoch": 2.21, "learning_rate": 2.3393708293612968e-06, "loss": 2.4402, "step": 11600 }, { "epoch": 2.21, "learning_rate": 2.339313632030505e-06, "loss": 4.3017, "step": 11601 }, { "epoch": 2.21, "learning_rate": 2.339256434699714e-06, "loss": 3.1648, "step": 11602 }, { "epoch": 2.21, "learning_rate": 2.3391992373689227e-06, "loss": 2.3297, "step": 11603 }, { "epoch": 2.21, "learning_rate": 2.3391420400381315e-06, "loss": 2.8206, "step": 11604 }, { "epoch": 2.21, "learning_rate": 2.3390848427073403e-06, "loss": 1.5915, "step": 11605 }, { "epoch": 2.21, "learning_rate": 2.339027645376549e-06, "loss": 2.6005, "step": 11606 }, { "epoch": 2.21, "learning_rate": 2.338970448045758e-06, "loss": 2.1079, "step": 11607 }, { "epoch": 2.21, "learning_rate": 2.3389132507149667e-06, "loss": 3.3258, "step": 11608 }, { "epoch": 2.21, "learning_rate": 2.3388560533841755e-06, "loss": 2.9927, "step": 11609 }, { "epoch": 2.21, "learning_rate": 2.3387988560533843e-06, "loss": 2.9663, "step": 11610 }, { "epoch": 2.21, "learning_rate": 2.338741658722593e-06, "loss": 2.8295, "step": 11611 }, { "epoch": 2.21, "learning_rate": 2.338684461391802e-06, "loss": 3.1936, "step": 11612 }, { "epoch": 2.21, "learning_rate": 2.3386272640610107e-06, "loss": 2.573, "step": 11613 }, { "epoch": 2.21, "learning_rate": 2.3385700667302195e-06, "loss": 2.7705, "step": 11614 }, { "epoch": 2.21, "learning_rate": 2.3385128693994283e-06, "loss": 2.8151, "step": 11615 }, { "epoch": 2.21, "learning_rate": 2.3384556720686366e-06, "loss": 2.6952, "step": 11616 }, { "epoch": 2.21, "learning_rate": 2.338398474737846e-06, "loss": 2.4972, "step": 11617 }, { "epoch": 2.21, "learning_rate": 2.3383412774070542e-06, "loss": 3.0285, "step": 11618 }, { "epoch": 2.21, "learning_rate": 2.338284080076263e-06, "loss": 3.0642, "step": 11619 }, { "epoch": 2.21, "learning_rate": 2.3382268827454722e-06, "loss": 2.3622, "step": 11620 }, { "epoch": 2.21, "learning_rate": 2.3381696854146806e-06, "loss": 3.1013, "step": 11621 }, { "epoch": 2.21, "learning_rate": 2.3381124880838894e-06, "loss": 1.6864, "step": 11622 }, { "epoch": 2.21, "learning_rate": 2.338055290753098e-06, "loss": 2.6245, "step": 11623 }, { "epoch": 2.21, "learning_rate": 2.337998093422307e-06, "loss": 3.4719, "step": 11624 }, { "epoch": 2.21, "learning_rate": 2.3379408960915158e-06, "loss": 2.7702, "step": 11625 }, { "epoch": 2.21, "learning_rate": 2.3378836987607246e-06, "loss": 3.2637, "step": 11626 }, { "epoch": 2.21, "learning_rate": 2.3378265014299334e-06, "loss": 2.8795, "step": 11627 }, { "epoch": 2.21, "learning_rate": 2.337769304099142e-06, "loss": 2.2546, "step": 11628 }, { "epoch": 2.22, "learning_rate": 2.337712106768351e-06, "loss": 2.463, "step": 11629 }, { "epoch": 2.22, "learning_rate": 2.3376549094375597e-06, "loss": 2.469, "step": 11630 }, { "epoch": 2.22, "learning_rate": 2.3375977121067685e-06, "loss": 2.9225, "step": 11631 }, { "epoch": 2.22, "learning_rate": 2.3375405147759773e-06, "loss": 2.9611, "step": 11632 }, { "epoch": 2.22, "learning_rate": 2.3374833174451857e-06, "loss": 3.3053, "step": 11633 }, { "epoch": 2.22, "learning_rate": 2.337426120114395e-06, "loss": 2.6605, "step": 11634 }, { "epoch": 2.22, "learning_rate": 2.3373689227836033e-06, "loss": 3.279, "step": 11635 }, { "epoch": 2.22, "learning_rate": 2.337311725452812e-06, "loss": 2.7479, "step": 11636 }, { "epoch": 2.22, "learning_rate": 2.3372545281220213e-06, "loss": 2.465, "step": 11637 }, { "epoch": 2.22, "learning_rate": 2.3371973307912297e-06, "loss": 3.1537, "step": 11638 }, { "epoch": 2.22, "learning_rate": 2.3371401334604385e-06, "loss": 2.888, "step": 11639 }, { "epoch": 2.22, "learning_rate": 2.3370829361296473e-06, "loss": 2.9181, "step": 11640 }, { "epoch": 2.22, "learning_rate": 2.337025738798856e-06, "loss": 2.8791, "step": 11641 }, { "epoch": 2.22, "learning_rate": 2.336968541468065e-06, "loss": 3.2559, "step": 11642 }, { "epoch": 2.22, "learning_rate": 2.3369113441372736e-06, "loss": 2.8376, "step": 11643 }, { "epoch": 2.22, "learning_rate": 2.3368541468064824e-06, "loss": 2.3716, "step": 11644 }, { "epoch": 2.22, "learning_rate": 2.3367969494756912e-06, "loss": 3.0149, "step": 11645 }, { "epoch": 2.22, "learning_rate": 2.3367397521449e-06, "loss": 2.3912, "step": 11646 }, { "epoch": 2.22, "learning_rate": 2.336682554814109e-06, "loss": 3.625, "step": 11647 }, { "epoch": 2.22, "learning_rate": 2.3366253574833176e-06, "loss": 2.8665, "step": 11648 }, { "epoch": 2.22, "learning_rate": 2.3365681601525264e-06, "loss": 2.7317, "step": 11649 }, { "epoch": 2.22, "learning_rate": 2.3365109628217348e-06, "loss": 2.1244, "step": 11650 }, { "epoch": 2.22, "learning_rate": 2.336453765490944e-06, "loss": 2.8688, "step": 11651 }, { "epoch": 2.22, "learning_rate": 2.3363965681601524e-06, "loss": 2.5535, "step": 11652 }, { "epoch": 2.22, "learning_rate": 2.336339370829361e-06, "loss": 2.264, "step": 11653 }, { "epoch": 2.22, "learning_rate": 2.3362821734985704e-06, "loss": 2.1326, "step": 11654 }, { "epoch": 2.22, "learning_rate": 2.3362249761677787e-06, "loss": 2.9563, "step": 11655 }, { "epoch": 2.22, "learning_rate": 2.3361677788369875e-06, "loss": 3.4308, "step": 11656 }, { "epoch": 2.22, "learning_rate": 2.3361105815061963e-06, "loss": 2.0718, "step": 11657 }, { "epoch": 2.22, "learning_rate": 2.336053384175405e-06, "loss": 3.6486, "step": 11658 }, { "epoch": 2.22, "learning_rate": 2.3359961868446143e-06, "loss": 2.1535, "step": 11659 }, { "epoch": 2.22, "learning_rate": 2.3359389895138227e-06, "loss": 2.1076, "step": 11660 }, { "epoch": 2.22, "learning_rate": 2.3358817921830315e-06, "loss": 3.2438, "step": 11661 }, { "epoch": 2.22, "learning_rate": 2.3358245948522403e-06, "loss": 3.2149, "step": 11662 }, { "epoch": 2.22, "learning_rate": 2.335767397521449e-06, "loss": 2.3914, "step": 11663 }, { "epoch": 2.22, "learning_rate": 2.335710200190658e-06, "loss": 1.8467, "step": 11664 }, { "epoch": 2.22, "learning_rate": 2.3356530028598667e-06, "loss": 3.2787, "step": 11665 }, { "epoch": 2.22, "learning_rate": 2.3355958055290755e-06, "loss": 3.3581, "step": 11666 }, { "epoch": 2.22, "learning_rate": 2.335538608198284e-06, "loss": 3.08, "step": 11667 }, { "epoch": 2.22, "learning_rate": 2.335481410867493e-06, "loss": 2.4982, "step": 11668 }, { "epoch": 2.22, "learning_rate": 2.335424213536702e-06, "loss": 2.8169, "step": 11669 }, { "epoch": 2.22, "learning_rate": 2.3353670162059102e-06, "loss": 2.0365, "step": 11670 }, { "epoch": 2.22, "learning_rate": 2.3353098188751194e-06, "loss": 2.9373, "step": 11671 }, { "epoch": 2.22, "learning_rate": 2.335252621544328e-06, "loss": 2.4087, "step": 11672 }, { "epoch": 2.22, "learning_rate": 2.3351954242135366e-06, "loss": 2.4447, "step": 11673 }, { "epoch": 2.22, "learning_rate": 2.3351382268827454e-06, "loss": 2.6119, "step": 11674 }, { "epoch": 2.22, "learning_rate": 2.335081029551954e-06, "loss": 4.1271, "step": 11675 }, { "epoch": 2.22, "learning_rate": 2.3350238322211634e-06, "loss": 2.9227, "step": 11676 }, { "epoch": 2.22, "learning_rate": 2.3349666348903718e-06, "loss": 3.1944, "step": 11677 }, { "epoch": 2.22, "learning_rate": 2.3349094375595806e-06, "loss": 2.0395, "step": 11678 }, { "epoch": 2.22, "learning_rate": 2.3348522402287894e-06, "loss": 2.4166, "step": 11679 }, { "epoch": 2.22, "learning_rate": 2.334795042897998e-06, "loss": 2.1408, "step": 11680 }, { "epoch": 2.22, "learning_rate": 2.334737845567207e-06, "loss": 2.628, "step": 11681 }, { "epoch": 2.23, "learning_rate": 2.3346806482364157e-06, "loss": 2.7106, "step": 11682 }, { "epoch": 2.23, "learning_rate": 2.3346234509056245e-06, "loss": 2.5567, "step": 11683 }, { "epoch": 2.23, "learning_rate": 2.334566253574833e-06, "loss": 2.0489, "step": 11684 }, { "epoch": 2.23, "learning_rate": 2.334509056244042e-06, "loss": 3.0064, "step": 11685 }, { "epoch": 2.23, "learning_rate": 2.334451858913251e-06, "loss": 3.2291, "step": 11686 }, { "epoch": 2.23, "learning_rate": 2.3343946615824593e-06, "loss": 2.8834, "step": 11687 }, { "epoch": 2.23, "learning_rate": 2.3343374642516685e-06, "loss": 1.7342, "step": 11688 }, { "epoch": 2.23, "learning_rate": 2.334280266920877e-06, "loss": 2.6635, "step": 11689 }, { "epoch": 2.23, "learning_rate": 2.3342230695900857e-06, "loss": 2.6465, "step": 11690 }, { "epoch": 2.23, "learning_rate": 2.3341658722592945e-06, "loss": 2.8668, "step": 11691 }, { "epoch": 2.23, "learning_rate": 2.3341086749285033e-06, "loss": 2.6812, "step": 11692 }, { "epoch": 2.23, "learning_rate": 2.3340514775977125e-06, "loss": 2.7228, "step": 11693 }, { "epoch": 2.23, "learning_rate": 2.333994280266921e-06, "loss": 2.969, "step": 11694 }, { "epoch": 2.23, "learning_rate": 2.3339370829361296e-06, "loss": 2.3196, "step": 11695 }, { "epoch": 2.23, "learning_rate": 2.3338798856053384e-06, "loss": 2.4652, "step": 11696 }, { "epoch": 2.23, "learning_rate": 2.3338226882745472e-06, "loss": 3.0192, "step": 11697 }, { "epoch": 2.23, "learning_rate": 2.333765490943756e-06, "loss": 3.194, "step": 11698 }, { "epoch": 2.23, "learning_rate": 2.333708293612965e-06, "loss": 2.3469, "step": 11699 }, { "epoch": 2.23, "learning_rate": 2.3336510962821736e-06, "loss": 2.1173, "step": 11700 }, { "epoch": 2.23, "learning_rate": 2.3335938989513824e-06, "loss": 3.1799, "step": 11701 }, { "epoch": 2.23, "learning_rate": 2.333536701620591e-06, "loss": 2.7782, "step": 11702 }, { "epoch": 2.23, "learning_rate": 2.3334795042898e-06, "loss": 2.2708, "step": 11703 }, { "epoch": 2.23, "learning_rate": 2.3334223069590084e-06, "loss": 2.2457, "step": 11704 }, { "epoch": 2.23, "learning_rate": 2.3333651096282176e-06, "loss": 1.9913, "step": 11705 }, { "epoch": 2.23, "learning_rate": 2.333307912297426e-06, "loss": 2.8057, "step": 11706 }, { "epoch": 2.23, "learning_rate": 2.333250714966635e-06, "loss": 2.2353, "step": 11707 }, { "epoch": 2.23, "learning_rate": 2.333193517635844e-06, "loss": 2.7279, "step": 11708 }, { "epoch": 2.23, "learning_rate": 2.3331363203050523e-06, "loss": 2.9232, "step": 11709 }, { "epoch": 2.23, "learning_rate": 2.3330791229742616e-06, "loss": 4.0976, "step": 11710 }, { "epoch": 2.23, "learning_rate": 2.33302192564347e-06, "loss": 2.8972, "step": 11711 }, { "epoch": 2.23, "learning_rate": 2.3329647283126787e-06, "loss": 2.3599, "step": 11712 }, { "epoch": 2.23, "learning_rate": 2.3329075309818875e-06, "loss": 3.0789, "step": 11713 }, { "epoch": 2.23, "learning_rate": 2.3328503336510963e-06, "loss": 2.2308, "step": 11714 }, { "epoch": 2.23, "learning_rate": 2.332793136320305e-06, "loss": 2.3549, "step": 11715 }, { "epoch": 2.23, "learning_rate": 2.332735938989514e-06, "loss": 1.9879, "step": 11716 }, { "epoch": 2.23, "learning_rate": 2.3326787416587227e-06, "loss": 2.4377, "step": 11717 }, { "epoch": 2.23, "learning_rate": 2.3326215443279315e-06, "loss": 2.8251, "step": 11718 }, { "epoch": 2.23, "learning_rate": 2.3325643469971403e-06, "loss": 2.5434, "step": 11719 }, { "epoch": 2.23, "learning_rate": 2.332507149666349e-06, "loss": 2.0475, "step": 11720 }, { "epoch": 2.23, "learning_rate": 2.3324499523355574e-06, "loss": 2.8764, "step": 11721 }, { "epoch": 2.23, "learning_rate": 2.3323927550047667e-06, "loss": 2.6091, "step": 11722 }, { "epoch": 2.23, "learning_rate": 2.332335557673975e-06, "loss": 3.4854, "step": 11723 }, { "epoch": 2.23, "learning_rate": 2.3322783603431842e-06, "loss": 3.3676, "step": 11724 }, { "epoch": 2.23, "learning_rate": 2.332221163012393e-06, "loss": 2.2101, "step": 11725 }, { "epoch": 2.23, "learning_rate": 2.3321639656816014e-06, "loss": 3.6269, "step": 11726 }, { "epoch": 2.23, "learning_rate": 2.3321067683508106e-06, "loss": 3.1761, "step": 11727 }, { "epoch": 2.23, "learning_rate": 2.332049571020019e-06, "loss": 3.2458, "step": 11728 }, { "epoch": 2.23, "learning_rate": 2.331992373689228e-06, "loss": 2.6591, "step": 11729 }, { "epoch": 2.23, "learning_rate": 2.3319351763584366e-06, "loss": 3.3738, "step": 11730 }, { "epoch": 2.23, "learning_rate": 2.3318779790276454e-06, "loss": 3.4381, "step": 11731 }, { "epoch": 2.23, "learning_rate": 2.331820781696854e-06, "loss": 2.451, "step": 11732 }, { "epoch": 2.23, "learning_rate": 2.331763584366063e-06, "loss": 3.996, "step": 11733 }, { "epoch": 2.24, "learning_rate": 2.3317063870352718e-06, "loss": 3.6751, "step": 11734 }, { "epoch": 2.24, "learning_rate": 2.3316491897044806e-06, "loss": 3.0634, "step": 11735 }, { "epoch": 2.24, "learning_rate": 2.3315919923736893e-06, "loss": 2.702, "step": 11736 }, { "epoch": 2.24, "learning_rate": 2.331534795042898e-06, "loss": 2.9695, "step": 11737 }, { "epoch": 2.24, "learning_rate": 2.3314775977121065e-06, "loss": 2.6216, "step": 11738 }, { "epoch": 2.24, "learning_rate": 2.3314204003813157e-06, "loss": 2.4806, "step": 11739 }, { "epoch": 2.24, "learning_rate": 2.331363203050524e-06, "loss": 3.6183, "step": 11740 }, { "epoch": 2.24, "learning_rate": 2.3313060057197333e-06, "loss": 3.0598, "step": 11741 }, { "epoch": 2.24, "learning_rate": 2.331248808388942e-06, "loss": 3.4133, "step": 11742 }, { "epoch": 2.24, "learning_rate": 2.3311916110581505e-06, "loss": 2.9482, "step": 11743 }, { "epoch": 2.24, "learning_rate": 2.3311344137273597e-06, "loss": 3.4074, "step": 11744 }, { "epoch": 2.24, "learning_rate": 2.331077216396568e-06, "loss": 2.7734, "step": 11745 }, { "epoch": 2.24, "learning_rate": 2.331020019065777e-06, "loss": 2.1256, "step": 11746 }, { "epoch": 2.24, "learning_rate": 2.330962821734986e-06, "loss": 3.0297, "step": 11747 }, { "epoch": 2.24, "learning_rate": 2.3309056244041944e-06, "loss": 3.4753, "step": 11748 }, { "epoch": 2.24, "learning_rate": 2.3308484270734032e-06, "loss": 3.0088, "step": 11749 }, { "epoch": 2.24, "learning_rate": 2.330791229742612e-06, "loss": 2.3217, "step": 11750 }, { "epoch": 2.24, "learning_rate": 2.330734032411821e-06, "loss": 2.6154, "step": 11751 }, { "epoch": 2.24, "learning_rate": 2.3306768350810296e-06, "loss": 3.1253, "step": 11752 }, { "epoch": 2.24, "learning_rate": 2.3306196377502384e-06, "loss": 2.4095, "step": 11753 }, { "epoch": 2.24, "learning_rate": 2.330562440419447e-06, "loss": 2.6838, "step": 11754 }, { "epoch": 2.24, "learning_rate": 2.3305052430886556e-06, "loss": 2.8096, "step": 11755 }, { "epoch": 2.24, "learning_rate": 2.330448045757865e-06, "loss": 2.9663, "step": 11756 }, { "epoch": 2.24, "learning_rate": 2.3303908484270736e-06, "loss": 2.0264, "step": 11757 }, { "epoch": 2.24, "learning_rate": 2.3303336510962824e-06, "loss": 2.6415, "step": 11758 }, { "epoch": 2.24, "learning_rate": 2.330276453765491e-06, "loss": 3.3549, "step": 11759 }, { "epoch": 2.24, "learning_rate": 2.3302192564346996e-06, "loss": 1.7734, "step": 11760 }, { "epoch": 2.24, "learning_rate": 2.3301620591039088e-06, "loss": 2.7403, "step": 11761 }, { "epoch": 2.24, "learning_rate": 2.330104861773117e-06, "loss": 2.8738, "step": 11762 }, { "epoch": 2.24, "learning_rate": 2.330047664442326e-06, "loss": 2.2823, "step": 11763 }, { "epoch": 2.24, "learning_rate": 2.329990467111535e-06, "loss": 2.2388, "step": 11764 }, { "epoch": 2.24, "learning_rate": 2.3299332697807435e-06, "loss": 2.9138, "step": 11765 }, { "epoch": 2.24, "learning_rate": 2.3298760724499523e-06, "loss": 2.5639, "step": 11766 }, { "epoch": 2.24, "learning_rate": 2.329818875119161e-06, "loss": 3.1613, "step": 11767 }, { "epoch": 2.24, "learning_rate": 2.32976167778837e-06, "loss": 3.8653, "step": 11768 }, { "epoch": 2.24, "learning_rate": 2.3297044804575787e-06, "loss": 2.0914, "step": 11769 }, { "epoch": 2.24, "learning_rate": 2.3296472831267875e-06, "loss": 3.3423, "step": 11770 }, { "epoch": 2.24, "learning_rate": 2.3295900857959963e-06, "loss": 2.6148, "step": 11771 }, { "epoch": 2.24, "learning_rate": 2.329532888465205e-06, "loss": 2.9327, "step": 11772 }, { "epoch": 2.24, "learning_rate": 2.329475691134414e-06, "loss": 3.7326, "step": 11773 }, { "epoch": 2.24, "learning_rate": 2.3294184938036227e-06, "loss": 2.5233, "step": 11774 }, { "epoch": 2.24, "learning_rate": 2.3293612964728315e-06, "loss": 3.1681, "step": 11775 }, { "epoch": 2.24, "learning_rate": 2.3293040991420403e-06, "loss": 2.0643, "step": 11776 }, { "epoch": 2.24, "learning_rate": 2.3292469018112486e-06, "loss": 2.189, "step": 11777 }, { "epoch": 2.24, "learning_rate": 2.329189704480458e-06, "loss": 3.1692, "step": 11778 }, { "epoch": 2.24, "learning_rate": 2.329132507149666e-06, "loss": 2.7168, "step": 11779 }, { "epoch": 2.24, "learning_rate": 2.329075309818875e-06, "loss": 2.728, "step": 11780 }, { "epoch": 2.24, "learning_rate": 2.3290181124880842e-06, "loss": 3.0386, "step": 11781 }, { "epoch": 2.24, "learning_rate": 2.3289609151572926e-06, "loss": 2.0117, "step": 11782 }, { "epoch": 2.24, "learning_rate": 2.3289037178265014e-06, "loss": 2.7846, "step": 11783 }, { "epoch": 2.24, "learning_rate": 2.32884652049571e-06, "loss": 2.8729, "step": 11784 }, { "epoch": 2.24, "learning_rate": 2.328789323164919e-06, "loss": 2.6365, "step": 11785 }, { "epoch": 2.24, "learning_rate": 2.3287321258341278e-06, "loss": 2.4686, "step": 11786 }, { "epoch": 2.25, "learning_rate": 2.3286749285033366e-06, "loss": 3.3935, "step": 11787 }, { "epoch": 2.25, "learning_rate": 2.3286177311725454e-06, "loss": 2.7046, "step": 11788 }, { "epoch": 2.25, "learning_rate": 2.328560533841754e-06, "loss": 2.2925, "step": 11789 }, { "epoch": 2.25, "learning_rate": 2.328503336510963e-06, "loss": 4.0799, "step": 11790 }, { "epoch": 2.25, "learning_rate": 2.3284461391801717e-06, "loss": 3.6335, "step": 11791 }, { "epoch": 2.25, "learning_rate": 2.3283889418493805e-06, "loss": 2.944, "step": 11792 }, { "epoch": 2.25, "learning_rate": 2.3283317445185893e-06, "loss": 2.7766, "step": 11793 }, { "epoch": 2.25, "learning_rate": 2.3282745471877977e-06, "loss": 2.5748, "step": 11794 }, { "epoch": 2.25, "learning_rate": 2.328217349857007e-06, "loss": 2.4394, "step": 11795 }, { "epoch": 2.25, "learning_rate": 2.3281601525262157e-06, "loss": 3.4352, "step": 11796 }, { "epoch": 2.25, "learning_rate": 2.328102955195424e-06, "loss": 2.7779, "step": 11797 }, { "epoch": 2.25, "learning_rate": 2.3280457578646333e-06, "loss": 3.4499, "step": 11798 }, { "epoch": 2.25, "learning_rate": 2.3279885605338417e-06, "loss": 2.5288, "step": 11799 }, { "epoch": 2.25, "learning_rate": 2.3279313632030505e-06, "loss": 3.0688, "step": 11800 }, { "epoch": 2.25, "learning_rate": 2.3278741658722592e-06, "loss": 3.3393, "step": 11801 }, { "epoch": 2.25, "learning_rate": 2.327816968541468e-06, "loss": 2.9142, "step": 11802 }, { "epoch": 2.25, "learning_rate": 2.327759771210677e-06, "loss": 2.0813, "step": 11803 }, { "epoch": 2.25, "learning_rate": 2.3277025738798856e-06, "loss": 2.6927, "step": 11804 }, { "epoch": 2.25, "learning_rate": 2.3276453765490944e-06, "loss": 1.6365, "step": 11805 }, { "epoch": 2.25, "learning_rate": 2.3275881792183032e-06, "loss": 3.5758, "step": 11806 }, { "epoch": 2.25, "learning_rate": 2.327530981887512e-06, "loss": 2.6592, "step": 11807 }, { "epoch": 2.25, "learning_rate": 2.327473784556721e-06, "loss": 2.4457, "step": 11808 }, { "epoch": 2.25, "learning_rate": 2.3274165872259296e-06, "loss": 2.3935, "step": 11809 }, { "epoch": 2.25, "learning_rate": 2.3273593898951384e-06, "loss": 3.4264, "step": 11810 }, { "epoch": 2.25, "learning_rate": 2.3273021925643468e-06, "loss": 2.8961, "step": 11811 }, { "epoch": 2.25, "learning_rate": 2.327244995233556e-06, "loss": 2.4374, "step": 11812 }, { "epoch": 2.25, "learning_rate": 2.3271877979027648e-06, "loss": 2.5314, "step": 11813 }, { "epoch": 2.25, "learning_rate": 2.327130600571973e-06, "loss": 2.6626, "step": 11814 }, { "epoch": 2.25, "learning_rate": 2.3270734032411824e-06, "loss": 2.9473, "step": 11815 }, { "epoch": 2.25, "learning_rate": 2.3270162059103907e-06, "loss": 1.8853, "step": 11816 }, { "epoch": 2.25, "learning_rate": 2.3269590085795995e-06, "loss": 2.8497, "step": 11817 }, { "epoch": 2.25, "learning_rate": 2.3269018112488083e-06, "loss": 2.6669, "step": 11818 }, { "epoch": 2.25, "learning_rate": 2.326844613918017e-06, "loss": 2.7443, "step": 11819 }, { "epoch": 2.25, "learning_rate": 2.326787416587226e-06, "loss": 2.8016, "step": 11820 }, { "epoch": 2.25, "learning_rate": 2.3267302192564347e-06, "loss": 2.9627, "step": 11821 }, { "epoch": 2.25, "learning_rate": 2.3266730219256435e-06, "loss": 2.8026, "step": 11822 }, { "epoch": 2.25, "learning_rate": 2.3266158245948523e-06, "loss": 2.2064, "step": 11823 }, { "epoch": 2.25, "learning_rate": 2.326558627264061e-06, "loss": 3.1523, "step": 11824 }, { "epoch": 2.25, "learning_rate": 2.32650142993327e-06, "loss": 2.2711, "step": 11825 }, { "epoch": 2.25, "learning_rate": 2.3264442326024787e-06, "loss": 2.4509, "step": 11826 }, { "epoch": 2.25, "learning_rate": 2.3263870352716875e-06, "loss": 2.3675, "step": 11827 }, { "epoch": 2.25, "learning_rate": 2.326329837940896e-06, "loss": 2.2769, "step": 11828 }, { "epoch": 2.25, "learning_rate": 2.326272640610105e-06, "loss": 3.3623, "step": 11829 }, { "epoch": 2.25, "learning_rate": 2.326215443279314e-06, "loss": 2.87, "step": 11830 }, { "epoch": 2.25, "learning_rate": 2.3261582459485222e-06, "loss": 3.1589, "step": 11831 }, { "epoch": 2.25, "learning_rate": 2.3261010486177314e-06, "loss": 2.3982, "step": 11832 }, { "epoch": 2.25, "learning_rate": 2.32604385128694e-06, "loss": 2.5545, "step": 11833 }, { "epoch": 2.25, "learning_rate": 2.3259866539561486e-06, "loss": 2.2439, "step": 11834 }, { "epoch": 2.25, "learning_rate": 2.325929456625358e-06, "loss": 2.7688, "step": 11835 }, { "epoch": 2.25, "learning_rate": 2.325872259294566e-06, "loss": 2.0107, "step": 11836 }, { "epoch": 2.25, "learning_rate": 2.3258150619637754e-06, "loss": 2.1256, "step": 11837 }, { "epoch": 2.25, "learning_rate": 2.3257578646329838e-06, "loss": 3.0499, "step": 11838 }, { "epoch": 2.26, "learning_rate": 2.3257006673021926e-06, "loss": 2.4055, "step": 11839 }, { "epoch": 2.26, "learning_rate": 2.3256434699714014e-06, "loss": 2.4265, "step": 11840 }, { "epoch": 2.26, "learning_rate": 2.32558627264061e-06, "loss": 2.2224, "step": 11841 }, { "epoch": 2.26, "learning_rate": 2.325529075309819e-06, "loss": 2.4696, "step": 11842 }, { "epoch": 2.26, "learning_rate": 2.3254718779790277e-06, "loss": 2.3561, "step": 11843 }, { "epoch": 2.26, "learning_rate": 2.3254146806482365e-06, "loss": 2.7847, "step": 11844 }, { "epoch": 2.26, "learning_rate": 2.3253574833174453e-06, "loss": 3.0338, "step": 11845 }, { "epoch": 2.26, "learning_rate": 2.325300285986654e-06, "loss": 3.8806, "step": 11846 }, { "epoch": 2.26, "learning_rate": 2.325243088655863e-06, "loss": 2.9899, "step": 11847 }, { "epoch": 2.26, "learning_rate": 2.3251858913250713e-06, "loss": 2.0841, "step": 11848 }, { "epoch": 2.26, "learning_rate": 2.3251286939942805e-06, "loss": 2.2107, "step": 11849 }, { "epoch": 2.26, "learning_rate": 2.325071496663489e-06, "loss": 2.5829, "step": 11850 }, { "epoch": 2.26, "learning_rate": 2.3250142993326977e-06, "loss": 3.885, "step": 11851 }, { "epoch": 2.26, "learning_rate": 2.324957102001907e-06, "loss": 2.9344, "step": 11852 }, { "epoch": 2.26, "learning_rate": 2.3248999046711153e-06, "loss": 2.6476, "step": 11853 }, { "epoch": 2.26, "learning_rate": 2.3248427073403245e-06, "loss": 2.635, "step": 11854 }, { "epoch": 2.26, "learning_rate": 2.324785510009533e-06, "loss": 2.3576, "step": 11855 }, { "epoch": 2.26, "learning_rate": 2.3247283126787416e-06, "loss": 2.377, "step": 11856 }, { "epoch": 2.26, "learning_rate": 2.3246711153479504e-06, "loss": 2.799, "step": 11857 }, { "epoch": 2.26, "learning_rate": 2.3246139180171592e-06, "loss": 2.283, "step": 11858 }, { "epoch": 2.26, "learning_rate": 2.324556720686368e-06, "loss": 2.2423, "step": 11859 }, { "epoch": 2.26, "learning_rate": 2.324499523355577e-06, "loss": 2.3764, "step": 11860 }, { "epoch": 2.26, "learning_rate": 2.3244423260247856e-06, "loss": 2.979, "step": 11861 }, { "epoch": 2.26, "learning_rate": 2.3243851286939944e-06, "loss": 3.4783, "step": 11862 }, { "epoch": 2.26, "learning_rate": 2.324327931363203e-06, "loss": 2.339, "step": 11863 }, { "epoch": 2.26, "learning_rate": 2.324270734032412e-06, "loss": 3.1761, "step": 11864 }, { "epoch": 2.26, "learning_rate": 2.3242135367016204e-06, "loss": 2.8624, "step": 11865 }, { "epoch": 2.26, "learning_rate": 2.3241563393708296e-06, "loss": 2.7707, "step": 11866 }, { "epoch": 2.26, "learning_rate": 2.324099142040038e-06, "loss": 2.6526, "step": 11867 }, { "epoch": 2.26, "learning_rate": 2.3240419447092467e-06, "loss": 3.2398, "step": 11868 }, { "epoch": 2.26, "learning_rate": 2.323984747378456e-06, "loss": 3.1312, "step": 11869 }, { "epoch": 2.26, "learning_rate": 2.3239275500476643e-06, "loss": 2.4637, "step": 11870 }, { "epoch": 2.26, "learning_rate": 2.3238703527168735e-06, "loss": 2.5659, "step": 11871 }, { "epoch": 2.26, "learning_rate": 2.323813155386082e-06, "loss": 2.4399, "step": 11872 }, { "epoch": 2.26, "learning_rate": 2.3237559580552907e-06, "loss": 2.6078, "step": 11873 }, { "epoch": 2.26, "learning_rate": 2.3236987607244995e-06, "loss": 2.6629, "step": 11874 }, { "epoch": 2.26, "learning_rate": 2.3236415633937083e-06, "loss": 2.5964, "step": 11875 }, { "epoch": 2.26, "learning_rate": 2.323584366062917e-06, "loss": 2.197, "step": 11876 }, { "epoch": 2.26, "learning_rate": 2.323527168732126e-06, "loss": 2.5896, "step": 11877 }, { "epoch": 2.26, "learning_rate": 2.3234699714013347e-06, "loss": 2.2196, "step": 11878 }, { "epoch": 2.26, "learning_rate": 2.3234127740705435e-06, "loss": 2.8821, "step": 11879 }, { "epoch": 2.26, "learning_rate": 2.3233555767397523e-06, "loss": 2.5788, "step": 11880 }, { "epoch": 2.26, "learning_rate": 2.323298379408961e-06, "loss": 2.2994, "step": 11881 }, { "epoch": 2.26, "learning_rate": 2.3232411820781694e-06, "loss": 2.7656, "step": 11882 }, { "epoch": 2.26, "learning_rate": 2.3231839847473786e-06, "loss": 2.9238, "step": 11883 }, { "epoch": 2.26, "learning_rate": 2.3231267874165874e-06, "loss": 3.5433, "step": 11884 }, { "epoch": 2.26, "learning_rate": 2.323069590085796e-06, "loss": 2.8182, "step": 11885 }, { "epoch": 2.26, "learning_rate": 2.323012392755005e-06, "loss": 3.3749, "step": 11886 }, { "epoch": 2.26, "learning_rate": 2.3229551954242134e-06, "loss": 3.0205, "step": 11887 }, { "epoch": 2.26, "learning_rate": 2.3228979980934226e-06, "loss": 2.7694, "step": 11888 }, { "epoch": 2.26, "learning_rate": 2.322840800762631e-06, "loss": 3.2295, "step": 11889 }, { "epoch": 2.26, "learning_rate": 2.3227836034318398e-06, "loss": 2.3207, "step": 11890 }, { "epoch": 2.26, "learning_rate": 2.322726406101049e-06, "loss": 1.9482, "step": 11891 }, { "epoch": 2.27, "learning_rate": 2.3226692087702574e-06, "loss": 1.9424, "step": 11892 }, { "epoch": 2.27, "learning_rate": 2.322612011439466e-06, "loss": 2.0536, "step": 11893 }, { "epoch": 2.27, "learning_rate": 2.322554814108675e-06, "loss": 2.8294, "step": 11894 }, { "epoch": 2.27, "learning_rate": 2.3224976167778838e-06, "loss": 2.58, "step": 11895 }, { "epoch": 2.27, "learning_rate": 2.3224404194470925e-06, "loss": 2.9312, "step": 11896 }, { "epoch": 2.27, "learning_rate": 2.3223832221163013e-06, "loss": 2.2986, "step": 11897 }, { "epoch": 2.27, "learning_rate": 2.32232602478551e-06, "loss": 3.0607, "step": 11898 }, { "epoch": 2.27, "learning_rate": 2.3222688274547185e-06, "loss": 2.9123, "step": 11899 }, { "epoch": 2.27, "learning_rate": 2.3222116301239277e-06, "loss": 3.0607, "step": 11900 }, { "epoch": 2.27, "learning_rate": 2.3221544327931365e-06, "loss": 2.2903, "step": 11901 }, { "epoch": 2.27, "learning_rate": 2.3220972354623453e-06, "loss": 3.4013, "step": 11902 }, { "epoch": 2.27, "learning_rate": 2.322040038131554e-06, "loss": 2.8033, "step": 11903 }, { "epoch": 2.27, "learning_rate": 2.3219828408007625e-06, "loss": 2.472, "step": 11904 }, { "epoch": 2.27, "learning_rate": 2.3219256434699717e-06, "loss": 2.731, "step": 11905 }, { "epoch": 2.27, "learning_rate": 2.32186844613918e-06, "loss": 2.8257, "step": 11906 }, { "epoch": 2.27, "learning_rate": 2.321811248808389e-06, "loss": 2.9699, "step": 11907 }, { "epoch": 2.27, "learning_rate": 2.321754051477598e-06, "loss": 2.7181, "step": 11908 }, { "epoch": 2.27, "learning_rate": 2.3216968541468064e-06, "loss": 2.4012, "step": 11909 }, { "epoch": 2.27, "learning_rate": 2.3216396568160152e-06, "loss": 3.0743, "step": 11910 }, { "epoch": 2.27, "learning_rate": 2.321582459485224e-06, "loss": 2.7029, "step": 11911 }, { "epoch": 2.27, "learning_rate": 2.321525262154433e-06, "loss": 2.8107, "step": 11912 }, { "epoch": 2.27, "learning_rate": 2.3214680648236416e-06, "loss": 1.695, "step": 11913 }, { "epoch": 2.27, "learning_rate": 2.3214108674928504e-06, "loss": 3.4912, "step": 11914 }, { "epoch": 2.27, "learning_rate": 2.321353670162059e-06, "loss": 3.1119, "step": 11915 }, { "epoch": 2.27, "learning_rate": 2.3212964728312676e-06, "loss": 2.5502, "step": 11916 }, { "epoch": 2.27, "learning_rate": 2.321239275500477e-06, "loss": 2.6114, "step": 11917 }, { "epoch": 2.27, "learning_rate": 2.3211820781696856e-06, "loss": 2.4533, "step": 11918 }, { "epoch": 2.27, "learning_rate": 2.3211248808388944e-06, "loss": 2.9394, "step": 11919 }, { "epoch": 2.27, "learning_rate": 2.321067683508103e-06, "loss": 2.9259, "step": 11920 }, { "epoch": 2.27, "learning_rate": 2.3210104861773115e-06, "loss": 3.0368, "step": 11921 }, { "epoch": 2.27, "learning_rate": 2.3209532888465208e-06, "loss": 2.8368, "step": 11922 }, { "epoch": 2.27, "learning_rate": 2.3208960915157296e-06, "loss": 2.881, "step": 11923 }, { "epoch": 2.27, "learning_rate": 2.320838894184938e-06, "loss": 2.5891, "step": 11924 }, { "epoch": 2.27, "learning_rate": 2.320781696854147e-06, "loss": 2.6042, "step": 11925 }, { "epoch": 2.27, "learning_rate": 2.3207244995233555e-06, "loss": 2.6269, "step": 11926 }, { "epoch": 2.27, "learning_rate": 2.3206673021925643e-06, "loss": 3.01, "step": 11927 }, { "epoch": 2.27, "learning_rate": 2.320610104861773e-06, "loss": 2.4867, "step": 11928 }, { "epoch": 2.27, "learning_rate": 2.320552907530982e-06, "loss": 1.764, "step": 11929 }, { "epoch": 2.27, "learning_rate": 2.3204957102001907e-06, "loss": 2.8351, "step": 11930 }, { "epoch": 2.27, "learning_rate": 2.3204385128693995e-06, "loss": 2.3817, "step": 11931 }, { "epoch": 2.27, "learning_rate": 2.3203813155386083e-06, "loss": 1.9982, "step": 11932 }, { "epoch": 2.27, "learning_rate": 2.320324118207817e-06, "loss": 3.6711, "step": 11933 }, { "epoch": 2.27, "learning_rate": 2.320266920877026e-06, "loss": 3.0092, "step": 11934 }, { "epoch": 2.27, "learning_rate": 2.3202097235462347e-06, "loss": 3.3142, "step": 11935 }, { "epoch": 2.27, "learning_rate": 2.3201525262154435e-06, "loss": 2.3557, "step": 11936 }, { "epoch": 2.27, "learning_rate": 2.3200953288846522e-06, "loss": 2.1099, "step": 11937 }, { "epoch": 2.27, "learning_rate": 2.3200381315538606e-06, "loss": 2.352, "step": 11938 }, { "epoch": 2.27, "learning_rate": 2.31998093422307e-06, "loss": 1.7082, "step": 11939 }, { "epoch": 2.27, "learning_rate": 2.3199237368922786e-06, "loss": 2.6497, "step": 11940 }, { "epoch": 2.27, "learning_rate": 2.319866539561487e-06, "loss": 3.4159, "step": 11941 }, { "epoch": 2.27, "learning_rate": 2.3198093422306962e-06, "loss": 3.0034, "step": 11942 }, { "epoch": 2.27, "learning_rate": 2.3197521448999046e-06, "loss": 3.1136, "step": 11943 }, { "epoch": 2.28, "learning_rate": 2.3196949475691134e-06, "loss": 2.5666, "step": 11944 }, { "epoch": 2.28, "learning_rate": 2.319637750238322e-06, "loss": 2.3687, "step": 11945 }, { "epoch": 2.28, "learning_rate": 2.319580552907531e-06, "loss": 2.9091, "step": 11946 }, { "epoch": 2.28, "learning_rate": 2.3195233555767398e-06, "loss": 3.6745, "step": 11947 }, { "epoch": 2.28, "learning_rate": 2.3194661582459486e-06, "loss": 2.6858, "step": 11948 }, { "epoch": 2.28, "learning_rate": 2.3194089609151573e-06, "loss": 2.2176, "step": 11949 }, { "epoch": 2.28, "learning_rate": 2.319351763584366e-06, "loss": 2.3988, "step": 11950 }, { "epoch": 2.28, "learning_rate": 2.319294566253575e-06, "loss": 1.8596, "step": 11951 }, { "epoch": 2.28, "learning_rate": 2.3192373689227837e-06, "loss": 2.8669, "step": 11952 }, { "epoch": 2.28, "learning_rate": 2.3191801715919925e-06, "loss": 3.2072, "step": 11953 }, { "epoch": 2.28, "learning_rate": 2.3191229742612013e-06, "loss": 2.8928, "step": 11954 }, { "epoch": 2.28, "learning_rate": 2.3190657769304097e-06, "loss": 2.7475, "step": 11955 }, { "epoch": 2.28, "learning_rate": 2.319008579599619e-06, "loss": 2.8308, "step": 11956 }, { "epoch": 2.28, "learning_rate": 2.3189513822688277e-06, "loss": 2.5856, "step": 11957 }, { "epoch": 2.28, "learning_rate": 2.318894184938036e-06, "loss": 3.1165, "step": 11958 }, { "epoch": 2.28, "learning_rate": 2.3188369876072453e-06, "loss": 2.7069, "step": 11959 }, { "epoch": 2.28, "learning_rate": 2.3187797902764537e-06, "loss": 2.9967, "step": 11960 }, { "epoch": 2.28, "learning_rate": 2.3187225929456624e-06, "loss": 3.0434, "step": 11961 }, { "epoch": 2.28, "learning_rate": 2.3186653956148712e-06, "loss": 2.5086, "step": 11962 }, { "epoch": 2.28, "learning_rate": 2.31860819828408e-06, "loss": 3.6381, "step": 11963 }, { "epoch": 2.28, "learning_rate": 2.318551000953289e-06, "loss": 2.8796, "step": 11964 }, { "epoch": 2.28, "learning_rate": 2.3184938036224976e-06, "loss": 2.7067, "step": 11965 }, { "epoch": 2.28, "learning_rate": 2.3184366062917064e-06, "loss": 2.2474, "step": 11966 }, { "epoch": 2.28, "learning_rate": 2.318379408960915e-06, "loss": 2.3137, "step": 11967 }, { "epoch": 2.28, "learning_rate": 2.318322211630124e-06, "loss": 2.2352, "step": 11968 }, { "epoch": 2.28, "learning_rate": 2.318265014299333e-06, "loss": 2.6486, "step": 11969 }, { "epoch": 2.28, "learning_rate": 2.3182078169685416e-06, "loss": 3.4348, "step": 11970 }, { "epoch": 2.28, "learning_rate": 2.3181506196377504e-06, "loss": 3.7181, "step": 11971 }, { "epoch": 2.28, "learning_rate": 2.318093422306959e-06, "loss": 2.6322, "step": 11972 }, { "epoch": 2.28, "learning_rate": 2.318036224976168e-06, "loss": 1.9632, "step": 11973 }, { "epoch": 2.28, "learning_rate": 2.3179790276453768e-06, "loss": 2.8157, "step": 11974 }, { "epoch": 2.28, "learning_rate": 2.317921830314585e-06, "loss": 2.8585, "step": 11975 }, { "epoch": 2.28, "learning_rate": 2.3178646329837944e-06, "loss": 3.0123, "step": 11976 }, { "epoch": 2.28, "learning_rate": 2.3178074356530027e-06, "loss": 2.5292, "step": 11977 }, { "epoch": 2.28, "learning_rate": 2.3177502383222115e-06, "loss": 2.9181, "step": 11978 }, { "epoch": 2.28, "learning_rate": 2.3176930409914207e-06, "loss": 2.056, "step": 11979 }, { "epoch": 2.28, "learning_rate": 2.317635843660629e-06, "loss": 2.3917, "step": 11980 }, { "epoch": 2.28, "learning_rate": 2.317578646329838e-06, "loss": 2.4503, "step": 11981 }, { "epoch": 2.28, "learning_rate": 2.3175214489990467e-06, "loss": 2.0179, "step": 11982 }, { "epoch": 2.28, "learning_rate": 2.3174642516682555e-06, "loss": 2.6312, "step": 11983 }, { "epoch": 2.28, "learning_rate": 2.3174070543374643e-06, "loss": 3.0162, "step": 11984 }, { "epoch": 2.28, "learning_rate": 2.317349857006673e-06, "loss": 2.8129, "step": 11985 }, { "epoch": 2.28, "learning_rate": 2.317292659675882e-06, "loss": 3.1496, "step": 11986 }, { "epoch": 2.28, "learning_rate": 2.3172354623450907e-06, "loss": 3.4527, "step": 11987 }, { "epoch": 2.28, "learning_rate": 2.3171782650142995e-06, "loss": 2.8901, "step": 11988 }, { "epoch": 2.28, "learning_rate": 2.3171210676835083e-06, "loss": 3.112, "step": 11989 }, { "epoch": 2.28, "learning_rate": 2.317063870352717e-06, "loss": 2.0538, "step": 11990 }, { "epoch": 2.28, "learning_rate": 2.317006673021926e-06, "loss": 3.7181, "step": 11991 }, { "epoch": 2.28, "learning_rate": 2.316949475691134e-06, "loss": 2.3441, "step": 11992 }, { "epoch": 2.28, "learning_rate": 2.3168922783603434e-06, "loss": 2.1014, "step": 11993 }, { "epoch": 2.28, "learning_rate": 2.316835081029552e-06, "loss": 2.5478, "step": 11994 }, { "epoch": 2.28, "learning_rate": 2.3167778836987606e-06, "loss": 2.9345, "step": 11995 }, { "epoch": 2.28, "learning_rate": 2.31672068636797e-06, "loss": 2.9047, "step": 11996 }, { "epoch": 2.29, "learning_rate": 2.316663489037178e-06, "loss": 3.337, "step": 11997 }, { "epoch": 2.29, "learning_rate": 2.316606291706387e-06, "loss": 3.0375, "step": 11998 }, { "epoch": 2.29, "learning_rate": 2.3165490943755958e-06, "loss": 2.6986, "step": 11999 }, { "epoch": 2.29, "learning_rate": 2.3164918970448046e-06, "loss": 2.4715, "step": 12000 }, { "epoch": 2.29, "learning_rate": 2.3164346997140134e-06, "loss": 2.9375, "step": 12001 }, { "epoch": 2.29, "learning_rate": 2.316377502383222e-06, "loss": 2.3018, "step": 12002 }, { "epoch": 2.29, "learning_rate": 2.316320305052431e-06, "loss": 2.2851, "step": 12003 }, { "epoch": 2.29, "learning_rate": 2.3162631077216397e-06, "loss": 3.1655, "step": 12004 }, { "epoch": 2.29, "learning_rate": 2.3162059103908485e-06, "loss": 2.6752, "step": 12005 }, { "epoch": 2.29, "learning_rate": 2.3161487130600573e-06, "loss": 2.6884, "step": 12006 }, { "epoch": 2.29, "learning_rate": 2.316091515729266e-06, "loss": 2.758, "step": 12007 }, { "epoch": 2.29, "learning_rate": 2.316034318398475e-06, "loss": 2.6374, "step": 12008 }, { "epoch": 2.29, "learning_rate": 2.3159771210676833e-06, "loss": 2.7493, "step": 12009 }, { "epoch": 2.29, "learning_rate": 2.3159199237368925e-06, "loss": 3.1422, "step": 12010 }, { "epoch": 2.29, "learning_rate": 2.3158627264061013e-06, "loss": 2.2805, "step": 12011 }, { "epoch": 2.29, "learning_rate": 2.3158055290753097e-06, "loss": 3.0259, "step": 12012 }, { "epoch": 2.29, "learning_rate": 2.315748331744519e-06, "loss": 3.0737, "step": 12013 }, { "epoch": 2.29, "learning_rate": 2.3156911344137273e-06, "loss": 2.4315, "step": 12014 }, { "epoch": 2.29, "learning_rate": 2.315633937082936e-06, "loss": 2.3205, "step": 12015 }, { "epoch": 2.29, "learning_rate": 2.315576739752145e-06, "loss": 2.4826, "step": 12016 }, { "epoch": 2.29, "learning_rate": 2.3155195424213536e-06, "loss": 2.6709, "step": 12017 }, { "epoch": 2.29, "learning_rate": 2.315462345090563e-06, "loss": 3.1506, "step": 12018 }, { "epoch": 2.29, "learning_rate": 2.3154051477597712e-06, "loss": 2.1468, "step": 12019 }, { "epoch": 2.29, "learning_rate": 2.31534795042898e-06, "loss": 3.647, "step": 12020 }, { "epoch": 2.29, "learning_rate": 2.315290753098189e-06, "loss": 2.8894, "step": 12021 }, { "epoch": 2.29, "learning_rate": 2.3152335557673976e-06, "loss": 2.3831, "step": 12022 }, { "epoch": 2.29, "learning_rate": 2.3151763584366064e-06, "loss": 2.2031, "step": 12023 }, { "epoch": 2.29, "learning_rate": 2.315119161105815e-06, "loss": 2.8118, "step": 12024 }, { "epoch": 2.29, "learning_rate": 2.315061963775024e-06, "loss": 3.0113, "step": 12025 }, { "epoch": 2.29, "learning_rate": 2.3150047664442324e-06, "loss": 3.0231, "step": 12026 }, { "epoch": 2.29, "learning_rate": 2.3149475691134416e-06, "loss": 2.554, "step": 12027 }, { "epoch": 2.29, "learning_rate": 2.3148903717826504e-06, "loss": 2.4058, "step": 12028 }, { "epoch": 2.29, "learning_rate": 2.3148331744518587e-06, "loss": 2.9789, "step": 12029 }, { "epoch": 2.29, "learning_rate": 2.314775977121068e-06, "loss": 2.5473, "step": 12030 }, { "epoch": 2.29, "learning_rate": 2.3147187797902763e-06, "loss": 2.2363, "step": 12031 }, { "epoch": 2.29, "learning_rate": 2.3146615824594855e-06, "loss": 2.2532, "step": 12032 }, { "epoch": 2.29, "learning_rate": 2.314604385128694e-06, "loss": 3.409, "step": 12033 }, { "epoch": 2.29, "learning_rate": 2.3145471877979027e-06, "loss": 2.6637, "step": 12034 }, { "epoch": 2.29, "learning_rate": 2.314489990467112e-06, "loss": 2.3492, "step": 12035 }, { "epoch": 2.29, "learning_rate": 2.3144327931363203e-06, "loss": 2.7628, "step": 12036 }, { "epoch": 2.29, "learning_rate": 2.314375595805529e-06, "loss": 3.3154, "step": 12037 }, { "epoch": 2.29, "learning_rate": 2.314318398474738e-06, "loss": 2.2699, "step": 12038 }, { "epoch": 2.29, "learning_rate": 2.3142612011439467e-06, "loss": 2.3415, "step": 12039 }, { "epoch": 2.29, "learning_rate": 2.3142040038131555e-06, "loss": 3.2389, "step": 12040 }, { "epoch": 2.29, "learning_rate": 2.3141468064823643e-06, "loss": 2.6814, "step": 12041 }, { "epoch": 2.29, "learning_rate": 2.314089609151573e-06, "loss": 2.6604, "step": 12042 }, { "epoch": 2.29, "learning_rate": 2.3140324118207814e-06, "loss": 2.9646, "step": 12043 }, { "epoch": 2.29, "learning_rate": 2.3139752144899906e-06, "loss": 2.0357, "step": 12044 }, { "epoch": 2.29, "learning_rate": 2.3139180171591994e-06, "loss": 3.1324, "step": 12045 }, { "epoch": 2.29, "learning_rate": 2.313860819828408e-06, "loss": 3.4445, "step": 12046 }, { "epoch": 2.29, "learning_rate": 2.313803622497617e-06, "loss": 2.9543, "step": 12047 }, { "epoch": 2.29, "learning_rate": 2.3137464251668254e-06, "loss": 2.4353, "step": 12048 }, { "epoch": 2.3, "learning_rate": 2.3136892278360346e-06, "loss": 3.0282, "step": 12049 }, { "epoch": 2.3, "learning_rate": 2.313632030505243e-06, "loss": 2.577, "step": 12050 }, { "epoch": 2.3, "learning_rate": 2.3135748331744518e-06, "loss": 2.4473, "step": 12051 }, { "epoch": 2.3, "learning_rate": 2.313517635843661e-06, "loss": 3.708, "step": 12052 }, { "epoch": 2.3, "learning_rate": 2.3134604385128694e-06, "loss": 2.0648, "step": 12053 }, { "epoch": 2.3, "learning_rate": 2.313403241182078e-06, "loss": 2.9037, "step": 12054 }, { "epoch": 2.3, "learning_rate": 2.313346043851287e-06, "loss": 2.7254, "step": 12055 }, { "epoch": 2.3, "learning_rate": 2.3132888465204957e-06, "loss": 2.3242, "step": 12056 }, { "epoch": 2.3, "learning_rate": 2.3132316491897045e-06, "loss": 2.5912, "step": 12057 }, { "epoch": 2.3, "learning_rate": 2.3131744518589133e-06, "loss": 3.7479, "step": 12058 }, { "epoch": 2.3, "learning_rate": 2.313117254528122e-06, "loss": 2.4332, "step": 12059 }, { "epoch": 2.3, "learning_rate": 2.313060057197331e-06, "loss": 3.1348, "step": 12060 }, { "epoch": 2.3, "learning_rate": 2.3130028598665397e-06, "loss": 2.7827, "step": 12061 }, { "epoch": 2.3, "learning_rate": 2.3129456625357485e-06, "loss": 3.1356, "step": 12062 }, { "epoch": 2.3, "learning_rate": 2.312888465204957e-06, "loss": 2.6566, "step": 12063 }, { "epoch": 2.3, "learning_rate": 2.312831267874166e-06, "loss": 2.7487, "step": 12064 }, { "epoch": 2.3, "learning_rate": 2.3127740705433745e-06, "loss": 3.1788, "step": 12065 }, { "epoch": 2.3, "learning_rate": 2.3127168732125837e-06, "loss": 3.9856, "step": 12066 }, { "epoch": 2.3, "learning_rate": 2.3126596758817925e-06, "loss": 2.2756, "step": 12067 }, { "epoch": 2.3, "learning_rate": 2.312602478551001e-06, "loss": 2.6453, "step": 12068 }, { "epoch": 2.3, "learning_rate": 2.31254528122021e-06, "loss": 2.672, "step": 12069 }, { "epoch": 2.3, "learning_rate": 2.3124880838894184e-06, "loss": 2.9016, "step": 12070 }, { "epoch": 2.3, "learning_rate": 2.3124308865586272e-06, "loss": 2.9415, "step": 12071 }, { "epoch": 2.3, "learning_rate": 2.312373689227836e-06, "loss": 2.9594, "step": 12072 }, { "epoch": 2.3, "learning_rate": 2.312316491897045e-06, "loss": 3.44, "step": 12073 }, { "epoch": 2.3, "learning_rate": 2.3122592945662536e-06, "loss": 2.3236, "step": 12074 }, { "epoch": 2.3, "learning_rate": 2.3122020972354624e-06, "loss": 2.6318, "step": 12075 }, { "epoch": 2.3, "learning_rate": 2.312144899904671e-06, "loss": 3.2719, "step": 12076 }, { "epoch": 2.3, "learning_rate": 2.31208770257388e-06, "loss": 2.4548, "step": 12077 }, { "epoch": 2.3, "learning_rate": 2.3120305052430888e-06, "loss": 4.1492, "step": 12078 }, { "epoch": 2.3, "learning_rate": 2.3119733079122976e-06, "loss": 2.6816, "step": 12079 }, { "epoch": 2.3, "learning_rate": 2.311916110581506e-06, "loss": 2.7736, "step": 12080 }, { "epoch": 2.3, "learning_rate": 2.311858913250715e-06, "loss": 3.4436, "step": 12081 }, { "epoch": 2.3, "learning_rate": 2.3118017159199235e-06, "loss": 2.5225, "step": 12082 }, { "epoch": 2.3, "learning_rate": 2.3117445185891328e-06, "loss": 3.0808, "step": 12083 }, { "epoch": 2.3, "learning_rate": 2.3116873212583415e-06, "loss": 2.7637, "step": 12084 }, { "epoch": 2.3, "learning_rate": 2.31163012392755e-06, "loss": 3.2923, "step": 12085 }, { "epoch": 2.3, "learning_rate": 2.311572926596759e-06, "loss": 2.2832, "step": 12086 }, { "epoch": 2.3, "learning_rate": 2.3115157292659675e-06, "loss": 3.2383, "step": 12087 }, { "epoch": 2.3, "learning_rate": 2.3114585319351763e-06, "loss": 3.3711, "step": 12088 }, { "epoch": 2.3, "learning_rate": 2.311401334604385e-06, "loss": 2.8964, "step": 12089 }, { "epoch": 2.3, "learning_rate": 2.311344137273594e-06, "loss": 2.8851, "step": 12090 }, { "epoch": 2.3, "learning_rate": 2.3112869399428027e-06, "loss": 3.239, "step": 12091 }, { "epoch": 2.3, "learning_rate": 2.3112297426120115e-06, "loss": 2.6554, "step": 12092 }, { "epoch": 2.3, "learning_rate": 2.3111725452812203e-06, "loss": 1.8659, "step": 12093 }, { "epoch": 2.3, "learning_rate": 2.311115347950429e-06, "loss": 2.255, "step": 12094 }, { "epoch": 2.3, "learning_rate": 2.311058150619638e-06, "loss": 3.5439, "step": 12095 }, { "epoch": 2.3, "learning_rate": 2.3110009532888466e-06, "loss": 3.3676, "step": 12096 }, { "epoch": 2.3, "learning_rate": 2.3109437559580554e-06, "loss": 1.6657, "step": 12097 }, { "epoch": 2.3, "learning_rate": 2.3108865586272642e-06, "loss": 2.9723, "step": 12098 }, { "epoch": 2.3, "learning_rate": 2.310829361296473e-06, "loss": 3.2074, "step": 12099 }, { "epoch": 2.3, "learning_rate": 2.310772163965682e-06, "loss": 2.706, "step": 12100 }, { "epoch": 2.3, "learning_rate": 2.3107149666348906e-06, "loss": 3.2803, "step": 12101 }, { "epoch": 2.31, "learning_rate": 2.310657769304099e-06, "loss": 3.1578, "step": 12102 }, { "epoch": 2.31, "learning_rate": 2.310600571973308e-06, "loss": 3.0228, "step": 12103 }, { "epoch": 2.31, "learning_rate": 2.3105433746425166e-06, "loss": 2.7625, "step": 12104 }, { "epoch": 2.31, "learning_rate": 2.3104861773117254e-06, "loss": 2.8735, "step": 12105 }, { "epoch": 2.31, "learning_rate": 2.3104289799809346e-06, "loss": 1.8204, "step": 12106 }, { "epoch": 2.31, "learning_rate": 2.310371782650143e-06, "loss": 1.642, "step": 12107 }, { "epoch": 2.31, "learning_rate": 2.3103145853193518e-06, "loss": 2.6037, "step": 12108 }, { "epoch": 2.31, "learning_rate": 2.3102573879885605e-06, "loss": 3.2084, "step": 12109 }, { "epoch": 2.31, "learning_rate": 2.3102001906577693e-06, "loss": 1.868, "step": 12110 }, { "epoch": 2.31, "learning_rate": 2.310142993326978e-06, "loss": 2.2678, "step": 12111 }, { "epoch": 2.31, "learning_rate": 2.310085795996187e-06, "loss": 2.8787, "step": 12112 }, { "epoch": 2.31, "learning_rate": 2.3100285986653957e-06, "loss": 3.3032, "step": 12113 }, { "epoch": 2.31, "learning_rate": 2.3099714013346045e-06, "loss": 2.69, "step": 12114 }, { "epoch": 2.31, "learning_rate": 2.3099142040038133e-06, "loss": 2.6358, "step": 12115 }, { "epoch": 2.31, "learning_rate": 2.309857006673022e-06, "loss": 2.9957, "step": 12116 }, { "epoch": 2.31, "learning_rate": 2.309799809342231e-06, "loss": 3.2573, "step": 12117 }, { "epoch": 2.31, "learning_rate": 2.3097426120114397e-06, "loss": 2.972, "step": 12118 }, { "epoch": 2.31, "learning_rate": 2.309685414680648e-06, "loss": 2.7174, "step": 12119 }, { "epoch": 2.31, "learning_rate": 2.3096282173498573e-06, "loss": 3.4208, "step": 12120 }, { "epoch": 2.31, "learning_rate": 2.3095710200190656e-06, "loss": 1.9328, "step": 12121 }, { "epoch": 2.31, "learning_rate": 2.3095138226882744e-06, "loss": 3.1522, "step": 12122 }, { "epoch": 2.31, "learning_rate": 2.3094566253574837e-06, "loss": 3.3386, "step": 12123 }, { "epoch": 2.31, "learning_rate": 2.309399428026692e-06, "loss": 1.7446, "step": 12124 }, { "epoch": 2.31, "learning_rate": 2.309342230695901e-06, "loss": 2.9284, "step": 12125 }, { "epoch": 2.31, "learning_rate": 2.3092850333651096e-06, "loss": 2.956, "step": 12126 }, { "epoch": 2.31, "learning_rate": 2.3092278360343184e-06, "loss": 3.5186, "step": 12127 }, { "epoch": 2.31, "learning_rate": 2.309170638703527e-06, "loss": 3.1709, "step": 12128 }, { "epoch": 2.31, "learning_rate": 2.309113441372736e-06, "loss": 2.7518, "step": 12129 }, { "epoch": 2.31, "learning_rate": 2.309056244041945e-06, "loss": 2.9987, "step": 12130 }, { "epoch": 2.31, "learning_rate": 2.3089990467111536e-06, "loss": 2.2037, "step": 12131 }, { "epoch": 2.31, "learning_rate": 2.3089418493803624e-06, "loss": 2.8499, "step": 12132 }, { "epoch": 2.31, "learning_rate": 2.308884652049571e-06, "loss": 2.0512, "step": 12133 }, { "epoch": 2.31, "learning_rate": 2.30882745471878e-06, "loss": 2.9075, "step": 12134 }, { "epoch": 2.31, "learning_rate": 2.3087702573879888e-06, "loss": 2.6334, "step": 12135 }, { "epoch": 2.31, "learning_rate": 2.308713060057197e-06, "loss": 2.4791, "step": 12136 }, { "epoch": 2.31, "learning_rate": 2.3086558627264063e-06, "loss": 2.3866, "step": 12137 }, { "epoch": 2.31, "learning_rate": 2.3085986653956147e-06, "loss": 2.0806, "step": 12138 }, { "epoch": 2.31, "learning_rate": 2.3085414680648235e-06, "loss": 2.8498, "step": 12139 }, { "epoch": 2.31, "learning_rate": 2.3084842707340327e-06, "loss": 1.9272, "step": 12140 }, { "epoch": 2.31, "learning_rate": 2.308427073403241e-06, "loss": 2.3081, "step": 12141 }, { "epoch": 2.31, "learning_rate": 2.30836987607245e-06, "loss": 2.2649, "step": 12142 }, { "epoch": 2.31, "learning_rate": 2.3083126787416587e-06, "loss": 3.0954, "step": 12143 }, { "epoch": 2.31, "learning_rate": 2.3082554814108675e-06, "loss": 2.5363, "step": 12144 }, { "epoch": 2.31, "learning_rate": 2.3081982840800763e-06, "loss": 2.3582, "step": 12145 }, { "epoch": 2.31, "learning_rate": 2.308141086749285e-06, "loss": 2.6295, "step": 12146 }, { "epoch": 2.31, "learning_rate": 2.308083889418494e-06, "loss": 2.9951, "step": 12147 }, { "epoch": 2.31, "learning_rate": 2.3080266920877027e-06, "loss": 3.366, "step": 12148 }, { "epoch": 2.31, "learning_rate": 2.3079694947569115e-06, "loss": 1.6581, "step": 12149 }, { "epoch": 2.31, "learning_rate": 2.3079122974261202e-06, "loss": 2.4393, "step": 12150 }, { "epoch": 2.31, "learning_rate": 2.307855100095329e-06, "loss": 2.9556, "step": 12151 }, { "epoch": 2.31, "learning_rate": 2.307797902764538e-06, "loss": 2.2097, "step": 12152 }, { "epoch": 2.31, "learning_rate": 2.307740705433746e-06, "loss": 2.6023, "step": 12153 }, { "epoch": 2.32, "learning_rate": 2.3076835081029554e-06, "loss": 2.9891, "step": 12154 }, { "epoch": 2.32, "learning_rate": 2.3076263107721642e-06, "loss": 2.805, "step": 12155 }, { "epoch": 2.32, "learning_rate": 2.3075691134413726e-06, "loss": 2.7922, "step": 12156 }, { "epoch": 2.32, "learning_rate": 2.307511916110582e-06, "loss": 1.5526, "step": 12157 }, { "epoch": 2.32, "learning_rate": 2.30745471877979e-06, "loss": 2.7939, "step": 12158 }, { "epoch": 2.32, "learning_rate": 2.307397521448999e-06, "loss": 3.327, "step": 12159 }, { "epoch": 2.32, "learning_rate": 2.3073403241182078e-06, "loss": 2.997, "step": 12160 }, { "epoch": 2.32, "learning_rate": 2.3072831267874166e-06, "loss": 2.2731, "step": 12161 }, { "epoch": 2.32, "learning_rate": 2.3072259294566258e-06, "loss": 3.2716, "step": 12162 }, { "epoch": 2.32, "learning_rate": 2.307168732125834e-06, "loss": 3.3821, "step": 12163 }, { "epoch": 2.32, "learning_rate": 2.307111534795043e-06, "loss": 2.6067, "step": 12164 }, { "epoch": 2.32, "learning_rate": 2.3070543374642517e-06, "loss": 2.5785, "step": 12165 }, { "epoch": 2.32, "learning_rate": 2.3069971401334605e-06, "loss": 3.0788, "step": 12166 }, { "epoch": 2.32, "learning_rate": 2.3069399428026693e-06, "loss": 2.4361, "step": 12167 }, { "epoch": 2.32, "learning_rate": 2.306882745471878e-06, "loss": 2.89, "step": 12168 }, { "epoch": 2.32, "learning_rate": 2.306825548141087e-06, "loss": 2.3682, "step": 12169 }, { "epoch": 2.32, "learning_rate": 2.3067683508102953e-06, "loss": 3.3593, "step": 12170 }, { "epoch": 2.32, "learning_rate": 2.3067111534795045e-06, "loss": 3.2117, "step": 12171 }, { "epoch": 2.32, "learning_rate": 2.3066539561487133e-06, "loss": 2.6069, "step": 12172 }, { "epoch": 2.32, "learning_rate": 2.3065967588179217e-06, "loss": 2.1681, "step": 12173 }, { "epoch": 2.32, "learning_rate": 2.306539561487131e-06, "loss": 2.9049, "step": 12174 }, { "epoch": 2.32, "learning_rate": 2.3064823641563392e-06, "loss": 3.1557, "step": 12175 }, { "epoch": 2.32, "learning_rate": 2.306425166825548e-06, "loss": 2.5661, "step": 12176 }, { "epoch": 2.32, "learning_rate": 2.306367969494757e-06, "loss": 2.7962, "step": 12177 }, { "epoch": 2.32, "learning_rate": 2.3063107721639656e-06, "loss": 2.0518, "step": 12178 }, { "epoch": 2.32, "learning_rate": 2.306253574833175e-06, "loss": 2.7491, "step": 12179 }, { "epoch": 2.32, "learning_rate": 2.3061963775023832e-06, "loss": 3.4621, "step": 12180 }, { "epoch": 2.32, "learning_rate": 2.306139180171592e-06, "loss": 2.318, "step": 12181 }, { "epoch": 2.32, "learning_rate": 2.306081982840801e-06, "loss": 2.2412, "step": 12182 }, { "epoch": 2.32, "learning_rate": 2.3060247855100096e-06, "loss": 2.9277, "step": 12183 }, { "epoch": 2.32, "learning_rate": 2.3059675881792184e-06, "loss": 3.6254, "step": 12184 }, { "epoch": 2.32, "learning_rate": 2.305910390848427e-06, "loss": 3.4645, "step": 12185 }, { "epoch": 2.32, "learning_rate": 2.305853193517636e-06, "loss": 2.5281, "step": 12186 }, { "epoch": 2.32, "learning_rate": 2.3057959961868443e-06, "loss": 2.8406, "step": 12187 }, { "epoch": 2.32, "learning_rate": 2.3057387988560536e-06, "loss": 2.3489, "step": 12188 }, { "epoch": 2.32, "learning_rate": 2.3056816015252624e-06, "loss": 2.6667, "step": 12189 }, { "epoch": 2.32, "learning_rate": 2.3056244041944707e-06, "loss": 2.7751, "step": 12190 }, { "epoch": 2.32, "learning_rate": 2.30556720686368e-06, "loss": 3.3154, "step": 12191 }, { "epoch": 2.32, "learning_rate": 2.3055100095328883e-06, "loss": 3.0385, "step": 12192 }, { "epoch": 2.32, "learning_rate": 2.305452812202097e-06, "loss": 4.2413, "step": 12193 }, { "epoch": 2.32, "learning_rate": 2.3053956148713063e-06, "loss": 2.2097, "step": 12194 }, { "epoch": 2.32, "learning_rate": 2.3053384175405147e-06, "loss": 3.1742, "step": 12195 }, { "epoch": 2.32, "learning_rate": 2.305281220209724e-06, "loss": 3.0527, "step": 12196 }, { "epoch": 2.32, "learning_rate": 2.3052240228789323e-06, "loss": 2.6906, "step": 12197 }, { "epoch": 2.32, "learning_rate": 2.305166825548141e-06, "loss": 3.3315, "step": 12198 }, { "epoch": 2.32, "learning_rate": 2.30510962821735e-06, "loss": 3.2416, "step": 12199 }, { "epoch": 2.32, "learning_rate": 2.3050524308865587e-06, "loss": 2.959, "step": 12200 }, { "epoch": 2.32, "learning_rate": 2.3049952335557675e-06, "loss": 2.7224, "step": 12201 }, { "epoch": 2.32, "learning_rate": 2.3049380362249763e-06, "loss": 2.3285, "step": 12202 }, { "epoch": 2.32, "learning_rate": 2.304880838894185e-06, "loss": 3.3746, "step": 12203 }, { "epoch": 2.32, "learning_rate": 2.304823641563394e-06, "loss": 3.261, "step": 12204 }, { "epoch": 2.32, "learning_rate": 2.3047664442326026e-06, "loss": 3.0855, "step": 12205 }, { "epoch": 2.32, "learning_rate": 2.3047092469018114e-06, "loss": 2.3778, "step": 12206 }, { "epoch": 2.33, "learning_rate": 2.30465204957102e-06, "loss": 2.5784, "step": 12207 }, { "epoch": 2.33, "learning_rate": 2.304594852240229e-06, "loss": 3.1413, "step": 12208 }, { "epoch": 2.33, "learning_rate": 2.3045376549094374e-06, "loss": 1.7951, "step": 12209 }, { "epoch": 2.33, "learning_rate": 2.304480457578646e-06, "loss": 2.3389, "step": 12210 }, { "epoch": 2.33, "learning_rate": 2.3044232602478554e-06, "loss": 2.9624, "step": 12211 }, { "epoch": 2.33, "learning_rate": 2.3043660629170638e-06, "loss": 2.2945, "step": 12212 }, { "epoch": 2.33, "learning_rate": 2.304308865586273e-06, "loss": 2.0648, "step": 12213 }, { "epoch": 2.33, "learning_rate": 2.3042516682554814e-06, "loss": 3.0479, "step": 12214 }, { "epoch": 2.33, "learning_rate": 2.30419447092469e-06, "loss": 3.3177, "step": 12215 }, { "epoch": 2.33, "learning_rate": 2.304137273593899e-06, "loss": 2.5389, "step": 12216 }, { "epoch": 2.33, "learning_rate": 2.3040800762631077e-06, "loss": 3.1266, "step": 12217 }, { "epoch": 2.33, "learning_rate": 2.3040228789323165e-06, "loss": 3.6103, "step": 12218 }, { "epoch": 2.33, "learning_rate": 2.3039656816015253e-06, "loss": 2.9267, "step": 12219 }, { "epoch": 2.33, "learning_rate": 2.303908484270734e-06, "loss": 1.9914, "step": 12220 }, { "epoch": 2.33, "learning_rate": 2.303851286939943e-06, "loss": 3.2311, "step": 12221 }, { "epoch": 2.33, "learning_rate": 2.3037940896091517e-06, "loss": 2.4371, "step": 12222 }, { "epoch": 2.33, "learning_rate": 2.3037368922783605e-06, "loss": 2.2066, "step": 12223 }, { "epoch": 2.33, "learning_rate": 2.303679694947569e-06, "loss": 2.1145, "step": 12224 }, { "epoch": 2.33, "learning_rate": 2.303622497616778e-06, "loss": 2.8411, "step": 12225 }, { "epoch": 2.33, "learning_rate": 2.3035653002859865e-06, "loss": 1.8433, "step": 12226 }, { "epoch": 2.33, "learning_rate": 2.3035081029551953e-06, "loss": 1.9915, "step": 12227 }, { "epoch": 2.33, "learning_rate": 2.3034509056244045e-06, "loss": 3.3044, "step": 12228 }, { "epoch": 2.33, "learning_rate": 2.303393708293613e-06, "loss": 2.3721, "step": 12229 }, { "epoch": 2.33, "learning_rate": 2.303336510962822e-06, "loss": 2.9633, "step": 12230 }, { "epoch": 2.33, "learning_rate": 2.3032793136320304e-06, "loss": 2.5379, "step": 12231 }, { "epoch": 2.33, "learning_rate": 2.3032221163012392e-06, "loss": 2.1872, "step": 12232 }, { "epoch": 2.33, "learning_rate": 2.3031649189704484e-06, "loss": 1.8263, "step": 12233 }, { "epoch": 2.33, "learning_rate": 2.303107721639657e-06, "loss": 2.9128, "step": 12234 }, { "epoch": 2.33, "learning_rate": 2.3030505243088656e-06, "loss": 2.6146, "step": 12235 }, { "epoch": 2.33, "learning_rate": 2.3029933269780744e-06, "loss": 3.3187, "step": 12236 }, { "epoch": 2.33, "learning_rate": 2.302936129647283e-06, "loss": 3.6084, "step": 12237 }, { "epoch": 2.33, "learning_rate": 2.302878932316492e-06, "loss": 1.8078, "step": 12238 }, { "epoch": 2.33, "learning_rate": 2.3028217349857008e-06, "loss": 1.8994, "step": 12239 }, { "epoch": 2.33, "learning_rate": 2.3027645376549096e-06, "loss": 4.2725, "step": 12240 }, { "epoch": 2.33, "learning_rate": 2.302707340324118e-06, "loss": 2.8396, "step": 12241 }, { "epoch": 2.33, "learning_rate": 2.302650142993327e-06, "loss": 2.9764, "step": 12242 }, { "epoch": 2.33, "learning_rate": 2.302592945662536e-06, "loss": 3.0057, "step": 12243 }, { "epoch": 2.33, "learning_rate": 2.3025357483317447e-06, "loss": 3.2004, "step": 12244 }, { "epoch": 2.33, "learning_rate": 2.3024785510009535e-06, "loss": 2.7345, "step": 12245 }, { "epoch": 2.33, "learning_rate": 2.302421353670162e-06, "loss": 2.8265, "step": 12246 }, { "epoch": 2.33, "learning_rate": 2.302364156339371e-06, "loss": 2.6941, "step": 12247 }, { "epoch": 2.33, "learning_rate": 2.3023069590085795e-06, "loss": 2.4909, "step": 12248 }, { "epoch": 2.33, "learning_rate": 2.3022497616777883e-06, "loss": 3.3072, "step": 12249 }, { "epoch": 2.33, "learning_rate": 2.3021925643469975e-06, "loss": 2.6797, "step": 12250 }, { "epoch": 2.33, "learning_rate": 2.302135367016206e-06, "loss": 2.7973, "step": 12251 }, { "epoch": 2.33, "learning_rate": 2.3020781696854147e-06, "loss": 2.3104, "step": 12252 }, { "epoch": 2.33, "learning_rate": 2.3020209723546235e-06, "loss": 3.1936, "step": 12253 }, { "epoch": 2.33, "learning_rate": 2.3019637750238323e-06, "loss": 3.8849, "step": 12254 }, { "epoch": 2.33, "learning_rate": 2.301906577693041e-06, "loss": 1.9949, "step": 12255 }, { "epoch": 2.33, "learning_rate": 2.30184938036225e-06, "loss": 3.062, "step": 12256 }, { "epoch": 2.33, "learning_rate": 2.3017921830314586e-06, "loss": 3.1815, "step": 12257 }, { "epoch": 2.33, "learning_rate": 2.301734985700667e-06, "loss": 2.7363, "step": 12258 }, { "epoch": 2.34, "learning_rate": 2.3016777883698762e-06, "loss": 2.7529, "step": 12259 }, { "epoch": 2.34, "learning_rate": 2.301620591039085e-06, "loss": 3.2184, "step": 12260 }, { "epoch": 2.34, "learning_rate": 2.301563393708294e-06, "loss": 2.798, "step": 12261 }, { "epoch": 2.34, "learning_rate": 2.3015061963775026e-06, "loss": 2.992, "step": 12262 }, { "epoch": 2.34, "learning_rate": 2.301448999046711e-06, "loss": 2.9351, "step": 12263 }, { "epoch": 2.34, "learning_rate": 2.30139180171592e-06, "loss": 3.0945, "step": 12264 }, { "epoch": 2.34, "learning_rate": 2.3013346043851286e-06, "loss": 3.4338, "step": 12265 }, { "epoch": 2.34, "learning_rate": 2.3012774070543374e-06, "loss": 2.3608, "step": 12266 }, { "epoch": 2.34, "learning_rate": 2.3012202097235466e-06, "loss": 2.2897, "step": 12267 }, { "epoch": 2.34, "learning_rate": 2.301163012392755e-06, "loss": 3.2898, "step": 12268 }, { "epoch": 2.34, "learning_rate": 2.3011058150619637e-06, "loss": 2.6733, "step": 12269 }, { "epoch": 2.34, "learning_rate": 2.3010486177311725e-06, "loss": 3.6413, "step": 12270 }, { "epoch": 2.34, "learning_rate": 2.3009914204003813e-06, "loss": 2.9337, "step": 12271 }, { "epoch": 2.34, "learning_rate": 2.30093422306959e-06, "loss": 2.7375, "step": 12272 }, { "epoch": 2.34, "learning_rate": 2.300877025738799e-06, "loss": 1.8886, "step": 12273 }, { "epoch": 2.34, "learning_rate": 2.3008198284080077e-06, "loss": 2.9754, "step": 12274 }, { "epoch": 2.34, "learning_rate": 2.300762631077216e-06, "loss": 2.9702, "step": 12275 }, { "epoch": 2.34, "learning_rate": 2.3007054337464253e-06, "loss": 2.7569, "step": 12276 }, { "epoch": 2.34, "learning_rate": 2.300648236415634e-06, "loss": 2.9369, "step": 12277 }, { "epoch": 2.34, "learning_rate": 2.300591039084843e-06, "loss": 2.8108, "step": 12278 }, { "epoch": 2.34, "learning_rate": 2.3005338417540517e-06, "loss": 2.3825, "step": 12279 }, { "epoch": 2.34, "learning_rate": 2.30047664442326e-06, "loss": 3.9695, "step": 12280 }, { "epoch": 2.34, "learning_rate": 2.3004194470924693e-06, "loss": 3.043, "step": 12281 }, { "epoch": 2.34, "learning_rate": 2.300362249761678e-06, "loss": 2.4649, "step": 12282 }, { "epoch": 2.34, "learning_rate": 2.3003050524308864e-06, "loss": 2.8017, "step": 12283 }, { "epoch": 2.34, "learning_rate": 2.3002478551000957e-06, "loss": 2.5094, "step": 12284 }, { "epoch": 2.34, "learning_rate": 2.300190657769304e-06, "loss": 2.479, "step": 12285 }, { "epoch": 2.34, "learning_rate": 2.300133460438513e-06, "loss": 3.3587, "step": 12286 }, { "epoch": 2.34, "learning_rate": 2.3000762631077216e-06, "loss": 2.0351, "step": 12287 }, { "epoch": 2.34, "learning_rate": 2.3000190657769304e-06, "loss": 2.9689, "step": 12288 }, { "epoch": 2.34, "learning_rate": 2.299961868446139e-06, "loss": 2.5818, "step": 12289 }, { "epoch": 2.34, "learning_rate": 2.299904671115348e-06, "loss": 3.0661, "step": 12290 }, { "epoch": 2.34, "learning_rate": 2.2998474737845568e-06, "loss": 2.0488, "step": 12291 }, { "epoch": 2.34, "learning_rate": 2.2997902764537656e-06, "loss": 1.8935, "step": 12292 }, { "epoch": 2.34, "learning_rate": 2.2997330791229744e-06, "loss": 2.8025, "step": 12293 }, { "epoch": 2.34, "learning_rate": 2.299675881792183e-06, "loss": 4.4122, "step": 12294 }, { "epoch": 2.34, "learning_rate": 2.299618684461392e-06, "loss": 2.3952, "step": 12295 }, { "epoch": 2.34, "learning_rate": 2.2995614871306008e-06, "loss": 2.5091, "step": 12296 }, { "epoch": 2.34, "learning_rate": 2.299504289799809e-06, "loss": 2.8259, "step": 12297 }, { "epoch": 2.34, "learning_rate": 2.2994470924690183e-06, "loss": 2.8734, "step": 12298 }, { "epoch": 2.34, "learning_rate": 2.299389895138227e-06, "loss": 2.3509, "step": 12299 }, { "epoch": 2.34, "learning_rate": 2.2993326978074355e-06, "loss": 2.6849, "step": 12300 }, { "epoch": 2.34, "learning_rate": 2.2992755004766447e-06, "loss": 3.091, "step": 12301 }, { "epoch": 2.34, "learning_rate": 2.299218303145853e-06, "loss": 2.6357, "step": 12302 }, { "epoch": 2.34, "learning_rate": 2.299161105815062e-06, "loss": 2.6073, "step": 12303 }, { "epoch": 2.34, "learning_rate": 2.2991039084842707e-06, "loss": 2.7541, "step": 12304 }, { "epoch": 2.34, "learning_rate": 2.2990467111534795e-06, "loss": 2.3447, "step": 12305 }, { "epoch": 2.34, "learning_rate": 2.2989895138226883e-06, "loss": 2.4761, "step": 12306 }, { "epoch": 2.34, "learning_rate": 2.298932316491897e-06, "loss": 3.081, "step": 12307 }, { "epoch": 2.34, "learning_rate": 2.298875119161106e-06, "loss": 2.5007, "step": 12308 }, { "epoch": 2.34, "learning_rate": 2.2988179218303147e-06, "loss": 2.7916, "step": 12309 }, { "epoch": 2.34, "learning_rate": 2.2987607244995234e-06, "loss": 1.8147, "step": 12310 }, { "epoch": 2.34, "learning_rate": 2.2987035271687322e-06, "loss": 2.8565, "step": 12311 }, { "epoch": 2.35, "learning_rate": 2.298646329837941e-06, "loss": 3.1791, "step": 12312 }, { "epoch": 2.35, "learning_rate": 2.29858913250715e-06, "loss": 2.8049, "step": 12313 }, { "epoch": 2.35, "learning_rate": 2.298531935176358e-06, "loss": 2.0209, "step": 12314 }, { "epoch": 2.35, "learning_rate": 2.2984747378455674e-06, "loss": 2.6117, "step": 12315 }, { "epoch": 2.35, "learning_rate": 2.298417540514776e-06, "loss": 2.688, "step": 12316 }, { "epoch": 2.35, "learning_rate": 2.2983603431839846e-06, "loss": 1.8881, "step": 12317 }, { "epoch": 2.35, "learning_rate": 2.298303145853194e-06, "loss": 2.6373, "step": 12318 }, { "epoch": 2.35, "learning_rate": 2.298245948522402e-06, "loss": 3.1003, "step": 12319 }, { "epoch": 2.35, "learning_rate": 2.298188751191611e-06, "loss": 3.0956, "step": 12320 }, { "epoch": 2.35, "learning_rate": 2.29813155386082e-06, "loss": 2.4937, "step": 12321 }, { "epoch": 2.35, "learning_rate": 2.2980743565300285e-06, "loss": 2.8766, "step": 12322 }, { "epoch": 2.35, "learning_rate": 2.2980171591992373e-06, "loss": 2.7028, "step": 12323 }, { "epoch": 2.35, "learning_rate": 2.297959961868446e-06, "loss": 3.0756, "step": 12324 }, { "epoch": 2.35, "learning_rate": 2.297902764537655e-06, "loss": 2.1457, "step": 12325 }, { "epoch": 2.35, "learning_rate": 2.2978455672068637e-06, "loss": 2.0993, "step": 12326 }, { "epoch": 2.35, "learning_rate": 2.2977883698760725e-06, "loss": 3.0458, "step": 12327 }, { "epoch": 2.35, "learning_rate": 2.2977311725452813e-06, "loss": 2.6439, "step": 12328 }, { "epoch": 2.35, "learning_rate": 2.29767397521449e-06, "loss": 3.1411, "step": 12329 }, { "epoch": 2.35, "learning_rate": 2.297616777883699e-06, "loss": 2.516, "step": 12330 }, { "epoch": 2.35, "learning_rate": 2.2975595805529077e-06, "loss": 2.7985, "step": 12331 }, { "epoch": 2.35, "learning_rate": 2.2975023832221165e-06, "loss": 2.7319, "step": 12332 }, { "epoch": 2.35, "learning_rate": 2.2974451858913253e-06, "loss": 3.4368, "step": 12333 }, { "epoch": 2.35, "learning_rate": 2.2973879885605336e-06, "loss": 2.0036, "step": 12334 }, { "epoch": 2.35, "learning_rate": 2.297330791229743e-06, "loss": 3.7588, "step": 12335 }, { "epoch": 2.35, "learning_rate": 2.2972735938989512e-06, "loss": 2.119, "step": 12336 }, { "epoch": 2.35, "learning_rate": 2.29721639656816e-06, "loss": 2.4704, "step": 12337 }, { "epoch": 2.35, "learning_rate": 2.2971591992373692e-06, "loss": 1.8824, "step": 12338 }, { "epoch": 2.35, "learning_rate": 2.2971020019065776e-06, "loss": 3.148, "step": 12339 }, { "epoch": 2.35, "learning_rate": 2.2970448045757864e-06, "loss": 2.6115, "step": 12340 }, { "epoch": 2.35, "learning_rate": 2.296987607244995e-06, "loss": 3.8299, "step": 12341 }, { "epoch": 2.35, "learning_rate": 2.296930409914204e-06, "loss": 2.7292, "step": 12342 }, { "epoch": 2.35, "learning_rate": 2.296873212583413e-06, "loss": 2.3872, "step": 12343 }, { "epoch": 2.35, "learning_rate": 2.2968160152526216e-06, "loss": 2.4064, "step": 12344 }, { "epoch": 2.35, "learning_rate": 2.2967588179218304e-06, "loss": 2.5219, "step": 12345 }, { "epoch": 2.35, "learning_rate": 2.296701620591039e-06, "loss": 3.2834, "step": 12346 }, { "epoch": 2.35, "learning_rate": 2.296644423260248e-06, "loss": 2.1107, "step": 12347 }, { "epoch": 2.35, "learning_rate": 2.2965872259294568e-06, "loss": 2.2318, "step": 12348 }, { "epoch": 2.35, "learning_rate": 2.2965300285986656e-06, "loss": 2.4593, "step": 12349 }, { "epoch": 2.35, "learning_rate": 2.2964728312678744e-06, "loss": 2.8981, "step": 12350 }, { "epoch": 2.35, "learning_rate": 2.2964156339370827e-06, "loss": 2.5894, "step": 12351 }, { "epoch": 2.35, "learning_rate": 2.296358436606292e-06, "loss": 2.3353, "step": 12352 }, { "epoch": 2.35, "learning_rate": 2.2963012392755003e-06, "loss": 2.5093, "step": 12353 }, { "epoch": 2.35, "learning_rate": 2.296244041944709e-06, "loss": 2.5626, "step": 12354 }, { "epoch": 2.35, "learning_rate": 2.2961868446139183e-06, "loss": 2.3793, "step": 12355 }, { "epoch": 2.35, "learning_rate": 2.2961296472831267e-06, "loss": 2.9268, "step": 12356 }, { "epoch": 2.35, "learning_rate": 2.2960724499523355e-06, "loss": 3.1281, "step": 12357 }, { "epoch": 2.35, "learning_rate": 2.2960152526215443e-06, "loss": 3.7046, "step": 12358 }, { "epoch": 2.35, "learning_rate": 2.295958055290753e-06, "loss": 2.576, "step": 12359 }, { "epoch": 2.35, "learning_rate": 2.295900857959962e-06, "loss": 2.5336, "step": 12360 }, { "epoch": 2.35, "learning_rate": 2.2958436606291707e-06, "loss": 2.7756, "step": 12361 }, { "epoch": 2.35, "learning_rate": 2.2957864632983795e-06, "loss": 2.2535, "step": 12362 }, { "epoch": 2.35, "learning_rate": 2.2957292659675882e-06, "loss": 3.1277, "step": 12363 }, { "epoch": 2.36, "learning_rate": 2.295672068636797e-06, "loss": 2.4004, "step": 12364 }, { "epoch": 2.36, "learning_rate": 2.295614871306006e-06, "loss": 2.8586, "step": 12365 }, { "epoch": 2.36, "learning_rate": 2.2955576739752146e-06, "loss": 2.6615, "step": 12366 }, { "epoch": 2.36, "learning_rate": 2.2955004766444234e-06, "loss": 3.8651, "step": 12367 }, { "epoch": 2.36, "learning_rate": 2.295443279313632e-06, "loss": 2.8154, "step": 12368 }, { "epoch": 2.36, "learning_rate": 2.295386081982841e-06, "loss": 2.5794, "step": 12369 }, { "epoch": 2.36, "learning_rate": 2.29532888465205e-06, "loss": 2.1015, "step": 12370 }, { "epoch": 2.36, "learning_rate": 2.295271687321258e-06, "loss": 2.5958, "step": 12371 }, { "epoch": 2.36, "learning_rate": 2.2952144899904674e-06, "loss": 2.8341, "step": 12372 }, { "epoch": 2.36, "learning_rate": 2.2951572926596758e-06, "loss": 2.9544, "step": 12373 }, { "epoch": 2.36, "learning_rate": 2.295100095328885e-06, "loss": 2.4975, "step": 12374 }, { "epoch": 2.36, "learning_rate": 2.2950428979980933e-06, "loss": 2.5085, "step": 12375 }, { "epoch": 2.36, "learning_rate": 2.294985700667302e-06, "loss": 2.9982, "step": 12376 }, { "epoch": 2.36, "learning_rate": 2.2949285033365114e-06, "loss": 2.6537, "step": 12377 }, { "epoch": 2.36, "learning_rate": 2.2948713060057197e-06, "loss": 2.4899, "step": 12378 }, { "epoch": 2.36, "learning_rate": 2.2948141086749285e-06, "loss": 3.272, "step": 12379 }, { "epoch": 2.36, "learning_rate": 2.2947569113441373e-06, "loss": 2.3145, "step": 12380 }, { "epoch": 2.36, "learning_rate": 2.294699714013346e-06, "loss": 3.2159, "step": 12381 }, { "epoch": 2.36, "learning_rate": 2.294642516682555e-06, "loss": 3.2943, "step": 12382 }, { "epoch": 2.36, "learning_rate": 2.2945853193517637e-06, "loss": 2.6952, "step": 12383 }, { "epoch": 2.36, "learning_rate": 2.2945281220209725e-06, "loss": 2.1813, "step": 12384 }, { "epoch": 2.36, "learning_rate": 2.294470924690181e-06, "loss": 3.2673, "step": 12385 }, { "epoch": 2.36, "learning_rate": 2.29441372735939e-06, "loss": 2.5957, "step": 12386 }, { "epoch": 2.36, "learning_rate": 2.294356530028599e-06, "loss": 2.4274, "step": 12387 }, { "epoch": 2.36, "learning_rate": 2.2942993326978072e-06, "loss": 2.7853, "step": 12388 }, { "epoch": 2.36, "learning_rate": 2.2942421353670165e-06, "loss": 3.4927, "step": 12389 }, { "epoch": 2.36, "learning_rate": 2.294184938036225e-06, "loss": 2.9969, "step": 12390 }, { "epoch": 2.36, "learning_rate": 2.294127740705434e-06, "loss": 3.6487, "step": 12391 }, { "epoch": 2.36, "learning_rate": 2.2940705433746424e-06, "loss": 2.4133, "step": 12392 }, { "epoch": 2.36, "learning_rate": 2.2940133460438512e-06, "loss": 2.7105, "step": 12393 }, { "epoch": 2.36, "learning_rate": 2.2939561487130604e-06, "loss": 2.5457, "step": 12394 }, { "epoch": 2.36, "learning_rate": 2.293898951382269e-06, "loss": 2.3331, "step": 12395 }, { "epoch": 2.36, "learning_rate": 2.2938417540514776e-06, "loss": 3.1613, "step": 12396 }, { "epoch": 2.36, "learning_rate": 2.2937845567206864e-06, "loss": 2.6333, "step": 12397 }, { "epoch": 2.36, "learning_rate": 2.293727359389895e-06, "loss": 3.0285, "step": 12398 }, { "epoch": 2.36, "learning_rate": 2.293670162059104e-06, "loss": 2.9826, "step": 12399 }, { "epoch": 2.36, "learning_rate": 2.2936129647283128e-06, "loss": 2.7813, "step": 12400 }, { "epoch": 2.36, "learning_rate": 2.2935557673975216e-06, "loss": 2.6531, "step": 12401 }, { "epoch": 2.36, "learning_rate": 2.29349857006673e-06, "loss": 1.6483, "step": 12402 }, { "epoch": 2.36, "learning_rate": 2.293441372735939e-06, "loss": 1.9961, "step": 12403 }, { "epoch": 2.36, "learning_rate": 2.293384175405148e-06, "loss": 3.3115, "step": 12404 }, { "epoch": 2.36, "learning_rate": 2.2933269780743563e-06, "loss": 2.0703, "step": 12405 }, { "epoch": 2.36, "learning_rate": 2.2932697807435655e-06, "loss": 2.038, "step": 12406 }, { "epoch": 2.36, "learning_rate": 2.293212583412774e-06, "loss": 3.3317, "step": 12407 }, { "epoch": 2.36, "learning_rate": 2.293155386081983e-06, "loss": 3.4327, "step": 12408 }, { "epoch": 2.36, "learning_rate": 2.293098188751192e-06, "loss": 2.6367, "step": 12409 }, { "epoch": 2.36, "learning_rate": 2.2930409914204003e-06, "loss": 3.5407, "step": 12410 }, { "epoch": 2.36, "learning_rate": 2.2929837940896095e-06, "loss": 2.7789, "step": 12411 }, { "epoch": 2.36, "learning_rate": 2.292926596758818e-06, "loss": 2.4105, "step": 12412 }, { "epoch": 2.36, "learning_rate": 2.2928693994280267e-06, "loss": 2.2413, "step": 12413 }, { "epoch": 2.36, "learning_rate": 2.2928122020972355e-06, "loss": 3.0652, "step": 12414 }, { "epoch": 2.36, "learning_rate": 2.2927550047664443e-06, "loss": 2.4633, "step": 12415 }, { "epoch": 2.36, "learning_rate": 2.292697807435653e-06, "loss": 3.2905, "step": 12416 }, { "epoch": 2.37, "learning_rate": 2.292640610104862e-06, "loss": 2.7666, "step": 12417 }, { "epoch": 2.37, "learning_rate": 2.2925834127740706e-06, "loss": 2.3837, "step": 12418 }, { "epoch": 2.37, "learning_rate": 2.2925262154432794e-06, "loss": 3.8239, "step": 12419 }, { "epoch": 2.37, "learning_rate": 2.2924690181124882e-06, "loss": 2.7593, "step": 12420 }, { "epoch": 2.37, "learning_rate": 2.292411820781697e-06, "loss": 2.3849, "step": 12421 }, { "epoch": 2.37, "learning_rate": 2.2923546234509054e-06, "loss": 3.4927, "step": 12422 }, { "epoch": 2.37, "learning_rate": 2.2922974261201146e-06, "loss": 1.7219, "step": 12423 }, { "epoch": 2.37, "learning_rate": 2.292240228789323e-06, "loss": 2.8803, "step": 12424 }, { "epoch": 2.37, "learning_rate": 2.292183031458532e-06, "loss": 2.6516, "step": 12425 }, { "epoch": 2.37, "learning_rate": 2.292125834127741e-06, "loss": 2.786, "step": 12426 }, { "epoch": 2.37, "learning_rate": 2.2920686367969494e-06, "loss": 2.7984, "step": 12427 }, { "epoch": 2.37, "learning_rate": 2.2920114394661586e-06, "loss": 2.8371, "step": 12428 }, { "epoch": 2.37, "learning_rate": 2.291954242135367e-06, "loss": 3.4513, "step": 12429 }, { "epoch": 2.37, "learning_rate": 2.2918970448045757e-06, "loss": 2.8272, "step": 12430 }, { "epoch": 2.37, "learning_rate": 2.2918398474737845e-06, "loss": 3.1351, "step": 12431 }, { "epoch": 2.37, "learning_rate": 2.2917826501429933e-06, "loss": 2.4159, "step": 12432 }, { "epoch": 2.37, "learning_rate": 2.291725452812202e-06, "loss": 3.0465, "step": 12433 }, { "epoch": 2.37, "learning_rate": 2.291668255481411e-06, "loss": 4.0257, "step": 12434 }, { "epoch": 2.37, "learning_rate": 2.2916110581506197e-06, "loss": 2.7038, "step": 12435 }, { "epoch": 2.37, "learning_rate": 2.2915538608198285e-06, "loss": 3.302, "step": 12436 }, { "epoch": 2.37, "learning_rate": 2.2914966634890373e-06, "loss": 3.0627, "step": 12437 }, { "epoch": 2.37, "learning_rate": 2.291439466158246e-06, "loss": 2.5469, "step": 12438 }, { "epoch": 2.37, "learning_rate": 2.291382268827455e-06, "loss": 2.5437, "step": 12439 }, { "epoch": 2.37, "learning_rate": 2.2913250714966637e-06, "loss": 2.4895, "step": 12440 }, { "epoch": 2.37, "learning_rate": 2.291267874165872e-06, "loss": 2.7794, "step": 12441 }, { "epoch": 2.37, "learning_rate": 2.2912106768350813e-06, "loss": 2.9742, "step": 12442 }, { "epoch": 2.37, "learning_rate": 2.29115347950429e-06, "loss": 2.456, "step": 12443 }, { "epoch": 2.37, "learning_rate": 2.2910962821734984e-06, "loss": 1.4991, "step": 12444 }, { "epoch": 2.37, "learning_rate": 2.2910390848427076e-06, "loss": 2.1138, "step": 12445 }, { "epoch": 2.37, "learning_rate": 2.290981887511916e-06, "loss": 2.2999, "step": 12446 }, { "epoch": 2.37, "learning_rate": 2.290924690181125e-06, "loss": 3.1421, "step": 12447 }, { "epoch": 2.37, "learning_rate": 2.2908674928503336e-06, "loss": 2.327, "step": 12448 }, { "epoch": 2.37, "learning_rate": 2.2908102955195424e-06, "loss": 3.4088, "step": 12449 }, { "epoch": 2.37, "learning_rate": 2.290753098188751e-06, "loss": 2.3757, "step": 12450 }, { "epoch": 2.37, "learning_rate": 2.29069590085796e-06, "loss": 2.4269, "step": 12451 }, { "epoch": 2.37, "learning_rate": 2.2906387035271688e-06, "loss": 2.4379, "step": 12452 }, { "epoch": 2.37, "learning_rate": 2.2905815061963776e-06, "loss": 3.2762, "step": 12453 }, { "epoch": 2.37, "learning_rate": 2.2905243088655864e-06, "loss": 3.8171, "step": 12454 }, { "epoch": 2.37, "learning_rate": 2.290467111534795e-06, "loss": 2.9777, "step": 12455 }, { "epoch": 2.37, "learning_rate": 2.290409914204004e-06, "loss": 1.9517, "step": 12456 }, { "epoch": 2.37, "learning_rate": 2.2903527168732127e-06, "loss": 2.1281, "step": 12457 }, { "epoch": 2.37, "learning_rate": 2.2902955195424215e-06, "loss": 3.1784, "step": 12458 }, { "epoch": 2.37, "learning_rate": 2.2902383222116303e-06, "loss": 2.4291, "step": 12459 }, { "epoch": 2.37, "learning_rate": 2.290181124880839e-06, "loss": 3.5668, "step": 12460 }, { "epoch": 2.37, "learning_rate": 2.2901239275500475e-06, "loss": 2.6494, "step": 12461 }, { "epoch": 2.37, "learning_rate": 2.2900667302192567e-06, "loss": 2.3889, "step": 12462 }, { "epoch": 2.37, "learning_rate": 2.290009532888465e-06, "loss": 2.5659, "step": 12463 }, { "epoch": 2.37, "learning_rate": 2.289952335557674e-06, "loss": 3.1223, "step": 12464 }, { "epoch": 2.37, "learning_rate": 2.289895138226883e-06, "loss": 2.7756, "step": 12465 }, { "epoch": 2.37, "learning_rate": 2.2898379408960915e-06, "loss": 2.5187, "step": 12466 }, { "epoch": 2.37, "learning_rate": 2.2897807435653003e-06, "loss": 3.0259, "step": 12467 }, { "epoch": 2.37, "learning_rate": 2.289723546234509e-06, "loss": 2.1853, "step": 12468 }, { "epoch": 2.38, "learning_rate": 2.289666348903718e-06, "loss": 2.9896, "step": 12469 }, { "epoch": 2.38, "learning_rate": 2.2896091515729266e-06, "loss": 3.3348, "step": 12470 }, { "epoch": 2.38, "learning_rate": 2.2895519542421354e-06, "loss": 3.4426, "step": 12471 }, { "epoch": 2.38, "learning_rate": 2.2894947569113442e-06, "loss": 2.1586, "step": 12472 }, { "epoch": 2.38, "learning_rate": 2.289437559580553e-06, "loss": 2.9341, "step": 12473 }, { "epoch": 2.38, "learning_rate": 2.289380362249762e-06, "loss": 2.4912, "step": 12474 }, { "epoch": 2.38, "learning_rate": 2.2893231649189706e-06, "loss": 2.7152, "step": 12475 }, { "epoch": 2.38, "learning_rate": 2.2892659675881794e-06, "loss": 3.0787, "step": 12476 }, { "epoch": 2.38, "learning_rate": 2.289208770257388e-06, "loss": 2.7579, "step": 12477 }, { "epoch": 2.38, "learning_rate": 2.2891515729265966e-06, "loss": 2.9945, "step": 12478 }, { "epoch": 2.38, "learning_rate": 2.2890943755958058e-06, "loss": 3.1855, "step": 12479 }, { "epoch": 2.38, "learning_rate": 2.289037178265014e-06, "loss": 1.8743, "step": 12480 }, { "epoch": 2.38, "learning_rate": 2.288979980934223e-06, "loss": 2.2451, "step": 12481 }, { "epoch": 2.38, "learning_rate": 2.288922783603432e-06, "loss": 2.5057, "step": 12482 }, { "epoch": 2.38, "learning_rate": 2.2888655862726405e-06, "loss": 2.2945, "step": 12483 }, { "epoch": 2.38, "learning_rate": 2.2888083889418493e-06, "loss": 3.0536, "step": 12484 }, { "epoch": 2.38, "learning_rate": 2.288751191611058e-06, "loss": 2.9286, "step": 12485 }, { "epoch": 2.38, "learning_rate": 2.288693994280267e-06, "loss": 2.0744, "step": 12486 }, { "epoch": 2.38, "learning_rate": 2.2886367969494757e-06, "loss": 3.2686, "step": 12487 }, { "epoch": 2.38, "learning_rate": 2.2885795996186845e-06, "loss": 3.3489, "step": 12488 }, { "epoch": 2.38, "learning_rate": 2.2885224022878933e-06, "loss": 3.3293, "step": 12489 }, { "epoch": 2.38, "learning_rate": 2.288465204957102e-06, "loss": 2.0587, "step": 12490 }, { "epoch": 2.38, "learning_rate": 2.288408007626311e-06, "loss": 2.3958, "step": 12491 }, { "epoch": 2.38, "learning_rate": 2.2883508102955197e-06, "loss": 2.8676, "step": 12492 }, { "epoch": 2.38, "learning_rate": 2.2882936129647285e-06, "loss": 2.4669, "step": 12493 }, { "epoch": 2.38, "learning_rate": 2.2882364156339373e-06, "loss": 3.0203, "step": 12494 }, { "epoch": 2.38, "learning_rate": 2.2881792183031456e-06, "loss": 1.8457, "step": 12495 }, { "epoch": 2.38, "learning_rate": 2.288122020972355e-06, "loss": 3.4097, "step": 12496 }, { "epoch": 2.38, "learning_rate": 2.2880648236415637e-06, "loss": 2.4612, "step": 12497 }, { "epoch": 2.38, "learning_rate": 2.288007626310772e-06, "loss": 2.4449, "step": 12498 }, { "epoch": 2.38, "learning_rate": 2.2879504289799812e-06, "loss": 2.93, "step": 12499 }, { "epoch": 2.38, "learning_rate": 2.2878932316491896e-06, "loss": 3.0877, "step": 12500 }, { "epoch": 2.38, "learning_rate": 2.2878360343183984e-06, "loss": 2.4766, "step": 12501 }, { "epoch": 2.38, "learning_rate": 2.287778836987607e-06, "loss": 2.8047, "step": 12502 }, { "epoch": 2.38, "learning_rate": 2.287721639656816e-06, "loss": 2.1976, "step": 12503 }, { "epoch": 2.38, "learning_rate": 2.287664442326025e-06, "loss": 3.0011, "step": 12504 }, { "epoch": 2.38, "learning_rate": 2.2876072449952336e-06, "loss": 3.0455, "step": 12505 }, { "epoch": 2.38, "learning_rate": 2.2875500476644424e-06, "loss": 2.472, "step": 12506 }, { "epoch": 2.38, "learning_rate": 2.287492850333651e-06, "loss": 2.8403, "step": 12507 }, { "epoch": 2.38, "learning_rate": 2.28743565300286e-06, "loss": 3.3346, "step": 12508 }, { "epoch": 2.38, "learning_rate": 2.2873784556720688e-06, "loss": 2.4523, "step": 12509 }, { "epoch": 2.38, "learning_rate": 2.2873212583412775e-06, "loss": 3.3745, "step": 12510 }, { "epoch": 2.38, "learning_rate": 2.2872640610104863e-06, "loss": 1.739, "step": 12511 }, { "epoch": 2.38, "learning_rate": 2.2872068636796947e-06, "loss": 2.4647, "step": 12512 }, { "epoch": 2.38, "learning_rate": 2.287149666348904e-06, "loss": 2.7464, "step": 12513 }, { "epoch": 2.38, "learning_rate": 2.2870924690181127e-06, "loss": 3.1787, "step": 12514 }, { "epoch": 2.38, "learning_rate": 2.287035271687321e-06, "loss": 2.1996, "step": 12515 }, { "epoch": 2.38, "learning_rate": 2.2869780743565303e-06, "loss": 2.5279, "step": 12516 }, { "epoch": 2.38, "learning_rate": 2.2869208770257387e-06, "loss": 1.7327, "step": 12517 }, { "epoch": 2.38, "learning_rate": 2.2868636796949475e-06, "loss": 2.1004, "step": 12518 }, { "epoch": 2.38, "learning_rate": 2.2868064823641563e-06, "loss": 3.4868, "step": 12519 }, { "epoch": 2.38, "learning_rate": 2.286749285033365e-06, "loss": 2.6462, "step": 12520 }, { "epoch": 2.38, "learning_rate": 2.2866920877025743e-06, "loss": 2.7247, "step": 12521 }, { "epoch": 2.39, "learning_rate": 2.2866348903717827e-06, "loss": 1.815, "step": 12522 }, { "epoch": 2.39, "learning_rate": 2.2865776930409914e-06, "loss": 3.3521, "step": 12523 }, { "epoch": 2.39, "learning_rate": 2.2865204957102002e-06, "loss": 1.9127, "step": 12524 }, { "epoch": 2.39, "learning_rate": 2.286463298379409e-06, "loss": 2.9841, "step": 12525 }, { "epoch": 2.39, "learning_rate": 2.286406101048618e-06, "loss": 2.4685, "step": 12526 }, { "epoch": 2.39, "learning_rate": 2.2863489037178266e-06, "loss": 2.7879, "step": 12527 }, { "epoch": 2.39, "learning_rate": 2.2862917063870354e-06, "loss": 3.3628, "step": 12528 }, { "epoch": 2.39, "learning_rate": 2.2862345090562438e-06, "loss": 2.5618, "step": 12529 }, { "epoch": 2.39, "learning_rate": 2.286177311725453e-06, "loss": 2.5625, "step": 12530 }, { "epoch": 2.39, "learning_rate": 2.286120114394662e-06, "loss": 2.9251, "step": 12531 }, { "epoch": 2.39, "learning_rate": 2.28606291706387e-06, "loss": 2.9882, "step": 12532 }, { "epoch": 2.39, "learning_rate": 2.2860057197330794e-06, "loss": 3.0301, "step": 12533 }, { "epoch": 2.39, "learning_rate": 2.2859485224022878e-06, "loss": 1.9837, "step": 12534 }, { "epoch": 2.39, "learning_rate": 2.2858913250714965e-06, "loss": 2.4388, "step": 12535 }, { "epoch": 2.39, "learning_rate": 2.2858341277407053e-06, "loss": 3.6224, "step": 12536 }, { "epoch": 2.39, "learning_rate": 2.285776930409914e-06, "loss": 3.0067, "step": 12537 }, { "epoch": 2.39, "learning_rate": 2.2857197330791234e-06, "loss": 3.3583, "step": 12538 }, { "epoch": 2.39, "learning_rate": 2.2856625357483317e-06, "loss": 2.7096, "step": 12539 }, { "epoch": 2.39, "learning_rate": 2.2856053384175405e-06, "loss": 3.4686, "step": 12540 }, { "epoch": 2.39, "learning_rate": 2.2855481410867493e-06, "loss": 2.7207, "step": 12541 }, { "epoch": 2.39, "learning_rate": 2.285490943755958e-06, "loss": 2.6871, "step": 12542 }, { "epoch": 2.39, "learning_rate": 2.285433746425167e-06, "loss": 2.9855, "step": 12543 }, { "epoch": 2.39, "learning_rate": 2.2853765490943757e-06, "loss": 2.4256, "step": 12544 }, { "epoch": 2.39, "learning_rate": 2.2853193517635845e-06, "loss": 2.9673, "step": 12545 }, { "epoch": 2.39, "learning_rate": 2.2852621544327933e-06, "loss": 3.0907, "step": 12546 }, { "epoch": 2.39, "learning_rate": 2.285204957102002e-06, "loss": 3.6519, "step": 12547 }, { "epoch": 2.39, "learning_rate": 2.285147759771211e-06, "loss": 2.4233, "step": 12548 }, { "epoch": 2.39, "learning_rate": 2.2850905624404192e-06, "loss": 2.4592, "step": 12549 }, { "epoch": 2.39, "learning_rate": 2.2850333651096285e-06, "loss": 3.7499, "step": 12550 }, { "epoch": 2.39, "learning_rate": 2.284976167778837e-06, "loss": 2.698, "step": 12551 }, { "epoch": 2.39, "learning_rate": 2.2849189704480456e-06, "loss": 2.7887, "step": 12552 }, { "epoch": 2.39, "learning_rate": 2.284861773117255e-06, "loss": 2.2402, "step": 12553 }, { "epoch": 2.39, "learning_rate": 2.284804575786463e-06, "loss": 2.0251, "step": 12554 }, { "epoch": 2.39, "learning_rate": 2.2847473784556724e-06, "loss": 3.3598, "step": 12555 }, { "epoch": 2.39, "learning_rate": 2.284690181124881e-06, "loss": 2.6368, "step": 12556 }, { "epoch": 2.39, "learning_rate": 2.2846329837940896e-06, "loss": 3.3362, "step": 12557 }, { "epoch": 2.39, "learning_rate": 2.2845757864632984e-06, "loss": 2.9827, "step": 12558 }, { "epoch": 2.39, "learning_rate": 2.284518589132507e-06, "loss": 2.2559, "step": 12559 }, { "epoch": 2.39, "learning_rate": 2.284461391801716e-06, "loss": 2.3212, "step": 12560 }, { "epoch": 2.39, "learning_rate": 2.2844041944709248e-06, "loss": 2.3364, "step": 12561 }, { "epoch": 2.39, "learning_rate": 2.2843469971401336e-06, "loss": 2.6213, "step": 12562 }, { "epoch": 2.39, "learning_rate": 2.2842897998093424e-06, "loss": 3.0024, "step": 12563 }, { "epoch": 2.39, "learning_rate": 2.284232602478551e-06, "loss": 2.8374, "step": 12564 }, { "epoch": 2.39, "learning_rate": 2.28417540514776e-06, "loss": 2.746, "step": 12565 }, { "epoch": 2.39, "learning_rate": 2.2841182078169683e-06, "loss": 2.8908, "step": 12566 }, { "epoch": 2.39, "learning_rate": 2.2840610104861775e-06, "loss": 3.2535, "step": 12567 }, { "epoch": 2.39, "learning_rate": 2.284003813155386e-06, "loss": 2.7842, "step": 12568 }, { "epoch": 2.39, "learning_rate": 2.283946615824595e-06, "loss": 3.1627, "step": 12569 }, { "epoch": 2.39, "learning_rate": 2.283889418493804e-06, "loss": 3.7593, "step": 12570 }, { "epoch": 2.39, "learning_rate": 2.2838322211630123e-06, "loss": 2.8144, "step": 12571 }, { "epoch": 2.39, "learning_rate": 2.2837750238322215e-06, "loss": 2.0579, "step": 12572 }, { "epoch": 2.39, "learning_rate": 2.28371782650143e-06, "loss": 3.109, "step": 12573 }, { "epoch": 2.4, "learning_rate": 2.2836606291706387e-06, "loss": 2.2422, "step": 12574 }, { "epoch": 2.4, "learning_rate": 2.2836034318398475e-06, "loss": 2.2676, "step": 12575 }, { "epoch": 2.4, "learning_rate": 2.2835462345090562e-06, "loss": 2.8551, "step": 12576 }, { "epoch": 2.4, "learning_rate": 2.283489037178265e-06, "loss": 1.9182, "step": 12577 }, { "epoch": 2.4, "learning_rate": 2.283431839847474e-06, "loss": 2.6095, "step": 12578 }, { "epoch": 2.4, "learning_rate": 2.2833746425166826e-06, "loss": 1.8124, "step": 12579 }, { "epoch": 2.4, "learning_rate": 2.2833174451858914e-06, "loss": 3.3988, "step": 12580 }, { "epoch": 2.4, "learning_rate": 2.2832602478551002e-06, "loss": 2.4636, "step": 12581 }, { "epoch": 2.4, "learning_rate": 2.283203050524309e-06, "loss": 2.3868, "step": 12582 }, { "epoch": 2.4, "learning_rate": 2.2831458531935174e-06, "loss": 2.7703, "step": 12583 }, { "epoch": 2.4, "learning_rate": 2.2830886558627266e-06, "loss": 2.0205, "step": 12584 }, { "epoch": 2.4, "learning_rate": 2.283031458531935e-06, "loss": 1.9532, "step": 12585 }, { "epoch": 2.4, "learning_rate": 2.282974261201144e-06, "loss": 2.3969, "step": 12586 }, { "epoch": 2.4, "learning_rate": 2.282917063870353e-06, "loss": 4.0322, "step": 12587 }, { "epoch": 2.4, "learning_rate": 2.2828598665395613e-06, "loss": 3.0124, "step": 12588 }, { "epoch": 2.4, "learning_rate": 2.2828026692087706e-06, "loss": 2.6928, "step": 12589 }, { "epoch": 2.4, "learning_rate": 2.282745471877979e-06, "loss": 2.9194, "step": 12590 }, { "epoch": 2.4, "learning_rate": 2.2826882745471877e-06, "loss": 1.4957, "step": 12591 }, { "epoch": 2.4, "learning_rate": 2.282631077216397e-06, "loss": 2.7966, "step": 12592 }, { "epoch": 2.4, "learning_rate": 2.2825738798856053e-06, "loss": 2.7025, "step": 12593 }, { "epoch": 2.4, "learning_rate": 2.282516682554814e-06, "loss": 2.9082, "step": 12594 }, { "epoch": 2.4, "learning_rate": 2.282459485224023e-06, "loss": 4.0864, "step": 12595 }, { "epoch": 2.4, "learning_rate": 2.2824022878932317e-06, "loss": 1.9402, "step": 12596 }, { "epoch": 2.4, "learning_rate": 2.2823450905624405e-06, "loss": 2.9441, "step": 12597 }, { "epoch": 2.4, "learning_rate": 2.2822878932316493e-06, "loss": 2.4863, "step": 12598 }, { "epoch": 2.4, "learning_rate": 2.282230695900858e-06, "loss": 3.5406, "step": 12599 }, { "epoch": 2.4, "learning_rate": 2.2821734985700665e-06, "loss": 2.3773, "step": 12600 }, { "epoch": 2.4, "learning_rate": 2.2821163012392757e-06, "loss": 2.3244, "step": 12601 }, { "epoch": 2.4, "learning_rate": 2.2820591039084845e-06, "loss": 2.2679, "step": 12602 }, { "epoch": 2.4, "learning_rate": 2.2820019065776933e-06, "loss": 2.6063, "step": 12603 }, { "epoch": 2.4, "learning_rate": 2.281944709246902e-06, "loss": 1.8829, "step": 12604 }, { "epoch": 2.4, "learning_rate": 2.2818875119161104e-06, "loss": 2.507, "step": 12605 }, { "epoch": 2.4, "learning_rate": 2.2818303145853196e-06, "loss": 2.9773, "step": 12606 }, { "epoch": 2.4, "learning_rate": 2.281773117254528e-06, "loss": 3.3091, "step": 12607 }, { "epoch": 2.4, "learning_rate": 2.281715919923737e-06, "loss": 2.3355, "step": 12608 }, { "epoch": 2.4, "learning_rate": 2.281658722592946e-06, "loss": 3.2559, "step": 12609 }, { "epoch": 2.4, "learning_rate": 2.2816015252621544e-06, "loss": 3.1089, "step": 12610 }, { "epoch": 2.4, "learning_rate": 2.281544327931363e-06, "loss": 3.0807, "step": 12611 }, { "epoch": 2.4, "learning_rate": 2.281487130600572e-06, "loss": 3.0214, "step": 12612 }, { "epoch": 2.4, "learning_rate": 2.2814299332697808e-06, "loss": 2.1427, "step": 12613 }, { "epoch": 2.4, "learning_rate": 2.2813727359389896e-06, "loss": 2.9117, "step": 12614 }, { "epoch": 2.4, "learning_rate": 2.2813155386081984e-06, "loss": 2.845, "step": 12615 }, { "epoch": 2.4, "learning_rate": 2.281258341277407e-06, "loss": 3.2028, "step": 12616 }, { "epoch": 2.4, "learning_rate": 2.2812011439466155e-06, "loss": 2.1358, "step": 12617 }, { "epoch": 2.4, "learning_rate": 2.2811439466158247e-06, "loss": 1.9301, "step": 12618 }, { "epoch": 2.4, "learning_rate": 2.2810867492850335e-06, "loss": 2.164, "step": 12619 }, { "epoch": 2.4, "learning_rate": 2.2810295519542423e-06, "loss": 2.1972, "step": 12620 }, { "epoch": 2.4, "learning_rate": 2.280972354623451e-06, "loss": 2.3314, "step": 12621 }, { "epoch": 2.4, "learning_rate": 2.2809151572926595e-06, "loss": 1.4604, "step": 12622 }, { "epoch": 2.4, "learning_rate": 2.2808579599618687e-06, "loss": 2.2624, "step": 12623 }, { "epoch": 2.4, "learning_rate": 2.280800762631077e-06, "loss": 2.4976, "step": 12624 }, { "epoch": 2.4, "learning_rate": 2.280743565300286e-06, "loss": 3.1091, "step": 12625 }, { "epoch": 2.4, "learning_rate": 2.280686367969495e-06, "loss": 3.008, "step": 12626 }, { "epoch": 2.41, "learning_rate": 2.2806291706387035e-06, "loss": 3.2934, "step": 12627 }, { "epoch": 2.41, "learning_rate": 2.2805719733079123e-06, "loss": 3.5601, "step": 12628 }, { "epoch": 2.41, "learning_rate": 2.280514775977121e-06, "loss": 3.6735, "step": 12629 }, { "epoch": 2.41, "learning_rate": 2.28045757864633e-06, "loss": 2.3635, "step": 12630 }, { "epoch": 2.41, "learning_rate": 2.2804003813155386e-06, "loss": 2.5532, "step": 12631 }, { "epoch": 2.41, "learning_rate": 2.2803431839847474e-06, "loss": 2.2338, "step": 12632 }, { "epoch": 2.41, "learning_rate": 2.2802859866539562e-06, "loss": 2.5516, "step": 12633 }, { "epoch": 2.41, "learning_rate": 2.280228789323165e-06, "loss": 2.5905, "step": 12634 }, { "epoch": 2.41, "learning_rate": 2.280171591992374e-06, "loss": 3.158, "step": 12635 }, { "epoch": 2.41, "learning_rate": 2.2801143946615826e-06, "loss": 2.758, "step": 12636 }, { "epoch": 2.41, "learning_rate": 2.2800571973307914e-06, "loss": 2.75, "step": 12637 }, { "epoch": 2.41, "learning_rate": 2.28e-06, "loss": 2.8522, "step": 12638 }, { "epoch": 2.41, "learning_rate": 2.2799428026692086e-06, "loss": 2.4129, "step": 12639 }, { "epoch": 2.41, "learning_rate": 2.2798856053384178e-06, "loss": 2.879, "step": 12640 }, { "epoch": 2.41, "learning_rate": 2.2798284080076266e-06, "loss": 3.0381, "step": 12641 }, { "epoch": 2.41, "learning_rate": 2.279771210676835e-06, "loss": 1.9549, "step": 12642 }, { "epoch": 2.41, "learning_rate": 2.279714013346044e-06, "loss": 1.8487, "step": 12643 }, { "epoch": 2.41, "learning_rate": 2.2796568160152525e-06, "loss": 3.2972, "step": 12644 }, { "epoch": 2.41, "learning_rate": 2.2795996186844613e-06, "loss": 2.6422, "step": 12645 }, { "epoch": 2.41, "learning_rate": 2.27954242135367e-06, "loss": 1.9263, "step": 12646 }, { "epoch": 2.41, "learning_rate": 2.279485224022879e-06, "loss": 3.0505, "step": 12647 }, { "epoch": 2.41, "learning_rate": 2.2794280266920877e-06, "loss": 3.1518, "step": 12648 }, { "epoch": 2.41, "learning_rate": 2.2793708293612965e-06, "loss": 2.3186, "step": 12649 }, { "epoch": 2.41, "learning_rate": 2.2793136320305053e-06, "loss": 2.4781, "step": 12650 }, { "epoch": 2.41, "learning_rate": 2.279256434699714e-06, "loss": 2.1549, "step": 12651 }, { "epoch": 2.41, "learning_rate": 2.279199237368923e-06, "loss": 3.02, "step": 12652 }, { "epoch": 2.41, "learning_rate": 2.2791420400381317e-06, "loss": 2.7996, "step": 12653 }, { "epoch": 2.41, "learning_rate": 2.2790848427073405e-06, "loss": 1.8711, "step": 12654 }, { "epoch": 2.41, "learning_rate": 2.2790276453765493e-06, "loss": 3.5408, "step": 12655 }, { "epoch": 2.41, "learning_rate": 2.2789704480457576e-06, "loss": 2.8935, "step": 12656 }, { "epoch": 2.41, "learning_rate": 2.278913250714967e-06, "loss": 2.3009, "step": 12657 }, { "epoch": 2.41, "learning_rate": 2.2788560533841756e-06, "loss": 3.1021, "step": 12658 }, { "epoch": 2.41, "learning_rate": 2.278798856053384e-06, "loss": 2.1784, "step": 12659 }, { "epoch": 2.41, "learning_rate": 2.2787416587225932e-06, "loss": 2.2621, "step": 12660 }, { "epoch": 2.41, "learning_rate": 2.2786844613918016e-06, "loss": 3.2445, "step": 12661 }, { "epoch": 2.41, "learning_rate": 2.2786272640610104e-06, "loss": 1.7608, "step": 12662 }, { "epoch": 2.41, "learning_rate": 2.278570066730219e-06, "loss": 3.2634, "step": 12663 }, { "epoch": 2.41, "learning_rate": 2.278512869399428e-06, "loss": 2.9992, "step": 12664 }, { "epoch": 2.41, "learning_rate": 2.2784556720686368e-06, "loss": 2.7792, "step": 12665 }, { "epoch": 2.41, "learning_rate": 2.2783984747378456e-06, "loss": 2.9152, "step": 12666 }, { "epoch": 2.41, "learning_rate": 2.2783412774070544e-06, "loss": 3.501, "step": 12667 }, { "epoch": 2.41, "learning_rate": 2.278284080076263e-06, "loss": 2.9292, "step": 12668 }, { "epoch": 2.41, "learning_rate": 2.278226882745472e-06, "loss": 2.5962, "step": 12669 }, { "epoch": 2.41, "learning_rate": 2.2781696854146807e-06, "loss": 3.0576, "step": 12670 }, { "epoch": 2.41, "learning_rate": 2.2781124880838895e-06, "loss": 3.0692, "step": 12671 }, { "epoch": 2.41, "learning_rate": 2.2780552907530983e-06, "loss": 2.2988, "step": 12672 }, { "epoch": 2.41, "learning_rate": 2.2779980934223067e-06, "loss": 2.5491, "step": 12673 }, { "epoch": 2.41, "learning_rate": 2.277940896091516e-06, "loss": 2.46, "step": 12674 }, { "epoch": 2.41, "learning_rate": 2.2778836987607247e-06, "loss": 2.2618, "step": 12675 }, { "epoch": 2.41, "learning_rate": 2.277826501429933e-06, "loss": 3.2106, "step": 12676 }, { "epoch": 2.41, "learning_rate": 2.2777693040991423e-06, "loss": 2.3303, "step": 12677 }, { "epoch": 2.41, "learning_rate": 2.2777121067683507e-06, "loss": 2.5463, "step": 12678 }, { "epoch": 2.42, "learning_rate": 2.2776549094375595e-06, "loss": 3.1722, "step": 12679 }, { "epoch": 2.42, "learning_rate": 2.2775977121067687e-06, "loss": 3.1157, "step": 12680 }, { "epoch": 2.42, "learning_rate": 2.277540514775977e-06, "loss": 3.1332, "step": 12681 }, { "epoch": 2.42, "learning_rate": 2.277483317445186e-06, "loss": 2.9452, "step": 12682 }, { "epoch": 2.42, "learning_rate": 2.2774261201143946e-06, "loss": 2.4763, "step": 12683 }, { "epoch": 2.42, "learning_rate": 2.2773689227836034e-06, "loss": 2.0853, "step": 12684 }, { "epoch": 2.42, "learning_rate": 2.2773117254528122e-06, "loss": 2.7743, "step": 12685 }, { "epoch": 2.42, "learning_rate": 2.277254528122021e-06, "loss": 2.3618, "step": 12686 }, { "epoch": 2.42, "learning_rate": 2.27719733079123e-06, "loss": 3.1337, "step": 12687 }, { "epoch": 2.42, "learning_rate": 2.2771401334604386e-06, "loss": 2.6198, "step": 12688 }, { "epoch": 2.42, "learning_rate": 2.2770829361296474e-06, "loss": 2.3849, "step": 12689 }, { "epoch": 2.42, "learning_rate": 2.277025738798856e-06, "loss": 2.726, "step": 12690 }, { "epoch": 2.42, "learning_rate": 2.276968541468065e-06, "loss": 3.426, "step": 12691 }, { "epoch": 2.42, "learning_rate": 2.276911344137274e-06, "loss": 2.3854, "step": 12692 }, { "epoch": 2.42, "learning_rate": 2.276854146806482e-06, "loss": 2.8498, "step": 12693 }, { "epoch": 2.42, "learning_rate": 2.2767969494756914e-06, "loss": 3.8972, "step": 12694 }, { "epoch": 2.42, "learning_rate": 2.2767397521448997e-06, "loss": 2.9813, "step": 12695 }, { "epoch": 2.42, "learning_rate": 2.2766825548141085e-06, "loss": 2.9144, "step": 12696 }, { "epoch": 2.42, "learning_rate": 2.2766253574833178e-06, "loss": 3.9071, "step": 12697 }, { "epoch": 2.42, "learning_rate": 2.276568160152526e-06, "loss": 2.4764, "step": 12698 }, { "epoch": 2.42, "learning_rate": 2.2765109628217353e-06, "loss": 2.1883, "step": 12699 }, { "epoch": 2.42, "learning_rate": 2.2764537654909437e-06, "loss": 3.3337, "step": 12700 }, { "epoch": 2.42, "learning_rate": 2.2763965681601525e-06, "loss": 3.4088, "step": 12701 }, { "epoch": 2.42, "learning_rate": 2.2763393708293613e-06, "loss": 2.7619, "step": 12702 }, { "epoch": 2.42, "learning_rate": 2.27628217349857e-06, "loss": 2.7235, "step": 12703 }, { "epoch": 2.42, "learning_rate": 2.276224976167779e-06, "loss": 2.1291, "step": 12704 }, { "epoch": 2.42, "learning_rate": 2.2761677788369877e-06, "loss": 2.1984, "step": 12705 }, { "epoch": 2.42, "learning_rate": 2.2761105815061965e-06, "loss": 2.0453, "step": 12706 }, { "epoch": 2.42, "learning_rate": 2.2760533841754053e-06, "loss": 2.554, "step": 12707 }, { "epoch": 2.42, "learning_rate": 2.275996186844614e-06, "loss": 2.8761, "step": 12708 }, { "epoch": 2.42, "learning_rate": 2.275938989513823e-06, "loss": 2.4645, "step": 12709 }, { "epoch": 2.42, "learning_rate": 2.2758817921830312e-06, "loss": 2.4995, "step": 12710 }, { "epoch": 2.42, "learning_rate": 2.2758245948522404e-06, "loss": 3.0837, "step": 12711 }, { "epoch": 2.42, "learning_rate": 2.275767397521449e-06, "loss": 1.3508, "step": 12712 }, { "epoch": 2.42, "learning_rate": 2.2757102001906576e-06, "loss": 2.9248, "step": 12713 }, { "epoch": 2.42, "learning_rate": 2.275653002859867e-06, "loss": 2.6947, "step": 12714 }, { "epoch": 2.42, "learning_rate": 2.275595805529075e-06, "loss": 3.7439, "step": 12715 }, { "epoch": 2.42, "learning_rate": 2.2755386081982844e-06, "loss": 1.829, "step": 12716 }, { "epoch": 2.42, "learning_rate": 2.2754814108674928e-06, "loss": 3.4433, "step": 12717 }, { "epoch": 2.42, "learning_rate": 2.2754242135367016e-06, "loss": 2.1159, "step": 12718 }, { "epoch": 2.42, "learning_rate": 2.275367016205911e-06, "loss": 3.0717, "step": 12719 }, { "epoch": 2.42, "learning_rate": 2.275309818875119e-06, "loss": 2.688, "step": 12720 }, { "epoch": 2.42, "learning_rate": 2.275252621544328e-06, "loss": 3.7624, "step": 12721 }, { "epoch": 2.42, "learning_rate": 2.2751954242135368e-06, "loss": 2.5559, "step": 12722 }, { "epoch": 2.42, "learning_rate": 2.2751382268827456e-06, "loss": 2.9503, "step": 12723 }, { "epoch": 2.42, "learning_rate": 2.2750810295519543e-06, "loss": 3.0921, "step": 12724 }, { "epoch": 2.42, "learning_rate": 2.275023832221163e-06, "loss": 2.6788, "step": 12725 }, { "epoch": 2.42, "learning_rate": 2.274966634890372e-06, "loss": 2.7896, "step": 12726 }, { "epoch": 2.42, "learning_rate": 2.2749094375595803e-06, "loss": 2.145, "step": 12727 }, { "epoch": 2.42, "learning_rate": 2.2748522402287895e-06, "loss": 2.443, "step": 12728 }, { "epoch": 2.42, "learning_rate": 2.2747950428979983e-06, "loss": 3.0065, "step": 12729 }, { "epoch": 2.42, "learning_rate": 2.2747378455672067e-06, "loss": 2.5566, "step": 12730 }, { "epoch": 2.42, "learning_rate": 2.274680648236416e-06, "loss": 2.0323, "step": 12731 }, { "epoch": 2.43, "learning_rate": 2.2746234509056243e-06, "loss": 2.1841, "step": 12732 }, { "epoch": 2.43, "learning_rate": 2.2745662535748335e-06, "loss": 2.6853, "step": 12733 }, { "epoch": 2.43, "learning_rate": 2.274509056244042e-06, "loss": 2.7216, "step": 12734 }, { "epoch": 2.43, "learning_rate": 2.2744518589132507e-06, "loss": 3.0601, "step": 12735 }, { "epoch": 2.43, "learning_rate": 2.27439466158246e-06, "loss": 3.0935, "step": 12736 }, { "epoch": 2.43, "learning_rate": 2.2743374642516682e-06, "loss": 2.7464, "step": 12737 }, { "epoch": 2.43, "learning_rate": 2.274280266920877e-06, "loss": 1.8623, "step": 12738 }, { "epoch": 2.43, "learning_rate": 2.274223069590086e-06, "loss": 2.6555, "step": 12739 }, { "epoch": 2.43, "learning_rate": 2.2741658722592946e-06, "loss": 3.2408, "step": 12740 }, { "epoch": 2.43, "learning_rate": 2.2741086749285034e-06, "loss": 3.9134, "step": 12741 }, { "epoch": 2.43, "learning_rate": 2.274051477597712e-06, "loss": 2.3695, "step": 12742 }, { "epoch": 2.43, "learning_rate": 2.273994280266921e-06, "loss": 2.2553, "step": 12743 }, { "epoch": 2.43, "learning_rate": 2.2739370829361294e-06, "loss": 2.1662, "step": 12744 }, { "epoch": 2.43, "learning_rate": 2.2738798856053386e-06, "loss": 2.2336, "step": 12745 }, { "epoch": 2.43, "learning_rate": 2.2738226882745474e-06, "loss": 2.2023, "step": 12746 }, { "epoch": 2.43, "learning_rate": 2.2737654909437558e-06, "loss": 2.7344, "step": 12747 }, { "epoch": 2.43, "learning_rate": 2.273708293612965e-06, "loss": 2.8618, "step": 12748 }, { "epoch": 2.43, "learning_rate": 2.2736510962821733e-06, "loss": 2.8468, "step": 12749 }, { "epoch": 2.43, "learning_rate": 2.2735938989513826e-06, "loss": 2.6985, "step": 12750 }, { "epoch": 2.43, "learning_rate": 2.273536701620591e-06, "loss": 2.423, "step": 12751 }, { "epoch": 2.43, "learning_rate": 2.2734795042897997e-06, "loss": 2.5274, "step": 12752 }, { "epoch": 2.43, "learning_rate": 2.273422306959009e-06, "loss": 2.7117, "step": 12753 }, { "epoch": 2.43, "learning_rate": 2.2733651096282173e-06, "loss": 3.7949, "step": 12754 }, { "epoch": 2.43, "learning_rate": 2.273307912297426e-06, "loss": 3.0994, "step": 12755 }, { "epoch": 2.43, "learning_rate": 2.273250714966635e-06, "loss": 2.4171, "step": 12756 }, { "epoch": 2.43, "learning_rate": 2.2731935176358437e-06, "loss": 2.5287, "step": 12757 }, { "epoch": 2.43, "learning_rate": 2.2731363203050525e-06, "loss": 2.0742, "step": 12758 }, { "epoch": 2.43, "learning_rate": 2.2730791229742613e-06, "loss": 3.3394, "step": 12759 }, { "epoch": 2.43, "learning_rate": 2.27302192564347e-06, "loss": 2.7862, "step": 12760 }, { "epoch": 2.43, "learning_rate": 2.2729647283126784e-06, "loss": 2.5433, "step": 12761 }, { "epoch": 2.43, "learning_rate": 2.2729075309818877e-06, "loss": 2.6719, "step": 12762 }, { "epoch": 2.43, "learning_rate": 2.2728503336510965e-06, "loss": 2.3387, "step": 12763 }, { "epoch": 2.43, "learning_rate": 2.2727931363203052e-06, "loss": 3.2966, "step": 12764 }, { "epoch": 2.43, "learning_rate": 2.272735938989514e-06, "loss": 2.1873, "step": 12765 }, { "epoch": 2.43, "learning_rate": 2.2726787416587224e-06, "loss": 2.1608, "step": 12766 }, { "epoch": 2.43, "learning_rate": 2.2726215443279316e-06, "loss": 2.9897, "step": 12767 }, { "epoch": 2.43, "learning_rate": 2.2725643469971404e-06, "loss": 3.3464, "step": 12768 }, { "epoch": 2.43, "learning_rate": 2.272507149666349e-06, "loss": 3.3832, "step": 12769 }, { "epoch": 2.43, "learning_rate": 2.272449952335558e-06, "loss": 2.613, "step": 12770 }, { "epoch": 2.43, "learning_rate": 2.2723927550047664e-06, "loss": 3.1184, "step": 12771 }, { "epoch": 2.43, "learning_rate": 2.272335557673975e-06, "loss": 2.283, "step": 12772 }, { "epoch": 2.43, "learning_rate": 2.272278360343184e-06, "loss": 3.5789, "step": 12773 }, { "epoch": 2.43, "learning_rate": 2.2722211630123928e-06, "loss": 1.7911, "step": 12774 }, { "epoch": 2.43, "learning_rate": 2.2721639656816016e-06, "loss": 2.977, "step": 12775 }, { "epoch": 2.43, "learning_rate": 2.2721067683508104e-06, "loss": 2.4181, "step": 12776 }, { "epoch": 2.43, "learning_rate": 2.272049571020019e-06, "loss": 2.7673, "step": 12777 }, { "epoch": 2.43, "learning_rate": 2.271992373689228e-06, "loss": 3.0178, "step": 12778 }, { "epoch": 2.43, "learning_rate": 2.2719351763584367e-06, "loss": 2.6948, "step": 12779 }, { "epoch": 2.43, "learning_rate": 2.2718779790276455e-06, "loss": 2.3921, "step": 12780 }, { "epoch": 2.43, "learning_rate": 2.2718207816968543e-06, "loss": 1.7773, "step": 12781 }, { "epoch": 2.43, "learning_rate": 2.271763584366063e-06, "loss": 4.4788, "step": 12782 }, { "epoch": 2.43, "learning_rate": 2.2717063870352715e-06, "loss": 2.0457, "step": 12783 }, { "epoch": 2.44, "learning_rate": 2.2716491897044807e-06, "loss": 3.2868, "step": 12784 }, { "epoch": 2.44, "learning_rate": 2.2715919923736895e-06, "loss": 2.1191, "step": 12785 }, { "epoch": 2.44, "learning_rate": 2.271534795042898e-06, "loss": 2.3506, "step": 12786 }, { "epoch": 2.44, "learning_rate": 2.271477597712107e-06, "loss": 2.8446, "step": 12787 }, { "epoch": 2.44, "learning_rate": 2.2714204003813155e-06, "loss": 2.0418, "step": 12788 }, { "epoch": 2.44, "learning_rate": 2.2713632030505242e-06, "loss": 2.1447, "step": 12789 }, { "epoch": 2.44, "learning_rate": 2.271306005719733e-06, "loss": 3.1367, "step": 12790 }, { "epoch": 2.44, "learning_rate": 2.271248808388942e-06, "loss": 1.8644, "step": 12791 }, { "epoch": 2.44, "learning_rate": 2.2711916110581506e-06, "loss": 2.2204, "step": 12792 }, { "epoch": 2.44, "learning_rate": 2.2711344137273594e-06, "loss": 1.8736, "step": 12793 }, { "epoch": 2.44, "learning_rate": 2.2710772163965682e-06, "loss": 3.4096, "step": 12794 }, { "epoch": 2.44, "learning_rate": 2.271020019065777e-06, "loss": 2.0102, "step": 12795 }, { "epoch": 2.44, "learning_rate": 2.270962821734986e-06, "loss": 2.0168, "step": 12796 }, { "epoch": 2.44, "learning_rate": 2.2709056244041946e-06, "loss": 3.0978, "step": 12797 }, { "epoch": 2.44, "learning_rate": 2.2708484270734034e-06, "loss": 1.8698, "step": 12798 }, { "epoch": 2.44, "learning_rate": 2.270791229742612e-06, "loss": 3.5115, "step": 12799 }, { "epoch": 2.44, "learning_rate": 2.2707340324118206e-06, "loss": 2.8565, "step": 12800 }, { "epoch": 2.44, "learning_rate": 2.2706768350810298e-06, "loss": 3.5245, "step": 12801 }, { "epoch": 2.44, "learning_rate": 2.2706196377502386e-06, "loss": 2.0405, "step": 12802 }, { "epoch": 2.44, "learning_rate": 2.270562440419447e-06, "loss": 3.1812, "step": 12803 }, { "epoch": 2.44, "learning_rate": 2.270505243088656e-06, "loss": 2.6982, "step": 12804 }, { "epoch": 2.44, "learning_rate": 2.2704480457578645e-06, "loss": 2.0674, "step": 12805 }, { "epoch": 2.44, "learning_rate": 2.2703908484270733e-06, "loss": 2.4648, "step": 12806 }, { "epoch": 2.44, "learning_rate": 2.2703336510962825e-06, "loss": 2.0794, "step": 12807 }, { "epoch": 2.44, "learning_rate": 2.270276453765491e-06, "loss": 2.1476, "step": 12808 }, { "epoch": 2.44, "learning_rate": 2.2702192564346997e-06, "loss": 2.7613, "step": 12809 }, { "epoch": 2.44, "learning_rate": 2.2701620591039085e-06, "loss": 3.5808, "step": 12810 }, { "epoch": 2.44, "learning_rate": 2.2701048617731173e-06, "loss": 2.3723, "step": 12811 }, { "epoch": 2.44, "learning_rate": 2.270047664442326e-06, "loss": 2.4708, "step": 12812 }, { "epoch": 2.44, "learning_rate": 2.269990467111535e-06, "loss": 3.3193, "step": 12813 }, { "epoch": 2.44, "learning_rate": 2.2699332697807437e-06, "loss": 2.623, "step": 12814 }, { "epoch": 2.44, "learning_rate": 2.2698760724499525e-06, "loss": 3.2176, "step": 12815 }, { "epoch": 2.44, "learning_rate": 2.2698188751191613e-06, "loss": 2.0631, "step": 12816 }, { "epoch": 2.44, "learning_rate": 2.26976167778837e-06, "loss": 4.0438, "step": 12817 }, { "epoch": 2.44, "learning_rate": 2.269704480457579e-06, "loss": 2.6814, "step": 12818 }, { "epoch": 2.44, "learning_rate": 2.2696472831267876e-06, "loss": 2.5926, "step": 12819 }, { "epoch": 2.44, "learning_rate": 2.269590085795996e-06, "loss": 1.2628, "step": 12820 }, { "epoch": 2.44, "learning_rate": 2.2695328884652052e-06, "loss": 3.1415, "step": 12821 }, { "epoch": 2.44, "learning_rate": 2.2694756911344136e-06, "loss": 2.7066, "step": 12822 }, { "epoch": 2.44, "learning_rate": 2.2694184938036224e-06, "loss": 2.9265, "step": 12823 }, { "epoch": 2.44, "learning_rate": 2.2693612964728316e-06, "loss": 1.3312, "step": 12824 }, { "epoch": 2.44, "learning_rate": 2.26930409914204e-06, "loss": 2.2294, "step": 12825 }, { "epoch": 2.44, "learning_rate": 2.2692469018112488e-06, "loss": 2.7188, "step": 12826 }, { "epoch": 2.44, "learning_rate": 2.2691897044804576e-06, "loss": 3.1288, "step": 12827 }, { "epoch": 2.44, "learning_rate": 2.2691325071496664e-06, "loss": 2.583, "step": 12828 }, { "epoch": 2.44, "learning_rate": 2.269075309818875e-06, "loss": 2.5004, "step": 12829 }, { "epoch": 2.44, "learning_rate": 2.269018112488084e-06, "loss": 4.2344, "step": 12830 }, { "epoch": 2.44, "learning_rate": 2.2689609151572927e-06, "loss": 3.5271, "step": 12831 }, { "epoch": 2.44, "learning_rate": 2.2689037178265015e-06, "loss": 2.3882, "step": 12832 }, { "epoch": 2.44, "learning_rate": 2.2688465204957103e-06, "loss": 3.0291, "step": 12833 }, { "epoch": 2.44, "learning_rate": 2.268789323164919e-06, "loss": 3.044, "step": 12834 }, { "epoch": 2.44, "learning_rate": 2.268732125834128e-06, "loss": 2.0818, "step": 12835 }, { "epoch": 2.44, "learning_rate": 2.2686749285033367e-06, "loss": 3.026, "step": 12836 }, { "epoch": 2.45, "learning_rate": 2.268617731172545e-06, "loss": 3.046, "step": 12837 }, { "epoch": 2.45, "learning_rate": 2.2685605338417543e-06, "loss": 2.9983, "step": 12838 }, { "epoch": 2.45, "learning_rate": 2.2685033365109627e-06, "loss": 2.057, "step": 12839 }, { "epoch": 2.45, "learning_rate": 2.2684461391801715e-06, "loss": 2.8026, "step": 12840 }, { "epoch": 2.45, "learning_rate": 2.2683889418493807e-06, "loss": 2.6257, "step": 12841 }, { "epoch": 2.45, "learning_rate": 2.268331744518589e-06, "loss": 2.2143, "step": 12842 }, { "epoch": 2.45, "learning_rate": 2.268274547187798e-06, "loss": 2.1494, "step": 12843 }, { "epoch": 2.45, "learning_rate": 2.2682173498570066e-06, "loss": 2.7327, "step": 12844 }, { "epoch": 2.45, "learning_rate": 2.2681601525262154e-06, "loss": 2.9315, "step": 12845 }, { "epoch": 2.45, "learning_rate": 2.2681029551954242e-06, "loss": 2.5231, "step": 12846 }, { "epoch": 2.45, "learning_rate": 2.268045757864633e-06, "loss": 2.8382, "step": 12847 }, { "epoch": 2.45, "learning_rate": 2.267988560533842e-06, "loss": 2.2861, "step": 12848 }, { "epoch": 2.45, "learning_rate": 2.2679313632030506e-06, "loss": 2.4213, "step": 12849 }, { "epoch": 2.45, "learning_rate": 2.2678741658722594e-06, "loss": 3.1359, "step": 12850 }, { "epoch": 2.45, "learning_rate": 2.267816968541468e-06, "loss": 2.4916, "step": 12851 }, { "epoch": 2.45, "learning_rate": 2.267759771210677e-06, "loss": 2.5531, "step": 12852 }, { "epoch": 2.45, "learning_rate": 2.2677025738798858e-06, "loss": 2.5298, "step": 12853 }, { "epoch": 2.45, "learning_rate": 2.267645376549094e-06, "loss": 3.4732, "step": 12854 }, { "epoch": 2.45, "learning_rate": 2.2675881792183034e-06, "loss": 2.8083, "step": 12855 }, { "epoch": 2.45, "learning_rate": 2.267530981887512e-06, "loss": 3.4773, "step": 12856 }, { "epoch": 2.45, "learning_rate": 2.2674737845567205e-06, "loss": 3.4068, "step": 12857 }, { "epoch": 2.45, "learning_rate": 2.2674165872259298e-06, "loss": 3.0135, "step": 12858 }, { "epoch": 2.45, "learning_rate": 2.267359389895138e-06, "loss": 1.8635, "step": 12859 }, { "epoch": 2.45, "learning_rate": 2.267302192564347e-06, "loss": 3.5339, "step": 12860 }, { "epoch": 2.45, "learning_rate": 2.2672449952335557e-06, "loss": 2.2581, "step": 12861 }, { "epoch": 2.45, "learning_rate": 2.2671877979027645e-06, "loss": 2.7681, "step": 12862 }, { "epoch": 2.45, "learning_rate": 2.2671306005719737e-06, "loss": 3.9486, "step": 12863 }, { "epoch": 2.45, "learning_rate": 2.267073403241182e-06, "loss": 2.3456, "step": 12864 }, { "epoch": 2.45, "learning_rate": 2.267016205910391e-06, "loss": 2.5353, "step": 12865 }, { "epoch": 2.45, "learning_rate": 2.2669590085795997e-06, "loss": 2.3865, "step": 12866 }, { "epoch": 2.45, "learning_rate": 2.2669018112488085e-06, "loss": 2.0607, "step": 12867 }, { "epoch": 2.45, "learning_rate": 2.2668446139180173e-06, "loss": 2.914, "step": 12868 }, { "epoch": 2.45, "learning_rate": 2.266787416587226e-06, "loss": 2.4924, "step": 12869 }, { "epoch": 2.45, "learning_rate": 2.266730219256435e-06, "loss": 2.9587, "step": 12870 }, { "epoch": 2.45, "learning_rate": 2.2666730219256432e-06, "loss": 3.4988, "step": 12871 }, { "epoch": 2.45, "learning_rate": 2.2666158245948524e-06, "loss": 2.627, "step": 12872 }, { "epoch": 2.45, "learning_rate": 2.2665586272640612e-06, "loss": 3.0678, "step": 12873 }, { "epoch": 2.45, "learning_rate": 2.2665014299332696e-06, "loss": 2.4248, "step": 12874 }, { "epoch": 2.45, "learning_rate": 2.266444232602479e-06, "loss": 2.1173, "step": 12875 }, { "epoch": 2.45, "learning_rate": 2.266387035271687e-06, "loss": 2.91, "step": 12876 }, { "epoch": 2.45, "learning_rate": 2.266329837940896e-06, "loss": 1.5858, "step": 12877 }, { "epoch": 2.45, "learning_rate": 2.2662726406101048e-06, "loss": 2.3451, "step": 12878 }, { "epoch": 2.45, "learning_rate": 2.2662154432793136e-06, "loss": 1.7198, "step": 12879 }, { "epoch": 2.45, "learning_rate": 2.266158245948523e-06, "loss": 2.8464, "step": 12880 }, { "epoch": 2.45, "learning_rate": 2.266101048617731e-06, "loss": 3.0074, "step": 12881 }, { "epoch": 2.45, "learning_rate": 2.26604385128694e-06, "loss": 2.2351, "step": 12882 }, { "epoch": 2.45, "learning_rate": 2.2659866539561487e-06, "loss": 2.413, "step": 12883 }, { "epoch": 2.45, "learning_rate": 2.2659294566253575e-06, "loss": 2.1722, "step": 12884 }, { "epoch": 2.45, "learning_rate": 2.2658722592945663e-06, "loss": 2.3353, "step": 12885 }, { "epoch": 2.45, "learning_rate": 2.265815061963775e-06, "loss": 3.1157, "step": 12886 }, { "epoch": 2.45, "learning_rate": 2.265757864632984e-06, "loss": 2.9992, "step": 12887 }, { "epoch": 2.45, "learning_rate": 2.2657006673021923e-06, "loss": 3.1561, "step": 12888 }, { "epoch": 2.46, "learning_rate": 2.2656434699714015e-06, "loss": 2.093, "step": 12889 }, { "epoch": 2.46, "learning_rate": 2.2655862726406103e-06, "loss": 2.4105, "step": 12890 }, { "epoch": 2.46, "learning_rate": 2.2655290753098187e-06, "loss": 2.1819, "step": 12891 }, { "epoch": 2.46, "learning_rate": 2.265471877979028e-06, "loss": 3.0813, "step": 12892 }, { "epoch": 2.46, "learning_rate": 2.2654146806482363e-06, "loss": 2.8431, "step": 12893 }, { "epoch": 2.46, "learning_rate": 2.265357483317445e-06, "loss": 2.6027, "step": 12894 }, { "epoch": 2.46, "learning_rate": 2.265300285986654e-06, "loss": 2.2832, "step": 12895 }, { "epoch": 2.46, "learning_rate": 2.2652430886558626e-06, "loss": 2.8725, "step": 12896 }, { "epoch": 2.46, "learning_rate": 2.265185891325072e-06, "loss": 1.8841, "step": 12897 }, { "epoch": 2.46, "learning_rate": 2.2651286939942802e-06, "loss": 1.6016, "step": 12898 }, { "epoch": 2.46, "learning_rate": 2.265071496663489e-06, "loss": 2.319, "step": 12899 }, { "epoch": 2.46, "learning_rate": 2.265014299332698e-06, "loss": 3.5196, "step": 12900 }, { "epoch": 2.46, "learning_rate": 2.2649571020019066e-06, "loss": 2.9522, "step": 12901 }, { "epoch": 2.46, "learning_rate": 2.2648999046711154e-06, "loss": 2.6775, "step": 12902 }, { "epoch": 2.46, "learning_rate": 2.264842707340324e-06, "loss": 3.0151, "step": 12903 }, { "epoch": 2.46, "learning_rate": 2.264785510009533e-06, "loss": 3.1137, "step": 12904 }, { "epoch": 2.46, "learning_rate": 2.264728312678742e-06, "loss": 2.2711, "step": 12905 }, { "epoch": 2.46, "learning_rate": 2.2646711153479506e-06, "loss": 2.8447, "step": 12906 }, { "epoch": 2.46, "learning_rate": 2.2646139180171594e-06, "loss": 1.8628, "step": 12907 }, { "epoch": 2.46, "learning_rate": 2.2645567206863677e-06, "loss": 2.0013, "step": 12908 }, { "epoch": 2.46, "learning_rate": 2.264499523355577e-06, "loss": 1.8452, "step": 12909 }, { "epoch": 2.46, "learning_rate": 2.2644423260247853e-06, "loss": 3.2693, "step": 12910 }, { "epoch": 2.46, "learning_rate": 2.2643851286939946e-06, "loss": 2.7239, "step": 12911 }, { "epoch": 2.46, "learning_rate": 2.2643279313632033e-06, "loss": 2.8663, "step": 12912 }, { "epoch": 2.46, "learning_rate": 2.2642707340324117e-06, "loss": 3.2481, "step": 12913 }, { "epoch": 2.46, "learning_rate": 2.264213536701621e-06, "loss": 2.6128, "step": 12914 }, { "epoch": 2.46, "learning_rate": 2.2641563393708293e-06, "loss": 2.864, "step": 12915 }, { "epoch": 2.46, "learning_rate": 2.264099142040038e-06, "loss": 3.1471, "step": 12916 }, { "epoch": 2.46, "learning_rate": 2.264041944709247e-06, "loss": 2.5255, "step": 12917 }, { "epoch": 2.46, "learning_rate": 2.2639847473784557e-06, "loss": 2.7808, "step": 12918 }, { "epoch": 2.46, "learning_rate": 2.2639275500476645e-06, "loss": 2.9212, "step": 12919 }, { "epoch": 2.46, "learning_rate": 2.2638703527168733e-06, "loss": 2.539, "step": 12920 }, { "epoch": 2.46, "learning_rate": 2.263813155386082e-06, "loss": 2.8415, "step": 12921 }, { "epoch": 2.46, "learning_rate": 2.263755958055291e-06, "loss": 1.6571, "step": 12922 }, { "epoch": 2.46, "learning_rate": 2.2636987607244997e-06, "loss": 2.3627, "step": 12923 }, { "epoch": 2.46, "learning_rate": 2.2636415633937084e-06, "loss": 3.5152, "step": 12924 }, { "epoch": 2.46, "learning_rate": 2.263584366062917e-06, "loss": 3.0567, "step": 12925 }, { "epoch": 2.46, "learning_rate": 2.263527168732126e-06, "loss": 2.5709, "step": 12926 }, { "epoch": 2.46, "learning_rate": 2.2634699714013344e-06, "loss": 2.5562, "step": 12927 }, { "epoch": 2.46, "learning_rate": 2.2634127740705436e-06, "loss": 2.6879, "step": 12928 }, { "epoch": 2.46, "learning_rate": 2.2633555767397524e-06, "loss": 2.9539, "step": 12929 }, { "epoch": 2.46, "learning_rate": 2.263298379408961e-06, "loss": 2.8687, "step": 12930 }, { "epoch": 2.46, "learning_rate": 2.26324118207817e-06, "loss": 2.6132, "step": 12931 }, { "epoch": 2.46, "learning_rate": 2.2631839847473784e-06, "loss": 3.3229, "step": 12932 }, { "epoch": 2.46, "learning_rate": 2.263126787416587e-06, "loss": 3.0546, "step": 12933 }, { "epoch": 2.46, "learning_rate": 2.263069590085796e-06, "loss": 2.4998, "step": 12934 }, { "epoch": 2.46, "learning_rate": 2.2630123927550048e-06, "loss": 2.7797, "step": 12935 }, { "epoch": 2.46, "learning_rate": 2.2629551954242136e-06, "loss": 1.7204, "step": 12936 }, { "epoch": 2.46, "learning_rate": 2.2628979980934223e-06, "loss": 2.7618, "step": 12937 }, { "epoch": 2.46, "learning_rate": 2.262840800762631e-06, "loss": 2.5333, "step": 12938 }, { "epoch": 2.46, "learning_rate": 2.26278360343184e-06, "loss": 3.4799, "step": 12939 }, { "epoch": 2.46, "learning_rate": 2.2627264061010487e-06, "loss": 2.8874, "step": 12940 }, { "epoch": 2.46, "learning_rate": 2.2626692087702575e-06, "loss": 2.5994, "step": 12941 }, { "epoch": 2.47, "learning_rate": 2.262612011439466e-06, "loss": 2.5026, "step": 12942 }, { "epoch": 2.47, "learning_rate": 2.262554814108675e-06, "loss": 2.4256, "step": 12943 }, { "epoch": 2.47, "learning_rate": 2.262497616777884e-06, "loss": 2.5868, "step": 12944 }, { "epoch": 2.47, "learning_rate": 2.2624404194470927e-06, "loss": 2.9556, "step": 12945 }, { "epoch": 2.47, "learning_rate": 2.2623832221163015e-06, "loss": 2.2879, "step": 12946 }, { "epoch": 2.47, "learning_rate": 2.26232602478551e-06, "loss": 2.8498, "step": 12947 }, { "epoch": 2.47, "learning_rate": 2.262268827454719e-06, "loss": 3.5093, "step": 12948 }, { "epoch": 2.47, "learning_rate": 2.2622116301239274e-06, "loss": 3.1726, "step": 12949 }, { "epoch": 2.47, "learning_rate": 2.2621544327931362e-06, "loss": 2.9836, "step": 12950 }, { "epoch": 2.47, "learning_rate": 2.2620972354623455e-06, "loss": 3.1895, "step": 12951 }, { "epoch": 2.47, "learning_rate": 2.262040038131554e-06, "loss": 2.6255, "step": 12952 }, { "epoch": 2.47, "learning_rate": 2.2619828408007626e-06, "loss": 3.2494, "step": 12953 }, { "epoch": 2.47, "learning_rate": 2.2619256434699714e-06, "loss": 2.6707, "step": 12954 }, { "epoch": 2.47, "learning_rate": 2.26186844613918e-06, "loss": 2.3827, "step": 12955 }, { "epoch": 2.47, "learning_rate": 2.261811248808389e-06, "loss": 2.803, "step": 12956 }, { "epoch": 2.47, "learning_rate": 2.261754051477598e-06, "loss": 2.3514, "step": 12957 }, { "epoch": 2.47, "learning_rate": 2.2616968541468066e-06, "loss": 1.9772, "step": 12958 }, { "epoch": 2.47, "learning_rate": 2.261639656816015e-06, "loss": 2.4422, "step": 12959 }, { "epoch": 2.47, "learning_rate": 2.261582459485224e-06, "loss": 2.6003, "step": 12960 }, { "epoch": 2.47, "learning_rate": 2.261525262154433e-06, "loss": 2.6359, "step": 12961 }, { "epoch": 2.47, "learning_rate": 2.2614680648236418e-06, "loss": 2.2951, "step": 12962 }, { "epoch": 2.47, "learning_rate": 2.2614108674928506e-06, "loss": 2.3289, "step": 12963 }, { "epoch": 2.47, "learning_rate": 2.261353670162059e-06, "loss": 3.105, "step": 12964 }, { "epoch": 2.47, "learning_rate": 2.261296472831268e-06, "loss": 2.9541, "step": 12965 }, { "epoch": 2.47, "learning_rate": 2.2612392755004765e-06, "loss": 2.1812, "step": 12966 }, { "epoch": 2.47, "learning_rate": 2.2611820781696853e-06, "loss": 3.4058, "step": 12967 }, { "epoch": 2.47, "learning_rate": 2.2611248808388945e-06, "loss": 2.5905, "step": 12968 }, { "epoch": 2.47, "learning_rate": 2.261067683508103e-06, "loss": 2.5702, "step": 12969 }, { "epoch": 2.47, "learning_rate": 2.2610104861773117e-06, "loss": 3.33, "step": 12970 }, { "epoch": 2.47, "learning_rate": 2.2609532888465205e-06, "loss": 3.0422, "step": 12971 }, { "epoch": 2.47, "learning_rate": 2.2608960915157293e-06, "loss": 2.7131, "step": 12972 }, { "epoch": 2.47, "learning_rate": 2.260838894184938e-06, "loss": 3.1574, "step": 12973 }, { "epoch": 2.47, "learning_rate": 2.260781696854147e-06, "loss": 2.5962, "step": 12974 }, { "epoch": 2.47, "learning_rate": 2.2607244995233557e-06, "loss": 3.2876, "step": 12975 }, { "epoch": 2.47, "learning_rate": 2.2606673021925645e-06, "loss": 2.6345, "step": 12976 }, { "epoch": 2.47, "learning_rate": 2.2606101048617733e-06, "loss": 2.0192, "step": 12977 }, { "epoch": 2.47, "learning_rate": 2.260552907530982e-06, "loss": 2.7805, "step": 12978 }, { "epoch": 2.47, "learning_rate": 2.260495710200191e-06, "loss": 2.6758, "step": 12979 }, { "epoch": 2.47, "learning_rate": 2.2604385128693996e-06, "loss": 2.7251, "step": 12980 }, { "epoch": 2.47, "learning_rate": 2.260381315538608e-06, "loss": 2.6831, "step": 12981 }, { "epoch": 2.47, "learning_rate": 2.2603241182078172e-06, "loss": 2.1253, "step": 12982 }, { "epoch": 2.47, "learning_rate": 2.2602669208770256e-06, "loss": 3.3393, "step": 12983 }, { "epoch": 2.47, "learning_rate": 2.2602097235462344e-06, "loss": 2.8486, "step": 12984 }, { "epoch": 2.47, "learning_rate": 2.2601525262154436e-06, "loss": 2.5746, "step": 12985 }, { "epoch": 2.47, "learning_rate": 2.260095328884652e-06, "loss": 2.6377, "step": 12986 }, { "epoch": 2.47, "learning_rate": 2.2600381315538608e-06, "loss": 2.4005, "step": 12987 }, { "epoch": 2.47, "learning_rate": 2.2599809342230696e-06, "loss": 1.6895, "step": 12988 }, { "epoch": 2.47, "learning_rate": 2.2599237368922784e-06, "loss": 3.3071, "step": 12989 }, { "epoch": 2.47, "learning_rate": 2.259866539561487e-06, "loss": 3.2624, "step": 12990 }, { "epoch": 2.47, "learning_rate": 2.259809342230696e-06, "loss": 2.1844, "step": 12991 }, { "epoch": 2.47, "learning_rate": 2.2597521448999047e-06, "loss": 2.5789, "step": 12992 }, { "epoch": 2.47, "learning_rate": 2.2596949475691135e-06, "loss": 2.3605, "step": 12993 }, { "epoch": 2.48, "learning_rate": 2.2596377502383223e-06, "loss": 1.9626, "step": 12994 }, { "epoch": 2.48, "learning_rate": 2.259580552907531e-06, "loss": 2.6353, "step": 12995 }, { "epoch": 2.48, "learning_rate": 2.25952335557674e-06, "loss": 3.3821, "step": 12996 }, { "epoch": 2.48, "learning_rate": 2.2594661582459487e-06, "loss": 2.334, "step": 12997 }, { "epoch": 2.48, "learning_rate": 2.259408960915157e-06, "loss": 2.4686, "step": 12998 }, { "epoch": 2.48, "learning_rate": 2.2593517635843663e-06, "loss": 2.2569, "step": 12999 }, { "epoch": 2.48, "learning_rate": 2.259294566253575e-06, "loss": 3.039, "step": 13000 }, { "epoch": 2.48, "learning_rate": 2.2592373689227835e-06, "loss": 3.105, "step": 13001 }, { "epoch": 2.48, "learning_rate": 2.2591801715919927e-06, "loss": 3.22, "step": 13002 }, { "epoch": 2.48, "learning_rate": 2.259122974261201e-06, "loss": 3.0603, "step": 13003 }, { "epoch": 2.48, "learning_rate": 2.25906577693041e-06, "loss": 2.9385, "step": 13004 }, { "epoch": 2.48, "learning_rate": 2.2590085795996186e-06, "loss": 4.1238, "step": 13005 }, { "epoch": 2.48, "learning_rate": 2.2589513822688274e-06, "loss": 2.4594, "step": 13006 }, { "epoch": 2.48, "learning_rate": 2.2588941849380362e-06, "loss": 2.7924, "step": 13007 }, { "epoch": 2.48, "learning_rate": 2.258836987607245e-06, "loss": 2.2112, "step": 13008 }, { "epoch": 2.48, "learning_rate": 2.258779790276454e-06, "loss": 2.6944, "step": 13009 }, { "epoch": 2.48, "learning_rate": 2.2587225929456626e-06, "loss": 2.5285, "step": 13010 }, { "epoch": 2.48, "learning_rate": 2.2586653956148714e-06, "loss": 2.2741, "step": 13011 }, { "epoch": 2.48, "learning_rate": 2.25860819828408e-06, "loss": 2.3868, "step": 13012 }, { "epoch": 2.48, "learning_rate": 2.258551000953289e-06, "loss": 2.9895, "step": 13013 }, { "epoch": 2.48, "learning_rate": 2.2584938036224978e-06, "loss": 2.4639, "step": 13014 }, { "epoch": 2.48, "learning_rate": 2.258436606291706e-06, "loss": 2.3561, "step": 13015 }, { "epoch": 2.48, "learning_rate": 2.2583794089609154e-06, "loss": 2.9831, "step": 13016 }, { "epoch": 2.48, "learning_rate": 2.258322211630124e-06, "loss": 3.5137, "step": 13017 }, { "epoch": 2.48, "learning_rate": 2.2582650142993325e-06, "loss": 1.9173, "step": 13018 }, { "epoch": 2.48, "learning_rate": 2.2582078169685417e-06, "loss": 3.0455, "step": 13019 }, { "epoch": 2.48, "learning_rate": 2.25815061963775e-06, "loss": 2.2738, "step": 13020 }, { "epoch": 2.48, "learning_rate": 2.258093422306959e-06, "loss": 3.1481, "step": 13021 }, { "epoch": 2.48, "learning_rate": 2.2580362249761677e-06, "loss": 3.2038, "step": 13022 }, { "epoch": 2.48, "learning_rate": 2.2579790276453765e-06, "loss": 3.1504, "step": 13023 }, { "epoch": 2.48, "learning_rate": 2.2579218303145853e-06, "loss": 2.5427, "step": 13024 }, { "epoch": 2.48, "learning_rate": 2.257864632983794e-06, "loss": 2.8858, "step": 13025 }, { "epoch": 2.48, "learning_rate": 2.257807435653003e-06, "loss": 1.9396, "step": 13026 }, { "epoch": 2.48, "learning_rate": 2.2577502383222117e-06, "loss": 2.1867, "step": 13027 }, { "epoch": 2.48, "learning_rate": 2.2576930409914205e-06, "loss": 2.7831, "step": 13028 }, { "epoch": 2.48, "learning_rate": 2.2576358436606293e-06, "loss": 2.584, "step": 13029 }, { "epoch": 2.48, "learning_rate": 2.257578646329838e-06, "loss": 3.6546, "step": 13030 }, { "epoch": 2.48, "learning_rate": 2.257521448999047e-06, "loss": 3.1439, "step": 13031 }, { "epoch": 2.48, "learning_rate": 2.2574642516682556e-06, "loss": 2.7839, "step": 13032 }, { "epoch": 2.48, "learning_rate": 2.2574070543374644e-06, "loss": 2.8548, "step": 13033 }, { "epoch": 2.48, "learning_rate": 2.2573498570066732e-06, "loss": 3.5702, "step": 13034 }, { "epoch": 2.48, "learning_rate": 2.2572926596758816e-06, "loss": 2.4869, "step": 13035 }, { "epoch": 2.48, "learning_rate": 2.257235462345091e-06, "loss": 2.968, "step": 13036 }, { "epoch": 2.48, "learning_rate": 2.257178265014299e-06, "loss": 1.8485, "step": 13037 }, { "epoch": 2.48, "learning_rate": 2.257121067683508e-06, "loss": 2.9208, "step": 13038 }, { "epoch": 2.48, "learning_rate": 2.257063870352717e-06, "loss": 2.8669, "step": 13039 }, { "epoch": 2.48, "learning_rate": 2.2570066730219256e-06, "loss": 1.88, "step": 13040 }, { "epoch": 2.48, "learning_rate": 2.2569494756911348e-06, "loss": 2.3381, "step": 13041 }, { "epoch": 2.48, "learning_rate": 2.256892278360343e-06, "loss": 2.8585, "step": 13042 }, { "epoch": 2.48, "learning_rate": 2.256835081029552e-06, "loss": 3.6648, "step": 13043 }, { "epoch": 2.48, "learning_rate": 2.2567778836987607e-06, "loss": 3.1399, "step": 13044 }, { "epoch": 2.48, "learning_rate": 2.2567206863679695e-06, "loss": 3.1117, "step": 13045 }, { "epoch": 2.48, "learning_rate": 2.2566634890371783e-06, "loss": 2.625, "step": 13046 }, { "epoch": 2.49, "learning_rate": 2.256606291706387e-06, "loss": 1.8766, "step": 13047 }, { "epoch": 2.49, "learning_rate": 2.256549094375596e-06, "loss": 2.7482, "step": 13048 }, { "epoch": 2.49, "learning_rate": 2.2564918970448047e-06, "loss": 2.6328, "step": 13049 }, { "epoch": 2.49, "learning_rate": 2.2564346997140135e-06, "loss": 2.806, "step": 13050 }, { "epoch": 2.49, "learning_rate": 2.2563775023832223e-06, "loss": 2.9718, "step": 13051 }, { "epoch": 2.49, "learning_rate": 2.2563203050524307e-06, "loss": 2.3658, "step": 13052 }, { "epoch": 2.49, "learning_rate": 2.25626310772164e-06, "loss": 2.5368, "step": 13053 }, { "epoch": 2.49, "learning_rate": 2.2562059103908483e-06, "loss": 3.9855, "step": 13054 }, { "epoch": 2.49, "learning_rate": 2.256148713060057e-06, "loss": 3.3616, "step": 13055 }, { "epoch": 2.49, "learning_rate": 2.2560915157292663e-06, "loss": 2.6283, "step": 13056 }, { "epoch": 2.49, "learning_rate": 2.2560343183984746e-06, "loss": 2.435, "step": 13057 }, { "epoch": 2.49, "learning_rate": 2.255977121067684e-06, "loss": 2.8233, "step": 13058 }, { "epoch": 2.49, "learning_rate": 2.2559199237368922e-06, "loss": 2.3858, "step": 13059 }, { "epoch": 2.49, "learning_rate": 2.255862726406101e-06, "loss": 2.1434, "step": 13060 }, { "epoch": 2.49, "learning_rate": 2.25580552907531e-06, "loss": 2.739, "step": 13061 }, { "epoch": 2.49, "learning_rate": 2.2557483317445186e-06, "loss": 3.243, "step": 13062 }, { "epoch": 2.49, "learning_rate": 2.2556911344137274e-06, "loss": 2.3665, "step": 13063 }, { "epoch": 2.49, "learning_rate": 2.255633937082936e-06, "loss": 3.6225, "step": 13064 }, { "epoch": 2.49, "learning_rate": 2.255576739752145e-06, "loss": 3.3336, "step": 13065 }, { "epoch": 2.49, "learning_rate": 2.2555195424213538e-06, "loss": 2.9947, "step": 13066 }, { "epoch": 2.49, "learning_rate": 2.2554623450905626e-06, "loss": 1.9629, "step": 13067 }, { "epoch": 2.49, "learning_rate": 2.2554051477597714e-06, "loss": 2.7864, "step": 13068 }, { "epoch": 2.49, "learning_rate": 2.2553479504289797e-06, "loss": 2.2668, "step": 13069 }, { "epoch": 2.49, "learning_rate": 2.255290753098189e-06, "loss": 2.8942, "step": 13070 }, { "epoch": 2.49, "learning_rate": 2.2552335557673973e-06, "loss": 3.7861, "step": 13071 }, { "epoch": 2.49, "learning_rate": 2.255176358436606e-06, "loss": 2.4734, "step": 13072 }, { "epoch": 2.49, "learning_rate": 2.2551191611058153e-06, "loss": 2.2708, "step": 13073 }, { "epoch": 2.49, "learning_rate": 2.2550619637750237e-06, "loss": 2.088, "step": 13074 }, { "epoch": 2.49, "learning_rate": 2.255004766444233e-06, "loss": 3.1478, "step": 13075 }, { "epoch": 2.49, "learning_rate": 2.2549475691134413e-06, "loss": 2.4606, "step": 13076 }, { "epoch": 2.49, "learning_rate": 2.25489037178265e-06, "loss": 3.4003, "step": 13077 }, { "epoch": 2.49, "learning_rate": 2.2548331744518593e-06, "loss": 3.9553, "step": 13078 }, { "epoch": 2.49, "learning_rate": 2.2547759771210677e-06, "loss": 2.7929, "step": 13079 }, { "epoch": 2.49, "learning_rate": 2.2547187797902765e-06, "loss": 2.9299, "step": 13080 }, { "epoch": 2.49, "learning_rate": 2.2546615824594853e-06, "loss": 2.7818, "step": 13081 }, { "epoch": 2.49, "learning_rate": 2.254604385128694e-06, "loss": 2.5377, "step": 13082 }, { "epoch": 2.49, "learning_rate": 2.254547187797903e-06, "loss": 2.1394, "step": 13083 }, { "epoch": 2.49, "learning_rate": 2.2544899904671116e-06, "loss": 2.4534, "step": 13084 }, { "epoch": 2.49, "learning_rate": 2.2544327931363204e-06, "loss": 2.3417, "step": 13085 }, { "epoch": 2.49, "learning_rate": 2.254375595805529e-06, "loss": 3.6741, "step": 13086 }, { "epoch": 2.49, "learning_rate": 2.254318398474738e-06, "loss": 2.8547, "step": 13087 }, { "epoch": 2.49, "learning_rate": 2.254261201143947e-06, "loss": 2.6836, "step": 13088 }, { "epoch": 2.49, "learning_rate": 2.254204003813155e-06, "loss": 2.9054, "step": 13089 }, { "epoch": 2.49, "learning_rate": 2.2541468064823644e-06, "loss": 2.7406, "step": 13090 }, { "epoch": 2.49, "learning_rate": 2.2540896091515728e-06, "loss": 2.3899, "step": 13091 }, { "epoch": 2.49, "learning_rate": 2.254032411820782e-06, "loss": 2.9325, "step": 13092 }, { "epoch": 2.49, "learning_rate": 2.2539752144899904e-06, "loss": 2.8339, "step": 13093 }, { "epoch": 2.49, "learning_rate": 2.253918017159199e-06, "loss": 3.4115, "step": 13094 }, { "epoch": 2.49, "learning_rate": 2.2538608198284084e-06, "loss": 2.7129, "step": 13095 }, { "epoch": 2.49, "learning_rate": 2.2538036224976168e-06, "loss": 2.5686, "step": 13096 }, { "epoch": 2.49, "learning_rate": 2.2537464251668255e-06, "loss": 1.8758, "step": 13097 }, { "epoch": 2.49, "learning_rate": 2.2536892278360343e-06, "loss": 2.9002, "step": 13098 }, { "epoch": 2.5, "learning_rate": 2.253632030505243e-06, "loss": 2.4001, "step": 13099 }, { "epoch": 2.5, "learning_rate": 2.253574833174452e-06, "loss": 3.1599, "step": 13100 }, { "epoch": 2.5, "learning_rate": 2.2535176358436607e-06, "loss": 2.4877, "step": 13101 }, { "epoch": 2.5, "learning_rate": 2.2534604385128695e-06, "loss": 2.9563, "step": 13102 }, { "epoch": 2.5, "learning_rate": 2.253403241182078e-06, "loss": 2.8354, "step": 13103 }, { "epoch": 2.5, "learning_rate": 2.253346043851287e-06, "loss": 2.748, "step": 13104 }, { "epoch": 2.5, "learning_rate": 2.253288846520496e-06, "loss": 2.1083, "step": 13105 }, { "epoch": 2.5, "learning_rate": 2.2532316491897047e-06, "loss": 1.9577, "step": 13106 }, { "epoch": 2.5, "learning_rate": 2.2531744518589135e-06, "loss": 2.8164, "step": 13107 }, { "epoch": 2.5, "learning_rate": 2.253117254528122e-06, "loss": 2.5281, "step": 13108 }, { "epoch": 2.5, "learning_rate": 2.253060057197331e-06, "loss": 2.4905, "step": 13109 }, { "epoch": 2.5, "learning_rate": 2.2530028598665394e-06, "loss": 2.3418, "step": 13110 }, { "epoch": 2.5, "learning_rate": 2.2529456625357482e-06, "loss": 2.7923, "step": 13111 }, { "epoch": 2.5, "learning_rate": 2.2528884652049575e-06, "loss": 3.0304, "step": 13112 }, { "epoch": 2.5, "learning_rate": 2.252831267874166e-06, "loss": 2.3456, "step": 13113 }, { "epoch": 2.5, "learning_rate": 2.2527740705433746e-06, "loss": 2.9479, "step": 13114 }, { "epoch": 2.5, "learning_rate": 2.2527168732125834e-06, "loss": 2.5647, "step": 13115 }, { "epoch": 2.5, "learning_rate": 2.252659675881792e-06, "loss": 3.3557, "step": 13116 }, { "epoch": 2.5, "learning_rate": 2.252602478551001e-06, "loss": 3.3073, "step": 13117 }, { "epoch": 2.5, "learning_rate": 2.25254528122021e-06, "loss": 2.336, "step": 13118 }, { "epoch": 2.5, "learning_rate": 2.2524880838894186e-06, "loss": 2.7736, "step": 13119 }, { "epoch": 2.5, "learning_rate": 2.2524308865586274e-06, "loss": 3.6654, "step": 13120 }, { "epoch": 2.5, "learning_rate": 2.252373689227836e-06, "loss": 2.5362, "step": 13121 }, { "epoch": 2.5, "learning_rate": 2.252316491897045e-06, "loss": 2.0622, "step": 13122 }, { "epoch": 2.5, "learning_rate": 2.2522592945662538e-06, "loss": 3.4173, "step": 13123 }, { "epoch": 2.5, "learning_rate": 2.2522020972354626e-06, "loss": 2.5199, "step": 13124 }, { "epoch": 2.5, "learning_rate": 2.252144899904671e-06, "loss": 3.622, "step": 13125 }, { "epoch": 2.5, "learning_rate": 2.25208770257388e-06, "loss": 2.6284, "step": 13126 }, { "epoch": 2.5, "learning_rate": 2.252030505243089e-06, "loss": 2.2676, "step": 13127 }, { "epoch": 2.5, "learning_rate": 2.2519733079122973e-06, "loss": 3.5481, "step": 13128 }, { "epoch": 2.5, "learning_rate": 2.2519161105815065e-06, "loss": 1.9766, "step": 13129 }, { "epoch": 2.5, "learning_rate": 2.251858913250715e-06, "loss": 2.5254, "step": 13130 }, { "epoch": 2.5, "learning_rate": 2.2518017159199237e-06, "loss": 2.6052, "step": 13131 }, { "epoch": 2.5, "learning_rate": 2.2517445185891325e-06, "loss": 2.5849, "step": 13132 }, { "epoch": 2.5, "learning_rate": 2.2516873212583413e-06, "loss": 1.8898, "step": 13133 }, { "epoch": 2.5, "learning_rate": 2.25163012392755e-06, "loss": 2.5382, "step": 13134 }, { "epoch": 2.5, "learning_rate": 2.251572926596759e-06, "loss": 3.2959, "step": 13135 }, { "epoch": 2.5, "learning_rate": 2.2515157292659677e-06, "loss": 2.121, "step": 13136 }, { "epoch": 2.5, "learning_rate": 2.2514585319351764e-06, "loss": 1.6881, "step": 13137 }, { "epoch": 2.5, "learning_rate": 2.2514013346043852e-06, "loss": 2.3151, "step": 13138 }, { "epoch": 2.5, "learning_rate": 2.251344137273594e-06, "loss": 2.6973, "step": 13139 }, { "epoch": 2.5, "learning_rate": 2.251286939942803e-06, "loss": 2.856, "step": 13140 }, { "epoch": 2.5, "learning_rate": 2.2512297426120116e-06, "loss": 2.5793, "step": 13141 }, { "epoch": 2.5, "learning_rate": 2.25117254528122e-06, "loss": 2.9962, "step": 13142 }, { "epoch": 2.5, "learning_rate": 2.2511153479504292e-06, "loss": 2.9421, "step": 13143 }, { "epoch": 2.5, "learning_rate": 2.251058150619638e-06, "loss": 2.5528, "step": 13144 }, { "epoch": 2.5, "learning_rate": 2.2510009532888464e-06, "loss": 2.3877, "step": 13145 }, { "epoch": 2.5, "learning_rate": 2.2509437559580556e-06, "loss": 2.7158, "step": 13146 }, { "epoch": 2.5, "learning_rate": 2.250886558627264e-06, "loss": 3.0263, "step": 13147 }, { "epoch": 2.5, "learning_rate": 2.2508293612964728e-06, "loss": 3.85, "step": 13148 }, { "epoch": 2.5, "learning_rate": 2.2507721639656816e-06, "loss": 2.0801, "step": 13149 }, { "epoch": 2.5, "learning_rate": 2.2507149666348903e-06, "loss": 2.5086, "step": 13150 }, { "epoch": 2.5, "learning_rate": 2.250657769304099e-06, "loss": 2.3432, "step": 13151 }, { "epoch": 2.51, "learning_rate": 2.250600571973308e-06, "loss": 2.8529, "step": 13152 }, { "epoch": 2.51, "learning_rate": 2.2505433746425167e-06, "loss": 2.2509, "step": 13153 }, { "epoch": 2.51, "learning_rate": 2.2504861773117255e-06, "loss": 3.6056, "step": 13154 }, { "epoch": 2.51, "learning_rate": 2.2504289799809343e-06, "loss": 2.4877, "step": 13155 }, { "epoch": 2.51, "learning_rate": 2.250371782650143e-06, "loss": 2.915, "step": 13156 }, { "epoch": 2.51, "learning_rate": 2.250314585319352e-06, "loss": 2.6897, "step": 13157 }, { "epoch": 2.51, "learning_rate": 2.2502573879885607e-06, "loss": 2.3156, "step": 13158 }, { "epoch": 2.51, "learning_rate": 2.250200190657769e-06, "loss": 1.8976, "step": 13159 }, { "epoch": 2.51, "learning_rate": 2.2501429933269783e-06, "loss": 2.0585, "step": 13160 }, { "epoch": 2.51, "learning_rate": 2.250085795996187e-06, "loss": 1.7303, "step": 13161 }, { "epoch": 2.51, "learning_rate": 2.2500285986653954e-06, "loss": 2.0283, "step": 13162 }, { "epoch": 2.51, "learning_rate": 2.2499714013346047e-06, "loss": 3.1773, "step": 13163 }, { "epoch": 2.51, "learning_rate": 2.249914204003813e-06, "loss": 2.787, "step": 13164 }, { "epoch": 2.51, "learning_rate": 2.249857006673022e-06, "loss": 2.7368, "step": 13165 }, { "epoch": 2.51, "learning_rate": 2.249799809342231e-06, "loss": 1.9422, "step": 13166 }, { "epoch": 2.51, "learning_rate": 2.2497426120114394e-06, "loss": 2.2662, "step": 13167 }, { "epoch": 2.51, "learning_rate": 2.249685414680648e-06, "loss": 3.6812, "step": 13168 }, { "epoch": 2.51, "learning_rate": 2.249628217349857e-06, "loss": 2.7999, "step": 13169 }, { "epoch": 2.51, "learning_rate": 2.249571020019066e-06, "loss": 2.9968, "step": 13170 }, { "epoch": 2.51, "learning_rate": 2.2495138226882746e-06, "loss": 3.3624, "step": 13171 }, { "epoch": 2.51, "learning_rate": 2.2494566253574834e-06, "loss": 2.7337, "step": 13172 }, { "epoch": 2.51, "learning_rate": 2.249399428026692e-06, "loss": 2.2248, "step": 13173 }, { "epoch": 2.51, "learning_rate": 2.249342230695901e-06, "loss": 1.6483, "step": 13174 }, { "epoch": 2.51, "learning_rate": 2.2492850333651098e-06, "loss": 2.6329, "step": 13175 }, { "epoch": 2.51, "learning_rate": 2.2492278360343186e-06, "loss": 3.2228, "step": 13176 }, { "epoch": 2.51, "learning_rate": 2.2491706387035274e-06, "loss": 2.6106, "step": 13177 }, { "epoch": 2.51, "learning_rate": 2.249113441372736e-06, "loss": 1.2972, "step": 13178 }, { "epoch": 2.51, "learning_rate": 2.2490562440419445e-06, "loss": 2.6666, "step": 13179 }, { "epoch": 2.51, "learning_rate": 2.2489990467111537e-06, "loss": 2.9223, "step": 13180 }, { "epoch": 2.51, "learning_rate": 2.248941849380362e-06, "loss": 2.4414, "step": 13181 }, { "epoch": 2.51, "learning_rate": 2.248884652049571e-06, "loss": 2.6116, "step": 13182 }, { "epoch": 2.51, "learning_rate": 2.24882745471878e-06, "loss": 2.4682, "step": 13183 }, { "epoch": 2.51, "learning_rate": 2.2487702573879885e-06, "loss": 2.5767, "step": 13184 }, { "epoch": 2.51, "learning_rate": 2.2487130600571973e-06, "loss": 2.3126, "step": 13185 }, { "epoch": 2.51, "learning_rate": 2.248655862726406e-06, "loss": 2.2781, "step": 13186 }, { "epoch": 2.51, "learning_rate": 2.248598665395615e-06, "loss": 2.4561, "step": 13187 }, { "epoch": 2.51, "learning_rate": 2.2485414680648237e-06, "loss": 2.8375, "step": 13188 }, { "epoch": 2.51, "learning_rate": 2.2484842707340325e-06, "loss": 2.1461, "step": 13189 }, { "epoch": 2.51, "learning_rate": 2.2484270734032413e-06, "loss": 2.5207, "step": 13190 }, { "epoch": 2.51, "learning_rate": 2.24836987607245e-06, "loss": 2.3459, "step": 13191 }, { "epoch": 2.51, "learning_rate": 2.248312678741659e-06, "loss": 2.69, "step": 13192 }, { "epoch": 2.51, "learning_rate": 2.2482554814108676e-06, "loss": 2.7572, "step": 13193 }, { "epoch": 2.51, "learning_rate": 2.2481982840800764e-06, "loss": 2.4751, "step": 13194 }, { "epoch": 2.51, "learning_rate": 2.2481410867492852e-06, "loss": 2.4301, "step": 13195 }, { "epoch": 2.51, "learning_rate": 2.2480838894184936e-06, "loss": 2.1194, "step": 13196 }, { "epoch": 2.51, "learning_rate": 2.248026692087703e-06, "loss": 2.5742, "step": 13197 }, { "epoch": 2.51, "learning_rate": 2.247969494756911e-06, "loss": 3.2795, "step": 13198 }, { "epoch": 2.51, "learning_rate": 2.24791229742612e-06, "loss": 3.1226, "step": 13199 }, { "epoch": 2.51, "learning_rate": 2.247855100095329e-06, "loss": 1.9136, "step": 13200 }, { "epoch": 2.51, "learning_rate": 2.2477979027645376e-06, "loss": 3.8564, "step": 13201 }, { "epoch": 2.51, "learning_rate": 2.2477407054337464e-06, "loss": 2.768, "step": 13202 }, { "epoch": 2.51, "learning_rate": 2.247683508102955e-06, "loss": 2.9001, "step": 13203 }, { "epoch": 2.52, "learning_rate": 2.247626310772164e-06, "loss": 2.8268, "step": 13204 }, { "epoch": 2.52, "learning_rate": 2.2475691134413727e-06, "loss": 3.8563, "step": 13205 }, { "epoch": 2.52, "learning_rate": 2.2475119161105815e-06, "loss": 2.9451, "step": 13206 }, { "epoch": 2.52, "learning_rate": 2.2474547187797903e-06, "loss": 2.2396, "step": 13207 }, { "epoch": 2.52, "learning_rate": 2.247397521448999e-06, "loss": 3.2562, "step": 13208 }, { "epoch": 2.52, "learning_rate": 2.247340324118208e-06, "loss": 3.0938, "step": 13209 }, { "epoch": 2.52, "learning_rate": 2.2472831267874167e-06, "loss": 2.2315, "step": 13210 }, { "epoch": 2.52, "learning_rate": 2.2472259294566255e-06, "loss": 1.9405, "step": 13211 }, { "epoch": 2.52, "learning_rate": 2.2471687321258343e-06, "loss": 2.6052, "step": 13212 }, { "epoch": 2.52, "learning_rate": 2.2471115347950427e-06, "loss": 3.4235, "step": 13213 }, { "epoch": 2.52, "learning_rate": 2.247054337464252e-06, "loss": 3.4109, "step": 13214 }, { "epoch": 2.52, "learning_rate": 2.2469971401334607e-06, "loss": 3.3056, "step": 13215 }, { "epoch": 2.52, "learning_rate": 2.246939942802669e-06, "loss": 2.8759, "step": 13216 }, { "epoch": 2.52, "learning_rate": 2.2468827454718783e-06, "loss": 3.1876, "step": 13217 }, { "epoch": 2.52, "learning_rate": 2.2468255481410866e-06, "loss": 2.1848, "step": 13218 }, { "epoch": 2.52, "learning_rate": 2.2467683508102954e-06, "loss": 2.2738, "step": 13219 }, { "epoch": 2.52, "learning_rate": 2.2467111534795042e-06, "loss": 3.562, "step": 13220 }, { "epoch": 2.52, "learning_rate": 2.246653956148713e-06, "loss": 2.8367, "step": 13221 }, { "epoch": 2.52, "learning_rate": 2.2465967588179222e-06, "loss": 2.9882, "step": 13222 }, { "epoch": 2.52, "learning_rate": 2.2465395614871306e-06, "loss": 3.2394, "step": 13223 }, { "epoch": 2.52, "learning_rate": 2.2464823641563394e-06, "loss": 2.8245, "step": 13224 }, { "epoch": 2.52, "learning_rate": 2.246425166825548e-06, "loss": 2.5077, "step": 13225 }, { "epoch": 2.52, "learning_rate": 2.246367969494757e-06, "loss": 3.0378, "step": 13226 }, { "epoch": 2.52, "learning_rate": 2.2463107721639658e-06, "loss": 2.2605, "step": 13227 }, { "epoch": 2.52, "learning_rate": 2.2462535748331746e-06, "loss": 2.9702, "step": 13228 }, { "epoch": 2.52, "learning_rate": 2.2461963775023834e-06, "loss": 2.1654, "step": 13229 }, { "epoch": 2.52, "learning_rate": 2.2461391801715917e-06, "loss": 2.2548, "step": 13230 }, { "epoch": 2.52, "learning_rate": 2.246081982840801e-06, "loss": 2.3281, "step": 13231 }, { "epoch": 2.52, "learning_rate": 2.2460247855100097e-06, "loss": 2.1762, "step": 13232 }, { "epoch": 2.52, "learning_rate": 2.245967588179218e-06, "loss": 2.57, "step": 13233 }, { "epoch": 2.52, "learning_rate": 2.2459103908484273e-06, "loss": 3.1902, "step": 13234 }, { "epoch": 2.52, "learning_rate": 2.2458531935176357e-06, "loss": 2.3422, "step": 13235 }, { "epoch": 2.52, "learning_rate": 2.245795996186845e-06, "loss": 3.1068, "step": 13236 }, { "epoch": 2.52, "learning_rate": 2.2457387988560533e-06, "loss": 2.4015, "step": 13237 }, { "epoch": 2.52, "learning_rate": 2.245681601525262e-06, "loss": 3.0499, "step": 13238 }, { "epoch": 2.52, "learning_rate": 2.2456244041944713e-06, "loss": 3.3038, "step": 13239 }, { "epoch": 2.52, "learning_rate": 2.2455672068636797e-06, "loss": 2.386, "step": 13240 }, { "epoch": 2.52, "learning_rate": 2.2455100095328885e-06, "loss": 2.8045, "step": 13241 }, { "epoch": 2.52, "learning_rate": 2.2454528122020973e-06, "loss": 1.4389, "step": 13242 }, { "epoch": 2.52, "learning_rate": 2.245395614871306e-06, "loss": 2.1623, "step": 13243 }, { "epoch": 2.52, "learning_rate": 2.245338417540515e-06, "loss": 2.5196, "step": 13244 }, { "epoch": 2.52, "learning_rate": 2.2452812202097236e-06, "loss": 3.1136, "step": 13245 }, { "epoch": 2.52, "learning_rate": 2.2452240228789324e-06, "loss": 3.8152, "step": 13246 }, { "epoch": 2.52, "learning_rate": 2.245166825548141e-06, "loss": 3.5698, "step": 13247 }, { "epoch": 2.52, "learning_rate": 2.24510962821735e-06, "loss": 2.864, "step": 13248 }, { "epoch": 2.52, "learning_rate": 2.245052430886559e-06, "loss": 2.7266, "step": 13249 }, { "epoch": 2.52, "learning_rate": 2.244995233555767e-06, "loss": 3.5856, "step": 13250 }, { "epoch": 2.52, "learning_rate": 2.2449380362249764e-06, "loss": 2.5524, "step": 13251 }, { "epoch": 2.52, "learning_rate": 2.2448808388941848e-06, "loss": 2.4448, "step": 13252 }, { "epoch": 2.52, "learning_rate": 2.244823641563394e-06, "loss": 2.7958, "step": 13253 }, { "epoch": 2.52, "learning_rate": 2.2447664442326028e-06, "loss": 2.3035, "step": 13254 }, { "epoch": 2.52, "learning_rate": 2.244709246901811e-06, "loss": 2.5597, "step": 13255 }, { "epoch": 2.52, "learning_rate": 2.2446520495710204e-06, "loss": 3.1944, "step": 13256 }, { "epoch": 2.53, "learning_rate": 2.2445948522402287e-06, "loss": 3.501, "step": 13257 }, { "epoch": 2.53, "learning_rate": 2.2445376549094375e-06, "loss": 2.6058, "step": 13258 }, { "epoch": 2.53, "learning_rate": 2.2444804575786463e-06, "loss": 3.4751, "step": 13259 }, { "epoch": 2.53, "learning_rate": 2.244423260247855e-06, "loss": 2.3535, "step": 13260 }, { "epoch": 2.53, "learning_rate": 2.244366062917064e-06, "loss": 2.1168, "step": 13261 }, { "epoch": 2.53, "learning_rate": 2.2443088655862727e-06, "loss": 2.8586, "step": 13262 }, { "epoch": 2.53, "learning_rate": 2.2442516682554815e-06, "loss": 2.7686, "step": 13263 }, { "epoch": 2.53, "learning_rate": 2.2441944709246903e-06, "loss": 2.7325, "step": 13264 }, { "epoch": 2.53, "learning_rate": 2.244137273593899e-06, "loss": 3.4781, "step": 13265 }, { "epoch": 2.53, "learning_rate": 2.244080076263108e-06, "loss": 2.2055, "step": 13266 }, { "epoch": 2.53, "learning_rate": 2.2440228789323163e-06, "loss": 3.2881, "step": 13267 }, { "epoch": 2.53, "learning_rate": 2.2439656816015255e-06, "loss": 3.8114, "step": 13268 }, { "epoch": 2.53, "learning_rate": 2.243908484270734e-06, "loss": 2.7709, "step": 13269 }, { "epoch": 2.53, "learning_rate": 2.243851286939943e-06, "loss": 2.6003, "step": 13270 }, { "epoch": 2.53, "learning_rate": 2.243794089609152e-06, "loss": 1.7626, "step": 13271 }, { "epoch": 2.53, "learning_rate": 2.2437368922783602e-06, "loss": 3.3117, "step": 13272 }, { "epoch": 2.53, "learning_rate": 2.2436796949475694e-06, "loss": 2.1229, "step": 13273 }, { "epoch": 2.53, "learning_rate": 2.243622497616778e-06, "loss": 3.0161, "step": 13274 }, { "epoch": 2.53, "learning_rate": 2.2435653002859866e-06, "loss": 3.1215, "step": 13275 }, { "epoch": 2.53, "learning_rate": 2.2435081029551954e-06, "loss": 3.0121, "step": 13276 }, { "epoch": 2.53, "learning_rate": 2.243450905624404e-06, "loss": 2.9707, "step": 13277 }, { "epoch": 2.53, "learning_rate": 2.243393708293613e-06, "loss": 3.4811, "step": 13278 }, { "epoch": 2.53, "learning_rate": 2.2433365109628218e-06, "loss": 2.8796, "step": 13279 }, { "epoch": 2.53, "learning_rate": 2.2432793136320306e-06, "loss": 3.0607, "step": 13280 }, { "epoch": 2.53, "learning_rate": 2.2432221163012394e-06, "loss": 2.5749, "step": 13281 }, { "epoch": 2.53, "learning_rate": 2.243164918970448e-06, "loss": 2.8137, "step": 13282 }, { "epoch": 2.53, "learning_rate": 2.243107721639657e-06, "loss": 2.7008, "step": 13283 }, { "epoch": 2.53, "learning_rate": 2.2430505243088653e-06, "loss": 2.7509, "step": 13284 }, { "epoch": 2.53, "learning_rate": 2.2429933269780745e-06, "loss": 2.4424, "step": 13285 }, { "epoch": 2.53, "learning_rate": 2.242936129647283e-06, "loss": 2.2933, "step": 13286 }, { "epoch": 2.53, "learning_rate": 2.242878932316492e-06, "loss": 2.159, "step": 13287 }, { "epoch": 2.53, "learning_rate": 2.242821734985701e-06, "loss": 2.3819, "step": 13288 }, { "epoch": 2.53, "learning_rate": 2.2427645376549093e-06, "loss": 3.1108, "step": 13289 }, { "epoch": 2.53, "learning_rate": 2.2427073403241185e-06, "loss": 2.3076, "step": 13290 }, { "epoch": 2.53, "learning_rate": 2.242650142993327e-06, "loss": 3.2789, "step": 13291 }, { "epoch": 2.53, "learning_rate": 2.2425929456625357e-06, "loss": 2.4311, "step": 13292 }, { "epoch": 2.53, "learning_rate": 2.2425357483317445e-06, "loss": 2.3276, "step": 13293 }, { "epoch": 2.53, "learning_rate": 2.2424785510009533e-06, "loss": 1.6182, "step": 13294 }, { "epoch": 2.53, "learning_rate": 2.242421353670162e-06, "loss": 2.527, "step": 13295 }, { "epoch": 2.53, "learning_rate": 2.242364156339371e-06, "loss": 2.2378, "step": 13296 }, { "epoch": 2.53, "learning_rate": 2.2423069590085796e-06, "loss": 3.6255, "step": 13297 }, { "epoch": 2.53, "learning_rate": 2.2422497616777884e-06, "loss": 3.0558, "step": 13298 }, { "epoch": 2.53, "learning_rate": 2.2421925643469972e-06, "loss": 3.0803, "step": 13299 }, { "epoch": 2.53, "learning_rate": 2.242135367016206e-06, "loss": 2.798, "step": 13300 }, { "epoch": 2.53, "learning_rate": 2.242078169685415e-06, "loss": 3.2436, "step": 13301 }, { "epoch": 2.53, "learning_rate": 2.2420209723546236e-06, "loss": 2.6467, "step": 13302 }, { "epoch": 2.53, "learning_rate": 2.2419637750238324e-06, "loss": 1.5867, "step": 13303 }, { "epoch": 2.53, "learning_rate": 2.241906577693041e-06, "loss": 2.4798, "step": 13304 }, { "epoch": 2.53, "learning_rate": 2.24184938036225e-06, "loss": 2.4638, "step": 13305 }, { "epoch": 2.53, "learning_rate": 2.2417921830314584e-06, "loss": 3.1481, "step": 13306 }, { "epoch": 2.53, "learning_rate": 2.2417349857006676e-06, "loss": 2.1837, "step": 13307 }, { "epoch": 2.53, "learning_rate": 2.241677788369876e-06, "loss": 3.2604, "step": 13308 }, { "epoch": 2.54, "learning_rate": 2.2416205910390848e-06, "loss": 1.9365, "step": 13309 }, { "epoch": 2.54, "learning_rate": 2.241563393708294e-06, "loss": 2.4832, "step": 13310 }, { "epoch": 2.54, "learning_rate": 2.2415061963775023e-06, "loss": 2.3409, "step": 13311 }, { "epoch": 2.54, "learning_rate": 2.241448999046711e-06, "loss": 3.0384, "step": 13312 }, { "epoch": 2.54, "learning_rate": 2.24139180171592e-06, "loss": 3.1937, "step": 13313 }, { "epoch": 2.54, "learning_rate": 2.2413346043851287e-06, "loss": 2.6293, "step": 13314 }, { "epoch": 2.54, "learning_rate": 2.2412774070543375e-06, "loss": 3.5837, "step": 13315 }, { "epoch": 2.54, "learning_rate": 2.2412202097235463e-06, "loss": 2.8624, "step": 13316 }, { "epoch": 2.54, "learning_rate": 2.241163012392755e-06, "loss": 2.7334, "step": 13317 }, { "epoch": 2.54, "learning_rate": 2.241105815061964e-06, "loss": 2.9193, "step": 13318 }, { "epoch": 2.54, "learning_rate": 2.2410486177311727e-06, "loss": 2.6032, "step": 13319 }, { "epoch": 2.54, "learning_rate": 2.2409914204003815e-06, "loss": 2.4317, "step": 13320 }, { "epoch": 2.54, "learning_rate": 2.2409342230695903e-06, "loss": 2.4108, "step": 13321 }, { "epoch": 2.54, "learning_rate": 2.240877025738799e-06, "loss": 3.6568, "step": 13322 }, { "epoch": 2.54, "learning_rate": 2.2408198284080074e-06, "loss": 2.9971, "step": 13323 }, { "epoch": 2.54, "learning_rate": 2.2407626310772167e-06, "loss": 3.4247, "step": 13324 }, { "epoch": 2.54, "learning_rate": 2.240705433746425e-06, "loss": 2.371, "step": 13325 }, { "epoch": 2.54, "learning_rate": 2.240648236415634e-06, "loss": 3.7978, "step": 13326 }, { "epoch": 2.54, "learning_rate": 2.240591039084843e-06, "loss": 2.4268, "step": 13327 }, { "epoch": 2.54, "learning_rate": 2.2405338417540514e-06, "loss": 1.8542, "step": 13328 }, { "epoch": 2.54, "learning_rate": 2.24047664442326e-06, "loss": 3.1598, "step": 13329 }, { "epoch": 2.54, "learning_rate": 2.240419447092469e-06, "loss": 3.8499, "step": 13330 }, { "epoch": 2.54, "learning_rate": 2.240362249761678e-06, "loss": 2.576, "step": 13331 }, { "epoch": 2.54, "learning_rate": 2.2403050524308866e-06, "loss": 3.5167, "step": 13332 }, { "epoch": 2.54, "learning_rate": 2.2402478551000954e-06, "loss": 2.688, "step": 13333 }, { "epoch": 2.54, "learning_rate": 2.240190657769304e-06, "loss": 3.4175, "step": 13334 }, { "epoch": 2.54, "learning_rate": 2.240133460438513e-06, "loss": 2.4433, "step": 13335 }, { "epoch": 2.54, "learning_rate": 2.2400762631077218e-06, "loss": 2.9862, "step": 13336 }, { "epoch": 2.54, "learning_rate": 2.2400190657769306e-06, "loss": 2.7579, "step": 13337 }, { "epoch": 2.54, "learning_rate": 2.2399618684461393e-06, "loss": 1.9669, "step": 13338 }, { "epoch": 2.54, "learning_rate": 2.239904671115348e-06, "loss": 2.4353, "step": 13339 }, { "epoch": 2.54, "learning_rate": 2.2398474737845565e-06, "loss": 2.9573, "step": 13340 }, { "epoch": 2.54, "learning_rate": 2.2397902764537657e-06, "loss": 3.6812, "step": 13341 }, { "epoch": 2.54, "learning_rate": 2.2397330791229745e-06, "loss": 2.374, "step": 13342 }, { "epoch": 2.54, "learning_rate": 2.239675881792183e-06, "loss": 2.218, "step": 13343 }, { "epoch": 2.54, "learning_rate": 2.239618684461392e-06, "loss": 2.3438, "step": 13344 }, { "epoch": 2.54, "learning_rate": 2.2395614871306005e-06, "loss": 2.1176, "step": 13345 }, { "epoch": 2.54, "learning_rate": 2.2395042897998093e-06, "loss": 3.765, "step": 13346 }, { "epoch": 2.54, "learning_rate": 2.239447092469018e-06, "loss": 2.3376, "step": 13347 }, { "epoch": 2.54, "learning_rate": 2.239389895138227e-06, "loss": 3.363, "step": 13348 }, { "epoch": 2.54, "learning_rate": 2.2393326978074357e-06, "loss": 2.9128, "step": 13349 }, { "epoch": 2.54, "learning_rate": 2.2392755004766445e-06, "loss": 2.9424, "step": 13350 }, { "epoch": 2.54, "learning_rate": 2.2392183031458532e-06, "loss": 3.2432, "step": 13351 }, { "epoch": 2.54, "learning_rate": 2.239161105815062e-06, "loss": 3.2043, "step": 13352 }, { "epoch": 2.54, "learning_rate": 2.239103908484271e-06, "loss": 2.3745, "step": 13353 }, { "epoch": 2.54, "learning_rate": 2.2390467111534796e-06, "loss": 3.2763, "step": 13354 }, { "epoch": 2.54, "learning_rate": 2.2389895138226884e-06, "loss": 2.808, "step": 13355 }, { "epoch": 2.54, "learning_rate": 2.2389323164918972e-06, "loss": 2.276, "step": 13356 }, { "epoch": 2.54, "learning_rate": 2.2388751191611056e-06, "loss": 2.775, "step": 13357 }, { "epoch": 2.54, "learning_rate": 2.238817921830315e-06, "loss": 3.414, "step": 13358 }, { "epoch": 2.54, "learning_rate": 2.2387607244995236e-06, "loss": 2.3402, "step": 13359 }, { "epoch": 2.54, "learning_rate": 2.238703527168732e-06, "loss": 2.8563, "step": 13360 }, { "epoch": 2.54, "learning_rate": 2.238646329837941e-06, "loss": 2.9128, "step": 13361 }, { "epoch": 2.55, "learning_rate": 2.2385891325071496e-06, "loss": 2.385, "step": 13362 }, { "epoch": 2.55, "learning_rate": 2.2385319351763583e-06, "loss": 1.9928, "step": 13363 }, { "epoch": 2.55, "learning_rate": 2.238474737845567e-06, "loss": 2.4904, "step": 13364 }, { "epoch": 2.55, "learning_rate": 2.238417540514776e-06, "loss": 2.7371, "step": 13365 }, { "epoch": 2.55, "learning_rate": 2.238360343183985e-06, "loss": 2.0992, "step": 13366 }, { "epoch": 2.55, "learning_rate": 2.2383031458531935e-06, "loss": 3.2126, "step": 13367 }, { "epoch": 2.55, "learning_rate": 2.2382459485224023e-06, "loss": 2.4924, "step": 13368 }, { "epoch": 2.55, "learning_rate": 2.238188751191611e-06, "loss": 3.3561, "step": 13369 }, { "epoch": 2.55, "learning_rate": 2.23813155386082e-06, "loss": 2.5138, "step": 13370 }, { "epoch": 2.55, "learning_rate": 2.2380743565300287e-06, "loss": 2.4465, "step": 13371 }, { "epoch": 2.55, "learning_rate": 2.2380171591992375e-06, "loss": 2.9491, "step": 13372 }, { "epoch": 2.55, "learning_rate": 2.2379599618684463e-06, "loss": 3.0458, "step": 13373 }, { "epoch": 2.55, "learning_rate": 2.2379027645376547e-06, "loss": 1.9144, "step": 13374 }, { "epoch": 2.55, "learning_rate": 2.237845567206864e-06, "loss": 2.9352, "step": 13375 }, { "epoch": 2.55, "learning_rate": 2.2377883698760727e-06, "loss": 2.4075, "step": 13376 }, { "epoch": 2.55, "learning_rate": 2.237731172545281e-06, "loss": 2.7024, "step": 13377 }, { "epoch": 2.55, "learning_rate": 2.2376739752144903e-06, "loss": 3.023, "step": 13378 }, { "epoch": 2.55, "learning_rate": 2.2376167778836986e-06, "loss": 1.9472, "step": 13379 }, { "epoch": 2.55, "learning_rate": 2.2375595805529074e-06, "loss": 2.6854, "step": 13380 }, { "epoch": 2.55, "learning_rate": 2.2375023832221162e-06, "loss": 2.5014, "step": 13381 }, { "epoch": 2.55, "learning_rate": 2.237445185891325e-06, "loss": 2.5, "step": 13382 }, { "epoch": 2.55, "learning_rate": 2.2373879885605342e-06, "loss": 1.9414, "step": 13383 }, { "epoch": 2.55, "learning_rate": 2.2373307912297426e-06, "loss": 2.9009, "step": 13384 }, { "epoch": 2.55, "learning_rate": 2.2372735938989514e-06, "loss": 3.2647, "step": 13385 }, { "epoch": 2.55, "learning_rate": 2.23721639656816e-06, "loss": 3.0865, "step": 13386 }, { "epoch": 2.55, "learning_rate": 2.237159199237369e-06, "loss": 3.1387, "step": 13387 }, { "epoch": 2.55, "learning_rate": 2.2371020019065778e-06, "loss": 3.505, "step": 13388 }, { "epoch": 2.55, "learning_rate": 2.2370448045757866e-06, "loss": 3.1966, "step": 13389 }, { "epoch": 2.55, "learning_rate": 2.2369876072449954e-06, "loss": 2.32, "step": 13390 }, { "epoch": 2.55, "learning_rate": 2.236930409914204e-06, "loss": 2.7412, "step": 13391 }, { "epoch": 2.55, "learning_rate": 2.236873212583413e-06, "loss": 2.3671, "step": 13392 }, { "epoch": 2.55, "learning_rate": 2.2368160152526217e-06, "loss": 2.4393, "step": 13393 }, { "epoch": 2.55, "learning_rate": 2.23675881792183e-06, "loss": 2.9646, "step": 13394 }, { "epoch": 2.55, "learning_rate": 2.2367016205910393e-06, "loss": 2.2418, "step": 13395 }, { "epoch": 2.55, "learning_rate": 2.2366444232602477e-06, "loss": 2.5581, "step": 13396 }, { "epoch": 2.55, "learning_rate": 2.2365872259294565e-06, "loss": 3.7745, "step": 13397 }, { "epoch": 2.55, "learning_rate": 2.2365300285986657e-06, "loss": 2.4883, "step": 13398 }, { "epoch": 2.55, "learning_rate": 2.236472831267874e-06, "loss": 2.4836, "step": 13399 }, { "epoch": 2.55, "learning_rate": 2.2364156339370833e-06, "loss": 2.4069, "step": 13400 }, { "epoch": 2.55, "learning_rate": 2.2363584366062917e-06, "loss": 3.6459, "step": 13401 }, { "epoch": 2.55, "learning_rate": 2.2363012392755005e-06, "loss": 2.9297, "step": 13402 }, { "epoch": 2.55, "learning_rate": 2.2362440419447093e-06, "loss": 3.7132, "step": 13403 }, { "epoch": 2.55, "learning_rate": 2.236186844613918e-06, "loss": 2.2879, "step": 13404 }, { "epoch": 2.55, "learning_rate": 2.236129647283127e-06, "loss": 1.7912, "step": 13405 }, { "epoch": 2.55, "learning_rate": 2.2360724499523356e-06, "loss": 2.4209, "step": 13406 }, { "epoch": 2.55, "learning_rate": 2.2360152526215444e-06, "loss": 2.2651, "step": 13407 }, { "epoch": 2.55, "learning_rate": 2.2359580552907532e-06, "loss": 3.8274, "step": 13408 }, { "epoch": 2.55, "learning_rate": 2.235900857959962e-06, "loss": 2.4356, "step": 13409 }, { "epoch": 2.55, "learning_rate": 2.235843660629171e-06, "loss": 2.3491, "step": 13410 }, { "epoch": 2.55, "learning_rate": 2.235786463298379e-06, "loss": 2.2283, "step": 13411 }, { "epoch": 2.55, "learning_rate": 2.2357292659675884e-06, "loss": 3.4536, "step": 13412 }, { "epoch": 2.55, "learning_rate": 2.2356720686367968e-06, "loss": 2.8942, "step": 13413 }, { "epoch": 2.56, "learning_rate": 2.2356148713060056e-06, "loss": 2.4287, "step": 13414 }, { "epoch": 2.56, "learning_rate": 2.2355576739752148e-06, "loss": 3.34, "step": 13415 }, { "epoch": 2.56, "learning_rate": 2.235500476644423e-06, "loss": 2.3483, "step": 13416 }, { "epoch": 2.56, "learning_rate": 2.2354432793136324e-06, "loss": 3.2618, "step": 13417 }, { "epoch": 2.56, "learning_rate": 2.2353860819828407e-06, "loss": 2.956, "step": 13418 }, { "epoch": 2.56, "learning_rate": 2.2353288846520495e-06, "loss": 2.6726, "step": 13419 }, { "epoch": 2.56, "learning_rate": 2.2352716873212583e-06, "loss": 2.3883, "step": 13420 }, { "epoch": 2.56, "learning_rate": 2.235214489990467e-06, "loss": 2.2306, "step": 13421 }, { "epoch": 2.56, "learning_rate": 2.235157292659676e-06, "loss": 3.0789, "step": 13422 }, { "epoch": 2.56, "learning_rate": 2.2351000953288847e-06, "loss": 2.4573, "step": 13423 }, { "epoch": 2.56, "learning_rate": 2.2350428979980935e-06, "loss": 3.2356, "step": 13424 }, { "epoch": 2.56, "learning_rate": 2.2349857006673023e-06, "loss": 1.5091, "step": 13425 }, { "epoch": 2.56, "learning_rate": 2.234928503336511e-06, "loss": 2.0766, "step": 13426 }, { "epoch": 2.56, "learning_rate": 2.23487130600572e-06, "loss": 1.8281, "step": 13427 }, { "epoch": 2.56, "learning_rate": 2.2348141086749283e-06, "loss": 2.4834, "step": 13428 }, { "epoch": 2.56, "learning_rate": 2.2347569113441375e-06, "loss": 2.8842, "step": 13429 }, { "epoch": 2.56, "learning_rate": 2.2346997140133463e-06, "loss": 2.2112, "step": 13430 }, { "epoch": 2.56, "learning_rate": 2.234642516682555e-06, "loss": 2.9007, "step": 13431 }, { "epoch": 2.56, "learning_rate": 2.234585319351764e-06, "loss": 2.4104, "step": 13432 }, { "epoch": 2.56, "learning_rate": 2.2345281220209722e-06, "loss": 2.6734, "step": 13433 }, { "epoch": 2.56, "learning_rate": 2.2344709246901814e-06, "loss": 2.6342, "step": 13434 }, { "epoch": 2.56, "learning_rate": 2.23441372735939e-06, "loss": 2.4508, "step": 13435 }, { "epoch": 2.56, "learning_rate": 2.2343565300285986e-06, "loss": 2.3061, "step": 13436 }, { "epoch": 2.56, "learning_rate": 2.234299332697808e-06, "loss": 2.2167, "step": 13437 }, { "epoch": 2.56, "learning_rate": 2.234242135367016e-06, "loss": 2.2102, "step": 13438 }, { "epoch": 2.56, "learning_rate": 2.234184938036225e-06, "loss": 3.4513, "step": 13439 }, { "epoch": 2.56, "learning_rate": 2.2341277407054338e-06, "loss": 3.1668, "step": 13440 }, { "epoch": 2.56, "learning_rate": 2.2340705433746426e-06, "loss": 2.1468, "step": 13441 }, { "epoch": 2.56, "learning_rate": 2.2340133460438514e-06, "loss": 2.3627, "step": 13442 }, { "epoch": 2.56, "learning_rate": 2.23395614871306e-06, "loss": 3.4798, "step": 13443 }, { "epoch": 2.56, "learning_rate": 2.233898951382269e-06, "loss": 3.0669, "step": 13444 }, { "epoch": 2.56, "learning_rate": 2.2338417540514773e-06, "loss": 3.2632, "step": 13445 }, { "epoch": 2.56, "learning_rate": 2.2337845567206865e-06, "loss": 1.8549, "step": 13446 }, { "epoch": 2.56, "learning_rate": 2.2337273593898953e-06, "loss": 3.1353, "step": 13447 }, { "epoch": 2.56, "learning_rate": 2.233670162059104e-06, "loss": 3.6268, "step": 13448 }, { "epoch": 2.56, "learning_rate": 2.233612964728313e-06, "loss": 1.819, "step": 13449 }, { "epoch": 2.56, "learning_rate": 2.2335557673975213e-06, "loss": 2.3558, "step": 13450 }, { "epoch": 2.56, "learning_rate": 2.2334985700667305e-06, "loss": 2.541, "step": 13451 }, { "epoch": 2.56, "learning_rate": 2.233441372735939e-06, "loss": 2.2672, "step": 13452 }, { "epoch": 2.56, "learning_rate": 2.2333841754051477e-06, "loss": 2.8646, "step": 13453 }, { "epoch": 2.56, "learning_rate": 2.233326978074357e-06, "loss": 2.1767, "step": 13454 }, { "epoch": 2.56, "learning_rate": 2.2332697807435653e-06, "loss": 2.2126, "step": 13455 }, { "epoch": 2.56, "learning_rate": 2.233212583412774e-06, "loss": 3.2945, "step": 13456 }, { "epoch": 2.56, "learning_rate": 2.233155386081983e-06, "loss": 3.055, "step": 13457 }, { "epoch": 2.56, "learning_rate": 2.2330981887511916e-06, "loss": 3.0264, "step": 13458 }, { "epoch": 2.56, "learning_rate": 2.2330409914204004e-06, "loss": 3.1558, "step": 13459 }, { "epoch": 2.56, "learning_rate": 2.2329837940896092e-06, "loss": 2.4237, "step": 13460 }, { "epoch": 2.56, "learning_rate": 2.232926596758818e-06, "loss": 2.195, "step": 13461 }, { "epoch": 2.56, "learning_rate": 2.2328693994280264e-06, "loss": 2.9599, "step": 13462 }, { "epoch": 2.56, "learning_rate": 2.2328122020972356e-06, "loss": 3.6592, "step": 13463 }, { "epoch": 2.56, "learning_rate": 2.2327550047664444e-06, "loss": 2.3116, "step": 13464 }, { "epoch": 2.56, "learning_rate": 2.232697807435653e-06, "loss": 2.9958, "step": 13465 }, { "epoch": 2.56, "learning_rate": 2.232640610104862e-06, "loss": 2.7734, "step": 13466 }, { "epoch": 2.57, "learning_rate": 2.2325834127740704e-06, "loss": 3.7479, "step": 13467 }, { "epoch": 2.57, "learning_rate": 2.2325262154432796e-06, "loss": 3.2917, "step": 13468 }, { "epoch": 2.57, "learning_rate": 2.232469018112488e-06, "loss": 2.5942, "step": 13469 }, { "epoch": 2.57, "learning_rate": 2.2324118207816967e-06, "loss": 2.5043, "step": 13470 }, { "epoch": 2.57, "learning_rate": 2.232354623450906e-06, "loss": 3.5388, "step": 13471 }, { "epoch": 2.57, "learning_rate": 2.2322974261201143e-06, "loss": 2.9067, "step": 13472 }, { "epoch": 2.57, "learning_rate": 2.232240228789323e-06, "loss": 2.3962, "step": 13473 }, { "epoch": 2.57, "learning_rate": 2.232183031458532e-06, "loss": 2.8215, "step": 13474 }, { "epoch": 2.57, "learning_rate": 2.2321258341277407e-06, "loss": 2.4448, "step": 13475 }, { "epoch": 2.57, "learning_rate": 2.2320686367969495e-06, "loss": 2.7046, "step": 13476 }, { "epoch": 2.57, "learning_rate": 2.2320114394661583e-06, "loss": 2.8363, "step": 13477 }, { "epoch": 2.57, "learning_rate": 2.231954242135367e-06, "loss": 2.879, "step": 13478 }, { "epoch": 2.57, "learning_rate": 2.231897044804576e-06, "loss": 3.1772, "step": 13479 }, { "epoch": 2.57, "learning_rate": 2.2318398474737847e-06, "loss": 1.9641, "step": 13480 }, { "epoch": 2.57, "learning_rate": 2.2317826501429935e-06, "loss": 2.3042, "step": 13481 }, { "epoch": 2.57, "learning_rate": 2.2317254528122023e-06, "loss": 2.0191, "step": 13482 }, { "epoch": 2.57, "learning_rate": 2.231668255481411e-06, "loss": 3.1991, "step": 13483 }, { "epoch": 2.57, "learning_rate": 2.2316110581506194e-06, "loss": 3.6347, "step": 13484 }, { "epoch": 2.57, "learning_rate": 2.2315538608198287e-06, "loss": 3.4476, "step": 13485 }, { "epoch": 2.57, "learning_rate": 2.2314966634890374e-06, "loss": 2.9321, "step": 13486 }, { "epoch": 2.57, "learning_rate": 2.231439466158246e-06, "loss": 2.9288, "step": 13487 }, { "epoch": 2.57, "learning_rate": 2.231382268827455e-06, "loss": 2.5648, "step": 13488 }, { "epoch": 2.57, "learning_rate": 2.2313250714966634e-06, "loss": 2.1854, "step": 13489 }, { "epoch": 2.57, "learning_rate": 2.231267874165872e-06, "loss": 1.836, "step": 13490 }, { "epoch": 2.57, "learning_rate": 2.231210676835081e-06, "loss": 3.5964, "step": 13491 }, { "epoch": 2.57, "learning_rate": 2.2311534795042898e-06, "loss": 2.3513, "step": 13492 }, { "epoch": 2.57, "learning_rate": 2.2310962821734986e-06, "loss": 3.5815, "step": 13493 }, { "epoch": 2.57, "learning_rate": 2.2310390848427074e-06, "loss": 2.5702, "step": 13494 }, { "epoch": 2.57, "learning_rate": 2.230981887511916e-06, "loss": 2.8533, "step": 13495 }, { "epoch": 2.57, "learning_rate": 2.230924690181125e-06, "loss": 2.6594, "step": 13496 }, { "epoch": 2.57, "learning_rate": 2.2308674928503338e-06, "loss": 3.2905, "step": 13497 }, { "epoch": 2.57, "learning_rate": 2.2308102955195425e-06, "loss": 2.3233, "step": 13498 }, { "epoch": 2.57, "learning_rate": 2.2307530981887513e-06, "loss": 3.0069, "step": 13499 }, { "epoch": 2.57, "learning_rate": 2.23069590085796e-06, "loss": 2.6759, "step": 13500 }, { "epoch": 2.57, "learning_rate": 2.2306387035271685e-06, "loss": 3.0969, "step": 13501 }, { "epoch": 2.57, "learning_rate": 2.2305815061963777e-06, "loss": 2.5905, "step": 13502 }, { "epoch": 2.57, "learning_rate": 2.2305243088655865e-06, "loss": 2.8153, "step": 13503 }, { "epoch": 2.57, "learning_rate": 2.230467111534795e-06, "loss": 2.3352, "step": 13504 }, { "epoch": 2.57, "learning_rate": 2.230409914204004e-06, "loss": 2.5458, "step": 13505 }, { "epoch": 2.57, "learning_rate": 2.2303527168732125e-06, "loss": 2.6527, "step": 13506 }, { "epoch": 2.57, "learning_rate": 2.2302955195424213e-06, "loss": 2.6398, "step": 13507 }, { "epoch": 2.57, "learning_rate": 2.23023832221163e-06, "loss": 1.711, "step": 13508 }, { "epoch": 2.57, "learning_rate": 2.230181124880839e-06, "loss": 2.3976, "step": 13509 }, { "epoch": 2.57, "learning_rate": 2.2301239275500477e-06, "loss": 2.2637, "step": 13510 }, { "epoch": 2.57, "learning_rate": 2.2300667302192564e-06, "loss": 3.3952, "step": 13511 }, { "epoch": 2.57, "learning_rate": 2.2300095328884652e-06, "loss": 2.5995, "step": 13512 }, { "epoch": 2.57, "learning_rate": 2.229952335557674e-06, "loss": 3.2916, "step": 13513 }, { "epoch": 2.57, "learning_rate": 2.229895138226883e-06, "loss": 2.9723, "step": 13514 }, { "epoch": 2.57, "learning_rate": 2.2298379408960916e-06, "loss": 2.0738, "step": 13515 }, { "epoch": 2.57, "learning_rate": 2.2297807435653004e-06, "loss": 3.5324, "step": 13516 }, { "epoch": 2.57, "learning_rate": 2.229723546234509e-06, "loss": 3.2674, "step": 13517 }, { "epoch": 2.57, "learning_rate": 2.2296663489037176e-06, "loss": 2.676, "step": 13518 }, { "epoch": 2.58, "learning_rate": 2.229609151572927e-06, "loss": 2.8172, "step": 13519 }, { "epoch": 2.58, "learning_rate": 2.2295519542421356e-06, "loss": 3.4004, "step": 13520 }, { "epoch": 2.58, "learning_rate": 2.229494756911344e-06, "loss": 2.8322, "step": 13521 }, { "epoch": 2.58, "learning_rate": 2.229437559580553e-06, "loss": 2.0733, "step": 13522 }, { "epoch": 2.58, "learning_rate": 2.2293803622497615e-06, "loss": 1.8069, "step": 13523 }, { "epoch": 2.58, "learning_rate": 2.2293231649189703e-06, "loss": 2.0605, "step": 13524 }, { "epoch": 2.58, "learning_rate": 2.2292659675881796e-06, "loss": 2.6574, "step": 13525 }, { "epoch": 2.58, "learning_rate": 2.229208770257388e-06, "loss": 2.7078, "step": 13526 }, { "epoch": 2.58, "learning_rate": 2.2291515729265967e-06, "loss": 3.3924, "step": 13527 }, { "epoch": 2.58, "learning_rate": 2.2290943755958055e-06, "loss": 3.0032, "step": 13528 }, { "epoch": 2.58, "learning_rate": 2.2290371782650143e-06, "loss": 4.0268, "step": 13529 }, { "epoch": 2.58, "learning_rate": 2.228979980934223e-06, "loss": 2.1075, "step": 13530 }, { "epoch": 2.58, "learning_rate": 2.228922783603432e-06, "loss": 3.4587, "step": 13531 }, { "epoch": 2.58, "learning_rate": 2.2288655862726407e-06, "loss": 3.2803, "step": 13532 }, { "epoch": 2.58, "learning_rate": 2.2288083889418495e-06, "loss": 2.4275, "step": 13533 }, { "epoch": 2.58, "learning_rate": 2.2287511916110583e-06, "loss": 1.8443, "step": 13534 }, { "epoch": 2.58, "learning_rate": 2.228693994280267e-06, "loss": 2.8066, "step": 13535 }, { "epoch": 2.58, "learning_rate": 2.228636796949476e-06, "loss": 2.8382, "step": 13536 }, { "epoch": 2.58, "learning_rate": 2.2285795996186847e-06, "loss": 2.7179, "step": 13537 }, { "epoch": 2.58, "learning_rate": 2.228522402287893e-06, "loss": 3.1237, "step": 13538 }, { "epoch": 2.58, "learning_rate": 2.2284652049571022e-06, "loss": 2.9762, "step": 13539 }, { "epoch": 2.58, "learning_rate": 2.2284080076263106e-06, "loss": 2.2479, "step": 13540 }, { "epoch": 2.58, "learning_rate": 2.2283508102955194e-06, "loss": 1.9062, "step": 13541 }, { "epoch": 2.58, "learning_rate": 2.2282936129647286e-06, "loss": 2.5998, "step": 13542 }, { "epoch": 2.58, "learning_rate": 2.228236415633937e-06, "loss": 2.0497, "step": 13543 }, { "epoch": 2.58, "learning_rate": 2.228179218303146e-06, "loss": 3.311, "step": 13544 }, { "epoch": 2.58, "learning_rate": 2.2281220209723546e-06, "loss": 3.6383, "step": 13545 }, { "epoch": 2.58, "learning_rate": 2.2280648236415634e-06, "loss": 1.7646, "step": 13546 }, { "epoch": 2.58, "learning_rate": 2.228007626310772e-06, "loss": 2.8236, "step": 13547 }, { "epoch": 2.58, "learning_rate": 2.227950428979981e-06, "loss": 2.2569, "step": 13548 }, { "epoch": 2.58, "learning_rate": 2.2278932316491898e-06, "loss": 3.1875, "step": 13549 }, { "epoch": 2.58, "learning_rate": 2.2278360343183986e-06, "loss": 2.7617, "step": 13550 }, { "epoch": 2.58, "learning_rate": 2.2277788369876073e-06, "loss": 2.0703, "step": 13551 }, { "epoch": 2.58, "learning_rate": 2.227721639656816e-06, "loss": 3.225, "step": 13552 }, { "epoch": 2.58, "learning_rate": 2.227664442326025e-06, "loss": 3.1332, "step": 13553 }, { "epoch": 2.58, "learning_rate": 2.2276072449952337e-06, "loss": 3.1934, "step": 13554 }, { "epoch": 2.58, "learning_rate": 2.227550047664442e-06, "loss": 2.2896, "step": 13555 }, { "epoch": 2.58, "learning_rate": 2.2274928503336513e-06, "loss": 3.0779, "step": 13556 }, { "epoch": 2.58, "learning_rate": 2.2274356530028597e-06, "loss": 2.482, "step": 13557 }, { "epoch": 2.58, "learning_rate": 2.2273784556720685e-06, "loss": 3.6646, "step": 13558 }, { "epoch": 2.58, "learning_rate": 2.2273212583412777e-06, "loss": 1.8475, "step": 13559 }, { "epoch": 2.58, "learning_rate": 2.227264061010486e-06, "loss": 2.3282, "step": 13560 }, { "epoch": 2.58, "learning_rate": 2.227206863679695e-06, "loss": 2.0038, "step": 13561 }, { "epoch": 2.58, "learning_rate": 2.2271496663489037e-06, "loss": 3.3097, "step": 13562 }, { "epoch": 2.58, "learning_rate": 2.2270924690181125e-06, "loss": 2.6603, "step": 13563 }, { "epoch": 2.58, "learning_rate": 2.2270352716873217e-06, "loss": 2.6903, "step": 13564 }, { "epoch": 2.58, "learning_rate": 2.22697807435653e-06, "loss": 2.9726, "step": 13565 }, { "epoch": 2.58, "learning_rate": 2.226920877025739e-06, "loss": 3.3157, "step": 13566 }, { "epoch": 2.58, "learning_rate": 2.2268636796949476e-06, "loss": 2.6212, "step": 13567 }, { "epoch": 2.58, "learning_rate": 2.2268064823641564e-06, "loss": 2.1213, "step": 13568 }, { "epoch": 2.58, "learning_rate": 2.2267492850333652e-06, "loss": 2.09, "step": 13569 }, { "epoch": 2.58, "learning_rate": 2.226692087702574e-06, "loss": 2.6872, "step": 13570 }, { "epoch": 2.58, "learning_rate": 2.226634890371783e-06, "loss": 2.6567, "step": 13571 }, { "epoch": 2.59, "learning_rate": 2.226577693040991e-06, "loss": 3.4421, "step": 13572 }, { "epoch": 2.59, "learning_rate": 2.2265204957102004e-06, "loss": 4.029, "step": 13573 }, { "epoch": 2.59, "learning_rate": 2.226463298379409e-06, "loss": 2.5859, "step": 13574 }, { "epoch": 2.59, "learning_rate": 2.2264061010486176e-06, "loss": 2.6988, "step": 13575 }, { "epoch": 2.59, "learning_rate": 2.2263489037178268e-06, "loss": 3.2026, "step": 13576 }, { "epoch": 2.59, "learning_rate": 2.226291706387035e-06, "loss": 2.2608, "step": 13577 }, { "epoch": 2.59, "learning_rate": 2.2262345090562444e-06, "loss": 2.6935, "step": 13578 }, { "epoch": 2.59, "learning_rate": 2.2261773117254527e-06, "loss": 2.1609, "step": 13579 }, { "epoch": 2.59, "learning_rate": 2.2261201143946615e-06, "loss": 3.5007, "step": 13580 }, { "epoch": 2.59, "learning_rate": 2.2260629170638707e-06, "loss": 3.1667, "step": 13581 }, { "epoch": 2.59, "learning_rate": 2.226005719733079e-06, "loss": 3.344, "step": 13582 }, { "epoch": 2.59, "learning_rate": 2.225948522402288e-06, "loss": 3.4036, "step": 13583 }, { "epoch": 2.59, "learning_rate": 2.2258913250714967e-06, "loss": 2.1544, "step": 13584 }, { "epoch": 2.59, "learning_rate": 2.2258341277407055e-06, "loss": 2.6949, "step": 13585 }, { "epoch": 2.59, "learning_rate": 2.2257769304099143e-06, "loss": 2.3345, "step": 13586 }, { "epoch": 2.59, "learning_rate": 2.225719733079123e-06, "loss": 2.2213, "step": 13587 }, { "epoch": 2.59, "learning_rate": 2.225662535748332e-06, "loss": 3.0183, "step": 13588 }, { "epoch": 2.59, "learning_rate": 2.2256053384175402e-06, "loss": 2.657, "step": 13589 }, { "epoch": 2.59, "learning_rate": 2.2255481410867495e-06, "loss": 2.5005, "step": 13590 }, { "epoch": 2.59, "learning_rate": 2.2254909437559583e-06, "loss": 2.6563, "step": 13591 }, { "epoch": 2.59, "learning_rate": 2.2254337464251666e-06, "loss": 2.315, "step": 13592 }, { "epoch": 2.59, "learning_rate": 2.225376549094376e-06, "loss": 3.7719, "step": 13593 }, { "epoch": 2.59, "learning_rate": 2.2253193517635842e-06, "loss": 3.3195, "step": 13594 }, { "epoch": 2.59, "learning_rate": 2.2252621544327934e-06, "loss": 2.5619, "step": 13595 }, { "epoch": 2.59, "learning_rate": 2.225204957102002e-06, "loss": 2.9563, "step": 13596 }, { "epoch": 2.59, "learning_rate": 2.2251477597712106e-06, "loss": 2.664, "step": 13597 }, { "epoch": 2.59, "learning_rate": 2.22509056244042e-06, "loss": 3.3413, "step": 13598 }, { "epoch": 2.59, "learning_rate": 2.225033365109628e-06, "loss": 3.0718, "step": 13599 }, { "epoch": 2.59, "learning_rate": 2.224976167778837e-06, "loss": 2.8666, "step": 13600 }, { "epoch": 2.59, "learning_rate": 2.2249189704480458e-06, "loss": 2.8904, "step": 13601 }, { "epoch": 2.59, "learning_rate": 2.2248617731172546e-06, "loss": 2.8027, "step": 13602 }, { "epoch": 2.59, "learning_rate": 2.2248045757864634e-06, "loss": 3.0196, "step": 13603 }, { "epoch": 2.59, "learning_rate": 2.224747378455672e-06, "loss": 3.416, "step": 13604 }, { "epoch": 2.59, "learning_rate": 2.224690181124881e-06, "loss": 2.8991, "step": 13605 }, { "epoch": 2.59, "learning_rate": 2.2246329837940893e-06, "loss": 3.565, "step": 13606 }, { "epoch": 2.59, "learning_rate": 2.2245757864632985e-06, "loss": 2.8518, "step": 13607 }, { "epoch": 2.59, "learning_rate": 2.2245185891325073e-06, "loss": 2.9655, "step": 13608 }, { "epoch": 2.59, "learning_rate": 2.2244613918017157e-06, "loss": 2.7532, "step": 13609 }, { "epoch": 2.59, "learning_rate": 2.224404194470925e-06, "loss": 1.5084, "step": 13610 }, { "epoch": 2.59, "learning_rate": 2.2243469971401333e-06, "loss": 3.5469, "step": 13611 }, { "epoch": 2.59, "learning_rate": 2.2242897998093425e-06, "loss": 3.5216, "step": 13612 }, { "epoch": 2.59, "learning_rate": 2.2242326024785513e-06, "loss": 2.2134, "step": 13613 }, { "epoch": 2.59, "learning_rate": 2.2241754051477597e-06, "loss": 2.2371, "step": 13614 }, { "epoch": 2.59, "learning_rate": 2.224118207816969e-06, "loss": 1.7996, "step": 13615 }, { "epoch": 2.59, "learning_rate": 2.2240610104861773e-06, "loss": 2.2273, "step": 13616 }, { "epoch": 2.59, "learning_rate": 2.224003813155386e-06, "loss": 3.1006, "step": 13617 }, { "epoch": 2.59, "learning_rate": 2.223946615824595e-06, "loss": 1.9727, "step": 13618 }, { "epoch": 2.59, "learning_rate": 2.2238894184938036e-06, "loss": 2.5111, "step": 13619 }, { "epoch": 2.59, "learning_rate": 2.2238322211630124e-06, "loss": 2.3853, "step": 13620 }, { "epoch": 2.59, "learning_rate": 2.2237750238322212e-06, "loss": 2.85, "step": 13621 }, { "epoch": 2.59, "learning_rate": 2.22371782650143e-06, "loss": 2.3561, "step": 13622 }, { "epoch": 2.59, "learning_rate": 2.223660629170639e-06, "loss": 1.9948, "step": 13623 }, { "epoch": 2.6, "learning_rate": 2.2236034318398476e-06, "loss": 3.3248, "step": 13624 }, { "epoch": 2.6, "learning_rate": 2.2235462345090564e-06, "loss": 1.8831, "step": 13625 }, { "epoch": 2.6, "learning_rate": 2.2234890371782648e-06, "loss": 2.6463, "step": 13626 }, { "epoch": 2.6, "learning_rate": 2.223431839847474e-06, "loss": 2.5523, "step": 13627 }, { "epoch": 2.6, "learning_rate": 2.2233746425166824e-06, "loss": 3.3033, "step": 13628 }, { "epoch": 2.6, "learning_rate": 2.2233174451858916e-06, "loss": 2.3875, "step": 13629 }, { "epoch": 2.6, "learning_rate": 2.2232602478551004e-06, "loss": 3.1029, "step": 13630 }, { "epoch": 2.6, "learning_rate": 2.2232030505243087e-06, "loss": 2.1107, "step": 13631 }, { "epoch": 2.6, "learning_rate": 2.223145853193518e-06, "loss": 2.9818, "step": 13632 }, { "epoch": 2.6, "learning_rate": 2.2230886558627263e-06, "loss": 2.2424, "step": 13633 }, { "epoch": 2.6, "learning_rate": 2.223031458531935e-06, "loss": 2.7857, "step": 13634 }, { "epoch": 2.6, "learning_rate": 2.222974261201144e-06, "loss": 2.8503, "step": 13635 }, { "epoch": 2.6, "learning_rate": 2.2229170638703527e-06, "loss": 2.956, "step": 13636 }, { "epoch": 2.6, "learning_rate": 2.2228598665395615e-06, "loss": 2.6861, "step": 13637 }, { "epoch": 2.6, "learning_rate": 2.2228026692087703e-06, "loss": 1.7451, "step": 13638 }, { "epoch": 2.6, "learning_rate": 2.222745471877979e-06, "loss": 2.4751, "step": 13639 }, { "epoch": 2.6, "learning_rate": 2.222688274547188e-06, "loss": 2.836, "step": 13640 }, { "epoch": 2.6, "learning_rate": 2.2226310772163967e-06, "loss": 3.3399, "step": 13641 }, { "epoch": 2.6, "learning_rate": 2.2225738798856055e-06, "loss": 2.3589, "step": 13642 }, { "epoch": 2.6, "learning_rate": 2.2225166825548143e-06, "loss": 2.4766, "step": 13643 }, { "epoch": 2.6, "learning_rate": 2.222459485224023e-06, "loss": 2.1227, "step": 13644 }, { "epoch": 2.6, "learning_rate": 2.2224022878932314e-06, "loss": 3.1545, "step": 13645 }, { "epoch": 2.6, "learning_rate": 2.2223450905624406e-06, "loss": 2.5273, "step": 13646 }, { "epoch": 2.6, "learning_rate": 2.2222878932316494e-06, "loss": 2.6892, "step": 13647 }, { "epoch": 2.6, "learning_rate": 2.222230695900858e-06, "loss": 2.4219, "step": 13648 }, { "epoch": 2.6, "learning_rate": 2.222173498570067e-06, "loss": 3.21, "step": 13649 }, { "epoch": 2.6, "learning_rate": 2.2221163012392754e-06, "loss": 3.8331, "step": 13650 }, { "epoch": 2.6, "learning_rate": 2.222059103908484e-06, "loss": 2.9388, "step": 13651 }, { "epoch": 2.6, "learning_rate": 2.2220019065776934e-06, "loss": 2.8466, "step": 13652 }, { "epoch": 2.6, "learning_rate": 2.2219447092469018e-06, "loss": 3.3756, "step": 13653 }, { "epoch": 2.6, "learning_rate": 2.2218875119161106e-06, "loss": 3.3863, "step": 13654 }, { "epoch": 2.6, "learning_rate": 2.2218303145853194e-06, "loss": 1.8791, "step": 13655 }, { "epoch": 2.6, "learning_rate": 2.221773117254528e-06, "loss": 2.7002, "step": 13656 }, { "epoch": 2.6, "learning_rate": 2.221715919923737e-06, "loss": 1.6535, "step": 13657 }, { "epoch": 2.6, "learning_rate": 2.2216587225929457e-06, "loss": 2.4684, "step": 13658 }, { "epoch": 2.6, "learning_rate": 2.2216015252621545e-06, "loss": 3.1155, "step": 13659 }, { "epoch": 2.6, "learning_rate": 2.2215443279313633e-06, "loss": 2.1522, "step": 13660 }, { "epoch": 2.6, "learning_rate": 2.221487130600572e-06, "loss": 2.4195, "step": 13661 }, { "epoch": 2.6, "learning_rate": 2.221429933269781e-06, "loss": 3.2942, "step": 13662 }, { "epoch": 2.6, "learning_rate": 2.2213727359389897e-06, "loss": 2.379, "step": 13663 }, { "epoch": 2.6, "learning_rate": 2.2213155386081985e-06, "loss": 2.863, "step": 13664 }, { "epoch": 2.6, "learning_rate": 2.221258341277407e-06, "loss": 2.8586, "step": 13665 }, { "epoch": 2.6, "learning_rate": 2.221201143946616e-06, "loss": 3.4343, "step": 13666 }, { "epoch": 2.6, "learning_rate": 2.2211439466158245e-06, "loss": 3.5975, "step": 13667 }, { "epoch": 2.6, "learning_rate": 2.2210867492850333e-06, "loss": 2.1389, "step": 13668 }, { "epoch": 2.6, "learning_rate": 2.2210295519542425e-06, "loss": 1.8793, "step": 13669 }, { "epoch": 2.6, "learning_rate": 2.220972354623451e-06, "loss": 2.4018, "step": 13670 }, { "epoch": 2.6, "learning_rate": 2.2209151572926596e-06, "loss": 2.3947, "step": 13671 }, { "epoch": 2.6, "learning_rate": 2.2208579599618684e-06, "loss": 3.0779, "step": 13672 }, { "epoch": 2.6, "learning_rate": 2.2208007626310772e-06, "loss": 3.1829, "step": 13673 }, { "epoch": 2.6, "learning_rate": 2.220743565300286e-06, "loss": 2.4032, "step": 13674 }, { "epoch": 2.6, "learning_rate": 2.220686367969495e-06, "loss": 2.1148, "step": 13675 }, { "epoch": 2.6, "learning_rate": 2.2206291706387036e-06, "loss": 2.528, "step": 13676 }, { "epoch": 2.61, "learning_rate": 2.2205719733079124e-06, "loss": 2.6689, "step": 13677 }, { "epoch": 2.61, "learning_rate": 2.220514775977121e-06, "loss": 2.1135, "step": 13678 }, { "epoch": 2.61, "learning_rate": 2.22045757864633e-06, "loss": 2.0686, "step": 13679 }, { "epoch": 2.61, "learning_rate": 2.2204003813155388e-06, "loss": 2.4923, "step": 13680 }, { "epoch": 2.61, "learning_rate": 2.2203431839847476e-06, "loss": 2.8111, "step": 13681 }, { "epoch": 2.61, "learning_rate": 2.220285986653956e-06, "loss": 1.4626, "step": 13682 }, { "epoch": 2.61, "learning_rate": 2.220228789323165e-06, "loss": 2.1631, "step": 13683 }, { "epoch": 2.61, "learning_rate": 2.2201715919923735e-06, "loss": 2.5607, "step": 13684 }, { "epoch": 2.61, "learning_rate": 2.2201143946615823e-06, "loss": 3.4528, "step": 13685 }, { "epoch": 2.61, "learning_rate": 2.2200571973307916e-06, "loss": 3.022, "step": 13686 }, { "epoch": 2.61, "learning_rate": 2.22e-06, "loss": 3.2883, "step": 13687 }, { "epoch": 2.61, "learning_rate": 2.2199428026692087e-06, "loss": 2.3957, "step": 13688 }, { "epoch": 2.61, "learning_rate": 2.2198856053384175e-06, "loss": 2.2822, "step": 13689 }, { "epoch": 2.61, "learning_rate": 2.2198284080076263e-06, "loss": 2.8625, "step": 13690 }, { "epoch": 2.61, "learning_rate": 2.219771210676835e-06, "loss": 2.6919, "step": 13691 }, { "epoch": 2.61, "learning_rate": 2.219714013346044e-06, "loss": 2.9908, "step": 13692 }, { "epoch": 2.61, "learning_rate": 2.2196568160152527e-06, "loss": 3.0352, "step": 13693 }, { "epoch": 2.61, "learning_rate": 2.2195996186844615e-06, "loss": 3.5201, "step": 13694 }, { "epoch": 2.61, "learning_rate": 2.2195424213536703e-06, "loss": 2.2696, "step": 13695 }, { "epoch": 2.61, "learning_rate": 2.219485224022879e-06, "loss": 2.0463, "step": 13696 }, { "epoch": 2.61, "learning_rate": 2.219428026692088e-06, "loss": 3.0919, "step": 13697 }, { "epoch": 2.61, "learning_rate": 2.2193708293612967e-06, "loss": 1.5337, "step": 13698 }, { "epoch": 2.61, "learning_rate": 2.219313632030505e-06, "loss": 2.0283, "step": 13699 }, { "epoch": 2.61, "learning_rate": 2.2192564346997142e-06, "loss": 2.6862, "step": 13700 }, { "epoch": 2.61, "learning_rate": 2.219199237368923e-06, "loss": 2.0284, "step": 13701 }, { "epoch": 2.61, "learning_rate": 2.2191420400381314e-06, "loss": 3.159, "step": 13702 }, { "epoch": 2.61, "learning_rate": 2.2190848427073406e-06, "loss": 2.146, "step": 13703 }, { "epoch": 2.61, "learning_rate": 2.219027645376549e-06, "loss": 3.203, "step": 13704 }, { "epoch": 2.61, "learning_rate": 2.2189704480457578e-06, "loss": 2.0021, "step": 13705 }, { "epoch": 2.61, "learning_rate": 2.2189132507149666e-06, "loss": 1.8762, "step": 13706 }, { "epoch": 2.61, "learning_rate": 2.2188560533841754e-06, "loss": 2.632, "step": 13707 }, { "epoch": 2.61, "learning_rate": 2.2187988560533846e-06, "loss": 2.7504, "step": 13708 }, { "epoch": 2.61, "learning_rate": 2.218741658722593e-06, "loss": 3.208, "step": 13709 }, { "epoch": 2.61, "learning_rate": 2.2186844613918018e-06, "loss": 2.7102, "step": 13710 }, { "epoch": 2.61, "learning_rate": 2.2186272640610105e-06, "loss": 1.915, "step": 13711 }, { "epoch": 2.61, "learning_rate": 2.2185700667302193e-06, "loss": 3.0049, "step": 13712 }, { "epoch": 2.61, "learning_rate": 2.218512869399428e-06, "loss": 3.2721, "step": 13713 }, { "epoch": 2.61, "learning_rate": 2.218455672068637e-06, "loss": 2.1338, "step": 13714 }, { "epoch": 2.61, "learning_rate": 2.2183984747378457e-06, "loss": 3.4242, "step": 13715 }, { "epoch": 2.61, "learning_rate": 2.218341277407054e-06, "loss": 3.5878, "step": 13716 }, { "epoch": 2.61, "learning_rate": 2.2182840800762633e-06, "loss": 3.412, "step": 13717 }, { "epoch": 2.61, "learning_rate": 2.218226882745472e-06, "loss": 2.4497, "step": 13718 }, { "epoch": 2.61, "learning_rate": 2.2181696854146805e-06, "loss": 2.257, "step": 13719 }, { "epoch": 2.61, "learning_rate": 2.2181124880838897e-06, "loss": 3.5713, "step": 13720 }, { "epoch": 2.61, "learning_rate": 2.218055290753098e-06, "loss": 3.4305, "step": 13721 }, { "epoch": 2.61, "learning_rate": 2.217998093422307e-06, "loss": 1.9401, "step": 13722 }, { "epoch": 2.61, "learning_rate": 2.2179408960915157e-06, "loss": 2.7367, "step": 13723 }, { "epoch": 2.61, "learning_rate": 2.2178836987607244e-06, "loss": 2.3533, "step": 13724 }, { "epoch": 2.61, "learning_rate": 2.2178265014299337e-06, "loss": 3.0745, "step": 13725 }, { "epoch": 2.61, "learning_rate": 2.217769304099142e-06, "loss": 2.1112, "step": 13726 }, { "epoch": 2.61, "learning_rate": 2.217712106768351e-06, "loss": 2.4421, "step": 13727 }, { "epoch": 2.61, "learning_rate": 2.2176549094375596e-06, "loss": 2.5884, "step": 13728 }, { "epoch": 2.62, "learning_rate": 2.2175977121067684e-06, "loss": 3.5801, "step": 13729 }, { "epoch": 2.62, "learning_rate": 2.217540514775977e-06, "loss": 3.0728, "step": 13730 }, { "epoch": 2.62, "learning_rate": 2.217483317445186e-06, "loss": 3.0116, "step": 13731 }, { "epoch": 2.62, "learning_rate": 2.217426120114395e-06, "loss": 3.4126, "step": 13732 }, { "epoch": 2.62, "learning_rate": 2.217368922783603e-06, "loss": 2.8424, "step": 13733 }, { "epoch": 2.62, "learning_rate": 2.2173117254528124e-06, "loss": 2.5494, "step": 13734 }, { "epoch": 2.62, "learning_rate": 2.217254528122021e-06, "loss": 1.8875, "step": 13735 }, { "epoch": 2.62, "learning_rate": 2.2171973307912295e-06, "loss": 3.331, "step": 13736 }, { "epoch": 2.62, "learning_rate": 2.2171401334604388e-06, "loss": 2.6925, "step": 13737 }, { "epoch": 2.62, "learning_rate": 2.217082936129647e-06, "loss": 2.0917, "step": 13738 }, { "epoch": 2.62, "learning_rate": 2.217025738798856e-06, "loss": 2.548, "step": 13739 }, { "epoch": 2.62, "learning_rate": 2.216968541468065e-06, "loss": 1.9566, "step": 13740 }, { "epoch": 2.62, "learning_rate": 2.2169113441372735e-06, "loss": 3.1736, "step": 13741 }, { "epoch": 2.62, "learning_rate": 2.2168541468064827e-06, "loss": 2.9964, "step": 13742 }, { "epoch": 2.62, "learning_rate": 2.216796949475691e-06, "loss": 3.12, "step": 13743 }, { "epoch": 2.62, "learning_rate": 2.2167397521449e-06, "loss": 2.607, "step": 13744 }, { "epoch": 2.62, "learning_rate": 2.2166825548141087e-06, "loss": 2.6335, "step": 13745 }, { "epoch": 2.62, "learning_rate": 2.2166253574833175e-06, "loss": 2.8544, "step": 13746 }, { "epoch": 2.62, "learning_rate": 2.2165681601525263e-06, "loss": 1.6205, "step": 13747 }, { "epoch": 2.62, "learning_rate": 2.216510962821735e-06, "loss": 3.5947, "step": 13748 }, { "epoch": 2.62, "learning_rate": 2.216453765490944e-06, "loss": 2.3929, "step": 13749 }, { "epoch": 2.62, "learning_rate": 2.2163965681601527e-06, "loss": 2.6984, "step": 13750 }, { "epoch": 2.62, "learning_rate": 2.2163393708293615e-06, "loss": 1.6853, "step": 13751 }, { "epoch": 2.62, "learning_rate": 2.2162821734985702e-06, "loss": 3.2935, "step": 13752 }, { "epoch": 2.62, "learning_rate": 2.2162249761677786e-06, "loss": 1.7258, "step": 13753 }, { "epoch": 2.62, "learning_rate": 2.216167778836988e-06, "loss": 2.3848, "step": 13754 }, { "epoch": 2.62, "learning_rate": 2.216110581506196e-06, "loss": 2.6226, "step": 13755 }, { "epoch": 2.62, "learning_rate": 2.216053384175405e-06, "loss": 2.3833, "step": 13756 }, { "epoch": 2.62, "learning_rate": 2.2159961868446142e-06, "loss": 3.2319, "step": 13757 }, { "epoch": 2.62, "learning_rate": 2.2159389895138226e-06, "loss": 2.0367, "step": 13758 }, { "epoch": 2.62, "learning_rate": 2.215881792183032e-06, "loss": 2.8113, "step": 13759 }, { "epoch": 2.62, "learning_rate": 2.21582459485224e-06, "loss": 3.4492, "step": 13760 }, { "epoch": 2.62, "learning_rate": 2.215767397521449e-06, "loss": 3.4327, "step": 13761 }, { "epoch": 2.62, "learning_rate": 2.2157102001906578e-06, "loss": 2.8691, "step": 13762 }, { "epoch": 2.62, "learning_rate": 2.2156530028598666e-06, "loss": 3.2116, "step": 13763 }, { "epoch": 2.62, "learning_rate": 2.2155958055290754e-06, "loss": 2.4298, "step": 13764 }, { "epoch": 2.62, "learning_rate": 2.215538608198284e-06, "loss": 3.2172, "step": 13765 }, { "epoch": 2.62, "learning_rate": 2.215481410867493e-06, "loss": 3.397, "step": 13766 }, { "epoch": 2.62, "learning_rate": 2.2154242135367017e-06, "loss": 3.091, "step": 13767 }, { "epoch": 2.62, "learning_rate": 2.2153670162059105e-06, "loss": 2.4024, "step": 13768 }, { "epoch": 2.62, "learning_rate": 2.2153098188751193e-06, "loss": 2.6009, "step": 13769 }, { "epoch": 2.62, "learning_rate": 2.2152526215443277e-06, "loss": 2.5054, "step": 13770 }, { "epoch": 2.62, "learning_rate": 2.215195424213537e-06, "loss": 2.5009, "step": 13771 }, { "epoch": 2.62, "learning_rate": 2.2151382268827453e-06, "loss": 2.0383, "step": 13772 }, { "epoch": 2.62, "learning_rate": 2.2150810295519545e-06, "loss": 2.6346, "step": 13773 }, { "epoch": 2.62, "learning_rate": 2.2150238322211633e-06, "loss": 3.8527, "step": 13774 }, { "epoch": 2.62, "learning_rate": 2.2149666348903717e-06, "loss": 3.3286, "step": 13775 }, { "epoch": 2.62, "learning_rate": 2.214909437559581e-06, "loss": 3.3512, "step": 13776 }, { "epoch": 2.62, "learning_rate": 2.2148522402287892e-06, "loss": 3.0381, "step": 13777 }, { "epoch": 2.62, "learning_rate": 2.214795042897998e-06, "loss": 2.6585, "step": 13778 }, { "epoch": 2.62, "learning_rate": 2.214737845567207e-06, "loss": 2.4587, "step": 13779 }, { "epoch": 2.62, "learning_rate": 2.2146806482364156e-06, "loss": 3.0343, "step": 13780 }, { "epoch": 2.62, "learning_rate": 2.2146234509056244e-06, "loss": 3.0612, "step": 13781 }, { "epoch": 2.63, "learning_rate": 2.2145662535748332e-06, "loss": 2.3526, "step": 13782 }, { "epoch": 2.63, "learning_rate": 2.214509056244042e-06, "loss": 2.271, "step": 13783 }, { "epoch": 2.63, "learning_rate": 2.214451858913251e-06, "loss": 1.8502, "step": 13784 }, { "epoch": 2.63, "learning_rate": 2.2143946615824596e-06, "loss": 2.5096, "step": 13785 }, { "epoch": 2.63, "learning_rate": 2.2143374642516684e-06, "loss": 2.2936, "step": 13786 }, { "epoch": 2.63, "learning_rate": 2.2142802669208768e-06, "loss": 2.5024, "step": 13787 }, { "epoch": 2.63, "learning_rate": 2.214223069590086e-06, "loss": 3.0323, "step": 13788 }, { "epoch": 2.63, "learning_rate": 2.2141658722592948e-06, "loss": 1.8887, "step": 13789 }, { "epoch": 2.63, "learning_rate": 2.2141086749285036e-06, "loss": 2.8787, "step": 13790 }, { "epoch": 2.63, "learning_rate": 2.2140514775977124e-06, "loss": 3.3867, "step": 13791 }, { "epoch": 2.63, "learning_rate": 2.2139942802669207e-06, "loss": 2.8886, "step": 13792 }, { "epoch": 2.63, "learning_rate": 2.21393708293613e-06, "loss": 2.4153, "step": 13793 }, { "epoch": 2.63, "learning_rate": 2.2138798856053383e-06, "loss": 3.3043, "step": 13794 }, { "epoch": 2.63, "learning_rate": 2.213822688274547e-06, "loss": 2.0937, "step": 13795 }, { "epoch": 2.63, "learning_rate": 2.2137654909437563e-06, "loss": 2.1374, "step": 13796 }, { "epoch": 2.63, "learning_rate": 2.2137082936129647e-06, "loss": 3.0037, "step": 13797 }, { "epoch": 2.63, "learning_rate": 2.2136510962821735e-06, "loss": 2.2932, "step": 13798 }, { "epoch": 2.63, "learning_rate": 2.2135938989513823e-06, "loss": 3.4026, "step": 13799 }, { "epoch": 2.63, "learning_rate": 2.213536701620591e-06, "loss": 2.2928, "step": 13800 }, { "epoch": 2.63, "learning_rate": 2.2134795042898e-06, "loss": 2.9151, "step": 13801 }, { "epoch": 2.63, "learning_rate": 2.2134223069590087e-06, "loss": 2.7934, "step": 13802 }, { "epoch": 2.63, "learning_rate": 2.2133651096282175e-06, "loss": 3.4377, "step": 13803 }, { "epoch": 2.63, "learning_rate": 2.213307912297426e-06, "loss": 3.2036, "step": 13804 }, { "epoch": 2.63, "learning_rate": 2.213250714966635e-06, "loss": 3.1125, "step": 13805 }, { "epoch": 2.63, "learning_rate": 2.213193517635844e-06, "loss": 2.1592, "step": 13806 }, { "epoch": 2.63, "learning_rate": 2.2131363203050526e-06, "loss": 2.631, "step": 13807 }, { "epoch": 2.63, "learning_rate": 2.2130791229742614e-06, "loss": 2.1058, "step": 13808 }, { "epoch": 2.63, "learning_rate": 2.21302192564347e-06, "loss": 3.6999, "step": 13809 }, { "epoch": 2.63, "learning_rate": 2.212964728312679e-06, "loss": 2.0643, "step": 13810 }, { "epoch": 2.63, "learning_rate": 2.2129075309818874e-06, "loss": 2.7742, "step": 13811 }, { "epoch": 2.63, "learning_rate": 2.212850333651096e-06, "loss": 2.4153, "step": 13812 }, { "epoch": 2.63, "learning_rate": 2.2127931363203054e-06, "loss": 2.1817, "step": 13813 }, { "epoch": 2.63, "learning_rate": 2.2127359389895138e-06, "loss": 2.4001, "step": 13814 }, { "epoch": 2.63, "learning_rate": 2.2126787416587226e-06, "loss": 3.2668, "step": 13815 }, { "epoch": 2.63, "learning_rate": 2.2126215443279314e-06, "loss": 3.4889, "step": 13816 }, { "epoch": 2.63, "learning_rate": 2.21256434699714e-06, "loss": 3.4108, "step": 13817 }, { "epoch": 2.63, "learning_rate": 2.212507149666349e-06, "loss": 2.7963, "step": 13818 }, { "epoch": 2.63, "learning_rate": 2.2124499523355577e-06, "loss": 3.2994, "step": 13819 }, { "epoch": 2.63, "learning_rate": 2.2123927550047665e-06, "loss": 3.1413, "step": 13820 }, { "epoch": 2.63, "learning_rate": 2.212335557673975e-06, "loss": 2.137, "step": 13821 }, { "epoch": 2.63, "learning_rate": 2.212278360343184e-06, "loss": 3.0594, "step": 13822 }, { "epoch": 2.63, "learning_rate": 2.212221163012393e-06, "loss": 1.9195, "step": 13823 }, { "epoch": 2.63, "learning_rate": 2.2121639656816017e-06, "loss": 2.5227, "step": 13824 }, { "epoch": 2.63, "learning_rate": 2.2121067683508105e-06, "loss": 2.6779, "step": 13825 }, { "epoch": 2.63, "learning_rate": 2.212049571020019e-06, "loss": 2.0217, "step": 13826 }, { "epoch": 2.63, "learning_rate": 2.211992373689228e-06, "loss": 2.1748, "step": 13827 }, { "epoch": 2.63, "learning_rate": 2.2119351763584365e-06, "loss": 1.794, "step": 13828 }, { "epoch": 2.63, "learning_rate": 2.2118779790276453e-06, "loss": 2.5709, "step": 13829 }, { "epoch": 2.63, "learning_rate": 2.2118207816968545e-06, "loss": 2.53, "step": 13830 }, { "epoch": 2.63, "learning_rate": 2.211763584366063e-06, "loss": 2.5081, "step": 13831 }, { "epoch": 2.63, "learning_rate": 2.2117063870352716e-06, "loss": 2.8739, "step": 13832 }, { "epoch": 2.63, "learning_rate": 2.2116491897044804e-06, "loss": 2.7965, "step": 13833 }, { "epoch": 2.64, "learning_rate": 2.2115919923736892e-06, "loss": 4.0645, "step": 13834 }, { "epoch": 2.64, "learning_rate": 2.211534795042898e-06, "loss": 3.4358, "step": 13835 }, { "epoch": 2.64, "learning_rate": 2.211477597712107e-06, "loss": 2.4533, "step": 13836 }, { "epoch": 2.64, "learning_rate": 2.2114204003813156e-06, "loss": 2.3929, "step": 13837 }, { "epoch": 2.64, "learning_rate": 2.2113632030505244e-06, "loss": 2.7948, "step": 13838 }, { "epoch": 2.64, "learning_rate": 2.211306005719733e-06, "loss": 3.1075, "step": 13839 }, { "epoch": 2.64, "learning_rate": 2.211248808388942e-06, "loss": 2.6081, "step": 13840 }, { "epoch": 2.64, "learning_rate": 2.2111916110581508e-06, "loss": 1.868, "step": 13841 }, { "epoch": 2.64, "learning_rate": 2.2111344137273596e-06, "loss": 3.3497, "step": 13842 }, { "epoch": 2.64, "learning_rate": 2.211077216396568e-06, "loss": 2.4625, "step": 13843 }, { "epoch": 2.64, "learning_rate": 2.211020019065777e-06, "loss": 3.2534, "step": 13844 }, { "epoch": 2.64, "learning_rate": 2.210962821734986e-06, "loss": 2.3754, "step": 13845 }, { "epoch": 2.64, "learning_rate": 2.2109056244041943e-06, "loss": 2.9513, "step": 13846 }, { "epoch": 2.64, "learning_rate": 2.2108484270734035e-06, "loss": 2.3108, "step": 13847 }, { "epoch": 2.64, "learning_rate": 2.210791229742612e-06, "loss": 2.0255, "step": 13848 }, { "epoch": 2.64, "learning_rate": 2.2107340324118207e-06, "loss": 3.0723, "step": 13849 }, { "epoch": 2.64, "learning_rate": 2.2106768350810295e-06, "loss": 1.9103, "step": 13850 }, { "epoch": 2.64, "learning_rate": 2.2106196377502383e-06, "loss": 3.0473, "step": 13851 }, { "epoch": 2.64, "learning_rate": 2.210562440419447e-06, "loss": 3.4788, "step": 13852 }, { "epoch": 2.64, "learning_rate": 2.210505243088656e-06, "loss": 2.4395, "step": 13853 }, { "epoch": 2.64, "learning_rate": 2.2104480457578647e-06, "loss": 3.1113, "step": 13854 }, { "epoch": 2.64, "learning_rate": 2.2103908484270735e-06, "loss": 2.4191, "step": 13855 }, { "epoch": 2.64, "learning_rate": 2.2103336510962823e-06, "loss": 1.8707, "step": 13856 }, { "epoch": 2.64, "learning_rate": 2.210276453765491e-06, "loss": 3.2, "step": 13857 }, { "epoch": 2.64, "learning_rate": 2.2102192564347e-06, "loss": 2.0296, "step": 13858 }, { "epoch": 2.64, "learning_rate": 2.2101620591039086e-06, "loss": 3.1145, "step": 13859 }, { "epoch": 2.64, "learning_rate": 2.210104861773117e-06, "loss": 2.4828, "step": 13860 }, { "epoch": 2.64, "learning_rate": 2.2100476644423262e-06, "loss": 3.2338, "step": 13861 }, { "epoch": 2.64, "learning_rate": 2.209990467111535e-06, "loss": 2.3722, "step": 13862 }, { "epoch": 2.64, "learning_rate": 2.2099332697807434e-06, "loss": 3.5046, "step": 13863 }, { "epoch": 2.64, "learning_rate": 2.2098760724499526e-06, "loss": 2.0376, "step": 13864 }, { "epoch": 2.64, "learning_rate": 2.209818875119161e-06, "loss": 2.0252, "step": 13865 }, { "epoch": 2.64, "learning_rate": 2.2097616777883698e-06, "loss": 2.9953, "step": 13866 }, { "epoch": 2.64, "learning_rate": 2.2097044804575786e-06, "loss": 2.9189, "step": 13867 }, { "epoch": 2.64, "learning_rate": 2.2096472831267874e-06, "loss": 2.7509, "step": 13868 }, { "epoch": 2.64, "learning_rate": 2.209590085795996e-06, "loss": 1.4183, "step": 13869 }, { "epoch": 2.64, "learning_rate": 2.209532888465205e-06, "loss": 2.6314, "step": 13870 }, { "epoch": 2.64, "learning_rate": 2.2094756911344137e-06, "loss": 2.2627, "step": 13871 }, { "epoch": 2.64, "learning_rate": 2.2094184938036225e-06, "loss": 2.3585, "step": 13872 }, { "epoch": 2.64, "learning_rate": 2.2093612964728313e-06, "loss": 3.121, "step": 13873 }, { "epoch": 2.64, "learning_rate": 2.20930409914204e-06, "loss": 3.164, "step": 13874 }, { "epoch": 2.64, "learning_rate": 2.209246901811249e-06, "loss": 1.7466, "step": 13875 }, { "epoch": 2.64, "learning_rate": 2.2091897044804577e-06, "loss": 1.2685, "step": 13876 }, { "epoch": 2.64, "learning_rate": 2.2091325071496665e-06, "loss": 2.4079, "step": 13877 }, { "epoch": 2.64, "learning_rate": 2.2090753098188753e-06, "loss": 2.9194, "step": 13878 }, { "epoch": 2.64, "learning_rate": 2.209018112488084e-06, "loss": 2.0716, "step": 13879 }, { "epoch": 2.64, "learning_rate": 2.2089609151572925e-06, "loss": 3.2461, "step": 13880 }, { "epoch": 2.64, "learning_rate": 2.2089037178265017e-06, "loss": 2.3692, "step": 13881 }, { "epoch": 2.64, "learning_rate": 2.20884652049571e-06, "loss": 2.695, "step": 13882 }, { "epoch": 2.64, "learning_rate": 2.208789323164919e-06, "loss": 2.2954, "step": 13883 }, { "epoch": 2.64, "learning_rate": 2.208732125834128e-06, "loss": 2.196, "step": 13884 }, { "epoch": 2.64, "learning_rate": 2.2086749285033364e-06, "loss": 3.2178, "step": 13885 }, { "epoch": 2.64, "learning_rate": 2.2086177311725452e-06, "loss": 3.0194, "step": 13886 }, { "epoch": 2.65, "learning_rate": 2.208560533841754e-06, "loss": 2.3969, "step": 13887 }, { "epoch": 2.65, "learning_rate": 2.208503336510963e-06, "loss": 2.4745, "step": 13888 }, { "epoch": 2.65, "learning_rate": 2.2084461391801716e-06, "loss": 2.8754, "step": 13889 }, { "epoch": 2.65, "learning_rate": 2.2083889418493804e-06, "loss": 2.4092, "step": 13890 }, { "epoch": 2.65, "learning_rate": 2.208331744518589e-06, "loss": 2.429, "step": 13891 }, { "epoch": 2.65, "learning_rate": 2.208274547187798e-06, "loss": 2.6377, "step": 13892 }, { "epoch": 2.65, "learning_rate": 2.208217349857007e-06, "loss": 2.7321, "step": 13893 }, { "epoch": 2.65, "learning_rate": 2.2081601525262156e-06, "loss": 2.6816, "step": 13894 }, { "epoch": 2.65, "learning_rate": 2.2081029551954244e-06, "loss": 1.8054, "step": 13895 }, { "epoch": 2.65, "learning_rate": 2.208045757864633e-06, "loss": 1.8659, "step": 13896 }, { "epoch": 2.65, "learning_rate": 2.2079885605338415e-06, "loss": 1.7135, "step": 13897 }, { "epoch": 2.65, "learning_rate": 2.2079313632030508e-06, "loss": 2.8724, "step": 13898 }, { "epoch": 2.65, "learning_rate": 2.207874165872259e-06, "loss": 3.057, "step": 13899 }, { "epoch": 2.65, "learning_rate": 2.207816968541468e-06, "loss": 3.555, "step": 13900 }, { "epoch": 2.65, "learning_rate": 2.207759771210677e-06, "loss": 1.7898, "step": 13901 }, { "epoch": 2.65, "learning_rate": 2.2077025738798855e-06, "loss": 3.3922, "step": 13902 }, { "epoch": 2.65, "learning_rate": 2.2076453765490947e-06, "loss": 1.796, "step": 13903 }, { "epoch": 2.65, "learning_rate": 2.207588179218303e-06, "loss": 2.8167, "step": 13904 }, { "epoch": 2.65, "learning_rate": 2.207530981887512e-06, "loss": 2.9539, "step": 13905 }, { "epoch": 2.65, "learning_rate": 2.2074737845567207e-06, "loss": 3.0036, "step": 13906 }, { "epoch": 2.65, "learning_rate": 2.2074165872259295e-06, "loss": 2.7157, "step": 13907 }, { "epoch": 2.65, "learning_rate": 2.2073593898951383e-06, "loss": 3.2728, "step": 13908 }, { "epoch": 2.65, "learning_rate": 2.207302192564347e-06, "loss": 3.1376, "step": 13909 }, { "epoch": 2.65, "learning_rate": 2.207244995233556e-06, "loss": 1.9778, "step": 13910 }, { "epoch": 2.65, "learning_rate": 2.2071877979027647e-06, "loss": 2.0177, "step": 13911 }, { "epoch": 2.65, "learning_rate": 2.2071306005719734e-06, "loss": 2.48, "step": 13912 }, { "epoch": 2.65, "learning_rate": 2.2070734032411822e-06, "loss": 3.0087, "step": 13913 }, { "epoch": 2.65, "learning_rate": 2.2070162059103906e-06, "loss": 2.347, "step": 13914 }, { "epoch": 2.65, "learning_rate": 2.2069590085796e-06, "loss": 3.3377, "step": 13915 }, { "epoch": 2.65, "learning_rate": 2.206901811248808e-06, "loss": 2.7579, "step": 13916 }, { "epoch": 2.65, "learning_rate": 2.206844613918017e-06, "loss": 2.0785, "step": 13917 }, { "epoch": 2.65, "learning_rate": 2.206787416587226e-06, "loss": 2.5255, "step": 13918 }, { "epoch": 2.65, "learning_rate": 2.2067302192564346e-06, "loss": 3.1134, "step": 13919 }, { "epoch": 2.65, "learning_rate": 2.206673021925644e-06, "loss": 2.3885, "step": 13920 }, { "epoch": 2.65, "learning_rate": 2.206615824594852e-06, "loss": 2.4181, "step": 13921 }, { "epoch": 2.65, "learning_rate": 2.206558627264061e-06, "loss": 1.7447, "step": 13922 }, { "epoch": 2.65, "learning_rate": 2.20650142993327e-06, "loss": 2.0834, "step": 13923 }, { "epoch": 2.65, "learning_rate": 2.2064442326024785e-06, "loss": 3.3696, "step": 13924 }, { "epoch": 2.65, "learning_rate": 2.2063870352716873e-06, "loss": 1.7989, "step": 13925 }, { "epoch": 2.65, "learning_rate": 2.206329837940896e-06, "loss": 2.3489, "step": 13926 }, { "epoch": 2.65, "learning_rate": 2.206272640610105e-06, "loss": 2.8814, "step": 13927 }, { "epoch": 2.65, "learning_rate": 2.2062154432793137e-06, "loss": 1.5133, "step": 13928 }, { "epoch": 2.65, "learning_rate": 2.2061582459485225e-06, "loss": 3.4956, "step": 13929 }, { "epoch": 2.65, "learning_rate": 2.2061010486177313e-06, "loss": 3.1256, "step": 13930 }, { "epoch": 2.65, "learning_rate": 2.2060438512869397e-06, "loss": 2.2873, "step": 13931 }, { "epoch": 2.65, "learning_rate": 2.205986653956149e-06, "loss": 2.5816, "step": 13932 }, { "epoch": 2.65, "learning_rate": 2.2059294566253577e-06, "loss": 3.1216, "step": 13933 }, { "epoch": 2.65, "learning_rate": 2.205872259294566e-06, "loss": 2.332, "step": 13934 }, { "epoch": 2.65, "learning_rate": 2.2058150619637753e-06, "loss": 2.5108, "step": 13935 }, { "epoch": 2.65, "learning_rate": 2.2057578646329837e-06, "loss": 2.7567, "step": 13936 }, { "epoch": 2.65, "learning_rate": 2.205700667302193e-06, "loss": 2.2294, "step": 13937 }, { "epoch": 2.65, "learning_rate": 2.2056434699714012e-06, "loss": 1.7086, "step": 13938 }, { "epoch": 2.66, "learning_rate": 2.20558627264061e-06, "loss": 2.5712, "step": 13939 }, { "epoch": 2.66, "learning_rate": 2.2055290753098193e-06, "loss": 4.0511, "step": 13940 }, { "epoch": 2.66, "learning_rate": 2.2054718779790276e-06, "loss": 2.3071, "step": 13941 }, { "epoch": 2.66, "learning_rate": 2.2054146806482364e-06, "loss": 3.0782, "step": 13942 }, { "epoch": 2.66, "learning_rate": 2.205357483317445e-06, "loss": 2.2026, "step": 13943 }, { "epoch": 2.66, "learning_rate": 2.205300285986654e-06, "loss": 2.0461, "step": 13944 }, { "epoch": 2.66, "learning_rate": 2.205243088655863e-06, "loss": 2.1665, "step": 13945 }, { "epoch": 2.66, "learning_rate": 2.2051858913250716e-06, "loss": 3.7378, "step": 13946 }, { "epoch": 2.66, "learning_rate": 2.2051286939942804e-06, "loss": 2.3056, "step": 13947 }, { "epoch": 2.66, "learning_rate": 2.2050714966634888e-06, "loss": 2.3424, "step": 13948 }, { "epoch": 2.66, "learning_rate": 2.205014299332698e-06, "loss": 3.1631, "step": 13949 }, { "epoch": 2.66, "learning_rate": 2.2049571020019068e-06, "loss": 2.975, "step": 13950 }, { "epoch": 2.66, "learning_rate": 2.204899904671115e-06, "loss": 2.8948, "step": 13951 }, { "epoch": 2.66, "learning_rate": 2.2048427073403244e-06, "loss": 2.4866, "step": 13952 }, { "epoch": 2.66, "learning_rate": 2.2047855100095327e-06, "loss": 1.8385, "step": 13953 }, { "epoch": 2.66, "learning_rate": 2.204728312678742e-06, "loss": 3.6515, "step": 13954 }, { "epoch": 2.66, "learning_rate": 2.2046711153479503e-06, "loss": 2.9612, "step": 13955 }, { "epoch": 2.66, "learning_rate": 2.204613918017159e-06, "loss": 2.6395, "step": 13956 }, { "epoch": 2.66, "learning_rate": 2.2045567206863683e-06, "loss": 2.9198, "step": 13957 }, { "epoch": 2.66, "learning_rate": 2.2044995233555767e-06, "loss": 2.6541, "step": 13958 }, { "epoch": 2.66, "learning_rate": 2.2044423260247855e-06, "loss": 2.6462, "step": 13959 }, { "epoch": 2.66, "learning_rate": 2.2043851286939943e-06, "loss": 3.726, "step": 13960 }, { "epoch": 2.66, "learning_rate": 2.204327931363203e-06, "loss": 3.0624, "step": 13961 }, { "epoch": 2.66, "learning_rate": 2.204270734032412e-06, "loss": 3.0265, "step": 13962 }, { "epoch": 2.66, "learning_rate": 2.2042135367016207e-06, "loss": 3.5358, "step": 13963 }, { "epoch": 2.66, "learning_rate": 2.2041563393708295e-06, "loss": 2.6729, "step": 13964 }, { "epoch": 2.66, "learning_rate": 2.2040991420400382e-06, "loss": 2.7121, "step": 13965 }, { "epoch": 2.66, "learning_rate": 2.204041944709247e-06, "loss": 3.7484, "step": 13966 }, { "epoch": 2.66, "learning_rate": 2.203984747378456e-06, "loss": 2.5171, "step": 13967 }, { "epoch": 2.66, "learning_rate": 2.2039275500476646e-06, "loss": 3.3745, "step": 13968 }, { "epoch": 2.66, "learning_rate": 2.2038703527168734e-06, "loss": 2.8352, "step": 13969 }, { "epoch": 2.66, "learning_rate": 2.203813155386082e-06, "loss": 1.8745, "step": 13970 }, { "epoch": 2.66, "learning_rate": 2.203755958055291e-06, "loss": 3.122, "step": 13971 }, { "epoch": 2.66, "learning_rate": 2.2036987607245e-06, "loss": 1.6065, "step": 13972 }, { "epoch": 2.66, "learning_rate": 2.203641563393708e-06, "loss": 3.427, "step": 13973 }, { "epoch": 2.66, "learning_rate": 2.2035843660629174e-06, "loss": 2.8541, "step": 13974 }, { "epoch": 2.66, "learning_rate": 2.2035271687321258e-06, "loss": 2.4282, "step": 13975 }, { "epoch": 2.66, "learning_rate": 2.2034699714013346e-06, "loss": 2.2981, "step": 13976 }, { "epoch": 2.66, "learning_rate": 2.2034127740705434e-06, "loss": 2.1298, "step": 13977 }, { "epoch": 2.66, "learning_rate": 2.203355576739752e-06, "loss": 2.9518, "step": 13978 }, { "epoch": 2.66, "learning_rate": 2.203298379408961e-06, "loss": 2.9852, "step": 13979 }, { "epoch": 2.66, "learning_rate": 2.2032411820781697e-06, "loss": 2.2261, "step": 13980 }, { "epoch": 2.66, "learning_rate": 2.2031839847473785e-06, "loss": 1.7921, "step": 13981 }, { "epoch": 2.66, "learning_rate": 2.2031267874165873e-06, "loss": 2.2054, "step": 13982 }, { "epoch": 2.66, "learning_rate": 2.203069590085796e-06, "loss": 2.3855, "step": 13983 }, { "epoch": 2.66, "learning_rate": 2.203012392755005e-06, "loss": 2.2669, "step": 13984 }, { "epoch": 2.66, "learning_rate": 2.2029551954242137e-06, "loss": 3.161, "step": 13985 }, { "epoch": 2.66, "learning_rate": 2.2028979980934225e-06, "loss": 3.4819, "step": 13986 }, { "epoch": 2.66, "learning_rate": 2.202840800762631e-06, "loss": 2.968, "step": 13987 }, { "epoch": 2.66, "learning_rate": 2.20278360343184e-06, "loss": 2.9823, "step": 13988 }, { "epoch": 2.66, "learning_rate": 2.202726406101049e-06, "loss": 2.9822, "step": 13989 }, { "epoch": 2.66, "learning_rate": 2.2026692087702572e-06, "loss": 1.9484, "step": 13990 }, { "epoch": 2.66, "learning_rate": 2.2026120114394665e-06, "loss": 3.2184, "step": 13991 }, { "epoch": 2.67, "learning_rate": 2.202554814108675e-06, "loss": 1.8042, "step": 13992 }, { "epoch": 2.67, "learning_rate": 2.2024976167778836e-06, "loss": 2.8931, "step": 13993 }, { "epoch": 2.67, "learning_rate": 2.2024404194470924e-06, "loss": 3.2814, "step": 13994 }, { "epoch": 2.67, "learning_rate": 2.2023832221163012e-06, "loss": 2.1824, "step": 13995 }, { "epoch": 2.67, "learning_rate": 2.20232602478551e-06, "loss": 3.0135, "step": 13996 }, { "epoch": 2.67, "learning_rate": 2.202268827454719e-06, "loss": 2.9223, "step": 13997 }, { "epoch": 2.67, "learning_rate": 2.2022116301239276e-06, "loss": 2.6921, "step": 13998 }, { "epoch": 2.67, "learning_rate": 2.2021544327931364e-06, "loss": 2.4534, "step": 13999 }, { "epoch": 2.67, "learning_rate": 2.202097235462345e-06, "loss": 2.8307, "step": 14000 }, { "epoch": 2.67, "learning_rate": 2.202040038131554e-06, "loss": 3.2194, "step": 14001 }, { "epoch": 2.67, "learning_rate": 2.2019828408007628e-06, "loss": 2.5177, "step": 14002 }, { "epoch": 2.67, "learning_rate": 2.2019256434699716e-06, "loss": 2.4253, "step": 14003 }, { "epoch": 2.67, "learning_rate": 2.20186844613918e-06, "loss": 2.6274, "step": 14004 }, { "epoch": 2.67, "learning_rate": 2.201811248808389e-06, "loss": 2.5116, "step": 14005 }, { "epoch": 2.67, "learning_rate": 2.201754051477598e-06, "loss": 2.8183, "step": 14006 }, { "epoch": 2.67, "learning_rate": 2.2016968541468063e-06, "loss": 2.4561, "step": 14007 }, { "epoch": 2.67, "learning_rate": 2.2016396568160155e-06, "loss": 2.2623, "step": 14008 }, { "epoch": 2.67, "learning_rate": 2.201582459485224e-06, "loss": 2.6694, "step": 14009 }, { "epoch": 2.67, "learning_rate": 2.2015252621544327e-06, "loss": 2.8823, "step": 14010 }, { "epoch": 2.67, "learning_rate": 2.201468064823642e-06, "loss": 2.3883, "step": 14011 }, { "epoch": 2.67, "learning_rate": 2.2014108674928503e-06, "loss": 2.7233, "step": 14012 }, { "epoch": 2.67, "learning_rate": 2.201353670162059e-06, "loss": 2.3964, "step": 14013 }, { "epoch": 2.67, "learning_rate": 2.201296472831268e-06, "loss": 2.0522, "step": 14014 }, { "epoch": 2.67, "learning_rate": 2.2012392755004767e-06, "loss": 3.3918, "step": 14015 }, { "epoch": 2.67, "learning_rate": 2.2011820781696855e-06, "loss": 2.5155, "step": 14016 }, { "epoch": 2.67, "learning_rate": 2.2011248808388943e-06, "loss": 2.9077, "step": 14017 }, { "epoch": 2.67, "learning_rate": 2.201067683508103e-06, "loss": 2.2941, "step": 14018 }, { "epoch": 2.67, "learning_rate": 2.201010486177312e-06, "loss": 2.2611, "step": 14019 }, { "epoch": 2.67, "learning_rate": 2.2009532888465206e-06, "loss": 2.3227, "step": 14020 }, { "epoch": 2.67, "learning_rate": 2.2008960915157294e-06, "loss": 2.8515, "step": 14021 }, { "epoch": 2.67, "learning_rate": 2.2008388941849382e-06, "loss": 2.1542, "step": 14022 }, { "epoch": 2.67, "learning_rate": 2.200781696854147e-06, "loss": 2.2467, "step": 14023 }, { "epoch": 2.67, "learning_rate": 2.2007244995233554e-06, "loss": 4.2044, "step": 14024 }, { "epoch": 2.67, "learning_rate": 2.2006673021925646e-06, "loss": 1.971, "step": 14025 }, { "epoch": 2.67, "learning_rate": 2.200610104861773e-06, "loss": 2.6762, "step": 14026 }, { "epoch": 2.67, "learning_rate": 2.2005529075309818e-06, "loss": 1.754, "step": 14027 }, { "epoch": 2.67, "learning_rate": 2.200495710200191e-06, "loss": 2.4383, "step": 14028 }, { "epoch": 2.67, "learning_rate": 2.2004385128693994e-06, "loss": 2.295, "step": 14029 }, { "epoch": 2.67, "learning_rate": 2.200381315538608e-06, "loss": 2.5592, "step": 14030 }, { "epoch": 2.67, "learning_rate": 2.200324118207817e-06, "loss": 2.7537, "step": 14031 }, { "epoch": 2.67, "learning_rate": 2.2002669208770257e-06, "loss": 3.8824, "step": 14032 }, { "epoch": 2.67, "learning_rate": 2.2002097235462345e-06, "loss": 3.0101, "step": 14033 }, { "epoch": 2.67, "learning_rate": 2.2001525262154433e-06, "loss": 2.5455, "step": 14034 }, { "epoch": 2.67, "learning_rate": 2.200095328884652e-06, "loss": 3.3952, "step": 14035 }, { "epoch": 2.67, "learning_rate": 2.200038131553861e-06, "loss": 2.055, "step": 14036 }, { "epoch": 2.67, "learning_rate": 2.1999809342230697e-06, "loss": 2.6696, "step": 14037 }, { "epoch": 2.67, "learning_rate": 2.1999237368922785e-06, "loss": 2.4345, "step": 14038 }, { "epoch": 2.67, "learning_rate": 2.1998665395614873e-06, "loss": 2.3346, "step": 14039 }, { "epoch": 2.67, "learning_rate": 2.199809342230696e-06, "loss": 2.9395, "step": 14040 }, { "epoch": 2.67, "learning_rate": 2.1997521448999045e-06, "loss": 2.7009, "step": 14041 }, { "epoch": 2.67, "learning_rate": 2.1996949475691137e-06, "loss": 1.8593, "step": 14042 }, { "epoch": 2.67, "learning_rate": 2.199637750238322e-06, "loss": 2.2443, "step": 14043 }, { "epoch": 2.68, "learning_rate": 2.199580552907531e-06, "loss": 3.9335, "step": 14044 }, { "epoch": 2.68, "learning_rate": 2.19952335557674e-06, "loss": 1.4296, "step": 14045 }, { "epoch": 2.68, "learning_rate": 2.1994661582459484e-06, "loss": 1.3914, "step": 14046 }, { "epoch": 2.68, "learning_rate": 2.1994089609151572e-06, "loss": 3.085, "step": 14047 }, { "epoch": 2.68, "learning_rate": 2.199351763584366e-06, "loss": 2.4618, "step": 14048 }, { "epoch": 2.68, "learning_rate": 2.199294566253575e-06, "loss": 2.3513, "step": 14049 }, { "epoch": 2.68, "learning_rate": 2.199237368922784e-06, "loss": 1.8739, "step": 14050 }, { "epoch": 2.68, "learning_rate": 2.1991801715919924e-06, "loss": 3.2189, "step": 14051 }, { "epoch": 2.68, "learning_rate": 2.199122974261201e-06, "loss": 3.0686, "step": 14052 }, { "epoch": 2.68, "learning_rate": 2.19906577693041e-06, "loss": 2.5988, "step": 14053 }, { "epoch": 2.68, "learning_rate": 2.1990085795996188e-06, "loss": 3.2333, "step": 14054 }, { "epoch": 2.68, "learning_rate": 2.1989513822688276e-06, "loss": 2.5558, "step": 14055 }, { "epoch": 2.68, "learning_rate": 2.1988941849380364e-06, "loss": 2.2518, "step": 14056 }, { "epoch": 2.68, "learning_rate": 2.198836987607245e-06, "loss": 2.8935, "step": 14057 }, { "epoch": 2.68, "learning_rate": 2.1987797902764535e-06, "loss": 3.2005, "step": 14058 }, { "epoch": 2.68, "learning_rate": 2.1987225929456628e-06, "loss": 3.4015, "step": 14059 }, { "epoch": 2.68, "learning_rate": 2.1986653956148715e-06, "loss": 2.8212, "step": 14060 }, { "epoch": 2.68, "learning_rate": 2.19860819828408e-06, "loss": 2.3064, "step": 14061 }, { "epoch": 2.68, "learning_rate": 2.198551000953289e-06, "loss": 2.255, "step": 14062 }, { "epoch": 2.68, "learning_rate": 2.1984938036224975e-06, "loss": 2.4445, "step": 14063 }, { "epoch": 2.68, "learning_rate": 2.1984366062917063e-06, "loss": 1.6265, "step": 14064 }, { "epoch": 2.68, "learning_rate": 2.198379408960915e-06, "loss": 2.4773, "step": 14065 }, { "epoch": 2.68, "learning_rate": 2.198322211630124e-06, "loss": 2.9052, "step": 14066 }, { "epoch": 2.68, "learning_rate": 2.198265014299333e-06, "loss": 2.4463, "step": 14067 }, { "epoch": 2.68, "learning_rate": 2.1982078169685415e-06, "loss": 2.9925, "step": 14068 }, { "epoch": 2.68, "learning_rate": 2.1981506196377503e-06, "loss": 2.999, "step": 14069 }, { "epoch": 2.68, "learning_rate": 2.198093422306959e-06, "loss": 2.6282, "step": 14070 }, { "epoch": 2.68, "learning_rate": 2.198036224976168e-06, "loss": 2.8618, "step": 14071 }, { "epoch": 2.68, "learning_rate": 2.1979790276453766e-06, "loss": 2.1512, "step": 14072 }, { "epoch": 2.68, "learning_rate": 2.1979218303145854e-06, "loss": 2.7053, "step": 14073 }, { "epoch": 2.68, "learning_rate": 2.1978646329837942e-06, "loss": 1.8478, "step": 14074 }, { "epoch": 2.68, "learning_rate": 2.1978074356530026e-06, "loss": 3.924, "step": 14075 }, { "epoch": 2.68, "learning_rate": 2.197750238322212e-06, "loss": 2.0537, "step": 14076 }, { "epoch": 2.68, "learning_rate": 2.1976930409914206e-06, "loss": 2.9251, "step": 14077 }, { "epoch": 2.68, "learning_rate": 2.197635843660629e-06, "loss": 2.8948, "step": 14078 }, { "epoch": 2.68, "learning_rate": 2.197578646329838e-06, "loss": 3.1803, "step": 14079 }, { "epoch": 2.68, "learning_rate": 2.1975214489990466e-06, "loss": 2.0432, "step": 14080 }, { "epoch": 2.68, "learning_rate": 2.1974642516682554e-06, "loss": 2.1321, "step": 14081 }, { "epoch": 2.68, "learning_rate": 2.197407054337464e-06, "loss": 3.5789, "step": 14082 }, { "epoch": 2.68, "learning_rate": 2.197349857006673e-06, "loss": 2.1876, "step": 14083 }, { "epoch": 2.68, "learning_rate": 2.197292659675882e-06, "loss": 1.743, "step": 14084 }, { "epoch": 2.68, "learning_rate": 2.1972354623450905e-06, "loss": 2.8961, "step": 14085 }, { "epoch": 2.68, "learning_rate": 2.1971782650142993e-06, "loss": 2.8868, "step": 14086 }, { "epoch": 2.68, "learning_rate": 2.197121067683508e-06, "loss": 3.3375, "step": 14087 }, { "epoch": 2.68, "learning_rate": 2.197063870352717e-06, "loss": 3.2249, "step": 14088 }, { "epoch": 2.68, "learning_rate": 2.1970066730219257e-06, "loss": 3.2685, "step": 14089 }, { "epoch": 2.68, "learning_rate": 2.1969494756911345e-06, "loss": 2.7805, "step": 14090 }, { "epoch": 2.68, "learning_rate": 2.1968922783603433e-06, "loss": 1.9912, "step": 14091 }, { "epoch": 2.68, "learning_rate": 2.1968350810295517e-06, "loss": 1.8151, "step": 14092 }, { "epoch": 2.68, "learning_rate": 2.196777883698761e-06, "loss": 2.8598, "step": 14093 }, { "epoch": 2.68, "learning_rate": 2.1967206863679697e-06, "loss": 1.973, "step": 14094 }, { "epoch": 2.68, "learning_rate": 2.196663489037178e-06, "loss": 3.0654, "step": 14095 }, { "epoch": 2.68, "learning_rate": 2.1966062917063873e-06, "loss": 3.2947, "step": 14096 }, { "epoch": 2.69, "learning_rate": 2.1965490943755956e-06, "loss": 2.3026, "step": 14097 }, { "epoch": 2.69, "learning_rate": 2.196491897044805e-06, "loss": 2.8227, "step": 14098 }, { "epoch": 2.69, "learning_rate": 2.1964346997140137e-06, "loss": 2.2086, "step": 14099 }, { "epoch": 2.69, "learning_rate": 2.196377502383222e-06, "loss": 3.2187, "step": 14100 }, { "epoch": 2.69, "learning_rate": 2.1963203050524312e-06, "loss": 2.4647, "step": 14101 }, { "epoch": 2.69, "learning_rate": 2.1962631077216396e-06, "loss": 2.8678, "step": 14102 }, { "epoch": 2.69, "learning_rate": 2.1962059103908484e-06, "loss": 3.0087, "step": 14103 }, { "epoch": 2.69, "learning_rate": 2.196148713060057e-06, "loss": 3.0825, "step": 14104 }, { "epoch": 2.69, "learning_rate": 2.196091515729266e-06, "loss": 2.3525, "step": 14105 }, { "epoch": 2.69, "learning_rate": 2.196034318398475e-06, "loss": 3.7609, "step": 14106 }, { "epoch": 2.69, "learning_rate": 2.1959771210676836e-06, "loss": 2.8529, "step": 14107 }, { "epoch": 2.69, "learning_rate": 2.1959199237368924e-06, "loss": 2.5035, "step": 14108 }, { "epoch": 2.69, "learning_rate": 2.195862726406101e-06, "loss": 1.8092, "step": 14109 }, { "epoch": 2.69, "learning_rate": 2.19580552907531e-06, "loss": 1.9571, "step": 14110 }, { "epoch": 2.69, "learning_rate": 2.1957483317445188e-06, "loss": 2.1277, "step": 14111 }, { "epoch": 2.69, "learning_rate": 2.195691134413727e-06, "loss": 2.7477, "step": 14112 }, { "epoch": 2.69, "learning_rate": 2.1956339370829363e-06, "loss": 3.5182, "step": 14113 }, { "epoch": 2.69, "learning_rate": 2.1955767397521447e-06, "loss": 2.9016, "step": 14114 }, { "epoch": 2.69, "learning_rate": 2.195519542421354e-06, "loss": 2.5203, "step": 14115 }, { "epoch": 2.69, "learning_rate": 2.1954623450905627e-06, "loss": 2.6995, "step": 14116 }, { "epoch": 2.69, "learning_rate": 2.195405147759771e-06, "loss": 2.3761, "step": 14117 }, { "epoch": 2.69, "learning_rate": 2.1953479504289803e-06, "loss": 3.5786, "step": 14118 }, { "epoch": 2.69, "learning_rate": 2.1952907530981887e-06, "loss": 2.6176, "step": 14119 }, { "epoch": 2.69, "learning_rate": 2.1952335557673975e-06, "loss": 2.6881, "step": 14120 }, { "epoch": 2.69, "learning_rate": 2.1951763584366063e-06, "loss": 2.5987, "step": 14121 }, { "epoch": 2.69, "learning_rate": 2.195119161105815e-06, "loss": 3.1464, "step": 14122 }, { "epoch": 2.69, "learning_rate": 2.195061963775024e-06, "loss": 2.294, "step": 14123 }, { "epoch": 2.69, "learning_rate": 2.1950047664442327e-06, "loss": 3.0504, "step": 14124 }, { "epoch": 2.69, "learning_rate": 2.1949475691134414e-06, "loss": 3.2482, "step": 14125 }, { "epoch": 2.69, "learning_rate": 2.1948903717826502e-06, "loss": 2.657, "step": 14126 }, { "epoch": 2.69, "learning_rate": 2.194833174451859e-06, "loss": 2.2348, "step": 14127 }, { "epoch": 2.69, "learning_rate": 2.194775977121068e-06, "loss": 2.9023, "step": 14128 }, { "epoch": 2.69, "learning_rate": 2.194718779790276e-06, "loss": 2.0805, "step": 14129 }, { "epoch": 2.69, "learning_rate": 2.1946615824594854e-06, "loss": 2.4937, "step": 14130 }, { "epoch": 2.69, "learning_rate": 2.1946043851286938e-06, "loss": 3.0528, "step": 14131 }, { "epoch": 2.69, "learning_rate": 2.194547187797903e-06, "loss": 1.4799, "step": 14132 }, { "epoch": 2.69, "learning_rate": 2.194489990467112e-06, "loss": 2.8582, "step": 14133 }, { "epoch": 2.69, "learning_rate": 2.19443279313632e-06, "loss": 2.5032, "step": 14134 }, { "epoch": 2.69, "learning_rate": 2.1943755958055294e-06, "loss": 2.6391, "step": 14135 }, { "epoch": 2.69, "learning_rate": 2.1943183984747378e-06, "loss": 2.1491, "step": 14136 }, { "epoch": 2.69, "learning_rate": 2.1942612011439466e-06, "loss": 1.5687, "step": 14137 }, { "epoch": 2.69, "learning_rate": 2.1942040038131558e-06, "loss": 2.6874, "step": 14138 }, { "epoch": 2.69, "learning_rate": 2.194146806482364e-06, "loss": 3.6975, "step": 14139 }, { "epoch": 2.69, "learning_rate": 2.194089609151573e-06, "loss": 3.1989, "step": 14140 }, { "epoch": 2.69, "learning_rate": 2.1940324118207817e-06, "loss": 3.24, "step": 14141 }, { "epoch": 2.69, "learning_rate": 2.1939752144899905e-06, "loss": 3.337, "step": 14142 }, { "epoch": 2.69, "learning_rate": 2.1939180171591993e-06, "loss": 2.0319, "step": 14143 }, { "epoch": 2.69, "learning_rate": 2.193860819828408e-06, "loss": 1.8297, "step": 14144 }, { "epoch": 2.69, "learning_rate": 2.193803622497617e-06, "loss": 2.5932, "step": 14145 }, { "epoch": 2.69, "learning_rate": 2.1937464251668253e-06, "loss": 2.9281, "step": 14146 }, { "epoch": 2.69, "learning_rate": 2.1936892278360345e-06, "loss": 2.3527, "step": 14147 }, { "epoch": 2.69, "learning_rate": 2.1936320305052433e-06, "loss": 2.0483, "step": 14148 }, { "epoch": 2.7, "learning_rate": 2.193574833174452e-06, "loss": 3.2762, "step": 14149 }, { "epoch": 2.7, "learning_rate": 2.193517635843661e-06, "loss": 2.1622, "step": 14150 }, { "epoch": 2.7, "learning_rate": 2.1934604385128692e-06, "loss": 2.113, "step": 14151 }, { "epoch": 2.7, "learning_rate": 2.1934032411820785e-06, "loss": 2.7888, "step": 14152 }, { "epoch": 2.7, "learning_rate": 2.193346043851287e-06, "loss": 3.3034, "step": 14153 }, { "epoch": 2.7, "learning_rate": 2.1932888465204956e-06, "loss": 2.3255, "step": 14154 }, { "epoch": 2.7, "learning_rate": 2.193231649189705e-06, "loss": 2.3982, "step": 14155 }, { "epoch": 2.7, "learning_rate": 2.193174451858913e-06, "loss": 3.5446, "step": 14156 }, { "epoch": 2.7, "learning_rate": 2.193117254528122e-06, "loss": 2.4032, "step": 14157 }, { "epoch": 2.7, "learning_rate": 2.193060057197331e-06, "loss": 3.1315, "step": 14158 }, { "epoch": 2.7, "learning_rate": 2.1930028598665396e-06, "loss": 1.9332, "step": 14159 }, { "epoch": 2.7, "learning_rate": 2.1929456625357484e-06, "loss": 2.8048, "step": 14160 }, { "epoch": 2.7, "learning_rate": 2.192888465204957e-06, "loss": 1.9976, "step": 14161 }, { "epoch": 2.7, "learning_rate": 2.192831267874166e-06, "loss": 3.1161, "step": 14162 }, { "epoch": 2.7, "learning_rate": 2.1927740705433743e-06, "loss": 2.9182, "step": 14163 }, { "epoch": 2.7, "learning_rate": 2.1927168732125836e-06, "loss": 3.2121, "step": 14164 }, { "epoch": 2.7, "learning_rate": 2.1926596758817924e-06, "loss": 1.9759, "step": 14165 }, { "epoch": 2.7, "learning_rate": 2.192602478551001e-06, "loss": 2.6334, "step": 14166 }, { "epoch": 2.7, "learning_rate": 2.19254528122021e-06, "loss": 3.0832, "step": 14167 }, { "epoch": 2.7, "learning_rate": 2.1924880838894183e-06, "loss": 3.028, "step": 14168 }, { "epoch": 2.7, "learning_rate": 2.1924308865586275e-06, "loss": 2.8312, "step": 14169 }, { "epoch": 2.7, "learning_rate": 2.192373689227836e-06, "loss": 2.7551, "step": 14170 }, { "epoch": 2.7, "learning_rate": 2.1923164918970447e-06, "loss": 2.2216, "step": 14171 }, { "epoch": 2.7, "learning_rate": 2.192259294566254e-06, "loss": 3.3457, "step": 14172 }, { "epoch": 2.7, "learning_rate": 2.1922020972354623e-06, "loss": 2.5744, "step": 14173 }, { "epoch": 2.7, "learning_rate": 2.192144899904671e-06, "loss": 3.1031, "step": 14174 }, { "epoch": 2.7, "learning_rate": 2.19208770257388e-06, "loss": 2.6294, "step": 14175 }, { "epoch": 2.7, "learning_rate": 2.1920305052430887e-06, "loss": 2.2209, "step": 14176 }, { "epoch": 2.7, "learning_rate": 2.1919733079122975e-06, "loss": 2.3471, "step": 14177 }, { "epoch": 2.7, "learning_rate": 2.1919161105815063e-06, "loss": 2.5728, "step": 14178 }, { "epoch": 2.7, "learning_rate": 2.191858913250715e-06, "loss": 2.0904, "step": 14179 }, { "epoch": 2.7, "learning_rate": 2.191801715919924e-06, "loss": 2.2905, "step": 14180 }, { "epoch": 2.7, "learning_rate": 2.1917445185891326e-06, "loss": 2.6833, "step": 14181 }, { "epoch": 2.7, "learning_rate": 2.1916873212583414e-06, "loss": 2.8333, "step": 14182 }, { "epoch": 2.7, "learning_rate": 2.1916301239275502e-06, "loss": 1.7, "step": 14183 }, { "epoch": 2.7, "learning_rate": 2.191572926596759e-06, "loss": 1.8525, "step": 14184 }, { "epoch": 2.7, "learning_rate": 2.1915157292659674e-06, "loss": 2.6361, "step": 14185 }, { "epoch": 2.7, "learning_rate": 2.1914585319351766e-06, "loss": 2.8016, "step": 14186 }, { "epoch": 2.7, "learning_rate": 2.1914013346043854e-06, "loss": 2.7494, "step": 14187 }, { "epoch": 2.7, "learning_rate": 2.1913441372735938e-06, "loss": 1.8385, "step": 14188 }, { "epoch": 2.7, "learning_rate": 2.191286939942803e-06, "loss": 2.8746, "step": 14189 }, { "epoch": 2.7, "learning_rate": 2.1912297426120114e-06, "loss": 3.1024, "step": 14190 }, { "epoch": 2.7, "learning_rate": 2.19117254528122e-06, "loss": 2.4289, "step": 14191 }, { "epoch": 2.7, "learning_rate": 2.191115347950429e-06, "loss": 2.3692, "step": 14192 }, { "epoch": 2.7, "learning_rate": 2.1910581506196377e-06, "loss": 3.4772, "step": 14193 }, { "epoch": 2.7, "learning_rate": 2.1910009532888465e-06, "loss": 1.98, "step": 14194 }, { "epoch": 2.7, "learning_rate": 2.1909437559580553e-06, "loss": 2.2833, "step": 14195 }, { "epoch": 2.7, "learning_rate": 2.190886558627264e-06, "loss": 3.3597, "step": 14196 }, { "epoch": 2.7, "learning_rate": 2.190829361296473e-06, "loss": 2.6829, "step": 14197 }, { "epoch": 2.7, "learning_rate": 2.1907721639656817e-06, "loss": 2.1679, "step": 14198 }, { "epoch": 2.7, "learning_rate": 2.1907149666348905e-06, "loss": 2.5104, "step": 14199 }, { "epoch": 2.7, "learning_rate": 2.1906577693040993e-06, "loss": 3.2256, "step": 14200 }, { "epoch": 2.7, "learning_rate": 2.190600571973308e-06, "loss": 2.493, "step": 14201 }, { "epoch": 2.71, "learning_rate": 2.1905433746425165e-06, "loss": 2.3841, "step": 14202 }, { "epoch": 2.71, "learning_rate": 2.1904861773117257e-06, "loss": 2.3147, "step": 14203 }, { "epoch": 2.71, "learning_rate": 2.1904289799809345e-06, "loss": 2.4793, "step": 14204 }, { "epoch": 2.71, "learning_rate": 2.190371782650143e-06, "loss": 2.1063, "step": 14205 }, { "epoch": 2.71, "learning_rate": 2.190314585319352e-06, "loss": 2.84, "step": 14206 }, { "epoch": 2.71, "learning_rate": 2.1902573879885604e-06, "loss": 1.9144, "step": 14207 }, { "epoch": 2.71, "learning_rate": 2.1902001906577692e-06, "loss": 2.8263, "step": 14208 }, { "epoch": 2.71, "learning_rate": 2.190142993326978e-06, "loss": 1.7509, "step": 14209 }, { "epoch": 2.71, "learning_rate": 2.190085795996187e-06, "loss": 2.7933, "step": 14210 }, { "epoch": 2.71, "learning_rate": 2.1900285986653956e-06, "loss": 2.9928, "step": 14211 }, { "epoch": 2.71, "learning_rate": 2.1899714013346044e-06, "loss": 2.0655, "step": 14212 }, { "epoch": 2.71, "learning_rate": 2.189914204003813e-06, "loss": 1.9133, "step": 14213 }, { "epoch": 2.71, "learning_rate": 2.189857006673022e-06, "loss": 1.949, "step": 14214 }, { "epoch": 2.71, "learning_rate": 2.1897998093422308e-06, "loss": 2.7591, "step": 14215 }, { "epoch": 2.71, "learning_rate": 2.1897426120114396e-06, "loss": 2.4144, "step": 14216 }, { "epoch": 2.71, "learning_rate": 2.1896854146806484e-06, "loss": 3.0552, "step": 14217 }, { "epoch": 2.71, "learning_rate": 2.189628217349857e-06, "loss": 2.3188, "step": 14218 }, { "epoch": 2.71, "learning_rate": 2.1895710200190655e-06, "loss": 2.5882, "step": 14219 }, { "epoch": 2.71, "learning_rate": 2.1895138226882747e-06, "loss": 2.5074, "step": 14220 }, { "epoch": 2.71, "learning_rate": 2.1894566253574835e-06, "loss": 3.6163, "step": 14221 }, { "epoch": 2.71, "learning_rate": 2.189399428026692e-06, "loss": 2.0154, "step": 14222 }, { "epoch": 2.71, "learning_rate": 2.189342230695901e-06, "loss": 2.3432, "step": 14223 }, { "epoch": 2.71, "learning_rate": 2.1892850333651095e-06, "loss": 2.6115, "step": 14224 }, { "epoch": 2.71, "learning_rate": 2.1892278360343183e-06, "loss": 2.2228, "step": 14225 }, { "epoch": 2.71, "learning_rate": 2.189170638703527e-06, "loss": 3.0082, "step": 14226 }, { "epoch": 2.71, "learning_rate": 2.189113441372736e-06, "loss": 2.8395, "step": 14227 }, { "epoch": 2.71, "learning_rate": 2.1890562440419447e-06, "loss": 1.7607, "step": 14228 }, { "epoch": 2.71, "learning_rate": 2.1889990467111535e-06, "loss": 1.8326, "step": 14229 }, { "epoch": 2.71, "learning_rate": 2.1889418493803623e-06, "loss": 1.8643, "step": 14230 }, { "epoch": 2.71, "learning_rate": 2.188884652049571e-06, "loss": 2.4749, "step": 14231 }, { "epoch": 2.71, "learning_rate": 2.18882745471878e-06, "loss": 2.7246, "step": 14232 }, { "epoch": 2.71, "learning_rate": 2.1887702573879886e-06, "loss": 2.5591, "step": 14233 }, { "epoch": 2.71, "learning_rate": 2.1887130600571974e-06, "loss": 3.0756, "step": 14234 }, { "epoch": 2.71, "learning_rate": 2.1886558627264062e-06, "loss": 2.5669, "step": 14235 }, { "epoch": 2.71, "learning_rate": 2.188598665395615e-06, "loss": 2.1132, "step": 14236 }, { "epoch": 2.71, "learning_rate": 2.188541468064824e-06, "loss": 3.3732, "step": 14237 }, { "epoch": 2.71, "learning_rate": 2.1884842707340326e-06, "loss": 2.3117, "step": 14238 }, { "epoch": 2.71, "learning_rate": 2.188427073403241e-06, "loss": 2.8347, "step": 14239 }, { "epoch": 2.71, "learning_rate": 2.18836987607245e-06, "loss": 2.3509, "step": 14240 }, { "epoch": 2.71, "learning_rate": 2.1883126787416586e-06, "loss": 1.7169, "step": 14241 }, { "epoch": 2.71, "learning_rate": 2.1882554814108674e-06, "loss": 2.1949, "step": 14242 }, { "epoch": 2.71, "learning_rate": 2.1881982840800766e-06, "loss": 2.0872, "step": 14243 }, { "epoch": 2.71, "learning_rate": 2.188141086749285e-06, "loss": 2.3418, "step": 14244 }, { "epoch": 2.71, "learning_rate": 2.188083889418494e-06, "loss": 2.1081, "step": 14245 }, { "epoch": 2.71, "learning_rate": 2.1880266920877025e-06, "loss": 2.7425, "step": 14246 }, { "epoch": 2.71, "learning_rate": 2.1879694947569113e-06, "loss": 2.809, "step": 14247 }, { "epoch": 2.71, "learning_rate": 2.18791229742612e-06, "loss": 2.6992, "step": 14248 }, { "epoch": 2.71, "learning_rate": 2.187855100095329e-06, "loss": 3.1073, "step": 14249 }, { "epoch": 2.71, "learning_rate": 2.1877979027645377e-06, "loss": 2.8127, "step": 14250 }, { "epoch": 2.71, "learning_rate": 2.1877407054337465e-06, "loss": 2.2586, "step": 14251 }, { "epoch": 2.71, "learning_rate": 2.1876835081029553e-06, "loss": 2.658, "step": 14252 }, { "epoch": 2.71, "learning_rate": 2.187626310772164e-06, "loss": 3.7651, "step": 14253 }, { "epoch": 2.72, "learning_rate": 2.187569113441373e-06, "loss": 2.0225, "step": 14254 }, { "epoch": 2.72, "learning_rate": 2.1875119161105817e-06, "loss": 2.6059, "step": 14255 }, { "epoch": 2.72, "learning_rate": 2.18745471877979e-06, "loss": 3.3095, "step": 14256 }, { "epoch": 2.72, "learning_rate": 2.1873975214489993e-06, "loss": 2.2524, "step": 14257 }, { "epoch": 2.72, "learning_rate": 2.1873403241182076e-06, "loss": 2.4515, "step": 14258 }, { "epoch": 2.72, "learning_rate": 2.1872831267874164e-06, "loss": 3.3529, "step": 14259 }, { "epoch": 2.72, "learning_rate": 2.1872259294566256e-06, "loss": 1.875, "step": 14260 }, { "epoch": 2.72, "learning_rate": 2.187168732125834e-06, "loss": 2.528, "step": 14261 }, { "epoch": 2.72, "learning_rate": 2.1871115347950432e-06, "loss": 2.348, "step": 14262 }, { "epoch": 2.72, "learning_rate": 2.1870543374642516e-06, "loss": 2.2773, "step": 14263 }, { "epoch": 2.72, "learning_rate": 2.1869971401334604e-06, "loss": 2.6421, "step": 14264 }, { "epoch": 2.72, "learning_rate": 2.186939942802669e-06, "loss": 2.9431, "step": 14265 }, { "epoch": 2.72, "learning_rate": 2.186882745471878e-06, "loss": 3.0432, "step": 14266 }, { "epoch": 2.72, "learning_rate": 2.1868255481410868e-06, "loss": 2.869, "step": 14267 }, { "epoch": 2.72, "learning_rate": 2.1867683508102956e-06, "loss": 2.6337, "step": 14268 }, { "epoch": 2.72, "learning_rate": 2.1867111534795044e-06, "loss": 2.5389, "step": 14269 }, { "epoch": 2.72, "learning_rate": 2.186653956148713e-06, "loss": 2.5943, "step": 14270 }, { "epoch": 2.72, "learning_rate": 2.186596758817922e-06, "loss": 2.0106, "step": 14271 }, { "epoch": 2.72, "learning_rate": 2.1865395614871308e-06, "loss": 2.2918, "step": 14272 }, { "epoch": 2.72, "learning_rate": 2.186482364156339e-06, "loss": 3.1025, "step": 14273 }, { "epoch": 2.72, "learning_rate": 2.1864251668255483e-06, "loss": 2.4209, "step": 14274 }, { "epoch": 2.72, "learning_rate": 2.186367969494757e-06, "loss": 3.0411, "step": 14275 }, { "epoch": 2.72, "learning_rate": 2.1863107721639655e-06, "loss": 2.2014, "step": 14276 }, { "epoch": 2.72, "learning_rate": 2.1862535748331747e-06, "loss": 3.3844, "step": 14277 }, { "epoch": 2.72, "learning_rate": 2.186196377502383e-06, "loss": 2.8376, "step": 14278 }, { "epoch": 2.72, "learning_rate": 2.1861391801715923e-06, "loss": 2.9296, "step": 14279 }, { "epoch": 2.72, "learning_rate": 2.1860819828408007e-06, "loss": 2.5145, "step": 14280 }, { "epoch": 2.72, "learning_rate": 2.1860247855100095e-06, "loss": 3.4745, "step": 14281 }, { "epoch": 2.72, "learning_rate": 2.1859675881792187e-06, "loss": 2.772, "step": 14282 }, { "epoch": 2.72, "learning_rate": 2.185910390848427e-06, "loss": 3.1661, "step": 14283 }, { "epoch": 2.72, "learning_rate": 2.185853193517636e-06, "loss": 2.8272, "step": 14284 }, { "epoch": 2.72, "learning_rate": 2.1857959961868446e-06, "loss": 2.0762, "step": 14285 }, { "epoch": 2.72, "learning_rate": 2.1857387988560534e-06, "loss": 3.4945, "step": 14286 }, { "epoch": 2.72, "learning_rate": 2.1856816015252622e-06, "loss": 2.5326, "step": 14287 }, { "epoch": 2.72, "learning_rate": 2.185624404194471e-06, "loss": 3.0363, "step": 14288 }, { "epoch": 2.72, "learning_rate": 2.18556720686368e-06, "loss": 2.1744, "step": 14289 }, { "epoch": 2.72, "learning_rate": 2.185510009532888e-06, "loss": 2.7939, "step": 14290 }, { "epoch": 2.72, "learning_rate": 2.1854528122020974e-06, "loss": 3.0023, "step": 14291 }, { "epoch": 2.72, "learning_rate": 2.185395614871306e-06, "loss": 2.1741, "step": 14292 }, { "epoch": 2.72, "learning_rate": 2.1853384175405146e-06, "loss": 2.3725, "step": 14293 }, { "epoch": 2.72, "learning_rate": 2.185281220209724e-06, "loss": 2.5647, "step": 14294 }, { "epoch": 2.72, "learning_rate": 2.185224022878932e-06, "loss": 3.1266, "step": 14295 }, { "epoch": 2.72, "learning_rate": 2.1851668255481414e-06, "loss": 2.5419, "step": 14296 }, { "epoch": 2.72, "learning_rate": 2.1851096282173498e-06, "loss": 3.9162, "step": 14297 }, { "epoch": 2.72, "learning_rate": 2.1850524308865585e-06, "loss": 3.0677, "step": 14298 }, { "epoch": 2.72, "learning_rate": 2.1849952335557678e-06, "loss": 2.6804, "step": 14299 }, { "epoch": 2.72, "learning_rate": 2.184938036224976e-06, "loss": 2.2877, "step": 14300 }, { "epoch": 2.72, "learning_rate": 2.184880838894185e-06, "loss": 3.1299, "step": 14301 }, { "epoch": 2.72, "learning_rate": 2.1848236415633937e-06, "loss": 1.5338, "step": 14302 }, { "epoch": 2.72, "learning_rate": 2.1847664442326025e-06, "loss": 1.9416, "step": 14303 }, { "epoch": 2.72, "learning_rate": 2.1847092469018113e-06, "loss": 2.2022, "step": 14304 }, { "epoch": 2.72, "learning_rate": 2.18465204957102e-06, "loss": 1.9066, "step": 14305 }, { "epoch": 2.72, "learning_rate": 2.184594852240229e-06, "loss": 3.3468, "step": 14306 }, { "epoch": 2.73, "learning_rate": 2.1845376549094373e-06, "loss": 2.7407, "step": 14307 }, { "epoch": 2.73, "learning_rate": 2.1844804575786465e-06, "loss": 2.977, "step": 14308 }, { "epoch": 2.73, "learning_rate": 2.1844232602478553e-06, "loss": 3.4849, "step": 14309 }, { "epoch": 2.73, "learning_rate": 2.184366062917064e-06, "loss": 2.9423, "step": 14310 }, { "epoch": 2.73, "learning_rate": 2.184308865586273e-06, "loss": 3.0538, "step": 14311 }, { "epoch": 2.73, "learning_rate": 2.1842516682554812e-06, "loss": 2.2307, "step": 14312 }, { "epoch": 2.73, "learning_rate": 2.1841944709246905e-06, "loss": 2.894, "step": 14313 }, { "epoch": 2.73, "learning_rate": 2.184137273593899e-06, "loss": 2.7887, "step": 14314 }, { "epoch": 2.73, "learning_rate": 2.1840800762631076e-06, "loss": 2.8941, "step": 14315 }, { "epoch": 2.73, "learning_rate": 2.184022878932317e-06, "loss": 1.7289, "step": 14316 }, { "epoch": 2.73, "learning_rate": 2.183965681601525e-06, "loss": 2.5923, "step": 14317 }, { "epoch": 2.73, "learning_rate": 2.183908484270734e-06, "loss": 1.6778, "step": 14318 }, { "epoch": 2.73, "learning_rate": 2.183851286939943e-06, "loss": 3.0511, "step": 14319 }, { "epoch": 2.73, "learning_rate": 2.1837940896091516e-06, "loss": 2.4729, "step": 14320 }, { "epoch": 2.73, "learning_rate": 2.1837368922783604e-06, "loss": 2.1148, "step": 14321 }, { "epoch": 2.73, "learning_rate": 2.183679694947569e-06, "loss": 2.3073, "step": 14322 }, { "epoch": 2.73, "learning_rate": 2.183622497616778e-06, "loss": 2.5564, "step": 14323 }, { "epoch": 2.73, "learning_rate": 2.1835653002859868e-06, "loss": 2.5956, "step": 14324 }, { "epoch": 2.73, "learning_rate": 2.1835081029551956e-06, "loss": 1.6547, "step": 14325 }, { "epoch": 2.73, "learning_rate": 2.1834509056244043e-06, "loss": 2.4063, "step": 14326 }, { "epoch": 2.73, "learning_rate": 2.183393708293613e-06, "loss": 2.9802, "step": 14327 }, { "epoch": 2.73, "learning_rate": 2.183336510962822e-06, "loss": 2.7673, "step": 14328 }, { "epoch": 2.73, "learning_rate": 2.1832793136320303e-06, "loss": 2.6409, "step": 14329 }, { "epoch": 2.73, "learning_rate": 2.1832221163012395e-06, "loss": 3.0117, "step": 14330 }, { "epoch": 2.73, "learning_rate": 2.1831649189704483e-06, "loss": 3.0967, "step": 14331 }, { "epoch": 2.73, "learning_rate": 2.1831077216396567e-06, "loss": 2.6908, "step": 14332 }, { "epoch": 2.73, "learning_rate": 2.183050524308866e-06, "loss": 2.7812, "step": 14333 }, { "epoch": 2.73, "learning_rate": 2.1829933269780743e-06, "loss": 2.0489, "step": 14334 }, { "epoch": 2.73, "learning_rate": 2.182936129647283e-06, "loss": 2.9069, "step": 14335 }, { "epoch": 2.73, "learning_rate": 2.182878932316492e-06, "loss": 2.6289, "step": 14336 }, { "epoch": 2.73, "learning_rate": 2.1828217349857007e-06, "loss": 2.2878, "step": 14337 }, { "epoch": 2.73, "learning_rate": 2.1827645376549094e-06, "loss": 2.8518, "step": 14338 }, { "epoch": 2.73, "learning_rate": 2.1827073403241182e-06, "loss": 2.412, "step": 14339 }, { "epoch": 2.73, "learning_rate": 2.182650142993327e-06, "loss": 1.5828, "step": 14340 }, { "epoch": 2.73, "learning_rate": 2.182592945662536e-06, "loss": 2.3984, "step": 14341 }, { "epoch": 2.73, "learning_rate": 2.1825357483317446e-06, "loss": 2.9775, "step": 14342 }, { "epoch": 2.73, "learning_rate": 2.1824785510009534e-06, "loss": 3.3261, "step": 14343 }, { "epoch": 2.73, "learning_rate": 2.1824213536701622e-06, "loss": 2.4217, "step": 14344 }, { "epoch": 2.73, "learning_rate": 2.182364156339371e-06, "loss": 2.7173, "step": 14345 }, { "epoch": 2.73, "learning_rate": 2.1823069590085794e-06, "loss": 2.9217, "step": 14346 }, { "epoch": 2.73, "learning_rate": 2.1822497616777886e-06, "loss": 2.529, "step": 14347 }, { "epoch": 2.73, "learning_rate": 2.1821925643469974e-06, "loss": 2.1666, "step": 14348 }, { "epoch": 2.73, "learning_rate": 2.1821353670162058e-06, "loss": 3.7454, "step": 14349 }, { "epoch": 2.73, "learning_rate": 2.182078169685415e-06, "loss": 2.0676, "step": 14350 }, { "epoch": 2.73, "learning_rate": 2.1820209723546233e-06, "loss": 2.2413, "step": 14351 }, { "epoch": 2.73, "learning_rate": 2.181963775023832e-06, "loss": 1.5333, "step": 14352 }, { "epoch": 2.73, "learning_rate": 2.181906577693041e-06, "loss": 3.8329, "step": 14353 }, { "epoch": 2.73, "learning_rate": 2.1818493803622497e-06, "loss": 2.2532, "step": 14354 }, { "epoch": 2.73, "learning_rate": 2.1817921830314585e-06, "loss": 3.4435, "step": 14355 }, { "epoch": 2.73, "learning_rate": 2.1817349857006673e-06, "loss": 2.35, "step": 14356 }, { "epoch": 2.73, "learning_rate": 2.181677788369876e-06, "loss": 1.58, "step": 14357 }, { "epoch": 2.73, "learning_rate": 2.181620591039085e-06, "loss": 2.0051, "step": 14358 }, { "epoch": 2.74, "learning_rate": 2.1815633937082937e-06, "loss": 2.3295, "step": 14359 }, { "epoch": 2.74, "learning_rate": 2.1815061963775025e-06, "loss": 3.2326, "step": 14360 }, { "epoch": 2.74, "learning_rate": 2.1814489990467113e-06, "loss": 1.7923, "step": 14361 }, { "epoch": 2.74, "learning_rate": 2.18139180171592e-06, "loss": 3.6501, "step": 14362 }, { "epoch": 2.74, "learning_rate": 2.181334604385129e-06, "loss": 2.6788, "step": 14363 }, { "epoch": 2.74, "learning_rate": 2.1812774070543377e-06, "loss": 2.1727, "step": 14364 }, { "epoch": 2.74, "learning_rate": 2.1812202097235465e-06, "loss": 1.6926, "step": 14365 }, { "epoch": 2.74, "learning_rate": 2.181163012392755e-06, "loss": 2.7168, "step": 14366 }, { "epoch": 2.74, "learning_rate": 2.181105815061964e-06, "loss": 2.4833, "step": 14367 }, { "epoch": 2.74, "learning_rate": 2.1810486177311724e-06, "loss": 3.2466, "step": 14368 }, { "epoch": 2.74, "learning_rate": 2.180991420400381e-06, "loss": 3.1019, "step": 14369 }, { "epoch": 2.74, "learning_rate": 2.1809342230695904e-06, "loss": 1.4698, "step": 14370 }, { "epoch": 2.74, "learning_rate": 2.180877025738799e-06, "loss": 2.0956, "step": 14371 }, { "epoch": 2.74, "learning_rate": 2.1808198284080076e-06, "loss": 2.895, "step": 14372 }, { "epoch": 2.74, "learning_rate": 2.1807626310772164e-06, "loss": 1.2798, "step": 14373 }, { "epoch": 2.74, "learning_rate": 2.180705433746425e-06, "loss": 2.9437, "step": 14374 }, { "epoch": 2.74, "learning_rate": 2.180648236415634e-06, "loss": 1.6486, "step": 14375 }, { "epoch": 2.74, "learning_rate": 2.1805910390848428e-06, "loss": 2.5229, "step": 14376 }, { "epoch": 2.74, "learning_rate": 2.1805338417540516e-06, "loss": 2.5267, "step": 14377 }, { "epoch": 2.74, "learning_rate": 2.1804766444232604e-06, "loss": 3.2333, "step": 14378 }, { "epoch": 2.74, "learning_rate": 2.180419447092469e-06, "loss": 2.4251, "step": 14379 }, { "epoch": 2.74, "learning_rate": 2.180362249761678e-06, "loss": 2.224, "step": 14380 }, { "epoch": 2.74, "learning_rate": 2.1803050524308867e-06, "loss": 3.1375, "step": 14381 }, { "epoch": 2.74, "learning_rate": 2.1802478551000955e-06, "loss": 2.9272, "step": 14382 }, { "epoch": 2.74, "learning_rate": 2.180190657769304e-06, "loss": 3.2949, "step": 14383 }, { "epoch": 2.74, "learning_rate": 2.180133460438513e-06, "loss": 2.4813, "step": 14384 }, { "epoch": 2.74, "learning_rate": 2.1800762631077215e-06, "loss": 1.8815, "step": 14385 }, { "epoch": 2.74, "learning_rate": 2.1800190657769303e-06, "loss": 2.207, "step": 14386 }, { "epoch": 2.74, "learning_rate": 2.1799618684461395e-06, "loss": 2.9364, "step": 14387 }, { "epoch": 2.74, "learning_rate": 2.179904671115348e-06, "loss": 2.3698, "step": 14388 }, { "epoch": 2.74, "learning_rate": 2.1798474737845567e-06, "loss": 2.3788, "step": 14389 }, { "epoch": 2.74, "learning_rate": 2.1797902764537655e-06, "loss": 3.0443, "step": 14390 }, { "epoch": 2.74, "learning_rate": 2.1797330791229743e-06, "loss": 2.9125, "step": 14391 }, { "epoch": 2.74, "learning_rate": 2.179675881792183e-06, "loss": 3.1309, "step": 14392 }, { "epoch": 2.74, "learning_rate": 2.179618684461392e-06, "loss": 1.7136, "step": 14393 }, { "epoch": 2.74, "learning_rate": 2.1795614871306006e-06, "loss": 2.4229, "step": 14394 }, { "epoch": 2.74, "learning_rate": 2.1795042897998094e-06, "loss": 2.2857, "step": 14395 }, { "epoch": 2.74, "learning_rate": 2.1794470924690182e-06, "loss": 2.8991, "step": 14396 }, { "epoch": 2.74, "learning_rate": 2.179389895138227e-06, "loss": 2.5048, "step": 14397 }, { "epoch": 2.74, "learning_rate": 2.179332697807436e-06, "loss": 2.4528, "step": 14398 }, { "epoch": 2.74, "learning_rate": 2.1792755004766446e-06, "loss": 2.5838, "step": 14399 }, { "epoch": 2.74, "learning_rate": 2.179218303145853e-06, "loss": 2.6608, "step": 14400 }, { "epoch": 2.74, "learning_rate": 2.179161105815062e-06, "loss": 2.7877, "step": 14401 }, { "epoch": 2.74, "learning_rate": 2.1791039084842706e-06, "loss": 2.4805, "step": 14402 }, { "epoch": 2.74, "learning_rate": 2.1790467111534794e-06, "loss": 2.7531, "step": 14403 }, { "epoch": 2.74, "learning_rate": 2.1789895138226886e-06, "loss": 2.4909, "step": 14404 }, { "epoch": 2.74, "learning_rate": 2.178932316491897e-06, "loss": 3.0442, "step": 14405 }, { "epoch": 2.74, "learning_rate": 2.1788751191611057e-06, "loss": 2.3708, "step": 14406 }, { "epoch": 2.74, "learning_rate": 2.1788179218303145e-06, "loss": 1.5505, "step": 14407 }, { "epoch": 2.74, "learning_rate": 2.1787607244995233e-06, "loss": 2.7491, "step": 14408 }, { "epoch": 2.74, "learning_rate": 2.1787035271687325e-06, "loss": 3.286, "step": 14409 }, { "epoch": 2.74, "learning_rate": 2.178646329837941e-06, "loss": 2.6102, "step": 14410 }, { "epoch": 2.74, "learning_rate": 2.1785891325071497e-06, "loss": 3.1838, "step": 14411 }, { "epoch": 2.75, "learning_rate": 2.1785319351763585e-06, "loss": 3.4517, "step": 14412 }, { "epoch": 2.75, "learning_rate": 2.1784747378455673e-06, "loss": 2.3533, "step": 14413 }, { "epoch": 2.75, "learning_rate": 2.178417540514776e-06, "loss": 2.2986, "step": 14414 }, { "epoch": 2.75, "learning_rate": 2.178360343183985e-06, "loss": 2.7015, "step": 14415 }, { "epoch": 2.75, "learning_rate": 2.1783031458531937e-06, "loss": 1.3898, "step": 14416 }, { "epoch": 2.75, "learning_rate": 2.178245948522402e-06, "loss": 2.0404, "step": 14417 }, { "epoch": 2.75, "learning_rate": 2.1781887511916113e-06, "loss": 2.9211, "step": 14418 }, { "epoch": 2.75, "learning_rate": 2.17813155386082e-06, "loss": 3.1008, "step": 14419 }, { "epoch": 2.75, "learning_rate": 2.1780743565300284e-06, "loss": 2.9915, "step": 14420 }, { "epoch": 2.75, "learning_rate": 2.1780171591992376e-06, "loss": 2.1958, "step": 14421 }, { "epoch": 2.75, "learning_rate": 2.177959961868446e-06, "loss": 2.8939, "step": 14422 }, { "epoch": 2.75, "learning_rate": 2.177902764537655e-06, "loss": 2.6171, "step": 14423 }, { "epoch": 2.75, "learning_rate": 2.1778455672068636e-06, "loss": 2.1612, "step": 14424 }, { "epoch": 2.75, "learning_rate": 2.1777883698760724e-06, "loss": 2.0619, "step": 14425 }, { "epoch": 2.75, "learning_rate": 2.1777311725452816e-06, "loss": 3.1364, "step": 14426 }, { "epoch": 2.75, "learning_rate": 2.17767397521449e-06, "loss": 1.9811, "step": 14427 }, { "epoch": 2.75, "learning_rate": 2.1776167778836988e-06, "loss": 2.2557, "step": 14428 }, { "epoch": 2.75, "learning_rate": 2.1775595805529076e-06, "loss": 2.5971, "step": 14429 }, { "epoch": 2.75, "learning_rate": 2.1775023832221164e-06, "loss": 3.1494, "step": 14430 }, { "epoch": 2.75, "learning_rate": 2.177445185891325e-06, "loss": 3.5078, "step": 14431 }, { "epoch": 2.75, "learning_rate": 2.177387988560534e-06, "loss": 1.9702, "step": 14432 }, { "epoch": 2.75, "learning_rate": 2.1773307912297427e-06, "loss": 2.0901, "step": 14433 }, { "epoch": 2.75, "learning_rate": 2.177273593898951e-06, "loss": 2.8516, "step": 14434 }, { "epoch": 2.75, "learning_rate": 2.1772163965681603e-06, "loss": 2.9362, "step": 14435 }, { "epoch": 2.75, "learning_rate": 2.177159199237369e-06, "loss": 2.5642, "step": 14436 }, { "epoch": 2.75, "learning_rate": 2.1771020019065775e-06, "loss": 3.0073, "step": 14437 }, { "epoch": 2.75, "learning_rate": 2.1770448045757867e-06, "loss": 2.6967, "step": 14438 }, { "epoch": 2.75, "learning_rate": 2.176987607244995e-06, "loss": 2.9403, "step": 14439 }, { "epoch": 2.75, "learning_rate": 2.1769304099142043e-06, "loss": 2.8384, "step": 14440 }, { "epoch": 2.75, "learning_rate": 2.1768732125834127e-06, "loss": 2.0497, "step": 14441 }, { "epoch": 2.75, "learning_rate": 2.1768160152526215e-06, "loss": 2.1951, "step": 14442 }, { "epoch": 2.75, "learning_rate": 2.1767588179218307e-06, "loss": 2.0459, "step": 14443 }, { "epoch": 2.75, "learning_rate": 2.176701620591039e-06, "loss": 1.9522, "step": 14444 }, { "epoch": 2.75, "learning_rate": 2.176644423260248e-06, "loss": 1.5178, "step": 14445 }, { "epoch": 2.75, "learning_rate": 2.1765872259294566e-06, "loss": 2.3383, "step": 14446 }, { "epoch": 2.75, "learning_rate": 2.1765300285986654e-06, "loss": 2.884, "step": 14447 }, { "epoch": 2.75, "learning_rate": 2.1764728312678742e-06, "loss": 2.1982, "step": 14448 }, { "epoch": 2.75, "learning_rate": 2.176415633937083e-06, "loss": 3.1264, "step": 14449 }, { "epoch": 2.75, "learning_rate": 2.176358436606292e-06, "loss": 4.0768, "step": 14450 }, { "epoch": 2.75, "learning_rate": 2.1763012392755006e-06, "loss": 2.5046, "step": 14451 }, { "epoch": 2.75, "learning_rate": 2.1762440419447094e-06, "loss": 2.6887, "step": 14452 }, { "epoch": 2.75, "learning_rate": 2.176186844613918e-06, "loss": 2.7741, "step": 14453 }, { "epoch": 2.75, "learning_rate": 2.1761296472831266e-06, "loss": 2.3268, "step": 14454 }, { "epoch": 2.75, "learning_rate": 2.1760724499523358e-06, "loss": 3.0596, "step": 14455 }, { "epoch": 2.75, "learning_rate": 2.176015252621544e-06, "loss": 2.4411, "step": 14456 }, { "epoch": 2.75, "learning_rate": 2.1759580552907534e-06, "loss": 1.0745, "step": 14457 }, { "epoch": 2.75, "learning_rate": 2.175900857959962e-06, "loss": 2.9929, "step": 14458 }, { "epoch": 2.75, "learning_rate": 2.1758436606291705e-06, "loss": 2.0478, "step": 14459 }, { "epoch": 2.75, "learning_rate": 2.1757864632983798e-06, "loss": 3.6459, "step": 14460 }, { "epoch": 2.75, "learning_rate": 2.175729265967588e-06, "loss": 1.64, "step": 14461 }, { "epoch": 2.75, "learning_rate": 2.175672068636797e-06, "loss": 2.6579, "step": 14462 }, { "epoch": 2.75, "learning_rate": 2.1756148713060057e-06, "loss": 1.7397, "step": 14463 }, { "epoch": 2.76, "learning_rate": 2.1755576739752145e-06, "loss": 2.6004, "step": 14464 }, { "epoch": 2.76, "learning_rate": 2.1755004766444233e-06, "loss": 2.0799, "step": 14465 }, { "epoch": 2.76, "learning_rate": 2.175443279313632e-06, "loss": 2.1539, "step": 14466 }, { "epoch": 2.76, "learning_rate": 2.175386081982841e-06, "loss": 2.6593, "step": 14467 }, { "epoch": 2.76, "learning_rate": 2.1753288846520497e-06, "loss": 3.3042, "step": 14468 }, { "epoch": 2.76, "learning_rate": 2.1752716873212585e-06, "loss": 3.9191, "step": 14469 }, { "epoch": 2.76, "learning_rate": 2.1752144899904673e-06, "loss": 2.7457, "step": 14470 }, { "epoch": 2.76, "learning_rate": 2.1751572926596756e-06, "loss": 2.7997, "step": 14471 }, { "epoch": 2.76, "learning_rate": 2.175100095328885e-06, "loss": 2.438, "step": 14472 }, { "epoch": 2.76, "learning_rate": 2.1750428979980932e-06, "loss": 1.6343, "step": 14473 }, { "epoch": 2.76, "learning_rate": 2.1749857006673024e-06, "loss": 2.5695, "step": 14474 }, { "epoch": 2.76, "learning_rate": 2.1749285033365112e-06, "loss": 3.8928, "step": 14475 }, { "epoch": 2.76, "learning_rate": 2.1748713060057196e-06, "loss": 1.5189, "step": 14476 }, { "epoch": 2.76, "learning_rate": 2.174814108674929e-06, "loss": 2.2873, "step": 14477 }, { "epoch": 2.76, "learning_rate": 2.174756911344137e-06, "loss": 2.561, "step": 14478 }, { "epoch": 2.76, "learning_rate": 2.174699714013346e-06, "loss": 2.2665, "step": 14479 }, { "epoch": 2.76, "learning_rate": 2.1746425166825548e-06, "loss": 2.8646, "step": 14480 }, { "epoch": 2.76, "learning_rate": 2.1745853193517636e-06, "loss": 3.0431, "step": 14481 }, { "epoch": 2.76, "learning_rate": 2.1745281220209724e-06, "loss": 3.7989, "step": 14482 }, { "epoch": 2.76, "learning_rate": 2.174470924690181e-06, "loss": 3.1836, "step": 14483 }, { "epoch": 2.76, "learning_rate": 2.17441372735939e-06, "loss": 2.4823, "step": 14484 }, { "epoch": 2.76, "learning_rate": 2.1743565300285988e-06, "loss": 2.5423, "step": 14485 }, { "epoch": 2.76, "learning_rate": 2.1742993326978075e-06, "loss": 2.0291, "step": 14486 }, { "epoch": 2.76, "learning_rate": 2.1742421353670163e-06, "loss": 3.38, "step": 14487 }, { "epoch": 2.76, "learning_rate": 2.1741849380362247e-06, "loss": 2.554, "step": 14488 }, { "epoch": 2.76, "learning_rate": 2.174127740705434e-06, "loss": 2.1715, "step": 14489 }, { "epoch": 2.76, "learning_rate": 2.1740705433746423e-06, "loss": 2.4549, "step": 14490 }, { "epoch": 2.76, "learning_rate": 2.1740133460438515e-06, "loss": 2.6823, "step": 14491 }, { "epoch": 2.76, "learning_rate": 2.1739561487130603e-06, "loss": 2.5061, "step": 14492 }, { "epoch": 2.76, "learning_rate": 2.1738989513822687e-06, "loss": 2.1418, "step": 14493 }, { "epoch": 2.76, "learning_rate": 2.173841754051478e-06, "loss": 2.5417, "step": 14494 }, { "epoch": 2.76, "learning_rate": 2.1737845567206863e-06, "loss": 1.3446, "step": 14495 }, { "epoch": 2.76, "learning_rate": 2.173727359389895e-06, "loss": 1.9429, "step": 14496 }, { "epoch": 2.76, "learning_rate": 2.1736701620591043e-06, "loss": 2.0558, "step": 14497 }, { "epoch": 2.76, "learning_rate": 2.1736129647283126e-06, "loss": 2.3478, "step": 14498 }, { "epoch": 2.76, "learning_rate": 2.1735557673975214e-06, "loss": 2.4454, "step": 14499 }, { "epoch": 2.76, "learning_rate": 2.1734985700667302e-06, "loss": 3.384, "step": 14500 }, { "epoch": 2.76, "learning_rate": 2.173441372735939e-06, "loss": 3.0698, "step": 14501 }, { "epoch": 2.76, "learning_rate": 2.173384175405148e-06, "loss": 2.1974, "step": 14502 }, { "epoch": 2.76, "learning_rate": 2.1733269780743566e-06, "loss": 2.5785, "step": 14503 }, { "epoch": 2.76, "learning_rate": 2.1732697807435654e-06, "loss": 3.1761, "step": 14504 }, { "epoch": 2.76, "learning_rate": 2.173212583412774e-06, "loss": 2.8084, "step": 14505 }, { "epoch": 2.76, "learning_rate": 2.173155386081983e-06, "loss": 3.0518, "step": 14506 }, { "epoch": 2.76, "learning_rate": 2.173098188751192e-06, "loss": 2.2377, "step": 14507 }, { "epoch": 2.76, "learning_rate": 2.1730409914204006e-06, "loss": 3.6246, "step": 14508 }, { "epoch": 2.76, "learning_rate": 2.1729837940896094e-06, "loss": 2.2492, "step": 14509 }, { "epoch": 2.76, "learning_rate": 2.1729265967588178e-06, "loss": 2.4354, "step": 14510 }, { "epoch": 2.76, "learning_rate": 2.172869399428027e-06, "loss": 2.6378, "step": 14511 }, { "epoch": 2.76, "learning_rate": 2.1728122020972353e-06, "loss": 3.1812, "step": 14512 }, { "epoch": 2.76, "learning_rate": 2.172755004766444e-06, "loss": 3.0933, "step": 14513 }, { "epoch": 2.76, "learning_rate": 2.1726978074356533e-06, "loss": 3.0062, "step": 14514 }, { "epoch": 2.76, "learning_rate": 2.1726406101048617e-06, "loss": 3.2413, "step": 14515 }, { "epoch": 2.76, "learning_rate": 2.1725834127740705e-06, "loss": 2.0485, "step": 14516 }, { "epoch": 2.77, "learning_rate": 2.1725262154432793e-06, "loss": 1.9783, "step": 14517 }, { "epoch": 2.77, "learning_rate": 2.172469018112488e-06, "loss": 2.4596, "step": 14518 }, { "epoch": 2.77, "learning_rate": 2.172411820781697e-06, "loss": 3.2516, "step": 14519 }, { "epoch": 2.77, "learning_rate": 2.1723546234509057e-06, "loss": 2.451, "step": 14520 }, { "epoch": 2.77, "learning_rate": 2.1722974261201145e-06, "loss": 3.238, "step": 14521 }, { "epoch": 2.77, "learning_rate": 2.1722402287893233e-06, "loss": 2.4623, "step": 14522 }, { "epoch": 2.77, "learning_rate": 2.172183031458532e-06, "loss": 3.2401, "step": 14523 }, { "epoch": 2.77, "learning_rate": 2.172125834127741e-06, "loss": 2.3153, "step": 14524 }, { "epoch": 2.77, "learning_rate": 2.1720686367969497e-06, "loss": 3.5359, "step": 14525 }, { "epoch": 2.77, "learning_rate": 2.1720114394661585e-06, "loss": 2.5524, "step": 14526 }, { "epoch": 2.77, "learning_rate": 2.171954242135367e-06, "loss": 1.8216, "step": 14527 }, { "epoch": 2.77, "learning_rate": 2.171897044804576e-06, "loss": 2.0496, "step": 14528 }, { "epoch": 2.77, "learning_rate": 2.1718398474737844e-06, "loss": 3.5021, "step": 14529 }, { "epoch": 2.77, "learning_rate": 2.171782650142993e-06, "loss": 3.014, "step": 14530 }, { "epoch": 2.77, "learning_rate": 2.1717254528122024e-06, "loss": 2.2352, "step": 14531 }, { "epoch": 2.77, "learning_rate": 2.171668255481411e-06, "loss": 2.9034, "step": 14532 }, { "epoch": 2.77, "learning_rate": 2.1716110581506196e-06, "loss": 2.5169, "step": 14533 }, { "epoch": 2.77, "learning_rate": 2.1715538608198284e-06, "loss": 2.743, "step": 14534 }, { "epoch": 2.77, "learning_rate": 2.171496663489037e-06, "loss": 1.9381, "step": 14535 }, { "epoch": 2.77, "learning_rate": 2.171439466158246e-06, "loss": 2.8942, "step": 14536 }, { "epoch": 2.77, "learning_rate": 2.1713822688274548e-06, "loss": 1.458, "step": 14537 }, { "epoch": 2.77, "learning_rate": 2.1713250714966636e-06, "loss": 2.294, "step": 14538 }, { "epoch": 2.77, "learning_rate": 2.1712678741658723e-06, "loss": 2.917, "step": 14539 }, { "epoch": 2.77, "learning_rate": 2.171210676835081e-06, "loss": 2.6349, "step": 14540 }, { "epoch": 2.77, "learning_rate": 2.17115347950429e-06, "loss": 1.7427, "step": 14541 }, { "epoch": 2.77, "learning_rate": 2.1710962821734987e-06, "loss": 2.8434, "step": 14542 }, { "epoch": 2.77, "learning_rate": 2.1710390848427075e-06, "loss": 3.2654, "step": 14543 }, { "epoch": 2.77, "learning_rate": 2.170981887511916e-06, "loss": 2.1891, "step": 14544 }, { "epoch": 2.77, "learning_rate": 2.170924690181125e-06, "loss": 3.43, "step": 14545 }, { "epoch": 2.77, "learning_rate": 2.170867492850334e-06, "loss": 3.506, "step": 14546 }, { "epoch": 2.77, "learning_rate": 2.1708102955195423e-06, "loss": 2.4508, "step": 14547 }, { "epoch": 2.77, "learning_rate": 2.1707530981887515e-06, "loss": 3.9018, "step": 14548 }, { "epoch": 2.77, "learning_rate": 2.17069590085796e-06, "loss": 2.7783, "step": 14549 }, { "epoch": 2.77, "learning_rate": 2.1706387035271687e-06, "loss": 2.5222, "step": 14550 }, { "epoch": 2.77, "learning_rate": 2.1705815061963775e-06, "loss": 3.1756, "step": 14551 }, { "epoch": 2.77, "learning_rate": 2.1705243088655862e-06, "loss": 2.0798, "step": 14552 }, { "epoch": 2.77, "learning_rate": 2.170467111534795e-06, "loss": 1.9148, "step": 14553 }, { "epoch": 2.77, "learning_rate": 2.170409914204004e-06, "loss": 2.5389, "step": 14554 }, { "epoch": 2.77, "learning_rate": 2.1703527168732126e-06, "loss": 1.6625, "step": 14555 }, { "epoch": 2.77, "learning_rate": 2.1702955195424214e-06, "loss": 2.9514, "step": 14556 }, { "epoch": 2.77, "learning_rate": 2.1702383222116302e-06, "loss": 2.1842, "step": 14557 }, { "epoch": 2.77, "learning_rate": 2.170181124880839e-06, "loss": 2.2793, "step": 14558 }, { "epoch": 2.77, "learning_rate": 2.170123927550048e-06, "loss": 2.8118, "step": 14559 }, { "epoch": 2.77, "learning_rate": 2.1700667302192566e-06, "loss": 2.3445, "step": 14560 }, { "epoch": 2.77, "learning_rate": 2.170009532888465e-06, "loss": 2.8924, "step": 14561 }, { "epoch": 2.77, "learning_rate": 2.169952335557674e-06, "loss": 2.4435, "step": 14562 }, { "epoch": 2.77, "learning_rate": 2.169895138226883e-06, "loss": 3.4829, "step": 14563 }, { "epoch": 2.77, "learning_rate": 2.1698379408960913e-06, "loss": 2.5606, "step": 14564 }, { "epoch": 2.77, "learning_rate": 2.1697807435653006e-06, "loss": 2.43, "step": 14565 }, { "epoch": 2.77, "learning_rate": 2.169723546234509e-06, "loss": 3.3786, "step": 14566 }, { "epoch": 2.77, "learning_rate": 2.1696663489037177e-06, "loss": 3.1697, "step": 14567 }, { "epoch": 2.77, "learning_rate": 2.1696091515729265e-06, "loss": 3.2849, "step": 14568 }, { "epoch": 2.78, "learning_rate": 2.1695519542421353e-06, "loss": 2.3677, "step": 14569 }, { "epoch": 2.78, "learning_rate": 2.1694947569113445e-06, "loss": 2.0096, "step": 14570 }, { "epoch": 2.78, "learning_rate": 2.169437559580553e-06, "loss": 2.6964, "step": 14571 }, { "epoch": 2.78, "learning_rate": 2.1693803622497617e-06, "loss": 1.7612, "step": 14572 }, { "epoch": 2.78, "learning_rate": 2.1693231649189705e-06, "loss": 4.2974, "step": 14573 }, { "epoch": 2.78, "learning_rate": 2.1692659675881793e-06, "loss": 1.9018, "step": 14574 }, { "epoch": 2.78, "learning_rate": 2.169208770257388e-06, "loss": 2.9181, "step": 14575 }, { "epoch": 2.78, "learning_rate": 2.169151572926597e-06, "loss": 1.5742, "step": 14576 }, { "epoch": 2.78, "learning_rate": 2.1690943755958057e-06, "loss": 3.0144, "step": 14577 }, { "epoch": 2.78, "learning_rate": 2.169037178265014e-06, "loss": 2.3182, "step": 14578 }, { "epoch": 2.78, "learning_rate": 2.1689799809342233e-06, "loss": 2.4263, "step": 14579 }, { "epoch": 2.78, "learning_rate": 2.168922783603432e-06, "loss": 2.8529, "step": 14580 }, { "epoch": 2.78, "learning_rate": 2.1688655862726404e-06, "loss": 2.4568, "step": 14581 }, { "epoch": 2.78, "learning_rate": 2.1688083889418496e-06, "loss": 3.506, "step": 14582 }, { "epoch": 2.78, "learning_rate": 2.168751191611058e-06, "loss": 2.7614, "step": 14583 }, { "epoch": 2.78, "learning_rate": 2.168693994280267e-06, "loss": 2.1407, "step": 14584 }, { "epoch": 2.78, "learning_rate": 2.168636796949476e-06, "loss": 2.2095, "step": 14585 }, { "epoch": 2.78, "learning_rate": 2.1685795996186844e-06, "loss": 2.7246, "step": 14586 }, { "epoch": 2.78, "learning_rate": 2.1685224022878936e-06, "loss": 3.1038, "step": 14587 }, { "epoch": 2.78, "learning_rate": 2.168465204957102e-06, "loss": 2.4426, "step": 14588 }, { "epoch": 2.78, "learning_rate": 2.1684080076263108e-06, "loss": 3.3435, "step": 14589 }, { "epoch": 2.78, "learning_rate": 2.1683508102955196e-06, "loss": 1.8173, "step": 14590 }, { "epoch": 2.78, "learning_rate": 2.1682936129647284e-06, "loss": 2.5888, "step": 14591 }, { "epoch": 2.78, "learning_rate": 2.168236415633937e-06, "loss": 3.2133, "step": 14592 }, { "epoch": 2.78, "learning_rate": 2.168179218303146e-06, "loss": 2.8748, "step": 14593 }, { "epoch": 2.78, "learning_rate": 2.1681220209723547e-06, "loss": 2.6356, "step": 14594 }, { "epoch": 2.78, "learning_rate": 2.1680648236415635e-06, "loss": 3.3222, "step": 14595 }, { "epoch": 2.78, "learning_rate": 2.1680076263107723e-06, "loss": 2.892, "step": 14596 }, { "epoch": 2.78, "learning_rate": 2.167950428979981e-06, "loss": 1.5159, "step": 14597 }, { "epoch": 2.78, "learning_rate": 2.1678932316491895e-06, "loss": 1.3978, "step": 14598 }, { "epoch": 2.78, "learning_rate": 2.1678360343183987e-06, "loss": 2.1974, "step": 14599 }, { "epoch": 2.78, "learning_rate": 2.167778836987607e-06, "loss": 2.0935, "step": 14600 }, { "epoch": 2.78, "learning_rate": 2.167721639656816e-06, "loss": 2.1117, "step": 14601 }, { "epoch": 2.78, "learning_rate": 2.167664442326025e-06, "loss": 2.1595, "step": 14602 }, { "epoch": 2.78, "learning_rate": 2.1676072449952335e-06, "loss": 2.3641, "step": 14603 }, { "epoch": 2.78, "learning_rate": 2.1675500476644427e-06, "loss": 3.5626, "step": 14604 }, { "epoch": 2.78, "learning_rate": 2.167492850333651e-06, "loss": 2.8709, "step": 14605 }, { "epoch": 2.78, "learning_rate": 2.16743565300286e-06, "loss": 2.0491, "step": 14606 }, { "epoch": 2.78, "learning_rate": 2.1673784556720686e-06, "loss": 1.6789, "step": 14607 }, { "epoch": 2.78, "learning_rate": 2.1673212583412774e-06, "loss": 2.1804, "step": 14608 }, { "epoch": 2.78, "learning_rate": 2.1672640610104862e-06, "loss": 2.1532, "step": 14609 }, { "epoch": 2.78, "learning_rate": 2.167206863679695e-06, "loss": 3.0258, "step": 14610 }, { "epoch": 2.78, "learning_rate": 2.167149666348904e-06, "loss": 2.7501, "step": 14611 }, { "epoch": 2.78, "learning_rate": 2.1670924690181126e-06, "loss": 2.6022, "step": 14612 }, { "epoch": 2.78, "learning_rate": 2.1670352716873214e-06, "loss": 1.9032, "step": 14613 }, { "epoch": 2.78, "learning_rate": 2.16697807435653e-06, "loss": 2.8526, "step": 14614 }, { "epoch": 2.78, "learning_rate": 2.1669208770257386e-06, "loss": 2.13, "step": 14615 }, { "epoch": 2.78, "learning_rate": 2.1668636796949478e-06, "loss": 1.7574, "step": 14616 }, { "epoch": 2.78, "learning_rate": 2.166806482364156e-06, "loss": 2.5982, "step": 14617 }, { "epoch": 2.78, "learning_rate": 2.166749285033365e-06, "loss": 2.7191, "step": 14618 }, { "epoch": 2.78, "learning_rate": 2.166692087702574e-06, "loss": 2.8009, "step": 14619 }, { "epoch": 2.78, "learning_rate": 2.1666348903717825e-06, "loss": 2.4486, "step": 14620 }, { "epoch": 2.78, "learning_rate": 2.1665776930409917e-06, "loss": 3.472, "step": 14621 }, { "epoch": 2.79, "learning_rate": 2.1665204957102e-06, "loss": 2.3938, "step": 14622 }, { "epoch": 2.79, "learning_rate": 2.166463298379409e-06, "loss": 2.8784, "step": 14623 }, { "epoch": 2.79, "learning_rate": 2.1664061010486177e-06, "loss": 3.4984, "step": 14624 }, { "epoch": 2.79, "learning_rate": 2.1663489037178265e-06, "loss": 2.4425, "step": 14625 }, { "epoch": 2.79, "learning_rate": 2.1662917063870353e-06, "loss": 2.7215, "step": 14626 }, { "epoch": 2.79, "learning_rate": 2.166234509056244e-06, "loss": 2.4867, "step": 14627 }, { "epoch": 2.79, "learning_rate": 2.166177311725453e-06, "loss": 2.3486, "step": 14628 }, { "epoch": 2.79, "learning_rate": 2.1661201143946617e-06, "loss": 2.4458, "step": 14629 }, { "epoch": 2.79, "learning_rate": 2.1660629170638705e-06, "loss": 3.3776, "step": 14630 }, { "epoch": 2.79, "learning_rate": 2.1660057197330793e-06, "loss": 2.3334, "step": 14631 }, { "epoch": 2.79, "learning_rate": 2.1659485224022876e-06, "loss": 2.3448, "step": 14632 }, { "epoch": 2.79, "learning_rate": 2.165891325071497e-06, "loss": 2.1311, "step": 14633 }, { "epoch": 2.79, "learning_rate": 2.1658341277407056e-06, "loss": 1.7158, "step": 14634 }, { "epoch": 2.79, "learning_rate": 2.1657769304099144e-06, "loss": 2.1802, "step": 14635 }, { "epoch": 2.79, "learning_rate": 2.1657197330791232e-06, "loss": 1.9145, "step": 14636 }, { "epoch": 2.79, "learning_rate": 2.1656625357483316e-06, "loss": 2.7993, "step": 14637 }, { "epoch": 2.79, "learning_rate": 2.165605338417541e-06, "loss": 2.3042, "step": 14638 }, { "epoch": 2.79, "learning_rate": 2.165548141086749e-06, "loss": 2.7353, "step": 14639 }, { "epoch": 2.79, "learning_rate": 2.165490943755958e-06, "loss": 1.8886, "step": 14640 }, { "epoch": 2.79, "learning_rate": 2.165433746425167e-06, "loss": 2.7056, "step": 14641 }, { "epoch": 2.79, "learning_rate": 2.1653765490943756e-06, "loss": 2.6323, "step": 14642 }, { "epoch": 2.79, "learning_rate": 2.1653193517635844e-06, "loss": 1.7559, "step": 14643 }, { "epoch": 2.79, "learning_rate": 2.165262154432793e-06, "loss": 4.1289, "step": 14644 }, { "epoch": 2.79, "learning_rate": 2.165204957102002e-06, "loss": 2.8341, "step": 14645 }, { "epoch": 2.79, "learning_rate": 2.1651477597712107e-06, "loss": 2.9461, "step": 14646 }, { "epoch": 2.79, "learning_rate": 2.1650905624404195e-06, "loss": 4.3481, "step": 14647 }, { "epoch": 2.79, "learning_rate": 2.1650333651096283e-06, "loss": 2.9209, "step": 14648 }, { "epoch": 2.79, "learning_rate": 2.1649761677788367e-06, "loss": 2.6575, "step": 14649 }, { "epoch": 2.79, "learning_rate": 2.164918970448046e-06, "loss": 1.7789, "step": 14650 }, { "epoch": 2.79, "learning_rate": 2.1648617731172547e-06, "loss": 1.9401, "step": 14651 }, { "epoch": 2.79, "learning_rate": 2.1648045757864635e-06, "loss": 2.2819, "step": 14652 }, { "epoch": 2.79, "learning_rate": 2.1647473784556723e-06, "loss": 2.2507, "step": 14653 }, { "epoch": 2.79, "learning_rate": 2.1646901811248807e-06, "loss": 1.6646, "step": 14654 }, { "epoch": 2.79, "learning_rate": 2.16463298379409e-06, "loss": 2.3423, "step": 14655 }, { "epoch": 2.79, "learning_rate": 2.1645757864632983e-06, "loss": 2.573, "step": 14656 }, { "epoch": 2.79, "learning_rate": 2.164518589132507e-06, "loss": 2.4543, "step": 14657 }, { "epoch": 2.79, "learning_rate": 2.1644613918017163e-06, "loss": 3.0079, "step": 14658 }, { "epoch": 2.79, "learning_rate": 2.1644041944709246e-06, "loss": 2.3995, "step": 14659 }, { "epoch": 2.79, "learning_rate": 2.1643469971401334e-06, "loss": 3.6783, "step": 14660 }, { "epoch": 2.79, "learning_rate": 2.1642897998093422e-06, "loss": 2.1324, "step": 14661 }, { "epoch": 2.79, "learning_rate": 2.164232602478551e-06, "loss": 3.5041, "step": 14662 }, { "epoch": 2.79, "learning_rate": 2.16417540514776e-06, "loss": 2.1619, "step": 14663 }, { "epoch": 2.79, "learning_rate": 2.1641182078169686e-06, "loss": 2.3922, "step": 14664 }, { "epoch": 2.79, "learning_rate": 2.1640610104861774e-06, "loss": 2.0847, "step": 14665 }, { "epoch": 2.79, "learning_rate": 2.1640038131553858e-06, "loss": 2.9138, "step": 14666 }, { "epoch": 2.79, "learning_rate": 2.163946615824595e-06, "loss": 3.1437, "step": 14667 }, { "epoch": 2.79, "learning_rate": 2.1638894184938038e-06, "loss": 1.8934, "step": 14668 }, { "epoch": 2.79, "learning_rate": 2.1638322211630126e-06, "loss": 2.2519, "step": 14669 }, { "epoch": 2.79, "learning_rate": 2.1637750238322214e-06, "loss": 2.1198, "step": 14670 }, { "epoch": 2.79, "learning_rate": 2.1637178265014297e-06, "loss": 2.3349, "step": 14671 }, { "epoch": 2.79, "learning_rate": 2.163660629170639e-06, "loss": 1.693, "step": 14672 }, { "epoch": 2.79, "learning_rate": 2.1636034318398478e-06, "loss": 2.0855, "step": 14673 }, { "epoch": 2.8, "learning_rate": 2.163546234509056e-06, "loss": 3.0148, "step": 14674 }, { "epoch": 2.8, "learning_rate": 2.1634890371782653e-06, "loss": 1.869, "step": 14675 }, { "epoch": 2.8, "learning_rate": 2.1634318398474737e-06, "loss": 2.4577, "step": 14676 }, { "epoch": 2.8, "learning_rate": 2.1633746425166825e-06, "loss": 2.2165, "step": 14677 }, { "epoch": 2.8, "learning_rate": 2.1633174451858913e-06, "loss": 2.6165, "step": 14678 }, { "epoch": 2.8, "learning_rate": 2.1632602478551e-06, "loss": 2.9346, "step": 14679 }, { "epoch": 2.8, "learning_rate": 2.163203050524309e-06, "loss": 2.4423, "step": 14680 }, { "epoch": 2.8, "learning_rate": 2.1631458531935177e-06, "loss": 2.5587, "step": 14681 }, { "epoch": 2.8, "learning_rate": 2.1630886558627265e-06, "loss": 2.7163, "step": 14682 }, { "epoch": 2.8, "learning_rate": 2.1630314585319353e-06, "loss": 2.9588, "step": 14683 }, { "epoch": 2.8, "learning_rate": 2.162974261201144e-06, "loss": 2.9073, "step": 14684 }, { "epoch": 2.8, "learning_rate": 2.162917063870353e-06, "loss": 3.01, "step": 14685 }, { "epoch": 2.8, "learning_rate": 2.1628598665395617e-06, "loss": 2.4653, "step": 14686 }, { "epoch": 2.8, "learning_rate": 2.1628026692087704e-06, "loss": 2.854, "step": 14687 }, { "epoch": 2.8, "learning_rate": 2.162745471877979e-06, "loss": 2.2233, "step": 14688 }, { "epoch": 2.8, "learning_rate": 2.162688274547188e-06, "loss": 2.2961, "step": 14689 }, { "epoch": 2.8, "learning_rate": 2.162631077216397e-06, "loss": 2.4025, "step": 14690 }, { "epoch": 2.8, "learning_rate": 2.162573879885605e-06, "loss": 2.1409, "step": 14691 }, { "epoch": 2.8, "learning_rate": 2.1625166825548144e-06, "loss": 3.5509, "step": 14692 }, { "epoch": 2.8, "learning_rate": 2.1624594852240228e-06, "loss": 3.0306, "step": 14693 }, { "epoch": 2.8, "learning_rate": 2.1624022878932316e-06, "loss": 2.2377, "step": 14694 }, { "epoch": 2.8, "learning_rate": 2.1623450905624404e-06, "loss": 3.2937, "step": 14695 }, { "epoch": 2.8, "learning_rate": 2.162287893231649e-06, "loss": 3.0267, "step": 14696 }, { "epoch": 2.8, "learning_rate": 2.162230695900858e-06, "loss": 2.6644, "step": 14697 }, { "epoch": 2.8, "learning_rate": 2.1621734985700668e-06, "loss": 2.1044, "step": 14698 }, { "epoch": 2.8, "learning_rate": 2.1621163012392755e-06, "loss": 2.522, "step": 14699 }, { "epoch": 2.8, "learning_rate": 2.1620591039084843e-06, "loss": 1.7011, "step": 14700 }, { "epoch": 2.8, "learning_rate": 2.162001906577693e-06, "loss": 3.2826, "step": 14701 }, { "epoch": 2.8, "learning_rate": 2.161944709246902e-06, "loss": 1.8782, "step": 14702 }, { "epoch": 2.8, "learning_rate": 2.1618875119161107e-06, "loss": 2.8194, "step": 14703 }, { "epoch": 2.8, "learning_rate": 2.1618303145853195e-06, "loss": 2.7355, "step": 14704 }, { "epoch": 2.8, "learning_rate": 2.161773117254528e-06, "loss": 2.4949, "step": 14705 }, { "epoch": 2.8, "learning_rate": 2.161715919923737e-06, "loss": 2.6618, "step": 14706 }, { "epoch": 2.8, "learning_rate": 2.161658722592946e-06, "loss": 2.9666, "step": 14707 }, { "epoch": 2.8, "learning_rate": 2.1616015252621543e-06, "loss": 3.6132, "step": 14708 }, { "epoch": 2.8, "learning_rate": 2.1615443279313635e-06, "loss": 2.6586, "step": 14709 }, { "epoch": 2.8, "learning_rate": 2.161487130600572e-06, "loss": 2.8972, "step": 14710 }, { "epoch": 2.8, "learning_rate": 2.1614299332697806e-06, "loss": 3.0758, "step": 14711 }, { "epoch": 2.8, "learning_rate": 2.1613727359389894e-06, "loss": 3.2001, "step": 14712 }, { "epoch": 2.8, "learning_rate": 2.1613155386081982e-06, "loss": 1.9188, "step": 14713 }, { "epoch": 2.8, "learning_rate": 2.161258341277407e-06, "loss": 2.1175, "step": 14714 }, { "epoch": 2.8, "learning_rate": 2.161201143946616e-06, "loss": 3.0894, "step": 14715 }, { "epoch": 2.8, "learning_rate": 2.1611439466158246e-06, "loss": 3.0865, "step": 14716 }, { "epoch": 2.8, "learning_rate": 2.1610867492850334e-06, "loss": 2.342, "step": 14717 }, { "epoch": 2.8, "learning_rate": 2.161029551954242e-06, "loss": 3.1141, "step": 14718 }, { "epoch": 2.8, "learning_rate": 2.160972354623451e-06, "loss": 2.7348, "step": 14719 }, { "epoch": 2.8, "learning_rate": 2.16091515729266e-06, "loss": 2.7362, "step": 14720 }, { "epoch": 2.8, "learning_rate": 2.1608579599618686e-06, "loss": 2.3648, "step": 14721 }, { "epoch": 2.8, "learning_rate": 2.1608007626310774e-06, "loss": 2.433, "step": 14722 }, { "epoch": 2.8, "learning_rate": 2.160743565300286e-06, "loss": 3.2589, "step": 14723 }, { "epoch": 2.8, "learning_rate": 2.160686367969495e-06, "loss": 2.9644, "step": 14724 }, { "epoch": 2.8, "learning_rate": 2.1606291706387033e-06, "loss": 2.2529, "step": 14725 }, { "epoch": 2.8, "learning_rate": 2.1605719733079126e-06, "loss": 3.2884, "step": 14726 }, { "epoch": 2.81, "learning_rate": 2.160514775977121e-06, "loss": 2.7135, "step": 14727 }, { "epoch": 2.81, "learning_rate": 2.1604575786463297e-06, "loss": 2.6051, "step": 14728 }, { "epoch": 2.81, "learning_rate": 2.160400381315539e-06, "loss": 2.9495, "step": 14729 }, { "epoch": 2.81, "learning_rate": 2.1603431839847473e-06, "loss": 1.8147, "step": 14730 }, { "epoch": 2.81, "learning_rate": 2.160285986653956e-06, "loss": 2.5657, "step": 14731 }, { "epoch": 2.81, "learning_rate": 2.160228789323165e-06, "loss": 3.7807, "step": 14732 }, { "epoch": 2.81, "learning_rate": 2.1601715919923737e-06, "loss": 1.6372, "step": 14733 }, { "epoch": 2.81, "learning_rate": 2.1601143946615825e-06, "loss": 2.9312, "step": 14734 }, { "epoch": 2.81, "learning_rate": 2.1600571973307913e-06, "loss": 2.9343, "step": 14735 }, { "epoch": 2.81, "learning_rate": 2.16e-06, "loss": 2.5808, "step": 14736 }, { "epoch": 2.81, "learning_rate": 2.159942802669209e-06, "loss": 2.4049, "step": 14737 }, { "epoch": 2.81, "learning_rate": 2.1598856053384177e-06, "loss": 1.8958, "step": 14738 }, { "epoch": 2.81, "learning_rate": 2.1598284080076265e-06, "loss": 3.3806, "step": 14739 }, { "epoch": 2.81, "learning_rate": 2.1597712106768352e-06, "loss": 1.9486, "step": 14740 }, { "epoch": 2.81, "learning_rate": 2.159714013346044e-06, "loss": 2.4881, "step": 14741 }, { "epoch": 2.81, "learning_rate": 2.1596568160152524e-06, "loss": 3.0217, "step": 14742 }, { "epoch": 2.81, "learning_rate": 2.1595996186844616e-06, "loss": 2.7261, "step": 14743 }, { "epoch": 2.81, "learning_rate": 2.15954242135367e-06, "loss": 2.1658, "step": 14744 }, { "epoch": 2.81, "learning_rate": 2.159485224022879e-06, "loss": 2.0182, "step": 14745 }, { "epoch": 2.81, "learning_rate": 2.159428026692088e-06, "loss": 2.9762, "step": 14746 }, { "epoch": 2.81, "learning_rate": 2.1593708293612964e-06, "loss": 2.9678, "step": 14747 }, { "epoch": 2.81, "learning_rate": 2.159313632030505e-06, "loss": 2.4296, "step": 14748 }, { "epoch": 2.81, "learning_rate": 2.159256434699714e-06, "loss": 3.4293, "step": 14749 }, { "epoch": 2.81, "learning_rate": 2.1591992373689228e-06, "loss": 2.6686, "step": 14750 }, { "epoch": 2.81, "learning_rate": 2.1591420400381316e-06, "loss": 2.9125, "step": 14751 }, { "epoch": 2.81, "learning_rate": 2.1590848427073403e-06, "loss": 1.9127, "step": 14752 }, { "epoch": 2.81, "learning_rate": 2.159027645376549e-06, "loss": 2.7116, "step": 14753 }, { "epoch": 2.81, "learning_rate": 2.158970448045758e-06, "loss": 3.4455, "step": 14754 }, { "epoch": 2.81, "learning_rate": 2.1589132507149667e-06, "loss": 1.9283, "step": 14755 }, { "epoch": 2.81, "learning_rate": 2.1588560533841755e-06, "loss": 2.4065, "step": 14756 }, { "epoch": 2.81, "learning_rate": 2.1587988560533843e-06, "loss": 2.8988, "step": 14757 }, { "epoch": 2.81, "learning_rate": 2.158741658722593e-06, "loss": 2.4362, "step": 14758 }, { "epoch": 2.81, "learning_rate": 2.1586844613918015e-06, "loss": 2.0777, "step": 14759 }, { "epoch": 2.81, "learning_rate": 2.1586272640610107e-06, "loss": 2.1243, "step": 14760 }, { "epoch": 2.81, "learning_rate": 2.1585700667302195e-06, "loss": 2.1808, "step": 14761 }, { "epoch": 2.81, "learning_rate": 2.158512869399428e-06, "loss": 2.239, "step": 14762 }, { "epoch": 2.81, "learning_rate": 2.158455672068637e-06, "loss": 3.2815, "step": 14763 }, { "epoch": 2.81, "learning_rate": 2.1583984747378455e-06, "loss": 2.4317, "step": 14764 }, { "epoch": 2.81, "learning_rate": 2.1583412774070547e-06, "loss": 2.9764, "step": 14765 }, { "epoch": 2.81, "learning_rate": 2.158284080076263e-06, "loss": 2.1404, "step": 14766 }, { "epoch": 2.81, "learning_rate": 2.158226882745472e-06, "loss": 3.5507, "step": 14767 }, { "epoch": 2.81, "learning_rate": 2.158169685414681e-06, "loss": 2.0067, "step": 14768 }, { "epoch": 2.81, "learning_rate": 2.1581124880838894e-06, "loss": 2.7716, "step": 14769 }, { "epoch": 2.81, "learning_rate": 2.1580552907530982e-06, "loss": 2.3962, "step": 14770 }, { "epoch": 2.81, "learning_rate": 2.157998093422307e-06, "loss": 2.6025, "step": 14771 }, { "epoch": 2.81, "learning_rate": 2.157940896091516e-06, "loss": 1.9498, "step": 14772 }, { "epoch": 2.81, "learning_rate": 2.1578836987607246e-06, "loss": 3.3932, "step": 14773 }, { "epoch": 2.81, "learning_rate": 2.1578265014299334e-06, "loss": 2.8489, "step": 14774 }, { "epoch": 2.81, "learning_rate": 2.157769304099142e-06, "loss": 2.3681, "step": 14775 }, { "epoch": 2.81, "learning_rate": 2.1577121067683506e-06, "loss": 2.6371, "step": 14776 }, { "epoch": 2.81, "learning_rate": 2.1576549094375598e-06, "loss": 2.2118, "step": 14777 }, { "epoch": 2.81, "learning_rate": 2.1575977121067686e-06, "loss": 2.7059, "step": 14778 }, { "epoch": 2.82, "learning_rate": 2.157540514775977e-06, "loss": 3.4969, "step": 14779 }, { "epoch": 2.82, "learning_rate": 2.157483317445186e-06, "loss": 3.2553, "step": 14780 }, { "epoch": 2.82, "learning_rate": 2.1574261201143945e-06, "loss": 3.1831, "step": 14781 }, { "epoch": 2.82, "learning_rate": 2.1573689227836037e-06, "loss": 2.8898, "step": 14782 }, { "epoch": 2.82, "learning_rate": 2.157311725452812e-06, "loss": 1.7689, "step": 14783 }, { "epoch": 2.82, "learning_rate": 2.157254528122021e-06, "loss": 3.887, "step": 14784 }, { "epoch": 2.82, "learning_rate": 2.15719733079123e-06, "loss": 2.8258, "step": 14785 }, { "epoch": 2.82, "learning_rate": 2.1571401334604385e-06, "loss": 2.9849, "step": 14786 }, { "epoch": 2.82, "learning_rate": 2.1570829361296473e-06, "loss": 2.7106, "step": 14787 }, { "epoch": 2.82, "learning_rate": 2.157025738798856e-06, "loss": 2.9328, "step": 14788 }, { "epoch": 2.82, "learning_rate": 2.156968541468065e-06, "loss": 2.0959, "step": 14789 }, { "epoch": 2.82, "learning_rate": 2.1569113441372737e-06, "loss": 2.8343, "step": 14790 }, { "epoch": 2.82, "learning_rate": 2.1568541468064825e-06, "loss": 3.046, "step": 14791 }, { "epoch": 2.82, "learning_rate": 2.1567969494756913e-06, "loss": 2.6299, "step": 14792 }, { "epoch": 2.82, "learning_rate": 2.1567397521448996e-06, "loss": 1.7859, "step": 14793 }, { "epoch": 2.82, "learning_rate": 2.156682554814109e-06, "loss": 2.5315, "step": 14794 }, { "epoch": 2.82, "learning_rate": 2.1566253574833176e-06, "loss": 2.1171, "step": 14795 }, { "epoch": 2.82, "learning_rate": 2.156568160152526e-06, "loss": 2.5719, "step": 14796 }, { "epoch": 2.82, "learning_rate": 2.1565109628217352e-06, "loss": 3.3623, "step": 14797 }, { "epoch": 2.82, "learning_rate": 2.1564537654909436e-06, "loss": 3.2163, "step": 14798 }, { "epoch": 2.82, "learning_rate": 2.156396568160153e-06, "loss": 2.0888, "step": 14799 }, { "epoch": 2.82, "learning_rate": 2.156339370829361e-06, "loss": 1.6095, "step": 14800 }, { "epoch": 2.82, "learning_rate": 2.15628217349857e-06, "loss": 2.6175, "step": 14801 }, { "epoch": 2.82, "learning_rate": 2.156224976167779e-06, "loss": 1.5752, "step": 14802 }, { "epoch": 2.82, "learning_rate": 2.1561677788369876e-06, "loss": 2.5206, "step": 14803 }, { "epoch": 2.82, "learning_rate": 2.1561105815061964e-06, "loss": 2.4926, "step": 14804 }, { "epoch": 2.82, "learning_rate": 2.156053384175405e-06, "loss": 3.3137, "step": 14805 }, { "epoch": 2.82, "learning_rate": 2.155996186844614e-06, "loss": 3.1734, "step": 14806 }, { "epoch": 2.82, "learning_rate": 2.1559389895138227e-06, "loss": 2.7274, "step": 14807 }, { "epoch": 2.82, "learning_rate": 2.1558817921830315e-06, "loss": 2.7065, "step": 14808 }, { "epoch": 2.82, "learning_rate": 2.1558245948522403e-06, "loss": 2.9004, "step": 14809 }, { "epoch": 2.82, "learning_rate": 2.155767397521449e-06, "loss": 2.6231, "step": 14810 }, { "epoch": 2.82, "learning_rate": 2.155710200190658e-06, "loss": 1.8468, "step": 14811 }, { "epoch": 2.82, "learning_rate": 2.1556530028598667e-06, "loss": 2.8196, "step": 14812 }, { "epoch": 2.82, "learning_rate": 2.155595805529075e-06, "loss": 3.1154, "step": 14813 }, { "epoch": 2.82, "learning_rate": 2.1555386081982843e-06, "loss": 2.0933, "step": 14814 }, { "epoch": 2.82, "learning_rate": 2.1554814108674927e-06, "loss": 2.7389, "step": 14815 }, { "epoch": 2.82, "learning_rate": 2.155424213536702e-06, "loss": 2.5678, "step": 14816 }, { "epoch": 2.82, "learning_rate": 2.1553670162059107e-06, "loss": 3.2599, "step": 14817 }, { "epoch": 2.82, "learning_rate": 2.155309818875119e-06, "loss": 2.7282, "step": 14818 }, { "epoch": 2.82, "learning_rate": 2.1552526215443283e-06, "loss": 1.7304, "step": 14819 }, { "epoch": 2.82, "learning_rate": 2.1551954242135366e-06, "loss": 3.0837, "step": 14820 }, { "epoch": 2.82, "learning_rate": 2.1551382268827454e-06, "loss": 2.2661, "step": 14821 }, { "epoch": 2.82, "learning_rate": 2.1550810295519542e-06, "loss": 3.3687, "step": 14822 }, { "epoch": 2.82, "learning_rate": 2.155023832221163e-06, "loss": 3.0481, "step": 14823 }, { "epoch": 2.82, "learning_rate": 2.154966634890372e-06, "loss": 2.3779, "step": 14824 }, { "epoch": 2.82, "learning_rate": 2.1549094375595806e-06, "loss": 2.3409, "step": 14825 }, { "epoch": 2.82, "learning_rate": 2.1548522402287894e-06, "loss": 2.5856, "step": 14826 }, { "epoch": 2.82, "learning_rate": 2.154795042897998e-06, "loss": 3.1252, "step": 14827 }, { "epoch": 2.82, "learning_rate": 2.154737845567207e-06, "loss": 2.7151, "step": 14828 }, { "epoch": 2.82, "learning_rate": 2.1546806482364158e-06, "loss": 4.3298, "step": 14829 }, { "epoch": 2.82, "learning_rate": 2.154623450905624e-06, "loss": 3.4754, "step": 14830 }, { "epoch": 2.82, "learning_rate": 2.1545662535748334e-06, "loss": 2.3223, "step": 14831 }, { "epoch": 2.83, "learning_rate": 2.1545090562440417e-06, "loss": 2.5639, "step": 14832 }, { "epoch": 2.83, "learning_rate": 2.154451858913251e-06, "loss": 2.2528, "step": 14833 }, { "epoch": 2.83, "learning_rate": 2.1543946615824597e-06, "loss": 2.1571, "step": 14834 }, { "epoch": 2.83, "learning_rate": 2.154337464251668e-06, "loss": 2.674, "step": 14835 }, { "epoch": 2.83, "learning_rate": 2.1542802669208773e-06, "loss": 2.2443, "step": 14836 }, { "epoch": 2.83, "learning_rate": 2.1542230695900857e-06, "loss": 2.7127, "step": 14837 }, { "epoch": 2.83, "learning_rate": 2.1541658722592945e-06, "loss": 2.6845, "step": 14838 }, { "epoch": 2.83, "learning_rate": 2.1541086749285033e-06, "loss": 2.6709, "step": 14839 }, { "epoch": 2.83, "learning_rate": 2.154051477597712e-06, "loss": 2.5633, "step": 14840 }, { "epoch": 2.83, "learning_rate": 2.153994280266921e-06, "loss": 2.5361, "step": 14841 }, { "epoch": 2.83, "learning_rate": 2.1539370829361297e-06, "loss": 1.7331, "step": 14842 }, { "epoch": 2.83, "learning_rate": 2.1538798856053385e-06, "loss": 3.1669, "step": 14843 }, { "epoch": 2.83, "learning_rate": 2.1538226882745473e-06, "loss": 2.7576, "step": 14844 }, { "epoch": 2.83, "learning_rate": 2.153765490943756e-06, "loss": 2.1096, "step": 14845 }, { "epoch": 2.83, "learning_rate": 2.153708293612965e-06, "loss": 1.7269, "step": 14846 }, { "epoch": 2.83, "learning_rate": 2.1536510962821736e-06, "loss": 2.5635, "step": 14847 }, { "epoch": 2.83, "learning_rate": 2.1535938989513824e-06, "loss": 3.1826, "step": 14848 }, { "epoch": 2.83, "learning_rate": 2.153536701620591e-06, "loss": 2.8665, "step": 14849 }, { "epoch": 2.83, "learning_rate": 2.1534795042898e-06, "loss": 2.1964, "step": 14850 }, { "epoch": 2.83, "learning_rate": 2.153422306959009e-06, "loss": 3.2343, "step": 14851 }, { "epoch": 2.83, "learning_rate": 2.153365109628217e-06, "loss": 2.5437, "step": 14852 }, { "epoch": 2.83, "learning_rate": 2.1533079122974264e-06, "loss": 3.0156, "step": 14853 }, { "epoch": 2.83, "learning_rate": 2.1532507149666348e-06, "loss": 2.018, "step": 14854 }, { "epoch": 2.83, "learning_rate": 2.1531935176358436e-06, "loss": 1.4861, "step": 14855 }, { "epoch": 2.83, "learning_rate": 2.153136320305053e-06, "loss": 2.6821, "step": 14856 }, { "epoch": 2.83, "learning_rate": 2.153079122974261e-06, "loss": 2.0641, "step": 14857 }, { "epoch": 2.83, "learning_rate": 2.15302192564347e-06, "loss": 2.5479, "step": 14858 }, { "epoch": 2.83, "learning_rate": 2.1529647283126787e-06, "loss": 2.6572, "step": 14859 }, { "epoch": 2.83, "learning_rate": 2.1529075309818875e-06, "loss": 1.873, "step": 14860 }, { "epoch": 2.83, "learning_rate": 2.1528503336510963e-06, "loss": 2.7702, "step": 14861 }, { "epoch": 2.83, "learning_rate": 2.152793136320305e-06, "loss": 2.304, "step": 14862 }, { "epoch": 2.83, "learning_rate": 2.152735938989514e-06, "loss": 3.6216, "step": 14863 }, { "epoch": 2.83, "learning_rate": 2.1526787416587227e-06, "loss": 2.7719, "step": 14864 }, { "epoch": 2.83, "learning_rate": 2.1526215443279315e-06, "loss": 2.3033, "step": 14865 }, { "epoch": 2.83, "learning_rate": 2.1525643469971403e-06, "loss": 2.4911, "step": 14866 }, { "epoch": 2.83, "learning_rate": 2.152507149666349e-06, "loss": 2.8885, "step": 14867 }, { "epoch": 2.83, "learning_rate": 2.152449952335558e-06, "loss": 2.0838, "step": 14868 }, { "epoch": 2.83, "learning_rate": 2.1523927550047663e-06, "loss": 2.1833, "step": 14869 }, { "epoch": 2.83, "learning_rate": 2.1523355576739755e-06, "loss": 2.5869, "step": 14870 }, { "epoch": 2.83, "learning_rate": 2.152278360343184e-06, "loss": 3.0117, "step": 14871 }, { "epoch": 2.83, "learning_rate": 2.1522211630123926e-06, "loss": 2.5195, "step": 14872 }, { "epoch": 2.83, "learning_rate": 2.152163965681602e-06, "loss": 2.9443, "step": 14873 }, { "epoch": 2.83, "learning_rate": 2.1521067683508102e-06, "loss": 2.3136, "step": 14874 }, { "epoch": 2.83, "learning_rate": 2.152049571020019e-06, "loss": 2.3294, "step": 14875 }, { "epoch": 2.83, "learning_rate": 2.151992373689228e-06, "loss": 2.2133, "step": 14876 }, { "epoch": 2.83, "learning_rate": 2.1519351763584366e-06, "loss": 2.8784, "step": 14877 }, { "epoch": 2.83, "learning_rate": 2.1518779790276454e-06, "loss": 2.7279, "step": 14878 }, { "epoch": 2.83, "learning_rate": 2.151820781696854e-06, "loss": 3.0795, "step": 14879 }, { "epoch": 2.83, "learning_rate": 2.151763584366063e-06, "loss": 2.0646, "step": 14880 }, { "epoch": 2.83, "learning_rate": 2.1517063870352718e-06, "loss": 3.1158, "step": 14881 }, { "epoch": 2.83, "learning_rate": 2.1516491897044806e-06, "loss": 3.4419, "step": 14882 }, { "epoch": 2.83, "learning_rate": 2.1515919923736894e-06, "loss": 2.5555, "step": 14883 }, { "epoch": 2.84, "learning_rate": 2.151534795042898e-06, "loss": 3.4495, "step": 14884 }, { "epoch": 2.84, "learning_rate": 2.151477597712107e-06, "loss": 3.7236, "step": 14885 }, { "epoch": 2.84, "learning_rate": 2.1514204003813153e-06, "loss": 3.2862, "step": 14886 }, { "epoch": 2.84, "learning_rate": 2.1513632030505246e-06, "loss": 2.9722, "step": 14887 }, { "epoch": 2.84, "learning_rate": 2.151306005719733e-06, "loss": 2.3212, "step": 14888 }, { "epoch": 2.84, "learning_rate": 2.1512488083889417e-06, "loss": 2.7421, "step": 14889 }, { "epoch": 2.84, "learning_rate": 2.151191611058151e-06, "loss": 2.8055, "step": 14890 }, { "epoch": 2.84, "learning_rate": 2.1511344137273593e-06, "loss": 2.2488, "step": 14891 }, { "epoch": 2.84, "learning_rate": 2.151077216396568e-06, "loss": 2.8289, "step": 14892 }, { "epoch": 2.84, "learning_rate": 2.151020019065777e-06, "loss": 3.0771, "step": 14893 }, { "epoch": 2.84, "learning_rate": 2.1509628217349857e-06, "loss": 2.7067, "step": 14894 }, { "epoch": 2.84, "learning_rate": 2.1509056244041945e-06, "loss": 3.3622, "step": 14895 }, { "epoch": 2.84, "learning_rate": 2.1508484270734033e-06, "loss": 3.2785, "step": 14896 }, { "epoch": 2.84, "learning_rate": 2.150791229742612e-06, "loss": 2.3826, "step": 14897 }, { "epoch": 2.84, "learning_rate": 2.150734032411821e-06, "loss": 2.3271, "step": 14898 }, { "epoch": 2.84, "learning_rate": 2.1506768350810297e-06, "loss": 2.4323, "step": 14899 }, { "epoch": 2.84, "learning_rate": 2.1506196377502384e-06, "loss": 2.3182, "step": 14900 }, { "epoch": 2.84, "learning_rate": 2.1505624404194472e-06, "loss": 2.3246, "step": 14901 }, { "epoch": 2.84, "learning_rate": 2.150505243088656e-06, "loss": 1.4734, "step": 14902 }, { "epoch": 2.84, "learning_rate": 2.1504480457578644e-06, "loss": 2.0731, "step": 14903 }, { "epoch": 2.84, "learning_rate": 2.1503908484270736e-06, "loss": 3.0812, "step": 14904 }, { "epoch": 2.84, "learning_rate": 2.1503336510962824e-06, "loss": 2.1906, "step": 14905 }, { "epoch": 2.84, "learning_rate": 2.1502764537654908e-06, "loss": 2.0781, "step": 14906 }, { "epoch": 2.84, "learning_rate": 2.1502192564347e-06, "loss": 3.2309, "step": 14907 }, { "epoch": 2.84, "learning_rate": 2.1501620591039084e-06, "loss": 2.3874, "step": 14908 }, { "epoch": 2.84, "learning_rate": 2.150104861773117e-06, "loss": 2.2505, "step": 14909 }, { "epoch": 2.84, "learning_rate": 2.150047664442326e-06, "loss": 2.9089, "step": 14910 }, { "epoch": 2.84, "learning_rate": 2.1499904671115348e-06, "loss": 3.3825, "step": 14911 }, { "epoch": 2.84, "learning_rate": 2.149933269780744e-06, "loss": 2.9185, "step": 14912 }, { "epoch": 2.84, "learning_rate": 2.1498760724499523e-06, "loss": 3.3479, "step": 14913 }, { "epoch": 2.84, "learning_rate": 2.149818875119161e-06, "loss": 2.8214, "step": 14914 }, { "epoch": 2.84, "learning_rate": 2.14976167778837e-06, "loss": 2.1711, "step": 14915 }, { "epoch": 2.84, "learning_rate": 2.1497044804575787e-06, "loss": 2.9307, "step": 14916 }, { "epoch": 2.84, "learning_rate": 2.1496472831267875e-06, "loss": 2.1641, "step": 14917 }, { "epoch": 2.84, "learning_rate": 2.1495900857959963e-06, "loss": 2.557, "step": 14918 }, { "epoch": 2.84, "learning_rate": 2.149532888465205e-06, "loss": 3.1569, "step": 14919 }, { "epoch": 2.84, "learning_rate": 2.1494756911344135e-06, "loss": 3.1129, "step": 14920 }, { "epoch": 2.84, "learning_rate": 2.1494184938036227e-06, "loss": 2.848, "step": 14921 }, { "epoch": 2.84, "learning_rate": 2.1493612964728315e-06, "loss": 2.2103, "step": 14922 }, { "epoch": 2.84, "learning_rate": 2.14930409914204e-06, "loss": 1.7414, "step": 14923 }, { "epoch": 2.84, "learning_rate": 2.149246901811249e-06, "loss": 2.0647, "step": 14924 }, { "epoch": 2.84, "learning_rate": 2.1491897044804574e-06, "loss": 1.8733, "step": 14925 }, { "epoch": 2.84, "learning_rate": 2.1491325071496662e-06, "loss": 3.2094, "step": 14926 }, { "epoch": 2.84, "learning_rate": 2.149075309818875e-06, "loss": 2.7128, "step": 14927 }, { "epoch": 2.84, "learning_rate": 2.149018112488084e-06, "loss": 2.8996, "step": 14928 }, { "epoch": 2.84, "learning_rate": 2.148960915157293e-06, "loss": 2.8722, "step": 14929 }, { "epoch": 2.84, "learning_rate": 2.1489037178265014e-06, "loss": 2.9438, "step": 14930 }, { "epoch": 2.84, "learning_rate": 2.14884652049571e-06, "loss": 3.0536, "step": 14931 }, { "epoch": 2.84, "learning_rate": 2.148789323164919e-06, "loss": 2.8898, "step": 14932 }, { "epoch": 2.84, "learning_rate": 2.148732125834128e-06, "loss": 2.2328, "step": 14933 }, { "epoch": 2.84, "learning_rate": 2.1486749285033366e-06, "loss": 3.5035, "step": 14934 }, { "epoch": 2.84, "learning_rate": 2.1486177311725454e-06, "loss": 3.4646, "step": 14935 }, { "epoch": 2.84, "learning_rate": 2.148560533841754e-06, "loss": 2.6344, "step": 14936 }, { "epoch": 2.85, "learning_rate": 2.1485033365109625e-06, "loss": 2.6716, "step": 14937 }, { "epoch": 2.85, "learning_rate": 2.1484461391801718e-06, "loss": 3.2563, "step": 14938 }, { "epoch": 2.85, "learning_rate": 2.1483889418493806e-06, "loss": 1.7138, "step": 14939 }, { "epoch": 2.85, "learning_rate": 2.148331744518589e-06, "loss": 2.6645, "step": 14940 }, { "epoch": 2.85, "learning_rate": 2.148274547187798e-06, "loss": 1.7595, "step": 14941 }, { "epoch": 2.85, "learning_rate": 2.1482173498570065e-06, "loss": 1.8359, "step": 14942 }, { "epoch": 2.85, "learning_rate": 2.1481601525262153e-06, "loss": 2.7042, "step": 14943 }, { "epoch": 2.85, "learning_rate": 2.1481029551954245e-06, "loss": 3.1312, "step": 14944 }, { "epoch": 2.85, "learning_rate": 2.148045757864633e-06, "loss": 2.9793, "step": 14945 }, { "epoch": 2.85, "learning_rate": 2.147988560533842e-06, "loss": 1.9311, "step": 14946 }, { "epoch": 2.85, "learning_rate": 2.1479313632030505e-06, "loss": 1.6794, "step": 14947 }, { "epoch": 2.85, "learning_rate": 2.1478741658722593e-06, "loss": 3.6719, "step": 14948 }, { "epoch": 2.85, "learning_rate": 2.147816968541468e-06, "loss": 2.591, "step": 14949 }, { "epoch": 2.85, "learning_rate": 2.147759771210677e-06, "loss": 2.8406, "step": 14950 }, { "epoch": 2.85, "learning_rate": 2.1477025738798857e-06, "loss": 3.5079, "step": 14951 }, { "epoch": 2.85, "learning_rate": 2.1476453765490945e-06, "loss": 2.3528, "step": 14952 }, { "epoch": 2.85, "learning_rate": 2.1475881792183032e-06, "loss": 2.9841, "step": 14953 }, { "epoch": 2.85, "learning_rate": 2.147530981887512e-06, "loss": 2.1498, "step": 14954 }, { "epoch": 2.85, "learning_rate": 2.147473784556721e-06, "loss": 1.9016, "step": 14955 }, { "epoch": 2.85, "learning_rate": 2.1474165872259296e-06, "loss": 1.5557, "step": 14956 }, { "epoch": 2.85, "learning_rate": 2.147359389895138e-06, "loss": 1.9269, "step": 14957 }, { "epoch": 2.85, "learning_rate": 2.1473021925643472e-06, "loss": 3.221, "step": 14958 }, { "epoch": 2.85, "learning_rate": 2.1472449952335556e-06, "loss": 3.1046, "step": 14959 }, { "epoch": 2.85, "learning_rate": 2.1471877979027644e-06, "loss": 2.295, "step": 14960 }, { "epoch": 2.85, "learning_rate": 2.1471306005719736e-06, "loss": 2.551, "step": 14961 }, { "epoch": 2.85, "learning_rate": 2.147073403241182e-06, "loss": 3.0069, "step": 14962 }, { "epoch": 2.85, "learning_rate": 2.147016205910391e-06, "loss": 3.4944, "step": 14963 }, { "epoch": 2.85, "learning_rate": 2.1469590085795996e-06, "loss": 2.5357, "step": 14964 }, { "epoch": 2.85, "learning_rate": 2.1469018112488084e-06, "loss": 2.8146, "step": 14965 }, { "epoch": 2.85, "learning_rate": 2.146844613918017e-06, "loss": 2.9222, "step": 14966 }, { "epoch": 2.85, "learning_rate": 2.146787416587226e-06, "loss": 1.7293, "step": 14967 }, { "epoch": 2.85, "learning_rate": 2.1467302192564347e-06, "loss": 1.9423, "step": 14968 }, { "epoch": 2.85, "learning_rate": 2.1466730219256435e-06, "loss": 2.3714, "step": 14969 }, { "epoch": 2.85, "learning_rate": 2.1466158245948523e-06, "loss": 2.2218, "step": 14970 }, { "epoch": 2.85, "learning_rate": 2.146558627264061e-06, "loss": 3.2389, "step": 14971 }, { "epoch": 2.85, "learning_rate": 2.14650142993327e-06, "loss": 3.0859, "step": 14972 }, { "epoch": 2.85, "learning_rate": 2.1464442326024787e-06, "loss": 2.4562, "step": 14973 }, { "epoch": 2.85, "learning_rate": 2.146387035271687e-06, "loss": 2.0461, "step": 14974 }, { "epoch": 2.85, "learning_rate": 2.1463298379408963e-06, "loss": 2.1286, "step": 14975 }, { "epoch": 2.85, "learning_rate": 2.1462726406101047e-06, "loss": 2.8613, "step": 14976 }, { "epoch": 2.85, "learning_rate": 2.146215443279314e-06, "loss": 1.574, "step": 14977 }, { "epoch": 2.85, "learning_rate": 2.1461582459485227e-06, "loss": 2.3829, "step": 14978 }, { "epoch": 2.85, "learning_rate": 2.146101048617731e-06, "loss": 2.1162, "step": 14979 }, { "epoch": 2.85, "learning_rate": 2.1460438512869403e-06, "loss": 2.6274, "step": 14980 }, { "epoch": 2.85, "learning_rate": 2.1459866539561486e-06, "loss": 2.2159, "step": 14981 }, { "epoch": 2.85, "learning_rate": 2.1459294566253574e-06, "loss": 1.4993, "step": 14982 }, { "epoch": 2.85, "learning_rate": 2.1458722592945666e-06, "loss": 3.0081, "step": 14983 }, { "epoch": 2.85, "learning_rate": 2.145815061963775e-06, "loss": 2.9081, "step": 14984 }, { "epoch": 2.85, "learning_rate": 2.145757864632984e-06, "loss": 1.4996, "step": 14985 }, { "epoch": 2.85, "learning_rate": 2.1457006673021926e-06, "loss": 2.7323, "step": 14986 }, { "epoch": 2.85, "learning_rate": 2.1456434699714014e-06, "loss": 3.4628, "step": 14987 }, { "epoch": 2.85, "learning_rate": 2.14558627264061e-06, "loss": 3.2312, "step": 14988 }, { "epoch": 2.86, "learning_rate": 2.145529075309819e-06, "loss": 2.3171, "step": 14989 }, { "epoch": 2.86, "learning_rate": 2.1454718779790278e-06, "loss": 2.6259, "step": 14990 }, { "epoch": 2.86, "learning_rate": 2.145414680648236e-06, "loss": 2.7489, "step": 14991 }, { "epoch": 2.86, "learning_rate": 2.1453574833174454e-06, "loss": 2.9927, "step": 14992 }, { "epoch": 2.86, "learning_rate": 2.145300285986654e-06, "loss": 1.7523, "step": 14993 }, { "epoch": 2.86, "learning_rate": 2.145243088655863e-06, "loss": 1.7399, "step": 14994 }, { "epoch": 2.86, "learning_rate": 2.1451858913250717e-06, "loss": 2.0535, "step": 14995 }, { "epoch": 2.86, "learning_rate": 2.14512869399428e-06, "loss": 2.1906, "step": 14996 }, { "epoch": 2.86, "learning_rate": 2.1450714966634893e-06, "loss": 1.7619, "step": 14997 }, { "epoch": 2.86, "learning_rate": 2.1450142993326977e-06, "loss": 2.2827, "step": 14998 }, { "epoch": 2.86, "learning_rate": 2.1449571020019065e-06, "loss": 3.1707, "step": 14999 }, { "epoch": 2.86, "learning_rate": 2.1448999046711157e-06, "loss": 2.9941, "step": 15000 }, { "epoch": 2.86, "learning_rate": 2.144842707340324e-06, "loss": 2.1012, "step": 15001 }, { "epoch": 2.86, "learning_rate": 2.144785510009533e-06, "loss": 2.6617, "step": 15002 }, { "epoch": 2.86, "learning_rate": 2.1447283126787417e-06, "loss": 2.2744, "step": 15003 }, { "epoch": 2.86, "learning_rate": 2.1446711153479505e-06, "loss": 2.0335, "step": 15004 }, { "epoch": 2.86, "learning_rate": 2.1446139180171593e-06, "loss": 1.9631, "step": 15005 }, { "epoch": 2.86, "learning_rate": 2.144556720686368e-06, "loss": 1.9781, "step": 15006 }, { "epoch": 2.86, "learning_rate": 2.144499523355577e-06, "loss": 2.8886, "step": 15007 }, { "epoch": 2.86, "learning_rate": 2.1444423260247852e-06, "loss": 2.367, "step": 15008 }, { "epoch": 2.86, "learning_rate": 2.1443851286939944e-06, "loss": 3.3146, "step": 15009 }, { "epoch": 2.86, "learning_rate": 2.1443279313632032e-06, "loss": 2.9693, "step": 15010 }, { "epoch": 2.86, "learning_rate": 2.144270734032412e-06, "loss": 2.5855, "step": 15011 }, { "epoch": 2.86, "learning_rate": 2.144213536701621e-06, "loss": 2.0615, "step": 15012 }, { "epoch": 2.86, "learning_rate": 2.144156339370829e-06, "loss": 2.4045, "step": 15013 }, { "epoch": 2.86, "learning_rate": 2.1440991420400384e-06, "loss": 2.5472, "step": 15014 }, { "epoch": 2.86, "learning_rate": 2.1440419447092468e-06, "loss": 2.3577, "step": 15015 }, { "epoch": 2.86, "learning_rate": 2.1439847473784556e-06, "loss": 1.4431, "step": 15016 }, { "epoch": 2.86, "learning_rate": 2.1439275500476648e-06, "loss": 2.6722, "step": 15017 }, { "epoch": 2.86, "learning_rate": 2.143870352716873e-06, "loss": 3.1771, "step": 15018 }, { "epoch": 2.86, "learning_rate": 2.143813155386082e-06, "loss": 3.1101, "step": 15019 }, { "epoch": 2.86, "learning_rate": 2.1437559580552907e-06, "loss": 2.6574, "step": 15020 }, { "epoch": 2.86, "learning_rate": 2.1436987607244995e-06, "loss": 3.1574, "step": 15021 }, { "epoch": 2.86, "learning_rate": 2.1436415633937083e-06, "loss": 2.035, "step": 15022 }, { "epoch": 2.86, "learning_rate": 2.143584366062917e-06, "loss": 2.659, "step": 15023 }, { "epoch": 2.86, "learning_rate": 2.143527168732126e-06, "loss": 3.0298, "step": 15024 }, { "epoch": 2.86, "learning_rate": 2.1434699714013343e-06, "loss": 1.3334, "step": 15025 }, { "epoch": 2.86, "learning_rate": 2.1434127740705435e-06, "loss": 2.0266, "step": 15026 }, { "epoch": 2.86, "learning_rate": 2.1433555767397523e-06, "loss": 2.0137, "step": 15027 }, { "epoch": 2.86, "learning_rate": 2.143298379408961e-06, "loss": 2.3033, "step": 15028 }, { "epoch": 2.86, "learning_rate": 2.14324118207817e-06, "loss": 2.8918, "step": 15029 }, { "epoch": 2.86, "learning_rate": 2.1431839847473783e-06, "loss": 2.4186, "step": 15030 }, { "epoch": 2.86, "learning_rate": 2.1431267874165875e-06, "loss": 2.3059, "step": 15031 }, { "epoch": 2.86, "learning_rate": 2.1430695900857963e-06, "loss": 1.8489, "step": 15032 }, { "epoch": 2.86, "learning_rate": 2.1430123927550046e-06, "loss": 2.0948, "step": 15033 }, { "epoch": 2.86, "learning_rate": 2.142955195424214e-06, "loss": 2.1741, "step": 15034 }, { "epoch": 2.86, "learning_rate": 2.1428979980934222e-06, "loss": 3.0209, "step": 15035 }, { "epoch": 2.86, "learning_rate": 2.142840800762631e-06, "loss": 3.0834, "step": 15036 }, { "epoch": 2.86, "learning_rate": 2.14278360343184e-06, "loss": 2.8439, "step": 15037 }, { "epoch": 2.86, "learning_rate": 2.1427264061010486e-06, "loss": 2.4425, "step": 15038 }, { "epoch": 2.86, "learning_rate": 2.1426692087702574e-06, "loss": 1.9199, "step": 15039 }, { "epoch": 2.86, "learning_rate": 2.142612011439466e-06, "loss": 2.6226, "step": 15040 }, { "epoch": 2.86, "learning_rate": 2.142554814108675e-06, "loss": 2.4762, "step": 15041 }, { "epoch": 2.87, "learning_rate": 2.1424976167778838e-06, "loss": 2.3132, "step": 15042 }, { "epoch": 2.87, "learning_rate": 2.1424404194470926e-06, "loss": 2.415, "step": 15043 }, { "epoch": 2.87, "learning_rate": 2.1423832221163014e-06, "loss": 3.1409, "step": 15044 }, { "epoch": 2.87, "learning_rate": 2.14232602478551e-06, "loss": 2.2055, "step": 15045 }, { "epoch": 2.87, "learning_rate": 2.142268827454719e-06, "loss": 3.2566, "step": 15046 }, { "epoch": 2.87, "learning_rate": 2.1422116301239273e-06, "loss": 4.1541, "step": 15047 }, { "epoch": 2.87, "learning_rate": 2.1421544327931365e-06, "loss": 2.0243, "step": 15048 }, { "epoch": 2.87, "learning_rate": 2.1420972354623453e-06, "loss": 1.5827, "step": 15049 }, { "epoch": 2.87, "learning_rate": 2.1420400381315537e-06, "loss": 2.3584, "step": 15050 }, { "epoch": 2.87, "learning_rate": 2.141982840800763e-06, "loss": 3.0016, "step": 15051 }, { "epoch": 2.87, "learning_rate": 2.1419256434699713e-06, "loss": 2.6779, "step": 15052 }, { "epoch": 2.87, "learning_rate": 2.14186844613918e-06, "loss": 2.9958, "step": 15053 }, { "epoch": 2.87, "learning_rate": 2.141811248808389e-06, "loss": 2.6231, "step": 15054 }, { "epoch": 2.87, "learning_rate": 2.1417540514775977e-06, "loss": 2.6459, "step": 15055 }, { "epoch": 2.87, "learning_rate": 2.1416968541468065e-06, "loss": 2.9971, "step": 15056 }, { "epoch": 2.87, "learning_rate": 2.1416396568160153e-06, "loss": 3.0428, "step": 15057 }, { "epoch": 2.87, "learning_rate": 2.141582459485224e-06, "loss": 2.6302, "step": 15058 }, { "epoch": 2.87, "learning_rate": 2.141525262154433e-06, "loss": 1.2877, "step": 15059 }, { "epoch": 2.87, "learning_rate": 2.1414680648236416e-06, "loss": 1.9049, "step": 15060 }, { "epoch": 2.87, "learning_rate": 2.1414108674928504e-06, "loss": 3.3693, "step": 15061 }, { "epoch": 2.87, "learning_rate": 2.1413536701620592e-06, "loss": 2.5023, "step": 15062 }, { "epoch": 2.87, "learning_rate": 2.141296472831268e-06, "loss": 2.5113, "step": 15063 }, { "epoch": 2.87, "learning_rate": 2.1412392755004764e-06, "loss": 2.3706, "step": 15064 }, { "epoch": 2.87, "learning_rate": 2.1411820781696856e-06, "loss": 3.0818, "step": 15065 }, { "epoch": 2.87, "learning_rate": 2.1411248808388944e-06, "loss": 3.0741, "step": 15066 }, { "epoch": 2.87, "learning_rate": 2.1410676835081028e-06, "loss": 1.9478, "step": 15067 }, { "epoch": 2.87, "learning_rate": 2.141010486177312e-06, "loss": 2.768, "step": 15068 }, { "epoch": 2.87, "learning_rate": 2.1409532888465204e-06, "loss": 2.9116, "step": 15069 }, { "epoch": 2.87, "learning_rate": 2.140896091515729e-06, "loss": 2.0602, "step": 15070 }, { "epoch": 2.87, "learning_rate": 2.1408388941849384e-06, "loss": 2.6097, "step": 15071 }, { "epoch": 2.87, "learning_rate": 2.1407816968541467e-06, "loss": 2.0468, "step": 15072 }, { "epoch": 2.87, "learning_rate": 2.1407244995233555e-06, "loss": 3.2869, "step": 15073 }, { "epoch": 2.87, "learning_rate": 2.1406673021925643e-06, "loss": 2.1076, "step": 15074 }, { "epoch": 2.87, "learning_rate": 2.140610104861773e-06, "loss": 2.851, "step": 15075 }, { "epoch": 2.87, "learning_rate": 2.140552907530982e-06, "loss": 1.9047, "step": 15076 }, { "epoch": 2.87, "learning_rate": 2.1404957102001907e-06, "loss": 3.489, "step": 15077 }, { "epoch": 2.87, "learning_rate": 2.1404385128693995e-06, "loss": 2.9481, "step": 15078 }, { "epoch": 2.87, "learning_rate": 2.1403813155386083e-06, "loss": 2.0302, "step": 15079 }, { "epoch": 2.87, "learning_rate": 2.140324118207817e-06, "loss": 2.0973, "step": 15080 }, { "epoch": 2.87, "learning_rate": 2.140266920877026e-06, "loss": 2.5107, "step": 15081 }, { "epoch": 2.87, "learning_rate": 2.1402097235462347e-06, "loss": 2.6745, "step": 15082 }, { "epoch": 2.87, "learning_rate": 2.1401525262154435e-06, "loss": 3.1274, "step": 15083 }, { "epoch": 2.87, "learning_rate": 2.140095328884652e-06, "loss": 1.9922, "step": 15084 }, { "epoch": 2.87, "learning_rate": 2.140038131553861e-06, "loss": 2.6537, "step": 15085 }, { "epoch": 2.87, "learning_rate": 2.1399809342230694e-06, "loss": 3.0084, "step": 15086 }, { "epoch": 2.87, "learning_rate": 2.1399237368922782e-06, "loss": 2.6735, "step": 15087 }, { "epoch": 2.87, "learning_rate": 2.1398665395614874e-06, "loss": 2.4872, "step": 15088 }, { "epoch": 2.87, "learning_rate": 2.139809342230696e-06, "loss": 3.129, "step": 15089 }, { "epoch": 2.87, "learning_rate": 2.1397521448999046e-06, "loss": 2.5992, "step": 15090 }, { "epoch": 2.87, "learning_rate": 2.1396949475691134e-06, "loss": 1.6944, "step": 15091 }, { "epoch": 2.87, "learning_rate": 2.139637750238322e-06, "loss": 2.137, "step": 15092 }, { "epoch": 2.87, "learning_rate": 2.139580552907531e-06, "loss": 2.9903, "step": 15093 }, { "epoch": 2.88, "learning_rate": 2.1395233555767398e-06, "loss": 2.4235, "step": 15094 }, { "epoch": 2.88, "learning_rate": 2.1394661582459486e-06, "loss": 2.4947, "step": 15095 }, { "epoch": 2.88, "learning_rate": 2.1394089609151574e-06, "loss": 1.9974, "step": 15096 }, { "epoch": 2.88, "learning_rate": 2.139351763584366e-06, "loss": 2.9109, "step": 15097 }, { "epoch": 2.88, "learning_rate": 2.139294566253575e-06, "loss": 2.7751, "step": 15098 }, { "epoch": 2.88, "learning_rate": 2.1392373689227838e-06, "loss": 2.0492, "step": 15099 }, { "epoch": 2.88, "learning_rate": 2.1391801715919926e-06, "loss": 1.701, "step": 15100 }, { "epoch": 2.88, "learning_rate": 2.139122974261201e-06, "loss": 2.0857, "step": 15101 }, { "epoch": 2.88, "learning_rate": 2.13906577693041e-06, "loss": 2.303, "step": 15102 }, { "epoch": 2.88, "learning_rate": 2.1390085795996185e-06, "loss": 2.041, "step": 15103 }, { "epoch": 2.88, "learning_rate": 2.1389513822688273e-06, "loss": 2.7428, "step": 15104 }, { "epoch": 2.88, "learning_rate": 2.1388941849380365e-06, "loss": 3.1323, "step": 15105 }, { "epoch": 2.88, "learning_rate": 2.138836987607245e-06, "loss": 3.7841, "step": 15106 }, { "epoch": 2.88, "learning_rate": 2.138779790276454e-06, "loss": 3.0536, "step": 15107 }, { "epoch": 2.88, "learning_rate": 2.1387225929456625e-06, "loss": 3.1658, "step": 15108 }, { "epoch": 2.88, "learning_rate": 2.1386653956148713e-06, "loss": 2.1351, "step": 15109 }, { "epoch": 2.88, "learning_rate": 2.13860819828408e-06, "loss": 2.6986, "step": 15110 }, { "epoch": 2.88, "learning_rate": 2.138551000953289e-06, "loss": 3.1081, "step": 15111 }, { "epoch": 2.88, "learning_rate": 2.1384938036224977e-06, "loss": 2.6286, "step": 15112 }, { "epoch": 2.88, "learning_rate": 2.1384366062917064e-06, "loss": 2.1901, "step": 15113 }, { "epoch": 2.88, "learning_rate": 2.1383794089609152e-06, "loss": 2.0963, "step": 15114 }, { "epoch": 2.88, "learning_rate": 2.138322211630124e-06, "loss": 2.6245, "step": 15115 }, { "epoch": 2.88, "learning_rate": 2.138265014299333e-06, "loss": 2.0619, "step": 15116 }, { "epoch": 2.88, "learning_rate": 2.1382078169685416e-06, "loss": 1.985, "step": 15117 }, { "epoch": 2.88, "learning_rate": 2.13815061963775e-06, "loss": 2.1287, "step": 15118 }, { "epoch": 2.88, "learning_rate": 2.138093422306959e-06, "loss": 3.1488, "step": 15119 }, { "epoch": 2.88, "learning_rate": 2.138036224976168e-06, "loss": 2.3948, "step": 15120 }, { "epoch": 2.88, "learning_rate": 2.1379790276453764e-06, "loss": 2.3236, "step": 15121 }, { "epoch": 2.88, "learning_rate": 2.1379218303145856e-06, "loss": 2.4444, "step": 15122 }, { "epoch": 2.88, "learning_rate": 2.137864632983794e-06, "loss": 1.8692, "step": 15123 }, { "epoch": 2.88, "learning_rate": 2.137807435653003e-06, "loss": 2.7351, "step": 15124 }, { "epoch": 2.88, "learning_rate": 2.1377502383222115e-06, "loss": 2.9892, "step": 15125 }, { "epoch": 2.88, "learning_rate": 2.1376930409914203e-06, "loss": 2.9306, "step": 15126 }, { "epoch": 2.88, "learning_rate": 2.1376358436606296e-06, "loss": 1.717, "step": 15127 }, { "epoch": 2.88, "learning_rate": 2.137578646329838e-06, "loss": 2.1932, "step": 15128 }, { "epoch": 2.88, "learning_rate": 2.1375214489990467e-06, "loss": 3.1923, "step": 15129 }, { "epoch": 2.88, "learning_rate": 2.1374642516682555e-06, "loss": 2.2419, "step": 15130 }, { "epoch": 2.88, "learning_rate": 2.1374070543374643e-06, "loss": 2.1499, "step": 15131 }, { "epoch": 2.88, "learning_rate": 2.137349857006673e-06, "loss": 2.1558, "step": 15132 }, { "epoch": 2.88, "learning_rate": 2.137292659675882e-06, "loss": 2.0762, "step": 15133 }, { "epoch": 2.88, "learning_rate": 2.1372354623450907e-06, "loss": 2.4718, "step": 15134 }, { "epoch": 2.88, "learning_rate": 2.137178265014299e-06, "loss": 2.4589, "step": 15135 }, { "epoch": 2.88, "learning_rate": 2.1371210676835083e-06, "loss": 2.8144, "step": 15136 }, { "epoch": 2.88, "learning_rate": 2.137063870352717e-06, "loss": 2.9936, "step": 15137 }, { "epoch": 2.88, "learning_rate": 2.1370066730219254e-06, "loss": 2.6605, "step": 15138 }, { "epoch": 2.88, "learning_rate": 2.1369494756911347e-06, "loss": 2.4237, "step": 15139 }, { "epoch": 2.88, "learning_rate": 2.136892278360343e-06, "loss": 3.0615, "step": 15140 }, { "epoch": 2.88, "learning_rate": 2.1368350810295523e-06, "loss": 2.6635, "step": 15141 }, { "epoch": 2.88, "learning_rate": 2.1367778836987606e-06, "loss": 2.0409, "step": 15142 }, { "epoch": 2.88, "learning_rate": 2.1367206863679694e-06, "loss": 2.5732, "step": 15143 }, { "epoch": 2.88, "learning_rate": 2.1366634890371786e-06, "loss": 1.7246, "step": 15144 }, { "epoch": 2.88, "learning_rate": 2.136606291706387e-06, "loss": 2.9995, "step": 15145 }, { "epoch": 2.88, "learning_rate": 2.136549094375596e-06, "loss": 3.5131, "step": 15146 }, { "epoch": 2.89, "learning_rate": 2.1364918970448046e-06, "loss": 1.873, "step": 15147 }, { "epoch": 2.89, "learning_rate": 2.1364346997140134e-06, "loss": 2.8981, "step": 15148 }, { "epoch": 2.89, "learning_rate": 2.136377502383222e-06, "loss": 2.2557, "step": 15149 }, { "epoch": 2.89, "learning_rate": 2.136320305052431e-06, "loss": 2.101, "step": 15150 }, { "epoch": 2.89, "learning_rate": 2.1362631077216398e-06, "loss": 2.1294, "step": 15151 }, { "epoch": 2.89, "learning_rate": 2.136205910390848e-06, "loss": 1.8262, "step": 15152 }, { "epoch": 2.89, "learning_rate": 2.1361487130600574e-06, "loss": 2.0378, "step": 15153 }, { "epoch": 2.89, "learning_rate": 2.136091515729266e-06, "loss": 2.1458, "step": 15154 }, { "epoch": 2.89, "learning_rate": 2.1360343183984745e-06, "loss": 3.0048, "step": 15155 }, { "epoch": 2.89, "learning_rate": 2.1359771210676837e-06, "loss": 2.7829, "step": 15156 }, { "epoch": 2.89, "learning_rate": 2.135919923736892e-06, "loss": 3.2121, "step": 15157 }, { "epoch": 2.89, "learning_rate": 2.1358627264061013e-06, "loss": 3.1319, "step": 15158 }, { "epoch": 2.89, "learning_rate": 2.1358055290753097e-06, "loss": 2.5013, "step": 15159 }, { "epoch": 2.89, "learning_rate": 2.1357483317445185e-06, "loss": 2.455, "step": 15160 }, { "epoch": 2.89, "learning_rate": 2.1356911344137277e-06, "loss": 3.4346, "step": 15161 }, { "epoch": 2.89, "learning_rate": 2.135633937082936e-06, "loss": 2.5461, "step": 15162 }, { "epoch": 2.89, "learning_rate": 2.135576739752145e-06, "loss": 2.554, "step": 15163 }, { "epoch": 2.89, "learning_rate": 2.1355195424213537e-06, "loss": 2.5418, "step": 15164 }, { "epoch": 2.89, "learning_rate": 2.1354623450905625e-06, "loss": 2.6775, "step": 15165 }, { "epoch": 2.89, "learning_rate": 2.1354051477597712e-06, "loss": 2.0613, "step": 15166 }, { "epoch": 2.89, "learning_rate": 2.13534795042898e-06, "loss": 2.4166, "step": 15167 }, { "epoch": 2.89, "learning_rate": 2.135290753098189e-06, "loss": 1.6629, "step": 15168 }, { "epoch": 2.89, "learning_rate": 2.1352335557673976e-06, "loss": 2.6754, "step": 15169 }, { "epoch": 2.89, "learning_rate": 2.1351763584366064e-06, "loss": 2.1837, "step": 15170 }, { "epoch": 2.89, "learning_rate": 2.1351191611058152e-06, "loss": 3.2241, "step": 15171 }, { "epoch": 2.89, "learning_rate": 2.135061963775024e-06, "loss": 2.6499, "step": 15172 }, { "epoch": 2.89, "learning_rate": 2.135004766444233e-06, "loss": 3.6112, "step": 15173 }, { "epoch": 2.89, "learning_rate": 2.134947569113441e-06, "loss": 2.8296, "step": 15174 }, { "epoch": 2.89, "learning_rate": 2.1348903717826504e-06, "loss": 2.2921, "step": 15175 }, { "epoch": 2.89, "learning_rate": 2.134833174451859e-06, "loss": 3.2708, "step": 15176 }, { "epoch": 2.89, "learning_rate": 2.1347759771210676e-06, "loss": 2.7553, "step": 15177 }, { "epoch": 2.89, "learning_rate": 2.1347187797902768e-06, "loss": 1.9648, "step": 15178 }, { "epoch": 2.89, "learning_rate": 2.134661582459485e-06, "loss": 3.4272, "step": 15179 }, { "epoch": 2.89, "learning_rate": 2.134604385128694e-06, "loss": 2.8455, "step": 15180 }, { "epoch": 2.89, "learning_rate": 2.1345471877979027e-06, "loss": 3.4853, "step": 15181 }, { "epoch": 2.89, "learning_rate": 2.1344899904671115e-06, "loss": 3.3218, "step": 15182 }, { "epoch": 2.89, "learning_rate": 2.1344327931363203e-06, "loss": 2.9879, "step": 15183 }, { "epoch": 2.89, "learning_rate": 2.134375595805529e-06, "loss": 2.6465, "step": 15184 }, { "epoch": 2.89, "learning_rate": 2.134318398474738e-06, "loss": 2.9469, "step": 15185 }, { "epoch": 2.89, "learning_rate": 2.1342612011439467e-06, "loss": 2.5983, "step": 15186 }, { "epoch": 2.89, "learning_rate": 2.1342040038131555e-06, "loss": 2.1118, "step": 15187 }, { "epoch": 2.89, "learning_rate": 2.1341468064823643e-06, "loss": 2.3499, "step": 15188 }, { "epoch": 2.89, "learning_rate": 2.134089609151573e-06, "loss": 2.7753, "step": 15189 }, { "epoch": 2.89, "learning_rate": 2.134032411820782e-06, "loss": 2.1406, "step": 15190 }, { "epoch": 2.89, "learning_rate": 2.1339752144899902e-06, "loss": 2.9736, "step": 15191 }, { "epoch": 2.89, "learning_rate": 2.1339180171591995e-06, "loss": 2.1845, "step": 15192 }, { "epoch": 2.89, "learning_rate": 2.1338608198284083e-06, "loss": 2.9357, "step": 15193 }, { "epoch": 2.89, "learning_rate": 2.1338036224976166e-06, "loss": 2.2933, "step": 15194 }, { "epoch": 2.89, "learning_rate": 2.133746425166826e-06, "loss": 1.9471, "step": 15195 }, { "epoch": 2.89, "learning_rate": 2.1336892278360342e-06, "loss": 2.9368, "step": 15196 }, { "epoch": 2.89, "learning_rate": 2.133632030505243e-06, "loss": 2.1545, "step": 15197 }, { "epoch": 2.89, "learning_rate": 2.133574833174452e-06, "loss": 1.5259, "step": 15198 }, { "epoch": 2.9, "learning_rate": 2.1335176358436606e-06, "loss": 2.9431, "step": 15199 }, { "epoch": 2.9, "learning_rate": 2.1334604385128694e-06, "loss": 2.3142, "step": 15200 }, { "epoch": 2.9, "learning_rate": 2.133403241182078e-06, "loss": 2.63, "step": 15201 }, { "epoch": 2.9, "learning_rate": 2.133346043851287e-06, "loss": 2.9153, "step": 15202 }, { "epoch": 2.9, "learning_rate": 2.1332888465204958e-06, "loss": 3.2367, "step": 15203 }, { "epoch": 2.9, "learning_rate": 2.1332316491897046e-06, "loss": 3.0615, "step": 15204 }, { "epoch": 2.9, "learning_rate": 2.1331744518589134e-06, "loss": 2.3252, "step": 15205 }, { "epoch": 2.9, "learning_rate": 2.133117254528122e-06, "loss": 2.8758, "step": 15206 }, { "epoch": 2.9, "learning_rate": 2.133060057197331e-06, "loss": 1.6574, "step": 15207 }, { "epoch": 2.9, "learning_rate": 2.1330028598665397e-06, "loss": 2.5057, "step": 15208 }, { "epoch": 2.9, "learning_rate": 2.1329456625357485e-06, "loss": 2.6813, "step": 15209 }, { "epoch": 2.9, "learning_rate": 2.1328884652049573e-06, "loss": 1.9788, "step": 15210 }, { "epoch": 2.9, "learning_rate": 2.1328312678741657e-06, "loss": 1.8231, "step": 15211 }, { "epoch": 2.9, "learning_rate": 2.132774070543375e-06, "loss": 2.8292, "step": 15212 }, { "epoch": 2.9, "learning_rate": 2.1327168732125833e-06, "loss": 1.9362, "step": 15213 }, { "epoch": 2.9, "learning_rate": 2.132659675881792e-06, "loss": 3.5021, "step": 15214 }, { "epoch": 2.9, "learning_rate": 2.1326024785510013e-06, "loss": 1.8529, "step": 15215 }, { "epoch": 2.9, "learning_rate": 2.1325452812202097e-06, "loss": 2.0579, "step": 15216 }, { "epoch": 2.9, "learning_rate": 2.1324880838894185e-06, "loss": 2.6356, "step": 15217 }, { "epoch": 2.9, "learning_rate": 2.1324308865586273e-06, "loss": 3.5289, "step": 15218 }, { "epoch": 2.9, "learning_rate": 2.132373689227836e-06, "loss": 1.4072, "step": 15219 }, { "epoch": 2.9, "learning_rate": 2.132316491897045e-06, "loss": 2.731, "step": 15220 }, { "epoch": 2.9, "learning_rate": 2.1322592945662536e-06, "loss": 2.2197, "step": 15221 }, { "epoch": 2.9, "learning_rate": 2.1322020972354624e-06, "loss": 2.2868, "step": 15222 }, { "epoch": 2.9, "learning_rate": 2.1321448999046712e-06, "loss": 3.1323, "step": 15223 }, { "epoch": 2.9, "learning_rate": 2.13208770257388e-06, "loss": 3.174, "step": 15224 }, { "epoch": 2.9, "learning_rate": 2.132030505243089e-06, "loss": 2.3119, "step": 15225 }, { "epoch": 2.9, "learning_rate": 2.1319733079122976e-06, "loss": 2.0386, "step": 15226 }, { "epoch": 2.9, "learning_rate": 2.1319161105815064e-06, "loss": 3.2401, "step": 15227 }, { "epoch": 2.9, "learning_rate": 2.1318589132507148e-06, "loss": 2.6809, "step": 15228 }, { "epoch": 2.9, "learning_rate": 2.131801715919924e-06, "loss": 2.9803, "step": 15229 }, { "epoch": 2.9, "learning_rate": 2.1317445185891324e-06, "loss": 2.2656, "step": 15230 }, { "epoch": 2.9, "learning_rate": 2.131687321258341e-06, "loss": 3.2559, "step": 15231 }, { "epoch": 2.9, "learning_rate": 2.1316301239275504e-06, "loss": 2.723, "step": 15232 }, { "epoch": 2.9, "learning_rate": 2.1315729265967587e-06, "loss": 2.7285, "step": 15233 }, { "epoch": 2.9, "learning_rate": 2.1315157292659675e-06, "loss": 1.6479, "step": 15234 }, { "epoch": 2.9, "learning_rate": 2.1314585319351763e-06, "loss": 2.1716, "step": 15235 }, { "epoch": 2.9, "learning_rate": 2.131401334604385e-06, "loss": 1.3015, "step": 15236 }, { "epoch": 2.9, "learning_rate": 2.131344137273594e-06, "loss": 2.525, "step": 15237 }, { "epoch": 2.9, "learning_rate": 2.1312869399428027e-06, "loss": 3.2193, "step": 15238 }, { "epoch": 2.9, "learning_rate": 2.1312297426120115e-06, "loss": 2.1839, "step": 15239 }, { "epoch": 2.9, "learning_rate": 2.1311725452812203e-06, "loss": 2.3311, "step": 15240 }, { "epoch": 2.9, "learning_rate": 2.131115347950429e-06, "loss": 3.2002, "step": 15241 }, { "epoch": 2.9, "learning_rate": 2.131058150619638e-06, "loss": 2.9921, "step": 15242 }, { "epoch": 2.9, "learning_rate": 2.1310009532888467e-06, "loss": 1.9377, "step": 15243 }, { "epoch": 2.9, "learning_rate": 2.1309437559580555e-06, "loss": 2.8653, "step": 15244 }, { "epoch": 2.9, "learning_rate": 2.130886558627264e-06, "loss": 3.0158, "step": 15245 }, { "epoch": 2.9, "learning_rate": 2.130829361296473e-06, "loss": 3.464, "step": 15246 }, { "epoch": 2.9, "learning_rate": 2.1307721639656814e-06, "loss": 1.7322, "step": 15247 }, { "epoch": 2.9, "learning_rate": 2.1307149666348902e-06, "loss": 2.9384, "step": 15248 }, { "epoch": 2.9, "learning_rate": 2.1306577693040994e-06, "loss": 2.9351, "step": 15249 }, { "epoch": 2.9, "learning_rate": 2.130600571973308e-06, "loss": 2.5411, "step": 15250 }, { "epoch": 2.9, "learning_rate": 2.1305433746425166e-06, "loss": 1.9148, "step": 15251 }, { "epoch": 2.91, "learning_rate": 2.1304861773117254e-06, "loss": 2.951, "step": 15252 }, { "epoch": 2.91, "learning_rate": 2.130428979980934e-06, "loss": 2.2731, "step": 15253 }, { "epoch": 2.91, "learning_rate": 2.1303717826501434e-06, "loss": 3.5096, "step": 15254 }, { "epoch": 2.91, "learning_rate": 2.1303145853193518e-06, "loss": 2.2843, "step": 15255 }, { "epoch": 2.91, "learning_rate": 2.1302573879885606e-06, "loss": 3.1539, "step": 15256 }, { "epoch": 2.91, "learning_rate": 2.1302001906577694e-06, "loss": 3.0747, "step": 15257 }, { "epoch": 2.91, "learning_rate": 2.130142993326978e-06, "loss": 3.0556, "step": 15258 }, { "epoch": 2.91, "learning_rate": 2.130085795996187e-06, "loss": 3.097, "step": 15259 }, { "epoch": 2.91, "learning_rate": 2.1300285986653958e-06, "loss": 3.0729, "step": 15260 }, { "epoch": 2.91, "learning_rate": 2.1299714013346045e-06, "loss": 3.4999, "step": 15261 }, { "epoch": 2.91, "learning_rate": 2.129914204003813e-06, "loss": 3.2797, "step": 15262 }, { "epoch": 2.91, "learning_rate": 2.129857006673022e-06, "loss": 1.8165, "step": 15263 }, { "epoch": 2.91, "learning_rate": 2.129799809342231e-06, "loss": 2.5025, "step": 15264 }, { "epoch": 2.91, "learning_rate": 2.1297426120114393e-06, "loss": 1.9947, "step": 15265 }, { "epoch": 2.91, "learning_rate": 2.1296854146806485e-06, "loss": 2.1827, "step": 15266 }, { "epoch": 2.91, "learning_rate": 2.129628217349857e-06, "loss": 2.9433, "step": 15267 }, { "epoch": 2.91, "learning_rate": 2.1295710200190657e-06, "loss": 2.11, "step": 15268 }, { "epoch": 2.91, "learning_rate": 2.1295138226882745e-06, "loss": 3.3844, "step": 15269 }, { "epoch": 2.91, "learning_rate": 2.1294566253574833e-06, "loss": 2.0491, "step": 15270 }, { "epoch": 2.91, "learning_rate": 2.1293994280266925e-06, "loss": 3.323, "step": 15271 }, { "epoch": 2.91, "learning_rate": 2.129342230695901e-06, "loss": 3.1776, "step": 15272 }, { "epoch": 2.91, "learning_rate": 2.1292850333651096e-06, "loss": 2.5975, "step": 15273 }, { "epoch": 2.91, "learning_rate": 2.1292278360343184e-06, "loss": 1.681, "step": 15274 }, { "epoch": 2.91, "learning_rate": 2.1291706387035272e-06, "loss": 2.6456, "step": 15275 }, { "epoch": 2.91, "learning_rate": 2.129113441372736e-06, "loss": 2.1387, "step": 15276 }, { "epoch": 2.91, "learning_rate": 2.129056244041945e-06, "loss": 2.911, "step": 15277 }, { "epoch": 2.91, "learning_rate": 2.1289990467111536e-06, "loss": 2.4741, "step": 15278 }, { "epoch": 2.91, "learning_rate": 2.128941849380362e-06, "loss": 2.1903, "step": 15279 }, { "epoch": 2.91, "learning_rate": 2.128884652049571e-06, "loss": 2.2538, "step": 15280 }, { "epoch": 2.91, "learning_rate": 2.12882745471878e-06, "loss": 3.1598, "step": 15281 }, { "epoch": 2.91, "learning_rate": 2.1287702573879884e-06, "loss": 3.4354, "step": 15282 }, { "epoch": 2.91, "learning_rate": 2.1287130600571976e-06, "loss": 3.0306, "step": 15283 }, { "epoch": 2.91, "learning_rate": 2.128655862726406e-06, "loss": 2.7621, "step": 15284 }, { "epoch": 2.91, "learning_rate": 2.1285986653956147e-06, "loss": 1.7575, "step": 15285 }, { "epoch": 2.91, "learning_rate": 2.1285414680648235e-06, "loss": 2.8055, "step": 15286 }, { "epoch": 2.91, "learning_rate": 2.1284842707340323e-06, "loss": 2.8252, "step": 15287 }, { "epoch": 2.91, "learning_rate": 2.1284270734032416e-06, "loss": 1.9315, "step": 15288 }, { "epoch": 2.91, "learning_rate": 2.12836987607245e-06, "loss": 2.2917, "step": 15289 }, { "epoch": 2.91, "learning_rate": 2.1283126787416587e-06, "loss": 1.234, "step": 15290 }, { "epoch": 2.91, "learning_rate": 2.1282554814108675e-06, "loss": 2.8441, "step": 15291 }, { "epoch": 2.91, "learning_rate": 2.1281982840800763e-06, "loss": 2.7991, "step": 15292 }, { "epoch": 2.91, "learning_rate": 2.128141086749285e-06, "loss": 2.722, "step": 15293 }, { "epoch": 2.91, "learning_rate": 2.128083889418494e-06, "loss": 3.2488, "step": 15294 }, { "epoch": 2.91, "learning_rate": 2.1280266920877027e-06, "loss": 2.7293, "step": 15295 }, { "epoch": 2.91, "learning_rate": 2.1279694947569115e-06, "loss": 2.1009, "step": 15296 }, { "epoch": 2.91, "learning_rate": 2.1279122974261203e-06, "loss": 2.3069, "step": 15297 }, { "epoch": 2.91, "learning_rate": 2.127855100095329e-06, "loss": 2.7625, "step": 15298 }, { "epoch": 2.91, "learning_rate": 2.1277979027645374e-06, "loss": 2.5941, "step": 15299 }, { "epoch": 2.91, "learning_rate": 2.1277407054337467e-06, "loss": 3.3274, "step": 15300 }, { "epoch": 2.91, "learning_rate": 2.127683508102955e-06, "loss": 1.2862, "step": 15301 }, { "epoch": 2.91, "learning_rate": 2.1276263107721642e-06, "loss": 2.1869, "step": 15302 }, { "epoch": 2.91, "learning_rate": 2.127569113441373e-06, "loss": 2.4955, "step": 15303 }, { "epoch": 2.92, "learning_rate": 2.1275119161105814e-06, "loss": 2.2715, "step": 15304 }, { "epoch": 2.92, "learning_rate": 2.1274547187797906e-06, "loss": 1.708, "step": 15305 }, { "epoch": 2.92, "learning_rate": 2.127397521448999e-06, "loss": 2.1539, "step": 15306 }, { "epoch": 2.92, "learning_rate": 2.127340324118208e-06, "loss": 1.8049, "step": 15307 }, { "epoch": 2.92, "learning_rate": 2.1272831267874166e-06, "loss": 2.6209, "step": 15308 }, { "epoch": 2.92, "learning_rate": 2.1272259294566254e-06, "loss": 2.5707, "step": 15309 }, { "epoch": 2.92, "learning_rate": 2.127168732125834e-06, "loss": 2.6352, "step": 15310 }, { "epoch": 2.92, "learning_rate": 2.127111534795043e-06, "loss": 2.4034, "step": 15311 }, { "epoch": 2.92, "learning_rate": 2.1270543374642518e-06, "loss": 3.406, "step": 15312 }, { "epoch": 2.92, "learning_rate": 2.1269971401334606e-06, "loss": 2.1763, "step": 15313 }, { "epoch": 2.92, "learning_rate": 2.1269399428026693e-06, "loss": 2.2765, "step": 15314 }, { "epoch": 2.92, "learning_rate": 2.126882745471878e-06, "loss": 2.4827, "step": 15315 }, { "epoch": 2.92, "learning_rate": 2.1268255481410865e-06, "loss": 2.3148, "step": 15316 }, { "epoch": 2.92, "learning_rate": 2.1267683508102957e-06, "loss": 2.2466, "step": 15317 }, { "epoch": 2.92, "learning_rate": 2.126711153479504e-06, "loss": 2.1153, "step": 15318 }, { "epoch": 2.92, "learning_rate": 2.1266539561487133e-06, "loss": 2.469, "step": 15319 }, { "epoch": 2.92, "learning_rate": 2.126596758817922e-06, "loss": 2.5303, "step": 15320 }, { "epoch": 2.92, "learning_rate": 2.1265395614871305e-06, "loss": 3.0818, "step": 15321 }, { "epoch": 2.92, "learning_rate": 2.1264823641563397e-06, "loss": 2.2057, "step": 15322 }, { "epoch": 2.92, "learning_rate": 2.126425166825548e-06, "loss": 3.3662, "step": 15323 }, { "epoch": 2.92, "learning_rate": 2.126367969494757e-06, "loss": 2.92, "step": 15324 }, { "epoch": 2.92, "learning_rate": 2.1263107721639657e-06, "loss": 3.0972, "step": 15325 }, { "epoch": 2.92, "learning_rate": 2.1262535748331744e-06, "loss": 3.6263, "step": 15326 }, { "epoch": 2.92, "learning_rate": 2.1261963775023832e-06, "loss": 2.4329, "step": 15327 }, { "epoch": 2.92, "learning_rate": 2.126139180171592e-06, "loss": 2.6503, "step": 15328 }, { "epoch": 2.92, "learning_rate": 2.126081982840801e-06, "loss": 1.5829, "step": 15329 }, { "epoch": 2.92, "learning_rate": 2.1260247855100096e-06, "loss": 2.8531, "step": 15330 }, { "epoch": 2.92, "learning_rate": 2.1259675881792184e-06, "loss": 3.5081, "step": 15331 }, { "epoch": 2.92, "learning_rate": 2.125910390848427e-06, "loss": 2.0748, "step": 15332 }, { "epoch": 2.92, "learning_rate": 2.1258531935176356e-06, "loss": 2.5024, "step": 15333 }, { "epoch": 2.92, "learning_rate": 2.125795996186845e-06, "loss": 3.0839, "step": 15334 }, { "epoch": 2.92, "learning_rate": 2.125738798856053e-06, "loss": 1.8947, "step": 15335 }, { "epoch": 2.92, "learning_rate": 2.1256816015252624e-06, "loss": 2.7817, "step": 15336 }, { "epoch": 2.92, "learning_rate": 2.125624404194471e-06, "loss": 2.2832, "step": 15337 }, { "epoch": 2.92, "learning_rate": 2.1255672068636796e-06, "loss": 3.9124, "step": 15338 }, { "epoch": 2.92, "learning_rate": 2.1255100095328888e-06, "loss": 1.8763, "step": 15339 }, { "epoch": 2.92, "learning_rate": 2.125452812202097e-06, "loss": 3.1322, "step": 15340 }, { "epoch": 2.92, "learning_rate": 2.125395614871306e-06, "loss": 2.458, "step": 15341 }, { "epoch": 2.92, "learning_rate": 2.125338417540515e-06, "loss": 1.367, "step": 15342 }, { "epoch": 2.92, "learning_rate": 2.1252812202097235e-06, "loss": 2.6476, "step": 15343 }, { "epoch": 2.92, "learning_rate": 2.1252240228789323e-06, "loss": 3.2508, "step": 15344 }, { "epoch": 2.92, "learning_rate": 2.125166825548141e-06, "loss": 1.6064, "step": 15345 }, { "epoch": 2.92, "learning_rate": 2.12510962821735e-06, "loss": 1.6647, "step": 15346 }, { "epoch": 2.92, "learning_rate": 2.1250524308865587e-06, "loss": 2.2022, "step": 15347 }, { "epoch": 2.92, "learning_rate": 2.1249952335557675e-06, "loss": 3.2134, "step": 15348 }, { "epoch": 2.92, "learning_rate": 2.1249380362249763e-06, "loss": 2.0735, "step": 15349 }, { "epoch": 2.92, "learning_rate": 2.1248808388941847e-06, "loss": 3.2735, "step": 15350 }, { "epoch": 2.92, "learning_rate": 2.124823641563394e-06, "loss": 2.2269, "step": 15351 }, { "epoch": 2.92, "learning_rate": 2.1247664442326027e-06, "loss": 3.0395, "step": 15352 }, { "epoch": 2.92, "learning_rate": 2.1247092469018115e-06, "loss": 2.7714, "step": 15353 }, { "epoch": 2.92, "learning_rate": 2.1246520495710203e-06, "loss": 3.5886, "step": 15354 }, { "epoch": 2.92, "learning_rate": 2.1245948522402286e-06, "loss": 3.0277, "step": 15355 }, { "epoch": 2.92, "learning_rate": 2.124537654909438e-06, "loss": 2.7991, "step": 15356 }, { "epoch": 2.93, "learning_rate": 2.124480457578646e-06, "loss": 2.5333, "step": 15357 }, { "epoch": 2.93, "learning_rate": 2.124423260247855e-06, "loss": 2.2918, "step": 15358 }, { "epoch": 2.93, "learning_rate": 2.1243660629170642e-06, "loss": 2.14, "step": 15359 }, { "epoch": 2.93, "learning_rate": 2.1243088655862726e-06, "loss": 3.3706, "step": 15360 }, { "epoch": 2.93, "learning_rate": 2.1242516682554814e-06, "loss": 2.6665, "step": 15361 }, { "epoch": 2.93, "learning_rate": 2.12419447092469e-06, "loss": 1.9117, "step": 15362 }, { "epoch": 2.93, "learning_rate": 2.124137273593899e-06, "loss": 2.5286, "step": 15363 }, { "epoch": 2.93, "learning_rate": 2.1240800762631078e-06, "loss": 1.8689, "step": 15364 }, { "epoch": 2.93, "learning_rate": 2.1240228789323166e-06, "loss": 2.6369, "step": 15365 }, { "epoch": 2.93, "learning_rate": 2.1239656816015254e-06, "loss": 3.5226, "step": 15366 }, { "epoch": 2.93, "learning_rate": 2.123908484270734e-06, "loss": 1.7484, "step": 15367 }, { "epoch": 2.93, "learning_rate": 2.123851286939943e-06, "loss": 3.5898, "step": 15368 }, { "epoch": 2.93, "learning_rate": 2.1237940896091517e-06, "loss": 2.75, "step": 15369 }, { "epoch": 2.93, "learning_rate": 2.1237368922783605e-06, "loss": 3.4557, "step": 15370 }, { "epoch": 2.93, "learning_rate": 2.1236796949475693e-06, "loss": 2.2879, "step": 15371 }, { "epoch": 2.93, "learning_rate": 2.1236224976167777e-06, "loss": 2.952, "step": 15372 }, { "epoch": 2.93, "learning_rate": 2.123565300285987e-06, "loss": 2.9063, "step": 15373 }, { "epoch": 2.93, "learning_rate": 2.1235081029551953e-06, "loss": 2.8301, "step": 15374 }, { "epoch": 2.93, "learning_rate": 2.123450905624404e-06, "loss": 1.6374, "step": 15375 }, { "epoch": 2.93, "learning_rate": 2.1233937082936133e-06, "loss": 2.8757, "step": 15376 }, { "epoch": 2.93, "learning_rate": 2.1233365109628217e-06, "loss": 2.3736, "step": 15377 }, { "epoch": 2.93, "learning_rate": 2.1232793136320305e-06, "loss": 2.1201, "step": 15378 }, { "epoch": 2.93, "learning_rate": 2.1232221163012392e-06, "loss": 2.8394, "step": 15379 }, { "epoch": 2.93, "learning_rate": 2.123164918970448e-06, "loss": 2.6904, "step": 15380 }, { "epoch": 2.93, "learning_rate": 2.123107721639657e-06, "loss": 1.3571, "step": 15381 }, { "epoch": 2.93, "learning_rate": 2.1230505243088656e-06, "loss": 1.4889, "step": 15382 }, { "epoch": 2.93, "learning_rate": 2.1229933269780744e-06, "loss": 2.1565, "step": 15383 }, { "epoch": 2.93, "learning_rate": 2.1229361296472832e-06, "loss": 2.7344, "step": 15384 }, { "epoch": 2.93, "learning_rate": 2.122878932316492e-06, "loss": 3.0118, "step": 15385 }, { "epoch": 2.93, "learning_rate": 2.122821734985701e-06, "loss": 2.8195, "step": 15386 }, { "epoch": 2.93, "learning_rate": 2.1227645376549096e-06, "loss": 3.1595, "step": 15387 }, { "epoch": 2.93, "learning_rate": 2.1227073403241184e-06, "loss": 2.2212, "step": 15388 }, { "epoch": 2.93, "learning_rate": 2.1226501429933268e-06, "loss": 1.4565, "step": 15389 }, { "epoch": 2.93, "learning_rate": 2.122592945662536e-06, "loss": 2.5073, "step": 15390 }, { "epoch": 2.93, "learning_rate": 2.1225357483317448e-06, "loss": 2.8981, "step": 15391 }, { "epoch": 2.93, "learning_rate": 2.122478551000953e-06, "loss": 3.0912, "step": 15392 }, { "epoch": 2.93, "learning_rate": 2.1224213536701624e-06, "loss": 2.6003, "step": 15393 }, { "epoch": 2.93, "learning_rate": 2.1223641563393707e-06, "loss": 2.8467, "step": 15394 }, { "epoch": 2.93, "learning_rate": 2.1223069590085795e-06, "loss": 2.9207, "step": 15395 }, { "epoch": 2.93, "learning_rate": 2.1222497616777883e-06, "loss": 1.9803, "step": 15396 }, { "epoch": 2.93, "learning_rate": 2.122192564346997e-06, "loss": 3.3629, "step": 15397 }, { "epoch": 2.93, "learning_rate": 2.122135367016206e-06, "loss": 2.5969, "step": 15398 }, { "epoch": 2.93, "learning_rate": 2.1220781696854147e-06, "loss": 3.3278, "step": 15399 }, { "epoch": 2.93, "learning_rate": 2.1220209723546235e-06, "loss": 2.3994, "step": 15400 }, { "epoch": 2.93, "learning_rate": 2.1219637750238323e-06, "loss": 2.0029, "step": 15401 }, { "epoch": 2.93, "learning_rate": 2.121906577693041e-06, "loss": 3.2458, "step": 15402 }, { "epoch": 2.93, "learning_rate": 2.12184938036225e-06, "loss": 2.8289, "step": 15403 }, { "epoch": 2.93, "learning_rate": 2.1217921830314587e-06, "loss": 3.2511, "step": 15404 }, { "epoch": 2.93, "learning_rate": 2.1217349857006675e-06, "loss": 2.0163, "step": 15405 }, { "epoch": 2.93, "learning_rate": 2.121677788369876e-06, "loss": 2.4684, "step": 15406 }, { "epoch": 2.93, "learning_rate": 2.121620591039085e-06, "loss": 2.6694, "step": 15407 }, { "epoch": 2.93, "learning_rate": 2.121563393708294e-06, "loss": 2.7321, "step": 15408 }, { "epoch": 2.94, "learning_rate": 2.1215061963775022e-06, "loss": 3.0327, "step": 15409 }, { "epoch": 2.94, "learning_rate": 2.1214489990467114e-06, "loss": 1.8658, "step": 15410 }, { "epoch": 2.94, "learning_rate": 2.12139180171592e-06, "loss": 2.3846, "step": 15411 }, { "epoch": 2.94, "learning_rate": 2.1213346043851286e-06, "loss": 2.9241, "step": 15412 }, { "epoch": 2.94, "learning_rate": 2.1212774070543374e-06, "loss": 3.0955, "step": 15413 }, { "epoch": 2.94, "learning_rate": 2.121220209723546e-06, "loss": 2.0107, "step": 15414 }, { "epoch": 2.94, "learning_rate": 2.121163012392755e-06, "loss": 1.5394, "step": 15415 }, { "epoch": 2.94, "learning_rate": 2.1211058150619638e-06, "loss": 2.4484, "step": 15416 }, { "epoch": 2.94, "learning_rate": 2.1210486177311726e-06, "loss": 3.6487, "step": 15417 }, { "epoch": 2.94, "learning_rate": 2.1209914204003814e-06, "loss": 2.7175, "step": 15418 }, { "epoch": 2.94, "learning_rate": 2.12093422306959e-06, "loss": 2.5731, "step": 15419 }, { "epoch": 2.94, "learning_rate": 2.120877025738799e-06, "loss": 2.2888, "step": 15420 }, { "epoch": 2.94, "learning_rate": 2.1208198284080077e-06, "loss": 1.7532, "step": 15421 }, { "epoch": 2.94, "learning_rate": 2.1207626310772165e-06, "loss": 3.1263, "step": 15422 }, { "epoch": 2.94, "learning_rate": 2.120705433746425e-06, "loss": 3.0034, "step": 15423 }, { "epoch": 2.94, "learning_rate": 2.120648236415634e-06, "loss": 1.7571, "step": 15424 }, { "epoch": 2.94, "learning_rate": 2.120591039084843e-06, "loss": 3.403, "step": 15425 }, { "epoch": 2.94, "learning_rate": 2.1205338417540513e-06, "loss": 3.7562, "step": 15426 }, { "epoch": 2.94, "learning_rate": 2.1204766444232605e-06, "loss": 1.9885, "step": 15427 }, { "epoch": 2.94, "learning_rate": 2.120419447092469e-06, "loss": 3.3769, "step": 15428 }, { "epoch": 2.94, "learning_rate": 2.1203622497616777e-06, "loss": 2.4229, "step": 15429 }, { "epoch": 2.94, "learning_rate": 2.120305052430887e-06, "loss": 2.409, "step": 15430 }, { "epoch": 2.94, "learning_rate": 2.1202478551000953e-06, "loss": 2.9065, "step": 15431 }, { "epoch": 2.94, "learning_rate": 2.1201906577693045e-06, "loss": 2.6368, "step": 15432 }, { "epoch": 2.94, "learning_rate": 2.120133460438513e-06, "loss": 3.8151, "step": 15433 }, { "epoch": 2.94, "learning_rate": 2.1200762631077216e-06, "loss": 3.3725, "step": 15434 }, { "epoch": 2.94, "learning_rate": 2.1200190657769304e-06, "loss": 2.2333, "step": 15435 }, { "epoch": 2.94, "learning_rate": 2.1199618684461392e-06, "loss": 2.3026, "step": 15436 }, { "epoch": 2.94, "learning_rate": 2.119904671115348e-06, "loss": 3.258, "step": 15437 }, { "epoch": 2.94, "learning_rate": 2.119847473784557e-06, "loss": 2.5144, "step": 15438 }, { "epoch": 2.94, "learning_rate": 2.1197902764537656e-06, "loss": 2.274, "step": 15439 }, { "epoch": 2.94, "learning_rate": 2.1197330791229744e-06, "loss": 2.2341, "step": 15440 }, { "epoch": 2.94, "learning_rate": 2.119675881792183e-06, "loss": 2.1322, "step": 15441 }, { "epoch": 2.94, "learning_rate": 2.119618684461392e-06, "loss": 2.5048, "step": 15442 }, { "epoch": 2.94, "learning_rate": 2.1195614871306004e-06, "loss": 2.8707, "step": 15443 }, { "epoch": 2.94, "learning_rate": 2.1195042897998096e-06, "loss": 2.9252, "step": 15444 }, { "epoch": 2.94, "learning_rate": 2.119447092469018e-06, "loss": 2.963, "step": 15445 }, { "epoch": 2.94, "learning_rate": 2.1193898951382267e-06, "loss": 2.0007, "step": 15446 }, { "epoch": 2.94, "learning_rate": 2.119332697807436e-06, "loss": 3.0174, "step": 15447 }, { "epoch": 2.94, "learning_rate": 2.1192755004766443e-06, "loss": 1.8624, "step": 15448 }, { "epoch": 2.94, "learning_rate": 2.1192183031458535e-06, "loss": 2.185, "step": 15449 }, { "epoch": 2.94, "learning_rate": 2.119161105815062e-06, "loss": 3.026, "step": 15450 }, { "epoch": 2.94, "learning_rate": 2.1191039084842707e-06, "loss": 1.8178, "step": 15451 }, { "epoch": 2.94, "learning_rate": 2.1190467111534795e-06, "loss": 2.4715, "step": 15452 }, { "epoch": 2.94, "learning_rate": 2.1189895138226883e-06, "loss": 2.1213, "step": 15453 }, { "epoch": 2.94, "learning_rate": 2.118932316491897e-06, "loss": 2.788, "step": 15454 }, { "epoch": 2.94, "learning_rate": 2.118875119161106e-06, "loss": 2.0991, "step": 15455 }, { "epoch": 2.94, "learning_rate": 2.1188179218303147e-06, "loss": 2.3993, "step": 15456 }, { "epoch": 2.94, "learning_rate": 2.1187607244995235e-06, "loss": 3.112, "step": 15457 }, { "epoch": 2.94, "learning_rate": 2.1187035271687323e-06, "loss": 3.5876, "step": 15458 }, { "epoch": 2.94, "learning_rate": 2.118646329837941e-06, "loss": 2.7506, "step": 15459 }, { "epoch": 2.94, "learning_rate": 2.1185891325071494e-06, "loss": 3.4495, "step": 15460 }, { "epoch": 2.94, "learning_rate": 2.1185319351763586e-06, "loss": 2.0707, "step": 15461 }, { "epoch": 2.95, "learning_rate": 2.118474737845567e-06, "loss": 2.3136, "step": 15462 }, { "epoch": 2.95, "learning_rate": 2.118417540514776e-06, "loss": 1.7797, "step": 15463 }, { "epoch": 2.95, "learning_rate": 2.118360343183985e-06, "loss": 3.5604, "step": 15464 }, { "epoch": 2.95, "learning_rate": 2.1183031458531934e-06, "loss": 2.8501, "step": 15465 }, { "epoch": 2.95, "learning_rate": 2.1182459485224026e-06, "loss": 2.9491, "step": 15466 }, { "epoch": 2.95, "learning_rate": 2.118188751191611e-06, "loss": 2.4399, "step": 15467 }, { "epoch": 2.95, "learning_rate": 2.1181315538608198e-06, "loss": 2.8612, "step": 15468 }, { "epoch": 2.95, "learning_rate": 2.118074356530029e-06, "loss": 2.9576, "step": 15469 }, { "epoch": 2.95, "learning_rate": 2.1180171591992374e-06, "loss": 2.329, "step": 15470 }, { "epoch": 2.95, "learning_rate": 2.117959961868446e-06, "loss": 2.1658, "step": 15471 }, { "epoch": 2.95, "learning_rate": 2.117902764537655e-06, "loss": 2.1502, "step": 15472 }, { "epoch": 2.95, "learning_rate": 2.1178455672068638e-06, "loss": 2.6942, "step": 15473 }, { "epoch": 2.95, "learning_rate": 2.1177883698760725e-06, "loss": 2.2587, "step": 15474 }, { "epoch": 2.95, "learning_rate": 2.1177311725452813e-06, "loss": 2.9136, "step": 15475 }, { "epoch": 2.95, "learning_rate": 2.11767397521449e-06, "loss": 2.8713, "step": 15476 }, { "epoch": 2.95, "learning_rate": 2.1176167778836985e-06, "loss": 2.4251, "step": 15477 }, { "epoch": 2.95, "learning_rate": 2.1175595805529077e-06, "loss": 1.9242, "step": 15478 }, { "epoch": 2.95, "learning_rate": 2.1175023832221165e-06, "loss": 2.6098, "step": 15479 }, { "epoch": 2.95, "learning_rate": 2.117445185891325e-06, "loss": 1.8562, "step": 15480 }, { "epoch": 2.95, "learning_rate": 2.117387988560534e-06, "loss": 1.9501, "step": 15481 }, { "epoch": 2.95, "learning_rate": 2.1173307912297425e-06, "loss": 2.0354, "step": 15482 }, { "epoch": 2.95, "learning_rate": 2.1172735938989517e-06, "loss": 1.6592, "step": 15483 }, { "epoch": 2.95, "learning_rate": 2.11721639656816e-06, "loss": 2.5129, "step": 15484 }, { "epoch": 2.95, "learning_rate": 2.117159199237369e-06, "loss": 2.6223, "step": 15485 }, { "epoch": 2.95, "learning_rate": 2.117102001906578e-06, "loss": 2.1101, "step": 15486 }, { "epoch": 2.95, "learning_rate": 2.1170448045757864e-06, "loss": 3.1313, "step": 15487 }, { "epoch": 2.95, "learning_rate": 2.1169876072449952e-06, "loss": 2.5437, "step": 15488 }, { "epoch": 2.95, "learning_rate": 2.116930409914204e-06, "loss": 2.0915, "step": 15489 }, { "epoch": 2.95, "learning_rate": 2.116873212583413e-06, "loss": 3.58, "step": 15490 }, { "epoch": 2.95, "learning_rate": 2.1168160152526216e-06, "loss": 2.8568, "step": 15491 }, { "epoch": 2.95, "learning_rate": 2.1167588179218304e-06, "loss": 3.255, "step": 15492 }, { "epoch": 2.95, "learning_rate": 2.116701620591039e-06, "loss": 2.9195, "step": 15493 }, { "epoch": 2.95, "learning_rate": 2.1166444232602476e-06, "loss": 1.7967, "step": 15494 }, { "epoch": 2.95, "learning_rate": 2.116587225929457e-06, "loss": 2.135, "step": 15495 }, { "epoch": 2.95, "learning_rate": 2.1165300285986656e-06, "loss": 2.0359, "step": 15496 }, { "epoch": 2.95, "learning_rate": 2.116472831267874e-06, "loss": 2.4534, "step": 15497 }, { "epoch": 2.95, "learning_rate": 2.116415633937083e-06, "loss": 3.3803, "step": 15498 }, { "epoch": 2.95, "learning_rate": 2.1163584366062915e-06, "loss": 2.903, "step": 15499 }, { "epoch": 2.95, "learning_rate": 2.1163012392755008e-06, "loss": 2.0424, "step": 15500 }, { "epoch": 2.95, "learning_rate": 2.116244041944709e-06, "loss": 2.0151, "step": 15501 }, { "epoch": 2.95, "learning_rate": 2.116186844613918e-06, "loss": 2.4753, "step": 15502 }, { "epoch": 2.95, "learning_rate": 2.116129647283127e-06, "loss": 3.5038, "step": 15503 }, { "epoch": 2.95, "learning_rate": 2.1160724499523355e-06, "loss": 1.8967, "step": 15504 }, { "epoch": 2.95, "learning_rate": 2.1160152526215443e-06, "loss": 2.4666, "step": 15505 }, { "epoch": 2.95, "learning_rate": 2.115958055290753e-06, "loss": 2.2373, "step": 15506 }, { "epoch": 2.95, "learning_rate": 2.115900857959962e-06, "loss": 2.2167, "step": 15507 }, { "epoch": 2.95, "learning_rate": 2.1158436606291707e-06, "loss": 2.2829, "step": 15508 }, { "epoch": 2.95, "learning_rate": 2.1157864632983795e-06, "loss": 2.2768, "step": 15509 }, { "epoch": 2.95, "learning_rate": 2.1157292659675883e-06, "loss": 1.8778, "step": 15510 }, { "epoch": 2.95, "learning_rate": 2.1156720686367966e-06, "loss": 2.6087, "step": 15511 }, { "epoch": 2.95, "learning_rate": 2.115614871306006e-06, "loss": 2.914, "step": 15512 }, { "epoch": 2.95, "learning_rate": 2.1155576739752147e-06, "loss": 2.233, "step": 15513 }, { "epoch": 2.96, "learning_rate": 2.1155004766444235e-06, "loss": 2.7274, "step": 15514 }, { "epoch": 2.96, "learning_rate": 2.1154432793136322e-06, "loss": 2.0993, "step": 15515 }, { "epoch": 2.96, "learning_rate": 2.1153860819828406e-06, "loss": 3.19, "step": 15516 }, { "epoch": 2.96, "learning_rate": 2.11532888465205e-06, "loss": 4.2922, "step": 15517 }, { "epoch": 2.96, "learning_rate": 2.1152716873212586e-06, "loss": 2.488, "step": 15518 }, { "epoch": 2.96, "learning_rate": 2.115214489990467e-06, "loss": 1.5322, "step": 15519 }, { "epoch": 2.96, "learning_rate": 2.1151572926596762e-06, "loss": 2.477, "step": 15520 }, { "epoch": 2.96, "learning_rate": 2.1151000953288846e-06, "loss": 2.5743, "step": 15521 }, { "epoch": 2.96, "learning_rate": 2.1150428979980934e-06, "loss": 2.7548, "step": 15522 }, { "epoch": 2.96, "learning_rate": 2.114985700667302e-06, "loss": 2.0775, "step": 15523 }, { "epoch": 2.96, "learning_rate": 2.114928503336511e-06, "loss": 2.222, "step": 15524 }, { "epoch": 2.96, "learning_rate": 2.1148713060057198e-06, "loss": 2.9843, "step": 15525 }, { "epoch": 2.96, "learning_rate": 2.1148141086749286e-06, "loss": 2.5213, "step": 15526 }, { "epoch": 2.96, "learning_rate": 2.1147569113441373e-06, "loss": 2.3399, "step": 15527 }, { "epoch": 2.96, "learning_rate": 2.114699714013346e-06, "loss": 2.0055, "step": 15528 }, { "epoch": 2.96, "learning_rate": 2.114642516682555e-06, "loss": 2.5497, "step": 15529 }, { "epoch": 2.96, "learning_rate": 2.1145853193517637e-06, "loss": 1.8014, "step": 15530 }, { "epoch": 2.96, "learning_rate": 2.1145281220209725e-06, "loss": 2.5963, "step": 15531 }, { "epoch": 2.96, "learning_rate": 2.1144709246901813e-06, "loss": 2.1096, "step": 15532 }, { "epoch": 2.96, "learning_rate": 2.1144137273593897e-06, "loss": 2.3457, "step": 15533 }, { "epoch": 2.96, "learning_rate": 2.114356530028599e-06, "loss": 2.0259, "step": 15534 }, { "epoch": 2.96, "learning_rate": 2.1142993326978077e-06, "loss": 3.4635, "step": 15535 }, { "epoch": 2.96, "learning_rate": 2.114242135367016e-06, "loss": 2.7237, "step": 15536 }, { "epoch": 2.96, "learning_rate": 2.1141849380362253e-06, "loss": 2.7979, "step": 15537 }, { "epoch": 2.96, "learning_rate": 2.1141277407054337e-06, "loss": 2.7829, "step": 15538 }, { "epoch": 2.96, "learning_rate": 2.1140705433746424e-06, "loss": 2.4928, "step": 15539 }, { "epoch": 2.96, "learning_rate": 2.1140133460438512e-06, "loss": 2.3728, "step": 15540 }, { "epoch": 2.96, "learning_rate": 2.11395614871306e-06, "loss": 2.1575, "step": 15541 }, { "epoch": 2.96, "learning_rate": 2.113898951382269e-06, "loss": 3.4917, "step": 15542 }, { "epoch": 2.96, "learning_rate": 2.1138417540514776e-06, "loss": 3.0692, "step": 15543 }, { "epoch": 2.96, "learning_rate": 2.1137845567206864e-06, "loss": 2.9615, "step": 15544 }, { "epoch": 2.96, "learning_rate": 2.1137273593898952e-06, "loss": 2.6745, "step": 15545 }, { "epoch": 2.96, "learning_rate": 2.113670162059104e-06, "loss": 2.7864, "step": 15546 }, { "epoch": 2.96, "learning_rate": 2.113612964728313e-06, "loss": 2.0483, "step": 15547 }, { "epoch": 2.96, "learning_rate": 2.1135557673975216e-06, "loss": 2.0189, "step": 15548 }, { "epoch": 2.96, "learning_rate": 2.1134985700667304e-06, "loss": 2.8561, "step": 15549 }, { "epoch": 2.96, "learning_rate": 2.1134413727359388e-06, "loss": 2.4235, "step": 15550 }, { "epoch": 2.96, "learning_rate": 2.113384175405148e-06, "loss": 3.1542, "step": 15551 }, { "epoch": 2.96, "learning_rate": 2.1133269780743568e-06, "loss": 2.1875, "step": 15552 }, { "epoch": 2.96, "learning_rate": 2.113269780743565e-06, "loss": 2.2827, "step": 15553 }, { "epoch": 2.96, "learning_rate": 2.1132125834127744e-06, "loss": 3.4208, "step": 15554 }, { "epoch": 2.96, "learning_rate": 2.1131553860819827e-06, "loss": 2.4532, "step": 15555 }, { "epoch": 2.96, "learning_rate": 2.1130981887511915e-06, "loss": 1.7279, "step": 15556 }, { "epoch": 2.96, "learning_rate": 2.1130409914204003e-06, "loss": 2.7889, "step": 15557 }, { "epoch": 2.96, "learning_rate": 2.112983794089609e-06, "loss": 2.5057, "step": 15558 }, { "epoch": 2.96, "learning_rate": 2.112926596758818e-06, "loss": 2.957, "step": 15559 }, { "epoch": 2.96, "learning_rate": 2.1128693994280267e-06, "loss": 2.4334, "step": 15560 }, { "epoch": 2.96, "learning_rate": 2.1128122020972355e-06, "loss": 2.6059, "step": 15561 }, { "epoch": 2.96, "learning_rate": 2.1127550047664443e-06, "loss": 3.5703, "step": 15562 }, { "epoch": 2.96, "learning_rate": 2.112697807435653e-06, "loss": 2.4225, "step": 15563 }, { "epoch": 2.96, "learning_rate": 2.112640610104862e-06, "loss": 2.574, "step": 15564 }, { "epoch": 2.96, "learning_rate": 2.1125834127740707e-06, "loss": 2.1388, "step": 15565 }, { "epoch": 2.96, "learning_rate": 2.1125262154432795e-06, "loss": 2.1245, "step": 15566 }, { "epoch": 2.97, "learning_rate": 2.1124690181124883e-06, "loss": 2.7873, "step": 15567 }, { "epoch": 2.97, "learning_rate": 2.112411820781697e-06, "loss": 2.5887, "step": 15568 }, { "epoch": 2.97, "learning_rate": 2.112354623450906e-06, "loss": 2.1633, "step": 15569 }, { "epoch": 2.97, "learning_rate": 2.112297426120114e-06, "loss": 2.2736, "step": 15570 }, { "epoch": 2.97, "learning_rate": 2.1122402287893234e-06, "loss": 2.3174, "step": 15571 }, { "epoch": 2.97, "learning_rate": 2.112183031458532e-06, "loss": 2.7151, "step": 15572 }, { "epoch": 2.97, "learning_rate": 2.1121258341277406e-06, "loss": 2.4434, "step": 15573 }, { "epoch": 2.97, "learning_rate": 2.11206863679695e-06, "loss": 3.2005, "step": 15574 }, { "epoch": 2.97, "learning_rate": 2.112011439466158e-06, "loss": 2.9567, "step": 15575 }, { "epoch": 2.97, "learning_rate": 2.111954242135367e-06, "loss": 2.6173, "step": 15576 }, { "epoch": 2.97, "learning_rate": 2.1118970448045758e-06, "loss": 3.2533, "step": 15577 }, { "epoch": 2.97, "learning_rate": 2.1118398474737846e-06, "loss": 2.5265, "step": 15578 }, { "epoch": 2.97, "learning_rate": 2.1117826501429934e-06, "loss": 2.9214, "step": 15579 }, { "epoch": 2.97, "learning_rate": 2.111725452812202e-06, "loss": 1.8084, "step": 15580 }, { "epoch": 2.97, "learning_rate": 2.111668255481411e-06, "loss": 2.8922, "step": 15581 }, { "epoch": 2.97, "learning_rate": 2.1116110581506197e-06, "loss": 3.3128, "step": 15582 }, { "epoch": 2.97, "learning_rate": 2.1115538608198285e-06, "loss": 3.5633, "step": 15583 }, { "epoch": 2.97, "learning_rate": 2.1114966634890373e-06, "loss": 3.2253, "step": 15584 }, { "epoch": 2.97, "learning_rate": 2.111439466158246e-06, "loss": 2.8821, "step": 15585 }, { "epoch": 2.97, "learning_rate": 2.111382268827455e-06, "loss": 2.4662, "step": 15586 }, { "epoch": 2.97, "learning_rate": 2.1113250714966633e-06, "loss": 2.2097, "step": 15587 }, { "epoch": 2.97, "learning_rate": 2.1112678741658725e-06, "loss": 3.1084, "step": 15588 }, { "epoch": 2.97, "learning_rate": 2.111210676835081e-06, "loss": 2.107, "step": 15589 }, { "epoch": 2.97, "learning_rate": 2.1111534795042897e-06, "loss": 2.293, "step": 15590 }, { "epoch": 2.97, "learning_rate": 2.111096282173499e-06, "loss": 4.0553, "step": 15591 }, { "epoch": 2.97, "learning_rate": 2.1110390848427073e-06, "loss": 2.4029, "step": 15592 }, { "epoch": 2.97, "learning_rate": 2.110981887511916e-06, "loss": 2.1089, "step": 15593 }, { "epoch": 2.97, "learning_rate": 2.110924690181125e-06, "loss": 3.4506, "step": 15594 }, { "epoch": 2.97, "learning_rate": 2.1108674928503336e-06, "loss": 3.2746, "step": 15595 }, { "epoch": 2.97, "learning_rate": 2.1108102955195424e-06, "loss": 1.935, "step": 15596 }, { "epoch": 2.97, "learning_rate": 2.1107530981887512e-06, "loss": 2.4034, "step": 15597 }, { "epoch": 2.97, "learning_rate": 2.11069590085796e-06, "loss": 2.4164, "step": 15598 }, { "epoch": 2.97, "learning_rate": 2.110638703527169e-06, "loss": 1.9696, "step": 15599 }, { "epoch": 2.97, "learning_rate": 2.1105815061963776e-06, "loss": 3.5231, "step": 15600 }, { "epoch": 2.97, "learning_rate": 2.1105243088655864e-06, "loss": 1.801, "step": 15601 }, { "epoch": 2.97, "learning_rate": 2.110467111534795e-06, "loss": 3.6496, "step": 15602 }, { "epoch": 2.97, "learning_rate": 2.110409914204004e-06, "loss": 3.0896, "step": 15603 }, { "epoch": 2.97, "learning_rate": 2.1103527168732124e-06, "loss": 2.4679, "step": 15604 }, { "epoch": 2.97, "learning_rate": 2.1102955195424216e-06, "loss": 2.2689, "step": 15605 }, { "epoch": 2.97, "learning_rate": 2.1102383222116304e-06, "loss": 2.6826, "step": 15606 }, { "epoch": 2.97, "learning_rate": 2.1101811248808387e-06, "loss": 1.9544, "step": 15607 }, { "epoch": 2.97, "learning_rate": 2.110123927550048e-06, "loss": 2.5334, "step": 15608 }, { "epoch": 2.97, "learning_rate": 2.1100667302192563e-06, "loss": 3.2368, "step": 15609 }, { "epoch": 2.97, "learning_rate": 2.110009532888465e-06, "loss": 2.4302, "step": 15610 }, { "epoch": 2.97, "learning_rate": 2.109952335557674e-06, "loss": 3.0875, "step": 15611 }, { "epoch": 2.97, "learning_rate": 2.1098951382268827e-06, "loss": 1.7357, "step": 15612 }, { "epoch": 2.97, "learning_rate": 2.109837940896092e-06, "loss": 2.8303, "step": 15613 }, { "epoch": 2.97, "learning_rate": 2.1097807435653003e-06, "loss": 2.6745, "step": 15614 }, { "epoch": 2.97, "learning_rate": 2.109723546234509e-06, "loss": 2.832, "step": 15615 }, { "epoch": 2.97, "learning_rate": 2.109666348903718e-06, "loss": 1.3475, "step": 15616 }, { "epoch": 2.97, "learning_rate": 2.1096091515729267e-06, "loss": 2.8508, "step": 15617 }, { "epoch": 2.97, "learning_rate": 2.1095519542421355e-06, "loss": 2.6727, "step": 15618 }, { "epoch": 2.98, "learning_rate": 2.1094947569113443e-06, "loss": 2.4408, "step": 15619 }, { "epoch": 2.98, "learning_rate": 2.109437559580553e-06, "loss": 2.7371, "step": 15620 }, { "epoch": 2.98, "learning_rate": 2.1093803622497614e-06, "loss": 2.8295, "step": 15621 }, { "epoch": 2.98, "learning_rate": 2.1093231649189706e-06, "loss": 1.4335, "step": 15622 }, { "epoch": 2.98, "learning_rate": 2.1092659675881794e-06, "loss": 1.9409, "step": 15623 }, { "epoch": 2.98, "learning_rate": 2.109208770257388e-06, "loss": 3.1292, "step": 15624 }, { "epoch": 2.98, "learning_rate": 2.109151572926597e-06, "loss": 2.262, "step": 15625 }, { "epoch": 2.98, "learning_rate": 2.1090943755958054e-06, "loss": 1.0926, "step": 15626 }, { "epoch": 2.98, "learning_rate": 2.109037178265014e-06, "loss": 2.3819, "step": 15627 }, { "epoch": 2.98, "learning_rate": 2.108979980934223e-06, "loss": 2.7409, "step": 15628 }, { "epoch": 2.98, "learning_rate": 2.1089227836034318e-06, "loss": 3.3301, "step": 15629 }, { "epoch": 2.98, "learning_rate": 2.108865586272641e-06, "loss": 2.7971, "step": 15630 }, { "epoch": 2.98, "learning_rate": 2.1088083889418494e-06, "loss": 2.1561, "step": 15631 }, { "epoch": 2.98, "learning_rate": 2.108751191611058e-06, "loss": 2.4771, "step": 15632 }, { "epoch": 2.98, "learning_rate": 2.108693994280267e-06, "loss": 3.3146, "step": 15633 }, { "epoch": 2.98, "learning_rate": 2.1086367969494757e-06, "loss": 2.4494, "step": 15634 }, { "epoch": 2.98, "learning_rate": 2.1085795996186845e-06, "loss": 2.4786, "step": 15635 }, { "epoch": 2.98, "learning_rate": 2.1085224022878933e-06, "loss": 2.4603, "step": 15636 }, { "epoch": 2.98, "learning_rate": 2.108465204957102e-06, "loss": 3.1373, "step": 15637 }, { "epoch": 2.98, "learning_rate": 2.1084080076263105e-06, "loss": 2.8665, "step": 15638 }, { "epoch": 2.98, "learning_rate": 2.1083508102955197e-06, "loss": 2.8039, "step": 15639 }, { "epoch": 2.98, "learning_rate": 2.1082936129647285e-06, "loss": 1.6812, "step": 15640 }, { "epoch": 2.98, "learning_rate": 2.108236415633937e-06, "loss": 2.4986, "step": 15641 }, { "epoch": 2.98, "learning_rate": 2.108179218303146e-06, "loss": 2.0607, "step": 15642 }, { "epoch": 2.98, "learning_rate": 2.1081220209723545e-06, "loss": 2.0373, "step": 15643 }, { "epoch": 2.98, "learning_rate": 2.1080648236415637e-06, "loss": 2.3996, "step": 15644 }, { "epoch": 2.98, "learning_rate": 2.108007626310772e-06, "loss": 2.3061, "step": 15645 }, { "epoch": 2.98, "learning_rate": 2.107950428979981e-06, "loss": 2.7878, "step": 15646 }, { "epoch": 2.98, "learning_rate": 2.10789323164919e-06, "loss": 2.397, "step": 15647 }, { "epoch": 2.98, "learning_rate": 2.1078360343183984e-06, "loss": 2.8728, "step": 15648 }, { "epoch": 2.98, "learning_rate": 2.1077788369876072e-06, "loss": 2.6053, "step": 15649 }, { "epoch": 2.98, "learning_rate": 2.107721639656816e-06, "loss": 3.5645, "step": 15650 }, { "epoch": 2.98, "learning_rate": 2.107664442326025e-06, "loss": 2.6683, "step": 15651 }, { "epoch": 2.98, "learning_rate": 2.1076072449952336e-06, "loss": 2.7061, "step": 15652 }, { "epoch": 2.98, "learning_rate": 2.1075500476644424e-06, "loss": 2.4795, "step": 15653 }, { "epoch": 2.98, "learning_rate": 2.107492850333651e-06, "loss": 2.4448, "step": 15654 }, { "epoch": 2.98, "learning_rate": 2.10743565300286e-06, "loss": 2.6322, "step": 15655 }, { "epoch": 2.98, "learning_rate": 2.1073784556720688e-06, "loss": 1.9041, "step": 15656 }, { "epoch": 2.98, "learning_rate": 2.1073212583412776e-06, "loss": 3.4944, "step": 15657 }, { "epoch": 2.98, "learning_rate": 2.107264061010486e-06, "loss": 3.1261, "step": 15658 }, { "epoch": 2.98, "learning_rate": 2.107206863679695e-06, "loss": 2.1672, "step": 15659 }, { "epoch": 2.98, "learning_rate": 2.1071496663489035e-06, "loss": 1.21, "step": 15660 }, { "epoch": 2.98, "learning_rate": 2.1070924690181128e-06, "loss": 2.7735, "step": 15661 }, { "epoch": 2.98, "learning_rate": 2.1070352716873215e-06, "loss": 2.5469, "step": 15662 }, { "epoch": 2.98, "learning_rate": 2.10697807435653e-06, "loss": 2.8401, "step": 15663 }, { "epoch": 2.98, "learning_rate": 2.106920877025739e-06, "loss": 2.5598, "step": 15664 }, { "epoch": 2.98, "learning_rate": 2.1068636796949475e-06, "loss": 1.6791, "step": 15665 }, { "epoch": 2.98, "learning_rate": 2.1068064823641563e-06, "loss": 2.6774, "step": 15666 }, { "epoch": 2.98, "learning_rate": 2.106749285033365e-06, "loss": 2.8876, "step": 15667 }, { "epoch": 2.98, "learning_rate": 2.106692087702574e-06, "loss": 2.6867, "step": 15668 }, { "epoch": 2.98, "learning_rate": 2.1066348903717827e-06, "loss": 2.4878, "step": 15669 }, { "epoch": 2.98, "learning_rate": 2.1065776930409915e-06, "loss": 1.4682, "step": 15670 }, { "epoch": 2.98, "learning_rate": 2.1065204957102003e-06, "loss": 2.2527, "step": 15671 }, { "epoch": 2.99, "learning_rate": 2.106463298379409e-06, "loss": 2.5586, "step": 15672 }, { "epoch": 2.99, "learning_rate": 2.106406101048618e-06, "loss": 3.4321, "step": 15673 }, { "epoch": 2.99, "learning_rate": 2.1063489037178266e-06, "loss": 2.8443, "step": 15674 }, { "epoch": 2.99, "learning_rate": 2.106291706387035e-06, "loss": 1.2757, "step": 15675 }, { "epoch": 2.99, "learning_rate": 2.1062345090562442e-06, "loss": 2.2843, "step": 15676 }, { "epoch": 2.99, "learning_rate": 2.1061773117254526e-06, "loss": 2.1899, "step": 15677 }, { "epoch": 2.99, "learning_rate": 2.106120114394662e-06, "loss": 3.5528, "step": 15678 }, { "epoch": 2.99, "learning_rate": 2.1060629170638706e-06, "loss": 3.1035, "step": 15679 }, { "epoch": 2.99, "learning_rate": 2.106005719733079e-06, "loss": 2.5288, "step": 15680 }, { "epoch": 2.99, "learning_rate": 2.105948522402288e-06, "loss": 2.0602, "step": 15681 }, { "epoch": 2.99, "learning_rate": 2.1058913250714966e-06, "loss": 1.7629, "step": 15682 }, { "epoch": 2.99, "learning_rate": 2.1058341277407054e-06, "loss": 1.711, "step": 15683 }, { "epoch": 2.99, "learning_rate": 2.105776930409914e-06, "loss": 1.6218, "step": 15684 }, { "epoch": 2.99, "learning_rate": 2.105719733079123e-06, "loss": 3.5351, "step": 15685 }, { "epoch": 2.99, "learning_rate": 2.1056625357483318e-06, "loss": 2.9888, "step": 15686 }, { "epoch": 2.99, "learning_rate": 2.1056053384175405e-06, "loss": 2.5582, "step": 15687 }, { "epoch": 2.99, "learning_rate": 2.1055481410867493e-06, "loss": 3.1969, "step": 15688 }, { "epoch": 2.99, "learning_rate": 2.105490943755958e-06, "loss": 2.1365, "step": 15689 }, { "epoch": 2.99, "learning_rate": 2.105433746425167e-06, "loss": 1.4043, "step": 15690 }, { "epoch": 2.99, "learning_rate": 2.1053765490943757e-06, "loss": 3.0795, "step": 15691 }, { "epoch": 2.99, "learning_rate": 2.105319351763584e-06, "loss": 2.205, "step": 15692 }, { "epoch": 2.99, "learning_rate": 2.1052621544327933e-06, "loss": 2.0967, "step": 15693 }, { "epoch": 2.99, "learning_rate": 2.105204957102002e-06, "loss": 2.6284, "step": 15694 }, { "epoch": 2.99, "learning_rate": 2.105147759771211e-06, "loss": 2.2549, "step": 15695 }, { "epoch": 2.99, "learning_rate": 2.1050905624404197e-06, "loss": 1.9027, "step": 15696 }, { "epoch": 2.99, "learning_rate": 2.105033365109628e-06, "loss": 3.7279, "step": 15697 }, { "epoch": 2.99, "learning_rate": 2.1049761677788373e-06, "loss": 1.6581, "step": 15698 }, { "epoch": 2.99, "learning_rate": 2.1049189704480456e-06, "loss": 2.434, "step": 15699 }, { "epoch": 2.99, "learning_rate": 2.1048617731172544e-06, "loss": 2.5627, "step": 15700 }, { "epoch": 2.99, "learning_rate": 2.1048045757864637e-06, "loss": 1.7471, "step": 15701 }, { "epoch": 2.99, "learning_rate": 2.104747378455672e-06, "loss": 2.2595, "step": 15702 }, { "epoch": 2.99, "learning_rate": 2.104690181124881e-06, "loss": 2.8747, "step": 15703 }, { "epoch": 2.99, "learning_rate": 2.1046329837940896e-06, "loss": 2.7823, "step": 15704 }, { "epoch": 2.99, "learning_rate": 2.1045757864632984e-06, "loss": 3.6021, "step": 15705 }, { "epoch": 2.99, "learning_rate": 2.104518589132507e-06, "loss": 2.633, "step": 15706 }, { "epoch": 2.99, "learning_rate": 2.104461391801716e-06, "loss": 3.7705, "step": 15707 }, { "epoch": 2.99, "learning_rate": 2.104404194470925e-06, "loss": 3.0336, "step": 15708 }, { "epoch": 2.99, "learning_rate": 2.1043469971401336e-06, "loss": 2.6979, "step": 15709 }, { "epoch": 2.99, "learning_rate": 2.1042897998093424e-06, "loss": 2.1363, "step": 15710 }, { "epoch": 2.99, "learning_rate": 2.104232602478551e-06, "loss": 2.2605, "step": 15711 }, { "epoch": 2.99, "learning_rate": 2.10417540514776e-06, "loss": 2.4408, "step": 15712 }, { "epoch": 2.99, "learning_rate": 2.1041182078169688e-06, "loss": 2.6595, "step": 15713 }, { "epoch": 2.99, "learning_rate": 2.104061010486177e-06, "loss": 2.1825, "step": 15714 }, { "epoch": 2.99, "learning_rate": 2.1040038131553863e-06, "loss": 2.1364, "step": 15715 }, { "epoch": 2.99, "learning_rate": 2.1039466158245947e-06, "loss": 2.0747, "step": 15716 }, { "epoch": 2.99, "learning_rate": 2.1038894184938035e-06, "loss": 2.485, "step": 15717 }, { "epoch": 2.99, "learning_rate": 2.1038322211630127e-06, "loss": 2.6469, "step": 15718 }, { "epoch": 2.99, "learning_rate": 2.103775023832221e-06, "loss": 1.9025, "step": 15719 }, { "epoch": 2.99, "learning_rate": 2.10371782650143e-06, "loss": 2.1717, "step": 15720 }, { "epoch": 2.99, "learning_rate": 2.1036606291706387e-06, "loss": 2.4941, "step": 15721 }, { "epoch": 2.99, "learning_rate": 2.1036034318398475e-06, "loss": 2.7125, "step": 15722 }, { "epoch": 2.99, "learning_rate": 2.1035462345090563e-06, "loss": 1.9716, "step": 15723 }, { "epoch": 3.0, "learning_rate": 2.103489037178265e-06, "loss": 2.4299, "step": 15724 }, { "epoch": 3.0, "learning_rate": 2.103431839847474e-06, "loss": 2.576, "step": 15725 }, { "epoch": 3.0, "learning_rate": 2.1033746425166827e-06, "loss": 2.8574, "step": 15726 }, { "epoch": 3.0, "learning_rate": 2.1033174451858915e-06, "loss": 1.5706, "step": 15727 }, { "epoch": 3.0, "learning_rate": 2.1032602478551002e-06, "loss": 2.7418, "step": 15728 }, { "epoch": 3.0, "learning_rate": 2.103203050524309e-06, "loss": 2.2799, "step": 15729 }, { "epoch": 3.0, "learning_rate": 2.103145853193518e-06, "loss": 2.2405, "step": 15730 }, { "epoch": 3.0, "learning_rate": 2.103088655862726e-06, "loss": 1.8258, "step": 15731 }, { "epoch": 3.0, "learning_rate": 2.1030314585319354e-06, "loss": 2.5095, "step": 15732 }, { "epoch": 3.0, "learning_rate": 2.102974261201144e-06, "loss": 2.7163, "step": 15733 }, { "epoch": 3.0, "learning_rate": 2.1029170638703526e-06, "loss": 2.2311, "step": 15734 }, { "epoch": 3.0, "learning_rate": 2.102859866539562e-06, "loss": 3.8078, "step": 15735 }, { "epoch": 3.0, "learning_rate": 2.10280266920877e-06, "loss": 1.728, "step": 15736 }, { "epoch": 3.0, "learning_rate": 2.102745471877979e-06, "loss": 1.5281, "step": 15737 }, { "epoch": 3.0, "learning_rate": 2.1026882745471878e-06, "loss": 2.2222, "step": 15738 }, { "epoch": 3.0, "learning_rate": 2.1026310772163966e-06, "loss": 3.5239, "step": 15739 }, { "epoch": 3.0, "learning_rate": 2.1025738798856053e-06, "loss": 3.275, "step": 15740 }, { "epoch": 3.0, "learning_rate": 2.102516682554814e-06, "loss": 1.8786, "step": 15741 }, { "epoch": 3.0, "learning_rate": 2.102459485224023e-06, "loss": 2.7758, "step": 15742 }, { "epoch": 3.0, "learning_rate": 2.1024022878932317e-06, "loss": 2.6896, "step": 15743 }, { "epoch": 3.0, "learning_rate": 2.1023450905624405e-06, "loss": 2.4462, "step": 15744 }, { "epoch": 3.0, "learning_rate": 2.1022878932316493e-06, "loss": 2.1572, "step": 15745 }, { "epoch": 3.0, "learning_rate": 2.102230695900858e-06, "loss": 3.4842, "step": 15746 }, { "epoch": 3.0, "learning_rate": 2.102173498570067e-06, "loss": 1.6057, "step": 15747 }, { "epoch": 3.0, "learning_rate": 2.1021163012392753e-06, "loss": 3.2226, "step": 15748 }, { "epoch": 3.0, "learning_rate": 2.1020591039084845e-06, "loss": 2.2246, "step": 15749 }, { "epoch": 3.0, "learning_rate": 2.1020019065776933e-06, "loss": 3.5558, "step": 15750 }, { "epoch": 3.0, "eval_accuracy": 0.3596228032576082, "eval_loss": 2.431623935699463, "eval_runtime": 125.2054, "eval_samples_per_second": 18.633, "eval_steps_per_second": 4.664, "step": 15750 }, { "epoch": 3.0, "learning_rate": 2.1019447092469017e-06, "loss": 3.3614, "step": 15751 }, { "epoch": 3.0, "learning_rate": 2.101887511916111e-06, "loss": 3.0599, "step": 15752 }, { "epoch": 3.0, "learning_rate": 2.1018303145853192e-06, "loss": 3.4827, "step": 15753 }, { "epoch": 3.0, "learning_rate": 2.101773117254528e-06, "loss": 2.6301, "step": 15754 }, { "epoch": 3.0, "learning_rate": 2.101715919923737e-06, "loss": 2.5774, "step": 15755 }, { "epoch": 3.0, "learning_rate": 2.1016587225929456e-06, "loss": 3.3713, "step": 15756 }, { "epoch": 3.0, "learning_rate": 2.1016015252621544e-06, "loss": 3.3031, "step": 15757 }, { "epoch": 3.0, "learning_rate": 2.1015443279313632e-06, "loss": 1.6124, "step": 15758 }, { "epoch": 3.0, "learning_rate": 2.101487130600572e-06, "loss": 2.6125, "step": 15759 }, { "epoch": 3.0, "learning_rate": 2.101429933269781e-06, "loss": 3.8195, "step": 15760 }, { "epoch": 3.0, "learning_rate": 2.1013727359389896e-06, "loss": 3.5009, "step": 15761 }, { "epoch": 3.0, "learning_rate": 2.1013155386081984e-06, "loss": 3.0313, "step": 15762 }, { "epoch": 3.0, "learning_rate": 2.101258341277407e-06, "loss": 2.802, "step": 15763 }, { "epoch": 3.0, "learning_rate": 2.101201143946616e-06, "loss": 2.7842, "step": 15764 }, { "epoch": 3.0, "learning_rate": 2.1011439466158243e-06, "loss": 2.6833, "step": 15765 }, { "epoch": 3.0, "learning_rate": 2.1010867492850336e-06, "loss": 3.3914, "step": 15766 }, { "epoch": 3.0, "learning_rate": 2.1010295519542424e-06, "loss": 2.2655, "step": 15767 }, { "epoch": 3.0, "learning_rate": 2.1009723546234507e-06, "loss": 2.3809, "step": 15768 }, { "epoch": 3.0, "learning_rate": 2.10091515729266e-06, "loss": 2.4876, "step": 15769 }, { "epoch": 3.0, "learning_rate": 2.1008579599618683e-06, "loss": 2.4943, "step": 15770 }, { "epoch": 3.0, "learning_rate": 2.100800762631077e-06, "loss": 2.5328, "step": 15771 }, { "epoch": 3.0, "learning_rate": 2.100743565300286e-06, "loss": 2.2634, "step": 15772 }, { "epoch": 3.0, "learning_rate": 2.1006863679694947e-06, "loss": 1.1636, "step": 15773 }, { "epoch": 3.0, "learning_rate": 2.100629170638704e-06, "loss": 2.6362, "step": 15774 }, { "epoch": 3.0, "learning_rate": 2.1005719733079123e-06, "loss": 2.0973, "step": 15775 }, { "epoch": 3.0, "learning_rate": 2.100514775977121e-06, "loss": 2.4896, "step": 15776 }, { "epoch": 3.01, "learning_rate": 2.10045757864633e-06, "loss": 2.3837, "step": 15777 }, { "epoch": 3.01, "learning_rate": 2.1004003813155387e-06, "loss": 1.6701, "step": 15778 }, { "epoch": 3.01, "learning_rate": 2.1003431839847475e-06, "loss": 2.7867, "step": 15779 }, { "epoch": 3.01, "learning_rate": 2.1002859866539563e-06, "loss": 3.1096, "step": 15780 }, { "epoch": 3.01, "learning_rate": 2.100228789323165e-06, "loss": 2.7573, "step": 15781 }, { "epoch": 3.01, "learning_rate": 2.100171591992374e-06, "loss": 1.8603, "step": 15782 }, { "epoch": 3.01, "learning_rate": 2.1001143946615826e-06, "loss": 2.5103, "step": 15783 }, { "epoch": 3.01, "learning_rate": 2.1000571973307914e-06, "loss": 2.3607, "step": 15784 }, { "epoch": 3.01, "learning_rate": 2.1e-06, "loss": 1.8255, "step": 15785 }, { "epoch": 3.01, "learning_rate": 2.099942802669209e-06, "loss": 3.1736, "step": 15786 }, { "epoch": 3.01, "learning_rate": 2.0998856053384174e-06, "loss": 2.8962, "step": 15787 }, { "epoch": 3.01, "learning_rate": 2.099828408007626e-06, "loss": 2.3764, "step": 15788 }, { "epoch": 3.01, "learning_rate": 2.0997712106768354e-06, "loss": 2.6265, "step": 15789 }, { "epoch": 3.01, "learning_rate": 2.0997140133460438e-06, "loss": 2.5419, "step": 15790 }, { "epoch": 3.01, "learning_rate": 2.099656816015253e-06, "loss": 2.4557, "step": 15791 }, { "epoch": 3.01, "learning_rate": 2.0995996186844614e-06, "loss": 3.0475, "step": 15792 }, { "epoch": 3.01, "learning_rate": 2.09954242135367e-06, "loss": 2.6742, "step": 15793 }, { "epoch": 3.01, "learning_rate": 2.099485224022879e-06, "loss": 2.6597, "step": 15794 }, { "epoch": 3.01, "learning_rate": 2.0994280266920877e-06, "loss": 2.1801, "step": 15795 }, { "epoch": 3.01, "learning_rate": 2.0993708293612965e-06, "loss": 2.5708, "step": 15796 }, { "epoch": 3.01, "learning_rate": 2.0993136320305053e-06, "loss": 2.0175, "step": 15797 }, { "epoch": 3.01, "learning_rate": 2.099256434699714e-06, "loss": 1.7056, "step": 15798 }, { "epoch": 3.01, "learning_rate": 2.099199237368923e-06, "loss": 2.8472, "step": 15799 }, { "epoch": 3.01, "learning_rate": 2.0991420400381317e-06, "loss": 2.2625, "step": 15800 }, { "epoch": 3.01, "learning_rate": 2.0990848427073405e-06, "loss": 3.0871, "step": 15801 }, { "epoch": 3.01, "learning_rate": 2.099027645376549e-06, "loss": 1.9914, "step": 15802 }, { "epoch": 3.01, "learning_rate": 2.098970448045758e-06, "loss": 2.16, "step": 15803 }, { "epoch": 3.01, "learning_rate": 2.0989132507149665e-06, "loss": 1.6234, "step": 15804 }, { "epoch": 3.01, "learning_rate": 2.0988560533841753e-06, "loss": 1.7069, "step": 15805 }, { "epoch": 3.01, "learning_rate": 2.0987988560533845e-06, "loss": 1.5281, "step": 15806 }, { "epoch": 3.01, "learning_rate": 2.098741658722593e-06, "loss": 2.2257, "step": 15807 }, { "epoch": 3.01, "learning_rate": 2.098684461391802e-06, "loss": 3.2065, "step": 15808 }, { "epoch": 3.01, "learning_rate": 2.0986272640610104e-06, "loss": 1.7675, "step": 15809 }, { "epoch": 3.01, "learning_rate": 2.0985700667302192e-06, "loss": 2.857, "step": 15810 }, { "epoch": 3.01, "learning_rate": 2.098512869399428e-06, "loss": 2.276, "step": 15811 }, { "epoch": 3.01, "learning_rate": 2.098455672068637e-06, "loss": 2.216, "step": 15812 }, { "epoch": 3.01, "learning_rate": 2.0983984747378456e-06, "loss": 2.6776, "step": 15813 }, { "epoch": 3.01, "learning_rate": 2.0983412774070544e-06, "loss": 1.9676, "step": 15814 }, { "epoch": 3.01, "learning_rate": 2.098284080076263e-06, "loss": 1.8246, "step": 15815 }, { "epoch": 3.01, "learning_rate": 2.098226882745472e-06, "loss": 2.6183, "step": 15816 }, { "epoch": 3.01, "learning_rate": 2.0981696854146808e-06, "loss": 2.6381, "step": 15817 }, { "epoch": 3.01, "learning_rate": 2.0981124880838896e-06, "loss": 2.0479, "step": 15818 }, { "epoch": 3.01, "learning_rate": 2.098055290753098e-06, "loss": 2.4868, "step": 15819 }, { "epoch": 3.01, "learning_rate": 2.097998093422307e-06, "loss": 1.6472, "step": 15820 }, { "epoch": 3.01, "learning_rate": 2.0979408960915155e-06, "loss": 1.9752, "step": 15821 }, { "epoch": 3.01, "learning_rate": 2.0978836987607243e-06, "loss": 2.5573, "step": 15822 }, { "epoch": 3.01, "learning_rate": 2.0978265014299335e-06, "loss": 2.0825, "step": 15823 }, { "epoch": 3.01, "learning_rate": 2.097769304099142e-06, "loss": 2.3238, "step": 15824 }, { "epoch": 3.01, "learning_rate": 2.097712106768351e-06, "loss": 2.031, "step": 15825 }, { "epoch": 3.01, "learning_rate": 2.0976549094375595e-06, "loss": 1.3897, "step": 15826 }, { "epoch": 3.01, "learning_rate": 2.0975977121067683e-06, "loss": 3.3521, "step": 15827 }, { "epoch": 3.01, "learning_rate": 2.0975405147759775e-06, "loss": 3.0299, "step": 15828 }, { "epoch": 3.02, "learning_rate": 2.097483317445186e-06, "loss": 1.2069, "step": 15829 }, { "epoch": 3.02, "learning_rate": 2.0974261201143947e-06, "loss": 2.4888, "step": 15830 }, { "epoch": 3.02, "learning_rate": 2.0973689227836035e-06, "loss": 1.2791, "step": 15831 }, { "epoch": 3.02, "learning_rate": 2.0973117254528123e-06, "loss": 2.4039, "step": 15832 }, { "epoch": 3.02, "learning_rate": 2.097254528122021e-06, "loss": 3.2717, "step": 15833 }, { "epoch": 3.02, "learning_rate": 2.09719733079123e-06, "loss": 1.8259, "step": 15834 }, { "epoch": 3.02, "learning_rate": 2.0971401334604386e-06, "loss": 1.5393, "step": 15835 }, { "epoch": 3.02, "learning_rate": 2.097082936129647e-06, "loss": 1.8098, "step": 15836 }, { "epoch": 3.02, "learning_rate": 2.0970257387988562e-06, "loss": 2.1657, "step": 15837 }, { "epoch": 3.02, "learning_rate": 2.096968541468065e-06, "loss": 1.46, "step": 15838 }, { "epoch": 3.02, "learning_rate": 2.096911344137274e-06, "loss": 2.6633, "step": 15839 }, { "epoch": 3.02, "learning_rate": 2.0968541468064826e-06, "loss": 2.1948, "step": 15840 }, { "epoch": 3.02, "learning_rate": 2.096796949475691e-06, "loss": 3.7704, "step": 15841 }, { "epoch": 3.02, "learning_rate": 2.0967397521449e-06, "loss": 2.3282, "step": 15842 }, { "epoch": 3.02, "learning_rate": 2.0966825548141086e-06, "loss": 1.3376, "step": 15843 }, { "epoch": 3.02, "learning_rate": 2.0966253574833174e-06, "loss": 3.1395, "step": 15844 }, { "epoch": 3.02, "learning_rate": 2.0965681601525266e-06, "loss": 2.7396, "step": 15845 }, { "epoch": 3.02, "learning_rate": 2.096510962821735e-06, "loss": 1.4029, "step": 15846 }, { "epoch": 3.02, "learning_rate": 2.0964537654909437e-06, "loss": 3.7318, "step": 15847 }, { "epoch": 3.02, "learning_rate": 2.0963965681601525e-06, "loss": 2.2423, "step": 15848 }, { "epoch": 3.02, "learning_rate": 2.0963393708293613e-06, "loss": 3.1681, "step": 15849 }, { "epoch": 3.02, "learning_rate": 2.09628217349857e-06, "loss": 2.4838, "step": 15850 }, { "epoch": 3.02, "learning_rate": 2.096224976167779e-06, "loss": 3.0342, "step": 15851 }, { "epoch": 3.02, "learning_rate": 2.0961677788369877e-06, "loss": 3.0286, "step": 15852 }, { "epoch": 3.02, "learning_rate": 2.096110581506196e-06, "loss": 2.068, "step": 15853 }, { "epoch": 3.02, "learning_rate": 2.0960533841754053e-06, "loss": 3.0762, "step": 15854 }, { "epoch": 3.02, "learning_rate": 2.095996186844614e-06, "loss": 2.3226, "step": 15855 }, { "epoch": 3.02, "learning_rate": 2.095938989513823e-06, "loss": 2.4495, "step": 15856 }, { "epoch": 3.02, "learning_rate": 2.0958817921830317e-06, "loss": 2.4117, "step": 15857 }, { "epoch": 3.02, "learning_rate": 2.09582459485224e-06, "loss": 2.9006, "step": 15858 }, { "epoch": 3.02, "learning_rate": 2.0957673975214493e-06, "loss": 2.1161, "step": 15859 }, { "epoch": 3.02, "learning_rate": 2.0957102001906576e-06, "loss": 2.8425, "step": 15860 }, { "epoch": 3.02, "learning_rate": 2.0956530028598664e-06, "loss": 1.4525, "step": 15861 }, { "epoch": 3.02, "learning_rate": 2.0955958055290757e-06, "loss": 1.645, "step": 15862 }, { "epoch": 3.02, "learning_rate": 2.095538608198284e-06, "loss": 2.1566, "step": 15863 }, { "epoch": 3.02, "learning_rate": 2.095481410867493e-06, "loss": 3.2522, "step": 15864 }, { "epoch": 3.02, "learning_rate": 2.0954242135367016e-06, "loss": 3.2442, "step": 15865 }, { "epoch": 3.02, "learning_rate": 2.0953670162059104e-06, "loss": 3.5081, "step": 15866 }, { "epoch": 3.02, "learning_rate": 2.095309818875119e-06, "loss": 2.9285, "step": 15867 }, { "epoch": 3.02, "learning_rate": 2.095252621544328e-06, "loss": 2.261, "step": 15868 }, { "epoch": 3.02, "learning_rate": 2.0951954242135368e-06, "loss": 2.9392, "step": 15869 }, { "epoch": 3.02, "learning_rate": 2.095138226882745e-06, "loss": 2.6499, "step": 15870 }, { "epoch": 3.02, "learning_rate": 2.0950810295519544e-06, "loss": 2.4964, "step": 15871 }, { "epoch": 3.02, "learning_rate": 2.095023832221163e-06, "loss": 2.261, "step": 15872 }, { "epoch": 3.02, "learning_rate": 2.094966634890372e-06, "loss": 2.8601, "step": 15873 }, { "epoch": 3.02, "learning_rate": 2.0949094375595808e-06, "loss": 2.3702, "step": 15874 }, { "epoch": 3.02, "learning_rate": 2.094852240228789e-06, "loss": 3.1278, "step": 15875 }, { "epoch": 3.02, "learning_rate": 2.0947950428979983e-06, "loss": 2.8801, "step": 15876 }, { "epoch": 3.02, "learning_rate": 2.094737845567207e-06, "loss": 2.7927, "step": 15877 }, { "epoch": 3.02, "learning_rate": 2.0946806482364155e-06, "loss": 2.7574, "step": 15878 }, { "epoch": 3.02, "learning_rate": 2.0946234509056247e-06, "loss": 2.0881, "step": 15879 }, { "epoch": 3.02, "learning_rate": 2.094566253574833e-06, "loss": 2.2757, "step": 15880 }, { "epoch": 3.02, "learning_rate": 2.094509056244042e-06, "loss": 2.9973, "step": 15881 }, { "epoch": 3.03, "learning_rate": 2.0944518589132507e-06, "loss": 2.8277, "step": 15882 }, { "epoch": 3.03, "learning_rate": 2.0943946615824595e-06, "loss": 2.4684, "step": 15883 }, { "epoch": 3.03, "learning_rate": 2.0943374642516683e-06, "loss": 2.2193, "step": 15884 }, { "epoch": 3.03, "learning_rate": 2.094280266920877e-06, "loss": 2.6633, "step": 15885 }, { "epoch": 3.03, "learning_rate": 2.094223069590086e-06, "loss": 2.1518, "step": 15886 }, { "epoch": 3.03, "learning_rate": 2.0941658722592947e-06, "loss": 2.7641, "step": 15887 }, { "epoch": 3.03, "learning_rate": 2.0941086749285034e-06, "loss": 2.3078, "step": 15888 }, { "epoch": 3.03, "learning_rate": 2.0940514775977122e-06, "loss": 2.3241, "step": 15889 }, { "epoch": 3.03, "learning_rate": 2.093994280266921e-06, "loss": 1.3069, "step": 15890 }, { "epoch": 3.03, "learning_rate": 2.09393708293613e-06, "loss": 2.358, "step": 15891 }, { "epoch": 3.03, "learning_rate": 2.093879885605338e-06, "loss": 2.1131, "step": 15892 }, { "epoch": 3.03, "learning_rate": 2.0938226882745474e-06, "loss": 3.9452, "step": 15893 }, { "epoch": 3.03, "learning_rate": 2.093765490943756e-06, "loss": 2.0826, "step": 15894 }, { "epoch": 3.03, "learning_rate": 2.0937082936129646e-06, "loss": 2.6442, "step": 15895 }, { "epoch": 3.03, "learning_rate": 2.093651096282174e-06, "loss": 3.0842, "step": 15896 }, { "epoch": 3.03, "learning_rate": 2.093593898951382e-06, "loss": 2.7182, "step": 15897 }, { "epoch": 3.03, "learning_rate": 2.093536701620591e-06, "loss": 2.1374, "step": 15898 }, { "epoch": 3.03, "learning_rate": 2.0934795042897998e-06, "loss": 2.5455, "step": 15899 }, { "epoch": 3.03, "learning_rate": 2.0934223069590085e-06, "loss": 3.0904, "step": 15900 }, { "epoch": 3.03, "learning_rate": 2.0933651096282173e-06, "loss": 1.8758, "step": 15901 }, { "epoch": 3.03, "learning_rate": 2.093307912297426e-06, "loss": 1.7479, "step": 15902 }, { "epoch": 3.03, "learning_rate": 2.093250714966635e-06, "loss": 3.2174, "step": 15903 }, { "epoch": 3.03, "learning_rate": 2.0931935176358437e-06, "loss": 1.8543, "step": 15904 }, { "epoch": 3.03, "learning_rate": 2.0931363203050525e-06, "loss": 2.5015, "step": 15905 }, { "epoch": 3.03, "learning_rate": 2.0930791229742613e-06, "loss": 3.449, "step": 15906 }, { "epoch": 3.03, "learning_rate": 2.09302192564347e-06, "loss": 2.8693, "step": 15907 }, { "epoch": 3.03, "learning_rate": 2.092964728312679e-06, "loss": 1.7613, "step": 15908 }, { "epoch": 3.03, "learning_rate": 2.0929075309818873e-06, "loss": 3.4132, "step": 15909 }, { "epoch": 3.03, "learning_rate": 2.0928503336510965e-06, "loss": 2.5505, "step": 15910 }, { "epoch": 3.03, "learning_rate": 2.0927931363203053e-06, "loss": 2.0095, "step": 15911 }, { "epoch": 3.03, "learning_rate": 2.0927359389895136e-06, "loss": 2.895, "step": 15912 }, { "epoch": 3.03, "learning_rate": 2.092678741658723e-06, "loss": 2.5892, "step": 15913 }, { "epoch": 3.03, "learning_rate": 2.0926215443279312e-06, "loss": 2.57, "step": 15914 }, { "epoch": 3.03, "learning_rate": 2.09256434699714e-06, "loss": 2.0234, "step": 15915 }, { "epoch": 3.03, "learning_rate": 2.0925071496663492e-06, "loss": 1.922, "step": 15916 }, { "epoch": 3.03, "learning_rate": 2.0924499523355576e-06, "loss": 2.2598, "step": 15917 }, { "epoch": 3.03, "learning_rate": 2.0923927550047664e-06, "loss": 1.9287, "step": 15918 }, { "epoch": 3.03, "learning_rate": 2.092335557673975e-06, "loss": 2.2945, "step": 15919 }, { "epoch": 3.03, "learning_rate": 2.092278360343184e-06, "loss": 1.6298, "step": 15920 }, { "epoch": 3.03, "learning_rate": 2.092221163012393e-06, "loss": 2.689, "step": 15921 }, { "epoch": 3.03, "learning_rate": 2.0921639656816016e-06, "loss": 3.8466, "step": 15922 }, { "epoch": 3.03, "learning_rate": 2.0921067683508104e-06, "loss": 2.3712, "step": 15923 }, { "epoch": 3.03, "learning_rate": 2.092049571020019e-06, "loss": 2.7889, "step": 15924 }, { "epoch": 3.03, "learning_rate": 2.091992373689228e-06, "loss": 1.3798, "step": 15925 }, { "epoch": 3.03, "learning_rate": 2.0919351763584368e-06, "loss": 3.1093, "step": 15926 }, { "epoch": 3.03, "learning_rate": 2.0918779790276456e-06, "loss": 1.3997, "step": 15927 }, { "epoch": 3.03, "learning_rate": 2.0918207816968544e-06, "loss": 2.3061, "step": 15928 }, { "epoch": 3.03, "learning_rate": 2.0917635843660627e-06, "loss": 2.4813, "step": 15929 }, { "epoch": 3.03, "learning_rate": 2.091706387035272e-06, "loss": 2.476, "step": 15930 }, { "epoch": 3.03, "learning_rate": 2.0916491897044803e-06, "loss": 2.4451, "step": 15931 }, { "epoch": 3.03, "learning_rate": 2.091591992373689e-06, "loss": 2.5118, "step": 15932 }, { "epoch": 3.03, "learning_rate": 2.0915347950428983e-06, "loss": 2.0057, "step": 15933 }, { "epoch": 3.04, "learning_rate": 2.0914775977121067e-06, "loss": 2.2429, "step": 15934 }, { "epoch": 3.04, "learning_rate": 2.0914204003813155e-06, "loss": 2.0125, "step": 15935 }, { "epoch": 3.04, "learning_rate": 2.0913632030505243e-06, "loss": 2.4093, "step": 15936 }, { "epoch": 3.04, "learning_rate": 2.091306005719733e-06, "loss": 3.1303, "step": 15937 }, { "epoch": 3.04, "learning_rate": 2.091248808388942e-06, "loss": 1.9519, "step": 15938 }, { "epoch": 3.04, "learning_rate": 2.0911916110581507e-06, "loss": 2.9372, "step": 15939 }, { "epoch": 3.04, "learning_rate": 2.0911344137273595e-06, "loss": 3.0386, "step": 15940 }, { "epoch": 3.04, "learning_rate": 2.0910772163965682e-06, "loss": 3.4835, "step": 15941 }, { "epoch": 3.04, "learning_rate": 2.091020019065777e-06, "loss": 2.9114, "step": 15942 }, { "epoch": 3.04, "learning_rate": 2.090962821734986e-06, "loss": 2.2362, "step": 15943 }, { "epoch": 3.04, "learning_rate": 2.0909056244041946e-06, "loss": 2.5021, "step": 15944 }, { "epoch": 3.04, "learning_rate": 2.0908484270734034e-06, "loss": 2.4065, "step": 15945 }, { "epoch": 3.04, "learning_rate": 2.090791229742612e-06, "loss": 3.2252, "step": 15946 }, { "epoch": 3.04, "learning_rate": 2.090734032411821e-06, "loss": 2.0411, "step": 15947 }, { "epoch": 3.04, "learning_rate": 2.0906768350810294e-06, "loss": 2.5037, "step": 15948 }, { "epoch": 3.04, "learning_rate": 2.090619637750238e-06, "loss": 2.5844, "step": 15949 }, { "epoch": 3.04, "learning_rate": 2.0905624404194474e-06, "loss": 2.0814, "step": 15950 }, { "epoch": 3.04, "learning_rate": 2.0905052430886558e-06, "loss": 2.2707, "step": 15951 }, { "epoch": 3.04, "learning_rate": 2.0904480457578646e-06, "loss": 2.1049, "step": 15952 }, { "epoch": 3.04, "learning_rate": 2.0903908484270733e-06, "loss": 2.6675, "step": 15953 }, { "epoch": 3.04, "learning_rate": 2.090333651096282e-06, "loss": 3.0377, "step": 15954 }, { "epoch": 3.04, "learning_rate": 2.090276453765491e-06, "loss": 1.9793, "step": 15955 }, { "epoch": 3.04, "learning_rate": 2.0902192564346997e-06, "loss": 2.7849, "step": 15956 }, { "epoch": 3.04, "learning_rate": 2.0901620591039085e-06, "loss": 2.3036, "step": 15957 }, { "epoch": 3.04, "learning_rate": 2.0901048617731173e-06, "loss": 3.3201, "step": 15958 }, { "epoch": 3.04, "learning_rate": 2.090047664442326e-06, "loss": 2.7918, "step": 15959 }, { "epoch": 3.04, "learning_rate": 2.089990467111535e-06, "loss": 1.9022, "step": 15960 }, { "epoch": 3.04, "learning_rate": 2.0899332697807437e-06, "loss": 2.6036, "step": 15961 }, { "epoch": 3.04, "learning_rate": 2.0898760724499525e-06, "loss": 2.2585, "step": 15962 }, { "epoch": 3.04, "learning_rate": 2.089818875119161e-06, "loss": 3.0216, "step": 15963 }, { "epoch": 3.04, "learning_rate": 2.08976167778837e-06, "loss": 2.7, "step": 15964 }, { "epoch": 3.04, "learning_rate": 2.089704480457579e-06, "loss": 2.8888, "step": 15965 }, { "epoch": 3.04, "learning_rate": 2.0896472831267872e-06, "loss": 3.5366, "step": 15966 }, { "epoch": 3.04, "learning_rate": 2.0895900857959965e-06, "loss": 2.6545, "step": 15967 }, { "epoch": 3.04, "learning_rate": 2.089532888465205e-06, "loss": 2.0772, "step": 15968 }, { "epoch": 3.04, "learning_rate": 2.089475691134414e-06, "loss": 2.2724, "step": 15969 }, { "epoch": 3.04, "learning_rate": 2.0894184938036224e-06, "loss": 2.3191, "step": 15970 }, { "epoch": 3.04, "learning_rate": 2.0893612964728312e-06, "loss": 2.2099, "step": 15971 }, { "epoch": 3.04, "learning_rate": 2.0893040991420404e-06, "loss": 2.2521, "step": 15972 }, { "epoch": 3.04, "learning_rate": 2.089246901811249e-06, "loss": 3.4461, "step": 15973 }, { "epoch": 3.04, "learning_rate": 2.0891897044804576e-06, "loss": 3.3976, "step": 15974 }, { "epoch": 3.04, "learning_rate": 2.0891325071496664e-06, "loss": 2.0131, "step": 15975 }, { "epoch": 3.04, "learning_rate": 2.089075309818875e-06, "loss": 2.0407, "step": 15976 }, { "epoch": 3.04, "learning_rate": 2.089018112488084e-06, "loss": 2.2251, "step": 15977 }, { "epoch": 3.04, "learning_rate": 2.0889609151572928e-06, "loss": 2.2909, "step": 15978 }, { "epoch": 3.04, "learning_rate": 2.0889037178265016e-06, "loss": 2.7324, "step": 15979 }, { "epoch": 3.04, "learning_rate": 2.08884652049571e-06, "loss": 3.499, "step": 15980 }, { "epoch": 3.04, "learning_rate": 2.088789323164919e-06, "loss": 1.6209, "step": 15981 }, { "epoch": 3.04, "learning_rate": 2.088732125834128e-06, "loss": 2.8936, "step": 15982 }, { "epoch": 3.04, "learning_rate": 2.0886749285033363e-06, "loss": 1.5927, "step": 15983 }, { "epoch": 3.04, "learning_rate": 2.0886177311725455e-06, "loss": 2.0401, "step": 15984 }, { "epoch": 3.04, "learning_rate": 2.088560533841754e-06, "loss": 3.2749, "step": 15985 }, { "epoch": 3.04, "learning_rate": 2.088503336510963e-06, "loss": 3.2061, "step": 15986 }, { "epoch": 3.05, "learning_rate": 2.0884461391801715e-06, "loss": 3.7028, "step": 15987 }, { "epoch": 3.05, "learning_rate": 2.0883889418493803e-06, "loss": 3.0749, "step": 15988 }, { "epoch": 3.05, "learning_rate": 2.0883317445185895e-06, "loss": 3.4709, "step": 15989 }, { "epoch": 3.05, "learning_rate": 2.088274547187798e-06, "loss": 2.2205, "step": 15990 }, { "epoch": 3.05, "learning_rate": 2.0882173498570067e-06, "loss": 2.9089, "step": 15991 }, { "epoch": 3.05, "learning_rate": 2.0881601525262155e-06, "loss": 2.4154, "step": 15992 }, { "epoch": 3.05, "learning_rate": 2.0881029551954243e-06, "loss": 2.6395, "step": 15993 }, { "epoch": 3.05, "learning_rate": 2.088045757864633e-06, "loss": 1.4179, "step": 15994 }, { "epoch": 3.05, "learning_rate": 2.087988560533842e-06, "loss": 2.9834, "step": 15995 }, { "epoch": 3.05, "learning_rate": 2.0879313632030506e-06, "loss": 2.694, "step": 15996 }, { "epoch": 3.05, "learning_rate": 2.087874165872259e-06, "loss": 2.5028, "step": 15997 }, { "epoch": 3.05, "learning_rate": 2.0878169685414682e-06, "loss": 3.6262, "step": 15998 }, { "epoch": 3.05, "learning_rate": 2.087759771210677e-06, "loss": 2.1137, "step": 15999 }, { "epoch": 3.05, "learning_rate": 2.0877025738798854e-06, "loss": 2.7932, "step": 16000 }, { "epoch": 3.05, "learning_rate": 2.0876453765490946e-06, "loss": 2.4555, "step": 16001 }, { "epoch": 3.05, "learning_rate": 2.087588179218303e-06, "loss": 2.2277, "step": 16002 }, { "epoch": 3.05, "learning_rate": 2.087530981887512e-06, "loss": 2.4998, "step": 16003 }, { "epoch": 3.05, "learning_rate": 2.087473784556721e-06, "loss": 2.5765, "step": 16004 }, { "epoch": 3.05, "learning_rate": 2.0874165872259294e-06, "loss": 2.2676, "step": 16005 }, { "epoch": 3.05, "learning_rate": 2.0873593898951386e-06, "loss": 3.3173, "step": 16006 }, { "epoch": 3.05, "learning_rate": 2.087302192564347e-06, "loss": 1.9807, "step": 16007 }, { "epoch": 3.05, "learning_rate": 2.0872449952335557e-06, "loss": 3.4694, "step": 16008 }, { "epoch": 3.05, "learning_rate": 2.0871877979027645e-06, "loss": 2.3107, "step": 16009 }, { "epoch": 3.05, "learning_rate": 2.0871306005719733e-06, "loss": 2.1548, "step": 16010 }, { "epoch": 3.05, "learning_rate": 2.087073403241182e-06, "loss": 2.6897, "step": 16011 }, { "epoch": 3.05, "learning_rate": 2.087016205910391e-06, "loss": 2.2759, "step": 16012 }, { "epoch": 3.05, "learning_rate": 2.0869590085795997e-06, "loss": 2.5477, "step": 16013 }, { "epoch": 3.05, "learning_rate": 2.0869018112488085e-06, "loss": 2.3833, "step": 16014 }, { "epoch": 3.05, "learning_rate": 2.0868446139180173e-06, "loss": 1.3493, "step": 16015 }, { "epoch": 3.05, "learning_rate": 2.086787416587226e-06, "loss": 2.8292, "step": 16016 }, { "epoch": 3.05, "learning_rate": 2.0867302192564345e-06, "loss": 1.7671, "step": 16017 }, { "epoch": 3.05, "learning_rate": 2.0866730219256437e-06, "loss": 2.72, "step": 16018 }, { "epoch": 3.05, "learning_rate": 2.086615824594852e-06, "loss": 2.5833, "step": 16019 }, { "epoch": 3.05, "learning_rate": 2.0865586272640613e-06, "loss": 1.9952, "step": 16020 }, { "epoch": 3.05, "learning_rate": 2.08650142993327e-06, "loss": 1.6853, "step": 16021 }, { "epoch": 3.05, "learning_rate": 2.0864442326024784e-06, "loss": 2.6571, "step": 16022 }, { "epoch": 3.05, "learning_rate": 2.0863870352716876e-06, "loss": 1.8896, "step": 16023 }, { "epoch": 3.05, "learning_rate": 2.086329837940896e-06, "loss": 2.1225, "step": 16024 }, { "epoch": 3.05, "learning_rate": 2.086272640610105e-06, "loss": 3.0526, "step": 16025 }, { "epoch": 3.05, "learning_rate": 2.0862154432793136e-06, "loss": 1.5173, "step": 16026 }, { "epoch": 3.05, "learning_rate": 2.0861582459485224e-06, "loss": 1.9986, "step": 16027 }, { "epoch": 3.05, "learning_rate": 2.086101048617731e-06, "loss": 2.0587, "step": 16028 }, { "epoch": 3.05, "learning_rate": 2.08604385128694e-06, "loss": 3.2895, "step": 16029 }, { "epoch": 3.05, "learning_rate": 2.0859866539561488e-06, "loss": 2.2275, "step": 16030 }, { "epoch": 3.05, "learning_rate": 2.0859294566253576e-06, "loss": 3.3245, "step": 16031 }, { "epoch": 3.05, "learning_rate": 2.0858722592945664e-06, "loss": 3.1023, "step": 16032 }, { "epoch": 3.05, "learning_rate": 2.085815061963775e-06, "loss": 3.0656, "step": 16033 }, { "epoch": 3.05, "learning_rate": 2.085757864632984e-06, "loss": 2.5516, "step": 16034 }, { "epoch": 3.05, "learning_rate": 2.0857006673021927e-06, "loss": 2.6733, "step": 16035 }, { "epoch": 3.05, "learning_rate": 2.085643469971401e-06, "loss": 2.1409, "step": 16036 }, { "epoch": 3.05, "learning_rate": 2.0855862726406103e-06, "loss": 3.8077, "step": 16037 }, { "epoch": 3.05, "learning_rate": 2.085529075309819e-06, "loss": 3.0514, "step": 16038 }, { "epoch": 3.06, "learning_rate": 2.0854718779790275e-06, "loss": 2.3123, "step": 16039 }, { "epoch": 3.06, "learning_rate": 2.0854146806482367e-06, "loss": 2.4364, "step": 16040 }, { "epoch": 3.06, "learning_rate": 2.085357483317445e-06, "loss": 3.1763, "step": 16041 }, { "epoch": 3.06, "learning_rate": 2.085300285986654e-06, "loss": 1.9963, "step": 16042 }, { "epoch": 3.06, "learning_rate": 2.0852430886558627e-06, "loss": 3.2094, "step": 16043 }, { "epoch": 3.06, "learning_rate": 2.0851858913250715e-06, "loss": 2.7174, "step": 16044 }, { "epoch": 3.06, "learning_rate": 2.0851286939942803e-06, "loss": 3.423, "step": 16045 }, { "epoch": 3.06, "learning_rate": 2.085071496663489e-06, "loss": 2.0305, "step": 16046 }, { "epoch": 3.06, "learning_rate": 2.085014299332698e-06, "loss": 2.3767, "step": 16047 }, { "epoch": 3.06, "learning_rate": 2.0849571020019066e-06, "loss": 3.0024, "step": 16048 }, { "epoch": 3.06, "learning_rate": 2.0848999046711154e-06, "loss": 3.2457, "step": 16049 }, { "epoch": 3.06, "learning_rate": 2.0848427073403242e-06, "loss": 2.4622, "step": 16050 }, { "epoch": 3.06, "learning_rate": 2.084785510009533e-06, "loss": 3.1464, "step": 16051 }, { "epoch": 3.06, "learning_rate": 2.084728312678742e-06, "loss": 1.7767, "step": 16052 }, { "epoch": 3.06, "learning_rate": 2.0846711153479506e-06, "loss": 1.7571, "step": 16053 }, { "epoch": 3.06, "learning_rate": 2.0846139180171594e-06, "loss": 2.799, "step": 16054 }, { "epoch": 3.06, "learning_rate": 2.084556720686368e-06, "loss": 3.0456, "step": 16055 }, { "epoch": 3.06, "learning_rate": 2.0844995233555766e-06, "loss": 2.2361, "step": 16056 }, { "epoch": 3.06, "learning_rate": 2.084442326024786e-06, "loss": 2.7489, "step": 16057 }, { "epoch": 3.06, "learning_rate": 2.084385128693994e-06, "loss": 3.3468, "step": 16058 }, { "epoch": 3.06, "learning_rate": 2.084327931363203e-06, "loss": 1.9205, "step": 16059 }, { "epoch": 3.06, "learning_rate": 2.084270734032412e-06, "loss": 1.8658, "step": 16060 }, { "epoch": 3.06, "learning_rate": 2.0842135367016205e-06, "loss": 1.8917, "step": 16061 }, { "epoch": 3.06, "learning_rate": 2.0841563393708293e-06, "loss": 2.265, "step": 16062 }, { "epoch": 3.06, "learning_rate": 2.084099142040038e-06, "loss": 3.328, "step": 16063 }, { "epoch": 3.06, "learning_rate": 2.084041944709247e-06, "loss": 2.6538, "step": 16064 }, { "epoch": 3.06, "learning_rate": 2.0839847473784557e-06, "loss": 3.0008, "step": 16065 }, { "epoch": 3.06, "learning_rate": 2.0839275500476645e-06, "loss": 2.2654, "step": 16066 }, { "epoch": 3.06, "learning_rate": 2.0838703527168733e-06, "loss": 1.8263, "step": 16067 }, { "epoch": 3.06, "learning_rate": 2.083813155386082e-06, "loss": 2.6823, "step": 16068 }, { "epoch": 3.06, "learning_rate": 2.083755958055291e-06, "loss": 2.5354, "step": 16069 }, { "epoch": 3.06, "learning_rate": 2.0836987607244997e-06, "loss": 2.4288, "step": 16070 }, { "epoch": 3.06, "learning_rate": 2.0836415633937085e-06, "loss": 2.0613, "step": 16071 }, { "epoch": 3.06, "learning_rate": 2.0835843660629173e-06, "loss": 1.8898, "step": 16072 }, { "epoch": 3.06, "learning_rate": 2.0835271687321256e-06, "loss": 2.26, "step": 16073 }, { "epoch": 3.06, "learning_rate": 2.083469971401335e-06, "loss": 4.1854, "step": 16074 }, { "epoch": 3.06, "learning_rate": 2.0834127740705432e-06, "loss": 1.4512, "step": 16075 }, { "epoch": 3.06, "learning_rate": 2.083355576739752e-06, "loss": 1.4091, "step": 16076 }, { "epoch": 3.06, "learning_rate": 2.0832983794089612e-06, "loss": 2.2024, "step": 16077 }, { "epoch": 3.06, "learning_rate": 2.0832411820781696e-06, "loss": 3.2942, "step": 16078 }, { "epoch": 3.06, "learning_rate": 2.0831839847473784e-06, "loss": 3.2083, "step": 16079 }, { "epoch": 3.06, "learning_rate": 2.083126787416587e-06, "loss": 3.282, "step": 16080 }, { "epoch": 3.06, "learning_rate": 2.083069590085796e-06, "loss": 2.6661, "step": 16081 }, { "epoch": 3.06, "learning_rate": 2.0830123927550048e-06, "loss": 3.2743, "step": 16082 }, { "epoch": 3.06, "learning_rate": 2.0829551954242136e-06, "loss": 2.4228, "step": 16083 }, { "epoch": 3.06, "learning_rate": 2.0828979980934224e-06, "loss": 3.0697, "step": 16084 }, { "epoch": 3.06, "learning_rate": 2.082840800762631e-06, "loss": 3.2059, "step": 16085 }, { "epoch": 3.06, "learning_rate": 2.08278360343184e-06, "loss": 3.2805, "step": 16086 }, { "epoch": 3.06, "learning_rate": 2.0827264061010488e-06, "loss": 2.2716, "step": 16087 }, { "epoch": 3.06, "learning_rate": 2.0826692087702575e-06, "loss": 2.5064, "step": 16088 }, { "epoch": 3.06, "learning_rate": 2.0826120114394663e-06, "loss": 3.1125, "step": 16089 }, { "epoch": 3.06, "learning_rate": 2.0825548141086747e-06, "loss": 2.3329, "step": 16090 }, { "epoch": 3.06, "learning_rate": 2.082497616777884e-06, "loss": 2.1616, "step": 16091 }, { "epoch": 3.07, "learning_rate": 2.0824404194470927e-06, "loss": 3.4259, "step": 16092 }, { "epoch": 3.07, "learning_rate": 2.082383222116301e-06, "loss": 1.6392, "step": 16093 }, { "epoch": 3.07, "learning_rate": 2.0823260247855103e-06, "loss": 2.6813, "step": 16094 }, { "epoch": 3.07, "learning_rate": 2.0822688274547187e-06, "loss": 2.9564, "step": 16095 }, { "epoch": 3.07, "learning_rate": 2.0822116301239275e-06, "loss": 1.4352, "step": 16096 }, { "epoch": 3.07, "learning_rate": 2.0821544327931363e-06, "loss": 2.0756, "step": 16097 }, { "epoch": 3.07, "learning_rate": 2.082097235462345e-06, "loss": 2.4693, "step": 16098 }, { "epoch": 3.07, "learning_rate": 2.0820400381315543e-06, "loss": 2.6508, "step": 16099 }, { "epoch": 3.07, "learning_rate": 2.0819828408007627e-06, "loss": 2.7416, "step": 16100 }, { "epoch": 3.07, "learning_rate": 2.0819256434699714e-06, "loss": 2.4064, "step": 16101 }, { "epoch": 3.07, "learning_rate": 2.0818684461391802e-06, "loss": 2.0515, "step": 16102 }, { "epoch": 3.07, "learning_rate": 2.081811248808389e-06, "loss": 2.085, "step": 16103 }, { "epoch": 3.07, "learning_rate": 2.081754051477598e-06, "loss": 2.2688, "step": 16104 }, { "epoch": 3.07, "learning_rate": 2.0816968541468066e-06, "loss": 3.063, "step": 16105 }, { "epoch": 3.07, "learning_rate": 2.0816396568160154e-06, "loss": 2.0207, "step": 16106 }, { "epoch": 3.07, "learning_rate": 2.0815824594852238e-06, "loss": 2.0962, "step": 16107 }, { "epoch": 3.07, "learning_rate": 2.081525262154433e-06, "loss": 1.8811, "step": 16108 }, { "epoch": 3.07, "learning_rate": 2.081468064823642e-06, "loss": 2.1006, "step": 16109 }, { "epoch": 3.07, "learning_rate": 2.08141086749285e-06, "loss": 2.1913, "step": 16110 }, { "epoch": 3.07, "learning_rate": 2.0813536701620594e-06, "loss": 3.1941, "step": 16111 }, { "epoch": 3.07, "learning_rate": 2.0812964728312678e-06, "loss": 2.7966, "step": 16112 }, { "epoch": 3.07, "learning_rate": 2.0812392755004765e-06, "loss": 3.1283, "step": 16113 }, { "epoch": 3.07, "learning_rate": 2.0811820781696853e-06, "loss": 3.1064, "step": 16114 }, { "epoch": 3.07, "learning_rate": 2.081124880838894e-06, "loss": 3.1835, "step": 16115 }, { "epoch": 3.07, "learning_rate": 2.0810676835081034e-06, "loss": 2.5223, "step": 16116 }, { "epoch": 3.07, "learning_rate": 2.0810104861773117e-06, "loss": 2.9658, "step": 16117 }, { "epoch": 3.07, "learning_rate": 2.0809532888465205e-06, "loss": 2.7502, "step": 16118 }, { "epoch": 3.07, "learning_rate": 2.0808960915157293e-06, "loss": 3.05, "step": 16119 }, { "epoch": 3.07, "learning_rate": 2.080838894184938e-06, "loss": 3.197, "step": 16120 }, { "epoch": 3.07, "learning_rate": 2.080781696854147e-06, "loss": 2.2589, "step": 16121 }, { "epoch": 3.07, "learning_rate": 2.0807244995233557e-06, "loss": 2.2364, "step": 16122 }, { "epoch": 3.07, "learning_rate": 2.0806673021925645e-06, "loss": 2.76, "step": 16123 }, { "epoch": 3.07, "learning_rate": 2.080610104861773e-06, "loss": 3.2038, "step": 16124 }, { "epoch": 3.07, "learning_rate": 2.080552907530982e-06, "loss": 2.7388, "step": 16125 }, { "epoch": 3.07, "learning_rate": 2.080495710200191e-06, "loss": 2.5769, "step": 16126 }, { "epoch": 3.07, "learning_rate": 2.0804385128693992e-06, "loss": 2.0967, "step": 16127 }, { "epoch": 3.07, "learning_rate": 2.0803813155386085e-06, "loss": 4.2213, "step": 16128 }, { "epoch": 3.07, "learning_rate": 2.080324118207817e-06, "loss": 2.7287, "step": 16129 }, { "epoch": 3.07, "learning_rate": 2.0802669208770256e-06, "loss": 3.2304, "step": 16130 }, { "epoch": 3.07, "learning_rate": 2.0802097235462344e-06, "loss": 1.72, "step": 16131 }, { "epoch": 3.07, "learning_rate": 2.080152526215443e-06, "loss": 1.6709, "step": 16132 }, { "epoch": 3.07, "learning_rate": 2.0800953288846524e-06, "loss": 2.4822, "step": 16133 }, { "epoch": 3.07, "learning_rate": 2.080038131553861e-06, "loss": 2.3071, "step": 16134 }, { "epoch": 3.07, "learning_rate": 2.0799809342230696e-06, "loss": 2.5907, "step": 16135 }, { "epoch": 3.07, "learning_rate": 2.0799237368922784e-06, "loss": 2.7006, "step": 16136 }, { "epoch": 3.07, "learning_rate": 2.079866539561487e-06, "loss": 1.9362, "step": 16137 }, { "epoch": 3.07, "learning_rate": 2.079809342230696e-06, "loss": 2.3554, "step": 16138 }, { "epoch": 3.07, "learning_rate": 2.0797521448999048e-06, "loss": 2.6026, "step": 16139 }, { "epoch": 3.07, "learning_rate": 2.0796949475691136e-06, "loss": 2.6593, "step": 16140 }, { "epoch": 3.07, "learning_rate": 2.0796377502383224e-06, "loss": 2.1993, "step": 16141 }, { "epoch": 3.07, "learning_rate": 2.079580552907531e-06, "loss": 1.8082, "step": 16142 }, { "epoch": 3.07, "learning_rate": 2.07952335557674e-06, "loss": 2.9969, "step": 16143 }, { "epoch": 3.08, "learning_rate": 2.0794661582459483e-06, "loss": 3.5887, "step": 16144 }, { "epoch": 3.08, "learning_rate": 2.0794089609151575e-06, "loss": 2.5606, "step": 16145 }, { "epoch": 3.08, "learning_rate": 2.079351763584366e-06, "loss": 2.2964, "step": 16146 }, { "epoch": 3.08, "learning_rate": 2.0792945662535747e-06, "loss": 2.952, "step": 16147 }, { "epoch": 3.08, "learning_rate": 2.079237368922784e-06, "loss": 1.8743, "step": 16148 }, { "epoch": 3.08, "learning_rate": 2.0791801715919923e-06, "loss": 2.6745, "step": 16149 }, { "epoch": 3.08, "learning_rate": 2.0791229742612015e-06, "loss": 3.6512, "step": 16150 }, { "epoch": 3.08, "learning_rate": 2.07906577693041e-06, "loss": 3.284, "step": 16151 }, { "epoch": 3.08, "learning_rate": 2.0790085795996187e-06, "loss": 1.9793, "step": 16152 }, { "epoch": 3.08, "learning_rate": 2.0789513822688275e-06, "loss": 2.9366, "step": 16153 }, { "epoch": 3.08, "learning_rate": 2.0788941849380362e-06, "loss": 2.3223, "step": 16154 }, { "epoch": 3.08, "learning_rate": 2.078836987607245e-06, "loss": 1.6136, "step": 16155 }, { "epoch": 3.08, "learning_rate": 2.078779790276454e-06, "loss": 2.1992, "step": 16156 }, { "epoch": 3.08, "learning_rate": 2.0787225929456626e-06, "loss": 2.1285, "step": 16157 }, { "epoch": 3.08, "learning_rate": 2.0786653956148714e-06, "loss": 3.1639, "step": 16158 }, { "epoch": 3.08, "learning_rate": 2.0786081982840802e-06, "loss": 2.218, "step": 16159 }, { "epoch": 3.08, "learning_rate": 2.078551000953289e-06, "loss": 3.1193, "step": 16160 }, { "epoch": 3.08, "learning_rate": 2.0784938036224974e-06, "loss": 1.8772, "step": 16161 }, { "epoch": 3.08, "learning_rate": 2.0784366062917066e-06, "loss": 2.1341, "step": 16162 }, { "epoch": 3.08, "learning_rate": 2.078379408960915e-06, "loss": 1.7099, "step": 16163 }, { "epoch": 3.08, "learning_rate": 2.0783222116301238e-06, "loss": 2.8314, "step": 16164 }, { "epoch": 3.08, "learning_rate": 2.078265014299333e-06, "loss": 1.8422, "step": 16165 }, { "epoch": 3.08, "learning_rate": 2.0782078169685413e-06, "loss": 4.0116, "step": 16166 }, { "epoch": 3.08, "learning_rate": 2.0781506196377506e-06, "loss": 2.1848, "step": 16167 }, { "epoch": 3.08, "learning_rate": 2.078093422306959e-06, "loss": 1.8726, "step": 16168 }, { "epoch": 3.08, "learning_rate": 2.0780362249761677e-06, "loss": 3.3554, "step": 16169 }, { "epoch": 3.08, "learning_rate": 2.0779790276453765e-06, "loss": 2.7694, "step": 16170 }, { "epoch": 3.08, "learning_rate": 2.0779218303145853e-06, "loss": 1.7971, "step": 16171 }, { "epoch": 3.08, "learning_rate": 2.077864632983794e-06, "loss": 2.2902, "step": 16172 }, { "epoch": 3.08, "learning_rate": 2.077807435653003e-06, "loss": 2.8212, "step": 16173 }, { "epoch": 3.08, "learning_rate": 2.0777502383222117e-06, "loss": 3.217, "step": 16174 }, { "epoch": 3.08, "learning_rate": 2.0776930409914205e-06, "loss": 2.3425, "step": 16175 }, { "epoch": 3.08, "learning_rate": 2.0776358436606293e-06, "loss": 3.2699, "step": 16176 }, { "epoch": 3.08, "learning_rate": 2.077578646329838e-06, "loss": 3.0069, "step": 16177 }, { "epoch": 3.08, "learning_rate": 2.0775214489990465e-06, "loss": 2.7427, "step": 16178 }, { "epoch": 3.08, "learning_rate": 2.0774642516682557e-06, "loss": 2.6718, "step": 16179 }, { "epoch": 3.08, "learning_rate": 2.077407054337464e-06, "loss": 2.8039, "step": 16180 }, { "epoch": 3.08, "learning_rate": 2.0773498570066733e-06, "loss": 2.4257, "step": 16181 }, { "epoch": 3.08, "learning_rate": 2.077292659675882e-06, "loss": 2.2893, "step": 16182 }, { "epoch": 3.08, "learning_rate": 2.0772354623450904e-06, "loss": 2.3367, "step": 16183 }, { "epoch": 3.08, "learning_rate": 2.0771782650142996e-06, "loss": 2.1373, "step": 16184 }, { "epoch": 3.08, "learning_rate": 2.077121067683508e-06, "loss": 2.8713, "step": 16185 }, { "epoch": 3.08, "learning_rate": 2.077063870352717e-06, "loss": 1.8836, "step": 16186 }, { "epoch": 3.08, "learning_rate": 2.077006673021926e-06, "loss": 3.4512, "step": 16187 }, { "epoch": 3.08, "learning_rate": 2.0769494756911344e-06, "loss": 1.7512, "step": 16188 }, { "epoch": 3.08, "learning_rate": 2.076892278360343e-06, "loss": 3.96, "step": 16189 }, { "epoch": 3.08, "learning_rate": 2.076835081029552e-06, "loss": 2.5072, "step": 16190 }, { "epoch": 3.08, "learning_rate": 2.0767778836987608e-06, "loss": 2.48, "step": 16191 }, { "epoch": 3.08, "learning_rate": 2.0767206863679696e-06, "loss": 2.4577, "step": 16192 }, { "epoch": 3.08, "learning_rate": 2.0766634890371784e-06, "loss": 3.3951, "step": 16193 }, { "epoch": 3.08, "learning_rate": 2.076606291706387e-06, "loss": 3.2692, "step": 16194 }, { "epoch": 3.08, "learning_rate": 2.0765490943755955e-06, "loss": 1.4616, "step": 16195 }, { "epoch": 3.08, "learning_rate": 2.0764918970448047e-06, "loss": 1.9543, "step": 16196 }, { "epoch": 3.09, "learning_rate": 2.0764346997140135e-06, "loss": 3.0845, "step": 16197 }, { "epoch": 3.09, "learning_rate": 2.0763775023832223e-06, "loss": 2.0645, "step": 16198 }, { "epoch": 3.09, "learning_rate": 2.076320305052431e-06, "loss": 2.4725, "step": 16199 }, { "epoch": 3.09, "learning_rate": 2.0762631077216395e-06, "loss": 2.0881, "step": 16200 }, { "epoch": 3.09, "learning_rate": 2.0762059103908487e-06, "loss": 1.6684, "step": 16201 }, { "epoch": 3.09, "learning_rate": 2.076148713060057e-06, "loss": 2.3044, "step": 16202 }, { "epoch": 3.09, "learning_rate": 2.076091515729266e-06, "loss": 2.3961, "step": 16203 }, { "epoch": 3.09, "learning_rate": 2.076034318398475e-06, "loss": 2.9377, "step": 16204 }, { "epoch": 3.09, "learning_rate": 2.0759771210676835e-06, "loss": 2.5885, "step": 16205 }, { "epoch": 3.09, "learning_rate": 2.0759199237368923e-06, "loss": 2.7706, "step": 16206 }, { "epoch": 3.09, "learning_rate": 2.075862726406101e-06, "loss": 1.4589, "step": 16207 }, { "epoch": 3.09, "learning_rate": 2.07580552907531e-06, "loss": 2.3294, "step": 16208 }, { "epoch": 3.09, "learning_rate": 2.0757483317445186e-06, "loss": 2.465, "step": 16209 }, { "epoch": 3.09, "learning_rate": 2.0756911344137274e-06, "loss": 2.9453, "step": 16210 }, { "epoch": 3.09, "learning_rate": 2.0756339370829362e-06, "loss": 2.2542, "step": 16211 }, { "epoch": 3.09, "learning_rate": 2.0755767397521446e-06, "loss": 1.9676, "step": 16212 }, { "epoch": 3.09, "learning_rate": 2.075519542421354e-06, "loss": 3.9104, "step": 16213 }, { "epoch": 3.09, "learning_rate": 2.0754623450905626e-06, "loss": 1.7984, "step": 16214 }, { "epoch": 3.09, "learning_rate": 2.0754051477597714e-06, "loss": 2.8749, "step": 16215 }, { "epoch": 3.09, "learning_rate": 2.07534795042898e-06, "loss": 1.7131, "step": 16216 }, { "epoch": 3.09, "learning_rate": 2.0752907530981886e-06, "loss": 2.347, "step": 16217 }, { "epoch": 3.09, "learning_rate": 2.0752335557673978e-06, "loss": 3.1312, "step": 16218 }, { "epoch": 3.09, "learning_rate": 2.075176358436606e-06, "loss": 2.4196, "step": 16219 }, { "epoch": 3.09, "learning_rate": 2.075119161105815e-06, "loss": 2.1983, "step": 16220 }, { "epoch": 3.09, "learning_rate": 2.075061963775024e-06, "loss": 3.3348, "step": 16221 }, { "epoch": 3.09, "learning_rate": 2.0750047664442325e-06, "loss": 1.4461, "step": 16222 }, { "epoch": 3.09, "learning_rate": 2.0749475691134413e-06, "loss": 2.3639, "step": 16223 }, { "epoch": 3.09, "learning_rate": 2.07489037178265e-06, "loss": 1.9999, "step": 16224 }, { "epoch": 3.09, "learning_rate": 2.074833174451859e-06, "loss": 2.8627, "step": 16225 }, { "epoch": 3.09, "learning_rate": 2.0747759771210677e-06, "loss": 2.0865, "step": 16226 }, { "epoch": 3.09, "learning_rate": 2.0747187797902765e-06, "loss": 2.046, "step": 16227 }, { "epoch": 3.09, "learning_rate": 2.0746615824594853e-06, "loss": 1.9588, "step": 16228 }, { "epoch": 3.09, "learning_rate": 2.074604385128694e-06, "loss": 2.8033, "step": 16229 }, { "epoch": 3.09, "learning_rate": 2.074547187797903e-06, "loss": 3.8752, "step": 16230 }, { "epoch": 3.09, "learning_rate": 2.0744899904671117e-06, "loss": 2.0014, "step": 16231 }, { "epoch": 3.09, "learning_rate": 2.0744327931363205e-06, "loss": 1.5423, "step": 16232 }, { "epoch": 3.09, "learning_rate": 2.0743755958055293e-06, "loss": 2.5573, "step": 16233 }, { "epoch": 3.09, "learning_rate": 2.0743183984747376e-06, "loss": 3.1077, "step": 16234 }, { "epoch": 3.09, "learning_rate": 2.074261201143947e-06, "loss": 2.6248, "step": 16235 }, { "epoch": 3.09, "learning_rate": 2.0742040038131556e-06, "loss": 2.2049, "step": 16236 }, { "epoch": 3.09, "learning_rate": 2.074146806482364e-06, "loss": 2.0198, "step": 16237 }, { "epoch": 3.09, "learning_rate": 2.0740896091515732e-06, "loss": 2.5902, "step": 16238 }, { "epoch": 3.09, "learning_rate": 2.0740324118207816e-06, "loss": 2.8549, "step": 16239 }, { "epoch": 3.09, "learning_rate": 2.0739752144899904e-06, "loss": 2.1643, "step": 16240 }, { "epoch": 3.09, "learning_rate": 2.073918017159199e-06, "loss": 3.4594, "step": 16241 }, { "epoch": 3.09, "learning_rate": 2.073860819828408e-06, "loss": 2.3607, "step": 16242 }, { "epoch": 3.09, "learning_rate": 2.0738036224976168e-06, "loss": 2.1193, "step": 16243 }, { "epoch": 3.09, "learning_rate": 2.0737464251668256e-06, "loss": 2.7583, "step": 16244 }, { "epoch": 3.09, "learning_rate": 2.0736892278360344e-06, "loss": 2.4793, "step": 16245 }, { "epoch": 3.09, "learning_rate": 2.073632030505243e-06, "loss": 2.6604, "step": 16246 }, { "epoch": 3.09, "learning_rate": 2.073574833174452e-06, "loss": 2.9838, "step": 16247 }, { "epoch": 3.09, "learning_rate": 2.0735176358436607e-06, "loss": 2.3848, "step": 16248 }, { "epoch": 3.1, "learning_rate": 2.0734604385128695e-06, "loss": 2.8468, "step": 16249 }, { "epoch": 3.1, "learning_rate": 2.0734032411820783e-06, "loss": 1.8562, "step": 16250 }, { "epoch": 3.1, "learning_rate": 2.0733460438512867e-06, "loss": 2.3836, "step": 16251 }, { "epoch": 3.1, "learning_rate": 2.073288846520496e-06, "loss": 2.8297, "step": 16252 }, { "epoch": 3.1, "learning_rate": 2.0732316491897047e-06, "loss": 1.5539, "step": 16253 }, { "epoch": 3.1, "learning_rate": 2.073174451858913e-06, "loss": 4.7047, "step": 16254 }, { "epoch": 3.1, "learning_rate": 2.0731172545281223e-06, "loss": 2.9339, "step": 16255 }, { "epoch": 3.1, "learning_rate": 2.0730600571973307e-06, "loss": 2.6446, "step": 16256 }, { "epoch": 3.1, "learning_rate": 2.0730028598665395e-06, "loss": 1.5639, "step": 16257 }, { "epoch": 3.1, "learning_rate": 2.0729456625357483e-06, "loss": 2.1479, "step": 16258 }, { "epoch": 3.1, "learning_rate": 2.072888465204957e-06, "loss": 1.8997, "step": 16259 }, { "epoch": 3.1, "learning_rate": 2.072831267874166e-06, "loss": 2.1756, "step": 16260 }, { "epoch": 3.1, "learning_rate": 2.0727740705433746e-06, "loss": 1.8033, "step": 16261 }, { "epoch": 3.1, "learning_rate": 2.0727168732125834e-06, "loss": 2.8831, "step": 16262 }, { "epoch": 3.1, "learning_rate": 2.0726596758817922e-06, "loss": 1.7645, "step": 16263 }, { "epoch": 3.1, "learning_rate": 2.072602478551001e-06, "loss": 2.8188, "step": 16264 }, { "epoch": 3.1, "learning_rate": 2.07254528122021e-06, "loss": 2.589, "step": 16265 }, { "epoch": 3.1, "learning_rate": 2.0724880838894186e-06, "loss": 2.2492, "step": 16266 }, { "epoch": 3.1, "learning_rate": 2.0724308865586274e-06, "loss": 1.7794, "step": 16267 }, { "epoch": 3.1, "learning_rate": 2.0723736892278358e-06, "loss": 2.0857, "step": 16268 }, { "epoch": 3.1, "learning_rate": 2.072316491897045e-06, "loss": 2.1527, "step": 16269 }, { "epoch": 3.1, "learning_rate": 2.072259294566254e-06, "loss": 1.9715, "step": 16270 }, { "epoch": 3.1, "learning_rate": 2.072202097235462e-06, "loss": 2.5424, "step": 16271 }, { "epoch": 3.1, "learning_rate": 2.0721448999046714e-06, "loss": 2.392, "step": 16272 }, { "epoch": 3.1, "learning_rate": 2.0720877025738797e-06, "loss": 2.1844, "step": 16273 }, { "epoch": 3.1, "learning_rate": 2.0720305052430885e-06, "loss": 1.4325, "step": 16274 }, { "epoch": 3.1, "learning_rate": 2.0719733079122978e-06, "loss": 2.9555, "step": 16275 }, { "epoch": 3.1, "learning_rate": 2.071916110581506e-06, "loss": 3.0864, "step": 16276 }, { "epoch": 3.1, "learning_rate": 2.071858913250715e-06, "loss": 2.0823, "step": 16277 }, { "epoch": 3.1, "learning_rate": 2.0718017159199237e-06, "loss": 2.4457, "step": 16278 }, { "epoch": 3.1, "learning_rate": 2.0717445185891325e-06, "loss": 3.2149, "step": 16279 }, { "epoch": 3.1, "learning_rate": 2.0716873212583413e-06, "loss": 2.0625, "step": 16280 }, { "epoch": 3.1, "learning_rate": 2.07163012392755e-06, "loss": 2.2264, "step": 16281 }, { "epoch": 3.1, "learning_rate": 2.071572926596759e-06, "loss": 3.3076, "step": 16282 }, { "epoch": 3.1, "learning_rate": 2.0715157292659677e-06, "loss": 2.3701, "step": 16283 }, { "epoch": 3.1, "learning_rate": 2.0714585319351765e-06, "loss": 2.779, "step": 16284 }, { "epoch": 3.1, "learning_rate": 2.0714013346043853e-06, "loss": 3.3579, "step": 16285 }, { "epoch": 3.1, "learning_rate": 2.071344137273594e-06, "loss": 2.0834, "step": 16286 }, { "epoch": 3.1, "learning_rate": 2.071286939942803e-06, "loss": 2.242, "step": 16287 }, { "epoch": 3.1, "learning_rate": 2.0712297426120112e-06, "loss": 2.1465, "step": 16288 }, { "epoch": 3.1, "learning_rate": 2.0711725452812204e-06, "loss": 2.8988, "step": 16289 }, { "epoch": 3.1, "learning_rate": 2.071115347950429e-06, "loss": 2.5643, "step": 16290 }, { "epoch": 3.1, "learning_rate": 2.0710581506196376e-06, "loss": 3.4396, "step": 16291 }, { "epoch": 3.1, "learning_rate": 2.071000953288847e-06, "loss": 2.4502, "step": 16292 }, { "epoch": 3.1, "learning_rate": 2.070943755958055e-06, "loss": 2.2812, "step": 16293 }, { "epoch": 3.1, "learning_rate": 2.070886558627264e-06, "loss": 2.9103, "step": 16294 }, { "epoch": 3.1, "learning_rate": 2.0708293612964728e-06, "loss": 2.6772, "step": 16295 }, { "epoch": 3.1, "learning_rate": 2.0707721639656816e-06, "loss": 2.6849, "step": 16296 }, { "epoch": 3.1, "learning_rate": 2.0707149666348904e-06, "loss": 1.8528, "step": 16297 }, { "epoch": 3.1, "learning_rate": 2.070657769304099e-06, "loss": 2.1044, "step": 16298 }, { "epoch": 3.1, "learning_rate": 2.070600571973308e-06, "loss": 3.0445, "step": 16299 }, { "epoch": 3.1, "learning_rate": 2.0705433746425168e-06, "loss": 2.0071, "step": 16300 }, { "epoch": 3.1, "learning_rate": 2.0704861773117256e-06, "loss": 3.0991, "step": 16301 }, { "epoch": 3.11, "learning_rate": 2.0704289799809343e-06, "loss": 2.8643, "step": 16302 }, { "epoch": 3.11, "learning_rate": 2.070371782650143e-06, "loss": 1.6227, "step": 16303 }, { "epoch": 3.11, "learning_rate": 2.070314585319352e-06, "loss": 2.3658, "step": 16304 }, { "epoch": 3.11, "learning_rate": 2.0702573879885603e-06, "loss": 3.201, "step": 16305 }, { "epoch": 3.11, "learning_rate": 2.0702001906577695e-06, "loss": 2.5847, "step": 16306 }, { "epoch": 3.11, "learning_rate": 2.070142993326978e-06, "loss": 2.9232, "step": 16307 }, { "epoch": 3.11, "learning_rate": 2.0700857959961867e-06, "loss": 2.1236, "step": 16308 }, { "epoch": 3.11, "learning_rate": 2.070028598665396e-06, "loss": 2.1005, "step": 16309 }, { "epoch": 3.11, "learning_rate": 2.0699714013346043e-06, "loss": 2.2552, "step": 16310 }, { "epoch": 3.11, "learning_rate": 2.0699142040038135e-06, "loss": 2.1455, "step": 16311 }, { "epoch": 3.11, "learning_rate": 2.069857006673022e-06, "loss": 2.0134, "step": 16312 }, { "epoch": 3.11, "learning_rate": 2.0697998093422307e-06, "loss": 3.0954, "step": 16313 }, { "epoch": 3.11, "learning_rate": 2.06974261201144e-06, "loss": 3.5659, "step": 16314 }, { "epoch": 3.11, "learning_rate": 2.0696854146806482e-06, "loss": 2.5905, "step": 16315 }, { "epoch": 3.11, "learning_rate": 2.069628217349857e-06, "loss": 1.9221, "step": 16316 }, { "epoch": 3.11, "learning_rate": 2.069571020019066e-06, "loss": 2.8402, "step": 16317 }, { "epoch": 3.11, "learning_rate": 2.0695138226882746e-06, "loss": 2.2007, "step": 16318 }, { "epoch": 3.11, "learning_rate": 2.0694566253574834e-06, "loss": 2.4616, "step": 16319 }, { "epoch": 3.11, "learning_rate": 2.069399428026692e-06, "loss": 3.5782, "step": 16320 }, { "epoch": 3.11, "learning_rate": 2.069342230695901e-06, "loss": 2.7107, "step": 16321 }, { "epoch": 3.11, "learning_rate": 2.0692850333651094e-06, "loss": 1.5768, "step": 16322 }, { "epoch": 3.11, "learning_rate": 2.0692278360343186e-06, "loss": 2.7579, "step": 16323 }, { "epoch": 3.11, "learning_rate": 2.0691706387035274e-06, "loss": 2.8333, "step": 16324 }, { "epoch": 3.11, "learning_rate": 2.0691134413727358e-06, "loss": 3.1661, "step": 16325 }, { "epoch": 3.11, "learning_rate": 2.069056244041945e-06, "loss": 2.7881, "step": 16326 }, { "epoch": 3.11, "learning_rate": 2.0689990467111533e-06, "loss": 2.7725, "step": 16327 }, { "epoch": 3.11, "learning_rate": 2.0689418493803626e-06, "loss": 2.4879, "step": 16328 }, { "epoch": 3.11, "learning_rate": 2.068884652049571e-06, "loss": 3.2764, "step": 16329 }, { "epoch": 3.11, "learning_rate": 2.0688274547187797e-06, "loss": 3.4393, "step": 16330 }, { "epoch": 3.11, "learning_rate": 2.068770257387989e-06, "loss": 2.3955, "step": 16331 }, { "epoch": 3.11, "learning_rate": 2.0687130600571973e-06, "loss": 3.0176, "step": 16332 }, { "epoch": 3.11, "learning_rate": 2.068655862726406e-06, "loss": 2.1272, "step": 16333 }, { "epoch": 3.11, "learning_rate": 2.068598665395615e-06, "loss": 2.2838, "step": 16334 }, { "epoch": 3.11, "learning_rate": 2.0685414680648237e-06, "loss": 1.7958, "step": 16335 }, { "epoch": 3.11, "learning_rate": 2.0684842707340325e-06, "loss": 1.936, "step": 16336 }, { "epoch": 3.11, "learning_rate": 2.0684270734032413e-06, "loss": 3.321, "step": 16337 }, { "epoch": 3.11, "learning_rate": 2.06836987607245e-06, "loss": 2.988, "step": 16338 }, { "epoch": 3.11, "learning_rate": 2.0683126787416584e-06, "loss": 1.5977, "step": 16339 }, { "epoch": 3.11, "learning_rate": 2.0682554814108677e-06, "loss": 2.7607, "step": 16340 }, { "epoch": 3.11, "learning_rate": 2.0681982840800765e-06, "loss": 2.0585, "step": 16341 }, { "epoch": 3.11, "learning_rate": 2.068141086749285e-06, "loss": 2.8914, "step": 16342 }, { "epoch": 3.11, "learning_rate": 2.068083889418494e-06, "loss": 1.6301, "step": 16343 }, { "epoch": 3.11, "learning_rate": 2.0680266920877024e-06, "loss": 1.793, "step": 16344 }, { "epoch": 3.11, "learning_rate": 2.0679694947569116e-06, "loss": 2.3282, "step": 16345 }, { "epoch": 3.11, "learning_rate": 2.06791229742612e-06, "loss": 2.736, "step": 16346 }, { "epoch": 3.11, "learning_rate": 2.067855100095329e-06, "loss": 2.1363, "step": 16347 }, { "epoch": 3.11, "learning_rate": 2.067797902764538e-06, "loss": 1.7364, "step": 16348 }, { "epoch": 3.11, "learning_rate": 2.0677407054337464e-06, "loss": 3.3973, "step": 16349 }, { "epoch": 3.11, "learning_rate": 2.067683508102955e-06, "loss": 3.0504, "step": 16350 }, { "epoch": 3.11, "learning_rate": 2.067626310772164e-06, "loss": 2.8101, "step": 16351 }, { "epoch": 3.11, "learning_rate": 2.0675691134413728e-06, "loss": 1.4712, "step": 16352 }, { "epoch": 3.11, "learning_rate": 2.0675119161105816e-06, "loss": 3.2667, "step": 16353 }, { "epoch": 3.12, "learning_rate": 2.0674547187797904e-06, "loss": 2.7503, "step": 16354 }, { "epoch": 3.12, "learning_rate": 2.067397521448999e-06, "loss": 2.9552, "step": 16355 }, { "epoch": 3.12, "learning_rate": 2.0673403241182075e-06, "loss": 1.8847, "step": 16356 }, { "epoch": 3.12, "learning_rate": 2.0672831267874167e-06, "loss": 2.9914, "step": 16357 }, { "epoch": 3.12, "learning_rate": 2.0672259294566255e-06, "loss": 2.6012, "step": 16358 }, { "epoch": 3.12, "learning_rate": 2.067168732125834e-06, "loss": 2.3079, "step": 16359 }, { "epoch": 3.12, "learning_rate": 2.067111534795043e-06, "loss": 1.46, "step": 16360 }, { "epoch": 3.12, "learning_rate": 2.0670543374642515e-06, "loss": 2.504, "step": 16361 }, { "epoch": 3.12, "learning_rate": 2.0669971401334607e-06, "loss": 2.414, "step": 16362 }, { "epoch": 3.12, "learning_rate": 2.0669399428026695e-06, "loss": 3.2367, "step": 16363 }, { "epoch": 3.12, "learning_rate": 2.066882745471878e-06, "loss": 2.5582, "step": 16364 }, { "epoch": 3.12, "learning_rate": 2.066825548141087e-06, "loss": 3.2227, "step": 16365 }, { "epoch": 3.12, "learning_rate": 2.0667683508102955e-06, "loss": 2.361, "step": 16366 }, { "epoch": 3.12, "learning_rate": 2.0667111534795042e-06, "loss": 2.1274, "step": 16367 }, { "epoch": 3.12, "learning_rate": 2.066653956148713e-06, "loss": 2.1693, "step": 16368 }, { "epoch": 3.12, "learning_rate": 2.066596758817922e-06, "loss": 2.4188, "step": 16369 }, { "epoch": 3.12, "learning_rate": 2.0665395614871306e-06, "loss": 2.2379, "step": 16370 }, { "epoch": 3.12, "learning_rate": 2.0664823641563394e-06, "loss": 2.0459, "step": 16371 }, { "epoch": 3.12, "learning_rate": 2.0664251668255482e-06, "loss": 2.3712, "step": 16372 }, { "epoch": 3.12, "learning_rate": 2.066367969494757e-06, "loss": 2.8729, "step": 16373 }, { "epoch": 3.12, "learning_rate": 2.066310772163966e-06, "loss": 1.438, "step": 16374 }, { "epoch": 3.12, "learning_rate": 2.0662535748331746e-06, "loss": 3.0756, "step": 16375 }, { "epoch": 3.12, "learning_rate": 2.0661963775023834e-06, "loss": 1.9378, "step": 16376 }, { "epoch": 3.12, "learning_rate": 2.066139180171592e-06, "loss": 3.0411, "step": 16377 }, { "epoch": 3.12, "learning_rate": 2.0660819828408006e-06, "loss": 3.625, "step": 16378 }, { "epoch": 3.12, "learning_rate": 2.0660247855100098e-06, "loss": 3.32, "step": 16379 }, { "epoch": 3.12, "learning_rate": 2.0659675881792186e-06, "loss": 2.8517, "step": 16380 }, { "epoch": 3.12, "learning_rate": 2.065910390848427e-06, "loss": 2.7143, "step": 16381 }, { "epoch": 3.12, "learning_rate": 2.065853193517636e-06, "loss": 2.1635, "step": 16382 }, { "epoch": 3.12, "learning_rate": 2.0657959961868445e-06, "loss": 3.154, "step": 16383 }, { "epoch": 3.12, "learning_rate": 2.0657387988560533e-06, "loss": 1.8457, "step": 16384 }, { "epoch": 3.12, "learning_rate": 2.065681601525262e-06, "loss": 2.3427, "step": 16385 }, { "epoch": 3.12, "learning_rate": 2.065624404194471e-06, "loss": 3.3, "step": 16386 }, { "epoch": 3.12, "learning_rate": 2.0655672068636797e-06, "loss": 1.5335, "step": 16387 }, { "epoch": 3.12, "learning_rate": 2.0655100095328885e-06, "loss": 2.2529, "step": 16388 }, { "epoch": 3.12, "learning_rate": 2.0654528122020973e-06, "loss": 2.9643, "step": 16389 }, { "epoch": 3.12, "learning_rate": 2.065395614871306e-06, "loss": 2.3019, "step": 16390 }, { "epoch": 3.12, "learning_rate": 2.065338417540515e-06, "loss": 2.0252, "step": 16391 }, { "epoch": 3.12, "learning_rate": 2.0652812202097237e-06, "loss": 3.0499, "step": 16392 }, { "epoch": 3.12, "learning_rate": 2.0652240228789325e-06, "loss": 1.7442, "step": 16393 }, { "epoch": 3.12, "learning_rate": 2.0651668255481413e-06, "loss": 2.195, "step": 16394 }, { "epoch": 3.12, "learning_rate": 2.0651096282173496e-06, "loss": 3.0941, "step": 16395 }, { "epoch": 3.12, "learning_rate": 2.065052430886559e-06, "loss": 2.522, "step": 16396 }, { "epoch": 3.12, "learning_rate": 2.0649952335557676e-06, "loss": 2.154, "step": 16397 }, { "epoch": 3.12, "learning_rate": 2.064938036224976e-06, "loss": 1.8839, "step": 16398 }, { "epoch": 3.12, "learning_rate": 2.0648808388941852e-06, "loss": 2.9792, "step": 16399 }, { "epoch": 3.12, "learning_rate": 2.0648236415633936e-06, "loss": 2.1343, "step": 16400 }, { "epoch": 3.12, "learning_rate": 2.0647664442326024e-06, "loss": 2.4435, "step": 16401 }, { "epoch": 3.12, "learning_rate": 2.0647092469018116e-06, "loss": 3.3091, "step": 16402 }, { "epoch": 3.12, "learning_rate": 2.06465204957102e-06, "loss": 3.1717, "step": 16403 }, { "epoch": 3.12, "learning_rate": 2.0645948522402288e-06, "loss": 2.259, "step": 16404 }, { "epoch": 3.12, "learning_rate": 2.0645376549094376e-06, "loss": 2.2497, "step": 16405 }, { "epoch": 3.12, "learning_rate": 2.0644804575786464e-06, "loss": 1.8995, "step": 16406 }, { "epoch": 3.13, "learning_rate": 2.064423260247855e-06, "loss": 1.9097, "step": 16407 }, { "epoch": 3.13, "learning_rate": 2.064366062917064e-06, "loss": 2.401, "step": 16408 }, { "epoch": 3.13, "learning_rate": 2.0643088655862727e-06, "loss": 2.3463, "step": 16409 }, { "epoch": 3.13, "learning_rate": 2.0642516682554815e-06, "loss": 2.788, "step": 16410 }, { "epoch": 3.13, "learning_rate": 2.0641944709246903e-06, "loss": 1.9149, "step": 16411 }, { "epoch": 3.13, "learning_rate": 2.064137273593899e-06, "loss": 1.6275, "step": 16412 }, { "epoch": 3.13, "learning_rate": 2.064080076263108e-06, "loss": 2.7891, "step": 16413 }, { "epoch": 3.13, "learning_rate": 2.0640228789323167e-06, "loss": 1.7012, "step": 16414 }, { "epoch": 3.13, "learning_rate": 2.063965681601525e-06, "loss": 1.7208, "step": 16415 }, { "epoch": 3.13, "learning_rate": 2.0639084842707343e-06, "loss": 2.0444, "step": 16416 }, { "epoch": 3.13, "learning_rate": 2.0638512869399427e-06, "loss": 2.3831, "step": 16417 }, { "epoch": 3.13, "learning_rate": 2.0637940896091515e-06, "loss": 1.8834, "step": 16418 }, { "epoch": 3.13, "learning_rate": 2.0637368922783607e-06, "loss": 2.3998, "step": 16419 }, { "epoch": 3.13, "learning_rate": 2.063679694947569e-06, "loss": 2.0958, "step": 16420 }, { "epoch": 3.13, "learning_rate": 2.063622497616778e-06, "loss": 3.0612, "step": 16421 }, { "epoch": 3.13, "learning_rate": 2.0635653002859866e-06, "loss": 3.0863, "step": 16422 }, { "epoch": 3.13, "learning_rate": 2.0635081029551954e-06, "loss": 2.4607, "step": 16423 }, { "epoch": 3.13, "learning_rate": 2.0634509056244042e-06, "loss": 3.0914, "step": 16424 }, { "epoch": 3.13, "learning_rate": 2.063393708293613e-06, "loss": 3.1322, "step": 16425 }, { "epoch": 3.13, "learning_rate": 2.063336510962822e-06, "loss": 1.6531, "step": 16426 }, { "epoch": 3.13, "learning_rate": 2.0632793136320306e-06, "loss": 2.2001, "step": 16427 }, { "epoch": 3.13, "learning_rate": 2.0632221163012394e-06, "loss": 3.0695, "step": 16428 }, { "epoch": 3.13, "learning_rate": 2.063164918970448e-06, "loss": 1.6931, "step": 16429 }, { "epoch": 3.13, "learning_rate": 2.063107721639657e-06, "loss": 2.9342, "step": 16430 }, { "epoch": 3.13, "learning_rate": 2.0630505243088658e-06, "loss": 3.5148, "step": 16431 }, { "epoch": 3.13, "learning_rate": 2.062993326978074e-06, "loss": 3.294, "step": 16432 }, { "epoch": 3.13, "learning_rate": 2.0629361296472834e-06, "loss": 2.857, "step": 16433 }, { "epoch": 3.13, "learning_rate": 2.0628789323164917e-06, "loss": 2.7561, "step": 16434 }, { "epoch": 3.13, "learning_rate": 2.0628217349857005e-06, "loss": 3.2026, "step": 16435 }, { "epoch": 3.13, "learning_rate": 2.0627645376549098e-06, "loss": 3.2208, "step": 16436 }, { "epoch": 3.13, "learning_rate": 2.062707340324118e-06, "loss": 1.443, "step": 16437 }, { "epoch": 3.13, "learning_rate": 2.062650142993327e-06, "loss": 2.4045, "step": 16438 }, { "epoch": 3.13, "learning_rate": 2.0625929456625357e-06, "loss": 3.7987, "step": 16439 }, { "epoch": 3.13, "learning_rate": 2.0625357483317445e-06, "loss": 2.468, "step": 16440 }, { "epoch": 3.13, "learning_rate": 2.0624785510009533e-06, "loss": 2.8293, "step": 16441 }, { "epoch": 3.13, "learning_rate": 2.062421353670162e-06, "loss": 2.1511, "step": 16442 }, { "epoch": 3.13, "learning_rate": 2.062364156339371e-06, "loss": 2.721, "step": 16443 }, { "epoch": 3.13, "learning_rate": 2.0623069590085797e-06, "loss": 2.3381, "step": 16444 }, { "epoch": 3.13, "learning_rate": 2.0622497616777885e-06, "loss": 2.6836, "step": 16445 }, { "epoch": 3.13, "learning_rate": 2.0621925643469973e-06, "loss": 2.2857, "step": 16446 }, { "epoch": 3.13, "learning_rate": 2.062135367016206e-06, "loss": 2.4145, "step": 16447 }, { "epoch": 3.13, "learning_rate": 2.062078169685415e-06, "loss": 3.1252, "step": 16448 }, { "epoch": 3.13, "learning_rate": 2.0620209723546232e-06, "loss": 2.4288, "step": 16449 }, { "epoch": 3.13, "learning_rate": 2.0619637750238324e-06, "loss": 2.0862, "step": 16450 }, { "epoch": 3.13, "learning_rate": 2.0619065776930412e-06, "loss": 2.6945, "step": 16451 }, { "epoch": 3.13, "learning_rate": 2.0618493803622496e-06, "loss": 2.3521, "step": 16452 }, { "epoch": 3.13, "learning_rate": 2.061792183031459e-06, "loss": 2.2749, "step": 16453 }, { "epoch": 3.13, "learning_rate": 2.061734985700667e-06, "loss": 1.2867, "step": 16454 }, { "epoch": 3.13, "learning_rate": 2.061677788369876e-06, "loss": 3.525, "step": 16455 }, { "epoch": 3.13, "learning_rate": 2.0616205910390848e-06, "loss": 2.9481, "step": 16456 }, { "epoch": 3.13, "learning_rate": 2.0615633937082936e-06, "loss": 2.4343, "step": 16457 }, { "epoch": 3.13, "learning_rate": 2.061506196377503e-06, "loss": 3.4258, "step": 16458 }, { "epoch": 3.14, "learning_rate": 2.061448999046711e-06, "loss": 2.5198, "step": 16459 }, { "epoch": 3.14, "learning_rate": 2.06139180171592e-06, "loss": 2.3482, "step": 16460 }, { "epoch": 3.14, "learning_rate": 2.0613346043851287e-06, "loss": 2.6372, "step": 16461 }, { "epoch": 3.14, "learning_rate": 2.0612774070543375e-06, "loss": 2.933, "step": 16462 }, { "epoch": 3.14, "learning_rate": 2.0612202097235463e-06, "loss": 2.9714, "step": 16463 }, { "epoch": 3.14, "learning_rate": 2.061163012392755e-06, "loss": 2.4928, "step": 16464 }, { "epoch": 3.14, "learning_rate": 2.061105815061964e-06, "loss": 2.0351, "step": 16465 }, { "epoch": 3.14, "learning_rate": 2.0610486177311723e-06, "loss": 2.5996, "step": 16466 }, { "epoch": 3.14, "learning_rate": 2.0609914204003815e-06, "loss": 2.1207, "step": 16467 }, { "epoch": 3.14, "learning_rate": 2.0609342230695903e-06, "loss": 2.9169, "step": 16468 }, { "epoch": 3.14, "learning_rate": 2.0608770257387987e-06, "loss": 2.2329, "step": 16469 }, { "epoch": 3.14, "learning_rate": 2.060819828408008e-06, "loss": 2.7107, "step": 16470 }, { "epoch": 3.14, "learning_rate": 2.0607626310772163e-06, "loss": 1.9966, "step": 16471 }, { "epoch": 3.14, "learning_rate": 2.060705433746425e-06, "loss": 2.0323, "step": 16472 }, { "epoch": 3.14, "learning_rate": 2.060648236415634e-06, "loss": 2.8081, "step": 16473 }, { "epoch": 3.14, "learning_rate": 2.0605910390848426e-06, "loss": 2.6262, "step": 16474 }, { "epoch": 3.14, "learning_rate": 2.060533841754052e-06, "loss": 3.8083, "step": 16475 }, { "epoch": 3.14, "learning_rate": 2.0604766444232602e-06, "loss": 1.9638, "step": 16476 }, { "epoch": 3.14, "learning_rate": 2.060419447092469e-06, "loss": 2.5996, "step": 16477 }, { "epoch": 3.14, "learning_rate": 2.060362249761678e-06, "loss": 2.7957, "step": 16478 }, { "epoch": 3.14, "learning_rate": 2.0603050524308866e-06, "loss": 2.9782, "step": 16479 }, { "epoch": 3.14, "learning_rate": 2.0602478551000954e-06, "loss": 2.438, "step": 16480 }, { "epoch": 3.14, "learning_rate": 2.060190657769304e-06, "loss": 3.4221, "step": 16481 }, { "epoch": 3.14, "learning_rate": 2.060133460438513e-06, "loss": 2.4014, "step": 16482 }, { "epoch": 3.14, "learning_rate": 2.0600762631077214e-06, "loss": 2.4254, "step": 16483 }, { "epoch": 3.14, "learning_rate": 2.0600190657769306e-06, "loss": 1.6255, "step": 16484 }, { "epoch": 3.14, "learning_rate": 2.0599618684461394e-06, "loss": 3.2515, "step": 16485 }, { "epoch": 3.14, "learning_rate": 2.0599046711153477e-06, "loss": 2.7324, "step": 16486 }, { "epoch": 3.14, "learning_rate": 2.059847473784557e-06, "loss": 2.2162, "step": 16487 }, { "epoch": 3.14, "learning_rate": 2.0597902764537653e-06, "loss": 2.1504, "step": 16488 }, { "epoch": 3.14, "learning_rate": 2.059733079122974e-06, "loss": 1.5824, "step": 16489 }, { "epoch": 3.14, "learning_rate": 2.059675881792183e-06, "loss": 2.981, "step": 16490 }, { "epoch": 3.14, "learning_rate": 2.0596186844613917e-06, "loss": 3.26, "step": 16491 }, { "epoch": 3.14, "learning_rate": 2.059561487130601e-06, "loss": 2.6898, "step": 16492 }, { "epoch": 3.14, "learning_rate": 2.0595042897998093e-06, "loss": 3.3795, "step": 16493 }, { "epoch": 3.14, "learning_rate": 2.059447092469018e-06, "loss": 1.9018, "step": 16494 }, { "epoch": 3.14, "learning_rate": 2.059389895138227e-06, "loss": 2.9663, "step": 16495 }, { "epoch": 3.14, "learning_rate": 2.0593326978074357e-06, "loss": 2.484, "step": 16496 }, { "epoch": 3.14, "learning_rate": 2.0592755004766445e-06, "loss": 2.1001, "step": 16497 }, { "epoch": 3.14, "learning_rate": 2.0592183031458533e-06, "loss": 2.578, "step": 16498 }, { "epoch": 3.14, "learning_rate": 2.059161105815062e-06, "loss": 1.9537, "step": 16499 }, { "epoch": 3.14, "learning_rate": 2.059103908484271e-06, "loss": 3.0333, "step": 16500 }, { "epoch": 3.14, "learning_rate": 2.0590467111534797e-06, "loss": 1.5717, "step": 16501 }, { "epoch": 3.14, "learning_rate": 2.0589895138226884e-06, "loss": 3.1107, "step": 16502 }, { "epoch": 3.14, "learning_rate": 2.058932316491897e-06, "loss": 2.0112, "step": 16503 }, { "epoch": 3.14, "learning_rate": 2.058875119161106e-06, "loss": 2.9045, "step": 16504 }, { "epoch": 3.14, "learning_rate": 2.0588179218303144e-06, "loss": 2.3225, "step": 16505 }, { "epoch": 3.14, "learning_rate": 2.0587607244995236e-06, "loss": 2.2155, "step": 16506 }, { "epoch": 3.14, "learning_rate": 2.0587035271687324e-06, "loss": 2.1858, "step": 16507 }, { "epoch": 3.14, "learning_rate": 2.058646329837941e-06, "loss": 2.6797, "step": 16508 }, { "epoch": 3.14, "learning_rate": 2.05858913250715e-06, "loss": 2.3537, "step": 16509 }, { "epoch": 3.14, "learning_rate": 2.0585319351763584e-06, "loss": 2.2184, "step": 16510 }, { "epoch": 3.14, "learning_rate": 2.058474737845567e-06, "loss": 2.5241, "step": 16511 }, { "epoch": 3.15, "learning_rate": 2.058417540514776e-06, "loss": 2.4802, "step": 16512 }, { "epoch": 3.15, "learning_rate": 2.0583603431839848e-06, "loss": 2.9582, "step": 16513 }, { "epoch": 3.15, "learning_rate": 2.0583031458531936e-06, "loss": 2.4918, "step": 16514 }, { "epoch": 3.15, "learning_rate": 2.0582459485224023e-06, "loss": 2.7399, "step": 16515 }, { "epoch": 3.15, "learning_rate": 2.058188751191611e-06, "loss": 2.4774, "step": 16516 }, { "epoch": 3.15, "learning_rate": 2.05813155386082e-06, "loss": 3.2185, "step": 16517 }, { "epoch": 3.15, "learning_rate": 2.0580743565300287e-06, "loss": 2.4065, "step": 16518 }, { "epoch": 3.15, "learning_rate": 2.0580171591992375e-06, "loss": 2.8664, "step": 16519 }, { "epoch": 3.15, "learning_rate": 2.057959961868446e-06, "loss": 2.0708, "step": 16520 }, { "epoch": 3.15, "learning_rate": 2.057902764537655e-06, "loss": 2.8094, "step": 16521 }, { "epoch": 3.15, "learning_rate": 2.0578455672068635e-06, "loss": 2.6752, "step": 16522 }, { "epoch": 3.15, "learning_rate": 2.0577883698760727e-06, "loss": 1.4, "step": 16523 }, { "epoch": 3.15, "learning_rate": 2.0577311725452815e-06, "loss": 3.139, "step": 16524 }, { "epoch": 3.15, "learning_rate": 2.05767397521449e-06, "loss": 2.8338, "step": 16525 }, { "epoch": 3.15, "learning_rate": 2.057616777883699e-06, "loss": 2.054, "step": 16526 }, { "epoch": 3.15, "learning_rate": 2.0575595805529074e-06, "loss": 2.7903, "step": 16527 }, { "epoch": 3.15, "learning_rate": 2.0575023832221162e-06, "loss": 2.6948, "step": 16528 }, { "epoch": 3.15, "learning_rate": 2.057445185891325e-06, "loss": 2.0061, "step": 16529 }, { "epoch": 3.15, "learning_rate": 2.057387988560534e-06, "loss": 2.9201, "step": 16530 }, { "epoch": 3.15, "learning_rate": 2.0573307912297426e-06, "loss": 2.0171, "step": 16531 }, { "epoch": 3.15, "learning_rate": 2.0572735938989514e-06, "loss": 3.3129, "step": 16532 }, { "epoch": 3.15, "learning_rate": 2.05721639656816e-06, "loss": 2.2987, "step": 16533 }, { "epoch": 3.15, "learning_rate": 2.057159199237369e-06, "loss": 2.4996, "step": 16534 }, { "epoch": 3.15, "learning_rate": 2.057102001906578e-06, "loss": 2.7306, "step": 16535 }, { "epoch": 3.15, "learning_rate": 2.0570448045757866e-06, "loss": 2.4511, "step": 16536 }, { "epoch": 3.15, "learning_rate": 2.056987607244995e-06, "loss": 1.7213, "step": 16537 }, { "epoch": 3.15, "learning_rate": 2.056930409914204e-06, "loss": 2.457, "step": 16538 }, { "epoch": 3.15, "learning_rate": 2.056873212583413e-06, "loss": 2.8569, "step": 16539 }, { "epoch": 3.15, "learning_rate": 2.0568160152526218e-06, "loss": 2.4718, "step": 16540 }, { "epoch": 3.15, "learning_rate": 2.0567588179218306e-06, "loss": 2.3891, "step": 16541 }, { "epoch": 3.15, "learning_rate": 2.056701620591039e-06, "loss": 2.2878, "step": 16542 }, { "epoch": 3.15, "learning_rate": 2.056644423260248e-06, "loss": 2.7824, "step": 16543 }, { "epoch": 3.15, "learning_rate": 2.0565872259294565e-06, "loss": 1.8708, "step": 16544 }, { "epoch": 3.15, "learning_rate": 2.0565300285986653e-06, "loss": 2.7231, "step": 16545 }, { "epoch": 3.15, "learning_rate": 2.0564728312678745e-06, "loss": 3.8796, "step": 16546 }, { "epoch": 3.15, "learning_rate": 2.056415633937083e-06, "loss": 2.4583, "step": 16547 }, { "epoch": 3.15, "learning_rate": 2.0563584366062917e-06, "loss": 1.6828, "step": 16548 }, { "epoch": 3.15, "learning_rate": 2.0563012392755005e-06, "loss": 2.7119, "step": 16549 }, { "epoch": 3.15, "learning_rate": 2.0562440419447093e-06, "loss": 1.6844, "step": 16550 }, { "epoch": 3.15, "learning_rate": 2.056186844613918e-06, "loss": 4.0409, "step": 16551 }, { "epoch": 3.15, "learning_rate": 2.056129647283127e-06, "loss": 1.8862, "step": 16552 }, { "epoch": 3.15, "learning_rate": 2.0560724499523357e-06, "loss": 3.3538, "step": 16553 }, { "epoch": 3.15, "learning_rate": 2.056015252621544e-06, "loss": 2.7253, "step": 16554 }, { "epoch": 3.15, "learning_rate": 2.0559580552907533e-06, "loss": 3.1397, "step": 16555 }, { "epoch": 3.15, "learning_rate": 2.055900857959962e-06, "loss": 3.4021, "step": 16556 }, { "epoch": 3.15, "learning_rate": 2.055843660629171e-06, "loss": 1.9445, "step": 16557 }, { "epoch": 3.15, "learning_rate": 2.0557864632983796e-06, "loss": 1.9811, "step": 16558 }, { "epoch": 3.15, "learning_rate": 2.055729265967588e-06, "loss": 2.4027, "step": 16559 }, { "epoch": 3.15, "learning_rate": 2.0556720686367972e-06, "loss": 3.409, "step": 16560 }, { "epoch": 3.15, "learning_rate": 2.0556148713060056e-06, "loss": 2.5251, "step": 16561 }, { "epoch": 3.15, "learning_rate": 2.0555576739752144e-06, "loss": 1.6493, "step": 16562 }, { "epoch": 3.15, "learning_rate": 2.0555004766444236e-06, "loss": 2.1368, "step": 16563 }, { "epoch": 3.16, "learning_rate": 2.055443279313632e-06, "loss": 2.446, "step": 16564 }, { "epoch": 3.16, "learning_rate": 2.0553860819828408e-06, "loss": 2.1562, "step": 16565 }, { "epoch": 3.16, "learning_rate": 2.0553288846520496e-06, "loss": 2.1983, "step": 16566 }, { "epoch": 3.16, "learning_rate": 2.0552716873212584e-06, "loss": 2.2227, "step": 16567 }, { "epoch": 3.16, "learning_rate": 2.055214489990467e-06, "loss": 2.0333, "step": 16568 }, { "epoch": 3.16, "learning_rate": 2.055157292659676e-06, "loss": 2.3563, "step": 16569 }, { "epoch": 3.16, "learning_rate": 2.0551000953288847e-06, "loss": 2.807, "step": 16570 }, { "epoch": 3.16, "learning_rate": 2.0550428979980935e-06, "loss": 3.6582, "step": 16571 }, { "epoch": 3.16, "learning_rate": 2.0549857006673023e-06, "loss": 1.9201, "step": 16572 }, { "epoch": 3.16, "learning_rate": 2.054928503336511e-06, "loss": 2.6365, "step": 16573 }, { "epoch": 3.16, "learning_rate": 2.05487130600572e-06, "loss": 3.5704, "step": 16574 }, { "epoch": 3.16, "learning_rate": 2.0548141086749287e-06, "loss": 1.9234, "step": 16575 }, { "epoch": 3.16, "learning_rate": 2.054756911344137e-06, "loss": 2.3498, "step": 16576 }, { "epoch": 3.16, "learning_rate": 2.0546997140133463e-06, "loss": 1.874, "step": 16577 }, { "epoch": 3.16, "learning_rate": 2.0546425166825547e-06, "loss": 2.2681, "step": 16578 }, { "epoch": 3.16, "learning_rate": 2.0545853193517635e-06, "loss": 1.6706, "step": 16579 }, { "epoch": 3.16, "learning_rate": 2.0545281220209727e-06, "loss": 2.664, "step": 16580 }, { "epoch": 3.16, "learning_rate": 2.054470924690181e-06, "loss": 2.1696, "step": 16581 }, { "epoch": 3.16, "learning_rate": 2.05441372735939e-06, "loss": 0.8712, "step": 16582 }, { "epoch": 3.16, "learning_rate": 2.0543565300285986e-06, "loss": 2.2252, "step": 16583 }, { "epoch": 3.16, "learning_rate": 2.0542993326978074e-06, "loss": 2.6586, "step": 16584 }, { "epoch": 3.16, "learning_rate": 2.0542421353670162e-06, "loss": 2.7419, "step": 16585 }, { "epoch": 3.16, "learning_rate": 2.054184938036225e-06, "loss": 1.5504, "step": 16586 }, { "epoch": 3.16, "learning_rate": 2.054127740705434e-06, "loss": 1.7293, "step": 16587 }, { "epoch": 3.16, "learning_rate": 2.0540705433746426e-06, "loss": 3.4873, "step": 16588 }, { "epoch": 3.16, "learning_rate": 2.0540133460438514e-06, "loss": 3.7482, "step": 16589 }, { "epoch": 3.16, "learning_rate": 2.05395614871306e-06, "loss": 1.7653, "step": 16590 }, { "epoch": 3.16, "learning_rate": 2.053898951382269e-06, "loss": 1.3575, "step": 16591 }, { "epoch": 3.16, "learning_rate": 2.0538417540514778e-06, "loss": 3.6063, "step": 16592 }, { "epoch": 3.16, "learning_rate": 2.053784556720686e-06, "loss": 1.3011, "step": 16593 }, { "epoch": 3.16, "learning_rate": 2.0537273593898954e-06, "loss": 2.8889, "step": 16594 }, { "epoch": 3.16, "learning_rate": 2.053670162059104e-06, "loss": 2.9088, "step": 16595 }, { "epoch": 3.16, "learning_rate": 2.0536129647283125e-06, "loss": 2.4043, "step": 16596 }, { "epoch": 3.16, "learning_rate": 2.0535557673975217e-06, "loss": 2.1678, "step": 16597 }, { "epoch": 3.16, "learning_rate": 2.05349857006673e-06, "loss": 1.6289, "step": 16598 }, { "epoch": 3.16, "learning_rate": 2.053441372735939e-06, "loss": 2.3327, "step": 16599 }, { "epoch": 3.16, "learning_rate": 2.0533841754051477e-06, "loss": 3.061, "step": 16600 }, { "epoch": 3.16, "learning_rate": 2.0533269780743565e-06, "loss": 2.5559, "step": 16601 }, { "epoch": 3.16, "learning_rate": 2.0532697807435653e-06, "loss": 1.6074, "step": 16602 }, { "epoch": 3.16, "learning_rate": 2.053212583412774e-06, "loss": 1.682, "step": 16603 }, { "epoch": 3.16, "learning_rate": 2.053155386081983e-06, "loss": 2.4336, "step": 16604 }, { "epoch": 3.16, "learning_rate": 2.0530981887511917e-06, "loss": 2.4352, "step": 16605 }, { "epoch": 3.16, "learning_rate": 2.0530409914204005e-06, "loss": 1.9657, "step": 16606 }, { "epoch": 3.16, "learning_rate": 2.0529837940896093e-06, "loss": 2.698, "step": 16607 }, { "epoch": 3.16, "learning_rate": 2.052926596758818e-06, "loss": 2.2703, "step": 16608 }, { "epoch": 3.16, "learning_rate": 2.052869399428027e-06, "loss": 3.237, "step": 16609 }, { "epoch": 3.16, "learning_rate": 2.0528122020972352e-06, "loss": 2.6572, "step": 16610 }, { "epoch": 3.16, "learning_rate": 2.0527550047664444e-06, "loss": 2.8691, "step": 16611 }, { "epoch": 3.16, "learning_rate": 2.0526978074356532e-06, "loss": 2.024, "step": 16612 }, { "epoch": 3.16, "learning_rate": 2.0526406101048616e-06, "loss": 2.8204, "step": 16613 }, { "epoch": 3.16, "learning_rate": 2.052583412774071e-06, "loss": 3.806, "step": 16614 }, { "epoch": 3.16, "learning_rate": 2.052526215443279e-06, "loss": 2.7466, "step": 16615 }, { "epoch": 3.16, "learning_rate": 2.052469018112488e-06, "loss": 1.5119, "step": 16616 }, { "epoch": 3.17, "learning_rate": 2.0524118207816968e-06, "loss": 3.7735, "step": 16617 }, { "epoch": 3.17, "learning_rate": 2.0523546234509056e-06, "loss": 1.9899, "step": 16618 }, { "epoch": 3.17, "learning_rate": 2.0522974261201144e-06, "loss": 1.8237, "step": 16619 }, { "epoch": 3.17, "learning_rate": 2.052240228789323e-06, "loss": 2.1434, "step": 16620 }, { "epoch": 3.17, "learning_rate": 2.052183031458532e-06, "loss": 3.9701, "step": 16621 }, { "epoch": 3.17, "learning_rate": 2.0521258341277407e-06, "loss": 2.4713, "step": 16622 }, { "epoch": 3.17, "learning_rate": 2.0520686367969495e-06, "loss": 1.4283, "step": 16623 }, { "epoch": 3.17, "learning_rate": 2.0520114394661583e-06, "loss": 2.7543, "step": 16624 }, { "epoch": 3.17, "learning_rate": 2.051954242135367e-06, "loss": 1.7999, "step": 16625 }, { "epoch": 3.17, "learning_rate": 2.051897044804576e-06, "loss": 2.1525, "step": 16626 }, { "epoch": 3.17, "learning_rate": 2.0518398474737847e-06, "loss": 3.6633, "step": 16627 }, { "epoch": 3.17, "learning_rate": 2.0517826501429935e-06, "loss": 2.4452, "step": 16628 }, { "epoch": 3.17, "learning_rate": 2.0517254528122023e-06, "loss": 2.29, "step": 16629 }, { "epoch": 3.17, "learning_rate": 2.0516682554814107e-06, "loss": 2.5389, "step": 16630 }, { "epoch": 3.17, "learning_rate": 2.05161105815062e-06, "loss": 2.0797, "step": 16631 }, { "epoch": 3.17, "learning_rate": 2.0515538608198283e-06, "loss": 3.4913, "step": 16632 }, { "epoch": 3.17, "learning_rate": 2.051496663489037e-06, "loss": 1.7588, "step": 16633 }, { "epoch": 3.17, "learning_rate": 2.0514394661582463e-06, "loss": 2.7463, "step": 16634 }, { "epoch": 3.17, "learning_rate": 2.0513822688274546e-06, "loss": 2.5986, "step": 16635 }, { "epoch": 3.17, "learning_rate": 2.051325071496664e-06, "loss": 2.7659, "step": 16636 }, { "epoch": 3.17, "learning_rate": 2.0512678741658722e-06, "loss": 2.0652, "step": 16637 }, { "epoch": 3.17, "learning_rate": 2.051210676835081e-06, "loss": 2.2964, "step": 16638 }, { "epoch": 3.17, "learning_rate": 2.05115347950429e-06, "loss": 2.0368, "step": 16639 }, { "epoch": 3.17, "learning_rate": 2.0510962821734986e-06, "loss": 2.6873, "step": 16640 }, { "epoch": 3.17, "learning_rate": 2.0510390848427074e-06, "loss": 2.8064, "step": 16641 }, { "epoch": 3.17, "learning_rate": 2.050981887511916e-06, "loss": 2.9318, "step": 16642 }, { "epoch": 3.17, "learning_rate": 2.050924690181125e-06, "loss": 2.7486, "step": 16643 }, { "epoch": 3.17, "learning_rate": 2.0508674928503338e-06, "loss": 2.9821, "step": 16644 }, { "epoch": 3.17, "learning_rate": 2.0508102955195426e-06, "loss": 2.2452, "step": 16645 }, { "epoch": 3.17, "learning_rate": 2.0507530981887514e-06, "loss": 1.9165, "step": 16646 }, { "epoch": 3.17, "learning_rate": 2.0506959008579597e-06, "loss": 2.5314, "step": 16647 }, { "epoch": 3.17, "learning_rate": 2.050638703527169e-06, "loss": 2.5794, "step": 16648 }, { "epoch": 3.17, "learning_rate": 2.0505815061963773e-06, "loss": 2.8744, "step": 16649 }, { "epoch": 3.17, "learning_rate": 2.050524308865586e-06, "loss": 2.5643, "step": 16650 }, { "epoch": 3.17, "learning_rate": 2.0504671115347953e-06, "loss": 1.5994, "step": 16651 }, { "epoch": 3.17, "learning_rate": 2.0504099142040037e-06, "loss": 2.4128, "step": 16652 }, { "epoch": 3.17, "learning_rate": 2.050352716873213e-06, "loss": 2.9295, "step": 16653 }, { "epoch": 3.17, "learning_rate": 2.0502955195424213e-06, "loss": 2.4108, "step": 16654 }, { "epoch": 3.17, "learning_rate": 2.05023832221163e-06, "loss": 1.5913, "step": 16655 }, { "epoch": 3.17, "learning_rate": 2.050181124880839e-06, "loss": 2.3933, "step": 16656 }, { "epoch": 3.17, "learning_rate": 2.0501239275500477e-06, "loss": 1.9756, "step": 16657 }, { "epoch": 3.17, "learning_rate": 2.0500667302192565e-06, "loss": 2.4691, "step": 16658 }, { "epoch": 3.17, "learning_rate": 2.0500095328884653e-06, "loss": 2.7938, "step": 16659 }, { "epoch": 3.17, "learning_rate": 2.049952335557674e-06, "loss": 2.6017, "step": 16660 }, { "epoch": 3.17, "learning_rate": 2.049895138226883e-06, "loss": 1.7754, "step": 16661 }, { "epoch": 3.17, "learning_rate": 2.0498379408960916e-06, "loss": 2.6949, "step": 16662 }, { "epoch": 3.17, "learning_rate": 2.0497807435653004e-06, "loss": 2.7185, "step": 16663 }, { "epoch": 3.17, "learning_rate": 2.049723546234509e-06, "loss": 1.4455, "step": 16664 }, { "epoch": 3.17, "learning_rate": 2.049666348903718e-06, "loss": 2.4395, "step": 16665 }, { "epoch": 3.17, "learning_rate": 2.0496091515729264e-06, "loss": 1.8554, "step": 16666 }, { "epoch": 3.17, "learning_rate": 2.049551954242135e-06, "loss": 3.4498, "step": 16667 }, { "epoch": 3.17, "learning_rate": 2.0494947569113444e-06, "loss": 1.7679, "step": 16668 }, { "epoch": 3.18, "learning_rate": 2.0494375595805528e-06, "loss": 2.8122, "step": 16669 }, { "epoch": 3.18, "learning_rate": 2.049380362249762e-06, "loss": 2.378, "step": 16670 }, { "epoch": 3.18, "learning_rate": 2.0493231649189704e-06, "loss": 2.5702, "step": 16671 }, { "epoch": 3.18, "learning_rate": 2.049265967588179e-06, "loss": 2.1179, "step": 16672 }, { "epoch": 3.18, "learning_rate": 2.0492087702573884e-06, "loss": 2.0795, "step": 16673 }, { "epoch": 3.18, "learning_rate": 2.0491515729265968e-06, "loss": 2.5518, "step": 16674 }, { "epoch": 3.18, "learning_rate": 2.0490943755958055e-06, "loss": 2.5374, "step": 16675 }, { "epoch": 3.18, "learning_rate": 2.0490371782650143e-06, "loss": 2.5346, "step": 16676 }, { "epoch": 3.18, "learning_rate": 2.048979980934223e-06, "loss": 2.1374, "step": 16677 }, { "epoch": 3.18, "learning_rate": 2.048922783603432e-06, "loss": 2.868, "step": 16678 }, { "epoch": 3.18, "learning_rate": 2.0488655862726407e-06, "loss": 2.1641, "step": 16679 }, { "epoch": 3.18, "learning_rate": 2.0488083889418495e-06, "loss": 3.1948, "step": 16680 }, { "epoch": 3.18, "learning_rate": 2.048751191611058e-06, "loss": 1.9163, "step": 16681 }, { "epoch": 3.18, "learning_rate": 2.048693994280267e-06, "loss": 1.1741, "step": 16682 }, { "epoch": 3.18, "learning_rate": 2.048636796949476e-06, "loss": 2.7608, "step": 16683 }, { "epoch": 3.18, "learning_rate": 2.0485795996186843e-06, "loss": 2.7167, "step": 16684 }, { "epoch": 3.18, "learning_rate": 2.0485224022878935e-06, "loss": 2.4351, "step": 16685 }, { "epoch": 3.18, "learning_rate": 2.048465204957102e-06, "loss": 2.5648, "step": 16686 }, { "epoch": 3.18, "learning_rate": 2.048408007626311e-06, "loss": 2.0451, "step": 16687 }, { "epoch": 3.18, "learning_rate": 2.0483508102955194e-06, "loss": 2.9228, "step": 16688 }, { "epoch": 3.18, "learning_rate": 2.0482936129647282e-06, "loss": 2.8205, "step": 16689 }, { "epoch": 3.18, "learning_rate": 2.0482364156339375e-06, "loss": 2.2476, "step": 16690 }, { "epoch": 3.18, "learning_rate": 2.048179218303146e-06, "loss": 2.007, "step": 16691 }, { "epoch": 3.18, "learning_rate": 2.0481220209723546e-06, "loss": 3.2215, "step": 16692 }, { "epoch": 3.18, "learning_rate": 2.0480648236415634e-06, "loss": 3.0935, "step": 16693 }, { "epoch": 3.18, "learning_rate": 2.048007626310772e-06, "loss": 2.9072, "step": 16694 }, { "epoch": 3.18, "learning_rate": 2.047950428979981e-06, "loss": 3.3468, "step": 16695 }, { "epoch": 3.18, "learning_rate": 2.04789323164919e-06, "loss": 1.6963, "step": 16696 }, { "epoch": 3.18, "learning_rate": 2.0478360343183986e-06, "loss": 3.2357, "step": 16697 }, { "epoch": 3.18, "learning_rate": 2.047778836987607e-06, "loss": 2.1851, "step": 16698 }, { "epoch": 3.18, "learning_rate": 2.047721639656816e-06, "loss": 3.3708, "step": 16699 }, { "epoch": 3.18, "learning_rate": 2.047664442326025e-06, "loss": 2.6012, "step": 16700 }, { "epoch": 3.18, "learning_rate": 2.0476072449952338e-06, "loss": 3.7633, "step": 16701 }, { "epoch": 3.18, "learning_rate": 2.0475500476644426e-06, "loss": 2.2294, "step": 16702 }, { "epoch": 3.18, "learning_rate": 2.047492850333651e-06, "loss": 1.9449, "step": 16703 }, { "epoch": 3.18, "learning_rate": 2.04743565300286e-06, "loss": 1.6889, "step": 16704 }, { "epoch": 3.18, "learning_rate": 2.0473784556720685e-06, "loss": 2.8028, "step": 16705 }, { "epoch": 3.18, "learning_rate": 2.0473212583412773e-06, "loss": 2.624, "step": 16706 }, { "epoch": 3.18, "learning_rate": 2.0472640610104865e-06, "loss": 2.5261, "step": 16707 }, { "epoch": 3.18, "learning_rate": 2.047206863679695e-06, "loss": 1.2918, "step": 16708 }, { "epoch": 3.18, "learning_rate": 2.0471496663489037e-06, "loss": 3.0291, "step": 16709 }, { "epoch": 3.18, "learning_rate": 2.0470924690181125e-06, "loss": 1.9318, "step": 16710 }, { "epoch": 3.18, "learning_rate": 2.0470352716873213e-06, "loss": 2.4108, "step": 16711 }, { "epoch": 3.18, "learning_rate": 2.04697807435653e-06, "loss": 3.0398, "step": 16712 }, { "epoch": 3.18, "learning_rate": 2.046920877025739e-06, "loss": 2.1286, "step": 16713 }, { "epoch": 3.18, "learning_rate": 2.0468636796949477e-06, "loss": 1.7662, "step": 16714 }, { "epoch": 3.18, "learning_rate": 2.0468064823641565e-06, "loss": 3.4887, "step": 16715 }, { "epoch": 3.18, "learning_rate": 2.0467492850333652e-06, "loss": 2.6894, "step": 16716 }, { "epoch": 3.18, "learning_rate": 2.046692087702574e-06, "loss": 2.2171, "step": 16717 }, { "epoch": 3.18, "learning_rate": 2.046634890371783e-06, "loss": 1.6691, "step": 16718 }, { "epoch": 3.18, "learning_rate": 2.0465776930409916e-06, "loss": 2.0715, "step": 16719 }, { "epoch": 3.18, "learning_rate": 2.0465204957102e-06, "loss": 1.5013, "step": 16720 }, { "epoch": 3.18, "learning_rate": 2.0464632983794092e-06, "loss": 1.8834, "step": 16721 }, { "epoch": 3.19, "learning_rate": 2.046406101048618e-06, "loss": 2.5583, "step": 16722 }, { "epoch": 3.19, "learning_rate": 2.0463489037178264e-06, "loss": 1.958, "step": 16723 }, { "epoch": 3.19, "learning_rate": 2.0462917063870356e-06, "loss": 2.488, "step": 16724 }, { "epoch": 3.19, "learning_rate": 2.046234509056244e-06, "loss": 2.7005, "step": 16725 }, { "epoch": 3.19, "learning_rate": 2.0461773117254528e-06, "loss": 2.77, "step": 16726 }, { "epoch": 3.19, "learning_rate": 2.0461201143946616e-06, "loss": 1.8887, "step": 16727 }, { "epoch": 3.19, "learning_rate": 2.0460629170638703e-06, "loss": 2.4508, "step": 16728 }, { "epoch": 3.19, "learning_rate": 2.046005719733079e-06, "loss": 3.2489, "step": 16729 }, { "epoch": 3.19, "learning_rate": 2.045948522402288e-06, "loss": 2.8493, "step": 16730 }, { "epoch": 3.19, "learning_rate": 2.0458913250714967e-06, "loss": 2.9373, "step": 16731 }, { "epoch": 3.19, "learning_rate": 2.0458341277407055e-06, "loss": 1.787, "step": 16732 }, { "epoch": 3.19, "learning_rate": 2.0457769304099143e-06, "loss": 2.201, "step": 16733 }, { "epoch": 3.19, "learning_rate": 2.045719733079123e-06, "loss": 2.1684, "step": 16734 }, { "epoch": 3.19, "learning_rate": 2.045662535748332e-06, "loss": 2.6297, "step": 16735 }, { "epoch": 3.19, "learning_rate": 2.0456053384175407e-06, "loss": 2.8931, "step": 16736 }, { "epoch": 3.19, "learning_rate": 2.045548141086749e-06, "loss": 2.7967, "step": 16737 }, { "epoch": 3.19, "learning_rate": 2.0454909437559583e-06, "loss": 1.9035, "step": 16738 }, { "epoch": 3.19, "learning_rate": 2.045433746425167e-06, "loss": 1.5174, "step": 16739 }, { "epoch": 3.19, "learning_rate": 2.0453765490943754e-06, "loss": 2.149, "step": 16740 }, { "epoch": 3.19, "learning_rate": 2.0453193517635847e-06, "loss": 2.3069, "step": 16741 }, { "epoch": 3.19, "learning_rate": 2.045262154432793e-06, "loss": 2.4679, "step": 16742 }, { "epoch": 3.19, "learning_rate": 2.045204957102002e-06, "loss": 2.8084, "step": 16743 }, { "epoch": 3.19, "learning_rate": 2.0451477597712106e-06, "loss": 2.2738, "step": 16744 }, { "epoch": 3.19, "learning_rate": 2.0450905624404194e-06, "loss": 2.1697, "step": 16745 }, { "epoch": 3.19, "learning_rate": 2.045033365109628e-06, "loss": 2.9571, "step": 16746 }, { "epoch": 3.19, "learning_rate": 2.044976167778837e-06, "loss": 2.9626, "step": 16747 }, { "epoch": 3.19, "learning_rate": 2.044918970448046e-06, "loss": 1.8827, "step": 16748 }, { "epoch": 3.19, "learning_rate": 2.0448617731172546e-06, "loss": 2.3691, "step": 16749 }, { "epoch": 3.19, "learning_rate": 2.0448045757864634e-06, "loss": 3.4956, "step": 16750 }, { "epoch": 3.19, "learning_rate": 2.044747378455672e-06, "loss": 2.525, "step": 16751 }, { "epoch": 3.19, "learning_rate": 2.044690181124881e-06, "loss": 2.5085, "step": 16752 }, { "epoch": 3.19, "learning_rate": 2.0446329837940898e-06, "loss": 1.7928, "step": 16753 }, { "epoch": 3.19, "learning_rate": 2.044575786463298e-06, "loss": 2.5916, "step": 16754 }, { "epoch": 3.19, "learning_rate": 2.0445185891325074e-06, "loss": 2.7473, "step": 16755 }, { "epoch": 3.19, "learning_rate": 2.044461391801716e-06, "loss": 2.3905, "step": 16756 }, { "epoch": 3.19, "learning_rate": 2.0444041944709245e-06, "loss": 2.0711, "step": 16757 }, { "epoch": 3.19, "learning_rate": 2.0443469971401337e-06, "loss": 2.8007, "step": 16758 }, { "epoch": 3.19, "learning_rate": 2.044289799809342e-06, "loss": 3.3099, "step": 16759 }, { "epoch": 3.19, "learning_rate": 2.044232602478551e-06, "loss": 2.5234, "step": 16760 }, { "epoch": 3.19, "learning_rate": 2.04417540514776e-06, "loss": 2.5718, "step": 16761 }, { "epoch": 3.19, "learning_rate": 2.0441182078169685e-06, "loss": 2.4753, "step": 16762 }, { "epoch": 3.19, "learning_rate": 2.0440610104861773e-06, "loss": 1.0773, "step": 16763 }, { "epoch": 3.19, "learning_rate": 2.044003813155386e-06, "loss": 2.4326, "step": 16764 }, { "epoch": 3.19, "learning_rate": 2.043946615824595e-06, "loss": 3.1099, "step": 16765 }, { "epoch": 3.19, "learning_rate": 2.0438894184938037e-06, "loss": 2.3271, "step": 16766 }, { "epoch": 3.19, "learning_rate": 2.0438322211630125e-06, "loss": 2.7343, "step": 16767 }, { "epoch": 3.19, "learning_rate": 2.0437750238322213e-06, "loss": 1.6762, "step": 16768 }, { "epoch": 3.19, "learning_rate": 2.04371782650143e-06, "loss": 2.5883, "step": 16769 }, { "epoch": 3.19, "learning_rate": 2.043660629170639e-06, "loss": 2.7839, "step": 16770 }, { "epoch": 3.19, "learning_rate": 2.0436034318398476e-06, "loss": 1.6441, "step": 16771 }, { "epoch": 3.19, "learning_rate": 2.0435462345090564e-06, "loss": 2.9848, "step": 16772 }, { "epoch": 3.19, "learning_rate": 2.0434890371782652e-06, "loss": 1.9843, "step": 16773 }, { "epoch": 3.2, "learning_rate": 2.0434318398474736e-06, "loss": 3.1121, "step": 16774 }, { "epoch": 3.2, "learning_rate": 2.043374642516683e-06, "loss": 1.4133, "step": 16775 }, { "epoch": 3.2, "learning_rate": 2.043317445185891e-06, "loss": 2.9697, "step": 16776 }, { "epoch": 3.2, "learning_rate": 2.0432602478551e-06, "loss": 1.7332, "step": 16777 }, { "epoch": 3.2, "learning_rate": 2.043203050524309e-06, "loss": 1.9212, "step": 16778 }, { "epoch": 3.2, "learning_rate": 2.0431458531935176e-06, "loss": 2.7517, "step": 16779 }, { "epoch": 3.2, "learning_rate": 2.0430886558627264e-06, "loss": 2.0973, "step": 16780 }, { "epoch": 3.2, "learning_rate": 2.043031458531935e-06, "loss": 2.4436, "step": 16781 }, { "epoch": 3.2, "learning_rate": 2.042974261201144e-06, "loss": 2.8706, "step": 16782 }, { "epoch": 3.2, "learning_rate": 2.0429170638703527e-06, "loss": 3.6649, "step": 16783 }, { "epoch": 3.2, "learning_rate": 2.0428598665395615e-06, "loss": 2.2966, "step": 16784 }, { "epoch": 3.2, "learning_rate": 2.0428026692087703e-06, "loss": 2.7437, "step": 16785 }, { "epoch": 3.2, "learning_rate": 2.042745471877979e-06, "loss": 3.318, "step": 16786 }, { "epoch": 3.2, "learning_rate": 2.042688274547188e-06, "loss": 1.6257, "step": 16787 }, { "epoch": 3.2, "learning_rate": 2.0426310772163967e-06, "loss": 3.0707, "step": 16788 }, { "epoch": 3.2, "learning_rate": 2.0425738798856055e-06, "loss": 2.795, "step": 16789 }, { "epoch": 3.2, "learning_rate": 2.0425166825548143e-06, "loss": 2.1353, "step": 16790 }, { "epoch": 3.2, "learning_rate": 2.0424594852240227e-06, "loss": 2.7438, "step": 16791 }, { "epoch": 3.2, "learning_rate": 2.042402287893232e-06, "loss": 2.8827, "step": 16792 }, { "epoch": 3.2, "learning_rate": 2.0423450905624403e-06, "loss": 2.0029, "step": 16793 }, { "epoch": 3.2, "learning_rate": 2.042287893231649e-06, "loss": 3.0683, "step": 16794 }, { "epoch": 3.2, "learning_rate": 2.0422306959008583e-06, "loss": 2.6447, "step": 16795 }, { "epoch": 3.2, "learning_rate": 2.0421734985700666e-06, "loss": 2.5689, "step": 16796 }, { "epoch": 3.2, "learning_rate": 2.0421163012392754e-06, "loss": 1.5746, "step": 16797 }, { "epoch": 3.2, "learning_rate": 2.0420591039084842e-06, "loss": 1.6857, "step": 16798 }, { "epoch": 3.2, "learning_rate": 2.042001906577693e-06, "loss": 2.6628, "step": 16799 }, { "epoch": 3.2, "learning_rate": 2.0419447092469022e-06, "loss": 2.4369, "step": 16800 }, { "epoch": 3.2, "learning_rate": 2.0418875119161106e-06, "loss": 2.3439, "step": 16801 }, { "epoch": 3.2, "learning_rate": 2.0418303145853194e-06, "loss": 2.4552, "step": 16802 }, { "epoch": 3.2, "learning_rate": 2.041773117254528e-06, "loss": 2.2799, "step": 16803 }, { "epoch": 3.2, "learning_rate": 2.041715919923737e-06, "loss": 1.6864, "step": 16804 }, { "epoch": 3.2, "learning_rate": 2.0416587225929458e-06, "loss": 1.8634, "step": 16805 }, { "epoch": 3.2, "learning_rate": 2.0416015252621546e-06, "loss": 2.6532, "step": 16806 }, { "epoch": 3.2, "learning_rate": 2.0415443279313634e-06, "loss": 2.1898, "step": 16807 }, { "epoch": 3.2, "learning_rate": 2.0414871306005717e-06, "loss": 2.4833, "step": 16808 }, { "epoch": 3.2, "learning_rate": 2.041429933269781e-06, "loss": 2.3544, "step": 16809 }, { "epoch": 3.2, "learning_rate": 2.0413727359389897e-06, "loss": 1.256, "step": 16810 }, { "epoch": 3.2, "learning_rate": 2.041315538608198e-06, "loss": 2.7419, "step": 16811 }, { "epoch": 3.2, "learning_rate": 2.0412583412774073e-06, "loss": 3.4975, "step": 16812 }, { "epoch": 3.2, "learning_rate": 2.0412011439466157e-06, "loss": 1.1153, "step": 16813 }, { "epoch": 3.2, "learning_rate": 2.0411439466158245e-06, "loss": 3.4966, "step": 16814 }, { "epoch": 3.2, "learning_rate": 2.0410867492850333e-06, "loss": 1.4998, "step": 16815 }, { "epoch": 3.2, "learning_rate": 2.041029551954242e-06, "loss": 2.8624, "step": 16816 }, { "epoch": 3.2, "learning_rate": 2.0409723546234513e-06, "loss": 2.7389, "step": 16817 }, { "epoch": 3.2, "learning_rate": 2.0409151572926597e-06, "loss": 2.5028, "step": 16818 }, { "epoch": 3.2, "learning_rate": 2.0408579599618685e-06, "loss": 2.1902, "step": 16819 }, { "epoch": 3.2, "learning_rate": 2.0408007626310773e-06, "loss": 3.0428, "step": 16820 }, { "epoch": 3.2, "learning_rate": 2.040743565300286e-06, "loss": 1.9253, "step": 16821 }, { "epoch": 3.2, "learning_rate": 2.040686367969495e-06, "loss": 2.7327, "step": 16822 }, { "epoch": 3.2, "learning_rate": 2.0406291706387036e-06, "loss": 1.4964, "step": 16823 }, { "epoch": 3.2, "learning_rate": 2.0405719733079124e-06, "loss": 2.0708, "step": 16824 }, { "epoch": 3.2, "learning_rate": 2.040514775977121e-06, "loss": 2.1961, "step": 16825 }, { "epoch": 3.2, "learning_rate": 2.04045757864633e-06, "loss": 2.3588, "step": 16826 }, { "epoch": 3.21, "learning_rate": 2.040400381315539e-06, "loss": 1.8624, "step": 16827 }, { "epoch": 3.21, "learning_rate": 2.040343183984747e-06, "loss": 2.343, "step": 16828 }, { "epoch": 3.21, "learning_rate": 2.0402859866539564e-06, "loss": 2.5181, "step": 16829 }, { "epoch": 3.21, "learning_rate": 2.0402287893231648e-06, "loss": 3.1466, "step": 16830 }, { "epoch": 3.21, "learning_rate": 2.0401715919923736e-06, "loss": 2.4992, "step": 16831 }, { "epoch": 3.21, "learning_rate": 2.0401143946615824e-06, "loss": 2.594, "step": 16832 }, { "epoch": 3.21, "learning_rate": 2.040057197330791e-06, "loss": 2.4957, "step": 16833 }, { "epoch": 3.21, "learning_rate": 2.0400000000000004e-06, "loss": 2.386, "step": 16834 }, { "epoch": 3.21, "learning_rate": 2.0399428026692087e-06, "loss": 2.0842, "step": 16835 }, { "epoch": 3.21, "learning_rate": 2.0398856053384175e-06, "loss": 2.9105, "step": 16836 }, { "epoch": 3.21, "learning_rate": 2.0398284080076263e-06, "loss": 2.4739, "step": 16837 }, { "epoch": 3.21, "learning_rate": 2.039771210676835e-06, "loss": 2.3077, "step": 16838 }, { "epoch": 3.21, "learning_rate": 2.039714013346044e-06, "loss": 1.9023, "step": 16839 }, { "epoch": 3.21, "learning_rate": 2.0396568160152527e-06, "loss": 1.9899, "step": 16840 }, { "epoch": 3.21, "learning_rate": 2.0395996186844615e-06, "loss": 2.5543, "step": 16841 }, { "epoch": 3.21, "learning_rate": 2.03954242135367e-06, "loss": 2.9171, "step": 16842 }, { "epoch": 3.21, "learning_rate": 2.039485224022879e-06, "loss": 1.4282, "step": 16843 }, { "epoch": 3.21, "learning_rate": 2.039428026692088e-06, "loss": 2.863, "step": 16844 }, { "epoch": 3.21, "learning_rate": 2.0393708293612963e-06, "loss": 2.4429, "step": 16845 }, { "epoch": 3.21, "learning_rate": 2.0393136320305055e-06, "loss": 2.5819, "step": 16846 }, { "epoch": 3.21, "learning_rate": 2.039256434699714e-06, "loss": 2.3873, "step": 16847 }, { "epoch": 3.21, "learning_rate": 2.039199237368923e-06, "loss": 2.604, "step": 16848 }, { "epoch": 3.21, "learning_rate": 2.039142040038132e-06, "loss": 2.2886, "step": 16849 }, { "epoch": 3.21, "learning_rate": 2.0390848427073402e-06, "loss": 2.7806, "step": 16850 }, { "epoch": 3.21, "learning_rate": 2.0390276453765494e-06, "loss": 1.9351, "step": 16851 }, { "epoch": 3.21, "learning_rate": 2.038970448045758e-06, "loss": 2.1554, "step": 16852 }, { "epoch": 3.21, "learning_rate": 2.0389132507149666e-06, "loss": 3.2593, "step": 16853 }, { "epoch": 3.21, "learning_rate": 2.0388560533841754e-06, "loss": 2.6736, "step": 16854 }, { "epoch": 3.21, "learning_rate": 2.038798856053384e-06, "loss": 1.9313, "step": 16855 }, { "epoch": 3.21, "learning_rate": 2.038741658722593e-06, "loss": 2.2998, "step": 16856 }, { "epoch": 3.21, "learning_rate": 2.0386844613918018e-06, "loss": 2.8678, "step": 16857 }, { "epoch": 3.21, "learning_rate": 2.0386272640610106e-06, "loss": 2.5548, "step": 16858 }, { "epoch": 3.21, "learning_rate": 2.0385700667302194e-06, "loss": 2.6915, "step": 16859 }, { "epoch": 3.21, "learning_rate": 2.038512869399428e-06, "loss": 2.3248, "step": 16860 }, { "epoch": 3.21, "learning_rate": 2.038455672068637e-06, "loss": 1.8174, "step": 16861 }, { "epoch": 3.21, "learning_rate": 2.0383984747378453e-06, "loss": 3.3836, "step": 16862 }, { "epoch": 3.21, "learning_rate": 2.0383412774070545e-06, "loss": 2.8854, "step": 16863 }, { "epoch": 3.21, "learning_rate": 2.038284080076263e-06, "loss": 2.5319, "step": 16864 }, { "epoch": 3.21, "learning_rate": 2.038226882745472e-06, "loss": 1.129, "step": 16865 }, { "epoch": 3.21, "learning_rate": 2.038169685414681e-06, "loss": 2.187, "step": 16866 }, { "epoch": 3.21, "learning_rate": 2.0381124880838893e-06, "loss": 2.8928, "step": 16867 }, { "epoch": 3.21, "learning_rate": 2.0380552907530985e-06, "loss": 2.0121, "step": 16868 }, { "epoch": 3.21, "learning_rate": 2.037998093422307e-06, "loss": 2.5582, "step": 16869 }, { "epoch": 3.21, "learning_rate": 2.0379408960915157e-06, "loss": 2.0354, "step": 16870 }, { "epoch": 3.21, "learning_rate": 2.0378836987607245e-06, "loss": 2.9404, "step": 16871 }, { "epoch": 3.21, "learning_rate": 2.0378265014299333e-06, "loss": 2.3135, "step": 16872 }, { "epoch": 3.21, "learning_rate": 2.037769304099142e-06, "loss": 2.3612, "step": 16873 }, { "epoch": 3.21, "learning_rate": 2.037712106768351e-06, "loss": 2.4544, "step": 16874 }, { "epoch": 3.21, "learning_rate": 2.0376549094375596e-06, "loss": 3.1454, "step": 16875 }, { "epoch": 3.21, "learning_rate": 2.0375977121067684e-06, "loss": 2.7966, "step": 16876 }, { "epoch": 3.21, "learning_rate": 2.0375405147759772e-06, "loss": 2.0379, "step": 16877 }, { "epoch": 3.21, "learning_rate": 2.037483317445186e-06, "loss": 2.2612, "step": 16878 }, { "epoch": 3.22, "learning_rate": 2.0374261201143944e-06, "loss": 2.3052, "step": 16879 }, { "epoch": 3.22, "learning_rate": 2.0373689227836036e-06, "loss": 2.1313, "step": 16880 }, { "epoch": 3.22, "learning_rate": 2.037311725452812e-06, "loss": 2.0935, "step": 16881 }, { "epoch": 3.22, "learning_rate": 2.037254528122021e-06, "loss": 2.4195, "step": 16882 }, { "epoch": 3.22, "learning_rate": 2.03719733079123e-06, "loss": 2.7672, "step": 16883 }, { "epoch": 3.22, "learning_rate": 2.0371401334604384e-06, "loss": 2.3355, "step": 16884 }, { "epoch": 3.22, "learning_rate": 2.0370829361296476e-06, "loss": 1.9599, "step": 16885 }, { "epoch": 3.22, "learning_rate": 2.037025738798856e-06, "loss": 3.3755, "step": 16886 }, { "epoch": 3.22, "learning_rate": 2.0369685414680648e-06, "loss": 2.0271, "step": 16887 }, { "epoch": 3.22, "learning_rate": 2.0369113441372735e-06, "loss": 1.91, "step": 16888 }, { "epoch": 3.22, "learning_rate": 2.0368541468064823e-06, "loss": 1.3033, "step": 16889 }, { "epoch": 3.22, "learning_rate": 2.036796949475691e-06, "loss": 1.4355, "step": 16890 }, { "epoch": 3.22, "learning_rate": 2.0367397521449e-06, "loss": 2.2845, "step": 16891 }, { "epoch": 3.22, "learning_rate": 2.0366825548141087e-06, "loss": 2.0776, "step": 16892 }, { "epoch": 3.22, "learning_rate": 2.0366253574833175e-06, "loss": 2.199, "step": 16893 }, { "epoch": 3.22, "learning_rate": 2.0365681601525263e-06, "loss": 2.6366, "step": 16894 }, { "epoch": 3.22, "learning_rate": 2.036510962821735e-06, "loss": 2.4312, "step": 16895 }, { "epoch": 3.22, "learning_rate": 2.0364537654909435e-06, "loss": 2.253, "step": 16896 }, { "epoch": 3.22, "learning_rate": 2.0363965681601527e-06, "loss": 2.3515, "step": 16897 }, { "epoch": 3.22, "learning_rate": 2.0363393708293615e-06, "loss": 3.5371, "step": 16898 }, { "epoch": 3.22, "learning_rate": 2.0362821734985703e-06, "loss": 1.2052, "step": 16899 }, { "epoch": 3.22, "learning_rate": 2.036224976167779e-06, "loss": 2.4762, "step": 16900 }, { "epoch": 3.22, "learning_rate": 2.0361677788369874e-06, "loss": 3.1563, "step": 16901 }, { "epoch": 3.22, "learning_rate": 2.0361105815061967e-06, "loss": 1.4211, "step": 16902 }, { "epoch": 3.22, "learning_rate": 2.036053384175405e-06, "loss": 3.1191, "step": 16903 }, { "epoch": 3.22, "learning_rate": 2.035996186844614e-06, "loss": 2.3109, "step": 16904 }, { "epoch": 3.22, "learning_rate": 2.035938989513823e-06, "loss": 1.9628, "step": 16905 }, { "epoch": 3.22, "learning_rate": 2.0358817921830314e-06, "loss": 2.0617, "step": 16906 }, { "epoch": 3.22, "learning_rate": 2.03582459485224e-06, "loss": 3.4432, "step": 16907 }, { "epoch": 3.22, "learning_rate": 2.035767397521449e-06, "loss": 3.0607, "step": 16908 }, { "epoch": 3.22, "learning_rate": 2.035710200190658e-06, "loss": 3.9265, "step": 16909 }, { "epoch": 3.22, "learning_rate": 2.0356530028598666e-06, "loss": 2.9113, "step": 16910 }, { "epoch": 3.22, "learning_rate": 2.0355958055290754e-06, "loss": 3.3658, "step": 16911 }, { "epoch": 3.22, "learning_rate": 2.035538608198284e-06, "loss": 2.6403, "step": 16912 }, { "epoch": 3.22, "learning_rate": 2.035481410867493e-06, "loss": 3.4468, "step": 16913 }, { "epoch": 3.22, "learning_rate": 2.0354242135367018e-06, "loss": 3.0616, "step": 16914 }, { "epoch": 3.22, "learning_rate": 2.0353670162059106e-06, "loss": 2.5635, "step": 16915 }, { "epoch": 3.22, "learning_rate": 2.0353098188751193e-06, "loss": 3.3683, "step": 16916 }, { "epoch": 3.22, "learning_rate": 2.035252621544328e-06, "loss": 2.4197, "step": 16917 }, { "epoch": 3.22, "learning_rate": 2.0351954242135365e-06, "loss": 3.1376, "step": 16918 }, { "epoch": 3.22, "learning_rate": 2.0351382268827457e-06, "loss": 2.5086, "step": 16919 }, { "epoch": 3.22, "learning_rate": 2.035081029551954e-06, "loss": 3.1892, "step": 16920 }, { "epoch": 3.22, "learning_rate": 2.035023832221163e-06, "loss": 2.4732, "step": 16921 }, { "epoch": 3.22, "learning_rate": 2.034966634890372e-06, "loss": 1.9692, "step": 16922 }, { "epoch": 3.22, "learning_rate": 2.0349094375595805e-06, "loss": 1.8472, "step": 16923 }, { "epoch": 3.22, "learning_rate": 2.0348522402287893e-06, "loss": 2.6806, "step": 16924 }, { "epoch": 3.22, "learning_rate": 2.034795042897998e-06, "loss": 2.2671, "step": 16925 }, { "epoch": 3.22, "learning_rate": 2.034737845567207e-06, "loss": 2.6733, "step": 16926 }, { "epoch": 3.22, "learning_rate": 2.0346806482364157e-06, "loss": 2.3045, "step": 16927 }, { "epoch": 3.22, "learning_rate": 2.0346234509056245e-06, "loss": 1.8369, "step": 16928 }, { "epoch": 3.22, "learning_rate": 2.0345662535748332e-06, "loss": 3.7035, "step": 16929 }, { "epoch": 3.22, "learning_rate": 2.034509056244042e-06, "loss": 3.66, "step": 16930 }, { "epoch": 3.22, "learning_rate": 2.034451858913251e-06, "loss": 2.6365, "step": 16931 }, { "epoch": 3.23, "learning_rate": 2.0343946615824596e-06, "loss": 3.0527, "step": 16932 }, { "epoch": 3.23, "learning_rate": 2.0343374642516684e-06, "loss": 1.7942, "step": 16933 }, { "epoch": 3.23, "learning_rate": 2.0342802669208772e-06, "loss": 2.209, "step": 16934 }, { "epoch": 3.23, "learning_rate": 2.0342230695900856e-06, "loss": 1.9369, "step": 16935 }, { "epoch": 3.23, "learning_rate": 2.034165872259295e-06, "loss": 2.6759, "step": 16936 }, { "epoch": 3.23, "learning_rate": 2.0341086749285036e-06, "loss": 2.823, "step": 16937 }, { "epoch": 3.23, "learning_rate": 2.034051477597712e-06, "loss": 1.777, "step": 16938 }, { "epoch": 3.23, "learning_rate": 2.033994280266921e-06, "loss": 2.4264, "step": 16939 }, { "epoch": 3.23, "learning_rate": 2.0339370829361296e-06, "loss": 2.1021, "step": 16940 }, { "epoch": 3.23, "learning_rate": 2.0338798856053383e-06, "loss": 2.2998, "step": 16941 }, { "epoch": 3.23, "learning_rate": 2.033822688274547e-06, "loss": 1.6074, "step": 16942 }, { "epoch": 3.23, "learning_rate": 2.033765490943756e-06, "loss": 2.1001, "step": 16943 }, { "epoch": 3.23, "learning_rate": 2.0337082936129647e-06, "loss": 2.887, "step": 16944 }, { "epoch": 3.23, "learning_rate": 2.0336510962821735e-06, "loss": 3.1046, "step": 16945 }, { "epoch": 3.23, "learning_rate": 2.0335938989513823e-06, "loss": 2.3616, "step": 16946 }, { "epoch": 3.23, "learning_rate": 2.033536701620591e-06, "loss": 2.7802, "step": 16947 }, { "epoch": 3.23, "learning_rate": 2.0334795042898e-06, "loss": 2.0705, "step": 16948 }, { "epoch": 3.23, "learning_rate": 2.0334223069590087e-06, "loss": 3.0954, "step": 16949 }, { "epoch": 3.23, "learning_rate": 2.0333651096282175e-06, "loss": 2.3115, "step": 16950 }, { "epoch": 3.23, "learning_rate": 2.0333079122974263e-06, "loss": 2.8217, "step": 16951 }, { "epoch": 3.23, "learning_rate": 2.0332507149666347e-06, "loss": 2.2924, "step": 16952 }, { "epoch": 3.23, "learning_rate": 2.033193517635844e-06, "loss": 2.6336, "step": 16953 }, { "epoch": 3.23, "learning_rate": 2.0331363203050527e-06, "loss": 2.0158, "step": 16954 }, { "epoch": 3.23, "learning_rate": 2.033079122974261e-06, "loss": 2.3213, "step": 16955 }, { "epoch": 3.23, "learning_rate": 2.0330219256434703e-06, "loss": 2.9644, "step": 16956 }, { "epoch": 3.23, "learning_rate": 2.0329647283126786e-06, "loss": 2.3621, "step": 16957 }, { "epoch": 3.23, "learning_rate": 2.0329075309818874e-06, "loss": 2.4485, "step": 16958 }, { "epoch": 3.23, "learning_rate": 2.0328503336510962e-06, "loss": 1.6534, "step": 16959 }, { "epoch": 3.23, "learning_rate": 2.032793136320305e-06, "loss": 2.4468, "step": 16960 }, { "epoch": 3.23, "learning_rate": 2.032735938989514e-06, "loss": 2.4799, "step": 16961 }, { "epoch": 3.23, "learning_rate": 2.0326787416587226e-06, "loss": 2.4207, "step": 16962 }, { "epoch": 3.23, "learning_rate": 2.0326215443279314e-06, "loss": 2.6865, "step": 16963 }, { "epoch": 3.23, "learning_rate": 2.03256434699714e-06, "loss": 1.6789, "step": 16964 }, { "epoch": 3.23, "learning_rate": 2.032507149666349e-06, "loss": 2.6165, "step": 16965 }, { "epoch": 3.23, "learning_rate": 2.0324499523355578e-06, "loss": 2.4588, "step": 16966 }, { "epoch": 3.23, "learning_rate": 2.0323927550047666e-06, "loss": 2.0665, "step": 16967 }, { "epoch": 3.23, "learning_rate": 2.0323355576739754e-06, "loss": 2.1543, "step": 16968 }, { "epoch": 3.23, "learning_rate": 2.0322783603431837e-06, "loss": 1.9393, "step": 16969 }, { "epoch": 3.23, "learning_rate": 2.032221163012393e-06, "loss": 2.656, "step": 16970 }, { "epoch": 3.23, "learning_rate": 2.0321639656816017e-06, "loss": 3.2465, "step": 16971 }, { "epoch": 3.23, "learning_rate": 2.03210676835081e-06, "loss": 2.5877, "step": 16972 }, { "epoch": 3.23, "learning_rate": 2.0320495710200193e-06, "loss": 2.846, "step": 16973 }, { "epoch": 3.23, "learning_rate": 2.0319923736892277e-06, "loss": 2.0889, "step": 16974 }, { "epoch": 3.23, "learning_rate": 2.0319351763584365e-06, "loss": 2.6697, "step": 16975 }, { "epoch": 3.23, "learning_rate": 2.0318779790276453e-06, "loss": 1.7648, "step": 16976 }, { "epoch": 3.23, "learning_rate": 2.031820781696854e-06, "loss": 2.5948, "step": 16977 }, { "epoch": 3.23, "learning_rate": 2.0317635843660633e-06, "loss": 2.7757, "step": 16978 }, { "epoch": 3.23, "learning_rate": 2.0317063870352717e-06, "loss": 3.2422, "step": 16979 }, { "epoch": 3.23, "learning_rate": 2.0316491897044805e-06, "loss": 1.6866, "step": 16980 }, { "epoch": 3.23, "learning_rate": 2.0315919923736893e-06, "loss": 2.9197, "step": 16981 }, { "epoch": 3.23, "learning_rate": 2.031534795042898e-06, "loss": 1.753, "step": 16982 }, { "epoch": 3.23, "learning_rate": 2.031477597712107e-06, "loss": 2.0044, "step": 16983 }, { "epoch": 3.24, "learning_rate": 2.0314204003813156e-06, "loss": 2.6515, "step": 16984 }, { "epoch": 3.24, "learning_rate": 2.0313632030505244e-06, "loss": 2.1, "step": 16985 }, { "epoch": 3.24, "learning_rate": 2.0313060057197332e-06, "loss": 2.1256, "step": 16986 }, { "epoch": 3.24, "learning_rate": 2.031248808388942e-06, "loss": 2.0057, "step": 16987 }, { "epoch": 3.24, "learning_rate": 2.031191611058151e-06, "loss": 2.8326, "step": 16988 }, { "epoch": 3.24, "learning_rate": 2.031134413727359e-06, "loss": 1.4401, "step": 16989 }, { "epoch": 3.24, "learning_rate": 2.0310772163965684e-06, "loss": 2.6231, "step": 16990 }, { "epoch": 3.24, "learning_rate": 2.0310200190657768e-06, "loss": 3.2042, "step": 16991 }, { "epoch": 3.24, "learning_rate": 2.0309628217349856e-06, "loss": 2.0148, "step": 16992 }, { "epoch": 3.24, "learning_rate": 2.0309056244041948e-06, "loss": 2.9744, "step": 16993 }, { "epoch": 3.24, "learning_rate": 2.030848427073403e-06, "loss": 4.1117, "step": 16994 }, { "epoch": 3.24, "learning_rate": 2.0307912297426124e-06, "loss": 1.999, "step": 16995 }, { "epoch": 3.24, "learning_rate": 2.0307340324118207e-06, "loss": 1.5314, "step": 16996 }, { "epoch": 3.24, "learning_rate": 2.0306768350810295e-06, "loss": 1.9657, "step": 16997 }, { "epoch": 3.24, "learning_rate": 2.0306196377502383e-06, "loss": 2.5961, "step": 16998 }, { "epoch": 3.24, "learning_rate": 2.030562440419447e-06, "loss": 3.5722, "step": 16999 }, { "epoch": 3.24, "learning_rate": 2.030505243088656e-06, "loss": 2.0042, "step": 17000 }, { "epoch": 3.24, "learning_rate": 2.0304480457578647e-06, "loss": 3.1454, "step": 17001 }, { "epoch": 3.24, "learning_rate": 2.0303908484270735e-06, "loss": 3.6918, "step": 17002 }, { "epoch": 3.24, "learning_rate": 2.0303336510962823e-06, "loss": 2.0287, "step": 17003 }, { "epoch": 3.24, "learning_rate": 2.030276453765491e-06, "loss": 2.4703, "step": 17004 }, { "epoch": 3.24, "learning_rate": 2.0302192564347e-06, "loss": 1.9825, "step": 17005 }, { "epoch": 3.24, "learning_rate": 2.0301620591039083e-06, "loss": 2.2115, "step": 17006 }, { "epoch": 3.24, "learning_rate": 2.0301048617731175e-06, "loss": 3.718, "step": 17007 }, { "epoch": 3.24, "learning_rate": 2.030047664442326e-06, "loss": 3.2177, "step": 17008 }, { "epoch": 3.24, "learning_rate": 2.0299904671115346e-06, "loss": 1.279, "step": 17009 }, { "epoch": 3.24, "learning_rate": 2.029933269780744e-06, "loss": 2.3211, "step": 17010 }, { "epoch": 3.24, "learning_rate": 2.0298760724499522e-06, "loss": 2.604, "step": 17011 }, { "epoch": 3.24, "learning_rate": 2.0298188751191614e-06, "loss": 2.9612, "step": 17012 }, { "epoch": 3.24, "learning_rate": 2.02976167778837e-06, "loss": 1.5265, "step": 17013 }, { "epoch": 3.24, "learning_rate": 2.0297044804575786e-06, "loss": 2.5742, "step": 17014 }, { "epoch": 3.24, "learning_rate": 2.0296472831267874e-06, "loss": 2.0956, "step": 17015 }, { "epoch": 3.24, "learning_rate": 2.029590085795996e-06, "loss": 3.3537, "step": 17016 }, { "epoch": 3.24, "learning_rate": 2.029532888465205e-06, "loss": 1.8301, "step": 17017 }, { "epoch": 3.24, "learning_rate": 2.0294756911344138e-06, "loss": 2.4285, "step": 17018 }, { "epoch": 3.24, "learning_rate": 2.0294184938036226e-06, "loss": 2.7259, "step": 17019 }, { "epoch": 3.24, "learning_rate": 2.0293612964728314e-06, "loss": 2.2626, "step": 17020 }, { "epoch": 3.24, "learning_rate": 2.02930409914204e-06, "loss": 2.7278, "step": 17021 }, { "epoch": 3.24, "learning_rate": 2.029246901811249e-06, "loss": 2.4067, "step": 17022 }, { "epoch": 3.24, "learning_rate": 2.0291897044804573e-06, "loss": 3.2522, "step": 17023 }, { "epoch": 3.24, "learning_rate": 2.0291325071496665e-06, "loss": 2.2022, "step": 17024 }, { "epoch": 3.24, "learning_rate": 2.0290753098188753e-06, "loss": 2.4981, "step": 17025 }, { "epoch": 3.24, "learning_rate": 2.0290181124880837e-06, "loss": 2.5992, "step": 17026 }, { "epoch": 3.24, "learning_rate": 2.028960915157293e-06, "loss": 1.8628, "step": 17027 }, { "epoch": 3.24, "learning_rate": 2.0289037178265013e-06, "loss": 3.2227, "step": 17028 }, { "epoch": 3.24, "learning_rate": 2.0288465204957105e-06, "loss": 2.5163, "step": 17029 }, { "epoch": 3.24, "learning_rate": 2.028789323164919e-06, "loss": 1.8434, "step": 17030 }, { "epoch": 3.24, "learning_rate": 2.0287321258341277e-06, "loss": 2.0361, "step": 17031 }, { "epoch": 3.24, "learning_rate": 2.028674928503337e-06, "loss": 2.4352, "step": 17032 }, { "epoch": 3.24, "learning_rate": 2.0286177311725453e-06, "loss": 2.5525, "step": 17033 }, { "epoch": 3.24, "learning_rate": 2.028560533841754e-06, "loss": 3.3237, "step": 17034 }, { "epoch": 3.24, "learning_rate": 2.028503336510963e-06, "loss": 2.7239, "step": 17035 }, { "epoch": 3.24, "learning_rate": 2.0284461391801716e-06, "loss": 2.0349, "step": 17036 }, { "epoch": 3.25, "learning_rate": 2.0283889418493804e-06, "loss": 2.2127, "step": 17037 }, { "epoch": 3.25, "learning_rate": 2.0283317445185892e-06, "loss": 2.3655, "step": 17038 }, { "epoch": 3.25, "learning_rate": 2.028274547187798e-06, "loss": 1.8061, "step": 17039 }, { "epoch": 3.25, "learning_rate": 2.0282173498570064e-06, "loss": 1.5168, "step": 17040 }, { "epoch": 3.25, "learning_rate": 2.0281601525262156e-06, "loss": 2.7081, "step": 17041 }, { "epoch": 3.25, "learning_rate": 2.0281029551954244e-06, "loss": 2.8796, "step": 17042 }, { "epoch": 3.25, "learning_rate": 2.028045757864633e-06, "loss": 1.8809, "step": 17043 }, { "epoch": 3.25, "learning_rate": 2.027988560533842e-06, "loss": 3.0647, "step": 17044 }, { "epoch": 3.25, "learning_rate": 2.0279313632030504e-06, "loss": 1.9535, "step": 17045 }, { "epoch": 3.25, "learning_rate": 2.0278741658722596e-06, "loss": 2.5025, "step": 17046 }, { "epoch": 3.25, "learning_rate": 2.027816968541468e-06, "loss": 1.9792, "step": 17047 }, { "epoch": 3.25, "learning_rate": 2.0277597712106767e-06, "loss": 3.4972, "step": 17048 }, { "epoch": 3.25, "learning_rate": 2.027702573879886e-06, "loss": 1.9542, "step": 17049 }, { "epoch": 3.25, "learning_rate": 2.0276453765490943e-06, "loss": 1.3829, "step": 17050 }, { "epoch": 3.25, "learning_rate": 2.027588179218303e-06, "loss": 1.7699, "step": 17051 }, { "epoch": 3.25, "learning_rate": 2.027530981887512e-06, "loss": 1.9659, "step": 17052 }, { "epoch": 3.25, "learning_rate": 2.0274737845567207e-06, "loss": 2.6429, "step": 17053 }, { "epoch": 3.25, "learning_rate": 2.0274165872259295e-06, "loss": 2.6467, "step": 17054 }, { "epoch": 3.25, "learning_rate": 2.0273593898951383e-06, "loss": 3.2767, "step": 17055 }, { "epoch": 3.25, "learning_rate": 2.027302192564347e-06, "loss": 3.1549, "step": 17056 }, { "epoch": 3.25, "learning_rate": 2.0272449952335555e-06, "loss": 2.8499, "step": 17057 }, { "epoch": 3.25, "learning_rate": 2.0271877979027647e-06, "loss": 2.1175, "step": 17058 }, { "epoch": 3.25, "learning_rate": 2.0271306005719735e-06, "loss": 2.9832, "step": 17059 }, { "epoch": 3.25, "learning_rate": 2.0270734032411823e-06, "loss": 2.1636, "step": 17060 }, { "epoch": 3.25, "learning_rate": 2.027016205910391e-06, "loss": 2.1731, "step": 17061 }, { "epoch": 3.25, "learning_rate": 2.0269590085795994e-06, "loss": 2.7847, "step": 17062 }, { "epoch": 3.25, "learning_rate": 2.0269018112488087e-06, "loss": 2.7493, "step": 17063 }, { "epoch": 3.25, "learning_rate": 2.026844613918017e-06, "loss": 3.1191, "step": 17064 }, { "epoch": 3.25, "learning_rate": 2.026787416587226e-06, "loss": 3.0157, "step": 17065 }, { "epoch": 3.25, "learning_rate": 2.026730219256435e-06, "loss": 2.3837, "step": 17066 }, { "epoch": 3.25, "learning_rate": 2.0266730219256434e-06, "loss": 1.6515, "step": 17067 }, { "epoch": 3.25, "learning_rate": 2.026615824594852e-06, "loss": 3.006, "step": 17068 }, { "epoch": 3.25, "learning_rate": 2.026558627264061e-06, "loss": 2.2005, "step": 17069 }, { "epoch": 3.25, "learning_rate": 2.0265014299332698e-06, "loss": 1.5487, "step": 17070 }, { "epoch": 3.25, "learning_rate": 2.0264442326024786e-06, "loss": 2.1644, "step": 17071 }, { "epoch": 3.25, "learning_rate": 2.0263870352716874e-06, "loss": 2.1576, "step": 17072 }, { "epoch": 3.25, "learning_rate": 2.026329837940896e-06, "loss": 3.8138, "step": 17073 }, { "epoch": 3.25, "learning_rate": 2.026272640610105e-06, "loss": 2.697, "step": 17074 }, { "epoch": 3.25, "learning_rate": 2.0262154432793138e-06, "loss": 2.3199, "step": 17075 }, { "epoch": 3.25, "learning_rate": 2.0261582459485225e-06, "loss": 1.7355, "step": 17076 }, { "epoch": 3.25, "learning_rate": 2.0261010486177313e-06, "loss": 3.4152, "step": 17077 }, { "epoch": 3.25, "learning_rate": 2.02604385128694e-06, "loss": 2.9823, "step": 17078 }, { "epoch": 3.25, "learning_rate": 2.0259866539561485e-06, "loss": 3.2492, "step": 17079 }, { "epoch": 3.25, "learning_rate": 2.0259294566253577e-06, "loss": 2.6908, "step": 17080 }, { "epoch": 3.25, "learning_rate": 2.0258722592945665e-06, "loss": 2.3361, "step": 17081 }, { "epoch": 3.25, "learning_rate": 2.025815061963775e-06, "loss": 1.9515, "step": 17082 }, { "epoch": 3.25, "learning_rate": 2.025757864632984e-06, "loss": 2.1831, "step": 17083 }, { "epoch": 3.25, "learning_rate": 2.0257006673021925e-06, "loss": 2.6594, "step": 17084 }, { "epoch": 3.25, "learning_rate": 2.0256434699714013e-06, "loss": 2.7891, "step": 17085 }, { "epoch": 3.25, "learning_rate": 2.02558627264061e-06, "loss": 2.4515, "step": 17086 }, { "epoch": 3.25, "learning_rate": 2.025529075309819e-06, "loss": 2.5096, "step": 17087 }, { "epoch": 3.25, "learning_rate": 2.0254718779790277e-06, "loss": 2.9392, "step": 17088 }, { "epoch": 3.26, "learning_rate": 2.0254146806482364e-06, "loss": 2.5179, "step": 17089 }, { "epoch": 3.26, "learning_rate": 2.0253574833174452e-06, "loss": 3.2636, "step": 17090 }, { "epoch": 3.26, "learning_rate": 2.025300285986654e-06, "loss": 2.4552, "step": 17091 }, { "epoch": 3.26, "learning_rate": 2.025243088655863e-06, "loss": 2.8439, "step": 17092 }, { "epoch": 3.26, "learning_rate": 2.0251858913250716e-06, "loss": 3.7096, "step": 17093 }, { "epoch": 3.26, "learning_rate": 2.0251286939942804e-06, "loss": 2.165, "step": 17094 }, { "epoch": 3.26, "learning_rate": 2.025071496663489e-06, "loss": 3.4308, "step": 17095 }, { "epoch": 3.26, "learning_rate": 2.0250142993326976e-06, "loss": 2.3478, "step": 17096 }, { "epoch": 3.26, "learning_rate": 2.024957102001907e-06, "loss": 2.9802, "step": 17097 }, { "epoch": 3.26, "learning_rate": 2.0248999046711156e-06, "loss": 2.1939, "step": 17098 }, { "epoch": 3.26, "learning_rate": 2.024842707340324e-06, "loss": 2.1334, "step": 17099 }, { "epoch": 3.26, "learning_rate": 2.024785510009533e-06, "loss": 2.9702, "step": 17100 }, { "epoch": 3.26, "learning_rate": 2.0247283126787415e-06, "loss": 2.612, "step": 17101 }, { "epoch": 3.26, "learning_rate": 2.0246711153479503e-06, "loss": 2.7626, "step": 17102 }, { "epoch": 3.26, "learning_rate": 2.024613918017159e-06, "loss": 2.6108, "step": 17103 }, { "epoch": 3.26, "learning_rate": 2.024556720686368e-06, "loss": 3.6432, "step": 17104 }, { "epoch": 3.26, "learning_rate": 2.0244995233555767e-06, "loss": 2.2061, "step": 17105 }, { "epoch": 3.26, "learning_rate": 2.0244423260247855e-06, "loss": 3.209, "step": 17106 }, { "epoch": 3.26, "learning_rate": 2.0243851286939943e-06, "loss": 2.7895, "step": 17107 }, { "epoch": 3.26, "learning_rate": 2.024327931363203e-06, "loss": 2.333, "step": 17108 }, { "epoch": 3.26, "learning_rate": 2.024270734032412e-06, "loss": 2.0042, "step": 17109 }, { "epoch": 3.26, "learning_rate": 2.0242135367016207e-06, "loss": 3.1618, "step": 17110 }, { "epoch": 3.26, "learning_rate": 2.0241563393708295e-06, "loss": 2.2511, "step": 17111 }, { "epoch": 3.26, "learning_rate": 2.0240991420400383e-06, "loss": 1.5284, "step": 17112 }, { "epoch": 3.26, "learning_rate": 2.024041944709247e-06, "loss": 3.0756, "step": 17113 }, { "epoch": 3.26, "learning_rate": 2.023984747378456e-06, "loss": 2.2258, "step": 17114 }, { "epoch": 3.26, "learning_rate": 2.0239275500476647e-06, "loss": 2.0654, "step": 17115 }, { "epoch": 3.26, "learning_rate": 2.023870352716873e-06, "loss": 3.2455, "step": 17116 }, { "epoch": 3.26, "learning_rate": 2.0238131553860822e-06, "loss": 3.5752, "step": 17117 }, { "epoch": 3.26, "learning_rate": 2.0237559580552906e-06, "loss": 2.5404, "step": 17118 }, { "epoch": 3.26, "learning_rate": 2.0236987607244994e-06, "loss": 2.1155, "step": 17119 }, { "epoch": 3.26, "learning_rate": 2.0236415633937086e-06, "loss": 2.9437, "step": 17120 }, { "epoch": 3.26, "learning_rate": 2.023584366062917e-06, "loss": 1.4861, "step": 17121 }, { "epoch": 3.26, "learning_rate": 2.023527168732126e-06, "loss": 2.3084, "step": 17122 }, { "epoch": 3.26, "learning_rate": 2.0234699714013346e-06, "loss": 3.5998, "step": 17123 }, { "epoch": 3.26, "learning_rate": 2.0234127740705434e-06, "loss": 1.9231, "step": 17124 }, { "epoch": 3.26, "learning_rate": 2.023355576739752e-06, "loss": 3.6046, "step": 17125 }, { "epoch": 3.26, "learning_rate": 2.023298379408961e-06, "loss": 2.6224, "step": 17126 }, { "epoch": 3.26, "learning_rate": 2.0232411820781698e-06, "loss": 2.2852, "step": 17127 }, { "epoch": 3.26, "learning_rate": 2.0231839847473786e-06, "loss": 2.4144, "step": 17128 }, { "epoch": 3.26, "learning_rate": 2.0231267874165874e-06, "loss": 2.9119, "step": 17129 }, { "epoch": 3.26, "learning_rate": 2.023069590085796e-06, "loss": 2.5001, "step": 17130 }, { "epoch": 3.26, "learning_rate": 2.023012392755005e-06, "loss": 2.545, "step": 17131 }, { "epoch": 3.26, "learning_rate": 2.0229551954242137e-06, "loss": 2.0847, "step": 17132 }, { "epoch": 3.26, "learning_rate": 2.022897998093422e-06, "loss": 2.9528, "step": 17133 }, { "epoch": 3.26, "learning_rate": 2.0228408007626313e-06, "loss": 3.2186, "step": 17134 }, { "epoch": 3.26, "learning_rate": 2.0227836034318397e-06, "loss": 3.5242, "step": 17135 }, { "epoch": 3.26, "learning_rate": 2.0227264061010485e-06, "loss": 1.8475, "step": 17136 }, { "epoch": 3.26, "learning_rate": 2.0226692087702577e-06, "loss": 2.0145, "step": 17137 }, { "epoch": 3.26, "learning_rate": 2.022612011439466e-06, "loss": 2.9113, "step": 17138 }, { "epoch": 3.26, "learning_rate": 2.022554814108675e-06, "loss": 2.2729, "step": 17139 }, { "epoch": 3.26, "learning_rate": 2.0224976167778837e-06, "loss": 2.7673, "step": 17140 }, { "epoch": 3.26, "learning_rate": 2.0224404194470925e-06, "loss": 2.3451, "step": 17141 }, { "epoch": 3.27, "learning_rate": 2.0223832221163012e-06, "loss": 3.5047, "step": 17142 }, { "epoch": 3.27, "learning_rate": 2.02232602478551e-06, "loss": 2.2341, "step": 17143 }, { "epoch": 3.27, "learning_rate": 2.022268827454719e-06, "loss": 2.7265, "step": 17144 }, { "epoch": 3.27, "learning_rate": 2.0222116301239276e-06, "loss": 2.0601, "step": 17145 }, { "epoch": 3.27, "learning_rate": 2.0221544327931364e-06, "loss": 3.2935, "step": 17146 }, { "epoch": 3.27, "learning_rate": 2.0220972354623452e-06, "loss": 2.2345, "step": 17147 }, { "epoch": 3.27, "learning_rate": 2.022040038131554e-06, "loss": 1.8395, "step": 17148 }, { "epoch": 3.27, "learning_rate": 2.021982840800763e-06, "loss": 2.5101, "step": 17149 }, { "epoch": 3.27, "learning_rate": 2.021925643469971e-06, "loss": 2.494, "step": 17150 }, { "epoch": 3.27, "learning_rate": 2.0218684461391804e-06, "loss": 1.6187, "step": 17151 }, { "epoch": 3.27, "learning_rate": 2.0218112488083888e-06, "loss": 3.1097, "step": 17152 }, { "epoch": 3.27, "learning_rate": 2.0217540514775976e-06, "loss": 1.7521, "step": 17153 }, { "epoch": 3.27, "learning_rate": 2.0216968541468068e-06, "loss": 1.974, "step": 17154 }, { "epoch": 3.27, "learning_rate": 2.021639656816015e-06, "loss": 1.6077, "step": 17155 }, { "epoch": 3.27, "learning_rate": 2.021582459485224e-06, "loss": 2.7845, "step": 17156 }, { "epoch": 3.27, "learning_rate": 2.0215252621544327e-06, "loss": 2.1537, "step": 17157 }, { "epoch": 3.27, "learning_rate": 2.0214680648236415e-06, "loss": 2.177, "step": 17158 }, { "epoch": 3.27, "learning_rate": 2.0214108674928507e-06, "loss": 2.4766, "step": 17159 }, { "epoch": 3.27, "learning_rate": 2.021353670162059e-06, "loss": 2.4098, "step": 17160 }, { "epoch": 3.27, "learning_rate": 2.021296472831268e-06, "loss": 3.0293, "step": 17161 }, { "epoch": 3.27, "learning_rate": 2.0212392755004767e-06, "loss": 1.7177, "step": 17162 }, { "epoch": 3.27, "learning_rate": 2.0211820781696855e-06, "loss": 2.4317, "step": 17163 }, { "epoch": 3.27, "learning_rate": 2.0211248808388943e-06, "loss": 2.4727, "step": 17164 }, { "epoch": 3.27, "learning_rate": 2.021067683508103e-06, "loss": 2.4211, "step": 17165 }, { "epoch": 3.27, "learning_rate": 2.021010486177312e-06, "loss": 1.9391, "step": 17166 }, { "epoch": 3.27, "learning_rate": 2.0209532888465202e-06, "loss": 2.1269, "step": 17167 }, { "epoch": 3.27, "learning_rate": 2.0208960915157295e-06, "loss": 1.7004, "step": 17168 }, { "epoch": 3.27, "learning_rate": 2.0208388941849383e-06, "loss": 2.2619, "step": 17169 }, { "epoch": 3.27, "learning_rate": 2.0207816968541466e-06, "loss": 3.4383, "step": 17170 }, { "epoch": 3.27, "learning_rate": 2.020724499523356e-06, "loss": 2.1434, "step": 17171 }, { "epoch": 3.27, "learning_rate": 2.0206673021925642e-06, "loss": 1.8743, "step": 17172 }, { "epoch": 3.27, "learning_rate": 2.0206101048617734e-06, "loss": 2.0485, "step": 17173 }, { "epoch": 3.27, "learning_rate": 2.020552907530982e-06, "loss": 3.2415, "step": 17174 }, { "epoch": 3.27, "learning_rate": 2.0204957102001906e-06, "loss": 3.8071, "step": 17175 }, { "epoch": 3.27, "learning_rate": 2.0204385128694e-06, "loss": 2.5806, "step": 17176 }, { "epoch": 3.27, "learning_rate": 2.020381315538608e-06, "loss": 3.1986, "step": 17177 }, { "epoch": 3.27, "learning_rate": 2.020324118207817e-06, "loss": 2.359, "step": 17178 }, { "epoch": 3.27, "learning_rate": 2.0202669208770258e-06, "loss": 1.413, "step": 17179 }, { "epoch": 3.27, "learning_rate": 2.0202097235462346e-06, "loss": 1.9494, "step": 17180 }, { "epoch": 3.27, "learning_rate": 2.0201525262154434e-06, "loss": 2.7526, "step": 17181 }, { "epoch": 3.27, "learning_rate": 2.020095328884652e-06, "loss": 3.0576, "step": 17182 }, { "epoch": 3.27, "learning_rate": 2.020038131553861e-06, "loss": 2.0713, "step": 17183 }, { "epoch": 3.27, "learning_rate": 2.0199809342230693e-06, "loss": 1.9816, "step": 17184 }, { "epoch": 3.27, "learning_rate": 2.0199237368922785e-06, "loss": 1.7124, "step": 17185 }, { "epoch": 3.27, "learning_rate": 2.0198665395614873e-06, "loss": 2.8879, "step": 17186 }, { "epoch": 3.27, "learning_rate": 2.0198093422306957e-06, "loss": 2.9129, "step": 17187 }, { "epoch": 3.27, "learning_rate": 2.019752144899905e-06, "loss": 2.4565, "step": 17188 }, { "epoch": 3.27, "learning_rate": 2.0196949475691133e-06, "loss": 1.9842, "step": 17189 }, { "epoch": 3.27, "learning_rate": 2.0196377502383225e-06, "loss": 3.1407, "step": 17190 }, { "epoch": 3.27, "learning_rate": 2.019580552907531e-06, "loss": 2.1598, "step": 17191 }, { "epoch": 3.27, "learning_rate": 2.0195233555767397e-06, "loss": 1.9881, "step": 17192 }, { "epoch": 3.27, "learning_rate": 2.019466158245949e-06, "loss": 3.289, "step": 17193 }, { "epoch": 3.28, "learning_rate": 2.0194089609151573e-06, "loss": 1.619, "step": 17194 }, { "epoch": 3.28, "learning_rate": 2.019351763584366e-06, "loss": 2.2158, "step": 17195 }, { "epoch": 3.28, "learning_rate": 2.019294566253575e-06, "loss": 2.4132, "step": 17196 }, { "epoch": 3.28, "learning_rate": 2.0192373689227836e-06, "loss": 2.1675, "step": 17197 }, { "epoch": 3.28, "learning_rate": 2.0191801715919924e-06, "loss": 2.141, "step": 17198 }, { "epoch": 3.28, "learning_rate": 2.0191229742612012e-06, "loss": 2.9615, "step": 17199 }, { "epoch": 3.28, "learning_rate": 2.01906577693041e-06, "loss": 2.9888, "step": 17200 }, { "epoch": 3.28, "learning_rate": 2.0190085795996184e-06, "loss": 1.8885, "step": 17201 }, { "epoch": 3.28, "learning_rate": 2.0189513822688276e-06, "loss": 2.2376, "step": 17202 }, { "epoch": 3.28, "learning_rate": 2.0188941849380364e-06, "loss": 2.2438, "step": 17203 }, { "epoch": 3.28, "learning_rate": 2.0188369876072448e-06, "loss": 2.2432, "step": 17204 }, { "epoch": 3.28, "learning_rate": 2.018779790276454e-06, "loss": 1.5748, "step": 17205 }, { "epoch": 3.28, "learning_rate": 2.0187225929456624e-06, "loss": 1.4339, "step": 17206 }, { "epoch": 3.28, "learning_rate": 2.0186653956148716e-06, "loss": 1.8112, "step": 17207 }, { "epoch": 3.28, "learning_rate": 2.0186081982840804e-06, "loss": 1.9278, "step": 17208 }, { "epoch": 3.28, "learning_rate": 2.0185510009532887e-06, "loss": 3.1681, "step": 17209 }, { "epoch": 3.28, "learning_rate": 2.018493803622498e-06, "loss": 3.0524, "step": 17210 }, { "epoch": 3.28, "learning_rate": 2.0184366062917063e-06, "loss": 1.8222, "step": 17211 }, { "epoch": 3.28, "learning_rate": 2.018379408960915e-06, "loss": 2.0838, "step": 17212 }, { "epoch": 3.28, "learning_rate": 2.018322211630124e-06, "loss": 1.2619, "step": 17213 }, { "epoch": 3.28, "learning_rate": 2.0182650142993327e-06, "loss": 2.4399, "step": 17214 }, { "epoch": 3.28, "learning_rate": 2.0182078169685415e-06, "loss": 2.7971, "step": 17215 }, { "epoch": 3.28, "learning_rate": 2.0181506196377503e-06, "loss": 2.6259, "step": 17216 }, { "epoch": 3.28, "learning_rate": 2.018093422306959e-06, "loss": 3.1771, "step": 17217 }, { "epoch": 3.28, "learning_rate": 2.018036224976168e-06, "loss": 2.3369, "step": 17218 }, { "epoch": 3.28, "learning_rate": 2.0179790276453767e-06, "loss": 2.3253, "step": 17219 }, { "epoch": 3.28, "learning_rate": 2.0179218303145855e-06, "loss": 2.1135, "step": 17220 }, { "epoch": 3.28, "learning_rate": 2.017864632983794e-06, "loss": 1.6254, "step": 17221 }, { "epoch": 3.28, "learning_rate": 2.017807435653003e-06, "loss": 2.2321, "step": 17222 }, { "epoch": 3.28, "learning_rate": 2.0177502383222114e-06, "loss": 3.4968, "step": 17223 }, { "epoch": 3.28, "learning_rate": 2.0176930409914206e-06, "loss": 3.4895, "step": 17224 }, { "epoch": 3.28, "learning_rate": 2.0176358436606294e-06, "loss": 1.9924, "step": 17225 }, { "epoch": 3.28, "learning_rate": 2.017578646329838e-06, "loss": 2.2799, "step": 17226 }, { "epoch": 3.28, "learning_rate": 2.017521448999047e-06, "loss": 2.7573, "step": 17227 }, { "epoch": 3.28, "learning_rate": 2.0174642516682554e-06, "loss": 2.6818, "step": 17228 }, { "epoch": 3.28, "learning_rate": 2.017407054337464e-06, "loss": 2.5971, "step": 17229 }, { "epoch": 3.28, "learning_rate": 2.017349857006673e-06, "loss": 3.1486, "step": 17230 }, { "epoch": 3.28, "learning_rate": 2.0172926596758818e-06, "loss": 2.5257, "step": 17231 }, { "epoch": 3.28, "learning_rate": 2.0172354623450906e-06, "loss": 1.8186, "step": 17232 }, { "epoch": 3.28, "learning_rate": 2.0171782650142994e-06, "loss": 1.6889, "step": 17233 }, { "epoch": 3.28, "learning_rate": 2.017121067683508e-06, "loss": 2.0864, "step": 17234 }, { "epoch": 3.28, "learning_rate": 2.017063870352717e-06, "loss": 3.1792, "step": 17235 }, { "epoch": 3.28, "learning_rate": 2.0170066730219257e-06, "loss": 2.2968, "step": 17236 }, { "epoch": 3.28, "learning_rate": 2.0169494756911345e-06, "loss": 2.6207, "step": 17237 }, { "epoch": 3.28, "learning_rate": 2.0168922783603433e-06, "loss": 3.2201, "step": 17238 }, { "epoch": 3.28, "learning_rate": 2.016835081029552e-06, "loss": 1.5773, "step": 17239 }, { "epoch": 3.28, "learning_rate": 2.0167778836987605e-06, "loss": 1.7993, "step": 17240 }, { "epoch": 3.28, "learning_rate": 2.0167206863679697e-06, "loss": 2.6205, "step": 17241 }, { "epoch": 3.28, "learning_rate": 2.0166634890371785e-06, "loss": 1.2647, "step": 17242 }, { "epoch": 3.28, "learning_rate": 2.016606291706387e-06, "loss": 3.7505, "step": 17243 }, { "epoch": 3.28, "learning_rate": 2.016549094375596e-06, "loss": 2.4428, "step": 17244 }, { "epoch": 3.28, "learning_rate": 2.0164918970448045e-06, "loss": 1.4902, "step": 17245 }, { "epoch": 3.28, "learning_rate": 2.0164346997140133e-06, "loss": 2.27, "step": 17246 }, { "epoch": 3.29, "learning_rate": 2.0163775023832225e-06, "loss": 2.7114, "step": 17247 }, { "epoch": 3.29, "learning_rate": 2.016320305052431e-06, "loss": 3.1605, "step": 17248 }, { "epoch": 3.29, "learning_rate": 2.0162631077216396e-06, "loss": 2.2673, "step": 17249 }, { "epoch": 3.29, "learning_rate": 2.0162059103908484e-06, "loss": 2.0372, "step": 17250 }, { "epoch": 3.29, "learning_rate": 2.0161487130600572e-06, "loss": 1.6202, "step": 17251 }, { "epoch": 3.29, "learning_rate": 2.016091515729266e-06, "loss": 2.17, "step": 17252 }, { "epoch": 3.29, "learning_rate": 2.016034318398475e-06, "loss": 3.256, "step": 17253 }, { "epoch": 3.29, "learning_rate": 2.0159771210676836e-06, "loss": 2.5489, "step": 17254 }, { "epoch": 3.29, "learning_rate": 2.0159199237368924e-06, "loss": 2.1013, "step": 17255 }, { "epoch": 3.29, "learning_rate": 2.015862726406101e-06, "loss": 2.9051, "step": 17256 }, { "epoch": 3.29, "learning_rate": 2.01580552907531e-06, "loss": 1.4709, "step": 17257 }, { "epoch": 3.29, "learning_rate": 2.0157483317445188e-06, "loss": 2.2356, "step": 17258 }, { "epoch": 3.29, "learning_rate": 2.0156911344137276e-06, "loss": 2.911, "step": 17259 }, { "epoch": 3.29, "learning_rate": 2.015633937082936e-06, "loss": 1.3917, "step": 17260 }, { "epoch": 3.29, "learning_rate": 2.015576739752145e-06, "loss": 2.4986, "step": 17261 }, { "epoch": 3.29, "learning_rate": 2.0155195424213535e-06, "loss": 3.5655, "step": 17262 }, { "epoch": 3.29, "learning_rate": 2.0154623450905623e-06, "loss": 1.6638, "step": 17263 }, { "epoch": 3.29, "learning_rate": 2.0154051477597716e-06, "loss": 3.3761, "step": 17264 }, { "epoch": 3.29, "learning_rate": 2.01534795042898e-06, "loss": 1.4723, "step": 17265 }, { "epoch": 3.29, "learning_rate": 2.0152907530981887e-06, "loss": 1.8956, "step": 17266 }, { "epoch": 3.29, "learning_rate": 2.0152335557673975e-06, "loss": 1.865, "step": 17267 }, { "epoch": 3.29, "learning_rate": 2.0151763584366063e-06, "loss": 1.9743, "step": 17268 }, { "epoch": 3.29, "learning_rate": 2.015119161105815e-06, "loss": 2.015, "step": 17269 }, { "epoch": 3.29, "learning_rate": 2.015061963775024e-06, "loss": 3.5016, "step": 17270 }, { "epoch": 3.29, "learning_rate": 2.0150047664442327e-06, "loss": 2.6179, "step": 17271 }, { "epoch": 3.29, "learning_rate": 2.0149475691134415e-06, "loss": 2.4687, "step": 17272 }, { "epoch": 3.29, "learning_rate": 2.0148903717826503e-06, "loss": 2.5399, "step": 17273 }, { "epoch": 3.29, "learning_rate": 2.014833174451859e-06, "loss": 2.3679, "step": 17274 }, { "epoch": 3.29, "learning_rate": 2.014775977121068e-06, "loss": 1.6376, "step": 17275 }, { "epoch": 3.29, "learning_rate": 2.0147187797902767e-06, "loss": 2.5776, "step": 17276 }, { "epoch": 3.29, "learning_rate": 2.014661582459485e-06, "loss": 2.1257, "step": 17277 }, { "epoch": 3.29, "learning_rate": 2.0146043851286942e-06, "loss": 3.4039, "step": 17278 }, { "epoch": 3.29, "learning_rate": 2.0145471877979026e-06, "loss": 2.9761, "step": 17279 }, { "epoch": 3.29, "learning_rate": 2.0144899904671114e-06, "loss": 2.4493, "step": 17280 }, { "epoch": 3.29, "learning_rate": 2.0144327931363206e-06, "loss": 1.9958, "step": 17281 }, { "epoch": 3.29, "learning_rate": 2.014375595805529e-06, "loss": 1.6893, "step": 17282 }, { "epoch": 3.29, "learning_rate": 2.0143183984747378e-06, "loss": 2.3257, "step": 17283 }, { "epoch": 3.29, "learning_rate": 2.0142612011439466e-06, "loss": 2.1329, "step": 17284 }, { "epoch": 3.29, "learning_rate": 2.0142040038131554e-06, "loss": 1.9219, "step": 17285 }, { "epoch": 3.29, "learning_rate": 2.014146806482364e-06, "loss": 2.5605, "step": 17286 }, { "epoch": 3.29, "learning_rate": 2.014089609151573e-06, "loss": 3.3484, "step": 17287 }, { "epoch": 3.29, "learning_rate": 2.0140324118207818e-06, "loss": 3.2317, "step": 17288 }, { "epoch": 3.29, "learning_rate": 2.0139752144899905e-06, "loss": 2.179, "step": 17289 }, { "epoch": 3.29, "learning_rate": 2.0139180171591993e-06, "loss": 3.1561, "step": 17290 }, { "epoch": 3.29, "learning_rate": 2.013860819828408e-06, "loss": 3.0618, "step": 17291 }, { "epoch": 3.29, "learning_rate": 2.013803622497617e-06, "loss": 2.5148, "step": 17292 }, { "epoch": 3.29, "learning_rate": 2.0137464251668257e-06, "loss": 2.2647, "step": 17293 }, { "epoch": 3.29, "learning_rate": 2.013689227836034e-06, "loss": 2.6279, "step": 17294 }, { "epoch": 3.29, "learning_rate": 2.0136320305052433e-06, "loss": 1.6084, "step": 17295 }, { "epoch": 3.29, "learning_rate": 2.013574833174452e-06, "loss": 2.5934, "step": 17296 }, { "epoch": 3.29, "learning_rate": 2.0135176358436605e-06, "loss": 3.1451, "step": 17297 }, { "epoch": 3.29, "learning_rate": 2.0134604385128697e-06, "loss": 2.7386, "step": 17298 }, { "epoch": 3.3, "learning_rate": 2.013403241182078e-06, "loss": 2.5214, "step": 17299 }, { "epoch": 3.3, "learning_rate": 2.013346043851287e-06, "loss": 2.6956, "step": 17300 }, { "epoch": 3.3, "learning_rate": 2.0132888465204957e-06, "loss": 2.4723, "step": 17301 }, { "epoch": 3.3, "learning_rate": 2.0132316491897044e-06, "loss": 2.625, "step": 17302 }, { "epoch": 3.3, "learning_rate": 2.0131744518589137e-06, "loss": 3.9897, "step": 17303 }, { "epoch": 3.3, "learning_rate": 2.013117254528122e-06, "loss": 1.8617, "step": 17304 }, { "epoch": 3.3, "learning_rate": 2.013060057197331e-06, "loss": 2.1865, "step": 17305 }, { "epoch": 3.3, "learning_rate": 2.0130028598665396e-06, "loss": 2.4545, "step": 17306 }, { "epoch": 3.3, "learning_rate": 2.0129456625357484e-06, "loss": 2.9077, "step": 17307 }, { "epoch": 3.3, "learning_rate": 2.012888465204957e-06, "loss": 2.29, "step": 17308 }, { "epoch": 3.3, "learning_rate": 2.012831267874166e-06, "loss": 3.3384, "step": 17309 }, { "epoch": 3.3, "learning_rate": 2.012774070543375e-06, "loss": 2.2446, "step": 17310 }, { "epoch": 3.3, "learning_rate": 2.012716873212583e-06, "loss": 2.8338, "step": 17311 }, { "epoch": 3.3, "learning_rate": 2.0126596758817924e-06, "loss": 1.3748, "step": 17312 }, { "epoch": 3.3, "learning_rate": 2.012602478551001e-06, "loss": 3.4457, "step": 17313 }, { "epoch": 3.3, "learning_rate": 2.0125452812202095e-06, "loss": 2.4218, "step": 17314 }, { "epoch": 3.3, "learning_rate": 2.0124880838894188e-06, "loss": 2.787, "step": 17315 }, { "epoch": 3.3, "learning_rate": 2.012430886558627e-06, "loss": 3.4983, "step": 17316 }, { "epoch": 3.3, "learning_rate": 2.012373689227836e-06, "loss": 1.8073, "step": 17317 }, { "epoch": 3.3, "learning_rate": 2.0123164918970447e-06, "loss": 2.354, "step": 17318 }, { "epoch": 3.3, "learning_rate": 2.0122592945662535e-06, "loss": 2.583, "step": 17319 }, { "epoch": 3.3, "learning_rate": 2.0122020972354627e-06, "loss": 3.004, "step": 17320 }, { "epoch": 3.3, "learning_rate": 2.012144899904671e-06, "loss": 2.8493, "step": 17321 }, { "epoch": 3.3, "learning_rate": 2.01208770257388e-06, "loss": 2.6251, "step": 17322 }, { "epoch": 3.3, "learning_rate": 2.0120305052430887e-06, "loss": 2.2308, "step": 17323 }, { "epoch": 3.3, "learning_rate": 2.0119733079122975e-06, "loss": 2.5143, "step": 17324 }, { "epoch": 3.3, "learning_rate": 2.0119161105815063e-06, "loss": 1.7064, "step": 17325 }, { "epoch": 3.3, "learning_rate": 2.011858913250715e-06, "loss": 3.7198, "step": 17326 }, { "epoch": 3.3, "learning_rate": 2.011801715919924e-06, "loss": 2.1803, "step": 17327 }, { "epoch": 3.3, "learning_rate": 2.0117445185891322e-06, "loss": 2.351, "step": 17328 }, { "epoch": 3.3, "learning_rate": 2.0116873212583415e-06, "loss": 2.0303, "step": 17329 }, { "epoch": 3.3, "learning_rate": 2.0116301239275502e-06, "loss": 1.617, "step": 17330 }, { "epoch": 3.3, "learning_rate": 2.0115729265967586e-06, "loss": 2.9622, "step": 17331 }, { "epoch": 3.3, "learning_rate": 2.011515729265968e-06, "loss": 1.9756, "step": 17332 }, { "epoch": 3.3, "learning_rate": 2.011458531935176e-06, "loss": 1.5708, "step": 17333 }, { "epoch": 3.3, "learning_rate": 2.011401334604385e-06, "loss": 3.0352, "step": 17334 }, { "epoch": 3.3, "learning_rate": 2.0113441372735942e-06, "loss": 2.2271, "step": 17335 }, { "epoch": 3.3, "learning_rate": 2.0112869399428026e-06, "loss": 2.2746, "step": 17336 }, { "epoch": 3.3, "learning_rate": 2.011229742612012e-06, "loss": 1.1939, "step": 17337 }, { "epoch": 3.3, "learning_rate": 2.01117254528122e-06, "loss": 2.2565, "step": 17338 }, { "epoch": 3.3, "learning_rate": 2.011115347950429e-06, "loss": 1.3984, "step": 17339 }, { "epoch": 3.3, "learning_rate": 2.0110581506196378e-06, "loss": 1.7922, "step": 17340 }, { "epoch": 3.3, "learning_rate": 2.0110009532888466e-06, "loss": 2.2927, "step": 17341 }, { "epoch": 3.3, "learning_rate": 2.0109437559580554e-06, "loss": 3.2045, "step": 17342 }, { "epoch": 3.3, "learning_rate": 2.010886558627264e-06, "loss": 1.3176, "step": 17343 }, { "epoch": 3.3, "learning_rate": 2.010829361296473e-06, "loss": 2.5446, "step": 17344 }, { "epoch": 3.3, "learning_rate": 2.0107721639656817e-06, "loss": 3.1659, "step": 17345 }, { "epoch": 3.3, "learning_rate": 2.0107149666348905e-06, "loss": 2.1468, "step": 17346 }, { "epoch": 3.3, "learning_rate": 2.0106577693040993e-06, "loss": 1.9903, "step": 17347 }, { "epoch": 3.3, "learning_rate": 2.0106005719733077e-06, "loss": 2.2605, "step": 17348 }, { "epoch": 3.3, "learning_rate": 2.010543374642517e-06, "loss": 1.8796, "step": 17349 }, { "epoch": 3.3, "learning_rate": 2.0104861773117253e-06, "loss": 2.5984, "step": 17350 }, { "epoch": 3.3, "learning_rate": 2.010428979980934e-06, "loss": 1.7808, "step": 17351 }, { "epoch": 3.31, "learning_rate": 2.0103717826501433e-06, "loss": 2.9111, "step": 17352 }, { "epoch": 3.31, "learning_rate": 2.0103145853193517e-06, "loss": 2.3747, "step": 17353 }, { "epoch": 3.31, "learning_rate": 2.010257387988561e-06, "loss": 3.0317, "step": 17354 }, { "epoch": 3.31, "learning_rate": 2.0102001906577692e-06, "loss": 2.3694, "step": 17355 }, { "epoch": 3.31, "learning_rate": 2.010142993326978e-06, "loss": 1.801, "step": 17356 }, { "epoch": 3.31, "learning_rate": 2.010085795996187e-06, "loss": 3.3378, "step": 17357 }, { "epoch": 3.31, "learning_rate": 2.0100285986653956e-06, "loss": 1.437, "step": 17358 }, { "epoch": 3.31, "learning_rate": 2.0099714013346044e-06, "loss": 3.4478, "step": 17359 }, { "epoch": 3.31, "learning_rate": 2.0099142040038132e-06, "loss": 1.4151, "step": 17360 }, { "epoch": 3.31, "learning_rate": 2.009857006673022e-06, "loss": 3.0388, "step": 17361 }, { "epoch": 3.31, "learning_rate": 2.009799809342231e-06, "loss": 2.8249, "step": 17362 }, { "epoch": 3.31, "learning_rate": 2.0097426120114396e-06, "loss": 2.4801, "step": 17363 }, { "epoch": 3.31, "learning_rate": 2.0096854146806484e-06, "loss": 2.7022, "step": 17364 }, { "epoch": 3.31, "learning_rate": 2.0096282173498568e-06, "loss": 2.0869, "step": 17365 }, { "epoch": 3.31, "learning_rate": 2.009571020019066e-06, "loss": 1.3454, "step": 17366 }, { "epoch": 3.31, "learning_rate": 2.0095138226882743e-06, "loss": 2.6135, "step": 17367 }, { "epoch": 3.31, "learning_rate": 2.0094566253574836e-06, "loss": 3.399, "step": 17368 }, { "epoch": 3.31, "learning_rate": 2.0093994280266924e-06, "loss": 3.0424, "step": 17369 }, { "epoch": 3.31, "learning_rate": 2.0093422306959007e-06, "loss": 2.3842, "step": 17370 }, { "epoch": 3.31, "learning_rate": 2.00928503336511e-06, "loss": 2.3254, "step": 17371 }, { "epoch": 3.31, "learning_rate": 2.0092278360343183e-06, "loss": 2.5914, "step": 17372 }, { "epoch": 3.31, "learning_rate": 2.009170638703527e-06, "loss": 3.1481, "step": 17373 }, { "epoch": 3.31, "learning_rate": 2.009113441372736e-06, "loss": 2.5414, "step": 17374 }, { "epoch": 3.31, "learning_rate": 2.0090562440419447e-06, "loss": 2.835, "step": 17375 }, { "epoch": 3.31, "learning_rate": 2.0089990467111535e-06, "loss": 2.9269, "step": 17376 }, { "epoch": 3.31, "learning_rate": 2.0089418493803623e-06, "loss": 2.4942, "step": 17377 }, { "epoch": 3.31, "learning_rate": 2.008884652049571e-06, "loss": 2.0748, "step": 17378 }, { "epoch": 3.31, "learning_rate": 2.00882745471878e-06, "loss": 2.9788, "step": 17379 }, { "epoch": 3.31, "learning_rate": 2.0087702573879887e-06, "loss": 2.9239, "step": 17380 }, { "epoch": 3.31, "learning_rate": 2.0087130600571975e-06, "loss": 1.2852, "step": 17381 }, { "epoch": 3.31, "learning_rate": 2.008655862726406e-06, "loss": 1.2548, "step": 17382 }, { "epoch": 3.31, "learning_rate": 2.008598665395615e-06, "loss": 2.1549, "step": 17383 }, { "epoch": 3.31, "learning_rate": 2.008541468064824e-06, "loss": 1.9968, "step": 17384 }, { "epoch": 3.31, "learning_rate": 2.0084842707340326e-06, "loss": 2.6377, "step": 17385 }, { "epoch": 3.31, "learning_rate": 2.0084270734032414e-06, "loss": 2.3599, "step": 17386 }, { "epoch": 3.31, "learning_rate": 2.00836987607245e-06, "loss": 2.5988, "step": 17387 }, { "epoch": 3.31, "learning_rate": 2.008312678741659e-06, "loss": 1.7147, "step": 17388 }, { "epoch": 3.31, "learning_rate": 2.0082554814108674e-06, "loss": 2.0152, "step": 17389 }, { "epoch": 3.31, "learning_rate": 2.008198284080076e-06, "loss": 3.1566, "step": 17390 }, { "epoch": 3.31, "learning_rate": 2.0081410867492854e-06, "loss": 1.6956, "step": 17391 }, { "epoch": 3.31, "learning_rate": 2.0080838894184938e-06, "loss": 2.8651, "step": 17392 }, { "epoch": 3.31, "learning_rate": 2.0080266920877026e-06, "loss": 2.1836, "step": 17393 }, { "epoch": 3.31, "learning_rate": 2.0079694947569114e-06, "loss": 3.7052, "step": 17394 }, { "epoch": 3.31, "learning_rate": 2.00791229742612e-06, "loss": 1.8433, "step": 17395 }, { "epoch": 3.31, "learning_rate": 2.007855100095329e-06, "loss": 2.4288, "step": 17396 }, { "epoch": 3.31, "learning_rate": 2.0077979027645377e-06, "loss": 2.1827, "step": 17397 }, { "epoch": 3.31, "learning_rate": 2.0077407054337465e-06, "loss": 2.8692, "step": 17398 }, { "epoch": 3.31, "learning_rate": 2.007683508102955e-06, "loss": 2.2976, "step": 17399 }, { "epoch": 3.31, "learning_rate": 2.007626310772164e-06, "loss": 1.8628, "step": 17400 }, { "epoch": 3.31, "learning_rate": 2.007569113441373e-06, "loss": 2.6044, "step": 17401 }, { "epoch": 3.31, "learning_rate": 2.0075119161105817e-06, "loss": 1.8668, "step": 17402 }, { "epoch": 3.31, "learning_rate": 2.0074547187797905e-06, "loss": 2.3501, "step": 17403 }, { "epoch": 3.32, "learning_rate": 2.007397521448999e-06, "loss": 2.6584, "step": 17404 }, { "epoch": 3.32, "learning_rate": 2.007340324118208e-06, "loss": 2.1616, "step": 17405 }, { "epoch": 3.32, "learning_rate": 2.0072831267874165e-06, "loss": 3.2954, "step": 17406 }, { "epoch": 3.32, "learning_rate": 2.0072259294566253e-06, "loss": 3.543, "step": 17407 }, { "epoch": 3.32, "learning_rate": 2.0071687321258345e-06, "loss": 2.8386, "step": 17408 }, { "epoch": 3.32, "learning_rate": 2.007111534795043e-06, "loss": 3.4695, "step": 17409 }, { "epoch": 3.32, "learning_rate": 2.0070543374642516e-06, "loss": 3.8912, "step": 17410 }, { "epoch": 3.32, "learning_rate": 2.0069971401334604e-06, "loss": 1.6823, "step": 17411 }, { "epoch": 3.32, "learning_rate": 2.0069399428026692e-06, "loss": 2.0451, "step": 17412 }, { "epoch": 3.32, "learning_rate": 2.006882745471878e-06, "loss": 3.6594, "step": 17413 }, { "epoch": 3.32, "learning_rate": 2.006825548141087e-06, "loss": 2.3382, "step": 17414 }, { "epoch": 3.32, "learning_rate": 2.0067683508102956e-06, "loss": 3.9874, "step": 17415 }, { "epoch": 3.32, "learning_rate": 2.006711153479504e-06, "loss": 2.806, "step": 17416 }, { "epoch": 3.32, "learning_rate": 2.006653956148713e-06, "loss": 1.599, "step": 17417 }, { "epoch": 3.32, "learning_rate": 2.006596758817922e-06, "loss": 2.1082, "step": 17418 }, { "epoch": 3.32, "learning_rate": 2.0065395614871308e-06, "loss": 2.1424, "step": 17419 }, { "epoch": 3.32, "learning_rate": 2.0064823641563396e-06, "loss": 2.3226, "step": 17420 }, { "epoch": 3.32, "learning_rate": 2.006425166825548e-06, "loss": 2.6407, "step": 17421 }, { "epoch": 3.32, "learning_rate": 2.006367969494757e-06, "loss": 2.8771, "step": 17422 }, { "epoch": 3.32, "learning_rate": 2.006310772163966e-06, "loss": 3.3966, "step": 17423 }, { "epoch": 3.32, "learning_rate": 2.0062535748331743e-06, "loss": 2.6572, "step": 17424 }, { "epoch": 3.32, "learning_rate": 2.0061963775023835e-06, "loss": 2.7904, "step": 17425 }, { "epoch": 3.32, "learning_rate": 2.006139180171592e-06, "loss": 3.4533, "step": 17426 }, { "epoch": 3.32, "learning_rate": 2.0060819828408007e-06, "loss": 2.2033, "step": 17427 }, { "epoch": 3.32, "learning_rate": 2.0060247855100095e-06, "loss": 2.721, "step": 17428 }, { "epoch": 3.32, "learning_rate": 2.0059675881792183e-06, "loss": 2.5161, "step": 17429 }, { "epoch": 3.32, "learning_rate": 2.005910390848427e-06, "loss": 1.6823, "step": 17430 }, { "epoch": 3.32, "learning_rate": 2.005853193517636e-06, "loss": 2.1171, "step": 17431 }, { "epoch": 3.32, "learning_rate": 2.0057959961868447e-06, "loss": 2.4512, "step": 17432 }, { "epoch": 3.32, "learning_rate": 2.0057387988560535e-06, "loss": 2.2752, "step": 17433 }, { "epoch": 3.32, "learning_rate": 2.0056816015252623e-06, "loss": 2.8941, "step": 17434 }, { "epoch": 3.32, "learning_rate": 2.005624404194471e-06, "loss": 2.2374, "step": 17435 }, { "epoch": 3.32, "learning_rate": 2.00556720686368e-06, "loss": 2.6983, "step": 17436 }, { "epoch": 3.32, "learning_rate": 2.0055100095328886e-06, "loss": 1.9859, "step": 17437 }, { "epoch": 3.32, "learning_rate": 2.005452812202097e-06, "loss": 2.502, "step": 17438 }, { "epoch": 3.32, "learning_rate": 2.0053956148713062e-06, "loss": 2.7539, "step": 17439 }, { "epoch": 3.32, "learning_rate": 2.005338417540515e-06, "loss": 3.132, "step": 17440 }, { "epoch": 3.32, "learning_rate": 2.0052812202097234e-06, "loss": 2.2946, "step": 17441 }, { "epoch": 3.32, "learning_rate": 2.0052240228789326e-06, "loss": 2.7475, "step": 17442 }, { "epoch": 3.32, "learning_rate": 2.005166825548141e-06, "loss": 1.7278, "step": 17443 }, { "epoch": 3.32, "learning_rate": 2.0051096282173498e-06, "loss": 3.6909, "step": 17444 }, { "epoch": 3.32, "learning_rate": 2.0050524308865586e-06, "loss": 3.1281, "step": 17445 }, { "epoch": 3.32, "learning_rate": 2.0049952335557674e-06, "loss": 2.9598, "step": 17446 }, { "epoch": 3.32, "learning_rate": 2.004938036224976e-06, "loss": 1.5806, "step": 17447 }, { "epoch": 3.32, "learning_rate": 2.004880838894185e-06, "loss": 1.7916, "step": 17448 }, { "epoch": 3.32, "learning_rate": 2.0048236415633937e-06, "loss": 2.1556, "step": 17449 }, { "epoch": 3.32, "learning_rate": 2.0047664442326025e-06, "loss": 1.6028, "step": 17450 }, { "epoch": 3.32, "learning_rate": 2.0047092469018113e-06, "loss": 2.2564, "step": 17451 }, { "epoch": 3.32, "learning_rate": 2.00465204957102e-06, "loss": 2.8696, "step": 17452 }, { "epoch": 3.32, "learning_rate": 2.004594852240229e-06, "loss": 3.3595, "step": 17453 }, { "epoch": 3.32, "learning_rate": 2.0045376549094377e-06, "loss": 2.5373, "step": 17454 }, { "epoch": 3.32, "learning_rate": 2.004480457578646e-06, "loss": 1.573, "step": 17455 }, { "epoch": 3.32, "learning_rate": 2.0044232602478553e-06, "loss": 2.8915, "step": 17456 }, { "epoch": 3.33, "learning_rate": 2.004366062917064e-06, "loss": 2.5264, "step": 17457 }, { "epoch": 3.33, "learning_rate": 2.0043088655862725e-06, "loss": 2.6075, "step": 17458 }, { "epoch": 3.33, "learning_rate": 2.0042516682554817e-06, "loss": 2.4093, "step": 17459 }, { "epoch": 3.33, "learning_rate": 2.00419447092469e-06, "loss": 1.345, "step": 17460 }, { "epoch": 3.33, "learning_rate": 2.004137273593899e-06, "loss": 2.7793, "step": 17461 }, { "epoch": 3.33, "learning_rate": 2.0040800762631076e-06, "loss": 2.2361, "step": 17462 }, { "epoch": 3.33, "learning_rate": 2.0040228789323164e-06, "loss": 3.2378, "step": 17463 }, { "epoch": 3.33, "learning_rate": 2.0039656816015252e-06, "loss": 2.6072, "step": 17464 }, { "epoch": 3.33, "learning_rate": 2.003908484270734e-06, "loss": 3.2591, "step": 17465 }, { "epoch": 3.33, "learning_rate": 2.003851286939943e-06, "loss": 1.9564, "step": 17466 }, { "epoch": 3.33, "learning_rate": 2.0037940896091516e-06, "loss": 1.9012, "step": 17467 }, { "epoch": 3.33, "learning_rate": 2.0037368922783604e-06, "loss": 2.6099, "step": 17468 }, { "epoch": 3.33, "learning_rate": 2.003679694947569e-06, "loss": 1.4581, "step": 17469 }, { "epoch": 3.33, "learning_rate": 2.003622497616778e-06, "loss": 1.5156, "step": 17470 }, { "epoch": 3.33, "learning_rate": 2.003565300285987e-06, "loss": 2.3393, "step": 17471 }, { "epoch": 3.33, "learning_rate": 2.0035081029551956e-06, "loss": 2.3604, "step": 17472 }, { "epoch": 3.33, "learning_rate": 2.0034509056244044e-06, "loss": 2.3535, "step": 17473 }, { "epoch": 3.33, "learning_rate": 2.003393708293613e-06, "loss": 2.2484, "step": 17474 }, { "epoch": 3.33, "learning_rate": 2.0033365109628215e-06, "loss": 2.6321, "step": 17475 }, { "epoch": 3.33, "learning_rate": 2.0032793136320308e-06, "loss": 2.7344, "step": 17476 }, { "epoch": 3.33, "learning_rate": 2.003222116301239e-06, "loss": 2.6762, "step": 17477 }, { "epoch": 3.33, "learning_rate": 2.003164918970448e-06, "loss": 2.2835, "step": 17478 }, { "epoch": 3.33, "learning_rate": 2.003107721639657e-06, "loss": 1.7934, "step": 17479 }, { "epoch": 3.33, "learning_rate": 2.0030505243088655e-06, "loss": 2.425, "step": 17480 }, { "epoch": 3.33, "learning_rate": 2.0029933269780743e-06, "loss": 2.3046, "step": 17481 }, { "epoch": 3.33, "learning_rate": 2.002936129647283e-06, "loss": 3.6303, "step": 17482 }, { "epoch": 3.33, "learning_rate": 2.002878932316492e-06, "loss": 1.7012, "step": 17483 }, { "epoch": 3.33, "learning_rate": 2.0028217349857007e-06, "loss": 2.0985, "step": 17484 }, { "epoch": 3.33, "learning_rate": 2.0027645376549095e-06, "loss": 1.8039, "step": 17485 }, { "epoch": 3.33, "learning_rate": 2.0027073403241183e-06, "loss": 2.0403, "step": 17486 }, { "epoch": 3.33, "learning_rate": 2.002650142993327e-06, "loss": 2.287, "step": 17487 }, { "epoch": 3.33, "learning_rate": 2.002592945662536e-06, "loss": 2.4686, "step": 17488 }, { "epoch": 3.33, "learning_rate": 2.0025357483317447e-06, "loss": 2.3032, "step": 17489 }, { "epoch": 3.33, "learning_rate": 2.0024785510009534e-06, "loss": 1.9213, "step": 17490 }, { "epoch": 3.33, "learning_rate": 2.0024213536701622e-06, "loss": 1.962, "step": 17491 }, { "epoch": 3.33, "learning_rate": 2.0023641563393706e-06, "loss": 2.9729, "step": 17492 }, { "epoch": 3.33, "learning_rate": 2.00230695900858e-06, "loss": 2.7073, "step": 17493 }, { "epoch": 3.33, "learning_rate": 2.002249761677788e-06, "loss": 2.7576, "step": 17494 }, { "epoch": 3.33, "learning_rate": 2.002192564346997e-06, "loss": 2.2444, "step": 17495 }, { "epoch": 3.33, "learning_rate": 2.002135367016206e-06, "loss": 3.2201, "step": 17496 }, { "epoch": 3.33, "learning_rate": 2.0020781696854146e-06, "loss": 2.9231, "step": 17497 }, { "epoch": 3.33, "learning_rate": 2.0020209723546234e-06, "loss": 2.8185, "step": 17498 }, { "epoch": 3.33, "learning_rate": 2.001963775023832e-06, "loss": 3.4395, "step": 17499 }, { "epoch": 3.33, "learning_rate": 2.001906577693041e-06, "loss": 2.0221, "step": 17500 }, { "epoch": 3.33, "learning_rate": 2.0018493803622498e-06, "loss": 2.3623, "step": 17501 }, { "epoch": 3.33, "learning_rate": 2.0017921830314586e-06, "loss": 2.7761, "step": 17502 }, { "epoch": 3.33, "learning_rate": 2.0017349857006673e-06, "loss": 2.1761, "step": 17503 }, { "epoch": 3.33, "learning_rate": 2.001677788369876e-06, "loss": 2.106, "step": 17504 }, { "epoch": 3.33, "learning_rate": 2.001620591039085e-06, "loss": 2.695, "step": 17505 }, { "epoch": 3.33, "learning_rate": 2.0015633937082937e-06, "loss": 2.9918, "step": 17506 }, { "epoch": 3.33, "learning_rate": 2.0015061963775025e-06, "loss": 2.3946, "step": 17507 }, { "epoch": 3.33, "learning_rate": 2.0014489990467113e-06, "loss": 1.9312, "step": 17508 }, { "epoch": 3.34, "learning_rate": 2.0013918017159197e-06, "loss": 2.6249, "step": 17509 }, { "epoch": 3.34, "learning_rate": 2.001334604385129e-06, "loss": 2.6092, "step": 17510 }, { "epoch": 3.34, "learning_rate": 2.0012774070543373e-06, "loss": 2.3538, "step": 17511 }, { "epoch": 3.34, "learning_rate": 2.001220209723546e-06, "loss": 2.158, "step": 17512 }, { "epoch": 3.34, "learning_rate": 2.0011630123927553e-06, "loss": 2.2439, "step": 17513 }, { "epoch": 3.34, "learning_rate": 2.0011058150619637e-06, "loss": 2.9132, "step": 17514 }, { "epoch": 3.34, "learning_rate": 2.001048617731173e-06, "loss": 2.2205, "step": 17515 }, { "epoch": 3.34, "learning_rate": 2.0009914204003812e-06, "loss": 2.0094, "step": 17516 }, { "epoch": 3.34, "learning_rate": 2.00093422306959e-06, "loss": 1.8079, "step": 17517 }, { "epoch": 3.34, "learning_rate": 2.0008770257387993e-06, "loss": 2.5963, "step": 17518 }, { "epoch": 3.34, "learning_rate": 2.0008198284080076e-06, "loss": 2.7968, "step": 17519 }, { "epoch": 3.34, "learning_rate": 2.0007626310772164e-06, "loss": 1.696, "step": 17520 }, { "epoch": 3.34, "learning_rate": 2.000705433746425e-06, "loss": 2.618, "step": 17521 }, { "epoch": 3.34, "learning_rate": 2.000648236415634e-06, "loss": 2.1005, "step": 17522 }, { "epoch": 3.34, "learning_rate": 2.000591039084843e-06, "loss": 3.0554, "step": 17523 }, { "epoch": 3.34, "learning_rate": 2.0005338417540516e-06, "loss": 2.2422, "step": 17524 }, { "epoch": 3.34, "learning_rate": 2.0004766444232604e-06, "loss": 2.2375, "step": 17525 }, { "epoch": 3.34, "learning_rate": 2.0004194470924688e-06, "loss": 2.17, "step": 17526 }, { "epoch": 3.34, "learning_rate": 2.000362249761678e-06, "loss": 2.2728, "step": 17527 }, { "epoch": 3.34, "learning_rate": 2.0003050524308868e-06, "loss": 3.1745, "step": 17528 }, { "epoch": 3.34, "learning_rate": 2.000247855100095e-06, "loss": 2.6959, "step": 17529 }, { "epoch": 3.34, "learning_rate": 2.0001906577693044e-06, "loss": 2.0795, "step": 17530 }, { "epoch": 3.34, "learning_rate": 2.0001334604385127e-06, "loss": 3.4528, "step": 17531 }, { "epoch": 3.34, "learning_rate": 2.000076263107722e-06, "loss": 2.8547, "step": 17532 }, { "epoch": 3.34, "learning_rate": 2.0000190657769303e-06, "loss": 2.1144, "step": 17533 }, { "epoch": 3.34, "learning_rate": 1.999961868446139e-06, "loss": 2.1763, "step": 17534 }, { "epoch": 3.34, "learning_rate": 1.9999046711153483e-06, "loss": 2.9391, "step": 17535 }, { "epoch": 3.34, "learning_rate": 1.9998474737845567e-06, "loss": 2.2163, "step": 17536 }, { "epoch": 3.34, "learning_rate": 1.9997902764537655e-06, "loss": 3.2522, "step": 17537 }, { "epoch": 3.34, "learning_rate": 1.9997330791229743e-06, "loss": 1.7271, "step": 17538 }, { "epoch": 3.34, "learning_rate": 1.999675881792183e-06, "loss": 2.7506, "step": 17539 }, { "epoch": 3.34, "learning_rate": 1.999618684461392e-06, "loss": 2.5167, "step": 17540 }, { "epoch": 3.34, "learning_rate": 1.9995614871306007e-06, "loss": 2.8978, "step": 17541 }, { "epoch": 3.34, "learning_rate": 1.9995042897998095e-06, "loss": 2.9409, "step": 17542 }, { "epoch": 3.34, "learning_rate": 1.999447092469018e-06, "loss": 1.8158, "step": 17543 }, { "epoch": 3.34, "learning_rate": 1.999389895138227e-06, "loss": 1.6621, "step": 17544 }, { "epoch": 3.34, "learning_rate": 1.999332697807436e-06, "loss": 2.0088, "step": 17545 }, { "epoch": 3.34, "learning_rate": 1.999275500476644e-06, "loss": 2.6783, "step": 17546 }, { "epoch": 3.34, "learning_rate": 1.9992183031458534e-06, "loss": 1.9996, "step": 17547 }, { "epoch": 3.34, "learning_rate": 1.999161105815062e-06, "loss": 2.7588, "step": 17548 }, { "epoch": 3.34, "learning_rate": 1.999103908484271e-06, "loss": 1.5476, "step": 17549 }, { "epoch": 3.34, "learning_rate": 1.9990467111534794e-06, "loss": 1.1845, "step": 17550 }, { "epoch": 3.34, "learning_rate": 1.998989513822688e-06, "loss": 2.925, "step": 17551 }, { "epoch": 3.34, "learning_rate": 1.9989323164918974e-06, "loss": 2.7567, "step": 17552 }, { "epoch": 3.34, "learning_rate": 1.9988751191611058e-06, "loss": 2.0327, "step": 17553 }, { "epoch": 3.34, "learning_rate": 1.9988179218303146e-06, "loss": 1.842, "step": 17554 }, { "epoch": 3.34, "learning_rate": 1.9987607244995234e-06, "loss": 1.5691, "step": 17555 }, { "epoch": 3.34, "learning_rate": 1.998703527168732e-06, "loss": 2.5051, "step": 17556 }, { "epoch": 3.34, "learning_rate": 1.998646329837941e-06, "loss": 1.8911, "step": 17557 }, { "epoch": 3.34, "learning_rate": 1.9985891325071497e-06, "loss": 2.0271, "step": 17558 }, { "epoch": 3.34, "learning_rate": 1.9985319351763585e-06, "loss": 2.4484, "step": 17559 }, { "epoch": 3.34, "learning_rate": 1.9984747378455673e-06, "loss": 3.0544, "step": 17560 }, { "epoch": 3.34, "learning_rate": 1.998417540514776e-06, "loss": 2.43, "step": 17561 }, { "epoch": 3.35, "learning_rate": 1.998360343183985e-06, "loss": 3.2658, "step": 17562 }, { "epoch": 3.35, "learning_rate": 1.9983031458531933e-06, "loss": 1.7858, "step": 17563 }, { "epoch": 3.35, "learning_rate": 1.9982459485224025e-06, "loss": 2.6903, "step": 17564 }, { "epoch": 3.35, "learning_rate": 1.998188751191611e-06, "loss": 2.5152, "step": 17565 }, { "epoch": 3.35, "learning_rate": 1.99813155386082e-06, "loss": 2.7059, "step": 17566 }, { "epoch": 3.35, "learning_rate": 1.998074356530029e-06, "loss": 1.8585, "step": 17567 }, { "epoch": 3.35, "learning_rate": 1.9980171591992372e-06, "loss": 3.5566, "step": 17568 }, { "epoch": 3.35, "learning_rate": 1.9979599618684465e-06, "loss": 2.5153, "step": 17569 }, { "epoch": 3.35, "learning_rate": 1.997902764537655e-06, "loss": 1.1869, "step": 17570 }, { "epoch": 3.35, "learning_rate": 1.9978455672068636e-06, "loss": 2.604, "step": 17571 }, { "epoch": 3.35, "learning_rate": 1.9977883698760724e-06, "loss": 1.9134, "step": 17572 }, { "epoch": 3.35, "learning_rate": 1.9977311725452812e-06, "loss": 2.651, "step": 17573 }, { "epoch": 3.35, "learning_rate": 1.99767397521449e-06, "loss": 1.7194, "step": 17574 }, { "epoch": 3.35, "learning_rate": 1.997616777883699e-06, "loss": 2.1203, "step": 17575 }, { "epoch": 3.35, "learning_rate": 1.9975595805529076e-06, "loss": 2.5697, "step": 17576 }, { "epoch": 3.35, "learning_rate": 1.9975023832221164e-06, "loss": 1.8226, "step": 17577 }, { "epoch": 3.35, "learning_rate": 1.997445185891325e-06, "loss": 2.2976, "step": 17578 }, { "epoch": 3.35, "learning_rate": 1.997387988560534e-06, "loss": 3.1478, "step": 17579 }, { "epoch": 3.35, "learning_rate": 1.9973307912297428e-06, "loss": 1.462, "step": 17580 }, { "epoch": 3.35, "learning_rate": 1.9972735938989516e-06, "loss": 2.2733, "step": 17581 }, { "epoch": 3.35, "learning_rate": 1.99721639656816e-06, "loss": 2.8007, "step": 17582 }, { "epoch": 3.35, "learning_rate": 1.997159199237369e-06, "loss": 3.3172, "step": 17583 }, { "epoch": 3.35, "learning_rate": 1.997102001906578e-06, "loss": 2.243, "step": 17584 }, { "epoch": 3.35, "learning_rate": 1.9970448045757863e-06, "loss": 1.6777, "step": 17585 }, { "epoch": 3.35, "learning_rate": 1.9969876072449955e-06, "loss": 3.1571, "step": 17586 }, { "epoch": 3.35, "learning_rate": 1.996930409914204e-06, "loss": 2.8047, "step": 17587 }, { "epoch": 3.35, "learning_rate": 1.9968732125834127e-06, "loss": 2.6384, "step": 17588 }, { "epoch": 3.35, "learning_rate": 1.9968160152526215e-06, "loss": 2.1171, "step": 17589 }, { "epoch": 3.35, "learning_rate": 1.9967588179218303e-06, "loss": 3.0118, "step": 17590 }, { "epoch": 3.35, "learning_rate": 1.996701620591039e-06, "loss": 2.3944, "step": 17591 }, { "epoch": 3.35, "learning_rate": 1.996644423260248e-06, "loss": 1.9927, "step": 17592 }, { "epoch": 3.35, "learning_rate": 1.9965872259294567e-06, "loss": 2.3055, "step": 17593 }, { "epoch": 3.35, "learning_rate": 1.9965300285986655e-06, "loss": 2.2031, "step": 17594 }, { "epoch": 3.35, "learning_rate": 1.9964728312678743e-06, "loss": 2.4777, "step": 17595 }, { "epoch": 3.35, "learning_rate": 1.996415633937083e-06, "loss": 2.4816, "step": 17596 }, { "epoch": 3.35, "learning_rate": 1.996358436606292e-06, "loss": 2.4594, "step": 17597 }, { "epoch": 3.35, "learning_rate": 1.9963012392755006e-06, "loss": 1.4947, "step": 17598 }, { "epoch": 3.35, "learning_rate": 1.996244041944709e-06, "loss": 1.9772, "step": 17599 }, { "epoch": 3.35, "learning_rate": 1.9961868446139182e-06, "loss": 2.5297, "step": 17600 }, { "epoch": 3.35, "learning_rate": 1.996129647283127e-06, "loss": 3.3731, "step": 17601 }, { "epoch": 3.35, "learning_rate": 1.9960724499523354e-06, "loss": 2.5015, "step": 17602 }, { "epoch": 3.35, "learning_rate": 1.9960152526215446e-06, "loss": 3.1321, "step": 17603 }, { "epoch": 3.35, "learning_rate": 1.995958055290753e-06, "loss": 1.2789, "step": 17604 }, { "epoch": 3.35, "learning_rate": 1.9959008579599618e-06, "loss": 2.004, "step": 17605 }, { "epoch": 3.35, "learning_rate": 1.995843660629171e-06, "loss": 2.3168, "step": 17606 }, { "epoch": 3.35, "learning_rate": 1.9957864632983794e-06, "loss": 2.3889, "step": 17607 }, { "epoch": 3.35, "learning_rate": 1.995729265967588e-06, "loss": 2.759, "step": 17608 }, { "epoch": 3.35, "learning_rate": 1.995672068636797e-06, "loss": 2.2703, "step": 17609 }, { "epoch": 3.35, "learning_rate": 1.9956148713060057e-06, "loss": 1.9086, "step": 17610 }, { "epoch": 3.35, "learning_rate": 1.9955576739752145e-06, "loss": 3.0372, "step": 17611 }, { "epoch": 3.35, "learning_rate": 1.9955004766444233e-06, "loss": 2.9771, "step": 17612 }, { "epoch": 3.35, "learning_rate": 1.995443279313632e-06, "loss": 2.5622, "step": 17613 }, { "epoch": 3.36, "learning_rate": 1.995386081982841e-06, "loss": 2.4572, "step": 17614 }, { "epoch": 3.36, "learning_rate": 1.9953288846520497e-06, "loss": 3.2217, "step": 17615 }, { "epoch": 3.36, "learning_rate": 1.9952716873212585e-06, "loss": 2.2238, "step": 17616 }, { "epoch": 3.36, "learning_rate": 1.9952144899904673e-06, "loss": 1.9826, "step": 17617 }, { "epoch": 3.36, "learning_rate": 1.995157292659676e-06, "loss": 2.6153, "step": 17618 }, { "epoch": 3.36, "learning_rate": 1.9951000953288845e-06, "loss": 3.5653, "step": 17619 }, { "epoch": 3.36, "learning_rate": 1.9950428979980937e-06, "loss": 2.9993, "step": 17620 }, { "epoch": 3.36, "learning_rate": 1.994985700667302e-06, "loss": 2.0657, "step": 17621 }, { "epoch": 3.36, "learning_rate": 1.994928503336511e-06, "loss": 2.4373, "step": 17622 }, { "epoch": 3.36, "learning_rate": 1.99487130600572e-06, "loss": 3.428, "step": 17623 }, { "epoch": 3.36, "learning_rate": 1.9948141086749284e-06, "loss": 2.1198, "step": 17624 }, { "epoch": 3.36, "learning_rate": 1.9947569113441372e-06, "loss": 3.0315, "step": 17625 }, { "epoch": 3.36, "learning_rate": 1.994699714013346e-06, "loss": 1.9598, "step": 17626 }, { "epoch": 3.36, "learning_rate": 1.994642516682555e-06, "loss": 2.5481, "step": 17627 }, { "epoch": 3.36, "learning_rate": 1.9945853193517636e-06, "loss": 1.9073, "step": 17628 }, { "epoch": 3.36, "learning_rate": 1.9945281220209724e-06, "loss": 2.0533, "step": 17629 }, { "epoch": 3.36, "learning_rate": 1.994470924690181e-06, "loss": 3.3189, "step": 17630 }, { "epoch": 3.36, "learning_rate": 1.99441372735939e-06, "loss": 2.2811, "step": 17631 }, { "epoch": 3.36, "learning_rate": 1.9943565300285988e-06, "loss": 2.646, "step": 17632 }, { "epoch": 3.36, "learning_rate": 1.9942993326978076e-06, "loss": 2.3441, "step": 17633 }, { "epoch": 3.36, "learning_rate": 1.9942421353670164e-06, "loss": 2.654, "step": 17634 }, { "epoch": 3.36, "learning_rate": 1.994184938036225e-06, "loss": 3.077, "step": 17635 }, { "epoch": 3.36, "learning_rate": 1.9941277407054335e-06, "loss": 2.2299, "step": 17636 }, { "epoch": 3.36, "learning_rate": 1.9940705433746428e-06, "loss": 1.8054, "step": 17637 }, { "epoch": 3.36, "learning_rate": 1.994013346043851e-06, "loss": 2.925, "step": 17638 }, { "epoch": 3.36, "learning_rate": 1.99395614871306e-06, "loss": 2.9111, "step": 17639 }, { "epoch": 3.36, "learning_rate": 1.993898951382269e-06, "loss": 1.8123, "step": 17640 }, { "epoch": 3.36, "learning_rate": 1.9938417540514775e-06, "loss": 1.9629, "step": 17641 }, { "epoch": 3.36, "learning_rate": 1.9937845567206863e-06, "loss": 1.7558, "step": 17642 }, { "epoch": 3.36, "learning_rate": 1.993727359389895e-06, "loss": 1.7019, "step": 17643 }, { "epoch": 3.36, "learning_rate": 1.993670162059104e-06, "loss": 3.1951, "step": 17644 }, { "epoch": 3.36, "learning_rate": 1.993612964728313e-06, "loss": 2.0504, "step": 17645 }, { "epoch": 3.36, "learning_rate": 1.9935557673975215e-06, "loss": 2.9732, "step": 17646 }, { "epoch": 3.36, "learning_rate": 1.9934985700667303e-06, "loss": 1.9282, "step": 17647 }, { "epoch": 3.36, "learning_rate": 1.993441372735939e-06, "loss": 0.8407, "step": 17648 }, { "epoch": 3.36, "learning_rate": 1.993384175405148e-06, "loss": 2.6435, "step": 17649 }, { "epoch": 3.36, "learning_rate": 1.9933269780743566e-06, "loss": 2.2659, "step": 17650 }, { "epoch": 3.36, "learning_rate": 1.9932697807435654e-06, "loss": 2.3675, "step": 17651 }, { "epoch": 3.36, "learning_rate": 1.9932125834127742e-06, "loss": 2.815, "step": 17652 }, { "epoch": 3.36, "learning_rate": 1.9931553860819826e-06, "loss": 2.2623, "step": 17653 }, { "epoch": 3.36, "learning_rate": 1.993098188751192e-06, "loss": 2.0577, "step": 17654 }, { "epoch": 3.36, "learning_rate": 1.9930409914204006e-06, "loss": 2.0715, "step": 17655 }, { "epoch": 3.36, "learning_rate": 1.992983794089609e-06, "loss": 2.4229, "step": 17656 }, { "epoch": 3.36, "learning_rate": 1.992926596758818e-06, "loss": 1.5452, "step": 17657 }, { "epoch": 3.36, "learning_rate": 1.9928693994280266e-06, "loss": 2.8176, "step": 17658 }, { "epoch": 3.36, "learning_rate": 1.9928122020972354e-06, "loss": 1.694, "step": 17659 }, { "epoch": 3.36, "learning_rate": 1.992755004766444e-06, "loss": 2.6746, "step": 17660 }, { "epoch": 3.36, "learning_rate": 1.992697807435653e-06, "loss": 2.1173, "step": 17661 }, { "epoch": 3.36, "learning_rate": 1.992640610104862e-06, "loss": 1.8058, "step": 17662 }, { "epoch": 3.36, "learning_rate": 1.9925834127740705e-06, "loss": 2.4795, "step": 17663 }, { "epoch": 3.36, "learning_rate": 1.9925262154432793e-06, "loss": 1.7276, "step": 17664 }, { "epoch": 3.36, "learning_rate": 1.992469018112488e-06, "loss": 3.5702, "step": 17665 }, { "epoch": 3.36, "learning_rate": 1.992411820781697e-06, "loss": 2.1794, "step": 17666 }, { "epoch": 3.37, "learning_rate": 1.9923546234509057e-06, "loss": 2.6102, "step": 17667 }, { "epoch": 3.37, "learning_rate": 1.9922974261201145e-06, "loss": 2.4824, "step": 17668 }, { "epoch": 3.37, "learning_rate": 1.9922402287893233e-06, "loss": 2.2218, "step": 17669 }, { "epoch": 3.37, "learning_rate": 1.9921830314585317e-06, "loss": 3.3265, "step": 17670 }, { "epoch": 3.37, "learning_rate": 1.992125834127741e-06, "loss": 2.7088, "step": 17671 }, { "epoch": 3.37, "learning_rate": 1.9920686367969497e-06, "loss": 2.001, "step": 17672 }, { "epoch": 3.37, "learning_rate": 1.992011439466158e-06, "loss": 1.8639, "step": 17673 }, { "epoch": 3.37, "learning_rate": 1.9919542421353673e-06, "loss": 3.4069, "step": 17674 }, { "epoch": 3.37, "learning_rate": 1.9918970448045756e-06, "loss": 2.54, "step": 17675 }, { "epoch": 3.37, "learning_rate": 1.9918398474737844e-06, "loss": 2.5778, "step": 17676 }, { "epoch": 3.37, "learning_rate": 1.9917826501429932e-06, "loss": 2.5997, "step": 17677 }, { "epoch": 3.37, "learning_rate": 1.991725452812202e-06, "loss": 2.8361, "step": 17678 }, { "epoch": 3.37, "learning_rate": 1.9916682554814112e-06, "loss": 1.9542, "step": 17679 }, { "epoch": 3.37, "learning_rate": 1.9916110581506196e-06, "loss": 3.5538, "step": 17680 }, { "epoch": 3.37, "learning_rate": 1.9915538608198284e-06, "loss": 2.8065, "step": 17681 }, { "epoch": 3.37, "learning_rate": 1.991496663489037e-06, "loss": 2.6315, "step": 17682 }, { "epoch": 3.37, "learning_rate": 1.991439466158246e-06, "loss": 2.0732, "step": 17683 }, { "epoch": 3.37, "learning_rate": 1.991382268827455e-06, "loss": 1.4018, "step": 17684 }, { "epoch": 3.37, "learning_rate": 1.9913250714966636e-06, "loss": 2.5198, "step": 17685 }, { "epoch": 3.37, "learning_rate": 1.9912678741658724e-06, "loss": 2.9203, "step": 17686 }, { "epoch": 3.37, "learning_rate": 1.9912106768350807e-06, "loss": 2.8046, "step": 17687 }, { "epoch": 3.37, "learning_rate": 1.99115347950429e-06, "loss": 2.3988, "step": 17688 }, { "epoch": 3.37, "learning_rate": 1.9910962821734988e-06, "loss": 2.8525, "step": 17689 }, { "epoch": 3.37, "learning_rate": 1.991039084842707e-06, "loss": 2.3782, "step": 17690 }, { "epoch": 3.37, "learning_rate": 1.9909818875119163e-06, "loss": 1.7485, "step": 17691 }, { "epoch": 3.37, "learning_rate": 1.9909246901811247e-06, "loss": 2.6142, "step": 17692 }, { "epoch": 3.37, "learning_rate": 1.9908674928503335e-06, "loss": 1.8417, "step": 17693 }, { "epoch": 3.37, "learning_rate": 1.9908102955195427e-06, "loss": 2.3553, "step": 17694 }, { "epoch": 3.37, "learning_rate": 1.990753098188751e-06, "loss": 3.0752, "step": 17695 }, { "epoch": 3.37, "learning_rate": 1.9906959008579603e-06, "loss": 2.2441, "step": 17696 }, { "epoch": 3.37, "learning_rate": 1.9906387035271687e-06, "loss": 3.2433, "step": 17697 }, { "epoch": 3.37, "learning_rate": 1.9905815061963775e-06, "loss": 2.2382, "step": 17698 }, { "epoch": 3.37, "learning_rate": 1.9905243088655863e-06, "loss": 2.3707, "step": 17699 }, { "epoch": 3.37, "learning_rate": 1.990467111534795e-06, "loss": 2.5774, "step": 17700 }, { "epoch": 3.37, "learning_rate": 1.990409914204004e-06, "loss": 1.7928, "step": 17701 }, { "epoch": 3.37, "learning_rate": 1.9903527168732127e-06, "loss": 1.8457, "step": 17702 }, { "epoch": 3.37, "learning_rate": 1.9902955195424214e-06, "loss": 2.9057, "step": 17703 }, { "epoch": 3.37, "learning_rate": 1.9902383222116302e-06, "loss": 2.2639, "step": 17704 }, { "epoch": 3.37, "learning_rate": 1.990181124880839e-06, "loss": 1.9653, "step": 17705 }, { "epoch": 3.37, "learning_rate": 1.990123927550048e-06, "loss": 2.8925, "step": 17706 }, { "epoch": 3.37, "learning_rate": 1.990066730219256e-06, "loss": 2.6957, "step": 17707 }, { "epoch": 3.37, "learning_rate": 1.9900095328884654e-06, "loss": 2.1928, "step": 17708 }, { "epoch": 3.37, "learning_rate": 1.989952335557674e-06, "loss": 1.7637, "step": 17709 }, { "epoch": 3.37, "learning_rate": 1.989895138226883e-06, "loss": 1.7137, "step": 17710 }, { "epoch": 3.37, "learning_rate": 1.989837940896092e-06, "loss": 3.4123, "step": 17711 }, { "epoch": 3.37, "learning_rate": 1.9897807435653e-06, "loss": 2.6761, "step": 17712 }, { "epoch": 3.37, "learning_rate": 1.9897235462345094e-06, "loss": 2.6652, "step": 17713 }, { "epoch": 3.37, "learning_rate": 1.9896663489037178e-06, "loss": 2.5365, "step": 17714 }, { "epoch": 3.37, "learning_rate": 1.9896091515729266e-06, "loss": 2.7443, "step": 17715 }, { "epoch": 3.37, "learning_rate": 1.9895519542421353e-06, "loss": 2.2489, "step": 17716 }, { "epoch": 3.37, "learning_rate": 1.989494756911344e-06, "loss": 1.5578, "step": 17717 }, { "epoch": 3.37, "learning_rate": 1.989437559580553e-06, "loss": 2.3004, "step": 17718 }, { "epoch": 3.38, "learning_rate": 1.9893803622497617e-06, "loss": 1.2231, "step": 17719 }, { "epoch": 3.38, "learning_rate": 1.9893231649189705e-06, "loss": 3.0585, "step": 17720 }, { "epoch": 3.38, "learning_rate": 1.9892659675881793e-06, "loss": 2.7075, "step": 17721 }, { "epoch": 3.38, "learning_rate": 1.989208770257388e-06, "loss": 1.9309, "step": 17722 }, { "epoch": 3.38, "learning_rate": 1.989151572926597e-06, "loss": 2.4799, "step": 17723 }, { "epoch": 3.38, "learning_rate": 1.9890943755958053e-06, "loss": 1.5117, "step": 17724 }, { "epoch": 3.38, "learning_rate": 1.9890371782650145e-06, "loss": 2.1071, "step": 17725 }, { "epoch": 3.38, "learning_rate": 1.988979980934223e-06, "loss": 2.8377, "step": 17726 }, { "epoch": 3.38, "learning_rate": 1.988922783603432e-06, "loss": 1.4139, "step": 17727 }, { "epoch": 3.38, "learning_rate": 1.988865586272641e-06, "loss": 2.5068, "step": 17728 }, { "epoch": 3.38, "learning_rate": 1.9888083889418492e-06, "loss": 2.5935, "step": 17729 }, { "epoch": 3.38, "learning_rate": 1.9887511916110585e-06, "loss": 2.1015, "step": 17730 }, { "epoch": 3.38, "learning_rate": 1.988693994280267e-06, "loss": 3.4497, "step": 17731 }, { "epoch": 3.38, "learning_rate": 1.9886367969494756e-06, "loss": 3.397, "step": 17732 }, { "epoch": 3.38, "learning_rate": 1.988579599618685e-06, "loss": 1.8659, "step": 17733 }, { "epoch": 3.38, "learning_rate": 1.988522402287893e-06, "loss": 2.8452, "step": 17734 }, { "epoch": 3.38, "learning_rate": 1.988465204957102e-06, "loss": 3.5762, "step": 17735 }, { "epoch": 3.38, "learning_rate": 1.988408007626311e-06, "loss": 2.5085, "step": 17736 }, { "epoch": 3.38, "learning_rate": 1.9883508102955196e-06, "loss": 2.1739, "step": 17737 }, { "epoch": 3.38, "learning_rate": 1.9882936129647284e-06, "loss": 2.8821, "step": 17738 }, { "epoch": 3.38, "learning_rate": 1.988236415633937e-06, "loss": 2.5434, "step": 17739 }, { "epoch": 3.38, "learning_rate": 1.988179218303146e-06, "loss": 2.5747, "step": 17740 }, { "epoch": 3.38, "learning_rate": 1.9881220209723543e-06, "loss": 2.9238, "step": 17741 }, { "epoch": 3.38, "learning_rate": 1.9880648236415636e-06, "loss": 3.3587, "step": 17742 }, { "epoch": 3.38, "learning_rate": 1.9880076263107724e-06, "loss": 2.9555, "step": 17743 }, { "epoch": 3.38, "learning_rate": 1.987950428979981e-06, "loss": 2.2884, "step": 17744 }, { "epoch": 3.38, "learning_rate": 1.98789323164919e-06, "loss": 2.0443, "step": 17745 }, { "epoch": 3.38, "learning_rate": 1.9878360343183983e-06, "loss": 1.4267, "step": 17746 }, { "epoch": 3.38, "learning_rate": 1.9877788369876075e-06, "loss": 2.9164, "step": 17747 }, { "epoch": 3.38, "learning_rate": 1.987721639656816e-06, "loss": 2.1637, "step": 17748 }, { "epoch": 3.38, "learning_rate": 1.9876644423260247e-06, "loss": 2.6751, "step": 17749 }, { "epoch": 3.38, "learning_rate": 1.987607244995234e-06, "loss": 2.6568, "step": 17750 }, { "epoch": 3.38, "learning_rate": 1.9875500476644423e-06, "loss": 2.3549, "step": 17751 }, { "epoch": 3.38, "learning_rate": 1.987492850333651e-06, "loss": 1.6554, "step": 17752 }, { "epoch": 3.38, "learning_rate": 1.98743565300286e-06, "loss": 1.9006, "step": 17753 }, { "epoch": 3.38, "learning_rate": 1.9873784556720687e-06, "loss": 1.2668, "step": 17754 }, { "epoch": 3.38, "learning_rate": 1.9873212583412775e-06, "loss": 3.038, "step": 17755 }, { "epoch": 3.38, "learning_rate": 1.9872640610104863e-06, "loss": 2.6444, "step": 17756 }, { "epoch": 3.38, "learning_rate": 1.987206863679695e-06, "loss": 2.0712, "step": 17757 }, { "epoch": 3.38, "learning_rate": 1.9871496663489034e-06, "loss": 3.3653, "step": 17758 }, { "epoch": 3.38, "learning_rate": 1.9870924690181126e-06, "loss": 1.4889, "step": 17759 }, { "epoch": 3.38, "learning_rate": 1.9870352716873214e-06, "loss": 1.5305, "step": 17760 }, { "epoch": 3.38, "learning_rate": 1.9869780743565302e-06, "loss": 2.7114, "step": 17761 }, { "epoch": 3.38, "learning_rate": 1.986920877025739e-06, "loss": 3.2097, "step": 17762 }, { "epoch": 3.38, "learning_rate": 1.9868636796949474e-06, "loss": 3.9256, "step": 17763 }, { "epoch": 3.38, "learning_rate": 1.9868064823641566e-06, "loss": 2.1496, "step": 17764 }, { "epoch": 3.38, "learning_rate": 1.986749285033365e-06, "loss": 1.8722, "step": 17765 }, { "epoch": 3.38, "learning_rate": 1.9866920877025738e-06, "loss": 1.6426, "step": 17766 }, { "epoch": 3.38, "learning_rate": 1.986634890371783e-06, "loss": 3.3836, "step": 17767 }, { "epoch": 3.38, "learning_rate": 1.9865776930409914e-06, "loss": 2.2315, "step": 17768 }, { "epoch": 3.38, "learning_rate": 1.9865204957102e-06, "loss": 2.5503, "step": 17769 }, { "epoch": 3.38, "learning_rate": 1.986463298379409e-06, "loss": 3.0696, "step": 17770 }, { "epoch": 3.38, "learning_rate": 1.9864061010486177e-06, "loss": 1.6734, "step": 17771 }, { "epoch": 3.39, "learning_rate": 1.9863489037178265e-06, "loss": 1.652, "step": 17772 }, { "epoch": 3.39, "learning_rate": 1.9862917063870353e-06, "loss": 2.381, "step": 17773 }, { "epoch": 3.39, "learning_rate": 1.986234509056244e-06, "loss": 3.2327, "step": 17774 }, { "epoch": 3.39, "learning_rate": 1.986177311725453e-06, "loss": 2.4796, "step": 17775 }, { "epoch": 3.39, "learning_rate": 1.9861201143946617e-06, "loss": 1.4031, "step": 17776 }, { "epoch": 3.39, "learning_rate": 1.9860629170638705e-06, "loss": 2.9838, "step": 17777 }, { "epoch": 3.39, "learning_rate": 1.9860057197330793e-06, "loss": 1.5977, "step": 17778 }, { "epoch": 3.39, "learning_rate": 1.985948522402288e-06, "loss": 1.5606, "step": 17779 }, { "epoch": 3.39, "learning_rate": 1.9858913250714965e-06, "loss": 2.689, "step": 17780 }, { "epoch": 3.39, "learning_rate": 1.9858341277407057e-06, "loss": 1.5583, "step": 17781 }, { "epoch": 3.39, "learning_rate": 1.9857769304099145e-06, "loss": 1.4487, "step": 17782 }, { "epoch": 3.39, "learning_rate": 1.985719733079123e-06, "loss": 2.6294, "step": 17783 }, { "epoch": 3.39, "learning_rate": 1.985662535748332e-06, "loss": 2.5401, "step": 17784 }, { "epoch": 3.39, "learning_rate": 1.9856053384175404e-06, "loss": 2.4433, "step": 17785 }, { "epoch": 3.39, "learning_rate": 1.9855481410867492e-06, "loss": 3.0273, "step": 17786 }, { "epoch": 3.39, "learning_rate": 1.985490943755958e-06, "loss": 1.5787, "step": 17787 }, { "epoch": 3.39, "learning_rate": 1.985433746425167e-06, "loss": 2.635, "step": 17788 }, { "epoch": 3.39, "learning_rate": 1.9853765490943756e-06, "loss": 2.6219, "step": 17789 }, { "epoch": 3.39, "learning_rate": 1.9853193517635844e-06, "loss": 1.8554, "step": 17790 }, { "epoch": 3.39, "learning_rate": 1.985262154432793e-06, "loss": 2.5536, "step": 17791 }, { "epoch": 3.39, "learning_rate": 1.985204957102002e-06, "loss": 2.5881, "step": 17792 }, { "epoch": 3.39, "learning_rate": 1.9851477597712108e-06, "loss": 1.569, "step": 17793 }, { "epoch": 3.39, "learning_rate": 1.9850905624404196e-06, "loss": 2.2258, "step": 17794 }, { "epoch": 3.39, "learning_rate": 1.9850333651096284e-06, "loss": 1.7026, "step": 17795 }, { "epoch": 3.39, "learning_rate": 1.984976167778837e-06, "loss": 3.1999, "step": 17796 }, { "epoch": 3.39, "learning_rate": 1.9849189704480455e-06, "loss": 2.5508, "step": 17797 }, { "epoch": 3.39, "learning_rate": 1.9848617731172547e-06, "loss": 2.6247, "step": 17798 }, { "epoch": 3.39, "learning_rate": 1.9848045757864635e-06, "loss": 2.3645, "step": 17799 }, { "epoch": 3.39, "learning_rate": 1.984747378455672e-06, "loss": 3.0332, "step": 17800 }, { "epoch": 3.39, "learning_rate": 1.984690181124881e-06, "loss": 1.8798, "step": 17801 }, { "epoch": 3.39, "learning_rate": 1.9846329837940895e-06, "loss": 2.6935, "step": 17802 }, { "epoch": 3.39, "learning_rate": 1.9845757864632983e-06, "loss": 4.0161, "step": 17803 }, { "epoch": 3.39, "learning_rate": 1.984518589132507e-06, "loss": 1.752, "step": 17804 }, { "epoch": 3.39, "learning_rate": 1.984461391801716e-06, "loss": 2.289, "step": 17805 }, { "epoch": 3.39, "learning_rate": 1.9844041944709247e-06, "loss": 1.771, "step": 17806 }, { "epoch": 3.39, "learning_rate": 1.9843469971401335e-06, "loss": 3.2647, "step": 17807 }, { "epoch": 3.39, "learning_rate": 1.9842897998093423e-06, "loss": 2.0021, "step": 17808 }, { "epoch": 3.39, "learning_rate": 1.984232602478551e-06, "loss": 2.7417, "step": 17809 }, { "epoch": 3.39, "learning_rate": 1.98417540514776e-06, "loss": 2.5748, "step": 17810 }, { "epoch": 3.39, "learning_rate": 1.9841182078169686e-06, "loss": 2.5225, "step": 17811 }, { "epoch": 3.39, "learning_rate": 1.9840610104861774e-06, "loss": 2.6131, "step": 17812 }, { "epoch": 3.39, "learning_rate": 1.9840038131553862e-06, "loss": 2.1955, "step": 17813 }, { "epoch": 3.39, "learning_rate": 1.9839466158245946e-06, "loss": 3.1398, "step": 17814 }, { "epoch": 3.39, "learning_rate": 1.983889418493804e-06, "loss": 2.2382, "step": 17815 }, { "epoch": 3.39, "learning_rate": 1.9838322211630126e-06, "loss": 1.7477, "step": 17816 }, { "epoch": 3.39, "learning_rate": 1.983775023832221e-06, "loss": 2.3812, "step": 17817 }, { "epoch": 3.39, "learning_rate": 1.98371782650143e-06, "loss": 2.3076, "step": 17818 }, { "epoch": 3.39, "learning_rate": 1.9836606291706386e-06, "loss": 3.0545, "step": 17819 }, { "epoch": 3.39, "learning_rate": 1.9836034318398474e-06, "loss": 2.378, "step": 17820 }, { "epoch": 3.39, "learning_rate": 1.983546234509056e-06, "loss": 1.6684, "step": 17821 }, { "epoch": 3.39, "learning_rate": 1.983489037178265e-06, "loss": 3.4051, "step": 17822 }, { "epoch": 3.39, "learning_rate": 1.9834318398474737e-06, "loss": 2.3932, "step": 17823 }, { "epoch": 3.4, "learning_rate": 1.9833746425166825e-06, "loss": 2.0096, "step": 17824 }, { "epoch": 3.4, "learning_rate": 1.9833174451858913e-06, "loss": 2.8172, "step": 17825 }, { "epoch": 3.4, "learning_rate": 1.9832602478551e-06, "loss": 2.5447, "step": 17826 }, { "epoch": 3.4, "learning_rate": 1.983203050524309e-06, "loss": 3.1424, "step": 17827 }, { "epoch": 3.4, "learning_rate": 1.9831458531935177e-06, "loss": 1.386, "step": 17828 }, { "epoch": 3.4, "learning_rate": 1.9830886558627265e-06, "loss": 1.8625, "step": 17829 }, { "epoch": 3.4, "learning_rate": 1.9830314585319353e-06, "loss": 3.3791, "step": 17830 }, { "epoch": 3.4, "learning_rate": 1.982974261201144e-06, "loss": 2.4574, "step": 17831 }, { "epoch": 3.4, "learning_rate": 1.982917063870353e-06, "loss": 2.6408, "step": 17832 }, { "epoch": 3.4, "learning_rate": 1.9828598665395617e-06, "loss": 2.1574, "step": 17833 }, { "epoch": 3.4, "learning_rate": 1.98280266920877e-06, "loss": 2.5753, "step": 17834 }, { "epoch": 3.4, "learning_rate": 1.9827454718779793e-06, "loss": 2.3952, "step": 17835 }, { "epoch": 3.4, "learning_rate": 1.9826882745471876e-06, "loss": 1.9421, "step": 17836 }, { "epoch": 3.4, "learning_rate": 1.9826310772163964e-06, "loss": 2.3199, "step": 17837 }, { "epoch": 3.4, "learning_rate": 1.9825738798856056e-06, "loss": 2.4017, "step": 17838 }, { "epoch": 3.4, "learning_rate": 1.982516682554814e-06, "loss": 3.3677, "step": 17839 }, { "epoch": 3.4, "learning_rate": 1.9824594852240232e-06, "loss": 2.9299, "step": 17840 }, { "epoch": 3.4, "learning_rate": 1.9824022878932316e-06, "loss": 2.2366, "step": 17841 }, { "epoch": 3.4, "learning_rate": 1.9823450905624404e-06, "loss": 2.4871, "step": 17842 }, { "epoch": 3.4, "learning_rate": 1.982287893231649e-06, "loss": 3.1337, "step": 17843 }, { "epoch": 3.4, "learning_rate": 1.982230695900858e-06, "loss": 1.4516, "step": 17844 }, { "epoch": 3.4, "learning_rate": 1.9821734985700668e-06, "loss": 2.6616, "step": 17845 }, { "epoch": 3.4, "learning_rate": 1.9821163012392756e-06, "loss": 2.0079, "step": 17846 }, { "epoch": 3.4, "learning_rate": 1.9820591039084844e-06, "loss": 2.5279, "step": 17847 }, { "epoch": 3.4, "learning_rate": 1.982001906577693e-06, "loss": 2.7272, "step": 17848 }, { "epoch": 3.4, "learning_rate": 1.981944709246902e-06, "loss": 2.8939, "step": 17849 }, { "epoch": 3.4, "learning_rate": 1.9818875119161108e-06, "loss": 2.8329, "step": 17850 }, { "epoch": 3.4, "learning_rate": 1.981830314585319e-06, "loss": 3.3878, "step": 17851 }, { "epoch": 3.4, "learning_rate": 1.9817731172545283e-06, "loss": 2.961, "step": 17852 }, { "epoch": 3.4, "learning_rate": 1.9817159199237367e-06, "loss": 2.4039, "step": 17853 }, { "epoch": 3.4, "learning_rate": 1.9816587225929455e-06, "loss": 3.1399, "step": 17854 }, { "epoch": 3.4, "learning_rate": 1.9816015252621547e-06, "loss": 1.9714, "step": 17855 }, { "epoch": 3.4, "learning_rate": 1.981544327931363e-06, "loss": 2.6038, "step": 17856 }, { "epoch": 3.4, "learning_rate": 1.9814871306005723e-06, "loss": 2.8224, "step": 17857 }, { "epoch": 3.4, "learning_rate": 1.9814299332697807e-06, "loss": 1.7302, "step": 17858 }, { "epoch": 3.4, "learning_rate": 1.9813727359389895e-06, "loss": 3.696, "step": 17859 }, { "epoch": 3.4, "learning_rate": 1.9813155386081983e-06, "loss": 1.6484, "step": 17860 }, { "epoch": 3.4, "learning_rate": 1.981258341277407e-06, "loss": 2.0152, "step": 17861 }, { "epoch": 3.4, "learning_rate": 1.981201143946616e-06, "loss": 2.1352, "step": 17862 }, { "epoch": 3.4, "learning_rate": 1.9811439466158246e-06, "loss": 2.8528, "step": 17863 }, { "epoch": 3.4, "learning_rate": 1.9810867492850334e-06, "loss": 1.8618, "step": 17864 }, { "epoch": 3.4, "learning_rate": 1.9810295519542422e-06, "loss": 2.0664, "step": 17865 }, { "epoch": 3.4, "learning_rate": 1.980972354623451e-06, "loss": 1.6472, "step": 17866 }, { "epoch": 3.4, "learning_rate": 1.98091515729266e-06, "loss": 2.5525, "step": 17867 }, { "epoch": 3.4, "learning_rate": 1.980857959961868e-06, "loss": 2.6898, "step": 17868 }, { "epoch": 3.4, "learning_rate": 1.9808007626310774e-06, "loss": 1.4429, "step": 17869 }, { "epoch": 3.4, "learning_rate": 1.980743565300286e-06, "loss": 1.9963, "step": 17870 }, { "epoch": 3.4, "learning_rate": 1.9806863679694946e-06, "loss": 2.3271, "step": 17871 }, { "epoch": 3.4, "learning_rate": 1.980629170638704e-06, "loss": 1.9598, "step": 17872 }, { "epoch": 3.4, "learning_rate": 1.980571973307912e-06, "loss": 1.4838, "step": 17873 }, { "epoch": 3.4, "learning_rate": 1.9805147759771214e-06, "loss": 2.1708, "step": 17874 }, { "epoch": 3.4, "learning_rate": 1.9804575786463298e-06, "loss": 3.141, "step": 17875 }, { "epoch": 3.4, "learning_rate": 1.9804003813155385e-06, "loss": 3.5698, "step": 17876 }, { "epoch": 3.41, "learning_rate": 1.9803431839847478e-06, "loss": 2.3177, "step": 17877 }, { "epoch": 3.41, "learning_rate": 1.980285986653956e-06, "loss": 2.1618, "step": 17878 }, { "epoch": 3.41, "learning_rate": 1.980228789323165e-06, "loss": 2.2802, "step": 17879 }, { "epoch": 3.41, "learning_rate": 1.9801715919923737e-06, "loss": 3.411, "step": 17880 }, { "epoch": 3.41, "learning_rate": 1.9801143946615825e-06, "loss": 2.1725, "step": 17881 }, { "epoch": 3.41, "learning_rate": 1.9800571973307913e-06, "loss": 2.6485, "step": 17882 }, { "epoch": 3.41, "learning_rate": 1.98e-06, "loss": 2.2507, "step": 17883 }, { "epoch": 3.41, "learning_rate": 1.979942802669209e-06, "loss": 2.3476, "step": 17884 }, { "epoch": 3.41, "learning_rate": 1.9798856053384173e-06, "loss": 2.2372, "step": 17885 }, { "epoch": 3.41, "learning_rate": 1.9798284080076265e-06, "loss": 2.5821, "step": 17886 }, { "epoch": 3.41, "learning_rate": 1.9797712106768353e-06, "loss": 2.3774, "step": 17887 }, { "epoch": 3.41, "learning_rate": 1.9797140133460436e-06, "loss": 2.6748, "step": 17888 }, { "epoch": 3.41, "learning_rate": 1.979656816015253e-06, "loss": 1.5309, "step": 17889 }, { "epoch": 3.41, "learning_rate": 1.9795996186844612e-06, "loss": 2.1426, "step": 17890 }, { "epoch": 3.41, "learning_rate": 1.9795424213536705e-06, "loss": 2.3963, "step": 17891 }, { "epoch": 3.41, "learning_rate": 1.979485224022879e-06, "loss": 2.27, "step": 17892 }, { "epoch": 3.41, "learning_rate": 1.9794280266920876e-06, "loss": 2.3301, "step": 17893 }, { "epoch": 3.41, "learning_rate": 1.979370829361297e-06, "loss": 2.4175, "step": 17894 }, { "epoch": 3.41, "learning_rate": 1.979313632030505e-06, "loss": 1.9909, "step": 17895 }, { "epoch": 3.41, "learning_rate": 1.979256434699714e-06, "loss": 2.1091, "step": 17896 }, { "epoch": 3.41, "learning_rate": 1.979199237368923e-06, "loss": 1.7919, "step": 17897 }, { "epoch": 3.41, "learning_rate": 1.9791420400381316e-06, "loss": 2.5205, "step": 17898 }, { "epoch": 3.41, "learning_rate": 1.9790848427073404e-06, "loss": 2.3333, "step": 17899 }, { "epoch": 3.41, "learning_rate": 1.979027645376549e-06, "loss": 1.4125, "step": 17900 }, { "epoch": 3.41, "learning_rate": 1.978970448045758e-06, "loss": 2.168, "step": 17901 }, { "epoch": 3.41, "learning_rate": 1.9789132507149663e-06, "loss": 1.4749, "step": 17902 }, { "epoch": 3.41, "learning_rate": 1.9788560533841756e-06, "loss": 2.1802, "step": 17903 }, { "epoch": 3.41, "learning_rate": 1.9787988560533843e-06, "loss": 1.9837, "step": 17904 }, { "epoch": 3.41, "learning_rate": 1.978741658722593e-06, "loss": 2.9377, "step": 17905 }, { "epoch": 3.41, "learning_rate": 1.978684461391802e-06, "loss": 3.3995, "step": 17906 }, { "epoch": 3.41, "learning_rate": 1.9786272640610103e-06, "loss": 2.5047, "step": 17907 }, { "epoch": 3.41, "learning_rate": 1.9785700667302195e-06, "loss": 3.5091, "step": 17908 }, { "epoch": 3.41, "learning_rate": 1.978512869399428e-06, "loss": 2.5166, "step": 17909 }, { "epoch": 3.41, "learning_rate": 1.9784556720686367e-06, "loss": 2.1363, "step": 17910 }, { "epoch": 3.41, "learning_rate": 1.978398474737846e-06, "loss": 2.4562, "step": 17911 }, { "epoch": 3.41, "learning_rate": 1.9783412774070543e-06, "loss": 2.6407, "step": 17912 }, { "epoch": 3.41, "learning_rate": 1.978284080076263e-06, "loss": 2.8487, "step": 17913 }, { "epoch": 3.41, "learning_rate": 1.978226882745472e-06, "loss": 3.5725, "step": 17914 }, { "epoch": 3.41, "learning_rate": 1.9781696854146807e-06, "loss": 2.5218, "step": 17915 }, { "epoch": 3.41, "learning_rate": 1.9781124880838894e-06, "loss": 2.3879, "step": 17916 }, { "epoch": 3.41, "learning_rate": 1.9780552907530982e-06, "loss": 1.9802, "step": 17917 }, { "epoch": 3.41, "learning_rate": 1.977998093422307e-06, "loss": 2.8062, "step": 17918 }, { "epoch": 3.41, "learning_rate": 1.977940896091516e-06, "loss": 2.4455, "step": 17919 }, { "epoch": 3.41, "learning_rate": 1.9778836987607246e-06, "loss": 2.7701, "step": 17920 }, { "epoch": 3.41, "learning_rate": 1.9778265014299334e-06, "loss": 1.4032, "step": 17921 }, { "epoch": 3.41, "learning_rate": 1.9777693040991422e-06, "loss": 2.1263, "step": 17922 }, { "epoch": 3.41, "learning_rate": 1.977712106768351e-06, "loss": 1.8354, "step": 17923 }, { "epoch": 3.41, "learning_rate": 1.9776549094375594e-06, "loss": 2.3288, "step": 17924 }, { "epoch": 3.41, "learning_rate": 1.9775977121067686e-06, "loss": 3.0018, "step": 17925 }, { "epoch": 3.41, "learning_rate": 1.9775405147759774e-06, "loss": 1.4896, "step": 17926 }, { "epoch": 3.41, "learning_rate": 1.9774833174451858e-06, "loss": 2.0653, "step": 17927 }, { "epoch": 3.41, "learning_rate": 1.977426120114395e-06, "loss": 1.7751, "step": 17928 }, { "epoch": 3.42, "learning_rate": 1.9773689227836033e-06, "loss": 2.4768, "step": 17929 }, { "epoch": 3.42, "learning_rate": 1.977311725452812e-06, "loss": 2.6829, "step": 17930 }, { "epoch": 3.42, "learning_rate": 1.977254528122021e-06, "loss": 2.6482, "step": 17931 }, { "epoch": 3.42, "learning_rate": 1.9771973307912297e-06, "loss": 0.963, "step": 17932 }, { "epoch": 3.42, "learning_rate": 1.9771401334604385e-06, "loss": 1.7664, "step": 17933 }, { "epoch": 3.42, "learning_rate": 1.9770829361296473e-06, "loss": 2.4111, "step": 17934 }, { "epoch": 3.42, "learning_rate": 1.977025738798856e-06, "loss": 1.7357, "step": 17935 }, { "epoch": 3.42, "learning_rate": 1.976968541468065e-06, "loss": 2.7091, "step": 17936 }, { "epoch": 3.42, "learning_rate": 1.9769113441372737e-06, "loss": 1.9268, "step": 17937 }, { "epoch": 3.42, "learning_rate": 1.9768541468064825e-06, "loss": 3.4568, "step": 17938 }, { "epoch": 3.42, "learning_rate": 1.9767969494756913e-06, "loss": 1.7305, "step": 17939 }, { "epoch": 3.42, "learning_rate": 1.9767397521449e-06, "loss": 3.2508, "step": 17940 }, { "epoch": 3.42, "learning_rate": 1.9766825548141084e-06, "loss": 2.9551, "step": 17941 }, { "epoch": 3.42, "learning_rate": 1.9766253574833177e-06, "loss": 2.9705, "step": 17942 }, { "epoch": 3.42, "learning_rate": 1.9765681601525265e-06, "loss": 2.5332, "step": 17943 }, { "epoch": 3.42, "learning_rate": 1.976510962821735e-06, "loss": 2.4774, "step": 17944 }, { "epoch": 3.42, "learning_rate": 1.976453765490944e-06, "loss": 1.968, "step": 17945 }, { "epoch": 3.42, "learning_rate": 1.9763965681601524e-06, "loss": 1.6762, "step": 17946 }, { "epoch": 3.42, "learning_rate": 1.976339370829361e-06, "loss": 2.4503, "step": 17947 }, { "epoch": 3.42, "learning_rate": 1.97628217349857e-06, "loss": 2.2345, "step": 17948 }, { "epoch": 3.42, "learning_rate": 1.976224976167779e-06, "loss": 2.8131, "step": 17949 }, { "epoch": 3.42, "learning_rate": 1.9761677788369876e-06, "loss": 2.811, "step": 17950 }, { "epoch": 3.42, "learning_rate": 1.9761105815061964e-06, "loss": 1.8694, "step": 17951 }, { "epoch": 3.42, "learning_rate": 1.976053384175405e-06, "loss": 3.1985, "step": 17952 }, { "epoch": 3.42, "learning_rate": 1.975996186844614e-06, "loss": 3.1659, "step": 17953 }, { "epoch": 3.42, "learning_rate": 1.9759389895138228e-06, "loss": 2.7061, "step": 17954 }, { "epoch": 3.42, "learning_rate": 1.9758817921830316e-06, "loss": 3.3175, "step": 17955 }, { "epoch": 3.42, "learning_rate": 1.9758245948522404e-06, "loss": 2.9597, "step": 17956 }, { "epoch": 3.42, "learning_rate": 1.975767397521449e-06, "loss": 2.2355, "step": 17957 }, { "epoch": 3.42, "learning_rate": 1.975710200190658e-06, "loss": 1.8824, "step": 17958 }, { "epoch": 3.42, "learning_rate": 1.9756530028598667e-06, "loss": 2.7381, "step": 17959 }, { "epoch": 3.42, "learning_rate": 1.9755958055290755e-06, "loss": 1.6557, "step": 17960 }, { "epoch": 3.42, "learning_rate": 1.975538608198284e-06, "loss": 2.2555, "step": 17961 }, { "epoch": 3.42, "learning_rate": 1.975481410867493e-06, "loss": 3.7535, "step": 17962 }, { "epoch": 3.42, "learning_rate": 1.9754242135367015e-06, "loss": 1.8768, "step": 17963 }, { "epoch": 3.42, "learning_rate": 1.9753670162059103e-06, "loss": 2.4918, "step": 17964 }, { "epoch": 3.42, "learning_rate": 1.9753098188751195e-06, "loss": 3.3163, "step": 17965 }, { "epoch": 3.42, "learning_rate": 1.975252621544328e-06, "loss": 1.835, "step": 17966 }, { "epoch": 3.42, "learning_rate": 1.9751954242135367e-06, "loss": 3.0517, "step": 17967 }, { "epoch": 3.42, "learning_rate": 1.9751382268827455e-06, "loss": 2.5191, "step": 17968 }, { "epoch": 3.42, "learning_rate": 1.9750810295519543e-06, "loss": 1.5295, "step": 17969 }, { "epoch": 3.42, "learning_rate": 1.975023832221163e-06, "loss": 2.1, "step": 17970 }, { "epoch": 3.42, "learning_rate": 1.974966634890372e-06, "loss": 1.1519, "step": 17971 }, { "epoch": 3.42, "learning_rate": 1.9749094375595806e-06, "loss": 2.2445, "step": 17972 }, { "epoch": 3.42, "learning_rate": 1.9748522402287894e-06, "loss": 3.0456, "step": 17973 }, { "epoch": 3.42, "learning_rate": 1.9747950428979982e-06, "loss": 2.7549, "step": 17974 }, { "epoch": 3.42, "learning_rate": 1.974737845567207e-06, "loss": 2.0542, "step": 17975 }, { "epoch": 3.42, "learning_rate": 1.974680648236416e-06, "loss": 2.3203, "step": 17976 }, { "epoch": 3.42, "learning_rate": 1.9746234509056246e-06, "loss": 2.6563, "step": 17977 }, { "epoch": 3.42, "learning_rate": 1.974566253574833e-06, "loss": 2.4988, "step": 17978 }, { "epoch": 3.42, "learning_rate": 1.974509056244042e-06, "loss": 2.5666, "step": 17979 }, { "epoch": 3.42, "learning_rate": 1.9744518589132506e-06, "loss": 2.5715, "step": 17980 }, { "epoch": 3.42, "learning_rate": 1.9743946615824594e-06, "loss": 2.4096, "step": 17981 }, { "epoch": 3.43, "learning_rate": 1.9743374642516686e-06, "loss": 1.7307, "step": 17982 }, { "epoch": 3.43, "learning_rate": 1.974280266920877e-06, "loss": 1.646, "step": 17983 }, { "epoch": 3.43, "learning_rate": 1.9742230695900857e-06, "loss": 1.5117, "step": 17984 }, { "epoch": 3.43, "learning_rate": 1.9741658722592945e-06, "loss": 2.999, "step": 17985 }, { "epoch": 3.43, "learning_rate": 1.9741086749285033e-06, "loss": 2.6958, "step": 17986 }, { "epoch": 3.43, "learning_rate": 1.974051477597712e-06, "loss": 2.9422, "step": 17987 }, { "epoch": 3.43, "learning_rate": 1.973994280266921e-06, "loss": 2.1489, "step": 17988 }, { "epoch": 3.43, "learning_rate": 1.9739370829361297e-06, "loss": 2.2439, "step": 17989 }, { "epoch": 3.43, "learning_rate": 1.9738798856053385e-06, "loss": 1.4658, "step": 17990 }, { "epoch": 3.43, "learning_rate": 1.9738226882745473e-06, "loss": 2.397, "step": 17991 }, { "epoch": 3.43, "learning_rate": 1.973765490943756e-06, "loss": 2.1938, "step": 17992 }, { "epoch": 3.43, "learning_rate": 1.973708293612965e-06, "loss": 2.1253, "step": 17993 }, { "epoch": 3.43, "learning_rate": 1.9736510962821737e-06, "loss": 2.0661, "step": 17994 }, { "epoch": 3.43, "learning_rate": 1.973593898951382e-06, "loss": 2.0989, "step": 17995 }, { "epoch": 3.43, "learning_rate": 1.9735367016205913e-06, "loss": 2.1817, "step": 17996 }, { "epoch": 3.43, "learning_rate": 1.9734795042897996e-06, "loss": 2.7919, "step": 17997 }, { "epoch": 3.43, "learning_rate": 1.9734223069590084e-06, "loss": 1.6395, "step": 17998 }, { "epoch": 3.43, "learning_rate": 1.9733651096282176e-06, "loss": 1.7989, "step": 17999 }, { "epoch": 3.43, "learning_rate": 1.973307912297426e-06, "loss": 3.1555, "step": 18000 }, { "epoch": 3.43, "learning_rate": 1.973250714966635e-06, "loss": 2.0831, "step": 18001 }, { "epoch": 3.43, "learning_rate": 1.9731935176358436e-06, "loss": 2.4985, "step": 18002 }, { "epoch": 3.43, "learning_rate": 1.9731363203050524e-06, "loss": 1.5849, "step": 18003 }, { "epoch": 3.43, "learning_rate": 1.9730791229742616e-06, "loss": 1.7977, "step": 18004 }, { "epoch": 3.43, "learning_rate": 1.97302192564347e-06, "loss": 2.6932, "step": 18005 }, { "epoch": 3.43, "learning_rate": 1.9729647283126788e-06, "loss": 1.8799, "step": 18006 }, { "epoch": 3.43, "learning_rate": 1.9729075309818876e-06, "loss": 2.7072, "step": 18007 }, { "epoch": 3.43, "learning_rate": 1.9728503336510964e-06, "loss": 2.188, "step": 18008 }, { "epoch": 3.43, "learning_rate": 1.972793136320305e-06, "loss": 2.1774, "step": 18009 }, { "epoch": 3.43, "learning_rate": 1.972735938989514e-06, "loss": 1.6332, "step": 18010 }, { "epoch": 3.43, "learning_rate": 1.9726787416587227e-06, "loss": 3.0062, "step": 18011 }, { "epoch": 3.43, "learning_rate": 1.972621544327931e-06, "loss": 2.7823, "step": 18012 }, { "epoch": 3.43, "learning_rate": 1.9725643469971403e-06, "loss": 2.3801, "step": 18013 }, { "epoch": 3.43, "learning_rate": 1.972507149666349e-06, "loss": 2.6983, "step": 18014 }, { "epoch": 3.43, "learning_rate": 1.9724499523355575e-06, "loss": 1.3113, "step": 18015 }, { "epoch": 3.43, "learning_rate": 1.9723927550047667e-06, "loss": 2.4925, "step": 18016 }, { "epoch": 3.43, "learning_rate": 1.972335557673975e-06, "loss": 2.8306, "step": 18017 }, { "epoch": 3.43, "learning_rate": 1.972278360343184e-06, "loss": 2.9019, "step": 18018 }, { "epoch": 3.43, "learning_rate": 1.9722211630123927e-06, "loss": 3.2866, "step": 18019 }, { "epoch": 3.43, "learning_rate": 1.9721639656816015e-06, "loss": 2.6837, "step": 18020 }, { "epoch": 3.43, "learning_rate": 1.9721067683508107e-06, "loss": 2.7347, "step": 18021 }, { "epoch": 3.43, "learning_rate": 1.972049571020019e-06, "loss": 2.7475, "step": 18022 }, { "epoch": 3.43, "learning_rate": 1.971992373689228e-06, "loss": 1.8917, "step": 18023 }, { "epoch": 3.43, "learning_rate": 1.9719351763584366e-06, "loss": 2.7733, "step": 18024 }, { "epoch": 3.43, "learning_rate": 1.9718779790276454e-06, "loss": 2.4118, "step": 18025 }, { "epoch": 3.43, "learning_rate": 1.9718207816968542e-06, "loss": 3.4478, "step": 18026 }, { "epoch": 3.43, "learning_rate": 1.971763584366063e-06, "loss": 2.0213, "step": 18027 }, { "epoch": 3.43, "learning_rate": 1.971706387035272e-06, "loss": 2.419, "step": 18028 }, { "epoch": 3.43, "learning_rate": 1.97164918970448e-06, "loss": 2.3969, "step": 18029 }, { "epoch": 3.43, "learning_rate": 1.9715919923736894e-06, "loss": 2.1218, "step": 18030 }, { "epoch": 3.43, "learning_rate": 1.971534795042898e-06, "loss": 1.9521, "step": 18031 }, { "epoch": 3.43, "learning_rate": 1.9714775977121066e-06, "loss": 2.1979, "step": 18032 }, { "epoch": 3.43, "learning_rate": 1.9714204003813158e-06, "loss": 3.5475, "step": 18033 }, { "epoch": 3.44, "learning_rate": 1.971363203050524e-06, "loss": 2.5251, "step": 18034 }, { "epoch": 3.44, "learning_rate": 1.9713060057197334e-06, "loss": 1.6171, "step": 18035 }, { "epoch": 3.44, "learning_rate": 1.9712488083889417e-06, "loss": 2.4483, "step": 18036 }, { "epoch": 3.44, "learning_rate": 1.9711916110581505e-06, "loss": 2.3476, "step": 18037 }, { "epoch": 3.44, "learning_rate": 1.9711344137273598e-06, "loss": 2.122, "step": 18038 }, { "epoch": 3.44, "learning_rate": 1.971077216396568e-06, "loss": 1.3661, "step": 18039 }, { "epoch": 3.44, "learning_rate": 1.971020019065777e-06, "loss": 1.8851, "step": 18040 }, { "epoch": 3.44, "learning_rate": 1.9709628217349857e-06, "loss": 3.5825, "step": 18041 }, { "epoch": 3.44, "learning_rate": 1.9709056244041945e-06, "loss": 2.4533, "step": 18042 }, { "epoch": 3.44, "learning_rate": 1.9708484270734033e-06, "loss": 1.974, "step": 18043 }, { "epoch": 3.44, "learning_rate": 1.970791229742612e-06, "loss": 2.862, "step": 18044 }, { "epoch": 3.44, "learning_rate": 1.970734032411821e-06, "loss": 2.7863, "step": 18045 }, { "epoch": 3.44, "learning_rate": 1.9706768350810297e-06, "loss": 1.5117, "step": 18046 }, { "epoch": 3.44, "learning_rate": 1.9706196377502385e-06, "loss": 2.7052, "step": 18047 }, { "epoch": 3.44, "learning_rate": 1.9705624404194473e-06, "loss": 2.6364, "step": 18048 }, { "epoch": 3.44, "learning_rate": 1.9705052430886556e-06, "loss": 1.8233, "step": 18049 }, { "epoch": 3.44, "learning_rate": 1.970448045757865e-06, "loss": 1.3103, "step": 18050 }, { "epoch": 3.44, "learning_rate": 1.9703908484270732e-06, "loss": 2.0063, "step": 18051 }, { "epoch": 3.44, "learning_rate": 1.9703336510962824e-06, "loss": 2.2784, "step": 18052 }, { "epoch": 3.44, "learning_rate": 1.9702764537654912e-06, "loss": 1.8938, "step": 18053 }, { "epoch": 3.44, "learning_rate": 1.9702192564346996e-06, "loss": 2.2341, "step": 18054 }, { "epoch": 3.44, "learning_rate": 1.970162059103909e-06, "loss": 2.5997, "step": 18055 }, { "epoch": 3.44, "learning_rate": 1.970104861773117e-06, "loss": 3.0246, "step": 18056 }, { "epoch": 3.44, "learning_rate": 1.970047664442326e-06, "loss": 2.2419, "step": 18057 }, { "epoch": 3.44, "learning_rate": 1.9699904671115348e-06, "loss": 1.5502, "step": 18058 }, { "epoch": 3.44, "learning_rate": 1.9699332697807436e-06, "loss": 1.9902, "step": 18059 }, { "epoch": 3.44, "learning_rate": 1.9698760724499524e-06, "loss": 2.2756, "step": 18060 }, { "epoch": 3.44, "learning_rate": 1.969818875119161e-06, "loss": 2.8012, "step": 18061 }, { "epoch": 3.44, "learning_rate": 1.96976167778837e-06, "loss": 1.6185, "step": 18062 }, { "epoch": 3.44, "learning_rate": 1.9697044804575788e-06, "loss": 2.8779, "step": 18063 }, { "epoch": 3.44, "learning_rate": 1.9696472831267875e-06, "loss": 1.6472, "step": 18064 }, { "epoch": 3.44, "learning_rate": 1.9695900857959963e-06, "loss": 2.6458, "step": 18065 }, { "epoch": 3.44, "learning_rate": 1.9695328884652047e-06, "loss": 1.6808, "step": 18066 }, { "epoch": 3.44, "learning_rate": 1.969475691134414e-06, "loss": 2.6987, "step": 18067 }, { "epoch": 3.44, "learning_rate": 1.9694184938036223e-06, "loss": 2.8658, "step": 18068 }, { "epoch": 3.44, "learning_rate": 1.9693612964728315e-06, "loss": 3.2798, "step": 18069 }, { "epoch": 3.44, "learning_rate": 1.9693040991420403e-06, "loss": 2.6129, "step": 18070 }, { "epoch": 3.44, "learning_rate": 1.9692469018112487e-06, "loss": 2.8997, "step": 18071 }, { "epoch": 3.44, "learning_rate": 1.969189704480458e-06, "loss": 2.8382, "step": 18072 }, { "epoch": 3.44, "learning_rate": 1.9691325071496663e-06, "loss": 1.888, "step": 18073 }, { "epoch": 3.44, "learning_rate": 1.969075309818875e-06, "loss": 1.9751, "step": 18074 }, { "epoch": 3.44, "learning_rate": 1.969018112488084e-06, "loss": 2.4104, "step": 18075 }, { "epoch": 3.44, "learning_rate": 1.9689609151572926e-06, "loss": 3.1367, "step": 18076 }, { "epoch": 3.44, "learning_rate": 1.9689037178265014e-06, "loss": 3.0152, "step": 18077 }, { "epoch": 3.44, "learning_rate": 1.9688465204957102e-06, "loss": 2.0285, "step": 18078 }, { "epoch": 3.44, "learning_rate": 1.968789323164919e-06, "loss": 2.2544, "step": 18079 }, { "epoch": 3.44, "learning_rate": 1.968732125834128e-06, "loss": 2.9334, "step": 18080 }, { "epoch": 3.44, "learning_rate": 1.9686749285033366e-06, "loss": 2.2925, "step": 18081 }, { "epoch": 3.44, "learning_rate": 1.9686177311725454e-06, "loss": 1.4781, "step": 18082 }, { "epoch": 3.44, "learning_rate": 1.9685605338417538e-06, "loss": 3.4191, "step": 18083 }, { "epoch": 3.44, "learning_rate": 1.968503336510963e-06, "loss": 2.6572, "step": 18084 }, { "epoch": 3.44, "learning_rate": 1.9684461391801714e-06, "loss": 2.6822, "step": 18085 }, { "epoch": 3.44, "learning_rate": 1.9683889418493806e-06, "loss": 1.6814, "step": 18086 }, { "epoch": 3.45, "learning_rate": 1.9683317445185894e-06, "loss": 2.5979, "step": 18087 }, { "epoch": 3.45, "learning_rate": 1.9682745471877978e-06, "loss": 2.1385, "step": 18088 }, { "epoch": 3.45, "learning_rate": 1.968217349857007e-06, "loss": 1.9513, "step": 18089 }, { "epoch": 3.45, "learning_rate": 1.9681601525262153e-06, "loss": 3.2118, "step": 18090 }, { "epoch": 3.45, "learning_rate": 1.968102955195424e-06, "loss": 1.4128, "step": 18091 }, { "epoch": 3.45, "learning_rate": 1.9680457578646334e-06, "loss": 1.4931, "step": 18092 }, { "epoch": 3.45, "learning_rate": 1.9679885605338417e-06, "loss": 2.9298, "step": 18093 }, { "epoch": 3.45, "learning_rate": 1.9679313632030505e-06, "loss": 2.8684, "step": 18094 }, { "epoch": 3.45, "learning_rate": 1.9678741658722593e-06, "loss": 2.7572, "step": 18095 }, { "epoch": 3.45, "learning_rate": 1.967816968541468e-06, "loss": 3.0902, "step": 18096 }, { "epoch": 3.45, "learning_rate": 1.967759771210677e-06, "loss": 1.4842, "step": 18097 }, { "epoch": 3.45, "learning_rate": 1.9677025738798857e-06, "loss": 1.4275, "step": 18098 }, { "epoch": 3.45, "learning_rate": 1.9676453765490945e-06, "loss": 2.7992, "step": 18099 }, { "epoch": 3.45, "learning_rate": 1.967588179218303e-06, "loss": 2.241, "step": 18100 }, { "epoch": 3.45, "learning_rate": 1.967530981887512e-06, "loss": 1.8082, "step": 18101 }, { "epoch": 3.45, "learning_rate": 1.967473784556721e-06, "loss": 2.1125, "step": 18102 }, { "epoch": 3.45, "learning_rate": 1.9674165872259297e-06, "loss": 3.0955, "step": 18103 }, { "epoch": 3.45, "learning_rate": 1.9673593898951385e-06, "loss": 2.0278, "step": 18104 }, { "epoch": 3.45, "learning_rate": 1.967302192564347e-06, "loss": 2.3216, "step": 18105 }, { "epoch": 3.45, "learning_rate": 1.967244995233556e-06, "loss": 2.9206, "step": 18106 }, { "epoch": 3.45, "learning_rate": 1.9671877979027644e-06, "loss": 2.0307, "step": 18107 }, { "epoch": 3.45, "learning_rate": 1.967130600571973e-06, "loss": 1.4919, "step": 18108 }, { "epoch": 3.45, "learning_rate": 1.9670734032411824e-06, "loss": 2.1354, "step": 18109 }, { "epoch": 3.45, "learning_rate": 1.967016205910391e-06, "loss": 1.9376, "step": 18110 }, { "epoch": 3.45, "learning_rate": 1.9669590085795996e-06, "loss": 1.7704, "step": 18111 }, { "epoch": 3.45, "learning_rate": 1.9669018112488084e-06, "loss": 1.8961, "step": 18112 }, { "epoch": 3.45, "learning_rate": 1.966844613918017e-06, "loss": 2.8793, "step": 18113 }, { "epoch": 3.45, "learning_rate": 1.966787416587226e-06, "loss": 1.7631, "step": 18114 }, { "epoch": 3.45, "learning_rate": 1.9667302192564348e-06, "loss": 2.1876, "step": 18115 }, { "epoch": 3.45, "learning_rate": 1.9666730219256436e-06, "loss": 3.5175, "step": 18116 }, { "epoch": 3.45, "learning_rate": 1.9666158245948523e-06, "loss": 1.9559, "step": 18117 }, { "epoch": 3.45, "learning_rate": 1.966558627264061e-06, "loss": 1.3055, "step": 18118 }, { "epoch": 3.45, "learning_rate": 1.96650142993327e-06, "loss": 2.8684, "step": 18119 }, { "epoch": 3.45, "learning_rate": 1.9664442326024787e-06, "loss": 1.3253, "step": 18120 }, { "epoch": 3.45, "learning_rate": 1.9663870352716875e-06, "loss": 2.4099, "step": 18121 }, { "epoch": 3.45, "learning_rate": 1.966329837940896e-06, "loss": 2.0933, "step": 18122 }, { "epoch": 3.45, "learning_rate": 1.966272640610105e-06, "loss": 2.7217, "step": 18123 }, { "epoch": 3.45, "learning_rate": 1.9662154432793135e-06, "loss": 2.43, "step": 18124 }, { "epoch": 3.45, "learning_rate": 1.9661582459485223e-06, "loss": 2.3086, "step": 18125 }, { "epoch": 3.45, "learning_rate": 1.9661010486177315e-06, "loss": 2.7937, "step": 18126 }, { "epoch": 3.45, "learning_rate": 1.96604385128694e-06, "loss": 2.5276, "step": 18127 }, { "epoch": 3.45, "learning_rate": 1.9659866539561487e-06, "loss": 2.0023, "step": 18128 }, { "epoch": 3.45, "learning_rate": 1.9659294566253575e-06, "loss": 2.2074, "step": 18129 }, { "epoch": 3.45, "learning_rate": 1.9658722592945662e-06, "loss": 3.0238, "step": 18130 }, { "epoch": 3.45, "learning_rate": 1.965815061963775e-06, "loss": 3.3156, "step": 18131 }, { "epoch": 3.45, "learning_rate": 1.965757864632984e-06, "loss": 3.2671, "step": 18132 }, { "epoch": 3.45, "learning_rate": 1.9657006673021926e-06, "loss": 2.5055, "step": 18133 }, { "epoch": 3.45, "learning_rate": 1.9656434699714014e-06, "loss": 1.0809, "step": 18134 }, { "epoch": 3.45, "learning_rate": 1.9655862726406102e-06, "loss": 2.5056, "step": 18135 }, { "epoch": 3.45, "learning_rate": 1.965529075309819e-06, "loss": 1.6251, "step": 18136 }, { "epoch": 3.45, "learning_rate": 1.965471877979028e-06, "loss": 2.3492, "step": 18137 }, { "epoch": 3.45, "learning_rate": 1.9654146806482366e-06, "loss": 2.2145, "step": 18138 }, { "epoch": 3.46, "learning_rate": 1.965357483317445e-06, "loss": 2.1834, "step": 18139 }, { "epoch": 3.46, "learning_rate": 1.965300285986654e-06, "loss": 1.438, "step": 18140 }, { "epoch": 3.46, "learning_rate": 1.965243088655863e-06, "loss": 1.3684, "step": 18141 }, { "epoch": 3.46, "learning_rate": 1.9651858913250713e-06, "loss": 1.4033, "step": 18142 }, { "epoch": 3.46, "learning_rate": 1.9651286939942806e-06, "loss": 3.0945, "step": 18143 }, { "epoch": 3.46, "learning_rate": 1.965071496663489e-06, "loss": 3.6686, "step": 18144 }, { "epoch": 3.46, "learning_rate": 1.9650142993326977e-06, "loss": 2.5087, "step": 18145 }, { "epoch": 3.46, "learning_rate": 1.9649571020019065e-06, "loss": 1.99, "step": 18146 }, { "epoch": 3.46, "learning_rate": 1.9648999046711153e-06, "loss": 2.9015, "step": 18147 }, { "epoch": 3.46, "learning_rate": 1.964842707340324e-06, "loss": 3.3616, "step": 18148 }, { "epoch": 3.46, "learning_rate": 1.964785510009533e-06, "loss": 2.5392, "step": 18149 }, { "epoch": 3.46, "learning_rate": 1.9647283126787417e-06, "loss": 2.5494, "step": 18150 }, { "epoch": 3.46, "learning_rate": 1.9646711153479505e-06, "loss": 1.569, "step": 18151 }, { "epoch": 3.46, "learning_rate": 1.9646139180171593e-06, "loss": 1.8859, "step": 18152 }, { "epoch": 3.46, "learning_rate": 1.964556720686368e-06, "loss": 3.2059, "step": 18153 }, { "epoch": 3.46, "learning_rate": 1.964499523355577e-06, "loss": 2.6154, "step": 18154 }, { "epoch": 3.46, "learning_rate": 1.9644423260247857e-06, "loss": 2.4948, "step": 18155 }, { "epoch": 3.46, "learning_rate": 1.964385128693994e-06, "loss": 2.274, "step": 18156 }, { "epoch": 3.46, "learning_rate": 1.9643279313632033e-06, "loss": 2.8029, "step": 18157 }, { "epoch": 3.46, "learning_rate": 1.964270734032412e-06, "loss": 2.2875, "step": 18158 }, { "epoch": 3.46, "learning_rate": 1.9642135367016204e-06, "loss": 3.994, "step": 18159 }, { "epoch": 3.46, "learning_rate": 1.9641563393708296e-06, "loss": 2.4787, "step": 18160 }, { "epoch": 3.46, "learning_rate": 1.964099142040038e-06, "loss": 2.4849, "step": 18161 }, { "epoch": 3.46, "learning_rate": 1.964041944709247e-06, "loss": 1.9113, "step": 18162 }, { "epoch": 3.46, "learning_rate": 1.9639847473784556e-06, "loss": 2.3412, "step": 18163 }, { "epoch": 3.46, "learning_rate": 1.9639275500476644e-06, "loss": 1.5799, "step": 18164 }, { "epoch": 3.46, "learning_rate": 1.963870352716873e-06, "loss": 2.2338, "step": 18165 }, { "epoch": 3.46, "learning_rate": 1.963813155386082e-06, "loss": 2.7748, "step": 18166 }, { "epoch": 3.46, "learning_rate": 1.9637559580552908e-06, "loss": 2.208, "step": 18167 }, { "epoch": 3.46, "learning_rate": 1.9636987607244996e-06, "loss": 2.5717, "step": 18168 }, { "epoch": 3.46, "learning_rate": 1.9636415633937084e-06, "loss": 2.3041, "step": 18169 }, { "epoch": 3.46, "learning_rate": 1.963584366062917e-06, "loss": 2.2156, "step": 18170 }, { "epoch": 3.46, "learning_rate": 1.963527168732126e-06, "loss": 1.9598, "step": 18171 }, { "epoch": 3.46, "learning_rate": 1.9634699714013347e-06, "loss": 3.3159, "step": 18172 }, { "epoch": 3.46, "learning_rate": 1.963412774070543e-06, "loss": 2.2086, "step": 18173 }, { "epoch": 3.46, "learning_rate": 1.9633555767397523e-06, "loss": 2.6198, "step": 18174 }, { "epoch": 3.46, "learning_rate": 1.963298379408961e-06, "loss": 2.3632, "step": 18175 }, { "epoch": 3.46, "learning_rate": 1.9632411820781695e-06, "loss": 2.169, "step": 18176 }, { "epoch": 3.46, "learning_rate": 1.9631839847473787e-06, "loss": 2.6884, "step": 18177 }, { "epoch": 3.46, "learning_rate": 1.963126787416587e-06, "loss": 1.7071, "step": 18178 }, { "epoch": 3.46, "learning_rate": 1.963069590085796e-06, "loss": 2.9183, "step": 18179 }, { "epoch": 3.46, "learning_rate": 1.963012392755005e-06, "loss": 2.4137, "step": 18180 }, { "epoch": 3.46, "learning_rate": 1.9629551954242135e-06, "loss": 2.4987, "step": 18181 }, { "epoch": 3.46, "learning_rate": 1.9628979980934227e-06, "loss": 2.6875, "step": 18182 }, { "epoch": 3.46, "learning_rate": 1.962840800762631e-06, "loss": 2.0507, "step": 18183 }, { "epoch": 3.46, "learning_rate": 1.96278360343184e-06, "loss": 2.1083, "step": 18184 }, { "epoch": 3.46, "learning_rate": 1.9627264061010486e-06, "loss": 2.6392, "step": 18185 }, { "epoch": 3.46, "learning_rate": 1.9626692087702574e-06, "loss": 1.2243, "step": 18186 }, { "epoch": 3.46, "learning_rate": 1.9626120114394662e-06, "loss": 2.5156, "step": 18187 }, { "epoch": 3.46, "learning_rate": 1.962554814108675e-06, "loss": 1.489, "step": 18188 }, { "epoch": 3.46, "learning_rate": 1.962497616777884e-06, "loss": 1.7542, "step": 18189 }, { "epoch": 3.46, "learning_rate": 1.9624404194470926e-06, "loss": 2.5349, "step": 18190 }, { "epoch": 3.46, "learning_rate": 1.9623832221163014e-06, "loss": 2.5096, "step": 18191 }, { "epoch": 3.47, "learning_rate": 1.96232602478551e-06, "loss": 2.3949, "step": 18192 }, { "epoch": 3.47, "learning_rate": 1.9622688274547186e-06, "loss": 3.3222, "step": 18193 }, { "epoch": 3.47, "learning_rate": 1.9622116301239278e-06, "loss": 2.9363, "step": 18194 }, { "epoch": 3.47, "learning_rate": 1.962154432793136e-06, "loss": 1.795, "step": 18195 }, { "epoch": 3.47, "learning_rate": 1.962097235462345e-06, "loss": 2.5648, "step": 18196 }, { "epoch": 3.47, "learning_rate": 1.962040038131554e-06, "loss": 2.0717, "step": 18197 }, { "epoch": 3.47, "learning_rate": 1.9619828408007625e-06, "loss": 3.7143, "step": 18198 }, { "epoch": 3.47, "learning_rate": 1.9619256434699717e-06, "loss": 3.2256, "step": 18199 }, { "epoch": 3.47, "learning_rate": 1.96186844613918e-06, "loss": 2.5131, "step": 18200 }, { "epoch": 3.47, "learning_rate": 1.961811248808389e-06, "loss": 2.608, "step": 18201 }, { "epoch": 3.47, "learning_rate": 1.9617540514775977e-06, "loss": 2.7185, "step": 18202 }, { "epoch": 3.47, "learning_rate": 1.9616968541468065e-06, "loss": 1.9392, "step": 18203 }, { "epoch": 3.47, "learning_rate": 1.9616396568160153e-06, "loss": 1.4019, "step": 18204 }, { "epoch": 3.47, "learning_rate": 1.961582459485224e-06, "loss": 2.2515, "step": 18205 }, { "epoch": 3.47, "learning_rate": 1.961525262154433e-06, "loss": 2.3262, "step": 18206 }, { "epoch": 3.47, "learning_rate": 1.9614680648236417e-06, "loss": 3.3044, "step": 18207 }, { "epoch": 3.47, "learning_rate": 1.9614108674928505e-06, "loss": 2.007, "step": 18208 }, { "epoch": 3.47, "learning_rate": 1.9613536701620593e-06, "loss": 1.7402, "step": 18209 }, { "epoch": 3.47, "learning_rate": 1.9612964728312676e-06, "loss": 2.3904, "step": 18210 }, { "epoch": 3.47, "learning_rate": 1.961239275500477e-06, "loss": 2.1219, "step": 18211 }, { "epoch": 3.47, "learning_rate": 1.9611820781696852e-06, "loss": 2.2281, "step": 18212 }, { "epoch": 3.47, "learning_rate": 1.961124880838894e-06, "loss": 2.3415, "step": 18213 }, { "epoch": 3.47, "learning_rate": 1.9610676835081032e-06, "loss": 2.1896, "step": 18214 }, { "epoch": 3.47, "learning_rate": 1.9610104861773116e-06, "loss": 1.6763, "step": 18215 }, { "epoch": 3.47, "learning_rate": 1.960953288846521e-06, "loss": 2.3204, "step": 18216 }, { "epoch": 3.47, "learning_rate": 1.960896091515729e-06, "loss": 2.4481, "step": 18217 }, { "epoch": 3.47, "learning_rate": 1.960838894184938e-06, "loss": 2.2824, "step": 18218 }, { "epoch": 3.47, "learning_rate": 1.9607816968541468e-06, "loss": 2.3587, "step": 18219 }, { "epoch": 3.47, "learning_rate": 1.9607244995233556e-06, "loss": 2.2642, "step": 18220 }, { "epoch": 3.47, "learning_rate": 1.9606673021925644e-06, "loss": 3.0574, "step": 18221 }, { "epoch": 3.47, "learning_rate": 1.960610104861773e-06, "loss": 2.2462, "step": 18222 }, { "epoch": 3.47, "learning_rate": 1.960552907530982e-06, "loss": 2.415, "step": 18223 }, { "epoch": 3.47, "learning_rate": 1.9604957102001907e-06, "loss": 3.1587, "step": 18224 }, { "epoch": 3.47, "learning_rate": 1.9604385128693995e-06, "loss": 2.239, "step": 18225 }, { "epoch": 3.47, "learning_rate": 1.9603813155386083e-06, "loss": 1.7704, "step": 18226 }, { "epoch": 3.47, "learning_rate": 1.9603241182078167e-06, "loss": 2.536, "step": 18227 }, { "epoch": 3.47, "learning_rate": 1.960266920877026e-06, "loss": 2.741, "step": 18228 }, { "epoch": 3.47, "learning_rate": 1.9602097235462347e-06, "loss": 3.5321, "step": 18229 }, { "epoch": 3.47, "learning_rate": 1.960152526215443e-06, "loss": 2.0692, "step": 18230 }, { "epoch": 3.47, "learning_rate": 1.9600953288846523e-06, "loss": 2.4139, "step": 18231 }, { "epoch": 3.47, "learning_rate": 1.9600381315538607e-06, "loss": 1.8533, "step": 18232 }, { "epoch": 3.47, "learning_rate": 1.95998093422307e-06, "loss": 1.7837, "step": 18233 }, { "epoch": 3.47, "learning_rate": 1.9599237368922783e-06, "loss": 2.6673, "step": 18234 }, { "epoch": 3.47, "learning_rate": 1.959866539561487e-06, "loss": 2.4106, "step": 18235 }, { "epoch": 3.47, "learning_rate": 1.9598093422306963e-06, "loss": 2.8274, "step": 18236 }, { "epoch": 3.47, "learning_rate": 1.9597521448999046e-06, "loss": 2.146, "step": 18237 }, { "epoch": 3.47, "learning_rate": 1.9596949475691134e-06, "loss": 2.6331, "step": 18238 }, { "epoch": 3.47, "learning_rate": 1.9596377502383222e-06, "loss": 2.241, "step": 18239 }, { "epoch": 3.47, "learning_rate": 1.959580552907531e-06, "loss": 2.5369, "step": 18240 }, { "epoch": 3.47, "learning_rate": 1.95952335557674e-06, "loss": 2.7745, "step": 18241 }, { "epoch": 3.47, "learning_rate": 1.9594661582459486e-06, "loss": 2.9468, "step": 18242 }, { "epoch": 3.47, "learning_rate": 1.9594089609151574e-06, "loss": 2.0961, "step": 18243 }, { "epoch": 3.48, "learning_rate": 1.9593517635843658e-06, "loss": 2.0089, "step": 18244 }, { "epoch": 3.48, "learning_rate": 1.959294566253575e-06, "loss": 1.4972, "step": 18245 }, { "epoch": 3.48, "learning_rate": 1.9592373689227838e-06, "loss": 2.3957, "step": 18246 }, { "epoch": 3.48, "learning_rate": 1.9591801715919926e-06, "loss": 3.4964, "step": 18247 }, { "epoch": 3.48, "learning_rate": 1.9591229742612014e-06, "loss": 2.1783, "step": 18248 }, { "epoch": 3.48, "learning_rate": 1.9590657769304097e-06, "loss": 2.1657, "step": 18249 }, { "epoch": 3.48, "learning_rate": 1.959008579599619e-06, "loss": 1.3391, "step": 18250 }, { "epoch": 3.48, "learning_rate": 1.9589513822688273e-06, "loss": 3.4526, "step": 18251 }, { "epoch": 3.48, "learning_rate": 1.958894184938036e-06, "loss": 2.114, "step": 18252 }, { "epoch": 3.48, "learning_rate": 1.9588369876072453e-06, "loss": 2.086, "step": 18253 }, { "epoch": 3.48, "learning_rate": 1.9587797902764537e-06, "loss": 2.7248, "step": 18254 }, { "epoch": 3.48, "learning_rate": 1.9587225929456625e-06, "loss": 3.2475, "step": 18255 }, { "epoch": 3.48, "learning_rate": 1.9586653956148713e-06, "loss": 2.2694, "step": 18256 }, { "epoch": 3.48, "learning_rate": 1.95860819828408e-06, "loss": 1.7407, "step": 18257 }, { "epoch": 3.48, "learning_rate": 1.958551000953289e-06, "loss": 2.5191, "step": 18258 }, { "epoch": 3.48, "learning_rate": 1.9584938036224977e-06, "loss": 2.6041, "step": 18259 }, { "epoch": 3.48, "learning_rate": 1.9584366062917065e-06, "loss": 1.1852, "step": 18260 }, { "epoch": 3.48, "learning_rate": 1.958379408960915e-06, "loss": 1.8758, "step": 18261 }, { "epoch": 3.48, "learning_rate": 1.958322211630124e-06, "loss": 3.0334, "step": 18262 }, { "epoch": 3.48, "learning_rate": 1.958265014299333e-06, "loss": 2.128, "step": 18263 }, { "epoch": 3.48, "learning_rate": 1.9582078169685417e-06, "loss": 2.4374, "step": 18264 }, { "epoch": 3.48, "learning_rate": 1.9581506196377504e-06, "loss": 2.4706, "step": 18265 }, { "epoch": 3.48, "learning_rate": 1.958093422306959e-06, "loss": 2.6045, "step": 18266 }, { "epoch": 3.48, "learning_rate": 1.958036224976168e-06, "loss": 3.3433, "step": 18267 }, { "epoch": 3.48, "learning_rate": 1.957979027645377e-06, "loss": 1.8851, "step": 18268 }, { "epoch": 3.48, "learning_rate": 1.957921830314585e-06, "loss": 2.0563, "step": 18269 }, { "epoch": 3.48, "learning_rate": 1.9578646329837944e-06, "loss": 2.1922, "step": 18270 }, { "epoch": 3.48, "learning_rate": 1.9578074356530028e-06, "loss": 2.5412, "step": 18271 }, { "epoch": 3.48, "learning_rate": 1.9577502383222116e-06, "loss": 1.526, "step": 18272 }, { "epoch": 3.48, "learning_rate": 1.9576930409914204e-06, "loss": 1.4042, "step": 18273 }, { "epoch": 3.48, "learning_rate": 1.957635843660629e-06, "loss": 2.55, "step": 18274 }, { "epoch": 3.48, "learning_rate": 1.957578646329838e-06, "loss": 3.0619, "step": 18275 }, { "epoch": 3.48, "learning_rate": 1.9575214489990468e-06, "loss": 2.4405, "step": 18276 }, { "epoch": 3.48, "learning_rate": 1.9574642516682555e-06, "loss": 2.262, "step": 18277 }, { "epoch": 3.48, "learning_rate": 1.9574070543374643e-06, "loss": 2.2062, "step": 18278 }, { "epoch": 3.48, "learning_rate": 1.957349857006673e-06, "loss": 1.8645, "step": 18279 }, { "epoch": 3.48, "learning_rate": 1.957292659675882e-06, "loss": 2.6166, "step": 18280 }, { "epoch": 3.48, "learning_rate": 1.9572354623450907e-06, "loss": 2.8228, "step": 18281 }, { "epoch": 3.48, "learning_rate": 1.9571782650142995e-06, "loss": 1.7456, "step": 18282 }, { "epoch": 3.48, "learning_rate": 1.957121067683508e-06, "loss": 1.8645, "step": 18283 }, { "epoch": 3.48, "learning_rate": 1.957063870352717e-06, "loss": 1.8352, "step": 18284 }, { "epoch": 3.48, "learning_rate": 1.957006673021926e-06, "loss": 2.6981, "step": 18285 }, { "epoch": 3.48, "learning_rate": 1.9569494756911343e-06, "loss": 1.8058, "step": 18286 }, { "epoch": 3.48, "learning_rate": 1.9568922783603435e-06, "loss": 2.8453, "step": 18287 }, { "epoch": 3.48, "learning_rate": 1.956835081029552e-06, "loss": 2.272, "step": 18288 }, { "epoch": 3.48, "learning_rate": 1.9567778836987607e-06, "loss": 2.9059, "step": 18289 }, { "epoch": 3.48, "learning_rate": 1.9567206863679694e-06, "loss": 3.2369, "step": 18290 }, { "epoch": 3.48, "learning_rate": 1.9566634890371782e-06, "loss": 1.7596, "step": 18291 }, { "epoch": 3.48, "learning_rate": 1.956606291706387e-06, "loss": 2.8399, "step": 18292 }, { "epoch": 3.48, "learning_rate": 1.956549094375596e-06, "loss": 1.2481, "step": 18293 }, { "epoch": 3.48, "learning_rate": 1.9564918970448046e-06, "loss": 2.6083, "step": 18294 }, { "epoch": 3.48, "learning_rate": 1.9564346997140134e-06, "loss": 2.313, "step": 18295 }, { "epoch": 3.48, "learning_rate": 1.956377502383222e-06, "loss": 1.88, "step": 18296 }, { "epoch": 3.49, "learning_rate": 1.956320305052431e-06, "loss": 2.8155, "step": 18297 }, { "epoch": 3.49, "learning_rate": 1.95626310772164e-06, "loss": 2.021, "step": 18298 }, { "epoch": 3.49, "learning_rate": 1.9562059103908486e-06, "loss": 1.6553, "step": 18299 }, { "epoch": 3.49, "learning_rate": 1.956148713060057e-06, "loss": 2.723, "step": 18300 }, { "epoch": 3.49, "learning_rate": 1.956091515729266e-06, "loss": 2.717, "step": 18301 }, { "epoch": 3.49, "learning_rate": 1.956034318398475e-06, "loss": 2.857, "step": 18302 }, { "epoch": 3.49, "learning_rate": 1.9559771210676833e-06, "loss": 2.8012, "step": 18303 }, { "epoch": 3.49, "learning_rate": 1.9559199237368926e-06, "loss": 2.7264, "step": 18304 }, { "epoch": 3.49, "learning_rate": 1.955862726406101e-06, "loss": 2.8915, "step": 18305 }, { "epoch": 3.49, "learning_rate": 1.9558055290753097e-06, "loss": 3.0049, "step": 18306 }, { "epoch": 3.49, "learning_rate": 1.9557483317445185e-06, "loss": 1.4607, "step": 18307 }, { "epoch": 3.49, "learning_rate": 1.9556911344137273e-06, "loss": 2.4083, "step": 18308 }, { "epoch": 3.49, "learning_rate": 1.955633937082936e-06, "loss": 2.2658, "step": 18309 }, { "epoch": 3.49, "learning_rate": 1.955576739752145e-06, "loss": 2.6829, "step": 18310 }, { "epoch": 3.49, "learning_rate": 1.9555195424213537e-06, "loss": 2.7874, "step": 18311 }, { "epoch": 3.49, "learning_rate": 1.9554623450905625e-06, "loss": 1.4117, "step": 18312 }, { "epoch": 3.49, "learning_rate": 1.9554051477597713e-06, "loss": 2.7313, "step": 18313 }, { "epoch": 3.49, "learning_rate": 1.95534795042898e-06, "loss": 1.5131, "step": 18314 }, { "epoch": 3.49, "learning_rate": 1.955290753098189e-06, "loss": 2.749, "step": 18315 }, { "epoch": 3.49, "learning_rate": 1.9552335557673977e-06, "loss": 1.9973, "step": 18316 }, { "epoch": 3.49, "learning_rate": 1.9551763584366065e-06, "loss": 2.9579, "step": 18317 }, { "epoch": 3.49, "learning_rate": 1.9551191611058152e-06, "loss": 2.5782, "step": 18318 }, { "epoch": 3.49, "learning_rate": 1.955061963775024e-06, "loss": 2.1544, "step": 18319 }, { "epoch": 3.49, "learning_rate": 1.9550047664442324e-06, "loss": 4.0573, "step": 18320 }, { "epoch": 3.49, "learning_rate": 1.9549475691134416e-06, "loss": 1.4138, "step": 18321 }, { "epoch": 3.49, "learning_rate": 1.95489037178265e-06, "loss": 3.3243, "step": 18322 }, { "epoch": 3.49, "learning_rate": 1.954833174451859e-06, "loss": 2.1283, "step": 18323 }, { "epoch": 3.49, "learning_rate": 1.954775977121068e-06, "loss": 2.3979, "step": 18324 }, { "epoch": 3.49, "learning_rate": 1.9547187797902764e-06, "loss": 2.8438, "step": 18325 }, { "epoch": 3.49, "learning_rate": 1.954661582459485e-06, "loss": 3.1125, "step": 18326 }, { "epoch": 3.49, "learning_rate": 1.954604385128694e-06, "loss": 2.7486, "step": 18327 }, { "epoch": 3.49, "learning_rate": 1.9545471877979028e-06, "loss": 2.4325, "step": 18328 }, { "epoch": 3.49, "learning_rate": 1.9544899904671116e-06, "loss": 3.025, "step": 18329 }, { "epoch": 3.49, "learning_rate": 1.9544327931363203e-06, "loss": 2.3138, "step": 18330 }, { "epoch": 3.49, "learning_rate": 1.954375595805529e-06, "loss": 1.4023, "step": 18331 }, { "epoch": 3.49, "learning_rate": 1.954318398474738e-06, "loss": 2.7107, "step": 18332 }, { "epoch": 3.49, "learning_rate": 1.9542612011439467e-06, "loss": 2.2704, "step": 18333 }, { "epoch": 3.49, "learning_rate": 1.9542040038131555e-06, "loss": 2.4474, "step": 18334 }, { "epoch": 3.49, "learning_rate": 1.9541468064823643e-06, "loss": 3.2506, "step": 18335 }, { "epoch": 3.49, "learning_rate": 1.954089609151573e-06, "loss": 2.2891, "step": 18336 }, { "epoch": 3.49, "learning_rate": 1.9540324118207815e-06, "loss": 2.1757, "step": 18337 }, { "epoch": 3.49, "learning_rate": 1.9539752144899907e-06, "loss": 2.988, "step": 18338 }, { "epoch": 3.49, "learning_rate": 1.953918017159199e-06, "loss": 2.8075, "step": 18339 }, { "epoch": 3.49, "learning_rate": 1.953860819828408e-06, "loss": 1.364, "step": 18340 }, { "epoch": 3.49, "learning_rate": 1.953803622497617e-06, "loss": 2.7865, "step": 18341 }, { "epoch": 3.49, "learning_rate": 1.9537464251668255e-06, "loss": 2.0962, "step": 18342 }, { "epoch": 3.49, "learning_rate": 1.9536892278360342e-06, "loss": 2.4579, "step": 18343 }, { "epoch": 3.49, "learning_rate": 1.953632030505243e-06, "loss": 1.638, "step": 18344 }, { "epoch": 3.49, "learning_rate": 1.953574833174452e-06, "loss": 1.4774, "step": 18345 }, { "epoch": 3.49, "learning_rate": 1.9535176358436606e-06, "loss": 2.3153, "step": 18346 }, { "epoch": 3.49, "learning_rate": 1.9534604385128694e-06, "loss": 2.8036, "step": 18347 }, { "epoch": 3.49, "learning_rate": 1.9534032411820782e-06, "loss": 2.8916, "step": 18348 }, { "epoch": 3.5, "learning_rate": 1.953346043851287e-06, "loss": 1.4195, "step": 18349 }, { "epoch": 3.5, "learning_rate": 1.953288846520496e-06, "loss": 3.5015, "step": 18350 }, { "epoch": 3.5, "learning_rate": 1.9532316491897046e-06, "loss": 2.3466, "step": 18351 }, { "epoch": 3.5, "learning_rate": 1.9531744518589134e-06, "loss": 2.7344, "step": 18352 }, { "epoch": 3.5, "learning_rate": 1.953117254528122e-06, "loss": 2.6926, "step": 18353 }, { "epoch": 3.5, "learning_rate": 1.9530600571973306e-06, "loss": 2.5257, "step": 18354 }, { "epoch": 3.5, "learning_rate": 1.9530028598665398e-06, "loss": 2.9893, "step": 18355 }, { "epoch": 3.5, "learning_rate": 1.9529456625357486e-06, "loss": 2.2933, "step": 18356 }, { "epoch": 3.5, "learning_rate": 1.952888465204957e-06, "loss": 1.9384, "step": 18357 }, { "epoch": 3.5, "learning_rate": 1.952831267874166e-06, "loss": 1.9928, "step": 18358 }, { "epoch": 3.5, "learning_rate": 1.9527740705433745e-06, "loss": 2.1319, "step": 18359 }, { "epoch": 3.5, "learning_rate": 1.9527168732125833e-06, "loss": 2.5762, "step": 18360 }, { "epoch": 3.5, "learning_rate": 1.952659675881792e-06, "loss": 3.3286, "step": 18361 }, { "epoch": 3.5, "learning_rate": 1.952602478551001e-06, "loss": 2.298, "step": 18362 }, { "epoch": 3.5, "learning_rate": 1.95254528122021e-06, "loss": 2.951, "step": 18363 }, { "epoch": 3.5, "learning_rate": 1.9524880838894185e-06, "loss": 2.0142, "step": 18364 }, { "epoch": 3.5, "learning_rate": 1.9524308865586273e-06, "loss": 2.3987, "step": 18365 }, { "epoch": 3.5, "learning_rate": 1.952373689227836e-06, "loss": 4.0538, "step": 18366 }, { "epoch": 3.5, "learning_rate": 1.952316491897045e-06, "loss": 1.8893, "step": 18367 }, { "epoch": 3.5, "learning_rate": 1.9522592945662537e-06, "loss": 1.524, "step": 18368 }, { "epoch": 3.5, "learning_rate": 1.9522020972354625e-06, "loss": 1.4912, "step": 18369 }, { "epoch": 3.5, "learning_rate": 1.9521448999046713e-06, "loss": 2.9155, "step": 18370 }, { "epoch": 3.5, "learning_rate": 1.9520877025738796e-06, "loss": 2.731, "step": 18371 }, { "epoch": 3.5, "learning_rate": 1.952030505243089e-06, "loss": 1.826, "step": 18372 }, { "epoch": 3.5, "learning_rate": 1.9519733079122976e-06, "loss": 2.7515, "step": 18373 }, { "epoch": 3.5, "learning_rate": 1.951916110581506e-06, "loss": 2.859, "step": 18374 }, { "epoch": 3.5, "learning_rate": 1.9518589132507152e-06, "loss": 2.1559, "step": 18375 }, { "epoch": 3.5, "learning_rate": 1.9518017159199236e-06, "loss": 2.8456, "step": 18376 }, { "epoch": 3.5, "learning_rate": 1.951744518589133e-06, "loss": 2.0453, "step": 18377 }, { "epoch": 3.5, "learning_rate": 1.951687321258341e-06, "loss": 2.196, "step": 18378 }, { "epoch": 3.5, "learning_rate": 1.95163012392755e-06, "loss": 2.7292, "step": 18379 }, { "epoch": 3.5, "learning_rate": 1.951572926596759e-06, "loss": 3.0739, "step": 18380 }, { "epoch": 3.5, "learning_rate": 1.9515157292659676e-06, "loss": 2.4464, "step": 18381 }, { "epoch": 3.5, "learning_rate": 1.9514585319351764e-06, "loss": 2.4445, "step": 18382 }, { "epoch": 3.5, "learning_rate": 1.951401334604385e-06, "loss": 1.8677, "step": 18383 }, { "epoch": 3.5, "learning_rate": 1.951344137273594e-06, "loss": 2.6693, "step": 18384 }, { "epoch": 3.5, "learning_rate": 1.9512869399428027e-06, "loss": 1.3286, "step": 18385 }, { "epoch": 3.5, "learning_rate": 1.9512297426120115e-06, "loss": 2.146, "step": 18386 }, { "epoch": 3.5, "learning_rate": 1.9511725452812203e-06, "loss": 3.4538, "step": 18387 }, { "epoch": 3.5, "learning_rate": 1.9511153479504287e-06, "loss": 2.0492, "step": 18388 }, { "epoch": 3.5, "learning_rate": 1.951058150619638e-06, "loss": 2.3665, "step": 18389 }, { "epoch": 3.5, "learning_rate": 1.9510009532888467e-06, "loss": 3.353, "step": 18390 }, { "epoch": 3.5, "learning_rate": 1.950943755958055e-06, "loss": 2.5509, "step": 18391 }, { "epoch": 3.5, "learning_rate": 1.9508865586272643e-06, "loss": 2.4325, "step": 18392 }, { "epoch": 3.5, "learning_rate": 1.9508293612964727e-06, "loss": 1.8532, "step": 18393 }, { "epoch": 3.5, "learning_rate": 1.950772163965682e-06, "loss": 2.1228, "step": 18394 }, { "epoch": 3.5, "learning_rate": 1.9507149666348903e-06, "loss": 1.8407, "step": 18395 }, { "epoch": 3.5, "learning_rate": 1.950657769304099e-06, "loss": 1.8817, "step": 18396 }, { "epoch": 3.5, "learning_rate": 1.9506005719733083e-06, "loss": 1.9631, "step": 18397 }, { "epoch": 3.5, "learning_rate": 1.9505433746425166e-06, "loss": 2.9096, "step": 18398 }, { "epoch": 3.5, "learning_rate": 1.9504861773117254e-06, "loss": 2.0032, "step": 18399 }, { "epoch": 3.5, "learning_rate": 1.9504289799809342e-06, "loss": 2.1622, "step": 18400 }, { "epoch": 3.5, "learning_rate": 1.950371782650143e-06, "loss": 2.1874, "step": 18401 }, { "epoch": 3.51, "learning_rate": 1.950314585319352e-06, "loss": 2.3167, "step": 18402 }, { "epoch": 3.51, "learning_rate": 1.9502573879885606e-06, "loss": 2.5935, "step": 18403 }, { "epoch": 3.51, "learning_rate": 1.9502001906577694e-06, "loss": 2.1322, "step": 18404 }, { "epoch": 3.51, "learning_rate": 1.950142993326978e-06, "loss": 2.4103, "step": 18405 }, { "epoch": 3.51, "learning_rate": 1.950085795996187e-06, "loss": 1.7683, "step": 18406 }, { "epoch": 3.51, "learning_rate": 1.9500285986653958e-06, "loss": 2.8267, "step": 18407 }, { "epoch": 3.51, "learning_rate": 1.949971401334604e-06, "loss": 2.0505, "step": 18408 }, { "epoch": 3.51, "learning_rate": 1.9499142040038134e-06, "loss": 3.7247, "step": 18409 }, { "epoch": 3.51, "learning_rate": 1.9498570066730217e-06, "loss": 2.5345, "step": 18410 }, { "epoch": 3.51, "learning_rate": 1.949799809342231e-06, "loss": 2.7319, "step": 18411 }, { "epoch": 3.51, "learning_rate": 1.9497426120114397e-06, "loss": 2.8403, "step": 18412 }, { "epoch": 3.51, "learning_rate": 1.949685414680648e-06, "loss": 1.7328, "step": 18413 }, { "epoch": 3.51, "learning_rate": 1.9496282173498573e-06, "loss": 2.7519, "step": 18414 }, { "epoch": 3.51, "learning_rate": 1.9495710200190657e-06, "loss": 2.668, "step": 18415 }, { "epoch": 3.51, "learning_rate": 1.9495138226882745e-06, "loss": 1.9983, "step": 18416 }, { "epoch": 3.51, "learning_rate": 1.9494566253574833e-06, "loss": 3.5084, "step": 18417 }, { "epoch": 3.51, "learning_rate": 1.949399428026692e-06, "loss": 2.4948, "step": 18418 }, { "epoch": 3.51, "learning_rate": 1.949342230695901e-06, "loss": 2.5326, "step": 18419 }, { "epoch": 3.51, "learning_rate": 1.9492850333651097e-06, "loss": 2.2803, "step": 18420 }, { "epoch": 3.51, "learning_rate": 1.9492278360343185e-06, "loss": 3.0067, "step": 18421 }, { "epoch": 3.51, "learning_rate": 1.9491706387035273e-06, "loss": 0.9706, "step": 18422 }, { "epoch": 3.51, "learning_rate": 1.949113441372736e-06, "loss": 2.2537, "step": 18423 }, { "epoch": 3.51, "learning_rate": 1.949056244041945e-06, "loss": 3.562, "step": 18424 }, { "epoch": 3.51, "learning_rate": 1.9489990467111532e-06, "loss": 0.9429, "step": 18425 }, { "epoch": 3.51, "learning_rate": 1.9489418493803624e-06, "loss": 2.3429, "step": 18426 }, { "epoch": 3.51, "learning_rate": 1.948884652049571e-06, "loss": 1.4918, "step": 18427 }, { "epoch": 3.51, "learning_rate": 1.94882745471878e-06, "loss": 2.0734, "step": 18428 }, { "epoch": 3.51, "learning_rate": 1.948770257387989e-06, "loss": 3.2619, "step": 18429 }, { "epoch": 3.51, "learning_rate": 1.948713060057197e-06, "loss": 2.3528, "step": 18430 }, { "epoch": 3.51, "learning_rate": 1.9486558627264064e-06, "loss": 1.412, "step": 18431 }, { "epoch": 3.51, "learning_rate": 1.9485986653956148e-06, "loss": 2.187, "step": 18432 }, { "epoch": 3.51, "learning_rate": 1.9485414680648236e-06, "loss": 2.0665, "step": 18433 }, { "epoch": 3.51, "learning_rate": 1.9484842707340324e-06, "loss": 1.6229, "step": 18434 }, { "epoch": 3.51, "learning_rate": 1.948427073403241e-06, "loss": 1.1594, "step": 18435 }, { "epoch": 3.51, "learning_rate": 1.94836987607245e-06, "loss": 1.4854, "step": 18436 }, { "epoch": 3.51, "learning_rate": 1.9483126787416587e-06, "loss": 2.0401, "step": 18437 }, { "epoch": 3.51, "learning_rate": 1.9482554814108675e-06, "loss": 3.4968, "step": 18438 }, { "epoch": 3.51, "learning_rate": 1.9481982840800763e-06, "loss": 2.9651, "step": 18439 }, { "epoch": 3.51, "learning_rate": 1.948141086749285e-06, "loss": 2.5577, "step": 18440 }, { "epoch": 3.51, "learning_rate": 1.948083889418494e-06, "loss": 3.6612, "step": 18441 }, { "epoch": 3.51, "learning_rate": 1.9480266920877027e-06, "loss": 2.4698, "step": 18442 }, { "epoch": 3.51, "learning_rate": 1.9479694947569115e-06, "loss": 2.7451, "step": 18443 }, { "epoch": 3.51, "learning_rate": 1.9479122974261203e-06, "loss": 2.4199, "step": 18444 }, { "epoch": 3.51, "learning_rate": 1.947855100095329e-06, "loss": 2.1654, "step": 18445 }, { "epoch": 3.51, "learning_rate": 1.947797902764538e-06, "loss": 3.0786, "step": 18446 }, { "epoch": 3.51, "learning_rate": 1.9477407054337463e-06, "loss": 1.4311, "step": 18447 }, { "epoch": 3.51, "learning_rate": 1.9476835081029555e-06, "loss": 1.7208, "step": 18448 }, { "epoch": 3.51, "learning_rate": 1.947626310772164e-06, "loss": 2.6964, "step": 18449 }, { "epoch": 3.51, "learning_rate": 1.9475691134413726e-06, "loss": 1.763, "step": 18450 }, { "epoch": 3.51, "learning_rate": 1.947511916110582e-06, "loss": 2.9564, "step": 18451 }, { "epoch": 3.51, "learning_rate": 1.9474547187797902e-06, "loss": 2.2549, "step": 18452 }, { "epoch": 3.51, "learning_rate": 1.947397521448999e-06, "loss": 3.4688, "step": 18453 }, { "epoch": 3.52, "learning_rate": 1.947340324118208e-06, "loss": 3.2433, "step": 18454 }, { "epoch": 3.52, "learning_rate": 1.9472831267874166e-06, "loss": 2.2047, "step": 18455 }, { "epoch": 3.52, "learning_rate": 1.9472259294566254e-06, "loss": 2.3901, "step": 18456 }, { "epoch": 3.52, "learning_rate": 1.947168732125834e-06, "loss": 1.7313, "step": 18457 }, { "epoch": 3.52, "learning_rate": 1.947111534795043e-06, "loss": 2.7523, "step": 18458 }, { "epoch": 3.52, "learning_rate": 1.9470543374642518e-06, "loss": 2.0125, "step": 18459 }, { "epoch": 3.52, "learning_rate": 1.9469971401334606e-06, "loss": 1.4864, "step": 18460 }, { "epoch": 3.52, "learning_rate": 1.9469399428026694e-06, "loss": 3.1458, "step": 18461 }, { "epoch": 3.52, "learning_rate": 1.946882745471878e-06, "loss": 2.0797, "step": 18462 }, { "epoch": 3.52, "learning_rate": 1.946825548141087e-06, "loss": 2.8406, "step": 18463 }, { "epoch": 3.52, "learning_rate": 1.9467683508102953e-06, "loss": 1.7266, "step": 18464 }, { "epoch": 3.52, "learning_rate": 1.9467111534795046e-06, "loss": 1.8744, "step": 18465 }, { "epoch": 3.52, "learning_rate": 1.946653956148713e-06, "loss": 2.6078, "step": 18466 }, { "epoch": 3.52, "learning_rate": 1.9465967588179217e-06, "loss": 1.5996, "step": 18467 }, { "epoch": 3.52, "learning_rate": 1.946539561487131e-06, "loss": 2.2556, "step": 18468 }, { "epoch": 3.52, "learning_rate": 1.9464823641563393e-06, "loss": 2.7679, "step": 18469 }, { "epoch": 3.52, "learning_rate": 1.946425166825548e-06, "loss": 1.8033, "step": 18470 }, { "epoch": 3.52, "learning_rate": 1.946367969494757e-06, "loss": 4.2306, "step": 18471 }, { "epoch": 3.52, "learning_rate": 1.9463107721639657e-06, "loss": 3.2029, "step": 18472 }, { "epoch": 3.52, "learning_rate": 1.9462535748331745e-06, "loss": 2.3618, "step": 18473 }, { "epoch": 3.52, "learning_rate": 1.9461963775023833e-06, "loss": 3.1847, "step": 18474 }, { "epoch": 3.52, "learning_rate": 1.946139180171592e-06, "loss": 1.7467, "step": 18475 }, { "epoch": 3.52, "learning_rate": 1.946081982840801e-06, "loss": 1.6707, "step": 18476 }, { "epoch": 3.52, "learning_rate": 1.9460247855100097e-06, "loss": 1.1332, "step": 18477 }, { "epoch": 3.52, "learning_rate": 1.9459675881792184e-06, "loss": 1.7726, "step": 18478 }, { "epoch": 3.52, "learning_rate": 1.9459103908484272e-06, "loss": 1.6565, "step": 18479 }, { "epoch": 3.52, "learning_rate": 1.945853193517636e-06, "loss": 2.6772, "step": 18480 }, { "epoch": 3.52, "learning_rate": 1.9457959961868444e-06, "loss": 2.2397, "step": 18481 }, { "epoch": 3.52, "learning_rate": 1.9457387988560536e-06, "loss": 2.0646, "step": 18482 }, { "epoch": 3.52, "learning_rate": 1.945681601525262e-06, "loss": 3.0331, "step": 18483 }, { "epoch": 3.52, "learning_rate": 1.9456244041944708e-06, "loss": 2.0651, "step": 18484 }, { "epoch": 3.52, "learning_rate": 1.94556720686368e-06, "loss": 2.2174, "step": 18485 }, { "epoch": 3.52, "learning_rate": 1.9455100095328884e-06, "loss": 2.7054, "step": 18486 }, { "epoch": 3.52, "learning_rate": 1.945452812202097e-06, "loss": 1.9487, "step": 18487 }, { "epoch": 3.52, "learning_rate": 1.945395614871306e-06, "loss": 2.2807, "step": 18488 }, { "epoch": 3.52, "learning_rate": 1.9453384175405148e-06, "loss": 2.4357, "step": 18489 }, { "epoch": 3.52, "learning_rate": 1.9452812202097235e-06, "loss": 1.9848, "step": 18490 }, { "epoch": 3.52, "learning_rate": 1.9452240228789323e-06, "loss": 1.9418, "step": 18491 }, { "epoch": 3.52, "learning_rate": 1.945166825548141e-06, "loss": 1.5863, "step": 18492 }, { "epoch": 3.52, "learning_rate": 1.94510962821735e-06, "loss": 2.5763, "step": 18493 }, { "epoch": 3.52, "learning_rate": 1.9450524308865587e-06, "loss": 2.5412, "step": 18494 }, { "epoch": 3.52, "learning_rate": 1.9449952335557675e-06, "loss": 2.2294, "step": 18495 }, { "epoch": 3.52, "learning_rate": 1.9449380362249763e-06, "loss": 1.5023, "step": 18496 }, { "epoch": 3.52, "learning_rate": 1.944880838894185e-06, "loss": 2.4704, "step": 18497 }, { "epoch": 3.52, "learning_rate": 1.9448236415633935e-06, "loss": 1.6322, "step": 18498 }, { "epoch": 3.52, "learning_rate": 1.9447664442326027e-06, "loss": 1.8483, "step": 18499 }, { "epoch": 3.52, "learning_rate": 1.9447092469018115e-06, "loss": 1.7443, "step": 18500 }, { "epoch": 3.52, "learning_rate": 1.94465204957102e-06, "loss": 2.461, "step": 18501 }, { "epoch": 3.52, "learning_rate": 1.944594852240229e-06, "loss": 3.9865, "step": 18502 }, { "epoch": 3.52, "learning_rate": 1.9445376549094374e-06, "loss": 2.9102, "step": 18503 }, { "epoch": 3.52, "learning_rate": 1.9444804575786462e-06, "loss": 2.3886, "step": 18504 }, { "epoch": 3.52, "learning_rate": 1.944423260247855e-06, "loss": 2.3626, "step": 18505 }, { "epoch": 3.52, "learning_rate": 1.944366062917064e-06, "loss": 1.7924, "step": 18506 }, { "epoch": 3.53, "learning_rate": 1.944308865586273e-06, "loss": 2.2583, "step": 18507 }, { "epoch": 3.53, "learning_rate": 1.9442516682554814e-06, "loss": 1.8893, "step": 18508 }, { "epoch": 3.53, "learning_rate": 1.94419447092469e-06, "loss": 2.4665, "step": 18509 }, { "epoch": 3.53, "learning_rate": 1.944137273593899e-06, "loss": 2.7834, "step": 18510 }, { "epoch": 3.53, "learning_rate": 1.944080076263108e-06, "loss": 2.7725, "step": 18511 }, { "epoch": 3.53, "learning_rate": 1.9440228789323166e-06, "loss": 2.2737, "step": 18512 }, { "epoch": 3.53, "learning_rate": 1.9439656816015254e-06, "loss": 2.0553, "step": 18513 }, { "epoch": 3.53, "learning_rate": 1.943908484270734e-06, "loss": 1.3268, "step": 18514 }, { "epoch": 3.53, "learning_rate": 1.9438512869399425e-06, "loss": 2.0309, "step": 18515 }, { "epoch": 3.53, "learning_rate": 1.9437940896091518e-06, "loss": 2.2173, "step": 18516 }, { "epoch": 3.53, "learning_rate": 1.9437368922783606e-06, "loss": 2.4471, "step": 18517 }, { "epoch": 3.53, "learning_rate": 1.943679694947569e-06, "loss": 1.862, "step": 18518 }, { "epoch": 3.53, "learning_rate": 1.943622497616778e-06, "loss": 2.3659, "step": 18519 }, { "epoch": 3.53, "learning_rate": 1.9435653002859865e-06, "loss": 1.9491, "step": 18520 }, { "epoch": 3.53, "learning_rate": 1.9435081029551953e-06, "loss": 1.628, "step": 18521 }, { "epoch": 3.53, "learning_rate": 1.943450905624404e-06, "loss": 1.323, "step": 18522 }, { "epoch": 3.53, "learning_rate": 1.943393708293613e-06, "loss": 2.5855, "step": 18523 }, { "epoch": 3.53, "learning_rate": 1.943336510962822e-06, "loss": 1.3486, "step": 18524 }, { "epoch": 3.53, "learning_rate": 1.9432793136320305e-06, "loss": 1.9535, "step": 18525 }, { "epoch": 3.53, "learning_rate": 1.9432221163012393e-06, "loss": 1.7551, "step": 18526 }, { "epoch": 3.53, "learning_rate": 1.943164918970448e-06, "loss": 2.257, "step": 18527 }, { "epoch": 3.53, "learning_rate": 1.943107721639657e-06, "loss": 1.3516, "step": 18528 }, { "epoch": 3.53, "learning_rate": 1.9430505243088657e-06, "loss": 1.3813, "step": 18529 }, { "epoch": 3.53, "learning_rate": 1.9429933269780745e-06, "loss": 3.2069, "step": 18530 }, { "epoch": 3.53, "learning_rate": 1.9429361296472832e-06, "loss": 2.5039, "step": 18531 }, { "epoch": 3.53, "learning_rate": 1.9428789323164916e-06, "loss": 2.6372, "step": 18532 }, { "epoch": 3.53, "learning_rate": 1.942821734985701e-06, "loss": 1.5146, "step": 18533 }, { "epoch": 3.53, "learning_rate": 1.9427645376549096e-06, "loss": 3.1495, "step": 18534 }, { "epoch": 3.53, "learning_rate": 1.942707340324118e-06, "loss": 2.8738, "step": 18535 }, { "epoch": 3.53, "learning_rate": 1.9426501429933272e-06, "loss": 2.8784, "step": 18536 }, { "epoch": 3.53, "learning_rate": 1.9425929456625356e-06, "loss": 2.9915, "step": 18537 }, { "epoch": 3.53, "learning_rate": 1.9425357483317444e-06, "loss": 2.7032, "step": 18538 }, { "epoch": 3.53, "learning_rate": 1.9424785510009536e-06, "loss": 2.8225, "step": 18539 }, { "epoch": 3.53, "learning_rate": 1.942421353670162e-06, "loss": 2.9517, "step": 18540 }, { "epoch": 3.53, "learning_rate": 1.942364156339371e-06, "loss": 1.8563, "step": 18541 }, { "epoch": 3.53, "learning_rate": 1.9423069590085796e-06, "loss": 3.2756, "step": 18542 }, { "epoch": 3.53, "learning_rate": 1.9422497616777884e-06, "loss": 2.1403, "step": 18543 }, { "epoch": 3.53, "learning_rate": 1.942192564346997e-06, "loss": 2.6882, "step": 18544 }, { "epoch": 3.53, "learning_rate": 1.942135367016206e-06, "loss": 1.9251, "step": 18545 }, { "epoch": 3.53, "learning_rate": 1.9420781696854147e-06, "loss": 2.1831, "step": 18546 }, { "epoch": 3.53, "learning_rate": 1.9420209723546235e-06, "loss": 2.2705, "step": 18547 }, { "epoch": 3.53, "learning_rate": 1.9419637750238323e-06, "loss": 2.5861, "step": 18548 }, { "epoch": 3.53, "learning_rate": 1.941906577693041e-06, "loss": 3.0432, "step": 18549 }, { "epoch": 3.53, "learning_rate": 1.94184938036225e-06, "loss": 1.9109, "step": 18550 }, { "epoch": 3.53, "learning_rate": 1.9417921830314587e-06, "loss": 3.5845, "step": 18551 }, { "epoch": 3.53, "learning_rate": 1.941734985700667e-06, "loss": 3.6819, "step": 18552 }, { "epoch": 3.53, "learning_rate": 1.9416777883698763e-06, "loss": 1.8542, "step": 18553 }, { "epoch": 3.53, "learning_rate": 1.9416205910390847e-06, "loss": 1.4673, "step": 18554 }, { "epoch": 3.53, "learning_rate": 1.9415633937082935e-06, "loss": 3.2172, "step": 18555 }, { "epoch": 3.53, "learning_rate": 1.9415061963775027e-06, "loss": 2.1783, "step": 18556 }, { "epoch": 3.53, "learning_rate": 1.941448999046711e-06, "loss": 2.5658, "step": 18557 }, { "epoch": 3.53, "learning_rate": 1.9413918017159203e-06, "loss": 1.469, "step": 18558 }, { "epoch": 3.54, "learning_rate": 1.9413346043851286e-06, "loss": 2.1945, "step": 18559 }, { "epoch": 3.54, "learning_rate": 1.9412774070543374e-06, "loss": 2.4078, "step": 18560 }, { "epoch": 3.54, "learning_rate": 1.9412202097235462e-06, "loss": 3.0517, "step": 18561 }, { "epoch": 3.54, "learning_rate": 1.941163012392755e-06, "loss": 2.7791, "step": 18562 }, { "epoch": 3.54, "learning_rate": 1.941105815061964e-06, "loss": 2.3054, "step": 18563 }, { "epoch": 3.54, "learning_rate": 1.9410486177311726e-06, "loss": 2.0439, "step": 18564 }, { "epoch": 3.54, "learning_rate": 1.9409914204003814e-06, "loss": 2.0749, "step": 18565 }, { "epoch": 3.54, "learning_rate": 1.94093422306959e-06, "loss": 2.8161, "step": 18566 }, { "epoch": 3.54, "learning_rate": 1.940877025738799e-06, "loss": 2.1017, "step": 18567 }, { "epoch": 3.54, "learning_rate": 1.9408198284080078e-06, "loss": 2.674, "step": 18568 }, { "epoch": 3.54, "learning_rate": 1.940762631077216e-06, "loss": 2.3641, "step": 18569 }, { "epoch": 3.54, "learning_rate": 1.9407054337464254e-06, "loss": 3.1927, "step": 18570 }, { "epoch": 3.54, "learning_rate": 1.9406482364156337e-06, "loss": 2.2436, "step": 18571 }, { "epoch": 3.54, "learning_rate": 1.940591039084843e-06, "loss": 1.9914, "step": 18572 }, { "epoch": 3.54, "learning_rate": 1.9405338417540517e-06, "loss": 2.4534, "step": 18573 }, { "epoch": 3.54, "learning_rate": 1.94047664442326e-06, "loss": 2.6202, "step": 18574 }, { "epoch": 3.54, "learning_rate": 1.9404194470924693e-06, "loss": 1.7989, "step": 18575 }, { "epoch": 3.54, "learning_rate": 1.9403622497616777e-06, "loss": 2.9723, "step": 18576 }, { "epoch": 3.54, "learning_rate": 1.9403050524308865e-06, "loss": 3.1308, "step": 18577 }, { "epoch": 3.54, "learning_rate": 1.9402478551000957e-06, "loss": 3.1165, "step": 18578 }, { "epoch": 3.54, "learning_rate": 1.940190657769304e-06, "loss": 1.4121, "step": 18579 }, { "epoch": 3.54, "learning_rate": 1.940133460438513e-06, "loss": 1.3908, "step": 18580 }, { "epoch": 3.54, "learning_rate": 1.9400762631077217e-06, "loss": 2.488, "step": 18581 }, { "epoch": 3.54, "learning_rate": 1.9400190657769305e-06, "loss": 2.5625, "step": 18582 }, { "epoch": 3.54, "learning_rate": 1.9399618684461393e-06, "loss": 2.2074, "step": 18583 }, { "epoch": 3.54, "learning_rate": 1.939904671115348e-06, "loss": 2.7011, "step": 18584 }, { "epoch": 3.54, "learning_rate": 1.939847473784557e-06, "loss": 3.1372, "step": 18585 }, { "epoch": 3.54, "learning_rate": 1.9397902764537652e-06, "loss": 2.922, "step": 18586 }, { "epoch": 3.54, "learning_rate": 1.9397330791229744e-06, "loss": 2.4909, "step": 18587 }, { "epoch": 3.54, "learning_rate": 1.9396758817921832e-06, "loss": 1.8433, "step": 18588 }, { "epoch": 3.54, "learning_rate": 1.939618684461392e-06, "loss": 1.6364, "step": 18589 }, { "epoch": 3.54, "learning_rate": 1.939561487130601e-06, "loss": 2.7142, "step": 18590 }, { "epoch": 3.54, "learning_rate": 1.939504289799809e-06, "loss": 2.2012, "step": 18591 }, { "epoch": 3.54, "learning_rate": 1.9394470924690184e-06, "loss": 2.1331, "step": 18592 }, { "epoch": 3.54, "learning_rate": 1.9393898951382268e-06, "loss": 2.7156, "step": 18593 }, { "epoch": 3.54, "learning_rate": 1.9393326978074356e-06, "loss": 1.2939, "step": 18594 }, { "epoch": 3.54, "learning_rate": 1.9392755004766448e-06, "loss": 3.3559, "step": 18595 }, { "epoch": 3.54, "learning_rate": 1.939218303145853e-06, "loss": 3.4736, "step": 18596 }, { "epoch": 3.54, "learning_rate": 1.939161105815062e-06, "loss": 1.9078, "step": 18597 }, { "epoch": 3.54, "learning_rate": 1.9391039084842707e-06, "loss": 1.867, "step": 18598 }, { "epoch": 3.54, "learning_rate": 1.9390467111534795e-06, "loss": 1.9797, "step": 18599 }, { "epoch": 3.54, "learning_rate": 1.9389895138226883e-06, "loss": 1.71, "step": 18600 }, { "epoch": 3.54, "learning_rate": 1.938932316491897e-06, "loss": 2.1044, "step": 18601 }, { "epoch": 3.54, "learning_rate": 1.938875119161106e-06, "loss": 1.8465, "step": 18602 }, { "epoch": 3.54, "learning_rate": 1.9388179218303143e-06, "loss": 2.1665, "step": 18603 }, { "epoch": 3.54, "learning_rate": 1.9387607244995235e-06, "loss": 2.4731, "step": 18604 }, { "epoch": 3.54, "learning_rate": 1.9387035271687323e-06, "loss": 3.6933, "step": 18605 }, { "epoch": 3.54, "learning_rate": 1.938646329837941e-06, "loss": 2.1114, "step": 18606 }, { "epoch": 3.54, "learning_rate": 1.93858913250715e-06, "loss": 2.9132, "step": 18607 }, { "epoch": 3.54, "learning_rate": 1.9385319351763583e-06, "loss": 1.2745, "step": 18608 }, { "epoch": 3.54, "learning_rate": 1.9384747378455675e-06, "loss": 2.7682, "step": 18609 }, { "epoch": 3.54, "learning_rate": 1.938417540514776e-06, "loss": 2.085, "step": 18610 }, { "epoch": 3.54, "learning_rate": 1.9383603431839846e-06, "loss": 2.9056, "step": 18611 }, { "epoch": 3.55, "learning_rate": 1.938303145853194e-06, "loss": 2.2836, "step": 18612 }, { "epoch": 3.55, "learning_rate": 1.9382459485224022e-06, "loss": 2.4409, "step": 18613 }, { "epoch": 3.55, "learning_rate": 1.938188751191611e-06, "loss": 2.0569, "step": 18614 }, { "epoch": 3.55, "learning_rate": 1.93813155386082e-06, "loss": 2.449, "step": 18615 }, { "epoch": 3.55, "learning_rate": 1.9380743565300286e-06, "loss": 1.5416, "step": 18616 }, { "epoch": 3.55, "learning_rate": 1.9380171591992374e-06, "loss": 2.2889, "step": 18617 }, { "epoch": 3.55, "learning_rate": 1.937959961868446e-06, "loss": 2.1157, "step": 18618 }, { "epoch": 3.55, "learning_rate": 1.937902764537655e-06, "loss": 3.035, "step": 18619 }, { "epoch": 3.55, "learning_rate": 1.9378455672068634e-06, "loss": 2.0942, "step": 18620 }, { "epoch": 3.55, "learning_rate": 1.9377883698760726e-06, "loss": 2.0849, "step": 18621 }, { "epoch": 3.55, "learning_rate": 1.9377311725452814e-06, "loss": 2.3232, "step": 18622 }, { "epoch": 3.55, "learning_rate": 1.93767397521449e-06, "loss": 1.3511, "step": 18623 }, { "epoch": 3.55, "learning_rate": 1.937616777883699e-06, "loss": 2.0488, "step": 18624 }, { "epoch": 3.55, "learning_rate": 1.9375595805529073e-06, "loss": 2.7392, "step": 18625 }, { "epoch": 3.55, "learning_rate": 1.9375023832221165e-06, "loss": 2.2931, "step": 18626 }, { "epoch": 3.55, "learning_rate": 1.9374451858913253e-06, "loss": 1.2737, "step": 18627 }, { "epoch": 3.55, "learning_rate": 1.9373879885605337e-06, "loss": 2.2586, "step": 18628 }, { "epoch": 3.55, "learning_rate": 1.937330791229743e-06, "loss": 2.3417, "step": 18629 }, { "epoch": 3.55, "learning_rate": 1.9372735938989513e-06, "loss": 1.356, "step": 18630 }, { "epoch": 3.55, "learning_rate": 1.93721639656816e-06, "loss": 3.1212, "step": 18631 }, { "epoch": 3.55, "learning_rate": 1.937159199237369e-06, "loss": 1.2754, "step": 18632 }, { "epoch": 3.55, "learning_rate": 1.9371020019065777e-06, "loss": 2.64, "step": 18633 }, { "epoch": 3.55, "learning_rate": 1.9370448045757865e-06, "loss": 3.0246, "step": 18634 }, { "epoch": 3.55, "learning_rate": 1.9369876072449953e-06, "loss": 2.7771, "step": 18635 }, { "epoch": 3.55, "learning_rate": 1.936930409914204e-06, "loss": 2.2686, "step": 18636 }, { "epoch": 3.55, "learning_rate": 1.936873212583413e-06, "loss": 4.4445, "step": 18637 }, { "epoch": 3.55, "learning_rate": 1.9368160152526216e-06, "loss": 1.9869, "step": 18638 }, { "epoch": 3.55, "learning_rate": 1.9367588179218304e-06, "loss": 3.6491, "step": 18639 }, { "epoch": 3.55, "learning_rate": 1.9367016205910392e-06, "loss": 1.5845, "step": 18640 }, { "epoch": 3.55, "learning_rate": 1.936644423260248e-06, "loss": 1.4193, "step": 18641 }, { "epoch": 3.55, "learning_rate": 1.9365872259294564e-06, "loss": 2.5417, "step": 18642 }, { "epoch": 3.55, "learning_rate": 1.9365300285986656e-06, "loss": 1.7157, "step": 18643 }, { "epoch": 3.55, "learning_rate": 1.9364728312678744e-06, "loss": 2.2382, "step": 18644 }, { "epoch": 3.55, "learning_rate": 1.9364156339370828e-06, "loss": 2.3854, "step": 18645 }, { "epoch": 3.55, "learning_rate": 1.936358436606292e-06, "loss": 2.8218, "step": 18646 }, { "epoch": 3.55, "learning_rate": 1.9363012392755004e-06, "loss": 2.7577, "step": 18647 }, { "epoch": 3.55, "learning_rate": 1.936244041944709e-06, "loss": 2.8198, "step": 18648 }, { "epoch": 3.55, "learning_rate": 1.936186844613918e-06, "loss": 2.7348, "step": 18649 }, { "epoch": 3.55, "learning_rate": 1.9361296472831267e-06, "loss": 2.2258, "step": 18650 }, { "epoch": 3.55, "learning_rate": 1.9360724499523355e-06, "loss": 2.3825, "step": 18651 }, { "epoch": 3.55, "learning_rate": 1.9360152526215443e-06, "loss": 0.8726, "step": 18652 }, { "epoch": 3.55, "learning_rate": 1.935958055290753e-06, "loss": 1.8105, "step": 18653 }, { "epoch": 3.55, "learning_rate": 1.935900857959962e-06, "loss": 3.2738, "step": 18654 }, { "epoch": 3.55, "learning_rate": 1.9358436606291707e-06, "loss": 2.7208, "step": 18655 }, { "epoch": 3.55, "learning_rate": 1.9357864632983795e-06, "loss": 2.1503, "step": 18656 }, { "epoch": 3.55, "learning_rate": 1.9357292659675883e-06, "loss": 2.5909, "step": 18657 }, { "epoch": 3.55, "learning_rate": 1.935672068636797e-06, "loss": 2.2234, "step": 18658 }, { "epoch": 3.55, "learning_rate": 1.9356148713060055e-06, "loss": 1.228, "step": 18659 }, { "epoch": 3.55, "learning_rate": 1.9355576739752147e-06, "loss": 2.3402, "step": 18660 }, { "epoch": 3.55, "learning_rate": 1.9355004766444235e-06, "loss": 2.6373, "step": 18661 }, { "epoch": 3.55, "learning_rate": 1.935443279313632e-06, "loss": 2.3441, "step": 18662 }, { "epoch": 3.55, "learning_rate": 1.935386081982841e-06, "loss": 2.9004, "step": 18663 }, { "epoch": 3.56, "learning_rate": 1.9353288846520494e-06, "loss": 2.9026, "step": 18664 }, { "epoch": 3.56, "learning_rate": 1.9352716873212582e-06, "loss": 1.132, "step": 18665 }, { "epoch": 3.56, "learning_rate": 1.9352144899904674e-06, "loss": 2.4627, "step": 18666 }, { "epoch": 3.56, "learning_rate": 1.935157292659676e-06, "loss": 1.5092, "step": 18667 }, { "epoch": 3.56, "learning_rate": 1.9351000953288846e-06, "loss": 3.0359, "step": 18668 }, { "epoch": 3.56, "learning_rate": 1.9350428979980934e-06, "loss": 3.1991, "step": 18669 }, { "epoch": 3.56, "learning_rate": 1.934985700667302e-06, "loss": 3.6857, "step": 18670 }, { "epoch": 3.56, "learning_rate": 1.934928503336511e-06, "loss": 1.9793, "step": 18671 }, { "epoch": 3.56, "learning_rate": 1.93487130600572e-06, "loss": 2.1056, "step": 18672 }, { "epoch": 3.56, "learning_rate": 1.9348141086749286e-06, "loss": 3.3847, "step": 18673 }, { "epoch": 3.56, "learning_rate": 1.9347569113441374e-06, "loss": 2.77, "step": 18674 }, { "epoch": 3.56, "learning_rate": 1.934699714013346e-06, "loss": 3.36, "step": 18675 }, { "epoch": 3.56, "learning_rate": 1.934642516682555e-06, "loss": 2.3284, "step": 18676 }, { "epoch": 3.56, "learning_rate": 1.9345853193517638e-06, "loss": 2.0915, "step": 18677 }, { "epoch": 3.56, "learning_rate": 1.9345281220209726e-06, "loss": 2.6316, "step": 18678 }, { "epoch": 3.56, "learning_rate": 1.934470924690181e-06, "loss": 1.7509, "step": 18679 }, { "epoch": 3.56, "learning_rate": 1.93441372735939e-06, "loss": 2.9626, "step": 18680 }, { "epoch": 3.56, "learning_rate": 1.9343565300285985e-06, "loss": 2.8479, "step": 18681 }, { "epoch": 3.56, "learning_rate": 1.9342993326978073e-06, "loss": 2.3765, "step": 18682 }, { "epoch": 3.56, "learning_rate": 1.9342421353670165e-06, "loss": 2.0859, "step": 18683 }, { "epoch": 3.56, "learning_rate": 1.934184938036225e-06, "loss": 2.6633, "step": 18684 }, { "epoch": 3.56, "learning_rate": 1.9341277407054337e-06, "loss": 2.7503, "step": 18685 }, { "epoch": 3.56, "learning_rate": 1.9340705433746425e-06, "loss": 1.623, "step": 18686 }, { "epoch": 3.56, "learning_rate": 1.9340133460438513e-06, "loss": 1.5865, "step": 18687 }, { "epoch": 3.56, "learning_rate": 1.93395614871306e-06, "loss": 1.736, "step": 18688 }, { "epoch": 3.56, "learning_rate": 1.933898951382269e-06, "loss": 3.1971, "step": 18689 }, { "epoch": 3.56, "learning_rate": 1.9338417540514777e-06, "loss": 1.7968, "step": 18690 }, { "epoch": 3.56, "learning_rate": 1.9337845567206864e-06, "loss": 1.509, "step": 18691 }, { "epoch": 3.56, "learning_rate": 1.9337273593898952e-06, "loss": 3.3761, "step": 18692 }, { "epoch": 3.56, "learning_rate": 1.933670162059104e-06, "loss": 2.3587, "step": 18693 }, { "epoch": 3.56, "learning_rate": 1.933612964728313e-06, "loss": 2.7612, "step": 18694 }, { "epoch": 3.56, "learning_rate": 1.9335557673975216e-06, "loss": 2.209, "step": 18695 }, { "epoch": 3.56, "learning_rate": 1.93349857006673e-06, "loss": 2.493, "step": 18696 }, { "epoch": 3.56, "learning_rate": 1.933441372735939e-06, "loss": 4.0358, "step": 18697 }, { "epoch": 3.56, "learning_rate": 1.9333841754051476e-06, "loss": 1.5979, "step": 18698 }, { "epoch": 3.56, "learning_rate": 1.9333269780743564e-06, "loss": 2.4787, "step": 18699 }, { "epoch": 3.56, "learning_rate": 1.9332697807435656e-06, "loss": 1.8713, "step": 18700 }, { "epoch": 3.56, "learning_rate": 1.933212583412774e-06, "loss": 2.4114, "step": 18701 }, { "epoch": 3.56, "learning_rate": 1.933155386081983e-06, "loss": 1.5987, "step": 18702 }, { "epoch": 3.56, "learning_rate": 1.9330981887511915e-06, "loss": 3.4395, "step": 18703 }, { "epoch": 3.56, "learning_rate": 1.9330409914204003e-06, "loss": 3.2446, "step": 18704 }, { "epoch": 3.56, "learning_rate": 1.932983794089609e-06, "loss": 1.8912, "step": 18705 }, { "epoch": 3.56, "learning_rate": 1.932926596758818e-06, "loss": 1.3632, "step": 18706 }, { "epoch": 3.56, "learning_rate": 1.9328693994280267e-06, "loss": 1.628, "step": 18707 }, { "epoch": 3.56, "learning_rate": 1.9328122020972355e-06, "loss": 2.2487, "step": 18708 }, { "epoch": 3.56, "learning_rate": 1.9327550047664443e-06, "loss": 1.3833, "step": 18709 }, { "epoch": 3.56, "learning_rate": 1.932697807435653e-06, "loss": 3.4159, "step": 18710 }, { "epoch": 3.56, "learning_rate": 1.932640610104862e-06, "loss": 2.2157, "step": 18711 }, { "epoch": 3.56, "learning_rate": 1.9325834127740707e-06, "loss": 2.5921, "step": 18712 }, { "epoch": 3.56, "learning_rate": 1.932526215443279e-06, "loss": 2.0643, "step": 18713 }, { "epoch": 3.56, "learning_rate": 1.9324690181124883e-06, "loss": 2.8894, "step": 18714 }, { "epoch": 3.56, "learning_rate": 1.932411820781697e-06, "loss": 1.6639, "step": 18715 }, { "epoch": 3.56, "learning_rate": 1.9323546234509054e-06, "loss": 2.662, "step": 18716 }, { "epoch": 3.57, "learning_rate": 1.9322974261201147e-06, "loss": 2.4989, "step": 18717 }, { "epoch": 3.57, "learning_rate": 1.932240228789323e-06, "loss": 2.9951, "step": 18718 }, { "epoch": 3.57, "learning_rate": 1.9321830314585323e-06, "loss": 2.7117, "step": 18719 }, { "epoch": 3.57, "learning_rate": 1.9321258341277406e-06, "loss": 2.0172, "step": 18720 }, { "epoch": 3.57, "learning_rate": 1.9320686367969494e-06, "loss": 2.3178, "step": 18721 }, { "epoch": 3.57, "learning_rate": 1.9320114394661586e-06, "loss": 2.3092, "step": 18722 }, { "epoch": 3.57, "learning_rate": 1.931954242135367e-06, "loss": 2.6602, "step": 18723 }, { "epoch": 3.57, "learning_rate": 1.931897044804576e-06, "loss": 0.8787, "step": 18724 }, { "epoch": 3.57, "learning_rate": 1.9318398474737846e-06, "loss": 1.8794, "step": 18725 }, { "epoch": 3.57, "learning_rate": 1.9317826501429934e-06, "loss": 3.2123, "step": 18726 }, { "epoch": 3.57, "learning_rate": 1.931725452812202e-06, "loss": 1.7468, "step": 18727 }, { "epoch": 3.57, "learning_rate": 1.931668255481411e-06, "loss": 2.4661, "step": 18728 }, { "epoch": 3.57, "learning_rate": 1.9316110581506198e-06, "loss": 1.2221, "step": 18729 }, { "epoch": 3.57, "learning_rate": 1.931553860819828e-06, "loss": 1.2348, "step": 18730 }, { "epoch": 3.57, "learning_rate": 1.9314966634890374e-06, "loss": 2.2789, "step": 18731 }, { "epoch": 3.57, "learning_rate": 1.931439466158246e-06, "loss": 2.8759, "step": 18732 }, { "epoch": 3.57, "learning_rate": 1.9313822688274545e-06, "loss": 1.9316, "step": 18733 }, { "epoch": 3.57, "learning_rate": 1.9313250714966637e-06, "loss": 3.5359, "step": 18734 }, { "epoch": 3.57, "learning_rate": 1.931267874165872e-06, "loss": 2.8713, "step": 18735 }, { "epoch": 3.57, "learning_rate": 1.9312106768350813e-06, "loss": 2.1986, "step": 18736 }, { "epoch": 3.57, "learning_rate": 1.9311534795042897e-06, "loss": 2.0426, "step": 18737 }, { "epoch": 3.57, "learning_rate": 1.9310962821734985e-06, "loss": 2.8317, "step": 18738 }, { "epoch": 3.57, "learning_rate": 1.9310390848427077e-06, "loss": 1.7091, "step": 18739 }, { "epoch": 3.57, "learning_rate": 1.930981887511916e-06, "loss": 1.6706, "step": 18740 }, { "epoch": 3.57, "learning_rate": 1.930924690181125e-06, "loss": 2.0406, "step": 18741 }, { "epoch": 3.57, "learning_rate": 1.9308674928503337e-06, "loss": 1.8508, "step": 18742 }, { "epoch": 3.57, "learning_rate": 1.9308102955195425e-06, "loss": 2.1251, "step": 18743 }, { "epoch": 3.57, "learning_rate": 1.9307530981887512e-06, "loss": 2.3068, "step": 18744 }, { "epoch": 3.57, "learning_rate": 1.93069590085796e-06, "loss": 2.3891, "step": 18745 }, { "epoch": 3.57, "learning_rate": 1.930638703527169e-06, "loss": 2.719, "step": 18746 }, { "epoch": 3.57, "learning_rate": 1.930581506196377e-06, "loss": 2.0093, "step": 18747 }, { "epoch": 3.57, "learning_rate": 1.9305243088655864e-06, "loss": 2.5907, "step": 18748 }, { "epoch": 3.57, "learning_rate": 1.9304671115347952e-06, "loss": 1.8631, "step": 18749 }, { "epoch": 3.57, "learning_rate": 1.9304099142040036e-06, "loss": 2.297, "step": 18750 }, { "epoch": 3.57, "learning_rate": 1.930352716873213e-06, "loss": 1.2713, "step": 18751 }, { "epoch": 3.57, "learning_rate": 1.930295519542421e-06, "loss": 2.1426, "step": 18752 }, { "epoch": 3.57, "learning_rate": 1.9302383222116304e-06, "loss": 2.9025, "step": 18753 }, { "epoch": 3.57, "learning_rate": 1.930181124880839e-06, "loss": 2.6963, "step": 18754 }, { "epoch": 3.57, "learning_rate": 1.9301239275500476e-06, "loss": 2.589, "step": 18755 }, { "epoch": 3.57, "learning_rate": 1.9300667302192568e-06, "loss": 2.3598, "step": 18756 }, { "epoch": 3.57, "learning_rate": 1.930009532888465e-06, "loss": 2.1314, "step": 18757 }, { "epoch": 3.57, "learning_rate": 1.929952335557674e-06, "loss": 3.432, "step": 18758 }, { "epoch": 3.57, "learning_rate": 1.9298951382268827e-06, "loss": 2.1293, "step": 18759 }, { "epoch": 3.57, "learning_rate": 1.9298379408960915e-06, "loss": 2.8617, "step": 18760 }, { "epoch": 3.57, "learning_rate": 1.9297807435653003e-06, "loss": 2.6476, "step": 18761 }, { "epoch": 3.57, "learning_rate": 1.929723546234509e-06, "loss": 2.7236, "step": 18762 }, { "epoch": 3.57, "learning_rate": 1.929666348903718e-06, "loss": 2.7685, "step": 18763 }, { "epoch": 3.57, "learning_rate": 1.9296091515729267e-06, "loss": 2.2492, "step": 18764 }, { "epoch": 3.57, "learning_rate": 1.9295519542421355e-06, "loss": 3.1448, "step": 18765 }, { "epoch": 3.57, "learning_rate": 1.9294947569113443e-06, "loss": 2.8213, "step": 18766 }, { "epoch": 3.57, "learning_rate": 1.9294375595805527e-06, "loss": 2.8384, "step": 18767 }, { "epoch": 3.57, "learning_rate": 1.929380362249762e-06, "loss": 2.9951, "step": 18768 }, { "epoch": 3.58, "learning_rate": 1.9293231649189702e-06, "loss": 2.3265, "step": 18769 }, { "epoch": 3.58, "learning_rate": 1.9292659675881795e-06, "loss": 3.0512, "step": 18770 }, { "epoch": 3.58, "learning_rate": 1.9292087702573883e-06, "loss": 1.9937, "step": 18771 }, { "epoch": 3.58, "learning_rate": 1.9291515729265966e-06, "loss": 2.5182, "step": 18772 }, { "epoch": 3.58, "learning_rate": 1.929094375595806e-06, "loss": 2.407, "step": 18773 }, { "epoch": 3.58, "learning_rate": 1.9290371782650142e-06, "loss": 2.5717, "step": 18774 }, { "epoch": 3.58, "learning_rate": 1.928979980934223e-06, "loss": 2.956, "step": 18775 }, { "epoch": 3.58, "learning_rate": 1.928922783603432e-06, "loss": 2.3936, "step": 18776 }, { "epoch": 3.58, "learning_rate": 1.9288655862726406e-06, "loss": 1.5704, "step": 18777 }, { "epoch": 3.58, "learning_rate": 1.9288083889418494e-06, "loss": 1.6148, "step": 18778 }, { "epoch": 3.58, "learning_rate": 1.928751191611058e-06, "loss": 4.0488, "step": 18779 }, { "epoch": 3.58, "learning_rate": 1.928693994280267e-06, "loss": 2.3652, "step": 18780 }, { "epoch": 3.58, "learning_rate": 1.9286367969494758e-06, "loss": 3.3084, "step": 18781 }, { "epoch": 3.58, "learning_rate": 1.9285795996186846e-06, "loss": 1.7132, "step": 18782 }, { "epoch": 3.58, "learning_rate": 1.9285224022878934e-06, "loss": 2.2522, "step": 18783 }, { "epoch": 3.58, "learning_rate": 1.928465204957102e-06, "loss": 1.8274, "step": 18784 }, { "epoch": 3.58, "learning_rate": 1.928408007626311e-06, "loss": 3.3161, "step": 18785 }, { "epoch": 3.58, "learning_rate": 1.9283508102955193e-06, "loss": 2.7295, "step": 18786 }, { "epoch": 3.58, "learning_rate": 1.9282936129647285e-06, "loss": 2.5104, "step": 18787 }, { "epoch": 3.58, "learning_rate": 1.9282364156339373e-06, "loss": 3.3803, "step": 18788 }, { "epoch": 3.58, "learning_rate": 1.9281792183031457e-06, "loss": 2.0728, "step": 18789 }, { "epoch": 3.58, "learning_rate": 1.928122020972355e-06, "loss": 2.8809, "step": 18790 }, { "epoch": 3.58, "learning_rate": 1.9280648236415633e-06, "loss": 1.7337, "step": 18791 }, { "epoch": 3.58, "learning_rate": 1.928007626310772e-06, "loss": 1.986, "step": 18792 }, { "epoch": 3.58, "learning_rate": 1.927950428979981e-06, "loss": 3.4674, "step": 18793 }, { "epoch": 3.58, "learning_rate": 1.9278932316491897e-06, "loss": 2.9423, "step": 18794 }, { "epoch": 3.58, "learning_rate": 1.9278360343183985e-06, "loss": 3.6739, "step": 18795 }, { "epoch": 3.58, "learning_rate": 1.9277788369876073e-06, "loss": 3.1384, "step": 18796 }, { "epoch": 3.58, "learning_rate": 1.927721639656816e-06, "loss": 1.7504, "step": 18797 }, { "epoch": 3.58, "learning_rate": 1.927664442326025e-06, "loss": 2.5805, "step": 18798 }, { "epoch": 3.58, "learning_rate": 1.9276072449952336e-06, "loss": 2.3591, "step": 18799 }, { "epoch": 3.58, "learning_rate": 1.9275500476644424e-06, "loss": 2.0069, "step": 18800 }, { "epoch": 3.58, "learning_rate": 1.9274928503336512e-06, "loss": 1.3099, "step": 18801 }, { "epoch": 3.58, "learning_rate": 1.92743565300286e-06, "loss": 2.6189, "step": 18802 }, { "epoch": 3.58, "learning_rate": 1.927378455672069e-06, "loss": 3.3938, "step": 18803 }, { "epoch": 3.58, "learning_rate": 1.9273212583412776e-06, "loss": 3.2368, "step": 18804 }, { "epoch": 3.58, "learning_rate": 1.9272640610104864e-06, "loss": 2.5293, "step": 18805 }, { "epoch": 3.58, "learning_rate": 1.9272068636796948e-06, "loss": 2.1459, "step": 18806 }, { "epoch": 3.58, "learning_rate": 1.927149666348904e-06, "loss": 3.0005, "step": 18807 }, { "epoch": 3.58, "learning_rate": 1.9270924690181124e-06, "loss": 2.3317, "step": 18808 }, { "epoch": 3.58, "learning_rate": 1.927035271687321e-06, "loss": 1.2841, "step": 18809 }, { "epoch": 3.58, "learning_rate": 1.9269780743565304e-06, "loss": 3.0672, "step": 18810 }, { "epoch": 3.58, "learning_rate": 1.9269208770257387e-06, "loss": 2.4714, "step": 18811 }, { "epoch": 3.58, "learning_rate": 1.9268636796949475e-06, "loss": 2.103, "step": 18812 }, { "epoch": 3.58, "learning_rate": 1.9268064823641563e-06, "loss": 2.566, "step": 18813 }, { "epoch": 3.58, "learning_rate": 1.926749285033365e-06, "loss": 2.7993, "step": 18814 }, { "epoch": 3.58, "learning_rate": 1.926692087702574e-06, "loss": 1.8705, "step": 18815 }, { "epoch": 3.58, "learning_rate": 1.9266348903717827e-06, "loss": 1.7699, "step": 18816 }, { "epoch": 3.58, "learning_rate": 1.9265776930409915e-06, "loss": 3.0048, "step": 18817 }, { "epoch": 3.58, "learning_rate": 1.9265204957102003e-06, "loss": 2.1239, "step": 18818 }, { "epoch": 3.58, "learning_rate": 1.926463298379409e-06, "loss": 2.504, "step": 18819 }, { "epoch": 3.58, "learning_rate": 1.926406101048618e-06, "loss": 1.8205, "step": 18820 }, { "epoch": 3.58, "learning_rate": 1.9263489037178267e-06, "loss": 2.2869, "step": 18821 }, { "epoch": 3.59, "learning_rate": 1.9262917063870355e-06, "loss": 2.5217, "step": 18822 }, { "epoch": 3.59, "learning_rate": 1.926234509056244e-06, "loss": 2.4812, "step": 18823 }, { "epoch": 3.59, "learning_rate": 1.926177311725453e-06, "loss": 1.9721, "step": 18824 }, { "epoch": 3.59, "learning_rate": 1.9261201143946614e-06, "loss": 2.4194, "step": 18825 }, { "epoch": 3.59, "learning_rate": 1.9260629170638702e-06, "loss": 2.8833, "step": 18826 }, { "epoch": 3.59, "learning_rate": 1.9260057197330794e-06, "loss": 3.5655, "step": 18827 }, { "epoch": 3.59, "learning_rate": 1.925948522402288e-06, "loss": 2.496, "step": 18828 }, { "epoch": 3.59, "learning_rate": 1.9258913250714966e-06, "loss": 2.2835, "step": 18829 }, { "epoch": 3.59, "learning_rate": 1.9258341277407054e-06, "loss": 2.1266, "step": 18830 }, { "epoch": 3.59, "learning_rate": 1.925776930409914e-06, "loss": 2.2541, "step": 18831 }, { "epoch": 3.59, "learning_rate": 1.925719733079123e-06, "loss": 2.6013, "step": 18832 }, { "epoch": 3.59, "learning_rate": 1.9256625357483318e-06, "loss": 2.2679, "step": 18833 }, { "epoch": 3.59, "learning_rate": 1.9256053384175406e-06, "loss": 1.8752, "step": 18834 }, { "epoch": 3.59, "learning_rate": 1.9255481410867494e-06, "loss": 2.2856, "step": 18835 }, { "epoch": 3.59, "learning_rate": 1.925490943755958e-06, "loss": 2.4918, "step": 18836 }, { "epoch": 3.59, "learning_rate": 1.925433746425167e-06, "loss": 3.1154, "step": 18837 }, { "epoch": 3.59, "learning_rate": 1.9253765490943758e-06, "loss": 3.1911, "step": 18838 }, { "epoch": 3.59, "learning_rate": 1.9253193517635845e-06, "loss": 2.8364, "step": 18839 }, { "epoch": 3.59, "learning_rate": 1.925262154432793e-06, "loss": 1.769, "step": 18840 }, { "epoch": 3.59, "learning_rate": 1.925204957102002e-06, "loss": 1.8059, "step": 18841 }, { "epoch": 3.59, "learning_rate": 1.9251477597712105e-06, "loss": 2.1624, "step": 18842 }, { "epoch": 3.59, "learning_rate": 1.9250905624404193e-06, "loss": 2.116, "step": 18843 }, { "epoch": 3.59, "learning_rate": 1.9250333651096285e-06, "loss": 2.4497, "step": 18844 }, { "epoch": 3.59, "learning_rate": 1.924976167778837e-06, "loss": 1.279, "step": 18845 }, { "epoch": 3.59, "learning_rate": 1.9249189704480457e-06, "loss": 2.0904, "step": 18846 }, { "epoch": 3.59, "learning_rate": 1.9248617731172545e-06, "loss": 3.1216, "step": 18847 }, { "epoch": 3.59, "learning_rate": 1.9248045757864633e-06, "loss": 2.0898, "step": 18848 }, { "epoch": 3.59, "learning_rate": 1.9247473784556725e-06, "loss": 1.9606, "step": 18849 }, { "epoch": 3.59, "learning_rate": 1.924690181124881e-06, "loss": 1.9726, "step": 18850 }, { "epoch": 3.59, "learning_rate": 1.9246329837940896e-06, "loss": 2.1116, "step": 18851 }, { "epoch": 3.59, "learning_rate": 1.9245757864632984e-06, "loss": 2.1511, "step": 18852 }, { "epoch": 3.59, "learning_rate": 1.9245185891325072e-06, "loss": 2.4692, "step": 18853 }, { "epoch": 3.59, "learning_rate": 1.924461391801716e-06, "loss": 1.8067, "step": 18854 }, { "epoch": 3.59, "learning_rate": 1.924404194470925e-06, "loss": 2.6251, "step": 18855 }, { "epoch": 3.59, "learning_rate": 1.9243469971401336e-06, "loss": 2.6363, "step": 18856 }, { "epoch": 3.59, "learning_rate": 1.924289799809342e-06, "loss": 2.2773, "step": 18857 }, { "epoch": 3.59, "learning_rate": 1.924232602478551e-06, "loss": 2.1687, "step": 18858 }, { "epoch": 3.59, "learning_rate": 1.92417540514776e-06, "loss": 2.2588, "step": 18859 }, { "epoch": 3.59, "learning_rate": 1.9241182078169684e-06, "loss": 1.851, "step": 18860 }, { "epoch": 3.59, "learning_rate": 1.9240610104861776e-06, "loss": 0.9148, "step": 18861 }, { "epoch": 3.59, "learning_rate": 1.924003813155386e-06, "loss": 2.5176, "step": 18862 }, { "epoch": 3.59, "learning_rate": 1.9239466158245947e-06, "loss": 1.226, "step": 18863 }, { "epoch": 3.59, "learning_rate": 1.9238894184938035e-06, "loss": 1.4959, "step": 18864 }, { "epoch": 3.59, "learning_rate": 1.9238322211630123e-06, "loss": 2.9653, "step": 18865 }, { "epoch": 3.59, "learning_rate": 1.9237750238322216e-06, "loss": 2.1492, "step": 18866 }, { "epoch": 3.59, "learning_rate": 1.92371782650143e-06, "loss": 1.7929, "step": 18867 }, { "epoch": 3.59, "learning_rate": 1.9236606291706387e-06, "loss": 2.5079, "step": 18868 }, { "epoch": 3.59, "learning_rate": 1.9236034318398475e-06, "loss": 1.1245, "step": 18869 }, { "epoch": 3.59, "learning_rate": 1.9235462345090563e-06, "loss": 2.3621, "step": 18870 }, { "epoch": 3.59, "learning_rate": 1.923489037178265e-06, "loss": 1.9432, "step": 18871 }, { "epoch": 3.59, "learning_rate": 1.923431839847474e-06, "loss": 2.5569, "step": 18872 }, { "epoch": 3.59, "learning_rate": 1.9233746425166827e-06, "loss": 3.6393, "step": 18873 }, { "epoch": 3.6, "learning_rate": 1.923317445185891e-06, "loss": 2.06, "step": 18874 }, { "epoch": 3.6, "learning_rate": 1.9232602478551003e-06, "loss": 2.0173, "step": 18875 }, { "epoch": 3.6, "learning_rate": 1.923203050524309e-06, "loss": 4.018, "step": 18876 }, { "epoch": 3.6, "learning_rate": 1.9231458531935174e-06, "loss": 1.9677, "step": 18877 }, { "epoch": 3.6, "learning_rate": 1.9230886558627267e-06, "loss": 2.0244, "step": 18878 }, { "epoch": 3.6, "learning_rate": 1.923031458531935e-06, "loss": 1.8172, "step": 18879 }, { "epoch": 3.6, "learning_rate": 1.922974261201144e-06, "loss": 1.4881, "step": 18880 }, { "epoch": 3.6, "learning_rate": 1.9229170638703526e-06, "loss": 2.4374, "step": 18881 }, { "epoch": 3.6, "learning_rate": 1.9228598665395614e-06, "loss": 2.7333, "step": 18882 }, { "epoch": 3.6, "learning_rate": 1.9228026692087706e-06, "loss": 2.1675, "step": 18883 }, { "epoch": 3.6, "learning_rate": 1.922745471877979e-06, "loss": 1.4616, "step": 18884 }, { "epoch": 3.6, "learning_rate": 1.922688274547188e-06, "loss": 2.3919, "step": 18885 }, { "epoch": 3.6, "learning_rate": 1.9226310772163966e-06, "loss": 2.4544, "step": 18886 }, { "epoch": 3.6, "learning_rate": 1.9225738798856054e-06, "loss": 1.7541, "step": 18887 }, { "epoch": 3.6, "learning_rate": 1.922516682554814e-06, "loss": 3.3125, "step": 18888 }, { "epoch": 3.6, "learning_rate": 1.922459485224023e-06, "loss": 2.6207, "step": 18889 }, { "epoch": 3.6, "learning_rate": 1.9224022878932318e-06, "loss": 1.7957, "step": 18890 }, { "epoch": 3.6, "learning_rate": 1.9223450905624406e-06, "loss": 1.9593, "step": 18891 }, { "epoch": 3.6, "learning_rate": 1.9222878932316493e-06, "loss": 3.3069, "step": 18892 }, { "epoch": 3.6, "learning_rate": 1.922230695900858e-06, "loss": 2.0116, "step": 18893 }, { "epoch": 3.6, "learning_rate": 1.9221734985700665e-06, "loss": 2.5045, "step": 18894 }, { "epoch": 3.6, "learning_rate": 1.9221163012392757e-06, "loss": 2.9407, "step": 18895 }, { "epoch": 3.6, "learning_rate": 1.922059103908484e-06, "loss": 2.8263, "step": 18896 }, { "epoch": 3.6, "learning_rate": 1.922001906577693e-06, "loss": 3.0188, "step": 18897 }, { "epoch": 3.6, "learning_rate": 1.921944709246902e-06, "loss": 2.6067, "step": 18898 }, { "epoch": 3.6, "learning_rate": 1.9218875119161105e-06, "loss": 2.4133, "step": 18899 }, { "epoch": 3.6, "learning_rate": 1.9218303145853197e-06, "loss": 2.6261, "step": 18900 }, { "epoch": 3.6, "learning_rate": 1.921773117254528e-06, "loss": 2.8099, "step": 18901 }, { "epoch": 3.6, "learning_rate": 1.921715919923737e-06, "loss": 2.7662, "step": 18902 }, { "epoch": 3.6, "learning_rate": 1.9216587225929457e-06, "loss": 2.1209, "step": 18903 }, { "epoch": 3.6, "learning_rate": 1.9216015252621544e-06, "loss": 2.6227, "step": 18904 }, { "epoch": 3.6, "learning_rate": 1.9215443279313632e-06, "loss": 2.2599, "step": 18905 }, { "epoch": 3.6, "learning_rate": 1.921487130600572e-06, "loss": 3.0147, "step": 18906 }, { "epoch": 3.6, "learning_rate": 1.921429933269781e-06, "loss": 2.0822, "step": 18907 }, { "epoch": 3.6, "learning_rate": 1.9213727359389896e-06, "loss": 1.9219, "step": 18908 }, { "epoch": 3.6, "learning_rate": 1.9213155386081984e-06, "loss": 2.8897, "step": 18909 }, { "epoch": 3.6, "learning_rate": 1.921258341277407e-06, "loss": 2.7775, "step": 18910 }, { "epoch": 3.6, "learning_rate": 1.9212011439466156e-06, "loss": 2.4575, "step": 18911 }, { "epoch": 3.6, "learning_rate": 1.921143946615825e-06, "loss": 1.5034, "step": 18912 }, { "epoch": 3.6, "learning_rate": 1.921086749285033e-06, "loss": 1.8063, "step": 18913 }, { "epoch": 3.6, "learning_rate": 1.9210295519542424e-06, "loss": 1.8118, "step": 18914 }, { "epoch": 3.6, "learning_rate": 1.920972354623451e-06, "loss": 3.1125, "step": 18915 }, { "epoch": 3.6, "learning_rate": 1.9209151572926596e-06, "loss": 2.4109, "step": 18916 }, { "epoch": 3.6, "learning_rate": 1.9208579599618688e-06, "loss": 1.5587, "step": 18917 }, { "epoch": 3.6, "learning_rate": 1.920800762631077e-06, "loss": 1.723, "step": 18918 }, { "epoch": 3.6, "learning_rate": 1.920743565300286e-06, "loss": 1.6156, "step": 18919 }, { "epoch": 3.6, "learning_rate": 1.9206863679694947e-06, "loss": 1.2551, "step": 18920 }, { "epoch": 3.6, "learning_rate": 1.9206291706387035e-06, "loss": 2.9386, "step": 18921 }, { "epoch": 3.6, "learning_rate": 1.9205719733079123e-06, "loss": 2.1023, "step": 18922 }, { "epoch": 3.6, "learning_rate": 1.920514775977121e-06, "loss": 3.4823, "step": 18923 }, { "epoch": 3.6, "learning_rate": 1.92045757864633e-06, "loss": 3.7069, "step": 18924 }, { "epoch": 3.6, "learning_rate": 1.9204003813155387e-06, "loss": 2.9143, "step": 18925 }, { "epoch": 3.6, "learning_rate": 1.9203431839847475e-06, "loss": 1.2216, "step": 18926 }, { "epoch": 3.61, "learning_rate": 1.9202859866539563e-06, "loss": 2.7708, "step": 18927 }, { "epoch": 3.61, "learning_rate": 1.9202287893231647e-06, "loss": 2.9699, "step": 18928 }, { "epoch": 3.61, "learning_rate": 1.920171591992374e-06, "loss": 1.8821, "step": 18929 }, { "epoch": 3.61, "learning_rate": 1.9201143946615822e-06, "loss": 1.7996, "step": 18930 }, { "epoch": 3.61, "learning_rate": 1.9200571973307915e-06, "loss": 2.7314, "step": 18931 }, { "epoch": 3.61, "learning_rate": 1.9200000000000003e-06, "loss": 2.1981, "step": 18932 }, { "epoch": 3.61, "learning_rate": 1.9199428026692086e-06, "loss": 2.3213, "step": 18933 }, { "epoch": 3.61, "learning_rate": 1.919885605338418e-06, "loss": 4.1571, "step": 18934 }, { "epoch": 3.61, "learning_rate": 1.919828408007626e-06, "loss": 2.0282, "step": 18935 }, { "epoch": 3.61, "learning_rate": 1.919771210676835e-06, "loss": 2.1235, "step": 18936 }, { "epoch": 3.61, "learning_rate": 1.9197140133460442e-06, "loss": 2.3047, "step": 18937 }, { "epoch": 3.61, "learning_rate": 1.9196568160152526e-06, "loss": 2.395, "step": 18938 }, { "epoch": 3.61, "learning_rate": 1.9195996186844614e-06, "loss": 2.0147, "step": 18939 }, { "epoch": 3.61, "learning_rate": 1.91954242135367e-06, "loss": 2.2031, "step": 18940 }, { "epoch": 3.61, "learning_rate": 1.919485224022879e-06, "loss": 2.7521, "step": 18941 }, { "epoch": 3.61, "learning_rate": 1.9194280266920878e-06, "loss": 2.4656, "step": 18942 }, { "epoch": 3.61, "learning_rate": 1.9193708293612966e-06, "loss": 2.3683, "step": 18943 }, { "epoch": 3.61, "learning_rate": 1.9193136320305054e-06, "loss": 3.4188, "step": 18944 }, { "epoch": 3.61, "learning_rate": 1.9192564346997137e-06, "loss": 2.1357, "step": 18945 }, { "epoch": 3.61, "learning_rate": 1.919199237368923e-06, "loss": 1.708, "step": 18946 }, { "epoch": 3.61, "learning_rate": 1.9191420400381317e-06, "loss": 2.7513, "step": 18947 }, { "epoch": 3.61, "learning_rate": 1.9190848427073405e-06, "loss": 1.6758, "step": 18948 }, { "epoch": 3.61, "learning_rate": 1.9190276453765493e-06, "loss": 2.5375, "step": 18949 }, { "epoch": 3.61, "learning_rate": 1.9189704480457577e-06, "loss": 2.8926, "step": 18950 }, { "epoch": 3.61, "learning_rate": 1.918913250714967e-06, "loss": 3.0992, "step": 18951 }, { "epoch": 3.61, "learning_rate": 1.9188560533841753e-06, "loss": 2.3787, "step": 18952 }, { "epoch": 3.61, "learning_rate": 1.918798856053384e-06, "loss": 3.4273, "step": 18953 }, { "epoch": 3.61, "learning_rate": 1.9187416587225933e-06, "loss": 2.9874, "step": 18954 }, { "epoch": 3.61, "learning_rate": 1.9186844613918017e-06, "loss": 2.0964, "step": 18955 }, { "epoch": 3.61, "learning_rate": 1.9186272640610105e-06, "loss": 2.3394, "step": 18956 }, { "epoch": 3.61, "learning_rate": 1.9185700667302193e-06, "loss": 2.4275, "step": 18957 }, { "epoch": 3.61, "learning_rate": 1.918512869399428e-06, "loss": 2.1642, "step": 18958 }, { "epoch": 3.61, "learning_rate": 1.918455672068637e-06, "loss": 2.0843, "step": 18959 }, { "epoch": 3.61, "learning_rate": 1.9183984747378456e-06, "loss": 3.3776, "step": 18960 }, { "epoch": 3.61, "learning_rate": 1.9183412774070544e-06, "loss": 1.7902, "step": 18961 }, { "epoch": 3.61, "learning_rate": 1.918284080076263e-06, "loss": 1.8394, "step": 18962 }, { "epoch": 3.61, "learning_rate": 1.918226882745472e-06, "loss": 1.8511, "step": 18963 }, { "epoch": 3.61, "learning_rate": 1.918169685414681e-06, "loss": 2.1934, "step": 18964 }, { "epoch": 3.61, "learning_rate": 1.9181124880838896e-06, "loss": 3.5298, "step": 18965 }, { "epoch": 3.61, "learning_rate": 1.9180552907530984e-06, "loss": 2.8249, "step": 18966 }, { "epoch": 3.61, "learning_rate": 1.9179980934223068e-06, "loss": 2.3508, "step": 18967 }, { "epoch": 3.61, "learning_rate": 1.917940896091516e-06, "loss": 2.8342, "step": 18968 }, { "epoch": 3.61, "learning_rate": 1.9178836987607244e-06, "loss": 1.4208, "step": 18969 }, { "epoch": 3.61, "learning_rate": 1.917826501429933e-06, "loss": 1.7344, "step": 18970 }, { "epoch": 3.61, "learning_rate": 1.9177693040991424e-06, "loss": 1.072, "step": 18971 }, { "epoch": 3.61, "learning_rate": 1.9177121067683507e-06, "loss": 2.7501, "step": 18972 }, { "epoch": 3.61, "learning_rate": 1.9176549094375595e-06, "loss": 1.3324, "step": 18973 }, { "epoch": 3.61, "learning_rate": 1.9175977121067683e-06, "loss": 1.3859, "step": 18974 }, { "epoch": 3.61, "learning_rate": 1.917540514775977e-06, "loss": 2.3862, "step": 18975 }, { "epoch": 3.61, "learning_rate": 1.917483317445186e-06, "loss": 1.6827, "step": 18976 }, { "epoch": 3.61, "learning_rate": 1.9174261201143947e-06, "loss": 2.8603, "step": 18977 }, { "epoch": 3.61, "learning_rate": 1.9173689227836035e-06, "loss": 3.485, "step": 18978 }, { "epoch": 3.62, "learning_rate": 1.9173117254528123e-06, "loss": 2.2999, "step": 18979 }, { "epoch": 3.62, "learning_rate": 1.917254528122021e-06, "loss": 2.233, "step": 18980 }, { "epoch": 3.62, "learning_rate": 1.91719733079123e-06, "loss": 3.0367, "step": 18981 }, { "epoch": 3.62, "learning_rate": 1.9171401334604387e-06, "loss": 2.2232, "step": 18982 }, { "epoch": 3.62, "learning_rate": 1.9170829361296475e-06, "loss": 2.5087, "step": 18983 }, { "epoch": 3.62, "learning_rate": 1.917025738798856e-06, "loss": 2.9541, "step": 18984 }, { "epoch": 3.62, "learning_rate": 1.916968541468065e-06, "loss": 2.3683, "step": 18985 }, { "epoch": 3.62, "learning_rate": 1.916911344137274e-06, "loss": 1.9199, "step": 18986 }, { "epoch": 3.62, "learning_rate": 1.9168541468064822e-06, "loss": 3.0482, "step": 18987 }, { "epoch": 3.62, "learning_rate": 1.9167969494756914e-06, "loss": 2.8504, "step": 18988 }, { "epoch": 3.62, "learning_rate": 1.9167397521449e-06, "loss": 1.5764, "step": 18989 }, { "epoch": 3.62, "learning_rate": 1.9166825548141086e-06, "loss": 1.9221, "step": 18990 }, { "epoch": 3.62, "learning_rate": 1.9166253574833174e-06, "loss": 2.313, "step": 18991 }, { "epoch": 3.62, "learning_rate": 1.916568160152526e-06, "loss": 1.6932, "step": 18992 }, { "epoch": 3.62, "learning_rate": 1.916510962821735e-06, "loss": 2.6498, "step": 18993 }, { "epoch": 3.62, "learning_rate": 1.9164537654909438e-06, "loss": 2.4843, "step": 18994 }, { "epoch": 3.62, "learning_rate": 1.9163965681601526e-06, "loss": 2.2871, "step": 18995 }, { "epoch": 3.62, "learning_rate": 1.9163393708293614e-06, "loss": 2.4278, "step": 18996 }, { "epoch": 3.62, "learning_rate": 1.91628217349857e-06, "loss": 3.2537, "step": 18997 }, { "epoch": 3.62, "learning_rate": 1.916224976167779e-06, "loss": 1.7248, "step": 18998 }, { "epoch": 3.62, "learning_rate": 1.9161677788369877e-06, "loss": 3.294, "step": 18999 }, { "epoch": 3.62, "learning_rate": 1.9161105815061965e-06, "loss": 2.8744, "step": 19000 }, { "epoch": 3.62, "learning_rate": 1.916053384175405e-06, "loss": 2.987, "step": 19001 }, { "epoch": 3.62, "learning_rate": 1.915996186844614e-06, "loss": 2.7145, "step": 19002 }, { "epoch": 3.62, "learning_rate": 1.915938989513823e-06, "loss": 3.0515, "step": 19003 }, { "epoch": 3.62, "learning_rate": 1.9158817921830313e-06, "loss": 2.768, "step": 19004 }, { "epoch": 3.62, "learning_rate": 1.9158245948522405e-06, "loss": 2.503, "step": 19005 }, { "epoch": 3.62, "learning_rate": 1.915767397521449e-06, "loss": 1.6765, "step": 19006 }, { "epoch": 3.62, "learning_rate": 1.9157102001906577e-06, "loss": 1.4343, "step": 19007 }, { "epoch": 3.62, "learning_rate": 1.9156530028598665e-06, "loss": 1.674, "step": 19008 }, { "epoch": 3.62, "learning_rate": 1.9155958055290753e-06, "loss": 1.4374, "step": 19009 }, { "epoch": 3.62, "learning_rate": 1.915538608198284e-06, "loss": 1.935, "step": 19010 }, { "epoch": 3.62, "learning_rate": 1.915481410867493e-06, "loss": 2.363, "step": 19011 }, { "epoch": 3.62, "learning_rate": 1.9154242135367016e-06, "loss": 1.9569, "step": 19012 }, { "epoch": 3.62, "learning_rate": 1.9153670162059104e-06, "loss": 2.4139, "step": 19013 }, { "epoch": 3.62, "learning_rate": 1.9153098188751192e-06, "loss": 2.0488, "step": 19014 }, { "epoch": 3.62, "learning_rate": 1.915252621544328e-06, "loss": 2.771, "step": 19015 }, { "epoch": 3.62, "learning_rate": 1.915195424213537e-06, "loss": 1.6274, "step": 19016 }, { "epoch": 3.62, "learning_rate": 1.9151382268827456e-06, "loss": 2.0504, "step": 19017 }, { "epoch": 3.62, "learning_rate": 1.915081029551954e-06, "loss": 1.993, "step": 19018 }, { "epoch": 3.62, "learning_rate": 1.915023832221163e-06, "loss": 2.7399, "step": 19019 }, { "epoch": 3.62, "learning_rate": 1.914966634890372e-06, "loss": 3.4846, "step": 19020 }, { "epoch": 3.62, "learning_rate": 1.9149094375595804e-06, "loss": 2.1513, "step": 19021 }, { "epoch": 3.62, "learning_rate": 1.9148522402287896e-06, "loss": 2.2496, "step": 19022 }, { "epoch": 3.62, "learning_rate": 1.914795042897998e-06, "loss": 3.1369, "step": 19023 }, { "epoch": 3.62, "learning_rate": 1.9147378455672067e-06, "loss": 1.5238, "step": 19024 }, { "epoch": 3.62, "learning_rate": 1.914680648236416e-06, "loss": 2.6932, "step": 19025 }, { "epoch": 3.62, "learning_rate": 1.9146234509056243e-06, "loss": 2.3996, "step": 19026 }, { "epoch": 3.62, "learning_rate": 1.914566253574833e-06, "loss": 1.8661, "step": 19027 }, { "epoch": 3.62, "learning_rate": 1.914509056244042e-06, "loss": 1.7516, "step": 19028 }, { "epoch": 3.62, "learning_rate": 1.9144518589132507e-06, "loss": 2.5506, "step": 19029 }, { "epoch": 3.62, "learning_rate": 1.9143946615824595e-06, "loss": 1.4774, "step": 19030 }, { "epoch": 3.62, "learning_rate": 1.9143374642516683e-06, "loss": 1.9764, "step": 19031 }, { "epoch": 3.63, "learning_rate": 1.914280266920877e-06, "loss": 2.1365, "step": 19032 }, { "epoch": 3.63, "learning_rate": 1.914223069590086e-06, "loss": 1.3505, "step": 19033 }, { "epoch": 3.63, "learning_rate": 1.9141658722592947e-06, "loss": 2.4256, "step": 19034 }, { "epoch": 3.63, "learning_rate": 1.9141086749285035e-06, "loss": 2.1939, "step": 19035 }, { "epoch": 3.63, "learning_rate": 1.9140514775977123e-06, "loss": 2.0649, "step": 19036 }, { "epoch": 3.63, "learning_rate": 1.913994280266921e-06, "loss": 1.4306, "step": 19037 }, { "epoch": 3.63, "learning_rate": 1.9139370829361294e-06, "loss": 1.8106, "step": 19038 }, { "epoch": 3.63, "learning_rate": 1.9138798856053386e-06, "loss": 3.1459, "step": 19039 }, { "epoch": 3.63, "learning_rate": 1.913822688274547e-06, "loss": 1.971, "step": 19040 }, { "epoch": 3.63, "learning_rate": 1.913765490943756e-06, "loss": 2.5553, "step": 19041 }, { "epoch": 3.63, "learning_rate": 1.913708293612965e-06, "loss": 2.2909, "step": 19042 }, { "epoch": 3.63, "learning_rate": 1.9136510962821734e-06, "loss": 1.1616, "step": 19043 }, { "epoch": 3.63, "learning_rate": 1.9135938989513826e-06, "loss": 2.3833, "step": 19044 }, { "epoch": 3.63, "learning_rate": 1.913536701620591e-06, "loss": 2.6909, "step": 19045 }, { "epoch": 3.63, "learning_rate": 1.9134795042897998e-06, "loss": 2.3308, "step": 19046 }, { "epoch": 3.63, "learning_rate": 1.9134223069590086e-06, "loss": 2.6101, "step": 19047 }, { "epoch": 3.63, "learning_rate": 1.9133651096282174e-06, "loss": 2.8407, "step": 19048 }, { "epoch": 3.63, "learning_rate": 1.913307912297426e-06, "loss": 3.0054, "step": 19049 }, { "epoch": 3.63, "learning_rate": 1.913250714966635e-06, "loss": 3.0599, "step": 19050 }, { "epoch": 3.63, "learning_rate": 1.9131935176358438e-06, "loss": 2.0894, "step": 19051 }, { "epoch": 3.63, "learning_rate": 1.9131363203050525e-06, "loss": 2.4495, "step": 19052 }, { "epoch": 3.63, "learning_rate": 1.9130791229742613e-06, "loss": 1.9742, "step": 19053 }, { "epoch": 3.63, "learning_rate": 1.91302192564347e-06, "loss": 2.4326, "step": 19054 }, { "epoch": 3.63, "learning_rate": 1.9129647283126785e-06, "loss": 2.0462, "step": 19055 }, { "epoch": 3.63, "learning_rate": 1.9129075309818877e-06, "loss": 2.5874, "step": 19056 }, { "epoch": 3.63, "learning_rate": 1.912850333651096e-06, "loss": 1.7216, "step": 19057 }, { "epoch": 3.63, "learning_rate": 1.912793136320305e-06, "loss": 2.668, "step": 19058 }, { "epoch": 3.63, "learning_rate": 1.912735938989514e-06, "loss": 1.0913, "step": 19059 }, { "epoch": 3.63, "learning_rate": 1.9126787416587225e-06, "loss": 2.5256, "step": 19060 }, { "epoch": 3.63, "learning_rate": 1.9126215443279317e-06, "loss": 1.4696, "step": 19061 }, { "epoch": 3.63, "learning_rate": 1.91256434699714e-06, "loss": 3.2919, "step": 19062 }, { "epoch": 3.63, "learning_rate": 1.912507149666349e-06, "loss": 1.0913, "step": 19063 }, { "epoch": 3.63, "learning_rate": 1.912449952335558e-06, "loss": 1.3791, "step": 19064 }, { "epoch": 3.63, "learning_rate": 1.9123927550047664e-06, "loss": 2.5976, "step": 19065 }, { "epoch": 3.63, "learning_rate": 1.9123355576739752e-06, "loss": 3.4732, "step": 19066 }, { "epoch": 3.63, "learning_rate": 1.912278360343184e-06, "loss": 2.2257, "step": 19067 }, { "epoch": 3.63, "learning_rate": 1.912221163012393e-06, "loss": 1.6385, "step": 19068 }, { "epoch": 3.63, "learning_rate": 1.9121639656816016e-06, "loss": 2.1924, "step": 19069 }, { "epoch": 3.63, "learning_rate": 1.9121067683508104e-06, "loss": 2.4967, "step": 19070 }, { "epoch": 3.63, "learning_rate": 1.912049571020019e-06, "loss": 2.1233, "step": 19071 }, { "epoch": 3.63, "learning_rate": 1.9119923736892276e-06, "loss": 2.7636, "step": 19072 }, { "epoch": 3.63, "learning_rate": 1.911935176358437e-06, "loss": 1.3982, "step": 19073 }, { "epoch": 3.63, "learning_rate": 1.9118779790276456e-06, "loss": 1.5444, "step": 19074 }, { "epoch": 3.63, "learning_rate": 1.911820781696854e-06, "loss": 2.8015, "step": 19075 }, { "epoch": 3.63, "learning_rate": 1.911763584366063e-06, "loss": 2.6465, "step": 19076 }, { "epoch": 3.63, "learning_rate": 1.9117063870352715e-06, "loss": 1.5028, "step": 19077 }, { "epoch": 3.63, "learning_rate": 1.9116491897044808e-06, "loss": 1.7853, "step": 19078 }, { "epoch": 3.63, "learning_rate": 1.911591992373689e-06, "loss": 2.424, "step": 19079 }, { "epoch": 3.63, "learning_rate": 1.911534795042898e-06, "loss": 0.899, "step": 19080 }, { "epoch": 3.63, "learning_rate": 1.911477597712107e-06, "loss": 2.5971, "step": 19081 }, { "epoch": 3.63, "learning_rate": 1.9114204003813155e-06, "loss": 2.199, "step": 19082 }, { "epoch": 3.63, "learning_rate": 1.9113632030505243e-06, "loss": 2.4663, "step": 19083 }, { "epoch": 3.64, "learning_rate": 1.911306005719733e-06, "loss": 3.3398, "step": 19084 }, { "epoch": 3.64, "learning_rate": 1.911248808388942e-06, "loss": 1.6365, "step": 19085 }, { "epoch": 3.64, "learning_rate": 1.9111916110581507e-06, "loss": 3.0068, "step": 19086 }, { "epoch": 3.64, "learning_rate": 1.9111344137273595e-06, "loss": 1.6808, "step": 19087 }, { "epoch": 3.64, "learning_rate": 1.9110772163965683e-06, "loss": 3.2742, "step": 19088 }, { "epoch": 3.64, "learning_rate": 1.9110200190657766e-06, "loss": 3.6576, "step": 19089 }, { "epoch": 3.64, "learning_rate": 1.910962821734986e-06, "loss": 2.6904, "step": 19090 }, { "epoch": 3.64, "learning_rate": 1.9109056244041947e-06, "loss": 2.5344, "step": 19091 }, { "epoch": 3.64, "learning_rate": 1.910848427073403e-06, "loss": 1.733, "step": 19092 }, { "epoch": 3.64, "learning_rate": 1.9107912297426122e-06, "loss": 4.0632, "step": 19093 }, { "epoch": 3.64, "learning_rate": 1.9107340324118206e-06, "loss": 2.8646, "step": 19094 }, { "epoch": 3.64, "learning_rate": 1.91067683508103e-06, "loss": 2.8486, "step": 19095 }, { "epoch": 3.64, "learning_rate": 1.910619637750238e-06, "loss": 2.4873, "step": 19096 }, { "epoch": 3.64, "learning_rate": 1.910562440419447e-06, "loss": 2.4003, "step": 19097 }, { "epoch": 3.64, "learning_rate": 1.9105052430886562e-06, "loss": 2.4529, "step": 19098 }, { "epoch": 3.64, "learning_rate": 1.9104480457578646e-06, "loss": 2.9523, "step": 19099 }, { "epoch": 3.64, "learning_rate": 1.9103908484270734e-06, "loss": 2.2211, "step": 19100 }, { "epoch": 3.64, "learning_rate": 1.910333651096282e-06, "loss": 2.8327, "step": 19101 }, { "epoch": 3.64, "learning_rate": 1.910276453765491e-06, "loss": 2.3068, "step": 19102 }, { "epoch": 3.64, "learning_rate": 1.9102192564346998e-06, "loss": 1.6055, "step": 19103 }, { "epoch": 3.64, "learning_rate": 1.9101620591039086e-06, "loss": 1.7447, "step": 19104 }, { "epoch": 3.64, "learning_rate": 1.9101048617731173e-06, "loss": 3.4076, "step": 19105 }, { "epoch": 3.64, "learning_rate": 1.9100476644423257e-06, "loss": 2.6244, "step": 19106 }, { "epoch": 3.64, "learning_rate": 1.909990467111535e-06, "loss": 2.5565, "step": 19107 }, { "epoch": 3.64, "learning_rate": 1.9099332697807437e-06, "loss": 2.0006, "step": 19108 }, { "epoch": 3.64, "learning_rate": 1.9098760724499525e-06, "loss": 2.8246, "step": 19109 }, { "epoch": 3.64, "learning_rate": 1.9098188751191613e-06, "loss": 2.9582, "step": 19110 }, { "epoch": 3.64, "learning_rate": 1.9097616777883697e-06, "loss": 1.0938, "step": 19111 }, { "epoch": 3.64, "learning_rate": 1.909704480457579e-06, "loss": 2.1696, "step": 19112 }, { "epoch": 3.64, "learning_rate": 1.9096472831267877e-06, "loss": 1.3422, "step": 19113 }, { "epoch": 3.64, "learning_rate": 1.909590085795996e-06, "loss": 3.8613, "step": 19114 }, { "epoch": 3.64, "learning_rate": 1.9095328884652053e-06, "loss": 2.2313, "step": 19115 }, { "epoch": 3.64, "learning_rate": 1.9094756911344137e-06, "loss": 3.3131, "step": 19116 }, { "epoch": 3.64, "learning_rate": 1.9094184938036224e-06, "loss": 2.5825, "step": 19117 }, { "epoch": 3.64, "learning_rate": 1.9093612964728312e-06, "loss": 1.9729, "step": 19118 }, { "epoch": 3.64, "learning_rate": 1.90930409914204e-06, "loss": 1.9339, "step": 19119 }, { "epoch": 3.64, "learning_rate": 1.909246901811249e-06, "loss": 2.9283, "step": 19120 }, { "epoch": 3.64, "learning_rate": 1.9091897044804576e-06, "loss": 2.1932, "step": 19121 }, { "epoch": 3.64, "learning_rate": 1.9091325071496664e-06, "loss": 2.9015, "step": 19122 }, { "epoch": 3.64, "learning_rate": 1.9090753098188752e-06, "loss": 2.1891, "step": 19123 }, { "epoch": 3.64, "learning_rate": 1.909018112488084e-06, "loss": 2.1208, "step": 19124 }, { "epoch": 3.64, "learning_rate": 1.908960915157293e-06, "loss": 1.0687, "step": 19125 }, { "epoch": 3.64, "learning_rate": 1.9089037178265016e-06, "loss": 1.9314, "step": 19126 }, { "epoch": 3.64, "learning_rate": 1.9088465204957104e-06, "loss": 2.6505, "step": 19127 }, { "epoch": 3.64, "learning_rate": 1.9087893231649188e-06, "loss": 1.7047, "step": 19128 }, { "epoch": 3.64, "learning_rate": 1.908732125834128e-06, "loss": 2.1837, "step": 19129 }, { "epoch": 3.64, "learning_rate": 1.9086749285033368e-06, "loss": 2.0215, "step": 19130 }, { "epoch": 3.64, "learning_rate": 1.908617731172545e-06, "loss": 1.2017, "step": 19131 }, { "epoch": 3.64, "learning_rate": 1.9085605338417544e-06, "loss": 2.6417, "step": 19132 }, { "epoch": 3.64, "learning_rate": 1.9085033365109627e-06, "loss": 2.2216, "step": 19133 }, { "epoch": 3.64, "learning_rate": 1.9084461391801715e-06, "loss": 2.0878, "step": 19134 }, { "epoch": 3.64, "learning_rate": 1.9083889418493803e-06, "loss": 2.3688, "step": 19135 }, { "epoch": 3.64, "learning_rate": 1.908331744518589e-06, "loss": 2.3914, "step": 19136 }, { "epoch": 3.65, "learning_rate": 1.908274547187798e-06, "loss": 2.0385, "step": 19137 }, { "epoch": 3.65, "learning_rate": 1.9082173498570067e-06, "loss": 3.2416, "step": 19138 }, { "epoch": 3.65, "learning_rate": 1.9081601525262155e-06, "loss": 2.6726, "step": 19139 }, { "epoch": 3.65, "learning_rate": 1.9081029551954243e-06, "loss": 2.7375, "step": 19140 }, { "epoch": 3.65, "learning_rate": 1.908045757864633e-06, "loss": 3.3218, "step": 19141 }, { "epoch": 3.65, "learning_rate": 1.907988560533842e-06, "loss": 1.9342, "step": 19142 }, { "epoch": 3.65, "learning_rate": 1.9079313632030507e-06, "loss": 2.5025, "step": 19143 }, { "epoch": 3.65, "learning_rate": 1.9078741658722595e-06, "loss": 2.1421, "step": 19144 }, { "epoch": 3.65, "learning_rate": 1.907816968541468e-06, "loss": 2.1925, "step": 19145 }, { "epoch": 3.65, "learning_rate": 1.907759771210677e-06, "loss": 2.4443, "step": 19146 }, { "epoch": 3.65, "learning_rate": 1.907702573879886e-06, "loss": 1.5055, "step": 19147 }, { "epoch": 3.65, "learning_rate": 1.907645376549094e-06, "loss": 2.5028, "step": 19148 }, { "epoch": 3.65, "learning_rate": 1.9075881792183034e-06, "loss": 2.584, "step": 19149 }, { "epoch": 3.65, "learning_rate": 1.907530981887512e-06, "loss": 3.1947, "step": 19150 }, { "epoch": 3.65, "learning_rate": 1.9074737845567206e-06, "loss": 1.6821, "step": 19151 }, { "epoch": 3.65, "learning_rate": 1.9074165872259294e-06, "loss": 3.175, "step": 19152 }, { "epoch": 3.65, "learning_rate": 1.907359389895138e-06, "loss": 1.9619, "step": 19153 }, { "epoch": 3.65, "learning_rate": 1.907302192564347e-06, "loss": 1.9471, "step": 19154 }, { "epoch": 3.65, "learning_rate": 1.907244995233556e-06, "loss": 2.6172, "step": 19155 }, { "epoch": 3.65, "learning_rate": 1.9071877979027646e-06, "loss": 2.5581, "step": 19156 }, { "epoch": 3.65, "learning_rate": 1.9071306005719731e-06, "loss": 2.6123, "step": 19157 }, { "epoch": 3.65, "learning_rate": 1.9070734032411821e-06, "loss": 2.4283, "step": 19158 }, { "epoch": 3.65, "learning_rate": 1.9070162059103907e-06, "loss": 1.7943, "step": 19159 }, { "epoch": 3.65, "learning_rate": 1.9069590085795997e-06, "loss": 2.9705, "step": 19160 }, { "epoch": 3.65, "learning_rate": 1.9069018112488085e-06, "loss": 2.1548, "step": 19161 }, { "epoch": 3.65, "learning_rate": 1.9068446139180171e-06, "loss": 3.895, "step": 19162 }, { "epoch": 3.65, "learning_rate": 1.9067874165872261e-06, "loss": 2.1778, "step": 19163 }, { "epoch": 3.65, "learning_rate": 1.9067302192564347e-06, "loss": 2.7657, "step": 19164 }, { "epoch": 3.65, "learning_rate": 1.9066730219256435e-06, "loss": 2.4293, "step": 19165 }, { "epoch": 3.65, "learning_rate": 1.9066158245948525e-06, "loss": 3.0372, "step": 19166 }, { "epoch": 3.65, "learning_rate": 1.906558627264061e-06, "loss": 2.0137, "step": 19167 }, { "epoch": 3.65, "learning_rate": 1.9065014299332697e-06, "loss": 2.6887, "step": 19168 }, { "epoch": 3.65, "learning_rate": 1.9064442326024787e-06, "loss": 1.6168, "step": 19169 }, { "epoch": 3.65, "learning_rate": 1.9063870352716873e-06, "loss": 3.0349, "step": 19170 }, { "epoch": 3.65, "learning_rate": 1.906329837940896e-06, "loss": 2.7159, "step": 19171 }, { "epoch": 3.65, "learning_rate": 1.906272640610105e-06, "loss": 2.4364, "step": 19172 }, { "epoch": 3.65, "learning_rate": 1.9062154432793136e-06, "loss": 1.8188, "step": 19173 }, { "epoch": 3.65, "learning_rate": 1.9061582459485226e-06, "loss": 3.3129, "step": 19174 }, { "epoch": 3.65, "learning_rate": 1.9061010486177312e-06, "loss": 2.8717, "step": 19175 }, { "epoch": 3.65, "learning_rate": 1.90604385128694e-06, "loss": 1.6445, "step": 19176 }, { "epoch": 3.65, "learning_rate": 1.9059866539561488e-06, "loss": 2.2246, "step": 19177 }, { "epoch": 3.65, "learning_rate": 1.9059294566253576e-06, "loss": 2.8352, "step": 19178 }, { "epoch": 3.65, "learning_rate": 1.9058722592945662e-06, "loss": 1.6528, "step": 19179 }, { "epoch": 3.65, "learning_rate": 1.9058150619637752e-06, "loss": 1.1107, "step": 19180 }, { "epoch": 3.65, "learning_rate": 1.9057578646329838e-06, "loss": 3.4913, "step": 19181 }, { "epoch": 3.65, "learning_rate": 1.9057006673021926e-06, "loss": 2.5827, "step": 19182 }, { "epoch": 3.65, "learning_rate": 1.9056434699714016e-06, "loss": 1.2478, "step": 19183 }, { "epoch": 3.65, "learning_rate": 1.9055862726406102e-06, "loss": 2.5112, "step": 19184 }, { "epoch": 3.65, "learning_rate": 1.9055290753098187e-06, "loss": 2.3486, "step": 19185 }, { "epoch": 3.65, "learning_rate": 1.9054718779790277e-06, "loss": 2.5819, "step": 19186 }, { "epoch": 3.65, "learning_rate": 1.9054146806482365e-06, "loss": 2.7035, "step": 19187 }, { "epoch": 3.65, "learning_rate": 1.9053574833174451e-06, "loss": 3.256, "step": 19188 }, { "epoch": 3.66, "learning_rate": 1.9053002859866541e-06, "loss": 2.9305, "step": 19189 }, { "epoch": 3.66, "learning_rate": 1.9052430886558627e-06, "loss": 1.2452, "step": 19190 }, { "epoch": 3.66, "learning_rate": 1.9051858913250717e-06, "loss": 2.6516, "step": 19191 }, { "epoch": 3.66, "learning_rate": 1.9051286939942803e-06, "loss": 1.5556, "step": 19192 }, { "epoch": 3.66, "learning_rate": 1.905071496663489e-06, "loss": 1.7163, "step": 19193 }, { "epoch": 3.66, "learning_rate": 1.905014299332698e-06, "loss": 2.2667, "step": 19194 }, { "epoch": 3.66, "learning_rate": 1.9049571020019067e-06, "loss": 2.8786, "step": 19195 }, { "epoch": 3.66, "learning_rate": 1.9048999046711153e-06, "loss": 1.8545, "step": 19196 }, { "epoch": 3.66, "learning_rate": 1.9048427073403243e-06, "loss": 3.3137, "step": 19197 }, { "epoch": 3.66, "learning_rate": 1.9047855100095328e-06, "loss": 2.2428, "step": 19198 }, { "epoch": 3.66, "learning_rate": 1.9047283126787416e-06, "loss": 2.2529, "step": 19199 }, { "epoch": 3.66, "learning_rate": 1.9046711153479506e-06, "loss": 2.2397, "step": 19200 }, { "epoch": 3.66, "learning_rate": 1.9046139180171592e-06, "loss": 1.8285, "step": 19201 }, { "epoch": 3.66, "learning_rate": 1.9045567206863678e-06, "loss": 2.1839, "step": 19202 }, { "epoch": 3.66, "learning_rate": 1.9044995233555768e-06, "loss": 2.1566, "step": 19203 }, { "epoch": 3.66, "learning_rate": 1.9044423260247856e-06, "loss": 2.2615, "step": 19204 }, { "epoch": 3.66, "learning_rate": 1.9043851286939942e-06, "loss": 2.3976, "step": 19205 }, { "epoch": 3.66, "learning_rate": 1.9043279313632032e-06, "loss": 2.4351, "step": 19206 }, { "epoch": 3.66, "learning_rate": 1.9042707340324118e-06, "loss": 1.7918, "step": 19207 }, { "epoch": 3.66, "learning_rate": 1.9042135367016208e-06, "loss": 2.537, "step": 19208 }, { "epoch": 3.66, "learning_rate": 1.9041563393708294e-06, "loss": 2.9949, "step": 19209 }, { "epoch": 3.66, "learning_rate": 1.9040991420400382e-06, "loss": 1.2248, "step": 19210 }, { "epoch": 3.66, "learning_rate": 1.9040419447092472e-06, "loss": 1.598, "step": 19211 }, { "epoch": 3.66, "learning_rate": 1.9039847473784557e-06, "loss": 1.7845, "step": 19212 }, { "epoch": 3.66, "learning_rate": 1.9039275500476643e-06, "loss": 2.1858, "step": 19213 }, { "epoch": 3.66, "learning_rate": 1.9038703527168733e-06, "loss": 2.5112, "step": 19214 }, { "epoch": 3.66, "learning_rate": 1.9038131553860821e-06, "loss": 3.064, "step": 19215 }, { "epoch": 3.66, "learning_rate": 1.9037559580552907e-06, "loss": 3.4855, "step": 19216 }, { "epoch": 3.66, "learning_rate": 1.9036987607244997e-06, "loss": 2.4062, "step": 19217 }, { "epoch": 3.66, "learning_rate": 1.9036415633937083e-06, "loss": 2.9009, "step": 19218 }, { "epoch": 3.66, "learning_rate": 1.9035843660629169e-06, "loss": 2.5442, "step": 19219 }, { "epoch": 3.66, "learning_rate": 1.9035271687321259e-06, "loss": 2.9147, "step": 19220 }, { "epoch": 3.66, "learning_rate": 1.9034699714013347e-06, "loss": 2.8313, "step": 19221 }, { "epoch": 3.66, "learning_rate": 1.9034127740705433e-06, "loss": 2.3764, "step": 19222 }, { "epoch": 3.66, "learning_rate": 1.9033555767397523e-06, "loss": 2.2902, "step": 19223 }, { "epoch": 3.66, "learning_rate": 1.9032983794089608e-06, "loss": 2.3259, "step": 19224 }, { "epoch": 3.66, "learning_rate": 1.9032411820781699e-06, "loss": 1.7411, "step": 19225 }, { "epoch": 3.66, "learning_rate": 1.9031839847473784e-06, "loss": 2.1, "step": 19226 }, { "epoch": 3.66, "learning_rate": 1.9031267874165872e-06, "loss": 2.4817, "step": 19227 }, { "epoch": 3.66, "learning_rate": 1.9030695900857962e-06, "loss": 2.3129, "step": 19228 }, { "epoch": 3.66, "learning_rate": 1.9030123927550048e-06, "loss": 3.0688, "step": 19229 }, { "epoch": 3.66, "learning_rate": 1.9029551954242134e-06, "loss": 2.4334, "step": 19230 }, { "epoch": 3.66, "learning_rate": 1.9028979980934224e-06, "loss": 2.4859, "step": 19231 }, { "epoch": 3.66, "learning_rate": 1.9028408007626312e-06, "loss": 2.3987, "step": 19232 }, { "epoch": 3.66, "learning_rate": 1.9027836034318398e-06, "loss": 2.8826, "step": 19233 }, { "epoch": 3.66, "learning_rate": 1.9027264061010488e-06, "loss": 1.7899, "step": 19234 }, { "epoch": 3.66, "learning_rate": 1.9026692087702574e-06, "loss": 2.0402, "step": 19235 }, { "epoch": 3.66, "learning_rate": 1.9026120114394662e-06, "loss": 2.3507, "step": 19236 }, { "epoch": 3.66, "learning_rate": 1.902554814108675e-06, "loss": 2.315, "step": 19237 }, { "epoch": 3.66, "learning_rate": 1.9024976167778837e-06, "loss": 2.8149, "step": 19238 }, { "epoch": 3.66, "learning_rate": 1.9024404194470928e-06, "loss": 2.2985, "step": 19239 }, { "epoch": 3.66, "learning_rate": 1.9023832221163013e-06, "loss": 2.3488, "step": 19240 }, { "epoch": 3.66, "learning_rate": 1.90232602478551e-06, "loss": 1.1563, "step": 19241 }, { "epoch": 3.67, "learning_rate": 1.902268827454719e-06, "loss": 3.0547, "step": 19242 }, { "epoch": 3.67, "learning_rate": 1.9022116301239277e-06, "loss": 2.185, "step": 19243 }, { "epoch": 3.67, "learning_rate": 1.9021544327931363e-06, "loss": 2.8153, "step": 19244 }, { "epoch": 3.67, "learning_rate": 1.9020972354623453e-06, "loss": 1.9753, "step": 19245 }, { "epoch": 3.67, "learning_rate": 1.9020400381315539e-06, "loss": 2.2185, "step": 19246 }, { "epoch": 3.67, "learning_rate": 1.9019828408007625e-06, "loss": 2.5842, "step": 19247 }, { "epoch": 3.67, "learning_rate": 1.9019256434699715e-06, "loss": 2.1313, "step": 19248 }, { "epoch": 3.67, "learning_rate": 1.9018684461391803e-06, "loss": 2.0088, "step": 19249 }, { "epoch": 3.67, "learning_rate": 1.9018112488083889e-06, "loss": 2.1556, "step": 19250 }, { "epoch": 3.67, "learning_rate": 1.9017540514775979e-06, "loss": 2.3584, "step": 19251 }, { "epoch": 3.67, "learning_rate": 1.9016968541468064e-06, "loss": 3.5317, "step": 19252 }, { "epoch": 3.67, "learning_rate": 1.9016396568160152e-06, "loss": 2.4964, "step": 19253 }, { "epoch": 3.67, "learning_rate": 1.9015824594852242e-06, "loss": 3.3066, "step": 19254 }, { "epoch": 3.67, "learning_rate": 1.9015252621544328e-06, "loss": 3.0825, "step": 19255 }, { "epoch": 3.67, "learning_rate": 1.9014680648236418e-06, "loss": 2.1665, "step": 19256 }, { "epoch": 3.67, "learning_rate": 1.9014108674928504e-06, "loss": 1.9953, "step": 19257 }, { "epoch": 3.67, "learning_rate": 1.901353670162059e-06, "loss": 2.7765, "step": 19258 }, { "epoch": 3.67, "learning_rate": 1.901296472831268e-06, "loss": 2.9367, "step": 19259 }, { "epoch": 3.67, "learning_rate": 1.9012392755004768e-06, "loss": 2.9512, "step": 19260 }, { "epoch": 3.67, "learning_rate": 1.9011820781696854e-06, "loss": 1.9776, "step": 19261 }, { "epoch": 3.67, "learning_rate": 1.9011248808388944e-06, "loss": 3.4446, "step": 19262 }, { "epoch": 3.67, "learning_rate": 1.901067683508103e-06, "loss": 2.8295, "step": 19263 }, { "epoch": 3.67, "learning_rate": 1.9010104861773118e-06, "loss": 2.7482, "step": 19264 }, { "epoch": 3.67, "learning_rate": 1.9009532888465205e-06, "loss": 2.2643, "step": 19265 }, { "epoch": 3.67, "learning_rate": 1.9008960915157293e-06, "loss": 2.8001, "step": 19266 }, { "epoch": 3.67, "learning_rate": 1.900838894184938e-06, "loss": 2.3585, "step": 19267 }, { "epoch": 3.67, "learning_rate": 1.900781696854147e-06, "loss": 2.6516, "step": 19268 }, { "epoch": 3.67, "learning_rate": 1.9007244995233555e-06, "loss": 3.3229, "step": 19269 }, { "epoch": 3.67, "learning_rate": 1.9006673021925643e-06, "loss": 2.6762, "step": 19270 }, { "epoch": 3.67, "learning_rate": 1.9006101048617733e-06, "loss": 2.1572, "step": 19271 }, { "epoch": 3.67, "learning_rate": 1.9005529075309819e-06, "loss": 1.8934, "step": 19272 }, { "epoch": 3.67, "learning_rate": 1.900495710200191e-06, "loss": 1.3325, "step": 19273 }, { "epoch": 3.67, "learning_rate": 1.9004385128693995e-06, "loss": 2.7167, "step": 19274 }, { "epoch": 3.67, "learning_rate": 1.900381315538608e-06, "loss": 1.5791, "step": 19275 }, { "epoch": 3.67, "learning_rate": 1.900324118207817e-06, "loss": 3.3315, "step": 19276 }, { "epoch": 3.67, "learning_rate": 1.9002669208770259e-06, "loss": 1.9653, "step": 19277 }, { "epoch": 3.67, "learning_rate": 1.9002097235462344e-06, "loss": 1.9481, "step": 19278 }, { "epoch": 3.67, "learning_rate": 1.9001525262154434e-06, "loss": 2.6001, "step": 19279 }, { "epoch": 3.67, "learning_rate": 1.900095328884652e-06, "loss": 1.4663, "step": 19280 }, { "epoch": 3.67, "learning_rate": 1.9000381315538608e-06, "loss": 2.5643, "step": 19281 }, { "epoch": 3.67, "learning_rate": 1.8999809342230698e-06, "loss": 2.3682, "step": 19282 }, { "epoch": 3.67, "learning_rate": 1.8999237368922784e-06, "loss": 2.0884, "step": 19283 }, { "epoch": 3.67, "learning_rate": 1.899866539561487e-06, "loss": 2.9811, "step": 19284 }, { "epoch": 3.67, "learning_rate": 1.899809342230696e-06, "loss": 2.1499, "step": 19285 }, { "epoch": 3.67, "learning_rate": 1.8997521448999046e-06, "loss": 2.9487, "step": 19286 }, { "epoch": 3.67, "learning_rate": 1.8996949475691134e-06, "loss": 3.2839, "step": 19287 }, { "epoch": 3.67, "learning_rate": 1.8996377502383224e-06, "loss": 2.9506, "step": 19288 }, { "epoch": 3.67, "learning_rate": 1.899580552907531e-06, "loss": 2.3929, "step": 19289 }, { "epoch": 3.67, "learning_rate": 1.89952335557674e-06, "loss": 1.3736, "step": 19290 }, { "epoch": 3.67, "learning_rate": 1.8994661582459486e-06, "loss": 1.8291, "step": 19291 }, { "epoch": 3.67, "learning_rate": 1.8994089609151573e-06, "loss": 2.3739, "step": 19292 }, { "epoch": 3.67, "learning_rate": 1.8993517635843661e-06, "loss": 1.731, "step": 19293 }, { "epoch": 3.68, "learning_rate": 1.899294566253575e-06, "loss": 3.174, "step": 19294 }, { "epoch": 3.68, "learning_rate": 1.8992373689227835e-06, "loss": 2.12, "step": 19295 }, { "epoch": 3.68, "learning_rate": 1.8991801715919925e-06, "loss": 2.3146, "step": 19296 }, { "epoch": 3.68, "learning_rate": 1.899122974261201e-06, "loss": 1.8697, "step": 19297 }, { "epoch": 3.68, "learning_rate": 1.89906577693041e-06, "loss": 1.3069, "step": 19298 }, { "epoch": 3.68, "learning_rate": 1.899008579599619e-06, "loss": 1.8602, "step": 19299 }, { "epoch": 3.68, "learning_rate": 1.8989513822688275e-06, "loss": 2.7588, "step": 19300 }, { "epoch": 3.68, "learning_rate": 1.898894184938036e-06, "loss": 2.5449, "step": 19301 }, { "epoch": 3.68, "learning_rate": 1.898836987607245e-06, "loss": 3.4612, "step": 19302 }, { "epoch": 3.68, "learning_rate": 1.8987797902764539e-06, "loss": 2.597, "step": 19303 }, { "epoch": 3.68, "learning_rate": 1.8987225929456627e-06, "loss": 2.1242, "step": 19304 }, { "epoch": 3.68, "learning_rate": 1.8986653956148715e-06, "loss": 2.4362, "step": 19305 }, { "epoch": 3.68, "learning_rate": 1.89860819828408e-06, "loss": 2.9508, "step": 19306 }, { "epoch": 3.68, "learning_rate": 1.898551000953289e-06, "loss": 2.775, "step": 19307 }, { "epoch": 3.68, "learning_rate": 1.8984938036224976e-06, "loss": 1.1459, "step": 19308 }, { "epoch": 3.68, "learning_rate": 1.8984366062917064e-06, "loss": 1.6869, "step": 19309 }, { "epoch": 3.68, "learning_rate": 1.8983794089609154e-06, "loss": 2.1422, "step": 19310 }, { "epoch": 3.68, "learning_rate": 1.898322211630124e-06, "loss": 1.9549, "step": 19311 }, { "epoch": 3.68, "learning_rate": 1.8982650142993326e-06, "loss": 3.2747, "step": 19312 }, { "epoch": 3.68, "learning_rate": 1.8982078169685416e-06, "loss": 2.8115, "step": 19313 }, { "epoch": 3.68, "learning_rate": 1.8981506196377502e-06, "loss": 2.0745, "step": 19314 }, { "epoch": 3.68, "learning_rate": 1.898093422306959e-06, "loss": 2.5497, "step": 19315 }, { "epoch": 3.68, "learning_rate": 1.898036224976168e-06, "loss": 3.3091, "step": 19316 }, { "epoch": 3.68, "learning_rate": 1.8979790276453766e-06, "loss": 1.9949, "step": 19317 }, { "epoch": 3.68, "learning_rate": 1.8979218303145851e-06, "loss": 2.4769, "step": 19318 }, { "epoch": 3.68, "learning_rate": 1.8978646329837941e-06, "loss": 2.5449, "step": 19319 }, { "epoch": 3.68, "learning_rate": 1.897807435653003e-06, "loss": 1.71, "step": 19320 }, { "epoch": 3.68, "learning_rate": 1.897750238322212e-06, "loss": 2.8183, "step": 19321 }, { "epoch": 3.68, "learning_rate": 1.8976930409914205e-06, "loss": 2.0337, "step": 19322 }, { "epoch": 3.68, "learning_rate": 1.897635843660629e-06, "loss": 2.6393, "step": 19323 }, { "epoch": 3.68, "learning_rate": 1.8975786463298381e-06, "loss": 2.9307, "step": 19324 }, { "epoch": 3.68, "learning_rate": 1.8975214489990467e-06, "loss": 2.8126, "step": 19325 }, { "epoch": 3.68, "learning_rate": 1.8974642516682555e-06, "loss": 2.0688, "step": 19326 }, { "epoch": 3.68, "learning_rate": 1.8974070543374645e-06, "loss": 2.8407, "step": 19327 }, { "epoch": 3.68, "learning_rate": 1.897349857006673e-06, "loss": 2.6253, "step": 19328 }, { "epoch": 3.68, "learning_rate": 1.8972926596758817e-06, "loss": 2.5516, "step": 19329 }, { "epoch": 3.68, "learning_rate": 1.8972354623450907e-06, "loss": 2.1819, "step": 19330 }, { "epoch": 3.68, "learning_rate": 1.8971782650142995e-06, "loss": 2.2729, "step": 19331 }, { "epoch": 3.68, "learning_rate": 1.897121067683508e-06, "loss": 2.6424, "step": 19332 }, { "epoch": 3.68, "learning_rate": 1.897063870352717e-06, "loss": 1.4922, "step": 19333 }, { "epoch": 3.68, "learning_rate": 1.8970066730219256e-06, "loss": 2.1512, "step": 19334 }, { "epoch": 3.68, "learning_rate": 1.8969494756911342e-06, "loss": 1.6321, "step": 19335 }, { "epoch": 3.68, "learning_rate": 1.8968922783603432e-06, "loss": 1.64, "step": 19336 }, { "epoch": 3.68, "learning_rate": 1.896835081029552e-06, "loss": 3.3104, "step": 19337 }, { "epoch": 3.68, "learning_rate": 1.896777883698761e-06, "loss": 1.3514, "step": 19338 }, { "epoch": 3.68, "learning_rate": 1.8967206863679696e-06, "loss": 2.5634, "step": 19339 }, { "epoch": 3.68, "learning_rate": 1.8966634890371782e-06, "loss": 2.3738, "step": 19340 }, { "epoch": 3.68, "learning_rate": 1.8966062917063872e-06, "loss": 2.0123, "step": 19341 }, { "epoch": 3.68, "learning_rate": 1.896549094375596e-06, "loss": 2.0242, "step": 19342 }, { "epoch": 3.68, "learning_rate": 1.8964918970448046e-06, "loss": 2.6774, "step": 19343 }, { "epoch": 3.68, "learning_rate": 1.8964346997140136e-06, "loss": 2.9333, "step": 19344 }, { "epoch": 3.68, "learning_rate": 1.8963775023832221e-06, "loss": 1.5479, "step": 19345 }, { "epoch": 3.68, "learning_rate": 1.8963203050524307e-06, "loss": 2.1064, "step": 19346 }, { "epoch": 3.69, "learning_rate": 1.8962631077216397e-06, "loss": 2.1257, "step": 19347 }, { "epoch": 3.69, "learning_rate": 1.8962059103908485e-06, "loss": 2.9094, "step": 19348 }, { "epoch": 3.69, "learning_rate": 1.8961487130600571e-06, "loss": 2.3211, "step": 19349 }, { "epoch": 3.69, "learning_rate": 1.8960915157292661e-06, "loss": 2.2591, "step": 19350 }, { "epoch": 3.69, "learning_rate": 1.8960343183984747e-06, "loss": 2.4648, "step": 19351 }, { "epoch": 3.69, "learning_rate": 1.8959771210676835e-06, "loss": 2.8044, "step": 19352 }, { "epoch": 3.69, "learning_rate": 1.8959199237368923e-06, "loss": 2.0742, "step": 19353 }, { "epoch": 3.69, "learning_rate": 1.895862726406101e-06, "loss": 2.6298, "step": 19354 }, { "epoch": 3.69, "learning_rate": 1.89580552907531e-06, "loss": 2.577, "step": 19355 }, { "epoch": 3.69, "learning_rate": 1.8957483317445187e-06, "loss": 1.7306, "step": 19356 }, { "epoch": 3.69, "learning_rate": 1.8956911344137272e-06, "loss": 2.0428, "step": 19357 }, { "epoch": 3.69, "learning_rate": 1.8956339370829363e-06, "loss": 3.1736, "step": 19358 }, { "epoch": 3.69, "learning_rate": 1.895576739752145e-06, "loss": 2.6107, "step": 19359 }, { "epoch": 3.69, "learning_rate": 1.8955195424213536e-06, "loss": 1.7475, "step": 19360 }, { "epoch": 3.69, "learning_rate": 1.8954623450905626e-06, "loss": 1.8589, "step": 19361 }, { "epoch": 3.69, "learning_rate": 1.8954051477597712e-06, "loss": 1.0963, "step": 19362 }, { "epoch": 3.69, "learning_rate": 1.8953479504289798e-06, "loss": 3.4807, "step": 19363 }, { "epoch": 3.69, "learning_rate": 1.8952907530981888e-06, "loss": 2.8389, "step": 19364 }, { "epoch": 3.69, "learning_rate": 1.8952335557673976e-06, "loss": 2.3234, "step": 19365 }, { "epoch": 3.69, "learning_rate": 1.8951763584366062e-06, "loss": 2.8907, "step": 19366 }, { "epoch": 3.69, "learning_rate": 1.8951191611058152e-06, "loss": 1.9605, "step": 19367 }, { "epoch": 3.69, "learning_rate": 1.8950619637750238e-06, "loss": 1.9617, "step": 19368 }, { "epoch": 3.69, "learning_rate": 1.8950047664442328e-06, "loss": 2.4168, "step": 19369 }, { "epoch": 3.69, "learning_rate": 1.8949475691134416e-06, "loss": 1.9739, "step": 19370 }, { "epoch": 3.69, "learning_rate": 1.8948903717826502e-06, "loss": 1.5152, "step": 19371 }, { "epoch": 3.69, "learning_rate": 1.8948331744518592e-06, "loss": 2.2956, "step": 19372 }, { "epoch": 3.69, "learning_rate": 1.8947759771210677e-06, "loss": 2.2596, "step": 19373 }, { "epoch": 3.69, "learning_rate": 1.8947187797902763e-06, "loss": 2.8748, "step": 19374 }, { "epoch": 3.69, "learning_rate": 1.8946615824594853e-06, "loss": 1.5554, "step": 19375 }, { "epoch": 3.69, "learning_rate": 1.8946043851286941e-06, "loss": 1.907, "step": 19376 }, { "epoch": 3.69, "learning_rate": 1.8945471877979027e-06, "loss": 1.0932, "step": 19377 }, { "epoch": 3.69, "learning_rate": 1.8944899904671117e-06, "loss": 2.3374, "step": 19378 }, { "epoch": 3.69, "learning_rate": 1.8944327931363203e-06, "loss": 2.4753, "step": 19379 }, { "epoch": 3.69, "learning_rate": 1.894375595805529e-06, "loss": 2.7033, "step": 19380 }, { "epoch": 3.69, "learning_rate": 1.8943183984747379e-06, "loss": 2.6535, "step": 19381 }, { "epoch": 3.69, "learning_rate": 1.8942612011439467e-06, "loss": 2.3245, "step": 19382 }, { "epoch": 3.69, "learning_rate": 1.8942040038131553e-06, "loss": 3.4066, "step": 19383 }, { "epoch": 3.69, "learning_rate": 1.8941468064823643e-06, "loss": 2.9062, "step": 19384 }, { "epoch": 3.69, "learning_rate": 1.8940896091515728e-06, "loss": 1.6168, "step": 19385 }, { "epoch": 3.69, "learning_rate": 1.8940324118207818e-06, "loss": 2.0528, "step": 19386 }, { "epoch": 3.69, "learning_rate": 1.8939752144899906e-06, "loss": 3.3144, "step": 19387 }, { "epoch": 3.69, "learning_rate": 1.8939180171591992e-06, "loss": 1.7519, "step": 19388 }, { "epoch": 3.69, "learning_rate": 1.8938608198284082e-06, "loss": 2.7959, "step": 19389 }, { "epoch": 3.69, "learning_rate": 1.8938036224976168e-06, "loss": 2.6598, "step": 19390 }, { "epoch": 3.69, "learning_rate": 1.8937464251668256e-06, "loss": 1.6645, "step": 19391 }, { "epoch": 3.69, "learning_rate": 1.8936892278360344e-06, "loss": 2.0287, "step": 19392 }, { "epoch": 3.69, "learning_rate": 1.8936320305052432e-06, "loss": 2.584, "step": 19393 }, { "epoch": 3.69, "learning_rate": 1.8935748331744518e-06, "loss": 1.7409, "step": 19394 }, { "epoch": 3.69, "learning_rate": 1.8935176358436608e-06, "loss": 2.4477, "step": 19395 }, { "epoch": 3.69, "learning_rate": 1.8934604385128694e-06, "loss": 1.8589, "step": 19396 }, { "epoch": 3.69, "learning_rate": 1.8934032411820782e-06, "loss": 1.6067, "step": 19397 }, { "epoch": 3.69, "learning_rate": 1.8933460438512872e-06, "loss": 2.4854, "step": 19398 }, { "epoch": 3.7, "learning_rate": 1.8932888465204957e-06, "loss": 2.4196, "step": 19399 }, { "epoch": 3.7, "learning_rate": 1.8932316491897043e-06, "loss": 2.7808, "step": 19400 }, { "epoch": 3.7, "learning_rate": 1.8931744518589133e-06, "loss": 2.7563, "step": 19401 }, { "epoch": 3.7, "learning_rate": 1.893117254528122e-06, "loss": 3.1126, "step": 19402 }, { "epoch": 3.7, "learning_rate": 1.893060057197331e-06, "loss": 1.1215, "step": 19403 }, { "epoch": 3.7, "learning_rate": 1.8930028598665397e-06, "loss": 2.2006, "step": 19404 }, { "epoch": 3.7, "learning_rate": 1.8929456625357483e-06, "loss": 2.6121, "step": 19405 }, { "epoch": 3.7, "learning_rate": 1.8928884652049573e-06, "loss": 1.8488, "step": 19406 }, { "epoch": 3.7, "learning_rate": 1.8928312678741659e-06, "loss": 1.4352, "step": 19407 }, { "epoch": 3.7, "learning_rate": 1.8927740705433747e-06, "loss": 2.5589, "step": 19408 }, { "epoch": 3.7, "learning_rate": 1.8927168732125837e-06, "loss": 2.4867, "step": 19409 }, { "epoch": 3.7, "learning_rate": 1.8926596758817923e-06, "loss": 2.9464, "step": 19410 }, { "epoch": 3.7, "learning_rate": 1.8926024785510008e-06, "loss": 1.2338, "step": 19411 }, { "epoch": 3.7, "learning_rate": 1.8925452812202098e-06, "loss": 2.3594, "step": 19412 }, { "epoch": 3.7, "learning_rate": 1.8924880838894184e-06, "loss": 2.0682, "step": 19413 }, { "epoch": 3.7, "learning_rate": 1.8924308865586272e-06, "loss": 2.6567, "step": 19414 }, { "epoch": 3.7, "learning_rate": 1.8923736892278362e-06, "loss": 3.2106, "step": 19415 }, { "epoch": 3.7, "learning_rate": 1.8923164918970448e-06, "loss": 2.5928, "step": 19416 }, { "epoch": 3.7, "learning_rate": 1.8922592945662534e-06, "loss": 2.3285, "step": 19417 }, { "epoch": 3.7, "learning_rate": 1.8922020972354624e-06, "loss": 1.5617, "step": 19418 }, { "epoch": 3.7, "learning_rate": 1.8921448999046712e-06, "loss": 1.8155, "step": 19419 }, { "epoch": 3.7, "learning_rate": 1.89208770257388e-06, "loss": 2.754, "step": 19420 }, { "epoch": 3.7, "learning_rate": 1.8920305052430888e-06, "loss": 2.526, "step": 19421 }, { "epoch": 3.7, "learning_rate": 1.8919733079122974e-06, "loss": 2.2122, "step": 19422 }, { "epoch": 3.7, "learning_rate": 1.8919161105815064e-06, "loss": 2.4746, "step": 19423 }, { "epoch": 3.7, "learning_rate": 1.891858913250715e-06, "loss": 1.9495, "step": 19424 }, { "epoch": 3.7, "learning_rate": 1.8918017159199237e-06, "loss": 2.0023, "step": 19425 }, { "epoch": 3.7, "learning_rate": 1.8917445185891328e-06, "loss": 3.0049, "step": 19426 }, { "epoch": 3.7, "learning_rate": 1.8916873212583413e-06, "loss": 1.7606, "step": 19427 }, { "epoch": 3.7, "learning_rate": 1.89163012392755e-06, "loss": 1.7694, "step": 19428 }, { "epoch": 3.7, "learning_rate": 1.891572926596759e-06, "loss": 2.0935, "step": 19429 }, { "epoch": 3.7, "learning_rate": 1.8915157292659675e-06, "loss": 2.1662, "step": 19430 }, { "epoch": 3.7, "learning_rate": 1.8914585319351763e-06, "loss": 2.8714, "step": 19431 }, { "epoch": 3.7, "learning_rate": 1.8914013346043853e-06, "loss": 2.5197, "step": 19432 }, { "epoch": 3.7, "learning_rate": 1.8913441372735939e-06, "loss": 2.1175, "step": 19433 }, { "epoch": 3.7, "learning_rate": 1.8912869399428025e-06, "loss": 2.0801, "step": 19434 }, { "epoch": 3.7, "learning_rate": 1.8912297426120115e-06, "loss": 3.0329, "step": 19435 }, { "epoch": 3.7, "learning_rate": 1.8911725452812203e-06, "loss": 2.9175, "step": 19436 }, { "epoch": 3.7, "learning_rate": 1.8911153479504293e-06, "loss": 2.1541, "step": 19437 }, { "epoch": 3.7, "learning_rate": 1.8910581506196379e-06, "loss": 2.6983, "step": 19438 }, { "epoch": 3.7, "learning_rate": 1.8910009532888464e-06, "loss": 2.6792, "step": 19439 }, { "epoch": 3.7, "learning_rate": 1.8909437559580554e-06, "loss": 1.1099, "step": 19440 }, { "epoch": 3.7, "learning_rate": 1.890886558627264e-06, "loss": 2.9067, "step": 19441 }, { "epoch": 3.7, "learning_rate": 1.8908293612964728e-06, "loss": 2.3474, "step": 19442 }, { "epoch": 3.7, "learning_rate": 1.8907721639656818e-06, "loss": 2.3374, "step": 19443 }, { "epoch": 3.7, "learning_rate": 1.8907149666348904e-06, "loss": 2.0638, "step": 19444 }, { "epoch": 3.7, "learning_rate": 1.890657769304099e-06, "loss": 1.7173, "step": 19445 }, { "epoch": 3.7, "learning_rate": 1.890600571973308e-06, "loss": 1.8743, "step": 19446 }, { "epoch": 3.7, "learning_rate": 1.8905433746425168e-06, "loss": 2.1209, "step": 19447 }, { "epoch": 3.7, "learning_rate": 1.8904861773117254e-06, "loss": 3.046, "step": 19448 }, { "epoch": 3.7, "learning_rate": 1.8904289799809344e-06, "loss": 1.9321, "step": 19449 }, { "epoch": 3.7, "learning_rate": 1.890371782650143e-06, "loss": 2.126, "step": 19450 }, { "epoch": 3.7, "learning_rate": 1.890314585319352e-06, "loss": 1.6025, "step": 19451 }, { "epoch": 3.71, "learning_rate": 1.8902573879885605e-06, "loss": 1.9314, "step": 19452 }, { "epoch": 3.71, "learning_rate": 1.8902001906577693e-06, "loss": 2.5111, "step": 19453 }, { "epoch": 3.71, "learning_rate": 1.8901429933269783e-06, "loss": 1.8974, "step": 19454 }, { "epoch": 3.71, "learning_rate": 1.890085795996187e-06, "loss": 3.6664, "step": 19455 }, { "epoch": 3.71, "learning_rate": 1.8900285986653955e-06, "loss": 2.6402, "step": 19456 }, { "epoch": 3.71, "learning_rate": 1.8899714013346045e-06, "loss": 3.1305, "step": 19457 }, { "epoch": 3.71, "learning_rate": 1.8899142040038133e-06, "loss": 2.9922, "step": 19458 }, { "epoch": 3.71, "learning_rate": 1.8898570066730219e-06, "loss": 2.9382, "step": 19459 }, { "epoch": 3.71, "learning_rate": 1.8897998093422309e-06, "loss": 2.5076, "step": 19460 }, { "epoch": 3.71, "learning_rate": 1.8897426120114395e-06, "loss": 2.9211, "step": 19461 }, { "epoch": 3.71, "learning_rate": 1.889685414680648e-06, "loss": 1.9502, "step": 19462 }, { "epoch": 3.71, "learning_rate": 1.889628217349857e-06, "loss": 1.2442, "step": 19463 }, { "epoch": 3.71, "learning_rate": 1.8895710200190659e-06, "loss": 1.5112, "step": 19464 }, { "epoch": 3.71, "learning_rate": 1.8895138226882744e-06, "loss": 0.8638, "step": 19465 }, { "epoch": 3.71, "learning_rate": 1.8894566253574834e-06, "loss": 2.2499, "step": 19466 }, { "epoch": 3.71, "learning_rate": 1.889399428026692e-06, "loss": 1.8537, "step": 19467 }, { "epoch": 3.71, "learning_rate": 1.889342230695901e-06, "loss": 2.3039, "step": 19468 }, { "epoch": 3.71, "learning_rate": 1.8892850333651096e-06, "loss": 2.5848, "step": 19469 }, { "epoch": 3.71, "learning_rate": 1.8892278360343184e-06, "loss": 2.5327, "step": 19470 }, { "epoch": 3.71, "learning_rate": 1.8891706387035274e-06, "loss": 2.2229, "step": 19471 }, { "epoch": 3.71, "learning_rate": 1.889113441372736e-06, "loss": 2.3693, "step": 19472 }, { "epoch": 3.71, "learning_rate": 1.8890562440419446e-06, "loss": 1.9239, "step": 19473 }, { "epoch": 3.71, "learning_rate": 1.8889990467111536e-06, "loss": 2.868, "step": 19474 }, { "epoch": 3.71, "learning_rate": 1.8889418493803624e-06, "loss": 1.6071, "step": 19475 }, { "epoch": 3.71, "learning_rate": 1.888884652049571e-06, "loss": 2.6739, "step": 19476 }, { "epoch": 3.71, "learning_rate": 1.88882745471878e-06, "loss": 4.0583, "step": 19477 }, { "epoch": 3.71, "learning_rate": 1.8887702573879885e-06, "loss": 1.5436, "step": 19478 }, { "epoch": 3.71, "learning_rate": 1.8887130600571973e-06, "loss": 1.536, "step": 19479 }, { "epoch": 3.71, "learning_rate": 1.8886558627264061e-06, "loss": 2.5724, "step": 19480 }, { "epoch": 3.71, "learning_rate": 1.888598665395615e-06, "loss": 1.6503, "step": 19481 }, { "epoch": 3.71, "learning_rate": 1.8885414680648235e-06, "loss": 2.1688, "step": 19482 }, { "epoch": 3.71, "learning_rate": 1.8884842707340325e-06, "loss": 2.1599, "step": 19483 }, { "epoch": 3.71, "learning_rate": 1.888427073403241e-06, "loss": 3.0973, "step": 19484 }, { "epoch": 3.71, "learning_rate": 1.88836987607245e-06, "loss": 2.0357, "step": 19485 }, { "epoch": 3.71, "learning_rate": 1.888312678741659e-06, "loss": 1.9681, "step": 19486 }, { "epoch": 3.71, "learning_rate": 1.8882554814108675e-06, "loss": 1.8046, "step": 19487 }, { "epoch": 3.71, "learning_rate": 1.8881982840800765e-06, "loss": 2.7342, "step": 19488 }, { "epoch": 3.71, "learning_rate": 1.888141086749285e-06, "loss": 1.9671, "step": 19489 }, { "epoch": 3.71, "learning_rate": 1.8880838894184936e-06, "loss": 2.737, "step": 19490 }, { "epoch": 3.71, "learning_rate": 1.8880266920877027e-06, "loss": 2.4084, "step": 19491 }, { "epoch": 3.71, "learning_rate": 1.8879694947569114e-06, "loss": 0.9038, "step": 19492 }, { "epoch": 3.71, "learning_rate": 1.88791229742612e-06, "loss": 2.3871, "step": 19493 }, { "epoch": 3.71, "learning_rate": 1.887855100095329e-06, "loss": 1.3702, "step": 19494 }, { "epoch": 3.71, "learning_rate": 1.8877979027645376e-06, "loss": 1.832, "step": 19495 }, { "epoch": 3.71, "learning_rate": 1.8877407054337464e-06, "loss": 1.9632, "step": 19496 }, { "epoch": 3.71, "learning_rate": 1.8876835081029554e-06, "loss": 2.5978, "step": 19497 }, { "epoch": 3.71, "learning_rate": 1.887626310772164e-06, "loss": 2.6076, "step": 19498 }, { "epoch": 3.71, "learning_rate": 1.8875691134413726e-06, "loss": 1.574, "step": 19499 }, { "epoch": 3.71, "learning_rate": 1.8875119161105816e-06, "loss": 2.8154, "step": 19500 }, { "epoch": 3.71, "learning_rate": 1.8874547187797902e-06, "loss": 1.332, "step": 19501 }, { "epoch": 3.71, "learning_rate": 1.8873975214489992e-06, "loss": 2.0026, "step": 19502 }, { "epoch": 3.71, "learning_rate": 1.887340324118208e-06, "loss": 1.6476, "step": 19503 }, { "epoch": 3.72, "learning_rate": 1.8872831267874166e-06, "loss": 3.7926, "step": 19504 }, { "epoch": 3.72, "learning_rate": 1.8872259294566256e-06, "loss": 2.5087, "step": 19505 }, { "epoch": 3.72, "learning_rate": 1.8871687321258341e-06, "loss": 2.0243, "step": 19506 }, { "epoch": 3.72, "learning_rate": 1.887111534795043e-06, "loss": 1.9739, "step": 19507 }, { "epoch": 3.72, "learning_rate": 1.8870543374642517e-06, "loss": 1.9769, "step": 19508 }, { "epoch": 3.72, "learning_rate": 1.8869971401334605e-06, "loss": 3.3612, "step": 19509 }, { "epoch": 3.72, "learning_rate": 1.886939942802669e-06, "loss": 2.8929, "step": 19510 }, { "epoch": 3.72, "learning_rate": 1.8868827454718781e-06, "loss": 3.0288, "step": 19511 }, { "epoch": 3.72, "learning_rate": 1.8868255481410867e-06, "loss": 2.1057, "step": 19512 }, { "epoch": 3.72, "learning_rate": 1.8867683508102955e-06, "loss": 1.9153, "step": 19513 }, { "epoch": 3.72, "learning_rate": 1.8867111534795045e-06, "loss": 2.5174, "step": 19514 }, { "epoch": 3.72, "learning_rate": 1.886653956148713e-06, "loss": 1.922, "step": 19515 }, { "epoch": 3.72, "learning_rate": 1.886596758817922e-06, "loss": 1.9517, "step": 19516 }, { "epoch": 3.72, "learning_rate": 1.8865395614871307e-06, "loss": 2.9361, "step": 19517 }, { "epoch": 3.72, "learning_rate": 1.8864823641563392e-06, "loss": 2.8729, "step": 19518 }, { "epoch": 3.72, "learning_rate": 1.8864251668255482e-06, "loss": 3.1465, "step": 19519 }, { "epoch": 3.72, "learning_rate": 1.886367969494757e-06, "loss": 2.106, "step": 19520 }, { "epoch": 3.72, "learning_rate": 1.8863107721639656e-06, "loss": 1.9735, "step": 19521 }, { "epoch": 3.72, "learning_rate": 1.8862535748331746e-06, "loss": 2.2241, "step": 19522 }, { "epoch": 3.72, "learning_rate": 1.8861963775023832e-06, "loss": 2.3751, "step": 19523 }, { "epoch": 3.72, "learning_rate": 1.886139180171592e-06, "loss": 1.3084, "step": 19524 }, { "epoch": 3.72, "learning_rate": 1.886081982840801e-06, "loss": 1.9403, "step": 19525 }, { "epoch": 3.72, "learning_rate": 1.8860247855100096e-06, "loss": 3.2579, "step": 19526 }, { "epoch": 3.72, "learning_rate": 1.8859675881792182e-06, "loss": 2.9171, "step": 19527 }, { "epoch": 3.72, "learning_rate": 1.8859103908484272e-06, "loss": 2.7935, "step": 19528 }, { "epoch": 3.72, "learning_rate": 1.8858531935176358e-06, "loss": 1.9862, "step": 19529 }, { "epoch": 3.72, "learning_rate": 1.8857959961868446e-06, "loss": 1.8307, "step": 19530 }, { "epoch": 3.72, "learning_rate": 1.8857387988560536e-06, "loss": 2.6639, "step": 19531 }, { "epoch": 3.72, "learning_rate": 1.8856816015252621e-06, "loss": 3.0553, "step": 19532 }, { "epoch": 3.72, "learning_rate": 1.8856244041944711e-06, "loss": 2.7669, "step": 19533 }, { "epoch": 3.72, "learning_rate": 1.8855672068636797e-06, "loss": 2.3596, "step": 19534 }, { "epoch": 3.72, "learning_rate": 1.8855100095328885e-06, "loss": 2.629, "step": 19535 }, { "epoch": 3.72, "learning_rate": 1.8854528122020973e-06, "loss": 1.6869, "step": 19536 }, { "epoch": 3.72, "learning_rate": 1.8853956148713061e-06, "loss": 2.6788, "step": 19537 }, { "epoch": 3.72, "learning_rate": 1.8853384175405147e-06, "loss": 2.918, "step": 19538 }, { "epoch": 3.72, "learning_rate": 1.8852812202097237e-06, "loss": 2.1214, "step": 19539 }, { "epoch": 3.72, "learning_rate": 1.8852240228789323e-06, "loss": 1.4523, "step": 19540 }, { "epoch": 3.72, "learning_rate": 1.885166825548141e-06, "loss": 2.3579, "step": 19541 }, { "epoch": 3.72, "learning_rate": 1.88510962821735e-06, "loss": 2.608, "step": 19542 }, { "epoch": 3.72, "learning_rate": 1.8850524308865587e-06, "loss": 2.7979, "step": 19543 }, { "epoch": 3.72, "learning_rate": 1.8849952335557672e-06, "loss": 2.1898, "step": 19544 }, { "epoch": 3.72, "learning_rate": 1.8849380362249763e-06, "loss": 1.6301, "step": 19545 }, { "epoch": 3.72, "learning_rate": 1.884880838894185e-06, "loss": 3.3038, "step": 19546 }, { "epoch": 3.72, "learning_rate": 1.8848236415633936e-06, "loss": 2.9564, "step": 19547 }, { "epoch": 3.72, "learning_rate": 1.8847664442326026e-06, "loss": 2.0848, "step": 19548 }, { "epoch": 3.72, "learning_rate": 1.8847092469018112e-06, "loss": 2.6751, "step": 19549 }, { "epoch": 3.72, "learning_rate": 1.8846520495710202e-06, "loss": 1.2132, "step": 19550 }, { "epoch": 3.72, "learning_rate": 1.8845948522402288e-06, "loss": 2.4225, "step": 19551 }, { "epoch": 3.72, "learning_rate": 1.8845376549094376e-06, "loss": 3.4633, "step": 19552 }, { "epoch": 3.72, "learning_rate": 1.8844804575786466e-06, "loss": 2.6798, "step": 19553 }, { "epoch": 3.72, "learning_rate": 1.8844232602478552e-06, "loss": 2.8635, "step": 19554 }, { "epoch": 3.72, "learning_rate": 1.8843660629170638e-06, "loss": 2.5692, "step": 19555 }, { "epoch": 3.72, "learning_rate": 1.8843088655862728e-06, "loss": 2.2548, "step": 19556 }, { "epoch": 3.73, "learning_rate": 1.8842516682554814e-06, "loss": 1.9906, "step": 19557 }, { "epoch": 3.73, "learning_rate": 1.8841944709246901e-06, "loss": 2.9797, "step": 19558 }, { "epoch": 3.73, "learning_rate": 1.8841372735938992e-06, "loss": 2.5541, "step": 19559 }, { "epoch": 3.73, "learning_rate": 1.8840800762631077e-06, "loss": 2.412, "step": 19560 }, { "epoch": 3.73, "learning_rate": 1.8840228789323163e-06, "loss": 2.8208, "step": 19561 }, { "epoch": 3.73, "learning_rate": 1.8839656816015253e-06, "loss": 1.972, "step": 19562 }, { "epoch": 3.73, "learning_rate": 1.8839084842707341e-06, "loss": 1.8773, "step": 19563 }, { "epoch": 3.73, "learning_rate": 1.8838512869399427e-06, "loss": 2.9741, "step": 19564 }, { "epoch": 3.73, "learning_rate": 1.8837940896091517e-06, "loss": 2.2057, "step": 19565 }, { "epoch": 3.73, "learning_rate": 1.8837368922783603e-06, "loss": 3.5436, "step": 19566 }, { "epoch": 3.73, "learning_rate": 1.8836796949475693e-06, "loss": 2.327, "step": 19567 }, { "epoch": 3.73, "learning_rate": 1.8836224976167779e-06, "loss": 3.0872, "step": 19568 }, { "epoch": 3.73, "learning_rate": 1.8835653002859867e-06, "loss": 2.6457, "step": 19569 }, { "epoch": 3.73, "learning_rate": 1.8835081029551957e-06, "loss": 1.6356, "step": 19570 }, { "epoch": 3.73, "learning_rate": 1.8834509056244043e-06, "loss": 1.9252, "step": 19571 }, { "epoch": 3.73, "learning_rate": 1.8833937082936128e-06, "loss": 3.169, "step": 19572 }, { "epoch": 3.73, "learning_rate": 1.8833365109628218e-06, "loss": 3.1574, "step": 19573 }, { "epoch": 3.73, "learning_rate": 1.8832793136320306e-06, "loss": 1.702, "step": 19574 }, { "epoch": 3.73, "learning_rate": 1.8832221163012392e-06, "loss": 2.6528, "step": 19575 }, { "epoch": 3.73, "learning_rate": 1.8831649189704482e-06, "loss": 2.5926, "step": 19576 }, { "epoch": 3.73, "learning_rate": 1.8831077216396568e-06, "loss": 2.9391, "step": 19577 }, { "epoch": 3.73, "learning_rate": 1.8830505243088654e-06, "loss": 2.017, "step": 19578 }, { "epoch": 3.73, "learning_rate": 1.8829933269780744e-06, "loss": 3.1429, "step": 19579 }, { "epoch": 3.73, "learning_rate": 1.8829361296472832e-06, "loss": 1.8143, "step": 19580 }, { "epoch": 3.73, "learning_rate": 1.8828789323164922e-06, "loss": 2.2991, "step": 19581 }, { "epoch": 3.73, "learning_rate": 1.8828217349857008e-06, "loss": 2.2306, "step": 19582 }, { "epoch": 3.73, "learning_rate": 1.8827645376549094e-06, "loss": 2.0525, "step": 19583 }, { "epoch": 3.73, "learning_rate": 1.8827073403241184e-06, "loss": 1.702, "step": 19584 }, { "epoch": 3.73, "learning_rate": 1.882650142993327e-06, "loss": 2.0082, "step": 19585 }, { "epoch": 3.73, "learning_rate": 1.8825929456625357e-06, "loss": 2.0521, "step": 19586 }, { "epoch": 3.73, "learning_rate": 1.8825357483317447e-06, "loss": 2.8293, "step": 19587 }, { "epoch": 3.73, "learning_rate": 1.8824785510009533e-06, "loss": 2.3417, "step": 19588 }, { "epoch": 3.73, "learning_rate": 1.882421353670162e-06, "loss": 1.4637, "step": 19589 }, { "epoch": 3.73, "learning_rate": 1.882364156339371e-06, "loss": 2.4352, "step": 19590 }, { "epoch": 3.73, "learning_rate": 1.8823069590085797e-06, "loss": 2.0422, "step": 19591 }, { "epoch": 3.73, "learning_rate": 1.8822497616777883e-06, "loss": 2.0545, "step": 19592 }, { "epoch": 3.73, "learning_rate": 1.8821925643469973e-06, "loss": 2.6872, "step": 19593 }, { "epoch": 3.73, "learning_rate": 1.8821353670162059e-06, "loss": 1.7334, "step": 19594 }, { "epoch": 3.73, "learning_rate": 1.8820781696854147e-06, "loss": 1.3206, "step": 19595 }, { "epoch": 3.73, "learning_rate": 1.8820209723546235e-06, "loss": 2.2644, "step": 19596 }, { "epoch": 3.73, "learning_rate": 1.8819637750238323e-06, "loss": 1.929, "step": 19597 }, { "epoch": 3.73, "learning_rate": 1.8819065776930413e-06, "loss": 3.809, "step": 19598 }, { "epoch": 3.73, "learning_rate": 1.8818493803622498e-06, "loss": 2.4401, "step": 19599 }, { "epoch": 3.73, "learning_rate": 1.8817921830314584e-06, "loss": 2.1433, "step": 19600 }, { "epoch": 3.73, "learning_rate": 1.8817349857006674e-06, "loss": 3.0135, "step": 19601 }, { "epoch": 3.73, "learning_rate": 1.8816777883698762e-06, "loss": 2.6153, "step": 19602 }, { "epoch": 3.73, "learning_rate": 1.8816205910390848e-06, "loss": 1.7075, "step": 19603 }, { "epoch": 3.73, "learning_rate": 1.8815633937082938e-06, "loss": 2.9399, "step": 19604 }, { "epoch": 3.73, "learning_rate": 1.8815061963775024e-06, "loss": 2.3407, "step": 19605 }, { "epoch": 3.73, "learning_rate": 1.881448999046711e-06, "loss": 2.4197, "step": 19606 }, { "epoch": 3.73, "learning_rate": 1.88139180171592e-06, "loss": 3.3545, "step": 19607 }, { "epoch": 3.73, "learning_rate": 1.8813346043851288e-06, "loss": 3.5131, "step": 19608 }, { "epoch": 3.74, "learning_rate": 1.8812774070543374e-06, "loss": 1.8443, "step": 19609 }, { "epoch": 3.74, "learning_rate": 1.8812202097235464e-06, "loss": 1.7902, "step": 19610 }, { "epoch": 3.74, "learning_rate": 1.881163012392755e-06, "loss": 2.964, "step": 19611 }, { "epoch": 3.74, "learning_rate": 1.8811058150619637e-06, "loss": 3.1762, "step": 19612 }, { "epoch": 3.74, "learning_rate": 1.8810486177311727e-06, "loss": 2.0987, "step": 19613 }, { "epoch": 3.74, "learning_rate": 1.8809914204003813e-06, "loss": 1.594, "step": 19614 }, { "epoch": 3.74, "learning_rate": 1.8809342230695903e-06, "loss": 1.8208, "step": 19615 }, { "epoch": 3.74, "learning_rate": 1.880877025738799e-06, "loss": 2.9481, "step": 19616 }, { "epoch": 3.74, "learning_rate": 1.8808198284080075e-06, "loss": 0.9147, "step": 19617 }, { "epoch": 3.74, "learning_rate": 1.8807626310772165e-06, "loss": 2.432, "step": 19618 }, { "epoch": 3.74, "learning_rate": 1.8807054337464253e-06, "loss": 2.6937, "step": 19619 }, { "epoch": 3.74, "learning_rate": 1.8806482364156339e-06, "loss": 3.001, "step": 19620 }, { "epoch": 3.74, "learning_rate": 1.8805910390848429e-06, "loss": 2.7961, "step": 19621 }, { "epoch": 3.74, "learning_rate": 1.8805338417540515e-06, "loss": 1.6374, "step": 19622 }, { "epoch": 3.74, "learning_rate": 1.8804766444232603e-06, "loss": 2.3751, "step": 19623 }, { "epoch": 3.74, "learning_rate": 1.880419447092469e-06, "loss": 1.2652, "step": 19624 }, { "epoch": 3.74, "learning_rate": 1.8803622497616779e-06, "loss": 2.2653, "step": 19625 }, { "epoch": 3.74, "learning_rate": 1.8803050524308864e-06, "loss": 1.898, "step": 19626 }, { "epoch": 3.74, "learning_rate": 1.8802478551000954e-06, "loss": 2.5408, "step": 19627 }, { "epoch": 3.74, "learning_rate": 1.880190657769304e-06, "loss": 2.4332, "step": 19628 }, { "epoch": 3.74, "learning_rate": 1.8801334604385128e-06, "loss": 1.8752, "step": 19629 }, { "epoch": 3.74, "learning_rate": 1.8800762631077218e-06, "loss": 2.4212, "step": 19630 }, { "epoch": 3.74, "learning_rate": 1.8800190657769304e-06, "loss": 2.4626, "step": 19631 }, { "epoch": 3.74, "learning_rate": 1.8799618684461394e-06, "loss": 2.5494, "step": 19632 }, { "epoch": 3.74, "learning_rate": 1.879904671115348e-06, "loss": 2.8289, "step": 19633 }, { "epoch": 3.74, "learning_rate": 1.8798474737845568e-06, "loss": 2.0687, "step": 19634 }, { "epoch": 3.74, "learning_rate": 1.8797902764537656e-06, "loss": 2.253, "step": 19635 }, { "epoch": 3.74, "learning_rate": 1.8797330791229744e-06, "loss": 2.0417, "step": 19636 }, { "epoch": 3.74, "learning_rate": 1.879675881792183e-06, "loss": 2.1899, "step": 19637 }, { "epoch": 3.74, "learning_rate": 1.879618684461392e-06, "loss": 2.5094, "step": 19638 }, { "epoch": 3.74, "learning_rate": 1.8795614871306005e-06, "loss": 2.7494, "step": 19639 }, { "epoch": 3.74, "learning_rate": 1.8795042897998093e-06, "loss": 2.2516, "step": 19640 }, { "epoch": 3.74, "learning_rate": 1.8794470924690183e-06, "loss": 1.8957, "step": 19641 }, { "epoch": 3.74, "learning_rate": 1.879389895138227e-06, "loss": 1.8588, "step": 19642 }, { "epoch": 3.74, "learning_rate": 1.8793326978074355e-06, "loss": 1.5208, "step": 19643 }, { "epoch": 3.74, "learning_rate": 1.8792755004766445e-06, "loss": 4.0504, "step": 19644 }, { "epoch": 3.74, "learning_rate": 1.879218303145853e-06, "loss": 2.0167, "step": 19645 }, { "epoch": 3.74, "learning_rate": 1.879161105815062e-06, "loss": 2.9474, "step": 19646 }, { "epoch": 3.74, "learning_rate": 1.8791039084842709e-06, "loss": 1.988, "step": 19647 }, { "epoch": 3.74, "learning_rate": 1.8790467111534795e-06, "loss": 2.5349, "step": 19648 }, { "epoch": 3.74, "learning_rate": 1.8789895138226885e-06, "loss": 3.3129, "step": 19649 }, { "epoch": 3.74, "learning_rate": 1.878932316491897e-06, "loss": 1.8666, "step": 19650 }, { "epoch": 3.74, "learning_rate": 1.8788751191611059e-06, "loss": 3.6911, "step": 19651 }, { "epoch": 3.74, "learning_rate": 1.8788179218303149e-06, "loss": 1.8833, "step": 19652 }, { "epoch": 3.74, "learning_rate": 1.8787607244995234e-06, "loss": 2.3537, "step": 19653 }, { "epoch": 3.74, "learning_rate": 1.878703527168732e-06, "loss": 2.1029, "step": 19654 }, { "epoch": 3.74, "learning_rate": 1.878646329837941e-06, "loss": 3.6068, "step": 19655 }, { "epoch": 3.74, "learning_rate": 1.8785891325071496e-06, "loss": 2.6908, "step": 19656 }, { "epoch": 3.74, "learning_rate": 1.8785319351763584e-06, "loss": 1.9697, "step": 19657 }, { "epoch": 3.74, "learning_rate": 1.8784747378455674e-06, "loss": 2.708, "step": 19658 }, { "epoch": 3.74, "learning_rate": 1.878417540514776e-06, "loss": 2.4383, "step": 19659 }, { "epoch": 3.74, "learning_rate": 1.8783603431839846e-06, "loss": 3.857, "step": 19660 }, { "epoch": 3.74, "learning_rate": 1.8783031458531936e-06, "loss": 2.084, "step": 19661 }, { "epoch": 3.75, "learning_rate": 1.8782459485224024e-06, "loss": 1.9333, "step": 19662 }, { "epoch": 3.75, "learning_rate": 1.8781887511916112e-06, "loss": 3.1712, "step": 19663 }, { "epoch": 3.75, "learning_rate": 1.87813155386082e-06, "loss": 1.7326, "step": 19664 }, { "epoch": 3.75, "learning_rate": 1.8780743565300285e-06, "loss": 1.8265, "step": 19665 }, { "epoch": 3.75, "learning_rate": 1.8780171591992376e-06, "loss": 2.1094, "step": 19666 }, { "epoch": 3.75, "learning_rate": 1.8779599618684461e-06, "loss": 3.5892, "step": 19667 }, { "epoch": 3.75, "learning_rate": 1.877902764537655e-06, "loss": 1.7824, "step": 19668 }, { "epoch": 3.75, "learning_rate": 1.877845567206864e-06, "loss": 2.9209, "step": 19669 }, { "epoch": 3.75, "learning_rate": 1.8777883698760725e-06, "loss": 2.3517, "step": 19670 }, { "epoch": 3.75, "learning_rate": 1.877731172545281e-06, "loss": 2.5419, "step": 19671 }, { "epoch": 3.75, "learning_rate": 1.87767397521449e-06, "loss": 1.5897, "step": 19672 }, { "epoch": 3.75, "learning_rate": 1.8776167778836987e-06, "loss": 2.1255, "step": 19673 }, { "epoch": 3.75, "learning_rate": 1.8775595805529075e-06, "loss": 2.0766, "step": 19674 }, { "epoch": 3.75, "learning_rate": 1.8775023832221165e-06, "loss": 1.7959, "step": 19675 }, { "epoch": 3.75, "learning_rate": 1.877445185891325e-06, "loss": 1.7127, "step": 19676 }, { "epoch": 3.75, "learning_rate": 1.8773879885605336e-06, "loss": 2.9137, "step": 19677 }, { "epoch": 3.75, "learning_rate": 1.8773307912297427e-06, "loss": 1.9189, "step": 19678 }, { "epoch": 3.75, "learning_rate": 1.8772735938989514e-06, "loss": 3.1042, "step": 19679 }, { "epoch": 3.75, "learning_rate": 1.8772163965681605e-06, "loss": 2.3301, "step": 19680 }, { "epoch": 3.75, "learning_rate": 1.877159199237369e-06, "loss": 2.8661, "step": 19681 }, { "epoch": 3.75, "learning_rate": 1.8771020019065776e-06, "loss": 2.3444, "step": 19682 }, { "epoch": 3.75, "learning_rate": 1.8770448045757866e-06, "loss": 1.4719, "step": 19683 }, { "epoch": 3.75, "learning_rate": 1.8769876072449952e-06, "loss": 2.1232, "step": 19684 }, { "epoch": 3.75, "learning_rate": 1.876930409914204e-06, "loss": 2.917, "step": 19685 }, { "epoch": 3.75, "learning_rate": 1.876873212583413e-06, "loss": 1.7986, "step": 19686 }, { "epoch": 3.75, "learning_rate": 1.8768160152526216e-06, "loss": 1.843, "step": 19687 }, { "epoch": 3.75, "learning_rate": 1.8767588179218302e-06, "loss": 2.0733, "step": 19688 }, { "epoch": 3.75, "learning_rate": 1.8767016205910392e-06, "loss": 2.7999, "step": 19689 }, { "epoch": 3.75, "learning_rate": 1.876644423260248e-06, "loss": 2.8136, "step": 19690 }, { "epoch": 3.75, "learning_rate": 1.8765872259294565e-06, "loss": 2.8717, "step": 19691 }, { "epoch": 3.75, "learning_rate": 1.8765300285986656e-06, "loss": 2.008, "step": 19692 }, { "epoch": 3.75, "learning_rate": 1.8764728312678741e-06, "loss": 2.314, "step": 19693 }, { "epoch": 3.75, "learning_rate": 1.8764156339370827e-06, "loss": 2.193, "step": 19694 }, { "epoch": 3.75, "learning_rate": 1.8763584366062917e-06, "loss": 1.8306, "step": 19695 }, { "epoch": 3.75, "learning_rate": 1.8763012392755005e-06, "loss": 1.9292, "step": 19696 }, { "epoch": 3.75, "learning_rate": 1.8762440419447095e-06, "loss": 2.1867, "step": 19697 }, { "epoch": 3.75, "learning_rate": 1.876186844613918e-06, "loss": 1.9134, "step": 19698 }, { "epoch": 3.75, "learning_rate": 1.8761296472831267e-06, "loss": 2.3355, "step": 19699 }, { "epoch": 3.75, "learning_rate": 1.8760724499523357e-06, "loss": 2.2078, "step": 19700 }, { "epoch": 3.75, "learning_rate": 1.8760152526215445e-06, "loss": 2.4369, "step": 19701 }, { "epoch": 3.75, "learning_rate": 1.875958055290753e-06, "loss": 2.4097, "step": 19702 }, { "epoch": 3.75, "learning_rate": 1.875900857959962e-06, "loss": 2.7753, "step": 19703 }, { "epoch": 3.75, "learning_rate": 1.8758436606291707e-06, "loss": 3.3339, "step": 19704 }, { "epoch": 3.75, "learning_rate": 1.8757864632983792e-06, "loss": 2.9228, "step": 19705 }, { "epoch": 3.75, "learning_rate": 1.8757292659675882e-06, "loss": 1.8797, "step": 19706 }, { "epoch": 3.75, "learning_rate": 1.875672068636797e-06, "loss": 2.3927, "step": 19707 }, { "epoch": 3.75, "learning_rate": 1.8756148713060056e-06, "loss": 2.514, "step": 19708 }, { "epoch": 3.75, "learning_rate": 1.8755576739752146e-06, "loss": 3.0132, "step": 19709 }, { "epoch": 3.75, "learning_rate": 1.8755004766444232e-06, "loss": 2.373, "step": 19710 }, { "epoch": 3.75, "learning_rate": 1.8754432793136322e-06, "loss": 1.786, "step": 19711 }, { "epoch": 3.75, "learning_rate": 1.8753860819828408e-06, "loss": 3.1717, "step": 19712 }, { "epoch": 3.75, "learning_rate": 1.8753288846520496e-06, "loss": 2.1011, "step": 19713 }, { "epoch": 3.76, "learning_rate": 1.8752716873212586e-06, "loss": 2.0978, "step": 19714 }, { "epoch": 3.76, "learning_rate": 1.8752144899904672e-06, "loss": 2.7141, "step": 19715 }, { "epoch": 3.76, "learning_rate": 1.8751572926596758e-06, "loss": 1.714, "step": 19716 }, { "epoch": 3.76, "learning_rate": 1.8751000953288848e-06, "loss": 2.2371, "step": 19717 }, { "epoch": 3.76, "learning_rate": 1.8750428979980936e-06, "loss": 2.2484, "step": 19718 }, { "epoch": 3.76, "learning_rate": 1.8749857006673021e-06, "loss": 2.4214, "step": 19719 }, { "epoch": 3.76, "learning_rate": 1.8749285033365111e-06, "loss": 2.5391, "step": 19720 }, { "epoch": 3.76, "learning_rate": 1.8748713060057197e-06, "loss": 2.8966, "step": 19721 }, { "epoch": 3.76, "learning_rate": 1.8748141086749285e-06, "loss": 2.2013, "step": 19722 }, { "epoch": 3.76, "learning_rate": 1.8747569113441373e-06, "loss": 1.4766, "step": 19723 }, { "epoch": 3.76, "learning_rate": 1.8746997140133461e-06, "loss": 1.676, "step": 19724 }, { "epoch": 3.76, "learning_rate": 1.8746425166825547e-06, "loss": 2.6248, "step": 19725 }, { "epoch": 3.76, "learning_rate": 1.8745853193517637e-06, "loss": 1.8234, "step": 19726 }, { "epoch": 3.76, "learning_rate": 1.8745281220209723e-06, "loss": 2.7584, "step": 19727 }, { "epoch": 3.76, "learning_rate": 1.8744709246901813e-06, "loss": 2.0217, "step": 19728 }, { "epoch": 3.76, "learning_rate": 1.87441372735939e-06, "loss": 1.7778, "step": 19729 }, { "epoch": 3.76, "learning_rate": 1.8743565300285987e-06, "loss": 1.8634, "step": 19730 }, { "epoch": 3.76, "learning_rate": 1.8742993326978077e-06, "loss": 2.9492, "step": 19731 }, { "epoch": 3.76, "learning_rate": 1.8742421353670162e-06, "loss": 1.3766, "step": 19732 }, { "epoch": 3.76, "learning_rate": 1.8741849380362248e-06, "loss": 1.7726, "step": 19733 }, { "epoch": 3.76, "learning_rate": 1.8741277407054338e-06, "loss": 2.7586, "step": 19734 }, { "epoch": 3.76, "learning_rate": 1.8740705433746426e-06, "loss": 2.9775, "step": 19735 }, { "epoch": 3.76, "learning_rate": 1.8740133460438512e-06, "loss": 1.7339, "step": 19736 }, { "epoch": 3.76, "learning_rate": 1.8739561487130602e-06, "loss": 1.3767, "step": 19737 }, { "epoch": 3.76, "learning_rate": 1.8738989513822688e-06, "loss": 2.2483, "step": 19738 }, { "epoch": 3.76, "learning_rate": 1.8738417540514776e-06, "loss": 3.1264, "step": 19739 }, { "epoch": 3.76, "learning_rate": 1.8737845567206864e-06, "loss": 0.9341, "step": 19740 }, { "epoch": 3.76, "learning_rate": 1.8737273593898952e-06, "loss": 1.8688, "step": 19741 }, { "epoch": 3.76, "learning_rate": 1.8736701620591038e-06, "loss": 2.4577, "step": 19742 }, { "epoch": 3.76, "learning_rate": 1.8736129647283128e-06, "loss": 3.1179, "step": 19743 }, { "epoch": 3.76, "learning_rate": 1.8735557673975214e-06, "loss": 2.9282, "step": 19744 }, { "epoch": 3.76, "learning_rate": 1.8734985700667304e-06, "loss": 1.6513, "step": 19745 }, { "epoch": 3.76, "learning_rate": 1.8734413727359391e-06, "loss": 1.7418, "step": 19746 }, { "epoch": 3.76, "learning_rate": 1.8733841754051477e-06, "loss": 1.4375, "step": 19747 }, { "epoch": 3.76, "learning_rate": 1.8733269780743567e-06, "loss": 1.1656, "step": 19748 }, { "epoch": 3.76, "learning_rate": 1.8732697807435653e-06, "loss": 2.0412, "step": 19749 }, { "epoch": 3.76, "learning_rate": 1.8732125834127741e-06, "loss": 2.1952, "step": 19750 }, { "epoch": 3.76, "learning_rate": 1.873155386081983e-06, "loss": 2.6133, "step": 19751 }, { "epoch": 3.76, "learning_rate": 1.8730981887511917e-06, "loss": 2.3403, "step": 19752 }, { "epoch": 3.76, "learning_rate": 1.8730409914204003e-06, "loss": 2.2512, "step": 19753 }, { "epoch": 3.76, "learning_rate": 1.8729837940896093e-06, "loss": 1.9337, "step": 19754 }, { "epoch": 3.76, "learning_rate": 1.8729265967588179e-06, "loss": 1.1413, "step": 19755 }, { "epoch": 3.76, "learning_rate": 1.8728693994280267e-06, "loss": 2.0771, "step": 19756 }, { "epoch": 3.76, "learning_rate": 1.8728122020972357e-06, "loss": 1.6797, "step": 19757 }, { "epoch": 3.76, "learning_rate": 1.8727550047664443e-06, "loss": 1.1113, "step": 19758 }, { "epoch": 3.76, "learning_rate": 1.8726978074356528e-06, "loss": 2.3462, "step": 19759 }, { "epoch": 3.76, "learning_rate": 1.8726406101048618e-06, "loss": 4.0009, "step": 19760 }, { "epoch": 3.76, "learning_rate": 1.8725834127740704e-06, "loss": 2.5535, "step": 19761 }, { "epoch": 3.76, "learning_rate": 1.8725262154432794e-06, "loss": 3.1124, "step": 19762 }, { "epoch": 3.76, "learning_rate": 1.8724690181124882e-06, "loss": 2.411, "step": 19763 }, { "epoch": 3.76, "learning_rate": 1.8724118207816968e-06, "loss": 2.0292, "step": 19764 }, { "epoch": 3.76, "learning_rate": 1.8723546234509058e-06, "loss": 2.8919, "step": 19765 }, { "epoch": 3.76, "learning_rate": 1.8722974261201144e-06, "loss": 2.3804, "step": 19766 }, { "epoch": 3.77, "learning_rate": 1.8722402287893232e-06, "loss": 3.1393, "step": 19767 }, { "epoch": 3.77, "learning_rate": 1.8721830314585322e-06, "loss": 2.1631, "step": 19768 }, { "epoch": 3.77, "learning_rate": 1.8721258341277408e-06, "loss": 1.6005, "step": 19769 }, { "epoch": 3.77, "learning_rate": 1.8720686367969494e-06, "loss": 2.8807, "step": 19770 }, { "epoch": 3.77, "learning_rate": 1.8720114394661584e-06, "loss": 2.8981, "step": 19771 }, { "epoch": 3.77, "learning_rate": 1.871954242135367e-06, "loss": 2.0744, "step": 19772 }, { "epoch": 3.77, "learning_rate": 1.8718970448045757e-06, "loss": 2.1239, "step": 19773 }, { "epoch": 3.77, "learning_rate": 1.8718398474737847e-06, "loss": 2.8407, "step": 19774 }, { "epoch": 3.77, "learning_rate": 1.8717826501429933e-06, "loss": 2.2388, "step": 19775 }, { "epoch": 3.77, "learning_rate": 1.8717254528122023e-06, "loss": 2.3633, "step": 19776 }, { "epoch": 3.77, "learning_rate": 1.871668255481411e-06, "loss": 2.5285, "step": 19777 }, { "epoch": 3.77, "learning_rate": 1.8716110581506197e-06, "loss": 2.4468, "step": 19778 }, { "epoch": 3.77, "learning_rate": 1.8715538608198285e-06, "loss": 1.5056, "step": 19779 }, { "epoch": 3.77, "learning_rate": 1.8714966634890373e-06, "loss": 1.7613, "step": 19780 }, { "epoch": 3.77, "learning_rate": 1.8714394661582459e-06, "loss": 2.7362, "step": 19781 }, { "epoch": 3.77, "learning_rate": 1.8713822688274549e-06, "loss": 2.8181, "step": 19782 }, { "epoch": 3.77, "learning_rate": 1.8713250714966635e-06, "loss": 2.9155, "step": 19783 }, { "epoch": 3.77, "learning_rate": 1.8712678741658723e-06, "loss": 3.0649, "step": 19784 }, { "epoch": 3.77, "learning_rate": 1.8712106768350813e-06, "loss": 2.8036, "step": 19785 }, { "epoch": 3.77, "learning_rate": 1.8711534795042898e-06, "loss": 1.7157, "step": 19786 }, { "epoch": 3.77, "learning_rate": 1.8710962821734984e-06, "loss": 2.892, "step": 19787 }, { "epoch": 3.77, "learning_rate": 1.8710390848427074e-06, "loss": 1.7332, "step": 19788 }, { "epoch": 3.77, "learning_rate": 1.8709818875119162e-06, "loss": 2.1418, "step": 19789 }, { "epoch": 3.77, "learning_rate": 1.8709246901811248e-06, "loss": 2.5349, "step": 19790 }, { "epoch": 3.77, "learning_rate": 1.8708674928503338e-06, "loss": 3.427, "step": 19791 }, { "epoch": 3.77, "learning_rate": 1.8708102955195424e-06, "loss": 4.3914, "step": 19792 }, { "epoch": 3.77, "learning_rate": 1.8707530981887514e-06, "loss": 2.8455, "step": 19793 }, { "epoch": 3.77, "learning_rate": 1.87069590085796e-06, "loss": 2.1473, "step": 19794 }, { "epoch": 3.77, "learning_rate": 1.8706387035271688e-06, "loss": 2.3354, "step": 19795 }, { "epoch": 3.77, "learning_rate": 1.8705815061963778e-06, "loss": 1.5934, "step": 19796 }, { "epoch": 3.77, "learning_rate": 1.8705243088655864e-06, "loss": 2.551, "step": 19797 }, { "epoch": 3.77, "learning_rate": 1.870467111534795e-06, "loss": 3.1069, "step": 19798 }, { "epoch": 3.77, "learning_rate": 1.870409914204004e-06, "loss": 3.8748, "step": 19799 }, { "epoch": 3.77, "learning_rate": 1.8703527168732125e-06, "loss": 2.2662, "step": 19800 }, { "epoch": 3.77, "learning_rate": 1.8702955195424213e-06, "loss": 2.0393, "step": 19801 }, { "epoch": 3.77, "learning_rate": 1.8702383222116303e-06, "loss": 2.4456, "step": 19802 }, { "epoch": 3.77, "learning_rate": 1.870181124880839e-06, "loss": 2.2796, "step": 19803 }, { "epoch": 3.77, "learning_rate": 1.8701239275500475e-06, "loss": 2.6404, "step": 19804 }, { "epoch": 3.77, "learning_rate": 1.8700667302192565e-06, "loss": 2.2421, "step": 19805 }, { "epoch": 3.77, "learning_rate": 1.8700095328884653e-06, "loss": 1.8416, "step": 19806 }, { "epoch": 3.77, "learning_rate": 1.8699523355576739e-06, "loss": 2.0252, "step": 19807 }, { "epoch": 3.77, "learning_rate": 1.8698951382268829e-06, "loss": 2.8551, "step": 19808 }, { "epoch": 3.77, "learning_rate": 1.8698379408960915e-06, "loss": 2.612, "step": 19809 }, { "epoch": 3.77, "learning_rate": 1.8697807435653005e-06, "loss": 1.7753, "step": 19810 }, { "epoch": 3.77, "learning_rate": 1.869723546234509e-06, "loss": 1.5846, "step": 19811 }, { "epoch": 3.77, "learning_rate": 1.8696663489037178e-06, "loss": 2.1394, "step": 19812 }, { "epoch": 3.77, "learning_rate": 1.8696091515729269e-06, "loss": 1.5382, "step": 19813 }, { "epoch": 3.77, "learning_rate": 1.8695519542421354e-06, "loss": 2.4829, "step": 19814 }, { "epoch": 3.77, "learning_rate": 1.869494756911344e-06, "loss": 1.7299, "step": 19815 }, { "epoch": 3.77, "learning_rate": 1.869437559580553e-06, "loss": 2.3255, "step": 19816 }, { "epoch": 3.77, "learning_rate": 1.8693803622497618e-06, "loss": 1.6363, "step": 19817 }, { "epoch": 3.77, "learning_rate": 1.8693231649189704e-06, "loss": 1.4374, "step": 19818 }, { "epoch": 3.78, "learning_rate": 1.8692659675881794e-06, "loss": 3.1706, "step": 19819 }, { "epoch": 3.78, "learning_rate": 1.869208770257388e-06, "loss": 2.6484, "step": 19820 }, { "epoch": 3.78, "learning_rate": 1.8691515729265966e-06, "loss": 1.5885, "step": 19821 }, { "epoch": 3.78, "learning_rate": 1.8690943755958056e-06, "loss": 2.4486, "step": 19822 }, { "epoch": 3.78, "learning_rate": 1.8690371782650144e-06, "loss": 1.9575, "step": 19823 }, { "epoch": 3.78, "learning_rate": 1.868979980934223e-06, "loss": 2.4653, "step": 19824 }, { "epoch": 3.78, "learning_rate": 1.868922783603432e-06, "loss": 2.0366, "step": 19825 }, { "epoch": 3.78, "learning_rate": 1.8688655862726405e-06, "loss": 1.2878, "step": 19826 }, { "epoch": 3.78, "learning_rate": 1.8688083889418495e-06, "loss": 1.5482, "step": 19827 }, { "epoch": 3.78, "learning_rate": 1.8687511916110581e-06, "loss": 3.5804, "step": 19828 }, { "epoch": 3.78, "learning_rate": 1.868693994280267e-06, "loss": 1.6822, "step": 19829 }, { "epoch": 3.78, "learning_rate": 1.868636796949476e-06, "loss": 3.3462, "step": 19830 }, { "epoch": 3.78, "learning_rate": 1.8685795996186845e-06, "loss": 1.8785, "step": 19831 }, { "epoch": 3.78, "learning_rate": 1.868522402287893e-06, "loss": 2.0612, "step": 19832 }, { "epoch": 3.78, "learning_rate": 1.868465204957102e-06, "loss": 2.4625, "step": 19833 }, { "epoch": 3.78, "learning_rate": 1.8684080076263109e-06, "loss": 2.823, "step": 19834 }, { "epoch": 3.78, "learning_rate": 1.8683508102955195e-06, "loss": 3.3599, "step": 19835 }, { "epoch": 3.78, "learning_rate": 1.8682936129647285e-06, "loss": 1.5988, "step": 19836 }, { "epoch": 3.78, "learning_rate": 1.868236415633937e-06, "loss": 1.9615, "step": 19837 }, { "epoch": 3.78, "learning_rate": 1.8681792183031459e-06, "loss": 1.2524, "step": 19838 }, { "epoch": 3.78, "learning_rate": 1.8681220209723546e-06, "loss": 2.0855, "step": 19839 }, { "epoch": 3.78, "learning_rate": 1.8680648236415634e-06, "loss": 3.6223, "step": 19840 }, { "epoch": 3.78, "learning_rate": 1.8680076263107724e-06, "loss": 1.7448, "step": 19841 }, { "epoch": 3.78, "learning_rate": 1.867950428979981e-06, "loss": 1.7942, "step": 19842 }, { "epoch": 3.78, "learning_rate": 1.8678932316491896e-06, "loss": 1.2691, "step": 19843 }, { "epoch": 3.78, "learning_rate": 1.8678360343183986e-06, "loss": 2.6649, "step": 19844 }, { "epoch": 3.78, "learning_rate": 1.8677788369876074e-06, "loss": 2.6072, "step": 19845 }, { "epoch": 3.78, "learning_rate": 1.867721639656816e-06, "loss": 2.3589, "step": 19846 }, { "epoch": 3.78, "learning_rate": 1.867664442326025e-06, "loss": 2.8314, "step": 19847 }, { "epoch": 3.78, "learning_rate": 1.8676072449952336e-06, "loss": 2.4653, "step": 19848 }, { "epoch": 3.78, "learning_rate": 1.8675500476644422e-06, "loss": 2.2028, "step": 19849 }, { "epoch": 3.78, "learning_rate": 1.8674928503336512e-06, "loss": 2.5222, "step": 19850 }, { "epoch": 3.78, "learning_rate": 1.86743565300286e-06, "loss": 1.6214, "step": 19851 }, { "epoch": 3.78, "learning_rate": 1.8673784556720685e-06, "loss": 1.8278, "step": 19852 }, { "epoch": 3.78, "learning_rate": 1.8673212583412775e-06, "loss": 3.2678, "step": 19853 }, { "epoch": 3.78, "learning_rate": 1.8672640610104861e-06, "loss": 3.5756, "step": 19854 }, { "epoch": 3.78, "learning_rate": 1.867206863679695e-06, "loss": 1.2587, "step": 19855 }, { "epoch": 3.78, "learning_rate": 1.867149666348904e-06, "loss": 2.9158, "step": 19856 }, { "epoch": 3.78, "learning_rate": 1.8670924690181125e-06, "loss": 2.2837, "step": 19857 }, { "epoch": 3.78, "learning_rate": 1.8670352716873215e-06, "loss": 2.7008, "step": 19858 }, { "epoch": 3.78, "learning_rate": 1.86697807435653e-06, "loss": 2.4713, "step": 19859 }, { "epoch": 3.78, "learning_rate": 1.8669208770257387e-06, "loss": 2.0014, "step": 19860 }, { "epoch": 3.78, "learning_rate": 1.8668636796949477e-06, "loss": 1.793, "step": 19861 }, { "epoch": 3.78, "learning_rate": 1.8668064823641565e-06, "loss": 3.1821, "step": 19862 }, { "epoch": 3.78, "learning_rate": 1.866749285033365e-06, "loss": 1.8303, "step": 19863 }, { "epoch": 3.78, "learning_rate": 1.866692087702574e-06, "loss": 3.2188, "step": 19864 }, { "epoch": 3.78, "learning_rate": 1.8666348903717826e-06, "loss": 1.5711, "step": 19865 }, { "epoch": 3.78, "learning_rate": 1.8665776930409914e-06, "loss": 2.9502, "step": 19866 }, { "epoch": 3.78, "learning_rate": 1.8665204957102002e-06, "loss": 2.0142, "step": 19867 }, { "epoch": 3.78, "learning_rate": 1.866463298379409e-06, "loss": 2.7087, "step": 19868 }, { "epoch": 3.78, "learning_rate": 1.8664061010486176e-06, "loss": 2.5868, "step": 19869 }, { "epoch": 3.78, "learning_rate": 1.8663489037178266e-06, "loss": 1.3447, "step": 19870 }, { "epoch": 3.78, "learning_rate": 1.8662917063870352e-06, "loss": 1.569, "step": 19871 }, { "epoch": 3.79, "learning_rate": 1.866234509056244e-06, "loss": 2.1861, "step": 19872 }, { "epoch": 3.79, "learning_rate": 1.866177311725453e-06, "loss": 2.9217, "step": 19873 }, { "epoch": 3.79, "learning_rate": 1.8661201143946616e-06, "loss": 2.7844, "step": 19874 }, { "epoch": 3.79, "learning_rate": 1.8660629170638706e-06, "loss": 1.5653, "step": 19875 }, { "epoch": 3.79, "learning_rate": 1.8660057197330792e-06, "loss": 2.0006, "step": 19876 }, { "epoch": 3.79, "learning_rate": 1.865948522402288e-06, "loss": 2.2961, "step": 19877 }, { "epoch": 3.79, "learning_rate": 1.8658913250714968e-06, "loss": 2.5404, "step": 19878 }, { "epoch": 3.79, "learning_rate": 1.8658341277407056e-06, "loss": 2.5605, "step": 19879 }, { "epoch": 3.79, "learning_rate": 1.8657769304099141e-06, "loss": 2.8792, "step": 19880 }, { "epoch": 3.79, "learning_rate": 1.8657197330791231e-06, "loss": 1.7693, "step": 19881 }, { "epoch": 3.79, "learning_rate": 1.8656625357483317e-06, "loss": 2.0178, "step": 19882 }, { "epoch": 3.79, "learning_rate": 1.8656053384175405e-06, "loss": 2.0299, "step": 19883 }, { "epoch": 3.79, "learning_rate": 1.8655481410867495e-06, "loss": 2.0254, "step": 19884 }, { "epoch": 3.79, "learning_rate": 1.865490943755958e-06, "loss": 1.6273, "step": 19885 }, { "epoch": 3.79, "learning_rate": 1.8654337464251667e-06, "loss": 2.0055, "step": 19886 }, { "epoch": 3.79, "learning_rate": 1.8653765490943757e-06, "loss": 1.5827, "step": 19887 }, { "epoch": 3.79, "learning_rate": 1.8653193517635843e-06, "loss": 2.3659, "step": 19888 }, { "epoch": 3.79, "learning_rate": 1.865262154432793e-06, "loss": 1.2758, "step": 19889 }, { "epoch": 3.79, "learning_rate": 1.865204957102002e-06, "loss": 2.3218, "step": 19890 }, { "epoch": 3.79, "learning_rate": 1.8651477597712107e-06, "loss": 3.5944, "step": 19891 }, { "epoch": 3.79, "learning_rate": 1.8650905624404197e-06, "loss": 1.7794, "step": 19892 }, { "epoch": 3.79, "learning_rate": 1.8650333651096282e-06, "loss": 1.9165, "step": 19893 }, { "epoch": 3.79, "learning_rate": 1.864976167778837e-06, "loss": 3.2939, "step": 19894 }, { "epoch": 3.79, "learning_rate": 1.864918970448046e-06, "loss": 2.5156, "step": 19895 }, { "epoch": 3.79, "learning_rate": 1.8648617731172546e-06, "loss": 2.2404, "step": 19896 }, { "epoch": 3.79, "learning_rate": 1.8648045757864632e-06, "loss": 2.5476, "step": 19897 }, { "epoch": 3.79, "learning_rate": 1.8647473784556722e-06, "loss": 2.7471, "step": 19898 }, { "epoch": 3.79, "learning_rate": 1.8646901811248808e-06, "loss": 1.2606, "step": 19899 }, { "epoch": 3.79, "learning_rate": 1.8646329837940896e-06, "loss": 1.676, "step": 19900 }, { "epoch": 3.79, "learning_rate": 1.8645757864632986e-06, "loss": 1.5217, "step": 19901 }, { "epoch": 3.79, "learning_rate": 1.8645185891325072e-06, "loss": 1.8455, "step": 19902 }, { "epoch": 3.79, "learning_rate": 1.8644613918017158e-06, "loss": 1.5244, "step": 19903 }, { "epoch": 3.79, "learning_rate": 1.8644041944709248e-06, "loss": 3.4278, "step": 19904 }, { "epoch": 3.79, "learning_rate": 1.8643469971401336e-06, "loss": 2.986, "step": 19905 }, { "epoch": 3.79, "learning_rate": 1.8642897998093423e-06, "loss": 2.2461, "step": 19906 }, { "epoch": 3.79, "learning_rate": 1.8642326024785511e-06, "loss": 2.1445, "step": 19907 }, { "epoch": 3.79, "learning_rate": 1.8641754051477597e-06, "loss": 1.2843, "step": 19908 }, { "epoch": 3.79, "learning_rate": 1.8641182078169687e-06, "loss": 1.992, "step": 19909 }, { "epoch": 3.79, "learning_rate": 1.8640610104861773e-06, "loss": 2.6652, "step": 19910 }, { "epoch": 3.79, "learning_rate": 1.8640038131553861e-06, "loss": 2.2574, "step": 19911 }, { "epoch": 3.79, "learning_rate": 1.8639466158245951e-06, "loss": 2.5082, "step": 19912 }, { "epoch": 3.79, "learning_rate": 1.8638894184938037e-06, "loss": 1.5825, "step": 19913 }, { "epoch": 3.79, "learning_rate": 1.8638322211630123e-06, "loss": 1.9008, "step": 19914 }, { "epoch": 3.79, "learning_rate": 1.8637750238322213e-06, "loss": 2.1914, "step": 19915 }, { "epoch": 3.79, "learning_rate": 1.8637178265014299e-06, "loss": 2.0227, "step": 19916 }, { "epoch": 3.79, "learning_rate": 1.8636606291706387e-06, "loss": 2.048, "step": 19917 }, { "epoch": 3.79, "learning_rate": 1.8636034318398477e-06, "loss": 2.3584, "step": 19918 }, { "epoch": 3.79, "learning_rate": 1.8635462345090562e-06, "loss": 2.6765, "step": 19919 }, { "epoch": 3.79, "learning_rate": 1.8634890371782648e-06, "loss": 1.7967, "step": 19920 }, { "epoch": 3.79, "learning_rate": 1.8634318398474738e-06, "loss": 2.3946, "step": 19921 }, { "epoch": 3.79, "learning_rate": 1.8633746425166826e-06, "loss": 2.5263, "step": 19922 }, { "epoch": 3.79, "learning_rate": 1.8633174451858916e-06, "loss": 2.2072, "step": 19923 }, { "epoch": 3.8, "learning_rate": 1.8632602478551002e-06, "loss": 3.9307, "step": 19924 }, { "epoch": 3.8, "learning_rate": 1.8632030505243088e-06, "loss": 1.4391, "step": 19925 }, { "epoch": 3.8, "learning_rate": 1.8631458531935178e-06, "loss": 2.1198, "step": 19926 }, { "epoch": 3.8, "learning_rate": 1.8630886558627264e-06, "loss": 2.3725, "step": 19927 }, { "epoch": 3.8, "learning_rate": 1.8630314585319352e-06, "loss": 2.416, "step": 19928 }, { "epoch": 3.8, "learning_rate": 1.8629742612011442e-06, "loss": 1.968, "step": 19929 }, { "epoch": 3.8, "learning_rate": 1.8629170638703528e-06, "loss": 2.0723, "step": 19930 }, { "epoch": 3.8, "learning_rate": 1.8628598665395613e-06, "loss": 2.19, "step": 19931 }, { "epoch": 3.8, "learning_rate": 1.8628026692087704e-06, "loss": 3.1344, "step": 19932 }, { "epoch": 3.8, "learning_rate": 1.8627454718779791e-06, "loss": 2.9396, "step": 19933 }, { "epoch": 3.8, "learning_rate": 1.8626882745471877e-06, "loss": 2.6246, "step": 19934 }, { "epoch": 3.8, "learning_rate": 1.8626310772163967e-06, "loss": 2.6517, "step": 19935 }, { "epoch": 3.8, "learning_rate": 1.8625738798856053e-06, "loss": 3.1491, "step": 19936 }, { "epoch": 3.8, "learning_rate": 1.862516682554814e-06, "loss": 1.6109, "step": 19937 }, { "epoch": 3.8, "learning_rate": 1.862459485224023e-06, "loss": 2.0995, "step": 19938 }, { "epoch": 3.8, "learning_rate": 1.8624022878932317e-06, "loss": 2.2069, "step": 19939 }, { "epoch": 3.8, "learning_rate": 1.8623450905624407e-06, "loss": 3.4725, "step": 19940 }, { "epoch": 3.8, "learning_rate": 1.8622878932316493e-06, "loss": 2.4456, "step": 19941 }, { "epoch": 3.8, "learning_rate": 1.8622306959008579e-06, "loss": 1.8863, "step": 19942 }, { "epoch": 3.8, "learning_rate": 1.8621734985700669e-06, "loss": 2.9629, "step": 19943 }, { "epoch": 3.8, "learning_rate": 1.8621163012392757e-06, "loss": 2.1488, "step": 19944 }, { "epoch": 3.8, "learning_rate": 1.8620591039084842e-06, "loss": 1.832, "step": 19945 }, { "epoch": 3.8, "learning_rate": 1.8620019065776933e-06, "loss": 2.9201, "step": 19946 }, { "epoch": 3.8, "learning_rate": 1.8619447092469018e-06, "loss": 1.2436, "step": 19947 }, { "epoch": 3.8, "learning_rate": 1.8618875119161104e-06, "loss": 1.7119, "step": 19948 }, { "epoch": 3.8, "learning_rate": 1.8618303145853194e-06, "loss": 2.4406, "step": 19949 }, { "epoch": 3.8, "learning_rate": 1.8617731172545282e-06, "loss": 1.831, "step": 19950 }, { "epoch": 3.8, "learning_rate": 1.8617159199237368e-06, "loss": 2.2968, "step": 19951 }, { "epoch": 3.8, "learning_rate": 1.8616587225929458e-06, "loss": 2.6893, "step": 19952 }, { "epoch": 3.8, "learning_rate": 1.8616015252621544e-06, "loss": 2.0647, "step": 19953 }, { "epoch": 3.8, "learning_rate": 1.8615443279313632e-06, "loss": 2.7328, "step": 19954 }, { "epoch": 3.8, "learning_rate": 1.861487130600572e-06, "loss": 1.8793, "step": 19955 }, { "epoch": 3.8, "learning_rate": 1.8614299332697808e-06, "loss": 2.7114, "step": 19956 }, { "epoch": 3.8, "learning_rate": 1.8613727359389898e-06, "loss": 1.2422, "step": 19957 }, { "epoch": 3.8, "learning_rate": 1.8613155386081984e-06, "loss": 2.1447, "step": 19958 }, { "epoch": 3.8, "learning_rate": 1.861258341277407e-06, "loss": 3.6103, "step": 19959 }, { "epoch": 3.8, "learning_rate": 1.861201143946616e-06, "loss": 2.0664, "step": 19960 }, { "epoch": 3.8, "learning_rate": 1.8611439466158247e-06, "loss": 3.4439, "step": 19961 }, { "epoch": 3.8, "learning_rate": 1.8610867492850333e-06, "loss": 1.7614, "step": 19962 }, { "epoch": 3.8, "learning_rate": 1.8610295519542423e-06, "loss": 2.122, "step": 19963 }, { "epoch": 3.8, "learning_rate": 1.860972354623451e-06, "loss": 2.5183, "step": 19964 }, { "epoch": 3.8, "learning_rate": 1.8609151572926597e-06, "loss": 1.7765, "step": 19965 }, { "epoch": 3.8, "learning_rate": 1.8608579599618685e-06, "loss": 2.0463, "step": 19966 }, { "epoch": 3.8, "learning_rate": 1.8608007626310773e-06, "loss": 1.9531, "step": 19967 }, { "epoch": 3.8, "learning_rate": 1.8607435653002859e-06, "loss": 1.8722, "step": 19968 }, { "epoch": 3.8, "learning_rate": 1.8606863679694949e-06, "loss": 1.0573, "step": 19969 }, { "epoch": 3.8, "learning_rate": 1.8606291706387035e-06, "loss": 2.7401, "step": 19970 }, { "epoch": 3.8, "learning_rate": 1.8605719733079125e-06, "loss": 2.8598, "step": 19971 }, { "epoch": 3.8, "learning_rate": 1.8605147759771213e-06, "loss": 2.8998, "step": 19972 }, { "epoch": 3.8, "learning_rate": 1.8604575786463298e-06, "loss": 2.7687, "step": 19973 }, { "epoch": 3.8, "learning_rate": 1.8604003813155388e-06, "loss": 2.3874, "step": 19974 }, { "epoch": 3.8, "learning_rate": 1.8603431839847474e-06, "loss": 2.6519, "step": 19975 }, { "epoch": 3.8, "learning_rate": 1.860285986653956e-06, "loss": 1.9083, "step": 19976 }, { "epoch": 3.81, "learning_rate": 1.860228789323165e-06, "loss": 2.2539, "step": 19977 }, { "epoch": 3.81, "learning_rate": 1.8601715919923738e-06, "loss": 3.2333, "step": 19978 }, { "epoch": 3.81, "learning_rate": 1.8601143946615824e-06, "loss": 1.9914, "step": 19979 }, { "epoch": 3.81, "learning_rate": 1.8600571973307914e-06, "loss": 2.0999, "step": 19980 }, { "epoch": 3.81, "learning_rate": 1.86e-06, "loss": 3.0269, "step": 19981 }, { "epoch": 3.81, "learning_rate": 1.8599428026692088e-06, "loss": 3.0808, "step": 19982 }, { "epoch": 3.81, "learning_rate": 1.8598856053384176e-06, "loss": 2.0096, "step": 19983 }, { "epoch": 3.81, "learning_rate": 1.8598284080076264e-06, "loss": 1.2596, "step": 19984 }, { "epoch": 3.81, "learning_rate": 1.859771210676835e-06, "loss": 2.3552, "step": 19985 }, { "epoch": 3.81, "learning_rate": 1.859714013346044e-06, "loss": 2.2496, "step": 19986 }, { "epoch": 3.81, "learning_rate": 1.8596568160152525e-06, "loss": 2.2537, "step": 19987 }, { "epoch": 3.81, "learning_rate": 1.8595996186844615e-06, "loss": 1.669, "step": 19988 }, { "epoch": 3.81, "learning_rate": 1.8595424213536703e-06, "loss": 3.5412, "step": 19989 }, { "epoch": 3.81, "learning_rate": 1.859485224022879e-06, "loss": 3.2862, "step": 19990 }, { "epoch": 3.81, "learning_rate": 1.859428026692088e-06, "loss": 1.6766, "step": 19991 }, { "epoch": 3.81, "learning_rate": 1.8593708293612965e-06, "loss": 2.9929, "step": 19992 }, { "epoch": 3.81, "learning_rate": 1.8593136320305053e-06, "loss": 3.7836, "step": 19993 }, { "epoch": 3.81, "learning_rate": 1.859256434699714e-06, "loss": 2.2043, "step": 19994 }, { "epoch": 3.81, "learning_rate": 1.8591992373689229e-06, "loss": 3.4385, "step": 19995 }, { "epoch": 3.81, "learning_rate": 1.8591420400381315e-06, "loss": 2.3433, "step": 19996 }, { "epoch": 3.81, "learning_rate": 1.8590848427073405e-06, "loss": 1.7373, "step": 19997 }, { "epoch": 3.81, "learning_rate": 1.859027645376549e-06, "loss": 2.3355, "step": 19998 }, { "epoch": 3.81, "learning_rate": 1.8589704480457578e-06, "loss": 2.4207, "step": 19999 }, { "epoch": 3.81, "learning_rate": 1.8589132507149669e-06, "loss": 2.0834, "step": 20000 }, { "epoch": 3.81, "learning_rate": 1.8588560533841754e-06, "loss": 3.1099, "step": 20001 }, { "epoch": 3.81, "learning_rate": 1.858798856053384e-06, "loss": 3.4169, "step": 20002 }, { "epoch": 3.81, "learning_rate": 1.858741658722593e-06, "loss": 1.3705, "step": 20003 }, { "epoch": 3.81, "learning_rate": 1.8586844613918016e-06, "loss": 2.1962, "step": 20004 }, { "epoch": 3.81, "learning_rate": 1.8586272640610106e-06, "loss": 1.7473, "step": 20005 }, { "epoch": 3.81, "learning_rate": 1.8585700667302194e-06, "loss": 1.9794, "step": 20006 }, { "epoch": 3.81, "learning_rate": 1.858512869399428e-06, "loss": 1.1927, "step": 20007 }, { "epoch": 3.81, "learning_rate": 1.858455672068637e-06, "loss": 2.582, "step": 20008 }, { "epoch": 3.81, "learning_rate": 1.8583984747378456e-06, "loss": 2.4466, "step": 20009 }, { "epoch": 3.81, "learning_rate": 1.8583412774070544e-06, "loss": 1.6083, "step": 20010 }, { "epoch": 3.81, "learning_rate": 1.8582840800762634e-06, "loss": 1.8971, "step": 20011 }, { "epoch": 3.81, "learning_rate": 1.858226882745472e-06, "loss": 2.3583, "step": 20012 }, { "epoch": 3.81, "learning_rate": 1.8581696854146805e-06, "loss": 1.9805, "step": 20013 }, { "epoch": 3.81, "learning_rate": 1.8581124880838895e-06, "loss": 3.1728, "step": 20014 }, { "epoch": 3.81, "learning_rate": 1.8580552907530981e-06, "loss": 1.7922, "step": 20015 }, { "epoch": 3.81, "learning_rate": 1.857998093422307e-06, "loss": 2.2952, "step": 20016 }, { "epoch": 3.81, "learning_rate": 1.857940896091516e-06, "loss": 1.5955, "step": 20017 }, { "epoch": 3.81, "learning_rate": 1.8578836987607245e-06, "loss": 2.3995, "step": 20018 }, { "epoch": 3.81, "learning_rate": 1.857826501429933e-06, "loss": 2.3253, "step": 20019 }, { "epoch": 3.81, "learning_rate": 1.857769304099142e-06, "loss": 1.8036, "step": 20020 }, { "epoch": 3.81, "learning_rate": 1.8577121067683509e-06, "loss": 2.4124, "step": 20021 }, { "epoch": 3.81, "learning_rate": 1.8576549094375597e-06, "loss": 1.6781, "step": 20022 }, { "epoch": 3.81, "learning_rate": 1.8575977121067685e-06, "loss": 1.733, "step": 20023 }, { "epoch": 3.81, "learning_rate": 1.857540514775977e-06, "loss": 2.113, "step": 20024 }, { "epoch": 3.81, "learning_rate": 1.857483317445186e-06, "loss": 0.9631, "step": 20025 }, { "epoch": 3.81, "learning_rate": 1.8574261201143946e-06, "loss": 2.0283, "step": 20026 }, { "epoch": 3.81, "learning_rate": 1.8573689227836034e-06, "loss": 2.1973, "step": 20027 }, { "epoch": 3.81, "learning_rate": 1.8573117254528124e-06, "loss": 2.9538, "step": 20028 }, { "epoch": 3.82, "learning_rate": 1.857254528122021e-06, "loss": 2.5459, "step": 20029 }, { "epoch": 3.82, "learning_rate": 1.8571973307912296e-06, "loss": 1.5507, "step": 20030 }, { "epoch": 3.82, "learning_rate": 1.8571401334604386e-06, "loss": 2.5531, "step": 20031 }, { "epoch": 3.82, "learning_rate": 1.8570829361296474e-06, "loss": 2.3696, "step": 20032 }, { "epoch": 3.82, "learning_rate": 1.857025738798856e-06, "loss": 3.282, "step": 20033 }, { "epoch": 3.82, "learning_rate": 1.856968541468065e-06, "loss": 1.3609, "step": 20034 }, { "epoch": 3.82, "learning_rate": 1.8569113441372736e-06, "loss": 2.4187, "step": 20035 }, { "epoch": 3.82, "learning_rate": 1.8568541468064826e-06, "loss": 1.939, "step": 20036 }, { "epoch": 3.82, "learning_rate": 1.8567969494756912e-06, "loss": 2.4217, "step": 20037 }, { "epoch": 3.82, "learning_rate": 1.8567397521449e-06, "loss": 1.4303, "step": 20038 }, { "epoch": 3.82, "learning_rate": 1.856682554814109e-06, "loss": 1.3072, "step": 20039 }, { "epoch": 3.82, "learning_rate": 1.8566253574833175e-06, "loss": 1.8555, "step": 20040 }, { "epoch": 3.82, "learning_rate": 1.8565681601525261e-06, "loss": 1.5897, "step": 20041 }, { "epoch": 3.82, "learning_rate": 1.8565109628217351e-06, "loss": 1.3952, "step": 20042 }, { "epoch": 3.82, "learning_rate": 1.8564537654909437e-06, "loss": 2.7421, "step": 20043 }, { "epoch": 3.82, "learning_rate": 1.8563965681601525e-06, "loss": 3.0171, "step": 20044 }, { "epoch": 3.82, "learning_rate": 1.8563393708293615e-06, "loss": 2.6482, "step": 20045 }, { "epoch": 3.82, "learning_rate": 1.85628217349857e-06, "loss": 2.4433, "step": 20046 }, { "epoch": 3.82, "learning_rate": 1.8562249761677787e-06, "loss": 2.76, "step": 20047 }, { "epoch": 3.82, "learning_rate": 1.8561677788369877e-06, "loss": 2.6337, "step": 20048 }, { "epoch": 3.82, "learning_rate": 1.8561105815061965e-06, "loss": 2.0532, "step": 20049 }, { "epoch": 3.82, "learning_rate": 1.856053384175405e-06, "loss": 2.2491, "step": 20050 }, { "epoch": 3.82, "learning_rate": 1.855996186844614e-06, "loss": 1.4535, "step": 20051 }, { "epoch": 3.82, "learning_rate": 1.8559389895138226e-06, "loss": 1.5854, "step": 20052 }, { "epoch": 3.82, "learning_rate": 1.8558817921830317e-06, "loss": 3.0741, "step": 20053 }, { "epoch": 3.82, "learning_rate": 1.8558245948522402e-06, "loss": 2.5153, "step": 20054 }, { "epoch": 3.82, "learning_rate": 1.855767397521449e-06, "loss": 1.5223, "step": 20055 }, { "epoch": 3.82, "learning_rate": 1.855710200190658e-06, "loss": 3.7841, "step": 20056 }, { "epoch": 3.82, "learning_rate": 1.8556530028598666e-06, "loss": 1.8452, "step": 20057 }, { "epoch": 3.82, "learning_rate": 1.8555958055290752e-06, "loss": 1.8079, "step": 20058 }, { "epoch": 3.82, "learning_rate": 1.8555386081982842e-06, "loss": 1.6552, "step": 20059 }, { "epoch": 3.82, "learning_rate": 1.855481410867493e-06, "loss": 2.8322, "step": 20060 }, { "epoch": 3.82, "learning_rate": 1.8554242135367016e-06, "loss": 1.6843, "step": 20061 }, { "epoch": 3.82, "learning_rate": 1.8553670162059106e-06, "loss": 1.8974, "step": 20062 }, { "epoch": 3.82, "learning_rate": 1.8553098188751192e-06, "loss": 1.6191, "step": 20063 }, { "epoch": 3.82, "learning_rate": 1.8552526215443277e-06, "loss": 1.0767, "step": 20064 }, { "epoch": 3.82, "learning_rate": 1.8551954242135368e-06, "loss": 1.9067, "step": 20065 }, { "epoch": 3.82, "learning_rate": 1.8551382268827455e-06, "loss": 3.4556, "step": 20066 }, { "epoch": 3.82, "learning_rate": 1.8550810295519541e-06, "loss": 2.4537, "step": 20067 }, { "epoch": 3.82, "learning_rate": 1.8550238322211631e-06, "loss": 2.6685, "step": 20068 }, { "epoch": 3.82, "learning_rate": 1.8549666348903717e-06, "loss": 2.6939, "step": 20069 }, { "epoch": 3.82, "learning_rate": 1.8549094375595807e-06, "loss": 2.837, "step": 20070 }, { "epoch": 3.82, "learning_rate": 1.8548522402287893e-06, "loss": 2.0478, "step": 20071 }, { "epoch": 3.82, "learning_rate": 1.854795042897998e-06, "loss": 2.88, "step": 20072 }, { "epoch": 3.82, "learning_rate": 1.854737845567207e-06, "loss": 3.0777, "step": 20073 }, { "epoch": 3.82, "learning_rate": 1.8546806482364157e-06, "loss": 2.7399, "step": 20074 }, { "epoch": 3.82, "learning_rate": 1.8546234509056243e-06, "loss": 2.4818, "step": 20075 }, { "epoch": 3.82, "learning_rate": 1.8545662535748333e-06, "loss": 2.9443, "step": 20076 }, { "epoch": 3.82, "learning_rate": 1.854509056244042e-06, "loss": 3.4268, "step": 20077 }, { "epoch": 3.82, "learning_rate": 1.8544518589132507e-06, "loss": 2.149, "step": 20078 }, { "epoch": 3.82, "learning_rate": 1.8543946615824597e-06, "loss": 2.7608, "step": 20079 }, { "epoch": 3.82, "learning_rate": 1.8543374642516682e-06, "loss": 1.1839, "step": 20080 }, { "epoch": 3.82, "learning_rate": 1.854280266920877e-06, "loss": 2.162, "step": 20081 }, { "epoch": 3.83, "learning_rate": 1.8542230695900858e-06, "loss": 2.0656, "step": 20082 }, { "epoch": 3.83, "learning_rate": 1.8541658722592946e-06, "loss": 1.8569, "step": 20083 }, { "epoch": 3.83, "learning_rate": 1.8541086749285032e-06, "loss": 2.4139, "step": 20084 }, { "epoch": 3.83, "learning_rate": 1.8540514775977122e-06, "loss": 2.1236, "step": 20085 }, { "epoch": 3.83, "learning_rate": 1.8539942802669208e-06, "loss": 2.3716, "step": 20086 }, { "epoch": 3.83, "learning_rate": 1.8539370829361298e-06, "loss": 2.175, "step": 20087 }, { "epoch": 3.83, "learning_rate": 1.8538798856053386e-06, "loss": 3.0835, "step": 20088 }, { "epoch": 3.83, "learning_rate": 1.8538226882745472e-06, "loss": 1.2587, "step": 20089 }, { "epoch": 3.83, "learning_rate": 1.8537654909437562e-06, "loss": 1.9644, "step": 20090 }, { "epoch": 3.83, "learning_rate": 1.8537082936129648e-06, "loss": 2.7622, "step": 20091 }, { "epoch": 3.83, "learning_rate": 1.8536510962821733e-06, "loss": 2.0716, "step": 20092 }, { "epoch": 3.83, "learning_rate": 1.8535938989513823e-06, "loss": 2.3105, "step": 20093 }, { "epoch": 3.83, "learning_rate": 1.8535367016205911e-06, "loss": 2.3737, "step": 20094 }, { "epoch": 3.83, "learning_rate": 1.8534795042897997e-06, "loss": 1.8508, "step": 20095 }, { "epoch": 3.83, "learning_rate": 1.8534223069590087e-06, "loss": 1.4118, "step": 20096 }, { "epoch": 3.83, "learning_rate": 1.8533651096282173e-06, "loss": 1.5802, "step": 20097 }, { "epoch": 3.83, "learning_rate": 1.853307912297426e-06, "loss": 1.1826, "step": 20098 }, { "epoch": 3.83, "learning_rate": 1.8532507149666351e-06, "loss": 1.5994, "step": 20099 }, { "epoch": 3.83, "learning_rate": 1.8531935176358437e-06, "loss": 1.6545, "step": 20100 }, { "epoch": 3.83, "learning_rate": 1.8531363203050523e-06, "loss": 3.1368, "step": 20101 }, { "epoch": 3.83, "learning_rate": 1.8530791229742613e-06, "loss": 3.1078, "step": 20102 }, { "epoch": 3.83, "learning_rate": 1.8530219256434699e-06, "loss": 2.2306, "step": 20103 }, { "epoch": 3.83, "learning_rate": 1.8529647283126789e-06, "loss": 1.5689, "step": 20104 }, { "epoch": 3.83, "learning_rate": 1.8529075309818877e-06, "loss": 1.9611, "step": 20105 }, { "epoch": 3.83, "learning_rate": 1.8528503336510962e-06, "loss": 2.6597, "step": 20106 }, { "epoch": 3.83, "learning_rate": 1.8527931363203052e-06, "loss": 2.0279, "step": 20107 }, { "epoch": 3.83, "learning_rate": 1.8527359389895138e-06, "loss": 1.4215, "step": 20108 }, { "epoch": 3.83, "learning_rate": 1.8526787416587226e-06, "loss": 2.3495, "step": 20109 }, { "epoch": 3.83, "learning_rate": 1.8526215443279314e-06, "loss": 2.4945, "step": 20110 }, { "epoch": 3.83, "learning_rate": 1.8525643469971402e-06, "loss": 1.8856, "step": 20111 }, { "epoch": 3.83, "learning_rate": 1.8525071496663488e-06, "loss": 1.4422, "step": 20112 }, { "epoch": 3.83, "learning_rate": 1.8524499523355578e-06, "loss": 1.5161, "step": 20113 }, { "epoch": 3.83, "learning_rate": 1.8523927550047664e-06, "loss": 3.3645, "step": 20114 }, { "epoch": 3.83, "learning_rate": 1.8523355576739752e-06, "loss": 2.5226, "step": 20115 }, { "epoch": 3.83, "learning_rate": 1.8522783603431842e-06, "loss": 2.4096, "step": 20116 }, { "epoch": 3.83, "learning_rate": 1.8522211630123928e-06, "loss": 2.1878, "step": 20117 }, { "epoch": 3.83, "learning_rate": 1.8521639656816018e-06, "loss": 2.2166, "step": 20118 }, { "epoch": 3.83, "learning_rate": 1.8521067683508103e-06, "loss": 2.5736, "step": 20119 }, { "epoch": 3.83, "learning_rate": 1.8520495710200191e-06, "loss": 2.0901, "step": 20120 }, { "epoch": 3.83, "learning_rate": 1.851992373689228e-06, "loss": 1.7695, "step": 20121 }, { "epoch": 3.83, "learning_rate": 1.8519351763584367e-06, "loss": 1.484, "step": 20122 }, { "epoch": 3.83, "learning_rate": 1.8518779790276453e-06, "loss": 2.5614, "step": 20123 }, { "epoch": 3.83, "learning_rate": 1.8518207816968543e-06, "loss": 2.0033, "step": 20124 }, { "epoch": 3.83, "learning_rate": 1.851763584366063e-06, "loss": 1.7984, "step": 20125 }, { "epoch": 3.83, "learning_rate": 1.8517063870352717e-06, "loss": 4.0094, "step": 20126 }, { "epoch": 3.83, "learning_rate": 1.8516491897044807e-06, "loss": 1.2306, "step": 20127 }, { "epoch": 3.83, "learning_rate": 1.8515919923736893e-06, "loss": 3.3056, "step": 20128 }, { "epoch": 3.83, "learning_rate": 1.8515347950428979e-06, "loss": 2.4417, "step": 20129 }, { "epoch": 3.83, "learning_rate": 1.8514775977121069e-06, "loss": 1.9914, "step": 20130 }, { "epoch": 3.83, "learning_rate": 1.8514204003813155e-06, "loss": 1.4756, "step": 20131 }, { "epoch": 3.83, "learning_rate": 1.8513632030505242e-06, "loss": 3.2648, "step": 20132 }, { "epoch": 3.83, "learning_rate": 1.8513060057197333e-06, "loss": 2.2686, "step": 20133 }, { "epoch": 3.84, "learning_rate": 1.8512488083889418e-06, "loss": 2.1206, "step": 20134 }, { "epoch": 3.84, "learning_rate": 1.8511916110581508e-06, "loss": 2.3749, "step": 20135 }, { "epoch": 3.84, "learning_rate": 1.8511344137273594e-06, "loss": 1.9726, "step": 20136 }, { "epoch": 3.84, "learning_rate": 1.8510772163965682e-06, "loss": 2.7352, "step": 20137 }, { "epoch": 3.84, "learning_rate": 1.851020019065777e-06, "loss": 1.8431, "step": 20138 }, { "epoch": 3.84, "learning_rate": 1.8509628217349858e-06, "loss": 2.7551, "step": 20139 }, { "epoch": 3.84, "learning_rate": 1.8509056244041944e-06, "loss": 2.463, "step": 20140 }, { "epoch": 3.84, "learning_rate": 1.8508484270734034e-06, "loss": 2.015, "step": 20141 }, { "epoch": 3.84, "learning_rate": 1.850791229742612e-06, "loss": 2.8907, "step": 20142 }, { "epoch": 3.84, "learning_rate": 1.8507340324118208e-06, "loss": 3.5851, "step": 20143 }, { "epoch": 3.84, "learning_rate": 1.8506768350810298e-06, "loss": 2.9016, "step": 20144 }, { "epoch": 3.84, "learning_rate": 1.8506196377502384e-06, "loss": 3.0366, "step": 20145 }, { "epoch": 3.84, "learning_rate": 1.850562440419447e-06, "loss": 2.2985, "step": 20146 }, { "epoch": 3.84, "learning_rate": 1.850505243088656e-06, "loss": 1.5131, "step": 20147 }, { "epoch": 3.84, "learning_rate": 1.8504480457578647e-06, "loss": 2.4996, "step": 20148 }, { "epoch": 3.84, "learning_rate": 1.8503908484270733e-06, "loss": 2.0626, "step": 20149 }, { "epoch": 3.84, "learning_rate": 1.8503336510962823e-06, "loss": 1.7325, "step": 20150 }, { "epoch": 3.84, "learning_rate": 1.850276453765491e-06, "loss": 2.4134, "step": 20151 }, { "epoch": 3.84, "learning_rate": 1.8502192564347e-06, "loss": 2.1793, "step": 20152 }, { "epoch": 3.84, "learning_rate": 1.8501620591039085e-06, "loss": 2.5715, "step": 20153 }, { "epoch": 3.84, "learning_rate": 1.8501048617731173e-06, "loss": 2.1727, "step": 20154 }, { "epoch": 3.84, "learning_rate": 1.8500476644423263e-06, "loss": 2.0169, "step": 20155 }, { "epoch": 3.84, "learning_rate": 1.8499904671115349e-06, "loss": 3.2411, "step": 20156 }, { "epoch": 3.84, "learning_rate": 1.8499332697807435e-06, "loss": 1.9874, "step": 20157 }, { "epoch": 3.84, "learning_rate": 1.8498760724499525e-06, "loss": 2.9971, "step": 20158 }, { "epoch": 3.84, "learning_rate": 1.849818875119161e-06, "loss": 2.7951, "step": 20159 }, { "epoch": 3.84, "learning_rate": 1.8497616777883698e-06, "loss": 3.1378, "step": 20160 }, { "epoch": 3.84, "learning_rate": 1.8497044804575788e-06, "loss": 2.1477, "step": 20161 }, { "epoch": 3.84, "learning_rate": 1.8496472831267874e-06, "loss": 2.2404, "step": 20162 }, { "epoch": 3.84, "learning_rate": 1.849590085795996e-06, "loss": 2.8726, "step": 20163 }, { "epoch": 3.84, "learning_rate": 1.849532888465205e-06, "loss": 2.6981, "step": 20164 }, { "epoch": 3.84, "learning_rate": 1.8494756911344138e-06, "loss": 2.8715, "step": 20165 }, { "epoch": 3.84, "learning_rate": 1.8494184938036224e-06, "loss": 2.9127, "step": 20166 }, { "epoch": 3.84, "learning_rate": 1.8493612964728314e-06, "loss": 2.3351, "step": 20167 }, { "epoch": 3.84, "learning_rate": 1.84930409914204e-06, "loss": 2.3589, "step": 20168 }, { "epoch": 3.84, "learning_rate": 1.849246901811249e-06, "loss": 1.71, "step": 20169 }, { "epoch": 3.84, "learning_rate": 1.8491897044804576e-06, "loss": 4.1992, "step": 20170 }, { "epoch": 3.84, "learning_rate": 1.8491325071496664e-06, "loss": 2.7414, "step": 20171 }, { "epoch": 3.84, "learning_rate": 1.8490753098188754e-06, "loss": 2.2854, "step": 20172 }, { "epoch": 3.84, "learning_rate": 1.849018112488084e-06, "loss": 2.3699, "step": 20173 }, { "epoch": 3.84, "learning_rate": 1.8489609151572925e-06, "loss": 2.5808, "step": 20174 }, { "epoch": 3.84, "learning_rate": 1.8489037178265015e-06, "loss": 2.7769, "step": 20175 }, { "epoch": 3.84, "learning_rate": 1.8488465204957103e-06, "loss": 1.9271, "step": 20176 }, { "epoch": 3.84, "learning_rate": 1.848789323164919e-06, "loss": 1.7823, "step": 20177 }, { "epoch": 3.84, "learning_rate": 1.848732125834128e-06, "loss": 2.5442, "step": 20178 }, { "epoch": 3.84, "learning_rate": 1.8486749285033365e-06, "loss": 2.8218, "step": 20179 }, { "epoch": 3.84, "learning_rate": 1.848617731172545e-06, "loss": 2.249, "step": 20180 }, { "epoch": 3.84, "learning_rate": 1.848560533841754e-06, "loss": 2.0541, "step": 20181 }, { "epoch": 3.84, "learning_rate": 1.8485033365109629e-06, "loss": 2.9354, "step": 20182 }, { "epoch": 3.84, "learning_rate": 1.8484461391801719e-06, "loss": 1.5221, "step": 20183 }, { "epoch": 3.84, "learning_rate": 1.8483889418493805e-06, "loss": 1.8037, "step": 20184 }, { "epoch": 3.84, "learning_rate": 1.848331744518589e-06, "loss": 1.6479, "step": 20185 }, { "epoch": 3.84, "learning_rate": 1.848274547187798e-06, "loss": 2.4936, "step": 20186 }, { "epoch": 3.85, "learning_rate": 1.8482173498570068e-06, "loss": 2.0676, "step": 20187 }, { "epoch": 3.85, "learning_rate": 1.8481601525262154e-06, "loss": 2.0247, "step": 20188 }, { "epoch": 3.85, "learning_rate": 1.8481029551954244e-06, "loss": 2.6289, "step": 20189 }, { "epoch": 3.85, "learning_rate": 1.848045757864633e-06, "loss": 2.9047, "step": 20190 }, { "epoch": 3.85, "learning_rate": 1.8479885605338416e-06, "loss": 1.8719, "step": 20191 }, { "epoch": 3.85, "learning_rate": 1.8479313632030506e-06, "loss": 2.6696, "step": 20192 }, { "epoch": 3.85, "learning_rate": 1.8478741658722594e-06, "loss": 2.464, "step": 20193 }, { "epoch": 3.85, "learning_rate": 1.847816968541468e-06, "loss": 3.1617, "step": 20194 }, { "epoch": 3.85, "learning_rate": 1.847759771210677e-06, "loss": 1.3988, "step": 20195 }, { "epoch": 3.85, "learning_rate": 1.8477025738798856e-06, "loss": 2.2887, "step": 20196 }, { "epoch": 3.85, "learning_rate": 1.8476453765490944e-06, "loss": 2.764, "step": 20197 }, { "epoch": 3.85, "learning_rate": 1.8475881792183032e-06, "loss": 2.5312, "step": 20198 }, { "epoch": 3.85, "learning_rate": 1.847530981887512e-06, "loss": 3.1839, "step": 20199 }, { "epoch": 3.85, "learning_rate": 1.847473784556721e-06, "loss": 1.8756, "step": 20200 }, { "epoch": 3.85, "learning_rate": 1.8474165872259295e-06, "loss": 2.6769, "step": 20201 }, { "epoch": 3.85, "learning_rate": 1.8473593898951381e-06, "loss": 2.3633, "step": 20202 }, { "epoch": 3.85, "learning_rate": 1.8473021925643471e-06, "loss": 2.598, "step": 20203 }, { "epoch": 3.85, "learning_rate": 1.847244995233556e-06, "loss": 1.2675, "step": 20204 }, { "epoch": 3.85, "learning_rate": 1.8471877979027645e-06, "loss": 2.994, "step": 20205 }, { "epoch": 3.85, "learning_rate": 1.8471306005719735e-06, "loss": 2.8819, "step": 20206 }, { "epoch": 3.85, "learning_rate": 1.847073403241182e-06, "loss": 3.3164, "step": 20207 }, { "epoch": 3.85, "learning_rate": 1.8470162059103907e-06, "loss": 2.7455, "step": 20208 }, { "epoch": 3.85, "learning_rate": 1.8469590085795997e-06, "loss": 3.5421, "step": 20209 }, { "epoch": 3.85, "learning_rate": 1.8469018112488085e-06, "loss": 3.3778, "step": 20210 }, { "epoch": 3.85, "learning_rate": 1.846844613918017e-06, "loss": 2.4345, "step": 20211 }, { "epoch": 3.85, "learning_rate": 1.846787416587226e-06, "loss": 2.7364, "step": 20212 }, { "epoch": 3.85, "learning_rate": 1.8467302192564346e-06, "loss": 2.5136, "step": 20213 }, { "epoch": 3.85, "learning_rate": 1.8466730219256434e-06, "loss": 2.4951, "step": 20214 }, { "epoch": 3.85, "learning_rate": 1.8466158245948524e-06, "loss": 2.4309, "step": 20215 }, { "epoch": 3.85, "learning_rate": 1.846558627264061e-06, "loss": 2.18, "step": 20216 }, { "epoch": 3.85, "learning_rate": 1.84650142993327e-06, "loss": 1.7767, "step": 20217 }, { "epoch": 3.85, "learning_rate": 1.8464442326024786e-06, "loss": 3.6336, "step": 20218 }, { "epoch": 3.85, "learning_rate": 1.8463870352716872e-06, "loss": 2.3541, "step": 20219 }, { "epoch": 3.85, "learning_rate": 1.8463298379408962e-06, "loss": 1.8993, "step": 20220 }, { "epoch": 3.85, "learning_rate": 1.846272640610105e-06, "loss": 2.8167, "step": 20221 }, { "epoch": 3.85, "learning_rate": 1.8462154432793136e-06, "loss": 3.401, "step": 20222 }, { "epoch": 3.85, "learning_rate": 1.8461582459485226e-06, "loss": 1.9327, "step": 20223 }, { "epoch": 3.85, "learning_rate": 1.8461010486177312e-06, "loss": 3.2568, "step": 20224 }, { "epoch": 3.85, "learning_rate": 1.84604385128694e-06, "loss": 1.8134, "step": 20225 }, { "epoch": 3.85, "learning_rate": 1.8459866539561487e-06, "loss": 2.0311, "step": 20226 }, { "epoch": 3.85, "learning_rate": 1.8459294566253575e-06, "loss": 2.5938, "step": 20227 }, { "epoch": 3.85, "learning_rate": 1.8458722592945661e-06, "loss": 1.8936, "step": 20228 }, { "epoch": 3.85, "learning_rate": 1.8458150619637751e-06, "loss": 1.9139, "step": 20229 }, { "epoch": 3.85, "learning_rate": 1.8457578646329837e-06, "loss": 2.9983, "step": 20230 }, { "epoch": 3.85, "learning_rate": 1.8457006673021925e-06, "loss": 2.8432, "step": 20231 }, { "epoch": 3.85, "learning_rate": 1.8456434699714015e-06, "loss": 2.5477, "step": 20232 }, { "epoch": 3.85, "learning_rate": 1.84558627264061e-06, "loss": 2.5244, "step": 20233 }, { "epoch": 3.85, "learning_rate": 1.845529075309819e-06, "loss": 2.9425, "step": 20234 }, { "epoch": 3.85, "learning_rate": 1.8454718779790277e-06, "loss": 2.7867, "step": 20235 }, { "epoch": 3.85, "learning_rate": 1.8454146806482365e-06, "loss": 3.5459, "step": 20236 }, { "epoch": 3.85, "learning_rate": 1.8453574833174453e-06, "loss": 1.8428, "step": 20237 }, { "epoch": 3.85, "learning_rate": 1.845300285986654e-06, "loss": 3.2915, "step": 20238 }, { "epoch": 3.86, "learning_rate": 1.8452430886558626e-06, "loss": 2.4716, "step": 20239 }, { "epoch": 3.86, "learning_rate": 1.8451858913250716e-06, "loss": 1.9734, "step": 20240 }, { "epoch": 3.86, "learning_rate": 1.8451286939942802e-06, "loss": 2.0929, "step": 20241 }, { "epoch": 3.86, "learning_rate": 1.845071496663489e-06, "loss": 1.8365, "step": 20242 }, { "epoch": 3.86, "learning_rate": 1.845014299332698e-06, "loss": 2.1379, "step": 20243 }, { "epoch": 3.86, "learning_rate": 1.8449571020019066e-06, "loss": 2.0783, "step": 20244 }, { "epoch": 3.86, "learning_rate": 1.8448999046711152e-06, "loss": 3.5572, "step": 20245 }, { "epoch": 3.86, "learning_rate": 1.8448427073403242e-06, "loss": 2.1681, "step": 20246 }, { "epoch": 3.86, "learning_rate": 1.8447855100095328e-06, "loss": 2.282, "step": 20247 }, { "epoch": 3.86, "learning_rate": 1.8447283126787418e-06, "loss": 1.3922, "step": 20248 }, { "epoch": 3.86, "learning_rate": 1.8446711153479506e-06, "loss": 2.3275, "step": 20249 }, { "epoch": 3.86, "learning_rate": 1.8446139180171592e-06, "loss": 2.6358, "step": 20250 }, { "epoch": 3.86, "learning_rate": 1.8445567206863682e-06, "loss": 2.7659, "step": 20251 }, { "epoch": 3.86, "learning_rate": 1.8444995233555768e-06, "loss": 1.5883, "step": 20252 }, { "epoch": 3.86, "learning_rate": 1.8444423260247855e-06, "loss": 2.8452, "step": 20253 }, { "epoch": 3.86, "learning_rate": 1.8443851286939946e-06, "loss": 3.1644, "step": 20254 }, { "epoch": 3.86, "learning_rate": 1.8443279313632031e-06, "loss": 2.3056, "step": 20255 }, { "epoch": 3.86, "learning_rate": 1.8442707340324117e-06, "loss": 2.0377, "step": 20256 }, { "epoch": 3.86, "learning_rate": 1.8442135367016207e-06, "loss": 3.0041, "step": 20257 }, { "epoch": 3.86, "learning_rate": 1.8441563393708293e-06, "loss": 2.2781, "step": 20258 }, { "epoch": 3.86, "learning_rate": 1.844099142040038e-06, "loss": 1.3085, "step": 20259 }, { "epoch": 3.86, "learning_rate": 1.844041944709247e-06, "loss": 1.7613, "step": 20260 }, { "epoch": 3.86, "learning_rate": 1.8439847473784557e-06, "loss": 1.6081, "step": 20261 }, { "epoch": 3.86, "learning_rate": 1.8439275500476643e-06, "loss": 1.6144, "step": 20262 }, { "epoch": 3.86, "learning_rate": 1.8438703527168733e-06, "loss": 2.5652, "step": 20263 }, { "epoch": 3.86, "learning_rate": 1.843813155386082e-06, "loss": 1.4288, "step": 20264 }, { "epoch": 3.86, "learning_rate": 1.8437559580552909e-06, "loss": 2.482, "step": 20265 }, { "epoch": 3.86, "learning_rate": 1.8436987607244997e-06, "loss": 2.0913, "step": 20266 }, { "epoch": 3.86, "learning_rate": 1.8436415633937082e-06, "loss": 2.3871, "step": 20267 }, { "epoch": 3.86, "learning_rate": 1.8435843660629172e-06, "loss": 2.3743, "step": 20268 }, { "epoch": 3.86, "learning_rate": 1.8435271687321258e-06, "loss": 2.3677, "step": 20269 }, { "epoch": 3.86, "learning_rate": 1.8434699714013346e-06, "loss": 2.3547, "step": 20270 }, { "epoch": 3.86, "learning_rate": 1.8434127740705436e-06, "loss": 2.7026, "step": 20271 }, { "epoch": 3.86, "learning_rate": 1.8433555767397522e-06, "loss": 2.143, "step": 20272 }, { "epoch": 3.86, "learning_rate": 1.8432983794089608e-06, "loss": 3.214, "step": 20273 }, { "epoch": 3.86, "learning_rate": 1.8432411820781698e-06, "loss": 2.2188, "step": 20274 }, { "epoch": 3.86, "learning_rate": 1.8431839847473786e-06, "loss": 2.1931, "step": 20275 }, { "epoch": 3.86, "learning_rate": 1.8431267874165872e-06, "loss": 2.4875, "step": 20276 }, { "epoch": 3.86, "learning_rate": 1.8430695900857962e-06, "loss": 2.0812, "step": 20277 }, { "epoch": 3.86, "learning_rate": 1.8430123927550048e-06, "loss": 2.283, "step": 20278 }, { "epoch": 3.86, "learning_rate": 1.8429551954242133e-06, "loss": 2.8146, "step": 20279 }, { "epoch": 3.86, "learning_rate": 1.8428979980934223e-06, "loss": 2.2184, "step": 20280 }, { "epoch": 3.86, "learning_rate": 1.8428408007626311e-06, "loss": 2.5706, "step": 20281 }, { "epoch": 3.86, "learning_rate": 1.8427836034318401e-06, "loss": 1.951, "step": 20282 }, { "epoch": 3.86, "learning_rate": 1.8427264061010487e-06, "loss": 2.5886, "step": 20283 }, { "epoch": 3.86, "learning_rate": 1.8426692087702573e-06, "loss": 1.6287, "step": 20284 }, { "epoch": 3.86, "learning_rate": 1.8426120114394663e-06, "loss": 3.046, "step": 20285 }, { "epoch": 3.86, "learning_rate": 1.8425548141086749e-06, "loss": 1.9532, "step": 20286 }, { "epoch": 3.86, "learning_rate": 1.8424976167778837e-06, "loss": 2.656, "step": 20287 }, { "epoch": 3.86, "learning_rate": 1.8424404194470927e-06, "loss": 1.5471, "step": 20288 }, { "epoch": 3.86, "learning_rate": 1.8423832221163013e-06, "loss": 1.9258, "step": 20289 }, { "epoch": 3.86, "learning_rate": 1.8423260247855099e-06, "loss": 2.4181, "step": 20290 }, { "epoch": 3.86, "learning_rate": 1.8422688274547189e-06, "loss": 2.2237, "step": 20291 }, { "epoch": 3.87, "learning_rate": 1.8422116301239277e-06, "loss": 2.3008, "step": 20292 }, { "epoch": 3.87, "learning_rate": 1.8421544327931362e-06, "loss": 2.414, "step": 20293 }, { "epoch": 3.87, "learning_rate": 1.8420972354623452e-06, "loss": 2.5313, "step": 20294 }, { "epoch": 3.87, "learning_rate": 1.8420400381315538e-06, "loss": 1.986, "step": 20295 }, { "epoch": 3.87, "learning_rate": 1.8419828408007624e-06, "loss": 2.8812, "step": 20296 }, { "epoch": 3.87, "learning_rate": 1.8419256434699714e-06, "loss": 1.4637, "step": 20297 }, { "epoch": 3.87, "learning_rate": 1.8418684461391802e-06, "loss": 2.8526, "step": 20298 }, { "epoch": 3.87, "learning_rate": 1.8418112488083892e-06, "loss": 1.2661, "step": 20299 }, { "epoch": 3.87, "learning_rate": 1.8417540514775978e-06, "loss": 2.6604, "step": 20300 }, { "epoch": 3.87, "learning_rate": 1.8416968541468064e-06, "loss": 2.5261, "step": 20301 }, { "epoch": 3.87, "learning_rate": 1.8416396568160154e-06, "loss": 2.2365, "step": 20302 }, { "epoch": 3.87, "learning_rate": 1.8415824594852242e-06, "loss": 2.6168, "step": 20303 }, { "epoch": 3.87, "learning_rate": 1.8415252621544328e-06, "loss": 1.6792, "step": 20304 }, { "epoch": 3.87, "learning_rate": 1.8414680648236418e-06, "loss": 1.8526, "step": 20305 }, { "epoch": 3.87, "learning_rate": 1.8414108674928503e-06, "loss": 2.7153, "step": 20306 }, { "epoch": 3.87, "learning_rate": 1.841353670162059e-06, "loss": 2.6025, "step": 20307 }, { "epoch": 3.87, "learning_rate": 1.841296472831268e-06, "loss": 1.9688, "step": 20308 }, { "epoch": 3.87, "learning_rate": 1.8412392755004767e-06, "loss": 2.5358, "step": 20309 }, { "epoch": 3.87, "learning_rate": 1.8411820781696853e-06, "loss": 1.8625, "step": 20310 }, { "epoch": 3.87, "learning_rate": 1.8411248808388943e-06, "loss": 2.0536, "step": 20311 }, { "epoch": 3.87, "learning_rate": 1.841067683508103e-06, "loss": 1.9491, "step": 20312 }, { "epoch": 3.87, "learning_rate": 1.841010486177312e-06, "loss": 2.7694, "step": 20313 }, { "epoch": 3.87, "learning_rate": 1.8409532888465205e-06, "loss": 2.9108, "step": 20314 }, { "epoch": 3.87, "learning_rate": 1.8408960915157293e-06, "loss": 2.0397, "step": 20315 }, { "epoch": 3.87, "learning_rate": 1.8408388941849383e-06, "loss": 2.3601, "step": 20316 }, { "epoch": 3.87, "learning_rate": 1.8407816968541469e-06, "loss": 3.039, "step": 20317 }, { "epoch": 3.87, "learning_rate": 1.8407244995233554e-06, "loss": 1.5525, "step": 20318 }, { "epoch": 3.87, "learning_rate": 1.8406673021925645e-06, "loss": 2.2745, "step": 20319 }, { "epoch": 3.87, "learning_rate": 1.8406101048617732e-06, "loss": 2.5432, "step": 20320 }, { "epoch": 3.87, "learning_rate": 1.8405529075309818e-06, "loss": 1.9615, "step": 20321 }, { "epoch": 3.87, "learning_rate": 1.8404957102001908e-06, "loss": 1.7321, "step": 20322 }, { "epoch": 3.87, "learning_rate": 1.8404385128693994e-06, "loss": 3.3962, "step": 20323 }, { "epoch": 3.87, "learning_rate": 1.8403813155386082e-06, "loss": 1.1513, "step": 20324 }, { "epoch": 3.87, "learning_rate": 1.840324118207817e-06, "loss": 2.4329, "step": 20325 }, { "epoch": 3.87, "learning_rate": 1.8402669208770258e-06, "loss": 2.7894, "step": 20326 }, { "epoch": 3.87, "learning_rate": 1.8402097235462344e-06, "loss": 1.4968, "step": 20327 }, { "epoch": 3.87, "learning_rate": 1.8401525262154434e-06, "loss": 2.2677, "step": 20328 }, { "epoch": 3.87, "learning_rate": 1.840095328884652e-06, "loss": 1.6781, "step": 20329 }, { "epoch": 3.87, "learning_rate": 1.840038131553861e-06, "loss": 2.7285, "step": 20330 }, { "epoch": 3.87, "learning_rate": 1.8399809342230698e-06, "loss": 3.3581, "step": 20331 }, { "epoch": 3.87, "learning_rate": 1.8399237368922784e-06, "loss": 3.3547, "step": 20332 }, { "epoch": 3.87, "learning_rate": 1.8398665395614874e-06, "loss": 1.2576, "step": 20333 }, { "epoch": 3.87, "learning_rate": 1.839809342230696e-06, "loss": 1.9004, "step": 20334 }, { "epoch": 3.87, "learning_rate": 1.8397521448999045e-06, "loss": 2.1175, "step": 20335 }, { "epoch": 3.87, "learning_rate": 1.8396949475691135e-06, "loss": 2.0992, "step": 20336 }, { "epoch": 3.87, "learning_rate": 1.8396377502383223e-06, "loss": 2.4071, "step": 20337 }, { "epoch": 3.87, "learning_rate": 1.839580552907531e-06, "loss": 1.5613, "step": 20338 }, { "epoch": 3.87, "learning_rate": 1.83952335557674e-06, "loss": 2.902, "step": 20339 }, { "epoch": 3.87, "learning_rate": 1.8394661582459485e-06, "loss": 1.4198, "step": 20340 }, { "epoch": 3.87, "learning_rate": 1.8394089609151573e-06, "loss": 2.5898, "step": 20341 }, { "epoch": 3.87, "learning_rate": 1.8393517635843663e-06, "loss": 2.4828, "step": 20342 }, { "epoch": 3.87, "learning_rate": 1.8392945662535749e-06, "loss": 2.1908, "step": 20343 }, { "epoch": 3.88, "learning_rate": 1.8392373689227835e-06, "loss": 2.163, "step": 20344 }, { "epoch": 3.88, "learning_rate": 1.8391801715919925e-06, "loss": 1.7127, "step": 20345 }, { "epoch": 3.88, "learning_rate": 1.839122974261201e-06, "loss": 1.5459, "step": 20346 }, { "epoch": 3.88, "learning_rate": 1.83906577693041e-06, "loss": 3.0366, "step": 20347 }, { "epoch": 3.88, "learning_rate": 1.8390085795996188e-06, "loss": 2.1728, "step": 20348 }, { "epoch": 3.88, "learning_rate": 1.8389513822688274e-06, "loss": 3.0976, "step": 20349 }, { "epoch": 3.88, "learning_rate": 1.8388941849380364e-06, "loss": 2.5054, "step": 20350 }, { "epoch": 3.88, "learning_rate": 1.838836987607245e-06, "loss": 2.0658, "step": 20351 }, { "epoch": 3.88, "learning_rate": 1.8387797902764538e-06, "loss": 2.886, "step": 20352 }, { "epoch": 3.88, "learning_rate": 1.8387225929456626e-06, "loss": 2.8932, "step": 20353 }, { "epoch": 3.88, "learning_rate": 1.8386653956148714e-06, "loss": 0.8745, "step": 20354 }, { "epoch": 3.88, "learning_rate": 1.83860819828408e-06, "loss": 1.9778, "step": 20355 }, { "epoch": 3.88, "learning_rate": 1.838551000953289e-06, "loss": 1.6387, "step": 20356 }, { "epoch": 3.88, "learning_rate": 1.8384938036224976e-06, "loss": 1.7102, "step": 20357 }, { "epoch": 3.88, "learning_rate": 1.8384366062917064e-06, "loss": 1.8116, "step": 20358 }, { "epoch": 3.88, "learning_rate": 1.8383794089609154e-06, "loss": 2.2464, "step": 20359 }, { "epoch": 3.88, "learning_rate": 1.838322211630124e-06, "loss": 1.6476, "step": 20360 }, { "epoch": 3.88, "learning_rate": 1.8382650142993325e-06, "loss": 2.6521, "step": 20361 }, { "epoch": 3.88, "learning_rate": 1.8382078169685415e-06, "loss": 1.087, "step": 20362 }, { "epoch": 3.88, "learning_rate": 1.8381506196377503e-06, "loss": 3.2765, "step": 20363 }, { "epoch": 3.88, "learning_rate": 1.8380934223069591e-06, "loss": 2.0682, "step": 20364 }, { "epoch": 3.88, "learning_rate": 1.838036224976168e-06, "loss": 2.6094, "step": 20365 }, { "epoch": 3.88, "learning_rate": 1.8379790276453765e-06, "loss": 1.5901, "step": 20366 }, { "epoch": 3.88, "learning_rate": 1.8379218303145855e-06, "loss": 0.9911, "step": 20367 }, { "epoch": 3.88, "learning_rate": 1.837864632983794e-06, "loss": 1.1662, "step": 20368 }, { "epoch": 3.88, "learning_rate": 1.8378074356530029e-06, "loss": 3.0467, "step": 20369 }, { "epoch": 3.88, "learning_rate": 1.8377502383222119e-06, "loss": 2.8442, "step": 20370 }, { "epoch": 3.88, "learning_rate": 1.8376930409914205e-06, "loss": 2.2069, "step": 20371 }, { "epoch": 3.88, "learning_rate": 1.837635843660629e-06, "loss": 1.257, "step": 20372 }, { "epoch": 3.88, "learning_rate": 1.837578646329838e-06, "loss": 1.5178, "step": 20373 }, { "epoch": 3.88, "learning_rate": 1.8375214489990466e-06, "loss": 1.5411, "step": 20374 }, { "epoch": 3.88, "learning_rate": 1.8374642516682554e-06, "loss": 2.5716, "step": 20375 }, { "epoch": 3.88, "learning_rate": 1.8374070543374644e-06, "loss": 1.3265, "step": 20376 }, { "epoch": 3.88, "learning_rate": 1.837349857006673e-06, "loss": 2.1803, "step": 20377 }, { "epoch": 3.88, "learning_rate": 1.837292659675882e-06, "loss": 2.1527, "step": 20378 }, { "epoch": 3.88, "learning_rate": 1.8372354623450906e-06, "loss": 2.9, "step": 20379 }, { "epoch": 3.88, "learning_rate": 1.8371782650142994e-06, "loss": 1.2717, "step": 20380 }, { "epoch": 3.88, "learning_rate": 1.8371210676835082e-06, "loss": 4.1541, "step": 20381 }, { "epoch": 3.88, "learning_rate": 1.837063870352717e-06, "loss": 2.0752, "step": 20382 }, { "epoch": 3.88, "learning_rate": 1.8370066730219256e-06, "loss": 2.1429, "step": 20383 }, { "epoch": 3.88, "learning_rate": 1.8369494756911346e-06, "loss": 1.8938, "step": 20384 }, { "epoch": 3.88, "learning_rate": 1.8368922783603432e-06, "loss": 2.1837, "step": 20385 }, { "epoch": 3.88, "learning_rate": 1.836835081029552e-06, "loss": 1.1142, "step": 20386 }, { "epoch": 3.88, "learning_rate": 1.836777883698761e-06, "loss": 1.7337, "step": 20387 }, { "epoch": 3.88, "learning_rate": 1.8367206863679695e-06, "loss": 2.6527, "step": 20388 }, { "epoch": 3.88, "learning_rate": 1.8366634890371781e-06, "loss": 2.3765, "step": 20389 }, { "epoch": 3.88, "learning_rate": 1.8366062917063871e-06, "loss": 1.5392, "step": 20390 }, { "epoch": 3.88, "learning_rate": 1.836549094375596e-06, "loss": 2.0313, "step": 20391 }, { "epoch": 3.88, "learning_rate": 1.8364918970448045e-06, "loss": 2.8664, "step": 20392 }, { "epoch": 3.88, "learning_rate": 1.8364346997140135e-06, "loss": 2.6003, "step": 20393 }, { "epoch": 3.88, "learning_rate": 1.836377502383222e-06, "loss": 2.0008, "step": 20394 }, { "epoch": 3.88, "learning_rate": 1.836320305052431e-06, "loss": 1.9913, "step": 20395 }, { "epoch": 3.88, "learning_rate": 1.8362631077216397e-06, "loss": 0.8678, "step": 20396 }, { "epoch": 3.89, "learning_rate": 1.8362059103908485e-06, "loss": 2.644, "step": 20397 }, { "epoch": 3.89, "learning_rate": 1.8361487130600575e-06, "loss": 2.0058, "step": 20398 }, { "epoch": 3.89, "learning_rate": 1.836091515729266e-06, "loss": 2.1222, "step": 20399 }, { "epoch": 3.89, "learning_rate": 1.8360343183984746e-06, "loss": 1.5378, "step": 20400 }, { "epoch": 3.89, "learning_rate": 1.8359771210676836e-06, "loss": 1.9651, "step": 20401 }, { "epoch": 3.89, "learning_rate": 1.8359199237368922e-06, "loss": 3.4591, "step": 20402 }, { "epoch": 3.89, "learning_rate": 1.835862726406101e-06, "loss": 1.5657, "step": 20403 }, { "epoch": 3.89, "learning_rate": 1.83580552907531e-06, "loss": 3.1148, "step": 20404 }, { "epoch": 3.89, "learning_rate": 1.8357483317445186e-06, "loss": 1.9615, "step": 20405 }, { "epoch": 3.89, "learning_rate": 1.8356911344137272e-06, "loss": 1.5543, "step": 20406 }, { "epoch": 3.89, "learning_rate": 1.8356339370829362e-06, "loss": 1.3756, "step": 20407 }, { "epoch": 3.89, "learning_rate": 1.835576739752145e-06, "loss": 2.2232, "step": 20408 }, { "epoch": 3.89, "learning_rate": 1.8355195424213536e-06, "loss": 2.3884, "step": 20409 }, { "epoch": 3.89, "learning_rate": 1.8354623450905626e-06, "loss": 2.4704, "step": 20410 }, { "epoch": 3.89, "learning_rate": 1.8354051477597712e-06, "loss": 2.9554, "step": 20411 }, { "epoch": 3.89, "learning_rate": 1.8353479504289802e-06, "loss": 2.6512, "step": 20412 }, { "epoch": 3.89, "learning_rate": 1.8352907530981887e-06, "loss": 1.356, "step": 20413 }, { "epoch": 3.89, "learning_rate": 1.8352335557673975e-06, "loss": 2.626, "step": 20414 }, { "epoch": 3.89, "learning_rate": 1.8351763584366065e-06, "loss": 1.4105, "step": 20415 }, { "epoch": 3.89, "learning_rate": 1.8351191611058151e-06, "loss": 2.2117, "step": 20416 }, { "epoch": 3.89, "learning_rate": 1.8350619637750237e-06, "loss": 1.8959, "step": 20417 }, { "epoch": 3.89, "learning_rate": 1.8350047664442327e-06, "loss": 2.7619, "step": 20418 }, { "epoch": 3.89, "learning_rate": 1.8349475691134415e-06, "loss": 2.1937, "step": 20419 }, { "epoch": 3.89, "learning_rate": 1.83489037178265e-06, "loss": 1.2068, "step": 20420 }, { "epoch": 3.89, "learning_rate": 1.834833174451859e-06, "loss": 1.9384, "step": 20421 }, { "epoch": 3.89, "learning_rate": 1.8347759771210677e-06, "loss": 2.3607, "step": 20422 }, { "epoch": 3.89, "learning_rate": 1.8347187797902763e-06, "loss": 1.6789, "step": 20423 }, { "epoch": 3.89, "learning_rate": 1.8346615824594853e-06, "loss": 2.3169, "step": 20424 }, { "epoch": 3.89, "learning_rate": 1.834604385128694e-06, "loss": 1.9644, "step": 20425 }, { "epoch": 3.89, "learning_rate": 1.8345471877979026e-06, "loss": 1.1513, "step": 20426 }, { "epoch": 3.89, "learning_rate": 1.8344899904671116e-06, "loss": 2.2099, "step": 20427 }, { "epoch": 3.89, "learning_rate": 1.8344327931363202e-06, "loss": 2.2559, "step": 20428 }, { "epoch": 3.89, "learning_rate": 1.8343755958055292e-06, "loss": 1.3297, "step": 20429 }, { "epoch": 3.89, "learning_rate": 1.834318398474738e-06, "loss": 2.7015, "step": 20430 }, { "epoch": 3.89, "learning_rate": 1.8342612011439466e-06, "loss": 2.0497, "step": 20431 }, { "epoch": 3.89, "learning_rate": 1.8342040038131556e-06, "loss": 2.1041, "step": 20432 }, { "epoch": 3.89, "learning_rate": 1.8341468064823642e-06, "loss": 3.2087, "step": 20433 }, { "epoch": 3.89, "learning_rate": 1.8340896091515728e-06, "loss": 2.453, "step": 20434 }, { "epoch": 3.89, "learning_rate": 1.8340324118207818e-06, "loss": 2.5473, "step": 20435 }, { "epoch": 3.89, "learning_rate": 1.8339752144899906e-06, "loss": 2.4572, "step": 20436 }, { "epoch": 3.89, "learning_rate": 1.8339180171591992e-06, "loss": 2.6991, "step": 20437 }, { "epoch": 3.89, "learning_rate": 1.8338608198284082e-06, "loss": 2.3537, "step": 20438 }, { "epoch": 3.89, "learning_rate": 1.8338036224976167e-06, "loss": 3.1588, "step": 20439 }, { "epoch": 3.89, "learning_rate": 1.8337464251668255e-06, "loss": 1.8082, "step": 20440 }, { "epoch": 3.89, "learning_rate": 1.8336892278360343e-06, "loss": 2.1299, "step": 20441 }, { "epoch": 3.89, "learning_rate": 1.8336320305052431e-06, "loss": 2.4993, "step": 20442 }, { "epoch": 3.89, "learning_rate": 1.8335748331744521e-06, "loss": 3.475, "step": 20443 }, { "epoch": 3.89, "learning_rate": 1.8335176358436607e-06, "loss": 1.9721, "step": 20444 }, { "epoch": 3.89, "learning_rate": 1.8334604385128693e-06, "loss": 2.3338, "step": 20445 }, { "epoch": 3.89, "learning_rate": 1.8334032411820783e-06, "loss": 1.8148, "step": 20446 }, { "epoch": 3.89, "learning_rate": 1.833346043851287e-06, "loss": 1.5108, "step": 20447 }, { "epoch": 3.89, "learning_rate": 1.8332888465204957e-06, "loss": 1.9606, "step": 20448 }, { "epoch": 3.9, "learning_rate": 1.8332316491897047e-06, "loss": 1.496, "step": 20449 }, { "epoch": 3.9, "learning_rate": 1.8331744518589133e-06, "loss": 2.1015, "step": 20450 }, { "epoch": 3.9, "learning_rate": 1.8331172545281219e-06, "loss": 1.8012, "step": 20451 }, { "epoch": 3.9, "learning_rate": 1.8330600571973309e-06, "loss": 3.2318, "step": 20452 }, { "epoch": 3.9, "learning_rate": 1.8330028598665396e-06, "loss": 1.617, "step": 20453 }, { "epoch": 3.9, "learning_rate": 1.8329456625357482e-06, "loss": 2.1076, "step": 20454 }, { "epoch": 3.9, "learning_rate": 1.8328884652049572e-06, "loss": 2.5766, "step": 20455 }, { "epoch": 3.9, "learning_rate": 1.8328312678741658e-06, "loss": 2.1828, "step": 20456 }, { "epoch": 3.9, "learning_rate": 1.8327740705433746e-06, "loss": 1.5677, "step": 20457 }, { "epoch": 3.9, "learning_rate": 1.8327168732125836e-06, "loss": 1.5851, "step": 20458 }, { "epoch": 3.9, "learning_rate": 1.8326596758817922e-06, "loss": 2.3583, "step": 20459 }, { "epoch": 3.9, "learning_rate": 1.8326024785510012e-06, "loss": 3.2325, "step": 20460 }, { "epoch": 3.9, "learning_rate": 1.8325452812202098e-06, "loss": 2.6218, "step": 20461 }, { "epoch": 3.9, "learning_rate": 1.8324880838894184e-06, "loss": 1.6653, "step": 20462 }, { "epoch": 3.9, "learning_rate": 1.8324308865586274e-06, "loss": 3.3767, "step": 20463 }, { "epoch": 3.9, "learning_rate": 1.8323736892278362e-06, "loss": 2.0131, "step": 20464 }, { "epoch": 3.9, "learning_rate": 1.8323164918970448e-06, "loss": 2.8043, "step": 20465 }, { "epoch": 3.9, "learning_rate": 1.8322592945662538e-06, "loss": 2.7297, "step": 20466 }, { "epoch": 3.9, "learning_rate": 1.8322020972354623e-06, "loss": 2.1766, "step": 20467 }, { "epoch": 3.9, "learning_rate": 1.8321448999046711e-06, "loss": 2.9639, "step": 20468 }, { "epoch": 3.9, "learning_rate": 1.83208770257388e-06, "loss": 2.0625, "step": 20469 }, { "epoch": 3.9, "learning_rate": 1.8320305052430887e-06, "loss": 2.9693, "step": 20470 }, { "epoch": 3.9, "learning_rate": 1.8319733079122973e-06, "loss": 3.2292, "step": 20471 }, { "epoch": 3.9, "learning_rate": 1.8319161105815063e-06, "loss": 1.7836, "step": 20472 }, { "epoch": 3.9, "learning_rate": 1.8318589132507149e-06, "loss": 1.7756, "step": 20473 }, { "epoch": 3.9, "learning_rate": 1.8318017159199237e-06, "loss": 2.0474, "step": 20474 }, { "epoch": 3.9, "learning_rate": 1.8317445185891327e-06, "loss": 3.6125, "step": 20475 }, { "epoch": 3.9, "learning_rate": 1.8316873212583413e-06, "loss": 2.6789, "step": 20476 }, { "epoch": 3.9, "learning_rate": 1.8316301239275503e-06, "loss": 2.1702, "step": 20477 }, { "epoch": 3.9, "learning_rate": 1.8315729265967589e-06, "loss": 3.3732, "step": 20478 }, { "epoch": 3.9, "learning_rate": 1.8315157292659677e-06, "loss": 3.8094, "step": 20479 }, { "epoch": 3.9, "learning_rate": 1.8314585319351764e-06, "loss": 2.0138, "step": 20480 }, { "epoch": 3.9, "learning_rate": 1.8314013346043852e-06, "loss": 3.8359, "step": 20481 }, { "epoch": 3.9, "learning_rate": 1.8313441372735938e-06, "loss": 2.5748, "step": 20482 }, { "epoch": 3.9, "learning_rate": 1.8312869399428028e-06, "loss": 3.3355, "step": 20483 }, { "epoch": 3.9, "learning_rate": 1.8312297426120114e-06, "loss": 2.6467, "step": 20484 }, { "epoch": 3.9, "learning_rate": 1.8311725452812202e-06, "loss": 1.8382, "step": 20485 }, { "epoch": 3.9, "learning_rate": 1.8311153479504292e-06, "loss": 3.0296, "step": 20486 }, { "epoch": 3.9, "learning_rate": 1.8310581506196378e-06, "loss": 2.3755, "step": 20487 }, { "epoch": 3.9, "learning_rate": 1.8310009532888464e-06, "loss": 2.3479, "step": 20488 }, { "epoch": 3.9, "learning_rate": 1.8309437559580554e-06, "loss": 2.9643, "step": 20489 }, { "epoch": 3.9, "learning_rate": 1.830886558627264e-06, "loss": 2.6026, "step": 20490 }, { "epoch": 3.9, "learning_rate": 1.8308293612964728e-06, "loss": 3.6733, "step": 20491 }, { "epoch": 3.9, "learning_rate": 1.8307721639656818e-06, "loss": 0.9945, "step": 20492 }, { "epoch": 3.9, "learning_rate": 1.8307149666348903e-06, "loss": 1.7762, "step": 20493 }, { "epoch": 3.9, "learning_rate": 1.8306577693040993e-06, "loss": 1.8382, "step": 20494 }, { "epoch": 3.9, "learning_rate": 1.830600571973308e-06, "loss": 1.8181, "step": 20495 }, { "epoch": 3.9, "learning_rate": 1.8305433746425167e-06, "loss": 2.07, "step": 20496 }, { "epoch": 3.9, "learning_rate": 1.8304861773117257e-06, "loss": 2.5828, "step": 20497 }, { "epoch": 3.9, "learning_rate": 1.8304289799809343e-06, "loss": 1.3567, "step": 20498 }, { "epoch": 3.9, "learning_rate": 1.830371782650143e-06, "loss": 2.0008, "step": 20499 }, { "epoch": 3.9, "learning_rate": 1.830314585319352e-06, "loss": 3.5305, "step": 20500 }, { "epoch": 3.9, "learning_rate": 1.8302573879885605e-06, "loss": 1.3953, "step": 20501 }, { "epoch": 3.91, "learning_rate": 1.8302001906577693e-06, "loss": 2.7052, "step": 20502 }, { "epoch": 3.91, "learning_rate": 1.8301429933269783e-06, "loss": 3.5107, "step": 20503 }, { "epoch": 3.91, "learning_rate": 1.8300857959961869e-06, "loss": 1.5813, "step": 20504 }, { "epoch": 3.91, "learning_rate": 1.8300285986653954e-06, "loss": 1.907, "step": 20505 }, { "epoch": 3.91, "learning_rate": 1.8299714013346045e-06, "loss": 1.7865, "step": 20506 }, { "epoch": 3.91, "learning_rate": 1.8299142040038132e-06, "loss": 2.5345, "step": 20507 }, { "epoch": 3.91, "learning_rate": 1.829857006673022e-06, "loss": 3.7061, "step": 20508 }, { "epoch": 3.91, "learning_rate": 1.8297998093422308e-06, "loss": 3.1525, "step": 20509 }, { "epoch": 3.91, "learning_rate": 1.8297426120114394e-06, "loss": 2.6752, "step": 20510 }, { "epoch": 3.91, "learning_rate": 1.8296854146806484e-06, "loss": 1.5206, "step": 20511 }, { "epoch": 3.91, "learning_rate": 1.829628217349857e-06, "loss": 2.075, "step": 20512 }, { "epoch": 3.91, "learning_rate": 1.8295710200190658e-06, "loss": 1.9248, "step": 20513 }, { "epoch": 3.91, "learning_rate": 1.8295138226882748e-06, "loss": 2.7445, "step": 20514 }, { "epoch": 3.91, "learning_rate": 1.8294566253574834e-06, "loss": 2.7338, "step": 20515 }, { "epoch": 3.91, "learning_rate": 1.829399428026692e-06, "loss": 2.6358, "step": 20516 }, { "epoch": 3.91, "learning_rate": 1.829342230695901e-06, "loss": 2.3696, "step": 20517 }, { "epoch": 3.91, "learning_rate": 1.8292850333651098e-06, "loss": 2.8581, "step": 20518 }, { "epoch": 3.91, "learning_rate": 1.8292278360343183e-06, "loss": 1.917, "step": 20519 }, { "epoch": 3.91, "learning_rate": 1.8291706387035274e-06, "loss": 1.4264, "step": 20520 }, { "epoch": 3.91, "learning_rate": 1.829113441372736e-06, "loss": 1.8055, "step": 20521 }, { "epoch": 3.91, "learning_rate": 1.8290562440419445e-06, "loss": 0.9937, "step": 20522 }, { "epoch": 3.91, "learning_rate": 1.8289990467111535e-06, "loss": 3.048, "step": 20523 }, { "epoch": 3.91, "learning_rate": 1.8289418493803623e-06, "loss": 2.6813, "step": 20524 }, { "epoch": 3.91, "learning_rate": 1.8288846520495713e-06, "loss": 1.668, "step": 20525 }, { "epoch": 3.91, "learning_rate": 1.82882745471878e-06, "loss": 3.3138, "step": 20526 }, { "epoch": 3.91, "learning_rate": 1.8287702573879885e-06, "loss": 2.4687, "step": 20527 }, { "epoch": 3.91, "learning_rate": 1.8287130600571975e-06, "loss": 3.2792, "step": 20528 }, { "epoch": 3.91, "learning_rate": 1.828655862726406e-06, "loss": 1.8415, "step": 20529 }, { "epoch": 3.91, "learning_rate": 1.8285986653956149e-06, "loss": 1.3391, "step": 20530 }, { "epoch": 3.91, "learning_rate": 1.8285414680648239e-06, "loss": 1.1823, "step": 20531 }, { "epoch": 3.91, "learning_rate": 1.8284842707340325e-06, "loss": 2.8716, "step": 20532 }, { "epoch": 3.91, "learning_rate": 1.828427073403241e-06, "loss": 2.2524, "step": 20533 }, { "epoch": 3.91, "learning_rate": 1.82836987607245e-06, "loss": 2.4803, "step": 20534 }, { "epoch": 3.91, "learning_rate": 1.8283126787416588e-06, "loss": 2.0889, "step": 20535 }, { "epoch": 3.91, "learning_rate": 1.8282554814108674e-06, "loss": 2.4741, "step": 20536 }, { "epoch": 3.91, "learning_rate": 1.8281982840800764e-06, "loss": 1.554, "step": 20537 }, { "epoch": 3.91, "learning_rate": 1.828141086749285e-06, "loss": 2.3595, "step": 20538 }, { "epoch": 3.91, "learning_rate": 1.8280838894184936e-06, "loss": 3.4852, "step": 20539 }, { "epoch": 3.91, "learning_rate": 1.8280266920877026e-06, "loss": 3.0586, "step": 20540 }, { "epoch": 3.91, "learning_rate": 1.8279694947569114e-06, "loss": 1.7736, "step": 20541 }, { "epoch": 3.91, "learning_rate": 1.8279122974261204e-06, "loss": 2.5228, "step": 20542 }, { "epoch": 3.91, "learning_rate": 1.827855100095329e-06, "loss": 2.1455, "step": 20543 }, { "epoch": 3.91, "learning_rate": 1.8277979027645376e-06, "loss": 2.8576, "step": 20544 }, { "epoch": 3.91, "learning_rate": 1.8277407054337466e-06, "loss": 1.5456, "step": 20545 }, { "epoch": 3.91, "learning_rate": 1.8276835081029554e-06, "loss": 2.8689, "step": 20546 }, { "epoch": 3.91, "learning_rate": 1.827626310772164e-06, "loss": 3.2582, "step": 20547 }, { "epoch": 3.91, "learning_rate": 1.827569113441373e-06, "loss": 2.7732, "step": 20548 }, { "epoch": 3.91, "learning_rate": 1.8275119161105815e-06, "loss": 1.8887, "step": 20549 }, { "epoch": 3.91, "learning_rate": 1.8274547187797901e-06, "loss": 4.0783, "step": 20550 }, { "epoch": 3.91, "learning_rate": 1.8273975214489991e-06, "loss": 2.2829, "step": 20551 }, { "epoch": 3.91, "learning_rate": 1.827340324118208e-06, "loss": 2.5385, "step": 20552 }, { "epoch": 3.91, "learning_rate": 1.8272831267874165e-06, "loss": 1.5042, "step": 20553 }, { "epoch": 3.92, "learning_rate": 1.8272259294566255e-06, "loss": 1.8868, "step": 20554 }, { "epoch": 3.92, "learning_rate": 1.827168732125834e-06, "loss": 2.4752, "step": 20555 }, { "epoch": 3.92, "learning_rate": 1.8271115347950429e-06, "loss": 2.8573, "step": 20556 }, { "epoch": 3.92, "learning_rate": 1.8270543374642517e-06, "loss": 2.192, "step": 20557 }, { "epoch": 3.92, "learning_rate": 1.8269971401334605e-06, "loss": 2.5182, "step": 20558 }, { "epoch": 3.92, "learning_rate": 1.8269399428026695e-06, "loss": 2.6292, "step": 20559 }, { "epoch": 3.92, "learning_rate": 1.826882745471878e-06, "loss": 2.6939, "step": 20560 }, { "epoch": 3.92, "learning_rate": 1.8268255481410866e-06, "loss": 1.8794, "step": 20561 }, { "epoch": 3.92, "learning_rate": 1.8267683508102956e-06, "loss": 2.9207, "step": 20562 }, { "epoch": 3.92, "learning_rate": 1.8267111534795044e-06, "loss": 3.009, "step": 20563 }, { "epoch": 3.92, "learning_rate": 1.826653956148713e-06, "loss": 1.7823, "step": 20564 }, { "epoch": 3.92, "learning_rate": 1.826596758817922e-06, "loss": 1.3832, "step": 20565 }, { "epoch": 3.92, "learning_rate": 1.8265395614871306e-06, "loss": 3.1616, "step": 20566 }, { "epoch": 3.92, "learning_rate": 1.8264823641563394e-06, "loss": 1.2209, "step": 20567 }, { "epoch": 3.92, "learning_rate": 1.8264251668255482e-06, "loss": 2.2502, "step": 20568 }, { "epoch": 3.92, "learning_rate": 1.826367969494757e-06, "loss": 1.0542, "step": 20569 }, { "epoch": 3.92, "learning_rate": 1.8263107721639656e-06, "loss": 2.1237, "step": 20570 }, { "epoch": 3.92, "learning_rate": 1.8262535748331746e-06, "loss": 2.1868, "step": 20571 }, { "epoch": 3.92, "learning_rate": 1.8261963775023831e-06, "loss": 2.7202, "step": 20572 }, { "epoch": 3.92, "learning_rate": 1.8261391801715922e-06, "loss": 1.8148, "step": 20573 }, { "epoch": 3.92, "learning_rate": 1.826081982840801e-06, "loss": 2.7579, "step": 20574 }, { "epoch": 3.92, "learning_rate": 1.8260247855100095e-06, "loss": 1.7059, "step": 20575 }, { "epoch": 3.92, "learning_rate": 1.8259675881792185e-06, "loss": 2.1911, "step": 20576 }, { "epoch": 3.92, "learning_rate": 1.8259103908484271e-06, "loss": 3.6972, "step": 20577 }, { "epoch": 3.92, "learning_rate": 1.8258531935176357e-06, "loss": 2.4656, "step": 20578 }, { "epoch": 3.92, "learning_rate": 1.8257959961868447e-06, "loss": 2.005, "step": 20579 }, { "epoch": 3.92, "learning_rate": 1.8257387988560535e-06, "loss": 1.5688, "step": 20580 }, { "epoch": 3.92, "learning_rate": 1.825681601525262e-06, "loss": 1.6177, "step": 20581 }, { "epoch": 3.92, "learning_rate": 1.825624404194471e-06, "loss": 2.0106, "step": 20582 }, { "epoch": 3.92, "learning_rate": 1.8255672068636797e-06, "loss": 1.1224, "step": 20583 }, { "epoch": 3.92, "learning_rate": 1.8255100095328885e-06, "loss": 2.8389, "step": 20584 }, { "epoch": 3.92, "learning_rate": 1.8254528122020975e-06, "loss": 2.2073, "step": 20585 }, { "epoch": 3.92, "learning_rate": 1.825395614871306e-06, "loss": 2.2313, "step": 20586 }, { "epoch": 3.92, "learning_rate": 1.8253384175405146e-06, "loss": 3.0396, "step": 20587 }, { "epoch": 3.92, "learning_rate": 1.8252812202097236e-06, "loss": 1.9415, "step": 20588 }, { "epoch": 3.92, "learning_rate": 1.8252240228789322e-06, "loss": 1.5861, "step": 20589 }, { "epoch": 3.92, "learning_rate": 1.8251668255481412e-06, "loss": 2.642, "step": 20590 }, { "epoch": 3.92, "learning_rate": 1.82510962821735e-06, "loss": 2.2035, "step": 20591 }, { "epoch": 3.92, "learning_rate": 1.8250524308865586e-06, "loss": 2.1173, "step": 20592 }, { "epoch": 3.92, "learning_rate": 1.8249952335557676e-06, "loss": 2.8878, "step": 20593 }, { "epoch": 3.92, "learning_rate": 1.8249380362249762e-06, "loss": 2.8616, "step": 20594 }, { "epoch": 3.92, "learning_rate": 1.824880838894185e-06, "loss": 2.6394, "step": 20595 }, { "epoch": 3.92, "learning_rate": 1.8248236415633938e-06, "loss": 2.6627, "step": 20596 }, { "epoch": 3.92, "learning_rate": 1.8247664442326026e-06, "loss": 2.8016, "step": 20597 }, { "epoch": 3.92, "learning_rate": 1.8247092469018112e-06, "loss": 2.6833, "step": 20598 }, { "epoch": 3.92, "learning_rate": 1.8246520495710202e-06, "loss": 1.7772, "step": 20599 }, { "epoch": 3.92, "learning_rate": 1.8245948522402287e-06, "loss": 2.6157, "step": 20600 }, { "epoch": 3.92, "learning_rate": 1.8245376549094375e-06, "loss": 2.0692, "step": 20601 }, { "epoch": 3.92, "learning_rate": 1.8244804575786465e-06, "loss": 1.4349, "step": 20602 }, { "epoch": 3.92, "learning_rate": 1.8244232602478551e-06, "loss": 2.5048, "step": 20603 }, { "epoch": 3.92, "learning_rate": 1.8243660629170637e-06, "loss": 1.0319, "step": 20604 }, { "epoch": 3.92, "learning_rate": 1.8243088655862727e-06, "loss": 1.213, "step": 20605 }, { "epoch": 3.92, "learning_rate": 1.8242516682554813e-06, "loss": 2.5728, "step": 20606 }, { "epoch": 3.93, "learning_rate": 1.8241944709246903e-06, "loss": 2.7355, "step": 20607 }, { "epoch": 3.93, "learning_rate": 1.824137273593899e-06, "loss": 1.6841, "step": 20608 }, { "epoch": 3.93, "learning_rate": 1.8240800762631077e-06, "loss": 1.5881, "step": 20609 }, { "epoch": 3.93, "learning_rate": 1.8240228789323167e-06, "loss": 2.1791, "step": 20610 }, { "epoch": 3.93, "learning_rate": 1.8239656816015253e-06, "loss": 2.4325, "step": 20611 }, { "epoch": 3.93, "learning_rate": 1.823908484270734e-06, "loss": 1.3827, "step": 20612 }, { "epoch": 3.93, "learning_rate": 1.823851286939943e-06, "loss": 1.6594, "step": 20613 }, { "epoch": 3.93, "learning_rate": 1.8237940896091516e-06, "loss": 1.5849, "step": 20614 }, { "epoch": 3.93, "learning_rate": 1.8237368922783602e-06, "loss": 2.4864, "step": 20615 }, { "epoch": 3.93, "learning_rate": 1.8236796949475692e-06, "loss": 1.4468, "step": 20616 }, { "epoch": 3.93, "learning_rate": 1.8236224976167778e-06, "loss": 2.9122, "step": 20617 }, { "epoch": 3.93, "learning_rate": 1.8235653002859866e-06, "loss": 3.1582, "step": 20618 }, { "epoch": 3.93, "learning_rate": 1.8235081029551956e-06, "loss": 2.3109, "step": 20619 }, { "epoch": 3.93, "learning_rate": 1.8234509056244042e-06, "loss": 2.0302, "step": 20620 }, { "epoch": 3.93, "learning_rate": 1.8233937082936128e-06, "loss": 0.7474, "step": 20621 }, { "epoch": 3.93, "learning_rate": 1.8233365109628218e-06, "loss": 2.8264, "step": 20622 }, { "epoch": 3.93, "learning_rate": 1.8232793136320306e-06, "loss": 2.0539, "step": 20623 }, { "epoch": 3.93, "learning_rate": 1.8232221163012394e-06, "loss": 2.0993, "step": 20624 }, { "epoch": 3.93, "learning_rate": 1.8231649189704482e-06, "loss": 2.8517, "step": 20625 }, { "epoch": 3.93, "learning_rate": 1.8231077216396567e-06, "loss": 2.261, "step": 20626 }, { "epoch": 3.93, "learning_rate": 1.8230505243088658e-06, "loss": 2.7702, "step": 20627 }, { "epoch": 3.93, "learning_rate": 1.8229933269780743e-06, "loss": 2.395, "step": 20628 }, { "epoch": 3.93, "learning_rate": 1.8229361296472831e-06, "loss": 1.6506, "step": 20629 }, { "epoch": 3.93, "learning_rate": 1.8228789323164921e-06, "loss": 1.8217, "step": 20630 }, { "epoch": 3.93, "learning_rate": 1.8228217349857007e-06, "loss": 2.7633, "step": 20631 }, { "epoch": 3.93, "learning_rate": 1.8227645376549093e-06, "loss": 1.6949, "step": 20632 }, { "epoch": 3.93, "learning_rate": 1.8227073403241183e-06, "loss": 2.2063, "step": 20633 }, { "epoch": 3.93, "learning_rate": 1.822650142993327e-06, "loss": 1.9132, "step": 20634 }, { "epoch": 3.93, "learning_rate": 1.8225929456625357e-06, "loss": 2.464, "step": 20635 }, { "epoch": 3.93, "learning_rate": 1.8225357483317447e-06, "loss": 2.2468, "step": 20636 }, { "epoch": 3.93, "learning_rate": 1.8224785510009533e-06, "loss": 2.3989, "step": 20637 }, { "epoch": 3.93, "learning_rate": 1.8224213536701623e-06, "loss": 1.8116, "step": 20638 }, { "epoch": 3.93, "learning_rate": 1.8223641563393709e-06, "loss": 1.1133, "step": 20639 }, { "epoch": 3.93, "learning_rate": 1.8223069590085796e-06, "loss": 3.0641, "step": 20640 }, { "epoch": 3.93, "learning_rate": 1.8222497616777887e-06, "loss": 2.232, "step": 20641 }, { "epoch": 3.93, "learning_rate": 1.8221925643469972e-06, "loss": 2.6488, "step": 20642 }, { "epoch": 3.93, "learning_rate": 1.8221353670162058e-06, "loss": 1.3065, "step": 20643 }, { "epoch": 3.93, "learning_rate": 1.8220781696854148e-06, "loss": 1.3903, "step": 20644 }, { "epoch": 3.93, "learning_rate": 1.8220209723546234e-06, "loss": 2.5122, "step": 20645 }, { "epoch": 3.93, "learning_rate": 1.8219637750238322e-06, "loss": 2.9624, "step": 20646 }, { "epoch": 3.93, "learning_rate": 1.8219065776930412e-06, "loss": 2.2769, "step": 20647 }, { "epoch": 3.93, "learning_rate": 1.8218493803622498e-06, "loss": 1.8085, "step": 20648 }, { "epoch": 3.93, "learning_rate": 1.8217921830314584e-06, "loss": 2.6036, "step": 20649 }, { "epoch": 3.93, "learning_rate": 1.8217349857006674e-06, "loss": 3.2933, "step": 20650 }, { "epoch": 3.93, "learning_rate": 1.8216777883698762e-06, "loss": 1.6948, "step": 20651 }, { "epoch": 3.93, "learning_rate": 1.8216205910390847e-06, "loss": 1.9653, "step": 20652 }, { "epoch": 3.93, "learning_rate": 1.8215633937082938e-06, "loss": 3.5606, "step": 20653 }, { "epoch": 3.93, "learning_rate": 1.8215061963775023e-06, "loss": 2.5808, "step": 20654 }, { "epoch": 3.93, "learning_rate": 1.8214489990467113e-06, "loss": 2.3816, "step": 20655 }, { "epoch": 3.93, "learning_rate": 1.82139180171592e-06, "loss": 3.3033, "step": 20656 }, { "epoch": 3.93, "learning_rate": 1.8213346043851287e-06, "loss": 2.2258, "step": 20657 }, { "epoch": 3.93, "learning_rate": 1.8212774070543377e-06, "loss": 1.4507, "step": 20658 }, { "epoch": 3.94, "learning_rate": 1.8212202097235463e-06, "loss": 3.4294, "step": 20659 }, { "epoch": 3.94, "learning_rate": 1.8211630123927549e-06, "loss": 2.6891, "step": 20660 }, { "epoch": 3.94, "learning_rate": 1.8211058150619639e-06, "loss": 1.9359, "step": 20661 }, { "epoch": 3.94, "learning_rate": 1.8210486177311727e-06, "loss": 2.6194, "step": 20662 }, { "epoch": 3.94, "learning_rate": 1.8209914204003813e-06, "loss": 3.4465, "step": 20663 }, { "epoch": 3.94, "learning_rate": 1.8209342230695903e-06, "loss": 2.4113, "step": 20664 }, { "epoch": 3.94, "learning_rate": 1.8208770257387989e-06, "loss": 2.2335, "step": 20665 }, { "epoch": 3.94, "learning_rate": 1.8208198284080074e-06, "loss": 1.7265, "step": 20666 }, { "epoch": 3.94, "learning_rate": 1.8207626310772164e-06, "loss": 1.6416, "step": 20667 }, { "epoch": 3.94, "learning_rate": 1.8207054337464252e-06, "loss": 1.5093, "step": 20668 }, { "epoch": 3.94, "learning_rate": 1.8206482364156338e-06, "loss": 1.8186, "step": 20669 }, { "epoch": 3.94, "learning_rate": 1.8205910390848428e-06, "loss": 1.6597, "step": 20670 }, { "epoch": 3.94, "learning_rate": 1.8205338417540514e-06, "loss": 1.9689, "step": 20671 }, { "epoch": 3.94, "learning_rate": 1.8204766444232604e-06, "loss": 1.2471, "step": 20672 }, { "epoch": 3.94, "learning_rate": 1.8204194470924692e-06, "loss": 2.2774, "step": 20673 }, { "epoch": 3.94, "learning_rate": 1.8203622497616778e-06, "loss": 2.5597, "step": 20674 }, { "epoch": 3.94, "learning_rate": 1.8203050524308868e-06, "loss": 1.9844, "step": 20675 }, { "epoch": 3.94, "learning_rate": 1.8202478551000954e-06, "loss": 2.8277, "step": 20676 }, { "epoch": 3.94, "learning_rate": 1.820190657769304e-06, "loss": 2.4637, "step": 20677 }, { "epoch": 3.94, "learning_rate": 1.820133460438513e-06, "loss": 3.4724, "step": 20678 }, { "epoch": 3.94, "learning_rate": 1.8200762631077218e-06, "loss": 2.4579, "step": 20679 }, { "epoch": 3.94, "learning_rate": 1.8200190657769303e-06, "loss": 2.5814, "step": 20680 }, { "epoch": 3.94, "learning_rate": 1.8199618684461393e-06, "loss": 1.9782, "step": 20681 }, { "epoch": 3.94, "learning_rate": 1.819904671115348e-06, "loss": 2.3734, "step": 20682 }, { "epoch": 3.94, "learning_rate": 1.8198474737845567e-06, "loss": 1.5541, "step": 20683 }, { "epoch": 3.94, "learning_rate": 1.8197902764537655e-06, "loss": 2.3437, "step": 20684 }, { "epoch": 3.94, "learning_rate": 1.8197330791229743e-06, "loss": 1.9054, "step": 20685 }, { "epoch": 3.94, "learning_rate": 1.8196758817921829e-06, "loss": 2.5452, "step": 20686 }, { "epoch": 3.94, "learning_rate": 1.819618684461392e-06, "loss": 1.8901, "step": 20687 }, { "epoch": 3.94, "learning_rate": 1.8195614871306005e-06, "loss": 2.5498, "step": 20688 }, { "epoch": 3.94, "learning_rate": 1.8195042897998095e-06, "loss": 1.6272, "step": 20689 }, { "epoch": 3.94, "learning_rate": 1.8194470924690183e-06, "loss": 2.9215, "step": 20690 }, { "epoch": 3.94, "learning_rate": 1.8193898951382269e-06, "loss": 2.6223, "step": 20691 }, { "epoch": 3.94, "learning_rate": 1.8193326978074359e-06, "loss": 1.8974, "step": 20692 }, { "epoch": 3.94, "learning_rate": 1.8192755004766444e-06, "loss": 1.2377, "step": 20693 }, { "epoch": 3.94, "learning_rate": 1.819218303145853e-06, "loss": 1.7152, "step": 20694 }, { "epoch": 3.94, "learning_rate": 1.819161105815062e-06, "loss": 2.0643, "step": 20695 }, { "epoch": 3.94, "learning_rate": 1.8191039084842708e-06, "loss": 3.4007, "step": 20696 }, { "epoch": 3.94, "learning_rate": 1.8190467111534794e-06, "loss": 2.826, "step": 20697 }, { "epoch": 3.94, "learning_rate": 1.8189895138226884e-06, "loss": 2.8785, "step": 20698 }, { "epoch": 3.94, "learning_rate": 1.818932316491897e-06, "loss": 1.9194, "step": 20699 }, { "epoch": 3.94, "learning_rate": 1.8188751191611058e-06, "loss": 2.0991, "step": 20700 }, { "epoch": 3.94, "learning_rate": 1.8188179218303148e-06, "loss": 1.7141, "step": 20701 }, { "epoch": 3.94, "learning_rate": 1.8187607244995234e-06, "loss": 3.0177, "step": 20702 }, { "epoch": 3.94, "learning_rate": 1.818703527168732e-06, "loss": 2.5736, "step": 20703 }, { "epoch": 3.94, "learning_rate": 1.818646329837941e-06, "loss": 2.1594, "step": 20704 }, { "epoch": 3.94, "learning_rate": 1.8185891325071496e-06, "loss": 2.0136, "step": 20705 }, { "epoch": 3.94, "learning_rate": 1.8185319351763586e-06, "loss": 2.1009, "step": 20706 }, { "epoch": 3.94, "learning_rate": 1.8184747378455674e-06, "loss": 2.4473, "step": 20707 }, { "epoch": 3.94, "learning_rate": 1.818417540514776e-06, "loss": 2.1335, "step": 20708 }, { "epoch": 3.94, "learning_rate": 1.818360343183985e-06, "loss": 1.539, "step": 20709 }, { "epoch": 3.94, "learning_rate": 1.8183031458531935e-06, "loss": 2.3462, "step": 20710 }, { "epoch": 3.94, "learning_rate": 1.8182459485224023e-06, "loss": 2.135, "step": 20711 }, { "epoch": 3.95, "learning_rate": 1.8181887511916111e-06, "loss": 3.501, "step": 20712 }, { "epoch": 3.95, "learning_rate": 1.81813155386082e-06, "loss": 2.3947, "step": 20713 }, { "epoch": 3.95, "learning_rate": 1.8180743565300285e-06, "loss": 2.0628, "step": 20714 }, { "epoch": 3.95, "learning_rate": 1.8180171591992375e-06, "loss": 1.7547, "step": 20715 }, { "epoch": 3.95, "learning_rate": 1.817959961868446e-06, "loss": 2.6235, "step": 20716 }, { "epoch": 3.95, "learning_rate": 1.8179027645376549e-06, "loss": 2.9214, "step": 20717 }, { "epoch": 3.95, "learning_rate": 1.8178455672068639e-06, "loss": 2.6347, "step": 20718 }, { "epoch": 3.95, "learning_rate": 1.8177883698760725e-06, "loss": 1.2562, "step": 20719 }, { "epoch": 3.95, "learning_rate": 1.8177311725452815e-06, "loss": 4.1025, "step": 20720 }, { "epoch": 3.95, "learning_rate": 1.81767397521449e-06, "loss": 1.4485, "step": 20721 }, { "epoch": 3.95, "learning_rate": 1.8176167778836988e-06, "loss": 1.7994, "step": 20722 }, { "epoch": 3.95, "learning_rate": 1.8175595805529076e-06, "loss": 2.5477, "step": 20723 }, { "epoch": 3.95, "learning_rate": 1.8175023832221164e-06, "loss": 1.1684, "step": 20724 }, { "epoch": 3.95, "learning_rate": 1.817445185891325e-06, "loss": 3.4127, "step": 20725 }, { "epoch": 3.95, "learning_rate": 1.817387988560534e-06, "loss": 1.728, "step": 20726 }, { "epoch": 3.95, "learning_rate": 1.8173307912297426e-06, "loss": 1.0418, "step": 20727 }, { "epoch": 3.95, "learning_rate": 1.8172735938989514e-06, "loss": 2.2997, "step": 20728 }, { "epoch": 3.95, "learning_rate": 1.8172163965681604e-06, "loss": 1.794, "step": 20729 }, { "epoch": 3.95, "learning_rate": 1.817159199237369e-06, "loss": 1.8402, "step": 20730 }, { "epoch": 3.95, "learning_rate": 1.8171020019065776e-06, "loss": 2.5476, "step": 20731 }, { "epoch": 3.95, "learning_rate": 1.8170448045757866e-06, "loss": 2.6165, "step": 20732 }, { "epoch": 3.95, "learning_rate": 1.8169876072449951e-06, "loss": 1.9843, "step": 20733 }, { "epoch": 3.95, "learning_rate": 1.816930409914204e-06, "loss": 3.1556, "step": 20734 }, { "epoch": 3.95, "learning_rate": 1.816873212583413e-06, "loss": 3.1917, "step": 20735 }, { "epoch": 3.95, "learning_rate": 1.8168160152526215e-06, "loss": 2.0768, "step": 20736 }, { "epoch": 3.95, "learning_rate": 1.8167588179218305e-06, "loss": 3.0474, "step": 20737 }, { "epoch": 3.95, "learning_rate": 1.8167016205910391e-06, "loss": 2.3894, "step": 20738 }, { "epoch": 3.95, "learning_rate": 1.816644423260248e-06, "loss": 1.975, "step": 20739 }, { "epoch": 3.95, "learning_rate": 1.816587225929457e-06, "loss": 1.9, "step": 20740 }, { "epoch": 3.95, "learning_rate": 1.8165300285986655e-06, "loss": 2.032, "step": 20741 }, { "epoch": 3.95, "learning_rate": 1.816472831267874e-06, "loss": 1.1908, "step": 20742 }, { "epoch": 3.95, "learning_rate": 1.816415633937083e-06, "loss": 2.013, "step": 20743 }, { "epoch": 3.95, "learning_rate": 1.8163584366062917e-06, "loss": 3.0602, "step": 20744 }, { "epoch": 3.95, "learning_rate": 1.8163012392755005e-06, "loss": 1.329, "step": 20745 }, { "epoch": 3.95, "learning_rate": 1.8162440419447095e-06, "loss": 1.9852, "step": 20746 }, { "epoch": 3.95, "learning_rate": 1.816186844613918e-06, "loss": 1.4151, "step": 20747 }, { "epoch": 3.95, "learning_rate": 1.8161296472831266e-06, "loss": 1.3746, "step": 20748 }, { "epoch": 3.95, "learning_rate": 1.8160724499523356e-06, "loss": 3.269, "step": 20749 }, { "epoch": 3.95, "learning_rate": 1.8160152526215444e-06, "loss": 2.7933, "step": 20750 }, { "epoch": 3.95, "learning_rate": 1.815958055290753e-06, "loss": 2.4186, "step": 20751 }, { "epoch": 3.95, "learning_rate": 1.815900857959962e-06, "loss": 0.6357, "step": 20752 }, { "epoch": 3.95, "learning_rate": 1.8158436606291706e-06, "loss": 1.8754, "step": 20753 }, { "epoch": 3.95, "learning_rate": 1.8157864632983796e-06, "loss": 3.3612, "step": 20754 }, { "epoch": 3.95, "learning_rate": 1.8157292659675882e-06, "loss": 2.7184, "step": 20755 }, { "epoch": 3.95, "learning_rate": 1.815672068636797e-06, "loss": 2.0407, "step": 20756 }, { "epoch": 3.95, "learning_rate": 1.815614871306006e-06, "loss": 0.7926, "step": 20757 }, { "epoch": 3.95, "learning_rate": 1.8155576739752146e-06, "loss": 1.8905, "step": 20758 }, { "epoch": 3.95, "learning_rate": 1.8155004766444231e-06, "loss": 2.5117, "step": 20759 }, { "epoch": 3.95, "learning_rate": 1.8154432793136322e-06, "loss": 1.5805, "step": 20760 }, { "epoch": 3.95, "learning_rate": 1.8153860819828407e-06, "loss": 2.5236, "step": 20761 }, { "epoch": 3.95, "learning_rate": 1.8153288846520495e-06, "loss": 2.7994, "step": 20762 }, { "epoch": 3.95, "learning_rate": 1.8152716873212585e-06, "loss": 1.9951, "step": 20763 }, { "epoch": 3.96, "learning_rate": 1.8152144899904671e-06, "loss": 2.8612, "step": 20764 }, { "epoch": 3.96, "learning_rate": 1.8151572926596757e-06, "loss": 1.7701, "step": 20765 }, { "epoch": 3.96, "learning_rate": 1.8151000953288847e-06, "loss": 1.8012, "step": 20766 }, { "epoch": 3.96, "learning_rate": 1.8150428979980935e-06, "loss": 0.7076, "step": 20767 }, { "epoch": 3.96, "learning_rate": 1.814985700667302e-06, "loss": 3.2815, "step": 20768 }, { "epoch": 3.96, "learning_rate": 1.814928503336511e-06, "loss": 2.4347, "step": 20769 }, { "epoch": 3.96, "learning_rate": 1.8148713060057197e-06, "loss": 2.1504, "step": 20770 }, { "epoch": 3.96, "learning_rate": 1.8148141086749287e-06, "loss": 0.854, "step": 20771 }, { "epoch": 3.96, "learning_rate": 1.8147569113441373e-06, "loss": 2.3415, "step": 20772 }, { "epoch": 3.96, "learning_rate": 1.814699714013346e-06, "loss": 2.7631, "step": 20773 }, { "epoch": 3.96, "learning_rate": 1.814642516682555e-06, "loss": 2.0561, "step": 20774 }, { "epoch": 3.96, "learning_rate": 1.8145853193517636e-06, "loss": 2.1436, "step": 20775 }, { "epoch": 3.96, "learning_rate": 1.8145281220209722e-06, "loss": 2.7792, "step": 20776 }, { "epoch": 3.96, "learning_rate": 1.8144709246901812e-06, "loss": 3.1661, "step": 20777 }, { "epoch": 3.96, "learning_rate": 1.81441372735939e-06, "loss": 2.7182, "step": 20778 }, { "epoch": 3.96, "learning_rate": 1.8143565300285986e-06, "loss": 3.0867, "step": 20779 }, { "epoch": 3.96, "learning_rate": 1.8142993326978076e-06, "loss": 1.9426, "step": 20780 }, { "epoch": 3.96, "learning_rate": 1.8142421353670162e-06, "loss": 1.2942, "step": 20781 }, { "epoch": 3.96, "learning_rate": 1.8141849380362248e-06, "loss": 1.9011, "step": 20782 }, { "epoch": 3.96, "learning_rate": 1.8141277407054338e-06, "loss": 3.3498, "step": 20783 }, { "epoch": 3.96, "learning_rate": 1.8140705433746426e-06, "loss": 1.6265, "step": 20784 }, { "epoch": 3.96, "learning_rate": 1.8140133460438516e-06, "loss": 1.9076, "step": 20785 }, { "epoch": 3.96, "learning_rate": 1.8139561487130602e-06, "loss": 3.4625, "step": 20786 }, { "epoch": 3.96, "learning_rate": 1.8138989513822687e-06, "loss": 1.2532, "step": 20787 }, { "epoch": 3.96, "learning_rate": 1.8138417540514777e-06, "loss": 3.1738, "step": 20788 }, { "epoch": 3.96, "learning_rate": 1.8137845567206865e-06, "loss": 2.9604, "step": 20789 }, { "epoch": 3.96, "learning_rate": 1.8137273593898951e-06, "loss": 2.6623, "step": 20790 }, { "epoch": 3.96, "learning_rate": 1.8136701620591041e-06, "loss": 3.1943, "step": 20791 }, { "epoch": 3.96, "learning_rate": 1.8136129647283127e-06, "loss": 2.5765, "step": 20792 }, { "epoch": 3.96, "learning_rate": 1.8135557673975213e-06, "loss": 1.8109, "step": 20793 }, { "epoch": 3.96, "learning_rate": 1.8134985700667303e-06, "loss": 1.4197, "step": 20794 }, { "epoch": 3.96, "learning_rate": 1.813441372735939e-06, "loss": 1.3728, "step": 20795 }, { "epoch": 3.96, "learning_rate": 1.8133841754051477e-06, "loss": 3.4084, "step": 20796 }, { "epoch": 3.96, "learning_rate": 1.8133269780743567e-06, "loss": 2.0782, "step": 20797 }, { "epoch": 3.96, "learning_rate": 1.8132697807435653e-06, "loss": 2.2461, "step": 20798 }, { "epoch": 3.96, "learning_rate": 1.813212583412774e-06, "loss": 2.6615, "step": 20799 }, { "epoch": 3.96, "learning_rate": 1.8131553860819828e-06, "loss": 2.6482, "step": 20800 }, { "epoch": 3.96, "learning_rate": 1.8130981887511916e-06, "loss": 2.2556, "step": 20801 }, { "epoch": 3.96, "learning_rate": 1.8130409914204006e-06, "loss": 2.0317, "step": 20802 }, { "epoch": 3.96, "learning_rate": 1.8129837940896092e-06, "loss": 3.6743, "step": 20803 }, { "epoch": 3.96, "learning_rate": 1.8129265967588178e-06, "loss": 2.929, "step": 20804 }, { "epoch": 3.96, "learning_rate": 1.8128693994280268e-06, "loss": 2.1436, "step": 20805 }, { "epoch": 3.96, "learning_rate": 1.8128122020972356e-06, "loss": 2.4089, "step": 20806 }, { "epoch": 3.96, "learning_rate": 1.8127550047664442e-06, "loss": 1.7632, "step": 20807 }, { "epoch": 3.96, "learning_rate": 1.8126978074356532e-06, "loss": 1.1639, "step": 20808 }, { "epoch": 3.96, "learning_rate": 1.8126406101048618e-06, "loss": 2.7442, "step": 20809 }, { "epoch": 3.96, "learning_rate": 1.8125834127740706e-06, "loss": 2.303, "step": 20810 }, { "epoch": 3.96, "learning_rate": 1.8125262154432794e-06, "loss": 3.6053, "step": 20811 }, { "epoch": 3.96, "learning_rate": 1.8124690181124882e-06, "loss": 2.0312, "step": 20812 }, { "epoch": 3.96, "learning_rate": 1.8124118207816967e-06, "loss": 3.0461, "step": 20813 }, { "epoch": 3.96, "learning_rate": 1.8123546234509057e-06, "loss": 2.5215, "step": 20814 }, { "epoch": 3.96, "learning_rate": 1.8122974261201143e-06, "loss": 2.1479, "step": 20815 }, { "epoch": 3.96, "learning_rate": 1.8122402287893231e-06, "loss": 3.0796, "step": 20816 }, { "epoch": 3.97, "learning_rate": 1.8121830314585321e-06, "loss": 1.556, "step": 20817 }, { "epoch": 3.97, "learning_rate": 1.8121258341277407e-06, "loss": 2.0517, "step": 20818 }, { "epoch": 3.97, "learning_rate": 1.8120686367969497e-06, "loss": 2.5803, "step": 20819 }, { "epoch": 3.97, "learning_rate": 1.8120114394661583e-06, "loss": 2.6462, "step": 20820 }, { "epoch": 3.97, "learning_rate": 1.8119542421353669e-06, "loss": 1.6986, "step": 20821 }, { "epoch": 3.97, "learning_rate": 1.8118970448045759e-06, "loss": 1.5309, "step": 20822 }, { "epoch": 3.97, "learning_rate": 1.8118398474737847e-06, "loss": 3.3766, "step": 20823 }, { "epoch": 3.97, "learning_rate": 1.8117826501429933e-06, "loss": 2.6547, "step": 20824 }, { "epoch": 3.97, "learning_rate": 1.8117254528122023e-06, "loss": 1.7987, "step": 20825 }, { "epoch": 3.97, "learning_rate": 1.8116682554814109e-06, "loss": 2.1736, "step": 20826 }, { "epoch": 3.97, "learning_rate": 1.8116110581506196e-06, "loss": 3.8784, "step": 20827 }, { "epoch": 3.97, "learning_rate": 1.8115538608198286e-06, "loss": 2.2644, "step": 20828 }, { "epoch": 3.97, "learning_rate": 1.8114966634890372e-06, "loss": 2.5347, "step": 20829 }, { "epoch": 3.97, "learning_rate": 1.8114394661582458e-06, "loss": 1.9822, "step": 20830 }, { "epoch": 3.97, "learning_rate": 1.8113822688274548e-06, "loss": 2.0967, "step": 20831 }, { "epoch": 3.97, "learning_rate": 1.8113250714966634e-06, "loss": 2.1618, "step": 20832 }, { "epoch": 3.97, "learning_rate": 1.8112678741658722e-06, "loss": 1.2976, "step": 20833 }, { "epoch": 3.97, "learning_rate": 1.8112106768350812e-06, "loss": 2.3501, "step": 20834 }, { "epoch": 3.97, "learning_rate": 1.8111534795042898e-06, "loss": 2.4165, "step": 20835 }, { "epoch": 3.97, "learning_rate": 1.8110962821734988e-06, "loss": 2.2459, "step": 20836 }, { "epoch": 3.97, "learning_rate": 1.8110390848427074e-06, "loss": 2.5198, "step": 20837 }, { "epoch": 3.97, "learning_rate": 1.8109818875119162e-06, "loss": 2.3568, "step": 20838 }, { "epoch": 3.97, "learning_rate": 1.810924690181125e-06, "loss": 2.8707, "step": 20839 }, { "epoch": 3.97, "learning_rate": 1.8108674928503338e-06, "loss": 2.3075, "step": 20840 }, { "epoch": 3.97, "learning_rate": 1.8108102955195423e-06, "loss": 2.147, "step": 20841 }, { "epoch": 3.97, "learning_rate": 1.8107530981887513e-06, "loss": 2.999, "step": 20842 }, { "epoch": 3.97, "learning_rate": 1.81069590085796e-06, "loss": 3.0254, "step": 20843 }, { "epoch": 3.97, "learning_rate": 1.8106387035271687e-06, "loss": 2.121, "step": 20844 }, { "epoch": 3.97, "learning_rate": 1.8105815061963777e-06, "loss": 2.5937, "step": 20845 }, { "epoch": 3.97, "learning_rate": 1.8105243088655863e-06, "loss": 2.3461, "step": 20846 }, { "epoch": 3.97, "learning_rate": 1.8104671115347949e-06, "loss": 2.3958, "step": 20847 }, { "epoch": 3.97, "learning_rate": 1.8104099142040039e-06, "loss": 2.3072, "step": 20848 }, { "epoch": 3.97, "learning_rate": 1.8103527168732125e-06, "loss": 0.8855, "step": 20849 }, { "epoch": 3.97, "learning_rate": 1.8102955195424215e-06, "loss": 1.9567, "step": 20850 }, { "epoch": 3.97, "learning_rate": 1.8102383222116303e-06, "loss": 1.2734, "step": 20851 }, { "epoch": 3.97, "learning_rate": 1.8101811248808389e-06, "loss": 2.0464, "step": 20852 }, { "epoch": 3.97, "learning_rate": 1.8101239275500479e-06, "loss": 1.4347, "step": 20853 }, { "epoch": 3.97, "learning_rate": 1.8100667302192564e-06, "loss": 2.1214, "step": 20854 }, { "epoch": 3.97, "learning_rate": 1.8100095328884652e-06, "loss": 1.8981, "step": 20855 }, { "epoch": 3.97, "learning_rate": 1.8099523355576742e-06, "loss": 2.0825, "step": 20856 }, { "epoch": 3.97, "learning_rate": 1.8098951382268828e-06, "loss": 3.3402, "step": 20857 }, { "epoch": 3.97, "learning_rate": 1.8098379408960914e-06, "loss": 1.7782, "step": 20858 }, { "epoch": 3.97, "learning_rate": 1.8097807435653004e-06, "loss": 2.5573, "step": 20859 }, { "epoch": 3.97, "learning_rate": 1.809723546234509e-06, "loss": 2.1826, "step": 20860 }, { "epoch": 3.97, "learning_rate": 1.8096663489037178e-06, "loss": 2.7276, "step": 20861 }, { "epoch": 3.97, "learning_rate": 1.8096091515729268e-06, "loss": 1.6268, "step": 20862 }, { "epoch": 3.97, "learning_rate": 1.8095519542421354e-06, "loss": 2.1585, "step": 20863 }, { "epoch": 3.97, "learning_rate": 1.809494756911344e-06, "loss": 2.1612, "step": 20864 }, { "epoch": 3.97, "learning_rate": 1.809437559580553e-06, "loss": 1.8826, "step": 20865 }, { "epoch": 3.97, "learning_rate": 1.8093803622497618e-06, "loss": 1.3049, "step": 20866 }, { "epoch": 3.97, "learning_rate": 1.8093231649189705e-06, "loss": 1.9493, "step": 20867 }, { "epoch": 3.97, "learning_rate": 1.8092659675881793e-06, "loss": 1.4941, "step": 20868 }, { "epoch": 3.98, "learning_rate": 1.809208770257388e-06, "loss": 2.3485, "step": 20869 }, { "epoch": 3.98, "learning_rate": 1.809151572926597e-06, "loss": 2.8883, "step": 20870 }, { "epoch": 3.98, "learning_rate": 1.8090943755958055e-06, "loss": 2.9656, "step": 20871 }, { "epoch": 3.98, "learning_rate": 1.8090371782650143e-06, "loss": 2.0819, "step": 20872 }, { "epoch": 3.98, "learning_rate": 1.8089799809342233e-06, "loss": 1.7982, "step": 20873 }, { "epoch": 3.98, "learning_rate": 1.808922783603432e-06, "loss": 1.9046, "step": 20874 }, { "epoch": 3.98, "learning_rate": 1.8088655862726405e-06, "loss": 1.7519, "step": 20875 }, { "epoch": 3.98, "learning_rate": 1.8088083889418495e-06, "loss": 1.7334, "step": 20876 }, { "epoch": 3.98, "learning_rate": 1.8087511916110583e-06, "loss": 3.7219, "step": 20877 }, { "epoch": 3.98, "learning_rate": 1.8086939942802669e-06, "loss": 2.3082, "step": 20878 }, { "epoch": 3.98, "learning_rate": 1.8086367969494759e-06, "loss": 2.699, "step": 20879 }, { "epoch": 3.98, "learning_rate": 1.8085795996186844e-06, "loss": 3.2358, "step": 20880 }, { "epoch": 3.98, "learning_rate": 1.808522402287893e-06, "loss": 2.0087, "step": 20881 }, { "epoch": 3.98, "learning_rate": 1.808465204957102e-06, "loss": 2.1984, "step": 20882 }, { "epoch": 3.98, "learning_rate": 1.8084080076263108e-06, "loss": 2.2808, "step": 20883 }, { "epoch": 3.98, "learning_rate": 1.8083508102955198e-06, "loss": 2.6219, "step": 20884 }, { "epoch": 3.98, "learning_rate": 1.8082936129647284e-06, "loss": 1.9688, "step": 20885 }, { "epoch": 3.98, "learning_rate": 1.808236415633937e-06, "loss": 2.3359, "step": 20886 }, { "epoch": 3.98, "learning_rate": 1.808179218303146e-06, "loss": 2.4077, "step": 20887 }, { "epoch": 3.98, "learning_rate": 1.8081220209723546e-06, "loss": 1.739, "step": 20888 }, { "epoch": 3.98, "learning_rate": 1.8080648236415634e-06, "loss": 2.6126, "step": 20889 }, { "epoch": 3.98, "learning_rate": 1.8080076263107724e-06, "loss": 1.9863, "step": 20890 }, { "epoch": 3.98, "learning_rate": 1.807950428979981e-06, "loss": 2.7051, "step": 20891 }, { "epoch": 3.98, "learning_rate": 1.8078932316491895e-06, "loss": 3.3201, "step": 20892 }, { "epoch": 3.98, "learning_rate": 1.8078360343183986e-06, "loss": 2.524, "step": 20893 }, { "epoch": 3.98, "learning_rate": 1.8077788369876073e-06, "loss": 1.8566, "step": 20894 }, { "epoch": 3.98, "learning_rate": 1.807721639656816e-06, "loss": 1.6939, "step": 20895 }, { "epoch": 3.98, "learning_rate": 1.807664442326025e-06, "loss": 2.1112, "step": 20896 }, { "epoch": 3.98, "learning_rate": 1.8076072449952335e-06, "loss": 2.1847, "step": 20897 }, { "epoch": 3.98, "learning_rate": 1.8075500476644423e-06, "loss": 2.4819, "step": 20898 }, { "epoch": 3.98, "learning_rate": 1.807492850333651e-06, "loss": 1.115, "step": 20899 }, { "epoch": 3.98, "learning_rate": 1.80743565300286e-06, "loss": 2.1932, "step": 20900 }, { "epoch": 3.98, "learning_rate": 1.807378455672069e-06, "loss": 2.7751, "step": 20901 }, { "epoch": 3.98, "learning_rate": 1.8073212583412775e-06, "loss": 2.8969, "step": 20902 }, { "epoch": 3.98, "learning_rate": 1.807264061010486e-06, "loss": 1.5275, "step": 20903 }, { "epoch": 3.98, "learning_rate": 1.807206863679695e-06, "loss": 2.6002, "step": 20904 }, { "epoch": 3.98, "learning_rate": 1.8071496663489039e-06, "loss": 2.6339, "step": 20905 }, { "epoch": 3.98, "learning_rate": 1.8070924690181124e-06, "loss": 2.3915, "step": 20906 }, { "epoch": 3.98, "learning_rate": 1.8070352716873215e-06, "loss": 2.3241, "step": 20907 }, { "epoch": 3.98, "learning_rate": 1.80697807435653e-06, "loss": 2.8408, "step": 20908 }, { "epoch": 3.98, "learning_rate": 1.8069208770257386e-06, "loss": 2.9582, "step": 20909 }, { "epoch": 3.98, "learning_rate": 1.8068636796949476e-06, "loss": 3.055, "step": 20910 }, { "epoch": 3.98, "learning_rate": 1.8068064823641564e-06, "loss": 2.2818, "step": 20911 }, { "epoch": 3.98, "learning_rate": 1.806749285033365e-06, "loss": 2.2694, "step": 20912 }, { "epoch": 3.98, "learning_rate": 1.806692087702574e-06, "loss": 1.3021, "step": 20913 }, { "epoch": 3.98, "learning_rate": 1.8066348903717826e-06, "loss": 1.508, "step": 20914 }, { "epoch": 3.98, "learning_rate": 1.8065776930409916e-06, "loss": 2.0293, "step": 20915 }, { "epoch": 3.98, "learning_rate": 1.8065204957102002e-06, "loss": 2.7317, "step": 20916 }, { "epoch": 3.98, "learning_rate": 1.806463298379409e-06, "loss": 2.0167, "step": 20917 }, { "epoch": 3.98, "learning_rate": 1.806406101048618e-06, "loss": 1.209, "step": 20918 }, { "epoch": 3.98, "learning_rate": 1.8063489037178266e-06, "loss": 2.6725, "step": 20919 }, { "epoch": 3.98, "learning_rate": 1.8062917063870351e-06, "loss": 1.5484, "step": 20920 }, { "epoch": 3.98, "learning_rate": 1.8062345090562441e-06, "loss": 1.6167, "step": 20921 }, { "epoch": 3.99, "learning_rate": 1.806177311725453e-06, "loss": 1.3452, "step": 20922 }, { "epoch": 3.99, "learning_rate": 1.8061201143946615e-06, "loss": 2.4909, "step": 20923 }, { "epoch": 3.99, "learning_rate": 1.8060629170638705e-06, "loss": 1.4846, "step": 20924 }, { "epoch": 3.99, "learning_rate": 1.8060057197330791e-06, "loss": 1.6977, "step": 20925 }, { "epoch": 3.99, "learning_rate": 1.805948522402288e-06, "loss": 2.0716, "step": 20926 }, { "epoch": 3.99, "learning_rate": 1.8058913250714967e-06, "loss": 3.1752, "step": 20927 }, { "epoch": 3.99, "learning_rate": 1.8058341277407055e-06, "loss": 3.0703, "step": 20928 }, { "epoch": 3.99, "learning_rate": 1.805776930409914e-06, "loss": 2.5358, "step": 20929 }, { "epoch": 3.99, "learning_rate": 1.805719733079123e-06, "loss": 2.2815, "step": 20930 }, { "epoch": 3.99, "learning_rate": 1.8056625357483317e-06, "loss": 2.4589, "step": 20931 }, { "epoch": 3.99, "learning_rate": 1.8056053384175407e-06, "loss": 0.9616, "step": 20932 }, { "epoch": 3.99, "learning_rate": 1.8055481410867495e-06, "loss": 2.2683, "step": 20933 }, { "epoch": 3.99, "learning_rate": 1.805490943755958e-06, "loss": 2.5996, "step": 20934 }, { "epoch": 3.99, "learning_rate": 1.805433746425167e-06, "loss": 1.7511, "step": 20935 }, { "epoch": 3.99, "learning_rate": 1.8053765490943756e-06, "loss": 1.9582, "step": 20936 }, { "epoch": 3.99, "learning_rate": 1.8053193517635842e-06, "loss": 1.7531, "step": 20937 }, { "epoch": 3.99, "learning_rate": 1.8052621544327932e-06, "loss": 1.6986, "step": 20938 }, { "epoch": 3.99, "learning_rate": 1.805204957102002e-06, "loss": 2.1464, "step": 20939 }, { "epoch": 3.99, "learning_rate": 1.8051477597712106e-06, "loss": 2.5076, "step": 20940 }, { "epoch": 3.99, "learning_rate": 1.8050905624404196e-06, "loss": 1.4512, "step": 20941 }, { "epoch": 3.99, "learning_rate": 1.8050333651096282e-06, "loss": 1.8865, "step": 20942 }, { "epoch": 3.99, "learning_rate": 1.804976167778837e-06, "loss": 2.3981, "step": 20943 }, { "epoch": 3.99, "learning_rate": 1.804918970448046e-06, "loss": 1.6829, "step": 20944 }, { "epoch": 3.99, "learning_rate": 1.8048617731172546e-06, "loss": 2.2681, "step": 20945 }, { "epoch": 3.99, "learning_rate": 1.8048045757864631e-06, "loss": 0.8431, "step": 20946 }, { "epoch": 3.99, "learning_rate": 1.8047473784556721e-06, "loss": 1.8574, "step": 20947 }, { "epoch": 3.99, "learning_rate": 1.8046901811248807e-06, "loss": 1.6576, "step": 20948 }, { "epoch": 3.99, "learning_rate": 1.8046329837940897e-06, "loss": 2.3067, "step": 20949 }, { "epoch": 3.99, "learning_rate": 1.8045757864632985e-06, "loss": 3.1115, "step": 20950 }, { "epoch": 3.99, "learning_rate": 1.8045185891325071e-06, "loss": 2.2613, "step": 20951 }, { "epoch": 3.99, "learning_rate": 1.8044613918017161e-06, "loss": 1.6051, "step": 20952 }, { "epoch": 3.99, "learning_rate": 1.8044041944709247e-06, "loss": 1.1584, "step": 20953 }, { "epoch": 3.99, "learning_rate": 1.8043469971401335e-06, "loss": 2.694, "step": 20954 }, { "epoch": 3.99, "learning_rate": 1.8042897998093423e-06, "loss": 1.2655, "step": 20955 }, { "epoch": 3.99, "learning_rate": 1.804232602478551e-06, "loss": 3.515, "step": 20956 }, { "epoch": 3.99, "learning_rate": 1.8041754051477597e-06, "loss": 2.7319, "step": 20957 }, { "epoch": 3.99, "learning_rate": 1.8041182078169687e-06, "loss": 2.2382, "step": 20958 }, { "epoch": 3.99, "learning_rate": 1.8040610104861773e-06, "loss": 2.5122, "step": 20959 }, { "epoch": 3.99, "learning_rate": 1.804003813155386e-06, "loss": 2.0722, "step": 20960 }, { "epoch": 3.99, "learning_rate": 1.803946615824595e-06, "loss": 2.2852, "step": 20961 }, { "epoch": 3.99, "learning_rate": 1.8038894184938036e-06, "loss": 1.8662, "step": 20962 }, { "epoch": 3.99, "learning_rate": 1.8038322211630122e-06, "loss": 2.0965, "step": 20963 }, { "epoch": 3.99, "learning_rate": 1.8037750238322212e-06, "loss": 1.6749, "step": 20964 }, { "epoch": 3.99, "learning_rate": 1.80371782650143e-06, "loss": 3.2155, "step": 20965 }, { "epoch": 3.99, "learning_rate": 1.8036606291706388e-06, "loss": 1.9181, "step": 20966 }, { "epoch": 3.99, "learning_rate": 1.8036034318398476e-06, "loss": 3.0619, "step": 20967 }, { "epoch": 3.99, "learning_rate": 1.8035462345090562e-06, "loss": 2.0757, "step": 20968 }, { "epoch": 3.99, "learning_rate": 1.8034890371782652e-06, "loss": 2.5441, "step": 20969 }, { "epoch": 3.99, "learning_rate": 1.8034318398474738e-06, "loss": 1.8923, "step": 20970 }, { "epoch": 3.99, "learning_rate": 1.8033746425166826e-06, "loss": 3.2666, "step": 20971 }, { "epoch": 3.99, "learning_rate": 1.8033174451858916e-06, "loss": 1.3163, "step": 20972 }, { "epoch": 3.99, "learning_rate": 1.8032602478551002e-06, "loss": 1.5598, "step": 20973 }, { "epoch": 4.0, "learning_rate": 1.8032030505243087e-06, "loss": 2.6456, "step": 20974 }, { "epoch": 4.0, "learning_rate": 1.8031458531935177e-06, "loss": 3.0459, "step": 20975 }, { "epoch": 4.0, "learning_rate": 1.8030886558627263e-06, "loss": 2.734, "step": 20976 }, { "epoch": 4.0, "learning_rate": 1.8030314585319351e-06, "loss": 1.581, "step": 20977 }, { "epoch": 4.0, "learning_rate": 1.8029742612011441e-06, "loss": 1.4076, "step": 20978 }, { "epoch": 4.0, "learning_rate": 1.8029170638703527e-06, "loss": 2.8973, "step": 20979 }, { "epoch": 4.0, "learning_rate": 1.8028598665395617e-06, "loss": 1.614, "step": 20980 }, { "epoch": 4.0, "learning_rate": 1.8028026692087703e-06, "loss": 2.263, "step": 20981 }, { "epoch": 4.0, "learning_rate": 1.802745471877979e-06, "loss": 1.9649, "step": 20982 }, { "epoch": 4.0, "learning_rate": 1.802688274547188e-06, "loss": 1.7919, "step": 20983 }, { "epoch": 4.0, "learning_rate": 1.8026310772163967e-06, "loss": 2.51, "step": 20984 }, { "epoch": 4.0, "learning_rate": 1.8025738798856053e-06, "loss": 1.8226, "step": 20985 }, { "epoch": 4.0, "learning_rate": 1.8025166825548143e-06, "loss": 3.2909, "step": 20986 }, { "epoch": 4.0, "learning_rate": 1.8024594852240228e-06, "loss": 1.0403, "step": 20987 }, { "epoch": 4.0, "learning_rate": 1.8024022878932316e-06, "loss": 3.4105, "step": 20988 }, { "epoch": 4.0, "learning_rate": 1.8023450905624406e-06, "loss": 1.4181, "step": 20989 }, { "epoch": 4.0, "learning_rate": 1.8022878932316492e-06, "loss": 2.0051, "step": 20990 }, { "epoch": 4.0, "learning_rate": 1.8022306959008578e-06, "loss": 1.8617, "step": 20991 }, { "epoch": 4.0, "learning_rate": 1.8021734985700668e-06, "loss": 1.76, "step": 20992 }, { "epoch": 4.0, "learning_rate": 1.8021163012392756e-06, "loss": 2.5157, "step": 20993 }, { "epoch": 4.0, "learning_rate": 1.8020591039084842e-06, "loss": 1.7974, "step": 20994 }, { "epoch": 4.0, "learning_rate": 1.8020019065776932e-06, "loss": 1.587, "step": 20995 }, { "epoch": 4.0, "learning_rate": 1.8019447092469018e-06, "loss": 2.8637, "step": 20996 }, { "epoch": 4.0, "learning_rate": 1.8018875119161108e-06, "loss": 2.6798, "step": 20997 }, { "epoch": 4.0, "learning_rate": 1.8018303145853194e-06, "loss": 1.3745, "step": 20998 }, { "epoch": 4.0, "learning_rate": 1.8017731172545282e-06, "loss": 1.5743, "step": 20999 }, { "epoch": 4.0, "learning_rate": 1.8017159199237372e-06, "loss": 4.1255, "step": 21000 }, { "epoch": 4.0, "eval_accuracy": 0.44192027432490355, "eval_loss": 2.127500057220459, "eval_runtime": 125.1034, "eval_samples_per_second": 18.649, "eval_steps_per_second": 4.668, "step": 21000 }, { "epoch": 4.0, "learning_rate": 1.8016587225929457e-06, "loss": 2.1671, "step": 21001 }, { "epoch": 4.0, "learning_rate": 1.8016015252621543e-06, "loss": 2.3027, "step": 21002 }, { "epoch": 4.0, "learning_rate": 1.8015443279313633e-06, "loss": 2.4984, "step": 21003 }, { "epoch": 4.0, "learning_rate": 1.801487130600572e-06, "loss": 1.6354, "step": 21004 }, { "epoch": 4.0, "learning_rate": 1.8014299332697807e-06, "loss": 1.468, "step": 21005 }, { "epoch": 4.0, "learning_rate": 1.8013727359389897e-06, "loss": 3.3246, "step": 21006 }, { "epoch": 4.0, "learning_rate": 1.8013155386081983e-06, "loss": 1.7632, "step": 21007 }, { "epoch": 4.0, "learning_rate": 1.8012583412774069e-06, "loss": 2.206, "step": 21008 }, { "epoch": 4.0, "learning_rate": 1.8012011439466159e-06, "loss": 2.6363, "step": 21009 }, { "epoch": 4.0, "learning_rate": 1.8011439466158247e-06, "loss": 2.512, "step": 21010 }, { "epoch": 4.0, "learning_rate": 1.8010867492850333e-06, "loss": 1.0088, "step": 21011 }, { "epoch": 4.0, "learning_rate": 1.8010295519542423e-06, "loss": 2.5021, "step": 21012 }, { "epoch": 4.0, "learning_rate": 1.8009723546234508e-06, "loss": 1.7404, "step": 21013 }, { "epoch": 4.0, "learning_rate": 1.8009151572926599e-06, "loss": 1.0325, "step": 21014 }, { "epoch": 4.0, "learning_rate": 1.8008579599618684e-06, "loss": 2.6169, "step": 21015 }, { "epoch": 4.0, "learning_rate": 1.8008007626310772e-06, "loss": 2.1647, "step": 21016 }, { "epoch": 4.0, "learning_rate": 1.8007435653002862e-06, "loss": 2.7453, "step": 21017 }, { "epoch": 4.0, "learning_rate": 1.8006863679694948e-06, "loss": 1.9275, "step": 21018 }, { "epoch": 4.0, "learning_rate": 1.8006291706387034e-06, "loss": 2.0192, "step": 21019 }, { "epoch": 4.0, "learning_rate": 1.8005719733079124e-06, "loss": 3.5497, "step": 21020 }, { "epoch": 4.0, "learning_rate": 1.8005147759771212e-06, "loss": 3.2334, "step": 21021 }, { "epoch": 4.0, "learning_rate": 1.8004575786463298e-06, "loss": 2.3402, "step": 21022 }, { "epoch": 4.0, "learning_rate": 1.8004003813155388e-06, "loss": 2.3126, "step": 21023 }, { "epoch": 4.0, "learning_rate": 1.8003431839847474e-06, "loss": 2.2868, "step": 21024 }, { "epoch": 4.0, "learning_rate": 1.800285986653956e-06, "loss": 2.4516, "step": 21025 }, { "epoch": 4.0, "learning_rate": 1.800228789323165e-06, "loss": 1.8718, "step": 21026 }, { "epoch": 4.01, "learning_rate": 1.8001715919923737e-06, "loss": 1.2791, "step": 21027 }, { "epoch": 4.01, "learning_rate": 1.8001143946615823e-06, "loss": 1.9234, "step": 21028 }, { "epoch": 4.01, "learning_rate": 1.8000571973307913e-06, "loss": 1.971, "step": 21029 }, { "epoch": 4.01, "learning_rate": 1.8e-06, "loss": 2.8588, "step": 21030 }, { "epoch": 4.01, "learning_rate": 1.799942802669209e-06, "loss": 2.5238, "step": 21031 }, { "epoch": 4.01, "learning_rate": 1.7998856053384177e-06, "loss": 3.0986, "step": 21032 }, { "epoch": 4.01, "learning_rate": 1.7998284080076263e-06, "loss": 1.873, "step": 21033 }, { "epoch": 4.01, "learning_rate": 1.7997712106768353e-06, "loss": 1.8312, "step": 21034 }, { "epoch": 4.01, "learning_rate": 1.7997140133460439e-06, "loss": 1.9821, "step": 21035 }, { "epoch": 4.01, "learning_rate": 1.7996568160152525e-06, "loss": 2.1305, "step": 21036 }, { "epoch": 4.01, "learning_rate": 1.7995996186844615e-06, "loss": 2.1978, "step": 21037 }, { "epoch": 4.01, "learning_rate": 1.7995424213536703e-06, "loss": 1.8684, "step": 21038 }, { "epoch": 4.01, "learning_rate": 1.7994852240228789e-06, "loss": 1.3735, "step": 21039 }, { "epoch": 4.01, "learning_rate": 1.7994280266920879e-06, "loss": 1.9824, "step": 21040 }, { "epoch": 4.01, "learning_rate": 1.7993708293612964e-06, "loss": 2.364, "step": 21041 }, { "epoch": 4.01, "learning_rate": 1.7993136320305052e-06, "loss": 2.271, "step": 21042 }, { "epoch": 4.01, "learning_rate": 1.799256434699714e-06, "loss": 3.5367, "step": 21043 }, { "epoch": 4.01, "learning_rate": 1.7991992373689228e-06, "loss": 1.7681, "step": 21044 }, { "epoch": 4.01, "learning_rate": 1.7991420400381318e-06, "loss": 2.046, "step": 21045 }, { "epoch": 4.01, "learning_rate": 1.7990848427073404e-06, "loss": 3.2381, "step": 21046 }, { "epoch": 4.01, "learning_rate": 1.799027645376549e-06, "loss": 1.2618, "step": 21047 }, { "epoch": 4.01, "learning_rate": 1.798970448045758e-06, "loss": 1.4689, "step": 21048 }, { "epoch": 4.01, "learning_rate": 1.7989132507149668e-06, "loss": 2.2321, "step": 21049 }, { "epoch": 4.01, "learning_rate": 1.7988560533841754e-06, "loss": 1.8314, "step": 21050 }, { "epoch": 4.01, "learning_rate": 1.7987988560533844e-06, "loss": 3.1766, "step": 21051 }, { "epoch": 4.01, "learning_rate": 1.798741658722593e-06, "loss": 2.3561, "step": 21052 }, { "epoch": 4.01, "learning_rate": 1.7986844613918018e-06, "loss": 2.2096, "step": 21053 }, { "epoch": 4.01, "learning_rate": 1.7986272640610105e-06, "loss": 2.9483, "step": 21054 }, { "epoch": 4.01, "learning_rate": 1.7985700667302193e-06, "loss": 2.0551, "step": 21055 }, { "epoch": 4.01, "learning_rate": 1.798512869399428e-06, "loss": 1.3217, "step": 21056 }, { "epoch": 4.01, "learning_rate": 1.798455672068637e-06, "loss": 2.2091, "step": 21057 }, { "epoch": 4.01, "learning_rate": 1.7983984747378455e-06, "loss": 2.0086, "step": 21058 }, { "epoch": 4.01, "learning_rate": 1.7983412774070543e-06, "loss": 2.6738, "step": 21059 }, { "epoch": 4.01, "learning_rate": 1.7982840800762633e-06, "loss": 2.2189, "step": 21060 }, { "epoch": 4.01, "learning_rate": 1.7982268827454719e-06, "loss": 4.0848, "step": 21061 }, { "epoch": 4.01, "learning_rate": 1.798169685414681e-06, "loss": 1.785, "step": 21062 }, { "epoch": 4.01, "learning_rate": 1.7981124880838895e-06, "loss": 2.8731, "step": 21063 }, { "epoch": 4.01, "learning_rate": 1.798055290753098e-06, "loss": 1.7391, "step": 21064 }, { "epoch": 4.01, "learning_rate": 1.797998093422307e-06, "loss": 1.7409, "step": 21065 }, { "epoch": 4.01, "learning_rate": 1.7979408960915159e-06, "loss": 2.5378, "step": 21066 }, { "epoch": 4.01, "learning_rate": 1.7978836987607244e-06, "loss": 1.4952, "step": 21067 }, { "epoch": 4.01, "learning_rate": 1.7978265014299334e-06, "loss": 2.4402, "step": 21068 }, { "epoch": 4.01, "learning_rate": 1.797769304099142e-06, "loss": 4.0098, "step": 21069 }, { "epoch": 4.01, "learning_rate": 1.7977121067683508e-06, "loss": 1.9293, "step": 21070 }, { "epoch": 4.01, "learning_rate": 1.7976549094375596e-06, "loss": 2.7113, "step": 21071 }, { "epoch": 4.01, "learning_rate": 1.7975977121067684e-06, "loss": 1.8293, "step": 21072 }, { "epoch": 4.01, "learning_rate": 1.797540514775977e-06, "loss": 3.0452, "step": 21073 }, { "epoch": 4.01, "learning_rate": 1.797483317445186e-06, "loss": 3.1262, "step": 21074 }, { "epoch": 4.01, "learning_rate": 1.7974261201143946e-06, "loss": 2.1796, "step": 21075 }, { "epoch": 4.01, "learning_rate": 1.7973689227836034e-06, "loss": 1.3421, "step": 21076 }, { "epoch": 4.01, "learning_rate": 1.7973117254528124e-06, "loss": 1.7879, "step": 21077 }, { "epoch": 4.01, "learning_rate": 1.797254528122021e-06, "loss": 2.6177, "step": 21078 }, { "epoch": 4.02, "learning_rate": 1.79719733079123e-06, "loss": 1.7425, "step": 21079 }, { "epoch": 4.02, "learning_rate": 1.7971401334604386e-06, "loss": 2.8623, "step": 21080 }, { "epoch": 4.02, "learning_rate": 1.7970829361296473e-06, "loss": 1.4704, "step": 21081 }, { "epoch": 4.02, "learning_rate": 1.7970257387988561e-06, "loss": 1.8793, "step": 21082 }, { "epoch": 4.02, "learning_rate": 1.796968541468065e-06, "loss": 2.0092, "step": 21083 }, { "epoch": 4.02, "learning_rate": 1.7969113441372735e-06, "loss": 1.6077, "step": 21084 }, { "epoch": 4.02, "learning_rate": 1.7968541468064825e-06, "loss": 2.8897, "step": 21085 }, { "epoch": 4.02, "learning_rate": 1.796796949475691e-06, "loss": 2.2292, "step": 21086 }, { "epoch": 4.02, "learning_rate": 1.7967397521449e-06, "loss": 2.6056, "step": 21087 }, { "epoch": 4.02, "learning_rate": 1.796682554814109e-06, "loss": 1.0598, "step": 21088 }, { "epoch": 4.02, "learning_rate": 1.7966253574833175e-06, "loss": 2.1825, "step": 21089 }, { "epoch": 4.02, "learning_rate": 1.796568160152526e-06, "loss": 2.6683, "step": 21090 }, { "epoch": 4.02, "learning_rate": 1.796510962821735e-06, "loss": 1.8884, "step": 21091 }, { "epoch": 4.02, "learning_rate": 1.7964537654909437e-06, "loss": 1.8702, "step": 21092 }, { "epoch": 4.02, "learning_rate": 1.7963965681601524e-06, "loss": 2.9073, "step": 21093 }, { "epoch": 4.02, "learning_rate": 1.7963393708293615e-06, "loss": 1.8195, "step": 21094 }, { "epoch": 4.02, "learning_rate": 1.79628217349857e-06, "loss": 3.7939, "step": 21095 }, { "epoch": 4.02, "learning_rate": 1.796224976167779e-06, "loss": 1.4214, "step": 21096 }, { "epoch": 4.02, "learning_rate": 1.7961677788369876e-06, "loss": 1.943, "step": 21097 }, { "epoch": 4.02, "learning_rate": 1.7961105815061964e-06, "loss": 2.1403, "step": 21098 }, { "epoch": 4.02, "learning_rate": 1.7960533841754054e-06, "loss": 2.4772, "step": 21099 }, { "epoch": 4.02, "learning_rate": 1.795996186844614e-06, "loss": 2.3653, "step": 21100 }, { "epoch": 4.02, "learning_rate": 1.7959389895138226e-06, "loss": 1.9969, "step": 21101 }, { "epoch": 4.02, "learning_rate": 1.7958817921830316e-06, "loss": 2.8289, "step": 21102 }, { "epoch": 4.02, "learning_rate": 1.7958245948522402e-06, "loss": 3.1773, "step": 21103 }, { "epoch": 4.02, "learning_rate": 1.795767397521449e-06, "loss": 2.8412, "step": 21104 }, { "epoch": 4.02, "learning_rate": 1.795710200190658e-06, "loss": 2.1517, "step": 21105 }, { "epoch": 4.02, "learning_rate": 1.7956530028598666e-06, "loss": 2.3552, "step": 21106 }, { "epoch": 4.02, "learning_rate": 1.7955958055290751e-06, "loss": 1.9223, "step": 21107 }, { "epoch": 4.02, "learning_rate": 1.7955386081982841e-06, "loss": 2.1418, "step": 21108 }, { "epoch": 4.02, "learning_rate": 1.795481410867493e-06, "loss": 2.3227, "step": 21109 }, { "epoch": 4.02, "learning_rate": 1.7954242135367017e-06, "loss": 2.0446, "step": 21110 }, { "epoch": 4.02, "learning_rate": 1.7953670162059105e-06, "loss": 2.276, "step": 21111 }, { "epoch": 4.02, "learning_rate": 1.795309818875119e-06, "loss": 1.3902, "step": 21112 }, { "epoch": 4.02, "learning_rate": 1.7952526215443281e-06, "loss": 2.9737, "step": 21113 }, { "epoch": 4.02, "learning_rate": 1.7951954242135367e-06, "loss": 1.6026, "step": 21114 }, { "epoch": 4.02, "learning_rate": 1.7951382268827455e-06, "loss": 2.6991, "step": 21115 }, { "epoch": 4.02, "learning_rate": 1.7950810295519545e-06, "loss": 1.8268, "step": 21116 }, { "epoch": 4.02, "learning_rate": 1.795023832221163e-06, "loss": 2.9276, "step": 21117 }, { "epoch": 4.02, "learning_rate": 1.7949666348903717e-06, "loss": 1.5798, "step": 21118 }, { "epoch": 4.02, "learning_rate": 1.7949094375595807e-06, "loss": 1.2556, "step": 21119 }, { "epoch": 4.02, "learning_rate": 1.7948522402287895e-06, "loss": 2.4941, "step": 21120 }, { "epoch": 4.02, "learning_rate": 1.794795042897998e-06, "loss": 3.1165, "step": 21121 }, { "epoch": 4.02, "learning_rate": 1.794737845567207e-06, "loss": 2.383, "step": 21122 }, { "epoch": 4.02, "learning_rate": 1.7946806482364156e-06, "loss": 3.6259, "step": 21123 }, { "epoch": 4.02, "learning_rate": 1.7946234509056242e-06, "loss": 3.2343, "step": 21124 }, { "epoch": 4.02, "learning_rate": 1.7945662535748332e-06, "loss": 2.3456, "step": 21125 }, { "epoch": 4.02, "learning_rate": 1.794509056244042e-06, "loss": 3.0867, "step": 21126 }, { "epoch": 4.02, "learning_rate": 1.794451858913251e-06, "loss": 1.8002, "step": 21127 }, { "epoch": 4.02, "learning_rate": 1.7943946615824596e-06, "loss": 3.2052, "step": 21128 }, { "epoch": 4.02, "learning_rate": 1.7943374642516682e-06, "loss": 2.4958, "step": 21129 }, { "epoch": 4.02, "learning_rate": 1.7942802669208772e-06, "loss": 2.1831, "step": 21130 }, { "epoch": 4.02, "learning_rate": 1.7942230695900858e-06, "loss": 2.4617, "step": 21131 }, { "epoch": 4.03, "learning_rate": 1.7941658722592946e-06, "loss": 1.5008, "step": 21132 }, { "epoch": 4.03, "learning_rate": 1.7941086749285036e-06, "loss": 2.9447, "step": 21133 }, { "epoch": 4.03, "learning_rate": 1.7940514775977121e-06, "loss": 2.9558, "step": 21134 }, { "epoch": 4.03, "learning_rate": 1.7939942802669207e-06, "loss": 2.2237, "step": 21135 }, { "epoch": 4.03, "learning_rate": 1.7939370829361297e-06, "loss": 2.0363, "step": 21136 }, { "epoch": 4.03, "learning_rate": 1.7938798856053385e-06, "loss": 2.0748, "step": 21137 }, { "epoch": 4.03, "learning_rate": 1.7938226882745471e-06, "loss": 1.7546, "step": 21138 }, { "epoch": 4.03, "learning_rate": 1.7937654909437561e-06, "loss": 1.3707, "step": 21139 }, { "epoch": 4.03, "learning_rate": 1.7937082936129647e-06, "loss": 2.5731, "step": 21140 }, { "epoch": 4.03, "learning_rate": 1.7936510962821735e-06, "loss": 1.911, "step": 21141 }, { "epoch": 4.03, "learning_rate": 1.7935938989513823e-06, "loss": 2.1058, "step": 21142 }, { "epoch": 4.03, "learning_rate": 1.793536701620591e-06, "loss": 2.4827, "step": 21143 }, { "epoch": 4.03, "learning_rate": 1.7934795042898e-06, "loss": 2.1969, "step": 21144 }, { "epoch": 4.03, "learning_rate": 1.7934223069590087e-06, "loss": 2.434, "step": 21145 }, { "epoch": 4.03, "learning_rate": 1.7933651096282172e-06, "loss": 2.3122, "step": 21146 }, { "epoch": 4.03, "learning_rate": 1.7933079122974263e-06, "loss": 1.9797, "step": 21147 }, { "epoch": 4.03, "learning_rate": 1.793250714966635e-06, "loss": 1.9488, "step": 21148 }, { "epoch": 4.03, "learning_rate": 1.7931935176358436e-06, "loss": 3.0983, "step": 21149 }, { "epoch": 4.03, "learning_rate": 1.7931363203050526e-06, "loss": 2.383, "step": 21150 }, { "epoch": 4.03, "learning_rate": 1.7930791229742612e-06, "loss": 3.2506, "step": 21151 }, { "epoch": 4.03, "learning_rate": 1.7930219256434698e-06, "loss": 2.0583, "step": 21152 }, { "epoch": 4.03, "learning_rate": 1.7929647283126788e-06, "loss": 2.739, "step": 21153 }, { "epoch": 4.03, "learning_rate": 1.7929075309818876e-06, "loss": 3.3541, "step": 21154 }, { "epoch": 4.03, "learning_rate": 1.7928503336510962e-06, "loss": 2.4824, "step": 21155 }, { "epoch": 4.03, "learning_rate": 1.7927931363203052e-06, "loss": 1.5722, "step": 21156 }, { "epoch": 4.03, "learning_rate": 1.7927359389895138e-06, "loss": 3.3057, "step": 21157 }, { "epoch": 4.03, "learning_rate": 1.7926787416587226e-06, "loss": 1.4247, "step": 21158 }, { "epoch": 4.03, "learning_rate": 1.7926215443279314e-06, "loss": 2.4307, "step": 21159 }, { "epoch": 4.03, "learning_rate": 1.7925643469971402e-06, "loss": 2.4635, "step": 21160 }, { "epoch": 4.03, "learning_rate": 1.7925071496663492e-06, "loss": 2.5494, "step": 21161 }, { "epoch": 4.03, "learning_rate": 1.7924499523355577e-06, "loss": 1.9474, "step": 21162 }, { "epoch": 4.03, "learning_rate": 1.7923927550047663e-06, "loss": 2.856, "step": 21163 }, { "epoch": 4.03, "learning_rate": 1.7923355576739753e-06, "loss": 2.6679, "step": 21164 }, { "epoch": 4.03, "learning_rate": 1.7922783603431841e-06, "loss": 2.6955, "step": 21165 }, { "epoch": 4.03, "learning_rate": 1.7922211630123927e-06, "loss": 1.4441, "step": 21166 }, { "epoch": 4.03, "learning_rate": 1.7921639656816017e-06, "loss": 1.6215, "step": 21167 }, { "epoch": 4.03, "learning_rate": 1.7921067683508103e-06, "loss": 2.0681, "step": 21168 }, { "epoch": 4.03, "learning_rate": 1.792049571020019e-06, "loss": 2.0303, "step": 21169 }, { "epoch": 4.03, "learning_rate": 1.7919923736892279e-06, "loss": 2.4835, "step": 21170 }, { "epoch": 4.03, "learning_rate": 1.7919351763584367e-06, "loss": 1.3034, "step": 21171 }, { "epoch": 4.03, "learning_rate": 1.7918779790276453e-06, "loss": 1.4428, "step": 21172 }, { "epoch": 4.03, "learning_rate": 1.7918207816968543e-06, "loss": 2.3703, "step": 21173 }, { "epoch": 4.03, "learning_rate": 1.7917635843660628e-06, "loss": 2.221, "step": 21174 }, { "epoch": 4.03, "learning_rate": 1.7917063870352718e-06, "loss": 2.199, "step": 21175 }, { "epoch": 4.03, "learning_rate": 1.7916491897044806e-06, "loss": 1.9231, "step": 21176 }, { "epoch": 4.03, "learning_rate": 1.7915919923736892e-06, "loss": 2.3511, "step": 21177 }, { "epoch": 4.03, "learning_rate": 1.7915347950428982e-06, "loss": 1.0903, "step": 21178 }, { "epoch": 4.03, "learning_rate": 1.7914775977121068e-06, "loss": 2.3491, "step": 21179 }, { "epoch": 4.03, "learning_rate": 1.7914204003813154e-06, "loss": 2.5696, "step": 21180 }, { "epoch": 4.03, "learning_rate": 1.7913632030505244e-06, "loss": 1.7064, "step": 21181 }, { "epoch": 4.03, "learning_rate": 1.7913060057197332e-06, "loss": 2.4353, "step": 21182 }, { "epoch": 4.03, "learning_rate": 1.7912488083889418e-06, "loss": 2.4915, "step": 21183 }, { "epoch": 4.04, "learning_rate": 1.7911916110581508e-06, "loss": 1.9706, "step": 21184 }, { "epoch": 4.04, "learning_rate": 1.7911344137273594e-06, "loss": 3.598, "step": 21185 }, { "epoch": 4.04, "learning_rate": 1.7910772163965682e-06, "loss": 2.5145, "step": 21186 }, { "epoch": 4.04, "learning_rate": 1.7910200190657772e-06, "loss": 1.9265, "step": 21187 }, { "epoch": 4.04, "learning_rate": 1.7909628217349857e-06, "loss": 2.61, "step": 21188 }, { "epoch": 4.04, "learning_rate": 1.7909056244041943e-06, "loss": 1.5787, "step": 21189 }, { "epoch": 4.04, "learning_rate": 1.7908484270734033e-06, "loss": 2.516, "step": 21190 }, { "epoch": 4.04, "learning_rate": 1.790791229742612e-06, "loss": 2.0538, "step": 21191 }, { "epoch": 4.04, "learning_rate": 1.790734032411821e-06, "loss": 2.6387, "step": 21192 }, { "epoch": 4.04, "learning_rate": 1.7906768350810297e-06, "loss": 2.8032, "step": 21193 }, { "epoch": 4.04, "learning_rate": 1.7906196377502383e-06, "loss": 1.9118, "step": 21194 }, { "epoch": 4.04, "learning_rate": 1.7905624404194473e-06, "loss": 1.0433, "step": 21195 }, { "epoch": 4.04, "learning_rate": 1.7905052430886559e-06, "loss": 2.1857, "step": 21196 }, { "epoch": 4.04, "learning_rate": 1.7904480457578647e-06, "loss": 2.4847, "step": 21197 }, { "epoch": 4.04, "learning_rate": 1.7903908484270735e-06, "loss": 2.1928, "step": 21198 }, { "epoch": 4.04, "learning_rate": 1.7903336510962823e-06, "loss": 2.2111, "step": 21199 }, { "epoch": 4.04, "learning_rate": 1.7902764537654908e-06, "loss": 3.2671, "step": 21200 }, { "epoch": 4.04, "learning_rate": 1.7902192564346998e-06, "loss": 1.4573, "step": 21201 }, { "epoch": 4.04, "learning_rate": 1.7901620591039084e-06, "loss": 1.9696, "step": 21202 }, { "epoch": 4.04, "learning_rate": 1.7901048617731172e-06, "loss": 1.648, "step": 21203 }, { "epoch": 4.04, "learning_rate": 1.7900476644423262e-06, "loss": 2.7566, "step": 21204 }, { "epoch": 4.04, "learning_rate": 1.7899904671115348e-06, "loss": 2.7767, "step": 21205 }, { "epoch": 4.04, "learning_rate": 1.7899332697807434e-06, "loss": 2.2123, "step": 21206 }, { "epoch": 4.04, "learning_rate": 1.7898760724499524e-06, "loss": 1.8171, "step": 21207 }, { "epoch": 4.04, "learning_rate": 1.7898188751191612e-06, "loss": 2.1863, "step": 21208 }, { "epoch": 4.04, "learning_rate": 1.78976167778837e-06, "loss": 1.4662, "step": 21209 }, { "epoch": 4.04, "learning_rate": 1.7897044804575788e-06, "loss": 1.3319, "step": 21210 }, { "epoch": 4.04, "learning_rate": 1.7896472831267874e-06, "loss": 3.0951, "step": 21211 }, { "epoch": 4.04, "learning_rate": 1.7895900857959964e-06, "loss": 2.3564, "step": 21212 }, { "epoch": 4.04, "learning_rate": 1.789532888465205e-06, "loss": 1.8525, "step": 21213 }, { "epoch": 4.04, "learning_rate": 1.7894756911344137e-06, "loss": 2.2456, "step": 21214 }, { "epoch": 4.04, "learning_rate": 1.7894184938036228e-06, "loss": 2.9298, "step": 21215 }, { "epoch": 4.04, "learning_rate": 1.7893612964728313e-06, "loss": 1.0926, "step": 21216 }, { "epoch": 4.04, "learning_rate": 1.78930409914204e-06, "loss": 2.7059, "step": 21217 }, { "epoch": 4.04, "learning_rate": 1.789246901811249e-06, "loss": 2.039, "step": 21218 }, { "epoch": 4.04, "learning_rate": 1.7891897044804575e-06, "loss": 2.5632, "step": 21219 }, { "epoch": 4.04, "learning_rate": 1.7891325071496663e-06, "loss": 1.5532, "step": 21220 }, { "epoch": 4.04, "learning_rate": 1.7890753098188753e-06, "loss": 2.5567, "step": 21221 }, { "epoch": 4.04, "learning_rate": 1.7890181124880839e-06, "loss": 2.6763, "step": 21222 }, { "epoch": 4.04, "learning_rate": 1.7889609151572925e-06, "loss": 2.0515, "step": 21223 }, { "epoch": 4.04, "learning_rate": 1.7889037178265015e-06, "loss": 2.2229, "step": 21224 }, { "epoch": 4.04, "learning_rate": 1.7888465204957103e-06, "loss": 2.3837, "step": 21225 }, { "epoch": 4.04, "learning_rate": 1.7887893231649193e-06, "loss": 2.9302, "step": 21226 }, { "epoch": 4.04, "learning_rate": 1.7887321258341279e-06, "loss": 3.3784, "step": 21227 }, { "epoch": 4.04, "learning_rate": 1.7886749285033364e-06, "loss": 2.2114, "step": 21228 }, { "epoch": 4.04, "learning_rate": 1.7886177311725454e-06, "loss": 3.1206, "step": 21229 }, { "epoch": 4.04, "learning_rate": 1.788560533841754e-06, "loss": 2.6992, "step": 21230 }, { "epoch": 4.04, "learning_rate": 1.7885033365109628e-06, "loss": 3.5219, "step": 21231 }, { "epoch": 4.04, "learning_rate": 1.7884461391801718e-06, "loss": 2.4945, "step": 21232 }, { "epoch": 4.04, "learning_rate": 1.7883889418493804e-06, "loss": 2.8215, "step": 21233 }, { "epoch": 4.04, "learning_rate": 1.788331744518589e-06, "loss": 3.019, "step": 21234 }, { "epoch": 4.04, "learning_rate": 1.788274547187798e-06, "loss": 1.1342, "step": 21235 }, { "epoch": 4.04, "learning_rate": 1.7882173498570068e-06, "loss": 2.7692, "step": 21236 }, { "epoch": 4.05, "learning_rate": 1.7881601525262154e-06, "loss": 3.2719, "step": 21237 }, { "epoch": 4.05, "learning_rate": 1.7881029551954244e-06, "loss": 2.1068, "step": 21238 }, { "epoch": 4.05, "learning_rate": 1.788045757864633e-06, "loss": 2.4028, "step": 21239 }, { "epoch": 4.05, "learning_rate": 1.787988560533842e-06, "loss": 2.1983, "step": 21240 }, { "epoch": 4.05, "learning_rate": 1.7879313632030505e-06, "loss": 2.3218, "step": 21241 }, { "epoch": 4.05, "learning_rate": 1.7878741658722593e-06, "loss": 3.2012, "step": 21242 }, { "epoch": 4.05, "learning_rate": 1.7878169685414683e-06, "loss": 1.2996, "step": 21243 }, { "epoch": 4.05, "learning_rate": 1.787759771210677e-06, "loss": 0.9546, "step": 21244 }, { "epoch": 4.05, "learning_rate": 1.7877025738798855e-06, "loss": 1.511, "step": 21245 }, { "epoch": 4.05, "learning_rate": 1.7876453765490945e-06, "loss": 1.7215, "step": 21246 }, { "epoch": 4.05, "learning_rate": 1.787588179218303e-06, "loss": 1.9417, "step": 21247 }, { "epoch": 4.05, "learning_rate": 1.7875309818875119e-06, "loss": 2.5576, "step": 21248 }, { "epoch": 4.05, "learning_rate": 1.787473784556721e-06, "loss": 2.943, "step": 21249 }, { "epoch": 4.05, "learning_rate": 1.7874165872259295e-06, "loss": 2.3963, "step": 21250 }, { "epoch": 4.05, "learning_rate": 1.787359389895138e-06, "loss": 1.8343, "step": 21251 }, { "epoch": 4.05, "learning_rate": 1.787302192564347e-06, "loss": 2.3694, "step": 21252 }, { "epoch": 4.05, "learning_rate": 1.7872449952335559e-06, "loss": 2.2986, "step": 21253 }, { "epoch": 4.05, "learning_rate": 1.7871877979027644e-06, "loss": 2.8096, "step": 21254 }, { "epoch": 4.05, "learning_rate": 1.7871306005719734e-06, "loss": 2.284, "step": 21255 }, { "epoch": 4.05, "learning_rate": 1.787073403241182e-06, "loss": 1.5958, "step": 21256 }, { "epoch": 4.05, "learning_rate": 1.787016205910391e-06, "loss": 3.1275, "step": 21257 }, { "epoch": 4.05, "learning_rate": 1.7869590085795996e-06, "loss": 2.7203, "step": 21258 }, { "epoch": 4.05, "learning_rate": 1.7869018112488084e-06, "loss": 1.7958, "step": 21259 }, { "epoch": 4.05, "learning_rate": 1.7868446139180174e-06, "loss": 2.115, "step": 21260 }, { "epoch": 4.05, "learning_rate": 1.786787416587226e-06, "loss": 3.2433, "step": 21261 }, { "epoch": 4.05, "learning_rate": 1.7867302192564346e-06, "loss": 3.9207, "step": 21262 }, { "epoch": 4.05, "learning_rate": 1.7866730219256436e-06, "loss": 1.32, "step": 21263 }, { "epoch": 4.05, "learning_rate": 1.7866158245948524e-06, "loss": 1.9354, "step": 21264 }, { "epoch": 4.05, "learning_rate": 1.786558627264061e-06, "loss": 1.5288, "step": 21265 }, { "epoch": 4.05, "learning_rate": 1.78650142993327e-06, "loss": 1.467, "step": 21266 }, { "epoch": 4.05, "learning_rate": 1.7864442326024785e-06, "loss": 2.0185, "step": 21267 }, { "epoch": 4.05, "learning_rate": 1.7863870352716871e-06, "loss": 2.9477, "step": 21268 }, { "epoch": 4.05, "learning_rate": 1.7863298379408961e-06, "loss": 2.1494, "step": 21269 }, { "epoch": 4.05, "learning_rate": 1.786272640610105e-06, "loss": 1.7642, "step": 21270 }, { "epoch": 4.05, "learning_rate": 1.7862154432793135e-06, "loss": 1.807, "step": 21271 }, { "epoch": 4.05, "learning_rate": 1.7861582459485225e-06, "loss": 1.6673, "step": 21272 }, { "epoch": 4.05, "learning_rate": 1.786101048617731e-06, "loss": 2.2335, "step": 21273 }, { "epoch": 4.05, "learning_rate": 1.78604385128694e-06, "loss": 2.067, "step": 21274 }, { "epoch": 4.05, "learning_rate": 1.785986653956149e-06, "loss": 1.6697, "step": 21275 }, { "epoch": 4.05, "learning_rate": 1.7859294566253575e-06, "loss": 2.459, "step": 21276 }, { "epoch": 4.05, "learning_rate": 1.7858722592945665e-06, "loss": 2.5403, "step": 21277 }, { "epoch": 4.05, "learning_rate": 1.785815061963775e-06, "loss": 3.0928, "step": 21278 }, { "epoch": 4.05, "learning_rate": 1.7857578646329836e-06, "loss": 2.4547, "step": 21279 }, { "epoch": 4.05, "learning_rate": 1.7857006673021927e-06, "loss": 1.3302, "step": 21280 }, { "epoch": 4.05, "learning_rate": 1.7856434699714014e-06, "loss": 3.2443, "step": 21281 }, { "epoch": 4.05, "learning_rate": 1.78558627264061e-06, "loss": 2.2868, "step": 21282 }, { "epoch": 4.05, "learning_rate": 1.785529075309819e-06, "loss": 3.2844, "step": 21283 }, { "epoch": 4.05, "learning_rate": 1.7854718779790276e-06, "loss": 4.1341, "step": 21284 }, { "epoch": 4.05, "learning_rate": 1.7854146806482364e-06, "loss": 2.4779, "step": 21285 }, { "epoch": 4.05, "learning_rate": 1.7853574833174452e-06, "loss": 2.2248, "step": 21286 }, { "epoch": 4.05, "learning_rate": 1.785300285986654e-06, "loss": 1.6605, "step": 21287 }, { "epoch": 4.05, "learning_rate": 1.7852430886558626e-06, "loss": 3.1572, "step": 21288 }, { "epoch": 4.06, "learning_rate": 1.7851858913250716e-06, "loss": 1.5394, "step": 21289 }, { "epoch": 4.06, "learning_rate": 1.7851286939942802e-06, "loss": 2.0117, "step": 21290 }, { "epoch": 4.06, "learning_rate": 1.7850714966634892e-06, "loss": 1.9151, "step": 21291 }, { "epoch": 4.06, "learning_rate": 1.785014299332698e-06, "loss": 2.521, "step": 21292 }, { "epoch": 4.06, "learning_rate": 1.7849571020019066e-06, "loss": 1.4125, "step": 21293 }, { "epoch": 4.06, "learning_rate": 1.7848999046711156e-06, "loss": 2.9477, "step": 21294 }, { "epoch": 4.06, "learning_rate": 1.7848427073403241e-06, "loss": 2.5353, "step": 21295 }, { "epoch": 4.06, "learning_rate": 1.784785510009533e-06, "loss": 1.5859, "step": 21296 }, { "epoch": 4.06, "learning_rate": 1.7847283126787417e-06, "loss": 1.1861, "step": 21297 }, { "epoch": 4.06, "learning_rate": 1.7846711153479505e-06, "loss": 2.3842, "step": 21298 }, { "epoch": 4.06, "learning_rate": 1.784613918017159e-06, "loss": 2.4474, "step": 21299 }, { "epoch": 4.06, "learning_rate": 1.7845567206863681e-06, "loss": 2.5654, "step": 21300 }, { "epoch": 4.06, "learning_rate": 1.7844995233555767e-06, "loss": 1.4081, "step": 21301 }, { "epoch": 4.06, "learning_rate": 1.7844423260247855e-06, "loss": 1.4055, "step": 21302 }, { "epoch": 4.06, "learning_rate": 1.7843851286939945e-06, "loss": 2.7817, "step": 21303 }, { "epoch": 4.06, "learning_rate": 1.784327931363203e-06, "loss": 2.3586, "step": 21304 }, { "epoch": 4.06, "learning_rate": 1.784270734032412e-06, "loss": 2.1247, "step": 21305 }, { "epoch": 4.06, "learning_rate": 1.7842135367016207e-06, "loss": 2.5684, "step": 21306 }, { "epoch": 4.06, "learning_rate": 1.7841563393708292e-06, "loss": 2.3626, "step": 21307 }, { "epoch": 4.06, "learning_rate": 1.7840991420400382e-06, "loss": 3.0619, "step": 21308 }, { "epoch": 4.06, "learning_rate": 1.784041944709247e-06, "loss": 1.2982, "step": 21309 }, { "epoch": 4.06, "learning_rate": 1.7839847473784556e-06, "loss": 2.8019, "step": 21310 }, { "epoch": 4.06, "learning_rate": 1.7839275500476646e-06, "loss": 2.4936, "step": 21311 }, { "epoch": 4.06, "learning_rate": 1.7838703527168732e-06, "loss": 1.7762, "step": 21312 }, { "epoch": 4.06, "learning_rate": 1.783813155386082e-06, "loss": 1.7521, "step": 21313 }, { "epoch": 4.06, "learning_rate": 1.7837559580552908e-06, "loss": 1.7646, "step": 21314 }, { "epoch": 4.06, "learning_rate": 1.7836987607244996e-06, "loss": 2.4134, "step": 21315 }, { "epoch": 4.06, "learning_rate": 1.7836415633937082e-06, "loss": 1.7316, "step": 21316 }, { "epoch": 4.06, "learning_rate": 1.7835843660629172e-06, "loss": 1.9438, "step": 21317 }, { "epoch": 4.06, "learning_rate": 1.7835271687321258e-06, "loss": 0.7772, "step": 21318 }, { "epoch": 4.06, "learning_rate": 1.7834699714013346e-06, "loss": 2.7065, "step": 21319 }, { "epoch": 4.06, "learning_rate": 1.7834127740705436e-06, "loss": 2.2778, "step": 21320 }, { "epoch": 4.06, "learning_rate": 1.7833555767397521e-06, "loss": 2.2536, "step": 21321 }, { "epoch": 4.06, "learning_rate": 1.7832983794089611e-06, "loss": 1.9428, "step": 21322 }, { "epoch": 4.06, "learning_rate": 1.7832411820781697e-06, "loss": 2.2206, "step": 21323 }, { "epoch": 4.06, "learning_rate": 1.7831839847473785e-06, "loss": 0.936, "step": 21324 }, { "epoch": 4.06, "learning_rate": 1.7831267874165873e-06, "loss": 1.5316, "step": 21325 }, { "epoch": 4.06, "learning_rate": 1.7830695900857961e-06, "loss": 2.5724, "step": 21326 }, { "epoch": 4.06, "learning_rate": 1.7830123927550047e-06, "loss": 1.3799, "step": 21327 }, { "epoch": 4.06, "learning_rate": 1.7829551954242137e-06, "loss": 3.0594, "step": 21328 }, { "epoch": 4.06, "learning_rate": 1.7828979980934223e-06, "loss": 2.5096, "step": 21329 }, { "epoch": 4.06, "learning_rate": 1.782840800762631e-06, "loss": 2.2155, "step": 21330 }, { "epoch": 4.06, "learning_rate": 1.78278360343184e-06, "loss": 2.3543, "step": 21331 }, { "epoch": 4.06, "learning_rate": 1.7827264061010487e-06, "loss": 2.7069, "step": 21332 }, { "epoch": 4.06, "learning_rate": 1.7826692087702572e-06, "loss": 1.7165, "step": 21333 }, { "epoch": 4.06, "learning_rate": 1.7826120114394663e-06, "loss": 2.7363, "step": 21334 }, { "epoch": 4.06, "learning_rate": 1.7825548141086748e-06, "loss": 1.2651, "step": 21335 }, { "epoch": 4.06, "learning_rate": 1.7824976167778836e-06, "loss": 1.4841, "step": 21336 }, { "epoch": 4.06, "learning_rate": 1.7824404194470926e-06, "loss": 1.9718, "step": 21337 }, { "epoch": 4.06, "learning_rate": 1.7823832221163012e-06, "loss": 1.2881, "step": 21338 }, { "epoch": 4.06, "learning_rate": 1.7823260247855102e-06, "loss": 2.3261, "step": 21339 }, { "epoch": 4.06, "learning_rate": 1.7822688274547188e-06, "loss": 2.6698, "step": 21340 }, { "epoch": 4.06, "learning_rate": 1.7822116301239276e-06, "loss": 2.5506, "step": 21341 }, { "epoch": 4.07, "learning_rate": 1.7821544327931366e-06, "loss": 2.3007, "step": 21342 }, { "epoch": 4.07, "learning_rate": 1.7820972354623452e-06, "loss": 2.1055, "step": 21343 }, { "epoch": 4.07, "learning_rate": 1.7820400381315538e-06, "loss": 2.0733, "step": 21344 }, { "epoch": 4.07, "learning_rate": 1.7819828408007628e-06, "loss": 1.7173, "step": 21345 }, { "epoch": 4.07, "learning_rate": 1.7819256434699714e-06, "loss": 3.3193, "step": 21346 }, { "epoch": 4.07, "learning_rate": 1.7818684461391801e-06, "loss": 2.3428, "step": 21347 }, { "epoch": 4.07, "learning_rate": 1.7818112488083892e-06, "loss": 2.3921, "step": 21348 }, { "epoch": 4.07, "learning_rate": 1.7817540514775977e-06, "loss": 1.5911, "step": 21349 }, { "epoch": 4.07, "learning_rate": 1.7816968541468063e-06, "loss": 1.6664, "step": 21350 }, { "epoch": 4.07, "learning_rate": 1.7816396568160153e-06, "loss": 2.6363, "step": 21351 }, { "epoch": 4.07, "learning_rate": 1.7815824594852241e-06, "loss": 2.4295, "step": 21352 }, { "epoch": 4.07, "learning_rate": 1.7815252621544327e-06, "loss": 2.5779, "step": 21353 }, { "epoch": 4.07, "learning_rate": 1.7814680648236417e-06, "loss": 1.697, "step": 21354 }, { "epoch": 4.07, "learning_rate": 1.7814108674928503e-06, "loss": 2.11, "step": 21355 }, { "epoch": 4.07, "learning_rate": 1.7813536701620593e-06, "loss": 1.3979, "step": 21356 }, { "epoch": 4.07, "learning_rate": 1.7812964728312679e-06, "loss": 2.1462, "step": 21357 }, { "epoch": 4.07, "learning_rate": 1.7812392755004767e-06, "loss": 2.9402, "step": 21358 }, { "epoch": 4.07, "learning_rate": 1.7811820781696857e-06, "loss": 3.2193, "step": 21359 }, { "epoch": 4.07, "learning_rate": 1.7811248808388943e-06, "loss": 2.1048, "step": 21360 }, { "epoch": 4.07, "learning_rate": 1.7810676835081028e-06, "loss": 2.2465, "step": 21361 }, { "epoch": 4.07, "learning_rate": 1.7810104861773118e-06, "loss": 1.7422, "step": 21362 }, { "epoch": 4.07, "learning_rate": 1.7809532888465206e-06, "loss": 3.3067, "step": 21363 }, { "epoch": 4.07, "learning_rate": 1.7808960915157292e-06, "loss": 2.0366, "step": 21364 }, { "epoch": 4.07, "learning_rate": 1.7808388941849382e-06, "loss": 2.8504, "step": 21365 }, { "epoch": 4.07, "learning_rate": 1.7807816968541468e-06, "loss": 3.2818, "step": 21366 }, { "epoch": 4.07, "learning_rate": 1.7807244995233554e-06, "loss": 2.6129, "step": 21367 }, { "epoch": 4.07, "learning_rate": 1.7806673021925644e-06, "loss": 1.9503, "step": 21368 }, { "epoch": 4.07, "learning_rate": 1.7806101048617732e-06, "loss": 2.8682, "step": 21369 }, { "epoch": 4.07, "learning_rate": 1.7805529075309818e-06, "loss": 1.6633, "step": 21370 }, { "epoch": 4.07, "learning_rate": 1.7804957102001908e-06, "loss": 2.4455, "step": 21371 }, { "epoch": 4.07, "learning_rate": 1.7804385128693994e-06, "loss": 2.0147, "step": 21372 }, { "epoch": 4.07, "learning_rate": 1.7803813155386084e-06, "loss": 2.6682, "step": 21373 }, { "epoch": 4.07, "learning_rate": 1.780324118207817e-06, "loss": 2.191, "step": 21374 }, { "epoch": 4.07, "learning_rate": 1.7802669208770257e-06, "loss": 2.8683, "step": 21375 }, { "epoch": 4.07, "learning_rate": 1.7802097235462347e-06, "loss": 1.4921, "step": 21376 }, { "epoch": 4.07, "learning_rate": 1.7801525262154433e-06, "loss": 1.4316, "step": 21377 }, { "epoch": 4.07, "learning_rate": 1.780095328884652e-06, "loss": 3.1264, "step": 21378 }, { "epoch": 4.07, "learning_rate": 1.780038131553861e-06, "loss": 1.8248, "step": 21379 }, { "epoch": 4.07, "learning_rate": 1.7799809342230697e-06, "loss": 2.1782, "step": 21380 }, { "epoch": 4.07, "learning_rate": 1.7799237368922783e-06, "loss": 3.7151, "step": 21381 }, { "epoch": 4.07, "learning_rate": 1.7798665395614873e-06, "loss": 1.4137, "step": 21382 }, { "epoch": 4.07, "learning_rate": 1.7798093422306959e-06, "loss": 3.5443, "step": 21383 }, { "epoch": 4.07, "learning_rate": 1.7797521448999045e-06, "loss": 1.08, "step": 21384 }, { "epoch": 4.07, "learning_rate": 1.7796949475691135e-06, "loss": 1.7033, "step": 21385 }, { "epoch": 4.07, "learning_rate": 1.7796377502383223e-06, "loss": 1.1483, "step": 21386 }, { "epoch": 4.07, "learning_rate": 1.7795805529075313e-06, "loss": 3.379, "step": 21387 }, { "epoch": 4.07, "learning_rate": 1.7795233555767398e-06, "loss": 1.8005, "step": 21388 }, { "epoch": 4.07, "learning_rate": 1.7794661582459484e-06, "loss": 2.5455, "step": 21389 }, { "epoch": 4.07, "learning_rate": 1.7794089609151574e-06, "loss": 1.7684, "step": 21390 }, { "epoch": 4.07, "learning_rate": 1.7793517635843662e-06, "loss": 1.7639, "step": 21391 }, { "epoch": 4.07, "learning_rate": 1.7792945662535748e-06, "loss": 3.1467, "step": 21392 }, { "epoch": 4.07, "learning_rate": 1.7792373689227838e-06, "loss": 1.6636, "step": 21393 }, { "epoch": 4.08, "learning_rate": 1.7791801715919924e-06, "loss": 2.8929, "step": 21394 }, { "epoch": 4.08, "learning_rate": 1.779122974261201e-06, "loss": 2.6829, "step": 21395 }, { "epoch": 4.08, "learning_rate": 1.77906577693041e-06, "loss": 1.2288, "step": 21396 }, { "epoch": 4.08, "learning_rate": 1.7790085795996188e-06, "loss": 2.5669, "step": 21397 }, { "epoch": 4.08, "learning_rate": 1.7789513822688274e-06, "loss": 2.9918, "step": 21398 }, { "epoch": 4.08, "learning_rate": 1.7788941849380364e-06, "loss": 3.1201, "step": 21399 }, { "epoch": 4.08, "learning_rate": 1.778836987607245e-06, "loss": 1.2067, "step": 21400 }, { "epoch": 4.08, "learning_rate": 1.7787797902764537e-06, "loss": 2.27, "step": 21401 }, { "epoch": 4.08, "learning_rate": 1.7787225929456625e-06, "loss": 2.386, "step": 21402 }, { "epoch": 4.08, "learning_rate": 1.7786653956148713e-06, "loss": 2.6809, "step": 21403 }, { "epoch": 4.08, "learning_rate": 1.7786081982840803e-06, "loss": 2.9926, "step": 21404 }, { "epoch": 4.08, "learning_rate": 1.778551000953289e-06, "loss": 2.4428, "step": 21405 }, { "epoch": 4.08, "learning_rate": 1.7784938036224975e-06, "loss": 2.53, "step": 21406 }, { "epoch": 4.08, "learning_rate": 1.7784366062917065e-06, "loss": 1.4838, "step": 21407 }, { "epoch": 4.08, "learning_rate": 1.7783794089609153e-06, "loss": 1.5203, "step": 21408 }, { "epoch": 4.08, "learning_rate": 1.7783222116301239e-06, "loss": 2.2663, "step": 21409 }, { "epoch": 4.08, "learning_rate": 1.7782650142993329e-06, "loss": 1.6178, "step": 21410 }, { "epoch": 4.08, "learning_rate": 1.7782078169685415e-06, "loss": 2.628, "step": 21411 }, { "epoch": 4.08, "learning_rate": 1.7781506196377503e-06, "loss": 1.4737, "step": 21412 }, { "epoch": 4.08, "learning_rate": 1.778093422306959e-06, "loss": 1.9812, "step": 21413 }, { "epoch": 4.08, "learning_rate": 1.7780362249761679e-06, "loss": 2.3328, "step": 21414 }, { "epoch": 4.08, "learning_rate": 1.7779790276453764e-06, "loss": 1.941, "step": 21415 }, { "epoch": 4.08, "learning_rate": 1.7779218303145854e-06, "loss": 2.1581, "step": 21416 }, { "epoch": 4.08, "learning_rate": 1.777864632983794e-06, "loss": 1.9235, "step": 21417 }, { "epoch": 4.08, "learning_rate": 1.7778074356530028e-06, "loss": 2.1795, "step": 21418 }, { "epoch": 4.08, "learning_rate": 1.7777502383222118e-06, "loss": 2.5684, "step": 21419 }, { "epoch": 4.08, "learning_rate": 1.7776930409914204e-06, "loss": 1.6776, "step": 21420 }, { "epoch": 4.08, "learning_rate": 1.7776358436606294e-06, "loss": 1.932, "step": 21421 }, { "epoch": 4.08, "learning_rate": 1.777578646329838e-06, "loss": 1.8915, "step": 21422 }, { "epoch": 4.08, "learning_rate": 1.7775214489990466e-06, "loss": 1.1814, "step": 21423 }, { "epoch": 4.08, "learning_rate": 1.7774642516682556e-06, "loss": 2.6604, "step": 21424 }, { "epoch": 4.08, "learning_rate": 1.7774070543374644e-06, "loss": 2.194, "step": 21425 }, { "epoch": 4.08, "learning_rate": 1.777349857006673e-06, "loss": 2.6151, "step": 21426 }, { "epoch": 4.08, "learning_rate": 1.777292659675882e-06, "loss": 1.8031, "step": 21427 }, { "epoch": 4.08, "learning_rate": 1.7772354623450905e-06, "loss": 2.1416, "step": 21428 }, { "epoch": 4.08, "learning_rate": 1.7771782650142993e-06, "loss": 1.9192, "step": 21429 }, { "epoch": 4.08, "learning_rate": 1.7771210676835083e-06, "loss": 1.4886, "step": 21430 }, { "epoch": 4.08, "learning_rate": 1.777063870352717e-06, "loss": 2.2425, "step": 21431 }, { "epoch": 4.08, "learning_rate": 1.7770066730219255e-06, "loss": 2.2585, "step": 21432 }, { "epoch": 4.08, "learning_rate": 1.7769494756911345e-06, "loss": 2.6287, "step": 21433 }, { "epoch": 4.08, "learning_rate": 1.776892278360343e-06, "loss": 3.2325, "step": 21434 }, { "epoch": 4.08, "learning_rate": 1.7768350810295519e-06, "loss": 2.0454, "step": 21435 }, { "epoch": 4.08, "learning_rate": 1.7767778836987609e-06, "loss": 3.3158, "step": 21436 }, { "epoch": 4.08, "learning_rate": 1.7767206863679695e-06, "loss": 1.412, "step": 21437 }, { "epoch": 4.08, "learning_rate": 1.7766634890371785e-06, "loss": 2.6062, "step": 21438 }, { "epoch": 4.08, "learning_rate": 1.776606291706387e-06, "loss": 2.4389, "step": 21439 }, { "epoch": 4.08, "learning_rate": 1.7765490943755959e-06, "loss": 2.8854, "step": 21440 }, { "epoch": 4.08, "learning_rate": 1.7764918970448046e-06, "loss": 1.2259, "step": 21441 }, { "epoch": 4.08, "learning_rate": 1.7764346997140134e-06, "loss": 3.8088, "step": 21442 }, { "epoch": 4.08, "learning_rate": 1.776377502383222e-06, "loss": 2.7419, "step": 21443 }, { "epoch": 4.08, "learning_rate": 1.776320305052431e-06, "loss": 2.3138, "step": 21444 }, { "epoch": 4.08, "learning_rate": 1.7762631077216396e-06, "loss": 2.8368, "step": 21445 }, { "epoch": 4.08, "learning_rate": 1.7762059103908484e-06, "loss": 2.4387, "step": 21446 }, { "epoch": 4.09, "learning_rate": 1.7761487130600574e-06, "loss": 1.7013, "step": 21447 }, { "epoch": 4.09, "learning_rate": 1.776091515729266e-06, "loss": 2.7074, "step": 21448 }, { "epoch": 4.09, "learning_rate": 1.7760343183984746e-06, "loss": 1.9975, "step": 21449 }, { "epoch": 4.09, "learning_rate": 1.7759771210676836e-06, "loss": 2.6809, "step": 21450 }, { "epoch": 4.09, "learning_rate": 1.7759199237368924e-06, "loss": 2.1314, "step": 21451 }, { "epoch": 4.09, "learning_rate": 1.7758627264061012e-06, "loss": 1.4845, "step": 21452 }, { "epoch": 4.09, "learning_rate": 1.77580552907531e-06, "loss": 1.5014, "step": 21453 }, { "epoch": 4.09, "learning_rate": 1.7757483317445185e-06, "loss": 1.4723, "step": 21454 }, { "epoch": 4.09, "learning_rate": 1.7756911344137276e-06, "loss": 2.5383, "step": 21455 }, { "epoch": 4.09, "learning_rate": 1.7756339370829361e-06, "loss": 1.4811, "step": 21456 }, { "epoch": 4.09, "learning_rate": 1.775576739752145e-06, "loss": 2.4697, "step": 21457 }, { "epoch": 4.09, "learning_rate": 1.775519542421354e-06, "loss": 2.265, "step": 21458 }, { "epoch": 4.09, "learning_rate": 1.7754623450905625e-06, "loss": 0.9732, "step": 21459 }, { "epoch": 4.09, "learning_rate": 1.775405147759771e-06, "loss": 1.6608, "step": 21460 }, { "epoch": 4.09, "learning_rate": 1.77534795042898e-06, "loss": 2.8853, "step": 21461 }, { "epoch": 4.09, "learning_rate": 1.7752907530981887e-06, "loss": 3.2768, "step": 21462 }, { "epoch": 4.09, "learning_rate": 1.7752335557673975e-06, "loss": 2.846, "step": 21463 }, { "epoch": 4.09, "learning_rate": 1.7751763584366065e-06, "loss": 2.4011, "step": 21464 }, { "epoch": 4.09, "learning_rate": 1.775119161105815e-06, "loss": 3.1119, "step": 21465 }, { "epoch": 4.09, "learning_rate": 1.7750619637750236e-06, "loss": 1.6218, "step": 21466 }, { "epoch": 4.09, "learning_rate": 1.7750047664442327e-06, "loss": 2.3091, "step": 21467 }, { "epoch": 4.09, "learning_rate": 1.7749475691134414e-06, "loss": 1.568, "step": 21468 }, { "epoch": 4.09, "learning_rate": 1.7748903717826502e-06, "loss": 2.6117, "step": 21469 }, { "epoch": 4.09, "learning_rate": 1.774833174451859e-06, "loss": 3.1675, "step": 21470 }, { "epoch": 4.09, "learning_rate": 1.7747759771210676e-06, "loss": 1.9288, "step": 21471 }, { "epoch": 4.09, "learning_rate": 1.7747187797902766e-06, "loss": 2.7151, "step": 21472 }, { "epoch": 4.09, "learning_rate": 1.7746615824594852e-06, "loss": 2.1722, "step": 21473 }, { "epoch": 4.09, "learning_rate": 1.774604385128694e-06, "loss": 2.6294, "step": 21474 }, { "epoch": 4.09, "learning_rate": 1.774547187797903e-06, "loss": 1.7279, "step": 21475 }, { "epoch": 4.09, "learning_rate": 1.7744899904671116e-06, "loss": 2.277, "step": 21476 }, { "epoch": 4.09, "learning_rate": 1.7744327931363202e-06, "loss": 1.7311, "step": 21477 }, { "epoch": 4.09, "learning_rate": 1.7743755958055292e-06, "loss": 3.1467, "step": 21478 }, { "epoch": 4.09, "learning_rate": 1.774318398474738e-06, "loss": 1.099, "step": 21479 }, { "epoch": 4.09, "learning_rate": 1.7742612011439465e-06, "loss": 1.4091, "step": 21480 }, { "epoch": 4.09, "learning_rate": 1.7742040038131556e-06, "loss": 1.6141, "step": 21481 }, { "epoch": 4.09, "learning_rate": 1.7741468064823641e-06, "loss": 2.9966, "step": 21482 }, { "epoch": 4.09, "learning_rate": 1.7740896091515727e-06, "loss": 2.1703, "step": 21483 }, { "epoch": 4.09, "learning_rate": 1.7740324118207817e-06, "loss": 2.3539, "step": 21484 }, { "epoch": 4.09, "learning_rate": 1.7739752144899905e-06, "loss": 2.8069, "step": 21485 }, { "epoch": 4.09, "learning_rate": 1.7739180171591995e-06, "loss": 1.1379, "step": 21486 }, { "epoch": 4.09, "learning_rate": 1.773860819828408e-06, "loss": 2.0301, "step": 21487 }, { "epoch": 4.09, "learning_rate": 1.7738036224976167e-06, "loss": 3.627, "step": 21488 }, { "epoch": 4.09, "learning_rate": 1.7737464251668257e-06, "loss": 1.1753, "step": 21489 }, { "epoch": 4.09, "learning_rate": 1.7736892278360343e-06, "loss": 1.6968, "step": 21490 }, { "epoch": 4.09, "learning_rate": 1.773632030505243e-06, "loss": 2.0367, "step": 21491 }, { "epoch": 4.09, "learning_rate": 1.773574833174452e-06, "loss": 1.6763, "step": 21492 }, { "epoch": 4.09, "learning_rate": 1.7735176358436607e-06, "loss": 1.8986, "step": 21493 }, { "epoch": 4.09, "learning_rate": 1.7734604385128692e-06, "loss": 2.897, "step": 21494 }, { "epoch": 4.09, "learning_rate": 1.7734032411820782e-06, "loss": 2.0887, "step": 21495 }, { "epoch": 4.09, "learning_rate": 1.773346043851287e-06, "loss": 1.5855, "step": 21496 }, { "epoch": 4.09, "learning_rate": 1.7732888465204956e-06, "loss": 1.8451, "step": 21497 }, { "epoch": 4.09, "learning_rate": 1.7732316491897046e-06, "loss": 2.8441, "step": 21498 }, { "epoch": 4.1, "learning_rate": 1.7731744518589132e-06, "loss": 2.1716, "step": 21499 }, { "epoch": 4.1, "learning_rate": 1.773117254528122e-06, "loss": 2.7588, "step": 21500 }, { "epoch": 4.1, "learning_rate": 1.7730600571973308e-06, "loss": 0.5536, "step": 21501 }, { "epoch": 4.1, "learning_rate": 1.7730028598665396e-06, "loss": 1.8142, "step": 21502 }, { "epoch": 4.1, "learning_rate": 1.7729456625357486e-06, "loss": 2.3663, "step": 21503 }, { "epoch": 4.1, "learning_rate": 1.7728884652049572e-06, "loss": 2.7527, "step": 21504 }, { "epoch": 4.1, "learning_rate": 1.7728312678741658e-06, "loss": 2.8036, "step": 21505 }, { "epoch": 4.1, "learning_rate": 1.7727740705433748e-06, "loss": 3.0167, "step": 21506 }, { "epoch": 4.1, "learning_rate": 1.7727168732125836e-06, "loss": 1.9096, "step": 21507 }, { "epoch": 4.1, "learning_rate": 1.7726596758817921e-06, "loss": 1.9422, "step": 21508 }, { "epoch": 4.1, "learning_rate": 1.7726024785510011e-06, "loss": 2.1216, "step": 21509 }, { "epoch": 4.1, "learning_rate": 1.7725452812202097e-06, "loss": 2.3306, "step": 21510 }, { "epoch": 4.1, "learning_rate": 1.7724880838894183e-06, "loss": 2.4656, "step": 21511 }, { "epoch": 4.1, "learning_rate": 1.7724308865586273e-06, "loss": 2.6696, "step": 21512 }, { "epoch": 4.1, "learning_rate": 1.7723736892278361e-06, "loss": 2.3867, "step": 21513 }, { "epoch": 4.1, "learning_rate": 1.7723164918970447e-06, "loss": 1.9607, "step": 21514 }, { "epoch": 4.1, "learning_rate": 1.7722592945662537e-06, "loss": 1.6156, "step": 21515 }, { "epoch": 4.1, "learning_rate": 1.7722020972354623e-06, "loss": 2.5336, "step": 21516 }, { "epoch": 4.1, "learning_rate": 1.7721448999046713e-06, "loss": 2.5122, "step": 21517 }, { "epoch": 4.1, "learning_rate": 1.77208770257388e-06, "loss": 3.125, "step": 21518 }, { "epoch": 4.1, "learning_rate": 1.7720305052430887e-06, "loss": 2.7998, "step": 21519 }, { "epoch": 4.1, "learning_rate": 1.7719733079122977e-06, "loss": 2.7537, "step": 21520 }, { "epoch": 4.1, "learning_rate": 1.7719161105815062e-06, "loss": 1.5069, "step": 21521 }, { "epoch": 4.1, "learning_rate": 1.7718589132507148e-06, "loss": 3.2396, "step": 21522 }, { "epoch": 4.1, "learning_rate": 1.7718017159199238e-06, "loss": 1.0569, "step": 21523 }, { "epoch": 4.1, "learning_rate": 1.7717445185891326e-06, "loss": 2.6641, "step": 21524 }, { "epoch": 4.1, "learning_rate": 1.7716873212583412e-06, "loss": 1.5712, "step": 21525 }, { "epoch": 4.1, "learning_rate": 1.7716301239275502e-06, "loss": 3.4511, "step": 21526 }, { "epoch": 4.1, "learning_rate": 1.7715729265967588e-06, "loss": 2.4403, "step": 21527 }, { "epoch": 4.1, "learning_rate": 1.7715157292659676e-06, "loss": 1.6347, "step": 21528 }, { "epoch": 4.1, "learning_rate": 1.7714585319351764e-06, "loss": 2.8792, "step": 21529 }, { "epoch": 4.1, "learning_rate": 1.7714013346043852e-06, "loss": 1.2134, "step": 21530 }, { "epoch": 4.1, "learning_rate": 1.7713441372735938e-06, "loss": 1.6925, "step": 21531 }, { "epoch": 4.1, "learning_rate": 1.7712869399428028e-06, "loss": 1.7837, "step": 21532 }, { "epoch": 4.1, "learning_rate": 1.7712297426120114e-06, "loss": 1.7869, "step": 21533 }, { "epoch": 4.1, "learning_rate": 1.7711725452812204e-06, "loss": 2.2948, "step": 21534 }, { "epoch": 4.1, "learning_rate": 1.7711153479504291e-06, "loss": 1.3361, "step": 21535 }, { "epoch": 4.1, "learning_rate": 1.7710581506196377e-06, "loss": 2.5134, "step": 21536 }, { "epoch": 4.1, "learning_rate": 1.7710009532888467e-06, "loss": 2.0287, "step": 21537 }, { "epoch": 4.1, "learning_rate": 1.7709437559580553e-06, "loss": 1.7649, "step": 21538 }, { "epoch": 4.1, "learning_rate": 1.770886558627264e-06, "loss": 3.0137, "step": 21539 }, { "epoch": 4.1, "learning_rate": 1.770829361296473e-06, "loss": 1.2179, "step": 21540 }, { "epoch": 4.1, "learning_rate": 1.7707721639656817e-06, "loss": 2.5102, "step": 21541 }, { "epoch": 4.1, "learning_rate": 1.7707149666348903e-06, "loss": 2.7237, "step": 21542 }, { "epoch": 4.1, "learning_rate": 1.7706577693040993e-06, "loss": 1.6771, "step": 21543 }, { "epoch": 4.1, "learning_rate": 1.7706005719733079e-06, "loss": 2.8, "step": 21544 }, { "epoch": 4.1, "learning_rate": 1.7705433746425167e-06, "loss": 1.7609, "step": 21545 }, { "epoch": 4.1, "learning_rate": 1.7704861773117257e-06, "loss": 2.9227, "step": 21546 }, { "epoch": 4.1, "learning_rate": 1.7704289799809343e-06, "loss": 2.6923, "step": 21547 }, { "epoch": 4.1, "learning_rate": 1.7703717826501428e-06, "loss": 2.1023, "step": 21548 }, { "epoch": 4.1, "learning_rate": 1.7703145853193518e-06, "loss": 1.9688, "step": 21549 }, { "epoch": 4.1, "learning_rate": 1.7702573879885604e-06, "loss": 3.3416, "step": 21550 }, { "epoch": 4.1, "learning_rate": 1.7702001906577694e-06, "loss": 2.0292, "step": 21551 }, { "epoch": 4.11, "learning_rate": 1.7701429933269782e-06, "loss": 2.5241, "step": 21552 }, { "epoch": 4.11, "learning_rate": 1.7700857959961868e-06, "loss": 2.6871, "step": 21553 }, { "epoch": 4.11, "learning_rate": 1.7700285986653958e-06, "loss": 2.2774, "step": 21554 }, { "epoch": 4.11, "learning_rate": 1.7699714013346044e-06, "loss": 3.1015, "step": 21555 }, { "epoch": 4.11, "learning_rate": 1.7699142040038132e-06, "loss": 1.8521, "step": 21556 }, { "epoch": 4.11, "learning_rate": 1.769857006673022e-06, "loss": 1.4256, "step": 21557 }, { "epoch": 4.11, "learning_rate": 1.7697998093422308e-06, "loss": 2.8039, "step": 21558 }, { "epoch": 4.11, "learning_rate": 1.7697426120114394e-06, "loss": 3.1947, "step": 21559 }, { "epoch": 4.11, "learning_rate": 1.7696854146806484e-06, "loss": 2.2643, "step": 21560 }, { "epoch": 4.11, "learning_rate": 1.769628217349857e-06, "loss": 2.1166, "step": 21561 }, { "epoch": 4.11, "learning_rate": 1.7695710200190657e-06, "loss": 2.1356, "step": 21562 }, { "epoch": 4.11, "learning_rate": 1.7695138226882747e-06, "loss": 3.0394, "step": 21563 }, { "epoch": 4.11, "learning_rate": 1.7694566253574833e-06, "loss": 3.1658, "step": 21564 }, { "epoch": 4.11, "learning_rate": 1.769399428026692e-06, "loss": 2.6575, "step": 21565 }, { "epoch": 4.11, "learning_rate": 1.769342230695901e-06, "loss": 1.2735, "step": 21566 }, { "epoch": 4.11, "learning_rate": 1.7692850333651097e-06, "loss": 1.8106, "step": 21567 }, { "epoch": 4.11, "learning_rate": 1.7692278360343185e-06, "loss": 2.7872, "step": 21568 }, { "epoch": 4.11, "learning_rate": 1.7691706387035273e-06, "loss": 2.295, "step": 21569 }, { "epoch": 4.11, "learning_rate": 1.7691134413727359e-06, "loss": 2.7212, "step": 21570 }, { "epoch": 4.11, "learning_rate": 1.7690562440419449e-06, "loss": 1.8683, "step": 21571 }, { "epoch": 4.11, "learning_rate": 1.7689990467111535e-06, "loss": 2.578, "step": 21572 }, { "epoch": 4.11, "learning_rate": 1.7689418493803623e-06, "loss": 2.1392, "step": 21573 }, { "epoch": 4.11, "learning_rate": 1.7688846520495713e-06, "loss": 1.4733, "step": 21574 }, { "epoch": 4.11, "learning_rate": 1.7688274547187798e-06, "loss": 2.0184, "step": 21575 }, { "epoch": 4.11, "learning_rate": 1.7687702573879884e-06, "loss": 2.3985, "step": 21576 }, { "epoch": 4.11, "learning_rate": 1.7687130600571974e-06, "loss": 2.4843, "step": 21577 }, { "epoch": 4.11, "learning_rate": 1.768655862726406e-06, "loss": 2.0954, "step": 21578 }, { "epoch": 4.11, "learning_rate": 1.7685986653956148e-06, "loss": 2.006, "step": 21579 }, { "epoch": 4.11, "learning_rate": 1.7685414680648238e-06, "loss": 2.6211, "step": 21580 }, { "epoch": 4.11, "learning_rate": 1.7684842707340324e-06, "loss": 2.3332, "step": 21581 }, { "epoch": 4.11, "learning_rate": 1.7684270734032414e-06, "loss": 1.7635, "step": 21582 }, { "epoch": 4.11, "learning_rate": 1.76836987607245e-06, "loss": 2.5895, "step": 21583 }, { "epoch": 4.11, "learning_rate": 1.7683126787416588e-06, "loss": 2.0002, "step": 21584 }, { "epoch": 4.11, "learning_rate": 1.7682554814108678e-06, "loss": 1.6916, "step": 21585 }, { "epoch": 4.11, "learning_rate": 1.7681982840800764e-06, "loss": 2.0199, "step": 21586 }, { "epoch": 4.11, "learning_rate": 1.768141086749285e-06, "loss": 1.6381, "step": 21587 }, { "epoch": 4.11, "learning_rate": 1.768083889418494e-06, "loss": 2.3169, "step": 21588 }, { "epoch": 4.11, "learning_rate": 1.7680266920877025e-06, "loss": 3.1901, "step": 21589 }, { "epoch": 4.11, "learning_rate": 1.7679694947569113e-06, "loss": 2.4756, "step": 21590 }, { "epoch": 4.11, "learning_rate": 1.7679122974261203e-06, "loss": 3.4344, "step": 21591 }, { "epoch": 4.11, "learning_rate": 1.767855100095329e-06, "loss": 2.6508, "step": 21592 }, { "epoch": 4.11, "learning_rate": 1.7677979027645375e-06, "loss": 2.6881, "step": 21593 }, { "epoch": 4.11, "learning_rate": 1.7677407054337465e-06, "loss": 2.836, "step": 21594 }, { "epoch": 4.11, "learning_rate": 1.7676835081029553e-06, "loss": 2.0065, "step": 21595 }, { "epoch": 4.11, "learning_rate": 1.7676263107721639e-06, "loss": 2.9616, "step": 21596 }, { "epoch": 4.11, "learning_rate": 1.7675691134413729e-06, "loss": 1.7165, "step": 21597 }, { "epoch": 4.11, "learning_rate": 1.7675119161105815e-06, "loss": 2.4497, "step": 21598 }, { "epoch": 4.11, "learning_rate": 1.7674547187797905e-06, "loss": 1.6606, "step": 21599 }, { "epoch": 4.11, "learning_rate": 1.767397521448999e-06, "loss": 2.6237, "step": 21600 }, { "epoch": 4.11, "learning_rate": 1.7673403241182078e-06, "loss": 3.0937, "step": 21601 }, { "epoch": 4.11, "learning_rate": 1.7672831267874169e-06, "loss": 1.6203, "step": 21602 }, { "epoch": 4.11, "learning_rate": 1.7672259294566254e-06, "loss": 2.6481, "step": 21603 }, { "epoch": 4.12, "learning_rate": 1.767168732125834e-06, "loss": 1.1049, "step": 21604 }, { "epoch": 4.12, "learning_rate": 1.767111534795043e-06, "loss": 2.376, "step": 21605 }, { "epoch": 4.12, "learning_rate": 1.7670543374642518e-06, "loss": 2.3915, "step": 21606 }, { "epoch": 4.12, "learning_rate": 1.7669971401334604e-06, "loss": 2.2201, "step": 21607 }, { "epoch": 4.12, "learning_rate": 1.7669399428026694e-06, "loss": 2.1166, "step": 21608 }, { "epoch": 4.12, "learning_rate": 1.766882745471878e-06, "loss": 2.8927, "step": 21609 }, { "epoch": 4.12, "learning_rate": 1.7668255481410866e-06, "loss": 1.4399, "step": 21610 }, { "epoch": 4.12, "learning_rate": 1.7667683508102956e-06, "loss": 1.4015, "step": 21611 }, { "epoch": 4.12, "learning_rate": 1.7667111534795044e-06, "loss": 2.7205, "step": 21612 }, { "epoch": 4.12, "learning_rate": 1.766653956148713e-06, "loss": 2.6068, "step": 21613 }, { "epoch": 4.12, "learning_rate": 1.766596758817922e-06, "loss": 2.7528, "step": 21614 }, { "epoch": 4.12, "learning_rate": 1.7665395614871305e-06, "loss": 2.6911, "step": 21615 }, { "epoch": 4.12, "learning_rate": 1.7664823641563395e-06, "loss": 1.7157, "step": 21616 }, { "epoch": 4.12, "learning_rate": 1.7664251668255481e-06, "loss": 2.8197, "step": 21617 }, { "epoch": 4.12, "learning_rate": 1.766367969494757e-06, "loss": 1.3921, "step": 21618 }, { "epoch": 4.12, "learning_rate": 1.766310772163966e-06, "loss": 1.9346, "step": 21619 }, { "epoch": 4.12, "learning_rate": 1.7662535748331745e-06, "loss": 1.8007, "step": 21620 }, { "epoch": 4.12, "learning_rate": 1.766196377502383e-06, "loss": 2.9999, "step": 21621 }, { "epoch": 4.12, "learning_rate": 1.766139180171592e-06, "loss": 1.6972, "step": 21622 }, { "epoch": 4.12, "learning_rate": 1.7660819828408009e-06, "loss": 2.2825, "step": 21623 }, { "epoch": 4.12, "learning_rate": 1.7660247855100095e-06, "loss": 2.9869, "step": 21624 }, { "epoch": 4.12, "learning_rate": 1.7659675881792185e-06, "loss": 0.973, "step": 21625 }, { "epoch": 4.12, "learning_rate": 1.765910390848427e-06, "loss": 1.9168, "step": 21626 }, { "epoch": 4.12, "learning_rate": 1.7658531935176356e-06, "loss": 0.898, "step": 21627 }, { "epoch": 4.12, "learning_rate": 1.7657959961868446e-06, "loss": 2.7821, "step": 21628 }, { "epoch": 4.12, "learning_rate": 1.7657387988560534e-06, "loss": 2.4911, "step": 21629 }, { "epoch": 4.12, "learning_rate": 1.765681601525262e-06, "loss": 3.3594, "step": 21630 }, { "epoch": 4.12, "learning_rate": 1.765624404194471e-06, "loss": 1.6296, "step": 21631 }, { "epoch": 4.12, "learning_rate": 1.7655672068636796e-06, "loss": 2.3649, "step": 21632 }, { "epoch": 4.12, "learning_rate": 1.7655100095328886e-06, "loss": 2.2465, "step": 21633 }, { "epoch": 4.12, "learning_rate": 1.7654528122020974e-06, "loss": 2.8009, "step": 21634 }, { "epoch": 4.12, "learning_rate": 1.765395614871306e-06, "loss": 1.5494, "step": 21635 }, { "epoch": 4.12, "learning_rate": 1.765338417540515e-06, "loss": 1.8062, "step": 21636 }, { "epoch": 4.12, "learning_rate": 1.7652812202097236e-06, "loss": 2.5323, "step": 21637 }, { "epoch": 4.12, "learning_rate": 1.7652240228789322e-06, "loss": 1.6943, "step": 21638 }, { "epoch": 4.12, "learning_rate": 1.7651668255481412e-06, "loss": 2.34, "step": 21639 }, { "epoch": 4.12, "learning_rate": 1.76510962821735e-06, "loss": 2.1997, "step": 21640 }, { "epoch": 4.12, "learning_rate": 1.7650524308865585e-06, "loss": 2.5028, "step": 21641 }, { "epoch": 4.12, "learning_rate": 1.7649952335557675e-06, "loss": 1.7506, "step": 21642 }, { "epoch": 4.12, "learning_rate": 1.7649380362249761e-06, "loss": 2.0798, "step": 21643 }, { "epoch": 4.12, "learning_rate": 1.764880838894185e-06, "loss": 2.6896, "step": 21644 }, { "epoch": 4.12, "learning_rate": 1.7648236415633937e-06, "loss": 3.1713, "step": 21645 }, { "epoch": 4.12, "learning_rate": 1.7647664442326025e-06, "loss": 2.2709, "step": 21646 }, { "epoch": 4.12, "learning_rate": 1.7647092469018115e-06, "loss": 2.1929, "step": 21647 }, { "epoch": 4.12, "learning_rate": 1.76465204957102e-06, "loss": 2.5893, "step": 21648 }, { "epoch": 4.12, "learning_rate": 1.7645948522402287e-06, "loss": 3.1229, "step": 21649 }, { "epoch": 4.12, "learning_rate": 1.7645376549094377e-06, "loss": 1.5637, "step": 21650 }, { "epoch": 4.12, "learning_rate": 1.7644804575786465e-06, "loss": 1.5851, "step": 21651 }, { "epoch": 4.12, "learning_rate": 1.764423260247855e-06, "loss": 2.2133, "step": 21652 }, { "epoch": 4.12, "learning_rate": 1.764366062917064e-06, "loss": 1.3338, "step": 21653 }, { "epoch": 4.12, "learning_rate": 1.7643088655862726e-06, "loss": 3.0092, "step": 21654 }, { "epoch": 4.12, "learning_rate": 1.7642516682554814e-06, "loss": 1.3922, "step": 21655 }, { "epoch": 4.12, "learning_rate": 1.7641944709246902e-06, "loss": 2.5757, "step": 21656 }, { "epoch": 4.13, "learning_rate": 1.764137273593899e-06, "loss": 2.3439, "step": 21657 }, { "epoch": 4.13, "learning_rate": 1.7640800762631076e-06, "loss": 1.9958, "step": 21658 }, { "epoch": 4.13, "learning_rate": 1.7640228789323166e-06, "loss": 1.8756, "step": 21659 }, { "epoch": 4.13, "learning_rate": 1.7639656816015252e-06, "loss": 1.647, "step": 21660 }, { "epoch": 4.13, "learning_rate": 1.763908484270734e-06, "loss": 2.6206, "step": 21661 }, { "epoch": 4.13, "learning_rate": 1.763851286939943e-06, "loss": 1.7076, "step": 21662 }, { "epoch": 4.13, "learning_rate": 1.7637940896091516e-06, "loss": 1.5152, "step": 21663 }, { "epoch": 4.13, "learning_rate": 1.7637368922783606e-06, "loss": 1.8664, "step": 21664 }, { "epoch": 4.13, "learning_rate": 1.7636796949475692e-06, "loss": 1.9352, "step": 21665 }, { "epoch": 4.13, "learning_rate": 1.7636224976167778e-06, "loss": 1.8024, "step": 21666 }, { "epoch": 4.13, "learning_rate": 1.7635653002859868e-06, "loss": 1.5019, "step": 21667 }, { "epoch": 4.13, "learning_rate": 1.7635081029551956e-06, "loss": 3.3978, "step": 21668 }, { "epoch": 4.13, "learning_rate": 1.7634509056244041e-06, "loss": 3.3828, "step": 21669 }, { "epoch": 4.13, "learning_rate": 1.7633937082936131e-06, "loss": 2.1018, "step": 21670 }, { "epoch": 4.13, "learning_rate": 1.7633365109628217e-06, "loss": 1.3641, "step": 21671 }, { "epoch": 4.13, "learning_rate": 1.7632793136320305e-06, "loss": 3.7185, "step": 21672 }, { "epoch": 4.13, "learning_rate": 1.7632221163012395e-06, "loss": 1.8427, "step": 21673 }, { "epoch": 4.13, "learning_rate": 1.763164918970448e-06, "loss": 2.4591, "step": 21674 }, { "epoch": 4.13, "learning_rate": 1.7631077216396567e-06, "loss": 2.733, "step": 21675 }, { "epoch": 4.13, "learning_rate": 1.7630505243088657e-06, "loss": 1.8844, "step": 21676 }, { "epoch": 4.13, "learning_rate": 1.7629933269780743e-06, "loss": 1.0217, "step": 21677 }, { "epoch": 4.13, "learning_rate": 1.762936129647283e-06, "loss": 2.4821, "step": 21678 }, { "epoch": 4.13, "learning_rate": 1.762878932316492e-06, "loss": 2.4456, "step": 21679 }, { "epoch": 4.13, "learning_rate": 1.7628217349857007e-06, "loss": 2.6658, "step": 21680 }, { "epoch": 4.13, "learning_rate": 1.7627645376549097e-06, "loss": 3.0797, "step": 21681 }, { "epoch": 4.13, "learning_rate": 1.7627073403241182e-06, "loss": 2.6377, "step": 21682 }, { "epoch": 4.13, "learning_rate": 1.762650142993327e-06, "loss": 2.5807, "step": 21683 }, { "epoch": 4.13, "learning_rate": 1.7625929456625358e-06, "loss": 1.9675, "step": 21684 }, { "epoch": 4.13, "learning_rate": 1.7625357483317446e-06, "loss": 1.8432, "step": 21685 }, { "epoch": 4.13, "learning_rate": 1.7624785510009532e-06, "loss": 2.2772, "step": 21686 }, { "epoch": 4.13, "learning_rate": 1.7624213536701622e-06, "loss": 2.6977, "step": 21687 }, { "epoch": 4.13, "learning_rate": 1.7623641563393708e-06, "loss": 2.3827, "step": 21688 }, { "epoch": 4.13, "learning_rate": 1.7623069590085796e-06, "loss": 1.5094, "step": 21689 }, { "epoch": 4.13, "learning_rate": 1.7622497616777886e-06, "loss": 1.4952, "step": 21690 }, { "epoch": 4.13, "learning_rate": 1.7621925643469972e-06, "loss": 2.1275, "step": 21691 }, { "epoch": 4.13, "learning_rate": 1.7621353670162058e-06, "loss": 1.1971, "step": 21692 }, { "epoch": 4.13, "learning_rate": 1.7620781696854148e-06, "loss": 1.646, "step": 21693 }, { "epoch": 4.13, "learning_rate": 1.7620209723546236e-06, "loss": 1.7064, "step": 21694 }, { "epoch": 4.13, "learning_rate": 1.7619637750238321e-06, "loss": 2.1539, "step": 21695 }, { "epoch": 4.13, "learning_rate": 1.7619065776930411e-06, "loss": 1.913, "step": 21696 }, { "epoch": 4.13, "learning_rate": 1.7618493803622497e-06, "loss": 2.6637, "step": 21697 }, { "epoch": 4.13, "learning_rate": 1.7617921830314587e-06, "loss": 2.1394, "step": 21698 }, { "epoch": 4.13, "learning_rate": 1.7617349857006673e-06, "loss": 1.994, "step": 21699 }, { "epoch": 4.13, "learning_rate": 1.7616777883698761e-06, "loss": 2.3305, "step": 21700 }, { "epoch": 4.13, "learning_rate": 1.7616205910390851e-06, "loss": 2.1886, "step": 21701 }, { "epoch": 4.13, "learning_rate": 1.7615633937082937e-06, "loss": 1.8018, "step": 21702 }, { "epoch": 4.13, "learning_rate": 1.7615061963775023e-06, "loss": 2.5621, "step": 21703 }, { "epoch": 4.13, "learning_rate": 1.7614489990467113e-06, "loss": 2.9333, "step": 21704 }, { "epoch": 4.13, "learning_rate": 1.7613918017159199e-06, "loss": 4.0888, "step": 21705 }, { "epoch": 4.13, "learning_rate": 1.7613346043851287e-06, "loss": 1.2539, "step": 21706 }, { "epoch": 4.13, "learning_rate": 1.7612774070543377e-06, "loss": 1.1949, "step": 21707 }, { "epoch": 4.13, "learning_rate": 1.7612202097235462e-06, "loss": 2.1387, "step": 21708 }, { "epoch": 4.14, "learning_rate": 1.7611630123927548e-06, "loss": 1.8425, "step": 21709 }, { "epoch": 4.14, "learning_rate": 1.7611058150619638e-06, "loss": 2.4887, "step": 21710 }, { "epoch": 4.14, "learning_rate": 1.7610486177311726e-06, "loss": 2.509, "step": 21711 }, { "epoch": 4.14, "learning_rate": 1.7609914204003814e-06, "loss": 1.005, "step": 21712 }, { "epoch": 4.14, "learning_rate": 1.7609342230695902e-06, "loss": 2.3573, "step": 21713 }, { "epoch": 4.14, "learning_rate": 1.7608770257387988e-06, "loss": 3.3924, "step": 21714 }, { "epoch": 4.14, "learning_rate": 1.7608198284080078e-06, "loss": 1.5592, "step": 21715 }, { "epoch": 4.14, "learning_rate": 1.7607626310772164e-06, "loss": 3.9511, "step": 21716 }, { "epoch": 4.14, "learning_rate": 1.7607054337464252e-06, "loss": 2.8537, "step": 21717 }, { "epoch": 4.14, "learning_rate": 1.7606482364156342e-06, "loss": 2.2203, "step": 21718 }, { "epoch": 4.14, "learning_rate": 1.7605910390848428e-06, "loss": 1.8671, "step": 21719 }, { "epoch": 4.14, "learning_rate": 1.7605338417540513e-06, "loss": 1.8382, "step": 21720 }, { "epoch": 4.14, "learning_rate": 1.7604766444232604e-06, "loss": 1.8854, "step": 21721 }, { "epoch": 4.14, "learning_rate": 1.7604194470924691e-06, "loss": 1.4343, "step": 21722 }, { "epoch": 4.14, "learning_rate": 1.7603622497616777e-06, "loss": 3.1727, "step": 21723 }, { "epoch": 4.14, "learning_rate": 1.7603050524308867e-06, "loss": 1.472, "step": 21724 }, { "epoch": 4.14, "learning_rate": 1.7602478551000953e-06, "loss": 1.2814, "step": 21725 }, { "epoch": 4.14, "learning_rate": 1.760190657769304e-06, "loss": 2.9718, "step": 21726 }, { "epoch": 4.14, "learning_rate": 1.760133460438513e-06, "loss": 0.6892, "step": 21727 }, { "epoch": 4.14, "learning_rate": 1.7600762631077217e-06, "loss": 2.378, "step": 21728 }, { "epoch": 4.14, "learning_rate": 1.7600190657769307e-06, "loss": 1.1481, "step": 21729 }, { "epoch": 4.14, "learning_rate": 1.7599618684461393e-06, "loss": 2.6094, "step": 21730 }, { "epoch": 4.14, "learning_rate": 1.7599046711153479e-06, "loss": 3.1427, "step": 21731 }, { "epoch": 4.14, "learning_rate": 1.7598474737845569e-06, "loss": 3.2427, "step": 21732 }, { "epoch": 4.14, "learning_rate": 1.7597902764537655e-06, "loss": 1.4342, "step": 21733 }, { "epoch": 4.14, "learning_rate": 1.7597330791229742e-06, "loss": 2.3056, "step": 21734 }, { "epoch": 4.14, "learning_rate": 1.7596758817921833e-06, "loss": 3.3101, "step": 21735 }, { "epoch": 4.14, "learning_rate": 1.7596186844613918e-06, "loss": 2.8279, "step": 21736 }, { "epoch": 4.14, "learning_rate": 1.7595614871306004e-06, "loss": 2.5585, "step": 21737 }, { "epoch": 4.14, "learning_rate": 1.7595042897998094e-06, "loss": 2.8689, "step": 21738 }, { "epoch": 4.14, "learning_rate": 1.7594470924690182e-06, "loss": 1.3288, "step": 21739 }, { "epoch": 4.14, "learning_rate": 1.7593898951382268e-06, "loss": 1.3135, "step": 21740 }, { "epoch": 4.14, "learning_rate": 1.7593326978074358e-06, "loss": 2.2573, "step": 21741 }, { "epoch": 4.14, "learning_rate": 1.7592755004766444e-06, "loss": 1.0076, "step": 21742 }, { "epoch": 4.14, "learning_rate": 1.7592183031458532e-06, "loss": 1.7852, "step": 21743 }, { "epoch": 4.14, "learning_rate": 1.759161105815062e-06, "loss": 2.2125, "step": 21744 }, { "epoch": 4.14, "learning_rate": 1.7591039084842708e-06, "loss": 2.4485, "step": 21745 }, { "epoch": 4.14, "learning_rate": 1.7590467111534798e-06, "loss": 1.6462, "step": 21746 }, { "epoch": 4.14, "learning_rate": 1.7589895138226884e-06, "loss": 2.8062, "step": 21747 }, { "epoch": 4.14, "learning_rate": 1.758932316491897e-06, "loss": 2.3859, "step": 21748 }, { "epoch": 4.14, "learning_rate": 1.758875119161106e-06, "loss": 1.7557, "step": 21749 }, { "epoch": 4.14, "learning_rate": 1.7588179218303147e-06, "loss": 2.1288, "step": 21750 }, { "epoch": 4.14, "learning_rate": 1.7587607244995233e-06, "loss": 2.4975, "step": 21751 }, { "epoch": 4.14, "learning_rate": 1.7587035271687323e-06, "loss": 1.7277, "step": 21752 }, { "epoch": 4.14, "learning_rate": 1.758646329837941e-06, "loss": 1.5835, "step": 21753 }, { "epoch": 4.14, "learning_rate": 1.7585891325071495e-06, "loss": 1.6003, "step": 21754 }, { "epoch": 4.14, "learning_rate": 1.7585319351763585e-06, "loss": 2.6603, "step": 21755 }, { "epoch": 4.14, "learning_rate": 1.7584747378455673e-06, "loss": 2.7736, "step": 21756 }, { "epoch": 4.14, "learning_rate": 1.7584175405147759e-06, "loss": 2.6076, "step": 21757 }, { "epoch": 4.14, "learning_rate": 1.7583603431839849e-06, "loss": 1.8344, "step": 21758 }, { "epoch": 4.14, "learning_rate": 1.7583031458531935e-06, "loss": 0.7118, "step": 21759 }, { "epoch": 4.14, "learning_rate": 1.7582459485224023e-06, "loss": 2.4593, "step": 21760 }, { "epoch": 4.14, "learning_rate": 1.7581887511916113e-06, "loss": 2.4323, "step": 21761 }, { "epoch": 4.15, "learning_rate": 1.7581315538608198e-06, "loss": 1.893, "step": 21762 }, { "epoch": 4.15, "learning_rate": 1.7580743565300288e-06, "loss": 1.939, "step": 21763 }, { "epoch": 4.15, "learning_rate": 1.7580171591992374e-06, "loss": 1.8089, "step": 21764 }, { "epoch": 4.15, "learning_rate": 1.757959961868446e-06, "loss": 1.2266, "step": 21765 }, { "epoch": 4.15, "learning_rate": 1.757902764537655e-06, "loss": 2.8673, "step": 21766 }, { "epoch": 4.15, "learning_rate": 1.7578455672068638e-06, "loss": 2.1756, "step": 21767 }, { "epoch": 4.15, "learning_rate": 1.7577883698760724e-06, "loss": 1.9228, "step": 21768 }, { "epoch": 4.15, "learning_rate": 1.7577311725452814e-06, "loss": 1.8367, "step": 21769 }, { "epoch": 4.15, "learning_rate": 1.75767397521449e-06, "loss": 2.7329, "step": 21770 }, { "epoch": 4.15, "learning_rate": 1.7576167778836988e-06, "loss": 1.9518, "step": 21771 }, { "epoch": 4.15, "learning_rate": 1.7575595805529076e-06, "loss": 2.6232, "step": 21772 }, { "epoch": 4.15, "learning_rate": 1.7575023832221164e-06, "loss": 0.9774, "step": 21773 }, { "epoch": 4.15, "learning_rate": 1.757445185891325e-06, "loss": 2.0364, "step": 21774 }, { "epoch": 4.15, "learning_rate": 1.757387988560534e-06, "loss": 2.3543, "step": 21775 }, { "epoch": 4.15, "learning_rate": 1.7573307912297425e-06, "loss": 2.5547, "step": 21776 }, { "epoch": 4.15, "learning_rate": 1.7572735938989515e-06, "loss": 2.9117, "step": 21777 }, { "epoch": 4.15, "learning_rate": 1.7572163965681603e-06, "loss": 1.1362, "step": 21778 }, { "epoch": 4.15, "learning_rate": 1.757159199237369e-06, "loss": 2.702, "step": 21779 }, { "epoch": 4.15, "learning_rate": 1.757102001906578e-06, "loss": 2.9657, "step": 21780 }, { "epoch": 4.15, "learning_rate": 1.7570448045757865e-06, "loss": 1.2503, "step": 21781 }, { "epoch": 4.15, "learning_rate": 1.756987607244995e-06, "loss": 2.5499, "step": 21782 }, { "epoch": 4.15, "learning_rate": 1.756930409914204e-06, "loss": 1.6199, "step": 21783 }, { "epoch": 4.15, "learning_rate": 1.7568732125834129e-06, "loss": 1.5363, "step": 21784 }, { "epoch": 4.15, "learning_rate": 1.7568160152526215e-06, "loss": 2.2173, "step": 21785 }, { "epoch": 4.15, "learning_rate": 1.7567588179218305e-06, "loss": 2.5141, "step": 21786 }, { "epoch": 4.15, "learning_rate": 1.756701620591039e-06, "loss": 2.5619, "step": 21787 }, { "epoch": 4.15, "learning_rate": 1.7566444232602478e-06, "loss": 1.6183, "step": 21788 }, { "epoch": 4.15, "learning_rate": 1.7565872259294569e-06, "loss": 1.9671, "step": 21789 }, { "epoch": 4.15, "learning_rate": 1.7565300285986654e-06, "loss": 2.0022, "step": 21790 }, { "epoch": 4.15, "learning_rate": 1.756472831267874e-06, "loss": 1.8939, "step": 21791 }, { "epoch": 4.15, "learning_rate": 1.756415633937083e-06, "loss": 2.2273, "step": 21792 }, { "epoch": 4.15, "learning_rate": 1.7563584366062916e-06, "loss": 2.0181, "step": 21793 }, { "epoch": 4.15, "learning_rate": 1.7563012392755006e-06, "loss": 2.1949, "step": 21794 }, { "epoch": 4.15, "learning_rate": 1.7562440419447094e-06, "loss": 2.204, "step": 21795 }, { "epoch": 4.15, "learning_rate": 1.756186844613918e-06, "loss": 1.7903, "step": 21796 }, { "epoch": 4.15, "learning_rate": 1.756129647283127e-06, "loss": 2.0352, "step": 21797 }, { "epoch": 4.15, "learning_rate": 1.7560724499523356e-06, "loss": 1.8787, "step": 21798 }, { "epoch": 4.15, "learning_rate": 1.7560152526215444e-06, "loss": 1.6336, "step": 21799 }, { "epoch": 4.15, "learning_rate": 1.7559580552907532e-06, "loss": 2.114, "step": 21800 }, { "epoch": 4.15, "learning_rate": 1.755900857959962e-06, "loss": 3.3517, "step": 21801 }, { "epoch": 4.15, "learning_rate": 1.7558436606291705e-06, "loss": 3.7599, "step": 21802 }, { "epoch": 4.15, "learning_rate": 1.7557864632983795e-06, "loss": 2.8763, "step": 21803 }, { "epoch": 4.15, "learning_rate": 1.7557292659675881e-06, "loss": 1.9416, "step": 21804 }, { "epoch": 4.15, "learning_rate": 1.755672068636797e-06, "loss": 1.3052, "step": 21805 }, { "epoch": 4.15, "learning_rate": 1.755614871306006e-06, "loss": 2.1769, "step": 21806 }, { "epoch": 4.15, "learning_rate": 1.7555576739752145e-06, "loss": 2.332, "step": 21807 }, { "epoch": 4.15, "learning_rate": 1.755500476644423e-06, "loss": 2.0261, "step": 21808 }, { "epoch": 4.15, "learning_rate": 1.755443279313632e-06, "loss": 1.3936, "step": 21809 }, { "epoch": 4.15, "learning_rate": 1.7553860819828409e-06, "loss": 1.511, "step": 21810 }, { "epoch": 4.15, "learning_rate": 1.7553288846520497e-06, "loss": 1.844, "step": 21811 }, { "epoch": 4.15, "learning_rate": 1.7552716873212585e-06, "loss": 1.0784, "step": 21812 }, { "epoch": 4.15, "learning_rate": 1.755214489990467e-06, "loss": 1.8356, "step": 21813 }, { "epoch": 4.16, "learning_rate": 1.755157292659676e-06, "loss": 2.3792, "step": 21814 }, { "epoch": 4.16, "learning_rate": 1.7551000953288846e-06, "loss": 2.8241, "step": 21815 }, { "epoch": 4.16, "learning_rate": 1.7550428979980934e-06, "loss": 3.0825, "step": 21816 }, { "epoch": 4.16, "learning_rate": 1.7549857006673024e-06, "loss": 2.5336, "step": 21817 }, { "epoch": 4.16, "learning_rate": 1.754928503336511e-06, "loss": 3.3514, "step": 21818 }, { "epoch": 4.16, "learning_rate": 1.7548713060057196e-06, "loss": 1.9332, "step": 21819 }, { "epoch": 4.16, "learning_rate": 1.7548141086749286e-06, "loss": 2.2923, "step": 21820 }, { "epoch": 4.16, "learning_rate": 1.7547569113441372e-06, "loss": 1.845, "step": 21821 }, { "epoch": 4.16, "learning_rate": 1.754699714013346e-06, "loss": 1.4362, "step": 21822 }, { "epoch": 4.16, "learning_rate": 1.754642516682555e-06, "loss": 2.0482, "step": 21823 }, { "epoch": 4.16, "learning_rate": 1.7545853193517636e-06, "loss": 2.4771, "step": 21824 }, { "epoch": 4.16, "learning_rate": 1.7545281220209722e-06, "loss": 1.67, "step": 21825 }, { "epoch": 4.16, "learning_rate": 1.7544709246901812e-06, "loss": 2.1599, "step": 21826 }, { "epoch": 4.16, "learning_rate": 1.75441372735939e-06, "loss": 2.7045, "step": 21827 }, { "epoch": 4.16, "learning_rate": 1.754356530028599e-06, "loss": 2.0367, "step": 21828 }, { "epoch": 4.16, "learning_rate": 1.7542993326978075e-06, "loss": 3.3155, "step": 21829 }, { "epoch": 4.16, "learning_rate": 1.7542421353670161e-06, "loss": 1.979, "step": 21830 }, { "epoch": 4.16, "learning_rate": 1.7541849380362251e-06, "loss": 1.858, "step": 21831 }, { "epoch": 4.16, "learning_rate": 1.7541277407054337e-06, "loss": 1.2147, "step": 21832 }, { "epoch": 4.16, "learning_rate": 1.7540705433746425e-06, "loss": 2.8281, "step": 21833 }, { "epoch": 4.16, "learning_rate": 1.7540133460438515e-06, "loss": 3.2175, "step": 21834 }, { "epoch": 4.16, "learning_rate": 1.75395614871306e-06, "loss": 1.7273, "step": 21835 }, { "epoch": 4.16, "learning_rate": 1.7538989513822687e-06, "loss": 1.6616, "step": 21836 }, { "epoch": 4.16, "learning_rate": 1.7538417540514777e-06, "loss": 1.8296, "step": 21837 }, { "epoch": 4.16, "learning_rate": 1.7537845567206865e-06, "loss": 2.8671, "step": 21838 }, { "epoch": 4.16, "learning_rate": 1.753727359389895e-06, "loss": 1.5054, "step": 21839 }, { "epoch": 4.16, "learning_rate": 1.753670162059104e-06, "loss": 1.6885, "step": 21840 }, { "epoch": 4.16, "learning_rate": 1.7536129647283126e-06, "loss": 2.162, "step": 21841 }, { "epoch": 4.16, "learning_rate": 1.7535557673975217e-06, "loss": 3.1092, "step": 21842 }, { "epoch": 4.16, "learning_rate": 1.7534985700667302e-06, "loss": 2.997, "step": 21843 }, { "epoch": 4.16, "learning_rate": 1.753441372735939e-06, "loss": 2.4603, "step": 21844 }, { "epoch": 4.16, "learning_rate": 1.753384175405148e-06, "loss": 1.8377, "step": 21845 }, { "epoch": 4.16, "learning_rate": 1.7533269780743566e-06, "loss": 2.2236, "step": 21846 }, { "epoch": 4.16, "learning_rate": 1.7532697807435652e-06, "loss": 1.7445, "step": 21847 }, { "epoch": 4.16, "learning_rate": 1.7532125834127742e-06, "loss": 2.5901, "step": 21848 }, { "epoch": 4.16, "learning_rate": 1.753155386081983e-06, "loss": 2.8303, "step": 21849 }, { "epoch": 4.16, "learning_rate": 1.7530981887511916e-06, "loss": 1.8656, "step": 21850 }, { "epoch": 4.16, "learning_rate": 1.7530409914204006e-06, "loss": 2.4504, "step": 21851 }, { "epoch": 4.16, "learning_rate": 1.7529837940896092e-06, "loss": 1.464, "step": 21852 }, { "epoch": 4.16, "learning_rate": 1.7529265967588177e-06, "loss": 2.6688, "step": 21853 }, { "epoch": 4.16, "learning_rate": 1.7528693994280268e-06, "loss": 2.5884, "step": 21854 }, { "epoch": 4.16, "learning_rate": 1.7528122020972355e-06, "loss": 2.6905, "step": 21855 }, { "epoch": 4.16, "learning_rate": 1.7527550047664441e-06, "loss": 2.1278, "step": 21856 }, { "epoch": 4.16, "learning_rate": 1.7526978074356531e-06, "loss": 1.6413, "step": 21857 }, { "epoch": 4.16, "learning_rate": 1.7526406101048617e-06, "loss": 2.3247, "step": 21858 }, { "epoch": 4.16, "learning_rate": 1.7525834127740707e-06, "loss": 1.656, "step": 21859 }, { "epoch": 4.16, "learning_rate": 1.7525262154432793e-06, "loss": 2.1451, "step": 21860 }, { "epoch": 4.16, "learning_rate": 1.752469018112488e-06, "loss": 1.9881, "step": 21861 }, { "epoch": 4.16, "learning_rate": 1.752411820781697e-06, "loss": 2.8394, "step": 21862 }, { "epoch": 4.16, "learning_rate": 1.7523546234509057e-06, "loss": 2.5063, "step": 21863 }, { "epoch": 4.16, "learning_rate": 1.7522974261201143e-06, "loss": 2.9053, "step": 21864 }, { "epoch": 4.16, "learning_rate": 1.7522402287893233e-06, "loss": 1.1365, "step": 21865 }, { "epoch": 4.16, "learning_rate": 1.752183031458532e-06, "loss": 1.7637, "step": 21866 }, { "epoch": 4.17, "learning_rate": 1.7521258341277407e-06, "loss": 1.3123, "step": 21867 }, { "epoch": 4.17, "learning_rate": 1.7520686367969497e-06, "loss": 2.43, "step": 21868 }, { "epoch": 4.17, "learning_rate": 1.7520114394661582e-06, "loss": 3.1624, "step": 21869 }, { "epoch": 4.17, "learning_rate": 1.7519542421353668e-06, "loss": 2.4041, "step": 21870 }, { "epoch": 4.17, "learning_rate": 1.7518970448045758e-06, "loss": 2.6293, "step": 21871 }, { "epoch": 4.17, "learning_rate": 1.7518398474737846e-06, "loss": 2.7769, "step": 21872 }, { "epoch": 4.17, "learning_rate": 1.7517826501429932e-06, "loss": 3.6379, "step": 21873 }, { "epoch": 4.17, "learning_rate": 1.7517254528122022e-06, "loss": 2.5663, "step": 21874 }, { "epoch": 4.17, "learning_rate": 1.7516682554814108e-06, "loss": 3.0634, "step": 21875 }, { "epoch": 4.17, "learning_rate": 1.7516110581506198e-06, "loss": 3.018, "step": 21876 }, { "epoch": 4.17, "learning_rate": 1.7515538608198286e-06, "loss": 2.404, "step": 21877 }, { "epoch": 4.17, "learning_rate": 1.7514966634890372e-06, "loss": 1.1928, "step": 21878 }, { "epoch": 4.17, "learning_rate": 1.7514394661582462e-06, "loss": 2.2977, "step": 21879 }, { "epoch": 4.17, "learning_rate": 1.7513822688274548e-06, "loss": 2.6977, "step": 21880 }, { "epoch": 4.17, "learning_rate": 1.7513250714966633e-06, "loss": 1.4488, "step": 21881 }, { "epoch": 4.17, "learning_rate": 1.7512678741658723e-06, "loss": 3.4073, "step": 21882 }, { "epoch": 4.17, "learning_rate": 1.7512106768350811e-06, "loss": 2.285, "step": 21883 }, { "epoch": 4.17, "learning_rate": 1.7511534795042897e-06, "loss": 2.2099, "step": 21884 }, { "epoch": 4.17, "learning_rate": 1.7510962821734987e-06, "loss": 2.401, "step": 21885 }, { "epoch": 4.17, "learning_rate": 1.7510390848427073e-06, "loss": 2.3742, "step": 21886 }, { "epoch": 4.17, "learning_rate": 1.750981887511916e-06, "loss": 2.8002, "step": 21887 }, { "epoch": 4.17, "learning_rate": 1.750924690181125e-06, "loss": 1.6876, "step": 21888 }, { "epoch": 4.17, "learning_rate": 1.7508674928503337e-06, "loss": 1.7902, "step": 21889 }, { "epoch": 4.17, "learning_rate": 1.7508102955195423e-06, "loss": 2.743, "step": 21890 }, { "epoch": 4.17, "learning_rate": 1.7507530981887513e-06, "loss": 2.9613, "step": 21891 }, { "epoch": 4.17, "learning_rate": 1.7506959008579599e-06, "loss": 1.6983, "step": 21892 }, { "epoch": 4.17, "learning_rate": 1.7506387035271689e-06, "loss": 2.4111, "step": 21893 }, { "epoch": 4.17, "learning_rate": 1.7505815061963777e-06, "loss": 2.3774, "step": 21894 }, { "epoch": 4.17, "learning_rate": 1.7505243088655862e-06, "loss": 1.8643, "step": 21895 }, { "epoch": 4.17, "learning_rate": 1.7504671115347952e-06, "loss": 2.5274, "step": 21896 }, { "epoch": 4.17, "learning_rate": 1.7504099142040038e-06, "loss": 2.3098, "step": 21897 }, { "epoch": 4.17, "learning_rate": 1.7503527168732126e-06, "loss": 1.3358, "step": 21898 }, { "epoch": 4.17, "learning_rate": 1.7502955195424214e-06, "loss": 1.1769, "step": 21899 }, { "epoch": 4.17, "learning_rate": 1.7502383222116302e-06, "loss": 4.3338, "step": 21900 }, { "epoch": 4.17, "learning_rate": 1.7501811248808388e-06, "loss": 2.0338, "step": 21901 }, { "epoch": 4.17, "learning_rate": 1.7501239275500478e-06, "loss": 1.9313, "step": 21902 }, { "epoch": 4.17, "learning_rate": 1.7500667302192564e-06, "loss": 3.4269, "step": 21903 }, { "epoch": 4.17, "learning_rate": 1.7500095328884652e-06, "loss": 2.0327, "step": 21904 }, { "epoch": 4.17, "learning_rate": 1.7499523355576742e-06, "loss": 3.7985, "step": 21905 }, { "epoch": 4.17, "learning_rate": 1.7498951382268828e-06, "loss": 1.9399, "step": 21906 }, { "epoch": 4.17, "learning_rate": 1.7498379408960918e-06, "loss": 1.6788, "step": 21907 }, { "epoch": 4.17, "learning_rate": 1.7497807435653004e-06, "loss": 2.2508, "step": 21908 }, { "epoch": 4.17, "learning_rate": 1.749723546234509e-06, "loss": 2.1679, "step": 21909 }, { "epoch": 4.17, "learning_rate": 1.749666348903718e-06, "loss": 1.9303, "step": 21910 }, { "epoch": 4.17, "learning_rate": 1.7496091515729267e-06, "loss": 1.7786, "step": 21911 }, { "epoch": 4.17, "learning_rate": 1.7495519542421353e-06, "loss": 1.9596, "step": 21912 }, { "epoch": 4.17, "learning_rate": 1.7494947569113443e-06, "loss": 1.4989, "step": 21913 }, { "epoch": 4.17, "learning_rate": 1.749437559580553e-06, "loss": 2.942, "step": 21914 }, { "epoch": 4.17, "learning_rate": 1.7493803622497617e-06, "loss": 2.5333, "step": 21915 }, { "epoch": 4.17, "learning_rate": 1.7493231649189707e-06, "loss": 1.7845, "step": 21916 }, { "epoch": 4.17, "learning_rate": 1.7492659675881793e-06, "loss": 1.7821, "step": 21917 }, { "epoch": 4.17, "learning_rate": 1.7492087702573879e-06, "loss": 3.0288, "step": 21918 }, { "epoch": 4.18, "learning_rate": 1.7491515729265969e-06, "loss": 2.0428, "step": 21919 }, { "epoch": 4.18, "learning_rate": 1.7490943755958055e-06, "loss": 1.5866, "step": 21920 }, { "epoch": 4.18, "learning_rate": 1.7490371782650142e-06, "loss": 2.2202, "step": 21921 }, { "epoch": 4.18, "learning_rate": 1.7489799809342233e-06, "loss": 1.596, "step": 21922 }, { "epoch": 4.18, "learning_rate": 1.7489227836034318e-06, "loss": 1.5666, "step": 21923 }, { "epoch": 4.18, "learning_rate": 1.7488655862726408e-06, "loss": 2.4414, "step": 21924 }, { "epoch": 4.18, "learning_rate": 1.7488083889418494e-06, "loss": 2.7544, "step": 21925 }, { "epoch": 4.18, "learning_rate": 1.7487511916110582e-06, "loss": 1.3159, "step": 21926 }, { "epoch": 4.18, "learning_rate": 1.748693994280267e-06, "loss": 0.9183, "step": 21927 }, { "epoch": 4.18, "learning_rate": 1.7486367969494758e-06, "loss": 2.2236, "step": 21928 }, { "epoch": 4.18, "learning_rate": 1.7485795996186844e-06, "loss": 1.7039, "step": 21929 }, { "epoch": 4.18, "learning_rate": 1.7485224022878934e-06, "loss": 3.2422, "step": 21930 }, { "epoch": 4.18, "learning_rate": 1.748465204957102e-06, "loss": 2.706, "step": 21931 }, { "epoch": 4.18, "learning_rate": 1.7484080076263108e-06, "loss": 1.5908, "step": 21932 }, { "epoch": 4.18, "learning_rate": 1.7483508102955198e-06, "loss": 2.0603, "step": 21933 }, { "epoch": 4.18, "learning_rate": 1.7482936129647284e-06, "loss": 3.0215, "step": 21934 }, { "epoch": 4.18, "learning_rate": 1.748236415633937e-06, "loss": 3.0571, "step": 21935 }, { "epoch": 4.18, "learning_rate": 1.748179218303146e-06, "loss": 1.3649, "step": 21936 }, { "epoch": 4.18, "learning_rate": 1.7481220209723545e-06, "loss": 1.8063, "step": 21937 }, { "epoch": 4.18, "learning_rate": 1.7480648236415633e-06, "loss": 2.6702, "step": 21938 }, { "epoch": 4.18, "learning_rate": 1.7480076263107723e-06, "loss": 2.6139, "step": 21939 }, { "epoch": 4.18, "learning_rate": 1.747950428979981e-06, "loss": 2.6561, "step": 21940 }, { "epoch": 4.18, "learning_rate": 1.74789323164919e-06, "loss": 2.7002, "step": 21941 }, { "epoch": 4.18, "learning_rate": 1.7478360343183985e-06, "loss": 2.4032, "step": 21942 }, { "epoch": 4.18, "learning_rate": 1.7477788369876073e-06, "loss": 2.1536, "step": 21943 }, { "epoch": 4.18, "learning_rate": 1.7477216396568163e-06, "loss": 0.9952, "step": 21944 }, { "epoch": 4.18, "learning_rate": 1.7476644423260249e-06, "loss": 1.2811, "step": 21945 }, { "epoch": 4.18, "learning_rate": 1.7476072449952335e-06, "loss": 1.2675, "step": 21946 }, { "epoch": 4.18, "learning_rate": 1.7475500476644425e-06, "loss": 1.5734, "step": 21947 }, { "epoch": 4.18, "learning_rate": 1.747492850333651e-06, "loss": 1.6194, "step": 21948 }, { "epoch": 4.18, "learning_rate": 1.7474356530028598e-06, "loss": 3.1514, "step": 21949 }, { "epoch": 4.18, "learning_rate": 1.7473784556720688e-06, "loss": 2.0986, "step": 21950 }, { "epoch": 4.18, "learning_rate": 1.7473212583412774e-06, "loss": 2.7631, "step": 21951 }, { "epoch": 4.18, "learning_rate": 1.747264061010486e-06, "loss": 2.3503, "step": 21952 }, { "epoch": 4.18, "learning_rate": 1.747206863679695e-06, "loss": 2.4197, "step": 21953 }, { "epoch": 4.18, "learning_rate": 1.7471496663489038e-06, "loss": 3.0497, "step": 21954 }, { "epoch": 4.18, "learning_rate": 1.7470924690181124e-06, "loss": 2.1774, "step": 21955 }, { "epoch": 4.18, "learning_rate": 1.7470352716873214e-06, "loss": 2.2457, "step": 21956 }, { "epoch": 4.18, "learning_rate": 1.74697807435653e-06, "loss": 1.5118, "step": 21957 }, { "epoch": 4.18, "learning_rate": 1.746920877025739e-06, "loss": 2.1482, "step": 21958 }, { "epoch": 4.18, "learning_rate": 1.7468636796949476e-06, "loss": 2.2443, "step": 21959 }, { "epoch": 4.18, "learning_rate": 1.7468064823641564e-06, "loss": 2.0126, "step": 21960 }, { "epoch": 4.18, "learning_rate": 1.7467492850333654e-06, "loss": 2.8585, "step": 21961 }, { "epoch": 4.18, "learning_rate": 1.746692087702574e-06, "loss": 2.0575, "step": 21962 }, { "epoch": 4.18, "learning_rate": 1.7466348903717825e-06, "loss": 1.5, "step": 21963 }, { "epoch": 4.18, "learning_rate": 1.7465776930409915e-06, "loss": 3.2287, "step": 21964 }, { "epoch": 4.18, "learning_rate": 1.7465204957102003e-06, "loss": 1.1887, "step": 21965 }, { "epoch": 4.18, "learning_rate": 1.746463298379409e-06, "loss": 1.1833, "step": 21966 }, { "epoch": 4.18, "learning_rate": 1.746406101048618e-06, "loss": 1.5982, "step": 21967 }, { "epoch": 4.18, "learning_rate": 1.7463489037178265e-06, "loss": 3.1801, "step": 21968 }, { "epoch": 4.18, "learning_rate": 1.746291706387035e-06, "loss": 2.7331, "step": 21969 }, { "epoch": 4.18, "learning_rate": 1.746234509056244e-06, "loss": 1.3168, "step": 21970 }, { "epoch": 4.18, "learning_rate": 1.7461773117254529e-06, "loss": 3.2109, "step": 21971 }, { "epoch": 4.19, "learning_rate": 1.7461201143946619e-06, "loss": 2.5237, "step": 21972 }, { "epoch": 4.19, "learning_rate": 1.7460629170638705e-06, "loss": 1.9525, "step": 21973 }, { "epoch": 4.19, "learning_rate": 1.746005719733079e-06, "loss": 3.0641, "step": 21974 }, { "epoch": 4.19, "learning_rate": 1.745948522402288e-06, "loss": 1.1066, "step": 21975 }, { "epoch": 4.19, "learning_rate": 1.7458913250714966e-06, "loss": 2.4404, "step": 21976 }, { "epoch": 4.19, "learning_rate": 1.7458341277407054e-06, "loss": 2.4944, "step": 21977 }, { "epoch": 4.19, "learning_rate": 1.7457769304099144e-06, "loss": 3.0044, "step": 21978 }, { "epoch": 4.19, "learning_rate": 1.745719733079123e-06, "loss": 3.3478, "step": 21979 }, { "epoch": 4.19, "learning_rate": 1.7456625357483316e-06, "loss": 2.7936, "step": 21980 }, { "epoch": 4.19, "learning_rate": 1.7456053384175406e-06, "loss": 2.1436, "step": 21981 }, { "epoch": 4.19, "learning_rate": 1.7455481410867494e-06, "loss": 1.9434, "step": 21982 }, { "epoch": 4.19, "learning_rate": 1.745490943755958e-06, "loss": 1.8663, "step": 21983 }, { "epoch": 4.19, "learning_rate": 1.745433746425167e-06, "loss": 3.0418, "step": 21984 }, { "epoch": 4.19, "learning_rate": 1.7453765490943756e-06, "loss": 1.6906, "step": 21985 }, { "epoch": 4.19, "learning_rate": 1.7453193517635844e-06, "loss": 1.2414, "step": 21986 }, { "epoch": 4.19, "learning_rate": 1.7452621544327932e-06, "loss": 1.898, "step": 21987 }, { "epoch": 4.19, "learning_rate": 1.745204957102002e-06, "loss": 2.7586, "step": 21988 }, { "epoch": 4.19, "learning_rate": 1.745147759771211e-06, "loss": 1.6513, "step": 21989 }, { "epoch": 4.19, "learning_rate": 1.7450905624404195e-06, "loss": 3.4327, "step": 21990 }, { "epoch": 4.19, "learning_rate": 1.7450333651096281e-06, "loss": 2.4145, "step": 21991 }, { "epoch": 4.19, "learning_rate": 1.7449761677788371e-06, "loss": 2.7888, "step": 21992 }, { "epoch": 4.19, "learning_rate": 1.744918970448046e-06, "loss": 2.0985, "step": 21993 }, { "epoch": 4.19, "learning_rate": 1.7448617731172545e-06, "loss": 2.5369, "step": 21994 }, { "epoch": 4.19, "learning_rate": 1.7448045757864635e-06, "loss": 2.3449, "step": 21995 }, { "epoch": 4.19, "learning_rate": 1.744747378455672e-06, "loss": 3.159, "step": 21996 }, { "epoch": 4.19, "learning_rate": 1.7446901811248807e-06, "loss": 0.789, "step": 21997 }, { "epoch": 4.19, "learning_rate": 1.7446329837940897e-06, "loss": 1.7973, "step": 21998 }, { "epoch": 4.19, "learning_rate": 1.7445757864632985e-06, "loss": 2.3014, "step": 21999 }, { "epoch": 4.19, "learning_rate": 1.744518589132507e-06, "loss": 2.0958, "step": 22000 }, { "epoch": 4.19, "learning_rate": 1.744461391801716e-06, "loss": 2.3778, "step": 22001 }, { "epoch": 4.19, "learning_rate": 1.7444041944709246e-06, "loss": 3.1433, "step": 22002 }, { "epoch": 4.19, "learning_rate": 1.7443469971401334e-06, "loss": 3.5083, "step": 22003 }, { "epoch": 4.19, "learning_rate": 1.7442897998093424e-06, "loss": 2.4474, "step": 22004 }, { "epoch": 4.19, "learning_rate": 1.744232602478551e-06, "loss": 2.6186, "step": 22005 }, { "epoch": 4.19, "learning_rate": 1.74417540514776e-06, "loss": 2.6166, "step": 22006 }, { "epoch": 4.19, "learning_rate": 1.7441182078169686e-06, "loss": 2.6991, "step": 22007 }, { "epoch": 4.19, "learning_rate": 1.7440610104861772e-06, "loss": 2.5843, "step": 22008 }, { "epoch": 4.19, "learning_rate": 1.7440038131553862e-06, "loss": 1.6361, "step": 22009 }, { "epoch": 4.19, "learning_rate": 1.743946615824595e-06, "loss": 1.33, "step": 22010 }, { "epoch": 4.19, "learning_rate": 1.7438894184938036e-06, "loss": 1.7653, "step": 22011 }, { "epoch": 4.19, "learning_rate": 1.7438322211630126e-06, "loss": 1.5753, "step": 22012 }, { "epoch": 4.19, "learning_rate": 1.7437750238322212e-06, "loss": 3.551, "step": 22013 }, { "epoch": 4.19, "learning_rate": 1.74371782650143e-06, "loss": 2.226, "step": 22014 }, { "epoch": 4.19, "learning_rate": 1.7436606291706387e-06, "loss": 2.7483, "step": 22015 }, { "epoch": 4.19, "learning_rate": 1.7436034318398475e-06, "loss": 1.7084, "step": 22016 }, { "epoch": 4.19, "learning_rate": 1.7435462345090561e-06, "loss": 1.5516, "step": 22017 }, { "epoch": 4.19, "learning_rate": 1.7434890371782651e-06, "loss": 1.8962, "step": 22018 }, { "epoch": 4.19, "learning_rate": 1.7434318398474737e-06, "loss": 3.1699, "step": 22019 }, { "epoch": 4.19, "learning_rate": 1.7433746425166825e-06, "loss": 2.1596, "step": 22020 }, { "epoch": 4.19, "learning_rate": 1.7433174451858915e-06, "loss": 3.3514, "step": 22021 }, { "epoch": 4.19, "learning_rate": 1.7432602478551e-06, "loss": 2.0657, "step": 22022 }, { "epoch": 4.19, "learning_rate": 1.743203050524309e-06, "loss": 1.7741, "step": 22023 }, { "epoch": 4.2, "learning_rate": 1.7431458531935177e-06, "loss": 2.2571, "step": 22024 }, { "epoch": 4.2, "learning_rate": 1.7430886558627263e-06, "loss": 3.1448, "step": 22025 }, { "epoch": 4.2, "learning_rate": 1.7430314585319353e-06, "loss": 2.1651, "step": 22026 }, { "epoch": 4.2, "learning_rate": 1.742974261201144e-06, "loss": 1.8608, "step": 22027 }, { "epoch": 4.2, "learning_rate": 1.7429170638703526e-06, "loss": 1.8352, "step": 22028 }, { "epoch": 4.2, "learning_rate": 1.7428598665395616e-06, "loss": 2.3779, "step": 22029 }, { "epoch": 4.2, "learning_rate": 1.7428026692087702e-06, "loss": 2.2883, "step": 22030 }, { "epoch": 4.2, "learning_rate": 1.742745471877979e-06, "loss": 1.7786, "step": 22031 }, { "epoch": 4.2, "learning_rate": 1.742688274547188e-06, "loss": 2.6054, "step": 22032 }, { "epoch": 4.2, "learning_rate": 1.7426310772163966e-06, "loss": 2.3073, "step": 22033 }, { "epoch": 4.2, "learning_rate": 1.7425738798856052e-06, "loss": 1.7047, "step": 22034 }, { "epoch": 4.2, "learning_rate": 1.7425166825548142e-06, "loss": 1.9596, "step": 22035 }, { "epoch": 4.2, "learning_rate": 1.7424594852240228e-06, "loss": 2.0511, "step": 22036 }, { "epoch": 4.2, "learning_rate": 1.7424022878932316e-06, "loss": 1.3483, "step": 22037 }, { "epoch": 4.2, "learning_rate": 1.7423450905624406e-06, "loss": 3.1697, "step": 22038 }, { "epoch": 4.2, "learning_rate": 1.7422878932316492e-06, "loss": 1.2601, "step": 22039 }, { "epoch": 4.2, "learning_rate": 1.7422306959008582e-06, "loss": 2.4818, "step": 22040 }, { "epoch": 4.2, "learning_rate": 1.7421734985700668e-06, "loss": 2.3884, "step": 22041 }, { "epoch": 4.2, "learning_rate": 1.7421163012392755e-06, "loss": 1.3296, "step": 22042 }, { "epoch": 4.2, "learning_rate": 1.7420591039084843e-06, "loss": 2.4921, "step": 22043 }, { "epoch": 4.2, "learning_rate": 1.7420019065776931e-06, "loss": 2.0656, "step": 22044 }, { "epoch": 4.2, "learning_rate": 1.7419447092469017e-06, "loss": 1.7224, "step": 22045 }, { "epoch": 4.2, "learning_rate": 1.7418875119161107e-06, "loss": 2.3134, "step": 22046 }, { "epoch": 4.2, "learning_rate": 1.7418303145853193e-06, "loss": 1.5624, "step": 22047 }, { "epoch": 4.2, "learning_rate": 1.741773117254528e-06, "loss": 2.0785, "step": 22048 }, { "epoch": 4.2, "learning_rate": 1.741715919923737e-06, "loss": 2.9073, "step": 22049 }, { "epoch": 4.2, "learning_rate": 1.7416587225929457e-06, "loss": 3.1752, "step": 22050 }, { "epoch": 4.2, "learning_rate": 1.7416015252621543e-06, "loss": 2.3875, "step": 22051 }, { "epoch": 4.2, "learning_rate": 1.7415443279313633e-06, "loss": 2.4137, "step": 22052 }, { "epoch": 4.2, "learning_rate": 1.741487130600572e-06, "loss": 3.618, "step": 22053 }, { "epoch": 4.2, "learning_rate": 1.7414299332697809e-06, "loss": 1.6293, "step": 22054 }, { "epoch": 4.2, "learning_rate": 1.7413727359389897e-06, "loss": 1.5137, "step": 22055 }, { "epoch": 4.2, "learning_rate": 1.7413155386081982e-06, "loss": 1.9406, "step": 22056 }, { "epoch": 4.2, "learning_rate": 1.7412583412774072e-06, "loss": 1.7559, "step": 22057 }, { "epoch": 4.2, "learning_rate": 1.7412011439466158e-06, "loss": 3.1078, "step": 22058 }, { "epoch": 4.2, "learning_rate": 1.7411439466158246e-06, "loss": 2.428, "step": 22059 }, { "epoch": 4.2, "learning_rate": 1.7410867492850336e-06, "loss": 1.1526, "step": 22060 }, { "epoch": 4.2, "learning_rate": 1.7410295519542422e-06, "loss": 2.6812, "step": 22061 }, { "epoch": 4.2, "learning_rate": 1.7409723546234508e-06, "loss": 3.0738, "step": 22062 }, { "epoch": 4.2, "learning_rate": 1.7409151572926598e-06, "loss": 2.2234, "step": 22063 }, { "epoch": 4.2, "learning_rate": 1.7408579599618684e-06, "loss": 3.775, "step": 22064 }, { "epoch": 4.2, "learning_rate": 1.7408007626310772e-06, "loss": 3.2174, "step": 22065 }, { "epoch": 4.2, "learning_rate": 1.7407435653002862e-06, "loss": 2.6918, "step": 22066 }, { "epoch": 4.2, "learning_rate": 1.7406863679694948e-06, "loss": 1.6707, "step": 22067 }, { "epoch": 4.2, "learning_rate": 1.7406291706387033e-06, "loss": 2.4435, "step": 22068 }, { "epoch": 4.2, "learning_rate": 1.7405719733079123e-06, "loss": 1.9598, "step": 22069 }, { "epoch": 4.2, "learning_rate": 1.7405147759771211e-06, "loss": 2.6076, "step": 22070 }, { "epoch": 4.2, "learning_rate": 1.7404575786463301e-06, "loss": 0.9392, "step": 22071 }, { "epoch": 4.2, "learning_rate": 1.7404003813155387e-06, "loss": 2.141, "step": 22072 }, { "epoch": 4.2, "learning_rate": 1.7403431839847473e-06, "loss": 2.2221, "step": 22073 }, { "epoch": 4.2, "learning_rate": 1.7402859866539563e-06, "loss": 2.0962, "step": 22074 }, { "epoch": 4.2, "learning_rate": 1.740228789323165e-06, "loss": 2.2584, "step": 22075 }, { "epoch": 4.2, "learning_rate": 1.7401715919923737e-06, "loss": 1.6216, "step": 22076 }, { "epoch": 4.21, "learning_rate": 1.7401143946615827e-06, "loss": 1.8305, "step": 22077 }, { "epoch": 4.21, "learning_rate": 1.7400571973307913e-06, "loss": 2.4605, "step": 22078 }, { "epoch": 4.21, "learning_rate": 1.7399999999999999e-06, "loss": 2.1043, "step": 22079 }, { "epoch": 4.21, "learning_rate": 1.7399428026692089e-06, "loss": 1.7391, "step": 22080 }, { "epoch": 4.21, "learning_rate": 1.7398856053384177e-06, "loss": 1.6869, "step": 22081 }, { "epoch": 4.21, "learning_rate": 1.7398284080076262e-06, "loss": 1.294, "step": 22082 }, { "epoch": 4.21, "learning_rate": 1.7397712106768352e-06, "loss": 2.3076, "step": 22083 }, { "epoch": 4.21, "learning_rate": 1.7397140133460438e-06, "loss": 1.6194, "step": 22084 }, { "epoch": 4.21, "learning_rate": 1.7396568160152524e-06, "loss": 2.1032, "step": 22085 }, { "epoch": 4.21, "learning_rate": 1.7395996186844614e-06, "loss": 2.616, "step": 22086 }, { "epoch": 4.21, "learning_rate": 1.7395424213536702e-06, "loss": 3.0732, "step": 22087 }, { "epoch": 4.21, "learning_rate": 1.7394852240228792e-06, "loss": 1.6992, "step": 22088 }, { "epoch": 4.21, "learning_rate": 1.7394280266920878e-06, "loss": 2.8944, "step": 22089 }, { "epoch": 4.21, "learning_rate": 1.7393708293612964e-06, "loss": 2.2916, "step": 22090 }, { "epoch": 4.21, "learning_rate": 1.7393136320305054e-06, "loss": 3.223, "step": 22091 }, { "epoch": 4.21, "learning_rate": 1.739256434699714e-06, "loss": 2.5355, "step": 22092 }, { "epoch": 4.21, "learning_rate": 1.7391992373689228e-06, "loss": 1.5616, "step": 22093 }, { "epoch": 4.21, "learning_rate": 1.7391420400381318e-06, "loss": 1.9285, "step": 22094 }, { "epoch": 4.21, "learning_rate": 1.7390848427073403e-06, "loss": 2.5613, "step": 22095 }, { "epoch": 4.21, "learning_rate": 1.739027645376549e-06, "loss": 2.1511, "step": 22096 }, { "epoch": 4.21, "learning_rate": 1.738970448045758e-06, "loss": 1.359, "step": 22097 }, { "epoch": 4.21, "learning_rate": 1.7389132507149667e-06, "loss": 2.9841, "step": 22098 }, { "epoch": 4.21, "learning_rate": 1.7388560533841753e-06, "loss": 2.6107, "step": 22099 }, { "epoch": 4.21, "learning_rate": 1.7387988560533843e-06, "loss": 2.3757, "step": 22100 }, { "epoch": 4.21, "learning_rate": 1.738741658722593e-06, "loss": 2.5845, "step": 22101 }, { "epoch": 4.21, "learning_rate": 1.7386844613918017e-06, "loss": 2.563, "step": 22102 }, { "epoch": 4.21, "learning_rate": 1.7386272640610105e-06, "loss": 2.1101, "step": 22103 }, { "epoch": 4.21, "learning_rate": 1.7385700667302193e-06, "loss": 2.3696, "step": 22104 }, { "epoch": 4.21, "learning_rate": 1.7385128693994283e-06, "loss": 2.1629, "step": 22105 }, { "epoch": 4.21, "learning_rate": 1.7384556720686369e-06, "loss": 1.7859, "step": 22106 }, { "epoch": 4.21, "learning_rate": 1.7383984747378454e-06, "loss": 3.0375, "step": 22107 }, { "epoch": 4.21, "learning_rate": 1.7383412774070545e-06, "loss": 1.8192, "step": 22108 }, { "epoch": 4.21, "learning_rate": 1.7382840800762632e-06, "loss": 2.4014, "step": 22109 }, { "epoch": 4.21, "learning_rate": 1.7382268827454718e-06, "loss": 2.6418, "step": 22110 }, { "epoch": 4.21, "learning_rate": 1.7381696854146808e-06, "loss": 2.0675, "step": 22111 }, { "epoch": 4.21, "learning_rate": 1.7381124880838894e-06, "loss": 2.3063, "step": 22112 }, { "epoch": 4.21, "learning_rate": 1.738055290753098e-06, "loss": 1.7029, "step": 22113 }, { "epoch": 4.21, "learning_rate": 1.737998093422307e-06, "loss": 2.5263, "step": 22114 }, { "epoch": 4.21, "learning_rate": 1.7379408960915158e-06, "loss": 3.8843, "step": 22115 }, { "epoch": 4.21, "learning_rate": 1.7378836987607244e-06, "loss": 1.7381, "step": 22116 }, { "epoch": 4.21, "learning_rate": 1.7378265014299334e-06, "loss": 2.7136, "step": 22117 }, { "epoch": 4.21, "learning_rate": 1.737769304099142e-06, "loss": 2.6455, "step": 22118 }, { "epoch": 4.21, "learning_rate": 1.737712106768351e-06, "loss": 3.3419, "step": 22119 }, { "epoch": 4.21, "learning_rate": 1.7376549094375598e-06, "loss": 2.8568, "step": 22120 }, { "epoch": 4.21, "learning_rate": 1.7375977121067684e-06, "loss": 3.2309, "step": 22121 }, { "epoch": 4.21, "learning_rate": 1.7375405147759774e-06, "loss": 3.6008, "step": 22122 }, { "epoch": 4.21, "learning_rate": 1.737483317445186e-06, "loss": 2.8108, "step": 22123 }, { "epoch": 4.21, "learning_rate": 1.7374261201143945e-06, "loss": 2.1126, "step": 22124 }, { "epoch": 4.21, "learning_rate": 1.7373689227836035e-06, "loss": 2.0058, "step": 22125 }, { "epoch": 4.21, "learning_rate": 1.7373117254528123e-06, "loss": 2.029, "step": 22126 }, { "epoch": 4.21, "learning_rate": 1.737254528122021e-06, "loss": 2.2582, "step": 22127 }, { "epoch": 4.21, "learning_rate": 1.73719733079123e-06, "loss": 2.7493, "step": 22128 }, { "epoch": 4.22, "learning_rate": 1.7371401334604385e-06, "loss": 1.5917, "step": 22129 }, { "epoch": 4.22, "learning_rate": 1.7370829361296473e-06, "loss": 2.9856, "step": 22130 }, { "epoch": 4.22, "learning_rate": 1.737025738798856e-06, "loss": 2.3522, "step": 22131 }, { "epoch": 4.22, "learning_rate": 1.7369685414680649e-06, "loss": 2.6509, "step": 22132 }, { "epoch": 4.22, "learning_rate": 1.7369113441372735e-06, "loss": 1.9377, "step": 22133 }, { "epoch": 4.22, "learning_rate": 1.7368541468064825e-06, "loss": 3.4972, "step": 22134 }, { "epoch": 4.22, "learning_rate": 1.736796949475691e-06, "loss": 2.4188, "step": 22135 }, { "epoch": 4.22, "learning_rate": 1.7367397521449e-06, "loss": 2.5782, "step": 22136 }, { "epoch": 4.22, "learning_rate": 1.7366825548141088e-06, "loss": 1.9943, "step": 22137 }, { "epoch": 4.22, "learning_rate": 1.7366253574833174e-06, "loss": 2.1452, "step": 22138 }, { "epoch": 4.22, "learning_rate": 1.7365681601525264e-06, "loss": 2.0511, "step": 22139 }, { "epoch": 4.22, "learning_rate": 1.736510962821735e-06, "loss": 2.7009, "step": 22140 }, { "epoch": 4.22, "learning_rate": 1.7364537654909438e-06, "loss": 2.2951, "step": 22141 }, { "epoch": 4.22, "learning_rate": 1.7363965681601526e-06, "loss": 2.2462, "step": 22142 }, { "epoch": 4.22, "learning_rate": 1.7363393708293614e-06, "loss": 2.48, "step": 22143 }, { "epoch": 4.22, "learning_rate": 1.73628217349857e-06, "loss": 1.4787, "step": 22144 }, { "epoch": 4.22, "learning_rate": 1.736224976167779e-06, "loss": 2.3988, "step": 22145 }, { "epoch": 4.22, "learning_rate": 1.7361677788369876e-06, "loss": 2.1908, "step": 22146 }, { "epoch": 4.22, "learning_rate": 1.7361105815061964e-06, "loss": 1.5161, "step": 22147 }, { "epoch": 4.22, "learning_rate": 1.7360533841754054e-06, "loss": 1.5794, "step": 22148 }, { "epoch": 4.22, "learning_rate": 1.735996186844614e-06, "loss": 1.1303, "step": 22149 }, { "epoch": 4.22, "learning_rate": 1.7359389895138225e-06, "loss": 2.1787, "step": 22150 }, { "epoch": 4.22, "learning_rate": 1.7358817921830315e-06, "loss": 2.072, "step": 22151 }, { "epoch": 4.22, "learning_rate": 1.7358245948522401e-06, "loss": 1.3101, "step": 22152 }, { "epoch": 4.22, "learning_rate": 1.7357673975214491e-06, "loss": 2.5968, "step": 22153 }, { "epoch": 4.22, "learning_rate": 1.735710200190658e-06, "loss": 2.0855, "step": 22154 }, { "epoch": 4.22, "learning_rate": 1.7356530028598665e-06, "loss": 2.6848, "step": 22155 }, { "epoch": 4.22, "learning_rate": 1.7355958055290755e-06, "loss": 2.133, "step": 22156 }, { "epoch": 4.22, "learning_rate": 1.735538608198284e-06, "loss": 1.2779, "step": 22157 }, { "epoch": 4.22, "learning_rate": 1.7354814108674929e-06, "loss": 3.605, "step": 22158 }, { "epoch": 4.22, "learning_rate": 1.7354242135367019e-06, "loss": 1.7162, "step": 22159 }, { "epoch": 4.22, "learning_rate": 1.7353670162059105e-06, "loss": 2.9674, "step": 22160 }, { "epoch": 4.22, "learning_rate": 1.735309818875119e-06, "loss": 1.342, "step": 22161 }, { "epoch": 4.22, "learning_rate": 1.735252621544328e-06, "loss": 2.4146, "step": 22162 }, { "epoch": 4.22, "learning_rate": 1.7351954242135366e-06, "loss": 1.4186, "step": 22163 }, { "epoch": 4.22, "learning_rate": 1.7351382268827454e-06, "loss": 1.7661, "step": 22164 }, { "epoch": 4.22, "learning_rate": 1.7350810295519544e-06, "loss": 2.2789, "step": 22165 }, { "epoch": 4.22, "learning_rate": 1.735023832221163e-06, "loss": 0.8329, "step": 22166 }, { "epoch": 4.22, "learning_rate": 1.7349666348903716e-06, "loss": 2.3347, "step": 22167 }, { "epoch": 4.22, "learning_rate": 1.7349094375595806e-06, "loss": 2.1034, "step": 22168 }, { "epoch": 4.22, "learning_rate": 1.7348522402287894e-06, "loss": 2.295, "step": 22169 }, { "epoch": 4.22, "learning_rate": 1.7347950428979982e-06, "loss": 2.1695, "step": 22170 }, { "epoch": 4.22, "learning_rate": 1.734737845567207e-06, "loss": 1.6677, "step": 22171 }, { "epoch": 4.22, "learning_rate": 1.7346806482364156e-06, "loss": 2.1069, "step": 22172 }, { "epoch": 4.22, "learning_rate": 1.7346234509056246e-06, "loss": 1.7087, "step": 22173 }, { "epoch": 4.22, "learning_rate": 1.7345662535748332e-06, "loss": 1.4597, "step": 22174 }, { "epoch": 4.22, "learning_rate": 1.734509056244042e-06, "loss": 2.8208, "step": 22175 }, { "epoch": 4.22, "learning_rate": 1.734451858913251e-06, "loss": 1.6464, "step": 22176 }, { "epoch": 4.22, "learning_rate": 1.7343946615824595e-06, "loss": 2.1732, "step": 22177 }, { "epoch": 4.22, "learning_rate": 1.7343374642516681e-06, "loss": 1.702, "step": 22178 }, { "epoch": 4.22, "learning_rate": 1.7342802669208771e-06, "loss": 3.0205, "step": 22179 }, { "epoch": 4.22, "learning_rate": 1.7342230695900857e-06, "loss": 1.8115, "step": 22180 }, { "epoch": 4.22, "learning_rate": 1.7341658722592945e-06, "loss": 1.3077, "step": 22181 }, { "epoch": 4.23, "learning_rate": 1.7341086749285035e-06, "loss": 2.4171, "step": 22182 }, { "epoch": 4.23, "learning_rate": 1.734051477597712e-06, "loss": 2.2023, "step": 22183 }, { "epoch": 4.23, "learning_rate": 1.733994280266921e-06, "loss": 1.9526, "step": 22184 }, { "epoch": 4.23, "learning_rate": 1.7339370829361297e-06, "loss": 1.5285, "step": 22185 }, { "epoch": 4.23, "learning_rate": 1.7338798856053385e-06, "loss": 1.4055, "step": 22186 }, { "epoch": 4.23, "learning_rate": 1.7338226882745475e-06, "loss": 2.5354, "step": 22187 }, { "epoch": 4.23, "learning_rate": 1.733765490943756e-06, "loss": 1.7949, "step": 22188 }, { "epoch": 4.23, "learning_rate": 1.7337082936129646e-06, "loss": 2.1145, "step": 22189 }, { "epoch": 4.23, "learning_rate": 1.7336510962821736e-06, "loss": 2.6814, "step": 22190 }, { "epoch": 4.23, "learning_rate": 1.7335938989513822e-06, "loss": 2.0647, "step": 22191 }, { "epoch": 4.23, "learning_rate": 1.733536701620591e-06, "loss": 2.0299, "step": 22192 }, { "epoch": 4.23, "learning_rate": 1.7334795042898e-06, "loss": 2.531, "step": 22193 }, { "epoch": 4.23, "learning_rate": 1.7334223069590086e-06, "loss": 2.9223, "step": 22194 }, { "epoch": 4.23, "learning_rate": 1.7333651096282172e-06, "loss": 2.1171, "step": 22195 }, { "epoch": 4.23, "learning_rate": 1.7333079122974262e-06, "loss": 3.2105, "step": 22196 }, { "epoch": 4.23, "learning_rate": 1.733250714966635e-06, "loss": 3.7342, "step": 22197 }, { "epoch": 4.23, "learning_rate": 1.7331935176358436e-06, "loss": 2.4203, "step": 22198 }, { "epoch": 4.23, "learning_rate": 1.7331363203050526e-06, "loss": 2.3058, "step": 22199 }, { "epoch": 4.23, "learning_rate": 1.7330791229742612e-06, "loss": 1.9294, "step": 22200 }, { "epoch": 4.23, "learning_rate": 1.7330219256434702e-06, "loss": 2.2418, "step": 22201 }, { "epoch": 4.23, "learning_rate": 1.7329647283126787e-06, "loss": 2.301, "step": 22202 }, { "epoch": 4.23, "learning_rate": 1.7329075309818875e-06, "loss": 3.0969, "step": 22203 }, { "epoch": 4.23, "learning_rate": 1.7328503336510965e-06, "loss": 2.1831, "step": 22204 }, { "epoch": 4.23, "learning_rate": 1.7327931363203051e-06, "loss": 2.1645, "step": 22205 }, { "epoch": 4.23, "learning_rate": 1.7327359389895137e-06, "loss": 1.2045, "step": 22206 }, { "epoch": 4.23, "learning_rate": 1.7326787416587227e-06, "loss": 1.7014, "step": 22207 }, { "epoch": 4.23, "learning_rate": 1.7326215443279315e-06, "loss": 3.1055, "step": 22208 }, { "epoch": 4.23, "learning_rate": 1.73256434699714e-06, "loss": 2.8191, "step": 22209 }, { "epoch": 4.23, "learning_rate": 1.732507149666349e-06, "loss": 1.6519, "step": 22210 }, { "epoch": 4.23, "learning_rate": 1.7324499523355577e-06, "loss": 2.4017, "step": 22211 }, { "epoch": 4.23, "learning_rate": 1.7323927550047663e-06, "loss": 2.0956, "step": 22212 }, { "epoch": 4.23, "learning_rate": 1.7323355576739753e-06, "loss": 1.4135, "step": 22213 }, { "epoch": 4.23, "learning_rate": 1.732278360343184e-06, "loss": 2.8013, "step": 22214 }, { "epoch": 4.23, "learning_rate": 1.7322211630123926e-06, "loss": 2.7207, "step": 22215 }, { "epoch": 4.23, "learning_rate": 1.7321639656816016e-06, "loss": 1.948, "step": 22216 }, { "epoch": 4.23, "learning_rate": 1.7321067683508102e-06, "loss": 1.2514, "step": 22217 }, { "epoch": 4.23, "learning_rate": 1.7320495710200192e-06, "loss": 1.8269, "step": 22218 }, { "epoch": 4.23, "learning_rate": 1.7319923736892278e-06, "loss": 2.3357, "step": 22219 }, { "epoch": 4.23, "learning_rate": 1.7319351763584366e-06, "loss": 2.405, "step": 22220 }, { "epoch": 4.23, "learning_rate": 1.7318779790276456e-06, "loss": 2.7924, "step": 22221 }, { "epoch": 4.23, "learning_rate": 1.7318207816968542e-06, "loss": 2.9686, "step": 22222 }, { "epoch": 4.23, "learning_rate": 1.7317635843660628e-06, "loss": 2.0362, "step": 22223 }, { "epoch": 4.23, "learning_rate": 1.7317063870352718e-06, "loss": 2.3573, "step": 22224 }, { "epoch": 4.23, "learning_rate": 1.7316491897044806e-06, "loss": 2.1902, "step": 22225 }, { "epoch": 4.23, "learning_rate": 1.7315919923736892e-06, "loss": 1.9432, "step": 22226 }, { "epoch": 4.23, "learning_rate": 1.7315347950428982e-06, "loss": 2.2707, "step": 22227 }, { "epoch": 4.23, "learning_rate": 1.7314775977121067e-06, "loss": 3.296, "step": 22228 }, { "epoch": 4.23, "learning_rate": 1.7314204003813155e-06, "loss": 2.7979, "step": 22229 }, { "epoch": 4.23, "learning_rate": 1.7313632030505243e-06, "loss": 2.6046, "step": 22230 }, { "epoch": 4.23, "learning_rate": 1.7313060057197331e-06, "loss": 2.2505, "step": 22231 }, { "epoch": 4.23, "learning_rate": 1.7312488083889417e-06, "loss": 1.9804, "step": 22232 }, { "epoch": 4.23, "learning_rate": 1.7311916110581507e-06, "loss": 2.2691, "step": 22233 }, { "epoch": 4.24, "learning_rate": 1.7311344137273593e-06, "loss": 2.3027, "step": 22234 }, { "epoch": 4.24, "learning_rate": 1.7310772163965683e-06, "loss": 1.9426, "step": 22235 }, { "epoch": 4.24, "learning_rate": 1.731020019065777e-06, "loss": 2.3693, "step": 22236 }, { "epoch": 4.24, "learning_rate": 1.7309628217349857e-06, "loss": 1.5062, "step": 22237 }, { "epoch": 4.24, "learning_rate": 1.7309056244041947e-06, "loss": 2.1109, "step": 22238 }, { "epoch": 4.24, "learning_rate": 1.7308484270734033e-06, "loss": 2.2445, "step": 22239 }, { "epoch": 4.24, "learning_rate": 1.7307912297426119e-06, "loss": 2.0427, "step": 22240 }, { "epoch": 4.24, "learning_rate": 1.7307340324118209e-06, "loss": 2.0388, "step": 22241 }, { "epoch": 4.24, "learning_rate": 1.7306768350810297e-06, "loss": 1.644, "step": 22242 }, { "epoch": 4.24, "learning_rate": 1.7306196377502382e-06, "loss": 2.6291, "step": 22243 }, { "epoch": 4.24, "learning_rate": 1.7305624404194472e-06, "loss": 2.8275, "step": 22244 }, { "epoch": 4.24, "learning_rate": 1.7305052430886558e-06, "loss": 1.6576, "step": 22245 }, { "epoch": 4.24, "learning_rate": 1.7304480457578646e-06, "loss": 1.3435, "step": 22246 }, { "epoch": 4.24, "learning_rate": 1.7303908484270734e-06, "loss": 2.6971, "step": 22247 }, { "epoch": 4.24, "learning_rate": 1.7303336510962822e-06, "loss": 1.4674, "step": 22248 }, { "epoch": 4.24, "learning_rate": 1.7302764537654912e-06, "loss": 2.8519, "step": 22249 }, { "epoch": 4.24, "learning_rate": 1.7302192564346998e-06, "loss": 1.2857, "step": 22250 }, { "epoch": 4.24, "learning_rate": 1.7301620591039084e-06, "loss": 1.4289, "step": 22251 }, { "epoch": 4.24, "learning_rate": 1.7301048617731174e-06, "loss": 1.6787, "step": 22252 }, { "epoch": 4.24, "learning_rate": 1.7300476644423262e-06, "loss": 2.2744, "step": 22253 }, { "epoch": 4.24, "learning_rate": 1.7299904671115348e-06, "loss": 3.4699, "step": 22254 }, { "epoch": 4.24, "learning_rate": 1.7299332697807438e-06, "loss": 1.8893, "step": 22255 }, { "epoch": 4.24, "learning_rate": 1.7298760724499523e-06, "loss": 1.7724, "step": 22256 }, { "epoch": 4.24, "learning_rate": 1.7298188751191611e-06, "loss": 1.4449, "step": 22257 }, { "epoch": 4.24, "learning_rate": 1.72976167778837e-06, "loss": 1.8083, "step": 22258 }, { "epoch": 4.24, "learning_rate": 1.7297044804575787e-06, "loss": 1.2995, "step": 22259 }, { "epoch": 4.24, "learning_rate": 1.7296472831267873e-06, "loss": 0.9651, "step": 22260 }, { "epoch": 4.24, "learning_rate": 1.7295900857959963e-06, "loss": 1.6685, "step": 22261 }, { "epoch": 4.24, "learning_rate": 1.7295328884652049e-06, "loss": 2.2039, "step": 22262 }, { "epoch": 4.24, "learning_rate": 1.7294756911344137e-06, "loss": 1.5113, "step": 22263 }, { "epoch": 4.24, "learning_rate": 1.7294184938036227e-06, "loss": 1.6038, "step": 22264 }, { "epoch": 4.24, "learning_rate": 1.7293612964728313e-06, "loss": 2.4572, "step": 22265 }, { "epoch": 4.24, "learning_rate": 1.7293040991420403e-06, "loss": 3.5565, "step": 22266 }, { "epoch": 4.24, "learning_rate": 1.7292469018112489e-06, "loss": 1.1113, "step": 22267 }, { "epoch": 4.24, "learning_rate": 1.7291897044804574e-06, "loss": 3.4692, "step": 22268 }, { "epoch": 4.24, "learning_rate": 1.7291325071496664e-06, "loss": 1.774, "step": 22269 }, { "epoch": 4.24, "learning_rate": 1.7290753098188752e-06, "loss": 2.0471, "step": 22270 }, { "epoch": 4.24, "learning_rate": 1.7290181124880838e-06, "loss": 1.3538, "step": 22271 }, { "epoch": 4.24, "learning_rate": 1.7289609151572928e-06, "loss": 2.2872, "step": 22272 }, { "epoch": 4.24, "learning_rate": 1.7289037178265014e-06, "loss": 1.5404, "step": 22273 }, { "epoch": 4.24, "learning_rate": 1.7288465204957102e-06, "loss": 3.8755, "step": 22274 }, { "epoch": 4.24, "learning_rate": 1.7287893231649192e-06, "loss": 3.9776, "step": 22275 }, { "epoch": 4.24, "learning_rate": 1.7287321258341278e-06, "loss": 2.7379, "step": 22276 }, { "epoch": 4.24, "learning_rate": 1.7286749285033364e-06, "loss": 2.2726, "step": 22277 }, { "epoch": 4.24, "learning_rate": 1.7286177311725454e-06, "loss": 3.1248, "step": 22278 }, { "epoch": 4.24, "learning_rate": 1.728560533841754e-06, "loss": 1.5158, "step": 22279 }, { "epoch": 4.24, "learning_rate": 1.7285033365109628e-06, "loss": 1.738, "step": 22280 }, { "epoch": 4.24, "learning_rate": 1.7284461391801718e-06, "loss": 2.4904, "step": 22281 }, { "epoch": 4.24, "learning_rate": 1.7283889418493803e-06, "loss": 2.4759, "step": 22282 }, { "epoch": 4.24, "learning_rate": 1.7283317445185893e-06, "loss": 1.6602, "step": 22283 }, { "epoch": 4.24, "learning_rate": 1.728274547187798e-06, "loss": 2.7972, "step": 22284 }, { "epoch": 4.24, "learning_rate": 1.7282173498570067e-06, "loss": 2.6873, "step": 22285 }, { "epoch": 4.24, "learning_rate": 1.7281601525262155e-06, "loss": 1.29, "step": 22286 }, { "epoch": 4.25, "learning_rate": 1.7281029551954243e-06, "loss": 1.8828, "step": 22287 }, { "epoch": 4.25, "learning_rate": 1.728045757864633e-06, "loss": 2.2758, "step": 22288 }, { "epoch": 4.25, "learning_rate": 1.727988560533842e-06, "loss": 2.0386, "step": 22289 }, { "epoch": 4.25, "learning_rate": 1.7279313632030505e-06, "loss": 1.649, "step": 22290 }, { "epoch": 4.25, "learning_rate": 1.7278741658722593e-06, "loss": 2.1413, "step": 22291 }, { "epoch": 4.25, "learning_rate": 1.7278169685414683e-06, "loss": 1.8129, "step": 22292 }, { "epoch": 4.25, "learning_rate": 1.7277597712106769e-06, "loss": 2.1227, "step": 22293 }, { "epoch": 4.25, "learning_rate": 1.7277025738798854e-06, "loss": 1.9823, "step": 22294 }, { "epoch": 4.25, "learning_rate": 1.7276453765490945e-06, "loss": 2.7647, "step": 22295 }, { "epoch": 4.25, "learning_rate": 1.7275881792183032e-06, "loss": 3.0217, "step": 22296 }, { "epoch": 4.25, "learning_rate": 1.7275309818875118e-06, "loss": 1.8286, "step": 22297 }, { "epoch": 4.25, "learning_rate": 1.7274737845567208e-06, "loss": 2.4189, "step": 22298 }, { "epoch": 4.25, "learning_rate": 1.7274165872259294e-06, "loss": 1.476, "step": 22299 }, { "epoch": 4.25, "learning_rate": 1.7273593898951384e-06, "loss": 2.52, "step": 22300 }, { "epoch": 4.25, "learning_rate": 1.727302192564347e-06, "loss": 1.8833, "step": 22301 }, { "epoch": 4.25, "learning_rate": 1.7272449952335558e-06, "loss": 1.7437, "step": 22302 }, { "epoch": 4.25, "learning_rate": 1.7271877979027648e-06, "loss": 2.0888, "step": 22303 }, { "epoch": 4.25, "learning_rate": 1.7271306005719734e-06, "loss": 1.9777, "step": 22304 }, { "epoch": 4.25, "learning_rate": 1.727073403241182e-06, "loss": 1.8449, "step": 22305 }, { "epoch": 4.25, "learning_rate": 1.727016205910391e-06, "loss": 2.7809, "step": 22306 }, { "epoch": 4.25, "learning_rate": 1.7269590085795996e-06, "loss": 2.196, "step": 22307 }, { "epoch": 4.25, "learning_rate": 1.7269018112488083e-06, "loss": 2.6949, "step": 22308 }, { "epoch": 4.25, "learning_rate": 1.7268446139180174e-06, "loss": 2.2437, "step": 22309 }, { "epoch": 4.25, "learning_rate": 1.726787416587226e-06, "loss": 1.5396, "step": 22310 }, { "epoch": 4.25, "learning_rate": 1.7267302192564345e-06, "loss": 2.4499, "step": 22311 }, { "epoch": 4.25, "learning_rate": 1.7266730219256435e-06, "loss": 2.5415, "step": 22312 }, { "epoch": 4.25, "learning_rate": 1.7266158245948523e-06, "loss": 2.3836, "step": 22313 }, { "epoch": 4.25, "learning_rate": 1.7265586272640613e-06, "loss": 1.8359, "step": 22314 }, { "epoch": 4.25, "learning_rate": 1.72650142993327e-06, "loss": 1.9199, "step": 22315 }, { "epoch": 4.25, "learning_rate": 1.7264442326024785e-06, "loss": 2.3402, "step": 22316 }, { "epoch": 4.25, "learning_rate": 1.7263870352716875e-06, "loss": 2.0313, "step": 22317 }, { "epoch": 4.25, "learning_rate": 1.726329837940896e-06, "loss": 2.4668, "step": 22318 }, { "epoch": 4.25, "learning_rate": 1.7262726406101049e-06, "loss": 2.4364, "step": 22319 }, { "epoch": 4.25, "learning_rate": 1.7262154432793139e-06, "loss": 3.7877, "step": 22320 }, { "epoch": 4.25, "learning_rate": 1.7261582459485225e-06, "loss": 2.2968, "step": 22321 }, { "epoch": 4.25, "learning_rate": 1.726101048617731e-06, "loss": 2.4957, "step": 22322 }, { "epoch": 4.25, "learning_rate": 1.72604385128694e-06, "loss": 1.8592, "step": 22323 }, { "epoch": 4.25, "learning_rate": 1.7259866539561488e-06, "loss": 1.715, "step": 22324 }, { "epoch": 4.25, "learning_rate": 1.7259294566253574e-06, "loss": 1.8187, "step": 22325 }, { "epoch": 4.25, "learning_rate": 1.7258722592945664e-06, "loss": 2.307, "step": 22326 }, { "epoch": 4.25, "learning_rate": 1.725815061963775e-06, "loss": 1.6775, "step": 22327 }, { "epoch": 4.25, "learning_rate": 1.7257578646329836e-06, "loss": 2.5847, "step": 22328 }, { "epoch": 4.25, "learning_rate": 1.7257006673021926e-06, "loss": 2.061, "step": 22329 }, { "epoch": 4.25, "learning_rate": 1.7256434699714014e-06, "loss": 2.1062, "step": 22330 }, { "epoch": 4.25, "learning_rate": 1.7255862726406104e-06, "loss": 2.4633, "step": 22331 }, { "epoch": 4.25, "learning_rate": 1.725529075309819e-06, "loss": 2.4752, "step": 22332 }, { "epoch": 4.25, "learning_rate": 1.7254718779790276e-06, "loss": 2.2112, "step": 22333 }, { "epoch": 4.25, "learning_rate": 1.7254146806482366e-06, "loss": 2.7534, "step": 22334 }, { "epoch": 4.25, "learning_rate": 1.7253574833174451e-06, "loss": 3.0439, "step": 22335 }, { "epoch": 4.25, "learning_rate": 1.725300285986654e-06, "loss": 1.7414, "step": 22336 }, { "epoch": 4.25, "learning_rate": 1.725243088655863e-06, "loss": 2.6811, "step": 22337 }, { "epoch": 4.25, "learning_rate": 1.7251858913250715e-06, "loss": 2.9342, "step": 22338 }, { "epoch": 4.26, "learning_rate": 1.7251286939942801e-06, "loss": 1.8621, "step": 22339 }, { "epoch": 4.26, "learning_rate": 1.7250714966634891e-06, "loss": 2.2868, "step": 22340 }, { "epoch": 4.26, "learning_rate": 1.725014299332698e-06, "loss": 2.4359, "step": 22341 }, { "epoch": 4.26, "learning_rate": 1.7249571020019065e-06, "loss": 2.6084, "step": 22342 }, { "epoch": 4.26, "learning_rate": 1.7248999046711155e-06, "loss": 1.6385, "step": 22343 }, { "epoch": 4.26, "learning_rate": 1.724842707340324e-06, "loss": 3.2633, "step": 22344 }, { "epoch": 4.26, "learning_rate": 1.7247855100095329e-06, "loss": 1.4033, "step": 22345 }, { "epoch": 4.26, "learning_rate": 1.7247283126787417e-06, "loss": 3.1601, "step": 22346 }, { "epoch": 4.26, "learning_rate": 1.7246711153479505e-06, "loss": 2.0248, "step": 22347 }, { "epoch": 4.26, "learning_rate": 1.7246139180171595e-06, "loss": 1.9419, "step": 22348 }, { "epoch": 4.26, "learning_rate": 1.724556720686368e-06, "loss": 1.9098, "step": 22349 }, { "epoch": 4.26, "learning_rate": 1.7244995233555766e-06, "loss": 1.1056, "step": 22350 }, { "epoch": 4.26, "learning_rate": 1.7244423260247856e-06, "loss": 1.7685, "step": 22351 }, { "epoch": 4.26, "learning_rate": 1.7243851286939944e-06, "loss": 2.4874, "step": 22352 }, { "epoch": 4.26, "learning_rate": 1.724327931363203e-06, "loss": 1.78, "step": 22353 }, { "epoch": 4.26, "learning_rate": 1.724270734032412e-06, "loss": 2.1564, "step": 22354 }, { "epoch": 4.26, "learning_rate": 1.7242135367016206e-06, "loss": 2.1022, "step": 22355 }, { "epoch": 4.26, "learning_rate": 1.7241563393708292e-06, "loss": 2.5395, "step": 22356 }, { "epoch": 4.26, "learning_rate": 1.7240991420400382e-06, "loss": 2.2973, "step": 22357 }, { "epoch": 4.26, "learning_rate": 1.724041944709247e-06, "loss": 1.8206, "step": 22358 }, { "epoch": 4.26, "learning_rate": 1.7239847473784556e-06, "loss": 1.7855, "step": 22359 }, { "epoch": 4.26, "learning_rate": 1.7239275500476646e-06, "loss": 0.7797, "step": 22360 }, { "epoch": 4.26, "learning_rate": 1.7238703527168731e-06, "loss": 1.7373, "step": 22361 }, { "epoch": 4.26, "learning_rate": 1.723813155386082e-06, "loss": 2.6004, "step": 22362 }, { "epoch": 4.26, "learning_rate": 1.723755958055291e-06, "loss": 3.1567, "step": 22363 }, { "epoch": 4.26, "learning_rate": 1.7236987607244995e-06, "loss": 2.5075, "step": 22364 }, { "epoch": 4.26, "learning_rate": 1.7236415633937085e-06, "loss": 2.2103, "step": 22365 }, { "epoch": 4.26, "learning_rate": 1.7235843660629171e-06, "loss": 1.8801, "step": 22366 }, { "epoch": 4.26, "learning_rate": 1.7235271687321257e-06, "loss": 2.3687, "step": 22367 }, { "epoch": 4.26, "learning_rate": 1.7234699714013347e-06, "loss": 2.1299, "step": 22368 }, { "epoch": 4.26, "learning_rate": 1.7234127740705435e-06, "loss": 2.0873, "step": 22369 }, { "epoch": 4.26, "learning_rate": 1.723355576739752e-06, "loss": 2.7798, "step": 22370 }, { "epoch": 4.26, "learning_rate": 1.723298379408961e-06, "loss": 1.7897, "step": 22371 }, { "epoch": 4.26, "learning_rate": 1.7232411820781697e-06, "loss": 2.1083, "step": 22372 }, { "epoch": 4.26, "learning_rate": 1.7231839847473785e-06, "loss": 2.4637, "step": 22373 }, { "epoch": 4.26, "learning_rate": 1.7231267874165873e-06, "loss": 2.1536, "step": 22374 }, { "epoch": 4.26, "learning_rate": 1.723069590085796e-06, "loss": 2.5017, "step": 22375 }, { "epoch": 4.26, "learning_rate": 1.7230123927550046e-06, "loss": 1.1551, "step": 22376 }, { "epoch": 4.26, "learning_rate": 1.7229551954242136e-06, "loss": 2.0603, "step": 22377 }, { "epoch": 4.26, "learning_rate": 1.7228979980934222e-06, "loss": 2.0989, "step": 22378 }, { "epoch": 4.26, "learning_rate": 1.7228408007626312e-06, "loss": 1.211, "step": 22379 }, { "epoch": 4.26, "learning_rate": 1.72278360343184e-06, "loss": 2.4386, "step": 22380 }, { "epoch": 4.26, "learning_rate": 1.7227264061010486e-06, "loss": 1.6182, "step": 22381 }, { "epoch": 4.26, "learning_rate": 1.7226692087702576e-06, "loss": 1.0229, "step": 22382 }, { "epoch": 4.26, "learning_rate": 1.7226120114394662e-06, "loss": 2.1128, "step": 22383 }, { "epoch": 4.26, "learning_rate": 1.722554814108675e-06, "loss": 3.2132, "step": 22384 }, { "epoch": 4.26, "learning_rate": 1.7224976167778838e-06, "loss": 2.7127, "step": 22385 }, { "epoch": 4.26, "learning_rate": 1.7224404194470926e-06, "loss": 1.1655, "step": 22386 }, { "epoch": 4.26, "learning_rate": 1.7223832221163012e-06, "loss": 2.5045, "step": 22387 }, { "epoch": 4.26, "learning_rate": 1.7223260247855102e-06, "loss": 2.1412, "step": 22388 }, { "epoch": 4.26, "learning_rate": 1.7222688274547187e-06, "loss": 2.2143, "step": 22389 }, { "epoch": 4.26, "learning_rate": 1.7222116301239275e-06, "loss": 1.3308, "step": 22390 }, { "epoch": 4.26, "learning_rate": 1.7221544327931365e-06, "loss": 1.4769, "step": 22391 }, { "epoch": 4.27, "learning_rate": 1.7220972354623451e-06, "loss": 1.6489, "step": 22392 }, { "epoch": 4.27, "learning_rate": 1.7220400381315537e-06, "loss": 1.7943, "step": 22393 }, { "epoch": 4.27, "learning_rate": 1.7219828408007627e-06, "loss": 2.5495, "step": 22394 }, { "epoch": 4.27, "learning_rate": 1.7219256434699713e-06, "loss": 1.9903, "step": 22395 }, { "epoch": 4.27, "learning_rate": 1.7218684461391803e-06, "loss": 3.3555, "step": 22396 }, { "epoch": 4.27, "learning_rate": 1.721811248808389e-06, "loss": 2.066, "step": 22397 }, { "epoch": 4.27, "learning_rate": 1.7217540514775977e-06, "loss": 0.9814, "step": 22398 }, { "epoch": 4.27, "learning_rate": 1.7216968541468067e-06, "loss": 2.1849, "step": 22399 }, { "epoch": 4.27, "learning_rate": 1.7216396568160153e-06, "loss": 1.4569, "step": 22400 }, { "epoch": 4.27, "learning_rate": 1.721582459485224e-06, "loss": 2.2607, "step": 22401 }, { "epoch": 4.27, "learning_rate": 1.7215252621544328e-06, "loss": 2.0575, "step": 22402 }, { "epoch": 4.27, "learning_rate": 1.7214680648236416e-06, "loss": 1.012, "step": 22403 }, { "epoch": 4.27, "learning_rate": 1.7214108674928502e-06, "loss": 2.4514, "step": 22404 }, { "epoch": 4.27, "learning_rate": 1.7213536701620592e-06, "loss": 3.0347, "step": 22405 }, { "epoch": 4.27, "learning_rate": 1.7212964728312678e-06, "loss": 2.1964, "step": 22406 }, { "epoch": 4.27, "learning_rate": 1.7212392755004766e-06, "loss": 1.2246, "step": 22407 }, { "epoch": 4.27, "learning_rate": 1.7211820781696856e-06, "loss": 2.1265, "step": 22408 }, { "epoch": 4.27, "learning_rate": 1.7211248808388942e-06, "loss": 2.158, "step": 22409 }, { "epoch": 4.27, "learning_rate": 1.7210676835081028e-06, "loss": 2.1854, "step": 22410 }, { "epoch": 4.27, "learning_rate": 1.7210104861773118e-06, "loss": 2.0004, "step": 22411 }, { "epoch": 4.27, "learning_rate": 1.7209532888465206e-06, "loss": 2.11, "step": 22412 }, { "epoch": 4.27, "learning_rate": 1.7208960915157294e-06, "loss": 1.9901, "step": 22413 }, { "epoch": 4.27, "learning_rate": 1.7208388941849382e-06, "loss": 1.6685, "step": 22414 }, { "epoch": 4.27, "learning_rate": 1.7207816968541467e-06, "loss": 1.9274, "step": 22415 }, { "epoch": 4.27, "learning_rate": 1.7207244995233558e-06, "loss": 2.5876, "step": 22416 }, { "epoch": 4.27, "learning_rate": 1.7206673021925643e-06, "loss": 1.7601, "step": 22417 }, { "epoch": 4.27, "learning_rate": 1.7206101048617731e-06, "loss": 1.7193, "step": 22418 }, { "epoch": 4.27, "learning_rate": 1.7205529075309821e-06, "loss": 2.9534, "step": 22419 }, { "epoch": 4.27, "learning_rate": 1.7204957102001907e-06, "loss": 2.822, "step": 22420 }, { "epoch": 4.27, "learning_rate": 1.7204385128693993e-06, "loss": 1.203, "step": 22421 }, { "epoch": 4.27, "learning_rate": 1.7203813155386083e-06, "loss": 2.3215, "step": 22422 }, { "epoch": 4.27, "learning_rate": 1.7203241182078169e-06, "loss": 1.7294, "step": 22423 }, { "epoch": 4.27, "learning_rate": 1.7202669208770257e-06, "loss": 2.6252, "step": 22424 }, { "epoch": 4.27, "learning_rate": 1.7202097235462347e-06, "loss": 1.8327, "step": 22425 }, { "epoch": 4.27, "learning_rate": 1.7201525262154433e-06, "loss": 3.1104, "step": 22426 }, { "epoch": 4.27, "learning_rate": 1.7200953288846518e-06, "loss": 2.1217, "step": 22427 }, { "epoch": 4.27, "learning_rate": 1.7200381315538609e-06, "loss": 1.5421, "step": 22428 }, { "epoch": 4.27, "learning_rate": 1.7199809342230696e-06, "loss": 2.6519, "step": 22429 }, { "epoch": 4.27, "learning_rate": 1.7199237368922787e-06, "loss": 1.7992, "step": 22430 }, { "epoch": 4.27, "learning_rate": 1.7198665395614872e-06, "loss": 1.4563, "step": 22431 }, { "epoch": 4.27, "learning_rate": 1.7198093422306958e-06, "loss": 1.4074, "step": 22432 }, { "epoch": 4.27, "learning_rate": 1.7197521448999048e-06, "loss": 2.1624, "step": 22433 }, { "epoch": 4.27, "learning_rate": 1.7196949475691134e-06, "loss": 3.4164, "step": 22434 }, { "epoch": 4.27, "learning_rate": 1.7196377502383222e-06, "loss": 1.6607, "step": 22435 }, { "epoch": 4.27, "learning_rate": 1.7195805529075312e-06, "loss": 2.6232, "step": 22436 }, { "epoch": 4.27, "learning_rate": 1.7195233555767398e-06, "loss": 1.2084, "step": 22437 }, { "epoch": 4.27, "learning_rate": 1.7194661582459484e-06, "loss": 2.2329, "step": 22438 }, { "epoch": 4.27, "learning_rate": 1.7194089609151574e-06, "loss": 2.4, "step": 22439 }, { "epoch": 4.27, "learning_rate": 1.7193517635843662e-06, "loss": 1.5691, "step": 22440 }, { "epoch": 4.27, "learning_rate": 1.7192945662535747e-06, "loss": 0.922, "step": 22441 }, { "epoch": 4.27, "learning_rate": 1.7192373689227838e-06, "loss": 1.4723, "step": 22442 }, { "epoch": 4.27, "learning_rate": 1.7191801715919923e-06, "loss": 2.1555, "step": 22443 }, { "epoch": 4.28, "learning_rate": 1.7191229742612013e-06, "loss": 2.6132, "step": 22444 }, { "epoch": 4.28, "learning_rate": 1.71906577693041e-06, "loss": 3.2074, "step": 22445 }, { "epoch": 4.28, "learning_rate": 1.7190085795996187e-06, "loss": 2.5448, "step": 22446 }, { "epoch": 4.28, "learning_rate": 1.7189513822688277e-06, "loss": 1.8043, "step": 22447 }, { "epoch": 4.28, "learning_rate": 1.7188941849380363e-06, "loss": 1.8133, "step": 22448 }, { "epoch": 4.28, "learning_rate": 1.7188369876072449e-06, "loss": 1.1582, "step": 22449 }, { "epoch": 4.28, "learning_rate": 1.7187797902764539e-06, "loss": 2.8314, "step": 22450 }, { "epoch": 4.28, "learning_rate": 1.7187225929456627e-06, "loss": 2.6706, "step": 22451 }, { "epoch": 4.28, "learning_rate": 1.7186653956148713e-06, "loss": 2.7959, "step": 22452 }, { "epoch": 4.28, "learning_rate": 1.7186081982840803e-06, "loss": 1.7671, "step": 22453 }, { "epoch": 4.28, "learning_rate": 1.7185510009532889e-06, "loss": 2.8108, "step": 22454 }, { "epoch": 4.28, "learning_rate": 1.7184938036224974e-06, "loss": 2.2856, "step": 22455 }, { "epoch": 4.28, "learning_rate": 1.7184366062917064e-06, "loss": 2.0344, "step": 22456 }, { "epoch": 4.28, "learning_rate": 1.7183794089609152e-06, "loss": 2.0023, "step": 22457 }, { "epoch": 4.28, "learning_rate": 1.7183222116301238e-06, "loss": 2.1179, "step": 22458 }, { "epoch": 4.28, "learning_rate": 1.7182650142993328e-06, "loss": 2.103, "step": 22459 }, { "epoch": 4.28, "learning_rate": 1.7182078169685414e-06, "loss": 3.2973, "step": 22460 }, { "epoch": 4.28, "learning_rate": 1.7181506196377504e-06, "loss": 1.1386, "step": 22461 }, { "epoch": 4.28, "learning_rate": 1.718093422306959e-06, "loss": 3.3596, "step": 22462 }, { "epoch": 4.28, "learning_rate": 1.7180362249761678e-06, "loss": 2.0665, "step": 22463 }, { "epoch": 4.28, "learning_rate": 1.7179790276453768e-06, "loss": 2.7831, "step": 22464 }, { "epoch": 4.28, "learning_rate": 1.7179218303145854e-06, "loss": 1.8072, "step": 22465 }, { "epoch": 4.28, "learning_rate": 1.717864632983794e-06, "loss": 2.8221, "step": 22466 }, { "epoch": 4.28, "learning_rate": 1.717807435653003e-06, "loss": 2.1645, "step": 22467 }, { "epoch": 4.28, "learning_rate": 1.7177502383222118e-06, "loss": 2.9166, "step": 22468 }, { "epoch": 4.28, "learning_rate": 1.7176930409914203e-06, "loss": 1.9229, "step": 22469 }, { "epoch": 4.28, "learning_rate": 1.7176358436606293e-06, "loss": 3.0908, "step": 22470 }, { "epoch": 4.28, "learning_rate": 1.717578646329838e-06, "loss": 2.3862, "step": 22471 }, { "epoch": 4.28, "learning_rate": 1.7175214489990467e-06, "loss": 2.764, "step": 22472 }, { "epoch": 4.28, "learning_rate": 1.7174642516682555e-06, "loss": 2.9648, "step": 22473 }, { "epoch": 4.28, "learning_rate": 1.7174070543374643e-06, "loss": 2.8022, "step": 22474 }, { "epoch": 4.28, "learning_rate": 1.7173498570066729e-06, "loss": 1.8476, "step": 22475 }, { "epoch": 4.28, "learning_rate": 1.717292659675882e-06, "loss": 2.8159, "step": 22476 }, { "epoch": 4.28, "learning_rate": 1.7172354623450905e-06, "loss": 3.0199, "step": 22477 }, { "epoch": 4.28, "learning_rate": 1.7171782650142995e-06, "loss": 1.6794, "step": 22478 }, { "epoch": 4.28, "learning_rate": 1.7171210676835083e-06, "loss": 2.385, "step": 22479 }, { "epoch": 4.28, "learning_rate": 1.7170638703527169e-06, "loss": 1.7354, "step": 22480 }, { "epoch": 4.28, "learning_rate": 1.7170066730219259e-06, "loss": 1.298, "step": 22481 }, { "epoch": 4.28, "learning_rate": 1.7169494756911344e-06, "loss": 1.7845, "step": 22482 }, { "epoch": 4.28, "learning_rate": 1.716892278360343e-06, "loss": 3.1189, "step": 22483 }, { "epoch": 4.28, "learning_rate": 1.716835081029552e-06, "loss": 1.5289, "step": 22484 }, { "epoch": 4.28, "learning_rate": 1.7167778836987608e-06, "loss": 2.7631, "step": 22485 }, { "epoch": 4.28, "learning_rate": 1.7167206863679694e-06, "loss": 2.8616, "step": 22486 }, { "epoch": 4.28, "learning_rate": 1.7166634890371784e-06, "loss": 2.0303, "step": 22487 }, { "epoch": 4.28, "learning_rate": 1.716606291706387e-06, "loss": 2.2016, "step": 22488 }, { "epoch": 4.28, "learning_rate": 1.7165490943755958e-06, "loss": 0.7686, "step": 22489 }, { "epoch": 4.28, "learning_rate": 1.7164918970448046e-06, "loss": 2.0471, "step": 22490 }, { "epoch": 4.28, "learning_rate": 1.7164346997140134e-06, "loss": 3.6693, "step": 22491 }, { "epoch": 4.28, "learning_rate": 1.716377502383222e-06, "loss": 2.4186, "step": 22492 }, { "epoch": 4.28, "learning_rate": 1.716320305052431e-06, "loss": 2.4521, "step": 22493 }, { "epoch": 4.28, "learning_rate": 1.7162631077216396e-06, "loss": 1.9947, "step": 22494 }, { "epoch": 4.28, "learning_rate": 1.7162059103908486e-06, "loss": 1.1765, "step": 22495 }, { "epoch": 4.28, "learning_rate": 1.7161487130600574e-06, "loss": 2.7117, "step": 22496 }, { "epoch": 4.29, "learning_rate": 1.716091515729266e-06, "loss": 1.3407, "step": 22497 }, { "epoch": 4.29, "learning_rate": 1.716034318398475e-06, "loss": 0.9758, "step": 22498 }, { "epoch": 4.29, "learning_rate": 1.7159771210676835e-06, "loss": 1.8235, "step": 22499 }, { "epoch": 4.29, "learning_rate": 1.7159199237368923e-06, "loss": 1.6077, "step": 22500 }, { "epoch": 4.29, "learning_rate": 1.7158627264061011e-06, "loss": 2.9624, "step": 22501 }, { "epoch": 4.29, "learning_rate": 1.71580552907531e-06, "loss": 3.1085, "step": 22502 }, { "epoch": 4.29, "learning_rate": 1.7157483317445185e-06, "loss": 1.6, "step": 22503 }, { "epoch": 4.29, "learning_rate": 1.7156911344137275e-06, "loss": 1.4416, "step": 22504 }, { "epoch": 4.29, "learning_rate": 1.715633937082936e-06, "loss": 3.0113, "step": 22505 }, { "epoch": 4.29, "learning_rate": 1.7155767397521449e-06, "loss": 0.8432, "step": 22506 }, { "epoch": 4.29, "learning_rate": 1.7155195424213539e-06, "loss": 2.6865, "step": 22507 }, { "epoch": 4.29, "learning_rate": 1.7154623450905625e-06, "loss": 2.3136, "step": 22508 }, { "epoch": 4.29, "learning_rate": 1.7154051477597715e-06, "loss": 1.6169, "step": 22509 }, { "epoch": 4.29, "learning_rate": 1.71534795042898e-06, "loss": 2.201, "step": 22510 }, { "epoch": 4.29, "learning_rate": 1.7152907530981886e-06, "loss": 2.022, "step": 22511 }, { "epoch": 4.29, "learning_rate": 1.7152335557673976e-06, "loss": 1.4915, "step": 22512 }, { "epoch": 4.29, "learning_rate": 1.7151763584366064e-06, "loss": 1.77, "step": 22513 }, { "epoch": 4.29, "learning_rate": 1.715119161105815e-06, "loss": 2.3314, "step": 22514 }, { "epoch": 4.29, "learning_rate": 1.715061963775024e-06, "loss": 1.4525, "step": 22515 }, { "epoch": 4.29, "learning_rate": 1.7150047664442326e-06, "loss": 0.9083, "step": 22516 }, { "epoch": 4.29, "learning_rate": 1.7149475691134414e-06, "loss": 1.6803, "step": 22517 }, { "epoch": 4.29, "learning_rate": 1.7148903717826504e-06, "loss": 2.3316, "step": 22518 }, { "epoch": 4.29, "learning_rate": 1.714833174451859e-06, "loss": 2.6599, "step": 22519 }, { "epoch": 4.29, "learning_rate": 1.7147759771210676e-06, "loss": 2.1503, "step": 22520 }, { "epoch": 4.29, "learning_rate": 1.7147187797902766e-06, "loss": 1.8948, "step": 22521 }, { "epoch": 4.29, "learning_rate": 1.7146615824594851e-06, "loss": 1.4899, "step": 22522 }, { "epoch": 4.29, "learning_rate": 1.714604385128694e-06, "loss": 2.0268, "step": 22523 }, { "epoch": 4.29, "learning_rate": 1.714547187797903e-06, "loss": 2.5987, "step": 22524 }, { "epoch": 4.29, "learning_rate": 1.7144899904671115e-06, "loss": 1.9424, "step": 22525 }, { "epoch": 4.29, "learning_rate": 1.7144327931363205e-06, "loss": 2.7726, "step": 22526 }, { "epoch": 4.29, "learning_rate": 1.7143755958055291e-06, "loss": 3.3254, "step": 22527 }, { "epoch": 4.29, "learning_rate": 1.714318398474738e-06, "loss": 2.9771, "step": 22528 }, { "epoch": 4.29, "learning_rate": 1.7142612011439467e-06, "loss": 1.7546, "step": 22529 }, { "epoch": 4.29, "learning_rate": 1.7142040038131555e-06, "loss": 1.8862, "step": 22530 }, { "epoch": 4.29, "learning_rate": 1.714146806482364e-06, "loss": 2.4926, "step": 22531 }, { "epoch": 4.29, "learning_rate": 1.714089609151573e-06, "loss": 1.8824, "step": 22532 }, { "epoch": 4.29, "learning_rate": 1.7140324118207817e-06, "loss": 1.5944, "step": 22533 }, { "epoch": 4.29, "learning_rate": 1.7139752144899905e-06, "loss": 3.0278, "step": 22534 }, { "epoch": 4.29, "learning_rate": 1.7139180171591995e-06, "loss": 2.0065, "step": 22535 }, { "epoch": 4.29, "learning_rate": 1.713860819828408e-06, "loss": 2.5915, "step": 22536 }, { "epoch": 4.29, "learning_rate": 1.7138036224976166e-06, "loss": 1.7706, "step": 22537 }, { "epoch": 4.29, "learning_rate": 1.7137464251668256e-06, "loss": 2.001, "step": 22538 }, { "epoch": 4.29, "learning_rate": 1.7136892278360344e-06, "loss": 2.4103, "step": 22539 }, { "epoch": 4.29, "learning_rate": 1.713632030505243e-06, "loss": 3.3868, "step": 22540 }, { "epoch": 4.29, "learning_rate": 1.713574833174452e-06, "loss": 2.5166, "step": 22541 }, { "epoch": 4.29, "learning_rate": 1.7135176358436606e-06, "loss": 2.1665, "step": 22542 }, { "epoch": 4.29, "learning_rate": 1.7134604385128696e-06, "loss": 1.6481, "step": 22543 }, { "epoch": 4.29, "learning_rate": 1.7134032411820782e-06, "loss": 3.0915, "step": 22544 }, { "epoch": 4.29, "learning_rate": 1.713346043851287e-06, "loss": 3.2035, "step": 22545 }, { "epoch": 4.29, "learning_rate": 1.713288846520496e-06, "loss": 2.4633, "step": 22546 }, { "epoch": 4.29, "learning_rate": 1.7132316491897046e-06, "loss": 1.4518, "step": 22547 }, { "epoch": 4.29, "learning_rate": 1.7131744518589131e-06, "loss": 1.1369, "step": 22548 }, { "epoch": 4.3, "learning_rate": 1.7131172545281222e-06, "loss": 1.8069, "step": 22549 }, { "epoch": 4.3, "learning_rate": 1.7130600571973307e-06, "loss": 1.6975, "step": 22550 }, { "epoch": 4.3, "learning_rate": 1.7130028598665395e-06, "loss": 2.1007, "step": 22551 }, { "epoch": 4.3, "learning_rate": 1.7129456625357485e-06, "loss": 3.7057, "step": 22552 }, { "epoch": 4.3, "learning_rate": 1.7128884652049571e-06, "loss": 2.4139, "step": 22553 }, { "epoch": 4.3, "learning_rate": 1.7128312678741657e-06, "loss": 2.6718, "step": 22554 }, { "epoch": 4.3, "learning_rate": 1.7127740705433747e-06, "loss": 2.7419, "step": 22555 }, { "epoch": 4.3, "learning_rate": 1.7127168732125835e-06, "loss": 3.0744, "step": 22556 }, { "epoch": 4.3, "learning_rate": 1.712659675881792e-06, "loss": 2.8654, "step": 22557 }, { "epoch": 4.3, "learning_rate": 1.712602478551001e-06, "loss": 1.6694, "step": 22558 }, { "epoch": 4.3, "learning_rate": 1.7125452812202097e-06, "loss": 1.9483, "step": 22559 }, { "epoch": 4.3, "learning_rate": 1.7124880838894187e-06, "loss": 3.233, "step": 22560 }, { "epoch": 4.3, "learning_rate": 1.7124308865586273e-06, "loss": 3.0996, "step": 22561 }, { "epoch": 4.3, "learning_rate": 1.712373689227836e-06, "loss": 1.8879, "step": 22562 }, { "epoch": 4.3, "learning_rate": 1.712316491897045e-06, "loss": 2.2614, "step": 22563 }, { "epoch": 4.3, "learning_rate": 1.7122592945662536e-06, "loss": 1.5086, "step": 22564 }, { "epoch": 4.3, "learning_rate": 1.7122020972354622e-06, "loss": 1.9321, "step": 22565 }, { "epoch": 4.3, "learning_rate": 1.7121448999046712e-06, "loss": 2.6767, "step": 22566 }, { "epoch": 4.3, "learning_rate": 1.71208770257388e-06, "loss": 1.4101, "step": 22567 }, { "epoch": 4.3, "learning_rate": 1.7120305052430886e-06, "loss": 2.7196, "step": 22568 }, { "epoch": 4.3, "learning_rate": 1.7119733079122976e-06, "loss": 1.329, "step": 22569 }, { "epoch": 4.3, "learning_rate": 1.7119161105815062e-06, "loss": 2.3725, "step": 22570 }, { "epoch": 4.3, "learning_rate": 1.7118589132507148e-06, "loss": 3.0878, "step": 22571 }, { "epoch": 4.3, "learning_rate": 1.7118017159199238e-06, "loss": 2.2242, "step": 22572 }, { "epoch": 4.3, "learning_rate": 1.7117445185891326e-06, "loss": 2.4262, "step": 22573 }, { "epoch": 4.3, "learning_rate": 1.7116873212583416e-06, "loss": 1.8203, "step": 22574 }, { "epoch": 4.3, "learning_rate": 1.7116301239275502e-06, "loss": 3.2854, "step": 22575 }, { "epoch": 4.3, "learning_rate": 1.7115729265967587e-06, "loss": 1.2219, "step": 22576 }, { "epoch": 4.3, "learning_rate": 1.7115157292659677e-06, "loss": 2.0789, "step": 22577 }, { "epoch": 4.3, "learning_rate": 1.7114585319351763e-06, "loss": 1.2705, "step": 22578 }, { "epoch": 4.3, "learning_rate": 1.7114013346043851e-06, "loss": 2.1979, "step": 22579 }, { "epoch": 4.3, "learning_rate": 1.7113441372735941e-06, "loss": 1.6331, "step": 22580 }, { "epoch": 4.3, "learning_rate": 1.7112869399428027e-06, "loss": 1.9917, "step": 22581 }, { "epoch": 4.3, "learning_rate": 1.7112297426120113e-06, "loss": 3.1016, "step": 22582 }, { "epoch": 4.3, "learning_rate": 1.7111725452812203e-06, "loss": 2.6798, "step": 22583 }, { "epoch": 4.3, "learning_rate": 1.711115347950429e-06, "loss": 2.848, "step": 22584 }, { "epoch": 4.3, "learning_rate": 1.7110581506196377e-06, "loss": 2.1967, "step": 22585 }, { "epoch": 4.3, "learning_rate": 1.7110009532888467e-06, "loss": 2.3041, "step": 22586 }, { "epoch": 4.3, "learning_rate": 1.7109437559580553e-06, "loss": 2.7568, "step": 22587 }, { "epoch": 4.3, "learning_rate": 1.710886558627264e-06, "loss": 2.1098, "step": 22588 }, { "epoch": 4.3, "learning_rate": 1.7108293612964728e-06, "loss": 1.3773, "step": 22589 }, { "epoch": 4.3, "learning_rate": 1.7107721639656816e-06, "loss": 2.6839, "step": 22590 }, { "epoch": 4.3, "learning_rate": 1.7107149666348906e-06, "loss": 2.6462, "step": 22591 }, { "epoch": 4.3, "learning_rate": 1.7106577693040992e-06, "loss": 2.0792, "step": 22592 }, { "epoch": 4.3, "learning_rate": 1.7106005719733078e-06, "loss": 2.0819, "step": 22593 }, { "epoch": 4.3, "learning_rate": 1.7105433746425168e-06, "loss": 1.9012, "step": 22594 }, { "epoch": 4.3, "learning_rate": 1.7104861773117256e-06, "loss": 2.2555, "step": 22595 }, { "epoch": 4.3, "learning_rate": 1.7104289799809342e-06, "loss": 3.1834, "step": 22596 }, { "epoch": 4.3, "learning_rate": 1.7103717826501432e-06, "loss": 1.7445, "step": 22597 }, { "epoch": 4.3, "learning_rate": 1.7103145853193518e-06, "loss": 3.1884, "step": 22598 }, { "epoch": 4.3, "learning_rate": 1.7102573879885604e-06, "loss": 1.5461, "step": 22599 }, { "epoch": 4.3, "learning_rate": 1.7102001906577694e-06, "loss": 2.9306, "step": 22600 }, { "epoch": 4.3, "learning_rate": 1.7101429933269782e-06, "loss": 1.7553, "step": 22601 }, { "epoch": 4.31, "learning_rate": 1.7100857959961867e-06, "loss": 2.4708, "step": 22602 }, { "epoch": 4.31, "learning_rate": 1.7100285986653957e-06, "loss": 2.7774, "step": 22603 }, { "epoch": 4.31, "learning_rate": 1.7099714013346043e-06, "loss": 1.8499, "step": 22604 }, { "epoch": 4.31, "learning_rate": 1.7099142040038131e-06, "loss": 3.5525, "step": 22605 }, { "epoch": 4.31, "learning_rate": 1.7098570066730221e-06, "loss": 1.3973, "step": 22606 }, { "epoch": 4.31, "learning_rate": 1.7097998093422307e-06, "loss": 1.6077, "step": 22607 }, { "epoch": 4.31, "learning_rate": 1.7097426120114397e-06, "loss": 2.8717, "step": 22608 }, { "epoch": 4.31, "learning_rate": 1.7096854146806483e-06, "loss": 2.9368, "step": 22609 }, { "epoch": 4.31, "learning_rate": 1.7096282173498569e-06, "loss": 1.3388, "step": 22610 }, { "epoch": 4.31, "learning_rate": 1.7095710200190659e-06, "loss": 2.381, "step": 22611 }, { "epoch": 4.31, "learning_rate": 1.7095138226882747e-06, "loss": 1.6019, "step": 22612 }, { "epoch": 4.31, "learning_rate": 1.7094566253574833e-06, "loss": 1.9135, "step": 22613 }, { "epoch": 4.31, "learning_rate": 1.7093994280266923e-06, "loss": 1.5359, "step": 22614 }, { "epoch": 4.31, "learning_rate": 1.7093422306959009e-06, "loss": 1.5179, "step": 22615 }, { "epoch": 4.31, "learning_rate": 1.7092850333651096e-06, "loss": 1.4715, "step": 22616 }, { "epoch": 4.31, "learning_rate": 1.7092278360343184e-06, "loss": 2.3631, "step": 22617 }, { "epoch": 4.31, "learning_rate": 1.7091706387035272e-06, "loss": 2.0957, "step": 22618 }, { "epoch": 4.31, "learning_rate": 1.7091134413727358e-06, "loss": 2.2121, "step": 22619 }, { "epoch": 4.31, "learning_rate": 1.7090562440419448e-06, "loss": 1.2616, "step": 22620 }, { "epoch": 4.31, "learning_rate": 1.7089990467111534e-06, "loss": 1.9798, "step": 22621 }, { "epoch": 4.31, "learning_rate": 1.7089418493803622e-06, "loss": 2.8454, "step": 22622 }, { "epoch": 4.31, "learning_rate": 1.7088846520495712e-06, "loss": 2.9104, "step": 22623 }, { "epoch": 4.31, "learning_rate": 1.7088274547187798e-06, "loss": 1.0768, "step": 22624 }, { "epoch": 4.31, "learning_rate": 1.7087702573879888e-06, "loss": 2.3002, "step": 22625 }, { "epoch": 4.31, "learning_rate": 1.7087130600571974e-06, "loss": 2.1315, "step": 22626 }, { "epoch": 4.31, "learning_rate": 1.7086558627264062e-06, "loss": 2.2348, "step": 22627 }, { "epoch": 4.31, "learning_rate": 1.708598665395615e-06, "loss": 2.9239, "step": 22628 }, { "epoch": 4.31, "learning_rate": 1.7085414680648238e-06, "loss": 1.059, "step": 22629 }, { "epoch": 4.31, "learning_rate": 1.7084842707340323e-06, "loss": 4.1792, "step": 22630 }, { "epoch": 4.31, "learning_rate": 1.7084270734032413e-06, "loss": 2.4764, "step": 22631 }, { "epoch": 4.31, "learning_rate": 1.70836987607245e-06, "loss": 1.9561, "step": 22632 }, { "epoch": 4.31, "learning_rate": 1.7083126787416587e-06, "loss": 1.5316, "step": 22633 }, { "epoch": 4.31, "learning_rate": 1.7082554814108677e-06, "loss": 1.8919, "step": 22634 }, { "epoch": 4.31, "learning_rate": 1.7081982840800763e-06, "loss": 2.2615, "step": 22635 }, { "epoch": 4.31, "learning_rate": 1.7081410867492849e-06, "loss": 2.7174, "step": 22636 }, { "epoch": 4.31, "learning_rate": 1.7080838894184939e-06, "loss": 1.3141, "step": 22637 }, { "epoch": 4.31, "learning_rate": 1.7080266920877025e-06, "loss": 2.713, "step": 22638 }, { "epoch": 4.31, "learning_rate": 1.7079694947569115e-06, "loss": 1.2558, "step": 22639 }, { "epoch": 4.31, "learning_rate": 1.7079122974261203e-06, "loss": 2.5696, "step": 22640 }, { "epoch": 4.31, "learning_rate": 1.7078551000953289e-06, "loss": 0.7869, "step": 22641 }, { "epoch": 4.31, "learning_rate": 1.7077979027645379e-06, "loss": 3.2178, "step": 22642 }, { "epoch": 4.31, "learning_rate": 1.7077407054337464e-06, "loss": 2.6799, "step": 22643 }, { "epoch": 4.31, "learning_rate": 1.7076835081029552e-06, "loss": 1.725, "step": 22644 }, { "epoch": 4.31, "learning_rate": 1.707626310772164e-06, "loss": 2.2964, "step": 22645 }, { "epoch": 4.31, "learning_rate": 1.7075691134413728e-06, "loss": 1.2181, "step": 22646 }, { "epoch": 4.31, "learning_rate": 1.7075119161105814e-06, "loss": 2.6696, "step": 22647 }, { "epoch": 4.31, "learning_rate": 1.7074547187797904e-06, "loss": 3.2042, "step": 22648 }, { "epoch": 4.31, "learning_rate": 1.707397521448999e-06, "loss": 1.9463, "step": 22649 }, { "epoch": 4.31, "learning_rate": 1.7073403241182078e-06, "loss": 2.4005, "step": 22650 }, { "epoch": 4.31, "learning_rate": 1.7072831267874168e-06, "loss": 1.7474, "step": 22651 }, { "epoch": 4.31, "learning_rate": 1.7072259294566254e-06, "loss": 3.202, "step": 22652 }, { "epoch": 4.31, "learning_rate": 1.707168732125834e-06, "loss": 1.6847, "step": 22653 }, { "epoch": 4.32, "learning_rate": 1.707111534795043e-06, "loss": 2.2661, "step": 22654 }, { "epoch": 4.32, "learning_rate": 1.7070543374642518e-06, "loss": 1.9793, "step": 22655 }, { "epoch": 4.32, "learning_rate": 1.7069971401334605e-06, "loss": 2.7164, "step": 22656 }, { "epoch": 4.32, "learning_rate": 1.7069399428026693e-06, "loss": 1.4831, "step": 22657 }, { "epoch": 4.32, "learning_rate": 1.706882745471878e-06, "loss": 1.9021, "step": 22658 }, { "epoch": 4.32, "learning_rate": 1.706825548141087e-06, "loss": 2.22, "step": 22659 }, { "epoch": 4.32, "learning_rate": 1.7067683508102955e-06, "loss": 1.2358, "step": 22660 }, { "epoch": 4.32, "learning_rate": 1.7067111534795043e-06, "loss": 3.2219, "step": 22661 }, { "epoch": 4.32, "learning_rate": 1.7066539561487133e-06, "loss": 3.2606, "step": 22662 }, { "epoch": 4.32, "learning_rate": 1.706596758817922e-06, "loss": 2.5786, "step": 22663 }, { "epoch": 4.32, "learning_rate": 1.7065395614871305e-06, "loss": 1.8287, "step": 22664 }, { "epoch": 4.32, "learning_rate": 1.7064823641563395e-06, "loss": 1.4437, "step": 22665 }, { "epoch": 4.32, "learning_rate": 1.706425166825548e-06, "loss": 3.0012, "step": 22666 }, { "epoch": 4.32, "learning_rate": 1.7063679694947569e-06, "loss": 2.0569, "step": 22667 }, { "epoch": 4.32, "learning_rate": 1.7063107721639659e-06, "loss": 2.8387, "step": 22668 }, { "epoch": 4.32, "learning_rate": 1.7062535748331744e-06, "loss": 2.038, "step": 22669 }, { "epoch": 4.32, "learning_rate": 1.706196377502383e-06, "loss": 1.2206, "step": 22670 }, { "epoch": 4.32, "learning_rate": 1.706139180171592e-06, "loss": 1.4229, "step": 22671 }, { "epoch": 4.32, "learning_rate": 1.7060819828408008e-06, "loss": 2.6916, "step": 22672 }, { "epoch": 4.32, "learning_rate": 1.7060247855100098e-06, "loss": 2.6639, "step": 22673 }, { "epoch": 4.32, "learning_rate": 1.7059675881792184e-06, "loss": 2.6743, "step": 22674 }, { "epoch": 4.32, "learning_rate": 1.705910390848427e-06, "loss": 1.36, "step": 22675 }, { "epoch": 4.32, "learning_rate": 1.705853193517636e-06, "loss": 1.4878, "step": 22676 }, { "epoch": 4.32, "learning_rate": 1.7057959961868446e-06, "loss": 1.8331, "step": 22677 }, { "epoch": 4.32, "learning_rate": 1.7057387988560534e-06, "loss": 1.8207, "step": 22678 }, { "epoch": 4.32, "learning_rate": 1.7056816015252624e-06, "loss": 1.5135, "step": 22679 }, { "epoch": 4.32, "learning_rate": 1.705624404194471e-06, "loss": 3.0894, "step": 22680 }, { "epoch": 4.32, "learning_rate": 1.7055672068636795e-06, "loss": 1.5146, "step": 22681 }, { "epoch": 4.32, "learning_rate": 1.7055100095328886e-06, "loss": 2.1585, "step": 22682 }, { "epoch": 4.32, "learning_rate": 1.7054528122020973e-06, "loss": 2.0084, "step": 22683 }, { "epoch": 4.32, "learning_rate": 1.705395614871306e-06, "loss": 1.9176, "step": 22684 }, { "epoch": 4.32, "learning_rate": 1.705338417540515e-06, "loss": 3.0576, "step": 22685 }, { "epoch": 4.32, "learning_rate": 1.7052812202097235e-06, "loss": 2.8047, "step": 22686 }, { "epoch": 4.32, "learning_rate": 1.705224022878932e-06, "loss": 2.2863, "step": 22687 }, { "epoch": 4.32, "learning_rate": 1.705166825548141e-06, "loss": 1.5201, "step": 22688 }, { "epoch": 4.32, "learning_rate": 1.70510962821735e-06, "loss": 1.9875, "step": 22689 }, { "epoch": 4.32, "learning_rate": 1.705052430886559e-06, "loss": 1.834, "step": 22690 }, { "epoch": 4.32, "learning_rate": 1.7049952335557675e-06, "loss": 1.537, "step": 22691 }, { "epoch": 4.32, "learning_rate": 1.704938036224976e-06, "loss": 1.7641, "step": 22692 }, { "epoch": 4.32, "learning_rate": 1.704880838894185e-06, "loss": 1.9629, "step": 22693 }, { "epoch": 4.32, "learning_rate": 1.7048236415633939e-06, "loss": 2.002, "step": 22694 }, { "epoch": 4.32, "learning_rate": 1.7047664442326024e-06, "loss": 2.824, "step": 22695 }, { "epoch": 4.32, "learning_rate": 1.7047092469018115e-06, "loss": 1.7111, "step": 22696 }, { "epoch": 4.32, "learning_rate": 1.70465204957102e-06, "loss": 1.5012, "step": 22697 }, { "epoch": 4.32, "learning_rate": 1.7045948522402286e-06, "loss": 1.8076, "step": 22698 }, { "epoch": 4.32, "learning_rate": 1.7045376549094376e-06, "loss": 1.1088, "step": 22699 }, { "epoch": 4.32, "learning_rate": 1.7044804575786464e-06, "loss": 2.0008, "step": 22700 }, { "epoch": 4.32, "learning_rate": 1.704423260247855e-06, "loss": 2.4574, "step": 22701 }, { "epoch": 4.32, "learning_rate": 1.704366062917064e-06, "loss": 2.2391, "step": 22702 }, { "epoch": 4.32, "learning_rate": 1.7043088655862726e-06, "loss": 1.0128, "step": 22703 }, { "epoch": 4.32, "learning_rate": 1.7042516682554814e-06, "loss": 2.7746, "step": 22704 }, { "epoch": 4.32, "learning_rate": 1.7041944709246902e-06, "loss": 1.6837, "step": 22705 }, { "epoch": 4.32, "learning_rate": 1.704137273593899e-06, "loss": 2.1685, "step": 22706 }, { "epoch": 4.33, "learning_rate": 1.704080076263108e-06, "loss": 1.1337, "step": 22707 }, { "epoch": 4.33, "learning_rate": 1.7040228789323166e-06, "loss": 2.1563, "step": 22708 }, { "epoch": 4.33, "learning_rate": 1.7039656816015251e-06, "loss": 1.9372, "step": 22709 }, { "epoch": 4.33, "learning_rate": 1.7039084842707341e-06, "loss": 1.811, "step": 22710 }, { "epoch": 4.33, "learning_rate": 1.703851286939943e-06, "loss": 1.7734, "step": 22711 }, { "epoch": 4.33, "learning_rate": 1.7037940896091515e-06, "loss": 2.499, "step": 22712 }, { "epoch": 4.33, "learning_rate": 1.7037368922783605e-06, "loss": 2.4977, "step": 22713 }, { "epoch": 4.33, "learning_rate": 1.7036796949475691e-06, "loss": 1.7864, "step": 22714 }, { "epoch": 4.33, "learning_rate": 1.7036224976167777e-06, "loss": 2.5894, "step": 22715 }, { "epoch": 4.33, "learning_rate": 1.7035653002859867e-06, "loss": 2.1889, "step": 22716 }, { "epoch": 4.33, "learning_rate": 1.7035081029551955e-06, "loss": 3.6153, "step": 22717 }, { "epoch": 4.33, "learning_rate": 1.703450905624404e-06, "loss": 2.7056, "step": 22718 }, { "epoch": 4.33, "learning_rate": 1.703393708293613e-06, "loss": 2.2188, "step": 22719 }, { "epoch": 4.33, "learning_rate": 1.7033365109628217e-06, "loss": 2.1025, "step": 22720 }, { "epoch": 4.33, "learning_rate": 1.7032793136320307e-06, "loss": 2.476, "step": 22721 }, { "epoch": 4.33, "learning_rate": 1.7032221163012395e-06, "loss": 1.3945, "step": 22722 }, { "epoch": 4.33, "learning_rate": 1.703164918970448e-06, "loss": 1.5775, "step": 22723 }, { "epoch": 4.33, "learning_rate": 1.703107721639657e-06, "loss": 2.9833, "step": 22724 }, { "epoch": 4.33, "learning_rate": 1.7030505243088656e-06, "loss": 2.5155, "step": 22725 }, { "epoch": 4.33, "learning_rate": 1.7029933269780742e-06, "loss": 2.8923, "step": 22726 }, { "epoch": 4.33, "learning_rate": 1.7029361296472832e-06, "loss": 1.8519, "step": 22727 }, { "epoch": 4.33, "learning_rate": 1.702878932316492e-06, "loss": 3.4634, "step": 22728 }, { "epoch": 4.33, "learning_rate": 1.7028217349857006e-06, "loss": 2.3926, "step": 22729 }, { "epoch": 4.33, "learning_rate": 1.7027645376549096e-06, "loss": 1.3964, "step": 22730 }, { "epoch": 4.33, "learning_rate": 1.7027073403241182e-06, "loss": 1.3201, "step": 22731 }, { "epoch": 4.33, "learning_rate": 1.702650142993327e-06, "loss": 1.9646, "step": 22732 }, { "epoch": 4.33, "learning_rate": 1.7025929456625358e-06, "loss": 0.6733, "step": 22733 }, { "epoch": 4.33, "learning_rate": 1.7025357483317446e-06, "loss": 2.2343, "step": 22734 }, { "epoch": 4.33, "learning_rate": 1.7024785510009531e-06, "loss": 2.4625, "step": 22735 }, { "epoch": 4.33, "learning_rate": 1.7024213536701621e-06, "loss": 2.025, "step": 22736 }, { "epoch": 4.33, "learning_rate": 1.7023641563393707e-06, "loss": 1.3061, "step": 22737 }, { "epoch": 4.33, "learning_rate": 1.7023069590085797e-06, "loss": 2.6832, "step": 22738 }, { "epoch": 4.33, "learning_rate": 1.7022497616777885e-06, "loss": 2.3853, "step": 22739 }, { "epoch": 4.33, "learning_rate": 1.7021925643469971e-06, "loss": 2.1313, "step": 22740 }, { "epoch": 4.33, "learning_rate": 1.7021353670162061e-06, "loss": 1.2287, "step": 22741 }, { "epoch": 4.33, "learning_rate": 1.7020781696854147e-06, "loss": 1.6334, "step": 22742 }, { "epoch": 4.33, "learning_rate": 1.7020209723546235e-06, "loss": 3.7152, "step": 22743 }, { "epoch": 4.33, "learning_rate": 1.7019637750238323e-06, "loss": 2.091, "step": 22744 }, { "epoch": 4.33, "learning_rate": 1.701906577693041e-06, "loss": 2.1916, "step": 22745 }, { "epoch": 4.33, "learning_rate": 1.7018493803622497e-06, "loss": 1.9523, "step": 22746 }, { "epoch": 4.33, "learning_rate": 1.7017921830314587e-06, "loss": 2.849, "step": 22747 }, { "epoch": 4.33, "learning_rate": 1.7017349857006673e-06, "loss": 1.7738, "step": 22748 }, { "epoch": 4.33, "learning_rate": 1.701677788369876e-06, "loss": 1.4115, "step": 22749 }, { "epoch": 4.33, "learning_rate": 1.701620591039085e-06, "loss": 1.8218, "step": 22750 }, { "epoch": 4.33, "learning_rate": 1.7015633937082936e-06, "loss": 3.0774, "step": 22751 }, { "epoch": 4.33, "learning_rate": 1.7015061963775022e-06, "loss": 3.4467, "step": 22752 }, { "epoch": 4.33, "learning_rate": 1.7014489990467112e-06, "loss": 2.1736, "step": 22753 }, { "epoch": 4.33, "learning_rate": 1.7013918017159198e-06, "loss": 1.9177, "step": 22754 }, { "epoch": 4.33, "learning_rate": 1.7013346043851288e-06, "loss": 1.9017, "step": 22755 }, { "epoch": 4.33, "learning_rate": 1.7012774070543376e-06, "loss": 2.3333, "step": 22756 }, { "epoch": 4.33, "learning_rate": 1.7012202097235462e-06, "loss": 1.9595, "step": 22757 }, { "epoch": 4.33, "learning_rate": 1.7011630123927552e-06, "loss": 1.9056, "step": 22758 }, { "epoch": 4.34, "learning_rate": 1.7011058150619638e-06, "loss": 1.3509, "step": 22759 }, { "epoch": 4.34, "learning_rate": 1.7010486177311726e-06, "loss": 1.2048, "step": 22760 }, { "epoch": 4.34, "learning_rate": 1.7009914204003816e-06, "loss": 1.6947, "step": 22761 }, { "epoch": 4.34, "learning_rate": 1.7009342230695902e-06, "loss": 2.6023, "step": 22762 }, { "epoch": 4.34, "learning_rate": 1.7008770257387987e-06, "loss": 2.9481, "step": 22763 }, { "epoch": 4.34, "learning_rate": 1.7008198284080077e-06, "loss": 1.7134, "step": 22764 }, { "epoch": 4.34, "learning_rate": 1.7007626310772163e-06, "loss": 2.7081, "step": 22765 }, { "epoch": 4.34, "learning_rate": 1.7007054337464251e-06, "loss": 1.5791, "step": 22766 }, { "epoch": 4.34, "learning_rate": 1.7006482364156341e-06, "loss": 2.5494, "step": 22767 }, { "epoch": 4.34, "learning_rate": 1.7005910390848427e-06, "loss": 2.2519, "step": 22768 }, { "epoch": 4.34, "learning_rate": 1.7005338417540513e-06, "loss": 2.0625, "step": 22769 }, { "epoch": 4.34, "learning_rate": 1.7004766444232603e-06, "loss": 1.1866, "step": 22770 }, { "epoch": 4.34, "learning_rate": 1.700419447092469e-06, "loss": 2.2296, "step": 22771 }, { "epoch": 4.34, "learning_rate": 1.7003622497616779e-06, "loss": 3.3554, "step": 22772 }, { "epoch": 4.34, "learning_rate": 1.7003050524308867e-06, "loss": 1.8004, "step": 22773 }, { "epoch": 4.34, "learning_rate": 1.7002478551000953e-06, "loss": 2.8905, "step": 22774 }, { "epoch": 4.34, "learning_rate": 1.7001906577693043e-06, "loss": 1.5412, "step": 22775 }, { "epoch": 4.34, "learning_rate": 1.7001334604385128e-06, "loss": 2.2992, "step": 22776 }, { "epoch": 4.34, "learning_rate": 1.7000762631077216e-06, "loss": 1.4657, "step": 22777 }, { "epoch": 4.34, "learning_rate": 1.7000190657769306e-06, "loss": 1.2946, "step": 22778 }, { "epoch": 4.34, "learning_rate": 1.6999618684461392e-06, "loss": 1.4939, "step": 22779 }, { "epoch": 4.34, "learning_rate": 1.6999046711153478e-06, "loss": 1.9722, "step": 22780 }, { "epoch": 4.34, "learning_rate": 1.6998474737845568e-06, "loss": 2.3534, "step": 22781 }, { "epoch": 4.34, "learning_rate": 1.6997902764537656e-06, "loss": 1.5871, "step": 22782 }, { "epoch": 4.34, "learning_rate": 1.6997330791229742e-06, "loss": 1.7447, "step": 22783 }, { "epoch": 4.34, "learning_rate": 1.6996758817921832e-06, "loss": 2.4122, "step": 22784 }, { "epoch": 4.34, "learning_rate": 1.6996186844613918e-06, "loss": 2.3087, "step": 22785 }, { "epoch": 4.34, "learning_rate": 1.6995614871306008e-06, "loss": 1.8504, "step": 22786 }, { "epoch": 4.34, "learning_rate": 1.6995042897998094e-06, "loss": 2.8829, "step": 22787 }, { "epoch": 4.34, "learning_rate": 1.6994470924690182e-06, "loss": 2.6115, "step": 22788 }, { "epoch": 4.34, "learning_rate": 1.6993898951382272e-06, "loss": 1.3587, "step": 22789 }, { "epoch": 4.34, "learning_rate": 1.6993326978074357e-06, "loss": 2.4691, "step": 22790 }, { "epoch": 4.34, "learning_rate": 1.6992755004766443e-06, "loss": 2.8775, "step": 22791 }, { "epoch": 4.34, "learning_rate": 1.6992183031458533e-06, "loss": 2.614, "step": 22792 }, { "epoch": 4.34, "learning_rate": 1.699161105815062e-06, "loss": 2.091, "step": 22793 }, { "epoch": 4.34, "learning_rate": 1.6991039084842707e-06, "loss": 2.6077, "step": 22794 }, { "epoch": 4.34, "learning_rate": 1.6990467111534797e-06, "loss": 2.0931, "step": 22795 }, { "epoch": 4.34, "learning_rate": 1.6989895138226883e-06, "loss": 2.5919, "step": 22796 }, { "epoch": 4.34, "learning_rate": 1.6989323164918969e-06, "loss": 2.0422, "step": 22797 }, { "epoch": 4.34, "learning_rate": 1.6988751191611059e-06, "loss": 0.5343, "step": 22798 }, { "epoch": 4.34, "learning_rate": 1.6988179218303147e-06, "loss": 1.6201, "step": 22799 }, { "epoch": 4.34, "learning_rate": 1.6987607244995233e-06, "loss": 1.5925, "step": 22800 }, { "epoch": 4.34, "learning_rate": 1.6987035271687323e-06, "loss": 3.1972, "step": 22801 }, { "epoch": 4.34, "learning_rate": 1.6986463298379408e-06, "loss": 2.0217, "step": 22802 }, { "epoch": 4.34, "learning_rate": 1.6985891325071499e-06, "loss": 2.287, "step": 22803 }, { "epoch": 4.34, "learning_rate": 1.6985319351763584e-06, "loss": 3.0131, "step": 22804 }, { "epoch": 4.34, "learning_rate": 1.6984747378455672e-06, "loss": 1.8084, "step": 22805 }, { "epoch": 4.34, "learning_rate": 1.6984175405147762e-06, "loss": 1.4182, "step": 22806 }, { "epoch": 4.34, "learning_rate": 1.6983603431839848e-06, "loss": 1.7689, "step": 22807 }, { "epoch": 4.34, "learning_rate": 1.6983031458531934e-06, "loss": 1.5793, "step": 22808 }, { "epoch": 4.34, "learning_rate": 1.6982459485224024e-06, "loss": 1.8968, "step": 22809 }, { "epoch": 4.34, "learning_rate": 1.6981887511916112e-06, "loss": 1.655, "step": 22810 }, { "epoch": 4.34, "learning_rate": 1.6981315538608198e-06, "loss": 1.8682, "step": 22811 }, { "epoch": 4.35, "learning_rate": 1.6980743565300288e-06, "loss": 1.7119, "step": 22812 }, { "epoch": 4.35, "learning_rate": 1.6980171591992374e-06, "loss": 3.5115, "step": 22813 }, { "epoch": 4.35, "learning_rate": 1.697959961868446e-06, "loss": 1.6075, "step": 22814 }, { "epoch": 4.35, "learning_rate": 1.697902764537655e-06, "loss": 2.6019, "step": 22815 }, { "epoch": 4.35, "learning_rate": 1.6978455672068637e-06, "loss": 2.1396, "step": 22816 }, { "epoch": 4.35, "learning_rate": 1.6977883698760723e-06, "loss": 2.4467, "step": 22817 }, { "epoch": 4.35, "learning_rate": 1.6977311725452813e-06, "loss": 1.4904, "step": 22818 }, { "epoch": 4.35, "learning_rate": 1.69767397521449e-06, "loss": 1.5349, "step": 22819 }, { "epoch": 4.35, "learning_rate": 1.697616777883699e-06, "loss": 1.9925, "step": 22820 }, { "epoch": 4.35, "learning_rate": 1.6975595805529075e-06, "loss": 1.8431, "step": 22821 }, { "epoch": 4.35, "learning_rate": 1.6975023832221163e-06, "loss": 1.7212, "step": 22822 }, { "epoch": 4.35, "learning_rate": 1.6974451858913253e-06, "loss": 1.6469, "step": 22823 }, { "epoch": 4.35, "learning_rate": 1.6973879885605339e-06, "loss": 2.632, "step": 22824 }, { "epoch": 4.35, "learning_rate": 1.6973307912297425e-06, "loss": 1.8903, "step": 22825 }, { "epoch": 4.35, "learning_rate": 1.6972735938989515e-06, "loss": 2.4269, "step": 22826 }, { "epoch": 4.35, "learning_rate": 1.6972163965681603e-06, "loss": 2.5467, "step": 22827 }, { "epoch": 4.35, "learning_rate": 1.6971591992373689e-06, "loss": 1.1272, "step": 22828 }, { "epoch": 4.35, "learning_rate": 1.6971020019065779e-06, "loss": 2.8246, "step": 22829 }, { "epoch": 4.35, "learning_rate": 1.6970448045757864e-06, "loss": 1.817, "step": 22830 }, { "epoch": 4.35, "learning_rate": 1.6969876072449952e-06, "loss": 2.6987, "step": 22831 }, { "epoch": 4.35, "learning_rate": 1.696930409914204e-06, "loss": 1.2889, "step": 22832 }, { "epoch": 4.35, "learning_rate": 1.6968732125834128e-06, "loss": 3.7618, "step": 22833 }, { "epoch": 4.35, "learning_rate": 1.6968160152526214e-06, "loss": 2.0708, "step": 22834 }, { "epoch": 4.35, "learning_rate": 1.6967588179218304e-06, "loss": 3.6602, "step": 22835 }, { "epoch": 4.35, "learning_rate": 1.696701620591039e-06, "loss": 2.0292, "step": 22836 }, { "epoch": 4.35, "learning_rate": 1.696644423260248e-06, "loss": 1.2866, "step": 22837 }, { "epoch": 4.35, "learning_rate": 1.6965872259294568e-06, "loss": 1.2062, "step": 22838 }, { "epoch": 4.35, "learning_rate": 1.6965300285986654e-06, "loss": 1.4124, "step": 22839 }, { "epoch": 4.35, "learning_rate": 1.6964728312678744e-06, "loss": 2.655, "step": 22840 }, { "epoch": 4.35, "learning_rate": 1.696415633937083e-06, "loss": 2.8727, "step": 22841 }, { "epoch": 4.35, "learning_rate": 1.6963584366062915e-06, "loss": 2.3538, "step": 22842 }, { "epoch": 4.35, "learning_rate": 1.6963012392755005e-06, "loss": 1.9751, "step": 22843 }, { "epoch": 4.35, "learning_rate": 1.6962440419447093e-06, "loss": 2.4786, "step": 22844 }, { "epoch": 4.35, "learning_rate": 1.696186844613918e-06, "loss": 2.4946, "step": 22845 }, { "epoch": 4.35, "learning_rate": 1.696129647283127e-06, "loss": 1.1502, "step": 22846 }, { "epoch": 4.35, "learning_rate": 1.6960724499523355e-06, "loss": 2.7953, "step": 22847 }, { "epoch": 4.35, "learning_rate": 1.6960152526215443e-06, "loss": 1.4462, "step": 22848 }, { "epoch": 4.35, "learning_rate": 1.6959580552907533e-06, "loss": 2.3052, "step": 22849 }, { "epoch": 4.35, "learning_rate": 1.6959008579599619e-06, "loss": 2.0043, "step": 22850 }, { "epoch": 4.35, "learning_rate": 1.695843660629171e-06, "loss": 1.677, "step": 22851 }, { "epoch": 4.35, "learning_rate": 1.6957864632983795e-06, "loss": 2.3966, "step": 22852 }, { "epoch": 4.35, "learning_rate": 1.695729265967588e-06, "loss": 2.2696, "step": 22853 }, { "epoch": 4.35, "learning_rate": 1.695672068636797e-06, "loss": 2.2255, "step": 22854 }, { "epoch": 4.35, "learning_rate": 1.6956148713060059e-06, "loss": 2.4281, "step": 22855 }, { "epoch": 4.35, "learning_rate": 1.6955576739752144e-06, "loss": 2.1765, "step": 22856 }, { "epoch": 4.35, "learning_rate": 1.6955004766444234e-06, "loss": 2.7046, "step": 22857 }, { "epoch": 4.35, "learning_rate": 1.695443279313632e-06, "loss": 1.8964, "step": 22858 }, { "epoch": 4.35, "learning_rate": 1.6953860819828408e-06, "loss": 1.4617, "step": 22859 }, { "epoch": 4.35, "learning_rate": 1.6953288846520496e-06, "loss": 2.387, "step": 22860 }, { "epoch": 4.35, "learning_rate": 1.6952716873212584e-06, "loss": 2.3489, "step": 22861 }, { "epoch": 4.35, "learning_rate": 1.695214489990467e-06, "loss": 1.6741, "step": 22862 }, { "epoch": 4.35, "learning_rate": 1.695157292659676e-06, "loss": 1.6435, "step": 22863 }, { "epoch": 4.36, "learning_rate": 1.6951000953288846e-06, "loss": 2.7178, "step": 22864 }, { "epoch": 4.36, "learning_rate": 1.6950428979980934e-06, "loss": 1.7525, "step": 22865 }, { "epoch": 4.36, "learning_rate": 1.6949857006673024e-06, "loss": 1.4486, "step": 22866 }, { "epoch": 4.36, "learning_rate": 1.694928503336511e-06, "loss": 1.0468, "step": 22867 }, { "epoch": 4.36, "learning_rate": 1.69487130600572e-06, "loss": 1.7426, "step": 22868 }, { "epoch": 4.36, "learning_rate": 1.6948141086749286e-06, "loss": 0.9946, "step": 22869 }, { "epoch": 4.36, "learning_rate": 1.6947569113441371e-06, "loss": 3.2603, "step": 22870 }, { "epoch": 4.36, "learning_rate": 1.6946997140133461e-06, "loss": 1.6236, "step": 22871 }, { "epoch": 4.36, "learning_rate": 1.694642516682555e-06, "loss": 1.6904, "step": 22872 }, { "epoch": 4.36, "learning_rate": 1.6945853193517635e-06, "loss": 2.7439, "step": 22873 }, { "epoch": 4.36, "learning_rate": 1.6945281220209725e-06, "loss": 2.2499, "step": 22874 }, { "epoch": 4.36, "learning_rate": 1.694470924690181e-06, "loss": 1.3108, "step": 22875 }, { "epoch": 4.36, "learning_rate": 1.69441372735939e-06, "loss": 2.8961, "step": 22876 }, { "epoch": 4.36, "learning_rate": 1.694356530028599e-06, "loss": 2.0466, "step": 22877 }, { "epoch": 4.36, "learning_rate": 1.6942993326978075e-06, "loss": 2.4074, "step": 22878 }, { "epoch": 4.36, "learning_rate": 1.694242135367016e-06, "loss": 1.6844, "step": 22879 }, { "epoch": 4.36, "learning_rate": 1.694184938036225e-06, "loss": 1.6985, "step": 22880 }, { "epoch": 4.36, "learning_rate": 1.6941277407054337e-06, "loss": 2.9281, "step": 22881 }, { "epoch": 4.36, "learning_rate": 1.6940705433746424e-06, "loss": 3.1033, "step": 22882 }, { "epoch": 4.36, "learning_rate": 1.6940133460438515e-06, "loss": 2.3909, "step": 22883 }, { "epoch": 4.36, "learning_rate": 1.69395614871306e-06, "loss": 2.6734, "step": 22884 }, { "epoch": 4.36, "learning_rate": 1.693898951382269e-06, "loss": 2.0138, "step": 22885 }, { "epoch": 4.36, "learning_rate": 1.6938417540514776e-06, "loss": 3.5089, "step": 22886 }, { "epoch": 4.36, "learning_rate": 1.6937845567206864e-06, "loss": 2.6715, "step": 22887 }, { "epoch": 4.36, "learning_rate": 1.6937273593898952e-06, "loss": 1.7955, "step": 22888 }, { "epoch": 4.36, "learning_rate": 1.693670162059104e-06, "loss": 2.0949, "step": 22889 }, { "epoch": 4.36, "learning_rate": 1.6936129647283126e-06, "loss": 1.9933, "step": 22890 }, { "epoch": 4.36, "learning_rate": 1.6935557673975216e-06, "loss": 2.3232, "step": 22891 }, { "epoch": 4.36, "learning_rate": 1.6934985700667302e-06, "loss": 1.495, "step": 22892 }, { "epoch": 4.36, "learning_rate": 1.693441372735939e-06, "loss": 2.0157, "step": 22893 }, { "epoch": 4.36, "learning_rate": 1.693384175405148e-06, "loss": 3.0806, "step": 22894 }, { "epoch": 4.36, "learning_rate": 1.6933269780743566e-06, "loss": 3.0643, "step": 22895 }, { "epoch": 4.36, "learning_rate": 1.6932697807435651e-06, "loss": 2.8346, "step": 22896 }, { "epoch": 4.36, "learning_rate": 1.6932125834127741e-06, "loss": 1.958, "step": 22897 }, { "epoch": 4.36, "learning_rate": 1.693155386081983e-06, "loss": 2.4239, "step": 22898 }, { "epoch": 4.36, "learning_rate": 1.6930981887511915e-06, "loss": 2.0136, "step": 22899 }, { "epoch": 4.36, "learning_rate": 1.6930409914204005e-06, "loss": 1.9859, "step": 22900 }, { "epoch": 4.36, "learning_rate": 1.6929837940896091e-06, "loss": 1.4409, "step": 22901 }, { "epoch": 4.36, "learning_rate": 1.6929265967588181e-06, "loss": 2.8744, "step": 22902 }, { "epoch": 4.36, "learning_rate": 1.6928693994280267e-06, "loss": 2.4479, "step": 22903 }, { "epoch": 4.36, "learning_rate": 1.6928122020972355e-06, "loss": 1.8533, "step": 22904 }, { "epoch": 4.36, "learning_rate": 1.6927550047664445e-06, "loss": 1.8719, "step": 22905 }, { "epoch": 4.36, "learning_rate": 1.692697807435653e-06, "loss": 2.0201, "step": 22906 }, { "epoch": 4.36, "learning_rate": 1.6926406101048617e-06, "loss": 1.5141, "step": 22907 }, { "epoch": 4.36, "learning_rate": 1.6925834127740707e-06, "loss": 2.689, "step": 22908 }, { "epoch": 4.36, "learning_rate": 1.6925262154432792e-06, "loss": 1.3518, "step": 22909 }, { "epoch": 4.36, "learning_rate": 1.692469018112488e-06, "loss": 1.7573, "step": 22910 }, { "epoch": 4.36, "learning_rate": 1.692411820781697e-06, "loss": 2.3903, "step": 22911 }, { "epoch": 4.36, "learning_rate": 1.6923546234509056e-06, "loss": 3.368, "step": 22912 }, { "epoch": 4.36, "learning_rate": 1.6922974261201142e-06, "loss": 1.246, "step": 22913 }, { "epoch": 4.36, "learning_rate": 1.6922402287893232e-06, "loss": 1.9118, "step": 22914 }, { "epoch": 4.36, "learning_rate": 1.692183031458532e-06, "loss": 0.9751, "step": 22915 }, { "epoch": 4.36, "learning_rate": 1.692125834127741e-06, "loss": 1.835, "step": 22916 }, { "epoch": 4.37, "learning_rate": 1.6920686367969496e-06, "loss": 2.0392, "step": 22917 }, { "epoch": 4.37, "learning_rate": 1.6920114394661582e-06, "loss": 2.7221, "step": 22918 }, { "epoch": 4.37, "learning_rate": 1.6919542421353672e-06, "loss": 1.8842, "step": 22919 }, { "epoch": 4.37, "learning_rate": 1.6918970448045758e-06, "loss": 2.348, "step": 22920 }, { "epoch": 4.37, "learning_rate": 1.6918398474737846e-06, "loss": 2.4185, "step": 22921 }, { "epoch": 4.37, "learning_rate": 1.6917826501429936e-06, "loss": 1.8465, "step": 22922 }, { "epoch": 4.37, "learning_rate": 1.6917254528122021e-06, "loss": 1.4055, "step": 22923 }, { "epoch": 4.37, "learning_rate": 1.6916682554814107e-06, "loss": 2.9091, "step": 22924 }, { "epoch": 4.37, "learning_rate": 1.6916110581506197e-06, "loss": 2.3883, "step": 22925 }, { "epoch": 4.37, "learning_rate": 1.6915538608198285e-06, "loss": 2.567, "step": 22926 }, { "epoch": 4.37, "learning_rate": 1.6914966634890371e-06, "loss": 1.3206, "step": 22927 }, { "epoch": 4.37, "learning_rate": 1.6914394661582461e-06, "loss": 2.8013, "step": 22928 }, { "epoch": 4.37, "learning_rate": 1.6913822688274547e-06, "loss": 1.0668, "step": 22929 }, { "epoch": 4.37, "learning_rate": 1.6913250714966633e-06, "loss": 3.5998, "step": 22930 }, { "epoch": 4.37, "learning_rate": 1.6912678741658723e-06, "loss": 2.3582, "step": 22931 }, { "epoch": 4.37, "learning_rate": 1.691210676835081e-06, "loss": 1.4259, "step": 22932 }, { "epoch": 4.37, "learning_rate": 1.69115347950429e-06, "loss": 2.6161, "step": 22933 }, { "epoch": 4.37, "learning_rate": 1.6910962821734987e-06, "loss": 2.4887, "step": 22934 }, { "epoch": 4.37, "learning_rate": 1.6910390848427072e-06, "loss": 2.4167, "step": 22935 }, { "epoch": 4.37, "learning_rate": 1.6909818875119163e-06, "loss": 1.8194, "step": 22936 }, { "epoch": 4.37, "learning_rate": 1.690924690181125e-06, "loss": 1.9098, "step": 22937 }, { "epoch": 4.37, "learning_rate": 1.6908674928503336e-06, "loss": 1.3722, "step": 22938 }, { "epoch": 4.37, "learning_rate": 1.6908102955195426e-06, "loss": 1.5245, "step": 22939 }, { "epoch": 4.37, "learning_rate": 1.6907530981887512e-06, "loss": 1.7781, "step": 22940 }, { "epoch": 4.37, "learning_rate": 1.6906959008579598e-06, "loss": 1.9444, "step": 22941 }, { "epoch": 4.37, "learning_rate": 1.6906387035271688e-06, "loss": 1.2663, "step": 22942 }, { "epoch": 4.37, "learning_rate": 1.6905815061963776e-06, "loss": 1.3193, "step": 22943 }, { "epoch": 4.37, "learning_rate": 1.6905243088655862e-06, "loss": 2.3578, "step": 22944 }, { "epoch": 4.37, "learning_rate": 1.6904671115347952e-06, "loss": 2.083, "step": 22945 }, { "epoch": 4.37, "learning_rate": 1.6904099142040038e-06, "loss": 2.1127, "step": 22946 }, { "epoch": 4.37, "learning_rate": 1.6903527168732126e-06, "loss": 1.3587, "step": 22947 }, { "epoch": 4.37, "learning_rate": 1.6902955195424214e-06, "loss": 1.5682, "step": 22948 }, { "epoch": 4.37, "learning_rate": 1.6902383222116302e-06, "loss": 2.5585, "step": 22949 }, { "epoch": 4.37, "learning_rate": 1.6901811248808392e-06, "loss": 2.8919, "step": 22950 }, { "epoch": 4.37, "learning_rate": 1.6901239275500477e-06, "loss": 1.5293, "step": 22951 }, { "epoch": 4.37, "learning_rate": 1.6900667302192563e-06, "loss": 2.3476, "step": 22952 }, { "epoch": 4.37, "learning_rate": 1.6900095328884653e-06, "loss": 2.3589, "step": 22953 }, { "epoch": 4.37, "learning_rate": 1.6899523355576741e-06, "loss": 2.152, "step": 22954 }, { "epoch": 4.37, "learning_rate": 1.6898951382268827e-06, "loss": 1.4082, "step": 22955 }, { "epoch": 4.37, "learning_rate": 1.6898379408960917e-06, "loss": 2.087, "step": 22956 }, { "epoch": 4.37, "learning_rate": 1.6897807435653003e-06, "loss": 2.1351, "step": 22957 }, { "epoch": 4.37, "learning_rate": 1.6897235462345089e-06, "loss": 2.2916, "step": 22958 }, { "epoch": 4.37, "learning_rate": 1.6896663489037179e-06, "loss": 1.005, "step": 22959 }, { "epoch": 4.37, "learning_rate": 1.6896091515729267e-06, "loss": 1.4325, "step": 22960 }, { "epoch": 4.37, "learning_rate": 1.6895519542421353e-06, "loss": 2.7952, "step": 22961 }, { "epoch": 4.37, "learning_rate": 1.6894947569113443e-06, "loss": 2.525, "step": 22962 }, { "epoch": 4.37, "learning_rate": 1.6894375595805528e-06, "loss": 1.9675, "step": 22963 }, { "epoch": 4.37, "learning_rate": 1.6893803622497616e-06, "loss": 2.3752, "step": 22964 }, { "epoch": 4.37, "learning_rate": 1.6893231649189706e-06, "loss": 2.2173, "step": 22965 }, { "epoch": 4.37, "learning_rate": 1.6892659675881792e-06, "loss": 2.3838, "step": 22966 }, { "epoch": 4.37, "learning_rate": 1.6892087702573882e-06, "loss": 1.9937, "step": 22967 }, { "epoch": 4.37, "learning_rate": 1.6891515729265968e-06, "loss": 1.9822, "step": 22968 }, { "epoch": 4.38, "learning_rate": 1.6890943755958054e-06, "loss": 2.1333, "step": 22969 }, { "epoch": 4.38, "learning_rate": 1.6890371782650144e-06, "loss": 2.1861, "step": 22970 }, { "epoch": 4.38, "learning_rate": 1.6889799809342232e-06, "loss": 1.7077, "step": 22971 }, { "epoch": 4.38, "learning_rate": 1.6889227836034318e-06, "loss": 2.217, "step": 22972 }, { "epoch": 4.38, "learning_rate": 1.6888655862726408e-06, "loss": 2.3673, "step": 22973 }, { "epoch": 4.38, "learning_rate": 1.6888083889418494e-06, "loss": 1.0001, "step": 22974 }, { "epoch": 4.38, "learning_rate": 1.6887511916110582e-06, "loss": 3.2617, "step": 22975 }, { "epoch": 4.38, "learning_rate": 1.688693994280267e-06, "loss": 3.4992, "step": 22976 }, { "epoch": 4.38, "learning_rate": 1.6886367969494757e-06, "loss": 0.7492, "step": 22977 }, { "epoch": 4.38, "learning_rate": 1.6885795996186843e-06, "loss": 1.1497, "step": 22978 }, { "epoch": 4.38, "learning_rate": 1.6885224022878933e-06, "loss": 1.686, "step": 22979 }, { "epoch": 4.38, "learning_rate": 1.688465204957102e-06, "loss": 0.956, "step": 22980 }, { "epoch": 4.38, "learning_rate": 1.688408007626311e-06, "loss": 1.6182, "step": 22981 }, { "epoch": 4.38, "learning_rate": 1.6883508102955197e-06, "loss": 2.4705, "step": 22982 }, { "epoch": 4.38, "learning_rate": 1.6882936129647283e-06, "loss": 1.2796, "step": 22983 }, { "epoch": 4.38, "learning_rate": 1.6882364156339373e-06, "loss": 3.5282, "step": 22984 }, { "epoch": 4.38, "learning_rate": 1.6881792183031459e-06, "loss": 1.4233, "step": 22985 }, { "epoch": 4.38, "learning_rate": 1.6881220209723547e-06, "loss": 1.8651, "step": 22986 }, { "epoch": 4.38, "learning_rate": 1.6880648236415635e-06, "loss": 2.6541, "step": 22987 }, { "epoch": 4.38, "learning_rate": 1.6880076263107723e-06, "loss": 2.0098, "step": 22988 }, { "epoch": 4.38, "learning_rate": 1.6879504289799808e-06, "loss": 2.0932, "step": 22989 }, { "epoch": 4.38, "learning_rate": 1.6878932316491898e-06, "loss": 1.8494, "step": 22990 }, { "epoch": 4.38, "learning_rate": 1.6878360343183984e-06, "loss": 2.0914, "step": 22991 }, { "epoch": 4.38, "learning_rate": 1.6877788369876072e-06, "loss": 2.3566, "step": 22992 }, { "epoch": 4.38, "learning_rate": 1.6877216396568162e-06, "loss": 2.6306, "step": 22993 }, { "epoch": 4.38, "learning_rate": 1.6876644423260248e-06, "loss": 2.2971, "step": 22994 }, { "epoch": 4.38, "learning_rate": 1.6876072449952334e-06, "loss": 1.6274, "step": 22995 }, { "epoch": 4.38, "learning_rate": 1.6875500476644424e-06, "loss": 3.2022, "step": 22996 }, { "epoch": 4.38, "learning_rate": 1.687492850333651e-06, "loss": 1.4505, "step": 22997 }, { "epoch": 4.38, "learning_rate": 1.68743565300286e-06, "loss": 2.3018, "step": 22998 }, { "epoch": 4.38, "learning_rate": 1.6873784556720688e-06, "loss": 2.6427, "step": 22999 }, { "epoch": 4.38, "learning_rate": 1.6873212583412774e-06, "loss": 2.6914, "step": 23000 }, { "epoch": 4.38, "learning_rate": 1.6872640610104864e-06, "loss": 2.0953, "step": 23001 }, { "epoch": 4.38, "learning_rate": 1.687206863679695e-06, "loss": 1.7221, "step": 23002 }, { "epoch": 4.38, "learning_rate": 1.6871496663489037e-06, "loss": 2.3842, "step": 23003 }, { "epoch": 4.38, "learning_rate": 1.6870924690181128e-06, "loss": 3.6851, "step": 23004 }, { "epoch": 4.38, "learning_rate": 1.6870352716873213e-06, "loss": 2.5307, "step": 23005 }, { "epoch": 4.38, "learning_rate": 1.68697807435653e-06, "loss": 2.6607, "step": 23006 }, { "epoch": 4.38, "learning_rate": 1.686920877025739e-06, "loss": 1.8484, "step": 23007 }, { "epoch": 4.38, "learning_rate": 1.6868636796949475e-06, "loss": 2.8727, "step": 23008 }, { "epoch": 4.38, "learning_rate": 1.6868064823641563e-06, "loss": 2.8352, "step": 23009 }, { "epoch": 4.38, "learning_rate": 1.6867492850333653e-06, "loss": 1.3512, "step": 23010 }, { "epoch": 4.38, "learning_rate": 1.6866920877025739e-06, "loss": 1.9696, "step": 23011 }, { "epoch": 4.38, "learning_rate": 1.6866348903717825e-06, "loss": 1.3156, "step": 23012 }, { "epoch": 4.38, "learning_rate": 1.6865776930409915e-06, "loss": 1.8718, "step": 23013 }, { "epoch": 4.38, "learning_rate": 1.6865204957102003e-06, "loss": 1.3749, "step": 23014 }, { "epoch": 4.38, "learning_rate": 1.686463298379409e-06, "loss": 1.9528, "step": 23015 }, { "epoch": 4.38, "learning_rate": 1.6864061010486179e-06, "loss": 2.7147, "step": 23016 }, { "epoch": 4.38, "learning_rate": 1.6863489037178264e-06, "loss": 2.7069, "step": 23017 }, { "epoch": 4.38, "learning_rate": 1.6862917063870354e-06, "loss": 2.1666, "step": 23018 }, { "epoch": 4.38, "learning_rate": 1.686234509056244e-06, "loss": 1.0338, "step": 23019 }, { "epoch": 4.38, "learning_rate": 1.6861773117254528e-06, "loss": 1.1525, "step": 23020 }, { "epoch": 4.38, "learning_rate": 1.6861201143946618e-06, "loss": 2.7476, "step": 23021 }, { "epoch": 4.39, "learning_rate": 1.6860629170638704e-06, "loss": 3.0267, "step": 23022 }, { "epoch": 4.39, "learning_rate": 1.686005719733079e-06, "loss": 1.7247, "step": 23023 }, { "epoch": 4.39, "learning_rate": 1.685948522402288e-06, "loss": 2.8352, "step": 23024 }, { "epoch": 4.39, "learning_rate": 1.6858913250714968e-06, "loss": 2.7581, "step": 23025 }, { "epoch": 4.39, "learning_rate": 1.6858341277407054e-06, "loss": 3.0599, "step": 23026 }, { "epoch": 4.39, "learning_rate": 1.6857769304099144e-06, "loss": 2.5043, "step": 23027 }, { "epoch": 4.39, "learning_rate": 1.685719733079123e-06, "loss": 2.6029, "step": 23028 }, { "epoch": 4.39, "learning_rate": 1.6856625357483315e-06, "loss": 2.9681, "step": 23029 }, { "epoch": 4.39, "learning_rate": 1.6856053384175405e-06, "loss": 3.1592, "step": 23030 }, { "epoch": 4.39, "learning_rate": 1.6855481410867493e-06, "loss": 1.6862, "step": 23031 }, { "epoch": 4.39, "learning_rate": 1.6854909437559583e-06, "loss": 3.4196, "step": 23032 }, { "epoch": 4.39, "learning_rate": 1.685433746425167e-06, "loss": 2.6077, "step": 23033 }, { "epoch": 4.39, "learning_rate": 1.6853765490943755e-06, "loss": 3.1845, "step": 23034 }, { "epoch": 4.39, "learning_rate": 1.6853193517635845e-06, "loss": 1.5639, "step": 23035 }, { "epoch": 4.39, "learning_rate": 1.685262154432793e-06, "loss": 1.5744, "step": 23036 }, { "epoch": 4.39, "learning_rate": 1.6852049571020019e-06, "loss": 1.6086, "step": 23037 }, { "epoch": 4.39, "learning_rate": 1.685147759771211e-06, "loss": 1.5155, "step": 23038 }, { "epoch": 4.39, "learning_rate": 1.6850905624404195e-06, "loss": 2.3316, "step": 23039 }, { "epoch": 4.39, "learning_rate": 1.685033365109628e-06, "loss": 1.0862, "step": 23040 }, { "epoch": 4.39, "learning_rate": 1.684976167778837e-06, "loss": 2.9014, "step": 23041 }, { "epoch": 4.39, "learning_rate": 1.6849189704480459e-06, "loss": 2.1125, "step": 23042 }, { "epoch": 4.39, "learning_rate": 1.6848617731172544e-06, "loss": 2.5945, "step": 23043 }, { "epoch": 4.39, "learning_rate": 1.6848045757864634e-06, "loss": 1.9867, "step": 23044 }, { "epoch": 4.39, "learning_rate": 1.684747378455672e-06, "loss": 2.288, "step": 23045 }, { "epoch": 4.39, "learning_rate": 1.684690181124881e-06, "loss": 3.128, "step": 23046 }, { "epoch": 4.39, "learning_rate": 1.6846329837940896e-06, "loss": 2.1768, "step": 23047 }, { "epoch": 4.39, "learning_rate": 1.6845757864632984e-06, "loss": 0.6604, "step": 23048 }, { "epoch": 4.39, "learning_rate": 1.6845185891325074e-06, "loss": 1.9721, "step": 23049 }, { "epoch": 4.39, "learning_rate": 1.684461391801716e-06, "loss": 2.9308, "step": 23050 }, { "epoch": 4.39, "learning_rate": 1.6844041944709246e-06, "loss": 1.5947, "step": 23051 }, { "epoch": 4.39, "learning_rate": 1.6843469971401336e-06, "loss": 1.404, "step": 23052 }, { "epoch": 4.39, "learning_rate": 1.6842897998093424e-06, "loss": 2.15, "step": 23053 }, { "epoch": 4.39, "learning_rate": 1.684232602478551e-06, "loss": 2.0093, "step": 23054 }, { "epoch": 4.39, "learning_rate": 1.68417540514776e-06, "loss": 1.873, "step": 23055 }, { "epoch": 4.39, "learning_rate": 1.6841182078169685e-06, "loss": 0.9554, "step": 23056 }, { "epoch": 4.39, "learning_rate": 1.6840610104861771e-06, "loss": 3.2219, "step": 23057 }, { "epoch": 4.39, "learning_rate": 1.6840038131553861e-06, "loss": 2.6052, "step": 23058 }, { "epoch": 4.39, "learning_rate": 1.683946615824595e-06, "loss": 2.5559, "step": 23059 }, { "epoch": 4.39, "learning_rate": 1.6838894184938035e-06, "loss": 2.9746, "step": 23060 }, { "epoch": 4.39, "learning_rate": 1.6838322211630125e-06, "loss": 2.4273, "step": 23061 }, { "epoch": 4.39, "learning_rate": 1.683775023832221e-06, "loss": 2.042, "step": 23062 }, { "epoch": 4.39, "learning_rate": 1.68371782650143e-06, "loss": 1.2269, "step": 23063 }, { "epoch": 4.39, "learning_rate": 1.6836606291706387e-06, "loss": 2.3595, "step": 23064 }, { "epoch": 4.39, "learning_rate": 1.6836034318398475e-06, "loss": 2.3284, "step": 23065 }, { "epoch": 4.39, "learning_rate": 1.6835462345090565e-06, "loss": 2.9822, "step": 23066 }, { "epoch": 4.39, "learning_rate": 1.683489037178265e-06, "loss": 2.2171, "step": 23067 }, { "epoch": 4.39, "learning_rate": 1.6834318398474737e-06, "loss": 1.7953, "step": 23068 }, { "epoch": 4.39, "learning_rate": 1.6833746425166827e-06, "loss": 1.502, "step": 23069 }, { "epoch": 4.39, "learning_rate": 1.6833174451858914e-06, "loss": 1.3681, "step": 23070 }, { "epoch": 4.39, "learning_rate": 1.6832602478551e-06, "loss": 1.8408, "step": 23071 }, { "epoch": 4.39, "learning_rate": 1.683203050524309e-06, "loss": 3.1813, "step": 23072 }, { "epoch": 4.39, "learning_rate": 1.6831458531935176e-06, "loss": 2.7969, "step": 23073 }, { "epoch": 4.4, "learning_rate": 1.6830886558627264e-06, "loss": 1.5498, "step": 23074 }, { "epoch": 4.4, "learning_rate": 1.6830314585319352e-06, "loss": 2.0, "step": 23075 }, { "epoch": 4.4, "learning_rate": 1.682974261201144e-06, "loss": 2.8447, "step": 23076 }, { "epoch": 4.4, "learning_rate": 1.6829170638703526e-06, "loss": 2.1701, "step": 23077 }, { "epoch": 4.4, "learning_rate": 1.6828598665395616e-06, "loss": 1.8171, "step": 23078 }, { "epoch": 4.4, "learning_rate": 1.6828026692087702e-06, "loss": 2.5244, "step": 23079 }, { "epoch": 4.4, "learning_rate": 1.6827454718779792e-06, "loss": 1.9889, "step": 23080 }, { "epoch": 4.4, "learning_rate": 1.682688274547188e-06, "loss": 2.5226, "step": 23081 }, { "epoch": 4.4, "learning_rate": 1.6826310772163966e-06, "loss": 2.2725, "step": 23082 }, { "epoch": 4.4, "learning_rate": 1.6825738798856056e-06, "loss": 1.8583, "step": 23083 }, { "epoch": 4.4, "learning_rate": 1.6825166825548141e-06, "loss": 3.1935, "step": 23084 }, { "epoch": 4.4, "learning_rate": 1.6824594852240227e-06, "loss": 2.4104, "step": 23085 }, { "epoch": 4.4, "learning_rate": 1.6824022878932317e-06, "loss": 2.0883, "step": 23086 }, { "epoch": 4.4, "learning_rate": 1.6823450905624405e-06, "loss": 3.0065, "step": 23087 }, { "epoch": 4.4, "learning_rate": 1.682287893231649e-06, "loss": 2.1211, "step": 23088 }, { "epoch": 4.4, "learning_rate": 1.6822306959008581e-06, "loss": 2.6476, "step": 23089 }, { "epoch": 4.4, "learning_rate": 1.6821734985700667e-06, "loss": 1.6196, "step": 23090 }, { "epoch": 4.4, "learning_rate": 1.6821163012392755e-06, "loss": 1.3323, "step": 23091 }, { "epoch": 4.4, "learning_rate": 1.6820591039084845e-06, "loss": 1.0596, "step": 23092 }, { "epoch": 4.4, "learning_rate": 1.682001906577693e-06, "loss": 1.7003, "step": 23093 }, { "epoch": 4.4, "learning_rate": 1.6819447092469017e-06, "loss": 1.9732, "step": 23094 }, { "epoch": 4.4, "learning_rate": 1.6818875119161107e-06, "loss": 2.9321, "step": 23095 }, { "epoch": 4.4, "learning_rate": 1.6818303145853192e-06, "loss": 2.6832, "step": 23096 }, { "epoch": 4.4, "learning_rate": 1.6817731172545282e-06, "loss": 2.228, "step": 23097 }, { "epoch": 4.4, "learning_rate": 1.681715919923737e-06, "loss": 2.0731, "step": 23098 }, { "epoch": 4.4, "learning_rate": 1.6816587225929456e-06, "loss": 2.5286, "step": 23099 }, { "epoch": 4.4, "learning_rate": 1.6816015252621546e-06, "loss": 2.5879, "step": 23100 }, { "epoch": 4.4, "learning_rate": 1.6815443279313632e-06, "loss": 1.2891, "step": 23101 }, { "epoch": 4.4, "learning_rate": 1.681487130600572e-06, "loss": 2.2533, "step": 23102 }, { "epoch": 4.4, "learning_rate": 1.6814299332697808e-06, "loss": 0.8697, "step": 23103 }, { "epoch": 4.4, "learning_rate": 1.6813727359389896e-06, "loss": 3.3013, "step": 23104 }, { "epoch": 4.4, "learning_rate": 1.6813155386081982e-06, "loss": 2.561, "step": 23105 }, { "epoch": 4.4, "learning_rate": 1.6812583412774072e-06, "loss": 1.9013, "step": 23106 }, { "epoch": 4.4, "learning_rate": 1.6812011439466158e-06, "loss": 2.8593, "step": 23107 }, { "epoch": 4.4, "learning_rate": 1.6811439466158246e-06, "loss": 2.0319, "step": 23108 }, { "epoch": 4.4, "learning_rate": 1.6810867492850336e-06, "loss": 1.5774, "step": 23109 }, { "epoch": 4.4, "learning_rate": 1.6810295519542421e-06, "loss": 2.0184, "step": 23110 }, { "epoch": 4.4, "learning_rate": 1.6809723546234511e-06, "loss": 2.9902, "step": 23111 }, { "epoch": 4.4, "learning_rate": 1.6809151572926597e-06, "loss": 2.9069, "step": 23112 }, { "epoch": 4.4, "learning_rate": 1.6808579599618683e-06, "loss": 2.3194, "step": 23113 }, { "epoch": 4.4, "learning_rate": 1.6808007626310773e-06, "loss": 1.7846, "step": 23114 }, { "epoch": 4.4, "learning_rate": 1.6807435653002861e-06, "loss": 1.5038, "step": 23115 }, { "epoch": 4.4, "learning_rate": 1.6806863679694947e-06, "loss": 1.7436, "step": 23116 }, { "epoch": 4.4, "learning_rate": 1.6806291706387037e-06, "loss": 2.6224, "step": 23117 }, { "epoch": 4.4, "learning_rate": 1.6805719733079123e-06, "loss": 1.8818, "step": 23118 }, { "epoch": 4.4, "learning_rate": 1.680514775977121e-06, "loss": 1.9471, "step": 23119 }, { "epoch": 4.4, "learning_rate": 1.68045757864633e-06, "loss": 1.5772, "step": 23120 }, { "epoch": 4.4, "learning_rate": 1.6804003813155387e-06, "loss": 1.1664, "step": 23121 }, { "epoch": 4.4, "learning_rate": 1.6803431839847472e-06, "loss": 2.86, "step": 23122 }, { "epoch": 4.4, "learning_rate": 1.6802859866539563e-06, "loss": 1.8018, "step": 23123 }, { "epoch": 4.4, "learning_rate": 1.6802287893231648e-06, "loss": 2.0231, "step": 23124 }, { "epoch": 4.4, "learning_rate": 1.6801715919923736e-06, "loss": 1.8538, "step": 23125 }, { "epoch": 4.4, "learning_rate": 1.6801143946615826e-06, "loss": 2.7052, "step": 23126 }, { "epoch": 4.41, "learning_rate": 1.6800571973307912e-06, "loss": 2.5039, "step": 23127 }, { "epoch": 4.41, "learning_rate": 1.6800000000000002e-06, "loss": 2.8684, "step": 23128 }, { "epoch": 4.41, "learning_rate": 1.6799428026692088e-06, "loss": 1.3431, "step": 23129 }, { "epoch": 4.41, "learning_rate": 1.6798856053384176e-06, "loss": 0.6511, "step": 23130 }, { "epoch": 4.41, "learning_rate": 1.6798284080076264e-06, "loss": 2.2492, "step": 23131 }, { "epoch": 4.41, "learning_rate": 1.6797712106768352e-06, "loss": 1.4436, "step": 23132 }, { "epoch": 4.41, "learning_rate": 1.6797140133460438e-06, "loss": 1.2366, "step": 23133 }, { "epoch": 4.41, "learning_rate": 1.6796568160152528e-06, "loss": 1.6241, "step": 23134 }, { "epoch": 4.41, "learning_rate": 1.6795996186844614e-06, "loss": 2.0445, "step": 23135 }, { "epoch": 4.41, "learning_rate": 1.6795424213536701e-06, "loss": 1.8139, "step": 23136 }, { "epoch": 4.41, "learning_rate": 1.6794852240228792e-06, "loss": 2.1872, "step": 23137 }, { "epoch": 4.41, "learning_rate": 1.6794280266920877e-06, "loss": 1.2459, "step": 23138 }, { "epoch": 4.41, "learning_rate": 1.6793708293612963e-06, "loss": 2.3082, "step": 23139 }, { "epoch": 4.41, "learning_rate": 1.6793136320305053e-06, "loss": 2.3282, "step": 23140 }, { "epoch": 4.41, "learning_rate": 1.6792564346997141e-06, "loss": 1.5821, "step": 23141 }, { "epoch": 4.41, "learning_rate": 1.6791992373689227e-06, "loss": 1.7986, "step": 23142 }, { "epoch": 4.41, "learning_rate": 1.6791420400381317e-06, "loss": 1.2965, "step": 23143 }, { "epoch": 4.41, "learning_rate": 1.6790848427073403e-06, "loss": 2.3591, "step": 23144 }, { "epoch": 4.41, "learning_rate": 1.6790276453765493e-06, "loss": 1.056, "step": 23145 }, { "epoch": 4.41, "learning_rate": 1.6789704480457579e-06, "loss": 2.3684, "step": 23146 }, { "epoch": 4.41, "learning_rate": 1.6789132507149667e-06, "loss": 2.0517, "step": 23147 }, { "epoch": 4.41, "learning_rate": 1.6788560533841757e-06, "loss": 2.9963, "step": 23148 }, { "epoch": 4.41, "learning_rate": 1.6787988560533843e-06, "loss": 2.6919, "step": 23149 }, { "epoch": 4.41, "learning_rate": 1.6787416587225928e-06, "loss": 1.4799, "step": 23150 }, { "epoch": 4.41, "learning_rate": 1.6786844613918018e-06, "loss": 1.0681, "step": 23151 }, { "epoch": 4.41, "learning_rate": 1.6786272640610104e-06, "loss": 1.7987, "step": 23152 }, { "epoch": 4.41, "learning_rate": 1.6785700667302192e-06, "loss": 1.6994, "step": 23153 }, { "epoch": 4.41, "learning_rate": 1.6785128693994282e-06, "loss": 2.7935, "step": 23154 }, { "epoch": 4.41, "learning_rate": 1.6784556720686368e-06, "loss": 2.7977, "step": 23155 }, { "epoch": 4.41, "learning_rate": 1.6783984747378454e-06, "loss": 1.9513, "step": 23156 }, { "epoch": 4.41, "learning_rate": 1.6783412774070544e-06, "loss": 2.2935, "step": 23157 }, { "epoch": 4.41, "learning_rate": 1.6782840800762632e-06, "loss": 1.6698, "step": 23158 }, { "epoch": 4.41, "learning_rate": 1.6782268827454718e-06, "loss": 1.8633, "step": 23159 }, { "epoch": 4.41, "learning_rate": 1.6781696854146808e-06, "loss": 2.4588, "step": 23160 }, { "epoch": 4.41, "learning_rate": 1.6781124880838894e-06, "loss": 3.1246, "step": 23161 }, { "epoch": 4.41, "learning_rate": 1.6780552907530984e-06, "loss": 1.0102, "step": 23162 }, { "epoch": 4.41, "learning_rate": 1.677998093422307e-06, "loss": 2.863, "step": 23163 }, { "epoch": 4.41, "learning_rate": 1.6779408960915157e-06, "loss": 2.5155, "step": 23164 }, { "epoch": 4.41, "learning_rate": 1.6778836987607247e-06, "loss": 1.9237, "step": 23165 }, { "epoch": 4.41, "learning_rate": 1.6778265014299333e-06, "loss": 2.8834, "step": 23166 }, { "epoch": 4.41, "learning_rate": 1.677769304099142e-06, "loss": 2.171, "step": 23167 }, { "epoch": 4.41, "learning_rate": 1.677712106768351e-06, "loss": 1.6976, "step": 23168 }, { "epoch": 4.41, "learning_rate": 1.6776549094375597e-06, "loss": 3.3731, "step": 23169 }, { "epoch": 4.41, "learning_rate": 1.6775977121067683e-06, "loss": 2.0609, "step": 23170 }, { "epoch": 4.41, "learning_rate": 1.6775405147759773e-06, "loss": 2.0075, "step": 23171 }, { "epoch": 4.41, "learning_rate": 1.6774833174451859e-06, "loss": 2.6653, "step": 23172 }, { "epoch": 4.41, "learning_rate": 1.6774261201143945e-06, "loss": 1.8793, "step": 23173 }, { "epoch": 4.41, "learning_rate": 1.6773689227836035e-06, "loss": 1.5448, "step": 23174 }, { "epoch": 4.41, "learning_rate": 1.6773117254528123e-06, "loss": 2.2664, "step": 23175 }, { "epoch": 4.41, "learning_rate": 1.6772545281220213e-06, "loss": 1.3005, "step": 23176 }, { "epoch": 4.41, "learning_rate": 1.6771973307912298e-06, "loss": 1.701, "step": 23177 }, { "epoch": 4.41, "learning_rate": 1.6771401334604384e-06, "loss": 1.3974, "step": 23178 }, { "epoch": 4.42, "learning_rate": 1.6770829361296474e-06, "loss": 3.1689, "step": 23179 }, { "epoch": 4.42, "learning_rate": 1.6770257387988562e-06, "loss": 3.3087, "step": 23180 }, { "epoch": 4.42, "learning_rate": 1.6769685414680648e-06, "loss": 1.7802, "step": 23181 }, { "epoch": 4.42, "learning_rate": 1.6769113441372738e-06, "loss": 1.5402, "step": 23182 }, { "epoch": 4.42, "learning_rate": 1.6768541468064824e-06, "loss": 1.9138, "step": 23183 }, { "epoch": 4.42, "learning_rate": 1.676796949475691e-06, "loss": 1.3459, "step": 23184 }, { "epoch": 4.42, "learning_rate": 1.6767397521449e-06, "loss": 3.0635, "step": 23185 }, { "epoch": 4.42, "learning_rate": 1.6766825548141088e-06, "loss": 2.4434, "step": 23186 }, { "epoch": 4.42, "learning_rate": 1.6766253574833174e-06, "loss": 2.1097, "step": 23187 }, { "epoch": 4.42, "learning_rate": 1.6765681601525264e-06, "loss": 2.6883, "step": 23188 }, { "epoch": 4.42, "learning_rate": 1.676510962821735e-06, "loss": 2.2005, "step": 23189 }, { "epoch": 4.42, "learning_rate": 1.6764537654909437e-06, "loss": 2.7035, "step": 23190 }, { "epoch": 4.42, "learning_rate": 1.6763965681601525e-06, "loss": 1.431, "step": 23191 }, { "epoch": 4.42, "learning_rate": 1.6763393708293613e-06, "loss": 1.5511, "step": 23192 }, { "epoch": 4.42, "learning_rate": 1.6762821734985703e-06, "loss": 1.362, "step": 23193 }, { "epoch": 4.42, "learning_rate": 1.676224976167779e-06, "loss": 1.3919, "step": 23194 }, { "epoch": 4.42, "learning_rate": 1.6761677788369875e-06, "loss": 1.1052, "step": 23195 }, { "epoch": 4.42, "learning_rate": 1.6761105815061965e-06, "loss": 1.4349, "step": 23196 }, { "epoch": 4.42, "learning_rate": 1.6760533841754053e-06, "loss": 2.2861, "step": 23197 }, { "epoch": 4.42, "learning_rate": 1.6759961868446139e-06, "loss": 2.7799, "step": 23198 }, { "epoch": 4.42, "learning_rate": 1.6759389895138229e-06, "loss": 1.9767, "step": 23199 }, { "epoch": 4.42, "learning_rate": 1.6758817921830315e-06, "loss": 2.3375, "step": 23200 }, { "epoch": 4.42, "learning_rate": 1.67582459485224e-06, "loss": 2.5397, "step": 23201 }, { "epoch": 4.42, "learning_rate": 1.675767397521449e-06, "loss": 2.2141, "step": 23202 }, { "epoch": 4.42, "learning_rate": 1.6757102001906579e-06, "loss": 1.4169, "step": 23203 }, { "epoch": 4.42, "learning_rate": 1.6756530028598664e-06, "loss": 3.0253, "step": 23204 }, { "epoch": 4.42, "learning_rate": 1.6755958055290754e-06, "loss": 1.9802, "step": 23205 }, { "epoch": 4.42, "learning_rate": 1.675538608198284e-06, "loss": 0.8692, "step": 23206 }, { "epoch": 4.42, "learning_rate": 1.6754814108674928e-06, "loss": 2.015, "step": 23207 }, { "epoch": 4.42, "learning_rate": 1.6754242135367018e-06, "loss": 1.6406, "step": 23208 }, { "epoch": 4.42, "learning_rate": 1.6753670162059104e-06, "loss": 2.4753, "step": 23209 }, { "epoch": 4.42, "learning_rate": 1.6753098188751194e-06, "loss": 2.662, "step": 23210 }, { "epoch": 4.42, "learning_rate": 1.675252621544328e-06, "loss": 1.5955, "step": 23211 }, { "epoch": 4.42, "learning_rate": 1.6751954242135366e-06, "loss": 2.7788, "step": 23212 }, { "epoch": 4.42, "learning_rate": 1.6751382268827456e-06, "loss": 2.4076, "step": 23213 }, { "epoch": 4.42, "learning_rate": 1.6750810295519544e-06, "loss": 2.6381, "step": 23214 }, { "epoch": 4.42, "learning_rate": 1.675023832221163e-06, "loss": 2.91, "step": 23215 }, { "epoch": 4.42, "learning_rate": 1.674966634890372e-06, "loss": 1.8972, "step": 23216 }, { "epoch": 4.42, "learning_rate": 1.6749094375595805e-06, "loss": 1.6396, "step": 23217 }, { "epoch": 4.42, "learning_rate": 1.6748522402287893e-06, "loss": 3.5044, "step": 23218 }, { "epoch": 4.42, "learning_rate": 1.6747950428979981e-06, "loss": 1.6173, "step": 23219 }, { "epoch": 4.42, "learning_rate": 1.674737845567207e-06, "loss": 1.8753, "step": 23220 }, { "epoch": 4.42, "learning_rate": 1.6746806482364155e-06, "loss": 2.8426, "step": 23221 }, { "epoch": 4.42, "learning_rate": 1.6746234509056245e-06, "loss": 1.4656, "step": 23222 }, { "epoch": 4.42, "learning_rate": 1.674566253574833e-06, "loss": 2.7896, "step": 23223 }, { "epoch": 4.42, "learning_rate": 1.6745090562440419e-06, "loss": 0.9718, "step": 23224 }, { "epoch": 4.42, "learning_rate": 1.6744518589132509e-06, "loss": 1.4202, "step": 23225 }, { "epoch": 4.42, "learning_rate": 1.6743946615824595e-06, "loss": 2.1647, "step": 23226 }, { "epoch": 4.42, "learning_rate": 1.6743374642516685e-06, "loss": 1.9318, "step": 23227 }, { "epoch": 4.42, "learning_rate": 1.674280266920877e-06, "loss": 1.9336, "step": 23228 }, { "epoch": 4.42, "learning_rate": 1.6742230695900859e-06, "loss": 3.9752, "step": 23229 }, { "epoch": 4.42, "learning_rate": 1.6741658722592946e-06, "loss": 2.7903, "step": 23230 }, { "epoch": 4.42, "learning_rate": 1.6741086749285034e-06, "loss": 2.3968, "step": 23231 }, { "epoch": 4.43, "learning_rate": 1.674051477597712e-06, "loss": 1.8238, "step": 23232 }, { "epoch": 4.43, "learning_rate": 1.673994280266921e-06, "loss": 2.6591, "step": 23233 }, { "epoch": 4.43, "learning_rate": 1.6739370829361296e-06, "loss": 1.7274, "step": 23234 }, { "epoch": 4.43, "learning_rate": 1.6738798856053384e-06, "loss": 2.8067, "step": 23235 }, { "epoch": 4.43, "learning_rate": 1.6738226882745474e-06, "loss": 1.4753, "step": 23236 }, { "epoch": 4.43, "learning_rate": 1.673765490943756e-06, "loss": 2.9546, "step": 23237 }, { "epoch": 4.43, "learning_rate": 1.6737082936129646e-06, "loss": 1.0451, "step": 23238 }, { "epoch": 4.43, "learning_rate": 1.6736510962821736e-06, "loss": 2.733, "step": 23239 }, { "epoch": 4.43, "learning_rate": 1.6735938989513822e-06, "loss": 1.8879, "step": 23240 }, { "epoch": 4.43, "learning_rate": 1.6735367016205912e-06, "loss": 2.0499, "step": 23241 }, { "epoch": 4.43, "learning_rate": 1.6734795042898e-06, "loss": 1.7049, "step": 23242 }, { "epoch": 4.43, "learning_rate": 1.6734223069590085e-06, "loss": 1.9945, "step": 23243 }, { "epoch": 4.43, "learning_rate": 1.6733651096282176e-06, "loss": 3.2036, "step": 23244 }, { "epoch": 4.43, "learning_rate": 1.6733079122974261e-06, "loss": 1.4168, "step": 23245 }, { "epoch": 4.43, "learning_rate": 1.673250714966635e-06, "loss": 2.2153, "step": 23246 }, { "epoch": 4.43, "learning_rate": 1.673193517635844e-06, "loss": 2.5252, "step": 23247 }, { "epoch": 4.43, "learning_rate": 1.6731363203050525e-06, "loss": 0.5962, "step": 23248 }, { "epoch": 4.43, "learning_rate": 1.673079122974261e-06, "loss": 1.9586, "step": 23249 }, { "epoch": 4.43, "learning_rate": 1.67302192564347e-06, "loss": 1.0301, "step": 23250 }, { "epoch": 4.43, "learning_rate": 1.6729647283126787e-06, "loss": 3.0822, "step": 23251 }, { "epoch": 4.43, "learning_rate": 1.6729075309818875e-06, "loss": 1.9213, "step": 23252 }, { "epoch": 4.43, "learning_rate": 1.6728503336510965e-06, "loss": 1.4925, "step": 23253 }, { "epoch": 4.43, "learning_rate": 1.672793136320305e-06, "loss": 3.102, "step": 23254 }, { "epoch": 4.43, "learning_rate": 1.6727359389895136e-06, "loss": 2.6164, "step": 23255 }, { "epoch": 4.43, "learning_rate": 1.6726787416587227e-06, "loss": 2.6626, "step": 23256 }, { "epoch": 4.43, "learning_rate": 1.6726215443279314e-06, "loss": 2.5677, "step": 23257 }, { "epoch": 4.43, "learning_rate": 1.6725643469971402e-06, "loss": 1.7811, "step": 23258 }, { "epoch": 4.43, "learning_rate": 1.672507149666349e-06, "loss": 1.5492, "step": 23259 }, { "epoch": 4.43, "learning_rate": 1.6724499523355576e-06, "loss": 3.5101, "step": 23260 }, { "epoch": 4.43, "learning_rate": 1.6723927550047666e-06, "loss": 1.6419, "step": 23261 }, { "epoch": 4.43, "learning_rate": 1.6723355576739752e-06, "loss": 2.6843, "step": 23262 }, { "epoch": 4.43, "learning_rate": 1.672278360343184e-06, "loss": 2.6757, "step": 23263 }, { "epoch": 4.43, "learning_rate": 1.672221163012393e-06, "loss": 3.3884, "step": 23264 }, { "epoch": 4.43, "learning_rate": 1.6721639656816016e-06, "loss": 1.5522, "step": 23265 }, { "epoch": 4.43, "learning_rate": 1.6721067683508102e-06, "loss": 1.5917, "step": 23266 }, { "epoch": 4.43, "learning_rate": 1.6720495710200192e-06, "loss": 1.5365, "step": 23267 }, { "epoch": 4.43, "learning_rate": 1.6719923736892278e-06, "loss": 1.5781, "step": 23268 }, { "epoch": 4.43, "learning_rate": 1.6719351763584365e-06, "loss": 2.4061, "step": 23269 }, { "epoch": 4.43, "learning_rate": 1.6718779790276456e-06, "loss": 1.2651, "step": 23270 }, { "epoch": 4.43, "learning_rate": 1.6718207816968541e-06, "loss": 2.0766, "step": 23271 }, { "epoch": 4.43, "learning_rate": 1.6717635843660627e-06, "loss": 2.3433, "step": 23272 }, { "epoch": 4.43, "learning_rate": 1.6717063870352717e-06, "loss": 2.0288, "step": 23273 }, { "epoch": 4.43, "learning_rate": 1.6716491897044805e-06, "loss": 1.6026, "step": 23274 }, { "epoch": 4.43, "learning_rate": 1.6715919923736895e-06, "loss": 3.1452, "step": 23275 }, { "epoch": 4.43, "learning_rate": 1.671534795042898e-06, "loss": 2.4207, "step": 23276 }, { "epoch": 4.43, "learning_rate": 1.6714775977121067e-06, "loss": 1.3251, "step": 23277 }, { "epoch": 4.43, "learning_rate": 1.6714204003813157e-06, "loss": 1.7893, "step": 23278 }, { "epoch": 4.43, "learning_rate": 1.6713632030505243e-06, "loss": 1.986, "step": 23279 }, { "epoch": 4.43, "learning_rate": 1.671306005719733e-06, "loss": 1.8149, "step": 23280 }, { "epoch": 4.43, "learning_rate": 1.671248808388942e-06, "loss": 2.0726, "step": 23281 }, { "epoch": 4.43, "learning_rate": 1.6711916110581507e-06, "loss": 2.0008, "step": 23282 }, { "epoch": 4.43, "learning_rate": 1.6711344137273592e-06, "loss": 3.214, "step": 23283 }, { "epoch": 4.44, "learning_rate": 1.6710772163965682e-06, "loss": 1.5759, "step": 23284 }, { "epoch": 4.44, "learning_rate": 1.671020019065777e-06, "loss": 2.5172, "step": 23285 }, { "epoch": 4.44, "learning_rate": 1.6709628217349856e-06, "loss": 1.6648, "step": 23286 }, { "epoch": 4.44, "learning_rate": 1.6709056244041946e-06, "loss": 3.1652, "step": 23287 }, { "epoch": 4.44, "learning_rate": 1.6708484270734032e-06, "loss": 2.0203, "step": 23288 }, { "epoch": 4.44, "learning_rate": 1.6707912297426118e-06, "loss": 1.8208, "step": 23289 }, { "epoch": 4.44, "learning_rate": 1.6707340324118208e-06, "loss": 1.7113, "step": 23290 }, { "epoch": 4.44, "learning_rate": 1.6706768350810296e-06, "loss": 2.0044, "step": 23291 }, { "epoch": 4.44, "learning_rate": 1.6706196377502386e-06, "loss": 2.3135, "step": 23292 }, { "epoch": 4.44, "learning_rate": 1.6705624404194472e-06, "loss": 2.7645, "step": 23293 }, { "epoch": 4.44, "learning_rate": 1.6705052430886558e-06, "loss": 1.4367, "step": 23294 }, { "epoch": 4.44, "learning_rate": 1.6704480457578648e-06, "loss": 1.3516, "step": 23295 }, { "epoch": 4.44, "learning_rate": 1.6703908484270736e-06, "loss": 1.9122, "step": 23296 }, { "epoch": 4.44, "learning_rate": 1.6703336510962821e-06, "loss": 3.4178, "step": 23297 }, { "epoch": 4.44, "learning_rate": 1.6702764537654911e-06, "loss": 1.2936, "step": 23298 }, { "epoch": 4.44, "learning_rate": 1.6702192564346997e-06, "loss": 1.9827, "step": 23299 }, { "epoch": 4.44, "learning_rate": 1.6701620591039083e-06, "loss": 1.8537, "step": 23300 }, { "epoch": 4.44, "learning_rate": 1.6701048617731173e-06, "loss": 1.7411, "step": 23301 }, { "epoch": 4.44, "learning_rate": 1.6700476644423261e-06, "loss": 2.9036, "step": 23302 }, { "epoch": 4.44, "learning_rate": 1.6699904671115347e-06, "loss": 1.566, "step": 23303 }, { "epoch": 4.44, "learning_rate": 1.6699332697807437e-06, "loss": 1.6564, "step": 23304 }, { "epoch": 4.44, "learning_rate": 1.6698760724499523e-06, "loss": 2.382, "step": 23305 }, { "epoch": 4.44, "learning_rate": 1.6698188751191613e-06, "loss": 1.0414, "step": 23306 }, { "epoch": 4.44, "learning_rate": 1.6697616777883699e-06, "loss": 2.5872, "step": 23307 }, { "epoch": 4.44, "learning_rate": 1.6697044804575787e-06, "loss": 3.173, "step": 23308 }, { "epoch": 4.44, "learning_rate": 1.6696472831267877e-06, "loss": 1.2125, "step": 23309 }, { "epoch": 4.44, "learning_rate": 1.6695900857959962e-06, "loss": 2.0359, "step": 23310 }, { "epoch": 4.44, "learning_rate": 1.6695328884652048e-06, "loss": 2.2421, "step": 23311 }, { "epoch": 4.44, "learning_rate": 1.6694756911344138e-06, "loss": 1.7069, "step": 23312 }, { "epoch": 4.44, "learning_rate": 1.6694184938036226e-06, "loss": 1.9106, "step": 23313 }, { "epoch": 4.44, "learning_rate": 1.6693612964728312e-06, "loss": 2.1257, "step": 23314 }, { "epoch": 4.44, "learning_rate": 1.6693040991420402e-06, "loss": 2.7949, "step": 23315 }, { "epoch": 4.44, "learning_rate": 1.6692469018112488e-06, "loss": 2.2686, "step": 23316 }, { "epoch": 4.44, "learning_rate": 1.6691897044804576e-06, "loss": 3.1157, "step": 23317 }, { "epoch": 4.44, "learning_rate": 1.6691325071496664e-06, "loss": 2.3793, "step": 23318 }, { "epoch": 4.44, "learning_rate": 1.6690753098188752e-06, "loss": 2.2062, "step": 23319 }, { "epoch": 4.44, "learning_rate": 1.6690181124880838e-06, "loss": 2.0242, "step": 23320 }, { "epoch": 4.44, "learning_rate": 1.6689609151572928e-06, "loss": 1.8169, "step": 23321 }, { "epoch": 4.44, "learning_rate": 1.6689037178265014e-06, "loss": 1.1242, "step": 23322 }, { "epoch": 4.44, "learning_rate": 1.6688465204957104e-06, "loss": 2.5626, "step": 23323 }, { "epoch": 4.44, "learning_rate": 1.6687893231649192e-06, "loss": 1.6491, "step": 23324 }, { "epoch": 4.44, "learning_rate": 1.6687321258341277e-06, "loss": 2.7092, "step": 23325 }, { "epoch": 4.44, "learning_rate": 1.6686749285033367e-06, "loss": 2.106, "step": 23326 }, { "epoch": 4.44, "learning_rate": 1.6686177311725453e-06, "loss": 1.2938, "step": 23327 }, { "epoch": 4.44, "learning_rate": 1.668560533841754e-06, "loss": 1.4632, "step": 23328 }, { "epoch": 4.44, "learning_rate": 1.668503336510963e-06, "loss": 1.9731, "step": 23329 }, { "epoch": 4.44, "learning_rate": 1.6684461391801717e-06, "loss": 2.1792, "step": 23330 }, { "epoch": 4.44, "learning_rate": 1.6683889418493803e-06, "loss": 2.5014, "step": 23331 }, { "epoch": 4.44, "learning_rate": 1.6683317445185893e-06, "loss": 2.5053, "step": 23332 }, { "epoch": 4.44, "learning_rate": 1.6682745471877979e-06, "loss": 3.0419, "step": 23333 }, { "epoch": 4.44, "learning_rate": 1.6682173498570067e-06, "loss": 2.9809, "step": 23334 }, { "epoch": 4.44, "learning_rate": 1.6681601525262157e-06, "loss": 2.466, "step": 23335 }, { "epoch": 4.44, "learning_rate": 1.6681029551954243e-06, "loss": 2.0088, "step": 23336 }, { "epoch": 4.45, "learning_rate": 1.6680457578646328e-06, "loss": 2.9865, "step": 23337 }, { "epoch": 4.45, "learning_rate": 1.6679885605338418e-06, "loss": 2.3646, "step": 23338 }, { "epoch": 4.45, "learning_rate": 1.6679313632030504e-06, "loss": 2.5938, "step": 23339 }, { "epoch": 4.45, "learning_rate": 1.6678741658722594e-06, "loss": 2.2735, "step": 23340 }, { "epoch": 4.45, "learning_rate": 1.6678169685414682e-06, "loss": 2.7578, "step": 23341 }, { "epoch": 4.45, "learning_rate": 1.6677597712106768e-06, "loss": 1.7244, "step": 23342 }, { "epoch": 4.45, "learning_rate": 1.6677025738798858e-06, "loss": 3.3578, "step": 23343 }, { "epoch": 4.45, "learning_rate": 1.6676453765490944e-06, "loss": 1.9917, "step": 23344 }, { "epoch": 4.45, "learning_rate": 1.6675881792183032e-06, "loss": 2.5476, "step": 23345 }, { "epoch": 4.45, "learning_rate": 1.667530981887512e-06, "loss": 2.8567, "step": 23346 }, { "epoch": 4.45, "learning_rate": 1.6674737845567208e-06, "loss": 1.072, "step": 23347 }, { "epoch": 4.45, "learning_rate": 1.6674165872259294e-06, "loss": 2.6876, "step": 23348 }, { "epoch": 4.45, "learning_rate": 1.6673593898951384e-06, "loss": 2.2301, "step": 23349 }, { "epoch": 4.45, "learning_rate": 1.667302192564347e-06, "loss": 2.5249, "step": 23350 }, { "epoch": 4.45, "learning_rate": 1.6672449952335557e-06, "loss": 2.9037, "step": 23351 }, { "epoch": 4.45, "learning_rate": 1.6671877979027647e-06, "loss": 2.1311, "step": 23352 }, { "epoch": 4.45, "learning_rate": 1.6671306005719733e-06, "loss": 1.6992, "step": 23353 }, { "epoch": 4.45, "learning_rate": 1.667073403241182e-06, "loss": 2.6304, "step": 23354 }, { "epoch": 4.45, "learning_rate": 1.667016205910391e-06, "loss": 1.8639, "step": 23355 }, { "epoch": 4.45, "learning_rate": 1.6669590085795995e-06, "loss": 2.3319, "step": 23356 }, { "epoch": 4.45, "learning_rate": 1.6669018112488085e-06, "loss": 2.0119, "step": 23357 }, { "epoch": 4.45, "learning_rate": 1.6668446139180173e-06, "loss": 2.5658, "step": 23358 }, { "epoch": 4.45, "learning_rate": 1.6667874165872259e-06, "loss": 1.6905, "step": 23359 }, { "epoch": 4.45, "learning_rate": 1.6667302192564349e-06, "loss": 3.8251, "step": 23360 }, { "epoch": 4.45, "learning_rate": 1.6666730219256435e-06, "loss": 2.6165, "step": 23361 }, { "epoch": 4.45, "learning_rate": 1.6666158245948523e-06, "loss": 1.119, "step": 23362 }, { "epoch": 4.45, "learning_rate": 1.6665586272640613e-06, "loss": 0.7604, "step": 23363 }, { "epoch": 4.45, "learning_rate": 1.6665014299332698e-06, "loss": 1.1164, "step": 23364 }, { "epoch": 4.45, "learning_rate": 1.6664442326024784e-06, "loss": 2.5738, "step": 23365 }, { "epoch": 4.45, "learning_rate": 1.6663870352716874e-06, "loss": 2.0928, "step": 23366 }, { "epoch": 4.45, "learning_rate": 1.666329837940896e-06, "loss": 3.0833, "step": 23367 }, { "epoch": 4.45, "learning_rate": 1.6662726406101048e-06, "loss": 2.7252, "step": 23368 }, { "epoch": 4.45, "learning_rate": 1.6662154432793138e-06, "loss": 1.7169, "step": 23369 }, { "epoch": 4.45, "learning_rate": 1.6661582459485224e-06, "loss": 0.9519, "step": 23370 }, { "epoch": 4.45, "learning_rate": 1.666101048617731e-06, "loss": 2.2995, "step": 23371 }, { "epoch": 4.45, "learning_rate": 1.66604385128694e-06, "loss": 1.7295, "step": 23372 }, { "epoch": 4.45, "learning_rate": 1.6659866539561488e-06, "loss": 2.0592, "step": 23373 }, { "epoch": 4.45, "learning_rate": 1.6659294566253576e-06, "loss": 2.3739, "step": 23374 }, { "epoch": 4.45, "learning_rate": 1.6658722592945664e-06, "loss": 1.8357, "step": 23375 }, { "epoch": 4.45, "learning_rate": 1.665815061963775e-06, "loss": 1.776, "step": 23376 }, { "epoch": 4.45, "learning_rate": 1.665757864632984e-06, "loss": 2.9712, "step": 23377 }, { "epoch": 4.45, "learning_rate": 1.6657006673021925e-06, "loss": 1.1572, "step": 23378 }, { "epoch": 4.45, "learning_rate": 1.6656434699714013e-06, "loss": 2.2269, "step": 23379 }, { "epoch": 4.45, "learning_rate": 1.6655862726406103e-06, "loss": 1.6756, "step": 23380 }, { "epoch": 4.45, "learning_rate": 1.665529075309819e-06, "loss": 2.5103, "step": 23381 }, { "epoch": 4.45, "learning_rate": 1.6654718779790275e-06, "loss": 1.7479, "step": 23382 }, { "epoch": 4.45, "learning_rate": 1.6654146806482365e-06, "loss": 2.1162, "step": 23383 }, { "epoch": 4.45, "learning_rate": 1.6653574833174453e-06, "loss": 1.3205, "step": 23384 }, { "epoch": 4.45, "learning_rate": 1.6653002859866539e-06, "loss": 1.4606, "step": 23385 }, { "epoch": 4.45, "learning_rate": 1.6652430886558629e-06, "loss": 2.0478, "step": 23386 }, { "epoch": 4.45, "learning_rate": 1.6651858913250715e-06, "loss": 1.1682, "step": 23387 }, { "epoch": 4.45, "learning_rate": 1.6651286939942805e-06, "loss": 1.7125, "step": 23388 }, { "epoch": 4.46, "learning_rate": 1.665071496663489e-06, "loss": 2.3282, "step": 23389 }, { "epoch": 4.46, "learning_rate": 1.6650142993326978e-06, "loss": 3.1385, "step": 23390 }, { "epoch": 4.46, "learning_rate": 1.6649571020019069e-06, "loss": 2.6308, "step": 23391 }, { "epoch": 4.46, "learning_rate": 1.6648999046711154e-06, "loss": 1.9183, "step": 23392 }, { "epoch": 4.46, "learning_rate": 1.664842707340324e-06, "loss": 1.2613, "step": 23393 }, { "epoch": 4.46, "learning_rate": 1.664785510009533e-06, "loss": 2.7485, "step": 23394 }, { "epoch": 4.46, "learning_rate": 1.6647283126787416e-06, "loss": 2.8133, "step": 23395 }, { "epoch": 4.46, "learning_rate": 1.6646711153479504e-06, "loss": 2.1249, "step": 23396 }, { "epoch": 4.46, "learning_rate": 1.6646139180171594e-06, "loss": 1.5613, "step": 23397 }, { "epoch": 4.46, "learning_rate": 1.664556720686368e-06, "loss": 2.2232, "step": 23398 }, { "epoch": 4.46, "learning_rate": 1.6644995233555766e-06, "loss": 1.6001, "step": 23399 }, { "epoch": 4.46, "learning_rate": 1.6644423260247856e-06, "loss": 1.6779, "step": 23400 }, { "epoch": 4.46, "learning_rate": 1.6643851286939944e-06, "loss": 2.6139, "step": 23401 }, { "epoch": 4.46, "learning_rate": 1.664327931363203e-06, "loss": 2.5245, "step": 23402 }, { "epoch": 4.46, "learning_rate": 1.664270734032412e-06, "loss": 2.9422, "step": 23403 }, { "epoch": 4.46, "learning_rate": 1.6642135367016205e-06, "loss": 2.291, "step": 23404 }, { "epoch": 4.46, "learning_rate": 1.6641563393708295e-06, "loss": 3.2078, "step": 23405 }, { "epoch": 4.46, "learning_rate": 1.6640991420400381e-06, "loss": 2.7446, "step": 23406 }, { "epoch": 4.46, "learning_rate": 1.664041944709247e-06, "loss": 0.411, "step": 23407 }, { "epoch": 4.46, "learning_rate": 1.663984747378456e-06, "loss": 3.0778, "step": 23408 }, { "epoch": 4.46, "learning_rate": 1.6639275500476645e-06, "loss": 1.8463, "step": 23409 }, { "epoch": 4.46, "learning_rate": 1.663870352716873e-06, "loss": 3.1021, "step": 23410 }, { "epoch": 4.46, "learning_rate": 1.663813155386082e-06, "loss": 2.6543, "step": 23411 }, { "epoch": 4.46, "learning_rate": 1.6637559580552909e-06, "loss": 2.3645, "step": 23412 }, { "epoch": 4.46, "learning_rate": 1.6636987607244995e-06, "loss": 2.0965, "step": 23413 }, { "epoch": 4.46, "learning_rate": 1.6636415633937085e-06, "loss": 3.3233, "step": 23414 }, { "epoch": 4.46, "learning_rate": 1.663584366062917e-06, "loss": 1.6262, "step": 23415 }, { "epoch": 4.46, "learning_rate": 1.6635271687321256e-06, "loss": 2.5024, "step": 23416 }, { "epoch": 4.46, "learning_rate": 1.6634699714013346e-06, "loss": 2.47, "step": 23417 }, { "epoch": 4.46, "learning_rate": 1.6634127740705434e-06, "loss": 0.9038, "step": 23418 }, { "epoch": 4.46, "learning_rate": 1.663355576739752e-06, "loss": 1.742, "step": 23419 }, { "epoch": 4.46, "learning_rate": 1.663298379408961e-06, "loss": 1.6063, "step": 23420 }, { "epoch": 4.46, "learning_rate": 1.6632411820781696e-06, "loss": 2.2549, "step": 23421 }, { "epoch": 4.46, "learning_rate": 1.6631839847473786e-06, "loss": 3.1216, "step": 23422 }, { "epoch": 4.46, "learning_rate": 1.6631267874165872e-06, "loss": 2.4051, "step": 23423 }, { "epoch": 4.46, "learning_rate": 1.663069590085796e-06, "loss": 1.878, "step": 23424 }, { "epoch": 4.46, "learning_rate": 1.663012392755005e-06, "loss": 2.6225, "step": 23425 }, { "epoch": 4.46, "learning_rate": 1.6629551954242136e-06, "loss": 2.4198, "step": 23426 }, { "epoch": 4.46, "learning_rate": 1.6628979980934222e-06, "loss": 1.2962, "step": 23427 }, { "epoch": 4.46, "learning_rate": 1.6628408007626312e-06, "loss": 1.6648, "step": 23428 }, { "epoch": 4.46, "learning_rate": 1.66278360343184e-06, "loss": 2.0815, "step": 23429 }, { "epoch": 4.46, "learning_rate": 1.6627264061010485e-06, "loss": 1.7643, "step": 23430 }, { "epoch": 4.46, "learning_rate": 1.6626692087702575e-06, "loss": 3.5968, "step": 23431 }, { "epoch": 4.46, "learning_rate": 1.6626120114394661e-06, "loss": 2.404, "step": 23432 }, { "epoch": 4.46, "learning_rate": 1.662554814108675e-06, "loss": 3.2701, "step": 23433 }, { "epoch": 4.46, "learning_rate": 1.6624976167778837e-06, "loss": 1.3601, "step": 23434 }, { "epoch": 4.46, "learning_rate": 1.6624404194470925e-06, "loss": 1.1266, "step": 23435 }, { "epoch": 4.46, "learning_rate": 1.662383222116301e-06, "loss": 2.0465, "step": 23436 }, { "epoch": 4.46, "learning_rate": 1.66232602478551e-06, "loss": 1.6678, "step": 23437 }, { "epoch": 4.46, "learning_rate": 1.6622688274547187e-06, "loss": 1.8358, "step": 23438 }, { "epoch": 4.46, "learning_rate": 1.6622116301239277e-06, "loss": 3.7373, "step": 23439 }, { "epoch": 4.46, "learning_rate": 1.6621544327931365e-06, "loss": 1.8744, "step": 23440 }, { "epoch": 4.46, "learning_rate": 1.662097235462345e-06, "loss": 2.7744, "step": 23441 }, { "epoch": 4.47, "learning_rate": 1.662040038131554e-06, "loss": 3.6149, "step": 23442 }, { "epoch": 4.47, "learning_rate": 1.6619828408007626e-06, "loss": 1.3886, "step": 23443 }, { "epoch": 4.47, "learning_rate": 1.6619256434699712e-06, "loss": 2.5281, "step": 23444 }, { "epoch": 4.47, "learning_rate": 1.6618684461391802e-06, "loss": 1.3893, "step": 23445 }, { "epoch": 4.47, "learning_rate": 1.661811248808389e-06, "loss": 0.8131, "step": 23446 }, { "epoch": 4.47, "learning_rate": 1.6617540514775976e-06, "loss": 1.5094, "step": 23447 }, { "epoch": 4.47, "learning_rate": 1.6616968541468066e-06, "loss": 1.894, "step": 23448 }, { "epoch": 4.47, "learning_rate": 1.6616396568160152e-06, "loss": 1.5742, "step": 23449 }, { "epoch": 4.47, "learning_rate": 1.661582459485224e-06, "loss": 1.6367, "step": 23450 }, { "epoch": 4.47, "learning_rate": 1.661525262154433e-06, "loss": 1.8809, "step": 23451 }, { "epoch": 4.47, "learning_rate": 1.6614680648236416e-06, "loss": 2.1039, "step": 23452 }, { "epoch": 4.47, "learning_rate": 1.6614108674928506e-06, "loss": 2.0619, "step": 23453 }, { "epoch": 4.47, "learning_rate": 1.6613536701620592e-06, "loss": 3.3215, "step": 23454 }, { "epoch": 4.47, "learning_rate": 1.6612964728312678e-06, "loss": 1.1044, "step": 23455 }, { "epoch": 4.47, "learning_rate": 1.6612392755004768e-06, "loss": 1.3588, "step": 23456 }, { "epoch": 4.47, "learning_rate": 1.6611820781696856e-06, "loss": 3.1705, "step": 23457 }, { "epoch": 4.47, "learning_rate": 1.6611248808388941e-06, "loss": 1.8582, "step": 23458 }, { "epoch": 4.47, "learning_rate": 1.6610676835081031e-06, "loss": 2.09, "step": 23459 }, { "epoch": 4.47, "learning_rate": 1.6610104861773117e-06, "loss": 1.9558, "step": 23460 }, { "epoch": 4.47, "learning_rate": 1.6609532888465205e-06, "loss": 1.9373, "step": 23461 }, { "epoch": 4.47, "learning_rate": 1.6608960915157293e-06, "loss": 1.6061, "step": 23462 }, { "epoch": 4.47, "learning_rate": 1.660838894184938e-06, "loss": 0.9629, "step": 23463 }, { "epoch": 4.47, "learning_rate": 1.6607816968541467e-06, "loss": 1.5019, "step": 23464 }, { "epoch": 4.47, "learning_rate": 1.6607244995233557e-06, "loss": 1.7329, "step": 23465 }, { "epoch": 4.47, "learning_rate": 1.6606673021925643e-06, "loss": 2.1355, "step": 23466 }, { "epoch": 4.47, "learning_rate": 1.660610104861773e-06, "loss": 2.2514, "step": 23467 }, { "epoch": 4.47, "learning_rate": 1.660552907530982e-06, "loss": 1.6698, "step": 23468 }, { "epoch": 4.47, "learning_rate": 1.6604957102001907e-06, "loss": 1.3756, "step": 23469 }, { "epoch": 4.47, "learning_rate": 1.6604385128693997e-06, "loss": 1.6781, "step": 23470 }, { "epoch": 4.47, "learning_rate": 1.6603813155386082e-06, "loss": 2.374, "step": 23471 }, { "epoch": 4.47, "learning_rate": 1.660324118207817e-06, "loss": 2.9782, "step": 23472 }, { "epoch": 4.47, "learning_rate": 1.6602669208770258e-06, "loss": 1.7921, "step": 23473 }, { "epoch": 4.47, "learning_rate": 1.6602097235462346e-06, "loss": 1.123, "step": 23474 }, { "epoch": 4.47, "learning_rate": 1.6601525262154432e-06, "loss": 1.5285, "step": 23475 }, { "epoch": 4.47, "learning_rate": 1.6600953288846522e-06, "loss": 2.187, "step": 23476 }, { "epoch": 4.47, "learning_rate": 1.6600381315538608e-06, "loss": 1.3548, "step": 23477 }, { "epoch": 4.47, "learning_rate": 1.6599809342230696e-06, "loss": 3.1658, "step": 23478 }, { "epoch": 4.47, "learning_rate": 1.6599237368922786e-06, "loss": 1.1304, "step": 23479 }, { "epoch": 4.47, "learning_rate": 1.6598665395614872e-06, "loss": 1.5211, "step": 23480 }, { "epoch": 4.47, "learning_rate": 1.6598093422306958e-06, "loss": 2.0354, "step": 23481 }, { "epoch": 4.47, "learning_rate": 1.6597521448999048e-06, "loss": 2.6892, "step": 23482 }, { "epoch": 4.47, "learning_rate": 1.6596949475691133e-06, "loss": 2.5912, "step": 23483 }, { "epoch": 4.47, "learning_rate": 1.6596377502383221e-06, "loss": 1.9491, "step": 23484 }, { "epoch": 4.47, "learning_rate": 1.6595805529075311e-06, "loss": 1.2559, "step": 23485 }, { "epoch": 4.47, "learning_rate": 1.6595233555767397e-06, "loss": 1.6359, "step": 23486 }, { "epoch": 4.47, "learning_rate": 1.6594661582459487e-06, "loss": 2.262, "step": 23487 }, { "epoch": 4.47, "learning_rate": 1.6594089609151573e-06, "loss": 2.5872, "step": 23488 }, { "epoch": 4.47, "learning_rate": 1.6593517635843661e-06, "loss": 3.8252, "step": 23489 }, { "epoch": 4.47, "learning_rate": 1.6592945662535751e-06, "loss": 1.893, "step": 23490 }, { "epoch": 4.47, "learning_rate": 1.6592373689227837e-06, "loss": 2.9319, "step": 23491 }, { "epoch": 4.47, "learning_rate": 1.6591801715919923e-06, "loss": 1.654, "step": 23492 }, { "epoch": 4.47, "learning_rate": 1.6591229742612013e-06, "loss": 1.9494, "step": 23493 }, { "epoch": 4.48, "learning_rate": 1.6590657769304099e-06, "loss": 1.724, "step": 23494 }, { "epoch": 4.48, "learning_rate": 1.6590085795996187e-06, "loss": 2.2766, "step": 23495 }, { "epoch": 4.48, "learning_rate": 1.6589513822688277e-06, "loss": 2.9106, "step": 23496 }, { "epoch": 4.48, "learning_rate": 1.6588941849380362e-06, "loss": 1.1179, "step": 23497 }, { "epoch": 4.48, "learning_rate": 1.6588369876072448e-06, "loss": 3.3643, "step": 23498 }, { "epoch": 4.48, "learning_rate": 1.6587797902764538e-06, "loss": 2.1694, "step": 23499 }, { "epoch": 4.48, "learning_rate": 1.6587225929456626e-06, "loss": 1.5387, "step": 23500 }, { "epoch": 4.48, "learning_rate": 1.6586653956148712e-06, "loss": 1.8343, "step": 23501 }, { "epoch": 4.48, "learning_rate": 1.6586081982840802e-06, "loss": 1.1001, "step": 23502 }, { "epoch": 4.48, "learning_rate": 1.6585510009532888e-06, "loss": 2.5365, "step": 23503 }, { "epoch": 4.48, "learning_rate": 1.6584938036224978e-06, "loss": 2.4252, "step": 23504 }, { "epoch": 4.48, "learning_rate": 1.6584366062917064e-06, "loss": 2.6249, "step": 23505 }, { "epoch": 4.48, "learning_rate": 1.6583794089609152e-06, "loss": 1.7609, "step": 23506 }, { "epoch": 4.48, "learning_rate": 1.6583222116301242e-06, "loss": 1.8543, "step": 23507 }, { "epoch": 4.48, "learning_rate": 1.6582650142993328e-06, "loss": 1.7093, "step": 23508 }, { "epoch": 4.48, "learning_rate": 1.6582078169685413e-06, "loss": 3.7659, "step": 23509 }, { "epoch": 4.48, "learning_rate": 1.6581506196377504e-06, "loss": 1.5922, "step": 23510 }, { "epoch": 4.48, "learning_rate": 1.658093422306959e-06, "loss": 1.8633, "step": 23511 }, { "epoch": 4.48, "learning_rate": 1.6580362249761677e-06, "loss": 1.8818, "step": 23512 }, { "epoch": 4.48, "learning_rate": 1.6579790276453767e-06, "loss": 1.7055, "step": 23513 }, { "epoch": 4.48, "learning_rate": 1.6579218303145853e-06, "loss": 1.8269, "step": 23514 }, { "epoch": 4.48, "learning_rate": 1.657864632983794e-06, "loss": 1.8575, "step": 23515 }, { "epoch": 4.48, "learning_rate": 1.657807435653003e-06, "loss": 2.0903, "step": 23516 }, { "epoch": 4.48, "learning_rate": 1.6577502383222117e-06, "loss": 2.4042, "step": 23517 }, { "epoch": 4.48, "learning_rate": 1.6576930409914207e-06, "loss": 1.7768, "step": 23518 }, { "epoch": 4.48, "learning_rate": 1.6576358436606293e-06, "loss": 2.6487, "step": 23519 }, { "epoch": 4.48, "learning_rate": 1.6575786463298379e-06, "loss": 2.2177, "step": 23520 }, { "epoch": 4.48, "learning_rate": 1.6575214489990469e-06, "loss": 1.901, "step": 23521 }, { "epoch": 4.48, "learning_rate": 1.6574642516682555e-06, "loss": 1.7297, "step": 23522 }, { "epoch": 4.48, "learning_rate": 1.6574070543374642e-06, "loss": 2.5259, "step": 23523 }, { "epoch": 4.48, "learning_rate": 1.6573498570066733e-06, "loss": 1.5604, "step": 23524 }, { "epoch": 4.48, "learning_rate": 1.6572926596758818e-06, "loss": 3.6405, "step": 23525 }, { "epoch": 4.48, "learning_rate": 1.6572354623450904e-06, "loss": 1.1209, "step": 23526 }, { "epoch": 4.48, "learning_rate": 1.6571782650142994e-06, "loss": 1.8591, "step": 23527 }, { "epoch": 4.48, "learning_rate": 1.6571210676835082e-06, "loss": 1.9161, "step": 23528 }, { "epoch": 4.48, "learning_rate": 1.6570638703527168e-06, "loss": 2.635, "step": 23529 }, { "epoch": 4.48, "learning_rate": 1.6570066730219258e-06, "loss": 1.9527, "step": 23530 }, { "epoch": 4.48, "learning_rate": 1.6569494756911344e-06, "loss": 2.9021, "step": 23531 }, { "epoch": 4.48, "learning_rate": 1.656892278360343e-06, "loss": 2.5767, "step": 23532 }, { "epoch": 4.48, "learning_rate": 1.656835081029552e-06, "loss": 2.3129, "step": 23533 }, { "epoch": 4.48, "learning_rate": 1.6567778836987608e-06, "loss": 2.1172, "step": 23534 }, { "epoch": 4.48, "learning_rate": 1.6567206863679698e-06, "loss": 2.1252, "step": 23535 }, { "epoch": 4.48, "learning_rate": 1.6566634890371784e-06, "loss": 1.5349, "step": 23536 }, { "epoch": 4.48, "learning_rate": 1.656606291706387e-06, "loss": 2.2331, "step": 23537 }, { "epoch": 4.48, "learning_rate": 1.656549094375596e-06, "loss": 1.6337, "step": 23538 }, { "epoch": 4.48, "learning_rate": 1.6564918970448047e-06, "loss": 1.6224, "step": 23539 }, { "epoch": 4.48, "learning_rate": 1.6564346997140133e-06, "loss": 2.8837, "step": 23540 }, { "epoch": 4.48, "learning_rate": 1.6563775023832223e-06, "loss": 1.5984, "step": 23541 }, { "epoch": 4.48, "learning_rate": 1.656320305052431e-06, "loss": 2.5378, "step": 23542 }, { "epoch": 4.48, "learning_rate": 1.6562631077216395e-06, "loss": 1.0633, "step": 23543 }, { "epoch": 4.48, "learning_rate": 1.6562059103908485e-06, "loss": 0.9251, "step": 23544 }, { "epoch": 4.48, "learning_rate": 1.6561487130600573e-06, "loss": 1.4576, "step": 23545 }, { "epoch": 4.48, "learning_rate": 1.6560915157292659e-06, "loss": 1.2118, "step": 23546 }, { "epoch": 4.49, "learning_rate": 1.6560343183984749e-06, "loss": 1.8266, "step": 23547 }, { "epoch": 4.49, "learning_rate": 1.6559771210676835e-06, "loss": 2.2713, "step": 23548 }, { "epoch": 4.49, "learning_rate": 1.6559199237368923e-06, "loss": 1.9249, "step": 23549 }, { "epoch": 4.49, "learning_rate": 1.655862726406101e-06, "loss": 1.72, "step": 23550 }, { "epoch": 4.49, "learning_rate": 1.6558055290753098e-06, "loss": 2.6971, "step": 23551 }, { "epoch": 4.49, "learning_rate": 1.6557483317445188e-06, "loss": 2.3473, "step": 23552 }, { "epoch": 4.49, "learning_rate": 1.6556911344137274e-06, "loss": 1.6688, "step": 23553 }, { "epoch": 4.49, "learning_rate": 1.655633937082936e-06, "loss": 2.7373, "step": 23554 }, { "epoch": 4.49, "learning_rate": 1.655576739752145e-06, "loss": 1.9669, "step": 23555 }, { "epoch": 4.49, "learning_rate": 1.6555195424213538e-06, "loss": 1.9753, "step": 23556 }, { "epoch": 4.49, "learning_rate": 1.6554623450905624e-06, "loss": 1.6583, "step": 23557 }, { "epoch": 4.49, "learning_rate": 1.6554051477597714e-06, "loss": 2.7044, "step": 23558 }, { "epoch": 4.49, "learning_rate": 1.65534795042898e-06, "loss": 2.7803, "step": 23559 }, { "epoch": 4.49, "learning_rate": 1.6552907530981888e-06, "loss": 0.958, "step": 23560 }, { "epoch": 4.49, "learning_rate": 1.6552335557673976e-06, "loss": 2.0589, "step": 23561 }, { "epoch": 4.49, "learning_rate": 1.6551763584366064e-06, "loss": 1.9476, "step": 23562 }, { "epoch": 4.49, "learning_rate": 1.655119161105815e-06, "loss": 1.8299, "step": 23563 }, { "epoch": 4.49, "learning_rate": 1.655061963775024e-06, "loss": 3.2171, "step": 23564 }, { "epoch": 4.49, "learning_rate": 1.6550047664442325e-06, "loss": 3.6137, "step": 23565 }, { "epoch": 4.49, "learning_rate": 1.6549475691134413e-06, "loss": 1.572, "step": 23566 }, { "epoch": 4.49, "learning_rate": 1.6548903717826503e-06, "loss": 2.7325, "step": 23567 }, { "epoch": 4.49, "learning_rate": 1.654833174451859e-06, "loss": 2.8029, "step": 23568 }, { "epoch": 4.49, "learning_rate": 1.654775977121068e-06, "loss": 2.1985, "step": 23569 }, { "epoch": 4.49, "learning_rate": 1.6547187797902765e-06, "loss": 1.3948, "step": 23570 }, { "epoch": 4.49, "learning_rate": 1.654661582459485e-06, "loss": 3.3824, "step": 23571 }, { "epoch": 4.49, "learning_rate": 1.654604385128694e-06, "loss": 2.2145, "step": 23572 }, { "epoch": 4.49, "learning_rate": 1.6545471877979029e-06, "loss": 3.3777, "step": 23573 }, { "epoch": 4.49, "learning_rate": 1.6544899904671115e-06, "loss": 2.6915, "step": 23574 }, { "epoch": 4.49, "learning_rate": 1.6544327931363205e-06, "loss": 2.4343, "step": 23575 }, { "epoch": 4.49, "learning_rate": 1.654375595805529e-06, "loss": 1.9315, "step": 23576 }, { "epoch": 4.49, "learning_rate": 1.6543183984747378e-06, "loss": 2.3525, "step": 23577 }, { "epoch": 4.49, "learning_rate": 1.6542612011439466e-06, "loss": 3.052, "step": 23578 }, { "epoch": 4.49, "learning_rate": 1.6542040038131554e-06, "loss": 2.2555, "step": 23579 }, { "epoch": 4.49, "learning_rate": 1.654146806482364e-06, "loss": 1.4831, "step": 23580 }, { "epoch": 4.49, "learning_rate": 1.654089609151573e-06, "loss": 1.4883, "step": 23581 }, { "epoch": 4.49, "learning_rate": 1.6540324118207816e-06, "loss": 1.4744, "step": 23582 }, { "epoch": 4.49, "learning_rate": 1.6539752144899906e-06, "loss": 1.4856, "step": 23583 }, { "epoch": 4.49, "learning_rate": 1.6539180171591994e-06, "loss": 2.5732, "step": 23584 }, { "epoch": 4.49, "learning_rate": 1.653860819828408e-06, "loss": 2.0586, "step": 23585 }, { "epoch": 4.49, "learning_rate": 1.653803622497617e-06, "loss": 1.7415, "step": 23586 }, { "epoch": 4.49, "learning_rate": 1.6537464251668256e-06, "loss": 2.4044, "step": 23587 }, { "epoch": 4.49, "learning_rate": 1.6536892278360344e-06, "loss": 2.4911, "step": 23588 }, { "epoch": 4.49, "learning_rate": 1.6536320305052432e-06, "loss": 1.7903, "step": 23589 }, { "epoch": 4.49, "learning_rate": 1.653574833174452e-06, "loss": 3.549, "step": 23590 }, { "epoch": 4.49, "learning_rate": 1.6535176358436605e-06, "loss": 0.9713, "step": 23591 }, { "epoch": 4.49, "learning_rate": 1.6534604385128695e-06, "loss": 2.2495, "step": 23592 }, { "epoch": 4.49, "learning_rate": 1.6534032411820781e-06, "loss": 2.7799, "step": 23593 }, { "epoch": 4.49, "learning_rate": 1.653346043851287e-06, "loss": 3.2617, "step": 23594 }, { "epoch": 4.49, "learning_rate": 1.653288846520496e-06, "loss": 2.5774, "step": 23595 }, { "epoch": 4.49, "learning_rate": 1.6532316491897045e-06, "loss": 1.6067, "step": 23596 }, { "epoch": 4.49, "learning_rate": 1.653174451858913e-06, "loss": 3.2983, "step": 23597 }, { "epoch": 4.49, "learning_rate": 1.653117254528122e-06, "loss": 1.5408, "step": 23598 }, { "epoch": 4.5, "learning_rate": 1.6530600571973307e-06, "loss": 2.112, "step": 23599 }, { "epoch": 4.5, "learning_rate": 1.6530028598665397e-06, "loss": 1.7802, "step": 23600 }, { "epoch": 4.5, "learning_rate": 1.6529456625357485e-06, "loss": 1.8221, "step": 23601 }, { "epoch": 4.5, "learning_rate": 1.652888465204957e-06, "loss": 2.3919, "step": 23602 }, { "epoch": 4.5, "learning_rate": 1.652831267874166e-06, "loss": 1.1066, "step": 23603 }, { "epoch": 4.5, "learning_rate": 1.6527740705433746e-06, "loss": 1.9426, "step": 23604 }, { "epoch": 4.5, "learning_rate": 1.6527168732125834e-06, "loss": 1.1158, "step": 23605 }, { "epoch": 4.5, "learning_rate": 1.6526596758817924e-06, "loss": 1.6333, "step": 23606 }, { "epoch": 4.5, "learning_rate": 1.652602478551001e-06, "loss": 1.7862, "step": 23607 }, { "epoch": 4.5, "learning_rate": 1.6525452812202096e-06, "loss": 2.3417, "step": 23608 }, { "epoch": 4.5, "learning_rate": 1.6524880838894186e-06, "loss": 0.8497, "step": 23609 }, { "epoch": 4.5, "learning_rate": 1.6524308865586272e-06, "loss": 2.2695, "step": 23610 }, { "epoch": 4.5, "learning_rate": 1.652373689227836e-06, "loss": 2.0843, "step": 23611 }, { "epoch": 4.5, "learning_rate": 1.652316491897045e-06, "loss": 2.8059, "step": 23612 }, { "epoch": 4.5, "learning_rate": 1.6522592945662536e-06, "loss": 2.3197, "step": 23613 }, { "epoch": 4.5, "learning_rate": 1.6522020972354622e-06, "loss": 2.1584, "step": 23614 }, { "epoch": 4.5, "learning_rate": 1.6521448999046712e-06, "loss": 2.3432, "step": 23615 }, { "epoch": 4.5, "learning_rate": 1.65208770257388e-06, "loss": 1.906, "step": 23616 }, { "epoch": 4.5, "learning_rate": 1.6520305052430888e-06, "loss": 1.5097, "step": 23617 }, { "epoch": 4.5, "learning_rate": 1.6519733079122975e-06, "loss": 1.8792, "step": 23618 }, { "epoch": 4.5, "learning_rate": 1.6519161105815061e-06, "loss": 2.3384, "step": 23619 }, { "epoch": 4.5, "learning_rate": 1.6518589132507151e-06, "loss": 2.1808, "step": 23620 }, { "epoch": 4.5, "learning_rate": 1.6518017159199237e-06, "loss": 2.4007, "step": 23621 }, { "epoch": 4.5, "learning_rate": 1.6517445185891325e-06, "loss": 2.6082, "step": 23622 }, { "epoch": 4.5, "learning_rate": 1.6516873212583415e-06, "loss": 1.2846, "step": 23623 }, { "epoch": 4.5, "learning_rate": 1.65163012392755e-06, "loss": 2.6853, "step": 23624 }, { "epoch": 4.5, "learning_rate": 1.6515729265967587e-06, "loss": 2.1899, "step": 23625 }, { "epoch": 4.5, "learning_rate": 1.6515157292659677e-06, "loss": 2.1056, "step": 23626 }, { "epoch": 4.5, "learning_rate": 1.6514585319351765e-06, "loss": 1.7961, "step": 23627 }, { "epoch": 4.5, "learning_rate": 1.651401334604385e-06, "loss": 2.3548, "step": 23628 }, { "epoch": 4.5, "learning_rate": 1.651344137273594e-06, "loss": 1.263, "step": 23629 }, { "epoch": 4.5, "learning_rate": 1.6512869399428026e-06, "loss": 2.4619, "step": 23630 }, { "epoch": 4.5, "learning_rate": 1.6512297426120112e-06, "loss": 1.3609, "step": 23631 }, { "epoch": 4.5, "learning_rate": 1.6511725452812202e-06, "loss": 1.5249, "step": 23632 }, { "epoch": 4.5, "learning_rate": 1.651115347950429e-06, "loss": 2.7031, "step": 23633 }, { "epoch": 4.5, "learning_rate": 1.651058150619638e-06, "loss": 3.2072, "step": 23634 }, { "epoch": 4.5, "learning_rate": 1.6510009532888466e-06, "loss": 1.6394, "step": 23635 }, { "epoch": 4.5, "learning_rate": 1.6509437559580552e-06, "loss": 3.0695, "step": 23636 }, { "epoch": 4.5, "learning_rate": 1.6508865586272642e-06, "loss": 2.2346, "step": 23637 }, { "epoch": 4.5, "learning_rate": 1.6508293612964728e-06, "loss": 2.1583, "step": 23638 }, { "epoch": 4.5, "learning_rate": 1.6507721639656816e-06, "loss": 2.4109, "step": 23639 }, { "epoch": 4.5, "learning_rate": 1.6507149666348906e-06, "loss": 2.1396, "step": 23640 }, { "epoch": 4.5, "learning_rate": 1.6506577693040992e-06, "loss": 1.7403, "step": 23641 }, { "epoch": 4.5, "learning_rate": 1.6506005719733077e-06, "loss": 2.134, "step": 23642 }, { "epoch": 4.5, "learning_rate": 1.6505433746425168e-06, "loss": 1.0573, "step": 23643 }, { "epoch": 4.5, "learning_rate": 1.6504861773117255e-06, "loss": 1.0816, "step": 23644 }, { "epoch": 4.5, "learning_rate": 1.6504289799809341e-06, "loss": 2.7213, "step": 23645 }, { "epoch": 4.5, "learning_rate": 1.6503717826501431e-06, "loss": 2.1707, "step": 23646 }, { "epoch": 4.5, "learning_rate": 1.6503145853193517e-06, "loss": 2.4162, "step": 23647 }, { "epoch": 4.5, "learning_rate": 1.6502573879885607e-06, "loss": 1.582, "step": 23648 }, { "epoch": 4.5, "learning_rate": 1.6502001906577693e-06, "loss": 1.9224, "step": 23649 }, { "epoch": 4.5, "learning_rate": 1.650142993326978e-06, "loss": 1.8424, "step": 23650 }, { "epoch": 4.5, "learning_rate": 1.650085795996187e-06, "loss": 2.1188, "step": 23651 }, { "epoch": 4.51, "learning_rate": 1.6500285986653957e-06, "loss": 2.1247, "step": 23652 }, { "epoch": 4.51, "learning_rate": 1.6499714013346043e-06, "loss": 2.2403, "step": 23653 }, { "epoch": 4.51, "learning_rate": 1.6499142040038133e-06, "loss": 2.3174, "step": 23654 }, { "epoch": 4.51, "learning_rate": 1.649857006673022e-06, "loss": 3.0296, "step": 23655 }, { "epoch": 4.51, "learning_rate": 1.6497998093422307e-06, "loss": 2.1153, "step": 23656 }, { "epoch": 4.51, "learning_rate": 1.6497426120114397e-06, "loss": 2.3651, "step": 23657 }, { "epoch": 4.51, "learning_rate": 1.6496854146806482e-06, "loss": 1.7741, "step": 23658 }, { "epoch": 4.51, "learning_rate": 1.6496282173498568e-06, "loss": 2.3845, "step": 23659 }, { "epoch": 4.51, "learning_rate": 1.6495710200190658e-06, "loss": 2.1129, "step": 23660 }, { "epoch": 4.51, "learning_rate": 1.6495138226882746e-06, "loss": 2.0199, "step": 23661 }, { "epoch": 4.51, "learning_rate": 1.6494566253574832e-06, "loss": 2.9503, "step": 23662 }, { "epoch": 4.51, "learning_rate": 1.6493994280266922e-06, "loss": 1.8131, "step": 23663 }, { "epoch": 4.51, "learning_rate": 1.6493422306959008e-06, "loss": 1.5129, "step": 23664 }, { "epoch": 4.51, "learning_rate": 1.6492850333651098e-06, "loss": 1.1571, "step": 23665 }, { "epoch": 4.51, "learning_rate": 1.6492278360343184e-06, "loss": 2.1903, "step": 23666 }, { "epoch": 4.51, "learning_rate": 1.6491706387035272e-06, "loss": 3.0047, "step": 23667 }, { "epoch": 4.51, "learning_rate": 1.6491134413727362e-06, "loss": 2.482, "step": 23668 }, { "epoch": 4.51, "learning_rate": 1.6490562440419448e-06, "loss": 2.1054, "step": 23669 }, { "epoch": 4.51, "learning_rate": 1.6489990467111533e-06, "loss": 2.2315, "step": 23670 }, { "epoch": 4.51, "learning_rate": 1.6489418493803623e-06, "loss": 3.1198, "step": 23671 }, { "epoch": 4.51, "learning_rate": 1.6488846520495711e-06, "loss": 2.7277, "step": 23672 }, { "epoch": 4.51, "learning_rate": 1.6488274547187797e-06, "loss": 1.4021, "step": 23673 }, { "epoch": 4.51, "learning_rate": 1.6487702573879887e-06, "loss": 2.6027, "step": 23674 }, { "epoch": 4.51, "learning_rate": 1.6487130600571973e-06, "loss": 2.393, "step": 23675 }, { "epoch": 4.51, "learning_rate": 1.648655862726406e-06, "loss": 2.0358, "step": 23676 }, { "epoch": 4.51, "learning_rate": 1.648598665395615e-06, "loss": 1.773, "step": 23677 }, { "epoch": 4.51, "learning_rate": 1.6485414680648237e-06, "loss": 2.7652, "step": 23678 }, { "epoch": 4.51, "learning_rate": 1.6484842707340323e-06, "loss": 1.7046, "step": 23679 }, { "epoch": 4.51, "learning_rate": 1.6484270734032413e-06, "loss": 2.6563, "step": 23680 }, { "epoch": 4.51, "learning_rate": 1.6483698760724499e-06, "loss": 2.3574, "step": 23681 }, { "epoch": 4.51, "learning_rate": 1.6483126787416589e-06, "loss": 2.7255, "step": 23682 }, { "epoch": 4.51, "learning_rate": 1.6482554814108677e-06, "loss": 1.878, "step": 23683 }, { "epoch": 4.51, "learning_rate": 1.6481982840800762e-06, "loss": 2.5673, "step": 23684 }, { "epoch": 4.51, "learning_rate": 1.6481410867492852e-06, "loss": 1.788, "step": 23685 }, { "epoch": 4.51, "learning_rate": 1.6480838894184938e-06, "loss": 2.7714, "step": 23686 }, { "epoch": 4.51, "learning_rate": 1.6480266920877024e-06, "loss": 2.1894, "step": 23687 }, { "epoch": 4.51, "learning_rate": 1.6479694947569114e-06, "loss": 1.9295, "step": 23688 }, { "epoch": 4.51, "learning_rate": 1.6479122974261202e-06, "loss": 2.1854, "step": 23689 }, { "epoch": 4.51, "learning_rate": 1.6478551000953288e-06, "loss": 1.0562, "step": 23690 }, { "epoch": 4.51, "learning_rate": 1.6477979027645378e-06, "loss": 1.3037, "step": 23691 }, { "epoch": 4.51, "learning_rate": 1.6477407054337464e-06, "loss": 1.7354, "step": 23692 }, { "epoch": 4.51, "learning_rate": 1.6476835081029552e-06, "loss": 3.3064, "step": 23693 }, { "epoch": 4.51, "learning_rate": 1.6476263107721642e-06, "loss": 1.36, "step": 23694 }, { "epoch": 4.51, "learning_rate": 1.6475691134413728e-06, "loss": 1.9245, "step": 23695 }, { "epoch": 4.51, "learning_rate": 1.6475119161105813e-06, "loss": 1.5856, "step": 23696 }, { "epoch": 4.51, "learning_rate": 1.6474547187797904e-06, "loss": 1.37, "step": 23697 }, { "epoch": 4.51, "learning_rate": 1.647397521448999e-06, "loss": 2.6162, "step": 23698 }, { "epoch": 4.51, "learning_rate": 1.647340324118208e-06, "loss": 1.0815, "step": 23699 }, { "epoch": 4.51, "learning_rate": 1.6472831267874167e-06, "loss": 1.7388, "step": 23700 }, { "epoch": 4.51, "learning_rate": 1.6472259294566253e-06, "loss": 1.9952, "step": 23701 }, { "epoch": 4.51, "learning_rate": 1.6471687321258343e-06, "loss": 2.0257, "step": 23702 }, { "epoch": 4.51, "learning_rate": 1.647111534795043e-06, "loss": 2.4502, "step": 23703 }, { "epoch": 4.52, "learning_rate": 1.6470543374642517e-06, "loss": 1.7332, "step": 23704 }, { "epoch": 4.52, "learning_rate": 1.6469971401334605e-06, "loss": 1.1604, "step": 23705 }, { "epoch": 4.52, "learning_rate": 1.6469399428026693e-06, "loss": 1.8725, "step": 23706 }, { "epoch": 4.52, "learning_rate": 1.6468827454718779e-06, "loss": 1.2983, "step": 23707 }, { "epoch": 4.52, "learning_rate": 1.6468255481410869e-06, "loss": 2.8555, "step": 23708 }, { "epoch": 4.52, "learning_rate": 1.6467683508102955e-06, "loss": 1.7019, "step": 23709 }, { "epoch": 4.52, "learning_rate": 1.6467111534795042e-06, "loss": 1.8072, "step": 23710 }, { "epoch": 4.52, "learning_rate": 1.6466539561487133e-06, "loss": 1.9259, "step": 23711 }, { "epoch": 4.52, "learning_rate": 1.6465967588179218e-06, "loss": 1.2066, "step": 23712 }, { "epoch": 4.52, "learning_rate": 1.6465395614871308e-06, "loss": 1.8772, "step": 23713 }, { "epoch": 4.52, "learning_rate": 1.6464823641563394e-06, "loss": 1.7084, "step": 23714 }, { "epoch": 4.52, "learning_rate": 1.6464251668255482e-06, "loss": 2.2602, "step": 23715 }, { "epoch": 4.52, "learning_rate": 1.646367969494757e-06, "loss": 2.5344, "step": 23716 }, { "epoch": 4.52, "learning_rate": 1.6463107721639658e-06, "loss": 1.6932, "step": 23717 }, { "epoch": 4.52, "learning_rate": 1.6462535748331744e-06, "loss": 2.2531, "step": 23718 }, { "epoch": 4.52, "learning_rate": 1.6461963775023834e-06, "loss": 3.1319, "step": 23719 }, { "epoch": 4.52, "learning_rate": 1.646139180171592e-06, "loss": 3.063, "step": 23720 }, { "epoch": 4.52, "learning_rate": 1.6460819828408008e-06, "loss": 1.8167, "step": 23721 }, { "epoch": 4.52, "learning_rate": 1.6460247855100098e-06, "loss": 2.5274, "step": 23722 }, { "epoch": 4.52, "learning_rate": 1.6459675881792184e-06, "loss": 1.9374, "step": 23723 }, { "epoch": 4.52, "learning_rate": 1.645910390848427e-06, "loss": 0.7073, "step": 23724 }, { "epoch": 4.52, "learning_rate": 1.645853193517636e-06, "loss": 1.8105, "step": 23725 }, { "epoch": 4.52, "learning_rate": 1.6457959961868445e-06, "loss": 1.8718, "step": 23726 }, { "epoch": 4.52, "learning_rate": 1.6457387988560533e-06, "loss": 2.0926, "step": 23727 }, { "epoch": 4.52, "learning_rate": 1.6456816015252623e-06, "loss": 1.6912, "step": 23728 }, { "epoch": 4.52, "learning_rate": 1.645624404194471e-06, "loss": 3.2015, "step": 23729 }, { "epoch": 4.52, "learning_rate": 1.64556720686368e-06, "loss": 2.4144, "step": 23730 }, { "epoch": 4.52, "learning_rate": 1.6455100095328885e-06, "loss": 2.145, "step": 23731 }, { "epoch": 4.52, "learning_rate": 1.6454528122020973e-06, "loss": 2.0568, "step": 23732 }, { "epoch": 4.52, "learning_rate": 1.645395614871306e-06, "loss": 1.2898, "step": 23733 }, { "epoch": 4.52, "learning_rate": 1.6453384175405149e-06, "loss": 2.7173, "step": 23734 }, { "epoch": 4.52, "learning_rate": 1.6452812202097235e-06, "loss": 0.94, "step": 23735 }, { "epoch": 4.52, "learning_rate": 1.6452240228789325e-06, "loss": 2.3991, "step": 23736 }, { "epoch": 4.52, "learning_rate": 1.645166825548141e-06, "loss": 2.4558, "step": 23737 }, { "epoch": 4.52, "learning_rate": 1.6451096282173498e-06, "loss": 2.8105, "step": 23738 }, { "epoch": 4.52, "learning_rate": 1.6450524308865588e-06, "loss": 1.1162, "step": 23739 }, { "epoch": 4.52, "learning_rate": 1.6449952335557674e-06, "loss": 2.786, "step": 23740 }, { "epoch": 4.52, "learning_rate": 1.644938036224976e-06, "loss": 3.2526, "step": 23741 }, { "epoch": 4.52, "learning_rate": 1.644880838894185e-06, "loss": 1.7352, "step": 23742 }, { "epoch": 4.52, "learning_rate": 1.6448236415633938e-06, "loss": 2.4666, "step": 23743 }, { "epoch": 4.52, "learning_rate": 1.6447664442326024e-06, "loss": 2.1967, "step": 23744 }, { "epoch": 4.52, "learning_rate": 1.6447092469018114e-06, "loss": 1.4438, "step": 23745 }, { "epoch": 4.52, "learning_rate": 1.64465204957102e-06, "loss": 2.3108, "step": 23746 }, { "epoch": 4.52, "learning_rate": 1.644594852240229e-06, "loss": 1.3203, "step": 23747 }, { "epoch": 4.52, "learning_rate": 1.6445376549094376e-06, "loss": 2.1831, "step": 23748 }, { "epoch": 4.52, "learning_rate": 1.6444804575786464e-06, "loss": 1.7372, "step": 23749 }, { "epoch": 4.52, "learning_rate": 1.6444232602478554e-06, "loss": 2.1119, "step": 23750 }, { "epoch": 4.52, "learning_rate": 1.644366062917064e-06, "loss": 1.2684, "step": 23751 }, { "epoch": 4.52, "learning_rate": 1.6443088655862725e-06, "loss": 1.1902, "step": 23752 }, { "epoch": 4.52, "learning_rate": 1.6442516682554815e-06, "loss": 3.2196, "step": 23753 }, { "epoch": 4.52, "learning_rate": 1.6441944709246901e-06, "loss": 3.1785, "step": 23754 }, { "epoch": 4.52, "learning_rate": 1.644137273593899e-06, "loss": 3.37, "step": 23755 }, { "epoch": 4.52, "learning_rate": 1.644080076263108e-06, "loss": 2.5969, "step": 23756 }, { "epoch": 4.53, "learning_rate": 1.6440228789323165e-06, "loss": 2.1391, "step": 23757 }, { "epoch": 4.53, "learning_rate": 1.643965681601525e-06, "loss": 1.9415, "step": 23758 }, { "epoch": 4.53, "learning_rate": 1.643908484270734e-06, "loss": 0.6233, "step": 23759 }, { "epoch": 4.53, "learning_rate": 1.6438512869399429e-06, "loss": 1.918, "step": 23760 }, { "epoch": 4.53, "learning_rate": 1.6437940896091515e-06, "loss": 2.3609, "step": 23761 }, { "epoch": 4.53, "learning_rate": 1.6437368922783605e-06, "loss": 1.9762, "step": 23762 }, { "epoch": 4.53, "learning_rate": 1.643679694947569e-06, "loss": 1.4619, "step": 23763 }, { "epoch": 4.53, "learning_rate": 1.643622497616778e-06, "loss": 2.0459, "step": 23764 }, { "epoch": 4.53, "learning_rate": 1.6435653002859866e-06, "loss": 1.7265, "step": 23765 }, { "epoch": 4.53, "learning_rate": 1.6435081029551954e-06, "loss": 1.5466, "step": 23766 }, { "epoch": 4.53, "learning_rate": 1.6434509056244044e-06, "loss": 2.7535, "step": 23767 }, { "epoch": 4.53, "learning_rate": 1.643393708293613e-06, "loss": 2.355, "step": 23768 }, { "epoch": 4.53, "learning_rate": 1.6433365109628216e-06, "loss": 2.6876, "step": 23769 }, { "epoch": 4.53, "learning_rate": 1.6432793136320306e-06, "loss": 3.0714, "step": 23770 }, { "epoch": 4.53, "learning_rate": 1.6432221163012394e-06, "loss": 1.6655, "step": 23771 }, { "epoch": 4.53, "learning_rate": 1.643164918970448e-06, "loss": 3.1755, "step": 23772 }, { "epoch": 4.53, "learning_rate": 1.643107721639657e-06, "loss": 2.28, "step": 23773 }, { "epoch": 4.53, "learning_rate": 1.6430505243088656e-06, "loss": 1.4477, "step": 23774 }, { "epoch": 4.53, "learning_rate": 1.6429933269780742e-06, "loss": 2.5853, "step": 23775 }, { "epoch": 4.53, "learning_rate": 1.6429361296472832e-06, "loss": 1.8008, "step": 23776 }, { "epoch": 4.53, "learning_rate": 1.642878932316492e-06, "loss": 2.9054, "step": 23777 }, { "epoch": 4.53, "learning_rate": 1.642821734985701e-06, "loss": 2.7664, "step": 23778 }, { "epoch": 4.53, "learning_rate": 1.6427645376549095e-06, "loss": 2.927, "step": 23779 }, { "epoch": 4.53, "learning_rate": 1.6427073403241181e-06, "loss": 1.2361, "step": 23780 }, { "epoch": 4.53, "learning_rate": 1.6426501429933271e-06, "loss": 2.0415, "step": 23781 }, { "epoch": 4.53, "learning_rate": 1.642592945662536e-06, "loss": 1.489, "step": 23782 }, { "epoch": 4.53, "learning_rate": 1.6425357483317445e-06, "loss": 2.1705, "step": 23783 }, { "epoch": 4.53, "learning_rate": 1.6424785510009535e-06, "loss": 2.6911, "step": 23784 }, { "epoch": 4.53, "learning_rate": 1.642421353670162e-06, "loss": 2.7348, "step": 23785 }, { "epoch": 4.53, "learning_rate": 1.6423641563393707e-06, "loss": 2.048, "step": 23786 }, { "epoch": 4.53, "learning_rate": 1.6423069590085797e-06, "loss": 1.478, "step": 23787 }, { "epoch": 4.53, "learning_rate": 1.6422497616777885e-06, "loss": 2.1363, "step": 23788 }, { "epoch": 4.53, "learning_rate": 1.642192564346997e-06, "loss": 1.981, "step": 23789 }, { "epoch": 4.53, "learning_rate": 1.642135367016206e-06, "loss": 2.5035, "step": 23790 }, { "epoch": 4.53, "learning_rate": 1.6420781696854146e-06, "loss": 1.1836, "step": 23791 }, { "epoch": 4.53, "learning_rate": 1.6420209723546234e-06, "loss": 1.4995, "step": 23792 }, { "epoch": 4.53, "learning_rate": 1.6419637750238322e-06, "loss": 2.2975, "step": 23793 }, { "epoch": 4.53, "learning_rate": 1.641906577693041e-06, "loss": 1.1677, "step": 23794 }, { "epoch": 4.53, "learning_rate": 1.64184938036225e-06, "loss": 2.5533, "step": 23795 }, { "epoch": 4.53, "learning_rate": 1.6417921830314586e-06, "loss": 1.5687, "step": 23796 }, { "epoch": 4.53, "learning_rate": 1.6417349857006672e-06, "loss": 3.2728, "step": 23797 }, { "epoch": 4.53, "learning_rate": 1.6416777883698762e-06, "loss": 1.6863, "step": 23798 }, { "epoch": 4.53, "learning_rate": 1.641620591039085e-06, "loss": 2.3708, "step": 23799 }, { "epoch": 4.53, "learning_rate": 1.6415633937082936e-06, "loss": 2.0699, "step": 23800 }, { "epoch": 4.53, "learning_rate": 1.6415061963775026e-06, "loss": 1.5995, "step": 23801 }, { "epoch": 4.53, "learning_rate": 1.6414489990467112e-06, "loss": 2.115, "step": 23802 }, { "epoch": 4.53, "learning_rate": 1.64139180171592e-06, "loss": 3.771, "step": 23803 }, { "epoch": 4.53, "learning_rate": 1.6413346043851287e-06, "loss": 2.4174, "step": 23804 }, { "epoch": 4.53, "learning_rate": 1.6412774070543375e-06, "loss": 2.1208, "step": 23805 }, { "epoch": 4.53, "learning_rate": 1.6412202097235461e-06, "loss": 2.4786, "step": 23806 }, { "epoch": 4.53, "learning_rate": 1.6411630123927551e-06, "loss": 2.4509, "step": 23807 }, { "epoch": 4.53, "learning_rate": 1.6411058150619637e-06, "loss": 1.1418, "step": 23808 }, { "epoch": 4.54, "learning_rate": 1.6410486177311725e-06, "loss": 3.315, "step": 23809 }, { "epoch": 4.54, "learning_rate": 1.6409914204003815e-06, "loss": 2.3093, "step": 23810 }, { "epoch": 4.54, "learning_rate": 1.64093422306959e-06, "loss": 2.5499, "step": 23811 }, { "epoch": 4.54, "learning_rate": 1.640877025738799e-06, "loss": 2.9188, "step": 23812 }, { "epoch": 4.54, "learning_rate": 1.6408198284080077e-06, "loss": 1.7605, "step": 23813 }, { "epoch": 4.54, "learning_rate": 1.6407626310772163e-06, "loss": 1.8148, "step": 23814 }, { "epoch": 4.54, "learning_rate": 1.6407054337464253e-06, "loss": 2.5542, "step": 23815 }, { "epoch": 4.54, "learning_rate": 1.640648236415634e-06, "loss": 1.0926, "step": 23816 }, { "epoch": 4.54, "learning_rate": 1.6405910390848426e-06, "loss": 2.0068, "step": 23817 }, { "epoch": 4.54, "learning_rate": 1.6405338417540516e-06, "loss": 2.3371, "step": 23818 }, { "epoch": 4.54, "learning_rate": 1.6404766444232602e-06, "loss": 2.407, "step": 23819 }, { "epoch": 4.54, "learning_rate": 1.640419447092469e-06, "loss": 0.804, "step": 23820 }, { "epoch": 4.54, "learning_rate": 1.6403622497616778e-06, "loss": 1.4759, "step": 23821 }, { "epoch": 4.54, "learning_rate": 1.6403050524308866e-06, "loss": 1.8202, "step": 23822 }, { "epoch": 4.54, "learning_rate": 1.6402478551000952e-06, "loss": 1.9271, "step": 23823 }, { "epoch": 4.54, "learning_rate": 1.6401906577693042e-06, "loss": 1.5937, "step": 23824 }, { "epoch": 4.54, "learning_rate": 1.6401334604385128e-06, "loss": 1.2025, "step": 23825 }, { "epoch": 4.54, "learning_rate": 1.6400762631077216e-06, "loss": 1.2854, "step": 23826 }, { "epoch": 4.54, "learning_rate": 1.6400190657769306e-06, "loss": 1.9128, "step": 23827 }, { "epoch": 4.54, "learning_rate": 1.6399618684461392e-06, "loss": 1.0696, "step": 23828 }, { "epoch": 4.54, "learning_rate": 1.6399046711153482e-06, "loss": 2.193, "step": 23829 }, { "epoch": 4.54, "learning_rate": 1.6398474737845568e-06, "loss": 1.3823, "step": 23830 }, { "epoch": 4.54, "learning_rate": 1.6397902764537655e-06, "loss": 2.1212, "step": 23831 }, { "epoch": 4.54, "learning_rate": 1.6397330791229743e-06, "loss": 2.029, "step": 23832 }, { "epoch": 4.54, "learning_rate": 1.6396758817921831e-06, "loss": 1.1984, "step": 23833 }, { "epoch": 4.54, "learning_rate": 1.6396186844613917e-06, "loss": 2.0827, "step": 23834 }, { "epoch": 4.54, "learning_rate": 1.6395614871306007e-06, "loss": 1.391, "step": 23835 }, { "epoch": 4.54, "learning_rate": 1.6395042897998093e-06, "loss": 2.4917, "step": 23836 }, { "epoch": 4.54, "learning_rate": 1.639447092469018e-06, "loss": 2.677, "step": 23837 }, { "epoch": 4.54, "learning_rate": 1.639389895138227e-06, "loss": 2.4164, "step": 23838 }, { "epoch": 4.54, "learning_rate": 1.6393326978074357e-06, "loss": 1.1024, "step": 23839 }, { "epoch": 4.54, "learning_rate": 1.6392755004766443e-06, "loss": 3.5884, "step": 23840 }, { "epoch": 4.54, "learning_rate": 1.6392183031458533e-06, "loss": 2.4119, "step": 23841 }, { "epoch": 4.54, "learning_rate": 1.6391611058150619e-06, "loss": 1.0075, "step": 23842 }, { "epoch": 4.54, "learning_rate": 1.6391039084842709e-06, "loss": 2.0685, "step": 23843 }, { "epoch": 4.54, "learning_rate": 1.6390467111534797e-06, "loss": 1.4524, "step": 23844 }, { "epoch": 4.54, "learning_rate": 1.6389895138226882e-06, "loss": 1.8876, "step": 23845 }, { "epoch": 4.54, "learning_rate": 1.6389323164918972e-06, "loss": 2.0128, "step": 23846 }, { "epoch": 4.54, "learning_rate": 1.6388751191611058e-06, "loss": 1.0785, "step": 23847 }, { "epoch": 4.54, "learning_rate": 1.6388179218303146e-06, "loss": 1.8052, "step": 23848 }, { "epoch": 4.54, "learning_rate": 1.6387607244995236e-06, "loss": 2.6715, "step": 23849 }, { "epoch": 4.54, "learning_rate": 1.6387035271687322e-06, "loss": 2.6012, "step": 23850 }, { "epoch": 4.54, "learning_rate": 1.6386463298379408e-06, "loss": 1.2035, "step": 23851 }, { "epoch": 4.54, "learning_rate": 1.6385891325071498e-06, "loss": 1.5414, "step": 23852 }, { "epoch": 4.54, "learning_rate": 1.6385319351763584e-06, "loss": 1.9029, "step": 23853 }, { "epoch": 4.54, "learning_rate": 1.6384747378455672e-06, "loss": 1.9038, "step": 23854 }, { "epoch": 4.54, "learning_rate": 1.6384175405147762e-06, "loss": 2.5861, "step": 23855 }, { "epoch": 4.54, "learning_rate": 1.6383603431839848e-06, "loss": 1.631, "step": 23856 }, { "epoch": 4.54, "learning_rate": 1.6383031458531933e-06, "loss": 2.2828, "step": 23857 }, { "epoch": 4.54, "learning_rate": 1.6382459485224023e-06, "loss": 1.3502, "step": 23858 }, { "epoch": 4.54, "learning_rate": 1.6381887511916111e-06, "loss": 2.1335, "step": 23859 }, { "epoch": 4.54, "learning_rate": 1.63813155386082e-06, "loss": 1.9998, "step": 23860 }, { "epoch": 4.54, "learning_rate": 1.6380743565300287e-06, "loss": 2.3087, "step": 23861 }, { "epoch": 4.55, "learning_rate": 1.6380171591992373e-06, "loss": 1.8876, "step": 23862 }, { "epoch": 4.55, "learning_rate": 1.6379599618684463e-06, "loss": 1.5532, "step": 23863 }, { "epoch": 4.55, "learning_rate": 1.637902764537655e-06, "loss": 3.0308, "step": 23864 }, { "epoch": 4.55, "learning_rate": 1.6378455672068637e-06, "loss": 1.2584, "step": 23865 }, { "epoch": 4.55, "learning_rate": 1.6377883698760727e-06, "loss": 1.5174, "step": 23866 }, { "epoch": 4.55, "learning_rate": 1.6377311725452813e-06, "loss": 1.8373, "step": 23867 }, { "epoch": 4.55, "learning_rate": 1.6376739752144899e-06, "loss": 1.7724, "step": 23868 }, { "epoch": 4.55, "learning_rate": 1.6376167778836989e-06, "loss": 2.3897, "step": 23869 }, { "epoch": 4.55, "learning_rate": 1.6375595805529077e-06, "loss": 3.0666, "step": 23870 }, { "epoch": 4.55, "learning_rate": 1.6375023832221162e-06, "loss": 1.3039, "step": 23871 }, { "epoch": 4.55, "learning_rate": 1.6374451858913252e-06, "loss": 2.1662, "step": 23872 }, { "epoch": 4.55, "learning_rate": 1.6373879885605338e-06, "loss": 2.4637, "step": 23873 }, { "epoch": 4.55, "learning_rate": 1.6373307912297424e-06, "loss": 1.5033, "step": 23874 }, { "epoch": 4.55, "learning_rate": 1.6372735938989514e-06, "loss": 1.578, "step": 23875 }, { "epoch": 4.55, "learning_rate": 1.6372163965681602e-06, "loss": 2.0903, "step": 23876 }, { "epoch": 4.55, "learning_rate": 1.6371591992373692e-06, "loss": 2.2485, "step": 23877 }, { "epoch": 4.55, "learning_rate": 1.6371020019065778e-06, "loss": 1.0068, "step": 23878 }, { "epoch": 4.55, "learning_rate": 1.6370448045757864e-06, "loss": 1.7777, "step": 23879 }, { "epoch": 4.55, "learning_rate": 1.6369876072449954e-06, "loss": 2.7296, "step": 23880 }, { "epoch": 4.55, "learning_rate": 1.636930409914204e-06, "loss": 2.3253, "step": 23881 }, { "epoch": 4.55, "learning_rate": 1.6368732125834128e-06, "loss": 2.468, "step": 23882 }, { "epoch": 4.55, "learning_rate": 1.6368160152526218e-06, "loss": 2.4494, "step": 23883 }, { "epoch": 4.55, "learning_rate": 1.6367588179218303e-06, "loss": 1.4117, "step": 23884 }, { "epoch": 4.55, "learning_rate": 1.636701620591039e-06, "loss": 3.3506, "step": 23885 }, { "epoch": 4.55, "learning_rate": 1.636644423260248e-06, "loss": 0.9933, "step": 23886 }, { "epoch": 4.55, "learning_rate": 1.6365872259294567e-06, "loss": 2.3246, "step": 23887 }, { "epoch": 4.55, "learning_rate": 1.6365300285986653e-06, "loss": 1.6368, "step": 23888 }, { "epoch": 4.55, "learning_rate": 1.6364728312678743e-06, "loss": 1.6727, "step": 23889 }, { "epoch": 4.55, "learning_rate": 1.636415633937083e-06, "loss": 2.5644, "step": 23890 }, { "epoch": 4.55, "learning_rate": 1.6363584366062915e-06, "loss": 1.8884, "step": 23891 }, { "epoch": 4.55, "learning_rate": 1.6363012392755005e-06, "loss": 1.871, "step": 23892 }, { "epoch": 4.55, "learning_rate": 1.6362440419447093e-06, "loss": 2.1138, "step": 23893 }, { "epoch": 4.55, "learning_rate": 1.6361868446139183e-06, "loss": 1.8286, "step": 23894 }, { "epoch": 4.55, "learning_rate": 1.6361296472831269e-06, "loss": 1.3501, "step": 23895 }, { "epoch": 4.55, "learning_rate": 1.6360724499523354e-06, "loss": 2.3166, "step": 23896 }, { "epoch": 4.55, "learning_rate": 1.6360152526215445e-06, "loss": 1.0843, "step": 23897 }, { "epoch": 4.55, "learning_rate": 1.6359580552907532e-06, "loss": 2.0261, "step": 23898 }, { "epoch": 4.55, "learning_rate": 1.6359008579599618e-06, "loss": 2.8823, "step": 23899 }, { "epoch": 4.55, "learning_rate": 1.6358436606291708e-06, "loss": 1.5697, "step": 23900 }, { "epoch": 4.55, "learning_rate": 1.6357864632983794e-06, "loss": 3.4975, "step": 23901 }, { "epoch": 4.55, "learning_rate": 1.635729265967588e-06, "loss": 1.8406, "step": 23902 }, { "epoch": 4.55, "learning_rate": 1.635672068636797e-06, "loss": 1.485, "step": 23903 }, { "epoch": 4.55, "learning_rate": 1.6356148713060058e-06, "loss": 2.9721, "step": 23904 }, { "epoch": 4.55, "learning_rate": 1.6355576739752144e-06, "loss": 0.8717, "step": 23905 }, { "epoch": 4.55, "learning_rate": 1.6355004766444234e-06, "loss": 2.3231, "step": 23906 }, { "epoch": 4.55, "learning_rate": 1.635443279313632e-06, "loss": 2.9159, "step": 23907 }, { "epoch": 4.55, "learning_rate": 1.635386081982841e-06, "loss": 1.9918, "step": 23908 }, { "epoch": 4.55, "learning_rate": 1.6353288846520496e-06, "loss": 3.6731, "step": 23909 }, { "epoch": 4.55, "learning_rate": 1.6352716873212584e-06, "loss": 2.1992, "step": 23910 }, { "epoch": 4.55, "learning_rate": 1.6352144899904674e-06, "loss": 2.6076, "step": 23911 }, { "epoch": 4.55, "learning_rate": 1.635157292659676e-06, "loss": 3.055, "step": 23912 }, { "epoch": 4.55, "learning_rate": 1.6351000953288845e-06, "loss": 1.9725, "step": 23913 }, { "epoch": 4.56, "learning_rate": 1.6350428979980935e-06, "loss": 1.3814, "step": 23914 }, { "epoch": 4.56, "learning_rate": 1.6349857006673023e-06, "loss": 2.1847, "step": 23915 }, { "epoch": 4.56, "learning_rate": 1.634928503336511e-06, "loss": 2.0984, "step": 23916 }, { "epoch": 4.56, "learning_rate": 1.63487130600572e-06, "loss": 3.1299, "step": 23917 }, { "epoch": 4.56, "learning_rate": 1.6348141086749285e-06, "loss": 1.5683, "step": 23918 }, { "epoch": 4.56, "learning_rate": 1.6347569113441373e-06, "loss": 2.8144, "step": 23919 }, { "epoch": 4.56, "learning_rate": 1.634699714013346e-06, "loss": 2.0488, "step": 23920 }, { "epoch": 4.56, "learning_rate": 1.6346425166825549e-06, "loss": 2.102, "step": 23921 }, { "epoch": 4.56, "learning_rate": 1.6345853193517635e-06, "loss": 1.8702, "step": 23922 }, { "epoch": 4.56, "learning_rate": 1.6345281220209725e-06, "loss": 1.9497, "step": 23923 }, { "epoch": 4.56, "learning_rate": 1.634470924690181e-06, "loss": 2.2482, "step": 23924 }, { "epoch": 4.56, "learning_rate": 1.63441372735939e-06, "loss": 1.8017, "step": 23925 }, { "epoch": 4.56, "learning_rate": 1.6343565300285988e-06, "loss": 2.7366, "step": 23926 }, { "epoch": 4.56, "learning_rate": 1.6342993326978074e-06, "loss": 2.0578, "step": 23927 }, { "epoch": 4.56, "learning_rate": 1.6342421353670164e-06, "loss": 1.9038, "step": 23928 }, { "epoch": 4.56, "learning_rate": 1.634184938036225e-06, "loss": 1.9122, "step": 23929 }, { "epoch": 4.56, "learning_rate": 1.6341277407054336e-06, "loss": 2.0831, "step": 23930 }, { "epoch": 4.56, "learning_rate": 1.6340705433746426e-06, "loss": 3.2099, "step": 23931 }, { "epoch": 4.56, "learning_rate": 1.6340133460438514e-06, "loss": 1.5092, "step": 23932 }, { "epoch": 4.56, "learning_rate": 1.63395614871306e-06, "loss": 3.4372, "step": 23933 }, { "epoch": 4.56, "learning_rate": 1.633898951382269e-06, "loss": 2.6008, "step": 23934 }, { "epoch": 4.56, "learning_rate": 1.6338417540514776e-06, "loss": 2.3386, "step": 23935 }, { "epoch": 4.56, "learning_rate": 1.6337845567206864e-06, "loss": 2.7486, "step": 23936 }, { "epoch": 4.56, "learning_rate": 1.6337273593898954e-06, "loss": 2.0977, "step": 23937 }, { "epoch": 4.56, "learning_rate": 1.633670162059104e-06, "loss": 2.289, "step": 23938 }, { "epoch": 4.56, "learning_rate": 1.6336129647283125e-06, "loss": 1.54, "step": 23939 }, { "epoch": 4.56, "learning_rate": 1.6335557673975215e-06, "loss": 1.7785, "step": 23940 }, { "epoch": 4.56, "learning_rate": 1.6334985700667301e-06, "loss": 2.2946, "step": 23941 }, { "epoch": 4.56, "learning_rate": 1.6334413727359391e-06, "loss": 2.8799, "step": 23942 }, { "epoch": 4.56, "learning_rate": 1.633384175405148e-06, "loss": 1.5493, "step": 23943 }, { "epoch": 4.56, "learning_rate": 1.6333269780743565e-06, "loss": 2.7726, "step": 23944 }, { "epoch": 4.56, "learning_rate": 1.6332697807435655e-06, "loss": 2.6541, "step": 23945 }, { "epoch": 4.56, "learning_rate": 1.633212583412774e-06, "loss": 2.3961, "step": 23946 }, { "epoch": 4.56, "learning_rate": 1.6331553860819829e-06, "loss": 2.4738, "step": 23947 }, { "epoch": 4.56, "learning_rate": 1.6330981887511917e-06, "loss": 2.5747, "step": 23948 }, { "epoch": 4.56, "learning_rate": 1.6330409914204005e-06, "loss": 2.256, "step": 23949 }, { "epoch": 4.56, "learning_rate": 1.632983794089609e-06, "loss": 2.3088, "step": 23950 }, { "epoch": 4.56, "learning_rate": 1.632926596758818e-06, "loss": 1.7515, "step": 23951 }, { "epoch": 4.56, "learning_rate": 1.6328693994280266e-06, "loss": 2.2186, "step": 23952 }, { "epoch": 4.56, "learning_rate": 1.6328122020972354e-06, "loss": 1.8421, "step": 23953 }, { "epoch": 4.56, "learning_rate": 1.6327550047664444e-06, "loss": 1.2016, "step": 23954 }, { "epoch": 4.56, "learning_rate": 1.632697807435653e-06, "loss": 1.9768, "step": 23955 }, { "epoch": 4.56, "learning_rate": 1.6326406101048616e-06, "loss": 1.9121, "step": 23956 }, { "epoch": 4.56, "learning_rate": 1.6325834127740706e-06, "loss": 3.5518, "step": 23957 }, { "epoch": 4.56, "learning_rate": 1.6325262154432794e-06, "loss": 1.6163, "step": 23958 }, { "epoch": 4.56, "learning_rate": 1.6324690181124882e-06, "loss": 2.5413, "step": 23959 }, { "epoch": 4.56, "learning_rate": 1.632411820781697e-06, "loss": 1.7066, "step": 23960 }, { "epoch": 4.56, "learning_rate": 1.6323546234509056e-06, "loss": 3.0907, "step": 23961 }, { "epoch": 4.56, "learning_rate": 1.6322974261201146e-06, "loss": 2.7581, "step": 23962 }, { "epoch": 4.56, "learning_rate": 1.6322402287893232e-06, "loss": 1.6573, "step": 23963 }, { "epoch": 4.56, "learning_rate": 1.632183031458532e-06, "loss": 2.1765, "step": 23964 }, { "epoch": 4.56, "learning_rate": 1.632125834127741e-06, "loss": 2.4489, "step": 23965 }, { "epoch": 4.56, "learning_rate": 1.6320686367969495e-06, "loss": 2.1899, "step": 23966 }, { "epoch": 4.57, "learning_rate": 1.6320114394661581e-06, "loss": 1.4825, "step": 23967 }, { "epoch": 4.57, "learning_rate": 1.6319542421353671e-06, "loss": 1.5087, "step": 23968 }, { "epoch": 4.57, "learning_rate": 1.6318970448045757e-06, "loss": 2.4317, "step": 23969 }, { "epoch": 4.57, "learning_rate": 1.6318398474737845e-06, "loss": 3.0063, "step": 23970 }, { "epoch": 4.57, "learning_rate": 1.6317826501429935e-06, "loss": 2.4554, "step": 23971 }, { "epoch": 4.57, "learning_rate": 1.631725452812202e-06, "loss": 1.9912, "step": 23972 }, { "epoch": 4.57, "learning_rate": 1.631668255481411e-06, "loss": 2.0145, "step": 23973 }, { "epoch": 4.57, "learning_rate": 1.6316110581506197e-06, "loss": 3.4882, "step": 23974 }, { "epoch": 4.57, "learning_rate": 1.6315538608198285e-06, "loss": 2.3026, "step": 23975 }, { "epoch": 4.57, "learning_rate": 1.6314966634890373e-06, "loss": 1.5901, "step": 23976 }, { "epoch": 4.57, "learning_rate": 1.631439466158246e-06, "loss": 1.7384, "step": 23977 }, { "epoch": 4.57, "learning_rate": 1.6313822688274546e-06, "loss": 3.3357, "step": 23978 }, { "epoch": 4.57, "learning_rate": 1.6313250714966636e-06, "loss": 1.8518, "step": 23979 }, { "epoch": 4.57, "learning_rate": 1.6312678741658722e-06, "loss": 1.6386, "step": 23980 }, { "epoch": 4.57, "learning_rate": 1.631210676835081e-06, "loss": 2.4435, "step": 23981 }, { "epoch": 4.57, "learning_rate": 1.63115347950429e-06, "loss": 2.4373, "step": 23982 }, { "epoch": 4.57, "learning_rate": 1.6310962821734986e-06, "loss": 1.7295, "step": 23983 }, { "epoch": 4.57, "learning_rate": 1.6310390848427072e-06, "loss": 2.3148, "step": 23984 }, { "epoch": 4.57, "learning_rate": 1.6309818875119162e-06, "loss": 2.7733, "step": 23985 }, { "epoch": 4.57, "learning_rate": 1.630924690181125e-06, "loss": 1.7872, "step": 23986 }, { "epoch": 4.57, "learning_rate": 1.6308674928503336e-06, "loss": 2.1826, "step": 23987 }, { "epoch": 4.57, "learning_rate": 1.6308102955195426e-06, "loss": 1.7077, "step": 23988 }, { "epoch": 4.57, "learning_rate": 1.6307530981887512e-06, "loss": 1.8383, "step": 23989 }, { "epoch": 4.57, "learning_rate": 1.6306959008579602e-06, "loss": 1.7151, "step": 23990 }, { "epoch": 4.57, "learning_rate": 1.6306387035271687e-06, "loss": 2.6051, "step": 23991 }, { "epoch": 4.57, "learning_rate": 1.6305815061963775e-06, "loss": 2.9108, "step": 23992 }, { "epoch": 4.57, "learning_rate": 1.6305243088655865e-06, "loss": 2.2994, "step": 23993 }, { "epoch": 4.57, "learning_rate": 1.6304671115347951e-06, "loss": 1.89, "step": 23994 }, { "epoch": 4.57, "learning_rate": 1.6304099142040037e-06, "loss": 2.3825, "step": 23995 }, { "epoch": 4.57, "learning_rate": 1.6303527168732127e-06, "loss": 1.7491, "step": 23996 }, { "epoch": 4.57, "learning_rate": 1.6302955195424213e-06, "loss": 2.2463, "step": 23997 }, { "epoch": 4.57, "learning_rate": 1.63023832221163e-06, "loss": 1.4766, "step": 23998 }, { "epoch": 4.57, "learning_rate": 1.630181124880839e-06, "loss": 1.5867, "step": 23999 }, { "epoch": 4.57, "learning_rate": 1.6301239275500477e-06, "loss": 1.5694, "step": 24000 }, { "epoch": 4.57, "learning_rate": 1.6300667302192563e-06, "loss": 1.7703, "step": 24001 }, { "epoch": 4.57, "learning_rate": 1.6300095328884653e-06, "loss": 2.5202, "step": 24002 }, { "epoch": 4.57, "learning_rate": 1.629952335557674e-06, "loss": 1.2615, "step": 24003 }, { "epoch": 4.57, "learning_rate": 1.6298951382268826e-06, "loss": 0.9624, "step": 24004 }, { "epoch": 4.57, "learning_rate": 1.6298379408960916e-06, "loss": 2.0066, "step": 24005 }, { "epoch": 4.57, "learning_rate": 1.6297807435653002e-06, "loss": 1.3868, "step": 24006 }, { "epoch": 4.57, "learning_rate": 1.6297235462345092e-06, "loss": 2.3817, "step": 24007 }, { "epoch": 4.57, "learning_rate": 1.6296663489037178e-06, "loss": 2.5297, "step": 24008 }, { "epoch": 4.57, "learning_rate": 1.6296091515729266e-06, "loss": 1.7773, "step": 24009 }, { "epoch": 4.57, "learning_rate": 1.6295519542421356e-06, "loss": 1.7096, "step": 24010 }, { "epoch": 4.57, "learning_rate": 1.6294947569113442e-06, "loss": 3.0767, "step": 24011 }, { "epoch": 4.57, "learning_rate": 1.6294375595805528e-06, "loss": 1.8861, "step": 24012 }, { "epoch": 4.57, "learning_rate": 1.6293803622497618e-06, "loss": 2.6019, "step": 24013 }, { "epoch": 4.57, "learning_rate": 1.6293231649189706e-06, "loss": 2.0424, "step": 24014 }, { "epoch": 4.57, "learning_rate": 1.6292659675881792e-06, "loss": 1.7664, "step": 24015 }, { "epoch": 4.57, "learning_rate": 1.6292087702573882e-06, "loss": 1.8639, "step": 24016 }, { "epoch": 4.57, "learning_rate": 1.6291515729265967e-06, "loss": 1.7475, "step": 24017 }, { "epoch": 4.57, "learning_rate": 1.6290943755958053e-06, "loss": 1.8283, "step": 24018 }, { "epoch": 4.58, "learning_rate": 1.6290371782650143e-06, "loss": 0.8643, "step": 24019 }, { "epoch": 4.58, "learning_rate": 1.6289799809342231e-06, "loss": 3.39, "step": 24020 }, { "epoch": 4.58, "learning_rate": 1.6289227836034317e-06, "loss": 1.1906, "step": 24021 }, { "epoch": 4.58, "learning_rate": 1.6288655862726407e-06, "loss": 1.5397, "step": 24022 }, { "epoch": 4.58, "learning_rate": 1.6288083889418493e-06, "loss": 1.0608, "step": 24023 }, { "epoch": 4.58, "learning_rate": 1.6287511916110583e-06, "loss": 1.8796, "step": 24024 }, { "epoch": 4.58, "learning_rate": 1.628693994280267e-06, "loss": 1.3767, "step": 24025 }, { "epoch": 4.58, "learning_rate": 1.6286367969494757e-06, "loss": 3.1235, "step": 24026 }, { "epoch": 4.58, "learning_rate": 1.6285795996186847e-06, "loss": 3.294, "step": 24027 }, { "epoch": 4.58, "learning_rate": 1.6285224022878933e-06, "loss": 1.6835, "step": 24028 }, { "epoch": 4.58, "learning_rate": 1.6284652049571019e-06, "loss": 1.6132, "step": 24029 }, { "epoch": 4.58, "learning_rate": 1.6284080076263109e-06, "loss": 2.9619, "step": 24030 }, { "epoch": 4.58, "learning_rate": 1.6283508102955197e-06, "loss": 2.4723, "step": 24031 }, { "epoch": 4.58, "learning_rate": 1.6282936129647282e-06, "loss": 2.9113, "step": 24032 }, { "epoch": 4.58, "learning_rate": 1.6282364156339372e-06, "loss": 1.4369, "step": 24033 }, { "epoch": 4.58, "learning_rate": 1.6281792183031458e-06, "loss": 3.3846, "step": 24034 }, { "epoch": 4.58, "learning_rate": 1.6281220209723546e-06, "loss": 2.3127, "step": 24035 }, { "epoch": 4.58, "learning_rate": 1.6280648236415634e-06, "loss": 2.571, "step": 24036 }, { "epoch": 4.58, "learning_rate": 1.6280076263107722e-06, "loss": 1.6969, "step": 24037 }, { "epoch": 4.58, "learning_rate": 1.6279504289799808e-06, "loss": 1.6668, "step": 24038 }, { "epoch": 4.58, "learning_rate": 1.6278932316491898e-06, "loss": 1.8695, "step": 24039 }, { "epoch": 4.58, "learning_rate": 1.6278360343183984e-06, "loss": 2.8415, "step": 24040 }, { "epoch": 4.58, "learning_rate": 1.6277788369876074e-06, "loss": 2.0827, "step": 24041 }, { "epoch": 4.58, "learning_rate": 1.6277216396568162e-06, "loss": 3.7579, "step": 24042 }, { "epoch": 4.58, "learning_rate": 1.6276644423260248e-06, "loss": 2.2643, "step": 24043 }, { "epoch": 4.58, "learning_rate": 1.6276072449952338e-06, "loss": 2.8028, "step": 24044 }, { "epoch": 4.58, "learning_rate": 1.6275500476644423e-06, "loss": 1.5319, "step": 24045 }, { "epoch": 4.58, "learning_rate": 1.627492850333651e-06, "loss": 1.3373, "step": 24046 }, { "epoch": 4.58, "learning_rate": 1.62743565300286e-06, "loss": 2.2503, "step": 24047 }, { "epoch": 4.58, "learning_rate": 1.6273784556720687e-06, "loss": 2.373, "step": 24048 }, { "epoch": 4.58, "learning_rate": 1.6273212583412773e-06, "loss": 1.1447, "step": 24049 }, { "epoch": 4.58, "learning_rate": 1.6272640610104863e-06, "loss": 1.7873, "step": 24050 }, { "epoch": 4.58, "learning_rate": 1.6272068636796949e-06, "loss": 2.5471, "step": 24051 }, { "epoch": 4.58, "learning_rate": 1.6271496663489037e-06, "loss": 1.7502, "step": 24052 }, { "epoch": 4.58, "learning_rate": 1.6270924690181127e-06, "loss": 1.9679, "step": 24053 }, { "epoch": 4.58, "learning_rate": 1.6270352716873213e-06, "loss": 2.0007, "step": 24054 }, { "epoch": 4.58, "learning_rate": 1.6269780743565303e-06, "loss": 1.6035, "step": 24055 }, { "epoch": 4.58, "learning_rate": 1.6269208770257389e-06, "loss": 2.3904, "step": 24056 }, { "epoch": 4.58, "learning_rate": 1.6268636796949474e-06, "loss": 1.9509, "step": 24057 }, { "epoch": 4.58, "learning_rate": 1.6268064823641564e-06, "loss": 1.8639, "step": 24058 }, { "epoch": 4.58, "learning_rate": 1.6267492850333652e-06, "loss": 1.3504, "step": 24059 }, { "epoch": 4.58, "learning_rate": 1.6266920877025738e-06, "loss": 2.3831, "step": 24060 }, { "epoch": 4.58, "learning_rate": 1.6266348903717828e-06, "loss": 2.6579, "step": 24061 }, { "epoch": 4.58, "learning_rate": 1.6265776930409914e-06, "loss": 1.6638, "step": 24062 }, { "epoch": 4.58, "learning_rate": 1.6265204957102002e-06, "loss": 1.8783, "step": 24063 }, { "epoch": 4.58, "learning_rate": 1.626463298379409e-06, "loss": 2.5807, "step": 24064 }, { "epoch": 4.58, "learning_rate": 1.6264061010486178e-06, "loss": 2.2919, "step": 24065 }, { "epoch": 4.58, "learning_rate": 1.6263489037178264e-06, "loss": 3.7891, "step": 24066 }, { "epoch": 4.58, "learning_rate": 1.6262917063870354e-06, "loss": 2.5986, "step": 24067 }, { "epoch": 4.58, "learning_rate": 1.626234509056244e-06, "loss": 2.773, "step": 24068 }, { "epoch": 4.58, "learning_rate": 1.6261773117254528e-06, "loss": 1.5481, "step": 24069 }, { "epoch": 4.58, "learning_rate": 1.6261201143946618e-06, "loss": 1.938, "step": 24070 }, { "epoch": 4.58, "learning_rate": 1.6260629170638703e-06, "loss": 2.6713, "step": 24071 }, { "epoch": 4.59, "learning_rate": 1.6260057197330793e-06, "loss": 2.2233, "step": 24072 }, { "epoch": 4.59, "learning_rate": 1.625948522402288e-06, "loss": 2.404, "step": 24073 }, { "epoch": 4.59, "learning_rate": 1.6258913250714967e-06, "loss": 1.7308, "step": 24074 }, { "epoch": 4.59, "learning_rate": 1.6258341277407055e-06, "loss": 1.7751, "step": 24075 }, { "epoch": 4.59, "learning_rate": 1.6257769304099143e-06, "loss": 1.9319, "step": 24076 }, { "epoch": 4.59, "learning_rate": 1.625719733079123e-06, "loss": 1.7617, "step": 24077 }, { "epoch": 4.59, "learning_rate": 1.625662535748332e-06, "loss": 1.3945, "step": 24078 }, { "epoch": 4.59, "learning_rate": 1.6256053384175405e-06, "loss": 2.2551, "step": 24079 }, { "epoch": 4.59, "learning_rate": 1.6255481410867493e-06, "loss": 3.2422, "step": 24080 }, { "epoch": 4.59, "learning_rate": 1.6254909437559583e-06, "loss": 0.882, "step": 24081 }, { "epoch": 4.59, "learning_rate": 1.6254337464251669e-06, "loss": 2.0283, "step": 24082 }, { "epoch": 4.59, "learning_rate": 1.6253765490943754e-06, "loss": 1.6283, "step": 24083 }, { "epoch": 4.59, "learning_rate": 1.6253193517635845e-06, "loss": 1.3382, "step": 24084 }, { "epoch": 4.59, "learning_rate": 1.625262154432793e-06, "loss": 2.1444, "step": 24085 }, { "epoch": 4.59, "learning_rate": 1.6252049571020018e-06, "loss": 2.197, "step": 24086 }, { "epoch": 4.59, "learning_rate": 1.6251477597712108e-06, "loss": 2.1273, "step": 24087 }, { "epoch": 4.59, "learning_rate": 1.6250905624404194e-06, "loss": 2.4024, "step": 24088 }, { "epoch": 4.59, "learning_rate": 1.6250333651096284e-06, "loss": 2.042, "step": 24089 }, { "epoch": 4.59, "learning_rate": 1.624976167778837e-06, "loss": 3.8012, "step": 24090 }, { "epoch": 4.59, "learning_rate": 1.6249189704480458e-06, "loss": 2.7373, "step": 24091 }, { "epoch": 4.59, "learning_rate": 1.6248617731172548e-06, "loss": 1.8475, "step": 24092 }, { "epoch": 4.59, "learning_rate": 1.6248045757864634e-06, "loss": 2.9752, "step": 24093 }, { "epoch": 4.59, "learning_rate": 1.624747378455672e-06, "loss": 2.4258, "step": 24094 }, { "epoch": 4.59, "learning_rate": 1.624690181124881e-06, "loss": 1.0919, "step": 24095 }, { "epoch": 4.59, "learning_rate": 1.6246329837940896e-06, "loss": 2.2001, "step": 24096 }, { "epoch": 4.59, "learning_rate": 1.6245757864632983e-06, "loss": 2.0902, "step": 24097 }, { "epoch": 4.59, "learning_rate": 1.6245185891325074e-06, "loss": 1.9336, "step": 24098 }, { "epoch": 4.59, "learning_rate": 1.624461391801716e-06, "loss": 2.0271, "step": 24099 }, { "epoch": 4.59, "learning_rate": 1.6244041944709245e-06, "loss": 1.5222, "step": 24100 }, { "epoch": 4.59, "learning_rate": 1.6243469971401335e-06, "loss": 2.68, "step": 24101 }, { "epoch": 4.59, "learning_rate": 1.6242897998093423e-06, "loss": 2.6484, "step": 24102 }, { "epoch": 4.59, "learning_rate": 1.624232602478551e-06, "loss": 2.9781, "step": 24103 }, { "epoch": 4.59, "learning_rate": 1.62417540514776e-06, "loss": 1.3565, "step": 24104 }, { "epoch": 4.59, "learning_rate": 1.6241182078169685e-06, "loss": 2.6064, "step": 24105 }, { "epoch": 4.59, "learning_rate": 1.6240610104861775e-06, "loss": 2.3067, "step": 24106 }, { "epoch": 4.59, "learning_rate": 1.624003813155386e-06, "loss": 2.0454, "step": 24107 }, { "epoch": 4.59, "learning_rate": 1.6239466158245949e-06, "loss": 2.4135, "step": 24108 }, { "epoch": 4.59, "learning_rate": 1.6238894184938039e-06, "loss": 2.0694, "step": 24109 }, { "epoch": 4.59, "learning_rate": 1.6238322211630125e-06, "loss": 2.3817, "step": 24110 }, { "epoch": 4.59, "learning_rate": 1.623775023832221e-06, "loss": 3.0264, "step": 24111 }, { "epoch": 4.59, "learning_rate": 1.62371782650143e-06, "loss": 1.4968, "step": 24112 }, { "epoch": 4.59, "learning_rate": 1.6236606291706388e-06, "loss": 1.6535, "step": 24113 }, { "epoch": 4.59, "learning_rate": 1.6236034318398474e-06, "loss": 2.3385, "step": 24114 }, { "epoch": 4.59, "learning_rate": 1.6235462345090564e-06, "loss": 2.1718, "step": 24115 }, { "epoch": 4.59, "learning_rate": 1.623489037178265e-06, "loss": 1.6818, "step": 24116 }, { "epoch": 4.59, "learning_rate": 1.6234318398474736e-06, "loss": 2.5349, "step": 24117 }, { "epoch": 4.59, "learning_rate": 1.6233746425166826e-06, "loss": 1.3039, "step": 24118 }, { "epoch": 4.59, "learning_rate": 1.6233174451858914e-06, "loss": 3.8206, "step": 24119 }, { "epoch": 4.59, "learning_rate": 1.6232602478551004e-06, "loss": 2.3124, "step": 24120 }, { "epoch": 4.59, "learning_rate": 1.623203050524309e-06, "loss": 2.3446, "step": 24121 }, { "epoch": 4.59, "learning_rate": 1.6231458531935176e-06, "loss": 1.7625, "step": 24122 }, { "epoch": 4.59, "learning_rate": 1.6230886558627266e-06, "loss": 1.3047, "step": 24123 }, { "epoch": 4.6, "learning_rate": 1.6230314585319351e-06, "loss": 2.8576, "step": 24124 }, { "epoch": 4.6, "learning_rate": 1.622974261201144e-06, "loss": 2.7284, "step": 24125 }, { "epoch": 4.6, "learning_rate": 1.622917063870353e-06, "loss": 3.0446, "step": 24126 }, { "epoch": 4.6, "learning_rate": 1.6228598665395615e-06, "loss": 1.5996, "step": 24127 }, { "epoch": 4.6, "learning_rate": 1.6228026692087701e-06, "loss": 1.2424, "step": 24128 }, { "epoch": 4.6, "learning_rate": 1.6227454718779791e-06, "loss": 2.5986, "step": 24129 }, { "epoch": 4.6, "learning_rate": 1.622688274547188e-06, "loss": 2.5293, "step": 24130 }, { "epoch": 4.6, "learning_rate": 1.6226310772163965e-06, "loss": 2.4951, "step": 24131 }, { "epoch": 4.6, "learning_rate": 1.6225738798856055e-06, "loss": 3.302, "step": 24132 }, { "epoch": 4.6, "learning_rate": 1.622516682554814e-06, "loss": 3.4359, "step": 24133 }, { "epoch": 4.6, "learning_rate": 1.6224594852240227e-06, "loss": 2.6001, "step": 24134 }, { "epoch": 4.6, "learning_rate": 1.6224022878932317e-06, "loss": 1.9773, "step": 24135 }, { "epoch": 4.6, "learning_rate": 1.6223450905624405e-06, "loss": 1.5153, "step": 24136 }, { "epoch": 4.6, "learning_rate": 1.6222878932316495e-06, "loss": 1.2586, "step": 24137 }, { "epoch": 4.6, "learning_rate": 1.622230695900858e-06, "loss": 1.7585, "step": 24138 }, { "epoch": 4.6, "learning_rate": 1.6221734985700666e-06, "loss": 1.6241, "step": 24139 }, { "epoch": 4.6, "learning_rate": 1.6221163012392756e-06, "loss": 0.8398, "step": 24140 }, { "epoch": 4.6, "learning_rate": 1.6220591039084844e-06, "loss": 1.8481, "step": 24141 }, { "epoch": 4.6, "learning_rate": 1.622001906577693e-06, "loss": 1.0431, "step": 24142 }, { "epoch": 4.6, "learning_rate": 1.621944709246902e-06, "loss": 2.7837, "step": 24143 }, { "epoch": 4.6, "learning_rate": 1.6218875119161106e-06, "loss": 2.05, "step": 24144 }, { "epoch": 4.6, "learning_rate": 1.6218303145853192e-06, "loss": 2.6521, "step": 24145 }, { "epoch": 4.6, "learning_rate": 1.6217731172545282e-06, "loss": 2.3713, "step": 24146 }, { "epoch": 4.6, "learning_rate": 1.621715919923737e-06, "loss": 2.256, "step": 24147 }, { "epoch": 4.6, "learning_rate": 1.6216587225929456e-06, "loss": 2.1406, "step": 24148 }, { "epoch": 4.6, "learning_rate": 1.6216015252621546e-06, "loss": 2.8712, "step": 24149 }, { "epoch": 4.6, "learning_rate": 1.6215443279313632e-06, "loss": 1.4189, "step": 24150 }, { "epoch": 4.6, "learning_rate": 1.621487130600572e-06, "loss": 1.9222, "step": 24151 }, { "epoch": 4.6, "learning_rate": 1.6214299332697807e-06, "loss": 1.6582, "step": 24152 }, { "epoch": 4.6, "learning_rate": 1.6213727359389895e-06, "loss": 1.7267, "step": 24153 }, { "epoch": 4.6, "learning_rate": 1.6213155386081985e-06, "loss": 2.4072, "step": 24154 }, { "epoch": 4.6, "learning_rate": 1.6212583412774071e-06, "loss": 1.0419, "step": 24155 }, { "epoch": 4.6, "learning_rate": 1.6212011439466157e-06, "loss": 2.272, "step": 24156 }, { "epoch": 4.6, "learning_rate": 1.6211439466158247e-06, "loss": 2.6641, "step": 24157 }, { "epoch": 4.6, "learning_rate": 1.6210867492850335e-06, "loss": 2.7162, "step": 24158 }, { "epoch": 4.6, "learning_rate": 1.621029551954242e-06, "loss": 2.4038, "step": 24159 }, { "epoch": 4.6, "learning_rate": 1.620972354623451e-06, "loss": 2.4181, "step": 24160 }, { "epoch": 4.6, "learning_rate": 1.6209151572926597e-06, "loss": 2.3635, "step": 24161 }, { "epoch": 4.6, "learning_rate": 1.6208579599618685e-06, "loss": 2.1833, "step": 24162 }, { "epoch": 4.6, "learning_rate": 1.6208007626310773e-06, "loss": 2.0808, "step": 24163 }, { "epoch": 4.6, "learning_rate": 1.620743565300286e-06, "loss": 3.3461, "step": 24164 }, { "epoch": 4.6, "learning_rate": 1.6206863679694946e-06, "loss": 1.3885, "step": 24165 }, { "epoch": 4.6, "learning_rate": 1.6206291706387036e-06, "loss": 1.4064, "step": 24166 }, { "epoch": 4.6, "learning_rate": 1.6205719733079122e-06, "loss": 1.8571, "step": 24167 }, { "epoch": 4.6, "learning_rate": 1.620514775977121e-06, "loss": 1.8916, "step": 24168 }, { "epoch": 4.6, "learning_rate": 1.62045757864633e-06, "loss": 2.2037, "step": 24169 }, { "epoch": 4.6, "learning_rate": 1.6204003813155386e-06, "loss": 0.9091, "step": 24170 }, { "epoch": 4.6, "learning_rate": 1.6203431839847476e-06, "loss": 2.1203, "step": 24171 }, { "epoch": 4.6, "learning_rate": 1.6202859866539562e-06, "loss": 1.4535, "step": 24172 }, { "epoch": 4.6, "learning_rate": 1.6202287893231648e-06, "loss": 2.9189, "step": 24173 }, { "epoch": 4.6, "learning_rate": 1.6201715919923738e-06, "loss": 0.9244, "step": 24174 }, { "epoch": 4.6, "learning_rate": 1.6201143946615826e-06, "loss": 1.734, "step": 24175 }, { "epoch": 4.6, "learning_rate": 1.6200571973307912e-06, "loss": 0.5079, "step": 24176 }, { "epoch": 4.61, "learning_rate": 1.6200000000000002e-06, "loss": 2.9936, "step": 24177 }, { "epoch": 4.61, "learning_rate": 1.6199428026692087e-06, "loss": 1.5817, "step": 24178 }, { "epoch": 4.61, "learning_rate": 1.6198856053384175e-06, "loss": 2.388, "step": 24179 }, { "epoch": 4.61, "learning_rate": 1.6198284080076265e-06, "loss": 2.6152, "step": 24180 }, { "epoch": 4.61, "learning_rate": 1.6197712106768351e-06, "loss": 2.7801, "step": 24181 }, { "epoch": 4.61, "learning_rate": 1.6197140133460437e-06, "loss": 1.8618, "step": 24182 }, { "epoch": 4.61, "learning_rate": 1.6196568160152527e-06, "loss": 2.5621, "step": 24183 }, { "epoch": 4.61, "learning_rate": 1.6195996186844613e-06, "loss": 1.5893, "step": 24184 }, { "epoch": 4.61, "learning_rate": 1.6195424213536703e-06, "loss": 1.3011, "step": 24185 }, { "epoch": 4.61, "learning_rate": 1.619485224022879e-06, "loss": 2.4214, "step": 24186 }, { "epoch": 4.61, "learning_rate": 1.6194280266920877e-06, "loss": 1.3039, "step": 24187 }, { "epoch": 4.61, "learning_rate": 1.6193708293612967e-06, "loss": 1.7327, "step": 24188 }, { "epoch": 4.61, "learning_rate": 1.6193136320305053e-06, "loss": 2.6582, "step": 24189 }, { "epoch": 4.61, "learning_rate": 1.619256434699714e-06, "loss": 2.1985, "step": 24190 }, { "epoch": 4.61, "learning_rate": 1.6191992373689228e-06, "loss": 2.4214, "step": 24191 }, { "epoch": 4.61, "learning_rate": 1.6191420400381316e-06, "loss": 2.0927, "step": 24192 }, { "epoch": 4.61, "learning_rate": 1.6190848427073402e-06, "loss": 1.1159, "step": 24193 }, { "epoch": 4.61, "learning_rate": 1.6190276453765492e-06, "loss": 3.199, "step": 24194 }, { "epoch": 4.61, "learning_rate": 1.6189704480457578e-06, "loss": 3.5646, "step": 24195 }, { "epoch": 4.61, "learning_rate": 1.6189132507149666e-06, "loss": 2.3618, "step": 24196 }, { "epoch": 4.61, "learning_rate": 1.6188560533841756e-06, "loss": 2.2153, "step": 24197 }, { "epoch": 4.61, "learning_rate": 1.6187988560533842e-06, "loss": 2.8131, "step": 24198 }, { "epoch": 4.61, "learning_rate": 1.6187416587225928e-06, "loss": 2.3904, "step": 24199 }, { "epoch": 4.61, "learning_rate": 1.6186844613918018e-06, "loss": 2.1454, "step": 24200 }, { "epoch": 4.61, "learning_rate": 1.6186272640610104e-06, "loss": 1.3378, "step": 24201 }, { "epoch": 4.61, "learning_rate": 1.6185700667302194e-06, "loss": 1.4695, "step": 24202 }, { "epoch": 4.61, "learning_rate": 1.6185128693994282e-06, "loss": 2.3256, "step": 24203 }, { "epoch": 4.61, "learning_rate": 1.6184556720686367e-06, "loss": 3.1198, "step": 24204 }, { "epoch": 4.61, "learning_rate": 1.6183984747378458e-06, "loss": 3.6235, "step": 24205 }, { "epoch": 4.61, "learning_rate": 1.6183412774070543e-06, "loss": 1.2818, "step": 24206 }, { "epoch": 4.61, "learning_rate": 1.6182840800762631e-06, "loss": 1.5485, "step": 24207 }, { "epoch": 4.61, "learning_rate": 1.6182268827454721e-06, "loss": 2.218, "step": 24208 }, { "epoch": 4.61, "learning_rate": 1.6181696854146807e-06, "loss": 1.4898, "step": 24209 }, { "epoch": 4.61, "learning_rate": 1.6181124880838893e-06, "loss": 2.4429, "step": 24210 }, { "epoch": 4.61, "learning_rate": 1.6180552907530983e-06, "loss": 1.8392, "step": 24211 }, { "epoch": 4.61, "learning_rate": 1.6179980934223069e-06, "loss": 2.4619, "step": 24212 }, { "epoch": 4.61, "learning_rate": 1.6179408960915157e-06, "loss": 1.8978, "step": 24213 }, { "epoch": 4.61, "learning_rate": 1.6178836987607247e-06, "loss": 2.1693, "step": 24214 }, { "epoch": 4.61, "learning_rate": 1.6178265014299333e-06, "loss": 2.0981, "step": 24215 }, { "epoch": 4.61, "learning_rate": 1.6177693040991418e-06, "loss": 1.2247, "step": 24216 }, { "epoch": 4.61, "learning_rate": 1.6177121067683509e-06, "loss": 0.5768, "step": 24217 }, { "epoch": 4.61, "learning_rate": 1.6176549094375596e-06, "loss": 1.8299, "step": 24218 }, { "epoch": 4.61, "learning_rate": 1.6175977121067684e-06, "loss": 2.236, "step": 24219 }, { "epoch": 4.61, "learning_rate": 1.6175405147759772e-06, "loss": 1.2018, "step": 24220 }, { "epoch": 4.61, "learning_rate": 1.6174833174451858e-06, "loss": 2.3909, "step": 24221 }, { "epoch": 4.61, "learning_rate": 1.6174261201143948e-06, "loss": 1.9671, "step": 24222 }, { "epoch": 4.61, "learning_rate": 1.6173689227836034e-06, "loss": 1.8226, "step": 24223 }, { "epoch": 4.61, "learning_rate": 1.6173117254528122e-06, "loss": 2.4284, "step": 24224 }, { "epoch": 4.61, "learning_rate": 1.6172545281220212e-06, "loss": 2.3086, "step": 24225 }, { "epoch": 4.61, "learning_rate": 1.6171973307912298e-06, "loss": 2.4781, "step": 24226 }, { "epoch": 4.61, "learning_rate": 1.6171401334604384e-06, "loss": 2.4564, "step": 24227 }, { "epoch": 4.61, "learning_rate": 1.6170829361296474e-06, "loss": 1.2583, "step": 24228 }, { "epoch": 4.62, "learning_rate": 1.6170257387988562e-06, "loss": 2.7365, "step": 24229 }, { "epoch": 4.62, "learning_rate": 1.6169685414680647e-06, "loss": 1.4914, "step": 24230 }, { "epoch": 4.62, "learning_rate": 1.6169113441372738e-06, "loss": 1.9198, "step": 24231 }, { "epoch": 4.62, "learning_rate": 1.6168541468064823e-06, "loss": 1.5168, "step": 24232 }, { "epoch": 4.62, "learning_rate": 1.616796949475691e-06, "loss": 1.6117, "step": 24233 }, { "epoch": 4.62, "learning_rate": 1.6167397521449e-06, "loss": 2.6478, "step": 24234 }, { "epoch": 4.62, "learning_rate": 1.6166825548141087e-06, "loss": 0.8055, "step": 24235 }, { "epoch": 4.62, "learning_rate": 1.6166253574833177e-06, "loss": 1.4608, "step": 24236 }, { "epoch": 4.62, "learning_rate": 1.6165681601525263e-06, "loss": 1.1218, "step": 24237 }, { "epoch": 4.62, "learning_rate": 1.6165109628217349e-06, "loss": 2.5338, "step": 24238 }, { "epoch": 4.62, "learning_rate": 1.6164537654909439e-06, "loss": 3.178, "step": 24239 }, { "epoch": 4.62, "learning_rate": 1.6163965681601525e-06, "loss": 2.3874, "step": 24240 }, { "epoch": 4.62, "learning_rate": 1.6163393708293613e-06, "loss": 3.6802, "step": 24241 }, { "epoch": 4.62, "learning_rate": 1.6162821734985703e-06, "loss": 2.0423, "step": 24242 }, { "epoch": 4.62, "learning_rate": 1.6162249761677789e-06, "loss": 1.7264, "step": 24243 }, { "epoch": 4.62, "learning_rate": 1.6161677788369874e-06, "loss": 2.2639, "step": 24244 }, { "epoch": 4.62, "learning_rate": 1.6161105815061964e-06, "loss": 2.2694, "step": 24245 }, { "epoch": 4.62, "learning_rate": 1.6160533841754052e-06, "loss": 1.8841, "step": 24246 }, { "epoch": 4.62, "learning_rate": 1.6159961868446138e-06, "loss": 2.3363, "step": 24247 }, { "epoch": 4.62, "learning_rate": 1.6159389895138228e-06, "loss": 1.8272, "step": 24248 }, { "epoch": 4.62, "learning_rate": 1.6158817921830314e-06, "loss": 2.0095, "step": 24249 }, { "epoch": 4.62, "learning_rate": 1.6158245948522404e-06, "loss": 2.6396, "step": 24250 }, { "epoch": 4.62, "learning_rate": 1.615767397521449e-06, "loss": 1.2091, "step": 24251 }, { "epoch": 4.62, "learning_rate": 1.6157102001906578e-06, "loss": 1.1118, "step": 24252 }, { "epoch": 4.62, "learning_rate": 1.6156530028598668e-06, "loss": 2.7527, "step": 24253 }, { "epoch": 4.62, "learning_rate": 1.6155958055290754e-06, "loss": 1.2745, "step": 24254 }, { "epoch": 4.62, "learning_rate": 1.615538608198284e-06, "loss": 1.2529, "step": 24255 }, { "epoch": 4.62, "learning_rate": 1.615481410867493e-06, "loss": 2.6607, "step": 24256 }, { "epoch": 4.62, "learning_rate": 1.6154242135367018e-06, "loss": 1.2234, "step": 24257 }, { "epoch": 4.62, "learning_rate": 1.6153670162059103e-06, "loss": 2.0004, "step": 24258 }, { "epoch": 4.62, "learning_rate": 1.6153098188751193e-06, "loss": 1.7199, "step": 24259 }, { "epoch": 4.62, "learning_rate": 1.615252621544328e-06, "loss": 2.8732, "step": 24260 }, { "epoch": 4.62, "learning_rate": 1.6151954242135365e-06, "loss": 1.0783, "step": 24261 }, { "epoch": 4.62, "learning_rate": 1.6151382268827455e-06, "loss": 2.4666, "step": 24262 }, { "epoch": 4.62, "learning_rate": 1.6150810295519543e-06, "loss": 1.009, "step": 24263 }, { "epoch": 4.62, "learning_rate": 1.6150238322211629e-06, "loss": 1.478, "step": 24264 }, { "epoch": 4.62, "learning_rate": 1.614966634890372e-06, "loss": 1.5958, "step": 24265 }, { "epoch": 4.62, "learning_rate": 1.6149094375595805e-06, "loss": 2.0925, "step": 24266 }, { "epoch": 4.62, "learning_rate": 1.6148522402287895e-06, "loss": 0.9177, "step": 24267 }, { "epoch": 4.62, "learning_rate": 1.6147950428979983e-06, "loss": 1.9615, "step": 24268 }, { "epoch": 4.62, "learning_rate": 1.6147378455672069e-06, "loss": 1.691, "step": 24269 }, { "epoch": 4.62, "learning_rate": 1.6146806482364159e-06, "loss": 1.7322, "step": 24270 }, { "epoch": 4.62, "learning_rate": 1.6146234509056244e-06, "loss": 1.0244, "step": 24271 }, { "epoch": 4.62, "learning_rate": 1.614566253574833e-06, "loss": 2.3414, "step": 24272 }, { "epoch": 4.62, "learning_rate": 1.614509056244042e-06, "loss": 1.085, "step": 24273 }, { "epoch": 4.62, "learning_rate": 1.6144518589132508e-06, "loss": 2.4546, "step": 24274 }, { "epoch": 4.62, "learning_rate": 1.6143946615824594e-06, "loss": 1.5109, "step": 24275 }, { "epoch": 4.62, "learning_rate": 1.6143374642516684e-06, "loss": 1.6501, "step": 24276 }, { "epoch": 4.62, "learning_rate": 1.614280266920877e-06, "loss": 3.1117, "step": 24277 }, { "epoch": 4.62, "learning_rate": 1.6142230695900858e-06, "loss": 1.8557, "step": 24278 }, { "epoch": 4.62, "learning_rate": 1.6141658722592946e-06, "loss": 1.7328, "step": 24279 }, { "epoch": 4.62, "learning_rate": 1.6141086749285034e-06, "loss": 2.2394, "step": 24280 }, { "epoch": 4.62, "learning_rate": 1.614051477597712e-06, "loss": 2.0629, "step": 24281 }, { "epoch": 4.63, "learning_rate": 1.613994280266921e-06, "loss": 1.1466, "step": 24282 }, { "epoch": 4.63, "learning_rate": 1.6139370829361296e-06, "loss": 2.2025, "step": 24283 }, { "epoch": 4.63, "learning_rate": 1.6138798856053386e-06, "loss": 1.606, "step": 24284 }, { "epoch": 4.63, "learning_rate": 1.6138226882745474e-06, "loss": 1.1899, "step": 24285 }, { "epoch": 4.63, "learning_rate": 1.613765490943756e-06, "loss": 1.4021, "step": 24286 }, { "epoch": 4.63, "learning_rate": 1.613708293612965e-06, "loss": 1.8649, "step": 24287 }, { "epoch": 4.63, "learning_rate": 1.6136510962821735e-06, "loss": 3.5998, "step": 24288 }, { "epoch": 4.63, "learning_rate": 1.613593898951382e-06, "loss": 2.2199, "step": 24289 }, { "epoch": 4.63, "learning_rate": 1.6135367016205911e-06, "loss": 2.1489, "step": 24290 }, { "epoch": 4.63, "learning_rate": 1.6134795042898e-06, "loss": 3.0025, "step": 24291 }, { "epoch": 4.63, "learning_rate": 1.6134223069590085e-06, "loss": 2.7515, "step": 24292 }, { "epoch": 4.63, "learning_rate": 1.6133651096282175e-06, "loss": 1.3954, "step": 24293 }, { "epoch": 4.63, "learning_rate": 1.613307912297426e-06, "loss": 2.3737, "step": 24294 }, { "epoch": 4.63, "learning_rate": 1.6132507149666349e-06, "loss": 2.9909, "step": 24295 }, { "epoch": 4.63, "learning_rate": 1.6131935176358439e-06, "loss": 2.4583, "step": 24296 }, { "epoch": 4.63, "learning_rate": 1.6131363203050525e-06, "loss": 1.122, "step": 24297 }, { "epoch": 4.63, "learning_rate": 1.613079122974261e-06, "loss": 0.9288, "step": 24298 }, { "epoch": 4.63, "learning_rate": 1.61302192564347e-06, "loss": 2.0659, "step": 24299 }, { "epoch": 4.63, "learning_rate": 1.6129647283126786e-06, "loss": 1.3462, "step": 24300 }, { "epoch": 4.63, "learning_rate": 1.6129075309818876e-06, "loss": 1.3857, "step": 24301 }, { "epoch": 4.63, "learning_rate": 1.6128503336510964e-06, "loss": 2.0011, "step": 24302 }, { "epoch": 4.63, "learning_rate": 1.612793136320305e-06, "loss": 1.2837, "step": 24303 }, { "epoch": 4.63, "learning_rate": 1.612735938989514e-06, "loss": 2.1038, "step": 24304 }, { "epoch": 4.63, "learning_rate": 1.6126787416587226e-06, "loss": 2.3739, "step": 24305 }, { "epoch": 4.63, "learning_rate": 1.6126215443279314e-06, "loss": 1.6684, "step": 24306 }, { "epoch": 4.63, "learning_rate": 1.6125643469971402e-06, "loss": 1.3314, "step": 24307 }, { "epoch": 4.63, "learning_rate": 1.612507149666349e-06, "loss": 2.6196, "step": 24308 }, { "epoch": 4.63, "learning_rate": 1.6124499523355576e-06, "loss": 2.9647, "step": 24309 }, { "epoch": 4.63, "learning_rate": 1.6123927550047666e-06, "loss": 2.4393, "step": 24310 }, { "epoch": 4.63, "learning_rate": 1.6123355576739751e-06, "loss": 1.8901, "step": 24311 }, { "epoch": 4.63, "learning_rate": 1.612278360343184e-06, "loss": 2.4066, "step": 24312 }, { "epoch": 4.63, "learning_rate": 1.612221163012393e-06, "loss": 1.9579, "step": 24313 }, { "epoch": 4.63, "learning_rate": 1.6121639656816015e-06, "loss": 1.8891, "step": 24314 }, { "epoch": 4.63, "learning_rate": 1.6121067683508105e-06, "loss": 3.0987, "step": 24315 }, { "epoch": 4.63, "learning_rate": 1.6120495710200191e-06, "loss": 2.3779, "step": 24316 }, { "epoch": 4.63, "learning_rate": 1.611992373689228e-06, "loss": 3.038, "step": 24317 }, { "epoch": 4.63, "learning_rate": 1.6119351763584367e-06, "loss": 1.5818, "step": 24318 }, { "epoch": 4.63, "learning_rate": 1.6118779790276455e-06, "loss": 1.9636, "step": 24319 }, { "epoch": 4.63, "learning_rate": 1.611820781696854e-06, "loss": 1.5895, "step": 24320 }, { "epoch": 4.63, "learning_rate": 1.611763584366063e-06, "loss": 1.3326, "step": 24321 }, { "epoch": 4.63, "learning_rate": 1.6117063870352717e-06, "loss": 2.0566, "step": 24322 }, { "epoch": 4.63, "learning_rate": 1.6116491897044805e-06, "loss": 2.1144, "step": 24323 }, { "epoch": 4.63, "learning_rate": 1.6115919923736895e-06, "loss": 1.916, "step": 24324 }, { "epoch": 4.63, "learning_rate": 1.611534795042898e-06, "loss": 1.6638, "step": 24325 }, { "epoch": 4.63, "learning_rate": 1.6114775977121066e-06, "loss": 1.0678, "step": 24326 }, { "epoch": 4.63, "learning_rate": 1.6114204003813156e-06, "loss": 3.028, "step": 24327 }, { "epoch": 4.63, "learning_rate": 1.6113632030505242e-06, "loss": 1.5703, "step": 24328 }, { "epoch": 4.63, "learning_rate": 1.611306005719733e-06, "loss": 1.2301, "step": 24329 }, { "epoch": 4.63, "learning_rate": 1.611248808388942e-06, "loss": 3.5151, "step": 24330 }, { "epoch": 4.63, "learning_rate": 1.6111916110581506e-06, "loss": 0.6621, "step": 24331 }, { "epoch": 4.63, "learning_rate": 1.6111344137273596e-06, "loss": 1.8101, "step": 24332 }, { "epoch": 4.63, "learning_rate": 1.6110772163965682e-06, "loss": 2.779, "step": 24333 }, { "epoch": 4.64, "learning_rate": 1.611020019065777e-06, "loss": 2.0152, "step": 24334 }, { "epoch": 4.64, "learning_rate": 1.610962821734986e-06, "loss": 1.4519, "step": 24335 }, { "epoch": 4.64, "learning_rate": 1.6109056244041946e-06, "loss": 2.1848, "step": 24336 }, { "epoch": 4.64, "learning_rate": 1.6108484270734031e-06, "loss": 2.4983, "step": 24337 }, { "epoch": 4.64, "learning_rate": 1.6107912297426122e-06, "loss": 2.419, "step": 24338 }, { "epoch": 4.64, "learning_rate": 1.6107340324118207e-06, "loss": 2.1443, "step": 24339 }, { "epoch": 4.64, "learning_rate": 1.6106768350810295e-06, "loss": 1.5973, "step": 24340 }, { "epoch": 4.64, "learning_rate": 1.6106196377502385e-06, "loss": 2.7104, "step": 24341 }, { "epoch": 4.64, "learning_rate": 1.6105624404194471e-06, "loss": 2.611, "step": 24342 }, { "epoch": 4.64, "learning_rate": 1.6105052430886557e-06, "loss": 2.2366, "step": 24343 }, { "epoch": 4.64, "learning_rate": 1.6104480457578647e-06, "loss": 1.1824, "step": 24344 }, { "epoch": 4.64, "learning_rate": 1.6103908484270735e-06, "loss": 1.7228, "step": 24345 }, { "epoch": 4.64, "learning_rate": 1.610333651096282e-06, "loss": 2.7196, "step": 24346 }, { "epoch": 4.64, "learning_rate": 1.610276453765491e-06, "loss": 1.7004, "step": 24347 }, { "epoch": 4.64, "learning_rate": 1.6102192564346997e-06, "loss": 2.3597, "step": 24348 }, { "epoch": 4.64, "learning_rate": 1.6101620591039087e-06, "loss": 2.0451, "step": 24349 }, { "epoch": 4.64, "learning_rate": 1.6101048617731173e-06, "loss": 1.7683, "step": 24350 }, { "epoch": 4.64, "learning_rate": 1.610047664442326e-06, "loss": 1.1021, "step": 24351 }, { "epoch": 4.64, "learning_rate": 1.609990467111535e-06, "loss": 2.114, "step": 24352 }, { "epoch": 4.64, "learning_rate": 1.6099332697807436e-06, "loss": 1.6726, "step": 24353 }, { "epoch": 4.64, "learning_rate": 1.6098760724499522e-06, "loss": 1.9265, "step": 24354 }, { "epoch": 4.64, "learning_rate": 1.6098188751191612e-06, "loss": 1.6347, "step": 24355 }, { "epoch": 4.64, "learning_rate": 1.60976167778837e-06, "loss": 1.9462, "step": 24356 }, { "epoch": 4.64, "learning_rate": 1.6097044804575786e-06, "loss": 1.9556, "step": 24357 }, { "epoch": 4.64, "learning_rate": 1.6096472831267876e-06, "loss": 2.2115, "step": 24358 }, { "epoch": 4.64, "learning_rate": 1.6095900857959962e-06, "loss": 2.354, "step": 24359 }, { "epoch": 4.64, "learning_rate": 1.6095328884652048e-06, "loss": 1.2794, "step": 24360 }, { "epoch": 4.64, "learning_rate": 1.6094756911344138e-06, "loss": 2.1784, "step": 24361 }, { "epoch": 4.64, "learning_rate": 1.6094184938036226e-06, "loss": 1.4263, "step": 24362 }, { "epoch": 4.64, "learning_rate": 1.6093612964728312e-06, "loss": 2.9701, "step": 24363 }, { "epoch": 4.64, "learning_rate": 1.6093040991420402e-06, "loss": 3.887, "step": 24364 }, { "epoch": 4.64, "learning_rate": 1.6092469018112487e-06, "loss": 2.0984, "step": 24365 }, { "epoch": 4.64, "learning_rate": 1.6091897044804577e-06, "loss": 1.7259, "step": 24366 }, { "epoch": 4.64, "learning_rate": 1.6091325071496663e-06, "loss": 1.6885, "step": 24367 }, { "epoch": 4.64, "learning_rate": 1.6090753098188751e-06, "loss": 2.1359, "step": 24368 }, { "epoch": 4.64, "learning_rate": 1.6090181124880841e-06, "loss": 2.2754, "step": 24369 }, { "epoch": 4.64, "learning_rate": 1.6089609151572927e-06, "loss": 2.0077, "step": 24370 }, { "epoch": 4.64, "learning_rate": 1.6089037178265013e-06, "loss": 1.9494, "step": 24371 }, { "epoch": 4.64, "learning_rate": 1.6088465204957103e-06, "loss": 2.2258, "step": 24372 }, { "epoch": 4.64, "learning_rate": 1.608789323164919e-06, "loss": 1.9405, "step": 24373 }, { "epoch": 4.64, "learning_rate": 1.6087321258341277e-06, "loss": 2.2284, "step": 24374 }, { "epoch": 4.64, "learning_rate": 1.6086749285033367e-06, "loss": 2.6127, "step": 24375 }, { "epoch": 4.64, "learning_rate": 1.6086177311725453e-06, "loss": 2.8492, "step": 24376 }, { "epoch": 4.64, "learning_rate": 1.6085605338417538e-06, "loss": 2.5099, "step": 24377 }, { "epoch": 4.64, "learning_rate": 1.6085033365109628e-06, "loss": 2.5245, "step": 24378 }, { "epoch": 4.64, "learning_rate": 1.6084461391801716e-06, "loss": 2.8801, "step": 24379 }, { "epoch": 4.64, "learning_rate": 1.6083889418493806e-06, "loss": 2.5842, "step": 24380 }, { "epoch": 4.64, "learning_rate": 1.6083317445185892e-06, "loss": 1.4536, "step": 24381 }, { "epoch": 4.64, "learning_rate": 1.6082745471877978e-06, "loss": 2.0049, "step": 24382 }, { "epoch": 4.64, "learning_rate": 1.6082173498570068e-06, "loss": 2.2769, "step": 24383 }, { "epoch": 4.64, "learning_rate": 1.6081601525262156e-06, "loss": 1.8113, "step": 24384 }, { "epoch": 4.64, "learning_rate": 1.6081029551954242e-06, "loss": 3.5261, "step": 24385 }, { "epoch": 4.64, "learning_rate": 1.6080457578646332e-06, "loss": 3.455, "step": 24386 }, { "epoch": 4.65, "learning_rate": 1.6079885605338418e-06, "loss": 2.3206, "step": 24387 }, { "epoch": 4.65, "learning_rate": 1.6079313632030504e-06, "loss": 2.6706, "step": 24388 }, { "epoch": 4.65, "learning_rate": 1.6078741658722594e-06, "loss": 2.7281, "step": 24389 }, { "epoch": 4.65, "learning_rate": 1.6078169685414682e-06, "loss": 2.3359, "step": 24390 }, { "epoch": 4.65, "learning_rate": 1.6077597712106767e-06, "loss": 1.9142, "step": 24391 }, { "epoch": 4.65, "learning_rate": 1.6077025738798857e-06, "loss": 2.7528, "step": 24392 }, { "epoch": 4.65, "learning_rate": 1.6076453765490943e-06, "loss": 2.6175, "step": 24393 }, { "epoch": 4.65, "learning_rate": 1.6075881792183031e-06, "loss": 1.1725, "step": 24394 }, { "epoch": 4.65, "learning_rate": 1.607530981887512e-06, "loss": 2.5277, "step": 24395 }, { "epoch": 4.65, "learning_rate": 1.6074737845567207e-06, "loss": 1.2322, "step": 24396 }, { "epoch": 4.65, "learning_rate": 1.6074165872259297e-06, "loss": 1.2094, "step": 24397 }, { "epoch": 4.65, "learning_rate": 1.6073593898951383e-06, "loss": 2.3656, "step": 24398 }, { "epoch": 4.65, "learning_rate": 1.6073021925643469e-06, "loss": 1.5961, "step": 24399 }, { "epoch": 4.65, "learning_rate": 1.6072449952335559e-06, "loss": 2.8674, "step": 24400 }, { "epoch": 4.65, "learning_rate": 1.6071877979027647e-06, "loss": 2.3336, "step": 24401 }, { "epoch": 4.65, "learning_rate": 1.6071306005719733e-06, "loss": 2.7679, "step": 24402 }, { "epoch": 4.65, "learning_rate": 1.6070734032411823e-06, "loss": 2.0229, "step": 24403 }, { "epoch": 4.65, "learning_rate": 1.6070162059103909e-06, "loss": 2.931, "step": 24404 }, { "epoch": 4.65, "learning_rate": 1.6069590085795996e-06, "loss": 0.737, "step": 24405 }, { "epoch": 4.65, "learning_rate": 1.6069018112488084e-06, "loss": 3.072, "step": 24406 }, { "epoch": 4.65, "learning_rate": 1.6068446139180172e-06, "loss": 1.8468, "step": 24407 }, { "epoch": 4.65, "learning_rate": 1.6067874165872258e-06, "loss": 2.4921, "step": 24408 }, { "epoch": 4.65, "learning_rate": 1.6067302192564348e-06, "loss": 1.6384, "step": 24409 }, { "epoch": 4.65, "learning_rate": 1.6066730219256434e-06, "loss": 2.6913, "step": 24410 }, { "epoch": 4.65, "learning_rate": 1.6066158245948522e-06, "loss": 2.885, "step": 24411 }, { "epoch": 4.65, "learning_rate": 1.6065586272640612e-06, "loss": 3.3766, "step": 24412 }, { "epoch": 4.65, "learning_rate": 1.6065014299332698e-06, "loss": 2.0464, "step": 24413 }, { "epoch": 4.65, "learning_rate": 1.6064442326024788e-06, "loss": 1.1295, "step": 24414 }, { "epoch": 4.65, "learning_rate": 1.6063870352716874e-06, "loss": 1.5982, "step": 24415 }, { "epoch": 4.65, "learning_rate": 1.606329837940896e-06, "loss": 1.8204, "step": 24416 }, { "epoch": 4.65, "learning_rate": 1.606272640610105e-06, "loss": 2.2838, "step": 24417 }, { "epoch": 4.65, "learning_rate": 1.6062154432793138e-06, "loss": 1.3548, "step": 24418 }, { "epoch": 4.65, "learning_rate": 1.6061582459485223e-06, "loss": 2.322, "step": 24419 }, { "epoch": 4.65, "learning_rate": 1.6061010486177313e-06, "loss": 1.611, "step": 24420 }, { "epoch": 4.65, "learning_rate": 1.60604385128694e-06, "loss": 2.8096, "step": 24421 }, { "epoch": 4.65, "learning_rate": 1.6059866539561487e-06, "loss": 2.3095, "step": 24422 }, { "epoch": 4.65, "learning_rate": 1.6059294566253577e-06, "loss": 1.7216, "step": 24423 }, { "epoch": 4.65, "learning_rate": 1.6058722592945663e-06, "loss": 1.7106, "step": 24424 }, { "epoch": 4.65, "learning_rate": 1.6058150619637749e-06, "loss": 0.9032, "step": 24425 }, { "epoch": 4.65, "learning_rate": 1.6057578646329839e-06, "loss": 2.4415, "step": 24426 }, { "epoch": 4.65, "learning_rate": 1.6057006673021925e-06, "loss": 2.5214, "step": 24427 }, { "epoch": 4.65, "learning_rate": 1.6056434699714013e-06, "loss": 3.0187, "step": 24428 }, { "epoch": 4.65, "learning_rate": 1.6055862726406103e-06, "loss": 1.6802, "step": 24429 }, { "epoch": 4.65, "learning_rate": 1.6055290753098189e-06, "loss": 2.1204, "step": 24430 }, { "epoch": 4.65, "learning_rate": 1.6054718779790279e-06, "loss": 3.4187, "step": 24431 }, { "epoch": 4.65, "learning_rate": 1.6054146806482364e-06, "loss": 1.592, "step": 24432 }, { "epoch": 4.65, "learning_rate": 1.6053574833174452e-06, "loss": 3.4287, "step": 24433 }, { "epoch": 4.65, "learning_rate": 1.605300285986654e-06, "loss": 1.9561, "step": 24434 }, { "epoch": 4.65, "learning_rate": 1.6052430886558628e-06, "loss": 1.6114, "step": 24435 }, { "epoch": 4.65, "learning_rate": 1.6051858913250714e-06, "loss": 2.0248, "step": 24436 }, { "epoch": 4.65, "learning_rate": 1.6051286939942804e-06, "loss": 2.0868, "step": 24437 }, { "epoch": 4.65, "learning_rate": 1.605071496663489e-06, "loss": 1.9625, "step": 24438 }, { "epoch": 4.66, "learning_rate": 1.6050142993326978e-06, "loss": 1.953, "step": 24439 }, { "epoch": 4.66, "learning_rate": 1.6049571020019068e-06, "loss": 1.851, "step": 24440 }, { "epoch": 4.66, "learning_rate": 1.6048999046711154e-06, "loss": 0.7398, "step": 24441 }, { "epoch": 4.66, "learning_rate": 1.604842707340324e-06, "loss": 2.9022, "step": 24442 }, { "epoch": 4.66, "learning_rate": 1.604785510009533e-06, "loss": 2.1512, "step": 24443 }, { "epoch": 4.66, "learning_rate": 1.6047283126787415e-06, "loss": 2.1244, "step": 24444 }, { "epoch": 4.66, "learning_rate": 1.6046711153479506e-06, "loss": 2.7212, "step": 24445 }, { "epoch": 4.66, "learning_rate": 1.6046139180171593e-06, "loss": 1.9266, "step": 24446 }, { "epoch": 4.66, "learning_rate": 1.604556720686368e-06, "loss": 2.7397, "step": 24447 }, { "epoch": 4.66, "learning_rate": 1.604499523355577e-06, "loss": 2.6366, "step": 24448 }, { "epoch": 4.66, "learning_rate": 1.6044423260247855e-06, "loss": 2.8261, "step": 24449 }, { "epoch": 4.66, "learning_rate": 1.6043851286939943e-06, "loss": 1.9698, "step": 24450 }, { "epoch": 4.66, "learning_rate": 1.6043279313632033e-06, "loss": 1.8207, "step": 24451 }, { "epoch": 4.66, "learning_rate": 1.604270734032412e-06, "loss": 1.3922, "step": 24452 }, { "epoch": 4.66, "learning_rate": 1.6042135367016205e-06, "loss": 2.6867, "step": 24453 }, { "epoch": 4.66, "learning_rate": 1.6041563393708295e-06, "loss": 1.7385, "step": 24454 }, { "epoch": 4.66, "learning_rate": 1.604099142040038e-06, "loss": 1.4574, "step": 24455 }, { "epoch": 4.66, "learning_rate": 1.6040419447092469e-06, "loss": 2.4062, "step": 24456 }, { "epoch": 4.66, "learning_rate": 1.6039847473784559e-06, "loss": 1.959, "step": 24457 }, { "epoch": 4.66, "learning_rate": 1.6039275500476644e-06, "loss": 2.0018, "step": 24458 }, { "epoch": 4.66, "learning_rate": 1.603870352716873e-06, "loss": 1.6441, "step": 24459 }, { "epoch": 4.66, "learning_rate": 1.603813155386082e-06, "loss": 2.164, "step": 24460 }, { "epoch": 4.66, "learning_rate": 1.6037559580552908e-06, "loss": 1.9668, "step": 24461 }, { "epoch": 4.66, "learning_rate": 1.6036987607244996e-06, "loss": 2.2988, "step": 24462 }, { "epoch": 4.66, "learning_rate": 1.6036415633937084e-06, "loss": 1.6353, "step": 24463 }, { "epoch": 4.66, "learning_rate": 1.603584366062917e-06, "loss": 2.7639, "step": 24464 }, { "epoch": 4.66, "learning_rate": 1.603527168732126e-06, "loss": 0.7393, "step": 24465 }, { "epoch": 4.66, "learning_rate": 1.6034699714013346e-06, "loss": 1.1958, "step": 24466 }, { "epoch": 4.66, "learning_rate": 1.6034127740705434e-06, "loss": 2.6577, "step": 24467 }, { "epoch": 4.66, "learning_rate": 1.6033555767397524e-06, "loss": 2.3688, "step": 24468 }, { "epoch": 4.66, "learning_rate": 1.603298379408961e-06, "loss": 1.8087, "step": 24469 }, { "epoch": 4.66, "learning_rate": 1.6032411820781695e-06, "loss": 2.0119, "step": 24470 }, { "epoch": 4.66, "learning_rate": 1.6031839847473786e-06, "loss": 1.5348, "step": 24471 }, { "epoch": 4.66, "learning_rate": 1.6031267874165873e-06, "loss": 2.5037, "step": 24472 }, { "epoch": 4.66, "learning_rate": 1.603069590085796e-06, "loss": 2.0832, "step": 24473 }, { "epoch": 4.66, "learning_rate": 1.603012392755005e-06, "loss": 2.249, "step": 24474 }, { "epoch": 4.66, "learning_rate": 1.6029551954242135e-06, "loss": 2.5953, "step": 24475 }, { "epoch": 4.66, "learning_rate": 1.602897998093422e-06, "loss": 1.8166, "step": 24476 }, { "epoch": 4.66, "learning_rate": 1.602840800762631e-06, "loss": 2.0497, "step": 24477 }, { "epoch": 4.66, "learning_rate": 1.60278360343184e-06, "loss": 3.0615, "step": 24478 }, { "epoch": 4.66, "learning_rate": 1.602726406101049e-06, "loss": 2.7228, "step": 24479 }, { "epoch": 4.66, "learning_rate": 1.6026692087702575e-06, "loss": 3.6928, "step": 24480 }, { "epoch": 4.66, "learning_rate": 1.602612011439466e-06, "loss": 0.9202, "step": 24481 }, { "epoch": 4.66, "learning_rate": 1.602554814108675e-06, "loss": 1.8938, "step": 24482 }, { "epoch": 4.66, "learning_rate": 1.6024976167778837e-06, "loss": 1.1174, "step": 24483 }, { "epoch": 4.66, "learning_rate": 1.6024404194470925e-06, "loss": 2.4141, "step": 24484 }, { "epoch": 4.66, "learning_rate": 1.6023832221163015e-06, "loss": 1.1269, "step": 24485 }, { "epoch": 4.66, "learning_rate": 1.60232602478551e-06, "loss": 1.3415, "step": 24486 }, { "epoch": 4.66, "learning_rate": 1.6022688274547186e-06, "loss": 1.3979, "step": 24487 }, { "epoch": 4.66, "learning_rate": 1.6022116301239276e-06, "loss": 1.6099, "step": 24488 }, { "epoch": 4.66, "learning_rate": 1.6021544327931364e-06, "loss": 2.774, "step": 24489 }, { "epoch": 4.66, "learning_rate": 1.602097235462345e-06, "loss": 1.4355, "step": 24490 }, { "epoch": 4.66, "learning_rate": 1.602040038131554e-06, "loss": 2.9782, "step": 24491 }, { "epoch": 4.67, "learning_rate": 1.6019828408007626e-06, "loss": 2.0038, "step": 24492 }, { "epoch": 4.67, "learning_rate": 1.6019256434699714e-06, "loss": 1.3646, "step": 24493 }, { "epoch": 4.67, "learning_rate": 1.6018684461391802e-06, "loss": 2.3218, "step": 24494 }, { "epoch": 4.67, "learning_rate": 1.601811248808389e-06, "loss": 2.6162, "step": 24495 }, { "epoch": 4.67, "learning_rate": 1.601754051477598e-06, "loss": 2.3118, "step": 24496 }, { "epoch": 4.67, "learning_rate": 1.6016968541468066e-06, "loss": 1.7555, "step": 24497 }, { "epoch": 4.67, "learning_rate": 1.6016396568160151e-06, "loss": 1.9763, "step": 24498 }, { "epoch": 4.67, "learning_rate": 1.6015824594852241e-06, "loss": 2.3376, "step": 24499 }, { "epoch": 4.67, "learning_rate": 1.601525262154433e-06, "loss": 3.7677, "step": 24500 }, { "epoch": 4.67, "learning_rate": 1.6014680648236415e-06, "loss": 0.9939, "step": 24501 }, { "epoch": 4.67, "learning_rate": 1.6014108674928505e-06, "loss": 1.5707, "step": 24502 }, { "epoch": 4.67, "learning_rate": 1.6013536701620591e-06, "loss": 1.3829, "step": 24503 }, { "epoch": 4.67, "learning_rate": 1.6012964728312677e-06, "loss": 1.9421, "step": 24504 }, { "epoch": 4.67, "learning_rate": 1.6012392755004767e-06, "loss": 2.1605, "step": 24505 }, { "epoch": 4.67, "learning_rate": 1.6011820781696855e-06, "loss": 2.0696, "step": 24506 }, { "epoch": 4.67, "learning_rate": 1.601124880838894e-06, "loss": 1.8244, "step": 24507 }, { "epoch": 4.67, "learning_rate": 1.601067683508103e-06, "loss": 2.2422, "step": 24508 }, { "epoch": 4.67, "learning_rate": 1.6010104861773117e-06, "loss": 2.8545, "step": 24509 }, { "epoch": 4.67, "learning_rate": 1.6009532888465207e-06, "loss": 2.1248, "step": 24510 }, { "epoch": 4.67, "learning_rate": 1.6008960915157295e-06, "loss": 1.3092, "step": 24511 }, { "epoch": 4.67, "learning_rate": 1.600838894184938e-06, "loss": 2.333, "step": 24512 }, { "epoch": 4.67, "learning_rate": 1.600781696854147e-06, "loss": 3.3687, "step": 24513 }, { "epoch": 4.67, "learning_rate": 1.6007244995233556e-06, "loss": 1.6956, "step": 24514 }, { "epoch": 4.67, "learning_rate": 1.6006673021925642e-06, "loss": 2.0092, "step": 24515 }, { "epoch": 4.67, "learning_rate": 1.6006101048617732e-06, "loss": 2.9026, "step": 24516 }, { "epoch": 4.67, "learning_rate": 1.600552907530982e-06, "loss": 1.1505, "step": 24517 }, { "epoch": 4.67, "learning_rate": 1.6004957102001906e-06, "loss": 2.5628, "step": 24518 }, { "epoch": 4.67, "learning_rate": 1.6004385128693996e-06, "loss": 1.5435, "step": 24519 }, { "epoch": 4.67, "learning_rate": 1.6003813155386082e-06, "loss": 2.045, "step": 24520 }, { "epoch": 4.67, "learning_rate": 1.600324118207817e-06, "loss": 1.8019, "step": 24521 }, { "epoch": 4.67, "learning_rate": 1.6002669208770258e-06, "loss": 2.0489, "step": 24522 }, { "epoch": 4.67, "learning_rate": 1.6002097235462346e-06, "loss": 1.0627, "step": 24523 }, { "epoch": 4.67, "learning_rate": 1.6001525262154431e-06, "loss": 1.9476, "step": 24524 }, { "epoch": 4.67, "learning_rate": 1.6000953288846521e-06, "loss": 1.6513, "step": 24525 }, { "epoch": 4.67, "learning_rate": 1.6000381315538607e-06, "loss": 2.1904, "step": 24526 }, { "epoch": 4.67, "learning_rate": 1.5999809342230697e-06, "loss": 1.8738, "step": 24527 }, { "epoch": 4.67, "learning_rate": 1.5999237368922785e-06, "loss": 1.9842, "step": 24528 }, { "epoch": 4.67, "learning_rate": 1.5998665395614871e-06, "loss": 3.4496, "step": 24529 }, { "epoch": 4.67, "learning_rate": 1.5998093422306961e-06, "loss": 2.91, "step": 24530 }, { "epoch": 4.67, "learning_rate": 1.5997521448999047e-06, "loss": 2.595, "step": 24531 }, { "epoch": 4.67, "learning_rate": 1.5996949475691133e-06, "loss": 1.8761, "step": 24532 }, { "epoch": 4.67, "learning_rate": 1.5996377502383223e-06, "loss": 1.2266, "step": 24533 }, { "epoch": 4.67, "learning_rate": 1.599580552907531e-06, "loss": 2.0238, "step": 24534 }, { "epoch": 4.67, "learning_rate": 1.5995233555767397e-06, "loss": 1.0409, "step": 24535 }, { "epoch": 4.67, "learning_rate": 1.5994661582459487e-06, "loss": 1.6571, "step": 24536 }, { "epoch": 4.67, "learning_rate": 1.5994089609151573e-06, "loss": 2.9295, "step": 24537 }, { "epoch": 4.67, "learning_rate": 1.599351763584366e-06, "loss": 3.3427, "step": 24538 }, { "epoch": 4.67, "learning_rate": 1.599294566253575e-06, "loss": 1.5432, "step": 24539 }, { "epoch": 4.67, "learning_rate": 1.5992373689227836e-06, "loss": 3.1861, "step": 24540 }, { "epoch": 4.67, "learning_rate": 1.5991801715919922e-06, "loss": 2.6367, "step": 24541 }, { "epoch": 4.67, "learning_rate": 1.5991229742612012e-06, "loss": 1.4997, "step": 24542 }, { "epoch": 4.67, "learning_rate": 1.5990657769304098e-06, "loss": 1.2337, "step": 24543 }, { "epoch": 4.68, "learning_rate": 1.5990085795996188e-06, "loss": 2.0245, "step": 24544 }, { "epoch": 4.68, "learning_rate": 1.5989513822688276e-06, "loss": 2.7458, "step": 24545 }, { "epoch": 4.68, "learning_rate": 1.5988941849380362e-06, "loss": 1.7033, "step": 24546 }, { "epoch": 4.68, "learning_rate": 1.5988369876072452e-06, "loss": 2.1076, "step": 24547 }, { "epoch": 4.68, "learning_rate": 1.5987797902764538e-06, "loss": 2.1918, "step": 24548 }, { "epoch": 4.68, "learning_rate": 1.5987225929456626e-06, "loss": 2.5867, "step": 24549 }, { "epoch": 4.68, "learning_rate": 1.5986653956148714e-06, "loss": 1.7054, "step": 24550 }, { "epoch": 4.68, "learning_rate": 1.5986081982840802e-06, "loss": 1.9534, "step": 24551 }, { "epoch": 4.68, "learning_rate": 1.5985510009532887e-06, "loss": 1.3969, "step": 24552 }, { "epoch": 4.68, "learning_rate": 1.5984938036224977e-06, "loss": 2.1746, "step": 24553 }, { "epoch": 4.68, "learning_rate": 1.5984366062917063e-06, "loss": 2.3583, "step": 24554 }, { "epoch": 4.68, "learning_rate": 1.5983794089609151e-06, "loss": 3.1752, "step": 24555 }, { "epoch": 4.68, "learning_rate": 1.5983222116301241e-06, "loss": 1.8171, "step": 24556 }, { "epoch": 4.68, "learning_rate": 1.5982650142993327e-06, "loss": 1.7087, "step": 24557 }, { "epoch": 4.68, "learning_rate": 1.5982078169685413e-06, "loss": 2.3895, "step": 24558 }, { "epoch": 4.68, "learning_rate": 1.5981506196377503e-06, "loss": 2.7481, "step": 24559 }, { "epoch": 4.68, "learning_rate": 1.598093422306959e-06, "loss": 1.9424, "step": 24560 }, { "epoch": 4.68, "learning_rate": 1.5980362249761679e-06, "loss": 1.3959, "step": 24561 }, { "epoch": 4.68, "learning_rate": 1.5979790276453767e-06, "loss": 1.6896, "step": 24562 }, { "epoch": 4.68, "learning_rate": 1.5979218303145853e-06, "loss": 2.1718, "step": 24563 }, { "epoch": 4.68, "learning_rate": 1.5978646329837943e-06, "loss": 1.873, "step": 24564 }, { "epoch": 4.68, "learning_rate": 1.5978074356530028e-06, "loss": 3.278, "step": 24565 }, { "epoch": 4.68, "learning_rate": 1.5977502383222116e-06, "loss": 1.9873, "step": 24566 }, { "epoch": 4.68, "learning_rate": 1.5976930409914206e-06, "loss": 1.1791, "step": 24567 }, { "epoch": 4.68, "learning_rate": 1.5976358436606292e-06, "loss": 2.512, "step": 24568 }, { "epoch": 4.68, "learning_rate": 1.5975786463298378e-06, "loss": 2.0566, "step": 24569 }, { "epoch": 4.68, "learning_rate": 1.5975214489990468e-06, "loss": 2.4967, "step": 24570 }, { "epoch": 4.68, "learning_rate": 1.5974642516682554e-06, "loss": 2.1766, "step": 24571 }, { "epoch": 4.68, "learning_rate": 1.5974070543374642e-06, "loss": 1.5085, "step": 24572 }, { "epoch": 4.68, "learning_rate": 1.5973498570066732e-06, "loss": 2.2689, "step": 24573 }, { "epoch": 4.68, "learning_rate": 1.5972926596758818e-06, "loss": 0.9594, "step": 24574 }, { "epoch": 4.68, "learning_rate": 1.5972354623450908e-06, "loss": 2.0608, "step": 24575 }, { "epoch": 4.68, "learning_rate": 1.5971782650142994e-06, "loss": 2.1949, "step": 24576 }, { "epoch": 4.68, "learning_rate": 1.5971210676835082e-06, "loss": 0.4804, "step": 24577 }, { "epoch": 4.68, "learning_rate": 1.5970638703527172e-06, "loss": 3.2713, "step": 24578 }, { "epoch": 4.68, "learning_rate": 1.5970066730219257e-06, "loss": 2.1808, "step": 24579 }, { "epoch": 4.68, "learning_rate": 1.5969494756911343e-06, "loss": 2.0366, "step": 24580 }, { "epoch": 4.68, "learning_rate": 1.5968922783603433e-06, "loss": 3.6821, "step": 24581 }, { "epoch": 4.68, "learning_rate": 1.596835081029552e-06, "loss": 1.5161, "step": 24582 }, { "epoch": 4.68, "learning_rate": 1.5967778836987607e-06, "loss": 1.5654, "step": 24583 }, { "epoch": 4.68, "learning_rate": 1.5967206863679697e-06, "loss": 3.7871, "step": 24584 }, { "epoch": 4.68, "learning_rate": 1.5966634890371783e-06, "loss": 1.463, "step": 24585 }, { "epoch": 4.68, "learning_rate": 1.5966062917063869e-06, "loss": 1.5443, "step": 24586 }, { "epoch": 4.68, "learning_rate": 1.5965490943755959e-06, "loss": 2.1962, "step": 24587 }, { "epoch": 4.68, "learning_rate": 1.5964918970448047e-06, "loss": 1.9696, "step": 24588 }, { "epoch": 4.68, "learning_rate": 1.5964346997140133e-06, "loss": 1.7069, "step": 24589 }, { "epoch": 4.68, "learning_rate": 1.5963775023832223e-06, "loss": 2.071, "step": 24590 }, { "epoch": 4.68, "learning_rate": 1.5963203050524308e-06, "loss": 2.3779, "step": 24591 }, { "epoch": 4.68, "learning_rate": 1.5962631077216399e-06, "loss": 1.3682, "step": 24592 }, { "epoch": 4.68, "learning_rate": 1.5962059103908484e-06, "loss": 2.5883, "step": 24593 }, { "epoch": 4.68, "learning_rate": 1.5961487130600572e-06, "loss": 2.843, "step": 24594 }, { "epoch": 4.68, "learning_rate": 1.5960915157292662e-06, "loss": 2.1701, "step": 24595 }, { "epoch": 4.68, "learning_rate": 1.5960343183984748e-06, "loss": 1.5611, "step": 24596 }, { "epoch": 4.69, "learning_rate": 1.5959771210676834e-06, "loss": 2.5724, "step": 24597 }, { "epoch": 4.69, "learning_rate": 1.5959199237368924e-06, "loss": 1.7409, "step": 24598 }, { "epoch": 4.69, "learning_rate": 1.595862726406101e-06, "loss": 1.7401, "step": 24599 }, { "epoch": 4.69, "learning_rate": 1.5958055290753098e-06, "loss": 2.3823, "step": 24600 }, { "epoch": 4.69, "learning_rate": 1.5957483317445188e-06, "loss": 1.8275, "step": 24601 }, { "epoch": 4.69, "learning_rate": 1.5956911344137274e-06, "loss": 2.5171, "step": 24602 }, { "epoch": 4.69, "learning_rate": 1.595633937082936e-06, "loss": 1.5525, "step": 24603 }, { "epoch": 4.69, "learning_rate": 1.595576739752145e-06, "loss": 2.9608, "step": 24604 }, { "epoch": 4.69, "learning_rate": 1.5955195424213537e-06, "loss": 1.9228, "step": 24605 }, { "epoch": 4.69, "learning_rate": 1.5954623450905623e-06, "loss": 2.3867, "step": 24606 }, { "epoch": 4.69, "learning_rate": 1.5954051477597713e-06, "loss": 2.6098, "step": 24607 }, { "epoch": 4.69, "learning_rate": 1.59534795042898e-06, "loss": 1.5616, "step": 24608 }, { "epoch": 4.69, "learning_rate": 1.595290753098189e-06, "loss": 2.0365, "step": 24609 }, { "epoch": 4.69, "learning_rate": 1.5952335557673975e-06, "loss": 2.0766, "step": 24610 }, { "epoch": 4.69, "learning_rate": 1.5951763584366063e-06, "loss": 3.0342, "step": 24611 }, { "epoch": 4.69, "learning_rate": 1.5951191611058153e-06, "loss": 1.4944, "step": 24612 }, { "epoch": 4.69, "learning_rate": 1.5950619637750239e-06, "loss": 1.3131, "step": 24613 }, { "epoch": 4.69, "learning_rate": 1.5950047664442325e-06, "loss": 1.9329, "step": 24614 }, { "epoch": 4.69, "learning_rate": 1.5949475691134415e-06, "loss": 1.4, "step": 24615 }, { "epoch": 4.69, "learning_rate": 1.5948903717826503e-06, "loss": 1.7268, "step": 24616 }, { "epoch": 4.69, "learning_rate": 1.5948331744518589e-06, "loss": 2.6183, "step": 24617 }, { "epoch": 4.69, "learning_rate": 1.5947759771210679e-06, "loss": 2.0559, "step": 24618 }, { "epoch": 4.69, "learning_rate": 1.5947187797902764e-06, "loss": 2.2964, "step": 24619 }, { "epoch": 4.69, "learning_rate": 1.594661582459485e-06, "loss": 1.95, "step": 24620 }, { "epoch": 4.69, "learning_rate": 1.594604385128694e-06, "loss": 2.2067, "step": 24621 }, { "epoch": 4.69, "learning_rate": 1.5945471877979028e-06, "loss": 2.3091, "step": 24622 }, { "epoch": 4.69, "learning_rate": 1.5944899904671114e-06, "loss": 1.2868, "step": 24623 }, { "epoch": 4.69, "learning_rate": 1.5944327931363204e-06, "loss": 1.0751, "step": 24624 }, { "epoch": 4.69, "learning_rate": 1.594375595805529e-06, "loss": 1.8806, "step": 24625 }, { "epoch": 4.69, "learning_rate": 1.594318398474738e-06, "loss": 2.2058, "step": 24626 }, { "epoch": 4.69, "learning_rate": 1.5942612011439468e-06, "loss": 2.2421, "step": 24627 }, { "epoch": 4.69, "learning_rate": 1.5942040038131554e-06, "loss": 0.7094, "step": 24628 }, { "epoch": 4.69, "learning_rate": 1.5941468064823644e-06, "loss": 1.8204, "step": 24629 }, { "epoch": 4.69, "learning_rate": 1.594089609151573e-06, "loss": 2.0944, "step": 24630 }, { "epoch": 4.69, "learning_rate": 1.5940324118207815e-06, "loss": 1.4761, "step": 24631 }, { "epoch": 4.69, "learning_rate": 1.5939752144899905e-06, "loss": 2.8759, "step": 24632 }, { "epoch": 4.69, "learning_rate": 1.5939180171591993e-06, "loss": 1.2413, "step": 24633 }, { "epoch": 4.69, "learning_rate": 1.593860819828408e-06, "loss": 1.9798, "step": 24634 }, { "epoch": 4.69, "learning_rate": 1.593803622497617e-06, "loss": 1.3849, "step": 24635 }, { "epoch": 4.69, "learning_rate": 1.5937464251668255e-06, "loss": 3.1481, "step": 24636 }, { "epoch": 4.69, "learning_rate": 1.5936892278360343e-06, "loss": 2.2335, "step": 24637 }, { "epoch": 4.69, "learning_rate": 1.593632030505243e-06, "loss": 2.5264, "step": 24638 }, { "epoch": 4.69, "learning_rate": 1.5935748331744519e-06, "loss": 1.6616, "step": 24639 }, { "epoch": 4.69, "learning_rate": 1.5935176358436605e-06, "loss": 1.7857, "step": 24640 }, { "epoch": 4.69, "learning_rate": 1.5934604385128695e-06, "loss": 2.634, "step": 24641 }, { "epoch": 4.69, "learning_rate": 1.593403241182078e-06, "loss": 1.3579, "step": 24642 }, { "epoch": 4.69, "learning_rate": 1.593346043851287e-06, "loss": 2.4194, "step": 24643 }, { "epoch": 4.69, "learning_rate": 1.5932888465204959e-06, "loss": 0.9476, "step": 24644 }, { "epoch": 4.69, "learning_rate": 1.5932316491897044e-06, "loss": 3.1541, "step": 24645 }, { "epoch": 4.69, "learning_rate": 1.5931744518589134e-06, "loss": 1.6729, "step": 24646 }, { "epoch": 4.69, "learning_rate": 1.593117254528122e-06, "loss": 2.8213, "step": 24647 }, { "epoch": 4.69, "learning_rate": 1.5930600571973308e-06, "loss": 3.4016, "step": 24648 }, { "epoch": 4.7, "learning_rate": 1.5930028598665396e-06, "loss": 1.5695, "step": 24649 }, { "epoch": 4.7, "learning_rate": 1.5929456625357484e-06, "loss": 2.0968, "step": 24650 }, { "epoch": 4.7, "learning_rate": 1.592888465204957e-06, "loss": 1.5768, "step": 24651 }, { "epoch": 4.7, "learning_rate": 1.592831267874166e-06, "loss": 1.3284, "step": 24652 }, { "epoch": 4.7, "learning_rate": 1.5927740705433746e-06, "loss": 2.5536, "step": 24653 }, { "epoch": 4.7, "learning_rate": 1.5927168732125834e-06, "loss": 2.2401, "step": 24654 }, { "epoch": 4.7, "learning_rate": 1.5926596758817924e-06, "loss": 1.8686, "step": 24655 }, { "epoch": 4.7, "learning_rate": 1.592602478551001e-06, "loss": 2.7443, "step": 24656 }, { "epoch": 4.7, "learning_rate": 1.59254528122021e-06, "loss": 1.5722, "step": 24657 }, { "epoch": 4.7, "learning_rate": 1.5924880838894186e-06, "loss": 1.7952, "step": 24658 }, { "epoch": 4.7, "learning_rate": 1.5924308865586271e-06, "loss": 1.4209, "step": 24659 }, { "epoch": 4.7, "learning_rate": 1.5923736892278361e-06, "loss": 3.0006, "step": 24660 }, { "epoch": 4.7, "learning_rate": 1.592316491897045e-06, "loss": 3.4455, "step": 24661 }, { "epoch": 4.7, "learning_rate": 1.5922592945662535e-06, "loss": 1.9918, "step": 24662 }, { "epoch": 4.7, "learning_rate": 1.5922020972354625e-06, "loss": 3.2088, "step": 24663 }, { "epoch": 4.7, "learning_rate": 1.592144899904671e-06, "loss": 1.554, "step": 24664 }, { "epoch": 4.7, "learning_rate": 1.59208770257388e-06, "loss": 2.713, "step": 24665 }, { "epoch": 4.7, "learning_rate": 1.592030505243089e-06, "loss": 1.6159, "step": 24666 }, { "epoch": 4.7, "learning_rate": 1.5919733079122975e-06, "loss": 2.0776, "step": 24667 }, { "epoch": 4.7, "learning_rate": 1.591916110581506e-06, "loss": 2.2355, "step": 24668 }, { "epoch": 4.7, "learning_rate": 1.591858913250715e-06, "loss": 1.5948, "step": 24669 }, { "epoch": 4.7, "learning_rate": 1.5918017159199237e-06, "loss": 0.8775, "step": 24670 }, { "epoch": 4.7, "learning_rate": 1.5917445185891324e-06, "loss": 0.8092, "step": 24671 }, { "epoch": 4.7, "learning_rate": 1.5916873212583415e-06, "loss": 1.2406, "step": 24672 }, { "epoch": 4.7, "learning_rate": 1.59163012392755e-06, "loss": 1.7666, "step": 24673 }, { "epoch": 4.7, "learning_rate": 1.591572926596759e-06, "loss": 1.6227, "step": 24674 }, { "epoch": 4.7, "learning_rate": 1.5915157292659676e-06, "loss": 2.5186, "step": 24675 }, { "epoch": 4.7, "learning_rate": 1.5914585319351764e-06, "loss": 2.3824, "step": 24676 }, { "epoch": 4.7, "learning_rate": 1.5914013346043852e-06, "loss": 2.3741, "step": 24677 }, { "epoch": 4.7, "learning_rate": 1.591344137273594e-06, "loss": 2.0147, "step": 24678 }, { "epoch": 4.7, "learning_rate": 1.5912869399428026e-06, "loss": 2.4484, "step": 24679 }, { "epoch": 4.7, "learning_rate": 1.5912297426120116e-06, "loss": 2.9128, "step": 24680 }, { "epoch": 4.7, "learning_rate": 1.5911725452812202e-06, "loss": 1.4181, "step": 24681 }, { "epoch": 4.7, "learning_rate": 1.591115347950429e-06, "loss": 2.6506, "step": 24682 }, { "epoch": 4.7, "learning_rate": 1.591058150619638e-06, "loss": 2.9003, "step": 24683 }, { "epoch": 4.7, "learning_rate": 1.5910009532888466e-06, "loss": 2.8159, "step": 24684 }, { "epoch": 4.7, "learning_rate": 1.5909437559580551e-06, "loss": 2.1257, "step": 24685 }, { "epoch": 4.7, "learning_rate": 1.5908865586272641e-06, "loss": 1.7391, "step": 24686 }, { "epoch": 4.7, "learning_rate": 1.5908293612964727e-06, "loss": 2.1143, "step": 24687 }, { "epoch": 4.7, "learning_rate": 1.5907721639656815e-06, "loss": 1.9614, "step": 24688 }, { "epoch": 4.7, "learning_rate": 1.5907149666348905e-06, "loss": 2.9093, "step": 24689 }, { "epoch": 4.7, "learning_rate": 1.5906577693040991e-06, "loss": 1.7099, "step": 24690 }, { "epoch": 4.7, "learning_rate": 1.5906005719733081e-06, "loss": 1.8277, "step": 24691 }, { "epoch": 4.7, "learning_rate": 1.5905433746425167e-06, "loss": 2.0954, "step": 24692 }, { "epoch": 4.7, "learning_rate": 1.5904861773117255e-06, "loss": 2.2211, "step": 24693 }, { "epoch": 4.7, "learning_rate": 1.5904289799809345e-06, "loss": 1.3568, "step": 24694 }, { "epoch": 4.7, "learning_rate": 1.590371782650143e-06, "loss": 1.3501, "step": 24695 }, { "epoch": 4.7, "learning_rate": 1.5903145853193517e-06, "loss": 1.6199, "step": 24696 }, { "epoch": 4.7, "learning_rate": 1.5902573879885607e-06, "loss": 2.2373, "step": 24697 }, { "epoch": 4.7, "learning_rate": 1.5902001906577692e-06, "loss": 1.5165, "step": 24698 }, { "epoch": 4.7, "learning_rate": 1.590142993326978e-06, "loss": 1.9059, "step": 24699 }, { "epoch": 4.7, "learning_rate": 1.590085795996187e-06, "loss": 1.6904, "step": 24700 }, { "epoch": 4.7, "learning_rate": 1.5900285986653956e-06, "loss": 1.6259, "step": 24701 }, { "epoch": 4.71, "learning_rate": 1.5899714013346042e-06, "loss": 1.7829, "step": 24702 }, { "epoch": 4.71, "learning_rate": 1.5899142040038132e-06, "loss": 2.3249, "step": 24703 }, { "epoch": 4.71, "learning_rate": 1.589857006673022e-06, "loss": 3.0181, "step": 24704 }, { "epoch": 4.71, "learning_rate": 1.5897998093422306e-06, "loss": 2.1418, "step": 24705 }, { "epoch": 4.71, "learning_rate": 1.5897426120114396e-06, "loss": 1.937, "step": 24706 }, { "epoch": 4.71, "learning_rate": 1.5896854146806482e-06, "loss": 1.6666, "step": 24707 }, { "epoch": 4.71, "learning_rate": 1.5896282173498572e-06, "loss": 3.0367, "step": 24708 }, { "epoch": 4.71, "learning_rate": 1.5895710200190658e-06, "loss": 2.1548, "step": 24709 }, { "epoch": 4.71, "learning_rate": 1.5895138226882746e-06, "loss": 1.5347, "step": 24710 }, { "epoch": 4.71, "learning_rate": 1.5894566253574836e-06, "loss": 1.4784, "step": 24711 }, { "epoch": 4.71, "learning_rate": 1.5893994280266921e-06, "loss": 1.4596, "step": 24712 }, { "epoch": 4.71, "learning_rate": 1.5893422306959007e-06, "loss": 1.7729, "step": 24713 }, { "epoch": 4.71, "learning_rate": 1.5892850333651097e-06, "loss": 1.2973, "step": 24714 }, { "epoch": 4.71, "learning_rate": 1.5892278360343185e-06, "loss": 1.1629, "step": 24715 }, { "epoch": 4.71, "learning_rate": 1.5891706387035271e-06, "loss": 2.1361, "step": 24716 }, { "epoch": 4.71, "learning_rate": 1.5891134413727361e-06, "loss": 1.2405, "step": 24717 }, { "epoch": 4.71, "learning_rate": 1.5890562440419447e-06, "loss": 2.2591, "step": 24718 }, { "epoch": 4.71, "learning_rate": 1.5889990467111533e-06, "loss": 0.9299, "step": 24719 }, { "epoch": 4.71, "learning_rate": 1.5889418493803623e-06, "loss": 1.3907, "step": 24720 }, { "epoch": 4.71, "learning_rate": 1.588884652049571e-06, "loss": 3.1276, "step": 24721 }, { "epoch": 4.71, "learning_rate": 1.58882745471878e-06, "loss": 1.9943, "step": 24722 }, { "epoch": 4.71, "learning_rate": 1.5887702573879887e-06, "loss": 1.7472, "step": 24723 }, { "epoch": 4.71, "learning_rate": 1.5887130600571972e-06, "loss": 1.6169, "step": 24724 }, { "epoch": 4.71, "learning_rate": 1.5886558627264063e-06, "loss": 3.0197, "step": 24725 }, { "epoch": 4.71, "learning_rate": 1.5885986653956148e-06, "loss": 2.0968, "step": 24726 }, { "epoch": 4.71, "learning_rate": 1.5885414680648236e-06, "loss": 2.5221, "step": 24727 }, { "epoch": 4.71, "learning_rate": 1.5884842707340326e-06, "loss": 1.3328, "step": 24728 }, { "epoch": 4.71, "learning_rate": 1.5884270734032412e-06, "loss": 2.1464, "step": 24729 }, { "epoch": 4.71, "learning_rate": 1.5883698760724498e-06, "loss": 1.4698, "step": 24730 }, { "epoch": 4.71, "learning_rate": 1.5883126787416588e-06, "loss": 2.2654, "step": 24731 }, { "epoch": 4.71, "learning_rate": 1.5882554814108676e-06, "loss": 1.9179, "step": 24732 }, { "epoch": 4.71, "learning_rate": 1.5881982840800762e-06, "loss": 2.6663, "step": 24733 }, { "epoch": 4.71, "learning_rate": 1.5881410867492852e-06, "loss": 1.3794, "step": 24734 }, { "epoch": 4.71, "learning_rate": 1.5880838894184938e-06, "loss": 1.7264, "step": 24735 }, { "epoch": 4.71, "learning_rate": 1.5880266920877026e-06, "loss": 2.3025, "step": 24736 }, { "epoch": 4.71, "learning_rate": 1.5879694947569114e-06, "loss": 1.6699, "step": 24737 }, { "epoch": 4.71, "learning_rate": 1.5879122974261202e-06, "loss": 2.6479, "step": 24738 }, { "epoch": 4.71, "learning_rate": 1.5878551000953292e-06, "loss": 1.7277, "step": 24739 }, { "epoch": 4.71, "learning_rate": 1.5877979027645377e-06, "loss": 1.2699, "step": 24740 }, { "epoch": 4.71, "learning_rate": 1.5877407054337463e-06, "loss": 1.4107, "step": 24741 }, { "epoch": 4.71, "learning_rate": 1.5876835081029553e-06, "loss": 2.2374, "step": 24742 }, { "epoch": 4.71, "learning_rate": 1.5876263107721641e-06, "loss": 1.3218, "step": 24743 }, { "epoch": 4.71, "learning_rate": 1.5875691134413727e-06, "loss": 2.8188, "step": 24744 }, { "epoch": 4.71, "learning_rate": 1.5875119161105817e-06, "loss": 2.4003, "step": 24745 }, { "epoch": 4.71, "learning_rate": 1.5874547187797903e-06, "loss": 1.3194, "step": 24746 }, { "epoch": 4.71, "learning_rate": 1.5873975214489989e-06, "loss": 3.1357, "step": 24747 }, { "epoch": 4.71, "learning_rate": 1.5873403241182079e-06, "loss": 1.4023, "step": 24748 }, { "epoch": 4.71, "learning_rate": 1.5872831267874167e-06, "loss": 1.6558, "step": 24749 }, { "epoch": 4.71, "learning_rate": 1.5872259294566253e-06, "loss": 1.9143, "step": 24750 }, { "epoch": 4.71, "learning_rate": 1.5871687321258343e-06, "loss": 2.2319, "step": 24751 }, { "epoch": 4.71, "learning_rate": 1.5871115347950428e-06, "loss": 2.1919, "step": 24752 }, { "epoch": 4.71, "learning_rate": 1.5870543374642516e-06, "loss": 1.7565, "step": 24753 }, { "epoch": 4.72, "learning_rate": 1.5869971401334604e-06, "loss": 2.8356, "step": 24754 }, { "epoch": 4.72, "learning_rate": 1.5869399428026692e-06, "loss": 1.7764, "step": 24755 }, { "epoch": 4.72, "learning_rate": 1.5868827454718782e-06, "loss": 1.5881, "step": 24756 }, { "epoch": 4.72, "learning_rate": 1.5868255481410868e-06, "loss": 1.6767, "step": 24757 }, { "epoch": 4.72, "learning_rate": 1.5867683508102954e-06, "loss": 1.3934, "step": 24758 }, { "epoch": 4.72, "learning_rate": 1.5867111534795044e-06, "loss": 2.3381, "step": 24759 }, { "epoch": 4.72, "learning_rate": 1.5866539561487132e-06, "loss": 1.5973, "step": 24760 }, { "epoch": 4.72, "learning_rate": 1.5865967588179218e-06, "loss": 2.1314, "step": 24761 }, { "epoch": 4.72, "learning_rate": 1.5865395614871308e-06, "loss": 2.659, "step": 24762 }, { "epoch": 4.72, "learning_rate": 1.5864823641563394e-06, "loss": 1.7825, "step": 24763 }, { "epoch": 4.72, "learning_rate": 1.5864251668255482e-06, "loss": 1.3844, "step": 24764 }, { "epoch": 4.72, "learning_rate": 1.586367969494757e-06, "loss": 2.2795, "step": 24765 }, { "epoch": 4.72, "learning_rate": 1.5863107721639657e-06, "loss": 1.3131, "step": 24766 }, { "epoch": 4.72, "learning_rate": 1.5862535748331743e-06, "loss": 1.4251, "step": 24767 }, { "epoch": 4.72, "learning_rate": 1.5861963775023833e-06, "loss": 2.4808, "step": 24768 }, { "epoch": 4.72, "learning_rate": 1.586139180171592e-06, "loss": 2.6109, "step": 24769 }, { "epoch": 4.72, "learning_rate": 1.5860819828408007e-06, "loss": 2.3862, "step": 24770 }, { "epoch": 4.72, "learning_rate": 1.5860247855100097e-06, "loss": 1.7183, "step": 24771 }, { "epoch": 4.72, "learning_rate": 1.5859675881792183e-06, "loss": 1.5255, "step": 24772 }, { "epoch": 4.72, "learning_rate": 1.5859103908484273e-06, "loss": 1.6298, "step": 24773 }, { "epoch": 4.72, "learning_rate": 1.5858531935176359e-06, "loss": 1.4629, "step": 24774 }, { "epoch": 4.72, "learning_rate": 1.5857959961868445e-06, "loss": 2.7718, "step": 24775 }, { "epoch": 4.72, "learning_rate": 1.5857387988560535e-06, "loss": 2.2975, "step": 24776 }, { "epoch": 4.72, "learning_rate": 1.5856816015252623e-06, "loss": 2.3269, "step": 24777 }, { "epoch": 4.72, "learning_rate": 1.5856244041944708e-06, "loss": 2.0107, "step": 24778 }, { "epoch": 4.72, "learning_rate": 1.5855672068636799e-06, "loss": 1.0951, "step": 24779 }, { "epoch": 4.72, "learning_rate": 1.5855100095328884e-06, "loss": 1.5179, "step": 24780 }, { "epoch": 4.72, "learning_rate": 1.5854528122020972e-06, "loss": 2.5227, "step": 24781 }, { "epoch": 4.72, "learning_rate": 1.5853956148713062e-06, "loss": 1.1407, "step": 24782 }, { "epoch": 4.72, "learning_rate": 1.5853384175405148e-06, "loss": 2.4807, "step": 24783 }, { "epoch": 4.72, "learning_rate": 1.5852812202097234e-06, "loss": 1.6725, "step": 24784 }, { "epoch": 4.72, "learning_rate": 1.5852240228789324e-06, "loss": 1.3166, "step": 24785 }, { "epoch": 4.72, "learning_rate": 1.585166825548141e-06, "loss": 1.8891, "step": 24786 }, { "epoch": 4.72, "learning_rate": 1.58510962821735e-06, "loss": 3.7914, "step": 24787 }, { "epoch": 4.72, "learning_rate": 1.5850524308865588e-06, "loss": 2.4272, "step": 24788 }, { "epoch": 4.72, "learning_rate": 1.5849952335557674e-06, "loss": 2.9156, "step": 24789 }, { "epoch": 4.72, "learning_rate": 1.5849380362249764e-06, "loss": 1.8218, "step": 24790 }, { "epoch": 4.72, "learning_rate": 1.584880838894185e-06, "loss": 2.6763, "step": 24791 }, { "epoch": 4.72, "learning_rate": 1.5848236415633937e-06, "loss": 2.8375, "step": 24792 }, { "epoch": 4.72, "learning_rate": 1.5847664442326025e-06, "loss": 2.5294, "step": 24793 }, { "epoch": 4.72, "learning_rate": 1.5847092469018113e-06, "loss": 2.3219, "step": 24794 }, { "epoch": 4.72, "learning_rate": 1.58465204957102e-06, "loss": 1.5134, "step": 24795 }, { "epoch": 4.72, "learning_rate": 1.584594852240229e-06, "loss": 1.3437, "step": 24796 }, { "epoch": 4.72, "learning_rate": 1.5845376549094375e-06, "loss": 1.7908, "step": 24797 }, { "epoch": 4.72, "learning_rate": 1.5844804575786463e-06, "loss": 2.0493, "step": 24798 }, { "epoch": 4.72, "learning_rate": 1.5844232602478553e-06, "loss": 2.6694, "step": 24799 }, { "epoch": 4.72, "learning_rate": 1.5843660629170639e-06, "loss": 2.1665, "step": 24800 }, { "epoch": 4.72, "learning_rate": 1.5843088655862725e-06, "loss": 3.3688, "step": 24801 }, { "epoch": 4.72, "learning_rate": 1.5842516682554815e-06, "loss": 1.831, "step": 24802 }, { "epoch": 4.72, "learning_rate": 1.5841944709246903e-06, "loss": 1.8265, "step": 24803 }, { "epoch": 4.72, "learning_rate": 1.584137273593899e-06, "loss": 2.4275, "step": 24804 }, { "epoch": 4.72, "learning_rate": 1.5840800762631079e-06, "loss": 1.6825, "step": 24805 }, { "epoch": 4.72, "learning_rate": 1.5840228789323164e-06, "loss": 2.6214, "step": 24806 }, { "epoch": 4.73, "learning_rate": 1.5839656816015254e-06, "loss": 0.8859, "step": 24807 }, { "epoch": 4.73, "learning_rate": 1.583908484270734e-06, "loss": 1.9128, "step": 24808 }, { "epoch": 4.73, "learning_rate": 1.5838512869399428e-06, "loss": 2.3602, "step": 24809 }, { "epoch": 4.73, "learning_rate": 1.5837940896091518e-06, "loss": 1.7229, "step": 24810 }, { "epoch": 4.73, "learning_rate": 1.5837368922783604e-06, "loss": 1.7922, "step": 24811 }, { "epoch": 4.73, "learning_rate": 1.583679694947569e-06, "loss": 2.4925, "step": 24812 }, { "epoch": 4.73, "learning_rate": 1.583622497616778e-06, "loss": 3.2445, "step": 24813 }, { "epoch": 4.73, "learning_rate": 1.5835653002859866e-06, "loss": 2.0733, "step": 24814 }, { "epoch": 4.73, "learning_rate": 1.5835081029551954e-06, "loss": 1.9261, "step": 24815 }, { "epoch": 4.73, "learning_rate": 1.5834509056244044e-06, "loss": 1.9376, "step": 24816 }, { "epoch": 4.73, "learning_rate": 1.583393708293613e-06, "loss": 1.6733, "step": 24817 }, { "epoch": 4.73, "learning_rate": 1.5833365109628215e-06, "loss": 2.6232, "step": 24818 }, { "epoch": 4.73, "learning_rate": 1.5832793136320305e-06, "loss": 1.8453, "step": 24819 }, { "epoch": 4.73, "learning_rate": 1.5832221163012393e-06, "loss": 0.9242, "step": 24820 }, { "epoch": 4.73, "learning_rate": 1.5831649189704483e-06, "loss": 4.0132, "step": 24821 }, { "epoch": 4.73, "learning_rate": 1.583107721639657e-06, "loss": 2.2158, "step": 24822 }, { "epoch": 4.73, "learning_rate": 1.5830505243088655e-06, "loss": 1.5053, "step": 24823 }, { "epoch": 4.73, "learning_rate": 1.5829933269780745e-06, "loss": 1.1888, "step": 24824 }, { "epoch": 4.73, "learning_rate": 1.582936129647283e-06, "loss": 2.5057, "step": 24825 }, { "epoch": 4.73, "learning_rate": 1.5828789323164919e-06, "loss": 2.7289, "step": 24826 }, { "epoch": 4.73, "learning_rate": 1.582821734985701e-06, "loss": 2.3011, "step": 24827 }, { "epoch": 4.73, "learning_rate": 1.5827645376549095e-06, "loss": 1.9676, "step": 24828 }, { "epoch": 4.73, "learning_rate": 1.582707340324118e-06, "loss": 1.6849, "step": 24829 }, { "epoch": 4.73, "learning_rate": 1.582650142993327e-06, "loss": 2.0481, "step": 24830 }, { "epoch": 4.73, "learning_rate": 1.5825929456625359e-06, "loss": 2.2111, "step": 24831 }, { "epoch": 4.73, "learning_rate": 1.5825357483317444e-06, "loss": 1.5639, "step": 24832 }, { "epoch": 4.73, "learning_rate": 1.5824785510009534e-06, "loss": 1.7021, "step": 24833 }, { "epoch": 4.73, "learning_rate": 1.582421353670162e-06, "loss": 2.5451, "step": 24834 }, { "epoch": 4.73, "learning_rate": 1.5823641563393706e-06, "loss": 1.873, "step": 24835 }, { "epoch": 4.73, "learning_rate": 1.5823069590085796e-06, "loss": 1.4318, "step": 24836 }, { "epoch": 4.73, "learning_rate": 1.5822497616777884e-06, "loss": 1.6712, "step": 24837 }, { "epoch": 4.73, "learning_rate": 1.5821925643469974e-06, "loss": 1.849, "step": 24838 }, { "epoch": 4.73, "learning_rate": 1.582135367016206e-06, "loss": 1.8608, "step": 24839 }, { "epoch": 4.73, "learning_rate": 1.5820781696854146e-06, "loss": 2.5737, "step": 24840 }, { "epoch": 4.73, "learning_rate": 1.5820209723546236e-06, "loss": 2.5728, "step": 24841 }, { "epoch": 4.73, "learning_rate": 1.5819637750238322e-06, "loss": 1.7107, "step": 24842 }, { "epoch": 4.73, "learning_rate": 1.581906577693041e-06, "loss": 2.3195, "step": 24843 }, { "epoch": 4.73, "learning_rate": 1.58184938036225e-06, "loss": 3.6948, "step": 24844 }, { "epoch": 4.73, "learning_rate": 1.5817921830314585e-06, "loss": 1.6881, "step": 24845 }, { "epoch": 4.73, "learning_rate": 1.5817349857006671e-06, "loss": 3.0764, "step": 24846 }, { "epoch": 4.73, "learning_rate": 1.5816777883698761e-06, "loss": 2.3476, "step": 24847 }, { "epoch": 4.73, "learning_rate": 1.581620591039085e-06, "loss": 1.0008, "step": 24848 }, { "epoch": 4.73, "learning_rate": 1.5815633937082935e-06, "loss": 3.3559, "step": 24849 }, { "epoch": 4.73, "learning_rate": 1.5815061963775025e-06, "loss": 2.8181, "step": 24850 }, { "epoch": 4.73, "learning_rate": 1.581448999046711e-06, "loss": 2.1515, "step": 24851 }, { "epoch": 4.73, "learning_rate": 1.58139180171592e-06, "loss": 1.8427, "step": 24852 }, { "epoch": 4.73, "learning_rate": 1.5813346043851287e-06, "loss": 2.6137, "step": 24853 }, { "epoch": 4.73, "learning_rate": 1.5812774070543375e-06, "loss": 3.15, "step": 24854 }, { "epoch": 4.73, "learning_rate": 1.5812202097235465e-06, "loss": 1.9141, "step": 24855 }, { "epoch": 4.73, "learning_rate": 1.581163012392755e-06, "loss": 2.4979, "step": 24856 }, { "epoch": 4.73, "learning_rate": 1.5811058150619637e-06, "loss": 1.4615, "step": 24857 }, { "epoch": 4.73, "learning_rate": 1.5810486177311727e-06, "loss": 3.266, "step": 24858 }, { "epoch": 4.74, "learning_rate": 1.5809914204003814e-06, "loss": 1.9179, "step": 24859 }, { "epoch": 4.74, "learning_rate": 1.58093422306959e-06, "loss": 2.5634, "step": 24860 }, { "epoch": 4.74, "learning_rate": 1.580877025738799e-06, "loss": 1.8615, "step": 24861 }, { "epoch": 4.74, "learning_rate": 1.5808198284080076e-06, "loss": 2.5865, "step": 24862 }, { "epoch": 4.74, "learning_rate": 1.5807626310772162e-06, "loss": 2.4325, "step": 24863 }, { "epoch": 4.74, "learning_rate": 1.5807054337464252e-06, "loss": 2.4062, "step": 24864 }, { "epoch": 4.74, "learning_rate": 1.580648236415634e-06, "loss": 2.4914, "step": 24865 }, { "epoch": 4.74, "learning_rate": 1.5805910390848426e-06, "loss": 1.5747, "step": 24866 }, { "epoch": 4.74, "learning_rate": 1.5805338417540516e-06, "loss": 2.5073, "step": 24867 }, { "epoch": 4.74, "learning_rate": 1.5804766444232602e-06, "loss": 1.7196, "step": 24868 }, { "epoch": 4.74, "learning_rate": 1.5804194470924692e-06, "loss": 1.74, "step": 24869 }, { "epoch": 4.74, "learning_rate": 1.580362249761678e-06, "loss": 2.3025, "step": 24870 }, { "epoch": 4.74, "learning_rate": 1.5803050524308866e-06, "loss": 1.6857, "step": 24871 }, { "epoch": 4.74, "learning_rate": 1.5802478551000956e-06, "loss": 1.9624, "step": 24872 }, { "epoch": 4.74, "learning_rate": 1.5801906577693041e-06, "loss": 2.0091, "step": 24873 }, { "epoch": 4.74, "learning_rate": 1.5801334604385127e-06, "loss": 1.826, "step": 24874 }, { "epoch": 4.74, "learning_rate": 1.5800762631077217e-06, "loss": 1.5862, "step": 24875 }, { "epoch": 4.74, "learning_rate": 1.5800190657769305e-06, "loss": 1.7388, "step": 24876 }, { "epoch": 4.74, "learning_rate": 1.579961868446139e-06, "loss": 2.0793, "step": 24877 }, { "epoch": 4.74, "learning_rate": 1.5799046711153481e-06, "loss": 1.9088, "step": 24878 }, { "epoch": 4.74, "learning_rate": 1.5798474737845567e-06, "loss": 2.3783, "step": 24879 }, { "epoch": 4.74, "learning_rate": 1.5797902764537655e-06, "loss": 3.7279, "step": 24880 }, { "epoch": 4.74, "learning_rate": 1.5797330791229743e-06, "loss": 2.2328, "step": 24881 }, { "epoch": 4.74, "learning_rate": 1.579675881792183e-06, "loss": 1.9707, "step": 24882 }, { "epoch": 4.74, "learning_rate": 1.5796186844613917e-06, "loss": 2.4847, "step": 24883 }, { "epoch": 4.74, "learning_rate": 1.5795614871306007e-06, "loss": 1.7213, "step": 24884 }, { "epoch": 4.74, "learning_rate": 1.5795042897998092e-06, "loss": 1.7316, "step": 24885 }, { "epoch": 4.74, "learning_rate": 1.5794470924690182e-06, "loss": 1.477, "step": 24886 }, { "epoch": 4.74, "learning_rate": 1.579389895138227e-06, "loss": 2.2009, "step": 24887 }, { "epoch": 4.74, "learning_rate": 1.5793326978074356e-06, "loss": 2.2567, "step": 24888 }, { "epoch": 4.74, "learning_rate": 1.5792755004766446e-06, "loss": 1.4094, "step": 24889 }, { "epoch": 4.74, "learning_rate": 1.5792183031458532e-06, "loss": 1.5472, "step": 24890 }, { "epoch": 4.74, "learning_rate": 1.579161105815062e-06, "loss": 1.7516, "step": 24891 }, { "epoch": 4.74, "learning_rate": 1.5791039084842708e-06, "loss": 1.5142, "step": 24892 }, { "epoch": 4.74, "learning_rate": 1.5790467111534796e-06, "loss": 2.0038, "step": 24893 }, { "epoch": 4.74, "learning_rate": 1.5789895138226882e-06, "loss": 2.2532, "step": 24894 }, { "epoch": 4.74, "learning_rate": 1.5789323164918972e-06, "loss": 1.6029, "step": 24895 }, { "epoch": 4.74, "learning_rate": 1.5788751191611058e-06, "loss": 2.2854, "step": 24896 }, { "epoch": 4.74, "learning_rate": 1.5788179218303146e-06, "loss": 1.412, "step": 24897 }, { "epoch": 4.74, "learning_rate": 1.5787607244995236e-06, "loss": 1.7963, "step": 24898 }, { "epoch": 4.74, "learning_rate": 1.5787035271687321e-06, "loss": 1.3375, "step": 24899 }, { "epoch": 4.74, "learning_rate": 1.5786463298379407e-06, "loss": 2.5195, "step": 24900 }, { "epoch": 4.74, "learning_rate": 1.5785891325071497e-06, "loss": 2.0028, "step": 24901 }, { "epoch": 4.74, "learning_rate": 1.5785319351763583e-06, "loss": 1.2608, "step": 24902 }, { "epoch": 4.74, "learning_rate": 1.5784747378455673e-06, "loss": 1.6945, "step": 24903 }, { "epoch": 4.74, "learning_rate": 1.5784175405147761e-06, "loss": 2.3217, "step": 24904 }, { "epoch": 4.74, "learning_rate": 1.5783603431839847e-06, "loss": 1.1809, "step": 24905 }, { "epoch": 4.74, "learning_rate": 1.5783031458531937e-06, "loss": 1.5212, "step": 24906 }, { "epoch": 4.74, "learning_rate": 1.5782459485224023e-06, "loss": 2.9608, "step": 24907 }, { "epoch": 4.74, "learning_rate": 1.578188751191611e-06, "loss": 1.9405, "step": 24908 }, { "epoch": 4.74, "learning_rate": 1.5781315538608199e-06, "loss": 1.7027, "step": 24909 }, { "epoch": 4.74, "learning_rate": 1.5780743565300287e-06, "loss": 2.4745, "step": 24910 }, { "epoch": 4.74, "learning_rate": 1.5780171591992372e-06, "loss": 1.4377, "step": 24911 }, { "epoch": 4.75, "learning_rate": 1.5779599618684463e-06, "loss": 2.9173, "step": 24912 }, { "epoch": 4.75, "learning_rate": 1.5779027645376548e-06, "loss": 2.9636, "step": 24913 }, { "epoch": 4.75, "learning_rate": 1.5778455672068636e-06, "loss": 1.2032, "step": 24914 }, { "epoch": 4.75, "learning_rate": 1.5777883698760726e-06, "loss": 1.6117, "step": 24915 }, { "epoch": 4.75, "learning_rate": 1.5777311725452812e-06, "loss": 2.5498, "step": 24916 }, { "epoch": 4.75, "learning_rate": 1.5776739752144902e-06, "loss": 2.2397, "step": 24917 }, { "epoch": 4.75, "learning_rate": 1.5776167778836988e-06, "loss": 1.1446, "step": 24918 }, { "epoch": 4.75, "learning_rate": 1.5775595805529076e-06, "loss": 1.9512, "step": 24919 }, { "epoch": 4.75, "learning_rate": 1.5775023832221164e-06, "loss": 2.6032, "step": 24920 }, { "epoch": 4.75, "learning_rate": 1.5774451858913252e-06, "loss": 2.9941, "step": 24921 }, { "epoch": 4.75, "learning_rate": 1.5773879885605338e-06, "loss": 2.0891, "step": 24922 }, { "epoch": 4.75, "learning_rate": 1.5773307912297428e-06, "loss": 1.1518, "step": 24923 }, { "epoch": 4.75, "learning_rate": 1.5772735938989514e-06, "loss": 2.6324, "step": 24924 }, { "epoch": 4.75, "learning_rate": 1.5772163965681601e-06, "loss": 0.9193, "step": 24925 }, { "epoch": 4.75, "learning_rate": 1.5771591992373692e-06, "loss": 1.8032, "step": 24926 }, { "epoch": 4.75, "learning_rate": 1.5771020019065777e-06, "loss": 1.7183, "step": 24927 }, { "epoch": 4.75, "learning_rate": 1.5770448045757863e-06, "loss": 0.5993, "step": 24928 }, { "epoch": 4.75, "learning_rate": 1.5769876072449953e-06, "loss": 2.189, "step": 24929 }, { "epoch": 4.75, "learning_rate": 1.576930409914204e-06, "loss": 2.7269, "step": 24930 }, { "epoch": 4.75, "learning_rate": 1.5768732125834127e-06, "loss": 2.1299, "step": 24931 }, { "epoch": 4.75, "learning_rate": 1.5768160152526217e-06, "loss": 2.7313, "step": 24932 }, { "epoch": 4.75, "learning_rate": 1.5767588179218303e-06, "loss": 2.2582, "step": 24933 }, { "epoch": 4.75, "learning_rate": 1.5767016205910393e-06, "loss": 2.7899, "step": 24934 }, { "epoch": 4.75, "learning_rate": 1.5766444232602479e-06, "loss": 1.9263, "step": 24935 }, { "epoch": 4.75, "learning_rate": 1.5765872259294567e-06, "loss": 1.9925, "step": 24936 }, { "epoch": 4.75, "learning_rate": 1.5765300285986657e-06, "loss": 2.1015, "step": 24937 }, { "epoch": 4.75, "learning_rate": 1.5764728312678743e-06, "loss": 2.4155, "step": 24938 }, { "epoch": 4.75, "learning_rate": 1.5764156339370828e-06, "loss": 2.2077, "step": 24939 }, { "epoch": 4.75, "learning_rate": 1.5763584366062918e-06, "loss": 1.2129, "step": 24940 }, { "epoch": 4.75, "learning_rate": 1.5763012392755004e-06, "loss": 2.7315, "step": 24941 }, { "epoch": 4.75, "learning_rate": 1.5762440419447092e-06, "loss": 2.0616, "step": 24942 }, { "epoch": 4.75, "learning_rate": 1.5761868446139182e-06, "loss": 1.5933, "step": 24943 }, { "epoch": 4.75, "learning_rate": 1.5761296472831268e-06, "loss": 1.4805, "step": 24944 }, { "epoch": 4.75, "learning_rate": 1.5760724499523354e-06, "loss": 2.9753, "step": 24945 }, { "epoch": 4.75, "learning_rate": 1.5760152526215444e-06, "loss": 1.4293, "step": 24946 }, { "epoch": 4.75, "learning_rate": 1.5759580552907532e-06, "loss": 2.9372, "step": 24947 }, { "epoch": 4.75, "learning_rate": 1.5759008579599618e-06, "loss": 2.4394, "step": 24948 }, { "epoch": 4.75, "learning_rate": 1.5758436606291708e-06, "loss": 2.8743, "step": 24949 }, { "epoch": 4.75, "learning_rate": 1.5757864632983794e-06, "loss": 2.4908, "step": 24950 }, { "epoch": 4.75, "learning_rate": 1.5757292659675884e-06, "loss": 2.0865, "step": 24951 }, { "epoch": 4.75, "learning_rate": 1.575672068636797e-06, "loss": 2.8015, "step": 24952 }, { "epoch": 4.75, "learning_rate": 1.5756148713060057e-06, "loss": 2.8075, "step": 24953 }, { "epoch": 4.75, "learning_rate": 1.5755576739752147e-06, "loss": 1.133, "step": 24954 }, { "epoch": 4.75, "learning_rate": 1.5755004766444233e-06, "loss": 2.5247, "step": 24955 }, { "epoch": 4.75, "learning_rate": 1.575443279313632e-06, "loss": 2.8683, "step": 24956 }, { "epoch": 4.75, "learning_rate": 1.575386081982841e-06, "loss": 2.5165, "step": 24957 }, { "epoch": 4.75, "learning_rate": 1.5753288846520497e-06, "loss": 1.6185, "step": 24958 }, { "epoch": 4.75, "learning_rate": 1.5752716873212583e-06, "loss": 2.4789, "step": 24959 }, { "epoch": 4.75, "learning_rate": 1.5752144899904673e-06, "loss": 1.9877, "step": 24960 }, { "epoch": 4.75, "learning_rate": 1.5751572926596759e-06, "loss": 1.9252, "step": 24961 }, { "epoch": 4.75, "learning_rate": 1.5751000953288845e-06, "loss": 2.4336, "step": 24962 }, { "epoch": 4.75, "learning_rate": 1.5750428979980935e-06, "loss": 2.0551, "step": 24963 }, { "epoch": 4.76, "learning_rate": 1.5749857006673023e-06, "loss": 1.5456, "step": 24964 }, { "epoch": 4.76, "learning_rate": 1.5749285033365108e-06, "loss": 2.6157, "step": 24965 }, { "epoch": 4.76, "learning_rate": 1.5748713060057198e-06, "loss": 2.4012, "step": 24966 }, { "epoch": 4.76, "learning_rate": 1.5748141086749284e-06, "loss": 2.5814, "step": 24967 }, { "epoch": 4.76, "learning_rate": 1.5747569113441374e-06, "loss": 2.1259, "step": 24968 }, { "epoch": 4.76, "learning_rate": 1.574699714013346e-06, "loss": 0.8479, "step": 24969 }, { "epoch": 4.76, "learning_rate": 1.5746425166825548e-06, "loss": 2.1215, "step": 24970 }, { "epoch": 4.76, "learning_rate": 1.5745853193517638e-06, "loss": 3.1397, "step": 24971 }, { "epoch": 4.76, "learning_rate": 1.5745281220209724e-06, "loss": 1.9124, "step": 24972 }, { "epoch": 4.76, "learning_rate": 1.574470924690181e-06, "loss": 2.0194, "step": 24973 }, { "epoch": 4.76, "learning_rate": 1.57441372735939e-06, "loss": 2.6449, "step": 24974 }, { "epoch": 4.76, "learning_rate": 1.5743565300285988e-06, "loss": 3.3415, "step": 24975 }, { "epoch": 4.76, "learning_rate": 1.5742993326978074e-06, "loss": 1.9017, "step": 24976 }, { "epoch": 4.76, "learning_rate": 1.5742421353670164e-06, "loss": 1.6186, "step": 24977 }, { "epoch": 4.76, "learning_rate": 1.574184938036225e-06, "loss": 2.432, "step": 24978 }, { "epoch": 4.76, "learning_rate": 1.5741277407054337e-06, "loss": 1.7857, "step": 24979 }, { "epoch": 4.76, "learning_rate": 1.5740705433746425e-06, "loss": 1.5508, "step": 24980 }, { "epoch": 4.76, "learning_rate": 1.5740133460438513e-06, "loss": 2.2714, "step": 24981 }, { "epoch": 4.76, "learning_rate": 1.5739561487130603e-06, "loss": 1.1431, "step": 24982 }, { "epoch": 4.76, "learning_rate": 1.573898951382269e-06, "loss": 1.0961, "step": 24983 }, { "epoch": 4.76, "learning_rate": 1.5738417540514775e-06, "loss": 3.0694, "step": 24984 }, { "epoch": 4.76, "learning_rate": 1.5737845567206865e-06, "loss": 1.7979, "step": 24985 }, { "epoch": 4.76, "learning_rate": 1.5737273593898953e-06, "loss": 1.6741, "step": 24986 }, { "epoch": 4.76, "learning_rate": 1.5736701620591039e-06, "loss": 2.2552, "step": 24987 }, { "epoch": 4.76, "learning_rate": 1.5736129647283129e-06, "loss": 1.1124, "step": 24988 }, { "epoch": 4.76, "learning_rate": 1.5735557673975215e-06, "loss": 1.1868, "step": 24989 }, { "epoch": 4.76, "learning_rate": 1.57349857006673e-06, "loss": 1.6077, "step": 24990 }, { "epoch": 4.76, "learning_rate": 1.573441372735939e-06, "loss": 1.3499, "step": 24991 }, { "epoch": 4.76, "learning_rate": 1.5733841754051479e-06, "loss": 1.301, "step": 24992 }, { "epoch": 4.76, "learning_rate": 1.5733269780743564e-06, "loss": 2.446, "step": 24993 }, { "epoch": 4.76, "learning_rate": 1.5732697807435654e-06, "loss": 1.1927, "step": 24994 }, { "epoch": 4.76, "learning_rate": 1.573212583412774e-06, "loss": 3.0888, "step": 24995 }, { "epoch": 4.76, "learning_rate": 1.5731553860819828e-06, "loss": 1.7773, "step": 24996 }, { "epoch": 4.76, "learning_rate": 1.5730981887511916e-06, "loss": 2.5147, "step": 24997 }, { "epoch": 4.76, "learning_rate": 1.5730409914204004e-06, "loss": 1.8548, "step": 24998 }, { "epoch": 4.76, "learning_rate": 1.5729837940896094e-06, "loss": 0.8344, "step": 24999 }, { "epoch": 4.76, "learning_rate": 1.572926596758818e-06, "loss": 2.2392, "step": 25000 }, { "epoch": 4.76, "learning_rate": 1.5728693994280266e-06, "loss": 1.4659, "step": 25001 }, { "epoch": 4.76, "learning_rate": 1.5728122020972356e-06, "loss": 1.3215, "step": 25002 }, { "epoch": 4.76, "learning_rate": 1.5727550047664444e-06, "loss": 2.3572, "step": 25003 }, { "epoch": 4.76, "learning_rate": 1.572697807435653e-06, "loss": 1.5098, "step": 25004 }, { "epoch": 4.76, "learning_rate": 1.572640610104862e-06, "loss": 1.839, "step": 25005 }, { "epoch": 4.76, "learning_rate": 1.5725834127740705e-06, "loss": 1.3432, "step": 25006 }, { "epoch": 4.76, "learning_rate": 1.5725262154432793e-06, "loss": 1.9584, "step": 25007 }, { "epoch": 4.76, "learning_rate": 1.5724690181124881e-06, "loss": 3.226, "step": 25008 }, { "epoch": 4.76, "learning_rate": 1.572411820781697e-06, "loss": 2.096, "step": 25009 }, { "epoch": 4.76, "learning_rate": 1.5723546234509055e-06, "loss": 1.1237, "step": 25010 }, { "epoch": 4.76, "learning_rate": 1.5722974261201145e-06, "loss": 2.2599, "step": 25011 }, { "epoch": 4.76, "learning_rate": 1.572240228789323e-06, "loss": 4.2614, "step": 25012 }, { "epoch": 4.76, "learning_rate": 1.5721830314585319e-06, "loss": 2.9577, "step": 25013 }, { "epoch": 4.76, "learning_rate": 1.5721258341277409e-06, "loss": 1.1488, "step": 25014 }, { "epoch": 4.76, "learning_rate": 1.5720686367969495e-06, "loss": 1.4634, "step": 25015 }, { "epoch": 4.76, "learning_rate": 1.5720114394661585e-06, "loss": 1.267, "step": 25016 }, { "epoch": 4.77, "learning_rate": 1.571954242135367e-06, "loss": 2.3939, "step": 25017 }, { "epoch": 4.77, "learning_rate": 1.5718970448045756e-06, "loss": 2.8581, "step": 25018 }, { "epoch": 4.77, "learning_rate": 1.5718398474737846e-06, "loss": 2.0929, "step": 25019 }, { "epoch": 4.77, "learning_rate": 1.5717826501429934e-06, "loss": 1.0871, "step": 25020 }, { "epoch": 4.77, "learning_rate": 1.571725452812202e-06, "loss": 2.6554, "step": 25021 }, { "epoch": 4.77, "learning_rate": 1.571668255481411e-06, "loss": 2.9815, "step": 25022 }, { "epoch": 4.77, "learning_rate": 1.5716110581506196e-06, "loss": 1.5879, "step": 25023 }, { "epoch": 4.77, "learning_rate": 1.5715538608198284e-06, "loss": 2.5203, "step": 25024 }, { "epoch": 4.77, "learning_rate": 1.5714966634890374e-06, "loss": 1.4796, "step": 25025 }, { "epoch": 4.77, "learning_rate": 1.571439466158246e-06, "loss": 2.8874, "step": 25026 }, { "epoch": 4.77, "learning_rate": 1.5713822688274546e-06, "loss": 2.8381, "step": 25027 }, { "epoch": 4.77, "learning_rate": 1.5713250714966636e-06, "loss": 3.1898, "step": 25028 }, { "epoch": 4.77, "learning_rate": 1.5712678741658722e-06, "loss": 2.5715, "step": 25029 }, { "epoch": 4.77, "learning_rate": 1.571210676835081e-06, "loss": 2.6035, "step": 25030 }, { "epoch": 4.77, "learning_rate": 1.57115347950429e-06, "loss": 1.8649, "step": 25031 }, { "epoch": 4.77, "learning_rate": 1.5710962821734985e-06, "loss": 2.131, "step": 25032 }, { "epoch": 4.77, "learning_rate": 1.5710390848427076e-06, "loss": 2.4735, "step": 25033 }, { "epoch": 4.77, "learning_rate": 1.5709818875119161e-06, "loss": 2.1289, "step": 25034 }, { "epoch": 4.77, "learning_rate": 1.570924690181125e-06, "loss": 1.9737, "step": 25035 }, { "epoch": 4.77, "learning_rate": 1.5708674928503337e-06, "loss": 2.5571, "step": 25036 }, { "epoch": 4.77, "learning_rate": 1.5708102955195425e-06, "loss": 2.0484, "step": 25037 }, { "epoch": 4.77, "learning_rate": 1.570753098188751e-06, "loss": 1.9194, "step": 25038 }, { "epoch": 4.77, "learning_rate": 1.57069590085796e-06, "loss": 2.6123, "step": 25039 }, { "epoch": 4.77, "learning_rate": 1.5706387035271687e-06, "loss": 2.455, "step": 25040 }, { "epoch": 4.77, "learning_rate": 1.5705815061963775e-06, "loss": 2.7477, "step": 25041 }, { "epoch": 4.77, "learning_rate": 1.5705243088655865e-06, "loss": 1.9095, "step": 25042 }, { "epoch": 4.77, "learning_rate": 1.570467111534795e-06, "loss": 2.7725, "step": 25043 }, { "epoch": 4.77, "learning_rate": 1.5704099142040036e-06, "loss": 2.1997, "step": 25044 }, { "epoch": 4.77, "learning_rate": 1.5703527168732127e-06, "loss": 1.7229, "step": 25045 }, { "epoch": 4.77, "learning_rate": 1.5702955195424214e-06, "loss": 1.9052, "step": 25046 }, { "epoch": 4.77, "learning_rate": 1.5702383222116302e-06, "loss": 2.1279, "step": 25047 }, { "epoch": 4.77, "learning_rate": 1.570181124880839e-06, "loss": 2.5007, "step": 25048 }, { "epoch": 4.77, "learning_rate": 1.5701239275500476e-06, "loss": 3.9717, "step": 25049 }, { "epoch": 4.77, "learning_rate": 1.5700667302192566e-06, "loss": 2.3449, "step": 25050 }, { "epoch": 4.77, "learning_rate": 1.5700095328884652e-06, "loss": 2.7931, "step": 25051 }, { "epoch": 4.77, "learning_rate": 1.569952335557674e-06, "loss": 2.1113, "step": 25052 }, { "epoch": 4.77, "learning_rate": 1.569895138226883e-06, "loss": 2.0787, "step": 25053 }, { "epoch": 4.77, "learning_rate": 1.5698379408960916e-06, "loss": 1.0449, "step": 25054 }, { "epoch": 4.77, "learning_rate": 1.5697807435653002e-06, "loss": 1.4998, "step": 25055 }, { "epoch": 4.77, "learning_rate": 1.5697235462345092e-06, "loss": 1.8603, "step": 25056 }, { "epoch": 4.77, "learning_rate": 1.5696663489037178e-06, "loss": 1.9977, "step": 25057 }, { "epoch": 4.77, "learning_rate": 1.5696091515729265e-06, "loss": 2.0109, "step": 25058 }, { "epoch": 4.77, "learning_rate": 1.5695519542421356e-06, "loss": 2.5707, "step": 25059 }, { "epoch": 4.77, "learning_rate": 1.5694947569113441e-06, "loss": 2.2265, "step": 25060 }, { "epoch": 4.77, "learning_rate": 1.5694375595805527e-06, "loss": 1.555, "step": 25061 }, { "epoch": 4.77, "learning_rate": 1.5693803622497617e-06, "loss": 1.3374, "step": 25062 }, { "epoch": 4.77, "learning_rate": 1.5693231649189705e-06, "loss": 2.2606, "step": 25063 }, { "epoch": 4.77, "learning_rate": 1.5692659675881793e-06, "loss": 2.2857, "step": 25064 }, { "epoch": 4.77, "learning_rate": 1.569208770257388e-06, "loss": 1.8504, "step": 25065 }, { "epoch": 4.77, "learning_rate": 1.5691515729265967e-06, "loss": 2.4424, "step": 25066 }, { "epoch": 4.77, "learning_rate": 1.5690943755958057e-06, "loss": 1.5339, "step": 25067 }, { "epoch": 4.77, "learning_rate": 1.5690371782650143e-06, "loss": 1.8405, "step": 25068 }, { "epoch": 4.78, "learning_rate": 1.568979980934223e-06, "loss": 2.2443, "step": 25069 }, { "epoch": 4.78, "learning_rate": 1.568922783603432e-06, "loss": 2.2176, "step": 25070 }, { "epoch": 4.78, "learning_rate": 1.5688655862726407e-06, "loss": 2.8147, "step": 25071 }, { "epoch": 4.78, "learning_rate": 1.5688083889418492e-06, "loss": 1.9213, "step": 25072 }, { "epoch": 4.78, "learning_rate": 1.5687511916110582e-06, "loss": 0.9654, "step": 25073 }, { "epoch": 4.78, "learning_rate": 1.568693994280267e-06, "loss": 2.9477, "step": 25074 }, { "epoch": 4.78, "learning_rate": 1.5686367969494756e-06, "loss": 1.461, "step": 25075 }, { "epoch": 4.78, "learning_rate": 1.5685795996186846e-06, "loss": 2.5008, "step": 25076 }, { "epoch": 4.78, "learning_rate": 1.5685224022878932e-06, "loss": 1.7389, "step": 25077 }, { "epoch": 4.78, "learning_rate": 1.5684652049571018e-06, "loss": 1.2446, "step": 25078 }, { "epoch": 4.78, "learning_rate": 1.5684080076263108e-06, "loss": 1.9336, "step": 25079 }, { "epoch": 4.78, "learning_rate": 1.5683508102955196e-06, "loss": 2.129, "step": 25080 }, { "epoch": 4.78, "learning_rate": 1.5682936129647286e-06, "loss": 1.3604, "step": 25081 }, { "epoch": 4.78, "learning_rate": 1.5682364156339372e-06, "loss": 3.1281, "step": 25082 }, { "epoch": 4.78, "learning_rate": 1.5681792183031458e-06, "loss": 1.7361, "step": 25083 }, { "epoch": 4.78, "learning_rate": 1.5681220209723548e-06, "loss": 2.1199, "step": 25084 }, { "epoch": 4.78, "learning_rate": 1.5680648236415633e-06, "loss": 2.4427, "step": 25085 }, { "epoch": 4.78, "learning_rate": 1.5680076263107721e-06, "loss": 2.0043, "step": 25086 }, { "epoch": 4.78, "learning_rate": 1.5679504289799811e-06, "loss": 2.8127, "step": 25087 }, { "epoch": 4.78, "learning_rate": 1.5678932316491897e-06, "loss": 1.6217, "step": 25088 }, { "epoch": 4.78, "learning_rate": 1.5678360343183983e-06, "loss": 1.772, "step": 25089 }, { "epoch": 4.78, "learning_rate": 1.5677788369876073e-06, "loss": 1.2629, "step": 25090 }, { "epoch": 4.78, "learning_rate": 1.5677216396568161e-06, "loss": 1.8061, "step": 25091 }, { "epoch": 4.78, "learning_rate": 1.5676644423260247e-06, "loss": 1.2531, "step": 25092 }, { "epoch": 4.78, "learning_rate": 1.5676072449952337e-06, "loss": 2.1008, "step": 25093 }, { "epoch": 4.78, "learning_rate": 1.5675500476644423e-06, "loss": 2.386, "step": 25094 }, { "epoch": 4.78, "learning_rate": 1.567492850333651e-06, "loss": 2.537, "step": 25095 }, { "epoch": 4.78, "learning_rate": 1.5674356530028599e-06, "loss": 1.904, "step": 25096 }, { "epoch": 4.78, "learning_rate": 1.5673784556720687e-06, "loss": 1.6658, "step": 25097 }, { "epoch": 4.78, "learning_rate": 1.5673212583412777e-06, "loss": 2.9139, "step": 25098 }, { "epoch": 4.78, "learning_rate": 1.5672640610104862e-06, "loss": 1.234, "step": 25099 }, { "epoch": 4.78, "learning_rate": 1.5672068636796948e-06, "loss": 2.0723, "step": 25100 }, { "epoch": 4.78, "learning_rate": 1.5671496663489038e-06, "loss": 1.5339, "step": 25101 }, { "epoch": 4.78, "learning_rate": 1.5670924690181126e-06, "loss": 3.2545, "step": 25102 }, { "epoch": 4.78, "learning_rate": 1.5670352716873212e-06, "loss": 1.6579, "step": 25103 }, { "epoch": 4.78, "learning_rate": 1.5669780743565302e-06, "loss": 1.7464, "step": 25104 }, { "epoch": 4.78, "learning_rate": 1.5669208770257388e-06, "loss": 1.7075, "step": 25105 }, { "epoch": 4.78, "learning_rate": 1.5668636796949474e-06, "loss": 1.7597, "step": 25106 }, { "epoch": 4.78, "learning_rate": 1.5668064823641564e-06, "loss": 2.4199, "step": 25107 }, { "epoch": 4.78, "learning_rate": 1.5667492850333652e-06, "loss": 2.1386, "step": 25108 }, { "epoch": 4.78, "learning_rate": 1.5666920877025738e-06, "loss": 1.6615, "step": 25109 }, { "epoch": 4.78, "learning_rate": 1.5666348903717828e-06, "loss": 2.5345, "step": 25110 }, { "epoch": 4.78, "learning_rate": 1.5665776930409914e-06, "loss": 2.2814, "step": 25111 }, { "epoch": 4.78, "learning_rate": 1.5665204957102004e-06, "loss": 3.1531, "step": 25112 }, { "epoch": 4.78, "learning_rate": 1.5664632983794092e-06, "loss": 2.4007, "step": 25113 }, { "epoch": 4.78, "learning_rate": 1.5664061010486177e-06, "loss": 1.697, "step": 25114 }, { "epoch": 4.78, "learning_rate": 1.5663489037178267e-06, "loss": 1.8963, "step": 25115 }, { "epoch": 4.78, "learning_rate": 1.5662917063870353e-06, "loss": 1.8264, "step": 25116 }, { "epoch": 4.78, "learning_rate": 1.566234509056244e-06, "loss": 2.5952, "step": 25117 }, { "epoch": 4.78, "learning_rate": 1.566177311725453e-06, "loss": 2.2723, "step": 25118 }, { "epoch": 4.78, "learning_rate": 1.5661201143946617e-06, "loss": 2.4869, "step": 25119 }, { "epoch": 4.78, "learning_rate": 1.5660629170638703e-06, "loss": 2.0256, "step": 25120 }, { "epoch": 4.78, "learning_rate": 1.5660057197330793e-06, "loss": 1.7175, "step": 25121 }, { "epoch": 4.79, "learning_rate": 1.5659485224022879e-06, "loss": 2.4, "step": 25122 }, { "epoch": 4.79, "learning_rate": 1.5658913250714967e-06, "loss": 2.2217, "step": 25123 }, { "epoch": 4.79, "learning_rate": 1.5658341277407055e-06, "loss": 2.7378, "step": 25124 }, { "epoch": 4.79, "learning_rate": 1.5657769304099143e-06, "loss": 3.0106, "step": 25125 }, { "epoch": 4.79, "learning_rate": 1.5657197330791228e-06, "loss": 2.7094, "step": 25126 }, { "epoch": 4.79, "learning_rate": 1.5656625357483318e-06, "loss": 1.1423, "step": 25127 }, { "epoch": 4.79, "learning_rate": 1.5656053384175404e-06, "loss": 1.8743, "step": 25128 }, { "epoch": 4.79, "learning_rate": 1.5655481410867494e-06, "loss": 2.9422, "step": 25129 }, { "epoch": 4.79, "learning_rate": 1.5654909437559582e-06, "loss": 1.5327, "step": 25130 }, { "epoch": 4.79, "learning_rate": 1.5654337464251668e-06, "loss": 2.9341, "step": 25131 }, { "epoch": 4.79, "learning_rate": 1.5653765490943758e-06, "loss": 1.9572, "step": 25132 }, { "epoch": 4.79, "learning_rate": 1.5653193517635844e-06, "loss": 1.3754, "step": 25133 }, { "epoch": 4.79, "learning_rate": 1.5652621544327932e-06, "loss": 1.7052, "step": 25134 }, { "epoch": 4.79, "learning_rate": 1.565204957102002e-06, "loss": 1.9853, "step": 25135 }, { "epoch": 4.79, "learning_rate": 1.5651477597712108e-06, "loss": 2.1937, "step": 25136 }, { "epoch": 4.79, "learning_rate": 1.5650905624404194e-06, "loss": 2.888, "step": 25137 }, { "epoch": 4.79, "learning_rate": 1.5650333651096284e-06, "loss": 2.5248, "step": 25138 }, { "epoch": 4.79, "learning_rate": 1.564976167778837e-06, "loss": 1.1215, "step": 25139 }, { "epoch": 4.79, "learning_rate": 1.5649189704480457e-06, "loss": 2.0986, "step": 25140 }, { "epoch": 4.79, "learning_rate": 1.5648617731172547e-06, "loss": 1.7529, "step": 25141 }, { "epoch": 4.79, "learning_rate": 1.5648045757864633e-06, "loss": 1.8657, "step": 25142 }, { "epoch": 4.79, "learning_rate": 1.564747378455672e-06, "loss": 3.4342, "step": 25143 }, { "epoch": 4.79, "learning_rate": 1.564690181124881e-06, "loss": 1.1166, "step": 25144 }, { "epoch": 4.79, "learning_rate": 1.5646329837940895e-06, "loss": 2.3877, "step": 25145 }, { "epoch": 4.79, "learning_rate": 1.5645757864632985e-06, "loss": 2.0017, "step": 25146 }, { "epoch": 4.79, "learning_rate": 1.5645185891325073e-06, "loss": 0.7407, "step": 25147 }, { "epoch": 4.79, "learning_rate": 1.5644613918017159e-06, "loss": 2.4247, "step": 25148 }, { "epoch": 4.79, "learning_rate": 1.5644041944709249e-06, "loss": 1.8389, "step": 25149 }, { "epoch": 4.79, "learning_rate": 1.5643469971401335e-06, "loss": 2.3258, "step": 25150 }, { "epoch": 4.79, "learning_rate": 1.5642897998093423e-06, "loss": 2.5819, "step": 25151 }, { "epoch": 4.79, "learning_rate": 1.564232602478551e-06, "loss": 1.6138, "step": 25152 }, { "epoch": 4.79, "learning_rate": 1.5641754051477598e-06, "loss": 2.7604, "step": 25153 }, { "epoch": 4.79, "learning_rate": 1.5641182078169684e-06, "loss": 2.3093, "step": 25154 }, { "epoch": 4.79, "learning_rate": 1.5640610104861774e-06, "loss": 1.4602, "step": 25155 }, { "epoch": 4.79, "learning_rate": 1.564003813155386e-06, "loss": 2.443, "step": 25156 }, { "epoch": 4.79, "learning_rate": 1.5639466158245948e-06, "loss": 2.4829, "step": 25157 }, { "epoch": 4.79, "learning_rate": 1.5638894184938038e-06, "loss": 1.6462, "step": 25158 }, { "epoch": 4.79, "learning_rate": 1.5638322211630124e-06, "loss": 1.5975, "step": 25159 }, { "epoch": 4.79, "learning_rate": 1.563775023832221e-06, "loss": 3.0095, "step": 25160 }, { "epoch": 4.79, "learning_rate": 1.56371782650143e-06, "loss": 1.2949, "step": 25161 }, { "epoch": 4.79, "learning_rate": 1.5636606291706388e-06, "loss": 1.2975, "step": 25162 }, { "epoch": 4.79, "learning_rate": 1.5636034318398476e-06, "loss": 2.6191, "step": 25163 }, { "epoch": 4.79, "learning_rate": 1.5635462345090564e-06, "loss": 1.4528, "step": 25164 }, { "epoch": 4.79, "learning_rate": 1.563489037178265e-06, "loss": 2.3508, "step": 25165 }, { "epoch": 4.79, "learning_rate": 1.563431839847474e-06, "loss": 1.6717, "step": 25166 }, { "epoch": 4.79, "learning_rate": 1.5633746425166825e-06, "loss": 1.587, "step": 25167 }, { "epoch": 4.79, "learning_rate": 1.5633174451858913e-06, "loss": 2.452, "step": 25168 }, { "epoch": 4.79, "learning_rate": 1.5632602478551003e-06, "loss": 2.0228, "step": 25169 }, { "epoch": 4.79, "learning_rate": 1.563203050524309e-06, "loss": 2.8907, "step": 25170 }, { "epoch": 4.79, "learning_rate": 1.5631458531935175e-06, "loss": 2.0082, "step": 25171 }, { "epoch": 4.79, "learning_rate": 1.5630886558627265e-06, "loss": 0.6137, "step": 25172 }, { "epoch": 4.79, "learning_rate": 1.563031458531935e-06, "loss": 1.3893, "step": 25173 }, { "epoch": 4.8, "learning_rate": 1.5629742612011439e-06, "loss": 1.8592, "step": 25174 }, { "epoch": 4.8, "learning_rate": 1.5629170638703529e-06, "loss": 1.2018, "step": 25175 }, { "epoch": 4.8, "learning_rate": 1.5628598665395615e-06, "loss": 1.9339, "step": 25176 }, { "epoch": 4.8, "learning_rate": 1.5628026692087705e-06, "loss": 3.2013, "step": 25177 }, { "epoch": 4.8, "learning_rate": 1.562745471877979e-06, "loss": 2.2293, "step": 25178 }, { "epoch": 4.8, "learning_rate": 1.5626882745471878e-06, "loss": 2.8107, "step": 25179 }, { "epoch": 4.8, "learning_rate": 1.5626310772163969e-06, "loss": 2.1161, "step": 25180 }, { "epoch": 4.8, "learning_rate": 1.5625738798856054e-06, "loss": 1.9371, "step": 25181 }, { "epoch": 4.8, "learning_rate": 1.562516682554814e-06, "loss": 2.0858, "step": 25182 }, { "epoch": 4.8, "learning_rate": 1.562459485224023e-06, "loss": 2.0163, "step": 25183 }, { "epoch": 4.8, "learning_rate": 1.5624022878932316e-06, "loss": 1.5829, "step": 25184 }, { "epoch": 4.8, "learning_rate": 1.5623450905624404e-06, "loss": 3.0488, "step": 25185 }, { "epoch": 4.8, "learning_rate": 1.5622878932316494e-06, "loss": 1.5565, "step": 25186 }, { "epoch": 4.8, "learning_rate": 1.562230695900858e-06, "loss": 1.7627, "step": 25187 }, { "epoch": 4.8, "learning_rate": 1.5621734985700666e-06, "loss": 1.1329, "step": 25188 }, { "epoch": 4.8, "learning_rate": 1.5621163012392756e-06, "loss": 1.9777, "step": 25189 }, { "epoch": 4.8, "learning_rate": 1.5620591039084844e-06, "loss": 2.1948, "step": 25190 }, { "epoch": 4.8, "learning_rate": 1.562001906577693e-06, "loss": 1.6814, "step": 25191 }, { "epoch": 4.8, "learning_rate": 1.561944709246902e-06, "loss": 1.8232, "step": 25192 }, { "epoch": 4.8, "learning_rate": 1.5618875119161105e-06, "loss": 2.7598, "step": 25193 }, { "epoch": 4.8, "learning_rate": 1.5618303145853195e-06, "loss": 2.2791, "step": 25194 }, { "epoch": 4.8, "learning_rate": 1.5617731172545281e-06, "loss": 2.3149, "step": 25195 }, { "epoch": 4.8, "learning_rate": 1.561715919923737e-06, "loss": 2.0622, "step": 25196 }, { "epoch": 4.8, "learning_rate": 1.561658722592946e-06, "loss": 1.4567, "step": 25197 }, { "epoch": 4.8, "learning_rate": 1.5616015252621545e-06, "loss": 2.3684, "step": 25198 }, { "epoch": 4.8, "learning_rate": 1.561544327931363e-06, "loss": 1.7748, "step": 25199 }, { "epoch": 4.8, "learning_rate": 1.561487130600572e-06, "loss": 1.9868, "step": 25200 }, { "epoch": 4.8, "learning_rate": 1.5614299332697809e-06, "loss": 2.3237, "step": 25201 }, { "epoch": 4.8, "learning_rate": 1.5613727359389895e-06, "loss": 2.8306, "step": 25202 }, { "epoch": 4.8, "learning_rate": 1.5613155386081985e-06, "loss": 1.7519, "step": 25203 }, { "epoch": 4.8, "learning_rate": 1.561258341277407e-06, "loss": 2.7214, "step": 25204 }, { "epoch": 4.8, "learning_rate": 1.5612011439466156e-06, "loss": 3.0869, "step": 25205 }, { "epoch": 4.8, "learning_rate": 1.5611439466158246e-06, "loss": 1.8735, "step": 25206 }, { "epoch": 4.8, "learning_rate": 1.5610867492850334e-06, "loss": 0.5505, "step": 25207 }, { "epoch": 4.8, "learning_rate": 1.561029551954242e-06, "loss": 1.2679, "step": 25208 }, { "epoch": 4.8, "learning_rate": 1.560972354623451e-06, "loss": 1.7627, "step": 25209 }, { "epoch": 4.8, "learning_rate": 1.5609151572926596e-06, "loss": 2.991, "step": 25210 }, { "epoch": 4.8, "learning_rate": 1.5608579599618686e-06, "loss": 2.763, "step": 25211 }, { "epoch": 4.8, "learning_rate": 1.5608007626310772e-06, "loss": 2.1973, "step": 25212 }, { "epoch": 4.8, "learning_rate": 1.560743565300286e-06, "loss": 1.8563, "step": 25213 }, { "epoch": 4.8, "learning_rate": 1.560686367969495e-06, "loss": 2.0103, "step": 25214 }, { "epoch": 4.8, "learning_rate": 1.5606291706387036e-06, "loss": 1.7467, "step": 25215 }, { "epoch": 4.8, "learning_rate": 1.5605719733079122e-06, "loss": 1.8971, "step": 25216 }, { "epoch": 4.8, "learning_rate": 1.5605147759771212e-06, "loss": 1.3698, "step": 25217 }, { "epoch": 4.8, "learning_rate": 1.56045757864633e-06, "loss": 2.2094, "step": 25218 }, { "epoch": 4.8, "learning_rate": 1.5604003813155385e-06, "loss": 1.4649, "step": 25219 }, { "epoch": 4.8, "learning_rate": 1.5603431839847475e-06, "loss": 1.9362, "step": 25220 }, { "epoch": 4.8, "learning_rate": 1.5602859866539561e-06, "loss": 2.658, "step": 25221 }, { "epoch": 4.8, "learning_rate": 1.5602287893231647e-06, "loss": 2.679, "step": 25222 }, { "epoch": 4.8, "learning_rate": 1.5601715919923737e-06, "loss": 1.6335, "step": 25223 }, { "epoch": 4.8, "learning_rate": 1.5601143946615825e-06, "loss": 1.8184, "step": 25224 }, { "epoch": 4.8, "learning_rate": 1.560057197330791e-06, "loss": 2.3728, "step": 25225 }, { "epoch": 4.8, "learning_rate": 1.56e-06, "loss": 2.074, "step": 25226 }, { "epoch": 4.81, "learning_rate": 1.5599428026692087e-06, "loss": 1.4331, "step": 25227 }, { "epoch": 4.81, "learning_rate": 1.5598856053384177e-06, "loss": 1.53, "step": 25228 }, { "epoch": 4.81, "learning_rate": 1.5598284080076265e-06, "loss": 2.191, "step": 25229 }, { "epoch": 4.81, "learning_rate": 1.559771210676835e-06, "loss": 2.2885, "step": 25230 }, { "epoch": 4.81, "learning_rate": 1.559714013346044e-06, "loss": 1.6629, "step": 25231 }, { "epoch": 4.81, "learning_rate": 1.5596568160152526e-06, "loss": 2.6044, "step": 25232 }, { "epoch": 4.81, "learning_rate": 1.5595996186844612e-06, "loss": 2.2956, "step": 25233 }, { "epoch": 4.81, "learning_rate": 1.5595424213536702e-06, "loss": 2.8401, "step": 25234 }, { "epoch": 4.81, "learning_rate": 1.559485224022879e-06, "loss": 2.4056, "step": 25235 }, { "epoch": 4.81, "learning_rate": 1.5594280266920876e-06, "loss": 1.6302, "step": 25236 }, { "epoch": 4.81, "learning_rate": 1.5593708293612966e-06, "loss": 3.6178, "step": 25237 }, { "epoch": 4.81, "learning_rate": 1.5593136320305052e-06, "loss": 0.9751, "step": 25238 }, { "epoch": 4.81, "learning_rate": 1.559256434699714e-06, "loss": 1.9175, "step": 25239 }, { "epoch": 4.81, "learning_rate": 1.5591992373689228e-06, "loss": 1.3638, "step": 25240 }, { "epoch": 4.81, "learning_rate": 1.5591420400381316e-06, "loss": 1.6288, "step": 25241 }, { "epoch": 4.81, "learning_rate": 1.5590848427073406e-06, "loss": 1.0379, "step": 25242 }, { "epoch": 4.81, "learning_rate": 1.5590276453765492e-06, "loss": 2.6783, "step": 25243 }, { "epoch": 4.81, "learning_rate": 1.5589704480457578e-06, "loss": 1.7393, "step": 25244 }, { "epoch": 4.81, "learning_rate": 1.5589132507149668e-06, "loss": 1.9287, "step": 25245 }, { "epoch": 4.81, "learning_rate": 1.5588560533841756e-06, "loss": 1.478, "step": 25246 }, { "epoch": 4.81, "learning_rate": 1.5587988560533841e-06, "loss": 1.8283, "step": 25247 }, { "epoch": 4.81, "learning_rate": 1.5587416587225931e-06, "loss": 3.2506, "step": 25248 }, { "epoch": 4.81, "learning_rate": 1.5586844613918017e-06, "loss": 1.5399, "step": 25249 }, { "epoch": 4.81, "learning_rate": 1.5586272640610105e-06, "loss": 3.9238, "step": 25250 }, { "epoch": 4.81, "learning_rate": 1.5585700667302193e-06, "loss": 2.1849, "step": 25251 }, { "epoch": 4.81, "learning_rate": 1.558512869399428e-06, "loss": 1.7606, "step": 25252 }, { "epoch": 4.81, "learning_rate": 1.5584556720686367e-06, "loss": 2.8514, "step": 25253 }, { "epoch": 4.81, "learning_rate": 1.5583984747378457e-06, "loss": 1.9716, "step": 25254 }, { "epoch": 4.81, "learning_rate": 1.5583412774070543e-06, "loss": 2.2015, "step": 25255 }, { "epoch": 4.81, "learning_rate": 1.558284080076263e-06, "loss": 2.8317, "step": 25256 }, { "epoch": 4.81, "learning_rate": 1.558226882745472e-06, "loss": 1.7814, "step": 25257 }, { "epoch": 4.81, "learning_rate": 1.5581696854146807e-06, "loss": 1.4339, "step": 25258 }, { "epoch": 4.81, "learning_rate": 1.5581124880838897e-06, "loss": 2.168, "step": 25259 }, { "epoch": 4.81, "learning_rate": 1.5580552907530982e-06, "loss": 3.9834, "step": 25260 }, { "epoch": 4.81, "learning_rate": 1.5579980934223068e-06, "loss": 2.1752, "step": 25261 }, { "epoch": 4.81, "learning_rate": 1.5579408960915158e-06, "loss": 1.177, "step": 25262 }, { "epoch": 4.81, "learning_rate": 1.5578836987607246e-06, "loss": 2.1431, "step": 25263 }, { "epoch": 4.81, "learning_rate": 1.5578265014299332e-06, "loss": 2.656, "step": 25264 }, { "epoch": 4.81, "learning_rate": 1.5577693040991422e-06, "loss": 3.5776, "step": 25265 }, { "epoch": 4.81, "learning_rate": 1.5577121067683508e-06, "loss": 3.2937, "step": 25266 }, { "epoch": 4.81, "learning_rate": 1.5576549094375596e-06, "loss": 1.7127, "step": 25267 }, { "epoch": 4.81, "learning_rate": 1.5575977121067686e-06, "loss": 3.1709, "step": 25268 }, { "epoch": 4.81, "learning_rate": 1.5575405147759772e-06, "loss": 2.1624, "step": 25269 }, { "epoch": 4.81, "learning_rate": 1.5574833174451858e-06, "loss": 1.0202, "step": 25270 }, { "epoch": 4.81, "learning_rate": 1.5574261201143948e-06, "loss": 2.4919, "step": 25271 }, { "epoch": 4.81, "learning_rate": 1.5573689227836033e-06, "loss": 0.8934, "step": 25272 }, { "epoch": 4.81, "learning_rate": 1.5573117254528121e-06, "loss": 1.9277, "step": 25273 }, { "epoch": 4.81, "learning_rate": 1.5572545281220211e-06, "loss": 3.3265, "step": 25274 }, { "epoch": 4.81, "learning_rate": 1.5571973307912297e-06, "loss": 1.2834, "step": 25275 }, { "epoch": 4.81, "learning_rate": 1.5571401334604387e-06, "loss": 2.2101, "step": 25276 }, { "epoch": 4.81, "learning_rate": 1.5570829361296473e-06, "loss": 1.9003, "step": 25277 }, { "epoch": 4.81, "learning_rate": 1.5570257387988561e-06, "loss": 2.0093, "step": 25278 }, { "epoch": 4.82, "learning_rate": 1.556968541468065e-06, "loss": 2.4254, "step": 25279 }, { "epoch": 4.82, "learning_rate": 1.5569113441372737e-06, "loss": 1.164, "step": 25280 }, { "epoch": 4.82, "learning_rate": 1.5568541468064823e-06, "loss": 2.0129, "step": 25281 }, { "epoch": 4.82, "learning_rate": 1.5567969494756913e-06, "loss": 2.3876, "step": 25282 }, { "epoch": 4.82, "learning_rate": 1.5567397521448999e-06, "loss": 1.3768, "step": 25283 }, { "epoch": 4.82, "learning_rate": 1.5566825548141087e-06, "loss": 2.2169, "step": 25284 }, { "epoch": 4.82, "learning_rate": 1.5566253574833177e-06, "loss": 1.9473, "step": 25285 }, { "epoch": 4.82, "learning_rate": 1.5565681601525262e-06, "loss": 2.3103, "step": 25286 }, { "epoch": 4.82, "learning_rate": 1.5565109628217348e-06, "loss": 2.6199, "step": 25287 }, { "epoch": 4.82, "learning_rate": 1.5564537654909438e-06, "loss": 1.9065, "step": 25288 }, { "epoch": 4.82, "learning_rate": 1.5563965681601526e-06, "loss": 1.8435, "step": 25289 }, { "epoch": 4.82, "learning_rate": 1.5563393708293612e-06, "loss": 1.6129, "step": 25290 }, { "epoch": 4.82, "learning_rate": 1.5562821734985702e-06, "loss": 2.2063, "step": 25291 }, { "epoch": 4.82, "learning_rate": 1.5562249761677788e-06, "loss": 3.0617, "step": 25292 }, { "epoch": 4.82, "learning_rate": 1.5561677788369878e-06, "loss": 2.4053, "step": 25293 }, { "epoch": 4.82, "learning_rate": 1.5561105815061964e-06, "loss": 2.0979, "step": 25294 }, { "epoch": 4.82, "learning_rate": 1.5560533841754052e-06, "loss": 2.5596, "step": 25295 }, { "epoch": 4.82, "learning_rate": 1.5559961868446142e-06, "loss": 1.6661, "step": 25296 }, { "epoch": 4.82, "learning_rate": 1.5559389895138228e-06, "loss": 1.96, "step": 25297 }, { "epoch": 4.82, "learning_rate": 1.5558817921830313e-06, "loss": 1.225, "step": 25298 }, { "epoch": 4.82, "learning_rate": 1.5558245948522404e-06, "loss": 3.6614, "step": 25299 }, { "epoch": 4.82, "learning_rate": 1.555767397521449e-06, "loss": 2.3449, "step": 25300 }, { "epoch": 4.82, "learning_rate": 1.5557102001906577e-06, "loss": 1.1298, "step": 25301 }, { "epoch": 4.82, "learning_rate": 1.5556530028598667e-06, "loss": 2.0891, "step": 25302 }, { "epoch": 4.82, "learning_rate": 1.5555958055290753e-06, "loss": 0.9821, "step": 25303 }, { "epoch": 4.82, "learning_rate": 1.555538608198284e-06, "loss": 2.1909, "step": 25304 }, { "epoch": 4.82, "learning_rate": 1.555481410867493e-06, "loss": 2.918, "step": 25305 }, { "epoch": 4.82, "learning_rate": 1.5554242135367017e-06, "loss": 2.6722, "step": 25306 }, { "epoch": 4.82, "learning_rate": 1.5553670162059103e-06, "loss": 2.3537, "step": 25307 }, { "epoch": 4.82, "learning_rate": 1.5553098188751193e-06, "loss": 1.7184, "step": 25308 }, { "epoch": 4.82, "learning_rate": 1.5552526215443279e-06, "loss": 1.4524, "step": 25309 }, { "epoch": 4.82, "learning_rate": 1.5551954242135369e-06, "loss": 2.3171, "step": 25310 }, { "epoch": 4.82, "learning_rate": 1.5551382268827455e-06, "loss": 1.9557, "step": 25311 }, { "epoch": 4.82, "learning_rate": 1.5550810295519542e-06, "loss": 3.0763, "step": 25312 }, { "epoch": 4.82, "learning_rate": 1.5550238322211633e-06, "loss": 2.1378, "step": 25313 }, { "epoch": 4.82, "learning_rate": 1.5549666348903718e-06, "loss": 2.5437, "step": 25314 }, { "epoch": 4.82, "learning_rate": 1.5549094375595804e-06, "loss": 2.4219, "step": 25315 }, { "epoch": 4.82, "learning_rate": 1.5548522402287894e-06, "loss": 1.6253, "step": 25316 }, { "epoch": 4.82, "learning_rate": 1.5547950428979982e-06, "loss": 1.1896, "step": 25317 }, { "epoch": 4.82, "learning_rate": 1.5547378455672068e-06, "loss": 1.126, "step": 25318 }, { "epoch": 4.82, "learning_rate": 1.5546806482364158e-06, "loss": 2.6937, "step": 25319 }, { "epoch": 4.82, "learning_rate": 1.5546234509056244e-06, "loss": 1.6679, "step": 25320 }, { "epoch": 4.82, "learning_rate": 1.554566253574833e-06, "loss": 2.4696, "step": 25321 }, { "epoch": 4.82, "learning_rate": 1.554509056244042e-06, "loss": 2.7916, "step": 25322 }, { "epoch": 4.82, "learning_rate": 1.5544518589132508e-06, "loss": 2.9911, "step": 25323 }, { "epoch": 4.82, "learning_rate": 1.5543946615824598e-06, "loss": 1.4773, "step": 25324 }, { "epoch": 4.82, "learning_rate": 1.5543374642516684e-06, "loss": 2.4912, "step": 25325 }, { "epoch": 4.82, "learning_rate": 1.554280266920877e-06, "loss": 2.4978, "step": 25326 }, { "epoch": 4.82, "learning_rate": 1.554223069590086e-06, "loss": 2.6703, "step": 25327 }, { "epoch": 4.82, "learning_rate": 1.5541658722592945e-06, "loss": 1.5527, "step": 25328 }, { "epoch": 4.82, "learning_rate": 1.5541086749285033e-06, "loss": 2.0206, "step": 25329 }, { "epoch": 4.82, "learning_rate": 1.5540514775977123e-06, "loss": 1.404, "step": 25330 }, { "epoch": 4.82, "learning_rate": 1.553994280266921e-06, "loss": 2.1958, "step": 25331 }, { "epoch": 4.83, "learning_rate": 1.5539370829361295e-06, "loss": 2.7195, "step": 25332 }, { "epoch": 4.83, "learning_rate": 1.5538798856053385e-06, "loss": 2.9547, "step": 25333 }, { "epoch": 4.83, "learning_rate": 1.5538226882745473e-06, "loss": 3.7677, "step": 25334 }, { "epoch": 4.83, "learning_rate": 1.5537654909437559e-06, "loss": 1.3288, "step": 25335 }, { "epoch": 4.83, "learning_rate": 1.5537082936129649e-06, "loss": 1.6714, "step": 25336 }, { "epoch": 4.83, "learning_rate": 1.5536510962821735e-06, "loss": 2.6088, "step": 25337 }, { "epoch": 4.83, "learning_rate": 1.5535938989513823e-06, "loss": 1.048, "step": 25338 }, { "epoch": 4.83, "learning_rate": 1.553536701620591e-06, "loss": 2.8022, "step": 25339 }, { "epoch": 4.83, "learning_rate": 1.5534795042897998e-06, "loss": 1.9561, "step": 25340 }, { "epoch": 4.83, "learning_rate": 1.5534223069590088e-06, "loss": 0.9041, "step": 25341 }, { "epoch": 4.83, "learning_rate": 1.5533651096282174e-06, "loss": 1.3601, "step": 25342 }, { "epoch": 4.83, "learning_rate": 1.553307912297426e-06, "loss": 1.4634, "step": 25343 }, { "epoch": 4.83, "learning_rate": 1.553250714966635e-06, "loss": 2.5774, "step": 25344 }, { "epoch": 4.83, "learning_rate": 1.5531935176358438e-06, "loss": 4.7203, "step": 25345 }, { "epoch": 4.83, "learning_rate": 1.5531363203050524e-06, "loss": 2.8607, "step": 25346 }, { "epoch": 4.83, "learning_rate": 1.5530791229742614e-06, "loss": 2.4255, "step": 25347 }, { "epoch": 4.83, "learning_rate": 1.55302192564347e-06, "loss": 1.5211, "step": 25348 }, { "epoch": 4.83, "learning_rate": 1.5529647283126786e-06, "loss": 2.0749, "step": 25349 }, { "epoch": 4.83, "learning_rate": 1.5529075309818876e-06, "loss": 1.4439, "step": 25350 }, { "epoch": 4.83, "learning_rate": 1.5528503336510964e-06, "loss": 3.737, "step": 25351 }, { "epoch": 4.83, "learning_rate": 1.552793136320305e-06, "loss": 2.7369, "step": 25352 }, { "epoch": 4.83, "learning_rate": 1.552735938989514e-06, "loss": 1.7871, "step": 25353 }, { "epoch": 4.83, "learning_rate": 1.5526787416587225e-06, "loss": 0.8531, "step": 25354 }, { "epoch": 4.83, "learning_rate": 1.5526215443279313e-06, "loss": 2.7315, "step": 25355 }, { "epoch": 4.83, "learning_rate": 1.5525643469971403e-06, "loss": 3.8654, "step": 25356 }, { "epoch": 4.83, "learning_rate": 1.552507149666349e-06, "loss": 1.7907, "step": 25357 }, { "epoch": 4.83, "learning_rate": 1.552449952335558e-06, "loss": 2.311, "step": 25358 }, { "epoch": 4.83, "learning_rate": 1.5523927550047665e-06, "loss": 1.5524, "step": 25359 }, { "epoch": 4.83, "learning_rate": 1.552335557673975e-06, "loss": 1.023, "step": 25360 }, { "epoch": 4.83, "learning_rate": 1.552278360343184e-06, "loss": 1.9432, "step": 25361 }, { "epoch": 4.83, "learning_rate": 1.5522211630123929e-06, "loss": 1.6647, "step": 25362 }, { "epoch": 4.83, "learning_rate": 1.5521639656816015e-06, "loss": 1.4361, "step": 25363 }, { "epoch": 4.83, "learning_rate": 1.5521067683508105e-06, "loss": 2.2799, "step": 25364 }, { "epoch": 4.83, "learning_rate": 1.552049571020019e-06, "loss": 2.24, "step": 25365 }, { "epoch": 4.83, "learning_rate": 1.5519923736892278e-06, "loss": 2.031, "step": 25366 }, { "epoch": 4.83, "learning_rate": 1.5519351763584366e-06, "loss": 1.5893, "step": 25367 }, { "epoch": 4.83, "learning_rate": 1.5518779790276454e-06, "loss": 1.2884, "step": 25368 }, { "epoch": 4.83, "learning_rate": 1.551820781696854e-06, "loss": 2.1444, "step": 25369 }, { "epoch": 4.83, "learning_rate": 1.551763584366063e-06, "loss": 2.7921, "step": 25370 }, { "epoch": 4.83, "learning_rate": 1.5517063870352716e-06, "loss": 2.2691, "step": 25371 }, { "epoch": 4.83, "learning_rate": 1.5516491897044804e-06, "loss": 1.0571, "step": 25372 }, { "epoch": 4.83, "learning_rate": 1.5515919923736894e-06, "loss": 1.413, "step": 25373 }, { "epoch": 4.83, "learning_rate": 1.551534795042898e-06, "loss": 2.9447, "step": 25374 }, { "epoch": 4.83, "learning_rate": 1.551477597712107e-06, "loss": 2.9298, "step": 25375 }, { "epoch": 4.83, "learning_rate": 1.5514204003813156e-06, "loss": 1.3808, "step": 25376 }, { "epoch": 4.83, "learning_rate": 1.5513632030505242e-06, "loss": 1.494, "step": 25377 }, { "epoch": 4.83, "learning_rate": 1.5513060057197332e-06, "loss": 1.4571, "step": 25378 }, { "epoch": 4.83, "learning_rate": 1.551248808388942e-06, "loss": 1.9436, "step": 25379 }, { "epoch": 4.83, "learning_rate": 1.5511916110581505e-06, "loss": 2.8559, "step": 25380 }, { "epoch": 4.83, "learning_rate": 1.5511344137273595e-06, "loss": 1.5678, "step": 25381 }, { "epoch": 4.83, "learning_rate": 1.5510772163965681e-06, "loss": 2.2509, "step": 25382 }, { "epoch": 4.83, "learning_rate": 1.551020019065777e-06, "loss": 2.8174, "step": 25383 }, { "epoch": 4.84, "learning_rate": 1.550962821734986e-06, "loss": 1.5938, "step": 25384 }, { "epoch": 4.84, "learning_rate": 1.5509056244041945e-06, "loss": 2.0271, "step": 25385 }, { "epoch": 4.84, "learning_rate": 1.550848427073403e-06, "loss": 1.2306, "step": 25386 }, { "epoch": 4.84, "learning_rate": 1.550791229742612e-06, "loss": 3.2564, "step": 25387 }, { "epoch": 4.84, "learning_rate": 1.5507340324118207e-06, "loss": 1.9438, "step": 25388 }, { "epoch": 4.84, "learning_rate": 1.5506768350810297e-06, "loss": 2.4139, "step": 25389 }, { "epoch": 4.84, "learning_rate": 1.5506196377502385e-06, "loss": 1.8454, "step": 25390 }, { "epoch": 4.84, "learning_rate": 1.550562440419447e-06, "loss": 1.7875, "step": 25391 }, { "epoch": 4.84, "learning_rate": 1.550505243088656e-06, "loss": 2.1412, "step": 25392 }, { "epoch": 4.84, "learning_rate": 1.5504480457578646e-06, "loss": 2.2177, "step": 25393 }, { "epoch": 4.84, "learning_rate": 1.5503908484270734e-06, "loss": 3.5137, "step": 25394 }, { "epoch": 4.84, "learning_rate": 1.5503336510962822e-06, "loss": 1.5674, "step": 25395 }, { "epoch": 4.84, "learning_rate": 1.550276453765491e-06, "loss": 2.0665, "step": 25396 }, { "epoch": 4.84, "learning_rate": 1.5502192564346996e-06, "loss": 2.6815, "step": 25397 }, { "epoch": 4.84, "learning_rate": 1.5501620591039086e-06, "loss": 2.9467, "step": 25398 }, { "epoch": 4.84, "learning_rate": 1.5501048617731172e-06, "loss": 0.9239, "step": 25399 }, { "epoch": 4.84, "learning_rate": 1.550047664442326e-06, "loss": 2.1685, "step": 25400 }, { "epoch": 4.84, "learning_rate": 1.549990467111535e-06, "loss": 3.7687, "step": 25401 }, { "epoch": 4.84, "learning_rate": 1.5499332697807436e-06, "loss": 1.6627, "step": 25402 }, { "epoch": 4.84, "learning_rate": 1.5498760724499522e-06, "loss": 2.1597, "step": 25403 }, { "epoch": 4.84, "learning_rate": 1.5498188751191612e-06, "loss": 2.686, "step": 25404 }, { "epoch": 4.84, "learning_rate": 1.54976167778837e-06, "loss": 2.401, "step": 25405 }, { "epoch": 4.84, "learning_rate": 1.5497044804575788e-06, "loss": 1.2373, "step": 25406 }, { "epoch": 4.84, "learning_rate": 1.5496472831267875e-06, "loss": 2.0678, "step": 25407 }, { "epoch": 4.84, "learning_rate": 1.5495900857959961e-06, "loss": 2.1411, "step": 25408 }, { "epoch": 4.84, "learning_rate": 1.5495328884652051e-06, "loss": 1.8315, "step": 25409 }, { "epoch": 4.84, "learning_rate": 1.5494756911344137e-06, "loss": 2.5909, "step": 25410 }, { "epoch": 4.84, "learning_rate": 1.5494184938036225e-06, "loss": 1.2665, "step": 25411 }, { "epoch": 4.84, "learning_rate": 1.5493612964728315e-06, "loss": 2.727, "step": 25412 }, { "epoch": 4.84, "learning_rate": 1.54930409914204e-06, "loss": 1.4364, "step": 25413 }, { "epoch": 4.84, "learning_rate": 1.5492469018112487e-06, "loss": 2.724, "step": 25414 }, { "epoch": 4.84, "learning_rate": 1.5491897044804577e-06, "loss": 1.8048, "step": 25415 }, { "epoch": 4.84, "learning_rate": 1.5491325071496663e-06, "loss": 2.566, "step": 25416 }, { "epoch": 4.84, "learning_rate": 1.549075309818875e-06, "loss": 1.0746, "step": 25417 }, { "epoch": 4.84, "learning_rate": 1.549018112488084e-06, "loss": 1.4655, "step": 25418 }, { "epoch": 4.84, "learning_rate": 1.5489609151572926e-06, "loss": 1.8286, "step": 25419 }, { "epoch": 4.84, "learning_rate": 1.5489037178265012e-06, "loss": 1.7251, "step": 25420 }, { "epoch": 4.84, "learning_rate": 1.5488465204957102e-06, "loss": 1.0897, "step": 25421 }, { "epoch": 4.84, "learning_rate": 1.548789323164919e-06, "loss": 1.9899, "step": 25422 }, { "epoch": 4.84, "learning_rate": 1.548732125834128e-06, "loss": 1.6262, "step": 25423 }, { "epoch": 4.84, "learning_rate": 1.5486749285033366e-06, "loss": 1.9239, "step": 25424 }, { "epoch": 4.84, "learning_rate": 1.5486177311725452e-06, "loss": 1.832, "step": 25425 }, { "epoch": 4.84, "learning_rate": 1.5485605338417542e-06, "loss": 1.6121, "step": 25426 }, { "epoch": 4.84, "learning_rate": 1.5485033365109628e-06, "loss": 1.4683, "step": 25427 }, { "epoch": 4.84, "learning_rate": 1.5484461391801716e-06, "loss": 1.9734, "step": 25428 }, { "epoch": 4.84, "learning_rate": 1.5483889418493806e-06, "loss": 2.7852, "step": 25429 }, { "epoch": 4.84, "learning_rate": 1.5483317445185892e-06, "loss": 1.7649, "step": 25430 }, { "epoch": 4.84, "learning_rate": 1.5482745471877977e-06, "loss": 1.1472, "step": 25431 }, { "epoch": 4.84, "learning_rate": 1.5482173498570068e-06, "loss": 2.3917, "step": 25432 }, { "epoch": 4.84, "learning_rate": 1.5481601525262155e-06, "loss": 1.1216, "step": 25433 }, { "epoch": 4.84, "learning_rate": 1.5481029551954241e-06, "loss": 2.1429, "step": 25434 }, { "epoch": 4.84, "learning_rate": 1.5480457578646331e-06, "loss": 2.4069, "step": 25435 }, { "epoch": 4.84, "learning_rate": 1.5479885605338417e-06, "loss": 1.6249, "step": 25436 }, { "epoch": 4.85, "learning_rate": 1.5479313632030503e-06, "loss": 2.443, "step": 25437 }, { "epoch": 4.85, "learning_rate": 1.5478741658722593e-06, "loss": 2.071, "step": 25438 }, { "epoch": 4.85, "learning_rate": 1.547816968541468e-06, "loss": 2.0685, "step": 25439 }, { "epoch": 4.85, "learning_rate": 1.547759771210677e-06, "loss": 2.3037, "step": 25440 }, { "epoch": 4.85, "learning_rate": 1.5477025738798857e-06, "loss": 3.0207, "step": 25441 }, { "epoch": 4.85, "learning_rate": 1.5476453765490943e-06, "loss": 1.127, "step": 25442 }, { "epoch": 4.85, "learning_rate": 1.5475881792183033e-06, "loss": 1.5757, "step": 25443 }, { "epoch": 4.85, "learning_rate": 1.547530981887512e-06, "loss": 3.335, "step": 25444 }, { "epoch": 4.85, "learning_rate": 1.5474737845567207e-06, "loss": 2.1751, "step": 25445 }, { "epoch": 4.85, "learning_rate": 1.5474165872259297e-06, "loss": 2.1227, "step": 25446 }, { "epoch": 4.85, "learning_rate": 1.5473593898951382e-06, "loss": 1.853, "step": 25447 }, { "epoch": 4.85, "learning_rate": 1.5473021925643468e-06, "loss": 2.8395, "step": 25448 }, { "epoch": 4.85, "learning_rate": 1.5472449952335558e-06, "loss": 1.8756, "step": 25449 }, { "epoch": 4.85, "learning_rate": 1.5471877979027646e-06, "loss": 3.2412, "step": 25450 }, { "epoch": 4.85, "learning_rate": 1.5471306005719732e-06, "loss": 1.6136, "step": 25451 }, { "epoch": 4.85, "learning_rate": 1.5470734032411822e-06, "loss": 3.3835, "step": 25452 }, { "epoch": 4.85, "learning_rate": 1.5470162059103908e-06, "loss": 2.9732, "step": 25453 }, { "epoch": 4.85, "learning_rate": 1.5469590085795998e-06, "loss": 2.2485, "step": 25454 }, { "epoch": 4.85, "learning_rate": 1.5469018112488084e-06, "loss": 2.1978, "step": 25455 }, { "epoch": 4.85, "learning_rate": 1.5468446139180172e-06, "loss": 1.8919, "step": 25456 }, { "epoch": 4.85, "learning_rate": 1.5467874165872262e-06, "loss": 1.7211, "step": 25457 }, { "epoch": 4.85, "learning_rate": 1.5467302192564348e-06, "loss": 1.4613, "step": 25458 }, { "epoch": 4.85, "learning_rate": 1.5466730219256433e-06, "loss": 2.0421, "step": 25459 }, { "epoch": 4.85, "learning_rate": 1.5466158245948523e-06, "loss": 2.6384, "step": 25460 }, { "epoch": 4.85, "learning_rate": 1.5465586272640611e-06, "loss": 2.6111, "step": 25461 }, { "epoch": 4.85, "learning_rate": 1.5465014299332697e-06, "loss": 0.3952, "step": 25462 }, { "epoch": 4.85, "learning_rate": 1.5464442326024787e-06, "loss": 2.3298, "step": 25463 }, { "epoch": 4.85, "learning_rate": 1.5463870352716873e-06, "loss": 1.9784, "step": 25464 }, { "epoch": 4.85, "learning_rate": 1.5463298379408959e-06, "loss": 2.2477, "step": 25465 }, { "epoch": 4.85, "learning_rate": 1.546272640610105e-06, "loss": 2.0836, "step": 25466 }, { "epoch": 4.85, "learning_rate": 1.5462154432793137e-06, "loss": 1.2275, "step": 25467 }, { "epoch": 4.85, "learning_rate": 1.5461582459485223e-06, "loss": 1.4929, "step": 25468 }, { "epoch": 4.85, "learning_rate": 1.5461010486177313e-06, "loss": 0.5399, "step": 25469 }, { "epoch": 4.85, "learning_rate": 1.5460438512869399e-06, "loss": 1.3781, "step": 25470 }, { "epoch": 4.85, "learning_rate": 1.5459866539561489e-06, "loss": 1.4805, "step": 25471 }, { "epoch": 4.85, "learning_rate": 1.5459294566253577e-06, "loss": 2.0211, "step": 25472 }, { "epoch": 4.85, "learning_rate": 1.5458722592945662e-06, "loss": 2.6853, "step": 25473 }, { "epoch": 4.85, "learning_rate": 1.5458150619637752e-06, "loss": 2.3486, "step": 25474 }, { "epoch": 4.85, "learning_rate": 1.5457578646329838e-06, "loss": 2.5136, "step": 25475 }, { "epoch": 4.85, "learning_rate": 1.5457006673021924e-06, "loss": 1.4053, "step": 25476 }, { "epoch": 4.85, "learning_rate": 1.5456434699714014e-06, "loss": 1.7776, "step": 25477 }, { "epoch": 4.85, "learning_rate": 1.5455862726406102e-06, "loss": 2.1785, "step": 25478 }, { "epoch": 4.85, "learning_rate": 1.5455290753098188e-06, "loss": 1.1964, "step": 25479 }, { "epoch": 4.85, "learning_rate": 1.5454718779790278e-06, "loss": 2.8263, "step": 25480 }, { "epoch": 4.85, "learning_rate": 1.5454146806482364e-06, "loss": 1.6437, "step": 25481 }, { "epoch": 4.85, "learning_rate": 1.5453574833174452e-06, "loss": 1.4599, "step": 25482 }, { "epoch": 4.85, "learning_rate": 1.545300285986654e-06, "loss": 1.3662, "step": 25483 }, { "epoch": 4.85, "learning_rate": 1.5452430886558628e-06, "loss": 1.8487, "step": 25484 }, { "epoch": 4.85, "learning_rate": 1.5451858913250713e-06, "loss": 2.739, "step": 25485 }, { "epoch": 4.85, "learning_rate": 1.5451286939942804e-06, "loss": 1.6004, "step": 25486 }, { "epoch": 4.85, "learning_rate": 1.545071496663489e-06, "loss": 2.3752, "step": 25487 }, { "epoch": 4.85, "learning_rate": 1.545014299332698e-06, "loss": 2.4381, "step": 25488 }, { "epoch": 4.86, "learning_rate": 1.5449571020019067e-06, "loss": 1.76, "step": 25489 }, { "epoch": 4.86, "learning_rate": 1.5448999046711153e-06, "loss": 2.3834, "step": 25490 }, { "epoch": 4.86, "learning_rate": 1.5448427073403243e-06, "loss": 1.5951, "step": 25491 }, { "epoch": 4.86, "learning_rate": 1.544785510009533e-06, "loss": 1.4722, "step": 25492 }, { "epoch": 4.86, "learning_rate": 1.5447283126787417e-06, "loss": 1.9981, "step": 25493 }, { "epoch": 4.86, "learning_rate": 1.5446711153479505e-06, "loss": 1.385, "step": 25494 }, { "epoch": 4.86, "learning_rate": 1.5446139180171593e-06, "loss": 1.6735, "step": 25495 }, { "epoch": 4.86, "learning_rate": 1.5445567206863679e-06, "loss": 1.5138, "step": 25496 }, { "epoch": 4.86, "learning_rate": 1.5444995233555769e-06, "loss": 1.5838, "step": 25497 }, { "epoch": 4.86, "learning_rate": 1.5444423260247855e-06, "loss": 2.3628, "step": 25498 }, { "epoch": 4.86, "learning_rate": 1.5443851286939942e-06, "loss": 0.8202, "step": 25499 }, { "epoch": 4.86, "learning_rate": 1.5443279313632033e-06, "loss": 1.8677, "step": 25500 }, { "epoch": 4.86, "learning_rate": 1.5442707340324118e-06, "loss": 1.9086, "step": 25501 }, { "epoch": 4.86, "learning_rate": 1.5442135367016204e-06, "loss": 1.4854, "step": 25502 }, { "epoch": 4.86, "learning_rate": 1.5441563393708294e-06, "loss": 1.8024, "step": 25503 }, { "epoch": 4.86, "learning_rate": 1.544099142040038e-06, "loss": 2.8206, "step": 25504 }, { "epoch": 4.86, "learning_rate": 1.544041944709247e-06, "loss": 1.8953, "step": 25505 }, { "epoch": 4.86, "learning_rate": 1.5439847473784558e-06, "loss": 1.9911, "step": 25506 }, { "epoch": 4.86, "learning_rate": 1.5439275500476644e-06, "loss": 1.7849, "step": 25507 }, { "epoch": 4.86, "learning_rate": 1.5438703527168734e-06, "loss": 1.0101, "step": 25508 }, { "epoch": 4.86, "learning_rate": 1.543813155386082e-06, "loss": 1.4745, "step": 25509 }, { "epoch": 4.86, "learning_rate": 1.5437559580552908e-06, "loss": 1.8557, "step": 25510 }, { "epoch": 4.86, "learning_rate": 1.5436987607244998e-06, "loss": 1.7908, "step": 25511 }, { "epoch": 4.86, "learning_rate": 1.5436415633937084e-06, "loss": 1.8093, "step": 25512 }, { "epoch": 4.86, "learning_rate": 1.543584366062917e-06, "loss": 0.7933, "step": 25513 }, { "epoch": 4.86, "learning_rate": 1.543527168732126e-06, "loss": 1.476, "step": 25514 }, { "epoch": 4.86, "learning_rate": 1.5434699714013345e-06, "loss": 1.8214, "step": 25515 }, { "epoch": 4.86, "learning_rate": 1.5434127740705433e-06, "loss": 1.8551, "step": 25516 }, { "epoch": 4.86, "learning_rate": 1.5433555767397523e-06, "loss": 2.7969, "step": 25517 }, { "epoch": 4.86, "learning_rate": 1.543298379408961e-06, "loss": 0.9667, "step": 25518 }, { "epoch": 4.86, "learning_rate": 1.54324118207817e-06, "loss": 1.7853, "step": 25519 }, { "epoch": 4.86, "learning_rate": 1.5431839847473785e-06, "loss": 2.3614, "step": 25520 }, { "epoch": 4.86, "learning_rate": 1.5431267874165873e-06, "loss": 3.2506, "step": 25521 }, { "epoch": 4.86, "learning_rate": 1.543069590085796e-06, "loss": 2.2512, "step": 25522 }, { "epoch": 4.86, "learning_rate": 1.5430123927550049e-06, "loss": 1.5486, "step": 25523 }, { "epoch": 4.86, "learning_rate": 1.5429551954242135e-06, "loss": 1.529, "step": 25524 }, { "epoch": 4.86, "learning_rate": 1.5428979980934225e-06, "loss": 1.4781, "step": 25525 }, { "epoch": 4.86, "learning_rate": 1.542840800762631e-06, "loss": 2.3145, "step": 25526 }, { "epoch": 4.86, "learning_rate": 1.5427836034318398e-06, "loss": 2.4143, "step": 25527 }, { "epoch": 4.86, "learning_rate": 1.5427264061010488e-06, "loss": 1.9916, "step": 25528 }, { "epoch": 4.86, "learning_rate": 1.5426692087702574e-06, "loss": 3.127, "step": 25529 }, { "epoch": 4.86, "learning_rate": 1.542612011439466e-06, "loss": 2.0972, "step": 25530 }, { "epoch": 4.86, "learning_rate": 1.542554814108675e-06, "loss": 3.1892, "step": 25531 }, { "epoch": 4.86, "learning_rate": 1.5424976167778836e-06, "loss": 2.1636, "step": 25532 }, { "epoch": 4.86, "learning_rate": 1.5424404194470924e-06, "loss": 1.7679, "step": 25533 }, { "epoch": 4.86, "learning_rate": 1.5423832221163014e-06, "loss": 2.7151, "step": 25534 }, { "epoch": 4.86, "learning_rate": 1.54232602478551e-06, "loss": 1.9546, "step": 25535 }, { "epoch": 4.86, "learning_rate": 1.542268827454719e-06, "loss": 2.4376, "step": 25536 }, { "epoch": 4.86, "learning_rate": 1.5422116301239276e-06, "loss": 2.507, "step": 25537 }, { "epoch": 4.86, "learning_rate": 1.5421544327931364e-06, "loss": 2.0573, "step": 25538 }, { "epoch": 4.86, "learning_rate": 1.5420972354623454e-06, "loss": 2.0353, "step": 25539 }, { "epoch": 4.86, "learning_rate": 1.542040038131554e-06, "loss": 2.5717, "step": 25540 }, { "epoch": 4.86, "learning_rate": 1.5419828408007625e-06, "loss": 1.8204, "step": 25541 }, { "epoch": 4.87, "learning_rate": 1.5419256434699715e-06, "loss": 1.5209, "step": 25542 }, { "epoch": 4.87, "learning_rate": 1.5418684461391801e-06, "loss": 1.885, "step": 25543 }, { "epoch": 4.87, "learning_rate": 1.541811248808389e-06, "loss": 3.0741, "step": 25544 }, { "epoch": 4.87, "learning_rate": 1.541754051477598e-06, "loss": 1.365, "step": 25545 }, { "epoch": 4.87, "learning_rate": 1.5416968541468065e-06, "loss": 2.9201, "step": 25546 }, { "epoch": 4.87, "learning_rate": 1.541639656816015e-06, "loss": 1.5026, "step": 25547 }, { "epoch": 4.87, "learning_rate": 1.541582459485224e-06, "loss": 2.139, "step": 25548 }, { "epoch": 4.87, "learning_rate": 1.5415252621544329e-06, "loss": 1.1515, "step": 25549 }, { "epoch": 4.87, "learning_rate": 1.5414680648236415e-06, "loss": 1.3465, "step": 25550 }, { "epoch": 4.87, "learning_rate": 1.5414108674928505e-06, "loss": 2.4156, "step": 25551 }, { "epoch": 4.87, "learning_rate": 1.541353670162059e-06, "loss": 2.9099, "step": 25552 }, { "epoch": 4.87, "learning_rate": 1.541296472831268e-06, "loss": 1.9139, "step": 25553 }, { "epoch": 4.87, "learning_rate": 1.5412392755004766e-06, "loss": 1.5391, "step": 25554 }, { "epoch": 4.87, "learning_rate": 1.5411820781696854e-06, "loss": 1.821, "step": 25555 }, { "epoch": 4.87, "learning_rate": 1.5411248808388944e-06, "loss": 3.4473, "step": 25556 }, { "epoch": 4.87, "learning_rate": 1.541067683508103e-06, "loss": 1.9712, "step": 25557 }, { "epoch": 4.87, "learning_rate": 1.5410104861773116e-06, "loss": 1.9037, "step": 25558 }, { "epoch": 4.87, "learning_rate": 1.5409532888465206e-06, "loss": 2.217, "step": 25559 }, { "epoch": 4.87, "learning_rate": 1.5408960915157294e-06, "loss": 2.6855, "step": 25560 }, { "epoch": 4.87, "learning_rate": 1.540838894184938e-06, "loss": 1.9728, "step": 25561 }, { "epoch": 4.87, "learning_rate": 1.540781696854147e-06, "loss": 2.4375, "step": 25562 }, { "epoch": 4.87, "learning_rate": 1.5407244995233556e-06, "loss": 1.6745, "step": 25563 }, { "epoch": 4.87, "learning_rate": 1.5406673021925642e-06, "loss": 2.0315, "step": 25564 }, { "epoch": 4.87, "learning_rate": 1.5406101048617732e-06, "loss": 1.8461, "step": 25565 }, { "epoch": 4.87, "learning_rate": 1.540552907530982e-06, "loss": 2.2342, "step": 25566 }, { "epoch": 4.87, "learning_rate": 1.5404957102001905e-06, "loss": 2.6362, "step": 25567 }, { "epoch": 4.87, "learning_rate": 1.5404385128693995e-06, "loss": 1.891, "step": 25568 }, { "epoch": 4.87, "learning_rate": 1.5403813155386081e-06, "loss": 2.4284, "step": 25569 }, { "epoch": 4.87, "learning_rate": 1.5403241182078171e-06, "loss": 2.2666, "step": 25570 }, { "epoch": 4.87, "learning_rate": 1.5402669208770257e-06, "loss": 2.1405, "step": 25571 }, { "epoch": 4.87, "learning_rate": 1.5402097235462345e-06, "loss": 1.1624, "step": 25572 }, { "epoch": 4.87, "learning_rate": 1.5401525262154435e-06, "loss": 1.9606, "step": 25573 }, { "epoch": 4.87, "learning_rate": 1.540095328884652e-06, "loss": 2.0591, "step": 25574 }, { "epoch": 4.87, "learning_rate": 1.5400381315538607e-06, "loss": 2.6913, "step": 25575 }, { "epoch": 4.87, "learning_rate": 1.5399809342230697e-06, "loss": 1.8, "step": 25576 }, { "epoch": 4.87, "learning_rate": 1.5399237368922785e-06, "loss": 2.9578, "step": 25577 }, { "epoch": 4.87, "learning_rate": 1.539866539561487e-06, "loss": 2.3115, "step": 25578 }, { "epoch": 4.87, "learning_rate": 1.539809342230696e-06, "loss": 2.4774, "step": 25579 }, { "epoch": 4.87, "learning_rate": 1.5397521448999046e-06, "loss": 1.1914, "step": 25580 }, { "epoch": 4.87, "learning_rate": 1.5396949475691134e-06, "loss": 1.6798, "step": 25581 }, { "epoch": 4.87, "learning_rate": 1.5396377502383222e-06, "loss": 2.0032, "step": 25582 }, { "epoch": 4.87, "learning_rate": 1.539580552907531e-06, "loss": 1.2539, "step": 25583 }, { "epoch": 4.87, "learning_rate": 1.53952335557674e-06, "loss": 2.1513, "step": 25584 }, { "epoch": 4.87, "learning_rate": 1.5394661582459486e-06, "loss": 1.9924, "step": 25585 }, { "epoch": 4.87, "learning_rate": 1.5394089609151572e-06, "loss": 1.2343, "step": 25586 }, { "epoch": 4.87, "learning_rate": 1.5393517635843662e-06, "loss": 2.2805, "step": 25587 }, { "epoch": 4.87, "learning_rate": 1.539294566253575e-06, "loss": 1.4663, "step": 25588 }, { "epoch": 4.87, "learning_rate": 1.5392373689227836e-06, "loss": 2.4118, "step": 25589 }, { "epoch": 4.87, "learning_rate": 1.5391801715919926e-06, "loss": 2.5618, "step": 25590 }, { "epoch": 4.87, "learning_rate": 1.5391229742612012e-06, "loss": 2.2074, "step": 25591 }, { "epoch": 4.87, "learning_rate": 1.5390657769304097e-06, "loss": 1.5681, "step": 25592 }, { "epoch": 4.87, "learning_rate": 1.5390085795996187e-06, "loss": 2.0999, "step": 25593 }, { "epoch": 4.88, "learning_rate": 1.5389513822688275e-06, "loss": 2.0922, "step": 25594 }, { "epoch": 4.88, "learning_rate": 1.5388941849380361e-06, "loss": 1.9341, "step": 25595 }, { "epoch": 4.88, "learning_rate": 1.5388369876072451e-06, "loss": 1.6323, "step": 25596 }, { "epoch": 4.88, "learning_rate": 1.5387797902764537e-06, "loss": 1.6407, "step": 25597 }, { "epoch": 4.88, "learning_rate": 1.5387225929456625e-06, "loss": 2.2066, "step": 25598 }, { "epoch": 4.88, "learning_rate": 1.5386653956148715e-06, "loss": 3.6113, "step": 25599 }, { "epoch": 4.88, "learning_rate": 1.53860819828408e-06, "loss": 2.2482, "step": 25600 }, { "epoch": 4.88, "learning_rate": 1.538551000953289e-06, "loss": 3.1216, "step": 25601 }, { "epoch": 4.88, "learning_rate": 1.5384938036224977e-06, "loss": 2.1414, "step": 25602 }, { "epoch": 4.88, "learning_rate": 1.5384366062917063e-06, "loss": 2.6294, "step": 25603 }, { "epoch": 4.88, "learning_rate": 1.5383794089609153e-06, "loss": 2.8411, "step": 25604 }, { "epoch": 4.88, "learning_rate": 1.538322211630124e-06, "loss": 1.5892, "step": 25605 }, { "epoch": 4.88, "learning_rate": 1.5382650142993326e-06, "loss": 2.6281, "step": 25606 }, { "epoch": 4.88, "learning_rate": 1.5382078169685416e-06, "loss": 2.2588, "step": 25607 }, { "epoch": 4.88, "learning_rate": 1.5381506196377502e-06, "loss": 2.5482, "step": 25608 }, { "epoch": 4.88, "learning_rate": 1.538093422306959e-06, "loss": 2.624, "step": 25609 }, { "epoch": 4.88, "learning_rate": 1.5380362249761678e-06, "loss": 2.36, "step": 25610 }, { "epoch": 4.88, "learning_rate": 1.5379790276453766e-06, "loss": 0.8871, "step": 25611 }, { "epoch": 4.88, "learning_rate": 1.5379218303145852e-06, "loss": 2.923, "step": 25612 }, { "epoch": 4.88, "learning_rate": 1.5378646329837942e-06, "loss": 2.4323, "step": 25613 }, { "epoch": 4.88, "learning_rate": 1.5378074356530028e-06, "loss": 2.253, "step": 25614 }, { "epoch": 4.88, "learning_rate": 1.5377502383222116e-06, "loss": 1.7006, "step": 25615 }, { "epoch": 4.88, "learning_rate": 1.5376930409914206e-06, "loss": 2.4233, "step": 25616 }, { "epoch": 4.88, "learning_rate": 1.5376358436606292e-06, "loss": 2.8266, "step": 25617 }, { "epoch": 4.88, "learning_rate": 1.5375786463298382e-06, "loss": 1.5594, "step": 25618 }, { "epoch": 4.88, "learning_rate": 1.5375214489990468e-06, "loss": 1.6558, "step": 25619 }, { "epoch": 4.88, "learning_rate": 1.5374642516682553e-06, "loss": 1.2623, "step": 25620 }, { "epoch": 4.88, "learning_rate": 1.5374070543374643e-06, "loss": 2.6737, "step": 25621 }, { "epoch": 4.88, "learning_rate": 1.5373498570066731e-06, "loss": 1.971, "step": 25622 }, { "epoch": 4.88, "learning_rate": 1.5372926596758817e-06, "loss": 1.0545, "step": 25623 }, { "epoch": 4.88, "learning_rate": 1.5372354623450907e-06, "loss": 2.2711, "step": 25624 }, { "epoch": 4.88, "learning_rate": 1.5371782650142993e-06, "loss": 2.6989, "step": 25625 }, { "epoch": 4.88, "learning_rate": 1.537121067683508e-06, "loss": 1.7664, "step": 25626 }, { "epoch": 4.88, "learning_rate": 1.537063870352717e-06, "loss": 2.2298, "step": 25627 }, { "epoch": 4.88, "learning_rate": 1.5370066730219257e-06, "loss": 3.1229, "step": 25628 }, { "epoch": 4.88, "learning_rate": 1.5369494756911343e-06, "loss": 0.7165, "step": 25629 }, { "epoch": 4.88, "learning_rate": 1.5368922783603433e-06, "loss": 1.9755, "step": 25630 }, { "epoch": 4.88, "learning_rate": 1.5368350810295519e-06, "loss": 1.1695, "step": 25631 }, { "epoch": 4.88, "learning_rate": 1.5367778836987606e-06, "loss": 1.6242, "step": 25632 }, { "epoch": 4.88, "learning_rate": 1.5367206863679697e-06, "loss": 2.3204, "step": 25633 }, { "epoch": 4.88, "learning_rate": 1.5366634890371782e-06, "loss": 3.7346, "step": 25634 }, { "epoch": 4.88, "learning_rate": 1.5366062917063872e-06, "loss": 1.8549, "step": 25635 }, { "epoch": 4.88, "learning_rate": 1.5365490943755958e-06, "loss": 2.423, "step": 25636 }, { "epoch": 4.88, "learning_rate": 1.5364918970448046e-06, "loss": 1.3117, "step": 25637 }, { "epoch": 4.88, "learning_rate": 1.5364346997140134e-06, "loss": 1.2863, "step": 25638 }, { "epoch": 4.88, "learning_rate": 1.5363775023832222e-06, "loss": 2.5617, "step": 25639 }, { "epoch": 4.88, "learning_rate": 1.5363203050524308e-06, "loss": 1.2246, "step": 25640 }, { "epoch": 4.88, "learning_rate": 1.5362631077216398e-06, "loss": 1.4767, "step": 25641 }, { "epoch": 4.88, "learning_rate": 1.5362059103908484e-06, "loss": 2.9327, "step": 25642 }, { "epoch": 4.88, "learning_rate": 1.5361487130600572e-06, "loss": 2.7451, "step": 25643 }, { "epoch": 4.88, "learning_rate": 1.5360915157292662e-06, "loss": 2.9102, "step": 25644 }, { "epoch": 4.88, "learning_rate": 1.5360343183984748e-06, "loss": 2.6559, "step": 25645 }, { "epoch": 4.88, "learning_rate": 1.5359771210676833e-06, "loss": 1.9195, "step": 25646 }, { "epoch": 4.89, "learning_rate": 1.5359199237368923e-06, "loss": 2.7206, "step": 25647 }, { "epoch": 4.89, "learning_rate": 1.5358627264061011e-06, "loss": 1.1729, "step": 25648 }, { "epoch": 4.89, "learning_rate": 1.53580552907531e-06, "loss": 1.2379, "step": 25649 }, { "epoch": 4.89, "learning_rate": 1.5357483317445187e-06, "loss": 3.4844, "step": 25650 }, { "epoch": 4.89, "learning_rate": 1.5356911344137273e-06, "loss": 2.5737, "step": 25651 }, { "epoch": 4.89, "learning_rate": 1.5356339370829363e-06, "loss": 1.7386, "step": 25652 }, { "epoch": 4.89, "learning_rate": 1.535576739752145e-06, "loss": 1.2661, "step": 25653 }, { "epoch": 4.89, "learning_rate": 1.5355195424213537e-06, "loss": 1.3895, "step": 25654 }, { "epoch": 4.89, "learning_rate": 1.5354623450905627e-06, "loss": 2.4996, "step": 25655 }, { "epoch": 4.89, "learning_rate": 1.5354051477597713e-06, "loss": 1.9202, "step": 25656 }, { "epoch": 4.89, "learning_rate": 1.5353479504289799e-06, "loss": 1.6906, "step": 25657 }, { "epoch": 4.89, "learning_rate": 1.5352907530981889e-06, "loss": 2.4717, "step": 25658 }, { "epoch": 4.89, "learning_rate": 1.5352335557673974e-06, "loss": 1.6456, "step": 25659 }, { "epoch": 4.89, "learning_rate": 1.5351763584366062e-06, "loss": 1.3006, "step": 25660 }, { "epoch": 4.89, "learning_rate": 1.5351191611058152e-06, "loss": 1.9674, "step": 25661 }, { "epoch": 4.89, "learning_rate": 1.5350619637750238e-06, "loss": 2.8472, "step": 25662 }, { "epoch": 4.89, "learning_rate": 1.5350047664442324e-06, "loss": 1.5899, "step": 25663 }, { "epoch": 4.89, "learning_rate": 1.5349475691134414e-06, "loss": 2.6208, "step": 25664 }, { "epoch": 4.89, "learning_rate": 1.5348903717826502e-06, "loss": 2.1088, "step": 25665 }, { "epoch": 4.89, "learning_rate": 1.5348331744518592e-06, "loss": 1.3038, "step": 25666 }, { "epoch": 4.89, "learning_rate": 1.5347759771210678e-06, "loss": 2.5946, "step": 25667 }, { "epoch": 4.89, "learning_rate": 1.5347187797902764e-06, "loss": 2.1302, "step": 25668 }, { "epoch": 4.89, "learning_rate": 1.5346615824594854e-06, "loss": 2.0633, "step": 25669 }, { "epoch": 4.89, "learning_rate": 1.534604385128694e-06, "loss": 2.2374, "step": 25670 }, { "epoch": 4.89, "learning_rate": 1.5345471877979028e-06, "loss": 3.3048, "step": 25671 }, { "epoch": 4.89, "learning_rate": 1.5344899904671118e-06, "loss": 1.4845, "step": 25672 }, { "epoch": 4.89, "learning_rate": 1.5344327931363203e-06, "loss": 2.2137, "step": 25673 }, { "epoch": 4.89, "learning_rate": 1.534375595805529e-06, "loss": 1.974, "step": 25674 }, { "epoch": 4.89, "learning_rate": 1.534318398474738e-06, "loss": 1.2029, "step": 25675 }, { "epoch": 4.89, "learning_rate": 1.5342612011439467e-06, "loss": 1.3508, "step": 25676 }, { "epoch": 4.89, "learning_rate": 1.5342040038131553e-06, "loss": 0.9675, "step": 25677 }, { "epoch": 4.89, "learning_rate": 1.5341468064823643e-06, "loss": 1.67, "step": 25678 }, { "epoch": 4.89, "learning_rate": 1.534089609151573e-06, "loss": 1.9474, "step": 25679 }, { "epoch": 4.89, "learning_rate": 1.5340324118207815e-06, "loss": 1.8301, "step": 25680 }, { "epoch": 4.89, "learning_rate": 1.5339752144899905e-06, "loss": 2.478, "step": 25681 }, { "epoch": 4.89, "learning_rate": 1.5339180171591993e-06, "loss": 1.7351, "step": 25682 }, { "epoch": 4.89, "learning_rate": 1.5338608198284083e-06, "loss": 0.8189, "step": 25683 }, { "epoch": 4.89, "learning_rate": 1.5338036224976169e-06, "loss": 2.6062, "step": 25684 }, { "epoch": 4.89, "learning_rate": 1.5337464251668254e-06, "loss": 1.7771, "step": 25685 }, { "epoch": 4.89, "learning_rate": 1.5336892278360345e-06, "loss": 1.0546, "step": 25686 }, { "epoch": 4.89, "learning_rate": 1.5336320305052432e-06, "loss": 1.3669, "step": 25687 }, { "epoch": 4.89, "learning_rate": 1.5335748331744518e-06, "loss": 2.6218, "step": 25688 }, { "epoch": 4.89, "learning_rate": 1.5335176358436608e-06, "loss": 2.9553, "step": 25689 }, { "epoch": 4.89, "learning_rate": 1.5334604385128694e-06, "loss": 2.8036, "step": 25690 }, { "epoch": 4.89, "learning_rate": 1.533403241182078e-06, "loss": 1.3276, "step": 25691 }, { "epoch": 4.89, "learning_rate": 1.533346043851287e-06, "loss": 1.714, "step": 25692 }, { "epoch": 4.89, "learning_rate": 1.5332888465204958e-06, "loss": 1.9097, "step": 25693 }, { "epoch": 4.89, "learning_rate": 1.5332316491897044e-06, "loss": 1.2619, "step": 25694 }, { "epoch": 4.89, "learning_rate": 1.5331744518589134e-06, "loss": 2.3576, "step": 25695 }, { "epoch": 4.89, "learning_rate": 1.533117254528122e-06, "loss": 1.824, "step": 25696 }, { "epoch": 4.89, "learning_rate": 1.5330600571973308e-06, "loss": 1.8756, "step": 25697 }, { "epoch": 4.89, "learning_rate": 1.5330028598665396e-06, "loss": 2.2543, "step": 25698 }, { "epoch": 4.9, "learning_rate": 1.5329456625357484e-06, "loss": 0.7003, "step": 25699 }, { "epoch": 4.9, "learning_rate": 1.5328884652049574e-06, "loss": 2.557, "step": 25700 }, { "epoch": 4.9, "learning_rate": 1.532831267874166e-06, "loss": 0.7513, "step": 25701 }, { "epoch": 4.9, "learning_rate": 1.5327740705433745e-06, "loss": 2.4984, "step": 25702 }, { "epoch": 4.9, "learning_rate": 1.5327168732125835e-06, "loss": 2.5088, "step": 25703 }, { "epoch": 4.9, "learning_rate": 1.5326596758817923e-06, "loss": 2.1028, "step": 25704 }, { "epoch": 4.9, "learning_rate": 1.532602478551001e-06, "loss": 1.8203, "step": 25705 }, { "epoch": 4.9, "learning_rate": 1.53254528122021e-06, "loss": 2.1803, "step": 25706 }, { "epoch": 4.9, "learning_rate": 1.5324880838894185e-06, "loss": 1.8588, "step": 25707 }, { "epoch": 4.9, "learning_rate": 1.532430886558627e-06, "loss": 3.1689, "step": 25708 }, { "epoch": 4.9, "learning_rate": 1.532373689227836e-06, "loss": 3.1185, "step": 25709 }, { "epoch": 4.9, "learning_rate": 1.5323164918970449e-06, "loss": 2.1422, "step": 25710 }, { "epoch": 4.9, "learning_rate": 1.5322592945662535e-06, "loss": 1.3401, "step": 25711 }, { "epoch": 4.9, "learning_rate": 1.5322020972354625e-06, "loss": 1.9474, "step": 25712 }, { "epoch": 4.9, "learning_rate": 1.532144899904671e-06, "loss": 1.5161, "step": 25713 }, { "epoch": 4.9, "learning_rate": 1.53208770257388e-06, "loss": 2.1815, "step": 25714 }, { "epoch": 4.9, "learning_rate": 1.5320305052430888e-06, "loss": 1.4481, "step": 25715 }, { "epoch": 4.9, "learning_rate": 1.5319733079122974e-06, "loss": 1.9558, "step": 25716 }, { "epoch": 4.9, "learning_rate": 1.5319161105815064e-06, "loss": 1.8113, "step": 25717 }, { "epoch": 4.9, "learning_rate": 1.531858913250715e-06, "loss": 1.9238, "step": 25718 }, { "epoch": 4.9, "learning_rate": 1.5318017159199236e-06, "loss": 1.3611, "step": 25719 }, { "epoch": 4.9, "learning_rate": 1.5317445185891326e-06, "loss": 1.7807, "step": 25720 }, { "epoch": 4.9, "learning_rate": 1.5316873212583414e-06, "loss": 1.3502, "step": 25721 }, { "epoch": 4.9, "learning_rate": 1.53163012392755e-06, "loss": 2.1689, "step": 25722 }, { "epoch": 4.9, "learning_rate": 1.531572926596759e-06, "loss": 1.442, "step": 25723 }, { "epoch": 4.9, "learning_rate": 1.5315157292659676e-06, "loss": 1.0156, "step": 25724 }, { "epoch": 4.9, "learning_rate": 1.5314585319351764e-06, "loss": 2.217, "step": 25725 }, { "epoch": 4.9, "learning_rate": 1.5314013346043851e-06, "loss": 2.5833, "step": 25726 }, { "epoch": 4.9, "learning_rate": 1.531344137273594e-06, "loss": 2.8212, "step": 25727 }, { "epoch": 4.9, "learning_rate": 1.5312869399428025e-06, "loss": 2.2041, "step": 25728 }, { "epoch": 4.9, "learning_rate": 1.5312297426120115e-06, "loss": 2.3979, "step": 25729 }, { "epoch": 4.9, "learning_rate": 1.5311725452812201e-06, "loss": 1.3645, "step": 25730 }, { "epoch": 4.9, "learning_rate": 1.5311153479504291e-06, "loss": 2.3281, "step": 25731 }, { "epoch": 4.9, "learning_rate": 1.531058150619638e-06, "loss": 1.9472, "step": 25732 }, { "epoch": 4.9, "learning_rate": 1.5310009532888465e-06, "loss": 2.633, "step": 25733 }, { "epoch": 4.9, "learning_rate": 1.5309437559580555e-06, "loss": 2.6361, "step": 25734 }, { "epoch": 4.9, "learning_rate": 1.530886558627264e-06, "loss": 1.9076, "step": 25735 }, { "epoch": 4.9, "learning_rate": 1.5308293612964729e-06, "loss": 1.2256, "step": 25736 }, { "epoch": 4.9, "learning_rate": 1.5307721639656817e-06, "loss": 1.7394, "step": 25737 }, { "epoch": 4.9, "learning_rate": 1.5307149666348905e-06, "loss": 2.708, "step": 25738 }, { "epoch": 4.9, "learning_rate": 1.530657769304099e-06, "loss": 2.7188, "step": 25739 }, { "epoch": 4.9, "learning_rate": 1.530600571973308e-06, "loss": 2.9276, "step": 25740 }, { "epoch": 4.9, "learning_rate": 1.5305433746425166e-06, "loss": 2.4792, "step": 25741 }, { "epoch": 4.9, "learning_rate": 1.5304861773117254e-06, "loss": 2.6729, "step": 25742 }, { "epoch": 4.9, "learning_rate": 1.5304289799809344e-06, "loss": 2.7223, "step": 25743 }, { "epoch": 4.9, "learning_rate": 1.530371782650143e-06, "loss": 1.3812, "step": 25744 }, { "epoch": 4.9, "learning_rate": 1.5303145853193516e-06, "loss": 2.625, "step": 25745 }, { "epoch": 4.9, "learning_rate": 1.5302573879885606e-06, "loss": 2.1158, "step": 25746 }, { "epoch": 4.9, "learning_rate": 1.5302001906577692e-06, "loss": 2.5956, "step": 25747 }, { "epoch": 4.9, "learning_rate": 1.5301429933269782e-06, "loss": 1.812, "step": 25748 }, { "epoch": 4.9, "learning_rate": 1.530085795996187e-06, "loss": 3.1897, "step": 25749 }, { "epoch": 4.9, "learning_rate": 1.5300285986653956e-06, "loss": 1.8909, "step": 25750 }, { "epoch": 4.9, "learning_rate": 1.5299714013346046e-06, "loss": 1.6199, "step": 25751 }, { "epoch": 4.91, "learning_rate": 1.5299142040038132e-06, "loss": 1.3288, "step": 25752 }, { "epoch": 4.91, "learning_rate": 1.529857006673022e-06, "loss": 2.6474, "step": 25753 }, { "epoch": 4.91, "learning_rate": 1.529799809342231e-06, "loss": 2.4847, "step": 25754 }, { "epoch": 4.91, "learning_rate": 1.5297426120114395e-06, "loss": 1.698, "step": 25755 }, { "epoch": 4.91, "learning_rate": 1.5296854146806481e-06, "loss": 1.9988, "step": 25756 }, { "epoch": 4.91, "learning_rate": 1.5296282173498571e-06, "loss": 1.3059, "step": 25757 }, { "epoch": 4.91, "learning_rate": 1.5295710200190657e-06, "loss": 2.4523, "step": 25758 }, { "epoch": 4.91, "learning_rate": 1.5295138226882745e-06, "loss": 0.8018, "step": 25759 }, { "epoch": 4.91, "learning_rate": 1.5294566253574835e-06, "loss": 1.8063, "step": 25760 }, { "epoch": 4.91, "learning_rate": 1.529399428026692e-06, "loss": 1.2869, "step": 25761 }, { "epoch": 4.91, "learning_rate": 1.5293422306959007e-06, "loss": 2.9976, "step": 25762 }, { "epoch": 4.91, "learning_rate": 1.5292850333651097e-06, "loss": 1.1459, "step": 25763 }, { "epoch": 4.91, "learning_rate": 1.5292278360343185e-06, "loss": 1.9418, "step": 25764 }, { "epoch": 4.91, "learning_rate": 1.5291706387035273e-06, "loss": 1.5951, "step": 25765 }, { "epoch": 4.91, "learning_rate": 1.529113441372736e-06, "loss": 1.643, "step": 25766 }, { "epoch": 4.91, "learning_rate": 1.5290562440419446e-06, "loss": 2.7011, "step": 25767 }, { "epoch": 4.91, "learning_rate": 1.5289990467111536e-06, "loss": 2.1456, "step": 25768 }, { "epoch": 4.91, "learning_rate": 1.5289418493803622e-06, "loss": 1.1011, "step": 25769 }, { "epoch": 4.91, "learning_rate": 1.528884652049571e-06, "loss": 1.8656, "step": 25770 }, { "epoch": 4.91, "learning_rate": 1.52882745471878e-06, "loss": 1.4191, "step": 25771 }, { "epoch": 4.91, "learning_rate": 1.5287702573879886e-06, "loss": 1.3956, "step": 25772 }, { "epoch": 4.91, "learning_rate": 1.5287130600571972e-06, "loss": 1.6091, "step": 25773 }, { "epoch": 4.91, "learning_rate": 1.5286558627264062e-06, "loss": 1.8179, "step": 25774 }, { "epoch": 4.91, "learning_rate": 1.5285986653956148e-06, "loss": 1.1288, "step": 25775 }, { "epoch": 4.91, "learning_rate": 1.5285414680648236e-06, "loss": 2.5162, "step": 25776 }, { "epoch": 4.91, "learning_rate": 1.5284842707340326e-06, "loss": 1.4895, "step": 25777 }, { "epoch": 4.91, "learning_rate": 1.5284270734032412e-06, "loss": 3.7947, "step": 25778 }, { "epoch": 4.91, "learning_rate": 1.5283698760724502e-06, "loss": 1.4531, "step": 25779 }, { "epoch": 4.91, "learning_rate": 1.5283126787416587e-06, "loss": 2.26, "step": 25780 }, { "epoch": 4.91, "learning_rate": 1.5282554814108675e-06, "loss": 2.1618, "step": 25781 }, { "epoch": 4.91, "learning_rate": 1.5281982840800765e-06, "loss": 1.9953, "step": 25782 }, { "epoch": 4.91, "learning_rate": 1.5281410867492851e-06, "loss": 1.7721, "step": 25783 }, { "epoch": 4.91, "learning_rate": 1.5280838894184937e-06, "loss": 2.0374, "step": 25784 }, { "epoch": 4.91, "learning_rate": 1.5280266920877027e-06, "loss": 2.3235, "step": 25785 }, { "epoch": 4.91, "learning_rate": 1.5279694947569113e-06, "loss": 2.4035, "step": 25786 }, { "epoch": 4.91, "learning_rate": 1.52791229742612e-06, "loss": 1.5558, "step": 25787 }, { "epoch": 4.91, "learning_rate": 1.527855100095329e-06, "loss": 2.5219, "step": 25788 }, { "epoch": 4.91, "learning_rate": 1.5277979027645377e-06, "loss": 1.5259, "step": 25789 }, { "epoch": 4.91, "learning_rate": 1.5277407054337463e-06, "loss": 1.0402, "step": 25790 }, { "epoch": 4.91, "learning_rate": 1.5276835081029553e-06, "loss": 1.653, "step": 25791 }, { "epoch": 4.91, "learning_rate": 1.527626310772164e-06, "loss": 1.772, "step": 25792 }, { "epoch": 4.91, "learning_rate": 1.5275691134413726e-06, "loss": 2.0353, "step": 25793 }, { "epoch": 4.91, "learning_rate": 1.5275119161105816e-06, "loss": 1.6245, "step": 25794 }, { "epoch": 4.91, "learning_rate": 1.5274547187797902e-06, "loss": 1.8577, "step": 25795 }, { "epoch": 4.91, "learning_rate": 1.5273975214489992e-06, "loss": 2.2359, "step": 25796 }, { "epoch": 4.91, "learning_rate": 1.5273403241182078e-06, "loss": 1.5768, "step": 25797 }, { "epoch": 4.91, "learning_rate": 1.5272831267874166e-06, "loss": 1.8375, "step": 25798 }, { "epoch": 4.91, "learning_rate": 1.5272259294566256e-06, "loss": 2.2557, "step": 25799 }, { "epoch": 4.91, "learning_rate": 1.5271687321258342e-06, "loss": 1.6604, "step": 25800 }, { "epoch": 4.91, "learning_rate": 1.5271115347950428e-06, "loss": 2.6563, "step": 25801 }, { "epoch": 4.91, "learning_rate": 1.5270543374642518e-06, "loss": 3.9148, "step": 25802 }, { "epoch": 4.91, "learning_rate": 1.5269971401334606e-06, "loss": 1.9429, "step": 25803 }, { "epoch": 4.92, "learning_rate": 1.5269399428026692e-06, "loss": 2.0045, "step": 25804 }, { "epoch": 4.92, "learning_rate": 1.5268827454718782e-06, "loss": 1.4248, "step": 25805 }, { "epoch": 4.92, "learning_rate": 1.5268255481410867e-06, "loss": 1.3229, "step": 25806 }, { "epoch": 4.92, "learning_rate": 1.5267683508102953e-06, "loss": 1.8083, "step": 25807 }, { "epoch": 4.92, "learning_rate": 1.5267111534795043e-06, "loss": 3.0936, "step": 25808 }, { "epoch": 4.92, "learning_rate": 1.5266539561487131e-06, "loss": 1.7594, "step": 25809 }, { "epoch": 4.92, "learning_rate": 1.5265967588179217e-06, "loss": 2.5497, "step": 25810 }, { "epoch": 4.92, "learning_rate": 1.5265395614871307e-06, "loss": 1.4654, "step": 25811 }, { "epoch": 4.92, "learning_rate": 1.5264823641563393e-06, "loss": 1.1615, "step": 25812 }, { "epoch": 4.92, "learning_rate": 1.5264251668255483e-06, "loss": 1.2004, "step": 25813 }, { "epoch": 4.92, "learning_rate": 1.5263679694947569e-06, "loss": 1.6112, "step": 25814 }, { "epoch": 4.92, "learning_rate": 1.5263107721639657e-06, "loss": 2.2749, "step": 25815 }, { "epoch": 4.92, "learning_rate": 1.5262535748331747e-06, "loss": 2.1091, "step": 25816 }, { "epoch": 4.92, "learning_rate": 1.5261963775023833e-06, "loss": 2.1167, "step": 25817 }, { "epoch": 4.92, "learning_rate": 1.5261391801715919e-06, "loss": 2.7886, "step": 25818 }, { "epoch": 4.92, "learning_rate": 1.5260819828408009e-06, "loss": 2.0616, "step": 25819 }, { "epoch": 4.92, "learning_rate": 1.5260247855100097e-06, "loss": 2.2216, "step": 25820 }, { "epoch": 4.92, "learning_rate": 1.5259675881792182e-06, "loss": 2.1136, "step": 25821 }, { "epoch": 4.92, "learning_rate": 1.5259103908484272e-06, "loss": 3.0138, "step": 25822 }, { "epoch": 4.92, "learning_rate": 1.5258531935176358e-06, "loss": 1.6248, "step": 25823 }, { "epoch": 4.92, "learning_rate": 1.5257959961868446e-06, "loss": 2.3862, "step": 25824 }, { "epoch": 4.92, "learning_rate": 1.5257387988560534e-06, "loss": 2.1341, "step": 25825 }, { "epoch": 4.92, "learning_rate": 1.5256816015252622e-06, "loss": 2.7271, "step": 25826 }, { "epoch": 4.92, "learning_rate": 1.5256244041944708e-06, "loss": 2.1885, "step": 25827 }, { "epoch": 4.92, "learning_rate": 1.5255672068636798e-06, "loss": 1.7329, "step": 25828 }, { "epoch": 4.92, "learning_rate": 1.5255100095328884e-06, "loss": 0.9713, "step": 25829 }, { "epoch": 4.92, "learning_rate": 1.5254528122020974e-06, "loss": 2.0712, "step": 25830 }, { "epoch": 4.92, "learning_rate": 1.5253956148713062e-06, "loss": 1.2623, "step": 25831 }, { "epoch": 4.92, "learning_rate": 1.5253384175405148e-06, "loss": 1.6334, "step": 25832 }, { "epoch": 4.92, "learning_rate": 1.5252812202097238e-06, "loss": 3.048, "step": 25833 }, { "epoch": 4.92, "learning_rate": 1.5252240228789323e-06, "loss": 2.6066, "step": 25834 }, { "epoch": 4.92, "learning_rate": 1.525166825548141e-06, "loss": 2.3007, "step": 25835 }, { "epoch": 4.92, "learning_rate": 1.52510962821735e-06, "loss": 2.2585, "step": 25836 }, { "epoch": 4.92, "learning_rate": 1.5250524308865587e-06, "loss": 1.9657, "step": 25837 }, { "epoch": 4.92, "learning_rate": 1.5249952335557673e-06, "loss": 2.6283, "step": 25838 }, { "epoch": 4.92, "learning_rate": 1.5249380362249763e-06, "loss": 2.2131, "step": 25839 }, { "epoch": 4.92, "learning_rate": 1.5248808388941849e-06, "loss": 3.2906, "step": 25840 }, { "epoch": 4.92, "learning_rate": 1.5248236415633937e-06, "loss": 1.4748, "step": 25841 }, { "epoch": 4.92, "learning_rate": 1.5247664442326027e-06, "loss": 2.4383, "step": 25842 }, { "epoch": 4.92, "learning_rate": 1.5247092469018113e-06, "loss": 2.4487, "step": 25843 }, { "epoch": 4.92, "learning_rate": 1.5246520495710203e-06, "loss": 1.5027, "step": 25844 }, { "epoch": 4.92, "learning_rate": 1.5245948522402289e-06, "loss": 2.7733, "step": 25845 }, { "epoch": 4.92, "learning_rate": 1.5245376549094374e-06, "loss": 2.4971, "step": 25846 }, { "epoch": 4.92, "learning_rate": 1.5244804575786464e-06, "loss": 2.2976, "step": 25847 }, { "epoch": 4.92, "learning_rate": 1.5244232602478552e-06, "loss": 1.9832, "step": 25848 }, { "epoch": 4.92, "learning_rate": 1.5243660629170638e-06, "loss": 2.3426, "step": 25849 }, { "epoch": 4.92, "learning_rate": 1.5243088655862728e-06, "loss": 3.4346, "step": 25850 }, { "epoch": 4.92, "learning_rate": 1.5242516682554814e-06, "loss": 1.4873, "step": 25851 }, { "epoch": 4.92, "learning_rate": 1.5241944709246902e-06, "loss": 2.103, "step": 25852 }, { "epoch": 4.92, "learning_rate": 1.524137273593899e-06, "loss": 1.848, "step": 25853 }, { "epoch": 4.92, "learning_rate": 1.5240800762631078e-06, "loss": 1.772, "step": 25854 }, { "epoch": 4.92, "learning_rate": 1.5240228789323164e-06, "loss": 2.2011, "step": 25855 }, { "epoch": 4.92, "learning_rate": 1.5239656816015254e-06, "loss": 1.5267, "step": 25856 }, { "epoch": 4.93, "learning_rate": 1.523908484270734e-06, "loss": 2.2308, "step": 25857 }, { "epoch": 4.93, "learning_rate": 1.5238512869399428e-06, "loss": 3.1275, "step": 25858 }, { "epoch": 4.93, "learning_rate": 1.5237940896091518e-06, "loss": 1.7474, "step": 25859 }, { "epoch": 4.93, "learning_rate": 1.5237368922783603e-06, "loss": 3.5572, "step": 25860 }, { "epoch": 4.93, "learning_rate": 1.5236796949475694e-06, "loss": 2.6356, "step": 25861 }, { "epoch": 4.93, "learning_rate": 1.523622497616778e-06, "loss": 2.1585, "step": 25862 }, { "epoch": 4.93, "learning_rate": 1.5235653002859865e-06, "loss": 3.0341, "step": 25863 }, { "epoch": 4.93, "learning_rate": 1.5235081029551955e-06, "loss": 2.5971, "step": 25864 }, { "epoch": 4.93, "learning_rate": 1.5234509056244043e-06, "loss": 1.8283, "step": 25865 }, { "epoch": 4.93, "learning_rate": 1.523393708293613e-06, "loss": 1.2065, "step": 25866 }, { "epoch": 4.93, "learning_rate": 1.523336510962822e-06, "loss": 1.9299, "step": 25867 }, { "epoch": 4.93, "learning_rate": 1.5232793136320305e-06, "loss": 2.3137, "step": 25868 }, { "epoch": 4.93, "learning_rate": 1.5232221163012393e-06, "loss": 1.6008, "step": 25869 }, { "epoch": 4.93, "learning_rate": 1.5231649189704483e-06, "loss": 1.5381, "step": 25870 }, { "epoch": 4.93, "learning_rate": 1.5231077216396569e-06, "loss": 2.6607, "step": 25871 }, { "epoch": 4.93, "learning_rate": 1.5230505243088654e-06, "loss": 1.8863, "step": 25872 }, { "epoch": 4.93, "learning_rate": 1.5229933269780745e-06, "loss": 1.5287, "step": 25873 }, { "epoch": 4.93, "learning_rate": 1.522936129647283e-06, "loss": 1.5987, "step": 25874 }, { "epoch": 4.93, "learning_rate": 1.5228789323164918e-06, "loss": 1.5265, "step": 25875 }, { "epoch": 4.93, "learning_rate": 1.5228217349857008e-06, "loss": 2.3801, "step": 25876 }, { "epoch": 4.93, "learning_rate": 1.5227645376549094e-06, "loss": 2.1854, "step": 25877 }, { "epoch": 4.93, "learning_rate": 1.5227073403241184e-06, "loss": 1.5575, "step": 25878 }, { "epoch": 4.93, "learning_rate": 1.522650142993327e-06, "loss": 2.6954, "step": 25879 }, { "epoch": 4.93, "learning_rate": 1.5225929456625358e-06, "loss": 1.273, "step": 25880 }, { "epoch": 4.93, "learning_rate": 1.5225357483317446e-06, "loss": 0.758, "step": 25881 }, { "epoch": 4.93, "learning_rate": 1.5224785510009534e-06, "loss": 1.7638, "step": 25882 }, { "epoch": 4.93, "learning_rate": 1.522421353670162e-06, "loss": 1.5807, "step": 25883 }, { "epoch": 4.93, "learning_rate": 1.522364156339371e-06, "loss": 3.226, "step": 25884 }, { "epoch": 4.93, "learning_rate": 1.5223069590085796e-06, "loss": 3.8462, "step": 25885 }, { "epoch": 4.93, "learning_rate": 1.5222497616777883e-06, "loss": 1.6151, "step": 25886 }, { "epoch": 4.93, "learning_rate": 1.5221925643469974e-06, "loss": 1.7799, "step": 25887 }, { "epoch": 4.93, "learning_rate": 1.522135367016206e-06, "loss": 2.1024, "step": 25888 }, { "epoch": 4.93, "learning_rate": 1.5220781696854145e-06, "loss": 3.2568, "step": 25889 }, { "epoch": 4.93, "learning_rate": 1.5220209723546235e-06, "loss": 2.7544, "step": 25890 }, { "epoch": 4.93, "learning_rate": 1.5219637750238323e-06, "loss": 0.8337, "step": 25891 }, { "epoch": 4.93, "learning_rate": 1.521906577693041e-06, "loss": 2.9405, "step": 25892 }, { "epoch": 4.93, "learning_rate": 1.52184938036225e-06, "loss": 1.408, "step": 25893 }, { "epoch": 4.93, "learning_rate": 1.5217921830314585e-06, "loss": 1.5267, "step": 25894 }, { "epoch": 4.93, "learning_rate": 1.5217349857006675e-06, "loss": 1.9637, "step": 25895 }, { "epoch": 4.93, "learning_rate": 1.521677788369876e-06, "loss": 1.7225, "step": 25896 }, { "epoch": 4.93, "learning_rate": 1.5216205910390849e-06, "loss": 2.1614, "step": 25897 }, { "epoch": 4.93, "learning_rate": 1.5215633937082939e-06, "loss": 2.7138, "step": 25898 }, { "epoch": 4.93, "learning_rate": 1.5215061963775025e-06, "loss": 3.2579, "step": 25899 }, { "epoch": 4.93, "learning_rate": 1.521448999046711e-06, "loss": 2.6818, "step": 25900 }, { "epoch": 4.93, "learning_rate": 1.52139180171592e-06, "loss": 1.8365, "step": 25901 }, { "epoch": 4.93, "learning_rate": 1.5213346043851286e-06, "loss": 2.6831, "step": 25902 }, { "epoch": 4.93, "learning_rate": 1.5212774070543374e-06, "loss": 3.3018, "step": 25903 }, { "epoch": 4.93, "learning_rate": 1.5212202097235464e-06, "loss": 1.2427, "step": 25904 }, { "epoch": 4.93, "learning_rate": 1.521163012392755e-06, "loss": 2.0635, "step": 25905 }, { "epoch": 4.93, "learning_rate": 1.5211058150619636e-06, "loss": 1.6092, "step": 25906 }, { "epoch": 4.93, "learning_rate": 1.5210486177311726e-06, "loss": 1.189, "step": 25907 }, { "epoch": 4.93, "learning_rate": 1.5209914204003814e-06, "loss": 3.0515, "step": 25908 }, { "epoch": 4.94, "learning_rate": 1.5209342230695904e-06, "loss": 1.9428, "step": 25909 }, { "epoch": 4.94, "learning_rate": 1.520877025738799e-06, "loss": 2.3277, "step": 25910 }, { "epoch": 4.94, "learning_rate": 1.5208198284080076e-06, "loss": 1.588, "step": 25911 }, { "epoch": 4.94, "learning_rate": 1.5207626310772166e-06, "loss": 1.0342, "step": 25912 }, { "epoch": 4.94, "learning_rate": 1.5207054337464251e-06, "loss": 1.8867, "step": 25913 }, { "epoch": 4.94, "learning_rate": 1.520648236415634e-06, "loss": 1.834, "step": 25914 }, { "epoch": 4.94, "learning_rate": 1.520591039084843e-06, "loss": 2.412, "step": 25915 }, { "epoch": 4.94, "learning_rate": 1.5205338417540515e-06, "loss": 2.0077, "step": 25916 }, { "epoch": 4.94, "learning_rate": 1.5204766444232601e-06, "loss": 2.2337, "step": 25917 }, { "epoch": 4.94, "learning_rate": 1.5204194470924691e-06, "loss": 2.3043, "step": 25918 }, { "epoch": 4.94, "learning_rate": 1.520362249761678e-06, "loss": 1.9437, "step": 25919 }, { "epoch": 4.94, "learning_rate": 1.5203050524308865e-06, "loss": 3.2432, "step": 25920 }, { "epoch": 4.94, "learning_rate": 1.5202478551000955e-06, "loss": 1.7192, "step": 25921 }, { "epoch": 4.94, "learning_rate": 1.520190657769304e-06, "loss": 2.8903, "step": 25922 }, { "epoch": 4.94, "learning_rate": 1.5201334604385127e-06, "loss": 2.2929, "step": 25923 }, { "epoch": 4.94, "learning_rate": 1.5200762631077217e-06, "loss": 2.2866, "step": 25924 }, { "epoch": 4.94, "learning_rate": 1.5200190657769305e-06, "loss": 1.9058, "step": 25925 }, { "epoch": 4.94, "learning_rate": 1.5199618684461395e-06, "loss": 3.6858, "step": 25926 }, { "epoch": 4.94, "learning_rate": 1.519904671115348e-06, "loss": 2.6594, "step": 25927 }, { "epoch": 4.94, "learning_rate": 1.5198474737845566e-06, "loss": 2.0941, "step": 25928 }, { "epoch": 4.94, "learning_rate": 1.5197902764537656e-06, "loss": 2.1585, "step": 25929 }, { "epoch": 4.94, "learning_rate": 1.5197330791229742e-06, "loss": 1.2608, "step": 25930 }, { "epoch": 4.94, "learning_rate": 1.519675881792183e-06, "loss": 2.7104, "step": 25931 }, { "epoch": 4.94, "learning_rate": 1.519618684461392e-06, "loss": 1.3961, "step": 25932 }, { "epoch": 4.94, "learning_rate": 1.5195614871306006e-06, "loss": 0.8872, "step": 25933 }, { "epoch": 4.94, "learning_rate": 1.5195042897998092e-06, "loss": 1.3999, "step": 25934 }, { "epoch": 4.94, "learning_rate": 1.5194470924690182e-06, "loss": 2.3742, "step": 25935 }, { "epoch": 4.94, "learning_rate": 1.519389895138227e-06, "loss": 0.8867, "step": 25936 }, { "epoch": 4.94, "learning_rate": 1.5193326978074356e-06, "loss": 2.2975, "step": 25937 }, { "epoch": 4.94, "learning_rate": 1.5192755004766446e-06, "loss": 1.807, "step": 25938 }, { "epoch": 4.94, "learning_rate": 1.5192183031458532e-06, "loss": 0.9401, "step": 25939 }, { "epoch": 4.94, "learning_rate": 1.519161105815062e-06, "loss": 2.2366, "step": 25940 }, { "epoch": 4.94, "learning_rate": 1.5191039084842707e-06, "loss": 2.5308, "step": 25941 }, { "epoch": 4.94, "learning_rate": 1.5190467111534795e-06, "loss": 1.8473, "step": 25942 }, { "epoch": 4.94, "learning_rate": 1.5189895138226885e-06, "loss": 1.8201, "step": 25943 }, { "epoch": 4.94, "learning_rate": 1.5189323164918971e-06, "loss": 3.1422, "step": 25944 }, { "epoch": 4.94, "learning_rate": 1.5188751191611057e-06, "loss": 1.8442, "step": 25945 }, { "epoch": 4.94, "learning_rate": 1.5188179218303147e-06, "loss": 1.3769, "step": 25946 }, { "epoch": 4.94, "learning_rate": 1.5187607244995235e-06, "loss": 2.5438, "step": 25947 }, { "epoch": 4.94, "learning_rate": 1.518703527168732e-06, "loss": 2.1149, "step": 25948 }, { "epoch": 4.94, "learning_rate": 1.518646329837941e-06, "loss": 3.4913, "step": 25949 }, { "epoch": 4.94, "learning_rate": 1.5185891325071497e-06, "loss": 1.7212, "step": 25950 }, { "epoch": 4.94, "learning_rate": 1.5185319351763583e-06, "loss": 2.9453, "step": 25951 }, { "epoch": 4.94, "learning_rate": 1.5184747378455673e-06, "loss": 3.306, "step": 25952 }, { "epoch": 4.94, "learning_rate": 1.518417540514776e-06, "loss": 1.5766, "step": 25953 }, { "epoch": 4.94, "learning_rate": 1.5183603431839846e-06, "loss": 1.2819, "step": 25954 }, { "epoch": 4.94, "learning_rate": 1.5183031458531936e-06, "loss": 2.0156, "step": 25955 }, { "epoch": 4.94, "learning_rate": 1.5182459485224022e-06, "loss": 2.5853, "step": 25956 }, { "epoch": 4.94, "learning_rate": 1.518188751191611e-06, "loss": 2.4727, "step": 25957 }, { "epoch": 4.94, "learning_rate": 1.51813155386082e-06, "loss": 2.9294, "step": 25958 }, { "epoch": 4.94, "learning_rate": 1.5180743565300286e-06, "loss": 1.3878, "step": 25959 }, { "epoch": 4.94, "learning_rate": 1.5180171591992376e-06, "loss": 2.2319, "step": 25960 }, { "epoch": 4.94, "learning_rate": 1.5179599618684462e-06, "loss": 2.5183, "step": 25961 }, { "epoch": 4.95, "learning_rate": 1.5179027645376548e-06, "loss": 1.6452, "step": 25962 }, { "epoch": 4.95, "learning_rate": 1.5178455672068638e-06, "loss": 2.8078, "step": 25963 }, { "epoch": 4.95, "learning_rate": 1.5177883698760726e-06, "loss": 2.4043, "step": 25964 }, { "epoch": 4.95, "learning_rate": 1.5177311725452812e-06, "loss": 2.0174, "step": 25965 }, { "epoch": 4.95, "learning_rate": 1.5176739752144902e-06, "loss": 1.4765, "step": 25966 }, { "epoch": 4.95, "learning_rate": 1.5176167778836987e-06, "loss": 1.9425, "step": 25967 }, { "epoch": 4.95, "learning_rate": 1.5175595805529075e-06, "loss": 1.7449, "step": 25968 }, { "epoch": 4.95, "learning_rate": 1.5175023832221163e-06, "loss": 3.1118, "step": 25969 }, { "epoch": 4.95, "learning_rate": 1.5174451858913251e-06, "loss": 1.5326, "step": 25970 }, { "epoch": 4.95, "learning_rate": 1.5173879885605337e-06, "loss": 1.3854, "step": 25971 }, { "epoch": 4.95, "learning_rate": 1.5173307912297427e-06, "loss": 1.2475, "step": 25972 }, { "epoch": 4.95, "learning_rate": 1.5172735938989513e-06, "loss": 1.8644, "step": 25973 }, { "epoch": 4.95, "learning_rate": 1.51721639656816e-06, "loss": 1.7396, "step": 25974 }, { "epoch": 4.95, "learning_rate": 1.517159199237369e-06, "loss": 2.8178, "step": 25975 }, { "epoch": 4.95, "learning_rate": 1.5171020019065777e-06, "loss": 2.5931, "step": 25976 }, { "epoch": 4.95, "learning_rate": 1.5170448045757867e-06, "loss": 0.8314, "step": 25977 }, { "epoch": 4.95, "learning_rate": 1.5169876072449953e-06, "loss": 2.4542, "step": 25978 }, { "epoch": 4.95, "learning_rate": 1.516930409914204e-06, "loss": 2.6924, "step": 25979 }, { "epoch": 4.95, "learning_rate": 1.5168732125834128e-06, "loss": 1.791, "step": 25980 }, { "epoch": 4.95, "learning_rate": 1.5168160152526216e-06, "loss": 1.8237, "step": 25981 }, { "epoch": 4.95, "learning_rate": 1.5167588179218302e-06, "loss": 2.3056, "step": 25982 }, { "epoch": 4.95, "learning_rate": 1.5167016205910392e-06, "loss": 2.3988, "step": 25983 }, { "epoch": 4.95, "learning_rate": 1.5166444232602478e-06, "loss": 1.9241, "step": 25984 }, { "epoch": 4.95, "learning_rate": 1.5165872259294566e-06, "loss": 2.3191, "step": 25985 }, { "epoch": 4.95, "learning_rate": 1.5165300285986656e-06, "loss": 2.3765, "step": 25986 }, { "epoch": 4.95, "learning_rate": 1.5164728312678742e-06, "loss": 1.736, "step": 25987 }, { "epoch": 4.95, "learning_rate": 1.5164156339370828e-06, "loss": 2.3006, "step": 25988 }, { "epoch": 4.95, "learning_rate": 1.5163584366062918e-06, "loss": 1.3794, "step": 25989 }, { "epoch": 4.95, "learning_rate": 1.5163012392755004e-06, "loss": 1.477, "step": 25990 }, { "epoch": 4.95, "learning_rate": 1.5162440419447094e-06, "loss": 2.0352, "step": 25991 }, { "epoch": 4.95, "learning_rate": 1.5161868446139182e-06, "loss": 1.9154, "step": 25992 }, { "epoch": 4.95, "learning_rate": 1.5161296472831267e-06, "loss": 1.6048, "step": 25993 }, { "epoch": 4.95, "learning_rate": 1.5160724499523358e-06, "loss": 0.8342, "step": 25994 }, { "epoch": 4.95, "learning_rate": 1.5160152526215443e-06, "loss": 2.2467, "step": 25995 }, { "epoch": 4.95, "learning_rate": 1.5159580552907531e-06, "loss": 1.0247, "step": 25996 }, { "epoch": 4.95, "learning_rate": 1.5159008579599621e-06, "loss": 1.7215, "step": 25997 }, { "epoch": 4.95, "learning_rate": 1.5158436606291707e-06, "loss": 1.165, "step": 25998 }, { "epoch": 4.95, "learning_rate": 1.5157864632983793e-06, "loss": 1.3331, "step": 25999 }, { "epoch": 4.95, "learning_rate": 1.5157292659675883e-06, "loss": 1.4446, "step": 26000 }, { "epoch": 4.95, "learning_rate": 1.5156720686367969e-06, "loss": 1.7975, "step": 26001 }, { "epoch": 4.95, "learning_rate": 1.5156148713060057e-06, "loss": 2.0161, "step": 26002 }, { "epoch": 4.95, "learning_rate": 1.5155576739752147e-06, "loss": 2.0544, "step": 26003 }, { "epoch": 4.95, "learning_rate": 1.5155004766444233e-06, "loss": 2.372, "step": 26004 }, { "epoch": 4.95, "learning_rate": 1.5154432793136318e-06, "loss": 1.5039, "step": 26005 }, { "epoch": 4.95, "learning_rate": 1.5153860819828409e-06, "loss": 1.794, "step": 26006 }, { "epoch": 4.95, "learning_rate": 1.5153288846520496e-06, "loss": 0.6818, "step": 26007 }, { "epoch": 4.95, "learning_rate": 1.5152716873212584e-06, "loss": 1.5616, "step": 26008 }, { "epoch": 4.95, "learning_rate": 1.5152144899904672e-06, "loss": 2.2114, "step": 26009 }, { "epoch": 4.95, "learning_rate": 1.5151572926596758e-06, "loss": 1.3239, "step": 26010 }, { "epoch": 4.95, "learning_rate": 1.5151000953288848e-06, "loss": 1.9515, "step": 26011 }, { "epoch": 4.95, "learning_rate": 1.5150428979980934e-06, "loss": 2.4598, "step": 26012 }, { "epoch": 4.95, "learning_rate": 1.5149857006673022e-06, "loss": 2.2655, "step": 26013 }, { "epoch": 4.96, "learning_rate": 1.5149285033365112e-06, "loss": 2.5229, "step": 26014 }, { "epoch": 4.96, "learning_rate": 1.5148713060057198e-06, "loss": 1.9048, "step": 26015 }, { "epoch": 4.96, "learning_rate": 1.5148141086749284e-06, "loss": 2.5843, "step": 26016 }, { "epoch": 4.96, "learning_rate": 1.5147569113441374e-06, "loss": 2.3658, "step": 26017 }, { "epoch": 4.96, "learning_rate": 1.514699714013346e-06, "loss": 1.5804, "step": 26018 }, { "epoch": 4.96, "learning_rate": 1.5146425166825547e-06, "loss": 2.2943, "step": 26019 }, { "epoch": 4.96, "learning_rate": 1.5145853193517638e-06, "loss": 1.0847, "step": 26020 }, { "epoch": 4.96, "learning_rate": 1.5145281220209723e-06, "loss": 2.3976, "step": 26021 }, { "epoch": 4.96, "learning_rate": 1.514470924690181e-06, "loss": 1.0075, "step": 26022 }, { "epoch": 4.96, "learning_rate": 1.51441372735939e-06, "loss": 2.7732, "step": 26023 }, { "epoch": 4.96, "learning_rate": 1.5143565300285987e-06, "loss": 1.8288, "step": 26024 }, { "epoch": 4.96, "learning_rate": 1.5142993326978077e-06, "loss": 2.1952, "step": 26025 }, { "epoch": 4.96, "learning_rate": 1.5142421353670163e-06, "loss": 2.4792, "step": 26026 }, { "epoch": 4.96, "learning_rate": 1.5141849380362249e-06, "loss": 1.5562, "step": 26027 }, { "epoch": 4.96, "learning_rate": 1.514127740705434e-06, "loss": 0.9539, "step": 26028 }, { "epoch": 4.96, "learning_rate": 1.5140705433746425e-06, "loss": 2.3765, "step": 26029 }, { "epoch": 4.96, "learning_rate": 1.5140133460438513e-06, "loss": 2.391, "step": 26030 }, { "epoch": 4.96, "learning_rate": 1.5139561487130603e-06, "loss": 2.4909, "step": 26031 }, { "epoch": 4.96, "learning_rate": 1.5138989513822689e-06, "loss": 2.4728, "step": 26032 }, { "epoch": 4.96, "learning_rate": 1.5138417540514774e-06, "loss": 2.0244, "step": 26033 }, { "epoch": 4.96, "learning_rate": 1.5137845567206864e-06, "loss": 2.9285, "step": 26034 }, { "epoch": 4.96, "learning_rate": 1.5137273593898952e-06, "loss": 2.4459, "step": 26035 }, { "epoch": 4.96, "learning_rate": 1.5136701620591038e-06, "loss": 1.66, "step": 26036 }, { "epoch": 4.96, "learning_rate": 1.5136129647283128e-06, "loss": 1.0184, "step": 26037 }, { "epoch": 4.96, "learning_rate": 1.5135557673975214e-06, "loss": 1.2711, "step": 26038 }, { "epoch": 4.96, "learning_rate": 1.51349857006673e-06, "loss": 3.3519, "step": 26039 }, { "epoch": 4.96, "learning_rate": 1.513441372735939e-06, "loss": 2.3583, "step": 26040 }, { "epoch": 4.96, "learning_rate": 1.5133841754051478e-06, "loss": 1.9118, "step": 26041 }, { "epoch": 4.96, "learning_rate": 1.5133269780743568e-06, "loss": 1.6612, "step": 26042 }, { "epoch": 4.96, "learning_rate": 1.5132697807435654e-06, "loss": 0.4903, "step": 26043 }, { "epoch": 4.96, "learning_rate": 1.513212583412774e-06, "loss": 2.2643, "step": 26044 }, { "epoch": 4.96, "learning_rate": 1.513155386081983e-06, "loss": 3.7058, "step": 26045 }, { "epoch": 4.96, "learning_rate": 1.5130981887511918e-06, "loss": 1.0294, "step": 26046 }, { "epoch": 4.96, "learning_rate": 1.5130409914204003e-06, "loss": 1.5301, "step": 26047 }, { "epoch": 4.96, "learning_rate": 1.5129837940896093e-06, "loss": 2.6674, "step": 26048 }, { "epoch": 4.96, "learning_rate": 1.512926596758818e-06, "loss": 1.186, "step": 26049 }, { "epoch": 4.96, "learning_rate": 1.5128693994280265e-06, "loss": 2.301, "step": 26050 }, { "epoch": 4.96, "learning_rate": 1.5128122020972355e-06, "loss": 1.4683, "step": 26051 }, { "epoch": 4.96, "learning_rate": 1.5127550047664443e-06, "loss": 2.7273, "step": 26052 }, { "epoch": 4.96, "learning_rate": 1.5126978074356529e-06, "loss": 2.1407, "step": 26053 }, { "epoch": 4.96, "learning_rate": 1.512640610104862e-06, "loss": 2.6262, "step": 26054 }, { "epoch": 4.96, "learning_rate": 1.5125834127740705e-06, "loss": 2.238, "step": 26055 }, { "epoch": 4.96, "learning_rate": 1.5125262154432795e-06, "loss": 1.2504, "step": 26056 }, { "epoch": 4.96, "learning_rate": 1.512469018112488e-06, "loss": 0.5048, "step": 26057 }, { "epoch": 4.96, "learning_rate": 1.5124118207816969e-06, "loss": 2.2243, "step": 26058 }, { "epoch": 4.96, "learning_rate": 1.5123546234509059e-06, "loss": 1.7202, "step": 26059 }, { "epoch": 4.96, "learning_rate": 1.5122974261201144e-06, "loss": 0.8306, "step": 26060 }, { "epoch": 4.96, "learning_rate": 1.512240228789323e-06, "loss": 1.336, "step": 26061 }, { "epoch": 4.96, "learning_rate": 1.512183031458532e-06, "loss": 2.2067, "step": 26062 }, { "epoch": 4.96, "learning_rate": 1.5121258341277408e-06, "loss": 1.3823, "step": 26063 }, { "epoch": 4.96, "learning_rate": 1.5120686367969494e-06, "loss": 3.7511, "step": 26064 }, { "epoch": 4.96, "learning_rate": 1.5120114394661584e-06, "loss": 1.1275, "step": 26065 }, { "epoch": 4.96, "learning_rate": 1.511954242135367e-06, "loss": 2.4662, "step": 26066 }, { "epoch": 4.97, "learning_rate": 1.5118970448045758e-06, "loss": 1.6538, "step": 26067 }, { "epoch": 4.97, "learning_rate": 1.5118398474737846e-06, "loss": 2.4346, "step": 26068 }, { "epoch": 4.97, "learning_rate": 1.5117826501429934e-06, "loss": 2.3779, "step": 26069 }, { "epoch": 4.97, "learning_rate": 1.511725452812202e-06, "loss": 2.1552, "step": 26070 }, { "epoch": 4.97, "learning_rate": 1.511668255481411e-06, "loss": 2.9699, "step": 26071 }, { "epoch": 4.97, "learning_rate": 1.5116110581506196e-06, "loss": 3.244, "step": 26072 }, { "epoch": 4.97, "learning_rate": 1.5115538608198286e-06, "loss": 2.0411, "step": 26073 }, { "epoch": 4.97, "learning_rate": 1.5114966634890374e-06, "loss": 2.0966, "step": 26074 }, { "epoch": 4.97, "learning_rate": 1.511439466158246e-06, "loss": 1.9826, "step": 26075 }, { "epoch": 4.97, "learning_rate": 1.511382268827455e-06, "loss": 1.5343, "step": 26076 }, { "epoch": 4.97, "learning_rate": 1.5113250714966635e-06, "loss": 1.2879, "step": 26077 }, { "epoch": 4.97, "learning_rate": 1.511267874165872e-06, "loss": 2.5259, "step": 26078 }, { "epoch": 4.97, "learning_rate": 1.5112106768350811e-06, "loss": 1.8995, "step": 26079 }, { "epoch": 4.97, "learning_rate": 1.51115347950429e-06, "loss": 1.0734, "step": 26080 }, { "epoch": 4.97, "learning_rate": 1.5110962821734985e-06, "loss": 2.0024, "step": 26081 }, { "epoch": 4.97, "learning_rate": 1.5110390848427075e-06, "loss": 1.9966, "step": 26082 }, { "epoch": 4.97, "learning_rate": 1.510981887511916e-06, "loss": 1.3154, "step": 26083 }, { "epoch": 4.97, "learning_rate": 1.5109246901811249e-06, "loss": 2.492, "step": 26084 }, { "epoch": 4.97, "learning_rate": 1.5108674928503337e-06, "loss": 2.1563, "step": 26085 }, { "epoch": 4.97, "learning_rate": 1.5108102955195425e-06, "loss": 2.282, "step": 26086 }, { "epoch": 4.97, "learning_rate": 1.510753098188751e-06, "loss": 4.0664, "step": 26087 }, { "epoch": 4.97, "learning_rate": 1.51069590085796e-06, "loss": 2.113, "step": 26088 }, { "epoch": 4.97, "learning_rate": 1.5106387035271686e-06, "loss": 2.1004, "step": 26089 }, { "epoch": 4.97, "learning_rate": 1.5105815061963776e-06, "loss": 2.206, "step": 26090 }, { "epoch": 4.97, "learning_rate": 1.5105243088655864e-06, "loss": 1.268, "step": 26091 }, { "epoch": 4.97, "learning_rate": 1.510467111534795e-06, "loss": 1.9553, "step": 26092 }, { "epoch": 4.97, "learning_rate": 1.510409914204004e-06, "loss": 1.0587, "step": 26093 }, { "epoch": 4.97, "learning_rate": 1.5103527168732126e-06, "loss": 1.7114, "step": 26094 }, { "epoch": 4.97, "learning_rate": 1.5102955195424214e-06, "loss": 1.4542, "step": 26095 }, { "epoch": 4.97, "learning_rate": 1.5102383222116302e-06, "loss": 2.9951, "step": 26096 }, { "epoch": 4.97, "learning_rate": 1.510181124880839e-06, "loss": 3.1436, "step": 26097 }, { "epoch": 4.97, "learning_rate": 1.5101239275500476e-06, "loss": 1.7064, "step": 26098 }, { "epoch": 4.97, "learning_rate": 1.5100667302192566e-06, "loss": 1.875, "step": 26099 }, { "epoch": 4.97, "learning_rate": 1.5100095328884651e-06, "loss": 2.631, "step": 26100 }, { "epoch": 4.97, "learning_rate": 1.509952335557674e-06, "loss": 1.1571, "step": 26101 }, { "epoch": 4.97, "learning_rate": 1.509895138226883e-06, "loss": 1.0634, "step": 26102 }, { "epoch": 4.97, "learning_rate": 1.5098379408960915e-06, "loss": 1.8994, "step": 26103 }, { "epoch": 4.97, "learning_rate": 1.5097807435653001e-06, "loss": 2.5672, "step": 26104 }, { "epoch": 4.97, "learning_rate": 1.5097235462345091e-06, "loss": 1.7927, "step": 26105 }, { "epoch": 4.97, "learning_rate": 1.5096663489037177e-06, "loss": 2.1631, "step": 26106 }, { "epoch": 4.97, "learning_rate": 1.5096091515729267e-06, "loss": 2.0717, "step": 26107 }, { "epoch": 4.97, "learning_rate": 1.5095519542421355e-06, "loss": 2.4126, "step": 26108 }, { "epoch": 4.97, "learning_rate": 1.509494756911344e-06, "loss": 2.0815, "step": 26109 }, { "epoch": 4.97, "learning_rate": 1.509437559580553e-06, "loss": 0.7884, "step": 26110 }, { "epoch": 4.97, "learning_rate": 1.5093803622497617e-06, "loss": 1.5444, "step": 26111 }, { "epoch": 4.97, "learning_rate": 1.5093231649189705e-06, "loss": 1.0339, "step": 26112 }, { "epoch": 4.97, "learning_rate": 1.5092659675881795e-06, "loss": 2.1415, "step": 26113 }, { "epoch": 4.97, "learning_rate": 1.509208770257388e-06, "loss": 2.069, "step": 26114 }, { "epoch": 4.97, "learning_rate": 1.5091515729265966e-06, "loss": 2.2232, "step": 26115 }, { "epoch": 4.97, "learning_rate": 1.5090943755958056e-06, "loss": 0.7744, "step": 26116 }, { "epoch": 4.97, "learning_rate": 1.5090371782650142e-06, "loss": 1.0473, "step": 26117 }, { "epoch": 4.97, "learning_rate": 1.508979980934223e-06, "loss": 3.3049, "step": 26118 }, { "epoch": 4.98, "learning_rate": 1.508922783603432e-06, "loss": 3.4374, "step": 26119 }, { "epoch": 4.98, "learning_rate": 1.5088655862726406e-06, "loss": 1.4357, "step": 26120 }, { "epoch": 4.98, "learning_rate": 1.5088083889418496e-06, "loss": 1.6275, "step": 26121 }, { "epoch": 4.98, "learning_rate": 1.5087511916110582e-06, "loss": 2.9552, "step": 26122 }, { "epoch": 4.98, "learning_rate": 1.508693994280267e-06, "loss": 2.5642, "step": 26123 }, { "epoch": 4.98, "learning_rate": 1.5086367969494758e-06, "loss": 3.8303, "step": 26124 }, { "epoch": 4.98, "learning_rate": 1.5085795996186846e-06, "loss": 3.0771, "step": 26125 }, { "epoch": 4.98, "learning_rate": 1.5085224022878931e-06, "loss": 1.6148, "step": 26126 }, { "epoch": 4.98, "learning_rate": 1.5084652049571022e-06, "loss": 1.9344, "step": 26127 }, { "epoch": 4.98, "learning_rate": 1.5084080076263107e-06, "loss": 2.7585, "step": 26128 }, { "epoch": 4.98, "learning_rate": 1.5083508102955195e-06, "loss": 1.5073, "step": 26129 }, { "epoch": 4.98, "learning_rate": 1.5082936129647285e-06, "loss": 1.2805, "step": 26130 }, { "epoch": 4.98, "learning_rate": 1.5082364156339371e-06, "loss": 2.082, "step": 26131 }, { "epoch": 4.98, "learning_rate": 1.5081792183031457e-06, "loss": 1.6671, "step": 26132 }, { "epoch": 4.98, "learning_rate": 1.5081220209723547e-06, "loss": 1.8896, "step": 26133 }, { "epoch": 4.98, "learning_rate": 1.5080648236415635e-06, "loss": 1.6995, "step": 26134 }, { "epoch": 4.98, "learning_rate": 1.508007626310772e-06, "loss": 2.2335, "step": 26135 }, { "epoch": 4.98, "learning_rate": 1.507950428979981e-06, "loss": 2.8286, "step": 26136 }, { "epoch": 4.98, "learning_rate": 1.5078932316491897e-06, "loss": 0.7442, "step": 26137 }, { "epoch": 4.98, "learning_rate": 1.5078360343183987e-06, "loss": 1.5132, "step": 26138 }, { "epoch": 4.98, "learning_rate": 1.5077788369876073e-06, "loss": 1.3828, "step": 26139 }, { "epoch": 4.98, "learning_rate": 1.507721639656816e-06, "loss": 1.1537, "step": 26140 }, { "epoch": 4.98, "learning_rate": 1.507664442326025e-06, "loss": 1.9006, "step": 26141 }, { "epoch": 4.98, "learning_rate": 1.5076072449952336e-06, "loss": 1.6666, "step": 26142 }, { "epoch": 4.98, "learning_rate": 1.5075500476644422e-06, "loss": 2.2282, "step": 26143 }, { "epoch": 4.98, "learning_rate": 1.5074928503336512e-06, "loss": 1.704, "step": 26144 }, { "epoch": 4.98, "learning_rate": 1.5074356530028598e-06, "loss": 3.0516, "step": 26145 }, { "epoch": 4.98, "learning_rate": 1.5073784556720686e-06, "loss": 2.7853, "step": 26146 }, { "epoch": 4.98, "learning_rate": 1.5073212583412776e-06, "loss": 3.1104, "step": 26147 }, { "epoch": 4.98, "learning_rate": 1.5072640610104862e-06, "loss": 2.3967, "step": 26148 }, { "epoch": 4.98, "learning_rate": 1.5072068636796948e-06, "loss": 2.1546, "step": 26149 }, { "epoch": 4.98, "learning_rate": 1.5071496663489038e-06, "loss": 2.8382, "step": 26150 }, { "epoch": 4.98, "learning_rate": 1.5070924690181126e-06, "loss": 1.5289, "step": 26151 }, { "epoch": 4.98, "learning_rate": 1.5070352716873212e-06, "loss": 1.2958, "step": 26152 }, { "epoch": 4.98, "learning_rate": 1.5069780743565302e-06, "loss": 0.7887, "step": 26153 }, { "epoch": 4.98, "learning_rate": 1.5069208770257387e-06, "loss": 0.9352, "step": 26154 }, { "epoch": 4.98, "learning_rate": 1.5068636796949477e-06, "loss": 0.9147, "step": 26155 }, { "epoch": 4.98, "learning_rate": 1.5068064823641563e-06, "loss": 2.5411, "step": 26156 }, { "epoch": 4.98, "learning_rate": 1.5067492850333651e-06, "loss": 1.4879, "step": 26157 }, { "epoch": 4.98, "learning_rate": 1.5066920877025741e-06, "loss": 1.792, "step": 26158 }, { "epoch": 4.98, "learning_rate": 1.5066348903717827e-06, "loss": 2.4285, "step": 26159 }, { "epoch": 4.98, "learning_rate": 1.5065776930409913e-06, "loss": 1.9465, "step": 26160 }, { "epoch": 4.98, "learning_rate": 1.5065204957102003e-06, "loss": 2.1395, "step": 26161 }, { "epoch": 4.98, "learning_rate": 1.506463298379409e-06, "loss": 3.3501, "step": 26162 }, { "epoch": 4.98, "learning_rate": 1.5064061010486177e-06, "loss": 1.1194, "step": 26163 }, { "epoch": 4.98, "learning_rate": 1.5063489037178267e-06, "loss": 2.2905, "step": 26164 }, { "epoch": 4.98, "learning_rate": 1.5062917063870353e-06, "loss": 2.264, "step": 26165 }, { "epoch": 4.98, "learning_rate": 1.5062345090562438e-06, "loss": 1.2762, "step": 26166 }, { "epoch": 4.98, "learning_rate": 1.5061773117254528e-06, "loss": 2.3407, "step": 26167 }, { "epoch": 4.98, "learning_rate": 1.5061201143946616e-06, "loss": 3.1233, "step": 26168 }, { "epoch": 4.98, "learning_rate": 1.5060629170638702e-06, "loss": 2.4097, "step": 26169 }, { "epoch": 4.98, "learning_rate": 1.5060057197330792e-06, "loss": 2.2102, "step": 26170 }, { "epoch": 4.98, "learning_rate": 1.5059485224022878e-06, "loss": 1.2574, "step": 26171 }, { "epoch": 4.99, "learning_rate": 1.5058913250714968e-06, "loss": 1.2306, "step": 26172 }, { "epoch": 4.99, "learning_rate": 1.5058341277407054e-06, "loss": 2.5668, "step": 26173 }, { "epoch": 4.99, "learning_rate": 1.5057769304099142e-06, "loss": 2.3881, "step": 26174 }, { "epoch": 4.99, "learning_rate": 1.5057197330791232e-06, "loss": 1.9114, "step": 26175 }, { "epoch": 4.99, "learning_rate": 1.5056625357483318e-06, "loss": 2.4572, "step": 26176 }, { "epoch": 4.99, "learning_rate": 1.5056053384175404e-06, "loss": 2.4407, "step": 26177 }, { "epoch": 4.99, "learning_rate": 1.5055481410867494e-06, "loss": 2.5069, "step": 26178 }, { "epoch": 4.99, "learning_rate": 1.5054909437559582e-06, "loss": 2.3065, "step": 26179 }, { "epoch": 4.99, "learning_rate": 1.5054337464251667e-06, "loss": 0.621, "step": 26180 }, { "epoch": 4.99, "learning_rate": 1.5053765490943757e-06, "loss": 0.7963, "step": 26181 }, { "epoch": 4.99, "learning_rate": 1.5053193517635843e-06, "loss": 2.5723, "step": 26182 }, { "epoch": 4.99, "learning_rate": 1.5052621544327931e-06, "loss": 1.4779, "step": 26183 }, { "epoch": 4.99, "learning_rate": 1.505204957102002e-06, "loss": 1.5156, "step": 26184 }, { "epoch": 4.99, "learning_rate": 1.5051477597712107e-06, "loss": 2.128, "step": 26185 }, { "epoch": 4.99, "learning_rate": 1.5050905624404197e-06, "loss": 0.9928, "step": 26186 }, { "epoch": 4.99, "learning_rate": 1.5050333651096283e-06, "loss": 2.9639, "step": 26187 }, { "epoch": 4.99, "learning_rate": 1.5049761677788369e-06, "loss": 2.8538, "step": 26188 }, { "epoch": 4.99, "learning_rate": 1.5049189704480459e-06, "loss": 1.8054, "step": 26189 }, { "epoch": 4.99, "learning_rate": 1.5048617731172547e-06, "loss": 1.1662, "step": 26190 }, { "epoch": 4.99, "learning_rate": 1.5048045757864633e-06, "loss": 1.8486, "step": 26191 }, { "epoch": 4.99, "learning_rate": 1.5047473784556723e-06, "loss": 1.6566, "step": 26192 }, { "epoch": 4.99, "learning_rate": 1.5046901811248809e-06, "loss": 2.2138, "step": 26193 }, { "epoch": 4.99, "learning_rate": 1.5046329837940894e-06, "loss": 2.2017, "step": 26194 }, { "epoch": 4.99, "learning_rate": 1.5045757864632984e-06, "loss": 2.3714, "step": 26195 }, { "epoch": 4.99, "learning_rate": 1.5045185891325072e-06, "loss": 3.441, "step": 26196 }, { "epoch": 4.99, "learning_rate": 1.5044613918017158e-06, "loss": 2.5592, "step": 26197 }, { "epoch": 4.99, "learning_rate": 1.5044041944709248e-06, "loss": 3.6328, "step": 26198 }, { "epoch": 4.99, "learning_rate": 1.5043469971401334e-06, "loss": 0.8004, "step": 26199 }, { "epoch": 4.99, "learning_rate": 1.5042897998093422e-06, "loss": 2.4813, "step": 26200 }, { "epoch": 4.99, "learning_rate": 1.5042326024785512e-06, "loss": 1.0849, "step": 26201 }, { "epoch": 4.99, "learning_rate": 1.5041754051477598e-06, "loss": 1.6254, "step": 26202 }, { "epoch": 4.99, "learning_rate": 1.5041182078169688e-06, "loss": 1.7459, "step": 26203 }, { "epoch": 4.99, "learning_rate": 1.5040610104861774e-06, "loss": 1.9073, "step": 26204 }, { "epoch": 4.99, "learning_rate": 1.504003813155386e-06, "loss": 1.6004, "step": 26205 }, { "epoch": 4.99, "learning_rate": 1.503946615824595e-06, "loss": 1.3394, "step": 26206 }, { "epoch": 4.99, "learning_rate": 1.5038894184938038e-06, "loss": 2.5233, "step": 26207 }, { "epoch": 4.99, "learning_rate": 1.5038322211630123e-06, "loss": 1.5113, "step": 26208 }, { "epoch": 4.99, "learning_rate": 1.5037750238322213e-06, "loss": 1.6035, "step": 26209 }, { "epoch": 4.99, "learning_rate": 1.50371782650143e-06, "loss": 1.9629, "step": 26210 }, { "epoch": 4.99, "learning_rate": 1.5036606291706387e-06, "loss": 1.1987, "step": 26211 }, { "epoch": 4.99, "learning_rate": 1.5036034318398475e-06, "loss": 2.3449, "step": 26212 }, { "epoch": 4.99, "learning_rate": 1.5035462345090563e-06, "loss": 2.0545, "step": 26213 }, { "epoch": 4.99, "learning_rate": 1.5034890371782649e-06, "loss": 2.2195, "step": 26214 }, { "epoch": 4.99, "learning_rate": 1.5034318398474739e-06, "loss": 1.3699, "step": 26215 }, { "epoch": 4.99, "learning_rate": 1.5033746425166825e-06, "loss": 1.36, "step": 26216 }, { "epoch": 4.99, "learning_rate": 1.5033174451858913e-06, "loss": 1.0033, "step": 26217 }, { "epoch": 4.99, "learning_rate": 1.5032602478551003e-06, "loss": 1.6358, "step": 26218 }, { "epoch": 4.99, "learning_rate": 1.5032030505243089e-06, "loss": 1.6326, "step": 26219 }, { "epoch": 4.99, "learning_rate": 1.5031458531935179e-06, "loss": 1.2513, "step": 26220 }, { "epoch": 4.99, "learning_rate": 1.5030886558627264e-06, "loss": 2.0874, "step": 26221 }, { "epoch": 4.99, "learning_rate": 1.5030314585319352e-06, "loss": 0.9674, "step": 26222 }, { "epoch": 4.99, "learning_rate": 1.502974261201144e-06, "loss": 2.2931, "step": 26223 }, { "epoch": 5.0, "learning_rate": 1.5029170638703528e-06, "loss": 1.8011, "step": 26224 }, { "epoch": 5.0, "learning_rate": 1.5028598665395614e-06, "loss": 3.2112, "step": 26225 }, { "epoch": 5.0, "learning_rate": 1.5028026692087704e-06, "loss": 2.3832, "step": 26226 }, { "epoch": 5.0, "learning_rate": 1.502745471877979e-06, "loss": 1.5292, "step": 26227 }, { "epoch": 5.0, "learning_rate": 1.5026882745471878e-06, "loss": 2.901, "step": 26228 }, { "epoch": 5.0, "learning_rate": 1.5026310772163968e-06, "loss": 2.7426, "step": 26229 }, { "epoch": 5.0, "learning_rate": 1.5025738798856054e-06, "loss": 2.5205, "step": 26230 }, { "epoch": 5.0, "learning_rate": 1.502516682554814e-06, "loss": 2.3225, "step": 26231 }, { "epoch": 5.0, "learning_rate": 1.502459485224023e-06, "loss": 3.6554, "step": 26232 }, { "epoch": 5.0, "learning_rate": 1.5024022878932315e-06, "loss": 2.7877, "step": 26233 }, { "epoch": 5.0, "learning_rate": 1.5023450905624403e-06, "loss": 2.3449, "step": 26234 }, { "epoch": 5.0, "learning_rate": 1.5022878932316493e-06, "loss": 2.0863, "step": 26235 }, { "epoch": 5.0, "learning_rate": 1.502230695900858e-06, "loss": 2.4715, "step": 26236 }, { "epoch": 5.0, "learning_rate": 1.502173498570067e-06, "loss": 2.1823, "step": 26237 }, { "epoch": 5.0, "learning_rate": 1.5021163012392755e-06, "loss": 1.7064, "step": 26238 }, { "epoch": 5.0, "learning_rate": 1.5020591039084843e-06, "loss": 2.5553, "step": 26239 }, { "epoch": 5.0, "learning_rate": 1.502001906577693e-06, "loss": 1.237, "step": 26240 }, { "epoch": 5.0, "learning_rate": 1.501944709246902e-06, "loss": 2.2211, "step": 26241 }, { "epoch": 5.0, "learning_rate": 1.5018875119161105e-06, "loss": 1.2408, "step": 26242 }, { "epoch": 5.0, "learning_rate": 1.5018303145853195e-06, "loss": 1.6262, "step": 26243 }, { "epoch": 5.0, "learning_rate": 1.501773117254528e-06, "loss": 2.9532, "step": 26244 }, { "epoch": 5.0, "learning_rate": 1.5017159199237369e-06, "loss": 2.7462, "step": 26245 }, { "epoch": 5.0, "learning_rate": 1.5016587225929459e-06, "loss": 1.8872, "step": 26246 }, { "epoch": 5.0, "learning_rate": 1.5016015252621544e-06, "loss": 2.3499, "step": 26247 }, { "epoch": 5.0, "learning_rate": 1.501544327931363e-06, "loss": 2.6947, "step": 26248 }, { "epoch": 5.0, "learning_rate": 1.501487130600572e-06, "loss": 1.7417, "step": 26249 }, { "epoch": 5.0, "learning_rate": 1.5014299332697808e-06, "loss": 1.9827, "step": 26250 }, { "epoch": 5.0, "eval_accuracy": 0.4989284183454779, "eval_loss": 1.9561748504638672, "eval_runtime": 124.8769, "eval_samples_per_second": 18.682, "eval_steps_per_second": 4.677, "step": 26250 }, { "epoch": 5.0, "learning_rate": 1.5013727359389896e-06, "loss": 1.7081, "step": 26251 }, { "epoch": 5.0, "learning_rate": 1.5013155386081984e-06, "loss": 2.5655, "step": 26252 }, { "epoch": 5.0, "learning_rate": 1.501258341277407e-06, "loss": 1.5445, "step": 26253 }, { "epoch": 5.0, "learning_rate": 1.501201143946616e-06, "loss": 2.1358, "step": 26254 }, { "epoch": 5.0, "learning_rate": 1.5011439466158246e-06, "loss": 2.7812, "step": 26255 }, { "epoch": 5.0, "learning_rate": 1.5010867492850334e-06, "loss": 1.4339, "step": 26256 }, { "epoch": 5.0, "learning_rate": 1.5010295519542424e-06, "loss": 2.096, "step": 26257 }, { "epoch": 5.0, "learning_rate": 1.500972354623451e-06, "loss": 3.2511, "step": 26258 }, { "epoch": 5.0, "learning_rate": 1.5009151572926595e-06, "loss": 0.9635, "step": 26259 }, { "epoch": 5.0, "learning_rate": 1.5008579599618686e-06, "loss": 1.5346, "step": 26260 }, { "epoch": 5.0, "learning_rate": 1.5008007626310771e-06, "loss": 2.0554, "step": 26261 }, { "epoch": 5.0, "learning_rate": 1.500743565300286e-06, "loss": 1.7309, "step": 26262 }, { "epoch": 5.0, "learning_rate": 1.500686367969495e-06, "loss": 2.7602, "step": 26263 }, { "epoch": 5.0, "learning_rate": 1.5006291706387035e-06, "loss": 3.4718, "step": 26264 }, { "epoch": 5.0, "learning_rate": 1.500571973307912e-06, "loss": 2.1763, "step": 26265 }, { "epoch": 5.0, "learning_rate": 1.500514775977121e-06, "loss": 2.002, "step": 26266 }, { "epoch": 5.0, "learning_rate": 1.50045757864633e-06, "loss": 3.0037, "step": 26267 }, { "epoch": 5.0, "learning_rate": 1.500400381315539e-06, "loss": 2.2687, "step": 26268 }, { "epoch": 5.0, "learning_rate": 1.5003431839847475e-06, "loss": 2.1294, "step": 26269 }, { "epoch": 5.0, "learning_rate": 1.500285986653956e-06, "loss": 2.8682, "step": 26270 }, { "epoch": 5.0, "learning_rate": 1.500228789323165e-06, "loss": 2.342, "step": 26271 }, { "epoch": 5.0, "learning_rate": 1.5001715919923737e-06, "loss": 1.6477, "step": 26272 }, { "epoch": 5.0, "learning_rate": 1.5001143946615825e-06, "loss": 0.8974, "step": 26273 }, { "epoch": 5.0, "learning_rate": 1.5000571973307915e-06, "loss": 1.8994, "step": 26274 }, { "epoch": 5.0, "learning_rate": 1.5e-06, "loss": 2.0593, "step": 26275 }, { "epoch": 5.0, "learning_rate": 1.4999428026692088e-06, "loss": 1.516, "step": 26276 }, { "epoch": 5.01, "learning_rate": 1.4998856053384174e-06, "loss": 2.1763, "step": 26277 }, { "epoch": 5.01, "learning_rate": 1.4998284080076264e-06, "loss": 2.4413, "step": 26278 }, { "epoch": 5.01, "learning_rate": 1.4997712106768352e-06, "loss": 3.08, "step": 26279 }, { "epoch": 5.01, "learning_rate": 1.499714013346044e-06, "loss": 2.1898, "step": 26280 }, { "epoch": 5.01, "learning_rate": 1.4996568160152526e-06, "loss": 1.7981, "step": 26281 }, { "epoch": 5.01, "learning_rate": 1.4995996186844614e-06, "loss": 2.1003, "step": 26282 }, { "epoch": 5.01, "learning_rate": 1.4995424213536702e-06, "loss": 1.73, "step": 26283 }, { "epoch": 5.01, "learning_rate": 1.499485224022879e-06, "loss": 2.2435, "step": 26284 }, { "epoch": 5.01, "learning_rate": 1.4994280266920878e-06, "loss": 1.7278, "step": 26285 }, { "epoch": 5.01, "learning_rate": 1.4993708293612966e-06, "loss": 3.1281, "step": 26286 }, { "epoch": 5.01, "learning_rate": 1.4993136320305051e-06, "loss": 2.0241, "step": 26287 }, { "epoch": 5.01, "learning_rate": 1.499256434699714e-06, "loss": 2.1003, "step": 26288 }, { "epoch": 5.01, "learning_rate": 1.499199237368923e-06, "loss": 2.3537, "step": 26289 }, { "epoch": 5.01, "learning_rate": 1.4991420400381317e-06, "loss": 1.4726, "step": 26290 }, { "epoch": 5.01, "learning_rate": 1.4990848427073403e-06, "loss": 2.0943, "step": 26291 }, { "epoch": 5.01, "learning_rate": 1.4990276453765491e-06, "loss": 1.6936, "step": 26292 }, { "epoch": 5.01, "learning_rate": 1.498970448045758e-06, "loss": 2.1501, "step": 26293 }, { "epoch": 5.01, "learning_rate": 1.4989132507149667e-06, "loss": 2.5609, "step": 26294 }, { "epoch": 5.01, "learning_rate": 1.4988560533841755e-06, "loss": 1.6333, "step": 26295 }, { "epoch": 5.01, "learning_rate": 1.4987988560533843e-06, "loss": 2.9993, "step": 26296 }, { "epoch": 5.01, "learning_rate": 1.498741658722593e-06, "loss": 2.3726, "step": 26297 }, { "epoch": 5.01, "learning_rate": 1.4986844613918017e-06, "loss": 2.7486, "step": 26298 }, { "epoch": 5.01, "learning_rate": 1.4986272640610105e-06, "loss": 2.1346, "step": 26299 }, { "epoch": 5.01, "learning_rate": 1.4985700667302192e-06, "loss": 2.3836, "step": 26300 }, { "epoch": 5.01, "learning_rate": 1.498512869399428e-06, "loss": 1.343, "step": 26301 }, { "epoch": 5.01, "learning_rate": 1.4984556720686368e-06, "loss": 3.341, "step": 26302 }, { "epoch": 5.01, "learning_rate": 1.4983984747378456e-06, "loss": 2.5058, "step": 26303 }, { "epoch": 5.01, "learning_rate": 1.4983412774070544e-06, "loss": 2.3672, "step": 26304 }, { "epoch": 5.01, "learning_rate": 1.498284080076263e-06, "loss": 1.0535, "step": 26305 }, { "epoch": 5.01, "learning_rate": 1.498226882745472e-06, "loss": 2.6402, "step": 26306 }, { "epoch": 5.01, "learning_rate": 1.4981696854146808e-06, "loss": 1.0817, "step": 26307 }, { "epoch": 5.01, "learning_rate": 1.4981124880838894e-06, "loss": 2.5645, "step": 26308 }, { "epoch": 5.01, "learning_rate": 1.4980552907530982e-06, "loss": 1.5601, "step": 26309 }, { "epoch": 5.01, "learning_rate": 1.497998093422307e-06, "loss": 2.2084, "step": 26310 }, { "epoch": 5.01, "learning_rate": 1.4979408960915158e-06, "loss": 1.9896, "step": 26311 }, { "epoch": 5.01, "learning_rate": 1.4978836987607246e-06, "loss": 0.8594, "step": 26312 }, { "epoch": 5.01, "learning_rate": 1.4978265014299334e-06, "loss": 2.0409, "step": 26313 }, { "epoch": 5.01, "learning_rate": 1.4977693040991421e-06, "loss": 1.0657, "step": 26314 }, { "epoch": 5.01, "learning_rate": 1.4977121067683507e-06, "loss": 1.5629, "step": 26315 }, { "epoch": 5.01, "learning_rate": 1.4976549094375595e-06, "loss": 2.7422, "step": 26316 }, { "epoch": 5.01, "learning_rate": 1.4975977121067685e-06, "loss": 2.4338, "step": 26317 }, { "epoch": 5.01, "learning_rate": 1.4975405147759771e-06, "loss": 2.3217, "step": 26318 }, { "epoch": 5.01, "learning_rate": 1.497483317445186e-06, "loss": 2.2108, "step": 26319 }, { "epoch": 5.01, "learning_rate": 1.4974261201143947e-06, "loss": 2.4786, "step": 26320 }, { "epoch": 5.01, "learning_rate": 1.4973689227836035e-06, "loss": 1.8414, "step": 26321 }, { "epoch": 5.01, "learning_rate": 1.4973117254528123e-06, "loss": 1.8165, "step": 26322 }, { "epoch": 5.01, "learning_rate": 1.497254528122021e-06, "loss": 1.7144, "step": 26323 }, { "epoch": 5.01, "learning_rate": 1.4971973307912299e-06, "loss": 1.5613, "step": 26324 }, { "epoch": 5.01, "learning_rate": 1.4971401334604385e-06, "loss": 1.6742, "step": 26325 }, { "epoch": 5.01, "learning_rate": 1.4970829361296473e-06, "loss": 1.7211, "step": 26326 }, { "epoch": 5.01, "learning_rate": 1.497025738798856e-06, "loss": 2.4525, "step": 26327 }, { "epoch": 5.01, "learning_rate": 1.4969685414680648e-06, "loss": 1.6383, "step": 26328 }, { "epoch": 5.02, "learning_rate": 1.4969113441372736e-06, "loss": 2.3117, "step": 26329 }, { "epoch": 5.02, "learning_rate": 1.4968541468064824e-06, "loss": 0.7384, "step": 26330 }, { "epoch": 5.02, "learning_rate": 1.4967969494756912e-06, "loss": 1.3202, "step": 26331 }, { "epoch": 5.02, "learning_rate": 1.4967397521448998e-06, "loss": 1.6024, "step": 26332 }, { "epoch": 5.02, "learning_rate": 1.4966825548141088e-06, "loss": 2.0345, "step": 26333 }, { "epoch": 5.02, "learning_rate": 1.4966253574833176e-06, "loss": 2.3448, "step": 26334 }, { "epoch": 5.02, "learning_rate": 1.4965681601525262e-06, "loss": 1.1951, "step": 26335 }, { "epoch": 5.02, "learning_rate": 1.496510962821735e-06, "loss": 2.176, "step": 26336 }, { "epoch": 5.02, "learning_rate": 1.4964537654909438e-06, "loss": 1.9893, "step": 26337 }, { "epoch": 5.02, "learning_rate": 1.4963965681601526e-06, "loss": 2.8486, "step": 26338 }, { "epoch": 5.02, "learning_rate": 1.4963393708293614e-06, "loss": 1.88, "step": 26339 }, { "epoch": 5.02, "learning_rate": 1.4962821734985702e-06, "loss": 1.8158, "step": 26340 }, { "epoch": 5.02, "learning_rate": 1.496224976167779e-06, "loss": 1.5205, "step": 26341 }, { "epoch": 5.02, "learning_rate": 1.4961677788369875e-06, "loss": 3.041, "step": 26342 }, { "epoch": 5.02, "learning_rate": 1.4961105815061963e-06, "loss": 1.3398, "step": 26343 }, { "epoch": 5.02, "learning_rate": 1.4960533841754051e-06, "loss": 1.7415, "step": 26344 }, { "epoch": 5.02, "learning_rate": 1.4959961868446141e-06, "loss": 1.6884, "step": 26345 }, { "epoch": 5.02, "learning_rate": 1.4959389895138227e-06, "loss": 2.228, "step": 26346 }, { "epoch": 5.02, "learning_rate": 1.4958817921830315e-06, "loss": 2.5385, "step": 26347 }, { "epoch": 5.02, "learning_rate": 1.4958245948522403e-06, "loss": 2.1421, "step": 26348 }, { "epoch": 5.02, "learning_rate": 1.4957673975214489e-06, "loss": 2.4659, "step": 26349 }, { "epoch": 5.02, "learning_rate": 1.4957102001906579e-06, "loss": 2.8312, "step": 26350 }, { "epoch": 5.02, "learning_rate": 1.4956530028598667e-06, "loss": 1.8738, "step": 26351 }, { "epoch": 5.02, "learning_rate": 1.4955958055290753e-06, "loss": 2.2965, "step": 26352 }, { "epoch": 5.02, "learning_rate": 1.495538608198284e-06, "loss": 3.344, "step": 26353 }, { "epoch": 5.02, "learning_rate": 1.4954814108674928e-06, "loss": 2.5221, "step": 26354 }, { "epoch": 5.02, "learning_rate": 1.4954242135367016e-06, "loss": 1.9039, "step": 26355 }, { "epoch": 5.02, "learning_rate": 1.4953670162059104e-06, "loss": 1.932, "step": 26356 }, { "epoch": 5.02, "learning_rate": 1.4953098188751192e-06, "loss": 1.6742, "step": 26357 }, { "epoch": 5.02, "learning_rate": 1.495252621544328e-06, "loss": 2.5886, "step": 26358 }, { "epoch": 5.02, "learning_rate": 1.4951954242135366e-06, "loss": 1.7027, "step": 26359 }, { "epoch": 5.02, "learning_rate": 1.4951382268827454e-06, "loss": 1.9124, "step": 26360 }, { "epoch": 5.02, "learning_rate": 1.4950810295519544e-06, "loss": 2.5932, "step": 26361 }, { "epoch": 5.02, "learning_rate": 1.4950238322211632e-06, "loss": 1.8563, "step": 26362 }, { "epoch": 5.02, "learning_rate": 1.4949666348903718e-06, "loss": 2.6051, "step": 26363 }, { "epoch": 5.02, "learning_rate": 1.4949094375595806e-06, "loss": 2.2362, "step": 26364 }, { "epoch": 5.02, "learning_rate": 1.4948522402287894e-06, "loss": 2.1306, "step": 26365 }, { "epoch": 5.02, "learning_rate": 1.4947950428979982e-06, "loss": 1.1959, "step": 26366 }, { "epoch": 5.02, "learning_rate": 1.494737845567207e-06, "loss": 1.5746, "step": 26367 }, { "epoch": 5.02, "learning_rate": 1.4946806482364157e-06, "loss": 2.2527, "step": 26368 }, { "epoch": 5.02, "learning_rate": 1.4946234509056245e-06, "loss": 2.3725, "step": 26369 }, { "epoch": 5.02, "learning_rate": 1.4945662535748331e-06, "loss": 1.3992, "step": 26370 }, { "epoch": 5.02, "learning_rate": 1.494509056244042e-06, "loss": 2.7719, "step": 26371 }, { "epoch": 5.02, "learning_rate": 1.4944518589132507e-06, "loss": 2.2856, "step": 26372 }, { "epoch": 5.02, "learning_rate": 1.4943946615824595e-06, "loss": 2.021, "step": 26373 }, { "epoch": 5.02, "learning_rate": 1.4943374642516683e-06, "loss": 1.7527, "step": 26374 }, { "epoch": 5.02, "learning_rate": 1.494280266920877e-06, "loss": 2.1882, "step": 26375 }, { "epoch": 5.02, "learning_rate": 1.4942230695900857e-06, "loss": 2.1061, "step": 26376 }, { "epoch": 5.02, "learning_rate": 1.4941658722592947e-06, "loss": 1.4463, "step": 26377 }, { "epoch": 5.02, "learning_rate": 1.4941086749285035e-06, "loss": 1.6148, "step": 26378 }, { "epoch": 5.02, "learning_rate": 1.4940514775977123e-06, "loss": 0.9639, "step": 26379 }, { "epoch": 5.02, "learning_rate": 1.4939942802669208e-06, "loss": 1.2141, "step": 26380 }, { "epoch": 5.02, "learning_rate": 1.4939370829361296e-06, "loss": 3.0139, "step": 26381 }, { "epoch": 5.03, "learning_rate": 1.4938798856053384e-06, "loss": 3.2322, "step": 26382 }, { "epoch": 5.03, "learning_rate": 1.4938226882745472e-06, "loss": 2.7883, "step": 26383 }, { "epoch": 5.03, "learning_rate": 1.493765490943756e-06, "loss": 1.3586, "step": 26384 }, { "epoch": 5.03, "learning_rate": 1.4937082936129648e-06, "loss": 2.0856, "step": 26385 }, { "epoch": 5.03, "learning_rate": 1.4936510962821736e-06, "loss": 2.4149, "step": 26386 }, { "epoch": 5.03, "learning_rate": 1.4935938989513822e-06, "loss": 1.6605, "step": 26387 }, { "epoch": 5.03, "learning_rate": 1.493536701620591e-06, "loss": 1.896, "step": 26388 }, { "epoch": 5.03, "learning_rate": 1.4934795042898e-06, "loss": 2.1081, "step": 26389 }, { "epoch": 5.03, "learning_rate": 1.4934223069590086e-06, "loss": 1.0455, "step": 26390 }, { "epoch": 5.03, "learning_rate": 1.4933651096282174e-06, "loss": 1.568, "step": 26391 }, { "epoch": 5.03, "learning_rate": 1.4933079122974262e-06, "loss": 1.7293, "step": 26392 }, { "epoch": 5.03, "learning_rate": 1.493250714966635e-06, "loss": 1.1554, "step": 26393 }, { "epoch": 5.03, "learning_rate": 1.4931935176358437e-06, "loss": 2.314, "step": 26394 }, { "epoch": 5.03, "learning_rate": 1.4931363203050525e-06, "loss": 1.3616, "step": 26395 }, { "epoch": 5.03, "learning_rate": 1.4930791229742613e-06, "loss": 2.5676, "step": 26396 }, { "epoch": 5.03, "learning_rate": 1.49302192564347e-06, "loss": 2.6467, "step": 26397 }, { "epoch": 5.03, "learning_rate": 1.4929647283126787e-06, "loss": 1.6823, "step": 26398 }, { "epoch": 5.03, "learning_rate": 1.4929075309818875e-06, "loss": 1.2556, "step": 26399 }, { "epoch": 5.03, "learning_rate": 1.4928503336510963e-06, "loss": 2.6013, "step": 26400 }, { "epoch": 5.03, "learning_rate": 1.492793136320305e-06, "loss": 1.3067, "step": 26401 }, { "epoch": 5.03, "learning_rate": 1.4927359389895139e-06, "loss": 0.8027, "step": 26402 }, { "epoch": 5.03, "learning_rate": 1.4926787416587227e-06, "loss": 2.4179, "step": 26403 }, { "epoch": 5.03, "learning_rate": 1.4926215443279313e-06, "loss": 1.5085, "step": 26404 }, { "epoch": 5.03, "learning_rate": 1.4925643469971403e-06, "loss": 1.8787, "step": 26405 }, { "epoch": 5.03, "learning_rate": 1.492507149666349e-06, "loss": 1.6594, "step": 26406 }, { "epoch": 5.03, "learning_rate": 1.4924499523355576e-06, "loss": 1.6244, "step": 26407 }, { "epoch": 5.03, "learning_rate": 1.4923927550047664e-06, "loss": 1.9963, "step": 26408 }, { "epoch": 5.03, "learning_rate": 1.4923355576739752e-06, "loss": 1.8462, "step": 26409 }, { "epoch": 5.03, "learning_rate": 1.492278360343184e-06, "loss": 1.226, "step": 26410 }, { "epoch": 5.03, "learning_rate": 1.4922211630123928e-06, "loss": 2.4002, "step": 26411 }, { "epoch": 5.03, "learning_rate": 1.4921639656816016e-06, "loss": 2.308, "step": 26412 }, { "epoch": 5.03, "learning_rate": 1.4921067683508104e-06, "loss": 1.5066, "step": 26413 }, { "epoch": 5.03, "learning_rate": 1.492049571020019e-06, "loss": 2.5818, "step": 26414 }, { "epoch": 5.03, "learning_rate": 1.4919923736892278e-06, "loss": 0.8431, "step": 26415 }, { "epoch": 5.03, "learning_rate": 1.4919351763584366e-06, "loss": 2.6032, "step": 26416 }, { "epoch": 5.03, "learning_rate": 1.4918779790276454e-06, "loss": 2.0223, "step": 26417 }, { "epoch": 5.03, "learning_rate": 1.4918207816968542e-06, "loss": 2.4349, "step": 26418 }, { "epoch": 5.03, "learning_rate": 1.491763584366063e-06, "loss": 3.1954, "step": 26419 }, { "epoch": 5.03, "learning_rate": 1.4917063870352718e-06, "loss": 2.1959, "step": 26420 }, { "epoch": 5.03, "learning_rate": 1.4916491897044805e-06, "loss": 2.3578, "step": 26421 }, { "epoch": 5.03, "learning_rate": 1.4915919923736893e-06, "loss": 1.8319, "step": 26422 }, { "epoch": 5.03, "learning_rate": 1.4915347950428981e-06, "loss": 1.4109, "step": 26423 }, { "epoch": 5.03, "learning_rate": 1.4914775977121067e-06, "loss": 0.7932, "step": 26424 }, { "epoch": 5.03, "learning_rate": 1.4914204003813155e-06, "loss": 2.8335, "step": 26425 }, { "epoch": 5.03, "learning_rate": 1.4913632030505243e-06, "loss": 1.8252, "step": 26426 }, { "epoch": 5.03, "learning_rate": 1.491306005719733e-06, "loss": 1.7221, "step": 26427 }, { "epoch": 5.03, "learning_rate": 1.4912488083889419e-06, "loss": 2.9502, "step": 26428 }, { "epoch": 5.03, "learning_rate": 1.4911916110581507e-06, "loss": 1.8886, "step": 26429 }, { "epoch": 5.03, "learning_rate": 1.4911344137273595e-06, "loss": 2.3825, "step": 26430 }, { "epoch": 5.03, "learning_rate": 1.491077216396568e-06, "loss": 1.2081, "step": 26431 }, { "epoch": 5.03, "learning_rate": 1.4910200190657769e-06, "loss": 2.4926, "step": 26432 }, { "epoch": 5.03, "learning_rate": 1.4909628217349859e-06, "loss": 1.8292, "step": 26433 }, { "epoch": 5.04, "learning_rate": 1.4909056244041947e-06, "loss": 2.578, "step": 26434 }, { "epoch": 5.04, "learning_rate": 1.4908484270734032e-06, "loss": 2.2989, "step": 26435 }, { "epoch": 5.04, "learning_rate": 1.490791229742612e-06, "loss": 3.0294, "step": 26436 }, { "epoch": 5.04, "learning_rate": 1.4907340324118208e-06, "loss": 4.247, "step": 26437 }, { "epoch": 5.04, "learning_rate": 1.4906768350810296e-06, "loss": 1.857, "step": 26438 }, { "epoch": 5.04, "learning_rate": 1.4906196377502384e-06, "loss": 1.7, "step": 26439 }, { "epoch": 5.04, "learning_rate": 1.4905624404194472e-06, "loss": 1.8129, "step": 26440 }, { "epoch": 5.04, "learning_rate": 1.4905052430886558e-06, "loss": 1.5179, "step": 26441 }, { "epoch": 5.04, "learning_rate": 1.4904480457578646e-06, "loss": 2.8241, "step": 26442 }, { "epoch": 5.04, "learning_rate": 1.4903908484270734e-06, "loss": 1.1729, "step": 26443 }, { "epoch": 5.04, "learning_rate": 1.4903336510962824e-06, "loss": 1.6603, "step": 26444 }, { "epoch": 5.04, "learning_rate": 1.490276453765491e-06, "loss": 2.7695, "step": 26445 }, { "epoch": 5.04, "learning_rate": 1.4902192564346998e-06, "loss": 1.2236, "step": 26446 }, { "epoch": 5.04, "learning_rate": 1.4901620591039086e-06, "loss": 1.9076, "step": 26447 }, { "epoch": 5.04, "learning_rate": 1.4901048617731171e-06, "loss": 2.7535, "step": 26448 }, { "epoch": 5.04, "learning_rate": 1.4900476644423261e-06, "loss": 2.0347, "step": 26449 }, { "epoch": 5.04, "learning_rate": 1.489990467111535e-06, "loss": 1.8543, "step": 26450 }, { "epoch": 5.04, "learning_rate": 1.4899332697807437e-06, "loss": 2.105, "step": 26451 }, { "epoch": 5.04, "learning_rate": 1.4898760724499523e-06, "loss": 1.4324, "step": 26452 }, { "epoch": 5.04, "learning_rate": 1.489818875119161e-06, "loss": 2.3903, "step": 26453 }, { "epoch": 5.04, "learning_rate": 1.48976167778837e-06, "loss": 2.529, "step": 26454 }, { "epoch": 5.04, "learning_rate": 1.4897044804575787e-06, "loss": 2.1647, "step": 26455 }, { "epoch": 5.04, "learning_rate": 1.4896472831267875e-06, "loss": 1.6754, "step": 26456 }, { "epoch": 5.04, "learning_rate": 1.4895900857959963e-06, "loss": 1.4908, "step": 26457 }, { "epoch": 5.04, "learning_rate": 1.4895328884652049e-06, "loss": 1.4245, "step": 26458 }, { "epoch": 5.04, "learning_rate": 1.4894756911344137e-06, "loss": 1.5182, "step": 26459 }, { "epoch": 5.04, "learning_rate": 1.4894184938036224e-06, "loss": 1.9296, "step": 26460 }, { "epoch": 5.04, "learning_rate": 1.4893612964728315e-06, "loss": 2.3811, "step": 26461 }, { "epoch": 5.04, "learning_rate": 1.48930409914204e-06, "loss": 1.3745, "step": 26462 }, { "epoch": 5.04, "learning_rate": 1.4892469018112488e-06, "loss": 3.1382, "step": 26463 }, { "epoch": 5.04, "learning_rate": 1.4891897044804576e-06, "loss": 2.7234, "step": 26464 }, { "epoch": 5.04, "learning_rate": 1.4891325071496664e-06, "loss": 1.6978, "step": 26465 }, { "epoch": 5.04, "learning_rate": 1.4890753098188752e-06, "loss": 3.0785, "step": 26466 }, { "epoch": 5.04, "learning_rate": 1.489018112488084e-06, "loss": 1.4899, "step": 26467 }, { "epoch": 5.04, "learning_rate": 1.4889609151572928e-06, "loss": 3.436, "step": 26468 }, { "epoch": 5.04, "learning_rate": 1.4889037178265014e-06, "loss": 3.4299, "step": 26469 }, { "epoch": 5.04, "learning_rate": 1.4888465204957102e-06, "loss": 0.8677, "step": 26470 }, { "epoch": 5.04, "learning_rate": 1.488789323164919e-06, "loss": 1.3765, "step": 26471 }, { "epoch": 5.04, "learning_rate": 1.4887321258341278e-06, "loss": 2.7489, "step": 26472 }, { "epoch": 5.04, "learning_rate": 1.4886749285033366e-06, "loss": 1.7278, "step": 26473 }, { "epoch": 5.04, "learning_rate": 1.4886177311725453e-06, "loss": 1.8332, "step": 26474 }, { "epoch": 5.04, "learning_rate": 1.4885605338417541e-06, "loss": 1.4154, "step": 26475 }, { "epoch": 5.04, "learning_rate": 1.4885033365109627e-06, "loss": 1.0732, "step": 26476 }, { "epoch": 5.04, "learning_rate": 1.4884461391801717e-06, "loss": 1.0383, "step": 26477 }, { "epoch": 5.04, "learning_rate": 1.4883889418493805e-06, "loss": 2.8659, "step": 26478 }, { "epoch": 5.04, "learning_rate": 1.4883317445185891e-06, "loss": 0.5125, "step": 26479 }, { "epoch": 5.04, "learning_rate": 1.488274547187798e-06, "loss": 1.9911, "step": 26480 }, { "epoch": 5.04, "learning_rate": 1.4882173498570067e-06, "loss": 3.4468, "step": 26481 }, { "epoch": 5.04, "learning_rate": 1.4881601525262155e-06, "loss": 1.3379, "step": 26482 }, { "epoch": 5.04, "learning_rate": 1.4881029551954243e-06, "loss": 2.495, "step": 26483 }, { "epoch": 5.04, "learning_rate": 1.488045757864633e-06, "loss": 2.4659, "step": 26484 }, { "epoch": 5.04, "learning_rate": 1.4879885605338419e-06, "loss": 1.9343, "step": 26485 }, { "epoch": 5.04, "learning_rate": 1.4879313632030505e-06, "loss": 2.1425, "step": 26486 }, { "epoch": 5.05, "learning_rate": 1.4878741658722592e-06, "loss": 1.1251, "step": 26487 }, { "epoch": 5.05, "learning_rate": 1.4878169685414683e-06, "loss": 2.2875, "step": 26488 }, { "epoch": 5.05, "learning_rate": 1.4877597712106768e-06, "loss": 1.6512, "step": 26489 }, { "epoch": 5.05, "learning_rate": 1.4877025738798856e-06, "loss": 1.5263, "step": 26490 }, { "epoch": 5.05, "learning_rate": 1.4876453765490944e-06, "loss": 2.8749, "step": 26491 }, { "epoch": 5.05, "learning_rate": 1.4875881792183032e-06, "loss": 1.0255, "step": 26492 }, { "epoch": 5.05, "learning_rate": 1.487530981887512e-06, "loss": 1.6807, "step": 26493 }, { "epoch": 5.05, "learning_rate": 1.4874737845567208e-06, "loss": 1.487, "step": 26494 }, { "epoch": 5.05, "learning_rate": 1.4874165872259296e-06, "loss": 2.7496, "step": 26495 }, { "epoch": 5.05, "learning_rate": 1.4873593898951382e-06, "loss": 2.6178, "step": 26496 }, { "epoch": 5.05, "learning_rate": 1.487302192564347e-06, "loss": 1.5889, "step": 26497 }, { "epoch": 5.05, "learning_rate": 1.4872449952335558e-06, "loss": 1.2477, "step": 26498 }, { "epoch": 5.05, "learning_rate": 1.4871877979027646e-06, "loss": 1.1291, "step": 26499 }, { "epoch": 5.05, "learning_rate": 1.4871306005719734e-06, "loss": 3.1533, "step": 26500 }, { "epoch": 5.05, "learning_rate": 1.4870734032411821e-06, "loss": 1.988, "step": 26501 }, { "epoch": 5.05, "learning_rate": 1.487016205910391e-06, "loss": 2.3131, "step": 26502 }, { "epoch": 5.05, "learning_rate": 1.4869590085795995e-06, "loss": 2.7337, "step": 26503 }, { "epoch": 5.05, "learning_rate": 1.4869018112488083e-06, "loss": 2.3713, "step": 26504 }, { "epoch": 5.05, "learning_rate": 1.4868446139180173e-06, "loss": 2.7126, "step": 26505 }, { "epoch": 5.05, "learning_rate": 1.486787416587226e-06, "loss": 2.3341, "step": 26506 }, { "epoch": 5.05, "learning_rate": 1.4867302192564347e-06, "loss": 2.0642, "step": 26507 }, { "epoch": 5.05, "learning_rate": 1.4866730219256435e-06, "loss": 2.2238, "step": 26508 }, { "epoch": 5.05, "learning_rate": 1.4866158245948523e-06, "loss": 1.7436, "step": 26509 }, { "epoch": 5.05, "learning_rate": 1.486558627264061e-06, "loss": 2.0819, "step": 26510 }, { "epoch": 5.05, "learning_rate": 1.4865014299332699e-06, "loss": 0.6906, "step": 26511 }, { "epoch": 5.05, "learning_rate": 1.4864442326024787e-06, "loss": 1.9848, "step": 26512 }, { "epoch": 5.05, "learning_rate": 1.4863870352716872e-06, "loss": 3.7793, "step": 26513 }, { "epoch": 5.05, "learning_rate": 1.486329837940896e-06, "loss": 1.7144, "step": 26514 }, { "epoch": 5.05, "learning_rate": 1.4862726406101048e-06, "loss": 1.8695, "step": 26515 }, { "epoch": 5.05, "learning_rate": 1.4862154432793138e-06, "loss": 1.8055, "step": 26516 }, { "epoch": 5.05, "learning_rate": 1.4861582459485224e-06, "loss": 2.6667, "step": 26517 }, { "epoch": 5.05, "learning_rate": 1.4861010486177312e-06, "loss": 1.2198, "step": 26518 }, { "epoch": 5.05, "learning_rate": 1.48604385128694e-06, "loss": 2.7735, "step": 26519 }, { "epoch": 5.05, "learning_rate": 1.4859866539561486e-06, "loss": 1.7362, "step": 26520 }, { "epoch": 5.05, "learning_rate": 1.4859294566253576e-06, "loss": 1.4887, "step": 26521 }, { "epoch": 5.05, "learning_rate": 1.4858722592945664e-06, "loss": 2.1496, "step": 26522 }, { "epoch": 5.05, "learning_rate": 1.485815061963775e-06, "loss": 2.2211, "step": 26523 }, { "epoch": 5.05, "learning_rate": 1.4857578646329838e-06, "loss": 1.4912, "step": 26524 }, { "epoch": 5.05, "learning_rate": 1.4857006673021926e-06, "loss": 1.3362, "step": 26525 }, { "epoch": 5.05, "learning_rate": 1.4856434699714014e-06, "loss": 0.9563, "step": 26526 }, { "epoch": 5.05, "learning_rate": 1.4855862726406102e-06, "loss": 2.1333, "step": 26527 }, { "epoch": 5.05, "learning_rate": 1.485529075309819e-06, "loss": 1.963, "step": 26528 }, { "epoch": 5.05, "learning_rate": 1.4854718779790277e-06, "loss": 1.4002, "step": 26529 }, { "epoch": 5.05, "learning_rate": 1.4854146806482363e-06, "loss": 1.6806, "step": 26530 }, { "epoch": 5.05, "learning_rate": 1.4853574833174451e-06, "loss": 1.1748, "step": 26531 }, { "epoch": 5.05, "learning_rate": 1.4853002859866541e-06, "loss": 0.5452, "step": 26532 }, { "epoch": 5.05, "learning_rate": 1.485243088655863e-06, "loss": 2.0987, "step": 26533 }, { "epoch": 5.05, "learning_rate": 1.4851858913250715e-06, "loss": 2.2338, "step": 26534 }, { "epoch": 5.05, "learning_rate": 1.4851286939942803e-06, "loss": 1.4578, "step": 26535 }, { "epoch": 5.05, "learning_rate": 1.485071496663489e-06, "loss": 2.4334, "step": 26536 }, { "epoch": 5.05, "learning_rate": 1.4850142993326979e-06, "loss": 1.8455, "step": 26537 }, { "epoch": 5.05, "learning_rate": 1.4849571020019067e-06, "loss": 2.4847, "step": 26538 }, { "epoch": 5.06, "learning_rate": 1.4848999046711155e-06, "loss": 2.2264, "step": 26539 }, { "epoch": 5.06, "learning_rate": 1.4848427073403243e-06, "loss": 1.5389, "step": 26540 }, { "epoch": 5.06, "learning_rate": 1.4847855100095328e-06, "loss": 1.2927, "step": 26541 }, { "epoch": 5.06, "learning_rate": 1.4847283126787416e-06, "loss": 0.6307, "step": 26542 }, { "epoch": 5.06, "learning_rate": 1.4846711153479504e-06, "loss": 1.9508, "step": 26543 }, { "epoch": 5.06, "learning_rate": 1.4846139180171592e-06, "loss": 2.1251, "step": 26544 }, { "epoch": 5.06, "learning_rate": 1.484556720686368e-06, "loss": 1.7349, "step": 26545 }, { "epoch": 5.06, "learning_rate": 1.4844995233555768e-06, "loss": 1.9875, "step": 26546 }, { "epoch": 5.06, "learning_rate": 1.4844423260247854e-06, "loss": 2.1049, "step": 26547 }, { "epoch": 5.06, "learning_rate": 1.4843851286939942e-06, "loss": 3.2119, "step": 26548 }, { "epoch": 5.06, "learning_rate": 1.4843279313632032e-06, "loss": 1.4235, "step": 26549 }, { "epoch": 5.06, "learning_rate": 1.484270734032412e-06, "loss": 1.2652, "step": 26550 }, { "epoch": 5.06, "learning_rate": 1.4842135367016206e-06, "loss": 3.0568, "step": 26551 }, { "epoch": 5.06, "learning_rate": 1.4841563393708294e-06, "loss": 2.7616, "step": 26552 }, { "epoch": 5.06, "learning_rate": 1.4840991420400382e-06, "loss": 2.0609, "step": 26553 }, { "epoch": 5.06, "learning_rate": 1.484041944709247e-06, "loss": 1.2693, "step": 26554 }, { "epoch": 5.06, "learning_rate": 1.4839847473784557e-06, "loss": 2.4594, "step": 26555 }, { "epoch": 5.06, "learning_rate": 1.4839275500476645e-06, "loss": 2.5656, "step": 26556 }, { "epoch": 5.06, "learning_rate": 1.4838703527168733e-06, "loss": 1.5152, "step": 26557 }, { "epoch": 5.06, "learning_rate": 1.483813155386082e-06, "loss": 1.85, "step": 26558 }, { "epoch": 5.06, "learning_rate": 1.4837559580552907e-06, "loss": 0.5506, "step": 26559 }, { "epoch": 5.06, "learning_rate": 1.4836987607244997e-06, "loss": 1.2399, "step": 26560 }, { "epoch": 5.06, "learning_rate": 1.4836415633937083e-06, "loss": 1.6005, "step": 26561 }, { "epoch": 5.06, "learning_rate": 1.483584366062917e-06, "loss": 2.2599, "step": 26562 }, { "epoch": 5.06, "learning_rate": 1.4835271687321259e-06, "loss": 1.8719, "step": 26563 }, { "epoch": 5.06, "learning_rate": 1.4834699714013347e-06, "loss": 2.8587, "step": 26564 }, { "epoch": 5.06, "learning_rate": 1.4834127740705435e-06, "loss": 1.712, "step": 26565 }, { "epoch": 5.06, "learning_rate": 1.4833555767397523e-06, "loss": 2.639, "step": 26566 }, { "epoch": 5.06, "learning_rate": 1.483298379408961e-06, "loss": 1.2465, "step": 26567 }, { "epoch": 5.06, "learning_rate": 1.4832411820781696e-06, "loss": 1.4887, "step": 26568 }, { "epoch": 5.06, "learning_rate": 1.4831839847473784e-06, "loss": 1.4958, "step": 26569 }, { "epoch": 5.06, "learning_rate": 1.4831267874165872e-06, "loss": 2.0949, "step": 26570 }, { "epoch": 5.06, "learning_rate": 1.483069590085796e-06, "loss": 2.3022, "step": 26571 }, { "epoch": 5.06, "learning_rate": 1.4830123927550048e-06, "loss": 1.6612, "step": 26572 }, { "epoch": 5.06, "learning_rate": 1.4829551954242136e-06, "loss": 2.6954, "step": 26573 }, { "epoch": 5.06, "learning_rate": 1.4828979980934224e-06, "loss": 2.1655, "step": 26574 }, { "epoch": 5.06, "learning_rate": 1.482840800762631e-06, "loss": 1.3428, "step": 26575 }, { "epoch": 5.06, "learning_rate": 1.48278360343184e-06, "loss": 2.0451, "step": 26576 }, { "epoch": 5.06, "learning_rate": 1.4827264061010488e-06, "loss": 2.7193, "step": 26577 }, { "epoch": 5.06, "learning_rate": 1.4826692087702574e-06, "loss": 0.9162, "step": 26578 }, { "epoch": 5.06, "learning_rate": 1.4826120114394662e-06, "loss": 1.2484, "step": 26579 }, { "epoch": 5.06, "learning_rate": 1.482554814108675e-06, "loss": 1.7389, "step": 26580 }, { "epoch": 5.06, "learning_rate": 1.4824976167778837e-06, "loss": 1.3951, "step": 26581 }, { "epoch": 5.06, "learning_rate": 1.4824404194470925e-06, "loss": 2.4026, "step": 26582 }, { "epoch": 5.06, "learning_rate": 1.4823832221163013e-06, "loss": 1.1504, "step": 26583 }, { "epoch": 5.06, "learning_rate": 1.4823260247855101e-06, "loss": 2.1438, "step": 26584 }, { "epoch": 5.06, "learning_rate": 1.4822688274547187e-06, "loss": 1.0326, "step": 26585 }, { "epoch": 5.06, "learning_rate": 1.4822116301239275e-06, "loss": 1.4503, "step": 26586 }, { "epoch": 5.06, "learning_rate": 1.4821544327931363e-06, "loss": 1.3286, "step": 26587 }, { "epoch": 5.06, "learning_rate": 1.482097235462345e-06, "loss": 2.436, "step": 26588 }, { "epoch": 5.06, "learning_rate": 1.4820400381315539e-06, "loss": 1.8356, "step": 26589 }, { "epoch": 5.06, "learning_rate": 1.4819828408007627e-06, "loss": 2.1352, "step": 26590 }, { "epoch": 5.06, "learning_rate": 1.4819256434699715e-06, "loss": 2.0076, "step": 26591 }, { "epoch": 5.07, "learning_rate": 1.48186844613918e-06, "loss": 1.813, "step": 26592 }, { "epoch": 5.07, "learning_rate": 1.481811248808389e-06, "loss": 1.8162, "step": 26593 }, { "epoch": 5.07, "learning_rate": 1.4817540514775979e-06, "loss": 2.0826, "step": 26594 }, { "epoch": 5.07, "learning_rate": 1.4816968541468064e-06, "loss": 2.7297, "step": 26595 }, { "epoch": 5.07, "learning_rate": 1.4816396568160152e-06, "loss": 1.6028, "step": 26596 }, { "epoch": 5.07, "learning_rate": 1.481582459485224e-06, "loss": 1.6551, "step": 26597 }, { "epoch": 5.07, "learning_rate": 1.4815252621544328e-06, "loss": 2.6813, "step": 26598 }, { "epoch": 5.07, "learning_rate": 1.4814680648236416e-06, "loss": 1.2153, "step": 26599 }, { "epoch": 5.07, "learning_rate": 1.4814108674928504e-06, "loss": 1.4136, "step": 26600 }, { "epoch": 5.07, "learning_rate": 1.4813536701620592e-06, "loss": 1.2685, "step": 26601 }, { "epoch": 5.07, "learning_rate": 1.4812964728312678e-06, "loss": 2.1139, "step": 26602 }, { "epoch": 5.07, "learning_rate": 1.4812392755004766e-06, "loss": 1.4565, "step": 26603 }, { "epoch": 5.07, "learning_rate": 1.4811820781696856e-06, "loss": 2.5746, "step": 26604 }, { "epoch": 5.07, "learning_rate": 1.4811248808388944e-06, "loss": 2.5504, "step": 26605 }, { "epoch": 5.07, "learning_rate": 1.481067683508103e-06, "loss": 1.4978, "step": 26606 }, { "epoch": 5.07, "learning_rate": 1.4810104861773118e-06, "loss": 2.034, "step": 26607 }, { "epoch": 5.07, "learning_rate": 1.4809532888465205e-06, "loss": 0.9014, "step": 26608 }, { "epoch": 5.07, "learning_rate": 1.4808960915157293e-06, "loss": 1.8025, "step": 26609 }, { "epoch": 5.07, "learning_rate": 1.4808388941849381e-06, "loss": 0.9227, "step": 26610 }, { "epoch": 5.07, "learning_rate": 1.480781696854147e-06, "loss": 2.9954, "step": 26611 }, { "epoch": 5.07, "learning_rate": 1.4807244995233555e-06, "loss": 0.7078, "step": 26612 }, { "epoch": 5.07, "learning_rate": 1.4806673021925643e-06, "loss": 2.264, "step": 26613 }, { "epoch": 5.07, "learning_rate": 1.480610104861773e-06, "loss": 2.0398, "step": 26614 }, { "epoch": 5.07, "learning_rate": 1.4805529075309819e-06, "loss": 1.4031, "step": 26615 }, { "epoch": 5.07, "learning_rate": 1.4804957102001907e-06, "loss": 2.3965, "step": 26616 }, { "epoch": 5.07, "learning_rate": 1.4804385128693995e-06, "loss": 2.6855, "step": 26617 }, { "epoch": 5.07, "learning_rate": 1.4803813155386083e-06, "loss": 1.0582, "step": 26618 }, { "epoch": 5.07, "learning_rate": 1.4803241182078169e-06, "loss": 1.8374, "step": 26619 }, { "epoch": 5.07, "learning_rate": 1.4802669208770259e-06, "loss": 1.5766, "step": 26620 }, { "epoch": 5.07, "learning_rate": 1.4802097235462347e-06, "loss": 1.3918, "step": 26621 }, { "epoch": 5.07, "learning_rate": 1.4801525262154434e-06, "loss": 2.6886, "step": 26622 }, { "epoch": 5.07, "learning_rate": 1.480095328884652e-06, "loss": 1.6243, "step": 26623 }, { "epoch": 5.07, "learning_rate": 1.4800381315538608e-06, "loss": 2.2226, "step": 26624 }, { "epoch": 5.07, "learning_rate": 1.4799809342230696e-06, "loss": 1.7833, "step": 26625 }, { "epoch": 5.07, "learning_rate": 1.4799237368922784e-06, "loss": 2.8282, "step": 26626 }, { "epoch": 5.07, "learning_rate": 1.4798665395614872e-06, "loss": 1.5186, "step": 26627 }, { "epoch": 5.07, "learning_rate": 1.479809342230696e-06, "loss": 1.806, "step": 26628 }, { "epoch": 5.07, "learning_rate": 1.4797521448999048e-06, "loss": 3.0162, "step": 26629 }, { "epoch": 5.07, "learning_rate": 1.4796949475691134e-06, "loss": 1.6867, "step": 26630 }, { "epoch": 5.07, "learning_rate": 1.4796377502383222e-06, "loss": 2.7693, "step": 26631 }, { "epoch": 5.07, "learning_rate": 1.4795805529075312e-06, "loss": 2.7079, "step": 26632 }, { "epoch": 5.07, "learning_rate": 1.4795233555767398e-06, "loss": 2.4571, "step": 26633 }, { "epoch": 5.07, "learning_rate": 1.4794661582459485e-06, "loss": 3.2267, "step": 26634 }, { "epoch": 5.07, "learning_rate": 1.4794089609151573e-06, "loss": 2.5882, "step": 26635 }, { "epoch": 5.07, "learning_rate": 1.479351763584366e-06, "loss": 1.7857, "step": 26636 }, { "epoch": 5.07, "learning_rate": 1.479294566253575e-06, "loss": 2.3966, "step": 26637 }, { "epoch": 5.07, "learning_rate": 1.4792373689227837e-06, "loss": 2.172, "step": 26638 }, { "epoch": 5.07, "learning_rate": 1.4791801715919925e-06, "loss": 0.9872, "step": 26639 }, { "epoch": 5.07, "learning_rate": 1.479122974261201e-06, "loss": 2.183, "step": 26640 }, { "epoch": 5.07, "learning_rate": 1.4790657769304099e-06, "loss": 1.2596, "step": 26641 }, { "epoch": 5.07, "learning_rate": 1.4790085795996187e-06, "loss": 1.3259, "step": 26642 }, { "epoch": 5.07, "learning_rate": 1.4789513822688275e-06, "loss": 1.212, "step": 26643 }, { "epoch": 5.08, "learning_rate": 1.4788941849380363e-06, "loss": 1.1807, "step": 26644 }, { "epoch": 5.08, "learning_rate": 1.478836987607245e-06, "loss": 2.2054, "step": 26645 }, { "epoch": 5.08, "learning_rate": 1.4787797902764539e-06, "loss": 2.7986, "step": 26646 }, { "epoch": 5.08, "learning_rate": 1.4787225929456624e-06, "loss": 2.6339, "step": 26647 }, { "epoch": 5.08, "learning_rate": 1.4786653956148714e-06, "loss": 1.557, "step": 26648 }, { "epoch": 5.08, "learning_rate": 1.4786081982840802e-06, "loss": 1.9374, "step": 26649 }, { "epoch": 5.08, "learning_rate": 1.4785510009532888e-06, "loss": 1.9783, "step": 26650 }, { "epoch": 5.08, "learning_rate": 1.4784938036224976e-06, "loss": 2.0313, "step": 26651 }, { "epoch": 5.08, "learning_rate": 1.4784366062917064e-06, "loss": 3.0773, "step": 26652 }, { "epoch": 5.08, "learning_rate": 1.4783794089609152e-06, "loss": 2.1622, "step": 26653 }, { "epoch": 5.08, "learning_rate": 1.478322211630124e-06, "loss": 3.0613, "step": 26654 }, { "epoch": 5.08, "learning_rate": 1.4782650142993328e-06, "loss": 2.475, "step": 26655 }, { "epoch": 5.08, "learning_rate": 1.4782078169685416e-06, "loss": 2.1751, "step": 26656 }, { "epoch": 5.08, "learning_rate": 1.4781506196377502e-06, "loss": 1.3152, "step": 26657 }, { "epoch": 5.08, "learning_rate": 1.478093422306959e-06, "loss": 3.362, "step": 26658 }, { "epoch": 5.08, "learning_rate": 1.4780362249761678e-06, "loss": 0.962, "step": 26659 }, { "epoch": 5.08, "learning_rate": 1.4779790276453766e-06, "loss": 1.8675, "step": 26660 }, { "epoch": 5.08, "learning_rate": 1.4779218303145853e-06, "loss": 2.0916, "step": 26661 }, { "epoch": 5.08, "learning_rate": 1.4778646329837941e-06, "loss": 1.7628, "step": 26662 }, { "epoch": 5.08, "learning_rate": 1.477807435653003e-06, "loss": 1.2803, "step": 26663 }, { "epoch": 5.08, "learning_rate": 1.4777502383222117e-06, "loss": 1.4101, "step": 26664 }, { "epoch": 5.08, "learning_rate": 1.4776930409914205e-06, "loss": 2.8488, "step": 26665 }, { "epoch": 5.08, "learning_rate": 1.4776358436606293e-06, "loss": 2.5532, "step": 26666 }, { "epoch": 5.08, "learning_rate": 1.477578646329838e-06, "loss": 2.3454, "step": 26667 }, { "epoch": 5.08, "learning_rate": 1.4775214489990467e-06, "loss": 2.1603, "step": 26668 }, { "epoch": 5.08, "learning_rate": 1.4774642516682555e-06, "loss": 2.2231, "step": 26669 }, { "epoch": 5.08, "learning_rate": 1.4774070543374643e-06, "loss": 2.9607, "step": 26670 }, { "epoch": 5.08, "learning_rate": 1.477349857006673e-06, "loss": 1.2552, "step": 26671 }, { "epoch": 5.08, "learning_rate": 1.4772926596758819e-06, "loss": 1.5733, "step": 26672 }, { "epoch": 5.08, "learning_rate": 1.4772354623450907e-06, "loss": 2.6757, "step": 26673 }, { "epoch": 5.08, "learning_rate": 1.4771782650142992e-06, "loss": 2.0169, "step": 26674 }, { "epoch": 5.08, "learning_rate": 1.477121067683508e-06, "loss": 2.7873, "step": 26675 }, { "epoch": 5.08, "learning_rate": 1.477063870352717e-06, "loss": 2.4118, "step": 26676 }, { "epoch": 5.08, "learning_rate": 1.4770066730219256e-06, "loss": 2.1554, "step": 26677 }, { "epoch": 5.08, "learning_rate": 1.4769494756911344e-06, "loss": 3.0295, "step": 26678 }, { "epoch": 5.08, "learning_rate": 1.4768922783603432e-06, "loss": 2.5442, "step": 26679 }, { "epoch": 5.08, "learning_rate": 1.476835081029552e-06, "loss": 2.3261, "step": 26680 }, { "epoch": 5.08, "learning_rate": 1.4767778836987608e-06, "loss": 1.9338, "step": 26681 }, { "epoch": 5.08, "learning_rate": 1.4767206863679696e-06, "loss": 2.5335, "step": 26682 }, { "epoch": 5.08, "learning_rate": 1.4766634890371784e-06, "loss": 1.6162, "step": 26683 }, { "epoch": 5.08, "learning_rate": 1.476606291706387e-06, "loss": 1.3571, "step": 26684 }, { "epoch": 5.08, "learning_rate": 1.4765490943755958e-06, "loss": 1.8773, "step": 26685 }, { "epoch": 5.08, "learning_rate": 1.4764918970448046e-06, "loss": 2.0661, "step": 26686 }, { "epoch": 5.08, "learning_rate": 1.4764346997140136e-06, "loss": 1.6895, "step": 26687 }, { "epoch": 5.08, "learning_rate": 1.4763775023832221e-06, "loss": 2.0038, "step": 26688 }, { "epoch": 5.08, "learning_rate": 1.476320305052431e-06, "loss": 2.2225, "step": 26689 }, { "epoch": 5.08, "learning_rate": 1.4762631077216397e-06, "loss": 1.0249, "step": 26690 }, { "epoch": 5.08, "learning_rate": 1.4762059103908483e-06, "loss": 2.0502, "step": 26691 }, { "epoch": 5.08, "learning_rate": 1.4761487130600573e-06, "loss": 2.2209, "step": 26692 }, { "epoch": 5.08, "learning_rate": 1.4760915157292661e-06, "loss": 2.7823, "step": 26693 }, { "epoch": 5.08, "learning_rate": 1.476034318398475e-06, "loss": 1.8015, "step": 26694 }, { "epoch": 5.08, "learning_rate": 1.4759771210676835e-06, "loss": 2.1711, "step": 26695 }, { "epoch": 5.08, "learning_rate": 1.4759199237368923e-06, "loss": 0.9595, "step": 26696 }, { "epoch": 5.09, "learning_rate": 1.475862726406101e-06, "loss": 1.764, "step": 26697 }, { "epoch": 5.09, "learning_rate": 1.4758055290753099e-06, "loss": 1.8175, "step": 26698 }, { "epoch": 5.09, "learning_rate": 1.4757483317445187e-06, "loss": 1.5811, "step": 26699 }, { "epoch": 5.09, "learning_rate": 1.4756911344137275e-06, "loss": 1.8972, "step": 26700 }, { "epoch": 5.09, "learning_rate": 1.475633937082936e-06, "loss": 2.7856, "step": 26701 }, { "epoch": 5.09, "learning_rate": 1.4755767397521448e-06, "loss": 1.7762, "step": 26702 }, { "epoch": 5.09, "learning_rate": 1.4755195424213536e-06, "loss": 2.7274, "step": 26703 }, { "epoch": 5.09, "learning_rate": 1.4754623450905626e-06, "loss": 2.1435, "step": 26704 }, { "epoch": 5.09, "learning_rate": 1.4754051477597712e-06, "loss": 2.5833, "step": 26705 }, { "epoch": 5.09, "learning_rate": 1.47534795042898e-06, "loss": 3.0798, "step": 26706 }, { "epoch": 5.09, "learning_rate": 1.4752907530981888e-06, "loss": 2.5887, "step": 26707 }, { "epoch": 5.09, "learning_rate": 1.4752335557673974e-06, "loss": 2.3365, "step": 26708 }, { "epoch": 5.09, "learning_rate": 1.4751763584366064e-06, "loss": 1.8556, "step": 26709 }, { "epoch": 5.09, "learning_rate": 1.4751191611058152e-06, "loss": 2.1009, "step": 26710 }, { "epoch": 5.09, "learning_rate": 1.475061963775024e-06, "loss": 1.7726, "step": 26711 }, { "epoch": 5.09, "learning_rate": 1.4750047664442326e-06, "loss": 0.9017, "step": 26712 }, { "epoch": 5.09, "learning_rate": 1.4749475691134414e-06, "loss": 1.8779, "step": 26713 }, { "epoch": 5.09, "learning_rate": 1.4748903717826501e-06, "loss": 1.3545, "step": 26714 }, { "epoch": 5.09, "learning_rate": 1.474833174451859e-06, "loss": 2.2848, "step": 26715 }, { "epoch": 5.09, "learning_rate": 1.4747759771210677e-06, "loss": 1.0617, "step": 26716 }, { "epoch": 5.09, "learning_rate": 1.4747187797902765e-06, "loss": 1.9081, "step": 26717 }, { "epoch": 5.09, "learning_rate": 1.4746615824594851e-06, "loss": 2.2278, "step": 26718 }, { "epoch": 5.09, "learning_rate": 1.474604385128694e-06, "loss": 1.8274, "step": 26719 }, { "epoch": 5.09, "learning_rate": 1.474547187797903e-06, "loss": 1.7559, "step": 26720 }, { "epoch": 5.09, "learning_rate": 1.4744899904671117e-06, "loss": 1.1143, "step": 26721 }, { "epoch": 5.09, "learning_rate": 1.4744327931363203e-06, "loss": 2.3761, "step": 26722 }, { "epoch": 5.09, "learning_rate": 1.474375595805529e-06, "loss": 1.9868, "step": 26723 }, { "epoch": 5.09, "learning_rate": 1.4743183984747379e-06, "loss": 1.9253, "step": 26724 }, { "epoch": 5.09, "learning_rate": 1.4742612011439467e-06, "loss": 2.6337, "step": 26725 }, { "epoch": 5.09, "learning_rate": 1.4742040038131555e-06, "loss": 0.7945, "step": 26726 }, { "epoch": 5.09, "learning_rate": 1.4741468064823643e-06, "loss": 1.653, "step": 26727 }, { "epoch": 5.09, "learning_rate": 1.474089609151573e-06, "loss": 2.823, "step": 26728 }, { "epoch": 5.09, "learning_rate": 1.4740324118207816e-06, "loss": 1.9567, "step": 26729 }, { "epoch": 5.09, "learning_rate": 1.4739752144899904e-06, "loss": 2.7577, "step": 26730 }, { "epoch": 5.09, "learning_rate": 1.4739180171591994e-06, "loss": 1.7691, "step": 26731 }, { "epoch": 5.09, "learning_rate": 1.473860819828408e-06, "loss": 1.9236, "step": 26732 }, { "epoch": 5.09, "learning_rate": 1.4738036224976168e-06, "loss": 1.653, "step": 26733 }, { "epoch": 5.09, "learning_rate": 1.4737464251668256e-06, "loss": 1.9, "step": 26734 }, { "epoch": 5.09, "learning_rate": 1.4736892278360344e-06, "loss": 1.8694, "step": 26735 }, { "epoch": 5.09, "learning_rate": 1.4736320305052432e-06, "loss": 2.6621, "step": 26736 }, { "epoch": 5.09, "learning_rate": 1.473574833174452e-06, "loss": 1.1536, "step": 26737 }, { "epoch": 5.09, "learning_rate": 1.4735176358436608e-06, "loss": 1.7539, "step": 26738 }, { "epoch": 5.09, "learning_rate": 1.4734604385128694e-06, "loss": 2.8743, "step": 26739 }, { "epoch": 5.09, "learning_rate": 1.4734032411820782e-06, "loss": 2.2427, "step": 26740 }, { "epoch": 5.09, "learning_rate": 1.473346043851287e-06, "loss": 2.7303, "step": 26741 }, { "epoch": 5.09, "learning_rate": 1.4732888465204957e-06, "loss": 3.3593, "step": 26742 }, { "epoch": 5.09, "learning_rate": 1.4732316491897045e-06, "loss": 1.3805, "step": 26743 }, { "epoch": 5.09, "learning_rate": 1.4731744518589133e-06, "loss": 2.3046, "step": 26744 }, { "epoch": 5.09, "learning_rate": 1.4731172545281221e-06, "loss": 2.3441, "step": 26745 }, { "epoch": 5.09, "learning_rate": 1.4730600571973307e-06, "loss": 1.5279, "step": 26746 }, { "epoch": 5.09, "learning_rate": 1.4730028598665395e-06, "loss": 2.5569, "step": 26747 }, { "epoch": 5.09, "learning_rate": 1.4729456625357485e-06, "loss": 2.8582, "step": 26748 }, { "epoch": 5.1, "learning_rate": 1.472888465204957e-06, "loss": 1.4651, "step": 26749 }, { "epoch": 5.1, "learning_rate": 1.4728312678741659e-06, "loss": 2.2894, "step": 26750 }, { "epoch": 5.1, "learning_rate": 1.4727740705433747e-06, "loss": 0.9249, "step": 26751 }, { "epoch": 5.1, "learning_rate": 1.4727168732125835e-06, "loss": 2.0955, "step": 26752 }, { "epoch": 5.1, "learning_rate": 1.4726596758817923e-06, "loss": 1.7361, "step": 26753 }, { "epoch": 5.1, "learning_rate": 1.472602478551001e-06, "loss": 1.7504, "step": 26754 }, { "epoch": 5.1, "learning_rate": 1.4725452812202098e-06, "loss": 1.7259, "step": 26755 }, { "epoch": 5.1, "learning_rate": 1.4724880838894184e-06, "loss": 1.2439, "step": 26756 }, { "epoch": 5.1, "learning_rate": 1.4724308865586272e-06, "loss": 2.0856, "step": 26757 }, { "epoch": 5.1, "learning_rate": 1.472373689227836e-06, "loss": 3.0788, "step": 26758 }, { "epoch": 5.1, "learning_rate": 1.472316491897045e-06, "loss": 1.961, "step": 26759 }, { "epoch": 5.1, "learning_rate": 1.4722592945662536e-06, "loss": 1.4326, "step": 26760 }, { "epoch": 5.1, "learning_rate": 1.4722020972354624e-06, "loss": 2.7378, "step": 26761 }, { "epoch": 5.1, "learning_rate": 1.4721448999046712e-06, "loss": 2.9911, "step": 26762 }, { "epoch": 5.1, "learning_rate": 1.4720877025738798e-06, "loss": 2.697, "step": 26763 }, { "epoch": 5.1, "learning_rate": 1.4720305052430888e-06, "loss": 1.6934, "step": 26764 }, { "epoch": 5.1, "learning_rate": 1.4719733079122976e-06, "loss": 1.2464, "step": 26765 }, { "epoch": 5.1, "learning_rate": 1.4719161105815062e-06, "loss": 2.7153, "step": 26766 }, { "epoch": 5.1, "learning_rate": 1.471858913250715e-06, "loss": 1.3403, "step": 26767 }, { "epoch": 5.1, "learning_rate": 1.4718017159199237e-06, "loss": 2.3667, "step": 26768 }, { "epoch": 5.1, "learning_rate": 1.4717445185891325e-06, "loss": 2.6685, "step": 26769 }, { "epoch": 5.1, "learning_rate": 1.4716873212583413e-06, "loss": 2.6495, "step": 26770 }, { "epoch": 5.1, "learning_rate": 1.4716301239275501e-06, "loss": 2.4646, "step": 26771 }, { "epoch": 5.1, "learning_rate": 1.471572926596759e-06, "loss": 1.5853, "step": 26772 }, { "epoch": 5.1, "learning_rate": 1.4715157292659675e-06, "loss": 2.125, "step": 26773 }, { "epoch": 5.1, "learning_rate": 1.4714585319351763e-06, "loss": 2.3445, "step": 26774 }, { "epoch": 5.1, "learning_rate": 1.4714013346043853e-06, "loss": 2.1779, "step": 26775 }, { "epoch": 5.1, "learning_rate": 1.471344137273594e-06, "loss": 1.9441, "step": 26776 }, { "epoch": 5.1, "learning_rate": 1.4712869399428027e-06, "loss": 3.1051, "step": 26777 }, { "epoch": 5.1, "learning_rate": 1.4712297426120115e-06, "loss": 1.4351, "step": 26778 }, { "epoch": 5.1, "learning_rate": 1.4711725452812203e-06, "loss": 2.1433, "step": 26779 }, { "epoch": 5.1, "learning_rate": 1.471115347950429e-06, "loss": 1.9833, "step": 26780 }, { "epoch": 5.1, "learning_rate": 1.4710581506196379e-06, "loss": 1.787, "step": 26781 }, { "epoch": 5.1, "learning_rate": 1.4710009532888466e-06, "loss": 0.5647, "step": 26782 }, { "epoch": 5.1, "learning_rate": 1.4709437559580552e-06, "loss": 1.2294, "step": 26783 }, { "epoch": 5.1, "learning_rate": 1.470886558627264e-06, "loss": 2.5776, "step": 26784 }, { "epoch": 5.1, "learning_rate": 1.4708293612964728e-06, "loss": 3.036, "step": 26785 }, { "epoch": 5.1, "learning_rate": 1.4707721639656816e-06, "loss": 2.0742, "step": 26786 }, { "epoch": 5.1, "learning_rate": 1.4707149666348904e-06, "loss": 2.4362, "step": 26787 }, { "epoch": 5.1, "learning_rate": 1.4706577693040992e-06, "loss": 2.7501, "step": 26788 }, { "epoch": 5.1, "learning_rate": 1.470600571973308e-06, "loss": 1.5118, "step": 26789 }, { "epoch": 5.1, "learning_rate": 1.4705433746425166e-06, "loss": 3.0915, "step": 26790 }, { "epoch": 5.1, "learning_rate": 1.4704861773117254e-06, "loss": 1.8456, "step": 26791 }, { "epoch": 5.1, "learning_rate": 1.4704289799809344e-06, "loss": 1.7965, "step": 26792 }, { "epoch": 5.1, "learning_rate": 1.4703717826501432e-06, "loss": 1.8279, "step": 26793 }, { "epoch": 5.1, "learning_rate": 1.4703145853193517e-06, "loss": 1.3416, "step": 26794 }, { "epoch": 5.1, "learning_rate": 1.4702573879885605e-06, "loss": 1.2848, "step": 26795 }, { "epoch": 5.1, "learning_rate": 1.4702001906577693e-06, "loss": 2.7402, "step": 26796 }, { "epoch": 5.1, "learning_rate": 1.4701429933269781e-06, "loss": 2.4568, "step": 26797 }, { "epoch": 5.1, "learning_rate": 1.470085795996187e-06, "loss": 2.6518, "step": 26798 }, { "epoch": 5.1, "learning_rate": 1.4700285986653957e-06, "loss": 3.158, "step": 26799 }, { "epoch": 5.1, "learning_rate": 1.4699714013346045e-06, "loss": 1.7277, "step": 26800 }, { "epoch": 5.1, "learning_rate": 1.469914204003813e-06, "loss": 3.1486, "step": 26801 }, { "epoch": 5.11, "learning_rate": 1.4698570066730219e-06, "loss": 0.9516, "step": 26802 }, { "epoch": 5.11, "learning_rate": 1.4697998093422309e-06, "loss": 2.5416, "step": 26803 }, { "epoch": 5.11, "learning_rate": 1.4697426120114395e-06, "loss": 2.9639, "step": 26804 }, { "epoch": 5.11, "learning_rate": 1.4696854146806483e-06, "loss": 3.3121, "step": 26805 }, { "epoch": 5.11, "learning_rate": 1.469628217349857e-06, "loss": 2.7854, "step": 26806 }, { "epoch": 5.11, "learning_rate": 1.4695710200190656e-06, "loss": 1.7965, "step": 26807 }, { "epoch": 5.11, "learning_rate": 1.4695138226882746e-06, "loss": 2.1295, "step": 26808 }, { "epoch": 5.11, "learning_rate": 1.4694566253574834e-06, "loss": 1.7355, "step": 26809 }, { "epoch": 5.11, "learning_rate": 1.4693994280266922e-06, "loss": 2.005, "step": 26810 }, { "epoch": 5.11, "learning_rate": 1.4693422306959008e-06, "loss": 2.1481, "step": 26811 }, { "epoch": 5.11, "learning_rate": 1.4692850333651096e-06, "loss": 2.0861, "step": 26812 }, { "epoch": 5.11, "learning_rate": 1.4692278360343184e-06, "loss": 2.0136, "step": 26813 }, { "epoch": 5.11, "learning_rate": 1.4691706387035272e-06, "loss": 1.9949, "step": 26814 }, { "epoch": 5.11, "learning_rate": 1.469113441372736e-06, "loss": 1.3654, "step": 26815 }, { "epoch": 5.11, "learning_rate": 1.4690562440419448e-06, "loss": 1.8027, "step": 26816 }, { "epoch": 5.11, "learning_rate": 1.4689990467111536e-06, "loss": 2.1054, "step": 26817 }, { "epoch": 5.11, "learning_rate": 1.4689418493803622e-06, "loss": 2.1089, "step": 26818 }, { "epoch": 5.11, "learning_rate": 1.4688846520495712e-06, "loss": 1.2555, "step": 26819 }, { "epoch": 5.11, "learning_rate": 1.46882745471878e-06, "loss": 2.9011, "step": 26820 }, { "epoch": 5.11, "learning_rate": 1.4687702573879885e-06, "loss": 2.5334, "step": 26821 }, { "epoch": 5.11, "learning_rate": 1.4687130600571973e-06, "loss": 1.4735, "step": 26822 }, { "epoch": 5.11, "learning_rate": 1.4686558627264061e-06, "loss": 2.2405, "step": 26823 }, { "epoch": 5.11, "learning_rate": 1.468598665395615e-06, "loss": 1.4779, "step": 26824 }, { "epoch": 5.11, "learning_rate": 1.4685414680648237e-06, "loss": 2.9422, "step": 26825 }, { "epoch": 5.11, "learning_rate": 1.4684842707340325e-06, "loss": 3.4107, "step": 26826 }, { "epoch": 5.11, "learning_rate": 1.4684270734032413e-06, "loss": 2.3853, "step": 26827 }, { "epoch": 5.11, "learning_rate": 1.4683698760724499e-06, "loss": 1.3453, "step": 26828 }, { "epoch": 5.11, "learning_rate": 1.4683126787416587e-06, "loss": 2.471, "step": 26829 }, { "epoch": 5.11, "learning_rate": 1.4682554814108675e-06, "loss": 1.8866, "step": 26830 }, { "epoch": 5.11, "learning_rate": 1.4681982840800763e-06, "loss": 1.1532, "step": 26831 }, { "epoch": 5.11, "learning_rate": 1.468141086749285e-06, "loss": 3.4749, "step": 26832 }, { "epoch": 5.11, "learning_rate": 1.4680838894184939e-06, "loss": 2.2496, "step": 26833 }, { "epoch": 5.11, "learning_rate": 1.4680266920877027e-06, "loss": 2.9638, "step": 26834 }, { "epoch": 5.11, "learning_rate": 1.4679694947569112e-06, "loss": 0.8666, "step": 26835 }, { "epoch": 5.11, "learning_rate": 1.4679122974261202e-06, "loss": 2.0004, "step": 26836 }, { "epoch": 5.11, "learning_rate": 1.467855100095329e-06, "loss": 1.3669, "step": 26837 }, { "epoch": 5.11, "learning_rate": 1.4677979027645376e-06, "loss": 1.2007, "step": 26838 }, { "epoch": 5.11, "learning_rate": 1.4677407054337464e-06, "loss": 2.4573, "step": 26839 }, { "epoch": 5.11, "learning_rate": 1.4676835081029552e-06, "loss": 1.6345, "step": 26840 }, { "epoch": 5.11, "learning_rate": 1.467626310772164e-06, "loss": 1.6617, "step": 26841 }, { "epoch": 5.11, "learning_rate": 1.4675691134413728e-06, "loss": 2.2842, "step": 26842 }, { "epoch": 5.11, "learning_rate": 1.4675119161105816e-06, "loss": 0.5564, "step": 26843 }, { "epoch": 5.11, "learning_rate": 1.4674547187797904e-06, "loss": 2.8875, "step": 26844 }, { "epoch": 5.11, "learning_rate": 1.467397521448999e-06, "loss": 2.6836, "step": 26845 }, { "epoch": 5.11, "learning_rate": 1.4673403241182078e-06, "loss": 1.2075, "step": 26846 }, { "epoch": 5.11, "learning_rate": 1.4672831267874168e-06, "loss": 2.2633, "step": 26847 }, { "epoch": 5.11, "learning_rate": 1.4672259294566253e-06, "loss": 1.7407, "step": 26848 }, { "epoch": 5.11, "learning_rate": 1.4671687321258341e-06, "loss": 2.7317, "step": 26849 }, { "epoch": 5.11, "learning_rate": 1.467111534795043e-06, "loss": 1.0285, "step": 26850 }, { "epoch": 5.11, "learning_rate": 1.4670543374642517e-06, "loss": 2.5132, "step": 26851 }, { "epoch": 5.11, "learning_rate": 1.4669971401334605e-06, "loss": 2.9827, "step": 26852 }, { "epoch": 5.11, "learning_rate": 1.4669399428026693e-06, "loss": 1.7866, "step": 26853 }, { "epoch": 5.12, "learning_rate": 1.4668827454718781e-06, "loss": 2.0666, "step": 26854 }, { "epoch": 5.12, "learning_rate": 1.4668255481410867e-06, "loss": 2.0127, "step": 26855 }, { "epoch": 5.12, "learning_rate": 1.4667683508102955e-06, "loss": 2.045, "step": 26856 }, { "epoch": 5.12, "learning_rate": 1.4667111534795043e-06, "loss": 2.3672, "step": 26857 }, { "epoch": 5.12, "learning_rate": 1.466653956148713e-06, "loss": 1.2795, "step": 26858 }, { "epoch": 5.12, "learning_rate": 1.4665967588179219e-06, "loss": 3.1038, "step": 26859 }, { "epoch": 5.12, "learning_rate": 1.4665395614871307e-06, "loss": 2.3641, "step": 26860 }, { "epoch": 5.12, "learning_rate": 1.4664823641563395e-06, "loss": 3.5097, "step": 26861 }, { "epoch": 5.12, "learning_rate": 1.466425166825548e-06, "loss": 1.4271, "step": 26862 }, { "epoch": 5.12, "learning_rate": 1.4663679694947568e-06, "loss": 2.0896, "step": 26863 }, { "epoch": 5.12, "learning_rate": 1.4663107721639658e-06, "loss": 1.5954, "step": 26864 }, { "epoch": 5.12, "learning_rate": 1.4662535748331746e-06, "loss": 1.7753, "step": 26865 }, { "epoch": 5.12, "learning_rate": 1.4661963775023832e-06, "loss": 1.9621, "step": 26866 }, { "epoch": 5.12, "learning_rate": 1.466139180171592e-06, "loss": 2.909, "step": 26867 }, { "epoch": 5.12, "learning_rate": 1.4660819828408008e-06, "loss": 1.7571, "step": 26868 }, { "epoch": 5.12, "learning_rate": 1.4660247855100096e-06, "loss": 2.3911, "step": 26869 }, { "epoch": 5.12, "learning_rate": 1.4659675881792184e-06, "loss": 1.7665, "step": 26870 }, { "epoch": 5.12, "learning_rate": 1.4659103908484272e-06, "loss": 2.0136, "step": 26871 }, { "epoch": 5.12, "learning_rate": 1.4658531935176358e-06, "loss": 1.8583, "step": 26872 }, { "epoch": 5.12, "learning_rate": 1.4657959961868446e-06, "loss": 2.3896, "step": 26873 }, { "epoch": 5.12, "learning_rate": 1.4657387988560533e-06, "loss": 1.9212, "step": 26874 }, { "epoch": 5.12, "learning_rate": 1.4656816015252624e-06, "loss": 1.2087, "step": 26875 }, { "epoch": 5.12, "learning_rate": 1.465624404194471e-06, "loss": 2.6935, "step": 26876 }, { "epoch": 5.12, "learning_rate": 1.4655672068636797e-06, "loss": 1.2656, "step": 26877 }, { "epoch": 5.12, "learning_rate": 1.4655100095328885e-06, "loss": 1.867, "step": 26878 }, { "epoch": 5.12, "learning_rate": 1.465452812202097e-06, "loss": 2.4956, "step": 26879 }, { "epoch": 5.12, "learning_rate": 1.4653956148713061e-06, "loss": 2.8408, "step": 26880 }, { "epoch": 5.12, "learning_rate": 1.465338417540515e-06, "loss": 3.1462, "step": 26881 }, { "epoch": 5.12, "learning_rate": 1.4652812202097237e-06, "loss": 2.8767, "step": 26882 }, { "epoch": 5.12, "learning_rate": 1.4652240228789323e-06, "loss": 2.6302, "step": 26883 }, { "epoch": 5.12, "learning_rate": 1.465166825548141e-06, "loss": 1.3271, "step": 26884 }, { "epoch": 5.12, "learning_rate": 1.4651096282173499e-06, "loss": 2.3899, "step": 26885 }, { "epoch": 5.12, "learning_rate": 1.4650524308865587e-06, "loss": 1.9559, "step": 26886 }, { "epoch": 5.12, "learning_rate": 1.4649952335557675e-06, "loss": 2.9818, "step": 26887 }, { "epoch": 5.12, "learning_rate": 1.4649380362249762e-06, "loss": 0.9687, "step": 26888 }, { "epoch": 5.12, "learning_rate": 1.4648808388941848e-06, "loss": 1.7892, "step": 26889 }, { "epoch": 5.12, "learning_rate": 1.4648236415633936e-06, "loss": 1.7667, "step": 26890 }, { "epoch": 5.12, "learning_rate": 1.4647664442326026e-06, "loss": 2.4383, "step": 26891 }, { "epoch": 5.12, "learning_rate": 1.4647092469018114e-06, "loss": 2.3995, "step": 26892 }, { "epoch": 5.12, "learning_rate": 1.46465204957102e-06, "loss": 2.4674, "step": 26893 }, { "epoch": 5.12, "learning_rate": 1.4645948522402288e-06, "loss": 2.3091, "step": 26894 }, { "epoch": 5.12, "learning_rate": 1.4645376549094376e-06, "loss": 2.1407, "step": 26895 }, { "epoch": 5.12, "learning_rate": 1.4644804575786464e-06, "loss": 0.7662, "step": 26896 }, { "epoch": 5.12, "learning_rate": 1.4644232602478552e-06, "loss": 2.5139, "step": 26897 }, { "epoch": 5.12, "learning_rate": 1.464366062917064e-06, "loss": 2.06, "step": 26898 }, { "epoch": 5.12, "learning_rate": 1.4643088655862728e-06, "loss": 0.9793, "step": 26899 }, { "epoch": 5.12, "learning_rate": 1.4642516682554814e-06, "loss": 1.739, "step": 26900 }, { "epoch": 5.12, "learning_rate": 1.4641944709246901e-06, "loss": 1.3221, "step": 26901 }, { "epoch": 5.12, "learning_rate": 1.464137273593899e-06, "loss": 2.5248, "step": 26902 }, { "epoch": 5.12, "learning_rate": 1.4640800762631077e-06, "loss": 2.3347, "step": 26903 }, { "epoch": 5.12, "learning_rate": 1.4640228789323165e-06, "loss": 0.9068, "step": 26904 }, { "epoch": 5.12, "learning_rate": 1.4639656816015253e-06, "loss": 1.8506, "step": 26905 }, { "epoch": 5.12, "learning_rate": 1.4639084842707341e-06, "loss": 3.0244, "step": 26906 }, { "epoch": 5.13, "learning_rate": 1.4638512869399427e-06, "loss": 3.132, "step": 26907 }, { "epoch": 5.13, "learning_rate": 1.4637940896091517e-06, "loss": 1.2186, "step": 26908 }, { "epoch": 5.13, "learning_rate": 1.4637368922783605e-06, "loss": 2.3709, "step": 26909 }, { "epoch": 5.13, "learning_rate": 1.463679694947569e-06, "loss": 2.5381, "step": 26910 }, { "epoch": 5.13, "learning_rate": 1.4636224976167779e-06, "loss": 1.7997, "step": 26911 }, { "epoch": 5.13, "learning_rate": 1.4635653002859867e-06, "loss": 3.538, "step": 26912 }, { "epoch": 5.13, "learning_rate": 1.4635081029551955e-06, "loss": 1.5407, "step": 26913 }, { "epoch": 5.13, "learning_rate": 1.4634509056244043e-06, "loss": 1.9434, "step": 26914 }, { "epoch": 5.13, "learning_rate": 1.463393708293613e-06, "loss": 1.7801, "step": 26915 }, { "epoch": 5.13, "learning_rate": 1.4633365109628218e-06, "loss": 2.6866, "step": 26916 }, { "epoch": 5.13, "learning_rate": 1.4632793136320304e-06, "loss": 1.5908, "step": 26917 }, { "epoch": 5.13, "learning_rate": 1.4632221163012392e-06, "loss": 2.1107, "step": 26918 }, { "epoch": 5.13, "learning_rate": 1.4631649189704482e-06, "loss": 1.8528, "step": 26919 }, { "epoch": 5.13, "learning_rate": 1.4631077216396568e-06, "loss": 1.3992, "step": 26920 }, { "epoch": 5.13, "learning_rate": 1.4630505243088656e-06, "loss": 3.0466, "step": 26921 }, { "epoch": 5.13, "learning_rate": 1.4629933269780744e-06, "loss": 1.0716, "step": 26922 }, { "epoch": 5.13, "learning_rate": 1.4629361296472832e-06, "loss": 2.1523, "step": 26923 }, { "epoch": 5.13, "learning_rate": 1.462878932316492e-06, "loss": 1.4196, "step": 26924 }, { "epoch": 5.13, "learning_rate": 1.4628217349857008e-06, "loss": 1.7268, "step": 26925 }, { "epoch": 5.13, "learning_rate": 1.4627645376549096e-06, "loss": 1.1415, "step": 26926 }, { "epoch": 5.13, "learning_rate": 1.4627073403241181e-06, "loss": 2.0921, "step": 26927 }, { "epoch": 5.13, "learning_rate": 1.462650142993327e-06, "loss": 1.7558, "step": 26928 }, { "epoch": 5.13, "learning_rate": 1.4625929456625357e-06, "loss": 2.0846, "step": 26929 }, { "epoch": 5.13, "learning_rate": 1.4625357483317447e-06, "loss": 2.1504, "step": 26930 }, { "epoch": 5.13, "learning_rate": 1.4624785510009533e-06, "loss": 2.6596, "step": 26931 }, { "epoch": 5.13, "learning_rate": 1.4624213536701621e-06, "loss": 1.0568, "step": 26932 }, { "epoch": 5.13, "learning_rate": 1.462364156339371e-06, "loss": 2.1887, "step": 26933 }, { "epoch": 5.13, "learning_rate": 1.4623069590085795e-06, "loss": 2.5838, "step": 26934 }, { "epoch": 5.13, "learning_rate": 1.4622497616777885e-06, "loss": 3.3698, "step": 26935 }, { "epoch": 5.13, "learning_rate": 1.4621925643469973e-06, "loss": 1.7831, "step": 26936 }, { "epoch": 5.13, "learning_rate": 1.4621353670162059e-06, "loss": 1.9518, "step": 26937 }, { "epoch": 5.13, "learning_rate": 1.4620781696854147e-06, "loss": 1.6992, "step": 26938 }, { "epoch": 5.13, "learning_rate": 1.4620209723546235e-06, "loss": 2.6341, "step": 26939 }, { "epoch": 5.13, "learning_rate": 1.4619637750238323e-06, "loss": 1.3092, "step": 26940 }, { "epoch": 5.13, "learning_rate": 1.461906577693041e-06, "loss": 1.9331, "step": 26941 }, { "epoch": 5.13, "learning_rate": 1.4618493803622498e-06, "loss": 1.718, "step": 26942 }, { "epoch": 5.13, "learning_rate": 1.4617921830314586e-06, "loss": 2.7757, "step": 26943 }, { "epoch": 5.13, "learning_rate": 1.4617349857006672e-06, "loss": 0.8552, "step": 26944 }, { "epoch": 5.13, "learning_rate": 1.461677788369876e-06, "loss": 1.5699, "step": 26945 }, { "epoch": 5.13, "learning_rate": 1.4616205910390848e-06, "loss": 2.3752, "step": 26946 }, { "epoch": 5.13, "learning_rate": 1.4615633937082938e-06, "loss": 2.472, "step": 26947 }, { "epoch": 5.13, "learning_rate": 1.4615061963775024e-06, "loss": 1.3411, "step": 26948 }, { "epoch": 5.13, "learning_rate": 1.4614489990467112e-06, "loss": 2.4563, "step": 26949 }, { "epoch": 5.13, "learning_rate": 1.46139180171592e-06, "loss": 1.3974, "step": 26950 }, { "epoch": 5.13, "learning_rate": 1.4613346043851286e-06, "loss": 2.6594, "step": 26951 }, { "epoch": 5.13, "learning_rate": 1.4612774070543376e-06, "loss": 2.812, "step": 26952 }, { "epoch": 5.13, "learning_rate": 1.4612202097235464e-06, "loss": 1.862, "step": 26953 }, { "epoch": 5.13, "learning_rate": 1.461163012392755e-06, "loss": 1.5534, "step": 26954 }, { "epoch": 5.13, "learning_rate": 1.4611058150619637e-06, "loss": 1.7909, "step": 26955 }, { "epoch": 5.13, "learning_rate": 1.4610486177311725e-06, "loss": 2.8598, "step": 26956 }, { "epoch": 5.13, "learning_rate": 1.4609914204003813e-06, "loss": 2.5286, "step": 26957 }, { "epoch": 5.13, "learning_rate": 1.4609342230695901e-06, "loss": 2.1388, "step": 26958 }, { "epoch": 5.14, "learning_rate": 1.460877025738799e-06, "loss": 1.1774, "step": 26959 }, { "epoch": 5.14, "learning_rate": 1.4608198284080077e-06, "loss": 1.2031, "step": 26960 }, { "epoch": 5.14, "learning_rate": 1.4607626310772163e-06, "loss": 1.9035, "step": 26961 }, { "epoch": 5.14, "learning_rate": 1.460705433746425e-06, "loss": 2.0083, "step": 26962 }, { "epoch": 5.14, "learning_rate": 1.460648236415634e-06, "loss": 1.5455, "step": 26963 }, { "epoch": 5.14, "learning_rate": 1.4605910390848429e-06, "loss": 0.7854, "step": 26964 }, { "epoch": 5.14, "learning_rate": 1.4605338417540515e-06, "loss": 2.3438, "step": 26965 }, { "epoch": 5.14, "learning_rate": 1.4604766444232603e-06, "loss": 1.6375, "step": 26966 }, { "epoch": 5.14, "learning_rate": 1.460419447092469e-06, "loss": 1.7836, "step": 26967 }, { "epoch": 5.14, "learning_rate": 1.4603622497616778e-06, "loss": 1.6812, "step": 26968 }, { "epoch": 5.14, "learning_rate": 1.4603050524308866e-06, "loss": 3.8817, "step": 26969 }, { "epoch": 5.14, "learning_rate": 1.4602478551000954e-06, "loss": 1.7812, "step": 26970 }, { "epoch": 5.14, "learning_rate": 1.4601906577693042e-06, "loss": 1.8497, "step": 26971 }, { "epoch": 5.14, "learning_rate": 1.4601334604385128e-06, "loss": 1.9724, "step": 26972 }, { "epoch": 5.14, "learning_rate": 1.4600762631077216e-06, "loss": 1.7809, "step": 26973 }, { "epoch": 5.14, "learning_rate": 1.4600190657769306e-06, "loss": 1.3752, "step": 26974 }, { "epoch": 5.14, "learning_rate": 1.4599618684461392e-06, "loss": 1.7847, "step": 26975 }, { "epoch": 5.14, "learning_rate": 1.459904671115348e-06, "loss": 2.2904, "step": 26976 }, { "epoch": 5.14, "learning_rate": 1.4598474737845568e-06, "loss": 1.9559, "step": 26977 }, { "epoch": 5.14, "learning_rate": 1.4597902764537654e-06, "loss": 1.6942, "step": 26978 }, { "epoch": 5.14, "learning_rate": 1.4597330791229744e-06, "loss": 1.3852, "step": 26979 }, { "epoch": 5.14, "learning_rate": 1.4596758817921832e-06, "loss": 3.1854, "step": 26980 }, { "epoch": 5.14, "learning_rate": 1.459618684461392e-06, "loss": 2.4002, "step": 26981 }, { "epoch": 5.14, "learning_rate": 1.4595614871306005e-06, "loss": 0.8124, "step": 26982 }, { "epoch": 5.14, "learning_rate": 1.4595042897998093e-06, "loss": 1.8942, "step": 26983 }, { "epoch": 5.14, "learning_rate": 1.4594470924690181e-06, "loss": 2.4322, "step": 26984 }, { "epoch": 5.14, "learning_rate": 1.459389895138227e-06, "loss": 2.3287, "step": 26985 }, { "epoch": 5.14, "learning_rate": 1.4593326978074357e-06, "loss": 1.8895, "step": 26986 }, { "epoch": 5.14, "learning_rate": 1.4592755004766445e-06, "loss": 2.3054, "step": 26987 }, { "epoch": 5.14, "learning_rate": 1.4592183031458533e-06, "loss": 1.4102, "step": 26988 }, { "epoch": 5.14, "learning_rate": 1.4591611058150619e-06, "loss": 1.2413, "step": 26989 }, { "epoch": 5.14, "learning_rate": 1.4591039084842707e-06, "loss": 2.613, "step": 26990 }, { "epoch": 5.14, "learning_rate": 1.4590467111534797e-06, "loss": 2.1765, "step": 26991 }, { "epoch": 5.14, "learning_rate": 1.4589895138226883e-06, "loss": 2.0365, "step": 26992 }, { "epoch": 5.14, "learning_rate": 1.458932316491897e-06, "loss": 1.9843, "step": 26993 }, { "epoch": 5.14, "learning_rate": 1.4588751191611059e-06, "loss": 2.2655, "step": 26994 }, { "epoch": 5.14, "learning_rate": 1.4588179218303146e-06, "loss": 1.0025, "step": 26995 }, { "epoch": 5.14, "learning_rate": 1.4587607244995234e-06, "loss": 3.4424, "step": 26996 }, { "epoch": 5.14, "learning_rate": 1.4587035271687322e-06, "loss": 0.8202, "step": 26997 }, { "epoch": 5.14, "learning_rate": 1.458646329837941e-06, "loss": 1.5902, "step": 26998 }, { "epoch": 5.14, "learning_rate": 1.4585891325071496e-06, "loss": 1.5577, "step": 26999 }, { "epoch": 5.14, "learning_rate": 1.4585319351763584e-06, "loss": 1.5166, "step": 27000 }, { "epoch": 5.14, "learning_rate": 1.4584747378455672e-06, "loss": 2.5659, "step": 27001 }, { "epoch": 5.14, "learning_rate": 1.458417540514776e-06, "loss": 2.2432, "step": 27002 }, { "epoch": 5.14, "learning_rate": 1.4583603431839848e-06, "loss": 1.9643, "step": 27003 }, { "epoch": 5.14, "learning_rate": 1.4583031458531936e-06, "loss": 2.0997, "step": 27004 }, { "epoch": 5.14, "learning_rate": 1.4582459485224024e-06, "loss": 1.5078, "step": 27005 }, { "epoch": 5.14, "learning_rate": 1.458188751191611e-06, "loss": 2.8549, "step": 27006 }, { "epoch": 5.14, "learning_rate": 1.45813155386082e-06, "loss": 1.405, "step": 27007 }, { "epoch": 5.14, "learning_rate": 1.4580743565300288e-06, "loss": 2.3153, "step": 27008 }, { "epoch": 5.14, "learning_rate": 1.4580171591992373e-06, "loss": 2.458, "step": 27009 }, { "epoch": 5.14, "learning_rate": 1.4579599618684461e-06, "loss": 1.7923, "step": 27010 }, { "epoch": 5.14, "learning_rate": 1.457902764537655e-06, "loss": 1.8904, "step": 27011 }, { "epoch": 5.15, "learning_rate": 1.4578455672068637e-06, "loss": 1.4588, "step": 27012 }, { "epoch": 5.15, "learning_rate": 1.4577883698760725e-06, "loss": 2.746, "step": 27013 }, { "epoch": 5.15, "learning_rate": 1.4577311725452813e-06, "loss": 2.2366, "step": 27014 }, { "epoch": 5.15, "learning_rate": 1.45767397521449e-06, "loss": 2.1739, "step": 27015 }, { "epoch": 5.15, "learning_rate": 1.4576167778836987e-06, "loss": 1.9803, "step": 27016 }, { "epoch": 5.15, "learning_rate": 1.4575595805529075e-06, "loss": 1.2344, "step": 27017 }, { "epoch": 5.15, "learning_rate": 1.4575023832221165e-06, "loss": 1.6408, "step": 27018 }, { "epoch": 5.15, "learning_rate": 1.457445185891325e-06, "loss": 2.0231, "step": 27019 }, { "epoch": 5.15, "learning_rate": 1.4573879885605339e-06, "loss": 1.5355, "step": 27020 }, { "epoch": 5.15, "learning_rate": 1.4573307912297427e-06, "loss": 1.8293, "step": 27021 }, { "epoch": 5.15, "learning_rate": 1.4572735938989514e-06, "loss": 3.1421, "step": 27022 }, { "epoch": 5.15, "learning_rate": 1.4572163965681602e-06, "loss": 1.8992, "step": 27023 }, { "epoch": 5.15, "learning_rate": 1.457159199237369e-06, "loss": 2.8418, "step": 27024 }, { "epoch": 5.15, "learning_rate": 1.4571020019065778e-06, "loss": 2.2285, "step": 27025 }, { "epoch": 5.15, "learning_rate": 1.4570448045757864e-06, "loss": 2.206, "step": 27026 }, { "epoch": 5.15, "learning_rate": 1.4569876072449952e-06, "loss": 1.1293, "step": 27027 }, { "epoch": 5.15, "learning_rate": 1.456930409914204e-06, "loss": 1.5556, "step": 27028 }, { "epoch": 5.15, "learning_rate": 1.4568732125834128e-06, "loss": 2.1167, "step": 27029 }, { "epoch": 5.15, "learning_rate": 1.4568160152526216e-06, "loss": 2.9558, "step": 27030 }, { "epoch": 5.15, "learning_rate": 1.4567588179218304e-06, "loss": 0.9645, "step": 27031 }, { "epoch": 5.15, "learning_rate": 1.4567016205910392e-06, "loss": 4.0427, "step": 27032 }, { "epoch": 5.15, "learning_rate": 1.4566444232602478e-06, "loss": 1.7244, "step": 27033 }, { "epoch": 5.15, "learning_rate": 1.4565872259294565e-06, "loss": 2.2072, "step": 27034 }, { "epoch": 5.15, "learning_rate": 1.4565300285986656e-06, "loss": 1.3317, "step": 27035 }, { "epoch": 5.15, "learning_rate": 1.4564728312678743e-06, "loss": 2.3185, "step": 27036 }, { "epoch": 5.15, "learning_rate": 1.456415633937083e-06, "loss": 1.1916, "step": 27037 }, { "epoch": 5.15, "learning_rate": 1.4563584366062917e-06, "loss": 0.8707, "step": 27038 }, { "epoch": 5.15, "learning_rate": 1.4563012392755005e-06, "loss": 1.4102, "step": 27039 }, { "epoch": 5.15, "learning_rate": 1.4562440419447093e-06, "loss": 2.7627, "step": 27040 }, { "epoch": 5.15, "learning_rate": 1.456186844613918e-06, "loss": 1.9094, "step": 27041 }, { "epoch": 5.15, "learning_rate": 1.456129647283127e-06, "loss": 2.5613, "step": 27042 }, { "epoch": 5.15, "learning_rate": 1.4560724499523355e-06, "loss": 2.9093, "step": 27043 }, { "epoch": 5.15, "learning_rate": 1.4560152526215443e-06, "loss": 1.3231, "step": 27044 }, { "epoch": 5.15, "learning_rate": 1.455958055290753e-06, "loss": 2.1413, "step": 27045 }, { "epoch": 5.15, "learning_rate": 1.455900857959962e-06, "loss": 1.7503, "step": 27046 }, { "epoch": 5.15, "learning_rate": 1.4558436606291707e-06, "loss": 1.9799, "step": 27047 }, { "epoch": 5.15, "learning_rate": 1.4557864632983794e-06, "loss": 1.3281, "step": 27048 }, { "epoch": 5.15, "learning_rate": 1.4557292659675882e-06, "loss": 1.9556, "step": 27049 }, { "epoch": 5.15, "learning_rate": 1.4556720686367968e-06, "loss": 2.4621, "step": 27050 }, { "epoch": 5.15, "learning_rate": 1.4556148713060058e-06, "loss": 1.5934, "step": 27051 }, { "epoch": 5.15, "learning_rate": 1.4555576739752146e-06, "loss": 2.2608, "step": 27052 }, { "epoch": 5.15, "learning_rate": 1.4555004766444234e-06, "loss": 1.3568, "step": 27053 }, { "epoch": 5.15, "learning_rate": 1.455443279313632e-06, "loss": 1.6943, "step": 27054 }, { "epoch": 5.15, "learning_rate": 1.4553860819828408e-06, "loss": 1.5538, "step": 27055 }, { "epoch": 5.15, "learning_rate": 1.4553288846520496e-06, "loss": 2.4828, "step": 27056 }, { "epoch": 5.15, "learning_rate": 1.4552716873212584e-06, "loss": 1.4865, "step": 27057 }, { "epoch": 5.15, "learning_rate": 1.4552144899904672e-06, "loss": 2.8024, "step": 27058 }, { "epoch": 5.15, "learning_rate": 1.455157292659676e-06, "loss": 1.871, "step": 27059 }, { "epoch": 5.15, "learning_rate": 1.4551000953288848e-06, "loss": 2.0975, "step": 27060 }, { "epoch": 5.15, "learning_rate": 1.4550428979980933e-06, "loss": 2.0258, "step": 27061 }, { "epoch": 5.15, "learning_rate": 1.4549857006673021e-06, "loss": 1.9828, "step": 27062 }, { "epoch": 5.15, "learning_rate": 1.4549285033365111e-06, "loss": 2.5537, "step": 27063 }, { "epoch": 5.16, "learning_rate": 1.4548713060057197e-06, "loss": 1.8813, "step": 27064 }, { "epoch": 5.16, "learning_rate": 1.4548141086749285e-06, "loss": 1.9072, "step": 27065 }, { "epoch": 5.16, "learning_rate": 1.4547569113441373e-06, "loss": 1.3132, "step": 27066 }, { "epoch": 5.16, "learning_rate": 1.4546997140133461e-06, "loss": 1.9451, "step": 27067 }, { "epoch": 5.16, "learning_rate": 1.454642516682555e-06, "loss": 1.6959, "step": 27068 }, { "epoch": 5.16, "learning_rate": 1.4545853193517637e-06, "loss": 1.8973, "step": 27069 }, { "epoch": 5.16, "learning_rate": 1.4545281220209725e-06, "loss": 1.4874, "step": 27070 }, { "epoch": 5.16, "learning_rate": 1.454470924690181e-06, "loss": 2.2717, "step": 27071 }, { "epoch": 5.16, "learning_rate": 1.4544137273593899e-06, "loss": 1.8355, "step": 27072 }, { "epoch": 5.16, "learning_rate": 1.4543565300285987e-06, "loss": 1.8452, "step": 27073 }, { "epoch": 5.16, "learning_rate": 1.4542993326978075e-06, "loss": 1.6154, "step": 27074 }, { "epoch": 5.16, "learning_rate": 1.4542421353670162e-06, "loss": 1.7872, "step": 27075 }, { "epoch": 5.16, "learning_rate": 1.454184938036225e-06, "loss": 2.5056, "step": 27076 }, { "epoch": 5.16, "learning_rate": 1.4541277407054338e-06, "loss": 0.985, "step": 27077 }, { "epoch": 5.16, "learning_rate": 1.4540705433746424e-06, "loss": 1.0245, "step": 27078 }, { "epoch": 5.16, "learning_rate": 1.4540133460438514e-06, "loss": 1.7449, "step": 27079 }, { "epoch": 5.16, "learning_rate": 1.4539561487130602e-06, "loss": 2.04, "step": 27080 }, { "epoch": 5.16, "learning_rate": 1.4538989513822688e-06, "loss": 0.8173, "step": 27081 }, { "epoch": 5.16, "learning_rate": 1.4538417540514776e-06, "loss": 1.9017, "step": 27082 }, { "epoch": 5.16, "learning_rate": 1.4537845567206864e-06, "loss": 2.2375, "step": 27083 }, { "epoch": 5.16, "learning_rate": 1.4537273593898952e-06, "loss": 1.2626, "step": 27084 }, { "epoch": 5.16, "learning_rate": 1.453670162059104e-06, "loss": 1.8186, "step": 27085 }, { "epoch": 5.16, "learning_rate": 1.4536129647283128e-06, "loss": 0.5564, "step": 27086 }, { "epoch": 5.16, "learning_rate": 1.4535557673975216e-06, "loss": 1.6135, "step": 27087 }, { "epoch": 5.16, "learning_rate": 1.4534985700667301e-06, "loss": 2.6928, "step": 27088 }, { "epoch": 5.16, "learning_rate": 1.453441372735939e-06, "loss": 2.8561, "step": 27089 }, { "epoch": 5.16, "learning_rate": 1.453384175405148e-06, "loss": 2.742, "step": 27090 }, { "epoch": 5.16, "learning_rate": 1.4533269780743565e-06, "loss": 1.5067, "step": 27091 }, { "epoch": 5.16, "learning_rate": 1.4532697807435653e-06, "loss": 1.7532, "step": 27092 }, { "epoch": 5.16, "learning_rate": 1.4532125834127741e-06, "loss": 2.0036, "step": 27093 }, { "epoch": 5.16, "learning_rate": 1.453155386081983e-06, "loss": 1.928, "step": 27094 }, { "epoch": 5.16, "learning_rate": 1.4530981887511917e-06, "loss": 1.5064, "step": 27095 }, { "epoch": 5.16, "learning_rate": 1.4530409914204005e-06, "loss": 2.286, "step": 27096 }, { "epoch": 5.16, "learning_rate": 1.4529837940896093e-06, "loss": 1.8245, "step": 27097 }, { "epoch": 5.16, "learning_rate": 1.4529265967588179e-06, "loss": 2.4727, "step": 27098 }, { "epoch": 5.16, "learning_rate": 1.4528693994280267e-06, "loss": 2.5755, "step": 27099 }, { "epoch": 5.16, "learning_rate": 1.4528122020972355e-06, "loss": 1.3877, "step": 27100 }, { "epoch": 5.16, "learning_rate": 1.4527550047664442e-06, "loss": 2.1703, "step": 27101 }, { "epoch": 5.16, "learning_rate": 1.452697807435653e-06, "loss": 2.4096, "step": 27102 }, { "epoch": 5.16, "learning_rate": 1.4526406101048618e-06, "loss": 2.2157, "step": 27103 }, { "epoch": 5.16, "learning_rate": 1.4525834127740706e-06, "loss": 1.8744, "step": 27104 }, { "epoch": 5.16, "learning_rate": 1.4525262154432792e-06, "loss": 0.68, "step": 27105 }, { "epoch": 5.16, "learning_rate": 1.452469018112488e-06, "loss": 2.7729, "step": 27106 }, { "epoch": 5.16, "learning_rate": 1.452411820781697e-06, "loss": 2.3566, "step": 27107 }, { "epoch": 5.16, "learning_rate": 1.4523546234509056e-06, "loss": 1.4245, "step": 27108 }, { "epoch": 5.16, "learning_rate": 1.4522974261201144e-06, "loss": 1.0384, "step": 27109 }, { "epoch": 5.16, "learning_rate": 1.4522402287893232e-06, "loss": 1.5071, "step": 27110 }, { "epoch": 5.16, "learning_rate": 1.452183031458532e-06, "loss": 0.9916, "step": 27111 }, { "epoch": 5.16, "learning_rate": 1.4521258341277408e-06, "loss": 1.8969, "step": 27112 }, { "epoch": 5.16, "learning_rate": 1.4520686367969496e-06, "loss": 2.786, "step": 27113 }, { "epoch": 5.16, "learning_rate": 1.4520114394661584e-06, "loss": 2.3178, "step": 27114 }, { "epoch": 5.16, "learning_rate": 1.451954242135367e-06, "loss": 1.5918, "step": 27115 }, { "epoch": 5.16, "learning_rate": 1.4518970448045757e-06, "loss": 3.2103, "step": 27116 }, { "epoch": 5.17, "learning_rate": 1.4518398474737845e-06, "loss": 1.18, "step": 27117 }, { "epoch": 5.17, "learning_rate": 1.4517826501429935e-06, "loss": 1.8874, "step": 27118 }, { "epoch": 5.17, "learning_rate": 1.4517254528122021e-06, "loss": 1.4914, "step": 27119 }, { "epoch": 5.17, "learning_rate": 1.451668255481411e-06, "loss": 3.4072, "step": 27120 }, { "epoch": 5.17, "learning_rate": 1.4516110581506197e-06, "loss": 1.4081, "step": 27121 }, { "epoch": 5.17, "learning_rate": 1.4515538608198283e-06, "loss": 2.0934, "step": 27122 }, { "epoch": 5.17, "learning_rate": 1.4514966634890373e-06, "loss": 1.6447, "step": 27123 }, { "epoch": 5.17, "learning_rate": 1.451439466158246e-06, "loss": 1.6434, "step": 27124 }, { "epoch": 5.17, "learning_rate": 1.4513822688274547e-06, "loss": 2.4652, "step": 27125 }, { "epoch": 5.17, "learning_rate": 1.4513250714966635e-06, "loss": 2.3741, "step": 27126 }, { "epoch": 5.17, "learning_rate": 1.4512678741658723e-06, "loss": 2.0277, "step": 27127 }, { "epoch": 5.17, "learning_rate": 1.451210676835081e-06, "loss": 2.2148, "step": 27128 }, { "epoch": 5.17, "learning_rate": 1.4511534795042898e-06, "loss": 2.7144, "step": 27129 }, { "epoch": 5.17, "learning_rate": 1.4510962821734986e-06, "loss": 2.1668, "step": 27130 }, { "epoch": 5.17, "learning_rate": 1.4510390848427074e-06, "loss": 2.9595, "step": 27131 }, { "epoch": 5.17, "learning_rate": 1.450981887511916e-06, "loss": 1.7344, "step": 27132 }, { "epoch": 5.17, "learning_rate": 1.4509246901811248e-06, "loss": 1.7519, "step": 27133 }, { "epoch": 5.17, "learning_rate": 1.4508674928503338e-06, "loss": 1.7676, "step": 27134 }, { "epoch": 5.17, "learning_rate": 1.4508102955195426e-06, "loss": 3.0659, "step": 27135 }, { "epoch": 5.17, "learning_rate": 1.4507530981887512e-06, "loss": 2.3426, "step": 27136 }, { "epoch": 5.17, "learning_rate": 1.45069590085796e-06, "loss": 1.579, "step": 27137 }, { "epoch": 5.17, "learning_rate": 1.4506387035271688e-06, "loss": 3.4057, "step": 27138 }, { "epoch": 5.17, "learning_rate": 1.4505815061963776e-06, "loss": 2.7848, "step": 27139 }, { "epoch": 5.17, "learning_rate": 1.4505243088655864e-06, "loss": 1.6805, "step": 27140 }, { "epoch": 5.17, "learning_rate": 1.4504671115347952e-06, "loss": 1.6158, "step": 27141 }, { "epoch": 5.17, "learning_rate": 1.450409914204004e-06, "loss": 1.3338, "step": 27142 }, { "epoch": 5.17, "learning_rate": 1.4503527168732125e-06, "loss": 0.9528, "step": 27143 }, { "epoch": 5.17, "learning_rate": 1.4502955195424213e-06, "loss": 1.2761, "step": 27144 }, { "epoch": 5.17, "learning_rate": 1.4502383222116301e-06, "loss": 2.1128, "step": 27145 }, { "epoch": 5.17, "learning_rate": 1.450181124880839e-06, "loss": 1.2988, "step": 27146 }, { "epoch": 5.17, "learning_rate": 1.4501239275500477e-06, "loss": 0.8998, "step": 27147 }, { "epoch": 5.17, "learning_rate": 1.4500667302192565e-06, "loss": 1.5506, "step": 27148 }, { "epoch": 5.17, "learning_rate": 1.450009532888465e-06, "loss": 1.7078, "step": 27149 }, { "epoch": 5.17, "learning_rate": 1.4499523355576739e-06, "loss": 1.3026, "step": 27150 }, { "epoch": 5.17, "learning_rate": 1.4498951382268829e-06, "loss": 3.1755, "step": 27151 }, { "epoch": 5.17, "learning_rate": 1.4498379408960917e-06, "loss": 1.7955, "step": 27152 }, { "epoch": 5.17, "learning_rate": 1.4497807435653003e-06, "loss": 1.3742, "step": 27153 }, { "epoch": 5.17, "learning_rate": 1.449723546234509e-06, "loss": 1.1017, "step": 27154 }, { "epoch": 5.17, "learning_rate": 1.4496663489037178e-06, "loss": 3.0759, "step": 27155 }, { "epoch": 5.17, "learning_rate": 1.4496091515729266e-06, "loss": 1.4648, "step": 27156 }, { "epoch": 5.17, "learning_rate": 1.4495519542421354e-06, "loss": 1.5575, "step": 27157 }, { "epoch": 5.17, "learning_rate": 1.4494947569113442e-06, "loss": 1.4596, "step": 27158 }, { "epoch": 5.17, "learning_rate": 1.449437559580553e-06, "loss": 2.7712, "step": 27159 }, { "epoch": 5.17, "learning_rate": 1.4493803622497616e-06, "loss": 2.1119, "step": 27160 }, { "epoch": 5.17, "learning_rate": 1.4493231649189704e-06, "loss": 2.5236, "step": 27161 }, { "epoch": 5.17, "learning_rate": 1.4492659675881794e-06, "loss": 1.2911, "step": 27162 }, { "epoch": 5.17, "learning_rate": 1.449208770257388e-06, "loss": 1.8292, "step": 27163 }, { "epoch": 5.17, "learning_rate": 1.4491515729265968e-06, "loss": 2.0456, "step": 27164 }, { "epoch": 5.17, "learning_rate": 1.4490943755958056e-06, "loss": 1.0879, "step": 27165 }, { "epoch": 5.17, "learning_rate": 1.4490371782650144e-06, "loss": 1.7934, "step": 27166 }, { "epoch": 5.17, "learning_rate": 1.4489799809342232e-06, "loss": 1.1679, "step": 27167 }, { "epoch": 5.17, "learning_rate": 1.448922783603432e-06, "loss": 1.7161, "step": 27168 }, { "epoch": 5.18, "learning_rate": 1.4488655862726407e-06, "loss": 2.2996, "step": 27169 }, { "epoch": 5.18, "learning_rate": 1.4488083889418493e-06, "loss": 3.4943, "step": 27170 }, { "epoch": 5.18, "learning_rate": 1.4487511916110581e-06, "loss": 1.8985, "step": 27171 }, { "epoch": 5.18, "learning_rate": 1.448693994280267e-06, "loss": 0.7632, "step": 27172 }, { "epoch": 5.18, "learning_rate": 1.4486367969494757e-06, "loss": 0.5623, "step": 27173 }, { "epoch": 5.18, "learning_rate": 1.4485795996186845e-06, "loss": 0.7185, "step": 27174 }, { "epoch": 5.18, "learning_rate": 1.4485224022878933e-06, "loss": 1.5455, "step": 27175 }, { "epoch": 5.18, "learning_rate": 1.448465204957102e-06, "loss": 3.1739, "step": 27176 }, { "epoch": 5.18, "learning_rate": 1.4484080076263107e-06, "loss": 2.3953, "step": 27177 }, { "epoch": 5.18, "learning_rate": 1.4483508102955197e-06, "loss": 2.1386, "step": 27178 }, { "epoch": 5.18, "learning_rate": 1.4482936129647285e-06, "loss": 1.9068, "step": 27179 }, { "epoch": 5.18, "learning_rate": 1.448236415633937e-06, "loss": 2.0491, "step": 27180 }, { "epoch": 5.18, "learning_rate": 1.4481792183031458e-06, "loss": 0.86, "step": 27181 }, { "epoch": 5.18, "learning_rate": 1.4481220209723546e-06, "loss": 1.6742, "step": 27182 }, { "epoch": 5.18, "learning_rate": 1.4480648236415634e-06, "loss": 2.2824, "step": 27183 }, { "epoch": 5.18, "learning_rate": 1.4480076263107722e-06, "loss": 2.7873, "step": 27184 }, { "epoch": 5.18, "learning_rate": 1.447950428979981e-06, "loss": 0.9043, "step": 27185 }, { "epoch": 5.18, "learning_rate": 1.4478932316491898e-06, "loss": 1.8351, "step": 27186 }, { "epoch": 5.18, "learning_rate": 1.4478360343183984e-06, "loss": 2.7793, "step": 27187 }, { "epoch": 5.18, "learning_rate": 1.4477788369876072e-06, "loss": 0.8183, "step": 27188 }, { "epoch": 5.18, "learning_rate": 1.447721639656816e-06, "loss": 1.1771, "step": 27189 }, { "epoch": 5.18, "learning_rate": 1.4476644423260248e-06, "loss": 2.0636, "step": 27190 }, { "epoch": 5.18, "learning_rate": 1.4476072449952336e-06, "loss": 1.8113, "step": 27191 }, { "epoch": 5.18, "learning_rate": 1.4475500476644424e-06, "loss": 1.414, "step": 27192 }, { "epoch": 5.18, "learning_rate": 1.4474928503336512e-06, "loss": 2.1372, "step": 27193 }, { "epoch": 5.18, "learning_rate": 1.4474356530028597e-06, "loss": 2.1421, "step": 27194 }, { "epoch": 5.18, "learning_rate": 1.4473784556720688e-06, "loss": 1.7149, "step": 27195 }, { "epoch": 5.18, "learning_rate": 1.4473212583412775e-06, "loss": 2.1496, "step": 27196 }, { "epoch": 5.18, "learning_rate": 1.4472640610104861e-06, "loss": 3.3207, "step": 27197 }, { "epoch": 5.18, "learning_rate": 1.447206863679695e-06, "loss": 2.1413, "step": 27198 }, { "epoch": 5.18, "learning_rate": 1.4471496663489037e-06, "loss": 2.6285, "step": 27199 }, { "epoch": 5.18, "learning_rate": 1.4470924690181125e-06, "loss": 2.3328, "step": 27200 }, { "epoch": 5.18, "learning_rate": 1.4470352716873213e-06, "loss": 2.1387, "step": 27201 }, { "epoch": 5.18, "learning_rate": 1.44697807435653e-06, "loss": 1.9965, "step": 27202 }, { "epoch": 5.18, "learning_rate": 1.4469208770257389e-06, "loss": 0.9307, "step": 27203 }, { "epoch": 5.18, "learning_rate": 1.4468636796949475e-06, "loss": 1.5046, "step": 27204 }, { "epoch": 5.18, "learning_rate": 1.4468064823641563e-06, "loss": 1.3875, "step": 27205 }, { "epoch": 5.18, "learning_rate": 1.4467492850333653e-06, "loss": 1.8311, "step": 27206 }, { "epoch": 5.18, "learning_rate": 1.446692087702574e-06, "loss": 1.745, "step": 27207 }, { "epoch": 5.18, "learning_rate": 1.4466348903717826e-06, "loss": 2.412, "step": 27208 }, { "epoch": 5.18, "learning_rate": 1.4465776930409914e-06, "loss": 1.9972, "step": 27209 }, { "epoch": 5.18, "learning_rate": 1.4465204957102002e-06, "loss": 2.4621, "step": 27210 }, { "epoch": 5.18, "learning_rate": 1.446463298379409e-06, "loss": 1.593, "step": 27211 }, { "epoch": 5.18, "learning_rate": 1.4464061010486178e-06, "loss": 2.4937, "step": 27212 }, { "epoch": 5.18, "learning_rate": 1.4463489037178266e-06, "loss": 1.1366, "step": 27213 }, { "epoch": 5.18, "learning_rate": 1.4462917063870352e-06, "loss": 2.558, "step": 27214 }, { "epoch": 5.18, "learning_rate": 1.446234509056244e-06, "loss": 1.7482, "step": 27215 }, { "epoch": 5.18, "learning_rate": 1.4461773117254528e-06, "loss": 2.6923, "step": 27216 }, { "epoch": 5.18, "learning_rate": 1.4461201143946616e-06, "loss": 0.724, "step": 27217 }, { "epoch": 5.18, "learning_rate": 1.4460629170638704e-06, "loss": 2.4279, "step": 27218 }, { "epoch": 5.18, "learning_rate": 1.4460057197330792e-06, "loss": 3.2202, "step": 27219 }, { "epoch": 5.18, "learning_rate": 1.445948522402288e-06, "loss": 1.4344, "step": 27220 }, { "epoch": 5.18, "learning_rate": 1.4458913250714965e-06, "loss": 2.0425, "step": 27221 }, { "epoch": 5.19, "learning_rate": 1.4458341277407055e-06, "loss": 1.5544, "step": 27222 }, { "epoch": 5.19, "learning_rate": 1.4457769304099143e-06, "loss": 1.6809, "step": 27223 }, { "epoch": 5.19, "learning_rate": 1.4457197330791231e-06, "loss": 1.2563, "step": 27224 }, { "epoch": 5.19, "learning_rate": 1.4456625357483317e-06, "loss": 3.4867, "step": 27225 }, { "epoch": 5.19, "learning_rate": 1.4456053384175405e-06, "loss": 2.7333, "step": 27226 }, { "epoch": 5.19, "learning_rate": 1.4455481410867493e-06, "loss": 2.8647, "step": 27227 }, { "epoch": 5.19, "learning_rate": 1.445490943755958e-06, "loss": 2.2427, "step": 27228 }, { "epoch": 5.19, "learning_rate": 1.4454337464251669e-06, "loss": 0.659, "step": 27229 }, { "epoch": 5.19, "learning_rate": 1.4453765490943757e-06, "loss": 1.4825, "step": 27230 }, { "epoch": 5.19, "learning_rate": 1.4453193517635845e-06, "loss": 0.5829, "step": 27231 }, { "epoch": 5.19, "learning_rate": 1.445262154432793e-06, "loss": 1.6078, "step": 27232 }, { "epoch": 5.19, "learning_rate": 1.4452049571020019e-06, "loss": 1.4388, "step": 27233 }, { "epoch": 5.19, "learning_rate": 1.4451477597712109e-06, "loss": 2.1389, "step": 27234 }, { "epoch": 5.19, "learning_rate": 1.4450905624404194e-06, "loss": 2.8465, "step": 27235 }, { "epoch": 5.19, "learning_rate": 1.4450333651096282e-06, "loss": 2.3422, "step": 27236 }, { "epoch": 5.19, "learning_rate": 1.444976167778837e-06, "loss": 0.6069, "step": 27237 }, { "epoch": 5.19, "learning_rate": 1.4449189704480456e-06, "loss": 1.863, "step": 27238 }, { "epoch": 5.19, "learning_rate": 1.4448617731172546e-06, "loss": 2.3595, "step": 27239 }, { "epoch": 5.19, "learning_rate": 1.4448045757864634e-06, "loss": 2.4012, "step": 27240 }, { "epoch": 5.19, "learning_rate": 1.4447473784556722e-06, "loss": 1.6869, "step": 27241 }, { "epoch": 5.19, "learning_rate": 1.4446901811248808e-06, "loss": 1.3967, "step": 27242 }, { "epoch": 5.19, "learning_rate": 1.4446329837940896e-06, "loss": 2.763, "step": 27243 }, { "epoch": 5.19, "learning_rate": 1.4445757864632984e-06, "loss": 2.0102, "step": 27244 }, { "epoch": 5.19, "learning_rate": 1.4445185891325072e-06, "loss": 1.5908, "step": 27245 }, { "epoch": 5.19, "learning_rate": 1.444461391801716e-06, "loss": 2.9067, "step": 27246 }, { "epoch": 5.19, "learning_rate": 1.4444041944709248e-06, "loss": 2.9455, "step": 27247 }, { "epoch": 5.19, "learning_rate": 1.4443469971401336e-06, "loss": 0.9541, "step": 27248 }, { "epoch": 5.19, "learning_rate": 1.4442897998093421e-06, "loss": 2.3894, "step": 27249 }, { "epoch": 5.19, "learning_rate": 1.4442326024785511e-06, "loss": 2.7999, "step": 27250 }, { "epoch": 5.19, "learning_rate": 1.44417540514776e-06, "loss": 2.2099, "step": 27251 }, { "epoch": 5.19, "learning_rate": 1.4441182078169685e-06, "loss": 1.9034, "step": 27252 }, { "epoch": 5.19, "learning_rate": 1.4440610104861773e-06, "loss": 1.0868, "step": 27253 }, { "epoch": 5.19, "learning_rate": 1.444003813155386e-06, "loss": 2.4671, "step": 27254 }, { "epoch": 5.19, "learning_rate": 1.443946615824595e-06, "loss": 1.4145, "step": 27255 }, { "epoch": 5.19, "learning_rate": 1.4438894184938037e-06, "loss": 2.2126, "step": 27256 }, { "epoch": 5.19, "learning_rate": 1.4438322211630125e-06, "loss": 1.6866, "step": 27257 }, { "epoch": 5.19, "learning_rate": 1.4437750238322213e-06, "loss": 2.3736, "step": 27258 }, { "epoch": 5.19, "learning_rate": 1.4437178265014299e-06, "loss": 1.8499, "step": 27259 }, { "epoch": 5.19, "learning_rate": 1.4436606291706387e-06, "loss": 1.5162, "step": 27260 }, { "epoch": 5.19, "learning_rate": 1.4436034318398474e-06, "loss": 1.8559, "step": 27261 }, { "epoch": 5.19, "learning_rate": 1.4435462345090562e-06, "loss": 1.9194, "step": 27262 }, { "epoch": 5.19, "learning_rate": 1.443489037178265e-06, "loss": 1.4532, "step": 27263 }, { "epoch": 5.19, "learning_rate": 1.4434318398474738e-06, "loss": 1.205, "step": 27264 }, { "epoch": 5.19, "learning_rate": 1.4433746425166826e-06, "loss": 2.3771, "step": 27265 }, { "epoch": 5.19, "learning_rate": 1.4433174451858914e-06, "loss": 2.1269, "step": 27266 }, { "epoch": 5.19, "learning_rate": 1.4432602478551002e-06, "loss": 3.5639, "step": 27267 }, { "epoch": 5.19, "learning_rate": 1.443203050524309e-06, "loss": 1.6269, "step": 27268 }, { "epoch": 5.19, "learning_rate": 1.4431458531935176e-06, "loss": 2.0666, "step": 27269 }, { "epoch": 5.19, "learning_rate": 1.4430886558627264e-06, "loss": 1.63, "step": 27270 }, { "epoch": 5.19, "learning_rate": 1.4430314585319352e-06, "loss": 1.7498, "step": 27271 }, { "epoch": 5.19, "learning_rate": 1.442974261201144e-06, "loss": 2.5351, "step": 27272 }, { "epoch": 5.19, "learning_rate": 1.4429170638703528e-06, "loss": 0.714, "step": 27273 }, { "epoch": 5.2, "learning_rate": 1.4428598665395616e-06, "loss": 1.6061, "step": 27274 }, { "epoch": 5.2, "learning_rate": 1.4428026692087704e-06, "loss": 2.7511, "step": 27275 }, { "epoch": 5.2, "learning_rate": 1.442745471877979e-06, "loss": 1.4992, "step": 27276 }, { "epoch": 5.2, "learning_rate": 1.4426882745471877e-06, "loss": 1.4937, "step": 27277 }, { "epoch": 5.2, "learning_rate": 1.4426310772163967e-06, "loss": 2.2994, "step": 27278 }, { "epoch": 5.2, "learning_rate": 1.4425738798856053e-06, "loss": 3.5941, "step": 27279 }, { "epoch": 5.2, "learning_rate": 1.4425166825548141e-06, "loss": 1.9505, "step": 27280 }, { "epoch": 5.2, "learning_rate": 1.442459485224023e-06, "loss": 0.9247, "step": 27281 }, { "epoch": 5.2, "learning_rate": 1.4424022878932317e-06, "loss": 1.7914, "step": 27282 }, { "epoch": 5.2, "learning_rate": 1.4423450905624405e-06, "loss": 1.5982, "step": 27283 }, { "epoch": 5.2, "learning_rate": 1.4422878932316493e-06, "loss": 1.535, "step": 27284 }, { "epoch": 5.2, "learning_rate": 1.442230695900858e-06, "loss": 2.0385, "step": 27285 }, { "epoch": 5.2, "learning_rate": 1.4421734985700667e-06, "loss": 2.019, "step": 27286 }, { "epoch": 5.2, "learning_rate": 1.4421163012392755e-06, "loss": 1.7515, "step": 27287 }, { "epoch": 5.2, "learning_rate": 1.4420591039084842e-06, "loss": 1.7869, "step": 27288 }, { "epoch": 5.2, "learning_rate": 1.4420019065776933e-06, "loss": 2.1941, "step": 27289 }, { "epoch": 5.2, "learning_rate": 1.4419447092469018e-06, "loss": 1.9998, "step": 27290 }, { "epoch": 5.2, "learning_rate": 1.4418875119161106e-06, "loss": 2.6209, "step": 27291 }, { "epoch": 5.2, "learning_rate": 1.4418303145853194e-06, "loss": 2.1981, "step": 27292 }, { "epoch": 5.2, "learning_rate": 1.441773117254528e-06, "loss": 0.8792, "step": 27293 }, { "epoch": 5.2, "learning_rate": 1.441715919923737e-06, "loss": 1.4825, "step": 27294 }, { "epoch": 5.2, "learning_rate": 1.4416587225929458e-06, "loss": 2.2565, "step": 27295 }, { "epoch": 5.2, "learning_rate": 1.4416015252621546e-06, "loss": 1.6062, "step": 27296 }, { "epoch": 5.2, "learning_rate": 1.4415443279313632e-06, "loss": 1.2896, "step": 27297 }, { "epoch": 5.2, "learning_rate": 1.441487130600572e-06, "loss": 1.5581, "step": 27298 }, { "epoch": 5.2, "learning_rate": 1.4414299332697808e-06, "loss": 3.0661, "step": 27299 }, { "epoch": 5.2, "learning_rate": 1.4413727359389896e-06, "loss": 1.9801, "step": 27300 }, { "epoch": 5.2, "learning_rate": 1.4413155386081984e-06, "loss": 2.8671, "step": 27301 }, { "epoch": 5.2, "learning_rate": 1.4412583412774071e-06, "loss": 0.7584, "step": 27302 }, { "epoch": 5.2, "learning_rate": 1.4412011439466157e-06, "loss": 1.4978, "step": 27303 }, { "epoch": 5.2, "learning_rate": 1.4411439466158245e-06, "loss": 2.0864, "step": 27304 }, { "epoch": 5.2, "learning_rate": 1.4410867492850333e-06, "loss": 1.0734, "step": 27305 }, { "epoch": 5.2, "learning_rate": 1.4410295519542423e-06, "loss": 3.552, "step": 27306 }, { "epoch": 5.2, "learning_rate": 1.440972354623451e-06, "loss": 1.7523, "step": 27307 }, { "epoch": 5.2, "learning_rate": 1.4409151572926597e-06, "loss": 1.7093, "step": 27308 }, { "epoch": 5.2, "learning_rate": 1.4408579599618685e-06, "loss": 3.0986, "step": 27309 }, { "epoch": 5.2, "learning_rate": 1.4408007626310773e-06, "loss": 2.3521, "step": 27310 }, { "epoch": 5.2, "learning_rate": 1.440743565300286e-06, "loss": 2.1537, "step": 27311 }, { "epoch": 5.2, "learning_rate": 1.4406863679694949e-06, "loss": 1.3772, "step": 27312 }, { "epoch": 5.2, "learning_rate": 1.4406291706387037e-06, "loss": 1.0423, "step": 27313 }, { "epoch": 5.2, "learning_rate": 1.4405719733079123e-06, "loss": 2.7636, "step": 27314 }, { "epoch": 5.2, "learning_rate": 1.440514775977121e-06, "loss": 1.3433, "step": 27315 }, { "epoch": 5.2, "learning_rate": 1.4404575786463298e-06, "loss": 2.7772, "step": 27316 }, { "epoch": 5.2, "learning_rate": 1.4404003813155386e-06, "loss": 2.8355, "step": 27317 }, { "epoch": 5.2, "learning_rate": 1.4403431839847474e-06, "loss": 1.0312, "step": 27318 }, { "epoch": 5.2, "learning_rate": 1.4402859866539562e-06, "loss": 2.2691, "step": 27319 }, { "epoch": 5.2, "learning_rate": 1.4402287893231648e-06, "loss": 1.8033, "step": 27320 }, { "epoch": 5.2, "learning_rate": 1.4401715919923736e-06, "loss": 1.5296, "step": 27321 }, { "epoch": 5.2, "learning_rate": 1.4401143946615826e-06, "loss": 0.704, "step": 27322 }, { "epoch": 5.2, "learning_rate": 1.4400571973307914e-06, "loss": 1.9789, "step": 27323 }, { "epoch": 5.2, "learning_rate": 1.44e-06, "loss": 2.0533, "step": 27324 }, { "epoch": 5.2, "learning_rate": 1.4399428026692088e-06, "loss": 3.4408, "step": 27325 }, { "epoch": 5.2, "learning_rate": 1.4398856053384176e-06, "loss": 1.5805, "step": 27326 }, { "epoch": 5.21, "learning_rate": 1.4398284080076264e-06, "loss": 0.8095, "step": 27327 }, { "epoch": 5.21, "learning_rate": 1.4397712106768352e-06, "loss": 2.1046, "step": 27328 }, { "epoch": 5.21, "learning_rate": 1.439714013346044e-06, "loss": 2.4623, "step": 27329 }, { "epoch": 5.21, "learning_rate": 1.4396568160152527e-06, "loss": 1.8175, "step": 27330 }, { "epoch": 5.21, "learning_rate": 1.4395996186844613e-06, "loss": 1.6558, "step": 27331 }, { "epoch": 5.21, "learning_rate": 1.4395424213536701e-06, "loss": 3.2626, "step": 27332 }, { "epoch": 5.21, "learning_rate": 1.4394852240228791e-06, "loss": 2.2844, "step": 27333 }, { "epoch": 5.21, "learning_rate": 1.4394280266920877e-06, "loss": 2.3411, "step": 27334 }, { "epoch": 5.21, "learning_rate": 1.4393708293612965e-06, "loss": 0.8237, "step": 27335 }, { "epoch": 5.21, "learning_rate": 1.4393136320305053e-06, "loss": 0.9466, "step": 27336 }, { "epoch": 5.21, "learning_rate": 1.439256434699714e-06, "loss": 1.9482, "step": 27337 }, { "epoch": 5.21, "learning_rate": 1.4391992373689229e-06, "loss": 3.3308, "step": 27338 }, { "epoch": 5.21, "learning_rate": 1.4391420400381317e-06, "loss": 3.676, "step": 27339 }, { "epoch": 5.21, "learning_rate": 1.4390848427073405e-06, "loss": 1.6504, "step": 27340 }, { "epoch": 5.21, "learning_rate": 1.439027645376549e-06, "loss": 2.3221, "step": 27341 }, { "epoch": 5.21, "learning_rate": 1.4389704480457578e-06, "loss": 1.9256, "step": 27342 }, { "epoch": 5.21, "learning_rate": 1.4389132507149666e-06, "loss": 1.4919, "step": 27343 }, { "epoch": 5.21, "learning_rate": 1.4388560533841754e-06, "loss": 0.7756, "step": 27344 }, { "epoch": 5.21, "learning_rate": 1.4387988560533842e-06, "loss": 3.0136, "step": 27345 }, { "epoch": 5.21, "learning_rate": 1.438741658722593e-06, "loss": 2.2035, "step": 27346 }, { "epoch": 5.21, "learning_rate": 1.4386844613918018e-06, "loss": 2.7937, "step": 27347 }, { "epoch": 5.21, "learning_rate": 1.4386272640610104e-06, "loss": 1.111, "step": 27348 }, { "epoch": 5.21, "learning_rate": 1.4385700667302192e-06, "loss": 0.7675, "step": 27349 }, { "epoch": 5.21, "learning_rate": 1.4385128693994282e-06, "loss": 2.3442, "step": 27350 }, { "epoch": 5.21, "learning_rate": 1.4384556720686368e-06, "loss": 1.8537, "step": 27351 }, { "epoch": 5.21, "learning_rate": 1.4383984747378456e-06, "loss": 1.6801, "step": 27352 }, { "epoch": 5.21, "learning_rate": 1.4383412774070544e-06, "loss": 2.5971, "step": 27353 }, { "epoch": 5.21, "learning_rate": 1.4382840800762632e-06, "loss": 1.9304, "step": 27354 }, { "epoch": 5.21, "learning_rate": 1.438226882745472e-06, "loss": 1.849, "step": 27355 }, { "epoch": 5.21, "learning_rate": 1.4381696854146807e-06, "loss": 1.9972, "step": 27356 }, { "epoch": 5.21, "learning_rate": 1.4381124880838895e-06, "loss": 1.0757, "step": 27357 }, { "epoch": 5.21, "learning_rate": 1.4380552907530981e-06, "loss": 1.3082, "step": 27358 }, { "epoch": 5.21, "learning_rate": 1.437998093422307e-06, "loss": 3.0711, "step": 27359 }, { "epoch": 5.21, "learning_rate": 1.4379408960915157e-06, "loss": 2.2375, "step": 27360 }, { "epoch": 5.21, "learning_rate": 1.4378836987607247e-06, "loss": 2.5143, "step": 27361 }, { "epoch": 5.21, "learning_rate": 1.4378265014299333e-06, "loss": 0.9659, "step": 27362 }, { "epoch": 5.21, "learning_rate": 1.437769304099142e-06, "loss": 2.0681, "step": 27363 }, { "epoch": 5.21, "learning_rate": 1.4377121067683509e-06, "loss": 2.3979, "step": 27364 }, { "epoch": 5.21, "learning_rate": 1.4376549094375595e-06, "loss": 2.999, "step": 27365 }, { "epoch": 5.21, "learning_rate": 1.4375977121067685e-06, "loss": 2.3119, "step": 27366 }, { "epoch": 5.21, "learning_rate": 1.4375405147759773e-06, "loss": 2.3627, "step": 27367 }, { "epoch": 5.21, "learning_rate": 1.4374833174451858e-06, "loss": 2.8156, "step": 27368 }, { "epoch": 5.21, "learning_rate": 1.4374261201143946e-06, "loss": 2.3535, "step": 27369 }, { "epoch": 5.21, "learning_rate": 1.4373689227836034e-06, "loss": 2.675, "step": 27370 }, { "epoch": 5.21, "learning_rate": 1.4373117254528122e-06, "loss": 1.8317, "step": 27371 }, { "epoch": 5.21, "learning_rate": 1.437254528122021e-06, "loss": 2.4857, "step": 27372 }, { "epoch": 5.21, "learning_rate": 1.4371973307912298e-06, "loss": 2.9781, "step": 27373 }, { "epoch": 5.21, "learning_rate": 1.4371401334604386e-06, "loss": 2.6619, "step": 27374 }, { "epoch": 5.21, "learning_rate": 1.4370829361296472e-06, "loss": 1.057, "step": 27375 }, { "epoch": 5.21, "learning_rate": 1.437025738798856e-06, "loss": 2.7015, "step": 27376 }, { "epoch": 5.21, "learning_rate": 1.436968541468065e-06, "loss": 3.1484, "step": 27377 }, { "epoch": 5.21, "learning_rate": 1.4369113441372738e-06, "loss": 2.0369, "step": 27378 }, { "epoch": 5.22, "learning_rate": 1.4368541468064824e-06, "loss": 1.9702, "step": 27379 }, { "epoch": 5.22, "learning_rate": 1.4367969494756912e-06, "loss": 1.7479, "step": 27380 }, { "epoch": 5.22, "learning_rate": 1.4367397521449e-06, "loss": 2.9235, "step": 27381 }, { "epoch": 5.22, "learning_rate": 1.4366825548141087e-06, "loss": 2.3319, "step": 27382 }, { "epoch": 5.22, "learning_rate": 1.4366253574833175e-06, "loss": 0.9856, "step": 27383 }, { "epoch": 5.22, "learning_rate": 1.4365681601525263e-06, "loss": 1.894, "step": 27384 }, { "epoch": 5.22, "learning_rate": 1.436510962821735e-06, "loss": 2.8531, "step": 27385 }, { "epoch": 5.22, "learning_rate": 1.4364537654909437e-06, "loss": 1.8212, "step": 27386 }, { "epoch": 5.22, "learning_rate": 1.4363965681601525e-06, "loss": 2.0572, "step": 27387 }, { "epoch": 5.22, "learning_rate": 1.4363393708293613e-06, "loss": 2.0143, "step": 27388 }, { "epoch": 5.22, "learning_rate": 1.43628217349857e-06, "loss": 1.7183, "step": 27389 }, { "epoch": 5.22, "learning_rate": 1.4362249761677789e-06, "loss": 2.7291, "step": 27390 }, { "epoch": 5.22, "learning_rate": 1.4361677788369877e-06, "loss": 2.6179, "step": 27391 }, { "epoch": 5.22, "learning_rate": 1.4361105815061963e-06, "loss": 1.7168, "step": 27392 }, { "epoch": 5.22, "learning_rate": 1.436053384175405e-06, "loss": 0.8453, "step": 27393 }, { "epoch": 5.22, "learning_rate": 1.435996186844614e-06, "loss": 2.5863, "step": 27394 }, { "epoch": 5.22, "learning_rate": 1.4359389895138229e-06, "loss": 1.742, "step": 27395 }, { "epoch": 5.22, "learning_rate": 1.4358817921830314e-06, "loss": 2.0678, "step": 27396 }, { "epoch": 5.22, "learning_rate": 1.4358245948522402e-06, "loss": 2.3175, "step": 27397 }, { "epoch": 5.22, "learning_rate": 1.435767397521449e-06, "loss": 1.9446, "step": 27398 }, { "epoch": 5.22, "learning_rate": 1.4357102001906578e-06, "loss": 2.3915, "step": 27399 }, { "epoch": 5.22, "learning_rate": 1.4356530028598666e-06, "loss": 1.829, "step": 27400 }, { "epoch": 5.22, "learning_rate": 1.4355958055290754e-06, "loss": 1.4803, "step": 27401 }, { "epoch": 5.22, "learning_rate": 1.4355386081982842e-06, "loss": 2.1635, "step": 27402 }, { "epoch": 5.22, "learning_rate": 1.4354814108674928e-06, "loss": 2.5695, "step": 27403 }, { "epoch": 5.22, "learning_rate": 1.4354242135367016e-06, "loss": 1.801, "step": 27404 }, { "epoch": 5.22, "learning_rate": 1.4353670162059106e-06, "loss": 1.7919, "step": 27405 }, { "epoch": 5.22, "learning_rate": 1.4353098188751192e-06, "loss": 1.034, "step": 27406 }, { "epoch": 5.22, "learning_rate": 1.435252621544328e-06, "loss": 2.0383, "step": 27407 }, { "epoch": 5.22, "learning_rate": 1.4351954242135368e-06, "loss": 1.3781, "step": 27408 }, { "epoch": 5.22, "learning_rate": 1.4351382268827453e-06, "loss": 1.3546, "step": 27409 }, { "epoch": 5.22, "learning_rate": 1.4350810295519543e-06, "loss": 2.732, "step": 27410 }, { "epoch": 5.22, "learning_rate": 1.4350238322211631e-06, "loss": 2.9373, "step": 27411 }, { "epoch": 5.22, "learning_rate": 1.434966634890372e-06, "loss": 1.8604, "step": 27412 }, { "epoch": 5.22, "learning_rate": 1.4349094375595805e-06, "loss": 2.778, "step": 27413 }, { "epoch": 5.22, "learning_rate": 1.4348522402287893e-06, "loss": 2.1442, "step": 27414 }, { "epoch": 5.22, "learning_rate": 1.434795042897998e-06, "loss": 1.2166, "step": 27415 }, { "epoch": 5.22, "learning_rate": 1.4347378455672069e-06, "loss": 2.475, "step": 27416 }, { "epoch": 5.22, "learning_rate": 1.4346806482364157e-06, "loss": 1.6728, "step": 27417 }, { "epoch": 5.22, "learning_rate": 1.4346234509056245e-06, "loss": 0.9932, "step": 27418 }, { "epoch": 5.22, "learning_rate": 1.4345662535748333e-06, "loss": 2.1005, "step": 27419 }, { "epoch": 5.22, "learning_rate": 1.4345090562440419e-06, "loss": 1.5225, "step": 27420 }, { "epoch": 5.22, "learning_rate": 1.4344518589132509e-06, "loss": 2.9681, "step": 27421 }, { "epoch": 5.22, "learning_rate": 1.4343946615824597e-06, "loss": 2.2574, "step": 27422 }, { "epoch": 5.22, "learning_rate": 1.4343374642516682e-06, "loss": 2.6306, "step": 27423 }, { "epoch": 5.22, "learning_rate": 1.434280266920877e-06, "loss": 1.9327, "step": 27424 }, { "epoch": 5.22, "learning_rate": 1.4342230695900858e-06, "loss": 2.1915, "step": 27425 }, { "epoch": 5.22, "learning_rate": 1.4341658722592946e-06, "loss": 2.0058, "step": 27426 }, { "epoch": 5.22, "learning_rate": 1.4341086749285034e-06, "loss": 2.5221, "step": 27427 }, { "epoch": 5.22, "learning_rate": 1.4340514775977122e-06, "loss": 1.9484, "step": 27428 }, { "epoch": 5.22, "learning_rate": 1.433994280266921e-06, "loss": 2.4778, "step": 27429 }, { "epoch": 5.22, "learning_rate": 1.4339370829361296e-06, "loss": 1.5926, "step": 27430 }, { "epoch": 5.22, "learning_rate": 1.4338798856053384e-06, "loss": 2.8807, "step": 27431 }, { "epoch": 5.23, "learning_rate": 1.4338226882745472e-06, "loss": 2.3773, "step": 27432 }, { "epoch": 5.23, "learning_rate": 1.433765490943756e-06, "loss": 1.4568, "step": 27433 }, { "epoch": 5.23, "learning_rate": 1.4337082936129648e-06, "loss": 1.7907, "step": 27434 }, { "epoch": 5.23, "learning_rate": 1.4336510962821735e-06, "loss": 0.4889, "step": 27435 }, { "epoch": 5.23, "learning_rate": 1.4335938989513823e-06, "loss": 1.3436, "step": 27436 }, { "epoch": 5.23, "learning_rate": 1.433536701620591e-06, "loss": 2.0903, "step": 27437 }, { "epoch": 5.23, "learning_rate": 1.4334795042898e-06, "loss": 2.1548, "step": 27438 }, { "epoch": 5.23, "learning_rate": 1.4334223069590087e-06, "loss": 2.5938, "step": 27439 }, { "epoch": 5.23, "learning_rate": 1.4333651096282173e-06, "loss": 1.5883, "step": 27440 }, { "epoch": 5.23, "learning_rate": 1.433307912297426e-06, "loss": 1.6875, "step": 27441 }, { "epoch": 5.23, "learning_rate": 1.433250714966635e-06, "loss": 1.4051, "step": 27442 }, { "epoch": 5.23, "learning_rate": 1.4331935176358437e-06, "loss": 1.2502, "step": 27443 }, { "epoch": 5.23, "learning_rate": 1.4331363203050525e-06, "loss": 2.2633, "step": 27444 }, { "epoch": 5.23, "learning_rate": 1.4330791229742613e-06, "loss": 2.3189, "step": 27445 }, { "epoch": 5.23, "learning_rate": 1.43302192564347e-06, "loss": 1.3086, "step": 27446 }, { "epoch": 5.23, "learning_rate": 1.4329647283126787e-06, "loss": 2.3941, "step": 27447 }, { "epoch": 5.23, "learning_rate": 1.4329075309818874e-06, "loss": 1.6731, "step": 27448 }, { "epoch": 5.23, "learning_rate": 1.4328503336510965e-06, "loss": 1.0468, "step": 27449 }, { "epoch": 5.23, "learning_rate": 1.432793136320305e-06, "loss": 1.6131, "step": 27450 }, { "epoch": 5.23, "learning_rate": 1.4327359389895138e-06, "loss": 1.6378, "step": 27451 }, { "epoch": 5.23, "learning_rate": 1.4326787416587226e-06, "loss": 1.2371, "step": 27452 }, { "epoch": 5.23, "learning_rate": 1.4326215443279314e-06, "loss": 0.6461, "step": 27453 }, { "epoch": 5.23, "learning_rate": 1.4325643469971402e-06, "loss": 2.3723, "step": 27454 }, { "epoch": 5.23, "learning_rate": 1.432507149666349e-06, "loss": 2.6944, "step": 27455 }, { "epoch": 5.23, "learning_rate": 1.4324499523355578e-06, "loss": 2.0973, "step": 27456 }, { "epoch": 5.23, "learning_rate": 1.4323927550047664e-06, "loss": 0.9578, "step": 27457 }, { "epoch": 5.23, "learning_rate": 1.4323355576739752e-06, "loss": 1.5648, "step": 27458 }, { "epoch": 5.23, "learning_rate": 1.432278360343184e-06, "loss": 2.7381, "step": 27459 }, { "epoch": 5.23, "learning_rate": 1.4322211630123928e-06, "loss": 0.9004, "step": 27460 }, { "epoch": 5.23, "learning_rate": 1.4321639656816016e-06, "loss": 1.514, "step": 27461 }, { "epoch": 5.23, "learning_rate": 1.4321067683508103e-06, "loss": 2.1627, "step": 27462 }, { "epoch": 5.23, "learning_rate": 1.4320495710200191e-06, "loss": 1.7042, "step": 27463 }, { "epoch": 5.23, "learning_rate": 1.4319923736892277e-06, "loss": 2.3181, "step": 27464 }, { "epoch": 5.23, "learning_rate": 1.4319351763584367e-06, "loss": 3.1498, "step": 27465 }, { "epoch": 5.23, "learning_rate": 1.4318779790276455e-06, "loss": 0.6717, "step": 27466 }, { "epoch": 5.23, "learning_rate": 1.4318207816968543e-06, "loss": 2.2498, "step": 27467 }, { "epoch": 5.23, "learning_rate": 1.431763584366063e-06, "loss": 2.7904, "step": 27468 }, { "epoch": 5.23, "learning_rate": 1.4317063870352717e-06, "loss": 2.4344, "step": 27469 }, { "epoch": 5.23, "learning_rate": 1.4316491897044805e-06, "loss": 1.3836, "step": 27470 }, { "epoch": 5.23, "learning_rate": 1.4315919923736893e-06, "loss": 2.1743, "step": 27471 }, { "epoch": 5.23, "learning_rate": 1.431534795042898e-06, "loss": 1.3691, "step": 27472 }, { "epoch": 5.23, "learning_rate": 1.4314775977121069e-06, "loss": 1.0879, "step": 27473 }, { "epoch": 5.23, "learning_rate": 1.4314204003813154e-06, "loss": 2.1786, "step": 27474 }, { "epoch": 5.23, "learning_rate": 1.4313632030505242e-06, "loss": 1.7135, "step": 27475 }, { "epoch": 5.23, "learning_rate": 1.431306005719733e-06, "loss": 2.119, "step": 27476 }, { "epoch": 5.23, "learning_rate": 1.431248808388942e-06, "loss": 0.7575, "step": 27477 }, { "epoch": 5.23, "learning_rate": 1.4311916110581506e-06, "loss": 2.1708, "step": 27478 }, { "epoch": 5.23, "learning_rate": 1.4311344137273594e-06, "loss": 2.5232, "step": 27479 }, { "epoch": 5.23, "learning_rate": 1.4310772163965682e-06, "loss": 1.7501, "step": 27480 }, { "epoch": 5.23, "learning_rate": 1.4310200190657768e-06, "loss": 1.8016, "step": 27481 }, { "epoch": 5.23, "learning_rate": 1.4309628217349858e-06, "loss": 2.3989, "step": 27482 }, { "epoch": 5.23, "learning_rate": 1.4309056244041946e-06, "loss": 0.9836, "step": 27483 }, { "epoch": 5.24, "learning_rate": 1.4308484270734034e-06, "loss": 1.6378, "step": 27484 }, { "epoch": 5.24, "learning_rate": 1.430791229742612e-06, "loss": 1.6531, "step": 27485 }, { "epoch": 5.24, "learning_rate": 1.4307340324118208e-06, "loss": 2.2753, "step": 27486 }, { "epoch": 5.24, "learning_rate": 1.4306768350810296e-06, "loss": 2.0801, "step": 27487 }, { "epoch": 5.24, "learning_rate": 1.4306196377502384e-06, "loss": 2.3737, "step": 27488 }, { "epoch": 5.24, "learning_rate": 1.4305624404194471e-06, "loss": 1.4576, "step": 27489 }, { "epoch": 5.24, "learning_rate": 1.430505243088656e-06, "loss": 1.733, "step": 27490 }, { "epoch": 5.24, "learning_rate": 1.4304480457578645e-06, "loss": 2.4242, "step": 27491 }, { "epoch": 5.24, "learning_rate": 1.4303908484270733e-06, "loss": 1.9548, "step": 27492 }, { "epoch": 5.24, "learning_rate": 1.4303336510962823e-06, "loss": 2.8142, "step": 27493 }, { "epoch": 5.24, "learning_rate": 1.4302764537654911e-06, "loss": 1.3241, "step": 27494 }, { "epoch": 5.24, "learning_rate": 1.4302192564346997e-06, "loss": 2.6149, "step": 27495 }, { "epoch": 5.24, "learning_rate": 1.4301620591039085e-06, "loss": 1.9578, "step": 27496 }, { "epoch": 5.24, "learning_rate": 1.4301048617731173e-06, "loss": 2.5481, "step": 27497 }, { "epoch": 5.24, "learning_rate": 1.430047664442326e-06, "loss": 2.0568, "step": 27498 }, { "epoch": 5.24, "learning_rate": 1.4299904671115349e-06, "loss": 1.4395, "step": 27499 }, { "epoch": 5.24, "learning_rate": 1.4299332697807437e-06, "loss": 2.9661, "step": 27500 }, { "epoch": 5.24, "learning_rate": 1.4298760724499525e-06, "loss": 2.9605, "step": 27501 }, { "epoch": 5.24, "learning_rate": 1.429818875119161e-06, "loss": 1.2328, "step": 27502 }, { "epoch": 5.24, "learning_rate": 1.4297616777883698e-06, "loss": 2.0584, "step": 27503 }, { "epoch": 5.24, "learning_rate": 1.4297044804575786e-06, "loss": 2.8197, "step": 27504 }, { "epoch": 5.24, "learning_rate": 1.4296472831267874e-06, "loss": 2.0218, "step": 27505 }, { "epoch": 5.24, "learning_rate": 1.4295900857959962e-06, "loss": 2.6397, "step": 27506 }, { "epoch": 5.24, "learning_rate": 1.429532888465205e-06, "loss": 3.0578, "step": 27507 }, { "epoch": 5.24, "learning_rate": 1.4294756911344138e-06, "loss": 1.2582, "step": 27508 }, { "epoch": 5.24, "learning_rate": 1.4294184938036226e-06, "loss": 1.1497, "step": 27509 }, { "epoch": 5.24, "learning_rate": 1.4293612964728314e-06, "loss": 2.0993, "step": 27510 }, { "epoch": 5.24, "learning_rate": 1.4293040991420402e-06, "loss": 3.0674, "step": 27511 }, { "epoch": 5.24, "learning_rate": 1.4292469018112488e-06, "loss": 1.4576, "step": 27512 }, { "epoch": 5.24, "learning_rate": 1.4291897044804576e-06, "loss": 3.018, "step": 27513 }, { "epoch": 5.24, "learning_rate": 1.4291325071496664e-06, "loss": 1.8562, "step": 27514 }, { "epoch": 5.24, "learning_rate": 1.4290753098188751e-06, "loss": 1.1468, "step": 27515 }, { "epoch": 5.24, "learning_rate": 1.429018112488084e-06, "loss": 2.3706, "step": 27516 }, { "epoch": 5.24, "learning_rate": 1.4289609151572927e-06, "loss": 2.4181, "step": 27517 }, { "epoch": 5.24, "learning_rate": 1.4289037178265015e-06, "loss": 1.0229, "step": 27518 }, { "epoch": 5.24, "learning_rate": 1.4288465204957101e-06, "loss": 2.3699, "step": 27519 }, { "epoch": 5.24, "learning_rate": 1.428789323164919e-06, "loss": 1.7918, "step": 27520 }, { "epoch": 5.24, "learning_rate": 1.428732125834128e-06, "loss": 1.5972, "step": 27521 }, { "epoch": 5.24, "learning_rate": 1.4286749285033365e-06, "loss": 1.8027, "step": 27522 }, { "epoch": 5.24, "learning_rate": 1.4286177311725453e-06, "loss": 1.6124, "step": 27523 }, { "epoch": 5.24, "learning_rate": 1.428560533841754e-06, "loss": 3.086, "step": 27524 }, { "epoch": 5.24, "learning_rate": 1.4285033365109629e-06, "loss": 0.8338, "step": 27525 }, { "epoch": 5.24, "learning_rate": 1.4284461391801717e-06, "loss": 1.654, "step": 27526 }, { "epoch": 5.24, "learning_rate": 1.4283889418493805e-06, "loss": 1.3243, "step": 27527 }, { "epoch": 5.24, "learning_rate": 1.4283317445185893e-06, "loss": 2.0961, "step": 27528 }, { "epoch": 5.24, "learning_rate": 1.4282745471877978e-06, "loss": 3.2679, "step": 27529 }, { "epoch": 5.24, "learning_rate": 1.4282173498570066e-06, "loss": 1.9375, "step": 27530 }, { "epoch": 5.24, "learning_rate": 1.4281601525262154e-06, "loss": 2.2312, "step": 27531 }, { "epoch": 5.24, "learning_rate": 1.4281029551954244e-06, "loss": 1.4756, "step": 27532 }, { "epoch": 5.24, "learning_rate": 1.428045757864633e-06, "loss": 1.5114, "step": 27533 }, { "epoch": 5.24, "learning_rate": 1.4279885605338418e-06, "loss": 2.6116, "step": 27534 }, { "epoch": 5.24, "learning_rate": 1.4279313632030506e-06, "loss": 1.1155, "step": 27535 }, { "epoch": 5.24, "learning_rate": 1.4278741658722592e-06, "loss": 1.2506, "step": 27536 }, { "epoch": 5.25, "learning_rate": 1.4278169685414682e-06, "loss": 1.6684, "step": 27537 }, { "epoch": 5.25, "learning_rate": 1.427759771210677e-06, "loss": 0.9699, "step": 27538 }, { "epoch": 5.25, "learning_rate": 1.4277025738798856e-06, "loss": 3.3249, "step": 27539 }, { "epoch": 5.25, "learning_rate": 1.4276453765490944e-06, "loss": 1.0174, "step": 27540 }, { "epoch": 5.25, "learning_rate": 1.4275881792183032e-06, "loss": 1.6745, "step": 27541 }, { "epoch": 5.25, "learning_rate": 1.427530981887512e-06, "loss": 2.5782, "step": 27542 }, { "epoch": 5.25, "learning_rate": 1.4274737845567207e-06, "loss": 3.5296, "step": 27543 }, { "epoch": 5.25, "learning_rate": 1.4274165872259295e-06, "loss": 2.9843, "step": 27544 }, { "epoch": 5.25, "learning_rate": 1.4273593898951383e-06, "loss": 1.4992, "step": 27545 }, { "epoch": 5.25, "learning_rate": 1.427302192564347e-06, "loss": 2.0139, "step": 27546 }, { "epoch": 5.25, "learning_rate": 1.4272449952335557e-06, "loss": 2.9653, "step": 27547 }, { "epoch": 5.25, "learning_rate": 1.4271877979027645e-06, "loss": 4.2751, "step": 27548 }, { "epoch": 5.25, "learning_rate": 1.4271306005719735e-06, "loss": 1.1719, "step": 27549 }, { "epoch": 5.25, "learning_rate": 1.427073403241182e-06, "loss": 1.5129, "step": 27550 }, { "epoch": 5.25, "learning_rate": 1.4270162059103909e-06, "loss": 2.0782, "step": 27551 }, { "epoch": 5.25, "learning_rate": 1.4269590085795997e-06, "loss": 1.6162, "step": 27552 }, { "epoch": 5.25, "learning_rate": 1.4269018112488085e-06, "loss": 2.615, "step": 27553 }, { "epoch": 5.25, "learning_rate": 1.4268446139180173e-06, "loss": 2.0236, "step": 27554 }, { "epoch": 5.25, "learning_rate": 1.426787416587226e-06, "loss": 2.475, "step": 27555 }, { "epoch": 5.25, "learning_rate": 1.4267302192564346e-06, "loss": 2.3619, "step": 27556 }, { "epoch": 5.25, "learning_rate": 1.4266730219256434e-06, "loss": 1.9945, "step": 27557 }, { "epoch": 5.25, "learning_rate": 1.4266158245948522e-06, "loss": 2.9006, "step": 27558 }, { "epoch": 5.25, "learning_rate": 1.426558627264061e-06, "loss": 0.9217, "step": 27559 }, { "epoch": 5.25, "learning_rate": 1.4265014299332698e-06, "loss": 2.9416, "step": 27560 }, { "epoch": 5.25, "learning_rate": 1.4264442326024786e-06, "loss": 2.8426, "step": 27561 }, { "epoch": 5.25, "learning_rate": 1.4263870352716874e-06, "loss": 2.6678, "step": 27562 }, { "epoch": 5.25, "learning_rate": 1.426329837940896e-06, "loss": 3.1326, "step": 27563 }, { "epoch": 5.25, "learning_rate": 1.4262726406101048e-06, "loss": 1.5575, "step": 27564 }, { "epoch": 5.25, "learning_rate": 1.4262154432793138e-06, "loss": 1.5885, "step": 27565 }, { "epoch": 5.25, "learning_rate": 1.4261582459485226e-06, "loss": 2.9948, "step": 27566 }, { "epoch": 5.25, "learning_rate": 1.4261010486177312e-06, "loss": 1.5242, "step": 27567 }, { "epoch": 5.25, "learning_rate": 1.42604385128694e-06, "loss": 1.596, "step": 27568 }, { "epoch": 5.25, "learning_rate": 1.4259866539561487e-06, "loss": 1.6606, "step": 27569 }, { "epoch": 5.25, "learning_rate": 1.4259294566253575e-06, "loss": 1.2519, "step": 27570 }, { "epoch": 5.25, "learning_rate": 1.4258722592945663e-06, "loss": 2.1557, "step": 27571 }, { "epoch": 5.25, "learning_rate": 1.4258150619637751e-06, "loss": 2.0076, "step": 27572 }, { "epoch": 5.25, "learning_rate": 1.425757864632984e-06, "loss": 2.3636, "step": 27573 }, { "epoch": 5.25, "learning_rate": 1.4257006673021925e-06, "loss": 1.557, "step": 27574 }, { "epoch": 5.25, "learning_rate": 1.4256434699714013e-06, "loss": 1.7319, "step": 27575 }, { "epoch": 5.25, "learning_rate": 1.4255862726406103e-06, "loss": 3.2406, "step": 27576 }, { "epoch": 5.25, "learning_rate": 1.4255290753098189e-06, "loss": 1.5744, "step": 27577 }, { "epoch": 5.25, "learning_rate": 1.4254718779790277e-06, "loss": 0.9504, "step": 27578 }, { "epoch": 5.25, "learning_rate": 1.4254146806482365e-06, "loss": 1.6491, "step": 27579 }, { "epoch": 5.25, "learning_rate": 1.425357483317445e-06, "loss": 1.8191, "step": 27580 }, { "epoch": 5.25, "learning_rate": 1.425300285986654e-06, "loss": 2.1616, "step": 27581 }, { "epoch": 5.25, "learning_rate": 1.4252430886558629e-06, "loss": 2.027, "step": 27582 }, { "epoch": 5.25, "learning_rate": 1.4251858913250716e-06, "loss": 1.5628, "step": 27583 }, { "epoch": 5.25, "learning_rate": 1.4251286939942802e-06, "loss": 1.6801, "step": 27584 }, { "epoch": 5.25, "learning_rate": 1.425071496663489e-06, "loss": 1.868, "step": 27585 }, { "epoch": 5.25, "learning_rate": 1.4250142993326978e-06, "loss": 1.9557, "step": 27586 }, { "epoch": 5.25, "learning_rate": 1.4249571020019066e-06, "loss": 1.4986, "step": 27587 }, { "epoch": 5.25, "learning_rate": 1.4248999046711154e-06, "loss": 2.4297, "step": 27588 }, { "epoch": 5.26, "learning_rate": 1.4248427073403242e-06, "loss": 1.2403, "step": 27589 }, { "epoch": 5.26, "learning_rate": 1.424785510009533e-06, "loss": 1.3788, "step": 27590 }, { "epoch": 5.26, "learning_rate": 1.4247283126787416e-06, "loss": 1.9767, "step": 27591 }, { "epoch": 5.26, "learning_rate": 1.4246711153479504e-06, "loss": 1.7002, "step": 27592 }, { "epoch": 5.26, "learning_rate": 1.4246139180171594e-06, "loss": 1.306, "step": 27593 }, { "epoch": 5.26, "learning_rate": 1.424556720686368e-06, "loss": 2.1848, "step": 27594 }, { "epoch": 5.26, "learning_rate": 1.4244995233555767e-06, "loss": 1.9654, "step": 27595 }, { "epoch": 5.26, "learning_rate": 1.4244423260247855e-06, "loss": 1.6128, "step": 27596 }, { "epoch": 5.26, "learning_rate": 1.4243851286939943e-06, "loss": 2.4107, "step": 27597 }, { "epoch": 5.26, "learning_rate": 1.4243279313632031e-06, "loss": 0.9847, "step": 27598 }, { "epoch": 5.26, "learning_rate": 1.424270734032412e-06, "loss": 1.5021, "step": 27599 }, { "epoch": 5.26, "learning_rate": 1.4242135367016207e-06, "loss": 2.5031, "step": 27600 }, { "epoch": 5.26, "learning_rate": 1.4241563393708293e-06, "loss": 1.3875, "step": 27601 }, { "epoch": 5.26, "learning_rate": 1.424099142040038e-06, "loss": 1.389, "step": 27602 }, { "epoch": 5.26, "learning_rate": 1.4240419447092469e-06, "loss": 0.6226, "step": 27603 }, { "epoch": 5.26, "learning_rate": 1.4239847473784557e-06, "loss": 2.3456, "step": 27604 }, { "epoch": 5.26, "learning_rate": 1.4239275500476645e-06, "loss": 1.82, "step": 27605 }, { "epoch": 5.26, "learning_rate": 1.4238703527168733e-06, "loss": 2.3241, "step": 27606 }, { "epoch": 5.26, "learning_rate": 1.423813155386082e-06, "loss": 2.6826, "step": 27607 }, { "epoch": 5.26, "learning_rate": 1.4237559580552906e-06, "loss": 1.8583, "step": 27608 }, { "epoch": 5.26, "learning_rate": 1.4236987607244997e-06, "loss": 1.2088, "step": 27609 }, { "epoch": 5.26, "learning_rate": 1.4236415633937084e-06, "loss": 2.3122, "step": 27610 }, { "epoch": 5.26, "learning_rate": 1.423584366062917e-06, "loss": 2.3241, "step": 27611 }, { "epoch": 5.26, "learning_rate": 1.4235271687321258e-06, "loss": 2.224, "step": 27612 }, { "epoch": 5.26, "learning_rate": 1.4234699714013346e-06, "loss": 2.8141, "step": 27613 }, { "epoch": 5.26, "learning_rate": 1.4234127740705434e-06, "loss": 1.9334, "step": 27614 }, { "epoch": 5.26, "learning_rate": 1.4233555767397522e-06, "loss": 2.0721, "step": 27615 }, { "epoch": 5.26, "learning_rate": 1.423298379408961e-06, "loss": 3.4421, "step": 27616 }, { "epoch": 5.26, "learning_rate": 1.4232411820781698e-06, "loss": 1.9022, "step": 27617 }, { "epoch": 5.26, "learning_rate": 1.4231839847473784e-06, "loss": 1.0714, "step": 27618 }, { "epoch": 5.26, "learning_rate": 1.4231267874165872e-06, "loss": 1.9018, "step": 27619 }, { "epoch": 5.26, "learning_rate": 1.4230695900857962e-06, "loss": 1.2675, "step": 27620 }, { "epoch": 5.26, "learning_rate": 1.4230123927550048e-06, "loss": 2.2843, "step": 27621 }, { "epoch": 5.26, "learning_rate": 1.4229551954242135e-06, "loss": 1.5478, "step": 27622 }, { "epoch": 5.26, "learning_rate": 1.4228979980934223e-06, "loss": 2.6539, "step": 27623 }, { "epoch": 5.26, "learning_rate": 1.4228408007626311e-06, "loss": 1.6979, "step": 27624 }, { "epoch": 5.26, "learning_rate": 1.42278360343184e-06, "loss": 1.8011, "step": 27625 }, { "epoch": 5.26, "learning_rate": 1.4227264061010487e-06, "loss": 2.3193, "step": 27626 }, { "epoch": 5.26, "learning_rate": 1.4226692087702575e-06, "loss": 1.5015, "step": 27627 }, { "epoch": 5.26, "learning_rate": 1.422612011439466e-06, "loss": 2.3967, "step": 27628 }, { "epoch": 5.26, "learning_rate": 1.4225548141086749e-06, "loss": 1.7631, "step": 27629 }, { "epoch": 5.26, "learning_rate": 1.4224976167778837e-06, "loss": 2.592, "step": 27630 }, { "epoch": 5.26, "learning_rate": 1.4224404194470925e-06, "loss": 2.8728, "step": 27631 }, { "epoch": 5.26, "learning_rate": 1.4223832221163013e-06, "loss": 2.8558, "step": 27632 }, { "epoch": 5.26, "learning_rate": 1.42232602478551e-06, "loss": 1.6521, "step": 27633 }, { "epoch": 5.26, "learning_rate": 1.4222688274547189e-06, "loss": 0.7647, "step": 27634 }, { "epoch": 5.26, "learning_rate": 1.4222116301239274e-06, "loss": 1.9818, "step": 27635 }, { "epoch": 5.26, "learning_rate": 1.4221544327931362e-06, "loss": 1.253, "step": 27636 }, { "epoch": 5.26, "learning_rate": 1.4220972354623452e-06, "loss": 2.8863, "step": 27637 }, { "epoch": 5.26, "learning_rate": 1.422040038131554e-06, "loss": 3.5138, "step": 27638 }, { "epoch": 5.26, "learning_rate": 1.4219828408007626e-06, "loss": 1.6921, "step": 27639 }, { "epoch": 5.26, "learning_rate": 1.4219256434699714e-06, "loss": 1.6338, "step": 27640 }, { "epoch": 5.26, "learning_rate": 1.4218684461391802e-06, "loss": 2.4059, "step": 27641 }, { "epoch": 5.27, "learning_rate": 1.421811248808389e-06, "loss": 1.0303, "step": 27642 }, { "epoch": 5.27, "learning_rate": 1.4217540514775978e-06, "loss": 2.3269, "step": 27643 }, { "epoch": 5.27, "learning_rate": 1.4216968541468066e-06, "loss": 1.5857, "step": 27644 }, { "epoch": 5.27, "learning_rate": 1.4216396568160152e-06, "loss": 1.9405, "step": 27645 }, { "epoch": 5.27, "learning_rate": 1.421582459485224e-06, "loss": 2.8442, "step": 27646 }, { "epoch": 5.27, "learning_rate": 1.4215252621544328e-06, "loss": 1.4962, "step": 27647 }, { "epoch": 5.27, "learning_rate": 1.4214680648236418e-06, "loss": 1.3382, "step": 27648 }, { "epoch": 5.27, "learning_rate": 1.4214108674928503e-06, "loss": 1.427, "step": 27649 }, { "epoch": 5.27, "learning_rate": 1.4213536701620591e-06, "loss": 3.2932, "step": 27650 }, { "epoch": 5.27, "learning_rate": 1.421296472831268e-06, "loss": 3.3483, "step": 27651 }, { "epoch": 5.27, "learning_rate": 1.4212392755004765e-06, "loss": 1.3791, "step": 27652 }, { "epoch": 5.27, "learning_rate": 1.4211820781696855e-06, "loss": 1.464, "step": 27653 }, { "epoch": 5.27, "learning_rate": 1.4211248808388943e-06, "loss": 1.0374, "step": 27654 }, { "epoch": 5.27, "learning_rate": 1.4210676835081031e-06, "loss": 1.3666, "step": 27655 }, { "epoch": 5.27, "learning_rate": 1.4210104861773117e-06, "loss": 2.1497, "step": 27656 }, { "epoch": 5.27, "learning_rate": 1.4209532888465205e-06, "loss": 1.1524, "step": 27657 }, { "epoch": 5.27, "learning_rate": 1.4208960915157293e-06, "loss": 1.8463, "step": 27658 }, { "epoch": 5.27, "learning_rate": 1.420838894184938e-06, "loss": 1.3356, "step": 27659 }, { "epoch": 5.27, "learning_rate": 1.4207816968541469e-06, "loss": 1.7834, "step": 27660 }, { "epoch": 5.27, "learning_rate": 1.4207244995233557e-06, "loss": 2.3353, "step": 27661 }, { "epoch": 5.27, "learning_rate": 1.4206673021925645e-06, "loss": 1.9542, "step": 27662 }, { "epoch": 5.27, "learning_rate": 1.420610104861773e-06, "loss": 1.1034, "step": 27663 }, { "epoch": 5.27, "learning_rate": 1.420552907530982e-06, "loss": 2.2738, "step": 27664 }, { "epoch": 5.27, "learning_rate": 1.4204957102001908e-06, "loss": 0.8917, "step": 27665 }, { "epoch": 5.27, "learning_rate": 1.4204385128693994e-06, "loss": 2.6656, "step": 27666 }, { "epoch": 5.27, "learning_rate": 1.4203813155386082e-06, "loss": 3.1879, "step": 27667 }, { "epoch": 5.27, "learning_rate": 1.420324118207817e-06, "loss": 2.5764, "step": 27668 }, { "epoch": 5.27, "learning_rate": 1.4202669208770258e-06, "loss": 1.0704, "step": 27669 }, { "epoch": 5.27, "learning_rate": 1.4202097235462346e-06, "loss": 2.5872, "step": 27670 }, { "epoch": 5.27, "learning_rate": 1.4201525262154434e-06, "loss": 1.3268, "step": 27671 }, { "epoch": 5.27, "learning_rate": 1.4200953288846522e-06, "loss": 0.8004, "step": 27672 }, { "epoch": 5.27, "learning_rate": 1.4200381315538608e-06, "loss": 2.02, "step": 27673 }, { "epoch": 5.27, "learning_rate": 1.4199809342230696e-06, "loss": 0.6905, "step": 27674 }, { "epoch": 5.27, "learning_rate": 1.4199237368922783e-06, "loss": 0.859, "step": 27675 }, { "epoch": 5.27, "learning_rate": 1.4198665395614871e-06, "loss": 1.8896, "step": 27676 }, { "epoch": 5.27, "learning_rate": 1.419809342230696e-06, "loss": 1.726, "step": 27677 }, { "epoch": 5.27, "learning_rate": 1.4197521448999047e-06, "loss": 1.998, "step": 27678 }, { "epoch": 5.27, "learning_rate": 1.4196949475691135e-06, "loss": 1.3565, "step": 27679 }, { "epoch": 5.27, "learning_rate": 1.4196377502383221e-06, "loss": 2.3629, "step": 27680 }, { "epoch": 5.27, "learning_rate": 1.4195805529075311e-06, "loss": 2.2121, "step": 27681 }, { "epoch": 5.27, "learning_rate": 1.41952335557674e-06, "loss": 1.8523, "step": 27682 }, { "epoch": 5.27, "learning_rate": 1.4194661582459485e-06, "loss": 1.6597, "step": 27683 }, { "epoch": 5.27, "learning_rate": 1.4194089609151573e-06, "loss": 1.7118, "step": 27684 }, { "epoch": 5.27, "learning_rate": 1.419351763584366e-06, "loss": 2.3454, "step": 27685 }, { "epoch": 5.27, "learning_rate": 1.4192945662535749e-06, "loss": 2.2982, "step": 27686 }, { "epoch": 5.27, "learning_rate": 1.4192373689227837e-06, "loss": 1.7895, "step": 27687 }, { "epoch": 5.27, "learning_rate": 1.4191801715919925e-06, "loss": 2.0963, "step": 27688 }, { "epoch": 5.27, "learning_rate": 1.4191229742612013e-06, "loss": 2.1342, "step": 27689 }, { "epoch": 5.27, "learning_rate": 1.4190657769304098e-06, "loss": 2.3081, "step": 27690 }, { "epoch": 5.27, "learning_rate": 1.4190085795996186e-06, "loss": 1.1124, "step": 27691 }, { "epoch": 5.27, "learning_rate": 1.4189513822688276e-06, "loss": 1.6129, "step": 27692 }, { "epoch": 5.27, "learning_rate": 1.4188941849380362e-06, "loss": 1.5361, "step": 27693 }, { "epoch": 5.28, "learning_rate": 1.418836987607245e-06, "loss": 2.9171, "step": 27694 }, { "epoch": 5.28, "learning_rate": 1.4187797902764538e-06, "loss": 2.9396, "step": 27695 }, { "epoch": 5.28, "learning_rate": 1.4187225929456626e-06, "loss": 1.796, "step": 27696 }, { "epoch": 5.28, "learning_rate": 1.4186653956148714e-06, "loss": 2.0127, "step": 27697 }, { "epoch": 5.28, "learning_rate": 1.4186081982840802e-06, "loss": 2.9743, "step": 27698 }, { "epoch": 5.28, "learning_rate": 1.418551000953289e-06, "loss": 2.006, "step": 27699 }, { "epoch": 5.28, "learning_rate": 1.4184938036224976e-06, "loss": 1.4039, "step": 27700 }, { "epoch": 5.28, "learning_rate": 1.4184366062917064e-06, "loss": 2.0925, "step": 27701 }, { "epoch": 5.28, "learning_rate": 1.4183794089609151e-06, "loss": 2.5077, "step": 27702 }, { "epoch": 5.28, "learning_rate": 1.418322211630124e-06, "loss": 2.0575, "step": 27703 }, { "epoch": 5.28, "learning_rate": 1.4182650142993327e-06, "loss": 2.3276, "step": 27704 }, { "epoch": 5.28, "learning_rate": 1.4182078169685415e-06, "loss": 2.1358, "step": 27705 }, { "epoch": 5.28, "learning_rate": 1.4181506196377503e-06, "loss": 1.424, "step": 27706 }, { "epoch": 5.28, "learning_rate": 1.418093422306959e-06, "loss": 3.7408, "step": 27707 }, { "epoch": 5.28, "learning_rate": 1.418036224976168e-06, "loss": 1.5164, "step": 27708 }, { "epoch": 5.28, "learning_rate": 1.4179790276453767e-06, "loss": 1.3039, "step": 27709 }, { "epoch": 5.28, "learning_rate": 1.4179218303145853e-06, "loss": 1.1736, "step": 27710 }, { "epoch": 5.28, "learning_rate": 1.417864632983794e-06, "loss": 1.99, "step": 27711 }, { "epoch": 5.28, "learning_rate": 1.4178074356530029e-06, "loss": 1.4718, "step": 27712 }, { "epoch": 5.28, "learning_rate": 1.4177502383222117e-06, "loss": 1.5476, "step": 27713 }, { "epoch": 5.28, "learning_rate": 1.4176930409914205e-06, "loss": 2.9206, "step": 27714 }, { "epoch": 5.28, "learning_rate": 1.4176358436606293e-06, "loss": 2.8324, "step": 27715 }, { "epoch": 5.28, "learning_rate": 1.417578646329838e-06, "loss": 2.6425, "step": 27716 }, { "epoch": 5.28, "learning_rate": 1.4175214489990466e-06, "loss": 2.8369, "step": 27717 }, { "epoch": 5.28, "learning_rate": 1.4174642516682554e-06, "loss": 1.7331, "step": 27718 }, { "epoch": 5.28, "learning_rate": 1.4174070543374642e-06, "loss": 1.7817, "step": 27719 }, { "epoch": 5.28, "learning_rate": 1.4173498570066732e-06, "loss": 2.7357, "step": 27720 }, { "epoch": 5.28, "learning_rate": 1.4172926596758818e-06, "loss": 2.4859, "step": 27721 }, { "epoch": 5.28, "learning_rate": 1.4172354623450906e-06, "loss": 2.8657, "step": 27722 }, { "epoch": 5.28, "learning_rate": 1.4171782650142994e-06, "loss": 1.2341, "step": 27723 }, { "epoch": 5.28, "learning_rate": 1.417121067683508e-06, "loss": 2.6268, "step": 27724 }, { "epoch": 5.28, "learning_rate": 1.417063870352717e-06, "loss": 1.7547, "step": 27725 }, { "epoch": 5.28, "learning_rate": 1.4170066730219258e-06, "loss": 1.7076, "step": 27726 }, { "epoch": 5.28, "learning_rate": 1.4169494756911346e-06, "loss": 1.7632, "step": 27727 }, { "epoch": 5.28, "learning_rate": 1.4168922783603432e-06, "loss": 2.0713, "step": 27728 }, { "epoch": 5.28, "learning_rate": 1.416835081029552e-06, "loss": 1.9677, "step": 27729 }, { "epoch": 5.28, "learning_rate": 1.4167778836987607e-06, "loss": 1.4453, "step": 27730 }, { "epoch": 5.28, "learning_rate": 1.4167206863679695e-06, "loss": 1.4106, "step": 27731 }, { "epoch": 5.28, "learning_rate": 1.4166634890371783e-06, "loss": 0.9865, "step": 27732 }, { "epoch": 5.28, "learning_rate": 1.4166062917063871e-06, "loss": 1.5521, "step": 27733 }, { "epoch": 5.28, "learning_rate": 1.4165490943755957e-06, "loss": 2.3511, "step": 27734 }, { "epoch": 5.28, "learning_rate": 1.4164918970448045e-06, "loss": 1.8579, "step": 27735 }, { "epoch": 5.28, "learning_rate": 1.4164346997140135e-06, "loss": 2.1029, "step": 27736 }, { "epoch": 5.28, "learning_rate": 1.4163775023832223e-06, "loss": 1.0718, "step": 27737 }, { "epoch": 5.28, "learning_rate": 1.4163203050524309e-06, "loss": 2.1374, "step": 27738 }, { "epoch": 5.28, "learning_rate": 1.4162631077216397e-06, "loss": 2.2392, "step": 27739 }, { "epoch": 5.28, "learning_rate": 1.4162059103908485e-06, "loss": 1.3481, "step": 27740 }, { "epoch": 5.28, "learning_rate": 1.4161487130600573e-06, "loss": 1.1078, "step": 27741 }, { "epoch": 5.28, "learning_rate": 1.416091515729266e-06, "loss": 2.2993, "step": 27742 }, { "epoch": 5.28, "learning_rate": 1.4160343183984748e-06, "loss": 1.4791, "step": 27743 }, { "epoch": 5.28, "learning_rate": 1.4159771210676836e-06, "loss": 3.0394, "step": 27744 }, { "epoch": 5.28, "learning_rate": 1.4159199237368922e-06, "loss": 2.9372, "step": 27745 }, { "epoch": 5.28, "learning_rate": 1.415862726406101e-06, "loss": 1.5726, "step": 27746 }, { "epoch": 5.29, "learning_rate": 1.4158055290753098e-06, "loss": 1.5575, "step": 27747 }, { "epoch": 5.29, "learning_rate": 1.4157483317445186e-06, "loss": 1.4851, "step": 27748 }, { "epoch": 5.29, "learning_rate": 1.4156911344137274e-06, "loss": 1.6116, "step": 27749 }, { "epoch": 5.29, "learning_rate": 1.4156339370829362e-06, "loss": 2.4853, "step": 27750 }, { "epoch": 5.29, "learning_rate": 1.4155767397521448e-06, "loss": 3.0362, "step": 27751 }, { "epoch": 5.29, "learning_rate": 1.4155195424213538e-06, "loss": 2.1889, "step": 27752 }, { "epoch": 5.29, "learning_rate": 1.4154623450905626e-06, "loss": 1.6245, "step": 27753 }, { "epoch": 5.29, "learning_rate": 1.4154051477597714e-06, "loss": 1.8028, "step": 27754 }, { "epoch": 5.29, "learning_rate": 1.41534795042898e-06, "loss": 1.2901, "step": 27755 }, { "epoch": 5.29, "learning_rate": 1.4152907530981887e-06, "loss": 1.3984, "step": 27756 }, { "epoch": 5.29, "learning_rate": 1.4152335557673975e-06, "loss": 2.4215, "step": 27757 }, { "epoch": 5.29, "learning_rate": 1.4151763584366063e-06, "loss": 1.3504, "step": 27758 }, { "epoch": 5.29, "learning_rate": 1.4151191611058151e-06, "loss": 2.2776, "step": 27759 }, { "epoch": 5.29, "learning_rate": 1.415061963775024e-06, "loss": 3.5857, "step": 27760 }, { "epoch": 5.29, "learning_rate": 1.4150047664442327e-06, "loss": 1.9404, "step": 27761 }, { "epoch": 5.29, "learning_rate": 1.4149475691134413e-06, "loss": 2.0514, "step": 27762 }, { "epoch": 5.29, "learning_rate": 1.41489037178265e-06, "loss": 2.9837, "step": 27763 }, { "epoch": 5.29, "learning_rate": 1.414833174451859e-06, "loss": 0.9307, "step": 27764 }, { "epoch": 5.29, "learning_rate": 1.4147759771210677e-06, "loss": 1.2339, "step": 27765 }, { "epoch": 5.29, "learning_rate": 1.4147187797902765e-06, "loss": 2.592, "step": 27766 }, { "epoch": 5.29, "learning_rate": 1.4146615824594853e-06, "loss": 1.8334, "step": 27767 }, { "epoch": 5.29, "learning_rate": 1.414604385128694e-06, "loss": 1.9538, "step": 27768 }, { "epoch": 5.29, "learning_rate": 1.4145471877979028e-06, "loss": 1.3571, "step": 27769 }, { "epoch": 5.29, "learning_rate": 1.4144899904671116e-06, "loss": 1.2815, "step": 27770 }, { "epoch": 5.29, "learning_rate": 1.4144327931363204e-06, "loss": 1.1224, "step": 27771 }, { "epoch": 5.29, "learning_rate": 1.414375595805529e-06, "loss": 2.2668, "step": 27772 }, { "epoch": 5.29, "learning_rate": 1.4143183984747378e-06, "loss": 2.6237, "step": 27773 }, { "epoch": 5.29, "learning_rate": 1.4142612011439466e-06, "loss": 1.3746, "step": 27774 }, { "epoch": 5.29, "learning_rate": 1.4142040038131554e-06, "loss": 2.232, "step": 27775 }, { "epoch": 5.29, "learning_rate": 1.4141468064823642e-06, "loss": 1.6743, "step": 27776 }, { "epoch": 5.29, "learning_rate": 1.414089609151573e-06, "loss": 1.6266, "step": 27777 }, { "epoch": 5.29, "learning_rate": 1.4140324118207818e-06, "loss": 1.0778, "step": 27778 }, { "epoch": 5.29, "learning_rate": 1.4139752144899904e-06, "loss": 1.939, "step": 27779 }, { "epoch": 5.29, "learning_rate": 1.4139180171591994e-06, "loss": 1.5353, "step": 27780 }, { "epoch": 5.29, "learning_rate": 1.4138608198284082e-06, "loss": 2.0126, "step": 27781 }, { "epoch": 5.29, "learning_rate": 1.4138036224976167e-06, "loss": 2.5207, "step": 27782 }, { "epoch": 5.29, "learning_rate": 1.4137464251668255e-06, "loss": 2.4723, "step": 27783 }, { "epoch": 5.29, "learning_rate": 1.4136892278360343e-06, "loss": 1.6519, "step": 27784 }, { "epoch": 5.29, "learning_rate": 1.4136320305052431e-06, "loss": 1.6555, "step": 27785 }, { "epoch": 5.29, "learning_rate": 1.413574833174452e-06, "loss": 2.0202, "step": 27786 }, { "epoch": 5.29, "learning_rate": 1.4135176358436607e-06, "loss": 1.789, "step": 27787 }, { "epoch": 5.29, "learning_rate": 1.4134604385128695e-06, "loss": 1.2363, "step": 27788 }, { "epoch": 5.29, "learning_rate": 1.413403241182078e-06, "loss": 1.7521, "step": 27789 }, { "epoch": 5.29, "learning_rate": 1.4133460438512869e-06, "loss": 1.9407, "step": 27790 }, { "epoch": 5.29, "learning_rate": 1.4132888465204957e-06, "loss": 1.6579, "step": 27791 }, { "epoch": 5.29, "learning_rate": 1.4132316491897045e-06, "loss": 1.0319, "step": 27792 }, { "epoch": 5.29, "learning_rate": 1.4131744518589133e-06, "loss": 2.7913, "step": 27793 }, { "epoch": 5.29, "learning_rate": 1.413117254528122e-06, "loss": 1.5603, "step": 27794 }, { "epoch": 5.29, "learning_rate": 1.4130600571973309e-06, "loss": 1.0877, "step": 27795 }, { "epoch": 5.29, "learning_rate": 1.4130028598665396e-06, "loss": 2.5366, "step": 27796 }, { "epoch": 5.29, "learning_rate": 1.4129456625357484e-06, "loss": 1.8793, "step": 27797 }, { "epoch": 5.29, "learning_rate": 1.4128884652049572e-06, "loss": 1.1814, "step": 27798 }, { "epoch": 5.3, "learning_rate": 1.4128312678741658e-06, "loss": 1.5521, "step": 27799 }, { "epoch": 5.3, "learning_rate": 1.4127740705433746e-06, "loss": 2.4265, "step": 27800 }, { "epoch": 5.3, "learning_rate": 1.4127168732125834e-06, "loss": 2.0452, "step": 27801 }, { "epoch": 5.3, "learning_rate": 1.4126596758817922e-06, "loss": 1.5966, "step": 27802 }, { "epoch": 5.3, "learning_rate": 1.412602478551001e-06, "loss": 3.4447, "step": 27803 }, { "epoch": 5.3, "learning_rate": 1.4125452812202098e-06, "loss": 0.9271, "step": 27804 }, { "epoch": 5.3, "learning_rate": 1.4124880838894186e-06, "loss": 2.5333, "step": 27805 }, { "epoch": 5.3, "learning_rate": 1.4124308865586272e-06, "loss": 1.6179, "step": 27806 }, { "epoch": 5.3, "learning_rate": 1.412373689227836e-06, "loss": 0.9189, "step": 27807 }, { "epoch": 5.3, "learning_rate": 1.412316491897045e-06, "loss": 2.3702, "step": 27808 }, { "epoch": 5.3, "learning_rate": 1.4122592945662538e-06, "loss": 2.3373, "step": 27809 }, { "epoch": 5.3, "learning_rate": 1.4122020972354623e-06, "loss": 1.8293, "step": 27810 }, { "epoch": 5.3, "learning_rate": 1.4121448999046711e-06, "loss": 1.893, "step": 27811 }, { "epoch": 5.3, "learning_rate": 1.41208770257388e-06, "loss": 1.9969, "step": 27812 }, { "epoch": 5.3, "learning_rate": 1.4120305052430887e-06, "loss": 1.5377, "step": 27813 }, { "epoch": 5.3, "learning_rate": 1.4119733079122975e-06, "loss": 1.3482, "step": 27814 }, { "epoch": 5.3, "learning_rate": 1.4119161105815063e-06, "loss": 1.7518, "step": 27815 }, { "epoch": 5.3, "learning_rate": 1.4118589132507149e-06, "loss": 1.8124, "step": 27816 }, { "epoch": 5.3, "learning_rate": 1.4118017159199237e-06, "loss": 2.1803, "step": 27817 }, { "epoch": 5.3, "learning_rate": 1.4117445185891325e-06, "loss": 1.0553, "step": 27818 }, { "epoch": 5.3, "learning_rate": 1.4116873212583415e-06, "loss": 2.6471, "step": 27819 }, { "epoch": 5.3, "learning_rate": 1.41163012392755e-06, "loss": 1.4383, "step": 27820 }, { "epoch": 5.3, "learning_rate": 1.4115729265967589e-06, "loss": 1.6846, "step": 27821 }, { "epoch": 5.3, "learning_rate": 1.4115157292659677e-06, "loss": 1.7618, "step": 27822 }, { "epoch": 5.3, "learning_rate": 1.4114585319351762e-06, "loss": 1.1418, "step": 27823 }, { "epoch": 5.3, "learning_rate": 1.4114013346043852e-06, "loss": 2.0857, "step": 27824 }, { "epoch": 5.3, "learning_rate": 1.411344137273594e-06, "loss": 0.8101, "step": 27825 }, { "epoch": 5.3, "learning_rate": 1.4112869399428028e-06, "loss": 1.5324, "step": 27826 }, { "epoch": 5.3, "learning_rate": 1.4112297426120114e-06, "loss": 2.4748, "step": 27827 }, { "epoch": 5.3, "learning_rate": 1.4111725452812202e-06, "loss": 1.6992, "step": 27828 }, { "epoch": 5.3, "learning_rate": 1.411115347950429e-06, "loss": 2.3369, "step": 27829 }, { "epoch": 5.3, "learning_rate": 1.4110581506196378e-06, "loss": 1.8946, "step": 27830 }, { "epoch": 5.3, "learning_rate": 1.4110009532888466e-06, "loss": 1.6189, "step": 27831 }, { "epoch": 5.3, "learning_rate": 1.4109437559580554e-06, "loss": 1.2653, "step": 27832 }, { "epoch": 5.3, "learning_rate": 1.4108865586272642e-06, "loss": 1.7756, "step": 27833 }, { "epoch": 5.3, "learning_rate": 1.4108293612964728e-06, "loss": 2.5849, "step": 27834 }, { "epoch": 5.3, "learning_rate": 1.4107721639656815e-06, "loss": 1.596, "step": 27835 }, { "epoch": 5.3, "learning_rate": 1.4107149666348906e-06, "loss": 2.9062, "step": 27836 }, { "epoch": 5.3, "learning_rate": 1.4106577693040991e-06, "loss": 2.6514, "step": 27837 }, { "epoch": 5.3, "learning_rate": 1.410600571973308e-06, "loss": 1.8861, "step": 27838 }, { "epoch": 5.3, "learning_rate": 1.4105433746425167e-06, "loss": 1.785, "step": 27839 }, { "epoch": 5.3, "learning_rate": 1.4104861773117255e-06, "loss": 0.9321, "step": 27840 }, { "epoch": 5.3, "learning_rate": 1.4104289799809343e-06, "loss": 1.7562, "step": 27841 }, { "epoch": 5.3, "learning_rate": 1.410371782650143e-06, "loss": 2.7981, "step": 27842 }, { "epoch": 5.3, "learning_rate": 1.410314585319352e-06, "loss": 2.0098, "step": 27843 }, { "epoch": 5.3, "learning_rate": 1.4102573879885605e-06, "loss": 2.7104, "step": 27844 }, { "epoch": 5.3, "learning_rate": 1.4102001906577693e-06, "loss": 2.3862, "step": 27845 }, { "epoch": 5.3, "learning_rate": 1.410142993326978e-06, "loss": 1.8163, "step": 27846 }, { "epoch": 5.3, "learning_rate": 1.4100857959961869e-06, "loss": 1.2724, "step": 27847 }, { "epoch": 5.3, "learning_rate": 1.4100285986653957e-06, "loss": 2.3261, "step": 27848 }, { "epoch": 5.3, "learning_rate": 1.4099714013346044e-06, "loss": 1.5303, "step": 27849 }, { "epoch": 5.3, "learning_rate": 1.4099142040038132e-06, "loss": 1.1068, "step": 27850 }, { "epoch": 5.3, "learning_rate": 1.4098570066730218e-06, "loss": 1.8518, "step": 27851 }, { "epoch": 5.31, "learning_rate": 1.4097998093422308e-06, "loss": 3.3855, "step": 27852 }, { "epoch": 5.31, "learning_rate": 1.4097426120114396e-06, "loss": 2.0024, "step": 27853 }, { "epoch": 5.31, "learning_rate": 1.4096854146806482e-06, "loss": 2.619, "step": 27854 }, { "epoch": 5.31, "learning_rate": 1.409628217349857e-06, "loss": 3.0039, "step": 27855 }, { "epoch": 5.31, "learning_rate": 1.4095710200190658e-06, "loss": 1.8232, "step": 27856 }, { "epoch": 5.31, "learning_rate": 1.4095138226882746e-06, "loss": 2.3615, "step": 27857 }, { "epoch": 5.31, "learning_rate": 1.4094566253574834e-06, "loss": 2.0297, "step": 27858 }, { "epoch": 5.31, "learning_rate": 1.4093994280266922e-06, "loss": 2.1188, "step": 27859 }, { "epoch": 5.31, "learning_rate": 1.409342230695901e-06, "loss": 2.0103, "step": 27860 }, { "epoch": 5.31, "learning_rate": 1.4092850333651096e-06, "loss": 1.454, "step": 27861 }, { "epoch": 5.31, "learning_rate": 1.4092278360343183e-06, "loss": 1.0611, "step": 27862 }, { "epoch": 5.31, "learning_rate": 1.4091706387035274e-06, "loss": 2.9685, "step": 27863 }, { "epoch": 5.31, "learning_rate": 1.409113441372736e-06, "loss": 1.6077, "step": 27864 }, { "epoch": 5.31, "learning_rate": 1.4090562440419447e-06, "loss": 2.3367, "step": 27865 }, { "epoch": 5.31, "learning_rate": 1.4089990467111535e-06, "loss": 3.4791, "step": 27866 }, { "epoch": 5.31, "learning_rate": 1.4089418493803623e-06, "loss": 1.7872, "step": 27867 }, { "epoch": 5.31, "learning_rate": 1.4088846520495711e-06, "loss": 2.2801, "step": 27868 }, { "epoch": 5.31, "learning_rate": 1.40882745471878e-06, "loss": 1.3479, "step": 27869 }, { "epoch": 5.31, "learning_rate": 1.4087702573879887e-06, "loss": 1.5816, "step": 27870 }, { "epoch": 5.31, "learning_rate": 1.4087130600571973e-06, "loss": 1.6697, "step": 27871 }, { "epoch": 5.31, "learning_rate": 1.408655862726406e-06, "loss": 2.2393, "step": 27872 }, { "epoch": 5.31, "learning_rate": 1.4085986653956149e-06, "loss": 0.8052, "step": 27873 }, { "epoch": 5.31, "learning_rate": 1.4085414680648237e-06, "loss": 1.6877, "step": 27874 }, { "epoch": 5.31, "learning_rate": 1.4084842707340325e-06, "loss": 1.8924, "step": 27875 }, { "epoch": 5.31, "learning_rate": 1.4084270734032412e-06, "loss": 1.5153, "step": 27876 }, { "epoch": 5.31, "learning_rate": 1.40836987607245e-06, "loss": 1.9037, "step": 27877 }, { "epoch": 5.31, "learning_rate": 1.4083126787416586e-06, "loss": 2.3556, "step": 27878 }, { "epoch": 5.31, "learning_rate": 1.4082554814108674e-06, "loss": 2.3371, "step": 27879 }, { "epoch": 5.31, "learning_rate": 1.4081982840800764e-06, "loss": 1.7585, "step": 27880 }, { "epoch": 5.31, "learning_rate": 1.408141086749285e-06, "loss": 1.885, "step": 27881 }, { "epoch": 5.31, "learning_rate": 1.4080838894184938e-06, "loss": 1.1039, "step": 27882 }, { "epoch": 5.31, "learning_rate": 1.4080266920877026e-06, "loss": 1.6135, "step": 27883 }, { "epoch": 5.31, "learning_rate": 1.4079694947569114e-06, "loss": 2.0488, "step": 27884 }, { "epoch": 5.31, "learning_rate": 1.4079122974261202e-06, "loss": 2.9643, "step": 27885 }, { "epoch": 5.31, "learning_rate": 1.407855100095329e-06, "loss": 2.9327, "step": 27886 }, { "epoch": 5.31, "learning_rate": 1.4077979027645378e-06, "loss": 2.4136, "step": 27887 }, { "epoch": 5.31, "learning_rate": 1.4077407054337463e-06, "loss": 1.012, "step": 27888 }, { "epoch": 5.31, "learning_rate": 1.4076835081029551e-06, "loss": 2.4921, "step": 27889 }, { "epoch": 5.31, "learning_rate": 1.407626310772164e-06, "loss": 1.4918, "step": 27890 }, { "epoch": 5.31, "learning_rate": 1.407569113441373e-06, "loss": 3.8885, "step": 27891 }, { "epoch": 5.31, "learning_rate": 1.4075119161105815e-06, "loss": 1.2184, "step": 27892 }, { "epoch": 5.31, "learning_rate": 1.4074547187797903e-06, "loss": 1.2232, "step": 27893 }, { "epoch": 5.31, "learning_rate": 1.4073975214489991e-06, "loss": 1.4867, "step": 27894 }, { "epoch": 5.31, "learning_rate": 1.4073403241182077e-06, "loss": 2.1055, "step": 27895 }, { "epoch": 5.31, "learning_rate": 1.4072831267874167e-06, "loss": 1.3668, "step": 27896 }, { "epoch": 5.31, "learning_rate": 1.4072259294566255e-06, "loss": 2.2705, "step": 27897 }, { "epoch": 5.31, "learning_rate": 1.4071687321258343e-06, "loss": 0.5072, "step": 27898 }, { "epoch": 5.31, "learning_rate": 1.4071115347950429e-06, "loss": 1.9129, "step": 27899 }, { "epoch": 5.31, "learning_rate": 1.4070543374642517e-06, "loss": 1.511, "step": 27900 }, { "epoch": 5.31, "learning_rate": 1.4069971401334605e-06, "loss": 2.4397, "step": 27901 }, { "epoch": 5.31, "learning_rate": 1.4069399428026693e-06, "loss": 1.9657, "step": 27902 }, { "epoch": 5.31, "learning_rate": 1.406882745471878e-06, "loss": 0.8783, "step": 27903 }, { "epoch": 5.32, "learning_rate": 1.4068255481410868e-06, "loss": 0.9548, "step": 27904 }, { "epoch": 5.32, "learning_rate": 1.4067683508102954e-06, "loss": 1.2342, "step": 27905 }, { "epoch": 5.32, "learning_rate": 1.4067111534795042e-06, "loss": 1.795, "step": 27906 }, { "epoch": 5.32, "learning_rate": 1.4066539561487132e-06, "loss": 2.8126, "step": 27907 }, { "epoch": 5.32, "learning_rate": 1.406596758817922e-06, "loss": 1.4033, "step": 27908 }, { "epoch": 5.32, "learning_rate": 1.4065395614871306e-06, "loss": 1.4113, "step": 27909 }, { "epoch": 5.32, "learning_rate": 1.4064823641563394e-06, "loss": 1.0402, "step": 27910 }, { "epoch": 5.32, "learning_rate": 1.4064251668255482e-06, "loss": 1.3586, "step": 27911 }, { "epoch": 5.32, "learning_rate": 1.406367969494757e-06, "loss": 1.9066, "step": 27912 }, { "epoch": 5.32, "learning_rate": 1.4063107721639658e-06, "loss": 1.3882, "step": 27913 }, { "epoch": 5.32, "learning_rate": 1.4062535748331746e-06, "loss": 2.4693, "step": 27914 }, { "epoch": 5.32, "learning_rate": 1.4061963775023834e-06, "loss": 1.3498, "step": 27915 }, { "epoch": 5.32, "learning_rate": 1.406139180171592e-06, "loss": 1.2015, "step": 27916 }, { "epoch": 5.32, "learning_rate": 1.4060819828408007e-06, "loss": 1.8545, "step": 27917 }, { "epoch": 5.32, "learning_rate": 1.4060247855100095e-06, "loss": 1.5018, "step": 27918 }, { "epoch": 5.32, "learning_rate": 1.4059675881792183e-06, "loss": 1.7143, "step": 27919 }, { "epoch": 5.32, "learning_rate": 1.4059103908484271e-06, "loss": 1.5775, "step": 27920 }, { "epoch": 5.32, "learning_rate": 1.405853193517636e-06, "loss": 2.3389, "step": 27921 }, { "epoch": 5.32, "learning_rate": 1.4057959961868445e-06, "loss": 2.5885, "step": 27922 }, { "epoch": 5.32, "learning_rate": 1.4057387988560533e-06, "loss": 2.7699, "step": 27923 }, { "epoch": 5.32, "learning_rate": 1.4056816015252623e-06, "loss": 0.8755, "step": 27924 }, { "epoch": 5.32, "learning_rate": 1.405624404194471e-06, "loss": 1.4129, "step": 27925 }, { "epoch": 5.32, "learning_rate": 1.4055672068636797e-06, "loss": 1.9533, "step": 27926 }, { "epoch": 5.32, "learning_rate": 1.4055100095328885e-06, "loss": 1.9727, "step": 27927 }, { "epoch": 5.32, "learning_rate": 1.4054528122020973e-06, "loss": 3.3052, "step": 27928 }, { "epoch": 5.32, "learning_rate": 1.405395614871306e-06, "loss": 2.2065, "step": 27929 }, { "epoch": 5.32, "learning_rate": 1.4053384175405148e-06, "loss": 0.5878, "step": 27930 }, { "epoch": 5.32, "learning_rate": 1.4052812202097236e-06, "loss": 1.3728, "step": 27931 }, { "epoch": 5.32, "learning_rate": 1.4052240228789324e-06, "loss": 2.2008, "step": 27932 }, { "epoch": 5.32, "learning_rate": 1.405166825548141e-06, "loss": 1.5235, "step": 27933 }, { "epoch": 5.32, "learning_rate": 1.4051096282173498e-06, "loss": 1.8318, "step": 27934 }, { "epoch": 5.32, "learning_rate": 1.4050524308865588e-06, "loss": 3.3721, "step": 27935 }, { "epoch": 5.32, "learning_rate": 1.4049952335557674e-06, "loss": 1.4741, "step": 27936 }, { "epoch": 5.32, "learning_rate": 1.4049380362249762e-06, "loss": 2.116, "step": 27937 }, { "epoch": 5.32, "learning_rate": 1.404880838894185e-06, "loss": 2.0889, "step": 27938 }, { "epoch": 5.32, "learning_rate": 1.4048236415633938e-06, "loss": 1.5595, "step": 27939 }, { "epoch": 5.32, "learning_rate": 1.4047664442326026e-06, "loss": 1.1052, "step": 27940 }, { "epoch": 5.32, "learning_rate": 1.4047092469018114e-06, "loss": 2.0157, "step": 27941 }, { "epoch": 5.32, "learning_rate": 1.4046520495710202e-06, "loss": 1.7003, "step": 27942 }, { "epoch": 5.32, "learning_rate": 1.4045948522402287e-06, "loss": 1.9256, "step": 27943 }, { "epoch": 5.32, "learning_rate": 1.4045376549094375e-06, "loss": 1.7811, "step": 27944 }, { "epoch": 5.32, "learning_rate": 1.4044804575786463e-06, "loss": 2.3661, "step": 27945 }, { "epoch": 5.32, "learning_rate": 1.4044232602478551e-06, "loss": 2.7004, "step": 27946 }, { "epoch": 5.32, "learning_rate": 1.404366062917064e-06, "loss": 1.3227, "step": 27947 }, { "epoch": 5.32, "learning_rate": 1.4043088655862727e-06, "loss": 1.1901, "step": 27948 }, { "epoch": 5.32, "learning_rate": 1.4042516682554815e-06, "loss": 2.3137, "step": 27949 }, { "epoch": 5.32, "learning_rate": 1.40419447092469e-06, "loss": 1.2199, "step": 27950 }, { "epoch": 5.32, "learning_rate": 1.404137273593899e-06, "loss": 1.6596, "step": 27951 }, { "epoch": 5.32, "learning_rate": 1.4040800762631079e-06, "loss": 2.0749, "step": 27952 }, { "epoch": 5.32, "learning_rate": 1.4040228789323165e-06, "loss": 1.8338, "step": 27953 }, { "epoch": 5.32, "learning_rate": 1.4039656816015253e-06, "loss": 2.7053, "step": 27954 }, { "epoch": 5.32, "learning_rate": 1.403908484270734e-06, "loss": 2.4996, "step": 27955 }, { "epoch": 5.32, "learning_rate": 1.4038512869399428e-06, "loss": 2.5641, "step": 27956 }, { "epoch": 5.33, "learning_rate": 1.4037940896091516e-06, "loss": 2.4567, "step": 27957 }, { "epoch": 5.33, "learning_rate": 1.4037368922783604e-06, "loss": 2.5801, "step": 27958 }, { "epoch": 5.33, "learning_rate": 1.4036796949475692e-06, "loss": 2.2395, "step": 27959 }, { "epoch": 5.33, "learning_rate": 1.4036224976167778e-06, "loss": 1.9825, "step": 27960 }, { "epoch": 5.33, "learning_rate": 1.4035653002859866e-06, "loss": 1.5034, "step": 27961 }, { "epoch": 5.33, "learning_rate": 1.4035081029551954e-06, "loss": 2.2706, "step": 27962 }, { "epoch": 5.33, "learning_rate": 1.4034509056244044e-06, "loss": 2.208, "step": 27963 }, { "epoch": 5.33, "learning_rate": 1.403393708293613e-06, "loss": 1.7481, "step": 27964 }, { "epoch": 5.33, "learning_rate": 1.4033365109628218e-06, "loss": 1.636, "step": 27965 }, { "epoch": 5.33, "learning_rate": 1.4032793136320306e-06, "loss": 1.6576, "step": 27966 }, { "epoch": 5.33, "learning_rate": 1.4032221163012392e-06, "loss": 1.8031, "step": 27967 }, { "epoch": 5.33, "learning_rate": 1.4031649189704482e-06, "loss": 2.415, "step": 27968 }, { "epoch": 5.33, "learning_rate": 1.403107721639657e-06, "loss": 1.6723, "step": 27969 }, { "epoch": 5.33, "learning_rate": 1.4030505243088655e-06, "loss": 0.9158, "step": 27970 }, { "epoch": 5.33, "learning_rate": 1.4029933269780743e-06, "loss": 0.906, "step": 27971 }, { "epoch": 5.33, "learning_rate": 1.4029361296472831e-06, "loss": 1.884, "step": 27972 }, { "epoch": 5.33, "learning_rate": 1.402878932316492e-06, "loss": 1.1711, "step": 27973 }, { "epoch": 5.33, "learning_rate": 1.4028217349857007e-06, "loss": 2.1381, "step": 27974 }, { "epoch": 5.33, "learning_rate": 1.4027645376549095e-06, "loss": 1.5662, "step": 27975 }, { "epoch": 5.33, "learning_rate": 1.4027073403241183e-06, "loss": 2.0951, "step": 27976 }, { "epoch": 5.33, "learning_rate": 1.4026501429933269e-06, "loss": 1.6832, "step": 27977 }, { "epoch": 5.33, "learning_rate": 1.4025929456625357e-06, "loss": 2.0285, "step": 27978 }, { "epoch": 5.33, "learning_rate": 1.4025357483317447e-06, "loss": 2.3593, "step": 27979 }, { "epoch": 5.33, "learning_rate": 1.4024785510009535e-06, "loss": 1.1619, "step": 27980 }, { "epoch": 5.33, "learning_rate": 1.402421353670162e-06, "loss": 2.6, "step": 27981 }, { "epoch": 5.33, "learning_rate": 1.4023641563393709e-06, "loss": 2.7736, "step": 27982 }, { "epoch": 5.33, "learning_rate": 1.4023069590085796e-06, "loss": 3.1608, "step": 27983 }, { "epoch": 5.33, "learning_rate": 1.4022497616777884e-06, "loss": 1.6761, "step": 27984 }, { "epoch": 5.33, "learning_rate": 1.4021925643469972e-06, "loss": 1.5821, "step": 27985 }, { "epoch": 5.33, "learning_rate": 1.402135367016206e-06, "loss": 1.649, "step": 27986 }, { "epoch": 5.33, "learning_rate": 1.4020781696854146e-06, "loss": 3.2601, "step": 27987 }, { "epoch": 5.33, "learning_rate": 1.4020209723546234e-06, "loss": 1.9231, "step": 27988 }, { "epoch": 5.33, "learning_rate": 1.4019637750238322e-06, "loss": 0.4285, "step": 27989 }, { "epoch": 5.33, "learning_rate": 1.401906577693041e-06, "loss": 2.4622, "step": 27990 }, { "epoch": 5.33, "learning_rate": 1.4018493803622498e-06, "loss": 0.7712, "step": 27991 }, { "epoch": 5.33, "learning_rate": 1.4017921830314586e-06, "loss": 1.3015, "step": 27992 }, { "epoch": 5.33, "learning_rate": 1.4017349857006674e-06, "loss": 1.3623, "step": 27993 }, { "epoch": 5.33, "learning_rate": 1.401677788369876e-06, "loss": 2.1358, "step": 27994 }, { "epoch": 5.33, "learning_rate": 1.401620591039085e-06, "loss": 3.033, "step": 27995 }, { "epoch": 5.33, "learning_rate": 1.4015633937082938e-06, "loss": 1.0904, "step": 27996 }, { "epoch": 5.33, "learning_rate": 1.4015061963775025e-06, "loss": 1.9724, "step": 27997 }, { "epoch": 5.33, "learning_rate": 1.4014489990467111e-06, "loss": 2.0204, "step": 27998 }, { "epoch": 5.33, "learning_rate": 1.40139180171592e-06, "loss": 2.3899, "step": 27999 }, { "epoch": 5.33, "learning_rate": 1.4013346043851287e-06, "loss": 3.4952, "step": 28000 }, { "epoch": 5.33, "learning_rate": 1.4012774070543375e-06, "loss": 1.4783, "step": 28001 }, { "epoch": 5.33, "learning_rate": 1.4012202097235463e-06, "loss": 3.7768, "step": 28002 }, { "epoch": 5.33, "learning_rate": 1.401163012392755e-06, "loss": 2.3799, "step": 28003 }, { "epoch": 5.33, "learning_rate": 1.4011058150619639e-06, "loss": 2.77, "step": 28004 }, { "epoch": 5.33, "learning_rate": 1.4010486177311725e-06, "loss": 1.5803, "step": 28005 }, { "epoch": 5.33, "learning_rate": 1.4009914204003813e-06, "loss": 2.9006, "step": 28006 }, { "epoch": 5.33, "learning_rate": 1.4009342230695903e-06, "loss": 2.4227, "step": 28007 }, { "epoch": 5.33, "learning_rate": 1.4008770257387989e-06, "loss": 2.0586, "step": 28008 }, { "epoch": 5.34, "learning_rate": 1.4008198284080076e-06, "loss": 1.4795, "step": 28009 }, { "epoch": 5.34, "learning_rate": 1.4007626310772164e-06, "loss": 1.6441, "step": 28010 }, { "epoch": 5.34, "learning_rate": 1.400705433746425e-06, "loss": 1.0263, "step": 28011 }, { "epoch": 5.34, "learning_rate": 1.400648236415634e-06, "loss": 3.0299, "step": 28012 }, { "epoch": 5.34, "learning_rate": 1.4005910390848428e-06, "loss": 1.3782, "step": 28013 }, { "epoch": 5.34, "learning_rate": 1.4005338417540516e-06, "loss": 1.3343, "step": 28014 }, { "epoch": 5.34, "learning_rate": 1.4004766444232602e-06, "loss": 2.1525, "step": 28015 }, { "epoch": 5.34, "learning_rate": 1.400419447092469e-06, "loss": 2.436, "step": 28016 }, { "epoch": 5.34, "learning_rate": 1.4003622497616778e-06, "loss": 1.2727, "step": 28017 }, { "epoch": 5.34, "learning_rate": 1.4003050524308866e-06, "loss": 2.0416, "step": 28018 }, { "epoch": 5.34, "learning_rate": 1.4002478551000954e-06, "loss": 2.2618, "step": 28019 }, { "epoch": 5.34, "learning_rate": 1.4001906577693042e-06, "loss": 0.8926, "step": 28020 }, { "epoch": 5.34, "learning_rate": 1.400133460438513e-06, "loss": 2.9974, "step": 28021 }, { "epoch": 5.34, "learning_rate": 1.4000762631077215e-06, "loss": 1.8536, "step": 28022 }, { "epoch": 5.34, "learning_rate": 1.4000190657769306e-06, "loss": 1.6646, "step": 28023 }, { "epoch": 5.34, "learning_rate": 1.3999618684461393e-06, "loss": 1.6776, "step": 28024 }, { "epoch": 5.34, "learning_rate": 1.399904671115348e-06, "loss": 1.0928, "step": 28025 }, { "epoch": 5.34, "learning_rate": 1.3998474737845567e-06, "loss": 2.0488, "step": 28026 }, { "epoch": 5.34, "learning_rate": 1.3997902764537655e-06, "loss": 2.768, "step": 28027 }, { "epoch": 5.34, "learning_rate": 1.3997330791229743e-06, "loss": 1.7532, "step": 28028 }, { "epoch": 5.34, "learning_rate": 1.399675881792183e-06, "loss": 1.2968, "step": 28029 }, { "epoch": 5.34, "learning_rate": 1.399618684461392e-06, "loss": 2.6986, "step": 28030 }, { "epoch": 5.34, "learning_rate": 1.3995614871306007e-06, "loss": 1.3694, "step": 28031 }, { "epoch": 5.34, "learning_rate": 1.3995042897998093e-06, "loss": 1.1199, "step": 28032 }, { "epoch": 5.34, "learning_rate": 1.399447092469018e-06, "loss": 2.0461, "step": 28033 }, { "epoch": 5.34, "learning_rate": 1.3993898951382269e-06, "loss": 1.8656, "step": 28034 }, { "epoch": 5.34, "learning_rate": 1.3993326978074357e-06, "loss": 2.3756, "step": 28035 }, { "epoch": 5.34, "learning_rate": 1.3992755004766444e-06, "loss": 2.1164, "step": 28036 }, { "epoch": 5.34, "learning_rate": 1.3992183031458532e-06, "loss": 2.496, "step": 28037 }, { "epoch": 5.34, "learning_rate": 1.399161105815062e-06, "loss": 1.0402, "step": 28038 }, { "epoch": 5.34, "learning_rate": 1.3991039084842706e-06, "loss": 1.5266, "step": 28039 }, { "epoch": 5.34, "learning_rate": 1.3990467111534796e-06, "loss": 2.2939, "step": 28040 }, { "epoch": 5.34, "learning_rate": 1.3989895138226884e-06, "loss": 2.1765, "step": 28041 }, { "epoch": 5.34, "learning_rate": 1.398932316491897e-06, "loss": 1.8014, "step": 28042 }, { "epoch": 5.34, "learning_rate": 1.3988751191611058e-06, "loss": 2.195, "step": 28043 }, { "epoch": 5.34, "learning_rate": 1.3988179218303146e-06, "loss": 4.4067, "step": 28044 }, { "epoch": 5.34, "learning_rate": 1.3987607244995234e-06, "loss": 3.105, "step": 28045 }, { "epoch": 5.34, "learning_rate": 1.3987035271687322e-06, "loss": 2.3231, "step": 28046 }, { "epoch": 5.34, "learning_rate": 1.398646329837941e-06, "loss": 2.5787, "step": 28047 }, { "epoch": 5.34, "learning_rate": 1.3985891325071498e-06, "loss": 2.5655, "step": 28048 }, { "epoch": 5.34, "learning_rate": 1.3985319351763583e-06, "loss": 2.108, "step": 28049 }, { "epoch": 5.34, "learning_rate": 1.3984747378455671e-06, "loss": 2.2099, "step": 28050 }, { "epoch": 5.34, "learning_rate": 1.3984175405147761e-06, "loss": 2.4794, "step": 28051 }, { "epoch": 5.34, "learning_rate": 1.3983603431839847e-06, "loss": 0.9432, "step": 28052 }, { "epoch": 5.34, "learning_rate": 1.3983031458531935e-06, "loss": 2.6825, "step": 28053 }, { "epoch": 5.34, "learning_rate": 1.3982459485224023e-06, "loss": 3.3468, "step": 28054 }, { "epoch": 5.34, "learning_rate": 1.398188751191611e-06, "loss": 1.8484, "step": 28055 }, { "epoch": 5.34, "learning_rate": 1.39813155386082e-06, "loss": 1.549, "step": 28056 }, { "epoch": 5.34, "learning_rate": 1.3980743565300287e-06, "loss": 1.3699, "step": 28057 }, { "epoch": 5.34, "learning_rate": 1.3980171591992375e-06, "loss": 1.7429, "step": 28058 }, { "epoch": 5.34, "learning_rate": 1.397959961868446e-06, "loss": 1.6904, "step": 28059 }, { "epoch": 5.34, "learning_rate": 1.3979027645376549e-06, "loss": 1.8675, "step": 28060 }, { "epoch": 5.34, "learning_rate": 1.3978455672068637e-06, "loss": 1.8439, "step": 28061 }, { "epoch": 5.35, "learning_rate": 1.3977883698760727e-06, "loss": 2.6134, "step": 28062 }, { "epoch": 5.35, "learning_rate": 1.3977311725452812e-06, "loss": 1.9347, "step": 28063 }, { "epoch": 5.35, "learning_rate": 1.39767397521449e-06, "loss": 2.5855, "step": 28064 }, { "epoch": 5.35, "learning_rate": 1.3976167778836988e-06, "loss": 1.5057, "step": 28065 }, { "epoch": 5.35, "learning_rate": 1.3975595805529074e-06, "loss": 1.2855, "step": 28066 }, { "epoch": 5.35, "learning_rate": 1.3975023832221164e-06, "loss": 0.6874, "step": 28067 }, { "epoch": 5.35, "learning_rate": 1.3974451858913252e-06, "loss": 0.9991, "step": 28068 }, { "epoch": 5.35, "learning_rate": 1.397387988560534e-06, "loss": 2.2791, "step": 28069 }, { "epoch": 5.35, "learning_rate": 1.3973307912297426e-06, "loss": 2.1679, "step": 28070 }, { "epoch": 5.35, "learning_rate": 1.3972735938989514e-06, "loss": 1.5751, "step": 28071 }, { "epoch": 5.35, "learning_rate": 1.3972163965681602e-06, "loss": 1.6601, "step": 28072 }, { "epoch": 5.35, "learning_rate": 1.397159199237369e-06, "loss": 2.135, "step": 28073 }, { "epoch": 5.35, "learning_rate": 1.3971020019065778e-06, "loss": 1.6971, "step": 28074 }, { "epoch": 5.35, "learning_rate": 1.3970448045757866e-06, "loss": 2.0533, "step": 28075 }, { "epoch": 5.35, "learning_rate": 1.3969876072449951e-06, "loss": 2.1287, "step": 28076 }, { "epoch": 5.35, "learning_rate": 1.396930409914204e-06, "loss": 2.6838, "step": 28077 }, { "epoch": 5.35, "learning_rate": 1.3968732125834127e-06, "loss": 1.8972, "step": 28078 }, { "epoch": 5.35, "learning_rate": 1.3968160152526217e-06, "loss": 1.524, "step": 28079 }, { "epoch": 5.35, "learning_rate": 1.3967588179218303e-06, "loss": 0.9478, "step": 28080 }, { "epoch": 5.35, "learning_rate": 1.3967016205910391e-06, "loss": 1.5915, "step": 28081 }, { "epoch": 5.35, "learning_rate": 1.396644423260248e-06, "loss": 1.6346, "step": 28082 }, { "epoch": 5.35, "learning_rate": 1.3965872259294565e-06, "loss": 1.666, "step": 28083 }, { "epoch": 5.35, "learning_rate": 1.3965300285986655e-06, "loss": 1.2407, "step": 28084 }, { "epoch": 5.35, "learning_rate": 1.3964728312678743e-06, "loss": 1.8817, "step": 28085 }, { "epoch": 5.35, "learning_rate": 1.396415633937083e-06, "loss": 1.4382, "step": 28086 }, { "epoch": 5.35, "learning_rate": 1.3963584366062917e-06, "loss": 2.575, "step": 28087 }, { "epoch": 5.35, "learning_rate": 1.3963012392755005e-06, "loss": 1.9817, "step": 28088 }, { "epoch": 5.35, "learning_rate": 1.3962440419447092e-06, "loss": 1.9309, "step": 28089 }, { "epoch": 5.35, "learning_rate": 1.396186844613918e-06, "loss": 2.2961, "step": 28090 }, { "epoch": 5.35, "learning_rate": 1.3961296472831268e-06, "loss": 1.8243, "step": 28091 }, { "epoch": 5.35, "learning_rate": 1.3960724499523356e-06, "loss": 2.6891, "step": 28092 }, { "epoch": 5.35, "learning_rate": 1.3960152526215442e-06, "loss": 1.9069, "step": 28093 }, { "epoch": 5.35, "learning_rate": 1.395958055290753e-06, "loss": 1.4951, "step": 28094 }, { "epoch": 5.35, "learning_rate": 1.395900857959962e-06, "loss": 1.6656, "step": 28095 }, { "epoch": 5.35, "learning_rate": 1.3958436606291708e-06, "loss": 1.9674, "step": 28096 }, { "epoch": 5.35, "learning_rate": 1.3957864632983794e-06, "loss": 1.7891, "step": 28097 }, { "epoch": 5.35, "learning_rate": 1.3957292659675882e-06, "loss": 2.3107, "step": 28098 }, { "epoch": 5.35, "learning_rate": 1.395672068636797e-06, "loss": 1.5931, "step": 28099 }, { "epoch": 5.35, "learning_rate": 1.3956148713060058e-06, "loss": 1.386, "step": 28100 }, { "epoch": 5.35, "learning_rate": 1.3955576739752146e-06, "loss": 2.592, "step": 28101 }, { "epoch": 5.35, "learning_rate": 1.3955004766444234e-06, "loss": 3.0753, "step": 28102 }, { "epoch": 5.35, "learning_rate": 1.3954432793136322e-06, "loss": 1.0651, "step": 28103 }, { "epoch": 5.35, "learning_rate": 1.3953860819828407e-06, "loss": 1.8239, "step": 28104 }, { "epoch": 5.35, "learning_rate": 1.3953288846520495e-06, "loss": 2.5959, "step": 28105 }, { "epoch": 5.35, "learning_rate": 1.3952716873212585e-06, "loss": 2.3015, "step": 28106 }, { "epoch": 5.35, "learning_rate": 1.3952144899904671e-06, "loss": 2.3324, "step": 28107 }, { "epoch": 5.35, "learning_rate": 1.395157292659676e-06, "loss": 2.1264, "step": 28108 }, { "epoch": 5.35, "learning_rate": 1.3951000953288847e-06, "loss": 1.7809, "step": 28109 }, { "epoch": 5.35, "learning_rate": 1.3950428979980935e-06, "loss": 1.2691, "step": 28110 }, { "epoch": 5.35, "learning_rate": 1.3949857006673023e-06, "loss": 1.5203, "step": 28111 }, { "epoch": 5.35, "learning_rate": 1.394928503336511e-06, "loss": 1.6777, "step": 28112 }, { "epoch": 5.35, "learning_rate": 1.3948713060057199e-06, "loss": 1.6744, "step": 28113 }, { "epoch": 5.36, "learning_rate": 1.3948141086749285e-06, "loss": 2.6483, "step": 28114 }, { "epoch": 5.36, "learning_rate": 1.3947569113441373e-06, "loss": 2.8374, "step": 28115 }, { "epoch": 5.36, "learning_rate": 1.394699714013346e-06, "loss": 2.7744, "step": 28116 }, { "epoch": 5.36, "learning_rate": 1.3946425166825548e-06, "loss": 0.9887, "step": 28117 }, { "epoch": 5.36, "learning_rate": 1.3945853193517636e-06, "loss": 1.0027, "step": 28118 }, { "epoch": 5.36, "learning_rate": 1.3945281220209724e-06, "loss": 1.2089, "step": 28119 }, { "epoch": 5.36, "learning_rate": 1.3944709246901812e-06, "loss": 1.4296, "step": 28120 }, { "epoch": 5.36, "learning_rate": 1.3944137273593898e-06, "loss": 2.0202, "step": 28121 }, { "epoch": 5.36, "learning_rate": 1.3943565300285986e-06, "loss": 2.0165, "step": 28122 }, { "epoch": 5.36, "learning_rate": 1.3942993326978076e-06, "loss": 2.0192, "step": 28123 }, { "epoch": 5.36, "learning_rate": 1.3942421353670162e-06, "loss": 2.7107, "step": 28124 }, { "epoch": 5.36, "learning_rate": 1.394184938036225e-06, "loss": 1.9682, "step": 28125 }, { "epoch": 5.36, "learning_rate": 1.3941277407054338e-06, "loss": 3.1514, "step": 28126 }, { "epoch": 5.36, "learning_rate": 1.3940705433746426e-06, "loss": 1.7747, "step": 28127 }, { "epoch": 5.36, "learning_rate": 1.3940133460438514e-06, "loss": 2.7355, "step": 28128 }, { "epoch": 5.36, "learning_rate": 1.3939561487130602e-06, "loss": 2.1937, "step": 28129 }, { "epoch": 5.36, "learning_rate": 1.393898951382269e-06, "loss": 1.6431, "step": 28130 }, { "epoch": 5.36, "learning_rate": 1.3938417540514775e-06, "loss": 2.7357, "step": 28131 }, { "epoch": 5.36, "learning_rate": 1.3937845567206863e-06, "loss": 2.2364, "step": 28132 }, { "epoch": 5.36, "learning_rate": 1.3937273593898951e-06, "loss": 1.7142, "step": 28133 }, { "epoch": 5.36, "learning_rate": 1.3936701620591041e-06, "loss": 3.1582, "step": 28134 }, { "epoch": 5.36, "learning_rate": 1.3936129647283127e-06, "loss": 1.3425, "step": 28135 }, { "epoch": 5.36, "learning_rate": 1.3935557673975215e-06, "loss": 2.4476, "step": 28136 }, { "epoch": 5.36, "learning_rate": 1.3934985700667303e-06, "loss": 2.5188, "step": 28137 }, { "epoch": 5.36, "learning_rate": 1.3934413727359389e-06, "loss": 1.7839, "step": 28138 }, { "epoch": 5.36, "learning_rate": 1.3933841754051479e-06, "loss": 2.6182, "step": 28139 }, { "epoch": 5.36, "learning_rate": 1.3933269780743567e-06, "loss": 2.6517, "step": 28140 }, { "epoch": 5.36, "learning_rate": 1.3932697807435653e-06, "loss": 2.134, "step": 28141 }, { "epoch": 5.36, "learning_rate": 1.393212583412774e-06, "loss": 1.8545, "step": 28142 }, { "epoch": 5.36, "learning_rate": 1.3931553860819828e-06, "loss": 1.7303, "step": 28143 }, { "epoch": 5.36, "learning_rate": 1.3930981887511916e-06, "loss": 1.6708, "step": 28144 }, { "epoch": 5.36, "learning_rate": 1.3930409914204004e-06, "loss": 2.9077, "step": 28145 }, { "epoch": 5.36, "learning_rate": 1.3929837940896092e-06, "loss": 2.7166, "step": 28146 }, { "epoch": 5.36, "learning_rate": 1.392926596758818e-06, "loss": 1.6068, "step": 28147 }, { "epoch": 5.36, "learning_rate": 1.3928693994280266e-06, "loss": 1.2766, "step": 28148 }, { "epoch": 5.36, "learning_rate": 1.3928122020972354e-06, "loss": 1.459, "step": 28149 }, { "epoch": 5.36, "learning_rate": 1.3927550047664444e-06, "loss": 1.7356, "step": 28150 }, { "epoch": 5.36, "learning_rate": 1.3926978074356532e-06, "loss": 2.0604, "step": 28151 }, { "epoch": 5.36, "learning_rate": 1.3926406101048618e-06, "loss": 1.7369, "step": 28152 }, { "epoch": 5.36, "learning_rate": 1.3925834127740706e-06, "loss": 1.2911, "step": 28153 }, { "epoch": 5.36, "learning_rate": 1.3925262154432794e-06, "loss": 2.7401, "step": 28154 }, { "epoch": 5.36, "learning_rate": 1.3924690181124882e-06, "loss": 2.6665, "step": 28155 }, { "epoch": 5.36, "learning_rate": 1.392411820781697e-06, "loss": 1.5898, "step": 28156 }, { "epoch": 5.36, "learning_rate": 1.3923546234509057e-06, "loss": 1.422, "step": 28157 }, { "epoch": 5.36, "learning_rate": 1.3922974261201143e-06, "loss": 0.8886, "step": 28158 }, { "epoch": 5.36, "learning_rate": 1.3922402287893231e-06, "loss": 0.6962, "step": 28159 }, { "epoch": 5.36, "learning_rate": 1.392183031458532e-06, "loss": 2.6321, "step": 28160 }, { "epoch": 5.36, "learning_rate": 1.3921258341277407e-06, "loss": 1.6075, "step": 28161 }, { "epoch": 5.36, "learning_rate": 1.3920686367969495e-06, "loss": 1.0583, "step": 28162 }, { "epoch": 5.36, "learning_rate": 1.3920114394661583e-06, "loss": 1.558, "step": 28163 }, { "epoch": 5.36, "learning_rate": 1.391954242135367e-06, "loss": 1.3191, "step": 28164 }, { "epoch": 5.36, "learning_rate": 1.3918970448045757e-06, "loss": 1.8835, "step": 28165 }, { "epoch": 5.36, "learning_rate": 1.3918398474737845e-06, "loss": 2.8932, "step": 28166 }, { "epoch": 5.37, "learning_rate": 1.3917826501429935e-06, "loss": 1.5789, "step": 28167 }, { "epoch": 5.37, "learning_rate": 1.3917254528122023e-06, "loss": 1.7776, "step": 28168 }, { "epoch": 5.37, "learning_rate": 1.3916682554814108e-06, "loss": 1.1425, "step": 28169 }, { "epoch": 5.37, "learning_rate": 1.3916110581506196e-06, "loss": 1.5856, "step": 28170 }, { "epoch": 5.37, "learning_rate": 1.3915538608198284e-06, "loss": 2.2377, "step": 28171 }, { "epoch": 5.37, "learning_rate": 1.3914966634890372e-06, "loss": 2.0188, "step": 28172 }, { "epoch": 5.37, "learning_rate": 1.391439466158246e-06, "loss": 1.8095, "step": 28173 }, { "epoch": 5.37, "learning_rate": 1.3913822688274548e-06, "loss": 0.7998, "step": 28174 }, { "epoch": 5.37, "learning_rate": 1.3913250714966636e-06, "loss": 1.2992, "step": 28175 }, { "epoch": 5.37, "learning_rate": 1.3912678741658722e-06, "loss": 1.3371, "step": 28176 }, { "epoch": 5.37, "learning_rate": 1.391210676835081e-06, "loss": 1.3755, "step": 28177 }, { "epoch": 5.37, "learning_rate": 1.39115347950429e-06, "loss": 1.397, "step": 28178 }, { "epoch": 5.37, "learning_rate": 1.3910962821734986e-06, "loss": 2.5351, "step": 28179 }, { "epoch": 5.37, "learning_rate": 1.3910390848427074e-06, "loss": 2.5707, "step": 28180 }, { "epoch": 5.37, "learning_rate": 1.3909818875119162e-06, "loss": 2.2514, "step": 28181 }, { "epoch": 5.37, "learning_rate": 1.3909246901811247e-06, "loss": 2.4597, "step": 28182 }, { "epoch": 5.37, "learning_rate": 1.3908674928503337e-06, "loss": 1.2302, "step": 28183 }, { "epoch": 5.37, "learning_rate": 1.3908102955195425e-06, "loss": 1.2841, "step": 28184 }, { "epoch": 5.37, "learning_rate": 1.3907530981887513e-06, "loss": 3.3057, "step": 28185 }, { "epoch": 5.37, "learning_rate": 1.39069590085796e-06, "loss": 2.3463, "step": 28186 }, { "epoch": 5.37, "learning_rate": 1.3906387035271687e-06, "loss": 2.1046, "step": 28187 }, { "epoch": 5.37, "learning_rate": 1.3905815061963775e-06, "loss": 1.1021, "step": 28188 }, { "epoch": 5.37, "learning_rate": 1.3905243088655863e-06, "loss": 2.0759, "step": 28189 }, { "epoch": 5.37, "learning_rate": 1.390467111534795e-06, "loss": 1.4378, "step": 28190 }, { "epoch": 5.37, "learning_rate": 1.3904099142040039e-06, "loss": 2.4834, "step": 28191 }, { "epoch": 5.37, "learning_rate": 1.3903527168732127e-06, "loss": 1.457, "step": 28192 }, { "epoch": 5.37, "learning_rate": 1.3902955195424213e-06, "loss": 1.6518, "step": 28193 }, { "epoch": 5.37, "learning_rate": 1.39023832221163e-06, "loss": 0.6643, "step": 28194 }, { "epoch": 5.37, "learning_rate": 1.390181124880839e-06, "loss": 1.449, "step": 28195 }, { "epoch": 5.37, "learning_rate": 1.3901239275500476e-06, "loss": 2.4699, "step": 28196 }, { "epoch": 5.37, "learning_rate": 1.3900667302192564e-06, "loss": 0.4129, "step": 28197 }, { "epoch": 5.37, "learning_rate": 1.3900095328884652e-06, "loss": 1.2532, "step": 28198 }, { "epoch": 5.37, "learning_rate": 1.389952335557674e-06, "loss": 1.7823, "step": 28199 }, { "epoch": 5.37, "learning_rate": 1.3898951382268828e-06, "loss": 2.2002, "step": 28200 }, { "epoch": 5.37, "learning_rate": 1.3898379408960916e-06, "loss": 1.144, "step": 28201 }, { "epoch": 5.37, "learning_rate": 1.3897807435653004e-06, "loss": 1.9694, "step": 28202 }, { "epoch": 5.37, "learning_rate": 1.389723546234509e-06, "loss": 2.176, "step": 28203 }, { "epoch": 5.37, "learning_rate": 1.3896663489037178e-06, "loss": 2.29, "step": 28204 }, { "epoch": 5.37, "learning_rate": 1.3896091515729266e-06, "loss": 1.5255, "step": 28205 }, { "epoch": 5.37, "learning_rate": 1.3895519542421354e-06, "loss": 2.1137, "step": 28206 }, { "epoch": 5.37, "learning_rate": 1.3894947569113442e-06, "loss": 2.3584, "step": 28207 }, { "epoch": 5.37, "learning_rate": 1.389437559580553e-06, "loss": 2.9547, "step": 28208 }, { "epoch": 5.37, "learning_rate": 1.3893803622497618e-06, "loss": 2.499, "step": 28209 }, { "epoch": 5.37, "learning_rate": 1.3893231649189703e-06, "loss": 2.0984, "step": 28210 }, { "epoch": 5.37, "learning_rate": 1.3892659675881793e-06, "loss": 1.8758, "step": 28211 }, { "epoch": 5.37, "learning_rate": 1.3892087702573881e-06, "loss": 1.3988, "step": 28212 }, { "epoch": 5.37, "learning_rate": 1.3891515729265967e-06, "loss": 2.3194, "step": 28213 }, { "epoch": 5.37, "learning_rate": 1.3890943755958055e-06, "loss": 1.9873, "step": 28214 }, { "epoch": 5.37, "learning_rate": 1.3890371782650143e-06, "loss": 1.6521, "step": 28215 }, { "epoch": 5.37, "learning_rate": 1.388979980934223e-06, "loss": 2.3594, "step": 28216 }, { "epoch": 5.37, "learning_rate": 1.3889227836034319e-06, "loss": 2.6115, "step": 28217 }, { "epoch": 5.37, "learning_rate": 1.3888655862726407e-06, "loss": 2.7204, "step": 28218 }, { "epoch": 5.38, "learning_rate": 1.3888083889418495e-06, "loss": 2.1142, "step": 28219 }, { "epoch": 5.38, "learning_rate": 1.388751191611058e-06, "loss": 1.8763, "step": 28220 }, { "epoch": 5.38, "learning_rate": 1.3886939942802669e-06, "loss": 3.0373, "step": 28221 }, { "epoch": 5.38, "learning_rate": 1.3886367969494759e-06, "loss": 1.9517, "step": 28222 }, { "epoch": 5.38, "learning_rate": 1.3885795996186844e-06, "loss": 2.8946, "step": 28223 }, { "epoch": 5.38, "learning_rate": 1.3885224022878932e-06, "loss": 2.5202, "step": 28224 }, { "epoch": 5.38, "learning_rate": 1.388465204957102e-06, "loss": 3.8378, "step": 28225 }, { "epoch": 5.38, "learning_rate": 1.3884080076263108e-06, "loss": 3.0611, "step": 28226 }, { "epoch": 5.38, "learning_rate": 1.3883508102955196e-06, "loss": 1.2776, "step": 28227 }, { "epoch": 5.38, "learning_rate": 1.3882936129647284e-06, "loss": 2.0716, "step": 28228 }, { "epoch": 5.38, "learning_rate": 1.3882364156339372e-06, "loss": 1.7295, "step": 28229 }, { "epoch": 5.38, "learning_rate": 1.3881792183031458e-06, "loss": 2.1512, "step": 28230 }, { "epoch": 5.38, "learning_rate": 1.3881220209723546e-06, "loss": 1.681, "step": 28231 }, { "epoch": 5.38, "learning_rate": 1.3880648236415634e-06, "loss": 2.1375, "step": 28232 }, { "epoch": 5.38, "learning_rate": 1.3880076263107722e-06, "loss": 0.9084, "step": 28233 }, { "epoch": 5.38, "learning_rate": 1.387950428979981e-06, "loss": 1.6649, "step": 28234 }, { "epoch": 5.38, "learning_rate": 1.3878932316491898e-06, "loss": 3.2856, "step": 28235 }, { "epoch": 5.38, "learning_rate": 1.3878360343183986e-06, "loss": 1.404, "step": 28236 }, { "epoch": 5.38, "learning_rate": 1.3877788369876071e-06, "loss": 1.675, "step": 28237 }, { "epoch": 5.38, "learning_rate": 1.387721639656816e-06, "loss": 2.631, "step": 28238 }, { "epoch": 5.38, "learning_rate": 1.387664442326025e-06, "loss": 1.647, "step": 28239 }, { "epoch": 5.38, "learning_rate": 1.3876072449952337e-06, "loss": 2.9044, "step": 28240 }, { "epoch": 5.38, "learning_rate": 1.3875500476644423e-06, "loss": 2.0135, "step": 28241 }, { "epoch": 5.38, "learning_rate": 1.387492850333651e-06, "loss": 2.2097, "step": 28242 }, { "epoch": 5.38, "learning_rate": 1.38743565300286e-06, "loss": 1.9855, "step": 28243 }, { "epoch": 5.38, "learning_rate": 1.3873784556720687e-06, "loss": 1.7331, "step": 28244 }, { "epoch": 5.38, "learning_rate": 1.3873212583412775e-06, "loss": 2.018, "step": 28245 }, { "epoch": 5.38, "learning_rate": 1.3872640610104863e-06, "loss": 1.3514, "step": 28246 }, { "epoch": 5.38, "learning_rate": 1.3872068636796949e-06, "loss": 0.5096, "step": 28247 }, { "epoch": 5.38, "learning_rate": 1.3871496663489037e-06, "loss": 1.4829, "step": 28248 }, { "epoch": 5.38, "learning_rate": 1.3870924690181124e-06, "loss": 1.5132, "step": 28249 }, { "epoch": 5.38, "learning_rate": 1.3870352716873215e-06, "loss": 2.2995, "step": 28250 }, { "epoch": 5.38, "learning_rate": 1.38697807435653e-06, "loss": 2.4851, "step": 28251 }, { "epoch": 5.38, "learning_rate": 1.3869208770257388e-06, "loss": 2.0844, "step": 28252 }, { "epoch": 5.38, "learning_rate": 1.3868636796949476e-06, "loss": 1.2285, "step": 28253 }, { "epoch": 5.38, "learning_rate": 1.3868064823641562e-06, "loss": 1.2675, "step": 28254 }, { "epoch": 5.38, "learning_rate": 1.3867492850333652e-06, "loss": 1.605, "step": 28255 }, { "epoch": 5.38, "learning_rate": 1.386692087702574e-06, "loss": 1.6587, "step": 28256 }, { "epoch": 5.38, "learning_rate": 1.3866348903717828e-06, "loss": 0.4731, "step": 28257 }, { "epoch": 5.38, "learning_rate": 1.3865776930409914e-06, "loss": 1.1022, "step": 28258 }, { "epoch": 5.38, "learning_rate": 1.3865204957102002e-06, "loss": 1.9709, "step": 28259 }, { "epoch": 5.38, "learning_rate": 1.386463298379409e-06, "loss": 1.8187, "step": 28260 }, { "epoch": 5.38, "learning_rate": 1.3864061010486178e-06, "loss": 2.0711, "step": 28261 }, { "epoch": 5.38, "learning_rate": 1.3863489037178266e-06, "loss": 1.286, "step": 28262 }, { "epoch": 5.38, "learning_rate": 1.3862917063870353e-06, "loss": 1.7882, "step": 28263 }, { "epoch": 5.38, "learning_rate": 1.3862345090562441e-06, "loss": 1.5535, "step": 28264 }, { "epoch": 5.38, "learning_rate": 1.3861773117254527e-06, "loss": 1.9711, "step": 28265 }, { "epoch": 5.38, "learning_rate": 1.3861201143946617e-06, "loss": 1.9016, "step": 28266 }, { "epoch": 5.38, "learning_rate": 1.3860629170638705e-06, "loss": 2.3998, "step": 28267 }, { "epoch": 5.38, "learning_rate": 1.3860057197330791e-06, "loss": 1.5527, "step": 28268 }, { "epoch": 5.38, "learning_rate": 1.385948522402288e-06, "loss": 2.392, "step": 28269 }, { "epoch": 5.38, "learning_rate": 1.3858913250714967e-06, "loss": 0.8995, "step": 28270 }, { "epoch": 5.38, "learning_rate": 1.3858341277407055e-06, "loss": 2.6275, "step": 28271 }, { "epoch": 5.39, "learning_rate": 1.3857769304099143e-06, "loss": 2.6349, "step": 28272 }, { "epoch": 5.39, "learning_rate": 1.385719733079123e-06, "loss": 1.4205, "step": 28273 }, { "epoch": 5.39, "learning_rate": 1.3856625357483319e-06, "loss": 2.6354, "step": 28274 }, { "epoch": 5.39, "learning_rate": 1.3856053384175405e-06, "loss": 1.8528, "step": 28275 }, { "epoch": 5.39, "learning_rate": 1.3855481410867492e-06, "loss": 2.0861, "step": 28276 }, { "epoch": 5.39, "learning_rate": 1.385490943755958e-06, "loss": 1.6103, "step": 28277 }, { "epoch": 5.39, "learning_rate": 1.3854337464251668e-06, "loss": 1.5551, "step": 28278 }, { "epoch": 5.39, "learning_rate": 1.3853765490943756e-06, "loss": 1.7338, "step": 28279 }, { "epoch": 5.39, "learning_rate": 1.3853193517635844e-06, "loss": 2.9819, "step": 28280 }, { "epoch": 5.39, "learning_rate": 1.3852621544327932e-06, "loss": 2.1982, "step": 28281 }, { "epoch": 5.39, "learning_rate": 1.3852049571020018e-06, "loss": 0.7962, "step": 28282 }, { "epoch": 5.39, "learning_rate": 1.3851477597712108e-06, "loss": 1.5453, "step": 28283 }, { "epoch": 5.39, "learning_rate": 1.3850905624404196e-06, "loss": 2.5924, "step": 28284 }, { "epoch": 5.39, "learning_rate": 1.3850333651096282e-06, "loss": 1.5288, "step": 28285 }, { "epoch": 5.39, "learning_rate": 1.384976167778837e-06, "loss": 3.2847, "step": 28286 }, { "epoch": 5.39, "learning_rate": 1.3849189704480458e-06, "loss": 1.4242, "step": 28287 }, { "epoch": 5.39, "learning_rate": 1.3848617731172546e-06, "loss": 1.3646, "step": 28288 }, { "epoch": 5.39, "learning_rate": 1.3848045757864634e-06, "loss": 2.1242, "step": 28289 }, { "epoch": 5.39, "learning_rate": 1.3847473784556721e-06, "loss": 2.6843, "step": 28290 }, { "epoch": 5.39, "learning_rate": 1.384690181124881e-06, "loss": 1.9668, "step": 28291 }, { "epoch": 5.39, "learning_rate": 1.3846329837940895e-06, "loss": 1.4562, "step": 28292 }, { "epoch": 5.39, "learning_rate": 1.3845757864632983e-06, "loss": 3.0975, "step": 28293 }, { "epoch": 5.39, "learning_rate": 1.3845185891325073e-06, "loss": 1.4471, "step": 28294 }, { "epoch": 5.39, "learning_rate": 1.384461391801716e-06, "loss": 2.133, "step": 28295 }, { "epoch": 5.39, "learning_rate": 1.3844041944709247e-06, "loss": 1.9508, "step": 28296 }, { "epoch": 5.39, "learning_rate": 1.3843469971401335e-06, "loss": 2.0493, "step": 28297 }, { "epoch": 5.39, "learning_rate": 1.3842897998093423e-06, "loss": 1.361, "step": 28298 }, { "epoch": 5.39, "learning_rate": 1.384232602478551e-06, "loss": 1.2841, "step": 28299 }, { "epoch": 5.39, "learning_rate": 1.3841754051477599e-06, "loss": 2.4067, "step": 28300 }, { "epoch": 5.39, "learning_rate": 1.3841182078169687e-06, "loss": 2.8006, "step": 28301 }, { "epoch": 5.39, "learning_rate": 1.3840610104861772e-06, "loss": 1.6508, "step": 28302 }, { "epoch": 5.39, "learning_rate": 1.384003813155386e-06, "loss": 2.1704, "step": 28303 }, { "epoch": 5.39, "learning_rate": 1.3839466158245948e-06, "loss": 1.8194, "step": 28304 }, { "epoch": 5.39, "learning_rate": 1.3838894184938038e-06, "loss": 2.3288, "step": 28305 }, { "epoch": 5.39, "learning_rate": 1.3838322211630124e-06, "loss": 0.9356, "step": 28306 }, { "epoch": 5.39, "learning_rate": 1.3837750238322212e-06, "loss": 1.617, "step": 28307 }, { "epoch": 5.39, "learning_rate": 1.38371782650143e-06, "loss": 1.8725, "step": 28308 }, { "epoch": 5.39, "learning_rate": 1.3836606291706386e-06, "loss": 2.1201, "step": 28309 }, { "epoch": 5.39, "learning_rate": 1.3836034318398476e-06, "loss": 1.5288, "step": 28310 }, { "epoch": 5.39, "learning_rate": 1.3835462345090564e-06, "loss": 2.1097, "step": 28311 }, { "epoch": 5.39, "learning_rate": 1.383489037178265e-06, "loss": 1.8518, "step": 28312 }, { "epoch": 5.39, "learning_rate": 1.3834318398474738e-06, "loss": 1.8157, "step": 28313 }, { "epoch": 5.39, "learning_rate": 1.3833746425166826e-06, "loss": 0.7474, "step": 28314 }, { "epoch": 5.39, "learning_rate": 1.3833174451858914e-06, "loss": 0.8766, "step": 28315 }, { "epoch": 5.39, "learning_rate": 1.3832602478551002e-06, "loss": 2.3291, "step": 28316 }, { "epoch": 5.39, "learning_rate": 1.383203050524309e-06, "loss": 2.5124, "step": 28317 }, { "epoch": 5.39, "learning_rate": 1.3831458531935177e-06, "loss": 2.5834, "step": 28318 }, { "epoch": 5.39, "learning_rate": 1.3830886558627263e-06, "loss": 0.6477, "step": 28319 }, { "epoch": 5.39, "learning_rate": 1.3830314585319351e-06, "loss": 1.3107, "step": 28320 }, { "epoch": 5.39, "learning_rate": 1.382974261201144e-06, "loss": 0.7716, "step": 28321 }, { "epoch": 5.39, "learning_rate": 1.382917063870353e-06, "loss": 1.8792, "step": 28322 }, { "epoch": 5.39, "learning_rate": 1.3828598665395615e-06, "loss": 3.395, "step": 28323 }, { "epoch": 5.4, "learning_rate": 1.3828026692087703e-06, "loss": 1.2814, "step": 28324 }, { "epoch": 5.4, "learning_rate": 1.382745471877979e-06, "loss": 2.5076, "step": 28325 }, { "epoch": 5.4, "learning_rate": 1.3826882745471877e-06, "loss": 0.7953, "step": 28326 }, { "epoch": 5.4, "learning_rate": 1.3826310772163967e-06, "loss": 1.0779, "step": 28327 }, { "epoch": 5.4, "learning_rate": 1.3825738798856055e-06, "loss": 1.9232, "step": 28328 }, { "epoch": 5.4, "learning_rate": 1.3825166825548143e-06, "loss": 2.8661, "step": 28329 }, { "epoch": 5.4, "learning_rate": 1.3824594852240228e-06, "loss": 1.4069, "step": 28330 }, { "epoch": 5.4, "learning_rate": 1.3824022878932316e-06, "loss": 2.6349, "step": 28331 }, { "epoch": 5.4, "learning_rate": 1.3823450905624404e-06, "loss": 2.4558, "step": 28332 }, { "epoch": 5.4, "learning_rate": 1.3822878932316492e-06, "loss": 2.8067, "step": 28333 }, { "epoch": 5.4, "learning_rate": 1.382230695900858e-06, "loss": 2.2242, "step": 28334 }, { "epoch": 5.4, "learning_rate": 1.3821734985700668e-06, "loss": 2.7258, "step": 28335 }, { "epoch": 5.4, "learning_rate": 1.3821163012392754e-06, "loss": 3.1746, "step": 28336 }, { "epoch": 5.4, "learning_rate": 1.3820591039084842e-06, "loss": 2.0231, "step": 28337 }, { "epoch": 5.4, "learning_rate": 1.3820019065776932e-06, "loss": 2.5752, "step": 28338 }, { "epoch": 5.4, "learning_rate": 1.381944709246902e-06, "loss": 2.9509, "step": 28339 }, { "epoch": 5.4, "learning_rate": 1.3818875119161106e-06, "loss": 2.4396, "step": 28340 }, { "epoch": 5.4, "learning_rate": 1.3818303145853194e-06, "loss": 1.4028, "step": 28341 }, { "epoch": 5.4, "learning_rate": 1.3817731172545282e-06, "loss": 1.7622, "step": 28342 }, { "epoch": 5.4, "learning_rate": 1.381715919923737e-06, "loss": 2.4528, "step": 28343 }, { "epoch": 5.4, "learning_rate": 1.3816587225929457e-06, "loss": 2.7188, "step": 28344 }, { "epoch": 5.4, "learning_rate": 1.3816015252621545e-06, "loss": 2.8116, "step": 28345 }, { "epoch": 5.4, "learning_rate": 1.3815443279313633e-06, "loss": 2.2329, "step": 28346 }, { "epoch": 5.4, "learning_rate": 1.381487130600572e-06, "loss": 2.1364, "step": 28347 }, { "epoch": 5.4, "learning_rate": 1.3814299332697807e-06, "loss": 1.9092, "step": 28348 }, { "epoch": 5.4, "learning_rate": 1.3813727359389897e-06, "loss": 2.9452, "step": 28349 }, { "epoch": 5.4, "learning_rate": 1.3813155386081983e-06, "loss": 1.5796, "step": 28350 }, { "epoch": 5.4, "learning_rate": 1.381258341277407e-06, "loss": 2.481, "step": 28351 }, { "epoch": 5.4, "learning_rate": 1.3812011439466159e-06, "loss": 1.3719, "step": 28352 }, { "epoch": 5.4, "learning_rate": 1.3811439466158245e-06, "loss": 2.6593, "step": 28353 }, { "epoch": 5.4, "learning_rate": 1.3810867492850335e-06, "loss": 2.587, "step": 28354 }, { "epoch": 5.4, "learning_rate": 1.3810295519542423e-06, "loss": 1.7481, "step": 28355 }, { "epoch": 5.4, "learning_rate": 1.380972354623451e-06, "loss": 1.9422, "step": 28356 }, { "epoch": 5.4, "learning_rate": 1.3809151572926596e-06, "loss": 2.3599, "step": 28357 }, { "epoch": 5.4, "learning_rate": 1.3808579599618684e-06, "loss": 2.0353, "step": 28358 }, { "epoch": 5.4, "learning_rate": 1.3808007626310772e-06, "loss": 1.7268, "step": 28359 }, { "epoch": 5.4, "learning_rate": 1.380743565300286e-06, "loss": 2.0678, "step": 28360 }, { "epoch": 5.4, "learning_rate": 1.3806863679694948e-06, "loss": 2.5464, "step": 28361 }, { "epoch": 5.4, "learning_rate": 1.3806291706387036e-06, "loss": 2.4806, "step": 28362 }, { "epoch": 5.4, "learning_rate": 1.3805719733079124e-06, "loss": 0.906, "step": 28363 }, { "epoch": 5.4, "learning_rate": 1.380514775977121e-06, "loss": 0.6458, "step": 28364 }, { "epoch": 5.4, "learning_rate": 1.3804575786463298e-06, "loss": 1.4805, "step": 28365 }, { "epoch": 5.4, "learning_rate": 1.3804003813155388e-06, "loss": 2.0936, "step": 28366 }, { "epoch": 5.4, "learning_rate": 1.3803431839847474e-06, "loss": 1.2896, "step": 28367 }, { "epoch": 5.4, "learning_rate": 1.3802859866539562e-06, "loss": 1.0511, "step": 28368 }, { "epoch": 5.4, "learning_rate": 1.380228789323165e-06, "loss": 0.6607, "step": 28369 }, { "epoch": 5.4, "learning_rate": 1.3801715919923737e-06, "loss": 1.9325, "step": 28370 }, { "epoch": 5.4, "learning_rate": 1.3801143946615825e-06, "loss": 1.5283, "step": 28371 }, { "epoch": 5.4, "learning_rate": 1.3800571973307913e-06, "loss": 2.3899, "step": 28372 }, { "epoch": 5.4, "learning_rate": 1.3800000000000001e-06, "loss": 1.8062, "step": 28373 }, { "epoch": 5.4, "learning_rate": 1.3799428026692087e-06, "loss": 1.2344, "step": 28374 }, { "epoch": 5.4, "learning_rate": 1.3798856053384175e-06, "loss": 1.7844, "step": 28375 }, { "epoch": 5.4, "learning_rate": 1.3798284080076263e-06, "loss": 2.0894, "step": 28376 }, { "epoch": 5.41, "learning_rate": 1.379771210676835e-06, "loss": 1.3353, "step": 28377 }, { "epoch": 5.41, "learning_rate": 1.3797140133460439e-06, "loss": 3.0667, "step": 28378 }, { "epoch": 5.41, "learning_rate": 1.3796568160152527e-06, "loss": 1.8685, "step": 28379 }, { "epoch": 5.41, "learning_rate": 1.3795996186844615e-06, "loss": 2.7681, "step": 28380 }, { "epoch": 5.41, "learning_rate": 1.37954242135367e-06, "loss": 2.23, "step": 28381 }, { "epoch": 5.41, "learning_rate": 1.379485224022879e-06, "loss": 1.9478, "step": 28382 }, { "epoch": 5.41, "learning_rate": 1.3794280266920879e-06, "loss": 0.9269, "step": 28383 }, { "epoch": 5.41, "learning_rate": 1.3793708293612964e-06, "loss": 1.2907, "step": 28384 }, { "epoch": 5.41, "learning_rate": 1.3793136320305052e-06, "loss": 2.0377, "step": 28385 }, { "epoch": 5.41, "learning_rate": 1.379256434699714e-06, "loss": 1.7133, "step": 28386 }, { "epoch": 5.41, "learning_rate": 1.3791992373689228e-06, "loss": 2.1345, "step": 28387 }, { "epoch": 5.41, "learning_rate": 1.3791420400381316e-06, "loss": 2.0632, "step": 28388 }, { "epoch": 5.41, "learning_rate": 1.3790848427073404e-06, "loss": 1.7646, "step": 28389 }, { "epoch": 5.41, "learning_rate": 1.3790276453765492e-06, "loss": 1.838, "step": 28390 }, { "epoch": 5.41, "learning_rate": 1.3789704480457578e-06, "loss": 2.7222, "step": 28391 }, { "epoch": 5.41, "learning_rate": 1.3789132507149666e-06, "loss": 3.3843, "step": 28392 }, { "epoch": 5.41, "learning_rate": 1.3788560533841754e-06, "loss": 2.1319, "step": 28393 }, { "epoch": 5.41, "learning_rate": 1.3787988560533844e-06, "loss": 1.7965, "step": 28394 }, { "epoch": 5.41, "learning_rate": 1.378741658722593e-06, "loss": 2.4804, "step": 28395 }, { "epoch": 5.41, "learning_rate": 1.3786844613918018e-06, "loss": 2.9822, "step": 28396 }, { "epoch": 5.41, "learning_rate": 1.3786272640610105e-06, "loss": 1.3061, "step": 28397 }, { "epoch": 5.41, "learning_rate": 1.3785700667302193e-06, "loss": 2.3852, "step": 28398 }, { "epoch": 5.41, "learning_rate": 1.3785128693994281e-06, "loss": 2.0974, "step": 28399 }, { "epoch": 5.41, "learning_rate": 1.378455672068637e-06, "loss": 1.7571, "step": 28400 }, { "epoch": 5.41, "learning_rate": 1.3783984747378455e-06, "loss": 1.8423, "step": 28401 }, { "epoch": 5.41, "learning_rate": 1.3783412774070543e-06, "loss": 1.927, "step": 28402 }, { "epoch": 5.41, "learning_rate": 1.378284080076263e-06, "loss": 2.7902, "step": 28403 }, { "epoch": 5.41, "learning_rate": 1.3782268827454719e-06, "loss": 0.6508, "step": 28404 }, { "epoch": 5.41, "learning_rate": 1.3781696854146807e-06, "loss": 2.6667, "step": 28405 }, { "epoch": 5.41, "learning_rate": 1.3781124880838895e-06, "loss": 2.2243, "step": 28406 }, { "epoch": 5.41, "learning_rate": 1.3780552907530983e-06, "loss": 1.4267, "step": 28407 }, { "epoch": 5.41, "learning_rate": 1.3779980934223069e-06, "loss": 2.0344, "step": 28408 }, { "epoch": 5.41, "learning_rate": 1.3779408960915156e-06, "loss": 2.5086, "step": 28409 }, { "epoch": 5.41, "learning_rate": 1.3778836987607247e-06, "loss": 2.1123, "step": 28410 }, { "epoch": 5.41, "learning_rate": 1.3778265014299334e-06, "loss": 1.2306, "step": 28411 }, { "epoch": 5.41, "learning_rate": 1.377769304099142e-06, "loss": 2.4082, "step": 28412 }, { "epoch": 5.41, "learning_rate": 1.3777121067683508e-06, "loss": 2.2563, "step": 28413 }, { "epoch": 5.41, "learning_rate": 1.3776549094375596e-06, "loss": 1.2541, "step": 28414 }, { "epoch": 5.41, "learning_rate": 1.3775977121067684e-06, "loss": 2.9821, "step": 28415 }, { "epoch": 5.41, "learning_rate": 1.3775405147759772e-06, "loss": 2.0766, "step": 28416 }, { "epoch": 5.41, "learning_rate": 1.377483317445186e-06, "loss": 1.7007, "step": 28417 }, { "epoch": 5.41, "learning_rate": 1.3774261201143946e-06, "loss": 2.4476, "step": 28418 }, { "epoch": 5.41, "learning_rate": 1.3773689227836034e-06, "loss": 2.0039, "step": 28419 }, { "epoch": 5.41, "learning_rate": 1.3773117254528122e-06, "loss": 1.49, "step": 28420 }, { "epoch": 5.41, "learning_rate": 1.3772545281220212e-06, "loss": 1.508, "step": 28421 }, { "epoch": 5.41, "learning_rate": 1.3771973307912298e-06, "loss": 1.2104, "step": 28422 }, { "epoch": 5.41, "learning_rate": 1.3771401334604385e-06, "loss": 1.5926, "step": 28423 }, { "epoch": 5.41, "learning_rate": 1.3770829361296473e-06, "loss": 1.2542, "step": 28424 }, { "epoch": 5.41, "learning_rate": 1.377025738798856e-06, "loss": 2.0774, "step": 28425 }, { "epoch": 5.41, "learning_rate": 1.376968541468065e-06, "loss": 3.0023, "step": 28426 }, { "epoch": 5.41, "learning_rate": 1.3769113441372737e-06, "loss": 2.5183, "step": 28427 }, { "epoch": 5.41, "learning_rate": 1.3768541468064825e-06, "loss": 2.573, "step": 28428 }, { "epoch": 5.42, "learning_rate": 1.376796949475691e-06, "loss": 2.5376, "step": 28429 }, { "epoch": 5.42, "learning_rate": 1.3767397521448999e-06, "loss": 2.1082, "step": 28430 }, { "epoch": 5.42, "learning_rate": 1.3766825548141087e-06, "loss": 3.1349, "step": 28431 }, { "epoch": 5.42, "learning_rate": 1.3766253574833175e-06, "loss": 1.2485, "step": 28432 }, { "epoch": 5.42, "learning_rate": 1.3765681601525263e-06, "loss": 1.926, "step": 28433 }, { "epoch": 5.42, "learning_rate": 1.376510962821735e-06, "loss": 1.5545, "step": 28434 }, { "epoch": 5.42, "learning_rate": 1.3764537654909439e-06, "loss": 1.3719, "step": 28435 }, { "epoch": 5.42, "learning_rate": 1.3763965681601524e-06, "loss": 1.6956, "step": 28436 }, { "epoch": 5.42, "learning_rate": 1.3763393708293612e-06, "loss": 1.8062, "step": 28437 }, { "epoch": 5.42, "learning_rate": 1.3762821734985702e-06, "loss": 1.8057, "step": 28438 }, { "epoch": 5.42, "learning_rate": 1.3762249761677788e-06, "loss": 2.389, "step": 28439 }, { "epoch": 5.42, "learning_rate": 1.3761677788369876e-06, "loss": 1.9314, "step": 28440 }, { "epoch": 5.42, "learning_rate": 1.3761105815061964e-06, "loss": 1.4394, "step": 28441 }, { "epoch": 5.42, "learning_rate": 1.3760533841754052e-06, "loss": 1.0999, "step": 28442 }, { "epoch": 5.42, "learning_rate": 1.375996186844614e-06, "loss": 2.6946, "step": 28443 }, { "epoch": 5.42, "learning_rate": 1.3759389895138228e-06, "loss": 2.8293, "step": 28444 }, { "epoch": 5.42, "learning_rate": 1.3758817921830316e-06, "loss": 2.7765, "step": 28445 }, { "epoch": 5.42, "learning_rate": 1.3758245948522402e-06, "loss": 1.9981, "step": 28446 }, { "epoch": 5.42, "learning_rate": 1.375767397521449e-06, "loss": 2.1831, "step": 28447 }, { "epoch": 5.42, "learning_rate": 1.3757102001906578e-06, "loss": 1.6003, "step": 28448 }, { "epoch": 5.42, "learning_rate": 1.3756530028598666e-06, "loss": 1.6166, "step": 28449 }, { "epoch": 5.42, "learning_rate": 1.3755958055290753e-06, "loss": 1.5409, "step": 28450 }, { "epoch": 5.42, "learning_rate": 1.3755386081982841e-06, "loss": 1.6047, "step": 28451 }, { "epoch": 5.42, "learning_rate": 1.375481410867493e-06, "loss": 2.4445, "step": 28452 }, { "epoch": 5.42, "learning_rate": 1.3754242135367015e-06, "loss": 2.4237, "step": 28453 }, { "epoch": 5.42, "learning_rate": 1.3753670162059105e-06, "loss": 1.0225, "step": 28454 }, { "epoch": 5.42, "learning_rate": 1.3753098188751193e-06, "loss": 2.6583, "step": 28455 }, { "epoch": 5.42, "learning_rate": 1.375252621544328e-06, "loss": 1.9877, "step": 28456 }, { "epoch": 5.42, "learning_rate": 1.3751954242135367e-06, "loss": 1.1273, "step": 28457 }, { "epoch": 5.42, "learning_rate": 1.3751382268827455e-06, "loss": 1.4244, "step": 28458 }, { "epoch": 5.42, "learning_rate": 1.3750810295519543e-06, "loss": 1.8126, "step": 28459 }, { "epoch": 5.42, "learning_rate": 1.375023832221163e-06, "loss": 1.0807, "step": 28460 }, { "epoch": 5.42, "learning_rate": 1.3749666348903719e-06, "loss": 1.7688, "step": 28461 }, { "epoch": 5.42, "learning_rate": 1.3749094375595807e-06, "loss": 1.0724, "step": 28462 }, { "epoch": 5.42, "learning_rate": 1.3748522402287892e-06, "loss": 1.0231, "step": 28463 }, { "epoch": 5.42, "learning_rate": 1.374795042897998e-06, "loss": 1.0288, "step": 28464 }, { "epoch": 5.42, "learning_rate": 1.374737845567207e-06, "loss": 1.5775, "step": 28465 }, { "epoch": 5.42, "learning_rate": 1.3746806482364156e-06, "loss": 0.861, "step": 28466 }, { "epoch": 5.42, "learning_rate": 1.3746234509056244e-06, "loss": 1.6556, "step": 28467 }, { "epoch": 5.42, "learning_rate": 1.3745662535748332e-06, "loss": 2.3571, "step": 28468 }, { "epoch": 5.42, "learning_rate": 1.374509056244042e-06, "loss": 1.8247, "step": 28469 }, { "epoch": 5.42, "learning_rate": 1.3744518589132508e-06, "loss": 1.7711, "step": 28470 }, { "epoch": 5.42, "learning_rate": 1.3743946615824596e-06, "loss": 1.3368, "step": 28471 }, { "epoch": 5.42, "learning_rate": 1.3743374642516684e-06, "loss": 1.8648, "step": 28472 }, { "epoch": 5.42, "learning_rate": 1.374280266920877e-06, "loss": 1.3726, "step": 28473 }, { "epoch": 5.42, "learning_rate": 1.3742230695900858e-06, "loss": 1.7319, "step": 28474 }, { "epoch": 5.42, "learning_rate": 1.3741658722592946e-06, "loss": 1.07, "step": 28475 }, { "epoch": 5.42, "learning_rate": 1.3741086749285034e-06, "loss": 2.6478, "step": 28476 }, { "epoch": 5.42, "learning_rate": 1.3740514775977121e-06, "loss": 2.2902, "step": 28477 }, { "epoch": 5.42, "learning_rate": 1.373994280266921e-06, "loss": 1.4247, "step": 28478 }, { "epoch": 5.42, "learning_rate": 1.3739370829361297e-06, "loss": 2.6315, "step": 28479 }, { "epoch": 5.42, "learning_rate": 1.3738798856053383e-06, "loss": 1.5268, "step": 28480 }, { "epoch": 5.42, "learning_rate": 1.3738226882745471e-06, "loss": 1.9449, "step": 28481 }, { "epoch": 5.43, "learning_rate": 1.3737654909437561e-06, "loss": 3.2186, "step": 28482 }, { "epoch": 5.43, "learning_rate": 1.3737082936129647e-06, "loss": 1.9131, "step": 28483 }, { "epoch": 5.43, "learning_rate": 1.3736510962821735e-06, "loss": 2.0133, "step": 28484 }, { "epoch": 5.43, "learning_rate": 1.3735938989513823e-06, "loss": 1.7245, "step": 28485 }, { "epoch": 5.43, "learning_rate": 1.373536701620591e-06, "loss": 3.1415, "step": 28486 }, { "epoch": 5.43, "learning_rate": 1.3734795042897999e-06, "loss": 2.1658, "step": 28487 }, { "epoch": 5.43, "learning_rate": 1.3734223069590087e-06, "loss": 2.766, "step": 28488 }, { "epoch": 5.43, "learning_rate": 1.3733651096282175e-06, "loss": 2.8386, "step": 28489 }, { "epoch": 5.43, "learning_rate": 1.373307912297426e-06, "loss": 2.6481, "step": 28490 }, { "epoch": 5.43, "learning_rate": 1.3732507149666348e-06, "loss": 1.4226, "step": 28491 }, { "epoch": 5.43, "learning_rate": 1.3731935176358436e-06, "loss": 1.976, "step": 28492 }, { "epoch": 5.43, "learning_rate": 1.3731363203050526e-06, "loss": 2.1149, "step": 28493 }, { "epoch": 5.43, "learning_rate": 1.3730791229742612e-06, "loss": 2.8175, "step": 28494 }, { "epoch": 5.43, "learning_rate": 1.37302192564347e-06, "loss": 2.6884, "step": 28495 }, { "epoch": 5.43, "learning_rate": 1.3729647283126788e-06, "loss": 1.3437, "step": 28496 }, { "epoch": 5.43, "learning_rate": 1.3729075309818874e-06, "loss": 2.1156, "step": 28497 }, { "epoch": 5.43, "learning_rate": 1.3728503336510964e-06, "loss": 1.2899, "step": 28498 }, { "epoch": 5.43, "learning_rate": 1.3727931363203052e-06, "loss": 2.3285, "step": 28499 }, { "epoch": 5.43, "learning_rate": 1.372735938989514e-06, "loss": 3.1469, "step": 28500 }, { "epoch": 5.43, "learning_rate": 1.3726787416587226e-06, "loss": 1.2193, "step": 28501 }, { "epoch": 5.43, "learning_rate": 1.3726215443279314e-06, "loss": 2.5912, "step": 28502 }, { "epoch": 5.43, "learning_rate": 1.3725643469971401e-06, "loss": 2.1133, "step": 28503 }, { "epoch": 5.43, "learning_rate": 1.372507149666349e-06, "loss": 3.1086, "step": 28504 }, { "epoch": 5.43, "learning_rate": 1.3724499523355577e-06, "loss": 1.6774, "step": 28505 }, { "epoch": 5.43, "learning_rate": 1.3723927550047665e-06, "loss": 1.9585, "step": 28506 }, { "epoch": 5.43, "learning_rate": 1.3723355576739751e-06, "loss": 1.3077, "step": 28507 }, { "epoch": 5.43, "learning_rate": 1.372278360343184e-06, "loss": 2.0368, "step": 28508 }, { "epoch": 5.43, "learning_rate": 1.372221163012393e-06, "loss": 1.8796, "step": 28509 }, { "epoch": 5.43, "learning_rate": 1.3721639656816017e-06, "loss": 1.762, "step": 28510 }, { "epoch": 5.43, "learning_rate": 1.3721067683508103e-06, "loss": 1.8005, "step": 28511 }, { "epoch": 5.43, "learning_rate": 1.372049571020019e-06, "loss": 1.2166, "step": 28512 }, { "epoch": 5.43, "learning_rate": 1.3719923736892279e-06, "loss": 2.3211, "step": 28513 }, { "epoch": 5.43, "learning_rate": 1.3719351763584367e-06, "loss": 3.3631, "step": 28514 }, { "epoch": 5.43, "learning_rate": 1.3718779790276455e-06, "loss": 2.3442, "step": 28515 }, { "epoch": 5.43, "learning_rate": 1.3718207816968543e-06, "loss": 1.836, "step": 28516 }, { "epoch": 5.43, "learning_rate": 1.371763584366063e-06, "loss": 1.5101, "step": 28517 }, { "epoch": 5.43, "learning_rate": 1.3717063870352716e-06, "loss": 2.117, "step": 28518 }, { "epoch": 5.43, "learning_rate": 1.3716491897044804e-06, "loss": 1.0188, "step": 28519 }, { "epoch": 5.43, "learning_rate": 1.3715919923736892e-06, "loss": 2.843, "step": 28520 }, { "epoch": 5.43, "learning_rate": 1.371534795042898e-06, "loss": 2.042, "step": 28521 }, { "epoch": 5.43, "learning_rate": 1.3714775977121068e-06, "loss": 1.244, "step": 28522 }, { "epoch": 5.43, "learning_rate": 1.3714204003813156e-06, "loss": 2.1878, "step": 28523 }, { "epoch": 5.43, "learning_rate": 1.3713632030505242e-06, "loss": 0.8043, "step": 28524 }, { "epoch": 5.43, "learning_rate": 1.371306005719733e-06, "loss": 1.8715, "step": 28525 }, { "epoch": 5.43, "learning_rate": 1.371248808388942e-06, "loss": 2.1672, "step": 28526 }, { "epoch": 5.43, "learning_rate": 1.3711916110581508e-06, "loss": 1.2295, "step": 28527 }, { "epoch": 5.43, "learning_rate": 1.3711344137273594e-06, "loss": 1.3155, "step": 28528 }, { "epoch": 5.43, "learning_rate": 1.3710772163965682e-06, "loss": 1.2869, "step": 28529 }, { "epoch": 5.43, "learning_rate": 1.371020019065777e-06, "loss": 2.2371, "step": 28530 }, { "epoch": 5.43, "learning_rate": 1.3709628217349857e-06, "loss": 2.1932, "step": 28531 }, { "epoch": 5.43, "learning_rate": 1.3709056244041945e-06, "loss": 2.0186, "step": 28532 }, { "epoch": 5.43, "learning_rate": 1.3708484270734033e-06, "loss": 1.2084, "step": 28533 }, { "epoch": 5.44, "learning_rate": 1.3707912297426121e-06, "loss": 2.892, "step": 28534 }, { "epoch": 5.44, "learning_rate": 1.3707340324118207e-06, "loss": 3.1668, "step": 28535 }, { "epoch": 5.44, "learning_rate": 1.3706768350810295e-06, "loss": 2.6806, "step": 28536 }, { "epoch": 5.44, "learning_rate": 1.3706196377502385e-06, "loss": 1.4488, "step": 28537 }, { "epoch": 5.44, "learning_rate": 1.370562440419447e-06, "loss": 1.9621, "step": 28538 }, { "epoch": 5.44, "learning_rate": 1.3705052430886559e-06, "loss": 2.2046, "step": 28539 }, { "epoch": 5.44, "learning_rate": 1.3704480457578647e-06, "loss": 1.9116, "step": 28540 }, { "epoch": 5.44, "learning_rate": 1.3703908484270735e-06, "loss": 1.4173, "step": 28541 }, { "epoch": 5.44, "learning_rate": 1.3703336510962823e-06, "loss": 1.518, "step": 28542 }, { "epoch": 5.44, "learning_rate": 1.370276453765491e-06, "loss": 1.4783, "step": 28543 }, { "epoch": 5.44, "learning_rate": 1.3702192564346998e-06, "loss": 1.1977, "step": 28544 }, { "epoch": 5.44, "learning_rate": 1.3701620591039084e-06, "loss": 1.5403, "step": 28545 }, { "epoch": 5.44, "learning_rate": 1.3701048617731172e-06, "loss": 1.83, "step": 28546 }, { "epoch": 5.44, "learning_rate": 1.370047664442326e-06, "loss": 2.2418, "step": 28547 }, { "epoch": 5.44, "learning_rate": 1.3699904671115348e-06, "loss": 2.5686, "step": 28548 }, { "epoch": 5.44, "learning_rate": 1.3699332697807436e-06, "loss": 1.5422, "step": 28549 }, { "epoch": 5.44, "learning_rate": 1.3698760724499524e-06, "loss": 1.6445, "step": 28550 }, { "epoch": 5.44, "learning_rate": 1.3698188751191612e-06, "loss": 1.3652, "step": 28551 }, { "epoch": 5.44, "learning_rate": 1.3697616777883698e-06, "loss": 0.649, "step": 28552 }, { "epoch": 5.44, "learning_rate": 1.3697044804575788e-06, "loss": 2.4935, "step": 28553 }, { "epoch": 5.44, "learning_rate": 1.3696472831267876e-06, "loss": 1.0073, "step": 28554 }, { "epoch": 5.44, "learning_rate": 1.3695900857959962e-06, "loss": 1.5204, "step": 28555 }, { "epoch": 5.44, "learning_rate": 1.369532888465205e-06, "loss": 2.0134, "step": 28556 }, { "epoch": 5.44, "learning_rate": 1.3694756911344137e-06, "loss": 1.1746, "step": 28557 }, { "epoch": 5.44, "learning_rate": 1.3694184938036225e-06, "loss": 2.9736, "step": 28558 }, { "epoch": 5.44, "learning_rate": 1.3693612964728313e-06, "loss": 2.0411, "step": 28559 }, { "epoch": 5.44, "learning_rate": 1.3693040991420401e-06, "loss": 1.4526, "step": 28560 }, { "epoch": 5.44, "learning_rate": 1.369246901811249e-06, "loss": 2.4889, "step": 28561 }, { "epoch": 5.44, "learning_rate": 1.3691897044804575e-06, "loss": 2.2028, "step": 28562 }, { "epoch": 5.44, "learning_rate": 1.3691325071496663e-06, "loss": 1.9507, "step": 28563 }, { "epoch": 5.44, "learning_rate": 1.369075309818875e-06, "loss": 0.9837, "step": 28564 }, { "epoch": 5.44, "learning_rate": 1.369018112488084e-06, "loss": 1.6808, "step": 28565 }, { "epoch": 5.44, "learning_rate": 1.3689609151572927e-06, "loss": 2.7488, "step": 28566 }, { "epoch": 5.44, "learning_rate": 1.3689037178265015e-06, "loss": 1.642, "step": 28567 }, { "epoch": 5.44, "learning_rate": 1.3688465204957103e-06, "loss": 0.9125, "step": 28568 }, { "epoch": 5.44, "learning_rate": 1.3687893231649188e-06, "loss": 1.4125, "step": 28569 }, { "epoch": 5.44, "learning_rate": 1.3687321258341279e-06, "loss": 1.4667, "step": 28570 }, { "epoch": 5.44, "learning_rate": 1.3686749285033366e-06, "loss": 2.1745, "step": 28571 }, { "epoch": 5.44, "learning_rate": 1.3686177311725452e-06, "loss": 2.1043, "step": 28572 }, { "epoch": 5.44, "learning_rate": 1.368560533841754e-06, "loss": 1.4514, "step": 28573 }, { "epoch": 5.44, "learning_rate": 1.3685033365109628e-06, "loss": 0.6198, "step": 28574 }, { "epoch": 5.44, "learning_rate": 1.3684461391801716e-06, "loss": 1.4409, "step": 28575 }, { "epoch": 5.44, "learning_rate": 1.3683889418493804e-06, "loss": 2.6374, "step": 28576 }, { "epoch": 5.44, "learning_rate": 1.3683317445185892e-06, "loss": 2.474, "step": 28577 }, { "epoch": 5.44, "learning_rate": 1.368274547187798e-06, "loss": 2.0268, "step": 28578 }, { "epoch": 5.44, "learning_rate": 1.3682173498570066e-06, "loss": 1.2607, "step": 28579 }, { "epoch": 5.44, "learning_rate": 1.3681601525262154e-06, "loss": 1.9447, "step": 28580 }, { "epoch": 5.44, "learning_rate": 1.3681029551954244e-06, "loss": 1.8326, "step": 28581 }, { "epoch": 5.44, "learning_rate": 1.3680457578646332e-06, "loss": 2.3571, "step": 28582 }, { "epoch": 5.44, "learning_rate": 1.3679885605338417e-06, "loss": 0.9594, "step": 28583 }, { "epoch": 5.44, "learning_rate": 1.3679313632030505e-06, "loss": 0.6904, "step": 28584 }, { "epoch": 5.44, "learning_rate": 1.3678741658722593e-06, "loss": 1.3419, "step": 28585 }, { "epoch": 5.44, "learning_rate": 1.3678169685414681e-06, "loss": 2.5055, "step": 28586 }, { "epoch": 5.45, "learning_rate": 1.367759771210677e-06, "loss": 1.2014, "step": 28587 }, { "epoch": 5.45, "learning_rate": 1.3677025738798857e-06, "loss": 2.1647, "step": 28588 }, { "epoch": 5.45, "learning_rate": 1.3676453765490943e-06, "loss": 2.2226, "step": 28589 }, { "epoch": 5.45, "learning_rate": 1.367588179218303e-06, "loss": 1.9121, "step": 28590 }, { "epoch": 5.45, "learning_rate": 1.3675309818875119e-06, "loss": 1.2325, "step": 28591 }, { "epoch": 5.45, "learning_rate": 1.3674737845567207e-06, "loss": 0.6851, "step": 28592 }, { "epoch": 5.45, "learning_rate": 1.3674165872259295e-06, "loss": 1.3278, "step": 28593 }, { "epoch": 5.45, "learning_rate": 1.3673593898951383e-06, "loss": 3.0034, "step": 28594 }, { "epoch": 5.45, "learning_rate": 1.367302192564347e-06, "loss": 1.6597, "step": 28595 }, { "epoch": 5.45, "learning_rate": 1.3672449952335556e-06, "loss": 2.7062, "step": 28596 }, { "epoch": 5.45, "learning_rate": 1.3671877979027646e-06, "loss": 2.1557, "step": 28597 }, { "epoch": 5.45, "learning_rate": 1.3671306005719734e-06, "loss": 1.7385, "step": 28598 }, { "epoch": 5.45, "learning_rate": 1.3670734032411822e-06, "loss": 2.3359, "step": 28599 }, { "epoch": 5.45, "learning_rate": 1.3670162059103908e-06, "loss": 1.3697, "step": 28600 }, { "epoch": 5.45, "learning_rate": 1.3669590085795996e-06, "loss": 3.2235, "step": 28601 }, { "epoch": 5.45, "learning_rate": 1.3669018112488084e-06, "loss": 2.0104, "step": 28602 }, { "epoch": 5.45, "learning_rate": 1.3668446139180172e-06, "loss": 0.433, "step": 28603 }, { "epoch": 5.45, "learning_rate": 1.366787416587226e-06, "loss": 2.1075, "step": 28604 }, { "epoch": 5.45, "learning_rate": 1.3667302192564348e-06, "loss": 1.1842, "step": 28605 }, { "epoch": 5.45, "learning_rate": 1.3666730219256436e-06, "loss": 1.6808, "step": 28606 }, { "epoch": 5.45, "learning_rate": 1.3666158245948522e-06, "loss": 2.3851, "step": 28607 }, { "epoch": 5.45, "learning_rate": 1.366558627264061e-06, "loss": 1.1638, "step": 28608 }, { "epoch": 5.45, "learning_rate": 1.36650142993327e-06, "loss": 1.3205, "step": 28609 }, { "epoch": 5.45, "learning_rate": 1.3664442326024785e-06, "loss": 1.2163, "step": 28610 }, { "epoch": 5.45, "learning_rate": 1.3663870352716873e-06, "loss": 0.8598, "step": 28611 }, { "epoch": 5.45, "learning_rate": 1.3663298379408961e-06, "loss": 2.1438, "step": 28612 }, { "epoch": 5.45, "learning_rate": 1.3662726406101047e-06, "loss": 0.8654, "step": 28613 }, { "epoch": 5.45, "learning_rate": 1.3662154432793137e-06, "loss": 0.5779, "step": 28614 }, { "epoch": 5.45, "learning_rate": 1.3661582459485225e-06, "loss": 2.4078, "step": 28615 }, { "epoch": 5.45, "learning_rate": 1.3661010486177313e-06, "loss": 3.3595, "step": 28616 }, { "epoch": 5.45, "learning_rate": 1.3660438512869399e-06, "loss": 1.6768, "step": 28617 }, { "epoch": 5.45, "learning_rate": 1.3659866539561487e-06, "loss": 1.5368, "step": 28618 }, { "epoch": 5.45, "learning_rate": 1.3659294566253575e-06, "loss": 2.0561, "step": 28619 }, { "epoch": 5.45, "learning_rate": 1.3658722592945663e-06, "loss": 1.097, "step": 28620 }, { "epoch": 5.45, "learning_rate": 1.365815061963775e-06, "loss": 1.305, "step": 28621 }, { "epoch": 5.45, "learning_rate": 1.3657578646329839e-06, "loss": 1.5319, "step": 28622 }, { "epoch": 5.45, "learning_rate": 1.3657006673021927e-06, "loss": 1.6234, "step": 28623 }, { "epoch": 5.45, "learning_rate": 1.3656434699714012e-06, "loss": 1.2694, "step": 28624 }, { "epoch": 5.45, "learning_rate": 1.3655862726406102e-06, "loss": 1.0986, "step": 28625 }, { "epoch": 5.45, "learning_rate": 1.365529075309819e-06, "loss": 2.3193, "step": 28626 }, { "epoch": 5.45, "learning_rate": 1.3654718779790276e-06, "loss": 1.4547, "step": 28627 }, { "epoch": 5.45, "learning_rate": 1.3654146806482364e-06, "loss": 1.9261, "step": 28628 }, { "epoch": 5.45, "learning_rate": 1.3653574833174452e-06, "loss": 3.0156, "step": 28629 }, { "epoch": 5.45, "learning_rate": 1.365300285986654e-06, "loss": 0.9608, "step": 28630 }, { "epoch": 5.45, "learning_rate": 1.3652430886558628e-06, "loss": 2.7862, "step": 28631 }, { "epoch": 5.45, "learning_rate": 1.3651858913250716e-06, "loss": 2.7524, "step": 28632 }, { "epoch": 5.45, "learning_rate": 1.3651286939942804e-06, "loss": 1.4375, "step": 28633 }, { "epoch": 5.45, "learning_rate": 1.365071496663489e-06, "loss": 1.6478, "step": 28634 }, { "epoch": 5.45, "learning_rate": 1.3650142993326978e-06, "loss": 0.7762, "step": 28635 }, { "epoch": 5.45, "learning_rate": 1.3649571020019065e-06, "loss": 0.6842, "step": 28636 }, { "epoch": 5.45, "learning_rate": 1.3648999046711153e-06, "loss": 2.0854, "step": 28637 }, { "epoch": 5.45, "learning_rate": 1.3648427073403241e-06, "loss": 1.9377, "step": 28638 }, { "epoch": 5.46, "learning_rate": 1.364785510009533e-06, "loss": 3.1314, "step": 28639 }, { "epoch": 5.46, "learning_rate": 1.3647283126787417e-06, "loss": 2.8953, "step": 28640 }, { "epoch": 5.46, "learning_rate": 1.3646711153479505e-06, "loss": 3.2553, "step": 28641 }, { "epoch": 5.46, "learning_rate": 1.3646139180171593e-06, "loss": 1.9368, "step": 28642 }, { "epoch": 5.46, "learning_rate": 1.3645567206863681e-06, "loss": 1.7881, "step": 28643 }, { "epoch": 5.46, "learning_rate": 1.3644995233555767e-06, "loss": 2.5524, "step": 28644 }, { "epoch": 5.46, "learning_rate": 1.3644423260247855e-06, "loss": 1.5794, "step": 28645 }, { "epoch": 5.46, "learning_rate": 1.3643851286939943e-06, "loss": 1.9668, "step": 28646 }, { "epoch": 5.46, "learning_rate": 1.364327931363203e-06, "loss": 1.4367, "step": 28647 }, { "epoch": 5.46, "learning_rate": 1.3642707340324119e-06, "loss": 2.5206, "step": 28648 }, { "epoch": 5.46, "learning_rate": 1.3642135367016207e-06, "loss": 1.879, "step": 28649 }, { "epoch": 5.46, "learning_rate": 1.3641563393708295e-06, "loss": 1.5104, "step": 28650 }, { "epoch": 5.46, "learning_rate": 1.364099142040038e-06, "loss": 2.239, "step": 28651 }, { "epoch": 5.46, "learning_rate": 1.3640419447092468e-06, "loss": 2.1398, "step": 28652 }, { "epoch": 5.46, "learning_rate": 1.3639847473784558e-06, "loss": 1.894, "step": 28653 }, { "epoch": 5.46, "learning_rate": 1.3639275500476644e-06, "loss": 3.1292, "step": 28654 }, { "epoch": 5.46, "learning_rate": 1.3638703527168732e-06, "loss": 2.0186, "step": 28655 }, { "epoch": 5.46, "learning_rate": 1.363813155386082e-06, "loss": 2.1586, "step": 28656 }, { "epoch": 5.46, "learning_rate": 1.3637559580552908e-06, "loss": 2.5467, "step": 28657 }, { "epoch": 5.46, "learning_rate": 1.3636987607244996e-06, "loss": 1.2284, "step": 28658 }, { "epoch": 5.46, "learning_rate": 1.3636415633937084e-06, "loss": 1.2512, "step": 28659 }, { "epoch": 5.46, "learning_rate": 1.3635843660629172e-06, "loss": 1.4037, "step": 28660 }, { "epoch": 5.46, "learning_rate": 1.3635271687321258e-06, "loss": 0.72, "step": 28661 }, { "epoch": 5.46, "learning_rate": 1.3634699714013346e-06, "loss": 2.0458, "step": 28662 }, { "epoch": 5.46, "learning_rate": 1.3634127740705433e-06, "loss": 1.8245, "step": 28663 }, { "epoch": 5.46, "learning_rate": 1.3633555767397524e-06, "loss": 1.5957, "step": 28664 }, { "epoch": 5.46, "learning_rate": 1.363298379408961e-06, "loss": 2.4708, "step": 28665 }, { "epoch": 5.46, "learning_rate": 1.3632411820781697e-06, "loss": 2.1254, "step": 28666 }, { "epoch": 5.46, "learning_rate": 1.3631839847473785e-06, "loss": 2.8607, "step": 28667 }, { "epoch": 5.46, "learning_rate": 1.363126787416587e-06, "loss": 1.989, "step": 28668 }, { "epoch": 5.46, "learning_rate": 1.3630695900857961e-06, "loss": 2.4867, "step": 28669 }, { "epoch": 5.46, "learning_rate": 1.363012392755005e-06, "loss": 2.5025, "step": 28670 }, { "epoch": 5.46, "learning_rate": 1.3629551954242137e-06, "loss": 2.2921, "step": 28671 }, { "epoch": 5.46, "learning_rate": 1.3628979980934223e-06, "loss": 1.6307, "step": 28672 }, { "epoch": 5.46, "learning_rate": 1.362840800762631e-06, "loss": 1.625, "step": 28673 }, { "epoch": 5.46, "learning_rate": 1.3627836034318399e-06, "loss": 3.0559, "step": 28674 }, { "epoch": 5.46, "learning_rate": 1.3627264061010487e-06, "loss": 1.4988, "step": 28675 }, { "epoch": 5.46, "learning_rate": 1.3626692087702575e-06, "loss": 2.0144, "step": 28676 }, { "epoch": 5.46, "learning_rate": 1.3626120114394662e-06, "loss": 1.1603, "step": 28677 }, { "epoch": 5.46, "learning_rate": 1.3625548141086748e-06, "loss": 1.6429, "step": 28678 }, { "epoch": 5.46, "learning_rate": 1.3624976167778836e-06, "loss": 2.6589, "step": 28679 }, { "epoch": 5.46, "learning_rate": 1.3624404194470924e-06, "loss": 1.1066, "step": 28680 }, { "epoch": 5.46, "learning_rate": 1.3623832221163014e-06, "loss": 2.0274, "step": 28681 }, { "epoch": 5.46, "learning_rate": 1.36232602478551e-06, "loss": 2.1349, "step": 28682 }, { "epoch": 5.46, "learning_rate": 1.3622688274547188e-06, "loss": 2.4435, "step": 28683 }, { "epoch": 5.46, "learning_rate": 1.3622116301239276e-06, "loss": 2.9959, "step": 28684 }, { "epoch": 5.46, "learning_rate": 1.3621544327931364e-06, "loss": 1.6883, "step": 28685 }, { "epoch": 5.46, "learning_rate": 1.3620972354623452e-06, "loss": 2.3469, "step": 28686 }, { "epoch": 5.46, "learning_rate": 1.362040038131554e-06, "loss": 1.6749, "step": 28687 }, { "epoch": 5.46, "learning_rate": 1.3619828408007628e-06, "loss": 2.3765, "step": 28688 }, { "epoch": 5.46, "learning_rate": 1.3619256434699714e-06, "loss": 0.944, "step": 28689 }, { "epoch": 5.46, "learning_rate": 1.3618684461391801e-06, "loss": 1.3175, "step": 28690 }, { "epoch": 5.46, "learning_rate": 1.361811248808389e-06, "loss": 4.1373, "step": 28691 }, { "epoch": 5.47, "learning_rate": 1.3617540514775977e-06, "loss": 2.5708, "step": 28692 }, { "epoch": 5.47, "learning_rate": 1.3616968541468065e-06, "loss": 1.6517, "step": 28693 }, { "epoch": 5.47, "learning_rate": 1.3616396568160153e-06, "loss": 2.2325, "step": 28694 }, { "epoch": 5.47, "learning_rate": 1.3615824594852241e-06, "loss": 1.7987, "step": 28695 }, { "epoch": 5.47, "learning_rate": 1.3615252621544327e-06, "loss": 2.4505, "step": 28696 }, { "epoch": 5.47, "learning_rate": 1.3614680648236417e-06, "loss": 2.4889, "step": 28697 }, { "epoch": 5.47, "learning_rate": 1.3614108674928505e-06, "loss": 1.8817, "step": 28698 }, { "epoch": 5.47, "learning_rate": 1.361353670162059e-06, "loss": 2.1929, "step": 28699 }, { "epoch": 5.47, "learning_rate": 1.3612964728312679e-06, "loss": 2.6529, "step": 28700 }, { "epoch": 5.47, "learning_rate": 1.3612392755004767e-06, "loss": 3.1783, "step": 28701 }, { "epoch": 5.47, "learning_rate": 1.3611820781696855e-06, "loss": 1.8817, "step": 28702 }, { "epoch": 5.47, "learning_rate": 1.3611248808388943e-06, "loss": 1.2387, "step": 28703 }, { "epoch": 5.47, "learning_rate": 1.361067683508103e-06, "loss": 2.4545, "step": 28704 }, { "epoch": 5.47, "learning_rate": 1.3610104861773118e-06, "loss": 2.7427, "step": 28705 }, { "epoch": 5.47, "learning_rate": 1.3609532888465204e-06, "loss": 2.5867, "step": 28706 }, { "epoch": 5.47, "learning_rate": 1.3608960915157292e-06, "loss": 2.8888, "step": 28707 }, { "epoch": 5.47, "learning_rate": 1.3608388941849382e-06, "loss": 2.1814, "step": 28708 }, { "epoch": 5.47, "learning_rate": 1.3607816968541468e-06, "loss": 1.9779, "step": 28709 }, { "epoch": 5.47, "learning_rate": 1.3607244995233556e-06, "loss": 1.4913, "step": 28710 }, { "epoch": 5.47, "learning_rate": 1.3606673021925644e-06, "loss": 1.0715, "step": 28711 }, { "epoch": 5.47, "learning_rate": 1.3606101048617732e-06, "loss": 2.4444, "step": 28712 }, { "epoch": 5.47, "learning_rate": 1.360552907530982e-06, "loss": 1.5952, "step": 28713 }, { "epoch": 5.47, "learning_rate": 1.3604957102001908e-06, "loss": 2.0985, "step": 28714 }, { "epoch": 5.47, "learning_rate": 1.3604385128693996e-06, "loss": 3.5933, "step": 28715 }, { "epoch": 5.47, "learning_rate": 1.3603813155386081e-06, "loss": 1.3474, "step": 28716 }, { "epoch": 5.47, "learning_rate": 1.360324118207817e-06, "loss": 1.6302, "step": 28717 }, { "epoch": 5.47, "learning_rate": 1.3602669208770257e-06, "loss": 1.6278, "step": 28718 }, { "epoch": 5.47, "learning_rate": 1.3602097235462345e-06, "loss": 1.6636, "step": 28719 }, { "epoch": 5.47, "learning_rate": 1.3601525262154433e-06, "loss": 3.007, "step": 28720 }, { "epoch": 5.47, "learning_rate": 1.3600953288846521e-06, "loss": 1.1205, "step": 28721 }, { "epoch": 5.47, "learning_rate": 1.360038131553861e-06, "loss": 1.7062, "step": 28722 }, { "epoch": 5.47, "learning_rate": 1.3599809342230695e-06, "loss": 1.9661, "step": 28723 }, { "epoch": 5.47, "learning_rate": 1.3599237368922783e-06, "loss": 1.8621, "step": 28724 }, { "epoch": 5.47, "learning_rate": 1.3598665395614873e-06, "loss": 2.4028, "step": 28725 }, { "epoch": 5.47, "learning_rate": 1.3598093422306959e-06, "loss": 2.4508, "step": 28726 }, { "epoch": 5.47, "learning_rate": 1.3597521448999047e-06, "loss": 3.8686, "step": 28727 }, { "epoch": 5.47, "learning_rate": 1.3596949475691135e-06, "loss": 1.7238, "step": 28728 }, { "epoch": 5.47, "learning_rate": 1.3596377502383223e-06, "loss": 1.6892, "step": 28729 }, { "epoch": 5.47, "learning_rate": 1.359580552907531e-06, "loss": 1.9806, "step": 28730 }, { "epoch": 5.47, "learning_rate": 1.3595233555767398e-06, "loss": 2.5767, "step": 28731 }, { "epoch": 5.47, "learning_rate": 1.3594661582459486e-06, "loss": 2.8891, "step": 28732 }, { "epoch": 5.47, "learning_rate": 1.3594089609151572e-06, "loss": 1.8659, "step": 28733 }, { "epoch": 5.47, "learning_rate": 1.359351763584366e-06, "loss": 1.0512, "step": 28734 }, { "epoch": 5.47, "learning_rate": 1.3592945662535748e-06, "loss": 3.4676, "step": 28735 }, { "epoch": 5.47, "learning_rate": 1.3592373689227838e-06, "loss": 0.4267, "step": 28736 }, { "epoch": 5.47, "learning_rate": 1.3591801715919924e-06, "loss": 1.9073, "step": 28737 }, { "epoch": 5.47, "learning_rate": 1.3591229742612012e-06, "loss": 3.2637, "step": 28738 }, { "epoch": 5.47, "learning_rate": 1.35906577693041e-06, "loss": 1.0031, "step": 28739 }, { "epoch": 5.47, "learning_rate": 1.3590085795996186e-06, "loss": 0.8281, "step": 28740 }, { "epoch": 5.47, "learning_rate": 1.3589513822688276e-06, "loss": 3.2206, "step": 28741 }, { "epoch": 5.47, "learning_rate": 1.3588941849380364e-06, "loss": 1.7891, "step": 28742 }, { "epoch": 5.47, "learning_rate": 1.358836987607245e-06, "loss": 2.5619, "step": 28743 }, { "epoch": 5.48, "learning_rate": 1.3587797902764537e-06, "loss": 1.7645, "step": 28744 }, { "epoch": 5.48, "learning_rate": 1.3587225929456625e-06, "loss": 1.0532, "step": 28745 }, { "epoch": 5.48, "learning_rate": 1.3586653956148713e-06, "loss": 0.6381, "step": 28746 }, { "epoch": 5.48, "learning_rate": 1.3586081982840801e-06, "loss": 3.0973, "step": 28747 }, { "epoch": 5.48, "learning_rate": 1.358551000953289e-06, "loss": 2.5571, "step": 28748 }, { "epoch": 5.48, "learning_rate": 1.3584938036224977e-06, "loss": 0.9298, "step": 28749 }, { "epoch": 5.48, "learning_rate": 1.3584366062917063e-06, "loss": 0.4087, "step": 28750 }, { "epoch": 5.48, "learning_rate": 1.358379408960915e-06, "loss": 1.0847, "step": 28751 }, { "epoch": 5.48, "learning_rate": 1.358322211630124e-06, "loss": 1.742, "step": 28752 }, { "epoch": 5.48, "learning_rate": 1.3582650142993329e-06, "loss": 2.491, "step": 28753 }, { "epoch": 5.48, "learning_rate": 1.3582078169685415e-06, "loss": 0.9158, "step": 28754 }, { "epoch": 5.48, "learning_rate": 1.3581506196377503e-06, "loss": 1.6889, "step": 28755 }, { "epoch": 5.48, "learning_rate": 1.358093422306959e-06, "loss": 2.2234, "step": 28756 }, { "epoch": 5.48, "learning_rate": 1.3580362249761678e-06, "loss": 1.9645, "step": 28757 }, { "epoch": 5.48, "learning_rate": 1.3579790276453766e-06, "loss": 1.4582, "step": 28758 }, { "epoch": 5.48, "learning_rate": 1.3579218303145854e-06, "loss": 1.0974, "step": 28759 }, { "epoch": 5.48, "learning_rate": 1.357864632983794e-06, "loss": 2.0365, "step": 28760 }, { "epoch": 5.48, "learning_rate": 1.3578074356530028e-06, "loss": 2.4328, "step": 28761 }, { "epoch": 5.48, "learning_rate": 1.3577502383222116e-06, "loss": 0.6261, "step": 28762 }, { "epoch": 5.48, "learning_rate": 1.3576930409914204e-06, "loss": 1.8867, "step": 28763 }, { "epoch": 5.48, "learning_rate": 1.3576358436606292e-06, "loss": 1.3798, "step": 28764 }, { "epoch": 5.48, "learning_rate": 1.357578646329838e-06, "loss": 2.3161, "step": 28765 }, { "epoch": 5.48, "learning_rate": 1.3575214489990468e-06, "loss": 1.0453, "step": 28766 }, { "epoch": 5.48, "learning_rate": 1.3574642516682554e-06, "loss": 2.9019, "step": 28767 }, { "epoch": 5.48, "learning_rate": 1.3574070543374642e-06, "loss": 0.6996, "step": 28768 }, { "epoch": 5.48, "learning_rate": 1.3573498570066732e-06, "loss": 1.5601, "step": 28769 }, { "epoch": 5.48, "learning_rate": 1.357292659675882e-06, "loss": 2.2847, "step": 28770 }, { "epoch": 5.48, "learning_rate": 1.3572354623450905e-06, "loss": 2.5512, "step": 28771 }, { "epoch": 5.48, "learning_rate": 1.3571782650142993e-06, "loss": 0.6833, "step": 28772 }, { "epoch": 5.48, "learning_rate": 1.3571210676835081e-06, "loss": 1.4909, "step": 28773 }, { "epoch": 5.48, "learning_rate": 1.357063870352717e-06, "loss": 0.6297, "step": 28774 }, { "epoch": 5.48, "learning_rate": 1.3570066730219257e-06, "loss": 2.078, "step": 28775 }, { "epoch": 5.48, "learning_rate": 1.3569494756911345e-06, "loss": 1.3887, "step": 28776 }, { "epoch": 5.48, "learning_rate": 1.3568922783603433e-06, "loss": 1.2554, "step": 28777 }, { "epoch": 5.48, "learning_rate": 1.3568350810295519e-06, "loss": 3.0029, "step": 28778 }, { "epoch": 5.48, "learning_rate": 1.3567778836987607e-06, "loss": 1.2319, "step": 28779 }, { "epoch": 5.48, "learning_rate": 1.3567206863679697e-06, "loss": 1.5695, "step": 28780 }, { "epoch": 5.48, "learning_rate": 1.3566634890371783e-06, "loss": 1.3862, "step": 28781 }, { "epoch": 5.48, "learning_rate": 1.356606291706387e-06, "loss": 1.7713, "step": 28782 }, { "epoch": 5.48, "learning_rate": 1.3565490943755959e-06, "loss": 2.1769, "step": 28783 }, { "epoch": 5.48, "learning_rate": 1.3564918970448044e-06, "loss": 1.0126, "step": 28784 }, { "epoch": 5.48, "learning_rate": 1.3564346997140134e-06, "loss": 1.4215, "step": 28785 }, { "epoch": 5.48, "learning_rate": 1.3563775023832222e-06, "loss": 2.6921, "step": 28786 }, { "epoch": 5.48, "learning_rate": 1.356320305052431e-06, "loss": 1.7898, "step": 28787 }, { "epoch": 5.48, "learning_rate": 1.3562631077216396e-06, "loss": 1.1475, "step": 28788 }, { "epoch": 5.48, "learning_rate": 1.3562059103908484e-06, "loss": 1.6181, "step": 28789 }, { "epoch": 5.48, "learning_rate": 1.3561487130600572e-06, "loss": 2.0683, "step": 28790 }, { "epoch": 5.48, "learning_rate": 1.356091515729266e-06, "loss": 2.3771, "step": 28791 }, { "epoch": 5.48, "learning_rate": 1.3560343183984748e-06, "loss": 1.039, "step": 28792 }, { "epoch": 5.48, "learning_rate": 1.3559771210676836e-06, "loss": 2.1937, "step": 28793 }, { "epoch": 5.48, "learning_rate": 1.3559199237368924e-06, "loss": 2.4625, "step": 28794 }, { "epoch": 5.48, "learning_rate": 1.355862726406101e-06, "loss": 2.3253, "step": 28795 }, { "epoch": 5.48, "learning_rate": 1.35580552907531e-06, "loss": 1.4511, "step": 28796 }, { "epoch": 5.49, "learning_rate": 1.3557483317445188e-06, "loss": 1.889, "step": 28797 }, { "epoch": 5.49, "learning_rate": 1.3556911344137273e-06, "loss": 1.9884, "step": 28798 }, { "epoch": 5.49, "learning_rate": 1.3556339370829361e-06, "loss": 1.7836, "step": 28799 }, { "epoch": 5.49, "learning_rate": 1.355576739752145e-06, "loss": 2.3417, "step": 28800 }, { "epoch": 5.49, "learning_rate": 1.3555195424213537e-06, "loss": 2.0301, "step": 28801 }, { "epoch": 5.49, "learning_rate": 1.3554623450905625e-06, "loss": 2.982, "step": 28802 }, { "epoch": 5.49, "learning_rate": 1.3554051477597713e-06, "loss": 1.1756, "step": 28803 }, { "epoch": 5.49, "learning_rate": 1.35534795042898e-06, "loss": 1.6665, "step": 28804 }, { "epoch": 5.49, "learning_rate": 1.3552907530981887e-06, "loss": 2.3791, "step": 28805 }, { "epoch": 5.49, "learning_rate": 1.3552335557673975e-06, "loss": 1.1833, "step": 28806 }, { "epoch": 5.49, "learning_rate": 1.3551763584366063e-06, "loss": 2.3584, "step": 28807 }, { "epoch": 5.49, "learning_rate": 1.355119161105815e-06, "loss": 2.4043, "step": 28808 }, { "epoch": 5.49, "learning_rate": 1.3550619637750239e-06, "loss": 2.8521, "step": 28809 }, { "epoch": 5.49, "learning_rate": 1.3550047664442327e-06, "loss": 1.5684, "step": 28810 }, { "epoch": 5.49, "learning_rate": 1.3549475691134414e-06, "loss": 1.0216, "step": 28811 }, { "epoch": 5.49, "learning_rate": 1.35489037178265e-06, "loss": 1.6066, "step": 28812 }, { "epoch": 5.49, "learning_rate": 1.354833174451859e-06, "loss": 1.6794, "step": 28813 }, { "epoch": 5.49, "learning_rate": 1.3547759771210678e-06, "loss": 1.9061, "step": 28814 }, { "epoch": 5.49, "learning_rate": 1.3547187797902764e-06, "loss": 3.021, "step": 28815 }, { "epoch": 5.49, "learning_rate": 1.3546615824594852e-06, "loss": 1.5512, "step": 28816 }, { "epoch": 5.49, "learning_rate": 1.354604385128694e-06, "loss": 2.4215, "step": 28817 }, { "epoch": 5.49, "learning_rate": 1.3545471877979028e-06, "loss": 1.2397, "step": 28818 }, { "epoch": 5.49, "learning_rate": 1.3544899904671116e-06, "loss": 1.4662, "step": 28819 }, { "epoch": 5.49, "learning_rate": 1.3544327931363204e-06, "loss": 3.9684, "step": 28820 }, { "epoch": 5.49, "learning_rate": 1.3543755958055292e-06, "loss": 1.879, "step": 28821 }, { "epoch": 5.49, "learning_rate": 1.3543183984747378e-06, "loss": 1.1348, "step": 28822 }, { "epoch": 5.49, "learning_rate": 1.3542612011439465e-06, "loss": 0.7462, "step": 28823 }, { "epoch": 5.49, "learning_rate": 1.3542040038131556e-06, "loss": 2.5095, "step": 28824 }, { "epoch": 5.49, "learning_rate": 1.3541468064823641e-06, "loss": 2.5981, "step": 28825 }, { "epoch": 5.49, "learning_rate": 1.354089609151573e-06, "loss": 2.5878, "step": 28826 }, { "epoch": 5.49, "learning_rate": 1.3540324118207817e-06, "loss": 1.9212, "step": 28827 }, { "epoch": 5.49, "learning_rate": 1.3539752144899905e-06, "loss": 1.2691, "step": 28828 }, { "epoch": 5.49, "learning_rate": 1.3539180171591993e-06, "loss": 3.2878, "step": 28829 }, { "epoch": 5.49, "learning_rate": 1.353860819828408e-06, "loss": 2.5727, "step": 28830 }, { "epoch": 5.49, "learning_rate": 1.353803622497617e-06, "loss": 2.7919, "step": 28831 }, { "epoch": 5.49, "learning_rate": 1.3537464251668255e-06, "loss": 3.3183, "step": 28832 }, { "epoch": 5.49, "learning_rate": 1.3536892278360343e-06, "loss": 2.3517, "step": 28833 }, { "epoch": 5.49, "learning_rate": 1.353632030505243e-06, "loss": 3.1588, "step": 28834 }, { "epoch": 5.49, "learning_rate": 1.3535748331744519e-06, "loss": 2.173, "step": 28835 }, { "epoch": 5.49, "learning_rate": 1.3535176358436607e-06, "loss": 2.5751, "step": 28836 }, { "epoch": 5.49, "learning_rate": 1.3534604385128694e-06, "loss": 1.4813, "step": 28837 }, { "epoch": 5.49, "learning_rate": 1.3534032411820782e-06, "loss": 1.9594, "step": 28838 }, { "epoch": 5.49, "learning_rate": 1.3533460438512868e-06, "loss": 2.7918, "step": 28839 }, { "epoch": 5.49, "learning_rate": 1.3532888465204958e-06, "loss": 0.5169, "step": 28840 }, { "epoch": 5.49, "learning_rate": 1.3532316491897046e-06, "loss": 1.6065, "step": 28841 }, { "epoch": 5.49, "learning_rate": 1.3531744518589134e-06, "loss": 2.4914, "step": 28842 }, { "epoch": 5.49, "learning_rate": 1.353117254528122e-06, "loss": 2.9234, "step": 28843 }, { "epoch": 5.49, "learning_rate": 1.3530600571973308e-06, "loss": 1.6327, "step": 28844 }, { "epoch": 5.49, "learning_rate": 1.3530028598665396e-06, "loss": 2.5746, "step": 28845 }, { "epoch": 5.49, "learning_rate": 1.3529456625357484e-06, "loss": 2.9004, "step": 28846 }, { "epoch": 5.49, "learning_rate": 1.3528884652049572e-06, "loss": 2.1934, "step": 28847 }, { "epoch": 5.49, "learning_rate": 1.352831267874166e-06, "loss": 2.1934, "step": 28848 }, { "epoch": 5.5, "learning_rate": 1.3527740705433746e-06, "loss": 2.854, "step": 28849 }, { "epoch": 5.5, "learning_rate": 1.3527168732125833e-06, "loss": 1.691, "step": 28850 }, { "epoch": 5.5, "learning_rate": 1.3526596758817921e-06, "loss": 1.3897, "step": 28851 }, { "epoch": 5.5, "learning_rate": 1.3526024785510011e-06, "loss": 2.8883, "step": 28852 }, { "epoch": 5.5, "learning_rate": 1.3525452812202097e-06, "loss": 1.1792, "step": 28853 }, { "epoch": 5.5, "learning_rate": 1.3524880838894185e-06, "loss": 2.1735, "step": 28854 }, { "epoch": 5.5, "learning_rate": 1.3524308865586273e-06, "loss": 1.6025, "step": 28855 }, { "epoch": 5.5, "learning_rate": 1.352373689227836e-06, "loss": 2.2663, "step": 28856 }, { "epoch": 5.5, "learning_rate": 1.352316491897045e-06, "loss": 1.9585, "step": 28857 }, { "epoch": 5.5, "learning_rate": 1.3522592945662537e-06, "loss": 1.9733, "step": 28858 }, { "epoch": 5.5, "learning_rate": 1.3522020972354625e-06, "loss": 1.3343, "step": 28859 }, { "epoch": 5.5, "learning_rate": 1.352144899904671e-06, "loss": 0.9451, "step": 28860 }, { "epoch": 5.5, "learning_rate": 1.3520877025738799e-06, "loss": 2.1669, "step": 28861 }, { "epoch": 5.5, "learning_rate": 1.3520305052430887e-06, "loss": 1.4911, "step": 28862 }, { "epoch": 5.5, "learning_rate": 1.3519733079122975e-06, "loss": 1.7587, "step": 28863 }, { "epoch": 5.5, "learning_rate": 1.3519161105815062e-06, "loss": 1.0086, "step": 28864 }, { "epoch": 5.5, "learning_rate": 1.351858913250715e-06, "loss": 2.7008, "step": 28865 }, { "epoch": 5.5, "learning_rate": 1.3518017159199238e-06, "loss": 3.4246, "step": 28866 }, { "epoch": 5.5, "learning_rate": 1.3517445185891324e-06, "loss": 1.3636, "step": 28867 }, { "epoch": 5.5, "learning_rate": 1.3516873212583414e-06, "loss": 1.5656, "step": 28868 }, { "epoch": 5.5, "learning_rate": 1.3516301239275502e-06, "loss": 2.0237, "step": 28869 }, { "epoch": 5.5, "learning_rate": 1.3515729265967588e-06, "loss": 1.6737, "step": 28870 }, { "epoch": 5.5, "learning_rate": 1.3515157292659676e-06, "loss": 2.9532, "step": 28871 }, { "epoch": 5.5, "learning_rate": 1.3514585319351764e-06, "loss": 2.4586, "step": 28872 }, { "epoch": 5.5, "learning_rate": 1.3514013346043852e-06, "loss": 1.8334, "step": 28873 }, { "epoch": 5.5, "learning_rate": 1.351344137273594e-06, "loss": 1.4543, "step": 28874 }, { "epoch": 5.5, "learning_rate": 1.3512869399428028e-06, "loss": 1.3684, "step": 28875 }, { "epoch": 5.5, "learning_rate": 1.3512297426120116e-06, "loss": 1.4182, "step": 28876 }, { "epoch": 5.5, "learning_rate": 1.3511725452812201e-06, "loss": 1.6903, "step": 28877 }, { "epoch": 5.5, "learning_rate": 1.351115347950429e-06, "loss": 1.544, "step": 28878 }, { "epoch": 5.5, "learning_rate": 1.3510581506196377e-06, "loss": 1.7206, "step": 28879 }, { "epoch": 5.5, "learning_rate": 1.3510009532888465e-06, "loss": 1.6421, "step": 28880 }, { "epoch": 5.5, "learning_rate": 1.3509437559580553e-06, "loss": 1.4493, "step": 28881 }, { "epoch": 5.5, "learning_rate": 1.3508865586272641e-06, "loss": 2.2979, "step": 28882 }, { "epoch": 5.5, "learning_rate": 1.350829361296473e-06, "loss": 2.1888, "step": 28883 }, { "epoch": 5.5, "learning_rate": 1.3507721639656817e-06, "loss": 0.811, "step": 28884 }, { "epoch": 5.5, "learning_rate": 1.3507149666348905e-06, "loss": 1.3302, "step": 28885 }, { "epoch": 5.5, "learning_rate": 1.3506577693040993e-06, "loss": 2.4236, "step": 28886 }, { "epoch": 5.5, "learning_rate": 1.3506005719733079e-06, "loss": 1.6262, "step": 28887 }, { "epoch": 5.5, "learning_rate": 1.3505433746425167e-06, "loss": 1.9136, "step": 28888 }, { "epoch": 5.5, "learning_rate": 1.3504861773117255e-06, "loss": 1.852, "step": 28889 }, { "epoch": 5.5, "learning_rate": 1.3504289799809342e-06, "loss": 1.609, "step": 28890 }, { "epoch": 5.5, "learning_rate": 1.350371782650143e-06, "loss": 2.9189, "step": 28891 }, { "epoch": 5.5, "learning_rate": 1.3503145853193518e-06, "loss": 2.7446, "step": 28892 }, { "epoch": 5.5, "learning_rate": 1.3502573879885606e-06, "loss": 1.7586, "step": 28893 }, { "epoch": 5.5, "learning_rate": 1.3502001906577692e-06, "loss": 2.1151, "step": 28894 }, { "epoch": 5.5, "learning_rate": 1.350142993326978e-06, "loss": 2.7994, "step": 28895 }, { "epoch": 5.5, "learning_rate": 1.350085795996187e-06, "loss": 0.6332, "step": 28896 }, { "epoch": 5.5, "learning_rate": 1.3500285986653956e-06, "loss": 1.9828, "step": 28897 }, { "epoch": 5.5, "learning_rate": 1.3499714013346044e-06, "loss": 1.3562, "step": 28898 }, { "epoch": 5.5, "learning_rate": 1.3499142040038132e-06, "loss": 1.1735, "step": 28899 }, { "epoch": 5.5, "learning_rate": 1.349857006673022e-06, "loss": 1.7796, "step": 28900 }, { "epoch": 5.5, "learning_rate": 1.3497998093422308e-06, "loss": 1.9445, "step": 28901 }, { "epoch": 5.51, "learning_rate": 1.3497426120114396e-06, "loss": 1.1698, "step": 28902 }, { "epoch": 5.51, "learning_rate": 1.3496854146806484e-06, "loss": 1.2295, "step": 28903 }, { "epoch": 5.51, "learning_rate": 1.349628217349857e-06, "loss": 1.5415, "step": 28904 }, { "epoch": 5.51, "learning_rate": 1.3495710200190657e-06, "loss": 2.0742, "step": 28905 }, { "epoch": 5.51, "learning_rate": 1.3495138226882745e-06, "loss": 1.5265, "step": 28906 }, { "epoch": 5.51, "learning_rate": 1.3494566253574835e-06, "loss": 3.4856, "step": 28907 }, { "epoch": 5.51, "learning_rate": 1.3493994280266921e-06, "loss": 1.0686, "step": 28908 }, { "epoch": 5.51, "learning_rate": 1.349342230695901e-06, "loss": 1.4, "step": 28909 }, { "epoch": 5.51, "learning_rate": 1.3492850333651097e-06, "loss": 1.73, "step": 28910 }, { "epoch": 5.51, "learning_rate": 1.3492278360343183e-06, "loss": 2.2921, "step": 28911 }, { "epoch": 5.51, "learning_rate": 1.3491706387035273e-06, "loss": 1.2229, "step": 28912 }, { "epoch": 5.51, "learning_rate": 1.349113441372736e-06, "loss": 2.0358, "step": 28913 }, { "epoch": 5.51, "learning_rate": 1.3490562440419447e-06, "loss": 1.676, "step": 28914 }, { "epoch": 5.51, "learning_rate": 1.3489990467111535e-06, "loss": 1.398, "step": 28915 }, { "epoch": 5.51, "learning_rate": 1.3489418493803623e-06, "loss": 1.3136, "step": 28916 }, { "epoch": 5.51, "learning_rate": 1.348884652049571e-06, "loss": 1.3662, "step": 28917 }, { "epoch": 5.51, "learning_rate": 1.3488274547187798e-06, "loss": 2.9291, "step": 28918 }, { "epoch": 5.51, "learning_rate": 1.3487702573879886e-06, "loss": 1.0225, "step": 28919 }, { "epoch": 5.51, "learning_rate": 1.3487130600571974e-06, "loss": 3.1527, "step": 28920 }, { "epoch": 5.51, "learning_rate": 1.348655862726406e-06, "loss": 2.9588, "step": 28921 }, { "epoch": 5.51, "learning_rate": 1.3485986653956148e-06, "loss": 1.4227, "step": 28922 }, { "epoch": 5.51, "learning_rate": 1.3485414680648236e-06, "loss": 2.3296, "step": 28923 }, { "epoch": 5.51, "learning_rate": 1.3484842707340326e-06, "loss": 2.0729, "step": 28924 }, { "epoch": 5.51, "learning_rate": 1.3484270734032412e-06, "loss": 1.6703, "step": 28925 }, { "epoch": 5.51, "learning_rate": 1.34836987607245e-06, "loss": 2.4137, "step": 28926 }, { "epoch": 5.51, "learning_rate": 1.3483126787416588e-06, "loss": 2.208, "step": 28927 }, { "epoch": 5.51, "learning_rate": 1.3482554814108676e-06, "loss": 2.1853, "step": 28928 }, { "epoch": 5.51, "learning_rate": 1.3481982840800764e-06, "loss": 1.5749, "step": 28929 }, { "epoch": 5.51, "learning_rate": 1.3481410867492852e-06, "loss": 1.9841, "step": 28930 }, { "epoch": 5.51, "learning_rate": 1.348083889418494e-06, "loss": 3.2958, "step": 28931 }, { "epoch": 5.51, "learning_rate": 1.3480266920877025e-06, "loss": 2.2558, "step": 28932 }, { "epoch": 5.51, "learning_rate": 1.3479694947569113e-06, "loss": 0.7908, "step": 28933 }, { "epoch": 5.51, "learning_rate": 1.3479122974261201e-06, "loss": 1.4899, "step": 28934 }, { "epoch": 5.51, "learning_rate": 1.347855100095329e-06, "loss": 2.7019, "step": 28935 }, { "epoch": 5.51, "learning_rate": 1.3477979027645377e-06, "loss": 2.4472, "step": 28936 }, { "epoch": 5.51, "learning_rate": 1.3477407054337465e-06, "loss": 2.792, "step": 28937 }, { "epoch": 5.51, "learning_rate": 1.347683508102955e-06, "loss": 2.6804, "step": 28938 }, { "epoch": 5.51, "learning_rate": 1.3476263107721639e-06, "loss": 0.8482, "step": 28939 }, { "epoch": 5.51, "learning_rate": 1.3475691134413729e-06, "loss": 3.2237, "step": 28940 }, { "epoch": 5.51, "learning_rate": 1.3475119161105817e-06, "loss": 2.8372, "step": 28941 }, { "epoch": 5.51, "learning_rate": 1.3474547187797903e-06, "loss": 2.2435, "step": 28942 }, { "epoch": 5.51, "learning_rate": 1.347397521448999e-06, "loss": 1.6838, "step": 28943 }, { "epoch": 5.51, "learning_rate": 1.3473403241182078e-06, "loss": 1.8035, "step": 28944 }, { "epoch": 5.51, "learning_rate": 1.3472831267874166e-06, "loss": 1.9793, "step": 28945 }, { "epoch": 5.51, "learning_rate": 1.3472259294566254e-06, "loss": 1.7009, "step": 28946 }, { "epoch": 5.51, "learning_rate": 1.3471687321258342e-06, "loss": 0.9108, "step": 28947 }, { "epoch": 5.51, "learning_rate": 1.347111534795043e-06, "loss": 3.133, "step": 28948 }, { "epoch": 5.51, "learning_rate": 1.3470543374642516e-06, "loss": 1.7637, "step": 28949 }, { "epoch": 5.51, "learning_rate": 1.3469971401334604e-06, "loss": 2.6666, "step": 28950 }, { "epoch": 5.51, "learning_rate": 1.3469399428026694e-06, "loss": 2.6038, "step": 28951 }, { "epoch": 5.51, "learning_rate": 1.346882745471878e-06, "loss": 1.6805, "step": 28952 }, { "epoch": 5.51, "learning_rate": 1.3468255481410868e-06, "loss": 2.1455, "step": 28953 }, { "epoch": 5.52, "learning_rate": 1.3467683508102956e-06, "loss": 2.8681, "step": 28954 }, { "epoch": 5.52, "learning_rate": 1.3467111534795042e-06, "loss": 2.6142, "step": 28955 }, { "epoch": 5.52, "learning_rate": 1.3466539561487132e-06, "loss": 3.0893, "step": 28956 }, { "epoch": 5.52, "learning_rate": 1.346596758817922e-06, "loss": 2.3855, "step": 28957 }, { "epoch": 5.52, "learning_rate": 1.3465395614871307e-06, "loss": 3.1942, "step": 28958 }, { "epoch": 5.52, "learning_rate": 1.3464823641563393e-06, "loss": 1.4247, "step": 28959 }, { "epoch": 5.52, "learning_rate": 1.3464251668255481e-06, "loss": 2.8025, "step": 28960 }, { "epoch": 5.52, "learning_rate": 1.346367969494757e-06, "loss": 1.3394, "step": 28961 }, { "epoch": 5.52, "learning_rate": 1.3463107721639657e-06, "loss": 1.8973, "step": 28962 }, { "epoch": 5.52, "learning_rate": 1.3462535748331745e-06, "loss": 2.0233, "step": 28963 }, { "epoch": 5.52, "learning_rate": 1.3461963775023833e-06, "loss": 1.7715, "step": 28964 }, { "epoch": 5.52, "learning_rate": 1.346139180171592e-06, "loss": 2.2482, "step": 28965 }, { "epoch": 5.52, "learning_rate": 1.3460819828408007e-06, "loss": 2.4354, "step": 28966 }, { "epoch": 5.52, "learning_rate": 1.3460247855100095e-06, "loss": 2.7392, "step": 28967 }, { "epoch": 5.52, "learning_rate": 1.3459675881792185e-06, "loss": 0.7879, "step": 28968 }, { "epoch": 5.52, "learning_rate": 1.345910390848427e-06, "loss": 1.7792, "step": 28969 }, { "epoch": 5.52, "learning_rate": 1.3458531935176358e-06, "loss": 1.5851, "step": 28970 }, { "epoch": 5.52, "learning_rate": 1.3457959961868446e-06, "loss": 0.5895, "step": 28971 }, { "epoch": 5.52, "learning_rate": 1.3457387988560534e-06, "loss": 2.1531, "step": 28972 }, { "epoch": 5.52, "learning_rate": 1.3456816015252622e-06, "loss": 2.2703, "step": 28973 }, { "epoch": 5.52, "learning_rate": 1.345624404194471e-06, "loss": 2.0763, "step": 28974 }, { "epoch": 5.52, "learning_rate": 1.3455672068636798e-06, "loss": 3.0929, "step": 28975 }, { "epoch": 5.52, "learning_rate": 1.3455100095328884e-06, "loss": 1.6984, "step": 28976 }, { "epoch": 5.52, "learning_rate": 1.3454528122020972e-06, "loss": 1.1429, "step": 28977 }, { "epoch": 5.52, "learning_rate": 1.345395614871306e-06, "loss": 2.4121, "step": 28978 }, { "epoch": 5.52, "learning_rate": 1.3453384175405148e-06, "loss": 1.1462, "step": 28979 }, { "epoch": 5.52, "learning_rate": 1.3452812202097236e-06, "loss": 1.36, "step": 28980 }, { "epoch": 5.52, "learning_rate": 1.3452240228789324e-06, "loss": 2.7718, "step": 28981 }, { "epoch": 5.52, "learning_rate": 1.3451668255481412e-06, "loss": 2.0086, "step": 28982 }, { "epoch": 5.52, "learning_rate": 1.3451096282173497e-06, "loss": 1.365, "step": 28983 }, { "epoch": 5.52, "learning_rate": 1.3450524308865588e-06, "loss": 2.1617, "step": 28984 }, { "epoch": 5.52, "learning_rate": 1.3449952335557675e-06, "loss": 2.4714, "step": 28985 }, { "epoch": 5.52, "learning_rate": 1.3449380362249761e-06, "loss": 1.495, "step": 28986 }, { "epoch": 5.52, "learning_rate": 1.344880838894185e-06, "loss": 2.5592, "step": 28987 }, { "epoch": 5.52, "learning_rate": 1.3448236415633937e-06, "loss": 2.992, "step": 28988 }, { "epoch": 5.52, "learning_rate": 1.3447664442326025e-06, "loss": 1.9803, "step": 28989 }, { "epoch": 5.52, "learning_rate": 1.3447092469018113e-06, "loss": 0.8184, "step": 28990 }, { "epoch": 5.52, "learning_rate": 1.34465204957102e-06, "loss": 2.9203, "step": 28991 }, { "epoch": 5.52, "learning_rate": 1.3445948522402289e-06, "loss": 1.3083, "step": 28992 }, { "epoch": 5.52, "learning_rate": 1.3445376549094375e-06, "loss": 1.9761, "step": 28993 }, { "epoch": 5.52, "learning_rate": 1.3444804575786463e-06, "loss": 1.5596, "step": 28994 }, { "epoch": 5.52, "learning_rate": 1.3444232602478553e-06, "loss": 1.6733, "step": 28995 }, { "epoch": 5.52, "learning_rate": 1.344366062917064e-06, "loss": 1.7598, "step": 28996 }, { "epoch": 5.52, "learning_rate": 1.3443088655862726e-06, "loss": 1.7508, "step": 28997 }, { "epoch": 5.52, "learning_rate": 1.3442516682554814e-06, "loss": 2.9733, "step": 28998 }, { "epoch": 5.52, "learning_rate": 1.3441944709246902e-06, "loss": 1.1185, "step": 28999 }, { "epoch": 5.52, "learning_rate": 1.344137273593899e-06, "loss": 1.9754, "step": 29000 }, { "epoch": 5.52, "learning_rate": 1.3440800762631078e-06, "loss": 2.4306, "step": 29001 }, { "epoch": 5.52, "learning_rate": 1.3440228789323166e-06, "loss": 0.7735, "step": 29002 }, { "epoch": 5.52, "learning_rate": 1.3439656816015252e-06, "loss": 1.4377, "step": 29003 }, { "epoch": 5.52, "learning_rate": 1.343908484270734e-06, "loss": 1.7182, "step": 29004 }, { "epoch": 5.52, "learning_rate": 1.3438512869399428e-06, "loss": 2.5513, "step": 29005 }, { "epoch": 5.52, "learning_rate": 1.3437940896091516e-06, "loss": 1.5923, "step": 29006 }, { "epoch": 5.53, "learning_rate": 1.3437368922783604e-06, "loss": 2.4506, "step": 29007 }, { "epoch": 5.53, "learning_rate": 1.3436796949475692e-06, "loss": 4.5747, "step": 29008 }, { "epoch": 5.53, "learning_rate": 1.343622497616778e-06, "loss": 1.1328, "step": 29009 }, { "epoch": 5.53, "learning_rate": 1.3435653002859865e-06, "loss": 3.052, "step": 29010 }, { "epoch": 5.53, "learning_rate": 1.3435081029551953e-06, "loss": 1.5994, "step": 29011 }, { "epoch": 5.53, "learning_rate": 1.3434509056244043e-06, "loss": 2.0689, "step": 29012 }, { "epoch": 5.53, "learning_rate": 1.3433937082936131e-06, "loss": 0.319, "step": 29013 }, { "epoch": 5.53, "learning_rate": 1.3433365109628217e-06, "loss": 1.6465, "step": 29014 }, { "epoch": 5.53, "learning_rate": 1.3432793136320305e-06, "loss": 1.2959, "step": 29015 }, { "epoch": 5.53, "learning_rate": 1.3432221163012393e-06, "loss": 1.7418, "step": 29016 }, { "epoch": 5.53, "learning_rate": 1.343164918970448e-06, "loss": 1.1831, "step": 29017 }, { "epoch": 5.53, "learning_rate": 1.3431077216396569e-06, "loss": 2.5206, "step": 29018 }, { "epoch": 5.53, "learning_rate": 1.3430505243088657e-06, "loss": 2.9212, "step": 29019 }, { "epoch": 5.53, "learning_rate": 1.3429933269780743e-06, "loss": 1.2615, "step": 29020 }, { "epoch": 5.53, "learning_rate": 1.342936129647283e-06, "loss": 1.8934, "step": 29021 }, { "epoch": 5.53, "learning_rate": 1.3428789323164919e-06, "loss": 1.6448, "step": 29022 }, { "epoch": 5.53, "learning_rate": 1.3428217349857009e-06, "loss": 2.5443, "step": 29023 }, { "epoch": 5.53, "learning_rate": 1.3427645376549094e-06, "loss": 0.5729, "step": 29024 }, { "epoch": 5.53, "learning_rate": 1.3427073403241182e-06, "loss": 3.7577, "step": 29025 }, { "epoch": 5.53, "learning_rate": 1.342650142993327e-06, "loss": 1.539, "step": 29026 }, { "epoch": 5.53, "learning_rate": 1.3425929456625356e-06, "loss": 2.0738, "step": 29027 }, { "epoch": 5.53, "learning_rate": 1.3425357483317446e-06, "loss": 2.7141, "step": 29028 }, { "epoch": 5.53, "learning_rate": 1.3424785510009534e-06, "loss": 1.0744, "step": 29029 }, { "epoch": 5.53, "learning_rate": 1.3424213536701622e-06, "loss": 3.543, "step": 29030 }, { "epoch": 5.53, "learning_rate": 1.3423641563393708e-06, "loss": 2.5378, "step": 29031 }, { "epoch": 5.53, "learning_rate": 1.3423069590085796e-06, "loss": 1.2159, "step": 29032 }, { "epoch": 5.53, "learning_rate": 1.3422497616777884e-06, "loss": 2.2755, "step": 29033 }, { "epoch": 5.53, "learning_rate": 1.3421925643469972e-06, "loss": 2.385, "step": 29034 }, { "epoch": 5.53, "learning_rate": 1.342135367016206e-06, "loss": 1.3038, "step": 29035 }, { "epoch": 5.53, "learning_rate": 1.3420781696854148e-06, "loss": 1.9951, "step": 29036 }, { "epoch": 5.53, "learning_rate": 1.3420209723546236e-06, "loss": 1.7949, "step": 29037 }, { "epoch": 5.53, "learning_rate": 1.3419637750238321e-06, "loss": 2.514, "step": 29038 }, { "epoch": 5.53, "learning_rate": 1.3419065776930411e-06, "loss": 1.8665, "step": 29039 }, { "epoch": 5.53, "learning_rate": 1.34184938036225e-06, "loss": 0.9619, "step": 29040 }, { "epoch": 5.53, "learning_rate": 1.3417921830314585e-06, "loss": 1.6877, "step": 29041 }, { "epoch": 5.53, "learning_rate": 1.3417349857006673e-06, "loss": 2.6291, "step": 29042 }, { "epoch": 5.53, "learning_rate": 1.341677788369876e-06, "loss": 3.4636, "step": 29043 }, { "epoch": 5.53, "learning_rate": 1.341620591039085e-06, "loss": 3.3783, "step": 29044 }, { "epoch": 5.53, "learning_rate": 1.3415633937082937e-06, "loss": 0.6475, "step": 29045 }, { "epoch": 5.53, "learning_rate": 1.3415061963775025e-06, "loss": 1.8714, "step": 29046 }, { "epoch": 5.53, "learning_rate": 1.3414489990467113e-06, "loss": 2.5159, "step": 29047 }, { "epoch": 5.53, "learning_rate": 1.3413918017159199e-06, "loss": 0.8822, "step": 29048 }, { "epoch": 5.53, "learning_rate": 1.3413346043851287e-06, "loss": 2.4342, "step": 29049 }, { "epoch": 5.53, "learning_rate": 1.3412774070543374e-06, "loss": 1.6496, "step": 29050 }, { "epoch": 5.53, "learning_rate": 1.3412202097235462e-06, "loss": 2.6698, "step": 29051 }, { "epoch": 5.53, "learning_rate": 1.341163012392755e-06, "loss": 1.9236, "step": 29052 }, { "epoch": 5.53, "learning_rate": 1.3411058150619638e-06, "loss": 0.6311, "step": 29053 }, { "epoch": 5.53, "learning_rate": 1.3410486177311726e-06, "loss": 2.6067, "step": 29054 }, { "epoch": 5.53, "learning_rate": 1.3409914204003812e-06, "loss": 1.4339, "step": 29055 }, { "epoch": 5.53, "learning_rate": 1.3409342230695902e-06, "loss": 1.1201, "step": 29056 }, { "epoch": 5.53, "learning_rate": 1.340877025738799e-06, "loss": 1.5145, "step": 29057 }, { "epoch": 5.53, "learning_rate": 1.3408198284080076e-06, "loss": 1.8713, "step": 29058 }, { "epoch": 5.54, "learning_rate": 1.3407626310772164e-06, "loss": 1.7487, "step": 29059 }, { "epoch": 5.54, "learning_rate": 1.3407054337464252e-06, "loss": 1.3324, "step": 29060 }, { "epoch": 5.54, "learning_rate": 1.340648236415634e-06, "loss": 2.6116, "step": 29061 }, { "epoch": 5.54, "learning_rate": 1.3405910390848428e-06, "loss": 2.129, "step": 29062 }, { "epoch": 5.54, "learning_rate": 1.3405338417540516e-06, "loss": 1.7647, "step": 29063 }, { "epoch": 5.54, "learning_rate": 1.3404766444232604e-06, "loss": 1.7951, "step": 29064 }, { "epoch": 5.54, "learning_rate": 1.340419447092469e-06, "loss": 1.776, "step": 29065 }, { "epoch": 5.54, "learning_rate": 1.3403622497616777e-06, "loss": 2.3248, "step": 29066 }, { "epoch": 5.54, "learning_rate": 1.3403050524308867e-06, "loss": 1.6995, "step": 29067 }, { "epoch": 5.54, "learning_rate": 1.3402478551000953e-06, "loss": 2.5938, "step": 29068 }, { "epoch": 5.54, "learning_rate": 1.3401906577693041e-06, "loss": 1.5836, "step": 29069 }, { "epoch": 5.54, "learning_rate": 1.340133460438513e-06, "loss": 1.7555, "step": 29070 }, { "epoch": 5.54, "learning_rate": 1.3400762631077217e-06, "loss": 1.7586, "step": 29071 }, { "epoch": 5.54, "learning_rate": 1.3400190657769305e-06, "loss": 1.9309, "step": 29072 }, { "epoch": 5.54, "learning_rate": 1.3399618684461393e-06, "loss": 2.7106, "step": 29073 }, { "epoch": 5.54, "learning_rate": 1.339904671115348e-06, "loss": 1.8942, "step": 29074 }, { "epoch": 5.54, "learning_rate": 1.3398474737845567e-06, "loss": 0.7261, "step": 29075 }, { "epoch": 5.54, "learning_rate": 1.3397902764537655e-06, "loss": 0.5664, "step": 29076 }, { "epoch": 5.54, "learning_rate": 1.3397330791229742e-06, "loss": 1.85, "step": 29077 }, { "epoch": 5.54, "learning_rate": 1.339675881792183e-06, "loss": 2.9501, "step": 29078 }, { "epoch": 5.54, "learning_rate": 1.3396186844613918e-06, "loss": 0.9989, "step": 29079 }, { "epoch": 5.54, "learning_rate": 1.3395614871306006e-06, "loss": 1.5335, "step": 29080 }, { "epoch": 5.54, "learning_rate": 1.3395042897998094e-06, "loss": 1.5727, "step": 29081 }, { "epoch": 5.54, "learning_rate": 1.339447092469018e-06, "loss": 1.0291, "step": 29082 }, { "epoch": 5.54, "learning_rate": 1.339389895138227e-06, "loss": 3.0985, "step": 29083 }, { "epoch": 5.54, "learning_rate": 1.3393326978074358e-06, "loss": 3.2274, "step": 29084 }, { "epoch": 5.54, "learning_rate": 1.3392755004766444e-06, "loss": 3.0173, "step": 29085 }, { "epoch": 5.54, "learning_rate": 1.3392183031458532e-06, "loss": 2.3233, "step": 29086 }, { "epoch": 5.54, "learning_rate": 1.339161105815062e-06, "loss": 2.5383, "step": 29087 }, { "epoch": 5.54, "learning_rate": 1.3391039084842708e-06, "loss": 2.3548, "step": 29088 }, { "epoch": 5.54, "learning_rate": 1.3390467111534796e-06, "loss": 1.1152, "step": 29089 }, { "epoch": 5.54, "learning_rate": 1.3389895138226884e-06, "loss": 2.743, "step": 29090 }, { "epoch": 5.54, "learning_rate": 1.3389323164918971e-06, "loss": 1.7826, "step": 29091 }, { "epoch": 5.54, "learning_rate": 1.3388751191611057e-06, "loss": 1.5377, "step": 29092 }, { "epoch": 5.54, "learning_rate": 1.3388179218303145e-06, "loss": 2.3349, "step": 29093 }, { "epoch": 5.54, "learning_rate": 1.3387607244995233e-06, "loss": 1.3642, "step": 29094 }, { "epoch": 5.54, "learning_rate": 1.3387035271687323e-06, "loss": 2.9772, "step": 29095 }, { "epoch": 5.54, "learning_rate": 1.338646329837941e-06, "loss": 3.15, "step": 29096 }, { "epoch": 5.54, "learning_rate": 1.3385891325071497e-06, "loss": 1.1078, "step": 29097 }, { "epoch": 5.54, "learning_rate": 1.3385319351763585e-06, "loss": 1.7189, "step": 29098 }, { "epoch": 5.54, "learning_rate": 1.338474737845567e-06, "loss": 2.1956, "step": 29099 }, { "epoch": 5.54, "learning_rate": 1.338417540514776e-06, "loss": 1.408, "step": 29100 }, { "epoch": 5.54, "learning_rate": 1.3383603431839849e-06, "loss": 1.6195, "step": 29101 }, { "epoch": 5.54, "learning_rate": 1.3383031458531937e-06, "loss": 2.8837, "step": 29102 }, { "epoch": 5.54, "learning_rate": 1.3382459485224023e-06, "loss": 2.5301, "step": 29103 }, { "epoch": 5.54, "learning_rate": 1.338188751191611e-06, "loss": 2.8104, "step": 29104 }, { "epoch": 5.54, "learning_rate": 1.3381315538608198e-06, "loss": 2.285, "step": 29105 }, { "epoch": 5.54, "learning_rate": 1.3380743565300286e-06, "loss": 3.4985, "step": 29106 }, { "epoch": 5.54, "learning_rate": 1.3380171591992374e-06, "loss": 2.1625, "step": 29107 }, { "epoch": 5.54, "learning_rate": 1.3379599618684462e-06, "loss": 2.4855, "step": 29108 }, { "epoch": 5.54, "learning_rate": 1.3379027645376548e-06, "loss": 2.7592, "step": 29109 }, { "epoch": 5.54, "learning_rate": 1.3378455672068636e-06, "loss": 1.9084, "step": 29110 }, { "epoch": 5.54, "learning_rate": 1.3377883698760726e-06, "loss": 1.2272, "step": 29111 }, { "epoch": 5.55, "learning_rate": 1.3377311725452814e-06, "loss": 2.7374, "step": 29112 }, { "epoch": 5.55, "learning_rate": 1.33767397521449e-06, "loss": 1.8983, "step": 29113 }, { "epoch": 5.55, "learning_rate": 1.3376167778836988e-06, "loss": 2.0435, "step": 29114 }, { "epoch": 5.55, "learning_rate": 1.3375595805529076e-06, "loss": 1.8432, "step": 29115 }, { "epoch": 5.55, "learning_rate": 1.3375023832221164e-06, "loss": 2.3965, "step": 29116 }, { "epoch": 5.55, "learning_rate": 1.3374451858913252e-06, "loss": 0.9557, "step": 29117 }, { "epoch": 5.55, "learning_rate": 1.337387988560534e-06, "loss": 1.3477, "step": 29118 }, { "epoch": 5.55, "learning_rate": 1.3373307912297427e-06, "loss": 1.7573, "step": 29119 }, { "epoch": 5.55, "learning_rate": 1.3372735938989513e-06, "loss": 2.4502, "step": 29120 }, { "epoch": 5.55, "learning_rate": 1.3372163965681601e-06, "loss": 1.593, "step": 29121 }, { "epoch": 5.55, "learning_rate": 1.337159199237369e-06, "loss": 1.6253, "step": 29122 }, { "epoch": 5.55, "learning_rate": 1.3371020019065777e-06, "loss": 1.1142, "step": 29123 }, { "epoch": 5.55, "learning_rate": 1.3370448045757865e-06, "loss": 3.1603, "step": 29124 }, { "epoch": 5.55, "learning_rate": 1.3369876072449953e-06, "loss": 2.2061, "step": 29125 }, { "epoch": 5.55, "learning_rate": 1.3369304099142039e-06, "loss": 2.0916, "step": 29126 }, { "epoch": 5.55, "learning_rate": 1.3368732125834129e-06, "loss": 2.9019, "step": 29127 }, { "epoch": 5.55, "learning_rate": 1.3368160152526217e-06, "loss": 2.41, "step": 29128 }, { "epoch": 5.55, "learning_rate": 1.3367588179218305e-06, "loss": 1.111, "step": 29129 }, { "epoch": 5.55, "learning_rate": 1.336701620591039e-06, "loss": 4.4039, "step": 29130 }, { "epoch": 5.55, "learning_rate": 1.3366444232602478e-06, "loss": 1.8576, "step": 29131 }, { "epoch": 5.55, "learning_rate": 1.3365872259294566e-06, "loss": 2.7376, "step": 29132 }, { "epoch": 5.55, "learning_rate": 1.3365300285986654e-06, "loss": 2.3648, "step": 29133 }, { "epoch": 5.55, "learning_rate": 1.3364728312678742e-06, "loss": 1.3346, "step": 29134 }, { "epoch": 5.55, "learning_rate": 1.336415633937083e-06, "loss": 2.6667, "step": 29135 }, { "epoch": 5.55, "learning_rate": 1.3363584366062918e-06, "loss": 2.0185, "step": 29136 }, { "epoch": 5.55, "learning_rate": 1.3363012392755004e-06, "loss": 1.6555, "step": 29137 }, { "epoch": 5.55, "learning_rate": 1.3362440419447092e-06, "loss": 0.873, "step": 29138 }, { "epoch": 5.55, "learning_rate": 1.3361868446139182e-06, "loss": 1.8089, "step": 29139 }, { "epoch": 5.55, "learning_rate": 1.3361296472831268e-06, "loss": 2.7565, "step": 29140 }, { "epoch": 5.55, "learning_rate": 1.3360724499523356e-06, "loss": 1.0782, "step": 29141 }, { "epoch": 5.55, "learning_rate": 1.3360152526215444e-06, "loss": 1.6623, "step": 29142 }, { "epoch": 5.55, "learning_rate": 1.3359580552907532e-06, "loss": 1.4115, "step": 29143 }, { "epoch": 5.55, "learning_rate": 1.335900857959962e-06, "loss": 2.7368, "step": 29144 }, { "epoch": 5.55, "learning_rate": 1.3358436606291707e-06, "loss": 0.8063, "step": 29145 }, { "epoch": 5.55, "learning_rate": 1.3357864632983795e-06, "loss": 1.5112, "step": 29146 }, { "epoch": 5.55, "learning_rate": 1.3357292659675881e-06, "loss": 0.9353, "step": 29147 }, { "epoch": 5.55, "learning_rate": 1.335672068636797e-06, "loss": 2.0232, "step": 29148 }, { "epoch": 5.55, "learning_rate": 1.3356148713060057e-06, "loss": 1.8468, "step": 29149 }, { "epoch": 5.55, "learning_rate": 1.3355576739752145e-06, "loss": 0.7758, "step": 29150 }, { "epoch": 5.55, "learning_rate": 1.3355004766444233e-06, "loss": 1.8835, "step": 29151 }, { "epoch": 5.55, "learning_rate": 1.335443279313632e-06, "loss": 1.579, "step": 29152 }, { "epoch": 5.55, "learning_rate": 1.3353860819828409e-06, "loss": 0.8272, "step": 29153 }, { "epoch": 5.55, "learning_rate": 1.3353288846520495e-06, "loss": 4.4909, "step": 29154 }, { "epoch": 5.55, "learning_rate": 1.3352716873212585e-06, "loss": 1.2866, "step": 29155 }, { "epoch": 5.55, "learning_rate": 1.3352144899904673e-06, "loss": 2.2331, "step": 29156 }, { "epoch": 5.55, "learning_rate": 1.3351572926596758e-06, "loss": 1.817, "step": 29157 }, { "epoch": 5.55, "learning_rate": 1.3351000953288846e-06, "loss": 2.1458, "step": 29158 }, { "epoch": 5.55, "learning_rate": 1.3350428979980934e-06, "loss": 1.306, "step": 29159 }, { "epoch": 5.55, "learning_rate": 1.3349857006673022e-06, "loss": 2.0111, "step": 29160 }, { "epoch": 5.55, "learning_rate": 1.334928503336511e-06, "loss": 1.9766, "step": 29161 }, { "epoch": 5.55, "learning_rate": 1.3348713060057198e-06, "loss": 2.3575, "step": 29162 }, { "epoch": 5.55, "learning_rate": 1.3348141086749286e-06, "loss": 1.9454, "step": 29163 }, { "epoch": 5.56, "learning_rate": 1.3347569113441372e-06, "loss": 2.1559, "step": 29164 }, { "epoch": 5.56, "learning_rate": 1.334699714013346e-06, "loss": 3.0625, "step": 29165 }, { "epoch": 5.56, "learning_rate": 1.3346425166825548e-06, "loss": 2.8004, "step": 29166 }, { "epoch": 5.56, "learning_rate": 1.3345853193517638e-06, "loss": 0.7733, "step": 29167 }, { "epoch": 5.56, "learning_rate": 1.3345281220209724e-06, "loss": 1.4577, "step": 29168 }, { "epoch": 5.56, "learning_rate": 1.3344709246901812e-06, "loss": 1.9281, "step": 29169 }, { "epoch": 5.56, "learning_rate": 1.33441372735939e-06, "loss": 2.7043, "step": 29170 }, { "epoch": 5.56, "learning_rate": 1.3343565300285987e-06, "loss": 1.3256, "step": 29171 }, { "epoch": 5.56, "learning_rate": 1.3342993326978075e-06, "loss": 1.8791, "step": 29172 }, { "epoch": 5.56, "learning_rate": 1.3342421353670163e-06, "loss": 2.5994, "step": 29173 }, { "epoch": 5.56, "learning_rate": 1.334184938036225e-06, "loss": 1.5358, "step": 29174 }, { "epoch": 5.56, "learning_rate": 1.3341277407054337e-06, "loss": 1.8668, "step": 29175 }, { "epoch": 5.56, "learning_rate": 1.3340705433746425e-06, "loss": 1.4332, "step": 29176 }, { "epoch": 5.56, "learning_rate": 1.3340133460438513e-06, "loss": 1.5141, "step": 29177 }, { "epoch": 5.56, "learning_rate": 1.33395614871306e-06, "loss": 1.9708, "step": 29178 }, { "epoch": 5.56, "learning_rate": 1.3338989513822689e-06, "loss": 1.4169, "step": 29179 }, { "epoch": 5.56, "learning_rate": 1.3338417540514777e-06, "loss": 2.6755, "step": 29180 }, { "epoch": 5.56, "learning_rate": 1.3337845567206863e-06, "loss": 2.8508, "step": 29181 }, { "epoch": 5.56, "learning_rate": 1.333727359389895e-06, "loss": 1.223, "step": 29182 }, { "epoch": 5.56, "learning_rate": 1.333670162059104e-06, "loss": 2.1986, "step": 29183 }, { "epoch": 5.56, "learning_rate": 1.3336129647283129e-06, "loss": 1.9946, "step": 29184 }, { "epoch": 5.56, "learning_rate": 1.3335557673975214e-06, "loss": 2.2479, "step": 29185 }, { "epoch": 5.56, "learning_rate": 1.3334985700667302e-06, "loss": 0.8635, "step": 29186 }, { "epoch": 5.56, "learning_rate": 1.333441372735939e-06, "loss": 2.0407, "step": 29187 }, { "epoch": 5.56, "learning_rate": 1.3333841754051478e-06, "loss": 2.4897, "step": 29188 }, { "epoch": 5.56, "learning_rate": 1.3333269780743566e-06, "loss": 1.6308, "step": 29189 }, { "epoch": 5.56, "learning_rate": 1.3332697807435654e-06, "loss": 2.1979, "step": 29190 }, { "epoch": 5.56, "learning_rate": 1.333212583412774e-06, "loss": 1.6733, "step": 29191 }, { "epoch": 5.56, "learning_rate": 1.3331553860819828e-06, "loss": 3.0422, "step": 29192 }, { "epoch": 5.56, "learning_rate": 1.3330981887511916e-06, "loss": 1.7895, "step": 29193 }, { "epoch": 5.56, "learning_rate": 1.3330409914204006e-06, "loss": 1.9088, "step": 29194 }, { "epoch": 5.56, "learning_rate": 1.3329837940896092e-06, "loss": 2.195, "step": 29195 }, { "epoch": 5.56, "learning_rate": 1.332926596758818e-06, "loss": 2.1745, "step": 29196 }, { "epoch": 5.56, "learning_rate": 1.3328693994280268e-06, "loss": 1.4878, "step": 29197 }, { "epoch": 5.56, "learning_rate": 1.3328122020972353e-06, "loss": 2.0321, "step": 29198 }, { "epoch": 5.56, "learning_rate": 1.3327550047664443e-06, "loss": 1.4643, "step": 29199 }, { "epoch": 5.56, "learning_rate": 1.3326978074356531e-06, "loss": 2.1808, "step": 29200 }, { "epoch": 5.56, "learning_rate": 1.332640610104862e-06, "loss": 1.0257, "step": 29201 }, { "epoch": 5.56, "learning_rate": 1.3325834127740705e-06, "loss": 2.6086, "step": 29202 }, { "epoch": 5.56, "learning_rate": 1.3325262154432793e-06, "loss": 1.8975, "step": 29203 }, { "epoch": 5.56, "learning_rate": 1.332469018112488e-06, "loss": 0.7512, "step": 29204 }, { "epoch": 5.56, "learning_rate": 1.3324118207816969e-06, "loss": 1.2427, "step": 29205 }, { "epoch": 5.56, "learning_rate": 1.3323546234509057e-06, "loss": 1.2031, "step": 29206 }, { "epoch": 5.56, "learning_rate": 1.3322974261201145e-06, "loss": 2.4124, "step": 29207 }, { "epoch": 5.56, "learning_rate": 1.3322402287893233e-06, "loss": 2.272, "step": 29208 }, { "epoch": 5.56, "learning_rate": 1.3321830314585319e-06, "loss": 1.661, "step": 29209 }, { "epoch": 5.56, "learning_rate": 1.3321258341277406e-06, "loss": 1.8627, "step": 29210 }, { "epoch": 5.56, "learning_rate": 1.3320686367969497e-06, "loss": 1.0754, "step": 29211 }, { "epoch": 5.56, "learning_rate": 1.3320114394661582e-06, "loss": 3.1778, "step": 29212 }, { "epoch": 5.56, "learning_rate": 1.331954242135367e-06, "loss": 2.7345, "step": 29213 }, { "epoch": 5.56, "learning_rate": 1.3318970448045758e-06, "loss": 2.9944, "step": 29214 }, { "epoch": 5.56, "learning_rate": 1.3318398474737844e-06, "loss": 2.0432, "step": 29215 }, { "epoch": 5.56, "learning_rate": 1.3317826501429934e-06, "loss": 2.0657, "step": 29216 }, { "epoch": 5.57, "learning_rate": 1.3317254528122022e-06, "loss": 1.7528, "step": 29217 }, { "epoch": 5.57, "learning_rate": 1.331668255481411e-06, "loss": 2.0026, "step": 29218 }, { "epoch": 5.57, "learning_rate": 1.3316110581506196e-06, "loss": 2.0109, "step": 29219 }, { "epoch": 5.57, "learning_rate": 1.3315538608198284e-06, "loss": 1.0839, "step": 29220 }, { "epoch": 5.57, "learning_rate": 1.3314966634890372e-06, "loss": 1.3505, "step": 29221 }, { "epoch": 5.57, "learning_rate": 1.331439466158246e-06, "loss": 2.9227, "step": 29222 }, { "epoch": 5.57, "learning_rate": 1.3313822688274548e-06, "loss": 2.5934, "step": 29223 }, { "epoch": 5.57, "learning_rate": 1.3313250714966636e-06, "loss": 2.2935, "step": 29224 }, { "epoch": 5.57, "learning_rate": 1.3312678741658723e-06, "loss": 1.4867, "step": 29225 }, { "epoch": 5.57, "learning_rate": 1.331210676835081e-06, "loss": 2.5404, "step": 29226 }, { "epoch": 5.57, "learning_rate": 1.33115347950429e-06, "loss": 1.376, "step": 29227 }, { "epoch": 5.57, "learning_rate": 1.3310962821734987e-06, "loss": 1.8104, "step": 29228 }, { "epoch": 5.57, "learning_rate": 1.3310390848427073e-06, "loss": 1.7613, "step": 29229 }, { "epoch": 5.57, "learning_rate": 1.330981887511916e-06, "loss": 2.3501, "step": 29230 }, { "epoch": 5.57, "learning_rate": 1.330924690181125e-06, "loss": 1.5961, "step": 29231 }, { "epoch": 5.57, "learning_rate": 1.3308674928503337e-06, "loss": 1.4566, "step": 29232 }, { "epoch": 5.57, "learning_rate": 1.3308102955195425e-06, "loss": 1.7207, "step": 29233 }, { "epoch": 5.57, "learning_rate": 1.3307530981887513e-06, "loss": 3.1899, "step": 29234 }, { "epoch": 5.57, "learning_rate": 1.33069590085796e-06, "loss": 2.379, "step": 29235 }, { "epoch": 5.57, "learning_rate": 1.3306387035271687e-06, "loss": 2.8847, "step": 29236 }, { "epoch": 5.57, "learning_rate": 1.3305815061963774e-06, "loss": 1.3363, "step": 29237 }, { "epoch": 5.57, "learning_rate": 1.3305243088655865e-06, "loss": 1.2124, "step": 29238 }, { "epoch": 5.57, "learning_rate": 1.330467111534795e-06, "loss": 1.9412, "step": 29239 }, { "epoch": 5.57, "learning_rate": 1.3304099142040038e-06, "loss": 2.8375, "step": 29240 }, { "epoch": 5.57, "learning_rate": 1.3303527168732126e-06, "loss": 1.5631, "step": 29241 }, { "epoch": 5.57, "learning_rate": 1.3302955195424214e-06, "loss": 1.6572, "step": 29242 }, { "epoch": 5.57, "learning_rate": 1.3302383222116302e-06, "loss": 1.5107, "step": 29243 }, { "epoch": 5.57, "learning_rate": 1.330181124880839e-06, "loss": 1.4892, "step": 29244 }, { "epoch": 5.57, "learning_rate": 1.3301239275500478e-06, "loss": 2.7132, "step": 29245 }, { "epoch": 5.57, "learning_rate": 1.3300667302192564e-06, "loss": 2.1049, "step": 29246 }, { "epoch": 5.57, "learning_rate": 1.3300095328884652e-06, "loss": 1.8283, "step": 29247 }, { "epoch": 5.57, "learning_rate": 1.329952335557674e-06, "loss": 1.6366, "step": 29248 }, { "epoch": 5.57, "learning_rate": 1.3298951382268828e-06, "loss": 2.3331, "step": 29249 }, { "epoch": 5.57, "learning_rate": 1.3298379408960916e-06, "loss": 3.735, "step": 29250 }, { "epoch": 5.57, "learning_rate": 1.3297807435653003e-06, "loss": 1.6415, "step": 29251 }, { "epoch": 5.57, "learning_rate": 1.3297235462345091e-06, "loss": 1.2247, "step": 29252 }, { "epoch": 5.57, "learning_rate": 1.3296663489037177e-06, "loss": 1.6311, "step": 29253 }, { "epoch": 5.57, "learning_rate": 1.3296091515729265e-06, "loss": 1.7906, "step": 29254 }, { "epoch": 5.57, "learning_rate": 1.3295519542421355e-06, "loss": 2.8241, "step": 29255 }, { "epoch": 5.57, "learning_rate": 1.329494756911344e-06, "loss": 2.7618, "step": 29256 }, { "epoch": 5.57, "learning_rate": 1.329437559580553e-06, "loss": 3.0285, "step": 29257 }, { "epoch": 5.57, "learning_rate": 1.3293803622497617e-06, "loss": 2.1588, "step": 29258 }, { "epoch": 5.57, "learning_rate": 1.3293231649189705e-06, "loss": 1.3393, "step": 29259 }, { "epoch": 5.57, "learning_rate": 1.3292659675881793e-06, "loss": 1.3639, "step": 29260 }, { "epoch": 5.57, "learning_rate": 1.329208770257388e-06, "loss": 2.5185, "step": 29261 }, { "epoch": 5.57, "learning_rate": 1.3291515729265969e-06, "loss": 1.1094, "step": 29262 }, { "epoch": 5.57, "learning_rate": 1.3290943755958055e-06, "loss": 1.9394, "step": 29263 }, { "epoch": 5.57, "learning_rate": 1.3290371782650142e-06, "loss": 2.7915, "step": 29264 }, { "epoch": 5.57, "learning_rate": 1.328979980934223e-06, "loss": 1.9278, "step": 29265 }, { "epoch": 5.57, "learning_rate": 1.328922783603432e-06, "loss": 2.4509, "step": 29266 }, { "epoch": 5.57, "learning_rate": 1.3288655862726406e-06, "loss": 3.308, "step": 29267 }, { "epoch": 5.57, "learning_rate": 1.3288083889418494e-06, "loss": 1.7316, "step": 29268 }, { "epoch": 5.58, "learning_rate": 1.3287511916110582e-06, "loss": 0.9656, "step": 29269 }, { "epoch": 5.58, "learning_rate": 1.3286939942802668e-06, "loss": 1.6873, "step": 29270 }, { "epoch": 5.58, "learning_rate": 1.3286367969494758e-06, "loss": 1.742, "step": 29271 }, { "epoch": 5.58, "learning_rate": 1.3285795996186846e-06, "loss": 1.476, "step": 29272 }, { "epoch": 5.58, "learning_rate": 1.3285224022878934e-06, "loss": 2.3226, "step": 29273 }, { "epoch": 5.58, "learning_rate": 1.328465204957102e-06, "loss": 2.307, "step": 29274 }, { "epoch": 5.58, "learning_rate": 1.3284080076263108e-06, "loss": 2.4055, "step": 29275 }, { "epoch": 5.58, "learning_rate": 1.3283508102955196e-06, "loss": 1.7374, "step": 29276 }, { "epoch": 5.58, "learning_rate": 1.3282936129647284e-06, "loss": 2.3014, "step": 29277 }, { "epoch": 5.58, "learning_rate": 1.3282364156339371e-06, "loss": 2.3688, "step": 29278 }, { "epoch": 5.58, "learning_rate": 1.328179218303146e-06, "loss": 2.0686, "step": 29279 }, { "epoch": 5.58, "learning_rate": 1.3281220209723545e-06, "loss": 1.7181, "step": 29280 }, { "epoch": 5.58, "learning_rate": 1.3280648236415633e-06, "loss": 2.1356, "step": 29281 }, { "epoch": 5.58, "learning_rate": 1.3280076263107723e-06, "loss": 1.7123, "step": 29282 }, { "epoch": 5.58, "learning_rate": 1.3279504289799811e-06, "loss": 2.6503, "step": 29283 }, { "epoch": 5.58, "learning_rate": 1.3278932316491897e-06, "loss": 1.5509, "step": 29284 }, { "epoch": 5.58, "learning_rate": 1.3278360343183985e-06, "loss": 1.3168, "step": 29285 }, { "epoch": 5.58, "learning_rate": 1.3277788369876073e-06, "loss": 0.9033, "step": 29286 }, { "epoch": 5.58, "learning_rate": 1.327721639656816e-06, "loss": 1.7956, "step": 29287 }, { "epoch": 5.58, "learning_rate": 1.3276644423260249e-06, "loss": 1.7049, "step": 29288 }, { "epoch": 5.58, "learning_rate": 1.3276072449952337e-06, "loss": 2.1151, "step": 29289 }, { "epoch": 5.58, "learning_rate": 1.3275500476644425e-06, "loss": 1.067, "step": 29290 }, { "epoch": 5.58, "learning_rate": 1.327492850333651e-06, "loss": 2.4333, "step": 29291 }, { "epoch": 5.58, "learning_rate": 1.3274356530028598e-06, "loss": 0.9928, "step": 29292 }, { "epoch": 5.58, "learning_rate": 1.3273784556720686e-06, "loss": 2.3822, "step": 29293 }, { "epoch": 5.58, "learning_rate": 1.3273212583412774e-06, "loss": 1.2936, "step": 29294 }, { "epoch": 5.58, "learning_rate": 1.3272640610104862e-06, "loss": 1.9561, "step": 29295 }, { "epoch": 5.58, "learning_rate": 1.327206863679695e-06, "loss": 1.3885, "step": 29296 }, { "epoch": 5.58, "learning_rate": 1.3271496663489038e-06, "loss": 1.6805, "step": 29297 }, { "epoch": 5.58, "learning_rate": 1.3270924690181124e-06, "loss": 2.2312, "step": 29298 }, { "epoch": 5.58, "learning_rate": 1.3270352716873214e-06, "loss": 1.3426, "step": 29299 }, { "epoch": 5.58, "learning_rate": 1.3269780743565302e-06, "loss": 2.7388, "step": 29300 }, { "epoch": 5.58, "learning_rate": 1.3269208770257388e-06, "loss": 1.6081, "step": 29301 }, { "epoch": 5.58, "learning_rate": 1.3268636796949476e-06, "loss": 1.6287, "step": 29302 }, { "epoch": 5.58, "learning_rate": 1.3268064823641564e-06, "loss": 3.2093, "step": 29303 }, { "epoch": 5.58, "learning_rate": 1.3267492850333651e-06, "loss": 1.5513, "step": 29304 }, { "epoch": 5.58, "learning_rate": 1.326692087702574e-06, "loss": 1.3131, "step": 29305 }, { "epoch": 5.58, "learning_rate": 1.3266348903717827e-06, "loss": 1.4828, "step": 29306 }, { "epoch": 5.58, "learning_rate": 1.3265776930409915e-06, "loss": 2.1358, "step": 29307 }, { "epoch": 5.58, "learning_rate": 1.3265204957102001e-06, "loss": 1.8981, "step": 29308 }, { "epoch": 5.58, "learning_rate": 1.326463298379409e-06, "loss": 1.4511, "step": 29309 }, { "epoch": 5.58, "learning_rate": 1.326406101048618e-06, "loss": 1.3308, "step": 29310 }, { "epoch": 5.58, "learning_rate": 1.3263489037178265e-06, "loss": 2.7551, "step": 29311 }, { "epoch": 5.58, "learning_rate": 1.3262917063870353e-06, "loss": 0.9482, "step": 29312 }, { "epoch": 5.58, "learning_rate": 1.326234509056244e-06, "loss": 2.3518, "step": 29313 }, { "epoch": 5.58, "learning_rate": 1.3261773117254529e-06, "loss": 1.4453, "step": 29314 }, { "epoch": 5.58, "learning_rate": 1.3261201143946617e-06, "loss": 2.907, "step": 29315 }, { "epoch": 5.58, "learning_rate": 1.3260629170638705e-06, "loss": 0.6767, "step": 29316 }, { "epoch": 5.58, "learning_rate": 1.3260057197330793e-06, "loss": 1.0447, "step": 29317 }, { "epoch": 5.58, "learning_rate": 1.3259485224022878e-06, "loss": 2.1674, "step": 29318 }, { "epoch": 5.58, "learning_rate": 1.3258913250714966e-06, "loss": 1.0988, "step": 29319 }, { "epoch": 5.58, "learning_rate": 1.3258341277407054e-06, "loss": 3.1033, "step": 29320 }, { "epoch": 5.58, "learning_rate": 1.3257769304099142e-06, "loss": 1.3979, "step": 29321 }, { "epoch": 5.59, "learning_rate": 1.325719733079123e-06, "loss": 2.9335, "step": 29322 }, { "epoch": 5.59, "learning_rate": 1.3256625357483318e-06, "loss": 1.6113, "step": 29323 }, { "epoch": 5.59, "learning_rate": 1.3256053384175406e-06, "loss": 0.7539, "step": 29324 }, { "epoch": 5.59, "learning_rate": 1.3255481410867492e-06, "loss": 0.9844, "step": 29325 }, { "epoch": 5.59, "learning_rate": 1.3254909437559582e-06, "loss": 1.7648, "step": 29326 }, { "epoch": 5.59, "learning_rate": 1.325433746425167e-06, "loss": 2.3328, "step": 29327 }, { "epoch": 5.59, "learning_rate": 1.3253765490943756e-06, "loss": 1.7444, "step": 29328 }, { "epoch": 5.59, "learning_rate": 1.3253193517635844e-06, "loss": 2.5673, "step": 29329 }, { "epoch": 5.59, "learning_rate": 1.3252621544327932e-06, "loss": 1.4825, "step": 29330 }, { "epoch": 5.59, "learning_rate": 1.325204957102002e-06, "loss": 0.9345, "step": 29331 }, { "epoch": 5.59, "learning_rate": 1.3251477597712107e-06, "loss": 1.9195, "step": 29332 }, { "epoch": 5.59, "learning_rate": 1.3250905624404195e-06, "loss": 2.1052, "step": 29333 }, { "epoch": 5.59, "learning_rate": 1.3250333651096283e-06, "loss": 2.2714, "step": 29334 }, { "epoch": 5.59, "learning_rate": 1.324976167778837e-06, "loss": 1.518, "step": 29335 }, { "epoch": 5.59, "learning_rate": 1.3249189704480457e-06, "loss": 2.4714, "step": 29336 }, { "epoch": 5.59, "learning_rate": 1.3248617731172545e-06, "loss": 2.1816, "step": 29337 }, { "epoch": 5.59, "learning_rate": 1.3248045757864635e-06, "loss": 1.5377, "step": 29338 }, { "epoch": 5.59, "learning_rate": 1.324747378455672e-06, "loss": 2.1394, "step": 29339 }, { "epoch": 5.59, "learning_rate": 1.3246901811248809e-06, "loss": 1.5484, "step": 29340 }, { "epoch": 5.59, "learning_rate": 1.3246329837940897e-06, "loss": 0.6964, "step": 29341 }, { "epoch": 5.59, "learning_rate": 1.3245757864632983e-06, "loss": 2.4446, "step": 29342 }, { "epoch": 5.59, "learning_rate": 1.3245185891325073e-06, "loss": 1.9236, "step": 29343 }, { "epoch": 5.59, "learning_rate": 1.324461391801716e-06, "loss": 2.5116, "step": 29344 }, { "epoch": 5.59, "learning_rate": 1.3244041944709246e-06, "loss": 2.1514, "step": 29345 }, { "epoch": 5.59, "learning_rate": 1.3243469971401334e-06, "loss": 2.1775, "step": 29346 }, { "epoch": 5.59, "learning_rate": 1.3242897998093422e-06, "loss": 1.817, "step": 29347 }, { "epoch": 5.59, "learning_rate": 1.324232602478551e-06, "loss": 1.4673, "step": 29348 }, { "epoch": 5.59, "learning_rate": 1.3241754051477598e-06, "loss": 1.509, "step": 29349 }, { "epoch": 5.59, "learning_rate": 1.3241182078169686e-06, "loss": 1.3144, "step": 29350 }, { "epoch": 5.59, "learning_rate": 1.3240610104861774e-06, "loss": 1.2438, "step": 29351 }, { "epoch": 5.59, "learning_rate": 1.324003813155386e-06, "loss": 3.6047, "step": 29352 }, { "epoch": 5.59, "learning_rate": 1.3239466158245948e-06, "loss": 2.0775, "step": 29353 }, { "epoch": 5.59, "learning_rate": 1.3238894184938038e-06, "loss": 1.6251, "step": 29354 }, { "epoch": 5.59, "learning_rate": 1.3238322211630126e-06, "loss": 1.8152, "step": 29355 }, { "epoch": 5.59, "learning_rate": 1.3237750238322212e-06, "loss": 1.0897, "step": 29356 }, { "epoch": 5.59, "learning_rate": 1.32371782650143e-06, "loss": 2.3007, "step": 29357 }, { "epoch": 5.59, "learning_rate": 1.3236606291706387e-06, "loss": 1.6474, "step": 29358 }, { "epoch": 5.59, "learning_rate": 1.3236034318398475e-06, "loss": 1.4308, "step": 29359 }, { "epoch": 5.59, "learning_rate": 1.3235462345090563e-06, "loss": 1.4949, "step": 29360 }, { "epoch": 5.59, "learning_rate": 1.3234890371782651e-06, "loss": 1.469, "step": 29361 }, { "epoch": 5.59, "learning_rate": 1.323431839847474e-06, "loss": 2.39, "step": 29362 }, { "epoch": 5.59, "learning_rate": 1.3233746425166825e-06, "loss": 2.368, "step": 29363 }, { "epoch": 5.59, "learning_rate": 1.3233174451858913e-06, "loss": 2.3721, "step": 29364 }, { "epoch": 5.59, "learning_rate": 1.3232602478551e-06, "loss": 1.5957, "step": 29365 }, { "epoch": 5.59, "learning_rate": 1.3232030505243089e-06, "loss": 2.4503, "step": 29366 }, { "epoch": 5.59, "learning_rate": 1.3231458531935177e-06, "loss": 1.0695, "step": 29367 }, { "epoch": 5.59, "learning_rate": 1.3230886558627265e-06, "loss": 1.5696, "step": 29368 }, { "epoch": 5.59, "learning_rate": 1.323031458531935e-06, "loss": 2.0341, "step": 29369 }, { "epoch": 5.59, "learning_rate": 1.3229742612011438e-06, "loss": 2.7125, "step": 29370 }, { "epoch": 5.59, "learning_rate": 1.3229170638703529e-06, "loss": 2.3343, "step": 29371 }, { "epoch": 5.59, "learning_rate": 1.3228598665395616e-06, "loss": 1.6638, "step": 29372 }, { "epoch": 5.59, "learning_rate": 1.3228026692087702e-06, "loss": 2.5904, "step": 29373 }, { "epoch": 5.6, "learning_rate": 1.322745471877979e-06, "loss": 1.4769, "step": 29374 }, { "epoch": 5.6, "learning_rate": 1.3226882745471878e-06, "loss": 1.7882, "step": 29375 }, { "epoch": 5.6, "learning_rate": 1.3226310772163966e-06, "loss": 1.7291, "step": 29376 }, { "epoch": 5.6, "learning_rate": 1.3225738798856054e-06, "loss": 1.5277, "step": 29377 }, { "epoch": 5.6, "learning_rate": 1.3225166825548142e-06, "loss": 2.0622, "step": 29378 }, { "epoch": 5.6, "learning_rate": 1.322459485224023e-06, "loss": 2.6919, "step": 29379 }, { "epoch": 5.6, "learning_rate": 1.3224022878932316e-06, "loss": 2.9172, "step": 29380 }, { "epoch": 5.6, "learning_rate": 1.3223450905624404e-06, "loss": 1.4996, "step": 29381 }, { "epoch": 5.6, "learning_rate": 1.3222878932316494e-06, "loss": 2.5763, "step": 29382 }, { "epoch": 5.6, "learning_rate": 1.322230695900858e-06, "loss": 1.7284, "step": 29383 }, { "epoch": 5.6, "learning_rate": 1.3221734985700667e-06, "loss": 1.1262, "step": 29384 }, { "epoch": 5.6, "learning_rate": 1.3221163012392755e-06, "loss": 1.5628, "step": 29385 }, { "epoch": 5.6, "learning_rate": 1.3220591039084841e-06, "loss": 0.8509, "step": 29386 }, { "epoch": 5.6, "learning_rate": 1.3220019065776931e-06, "loss": 1.9264, "step": 29387 }, { "epoch": 5.6, "learning_rate": 1.321944709246902e-06, "loss": 2.2201, "step": 29388 }, { "epoch": 5.6, "learning_rate": 1.3218875119161107e-06, "loss": 1.9562, "step": 29389 }, { "epoch": 5.6, "learning_rate": 1.3218303145853193e-06, "loss": 1.8908, "step": 29390 }, { "epoch": 5.6, "learning_rate": 1.321773117254528e-06, "loss": 0.687, "step": 29391 }, { "epoch": 5.6, "learning_rate": 1.3217159199237369e-06, "loss": 1.8591, "step": 29392 }, { "epoch": 5.6, "learning_rate": 1.3216587225929457e-06, "loss": 1.5994, "step": 29393 }, { "epoch": 5.6, "learning_rate": 1.3216015252621545e-06, "loss": 0.6423, "step": 29394 }, { "epoch": 5.6, "learning_rate": 1.3215443279313633e-06, "loss": 1.3076, "step": 29395 }, { "epoch": 5.6, "learning_rate": 1.321487130600572e-06, "loss": 3.4432, "step": 29396 }, { "epoch": 5.6, "learning_rate": 1.3214299332697806e-06, "loss": 1.4738, "step": 29397 }, { "epoch": 5.6, "learning_rate": 1.3213727359389897e-06, "loss": 1.4851, "step": 29398 }, { "epoch": 5.6, "learning_rate": 1.3213155386081984e-06, "loss": 1.1062, "step": 29399 }, { "epoch": 5.6, "learning_rate": 1.321258341277407e-06, "loss": 1.4703, "step": 29400 }, { "epoch": 5.6, "learning_rate": 1.3212011439466158e-06, "loss": 1.6098, "step": 29401 }, { "epoch": 5.6, "learning_rate": 1.3211439466158246e-06, "loss": 2.1435, "step": 29402 }, { "epoch": 5.6, "learning_rate": 1.3210867492850334e-06, "loss": 1.6298, "step": 29403 }, { "epoch": 5.6, "learning_rate": 1.3210295519542422e-06, "loss": 1.7971, "step": 29404 }, { "epoch": 5.6, "learning_rate": 1.320972354623451e-06, "loss": 1.6938, "step": 29405 }, { "epoch": 5.6, "learning_rate": 1.3209151572926598e-06, "loss": 1.8732, "step": 29406 }, { "epoch": 5.6, "learning_rate": 1.3208579599618684e-06, "loss": 1.7382, "step": 29407 }, { "epoch": 5.6, "learning_rate": 1.3208007626310772e-06, "loss": 2.2443, "step": 29408 }, { "epoch": 5.6, "learning_rate": 1.320743565300286e-06, "loss": 2.1426, "step": 29409 }, { "epoch": 5.6, "learning_rate": 1.3206863679694948e-06, "loss": 1.8722, "step": 29410 }, { "epoch": 5.6, "learning_rate": 1.3206291706387035e-06, "loss": 2.2314, "step": 29411 }, { "epoch": 5.6, "learning_rate": 1.3205719733079123e-06, "loss": 1.4457, "step": 29412 }, { "epoch": 5.6, "learning_rate": 1.3205147759771211e-06, "loss": 1.8803, "step": 29413 }, { "epoch": 5.6, "learning_rate": 1.3204575786463297e-06, "loss": 3.461, "step": 29414 }, { "epoch": 5.6, "learning_rate": 1.3204003813155387e-06, "loss": 1.8714, "step": 29415 }, { "epoch": 5.6, "learning_rate": 1.3203431839847475e-06, "loss": 3.9214, "step": 29416 }, { "epoch": 5.6, "learning_rate": 1.320285986653956e-06, "loss": 1.5532, "step": 29417 }, { "epoch": 5.6, "learning_rate": 1.3202287893231649e-06, "loss": 1.896, "step": 29418 }, { "epoch": 5.6, "learning_rate": 1.3201715919923737e-06, "loss": 3.032, "step": 29419 }, { "epoch": 5.6, "learning_rate": 1.3201143946615825e-06, "loss": 1.6203, "step": 29420 }, { "epoch": 5.6, "learning_rate": 1.3200571973307913e-06, "loss": 1.9416, "step": 29421 }, { "epoch": 5.6, "learning_rate": 1.32e-06, "loss": 2.0701, "step": 29422 }, { "epoch": 5.6, "learning_rate": 1.3199428026692089e-06, "loss": 1.6526, "step": 29423 }, { "epoch": 5.6, "learning_rate": 1.3198856053384174e-06, "loss": 2.7864, "step": 29424 }, { "epoch": 5.6, "learning_rate": 1.3198284080076262e-06, "loss": 2.5682, "step": 29425 }, { "epoch": 5.6, "learning_rate": 1.3197712106768352e-06, "loss": 1.3803, "step": 29426 }, { "epoch": 5.61, "learning_rate": 1.3197140133460438e-06, "loss": 0.9719, "step": 29427 }, { "epoch": 5.61, "learning_rate": 1.3196568160152526e-06, "loss": 2.374, "step": 29428 }, { "epoch": 5.61, "learning_rate": 1.3195996186844614e-06, "loss": 1.8988, "step": 29429 }, { "epoch": 5.61, "learning_rate": 1.3195424213536702e-06, "loss": 2.2201, "step": 29430 }, { "epoch": 5.61, "learning_rate": 1.319485224022879e-06, "loss": 3.6785, "step": 29431 }, { "epoch": 5.61, "learning_rate": 1.3194280266920878e-06, "loss": 1.7086, "step": 29432 }, { "epoch": 5.61, "learning_rate": 1.3193708293612966e-06, "loss": 0.8724, "step": 29433 }, { "epoch": 5.61, "learning_rate": 1.3193136320305052e-06, "loss": 1.8604, "step": 29434 }, { "epoch": 5.61, "learning_rate": 1.319256434699714e-06, "loss": 2.7497, "step": 29435 }, { "epoch": 5.61, "learning_rate": 1.3191992373689228e-06, "loss": 1.8779, "step": 29436 }, { "epoch": 5.61, "learning_rate": 1.3191420400381318e-06, "loss": 2.2834, "step": 29437 }, { "epoch": 5.61, "learning_rate": 1.3190848427073403e-06, "loss": 2.0554, "step": 29438 }, { "epoch": 5.61, "learning_rate": 1.3190276453765491e-06, "loss": 1.8617, "step": 29439 }, { "epoch": 5.61, "learning_rate": 1.318970448045758e-06, "loss": 1.6867, "step": 29440 }, { "epoch": 5.61, "learning_rate": 1.3189132507149665e-06, "loss": 2.2903, "step": 29441 }, { "epoch": 5.61, "learning_rate": 1.3188560533841755e-06, "loss": 0.8353, "step": 29442 }, { "epoch": 5.61, "learning_rate": 1.3187988560533843e-06, "loss": 2.6148, "step": 29443 }, { "epoch": 5.61, "learning_rate": 1.3187416587225931e-06, "loss": 1.734, "step": 29444 }, { "epoch": 5.61, "learning_rate": 1.3186844613918017e-06, "loss": 1.1086, "step": 29445 }, { "epoch": 5.61, "learning_rate": 1.3186272640610105e-06, "loss": 2.0756, "step": 29446 }, { "epoch": 5.61, "learning_rate": 1.3185700667302193e-06, "loss": 0.8503, "step": 29447 }, { "epoch": 5.61, "learning_rate": 1.318512869399428e-06, "loss": 1.5234, "step": 29448 }, { "epoch": 5.61, "learning_rate": 1.3184556720686369e-06, "loss": 1.8372, "step": 29449 }, { "epoch": 5.61, "learning_rate": 1.3183984747378457e-06, "loss": 1.0597, "step": 29450 }, { "epoch": 5.61, "learning_rate": 1.3183412774070542e-06, "loss": 2.8426, "step": 29451 }, { "epoch": 5.61, "learning_rate": 1.318284080076263e-06, "loss": 2.3953, "step": 29452 }, { "epoch": 5.61, "learning_rate": 1.3182268827454718e-06, "loss": 3.0891, "step": 29453 }, { "epoch": 5.61, "learning_rate": 1.3181696854146808e-06, "loss": 2.1725, "step": 29454 }, { "epoch": 5.61, "learning_rate": 1.3181124880838894e-06, "loss": 2.9078, "step": 29455 }, { "epoch": 5.61, "learning_rate": 1.3180552907530982e-06, "loss": 2.4212, "step": 29456 }, { "epoch": 5.61, "learning_rate": 1.317998093422307e-06, "loss": 2.2517, "step": 29457 }, { "epoch": 5.61, "learning_rate": 1.3179408960915156e-06, "loss": 2.3261, "step": 29458 }, { "epoch": 5.61, "learning_rate": 1.3178836987607246e-06, "loss": 1.6801, "step": 29459 }, { "epoch": 5.61, "learning_rate": 1.3178265014299334e-06, "loss": 2.618, "step": 29460 }, { "epoch": 5.61, "learning_rate": 1.3177693040991422e-06, "loss": 0.6624, "step": 29461 }, { "epoch": 5.61, "learning_rate": 1.3177121067683508e-06, "loss": 1.0971, "step": 29462 }, { "epoch": 5.61, "learning_rate": 1.3176549094375596e-06, "loss": 2.732, "step": 29463 }, { "epoch": 5.61, "learning_rate": 1.3175977121067683e-06, "loss": 1.9979, "step": 29464 }, { "epoch": 5.61, "learning_rate": 1.3175405147759771e-06, "loss": 1.9913, "step": 29465 }, { "epoch": 5.61, "learning_rate": 1.317483317445186e-06, "loss": 3.1202, "step": 29466 }, { "epoch": 5.61, "learning_rate": 1.3174261201143947e-06, "loss": 1.6407, "step": 29467 }, { "epoch": 5.61, "learning_rate": 1.3173689227836035e-06, "loss": 2.7832, "step": 29468 }, { "epoch": 5.61, "learning_rate": 1.3173117254528121e-06, "loss": 2.1976, "step": 29469 }, { "epoch": 5.61, "learning_rate": 1.3172545281220211e-06, "loss": 1.9536, "step": 29470 }, { "epoch": 5.61, "learning_rate": 1.31719733079123e-06, "loss": 3.0414, "step": 29471 }, { "epoch": 5.61, "learning_rate": 1.3171401334604385e-06, "loss": 0.7356, "step": 29472 }, { "epoch": 5.61, "learning_rate": 1.3170829361296473e-06, "loss": 2.4811, "step": 29473 }, { "epoch": 5.61, "learning_rate": 1.317025738798856e-06, "loss": 3.3999, "step": 29474 }, { "epoch": 5.61, "learning_rate": 1.3169685414680649e-06, "loss": 2.0111, "step": 29475 }, { "epoch": 5.61, "learning_rate": 1.3169113441372737e-06, "loss": 1.9748, "step": 29476 }, { "epoch": 5.61, "learning_rate": 1.3168541468064825e-06, "loss": 1.9805, "step": 29477 }, { "epoch": 5.61, "learning_rate": 1.3167969494756913e-06, "loss": 2.5329, "step": 29478 }, { "epoch": 5.62, "learning_rate": 1.3167397521448998e-06, "loss": 2.6786, "step": 29479 }, { "epoch": 5.62, "learning_rate": 1.3166825548141086e-06, "loss": 1.1601, "step": 29480 }, { "epoch": 5.62, "learning_rate": 1.3166253574833176e-06, "loss": 1.3862, "step": 29481 }, { "epoch": 5.62, "learning_rate": 1.3165681601525262e-06, "loss": 2.9284, "step": 29482 }, { "epoch": 5.62, "learning_rate": 1.316510962821735e-06, "loss": 1.598, "step": 29483 }, { "epoch": 5.62, "learning_rate": 1.3164537654909438e-06, "loss": 1.8724, "step": 29484 }, { "epoch": 5.62, "learning_rate": 1.3163965681601526e-06, "loss": 2.3832, "step": 29485 }, { "epoch": 5.62, "learning_rate": 1.3163393708293614e-06, "loss": 1.9944, "step": 29486 }, { "epoch": 5.62, "learning_rate": 1.3162821734985702e-06, "loss": 2.7019, "step": 29487 }, { "epoch": 5.62, "learning_rate": 1.316224976167779e-06, "loss": 1.0715, "step": 29488 }, { "epoch": 5.62, "learning_rate": 1.3161677788369876e-06, "loss": 0.8243, "step": 29489 }, { "epoch": 5.62, "learning_rate": 1.3161105815061964e-06, "loss": 2.5661, "step": 29490 }, { "epoch": 5.62, "learning_rate": 1.3160533841754051e-06, "loss": 2.3195, "step": 29491 }, { "epoch": 5.62, "learning_rate": 1.315996186844614e-06, "loss": 2.1806, "step": 29492 }, { "epoch": 5.62, "learning_rate": 1.3159389895138227e-06, "loss": 2.5881, "step": 29493 }, { "epoch": 5.62, "learning_rate": 1.3158817921830315e-06, "loss": 1.7793, "step": 29494 }, { "epoch": 5.62, "learning_rate": 1.3158245948522403e-06, "loss": 2.8184, "step": 29495 }, { "epoch": 5.62, "learning_rate": 1.315767397521449e-06, "loss": 0.9734, "step": 29496 }, { "epoch": 5.62, "learning_rate": 1.3157102001906577e-06, "loss": 0.6931, "step": 29497 }, { "epoch": 5.62, "learning_rate": 1.3156530028598667e-06, "loss": 2.7788, "step": 29498 }, { "epoch": 5.62, "learning_rate": 1.3155958055290753e-06, "loss": 2.3637, "step": 29499 }, { "epoch": 5.62, "learning_rate": 1.315538608198284e-06, "loss": 1.4782, "step": 29500 }, { "epoch": 5.62, "learning_rate": 1.3154814108674929e-06, "loss": 2.5682, "step": 29501 }, { "epoch": 5.62, "learning_rate": 1.3154242135367017e-06, "loss": 2.8661, "step": 29502 }, { "epoch": 5.62, "learning_rate": 1.3153670162059105e-06, "loss": 1.1651, "step": 29503 }, { "epoch": 5.62, "learning_rate": 1.3153098188751193e-06, "loss": 1.1658, "step": 29504 }, { "epoch": 5.62, "learning_rate": 1.315252621544328e-06, "loss": 2.5709, "step": 29505 }, { "epoch": 5.62, "learning_rate": 1.3151954242135366e-06, "loss": 1.5979, "step": 29506 }, { "epoch": 5.62, "learning_rate": 1.3151382268827454e-06, "loss": 3.2104, "step": 29507 }, { "epoch": 5.62, "learning_rate": 1.3150810295519542e-06, "loss": 1.1853, "step": 29508 }, { "epoch": 5.62, "learning_rate": 1.3150238322211632e-06, "loss": 0.8609, "step": 29509 }, { "epoch": 5.62, "learning_rate": 1.3149666348903718e-06, "loss": 1.6949, "step": 29510 }, { "epoch": 5.62, "learning_rate": 1.3149094375595806e-06, "loss": 2.5682, "step": 29511 }, { "epoch": 5.62, "learning_rate": 1.3148522402287894e-06, "loss": 2.5619, "step": 29512 }, { "epoch": 5.62, "learning_rate": 1.314795042897998e-06, "loss": 1.7707, "step": 29513 }, { "epoch": 5.62, "learning_rate": 1.314737845567207e-06, "loss": 2.0986, "step": 29514 }, { "epoch": 5.62, "learning_rate": 1.3146806482364158e-06, "loss": 1.3847, "step": 29515 }, { "epoch": 5.62, "learning_rate": 1.3146234509056244e-06, "loss": 1.1927, "step": 29516 }, { "epoch": 5.62, "learning_rate": 1.3145662535748332e-06, "loss": 1.1245, "step": 29517 }, { "epoch": 5.62, "learning_rate": 1.314509056244042e-06, "loss": 2.2047, "step": 29518 }, { "epoch": 5.62, "learning_rate": 1.3144518589132507e-06, "loss": 1.2843, "step": 29519 }, { "epoch": 5.62, "learning_rate": 1.3143946615824595e-06, "loss": 1.8603, "step": 29520 }, { "epoch": 5.62, "learning_rate": 1.3143374642516683e-06, "loss": 1.4764, "step": 29521 }, { "epoch": 5.62, "learning_rate": 1.3142802669208771e-06, "loss": 2.8594, "step": 29522 }, { "epoch": 5.62, "learning_rate": 1.3142230695900857e-06, "loss": 1.0891, "step": 29523 }, { "epoch": 5.62, "learning_rate": 1.3141658722592945e-06, "loss": 1.4267, "step": 29524 }, { "epoch": 5.62, "learning_rate": 1.3141086749285033e-06, "loss": 1.4859, "step": 29525 }, { "epoch": 5.62, "learning_rate": 1.3140514775977123e-06, "loss": 2.4681, "step": 29526 }, { "epoch": 5.62, "learning_rate": 1.3139942802669209e-06, "loss": 1.5533, "step": 29527 }, { "epoch": 5.62, "learning_rate": 1.3139370829361297e-06, "loss": 1.4225, "step": 29528 }, { "epoch": 5.62, "learning_rate": 1.3138798856053385e-06, "loss": 1.746, "step": 29529 }, { "epoch": 5.62, "learning_rate": 1.3138226882745473e-06, "loss": 1.5334, "step": 29530 }, { "epoch": 5.62, "learning_rate": 1.313765490943756e-06, "loss": 1.4004, "step": 29531 }, { "epoch": 5.63, "learning_rate": 1.3137082936129648e-06, "loss": 1.625, "step": 29532 }, { "epoch": 5.63, "learning_rate": 1.3136510962821736e-06, "loss": 1.1974, "step": 29533 }, { "epoch": 5.63, "learning_rate": 1.3135938989513822e-06, "loss": 0.829, "step": 29534 }, { "epoch": 5.63, "learning_rate": 1.313536701620591e-06, "loss": 1.7878, "step": 29535 }, { "epoch": 5.63, "learning_rate": 1.3134795042897998e-06, "loss": 1.7435, "step": 29536 }, { "epoch": 5.63, "learning_rate": 1.3134223069590086e-06, "loss": 0.6732, "step": 29537 }, { "epoch": 5.63, "learning_rate": 1.3133651096282174e-06, "loss": 0.3595, "step": 29538 }, { "epoch": 5.63, "learning_rate": 1.3133079122974262e-06, "loss": 1.7377, "step": 29539 }, { "epoch": 5.63, "learning_rate": 1.3132507149666348e-06, "loss": 1.7236, "step": 29540 }, { "epoch": 5.63, "learning_rate": 1.3131935176358436e-06, "loss": 2.3124, "step": 29541 }, { "epoch": 5.63, "learning_rate": 1.3131363203050526e-06, "loss": 2.0446, "step": 29542 }, { "epoch": 5.63, "learning_rate": 1.3130791229742614e-06, "loss": 2.313, "step": 29543 }, { "epoch": 5.63, "learning_rate": 1.31302192564347e-06, "loss": 2.298, "step": 29544 }, { "epoch": 5.63, "learning_rate": 1.3129647283126787e-06, "loss": 0.8517, "step": 29545 }, { "epoch": 5.63, "learning_rate": 1.3129075309818875e-06, "loss": 0.7236, "step": 29546 }, { "epoch": 5.63, "learning_rate": 1.3128503336510963e-06, "loss": 1.1331, "step": 29547 }, { "epoch": 5.63, "learning_rate": 1.3127931363203051e-06, "loss": 1.0009, "step": 29548 }, { "epoch": 5.63, "learning_rate": 1.312735938989514e-06, "loss": 1.8769, "step": 29549 }, { "epoch": 5.63, "learning_rate": 1.3126787416587227e-06, "loss": 1.9047, "step": 29550 }, { "epoch": 5.63, "learning_rate": 1.3126215443279313e-06, "loss": 2.1049, "step": 29551 }, { "epoch": 5.63, "learning_rate": 1.31256434699714e-06, "loss": 2.2866, "step": 29552 }, { "epoch": 5.63, "learning_rate": 1.312507149666349e-06, "loss": 1.2522, "step": 29553 }, { "epoch": 5.63, "learning_rate": 1.3124499523355577e-06, "loss": 1.3714, "step": 29554 }, { "epoch": 5.63, "learning_rate": 1.3123927550047665e-06, "loss": 2.2427, "step": 29555 }, { "epoch": 5.63, "learning_rate": 1.3123355576739753e-06, "loss": 2.0193, "step": 29556 }, { "epoch": 5.63, "learning_rate": 1.3122783603431838e-06, "loss": 2.2034, "step": 29557 }, { "epoch": 5.63, "learning_rate": 1.3122211630123929e-06, "loss": 1.0999, "step": 29558 }, { "epoch": 5.63, "learning_rate": 1.3121639656816016e-06, "loss": 2.8097, "step": 29559 }, { "epoch": 5.63, "learning_rate": 1.3121067683508104e-06, "loss": 1.6587, "step": 29560 }, { "epoch": 5.63, "learning_rate": 1.312049571020019e-06, "loss": 0.9517, "step": 29561 }, { "epoch": 5.63, "learning_rate": 1.3119923736892278e-06, "loss": 1.9741, "step": 29562 }, { "epoch": 5.63, "learning_rate": 1.3119351763584366e-06, "loss": 1.2091, "step": 29563 }, { "epoch": 5.63, "learning_rate": 1.3118779790276454e-06, "loss": 1.9005, "step": 29564 }, { "epoch": 5.63, "learning_rate": 1.3118207816968542e-06, "loss": 3.529, "step": 29565 }, { "epoch": 5.63, "learning_rate": 1.311763584366063e-06, "loss": 1.2254, "step": 29566 }, { "epoch": 5.63, "learning_rate": 1.3117063870352718e-06, "loss": 2.9349, "step": 29567 }, { "epoch": 5.63, "learning_rate": 1.3116491897044804e-06, "loss": 1.3858, "step": 29568 }, { "epoch": 5.63, "learning_rate": 1.3115919923736892e-06, "loss": 1.59, "step": 29569 }, { "epoch": 5.63, "learning_rate": 1.3115347950428982e-06, "loss": 2.2224, "step": 29570 }, { "epoch": 5.63, "learning_rate": 1.3114775977121067e-06, "loss": 1.9665, "step": 29571 }, { "epoch": 5.63, "learning_rate": 1.3114204003813155e-06, "loss": 3.3984, "step": 29572 }, { "epoch": 5.63, "learning_rate": 1.3113632030505243e-06, "loss": 2.6371, "step": 29573 }, { "epoch": 5.63, "learning_rate": 1.3113060057197331e-06, "loss": 2.5839, "step": 29574 }, { "epoch": 5.63, "learning_rate": 1.311248808388942e-06, "loss": 1.6152, "step": 29575 }, { "epoch": 5.63, "learning_rate": 1.3111916110581507e-06, "loss": 2.1116, "step": 29576 }, { "epoch": 5.63, "learning_rate": 1.3111344137273595e-06, "loss": 1.6132, "step": 29577 }, { "epoch": 5.63, "learning_rate": 1.311077216396568e-06, "loss": 1.2793, "step": 29578 }, { "epoch": 5.63, "learning_rate": 1.3110200190657769e-06, "loss": 2.1018, "step": 29579 }, { "epoch": 5.63, "learning_rate": 1.3109628217349857e-06, "loss": 1.9058, "step": 29580 }, { "epoch": 5.63, "learning_rate": 1.3109056244041945e-06, "loss": 2.8654, "step": 29581 }, { "epoch": 5.63, "learning_rate": 1.3108484270734033e-06, "loss": 0.9908, "step": 29582 }, { "epoch": 5.63, "learning_rate": 1.310791229742612e-06, "loss": 2.4815, "step": 29583 }, { "epoch": 5.64, "learning_rate": 1.3107340324118209e-06, "loss": 2.5433, "step": 29584 }, { "epoch": 5.64, "learning_rate": 1.3106768350810294e-06, "loss": 2.3734, "step": 29585 }, { "epoch": 5.64, "learning_rate": 1.3106196377502384e-06, "loss": 1.0126, "step": 29586 }, { "epoch": 5.64, "learning_rate": 1.3105624404194472e-06, "loss": 1.3677, "step": 29587 }, { "epoch": 5.64, "learning_rate": 1.3105052430886558e-06, "loss": 1.4652, "step": 29588 }, { "epoch": 5.64, "learning_rate": 1.3104480457578646e-06, "loss": 0.8505, "step": 29589 }, { "epoch": 5.64, "learning_rate": 1.3103908484270734e-06, "loss": 1.3295, "step": 29590 }, { "epoch": 5.64, "learning_rate": 1.3103336510962822e-06, "loss": 2.0386, "step": 29591 }, { "epoch": 5.64, "learning_rate": 1.310276453765491e-06, "loss": 1.5497, "step": 29592 }, { "epoch": 5.64, "learning_rate": 1.3102192564346998e-06, "loss": 1.2984, "step": 29593 }, { "epoch": 5.64, "learning_rate": 1.3101620591039086e-06, "loss": 3.6787, "step": 29594 }, { "epoch": 5.64, "learning_rate": 1.3101048617731172e-06, "loss": 1.1257, "step": 29595 }, { "epoch": 5.64, "learning_rate": 1.310047664442326e-06, "loss": 2.8458, "step": 29596 }, { "epoch": 5.64, "learning_rate": 1.309990467111535e-06, "loss": 1.9443, "step": 29597 }, { "epoch": 5.64, "learning_rate": 1.3099332697807438e-06, "loss": 0.8507, "step": 29598 }, { "epoch": 5.64, "learning_rate": 1.3098760724499523e-06, "loss": 2.3094, "step": 29599 }, { "epoch": 5.64, "learning_rate": 1.3098188751191611e-06, "loss": 1.6599, "step": 29600 }, { "epoch": 5.64, "learning_rate": 1.30976167778837e-06, "loss": 2.5325, "step": 29601 }, { "epoch": 5.64, "learning_rate": 1.3097044804575787e-06, "loss": 3.543, "step": 29602 }, { "epoch": 5.64, "learning_rate": 1.3096472831267875e-06, "loss": 2.811, "step": 29603 }, { "epoch": 5.64, "learning_rate": 1.3095900857959963e-06, "loss": 3.8255, "step": 29604 }, { "epoch": 5.64, "learning_rate": 1.3095328884652049e-06, "loss": 2.3893, "step": 29605 }, { "epoch": 5.64, "learning_rate": 1.3094756911344137e-06, "loss": 1.9407, "step": 29606 }, { "epoch": 5.64, "learning_rate": 1.3094184938036225e-06, "loss": 2.1261, "step": 29607 }, { "epoch": 5.64, "learning_rate": 1.3093612964728313e-06, "loss": 2.683, "step": 29608 }, { "epoch": 5.64, "learning_rate": 1.30930409914204e-06, "loss": 2.3609, "step": 29609 }, { "epoch": 5.64, "learning_rate": 1.3092469018112489e-06, "loss": 1.6011, "step": 29610 }, { "epoch": 5.64, "learning_rate": 1.3091897044804577e-06, "loss": 2.2936, "step": 29611 }, { "epoch": 5.64, "learning_rate": 1.3091325071496662e-06, "loss": 1.9402, "step": 29612 }, { "epoch": 5.64, "learning_rate": 1.309075309818875e-06, "loss": 1.2392, "step": 29613 }, { "epoch": 5.64, "learning_rate": 1.309018112488084e-06, "loss": 2.3852, "step": 29614 }, { "epoch": 5.64, "learning_rate": 1.3089609151572928e-06, "loss": 2.0088, "step": 29615 }, { "epoch": 5.64, "learning_rate": 1.3089037178265014e-06, "loss": 1.4701, "step": 29616 }, { "epoch": 5.64, "learning_rate": 1.3088465204957102e-06, "loss": 1.8851, "step": 29617 }, { "epoch": 5.64, "learning_rate": 1.308789323164919e-06, "loss": 1.8838, "step": 29618 }, { "epoch": 5.64, "learning_rate": 1.3087321258341278e-06, "loss": 1.8097, "step": 29619 }, { "epoch": 5.64, "learning_rate": 1.3086749285033366e-06, "loss": 1.2019, "step": 29620 }, { "epoch": 5.64, "learning_rate": 1.3086177311725454e-06, "loss": 2.9605, "step": 29621 }, { "epoch": 5.64, "learning_rate": 1.308560533841754e-06, "loss": 1.7313, "step": 29622 }, { "epoch": 5.64, "learning_rate": 1.3085033365109628e-06, "loss": 2.4434, "step": 29623 }, { "epoch": 5.64, "learning_rate": 1.3084461391801715e-06, "loss": 1.5182, "step": 29624 }, { "epoch": 5.64, "learning_rate": 1.3083889418493806e-06, "loss": 2.0235, "step": 29625 }, { "epoch": 5.64, "learning_rate": 1.3083317445185891e-06, "loss": 1.4506, "step": 29626 }, { "epoch": 5.64, "learning_rate": 1.308274547187798e-06, "loss": 0.8653, "step": 29627 }, { "epoch": 5.64, "learning_rate": 1.3082173498570067e-06, "loss": 2.0178, "step": 29628 }, { "epoch": 5.64, "learning_rate": 1.3081601525262153e-06, "loss": 2.1638, "step": 29629 }, { "epoch": 5.64, "learning_rate": 1.3081029551954243e-06, "loss": 0.9503, "step": 29630 }, { "epoch": 5.64, "learning_rate": 1.308045757864633e-06, "loss": 1.4567, "step": 29631 }, { "epoch": 5.64, "learning_rate": 1.307988560533842e-06, "loss": 1.5079, "step": 29632 }, { "epoch": 5.64, "learning_rate": 1.3079313632030505e-06, "loss": 2.2405, "step": 29633 }, { "epoch": 5.64, "learning_rate": 1.3078741658722593e-06, "loss": 2.6887, "step": 29634 }, { "epoch": 5.64, "learning_rate": 1.307816968541468e-06, "loss": 0.9091, "step": 29635 }, { "epoch": 5.64, "learning_rate": 1.3077597712106769e-06, "loss": 2.0309, "step": 29636 }, { "epoch": 5.65, "learning_rate": 1.3077025738798857e-06, "loss": 1.5534, "step": 29637 }, { "epoch": 5.65, "learning_rate": 1.3076453765490944e-06, "loss": 1.6767, "step": 29638 }, { "epoch": 5.65, "learning_rate": 1.3075881792183032e-06, "loss": 2.7109, "step": 29639 }, { "epoch": 5.65, "learning_rate": 1.3075309818875118e-06, "loss": 1.4326, "step": 29640 }, { "epoch": 5.65, "learning_rate": 1.3074737845567208e-06, "loss": 2.1213, "step": 29641 }, { "epoch": 5.65, "learning_rate": 1.3074165872259296e-06, "loss": 2.0129, "step": 29642 }, { "epoch": 5.65, "learning_rate": 1.3073593898951382e-06, "loss": 1.6425, "step": 29643 }, { "epoch": 5.65, "learning_rate": 1.307302192564347e-06, "loss": 2.987, "step": 29644 }, { "epoch": 5.65, "learning_rate": 1.3072449952335558e-06, "loss": 1.3493, "step": 29645 }, { "epoch": 5.65, "learning_rate": 1.3071877979027646e-06, "loss": 1.1707, "step": 29646 }, { "epoch": 5.65, "learning_rate": 1.3071306005719734e-06, "loss": 1.8553, "step": 29647 }, { "epoch": 5.65, "learning_rate": 1.3070734032411822e-06, "loss": 1.8737, "step": 29648 }, { "epoch": 5.65, "learning_rate": 1.307016205910391e-06, "loss": 1.5483, "step": 29649 }, { "epoch": 5.65, "learning_rate": 1.3069590085795996e-06, "loss": 0.4026, "step": 29650 }, { "epoch": 5.65, "learning_rate": 1.3069018112488083e-06, "loss": 1.1633, "step": 29651 }, { "epoch": 5.65, "learning_rate": 1.3068446139180171e-06, "loss": 1.6319, "step": 29652 }, { "epoch": 5.65, "learning_rate": 1.306787416587226e-06, "loss": 1.2497, "step": 29653 }, { "epoch": 5.65, "learning_rate": 1.3067302192564347e-06, "loss": 2.6656, "step": 29654 }, { "epoch": 5.65, "learning_rate": 1.3066730219256435e-06, "loss": 1.6373, "step": 29655 }, { "epoch": 5.65, "learning_rate": 1.3066158245948523e-06, "loss": 2.4199, "step": 29656 }, { "epoch": 5.65, "learning_rate": 1.306558627264061e-06, "loss": 2.3406, "step": 29657 }, { "epoch": 5.65, "learning_rate": 1.30650142993327e-06, "loss": 1.6561, "step": 29658 }, { "epoch": 5.65, "learning_rate": 1.3064442326024787e-06, "loss": 0.6416, "step": 29659 }, { "epoch": 5.65, "learning_rate": 1.3063870352716873e-06, "loss": 2.9392, "step": 29660 }, { "epoch": 5.65, "learning_rate": 1.306329837940896e-06, "loss": 2.3141, "step": 29661 }, { "epoch": 5.65, "learning_rate": 1.3062726406101049e-06, "loss": 2.0104, "step": 29662 }, { "epoch": 5.65, "learning_rate": 1.3062154432793137e-06, "loss": 2.4087, "step": 29663 }, { "epoch": 5.65, "learning_rate": 1.3061582459485225e-06, "loss": 2.4958, "step": 29664 }, { "epoch": 5.65, "learning_rate": 1.3061010486177312e-06, "loss": 2.9773, "step": 29665 }, { "epoch": 5.65, "learning_rate": 1.30604385128694e-06, "loss": 2.5858, "step": 29666 }, { "epoch": 5.65, "learning_rate": 1.3059866539561486e-06, "loss": 1.9219, "step": 29667 }, { "epoch": 5.65, "learning_rate": 1.3059294566253574e-06, "loss": 2.3538, "step": 29668 }, { "epoch": 5.65, "learning_rate": 1.3058722592945664e-06, "loss": 3.0872, "step": 29669 }, { "epoch": 5.65, "learning_rate": 1.305815061963775e-06, "loss": 0.5759, "step": 29670 }, { "epoch": 5.65, "learning_rate": 1.3057578646329838e-06, "loss": 1.9497, "step": 29671 }, { "epoch": 5.65, "learning_rate": 1.3057006673021926e-06, "loss": 1.8228, "step": 29672 }, { "epoch": 5.65, "learning_rate": 1.3056434699714014e-06, "loss": 1.9898, "step": 29673 }, { "epoch": 5.65, "learning_rate": 1.3055862726406102e-06, "loss": 2.4178, "step": 29674 }, { "epoch": 5.65, "learning_rate": 1.305529075309819e-06, "loss": 1.7774, "step": 29675 }, { "epoch": 5.65, "learning_rate": 1.3054718779790278e-06, "loss": 0.4607, "step": 29676 }, { "epoch": 5.65, "learning_rate": 1.3054146806482363e-06, "loss": 2.8513, "step": 29677 }, { "epoch": 5.65, "learning_rate": 1.3053574833174451e-06, "loss": 1.9308, "step": 29678 }, { "epoch": 5.65, "learning_rate": 1.305300285986654e-06, "loss": 1.8086, "step": 29679 }, { "epoch": 5.65, "learning_rate": 1.305243088655863e-06, "loss": 1.4935, "step": 29680 }, { "epoch": 5.65, "learning_rate": 1.3051858913250715e-06, "loss": 1.6759, "step": 29681 }, { "epoch": 5.65, "learning_rate": 1.3051286939942803e-06, "loss": 1.6527, "step": 29682 }, { "epoch": 5.65, "learning_rate": 1.3050714966634891e-06, "loss": 2.3426, "step": 29683 }, { "epoch": 5.65, "learning_rate": 1.3050142993326977e-06, "loss": 1.2455, "step": 29684 }, { "epoch": 5.65, "learning_rate": 1.3049571020019067e-06, "loss": 1.8266, "step": 29685 }, { "epoch": 5.65, "learning_rate": 1.3048999046711155e-06, "loss": 1.3851, "step": 29686 }, { "epoch": 5.65, "learning_rate": 1.304842707340324e-06, "loss": 1.5293, "step": 29687 }, { "epoch": 5.65, "learning_rate": 1.3047855100095329e-06, "loss": 2.0802, "step": 29688 }, { "epoch": 5.66, "learning_rate": 1.3047283126787417e-06, "loss": 1.3464, "step": 29689 }, { "epoch": 5.66, "learning_rate": 1.3046711153479505e-06, "loss": 2.7647, "step": 29690 }, { "epoch": 5.66, "learning_rate": 1.3046139180171593e-06, "loss": 2.2469, "step": 29691 }, { "epoch": 5.66, "learning_rate": 1.304556720686368e-06, "loss": 1.4623, "step": 29692 }, { "epoch": 5.66, "learning_rate": 1.3044995233555768e-06, "loss": 1.4164, "step": 29693 }, { "epoch": 5.66, "learning_rate": 1.3044423260247854e-06, "loss": 1.2456, "step": 29694 }, { "epoch": 5.66, "learning_rate": 1.3043851286939942e-06, "loss": 1.3354, "step": 29695 }, { "epoch": 5.66, "learning_rate": 1.304327931363203e-06, "loss": 1.0695, "step": 29696 }, { "epoch": 5.66, "learning_rate": 1.304270734032412e-06, "loss": 2.3813, "step": 29697 }, { "epoch": 5.66, "learning_rate": 1.3042135367016206e-06, "loss": 3.0762, "step": 29698 }, { "epoch": 5.66, "learning_rate": 1.3041563393708294e-06, "loss": 2.1613, "step": 29699 }, { "epoch": 5.66, "learning_rate": 1.3040991420400382e-06, "loss": 2.2906, "step": 29700 }, { "epoch": 5.66, "learning_rate": 1.3040419447092468e-06, "loss": 0.9981, "step": 29701 }, { "epoch": 5.66, "learning_rate": 1.3039847473784558e-06, "loss": 2.178, "step": 29702 }, { "epoch": 5.66, "learning_rate": 1.3039275500476646e-06, "loss": 1.3897, "step": 29703 }, { "epoch": 5.66, "learning_rate": 1.3038703527168734e-06, "loss": 1.6369, "step": 29704 }, { "epoch": 5.66, "learning_rate": 1.303813155386082e-06, "loss": 2.1675, "step": 29705 }, { "epoch": 5.66, "learning_rate": 1.3037559580552907e-06, "loss": 1.9523, "step": 29706 }, { "epoch": 5.66, "learning_rate": 1.3036987607244995e-06, "loss": 0.9749, "step": 29707 }, { "epoch": 5.66, "learning_rate": 1.3036415633937083e-06, "loss": 2.1639, "step": 29708 }, { "epoch": 5.66, "learning_rate": 1.3035843660629171e-06, "loss": 1.3862, "step": 29709 }, { "epoch": 5.66, "learning_rate": 1.303527168732126e-06, "loss": 1.5402, "step": 29710 }, { "epoch": 5.66, "learning_rate": 1.3034699714013345e-06, "loss": 2.2455, "step": 29711 }, { "epoch": 5.66, "learning_rate": 1.3034127740705433e-06, "loss": 1.3988, "step": 29712 }, { "epoch": 5.66, "learning_rate": 1.3033555767397523e-06, "loss": 3.1393, "step": 29713 }, { "epoch": 5.66, "learning_rate": 1.303298379408961e-06, "loss": 1.1434, "step": 29714 }, { "epoch": 5.66, "learning_rate": 1.3032411820781697e-06, "loss": 1.888, "step": 29715 }, { "epoch": 5.66, "learning_rate": 1.3031839847473785e-06, "loss": 0.801, "step": 29716 }, { "epoch": 5.66, "learning_rate": 1.3031267874165873e-06, "loss": 0.9771, "step": 29717 }, { "epoch": 5.66, "learning_rate": 1.303069590085796e-06, "loss": 1.0197, "step": 29718 }, { "epoch": 5.66, "learning_rate": 1.3030123927550048e-06, "loss": 3.2514, "step": 29719 }, { "epoch": 5.66, "learning_rate": 1.3029551954242136e-06, "loss": 2.388, "step": 29720 }, { "epoch": 5.66, "learning_rate": 1.3028979980934224e-06, "loss": 2.6225, "step": 29721 }, { "epoch": 5.66, "learning_rate": 1.302840800762631e-06, "loss": 1.444, "step": 29722 }, { "epoch": 5.66, "learning_rate": 1.3027836034318398e-06, "loss": 1.9681, "step": 29723 }, { "epoch": 5.66, "learning_rate": 1.3027264061010486e-06, "loss": 2.6014, "step": 29724 }, { "epoch": 5.66, "learning_rate": 1.3026692087702574e-06, "loss": 2.1078, "step": 29725 }, { "epoch": 5.66, "learning_rate": 1.3026120114394662e-06, "loss": 1.5756, "step": 29726 }, { "epoch": 5.66, "learning_rate": 1.302554814108675e-06, "loss": 1.8033, "step": 29727 }, { "epoch": 5.66, "learning_rate": 1.3024976167778836e-06, "loss": 1.9222, "step": 29728 }, { "epoch": 5.66, "learning_rate": 1.3024404194470926e-06, "loss": 2.5329, "step": 29729 }, { "epoch": 5.66, "learning_rate": 1.3023832221163014e-06, "loss": 1.0807, "step": 29730 }, { "epoch": 5.66, "learning_rate": 1.3023260247855102e-06, "loss": 1.0225, "step": 29731 }, { "epoch": 5.66, "learning_rate": 1.3022688274547187e-06, "loss": 1.7158, "step": 29732 }, { "epoch": 5.66, "learning_rate": 1.3022116301239275e-06, "loss": 1.384, "step": 29733 }, { "epoch": 5.66, "learning_rate": 1.3021544327931363e-06, "loss": 0.6707, "step": 29734 }, { "epoch": 5.66, "learning_rate": 1.3020972354623451e-06, "loss": 1.2224, "step": 29735 }, { "epoch": 5.66, "learning_rate": 1.302040038131554e-06, "loss": 1.215, "step": 29736 }, { "epoch": 5.66, "learning_rate": 1.3019828408007627e-06, "loss": 1.0043, "step": 29737 }, { "epoch": 5.66, "learning_rate": 1.3019256434699715e-06, "loss": 1.3232, "step": 29738 }, { "epoch": 5.66, "learning_rate": 1.30186844613918e-06, "loss": 2.3154, "step": 29739 }, { "epoch": 5.66, "learning_rate": 1.3018112488083889e-06, "loss": 1.8037, "step": 29740 }, { "epoch": 5.66, "learning_rate": 1.3017540514775979e-06, "loss": 2.0358, "step": 29741 }, { "epoch": 5.67, "learning_rate": 1.3016968541468065e-06, "loss": 1.0499, "step": 29742 }, { "epoch": 5.67, "learning_rate": 1.3016396568160153e-06, "loss": 2.1387, "step": 29743 }, { "epoch": 5.67, "learning_rate": 1.301582459485224e-06, "loss": 2.7753, "step": 29744 }, { "epoch": 5.67, "learning_rate": 1.3015252621544328e-06, "loss": 2.1735, "step": 29745 }, { "epoch": 5.67, "learning_rate": 1.3014680648236416e-06, "loss": 2.9382, "step": 29746 }, { "epoch": 5.67, "learning_rate": 1.3014108674928504e-06, "loss": 1.7937, "step": 29747 }, { "epoch": 5.67, "learning_rate": 1.3013536701620592e-06, "loss": 0.7323, "step": 29748 }, { "epoch": 5.67, "learning_rate": 1.3012964728312678e-06, "loss": 2.3644, "step": 29749 }, { "epoch": 5.67, "learning_rate": 1.3012392755004766e-06, "loss": 2.4324, "step": 29750 }, { "epoch": 5.67, "learning_rate": 1.3011820781696854e-06, "loss": 1.5869, "step": 29751 }, { "epoch": 5.67, "learning_rate": 1.3011248808388942e-06, "loss": 1.5264, "step": 29752 }, { "epoch": 5.67, "learning_rate": 1.301067683508103e-06, "loss": 1.3019, "step": 29753 }, { "epoch": 5.67, "learning_rate": 1.3010104861773118e-06, "loss": 1.8234, "step": 29754 }, { "epoch": 5.67, "learning_rate": 1.3009532888465206e-06, "loss": 1.1972, "step": 29755 }, { "epoch": 5.67, "learning_rate": 1.3008960915157292e-06, "loss": 1.289, "step": 29756 }, { "epoch": 5.67, "learning_rate": 1.3008388941849382e-06, "loss": 1.7868, "step": 29757 }, { "epoch": 5.67, "learning_rate": 1.300781696854147e-06, "loss": 1.7674, "step": 29758 }, { "epoch": 5.67, "learning_rate": 1.3007244995233555e-06, "loss": 1.093, "step": 29759 }, { "epoch": 5.67, "learning_rate": 1.3006673021925643e-06, "loss": 2.6733, "step": 29760 }, { "epoch": 5.67, "learning_rate": 1.3006101048617731e-06, "loss": 2.1839, "step": 29761 }, { "epoch": 5.67, "learning_rate": 1.300552907530982e-06, "loss": 2.3704, "step": 29762 }, { "epoch": 5.67, "learning_rate": 1.3004957102001907e-06, "loss": 1.3008, "step": 29763 }, { "epoch": 5.67, "learning_rate": 1.3004385128693995e-06, "loss": 2.5415, "step": 29764 }, { "epoch": 5.67, "learning_rate": 1.3003813155386083e-06, "loss": 2.9704, "step": 29765 }, { "epoch": 5.67, "learning_rate": 1.3003241182078169e-06, "loss": 2.0729, "step": 29766 }, { "epoch": 5.67, "learning_rate": 1.3002669208770257e-06, "loss": 2.4209, "step": 29767 }, { "epoch": 5.67, "learning_rate": 1.3002097235462345e-06, "loss": 1.6507, "step": 29768 }, { "epoch": 5.67, "learning_rate": 1.3001525262154435e-06, "loss": 3.4201, "step": 29769 }, { "epoch": 5.67, "learning_rate": 1.300095328884652e-06, "loss": 1.8657, "step": 29770 }, { "epoch": 5.67, "learning_rate": 1.3000381315538609e-06, "loss": 1.9342, "step": 29771 }, { "epoch": 5.67, "learning_rate": 1.2999809342230696e-06, "loss": 1.657, "step": 29772 }, { "epoch": 5.67, "learning_rate": 1.2999237368922784e-06, "loss": 1.1118, "step": 29773 }, { "epoch": 5.67, "learning_rate": 1.2998665395614872e-06, "loss": 1.7677, "step": 29774 }, { "epoch": 5.67, "learning_rate": 1.299809342230696e-06, "loss": 1.4741, "step": 29775 }, { "epoch": 5.67, "learning_rate": 1.2997521448999046e-06, "loss": 2.3557, "step": 29776 }, { "epoch": 5.67, "learning_rate": 1.2996949475691134e-06, "loss": 1.3589, "step": 29777 }, { "epoch": 5.67, "learning_rate": 1.2996377502383222e-06, "loss": 1.6634, "step": 29778 }, { "epoch": 5.67, "learning_rate": 1.299580552907531e-06, "loss": 2.1019, "step": 29779 }, { "epoch": 5.67, "learning_rate": 1.2995233555767398e-06, "loss": 2.4513, "step": 29780 }, { "epoch": 5.67, "learning_rate": 1.2994661582459486e-06, "loss": 1.7911, "step": 29781 }, { "epoch": 5.67, "learning_rate": 1.2994089609151574e-06, "loss": 1.8541, "step": 29782 }, { "epoch": 5.67, "learning_rate": 1.299351763584366e-06, "loss": 2.189, "step": 29783 }, { "epoch": 5.67, "learning_rate": 1.2992945662535747e-06, "loss": 2.4107, "step": 29784 }, { "epoch": 5.67, "learning_rate": 1.2992373689227838e-06, "loss": 2.3979, "step": 29785 }, { "epoch": 5.67, "learning_rate": 1.2991801715919925e-06, "loss": 3.1549, "step": 29786 }, { "epoch": 5.67, "learning_rate": 1.2991229742612011e-06, "loss": 1.1462, "step": 29787 }, { "epoch": 5.67, "learning_rate": 1.29906577693041e-06, "loss": 1.8734, "step": 29788 }, { "epoch": 5.67, "learning_rate": 1.2990085795996187e-06, "loss": 2.6661, "step": 29789 }, { "epoch": 5.67, "learning_rate": 1.2989513822688275e-06, "loss": 1.2052, "step": 29790 }, { "epoch": 5.67, "learning_rate": 1.2988941849380363e-06, "loss": 1.7101, "step": 29791 }, { "epoch": 5.67, "learning_rate": 1.298836987607245e-06, "loss": 1.0707, "step": 29792 }, { "epoch": 5.67, "learning_rate": 1.2987797902764537e-06, "loss": 2.7927, "step": 29793 }, { "epoch": 5.68, "learning_rate": 1.2987225929456625e-06, "loss": 3.4328, "step": 29794 }, { "epoch": 5.68, "learning_rate": 1.2986653956148713e-06, "loss": 2.6674, "step": 29795 }, { "epoch": 5.68, "learning_rate": 1.2986081982840803e-06, "loss": 1.5626, "step": 29796 }, { "epoch": 5.68, "learning_rate": 1.2985510009532889e-06, "loss": 1.6548, "step": 29797 }, { "epoch": 5.68, "learning_rate": 1.2984938036224976e-06, "loss": 2.458, "step": 29798 }, { "epoch": 5.68, "learning_rate": 1.2984366062917064e-06, "loss": 0.7922, "step": 29799 }, { "epoch": 5.68, "learning_rate": 1.298379408960915e-06, "loss": 3.0778, "step": 29800 }, { "epoch": 5.68, "learning_rate": 1.298322211630124e-06, "loss": 1.3343, "step": 29801 }, { "epoch": 5.68, "learning_rate": 1.2982650142993328e-06, "loss": 2.7985, "step": 29802 }, { "epoch": 5.68, "learning_rate": 1.2982078169685416e-06, "loss": 2.0027, "step": 29803 }, { "epoch": 5.68, "learning_rate": 1.2981506196377502e-06, "loss": 1.6918, "step": 29804 }, { "epoch": 5.68, "learning_rate": 1.298093422306959e-06, "loss": 3.1596, "step": 29805 }, { "epoch": 5.68, "learning_rate": 1.2980362249761678e-06, "loss": 1.6838, "step": 29806 }, { "epoch": 5.68, "learning_rate": 1.2979790276453766e-06, "loss": 2.2621, "step": 29807 }, { "epoch": 5.68, "learning_rate": 1.2979218303145854e-06, "loss": 2.9025, "step": 29808 }, { "epoch": 5.68, "learning_rate": 1.2978646329837942e-06, "loss": 2.1001, "step": 29809 }, { "epoch": 5.68, "learning_rate": 1.297807435653003e-06, "loss": 1.4954, "step": 29810 }, { "epoch": 5.68, "learning_rate": 1.2977502383222115e-06, "loss": 1.8139, "step": 29811 }, { "epoch": 5.68, "learning_rate": 1.2976930409914203e-06, "loss": 1.4244, "step": 29812 }, { "epoch": 5.68, "learning_rate": 1.2976358436606293e-06, "loss": 2.1344, "step": 29813 }, { "epoch": 5.68, "learning_rate": 1.297578646329838e-06, "loss": 1.3976, "step": 29814 }, { "epoch": 5.68, "learning_rate": 1.2975214489990467e-06, "loss": 2.8062, "step": 29815 }, { "epoch": 5.68, "learning_rate": 1.2974642516682555e-06, "loss": 2.1012, "step": 29816 }, { "epoch": 5.68, "learning_rate": 1.2974070543374643e-06, "loss": 2.2757, "step": 29817 }, { "epoch": 5.68, "learning_rate": 1.297349857006673e-06, "loss": 1.221, "step": 29818 }, { "epoch": 5.68, "learning_rate": 1.297292659675882e-06, "loss": 0.9548, "step": 29819 }, { "epoch": 5.68, "learning_rate": 1.2972354623450907e-06, "loss": 2.7084, "step": 29820 }, { "epoch": 5.68, "learning_rate": 1.2971782650142993e-06, "loss": 2.1343, "step": 29821 }, { "epoch": 5.68, "learning_rate": 1.297121067683508e-06, "loss": 1.2841, "step": 29822 }, { "epoch": 5.68, "learning_rate": 1.2970638703527169e-06, "loss": 1.5757, "step": 29823 }, { "epoch": 5.68, "learning_rate": 1.2970066730219257e-06, "loss": 0.3634, "step": 29824 }, { "epoch": 5.68, "learning_rate": 1.2969494756911344e-06, "loss": 2.1731, "step": 29825 }, { "epoch": 5.68, "learning_rate": 1.2968922783603432e-06, "loss": 1.8048, "step": 29826 }, { "epoch": 5.68, "learning_rate": 1.296835081029552e-06, "loss": 1.3305, "step": 29827 }, { "epoch": 5.68, "learning_rate": 1.2967778836987606e-06, "loss": 0.9289, "step": 29828 }, { "epoch": 5.68, "learning_rate": 1.2967206863679696e-06, "loss": 2.6876, "step": 29829 }, { "epoch": 5.68, "learning_rate": 1.2966634890371784e-06, "loss": 2.1406, "step": 29830 }, { "epoch": 5.68, "learning_rate": 1.296606291706387e-06, "loss": 3.6675, "step": 29831 }, { "epoch": 5.68, "learning_rate": 1.2965490943755958e-06, "loss": 0.831, "step": 29832 }, { "epoch": 5.68, "learning_rate": 1.2964918970448046e-06, "loss": 2.6964, "step": 29833 }, { "epoch": 5.68, "learning_rate": 1.2964346997140134e-06, "loss": 1.9427, "step": 29834 }, { "epoch": 5.68, "learning_rate": 1.2963775023832222e-06, "loss": 1.6743, "step": 29835 }, { "epoch": 5.68, "learning_rate": 1.296320305052431e-06, "loss": 1.9018, "step": 29836 }, { "epoch": 5.68, "learning_rate": 1.2962631077216398e-06, "loss": 2.5859, "step": 29837 }, { "epoch": 5.68, "learning_rate": 1.2962059103908483e-06, "loss": 0.8048, "step": 29838 }, { "epoch": 5.68, "learning_rate": 1.2961487130600571e-06, "loss": 1.8568, "step": 29839 }, { "epoch": 5.68, "learning_rate": 1.2960915157292661e-06, "loss": 1.6946, "step": 29840 }, { "epoch": 5.68, "learning_rate": 1.2960343183984747e-06, "loss": 2.6423, "step": 29841 }, { "epoch": 5.68, "learning_rate": 1.2959771210676835e-06, "loss": 1.8242, "step": 29842 }, { "epoch": 5.68, "learning_rate": 1.2959199237368923e-06, "loss": 1.3915, "step": 29843 }, { "epoch": 5.68, "learning_rate": 1.295862726406101e-06, "loss": 2.1551, "step": 29844 }, { "epoch": 5.68, "learning_rate": 1.29580552907531e-06, "loss": 1.3065, "step": 29845 }, { "epoch": 5.68, "learning_rate": 1.2957483317445187e-06, "loss": 1.6636, "step": 29846 }, { "epoch": 5.69, "learning_rate": 1.2956911344137275e-06, "loss": 2.8948, "step": 29847 }, { "epoch": 5.69, "learning_rate": 1.295633937082936e-06, "loss": 1.2332, "step": 29848 }, { "epoch": 5.69, "learning_rate": 1.2955767397521449e-06, "loss": 2.4359, "step": 29849 }, { "epoch": 5.69, "learning_rate": 1.2955195424213537e-06, "loss": 2.3596, "step": 29850 }, { "epoch": 5.69, "learning_rate": 1.2954623450905625e-06, "loss": 1.2934, "step": 29851 }, { "epoch": 5.69, "learning_rate": 1.2954051477597712e-06, "loss": 2.6525, "step": 29852 }, { "epoch": 5.69, "learning_rate": 1.29534795042898e-06, "loss": 2.291, "step": 29853 }, { "epoch": 5.69, "learning_rate": 1.2952907530981888e-06, "loss": 3.9566, "step": 29854 }, { "epoch": 5.69, "learning_rate": 1.2952335557673974e-06, "loss": 2.1025, "step": 29855 }, { "epoch": 5.69, "learning_rate": 1.2951763584366062e-06, "loss": 2.2677, "step": 29856 }, { "epoch": 5.69, "learning_rate": 1.2951191611058152e-06, "loss": 2.5058, "step": 29857 }, { "epoch": 5.69, "learning_rate": 1.2950619637750238e-06, "loss": 2.0302, "step": 29858 }, { "epoch": 5.69, "learning_rate": 1.2950047664442326e-06, "loss": 1.3005, "step": 29859 }, { "epoch": 5.69, "learning_rate": 1.2949475691134414e-06, "loss": 1.6997, "step": 29860 }, { "epoch": 5.69, "learning_rate": 1.2948903717826502e-06, "loss": 1.8106, "step": 29861 }, { "epoch": 5.69, "learning_rate": 1.294833174451859e-06, "loss": 1.4325, "step": 29862 }, { "epoch": 5.69, "learning_rate": 1.2947759771210678e-06, "loss": 3.2047, "step": 29863 }, { "epoch": 5.69, "learning_rate": 1.2947187797902766e-06, "loss": 1.567, "step": 29864 }, { "epoch": 5.69, "learning_rate": 1.2946615824594851e-06, "loss": 2.669, "step": 29865 }, { "epoch": 5.69, "learning_rate": 1.294604385128694e-06, "loss": 1.7496, "step": 29866 }, { "epoch": 5.69, "learning_rate": 1.2945471877979027e-06, "loss": 1.0269, "step": 29867 }, { "epoch": 5.69, "learning_rate": 1.2944899904671117e-06, "loss": 1.3893, "step": 29868 }, { "epoch": 5.69, "learning_rate": 1.2944327931363203e-06, "loss": 2.1124, "step": 29869 }, { "epoch": 5.69, "learning_rate": 1.2943755958055291e-06, "loss": 1.819, "step": 29870 }, { "epoch": 5.69, "learning_rate": 1.294318398474738e-06, "loss": 0.9752, "step": 29871 }, { "epoch": 5.69, "learning_rate": 1.2942612011439465e-06, "loss": 1.5171, "step": 29872 }, { "epoch": 5.69, "learning_rate": 1.2942040038131555e-06, "loss": 1.1398, "step": 29873 }, { "epoch": 5.69, "learning_rate": 1.2941468064823643e-06, "loss": 1.879, "step": 29874 }, { "epoch": 5.69, "learning_rate": 1.294089609151573e-06, "loss": 2.5075, "step": 29875 }, { "epoch": 5.69, "learning_rate": 1.2940324118207817e-06, "loss": 1.4678, "step": 29876 }, { "epoch": 5.69, "learning_rate": 1.2939752144899905e-06, "loss": 1.2524, "step": 29877 }, { "epoch": 5.69, "learning_rate": 1.2939180171591992e-06, "loss": 2.0086, "step": 29878 }, { "epoch": 5.69, "learning_rate": 1.293860819828408e-06, "loss": 2.3031, "step": 29879 }, { "epoch": 5.69, "learning_rate": 1.2938036224976168e-06, "loss": 1.9281, "step": 29880 }, { "epoch": 5.69, "learning_rate": 1.2937464251668256e-06, "loss": 2.392, "step": 29881 }, { "epoch": 5.69, "learning_rate": 1.2936892278360342e-06, "loss": 0.479, "step": 29882 }, { "epoch": 5.69, "learning_rate": 1.293632030505243e-06, "loss": 1.335, "step": 29883 }, { "epoch": 5.69, "learning_rate": 1.293574833174452e-06, "loss": 2.0941, "step": 29884 }, { "epoch": 5.69, "learning_rate": 1.2935176358436608e-06, "loss": 2.2726, "step": 29885 }, { "epoch": 5.69, "learning_rate": 1.2934604385128694e-06, "loss": 0.9071, "step": 29886 }, { "epoch": 5.69, "learning_rate": 1.2934032411820782e-06, "loss": 2.0748, "step": 29887 }, { "epoch": 5.69, "learning_rate": 1.293346043851287e-06, "loss": 0.834, "step": 29888 }, { "epoch": 5.69, "learning_rate": 1.2932888465204958e-06, "loss": 2.35, "step": 29889 }, { "epoch": 5.69, "learning_rate": 1.2932316491897046e-06, "loss": 1.1099, "step": 29890 }, { "epoch": 5.69, "learning_rate": 1.2931744518589134e-06, "loss": 2.7335, "step": 29891 }, { "epoch": 5.69, "learning_rate": 1.2931172545281222e-06, "loss": 2.8422, "step": 29892 }, { "epoch": 5.69, "learning_rate": 1.2930600571973307e-06, "loss": 2.0977, "step": 29893 }, { "epoch": 5.69, "learning_rate": 1.2930028598665395e-06, "loss": 1.9506, "step": 29894 }, { "epoch": 5.69, "learning_rate": 1.2929456625357483e-06, "loss": 1.9882, "step": 29895 }, { "epoch": 5.69, "learning_rate": 1.2928884652049571e-06, "loss": 1.0638, "step": 29896 }, { "epoch": 5.69, "learning_rate": 1.292831267874166e-06, "loss": 2.0087, "step": 29897 }, { "epoch": 5.69, "learning_rate": 1.2927740705433747e-06, "loss": 1.044, "step": 29898 }, { "epoch": 5.7, "learning_rate": 1.2927168732125835e-06, "loss": 2.2715, "step": 29899 }, { "epoch": 5.7, "learning_rate": 1.292659675881792e-06, "loss": 3.0762, "step": 29900 }, { "epoch": 5.7, "learning_rate": 1.292602478551001e-06, "loss": 1.887, "step": 29901 }, { "epoch": 5.7, "learning_rate": 1.2925452812202099e-06, "loss": 3.2704, "step": 29902 }, { "epoch": 5.7, "learning_rate": 1.2924880838894185e-06, "loss": 2.5008, "step": 29903 }, { "epoch": 5.7, "learning_rate": 1.2924308865586273e-06, "loss": 2.2739, "step": 29904 }, { "epoch": 5.7, "learning_rate": 1.292373689227836e-06, "loss": 1.0309, "step": 29905 }, { "epoch": 5.7, "learning_rate": 1.2923164918970448e-06, "loss": 2.1557, "step": 29906 }, { "epoch": 5.7, "learning_rate": 1.2922592945662536e-06, "loss": 2.6723, "step": 29907 }, { "epoch": 5.7, "learning_rate": 1.2922020972354624e-06, "loss": 2.512, "step": 29908 }, { "epoch": 5.7, "learning_rate": 1.2921448999046712e-06, "loss": 1.5975, "step": 29909 }, { "epoch": 5.7, "learning_rate": 1.2920877025738798e-06, "loss": 1.7856, "step": 29910 }, { "epoch": 5.7, "learning_rate": 1.2920305052430886e-06, "loss": 1.9212, "step": 29911 }, { "epoch": 5.7, "learning_rate": 1.2919733079122976e-06, "loss": 1.9728, "step": 29912 }, { "epoch": 5.7, "learning_rate": 1.2919161105815062e-06, "loss": 1.7969, "step": 29913 }, { "epoch": 5.7, "learning_rate": 1.291858913250715e-06, "loss": 1.3484, "step": 29914 }, { "epoch": 5.7, "learning_rate": 1.2918017159199238e-06, "loss": 1.622, "step": 29915 }, { "epoch": 5.7, "learning_rate": 1.2917445185891326e-06, "loss": 3.0313, "step": 29916 }, { "epoch": 5.7, "learning_rate": 1.2916873212583414e-06, "loss": 2.9494, "step": 29917 }, { "epoch": 5.7, "learning_rate": 1.2916301239275502e-06, "loss": 2.6096, "step": 29918 }, { "epoch": 5.7, "learning_rate": 1.291572926596759e-06, "loss": 2.2874, "step": 29919 }, { "epoch": 5.7, "learning_rate": 1.2915157292659675e-06, "loss": 0.9878, "step": 29920 }, { "epoch": 5.7, "learning_rate": 1.2914585319351763e-06, "loss": 2.1468, "step": 29921 }, { "epoch": 5.7, "learning_rate": 1.2914013346043851e-06, "loss": 1.9113, "step": 29922 }, { "epoch": 5.7, "learning_rate": 1.291344137273594e-06, "loss": 2.2255, "step": 29923 }, { "epoch": 5.7, "learning_rate": 1.2912869399428027e-06, "loss": 0.6122, "step": 29924 }, { "epoch": 5.7, "learning_rate": 1.2912297426120115e-06, "loss": 1.6419, "step": 29925 }, { "epoch": 5.7, "learning_rate": 1.2911725452812203e-06, "loss": 1.7108, "step": 29926 }, { "epoch": 5.7, "learning_rate": 1.2911153479504289e-06, "loss": 1.676, "step": 29927 }, { "epoch": 5.7, "learning_rate": 1.2910581506196379e-06, "loss": 2.0431, "step": 29928 }, { "epoch": 5.7, "learning_rate": 1.2910009532888467e-06, "loss": 1.2228, "step": 29929 }, { "epoch": 5.7, "learning_rate": 1.2909437559580553e-06, "loss": 1.6503, "step": 29930 }, { "epoch": 5.7, "learning_rate": 1.290886558627264e-06, "loss": 2.2086, "step": 29931 }, { "epoch": 5.7, "learning_rate": 1.2908293612964728e-06, "loss": 1.4892, "step": 29932 }, { "epoch": 5.7, "learning_rate": 1.2907721639656816e-06, "loss": 1.767, "step": 29933 }, { "epoch": 5.7, "learning_rate": 1.2907149666348904e-06, "loss": 3.0783, "step": 29934 }, { "epoch": 5.7, "learning_rate": 1.2906577693040992e-06, "loss": 2.3212, "step": 29935 }, { "epoch": 5.7, "learning_rate": 1.290600571973308e-06, "loss": 2.3079, "step": 29936 }, { "epoch": 5.7, "learning_rate": 1.2905433746425166e-06, "loss": 2.2785, "step": 29937 }, { "epoch": 5.7, "learning_rate": 1.2904861773117254e-06, "loss": 2.4118, "step": 29938 }, { "epoch": 5.7, "learning_rate": 1.2904289799809342e-06, "loss": 1.6899, "step": 29939 }, { "epoch": 5.7, "learning_rate": 1.2903717826501432e-06, "loss": 2.1355, "step": 29940 }, { "epoch": 5.7, "learning_rate": 1.2903145853193518e-06, "loss": 2.3132, "step": 29941 }, { "epoch": 5.7, "learning_rate": 1.2902573879885606e-06, "loss": 2.2514, "step": 29942 }, { "epoch": 5.7, "learning_rate": 1.2902001906577694e-06, "loss": 2.2577, "step": 29943 }, { "epoch": 5.7, "learning_rate": 1.290142993326978e-06, "loss": 0.8834, "step": 29944 }, { "epoch": 5.7, "learning_rate": 1.290085795996187e-06, "loss": 1.6944, "step": 29945 }, { "epoch": 5.7, "learning_rate": 1.2900285986653957e-06, "loss": 2.1553, "step": 29946 }, { "epoch": 5.7, "learning_rate": 1.2899714013346043e-06, "loss": 1.7275, "step": 29947 }, { "epoch": 5.7, "learning_rate": 1.2899142040038131e-06, "loss": 2.0809, "step": 29948 }, { "epoch": 5.7, "learning_rate": 1.289857006673022e-06, "loss": 1.4995, "step": 29949 }, { "epoch": 5.7, "learning_rate": 1.2897998093422307e-06, "loss": 2.2043, "step": 29950 }, { "epoch": 5.7, "learning_rate": 1.2897426120114395e-06, "loss": 2.2744, "step": 29951 }, { "epoch": 5.71, "learning_rate": 1.2896854146806483e-06, "loss": 2.3232, "step": 29952 }, { "epoch": 5.71, "learning_rate": 1.289628217349857e-06, "loss": 1.3871, "step": 29953 }, { "epoch": 5.71, "learning_rate": 1.2895710200190657e-06, "loss": 2.0517, "step": 29954 }, { "epoch": 5.71, "learning_rate": 1.2895138226882745e-06, "loss": 1.3362, "step": 29955 }, { "epoch": 5.71, "learning_rate": 1.2894566253574835e-06, "loss": 2.3021, "step": 29956 }, { "epoch": 5.71, "learning_rate": 1.2893994280266923e-06, "loss": 1.7922, "step": 29957 }, { "epoch": 5.71, "learning_rate": 1.2893422306959008e-06, "loss": 1.1232, "step": 29958 }, { "epoch": 5.71, "learning_rate": 1.2892850333651096e-06, "loss": 1.2042, "step": 29959 }, { "epoch": 5.71, "learning_rate": 1.2892278360343184e-06, "loss": 1.4944, "step": 29960 }, { "epoch": 5.71, "learning_rate": 1.2891706387035272e-06, "loss": 1.1376, "step": 29961 }, { "epoch": 5.71, "learning_rate": 1.289113441372736e-06, "loss": 2.4085, "step": 29962 }, { "epoch": 5.71, "learning_rate": 1.2890562440419448e-06, "loss": 2.9052, "step": 29963 }, { "epoch": 5.71, "learning_rate": 1.2889990467111536e-06, "loss": 2.1412, "step": 29964 }, { "epoch": 5.71, "learning_rate": 1.2889418493803622e-06, "loss": 0.9266, "step": 29965 }, { "epoch": 5.71, "learning_rate": 1.288884652049571e-06, "loss": 1.7974, "step": 29966 }, { "epoch": 5.71, "learning_rate": 1.2888274547187798e-06, "loss": 1.4199, "step": 29967 }, { "epoch": 5.71, "learning_rate": 1.2887702573879886e-06, "loss": 1.1536, "step": 29968 }, { "epoch": 5.71, "learning_rate": 1.2887130600571974e-06, "loss": 2.8253, "step": 29969 }, { "epoch": 5.71, "learning_rate": 1.2886558627264062e-06, "loss": 1.6233, "step": 29970 }, { "epoch": 5.71, "learning_rate": 1.2885986653956147e-06, "loss": 1.7995, "step": 29971 }, { "epoch": 5.71, "learning_rate": 1.2885414680648237e-06, "loss": 0.9632, "step": 29972 }, { "epoch": 5.71, "learning_rate": 1.2884842707340325e-06, "loss": 2.6751, "step": 29973 }, { "epoch": 5.71, "learning_rate": 1.2884270734032413e-06, "loss": 1.7318, "step": 29974 }, { "epoch": 5.71, "learning_rate": 1.28836987607245e-06, "loss": 1.5973, "step": 29975 }, { "epoch": 5.71, "learning_rate": 1.2883126787416587e-06, "loss": 1.1916, "step": 29976 }, { "epoch": 5.71, "learning_rate": 1.2882554814108675e-06, "loss": 2.6752, "step": 29977 }, { "epoch": 5.71, "learning_rate": 1.2881982840800763e-06, "loss": 1.8495, "step": 29978 }, { "epoch": 5.71, "learning_rate": 1.288141086749285e-06, "loss": 1.9355, "step": 29979 }, { "epoch": 5.71, "learning_rate": 1.2880838894184939e-06, "loss": 1.917, "step": 29980 }, { "epoch": 5.71, "learning_rate": 1.2880266920877027e-06, "loss": 1.8876, "step": 29981 }, { "epoch": 5.71, "learning_rate": 1.2879694947569113e-06, "loss": 2.2321, "step": 29982 }, { "epoch": 5.71, "learning_rate": 1.28791229742612e-06, "loss": 1.072, "step": 29983 }, { "epoch": 5.71, "learning_rate": 1.287855100095329e-06, "loss": 3.2608, "step": 29984 }, { "epoch": 5.71, "learning_rate": 1.2877979027645376e-06, "loss": 1.8983, "step": 29985 }, { "epoch": 5.71, "learning_rate": 1.2877407054337464e-06, "loss": 1.0401, "step": 29986 }, { "epoch": 5.71, "learning_rate": 1.2876835081029552e-06, "loss": 2.2576, "step": 29987 }, { "epoch": 5.71, "learning_rate": 1.2876263107721638e-06, "loss": 1.2348, "step": 29988 }, { "epoch": 5.71, "learning_rate": 1.2875691134413728e-06, "loss": 1.8969, "step": 29989 }, { "epoch": 5.71, "learning_rate": 1.2875119161105816e-06, "loss": 3.1315, "step": 29990 }, { "epoch": 5.71, "learning_rate": 1.2874547187797904e-06, "loss": 2.3686, "step": 29991 }, { "epoch": 5.71, "learning_rate": 1.287397521448999e-06, "loss": 2.7381, "step": 29992 }, { "epoch": 5.71, "learning_rate": 1.2873403241182078e-06, "loss": 1.856, "step": 29993 }, { "epoch": 5.71, "learning_rate": 1.2872831267874166e-06, "loss": 1.7317, "step": 29994 }, { "epoch": 5.71, "learning_rate": 1.2872259294566254e-06, "loss": 2.0502, "step": 29995 }, { "epoch": 5.71, "learning_rate": 1.2871687321258342e-06, "loss": 2.4003, "step": 29996 }, { "epoch": 5.71, "learning_rate": 1.287111534795043e-06, "loss": 2.0963, "step": 29997 }, { "epoch": 5.71, "learning_rate": 1.2870543374642518e-06, "loss": 1.2084, "step": 29998 }, { "epoch": 5.71, "learning_rate": 1.2869971401334603e-06, "loss": 2.8467, "step": 29999 }, { "epoch": 5.71, "learning_rate": 1.2869399428026693e-06, "loss": 1.5476, "step": 30000 }, { "epoch": 5.71, "learning_rate": 1.2868827454718781e-06, "loss": 1.234, "step": 30001 }, { "epoch": 5.71, "learning_rate": 1.2868255481410867e-06, "loss": 1.0687, "step": 30002 }, { "epoch": 5.71, "learning_rate": 1.2867683508102955e-06, "loss": 2.2123, "step": 30003 }, { "epoch": 5.72, "learning_rate": 1.2867111534795043e-06, "loss": 1.8905, "step": 30004 }, { "epoch": 5.72, "learning_rate": 1.286653956148713e-06, "loss": 2.6557, "step": 30005 }, { "epoch": 5.72, "learning_rate": 1.2865967588179219e-06, "loss": 2.5065, "step": 30006 }, { "epoch": 5.72, "learning_rate": 1.2865395614871307e-06, "loss": 1.9181, "step": 30007 }, { "epoch": 5.72, "learning_rate": 1.2864823641563395e-06, "loss": 1.3117, "step": 30008 }, { "epoch": 5.72, "learning_rate": 1.286425166825548e-06, "loss": 2.8749, "step": 30009 }, { "epoch": 5.72, "learning_rate": 1.2863679694947569e-06, "loss": 1.5237, "step": 30010 }, { "epoch": 5.72, "learning_rate": 1.2863107721639656e-06, "loss": 3.0675, "step": 30011 }, { "epoch": 5.72, "learning_rate": 1.2862535748331744e-06, "loss": 1.3861, "step": 30012 }, { "epoch": 5.72, "learning_rate": 1.2861963775023832e-06, "loss": 1.3323, "step": 30013 }, { "epoch": 5.72, "learning_rate": 1.286139180171592e-06, "loss": 2.8813, "step": 30014 }, { "epoch": 5.72, "learning_rate": 1.2860819828408008e-06, "loss": 1.2864, "step": 30015 }, { "epoch": 5.72, "learning_rate": 1.2860247855100096e-06, "loss": 0.9603, "step": 30016 }, { "epoch": 5.72, "learning_rate": 1.2859675881792184e-06, "loss": 1.4826, "step": 30017 }, { "epoch": 5.72, "learning_rate": 1.2859103908484272e-06, "loss": 0.7943, "step": 30018 }, { "epoch": 5.72, "learning_rate": 1.2858531935176358e-06, "loss": 1.4655, "step": 30019 }, { "epoch": 5.72, "learning_rate": 1.2857959961868446e-06, "loss": 1.1329, "step": 30020 }, { "epoch": 5.72, "learning_rate": 1.2857387988560534e-06, "loss": 2.0427, "step": 30021 }, { "epoch": 5.72, "learning_rate": 1.2856816015252622e-06, "loss": 1.7056, "step": 30022 }, { "epoch": 5.72, "learning_rate": 1.285624404194471e-06, "loss": 0.8167, "step": 30023 }, { "epoch": 5.72, "learning_rate": 1.2855672068636798e-06, "loss": 2.2017, "step": 30024 }, { "epoch": 5.72, "learning_rate": 1.2855100095328886e-06, "loss": 2.1134, "step": 30025 }, { "epoch": 5.72, "learning_rate": 1.2854528122020971e-06, "loss": 2.9766, "step": 30026 }, { "epoch": 5.72, "learning_rate": 1.285395614871306e-06, "loss": 2.237, "step": 30027 }, { "epoch": 5.72, "learning_rate": 1.285338417540515e-06, "loss": 2.726, "step": 30028 }, { "epoch": 5.72, "learning_rate": 1.2852812202097237e-06, "loss": 1.8553, "step": 30029 }, { "epoch": 5.72, "learning_rate": 1.2852240228789323e-06, "loss": 1.5754, "step": 30030 }, { "epoch": 5.72, "learning_rate": 1.285166825548141e-06, "loss": 0.8464, "step": 30031 }, { "epoch": 5.72, "learning_rate": 1.28510962821735e-06, "loss": 1.0221, "step": 30032 }, { "epoch": 5.72, "learning_rate": 1.2850524308865587e-06, "loss": 1.3825, "step": 30033 }, { "epoch": 5.72, "learning_rate": 1.2849952335557675e-06, "loss": 0.9713, "step": 30034 }, { "epoch": 5.72, "learning_rate": 1.2849380362249763e-06, "loss": 1.3965, "step": 30035 }, { "epoch": 5.72, "learning_rate": 1.2848808388941849e-06, "loss": 1.7897, "step": 30036 }, { "epoch": 5.72, "learning_rate": 1.2848236415633937e-06, "loss": 2.1044, "step": 30037 }, { "epoch": 5.72, "learning_rate": 1.2847664442326024e-06, "loss": 1.0752, "step": 30038 }, { "epoch": 5.72, "learning_rate": 1.2847092469018115e-06, "loss": 1.386, "step": 30039 }, { "epoch": 5.72, "learning_rate": 1.28465204957102e-06, "loss": 2.2934, "step": 30040 }, { "epoch": 5.72, "learning_rate": 1.2845948522402288e-06, "loss": 1.186, "step": 30041 }, { "epoch": 5.72, "learning_rate": 1.2845376549094376e-06, "loss": 1.0069, "step": 30042 }, { "epoch": 5.72, "learning_rate": 1.2844804575786462e-06, "loss": 3.0824, "step": 30043 }, { "epoch": 5.72, "learning_rate": 1.2844232602478552e-06, "loss": 2.3383, "step": 30044 }, { "epoch": 5.72, "learning_rate": 1.284366062917064e-06, "loss": 1.5827, "step": 30045 }, { "epoch": 5.72, "learning_rate": 1.2843088655862728e-06, "loss": 1.7614, "step": 30046 }, { "epoch": 5.72, "learning_rate": 1.2842516682554814e-06, "loss": 0.6107, "step": 30047 }, { "epoch": 5.72, "learning_rate": 1.2841944709246902e-06, "loss": 1.4109, "step": 30048 }, { "epoch": 5.72, "learning_rate": 1.284137273593899e-06, "loss": 2.1716, "step": 30049 }, { "epoch": 5.72, "learning_rate": 1.2840800762631078e-06, "loss": 1.6669, "step": 30050 }, { "epoch": 5.72, "learning_rate": 1.2840228789323166e-06, "loss": 2.7487, "step": 30051 }, { "epoch": 5.72, "learning_rate": 1.2839656816015253e-06, "loss": 1.7387, "step": 30052 }, { "epoch": 5.72, "learning_rate": 1.283908484270734e-06, "loss": 2.5955, "step": 30053 }, { "epoch": 5.72, "learning_rate": 1.2838512869399427e-06, "loss": 2.8004, "step": 30054 }, { "epoch": 5.72, "learning_rate": 1.2837940896091515e-06, "loss": 2.2564, "step": 30055 }, { "epoch": 5.72, "learning_rate": 1.2837368922783605e-06, "loss": 0.7559, "step": 30056 }, { "epoch": 5.73, "learning_rate": 1.2836796949475691e-06, "loss": 1.4508, "step": 30057 }, { "epoch": 5.73, "learning_rate": 1.283622497616778e-06, "loss": 2.1842, "step": 30058 }, { "epoch": 5.73, "learning_rate": 1.2835653002859867e-06, "loss": 2.1147, "step": 30059 }, { "epoch": 5.73, "learning_rate": 1.2835081029551955e-06, "loss": 1.3209, "step": 30060 }, { "epoch": 5.73, "learning_rate": 1.2834509056244043e-06, "loss": 1.0375, "step": 30061 }, { "epoch": 5.73, "learning_rate": 1.283393708293613e-06, "loss": 2.9048, "step": 30062 }, { "epoch": 5.73, "learning_rate": 1.2833365109628219e-06, "loss": 2.3457, "step": 30063 }, { "epoch": 5.73, "learning_rate": 1.2832793136320305e-06, "loss": 3.6274, "step": 30064 }, { "epoch": 5.73, "learning_rate": 1.2832221163012392e-06, "loss": 1.9341, "step": 30065 }, { "epoch": 5.73, "learning_rate": 1.283164918970448e-06, "loss": 0.8125, "step": 30066 }, { "epoch": 5.73, "learning_rate": 1.2831077216396568e-06, "loss": 0.7221, "step": 30067 }, { "epoch": 5.73, "learning_rate": 1.2830505243088656e-06, "loss": 1.6582, "step": 30068 }, { "epoch": 5.73, "learning_rate": 1.2829933269780744e-06, "loss": 3.8669, "step": 30069 }, { "epoch": 5.73, "learning_rate": 1.2829361296472832e-06, "loss": 2.0634, "step": 30070 }, { "epoch": 5.73, "learning_rate": 1.2828789323164918e-06, "loss": 1.4399, "step": 30071 }, { "epoch": 5.73, "learning_rate": 1.2828217349857008e-06, "loss": 1.9226, "step": 30072 }, { "epoch": 5.73, "learning_rate": 1.2827645376549096e-06, "loss": 1.5297, "step": 30073 }, { "epoch": 5.73, "learning_rate": 1.2827073403241182e-06, "loss": 2.1842, "step": 30074 }, { "epoch": 5.73, "learning_rate": 1.282650142993327e-06, "loss": 1.5759, "step": 30075 }, { "epoch": 5.73, "learning_rate": 1.2825929456625358e-06, "loss": 1.3604, "step": 30076 }, { "epoch": 5.73, "learning_rate": 1.2825357483317446e-06, "loss": 1.6471, "step": 30077 }, { "epoch": 5.73, "learning_rate": 1.2824785510009534e-06, "loss": 1.6298, "step": 30078 }, { "epoch": 5.73, "learning_rate": 1.2824213536701621e-06, "loss": 1.4066, "step": 30079 }, { "epoch": 5.73, "learning_rate": 1.282364156339371e-06, "loss": 1.7766, "step": 30080 }, { "epoch": 5.73, "learning_rate": 1.2823069590085795e-06, "loss": 2.0871, "step": 30081 }, { "epoch": 5.73, "learning_rate": 1.2822497616777883e-06, "loss": 2.2253, "step": 30082 }, { "epoch": 5.73, "learning_rate": 1.2821925643469973e-06, "loss": 1.6225, "step": 30083 }, { "epoch": 5.73, "learning_rate": 1.282135367016206e-06, "loss": 1.3171, "step": 30084 }, { "epoch": 5.73, "learning_rate": 1.2820781696854147e-06, "loss": 1.5291, "step": 30085 }, { "epoch": 5.73, "learning_rate": 1.2820209723546235e-06, "loss": 1.3407, "step": 30086 }, { "epoch": 5.73, "learning_rate": 1.2819637750238323e-06, "loss": 1.5825, "step": 30087 }, { "epoch": 5.73, "learning_rate": 1.281906577693041e-06, "loss": 1.5935, "step": 30088 }, { "epoch": 5.73, "learning_rate": 1.2818493803622499e-06, "loss": 2.078, "step": 30089 }, { "epoch": 5.73, "learning_rate": 1.2817921830314587e-06, "loss": 0.8861, "step": 30090 }, { "epoch": 5.73, "learning_rate": 1.2817349857006672e-06, "loss": 1.2643, "step": 30091 }, { "epoch": 5.73, "learning_rate": 1.281677788369876e-06, "loss": 3.8572, "step": 30092 }, { "epoch": 5.73, "learning_rate": 1.2816205910390848e-06, "loss": 2.3367, "step": 30093 }, { "epoch": 5.73, "learning_rate": 1.2815633937082936e-06, "loss": 0.7153, "step": 30094 }, { "epoch": 5.73, "learning_rate": 1.2815061963775024e-06, "loss": 1.3003, "step": 30095 }, { "epoch": 5.73, "learning_rate": 1.2814489990467112e-06, "loss": 1.6928, "step": 30096 }, { "epoch": 5.73, "learning_rate": 1.28139180171592e-06, "loss": 1.4124, "step": 30097 }, { "epoch": 5.73, "learning_rate": 1.2813346043851286e-06, "loss": 1.571, "step": 30098 }, { "epoch": 5.73, "learning_rate": 1.2812774070543374e-06, "loss": 1.9087, "step": 30099 }, { "epoch": 5.73, "learning_rate": 1.2812202097235464e-06, "loss": 1.3818, "step": 30100 }, { "epoch": 5.73, "learning_rate": 1.281163012392755e-06, "loss": 1.9521, "step": 30101 }, { "epoch": 5.73, "learning_rate": 1.2811058150619638e-06, "loss": 1.7025, "step": 30102 }, { "epoch": 5.73, "learning_rate": 1.2810486177311726e-06, "loss": 2.0266, "step": 30103 }, { "epoch": 5.73, "learning_rate": 1.2809914204003814e-06, "loss": 2.5686, "step": 30104 }, { "epoch": 5.73, "learning_rate": 1.2809342230695902e-06, "loss": 1.4555, "step": 30105 }, { "epoch": 5.73, "learning_rate": 1.280877025738799e-06, "loss": 1.7912, "step": 30106 }, { "epoch": 5.73, "learning_rate": 1.2808198284080077e-06, "loss": 0.7956, "step": 30107 }, { "epoch": 5.73, "learning_rate": 1.2807626310772163e-06, "loss": 3.0458, "step": 30108 }, { "epoch": 5.74, "learning_rate": 1.2807054337464251e-06, "loss": 2.5192, "step": 30109 }, { "epoch": 5.74, "learning_rate": 1.280648236415634e-06, "loss": 2.1186, "step": 30110 }, { "epoch": 5.74, "learning_rate": 1.280591039084843e-06, "loss": 1.3097, "step": 30111 }, { "epoch": 5.74, "learning_rate": 1.2805338417540515e-06, "loss": 2.645, "step": 30112 }, { "epoch": 5.74, "learning_rate": 1.2804766444232603e-06, "loss": 1.9191, "step": 30113 }, { "epoch": 5.74, "learning_rate": 1.280419447092469e-06, "loss": 2.4269, "step": 30114 }, { "epoch": 5.74, "learning_rate": 1.2803622497616777e-06, "loss": 1.2781, "step": 30115 }, { "epoch": 5.74, "learning_rate": 1.2803050524308867e-06, "loss": 3.4639, "step": 30116 }, { "epoch": 5.74, "learning_rate": 1.2802478551000955e-06, "loss": 1.186, "step": 30117 }, { "epoch": 5.74, "learning_rate": 1.280190657769304e-06, "loss": 3.4263, "step": 30118 }, { "epoch": 5.74, "learning_rate": 1.2801334604385128e-06, "loss": 1.6042, "step": 30119 }, { "epoch": 5.74, "learning_rate": 1.2800762631077216e-06, "loss": 1.3673, "step": 30120 }, { "epoch": 5.74, "learning_rate": 1.2800190657769304e-06, "loss": 1.8191, "step": 30121 }, { "epoch": 5.74, "learning_rate": 1.2799618684461392e-06, "loss": 0.9419, "step": 30122 }, { "epoch": 5.74, "learning_rate": 1.279904671115348e-06, "loss": 1.7582, "step": 30123 }, { "epoch": 5.74, "learning_rate": 1.2798474737845568e-06, "loss": 1.5878, "step": 30124 }, { "epoch": 5.74, "learning_rate": 1.2797902764537654e-06, "loss": 1.7316, "step": 30125 }, { "epoch": 5.74, "learning_rate": 1.2797330791229742e-06, "loss": 0.6368, "step": 30126 }, { "epoch": 5.74, "learning_rate": 1.2796758817921832e-06, "loss": 1.7012, "step": 30127 }, { "epoch": 5.74, "learning_rate": 1.279618684461392e-06, "loss": 1.4536, "step": 30128 }, { "epoch": 5.74, "learning_rate": 1.2795614871306006e-06, "loss": 1.7896, "step": 30129 }, { "epoch": 5.74, "learning_rate": 1.2795042897998094e-06, "loss": 1.5487, "step": 30130 }, { "epoch": 5.74, "learning_rate": 1.2794470924690182e-06, "loss": 2.4082, "step": 30131 }, { "epoch": 5.74, "learning_rate": 1.279389895138227e-06, "loss": 2.3101, "step": 30132 }, { "epoch": 5.74, "learning_rate": 1.2793326978074357e-06, "loss": 1.6238, "step": 30133 }, { "epoch": 5.74, "learning_rate": 1.2792755004766445e-06, "loss": 2.6812, "step": 30134 }, { "epoch": 5.74, "learning_rate": 1.2792183031458533e-06, "loss": 2.0859, "step": 30135 }, { "epoch": 5.74, "learning_rate": 1.279161105815062e-06, "loss": 2.609, "step": 30136 }, { "epoch": 5.74, "learning_rate": 1.2791039084842707e-06, "loss": 0.499, "step": 30137 }, { "epoch": 5.74, "learning_rate": 1.2790467111534795e-06, "loss": 1.1737, "step": 30138 }, { "epoch": 5.74, "learning_rate": 1.2789895138226883e-06, "loss": 1.5727, "step": 30139 }, { "epoch": 5.74, "learning_rate": 1.278932316491897e-06, "loss": 1.8465, "step": 30140 }, { "epoch": 5.74, "learning_rate": 1.2788751191611059e-06, "loss": 2.5055, "step": 30141 }, { "epoch": 5.74, "learning_rate": 1.2788179218303145e-06, "loss": 2.2304, "step": 30142 }, { "epoch": 5.74, "learning_rate": 1.2787607244995233e-06, "loss": 1.6208, "step": 30143 }, { "epoch": 5.74, "learning_rate": 1.2787035271687323e-06, "loss": 1.2526, "step": 30144 }, { "epoch": 5.74, "learning_rate": 1.278646329837941e-06, "loss": 2.507, "step": 30145 }, { "epoch": 5.74, "learning_rate": 1.2785891325071496e-06, "loss": 1.2694, "step": 30146 }, { "epoch": 5.74, "learning_rate": 1.2785319351763584e-06, "loss": 1.5007, "step": 30147 }, { "epoch": 5.74, "learning_rate": 1.2784747378455672e-06, "loss": 0.6068, "step": 30148 }, { "epoch": 5.74, "learning_rate": 1.278417540514776e-06, "loss": 2.5723, "step": 30149 }, { "epoch": 5.74, "learning_rate": 1.2783603431839848e-06, "loss": 1.7945, "step": 30150 }, { "epoch": 5.74, "learning_rate": 1.2783031458531936e-06, "loss": 3.2172, "step": 30151 }, { "epoch": 5.74, "learning_rate": 1.2782459485224024e-06, "loss": 1.0093, "step": 30152 }, { "epoch": 5.74, "learning_rate": 1.278188751191611e-06, "loss": 1.8025, "step": 30153 }, { "epoch": 5.74, "learning_rate": 1.2781315538608198e-06, "loss": 2.0567, "step": 30154 }, { "epoch": 5.74, "learning_rate": 1.2780743565300288e-06, "loss": 2.4466, "step": 30155 }, { "epoch": 5.74, "learning_rate": 1.2780171591992374e-06, "loss": 2.9644, "step": 30156 }, { "epoch": 5.74, "learning_rate": 1.2779599618684462e-06, "loss": 2.2655, "step": 30157 }, { "epoch": 5.74, "learning_rate": 1.277902764537655e-06, "loss": 2.7119, "step": 30158 }, { "epoch": 5.74, "learning_rate": 1.2778455672068635e-06, "loss": 0.7205, "step": 30159 }, { "epoch": 5.74, "learning_rate": 1.2777883698760725e-06, "loss": 0.844, "step": 30160 }, { "epoch": 5.74, "learning_rate": 1.2777311725452813e-06, "loss": 1.9101, "step": 30161 }, { "epoch": 5.75, "learning_rate": 1.2776739752144901e-06, "loss": 2.2592, "step": 30162 }, { "epoch": 5.75, "learning_rate": 1.2776167778836987e-06, "loss": 2.659, "step": 30163 }, { "epoch": 5.75, "learning_rate": 1.2775595805529075e-06, "loss": 2.438, "step": 30164 }, { "epoch": 5.75, "learning_rate": 1.2775023832221163e-06, "loss": 1.8482, "step": 30165 }, { "epoch": 5.75, "learning_rate": 1.277445185891325e-06, "loss": 1.5441, "step": 30166 }, { "epoch": 5.75, "learning_rate": 1.2773879885605339e-06, "loss": 1.9078, "step": 30167 }, { "epoch": 5.75, "learning_rate": 1.2773307912297427e-06, "loss": 1.6377, "step": 30168 }, { "epoch": 5.75, "learning_rate": 1.2772735938989515e-06, "loss": 1.3843, "step": 30169 }, { "epoch": 5.75, "learning_rate": 1.27721639656816e-06, "loss": 1.6193, "step": 30170 }, { "epoch": 5.75, "learning_rate": 1.277159199237369e-06, "loss": 2.2518, "step": 30171 }, { "epoch": 5.75, "learning_rate": 1.2771020019065779e-06, "loss": 1.8573, "step": 30172 }, { "epoch": 5.75, "learning_rate": 1.2770448045757864e-06, "loss": 1.6748, "step": 30173 }, { "epoch": 5.75, "learning_rate": 1.2769876072449952e-06, "loss": 1.7924, "step": 30174 }, { "epoch": 5.75, "learning_rate": 1.276930409914204e-06, "loss": 0.5998, "step": 30175 }, { "epoch": 5.75, "learning_rate": 1.2768732125834128e-06, "loss": 2.1715, "step": 30176 }, { "epoch": 5.75, "learning_rate": 1.2768160152526216e-06, "loss": 1.0388, "step": 30177 }, { "epoch": 5.75, "learning_rate": 1.2767588179218304e-06, "loss": 0.9094, "step": 30178 }, { "epoch": 5.75, "learning_rate": 1.2767016205910392e-06, "loss": 1.4556, "step": 30179 }, { "epoch": 5.75, "learning_rate": 1.2766444232602478e-06, "loss": 2.1062, "step": 30180 }, { "epoch": 5.75, "learning_rate": 1.2765872259294566e-06, "loss": 2.7614, "step": 30181 }, { "epoch": 5.75, "learning_rate": 1.2765300285986654e-06, "loss": 1.6611, "step": 30182 }, { "epoch": 5.75, "learning_rate": 1.2764728312678742e-06, "loss": 2.4611, "step": 30183 }, { "epoch": 5.75, "learning_rate": 1.276415633937083e-06, "loss": 1.1059, "step": 30184 }, { "epoch": 5.75, "learning_rate": 1.2763584366062918e-06, "loss": 2.5137, "step": 30185 }, { "epoch": 5.75, "learning_rate": 1.2763012392755005e-06, "loss": 2.2439, "step": 30186 }, { "epoch": 5.75, "learning_rate": 1.2762440419447091e-06, "loss": 1.4695, "step": 30187 }, { "epoch": 5.75, "learning_rate": 1.2761868446139181e-06, "loss": 1.0172, "step": 30188 }, { "epoch": 5.75, "learning_rate": 1.276129647283127e-06, "loss": 1.3637, "step": 30189 }, { "epoch": 5.75, "learning_rate": 1.2760724499523355e-06, "loss": 0.7963, "step": 30190 }, { "epoch": 5.75, "learning_rate": 1.2760152526215443e-06, "loss": 1.8772, "step": 30191 }, { "epoch": 5.75, "learning_rate": 1.275958055290753e-06, "loss": 2.6544, "step": 30192 }, { "epoch": 5.75, "learning_rate": 1.2759008579599619e-06, "loss": 1.9446, "step": 30193 }, { "epoch": 5.75, "learning_rate": 1.2758436606291707e-06, "loss": 2.2284, "step": 30194 }, { "epoch": 5.75, "learning_rate": 1.2757864632983795e-06, "loss": 2.4304, "step": 30195 }, { "epoch": 5.75, "learning_rate": 1.2757292659675883e-06, "loss": 1.6229, "step": 30196 }, { "epoch": 5.75, "learning_rate": 1.2756720686367969e-06, "loss": 2.9343, "step": 30197 }, { "epoch": 5.75, "learning_rate": 1.2756148713060056e-06, "loss": 0.5411, "step": 30198 }, { "epoch": 5.75, "learning_rate": 1.2755576739752147e-06, "loss": 2.3734, "step": 30199 }, { "epoch": 5.75, "learning_rate": 1.2755004766444234e-06, "loss": 1.4226, "step": 30200 }, { "epoch": 5.75, "learning_rate": 1.275443279313632e-06, "loss": 1.867, "step": 30201 }, { "epoch": 5.75, "learning_rate": 1.2753860819828408e-06, "loss": 2.051, "step": 30202 }, { "epoch": 5.75, "learning_rate": 1.2753288846520496e-06, "loss": 1.6697, "step": 30203 }, { "epoch": 5.75, "learning_rate": 1.2752716873212584e-06, "loss": 1.9041, "step": 30204 }, { "epoch": 5.75, "learning_rate": 1.2752144899904672e-06, "loss": 0.4435, "step": 30205 }, { "epoch": 5.75, "learning_rate": 1.275157292659676e-06, "loss": 1.5186, "step": 30206 }, { "epoch": 5.75, "learning_rate": 1.2751000953288846e-06, "loss": 2.3004, "step": 30207 }, { "epoch": 5.75, "learning_rate": 1.2750428979980934e-06, "loss": 2.784, "step": 30208 }, { "epoch": 5.75, "learning_rate": 1.2749857006673022e-06, "loss": 1.7652, "step": 30209 }, { "epoch": 5.75, "learning_rate": 1.274928503336511e-06, "loss": 2.9305, "step": 30210 }, { "epoch": 5.75, "learning_rate": 1.2748713060057198e-06, "loss": 1.0706, "step": 30211 }, { "epoch": 5.75, "learning_rate": 1.2748141086749285e-06, "loss": 3.4465, "step": 30212 }, { "epoch": 5.75, "learning_rate": 1.2747569113441373e-06, "loss": 1.9744, "step": 30213 }, { "epoch": 5.76, "learning_rate": 1.274699714013346e-06, "loss": 1.6096, "step": 30214 }, { "epoch": 5.76, "learning_rate": 1.274642516682555e-06, "loss": 1.7492, "step": 30215 }, { "epoch": 5.76, "learning_rate": 1.2745853193517637e-06, "loss": 2.1017, "step": 30216 }, { "epoch": 5.76, "learning_rate": 1.2745281220209725e-06, "loss": 2.2258, "step": 30217 }, { "epoch": 5.76, "learning_rate": 1.274470924690181e-06, "loss": 1.0804, "step": 30218 }, { "epoch": 5.76, "learning_rate": 1.2744137273593899e-06, "loss": 2.0261, "step": 30219 }, { "epoch": 5.76, "learning_rate": 1.2743565300285987e-06, "loss": 2.6569, "step": 30220 }, { "epoch": 5.76, "learning_rate": 1.2742993326978075e-06, "loss": 2.7589, "step": 30221 }, { "epoch": 5.76, "learning_rate": 1.2742421353670163e-06, "loss": 1.9749, "step": 30222 }, { "epoch": 5.76, "learning_rate": 1.274184938036225e-06, "loss": 1.421, "step": 30223 }, { "epoch": 5.76, "learning_rate": 1.2741277407054337e-06, "loss": 1.855, "step": 30224 }, { "epoch": 5.76, "learning_rate": 1.2740705433746424e-06, "loss": 1.9502, "step": 30225 }, { "epoch": 5.76, "learning_rate": 1.2740133460438512e-06, "loss": 0.9105, "step": 30226 }, { "epoch": 5.76, "learning_rate": 1.2739561487130602e-06, "loss": 2.6932, "step": 30227 }, { "epoch": 5.76, "learning_rate": 1.2738989513822688e-06, "loss": 2.7933, "step": 30228 }, { "epoch": 5.76, "learning_rate": 1.2738417540514776e-06, "loss": 2.1641, "step": 30229 }, { "epoch": 5.76, "learning_rate": 1.2737845567206864e-06, "loss": 1.853, "step": 30230 }, { "epoch": 5.76, "learning_rate": 1.273727359389895e-06, "loss": 1.9074, "step": 30231 }, { "epoch": 5.76, "learning_rate": 1.273670162059104e-06, "loss": 1.9998, "step": 30232 }, { "epoch": 5.76, "learning_rate": 1.2736129647283128e-06, "loss": 1.6834, "step": 30233 }, { "epoch": 5.76, "learning_rate": 1.2735557673975216e-06, "loss": 3.2476, "step": 30234 }, { "epoch": 5.76, "learning_rate": 1.2734985700667302e-06, "loss": 3.1437, "step": 30235 }, { "epoch": 5.76, "learning_rate": 1.273441372735939e-06, "loss": 1.5587, "step": 30236 }, { "epoch": 5.76, "learning_rate": 1.2733841754051478e-06, "loss": 1.5427, "step": 30237 }, { "epoch": 5.76, "learning_rate": 1.2733269780743566e-06, "loss": 1.3956, "step": 30238 }, { "epoch": 5.76, "learning_rate": 1.2732697807435653e-06, "loss": 2.9922, "step": 30239 }, { "epoch": 5.76, "learning_rate": 1.2732125834127741e-06, "loss": 2.2598, "step": 30240 }, { "epoch": 5.76, "learning_rate": 1.273155386081983e-06, "loss": 2.3629, "step": 30241 }, { "epoch": 5.76, "learning_rate": 1.2730981887511915e-06, "loss": 2.3941, "step": 30242 }, { "epoch": 5.76, "learning_rate": 1.2730409914204005e-06, "loss": 2.5249, "step": 30243 }, { "epoch": 5.76, "learning_rate": 1.2729837940896093e-06, "loss": 1.6849, "step": 30244 }, { "epoch": 5.76, "learning_rate": 1.272926596758818e-06, "loss": 2.3955, "step": 30245 }, { "epoch": 5.76, "learning_rate": 1.2728693994280267e-06, "loss": 1.3864, "step": 30246 }, { "epoch": 5.76, "learning_rate": 1.2728122020972355e-06, "loss": 3.8873, "step": 30247 }, { "epoch": 5.76, "learning_rate": 1.2727550047664443e-06, "loss": 2.4547, "step": 30248 }, { "epoch": 5.76, "learning_rate": 1.272697807435653e-06, "loss": 2.0459, "step": 30249 }, { "epoch": 5.76, "learning_rate": 1.2726406101048619e-06, "loss": 2.2471, "step": 30250 }, { "epoch": 5.76, "learning_rate": 1.2725834127740707e-06, "loss": 2.0783, "step": 30251 }, { "epoch": 5.76, "learning_rate": 1.2725262154432792e-06, "loss": 1.3195, "step": 30252 }, { "epoch": 5.76, "learning_rate": 1.272469018112488e-06, "loss": 1.2247, "step": 30253 }, { "epoch": 5.76, "learning_rate": 1.2724118207816968e-06, "loss": 1.5837, "step": 30254 }, { "epoch": 5.76, "learning_rate": 1.2723546234509056e-06, "loss": 1.6711, "step": 30255 }, { "epoch": 5.76, "learning_rate": 1.2722974261201144e-06, "loss": 1.7358, "step": 30256 }, { "epoch": 5.76, "learning_rate": 1.2722402287893232e-06, "loss": 1.3946, "step": 30257 }, { "epoch": 5.76, "learning_rate": 1.272183031458532e-06, "loss": 2.5772, "step": 30258 }, { "epoch": 5.76, "learning_rate": 1.2721258341277408e-06, "loss": 1.6997, "step": 30259 }, { "epoch": 5.76, "learning_rate": 1.2720686367969496e-06, "loss": 1.8517, "step": 30260 }, { "epoch": 5.76, "learning_rate": 1.2720114394661584e-06, "loss": 1.7465, "step": 30261 }, { "epoch": 5.76, "learning_rate": 1.271954242135367e-06, "loss": 2.0333, "step": 30262 }, { "epoch": 5.76, "learning_rate": 1.2718970448045758e-06, "loss": 3.6847, "step": 30263 }, { "epoch": 5.76, "learning_rate": 1.2718398474737846e-06, "loss": 1.6589, "step": 30264 }, { "epoch": 5.76, "learning_rate": 1.2717826501429934e-06, "loss": 0.7757, "step": 30265 }, { "epoch": 5.76, "learning_rate": 1.2717254528122021e-06, "loss": 1.6953, "step": 30266 }, { "epoch": 5.77, "learning_rate": 1.271668255481411e-06, "loss": 0.9241, "step": 30267 }, { "epoch": 5.77, "learning_rate": 1.2716110581506197e-06, "loss": 1.5235, "step": 30268 }, { "epoch": 5.77, "learning_rate": 1.2715538608198283e-06, "loss": 0.9657, "step": 30269 }, { "epoch": 5.77, "learning_rate": 1.2714966634890371e-06, "loss": 2.9453, "step": 30270 }, { "epoch": 5.77, "learning_rate": 1.2714394661582461e-06, "loss": 1.0696, "step": 30271 }, { "epoch": 5.77, "learning_rate": 1.2713822688274547e-06, "loss": 3.2593, "step": 30272 }, { "epoch": 5.77, "learning_rate": 1.2713250714966635e-06, "loss": 1.206, "step": 30273 }, { "epoch": 5.77, "learning_rate": 1.2712678741658723e-06, "loss": 2.4465, "step": 30274 }, { "epoch": 5.77, "learning_rate": 1.271210676835081e-06, "loss": 1.5791, "step": 30275 }, { "epoch": 5.77, "learning_rate": 1.2711534795042899e-06, "loss": 1.0028, "step": 30276 }, { "epoch": 5.77, "learning_rate": 1.2710962821734987e-06, "loss": 1.478, "step": 30277 }, { "epoch": 5.77, "learning_rate": 1.2710390848427075e-06, "loss": 2.5675, "step": 30278 }, { "epoch": 5.77, "learning_rate": 1.270981887511916e-06, "loss": 2.9365, "step": 30279 }, { "epoch": 5.77, "learning_rate": 1.2709246901811248e-06, "loss": 2.7053, "step": 30280 }, { "epoch": 5.77, "learning_rate": 1.2708674928503336e-06, "loss": 1.2533, "step": 30281 }, { "epoch": 5.77, "learning_rate": 1.2708102955195426e-06, "loss": 1.2549, "step": 30282 }, { "epoch": 5.77, "learning_rate": 1.2707530981887512e-06, "loss": 2.1586, "step": 30283 }, { "epoch": 5.77, "learning_rate": 1.27069590085796e-06, "loss": 1.8934, "step": 30284 }, { "epoch": 5.77, "learning_rate": 1.2706387035271688e-06, "loss": 1.4754, "step": 30285 }, { "epoch": 5.77, "learning_rate": 1.2705815061963774e-06, "loss": 1.8765, "step": 30286 }, { "epoch": 5.77, "learning_rate": 1.2705243088655864e-06, "loss": 0.9737, "step": 30287 }, { "epoch": 5.77, "learning_rate": 1.2704671115347952e-06, "loss": 0.7307, "step": 30288 }, { "epoch": 5.77, "learning_rate": 1.2704099142040038e-06, "loss": 1.1575, "step": 30289 }, { "epoch": 5.77, "learning_rate": 1.2703527168732126e-06, "loss": 1.7095, "step": 30290 }, { "epoch": 5.77, "learning_rate": 1.2702955195424214e-06, "loss": 3.0039, "step": 30291 }, { "epoch": 5.77, "learning_rate": 1.2702383222116301e-06, "loss": 1.5624, "step": 30292 }, { "epoch": 5.77, "learning_rate": 1.270181124880839e-06, "loss": 1.3296, "step": 30293 }, { "epoch": 5.77, "learning_rate": 1.2701239275500477e-06, "loss": 2.4059, "step": 30294 }, { "epoch": 5.77, "learning_rate": 1.2700667302192565e-06, "loss": 1.1197, "step": 30295 }, { "epoch": 5.77, "learning_rate": 1.2700095328884651e-06, "loss": 1.701, "step": 30296 }, { "epoch": 5.77, "learning_rate": 1.269952335557674e-06, "loss": 2.0542, "step": 30297 }, { "epoch": 5.77, "learning_rate": 1.2698951382268827e-06, "loss": 2.1697, "step": 30298 }, { "epoch": 5.77, "learning_rate": 1.2698379408960917e-06, "loss": 2.6943, "step": 30299 }, { "epoch": 5.77, "learning_rate": 1.2697807435653003e-06, "loss": 1.5551, "step": 30300 }, { "epoch": 5.77, "learning_rate": 1.269723546234509e-06, "loss": 2.8257, "step": 30301 }, { "epoch": 5.77, "learning_rate": 1.2696663489037179e-06, "loss": 1.9981, "step": 30302 }, { "epoch": 5.77, "learning_rate": 1.2696091515729267e-06, "loss": 1.4434, "step": 30303 }, { "epoch": 5.77, "learning_rate": 1.2695519542421355e-06, "loss": 0.7881, "step": 30304 }, { "epoch": 5.77, "learning_rate": 1.2694947569113443e-06, "loss": 1.1514, "step": 30305 }, { "epoch": 5.77, "learning_rate": 1.269437559580553e-06, "loss": 1.4214, "step": 30306 }, { "epoch": 5.77, "learning_rate": 1.2693803622497616e-06, "loss": 1.6469, "step": 30307 }, { "epoch": 5.77, "learning_rate": 1.2693231649189704e-06, "loss": 1.6895, "step": 30308 }, { "epoch": 5.77, "learning_rate": 1.2692659675881792e-06, "loss": 1.8407, "step": 30309 }, { "epoch": 5.77, "learning_rate": 1.269208770257388e-06, "loss": 3.2578, "step": 30310 }, { "epoch": 5.77, "learning_rate": 1.2691515729265968e-06, "loss": 1.481, "step": 30311 }, { "epoch": 5.77, "learning_rate": 1.2690943755958056e-06, "loss": 1.3918, "step": 30312 }, { "epoch": 5.77, "learning_rate": 1.2690371782650142e-06, "loss": 3.0245, "step": 30313 }, { "epoch": 5.77, "learning_rate": 1.268979980934223e-06, "loss": 1.8493, "step": 30314 }, { "epoch": 5.77, "learning_rate": 1.268922783603432e-06, "loss": 1.5887, "step": 30315 }, { "epoch": 5.77, "learning_rate": 1.2688655862726408e-06, "loss": 2.1243, "step": 30316 }, { "epoch": 5.77, "learning_rate": 1.2688083889418494e-06, "loss": 1.9485, "step": 30317 }, { "epoch": 5.77, "learning_rate": 1.2687511916110582e-06, "loss": 1.597, "step": 30318 }, { "epoch": 5.78, "learning_rate": 1.268693994280267e-06, "loss": 2.5479, "step": 30319 }, { "epoch": 5.78, "learning_rate": 1.2686367969494757e-06, "loss": 1.6555, "step": 30320 }, { "epoch": 5.78, "learning_rate": 1.2685795996186845e-06, "loss": 2.2715, "step": 30321 }, { "epoch": 5.78, "learning_rate": 1.2685224022878933e-06, "loss": 2.7035, "step": 30322 }, { "epoch": 5.78, "learning_rate": 1.2684652049571021e-06, "loss": 1.7173, "step": 30323 }, { "epoch": 5.78, "learning_rate": 1.2684080076263107e-06, "loss": 1.9562, "step": 30324 }, { "epoch": 5.78, "learning_rate": 1.2683508102955195e-06, "loss": 1.3493, "step": 30325 }, { "epoch": 5.78, "learning_rate": 1.2682936129647285e-06, "loss": 1.3371, "step": 30326 }, { "epoch": 5.78, "learning_rate": 1.268236415633937e-06, "loss": 1.1809, "step": 30327 }, { "epoch": 5.78, "learning_rate": 1.2681792183031459e-06, "loss": 1.9376, "step": 30328 }, { "epoch": 5.78, "learning_rate": 1.2681220209723547e-06, "loss": 1.994, "step": 30329 }, { "epoch": 5.78, "learning_rate": 1.2680648236415635e-06, "loss": 2.5342, "step": 30330 }, { "epoch": 5.78, "learning_rate": 1.2680076263107723e-06, "loss": 2.8146, "step": 30331 }, { "epoch": 5.78, "learning_rate": 1.267950428979981e-06, "loss": 1.3167, "step": 30332 }, { "epoch": 5.78, "learning_rate": 1.2678932316491898e-06, "loss": 0.8747, "step": 30333 }, { "epoch": 5.78, "learning_rate": 1.2678360343183984e-06, "loss": 1.8407, "step": 30334 }, { "epoch": 5.78, "learning_rate": 1.2677788369876072e-06, "loss": 2.0068, "step": 30335 }, { "epoch": 5.78, "learning_rate": 1.267721639656816e-06, "loss": 0.7319, "step": 30336 }, { "epoch": 5.78, "learning_rate": 1.2676644423260248e-06, "loss": 1.3532, "step": 30337 }, { "epoch": 5.78, "learning_rate": 1.2676072449952336e-06, "loss": 1.8744, "step": 30338 }, { "epoch": 5.78, "learning_rate": 1.2675500476644424e-06, "loss": 1.5341, "step": 30339 }, { "epoch": 5.78, "learning_rate": 1.2674928503336512e-06, "loss": 1.8889, "step": 30340 }, { "epoch": 5.78, "learning_rate": 1.2674356530028598e-06, "loss": 1.4468, "step": 30341 }, { "epoch": 5.78, "learning_rate": 1.2673784556720686e-06, "loss": 3.3609, "step": 30342 }, { "epoch": 5.78, "learning_rate": 1.2673212583412776e-06, "loss": 1.8552, "step": 30343 }, { "epoch": 5.78, "learning_rate": 1.2672640610104862e-06, "loss": 2.237, "step": 30344 }, { "epoch": 5.78, "learning_rate": 1.267206863679695e-06, "loss": 1.5383, "step": 30345 }, { "epoch": 5.78, "learning_rate": 1.2671496663489037e-06, "loss": 2.3964, "step": 30346 }, { "epoch": 5.78, "learning_rate": 1.2670924690181125e-06, "loss": 1.9959, "step": 30347 }, { "epoch": 5.78, "learning_rate": 1.2670352716873213e-06, "loss": 2.6784, "step": 30348 }, { "epoch": 5.78, "learning_rate": 1.2669780743565301e-06, "loss": 3.2668, "step": 30349 }, { "epoch": 5.78, "learning_rate": 1.266920877025739e-06, "loss": 1.7494, "step": 30350 }, { "epoch": 5.78, "learning_rate": 1.2668636796949475e-06, "loss": 1.9769, "step": 30351 }, { "epoch": 5.78, "learning_rate": 1.2668064823641563e-06, "loss": 2.4493, "step": 30352 }, { "epoch": 5.78, "learning_rate": 1.266749285033365e-06, "loss": 2.9045, "step": 30353 }, { "epoch": 5.78, "learning_rate": 1.2666920877025739e-06, "loss": 2.4757, "step": 30354 }, { "epoch": 5.78, "learning_rate": 1.2666348903717827e-06, "loss": 0.9484, "step": 30355 }, { "epoch": 5.78, "learning_rate": 1.2665776930409915e-06, "loss": 1.246, "step": 30356 }, { "epoch": 5.78, "learning_rate": 1.2665204957102003e-06, "loss": 1.1508, "step": 30357 }, { "epoch": 5.78, "learning_rate": 1.2664632983794088e-06, "loss": 0.5093, "step": 30358 }, { "epoch": 5.78, "learning_rate": 1.2664061010486179e-06, "loss": 2.3535, "step": 30359 }, { "epoch": 5.78, "learning_rate": 1.2663489037178266e-06, "loss": 1.4535, "step": 30360 }, { "epoch": 5.78, "learning_rate": 1.2662917063870352e-06, "loss": 2.2611, "step": 30361 }, { "epoch": 5.78, "learning_rate": 1.266234509056244e-06, "loss": 0.9401, "step": 30362 }, { "epoch": 5.78, "learning_rate": 1.2661773117254528e-06, "loss": 0.7909, "step": 30363 }, { "epoch": 5.78, "learning_rate": 1.2661201143946616e-06, "loss": 2.7872, "step": 30364 }, { "epoch": 5.78, "learning_rate": 1.2660629170638704e-06, "loss": 2.5071, "step": 30365 }, { "epoch": 5.78, "learning_rate": 1.2660057197330792e-06, "loss": 1.674, "step": 30366 }, { "epoch": 5.78, "learning_rate": 1.265948522402288e-06, "loss": 2.3847, "step": 30367 }, { "epoch": 5.78, "learning_rate": 1.2658913250714966e-06, "loss": 1.9781, "step": 30368 }, { "epoch": 5.78, "learning_rate": 1.2658341277407054e-06, "loss": 1.8893, "step": 30369 }, { "epoch": 5.78, "learning_rate": 1.2657769304099144e-06, "loss": 0.8948, "step": 30370 }, { "epoch": 5.78, "learning_rate": 1.2657197330791232e-06, "loss": 2.0142, "step": 30371 }, { "epoch": 5.79, "learning_rate": 1.2656625357483317e-06, "loss": 1.2431, "step": 30372 }, { "epoch": 5.79, "learning_rate": 1.2656053384175405e-06, "loss": 3.317, "step": 30373 }, { "epoch": 5.79, "learning_rate": 1.2655481410867493e-06, "loss": 2.6278, "step": 30374 }, { "epoch": 5.79, "learning_rate": 1.2654909437559581e-06, "loss": 1.5703, "step": 30375 }, { "epoch": 5.79, "learning_rate": 1.265433746425167e-06, "loss": 1.6518, "step": 30376 }, { "epoch": 5.79, "learning_rate": 1.2653765490943757e-06, "loss": 2.0383, "step": 30377 }, { "epoch": 5.79, "learning_rate": 1.2653193517635843e-06, "loss": 2.9101, "step": 30378 }, { "epoch": 5.79, "learning_rate": 1.265262154432793e-06, "loss": 2.0428, "step": 30379 }, { "epoch": 5.79, "learning_rate": 1.2652049571020019e-06, "loss": 2.1158, "step": 30380 }, { "epoch": 5.79, "learning_rate": 1.2651477597712107e-06, "loss": 1.1329, "step": 30381 }, { "epoch": 5.79, "learning_rate": 1.2650905624404195e-06, "loss": 2.4125, "step": 30382 }, { "epoch": 5.79, "learning_rate": 1.2650333651096283e-06, "loss": 1.4656, "step": 30383 }, { "epoch": 5.79, "learning_rate": 1.264976167778837e-06, "loss": 1.2637, "step": 30384 }, { "epoch": 5.79, "learning_rate": 1.2649189704480456e-06, "loss": 1.4477, "step": 30385 }, { "epoch": 5.79, "learning_rate": 1.2648617731172544e-06, "loss": 2.4457, "step": 30386 }, { "epoch": 5.79, "learning_rate": 1.2648045757864634e-06, "loss": 1.625, "step": 30387 }, { "epoch": 5.79, "learning_rate": 1.2647473784556722e-06, "loss": 1.6912, "step": 30388 }, { "epoch": 5.79, "learning_rate": 1.2646901811248808e-06, "loss": 1.3442, "step": 30389 }, { "epoch": 5.79, "learning_rate": 1.2646329837940896e-06, "loss": 2.0463, "step": 30390 }, { "epoch": 5.79, "learning_rate": 1.2645757864632984e-06, "loss": 1.4186, "step": 30391 }, { "epoch": 5.79, "learning_rate": 1.2645185891325072e-06, "loss": 3.4686, "step": 30392 }, { "epoch": 5.79, "learning_rate": 1.264461391801716e-06, "loss": 3.5216, "step": 30393 }, { "epoch": 5.79, "learning_rate": 1.2644041944709248e-06, "loss": 1.7001, "step": 30394 }, { "epoch": 5.79, "learning_rate": 1.2643469971401334e-06, "loss": 1.8351, "step": 30395 }, { "epoch": 5.79, "learning_rate": 1.2642897998093422e-06, "loss": 0.8942, "step": 30396 }, { "epoch": 5.79, "learning_rate": 1.264232602478551e-06, "loss": 2.0911, "step": 30397 }, { "epoch": 5.79, "learning_rate": 1.26417540514776e-06, "loss": 1.7289, "step": 30398 }, { "epoch": 5.79, "learning_rate": 1.2641182078169685e-06, "loss": 3.0373, "step": 30399 }, { "epoch": 5.79, "learning_rate": 1.2640610104861773e-06, "loss": 2.5184, "step": 30400 }, { "epoch": 5.79, "learning_rate": 1.2640038131553861e-06, "loss": 2.7343, "step": 30401 }, { "epoch": 5.79, "learning_rate": 1.2639466158245947e-06, "loss": 1.0112, "step": 30402 }, { "epoch": 5.79, "learning_rate": 1.2638894184938037e-06, "loss": 2.3306, "step": 30403 }, { "epoch": 5.79, "learning_rate": 1.2638322211630125e-06, "loss": 2.1725, "step": 30404 }, { "epoch": 5.79, "learning_rate": 1.2637750238322213e-06, "loss": 1.8598, "step": 30405 }, { "epoch": 5.79, "learning_rate": 1.2637178265014299e-06, "loss": 1.5704, "step": 30406 }, { "epoch": 5.79, "learning_rate": 1.2636606291706387e-06, "loss": 0.5871, "step": 30407 }, { "epoch": 5.79, "learning_rate": 1.2636034318398475e-06, "loss": 1.6282, "step": 30408 }, { "epoch": 5.79, "learning_rate": 1.2635462345090563e-06, "loss": 0.4624, "step": 30409 }, { "epoch": 5.79, "learning_rate": 1.263489037178265e-06, "loss": 2.3192, "step": 30410 }, { "epoch": 5.79, "learning_rate": 1.2634318398474739e-06, "loss": 1.7696, "step": 30411 }, { "epoch": 5.79, "learning_rate": 1.2633746425166827e-06, "loss": 2.3613, "step": 30412 }, { "epoch": 5.79, "learning_rate": 1.2633174451858912e-06, "loss": 1.4553, "step": 30413 }, { "epoch": 5.79, "learning_rate": 1.2632602478551002e-06, "loss": 3.6482, "step": 30414 }, { "epoch": 5.79, "learning_rate": 1.263203050524309e-06, "loss": 2.2318, "step": 30415 }, { "epoch": 5.79, "learning_rate": 1.2631458531935176e-06, "loss": 1.7855, "step": 30416 }, { "epoch": 5.79, "learning_rate": 1.2630886558627264e-06, "loss": 3.4959, "step": 30417 }, { "epoch": 5.79, "learning_rate": 1.2630314585319352e-06, "loss": 1.1911, "step": 30418 }, { "epoch": 5.79, "learning_rate": 1.262974261201144e-06, "loss": 1.4706, "step": 30419 }, { "epoch": 5.79, "learning_rate": 1.2629170638703528e-06, "loss": 0.6151, "step": 30420 }, { "epoch": 5.79, "learning_rate": 1.2628598665395616e-06, "loss": 2.3782, "step": 30421 }, { "epoch": 5.79, "learning_rate": 1.2628026692087704e-06, "loss": 2.9828, "step": 30422 }, { "epoch": 5.79, "learning_rate": 1.262745471877979e-06, "loss": 1.5778, "step": 30423 }, { "epoch": 5.8, "learning_rate": 1.2626882745471878e-06, "loss": 1.8189, "step": 30424 }, { "epoch": 5.8, "learning_rate": 1.2626310772163965e-06, "loss": 2.4505, "step": 30425 }, { "epoch": 5.8, "learning_rate": 1.2625738798856053e-06, "loss": 1.1657, "step": 30426 }, { "epoch": 5.8, "learning_rate": 1.2625166825548141e-06, "loss": 2.3773, "step": 30427 }, { "epoch": 5.8, "learning_rate": 1.262459485224023e-06, "loss": 1.2215, "step": 30428 }, { "epoch": 5.8, "learning_rate": 1.2624022878932317e-06, "loss": 2.3759, "step": 30429 }, { "epoch": 5.8, "learning_rate": 1.2623450905624403e-06, "loss": 2.3938, "step": 30430 }, { "epoch": 5.8, "learning_rate": 1.2622878932316493e-06, "loss": 1.4696, "step": 30431 }, { "epoch": 5.8, "learning_rate": 1.2622306959008581e-06, "loss": 1.5173, "step": 30432 }, { "epoch": 5.8, "learning_rate": 1.2621734985700667e-06, "loss": 2.3722, "step": 30433 }, { "epoch": 5.8, "learning_rate": 1.2621163012392755e-06, "loss": 2.6356, "step": 30434 }, { "epoch": 5.8, "learning_rate": 1.2620591039084843e-06, "loss": 2.3087, "step": 30435 }, { "epoch": 5.8, "learning_rate": 1.262001906577693e-06, "loss": 0.9357, "step": 30436 }, { "epoch": 5.8, "learning_rate": 1.2619447092469019e-06, "loss": 1.4883, "step": 30437 }, { "epoch": 5.8, "learning_rate": 1.2618875119161107e-06, "loss": 2.7241, "step": 30438 }, { "epoch": 5.8, "learning_rate": 1.2618303145853195e-06, "loss": 1.5648, "step": 30439 }, { "epoch": 5.8, "learning_rate": 1.261773117254528e-06, "loss": 1.4484, "step": 30440 }, { "epoch": 5.8, "learning_rate": 1.2617159199237368e-06, "loss": 0.8939, "step": 30441 }, { "epoch": 5.8, "learning_rate": 1.2616587225929458e-06, "loss": 1.6297, "step": 30442 }, { "epoch": 5.8, "learning_rate": 1.2616015252621544e-06, "loss": 1.9717, "step": 30443 }, { "epoch": 5.8, "learning_rate": 1.2615443279313632e-06, "loss": 1.677, "step": 30444 }, { "epoch": 5.8, "learning_rate": 1.261487130600572e-06, "loss": 2.9572, "step": 30445 }, { "epoch": 5.8, "learning_rate": 1.2614299332697808e-06, "loss": 1.2581, "step": 30446 }, { "epoch": 5.8, "learning_rate": 1.2613727359389896e-06, "loss": 0.9523, "step": 30447 }, { "epoch": 5.8, "learning_rate": 1.2613155386081984e-06, "loss": 1.0447, "step": 30448 }, { "epoch": 5.8, "learning_rate": 1.2612583412774072e-06, "loss": 2.5526, "step": 30449 }, { "epoch": 5.8, "learning_rate": 1.2612011439466158e-06, "loss": 1.1761, "step": 30450 }, { "epoch": 5.8, "learning_rate": 1.2611439466158246e-06, "loss": 1.1452, "step": 30451 }, { "epoch": 5.8, "learning_rate": 1.2610867492850333e-06, "loss": 1.1793, "step": 30452 }, { "epoch": 5.8, "learning_rate": 1.2610295519542421e-06, "loss": 2.1658, "step": 30453 }, { "epoch": 5.8, "learning_rate": 1.260972354623451e-06, "loss": 2.7388, "step": 30454 }, { "epoch": 5.8, "learning_rate": 1.2609151572926597e-06, "loss": 2.0045, "step": 30455 }, { "epoch": 5.8, "learning_rate": 1.2608579599618685e-06, "loss": 2.1804, "step": 30456 }, { "epoch": 5.8, "learning_rate": 1.260800762631077e-06, "loss": 2.4618, "step": 30457 }, { "epoch": 5.8, "learning_rate": 1.2607435653002861e-06, "loss": 2.4469, "step": 30458 }, { "epoch": 5.8, "learning_rate": 1.260686367969495e-06, "loss": 2.5963, "step": 30459 }, { "epoch": 5.8, "learning_rate": 1.2606291706387035e-06, "loss": 2.2285, "step": 30460 }, { "epoch": 5.8, "learning_rate": 1.2605719733079123e-06, "loss": 0.852, "step": 30461 }, { "epoch": 5.8, "learning_rate": 1.260514775977121e-06, "loss": 1.1954, "step": 30462 }, { "epoch": 5.8, "learning_rate": 1.2604575786463299e-06, "loss": 1.2381, "step": 30463 }, { "epoch": 5.8, "learning_rate": 1.2604003813155387e-06, "loss": 2.1374, "step": 30464 }, { "epoch": 5.8, "learning_rate": 1.2603431839847475e-06, "loss": 2.3351, "step": 30465 }, { "epoch": 5.8, "learning_rate": 1.2602859866539562e-06, "loss": 2.2996, "step": 30466 }, { "epoch": 5.8, "learning_rate": 1.2602287893231648e-06, "loss": 0.5563, "step": 30467 }, { "epoch": 5.8, "learning_rate": 1.2601715919923736e-06, "loss": 1.779, "step": 30468 }, { "epoch": 5.8, "learning_rate": 1.2601143946615824e-06, "loss": 1.6858, "step": 30469 }, { "epoch": 5.8, "learning_rate": 1.2600571973307914e-06, "loss": 0.8731, "step": 30470 }, { "epoch": 5.8, "learning_rate": 1.26e-06, "loss": 1.7309, "step": 30471 }, { "epoch": 5.8, "learning_rate": 1.2599428026692088e-06, "loss": 2.1893, "step": 30472 }, { "epoch": 5.8, "learning_rate": 1.2598856053384176e-06, "loss": 1.9003, "step": 30473 }, { "epoch": 5.8, "learning_rate": 1.2598284080076262e-06, "loss": 2.3492, "step": 30474 }, { "epoch": 5.8, "learning_rate": 1.2597712106768352e-06, "loss": 1.5428, "step": 30475 }, { "epoch": 5.8, "learning_rate": 1.259714013346044e-06, "loss": 2.3164, "step": 30476 }, { "epoch": 5.81, "learning_rate": 1.2596568160152528e-06, "loss": 3.5086, "step": 30477 }, { "epoch": 5.81, "learning_rate": 1.2595996186844614e-06, "loss": 2.526, "step": 30478 }, { "epoch": 5.81, "learning_rate": 1.2595424213536701e-06, "loss": 2.1503, "step": 30479 }, { "epoch": 5.81, "learning_rate": 1.259485224022879e-06, "loss": 1.4361, "step": 30480 }, { "epoch": 5.81, "learning_rate": 1.2594280266920877e-06, "loss": 1.8414, "step": 30481 }, { "epoch": 5.81, "learning_rate": 1.2593708293612965e-06, "loss": 1.3176, "step": 30482 }, { "epoch": 5.81, "learning_rate": 1.2593136320305053e-06, "loss": 1.5954, "step": 30483 }, { "epoch": 5.81, "learning_rate": 1.259256434699714e-06, "loss": 0.8276, "step": 30484 }, { "epoch": 5.81, "learning_rate": 1.2591992373689227e-06, "loss": 1.6801, "step": 30485 }, { "epoch": 5.81, "learning_rate": 1.2591420400381317e-06, "loss": 1.6425, "step": 30486 }, { "epoch": 5.81, "learning_rate": 1.2590848427073405e-06, "loss": 2.3259, "step": 30487 }, { "epoch": 5.81, "learning_rate": 1.259027645376549e-06, "loss": 1.8201, "step": 30488 }, { "epoch": 5.81, "learning_rate": 1.2589704480457579e-06, "loss": 1.0803, "step": 30489 }, { "epoch": 5.81, "learning_rate": 1.2589132507149667e-06, "loss": 1.9992, "step": 30490 }, { "epoch": 5.81, "learning_rate": 1.2588560533841755e-06, "loss": 1.0294, "step": 30491 }, { "epoch": 5.81, "learning_rate": 1.2587988560533843e-06, "loss": 2.0832, "step": 30492 }, { "epoch": 5.81, "learning_rate": 1.258741658722593e-06, "loss": 1.9692, "step": 30493 }, { "epoch": 5.81, "learning_rate": 1.2586844613918018e-06, "loss": 2.896, "step": 30494 }, { "epoch": 5.81, "learning_rate": 1.2586272640610104e-06, "loss": 2.8025, "step": 30495 }, { "epoch": 5.81, "learning_rate": 1.2585700667302192e-06, "loss": 2.5452, "step": 30496 }, { "epoch": 5.81, "learning_rate": 1.258512869399428e-06, "loss": 0.8881, "step": 30497 }, { "epoch": 5.81, "learning_rate": 1.2584556720686368e-06, "loss": 3.8668, "step": 30498 }, { "epoch": 5.81, "learning_rate": 1.2583984747378456e-06, "loss": 1.5169, "step": 30499 }, { "epoch": 5.81, "learning_rate": 1.2583412774070544e-06, "loss": 2.0057, "step": 30500 }, { "epoch": 5.81, "learning_rate": 1.2582840800762632e-06, "loss": 2.7115, "step": 30501 }, { "epoch": 5.81, "learning_rate": 1.258226882745472e-06, "loss": 1.0378, "step": 30502 }, { "epoch": 5.81, "learning_rate": 1.2581696854146808e-06, "loss": 1.2724, "step": 30503 }, { "epoch": 5.81, "learning_rate": 1.2581124880838896e-06, "loss": 1.4603, "step": 30504 }, { "epoch": 5.81, "learning_rate": 1.2580552907530981e-06, "loss": 1.3258, "step": 30505 }, { "epoch": 5.81, "learning_rate": 1.257998093422307e-06, "loss": 0.9862, "step": 30506 }, { "epoch": 5.81, "learning_rate": 1.2579408960915157e-06, "loss": 2.2409, "step": 30507 }, { "epoch": 5.81, "learning_rate": 1.2578836987607245e-06, "loss": 2.4695, "step": 30508 }, { "epoch": 5.81, "learning_rate": 1.2578265014299333e-06, "loss": 2.8387, "step": 30509 }, { "epoch": 5.81, "learning_rate": 1.2577693040991421e-06, "loss": 1.4676, "step": 30510 }, { "epoch": 5.81, "learning_rate": 1.257712106768351e-06, "loss": 2.6343, "step": 30511 }, { "epoch": 5.81, "learning_rate": 1.2576549094375595e-06, "loss": 1.3085, "step": 30512 }, { "epoch": 5.81, "learning_rate": 1.2575977121067683e-06, "loss": 1.7313, "step": 30513 }, { "epoch": 5.81, "learning_rate": 1.2575405147759773e-06, "loss": 1.9364, "step": 30514 }, { "epoch": 5.81, "learning_rate": 1.2574833174451859e-06, "loss": 0.7613, "step": 30515 }, { "epoch": 5.81, "learning_rate": 1.2574261201143947e-06, "loss": 1.3492, "step": 30516 }, { "epoch": 5.81, "learning_rate": 1.2573689227836035e-06, "loss": 1.1011, "step": 30517 }, { "epoch": 5.81, "learning_rate": 1.2573117254528123e-06, "loss": 2.9904, "step": 30518 }, { "epoch": 5.81, "learning_rate": 1.257254528122021e-06, "loss": 1.338, "step": 30519 }, { "epoch": 5.81, "learning_rate": 1.2571973307912298e-06, "loss": 2.106, "step": 30520 }, { "epoch": 5.81, "learning_rate": 1.2571401334604386e-06, "loss": 1.5923, "step": 30521 }, { "epoch": 5.81, "learning_rate": 1.2570829361296472e-06, "loss": 2.0197, "step": 30522 }, { "epoch": 5.81, "learning_rate": 1.257025738798856e-06, "loss": 1.8782, "step": 30523 }, { "epoch": 5.81, "learning_rate": 1.2569685414680648e-06, "loss": 2.5977, "step": 30524 }, { "epoch": 5.81, "learning_rate": 1.2569113441372736e-06, "loss": 2.4953, "step": 30525 }, { "epoch": 5.81, "learning_rate": 1.2568541468064824e-06, "loss": 0.2824, "step": 30526 }, { "epoch": 5.81, "learning_rate": 1.2567969494756912e-06, "loss": 2.6015, "step": 30527 }, { "epoch": 5.81, "learning_rate": 1.2567397521449e-06, "loss": 2.3782, "step": 30528 }, { "epoch": 5.82, "learning_rate": 1.2566825548141086e-06, "loss": 2.8704, "step": 30529 }, { "epoch": 5.82, "learning_rate": 1.2566253574833176e-06, "loss": 1.9003, "step": 30530 }, { "epoch": 5.82, "learning_rate": 1.2565681601525264e-06, "loss": 1.2275, "step": 30531 }, { "epoch": 5.82, "learning_rate": 1.256510962821735e-06, "loss": 2.3046, "step": 30532 }, { "epoch": 5.82, "learning_rate": 1.2564537654909437e-06, "loss": 1.5696, "step": 30533 }, { "epoch": 5.82, "learning_rate": 1.2563965681601525e-06, "loss": 1.9627, "step": 30534 }, { "epoch": 5.82, "learning_rate": 1.2563393708293613e-06, "loss": 2.3964, "step": 30535 }, { "epoch": 5.82, "learning_rate": 1.2562821734985701e-06, "loss": 1.4287, "step": 30536 }, { "epoch": 5.82, "learning_rate": 1.256224976167779e-06, "loss": 1.6761, "step": 30537 }, { "epoch": 5.82, "learning_rate": 1.2561677788369877e-06, "loss": 2.576, "step": 30538 }, { "epoch": 5.82, "learning_rate": 1.2561105815061963e-06, "loss": 1.4854, "step": 30539 }, { "epoch": 5.82, "learning_rate": 1.256053384175405e-06, "loss": 1.9751, "step": 30540 }, { "epoch": 5.82, "learning_rate": 1.2559961868446139e-06, "loss": 0.947, "step": 30541 }, { "epoch": 5.82, "learning_rate": 1.2559389895138229e-06, "loss": 2.6527, "step": 30542 }, { "epoch": 5.82, "learning_rate": 1.2558817921830315e-06, "loss": 1.3825, "step": 30543 }, { "epoch": 5.82, "learning_rate": 1.2558245948522403e-06, "loss": 1.627, "step": 30544 }, { "epoch": 5.82, "learning_rate": 1.255767397521449e-06, "loss": 2.5287, "step": 30545 }, { "epoch": 5.82, "learning_rate": 1.2557102001906576e-06, "loss": 1.8549, "step": 30546 }, { "epoch": 5.82, "learning_rate": 1.2556530028598666e-06, "loss": 1.7402, "step": 30547 }, { "epoch": 5.82, "learning_rate": 1.2555958055290754e-06, "loss": 3.1188, "step": 30548 }, { "epoch": 5.82, "learning_rate": 1.255538608198284e-06, "loss": 3.1811, "step": 30549 }, { "epoch": 5.82, "learning_rate": 1.2554814108674928e-06, "loss": 0.5871, "step": 30550 }, { "epoch": 5.82, "learning_rate": 1.2554242135367016e-06, "loss": 1.4923, "step": 30551 }, { "epoch": 5.82, "learning_rate": 1.2553670162059104e-06, "loss": 1.4249, "step": 30552 }, { "epoch": 5.82, "learning_rate": 1.2553098188751192e-06, "loss": 1.2808, "step": 30553 }, { "epoch": 5.82, "learning_rate": 1.255252621544328e-06, "loss": 1.0697, "step": 30554 }, { "epoch": 5.82, "learning_rate": 1.2551954242135368e-06, "loss": 2.0066, "step": 30555 }, { "epoch": 5.82, "learning_rate": 1.2551382268827454e-06, "loss": 1.8009, "step": 30556 }, { "epoch": 5.82, "learning_rate": 1.2550810295519542e-06, "loss": 2.3378, "step": 30557 }, { "epoch": 5.82, "learning_rate": 1.2550238322211632e-06, "loss": 0.7981, "step": 30558 }, { "epoch": 5.82, "learning_rate": 1.254966634890372e-06, "loss": 1.187, "step": 30559 }, { "epoch": 5.82, "learning_rate": 1.2549094375595805e-06, "loss": 2.4288, "step": 30560 }, { "epoch": 5.82, "learning_rate": 1.2548522402287893e-06, "loss": 0.7263, "step": 30561 }, { "epoch": 5.82, "learning_rate": 1.2547950428979981e-06, "loss": 1.995, "step": 30562 }, { "epoch": 5.82, "learning_rate": 1.254737845567207e-06, "loss": 1.6071, "step": 30563 }, { "epoch": 5.82, "learning_rate": 1.2546806482364157e-06, "loss": 1.7022, "step": 30564 }, { "epoch": 5.82, "learning_rate": 1.2546234509056245e-06, "loss": 2.489, "step": 30565 }, { "epoch": 5.82, "learning_rate": 1.2545662535748333e-06, "loss": 3.0465, "step": 30566 }, { "epoch": 5.82, "learning_rate": 1.2545090562440419e-06, "loss": 3.0189, "step": 30567 }, { "epoch": 5.82, "learning_rate": 1.2544518589132507e-06, "loss": 1.5463, "step": 30568 }, { "epoch": 5.82, "learning_rate": 1.2543946615824597e-06, "loss": 2.7818, "step": 30569 }, { "epoch": 5.82, "learning_rate": 1.2543374642516683e-06, "loss": 2.4192, "step": 30570 }, { "epoch": 5.82, "learning_rate": 1.254280266920877e-06, "loss": 3.3424, "step": 30571 }, { "epoch": 5.82, "learning_rate": 1.2542230695900859e-06, "loss": 0.7731, "step": 30572 }, { "epoch": 5.82, "learning_rate": 1.2541658722592944e-06, "loss": 1.8349, "step": 30573 }, { "epoch": 5.82, "learning_rate": 1.2541086749285034e-06, "loss": 2.0308, "step": 30574 }, { "epoch": 5.82, "learning_rate": 1.2540514775977122e-06, "loss": 2.681, "step": 30575 }, { "epoch": 5.82, "learning_rate": 1.253994280266921e-06, "loss": 1.9022, "step": 30576 }, { "epoch": 5.82, "learning_rate": 1.2539370829361296e-06, "loss": 2.2533, "step": 30577 }, { "epoch": 5.82, "learning_rate": 1.2538798856053384e-06, "loss": 1.8134, "step": 30578 }, { "epoch": 5.82, "learning_rate": 1.2538226882745472e-06, "loss": 2.1718, "step": 30579 }, { "epoch": 5.82, "learning_rate": 1.253765490943756e-06, "loss": 2.8073, "step": 30580 }, { "epoch": 5.82, "learning_rate": 1.2537082936129648e-06, "loss": 1.7909, "step": 30581 }, { "epoch": 5.83, "learning_rate": 1.2536510962821736e-06, "loss": 1.8554, "step": 30582 }, { "epoch": 5.83, "learning_rate": 1.2535938989513824e-06, "loss": 0.9663, "step": 30583 }, { "epoch": 5.83, "learning_rate": 1.253536701620591e-06, "loss": 1.9058, "step": 30584 }, { "epoch": 5.83, "learning_rate": 1.2534795042897997e-06, "loss": 1.9994, "step": 30585 }, { "epoch": 5.83, "learning_rate": 1.2534223069590088e-06, "loss": 2.8097, "step": 30586 }, { "epoch": 5.83, "learning_rate": 1.2533651096282173e-06, "loss": 2.5611, "step": 30587 }, { "epoch": 5.83, "learning_rate": 1.2533079122974261e-06, "loss": 1.9948, "step": 30588 }, { "epoch": 5.83, "learning_rate": 1.253250714966635e-06, "loss": 1.7765, "step": 30589 }, { "epoch": 5.83, "learning_rate": 1.2531935176358435e-06, "loss": 1.3092, "step": 30590 }, { "epoch": 5.83, "learning_rate": 1.2531363203050525e-06, "loss": 2.8517, "step": 30591 }, { "epoch": 5.83, "learning_rate": 1.2530791229742613e-06, "loss": 2.0131, "step": 30592 }, { "epoch": 5.83, "learning_rate": 1.25302192564347e-06, "loss": 2.7542, "step": 30593 }, { "epoch": 5.83, "learning_rate": 1.2529647283126787e-06, "loss": 0.5099, "step": 30594 }, { "epoch": 5.83, "learning_rate": 1.2529075309818875e-06, "loss": 1.1945, "step": 30595 }, { "epoch": 5.83, "learning_rate": 1.2528503336510963e-06, "loss": 2.5239, "step": 30596 }, { "epoch": 5.83, "learning_rate": 1.252793136320305e-06, "loss": 1.4591, "step": 30597 }, { "epoch": 5.83, "learning_rate": 1.2527359389895139e-06, "loss": 2.8365, "step": 30598 }, { "epoch": 5.83, "learning_rate": 1.2526787416587227e-06, "loss": 1.205, "step": 30599 }, { "epoch": 5.83, "learning_rate": 1.2526215443279314e-06, "loss": 1.8753, "step": 30600 }, { "epoch": 5.83, "learning_rate": 1.25256434699714e-06, "loss": 2.304, "step": 30601 }, { "epoch": 5.83, "learning_rate": 1.252507149666349e-06, "loss": 1.0358, "step": 30602 }, { "epoch": 5.83, "learning_rate": 1.2524499523355578e-06, "loss": 1.3223, "step": 30603 }, { "epoch": 5.83, "learning_rate": 1.2523927550047664e-06, "loss": 2.6293, "step": 30604 }, { "epoch": 5.83, "learning_rate": 1.2523355576739752e-06, "loss": 1.6975, "step": 30605 }, { "epoch": 5.83, "learning_rate": 1.252278360343184e-06, "loss": 0.9163, "step": 30606 }, { "epoch": 5.83, "learning_rate": 1.2522211630123928e-06, "loss": 2.3334, "step": 30607 }, { "epoch": 5.83, "learning_rate": 1.2521639656816016e-06, "loss": 1.3409, "step": 30608 }, { "epoch": 5.83, "learning_rate": 1.2521067683508104e-06, "loss": 2.3045, "step": 30609 }, { "epoch": 5.83, "learning_rate": 1.2520495710200192e-06, "loss": 2.0835, "step": 30610 }, { "epoch": 5.83, "learning_rate": 1.2519923736892278e-06, "loss": 2.7383, "step": 30611 }, { "epoch": 5.83, "learning_rate": 1.2519351763584365e-06, "loss": 0.7161, "step": 30612 }, { "epoch": 5.83, "learning_rate": 1.2518779790276456e-06, "loss": 1.2317, "step": 30613 }, { "epoch": 5.83, "learning_rate": 1.2518207816968541e-06, "loss": 1.559, "step": 30614 }, { "epoch": 5.83, "learning_rate": 1.251763584366063e-06, "loss": 2.1749, "step": 30615 }, { "epoch": 5.83, "learning_rate": 1.2517063870352717e-06, "loss": 1.8043, "step": 30616 }, { "epoch": 5.83, "learning_rate": 1.2516491897044805e-06, "loss": 2.4829, "step": 30617 }, { "epoch": 5.83, "learning_rate": 1.2515919923736893e-06, "loss": 2.1166, "step": 30618 }, { "epoch": 5.83, "learning_rate": 1.251534795042898e-06, "loss": 2.0422, "step": 30619 }, { "epoch": 5.83, "learning_rate": 1.251477597712107e-06, "loss": 0.7529, "step": 30620 }, { "epoch": 5.83, "learning_rate": 1.2514204003813155e-06, "loss": 1.5892, "step": 30621 }, { "epoch": 5.83, "learning_rate": 1.2513632030505243e-06, "loss": 0.8094, "step": 30622 }, { "epoch": 5.83, "learning_rate": 1.251306005719733e-06, "loss": 2.4896, "step": 30623 }, { "epoch": 5.83, "learning_rate": 1.2512488083889419e-06, "loss": 2.0202, "step": 30624 }, { "epoch": 5.83, "learning_rate": 1.2511916110581507e-06, "loss": 2.5405, "step": 30625 }, { "epoch": 5.83, "learning_rate": 1.2511344137273594e-06, "loss": 1.465, "step": 30626 }, { "epoch": 5.83, "learning_rate": 1.2510772163965682e-06, "loss": 2.4064, "step": 30627 }, { "epoch": 5.83, "learning_rate": 1.2510200190657768e-06, "loss": 1.5004, "step": 30628 }, { "epoch": 5.83, "learning_rate": 1.2509628217349856e-06, "loss": 1.3105, "step": 30629 }, { "epoch": 5.83, "learning_rate": 1.2509056244041946e-06, "loss": 0.8086, "step": 30630 }, { "epoch": 5.83, "learning_rate": 1.2508484270734034e-06, "loss": 1.3401, "step": 30631 }, { "epoch": 5.83, "learning_rate": 1.250791229742612e-06, "loss": 2.2391, "step": 30632 }, { "epoch": 5.83, "learning_rate": 1.2507340324118208e-06, "loss": 1.307, "step": 30633 }, { "epoch": 5.84, "learning_rate": 1.2506768350810296e-06, "loss": 1.6356, "step": 30634 }, { "epoch": 5.84, "learning_rate": 1.2506196377502384e-06, "loss": 1.7673, "step": 30635 }, { "epoch": 5.84, "learning_rate": 1.2505624404194472e-06, "loss": 2.3969, "step": 30636 }, { "epoch": 5.84, "learning_rate": 1.250505243088656e-06, "loss": 1.5964, "step": 30637 }, { "epoch": 5.84, "learning_rate": 1.2504480457578646e-06, "loss": 2.7149, "step": 30638 }, { "epoch": 5.84, "learning_rate": 1.2503908484270733e-06, "loss": 1.2287, "step": 30639 }, { "epoch": 5.84, "learning_rate": 1.2503336510962821e-06, "loss": 2.1224, "step": 30640 }, { "epoch": 5.84, "learning_rate": 1.2502764537654911e-06, "loss": 1.5916, "step": 30641 }, { "epoch": 5.84, "learning_rate": 1.2502192564346997e-06, "loss": 2.3637, "step": 30642 }, { "epoch": 5.84, "learning_rate": 1.2501620591039085e-06, "loss": 0.4463, "step": 30643 }, { "epoch": 5.84, "learning_rate": 1.2501048617731173e-06, "loss": 2.0363, "step": 30644 }, { "epoch": 5.84, "learning_rate": 1.250047664442326e-06, "loss": 2.8288, "step": 30645 }, { "epoch": 5.84, "learning_rate": 1.249990467111535e-06, "loss": 2.1893, "step": 30646 }, { "epoch": 5.84, "learning_rate": 1.2499332697807437e-06, "loss": 1.7349, "step": 30647 }, { "epoch": 5.84, "learning_rate": 1.2498760724499525e-06, "loss": 2.8431, "step": 30648 }, { "epoch": 5.84, "learning_rate": 1.249818875119161e-06, "loss": 2.2202, "step": 30649 }, { "epoch": 5.84, "learning_rate": 1.2497616777883699e-06, "loss": 1.8383, "step": 30650 }, { "epoch": 5.84, "learning_rate": 1.2497044804575787e-06, "loss": 1.1348, "step": 30651 }, { "epoch": 5.84, "learning_rate": 1.2496472831267875e-06, "loss": 1.3819, "step": 30652 }, { "epoch": 5.84, "learning_rate": 1.2495900857959962e-06, "loss": 2.4431, "step": 30653 }, { "epoch": 5.84, "learning_rate": 1.249532888465205e-06, "loss": 3.0255, "step": 30654 }, { "epoch": 5.84, "learning_rate": 1.2494756911344136e-06, "loss": 1.1229, "step": 30655 }, { "epoch": 5.84, "learning_rate": 1.2494184938036224e-06, "loss": 1.5531, "step": 30656 }, { "epoch": 5.84, "learning_rate": 1.2493612964728314e-06, "loss": 1.1581, "step": 30657 }, { "epoch": 5.84, "learning_rate": 1.2493040991420402e-06, "loss": 1.8473, "step": 30658 }, { "epoch": 5.84, "learning_rate": 1.2492469018112488e-06, "loss": 3.0191, "step": 30659 }, { "epoch": 5.84, "learning_rate": 1.2491897044804576e-06, "loss": 1.2055, "step": 30660 }, { "epoch": 5.84, "learning_rate": 1.2491325071496664e-06, "loss": 2.7951, "step": 30661 }, { "epoch": 5.84, "learning_rate": 1.2490753098188752e-06, "loss": 2.2516, "step": 30662 }, { "epoch": 5.84, "learning_rate": 1.249018112488084e-06, "loss": 3.1286, "step": 30663 }, { "epoch": 5.84, "learning_rate": 1.2489609151572928e-06, "loss": 1.7501, "step": 30664 }, { "epoch": 5.84, "learning_rate": 1.2489037178265016e-06, "loss": 0.9976, "step": 30665 }, { "epoch": 5.84, "learning_rate": 1.2488465204957101e-06, "loss": 0.9463, "step": 30666 }, { "epoch": 5.84, "learning_rate": 1.248789323164919e-06, "loss": 1.6103, "step": 30667 }, { "epoch": 5.84, "learning_rate": 1.2487321258341277e-06, "loss": 1.7995, "step": 30668 }, { "epoch": 5.84, "learning_rate": 1.2486749285033365e-06, "loss": 2.8829, "step": 30669 }, { "epoch": 5.84, "learning_rate": 1.2486177311725453e-06, "loss": 0.673, "step": 30670 }, { "epoch": 5.84, "learning_rate": 1.2485605338417541e-06, "loss": 3.3085, "step": 30671 }, { "epoch": 5.84, "learning_rate": 1.248503336510963e-06, "loss": 1.1332, "step": 30672 }, { "epoch": 5.84, "learning_rate": 1.2484461391801715e-06, "loss": 0.8378, "step": 30673 }, { "epoch": 5.84, "learning_rate": 1.2483889418493805e-06, "loss": 2.3065, "step": 30674 }, { "epoch": 5.84, "learning_rate": 1.2483317445185893e-06, "loss": 1.5062, "step": 30675 }, { "epoch": 5.84, "learning_rate": 1.2482745471877979e-06, "loss": 1.8326, "step": 30676 }, { "epoch": 5.84, "learning_rate": 1.2482173498570067e-06, "loss": 2.3112, "step": 30677 }, { "epoch": 5.84, "learning_rate": 1.2481601525262155e-06, "loss": 2.1984, "step": 30678 }, { "epoch": 5.84, "learning_rate": 1.2481029551954243e-06, "loss": 1.1196, "step": 30679 }, { "epoch": 5.84, "learning_rate": 1.248045757864633e-06, "loss": 1.9772, "step": 30680 }, { "epoch": 5.84, "learning_rate": 1.2479885605338418e-06, "loss": 1.9201, "step": 30681 }, { "epoch": 5.84, "learning_rate": 1.2479313632030506e-06, "loss": 1.9001, "step": 30682 }, { "epoch": 5.84, "learning_rate": 1.2478741658722592e-06, "loss": 1.757, "step": 30683 }, { "epoch": 5.84, "learning_rate": 1.247816968541468e-06, "loss": 1.3341, "step": 30684 }, { "epoch": 5.84, "learning_rate": 1.247759771210677e-06, "loss": 2.4906, "step": 30685 }, { "epoch": 5.84, "learning_rate": 1.2477025738798856e-06, "loss": 0.7722, "step": 30686 }, { "epoch": 5.85, "learning_rate": 1.2476453765490944e-06, "loss": 1.337, "step": 30687 }, { "epoch": 5.85, "learning_rate": 1.2475881792183032e-06, "loss": 1.6084, "step": 30688 }, { "epoch": 5.85, "learning_rate": 1.247530981887512e-06, "loss": 1.8441, "step": 30689 }, { "epoch": 5.85, "learning_rate": 1.2474737845567208e-06, "loss": 1.3334, "step": 30690 }, { "epoch": 5.85, "learning_rate": 1.2474165872259296e-06, "loss": 1.488, "step": 30691 }, { "epoch": 5.85, "learning_rate": 1.2473593898951384e-06, "loss": 2.1686, "step": 30692 }, { "epoch": 5.85, "learning_rate": 1.247302192564347e-06, "loss": 2.1408, "step": 30693 }, { "epoch": 5.85, "learning_rate": 1.2472449952335557e-06, "loss": 0.9023, "step": 30694 }, { "epoch": 5.85, "learning_rate": 1.2471877979027645e-06, "loss": 2.6002, "step": 30695 }, { "epoch": 5.85, "learning_rate": 1.2471306005719733e-06, "loss": 3.0619, "step": 30696 }, { "epoch": 5.85, "learning_rate": 1.2470734032411821e-06, "loss": 1.7071, "step": 30697 }, { "epoch": 5.85, "learning_rate": 1.247016205910391e-06, "loss": 1.1018, "step": 30698 }, { "epoch": 5.85, "learning_rate": 1.2469590085795997e-06, "loss": 1.017, "step": 30699 }, { "epoch": 5.85, "learning_rate": 1.2469018112488083e-06, "loss": 1.6772, "step": 30700 }, { "epoch": 5.85, "learning_rate": 1.246844613918017e-06, "loss": 1.7686, "step": 30701 }, { "epoch": 5.85, "learning_rate": 1.246787416587226e-06, "loss": 1.0221, "step": 30702 }, { "epoch": 5.85, "learning_rate": 1.2467302192564347e-06, "loss": 1.7628, "step": 30703 }, { "epoch": 5.85, "learning_rate": 1.2466730219256435e-06, "loss": 1.8342, "step": 30704 }, { "epoch": 5.85, "learning_rate": 1.2466158245948523e-06, "loss": 1.2251, "step": 30705 }, { "epoch": 5.85, "learning_rate": 1.246558627264061e-06, "loss": 3.1958, "step": 30706 }, { "epoch": 5.85, "learning_rate": 1.2465014299332698e-06, "loss": 2.6183, "step": 30707 }, { "epoch": 5.85, "learning_rate": 1.2464442326024786e-06, "loss": 2.4203, "step": 30708 }, { "epoch": 5.85, "learning_rate": 1.2463870352716874e-06, "loss": 1.2798, "step": 30709 }, { "epoch": 5.85, "learning_rate": 1.246329837940896e-06, "loss": 1.1171, "step": 30710 }, { "epoch": 5.85, "learning_rate": 1.2462726406101048e-06, "loss": 2.1525, "step": 30711 }, { "epoch": 5.85, "learning_rate": 1.2462154432793136e-06, "loss": 3.5067, "step": 30712 }, { "epoch": 5.85, "learning_rate": 1.2461582459485226e-06, "loss": 1.298, "step": 30713 }, { "epoch": 5.85, "learning_rate": 1.2461010486177312e-06, "loss": 2.0733, "step": 30714 }, { "epoch": 5.85, "learning_rate": 1.24604385128694e-06, "loss": 2.0488, "step": 30715 }, { "epoch": 5.85, "learning_rate": 1.2459866539561488e-06, "loss": 1.7662, "step": 30716 }, { "epoch": 5.85, "learning_rate": 1.2459294566253574e-06, "loss": 0.4843, "step": 30717 }, { "epoch": 5.85, "learning_rate": 1.2458722592945664e-06, "loss": 2.2591, "step": 30718 }, { "epoch": 5.85, "learning_rate": 1.2458150619637752e-06, "loss": 2.2437, "step": 30719 }, { "epoch": 5.85, "learning_rate": 1.2457578646329837e-06, "loss": 2.0006, "step": 30720 }, { "epoch": 5.85, "learning_rate": 1.2457006673021925e-06, "loss": 1.5684, "step": 30721 }, { "epoch": 5.85, "learning_rate": 1.2456434699714013e-06, "loss": 1.3381, "step": 30722 }, { "epoch": 5.85, "learning_rate": 1.2455862726406101e-06, "loss": 2.197, "step": 30723 }, { "epoch": 5.85, "learning_rate": 1.245529075309819e-06, "loss": 1.6951, "step": 30724 }, { "epoch": 5.85, "learning_rate": 1.2454718779790277e-06, "loss": 2.3238, "step": 30725 }, { "epoch": 5.85, "learning_rate": 1.2454146806482365e-06, "loss": 1.3279, "step": 30726 }, { "epoch": 5.85, "learning_rate": 1.245357483317445e-06, "loss": 1.4005, "step": 30727 }, { "epoch": 5.85, "learning_rate": 1.2453002859866539e-06, "loss": 1.2862, "step": 30728 }, { "epoch": 5.85, "learning_rate": 1.2452430886558629e-06, "loss": 1.4594, "step": 30729 }, { "epoch": 5.85, "learning_rate": 1.2451858913250717e-06, "loss": 0.9794, "step": 30730 }, { "epoch": 5.85, "learning_rate": 1.2451286939942803e-06, "loss": 3.1457, "step": 30731 }, { "epoch": 5.85, "learning_rate": 1.245071496663489e-06, "loss": 2.1467, "step": 30732 }, { "epoch": 5.85, "learning_rate": 1.2450142993326978e-06, "loss": 1.2476, "step": 30733 }, { "epoch": 5.85, "learning_rate": 1.2449571020019066e-06, "loss": 3.4562, "step": 30734 }, { "epoch": 5.85, "learning_rate": 1.2448999046711154e-06, "loss": 3.6859, "step": 30735 }, { "epoch": 5.85, "learning_rate": 1.2448427073403242e-06, "loss": 2.0584, "step": 30736 }, { "epoch": 5.85, "learning_rate": 1.244785510009533e-06, "loss": 3.198, "step": 30737 }, { "epoch": 5.85, "learning_rate": 1.2447283126787416e-06, "loss": 2.1127, "step": 30738 }, { "epoch": 5.86, "learning_rate": 1.2446711153479504e-06, "loss": 1.7089, "step": 30739 }, { "epoch": 5.86, "learning_rate": 1.2446139180171592e-06, "loss": 3.7097, "step": 30740 }, { "epoch": 5.86, "learning_rate": 1.244556720686368e-06, "loss": 1.7459, "step": 30741 }, { "epoch": 5.86, "learning_rate": 1.2444995233555768e-06, "loss": 1.5309, "step": 30742 }, { "epoch": 5.86, "learning_rate": 1.2444423260247856e-06, "loss": 1.2765, "step": 30743 }, { "epoch": 5.86, "learning_rate": 1.2443851286939942e-06, "loss": 1.7751, "step": 30744 }, { "epoch": 5.86, "learning_rate": 1.244327931363203e-06, "loss": 1.8484, "step": 30745 }, { "epoch": 5.86, "learning_rate": 1.244270734032412e-06, "loss": 1.7435, "step": 30746 }, { "epoch": 5.86, "learning_rate": 1.2442135367016207e-06, "loss": 1.5577, "step": 30747 }, { "epoch": 5.86, "learning_rate": 1.2441563393708293e-06, "loss": 0.9895, "step": 30748 }, { "epoch": 5.86, "learning_rate": 1.2440991420400381e-06, "loss": 1.1453, "step": 30749 }, { "epoch": 5.86, "learning_rate": 1.244041944709247e-06, "loss": 3.3128, "step": 30750 }, { "epoch": 5.86, "learning_rate": 1.2439847473784557e-06, "loss": 1.0099, "step": 30751 }, { "epoch": 5.86, "learning_rate": 1.2439275500476645e-06, "loss": 1.2046, "step": 30752 }, { "epoch": 5.86, "learning_rate": 1.2438703527168733e-06, "loss": 2.7174, "step": 30753 }, { "epoch": 5.86, "learning_rate": 1.243813155386082e-06, "loss": 1.8423, "step": 30754 }, { "epoch": 5.86, "learning_rate": 1.2437559580552907e-06, "loss": 1.3972, "step": 30755 }, { "epoch": 5.86, "learning_rate": 1.2436987607244995e-06, "loss": 1.5702, "step": 30756 }, { "epoch": 5.86, "learning_rate": 1.2436415633937085e-06, "loss": 1.9374, "step": 30757 }, { "epoch": 5.86, "learning_rate": 1.243584366062917e-06, "loss": 1.6713, "step": 30758 }, { "epoch": 5.86, "learning_rate": 1.2435271687321258e-06, "loss": 1.9765, "step": 30759 }, { "epoch": 5.86, "learning_rate": 1.2434699714013346e-06, "loss": 1.8428, "step": 30760 }, { "epoch": 5.86, "learning_rate": 1.2434127740705432e-06, "loss": 1.9231, "step": 30761 }, { "epoch": 5.86, "learning_rate": 1.2433555767397522e-06, "loss": 2.7292, "step": 30762 }, { "epoch": 5.86, "learning_rate": 1.243298379408961e-06, "loss": 1.9127, "step": 30763 }, { "epoch": 5.86, "learning_rate": 1.2432411820781698e-06, "loss": 2.579, "step": 30764 }, { "epoch": 5.86, "learning_rate": 1.2431839847473784e-06, "loss": 2.025, "step": 30765 }, { "epoch": 5.86, "learning_rate": 1.2431267874165872e-06, "loss": 1.4421, "step": 30766 }, { "epoch": 5.86, "learning_rate": 1.243069590085796e-06, "loss": 1.6296, "step": 30767 }, { "epoch": 5.86, "learning_rate": 1.2430123927550048e-06, "loss": 2.2937, "step": 30768 }, { "epoch": 5.86, "learning_rate": 1.2429551954242136e-06, "loss": 1.257, "step": 30769 }, { "epoch": 5.86, "learning_rate": 1.2428979980934224e-06, "loss": 1.9544, "step": 30770 }, { "epoch": 5.86, "learning_rate": 1.2428408007626312e-06, "loss": 2.2601, "step": 30771 }, { "epoch": 5.86, "learning_rate": 1.2427836034318397e-06, "loss": 2.3443, "step": 30772 }, { "epoch": 5.86, "learning_rate": 1.2427264061010488e-06, "loss": 0.9735, "step": 30773 }, { "epoch": 5.86, "learning_rate": 1.2426692087702575e-06, "loss": 1.3255, "step": 30774 }, { "epoch": 5.86, "learning_rate": 1.2426120114394661e-06, "loss": 1.8237, "step": 30775 }, { "epoch": 5.86, "learning_rate": 1.242554814108675e-06, "loss": 0.558, "step": 30776 }, { "epoch": 5.86, "learning_rate": 1.2424976167778837e-06, "loss": 1.9899, "step": 30777 }, { "epoch": 5.86, "learning_rate": 1.2424404194470925e-06, "loss": 1.5581, "step": 30778 }, { "epoch": 5.86, "learning_rate": 1.2423832221163013e-06, "loss": 2.4209, "step": 30779 }, { "epoch": 5.86, "learning_rate": 1.24232602478551e-06, "loss": 2.1085, "step": 30780 }, { "epoch": 5.86, "learning_rate": 1.2422688274547189e-06, "loss": 2.6442, "step": 30781 }, { "epoch": 5.86, "learning_rate": 1.2422116301239275e-06, "loss": 1.3142, "step": 30782 }, { "epoch": 5.86, "learning_rate": 1.2421544327931363e-06, "loss": 1.6482, "step": 30783 }, { "epoch": 5.86, "learning_rate": 1.242097235462345e-06, "loss": 2.066, "step": 30784 }, { "epoch": 5.86, "learning_rate": 1.2420400381315539e-06, "loss": 1.1382, "step": 30785 }, { "epoch": 5.86, "learning_rate": 1.2419828408007626e-06, "loss": 2.3285, "step": 30786 }, { "epoch": 5.86, "learning_rate": 1.2419256434699714e-06, "loss": 2.0639, "step": 30787 }, { "epoch": 5.86, "learning_rate": 1.2418684461391802e-06, "loss": 1.7165, "step": 30788 }, { "epoch": 5.86, "learning_rate": 1.2418112488083888e-06, "loss": 3.0538, "step": 30789 }, { "epoch": 5.86, "learning_rate": 1.2417540514775978e-06, "loss": 3.7595, "step": 30790 }, { "epoch": 5.86, "learning_rate": 1.2416968541468066e-06, "loss": 1.4857, "step": 30791 }, { "epoch": 5.87, "learning_rate": 1.2416396568160152e-06, "loss": 2.4781, "step": 30792 }, { "epoch": 5.87, "learning_rate": 1.241582459485224e-06, "loss": 1.9108, "step": 30793 }, { "epoch": 5.87, "learning_rate": 1.2415252621544328e-06, "loss": 1.3979, "step": 30794 }, { "epoch": 5.87, "learning_rate": 1.2414680648236416e-06, "loss": 2.5215, "step": 30795 }, { "epoch": 5.87, "learning_rate": 1.2414108674928504e-06, "loss": 1.8707, "step": 30796 }, { "epoch": 5.87, "learning_rate": 1.2413536701620592e-06, "loss": 1.3812, "step": 30797 }, { "epoch": 5.87, "learning_rate": 1.241296472831268e-06, "loss": 2.0084, "step": 30798 }, { "epoch": 5.87, "learning_rate": 1.2412392755004765e-06, "loss": 3.0374, "step": 30799 }, { "epoch": 5.87, "learning_rate": 1.2411820781696853e-06, "loss": 2.2897, "step": 30800 }, { "epoch": 5.87, "learning_rate": 1.2411248808388943e-06, "loss": 3.4929, "step": 30801 }, { "epoch": 5.87, "learning_rate": 1.2410676835081031e-06, "loss": 1.319, "step": 30802 }, { "epoch": 5.87, "learning_rate": 1.2410104861773117e-06, "loss": 0.8138, "step": 30803 }, { "epoch": 5.87, "learning_rate": 1.2409532888465205e-06, "loss": 1.7763, "step": 30804 }, { "epoch": 5.87, "learning_rate": 1.2408960915157293e-06, "loss": 1.9068, "step": 30805 }, { "epoch": 5.87, "learning_rate": 1.240838894184938e-06, "loss": 1.8909, "step": 30806 }, { "epoch": 5.87, "learning_rate": 1.240781696854147e-06, "loss": 1.5221, "step": 30807 }, { "epoch": 5.87, "learning_rate": 1.2407244995233557e-06, "loss": 1.6884, "step": 30808 }, { "epoch": 5.87, "learning_rate": 1.2406673021925643e-06, "loss": 2.1696, "step": 30809 }, { "epoch": 5.87, "learning_rate": 1.240610104861773e-06, "loss": 1.6796, "step": 30810 }, { "epoch": 5.87, "learning_rate": 1.2405529075309819e-06, "loss": 2.6003, "step": 30811 }, { "epoch": 5.87, "learning_rate": 1.2404957102001909e-06, "loss": 2.7917, "step": 30812 }, { "epoch": 5.87, "learning_rate": 1.2404385128693994e-06, "loss": 0.6685, "step": 30813 }, { "epoch": 5.87, "learning_rate": 1.2403813155386082e-06, "loss": 2.0732, "step": 30814 }, { "epoch": 5.87, "learning_rate": 1.240324118207817e-06, "loss": 1.5045, "step": 30815 }, { "epoch": 5.87, "learning_rate": 1.2402669208770256e-06, "loss": 2.597, "step": 30816 }, { "epoch": 5.87, "learning_rate": 1.2402097235462346e-06, "loss": 1.5037, "step": 30817 }, { "epoch": 5.87, "learning_rate": 1.2401525262154434e-06, "loss": 0.9165, "step": 30818 }, { "epoch": 5.87, "learning_rate": 1.2400953288846522e-06, "loss": 0.8028, "step": 30819 }, { "epoch": 5.87, "learning_rate": 1.2400381315538608e-06, "loss": 3.2387, "step": 30820 }, { "epoch": 5.87, "learning_rate": 1.2399809342230696e-06, "loss": 1.5642, "step": 30821 }, { "epoch": 5.87, "learning_rate": 1.2399237368922784e-06, "loss": 1.1868, "step": 30822 }, { "epoch": 5.87, "learning_rate": 1.2398665395614872e-06, "loss": 0.362, "step": 30823 }, { "epoch": 5.87, "learning_rate": 1.239809342230696e-06, "loss": 2.6217, "step": 30824 }, { "epoch": 5.87, "learning_rate": 1.2397521448999048e-06, "loss": 1.4055, "step": 30825 }, { "epoch": 5.87, "learning_rate": 1.2396949475691133e-06, "loss": 2.5663, "step": 30826 }, { "epoch": 5.87, "learning_rate": 1.2396377502383221e-06, "loss": 1.9335, "step": 30827 }, { "epoch": 5.87, "learning_rate": 1.239580552907531e-06, "loss": 1.3468, "step": 30828 }, { "epoch": 5.87, "learning_rate": 1.23952335557674e-06, "loss": 3.4027, "step": 30829 }, { "epoch": 5.87, "learning_rate": 1.2394661582459485e-06, "loss": 1.9089, "step": 30830 }, { "epoch": 5.87, "learning_rate": 1.2394089609151573e-06, "loss": 1.8028, "step": 30831 }, { "epoch": 5.87, "learning_rate": 1.239351763584366e-06, "loss": 1.964, "step": 30832 }, { "epoch": 5.87, "learning_rate": 1.2392945662535747e-06, "loss": 1.5548, "step": 30833 }, { "epoch": 5.87, "learning_rate": 1.2392373689227837e-06, "loss": 1.9264, "step": 30834 }, { "epoch": 5.87, "learning_rate": 1.2391801715919925e-06, "loss": 0.9209, "step": 30835 }, { "epoch": 5.87, "learning_rate": 1.2391229742612013e-06, "loss": 1.4061, "step": 30836 }, { "epoch": 5.87, "learning_rate": 1.2390657769304099e-06, "loss": 3.1287, "step": 30837 }, { "epoch": 5.87, "learning_rate": 1.2390085795996187e-06, "loss": 2.7628, "step": 30838 }, { "epoch": 5.87, "learning_rate": 1.2389513822688274e-06, "loss": 1.8326, "step": 30839 }, { "epoch": 5.87, "learning_rate": 1.2388941849380362e-06, "loss": 1.7327, "step": 30840 }, { "epoch": 5.87, "learning_rate": 1.238836987607245e-06, "loss": 2.2009, "step": 30841 }, { "epoch": 5.87, "learning_rate": 1.2387797902764538e-06, "loss": 1.2441, "step": 30842 }, { "epoch": 5.87, "learning_rate": 1.2387225929456626e-06, "loss": 1.342, "step": 30843 }, { "epoch": 5.88, "learning_rate": 1.2386653956148712e-06, "loss": 1.9481, "step": 30844 }, { "epoch": 5.88, "learning_rate": 1.2386081982840802e-06, "loss": 2.5764, "step": 30845 }, { "epoch": 5.88, "learning_rate": 1.238551000953289e-06, "loss": 0.7993, "step": 30846 }, { "epoch": 5.88, "learning_rate": 1.2384938036224976e-06, "loss": 1.8552, "step": 30847 }, { "epoch": 5.88, "learning_rate": 1.2384366062917064e-06, "loss": 2.3951, "step": 30848 }, { "epoch": 5.88, "learning_rate": 1.2383794089609152e-06, "loss": 1.9793, "step": 30849 }, { "epoch": 5.88, "learning_rate": 1.238322211630124e-06, "loss": 2.6035, "step": 30850 }, { "epoch": 5.88, "learning_rate": 1.2382650142993328e-06, "loss": 0.8149, "step": 30851 }, { "epoch": 5.88, "learning_rate": 1.2382078169685416e-06, "loss": 2.5361, "step": 30852 }, { "epoch": 5.88, "learning_rate": 1.2381506196377504e-06, "loss": 1.4361, "step": 30853 }, { "epoch": 5.88, "learning_rate": 1.238093422306959e-06, "loss": 1.1135, "step": 30854 }, { "epoch": 5.88, "learning_rate": 1.2380362249761677e-06, "loss": 2.0002, "step": 30855 }, { "epoch": 5.88, "learning_rate": 1.2379790276453765e-06, "loss": 0.4596, "step": 30856 }, { "epoch": 5.88, "learning_rate": 1.2379218303145853e-06, "loss": 1.567, "step": 30857 }, { "epoch": 5.88, "learning_rate": 1.2378646329837941e-06, "loss": 1.5666, "step": 30858 }, { "epoch": 5.88, "learning_rate": 1.237807435653003e-06, "loss": 0.7511, "step": 30859 }, { "epoch": 5.88, "learning_rate": 1.2377502383222117e-06, "loss": 1.7553, "step": 30860 }, { "epoch": 5.88, "learning_rate": 1.2376930409914205e-06, "loss": 2.5125, "step": 30861 }, { "epoch": 5.88, "learning_rate": 1.2376358436606293e-06, "loss": 1.302, "step": 30862 }, { "epoch": 5.88, "learning_rate": 1.237578646329838e-06, "loss": 0.8556, "step": 30863 }, { "epoch": 5.88, "learning_rate": 1.2375214489990467e-06, "loss": 1.3647, "step": 30864 }, { "epoch": 5.88, "learning_rate": 1.2374642516682555e-06, "loss": 1.3011, "step": 30865 }, { "epoch": 5.88, "learning_rate": 1.2374070543374642e-06, "loss": 2.7838, "step": 30866 }, { "epoch": 5.88, "learning_rate": 1.237349857006673e-06, "loss": 1.83, "step": 30867 }, { "epoch": 5.88, "learning_rate": 1.2372926596758818e-06, "loss": 3.0019, "step": 30868 }, { "epoch": 5.88, "learning_rate": 1.2372354623450906e-06, "loss": 2.4374, "step": 30869 }, { "epoch": 5.88, "learning_rate": 1.2371782650142994e-06, "loss": 2.8686, "step": 30870 }, { "epoch": 5.88, "learning_rate": 1.237121067683508e-06, "loss": 3.3684, "step": 30871 }, { "epoch": 5.88, "learning_rate": 1.2370638703527168e-06, "loss": 0.4422, "step": 30872 }, { "epoch": 5.88, "learning_rate": 1.2370066730219258e-06, "loss": 1.4914, "step": 30873 }, { "epoch": 5.88, "learning_rate": 1.2369494756911344e-06, "loss": 1.5997, "step": 30874 }, { "epoch": 5.88, "learning_rate": 1.2368922783603432e-06, "loss": 1.5919, "step": 30875 }, { "epoch": 5.88, "learning_rate": 1.236835081029552e-06, "loss": 0.6831, "step": 30876 }, { "epoch": 5.88, "learning_rate": 1.2367778836987608e-06, "loss": 2.9486, "step": 30877 }, { "epoch": 5.88, "learning_rate": 1.2367206863679696e-06, "loss": 2.8752, "step": 30878 }, { "epoch": 5.88, "learning_rate": 1.2366634890371784e-06, "loss": 1.7422, "step": 30879 }, { "epoch": 5.88, "learning_rate": 1.2366062917063871e-06, "loss": 1.6374, "step": 30880 }, { "epoch": 5.88, "learning_rate": 1.2365490943755957e-06, "loss": 1.6677, "step": 30881 }, { "epoch": 5.88, "learning_rate": 1.2364918970448045e-06, "loss": 1.2795, "step": 30882 }, { "epoch": 5.88, "learning_rate": 1.2364346997140133e-06, "loss": 1.8612, "step": 30883 }, { "epoch": 5.88, "learning_rate": 1.2363775023832223e-06, "loss": 1.8464, "step": 30884 }, { "epoch": 5.88, "learning_rate": 1.236320305052431e-06, "loss": 1.2011, "step": 30885 }, { "epoch": 5.88, "learning_rate": 1.2362631077216397e-06, "loss": 1.733, "step": 30886 }, { "epoch": 5.88, "learning_rate": 1.2362059103908485e-06, "loss": 1.4799, "step": 30887 }, { "epoch": 5.88, "learning_rate": 1.236148713060057e-06, "loss": 1.6021, "step": 30888 }, { "epoch": 5.88, "learning_rate": 1.236091515729266e-06, "loss": 0.9198, "step": 30889 }, { "epoch": 5.88, "learning_rate": 1.2360343183984749e-06, "loss": 2.689, "step": 30890 }, { "epoch": 5.88, "learning_rate": 1.2359771210676835e-06, "loss": 1.0442, "step": 30891 }, { "epoch": 5.88, "learning_rate": 1.2359199237368923e-06, "loss": 3.7231, "step": 30892 }, { "epoch": 5.88, "learning_rate": 1.235862726406101e-06, "loss": 2.109, "step": 30893 }, { "epoch": 5.88, "learning_rate": 1.2358055290753098e-06, "loss": 2.7307, "step": 30894 }, { "epoch": 5.88, "learning_rate": 1.2357483317445186e-06, "loss": 1.8614, "step": 30895 }, { "epoch": 5.88, "learning_rate": 1.2356911344137274e-06, "loss": 2.2863, "step": 30896 }, { "epoch": 5.89, "learning_rate": 1.2356339370829362e-06, "loss": 1.3319, "step": 30897 }, { "epoch": 5.89, "learning_rate": 1.2355767397521448e-06, "loss": 1.8539, "step": 30898 }, { "epoch": 5.89, "learning_rate": 1.2355195424213536e-06, "loss": 3.3089, "step": 30899 }, { "epoch": 5.89, "learning_rate": 1.2354623450905624e-06, "loss": 0.9698, "step": 30900 }, { "epoch": 5.89, "learning_rate": 1.2354051477597714e-06, "loss": 2.8067, "step": 30901 }, { "epoch": 5.89, "learning_rate": 1.23534795042898e-06, "loss": 0.8408, "step": 30902 }, { "epoch": 5.89, "learning_rate": 1.2352907530981888e-06, "loss": 1.8836, "step": 30903 }, { "epoch": 5.89, "learning_rate": 1.2352335557673976e-06, "loss": 2.2397, "step": 30904 }, { "epoch": 5.89, "learning_rate": 1.2351763584366064e-06, "loss": 2.1111, "step": 30905 }, { "epoch": 5.89, "learning_rate": 1.2351191611058152e-06, "loss": 2.7339, "step": 30906 }, { "epoch": 5.89, "learning_rate": 1.235061963775024e-06, "loss": 1.1794, "step": 30907 }, { "epoch": 5.89, "learning_rate": 1.2350047664442327e-06, "loss": 3.14, "step": 30908 }, { "epoch": 5.89, "learning_rate": 1.2349475691134413e-06, "loss": 2.3933, "step": 30909 }, { "epoch": 5.89, "learning_rate": 1.2348903717826501e-06, "loss": 2.4653, "step": 30910 }, { "epoch": 5.89, "learning_rate": 1.234833174451859e-06, "loss": 2.1331, "step": 30911 }, { "epoch": 5.89, "learning_rate": 1.2347759771210677e-06, "loss": 1.1583, "step": 30912 }, { "epoch": 5.89, "learning_rate": 1.2347187797902765e-06, "loss": 2.0111, "step": 30913 }, { "epoch": 5.89, "learning_rate": 1.2346615824594853e-06, "loss": 1.8439, "step": 30914 }, { "epoch": 5.89, "learning_rate": 1.2346043851286939e-06, "loss": 2.9684, "step": 30915 }, { "epoch": 5.89, "learning_rate": 1.2345471877979027e-06, "loss": 1.7009, "step": 30916 }, { "epoch": 5.89, "learning_rate": 1.2344899904671117e-06, "loss": 1.1688, "step": 30917 }, { "epoch": 5.89, "learning_rate": 1.2344327931363205e-06, "loss": 0.7749, "step": 30918 }, { "epoch": 5.89, "learning_rate": 1.234375595805529e-06, "loss": 2.361, "step": 30919 }, { "epoch": 5.89, "learning_rate": 1.2343183984747378e-06, "loss": 0.655, "step": 30920 }, { "epoch": 5.89, "learning_rate": 1.2342612011439466e-06, "loss": 2.5588, "step": 30921 }, { "epoch": 5.89, "learning_rate": 1.2342040038131554e-06, "loss": 2.297, "step": 30922 }, { "epoch": 5.89, "learning_rate": 1.2341468064823642e-06, "loss": 0.8335, "step": 30923 }, { "epoch": 5.89, "learning_rate": 1.234089609151573e-06, "loss": 1.0558, "step": 30924 }, { "epoch": 5.89, "learning_rate": 1.2340324118207818e-06, "loss": 1.7845, "step": 30925 }, { "epoch": 5.89, "learning_rate": 1.2339752144899904e-06, "loss": 1.8053, "step": 30926 }, { "epoch": 5.89, "learning_rate": 1.2339180171591992e-06, "loss": 1.5956, "step": 30927 }, { "epoch": 5.89, "learning_rate": 1.2338608198284082e-06, "loss": 1.6921, "step": 30928 }, { "epoch": 5.89, "learning_rate": 1.2338036224976168e-06, "loss": 1.6428, "step": 30929 }, { "epoch": 5.89, "learning_rate": 1.2337464251668256e-06, "loss": 2.2164, "step": 30930 }, { "epoch": 5.89, "learning_rate": 1.2336892278360344e-06, "loss": 2.0148, "step": 30931 }, { "epoch": 5.89, "learning_rate": 1.2336320305052432e-06, "loss": 1.9129, "step": 30932 }, { "epoch": 5.89, "learning_rate": 1.233574833174452e-06, "loss": 2.0264, "step": 30933 }, { "epoch": 5.89, "learning_rate": 1.2335176358436607e-06, "loss": 2.3914, "step": 30934 }, { "epoch": 5.89, "learning_rate": 1.2334604385128695e-06, "loss": 2.6499, "step": 30935 }, { "epoch": 5.89, "learning_rate": 1.2334032411820781e-06, "loss": 1.9171, "step": 30936 }, { "epoch": 5.89, "learning_rate": 1.233346043851287e-06, "loss": 3.2806, "step": 30937 }, { "epoch": 5.89, "learning_rate": 1.2332888465204957e-06, "loss": 1.3567, "step": 30938 }, { "epoch": 5.89, "learning_rate": 1.2332316491897045e-06, "loss": 0.6959, "step": 30939 }, { "epoch": 5.89, "learning_rate": 1.2331744518589133e-06, "loss": 1.9573, "step": 30940 }, { "epoch": 5.89, "learning_rate": 1.233117254528122e-06, "loss": 1.9552, "step": 30941 }, { "epoch": 5.89, "learning_rate": 1.2330600571973309e-06, "loss": 1.2717, "step": 30942 }, { "epoch": 5.89, "learning_rate": 1.2330028598665395e-06, "loss": 2.463, "step": 30943 }, { "epoch": 5.89, "learning_rate": 1.2329456625357483e-06, "loss": 1.6998, "step": 30944 }, { "epoch": 5.89, "learning_rate": 1.2328884652049573e-06, "loss": 1.4812, "step": 30945 }, { "epoch": 5.89, "learning_rate": 1.2328312678741658e-06, "loss": 2.0002, "step": 30946 }, { "epoch": 5.89, "learning_rate": 1.2327740705433746e-06, "loss": 2.326, "step": 30947 }, { "epoch": 5.89, "learning_rate": 1.2327168732125834e-06, "loss": 1.7124, "step": 30948 }, { "epoch": 5.9, "learning_rate": 1.2326596758817922e-06, "loss": 1.0957, "step": 30949 }, { "epoch": 5.9, "learning_rate": 1.232602478551001e-06, "loss": 0.6397, "step": 30950 }, { "epoch": 5.9, "learning_rate": 1.2325452812202098e-06, "loss": 2.9616, "step": 30951 }, { "epoch": 5.9, "learning_rate": 1.2324880838894186e-06, "loss": 1.3814, "step": 30952 }, { "epoch": 5.9, "learning_rate": 1.2324308865586272e-06, "loss": 2.3841, "step": 30953 }, { "epoch": 5.9, "learning_rate": 1.232373689227836e-06, "loss": 3.0843, "step": 30954 }, { "epoch": 5.9, "learning_rate": 1.2323164918970448e-06, "loss": 2.7138, "step": 30955 }, { "epoch": 5.9, "learning_rate": 1.2322592945662536e-06, "loss": 1.5384, "step": 30956 }, { "epoch": 5.9, "learning_rate": 1.2322020972354624e-06, "loss": 0.9775, "step": 30957 }, { "epoch": 5.9, "learning_rate": 1.2321448999046712e-06, "loss": 1.4693, "step": 30958 }, { "epoch": 5.9, "learning_rate": 1.23208770257388e-06, "loss": 0.6254, "step": 30959 }, { "epoch": 5.9, "learning_rate": 1.2320305052430885e-06, "loss": 1.4997, "step": 30960 }, { "epoch": 5.9, "learning_rate": 1.2319733079122975e-06, "loss": 1.4658, "step": 30961 }, { "epoch": 5.9, "learning_rate": 1.2319161105815063e-06, "loss": 1.7092, "step": 30962 }, { "epoch": 5.9, "learning_rate": 1.231858913250715e-06, "loss": 1.6169, "step": 30963 }, { "epoch": 5.9, "learning_rate": 1.2318017159199237e-06, "loss": 1.708, "step": 30964 }, { "epoch": 5.9, "learning_rate": 1.2317445185891325e-06, "loss": 1.9306, "step": 30965 }, { "epoch": 5.9, "learning_rate": 1.2316873212583413e-06, "loss": 1.3072, "step": 30966 }, { "epoch": 5.9, "learning_rate": 1.23163012392755e-06, "loss": 1.1791, "step": 30967 }, { "epoch": 5.9, "learning_rate": 1.2315729265967589e-06, "loss": 2.5938, "step": 30968 }, { "epoch": 5.9, "learning_rate": 1.2315157292659677e-06, "loss": 1.0501, "step": 30969 }, { "epoch": 5.9, "learning_rate": 1.2314585319351763e-06, "loss": 1.4158, "step": 30970 }, { "epoch": 5.9, "learning_rate": 1.231401334604385e-06, "loss": 2.1336, "step": 30971 }, { "epoch": 5.9, "learning_rate": 1.231344137273594e-06, "loss": 0.9977, "step": 30972 }, { "epoch": 5.9, "learning_rate": 1.2312869399428029e-06, "loss": 2.3133, "step": 30973 }, { "epoch": 5.9, "learning_rate": 1.2312297426120114e-06, "loss": 2.5284, "step": 30974 }, { "epoch": 5.9, "learning_rate": 1.2311725452812202e-06, "loss": 2.0752, "step": 30975 }, { "epoch": 5.9, "learning_rate": 1.231115347950429e-06, "loss": 1.2295, "step": 30976 }, { "epoch": 5.9, "learning_rate": 1.2310581506196378e-06, "loss": 1.0513, "step": 30977 }, { "epoch": 5.9, "learning_rate": 1.2310009532888466e-06, "loss": 0.8097, "step": 30978 }, { "epoch": 5.9, "learning_rate": 1.2309437559580554e-06, "loss": 2.2223, "step": 30979 }, { "epoch": 5.9, "learning_rate": 1.230886558627264e-06, "loss": 1.4308, "step": 30980 }, { "epoch": 5.9, "learning_rate": 1.2308293612964728e-06, "loss": 1.6308, "step": 30981 }, { "epoch": 5.9, "learning_rate": 1.2307721639656816e-06, "loss": 2.0193, "step": 30982 }, { "epoch": 5.9, "learning_rate": 1.2307149666348904e-06, "loss": 1.4681, "step": 30983 }, { "epoch": 5.9, "learning_rate": 1.2306577693040992e-06, "loss": 0.8758, "step": 30984 }, { "epoch": 5.9, "learning_rate": 1.230600571973308e-06, "loss": 1.9475, "step": 30985 }, { "epoch": 5.9, "learning_rate": 1.2305433746425168e-06, "loss": 1.1487, "step": 30986 }, { "epoch": 5.9, "learning_rate": 1.2304861773117253e-06, "loss": 2.2985, "step": 30987 }, { "epoch": 5.9, "learning_rate": 1.2304289799809341e-06, "loss": 2.7586, "step": 30988 }, { "epoch": 5.9, "learning_rate": 1.2303717826501431e-06, "loss": 0.6068, "step": 30989 }, { "epoch": 5.9, "learning_rate": 1.230314585319352e-06, "loss": 1.5591, "step": 30990 }, { "epoch": 5.9, "learning_rate": 1.2302573879885605e-06, "loss": 2.8117, "step": 30991 }, { "epoch": 5.9, "learning_rate": 1.2302001906577693e-06, "loss": 2.1906, "step": 30992 }, { "epoch": 5.9, "learning_rate": 1.230142993326978e-06, "loss": 0.6815, "step": 30993 }, { "epoch": 5.9, "learning_rate": 1.2300857959961869e-06, "loss": 2.6362, "step": 30994 }, { "epoch": 5.9, "learning_rate": 1.2300285986653957e-06, "loss": 3.1142, "step": 30995 }, { "epoch": 5.9, "learning_rate": 1.2299714013346045e-06, "loss": 2.4416, "step": 30996 }, { "epoch": 5.9, "learning_rate": 1.2299142040038133e-06, "loss": 3.0485, "step": 30997 }, { "epoch": 5.9, "learning_rate": 1.2298570066730219e-06, "loss": 1.9944, "step": 30998 }, { "epoch": 5.9, "learning_rate": 1.2297998093422306e-06, "loss": 2.593, "step": 30999 }, { "epoch": 5.9, "learning_rate": 1.2297426120114397e-06, "loss": 0.8716, "step": 31000 }, { "epoch": 5.9, "learning_rate": 1.2296854146806482e-06, "loss": 3.1974, "step": 31001 }, { "epoch": 5.91, "learning_rate": 1.229628217349857e-06, "loss": 1.9474, "step": 31002 }, { "epoch": 5.91, "learning_rate": 1.2295710200190658e-06, "loss": 2.0099, "step": 31003 }, { "epoch": 5.91, "learning_rate": 1.2295138226882744e-06, "loss": 1.3832, "step": 31004 }, { "epoch": 5.91, "learning_rate": 1.2294566253574834e-06, "loss": 2.3684, "step": 31005 }, { "epoch": 5.91, "learning_rate": 1.2293994280266922e-06, "loss": 1.8856, "step": 31006 }, { "epoch": 5.91, "learning_rate": 1.229342230695901e-06, "loss": 2.2219, "step": 31007 }, { "epoch": 5.91, "learning_rate": 1.2292850333651096e-06, "loss": 2.3314, "step": 31008 }, { "epoch": 5.91, "learning_rate": 1.2292278360343184e-06, "loss": 1.9009, "step": 31009 }, { "epoch": 5.91, "learning_rate": 1.2291706387035272e-06, "loss": 0.6901, "step": 31010 }, { "epoch": 5.91, "learning_rate": 1.229113441372736e-06, "loss": 1.014, "step": 31011 }, { "epoch": 5.91, "learning_rate": 1.2290562440419448e-06, "loss": 2.0734, "step": 31012 }, { "epoch": 5.91, "learning_rate": 1.2289990467111536e-06, "loss": 1.1738, "step": 31013 }, { "epoch": 5.91, "learning_rate": 1.2289418493803623e-06, "loss": 2.3678, "step": 31014 }, { "epoch": 5.91, "learning_rate": 1.228884652049571e-06, "loss": 2.5252, "step": 31015 }, { "epoch": 5.91, "learning_rate": 1.22882745471878e-06, "loss": 1.6266, "step": 31016 }, { "epoch": 5.91, "learning_rate": 1.2287702573879887e-06, "loss": 2.2198, "step": 31017 }, { "epoch": 5.91, "learning_rate": 1.2287130600571973e-06, "loss": 1.3815, "step": 31018 }, { "epoch": 5.91, "learning_rate": 1.228655862726406e-06, "loss": 1.649, "step": 31019 }, { "epoch": 5.91, "learning_rate": 1.228598665395615e-06, "loss": 2.4048, "step": 31020 }, { "epoch": 5.91, "learning_rate": 1.2285414680648237e-06, "loss": 1.4426, "step": 31021 }, { "epoch": 5.91, "learning_rate": 1.2284842707340325e-06, "loss": 2.1009, "step": 31022 }, { "epoch": 5.91, "learning_rate": 1.2284270734032413e-06, "loss": 2.2383, "step": 31023 }, { "epoch": 5.91, "learning_rate": 1.22836987607245e-06, "loss": 2.3447, "step": 31024 }, { "epoch": 5.91, "learning_rate": 1.2283126787416587e-06, "loss": 2.2444, "step": 31025 }, { "epoch": 5.91, "learning_rate": 1.2282554814108674e-06, "loss": 1.747, "step": 31026 }, { "epoch": 5.91, "learning_rate": 1.2281982840800762e-06, "loss": 2.4952, "step": 31027 }, { "epoch": 5.91, "learning_rate": 1.228141086749285e-06, "loss": 1.9812, "step": 31028 }, { "epoch": 5.91, "learning_rate": 1.2280838894184938e-06, "loss": 1.5642, "step": 31029 }, { "epoch": 5.91, "learning_rate": 1.2280266920877026e-06, "loss": 1.2585, "step": 31030 }, { "epoch": 5.91, "learning_rate": 1.2279694947569114e-06, "loss": 2.2471, "step": 31031 }, { "epoch": 5.91, "learning_rate": 1.22791229742612e-06, "loss": 2.0192, "step": 31032 }, { "epoch": 5.91, "learning_rate": 1.227855100095329e-06, "loss": 2.0716, "step": 31033 }, { "epoch": 5.91, "learning_rate": 1.2277979027645378e-06, "loss": 2.9385, "step": 31034 }, { "epoch": 5.91, "learning_rate": 1.2277407054337464e-06, "loss": 2.1128, "step": 31035 }, { "epoch": 5.91, "learning_rate": 1.2276835081029552e-06, "loss": 3.3693, "step": 31036 }, { "epoch": 5.91, "learning_rate": 1.227626310772164e-06, "loss": 2.3711, "step": 31037 }, { "epoch": 5.91, "learning_rate": 1.2275691134413728e-06, "loss": 2.5811, "step": 31038 }, { "epoch": 5.91, "learning_rate": 1.2275119161105816e-06, "loss": 1.5078, "step": 31039 }, { "epoch": 5.91, "learning_rate": 1.2274547187797903e-06, "loss": 2.3464, "step": 31040 }, { "epoch": 5.91, "learning_rate": 1.2273975214489991e-06, "loss": 1.8604, "step": 31041 }, { "epoch": 5.91, "learning_rate": 1.2273403241182077e-06, "loss": 1.212, "step": 31042 }, { "epoch": 5.91, "learning_rate": 1.2272831267874165e-06, "loss": 1.3582, "step": 31043 }, { "epoch": 5.91, "learning_rate": 1.2272259294566255e-06, "loss": 1.3522, "step": 31044 }, { "epoch": 5.91, "learning_rate": 1.227168732125834e-06, "loss": 2.275, "step": 31045 }, { "epoch": 5.91, "learning_rate": 1.227111534795043e-06, "loss": 1.8843, "step": 31046 }, { "epoch": 5.91, "learning_rate": 1.2270543374642517e-06, "loss": 1.9267, "step": 31047 }, { "epoch": 5.91, "learning_rate": 1.2269971401334605e-06, "loss": 2.2039, "step": 31048 }, { "epoch": 5.91, "learning_rate": 1.2269399428026693e-06, "loss": 2.0063, "step": 31049 }, { "epoch": 5.91, "learning_rate": 1.226882745471878e-06, "loss": 2.8327, "step": 31050 }, { "epoch": 5.91, "learning_rate": 1.2268255481410869e-06, "loss": 2.2524, "step": 31051 }, { "epoch": 5.91, "learning_rate": 1.2267683508102955e-06, "loss": 1.4458, "step": 31052 }, { "epoch": 5.91, "learning_rate": 1.2267111534795042e-06, "loss": 2.1845, "step": 31053 }, { "epoch": 5.92, "learning_rate": 1.226653956148713e-06, "loss": 2.0203, "step": 31054 }, { "epoch": 5.92, "learning_rate": 1.2265967588179218e-06, "loss": 1.27, "step": 31055 }, { "epoch": 5.92, "learning_rate": 1.2265395614871306e-06, "loss": 2.673, "step": 31056 }, { "epoch": 5.92, "learning_rate": 1.2264823641563394e-06, "loss": 2.7273, "step": 31057 }, { "epoch": 5.92, "learning_rate": 1.2264251668255482e-06, "loss": 2.2025, "step": 31058 }, { "epoch": 5.92, "learning_rate": 1.2263679694947568e-06, "loss": 0.7048, "step": 31059 }, { "epoch": 5.92, "learning_rate": 1.2263107721639658e-06, "loss": 2.8519, "step": 31060 }, { "epoch": 5.92, "learning_rate": 1.2262535748331746e-06, "loss": 1.7653, "step": 31061 }, { "epoch": 5.92, "learning_rate": 1.2261963775023832e-06, "loss": 1.8436, "step": 31062 }, { "epoch": 5.92, "learning_rate": 1.226139180171592e-06, "loss": 1.0947, "step": 31063 }, { "epoch": 5.92, "learning_rate": 1.2260819828408008e-06, "loss": 2.4498, "step": 31064 }, { "epoch": 5.92, "learning_rate": 1.2260247855100096e-06, "loss": 2.5983, "step": 31065 }, { "epoch": 5.92, "learning_rate": 1.2259675881792184e-06, "loss": 1.3989, "step": 31066 }, { "epoch": 5.92, "learning_rate": 1.2259103908484271e-06, "loss": 2.3901, "step": 31067 }, { "epoch": 5.92, "learning_rate": 1.225853193517636e-06, "loss": 2.165, "step": 31068 }, { "epoch": 5.92, "learning_rate": 1.2257959961868445e-06, "loss": 2.9942, "step": 31069 }, { "epoch": 5.92, "learning_rate": 1.2257387988560533e-06, "loss": 1.651, "step": 31070 }, { "epoch": 5.92, "learning_rate": 1.2256816015252621e-06, "loss": 1.6761, "step": 31071 }, { "epoch": 5.92, "learning_rate": 1.2256244041944711e-06, "loss": 1.3837, "step": 31072 }, { "epoch": 5.92, "learning_rate": 1.2255672068636797e-06, "loss": 1.3798, "step": 31073 }, { "epoch": 5.92, "learning_rate": 1.2255100095328885e-06, "loss": 2.4523, "step": 31074 }, { "epoch": 5.92, "learning_rate": 1.2254528122020973e-06, "loss": 2.2542, "step": 31075 }, { "epoch": 5.92, "learning_rate": 1.2253956148713059e-06, "loss": 2.1478, "step": 31076 }, { "epoch": 5.92, "learning_rate": 1.2253384175405149e-06, "loss": 2.4667, "step": 31077 }, { "epoch": 5.92, "learning_rate": 1.2252812202097237e-06, "loss": 2.7124, "step": 31078 }, { "epoch": 5.92, "learning_rate": 1.2252240228789325e-06, "loss": 3.9292, "step": 31079 }, { "epoch": 5.92, "learning_rate": 1.225166825548141e-06, "loss": 2.0584, "step": 31080 }, { "epoch": 5.92, "learning_rate": 1.2251096282173498e-06, "loss": 1.5772, "step": 31081 }, { "epoch": 5.92, "learning_rate": 1.2250524308865586e-06, "loss": 2.1793, "step": 31082 }, { "epoch": 5.92, "learning_rate": 1.2249952335557674e-06, "loss": 2.183, "step": 31083 }, { "epoch": 5.92, "learning_rate": 1.2249380362249762e-06, "loss": 2.413, "step": 31084 }, { "epoch": 5.92, "learning_rate": 1.224880838894185e-06, "loss": 1.4291, "step": 31085 }, { "epoch": 5.92, "learning_rate": 1.2248236415633936e-06, "loss": 1.8194, "step": 31086 }, { "epoch": 5.92, "learning_rate": 1.2247664442326024e-06, "loss": 2.1929, "step": 31087 }, { "epoch": 5.92, "learning_rate": 1.2247092469018114e-06, "loss": 1.8588, "step": 31088 }, { "epoch": 5.92, "learning_rate": 1.2246520495710202e-06, "loss": 1.6434, "step": 31089 }, { "epoch": 5.92, "learning_rate": 1.2245948522402288e-06, "loss": 1.4834, "step": 31090 }, { "epoch": 5.92, "learning_rate": 1.2245376549094376e-06, "loss": 2.442, "step": 31091 }, { "epoch": 5.92, "learning_rate": 1.2244804575786464e-06, "loss": 1.9348, "step": 31092 }, { "epoch": 5.92, "learning_rate": 1.2244232602478551e-06, "loss": 1.6387, "step": 31093 }, { "epoch": 5.92, "learning_rate": 1.224366062917064e-06, "loss": 2.4214, "step": 31094 }, { "epoch": 5.92, "learning_rate": 1.2243088655862727e-06, "loss": 2.702, "step": 31095 }, { "epoch": 5.92, "learning_rate": 1.2242516682554815e-06, "loss": 1.8953, "step": 31096 }, { "epoch": 5.92, "learning_rate": 1.2241944709246901e-06, "loss": 1.7875, "step": 31097 }, { "epoch": 5.92, "learning_rate": 1.224137273593899e-06, "loss": 1.6626, "step": 31098 }, { "epoch": 5.92, "learning_rate": 1.2240800762631077e-06, "loss": 1.3437, "step": 31099 }, { "epoch": 5.92, "learning_rate": 1.2240228789323165e-06, "loss": 2.1989, "step": 31100 }, { "epoch": 5.92, "learning_rate": 1.2239656816015253e-06, "loss": 2.7341, "step": 31101 }, { "epoch": 5.92, "learning_rate": 1.223908484270734e-06, "loss": 2.9468, "step": 31102 }, { "epoch": 5.92, "learning_rate": 1.2238512869399429e-06, "loss": 1.0739, "step": 31103 }, { "epoch": 5.92, "learning_rate": 1.2237940896091517e-06, "loss": 1.9897, "step": 31104 }, { "epoch": 5.92, "learning_rate": 1.2237368922783605e-06, "loss": 0.7841, "step": 31105 }, { "epoch": 5.92, "learning_rate": 1.2236796949475693e-06, "loss": 2.5359, "step": 31106 }, { "epoch": 5.93, "learning_rate": 1.2236224976167778e-06, "loss": 3.3527, "step": 31107 }, { "epoch": 5.93, "learning_rate": 1.2235653002859866e-06, "loss": 3.33, "step": 31108 }, { "epoch": 5.93, "learning_rate": 1.2235081029551954e-06, "loss": 2.9698, "step": 31109 }, { "epoch": 5.93, "learning_rate": 1.2234509056244042e-06, "loss": 1.7308, "step": 31110 }, { "epoch": 5.93, "learning_rate": 1.223393708293613e-06, "loss": 2.8828, "step": 31111 }, { "epoch": 5.93, "learning_rate": 1.2233365109628218e-06, "loss": 2.1913, "step": 31112 }, { "epoch": 5.93, "learning_rate": 1.2232793136320306e-06, "loss": 1.6559, "step": 31113 }, { "epoch": 5.93, "learning_rate": 1.2232221163012392e-06, "loss": 0.7531, "step": 31114 }, { "epoch": 5.93, "learning_rate": 1.223164918970448e-06, "loss": 1.0477, "step": 31115 }, { "epoch": 5.93, "learning_rate": 1.223107721639657e-06, "loss": 3.0517, "step": 31116 }, { "epoch": 5.93, "learning_rate": 1.2230505243088656e-06, "loss": 1.5352, "step": 31117 }, { "epoch": 5.93, "learning_rate": 1.2229933269780744e-06, "loss": 3.8595, "step": 31118 }, { "epoch": 5.93, "learning_rate": 1.2229361296472832e-06, "loss": 2.5236, "step": 31119 }, { "epoch": 5.93, "learning_rate": 1.222878932316492e-06, "loss": 1.2842, "step": 31120 }, { "epoch": 5.93, "learning_rate": 1.2228217349857007e-06, "loss": 1.3691, "step": 31121 }, { "epoch": 5.93, "learning_rate": 1.2227645376549095e-06, "loss": 0.7426, "step": 31122 }, { "epoch": 5.93, "learning_rate": 1.2227073403241183e-06, "loss": 2.8299, "step": 31123 }, { "epoch": 5.93, "learning_rate": 1.222650142993327e-06, "loss": 1.536, "step": 31124 }, { "epoch": 5.93, "learning_rate": 1.2225929456625357e-06, "loss": 1.7978, "step": 31125 }, { "epoch": 5.93, "learning_rate": 1.2225357483317445e-06, "loss": 1.2856, "step": 31126 }, { "epoch": 5.93, "learning_rate": 1.2224785510009533e-06, "loss": 1.8586, "step": 31127 }, { "epoch": 5.93, "learning_rate": 1.222421353670162e-06, "loss": 2.5219, "step": 31128 }, { "epoch": 5.93, "learning_rate": 1.2223641563393709e-06, "loss": 1.4526, "step": 31129 }, { "epoch": 5.93, "learning_rate": 1.2223069590085797e-06, "loss": 1.1577, "step": 31130 }, { "epoch": 5.93, "learning_rate": 1.2222497616777883e-06, "loss": 0.9633, "step": 31131 }, { "epoch": 5.93, "learning_rate": 1.2221925643469973e-06, "loss": 1.4196, "step": 31132 }, { "epoch": 5.93, "learning_rate": 1.222135367016206e-06, "loss": 2.5312, "step": 31133 }, { "epoch": 5.93, "learning_rate": 1.2220781696854146e-06, "loss": 1.041, "step": 31134 }, { "epoch": 5.93, "learning_rate": 1.2220209723546234e-06, "loss": 1.5354, "step": 31135 }, { "epoch": 5.93, "learning_rate": 1.2219637750238322e-06, "loss": 2.0685, "step": 31136 }, { "epoch": 5.93, "learning_rate": 1.221906577693041e-06, "loss": 1.9127, "step": 31137 }, { "epoch": 5.93, "learning_rate": 1.2218493803622498e-06, "loss": 2.2939, "step": 31138 }, { "epoch": 5.93, "learning_rate": 1.2217921830314586e-06, "loss": 2.3074, "step": 31139 }, { "epoch": 5.93, "learning_rate": 1.2217349857006674e-06, "loss": 0.8078, "step": 31140 }, { "epoch": 5.93, "learning_rate": 1.221677788369876e-06, "loss": 2.2706, "step": 31141 }, { "epoch": 5.93, "learning_rate": 1.2216205910390848e-06, "loss": 1.7211, "step": 31142 }, { "epoch": 5.93, "learning_rate": 1.2215633937082936e-06, "loss": 3.0107, "step": 31143 }, { "epoch": 5.93, "learning_rate": 1.2215061963775026e-06, "loss": 1.3966, "step": 31144 }, { "epoch": 5.93, "learning_rate": 1.2214489990467112e-06, "loss": 1.3347, "step": 31145 }, { "epoch": 5.93, "learning_rate": 1.22139180171592e-06, "loss": 1.5406, "step": 31146 }, { "epoch": 5.93, "learning_rate": 1.2213346043851287e-06, "loss": 1.0747, "step": 31147 }, { "epoch": 5.93, "learning_rate": 1.2212774070543375e-06, "loss": 1.0248, "step": 31148 }, { "epoch": 5.93, "learning_rate": 1.2212202097235463e-06, "loss": 1.1544, "step": 31149 }, { "epoch": 5.93, "learning_rate": 1.2211630123927551e-06, "loss": 2.2366, "step": 31150 }, { "epoch": 5.93, "learning_rate": 1.2211058150619637e-06, "loss": 1.348, "step": 31151 }, { "epoch": 5.93, "learning_rate": 1.2210486177311725e-06, "loss": 2.1825, "step": 31152 }, { "epoch": 5.93, "learning_rate": 1.2209914204003813e-06, "loss": 3.4926, "step": 31153 }, { "epoch": 5.93, "learning_rate": 1.22093422306959e-06, "loss": 1.9389, "step": 31154 }, { "epoch": 5.93, "learning_rate": 1.2208770257387989e-06, "loss": 1.0908, "step": 31155 }, { "epoch": 5.93, "learning_rate": 1.2208198284080077e-06, "loss": 0.9823, "step": 31156 }, { "epoch": 5.93, "learning_rate": 1.2207626310772165e-06, "loss": 1.4589, "step": 31157 }, { "epoch": 5.93, "learning_rate": 1.220705433746425e-06, "loss": 1.5262, "step": 31158 }, { "epoch": 5.94, "learning_rate": 1.2206482364156338e-06, "loss": 2.2184, "step": 31159 }, { "epoch": 5.94, "learning_rate": 1.2205910390848429e-06, "loss": 1.613, "step": 31160 }, { "epoch": 5.94, "learning_rate": 1.2205338417540516e-06, "loss": 1.9368, "step": 31161 }, { "epoch": 5.94, "learning_rate": 1.2204766444232602e-06, "loss": 1.6144, "step": 31162 }, { "epoch": 5.94, "learning_rate": 1.220419447092469e-06, "loss": 2.4327, "step": 31163 }, { "epoch": 5.94, "learning_rate": 1.2203622497616778e-06, "loss": 0.7197, "step": 31164 }, { "epoch": 5.94, "learning_rate": 1.2203050524308866e-06, "loss": 3.2694, "step": 31165 }, { "epoch": 5.94, "learning_rate": 1.2202478551000954e-06, "loss": 1.897, "step": 31166 }, { "epoch": 5.94, "learning_rate": 1.2201906577693042e-06, "loss": 3.48, "step": 31167 }, { "epoch": 5.94, "learning_rate": 1.220133460438513e-06, "loss": 1.2324, "step": 31168 }, { "epoch": 5.94, "learning_rate": 1.2200762631077216e-06, "loss": 1.5115, "step": 31169 }, { "epoch": 5.94, "learning_rate": 1.2200190657769304e-06, "loss": 2.2028, "step": 31170 }, { "epoch": 5.94, "learning_rate": 1.2199618684461394e-06, "loss": 1.3998, "step": 31171 }, { "epoch": 5.94, "learning_rate": 1.219904671115348e-06, "loss": 0.7847, "step": 31172 }, { "epoch": 5.94, "learning_rate": 1.2198474737845567e-06, "loss": 1.926, "step": 31173 }, { "epoch": 5.94, "learning_rate": 1.2197902764537655e-06, "loss": 1.1951, "step": 31174 }, { "epoch": 5.94, "learning_rate": 1.2197330791229741e-06, "loss": 2.213, "step": 31175 }, { "epoch": 5.94, "learning_rate": 1.2196758817921831e-06, "loss": 2.5649, "step": 31176 }, { "epoch": 5.94, "learning_rate": 1.219618684461392e-06, "loss": 1.7089, "step": 31177 }, { "epoch": 5.94, "learning_rate": 1.2195614871306007e-06, "loss": 1.5819, "step": 31178 }, { "epoch": 5.94, "learning_rate": 1.2195042897998093e-06, "loss": 2.4766, "step": 31179 }, { "epoch": 5.94, "learning_rate": 1.219447092469018e-06, "loss": 1.8356, "step": 31180 }, { "epoch": 5.94, "learning_rate": 1.2193898951382269e-06, "loss": 2.4758, "step": 31181 }, { "epoch": 5.94, "learning_rate": 1.2193326978074357e-06, "loss": 1.912, "step": 31182 }, { "epoch": 5.94, "learning_rate": 1.2192755004766445e-06, "loss": 2.2084, "step": 31183 }, { "epoch": 5.94, "learning_rate": 1.2192183031458533e-06, "loss": 0.4019, "step": 31184 }, { "epoch": 5.94, "learning_rate": 1.219161105815062e-06, "loss": 3.6413, "step": 31185 }, { "epoch": 5.94, "learning_rate": 1.2191039084842706e-06, "loss": 1.8067, "step": 31186 }, { "epoch": 5.94, "learning_rate": 1.2190467111534794e-06, "loss": 0.695, "step": 31187 }, { "epoch": 5.94, "learning_rate": 1.2189895138226884e-06, "loss": 2.4421, "step": 31188 }, { "epoch": 5.94, "learning_rate": 1.218932316491897e-06, "loss": 2.0471, "step": 31189 }, { "epoch": 5.94, "learning_rate": 1.2188751191611058e-06, "loss": 2.8085, "step": 31190 }, { "epoch": 5.94, "learning_rate": 1.2188179218303146e-06, "loss": 0.7199, "step": 31191 }, { "epoch": 5.94, "learning_rate": 1.2187607244995234e-06, "loss": 0.938, "step": 31192 }, { "epoch": 5.94, "learning_rate": 1.2187035271687322e-06, "loss": 1.9312, "step": 31193 }, { "epoch": 5.94, "learning_rate": 1.218646329837941e-06, "loss": 1.8251, "step": 31194 }, { "epoch": 5.94, "learning_rate": 1.2185891325071498e-06, "loss": 2.6125, "step": 31195 }, { "epoch": 5.94, "learning_rate": 1.2185319351763584e-06, "loss": 1.4322, "step": 31196 }, { "epoch": 5.94, "learning_rate": 1.2184747378455672e-06, "loss": 2.7164, "step": 31197 }, { "epoch": 5.94, "learning_rate": 1.218417540514776e-06, "loss": 2.0717, "step": 31198 }, { "epoch": 5.94, "learning_rate": 1.2183603431839848e-06, "loss": 1.8831, "step": 31199 }, { "epoch": 5.94, "learning_rate": 1.2183031458531935e-06, "loss": 1.2446, "step": 31200 }, { "epoch": 5.94, "learning_rate": 1.2182459485224023e-06, "loss": 1.941, "step": 31201 }, { "epoch": 5.94, "learning_rate": 1.2181887511916111e-06, "loss": 2.325, "step": 31202 }, { "epoch": 5.94, "learning_rate": 1.2181315538608197e-06, "loss": 2.5123, "step": 31203 }, { "epoch": 5.94, "learning_rate": 1.2180743565300287e-06, "loss": 1.2162, "step": 31204 }, { "epoch": 5.94, "learning_rate": 1.2180171591992375e-06, "loss": 1.5927, "step": 31205 }, { "epoch": 5.94, "learning_rate": 1.217959961868446e-06, "loss": 2.0478, "step": 31206 }, { "epoch": 5.94, "learning_rate": 1.2179027645376549e-06, "loss": 1.4153, "step": 31207 }, { "epoch": 5.94, "learning_rate": 1.2178455672068637e-06, "loss": 1.8241, "step": 31208 }, { "epoch": 5.94, "learning_rate": 1.2177883698760725e-06, "loss": 3.2963, "step": 31209 }, { "epoch": 5.94, "learning_rate": 1.2177311725452813e-06, "loss": 1.4126, "step": 31210 }, { "epoch": 5.94, "learning_rate": 1.21767397521449e-06, "loss": 1.0201, "step": 31211 }, { "epoch": 5.95, "learning_rate": 1.2176167778836989e-06, "loss": 1.686, "step": 31212 }, { "epoch": 5.95, "learning_rate": 1.2175595805529074e-06, "loss": 1.3998, "step": 31213 }, { "epoch": 5.95, "learning_rate": 1.2175023832221162e-06, "loss": 2.016, "step": 31214 }, { "epoch": 5.95, "learning_rate": 1.2174451858913252e-06, "loss": 1.6385, "step": 31215 }, { "epoch": 5.95, "learning_rate": 1.2173879885605338e-06, "loss": 0.9062, "step": 31216 }, { "epoch": 5.95, "learning_rate": 1.2173307912297426e-06, "loss": 2.0726, "step": 31217 }, { "epoch": 5.95, "learning_rate": 1.2172735938989514e-06, "loss": 1.5032, "step": 31218 }, { "epoch": 5.95, "learning_rate": 1.2172163965681602e-06, "loss": 1.8284, "step": 31219 }, { "epoch": 5.95, "learning_rate": 1.217159199237369e-06, "loss": 1.9163, "step": 31220 }, { "epoch": 5.95, "learning_rate": 1.2171020019065778e-06, "loss": 2.1727, "step": 31221 }, { "epoch": 5.95, "learning_rate": 1.2170448045757866e-06, "loss": 2.0014, "step": 31222 }, { "epoch": 5.95, "learning_rate": 1.2169876072449952e-06, "loss": 1.5532, "step": 31223 }, { "epoch": 5.95, "learning_rate": 1.216930409914204e-06, "loss": 2.0975, "step": 31224 }, { "epoch": 5.95, "learning_rate": 1.2168732125834128e-06, "loss": 2.0811, "step": 31225 }, { "epoch": 5.95, "learning_rate": 1.2168160152526216e-06, "loss": 1.6931, "step": 31226 }, { "epoch": 5.95, "learning_rate": 1.2167588179218303e-06, "loss": 1.2717, "step": 31227 }, { "epoch": 5.95, "learning_rate": 1.2167016205910391e-06, "loss": 1.7436, "step": 31228 }, { "epoch": 5.95, "learning_rate": 1.216644423260248e-06, "loss": 2.6678, "step": 31229 }, { "epoch": 5.95, "learning_rate": 1.2165872259294565e-06, "loss": 1.43, "step": 31230 }, { "epoch": 5.95, "learning_rate": 1.2165300285986653e-06, "loss": 1.1962, "step": 31231 }, { "epoch": 5.95, "learning_rate": 1.2164728312678743e-06, "loss": 2.2594, "step": 31232 }, { "epoch": 5.95, "learning_rate": 1.2164156339370831e-06, "loss": 1.0459, "step": 31233 }, { "epoch": 5.95, "learning_rate": 1.2163584366062917e-06, "loss": 1.0334, "step": 31234 }, { "epoch": 5.95, "learning_rate": 1.2163012392755005e-06, "loss": 1.7528, "step": 31235 }, { "epoch": 5.95, "learning_rate": 1.2162440419447093e-06, "loss": 2.1898, "step": 31236 }, { "epoch": 5.95, "learning_rate": 1.216186844613918e-06, "loss": 2.0791, "step": 31237 }, { "epoch": 5.95, "learning_rate": 1.2161296472831269e-06, "loss": 1.8115, "step": 31238 }, { "epoch": 5.95, "learning_rate": 1.2160724499523357e-06, "loss": 2.0715, "step": 31239 }, { "epoch": 5.95, "learning_rate": 1.2160152526215442e-06, "loss": 0.5563, "step": 31240 }, { "epoch": 5.95, "learning_rate": 1.215958055290753e-06, "loss": 2.4189, "step": 31241 }, { "epoch": 5.95, "learning_rate": 1.2159008579599618e-06, "loss": 4.0455, "step": 31242 }, { "epoch": 5.95, "learning_rate": 1.2158436606291708e-06, "loss": 2.7618, "step": 31243 }, { "epoch": 5.95, "learning_rate": 1.2157864632983794e-06, "loss": 1.658, "step": 31244 }, { "epoch": 5.95, "learning_rate": 1.2157292659675882e-06, "loss": 2.77, "step": 31245 }, { "epoch": 5.95, "learning_rate": 1.215672068636797e-06, "loss": 3.882, "step": 31246 }, { "epoch": 5.95, "learning_rate": 1.2156148713060056e-06, "loss": 3.433, "step": 31247 }, { "epoch": 5.95, "learning_rate": 1.2155576739752146e-06, "loss": 1.5793, "step": 31248 }, { "epoch": 5.95, "learning_rate": 1.2155004766444234e-06, "loss": 1.4363, "step": 31249 }, { "epoch": 5.95, "learning_rate": 1.2154432793136322e-06, "loss": 1.379, "step": 31250 }, { "epoch": 5.95, "learning_rate": 1.2153860819828408e-06, "loss": 0.6625, "step": 31251 }, { "epoch": 5.95, "learning_rate": 1.2153288846520496e-06, "loss": 1.9074, "step": 31252 }, { "epoch": 5.95, "learning_rate": 1.2152716873212583e-06, "loss": 2.7556, "step": 31253 }, { "epoch": 5.95, "learning_rate": 1.2152144899904671e-06, "loss": 1.1605, "step": 31254 }, { "epoch": 5.95, "learning_rate": 1.215157292659676e-06, "loss": 2.9528, "step": 31255 }, { "epoch": 5.95, "learning_rate": 1.2151000953288847e-06, "loss": 2.0415, "step": 31256 }, { "epoch": 5.95, "learning_rate": 1.2150428979980933e-06, "loss": 2.745, "step": 31257 }, { "epoch": 5.95, "learning_rate": 1.2149857006673021e-06, "loss": 1.294, "step": 31258 }, { "epoch": 5.95, "learning_rate": 1.2149285033365111e-06, "loss": 2.3931, "step": 31259 }, { "epoch": 5.95, "learning_rate": 1.21487130600572e-06, "loss": 1.3637, "step": 31260 }, { "epoch": 5.95, "learning_rate": 1.2148141086749285e-06, "loss": 2.179, "step": 31261 }, { "epoch": 5.95, "learning_rate": 1.2147569113441373e-06, "loss": 1.2211, "step": 31262 }, { "epoch": 5.95, "learning_rate": 1.214699714013346e-06, "loss": 1.2428, "step": 31263 }, { "epoch": 5.96, "learning_rate": 1.2146425166825549e-06, "loss": 1.7189, "step": 31264 }, { "epoch": 5.96, "learning_rate": 1.2145853193517637e-06, "loss": 1.916, "step": 31265 }, { "epoch": 5.96, "learning_rate": 1.2145281220209725e-06, "loss": 2.5856, "step": 31266 }, { "epoch": 5.96, "learning_rate": 1.2144709246901813e-06, "loss": 1.3065, "step": 31267 }, { "epoch": 5.96, "learning_rate": 1.2144137273593898e-06, "loss": 1.724, "step": 31268 }, { "epoch": 5.96, "learning_rate": 1.2143565300285986e-06, "loss": 1.7339, "step": 31269 }, { "epoch": 5.96, "learning_rate": 1.2142993326978074e-06, "loss": 2.5072, "step": 31270 }, { "epoch": 5.96, "learning_rate": 1.2142421353670162e-06, "loss": 2.0791, "step": 31271 }, { "epoch": 5.96, "learning_rate": 1.214184938036225e-06, "loss": 1.8327, "step": 31272 }, { "epoch": 5.96, "learning_rate": 1.2141277407054338e-06, "loss": 1.6235, "step": 31273 }, { "epoch": 5.96, "learning_rate": 1.2140705433746426e-06, "loss": 1.1495, "step": 31274 }, { "epoch": 5.96, "learning_rate": 1.2140133460438512e-06, "loss": 0.6163, "step": 31275 }, { "epoch": 5.96, "learning_rate": 1.2139561487130602e-06, "loss": 1.2063, "step": 31276 }, { "epoch": 5.96, "learning_rate": 1.213898951382269e-06, "loss": 2.0443, "step": 31277 }, { "epoch": 5.96, "learning_rate": 1.2138417540514776e-06, "loss": 1.0532, "step": 31278 }, { "epoch": 5.96, "learning_rate": 1.2137845567206864e-06, "loss": 2.0945, "step": 31279 }, { "epoch": 5.96, "learning_rate": 1.2137273593898951e-06, "loss": 2.2839, "step": 31280 }, { "epoch": 5.96, "learning_rate": 1.213670162059104e-06, "loss": 2.5695, "step": 31281 }, { "epoch": 5.96, "learning_rate": 1.2136129647283127e-06, "loss": 1.4667, "step": 31282 }, { "epoch": 5.96, "learning_rate": 1.2135557673975215e-06, "loss": 1.1109, "step": 31283 }, { "epoch": 5.96, "learning_rate": 1.2134985700667303e-06, "loss": 1.9579, "step": 31284 }, { "epoch": 5.96, "learning_rate": 1.213441372735939e-06, "loss": 2.6918, "step": 31285 }, { "epoch": 5.96, "learning_rate": 1.2133841754051477e-06, "loss": 0.9362, "step": 31286 }, { "epoch": 5.96, "learning_rate": 1.2133269780743567e-06, "loss": 2.0663, "step": 31287 }, { "epoch": 5.96, "learning_rate": 1.2132697807435653e-06, "loss": 1.5151, "step": 31288 }, { "epoch": 5.96, "learning_rate": 1.213212583412774e-06, "loss": 2.8709, "step": 31289 }, { "epoch": 5.96, "learning_rate": 1.2131553860819829e-06, "loss": 2.0107, "step": 31290 }, { "epoch": 5.96, "learning_rate": 1.2130981887511917e-06, "loss": 1.7707, "step": 31291 }, { "epoch": 5.96, "learning_rate": 1.2130409914204005e-06, "loss": 3.0061, "step": 31292 }, { "epoch": 5.96, "learning_rate": 1.2129837940896093e-06, "loss": 1.448, "step": 31293 }, { "epoch": 5.96, "learning_rate": 1.212926596758818e-06, "loss": 1.7759, "step": 31294 }, { "epoch": 5.96, "learning_rate": 1.2128693994280266e-06, "loss": 1.2697, "step": 31295 }, { "epoch": 5.96, "learning_rate": 1.2128122020972354e-06, "loss": 1.9248, "step": 31296 }, { "epoch": 5.96, "learning_rate": 1.2127550047664442e-06, "loss": 1.4025, "step": 31297 }, { "epoch": 5.96, "learning_rate": 1.212697807435653e-06, "loss": 1.6818, "step": 31298 }, { "epoch": 5.96, "learning_rate": 1.2126406101048618e-06, "loss": 0.3185, "step": 31299 }, { "epoch": 5.96, "learning_rate": 1.2125834127740706e-06, "loss": 0.7843, "step": 31300 }, { "epoch": 5.96, "learning_rate": 1.2125262154432794e-06, "loss": 1.9234, "step": 31301 }, { "epoch": 5.96, "learning_rate": 1.212469018112488e-06, "loss": 1.7136, "step": 31302 }, { "epoch": 5.96, "learning_rate": 1.212411820781697e-06, "loss": 2.5967, "step": 31303 }, { "epoch": 5.96, "learning_rate": 1.2123546234509058e-06, "loss": 2.3309, "step": 31304 }, { "epoch": 5.96, "learning_rate": 1.2122974261201144e-06, "loss": 1.7603, "step": 31305 }, { "epoch": 5.96, "learning_rate": 1.2122402287893232e-06, "loss": 2.7409, "step": 31306 }, { "epoch": 5.96, "learning_rate": 1.212183031458532e-06, "loss": 1.5405, "step": 31307 }, { "epoch": 5.96, "learning_rate": 1.2121258341277407e-06, "loss": 1.6589, "step": 31308 }, { "epoch": 5.96, "learning_rate": 1.2120686367969495e-06, "loss": 1.7744, "step": 31309 }, { "epoch": 5.96, "learning_rate": 1.2120114394661583e-06, "loss": 1.4417, "step": 31310 }, { "epoch": 5.96, "learning_rate": 1.2119542421353671e-06, "loss": 2.5238, "step": 31311 }, { "epoch": 5.96, "learning_rate": 1.2118970448045757e-06, "loss": 1.4375, "step": 31312 }, { "epoch": 5.96, "learning_rate": 1.2118398474737845e-06, "loss": 1.0944, "step": 31313 }, { "epoch": 5.96, "learning_rate": 1.2117826501429933e-06, "loss": 1.6551, "step": 31314 }, { "epoch": 5.96, "learning_rate": 1.2117254528122023e-06, "loss": 1.4189, "step": 31315 }, { "epoch": 5.96, "learning_rate": 1.2116682554814109e-06, "loss": 2.5735, "step": 31316 }, { "epoch": 5.97, "learning_rate": 1.2116110581506197e-06, "loss": 2.8823, "step": 31317 }, { "epoch": 5.97, "learning_rate": 1.2115538608198285e-06, "loss": 1.2767, "step": 31318 }, { "epoch": 5.97, "learning_rate": 1.211496663489037e-06, "loss": 1.6334, "step": 31319 }, { "epoch": 5.97, "learning_rate": 1.211439466158246e-06, "loss": 1.8034, "step": 31320 }, { "epoch": 5.97, "learning_rate": 1.2113822688274548e-06, "loss": 2.5163, "step": 31321 }, { "epoch": 5.97, "learning_rate": 1.2113250714966634e-06, "loss": 3.057, "step": 31322 }, { "epoch": 5.97, "learning_rate": 1.2112678741658722e-06, "loss": 1.8099, "step": 31323 }, { "epoch": 5.97, "learning_rate": 1.211210676835081e-06, "loss": 2.7259, "step": 31324 }, { "epoch": 5.97, "learning_rate": 1.2111534795042898e-06, "loss": 1.5386, "step": 31325 }, { "epoch": 5.97, "learning_rate": 1.2110962821734986e-06, "loss": 1.2175, "step": 31326 }, { "epoch": 5.97, "learning_rate": 1.2110390848427074e-06, "loss": 2.335, "step": 31327 }, { "epoch": 5.97, "learning_rate": 1.2109818875119162e-06, "loss": 1.3304, "step": 31328 }, { "epoch": 5.97, "learning_rate": 1.2109246901811248e-06, "loss": 0.8702, "step": 31329 }, { "epoch": 5.97, "learning_rate": 1.2108674928503336e-06, "loss": 1.024, "step": 31330 }, { "epoch": 5.97, "learning_rate": 1.2108102955195426e-06, "loss": 1.5214, "step": 31331 }, { "epoch": 5.97, "learning_rate": 1.2107530981887514e-06, "loss": 1.7352, "step": 31332 }, { "epoch": 5.97, "learning_rate": 1.21069590085796e-06, "loss": 2.7857, "step": 31333 }, { "epoch": 5.97, "learning_rate": 1.2106387035271687e-06, "loss": 1.845, "step": 31334 }, { "epoch": 5.97, "learning_rate": 1.2105815061963775e-06, "loss": 1.2842, "step": 31335 }, { "epoch": 5.97, "learning_rate": 1.2105243088655863e-06, "loss": 2.0186, "step": 31336 }, { "epoch": 5.97, "learning_rate": 1.2104671115347951e-06, "loss": 1.5389, "step": 31337 }, { "epoch": 5.97, "learning_rate": 1.210409914204004e-06, "loss": 1.5719, "step": 31338 }, { "epoch": 5.97, "learning_rate": 1.2103527168732127e-06, "loss": 1.5179, "step": 31339 }, { "epoch": 5.97, "learning_rate": 1.2102955195424213e-06, "loss": 1.2529, "step": 31340 }, { "epoch": 5.97, "learning_rate": 1.21023832221163e-06, "loss": 2.0518, "step": 31341 }, { "epoch": 5.97, "learning_rate": 1.2101811248808389e-06, "loss": 1.4248, "step": 31342 }, { "epoch": 5.97, "learning_rate": 1.2101239275500477e-06, "loss": 1.5989, "step": 31343 }, { "epoch": 5.97, "learning_rate": 1.2100667302192565e-06, "loss": 2.5012, "step": 31344 }, { "epoch": 5.97, "learning_rate": 1.2100095328884653e-06, "loss": 1.6368, "step": 31345 }, { "epoch": 5.97, "learning_rate": 1.2099523355576738e-06, "loss": 2.13, "step": 31346 }, { "epoch": 5.97, "learning_rate": 1.2098951382268829e-06, "loss": 1.8017, "step": 31347 }, { "epoch": 5.97, "learning_rate": 1.2098379408960916e-06, "loss": 1.5553, "step": 31348 }, { "epoch": 5.97, "learning_rate": 1.2097807435653004e-06, "loss": 1.2016, "step": 31349 }, { "epoch": 5.97, "learning_rate": 1.209723546234509e-06, "loss": 1.1494, "step": 31350 }, { "epoch": 5.97, "learning_rate": 1.2096663489037178e-06, "loss": 1.9462, "step": 31351 }, { "epoch": 5.97, "learning_rate": 1.2096091515729266e-06, "loss": 1.48, "step": 31352 }, { "epoch": 5.97, "learning_rate": 1.2095519542421354e-06, "loss": 3.08, "step": 31353 }, { "epoch": 5.97, "learning_rate": 1.2094947569113442e-06, "loss": 1.5828, "step": 31354 }, { "epoch": 5.97, "learning_rate": 1.209437559580553e-06, "loss": 0.8979, "step": 31355 }, { "epoch": 5.97, "learning_rate": 1.2093803622497618e-06, "loss": 2.6585, "step": 31356 }, { "epoch": 5.97, "learning_rate": 1.2093231649189704e-06, "loss": 2.6637, "step": 31357 }, { "epoch": 5.97, "learning_rate": 1.2092659675881792e-06, "loss": 2.5478, "step": 31358 }, { "epoch": 5.97, "learning_rate": 1.2092087702573882e-06, "loss": 2.5069, "step": 31359 }, { "epoch": 5.97, "learning_rate": 1.2091515729265967e-06, "loss": 1.3166, "step": 31360 }, { "epoch": 5.97, "learning_rate": 1.2090943755958055e-06, "loss": 1.7875, "step": 31361 }, { "epoch": 5.97, "learning_rate": 1.2090371782650143e-06, "loss": 3.0745, "step": 31362 }, { "epoch": 5.97, "learning_rate": 1.208979980934223e-06, "loss": 1.836, "step": 31363 }, { "epoch": 5.97, "learning_rate": 1.208922783603432e-06, "loss": 1.2279, "step": 31364 }, { "epoch": 5.97, "learning_rate": 1.2088655862726407e-06, "loss": 0.5951, "step": 31365 }, { "epoch": 5.97, "learning_rate": 1.2088083889418495e-06, "loss": 1.9285, "step": 31366 }, { "epoch": 5.97, "learning_rate": 1.208751191611058e-06, "loss": 1.5498, "step": 31367 }, { "epoch": 5.97, "learning_rate": 1.2086939942802669e-06, "loss": 0.4794, "step": 31368 }, { "epoch": 5.98, "learning_rate": 1.2086367969494757e-06, "loss": 1.9386, "step": 31369 }, { "epoch": 5.98, "learning_rate": 1.2085795996186845e-06, "loss": 1.255, "step": 31370 }, { "epoch": 5.98, "learning_rate": 1.2085224022878933e-06, "loss": 2.7936, "step": 31371 }, { "epoch": 5.98, "learning_rate": 1.208465204957102e-06, "loss": 2.0693, "step": 31372 }, { "epoch": 5.98, "learning_rate": 1.2084080076263109e-06, "loss": 1.7649, "step": 31373 }, { "epoch": 5.98, "learning_rate": 1.2083508102955194e-06, "loss": 1.1296, "step": 31374 }, { "epoch": 5.98, "learning_rate": 1.2082936129647284e-06, "loss": 0.8553, "step": 31375 }, { "epoch": 5.98, "learning_rate": 1.2082364156339372e-06, "loss": 0.903, "step": 31376 }, { "epoch": 5.98, "learning_rate": 1.2081792183031458e-06, "loss": 2.2595, "step": 31377 }, { "epoch": 5.98, "learning_rate": 1.2081220209723546e-06, "loss": 0.3518, "step": 31378 }, { "epoch": 5.98, "learning_rate": 1.2080648236415634e-06, "loss": 0.6437, "step": 31379 }, { "epoch": 5.98, "learning_rate": 1.2080076263107722e-06, "loss": 1.8353, "step": 31380 }, { "epoch": 5.98, "learning_rate": 1.207950428979981e-06, "loss": 0.813, "step": 31381 }, { "epoch": 5.98, "learning_rate": 1.2078932316491898e-06, "loss": 1.7654, "step": 31382 }, { "epoch": 5.98, "learning_rate": 1.2078360343183986e-06, "loss": 1.9758, "step": 31383 }, { "epoch": 5.98, "learning_rate": 1.2077788369876072e-06, "loss": 1.6018, "step": 31384 }, { "epoch": 5.98, "learning_rate": 1.207721639656816e-06, "loss": 3.3471, "step": 31385 }, { "epoch": 5.98, "learning_rate": 1.2076644423260248e-06, "loss": 1.5613, "step": 31386 }, { "epoch": 5.98, "learning_rate": 1.2076072449952335e-06, "loss": 1.1539, "step": 31387 }, { "epoch": 5.98, "learning_rate": 1.2075500476644423e-06, "loss": 2.2, "step": 31388 }, { "epoch": 5.98, "learning_rate": 1.2074928503336511e-06, "loss": 2.4278, "step": 31389 }, { "epoch": 5.98, "learning_rate": 1.20743565300286e-06, "loss": 2.5381, "step": 31390 }, { "epoch": 5.98, "learning_rate": 1.2073784556720687e-06, "loss": 1.1609, "step": 31391 }, { "epoch": 5.98, "learning_rate": 1.2073212583412775e-06, "loss": 2.2527, "step": 31392 }, { "epoch": 5.98, "learning_rate": 1.2072640610104863e-06, "loss": 3.1088, "step": 31393 }, { "epoch": 5.98, "learning_rate": 1.2072068636796949e-06, "loss": 1.9867, "step": 31394 }, { "epoch": 5.98, "learning_rate": 1.2071496663489037e-06, "loss": 2.9812, "step": 31395 }, { "epoch": 5.98, "learning_rate": 1.2070924690181125e-06, "loss": 1.5654, "step": 31396 }, { "epoch": 5.98, "learning_rate": 1.2070352716873213e-06, "loss": 2.697, "step": 31397 }, { "epoch": 5.98, "learning_rate": 1.20697807435653e-06, "loss": 1.9783, "step": 31398 }, { "epoch": 5.98, "learning_rate": 1.2069208770257389e-06, "loss": 1.951, "step": 31399 }, { "epoch": 5.98, "learning_rate": 1.2068636796949477e-06, "loss": 3.5315, "step": 31400 }, { "epoch": 5.98, "learning_rate": 1.2068064823641562e-06, "loss": 2.3869, "step": 31401 }, { "epoch": 5.98, "learning_rate": 1.206749285033365e-06, "loss": 1.3714, "step": 31402 }, { "epoch": 5.98, "learning_rate": 1.206692087702574e-06, "loss": 1.7021, "step": 31403 }, { "epoch": 5.98, "learning_rate": 1.2066348903717828e-06, "loss": 0.589, "step": 31404 }, { "epoch": 5.98, "learning_rate": 1.2065776930409914e-06, "loss": 3.0086, "step": 31405 }, { "epoch": 5.98, "learning_rate": 1.2065204957102002e-06, "loss": 2.0279, "step": 31406 }, { "epoch": 5.98, "learning_rate": 1.206463298379409e-06, "loss": 1.8345, "step": 31407 }, { "epoch": 5.98, "learning_rate": 1.2064061010486178e-06, "loss": 1.1322, "step": 31408 }, { "epoch": 5.98, "learning_rate": 1.2063489037178266e-06, "loss": 1.6518, "step": 31409 }, { "epoch": 5.98, "learning_rate": 1.2062917063870354e-06, "loss": 1.2486, "step": 31410 }, { "epoch": 5.98, "learning_rate": 1.206234509056244e-06, "loss": 3.4219, "step": 31411 }, { "epoch": 5.98, "learning_rate": 1.2061773117254528e-06, "loss": 2.3301, "step": 31412 }, { "epoch": 5.98, "learning_rate": 1.2061201143946615e-06, "loss": 2.3678, "step": 31413 }, { "epoch": 5.98, "learning_rate": 1.2060629170638706e-06, "loss": 1.5201, "step": 31414 }, { "epoch": 5.98, "learning_rate": 1.2060057197330791e-06, "loss": 0.9059, "step": 31415 }, { "epoch": 5.98, "learning_rate": 1.205948522402288e-06, "loss": 2.3561, "step": 31416 }, { "epoch": 5.98, "learning_rate": 1.2058913250714967e-06, "loss": 1.9072, "step": 31417 }, { "epoch": 5.98, "learning_rate": 1.2058341277407053e-06, "loss": 2.8929, "step": 31418 }, { "epoch": 5.98, "learning_rate": 1.2057769304099143e-06, "loss": 2.7405, "step": 31419 }, { "epoch": 5.98, "learning_rate": 1.205719733079123e-06, "loss": 1.778, "step": 31420 }, { "epoch": 5.98, "learning_rate": 1.205662535748332e-06, "loss": 1.2824, "step": 31421 }, { "epoch": 5.99, "learning_rate": 1.2056053384175405e-06, "loss": 1.974, "step": 31422 }, { "epoch": 5.99, "learning_rate": 1.2055481410867493e-06, "loss": 1.2878, "step": 31423 }, { "epoch": 5.99, "learning_rate": 1.205490943755958e-06, "loss": 0.9394, "step": 31424 }, { "epoch": 5.99, "learning_rate": 1.2054337464251669e-06, "loss": 1.524, "step": 31425 }, { "epoch": 5.99, "learning_rate": 1.2053765490943757e-06, "loss": 1.166, "step": 31426 }, { "epoch": 5.99, "learning_rate": 1.2053193517635844e-06, "loss": 2.0516, "step": 31427 }, { "epoch": 5.99, "learning_rate": 1.205262154432793e-06, "loss": 1.7759, "step": 31428 }, { "epoch": 5.99, "learning_rate": 1.2052049571020018e-06, "loss": 1.076, "step": 31429 }, { "epoch": 5.99, "learning_rate": 1.2051477597712106e-06, "loss": 2.7266, "step": 31430 }, { "epoch": 5.99, "learning_rate": 1.2050905624404196e-06, "loss": 3.0938, "step": 31431 }, { "epoch": 5.99, "learning_rate": 1.2050333651096282e-06, "loss": 2.8334, "step": 31432 }, { "epoch": 5.99, "learning_rate": 1.204976167778837e-06, "loss": 2.0785, "step": 31433 }, { "epoch": 5.99, "learning_rate": 1.2049189704480458e-06, "loss": 2.1018, "step": 31434 }, { "epoch": 5.99, "learning_rate": 1.2048617731172546e-06, "loss": 1.0397, "step": 31435 }, { "epoch": 5.99, "learning_rate": 1.2048045757864634e-06, "loss": 0.6526, "step": 31436 }, { "epoch": 5.99, "learning_rate": 1.2047473784556722e-06, "loss": 2.0224, "step": 31437 }, { "epoch": 5.99, "learning_rate": 1.204690181124881e-06, "loss": 2.8129, "step": 31438 }, { "epoch": 5.99, "learning_rate": 1.2046329837940896e-06, "loss": 0.9988, "step": 31439 }, { "epoch": 5.99, "learning_rate": 1.2045757864632983e-06, "loss": 1.0046, "step": 31440 }, { "epoch": 5.99, "learning_rate": 1.2045185891325071e-06, "loss": 1.1337, "step": 31441 }, { "epoch": 5.99, "learning_rate": 1.204461391801716e-06, "loss": 1.6945, "step": 31442 }, { "epoch": 5.99, "learning_rate": 1.2044041944709247e-06, "loss": 0.6019, "step": 31443 }, { "epoch": 5.99, "learning_rate": 1.2043469971401335e-06, "loss": 0.9887, "step": 31444 }, { "epoch": 5.99, "learning_rate": 1.2042897998093423e-06, "loss": 1.0203, "step": 31445 }, { "epoch": 5.99, "learning_rate": 1.204232602478551e-06, "loss": 0.7746, "step": 31446 }, { "epoch": 5.99, "learning_rate": 1.20417540514776e-06, "loss": 1.0505, "step": 31447 }, { "epoch": 5.99, "learning_rate": 1.2041182078169687e-06, "loss": 1.9083, "step": 31448 }, { "epoch": 5.99, "learning_rate": 1.2040610104861773e-06, "loss": 0.8175, "step": 31449 }, { "epoch": 5.99, "learning_rate": 1.204003813155386e-06, "loss": 2.7161, "step": 31450 }, { "epoch": 5.99, "learning_rate": 1.2039466158245949e-06, "loss": 1.3825, "step": 31451 }, { "epoch": 5.99, "learning_rate": 1.2038894184938037e-06, "loss": 0.9738, "step": 31452 }, { "epoch": 5.99, "learning_rate": 1.2038322211630125e-06, "loss": 2.5281, "step": 31453 }, { "epoch": 5.99, "learning_rate": 1.2037750238322212e-06, "loss": 1.9706, "step": 31454 }, { "epoch": 5.99, "learning_rate": 1.20371782650143e-06, "loss": 1.9394, "step": 31455 }, { "epoch": 5.99, "learning_rate": 1.2036606291706386e-06, "loss": 2.0353, "step": 31456 }, { "epoch": 5.99, "learning_rate": 1.2036034318398474e-06, "loss": 1.9568, "step": 31457 }, { "epoch": 5.99, "learning_rate": 1.2035462345090564e-06, "loss": 1.7025, "step": 31458 }, { "epoch": 5.99, "learning_rate": 1.203489037178265e-06, "loss": 0.7409, "step": 31459 }, { "epoch": 5.99, "learning_rate": 1.2034318398474738e-06, "loss": 0.9664, "step": 31460 }, { "epoch": 5.99, "learning_rate": 1.2033746425166826e-06, "loss": 2.2674, "step": 31461 }, { "epoch": 5.99, "learning_rate": 1.2033174451858914e-06, "loss": 1.8899, "step": 31462 }, { "epoch": 5.99, "learning_rate": 1.2032602478551002e-06, "loss": 2.3721, "step": 31463 }, { "epoch": 5.99, "learning_rate": 1.203203050524309e-06, "loss": 2.1384, "step": 31464 }, { "epoch": 5.99, "learning_rate": 1.2031458531935178e-06, "loss": 2.4692, "step": 31465 }, { "epoch": 5.99, "learning_rate": 1.2030886558627264e-06, "loss": 1.1601, "step": 31466 }, { "epoch": 5.99, "learning_rate": 1.2030314585319351e-06, "loss": 2.0405, "step": 31467 }, { "epoch": 5.99, "learning_rate": 1.202974261201144e-06, "loss": 1.3433, "step": 31468 }, { "epoch": 5.99, "learning_rate": 1.2029170638703527e-06, "loss": 1.0182, "step": 31469 }, { "epoch": 5.99, "learning_rate": 1.2028598665395615e-06, "loss": 2.8615, "step": 31470 }, { "epoch": 5.99, "learning_rate": 1.2028026692087703e-06, "loss": 1.4306, "step": 31471 }, { "epoch": 5.99, "learning_rate": 1.2027454718779791e-06, "loss": 1.4308, "step": 31472 }, { "epoch": 5.99, "learning_rate": 1.2026882745471877e-06, "loss": 2.0669, "step": 31473 }, { "epoch": 6.0, "learning_rate": 1.2026310772163965e-06, "loss": 1.7397, "step": 31474 }, { "epoch": 6.0, "learning_rate": 1.2025738798856055e-06, "loss": 1.0879, "step": 31475 }, { "epoch": 6.0, "learning_rate": 1.202516682554814e-06, "loss": 2.1532, "step": 31476 }, { "epoch": 6.0, "learning_rate": 1.2024594852240229e-06, "loss": 2.215, "step": 31477 }, { "epoch": 6.0, "learning_rate": 1.2024022878932317e-06, "loss": 0.9984, "step": 31478 }, { "epoch": 6.0, "learning_rate": 1.2023450905624405e-06, "loss": 1.7052, "step": 31479 }, { "epoch": 6.0, "learning_rate": 1.2022878932316493e-06, "loss": 1.3202, "step": 31480 }, { "epoch": 6.0, "learning_rate": 1.202230695900858e-06, "loss": 1.8836, "step": 31481 }, { "epoch": 6.0, "learning_rate": 1.2021734985700668e-06, "loss": 1.7505, "step": 31482 }, { "epoch": 6.0, "learning_rate": 1.2021163012392754e-06, "loss": 1.9956, "step": 31483 }, { "epoch": 6.0, "learning_rate": 1.2020591039084842e-06, "loss": 2.3581, "step": 31484 }, { "epoch": 6.0, "learning_rate": 1.202001906577693e-06, "loss": 2.2078, "step": 31485 }, { "epoch": 6.0, "learning_rate": 1.201944709246902e-06, "loss": 3.8625, "step": 31486 }, { "epoch": 6.0, "learning_rate": 1.2018875119161106e-06, "loss": 2.0709, "step": 31487 }, { "epoch": 6.0, "learning_rate": 1.2018303145853194e-06, "loss": 2.4819, "step": 31488 }, { "epoch": 6.0, "learning_rate": 1.2017731172545282e-06, "loss": 1.8463, "step": 31489 }, { "epoch": 6.0, "learning_rate": 1.2017159199237368e-06, "loss": 1.9028, "step": 31490 }, { "epoch": 6.0, "learning_rate": 1.2016587225929458e-06, "loss": 2.2333, "step": 31491 }, { "epoch": 6.0, "learning_rate": 1.2016015252621546e-06, "loss": 3.1825, "step": 31492 }, { "epoch": 6.0, "learning_rate": 1.2015443279313631e-06, "loss": 3.044, "step": 31493 }, { "epoch": 6.0, "learning_rate": 1.201487130600572e-06, "loss": 1.4507, "step": 31494 }, { "epoch": 6.0, "learning_rate": 1.2014299332697807e-06, "loss": 0.6709, "step": 31495 }, { "epoch": 6.0, "learning_rate": 1.2013727359389895e-06, "loss": 2.4611, "step": 31496 }, { "epoch": 6.0, "learning_rate": 1.2013155386081983e-06, "loss": 1.739, "step": 31497 }, { "epoch": 6.0, "learning_rate": 1.2012583412774071e-06, "loss": 2.1508, "step": 31498 }, { "epoch": 6.0, "learning_rate": 1.201201143946616e-06, "loss": 0.9895, "step": 31499 }, { "epoch": 6.0, "learning_rate": 1.2011439466158245e-06, "loss": 4.1185, "step": 31500 }, { "epoch": 6.0, "eval_accuracy": 0.5435062151735962, "eval_loss": 1.735284447669983, "eval_runtime": 126.2038, "eval_samples_per_second": 18.486, "eval_steps_per_second": 4.627, "step": 31500 }, { "epoch": 6.0, "learning_rate": 1.2010867492850333e-06, "loss": 0.7319, "step": 31501 }, { "epoch": 6.0, "learning_rate": 1.2010295519542423e-06, "loss": 2.6418, "step": 31502 }, { "epoch": 6.0, "learning_rate": 1.200972354623451e-06, "loss": 1.4912, "step": 31503 }, { "epoch": 6.0, "learning_rate": 1.2009151572926597e-06, "loss": 1.8824, "step": 31504 }, { "epoch": 6.0, "learning_rate": 1.2008579599618685e-06, "loss": 1.8578, "step": 31505 }, { "epoch": 6.0, "learning_rate": 1.2008007626310773e-06, "loss": 1.4437, "step": 31506 }, { "epoch": 6.0, "learning_rate": 1.200743565300286e-06, "loss": 1.0723, "step": 31507 }, { "epoch": 6.0, "learning_rate": 1.2006863679694948e-06, "loss": 1.3901, "step": 31508 }, { "epoch": 6.0, "learning_rate": 1.2006291706387036e-06, "loss": 1.7267, "step": 31509 }, { "epoch": 6.0, "learning_rate": 1.2005719733079124e-06, "loss": 1.2245, "step": 31510 }, { "epoch": 6.0, "learning_rate": 1.200514775977121e-06, "loss": 1.5166, "step": 31511 }, { "epoch": 6.0, "learning_rate": 1.2004575786463298e-06, "loss": 2.1826, "step": 31512 }, { "epoch": 6.0, "learning_rate": 1.2004003813155386e-06, "loss": 2.3567, "step": 31513 }, { "epoch": 6.0, "learning_rate": 1.2003431839847474e-06, "loss": 1.6277, "step": 31514 }, { "epoch": 6.0, "learning_rate": 1.2002859866539562e-06, "loss": 2.1096, "step": 31515 }, { "epoch": 6.0, "learning_rate": 1.200228789323165e-06, "loss": 1.1862, "step": 31516 }, { "epoch": 6.0, "learning_rate": 1.2001715919923736e-06, "loss": 2.2421, "step": 31517 }, { "epoch": 6.0, "learning_rate": 1.2001143946615824e-06, "loss": 1.4973, "step": 31518 }, { "epoch": 6.0, "learning_rate": 1.2000571973307914e-06, "loss": 1.0289, "step": 31519 }, { "epoch": 6.0, "learning_rate": 1.2000000000000002e-06, "loss": 1.1662, "step": 31520 }, { "epoch": 6.0, "learning_rate": 1.1999428026692087e-06, "loss": 2.2922, "step": 31521 }, { "epoch": 6.0, "learning_rate": 1.1998856053384175e-06, "loss": 1.7843, "step": 31522 }, { "epoch": 6.0, "learning_rate": 1.1998284080076263e-06, "loss": 1.9269, "step": 31523 }, { "epoch": 6.0, "learning_rate": 1.1997712106768351e-06, "loss": 2.2598, "step": 31524 }, { "epoch": 6.0, "learning_rate": 1.199714013346044e-06, "loss": 1.742, "step": 31525 }, { "epoch": 6.0, "learning_rate": 1.1996568160152527e-06, "loss": 2.1697, "step": 31526 }, { "epoch": 6.01, "learning_rate": 1.1995996186844615e-06, "loss": 0.4223, "step": 31527 }, { "epoch": 6.01, "learning_rate": 1.19954242135367e-06, "loss": 0.9312, "step": 31528 }, { "epoch": 6.01, "learning_rate": 1.1994852240228789e-06, "loss": 1.6476, "step": 31529 }, { "epoch": 6.01, "learning_rate": 1.1994280266920879e-06, "loss": 1.9837, "step": 31530 }, { "epoch": 6.01, "learning_rate": 1.1993708293612965e-06, "loss": 2.8121, "step": 31531 }, { "epoch": 6.01, "learning_rate": 1.1993136320305053e-06, "loss": 1.0145, "step": 31532 }, { "epoch": 6.01, "learning_rate": 1.199256434699714e-06, "loss": 1.6045, "step": 31533 }, { "epoch": 6.01, "learning_rate": 1.1991992373689228e-06, "loss": 1.4763, "step": 31534 }, { "epoch": 6.01, "learning_rate": 1.1991420400381316e-06, "loss": 2.3688, "step": 31535 }, { "epoch": 6.01, "learning_rate": 1.1990848427073404e-06, "loss": 1.2202, "step": 31536 }, { "epoch": 6.01, "learning_rate": 1.1990276453765492e-06, "loss": 1.7308, "step": 31537 }, { "epoch": 6.01, "learning_rate": 1.1989704480457578e-06, "loss": 1.8461, "step": 31538 }, { "epoch": 6.01, "learning_rate": 1.1989132507149666e-06, "loss": 0.9368, "step": 31539 }, { "epoch": 6.01, "learning_rate": 1.1988560533841754e-06, "loss": 0.7205, "step": 31540 }, { "epoch": 6.01, "learning_rate": 1.1987988560533842e-06, "loss": 1.5055, "step": 31541 }, { "epoch": 6.01, "learning_rate": 1.198741658722593e-06, "loss": 2.3122, "step": 31542 }, { "epoch": 6.01, "learning_rate": 1.1986844613918018e-06, "loss": 0.9451, "step": 31543 }, { "epoch": 6.01, "learning_rate": 1.1986272640610106e-06, "loss": 1.4298, "step": 31544 }, { "epoch": 6.01, "learning_rate": 1.1985700667302192e-06, "loss": 1.7493, "step": 31545 }, { "epoch": 6.01, "learning_rate": 1.1985128693994282e-06, "loss": 3.4987, "step": 31546 }, { "epoch": 6.01, "learning_rate": 1.198455672068637e-06, "loss": 1.2479, "step": 31547 }, { "epoch": 6.01, "learning_rate": 1.1983984747378455e-06, "loss": 3.088, "step": 31548 }, { "epoch": 6.01, "learning_rate": 1.1983412774070543e-06, "loss": 1.7697, "step": 31549 }, { "epoch": 6.01, "learning_rate": 1.1982840800762631e-06, "loss": 2.8818, "step": 31550 }, { "epoch": 6.01, "learning_rate": 1.198226882745472e-06, "loss": 1.6397, "step": 31551 }, { "epoch": 6.01, "learning_rate": 1.1981696854146807e-06, "loss": 1.9342, "step": 31552 }, { "epoch": 6.01, "learning_rate": 1.1981124880838895e-06, "loss": 1.1776, "step": 31553 }, { "epoch": 6.01, "learning_rate": 1.1980552907530983e-06, "loss": 1.1035, "step": 31554 }, { "epoch": 6.01, "learning_rate": 1.1979980934223069e-06, "loss": 1.7938, "step": 31555 }, { "epoch": 6.01, "learning_rate": 1.1979408960915157e-06, "loss": 0.8965, "step": 31556 }, { "epoch": 6.01, "learning_rate": 1.1978836987607245e-06, "loss": 1.8853, "step": 31557 }, { "epoch": 6.01, "learning_rate": 1.1978265014299333e-06, "loss": 2.793, "step": 31558 }, { "epoch": 6.01, "learning_rate": 1.197769304099142e-06, "loss": 0.7766, "step": 31559 }, { "epoch": 6.01, "learning_rate": 1.1977121067683509e-06, "loss": 1.8978, "step": 31560 }, { "epoch": 6.01, "learning_rate": 1.1976549094375596e-06, "loss": 1.5915, "step": 31561 }, { "epoch": 6.01, "learning_rate": 1.1975977121067682e-06, "loss": 1.1609, "step": 31562 }, { "epoch": 6.01, "learning_rate": 1.1975405147759772e-06, "loss": 1.2002, "step": 31563 }, { "epoch": 6.01, "learning_rate": 1.197483317445186e-06, "loss": 2.3139, "step": 31564 }, { "epoch": 6.01, "learning_rate": 1.1974261201143946e-06, "loss": 2.3349, "step": 31565 }, { "epoch": 6.01, "learning_rate": 1.1973689227836034e-06, "loss": 2.2173, "step": 31566 }, { "epoch": 6.01, "learning_rate": 1.1973117254528122e-06, "loss": 1.1034, "step": 31567 }, { "epoch": 6.01, "learning_rate": 1.197254528122021e-06, "loss": 1.8759, "step": 31568 }, { "epoch": 6.01, "learning_rate": 1.1971973307912298e-06, "loss": 2.042, "step": 31569 }, { "epoch": 6.01, "learning_rate": 1.1971401334604386e-06, "loss": 1.188, "step": 31570 }, { "epoch": 6.01, "learning_rate": 1.1970829361296474e-06, "loss": 1.8117, "step": 31571 }, { "epoch": 6.01, "learning_rate": 1.197025738798856e-06, "loss": 2.3357, "step": 31572 }, { "epoch": 6.01, "learning_rate": 1.1969685414680647e-06, "loss": 1.7172, "step": 31573 }, { "epoch": 6.01, "learning_rate": 1.1969113441372738e-06, "loss": 1.8971, "step": 31574 }, { "epoch": 6.01, "learning_rate": 1.1968541468064825e-06, "loss": 1.2752, "step": 31575 }, { "epoch": 6.01, "learning_rate": 1.1967969494756911e-06, "loss": 1.2684, "step": 31576 }, { "epoch": 6.01, "learning_rate": 1.1967397521449e-06, "loss": 3.0369, "step": 31577 }, { "epoch": 6.01, "learning_rate": 1.1966825548141087e-06, "loss": 2.9537, "step": 31578 }, { "epoch": 6.02, "learning_rate": 1.1966253574833175e-06, "loss": 1.5618, "step": 31579 }, { "epoch": 6.02, "learning_rate": 1.1965681601525263e-06, "loss": 2.0011, "step": 31580 }, { "epoch": 6.02, "learning_rate": 1.196510962821735e-06, "loss": 1.46, "step": 31581 }, { "epoch": 6.02, "learning_rate": 1.1964537654909437e-06, "loss": 0.8188, "step": 31582 }, { "epoch": 6.02, "learning_rate": 1.1963965681601525e-06, "loss": 2.3959, "step": 31583 }, { "epoch": 6.02, "learning_rate": 1.1963393708293613e-06, "loss": 1.4031, "step": 31584 }, { "epoch": 6.02, "learning_rate": 1.19628217349857e-06, "loss": 2.5888, "step": 31585 }, { "epoch": 6.02, "learning_rate": 1.1962249761677789e-06, "loss": 0.7765, "step": 31586 }, { "epoch": 6.02, "learning_rate": 1.1961677788369876e-06, "loss": 1.5052, "step": 31587 }, { "epoch": 6.02, "learning_rate": 1.1961105815061964e-06, "loss": 3.1231, "step": 31588 }, { "epoch": 6.02, "learning_rate": 1.196053384175405e-06, "loss": 0.4245, "step": 31589 }, { "epoch": 6.02, "learning_rate": 1.195996186844614e-06, "loss": 1.8605, "step": 31590 }, { "epoch": 6.02, "learning_rate": 1.1959389895138228e-06, "loss": 1.1831, "step": 31591 }, { "epoch": 6.02, "learning_rate": 1.1958817921830316e-06, "loss": 1.4875, "step": 31592 }, { "epoch": 6.02, "learning_rate": 1.1958245948522402e-06, "loss": 2.9389, "step": 31593 }, { "epoch": 6.02, "learning_rate": 1.195767397521449e-06, "loss": 3.0863, "step": 31594 }, { "epoch": 6.02, "learning_rate": 1.1957102001906578e-06, "loss": 2.6113, "step": 31595 }, { "epoch": 6.02, "learning_rate": 1.1956530028598666e-06, "loss": 0.8519, "step": 31596 }, { "epoch": 6.02, "learning_rate": 1.1955958055290754e-06, "loss": 1.0977, "step": 31597 }, { "epoch": 6.02, "learning_rate": 1.1955386081982842e-06, "loss": 1.2698, "step": 31598 }, { "epoch": 6.02, "learning_rate": 1.195481410867493e-06, "loss": 2.2173, "step": 31599 }, { "epoch": 6.02, "learning_rate": 1.1954242135367015e-06, "loss": 1.8255, "step": 31600 }, { "epoch": 6.02, "learning_rate": 1.1953670162059103e-06, "loss": 2.6769, "step": 31601 }, { "epoch": 6.02, "learning_rate": 1.1953098188751193e-06, "loss": 1.8301, "step": 31602 }, { "epoch": 6.02, "learning_rate": 1.195252621544328e-06, "loss": 1.9611, "step": 31603 }, { "epoch": 6.02, "learning_rate": 1.1951954242135367e-06, "loss": 0.8592, "step": 31604 }, { "epoch": 6.02, "learning_rate": 1.1951382268827455e-06, "loss": 2.1369, "step": 31605 }, { "epoch": 6.02, "learning_rate": 1.195081029551954e-06, "loss": 1.3675, "step": 31606 }, { "epoch": 6.02, "learning_rate": 1.195023832221163e-06, "loss": 3.1548, "step": 31607 }, { "epoch": 6.02, "learning_rate": 1.194966634890372e-06, "loss": 0.7194, "step": 31608 }, { "epoch": 6.02, "learning_rate": 1.1949094375595807e-06, "loss": 3.5184, "step": 31609 }, { "epoch": 6.02, "learning_rate": 1.1948522402287893e-06, "loss": 2.1711, "step": 31610 }, { "epoch": 6.02, "learning_rate": 1.194795042897998e-06, "loss": 1.4777, "step": 31611 }, { "epoch": 6.02, "learning_rate": 1.1947378455672069e-06, "loss": 1.2364, "step": 31612 }, { "epoch": 6.02, "learning_rate": 1.1946806482364157e-06, "loss": 1.9045, "step": 31613 }, { "epoch": 6.02, "learning_rate": 1.1946234509056244e-06, "loss": 1.1833, "step": 31614 }, { "epoch": 6.02, "learning_rate": 1.1945662535748332e-06, "loss": 1.3423, "step": 31615 }, { "epoch": 6.02, "learning_rate": 1.194509056244042e-06, "loss": 1.9155, "step": 31616 }, { "epoch": 6.02, "learning_rate": 1.1944518589132506e-06, "loss": 1.21, "step": 31617 }, { "epoch": 6.02, "learning_rate": 1.1943946615824596e-06, "loss": 0.5706, "step": 31618 }, { "epoch": 6.02, "learning_rate": 1.1943374642516684e-06, "loss": 1.8459, "step": 31619 }, { "epoch": 6.02, "learning_rate": 1.194280266920877e-06, "loss": 3.1039, "step": 31620 }, { "epoch": 6.02, "learning_rate": 1.1942230695900858e-06, "loss": 2.5739, "step": 31621 }, { "epoch": 6.02, "learning_rate": 1.1941658722592946e-06, "loss": 1.2555, "step": 31622 }, { "epoch": 6.02, "learning_rate": 1.1941086749285034e-06, "loss": 2.253, "step": 31623 }, { "epoch": 6.02, "learning_rate": 1.1940514775977122e-06, "loss": 1.8268, "step": 31624 }, { "epoch": 6.02, "learning_rate": 1.193994280266921e-06, "loss": 2.5119, "step": 31625 }, { "epoch": 6.02, "learning_rate": 1.1939370829361298e-06, "loss": 1.708, "step": 31626 }, { "epoch": 6.02, "learning_rate": 1.1938798856053383e-06, "loss": 1.0493, "step": 31627 }, { "epoch": 6.02, "learning_rate": 1.1938226882745471e-06, "loss": 1.7856, "step": 31628 }, { "epoch": 6.02, "learning_rate": 1.193765490943756e-06, "loss": 1.5643, "step": 31629 }, { "epoch": 6.02, "learning_rate": 1.1937082936129647e-06, "loss": 2.6185, "step": 31630 }, { "epoch": 6.02, "learning_rate": 1.1936510962821735e-06, "loss": 1.3425, "step": 31631 }, { "epoch": 6.03, "learning_rate": 1.1935938989513823e-06, "loss": 1.0476, "step": 31632 }, { "epoch": 6.03, "learning_rate": 1.1935367016205911e-06, "loss": 2.6095, "step": 31633 }, { "epoch": 6.03, "learning_rate": 1.1934795042898e-06, "loss": 1.8888, "step": 31634 }, { "epoch": 6.03, "learning_rate": 1.1934223069590087e-06, "loss": 1.3721, "step": 31635 }, { "epoch": 6.03, "learning_rate": 1.1933651096282175e-06, "loss": 2.3096, "step": 31636 }, { "epoch": 6.03, "learning_rate": 1.193307912297426e-06, "loss": 1.473, "step": 31637 }, { "epoch": 6.03, "learning_rate": 1.1932507149666349e-06, "loss": 1.0679, "step": 31638 }, { "epoch": 6.03, "learning_rate": 1.1931935176358437e-06, "loss": 2.6306, "step": 31639 }, { "epoch": 6.03, "learning_rate": 1.1931363203050525e-06, "loss": 2.934, "step": 31640 }, { "epoch": 6.03, "learning_rate": 1.1930791229742612e-06, "loss": 2.9301, "step": 31641 }, { "epoch": 6.03, "learning_rate": 1.19302192564347e-06, "loss": 1.4116, "step": 31642 }, { "epoch": 6.03, "learning_rate": 1.1929647283126788e-06, "loss": 0.7255, "step": 31643 }, { "epoch": 6.03, "learning_rate": 1.1929075309818874e-06, "loss": 1.8251, "step": 31644 }, { "epoch": 6.03, "learning_rate": 1.1928503336510962e-06, "loss": 1.1065, "step": 31645 }, { "epoch": 6.03, "learning_rate": 1.1927931363203052e-06, "loss": 1.9236, "step": 31646 }, { "epoch": 6.03, "learning_rate": 1.1927359389895138e-06, "loss": 1.2715, "step": 31647 }, { "epoch": 6.03, "learning_rate": 1.1926787416587226e-06, "loss": 1.2367, "step": 31648 }, { "epoch": 6.03, "learning_rate": 1.1926215443279314e-06, "loss": 1.0401, "step": 31649 }, { "epoch": 6.03, "learning_rate": 1.1925643469971402e-06, "loss": 2.0858, "step": 31650 }, { "epoch": 6.03, "learning_rate": 1.192507149666349e-06, "loss": 2.3843, "step": 31651 }, { "epoch": 6.03, "learning_rate": 1.1924499523355578e-06, "loss": 2.9096, "step": 31652 }, { "epoch": 6.03, "learning_rate": 1.1923927550047666e-06, "loss": 1.4261, "step": 31653 }, { "epoch": 6.03, "learning_rate": 1.1923355576739751e-06, "loss": 2.5145, "step": 31654 }, { "epoch": 6.03, "learning_rate": 1.192278360343184e-06, "loss": 1.7587, "step": 31655 }, { "epoch": 6.03, "learning_rate": 1.1922211630123927e-06, "loss": 1.5703, "step": 31656 }, { "epoch": 6.03, "learning_rate": 1.1921639656816017e-06, "loss": 2.1855, "step": 31657 }, { "epoch": 6.03, "learning_rate": 1.1921067683508103e-06, "loss": 1.9162, "step": 31658 }, { "epoch": 6.03, "learning_rate": 1.1920495710200191e-06, "loss": 1.2283, "step": 31659 }, { "epoch": 6.03, "learning_rate": 1.191992373689228e-06, "loss": 1.1467, "step": 31660 }, { "epoch": 6.03, "learning_rate": 1.1919351763584365e-06, "loss": 1.6613, "step": 31661 }, { "epoch": 6.03, "learning_rate": 1.1918779790276455e-06, "loss": 0.4553, "step": 31662 }, { "epoch": 6.03, "learning_rate": 1.1918207816968543e-06, "loss": 2.0524, "step": 31663 }, { "epoch": 6.03, "learning_rate": 1.191763584366063e-06, "loss": 1.0876, "step": 31664 }, { "epoch": 6.03, "learning_rate": 1.1917063870352717e-06, "loss": 2.9009, "step": 31665 }, { "epoch": 6.03, "learning_rate": 1.1916491897044805e-06, "loss": 2.4441, "step": 31666 }, { "epoch": 6.03, "learning_rate": 1.1915919923736892e-06, "loss": 2.0293, "step": 31667 }, { "epoch": 6.03, "learning_rate": 1.191534795042898e-06, "loss": 1.3412, "step": 31668 }, { "epoch": 6.03, "learning_rate": 1.1914775977121068e-06, "loss": 1.7104, "step": 31669 }, { "epoch": 6.03, "learning_rate": 1.1914204003813156e-06, "loss": 3.0285, "step": 31670 }, { "epoch": 6.03, "learning_rate": 1.1913632030505242e-06, "loss": 2.5065, "step": 31671 }, { "epoch": 6.03, "learning_rate": 1.191306005719733e-06, "loss": 2.0601, "step": 31672 }, { "epoch": 6.03, "learning_rate": 1.1912488083889418e-06, "loss": 1.5777, "step": 31673 }, { "epoch": 6.03, "learning_rate": 1.1911916110581508e-06, "loss": 1.48, "step": 31674 }, { "epoch": 6.03, "learning_rate": 1.1911344137273594e-06, "loss": 0.6285, "step": 31675 }, { "epoch": 6.03, "learning_rate": 1.1910772163965682e-06, "loss": 0.8288, "step": 31676 }, { "epoch": 6.03, "learning_rate": 1.191020019065777e-06, "loss": 2.3873, "step": 31677 }, { "epoch": 6.03, "learning_rate": 1.1909628217349856e-06, "loss": 1.4408, "step": 31678 }, { "epoch": 6.03, "learning_rate": 1.1909056244041946e-06, "loss": 1.3518, "step": 31679 }, { "epoch": 6.03, "learning_rate": 1.1908484270734034e-06, "loss": 3.6538, "step": 31680 }, { "epoch": 6.03, "learning_rate": 1.1907912297426122e-06, "loss": 2.976, "step": 31681 }, { "epoch": 6.03, "learning_rate": 1.1907340324118207e-06, "loss": 1.6454, "step": 31682 }, { "epoch": 6.03, "learning_rate": 1.1906768350810295e-06, "loss": 2.8291, "step": 31683 }, { "epoch": 6.04, "learning_rate": 1.1906196377502383e-06, "loss": 1.5265, "step": 31684 }, { "epoch": 6.04, "learning_rate": 1.1905624404194471e-06, "loss": 2.3377, "step": 31685 }, { "epoch": 6.04, "learning_rate": 1.190505243088656e-06, "loss": 1.0823, "step": 31686 }, { "epoch": 6.04, "learning_rate": 1.1904480457578647e-06, "loss": 0.333, "step": 31687 }, { "epoch": 6.04, "learning_rate": 1.1903908484270733e-06, "loss": 2.2526, "step": 31688 }, { "epoch": 6.04, "learning_rate": 1.190333651096282e-06, "loss": 1.3871, "step": 31689 }, { "epoch": 6.04, "learning_rate": 1.190276453765491e-06, "loss": 2.0336, "step": 31690 }, { "epoch": 6.04, "learning_rate": 1.1902192564346999e-06, "loss": 0.8367, "step": 31691 }, { "epoch": 6.04, "learning_rate": 1.1901620591039085e-06, "loss": 2.4076, "step": 31692 }, { "epoch": 6.04, "learning_rate": 1.1901048617731173e-06, "loss": 2.2538, "step": 31693 }, { "epoch": 6.04, "learning_rate": 1.190047664442326e-06, "loss": 1.5021, "step": 31694 }, { "epoch": 6.04, "learning_rate": 1.1899904671115348e-06, "loss": 2.4618, "step": 31695 }, { "epoch": 6.04, "learning_rate": 1.1899332697807436e-06, "loss": 1.361, "step": 31696 }, { "epoch": 6.04, "learning_rate": 1.1898760724499524e-06, "loss": 1.2497, "step": 31697 }, { "epoch": 6.04, "learning_rate": 1.1898188751191612e-06, "loss": 1.0051, "step": 31698 }, { "epoch": 6.04, "learning_rate": 1.1897616777883698e-06, "loss": 1.3457, "step": 31699 }, { "epoch": 6.04, "learning_rate": 1.1897044804575786e-06, "loss": 3.0301, "step": 31700 }, { "epoch": 6.04, "learning_rate": 1.1896472831267876e-06, "loss": 0.9218, "step": 31701 }, { "epoch": 6.04, "learning_rate": 1.1895900857959962e-06, "loss": 1.8249, "step": 31702 }, { "epoch": 6.04, "learning_rate": 1.189532888465205e-06, "loss": 1.3302, "step": 31703 }, { "epoch": 6.04, "learning_rate": 1.1894756911344138e-06, "loss": 1.5782, "step": 31704 }, { "epoch": 6.04, "learning_rate": 1.1894184938036226e-06, "loss": 1.3984, "step": 31705 }, { "epoch": 6.04, "learning_rate": 1.1893612964728314e-06, "loss": 0.3527, "step": 31706 }, { "epoch": 6.04, "learning_rate": 1.1893040991420402e-06, "loss": 1.0939, "step": 31707 }, { "epoch": 6.04, "learning_rate": 1.189246901811249e-06, "loss": 2.4614, "step": 31708 }, { "epoch": 6.04, "learning_rate": 1.1891897044804575e-06, "loss": 2.2507, "step": 31709 }, { "epoch": 6.04, "learning_rate": 1.1891325071496663e-06, "loss": 1.255, "step": 31710 }, { "epoch": 6.04, "learning_rate": 1.1890753098188751e-06, "loss": 1.0351, "step": 31711 }, { "epoch": 6.04, "learning_rate": 1.189018112488084e-06, "loss": 1.3198, "step": 31712 }, { "epoch": 6.04, "learning_rate": 1.1889609151572927e-06, "loss": 2.2397, "step": 31713 }, { "epoch": 6.04, "learning_rate": 1.1889037178265015e-06, "loss": 2.4172, "step": 31714 }, { "epoch": 6.04, "learning_rate": 1.1888465204957103e-06, "loss": 2.4247, "step": 31715 }, { "epoch": 6.04, "learning_rate": 1.1887893231649189e-06, "loss": 0.9937, "step": 31716 }, { "epoch": 6.04, "learning_rate": 1.1887321258341277e-06, "loss": 1.0925, "step": 31717 }, { "epoch": 6.04, "learning_rate": 1.1886749285033367e-06, "loss": 1.8863, "step": 31718 }, { "epoch": 6.04, "learning_rate": 1.1886177311725453e-06, "loss": 1.372, "step": 31719 }, { "epoch": 6.04, "learning_rate": 1.188560533841754e-06, "loss": 2.5053, "step": 31720 }, { "epoch": 6.04, "learning_rate": 1.1885033365109628e-06, "loss": 2.792, "step": 31721 }, { "epoch": 6.04, "learning_rate": 1.1884461391801716e-06, "loss": 1.8515, "step": 31722 }, { "epoch": 6.04, "learning_rate": 1.1883889418493804e-06, "loss": 1.2183, "step": 31723 }, { "epoch": 6.04, "learning_rate": 1.1883317445185892e-06, "loss": 1.5042, "step": 31724 }, { "epoch": 6.04, "learning_rate": 1.188274547187798e-06, "loss": 1.4445, "step": 31725 }, { "epoch": 6.04, "learning_rate": 1.1882173498570066e-06, "loss": 2.5958, "step": 31726 }, { "epoch": 6.04, "learning_rate": 1.1881601525262154e-06, "loss": 1.2329, "step": 31727 }, { "epoch": 6.04, "learning_rate": 1.1881029551954242e-06, "loss": 1.4251, "step": 31728 }, { "epoch": 6.04, "learning_rate": 1.188045757864633e-06, "loss": 2.2495, "step": 31729 }, { "epoch": 6.04, "learning_rate": 1.1879885605338418e-06, "loss": 1.09, "step": 31730 }, { "epoch": 6.04, "learning_rate": 1.1879313632030506e-06, "loss": 0.5389, "step": 31731 }, { "epoch": 6.04, "learning_rate": 1.1878741658722594e-06, "loss": 1.2748, "step": 31732 }, { "epoch": 6.04, "learning_rate": 1.187816968541468e-06, "loss": 2.1889, "step": 31733 }, { "epoch": 6.04, "learning_rate": 1.187759771210677e-06, "loss": 2.5691, "step": 31734 }, { "epoch": 6.04, "learning_rate": 1.1877025738798857e-06, "loss": 1.3158, "step": 31735 }, { "epoch": 6.04, "learning_rate": 1.1876453765490943e-06, "loss": 1.6919, "step": 31736 }, { "epoch": 6.05, "learning_rate": 1.1875881792183031e-06, "loss": 1.4359, "step": 31737 }, { "epoch": 6.05, "learning_rate": 1.187530981887512e-06, "loss": 2.1831, "step": 31738 }, { "epoch": 6.05, "learning_rate": 1.1874737845567207e-06, "loss": 2.0147, "step": 31739 }, { "epoch": 6.05, "learning_rate": 1.1874165872259295e-06, "loss": 1.7666, "step": 31740 }, { "epoch": 6.05, "learning_rate": 1.1873593898951383e-06, "loss": 1.083, "step": 31741 }, { "epoch": 6.05, "learning_rate": 1.187302192564347e-06, "loss": 1.6704, "step": 31742 }, { "epoch": 6.05, "learning_rate": 1.1872449952335557e-06, "loss": 3.8207, "step": 31743 }, { "epoch": 6.05, "learning_rate": 1.1871877979027645e-06, "loss": 1.8969, "step": 31744 }, { "epoch": 6.05, "learning_rate": 1.1871306005719735e-06, "loss": 2.2176, "step": 31745 }, { "epoch": 6.05, "learning_rate": 1.1870734032411823e-06, "loss": 1.4597, "step": 31746 }, { "epoch": 6.05, "learning_rate": 1.1870162059103908e-06, "loss": 0.5481, "step": 31747 }, { "epoch": 6.05, "learning_rate": 1.1869590085795996e-06, "loss": 2.609, "step": 31748 }, { "epoch": 6.05, "learning_rate": 1.1869018112488084e-06, "loss": 1.9081, "step": 31749 }, { "epoch": 6.05, "learning_rate": 1.1868446139180172e-06, "loss": 2.4662, "step": 31750 }, { "epoch": 6.05, "learning_rate": 1.186787416587226e-06, "loss": 1.527, "step": 31751 }, { "epoch": 6.05, "learning_rate": 1.1867302192564348e-06, "loss": 1.2607, "step": 31752 }, { "epoch": 6.05, "learning_rate": 1.1866730219256434e-06, "loss": 1.9706, "step": 31753 }, { "epoch": 6.05, "learning_rate": 1.1866158245948522e-06, "loss": 0.7282, "step": 31754 }, { "epoch": 6.05, "learning_rate": 1.186558627264061e-06, "loss": 1.3746, "step": 31755 }, { "epoch": 6.05, "learning_rate": 1.1865014299332698e-06, "loss": 2.1222, "step": 31756 }, { "epoch": 6.05, "learning_rate": 1.1864442326024786e-06, "loss": 1.034, "step": 31757 }, { "epoch": 6.05, "learning_rate": 1.1863870352716874e-06, "loss": 1.263, "step": 31758 }, { "epoch": 6.05, "learning_rate": 1.1863298379408962e-06, "loss": 2.2522, "step": 31759 }, { "epoch": 6.05, "learning_rate": 1.1862726406101047e-06, "loss": 1.0585, "step": 31760 }, { "epoch": 6.05, "learning_rate": 1.1862154432793135e-06, "loss": 1.2652, "step": 31761 }, { "epoch": 6.05, "learning_rate": 1.1861582459485225e-06, "loss": 0.9978, "step": 31762 }, { "epoch": 6.05, "learning_rate": 1.1861010486177313e-06, "loss": 1.7927, "step": 31763 }, { "epoch": 6.05, "learning_rate": 1.18604385128694e-06, "loss": 1.7615, "step": 31764 }, { "epoch": 6.05, "learning_rate": 1.1859866539561487e-06, "loss": 3.0424, "step": 31765 }, { "epoch": 6.05, "learning_rate": 1.1859294566253575e-06, "loss": 2.3205, "step": 31766 }, { "epoch": 6.05, "learning_rate": 1.1858722592945663e-06, "loss": 2.2557, "step": 31767 }, { "epoch": 6.05, "learning_rate": 1.185815061963775e-06, "loss": 1.7701, "step": 31768 }, { "epoch": 6.05, "learning_rate": 1.1857578646329839e-06, "loss": 2.1006, "step": 31769 }, { "epoch": 6.05, "learning_rate": 1.1857006673021927e-06, "loss": 1.0689, "step": 31770 }, { "epoch": 6.05, "learning_rate": 1.1856434699714013e-06, "loss": 1.7432, "step": 31771 }, { "epoch": 6.05, "learning_rate": 1.18558627264061e-06, "loss": 1.1713, "step": 31772 }, { "epoch": 6.05, "learning_rate": 1.185529075309819e-06, "loss": 0.657, "step": 31773 }, { "epoch": 6.05, "learning_rate": 1.1854718779790276e-06, "loss": 1.4299, "step": 31774 }, { "epoch": 6.05, "learning_rate": 1.1854146806482364e-06, "loss": 1.4985, "step": 31775 }, { "epoch": 6.05, "learning_rate": 1.1853574833174452e-06, "loss": 1.8795, "step": 31776 }, { "epoch": 6.05, "learning_rate": 1.1853002859866538e-06, "loss": 0.5692, "step": 31777 }, { "epoch": 6.05, "learning_rate": 1.1852430886558628e-06, "loss": 1.433, "step": 31778 }, { "epoch": 6.05, "learning_rate": 1.1851858913250716e-06, "loss": 1.8398, "step": 31779 }, { "epoch": 6.05, "learning_rate": 1.1851286939942804e-06, "loss": 3.4147, "step": 31780 }, { "epoch": 6.05, "learning_rate": 1.185071496663489e-06, "loss": 1.7035, "step": 31781 }, { "epoch": 6.05, "learning_rate": 1.1850142993326978e-06, "loss": 1.9647, "step": 31782 }, { "epoch": 6.05, "learning_rate": 1.1849571020019066e-06, "loss": 1.8684, "step": 31783 }, { "epoch": 6.05, "learning_rate": 1.1848999046711154e-06, "loss": 2.5246, "step": 31784 }, { "epoch": 6.05, "learning_rate": 1.1848427073403242e-06, "loss": 1.273, "step": 31785 }, { "epoch": 6.05, "learning_rate": 1.184785510009533e-06, "loss": 1.9709, "step": 31786 }, { "epoch": 6.05, "learning_rate": 1.1847283126787418e-06, "loss": 1.4977, "step": 31787 }, { "epoch": 6.05, "learning_rate": 1.1846711153479503e-06, "loss": 2.3319, "step": 31788 }, { "epoch": 6.06, "learning_rate": 1.1846139180171593e-06, "loss": 1.8776, "step": 31789 }, { "epoch": 6.06, "learning_rate": 1.1845567206863681e-06, "loss": 2.7894, "step": 31790 }, { "epoch": 6.06, "learning_rate": 1.1844995233555767e-06, "loss": 1.0509, "step": 31791 }, { "epoch": 6.06, "learning_rate": 1.1844423260247855e-06, "loss": 2.2699, "step": 31792 }, { "epoch": 6.06, "learning_rate": 1.1843851286939943e-06, "loss": 1.8729, "step": 31793 }, { "epoch": 6.06, "learning_rate": 1.184327931363203e-06, "loss": 1.4596, "step": 31794 }, { "epoch": 6.06, "learning_rate": 1.1842707340324119e-06, "loss": 1.8031, "step": 31795 }, { "epoch": 6.06, "learning_rate": 1.1842135367016207e-06, "loss": 2.1262, "step": 31796 }, { "epoch": 6.06, "learning_rate": 1.1841563393708295e-06, "loss": 2.3705, "step": 31797 }, { "epoch": 6.06, "learning_rate": 1.184099142040038e-06, "loss": 1.2578, "step": 31798 }, { "epoch": 6.06, "learning_rate": 1.1840419447092469e-06, "loss": 2.2424, "step": 31799 }, { "epoch": 6.06, "learning_rate": 1.1839847473784557e-06, "loss": 2.3913, "step": 31800 }, { "epoch": 6.06, "learning_rate": 1.1839275500476644e-06, "loss": 2.4679, "step": 31801 }, { "epoch": 6.06, "learning_rate": 1.1838703527168732e-06, "loss": 1.2132, "step": 31802 }, { "epoch": 6.06, "learning_rate": 1.183813155386082e-06, "loss": 2.3687, "step": 31803 }, { "epoch": 6.06, "learning_rate": 1.1837559580552908e-06, "loss": 1.3433, "step": 31804 }, { "epoch": 6.06, "learning_rate": 1.1836987607244994e-06, "loss": 1.1706, "step": 31805 }, { "epoch": 6.06, "learning_rate": 1.1836415633937084e-06, "loss": 1.0688, "step": 31806 }, { "epoch": 6.06, "learning_rate": 1.1835843660629172e-06, "loss": 1.753, "step": 31807 }, { "epoch": 6.06, "learning_rate": 1.1835271687321258e-06, "loss": 2.057, "step": 31808 }, { "epoch": 6.06, "learning_rate": 1.1834699714013346e-06, "loss": 1.5478, "step": 31809 }, { "epoch": 6.06, "learning_rate": 1.1834127740705434e-06, "loss": 1.5401, "step": 31810 }, { "epoch": 6.06, "learning_rate": 1.1833555767397522e-06, "loss": 3.015, "step": 31811 }, { "epoch": 6.06, "learning_rate": 1.183298379408961e-06, "loss": 1.6797, "step": 31812 }, { "epoch": 6.06, "learning_rate": 1.1832411820781698e-06, "loss": 1.7311, "step": 31813 }, { "epoch": 6.06, "learning_rate": 1.1831839847473786e-06, "loss": 2.9893, "step": 31814 }, { "epoch": 6.06, "learning_rate": 1.1831267874165871e-06, "loss": 2.3498, "step": 31815 }, { "epoch": 6.06, "learning_rate": 1.183069590085796e-06, "loss": 3.1033, "step": 31816 }, { "epoch": 6.06, "learning_rate": 1.183012392755005e-06, "loss": 1.7783, "step": 31817 }, { "epoch": 6.06, "learning_rate": 1.1829551954242135e-06, "loss": 1.5345, "step": 31818 }, { "epoch": 6.06, "learning_rate": 1.1828979980934223e-06, "loss": 2.8894, "step": 31819 }, { "epoch": 6.06, "learning_rate": 1.182840800762631e-06, "loss": 2.9698, "step": 31820 }, { "epoch": 6.06, "learning_rate": 1.18278360343184e-06, "loss": 0.8696, "step": 31821 }, { "epoch": 6.06, "learning_rate": 1.1827264061010487e-06, "loss": 1.6, "step": 31822 }, { "epoch": 6.06, "learning_rate": 1.1826692087702575e-06, "loss": 2.4772, "step": 31823 }, { "epoch": 6.06, "learning_rate": 1.1826120114394663e-06, "loss": 1.6376, "step": 31824 }, { "epoch": 6.06, "learning_rate": 1.1825548141086749e-06, "loss": 2.0672, "step": 31825 }, { "epoch": 6.06, "learning_rate": 1.1824976167778837e-06, "loss": 2.3553, "step": 31826 }, { "epoch": 6.06, "learning_rate": 1.1824404194470924e-06, "loss": 1.3986, "step": 31827 }, { "epoch": 6.06, "learning_rate": 1.1823832221163012e-06, "loss": 1.9871, "step": 31828 }, { "epoch": 6.06, "learning_rate": 1.18232602478551e-06, "loss": 1.7594, "step": 31829 }, { "epoch": 6.06, "learning_rate": 1.1822688274547188e-06, "loss": 1.9676, "step": 31830 }, { "epoch": 6.06, "learning_rate": 1.1822116301239276e-06, "loss": 2.5316, "step": 31831 }, { "epoch": 6.06, "learning_rate": 1.1821544327931362e-06, "loss": 1.7734, "step": 31832 }, { "epoch": 6.06, "learning_rate": 1.1820972354623452e-06, "loss": 1.0323, "step": 31833 }, { "epoch": 6.06, "learning_rate": 1.182040038131554e-06, "loss": 1.1955, "step": 31834 }, { "epoch": 6.06, "learning_rate": 1.1819828408007628e-06, "loss": 1.2175, "step": 31835 }, { "epoch": 6.06, "learning_rate": 1.1819256434699714e-06, "loss": 1.2933, "step": 31836 }, { "epoch": 6.06, "learning_rate": 1.1818684461391802e-06, "loss": 2.0147, "step": 31837 }, { "epoch": 6.06, "learning_rate": 1.181811248808389e-06, "loss": 1.9705, "step": 31838 }, { "epoch": 6.06, "learning_rate": 1.1817540514775978e-06, "loss": 1.924, "step": 31839 }, { "epoch": 6.06, "learning_rate": 1.1816968541468066e-06, "loss": 0.9128, "step": 31840 }, { "epoch": 6.06, "learning_rate": 1.1816396568160153e-06, "loss": 1.6735, "step": 31841 }, { "epoch": 6.07, "learning_rate": 1.181582459485224e-06, "loss": 1.8399, "step": 31842 }, { "epoch": 6.07, "learning_rate": 1.1815252621544327e-06, "loss": 1.2335, "step": 31843 }, { "epoch": 6.07, "learning_rate": 1.1814680648236415e-06, "loss": 0.412, "step": 31844 }, { "epoch": 6.07, "learning_rate": 1.1814108674928505e-06, "loss": 0.8589, "step": 31845 }, { "epoch": 6.07, "learning_rate": 1.1813536701620591e-06, "loss": 1.8893, "step": 31846 }, { "epoch": 6.07, "learning_rate": 1.181296472831268e-06, "loss": 1.4074, "step": 31847 }, { "epoch": 6.07, "learning_rate": 1.1812392755004767e-06, "loss": 1.4862, "step": 31848 }, { "epoch": 6.07, "learning_rate": 1.1811820781696853e-06, "loss": 0.867, "step": 31849 }, { "epoch": 6.07, "learning_rate": 1.1811248808388943e-06, "loss": 1.6374, "step": 31850 }, { "epoch": 6.07, "learning_rate": 1.181067683508103e-06, "loss": 1.2151, "step": 31851 }, { "epoch": 6.07, "learning_rate": 1.1810104861773119e-06, "loss": 1.7021, "step": 31852 }, { "epoch": 6.07, "learning_rate": 1.1809532888465205e-06, "loss": 2.5612, "step": 31853 }, { "epoch": 6.07, "learning_rate": 1.1808960915157292e-06, "loss": 1.6563, "step": 31854 }, { "epoch": 6.07, "learning_rate": 1.180838894184938e-06, "loss": 1.7755, "step": 31855 }, { "epoch": 6.07, "learning_rate": 1.1807816968541468e-06, "loss": 2.5913, "step": 31856 }, { "epoch": 6.07, "learning_rate": 1.1807244995233556e-06, "loss": 1.2615, "step": 31857 }, { "epoch": 6.07, "learning_rate": 1.1806673021925644e-06, "loss": 1.3488, "step": 31858 }, { "epoch": 6.07, "learning_rate": 1.180610104861773e-06, "loss": 1.7877, "step": 31859 }, { "epoch": 6.07, "learning_rate": 1.1805529075309818e-06, "loss": 2.0554, "step": 31860 }, { "epoch": 6.07, "learning_rate": 1.1804957102001908e-06, "loss": 3.5281, "step": 31861 }, { "epoch": 6.07, "learning_rate": 1.1804385128693996e-06, "loss": 2.0324, "step": 31862 }, { "epoch": 6.07, "learning_rate": 1.1803813155386082e-06, "loss": 1.8212, "step": 31863 }, { "epoch": 6.07, "learning_rate": 1.180324118207817e-06, "loss": 1.7173, "step": 31864 }, { "epoch": 6.07, "learning_rate": 1.1802669208770258e-06, "loss": 2.8813, "step": 31865 }, { "epoch": 6.07, "learning_rate": 1.1802097235462346e-06, "loss": 2.5076, "step": 31866 }, { "epoch": 6.07, "learning_rate": 1.1801525262154434e-06, "loss": 2.1329, "step": 31867 }, { "epoch": 6.07, "learning_rate": 1.1800953288846521e-06, "loss": 3.3058, "step": 31868 }, { "epoch": 6.07, "learning_rate": 1.180038131553861e-06, "loss": 2.3274, "step": 31869 }, { "epoch": 6.07, "learning_rate": 1.1799809342230695e-06, "loss": 1.4103, "step": 31870 }, { "epoch": 6.07, "learning_rate": 1.1799237368922783e-06, "loss": 2.4512, "step": 31871 }, { "epoch": 6.07, "learning_rate": 1.1798665395614871e-06, "loss": 2.1702, "step": 31872 }, { "epoch": 6.07, "learning_rate": 1.179809342230696e-06, "loss": 1.7756, "step": 31873 }, { "epoch": 6.07, "learning_rate": 1.1797521448999047e-06, "loss": 1.8088, "step": 31874 }, { "epoch": 6.07, "learning_rate": 1.1796949475691135e-06, "loss": 2.183, "step": 31875 }, { "epoch": 6.07, "learning_rate": 1.1796377502383223e-06, "loss": 2.1902, "step": 31876 }, { "epoch": 6.07, "learning_rate": 1.1795805529075309e-06, "loss": 2.0139, "step": 31877 }, { "epoch": 6.07, "learning_rate": 1.1795233555767399e-06, "loss": 1.0485, "step": 31878 }, { "epoch": 6.07, "learning_rate": 1.1794661582459487e-06, "loss": 3.3282, "step": 31879 }, { "epoch": 6.07, "learning_rate": 1.1794089609151572e-06, "loss": 1.5117, "step": 31880 }, { "epoch": 6.07, "learning_rate": 1.179351763584366e-06, "loss": 0.6964, "step": 31881 }, { "epoch": 6.07, "learning_rate": 1.1792945662535748e-06, "loss": 1.9069, "step": 31882 }, { "epoch": 6.07, "learning_rate": 1.1792373689227836e-06, "loss": 2.1877, "step": 31883 }, { "epoch": 6.07, "learning_rate": 1.1791801715919924e-06, "loss": 2.2298, "step": 31884 }, { "epoch": 6.07, "learning_rate": 1.1791229742612012e-06, "loss": 2.2212, "step": 31885 }, { "epoch": 6.07, "learning_rate": 1.17906577693041e-06, "loss": 1.2653, "step": 31886 }, { "epoch": 6.07, "learning_rate": 1.1790085795996186e-06, "loss": 1.2738, "step": 31887 }, { "epoch": 6.07, "learning_rate": 1.1789513822688274e-06, "loss": 2.9188, "step": 31888 }, { "epoch": 6.07, "learning_rate": 1.1788941849380364e-06, "loss": 2.3598, "step": 31889 }, { "epoch": 6.07, "learning_rate": 1.178836987607245e-06, "loss": 1.0078, "step": 31890 }, { "epoch": 6.07, "learning_rate": 1.1787797902764538e-06, "loss": 1.2665, "step": 31891 }, { "epoch": 6.07, "learning_rate": 1.1787225929456626e-06, "loss": 1.6167, "step": 31892 }, { "epoch": 6.07, "learning_rate": 1.1786653956148714e-06, "loss": 2.2558, "step": 31893 }, { "epoch": 6.08, "learning_rate": 1.1786081982840802e-06, "loss": 0.8587, "step": 31894 }, { "epoch": 6.08, "learning_rate": 1.178551000953289e-06, "loss": 2.0401, "step": 31895 }, { "epoch": 6.08, "learning_rate": 1.1784938036224977e-06, "loss": 1.903, "step": 31896 }, { "epoch": 6.08, "learning_rate": 1.1784366062917063e-06, "loss": 2.6774, "step": 31897 }, { "epoch": 6.08, "learning_rate": 1.1783794089609151e-06, "loss": 0.3544, "step": 31898 }, { "epoch": 6.08, "learning_rate": 1.178322211630124e-06, "loss": 1.696, "step": 31899 }, { "epoch": 6.08, "learning_rate": 1.178265014299333e-06, "loss": 1.7188, "step": 31900 }, { "epoch": 6.08, "learning_rate": 1.1782078169685415e-06, "loss": 1.5212, "step": 31901 }, { "epoch": 6.08, "learning_rate": 1.1781506196377503e-06, "loss": 0.8524, "step": 31902 }, { "epoch": 6.08, "learning_rate": 1.178093422306959e-06, "loss": 1.8868, "step": 31903 }, { "epoch": 6.08, "learning_rate": 1.1780362249761677e-06, "loss": 1.835, "step": 31904 }, { "epoch": 6.08, "learning_rate": 1.1779790276453767e-06, "loss": 1.3314, "step": 31905 }, { "epoch": 6.08, "learning_rate": 1.1779218303145855e-06, "loss": 0.6553, "step": 31906 }, { "epoch": 6.08, "learning_rate": 1.177864632983794e-06, "loss": 1.2078, "step": 31907 }, { "epoch": 6.08, "learning_rate": 1.1778074356530028e-06, "loss": 2.0263, "step": 31908 }, { "epoch": 6.08, "learning_rate": 1.1777502383222116e-06, "loss": 1.2445, "step": 31909 }, { "epoch": 6.08, "learning_rate": 1.1776930409914204e-06, "loss": 1.2939, "step": 31910 }, { "epoch": 6.08, "learning_rate": 1.1776358436606292e-06, "loss": 2.6127, "step": 31911 }, { "epoch": 6.08, "learning_rate": 1.177578646329838e-06, "loss": 1.3955, "step": 31912 }, { "epoch": 6.08, "learning_rate": 1.1775214489990468e-06, "loss": 1.5792, "step": 31913 }, { "epoch": 6.08, "learning_rate": 1.1774642516682554e-06, "loss": 1.8382, "step": 31914 }, { "epoch": 6.08, "learning_rate": 1.1774070543374642e-06, "loss": 2.0402, "step": 31915 }, { "epoch": 6.08, "learning_rate": 1.177349857006673e-06, "loss": 0.4055, "step": 31916 }, { "epoch": 6.08, "learning_rate": 1.177292659675882e-06, "loss": 1.4655, "step": 31917 }, { "epoch": 6.08, "learning_rate": 1.1772354623450906e-06, "loss": 1.0666, "step": 31918 }, { "epoch": 6.08, "learning_rate": 1.1771782650142994e-06, "loss": 1.1403, "step": 31919 }, { "epoch": 6.08, "learning_rate": 1.1771210676835082e-06, "loss": 0.839, "step": 31920 }, { "epoch": 6.08, "learning_rate": 1.1770638703527167e-06, "loss": 1.3664, "step": 31921 }, { "epoch": 6.08, "learning_rate": 1.1770066730219257e-06, "loss": 1.3336, "step": 31922 }, { "epoch": 6.08, "learning_rate": 1.1769494756911345e-06, "loss": 1.4095, "step": 31923 }, { "epoch": 6.08, "learning_rate": 1.1768922783603431e-06, "loss": 2.0522, "step": 31924 }, { "epoch": 6.08, "learning_rate": 1.176835081029552e-06, "loss": 2.4083, "step": 31925 }, { "epoch": 6.08, "learning_rate": 1.1767778836987607e-06, "loss": 1.554, "step": 31926 }, { "epoch": 6.08, "learning_rate": 1.1767206863679695e-06, "loss": 1.1197, "step": 31927 }, { "epoch": 6.08, "learning_rate": 1.1766634890371783e-06, "loss": 1.3656, "step": 31928 }, { "epoch": 6.08, "learning_rate": 1.176606291706387e-06, "loss": 1.5656, "step": 31929 }, { "epoch": 6.08, "learning_rate": 1.1765490943755959e-06, "loss": 1.859, "step": 31930 }, { "epoch": 6.08, "learning_rate": 1.1764918970448045e-06, "loss": 1.5215, "step": 31931 }, { "epoch": 6.08, "learning_rate": 1.1764346997140133e-06, "loss": 2.0232, "step": 31932 }, { "epoch": 6.08, "learning_rate": 1.1763775023832223e-06, "loss": 0.3677, "step": 31933 }, { "epoch": 6.08, "learning_rate": 1.176320305052431e-06, "loss": 1.7045, "step": 31934 }, { "epoch": 6.08, "learning_rate": 1.1762631077216396e-06, "loss": 2.0589, "step": 31935 }, { "epoch": 6.08, "learning_rate": 1.1762059103908484e-06, "loss": 1.2612, "step": 31936 }, { "epoch": 6.08, "learning_rate": 1.1761487130600572e-06, "loss": 1.6619, "step": 31937 }, { "epoch": 6.08, "learning_rate": 1.176091515729266e-06, "loss": 2.2954, "step": 31938 }, { "epoch": 6.08, "learning_rate": 1.1760343183984748e-06, "loss": 1.8417, "step": 31939 }, { "epoch": 6.08, "learning_rate": 1.1759771210676836e-06, "loss": 2.0452, "step": 31940 }, { "epoch": 6.08, "learning_rate": 1.1759199237368924e-06, "loss": 1.5249, "step": 31941 }, { "epoch": 6.08, "learning_rate": 1.175862726406101e-06, "loss": 2.8807, "step": 31942 }, { "epoch": 6.08, "learning_rate": 1.1758055290753098e-06, "loss": 2.0531, "step": 31943 }, { "epoch": 6.08, "learning_rate": 1.1757483317445188e-06, "loss": 2.8146, "step": 31944 }, { "epoch": 6.08, "learning_rate": 1.1756911344137274e-06, "loss": 1.7584, "step": 31945 }, { "epoch": 6.08, "learning_rate": 1.1756339370829362e-06, "loss": 2.1309, "step": 31946 }, { "epoch": 6.09, "learning_rate": 1.175576739752145e-06, "loss": 1.7438, "step": 31947 }, { "epoch": 6.09, "learning_rate": 1.1755195424213535e-06, "loss": 1.6262, "step": 31948 }, { "epoch": 6.09, "learning_rate": 1.1754623450905625e-06, "loss": 1.7471, "step": 31949 }, { "epoch": 6.09, "learning_rate": 1.1754051477597713e-06, "loss": 2.6688, "step": 31950 }, { "epoch": 6.09, "learning_rate": 1.1753479504289801e-06, "loss": 2.0696, "step": 31951 }, { "epoch": 6.09, "learning_rate": 1.1752907530981887e-06, "loss": 1.5589, "step": 31952 }, { "epoch": 6.09, "learning_rate": 1.1752335557673975e-06, "loss": 1.2348, "step": 31953 }, { "epoch": 6.09, "learning_rate": 1.1751763584366063e-06, "loss": 1.1287, "step": 31954 }, { "epoch": 6.09, "learning_rate": 1.175119161105815e-06, "loss": 2.8453, "step": 31955 }, { "epoch": 6.09, "learning_rate": 1.1750619637750239e-06, "loss": 2.5345, "step": 31956 }, { "epoch": 6.09, "learning_rate": 1.1750047664442327e-06, "loss": 1.3339, "step": 31957 }, { "epoch": 6.09, "learning_rate": 1.1749475691134415e-06, "loss": 2.9106, "step": 31958 }, { "epoch": 6.09, "learning_rate": 1.17489037178265e-06, "loss": 1.3625, "step": 31959 }, { "epoch": 6.09, "learning_rate": 1.1748331744518588e-06, "loss": 1.2999, "step": 31960 }, { "epoch": 6.09, "learning_rate": 1.1747759771210679e-06, "loss": 2.2111, "step": 31961 }, { "epoch": 6.09, "learning_rate": 1.1747187797902764e-06, "loss": 1.6776, "step": 31962 }, { "epoch": 6.09, "learning_rate": 1.1746615824594852e-06, "loss": 0.835, "step": 31963 }, { "epoch": 6.09, "learning_rate": 1.174604385128694e-06, "loss": 1.3557, "step": 31964 }, { "epoch": 6.09, "learning_rate": 1.1745471877979028e-06, "loss": 3.6275, "step": 31965 }, { "epoch": 6.09, "learning_rate": 1.1744899904671116e-06, "loss": 0.7308, "step": 31966 }, { "epoch": 6.09, "learning_rate": 1.1744327931363204e-06, "loss": 2.4585, "step": 31967 }, { "epoch": 6.09, "learning_rate": 1.1743755958055292e-06, "loss": 1.3739, "step": 31968 }, { "epoch": 6.09, "learning_rate": 1.1743183984747378e-06, "loss": 2.2705, "step": 31969 }, { "epoch": 6.09, "learning_rate": 1.1742612011439466e-06, "loss": 1.341, "step": 31970 }, { "epoch": 6.09, "learning_rate": 1.1742040038131554e-06, "loss": 1.4911, "step": 31971 }, { "epoch": 6.09, "learning_rate": 1.1741468064823642e-06, "loss": 2.7177, "step": 31972 }, { "epoch": 6.09, "learning_rate": 1.174089609151573e-06, "loss": 1.5203, "step": 31973 }, { "epoch": 6.09, "learning_rate": 1.1740324118207818e-06, "loss": 1.3737, "step": 31974 }, { "epoch": 6.09, "learning_rate": 1.1739752144899905e-06, "loss": 2.6791, "step": 31975 }, { "epoch": 6.09, "learning_rate": 1.1739180171591991e-06, "loss": 1.7337, "step": 31976 }, { "epoch": 6.09, "learning_rate": 1.1738608198284081e-06, "loss": 1.4672, "step": 31977 }, { "epoch": 6.09, "learning_rate": 1.173803622497617e-06, "loss": 1.9554, "step": 31978 }, { "epoch": 6.09, "learning_rate": 1.1737464251668255e-06, "loss": 0.976, "step": 31979 }, { "epoch": 6.09, "learning_rate": 1.1736892278360343e-06, "loss": 1.55, "step": 31980 }, { "epoch": 6.09, "learning_rate": 1.173632030505243e-06, "loss": 1.6113, "step": 31981 }, { "epoch": 6.09, "learning_rate": 1.1735748331744519e-06, "loss": 1.8831, "step": 31982 }, { "epoch": 6.09, "learning_rate": 1.1735176358436607e-06, "loss": 1.3702, "step": 31983 }, { "epoch": 6.09, "learning_rate": 1.1734604385128695e-06, "loss": 2.1985, "step": 31984 }, { "epoch": 6.09, "learning_rate": 1.1734032411820783e-06, "loss": 1.7946, "step": 31985 }, { "epoch": 6.09, "learning_rate": 1.1733460438512869e-06, "loss": 0.5982, "step": 31986 }, { "epoch": 6.09, "learning_rate": 1.1732888465204956e-06, "loss": 2.0286, "step": 31987 }, { "epoch": 6.09, "learning_rate": 1.1732316491897047e-06, "loss": 0.8519, "step": 31988 }, { "epoch": 6.09, "learning_rate": 1.1731744518589132e-06, "loss": 1.5256, "step": 31989 }, { "epoch": 6.09, "learning_rate": 1.173117254528122e-06, "loss": 1.2966, "step": 31990 }, { "epoch": 6.09, "learning_rate": 1.1730600571973308e-06, "loss": 1.9442, "step": 31991 }, { "epoch": 6.09, "learning_rate": 1.1730028598665396e-06, "loss": 1.0094, "step": 31992 }, { "epoch": 6.09, "learning_rate": 1.1729456625357484e-06, "loss": 1.3137, "step": 31993 }, { "epoch": 6.09, "learning_rate": 1.1728884652049572e-06, "loss": 3.2475, "step": 31994 }, { "epoch": 6.09, "learning_rate": 1.172831267874166e-06, "loss": 0.8798, "step": 31995 }, { "epoch": 6.09, "learning_rate": 1.1727740705433746e-06, "loss": 1.6896, "step": 31996 }, { "epoch": 6.09, "learning_rate": 1.1727168732125834e-06, "loss": 1.2778, "step": 31997 }, { "epoch": 6.09, "learning_rate": 1.1726596758817922e-06, "loss": 1.8585, "step": 31998 }, { "epoch": 6.1, "learning_rate": 1.172602478551001e-06, "loss": 1.5088, "step": 31999 }, { "epoch": 6.1, "learning_rate": 1.1725452812202098e-06, "loss": 1.0129, "step": 32000 }, { "epoch": 6.1, "learning_rate": 1.1724880838894185e-06, "loss": 2.3286, "step": 32001 }, { "epoch": 6.1, "learning_rate": 1.1724308865586273e-06, "loss": 2.7224, "step": 32002 }, { "epoch": 6.1, "learning_rate": 1.172373689227836e-06, "loss": 0.8008, "step": 32003 }, { "epoch": 6.1, "learning_rate": 1.1723164918970447e-06, "loss": 1.333, "step": 32004 }, { "epoch": 6.1, "learning_rate": 1.1722592945662537e-06, "loss": 1.7689, "step": 32005 }, { "epoch": 6.1, "learning_rate": 1.1722020972354625e-06, "loss": 3.2735, "step": 32006 }, { "epoch": 6.1, "learning_rate": 1.172144899904671e-06, "loss": 2.2272, "step": 32007 }, { "epoch": 6.1, "learning_rate": 1.1720877025738799e-06, "loss": 2.1103, "step": 32008 }, { "epoch": 6.1, "learning_rate": 1.1720305052430887e-06, "loss": 1.2363, "step": 32009 }, { "epoch": 6.1, "learning_rate": 1.1719733079122975e-06, "loss": 1.3762, "step": 32010 }, { "epoch": 6.1, "learning_rate": 1.1719161105815063e-06, "loss": 1.2355, "step": 32011 }, { "epoch": 6.1, "learning_rate": 1.171858913250715e-06, "loss": 2.1294, "step": 32012 }, { "epoch": 6.1, "learning_rate": 1.1718017159199237e-06, "loss": 2.6416, "step": 32013 }, { "epoch": 6.1, "learning_rate": 1.1717445185891324e-06, "loss": 2.3239, "step": 32014 }, { "epoch": 6.1, "learning_rate": 1.1716873212583412e-06, "loss": 1.8887, "step": 32015 }, { "epoch": 6.1, "learning_rate": 1.1716301239275502e-06, "loss": 2.9567, "step": 32016 }, { "epoch": 6.1, "learning_rate": 1.1715729265967588e-06, "loss": 1.0806, "step": 32017 }, { "epoch": 6.1, "learning_rate": 1.1715157292659676e-06, "loss": 1.7922, "step": 32018 }, { "epoch": 6.1, "learning_rate": 1.1714585319351764e-06, "loss": 1.459, "step": 32019 }, { "epoch": 6.1, "learning_rate": 1.171401334604385e-06, "loss": 1.5088, "step": 32020 }, { "epoch": 6.1, "learning_rate": 1.171344137273594e-06, "loss": 1.847, "step": 32021 }, { "epoch": 6.1, "learning_rate": 1.1712869399428028e-06, "loss": 2.5855, "step": 32022 }, { "epoch": 6.1, "learning_rate": 1.1712297426120116e-06, "loss": 2.0441, "step": 32023 }, { "epoch": 6.1, "learning_rate": 1.1711725452812202e-06, "loss": 1.1331, "step": 32024 }, { "epoch": 6.1, "learning_rate": 1.171115347950429e-06, "loss": 1.3548, "step": 32025 }, { "epoch": 6.1, "learning_rate": 1.1710581506196378e-06, "loss": 0.8615, "step": 32026 }, { "epoch": 6.1, "learning_rate": 1.1710009532888466e-06, "loss": 2.0265, "step": 32027 }, { "epoch": 6.1, "learning_rate": 1.1709437559580553e-06, "loss": 1.7562, "step": 32028 }, { "epoch": 6.1, "learning_rate": 1.1708865586272641e-06, "loss": 1.3752, "step": 32029 }, { "epoch": 6.1, "learning_rate": 1.1708293612964727e-06, "loss": 1.5056, "step": 32030 }, { "epoch": 6.1, "learning_rate": 1.1707721639656815e-06, "loss": 2.3836, "step": 32031 }, { "epoch": 6.1, "learning_rate": 1.1707149666348903e-06, "loss": 2.9921, "step": 32032 }, { "epoch": 6.1, "learning_rate": 1.1706577693040993e-06, "loss": 1.3605, "step": 32033 }, { "epoch": 6.1, "learning_rate": 1.170600571973308e-06, "loss": 1.6712, "step": 32034 }, { "epoch": 6.1, "learning_rate": 1.1705433746425167e-06, "loss": 1.586, "step": 32035 }, { "epoch": 6.1, "learning_rate": 1.1704861773117255e-06, "loss": 1.4291, "step": 32036 }, { "epoch": 6.1, "learning_rate": 1.1704289799809343e-06, "loss": 1.601, "step": 32037 }, { "epoch": 6.1, "learning_rate": 1.170371782650143e-06, "loss": 2.3794, "step": 32038 }, { "epoch": 6.1, "learning_rate": 1.1703145853193519e-06, "loss": 2.2274, "step": 32039 }, { "epoch": 6.1, "learning_rate": 1.1702573879885607e-06, "loss": 2.1553, "step": 32040 }, { "epoch": 6.1, "learning_rate": 1.1702001906577692e-06, "loss": 1.3835, "step": 32041 }, { "epoch": 6.1, "learning_rate": 1.170142993326978e-06, "loss": 1.3805, "step": 32042 }, { "epoch": 6.1, "learning_rate": 1.1700857959961868e-06, "loss": 2.7268, "step": 32043 }, { "epoch": 6.1, "learning_rate": 1.1700285986653956e-06, "loss": 2.3248, "step": 32044 }, { "epoch": 6.1, "learning_rate": 1.1699714013346044e-06, "loss": 2.7859, "step": 32045 }, { "epoch": 6.1, "learning_rate": 1.1699142040038132e-06, "loss": 1.5066, "step": 32046 }, { "epoch": 6.1, "learning_rate": 1.169857006673022e-06, "loss": 1.7783, "step": 32047 }, { "epoch": 6.1, "learning_rate": 1.1697998093422306e-06, "loss": 1.2448, "step": 32048 }, { "epoch": 6.1, "learning_rate": 1.1697426120114396e-06, "loss": 1.4938, "step": 32049 }, { "epoch": 6.1, "learning_rate": 1.1696854146806484e-06, "loss": 1.1811, "step": 32050 }, { "epoch": 6.1, "learning_rate": 1.169628217349857e-06, "loss": 1.0036, "step": 32051 }, { "epoch": 6.11, "learning_rate": 1.1695710200190658e-06, "loss": 1.34, "step": 32052 }, { "epoch": 6.11, "learning_rate": 1.1695138226882746e-06, "loss": 0.8756, "step": 32053 }, { "epoch": 6.11, "learning_rate": 1.1694566253574834e-06, "loss": 1.2771, "step": 32054 }, { "epoch": 6.11, "learning_rate": 1.1693994280266921e-06, "loss": 1.1737, "step": 32055 }, { "epoch": 6.11, "learning_rate": 1.169342230695901e-06, "loss": 1.5038, "step": 32056 }, { "epoch": 6.11, "learning_rate": 1.1692850333651097e-06, "loss": 2.1962, "step": 32057 }, { "epoch": 6.11, "learning_rate": 1.1692278360343183e-06, "loss": 0.6552, "step": 32058 }, { "epoch": 6.11, "learning_rate": 1.1691706387035271e-06, "loss": 1.3061, "step": 32059 }, { "epoch": 6.11, "learning_rate": 1.1691134413727361e-06, "loss": 0.6956, "step": 32060 }, { "epoch": 6.11, "learning_rate": 1.1690562440419447e-06, "loss": 2.6322, "step": 32061 }, { "epoch": 6.11, "learning_rate": 1.1689990467111535e-06, "loss": 1.7844, "step": 32062 }, { "epoch": 6.11, "learning_rate": 1.1689418493803623e-06, "loss": 2.9197, "step": 32063 }, { "epoch": 6.11, "learning_rate": 1.168884652049571e-06, "loss": 2.3004, "step": 32064 }, { "epoch": 6.11, "learning_rate": 1.1688274547187799e-06, "loss": 1.9729, "step": 32065 }, { "epoch": 6.11, "learning_rate": 1.1687702573879887e-06, "loss": 1.9645, "step": 32066 }, { "epoch": 6.11, "learning_rate": 1.1687130600571975e-06, "loss": 0.5082, "step": 32067 }, { "epoch": 6.11, "learning_rate": 1.168655862726406e-06, "loss": 1.1654, "step": 32068 }, { "epoch": 6.11, "learning_rate": 1.1685986653956148e-06, "loss": 1.9337, "step": 32069 }, { "epoch": 6.11, "learning_rate": 1.1685414680648236e-06, "loss": 2.5362, "step": 32070 }, { "epoch": 6.11, "learning_rate": 1.1684842707340324e-06, "loss": 2.278, "step": 32071 }, { "epoch": 6.11, "learning_rate": 1.1684270734032412e-06, "loss": 2.6421, "step": 32072 }, { "epoch": 6.11, "learning_rate": 1.16836987607245e-06, "loss": 0.8625, "step": 32073 }, { "epoch": 6.11, "learning_rate": 1.1683126787416588e-06, "loss": 2.558, "step": 32074 }, { "epoch": 6.11, "learning_rate": 1.1682554814108674e-06, "loss": 1.2865, "step": 32075 }, { "epoch": 6.11, "learning_rate": 1.1681982840800762e-06, "loss": 2.5711, "step": 32076 }, { "epoch": 6.11, "learning_rate": 1.1681410867492852e-06, "loss": 1.8757, "step": 32077 }, { "epoch": 6.11, "learning_rate": 1.1680838894184938e-06, "loss": 0.9123, "step": 32078 }, { "epoch": 6.11, "learning_rate": 1.1680266920877026e-06, "loss": 1.0066, "step": 32079 }, { "epoch": 6.11, "learning_rate": 1.1679694947569114e-06, "loss": 1.7764, "step": 32080 }, { "epoch": 6.11, "learning_rate": 1.1679122974261201e-06, "loss": 1.1854, "step": 32081 }, { "epoch": 6.11, "learning_rate": 1.167855100095329e-06, "loss": 2.9887, "step": 32082 }, { "epoch": 6.11, "learning_rate": 1.1677979027645377e-06, "loss": 1.9073, "step": 32083 }, { "epoch": 6.11, "learning_rate": 1.1677407054337465e-06, "loss": 2.729, "step": 32084 }, { "epoch": 6.11, "learning_rate": 1.1676835081029551e-06, "loss": 1.5003, "step": 32085 }, { "epoch": 6.11, "learning_rate": 1.167626310772164e-06, "loss": 2.0592, "step": 32086 }, { "epoch": 6.11, "learning_rate": 1.1675691134413727e-06, "loss": 1.1338, "step": 32087 }, { "epoch": 6.11, "learning_rate": 1.1675119161105817e-06, "loss": 1.2816, "step": 32088 }, { "epoch": 6.11, "learning_rate": 1.1674547187797903e-06, "loss": 2.8987, "step": 32089 }, { "epoch": 6.11, "learning_rate": 1.167397521448999e-06, "loss": 2.3586, "step": 32090 }, { "epoch": 6.11, "learning_rate": 1.1673403241182079e-06, "loss": 0.9633, "step": 32091 }, { "epoch": 6.11, "learning_rate": 1.1672831267874165e-06, "loss": 1.9091, "step": 32092 }, { "epoch": 6.11, "learning_rate": 1.1672259294566255e-06, "loss": 1.7182, "step": 32093 }, { "epoch": 6.11, "learning_rate": 1.1671687321258343e-06, "loss": 1.0719, "step": 32094 }, { "epoch": 6.11, "learning_rate": 1.1671115347950428e-06, "loss": 1.3098, "step": 32095 }, { "epoch": 6.11, "learning_rate": 1.1670543374642516e-06, "loss": 1.6465, "step": 32096 }, { "epoch": 6.11, "learning_rate": 1.1669971401334604e-06, "loss": 1.3651, "step": 32097 }, { "epoch": 6.11, "learning_rate": 1.1669399428026692e-06, "loss": 0.6571, "step": 32098 }, { "epoch": 6.11, "learning_rate": 1.166882745471878e-06, "loss": 2.468, "step": 32099 }, { "epoch": 6.11, "learning_rate": 1.1668255481410868e-06, "loss": 2.3847, "step": 32100 }, { "epoch": 6.11, "learning_rate": 1.1667683508102956e-06, "loss": 0.986, "step": 32101 }, { "epoch": 6.11, "learning_rate": 1.1667111534795042e-06, "loss": 1.3609, "step": 32102 }, { "epoch": 6.11, "learning_rate": 1.166653956148713e-06, "loss": 1.8321, "step": 32103 }, { "epoch": 6.12, "learning_rate": 1.166596758817922e-06, "loss": 1.1242, "step": 32104 }, { "epoch": 6.12, "learning_rate": 1.1665395614871308e-06, "loss": 1.8234, "step": 32105 }, { "epoch": 6.12, "learning_rate": 1.1664823641563394e-06, "loss": 2.2855, "step": 32106 }, { "epoch": 6.12, "learning_rate": 1.1664251668255482e-06, "loss": 2.8898, "step": 32107 }, { "epoch": 6.12, "learning_rate": 1.166367969494757e-06, "loss": 1.7617, "step": 32108 }, { "epoch": 6.12, "learning_rate": 1.1663107721639657e-06, "loss": 2.3377, "step": 32109 }, { "epoch": 6.12, "learning_rate": 1.1662535748331745e-06, "loss": 2.0363, "step": 32110 }, { "epoch": 6.12, "learning_rate": 1.1661963775023833e-06, "loss": 1.351, "step": 32111 }, { "epoch": 6.12, "learning_rate": 1.1661391801715921e-06, "loss": 1.3479, "step": 32112 }, { "epoch": 6.12, "learning_rate": 1.1660819828408007e-06, "loss": 1.8445, "step": 32113 }, { "epoch": 6.12, "learning_rate": 1.1660247855100095e-06, "loss": 1.4439, "step": 32114 }, { "epoch": 6.12, "learning_rate": 1.1659675881792183e-06, "loss": 1.0462, "step": 32115 }, { "epoch": 6.12, "learning_rate": 1.165910390848427e-06, "loss": 1.5918, "step": 32116 }, { "epoch": 6.12, "learning_rate": 1.1658531935176359e-06, "loss": 1.2652, "step": 32117 }, { "epoch": 6.12, "learning_rate": 1.1657959961868447e-06, "loss": 1.2523, "step": 32118 }, { "epoch": 6.12, "learning_rate": 1.1657387988560533e-06, "loss": 0.9689, "step": 32119 }, { "epoch": 6.12, "learning_rate": 1.165681601525262e-06, "loss": 1.124, "step": 32120 }, { "epoch": 6.12, "learning_rate": 1.165624404194471e-06, "loss": 1.2909, "step": 32121 }, { "epoch": 6.12, "learning_rate": 1.1655672068636798e-06, "loss": 0.4365, "step": 32122 }, { "epoch": 6.12, "learning_rate": 1.1655100095328884e-06, "loss": 0.5272, "step": 32123 }, { "epoch": 6.12, "learning_rate": 1.1654528122020972e-06, "loss": 2.7597, "step": 32124 }, { "epoch": 6.12, "learning_rate": 1.165395614871306e-06, "loss": 2.44, "step": 32125 }, { "epoch": 6.12, "learning_rate": 1.1653384175405148e-06, "loss": 2.48, "step": 32126 }, { "epoch": 6.12, "learning_rate": 1.1652812202097236e-06, "loss": 2.0701, "step": 32127 }, { "epoch": 6.12, "learning_rate": 1.1652240228789324e-06, "loss": 1.8693, "step": 32128 }, { "epoch": 6.12, "learning_rate": 1.1651668255481412e-06, "loss": 1.1835, "step": 32129 }, { "epoch": 6.12, "learning_rate": 1.1651096282173498e-06, "loss": 2.9076, "step": 32130 }, { "epoch": 6.12, "learning_rate": 1.1650524308865586e-06, "loss": 2.0292, "step": 32131 }, { "epoch": 6.12, "learning_rate": 1.1649952335557676e-06, "loss": 1.411, "step": 32132 }, { "epoch": 6.12, "learning_rate": 1.1649380362249762e-06, "loss": 3.1318, "step": 32133 }, { "epoch": 6.12, "learning_rate": 1.164880838894185e-06, "loss": 1.6302, "step": 32134 }, { "epoch": 6.12, "learning_rate": 1.1648236415633937e-06, "loss": 1.4369, "step": 32135 }, { "epoch": 6.12, "learning_rate": 1.1647664442326025e-06, "loss": 2.3042, "step": 32136 }, { "epoch": 6.12, "learning_rate": 1.1647092469018113e-06, "loss": 1.3374, "step": 32137 }, { "epoch": 6.12, "learning_rate": 1.1646520495710201e-06, "loss": 2.0997, "step": 32138 }, { "epoch": 6.12, "learning_rate": 1.164594852240229e-06, "loss": 2.8607, "step": 32139 }, { "epoch": 6.12, "learning_rate": 1.1645376549094375e-06, "loss": 1.5939, "step": 32140 }, { "epoch": 6.12, "learning_rate": 1.1644804575786463e-06, "loss": 2.1953, "step": 32141 }, { "epoch": 6.12, "learning_rate": 1.164423260247855e-06, "loss": 2.3017, "step": 32142 }, { "epoch": 6.12, "learning_rate": 1.1643660629170639e-06, "loss": 3.5939, "step": 32143 }, { "epoch": 6.12, "learning_rate": 1.1643088655862727e-06, "loss": 0.9937, "step": 32144 }, { "epoch": 6.12, "learning_rate": 1.1642516682554815e-06, "loss": 2.7585, "step": 32145 }, { "epoch": 6.12, "learning_rate": 1.1641944709246903e-06, "loss": 0.7646, "step": 32146 }, { "epoch": 6.12, "learning_rate": 1.1641372735938988e-06, "loss": 1.4225, "step": 32147 }, { "epoch": 6.12, "learning_rate": 1.1640800762631079e-06, "loss": 1.3005, "step": 32148 }, { "epoch": 6.12, "learning_rate": 1.1640228789323166e-06, "loss": 1.4257, "step": 32149 }, { "epoch": 6.12, "learning_rate": 1.1639656816015252e-06, "loss": 1.9225, "step": 32150 }, { "epoch": 6.12, "learning_rate": 1.163908484270734e-06, "loss": 1.8553, "step": 32151 }, { "epoch": 6.12, "learning_rate": 1.1638512869399428e-06, "loss": 1.4019, "step": 32152 }, { "epoch": 6.12, "learning_rate": 1.1637940896091516e-06, "loss": 1.3021, "step": 32153 }, { "epoch": 6.12, "learning_rate": 1.1637368922783604e-06, "loss": 1.8688, "step": 32154 }, { "epoch": 6.12, "learning_rate": 1.1636796949475692e-06, "loss": 1.6945, "step": 32155 }, { "epoch": 6.12, "learning_rate": 1.163622497616778e-06, "loss": 1.0502, "step": 32156 }, { "epoch": 6.13, "learning_rate": 1.1635653002859866e-06, "loss": 1.5313, "step": 32157 }, { "epoch": 6.13, "learning_rate": 1.1635081029551954e-06, "loss": 2.8665, "step": 32158 }, { "epoch": 6.13, "learning_rate": 1.1634509056244042e-06, "loss": 1.0862, "step": 32159 }, { "epoch": 6.13, "learning_rate": 1.163393708293613e-06, "loss": 0.9832, "step": 32160 }, { "epoch": 6.13, "learning_rate": 1.1633365109628217e-06, "loss": 1.7386, "step": 32161 }, { "epoch": 6.13, "learning_rate": 1.1632793136320305e-06, "loss": 2.2243, "step": 32162 }, { "epoch": 6.13, "learning_rate": 1.1632221163012393e-06, "loss": 1.9113, "step": 32163 }, { "epoch": 6.13, "learning_rate": 1.163164918970448e-06, "loss": 2.3403, "step": 32164 }, { "epoch": 6.13, "learning_rate": 1.163107721639657e-06, "loss": 2.935, "step": 32165 }, { "epoch": 6.13, "learning_rate": 1.1630505243088657e-06, "loss": 0.7288, "step": 32166 }, { "epoch": 6.13, "learning_rate": 1.1629933269780743e-06, "loss": 2.6299, "step": 32167 }, { "epoch": 6.13, "learning_rate": 1.162936129647283e-06, "loss": 0.5643, "step": 32168 }, { "epoch": 6.13, "learning_rate": 1.1628789323164919e-06, "loss": 2.3104, "step": 32169 }, { "epoch": 6.13, "learning_rate": 1.1628217349857007e-06, "loss": 1.9819, "step": 32170 }, { "epoch": 6.13, "learning_rate": 1.1627645376549095e-06, "loss": 1.4686, "step": 32171 }, { "epoch": 6.13, "learning_rate": 1.1627073403241183e-06, "loss": 1.8869, "step": 32172 }, { "epoch": 6.13, "learning_rate": 1.162650142993327e-06, "loss": 2.41, "step": 32173 }, { "epoch": 6.13, "learning_rate": 1.1625929456625356e-06, "loss": 1.7922, "step": 32174 }, { "epoch": 6.13, "learning_rate": 1.1625357483317444e-06, "loss": 1.0761, "step": 32175 }, { "epoch": 6.13, "learning_rate": 1.1624785510009534e-06, "loss": 2.1935, "step": 32176 }, { "epoch": 6.13, "learning_rate": 1.1624213536701622e-06, "loss": 1.1496, "step": 32177 }, { "epoch": 6.13, "learning_rate": 1.1623641563393708e-06, "loss": 2.6971, "step": 32178 }, { "epoch": 6.13, "learning_rate": 1.1623069590085796e-06, "loss": 0.9517, "step": 32179 }, { "epoch": 6.13, "learning_rate": 1.1622497616777884e-06, "loss": 1.1567, "step": 32180 }, { "epoch": 6.13, "learning_rate": 1.1621925643469972e-06, "loss": 1.7316, "step": 32181 }, { "epoch": 6.13, "learning_rate": 1.162135367016206e-06, "loss": 1.93, "step": 32182 }, { "epoch": 6.13, "learning_rate": 1.1620781696854148e-06, "loss": 2.4863, "step": 32183 }, { "epoch": 6.13, "learning_rate": 1.1620209723546234e-06, "loss": 0.8352, "step": 32184 }, { "epoch": 6.13, "learning_rate": 1.1619637750238322e-06, "loss": 1.8458, "step": 32185 }, { "epoch": 6.13, "learning_rate": 1.161906577693041e-06, "loss": 2.7016, "step": 32186 }, { "epoch": 6.13, "learning_rate": 1.1618493803622498e-06, "loss": 0.524, "step": 32187 }, { "epoch": 6.13, "learning_rate": 1.1617921830314585e-06, "loss": 1.5348, "step": 32188 }, { "epoch": 6.13, "learning_rate": 1.1617349857006673e-06, "loss": 0.6418, "step": 32189 }, { "epoch": 6.13, "learning_rate": 1.1616777883698761e-06, "loss": 1.197, "step": 32190 }, { "epoch": 6.13, "learning_rate": 1.1616205910390847e-06, "loss": 2.5082, "step": 32191 }, { "epoch": 6.13, "learning_rate": 1.1615633937082937e-06, "loss": 1.8122, "step": 32192 }, { "epoch": 6.13, "learning_rate": 1.1615061963775025e-06, "loss": 1.6101, "step": 32193 }, { "epoch": 6.13, "learning_rate": 1.1614489990467113e-06, "loss": 1.3385, "step": 32194 }, { "epoch": 6.13, "learning_rate": 1.1613918017159199e-06, "loss": 1.3294, "step": 32195 }, { "epoch": 6.13, "learning_rate": 1.1613346043851287e-06, "loss": 2.3574, "step": 32196 }, { "epoch": 6.13, "learning_rate": 1.1612774070543375e-06, "loss": 1.5996, "step": 32197 }, { "epoch": 6.13, "learning_rate": 1.1612202097235463e-06, "loss": 1.1894, "step": 32198 }, { "epoch": 6.13, "learning_rate": 1.161163012392755e-06, "loss": 2.2365, "step": 32199 }, { "epoch": 6.13, "learning_rate": 1.1611058150619639e-06, "loss": 1.4125, "step": 32200 }, { "epoch": 6.13, "learning_rate": 1.1610486177311727e-06, "loss": 2.3386, "step": 32201 }, { "epoch": 6.13, "learning_rate": 1.1609914204003812e-06, "loss": 0.7549, "step": 32202 }, { "epoch": 6.13, "learning_rate": 1.16093422306959e-06, "loss": 1.4973, "step": 32203 }, { "epoch": 6.13, "learning_rate": 1.160877025738799e-06, "loss": 2.1202, "step": 32204 }, { "epoch": 6.13, "learning_rate": 1.1608198284080076e-06, "loss": 1.7512, "step": 32205 }, { "epoch": 6.13, "learning_rate": 1.1607626310772164e-06, "loss": 1.3502, "step": 32206 }, { "epoch": 6.13, "learning_rate": 1.1607054337464252e-06, "loss": 1.089, "step": 32207 }, { "epoch": 6.13, "learning_rate": 1.1606482364156338e-06, "loss": 0.8363, "step": 32208 }, { "epoch": 6.14, "learning_rate": 1.1605910390848428e-06, "loss": 3.1954, "step": 32209 }, { "epoch": 6.14, "learning_rate": 1.1605338417540516e-06, "loss": 1.8764, "step": 32210 }, { "epoch": 6.14, "learning_rate": 1.1604766444232604e-06, "loss": 1.3653, "step": 32211 }, { "epoch": 6.14, "learning_rate": 1.160419447092469e-06, "loss": 1.3593, "step": 32212 }, { "epoch": 6.14, "learning_rate": 1.1603622497616778e-06, "loss": 2.5209, "step": 32213 }, { "epoch": 6.14, "learning_rate": 1.1603050524308865e-06, "loss": 0.9008, "step": 32214 }, { "epoch": 6.14, "learning_rate": 1.1602478551000953e-06, "loss": 1.996, "step": 32215 }, { "epoch": 6.14, "learning_rate": 1.1601906577693041e-06, "loss": 1.1477, "step": 32216 }, { "epoch": 6.14, "learning_rate": 1.160133460438513e-06, "loss": 1.3764, "step": 32217 }, { "epoch": 6.14, "learning_rate": 1.1600762631077217e-06, "loss": 1.2273, "step": 32218 }, { "epoch": 6.14, "learning_rate": 1.1600190657769303e-06, "loss": 1.3102, "step": 32219 }, { "epoch": 6.14, "learning_rate": 1.1599618684461393e-06, "loss": 2.1343, "step": 32220 }, { "epoch": 6.14, "learning_rate": 1.1599046711153481e-06, "loss": 1.9764, "step": 32221 }, { "epoch": 6.14, "learning_rate": 1.1598474737845567e-06, "loss": 1.382, "step": 32222 }, { "epoch": 6.14, "learning_rate": 1.1597902764537655e-06, "loss": 1.3053, "step": 32223 }, { "epoch": 6.14, "learning_rate": 1.1597330791229743e-06, "loss": 1.912, "step": 32224 }, { "epoch": 6.14, "learning_rate": 1.159675881792183e-06, "loss": 3.0937, "step": 32225 }, { "epoch": 6.14, "learning_rate": 1.1596186844613919e-06, "loss": 1.7747, "step": 32226 }, { "epoch": 6.14, "learning_rate": 1.1595614871306007e-06, "loss": 1.1445, "step": 32227 }, { "epoch": 6.14, "learning_rate": 1.1595042897998095e-06, "loss": 1.1951, "step": 32228 }, { "epoch": 6.14, "learning_rate": 1.159447092469018e-06, "loss": 2.4411, "step": 32229 }, { "epoch": 6.14, "learning_rate": 1.1593898951382268e-06, "loss": 1.5759, "step": 32230 }, { "epoch": 6.14, "learning_rate": 1.1593326978074356e-06, "loss": 1.9049, "step": 32231 }, { "epoch": 6.14, "learning_rate": 1.1592755004766444e-06, "loss": 1.8697, "step": 32232 }, { "epoch": 6.14, "learning_rate": 1.1592183031458532e-06, "loss": 3.2918, "step": 32233 }, { "epoch": 6.14, "learning_rate": 1.159161105815062e-06, "loss": 2.8454, "step": 32234 }, { "epoch": 6.14, "learning_rate": 1.1591039084842708e-06, "loss": 0.6082, "step": 32235 }, { "epoch": 6.14, "learning_rate": 1.1590467111534796e-06, "loss": 1.5098, "step": 32236 }, { "epoch": 6.14, "learning_rate": 1.1589895138226884e-06, "loss": 1.5659, "step": 32237 }, { "epoch": 6.14, "learning_rate": 1.1589323164918972e-06, "loss": 1.3913, "step": 32238 }, { "epoch": 6.14, "learning_rate": 1.1588751191611058e-06, "loss": 1.9412, "step": 32239 }, { "epoch": 6.14, "learning_rate": 1.1588179218303146e-06, "loss": 1.9934, "step": 32240 }, { "epoch": 6.14, "learning_rate": 1.1587607244995233e-06, "loss": 2.5648, "step": 32241 }, { "epoch": 6.14, "learning_rate": 1.1587035271687321e-06, "loss": 1.9385, "step": 32242 }, { "epoch": 6.14, "learning_rate": 1.158646329837941e-06, "loss": 1.6706, "step": 32243 }, { "epoch": 6.14, "learning_rate": 1.1585891325071497e-06, "loss": 2.2931, "step": 32244 }, { "epoch": 6.14, "learning_rate": 1.1585319351763585e-06, "loss": 2.926, "step": 32245 }, { "epoch": 6.14, "learning_rate": 1.158474737845567e-06, "loss": 1.8736, "step": 32246 }, { "epoch": 6.14, "learning_rate": 1.158417540514776e-06, "loss": 1.6321, "step": 32247 }, { "epoch": 6.14, "learning_rate": 1.158360343183985e-06, "loss": 1.8807, "step": 32248 }, { "epoch": 6.14, "learning_rate": 1.1583031458531935e-06, "loss": 1.6994, "step": 32249 }, { "epoch": 6.14, "learning_rate": 1.1582459485224023e-06, "loss": 1.9009, "step": 32250 }, { "epoch": 6.14, "learning_rate": 1.158188751191611e-06, "loss": 2.8736, "step": 32251 }, { "epoch": 6.14, "learning_rate": 1.1581315538608199e-06, "loss": 1.7301, "step": 32252 }, { "epoch": 6.14, "learning_rate": 1.1580743565300287e-06, "loss": 1.4761, "step": 32253 }, { "epoch": 6.14, "learning_rate": 1.1580171591992375e-06, "loss": 2.065, "step": 32254 }, { "epoch": 6.14, "learning_rate": 1.1579599618684462e-06, "loss": 1.0454, "step": 32255 }, { "epoch": 6.14, "learning_rate": 1.1579027645376548e-06, "loss": 1.853, "step": 32256 }, { "epoch": 6.14, "learning_rate": 1.1578455672068636e-06, "loss": 1.7967, "step": 32257 }, { "epoch": 6.14, "learning_rate": 1.1577883698760724e-06, "loss": 1.6337, "step": 32258 }, { "epoch": 6.14, "learning_rate": 1.1577311725452814e-06, "loss": 2.2097, "step": 32259 }, { "epoch": 6.14, "learning_rate": 1.15767397521449e-06, "loss": 1.5831, "step": 32260 }, { "epoch": 6.14, "learning_rate": 1.1576167778836988e-06, "loss": 0.7861, "step": 32261 }, { "epoch": 6.15, "learning_rate": 1.1575595805529076e-06, "loss": 1.4999, "step": 32262 }, { "epoch": 6.15, "learning_rate": 1.1575023832221162e-06, "loss": 1.3321, "step": 32263 }, { "epoch": 6.15, "learning_rate": 1.1574451858913252e-06, "loss": 1.8486, "step": 32264 }, { "epoch": 6.15, "learning_rate": 1.157387988560534e-06, "loss": 2.8995, "step": 32265 }, { "epoch": 6.15, "learning_rate": 1.1573307912297428e-06, "loss": 3.3116, "step": 32266 }, { "epoch": 6.15, "learning_rate": 1.1572735938989514e-06, "loss": 1.4468, "step": 32267 }, { "epoch": 6.15, "learning_rate": 1.1572163965681601e-06, "loss": 0.8566, "step": 32268 }, { "epoch": 6.15, "learning_rate": 1.157159199237369e-06, "loss": 1.2819, "step": 32269 }, { "epoch": 6.15, "learning_rate": 1.1571020019065777e-06, "loss": 2.3335, "step": 32270 }, { "epoch": 6.15, "learning_rate": 1.1570448045757865e-06, "loss": 0.6766, "step": 32271 }, { "epoch": 6.15, "learning_rate": 1.1569876072449953e-06, "loss": 2.3746, "step": 32272 }, { "epoch": 6.15, "learning_rate": 1.156930409914204e-06, "loss": 2.3432, "step": 32273 }, { "epoch": 6.15, "learning_rate": 1.1568732125834127e-06, "loss": 2.8513, "step": 32274 }, { "epoch": 6.15, "learning_rate": 1.1568160152526215e-06, "loss": 2.4922, "step": 32275 }, { "epoch": 6.15, "learning_rate": 1.1567588179218305e-06, "loss": 1.7147, "step": 32276 }, { "epoch": 6.15, "learning_rate": 1.156701620591039e-06, "loss": 1.6005, "step": 32277 }, { "epoch": 6.15, "learning_rate": 1.1566444232602479e-06, "loss": 1.477, "step": 32278 }, { "epoch": 6.15, "learning_rate": 1.1565872259294567e-06, "loss": 2.7554, "step": 32279 }, { "epoch": 6.15, "learning_rate": 1.1565300285986655e-06, "loss": 2.1701, "step": 32280 }, { "epoch": 6.15, "learning_rate": 1.1564728312678743e-06, "loss": 2.0635, "step": 32281 }, { "epoch": 6.15, "learning_rate": 1.156415633937083e-06, "loss": 2.0217, "step": 32282 }, { "epoch": 6.15, "learning_rate": 1.1563584366062918e-06, "loss": 1.2116, "step": 32283 }, { "epoch": 6.15, "learning_rate": 1.1563012392755004e-06, "loss": 1.21, "step": 32284 }, { "epoch": 6.15, "learning_rate": 1.1562440419447092e-06, "loss": 1.6655, "step": 32285 }, { "epoch": 6.15, "learning_rate": 1.156186844613918e-06, "loss": 1.3018, "step": 32286 }, { "epoch": 6.15, "learning_rate": 1.1561296472831268e-06, "loss": 2.9312, "step": 32287 }, { "epoch": 6.15, "learning_rate": 1.1560724499523356e-06, "loss": 1.7998, "step": 32288 }, { "epoch": 6.15, "learning_rate": 1.1560152526215444e-06, "loss": 1.4107, "step": 32289 }, { "epoch": 6.15, "learning_rate": 1.155958055290753e-06, "loss": 2.111, "step": 32290 }, { "epoch": 6.15, "learning_rate": 1.1559008579599618e-06, "loss": 2.1601, "step": 32291 }, { "epoch": 6.15, "learning_rate": 1.1558436606291708e-06, "loss": 2.8875, "step": 32292 }, { "epoch": 6.15, "learning_rate": 1.1557864632983796e-06, "loss": 1.2053, "step": 32293 }, { "epoch": 6.15, "learning_rate": 1.1557292659675881e-06, "loss": 2.4018, "step": 32294 }, { "epoch": 6.15, "learning_rate": 1.155672068636797e-06, "loss": 1.8521, "step": 32295 }, { "epoch": 6.15, "learning_rate": 1.1556148713060057e-06, "loss": 2.807, "step": 32296 }, { "epoch": 6.15, "learning_rate": 1.1555576739752145e-06, "loss": 2.5977, "step": 32297 }, { "epoch": 6.15, "learning_rate": 1.1555004766444233e-06, "loss": 1.1829, "step": 32298 }, { "epoch": 6.15, "learning_rate": 1.1554432793136321e-06, "loss": 2.1239, "step": 32299 }, { "epoch": 6.15, "learning_rate": 1.155386081982841e-06, "loss": 1.5859, "step": 32300 }, { "epoch": 6.15, "learning_rate": 1.1553288846520495e-06, "loss": 2.0899, "step": 32301 }, { "epoch": 6.15, "learning_rate": 1.1552716873212583e-06, "loss": 1.4564, "step": 32302 }, { "epoch": 6.15, "learning_rate": 1.1552144899904673e-06, "loss": 2.1029, "step": 32303 }, { "epoch": 6.15, "learning_rate": 1.1551572926596759e-06, "loss": 1.0258, "step": 32304 }, { "epoch": 6.15, "learning_rate": 1.1551000953288847e-06, "loss": 2.6512, "step": 32305 }, { "epoch": 6.15, "learning_rate": 1.1550428979980935e-06, "loss": 1.4767, "step": 32306 }, { "epoch": 6.15, "learning_rate": 1.1549857006673023e-06, "loss": 0.8819, "step": 32307 }, { "epoch": 6.15, "learning_rate": 1.154928503336511e-06, "loss": 1.5628, "step": 32308 }, { "epoch": 6.15, "learning_rate": 1.1548713060057198e-06, "loss": 1.4594, "step": 32309 }, { "epoch": 6.15, "learning_rate": 1.1548141086749286e-06, "loss": 2.2319, "step": 32310 }, { "epoch": 6.15, "learning_rate": 1.1547569113441372e-06, "loss": 3.6101, "step": 32311 }, { "epoch": 6.15, "learning_rate": 1.154699714013346e-06, "loss": 2.0496, "step": 32312 }, { "epoch": 6.15, "learning_rate": 1.1546425166825548e-06, "loss": 0.9284, "step": 32313 }, { "epoch": 6.16, "learning_rate": 1.1545853193517636e-06, "loss": 2.0221, "step": 32314 }, { "epoch": 6.16, "learning_rate": 1.1545281220209724e-06, "loss": 1.2765, "step": 32315 }, { "epoch": 6.16, "learning_rate": 1.1544709246901812e-06, "loss": 1.5328, "step": 32316 }, { "epoch": 6.16, "learning_rate": 1.15441372735939e-06, "loss": 1.1609, "step": 32317 }, { "epoch": 6.16, "learning_rate": 1.1543565300285986e-06, "loss": 2.0336, "step": 32318 }, { "epoch": 6.16, "learning_rate": 1.1542993326978074e-06, "loss": 1.2061, "step": 32319 }, { "epoch": 6.16, "learning_rate": 1.1542421353670164e-06, "loss": 1.6132, "step": 32320 }, { "epoch": 6.16, "learning_rate": 1.154184938036225e-06, "loss": 1.0642, "step": 32321 }, { "epoch": 6.16, "learning_rate": 1.1541277407054337e-06, "loss": 2.1864, "step": 32322 }, { "epoch": 6.16, "learning_rate": 1.1540705433746425e-06, "loss": 1.2913, "step": 32323 }, { "epoch": 6.16, "learning_rate": 1.1540133460438513e-06, "loss": 0.6737, "step": 32324 }, { "epoch": 6.16, "learning_rate": 1.1539561487130601e-06, "loss": 1.3017, "step": 32325 }, { "epoch": 6.16, "learning_rate": 1.153898951382269e-06, "loss": 2.5343, "step": 32326 }, { "epoch": 6.16, "learning_rate": 1.1538417540514777e-06, "loss": 1.6047, "step": 32327 }, { "epoch": 6.16, "learning_rate": 1.1537845567206863e-06, "loss": 2.2392, "step": 32328 }, { "epoch": 6.16, "learning_rate": 1.153727359389895e-06, "loss": 3.5419, "step": 32329 }, { "epoch": 6.16, "learning_rate": 1.1536701620591039e-06, "loss": 1.4086, "step": 32330 }, { "epoch": 6.16, "learning_rate": 1.1536129647283129e-06, "loss": 1.8349, "step": 32331 }, { "epoch": 6.16, "learning_rate": 1.1535557673975215e-06, "loss": 1.5873, "step": 32332 }, { "epoch": 6.16, "learning_rate": 1.1534985700667303e-06, "loss": 2.3575, "step": 32333 }, { "epoch": 6.16, "learning_rate": 1.153441372735939e-06, "loss": 0.942, "step": 32334 }, { "epoch": 6.16, "learning_rate": 1.1533841754051476e-06, "loss": 0.9032, "step": 32335 }, { "epoch": 6.16, "learning_rate": 1.1533269780743566e-06, "loss": 0.519, "step": 32336 }, { "epoch": 6.16, "learning_rate": 1.1532697807435654e-06, "loss": 1.4279, "step": 32337 }, { "epoch": 6.16, "learning_rate": 1.153212583412774e-06, "loss": 4.1944, "step": 32338 }, { "epoch": 6.16, "learning_rate": 1.1531553860819828e-06, "loss": 1.2771, "step": 32339 }, { "epoch": 6.16, "learning_rate": 1.1530981887511916e-06, "loss": 1.2292, "step": 32340 }, { "epoch": 6.16, "learning_rate": 1.1530409914204004e-06, "loss": 1.72, "step": 32341 }, { "epoch": 6.16, "learning_rate": 1.1529837940896092e-06, "loss": 2.3173, "step": 32342 }, { "epoch": 6.16, "learning_rate": 1.152926596758818e-06, "loss": 1.8993, "step": 32343 }, { "epoch": 6.16, "learning_rate": 1.1528693994280268e-06, "loss": 1.9477, "step": 32344 }, { "epoch": 6.16, "learning_rate": 1.1528122020972354e-06, "loss": 1.3816, "step": 32345 }, { "epoch": 6.16, "learning_rate": 1.1527550047664442e-06, "loss": 1.1139, "step": 32346 }, { "epoch": 6.16, "learning_rate": 1.1526978074356532e-06, "loss": 0.4532, "step": 32347 }, { "epoch": 6.16, "learning_rate": 1.152640610104862e-06, "loss": 2.9285, "step": 32348 }, { "epoch": 6.16, "learning_rate": 1.1525834127740705e-06, "loss": 2.1696, "step": 32349 }, { "epoch": 6.16, "learning_rate": 1.1525262154432793e-06, "loss": 0.5889, "step": 32350 }, { "epoch": 6.16, "learning_rate": 1.1524690181124881e-06, "loss": 3.0083, "step": 32351 }, { "epoch": 6.16, "learning_rate": 1.152411820781697e-06, "loss": 1.8737, "step": 32352 }, { "epoch": 6.16, "learning_rate": 1.1523546234509057e-06, "loss": 2.239, "step": 32353 }, { "epoch": 6.16, "learning_rate": 1.1522974261201145e-06, "loss": 2.3561, "step": 32354 }, { "epoch": 6.16, "learning_rate": 1.152240228789323e-06, "loss": 0.7406, "step": 32355 }, { "epoch": 6.16, "learning_rate": 1.1521830314585319e-06, "loss": 1.4839, "step": 32356 }, { "epoch": 6.16, "learning_rate": 1.1521258341277407e-06, "loss": 1.2061, "step": 32357 }, { "epoch": 6.16, "learning_rate": 1.1520686367969495e-06, "loss": 2.7732, "step": 32358 }, { "epoch": 6.16, "learning_rate": 1.1520114394661583e-06, "loss": 1.3864, "step": 32359 }, { "epoch": 6.16, "learning_rate": 1.151954242135367e-06, "loss": 2.147, "step": 32360 }, { "epoch": 6.16, "learning_rate": 1.1518970448045759e-06, "loss": 1.686, "step": 32361 }, { "epoch": 6.16, "learning_rate": 1.1518398474737844e-06, "loss": 1.1016, "step": 32362 }, { "epoch": 6.16, "learning_rate": 1.1517826501429932e-06, "loss": 1.4299, "step": 32363 }, { "epoch": 6.16, "learning_rate": 1.1517254528122022e-06, "loss": 2.5627, "step": 32364 }, { "epoch": 6.16, "learning_rate": 1.151668255481411e-06, "loss": 1.6206, "step": 32365 }, { "epoch": 6.16, "learning_rate": 1.1516110581506196e-06, "loss": 1.394, "step": 32366 }, { "epoch": 6.17, "learning_rate": 1.1515538608198284e-06, "loss": 1.7301, "step": 32367 }, { "epoch": 6.17, "learning_rate": 1.1514966634890372e-06, "loss": 2.2131, "step": 32368 }, { "epoch": 6.17, "learning_rate": 1.151439466158246e-06, "loss": 1.3124, "step": 32369 }, { "epoch": 6.17, "learning_rate": 1.1513822688274548e-06, "loss": 1.7929, "step": 32370 }, { "epoch": 6.17, "learning_rate": 1.1513250714966636e-06, "loss": 0.872, "step": 32371 }, { "epoch": 6.17, "learning_rate": 1.1512678741658724e-06, "loss": 3.0522, "step": 32372 }, { "epoch": 6.17, "learning_rate": 1.151210676835081e-06, "loss": 1.5636, "step": 32373 }, { "epoch": 6.17, "learning_rate": 1.1511534795042897e-06, "loss": 2.8028, "step": 32374 }, { "epoch": 6.17, "learning_rate": 1.1510962821734988e-06, "loss": 2.4281, "step": 32375 }, { "epoch": 6.17, "learning_rate": 1.1510390848427073e-06, "loss": 1.9192, "step": 32376 }, { "epoch": 6.17, "learning_rate": 1.1509818875119161e-06, "loss": 2.0002, "step": 32377 }, { "epoch": 6.17, "learning_rate": 1.150924690181125e-06, "loss": 2.2234, "step": 32378 }, { "epoch": 6.17, "learning_rate": 1.1508674928503335e-06, "loss": 1.8624, "step": 32379 }, { "epoch": 6.17, "learning_rate": 1.1508102955195425e-06, "loss": 0.6511, "step": 32380 }, { "epoch": 6.17, "learning_rate": 1.1507530981887513e-06, "loss": 2.8034, "step": 32381 }, { "epoch": 6.17, "learning_rate": 1.15069590085796e-06, "loss": 2.7514, "step": 32382 }, { "epoch": 6.17, "learning_rate": 1.1506387035271687e-06, "loss": 2.1215, "step": 32383 }, { "epoch": 6.17, "learning_rate": 1.1505815061963775e-06, "loss": 1.6124, "step": 32384 }, { "epoch": 6.17, "learning_rate": 1.1505243088655863e-06, "loss": 2.3925, "step": 32385 }, { "epoch": 6.17, "learning_rate": 1.150467111534795e-06, "loss": 1.6659, "step": 32386 }, { "epoch": 6.17, "learning_rate": 1.1504099142040039e-06, "loss": 1.8926, "step": 32387 }, { "epoch": 6.17, "learning_rate": 1.1503527168732127e-06, "loss": 2.1086, "step": 32388 }, { "epoch": 6.17, "learning_rate": 1.1502955195424214e-06, "loss": 1.3217, "step": 32389 }, { "epoch": 6.17, "learning_rate": 1.15023832221163e-06, "loss": 2.2493, "step": 32390 }, { "epoch": 6.17, "learning_rate": 1.150181124880839e-06, "loss": 1.5999, "step": 32391 }, { "epoch": 6.17, "learning_rate": 1.1501239275500478e-06, "loss": 1.0885, "step": 32392 }, { "epoch": 6.17, "learning_rate": 1.1500667302192564e-06, "loss": 1.467, "step": 32393 }, { "epoch": 6.17, "learning_rate": 1.1500095328884652e-06, "loss": 2.0206, "step": 32394 }, { "epoch": 6.17, "learning_rate": 1.149952335557674e-06, "loss": 0.8908, "step": 32395 }, { "epoch": 6.17, "learning_rate": 1.1498951382268828e-06, "loss": 1.6877, "step": 32396 }, { "epoch": 6.17, "learning_rate": 1.1498379408960916e-06, "loss": 1.7551, "step": 32397 }, { "epoch": 6.17, "learning_rate": 1.1497807435653004e-06, "loss": 1.1537, "step": 32398 }, { "epoch": 6.17, "learning_rate": 1.1497235462345092e-06, "loss": 2.7661, "step": 32399 }, { "epoch": 6.17, "learning_rate": 1.1496663489037178e-06, "loss": 1.9726, "step": 32400 }, { "epoch": 6.17, "learning_rate": 1.1496091515729265e-06, "loss": 1.9915, "step": 32401 }, { "epoch": 6.17, "learning_rate": 1.1495519542421353e-06, "loss": 1.0593, "step": 32402 }, { "epoch": 6.17, "learning_rate": 1.1494947569113441e-06, "loss": 0.8695, "step": 32403 }, { "epoch": 6.17, "learning_rate": 1.149437559580553e-06, "loss": 1.1717, "step": 32404 }, { "epoch": 6.17, "learning_rate": 1.1493803622497617e-06, "loss": 1.4419, "step": 32405 }, { "epoch": 6.17, "learning_rate": 1.1493231649189705e-06, "loss": 0.9769, "step": 32406 }, { "epoch": 6.17, "learning_rate": 1.149265967588179e-06, "loss": 3.3957, "step": 32407 }, { "epoch": 6.17, "learning_rate": 1.149208770257388e-06, "loss": 0.8298, "step": 32408 }, { "epoch": 6.17, "learning_rate": 1.149151572926597e-06, "loss": 1.9531, "step": 32409 }, { "epoch": 6.17, "learning_rate": 1.1490943755958055e-06, "loss": 1.8528, "step": 32410 }, { "epoch": 6.17, "learning_rate": 1.1490371782650143e-06, "loss": 2.5398, "step": 32411 }, { "epoch": 6.17, "learning_rate": 1.148979980934223e-06, "loss": 2.149, "step": 32412 }, { "epoch": 6.17, "learning_rate": 1.1489227836034319e-06, "loss": 1.4829, "step": 32413 }, { "epoch": 6.17, "learning_rate": 1.1488655862726407e-06, "loss": 1.5483, "step": 32414 }, { "epoch": 6.17, "learning_rate": 1.1488083889418494e-06, "loss": 2.3576, "step": 32415 }, { "epoch": 6.17, "learning_rate": 1.1487511916110582e-06, "loss": 2.4733, "step": 32416 }, { "epoch": 6.17, "learning_rate": 1.1486939942802668e-06, "loss": 0.7194, "step": 32417 }, { "epoch": 6.17, "learning_rate": 1.1486367969494756e-06, "loss": 1.5691, "step": 32418 }, { "epoch": 6.18, "learning_rate": 1.1485795996186846e-06, "loss": 2.6318, "step": 32419 }, { "epoch": 6.18, "learning_rate": 1.1485224022878932e-06, "loss": 1.1307, "step": 32420 }, { "epoch": 6.18, "learning_rate": 1.148465204957102e-06, "loss": 0.4955, "step": 32421 }, { "epoch": 6.18, "learning_rate": 1.1484080076263108e-06, "loss": 2.5804, "step": 32422 }, { "epoch": 6.18, "learning_rate": 1.1483508102955196e-06, "loss": 2.0483, "step": 32423 }, { "epoch": 6.18, "learning_rate": 1.1482936129647284e-06, "loss": 1.1314, "step": 32424 }, { "epoch": 6.18, "learning_rate": 1.1482364156339372e-06, "loss": 2.4961, "step": 32425 }, { "epoch": 6.18, "learning_rate": 1.148179218303146e-06, "loss": 1.4737, "step": 32426 }, { "epoch": 6.18, "learning_rate": 1.1481220209723546e-06, "loss": 2.5317, "step": 32427 }, { "epoch": 6.18, "learning_rate": 1.1480648236415633e-06, "loss": 2.721, "step": 32428 }, { "epoch": 6.18, "learning_rate": 1.1480076263107721e-06, "loss": 2.625, "step": 32429 }, { "epoch": 6.18, "learning_rate": 1.147950428979981e-06, "loss": 1.2523, "step": 32430 }, { "epoch": 6.18, "learning_rate": 1.1478932316491897e-06, "loss": 2.3694, "step": 32431 }, { "epoch": 6.18, "learning_rate": 1.1478360343183985e-06, "loss": 2.0555, "step": 32432 }, { "epoch": 6.18, "learning_rate": 1.1477788369876073e-06, "loss": 1.1279, "step": 32433 }, { "epoch": 6.18, "learning_rate": 1.147721639656816e-06, "loss": 1.448, "step": 32434 }, { "epoch": 6.18, "learning_rate": 1.147664442326025e-06, "loss": 3.0151, "step": 32435 }, { "epoch": 6.18, "learning_rate": 1.1476072449952337e-06, "loss": 2.4621, "step": 32436 }, { "epoch": 6.18, "learning_rate": 1.1475500476644425e-06, "loss": 0.8434, "step": 32437 }, { "epoch": 6.18, "learning_rate": 1.147492850333651e-06, "loss": 1.6006, "step": 32438 }, { "epoch": 6.18, "learning_rate": 1.1474356530028599e-06, "loss": 1.625, "step": 32439 }, { "epoch": 6.18, "learning_rate": 1.1473784556720687e-06, "loss": 2.8468, "step": 32440 }, { "epoch": 6.18, "learning_rate": 1.1473212583412775e-06, "loss": 2.0409, "step": 32441 }, { "epoch": 6.18, "learning_rate": 1.1472640610104862e-06, "loss": 2.151, "step": 32442 }, { "epoch": 6.18, "learning_rate": 1.147206863679695e-06, "loss": 0.8704, "step": 32443 }, { "epoch": 6.18, "learning_rate": 1.1471496663489036e-06, "loss": 0.3895, "step": 32444 }, { "epoch": 6.18, "learning_rate": 1.1470924690181124e-06, "loss": 2.244, "step": 32445 }, { "epoch": 6.18, "learning_rate": 1.1470352716873212e-06, "loss": 1.7452, "step": 32446 }, { "epoch": 6.18, "learning_rate": 1.1469780743565302e-06, "loss": 1.7399, "step": 32447 }, { "epoch": 6.18, "learning_rate": 1.1469208770257388e-06, "loss": 1.3011, "step": 32448 }, { "epoch": 6.18, "learning_rate": 1.1468636796949476e-06, "loss": 1.5822, "step": 32449 }, { "epoch": 6.18, "learning_rate": 1.1468064823641564e-06, "loss": 2.7082, "step": 32450 }, { "epoch": 6.18, "learning_rate": 1.146749285033365e-06, "loss": 0.7993, "step": 32451 }, { "epoch": 6.18, "learning_rate": 1.146692087702574e-06, "loss": 2.2042, "step": 32452 }, { "epoch": 6.18, "learning_rate": 1.1466348903717828e-06, "loss": 1.2649, "step": 32453 }, { "epoch": 6.18, "learning_rate": 1.1465776930409916e-06, "loss": 1.3656, "step": 32454 }, { "epoch": 6.18, "learning_rate": 1.1465204957102001e-06, "loss": 3.2803, "step": 32455 }, { "epoch": 6.18, "learning_rate": 1.146463298379409e-06, "loss": 0.5453, "step": 32456 }, { "epoch": 6.18, "learning_rate": 1.1464061010486177e-06, "loss": 1.8323, "step": 32457 }, { "epoch": 6.18, "learning_rate": 1.1463489037178265e-06, "loss": 2.2882, "step": 32458 }, { "epoch": 6.18, "learning_rate": 1.1462917063870353e-06, "loss": 1.6012, "step": 32459 }, { "epoch": 6.18, "learning_rate": 1.1462345090562441e-06, "loss": 2.2271, "step": 32460 }, { "epoch": 6.18, "learning_rate": 1.1461773117254527e-06, "loss": 1.8709, "step": 32461 }, { "epoch": 6.18, "learning_rate": 1.1461201143946615e-06, "loss": 1.5776, "step": 32462 }, { "epoch": 6.18, "learning_rate": 1.1460629170638705e-06, "loss": 2.4509, "step": 32463 }, { "epoch": 6.18, "learning_rate": 1.1460057197330793e-06, "loss": 1.5657, "step": 32464 }, { "epoch": 6.18, "learning_rate": 1.1459485224022879e-06, "loss": 1.0098, "step": 32465 }, { "epoch": 6.18, "learning_rate": 1.1458913250714967e-06, "loss": 2.2413, "step": 32466 }, { "epoch": 6.18, "learning_rate": 1.1458341277407055e-06, "loss": 1.0638, "step": 32467 }, { "epoch": 6.18, "learning_rate": 1.1457769304099143e-06, "loss": 1.6706, "step": 32468 }, { "epoch": 6.18, "learning_rate": 1.145719733079123e-06, "loss": 1.4264, "step": 32469 }, { "epoch": 6.18, "learning_rate": 1.1456625357483318e-06, "loss": 1.1514, "step": 32470 }, { "epoch": 6.18, "learning_rate": 1.1456053384175406e-06, "loss": 1.4156, "step": 32471 }, { "epoch": 6.19, "learning_rate": 1.1455481410867492e-06, "loss": 0.6026, "step": 32472 }, { "epoch": 6.19, "learning_rate": 1.145490943755958e-06, "loss": 1.4648, "step": 32473 }, { "epoch": 6.19, "learning_rate": 1.1454337464251668e-06, "loss": 1.9865, "step": 32474 }, { "epoch": 6.19, "learning_rate": 1.1453765490943756e-06, "loss": 1.8913, "step": 32475 }, { "epoch": 6.19, "learning_rate": 1.1453193517635844e-06, "loss": 1.3789, "step": 32476 }, { "epoch": 6.19, "learning_rate": 1.1452621544327932e-06, "loss": 1.3389, "step": 32477 }, { "epoch": 6.19, "learning_rate": 1.145204957102002e-06, "loss": 2.6727, "step": 32478 }, { "epoch": 6.19, "learning_rate": 1.1451477597712108e-06, "loss": 1.8277, "step": 32479 }, { "epoch": 6.19, "learning_rate": 1.1450905624404196e-06, "loss": 2.7645, "step": 32480 }, { "epoch": 6.19, "learning_rate": 1.1450333651096284e-06, "loss": 2.0278, "step": 32481 }, { "epoch": 6.19, "learning_rate": 1.144976167778837e-06, "loss": 1.9979, "step": 32482 }, { "epoch": 6.19, "learning_rate": 1.1449189704480457e-06, "loss": 1.6231, "step": 32483 }, { "epoch": 6.19, "learning_rate": 1.1448617731172545e-06, "loss": 1.6866, "step": 32484 }, { "epoch": 6.19, "learning_rate": 1.1448045757864633e-06, "loss": 1.7047, "step": 32485 }, { "epoch": 6.19, "learning_rate": 1.1447473784556721e-06, "loss": 1.6634, "step": 32486 }, { "epoch": 6.19, "learning_rate": 1.144690181124881e-06, "loss": 1.4635, "step": 32487 }, { "epoch": 6.19, "learning_rate": 1.1446329837940897e-06, "loss": 0.8233, "step": 32488 }, { "epoch": 6.19, "learning_rate": 1.1445757864632983e-06, "loss": 2.7408, "step": 32489 }, { "epoch": 6.19, "learning_rate": 1.144518589132507e-06, "loss": 0.9571, "step": 32490 }, { "epoch": 6.19, "learning_rate": 1.144461391801716e-06, "loss": 2.691, "step": 32491 }, { "epoch": 6.19, "learning_rate": 1.1444041944709247e-06, "loss": 3.0305, "step": 32492 }, { "epoch": 6.19, "learning_rate": 1.1443469971401335e-06, "loss": 2.7922, "step": 32493 }, { "epoch": 6.19, "learning_rate": 1.1442897998093423e-06, "loss": 2.1265, "step": 32494 }, { "epoch": 6.19, "learning_rate": 1.144232602478551e-06, "loss": 2.7063, "step": 32495 }, { "epoch": 6.19, "learning_rate": 1.1441754051477598e-06, "loss": 0.5915, "step": 32496 }, { "epoch": 6.19, "learning_rate": 1.1441182078169686e-06, "loss": 2.4554, "step": 32497 }, { "epoch": 6.19, "learning_rate": 1.1440610104861774e-06, "loss": 1.853, "step": 32498 }, { "epoch": 6.19, "learning_rate": 1.144003813155386e-06, "loss": 3.189, "step": 32499 }, { "epoch": 6.19, "learning_rate": 1.1439466158245948e-06, "loss": 1.9807, "step": 32500 }, { "epoch": 6.19, "learning_rate": 1.1438894184938036e-06, "loss": 3.19, "step": 32501 }, { "epoch": 6.19, "learning_rate": 1.1438322211630126e-06, "loss": 0.5283, "step": 32502 }, { "epoch": 6.19, "learning_rate": 1.1437750238322212e-06, "loss": 2.6148, "step": 32503 }, { "epoch": 6.19, "learning_rate": 1.14371782650143e-06, "loss": 0.5881, "step": 32504 }, { "epoch": 6.19, "learning_rate": 1.1436606291706388e-06, "loss": 1.9093, "step": 32505 }, { "epoch": 6.19, "learning_rate": 1.1436034318398474e-06, "loss": 2.5571, "step": 32506 }, { "epoch": 6.19, "learning_rate": 1.1435462345090564e-06, "loss": 3.2928, "step": 32507 }, { "epoch": 6.19, "learning_rate": 1.1434890371782652e-06, "loss": 2.1072, "step": 32508 }, { "epoch": 6.19, "learning_rate": 1.1434318398474737e-06, "loss": 2.9943, "step": 32509 }, { "epoch": 6.19, "learning_rate": 1.1433746425166825e-06, "loss": 1.3968, "step": 32510 }, { "epoch": 6.19, "learning_rate": 1.1433174451858913e-06, "loss": 1.2173, "step": 32511 }, { "epoch": 6.19, "learning_rate": 1.1432602478551001e-06, "loss": 1.6056, "step": 32512 }, { "epoch": 6.19, "learning_rate": 1.143203050524309e-06, "loss": 1.612, "step": 32513 }, { "epoch": 6.19, "learning_rate": 1.1431458531935177e-06, "loss": 1.7742, "step": 32514 }, { "epoch": 6.19, "learning_rate": 1.1430886558627265e-06, "loss": 1.2423, "step": 32515 }, { "epoch": 6.19, "learning_rate": 1.143031458531935e-06, "loss": 1.0608, "step": 32516 }, { "epoch": 6.19, "learning_rate": 1.1429742612011439e-06, "loss": 3.217, "step": 32517 }, { "epoch": 6.19, "learning_rate": 1.1429170638703527e-06, "loss": 2.5673, "step": 32518 }, { "epoch": 6.19, "learning_rate": 1.1428598665395617e-06, "loss": 1.6167, "step": 32519 }, { "epoch": 6.19, "learning_rate": 1.1428026692087703e-06, "loss": 2.3327, "step": 32520 }, { "epoch": 6.19, "learning_rate": 1.142745471877979e-06, "loss": 1.603, "step": 32521 }, { "epoch": 6.19, "learning_rate": 1.1426882745471878e-06, "loss": 1.2871, "step": 32522 }, { "epoch": 6.19, "learning_rate": 1.1426310772163966e-06, "loss": 2.736, "step": 32523 }, { "epoch": 6.2, "learning_rate": 1.1425738798856054e-06, "loss": 1.4003, "step": 32524 }, { "epoch": 6.2, "learning_rate": 1.1425166825548142e-06, "loss": 1.5586, "step": 32525 }, { "epoch": 6.2, "learning_rate": 1.1424594852240228e-06, "loss": 0.9933, "step": 32526 }, { "epoch": 6.2, "learning_rate": 1.1424022878932316e-06, "loss": 2.0336, "step": 32527 }, { "epoch": 6.2, "learning_rate": 1.1423450905624404e-06, "loss": 0.5633, "step": 32528 }, { "epoch": 6.2, "learning_rate": 1.1422878932316492e-06, "loss": 0.3682, "step": 32529 }, { "epoch": 6.2, "learning_rate": 1.142230695900858e-06, "loss": 3.5695, "step": 32530 }, { "epoch": 6.2, "learning_rate": 1.1421734985700668e-06, "loss": 2.4029, "step": 32531 }, { "epoch": 6.2, "learning_rate": 1.1421163012392756e-06, "loss": 2.0052, "step": 32532 }, { "epoch": 6.2, "learning_rate": 1.1420591039084842e-06, "loss": 1.9086, "step": 32533 }, { "epoch": 6.2, "learning_rate": 1.142001906577693e-06, "loss": 0.9143, "step": 32534 }, { "epoch": 6.2, "learning_rate": 1.141944709246902e-06, "loss": 3.0205, "step": 32535 }, { "epoch": 6.2, "learning_rate": 1.1418875119161107e-06, "loss": 2.0473, "step": 32536 }, { "epoch": 6.2, "learning_rate": 1.1418303145853193e-06, "loss": 2.4244, "step": 32537 }, { "epoch": 6.2, "learning_rate": 1.1417731172545281e-06, "loss": 1.2323, "step": 32538 }, { "epoch": 6.2, "learning_rate": 1.141715919923737e-06, "loss": 1.6777, "step": 32539 }, { "epoch": 6.2, "learning_rate": 1.1416587225929457e-06, "loss": 1.6549, "step": 32540 }, { "epoch": 6.2, "learning_rate": 1.1416015252621545e-06, "loss": 1.553, "step": 32541 }, { "epoch": 6.2, "learning_rate": 1.1415443279313633e-06, "loss": 0.8663, "step": 32542 }, { "epoch": 6.2, "learning_rate": 1.141487130600572e-06, "loss": 2.0277, "step": 32543 }, { "epoch": 6.2, "learning_rate": 1.1414299332697807e-06, "loss": 1.9513, "step": 32544 }, { "epoch": 6.2, "learning_rate": 1.1413727359389895e-06, "loss": 1.3586, "step": 32545 }, { "epoch": 6.2, "learning_rate": 1.1413155386081985e-06, "loss": 1.8091, "step": 32546 }, { "epoch": 6.2, "learning_rate": 1.141258341277407e-06, "loss": 1.3324, "step": 32547 }, { "epoch": 6.2, "learning_rate": 1.1412011439466158e-06, "loss": 2.6491, "step": 32548 }, { "epoch": 6.2, "learning_rate": 1.1411439466158246e-06, "loss": 3.3982, "step": 32549 }, { "epoch": 6.2, "learning_rate": 1.1410867492850332e-06, "loss": 1.3819, "step": 32550 }, { "epoch": 6.2, "learning_rate": 1.1410295519542422e-06, "loss": 2.4377, "step": 32551 }, { "epoch": 6.2, "learning_rate": 1.140972354623451e-06, "loss": 2.1914, "step": 32552 }, { "epoch": 6.2, "learning_rate": 1.1409151572926598e-06, "loss": 1.6703, "step": 32553 }, { "epoch": 6.2, "learning_rate": 1.1408579599618684e-06, "loss": 0.9115, "step": 32554 }, { "epoch": 6.2, "learning_rate": 1.1408007626310772e-06, "loss": 1.2615, "step": 32555 }, { "epoch": 6.2, "learning_rate": 1.140743565300286e-06, "loss": 2.0086, "step": 32556 }, { "epoch": 6.2, "learning_rate": 1.1406863679694948e-06, "loss": 1.205, "step": 32557 }, { "epoch": 6.2, "learning_rate": 1.1406291706387036e-06, "loss": 2.2832, "step": 32558 }, { "epoch": 6.2, "learning_rate": 1.1405719733079124e-06, "loss": 1.9821, "step": 32559 }, { "epoch": 6.2, "learning_rate": 1.1405147759771212e-06, "loss": 0.9302, "step": 32560 }, { "epoch": 6.2, "learning_rate": 1.1404575786463297e-06, "loss": 2.677, "step": 32561 }, { "epoch": 6.2, "learning_rate": 1.1404003813155385e-06, "loss": 1.9183, "step": 32562 }, { "epoch": 6.2, "learning_rate": 1.1403431839847475e-06, "loss": 3.3955, "step": 32563 }, { "epoch": 6.2, "learning_rate": 1.1402859866539561e-06, "loss": 1.4965, "step": 32564 }, { "epoch": 6.2, "learning_rate": 1.140228789323165e-06, "loss": 1.2461, "step": 32565 }, { "epoch": 6.2, "learning_rate": 1.1401715919923737e-06, "loss": 1.8605, "step": 32566 }, { "epoch": 6.2, "learning_rate": 1.1401143946615825e-06, "loss": 1.3978, "step": 32567 }, { "epoch": 6.2, "learning_rate": 1.1400571973307913e-06, "loss": 0.8983, "step": 32568 }, { "epoch": 6.2, "learning_rate": 1.14e-06, "loss": 1.5548, "step": 32569 }, { "epoch": 6.2, "learning_rate": 1.1399428026692089e-06, "loss": 1.3025, "step": 32570 }, { "epoch": 6.2, "learning_rate": 1.1398856053384175e-06, "loss": 1.7388, "step": 32571 }, { "epoch": 6.2, "learning_rate": 1.1398284080076263e-06, "loss": 1.2714, "step": 32572 }, { "epoch": 6.2, "learning_rate": 1.139771210676835e-06, "loss": 0.7601, "step": 32573 }, { "epoch": 6.2, "learning_rate": 1.1397140133460439e-06, "loss": 1.7577, "step": 32574 }, { "epoch": 6.2, "learning_rate": 1.1396568160152526e-06, "loss": 1.8729, "step": 32575 }, { "epoch": 6.2, "learning_rate": 1.1395996186844614e-06, "loss": 2.5136, "step": 32576 }, { "epoch": 6.21, "learning_rate": 1.1395424213536702e-06, "loss": 1.638, "step": 32577 }, { "epoch": 6.21, "learning_rate": 1.1394852240228788e-06, "loss": 0.7951, "step": 32578 }, { "epoch": 6.21, "learning_rate": 1.1394280266920878e-06, "loss": 1.9974, "step": 32579 }, { "epoch": 6.21, "learning_rate": 1.1393708293612966e-06, "loss": 1.4639, "step": 32580 }, { "epoch": 6.21, "learning_rate": 1.1393136320305052e-06, "loss": 1.9339, "step": 32581 }, { "epoch": 6.21, "learning_rate": 1.139256434699714e-06, "loss": 1.7564, "step": 32582 }, { "epoch": 6.21, "learning_rate": 1.1391992373689228e-06, "loss": 1.6182, "step": 32583 }, { "epoch": 6.21, "learning_rate": 1.1391420400381316e-06, "loss": 0.8609, "step": 32584 }, { "epoch": 6.21, "learning_rate": 1.1390848427073404e-06, "loss": 2.0159, "step": 32585 }, { "epoch": 6.21, "learning_rate": 1.1390276453765492e-06, "loss": 0.3557, "step": 32586 }, { "epoch": 6.21, "learning_rate": 1.138970448045758e-06, "loss": 1.3359, "step": 32587 }, { "epoch": 6.21, "learning_rate": 1.1389132507149665e-06, "loss": 1.0377, "step": 32588 }, { "epoch": 6.21, "learning_rate": 1.1388560533841753e-06, "loss": 2.4752, "step": 32589 }, { "epoch": 6.21, "learning_rate": 1.1387988560533843e-06, "loss": 1.2322, "step": 32590 }, { "epoch": 6.21, "learning_rate": 1.138741658722593e-06, "loss": 1.4095, "step": 32591 }, { "epoch": 6.21, "learning_rate": 1.1386844613918017e-06, "loss": 3.1907, "step": 32592 }, { "epoch": 6.21, "learning_rate": 1.1386272640610105e-06, "loss": 1.3803, "step": 32593 }, { "epoch": 6.21, "learning_rate": 1.1385700667302193e-06, "loss": 0.8619, "step": 32594 }, { "epoch": 6.21, "learning_rate": 1.138512869399428e-06, "loss": 1.0265, "step": 32595 }, { "epoch": 6.21, "learning_rate": 1.138455672068637e-06, "loss": 0.616, "step": 32596 }, { "epoch": 6.21, "learning_rate": 1.1383984747378457e-06, "loss": 1.4646, "step": 32597 }, { "epoch": 6.21, "learning_rate": 1.1383412774070543e-06, "loss": 1.4431, "step": 32598 }, { "epoch": 6.21, "learning_rate": 1.138284080076263e-06, "loss": 1.5695, "step": 32599 }, { "epoch": 6.21, "learning_rate": 1.1382268827454719e-06, "loss": 2.2562, "step": 32600 }, { "epoch": 6.21, "learning_rate": 1.1381696854146807e-06, "loss": 1.6218, "step": 32601 }, { "epoch": 6.21, "learning_rate": 1.1381124880838894e-06, "loss": 1.6408, "step": 32602 }, { "epoch": 6.21, "learning_rate": 1.1380552907530982e-06, "loss": 1.6714, "step": 32603 }, { "epoch": 6.21, "learning_rate": 1.137998093422307e-06, "loss": 3.1251, "step": 32604 }, { "epoch": 6.21, "learning_rate": 1.1379408960915156e-06, "loss": 2.4007, "step": 32605 }, { "epoch": 6.21, "learning_rate": 1.1378836987607244e-06, "loss": 1.5719, "step": 32606 }, { "epoch": 6.21, "learning_rate": 1.1378265014299334e-06, "loss": 1.536, "step": 32607 }, { "epoch": 6.21, "learning_rate": 1.1377693040991422e-06, "loss": 3.5477, "step": 32608 }, { "epoch": 6.21, "learning_rate": 1.1377121067683508e-06, "loss": 0.8221, "step": 32609 }, { "epoch": 6.21, "learning_rate": 1.1376549094375596e-06, "loss": 1.9487, "step": 32610 }, { "epoch": 6.21, "learning_rate": 1.1375977121067684e-06, "loss": 2.2114, "step": 32611 }, { "epoch": 6.21, "learning_rate": 1.1375405147759772e-06, "loss": 2.7667, "step": 32612 }, { "epoch": 6.21, "learning_rate": 1.137483317445186e-06, "loss": 2.0984, "step": 32613 }, { "epoch": 6.21, "learning_rate": 1.1374261201143948e-06, "loss": 1.9322, "step": 32614 }, { "epoch": 6.21, "learning_rate": 1.1373689227836033e-06, "loss": 2.32, "step": 32615 }, { "epoch": 6.21, "learning_rate": 1.1373117254528121e-06, "loss": 1.5788, "step": 32616 }, { "epoch": 6.21, "learning_rate": 1.137254528122021e-06, "loss": 1.9146, "step": 32617 }, { "epoch": 6.21, "learning_rate": 1.13719733079123e-06, "loss": 1.8996, "step": 32618 }, { "epoch": 6.21, "learning_rate": 1.1371401334604385e-06, "loss": 1.1914, "step": 32619 }, { "epoch": 6.21, "learning_rate": 1.1370829361296473e-06, "loss": 1.9586, "step": 32620 }, { "epoch": 6.21, "learning_rate": 1.137025738798856e-06, "loss": 1.5169, "step": 32621 }, { "epoch": 6.21, "learning_rate": 1.1369685414680647e-06, "loss": 2.3432, "step": 32622 }, { "epoch": 6.21, "learning_rate": 1.1369113441372737e-06, "loss": 2.7842, "step": 32623 }, { "epoch": 6.21, "learning_rate": 1.1368541468064825e-06, "loss": 1.0776, "step": 32624 }, { "epoch": 6.21, "learning_rate": 1.1367969494756913e-06, "loss": 2.8238, "step": 32625 }, { "epoch": 6.21, "learning_rate": 1.1367397521448999e-06, "loss": 1.9526, "step": 32626 }, { "epoch": 6.21, "learning_rate": 1.1366825548141087e-06, "loss": 2.3086, "step": 32627 }, { "epoch": 6.21, "learning_rate": 1.1366253574833174e-06, "loss": 1.3698, "step": 32628 }, { "epoch": 6.22, "learning_rate": 1.1365681601525262e-06, "loss": 2.4491, "step": 32629 }, { "epoch": 6.22, "learning_rate": 1.136510962821735e-06, "loss": 1.0573, "step": 32630 }, { "epoch": 6.22, "learning_rate": 1.1364537654909438e-06, "loss": 1.3631, "step": 32631 }, { "epoch": 6.22, "learning_rate": 1.1363965681601526e-06, "loss": 2.9483, "step": 32632 }, { "epoch": 6.22, "learning_rate": 1.1363393708293612e-06, "loss": 2.865, "step": 32633 }, { "epoch": 6.22, "learning_rate": 1.1362821734985702e-06, "loss": 1.6039, "step": 32634 }, { "epoch": 6.22, "learning_rate": 1.136224976167779e-06, "loss": 2.7862, "step": 32635 }, { "epoch": 6.22, "learning_rate": 1.1361677788369876e-06, "loss": 0.522, "step": 32636 }, { "epoch": 6.22, "learning_rate": 1.1361105815061964e-06, "loss": 1.8591, "step": 32637 }, { "epoch": 6.22, "learning_rate": 1.1360533841754052e-06, "loss": 2.023, "step": 32638 }, { "epoch": 6.22, "learning_rate": 1.135996186844614e-06, "loss": 0.9049, "step": 32639 }, { "epoch": 6.22, "learning_rate": 1.1359389895138228e-06, "loss": 1.3813, "step": 32640 }, { "epoch": 6.22, "learning_rate": 1.1358817921830316e-06, "loss": 1.9609, "step": 32641 }, { "epoch": 6.22, "learning_rate": 1.1358245948522404e-06, "loss": 3.1389, "step": 32642 }, { "epoch": 6.22, "learning_rate": 1.135767397521449e-06, "loss": 2.1812, "step": 32643 }, { "epoch": 6.22, "learning_rate": 1.1357102001906577e-06, "loss": 1.3359, "step": 32644 }, { "epoch": 6.22, "learning_rate": 1.1356530028598665e-06, "loss": 3.197, "step": 32645 }, { "epoch": 6.22, "learning_rate": 1.1355958055290753e-06, "loss": 2.3708, "step": 32646 }, { "epoch": 6.22, "learning_rate": 1.1355386081982841e-06, "loss": 1.2866, "step": 32647 }, { "epoch": 6.22, "learning_rate": 1.135481410867493e-06, "loss": 2.1975, "step": 32648 }, { "epoch": 6.22, "learning_rate": 1.1354242135367017e-06, "loss": 2.4839, "step": 32649 }, { "epoch": 6.22, "learning_rate": 1.1353670162059103e-06, "loss": 2.0218, "step": 32650 }, { "epoch": 6.22, "learning_rate": 1.1353098188751193e-06, "loss": 0.9572, "step": 32651 }, { "epoch": 6.22, "learning_rate": 1.135252621544328e-06, "loss": 0.6797, "step": 32652 }, { "epoch": 6.22, "learning_rate": 1.1351954242135367e-06, "loss": 2.867, "step": 32653 }, { "epoch": 6.22, "learning_rate": 1.1351382268827455e-06, "loss": 2.607, "step": 32654 }, { "epoch": 6.22, "learning_rate": 1.1350810295519542e-06, "loss": 1.062, "step": 32655 }, { "epoch": 6.22, "learning_rate": 1.135023832221163e-06, "loss": 0.6489, "step": 32656 }, { "epoch": 6.22, "learning_rate": 1.1349666348903718e-06, "loss": 1.4253, "step": 32657 }, { "epoch": 6.22, "learning_rate": 1.1349094375595806e-06, "loss": 1.7873, "step": 32658 }, { "epoch": 6.22, "learning_rate": 1.1348522402287894e-06, "loss": 1.5392, "step": 32659 }, { "epoch": 6.22, "learning_rate": 1.134795042897998e-06, "loss": 0.5659, "step": 32660 }, { "epoch": 6.22, "learning_rate": 1.1347378455672068e-06, "loss": 2.1986, "step": 32661 }, { "epoch": 6.22, "learning_rate": 1.1346806482364158e-06, "loss": 1.887, "step": 32662 }, { "epoch": 6.22, "learning_rate": 1.1346234509056244e-06, "loss": 0.843, "step": 32663 }, { "epoch": 6.22, "learning_rate": 1.1345662535748332e-06, "loss": 1.0098, "step": 32664 }, { "epoch": 6.22, "learning_rate": 1.134509056244042e-06, "loss": 2.1729, "step": 32665 }, { "epoch": 6.22, "learning_rate": 1.1344518589132508e-06, "loss": 1.8587, "step": 32666 }, { "epoch": 6.22, "learning_rate": 1.1343946615824596e-06, "loss": 0.4001, "step": 32667 }, { "epoch": 6.22, "learning_rate": 1.1343374642516684e-06, "loss": 3.2092, "step": 32668 }, { "epoch": 6.22, "learning_rate": 1.1342802669208771e-06, "loss": 1.6537, "step": 32669 }, { "epoch": 6.22, "learning_rate": 1.1342230695900857e-06, "loss": 1.7024, "step": 32670 }, { "epoch": 6.22, "learning_rate": 1.1341658722592945e-06, "loss": 2.8677, "step": 32671 }, { "epoch": 6.22, "learning_rate": 1.1341086749285033e-06, "loss": 1.7311, "step": 32672 }, { "epoch": 6.22, "learning_rate": 1.1340514775977121e-06, "loss": 0.9949, "step": 32673 }, { "epoch": 6.22, "learning_rate": 1.133994280266921e-06, "loss": 1.0264, "step": 32674 }, { "epoch": 6.22, "learning_rate": 1.1339370829361297e-06, "loss": 0.848, "step": 32675 }, { "epoch": 6.22, "learning_rate": 1.1338798856053385e-06, "loss": 1.97, "step": 32676 }, { "epoch": 6.22, "learning_rate": 1.133822688274547e-06, "loss": 0.9308, "step": 32677 }, { "epoch": 6.22, "learning_rate": 1.133765490943756e-06, "loss": 1.7791, "step": 32678 }, { "epoch": 6.22, "learning_rate": 1.1337082936129649e-06, "loss": 1.5926, "step": 32679 }, { "epoch": 6.22, "learning_rate": 1.1336510962821735e-06, "loss": 1.6147, "step": 32680 }, { "epoch": 6.22, "learning_rate": 1.1335938989513823e-06, "loss": 1.2695, "step": 32681 }, { "epoch": 6.23, "learning_rate": 1.133536701620591e-06, "loss": 1.3054, "step": 32682 }, { "epoch": 6.23, "learning_rate": 1.1334795042897998e-06, "loss": 1.7823, "step": 32683 }, { "epoch": 6.23, "learning_rate": 1.1334223069590086e-06, "loss": 1.7948, "step": 32684 }, { "epoch": 6.23, "learning_rate": 1.1333651096282174e-06, "loss": 1.4942, "step": 32685 }, { "epoch": 6.23, "learning_rate": 1.1333079122974262e-06, "loss": 2.5831, "step": 32686 }, { "epoch": 6.23, "learning_rate": 1.1332507149666348e-06, "loss": 1.4146, "step": 32687 }, { "epoch": 6.23, "learning_rate": 1.1331935176358436e-06, "loss": 1.5124, "step": 32688 }, { "epoch": 6.23, "learning_rate": 1.1331363203050524e-06, "loss": 3.9719, "step": 32689 }, { "epoch": 6.23, "learning_rate": 1.1330791229742614e-06, "loss": 0.7323, "step": 32690 }, { "epoch": 6.23, "learning_rate": 1.13302192564347e-06, "loss": 2.2612, "step": 32691 }, { "epoch": 6.23, "learning_rate": 1.1329647283126788e-06, "loss": 1.0162, "step": 32692 }, { "epoch": 6.23, "learning_rate": 1.1329075309818876e-06, "loss": 1.4441, "step": 32693 }, { "epoch": 6.23, "learning_rate": 1.1328503336510961e-06, "loss": 1.4285, "step": 32694 }, { "epoch": 6.23, "learning_rate": 1.1327931363203052e-06, "loss": 0.9266, "step": 32695 }, { "epoch": 6.23, "learning_rate": 1.132735938989514e-06, "loss": 2.5503, "step": 32696 }, { "epoch": 6.23, "learning_rate": 1.1326787416587225e-06, "loss": 2.578, "step": 32697 }, { "epoch": 6.23, "learning_rate": 1.1326215443279313e-06, "loss": 0.8152, "step": 32698 }, { "epoch": 6.23, "learning_rate": 1.1325643469971401e-06, "loss": 2.6777, "step": 32699 }, { "epoch": 6.23, "learning_rate": 1.132507149666349e-06, "loss": 1.8884, "step": 32700 }, { "epoch": 6.23, "learning_rate": 1.1324499523355577e-06, "loss": 1.4622, "step": 32701 }, { "epoch": 6.23, "learning_rate": 1.1323927550047665e-06, "loss": 1.7268, "step": 32702 }, { "epoch": 6.23, "learning_rate": 1.1323355576739753e-06, "loss": 2.2307, "step": 32703 }, { "epoch": 6.23, "learning_rate": 1.1322783603431839e-06, "loss": 2.009, "step": 32704 }, { "epoch": 6.23, "learning_rate": 1.1322211630123927e-06, "loss": 2.4352, "step": 32705 }, { "epoch": 6.23, "learning_rate": 1.1321639656816017e-06, "loss": 1.9827, "step": 32706 }, { "epoch": 6.23, "learning_rate": 1.1321067683508105e-06, "loss": 1.0305, "step": 32707 }, { "epoch": 6.23, "learning_rate": 1.132049571020019e-06, "loss": 1.6379, "step": 32708 }, { "epoch": 6.23, "learning_rate": 1.1319923736892278e-06, "loss": 1.4542, "step": 32709 }, { "epoch": 6.23, "learning_rate": 1.1319351763584366e-06, "loss": 2.0956, "step": 32710 }, { "epoch": 6.23, "learning_rate": 1.1318779790276454e-06, "loss": 1.4385, "step": 32711 }, { "epoch": 6.23, "learning_rate": 1.1318207816968542e-06, "loss": 1.3289, "step": 32712 }, { "epoch": 6.23, "learning_rate": 1.131763584366063e-06, "loss": 1.8019, "step": 32713 }, { "epoch": 6.23, "learning_rate": 1.1317063870352718e-06, "loss": 0.9965, "step": 32714 }, { "epoch": 6.23, "learning_rate": 1.1316491897044804e-06, "loss": 1.6554, "step": 32715 }, { "epoch": 6.23, "learning_rate": 1.1315919923736892e-06, "loss": 2.3606, "step": 32716 }, { "epoch": 6.23, "learning_rate": 1.131534795042898e-06, "loss": 1.129, "step": 32717 }, { "epoch": 6.23, "learning_rate": 1.1314775977121068e-06, "loss": 1.7721, "step": 32718 }, { "epoch": 6.23, "learning_rate": 1.1314204003813156e-06, "loss": 2.0261, "step": 32719 }, { "epoch": 6.23, "learning_rate": 1.1313632030505244e-06, "loss": 1.8452, "step": 32720 }, { "epoch": 6.23, "learning_rate": 1.131306005719733e-06, "loss": 1.1204, "step": 32721 }, { "epoch": 6.23, "learning_rate": 1.131248808388942e-06, "loss": 3.1882, "step": 32722 }, { "epoch": 6.23, "learning_rate": 1.1311916110581507e-06, "loss": 1.8549, "step": 32723 }, { "epoch": 6.23, "learning_rate": 1.1311344137273595e-06, "loss": 1.2205, "step": 32724 }, { "epoch": 6.23, "learning_rate": 1.1310772163965681e-06, "loss": 1.5732, "step": 32725 }, { "epoch": 6.23, "learning_rate": 1.131020019065777e-06, "loss": 1.0019, "step": 32726 }, { "epoch": 6.23, "learning_rate": 1.1309628217349857e-06, "loss": 2.6864, "step": 32727 }, { "epoch": 6.23, "learning_rate": 1.1309056244041945e-06, "loss": 0.9664, "step": 32728 }, { "epoch": 6.23, "learning_rate": 1.1308484270734033e-06, "loss": 1.8772, "step": 32729 }, { "epoch": 6.23, "learning_rate": 1.130791229742612e-06, "loss": 2.1283, "step": 32730 }, { "epoch": 6.23, "learning_rate": 1.1307340324118209e-06, "loss": 1.7835, "step": 32731 }, { "epoch": 6.23, "learning_rate": 1.1306768350810295e-06, "loss": 1.6992, "step": 32732 }, { "epoch": 6.23, "learning_rate": 1.1306196377502383e-06, "loss": 2.9091, "step": 32733 }, { "epoch": 6.24, "learning_rate": 1.1305624404194473e-06, "loss": 2.4717, "step": 32734 }, { "epoch": 6.24, "learning_rate": 1.1305052430886558e-06, "loss": 0.8655, "step": 32735 }, { "epoch": 6.24, "learning_rate": 1.1304480457578646e-06, "loss": 1.7496, "step": 32736 }, { "epoch": 6.24, "learning_rate": 1.1303908484270734e-06, "loss": 1.6545, "step": 32737 }, { "epoch": 6.24, "learning_rate": 1.1303336510962822e-06, "loss": 2.9604, "step": 32738 }, { "epoch": 6.24, "learning_rate": 1.130276453765491e-06, "loss": 1.5432, "step": 32739 }, { "epoch": 6.24, "learning_rate": 1.1302192564346998e-06, "loss": 1.5602, "step": 32740 }, { "epoch": 6.24, "learning_rate": 1.1301620591039086e-06, "loss": 0.5323, "step": 32741 }, { "epoch": 6.24, "learning_rate": 1.1301048617731172e-06, "loss": 2.1178, "step": 32742 }, { "epoch": 6.24, "learning_rate": 1.130047664442326e-06, "loss": 1.6728, "step": 32743 }, { "epoch": 6.24, "learning_rate": 1.1299904671115348e-06, "loss": 3.778, "step": 32744 }, { "epoch": 6.24, "learning_rate": 1.1299332697807436e-06, "loss": 0.9689, "step": 32745 }, { "epoch": 6.24, "learning_rate": 1.1298760724499524e-06, "loss": 1.1691, "step": 32746 }, { "epoch": 6.24, "learning_rate": 1.1298188751191612e-06, "loss": 2.1473, "step": 32747 }, { "epoch": 6.24, "learning_rate": 1.12976167778837e-06, "loss": 1.5515, "step": 32748 }, { "epoch": 6.24, "learning_rate": 1.1297044804575785e-06, "loss": 1.282, "step": 32749 }, { "epoch": 6.24, "learning_rate": 1.1296472831267875e-06, "loss": 0.947, "step": 32750 }, { "epoch": 6.24, "learning_rate": 1.1295900857959963e-06, "loss": 2.63, "step": 32751 }, { "epoch": 6.24, "learning_rate": 1.129532888465205e-06, "loss": 1.6569, "step": 32752 }, { "epoch": 6.24, "learning_rate": 1.1294756911344137e-06, "loss": 2.109, "step": 32753 }, { "epoch": 6.24, "learning_rate": 1.1294184938036225e-06, "loss": 1.1363, "step": 32754 }, { "epoch": 6.24, "learning_rate": 1.1293612964728313e-06, "loss": 1.7634, "step": 32755 }, { "epoch": 6.24, "learning_rate": 1.12930409914204e-06, "loss": 1.7631, "step": 32756 }, { "epoch": 6.24, "learning_rate": 1.1292469018112489e-06, "loss": 0.6928, "step": 32757 }, { "epoch": 6.24, "learning_rate": 1.1291897044804577e-06, "loss": 1.9518, "step": 32758 }, { "epoch": 6.24, "learning_rate": 1.1291325071496663e-06, "loss": 0.3603, "step": 32759 }, { "epoch": 6.24, "learning_rate": 1.129075309818875e-06, "loss": 1.5237, "step": 32760 }, { "epoch": 6.24, "learning_rate": 1.1290181124880839e-06, "loss": 1.7831, "step": 32761 }, { "epoch": 6.24, "learning_rate": 1.1289609151572926e-06, "loss": 2.0914, "step": 32762 }, { "epoch": 6.24, "learning_rate": 1.1289037178265014e-06, "loss": 2.6972, "step": 32763 }, { "epoch": 6.24, "learning_rate": 1.1288465204957102e-06, "loss": 1.6809, "step": 32764 }, { "epoch": 6.24, "learning_rate": 1.128789323164919e-06, "loss": 1.5816, "step": 32765 }, { "epoch": 6.24, "learning_rate": 1.1287321258341278e-06, "loss": 2.3313, "step": 32766 }, { "epoch": 6.24, "learning_rate": 1.1286749285033366e-06, "loss": 1.7193, "step": 32767 }, { "epoch": 6.24, "learning_rate": 1.1286177311725454e-06, "loss": 1.3674, "step": 32768 }, { "epoch": 6.24, "learning_rate": 1.128560533841754e-06, "loss": 1.8557, "step": 32769 }, { "epoch": 6.24, "learning_rate": 1.1285033365109628e-06, "loss": 2.3669, "step": 32770 }, { "epoch": 6.24, "learning_rate": 1.1284461391801716e-06, "loss": 1.6744, "step": 32771 }, { "epoch": 6.24, "learning_rate": 1.1283889418493804e-06, "loss": 0.7028, "step": 32772 }, { "epoch": 6.24, "learning_rate": 1.1283317445185892e-06, "loss": 1.5812, "step": 32773 }, { "epoch": 6.24, "learning_rate": 1.128274547187798e-06, "loss": 0.9658, "step": 32774 }, { "epoch": 6.24, "learning_rate": 1.1282173498570068e-06, "loss": 2.3649, "step": 32775 }, { "epoch": 6.24, "learning_rate": 1.1281601525262153e-06, "loss": 2.5326, "step": 32776 }, { "epoch": 6.24, "learning_rate": 1.1281029551954241e-06, "loss": 1.7941, "step": 32777 }, { "epoch": 6.24, "learning_rate": 1.1280457578646331e-06, "loss": 2.1403, "step": 32778 }, { "epoch": 6.24, "learning_rate": 1.127988560533842e-06, "loss": 2.6694, "step": 32779 }, { "epoch": 6.24, "learning_rate": 1.1279313632030505e-06, "loss": 1.3265, "step": 32780 }, { "epoch": 6.24, "learning_rate": 1.1278741658722593e-06, "loss": 2.5623, "step": 32781 }, { "epoch": 6.24, "learning_rate": 1.127816968541468e-06, "loss": 2.2574, "step": 32782 }, { "epoch": 6.24, "learning_rate": 1.1277597712106769e-06, "loss": 1.4666, "step": 32783 }, { "epoch": 6.24, "learning_rate": 1.1277025738798857e-06, "loss": 1.9498, "step": 32784 }, { "epoch": 6.24, "learning_rate": 1.1276453765490945e-06, "loss": 2.1894, "step": 32785 }, { "epoch": 6.24, "learning_rate": 1.127588179218303e-06, "loss": 1.8378, "step": 32786 }, { "epoch": 6.25, "learning_rate": 1.1275309818875119e-06, "loss": 2.9227, "step": 32787 }, { "epoch": 6.25, "learning_rate": 1.1274737845567206e-06, "loss": 2.1988, "step": 32788 }, { "epoch": 6.25, "learning_rate": 1.1274165872259297e-06, "loss": 1.2426, "step": 32789 }, { "epoch": 6.25, "learning_rate": 1.1273593898951382e-06, "loss": 1.6032, "step": 32790 }, { "epoch": 6.25, "learning_rate": 1.127302192564347e-06, "loss": 1.2285, "step": 32791 }, { "epoch": 6.25, "learning_rate": 1.1272449952335558e-06, "loss": 2.1157, "step": 32792 }, { "epoch": 6.25, "learning_rate": 1.1271877979027644e-06, "loss": 1.4732, "step": 32793 }, { "epoch": 6.25, "learning_rate": 1.1271306005719734e-06, "loss": 0.8152, "step": 32794 }, { "epoch": 6.25, "learning_rate": 1.1270734032411822e-06, "loss": 2.5179, "step": 32795 }, { "epoch": 6.25, "learning_rate": 1.127016205910391e-06, "loss": 1.1849, "step": 32796 }, { "epoch": 6.25, "learning_rate": 1.1269590085795996e-06, "loss": 1.4582, "step": 32797 }, { "epoch": 6.25, "learning_rate": 1.1269018112488084e-06, "loss": 0.9849, "step": 32798 }, { "epoch": 6.25, "learning_rate": 1.1268446139180172e-06, "loss": 1.2876, "step": 32799 }, { "epoch": 6.25, "learning_rate": 1.126787416587226e-06, "loss": 1.6523, "step": 32800 }, { "epoch": 6.25, "learning_rate": 1.1267302192564348e-06, "loss": 1.8305, "step": 32801 }, { "epoch": 6.25, "learning_rate": 1.1266730219256436e-06, "loss": 1.5228, "step": 32802 }, { "epoch": 6.25, "learning_rate": 1.1266158245948523e-06, "loss": 1.4963, "step": 32803 }, { "epoch": 6.25, "learning_rate": 1.126558627264061e-06, "loss": 1.4049, "step": 32804 }, { "epoch": 6.25, "learning_rate": 1.1265014299332697e-06, "loss": 1.2498, "step": 32805 }, { "epoch": 6.25, "learning_rate": 1.1264442326024787e-06, "loss": 1.5061, "step": 32806 }, { "epoch": 6.25, "learning_rate": 1.1263870352716873e-06, "loss": 0.4734, "step": 32807 }, { "epoch": 6.25, "learning_rate": 1.126329837940896e-06, "loss": 0.6821, "step": 32808 }, { "epoch": 6.25, "learning_rate": 1.126272640610105e-06, "loss": 3.2898, "step": 32809 }, { "epoch": 6.25, "learning_rate": 1.1262154432793137e-06, "loss": 2.7732, "step": 32810 }, { "epoch": 6.25, "learning_rate": 1.1261582459485225e-06, "loss": 0.5402, "step": 32811 }, { "epoch": 6.25, "learning_rate": 1.1261010486177313e-06, "loss": 2.1572, "step": 32812 }, { "epoch": 6.25, "learning_rate": 1.12604385128694e-06, "loss": 1.9235, "step": 32813 }, { "epoch": 6.25, "learning_rate": 1.1259866539561487e-06, "loss": 1.4076, "step": 32814 }, { "epoch": 6.25, "learning_rate": 1.1259294566253574e-06, "loss": 1.7669, "step": 32815 }, { "epoch": 6.25, "learning_rate": 1.1258722592945662e-06, "loss": 1.2223, "step": 32816 }, { "epoch": 6.25, "learning_rate": 1.125815061963775e-06, "loss": 2.4246, "step": 32817 }, { "epoch": 6.25, "learning_rate": 1.1257578646329838e-06, "loss": 1.8703, "step": 32818 }, { "epoch": 6.25, "learning_rate": 1.1257006673021926e-06, "loss": 2.7065, "step": 32819 }, { "epoch": 6.25, "learning_rate": 1.1256434699714014e-06, "loss": 1.3518, "step": 32820 }, { "epoch": 6.25, "learning_rate": 1.12558627264061e-06, "loss": 2.0299, "step": 32821 }, { "epoch": 6.25, "learning_rate": 1.125529075309819e-06, "loss": 1.5609, "step": 32822 }, { "epoch": 6.25, "learning_rate": 1.1254718779790278e-06, "loss": 1.879, "step": 32823 }, { "epoch": 6.25, "learning_rate": 1.1254146806482364e-06, "loss": 1.859, "step": 32824 }, { "epoch": 6.25, "learning_rate": 1.1253574833174452e-06, "loss": 1.3673, "step": 32825 }, { "epoch": 6.25, "learning_rate": 1.125300285986654e-06, "loss": 1.47, "step": 32826 }, { "epoch": 6.25, "learning_rate": 1.1252430886558628e-06, "loss": 1.5568, "step": 32827 }, { "epoch": 6.25, "learning_rate": 1.1251858913250716e-06, "loss": 1.1523, "step": 32828 }, { "epoch": 6.25, "learning_rate": 1.1251286939942803e-06, "loss": 1.2309, "step": 32829 }, { "epoch": 6.25, "learning_rate": 1.1250714966634891e-06, "loss": 1.2298, "step": 32830 }, { "epoch": 6.25, "learning_rate": 1.1250142993326977e-06, "loss": 2.0866, "step": 32831 }, { "epoch": 6.25, "learning_rate": 1.1249571020019065e-06, "loss": 1.8563, "step": 32832 }, { "epoch": 6.25, "learning_rate": 1.1248999046711155e-06, "loss": 3.205, "step": 32833 }, { "epoch": 6.25, "learning_rate": 1.124842707340324e-06, "loss": 1.7479, "step": 32834 }, { "epoch": 6.25, "learning_rate": 1.124785510009533e-06, "loss": 2.8148, "step": 32835 }, { "epoch": 6.25, "learning_rate": 1.1247283126787417e-06, "loss": 1.6368, "step": 32836 }, { "epoch": 6.25, "learning_rate": 1.1246711153479505e-06, "loss": 1.8175, "step": 32837 }, { "epoch": 6.25, "learning_rate": 1.1246139180171593e-06, "loss": 1.8587, "step": 32838 }, { "epoch": 6.26, "learning_rate": 1.124556720686368e-06, "loss": 2.3819, "step": 32839 }, { "epoch": 6.26, "learning_rate": 1.1244995233555769e-06, "loss": 3.529, "step": 32840 }, { "epoch": 6.26, "learning_rate": 1.1244423260247855e-06, "loss": 1.7459, "step": 32841 }, { "epoch": 6.26, "learning_rate": 1.1243851286939942e-06, "loss": 2.2427, "step": 32842 }, { "epoch": 6.26, "learning_rate": 1.124327931363203e-06, "loss": 2.1247, "step": 32843 }, { "epoch": 6.26, "learning_rate": 1.1242707340324118e-06, "loss": 2.4658, "step": 32844 }, { "epoch": 6.26, "learning_rate": 1.1242135367016206e-06, "loss": 1.0913, "step": 32845 }, { "epoch": 6.26, "learning_rate": 1.1241563393708294e-06, "loss": 1.8764, "step": 32846 }, { "epoch": 6.26, "learning_rate": 1.1240991420400382e-06, "loss": 1.7639, "step": 32847 }, { "epoch": 6.26, "learning_rate": 1.1240419447092468e-06, "loss": 1.4914, "step": 32848 }, { "epoch": 6.26, "learning_rate": 1.1239847473784556e-06, "loss": 1.5212, "step": 32849 }, { "epoch": 6.26, "learning_rate": 1.1239275500476646e-06, "loss": 1.829, "step": 32850 }, { "epoch": 6.26, "learning_rate": 1.1238703527168732e-06, "loss": 1.8868, "step": 32851 }, { "epoch": 6.26, "learning_rate": 1.123813155386082e-06, "loss": 2.1125, "step": 32852 }, { "epoch": 6.26, "learning_rate": 1.1237559580552908e-06, "loss": 1.51, "step": 32853 }, { "epoch": 6.26, "learning_rate": 1.1236987607244996e-06, "loss": 1.2787, "step": 32854 }, { "epoch": 6.26, "learning_rate": 1.1236415633937084e-06, "loss": 1.2205, "step": 32855 }, { "epoch": 6.26, "learning_rate": 1.1235843660629171e-06, "loss": 0.7566, "step": 32856 }, { "epoch": 6.26, "learning_rate": 1.123527168732126e-06, "loss": 2.063, "step": 32857 }, { "epoch": 6.26, "learning_rate": 1.1234699714013345e-06, "loss": 1.4678, "step": 32858 }, { "epoch": 6.26, "learning_rate": 1.1234127740705433e-06, "loss": 2.3097, "step": 32859 }, { "epoch": 6.26, "learning_rate": 1.1233555767397521e-06, "loss": 1.1487, "step": 32860 }, { "epoch": 6.26, "learning_rate": 1.1232983794089611e-06, "loss": 2.0412, "step": 32861 }, { "epoch": 6.26, "learning_rate": 1.1232411820781697e-06, "loss": 2.9721, "step": 32862 }, { "epoch": 6.26, "learning_rate": 1.1231839847473785e-06, "loss": 1.8052, "step": 32863 }, { "epoch": 6.26, "learning_rate": 1.1231267874165873e-06, "loss": 1.6288, "step": 32864 }, { "epoch": 6.26, "learning_rate": 1.1230695900857959e-06, "loss": 1.0451, "step": 32865 }, { "epoch": 6.26, "learning_rate": 1.1230123927550049e-06, "loss": 2.0076, "step": 32866 }, { "epoch": 6.26, "learning_rate": 1.1229551954242137e-06, "loss": 1.2718, "step": 32867 }, { "epoch": 6.26, "learning_rate": 1.1228979980934225e-06, "loss": 1.319, "step": 32868 }, { "epoch": 6.26, "learning_rate": 1.122840800762631e-06, "loss": 2.0654, "step": 32869 }, { "epoch": 6.26, "learning_rate": 1.1227836034318398e-06, "loss": 2.2727, "step": 32870 }, { "epoch": 6.26, "learning_rate": 1.1227264061010486e-06, "loss": 3.3565, "step": 32871 }, { "epoch": 6.26, "learning_rate": 1.1226692087702574e-06, "loss": 1.3469, "step": 32872 }, { "epoch": 6.26, "learning_rate": 1.1226120114394662e-06, "loss": 1.5318, "step": 32873 }, { "epoch": 6.26, "learning_rate": 1.122554814108675e-06, "loss": 1.7995, "step": 32874 }, { "epoch": 6.26, "learning_rate": 1.1224976167778836e-06, "loss": 2.3569, "step": 32875 }, { "epoch": 6.26, "learning_rate": 1.1224404194470924e-06, "loss": 1.0517, "step": 32876 }, { "epoch": 6.26, "learning_rate": 1.1223832221163014e-06, "loss": 1.0047, "step": 32877 }, { "epoch": 6.26, "learning_rate": 1.1223260247855102e-06, "loss": 1.0773, "step": 32878 }, { "epoch": 6.26, "learning_rate": 1.1222688274547188e-06, "loss": 2.3835, "step": 32879 }, { "epoch": 6.26, "learning_rate": 1.1222116301239276e-06, "loss": 0.8288, "step": 32880 }, { "epoch": 6.26, "learning_rate": 1.1221544327931364e-06, "loss": 1.5455, "step": 32881 }, { "epoch": 6.26, "learning_rate": 1.1220972354623452e-06, "loss": 1.2574, "step": 32882 }, { "epoch": 6.26, "learning_rate": 1.122040038131554e-06, "loss": 2.9491, "step": 32883 }, { "epoch": 6.26, "learning_rate": 1.1219828408007627e-06, "loss": 1.6383, "step": 32884 }, { "epoch": 6.26, "learning_rate": 1.1219256434699715e-06, "loss": 1.0683, "step": 32885 }, { "epoch": 6.26, "learning_rate": 1.1218684461391801e-06, "loss": 1.5957, "step": 32886 }, { "epoch": 6.26, "learning_rate": 1.121811248808389e-06, "loss": 1.3946, "step": 32887 }, { "epoch": 6.26, "learning_rate": 1.1217540514775977e-06, "loss": 0.658, "step": 32888 }, { "epoch": 6.26, "learning_rate": 1.1216968541468065e-06, "loss": 1.9014, "step": 32889 }, { "epoch": 6.26, "learning_rate": 1.1216396568160153e-06, "loss": 0.865, "step": 32890 }, { "epoch": 6.26, "learning_rate": 1.121582459485224e-06, "loss": 3.2695, "step": 32891 }, { "epoch": 6.27, "learning_rate": 1.1215252621544327e-06, "loss": 1.5698, "step": 32892 }, { "epoch": 6.27, "learning_rate": 1.1214680648236415e-06, "loss": 2.1234, "step": 32893 }, { "epoch": 6.27, "learning_rate": 1.1214108674928505e-06, "loss": 2.8874, "step": 32894 }, { "epoch": 6.27, "learning_rate": 1.1213536701620593e-06, "loss": 1.7288, "step": 32895 }, { "epoch": 6.27, "learning_rate": 1.1212964728312678e-06, "loss": 1.6115, "step": 32896 }, { "epoch": 6.27, "learning_rate": 1.1212392755004766e-06, "loss": 1.8863, "step": 32897 }, { "epoch": 6.27, "learning_rate": 1.1211820781696854e-06, "loss": 1.0415, "step": 32898 }, { "epoch": 6.27, "learning_rate": 1.1211248808388942e-06, "loss": 2.7094, "step": 32899 }, { "epoch": 6.27, "learning_rate": 1.121067683508103e-06, "loss": 2.3687, "step": 32900 }, { "epoch": 6.27, "learning_rate": 1.1210104861773118e-06, "loss": 1.7482, "step": 32901 }, { "epoch": 6.27, "learning_rate": 1.1209532888465206e-06, "loss": 2.7302, "step": 32902 }, { "epoch": 6.27, "learning_rate": 1.1208960915157292e-06, "loss": 2.8198, "step": 32903 }, { "epoch": 6.27, "learning_rate": 1.120838894184938e-06, "loss": 2.2933, "step": 32904 }, { "epoch": 6.27, "learning_rate": 1.120781696854147e-06, "loss": 2.8287, "step": 32905 }, { "epoch": 6.27, "learning_rate": 1.1207244995233556e-06, "loss": 1.8414, "step": 32906 }, { "epoch": 6.27, "learning_rate": 1.1206673021925644e-06, "loss": 1.7704, "step": 32907 }, { "epoch": 6.27, "learning_rate": 1.1206101048617732e-06, "loss": 2.3235, "step": 32908 }, { "epoch": 6.27, "learning_rate": 1.120552907530982e-06, "loss": 1.6733, "step": 32909 }, { "epoch": 6.27, "learning_rate": 1.1204957102001907e-06, "loss": 1.9701, "step": 32910 }, { "epoch": 6.27, "learning_rate": 1.1204385128693995e-06, "loss": 1.8117, "step": 32911 }, { "epoch": 6.27, "learning_rate": 1.1203813155386083e-06, "loss": 3.0942, "step": 32912 }, { "epoch": 6.27, "learning_rate": 1.120324118207817e-06, "loss": 1.6884, "step": 32913 }, { "epoch": 6.27, "learning_rate": 1.1202669208770257e-06, "loss": 1.4591, "step": 32914 }, { "epoch": 6.27, "learning_rate": 1.1202097235462345e-06, "loss": 1.7889, "step": 32915 }, { "epoch": 6.27, "learning_rate": 1.1201525262154433e-06, "loss": 1.1582, "step": 32916 }, { "epoch": 6.27, "learning_rate": 1.120095328884652e-06, "loss": 1.4027, "step": 32917 }, { "epoch": 6.27, "learning_rate": 1.1200381315538609e-06, "loss": 1.7394, "step": 32918 }, { "epoch": 6.27, "learning_rate": 1.1199809342230697e-06, "loss": 1.3656, "step": 32919 }, { "epoch": 6.27, "learning_rate": 1.1199237368922783e-06, "loss": 3.0189, "step": 32920 }, { "epoch": 6.27, "learning_rate": 1.1198665395614873e-06, "loss": 3.0033, "step": 32921 }, { "epoch": 6.27, "learning_rate": 1.119809342230696e-06, "loss": 1.9934, "step": 32922 }, { "epoch": 6.27, "learning_rate": 1.1197521448999046e-06, "loss": 1.156, "step": 32923 }, { "epoch": 6.27, "learning_rate": 1.1196949475691134e-06, "loss": 2.5706, "step": 32924 }, { "epoch": 6.27, "learning_rate": 1.1196377502383222e-06, "loss": 2.1051, "step": 32925 }, { "epoch": 6.27, "learning_rate": 1.119580552907531e-06, "loss": 3.1885, "step": 32926 }, { "epoch": 6.27, "learning_rate": 1.1195233555767398e-06, "loss": 1.4551, "step": 32927 }, { "epoch": 6.27, "learning_rate": 1.1194661582459486e-06, "loss": 2.2714, "step": 32928 }, { "epoch": 6.27, "learning_rate": 1.1194089609151574e-06, "loss": 1.9891, "step": 32929 }, { "epoch": 6.27, "learning_rate": 1.119351763584366e-06, "loss": 1.0948, "step": 32930 }, { "epoch": 6.27, "learning_rate": 1.1192945662535748e-06, "loss": 1.8628, "step": 32931 }, { "epoch": 6.27, "learning_rate": 1.1192373689227836e-06, "loss": 1.0074, "step": 32932 }, { "epoch": 6.27, "learning_rate": 1.1191801715919926e-06, "loss": 2.5844, "step": 32933 }, { "epoch": 6.27, "learning_rate": 1.1191229742612012e-06, "loss": 3.4453, "step": 32934 }, { "epoch": 6.27, "learning_rate": 1.11906577693041e-06, "loss": 0.4253, "step": 32935 }, { "epoch": 6.27, "learning_rate": 1.1190085795996187e-06, "loss": 1.9436, "step": 32936 }, { "epoch": 6.27, "learning_rate": 1.1189513822688273e-06, "loss": 2.0314, "step": 32937 }, { "epoch": 6.27, "learning_rate": 1.1188941849380363e-06, "loss": 1.5594, "step": 32938 }, { "epoch": 6.27, "learning_rate": 1.1188369876072451e-06, "loss": 1.9376, "step": 32939 }, { "epoch": 6.27, "learning_rate": 1.1187797902764537e-06, "loss": 2.1691, "step": 32940 }, { "epoch": 6.27, "learning_rate": 1.1187225929456625e-06, "loss": 0.7223, "step": 32941 }, { "epoch": 6.27, "learning_rate": 1.1186653956148713e-06, "loss": 1.559, "step": 32942 }, { "epoch": 6.27, "learning_rate": 1.11860819828408e-06, "loss": 1.1331, "step": 32943 }, { "epoch": 6.28, "learning_rate": 1.1185510009532889e-06, "loss": 1.7138, "step": 32944 }, { "epoch": 6.28, "learning_rate": 1.1184938036224977e-06, "loss": 1.699, "step": 32945 }, { "epoch": 6.28, "learning_rate": 1.1184366062917065e-06, "loss": 1.8236, "step": 32946 }, { "epoch": 6.28, "learning_rate": 1.118379408960915e-06, "loss": 1.3548, "step": 32947 }, { "epoch": 6.28, "learning_rate": 1.1183222116301238e-06, "loss": 1.365, "step": 32948 }, { "epoch": 6.28, "learning_rate": 1.1182650142993329e-06, "loss": 3.0151, "step": 32949 }, { "epoch": 6.28, "learning_rate": 1.1182078169685416e-06, "loss": 1.958, "step": 32950 }, { "epoch": 6.28, "learning_rate": 1.1181506196377502e-06, "loss": 2.4108, "step": 32951 }, { "epoch": 6.28, "learning_rate": 1.118093422306959e-06, "loss": 1.3818, "step": 32952 }, { "epoch": 6.28, "learning_rate": 1.1180362249761678e-06, "loss": 1.7676, "step": 32953 }, { "epoch": 6.28, "learning_rate": 1.1179790276453766e-06, "loss": 1.334, "step": 32954 }, { "epoch": 6.28, "learning_rate": 1.1179218303145854e-06, "loss": 1.2169, "step": 32955 }, { "epoch": 6.28, "learning_rate": 1.1178646329837942e-06, "loss": 0.6219, "step": 32956 }, { "epoch": 6.28, "learning_rate": 1.1178074356530028e-06, "loss": 1.1155, "step": 32957 }, { "epoch": 6.28, "learning_rate": 1.1177502383222116e-06, "loss": 1.2931, "step": 32958 }, { "epoch": 6.28, "learning_rate": 1.1176930409914204e-06, "loss": 1.3738, "step": 32959 }, { "epoch": 6.28, "learning_rate": 1.1176358436606292e-06, "loss": 1.7446, "step": 32960 }, { "epoch": 6.28, "learning_rate": 1.117578646329838e-06, "loss": 2.0505, "step": 32961 }, { "epoch": 6.28, "learning_rate": 1.1175214489990467e-06, "loss": 1.4083, "step": 32962 }, { "epoch": 6.28, "learning_rate": 1.1174642516682555e-06, "loss": 3.2639, "step": 32963 }, { "epoch": 6.28, "learning_rate": 1.1174070543374641e-06, "loss": 1.6393, "step": 32964 }, { "epoch": 6.28, "learning_rate": 1.1173498570066731e-06, "loss": 1.4203, "step": 32965 }, { "epoch": 6.28, "learning_rate": 1.117292659675882e-06, "loss": 1.2213, "step": 32966 }, { "epoch": 6.28, "learning_rate": 1.1172354623450907e-06, "loss": 1.895, "step": 32967 }, { "epoch": 6.28, "learning_rate": 1.1171782650142993e-06, "loss": 1.3509, "step": 32968 }, { "epoch": 6.28, "learning_rate": 1.117121067683508e-06, "loss": 0.7844, "step": 32969 }, { "epoch": 6.28, "learning_rate": 1.1170638703527169e-06, "loss": 3.2096, "step": 32970 }, { "epoch": 6.28, "learning_rate": 1.1170066730219257e-06, "loss": 2.2366, "step": 32971 }, { "epoch": 6.28, "learning_rate": 1.1169494756911345e-06, "loss": 0.9045, "step": 32972 }, { "epoch": 6.28, "learning_rate": 1.1168922783603433e-06, "loss": 1.6801, "step": 32973 }, { "epoch": 6.28, "learning_rate": 1.116835081029552e-06, "loss": 0.4833, "step": 32974 }, { "epoch": 6.28, "learning_rate": 1.1167778836987606e-06, "loss": 1.7043, "step": 32975 }, { "epoch": 6.28, "learning_rate": 1.1167206863679694e-06, "loss": 3.0259, "step": 32976 }, { "epoch": 6.28, "learning_rate": 1.1166634890371784e-06, "loss": 2.0587, "step": 32977 }, { "epoch": 6.28, "learning_rate": 1.116606291706387e-06, "loss": 1.574, "step": 32978 }, { "epoch": 6.28, "learning_rate": 1.1165490943755958e-06, "loss": 1.5043, "step": 32979 }, { "epoch": 6.28, "learning_rate": 1.1164918970448046e-06, "loss": 1.7576, "step": 32980 }, { "epoch": 6.28, "learning_rate": 1.1164346997140132e-06, "loss": 2.6584, "step": 32981 }, { "epoch": 6.28, "learning_rate": 1.1163775023832222e-06, "loss": 2.2474, "step": 32982 }, { "epoch": 6.28, "learning_rate": 1.116320305052431e-06, "loss": 1.7697, "step": 32983 }, { "epoch": 6.28, "learning_rate": 1.1162631077216398e-06, "loss": 3.1363, "step": 32984 }, { "epoch": 6.28, "learning_rate": 1.1162059103908484e-06, "loss": 1.5118, "step": 32985 }, { "epoch": 6.28, "learning_rate": 1.1161487130600572e-06, "loss": 2.5967, "step": 32986 }, { "epoch": 6.28, "learning_rate": 1.116091515729266e-06, "loss": 1.7771, "step": 32987 }, { "epoch": 6.28, "learning_rate": 1.1160343183984748e-06, "loss": 2.5973, "step": 32988 }, { "epoch": 6.28, "learning_rate": 1.1159771210676835e-06, "loss": 0.8778, "step": 32989 }, { "epoch": 6.28, "learning_rate": 1.1159199237368923e-06, "loss": 1.2058, "step": 32990 }, { "epoch": 6.28, "learning_rate": 1.1158627264061011e-06, "loss": 0.522, "step": 32991 }, { "epoch": 6.28, "learning_rate": 1.1158055290753097e-06, "loss": 1.5779, "step": 32992 }, { "epoch": 6.28, "learning_rate": 1.1157483317445187e-06, "loss": 1.3259, "step": 32993 }, { "epoch": 6.28, "learning_rate": 1.1156911344137275e-06, "loss": 2.4213, "step": 32994 }, { "epoch": 6.28, "learning_rate": 1.115633937082936e-06, "loss": 3.0771, "step": 32995 }, { "epoch": 6.28, "learning_rate": 1.1155767397521449e-06, "loss": 1.9735, "step": 32996 }, { "epoch": 6.29, "learning_rate": 1.1155195424213537e-06, "loss": 1.846, "step": 32997 }, { "epoch": 6.29, "learning_rate": 1.1154623450905625e-06, "loss": 0.4794, "step": 32998 }, { "epoch": 6.29, "learning_rate": 1.1154051477597713e-06, "loss": 0.7718, "step": 32999 }, { "epoch": 6.29, "learning_rate": 1.11534795042898e-06, "loss": 1.3135, "step": 33000 }, { "epoch": 6.29, "learning_rate": 1.1152907530981889e-06, "loss": 2.475, "step": 33001 }, { "epoch": 6.29, "learning_rate": 1.1152335557673974e-06, "loss": 2.8739, "step": 33002 }, { "epoch": 6.29, "learning_rate": 1.1151763584366062e-06, "loss": 1.6321, "step": 33003 }, { "epoch": 6.29, "learning_rate": 1.115119161105815e-06, "loss": 1.5228, "step": 33004 }, { "epoch": 6.29, "learning_rate": 1.1150619637750238e-06, "loss": 1.9787, "step": 33005 }, { "epoch": 6.29, "learning_rate": 1.1150047664442326e-06, "loss": 1.236, "step": 33006 }, { "epoch": 6.29, "learning_rate": 1.1149475691134414e-06, "loss": 0.648, "step": 33007 }, { "epoch": 6.29, "learning_rate": 1.1148903717826502e-06, "loss": 2.4916, "step": 33008 }, { "epoch": 6.29, "learning_rate": 1.1148331744518588e-06, "loss": 0.9598, "step": 33009 }, { "epoch": 6.29, "learning_rate": 1.1147759771210678e-06, "loss": 1.7838, "step": 33010 }, { "epoch": 6.29, "learning_rate": 1.1147187797902766e-06, "loss": 1.6786, "step": 33011 }, { "epoch": 6.29, "learning_rate": 1.1146615824594852e-06, "loss": 2.6814, "step": 33012 }, { "epoch": 6.29, "learning_rate": 1.114604385128694e-06, "loss": 1.0323, "step": 33013 }, { "epoch": 6.29, "learning_rate": 1.1145471877979028e-06, "loss": 2.2633, "step": 33014 }, { "epoch": 6.29, "learning_rate": 1.1144899904671116e-06, "loss": 1.5706, "step": 33015 }, { "epoch": 6.29, "learning_rate": 1.1144327931363203e-06, "loss": 2.5378, "step": 33016 }, { "epoch": 6.29, "learning_rate": 1.1143755958055291e-06, "loss": 1.5199, "step": 33017 }, { "epoch": 6.29, "learning_rate": 1.114318398474738e-06, "loss": 1.3141, "step": 33018 }, { "epoch": 6.29, "learning_rate": 1.1142612011439465e-06, "loss": 1.3785, "step": 33019 }, { "epoch": 6.29, "learning_rate": 1.1142040038131553e-06, "loss": 2.522, "step": 33020 }, { "epoch": 6.29, "learning_rate": 1.1141468064823643e-06, "loss": 1.8751, "step": 33021 }, { "epoch": 6.29, "learning_rate": 1.114089609151573e-06, "loss": 2.656, "step": 33022 }, { "epoch": 6.29, "learning_rate": 1.1140324118207817e-06, "loss": 1.6319, "step": 33023 }, { "epoch": 6.29, "learning_rate": 1.1139752144899905e-06, "loss": 2.122, "step": 33024 }, { "epoch": 6.29, "learning_rate": 1.1139180171591993e-06, "loss": 1.5308, "step": 33025 }, { "epoch": 6.29, "learning_rate": 1.113860819828408e-06, "loss": 1.2434, "step": 33026 }, { "epoch": 6.29, "learning_rate": 1.1138036224976169e-06, "loss": 1.0689, "step": 33027 }, { "epoch": 6.29, "learning_rate": 1.1137464251668257e-06, "loss": 1.3036, "step": 33028 }, { "epoch": 6.29, "learning_rate": 1.1136892278360342e-06, "loss": 2.8579, "step": 33029 }, { "epoch": 6.29, "learning_rate": 1.113632030505243e-06, "loss": 2.9651, "step": 33030 }, { "epoch": 6.29, "learning_rate": 1.1135748331744518e-06, "loss": 2.6113, "step": 33031 }, { "epoch": 6.29, "learning_rate": 1.1135176358436608e-06, "loss": 1.5005, "step": 33032 }, { "epoch": 6.29, "learning_rate": 1.1134604385128694e-06, "loss": 1.9365, "step": 33033 }, { "epoch": 6.29, "learning_rate": 1.1134032411820782e-06, "loss": 2.2611, "step": 33034 }, { "epoch": 6.29, "learning_rate": 1.113346043851287e-06, "loss": 0.9267, "step": 33035 }, { "epoch": 6.29, "learning_rate": 1.1132888465204956e-06, "loss": 1.9473, "step": 33036 }, { "epoch": 6.29, "learning_rate": 1.1132316491897046e-06, "loss": 0.7375, "step": 33037 }, { "epoch": 6.29, "learning_rate": 1.1131744518589134e-06, "loss": 1.0588, "step": 33038 }, { "epoch": 6.29, "learning_rate": 1.1131172545281222e-06, "loss": 1.183, "step": 33039 }, { "epoch": 6.29, "learning_rate": 1.1130600571973308e-06, "loss": 0.7725, "step": 33040 }, { "epoch": 6.29, "learning_rate": 1.1130028598665396e-06, "loss": 2.0308, "step": 33041 }, { "epoch": 6.29, "learning_rate": 1.1129456625357483e-06, "loss": 1.1422, "step": 33042 }, { "epoch": 6.29, "learning_rate": 1.1128884652049571e-06, "loss": 2.196, "step": 33043 }, { "epoch": 6.29, "learning_rate": 1.112831267874166e-06, "loss": 1.8078, "step": 33044 }, { "epoch": 6.29, "learning_rate": 1.1127740705433747e-06, "loss": 2.8566, "step": 33045 }, { "epoch": 6.29, "learning_rate": 1.1127168732125833e-06, "loss": 2.2057, "step": 33046 }, { "epoch": 6.29, "learning_rate": 1.1126596758817921e-06, "loss": 2.1911, "step": 33047 }, { "epoch": 6.29, "learning_rate": 1.112602478551001e-06, "loss": 1.2197, "step": 33048 }, { "epoch": 6.3, "learning_rate": 1.11254528122021e-06, "loss": 1.9739, "step": 33049 }, { "epoch": 6.3, "learning_rate": 1.1124880838894185e-06, "loss": 1.1564, "step": 33050 }, { "epoch": 6.3, "learning_rate": 1.1124308865586273e-06, "loss": 1.2914, "step": 33051 }, { "epoch": 6.3, "learning_rate": 1.112373689227836e-06, "loss": 1.6167, "step": 33052 }, { "epoch": 6.3, "learning_rate": 1.1123164918970447e-06, "loss": 1.0179, "step": 33053 }, { "epoch": 6.3, "learning_rate": 1.1122592945662537e-06, "loss": 1.4426, "step": 33054 }, { "epoch": 6.3, "learning_rate": 1.1122020972354625e-06, "loss": 1.4647, "step": 33055 }, { "epoch": 6.3, "learning_rate": 1.1121448999046713e-06, "loss": 1.4592, "step": 33056 }, { "epoch": 6.3, "learning_rate": 1.1120877025738798e-06, "loss": 1.4986, "step": 33057 }, { "epoch": 6.3, "learning_rate": 1.1120305052430886e-06, "loss": 2.1275, "step": 33058 }, { "epoch": 6.3, "learning_rate": 1.1119733079122974e-06, "loss": 1.652, "step": 33059 }, { "epoch": 6.3, "learning_rate": 1.1119161105815062e-06, "loss": 2.9082, "step": 33060 }, { "epoch": 6.3, "learning_rate": 1.111858913250715e-06, "loss": 2.4518, "step": 33061 }, { "epoch": 6.3, "learning_rate": 1.1118017159199238e-06, "loss": 0.9351, "step": 33062 }, { "epoch": 6.3, "learning_rate": 1.1117445185891324e-06, "loss": 1.9039, "step": 33063 }, { "epoch": 6.3, "learning_rate": 1.1116873212583412e-06, "loss": 2.9487, "step": 33064 }, { "epoch": 6.3, "learning_rate": 1.1116301239275502e-06, "loss": 2.6404, "step": 33065 }, { "epoch": 6.3, "learning_rate": 1.111572926596759e-06, "loss": 1.7708, "step": 33066 }, { "epoch": 6.3, "learning_rate": 1.1115157292659676e-06, "loss": 1.9366, "step": 33067 }, { "epoch": 6.3, "learning_rate": 1.1114585319351764e-06, "loss": 1.2535, "step": 33068 }, { "epoch": 6.3, "learning_rate": 1.1114013346043851e-06, "loss": 1.4545, "step": 33069 }, { "epoch": 6.3, "learning_rate": 1.111344137273594e-06, "loss": 2.0526, "step": 33070 }, { "epoch": 6.3, "learning_rate": 1.1112869399428027e-06, "loss": 1.8161, "step": 33071 }, { "epoch": 6.3, "learning_rate": 1.1112297426120115e-06, "loss": 0.6723, "step": 33072 }, { "epoch": 6.3, "learning_rate": 1.1111725452812203e-06, "loss": 0.4942, "step": 33073 }, { "epoch": 6.3, "learning_rate": 1.111115347950429e-06, "loss": 1.1936, "step": 33074 }, { "epoch": 6.3, "learning_rate": 1.1110581506196377e-06, "loss": 1.6835, "step": 33075 }, { "epoch": 6.3, "learning_rate": 1.1110009532888467e-06, "loss": 2.0409, "step": 33076 }, { "epoch": 6.3, "learning_rate": 1.1109437559580553e-06, "loss": 1.1356, "step": 33077 }, { "epoch": 6.3, "learning_rate": 1.110886558627264e-06, "loss": 2.4111, "step": 33078 }, { "epoch": 6.3, "learning_rate": 1.1108293612964729e-06, "loss": 2.3728, "step": 33079 }, { "epoch": 6.3, "learning_rate": 1.1107721639656817e-06, "loss": 2.4102, "step": 33080 }, { "epoch": 6.3, "learning_rate": 1.1107149666348905e-06, "loss": 1.2951, "step": 33081 }, { "epoch": 6.3, "learning_rate": 1.1106577693040993e-06, "loss": 1.8805, "step": 33082 }, { "epoch": 6.3, "learning_rate": 1.110600571973308e-06, "loss": 1.9404, "step": 33083 }, { "epoch": 6.3, "learning_rate": 1.1105433746425166e-06, "loss": 0.6042, "step": 33084 }, { "epoch": 6.3, "learning_rate": 1.1104861773117254e-06, "loss": 1.4781, "step": 33085 }, { "epoch": 6.3, "learning_rate": 1.1104289799809342e-06, "loss": 1.7162, "step": 33086 }, { "epoch": 6.3, "learning_rate": 1.110371782650143e-06, "loss": 3.1379, "step": 33087 }, { "epoch": 6.3, "learning_rate": 1.1103145853193518e-06, "loss": 1.0165, "step": 33088 }, { "epoch": 6.3, "learning_rate": 1.1102573879885606e-06, "loss": 0.4304, "step": 33089 }, { "epoch": 6.3, "learning_rate": 1.1102001906577694e-06, "loss": 2.1507, "step": 33090 }, { "epoch": 6.3, "learning_rate": 1.110142993326978e-06, "loss": 1.448, "step": 33091 }, { "epoch": 6.3, "learning_rate": 1.1100857959961868e-06, "loss": 0.8907, "step": 33092 }, { "epoch": 6.3, "learning_rate": 1.1100285986653958e-06, "loss": 1.3632, "step": 33093 }, { "epoch": 6.3, "learning_rate": 1.1099714013346044e-06, "loss": 2.0768, "step": 33094 }, { "epoch": 6.3, "learning_rate": 1.1099142040038132e-06, "loss": 2.0729, "step": 33095 }, { "epoch": 6.3, "learning_rate": 1.109857006673022e-06, "loss": 1.594, "step": 33096 }, { "epoch": 6.3, "learning_rate": 1.1097998093422307e-06, "loss": 2.2318, "step": 33097 }, { "epoch": 6.3, "learning_rate": 1.1097426120114395e-06, "loss": 1.0071, "step": 33098 }, { "epoch": 6.3, "learning_rate": 1.1096854146806483e-06, "loss": 2.2303, "step": 33099 }, { "epoch": 6.3, "learning_rate": 1.1096282173498571e-06, "loss": 3.0069, "step": 33100 }, { "epoch": 6.3, "learning_rate": 1.1095710200190657e-06, "loss": 1.7676, "step": 33101 }, { "epoch": 6.31, "learning_rate": 1.1095138226882745e-06, "loss": 2.3359, "step": 33102 }, { "epoch": 6.31, "learning_rate": 1.1094566253574833e-06, "loss": 2.6525, "step": 33103 }, { "epoch": 6.31, "learning_rate": 1.1093994280266923e-06, "loss": 2.3234, "step": 33104 }, { "epoch": 6.31, "learning_rate": 1.1093422306959009e-06, "loss": 1.3665, "step": 33105 }, { "epoch": 6.31, "learning_rate": 1.1092850333651097e-06, "loss": 0.9201, "step": 33106 }, { "epoch": 6.31, "learning_rate": 1.1092278360343185e-06, "loss": 2.2, "step": 33107 }, { "epoch": 6.31, "learning_rate": 1.109170638703527e-06, "loss": 0.7541, "step": 33108 }, { "epoch": 6.31, "learning_rate": 1.109113441372736e-06, "loss": 0.7944, "step": 33109 }, { "epoch": 6.31, "learning_rate": 1.1090562440419448e-06, "loss": 1.3851, "step": 33110 }, { "epoch": 6.31, "learning_rate": 1.1089990467111534e-06, "loss": 0.984, "step": 33111 }, { "epoch": 6.31, "learning_rate": 1.1089418493803622e-06, "loss": 2.2527, "step": 33112 }, { "epoch": 6.31, "learning_rate": 1.108884652049571e-06, "loss": 2.5857, "step": 33113 }, { "epoch": 6.31, "learning_rate": 1.1088274547187798e-06, "loss": 1.2078, "step": 33114 }, { "epoch": 6.31, "learning_rate": 1.1087702573879886e-06, "loss": 2.6254, "step": 33115 }, { "epoch": 6.31, "learning_rate": 1.1087130600571974e-06, "loss": 1.2602, "step": 33116 }, { "epoch": 6.31, "learning_rate": 1.1086558627264062e-06, "loss": 2.2511, "step": 33117 }, { "epoch": 6.31, "learning_rate": 1.1085986653956148e-06, "loss": 1.4556, "step": 33118 }, { "epoch": 6.31, "learning_rate": 1.1085414680648236e-06, "loss": 1.944, "step": 33119 }, { "epoch": 6.31, "learning_rate": 1.1084842707340326e-06, "loss": 2.5762, "step": 33120 }, { "epoch": 6.31, "learning_rate": 1.1084270734032414e-06, "loss": 2.3678, "step": 33121 }, { "epoch": 6.31, "learning_rate": 1.10836987607245e-06, "loss": 3.037, "step": 33122 }, { "epoch": 6.31, "learning_rate": 1.1083126787416587e-06, "loss": 4.2039, "step": 33123 }, { "epoch": 6.31, "learning_rate": 1.1082554814108675e-06, "loss": 2.3036, "step": 33124 }, { "epoch": 6.31, "learning_rate": 1.1081982840800763e-06, "loss": 1.4388, "step": 33125 }, { "epoch": 6.31, "learning_rate": 1.1081410867492851e-06, "loss": 1.5763, "step": 33126 }, { "epoch": 6.31, "learning_rate": 1.108083889418494e-06, "loss": 1.7895, "step": 33127 }, { "epoch": 6.31, "learning_rate": 1.1080266920877025e-06, "loss": 1.3884, "step": 33128 }, { "epoch": 6.31, "learning_rate": 1.1079694947569113e-06, "loss": 2.3507, "step": 33129 }, { "epoch": 6.31, "learning_rate": 1.10791229742612e-06, "loss": 1.4726, "step": 33130 }, { "epoch": 6.31, "learning_rate": 1.1078551000953289e-06, "loss": 1.0665, "step": 33131 }, { "epoch": 6.31, "learning_rate": 1.1077979027645377e-06, "loss": 1.8377, "step": 33132 }, { "epoch": 6.31, "learning_rate": 1.1077407054337465e-06, "loss": 3.0274, "step": 33133 }, { "epoch": 6.31, "learning_rate": 1.1076835081029553e-06, "loss": 0.9777, "step": 33134 }, { "epoch": 6.31, "learning_rate": 1.1076263107721638e-06, "loss": 0.8527, "step": 33135 }, { "epoch": 6.31, "learning_rate": 1.1075691134413726e-06, "loss": 2.6271, "step": 33136 }, { "epoch": 6.31, "learning_rate": 1.1075119161105816e-06, "loss": 0.6336, "step": 33137 }, { "epoch": 6.31, "learning_rate": 1.1074547187797904e-06, "loss": 1.0597, "step": 33138 }, { "epoch": 6.31, "learning_rate": 1.107397521448999e-06, "loss": 1.6448, "step": 33139 }, { "epoch": 6.31, "learning_rate": 1.1073403241182078e-06, "loss": 1.1587, "step": 33140 }, { "epoch": 6.31, "learning_rate": 1.1072831267874166e-06, "loss": 2.6816, "step": 33141 }, { "epoch": 6.31, "learning_rate": 1.1072259294566254e-06, "loss": 2.7052, "step": 33142 }, { "epoch": 6.31, "learning_rate": 1.1071687321258342e-06, "loss": 2.7221, "step": 33143 }, { "epoch": 6.31, "learning_rate": 1.107111534795043e-06, "loss": 2.0964, "step": 33144 }, { "epoch": 6.31, "learning_rate": 1.1070543374642518e-06, "loss": 0.9147, "step": 33145 }, { "epoch": 6.31, "learning_rate": 1.1069971401334604e-06, "loss": 1.6719, "step": 33146 }, { "epoch": 6.31, "learning_rate": 1.1069399428026692e-06, "loss": 1.9195, "step": 33147 }, { "epoch": 6.31, "learning_rate": 1.1068827454718782e-06, "loss": 1.1376, "step": 33148 }, { "epoch": 6.31, "learning_rate": 1.1068255481410867e-06, "loss": 0.6791, "step": 33149 }, { "epoch": 6.31, "learning_rate": 1.1067683508102955e-06, "loss": 2.791, "step": 33150 }, { "epoch": 6.31, "learning_rate": 1.1067111534795043e-06, "loss": 1.0074, "step": 33151 }, { "epoch": 6.31, "learning_rate": 1.106653956148713e-06, "loss": 1.2351, "step": 33152 }, { "epoch": 6.31, "learning_rate": 1.106596758817922e-06, "loss": 1.8721, "step": 33153 }, { "epoch": 6.32, "learning_rate": 1.1065395614871307e-06, "loss": 1.7005, "step": 33154 }, { "epoch": 6.32, "learning_rate": 1.1064823641563395e-06, "loss": 1.9762, "step": 33155 }, { "epoch": 6.32, "learning_rate": 1.106425166825548e-06, "loss": 2.3506, "step": 33156 }, { "epoch": 6.32, "learning_rate": 1.1063679694947569e-06, "loss": 1.4657, "step": 33157 }, { "epoch": 6.32, "learning_rate": 1.1063107721639657e-06, "loss": 1.4982, "step": 33158 }, { "epoch": 6.32, "learning_rate": 1.1062535748331745e-06, "loss": 1.1309, "step": 33159 }, { "epoch": 6.32, "learning_rate": 1.1061963775023833e-06, "loss": 0.7436, "step": 33160 }, { "epoch": 6.32, "learning_rate": 1.106139180171592e-06, "loss": 2.3658, "step": 33161 }, { "epoch": 6.32, "learning_rate": 1.1060819828408009e-06, "loss": 2.2374, "step": 33162 }, { "epoch": 6.32, "learning_rate": 1.1060247855100094e-06, "loss": 1.962, "step": 33163 }, { "epoch": 6.32, "learning_rate": 1.1059675881792182e-06, "loss": 1.6233, "step": 33164 }, { "epoch": 6.32, "learning_rate": 1.1059103908484272e-06, "loss": 2.6143, "step": 33165 }, { "epoch": 6.32, "learning_rate": 1.1058531935176358e-06, "loss": 0.9219, "step": 33166 }, { "epoch": 6.32, "learning_rate": 1.1057959961868446e-06, "loss": 1.6127, "step": 33167 }, { "epoch": 6.32, "learning_rate": 1.1057387988560534e-06, "loss": 1.0255, "step": 33168 }, { "epoch": 6.32, "learning_rate": 1.1056816015252622e-06, "loss": 1.5182, "step": 33169 }, { "epoch": 6.32, "learning_rate": 1.105624404194471e-06, "loss": 2.0592, "step": 33170 }, { "epoch": 6.32, "learning_rate": 1.1055672068636798e-06, "loss": 1.4535, "step": 33171 }, { "epoch": 6.32, "learning_rate": 1.1055100095328886e-06, "loss": 2.0059, "step": 33172 }, { "epoch": 6.32, "learning_rate": 1.1054528122020972e-06, "loss": 0.8209, "step": 33173 }, { "epoch": 6.32, "learning_rate": 1.105395614871306e-06, "loss": 1.9449, "step": 33174 }, { "epoch": 6.32, "learning_rate": 1.1053384175405148e-06, "loss": 2.7834, "step": 33175 }, { "epoch": 6.32, "learning_rate": 1.1052812202097235e-06, "loss": 1.5895, "step": 33176 }, { "epoch": 6.32, "learning_rate": 1.1052240228789323e-06, "loss": 1.3551, "step": 33177 }, { "epoch": 6.32, "learning_rate": 1.1051668255481411e-06, "loss": 1.7204, "step": 33178 }, { "epoch": 6.32, "learning_rate": 1.10510962821735e-06, "loss": 2.9502, "step": 33179 }, { "epoch": 6.32, "learning_rate": 1.1050524308865585e-06, "loss": 2.137, "step": 33180 }, { "epoch": 6.32, "learning_rate": 1.1049952335557675e-06, "loss": 2.8106, "step": 33181 }, { "epoch": 6.32, "learning_rate": 1.1049380362249763e-06, "loss": 2.6858, "step": 33182 }, { "epoch": 6.32, "learning_rate": 1.1048808388941849e-06, "loss": 1.6706, "step": 33183 }, { "epoch": 6.32, "learning_rate": 1.1048236415633937e-06, "loss": 1.7481, "step": 33184 }, { "epoch": 6.32, "learning_rate": 1.1047664442326025e-06, "loss": 1.519, "step": 33185 }, { "epoch": 6.32, "learning_rate": 1.1047092469018113e-06, "loss": 2.5303, "step": 33186 }, { "epoch": 6.32, "learning_rate": 1.10465204957102e-06, "loss": 2.0339, "step": 33187 }, { "epoch": 6.32, "learning_rate": 1.1045948522402289e-06, "loss": 3.7068, "step": 33188 }, { "epoch": 6.32, "learning_rate": 1.1045376549094377e-06, "loss": 2.0289, "step": 33189 }, { "epoch": 6.32, "learning_rate": 1.1044804575786462e-06, "loss": 2.0336, "step": 33190 }, { "epoch": 6.32, "learning_rate": 1.104423260247855e-06, "loss": 1.8987, "step": 33191 }, { "epoch": 6.32, "learning_rate": 1.104366062917064e-06, "loss": 2.9639, "step": 33192 }, { "epoch": 6.32, "learning_rate": 1.1043088655862726e-06, "loss": 0.8323, "step": 33193 }, { "epoch": 6.32, "learning_rate": 1.1042516682554814e-06, "loss": 2.8619, "step": 33194 }, { "epoch": 6.32, "learning_rate": 1.1041944709246902e-06, "loss": 2.1736, "step": 33195 }, { "epoch": 6.32, "learning_rate": 1.104137273593899e-06, "loss": 0.9788, "step": 33196 }, { "epoch": 6.32, "learning_rate": 1.1040800762631078e-06, "loss": 2.5317, "step": 33197 }, { "epoch": 6.32, "learning_rate": 1.1040228789323166e-06, "loss": 1.2927, "step": 33198 }, { "epoch": 6.32, "learning_rate": 1.1039656816015254e-06, "loss": 2.4141, "step": 33199 }, { "epoch": 6.32, "learning_rate": 1.103908484270734e-06, "loss": 0.9341, "step": 33200 }, { "epoch": 6.32, "learning_rate": 1.1038512869399428e-06, "loss": 2.5046, "step": 33201 }, { "epoch": 6.32, "learning_rate": 1.1037940896091515e-06, "loss": 1.3034, "step": 33202 }, { "epoch": 6.32, "learning_rate": 1.1037368922783603e-06, "loss": 2.7174, "step": 33203 }, { "epoch": 6.32, "learning_rate": 1.1036796949475691e-06, "loss": 1.3788, "step": 33204 }, { "epoch": 6.32, "learning_rate": 1.103622497616778e-06, "loss": 2.1696, "step": 33205 }, { "epoch": 6.32, "learning_rate": 1.1035653002859867e-06, "loss": 2.0066, "step": 33206 }, { "epoch": 6.33, "learning_rate": 1.1035081029551953e-06, "loss": 1.8022, "step": 33207 }, { "epoch": 6.33, "learning_rate": 1.103450905624404e-06, "loss": 1.4495, "step": 33208 }, { "epoch": 6.33, "learning_rate": 1.103393708293613e-06, "loss": 0.7538, "step": 33209 }, { "epoch": 6.33, "learning_rate": 1.103336510962822e-06, "loss": 0.6913, "step": 33210 }, { "epoch": 6.33, "learning_rate": 1.1032793136320305e-06, "loss": 1.6354, "step": 33211 }, { "epoch": 6.33, "learning_rate": 1.1032221163012393e-06, "loss": 2.2841, "step": 33212 }, { "epoch": 6.33, "learning_rate": 1.103164918970448e-06, "loss": 2.0873, "step": 33213 }, { "epoch": 6.33, "learning_rate": 1.1031077216396569e-06, "loss": 1.8266, "step": 33214 }, { "epoch": 6.33, "learning_rate": 1.1030505243088657e-06, "loss": 2.0061, "step": 33215 }, { "epoch": 6.33, "learning_rate": 1.1029933269780745e-06, "loss": 0.8612, "step": 33216 }, { "epoch": 6.33, "learning_rate": 1.102936129647283e-06, "loss": 2.0208, "step": 33217 }, { "epoch": 6.33, "learning_rate": 1.1028789323164918e-06, "loss": 0.9128, "step": 33218 }, { "epoch": 6.33, "learning_rate": 1.1028217349857006e-06, "loss": 2.5194, "step": 33219 }, { "epoch": 6.33, "learning_rate": 1.1027645376549096e-06, "loss": 1.31, "step": 33220 }, { "epoch": 6.33, "learning_rate": 1.1027073403241182e-06, "loss": 1.6346, "step": 33221 }, { "epoch": 6.33, "learning_rate": 1.102650142993327e-06, "loss": 0.8705, "step": 33222 }, { "epoch": 6.33, "learning_rate": 1.1025929456625358e-06, "loss": 1.7337, "step": 33223 }, { "epoch": 6.33, "learning_rate": 1.1025357483317444e-06, "loss": 3.1788, "step": 33224 }, { "epoch": 6.33, "learning_rate": 1.1024785510009534e-06, "loss": 1.9928, "step": 33225 }, { "epoch": 6.33, "learning_rate": 1.1024213536701622e-06, "loss": 2.2923, "step": 33226 }, { "epoch": 6.33, "learning_rate": 1.102364156339371e-06, "loss": 1.5645, "step": 33227 }, { "epoch": 6.33, "learning_rate": 1.1023069590085796e-06, "loss": 2.0175, "step": 33228 }, { "epoch": 6.33, "learning_rate": 1.1022497616777883e-06, "loss": 0.7401, "step": 33229 }, { "epoch": 6.33, "learning_rate": 1.1021925643469971e-06, "loss": 1.7743, "step": 33230 }, { "epoch": 6.33, "learning_rate": 1.102135367016206e-06, "loss": 0.8079, "step": 33231 }, { "epoch": 6.33, "learning_rate": 1.1020781696854147e-06, "loss": 3.0398, "step": 33232 }, { "epoch": 6.33, "learning_rate": 1.1020209723546235e-06, "loss": 1.6964, "step": 33233 }, { "epoch": 6.33, "learning_rate": 1.1019637750238323e-06, "loss": 1.6814, "step": 33234 }, { "epoch": 6.33, "learning_rate": 1.101906577693041e-06, "loss": 1.4401, "step": 33235 }, { "epoch": 6.33, "learning_rate": 1.10184938036225e-06, "loss": 1.4011, "step": 33236 }, { "epoch": 6.33, "learning_rate": 1.1017921830314587e-06, "loss": 1.2627, "step": 33237 }, { "epoch": 6.33, "learning_rate": 1.1017349857006673e-06, "loss": 2.9485, "step": 33238 }, { "epoch": 6.33, "learning_rate": 1.101677788369876e-06, "loss": 1.4998, "step": 33239 }, { "epoch": 6.33, "learning_rate": 1.1016205910390849e-06, "loss": 3.1734, "step": 33240 }, { "epoch": 6.33, "learning_rate": 1.1015633937082937e-06, "loss": 0.9254, "step": 33241 }, { "epoch": 6.33, "learning_rate": 1.1015061963775025e-06, "loss": 1.6175, "step": 33242 }, { "epoch": 6.33, "learning_rate": 1.1014489990467112e-06, "loss": 0.8537, "step": 33243 }, { "epoch": 6.33, "learning_rate": 1.10139180171592e-06, "loss": 3.8844, "step": 33244 }, { "epoch": 6.33, "learning_rate": 1.1013346043851286e-06, "loss": 0.8443, "step": 33245 }, { "epoch": 6.33, "learning_rate": 1.1012774070543374e-06, "loss": 1.4143, "step": 33246 }, { "epoch": 6.33, "learning_rate": 1.1012202097235462e-06, "loss": 2.0117, "step": 33247 }, { "epoch": 6.33, "learning_rate": 1.101163012392755e-06, "loss": 1.272, "step": 33248 }, { "epoch": 6.33, "learning_rate": 1.1011058150619638e-06, "loss": 1.9642, "step": 33249 }, { "epoch": 6.33, "learning_rate": 1.1010486177311726e-06, "loss": 2.129, "step": 33250 }, { "epoch": 6.33, "learning_rate": 1.1009914204003814e-06, "loss": 2.3204, "step": 33251 }, { "epoch": 6.33, "learning_rate": 1.10093422306959e-06, "loss": 1.8996, "step": 33252 }, { "epoch": 6.33, "learning_rate": 1.100877025738799e-06, "loss": 2.177, "step": 33253 }, { "epoch": 6.33, "learning_rate": 1.1008198284080078e-06, "loss": 1.5893, "step": 33254 }, { "epoch": 6.33, "learning_rate": 1.1007626310772164e-06, "loss": 1.5569, "step": 33255 }, { "epoch": 6.33, "learning_rate": 1.1007054337464251e-06, "loss": 1.6887, "step": 33256 }, { "epoch": 6.33, "learning_rate": 1.100648236415634e-06, "loss": 0.711, "step": 33257 }, { "epoch": 6.33, "learning_rate": 1.1005910390848427e-06, "loss": 1.7709, "step": 33258 }, { "epoch": 6.34, "learning_rate": 1.1005338417540515e-06, "loss": 1.7229, "step": 33259 }, { "epoch": 6.34, "learning_rate": 1.1004766444232603e-06, "loss": 0.8628, "step": 33260 }, { "epoch": 6.34, "learning_rate": 1.1004194470924691e-06, "loss": 0.637, "step": 33261 }, { "epoch": 6.34, "learning_rate": 1.1003622497616777e-06, "loss": 1.8518, "step": 33262 }, { "epoch": 6.34, "learning_rate": 1.1003050524308865e-06, "loss": 1.9943, "step": 33263 }, { "epoch": 6.34, "learning_rate": 1.1002478551000955e-06, "loss": 1.9443, "step": 33264 }, { "epoch": 6.34, "learning_rate": 1.100190657769304e-06, "loss": 1.4685, "step": 33265 }, { "epoch": 6.34, "learning_rate": 1.1001334604385129e-06, "loss": 2.6111, "step": 33266 }, { "epoch": 6.34, "learning_rate": 1.1000762631077217e-06, "loss": 2.5244, "step": 33267 }, { "epoch": 6.34, "learning_rate": 1.1000190657769305e-06, "loss": 1.9192, "step": 33268 }, { "epoch": 6.34, "learning_rate": 1.0999618684461393e-06, "loss": 2.1751, "step": 33269 }, { "epoch": 6.34, "learning_rate": 1.099904671115348e-06, "loss": 2.0282, "step": 33270 }, { "epoch": 6.34, "learning_rate": 1.0998474737845568e-06, "loss": 2.3931, "step": 33271 }, { "epoch": 6.34, "learning_rate": 1.0997902764537654e-06, "loss": 1.4183, "step": 33272 }, { "epoch": 6.34, "learning_rate": 1.0997330791229742e-06, "loss": 1.3274, "step": 33273 }, { "epoch": 6.34, "learning_rate": 1.099675881792183e-06, "loss": 2.6111, "step": 33274 }, { "epoch": 6.34, "learning_rate": 1.099618684461392e-06, "loss": 2.4238, "step": 33275 }, { "epoch": 6.34, "learning_rate": 1.0995614871306006e-06, "loss": 2.299, "step": 33276 }, { "epoch": 6.34, "learning_rate": 1.0995042897998094e-06, "loss": 1.3615, "step": 33277 }, { "epoch": 6.34, "learning_rate": 1.0994470924690182e-06, "loss": 2.3921, "step": 33278 }, { "epoch": 6.34, "learning_rate": 1.0993898951382268e-06, "loss": 1.674, "step": 33279 }, { "epoch": 6.34, "learning_rate": 1.0993326978074358e-06, "loss": 1.3146, "step": 33280 }, { "epoch": 6.34, "learning_rate": 1.0992755004766446e-06, "loss": 2.551, "step": 33281 }, { "epoch": 6.34, "learning_rate": 1.0992183031458531e-06, "loss": 2.8622, "step": 33282 }, { "epoch": 6.34, "learning_rate": 1.099161105815062e-06, "loss": 1.1045, "step": 33283 }, { "epoch": 6.34, "learning_rate": 1.0991039084842707e-06, "loss": 1.4196, "step": 33284 }, { "epoch": 6.34, "learning_rate": 1.0990467111534795e-06, "loss": 1.3787, "step": 33285 }, { "epoch": 6.34, "learning_rate": 1.0989895138226883e-06, "loss": 2.7283, "step": 33286 }, { "epoch": 6.34, "learning_rate": 1.0989323164918971e-06, "loss": 1.8854, "step": 33287 }, { "epoch": 6.34, "learning_rate": 1.098875119161106e-06, "loss": 1.4172, "step": 33288 }, { "epoch": 6.34, "learning_rate": 1.0988179218303145e-06, "loss": 1.4378, "step": 33289 }, { "epoch": 6.34, "learning_rate": 1.0987607244995233e-06, "loss": 2.889, "step": 33290 }, { "epoch": 6.34, "learning_rate": 1.098703527168732e-06, "loss": 2.1236, "step": 33291 }, { "epoch": 6.34, "learning_rate": 1.098646329837941e-06, "loss": 2.002, "step": 33292 }, { "epoch": 6.34, "learning_rate": 1.0985891325071497e-06, "loss": 2.1576, "step": 33293 }, { "epoch": 6.34, "learning_rate": 1.0985319351763585e-06, "loss": 3.4966, "step": 33294 }, { "epoch": 6.34, "learning_rate": 1.0984747378455673e-06, "loss": 1.743, "step": 33295 }, { "epoch": 6.34, "learning_rate": 1.0984175405147758e-06, "loss": 1.9812, "step": 33296 }, { "epoch": 6.34, "learning_rate": 1.0983603431839848e-06, "loss": 3.1577, "step": 33297 }, { "epoch": 6.34, "learning_rate": 1.0983031458531936e-06, "loss": 1.0362, "step": 33298 }, { "epoch": 6.34, "learning_rate": 1.0982459485224024e-06, "loss": 1.9217, "step": 33299 }, { "epoch": 6.34, "learning_rate": 1.098188751191611e-06, "loss": 1.5783, "step": 33300 }, { "epoch": 6.34, "learning_rate": 1.0981315538608198e-06, "loss": 2.27, "step": 33301 }, { "epoch": 6.34, "learning_rate": 1.0980743565300286e-06, "loss": 1.9116, "step": 33302 }, { "epoch": 6.34, "learning_rate": 1.0980171591992374e-06, "loss": 1.6866, "step": 33303 }, { "epoch": 6.34, "learning_rate": 1.0979599618684462e-06, "loss": 1.2191, "step": 33304 }, { "epoch": 6.34, "learning_rate": 1.097902764537655e-06, "loss": 2.606, "step": 33305 }, { "epoch": 6.34, "learning_rate": 1.0978455672068636e-06, "loss": 1.7362, "step": 33306 }, { "epoch": 6.34, "learning_rate": 1.0977883698760724e-06, "loss": 0.7034, "step": 33307 }, { "epoch": 6.34, "learning_rate": 1.0977311725452814e-06, "loss": 2.3777, "step": 33308 }, { "epoch": 6.34, "learning_rate": 1.0976739752144902e-06, "loss": 0.8699, "step": 33309 }, { "epoch": 6.34, "learning_rate": 1.0976167778836987e-06, "loss": 1.544, "step": 33310 }, { "epoch": 6.34, "learning_rate": 1.0975595805529075e-06, "loss": 2.8394, "step": 33311 }, { "epoch": 6.35, "learning_rate": 1.0975023832221163e-06, "loss": 1.616, "step": 33312 }, { "epoch": 6.35, "learning_rate": 1.0974451858913251e-06, "loss": 1.4048, "step": 33313 }, { "epoch": 6.35, "learning_rate": 1.097387988560534e-06, "loss": 1.0975, "step": 33314 }, { "epoch": 6.35, "learning_rate": 1.0973307912297427e-06, "loss": 2.4807, "step": 33315 }, { "epoch": 6.35, "learning_rate": 1.0972735938989515e-06, "loss": 1.9654, "step": 33316 }, { "epoch": 6.35, "learning_rate": 1.09721639656816e-06, "loss": 1.2501, "step": 33317 }, { "epoch": 6.35, "learning_rate": 1.0971591992373689e-06, "loss": 1.1904, "step": 33318 }, { "epoch": 6.35, "learning_rate": 1.0971020019065779e-06, "loss": 0.7609, "step": 33319 }, { "epoch": 6.35, "learning_rate": 1.0970448045757865e-06, "loss": 2.8297, "step": 33320 }, { "epoch": 6.35, "learning_rate": 1.0969876072449953e-06, "loss": 0.8319, "step": 33321 }, { "epoch": 6.35, "learning_rate": 1.096930409914204e-06, "loss": 1.2972, "step": 33322 }, { "epoch": 6.35, "learning_rate": 1.0968732125834126e-06, "loss": 1.713, "step": 33323 }, { "epoch": 6.35, "learning_rate": 1.0968160152526216e-06, "loss": 1.1181, "step": 33324 }, { "epoch": 6.35, "learning_rate": 1.0967588179218304e-06, "loss": 1.3794, "step": 33325 }, { "epoch": 6.35, "learning_rate": 1.0967016205910392e-06, "loss": 2.1884, "step": 33326 }, { "epoch": 6.35, "learning_rate": 1.0966444232602478e-06, "loss": 2.7194, "step": 33327 }, { "epoch": 6.35, "learning_rate": 1.0965872259294566e-06, "loss": 1.916, "step": 33328 }, { "epoch": 6.35, "learning_rate": 1.0965300285986654e-06, "loss": 0.9825, "step": 33329 }, { "epoch": 6.35, "learning_rate": 1.0964728312678742e-06, "loss": 1.5324, "step": 33330 }, { "epoch": 6.35, "learning_rate": 1.096415633937083e-06, "loss": 2.2078, "step": 33331 }, { "epoch": 6.35, "learning_rate": 1.0963584366062918e-06, "loss": 1.0203, "step": 33332 }, { "epoch": 6.35, "learning_rate": 1.0963012392755006e-06, "loss": 1.31, "step": 33333 }, { "epoch": 6.35, "learning_rate": 1.0962440419447092e-06, "loss": 1.8186, "step": 33334 }, { "epoch": 6.35, "learning_rate": 1.096186844613918e-06, "loss": 1.6048, "step": 33335 }, { "epoch": 6.35, "learning_rate": 1.096129647283127e-06, "loss": 3.6144, "step": 33336 }, { "epoch": 6.35, "learning_rate": 1.0960724499523355e-06, "loss": 2.2241, "step": 33337 }, { "epoch": 6.35, "learning_rate": 1.0960152526215443e-06, "loss": 2.6397, "step": 33338 }, { "epoch": 6.35, "learning_rate": 1.0959580552907531e-06, "loss": 1.2978, "step": 33339 }, { "epoch": 6.35, "learning_rate": 1.095900857959962e-06, "loss": 3.748, "step": 33340 }, { "epoch": 6.35, "learning_rate": 1.0958436606291707e-06, "loss": 2.0247, "step": 33341 }, { "epoch": 6.35, "learning_rate": 1.0957864632983795e-06, "loss": 1.2765, "step": 33342 }, { "epoch": 6.35, "learning_rate": 1.0957292659675883e-06, "loss": 1.1633, "step": 33343 }, { "epoch": 6.35, "learning_rate": 1.0956720686367969e-06, "loss": 0.9032, "step": 33344 }, { "epoch": 6.35, "learning_rate": 1.0956148713060057e-06, "loss": 2.4178, "step": 33345 }, { "epoch": 6.35, "learning_rate": 1.0955576739752145e-06, "loss": 2.1897, "step": 33346 }, { "epoch": 6.35, "learning_rate": 1.0955004766444233e-06, "loss": 1.479, "step": 33347 }, { "epoch": 6.35, "learning_rate": 1.095443279313632e-06, "loss": 1.5674, "step": 33348 }, { "epoch": 6.35, "learning_rate": 1.0953860819828409e-06, "loss": 2.6834, "step": 33349 }, { "epoch": 6.35, "learning_rate": 1.0953288846520496e-06, "loss": 1.1962, "step": 33350 }, { "epoch": 6.35, "learning_rate": 1.0952716873212582e-06, "loss": 1.4223, "step": 33351 }, { "epoch": 6.35, "learning_rate": 1.0952144899904672e-06, "loss": 0.6892, "step": 33352 }, { "epoch": 6.35, "learning_rate": 1.095157292659676e-06, "loss": 1.6521, "step": 33353 }, { "epoch": 6.35, "learning_rate": 1.0951000953288846e-06, "loss": 1.9349, "step": 33354 }, { "epoch": 6.35, "learning_rate": 1.0950428979980934e-06, "loss": 1.8474, "step": 33355 }, { "epoch": 6.35, "learning_rate": 1.0949857006673022e-06, "loss": 1.5279, "step": 33356 }, { "epoch": 6.35, "learning_rate": 1.094928503336511e-06, "loss": 1.7107, "step": 33357 }, { "epoch": 6.35, "learning_rate": 1.0948713060057198e-06, "loss": 2.3461, "step": 33358 }, { "epoch": 6.35, "learning_rate": 1.0948141086749286e-06, "loss": 2.2317, "step": 33359 }, { "epoch": 6.35, "learning_rate": 1.0947569113441374e-06, "loss": 1.8325, "step": 33360 }, { "epoch": 6.35, "learning_rate": 1.094699714013346e-06, "loss": 2.3608, "step": 33361 }, { "epoch": 6.35, "learning_rate": 1.0946425166825547e-06, "loss": 2.1246, "step": 33362 }, { "epoch": 6.35, "learning_rate": 1.0945853193517635e-06, "loss": 1.9327, "step": 33363 }, { "epoch": 6.36, "learning_rate": 1.0945281220209723e-06, "loss": 1.2236, "step": 33364 }, { "epoch": 6.36, "learning_rate": 1.0944709246901811e-06, "loss": 2.59, "step": 33365 }, { "epoch": 6.36, "learning_rate": 1.09441372735939e-06, "loss": 0.966, "step": 33366 }, { "epoch": 6.36, "learning_rate": 1.0943565300285987e-06, "loss": 1.9504, "step": 33367 }, { "epoch": 6.36, "learning_rate": 1.0942993326978075e-06, "loss": 1.8426, "step": 33368 }, { "epoch": 6.36, "learning_rate": 1.0942421353670163e-06, "loss": 0.9636, "step": 33369 }, { "epoch": 6.36, "learning_rate": 1.094184938036225e-06, "loss": 2.5207, "step": 33370 }, { "epoch": 6.36, "learning_rate": 1.0941277407054337e-06, "loss": 1.1875, "step": 33371 }, { "epoch": 6.36, "learning_rate": 1.0940705433746425e-06, "loss": 1.739, "step": 33372 }, { "epoch": 6.36, "learning_rate": 1.0940133460438513e-06, "loss": 2.7475, "step": 33373 }, { "epoch": 6.36, "learning_rate": 1.09395614871306e-06, "loss": 2.4517, "step": 33374 }, { "epoch": 6.36, "learning_rate": 1.0938989513822689e-06, "loss": 1.6012, "step": 33375 }, { "epoch": 6.36, "learning_rate": 1.0938417540514776e-06, "loss": 1.2609, "step": 33376 }, { "epoch": 6.36, "learning_rate": 1.0937845567206864e-06, "loss": 1.9098, "step": 33377 }, { "epoch": 6.36, "learning_rate": 1.093727359389895e-06, "loss": 2.3212, "step": 33378 }, { "epoch": 6.36, "learning_rate": 1.0936701620591038e-06, "loss": 1.9909, "step": 33379 }, { "epoch": 6.36, "learning_rate": 1.0936129647283128e-06, "loss": 2.3952, "step": 33380 }, { "epoch": 6.36, "learning_rate": 1.0935557673975216e-06, "loss": 0.9204, "step": 33381 }, { "epoch": 6.36, "learning_rate": 1.0934985700667302e-06, "loss": 0.2874, "step": 33382 }, { "epoch": 6.36, "learning_rate": 1.093441372735939e-06, "loss": 0.9829, "step": 33383 }, { "epoch": 6.36, "learning_rate": 1.0933841754051478e-06, "loss": 2.5661, "step": 33384 }, { "epoch": 6.36, "learning_rate": 1.0933269780743566e-06, "loss": 1.7452, "step": 33385 }, { "epoch": 6.36, "learning_rate": 1.0932697807435654e-06, "loss": 3.1011, "step": 33386 }, { "epoch": 6.36, "learning_rate": 1.0932125834127742e-06, "loss": 2.2971, "step": 33387 }, { "epoch": 6.36, "learning_rate": 1.0931553860819828e-06, "loss": 0.9268, "step": 33388 }, { "epoch": 6.36, "learning_rate": 1.0930981887511915e-06, "loss": 2.8938, "step": 33389 }, { "epoch": 6.36, "learning_rate": 1.0930409914204003e-06, "loss": 1.2808, "step": 33390 }, { "epoch": 6.36, "learning_rate": 1.0929837940896093e-06, "loss": 2.3168, "step": 33391 }, { "epoch": 6.36, "learning_rate": 1.092926596758818e-06, "loss": 1.0723, "step": 33392 }, { "epoch": 6.36, "learning_rate": 1.0928693994280267e-06, "loss": 1.4345, "step": 33393 }, { "epoch": 6.36, "learning_rate": 1.0928122020972355e-06, "loss": 2.3511, "step": 33394 }, { "epoch": 6.36, "learning_rate": 1.092755004766444e-06, "loss": 2.8383, "step": 33395 }, { "epoch": 6.36, "learning_rate": 1.092697807435653e-06, "loss": 2.1227, "step": 33396 }, { "epoch": 6.36, "learning_rate": 1.092640610104862e-06, "loss": 1.6624, "step": 33397 }, { "epoch": 6.36, "learning_rate": 1.0925834127740707e-06, "loss": 1.5063, "step": 33398 }, { "epoch": 6.36, "learning_rate": 1.0925262154432793e-06, "loss": 1.1893, "step": 33399 }, { "epoch": 6.36, "learning_rate": 1.092469018112488e-06, "loss": 2.1539, "step": 33400 }, { "epoch": 6.36, "learning_rate": 1.0924118207816969e-06, "loss": 1.4908, "step": 33401 }, { "epoch": 6.36, "learning_rate": 1.0923546234509057e-06, "loss": 1.9167, "step": 33402 }, { "epoch": 6.36, "learning_rate": 1.0922974261201144e-06, "loss": 2.5128, "step": 33403 }, { "epoch": 6.36, "learning_rate": 1.0922402287893232e-06, "loss": 0.8323, "step": 33404 }, { "epoch": 6.36, "learning_rate": 1.092183031458532e-06, "loss": 1.3282, "step": 33405 }, { "epoch": 6.36, "learning_rate": 1.0921258341277406e-06, "loss": 1.335, "step": 33406 }, { "epoch": 6.36, "learning_rate": 1.0920686367969494e-06, "loss": 1.498, "step": 33407 }, { "epoch": 6.36, "learning_rate": 1.0920114394661584e-06, "loss": 2.2852, "step": 33408 }, { "epoch": 6.36, "learning_rate": 1.091954242135367e-06, "loss": 2.3372, "step": 33409 }, { "epoch": 6.36, "learning_rate": 1.0918970448045758e-06, "loss": 0.4481, "step": 33410 }, { "epoch": 6.36, "learning_rate": 1.0918398474737846e-06, "loss": 2.6672, "step": 33411 }, { "epoch": 6.36, "learning_rate": 1.0917826501429934e-06, "loss": 1.4784, "step": 33412 }, { "epoch": 6.36, "learning_rate": 1.0917254528122022e-06, "loss": 3.0877, "step": 33413 }, { "epoch": 6.36, "learning_rate": 1.091668255481411e-06, "loss": 2.9245, "step": 33414 }, { "epoch": 6.36, "learning_rate": 1.0916110581506198e-06, "loss": 1.5454, "step": 33415 }, { "epoch": 6.36, "learning_rate": 1.0915538608198283e-06, "loss": 0.4685, "step": 33416 }, { "epoch": 6.37, "learning_rate": 1.0914966634890371e-06, "loss": 1.3448, "step": 33417 }, { "epoch": 6.37, "learning_rate": 1.091439466158246e-06, "loss": 2.0262, "step": 33418 }, { "epoch": 6.37, "learning_rate": 1.0913822688274547e-06, "loss": 1.9625, "step": 33419 }, { "epoch": 6.37, "learning_rate": 1.0913250714966635e-06, "loss": 1.3799, "step": 33420 }, { "epoch": 6.37, "learning_rate": 1.0912678741658723e-06, "loss": 1.536, "step": 33421 }, { "epoch": 6.37, "learning_rate": 1.0912106768350811e-06, "loss": 2.4608, "step": 33422 }, { "epoch": 6.37, "learning_rate": 1.0911534795042897e-06, "loss": 2.2058, "step": 33423 }, { "epoch": 6.37, "learning_rate": 1.0910962821734987e-06, "loss": 2.3863, "step": 33424 }, { "epoch": 6.37, "learning_rate": 1.0910390848427075e-06, "loss": 2.7937, "step": 33425 }, { "epoch": 6.37, "learning_rate": 1.090981887511916e-06, "loss": 1.1949, "step": 33426 }, { "epoch": 6.37, "learning_rate": 1.0909246901811249e-06, "loss": 1.6491, "step": 33427 }, { "epoch": 6.37, "learning_rate": 1.0908674928503337e-06, "loss": 1.4659, "step": 33428 }, { "epoch": 6.37, "learning_rate": 1.0908102955195425e-06, "loss": 3.3767, "step": 33429 }, { "epoch": 6.37, "learning_rate": 1.0907530981887512e-06, "loss": 2.7133, "step": 33430 }, { "epoch": 6.37, "learning_rate": 1.09069590085796e-06, "loss": 2.2416, "step": 33431 }, { "epoch": 6.37, "learning_rate": 1.0906387035271688e-06, "loss": 1.3826, "step": 33432 }, { "epoch": 6.37, "learning_rate": 1.0905815061963774e-06, "loss": 2.7537, "step": 33433 }, { "epoch": 6.37, "learning_rate": 1.0905243088655862e-06, "loss": 2.2392, "step": 33434 }, { "epoch": 6.37, "learning_rate": 1.0904671115347952e-06, "loss": 1.8913, "step": 33435 }, { "epoch": 6.37, "learning_rate": 1.0904099142040038e-06, "loss": 2.9861, "step": 33436 }, { "epoch": 6.37, "learning_rate": 1.0903527168732126e-06, "loss": 1.7065, "step": 33437 }, { "epoch": 6.37, "learning_rate": 1.0902955195424214e-06, "loss": 1.3733, "step": 33438 }, { "epoch": 6.37, "learning_rate": 1.0902383222116302e-06, "loss": 2.0513, "step": 33439 }, { "epoch": 6.37, "learning_rate": 1.090181124880839e-06, "loss": 2.221, "step": 33440 }, { "epoch": 6.37, "learning_rate": 1.0901239275500478e-06, "loss": 1.5333, "step": 33441 }, { "epoch": 6.37, "learning_rate": 1.0900667302192566e-06, "loss": 0.924, "step": 33442 }, { "epoch": 6.37, "learning_rate": 1.0900095328884651e-06, "loss": 1.8607, "step": 33443 }, { "epoch": 6.37, "learning_rate": 1.089952335557674e-06, "loss": 1.2727, "step": 33444 }, { "epoch": 6.37, "learning_rate": 1.0898951382268827e-06, "loss": 1.8613, "step": 33445 }, { "epoch": 6.37, "learning_rate": 1.0898379408960915e-06, "loss": 1.9157, "step": 33446 }, { "epoch": 6.37, "learning_rate": 1.0897807435653003e-06, "loss": 1.0474, "step": 33447 }, { "epoch": 6.37, "learning_rate": 1.0897235462345091e-06, "loss": 1.8329, "step": 33448 }, { "epoch": 6.37, "learning_rate": 1.089666348903718e-06, "loss": 1.7458, "step": 33449 }, { "epoch": 6.37, "learning_rate": 1.0896091515729265e-06, "loss": 2.1489, "step": 33450 }, { "epoch": 6.37, "learning_rate": 1.0895519542421353e-06, "loss": 2.1364, "step": 33451 }, { "epoch": 6.37, "learning_rate": 1.0894947569113443e-06, "loss": 1.8751, "step": 33452 }, { "epoch": 6.37, "learning_rate": 1.0894375595805529e-06, "loss": 2.5752, "step": 33453 }, { "epoch": 6.37, "learning_rate": 1.0893803622497617e-06, "loss": 1.8697, "step": 33454 }, { "epoch": 6.37, "learning_rate": 1.0893231649189705e-06, "loss": 1.2232, "step": 33455 }, { "epoch": 6.37, "learning_rate": 1.0892659675881792e-06, "loss": 1.8141, "step": 33456 }, { "epoch": 6.37, "learning_rate": 1.089208770257388e-06, "loss": 1.0776, "step": 33457 }, { "epoch": 6.37, "learning_rate": 1.0891515729265968e-06, "loss": 1.9333, "step": 33458 }, { "epoch": 6.37, "learning_rate": 1.0890943755958056e-06, "loss": 1.4879, "step": 33459 }, { "epoch": 6.37, "learning_rate": 1.0890371782650142e-06, "loss": 1.4713, "step": 33460 }, { "epoch": 6.37, "learning_rate": 1.088979980934223e-06, "loss": 4.6969, "step": 33461 }, { "epoch": 6.37, "learning_rate": 1.0889227836034318e-06, "loss": 1.3843, "step": 33462 }, { "epoch": 6.37, "learning_rate": 1.0888655862726408e-06, "loss": 1.0957, "step": 33463 }, { "epoch": 6.37, "learning_rate": 1.0888083889418494e-06, "loss": 1.9395, "step": 33464 }, { "epoch": 6.37, "learning_rate": 1.0887511916110582e-06, "loss": 2.4137, "step": 33465 }, { "epoch": 6.37, "learning_rate": 1.088693994280267e-06, "loss": 1.8434, "step": 33466 }, { "epoch": 6.37, "learning_rate": 1.0886367969494756e-06, "loss": 1.7777, "step": 33467 }, { "epoch": 6.37, "learning_rate": 1.0885795996186846e-06, "loss": 2.3477, "step": 33468 }, { "epoch": 6.38, "learning_rate": 1.0885224022878934e-06, "loss": 1.2575, "step": 33469 }, { "epoch": 6.38, "learning_rate": 1.0884652049571022e-06, "loss": 2.2775, "step": 33470 }, { "epoch": 6.38, "learning_rate": 1.0884080076263107e-06, "loss": 2.137, "step": 33471 }, { "epoch": 6.38, "learning_rate": 1.0883508102955195e-06, "loss": 1.4287, "step": 33472 }, { "epoch": 6.38, "learning_rate": 1.0882936129647283e-06, "loss": 0.6248, "step": 33473 }, { "epoch": 6.38, "learning_rate": 1.0882364156339371e-06, "loss": 1.1856, "step": 33474 }, { "epoch": 6.38, "learning_rate": 1.088179218303146e-06, "loss": 1.5983, "step": 33475 }, { "epoch": 6.38, "learning_rate": 1.0881220209723547e-06, "loss": 1.7645, "step": 33476 }, { "epoch": 6.38, "learning_rate": 1.0880648236415633e-06, "loss": 1.871, "step": 33477 }, { "epoch": 6.38, "learning_rate": 1.088007626310772e-06, "loss": 2.8081, "step": 33478 }, { "epoch": 6.38, "learning_rate": 1.087950428979981e-06, "loss": 0.9596, "step": 33479 }, { "epoch": 6.38, "learning_rate": 1.0878932316491899e-06, "loss": 1.9529, "step": 33480 }, { "epoch": 6.38, "learning_rate": 1.0878360343183985e-06, "loss": 1.8621, "step": 33481 }, { "epoch": 6.38, "learning_rate": 1.0877788369876073e-06, "loss": 1.5693, "step": 33482 }, { "epoch": 6.38, "learning_rate": 1.087721639656816e-06, "loss": 2.4661, "step": 33483 }, { "epoch": 6.38, "learning_rate": 1.0876644423260248e-06, "loss": 1.3947, "step": 33484 }, { "epoch": 6.38, "learning_rate": 1.0876072449952336e-06, "loss": 3.1018, "step": 33485 }, { "epoch": 6.38, "learning_rate": 1.0875500476644424e-06, "loss": 1.1379, "step": 33486 }, { "epoch": 6.38, "learning_rate": 1.0874928503336512e-06, "loss": 2.71, "step": 33487 }, { "epoch": 6.38, "learning_rate": 1.0874356530028598e-06, "loss": 1.8297, "step": 33488 }, { "epoch": 6.38, "learning_rate": 1.0873784556720686e-06, "loss": 1.8436, "step": 33489 }, { "epoch": 6.38, "learning_rate": 1.0873212583412774e-06, "loss": 1.773, "step": 33490 }, { "epoch": 6.38, "learning_rate": 1.0872640610104862e-06, "loss": 0.9512, "step": 33491 }, { "epoch": 6.38, "learning_rate": 1.087206863679695e-06, "loss": 1.727, "step": 33492 }, { "epoch": 6.38, "learning_rate": 1.0871496663489038e-06, "loss": 1.3781, "step": 33493 }, { "epoch": 6.38, "learning_rate": 1.0870924690181124e-06, "loss": 2.4018, "step": 33494 }, { "epoch": 6.38, "learning_rate": 1.0870352716873211e-06, "loss": 2.305, "step": 33495 }, { "epoch": 6.38, "learning_rate": 1.0869780743565302e-06, "loss": 1.772, "step": 33496 }, { "epoch": 6.38, "learning_rate": 1.086920877025739e-06, "loss": 1.4194, "step": 33497 }, { "epoch": 6.38, "learning_rate": 1.0868636796949475e-06, "loss": 1.5665, "step": 33498 }, { "epoch": 6.38, "learning_rate": 1.0868064823641563e-06, "loss": 2.3462, "step": 33499 }, { "epoch": 6.38, "learning_rate": 1.0867492850333651e-06, "loss": 0.5697, "step": 33500 }, { "epoch": 6.38, "learning_rate": 1.086692087702574e-06, "loss": 1.6794, "step": 33501 }, { "epoch": 6.38, "learning_rate": 1.0866348903717827e-06, "loss": 2.4265, "step": 33502 }, { "epoch": 6.38, "learning_rate": 1.0865776930409915e-06, "loss": 2.4973, "step": 33503 }, { "epoch": 6.38, "learning_rate": 1.0865204957102003e-06, "loss": 1.0451, "step": 33504 }, { "epoch": 6.38, "learning_rate": 1.0864632983794089e-06, "loss": 1.8412, "step": 33505 }, { "epoch": 6.38, "learning_rate": 1.0864061010486177e-06, "loss": 1.8339, "step": 33506 }, { "epoch": 6.38, "learning_rate": 1.0863489037178267e-06, "loss": 2.0161, "step": 33507 }, { "epoch": 6.38, "learning_rate": 1.0862917063870353e-06, "loss": 1.7137, "step": 33508 }, { "epoch": 6.38, "learning_rate": 1.086234509056244e-06, "loss": 1.3823, "step": 33509 }, { "epoch": 6.38, "learning_rate": 1.0861773117254528e-06, "loss": 1.8051, "step": 33510 }, { "epoch": 6.38, "learning_rate": 1.0861201143946616e-06, "loss": 3.4034, "step": 33511 }, { "epoch": 6.38, "learning_rate": 1.0860629170638704e-06, "loss": 1.5953, "step": 33512 }, { "epoch": 6.38, "learning_rate": 1.0860057197330792e-06, "loss": 1.9362, "step": 33513 }, { "epoch": 6.38, "learning_rate": 1.085948522402288e-06, "loss": 2.9827, "step": 33514 }, { "epoch": 6.38, "learning_rate": 1.0858913250714966e-06, "loss": 2.4914, "step": 33515 }, { "epoch": 6.38, "learning_rate": 1.0858341277407054e-06, "loss": 2.3283, "step": 33516 }, { "epoch": 6.38, "learning_rate": 1.0857769304099142e-06, "loss": 2.0681, "step": 33517 }, { "epoch": 6.38, "learning_rate": 1.085719733079123e-06, "loss": 2.3328, "step": 33518 }, { "epoch": 6.38, "learning_rate": 1.0856625357483318e-06, "loss": 0.8652, "step": 33519 }, { "epoch": 6.38, "learning_rate": 1.0856053384175406e-06, "loss": 0.6829, "step": 33520 }, { "epoch": 6.38, "learning_rate": 1.0855481410867494e-06, "loss": 2.0757, "step": 33521 }, { "epoch": 6.39, "learning_rate": 1.085490943755958e-06, "loss": 0.7536, "step": 33522 }, { "epoch": 6.39, "learning_rate": 1.085433746425167e-06, "loss": 1.4284, "step": 33523 }, { "epoch": 6.39, "learning_rate": 1.0853765490943757e-06, "loss": 2.1711, "step": 33524 }, { "epoch": 6.39, "learning_rate": 1.0853193517635843e-06, "loss": 1.4148, "step": 33525 }, { "epoch": 6.39, "learning_rate": 1.0852621544327931e-06, "loss": 2.2768, "step": 33526 }, { "epoch": 6.39, "learning_rate": 1.085204957102002e-06, "loss": 1.9916, "step": 33527 }, { "epoch": 6.39, "learning_rate": 1.0851477597712107e-06, "loss": 1.4166, "step": 33528 }, { "epoch": 6.39, "learning_rate": 1.0850905624404195e-06, "loss": 2.8152, "step": 33529 }, { "epoch": 6.39, "learning_rate": 1.0850333651096283e-06, "loss": 2.3156, "step": 33530 }, { "epoch": 6.39, "learning_rate": 1.084976167778837e-06, "loss": 0.4856, "step": 33531 }, { "epoch": 6.39, "learning_rate": 1.0849189704480457e-06, "loss": 1.5977, "step": 33532 }, { "epoch": 6.39, "learning_rate": 1.0848617731172545e-06, "loss": 1.5916, "step": 33533 }, { "epoch": 6.39, "learning_rate": 1.0848045757864633e-06, "loss": 1.3022, "step": 33534 }, { "epoch": 6.39, "learning_rate": 1.0847473784556723e-06, "loss": 2.1078, "step": 33535 }, { "epoch": 6.39, "learning_rate": 1.0846901811248808e-06, "loss": 2.1824, "step": 33536 }, { "epoch": 6.39, "learning_rate": 1.0846329837940896e-06, "loss": 2.002, "step": 33537 }, { "epoch": 6.39, "learning_rate": 1.0845757864632984e-06, "loss": 3.6901, "step": 33538 }, { "epoch": 6.39, "learning_rate": 1.084518589132507e-06, "loss": 2.5287, "step": 33539 }, { "epoch": 6.39, "learning_rate": 1.084461391801716e-06, "loss": 1.3036, "step": 33540 }, { "epoch": 6.39, "learning_rate": 1.0844041944709248e-06, "loss": 1.9652, "step": 33541 }, { "epoch": 6.39, "learning_rate": 1.0843469971401334e-06, "loss": 1.4888, "step": 33542 }, { "epoch": 6.39, "learning_rate": 1.0842897998093422e-06, "loss": 2.2708, "step": 33543 }, { "epoch": 6.39, "learning_rate": 1.084232602478551e-06, "loss": 2.3487, "step": 33544 }, { "epoch": 6.39, "learning_rate": 1.0841754051477598e-06, "loss": 2.1607, "step": 33545 }, { "epoch": 6.39, "learning_rate": 1.0841182078169686e-06, "loss": 0.882, "step": 33546 }, { "epoch": 6.39, "learning_rate": 1.0840610104861774e-06, "loss": 2.3441, "step": 33547 }, { "epoch": 6.39, "learning_rate": 1.0840038131553862e-06, "loss": 1.1201, "step": 33548 }, { "epoch": 6.39, "learning_rate": 1.0839466158245947e-06, "loss": 1.4714, "step": 33549 }, { "epoch": 6.39, "learning_rate": 1.0838894184938035e-06, "loss": 1.9543, "step": 33550 }, { "epoch": 6.39, "learning_rate": 1.0838322211630125e-06, "loss": 1.0756, "step": 33551 }, { "epoch": 6.39, "learning_rate": 1.0837750238322213e-06, "loss": 0.9147, "step": 33552 }, { "epoch": 6.39, "learning_rate": 1.08371782650143e-06, "loss": 1.1642, "step": 33553 }, { "epoch": 6.39, "learning_rate": 1.0836606291706387e-06, "loss": 1.9666, "step": 33554 }, { "epoch": 6.39, "learning_rate": 1.0836034318398475e-06, "loss": 1.2637, "step": 33555 }, { "epoch": 6.39, "learning_rate": 1.0835462345090563e-06, "loss": 3.4435, "step": 33556 }, { "epoch": 6.39, "learning_rate": 1.083489037178265e-06, "loss": 1.7214, "step": 33557 }, { "epoch": 6.39, "learning_rate": 1.0834318398474739e-06, "loss": 2.9729, "step": 33558 }, { "epoch": 6.39, "learning_rate": 1.0833746425166825e-06, "loss": 2.0559, "step": 33559 }, { "epoch": 6.39, "learning_rate": 1.0833174451858913e-06, "loss": 2.3783, "step": 33560 }, { "epoch": 6.39, "learning_rate": 1.0832602478551e-06, "loss": 4.0433, "step": 33561 }, { "epoch": 6.39, "learning_rate": 1.0832030505243089e-06, "loss": 1.7306, "step": 33562 }, { "epoch": 6.39, "learning_rate": 1.0831458531935176e-06, "loss": 1.7101, "step": 33563 }, { "epoch": 6.39, "learning_rate": 1.0830886558627264e-06, "loss": 2.2548, "step": 33564 }, { "epoch": 6.39, "learning_rate": 1.0830314585319352e-06, "loss": 1.1007, "step": 33565 }, { "epoch": 6.39, "learning_rate": 1.0829742612011438e-06, "loss": 2.0878, "step": 33566 }, { "epoch": 6.39, "learning_rate": 1.0829170638703528e-06, "loss": 1.4507, "step": 33567 }, { "epoch": 6.39, "learning_rate": 1.0828598665395616e-06, "loss": 1.1103, "step": 33568 }, { "epoch": 6.39, "learning_rate": 1.0828026692087704e-06, "loss": 1.2971, "step": 33569 }, { "epoch": 6.39, "learning_rate": 1.082745471877979e-06, "loss": 2.1552, "step": 33570 }, { "epoch": 6.39, "learning_rate": 1.0826882745471878e-06, "loss": 1.9744, "step": 33571 }, { "epoch": 6.39, "learning_rate": 1.0826310772163966e-06, "loss": 0.923, "step": 33572 }, { "epoch": 6.39, "learning_rate": 1.0825738798856054e-06, "loss": 3.3899, "step": 33573 }, { "epoch": 6.4, "learning_rate": 1.0825166825548142e-06, "loss": 2.4525, "step": 33574 }, { "epoch": 6.4, "learning_rate": 1.082459485224023e-06, "loss": 1.05, "step": 33575 }, { "epoch": 6.4, "learning_rate": 1.0824022878932318e-06, "loss": 1.912, "step": 33576 }, { "epoch": 6.4, "learning_rate": 1.0823450905624403e-06, "loss": 1.9897, "step": 33577 }, { "epoch": 6.4, "learning_rate": 1.0822878932316491e-06, "loss": 0.7019, "step": 33578 }, { "epoch": 6.4, "learning_rate": 1.0822306959008581e-06, "loss": 2.3278, "step": 33579 }, { "epoch": 6.4, "learning_rate": 1.0821734985700667e-06, "loss": 1.755, "step": 33580 }, { "epoch": 6.4, "learning_rate": 1.0821163012392755e-06, "loss": 1.6299, "step": 33581 }, { "epoch": 6.4, "learning_rate": 1.0820591039084843e-06, "loss": 1.1997, "step": 33582 }, { "epoch": 6.4, "learning_rate": 1.0820019065776929e-06, "loss": 1.2105, "step": 33583 }, { "epoch": 6.4, "learning_rate": 1.0819447092469019e-06, "loss": 1.0154, "step": 33584 }, { "epoch": 6.4, "learning_rate": 1.0818875119161107e-06, "loss": 1.2663, "step": 33585 }, { "epoch": 6.4, "learning_rate": 1.0818303145853195e-06, "loss": 1.7161, "step": 33586 }, { "epoch": 6.4, "learning_rate": 1.081773117254528e-06, "loss": 2.1945, "step": 33587 }, { "epoch": 6.4, "learning_rate": 1.0817159199237369e-06, "loss": 1.6837, "step": 33588 }, { "epoch": 6.4, "learning_rate": 1.0816587225929457e-06, "loss": 1.1719, "step": 33589 }, { "epoch": 6.4, "learning_rate": 1.0816015252621544e-06, "loss": 0.8725, "step": 33590 }, { "epoch": 6.4, "learning_rate": 1.0815443279313632e-06, "loss": 2.0216, "step": 33591 }, { "epoch": 6.4, "learning_rate": 1.081487130600572e-06, "loss": 2.3212, "step": 33592 }, { "epoch": 6.4, "learning_rate": 1.0814299332697808e-06, "loss": 2.0658, "step": 33593 }, { "epoch": 6.4, "learning_rate": 1.0813727359389894e-06, "loss": 2.481, "step": 33594 }, { "epoch": 6.4, "learning_rate": 1.0813155386081984e-06, "loss": 1.6045, "step": 33595 }, { "epoch": 6.4, "learning_rate": 1.0812583412774072e-06, "loss": 1.4521, "step": 33596 }, { "epoch": 6.4, "learning_rate": 1.0812011439466158e-06, "loss": 2.2023, "step": 33597 }, { "epoch": 6.4, "learning_rate": 1.0811439466158246e-06, "loss": 1.6521, "step": 33598 }, { "epoch": 6.4, "learning_rate": 1.0810867492850334e-06, "loss": 2.5955, "step": 33599 }, { "epoch": 6.4, "learning_rate": 1.0810295519542422e-06, "loss": 3.094, "step": 33600 }, { "epoch": 6.4, "learning_rate": 1.080972354623451e-06, "loss": 1.0531, "step": 33601 }, { "epoch": 6.4, "learning_rate": 1.0809151572926598e-06, "loss": 1.1753, "step": 33602 }, { "epoch": 6.4, "learning_rate": 1.0808579599618686e-06, "loss": 1.044, "step": 33603 }, { "epoch": 6.4, "learning_rate": 1.0808007626310771e-06, "loss": 1.9332, "step": 33604 }, { "epoch": 6.4, "learning_rate": 1.080743565300286e-06, "loss": 1.066, "step": 33605 }, { "epoch": 6.4, "learning_rate": 1.0806863679694947e-06, "loss": 1.6255, "step": 33606 }, { "epoch": 6.4, "learning_rate": 1.0806291706387035e-06, "loss": 1.4125, "step": 33607 }, { "epoch": 6.4, "learning_rate": 1.0805719733079123e-06, "loss": 1.7838, "step": 33608 }, { "epoch": 6.4, "learning_rate": 1.080514775977121e-06, "loss": 1.4043, "step": 33609 }, { "epoch": 6.4, "learning_rate": 1.08045757864633e-06, "loss": 2.5549, "step": 33610 }, { "epoch": 6.4, "learning_rate": 1.0804003813155387e-06, "loss": 1.7768, "step": 33611 }, { "epoch": 6.4, "learning_rate": 1.0803431839847475e-06, "loss": 2.6365, "step": 33612 }, { "epoch": 6.4, "learning_rate": 1.0802859866539563e-06, "loss": 1.3755, "step": 33613 }, { "epoch": 6.4, "learning_rate": 1.0802287893231649e-06, "loss": 2.984, "step": 33614 }, { "epoch": 6.4, "learning_rate": 1.0801715919923737e-06, "loss": 1.3142, "step": 33615 }, { "epoch": 6.4, "learning_rate": 1.0801143946615824e-06, "loss": 1.4566, "step": 33616 }, { "epoch": 6.4, "learning_rate": 1.0800571973307912e-06, "loss": 1.7615, "step": 33617 }, { "epoch": 6.4, "learning_rate": 1.08e-06, "loss": 2.0875, "step": 33618 }, { "epoch": 6.4, "learning_rate": 1.0799428026692088e-06, "loss": 3.0532, "step": 33619 }, { "epoch": 6.4, "learning_rate": 1.0798856053384176e-06, "loss": 2.0199, "step": 33620 }, { "epoch": 6.4, "learning_rate": 1.0798284080076262e-06, "loss": 2.1084, "step": 33621 }, { "epoch": 6.4, "learning_rate": 1.079771210676835e-06, "loss": 1.5978, "step": 33622 }, { "epoch": 6.4, "learning_rate": 1.079714013346044e-06, "loss": 1.6846, "step": 33623 }, { "epoch": 6.4, "learning_rate": 1.0796568160152526e-06, "loss": 1.0007, "step": 33624 }, { "epoch": 6.4, "learning_rate": 1.0795996186844614e-06, "loss": 2.254, "step": 33625 }, { "epoch": 6.4, "learning_rate": 1.0795424213536702e-06, "loss": 1.1756, "step": 33626 }, { "epoch": 6.41, "learning_rate": 1.079485224022879e-06, "loss": 1.421, "step": 33627 }, { "epoch": 6.41, "learning_rate": 1.0794280266920878e-06, "loss": 2.0323, "step": 33628 }, { "epoch": 6.41, "learning_rate": 1.0793708293612966e-06, "loss": 1.595, "step": 33629 }, { "epoch": 6.41, "learning_rate": 1.0793136320305053e-06, "loss": 2.8097, "step": 33630 }, { "epoch": 6.41, "learning_rate": 1.079256434699714e-06, "loss": 1.8938, "step": 33631 }, { "epoch": 6.41, "learning_rate": 1.0791992373689227e-06, "loss": 2.233, "step": 33632 }, { "epoch": 6.41, "learning_rate": 1.0791420400381315e-06, "loss": 2.2637, "step": 33633 }, { "epoch": 6.41, "learning_rate": 1.0790848427073405e-06, "loss": 1.7804, "step": 33634 }, { "epoch": 6.41, "learning_rate": 1.0790276453765491e-06, "loss": 2.1611, "step": 33635 }, { "epoch": 6.41, "learning_rate": 1.078970448045758e-06, "loss": 1.7151, "step": 33636 }, { "epoch": 6.41, "learning_rate": 1.0789132507149667e-06, "loss": 1.6352, "step": 33637 }, { "epoch": 6.41, "learning_rate": 1.0788560533841753e-06, "loss": 1.3574, "step": 33638 }, { "epoch": 6.41, "learning_rate": 1.0787988560533843e-06, "loss": 2.524, "step": 33639 }, { "epoch": 6.41, "learning_rate": 1.078741658722593e-06, "loss": 1.4403, "step": 33640 }, { "epoch": 6.41, "learning_rate": 1.0786844613918019e-06, "loss": 1.9972, "step": 33641 }, { "epoch": 6.41, "learning_rate": 1.0786272640610105e-06, "loss": 2.3707, "step": 33642 }, { "epoch": 6.41, "learning_rate": 1.0785700667302192e-06, "loss": 2.1997, "step": 33643 }, { "epoch": 6.41, "learning_rate": 1.078512869399428e-06, "loss": 1.4224, "step": 33644 }, { "epoch": 6.41, "learning_rate": 1.0784556720686368e-06, "loss": 1.3349, "step": 33645 }, { "epoch": 6.41, "learning_rate": 1.0783984747378456e-06, "loss": 2.329, "step": 33646 }, { "epoch": 6.41, "learning_rate": 1.0783412774070544e-06, "loss": 1.2088, "step": 33647 }, { "epoch": 6.41, "learning_rate": 1.078284080076263e-06, "loss": 1.0516, "step": 33648 }, { "epoch": 6.41, "learning_rate": 1.0782268827454718e-06, "loss": 2.8148, "step": 33649 }, { "epoch": 6.41, "learning_rate": 1.0781696854146806e-06, "loss": 1.0318, "step": 33650 }, { "epoch": 6.41, "learning_rate": 1.0781124880838896e-06, "loss": 1.678, "step": 33651 }, { "epoch": 6.41, "learning_rate": 1.0780552907530982e-06, "loss": 1.0468, "step": 33652 }, { "epoch": 6.41, "learning_rate": 1.077998093422307e-06, "loss": 2.3112, "step": 33653 }, { "epoch": 6.41, "learning_rate": 1.0779408960915158e-06, "loss": 2.3346, "step": 33654 }, { "epoch": 6.41, "learning_rate": 1.0778836987607246e-06, "loss": 1.7815, "step": 33655 }, { "epoch": 6.41, "learning_rate": 1.0778265014299334e-06, "loss": 2.6862, "step": 33656 }, { "epoch": 6.41, "learning_rate": 1.0777693040991421e-06, "loss": 2.2514, "step": 33657 }, { "epoch": 6.41, "learning_rate": 1.077712106768351e-06, "loss": 1.601, "step": 33658 }, { "epoch": 6.41, "learning_rate": 1.0776549094375595e-06, "loss": 1.3716, "step": 33659 }, { "epoch": 6.41, "learning_rate": 1.0775977121067683e-06, "loss": 2.511, "step": 33660 }, { "epoch": 6.41, "learning_rate": 1.0775405147759771e-06, "loss": 1.1935, "step": 33661 }, { "epoch": 6.41, "learning_rate": 1.077483317445186e-06, "loss": 0.9107, "step": 33662 }, { "epoch": 6.41, "learning_rate": 1.0774261201143947e-06, "loss": 2.5138, "step": 33663 }, { "epoch": 6.41, "learning_rate": 1.0773689227836035e-06, "loss": 2.202, "step": 33664 }, { "epoch": 6.41, "learning_rate": 1.077311725452812e-06, "loss": 3.4778, "step": 33665 }, { "epoch": 6.41, "learning_rate": 1.0772545281220209e-06, "loss": 2.6318, "step": 33666 }, { "epoch": 6.41, "learning_rate": 1.0771973307912299e-06, "loss": 1.2515, "step": 33667 }, { "epoch": 6.41, "learning_rate": 1.0771401334604387e-06, "loss": 2.1736, "step": 33668 }, { "epoch": 6.41, "learning_rate": 1.0770829361296472e-06, "loss": 2.6443, "step": 33669 }, { "epoch": 6.41, "learning_rate": 1.077025738798856e-06, "loss": 1.8065, "step": 33670 }, { "epoch": 6.41, "learning_rate": 1.0769685414680648e-06, "loss": 1.1926, "step": 33671 }, { "epoch": 6.41, "learning_rate": 1.0769113441372736e-06, "loss": 2.5453, "step": 33672 }, { "epoch": 6.41, "learning_rate": 1.0768541468064824e-06, "loss": 1.7481, "step": 33673 }, { "epoch": 6.41, "learning_rate": 1.0767969494756912e-06, "loss": 0.9942, "step": 33674 }, { "epoch": 6.41, "learning_rate": 1.0767397521449e-06, "loss": 0.9601, "step": 33675 }, { "epoch": 6.41, "learning_rate": 1.0766825548141086e-06, "loss": 1.4378, "step": 33676 }, { "epoch": 6.41, "learning_rate": 1.0766253574833174e-06, "loss": 2.3339, "step": 33677 }, { "epoch": 6.41, "learning_rate": 1.0765681601525264e-06, "loss": 2.3069, "step": 33678 }, { "epoch": 6.42, "learning_rate": 1.076510962821735e-06, "loss": 1.5265, "step": 33679 }, { "epoch": 6.42, "learning_rate": 1.0764537654909438e-06, "loss": 3.0789, "step": 33680 }, { "epoch": 6.42, "learning_rate": 1.0763965681601526e-06, "loss": 2.1434, "step": 33681 }, { "epoch": 6.42, "learning_rate": 1.0763393708293614e-06, "loss": 2.4509, "step": 33682 }, { "epoch": 6.42, "learning_rate": 1.0762821734985702e-06, "loss": 1.0077, "step": 33683 }, { "epoch": 6.42, "learning_rate": 1.076224976167779e-06, "loss": 2.745, "step": 33684 }, { "epoch": 6.42, "learning_rate": 1.0761677788369877e-06, "loss": 2.3333, "step": 33685 }, { "epoch": 6.42, "learning_rate": 1.0761105815061963e-06, "loss": 1.0344, "step": 33686 }, { "epoch": 6.42, "learning_rate": 1.0760533841754051e-06, "loss": 1.1102, "step": 33687 }, { "epoch": 6.42, "learning_rate": 1.075996186844614e-06, "loss": 2.541, "step": 33688 }, { "epoch": 6.42, "learning_rate": 1.0759389895138227e-06, "loss": 2.195, "step": 33689 }, { "epoch": 6.42, "learning_rate": 1.0758817921830315e-06, "loss": 1.3839, "step": 33690 }, { "epoch": 6.42, "learning_rate": 1.0758245948522403e-06, "loss": 0.8661, "step": 33691 }, { "epoch": 6.42, "learning_rate": 1.075767397521449e-06, "loss": 1.2146, "step": 33692 }, { "epoch": 6.42, "learning_rate": 1.0757102001906577e-06, "loss": 1.9535, "step": 33693 }, { "epoch": 6.42, "learning_rate": 1.0756530028598665e-06, "loss": 0.77, "step": 33694 }, { "epoch": 6.42, "learning_rate": 1.0755958055290755e-06, "loss": 2.2946, "step": 33695 }, { "epoch": 6.42, "learning_rate": 1.075538608198284e-06, "loss": 1.0272, "step": 33696 }, { "epoch": 6.42, "learning_rate": 1.0754814108674928e-06, "loss": 2.5119, "step": 33697 }, { "epoch": 6.42, "learning_rate": 1.0754242135367016e-06, "loss": 0.5565, "step": 33698 }, { "epoch": 6.42, "learning_rate": 1.0753670162059104e-06, "loss": 2.2677, "step": 33699 }, { "epoch": 6.42, "learning_rate": 1.0753098188751192e-06, "loss": 2.0593, "step": 33700 }, { "epoch": 6.42, "learning_rate": 1.075252621544328e-06, "loss": 1.3407, "step": 33701 }, { "epoch": 6.42, "learning_rate": 1.0751954242135368e-06, "loss": 0.5283, "step": 33702 }, { "epoch": 6.42, "learning_rate": 1.0751382268827454e-06, "loss": 2.2204, "step": 33703 }, { "epoch": 6.42, "learning_rate": 1.0750810295519542e-06, "loss": 2.3711, "step": 33704 }, { "epoch": 6.42, "learning_rate": 1.075023832221163e-06, "loss": 2.535, "step": 33705 }, { "epoch": 6.42, "learning_rate": 1.074966634890372e-06, "loss": 1.215, "step": 33706 }, { "epoch": 6.42, "learning_rate": 1.0749094375595806e-06, "loss": 1.4053, "step": 33707 }, { "epoch": 6.42, "learning_rate": 1.0748522402287894e-06, "loss": 1.9126, "step": 33708 }, { "epoch": 6.42, "learning_rate": 1.0747950428979982e-06, "loss": 1.2451, "step": 33709 }, { "epoch": 6.42, "learning_rate": 1.0747378455672067e-06, "loss": 1.9309, "step": 33710 }, { "epoch": 6.42, "learning_rate": 1.0746806482364157e-06, "loss": 1.3218, "step": 33711 }, { "epoch": 6.42, "learning_rate": 1.0746234509056245e-06, "loss": 1.2739, "step": 33712 }, { "epoch": 6.42, "learning_rate": 1.0745662535748331e-06, "loss": 1.4155, "step": 33713 }, { "epoch": 6.42, "learning_rate": 1.074509056244042e-06, "loss": 1.4022, "step": 33714 }, { "epoch": 6.42, "learning_rate": 1.0744518589132507e-06, "loss": 2.3753, "step": 33715 }, { "epoch": 6.42, "learning_rate": 1.0743946615824595e-06, "loss": 1.2071, "step": 33716 }, { "epoch": 6.42, "learning_rate": 1.0743374642516683e-06, "loss": 0.8741, "step": 33717 }, { "epoch": 6.42, "learning_rate": 1.074280266920877e-06, "loss": 1.9404, "step": 33718 }, { "epoch": 6.42, "learning_rate": 1.0742230695900859e-06, "loss": 1.1159, "step": 33719 }, { "epoch": 6.42, "learning_rate": 1.0741658722592945e-06, "loss": 2.0194, "step": 33720 }, { "epoch": 6.42, "learning_rate": 1.0741086749285033e-06, "loss": 1.9054, "step": 33721 }, { "epoch": 6.42, "learning_rate": 1.0740514775977123e-06, "loss": 2.7459, "step": 33722 }, { "epoch": 6.42, "learning_rate": 1.073994280266921e-06, "loss": 2.8469, "step": 33723 }, { "epoch": 6.42, "learning_rate": 1.0739370829361296e-06, "loss": 1.7238, "step": 33724 }, { "epoch": 6.42, "learning_rate": 1.0738798856053384e-06, "loss": 2.3969, "step": 33725 }, { "epoch": 6.42, "learning_rate": 1.0738226882745472e-06, "loss": 1.8486, "step": 33726 }, { "epoch": 6.42, "learning_rate": 1.073765490943756e-06, "loss": 0.6406, "step": 33727 }, { "epoch": 6.42, "learning_rate": 1.0737082936129648e-06, "loss": 2.7054, "step": 33728 }, { "epoch": 6.42, "learning_rate": 1.0736510962821736e-06, "loss": 2.007, "step": 33729 }, { "epoch": 6.42, "learning_rate": 1.0735938989513822e-06, "loss": 2.6256, "step": 33730 }, { "epoch": 6.42, "learning_rate": 1.073536701620591e-06, "loss": 3.1165, "step": 33731 }, { "epoch": 6.43, "learning_rate": 1.0734795042897998e-06, "loss": 1.4363, "step": 33732 }, { "epoch": 6.43, "learning_rate": 1.0734223069590086e-06, "loss": 1.2671, "step": 33733 }, { "epoch": 6.43, "learning_rate": 1.0733651096282174e-06, "loss": 1.7214, "step": 33734 }, { "epoch": 6.43, "learning_rate": 1.0733079122974262e-06, "loss": 1.8808, "step": 33735 }, { "epoch": 6.43, "learning_rate": 1.073250714966635e-06, "loss": 1.8522, "step": 33736 }, { "epoch": 6.43, "learning_rate": 1.0731935176358435e-06, "loss": 1.9098, "step": 33737 }, { "epoch": 6.43, "learning_rate": 1.0731363203050523e-06, "loss": 2.6142, "step": 33738 }, { "epoch": 6.43, "learning_rate": 1.0730791229742613e-06, "loss": 2.1364, "step": 33739 }, { "epoch": 6.43, "learning_rate": 1.0730219256434701e-06, "loss": 2.5329, "step": 33740 }, { "epoch": 6.43, "learning_rate": 1.0729647283126787e-06, "loss": 2.42, "step": 33741 }, { "epoch": 6.43, "learning_rate": 1.0729075309818875e-06, "loss": 3.8959, "step": 33742 }, { "epoch": 6.43, "learning_rate": 1.0728503336510963e-06, "loss": 2.5264, "step": 33743 }, { "epoch": 6.43, "learning_rate": 1.072793136320305e-06, "loss": 2.57, "step": 33744 }, { "epoch": 6.43, "learning_rate": 1.0727359389895139e-06, "loss": 2.2565, "step": 33745 }, { "epoch": 6.43, "learning_rate": 1.0726787416587227e-06, "loss": 1.3716, "step": 33746 }, { "epoch": 6.43, "learning_rate": 1.0726215443279315e-06, "loss": 2.5484, "step": 33747 }, { "epoch": 6.43, "learning_rate": 1.07256434699714e-06, "loss": 2.6623, "step": 33748 }, { "epoch": 6.43, "learning_rate": 1.0725071496663488e-06, "loss": 1.5258, "step": 33749 }, { "epoch": 6.43, "learning_rate": 1.0724499523355579e-06, "loss": 2.6468, "step": 33750 }, { "epoch": 6.43, "learning_rate": 1.0723927550047664e-06, "loss": 1.1094, "step": 33751 }, { "epoch": 6.43, "learning_rate": 1.0723355576739752e-06, "loss": 2.7619, "step": 33752 }, { "epoch": 6.43, "learning_rate": 1.072278360343184e-06, "loss": 1.8029, "step": 33753 }, { "epoch": 6.43, "learning_rate": 1.0722211630123926e-06, "loss": 2.991, "step": 33754 }, { "epoch": 6.43, "learning_rate": 1.0721639656816016e-06, "loss": 1.8862, "step": 33755 }, { "epoch": 6.43, "learning_rate": 1.0721067683508104e-06, "loss": 0.9664, "step": 33756 }, { "epoch": 6.43, "learning_rate": 1.0720495710200192e-06, "loss": 0.9159, "step": 33757 }, { "epoch": 6.43, "learning_rate": 1.0719923736892278e-06, "loss": 1.1761, "step": 33758 }, { "epoch": 6.43, "learning_rate": 1.0719351763584366e-06, "loss": 1.4509, "step": 33759 }, { "epoch": 6.43, "learning_rate": 1.0718779790276454e-06, "loss": 1.3453, "step": 33760 }, { "epoch": 6.43, "learning_rate": 1.0718207816968542e-06, "loss": 2.4459, "step": 33761 }, { "epoch": 6.43, "learning_rate": 1.071763584366063e-06, "loss": 1.1644, "step": 33762 }, { "epoch": 6.43, "learning_rate": 1.0717063870352718e-06, "loss": 1.5746, "step": 33763 }, { "epoch": 6.43, "learning_rate": 1.0716491897044805e-06, "loss": 1.1707, "step": 33764 }, { "epoch": 6.43, "learning_rate": 1.0715919923736891e-06, "loss": 1.592, "step": 33765 }, { "epoch": 6.43, "learning_rate": 1.0715347950428981e-06, "loss": 2.0289, "step": 33766 }, { "epoch": 6.43, "learning_rate": 1.071477597712107e-06, "loss": 2.0357, "step": 33767 }, { "epoch": 6.43, "learning_rate": 1.0714204003813155e-06, "loss": 0.3626, "step": 33768 }, { "epoch": 6.43, "learning_rate": 1.0713632030505243e-06, "loss": 1.7016, "step": 33769 }, { "epoch": 6.43, "learning_rate": 1.071306005719733e-06, "loss": 2.4004, "step": 33770 }, { "epoch": 6.43, "learning_rate": 1.0712488083889419e-06, "loss": 2.9684, "step": 33771 }, { "epoch": 6.43, "learning_rate": 1.0711916110581507e-06, "loss": 1.5148, "step": 33772 }, { "epoch": 6.43, "learning_rate": 1.0711344137273595e-06, "loss": 1.5987, "step": 33773 }, { "epoch": 6.43, "learning_rate": 1.0710772163965683e-06, "loss": 1.5851, "step": 33774 }, { "epoch": 6.43, "learning_rate": 1.0710200190657769e-06, "loss": 0.7978, "step": 33775 }, { "epoch": 6.43, "learning_rate": 1.0709628217349856e-06, "loss": 1.5883, "step": 33776 }, { "epoch": 6.43, "learning_rate": 1.0709056244041944e-06, "loss": 2.7846, "step": 33777 }, { "epoch": 6.43, "learning_rate": 1.0708484270734032e-06, "loss": 2.5049, "step": 33778 }, { "epoch": 6.43, "learning_rate": 1.070791229742612e-06, "loss": 1.3031, "step": 33779 }, { "epoch": 6.43, "learning_rate": 1.0707340324118208e-06, "loss": 0.91, "step": 33780 }, { "epoch": 6.43, "learning_rate": 1.0706768350810296e-06, "loss": 1.2189, "step": 33781 }, { "epoch": 6.43, "learning_rate": 1.0706196377502382e-06, "loss": 1.6573, "step": 33782 }, { "epoch": 6.43, "learning_rate": 1.0705624404194472e-06, "loss": 3.0061, "step": 33783 }, { "epoch": 6.44, "learning_rate": 1.070505243088656e-06, "loss": 1.9523, "step": 33784 }, { "epoch": 6.44, "learning_rate": 1.0704480457578646e-06, "loss": 1.433, "step": 33785 }, { "epoch": 6.44, "learning_rate": 1.0703908484270734e-06, "loss": 1.7748, "step": 33786 }, { "epoch": 6.44, "learning_rate": 1.0703336510962822e-06, "loss": 0.9325, "step": 33787 }, { "epoch": 6.44, "learning_rate": 1.070276453765491e-06, "loss": 2.9849, "step": 33788 }, { "epoch": 6.44, "learning_rate": 1.0702192564346998e-06, "loss": 1.3731, "step": 33789 }, { "epoch": 6.44, "learning_rate": 1.0701620591039085e-06, "loss": 3.2677, "step": 33790 }, { "epoch": 6.44, "learning_rate": 1.0701048617731173e-06, "loss": 2.1754, "step": 33791 }, { "epoch": 6.44, "learning_rate": 1.070047664442326e-06, "loss": 1.9847, "step": 33792 }, { "epoch": 6.44, "learning_rate": 1.0699904671115347e-06, "loss": 1.4288, "step": 33793 }, { "epoch": 6.44, "learning_rate": 1.0699332697807437e-06, "loss": 1.6346, "step": 33794 }, { "epoch": 6.44, "learning_rate": 1.0698760724499523e-06, "loss": 1.3326, "step": 33795 }, { "epoch": 6.44, "learning_rate": 1.069818875119161e-06, "loss": 1.5867, "step": 33796 }, { "epoch": 6.44, "learning_rate": 1.0697616777883699e-06, "loss": 1.5729, "step": 33797 }, { "epoch": 6.44, "learning_rate": 1.0697044804575787e-06, "loss": 1.9952, "step": 33798 }, { "epoch": 6.44, "learning_rate": 1.0696472831267875e-06, "loss": 2.2071, "step": 33799 }, { "epoch": 6.44, "learning_rate": 1.0695900857959963e-06, "loss": 2.1393, "step": 33800 }, { "epoch": 6.44, "learning_rate": 1.069532888465205e-06, "loss": 1.976, "step": 33801 }, { "epoch": 6.44, "learning_rate": 1.0694756911344137e-06, "loss": 0.9956, "step": 33802 }, { "epoch": 6.44, "learning_rate": 1.0694184938036224e-06, "loss": 2.5457, "step": 33803 }, { "epoch": 6.44, "learning_rate": 1.0693612964728312e-06, "loss": 2.2082, "step": 33804 }, { "epoch": 6.44, "learning_rate": 1.06930409914204e-06, "loss": 1.8739, "step": 33805 }, { "epoch": 6.44, "learning_rate": 1.0692469018112488e-06, "loss": 0.8089, "step": 33806 }, { "epoch": 6.44, "learning_rate": 1.0691897044804576e-06, "loss": 1.0886, "step": 33807 }, { "epoch": 6.44, "learning_rate": 1.0691325071496664e-06, "loss": 1.2513, "step": 33808 }, { "epoch": 6.44, "learning_rate": 1.069075309818875e-06, "loss": 1.9472, "step": 33809 }, { "epoch": 6.44, "learning_rate": 1.069018112488084e-06, "loss": 1.1727, "step": 33810 }, { "epoch": 6.44, "learning_rate": 1.0689609151572928e-06, "loss": 2.7032, "step": 33811 }, { "epoch": 6.44, "learning_rate": 1.0689037178265016e-06, "loss": 1.7262, "step": 33812 }, { "epoch": 6.44, "learning_rate": 1.0688465204957102e-06, "loss": 2.0632, "step": 33813 }, { "epoch": 6.44, "learning_rate": 1.068789323164919e-06, "loss": 2.085, "step": 33814 }, { "epoch": 6.44, "learning_rate": 1.0687321258341278e-06, "loss": 1.9497, "step": 33815 }, { "epoch": 6.44, "learning_rate": 1.0686749285033366e-06, "loss": 3.3294, "step": 33816 }, { "epoch": 6.44, "learning_rate": 1.0686177311725453e-06, "loss": 1.3902, "step": 33817 }, { "epoch": 6.44, "learning_rate": 1.0685605338417541e-06, "loss": 1.1167, "step": 33818 }, { "epoch": 6.44, "learning_rate": 1.0685033365109627e-06, "loss": 1.6627, "step": 33819 }, { "epoch": 6.44, "learning_rate": 1.0684461391801715e-06, "loss": 0.6955, "step": 33820 }, { "epoch": 6.44, "learning_rate": 1.0683889418493803e-06, "loss": 1.9895, "step": 33821 }, { "epoch": 6.44, "learning_rate": 1.0683317445185893e-06, "loss": 1.8474, "step": 33822 }, { "epoch": 6.44, "learning_rate": 1.068274547187798e-06, "loss": 2.1349, "step": 33823 }, { "epoch": 6.44, "learning_rate": 1.0682173498570067e-06, "loss": 1.4611, "step": 33824 }, { "epoch": 6.44, "learning_rate": 1.0681601525262155e-06, "loss": 1.4864, "step": 33825 }, { "epoch": 6.44, "learning_rate": 1.068102955195424e-06, "loss": 0.9835, "step": 33826 }, { "epoch": 6.44, "learning_rate": 1.068045757864633e-06, "loss": 1.5693, "step": 33827 }, { "epoch": 6.44, "learning_rate": 1.0679885605338419e-06, "loss": 3.0929, "step": 33828 }, { "epoch": 6.44, "learning_rate": 1.0679313632030507e-06, "loss": 2.8297, "step": 33829 }, { "epoch": 6.44, "learning_rate": 1.0678741658722592e-06, "loss": 1.1483, "step": 33830 }, { "epoch": 6.44, "learning_rate": 1.067816968541468e-06, "loss": 0.8759, "step": 33831 }, { "epoch": 6.44, "learning_rate": 1.0677597712106768e-06, "loss": 1.9591, "step": 33832 }, { "epoch": 6.44, "learning_rate": 1.0677025738798856e-06, "loss": 0.9896, "step": 33833 }, { "epoch": 6.44, "learning_rate": 1.0676453765490944e-06, "loss": 1.7456, "step": 33834 }, { "epoch": 6.44, "learning_rate": 1.0675881792183032e-06, "loss": 1.3936, "step": 33835 }, { "epoch": 6.44, "learning_rate": 1.067530981887512e-06, "loss": 3.2275, "step": 33836 }, { "epoch": 6.45, "learning_rate": 1.0674737845567206e-06, "loss": 1.1768, "step": 33837 }, { "epoch": 6.45, "learning_rate": 1.0674165872259296e-06, "loss": 1.4069, "step": 33838 }, { "epoch": 6.45, "learning_rate": 1.0673593898951384e-06, "loss": 0.7812, "step": 33839 }, { "epoch": 6.45, "learning_rate": 1.067302192564347e-06, "loss": 3.2212, "step": 33840 }, { "epoch": 6.45, "learning_rate": 1.0672449952335558e-06, "loss": 1.7026, "step": 33841 }, { "epoch": 6.45, "learning_rate": 1.0671877979027646e-06, "loss": 2.3387, "step": 33842 }, { "epoch": 6.45, "learning_rate": 1.0671306005719734e-06, "loss": 0.9089, "step": 33843 }, { "epoch": 6.45, "learning_rate": 1.0670734032411821e-06, "loss": 1.6146, "step": 33844 }, { "epoch": 6.45, "learning_rate": 1.067016205910391e-06, "loss": 2.3721, "step": 33845 }, { "epoch": 6.45, "learning_rate": 1.0669590085795997e-06, "loss": 2.3328, "step": 33846 }, { "epoch": 6.45, "learning_rate": 1.0669018112488083e-06, "loss": 1.3256, "step": 33847 }, { "epoch": 6.45, "learning_rate": 1.0668446139180171e-06, "loss": 1.0542, "step": 33848 }, { "epoch": 6.45, "learning_rate": 1.066787416587226e-06, "loss": 1.9434, "step": 33849 }, { "epoch": 6.45, "learning_rate": 1.0667302192564347e-06, "loss": 1.6408, "step": 33850 }, { "epoch": 6.45, "learning_rate": 1.0666730219256435e-06, "loss": 1.139, "step": 33851 }, { "epoch": 6.45, "learning_rate": 1.0666158245948523e-06, "loss": 3.097, "step": 33852 }, { "epoch": 6.45, "learning_rate": 1.066558627264061e-06, "loss": 2.1944, "step": 33853 }, { "epoch": 6.45, "learning_rate": 1.0665014299332699e-06, "loss": 0.8183, "step": 33854 }, { "epoch": 6.45, "learning_rate": 1.0664442326024787e-06, "loss": 1.9251, "step": 33855 }, { "epoch": 6.45, "learning_rate": 1.0663870352716875e-06, "loss": 2.1956, "step": 33856 }, { "epoch": 6.45, "learning_rate": 1.066329837940896e-06, "loss": 1.9383, "step": 33857 }, { "epoch": 6.45, "learning_rate": 1.0662726406101048e-06, "loss": 0.5618, "step": 33858 }, { "epoch": 6.45, "learning_rate": 1.0662154432793136e-06, "loss": 3.1375, "step": 33859 }, { "epoch": 6.45, "learning_rate": 1.0661582459485224e-06, "loss": 2.5534, "step": 33860 }, { "epoch": 6.45, "learning_rate": 1.0661010486177312e-06, "loss": 1.0367, "step": 33861 }, { "epoch": 6.45, "learning_rate": 1.06604385128694e-06, "loss": 1.4448, "step": 33862 }, { "epoch": 6.45, "learning_rate": 1.0659866539561488e-06, "loss": 1.4748, "step": 33863 }, { "epoch": 6.45, "learning_rate": 1.0659294566253574e-06, "loss": 2.5743, "step": 33864 }, { "epoch": 6.45, "learning_rate": 1.0658722592945662e-06, "loss": 2.2942, "step": 33865 }, { "epoch": 6.45, "learning_rate": 1.0658150619637752e-06, "loss": 0.9864, "step": 33866 }, { "epoch": 6.45, "learning_rate": 1.0657578646329838e-06, "loss": 1.3369, "step": 33867 }, { "epoch": 6.45, "learning_rate": 1.0657006673021926e-06, "loss": 0.9669, "step": 33868 }, { "epoch": 6.45, "learning_rate": 1.0656434699714014e-06, "loss": 1.7686, "step": 33869 }, { "epoch": 6.45, "learning_rate": 1.0655862726406101e-06, "loss": 1.3683, "step": 33870 }, { "epoch": 6.45, "learning_rate": 1.065529075309819e-06, "loss": 0.9345, "step": 33871 }, { "epoch": 6.45, "learning_rate": 1.0654718779790277e-06, "loss": 0.8672, "step": 33872 }, { "epoch": 6.45, "learning_rate": 1.0654146806482365e-06, "loss": 1.0282, "step": 33873 }, { "epoch": 6.45, "learning_rate": 1.0653574833174451e-06, "loss": 1.7271, "step": 33874 }, { "epoch": 6.45, "learning_rate": 1.065300285986654e-06, "loss": 2.7416, "step": 33875 }, { "epoch": 6.45, "learning_rate": 1.0652430886558627e-06, "loss": 1.6724, "step": 33876 }, { "epoch": 6.45, "learning_rate": 1.0651858913250717e-06, "loss": 1.9143, "step": 33877 }, { "epoch": 6.45, "learning_rate": 1.0651286939942803e-06, "loss": 1.2798, "step": 33878 }, { "epoch": 6.45, "learning_rate": 1.065071496663489e-06, "loss": 2.5168, "step": 33879 }, { "epoch": 6.45, "learning_rate": 1.0650142993326979e-06, "loss": 2.3239, "step": 33880 }, { "epoch": 6.45, "learning_rate": 1.0649571020019065e-06, "loss": 2.1415, "step": 33881 }, { "epoch": 6.45, "learning_rate": 1.0648999046711155e-06, "loss": 1.6627, "step": 33882 }, { "epoch": 6.45, "learning_rate": 1.0648427073403243e-06, "loss": 1.6044, "step": 33883 }, { "epoch": 6.45, "learning_rate": 1.0647855100095328e-06, "loss": 2.1764, "step": 33884 }, { "epoch": 6.45, "learning_rate": 1.0647283126787416e-06, "loss": 2.1254, "step": 33885 }, { "epoch": 6.45, "learning_rate": 1.0646711153479504e-06, "loss": 1.3292, "step": 33886 }, { "epoch": 6.45, "learning_rate": 1.0646139180171592e-06, "loss": 1.6036, "step": 33887 }, { "epoch": 6.45, "learning_rate": 1.064556720686368e-06, "loss": 1.3414, "step": 33888 }, { "epoch": 6.46, "learning_rate": 1.0644995233555768e-06, "loss": 1.2133, "step": 33889 }, { "epoch": 6.46, "learning_rate": 1.0644423260247856e-06, "loss": 1.2802, "step": 33890 }, { "epoch": 6.46, "learning_rate": 1.0643851286939942e-06, "loss": 1.9214, "step": 33891 }, { "epoch": 6.46, "learning_rate": 1.064327931363203e-06, "loss": 2.7344, "step": 33892 }, { "epoch": 6.46, "learning_rate": 1.0642707340324118e-06, "loss": 1.4391, "step": 33893 }, { "epoch": 6.46, "learning_rate": 1.0642135367016208e-06, "loss": 1.6306, "step": 33894 }, { "epoch": 6.46, "learning_rate": 1.0641563393708294e-06, "loss": 1.3912, "step": 33895 }, { "epoch": 6.46, "learning_rate": 1.0640991420400382e-06, "loss": 1.1912, "step": 33896 }, { "epoch": 6.46, "learning_rate": 1.064041944709247e-06, "loss": 1.08, "step": 33897 }, { "epoch": 6.46, "learning_rate": 1.0639847473784557e-06, "loss": 1.5791, "step": 33898 }, { "epoch": 6.46, "learning_rate": 1.0639275500476645e-06, "loss": 1.8916, "step": 33899 }, { "epoch": 6.46, "learning_rate": 1.0638703527168733e-06, "loss": 1.1186, "step": 33900 }, { "epoch": 6.46, "learning_rate": 1.0638131553860821e-06, "loss": 1.0975, "step": 33901 }, { "epoch": 6.46, "learning_rate": 1.0637559580552907e-06, "loss": 1.9357, "step": 33902 }, { "epoch": 6.46, "learning_rate": 1.0636987607244995e-06, "loss": 1.2935, "step": 33903 }, { "epoch": 6.46, "learning_rate": 1.0636415633937083e-06, "loss": 1.9209, "step": 33904 }, { "epoch": 6.46, "learning_rate": 1.063584366062917e-06, "loss": 2.6468, "step": 33905 }, { "epoch": 6.46, "learning_rate": 1.0635271687321259e-06, "loss": 1.3303, "step": 33906 }, { "epoch": 6.46, "learning_rate": 1.0634699714013347e-06, "loss": 1.5759, "step": 33907 }, { "epoch": 6.46, "learning_rate": 1.0634127740705433e-06, "loss": 2.0189, "step": 33908 }, { "epoch": 6.46, "learning_rate": 1.063355576739752e-06, "loss": 2.1749, "step": 33909 }, { "epoch": 6.46, "learning_rate": 1.063298379408961e-06, "loss": 1.6196, "step": 33910 }, { "epoch": 6.46, "learning_rate": 1.0632411820781698e-06, "loss": 1.8794, "step": 33911 }, { "epoch": 6.46, "learning_rate": 1.0631839847473784e-06, "loss": 2.3933, "step": 33912 }, { "epoch": 6.46, "learning_rate": 1.0631267874165872e-06, "loss": 2.8734, "step": 33913 }, { "epoch": 6.46, "learning_rate": 1.063069590085796e-06, "loss": 0.5791, "step": 33914 }, { "epoch": 6.46, "learning_rate": 1.0630123927550048e-06, "loss": 0.7643, "step": 33915 }, { "epoch": 6.46, "learning_rate": 1.0629551954242136e-06, "loss": 1.8652, "step": 33916 }, { "epoch": 6.46, "learning_rate": 1.0628979980934224e-06, "loss": 1.8649, "step": 33917 }, { "epoch": 6.46, "learning_rate": 1.0628408007626312e-06, "loss": 1.1598, "step": 33918 }, { "epoch": 6.46, "learning_rate": 1.0627836034318398e-06, "loss": 1.9493, "step": 33919 }, { "epoch": 6.46, "learning_rate": 1.0627264061010486e-06, "loss": 2.6129, "step": 33920 }, { "epoch": 6.46, "learning_rate": 1.0626692087702576e-06, "loss": 1.3458, "step": 33921 }, { "epoch": 6.46, "learning_rate": 1.0626120114394662e-06, "loss": 1.3668, "step": 33922 }, { "epoch": 6.46, "learning_rate": 1.062554814108675e-06, "loss": 1.7556, "step": 33923 }, { "epoch": 6.46, "learning_rate": 1.0624976167778837e-06, "loss": 1.9515, "step": 33924 }, { "epoch": 6.46, "learning_rate": 1.0624404194470923e-06, "loss": 1.2633, "step": 33925 }, { "epoch": 6.46, "learning_rate": 1.0623832221163013e-06, "loss": 2.308, "step": 33926 }, { "epoch": 6.46, "learning_rate": 1.0623260247855101e-06, "loss": 2.5635, "step": 33927 }, { "epoch": 6.46, "learning_rate": 1.062268827454719e-06, "loss": 1.3029, "step": 33928 }, { "epoch": 6.46, "learning_rate": 1.0622116301239275e-06, "loss": 1.9322, "step": 33929 }, { "epoch": 6.46, "learning_rate": 1.0621544327931363e-06, "loss": 1.5978, "step": 33930 }, { "epoch": 6.46, "learning_rate": 1.062097235462345e-06, "loss": 0.9157, "step": 33931 }, { "epoch": 6.46, "learning_rate": 1.0620400381315539e-06, "loss": 2.4409, "step": 33932 }, { "epoch": 6.46, "learning_rate": 1.0619828408007627e-06, "loss": 1.659, "step": 33933 }, { "epoch": 6.46, "learning_rate": 1.0619256434699715e-06, "loss": 1.9885, "step": 33934 }, { "epoch": 6.46, "learning_rate": 1.0618684461391803e-06, "loss": 2.3103, "step": 33935 }, { "epoch": 6.46, "learning_rate": 1.0618112488083888e-06, "loss": 2.4596, "step": 33936 }, { "epoch": 6.46, "learning_rate": 1.0617540514775976e-06, "loss": 1.7645, "step": 33937 }, { "epoch": 6.46, "learning_rate": 1.0616968541468066e-06, "loss": 1.8422, "step": 33938 }, { "epoch": 6.46, "learning_rate": 1.0616396568160152e-06, "loss": 1.645, "step": 33939 }, { "epoch": 6.46, "learning_rate": 1.061582459485224e-06, "loss": 1.105, "step": 33940 }, { "epoch": 6.46, "learning_rate": 1.0615252621544328e-06, "loss": 1.9608, "step": 33941 }, { "epoch": 6.47, "learning_rate": 1.0614680648236416e-06, "loss": 1.7583, "step": 33942 }, { "epoch": 6.47, "learning_rate": 1.0614108674928504e-06, "loss": 1.2948, "step": 33943 }, { "epoch": 6.47, "learning_rate": 1.0613536701620592e-06, "loss": 1.9643, "step": 33944 }, { "epoch": 6.47, "learning_rate": 1.061296472831268e-06, "loss": 1.6965, "step": 33945 }, { "epoch": 6.47, "learning_rate": 1.0612392755004766e-06, "loss": 1.2249, "step": 33946 }, { "epoch": 6.47, "learning_rate": 1.0611820781696854e-06, "loss": 0.8015, "step": 33947 }, { "epoch": 6.47, "learning_rate": 1.0611248808388942e-06, "loss": 1.8124, "step": 33948 }, { "epoch": 6.47, "learning_rate": 1.061067683508103e-06, "loss": 1.2746, "step": 33949 }, { "epoch": 6.47, "learning_rate": 1.0610104861773117e-06, "loss": 2.0097, "step": 33950 }, { "epoch": 6.47, "learning_rate": 1.0609532888465205e-06, "loss": 1.4584, "step": 33951 }, { "epoch": 6.47, "learning_rate": 1.0608960915157293e-06, "loss": 3.7464, "step": 33952 }, { "epoch": 6.47, "learning_rate": 1.060838894184938e-06, "loss": 2.3201, "step": 33953 }, { "epoch": 6.47, "learning_rate": 1.060781696854147e-06, "loss": 0.8449, "step": 33954 }, { "epoch": 6.47, "learning_rate": 1.0607244995233557e-06, "loss": 1.1705, "step": 33955 }, { "epoch": 6.47, "learning_rate": 1.0606673021925643e-06, "loss": 2.0899, "step": 33956 }, { "epoch": 6.47, "learning_rate": 1.060610104861773e-06, "loss": 2.585, "step": 33957 }, { "epoch": 6.47, "learning_rate": 1.0605529075309819e-06, "loss": 1.0735, "step": 33958 }, { "epoch": 6.47, "learning_rate": 1.0604957102001907e-06, "loss": 1.6873, "step": 33959 }, { "epoch": 6.47, "learning_rate": 1.0604385128693995e-06, "loss": 2.3939, "step": 33960 }, { "epoch": 6.47, "learning_rate": 1.0603813155386083e-06, "loss": 1.5759, "step": 33961 }, { "epoch": 6.47, "learning_rate": 1.060324118207817e-06, "loss": 1.5783, "step": 33962 }, { "epoch": 6.47, "learning_rate": 1.0602669208770256e-06, "loss": 1.5926, "step": 33963 }, { "epoch": 6.47, "learning_rate": 1.0602097235462344e-06, "loss": 1.7173, "step": 33964 }, { "epoch": 6.47, "learning_rate": 1.0601525262154434e-06, "loss": 1.0529, "step": 33965 }, { "epoch": 6.47, "learning_rate": 1.0600953288846522e-06, "loss": 1.9248, "step": 33966 }, { "epoch": 6.47, "learning_rate": 1.0600381315538608e-06, "loss": 1.6433, "step": 33967 }, { "epoch": 6.47, "learning_rate": 1.0599809342230696e-06, "loss": 1.2887, "step": 33968 }, { "epoch": 6.47, "learning_rate": 1.0599237368922784e-06, "loss": 1.2767, "step": 33969 }, { "epoch": 6.47, "learning_rate": 1.0598665395614872e-06, "loss": 1.986, "step": 33970 }, { "epoch": 6.47, "learning_rate": 1.059809342230696e-06, "loss": 1.2387, "step": 33971 }, { "epoch": 6.47, "learning_rate": 1.0597521448999048e-06, "loss": 1.6326, "step": 33972 }, { "epoch": 6.47, "learning_rate": 1.0596949475691134e-06, "loss": 2.4181, "step": 33973 }, { "epoch": 6.47, "learning_rate": 1.0596377502383222e-06, "loss": 1.7497, "step": 33974 }, { "epoch": 6.47, "learning_rate": 1.059580552907531e-06, "loss": 3.7855, "step": 33975 }, { "epoch": 6.47, "learning_rate": 1.0595233555767398e-06, "loss": 1.8604, "step": 33976 }, { "epoch": 6.47, "learning_rate": 1.0594661582459485e-06, "loss": 1.1858, "step": 33977 }, { "epoch": 6.47, "learning_rate": 1.0594089609151573e-06, "loss": 2.4263, "step": 33978 }, { "epoch": 6.47, "learning_rate": 1.0593517635843661e-06, "loss": 2.8167, "step": 33979 }, { "epoch": 6.47, "learning_rate": 1.0592945662535747e-06, "loss": 0.4199, "step": 33980 }, { "epoch": 6.47, "learning_rate": 1.0592373689227835e-06, "loss": 2.2797, "step": 33981 }, { "epoch": 6.47, "learning_rate": 1.0591801715919925e-06, "loss": 2.4983, "step": 33982 }, { "epoch": 6.47, "learning_rate": 1.0591229742612013e-06, "loss": 1.4587, "step": 33983 }, { "epoch": 6.47, "learning_rate": 1.0590657769304099e-06, "loss": 1.2649, "step": 33984 }, { "epoch": 6.47, "learning_rate": 1.0590085795996187e-06, "loss": 1.6689, "step": 33985 }, { "epoch": 6.47, "learning_rate": 1.0589513822688275e-06, "loss": 2.6181, "step": 33986 }, { "epoch": 6.47, "learning_rate": 1.0588941849380363e-06, "loss": 1.6471, "step": 33987 }, { "epoch": 6.47, "learning_rate": 1.058836987607245e-06, "loss": 2.3323, "step": 33988 }, { "epoch": 6.47, "learning_rate": 1.0587797902764539e-06, "loss": 2.7831, "step": 33989 }, { "epoch": 6.47, "learning_rate": 1.0587225929456624e-06, "loss": 2.9442, "step": 33990 }, { "epoch": 6.47, "learning_rate": 1.0586653956148712e-06, "loss": 2.3407, "step": 33991 }, { "epoch": 6.47, "learning_rate": 1.05860819828408e-06, "loss": 2.4969, "step": 33992 }, { "epoch": 6.47, "learning_rate": 1.058551000953289e-06, "loss": 1.2587, "step": 33993 }, { "epoch": 6.48, "learning_rate": 1.0584938036224976e-06, "loss": 1.7938, "step": 33994 }, { "epoch": 6.48, "learning_rate": 1.0584366062917064e-06, "loss": 1.6601, "step": 33995 }, { "epoch": 6.48, "learning_rate": 1.0583794089609152e-06, "loss": 1.8581, "step": 33996 }, { "epoch": 6.48, "learning_rate": 1.0583222116301238e-06, "loss": 2.0304, "step": 33997 }, { "epoch": 6.48, "learning_rate": 1.0582650142993328e-06, "loss": 1.5892, "step": 33998 }, { "epoch": 6.48, "learning_rate": 1.0582078169685416e-06, "loss": 2.737, "step": 33999 }, { "epoch": 6.48, "learning_rate": 1.0581506196377504e-06, "loss": 3.0162, "step": 34000 }, { "epoch": 6.48, "learning_rate": 1.058093422306959e-06, "loss": 3.6594, "step": 34001 }, { "epoch": 6.48, "learning_rate": 1.0580362249761678e-06, "loss": 0.7381, "step": 34002 }, { "epoch": 6.48, "learning_rate": 1.0579790276453766e-06, "loss": 2.0909, "step": 34003 }, { "epoch": 6.48, "learning_rate": 1.0579218303145853e-06, "loss": 1.3707, "step": 34004 }, { "epoch": 6.48, "learning_rate": 1.0578646329837941e-06, "loss": 1.6533, "step": 34005 }, { "epoch": 6.48, "learning_rate": 1.057807435653003e-06, "loss": 2.3665, "step": 34006 }, { "epoch": 6.48, "learning_rate": 1.0577502383222117e-06, "loss": 1.706, "step": 34007 }, { "epoch": 6.48, "learning_rate": 1.0576930409914203e-06, "loss": 1.2174, "step": 34008 }, { "epoch": 6.48, "learning_rate": 1.0576358436606293e-06, "loss": 0.8588, "step": 34009 }, { "epoch": 6.48, "learning_rate": 1.0575786463298381e-06, "loss": 1.2126, "step": 34010 }, { "epoch": 6.48, "learning_rate": 1.0575214489990467e-06, "loss": 1.3846, "step": 34011 }, { "epoch": 6.48, "learning_rate": 1.0574642516682555e-06, "loss": 2.9032, "step": 34012 }, { "epoch": 6.48, "learning_rate": 1.0574070543374643e-06, "loss": 2.5979, "step": 34013 }, { "epoch": 6.48, "learning_rate": 1.057349857006673e-06, "loss": 0.606, "step": 34014 }, { "epoch": 6.48, "learning_rate": 1.0572926596758819e-06, "loss": 1.7812, "step": 34015 }, { "epoch": 6.48, "learning_rate": 1.0572354623450907e-06, "loss": 0.7085, "step": 34016 }, { "epoch": 6.48, "learning_rate": 1.0571782650142995e-06, "loss": 2.0272, "step": 34017 }, { "epoch": 6.48, "learning_rate": 1.057121067683508e-06, "loss": 0.8674, "step": 34018 }, { "epoch": 6.48, "learning_rate": 1.0570638703527168e-06, "loss": 1.284, "step": 34019 }, { "epoch": 6.48, "learning_rate": 1.0570066730219256e-06, "loss": 2.3132, "step": 34020 }, { "epoch": 6.48, "learning_rate": 1.0569494756911344e-06, "loss": 2.3108, "step": 34021 }, { "epoch": 6.48, "learning_rate": 1.0568922783603432e-06, "loss": 1.7829, "step": 34022 }, { "epoch": 6.48, "learning_rate": 1.056835081029552e-06, "loss": 2.1258, "step": 34023 }, { "epoch": 6.48, "learning_rate": 1.0567778836987608e-06, "loss": 0.9687, "step": 34024 }, { "epoch": 6.48, "learning_rate": 1.0567206863679694e-06, "loss": 2.5348, "step": 34025 }, { "epoch": 6.48, "learning_rate": 1.0566634890371784e-06, "loss": 2.577, "step": 34026 }, { "epoch": 6.48, "learning_rate": 1.0566062917063872e-06, "loss": 1.0756, "step": 34027 }, { "epoch": 6.48, "learning_rate": 1.0565490943755958e-06, "loss": 0.9407, "step": 34028 }, { "epoch": 6.48, "learning_rate": 1.0564918970448046e-06, "loss": 0.9464, "step": 34029 }, { "epoch": 6.48, "learning_rate": 1.0564346997140133e-06, "loss": 3.2541, "step": 34030 }, { "epoch": 6.48, "learning_rate": 1.0563775023832221e-06, "loss": 2.7442, "step": 34031 }, { "epoch": 6.48, "learning_rate": 1.056320305052431e-06, "loss": 1.4558, "step": 34032 }, { "epoch": 6.48, "learning_rate": 1.0562631077216397e-06, "loss": 2.3661, "step": 34033 }, { "epoch": 6.48, "learning_rate": 1.0562059103908485e-06, "loss": 1.4363, "step": 34034 }, { "epoch": 6.48, "learning_rate": 1.056148713060057e-06, "loss": 1.1762, "step": 34035 }, { "epoch": 6.48, "learning_rate": 1.056091515729266e-06, "loss": 1.281, "step": 34036 }, { "epoch": 6.48, "learning_rate": 1.056034318398475e-06, "loss": 2.4135, "step": 34037 }, { "epoch": 6.48, "learning_rate": 1.0559771210676835e-06, "loss": 1.5604, "step": 34038 }, { "epoch": 6.48, "learning_rate": 1.0559199237368923e-06, "loss": 1.4173, "step": 34039 }, { "epoch": 6.48, "learning_rate": 1.055862726406101e-06, "loss": 1.841, "step": 34040 }, { "epoch": 6.48, "learning_rate": 1.0558055290753099e-06, "loss": 0.9136, "step": 34041 }, { "epoch": 6.48, "learning_rate": 1.0557483317445187e-06, "loss": 2.2068, "step": 34042 }, { "epoch": 6.48, "learning_rate": 1.0556911344137275e-06, "loss": 2.791, "step": 34043 }, { "epoch": 6.48, "learning_rate": 1.0556339370829362e-06, "loss": 2.8217, "step": 34044 }, { "epoch": 6.48, "learning_rate": 1.0555767397521448e-06, "loss": 2.2066, "step": 34045 }, { "epoch": 6.48, "learning_rate": 1.0555195424213536e-06, "loss": 2.2588, "step": 34046 }, { "epoch": 6.49, "learning_rate": 1.0554623450905624e-06, "loss": 2.1963, "step": 34047 }, { "epoch": 6.49, "learning_rate": 1.0554051477597712e-06, "loss": 2.7053, "step": 34048 }, { "epoch": 6.49, "learning_rate": 1.05534795042898e-06, "loss": 1.2101, "step": 34049 }, { "epoch": 6.49, "learning_rate": 1.0552907530981888e-06, "loss": 0.9381, "step": 34050 }, { "epoch": 6.49, "learning_rate": 1.0552335557673976e-06, "loss": 1.4484, "step": 34051 }, { "epoch": 6.49, "learning_rate": 1.0551763584366062e-06, "loss": 1.4263, "step": 34052 }, { "epoch": 6.49, "learning_rate": 1.0551191611058152e-06, "loss": 2.4052, "step": 34053 }, { "epoch": 6.49, "learning_rate": 1.055061963775024e-06, "loss": 3.1196, "step": 34054 }, { "epoch": 6.49, "learning_rate": 1.0550047664442326e-06, "loss": 2.0201, "step": 34055 }, { "epoch": 6.49, "learning_rate": 1.0549475691134414e-06, "loss": 1.3109, "step": 34056 }, { "epoch": 6.49, "learning_rate": 1.0548903717826501e-06, "loss": 0.9254, "step": 34057 }, { "epoch": 6.49, "learning_rate": 1.054833174451859e-06, "loss": 2.659, "step": 34058 }, { "epoch": 6.49, "learning_rate": 1.0547759771210677e-06, "loss": 1.6091, "step": 34059 }, { "epoch": 6.49, "learning_rate": 1.0547187797902765e-06, "loss": 1.2533, "step": 34060 }, { "epoch": 6.49, "learning_rate": 1.0546615824594853e-06, "loss": 0.659, "step": 34061 }, { "epoch": 6.49, "learning_rate": 1.054604385128694e-06, "loss": 1.4248, "step": 34062 }, { "epoch": 6.49, "learning_rate": 1.0545471877979027e-06, "loss": 2.2428, "step": 34063 }, { "epoch": 6.49, "learning_rate": 1.0544899904671115e-06, "loss": 0.9497, "step": 34064 }, { "epoch": 6.49, "learning_rate": 1.0544327931363205e-06, "loss": 1.9773, "step": 34065 }, { "epoch": 6.49, "learning_rate": 1.054375595805529e-06, "loss": 1.3874, "step": 34066 }, { "epoch": 6.49, "learning_rate": 1.0543183984747379e-06, "loss": 1.0391, "step": 34067 }, { "epoch": 6.49, "learning_rate": 1.0542612011439467e-06, "loss": 2.3299, "step": 34068 }, { "epoch": 6.49, "learning_rate": 1.0542040038131552e-06, "loss": 1.4009, "step": 34069 }, { "epoch": 6.49, "learning_rate": 1.0541468064823643e-06, "loss": 1.3304, "step": 34070 }, { "epoch": 6.49, "learning_rate": 1.054089609151573e-06, "loss": 1.2756, "step": 34071 }, { "epoch": 6.49, "learning_rate": 1.0540324118207818e-06, "loss": 2.6605, "step": 34072 }, { "epoch": 6.49, "learning_rate": 1.0539752144899904e-06, "loss": 1.8777, "step": 34073 }, { "epoch": 6.49, "learning_rate": 1.0539180171591992e-06, "loss": 1.203, "step": 34074 }, { "epoch": 6.49, "learning_rate": 1.053860819828408e-06, "loss": 1.6074, "step": 34075 }, { "epoch": 6.49, "learning_rate": 1.0538036224976168e-06, "loss": 2.7092, "step": 34076 }, { "epoch": 6.49, "learning_rate": 1.0537464251668256e-06, "loss": 2.5565, "step": 34077 }, { "epoch": 6.49, "learning_rate": 1.0536892278360344e-06, "loss": 2.0547, "step": 34078 }, { "epoch": 6.49, "learning_rate": 1.053632030505243e-06, "loss": 1.9183, "step": 34079 }, { "epoch": 6.49, "learning_rate": 1.0535748331744518e-06, "loss": 1.8501, "step": 34080 }, { "epoch": 6.49, "learning_rate": 1.0535176358436608e-06, "loss": 1.3743, "step": 34081 }, { "epoch": 6.49, "learning_rate": 1.0534604385128696e-06, "loss": 2.0812, "step": 34082 }, { "epoch": 6.49, "learning_rate": 1.0534032411820781e-06, "loss": 1.7341, "step": 34083 }, { "epoch": 6.49, "learning_rate": 1.053346043851287e-06, "loss": 2.9164, "step": 34084 }, { "epoch": 6.49, "learning_rate": 1.0532888465204957e-06, "loss": 1.5245, "step": 34085 }, { "epoch": 6.49, "learning_rate": 1.0532316491897045e-06, "loss": 1.6947, "step": 34086 }, { "epoch": 6.49, "learning_rate": 1.0531744518589133e-06, "loss": 1.711, "step": 34087 }, { "epoch": 6.49, "learning_rate": 1.0531172545281221e-06, "loss": 2.422, "step": 34088 }, { "epoch": 6.49, "learning_rate": 1.053060057197331e-06, "loss": 2.4778, "step": 34089 }, { "epoch": 6.49, "learning_rate": 1.0530028598665395e-06, "loss": 2.2273, "step": 34090 }, { "epoch": 6.49, "learning_rate": 1.0529456625357483e-06, "loss": 2.1153, "step": 34091 }, { "epoch": 6.49, "learning_rate": 1.052888465204957e-06, "loss": 2.4346, "step": 34092 }, { "epoch": 6.49, "learning_rate": 1.0528312678741659e-06, "loss": 1.9747, "step": 34093 }, { "epoch": 6.49, "learning_rate": 1.0527740705433747e-06, "loss": 2.8077, "step": 34094 }, { "epoch": 6.49, "learning_rate": 1.0527168732125835e-06, "loss": 1.2738, "step": 34095 }, { "epoch": 6.49, "learning_rate": 1.052659675881792e-06, "loss": 1.5965, "step": 34096 }, { "epoch": 6.49, "learning_rate": 1.052602478551001e-06, "loss": 2.1324, "step": 34097 }, { "epoch": 6.49, "learning_rate": 1.0525452812202098e-06, "loss": 2.4009, "step": 34098 }, { "epoch": 6.5, "learning_rate": 1.0524880838894186e-06, "loss": 1.8062, "step": 34099 }, { "epoch": 6.5, "learning_rate": 1.0524308865586272e-06, "loss": 1.2374, "step": 34100 }, { "epoch": 6.5, "learning_rate": 1.052373689227836e-06, "loss": 3.3097, "step": 34101 }, { "epoch": 6.5, "learning_rate": 1.0523164918970448e-06, "loss": 3.002, "step": 34102 }, { "epoch": 6.5, "learning_rate": 1.0522592945662536e-06, "loss": 2.8266, "step": 34103 }, { "epoch": 6.5, "learning_rate": 1.0522020972354624e-06, "loss": 1.4226, "step": 34104 }, { "epoch": 6.5, "learning_rate": 1.0521448999046712e-06, "loss": 1.248, "step": 34105 }, { "epoch": 6.5, "learning_rate": 1.05208770257388e-06, "loss": 2.3855, "step": 34106 }, { "epoch": 6.5, "learning_rate": 1.0520305052430886e-06, "loss": 1.2167, "step": 34107 }, { "epoch": 6.5, "learning_rate": 1.0519733079122974e-06, "loss": 2.3398, "step": 34108 }, { "epoch": 6.5, "learning_rate": 1.0519161105815064e-06, "loss": 2.3423, "step": 34109 }, { "epoch": 6.5, "learning_rate": 1.051858913250715e-06, "loss": 1.4725, "step": 34110 }, { "epoch": 6.5, "learning_rate": 1.0518017159199237e-06, "loss": 0.5703, "step": 34111 }, { "epoch": 6.5, "learning_rate": 1.0517445185891325e-06, "loss": 1.9806, "step": 34112 }, { "epoch": 6.5, "learning_rate": 1.0516873212583413e-06, "loss": 1.1547, "step": 34113 }, { "epoch": 6.5, "learning_rate": 1.0516301239275501e-06, "loss": 1.1661, "step": 34114 }, { "epoch": 6.5, "learning_rate": 1.051572926596759e-06, "loss": 1.7189, "step": 34115 }, { "epoch": 6.5, "learning_rate": 1.0515157292659677e-06, "loss": 1.4064, "step": 34116 }, { "epoch": 6.5, "learning_rate": 1.0514585319351763e-06, "loss": 1.2652, "step": 34117 }, { "epoch": 6.5, "learning_rate": 1.051401334604385e-06, "loss": 0.647, "step": 34118 }, { "epoch": 6.5, "learning_rate": 1.0513441372735939e-06, "loss": 2.1182, "step": 34119 }, { "epoch": 6.5, "learning_rate": 1.0512869399428027e-06, "loss": 2.1632, "step": 34120 }, { "epoch": 6.5, "learning_rate": 1.0512297426120115e-06, "loss": 2.2196, "step": 34121 }, { "epoch": 6.5, "learning_rate": 1.0511725452812203e-06, "loss": 1.9768, "step": 34122 }, { "epoch": 6.5, "learning_rate": 1.051115347950429e-06, "loss": 1.3124, "step": 34123 }, { "epoch": 6.5, "learning_rate": 1.0510581506196376e-06, "loss": 1.0699, "step": 34124 }, { "epoch": 6.5, "learning_rate": 1.0510009532888466e-06, "loss": 2.3369, "step": 34125 }, { "epoch": 6.5, "learning_rate": 1.0509437559580554e-06, "loss": 2.548, "step": 34126 }, { "epoch": 6.5, "learning_rate": 1.050886558627264e-06, "loss": 2.8589, "step": 34127 }, { "epoch": 6.5, "learning_rate": 1.0508293612964728e-06, "loss": 1.5771, "step": 34128 }, { "epoch": 6.5, "learning_rate": 1.0507721639656816e-06, "loss": 2.1204, "step": 34129 }, { "epoch": 6.5, "learning_rate": 1.0507149666348904e-06, "loss": 2.3508, "step": 34130 }, { "epoch": 6.5, "learning_rate": 1.0506577693040992e-06, "loss": 1.7839, "step": 34131 }, { "epoch": 6.5, "learning_rate": 1.050600571973308e-06, "loss": 1.4962, "step": 34132 }, { "epoch": 6.5, "learning_rate": 1.0505433746425168e-06, "loss": 1.1737, "step": 34133 }, { "epoch": 6.5, "learning_rate": 1.0504861773117254e-06, "loss": 2.1255, "step": 34134 }, { "epoch": 6.5, "learning_rate": 1.0504289799809342e-06, "loss": 1.6989, "step": 34135 }, { "epoch": 6.5, "learning_rate": 1.050371782650143e-06, "loss": 1.2845, "step": 34136 }, { "epoch": 6.5, "learning_rate": 1.050314585319352e-06, "loss": 1.2518, "step": 34137 }, { "epoch": 6.5, "learning_rate": 1.0502573879885605e-06, "loss": 1.3395, "step": 34138 }, { "epoch": 6.5, "learning_rate": 1.0502001906577693e-06, "loss": 1.8391, "step": 34139 }, { "epoch": 6.5, "learning_rate": 1.0501429933269781e-06, "loss": 2.4519, "step": 34140 }, { "epoch": 6.5, "learning_rate": 1.050085795996187e-06, "loss": 2.3325, "step": 34141 }, { "epoch": 6.5, "learning_rate": 1.0500285986653957e-06, "loss": 2.2493, "step": 34142 }, { "epoch": 6.5, "learning_rate": 1.0499714013346045e-06, "loss": 2.2549, "step": 34143 }, { "epoch": 6.5, "learning_rate": 1.049914204003813e-06, "loss": 1.6704, "step": 34144 }, { "epoch": 6.5, "learning_rate": 1.0498570066730219e-06, "loss": 2.1265, "step": 34145 }, { "epoch": 6.5, "learning_rate": 1.0497998093422307e-06, "loss": 1.2363, "step": 34146 }, { "epoch": 6.5, "learning_rate": 1.0497426120114395e-06, "loss": 1.198, "step": 34147 }, { "epoch": 6.5, "learning_rate": 1.0496854146806483e-06, "loss": 2.5555, "step": 34148 }, { "epoch": 6.5, "learning_rate": 1.049628217349857e-06, "loss": 1.6981, "step": 34149 }, { "epoch": 6.5, "learning_rate": 1.0495710200190659e-06, "loss": 0.8698, "step": 34150 }, { "epoch": 6.5, "learning_rate": 1.0495138226882744e-06, "loss": 0.916, "step": 34151 }, { "epoch": 6.51, "learning_rate": 1.0494566253574832e-06, "loss": 1.8241, "step": 34152 }, { "epoch": 6.51, "learning_rate": 1.0493994280266922e-06, "loss": 1.1583, "step": 34153 }, { "epoch": 6.51, "learning_rate": 1.049342230695901e-06, "loss": 1.5917, "step": 34154 }, { "epoch": 6.51, "learning_rate": 1.0492850333651096e-06, "loss": 1.4904, "step": 34155 }, { "epoch": 6.51, "learning_rate": 1.0492278360343184e-06, "loss": 1.7774, "step": 34156 }, { "epoch": 6.51, "learning_rate": 1.0491706387035272e-06, "loss": 1.8069, "step": 34157 }, { "epoch": 6.51, "learning_rate": 1.049113441372736e-06, "loss": 1.6978, "step": 34158 }, { "epoch": 6.51, "learning_rate": 1.0490562440419448e-06, "loss": 2.0561, "step": 34159 }, { "epoch": 6.51, "learning_rate": 1.0489990467111536e-06, "loss": 1.497, "step": 34160 }, { "epoch": 6.51, "learning_rate": 1.0489418493803622e-06, "loss": 1.6294, "step": 34161 }, { "epoch": 6.51, "learning_rate": 1.048884652049571e-06, "loss": 1.0665, "step": 34162 }, { "epoch": 6.51, "learning_rate": 1.0488274547187797e-06, "loss": 1.6308, "step": 34163 }, { "epoch": 6.51, "learning_rate": 1.0487702573879888e-06, "loss": 2.1391, "step": 34164 }, { "epoch": 6.51, "learning_rate": 1.0487130600571973e-06, "loss": 1.6291, "step": 34165 }, { "epoch": 6.51, "learning_rate": 1.0486558627264061e-06, "loss": 3.6321, "step": 34166 }, { "epoch": 6.51, "learning_rate": 1.048598665395615e-06, "loss": 0.3957, "step": 34167 }, { "epoch": 6.51, "learning_rate": 1.0485414680648235e-06, "loss": 1.4689, "step": 34168 }, { "epoch": 6.51, "learning_rate": 1.0484842707340325e-06, "loss": 1.5771, "step": 34169 }, { "epoch": 6.51, "learning_rate": 1.0484270734032413e-06, "loss": 2.2985, "step": 34170 }, { "epoch": 6.51, "learning_rate": 1.04836987607245e-06, "loss": 0.9389, "step": 34171 }, { "epoch": 6.51, "learning_rate": 1.0483126787416587e-06, "loss": 2.0371, "step": 34172 }, { "epoch": 6.51, "learning_rate": 1.0482554814108675e-06, "loss": 2.5052, "step": 34173 }, { "epoch": 6.51, "learning_rate": 1.0481982840800763e-06, "loss": 1.0846, "step": 34174 }, { "epoch": 6.51, "learning_rate": 1.048141086749285e-06, "loss": 1.6038, "step": 34175 }, { "epoch": 6.51, "learning_rate": 1.0480838894184939e-06, "loss": 1.6937, "step": 34176 }, { "epoch": 6.51, "learning_rate": 1.0480266920877027e-06, "loss": 2.5095, "step": 34177 }, { "epoch": 6.51, "learning_rate": 1.0479694947569114e-06, "loss": 1.6344, "step": 34178 }, { "epoch": 6.51, "learning_rate": 1.04791229742612e-06, "loss": 1.155, "step": 34179 }, { "epoch": 6.51, "learning_rate": 1.0478551000953288e-06, "loss": 2.5824, "step": 34180 }, { "epoch": 6.51, "learning_rate": 1.0477979027645378e-06, "loss": 1.0895, "step": 34181 }, { "epoch": 6.51, "learning_rate": 1.0477407054337464e-06, "loss": 1.6234, "step": 34182 }, { "epoch": 6.51, "learning_rate": 1.0476835081029552e-06, "loss": 1.4204, "step": 34183 }, { "epoch": 6.51, "learning_rate": 1.047626310772164e-06, "loss": 1.524, "step": 34184 }, { "epoch": 6.51, "learning_rate": 1.0475691134413726e-06, "loss": 0.9315, "step": 34185 }, { "epoch": 6.51, "learning_rate": 1.0475119161105816e-06, "loss": 1.442, "step": 34186 }, { "epoch": 6.51, "learning_rate": 1.0474547187797904e-06, "loss": 1.8747, "step": 34187 }, { "epoch": 6.51, "learning_rate": 1.0473975214489992e-06, "loss": 2.0692, "step": 34188 }, { "epoch": 6.51, "learning_rate": 1.0473403241182078e-06, "loss": 1.9772, "step": 34189 }, { "epoch": 6.51, "learning_rate": 1.0472831267874165e-06, "loss": 4.0765, "step": 34190 }, { "epoch": 6.51, "learning_rate": 1.0472259294566253e-06, "loss": 1.3383, "step": 34191 }, { "epoch": 6.51, "learning_rate": 1.0471687321258341e-06, "loss": 0.8885, "step": 34192 }, { "epoch": 6.51, "learning_rate": 1.047111534795043e-06, "loss": 1.8814, "step": 34193 }, { "epoch": 6.51, "learning_rate": 1.0470543374642517e-06, "loss": 1.2476, "step": 34194 }, { "epoch": 6.51, "learning_rate": 1.0469971401334605e-06, "loss": 1.5089, "step": 34195 }, { "epoch": 6.51, "learning_rate": 1.046939942802669e-06, "loss": 0.7126, "step": 34196 }, { "epoch": 6.51, "learning_rate": 1.046882745471878e-06, "loss": 2.4375, "step": 34197 }, { "epoch": 6.51, "learning_rate": 1.046825548141087e-06, "loss": 2.2993, "step": 34198 }, { "epoch": 6.51, "learning_rate": 1.0467683508102955e-06, "loss": 1.0931, "step": 34199 }, { "epoch": 6.51, "learning_rate": 1.0467111534795043e-06, "loss": 2.1307, "step": 34200 }, { "epoch": 6.51, "learning_rate": 1.046653956148713e-06, "loss": 1.2875, "step": 34201 }, { "epoch": 6.51, "learning_rate": 1.0465967588179219e-06, "loss": 1.1248, "step": 34202 }, { "epoch": 6.51, "learning_rate": 1.0465395614871307e-06, "loss": 1.181, "step": 34203 }, { "epoch": 6.52, "learning_rate": 1.0464823641563394e-06, "loss": 2.5847, "step": 34204 }, { "epoch": 6.52, "learning_rate": 1.0464251668255482e-06, "loss": 1.4113, "step": 34205 }, { "epoch": 6.52, "learning_rate": 1.0463679694947568e-06, "loss": 2.4715, "step": 34206 }, { "epoch": 6.52, "learning_rate": 1.0463107721639656e-06, "loss": 1.9003, "step": 34207 }, { "epoch": 6.52, "learning_rate": 1.0462535748331746e-06, "loss": 1.2074, "step": 34208 }, { "epoch": 6.52, "learning_rate": 1.0461963775023832e-06, "loss": 1.7051, "step": 34209 }, { "epoch": 6.52, "learning_rate": 1.046139180171592e-06, "loss": 2.5136, "step": 34210 }, { "epoch": 6.52, "learning_rate": 1.0460819828408008e-06, "loss": 0.8049, "step": 34211 }, { "epoch": 6.52, "learning_rate": 1.0460247855100096e-06, "loss": 1.5605, "step": 34212 }, { "epoch": 6.52, "learning_rate": 1.0459675881792184e-06, "loss": 0.6325, "step": 34213 }, { "epoch": 6.52, "learning_rate": 1.0459103908484272e-06, "loss": 2.2409, "step": 34214 }, { "epoch": 6.52, "learning_rate": 1.045853193517636e-06, "loss": 1.2311, "step": 34215 }, { "epoch": 6.52, "learning_rate": 1.0457959961868446e-06, "loss": 2.092, "step": 34216 }, { "epoch": 6.52, "learning_rate": 1.0457387988560533e-06, "loss": 1.5341, "step": 34217 }, { "epoch": 6.52, "learning_rate": 1.0456816015252621e-06, "loss": 2.5738, "step": 34218 }, { "epoch": 6.52, "learning_rate": 1.045624404194471e-06, "loss": 2.4125, "step": 34219 }, { "epoch": 6.52, "learning_rate": 1.0455672068636797e-06, "loss": 1.528, "step": 34220 }, { "epoch": 6.52, "learning_rate": 1.0455100095328885e-06, "loss": 2.2327, "step": 34221 }, { "epoch": 6.52, "learning_rate": 1.0454528122020973e-06, "loss": 2.2348, "step": 34222 }, { "epoch": 6.52, "learning_rate": 1.045395614871306e-06, "loss": 2.6778, "step": 34223 }, { "epoch": 6.52, "learning_rate": 1.0453384175405147e-06, "loss": 1.6299, "step": 34224 }, { "epoch": 6.52, "learning_rate": 1.0452812202097237e-06, "loss": 1.8462, "step": 34225 }, { "epoch": 6.52, "learning_rate": 1.0452240228789323e-06, "loss": 1.6402, "step": 34226 }, { "epoch": 6.52, "learning_rate": 1.045166825548141e-06, "loss": 3.0514, "step": 34227 }, { "epoch": 6.52, "learning_rate": 1.0451096282173499e-06, "loss": 1.685, "step": 34228 }, { "epoch": 6.52, "learning_rate": 1.0450524308865587e-06, "loss": 1.3727, "step": 34229 }, { "epoch": 6.52, "learning_rate": 1.0449952335557675e-06, "loss": 0.5714, "step": 34230 }, { "epoch": 6.52, "learning_rate": 1.0449380362249762e-06, "loss": 2.3987, "step": 34231 }, { "epoch": 6.52, "learning_rate": 1.044880838894185e-06, "loss": 1.3078, "step": 34232 }, { "epoch": 6.52, "learning_rate": 1.0448236415633936e-06, "loss": 3.18, "step": 34233 }, { "epoch": 6.52, "learning_rate": 1.0447664442326024e-06, "loss": 2.3781, "step": 34234 }, { "epoch": 6.52, "learning_rate": 1.0447092469018112e-06, "loss": 3.0991, "step": 34235 }, { "epoch": 6.52, "learning_rate": 1.0446520495710202e-06, "loss": 2.0315, "step": 34236 }, { "epoch": 6.52, "learning_rate": 1.0445948522402288e-06, "loss": 2.2264, "step": 34237 }, { "epoch": 6.52, "learning_rate": 1.0445376549094376e-06, "loss": 3.0156, "step": 34238 }, { "epoch": 6.52, "learning_rate": 1.0444804575786464e-06, "loss": 0.671, "step": 34239 }, { "epoch": 6.52, "learning_rate": 1.044423260247855e-06, "loss": 1.6756, "step": 34240 }, { "epoch": 6.52, "learning_rate": 1.044366062917064e-06, "loss": 1.402, "step": 34241 }, { "epoch": 6.52, "learning_rate": 1.0443088655862728e-06, "loss": 1.1583, "step": 34242 }, { "epoch": 6.52, "learning_rate": 1.0442516682554816e-06, "loss": 0.8193, "step": 34243 }, { "epoch": 6.52, "learning_rate": 1.0441944709246901e-06, "loss": 1.8645, "step": 34244 }, { "epoch": 6.52, "learning_rate": 1.044137273593899e-06, "loss": 2.0351, "step": 34245 }, { "epoch": 6.52, "learning_rate": 1.0440800762631077e-06, "loss": 1.4476, "step": 34246 }, { "epoch": 6.52, "learning_rate": 1.0440228789323165e-06, "loss": 1.7959, "step": 34247 }, { "epoch": 6.52, "learning_rate": 1.0439656816015253e-06, "loss": 1.8919, "step": 34248 }, { "epoch": 6.52, "learning_rate": 1.0439084842707341e-06, "loss": 2.0792, "step": 34249 }, { "epoch": 6.52, "learning_rate": 1.0438512869399427e-06, "loss": 0.7166, "step": 34250 }, { "epoch": 6.52, "learning_rate": 1.0437940896091515e-06, "loss": 1.1078, "step": 34251 }, { "epoch": 6.52, "learning_rate": 1.0437368922783605e-06, "loss": 1.3686, "step": 34252 }, { "epoch": 6.52, "learning_rate": 1.0436796949475693e-06, "loss": 2.0168, "step": 34253 }, { "epoch": 6.52, "learning_rate": 1.0436224976167779e-06, "loss": 1.4034, "step": 34254 }, { "epoch": 6.52, "learning_rate": 1.0435653002859867e-06, "loss": 1.402, "step": 34255 }, { "epoch": 6.52, "learning_rate": 1.0435081029551955e-06, "loss": 1.6609, "step": 34256 }, { "epoch": 6.53, "learning_rate": 1.0434509056244043e-06, "loss": 1.327, "step": 34257 }, { "epoch": 6.53, "learning_rate": 1.043393708293613e-06, "loss": 1.7811, "step": 34258 }, { "epoch": 6.53, "learning_rate": 1.0433365109628218e-06, "loss": 1.2198, "step": 34259 }, { "epoch": 6.53, "learning_rate": 1.0432793136320306e-06, "loss": 1.8013, "step": 34260 }, { "epoch": 6.53, "learning_rate": 1.0432221163012392e-06, "loss": 1.4618, "step": 34261 }, { "epoch": 6.53, "learning_rate": 1.043164918970448e-06, "loss": 1.5326, "step": 34262 }, { "epoch": 6.53, "learning_rate": 1.0431077216396568e-06, "loss": 1.0838, "step": 34263 }, { "epoch": 6.53, "learning_rate": 1.0430505243088656e-06, "loss": 2.2292, "step": 34264 }, { "epoch": 6.53, "learning_rate": 1.0429933269780744e-06, "loss": 1.3613, "step": 34265 }, { "epoch": 6.53, "learning_rate": 1.0429361296472832e-06, "loss": 0.5252, "step": 34266 }, { "epoch": 6.53, "learning_rate": 1.042878932316492e-06, "loss": 0.8914, "step": 34267 }, { "epoch": 6.53, "learning_rate": 1.0428217349857006e-06, "loss": 2.7675, "step": 34268 }, { "epoch": 6.53, "learning_rate": 1.0427645376549096e-06, "loss": 2.4269, "step": 34269 }, { "epoch": 6.53, "learning_rate": 1.0427073403241184e-06, "loss": 2.0972, "step": 34270 }, { "epoch": 6.53, "learning_rate": 1.042650142993327e-06, "loss": 0.6068, "step": 34271 }, { "epoch": 6.53, "learning_rate": 1.0425929456625357e-06, "loss": 1.5099, "step": 34272 }, { "epoch": 6.53, "learning_rate": 1.0425357483317445e-06, "loss": 1.4282, "step": 34273 }, { "epoch": 6.53, "learning_rate": 1.0424785510009533e-06, "loss": 1.6827, "step": 34274 }, { "epoch": 6.53, "learning_rate": 1.0424213536701621e-06, "loss": 2.5561, "step": 34275 }, { "epoch": 6.53, "learning_rate": 1.042364156339371e-06, "loss": 1.903, "step": 34276 }, { "epoch": 6.53, "learning_rate": 1.0423069590085797e-06, "loss": 2.633, "step": 34277 }, { "epoch": 6.53, "learning_rate": 1.0422497616777883e-06, "loss": 1.1422, "step": 34278 }, { "epoch": 6.53, "learning_rate": 1.042192564346997e-06, "loss": 1.4937, "step": 34279 }, { "epoch": 6.53, "learning_rate": 1.042135367016206e-06, "loss": 0.3865, "step": 34280 }, { "epoch": 6.53, "learning_rate": 1.0420781696854147e-06, "loss": 0.9494, "step": 34281 }, { "epoch": 6.53, "learning_rate": 1.0420209723546235e-06, "loss": 1.4344, "step": 34282 }, { "epoch": 6.53, "learning_rate": 1.0419637750238323e-06, "loss": 1.449, "step": 34283 }, { "epoch": 6.53, "learning_rate": 1.041906577693041e-06, "loss": 0.7289, "step": 34284 }, { "epoch": 6.53, "learning_rate": 1.0418493803622498e-06, "loss": 1.3147, "step": 34285 }, { "epoch": 6.53, "learning_rate": 1.0417921830314586e-06, "loss": 2.6396, "step": 34286 }, { "epoch": 6.53, "learning_rate": 1.0417349857006674e-06, "loss": 1.6717, "step": 34287 }, { "epoch": 6.53, "learning_rate": 1.041677788369876e-06, "loss": 2.553, "step": 34288 }, { "epoch": 6.53, "learning_rate": 1.0416205910390848e-06, "loss": 0.9633, "step": 34289 }, { "epoch": 6.53, "learning_rate": 1.0415633937082936e-06, "loss": 1.4807, "step": 34290 }, { "epoch": 6.53, "learning_rate": 1.0415061963775024e-06, "loss": 2.0042, "step": 34291 }, { "epoch": 6.53, "learning_rate": 1.0414489990467112e-06, "loss": 2.2958, "step": 34292 }, { "epoch": 6.53, "learning_rate": 1.04139180171592e-06, "loss": 2.7225, "step": 34293 }, { "epoch": 6.53, "learning_rate": 1.0413346043851288e-06, "loss": 1.1303, "step": 34294 }, { "epoch": 6.53, "learning_rate": 1.0412774070543374e-06, "loss": 2.5413, "step": 34295 }, { "epoch": 6.53, "learning_rate": 1.0412202097235464e-06, "loss": 1.4155, "step": 34296 }, { "epoch": 6.53, "learning_rate": 1.0411630123927552e-06, "loss": 2.4563, "step": 34297 }, { "epoch": 6.53, "learning_rate": 1.0411058150619637e-06, "loss": 2.1414, "step": 34298 }, { "epoch": 6.53, "learning_rate": 1.0410486177311725e-06, "loss": 2.7848, "step": 34299 }, { "epoch": 6.53, "learning_rate": 1.0409914204003813e-06, "loss": 1.5053, "step": 34300 }, { "epoch": 6.53, "learning_rate": 1.0409342230695901e-06, "loss": 2.2119, "step": 34301 }, { "epoch": 6.53, "learning_rate": 1.040877025738799e-06, "loss": 1.7304, "step": 34302 }, { "epoch": 6.53, "learning_rate": 1.0408198284080077e-06, "loss": 1.0288, "step": 34303 }, { "epoch": 6.53, "learning_rate": 1.0407626310772165e-06, "loss": 1.6564, "step": 34304 }, { "epoch": 6.53, "learning_rate": 1.040705433746425e-06, "loss": 1.3875, "step": 34305 }, { "epoch": 6.53, "learning_rate": 1.0406482364156339e-06, "loss": 0.6401, "step": 34306 }, { "epoch": 6.53, "learning_rate": 1.0405910390848427e-06, "loss": 0.9512, "step": 34307 }, { "epoch": 6.53, "learning_rate": 1.0405338417540517e-06, "loss": 1.3292, "step": 34308 }, { "epoch": 6.54, "learning_rate": 1.0404766444232603e-06, "loss": 2.0961, "step": 34309 }, { "epoch": 6.54, "learning_rate": 1.040419447092469e-06, "loss": 2.2491, "step": 34310 }, { "epoch": 6.54, "learning_rate": 1.0403622497616778e-06, "loss": 2.3378, "step": 34311 }, { "epoch": 6.54, "learning_rate": 1.0403050524308864e-06, "loss": 1.6616, "step": 34312 }, { "epoch": 6.54, "learning_rate": 1.0402478551000954e-06, "loss": 1.5758, "step": 34313 }, { "epoch": 6.54, "learning_rate": 1.0401906577693042e-06, "loss": 0.8441, "step": 34314 }, { "epoch": 6.54, "learning_rate": 1.0401334604385128e-06, "loss": 2.2793, "step": 34315 }, { "epoch": 6.54, "learning_rate": 1.0400762631077216e-06, "loss": 2.1264, "step": 34316 }, { "epoch": 6.54, "learning_rate": 1.0400190657769304e-06, "loss": 1.3618, "step": 34317 }, { "epoch": 6.54, "learning_rate": 1.0399618684461392e-06, "loss": 1.9937, "step": 34318 }, { "epoch": 6.54, "learning_rate": 1.039904671115348e-06, "loss": 4.0944, "step": 34319 }, { "epoch": 6.54, "learning_rate": 1.0398474737845568e-06, "loss": 1.9966, "step": 34320 }, { "epoch": 6.54, "learning_rate": 1.0397902764537656e-06, "loss": 1.4007, "step": 34321 }, { "epoch": 6.54, "learning_rate": 1.0397330791229742e-06, "loss": 1.8714, "step": 34322 }, { "epoch": 6.54, "learning_rate": 1.039675881792183e-06, "loss": 1.3596, "step": 34323 }, { "epoch": 6.54, "learning_rate": 1.039618684461392e-06, "loss": 1.2183, "step": 34324 }, { "epoch": 6.54, "learning_rate": 1.0395614871306007e-06, "loss": 1.1458, "step": 34325 }, { "epoch": 6.54, "learning_rate": 1.0395042897998093e-06, "loss": 1.9639, "step": 34326 }, { "epoch": 6.54, "learning_rate": 1.0394470924690181e-06, "loss": 1.5821, "step": 34327 }, { "epoch": 6.54, "learning_rate": 1.039389895138227e-06, "loss": 1.9137, "step": 34328 }, { "epoch": 6.54, "learning_rate": 1.0393326978074357e-06, "loss": 1.7587, "step": 34329 }, { "epoch": 6.54, "learning_rate": 1.0392755004766445e-06, "loss": 2.3233, "step": 34330 }, { "epoch": 6.54, "learning_rate": 1.0392183031458533e-06, "loss": 1.6474, "step": 34331 }, { "epoch": 6.54, "learning_rate": 1.0391611058150619e-06, "loss": 1.0299, "step": 34332 }, { "epoch": 6.54, "learning_rate": 1.0391039084842707e-06, "loss": 1.5948, "step": 34333 }, { "epoch": 6.54, "learning_rate": 1.0390467111534795e-06, "loss": 1.9642, "step": 34334 }, { "epoch": 6.54, "learning_rate": 1.0389895138226883e-06, "loss": 0.6888, "step": 34335 }, { "epoch": 6.54, "learning_rate": 1.038932316491897e-06, "loss": 1.6352, "step": 34336 }, { "epoch": 6.54, "learning_rate": 1.0388751191611059e-06, "loss": 2.5503, "step": 34337 }, { "epoch": 6.54, "learning_rate": 1.0388179218303146e-06, "loss": 1.235, "step": 34338 }, { "epoch": 6.54, "learning_rate": 1.0387607244995232e-06, "loss": 2.2086, "step": 34339 }, { "epoch": 6.54, "learning_rate": 1.038703527168732e-06, "loss": 1.5691, "step": 34340 }, { "epoch": 6.54, "learning_rate": 1.038646329837941e-06, "loss": 2.5554, "step": 34341 }, { "epoch": 6.54, "learning_rate": 1.0385891325071498e-06, "loss": 1.2357, "step": 34342 }, { "epoch": 6.54, "learning_rate": 1.0385319351763584e-06, "loss": 1.7666, "step": 34343 }, { "epoch": 6.54, "learning_rate": 1.0384747378455672e-06, "loss": 3.0273, "step": 34344 }, { "epoch": 6.54, "learning_rate": 1.038417540514776e-06, "loss": 2.1984, "step": 34345 }, { "epoch": 6.54, "learning_rate": 1.0383603431839848e-06, "loss": 1.6263, "step": 34346 }, { "epoch": 6.54, "learning_rate": 1.0383031458531936e-06, "loss": 2.0966, "step": 34347 }, { "epoch": 6.54, "learning_rate": 1.0382459485224024e-06, "loss": 2.405, "step": 34348 }, { "epoch": 6.54, "learning_rate": 1.0381887511916112e-06, "loss": 2.4423, "step": 34349 }, { "epoch": 6.54, "learning_rate": 1.0381315538608197e-06, "loss": 2.316, "step": 34350 }, { "epoch": 6.54, "learning_rate": 1.0380743565300285e-06, "loss": 3.6907, "step": 34351 }, { "epoch": 6.54, "learning_rate": 1.0380171591992375e-06, "loss": 2.5809, "step": 34352 }, { "epoch": 6.54, "learning_rate": 1.0379599618684461e-06, "loss": 1.0892, "step": 34353 }, { "epoch": 6.54, "learning_rate": 1.037902764537655e-06, "loss": 1.3188, "step": 34354 }, { "epoch": 6.54, "learning_rate": 1.0378455672068637e-06, "loss": 1.6408, "step": 34355 }, { "epoch": 6.54, "learning_rate": 1.0377883698760723e-06, "loss": 1.8209, "step": 34356 }, { "epoch": 6.54, "learning_rate": 1.0377311725452813e-06, "loss": 0.5994, "step": 34357 }, { "epoch": 6.54, "learning_rate": 1.03767397521449e-06, "loss": 1.4924, "step": 34358 }, { "epoch": 6.54, "learning_rate": 1.0376167778836989e-06, "loss": 1.3411, "step": 34359 }, { "epoch": 6.54, "learning_rate": 1.0375595805529075e-06, "loss": 2.6997, "step": 34360 }, { "epoch": 6.54, "learning_rate": 1.0375023832221163e-06, "loss": 1.0725, "step": 34361 }, { "epoch": 6.55, "learning_rate": 1.037445185891325e-06, "loss": 2.4974, "step": 34362 }, { "epoch": 6.55, "learning_rate": 1.0373879885605339e-06, "loss": 2.1104, "step": 34363 }, { "epoch": 6.55, "learning_rate": 1.0373307912297426e-06, "loss": 2.3964, "step": 34364 }, { "epoch": 6.55, "learning_rate": 1.0372735938989514e-06, "loss": 2.0328, "step": 34365 }, { "epoch": 6.55, "learning_rate": 1.0372163965681602e-06, "loss": 3.2178, "step": 34366 }, { "epoch": 6.55, "learning_rate": 1.0371591992373688e-06, "loss": 1.0204, "step": 34367 }, { "epoch": 6.55, "learning_rate": 1.0371020019065778e-06, "loss": 3.0906, "step": 34368 }, { "epoch": 6.55, "learning_rate": 1.0370448045757866e-06, "loss": 1.2172, "step": 34369 }, { "epoch": 6.55, "learning_rate": 1.0369876072449952e-06, "loss": 1.1002, "step": 34370 }, { "epoch": 6.55, "learning_rate": 1.036930409914204e-06, "loss": 2.2905, "step": 34371 }, { "epoch": 6.55, "learning_rate": 1.0368732125834128e-06, "loss": 1.7976, "step": 34372 }, { "epoch": 6.55, "learning_rate": 1.0368160152526216e-06, "loss": 2.2574, "step": 34373 }, { "epoch": 6.55, "learning_rate": 1.0367588179218304e-06, "loss": 2.1609, "step": 34374 }, { "epoch": 6.55, "learning_rate": 1.0367016205910392e-06, "loss": 2.1921, "step": 34375 }, { "epoch": 6.55, "learning_rate": 1.036644423260248e-06, "loss": 1.4183, "step": 34376 }, { "epoch": 6.55, "learning_rate": 1.0365872259294565e-06, "loss": 2.2893, "step": 34377 }, { "epoch": 6.55, "learning_rate": 1.0365300285986653e-06, "loss": 1.9358, "step": 34378 }, { "epoch": 6.55, "learning_rate": 1.0364728312678741e-06, "loss": 2.1639, "step": 34379 }, { "epoch": 6.55, "learning_rate": 1.036415633937083e-06, "loss": 2.9965, "step": 34380 }, { "epoch": 6.55, "learning_rate": 1.0363584366062917e-06, "loss": 1.3395, "step": 34381 }, { "epoch": 6.55, "learning_rate": 1.0363012392755005e-06, "loss": 2.2743, "step": 34382 }, { "epoch": 6.55, "learning_rate": 1.0362440419447093e-06, "loss": 0.3612, "step": 34383 }, { "epoch": 6.55, "learning_rate": 1.0361868446139179e-06, "loss": 1.5102, "step": 34384 }, { "epoch": 6.55, "learning_rate": 1.036129647283127e-06, "loss": 1.5258, "step": 34385 }, { "epoch": 6.55, "learning_rate": 1.0360724499523357e-06, "loss": 3.3172, "step": 34386 }, { "epoch": 6.55, "learning_rate": 1.0360152526215443e-06, "loss": 2.035, "step": 34387 }, { "epoch": 6.55, "learning_rate": 1.035958055290753e-06, "loss": 0.9724, "step": 34388 }, { "epoch": 6.55, "learning_rate": 1.0359008579599619e-06, "loss": 2.6878, "step": 34389 }, { "epoch": 6.55, "learning_rate": 1.0358436606291707e-06, "loss": 1.2132, "step": 34390 }, { "epoch": 6.55, "learning_rate": 1.0357864632983794e-06, "loss": 2.336, "step": 34391 }, { "epoch": 6.55, "learning_rate": 1.0357292659675882e-06, "loss": 2.1561, "step": 34392 }, { "epoch": 6.55, "learning_rate": 1.035672068636797e-06, "loss": 1.6345, "step": 34393 }, { "epoch": 6.55, "learning_rate": 1.0356148713060056e-06, "loss": 2.1516, "step": 34394 }, { "epoch": 6.55, "learning_rate": 1.0355576739752144e-06, "loss": 1.4082, "step": 34395 }, { "epoch": 6.55, "learning_rate": 1.0355004766444234e-06, "loss": 1.8475, "step": 34396 }, { "epoch": 6.55, "learning_rate": 1.035443279313632e-06, "loss": 0.6976, "step": 34397 }, { "epoch": 6.55, "learning_rate": 1.0353860819828408e-06, "loss": 1.9793, "step": 34398 }, { "epoch": 6.55, "learning_rate": 1.0353288846520496e-06, "loss": 2.8857, "step": 34399 }, { "epoch": 6.55, "learning_rate": 1.0352716873212584e-06, "loss": 1.9393, "step": 34400 }, { "epoch": 6.55, "learning_rate": 1.0352144899904672e-06, "loss": 1.8732, "step": 34401 }, { "epoch": 6.55, "learning_rate": 1.035157292659676e-06, "loss": 1.8018, "step": 34402 }, { "epoch": 6.55, "learning_rate": 1.0351000953288848e-06, "loss": 2.0483, "step": 34403 }, { "epoch": 6.55, "learning_rate": 1.0350428979980933e-06, "loss": 2.9816, "step": 34404 }, { "epoch": 6.55, "learning_rate": 1.0349857006673021e-06, "loss": 2.6803, "step": 34405 }, { "epoch": 6.55, "learning_rate": 1.034928503336511e-06, "loss": 1.2422, "step": 34406 }, { "epoch": 6.55, "learning_rate": 1.03487130600572e-06, "loss": 1.1586, "step": 34407 }, { "epoch": 6.55, "learning_rate": 1.0348141086749285e-06, "loss": 1.6206, "step": 34408 }, { "epoch": 6.55, "learning_rate": 1.0347569113441373e-06, "loss": 2.0126, "step": 34409 }, { "epoch": 6.55, "learning_rate": 1.034699714013346e-06, "loss": 1.462, "step": 34410 }, { "epoch": 6.55, "learning_rate": 1.0346425166825547e-06, "loss": 0.569, "step": 34411 }, { "epoch": 6.55, "learning_rate": 1.0345853193517637e-06, "loss": 1.4773, "step": 34412 }, { "epoch": 6.55, "learning_rate": 1.0345281220209725e-06, "loss": 2.3155, "step": 34413 }, { "epoch": 6.56, "learning_rate": 1.0344709246901813e-06, "loss": 1.86, "step": 34414 }, { "epoch": 6.56, "learning_rate": 1.0344137273593899e-06, "loss": 2.6562, "step": 34415 }, { "epoch": 6.56, "learning_rate": 1.0343565300285987e-06, "loss": 1.3466, "step": 34416 }, { "epoch": 6.56, "learning_rate": 1.0342993326978074e-06, "loss": 2.2743, "step": 34417 }, { "epoch": 6.56, "learning_rate": 1.0342421353670162e-06, "loss": 2.1587, "step": 34418 }, { "epoch": 6.56, "learning_rate": 1.034184938036225e-06, "loss": 0.9447, "step": 34419 }, { "epoch": 6.56, "learning_rate": 1.0341277407054338e-06, "loss": 2.2954, "step": 34420 }, { "epoch": 6.56, "learning_rate": 1.0340705433746424e-06, "loss": 1.709, "step": 34421 }, { "epoch": 6.56, "learning_rate": 1.0340133460438512e-06, "loss": 2.029, "step": 34422 }, { "epoch": 6.56, "learning_rate": 1.03395614871306e-06, "loss": 1.5777, "step": 34423 }, { "epoch": 6.56, "learning_rate": 1.033898951382269e-06, "loss": 1.8048, "step": 34424 }, { "epoch": 6.56, "learning_rate": 1.0338417540514776e-06, "loss": 0.8825, "step": 34425 }, { "epoch": 6.56, "learning_rate": 1.0337845567206864e-06, "loss": 1.2855, "step": 34426 }, { "epoch": 6.56, "learning_rate": 1.0337273593898952e-06, "loss": 1.6202, "step": 34427 }, { "epoch": 6.56, "learning_rate": 1.0336701620591038e-06, "loss": 1.8803, "step": 34428 }, { "epoch": 6.56, "learning_rate": 1.0336129647283128e-06, "loss": 0.6978, "step": 34429 }, { "epoch": 6.56, "learning_rate": 1.0335557673975216e-06, "loss": 0.7476, "step": 34430 }, { "epoch": 6.56, "learning_rate": 1.0334985700667304e-06, "loss": 2.2455, "step": 34431 }, { "epoch": 6.56, "learning_rate": 1.033441372735939e-06, "loss": 1.7325, "step": 34432 }, { "epoch": 6.56, "learning_rate": 1.0333841754051477e-06, "loss": 0.5621, "step": 34433 }, { "epoch": 6.56, "learning_rate": 1.0333269780743565e-06, "loss": 2.157, "step": 34434 }, { "epoch": 6.56, "learning_rate": 1.0332697807435653e-06, "loss": 1.2006, "step": 34435 }, { "epoch": 6.56, "learning_rate": 1.0332125834127741e-06, "loss": 2.2096, "step": 34436 }, { "epoch": 6.56, "learning_rate": 1.033155386081983e-06, "loss": 3.1025, "step": 34437 }, { "epoch": 6.56, "learning_rate": 1.0330981887511917e-06, "loss": 1.4757, "step": 34438 }, { "epoch": 6.56, "learning_rate": 1.0330409914204003e-06, "loss": 0.912, "step": 34439 }, { "epoch": 6.56, "learning_rate": 1.0329837940896093e-06, "loss": 3.2639, "step": 34440 }, { "epoch": 6.56, "learning_rate": 1.032926596758818e-06, "loss": 1.725, "step": 34441 }, { "epoch": 6.56, "learning_rate": 1.0328693994280267e-06, "loss": 2.3253, "step": 34442 }, { "epoch": 6.56, "learning_rate": 1.0328122020972355e-06, "loss": 1.8375, "step": 34443 }, { "epoch": 6.56, "learning_rate": 1.0327550047664442e-06, "loss": 1.6763, "step": 34444 }, { "epoch": 6.56, "learning_rate": 1.032697807435653e-06, "loss": 0.9607, "step": 34445 }, { "epoch": 6.56, "learning_rate": 1.0326406101048618e-06, "loss": 2.7986, "step": 34446 }, { "epoch": 6.56, "learning_rate": 1.0325834127740706e-06, "loss": 1.6576, "step": 34447 }, { "epoch": 6.56, "learning_rate": 1.0325262154432794e-06, "loss": 2.158, "step": 34448 }, { "epoch": 6.56, "learning_rate": 1.032469018112488e-06, "loss": 1.7724, "step": 34449 }, { "epoch": 6.56, "learning_rate": 1.0324118207816968e-06, "loss": 2.6607, "step": 34450 }, { "epoch": 6.56, "learning_rate": 1.0323546234509058e-06, "loss": 1.482, "step": 34451 }, { "epoch": 6.56, "learning_rate": 1.0322974261201144e-06, "loss": 2.3126, "step": 34452 }, { "epoch": 6.56, "learning_rate": 1.0322402287893232e-06, "loss": 1.4087, "step": 34453 }, { "epoch": 6.56, "learning_rate": 1.032183031458532e-06, "loss": 3.4593, "step": 34454 }, { "epoch": 6.56, "learning_rate": 1.0321258341277408e-06, "loss": 1.363, "step": 34455 }, { "epoch": 6.56, "learning_rate": 1.0320686367969496e-06, "loss": 1.6942, "step": 34456 }, { "epoch": 6.56, "learning_rate": 1.0320114394661584e-06, "loss": 2.305, "step": 34457 }, { "epoch": 6.56, "learning_rate": 1.0319542421353671e-06, "loss": 1.9907, "step": 34458 }, { "epoch": 6.56, "learning_rate": 1.0318970448045757e-06, "loss": 2.6651, "step": 34459 }, { "epoch": 6.56, "learning_rate": 1.0318398474737845e-06, "loss": 2.2857, "step": 34460 }, { "epoch": 6.56, "learning_rate": 1.0317826501429933e-06, "loss": 1.1295, "step": 34461 }, { "epoch": 6.56, "learning_rate": 1.0317254528122021e-06, "loss": 2.6756, "step": 34462 }, { "epoch": 6.56, "learning_rate": 1.031668255481411e-06, "loss": 0.8123, "step": 34463 }, { "epoch": 6.56, "learning_rate": 1.0316110581506197e-06, "loss": 1.5179, "step": 34464 }, { "epoch": 6.56, "learning_rate": 1.0315538608198285e-06, "loss": 1.9147, "step": 34465 }, { "epoch": 6.56, "learning_rate": 1.031496663489037e-06, "loss": 1.0635, "step": 34466 }, { "epoch": 6.57, "learning_rate": 1.0314394661582459e-06, "loss": 0.8869, "step": 34467 }, { "epoch": 6.57, "learning_rate": 1.0313822688274549e-06, "loss": 1.2385, "step": 34468 }, { "epoch": 6.57, "learning_rate": 1.0313250714966635e-06, "loss": 1.1081, "step": 34469 }, { "epoch": 6.57, "learning_rate": 1.0312678741658723e-06, "loss": 2.507, "step": 34470 }, { "epoch": 6.57, "learning_rate": 1.031210676835081e-06, "loss": 1.7864, "step": 34471 }, { "epoch": 6.57, "learning_rate": 1.0311534795042898e-06, "loss": 1.7058, "step": 34472 }, { "epoch": 6.57, "learning_rate": 1.0310962821734986e-06, "loss": 2.4817, "step": 34473 }, { "epoch": 6.57, "learning_rate": 1.0310390848427074e-06, "loss": 1.4456, "step": 34474 }, { "epoch": 6.57, "learning_rate": 1.0309818875119162e-06, "loss": 1.7205, "step": 34475 }, { "epoch": 6.57, "learning_rate": 1.0309246901811248e-06, "loss": 1.0045, "step": 34476 }, { "epoch": 6.57, "learning_rate": 1.0308674928503336e-06, "loss": 1.4822, "step": 34477 }, { "epoch": 6.57, "learning_rate": 1.0308102955195424e-06, "loss": 2.5936, "step": 34478 }, { "epoch": 6.57, "learning_rate": 1.0307530981887514e-06, "loss": 1.9837, "step": 34479 }, { "epoch": 6.57, "learning_rate": 1.03069590085796e-06, "loss": 2.5467, "step": 34480 }, { "epoch": 6.57, "learning_rate": 1.0306387035271688e-06, "loss": 1.0522, "step": 34481 }, { "epoch": 6.57, "learning_rate": 1.0305815061963776e-06, "loss": 0.7963, "step": 34482 }, { "epoch": 6.57, "learning_rate": 1.0305243088655861e-06, "loss": 1.5341, "step": 34483 }, { "epoch": 6.57, "learning_rate": 1.0304671115347952e-06, "loss": 1.6962, "step": 34484 }, { "epoch": 6.57, "learning_rate": 1.030409914204004e-06, "loss": 0.6983, "step": 34485 }, { "epoch": 6.57, "learning_rate": 1.0303527168732125e-06, "loss": 2.1061, "step": 34486 }, { "epoch": 6.57, "learning_rate": 1.0302955195424213e-06, "loss": 2.3238, "step": 34487 }, { "epoch": 6.57, "learning_rate": 1.0302383222116301e-06, "loss": 0.4011, "step": 34488 }, { "epoch": 6.57, "learning_rate": 1.030181124880839e-06, "loss": 1.1064, "step": 34489 }, { "epoch": 6.57, "learning_rate": 1.0301239275500477e-06, "loss": 2.3062, "step": 34490 }, { "epoch": 6.57, "learning_rate": 1.0300667302192565e-06, "loss": 2.0998, "step": 34491 }, { "epoch": 6.57, "learning_rate": 1.0300095328884653e-06, "loss": 2.0458, "step": 34492 }, { "epoch": 6.57, "learning_rate": 1.0299523355576739e-06, "loss": 2.2161, "step": 34493 }, { "epoch": 6.57, "learning_rate": 1.0298951382268827e-06, "loss": 0.9827, "step": 34494 }, { "epoch": 6.57, "learning_rate": 1.0298379408960915e-06, "loss": 2.939, "step": 34495 }, { "epoch": 6.57, "learning_rate": 1.0297807435653005e-06, "loss": 2.2573, "step": 34496 }, { "epoch": 6.57, "learning_rate": 1.029723546234509e-06, "loss": 2.1888, "step": 34497 }, { "epoch": 6.57, "learning_rate": 1.0296663489037178e-06, "loss": 1.2624, "step": 34498 }, { "epoch": 6.57, "learning_rate": 1.0296091515729266e-06, "loss": 1.2891, "step": 34499 }, { "epoch": 6.57, "learning_rate": 1.0295519542421354e-06, "loss": 1.4196, "step": 34500 }, { "epoch": 6.57, "learning_rate": 1.0294947569113442e-06, "loss": 1.3124, "step": 34501 }, { "epoch": 6.57, "learning_rate": 1.029437559580553e-06, "loss": 1.9995, "step": 34502 }, { "epoch": 6.57, "learning_rate": 1.0293803622497618e-06, "loss": 3.1755, "step": 34503 }, { "epoch": 6.57, "learning_rate": 1.0293231649189704e-06, "loss": 1.5439, "step": 34504 }, { "epoch": 6.57, "learning_rate": 1.0292659675881792e-06, "loss": 2.7205, "step": 34505 }, { "epoch": 6.57, "learning_rate": 1.029208770257388e-06, "loss": 1.0059, "step": 34506 }, { "epoch": 6.57, "learning_rate": 1.0291515729265968e-06, "loss": 1.5848, "step": 34507 }, { "epoch": 6.57, "learning_rate": 1.0290943755958056e-06, "loss": 1.4507, "step": 34508 }, { "epoch": 6.57, "learning_rate": 1.0290371782650144e-06, "loss": 0.3892, "step": 34509 }, { "epoch": 6.57, "learning_rate": 1.028979980934223e-06, "loss": 2.0334, "step": 34510 }, { "epoch": 6.57, "learning_rate": 1.0289227836034317e-06, "loss": 1.5544, "step": 34511 }, { "epoch": 6.57, "learning_rate": 1.0288655862726407e-06, "loss": 0.7607, "step": 34512 }, { "epoch": 6.57, "learning_rate": 1.0288083889418495e-06, "loss": 2.9672, "step": 34513 }, { "epoch": 6.57, "learning_rate": 1.0287511916110581e-06, "loss": 2.11, "step": 34514 }, { "epoch": 6.57, "learning_rate": 1.028693994280267e-06, "loss": 2.3502, "step": 34515 }, { "epoch": 6.57, "learning_rate": 1.0286367969494757e-06, "loss": 1.5031, "step": 34516 }, { "epoch": 6.57, "learning_rate": 1.0285795996186845e-06, "loss": 0.8002, "step": 34517 }, { "epoch": 6.57, "learning_rate": 1.0285224022878933e-06, "loss": 0.73, "step": 34518 }, { "epoch": 6.58, "learning_rate": 1.028465204957102e-06, "loss": 2.2699, "step": 34519 }, { "epoch": 6.58, "learning_rate": 1.0284080076263109e-06, "loss": 0.5055, "step": 34520 }, { "epoch": 6.58, "learning_rate": 1.0283508102955195e-06, "loss": 1.6317, "step": 34521 }, { "epoch": 6.58, "learning_rate": 1.0282936129647283e-06, "loss": 1.0306, "step": 34522 }, { "epoch": 6.58, "learning_rate": 1.0282364156339373e-06, "loss": 0.6264, "step": 34523 }, { "epoch": 6.58, "learning_rate": 1.0281792183031458e-06, "loss": 1.223, "step": 34524 }, { "epoch": 6.58, "learning_rate": 1.0281220209723546e-06, "loss": 1.8664, "step": 34525 }, { "epoch": 6.58, "learning_rate": 1.0280648236415634e-06, "loss": 0.8524, "step": 34526 }, { "epoch": 6.58, "learning_rate": 1.028007626310772e-06, "loss": 1.7604, "step": 34527 }, { "epoch": 6.58, "learning_rate": 1.027950428979981e-06, "loss": 0.7951, "step": 34528 }, { "epoch": 6.58, "learning_rate": 1.0278932316491898e-06, "loss": 2.2039, "step": 34529 }, { "epoch": 6.58, "learning_rate": 1.0278360343183986e-06, "loss": 3.5886, "step": 34530 }, { "epoch": 6.58, "learning_rate": 1.0277788369876072e-06, "loss": 2.2416, "step": 34531 }, { "epoch": 6.58, "learning_rate": 1.027721639656816e-06, "loss": 1.7761, "step": 34532 }, { "epoch": 6.58, "learning_rate": 1.0276644423260248e-06, "loss": 1.731, "step": 34533 }, { "epoch": 6.58, "learning_rate": 1.0276072449952336e-06, "loss": 1.4516, "step": 34534 }, { "epoch": 6.58, "learning_rate": 1.0275500476644424e-06, "loss": 2.3617, "step": 34535 }, { "epoch": 6.58, "learning_rate": 1.0274928503336512e-06, "loss": 2.3154, "step": 34536 }, { "epoch": 6.58, "learning_rate": 1.02743565300286e-06, "loss": 1.7107, "step": 34537 }, { "epoch": 6.58, "learning_rate": 1.0273784556720685e-06, "loss": 2.6219, "step": 34538 }, { "epoch": 6.58, "learning_rate": 1.0273212583412773e-06, "loss": 1.3827, "step": 34539 }, { "epoch": 6.58, "learning_rate": 1.0272640610104863e-06, "loss": 2.411, "step": 34540 }, { "epoch": 6.58, "learning_rate": 1.027206863679695e-06, "loss": 1.0085, "step": 34541 }, { "epoch": 6.58, "learning_rate": 1.0271496663489037e-06, "loss": 1.6496, "step": 34542 }, { "epoch": 6.58, "learning_rate": 1.0270924690181125e-06, "loss": 1.1734, "step": 34543 }, { "epoch": 6.58, "learning_rate": 1.0270352716873213e-06, "loss": 0.8363, "step": 34544 }, { "epoch": 6.58, "learning_rate": 1.02697807435653e-06, "loss": 1.7764, "step": 34545 }, { "epoch": 6.58, "learning_rate": 1.0269208770257389e-06, "loss": 0.865, "step": 34546 }, { "epoch": 6.58, "learning_rate": 1.0268636796949477e-06, "loss": 1.1112, "step": 34547 }, { "epoch": 6.58, "learning_rate": 1.0268064823641563e-06, "loss": 1.8823, "step": 34548 }, { "epoch": 6.58, "learning_rate": 1.026749285033365e-06, "loss": 0.4182, "step": 34549 }, { "epoch": 6.58, "learning_rate": 1.0266920877025739e-06, "loss": 1.2113, "step": 34550 }, { "epoch": 6.58, "learning_rate": 1.0266348903717826e-06, "loss": 1.0103, "step": 34551 }, { "epoch": 6.58, "learning_rate": 1.0265776930409914e-06, "loss": 2.4552, "step": 34552 }, { "epoch": 6.58, "learning_rate": 1.0265204957102002e-06, "loss": 0.6627, "step": 34553 }, { "epoch": 6.58, "learning_rate": 1.026463298379409e-06, "loss": 1.1258, "step": 34554 }, { "epoch": 6.58, "learning_rate": 1.0264061010486176e-06, "loss": 1.4156, "step": 34555 }, { "epoch": 6.58, "learning_rate": 1.0263489037178266e-06, "loss": 1.5369, "step": 34556 }, { "epoch": 6.58, "learning_rate": 1.0262917063870354e-06, "loss": 2.3173, "step": 34557 }, { "epoch": 6.58, "learning_rate": 1.026234509056244e-06, "loss": 1.2667, "step": 34558 }, { "epoch": 6.58, "learning_rate": 1.0261773117254528e-06, "loss": 1.9894, "step": 34559 }, { "epoch": 6.58, "learning_rate": 1.0261201143946616e-06, "loss": 1.6476, "step": 34560 }, { "epoch": 6.58, "learning_rate": 1.0260629170638704e-06, "loss": 1.9391, "step": 34561 }, { "epoch": 6.58, "learning_rate": 1.0260057197330792e-06, "loss": 1.0894, "step": 34562 }, { "epoch": 6.58, "learning_rate": 1.025948522402288e-06, "loss": 2.2371, "step": 34563 }, { "epoch": 6.58, "learning_rate": 1.0258913250714968e-06, "loss": 2.0525, "step": 34564 }, { "epoch": 6.58, "learning_rate": 1.0258341277407053e-06, "loss": 1.8648, "step": 34565 }, { "epoch": 6.58, "learning_rate": 1.0257769304099141e-06, "loss": 1.1978, "step": 34566 }, { "epoch": 6.58, "learning_rate": 1.0257197330791231e-06, "loss": 0.4683, "step": 34567 }, { "epoch": 6.58, "learning_rate": 1.025662535748332e-06, "loss": 1.0998, "step": 34568 }, { "epoch": 6.58, "learning_rate": 1.0256053384175405e-06, "loss": 1.3463, "step": 34569 }, { "epoch": 6.58, "learning_rate": 1.0255481410867493e-06, "loss": 1.938, "step": 34570 }, { "epoch": 6.58, "learning_rate": 1.025490943755958e-06, "loss": 1.3972, "step": 34571 }, { "epoch": 6.59, "learning_rate": 1.0254337464251669e-06, "loss": 1.3927, "step": 34572 }, { "epoch": 6.59, "learning_rate": 1.0253765490943757e-06, "loss": 1.3143, "step": 34573 }, { "epoch": 6.59, "learning_rate": 1.0253193517635845e-06, "loss": 1.4475, "step": 34574 }, { "epoch": 6.59, "learning_rate": 1.025262154432793e-06, "loss": 2.7334, "step": 34575 }, { "epoch": 6.59, "learning_rate": 1.0252049571020019e-06, "loss": 1.8066, "step": 34576 }, { "epoch": 6.59, "learning_rate": 1.0251477597712106e-06, "loss": 2.4186, "step": 34577 }, { "epoch": 6.59, "learning_rate": 1.0250905624404194e-06, "loss": 2.5672, "step": 34578 }, { "epoch": 6.59, "learning_rate": 1.0250333651096282e-06, "loss": 3.2662, "step": 34579 }, { "epoch": 6.59, "learning_rate": 1.024976167778837e-06, "loss": 2.5319, "step": 34580 }, { "epoch": 6.59, "learning_rate": 1.0249189704480458e-06, "loss": 3.0529, "step": 34581 }, { "epoch": 6.59, "learning_rate": 1.0248617731172544e-06, "loss": 2.1807, "step": 34582 }, { "epoch": 6.59, "learning_rate": 1.0248045757864632e-06, "loss": 2.2457, "step": 34583 }, { "epoch": 6.59, "learning_rate": 1.0247473784556722e-06, "loss": 1.3409, "step": 34584 }, { "epoch": 6.59, "learning_rate": 1.024690181124881e-06, "loss": 2.1939, "step": 34585 }, { "epoch": 6.59, "learning_rate": 1.0246329837940896e-06, "loss": 3.3057, "step": 34586 }, { "epoch": 6.59, "learning_rate": 1.0245757864632984e-06, "loss": 2.1652, "step": 34587 }, { "epoch": 6.59, "learning_rate": 1.0245185891325072e-06, "loss": 1.5659, "step": 34588 }, { "epoch": 6.59, "learning_rate": 1.024461391801716e-06, "loss": 2.1982, "step": 34589 }, { "epoch": 6.59, "learning_rate": 1.0244041944709248e-06, "loss": 1.7725, "step": 34590 }, { "epoch": 6.59, "learning_rate": 1.0243469971401336e-06, "loss": 2.049, "step": 34591 }, { "epoch": 6.59, "learning_rate": 1.0242897998093421e-06, "loss": 2.4814, "step": 34592 }, { "epoch": 6.59, "learning_rate": 1.024232602478551e-06, "loss": 2.0241, "step": 34593 }, { "epoch": 6.59, "learning_rate": 1.0241754051477597e-06, "loss": 0.9384, "step": 34594 }, { "epoch": 6.59, "learning_rate": 1.0241182078169687e-06, "loss": 1.2284, "step": 34595 }, { "epoch": 6.59, "learning_rate": 1.0240610104861773e-06, "loss": 2.7103, "step": 34596 }, { "epoch": 6.59, "learning_rate": 1.024003813155386e-06, "loss": 1.1756, "step": 34597 }, { "epoch": 6.59, "learning_rate": 1.023946615824595e-06, "loss": 2.7376, "step": 34598 }, { "epoch": 6.59, "learning_rate": 1.0238894184938035e-06, "loss": 1.7576, "step": 34599 }, { "epoch": 6.59, "learning_rate": 1.0238322211630125e-06, "loss": 3.01, "step": 34600 }, { "epoch": 6.59, "learning_rate": 1.0237750238322213e-06, "loss": 2.1172, "step": 34601 }, { "epoch": 6.59, "learning_rate": 1.02371782650143e-06, "loss": 2.1803, "step": 34602 }, { "epoch": 6.59, "learning_rate": 1.0236606291706387e-06, "loss": 2.7433, "step": 34603 }, { "epoch": 6.59, "learning_rate": 1.0236034318398474e-06, "loss": 1.8303, "step": 34604 }, { "epoch": 6.59, "learning_rate": 1.0235462345090562e-06, "loss": 1.6265, "step": 34605 }, { "epoch": 6.59, "learning_rate": 1.023489037178265e-06, "loss": 1.4651, "step": 34606 }, { "epoch": 6.59, "learning_rate": 1.0234318398474738e-06, "loss": 1.3993, "step": 34607 }, { "epoch": 6.59, "learning_rate": 1.0233746425166826e-06, "loss": 1.2591, "step": 34608 }, { "epoch": 6.59, "learning_rate": 1.0233174451858914e-06, "loss": 1.5264, "step": 34609 }, { "epoch": 6.59, "learning_rate": 1.0232602478551e-06, "loss": 1.9443, "step": 34610 }, { "epoch": 6.59, "learning_rate": 1.023203050524309e-06, "loss": 2.4738, "step": 34611 }, { "epoch": 6.59, "learning_rate": 1.0231458531935178e-06, "loss": 2.5716, "step": 34612 }, { "epoch": 6.59, "learning_rate": 1.0230886558627264e-06, "loss": 2.1956, "step": 34613 }, { "epoch": 6.59, "learning_rate": 1.0230314585319352e-06, "loss": 3.1034, "step": 34614 }, { "epoch": 6.59, "learning_rate": 1.022974261201144e-06, "loss": 2.1236, "step": 34615 }, { "epoch": 6.59, "learning_rate": 1.0229170638703528e-06, "loss": 2.0349, "step": 34616 }, { "epoch": 6.59, "learning_rate": 1.0228598665395616e-06, "loss": 1.7162, "step": 34617 }, { "epoch": 6.59, "learning_rate": 1.0228026692087703e-06, "loss": 3.1887, "step": 34618 }, { "epoch": 6.59, "learning_rate": 1.0227454718779791e-06, "loss": 2.3286, "step": 34619 }, { "epoch": 6.59, "learning_rate": 1.0226882745471877e-06, "loss": 2.7986, "step": 34620 }, { "epoch": 6.59, "learning_rate": 1.0226310772163965e-06, "loss": 0.9297, "step": 34621 }, { "epoch": 6.59, "learning_rate": 1.0225738798856053e-06, "loss": 1.9524, "step": 34622 }, { "epoch": 6.59, "learning_rate": 1.022516682554814e-06, "loss": 1.6418, "step": 34623 }, { "epoch": 6.6, "learning_rate": 1.022459485224023e-06, "loss": 2.342, "step": 34624 }, { "epoch": 6.6, "learning_rate": 1.0224022878932317e-06, "loss": 2.0884, "step": 34625 }, { "epoch": 6.6, "learning_rate": 1.0223450905624405e-06, "loss": 2.5697, "step": 34626 }, { "epoch": 6.6, "learning_rate": 1.022287893231649e-06, "loss": 1.2038, "step": 34627 }, { "epoch": 6.6, "learning_rate": 1.022230695900858e-06, "loss": 0.363, "step": 34628 }, { "epoch": 6.6, "learning_rate": 1.0221734985700669e-06, "loss": 2.1883, "step": 34629 }, { "epoch": 6.6, "learning_rate": 1.0221163012392755e-06, "loss": 2.9679, "step": 34630 }, { "epoch": 6.6, "learning_rate": 1.0220591039084842e-06, "loss": 1.0405, "step": 34631 }, { "epoch": 6.6, "learning_rate": 1.022001906577693e-06, "loss": 2.8045, "step": 34632 }, { "epoch": 6.6, "learning_rate": 1.0219447092469018e-06, "loss": 1.0606, "step": 34633 }, { "epoch": 6.6, "learning_rate": 1.0218875119161106e-06, "loss": 1.215, "step": 34634 }, { "epoch": 6.6, "learning_rate": 1.0218303145853194e-06, "loss": 2.4478, "step": 34635 }, { "epoch": 6.6, "learning_rate": 1.0217731172545282e-06, "loss": 1.7569, "step": 34636 }, { "epoch": 6.6, "learning_rate": 1.0217159199237368e-06, "loss": 2.2917, "step": 34637 }, { "epoch": 6.6, "learning_rate": 1.0216587225929456e-06, "loss": 1.9071, "step": 34638 }, { "epoch": 6.6, "learning_rate": 1.0216015252621546e-06, "loss": 1.6848, "step": 34639 }, { "epoch": 6.6, "learning_rate": 1.0215443279313632e-06, "loss": 3.0137, "step": 34640 }, { "epoch": 6.6, "learning_rate": 1.021487130600572e-06, "loss": 2.471, "step": 34641 }, { "epoch": 6.6, "learning_rate": 1.0214299332697808e-06, "loss": 3.0245, "step": 34642 }, { "epoch": 6.6, "learning_rate": 1.0213727359389896e-06, "loss": 1.5327, "step": 34643 }, { "epoch": 6.6, "learning_rate": 1.0213155386081984e-06, "loss": 3.2055, "step": 34644 }, { "epoch": 6.6, "learning_rate": 1.0212583412774071e-06, "loss": 2.8145, "step": 34645 }, { "epoch": 6.6, "learning_rate": 1.021201143946616e-06, "loss": 2.436, "step": 34646 }, { "epoch": 6.6, "learning_rate": 1.0211439466158245e-06, "loss": 1.9099, "step": 34647 }, { "epoch": 6.6, "learning_rate": 1.0210867492850333e-06, "loss": 0.8981, "step": 34648 }, { "epoch": 6.6, "learning_rate": 1.0210295519542421e-06, "loss": 1.68, "step": 34649 }, { "epoch": 6.6, "learning_rate": 1.0209723546234511e-06, "loss": 2.6145, "step": 34650 }, { "epoch": 6.6, "learning_rate": 1.0209151572926597e-06, "loss": 0.6786, "step": 34651 }, { "epoch": 6.6, "learning_rate": 1.0208579599618685e-06, "loss": 2.8899, "step": 34652 }, { "epoch": 6.6, "learning_rate": 1.0208007626310773e-06, "loss": 1.1203, "step": 34653 }, { "epoch": 6.6, "learning_rate": 1.0207435653002859e-06, "loss": 1.2821, "step": 34654 }, { "epoch": 6.6, "learning_rate": 1.0206863679694949e-06, "loss": 1.7047, "step": 34655 }, { "epoch": 6.6, "learning_rate": 1.0206291706387037e-06, "loss": 1.9193, "step": 34656 }, { "epoch": 6.6, "learning_rate": 1.0205719733079122e-06, "loss": 1.4401, "step": 34657 }, { "epoch": 6.6, "learning_rate": 1.020514775977121e-06, "loss": 1.5362, "step": 34658 }, { "epoch": 6.6, "learning_rate": 1.0204575786463298e-06, "loss": 2.3117, "step": 34659 }, { "epoch": 6.6, "learning_rate": 1.0204003813155386e-06, "loss": 3.5156, "step": 34660 }, { "epoch": 6.6, "learning_rate": 1.0203431839847474e-06, "loss": 0.8528, "step": 34661 }, { "epoch": 6.6, "learning_rate": 1.0202859866539562e-06, "loss": 1.7, "step": 34662 }, { "epoch": 6.6, "learning_rate": 1.020228789323165e-06, "loss": 3.0944, "step": 34663 }, { "epoch": 6.6, "learning_rate": 1.0201715919923736e-06, "loss": 1.3775, "step": 34664 }, { "epoch": 6.6, "learning_rate": 1.0201143946615824e-06, "loss": 0.8359, "step": 34665 }, { "epoch": 6.6, "learning_rate": 1.0200571973307912e-06, "loss": 2.2085, "step": 34666 }, { "epoch": 6.6, "learning_rate": 1.0200000000000002e-06, "loss": 2.7675, "step": 34667 }, { "epoch": 6.6, "learning_rate": 1.0199428026692088e-06, "loss": 1.506, "step": 34668 }, { "epoch": 6.6, "learning_rate": 1.0198856053384176e-06, "loss": 1.0897, "step": 34669 }, { "epoch": 6.6, "learning_rate": 1.0198284080076264e-06, "loss": 1.2339, "step": 34670 }, { "epoch": 6.6, "learning_rate": 1.019771210676835e-06, "loss": 2.1971, "step": 34671 }, { "epoch": 6.6, "learning_rate": 1.019714013346044e-06, "loss": 1.6773, "step": 34672 }, { "epoch": 6.6, "learning_rate": 1.0196568160152527e-06, "loss": 0.829, "step": 34673 }, { "epoch": 6.6, "learning_rate": 1.0195996186844615e-06, "loss": 1.7137, "step": 34674 }, { "epoch": 6.6, "learning_rate": 1.0195424213536701e-06, "loss": 2.3267, "step": 34675 }, { "epoch": 6.6, "learning_rate": 1.019485224022879e-06, "loss": 2.6629, "step": 34676 }, { "epoch": 6.61, "learning_rate": 1.0194280266920877e-06, "loss": 1.8379, "step": 34677 }, { "epoch": 6.61, "learning_rate": 1.0193708293612965e-06, "loss": 2.8817, "step": 34678 }, { "epoch": 6.61, "learning_rate": 1.0193136320305053e-06, "loss": 0.8479, "step": 34679 }, { "epoch": 6.61, "learning_rate": 1.019256434699714e-06, "loss": 3.5333, "step": 34680 }, { "epoch": 6.61, "learning_rate": 1.0191992373689227e-06, "loss": 1.0221, "step": 34681 }, { "epoch": 6.61, "learning_rate": 1.0191420400381315e-06, "loss": 2.0525, "step": 34682 }, { "epoch": 6.61, "learning_rate": 1.0190848427073405e-06, "loss": 2.0492, "step": 34683 }, { "epoch": 6.61, "learning_rate": 1.0190276453765493e-06, "loss": 1.7053, "step": 34684 }, { "epoch": 6.61, "learning_rate": 1.0189704480457578e-06, "loss": 1.0526, "step": 34685 }, { "epoch": 6.61, "learning_rate": 1.0189132507149666e-06, "loss": 1.6139, "step": 34686 }, { "epoch": 6.61, "learning_rate": 1.0188560533841754e-06, "loss": 0.4676, "step": 34687 }, { "epoch": 6.61, "learning_rate": 1.0187988560533842e-06, "loss": 0.6832, "step": 34688 }, { "epoch": 6.61, "learning_rate": 1.018741658722593e-06, "loss": 2.0789, "step": 34689 }, { "epoch": 6.61, "learning_rate": 1.0186844613918018e-06, "loss": 1.7526, "step": 34690 }, { "epoch": 6.61, "learning_rate": 1.0186272640610106e-06, "loss": 1.2436, "step": 34691 }, { "epoch": 6.61, "learning_rate": 1.0185700667302192e-06, "loss": 2.8514, "step": 34692 }, { "epoch": 6.61, "learning_rate": 1.018512869399428e-06, "loss": 1.4802, "step": 34693 }, { "epoch": 6.61, "learning_rate": 1.0184556720686368e-06, "loss": 2.5389, "step": 34694 }, { "epoch": 6.61, "learning_rate": 1.0183984747378456e-06, "loss": 1.2013, "step": 34695 }, { "epoch": 6.61, "learning_rate": 1.0183412774070544e-06, "loss": 3.3113, "step": 34696 }, { "epoch": 6.61, "learning_rate": 1.0182840800762632e-06, "loss": 0.2752, "step": 34697 }, { "epoch": 6.61, "learning_rate": 1.0182268827454717e-06, "loss": 2.1436, "step": 34698 }, { "epoch": 6.61, "learning_rate": 1.0181696854146807e-06, "loss": 1.6464, "step": 34699 }, { "epoch": 6.61, "learning_rate": 1.0181124880838895e-06, "loss": 0.8969, "step": 34700 }, { "epoch": 6.61, "learning_rate": 1.0180552907530983e-06, "loss": 1.6648, "step": 34701 }, { "epoch": 6.61, "learning_rate": 1.017998093422307e-06, "loss": 2.7629, "step": 34702 }, { "epoch": 6.61, "learning_rate": 1.0179408960915157e-06, "loss": 1.1968, "step": 34703 }, { "epoch": 6.61, "learning_rate": 1.0178836987607245e-06, "loss": 0.896, "step": 34704 }, { "epoch": 6.61, "learning_rate": 1.0178265014299333e-06, "loss": 1.6491, "step": 34705 }, { "epoch": 6.61, "learning_rate": 1.017769304099142e-06, "loss": 1.7504, "step": 34706 }, { "epoch": 6.61, "learning_rate": 1.0177121067683509e-06, "loss": 0.7464, "step": 34707 }, { "epoch": 6.61, "learning_rate": 1.0176549094375597e-06, "loss": 1.7873, "step": 34708 }, { "epoch": 6.61, "learning_rate": 1.0175977121067683e-06, "loss": 0.5295, "step": 34709 }, { "epoch": 6.61, "learning_rate": 1.017540514775977e-06, "loss": 0.7278, "step": 34710 }, { "epoch": 6.61, "learning_rate": 1.017483317445186e-06, "loss": 2.7669, "step": 34711 }, { "epoch": 6.61, "learning_rate": 1.0174261201143946e-06, "loss": 2.2151, "step": 34712 }, { "epoch": 6.61, "learning_rate": 1.0173689227836034e-06, "loss": 1.5325, "step": 34713 }, { "epoch": 6.61, "learning_rate": 1.0173117254528122e-06, "loss": 1.0662, "step": 34714 }, { "epoch": 6.61, "learning_rate": 1.017254528122021e-06, "loss": 1.8503, "step": 34715 }, { "epoch": 6.61, "learning_rate": 1.0171973307912298e-06, "loss": 1.2286, "step": 34716 }, { "epoch": 6.61, "learning_rate": 1.0171401334604386e-06, "loss": 2.8169, "step": 34717 }, { "epoch": 6.61, "learning_rate": 1.0170829361296474e-06, "loss": 2.2331, "step": 34718 }, { "epoch": 6.61, "learning_rate": 1.017025738798856e-06, "loss": 2.3381, "step": 34719 }, { "epoch": 6.61, "learning_rate": 1.0169685414680648e-06, "loss": 1.24, "step": 34720 }, { "epoch": 6.61, "learning_rate": 1.0169113441372736e-06, "loss": 0.9754, "step": 34721 }, { "epoch": 6.61, "learning_rate": 1.0168541468064824e-06, "loss": 1.7812, "step": 34722 }, { "epoch": 6.61, "learning_rate": 1.0167969494756912e-06, "loss": 0.9012, "step": 34723 }, { "epoch": 6.61, "learning_rate": 1.0167397521449e-06, "loss": 1.1821, "step": 34724 }, { "epoch": 6.61, "learning_rate": 1.0166825548141087e-06, "loss": 2.4288, "step": 34725 }, { "epoch": 6.61, "learning_rate": 1.0166253574833173e-06, "loss": 2.0428, "step": 34726 }, { "epoch": 6.61, "learning_rate": 1.0165681601525263e-06, "loss": 1.3335, "step": 34727 }, { "epoch": 6.61, "learning_rate": 1.0165109628217351e-06, "loss": 1.5746, "step": 34728 }, { "epoch": 6.62, "learning_rate": 1.0164537654909437e-06, "loss": 3.3258, "step": 34729 }, { "epoch": 6.62, "learning_rate": 1.0163965681601525e-06, "loss": 1.3831, "step": 34730 }, { "epoch": 6.62, "learning_rate": 1.0163393708293613e-06, "loss": 0.8252, "step": 34731 }, { "epoch": 6.62, "learning_rate": 1.01628217349857e-06, "loss": 1.751, "step": 34732 }, { "epoch": 6.62, "learning_rate": 1.0162249761677789e-06, "loss": 2.3301, "step": 34733 }, { "epoch": 6.62, "learning_rate": 1.0161677788369877e-06, "loss": 1.2651, "step": 34734 }, { "epoch": 6.62, "learning_rate": 1.0161105815061965e-06, "loss": 1.7359, "step": 34735 }, { "epoch": 6.62, "learning_rate": 1.016053384175405e-06, "loss": 0.9281, "step": 34736 }, { "epoch": 6.62, "learning_rate": 1.0159961868446138e-06, "loss": 1.207, "step": 34737 }, { "epoch": 6.62, "learning_rate": 1.0159389895138226e-06, "loss": 1.0273, "step": 34738 }, { "epoch": 6.62, "learning_rate": 1.0158817921830316e-06, "loss": 2.2933, "step": 34739 }, { "epoch": 6.62, "learning_rate": 1.0158245948522402e-06, "loss": 1.4712, "step": 34740 }, { "epoch": 6.62, "learning_rate": 1.015767397521449e-06, "loss": 0.8422, "step": 34741 }, { "epoch": 6.62, "learning_rate": 1.0157102001906578e-06, "loss": 1.5177, "step": 34742 }, { "epoch": 6.62, "learning_rate": 1.0156530028598666e-06, "loss": 0.7942, "step": 34743 }, { "epoch": 6.62, "learning_rate": 1.0155958055290754e-06, "loss": 1.5625, "step": 34744 }, { "epoch": 6.62, "learning_rate": 1.0155386081982842e-06, "loss": 2.4684, "step": 34745 }, { "epoch": 6.62, "learning_rate": 1.0154814108674928e-06, "loss": 0.9882, "step": 34746 }, { "epoch": 6.62, "learning_rate": 1.0154242135367016e-06, "loss": 2.6519, "step": 34747 }, { "epoch": 6.62, "learning_rate": 1.0153670162059104e-06, "loss": 1.3099, "step": 34748 }, { "epoch": 6.62, "learning_rate": 1.0153098188751192e-06, "loss": 1.525, "step": 34749 }, { "epoch": 6.62, "learning_rate": 1.015252621544328e-06, "loss": 1.3216, "step": 34750 }, { "epoch": 6.62, "learning_rate": 1.0151954242135367e-06, "loss": 1.4271, "step": 34751 }, { "epoch": 6.62, "learning_rate": 1.0151382268827455e-06, "loss": 2.5694, "step": 34752 }, { "epoch": 6.62, "learning_rate": 1.0150810295519541e-06, "loss": 1.5867, "step": 34753 }, { "epoch": 6.62, "learning_rate": 1.015023832221163e-06, "loss": 0.6386, "step": 34754 }, { "epoch": 6.62, "learning_rate": 1.014966634890372e-06, "loss": 2.1842, "step": 34755 }, { "epoch": 6.62, "learning_rate": 1.0149094375595807e-06, "loss": 1.9567, "step": 34756 }, { "epoch": 6.62, "learning_rate": 1.0148522402287893e-06, "loss": 2.1153, "step": 34757 }, { "epoch": 6.62, "learning_rate": 1.014795042897998e-06, "loss": 1.8242, "step": 34758 }, { "epoch": 6.62, "learning_rate": 1.0147378455672069e-06, "loss": 2.1448, "step": 34759 }, { "epoch": 6.62, "learning_rate": 1.0146806482364157e-06, "loss": 0.8385, "step": 34760 }, { "epoch": 6.62, "learning_rate": 1.0146234509056245e-06, "loss": 2.0363, "step": 34761 }, { "epoch": 6.62, "learning_rate": 1.0145662535748333e-06, "loss": 1.5504, "step": 34762 }, { "epoch": 6.62, "learning_rate": 1.0145090562440419e-06, "loss": 1.8827, "step": 34763 }, { "epoch": 6.62, "learning_rate": 1.0144518589132506e-06, "loss": 1.2494, "step": 34764 }, { "epoch": 6.62, "learning_rate": 1.0143946615824594e-06, "loss": 2.0209, "step": 34765 }, { "epoch": 6.62, "learning_rate": 1.0143374642516684e-06, "loss": 1.5868, "step": 34766 }, { "epoch": 6.62, "learning_rate": 1.014280266920877e-06, "loss": 1.7161, "step": 34767 }, { "epoch": 6.62, "learning_rate": 1.0142230695900858e-06, "loss": 2.739, "step": 34768 }, { "epoch": 6.62, "learning_rate": 1.0141658722592946e-06, "loss": 1.0983, "step": 34769 }, { "epoch": 6.62, "learning_rate": 1.0141086749285032e-06, "loss": 1.6631, "step": 34770 }, { "epoch": 6.62, "learning_rate": 1.0140514775977122e-06, "loss": 1.8394, "step": 34771 }, { "epoch": 6.62, "learning_rate": 1.013994280266921e-06, "loss": 2.302, "step": 34772 }, { "epoch": 6.62, "learning_rate": 1.0139370829361298e-06, "loss": 1.415, "step": 34773 }, { "epoch": 6.62, "learning_rate": 1.0138798856053384e-06, "loss": 1.2366, "step": 34774 }, { "epoch": 6.62, "learning_rate": 1.0138226882745472e-06, "loss": 2.4035, "step": 34775 }, { "epoch": 6.62, "learning_rate": 1.013765490943756e-06, "loss": 1.919, "step": 34776 }, { "epoch": 6.62, "learning_rate": 1.0137082936129648e-06, "loss": 2.3543, "step": 34777 }, { "epoch": 6.62, "learning_rate": 1.0136510962821735e-06, "loss": 1.4745, "step": 34778 }, { "epoch": 6.62, "learning_rate": 1.0135938989513823e-06, "loss": 1.7282, "step": 34779 }, { "epoch": 6.62, "learning_rate": 1.0135367016205911e-06, "loss": 0.9355, "step": 34780 }, { "epoch": 6.62, "learning_rate": 1.0134795042897997e-06, "loss": 1.2841, "step": 34781 }, { "epoch": 6.63, "learning_rate": 1.0134223069590085e-06, "loss": 2.5106, "step": 34782 }, { "epoch": 6.63, "learning_rate": 1.0133651096282175e-06, "loss": 1.9553, "step": 34783 }, { "epoch": 6.63, "learning_rate": 1.013307912297426e-06, "loss": 1.9381, "step": 34784 }, { "epoch": 6.63, "learning_rate": 1.0132507149666349e-06, "loss": 1.8131, "step": 34785 }, { "epoch": 6.63, "learning_rate": 1.0131935176358437e-06, "loss": 1.9097, "step": 34786 }, { "epoch": 6.63, "learning_rate": 1.0131363203050525e-06, "loss": 2.3104, "step": 34787 }, { "epoch": 6.63, "learning_rate": 1.0130791229742613e-06, "loss": 1.3537, "step": 34788 }, { "epoch": 6.63, "learning_rate": 1.01302192564347e-06, "loss": 2.9864, "step": 34789 }, { "epoch": 6.63, "learning_rate": 1.0129647283126789e-06, "loss": 2.9682, "step": 34790 }, { "epoch": 6.63, "learning_rate": 1.0129075309818874e-06, "loss": 1.2311, "step": 34791 }, { "epoch": 6.63, "learning_rate": 1.0128503336510962e-06, "loss": 1.3958, "step": 34792 }, { "epoch": 6.63, "learning_rate": 1.012793136320305e-06, "loss": 3.1566, "step": 34793 }, { "epoch": 6.63, "learning_rate": 1.0127359389895138e-06, "loss": 3.1621, "step": 34794 }, { "epoch": 6.63, "learning_rate": 1.0126787416587226e-06, "loss": 2.1008, "step": 34795 }, { "epoch": 6.63, "learning_rate": 1.0126215443279314e-06, "loss": 1.0597, "step": 34796 }, { "epoch": 6.63, "learning_rate": 1.0125643469971402e-06, "loss": 1.4454, "step": 34797 }, { "epoch": 6.63, "learning_rate": 1.0125071496663488e-06, "loss": 2.5791, "step": 34798 }, { "epoch": 6.63, "learning_rate": 1.0124499523355578e-06, "loss": 1.1017, "step": 34799 }, { "epoch": 6.63, "learning_rate": 1.0123927550047666e-06, "loss": 1.7976, "step": 34800 }, { "epoch": 6.63, "learning_rate": 1.0123355576739752e-06, "loss": 2.0265, "step": 34801 }, { "epoch": 6.63, "learning_rate": 1.012278360343184e-06, "loss": 1.4299, "step": 34802 }, { "epoch": 6.63, "learning_rate": 1.0122211630123928e-06, "loss": 2.0867, "step": 34803 }, { "epoch": 6.63, "learning_rate": 1.0121639656816016e-06, "loss": 2.0932, "step": 34804 }, { "epoch": 6.63, "learning_rate": 1.0121067683508103e-06, "loss": 2.3784, "step": 34805 }, { "epoch": 6.63, "learning_rate": 1.0120495710200191e-06, "loss": 2.5688, "step": 34806 }, { "epoch": 6.63, "learning_rate": 1.011992373689228e-06, "loss": 1.7633, "step": 34807 }, { "epoch": 6.63, "learning_rate": 1.0119351763584365e-06, "loss": 2.0671, "step": 34808 }, { "epoch": 6.63, "learning_rate": 1.0118779790276453e-06, "loss": 0.7444, "step": 34809 }, { "epoch": 6.63, "learning_rate": 1.0118207816968543e-06, "loss": 1.2671, "step": 34810 }, { "epoch": 6.63, "learning_rate": 1.011763584366063e-06, "loss": 3.2541, "step": 34811 }, { "epoch": 6.63, "learning_rate": 1.0117063870352717e-06, "loss": 2.2105, "step": 34812 }, { "epoch": 6.63, "learning_rate": 1.0116491897044805e-06, "loss": 2.8223, "step": 34813 }, { "epoch": 6.63, "learning_rate": 1.0115919923736893e-06, "loss": 2.1313, "step": 34814 }, { "epoch": 6.63, "learning_rate": 1.011534795042898e-06, "loss": 2.8554, "step": 34815 }, { "epoch": 6.63, "learning_rate": 1.0114775977121069e-06, "loss": 2.0104, "step": 34816 }, { "epoch": 6.63, "learning_rate": 1.0114204003813157e-06, "loss": 2.9257, "step": 34817 }, { "epoch": 6.63, "learning_rate": 1.0113632030505242e-06, "loss": 1.0846, "step": 34818 }, { "epoch": 6.63, "learning_rate": 1.011306005719733e-06, "loss": 2.4664, "step": 34819 }, { "epoch": 6.63, "learning_rate": 1.0112488083889418e-06, "loss": 1.0748, "step": 34820 }, { "epoch": 6.63, "learning_rate": 1.0111916110581506e-06, "loss": 2.0223, "step": 34821 }, { "epoch": 6.63, "learning_rate": 1.0111344137273594e-06, "loss": 1.4107, "step": 34822 }, { "epoch": 6.63, "learning_rate": 1.0110772163965682e-06, "loss": 1.3837, "step": 34823 }, { "epoch": 6.63, "learning_rate": 1.011020019065777e-06, "loss": 1.3638, "step": 34824 }, { "epoch": 6.63, "learning_rate": 1.0109628217349856e-06, "loss": 2.0431, "step": 34825 }, { "epoch": 6.63, "learning_rate": 1.0109056244041944e-06, "loss": 1.6354, "step": 34826 }, { "epoch": 6.63, "learning_rate": 1.0108484270734034e-06, "loss": 0.9196, "step": 34827 }, { "epoch": 6.63, "learning_rate": 1.010791229742612e-06, "loss": 2.0895, "step": 34828 }, { "epoch": 6.63, "learning_rate": 1.0107340324118208e-06, "loss": 0.8968, "step": 34829 }, { "epoch": 6.63, "learning_rate": 1.0106768350810296e-06, "loss": 1.0024, "step": 34830 }, { "epoch": 6.63, "learning_rate": 1.0106196377502383e-06, "loss": 1.7504, "step": 34831 }, { "epoch": 6.63, "learning_rate": 1.0105624404194471e-06, "loss": 1.4945, "step": 34832 }, { "epoch": 6.63, "learning_rate": 1.010505243088656e-06, "loss": 1.4426, "step": 34833 }, { "epoch": 6.64, "learning_rate": 1.0104480457578647e-06, "loss": 1.2252, "step": 34834 }, { "epoch": 6.64, "learning_rate": 1.0103908484270733e-06, "loss": 1.4818, "step": 34835 }, { "epoch": 6.64, "learning_rate": 1.0103336510962821e-06, "loss": 1.6828, "step": 34836 }, { "epoch": 6.64, "learning_rate": 1.010276453765491e-06, "loss": 1.8415, "step": 34837 }, { "epoch": 6.64, "learning_rate": 1.0102192564347e-06, "loss": 1.6269, "step": 34838 }, { "epoch": 6.64, "learning_rate": 1.0101620591039085e-06, "loss": 2.0445, "step": 34839 }, { "epoch": 6.64, "learning_rate": 1.0101048617731173e-06, "loss": 3.3478, "step": 34840 }, { "epoch": 6.64, "learning_rate": 1.010047664442326e-06, "loss": 2.36, "step": 34841 }, { "epoch": 6.64, "learning_rate": 1.0099904671115347e-06, "loss": 1.7379, "step": 34842 }, { "epoch": 6.64, "learning_rate": 1.0099332697807437e-06, "loss": 0.9005, "step": 34843 }, { "epoch": 6.64, "learning_rate": 1.0098760724499525e-06, "loss": 1.0809, "step": 34844 }, { "epoch": 6.64, "learning_rate": 1.0098188751191613e-06, "loss": 0.9275, "step": 34845 }, { "epoch": 6.64, "learning_rate": 1.0097616777883698e-06, "loss": 1.3518, "step": 34846 }, { "epoch": 6.64, "learning_rate": 1.0097044804575786e-06, "loss": 1.9059, "step": 34847 }, { "epoch": 6.64, "learning_rate": 1.0096472831267874e-06, "loss": 0.5427, "step": 34848 }, { "epoch": 6.64, "learning_rate": 1.0095900857959962e-06, "loss": 3.2035, "step": 34849 }, { "epoch": 6.64, "learning_rate": 1.009532888465205e-06, "loss": 1.1025, "step": 34850 }, { "epoch": 6.64, "learning_rate": 1.0094756911344138e-06, "loss": 3.0332, "step": 34851 }, { "epoch": 6.64, "learning_rate": 1.0094184938036224e-06, "loss": 2.6107, "step": 34852 }, { "epoch": 6.64, "learning_rate": 1.0093612964728312e-06, "loss": 2.9777, "step": 34853 }, { "epoch": 6.64, "learning_rate": 1.0093040991420402e-06, "loss": 1.5824, "step": 34854 }, { "epoch": 6.64, "learning_rate": 1.009246901811249e-06, "loss": 0.9512, "step": 34855 }, { "epoch": 6.64, "learning_rate": 1.0091897044804576e-06, "loss": 1.4362, "step": 34856 }, { "epoch": 6.64, "learning_rate": 1.0091325071496664e-06, "loss": 2.8232, "step": 34857 }, { "epoch": 6.64, "learning_rate": 1.0090753098188751e-06, "loss": 2.3551, "step": 34858 }, { "epoch": 6.64, "learning_rate": 1.009018112488084e-06, "loss": 2.1379, "step": 34859 }, { "epoch": 6.64, "learning_rate": 1.0089609151572927e-06, "loss": 1.7606, "step": 34860 }, { "epoch": 6.64, "learning_rate": 1.0089037178265015e-06, "loss": 1.4694, "step": 34861 }, { "epoch": 6.64, "learning_rate": 1.0088465204957103e-06, "loss": 1.6688, "step": 34862 }, { "epoch": 6.64, "learning_rate": 1.008789323164919e-06, "loss": 1.3478, "step": 34863 }, { "epoch": 6.64, "learning_rate": 1.0087321258341277e-06, "loss": 0.3686, "step": 34864 }, { "epoch": 6.64, "learning_rate": 1.0086749285033365e-06, "loss": 1.594, "step": 34865 }, { "epoch": 6.64, "learning_rate": 1.0086177311725453e-06, "loss": 1.1867, "step": 34866 }, { "epoch": 6.64, "learning_rate": 1.008560533841754e-06, "loss": 0.6917, "step": 34867 }, { "epoch": 6.64, "learning_rate": 1.0085033365109629e-06, "loss": 1.1384, "step": 34868 }, { "epoch": 6.64, "learning_rate": 1.0084461391801717e-06, "loss": 2.6498, "step": 34869 }, { "epoch": 6.64, "learning_rate": 1.0083889418493802e-06, "loss": 0.8606, "step": 34870 }, { "epoch": 6.64, "learning_rate": 1.0083317445185893e-06, "loss": 1.4973, "step": 34871 }, { "epoch": 6.64, "learning_rate": 1.008274547187798e-06, "loss": 1.0497, "step": 34872 }, { "epoch": 6.64, "learning_rate": 1.0082173498570066e-06, "loss": 1.7993, "step": 34873 }, { "epoch": 6.64, "learning_rate": 1.0081601525262154e-06, "loss": 1.4903, "step": 34874 }, { "epoch": 6.64, "learning_rate": 1.0081029551954242e-06, "loss": 0.5995, "step": 34875 }, { "epoch": 6.64, "learning_rate": 1.008045757864633e-06, "loss": 2.3625, "step": 34876 }, { "epoch": 6.64, "learning_rate": 1.0079885605338418e-06, "loss": 1.1611, "step": 34877 }, { "epoch": 6.64, "learning_rate": 1.0079313632030506e-06, "loss": 1.1356, "step": 34878 }, { "epoch": 6.64, "learning_rate": 1.0078741658722594e-06, "loss": 2.0466, "step": 34879 }, { "epoch": 6.64, "learning_rate": 1.007816968541468e-06, "loss": 2.0175, "step": 34880 }, { "epoch": 6.64, "learning_rate": 1.0077597712106768e-06, "loss": 1.242, "step": 34881 }, { "epoch": 6.64, "learning_rate": 1.0077025738798858e-06, "loss": 1.4801, "step": 34882 }, { "epoch": 6.64, "learning_rate": 1.0076453765490944e-06, "loss": 2.1836, "step": 34883 }, { "epoch": 6.64, "learning_rate": 1.0075881792183032e-06, "loss": 2.9375, "step": 34884 }, { "epoch": 6.64, "learning_rate": 1.007530981887512e-06, "loss": 0.2375, "step": 34885 }, { "epoch": 6.64, "learning_rate": 1.0074737845567207e-06, "loss": 1.9161, "step": 34886 }, { "epoch": 6.65, "learning_rate": 1.0074165872259295e-06, "loss": 1.3199, "step": 34887 }, { "epoch": 6.65, "learning_rate": 1.0073593898951383e-06, "loss": 1.494, "step": 34888 }, { "epoch": 6.65, "learning_rate": 1.0073021925643471e-06, "loss": 1.1543, "step": 34889 }, { "epoch": 6.65, "learning_rate": 1.0072449952335557e-06, "loss": 1.4145, "step": 34890 }, { "epoch": 6.65, "learning_rate": 1.0071877979027645e-06, "loss": 1.4037, "step": 34891 }, { "epoch": 6.65, "learning_rate": 1.0071306005719733e-06, "loss": 1.8247, "step": 34892 }, { "epoch": 6.65, "learning_rate": 1.007073403241182e-06, "loss": 2.1872, "step": 34893 }, { "epoch": 6.65, "learning_rate": 1.0070162059103909e-06, "loss": 1.8174, "step": 34894 }, { "epoch": 6.65, "learning_rate": 1.0069590085795997e-06, "loss": 0.913, "step": 34895 }, { "epoch": 6.65, "learning_rate": 1.0069018112488085e-06, "loss": 1.5532, "step": 34896 }, { "epoch": 6.65, "learning_rate": 1.006844613918017e-06, "loss": 2.4019, "step": 34897 }, { "epoch": 6.65, "learning_rate": 1.006787416587226e-06, "loss": 3.6441, "step": 34898 }, { "epoch": 6.65, "learning_rate": 1.0067302192564348e-06, "loss": 2.4979, "step": 34899 }, { "epoch": 6.65, "learning_rate": 1.0066730219256434e-06, "loss": 1.2754, "step": 34900 }, { "epoch": 6.65, "learning_rate": 1.0066158245948522e-06, "loss": 2.7133, "step": 34901 }, { "epoch": 6.65, "learning_rate": 1.006558627264061e-06, "loss": 2.0426, "step": 34902 }, { "epoch": 6.65, "learning_rate": 1.0065014299332698e-06, "loss": 2.0069, "step": 34903 }, { "epoch": 6.65, "learning_rate": 1.0064442326024786e-06, "loss": 2.5129, "step": 34904 }, { "epoch": 6.65, "learning_rate": 1.0063870352716874e-06, "loss": 1.6489, "step": 34905 }, { "epoch": 6.65, "learning_rate": 1.0063298379408962e-06, "loss": 1.123, "step": 34906 }, { "epoch": 6.65, "learning_rate": 1.0062726406101048e-06, "loss": 1.8844, "step": 34907 }, { "epoch": 6.65, "learning_rate": 1.0062154432793136e-06, "loss": 2.1234, "step": 34908 }, { "epoch": 6.65, "learning_rate": 1.0061582459485224e-06, "loss": 1.2873, "step": 34909 }, { "epoch": 6.65, "learning_rate": 1.0061010486177314e-06, "loss": 1.684, "step": 34910 }, { "epoch": 6.65, "learning_rate": 1.00604385128694e-06, "loss": 1.5486, "step": 34911 }, { "epoch": 6.65, "learning_rate": 1.0059866539561487e-06, "loss": 2.8581, "step": 34912 }, { "epoch": 6.65, "learning_rate": 1.0059294566253575e-06, "loss": 1.912, "step": 34913 }, { "epoch": 6.65, "learning_rate": 1.0058722592945661e-06, "loss": 1.8486, "step": 34914 }, { "epoch": 6.65, "learning_rate": 1.0058150619637751e-06, "loss": 2.4556, "step": 34915 }, { "epoch": 6.65, "learning_rate": 1.005757864632984e-06, "loss": 2.0908, "step": 34916 }, { "epoch": 6.65, "learning_rate": 1.0057006673021925e-06, "loss": 2.7091, "step": 34917 }, { "epoch": 6.65, "learning_rate": 1.0056434699714013e-06, "loss": 2.291, "step": 34918 }, { "epoch": 6.65, "learning_rate": 1.00558627264061e-06, "loss": 2.287, "step": 34919 }, { "epoch": 6.65, "learning_rate": 1.0055290753098189e-06, "loss": 1.6217, "step": 34920 }, { "epoch": 6.65, "learning_rate": 1.0054718779790277e-06, "loss": 1.5071, "step": 34921 }, { "epoch": 6.65, "learning_rate": 1.0054146806482365e-06, "loss": 2.9493, "step": 34922 }, { "epoch": 6.65, "learning_rate": 1.0053574833174453e-06, "loss": 4.3838, "step": 34923 }, { "epoch": 6.65, "learning_rate": 1.0053002859866538e-06, "loss": 2.0823, "step": 34924 }, { "epoch": 6.65, "learning_rate": 1.0052430886558626e-06, "loss": 1.2698, "step": 34925 }, { "epoch": 6.65, "learning_rate": 1.0051858913250716e-06, "loss": 1.4068, "step": 34926 }, { "epoch": 6.65, "learning_rate": 1.0051286939942804e-06, "loss": 0.3161, "step": 34927 }, { "epoch": 6.65, "learning_rate": 1.005071496663489e-06, "loss": 1.2793, "step": 34928 }, { "epoch": 6.65, "learning_rate": 1.0050142993326978e-06, "loss": 2.0292, "step": 34929 }, { "epoch": 6.65, "learning_rate": 1.0049571020019066e-06, "loss": 1.3904, "step": 34930 }, { "epoch": 6.65, "learning_rate": 1.0048999046711154e-06, "loss": 1.5848, "step": 34931 }, { "epoch": 6.65, "learning_rate": 1.0048427073403242e-06, "loss": 1.8694, "step": 34932 }, { "epoch": 6.65, "learning_rate": 1.004785510009533e-06, "loss": 2.4993, "step": 34933 }, { "epoch": 6.65, "learning_rate": 1.0047283126787418e-06, "loss": 1.3527, "step": 34934 }, { "epoch": 6.65, "learning_rate": 1.0046711153479504e-06, "loss": 1.4126, "step": 34935 }, { "epoch": 6.65, "learning_rate": 1.0046139180171592e-06, "loss": 2.3637, "step": 34936 }, { "epoch": 6.65, "learning_rate": 1.004556720686368e-06, "loss": 2.637, "step": 34937 }, { "epoch": 6.65, "learning_rate": 1.0044995233555767e-06, "loss": 1.2907, "step": 34938 }, { "epoch": 6.66, "learning_rate": 1.0044423260247855e-06, "loss": 0.8508, "step": 34939 }, { "epoch": 6.66, "learning_rate": 1.0043851286939943e-06, "loss": 1.328, "step": 34940 }, { "epoch": 6.66, "learning_rate": 1.004327931363203e-06, "loss": 0.9189, "step": 34941 }, { "epoch": 6.66, "learning_rate": 1.004270734032412e-06, "loss": 1.0301, "step": 34942 }, { "epoch": 6.66, "learning_rate": 1.0042135367016207e-06, "loss": 1.8725, "step": 34943 }, { "epoch": 6.66, "learning_rate": 1.0041563393708295e-06, "loss": 0.7565, "step": 34944 }, { "epoch": 6.66, "learning_rate": 1.004099142040038e-06, "loss": 1.6286, "step": 34945 }, { "epoch": 6.66, "learning_rate": 1.0040419447092469e-06, "loss": 2.0313, "step": 34946 }, { "epoch": 6.66, "learning_rate": 1.0039847473784557e-06, "loss": 1.3595, "step": 34947 }, { "epoch": 6.66, "learning_rate": 1.0039275500476645e-06, "loss": 1.6099, "step": 34948 }, { "epoch": 6.66, "learning_rate": 1.0038703527168733e-06, "loss": 2.6402, "step": 34949 }, { "epoch": 6.66, "learning_rate": 1.003813155386082e-06, "loss": 2.0527, "step": 34950 }, { "epoch": 6.66, "learning_rate": 1.0037559580552909e-06, "loss": 1.7024, "step": 34951 }, { "epoch": 6.66, "learning_rate": 1.0036987607244994e-06, "loss": 0.5255, "step": 34952 }, { "epoch": 6.66, "learning_rate": 1.0036415633937082e-06, "loss": 1.7716, "step": 34953 }, { "epoch": 6.66, "learning_rate": 1.0035843660629172e-06, "loss": 3.8167, "step": 34954 }, { "epoch": 6.66, "learning_rate": 1.0035271687321258e-06, "loss": 1.8805, "step": 34955 }, { "epoch": 6.66, "learning_rate": 1.0034699714013346e-06, "loss": 1.1161, "step": 34956 }, { "epoch": 6.66, "learning_rate": 1.0034127740705434e-06, "loss": 2.5945, "step": 34957 }, { "epoch": 6.66, "learning_rate": 1.003355576739752e-06, "loss": 1.6612, "step": 34958 }, { "epoch": 6.66, "learning_rate": 1.003298379408961e-06, "loss": 1.1725, "step": 34959 }, { "epoch": 6.66, "learning_rate": 1.0032411820781698e-06, "loss": 1.0406, "step": 34960 }, { "epoch": 6.66, "learning_rate": 1.0031839847473786e-06, "loss": 1.1968, "step": 34961 }, { "epoch": 6.66, "learning_rate": 1.0031267874165872e-06, "loss": 2.7693, "step": 34962 }, { "epoch": 6.66, "learning_rate": 1.003069590085796e-06, "loss": 2.0214, "step": 34963 }, { "epoch": 6.66, "learning_rate": 1.0030123927550048e-06, "loss": 2.5086, "step": 34964 }, { "epoch": 6.66, "learning_rate": 1.0029551954242135e-06, "loss": 2.4123, "step": 34965 }, { "epoch": 6.66, "learning_rate": 1.0028979980934223e-06, "loss": 1.4218, "step": 34966 }, { "epoch": 6.66, "learning_rate": 1.0028408007626311e-06, "loss": 1.4768, "step": 34967 }, { "epoch": 6.66, "learning_rate": 1.00278360343184e-06, "loss": 2.7828, "step": 34968 }, { "epoch": 6.66, "learning_rate": 1.0027264061010485e-06, "loss": 2.1874, "step": 34969 }, { "epoch": 6.66, "learning_rate": 1.0026692087702575e-06, "loss": 1.6626, "step": 34970 }, { "epoch": 6.66, "learning_rate": 1.0026120114394663e-06, "loss": 1.8849, "step": 34971 }, { "epoch": 6.66, "learning_rate": 1.0025548141086749e-06, "loss": 1.9633, "step": 34972 }, { "epoch": 6.66, "learning_rate": 1.0024976167778837e-06, "loss": 1.7798, "step": 34973 }, { "epoch": 6.66, "learning_rate": 1.0024404194470925e-06, "loss": 2.1139, "step": 34974 }, { "epoch": 6.66, "learning_rate": 1.0023832221163013e-06, "loss": 2.5667, "step": 34975 }, { "epoch": 6.66, "learning_rate": 1.00232602478551e-06, "loss": 2.553, "step": 34976 }, { "epoch": 6.66, "learning_rate": 1.0022688274547189e-06, "loss": 0.6819, "step": 34977 }, { "epoch": 6.66, "learning_rate": 1.0022116301239277e-06, "loss": 0.6432, "step": 34978 }, { "epoch": 6.66, "learning_rate": 1.0021544327931362e-06, "loss": 1.5114, "step": 34979 }, { "epoch": 6.66, "learning_rate": 1.002097235462345e-06, "loss": 1.3137, "step": 34980 }, { "epoch": 6.66, "learning_rate": 1.0020400381315538e-06, "loss": 1.9877, "step": 34981 }, { "epoch": 6.66, "learning_rate": 1.0019828408007626e-06, "loss": 2.1946, "step": 34982 }, { "epoch": 6.66, "learning_rate": 1.0019256434699714e-06, "loss": 2.0216, "step": 34983 }, { "epoch": 6.66, "learning_rate": 1.0018684461391802e-06, "loss": 2.0853, "step": 34984 }, { "epoch": 6.66, "learning_rate": 1.001811248808389e-06, "loss": 0.4905, "step": 34985 }, { "epoch": 6.66, "learning_rate": 1.0017540514775978e-06, "loss": 1.4394, "step": 34986 }, { "epoch": 6.66, "learning_rate": 1.0016968541468066e-06, "loss": 1.9762, "step": 34987 }, { "epoch": 6.66, "learning_rate": 1.0016396568160154e-06, "loss": 1.976, "step": 34988 }, { "epoch": 6.66, "learning_rate": 1.001582459485224e-06, "loss": 0.7475, "step": 34989 }, { "epoch": 6.66, "learning_rate": 1.0015252621544328e-06, "loss": 3.0326, "step": 34990 }, { "epoch": 6.66, "learning_rate": 1.0014680648236415e-06, "loss": 2.9698, "step": 34991 }, { "epoch": 6.67, "learning_rate": 1.0014108674928503e-06, "loss": 1.8051, "step": 34992 }, { "epoch": 6.67, "learning_rate": 1.0013536701620591e-06, "loss": 1.8601, "step": 34993 }, { "epoch": 6.67, "learning_rate": 1.001296472831268e-06, "loss": 3.4874, "step": 34994 }, { "epoch": 6.67, "learning_rate": 1.0012392755004767e-06, "loss": 0.9716, "step": 34995 }, { "epoch": 6.67, "learning_rate": 1.0011820781696853e-06, "loss": 2.588, "step": 34996 }, { "epoch": 6.67, "learning_rate": 1.001124880838894e-06, "loss": 2.2046, "step": 34997 }, { "epoch": 6.67, "learning_rate": 1.001067683508103e-06, "loss": 1.2432, "step": 34998 }, { "epoch": 6.67, "learning_rate": 1.0010104861773117e-06, "loss": 1.2513, "step": 34999 }, { "epoch": 6.67, "learning_rate": 1.0009532888465205e-06, "loss": 1.4848, "step": 35000 }, { "epoch": 6.67, "learning_rate": 1.0008960915157293e-06, "loss": 1.7645, "step": 35001 }, { "epoch": 6.67, "learning_rate": 1.000838894184938e-06, "loss": 1.357, "step": 35002 }, { "epoch": 6.67, "learning_rate": 1.0007816968541469e-06, "loss": 1.2779, "step": 35003 }, { "epoch": 6.67, "learning_rate": 1.0007244995233557e-06, "loss": 1.4529, "step": 35004 }, { "epoch": 6.67, "learning_rate": 1.0006673021925645e-06, "loss": 2.9906, "step": 35005 }, { "epoch": 6.67, "learning_rate": 1.000610104861773e-06, "loss": 2.1413, "step": 35006 }, { "epoch": 6.67, "learning_rate": 1.0005529075309818e-06, "loss": 1.2456, "step": 35007 }, { "epoch": 6.67, "learning_rate": 1.0004957102001906e-06, "loss": 2.3736, "step": 35008 }, { "epoch": 6.67, "learning_rate": 1.0004385128693996e-06, "loss": 0.3915, "step": 35009 }, { "epoch": 6.67, "learning_rate": 1.0003813155386082e-06, "loss": 2.4941, "step": 35010 }, { "epoch": 6.67, "learning_rate": 1.000324118207817e-06, "loss": 1.4185, "step": 35011 }, { "epoch": 6.67, "learning_rate": 1.0002669208770258e-06, "loss": 2.3697, "step": 35012 }, { "epoch": 6.67, "learning_rate": 1.0002097235462344e-06, "loss": 1.1603, "step": 35013 }, { "epoch": 6.67, "learning_rate": 1.0001525262154434e-06, "loss": 1.695, "step": 35014 }, { "epoch": 6.67, "learning_rate": 1.0000953288846522e-06, "loss": 0.7269, "step": 35015 }, { "epoch": 6.67, "learning_rate": 1.000038131553861e-06, "loss": 0.7571, "step": 35016 }, { "epoch": 6.67, "learning_rate": 9.999809342230696e-07, "loss": 1.0468, "step": 35017 }, { "epoch": 6.67, "learning_rate": 9.999237368922783e-07, "loss": 2.4237, "step": 35018 }, { "epoch": 6.67, "learning_rate": 9.998665395614871e-07, "loss": 1.8943, "step": 35019 }, { "epoch": 6.67, "learning_rate": 9.99809342230696e-07, "loss": 1.6055, "step": 35020 }, { "epoch": 6.67, "learning_rate": 9.997521448999047e-07, "loss": 1.3175, "step": 35021 }, { "epoch": 6.67, "learning_rate": 9.996949475691135e-07, "loss": 2.4584, "step": 35022 }, { "epoch": 6.67, "learning_rate": 9.99637750238322e-07, "loss": 1.4039, "step": 35023 }, { "epoch": 6.67, "learning_rate": 9.99580552907531e-07, "loss": 0.9306, "step": 35024 }, { "epoch": 6.67, "learning_rate": 9.995233555767397e-07, "loss": 1.2803, "step": 35025 }, { "epoch": 6.67, "learning_rate": 9.994661582459487e-07, "loss": 1.1704, "step": 35026 }, { "epoch": 6.67, "learning_rate": 9.994089609151573e-07, "loss": 0.8915, "step": 35027 }, { "epoch": 6.67, "learning_rate": 9.99351763584366e-07, "loss": 1.6191, "step": 35028 }, { "epoch": 6.67, "learning_rate": 9.992945662535749e-07, "loss": 0.7947, "step": 35029 }, { "epoch": 6.67, "learning_rate": 9.992373689227837e-07, "loss": 1.4969, "step": 35030 }, { "epoch": 6.67, "learning_rate": 9.991801715919925e-07, "loss": 1.8885, "step": 35031 }, { "epoch": 6.67, "learning_rate": 9.991229742612012e-07, "loss": 1.1516, "step": 35032 }, { "epoch": 6.67, "learning_rate": 9.9906577693041e-07, "loss": 1.6574, "step": 35033 }, { "epoch": 6.67, "learning_rate": 9.990085795996186e-07, "loss": 1.2054, "step": 35034 }, { "epoch": 6.67, "learning_rate": 9.989513822688274e-07, "loss": 2.8809, "step": 35035 }, { "epoch": 6.67, "learning_rate": 9.988941849380362e-07, "loss": 1.8672, "step": 35036 }, { "epoch": 6.67, "learning_rate": 9.98836987607245e-07, "loss": 3.2402, "step": 35037 }, { "epoch": 6.67, "learning_rate": 9.987797902764538e-07, "loss": 1.7944, "step": 35038 }, { "epoch": 6.67, "learning_rate": 9.987225929456626e-07, "loss": 0.7864, "step": 35039 }, { "epoch": 6.67, "learning_rate": 9.986653956148714e-07, "loss": 1.7404, "step": 35040 }, { "epoch": 6.67, "learning_rate": 9.9860819828408e-07, "loss": 2.8455, "step": 35041 }, { "epoch": 6.67, "learning_rate": 9.98551000953289e-07, "loss": 1.9765, "step": 35042 }, { "epoch": 6.67, "learning_rate": 9.984938036224978e-07, "loss": 1.6578, "step": 35043 }, { "epoch": 6.68, "learning_rate": 9.984366062917064e-07, "loss": 1.2208, "step": 35044 }, { "epoch": 6.68, "learning_rate": 9.983794089609151e-07, "loss": 2.9089, "step": 35045 }, { "epoch": 6.68, "learning_rate": 9.98322211630124e-07, "loss": 1.233, "step": 35046 }, { "epoch": 6.68, "learning_rate": 9.982650142993327e-07, "loss": 1.0972, "step": 35047 }, { "epoch": 6.68, "learning_rate": 9.982078169685415e-07, "loss": 1.1006, "step": 35048 }, { "epoch": 6.68, "learning_rate": 9.981506196377503e-07, "loss": 1.2859, "step": 35049 }, { "epoch": 6.68, "learning_rate": 9.980934223069591e-07, "loss": 2.2087, "step": 35050 }, { "epoch": 6.68, "learning_rate": 9.980362249761677e-07, "loss": 2.2541, "step": 35051 }, { "epoch": 6.68, "learning_rate": 9.979790276453765e-07, "loss": 3.2459, "step": 35052 }, { "epoch": 6.68, "learning_rate": 9.979218303145855e-07, "loss": 0.4559, "step": 35053 }, { "epoch": 6.68, "learning_rate": 9.97864632983794e-07, "loss": 1.9209, "step": 35054 }, { "epoch": 6.68, "learning_rate": 9.978074356530029e-07, "loss": 0.9136, "step": 35055 }, { "epoch": 6.68, "learning_rate": 9.977502383222117e-07, "loss": 1.3251, "step": 35056 }, { "epoch": 6.68, "learning_rate": 9.976930409914205e-07, "loss": 1.5596, "step": 35057 }, { "epoch": 6.68, "learning_rate": 9.976358436606293e-07, "loss": 2.4631, "step": 35058 }, { "epoch": 6.68, "learning_rate": 9.97578646329838e-07, "loss": 1.8379, "step": 35059 }, { "epoch": 6.68, "learning_rate": 9.975214489990468e-07, "loss": 1.8713, "step": 35060 }, { "epoch": 6.68, "learning_rate": 9.974642516682554e-07, "loss": 1.0047, "step": 35061 }, { "epoch": 6.68, "learning_rate": 9.974070543374642e-07, "loss": 1.2737, "step": 35062 }, { "epoch": 6.68, "learning_rate": 9.97349857006673e-07, "loss": 1.3645, "step": 35063 }, { "epoch": 6.68, "learning_rate": 9.972926596758818e-07, "loss": 1.2044, "step": 35064 }, { "epoch": 6.68, "learning_rate": 9.972354623450906e-07, "loss": 1.2171, "step": 35065 }, { "epoch": 6.68, "learning_rate": 9.971782650142994e-07, "loss": 1.0258, "step": 35066 }, { "epoch": 6.68, "learning_rate": 9.971210676835082e-07, "loss": 2.1109, "step": 35067 }, { "epoch": 6.68, "learning_rate": 9.970638703527168e-07, "loss": 2.0195, "step": 35068 }, { "epoch": 6.68, "learning_rate": 9.970066730219256e-07, "loss": 1.4774, "step": 35069 }, { "epoch": 6.68, "learning_rate": 9.969494756911346e-07, "loss": 1.1387, "step": 35070 }, { "epoch": 6.68, "learning_rate": 9.968922783603431e-07, "loss": 1.7812, "step": 35071 }, { "epoch": 6.68, "learning_rate": 9.96835081029552e-07, "loss": 0.8633, "step": 35072 }, { "epoch": 6.68, "learning_rate": 9.967778836987607e-07, "loss": 2.0739, "step": 35073 }, { "epoch": 6.68, "learning_rate": 9.967206863679695e-07, "loss": 0.3964, "step": 35074 }, { "epoch": 6.68, "learning_rate": 9.966634890371783e-07, "loss": 1.5912, "step": 35075 }, { "epoch": 6.68, "learning_rate": 9.966062917063871e-07, "loss": 0.6598, "step": 35076 }, { "epoch": 6.68, "learning_rate": 9.96549094375596e-07, "loss": 1.3547, "step": 35077 }, { "epoch": 6.68, "learning_rate": 9.964918970448045e-07, "loss": 3.2072, "step": 35078 }, { "epoch": 6.68, "learning_rate": 9.964346997140133e-07, "loss": 0.7407, "step": 35079 }, { "epoch": 6.68, "learning_rate": 9.96377502383222e-07, "loss": 2.2238, "step": 35080 }, { "epoch": 6.68, "learning_rate": 9.96320305052431e-07, "loss": 2.2127, "step": 35081 }, { "epoch": 6.68, "learning_rate": 9.962631077216397e-07, "loss": 1.0878, "step": 35082 }, { "epoch": 6.68, "learning_rate": 9.962059103908485e-07, "loss": 1.9054, "step": 35083 }, { "epoch": 6.68, "learning_rate": 9.961487130600573e-07, "loss": 0.4955, "step": 35084 }, { "epoch": 6.68, "learning_rate": 9.960915157292658e-07, "loss": 1.4273, "step": 35085 }, { "epoch": 6.68, "learning_rate": 9.960343183984748e-07, "loss": 1.591, "step": 35086 }, { "epoch": 6.68, "learning_rate": 9.959771210676836e-07, "loss": 0.774, "step": 35087 }, { "epoch": 6.68, "learning_rate": 9.959199237368922e-07, "loss": 2.333, "step": 35088 }, { "epoch": 6.68, "learning_rate": 9.95862726406101e-07, "loss": 1.9091, "step": 35089 }, { "epoch": 6.68, "learning_rate": 9.958055290753098e-07, "loss": 1.5448, "step": 35090 }, { "epoch": 6.68, "learning_rate": 9.957483317445186e-07, "loss": 2.7105, "step": 35091 }, { "epoch": 6.68, "learning_rate": 9.956911344137274e-07, "loss": 1.9363, "step": 35092 }, { "epoch": 6.68, "learning_rate": 9.956339370829362e-07, "loss": 1.3166, "step": 35093 }, { "epoch": 6.68, "learning_rate": 9.95576739752145e-07, "loss": 1.378, "step": 35094 }, { "epoch": 6.68, "learning_rate": 9.955195424213536e-07, "loss": 1.9877, "step": 35095 }, { "epoch": 6.68, "learning_rate": 9.954623450905624e-07, "loss": 1.9014, "step": 35096 }, { "epoch": 6.69, "learning_rate": 9.954051477597714e-07, "loss": 2.85, "step": 35097 }, { "epoch": 6.69, "learning_rate": 9.953479504289802e-07, "loss": 2.2759, "step": 35098 }, { "epoch": 6.69, "learning_rate": 9.952907530981887e-07, "loss": 0.7916, "step": 35099 }, { "epoch": 6.69, "learning_rate": 9.952335557673975e-07, "loss": 0.5194, "step": 35100 }, { "epoch": 6.69, "learning_rate": 9.951763584366063e-07, "loss": 1.7708, "step": 35101 }, { "epoch": 6.69, "learning_rate": 9.951191611058151e-07, "loss": 1.8922, "step": 35102 }, { "epoch": 6.69, "learning_rate": 9.95061963775024e-07, "loss": 1.1095, "step": 35103 }, { "epoch": 6.69, "learning_rate": 9.950047664442327e-07, "loss": 4.1046, "step": 35104 }, { "epoch": 6.69, "learning_rate": 9.949475691134415e-07, "loss": 1.1781, "step": 35105 }, { "epoch": 6.69, "learning_rate": 9.9489037178265e-07, "loss": 1.2175, "step": 35106 }, { "epoch": 6.69, "learning_rate": 9.948331744518589e-07, "loss": 1.4211, "step": 35107 }, { "epoch": 6.69, "learning_rate": 9.947759771210677e-07, "loss": 1.4706, "step": 35108 }, { "epoch": 6.69, "learning_rate": 9.947187797902765e-07, "loss": 1.8694, "step": 35109 }, { "epoch": 6.69, "learning_rate": 9.946615824594853e-07, "loss": 0.4778, "step": 35110 }, { "epoch": 6.69, "learning_rate": 9.94604385128694e-07, "loss": 1.9908, "step": 35111 }, { "epoch": 6.69, "learning_rate": 9.945471877979026e-07, "loss": 0.8825, "step": 35112 }, { "epoch": 6.69, "learning_rate": 9.944899904671114e-07, "loss": 2.0398, "step": 35113 }, { "epoch": 6.69, "learning_rate": 9.944327931363204e-07, "loss": 1.0106, "step": 35114 }, { "epoch": 6.69, "learning_rate": 9.943755958055292e-07, "loss": 1.1837, "step": 35115 }, { "epoch": 6.69, "learning_rate": 9.943183984747378e-07, "loss": 1.6402, "step": 35116 }, { "epoch": 6.69, "learning_rate": 9.942612011439466e-07, "loss": 1.4961, "step": 35117 }, { "epoch": 6.69, "learning_rate": 9.942040038131554e-07, "loss": 3.4298, "step": 35118 }, { "epoch": 6.69, "learning_rate": 9.941468064823642e-07, "loss": 1.7375, "step": 35119 }, { "epoch": 6.69, "learning_rate": 9.94089609151573e-07, "loss": 1.4243, "step": 35120 }, { "epoch": 6.69, "learning_rate": 9.940324118207818e-07, "loss": 1.8942, "step": 35121 }, { "epoch": 6.69, "learning_rate": 9.939752144899906e-07, "loss": 2.5863, "step": 35122 }, { "epoch": 6.69, "learning_rate": 9.939180171591992e-07, "loss": 0.6897, "step": 35123 }, { "epoch": 6.69, "learning_rate": 9.93860819828408e-07, "loss": 1.7141, "step": 35124 }, { "epoch": 6.69, "learning_rate": 9.93803622497617e-07, "loss": 1.7008, "step": 35125 }, { "epoch": 6.69, "learning_rate": 9.937464251668255e-07, "loss": 1.1981, "step": 35126 }, { "epoch": 6.69, "learning_rate": 9.936892278360343e-07, "loss": 2.2617, "step": 35127 }, { "epoch": 6.69, "learning_rate": 9.936320305052431e-07, "loss": 2.9886, "step": 35128 }, { "epoch": 6.69, "learning_rate": 9.935748331744517e-07, "loss": 2.367, "step": 35129 }, { "epoch": 6.69, "learning_rate": 9.935176358436607e-07, "loss": 1.6969, "step": 35130 }, { "epoch": 6.69, "learning_rate": 9.934604385128695e-07, "loss": 2.2066, "step": 35131 }, { "epoch": 6.69, "learning_rate": 9.934032411820783e-07, "loss": 1.2898, "step": 35132 }, { "epoch": 6.69, "learning_rate": 9.933460438512869e-07, "loss": 1.8994, "step": 35133 }, { "epoch": 6.69, "learning_rate": 9.932888465204957e-07, "loss": 0.7956, "step": 35134 }, { "epoch": 6.69, "learning_rate": 9.932316491897045e-07, "loss": 3.3389, "step": 35135 }, { "epoch": 6.69, "learning_rate": 9.931744518589133e-07, "loss": 1.8628, "step": 35136 }, { "epoch": 6.69, "learning_rate": 9.93117254528122e-07, "loss": 1.1731, "step": 35137 }, { "epoch": 6.69, "learning_rate": 9.930600571973309e-07, "loss": 1.5852, "step": 35138 }, { "epoch": 6.69, "learning_rate": 9.930028598665396e-07, "loss": 2.3311, "step": 35139 }, { "epoch": 6.69, "learning_rate": 9.929456625357482e-07, "loss": 1.9707, "step": 35140 }, { "epoch": 6.69, "learning_rate": 9.928884652049572e-07, "loss": 2.5481, "step": 35141 }, { "epoch": 6.69, "learning_rate": 9.92831267874166e-07, "loss": 2.6269, "step": 35142 }, { "epoch": 6.69, "learning_rate": 9.927740705433746e-07, "loss": 2.0809, "step": 35143 }, { "epoch": 6.69, "learning_rate": 9.927168732125834e-07, "loss": 1.1827, "step": 35144 }, { "epoch": 6.69, "learning_rate": 9.926596758817922e-07, "loss": 1.8946, "step": 35145 }, { "epoch": 6.69, "learning_rate": 9.92602478551001e-07, "loss": 0.8412, "step": 35146 }, { "epoch": 6.69, "learning_rate": 9.925452812202098e-07, "loss": 2.3453, "step": 35147 }, { "epoch": 6.69, "learning_rate": 9.924880838894186e-07, "loss": 2.4156, "step": 35148 }, { "epoch": 6.7, "learning_rate": 9.924308865586274e-07, "loss": 1.4709, "step": 35149 }, { "epoch": 6.7, "learning_rate": 9.92373689227836e-07, "loss": 1.7741, "step": 35150 }, { "epoch": 6.7, "learning_rate": 9.923164918970447e-07, "loss": 0.9417, "step": 35151 }, { "epoch": 6.7, "learning_rate": 9.922592945662535e-07, "loss": 0.9081, "step": 35152 }, { "epoch": 6.7, "learning_rate": 9.922020972354623e-07, "loss": 2.1301, "step": 35153 }, { "epoch": 6.7, "learning_rate": 9.921448999046711e-07, "loss": 1.8615, "step": 35154 }, { "epoch": 6.7, "learning_rate": 9.9208770257388e-07, "loss": 1.2318, "step": 35155 }, { "epoch": 6.7, "learning_rate": 9.920305052430887e-07, "loss": 2.0668, "step": 35156 }, { "epoch": 6.7, "learning_rate": 9.919733079122973e-07, "loss": 0.9204, "step": 35157 }, { "epoch": 6.7, "learning_rate": 9.919161105815063e-07, "loss": 1.7079, "step": 35158 }, { "epoch": 6.7, "learning_rate": 9.91858913250715e-07, "loss": 2.1506, "step": 35159 }, { "epoch": 6.7, "learning_rate": 9.918017159199237e-07, "loss": 2.2857, "step": 35160 }, { "epoch": 6.7, "learning_rate": 9.917445185891325e-07, "loss": 1.7726, "step": 35161 }, { "epoch": 6.7, "learning_rate": 9.916873212583413e-07, "loss": 2.1199, "step": 35162 }, { "epoch": 6.7, "learning_rate": 9.9163012392755e-07, "loss": 1.5495, "step": 35163 }, { "epoch": 6.7, "learning_rate": 9.915729265967589e-07, "loss": 1.388, "step": 35164 }, { "epoch": 6.7, "learning_rate": 9.915157292659676e-07, "loss": 1.8595, "step": 35165 }, { "epoch": 6.7, "learning_rate": 9.914585319351764e-07, "loss": 1.599, "step": 35166 }, { "epoch": 6.7, "learning_rate": 9.91401334604385e-07, "loss": 1.5349, "step": 35167 }, { "epoch": 6.7, "learning_rate": 9.913441372735938e-07, "loss": 1.5346, "step": 35168 }, { "epoch": 6.7, "learning_rate": 9.912869399428028e-07, "loss": 1.5413, "step": 35169 }, { "epoch": 6.7, "learning_rate": 9.912297426120116e-07, "loss": 1.7533, "step": 35170 }, { "epoch": 6.7, "learning_rate": 9.911725452812202e-07, "loss": 3.534, "step": 35171 }, { "epoch": 6.7, "learning_rate": 9.91115347950429e-07, "loss": 1.3363, "step": 35172 }, { "epoch": 6.7, "learning_rate": 9.910581506196378e-07, "loss": 1.736, "step": 35173 }, { "epoch": 6.7, "learning_rate": 9.910009532888466e-07, "loss": 2.2487, "step": 35174 }, { "epoch": 6.7, "learning_rate": 9.909437559580554e-07, "loss": 1.1223, "step": 35175 }, { "epoch": 6.7, "learning_rate": 9.908865586272642e-07, "loss": 2.428, "step": 35176 }, { "epoch": 6.7, "learning_rate": 9.908293612964728e-07, "loss": 0.8342, "step": 35177 }, { "epoch": 6.7, "learning_rate": 9.907721639656815e-07, "loss": 0.2148, "step": 35178 }, { "epoch": 6.7, "learning_rate": 9.907149666348903e-07, "loss": 0.7809, "step": 35179 }, { "epoch": 6.7, "learning_rate": 9.906577693040991e-07, "loss": 2.6267, "step": 35180 }, { "epoch": 6.7, "learning_rate": 9.90600571973308e-07, "loss": 1.7428, "step": 35181 }, { "epoch": 6.7, "learning_rate": 9.905433746425167e-07, "loss": 0.5974, "step": 35182 }, { "epoch": 6.7, "learning_rate": 9.904861773117255e-07, "loss": 1.8413, "step": 35183 }, { "epoch": 6.7, "learning_rate": 9.90428979980934e-07, "loss": 1.9517, "step": 35184 }, { "epoch": 6.7, "learning_rate": 9.90371782650143e-07, "loss": 1.6921, "step": 35185 }, { "epoch": 6.7, "learning_rate": 9.90314585319352e-07, "loss": 2.4295, "step": 35186 }, { "epoch": 6.7, "learning_rate": 9.902573879885607e-07, "loss": 2.0389, "step": 35187 }, { "epoch": 6.7, "learning_rate": 9.902001906577693e-07, "loss": 0.3762, "step": 35188 }, { "epoch": 6.7, "learning_rate": 9.90142993326978e-07, "loss": 2.9726, "step": 35189 }, { "epoch": 6.7, "learning_rate": 9.900857959961869e-07, "loss": 1.3938, "step": 35190 }, { "epoch": 6.7, "learning_rate": 9.900285986653957e-07, "loss": 1.3509, "step": 35191 }, { "epoch": 6.7, "learning_rate": 9.899714013346044e-07, "loss": 2.472, "step": 35192 }, { "epoch": 6.7, "learning_rate": 9.899142040038132e-07, "loss": 0.9054, "step": 35193 }, { "epoch": 6.7, "learning_rate": 9.898570066730218e-07, "loss": 0.9727, "step": 35194 }, { "epoch": 6.7, "learning_rate": 9.897998093422306e-07, "loss": 1.0538, "step": 35195 }, { "epoch": 6.7, "learning_rate": 9.897426120114394e-07, "loss": 1.8655, "step": 35196 }, { "epoch": 6.7, "learning_rate": 9.896854146806484e-07, "loss": 0.9159, "step": 35197 }, { "epoch": 6.7, "learning_rate": 9.89628217349857e-07, "loss": 2.1777, "step": 35198 }, { "epoch": 6.7, "learning_rate": 9.895710200190658e-07, "loss": 2.3055, "step": 35199 }, { "epoch": 6.7, "learning_rate": 9.895138226882746e-07, "loss": 1.8447, "step": 35200 }, { "epoch": 6.7, "learning_rate": 9.894566253574832e-07, "loss": 0.5685, "step": 35201 }, { "epoch": 6.71, "learning_rate": 9.893994280266922e-07, "loss": 2.337, "step": 35202 }, { "epoch": 6.71, "learning_rate": 9.89342230695901e-07, "loss": 1.7527, "step": 35203 }, { "epoch": 6.71, "learning_rate": 9.892850333651098e-07, "loss": 1.1826, "step": 35204 }, { "epoch": 6.71, "learning_rate": 9.892278360343183e-07, "loss": 2.1057, "step": 35205 }, { "epoch": 6.71, "learning_rate": 9.891706387035271e-07, "loss": 1.9686, "step": 35206 }, { "epoch": 6.71, "learning_rate": 9.89113441372736e-07, "loss": 1.507, "step": 35207 }, { "epoch": 6.71, "learning_rate": 9.890562440419447e-07, "loss": 1.8517, "step": 35208 }, { "epoch": 6.71, "learning_rate": 9.889990467111535e-07, "loss": 1.4885, "step": 35209 }, { "epoch": 6.71, "learning_rate": 9.889418493803623e-07, "loss": 1.4212, "step": 35210 }, { "epoch": 6.71, "learning_rate": 9.888846520495711e-07, "loss": 0.7609, "step": 35211 }, { "epoch": 6.71, "learning_rate": 9.888274547187797e-07, "loss": 1.7536, "step": 35212 }, { "epoch": 6.71, "learning_rate": 9.887702573879887e-07, "loss": 2.0629, "step": 35213 }, { "epoch": 6.71, "learning_rate": 9.887130600571975e-07, "loss": 3.2311, "step": 35214 }, { "epoch": 6.71, "learning_rate": 9.88655862726406e-07, "loss": 0.8048, "step": 35215 }, { "epoch": 6.71, "learning_rate": 9.885986653956149e-07, "loss": 1.8381, "step": 35216 }, { "epoch": 6.71, "learning_rate": 9.885414680648237e-07, "loss": 1.5484, "step": 35217 }, { "epoch": 6.71, "learning_rate": 9.884842707340325e-07, "loss": 2.2117, "step": 35218 }, { "epoch": 6.71, "learning_rate": 9.884270734032412e-07, "loss": 1.8472, "step": 35219 }, { "epoch": 6.71, "learning_rate": 9.8836987607245e-07, "loss": 2.657, "step": 35220 }, { "epoch": 6.71, "learning_rate": 9.883126787416588e-07, "loss": 1.6873, "step": 35221 }, { "epoch": 6.71, "learning_rate": 9.882554814108674e-07, "loss": 2.0001, "step": 35222 }, { "epoch": 6.71, "learning_rate": 9.881982840800762e-07, "loss": 1.693, "step": 35223 }, { "epoch": 6.71, "learning_rate": 9.88141086749285e-07, "loss": 1.6081, "step": 35224 }, { "epoch": 6.71, "learning_rate": 9.880838894184938e-07, "loss": 0.6591, "step": 35225 }, { "epoch": 6.71, "learning_rate": 9.880266920877026e-07, "loss": 0.3886, "step": 35226 }, { "epoch": 6.71, "learning_rate": 9.879694947569114e-07, "loss": 2.0068, "step": 35227 }, { "epoch": 6.71, "learning_rate": 9.879122974261202e-07, "loss": 2.8809, "step": 35228 }, { "epoch": 6.71, "learning_rate": 9.87855100095329e-07, "loss": 2.3535, "step": 35229 }, { "epoch": 6.71, "learning_rate": 9.877979027645378e-07, "loss": 1.7093, "step": 35230 }, { "epoch": 6.71, "learning_rate": 9.877407054337466e-07, "loss": 2.0024, "step": 35231 }, { "epoch": 6.71, "learning_rate": 9.876835081029551e-07, "loss": 0.789, "step": 35232 }, { "epoch": 6.71, "learning_rate": 9.87626310772164e-07, "loss": 1.9542, "step": 35233 }, { "epoch": 6.71, "learning_rate": 9.875691134413727e-07, "loss": 0.9437, "step": 35234 }, { "epoch": 6.71, "learning_rate": 9.875119161105815e-07, "loss": 1.6844, "step": 35235 }, { "epoch": 6.71, "learning_rate": 9.874547187797903e-07, "loss": 1.8275, "step": 35236 }, { "epoch": 6.71, "learning_rate": 9.873975214489991e-07, "loss": 2.9009, "step": 35237 }, { "epoch": 6.71, "learning_rate": 9.87340324118208e-07, "loss": 3.0312, "step": 35238 }, { "epoch": 6.71, "learning_rate": 9.872831267874165e-07, "loss": 1.424, "step": 35239 }, { "epoch": 6.71, "learning_rate": 9.872259294566253e-07, "loss": 0.8974, "step": 35240 }, { "epoch": 6.71, "learning_rate": 9.871687321258343e-07, "loss": 2.0408, "step": 35241 }, { "epoch": 6.71, "learning_rate": 9.871115347950429e-07, "loss": 1.3253, "step": 35242 }, { "epoch": 6.71, "learning_rate": 9.870543374642517e-07, "loss": 3.0105, "step": 35243 }, { "epoch": 6.71, "learning_rate": 9.869971401334605e-07, "loss": 0.8064, "step": 35244 }, { "epoch": 6.71, "learning_rate": 9.869399428026692e-07, "loss": 2.165, "step": 35245 }, { "epoch": 6.71, "learning_rate": 9.86882745471878e-07, "loss": 2.3758, "step": 35246 }, { "epoch": 6.71, "learning_rate": 9.868255481410868e-07, "loss": 2.5426, "step": 35247 }, { "epoch": 6.71, "learning_rate": 9.867683508102956e-07, "loss": 1.5715, "step": 35248 }, { "epoch": 6.71, "learning_rate": 9.867111534795042e-07, "loss": 1.0909, "step": 35249 }, { "epoch": 6.71, "learning_rate": 9.86653956148713e-07, "loss": 2.4145, "step": 35250 }, { "epoch": 6.71, "learning_rate": 9.865967588179218e-07, "loss": 1.3407, "step": 35251 }, { "epoch": 6.71, "learning_rate": 9.865395614871308e-07, "loss": 2.1111, "step": 35252 }, { "epoch": 6.71, "learning_rate": 9.864823641563394e-07, "loss": 0.778, "step": 35253 }, { "epoch": 6.72, "learning_rate": 9.864251668255482e-07, "loss": 1.1042, "step": 35254 }, { "epoch": 6.72, "learning_rate": 9.86367969494757e-07, "loss": 1.2915, "step": 35255 }, { "epoch": 6.72, "learning_rate": 9.863107721639656e-07, "loss": 1.8139, "step": 35256 }, { "epoch": 6.72, "learning_rate": 9.862535748331746e-07, "loss": 2.2542, "step": 35257 }, { "epoch": 6.72, "learning_rate": 9.861963775023834e-07, "loss": 2.8012, "step": 35258 }, { "epoch": 6.72, "learning_rate": 9.86139180171592e-07, "loss": 0.7077, "step": 35259 }, { "epoch": 6.72, "learning_rate": 9.860819828408007e-07, "loss": 1.5168, "step": 35260 }, { "epoch": 6.72, "learning_rate": 9.860247855100095e-07, "loss": 2.148, "step": 35261 }, { "epoch": 6.72, "learning_rate": 9.859675881792183e-07, "loss": 1.724, "step": 35262 }, { "epoch": 6.72, "learning_rate": 9.859103908484271e-07, "loss": 2.9741, "step": 35263 }, { "epoch": 6.72, "learning_rate": 9.85853193517636e-07, "loss": 1.0278, "step": 35264 }, { "epoch": 6.72, "learning_rate": 9.857959961868447e-07, "loss": 1.6907, "step": 35265 }, { "epoch": 6.72, "learning_rate": 9.857387988560533e-07, "loss": 1.1153, "step": 35266 }, { "epoch": 6.72, "learning_rate": 9.85681601525262e-07, "loss": 1.1382, "step": 35267 }, { "epoch": 6.72, "learning_rate": 9.856244041944709e-07, "loss": 1.6348, "step": 35268 }, { "epoch": 6.72, "learning_rate": 9.855672068636799e-07, "loss": 2.148, "step": 35269 }, { "epoch": 6.72, "learning_rate": 9.855100095328885e-07, "loss": 1.0773, "step": 35270 }, { "epoch": 6.72, "learning_rate": 9.854528122020973e-07, "loss": 1.1047, "step": 35271 }, { "epoch": 6.72, "learning_rate": 9.85395614871306e-07, "loss": 0.5867, "step": 35272 }, { "epoch": 6.72, "learning_rate": 9.853384175405148e-07, "loss": 1.6772, "step": 35273 }, { "epoch": 6.72, "learning_rate": 9.852812202097236e-07, "loss": 1.1309, "step": 35274 }, { "epoch": 6.72, "learning_rate": 9.852240228789324e-07, "loss": 1.5916, "step": 35275 }, { "epoch": 6.72, "learning_rate": 9.851668255481412e-07, "loss": 1.9499, "step": 35276 }, { "epoch": 6.72, "learning_rate": 9.851096282173498e-07, "loss": 1.0229, "step": 35277 }, { "epoch": 6.72, "learning_rate": 9.850524308865586e-07, "loss": 1.2838, "step": 35278 }, { "epoch": 6.72, "learning_rate": 9.849952335557674e-07, "loss": 1.3, "step": 35279 }, { "epoch": 6.72, "learning_rate": 9.849380362249762e-07, "loss": 0.8729, "step": 35280 }, { "epoch": 6.72, "learning_rate": 9.84880838894185e-07, "loss": 2.2872, "step": 35281 }, { "epoch": 6.72, "learning_rate": 9.848236415633938e-07, "loss": 1.5617, "step": 35282 }, { "epoch": 6.72, "learning_rate": 9.847664442326024e-07, "loss": 2.9621, "step": 35283 }, { "epoch": 6.72, "learning_rate": 9.847092469018111e-07, "loss": 1.7999, "step": 35284 }, { "epoch": 6.72, "learning_rate": 9.846520495710202e-07, "loss": 1.5619, "step": 35285 }, { "epoch": 6.72, "learning_rate": 9.84594852240229e-07, "loss": 1.0448, "step": 35286 }, { "epoch": 6.72, "learning_rate": 9.845376549094375e-07, "loss": 1.6063, "step": 35287 }, { "epoch": 6.72, "learning_rate": 9.844804575786463e-07, "loss": 2.2365, "step": 35288 }, { "epoch": 6.72, "learning_rate": 9.844232602478551e-07, "loss": 1.359, "step": 35289 }, { "epoch": 6.72, "learning_rate": 9.84366062917064e-07, "loss": 1.5622, "step": 35290 }, { "epoch": 6.72, "learning_rate": 9.843088655862727e-07, "loss": 2.4072, "step": 35291 }, { "epoch": 6.72, "learning_rate": 9.842516682554815e-07, "loss": 0.7554, "step": 35292 }, { "epoch": 6.72, "learning_rate": 9.841944709246903e-07, "loss": 1.6716, "step": 35293 }, { "epoch": 6.72, "learning_rate": 9.841372735938989e-07, "loss": 1.2131, "step": 35294 }, { "epoch": 6.72, "learning_rate": 9.840800762631077e-07, "loss": 3.5395, "step": 35295 }, { "epoch": 6.72, "learning_rate": 9.840228789323167e-07, "loss": 2.5444, "step": 35296 }, { "epoch": 6.72, "learning_rate": 9.839656816015253e-07, "loss": 1.5643, "step": 35297 }, { "epoch": 6.72, "learning_rate": 9.83908484270734e-07, "loss": 0.9708, "step": 35298 }, { "epoch": 6.72, "learning_rate": 9.838512869399428e-07, "loss": 2.4714, "step": 35299 }, { "epoch": 6.72, "learning_rate": 9.837940896091514e-07, "loss": 2.2064, "step": 35300 }, { "epoch": 6.72, "learning_rate": 9.837368922783604e-07, "loss": 1.4413, "step": 35301 }, { "epoch": 6.72, "learning_rate": 9.836796949475692e-07, "loss": 1.3724, "step": 35302 }, { "epoch": 6.72, "learning_rate": 9.83622497616778e-07, "loss": 2.1738, "step": 35303 }, { "epoch": 6.72, "learning_rate": 9.835653002859866e-07, "loss": 1.9136, "step": 35304 }, { "epoch": 6.72, "learning_rate": 9.835081029551954e-07, "loss": 1.1392, "step": 35305 }, { "epoch": 6.72, "learning_rate": 9.834509056244042e-07, "loss": 0.8418, "step": 35306 }, { "epoch": 6.73, "learning_rate": 9.83393708293613e-07, "loss": 1.077, "step": 35307 }, { "epoch": 6.73, "learning_rate": 9.833365109628218e-07, "loss": 1.2452, "step": 35308 }, { "epoch": 6.73, "learning_rate": 9.832793136320306e-07, "loss": 2.6449, "step": 35309 }, { "epoch": 6.73, "learning_rate": 9.832221163012394e-07, "loss": 1.4626, "step": 35310 }, { "epoch": 6.73, "learning_rate": 9.83164918970448e-07, "loss": 1.9211, "step": 35311 }, { "epoch": 6.73, "learning_rate": 9.831077216396567e-07, "loss": 0.803, "step": 35312 }, { "epoch": 6.73, "learning_rate": 9.830505243088657e-07, "loss": 2.5297, "step": 35313 }, { "epoch": 6.73, "learning_rate": 9.829933269780743e-07, "loss": 1.8129, "step": 35314 }, { "epoch": 6.73, "learning_rate": 9.829361296472831e-07, "loss": 0.7681, "step": 35315 }, { "epoch": 6.73, "learning_rate": 9.82878932316492e-07, "loss": 0.6249, "step": 35316 }, { "epoch": 6.73, "learning_rate": 9.828217349857007e-07, "loss": 2.3141, "step": 35317 }, { "epoch": 6.73, "learning_rate": 9.827645376549095e-07, "loss": 1.6659, "step": 35318 }, { "epoch": 6.73, "learning_rate": 9.827073403241183e-07, "loss": 0.6274, "step": 35319 }, { "epoch": 6.73, "learning_rate": 9.82650142993327e-07, "loss": 1.9737, "step": 35320 }, { "epoch": 6.73, "learning_rate": 9.825929456625357e-07, "loss": 1.2581, "step": 35321 }, { "epoch": 6.73, "learning_rate": 9.825357483317445e-07, "loss": 1.693, "step": 35322 }, { "epoch": 6.73, "learning_rate": 9.824785510009533e-07, "loss": 1.6043, "step": 35323 }, { "epoch": 6.73, "learning_rate": 9.82421353670162e-07, "loss": 0.7737, "step": 35324 }, { "epoch": 6.73, "learning_rate": 9.823641563393708e-07, "loss": 1.5809, "step": 35325 }, { "epoch": 6.73, "learning_rate": 9.823069590085796e-07, "loss": 2.4112, "step": 35326 }, { "epoch": 6.73, "learning_rate": 9.822497616777884e-07, "loss": 2.5081, "step": 35327 }, { "epoch": 6.73, "learning_rate": 9.82192564346997e-07, "loss": 0.6108, "step": 35328 }, { "epoch": 6.73, "learning_rate": 9.82135367016206e-07, "loss": 2.9618, "step": 35329 }, { "epoch": 6.73, "learning_rate": 9.820781696854148e-07, "loss": 0.9671, "step": 35330 }, { "epoch": 6.73, "learning_rate": 9.820209723546234e-07, "loss": 2.4438, "step": 35331 }, { "epoch": 6.73, "learning_rate": 9.819637750238322e-07, "loss": 1.5977, "step": 35332 }, { "epoch": 6.73, "learning_rate": 9.81906577693041e-07, "loss": 2.2248, "step": 35333 }, { "epoch": 6.73, "learning_rate": 9.818493803622498e-07, "loss": 1.6245, "step": 35334 }, { "epoch": 6.73, "learning_rate": 9.817921830314586e-07, "loss": 0.675, "step": 35335 }, { "epoch": 6.73, "learning_rate": 9.817349857006674e-07, "loss": 2.6069, "step": 35336 }, { "epoch": 6.73, "learning_rate": 9.816777883698762e-07, "loss": 1.3324, "step": 35337 }, { "epoch": 6.73, "learning_rate": 9.816205910390847e-07, "loss": 1.3154, "step": 35338 }, { "epoch": 6.73, "learning_rate": 9.815633937082935e-07, "loss": 3.1633, "step": 35339 }, { "epoch": 6.73, "learning_rate": 9.815061963775025e-07, "loss": 2.1566, "step": 35340 }, { "epoch": 6.73, "learning_rate": 9.814489990467113e-07, "loss": 1.4881, "step": 35341 }, { "epoch": 6.73, "learning_rate": 9.8139180171592e-07, "loss": 3.0871, "step": 35342 }, { "epoch": 6.73, "learning_rate": 9.813346043851287e-07, "loss": 1.9981, "step": 35343 }, { "epoch": 6.73, "learning_rate": 9.812774070543375e-07, "loss": 2.5699, "step": 35344 }, { "epoch": 6.73, "learning_rate": 9.812202097235463e-07, "loss": 0.7569, "step": 35345 }, { "epoch": 6.73, "learning_rate": 9.81163012392755e-07, "loss": 1.419, "step": 35346 }, { "epoch": 6.73, "learning_rate": 9.811058150619639e-07, "loss": 0.6206, "step": 35347 }, { "epoch": 6.73, "learning_rate": 9.810486177311725e-07, "loss": 1.6289, "step": 35348 }, { "epoch": 6.73, "learning_rate": 9.809914204003813e-07, "loss": 2.7874, "step": 35349 }, { "epoch": 6.73, "learning_rate": 9.8093422306959e-07, "loss": 2.1504, "step": 35350 }, { "epoch": 6.73, "learning_rate": 9.808770257387989e-07, "loss": 0.7851, "step": 35351 }, { "epoch": 6.73, "learning_rate": 9.808198284080076e-07, "loss": 0.7023, "step": 35352 }, { "epoch": 6.73, "learning_rate": 9.807626310772164e-07, "loss": 1.8749, "step": 35353 }, { "epoch": 6.73, "learning_rate": 9.807054337464252e-07, "loss": 2.5534, "step": 35354 }, { "epoch": 6.73, "learning_rate": 9.806482364156338e-07, "loss": 1.7999, "step": 35355 }, { "epoch": 6.73, "learning_rate": 9.805910390848426e-07, "loss": 1.7209, "step": 35356 }, { "epoch": 6.73, "learning_rate": 9.805338417540516e-07, "loss": 1.1899, "step": 35357 }, { "epoch": 6.73, "learning_rate": 9.804766444232604e-07, "loss": 1.1456, "step": 35358 }, { "epoch": 6.74, "learning_rate": 9.80419447092469e-07, "loss": 3.0597, "step": 35359 }, { "epoch": 6.74, "learning_rate": 9.803622497616778e-07, "loss": 0.9382, "step": 35360 }, { "epoch": 6.74, "learning_rate": 9.803050524308866e-07, "loss": 2.4401, "step": 35361 }, { "epoch": 6.74, "learning_rate": 9.802478551000954e-07, "loss": 1.5221, "step": 35362 }, { "epoch": 6.74, "learning_rate": 9.801906577693042e-07, "loss": 1.439, "step": 35363 }, { "epoch": 6.74, "learning_rate": 9.80133460438513e-07, "loss": 1.4311, "step": 35364 }, { "epoch": 6.74, "learning_rate": 9.800762631077215e-07, "loss": 1.0392, "step": 35365 }, { "epoch": 6.74, "learning_rate": 9.800190657769303e-07, "loss": 1.7072, "step": 35366 }, { "epoch": 6.74, "learning_rate": 9.799618684461391e-07, "loss": 1.7198, "step": 35367 }, { "epoch": 6.74, "learning_rate": 9.799046711153481e-07, "loss": 1.6496, "step": 35368 }, { "epoch": 6.74, "learning_rate": 9.798474737845567e-07, "loss": 1.9965, "step": 35369 }, { "epoch": 6.74, "learning_rate": 9.797902764537655e-07, "loss": 2.3736, "step": 35370 }, { "epoch": 6.74, "learning_rate": 9.797330791229743e-07, "loss": 1.761, "step": 35371 }, { "epoch": 6.74, "learning_rate": 9.796758817921829e-07, "loss": 2.8448, "step": 35372 }, { "epoch": 6.74, "learning_rate": 9.796186844613919e-07, "loss": 0.4925, "step": 35373 }, { "epoch": 6.74, "learning_rate": 9.795614871306007e-07, "loss": 2.2083, "step": 35374 }, { "epoch": 6.74, "learning_rate": 9.795042897998095e-07, "loss": 2.0648, "step": 35375 }, { "epoch": 6.74, "learning_rate": 9.79447092469018e-07, "loss": 2.2258, "step": 35376 }, { "epoch": 6.74, "learning_rate": 9.793898951382269e-07, "loss": 0.4457, "step": 35377 }, { "epoch": 6.74, "learning_rate": 9.793326978074357e-07, "loss": 1.5386, "step": 35378 }, { "epoch": 6.74, "learning_rate": 9.792755004766444e-07, "loss": 1.7615, "step": 35379 }, { "epoch": 6.74, "learning_rate": 9.792183031458532e-07, "loss": 1.2346, "step": 35380 }, { "epoch": 6.74, "learning_rate": 9.79161105815062e-07, "loss": 0.6641, "step": 35381 }, { "epoch": 6.74, "learning_rate": 9.791039084842708e-07, "loss": 2.1183, "step": 35382 }, { "epoch": 6.74, "learning_rate": 9.790467111534794e-07, "loss": 2.8322, "step": 35383 }, { "epoch": 6.74, "learning_rate": 9.789895138226884e-07, "loss": 1.9964, "step": 35384 }, { "epoch": 6.74, "learning_rate": 9.789323164918972e-07, "loss": 1.2565, "step": 35385 }, { "epoch": 6.74, "learning_rate": 9.788751191611058e-07, "loss": 0.3869, "step": 35386 }, { "epoch": 6.74, "learning_rate": 9.788179218303146e-07, "loss": 2.4414, "step": 35387 }, { "epoch": 6.74, "learning_rate": 9.787607244995234e-07, "loss": 1.2034, "step": 35388 }, { "epoch": 6.74, "learning_rate": 9.787035271687322e-07, "loss": 2.2365, "step": 35389 }, { "epoch": 6.74, "learning_rate": 9.78646329837941e-07, "loss": 1.0859, "step": 35390 }, { "epoch": 6.74, "learning_rate": 9.785891325071498e-07, "loss": 2.1048, "step": 35391 }, { "epoch": 6.74, "learning_rate": 9.785319351763586e-07, "loss": 1.5746, "step": 35392 }, { "epoch": 6.74, "learning_rate": 9.784747378455671e-07, "loss": 0.5273, "step": 35393 }, { "epoch": 6.74, "learning_rate": 9.78417540514776e-07, "loss": 2.1306, "step": 35394 }, { "epoch": 6.74, "learning_rate": 9.783603431839847e-07, "loss": 2.7118, "step": 35395 }, { "epoch": 6.74, "learning_rate": 9.783031458531935e-07, "loss": 0.8827, "step": 35396 }, { "epoch": 6.74, "learning_rate": 9.782459485224023e-07, "loss": 2.1443, "step": 35397 }, { "epoch": 6.74, "learning_rate": 9.78188751191611e-07, "loss": 2.8453, "step": 35398 }, { "epoch": 6.74, "learning_rate": 9.7813155386082e-07, "loss": 1.9057, "step": 35399 }, { "epoch": 6.74, "learning_rate": 9.780743565300285e-07, "loss": 2.2337, "step": 35400 }, { "epoch": 6.74, "learning_rate": 9.780171591992375e-07, "loss": 2.8153, "step": 35401 }, { "epoch": 6.74, "learning_rate": 9.779599618684463e-07, "loss": 1.6398, "step": 35402 }, { "epoch": 6.74, "learning_rate": 9.779027645376549e-07, "loss": 1.6355, "step": 35403 }, { "epoch": 6.74, "learning_rate": 9.778455672068637e-07, "loss": 2.204, "step": 35404 }, { "epoch": 6.74, "learning_rate": 9.777883698760724e-07, "loss": 2.154, "step": 35405 }, { "epoch": 6.74, "learning_rate": 9.777311725452812e-07, "loss": 1.8397, "step": 35406 }, { "epoch": 6.74, "learning_rate": 9.7767397521449e-07, "loss": 0.7665, "step": 35407 }, { "epoch": 6.74, "learning_rate": 9.776167778836988e-07, "loss": 2.6625, "step": 35408 }, { "epoch": 6.74, "learning_rate": 9.775595805529076e-07, "loss": 1.551, "step": 35409 }, { "epoch": 6.74, "learning_rate": 9.775023832221162e-07, "loss": 1.3378, "step": 35410 }, { "epoch": 6.74, "learning_rate": 9.77445185891325e-07, "loss": 2.1245, "step": 35411 }, { "epoch": 6.75, "learning_rate": 9.77387988560534e-07, "loss": 1.6907, "step": 35412 }, { "epoch": 6.75, "learning_rate": 9.773307912297426e-07, "loss": 0.862, "step": 35413 }, { "epoch": 6.75, "learning_rate": 9.772735938989514e-07, "loss": 1.5809, "step": 35414 }, { "epoch": 6.75, "learning_rate": 9.772163965681602e-07, "loss": 1.9209, "step": 35415 }, { "epoch": 6.75, "learning_rate": 9.77159199237369e-07, "loss": 0.3525, "step": 35416 }, { "epoch": 6.75, "learning_rate": 9.771020019065778e-07, "loss": 1.1999, "step": 35417 }, { "epoch": 6.75, "learning_rate": 9.770448045757866e-07, "loss": 2.2706, "step": 35418 }, { "epoch": 6.75, "learning_rate": 9.769876072449954e-07, "loss": 1.6511, "step": 35419 }, { "epoch": 6.75, "learning_rate": 9.76930409914204e-07, "loss": 2.6677, "step": 35420 }, { "epoch": 6.75, "learning_rate": 9.768732125834127e-07, "loss": 0.6446, "step": 35421 }, { "epoch": 6.75, "learning_rate": 9.768160152526215e-07, "loss": 1.7495, "step": 35422 }, { "epoch": 6.75, "learning_rate": 9.767588179218303e-07, "loss": 1.3477, "step": 35423 }, { "epoch": 6.75, "learning_rate": 9.767016205910391e-07, "loss": 1.4041, "step": 35424 }, { "epoch": 6.75, "learning_rate": 9.76644423260248e-07, "loss": 1.8306, "step": 35425 }, { "epoch": 6.75, "learning_rate": 9.765872259294567e-07, "loss": 1.6567, "step": 35426 }, { "epoch": 6.75, "learning_rate": 9.765300285986653e-07, "loss": 2.7886, "step": 35427 }, { "epoch": 6.75, "learning_rate": 9.764728312678743e-07, "loss": 1.7726, "step": 35428 }, { "epoch": 6.75, "learning_rate": 9.76415633937083e-07, "loss": 1.5885, "step": 35429 }, { "epoch": 6.75, "learning_rate": 9.763584366062917e-07, "loss": 1.9644, "step": 35430 }, { "epoch": 6.75, "learning_rate": 9.763012392755005e-07, "loss": 0.703, "step": 35431 }, { "epoch": 6.75, "learning_rate": 9.762440419447092e-07, "loss": 0.8853, "step": 35432 }, { "epoch": 6.75, "learning_rate": 9.76186844613918e-07, "loss": 0.856, "step": 35433 }, { "epoch": 6.75, "learning_rate": 9.761296472831268e-07, "loss": 1.7213, "step": 35434 }, { "epoch": 6.75, "learning_rate": 9.760724499523356e-07, "loss": 1.0731, "step": 35435 }, { "epoch": 6.75, "learning_rate": 9.760152526215444e-07, "loss": 1.8077, "step": 35436 }, { "epoch": 6.75, "learning_rate": 9.75958055290753e-07, "loss": 1.9546, "step": 35437 }, { "epoch": 6.75, "learning_rate": 9.759008579599618e-07, "loss": 0.8281, "step": 35438 }, { "epoch": 6.75, "learning_rate": 9.758436606291706e-07, "loss": 2.7986, "step": 35439 }, { "epoch": 6.75, "learning_rate": 9.757864632983796e-07, "loss": 1.6886, "step": 35440 }, { "epoch": 6.75, "learning_rate": 9.757292659675882e-07, "loss": 1.0006, "step": 35441 }, { "epoch": 6.75, "learning_rate": 9.75672068636797e-07, "loss": 1.0668, "step": 35442 }, { "epoch": 6.75, "learning_rate": 9.756148713060058e-07, "loss": 2.4605, "step": 35443 }, { "epoch": 6.75, "learning_rate": 9.755576739752143e-07, "loss": 1.2546, "step": 35444 }, { "epoch": 6.75, "learning_rate": 9.755004766444234e-07, "loss": 2.1654, "step": 35445 }, { "epoch": 6.75, "learning_rate": 9.754432793136321e-07, "loss": 2.3596, "step": 35446 }, { "epoch": 6.75, "learning_rate": 9.75386081982841e-07, "loss": 1.083, "step": 35447 }, { "epoch": 6.75, "learning_rate": 9.753288846520495e-07, "loss": 1.939, "step": 35448 }, { "epoch": 6.75, "learning_rate": 9.752716873212583e-07, "loss": 1.8593, "step": 35449 }, { "epoch": 6.75, "learning_rate": 9.752144899904671e-07, "loss": 2.442, "step": 35450 }, { "epoch": 6.75, "learning_rate": 9.75157292659676e-07, "loss": 1.4597, "step": 35451 }, { "epoch": 6.75, "learning_rate": 9.751000953288847e-07, "loss": 1.6141, "step": 35452 }, { "epoch": 6.75, "learning_rate": 9.750428979980935e-07, "loss": 0.968, "step": 35453 }, { "epoch": 6.75, "learning_rate": 9.74985700667302e-07, "loss": 1.3976, "step": 35454 }, { "epoch": 6.75, "learning_rate": 9.749285033365109e-07, "loss": 1.5442, "step": 35455 }, { "epoch": 6.75, "learning_rate": 9.748713060057199e-07, "loss": 1.6245, "step": 35456 }, { "epoch": 6.75, "learning_rate": 9.748141086749287e-07, "loss": 3.0384, "step": 35457 }, { "epoch": 6.75, "learning_rate": 9.747569113441373e-07, "loss": 1.7292, "step": 35458 }, { "epoch": 6.75, "learning_rate": 9.74699714013346e-07, "loss": 2.3456, "step": 35459 }, { "epoch": 6.75, "learning_rate": 9.746425166825548e-07, "loss": 2.0661, "step": 35460 }, { "epoch": 6.75, "learning_rate": 9.745853193517636e-07, "loss": 0.9317, "step": 35461 }, { "epoch": 6.75, "learning_rate": 9.745281220209724e-07, "loss": 1.9249, "step": 35462 }, { "epoch": 6.75, "learning_rate": 9.744709246901812e-07, "loss": 0.8194, "step": 35463 }, { "epoch": 6.76, "learning_rate": 9.7441372735939e-07, "loss": 2.6953, "step": 35464 }, { "epoch": 6.76, "learning_rate": 9.743565300285986e-07, "loss": 0.6527, "step": 35465 }, { "epoch": 6.76, "learning_rate": 9.742993326978074e-07, "loss": 1.4684, "step": 35466 }, { "epoch": 6.76, "learning_rate": 9.742421353670162e-07, "loss": 1.0378, "step": 35467 }, { "epoch": 6.76, "learning_rate": 9.74184938036225e-07, "loss": 1.5894, "step": 35468 }, { "epoch": 6.76, "learning_rate": 9.741277407054338e-07, "loss": 1.2126, "step": 35469 }, { "epoch": 6.76, "learning_rate": 9.740705433746426e-07, "loss": 2.3287, "step": 35470 }, { "epoch": 6.76, "learning_rate": 9.740133460438514e-07, "loss": 1.5023, "step": 35471 }, { "epoch": 6.76, "learning_rate": 9.739561487130602e-07, "loss": 2.8196, "step": 35472 }, { "epoch": 6.76, "learning_rate": 9.73898951382269e-07, "loss": 1.6991, "step": 35473 }, { "epoch": 6.76, "learning_rate": 9.738417540514777e-07, "loss": 2.5402, "step": 35474 }, { "epoch": 6.76, "learning_rate": 9.737845567206863e-07, "loss": 1.4701, "step": 35475 }, { "epoch": 6.76, "learning_rate": 9.737273593898951e-07, "loss": 1.6527, "step": 35476 }, { "epoch": 6.76, "learning_rate": 9.73670162059104e-07, "loss": 3.4591, "step": 35477 }, { "epoch": 6.76, "learning_rate": 9.736129647283127e-07, "loss": 0.7544, "step": 35478 }, { "epoch": 6.76, "learning_rate": 9.735557673975215e-07, "loss": 1.7431, "step": 35479 }, { "epoch": 6.76, "learning_rate": 9.734985700667303e-07, "loss": 1.2465, "step": 35480 }, { "epoch": 6.76, "learning_rate": 9.73441372735939e-07, "loss": 2.4599, "step": 35481 }, { "epoch": 6.76, "learning_rate": 9.733841754051477e-07, "loss": 2.5685, "step": 35482 }, { "epoch": 6.76, "learning_rate": 9.733269780743565e-07, "loss": 1.3045, "step": 35483 }, { "epoch": 6.76, "learning_rate": 9.732697807435655e-07, "loss": 2.5558, "step": 35484 }, { "epoch": 6.76, "learning_rate": 9.73212583412774e-07, "loss": 3.1055, "step": 35485 }, { "epoch": 6.76, "learning_rate": 9.731553860819828e-07, "loss": 2.5164, "step": 35486 }, { "epoch": 6.76, "learning_rate": 9.730981887511916e-07, "loss": 1.4307, "step": 35487 }, { "epoch": 6.76, "learning_rate": 9.730409914204004e-07, "loss": 2.0136, "step": 35488 }, { "epoch": 6.76, "learning_rate": 9.729837940896092e-07, "loss": 2.9161, "step": 35489 }, { "epoch": 6.76, "learning_rate": 9.72926596758818e-07, "loss": 0.7244, "step": 35490 }, { "epoch": 6.76, "learning_rate": 9.728693994280268e-07, "loss": 1.51, "step": 35491 }, { "epoch": 6.76, "learning_rate": 9.728122020972354e-07, "loss": 1.1791, "step": 35492 }, { "epoch": 6.76, "learning_rate": 9.727550047664442e-07, "loss": 2.0625, "step": 35493 }, { "epoch": 6.76, "learning_rate": 9.72697807435653e-07, "loss": 2.6518, "step": 35494 }, { "epoch": 6.76, "learning_rate": 9.726406101048618e-07, "loss": 1.0419, "step": 35495 }, { "epoch": 6.76, "learning_rate": 9.725834127740706e-07, "loss": 1.5693, "step": 35496 }, { "epoch": 6.76, "learning_rate": 9.725262154432794e-07, "loss": 0.7441, "step": 35497 }, { "epoch": 6.76, "learning_rate": 9.724690181124882e-07, "loss": 1.4919, "step": 35498 }, { "epoch": 6.76, "learning_rate": 9.724118207816967e-07, "loss": 1.3799, "step": 35499 }, { "epoch": 6.76, "learning_rate": 9.723546234509057e-07, "loss": 2.9921, "step": 35500 }, { "epoch": 6.76, "learning_rate": 9.722974261201145e-07, "loss": 1.9031, "step": 35501 }, { "epoch": 6.76, "learning_rate": 9.722402287893231e-07, "loss": 2.2644, "step": 35502 }, { "epoch": 6.76, "learning_rate": 9.72183031458532e-07, "loss": 0.9195, "step": 35503 }, { "epoch": 6.76, "learning_rate": 9.721258341277407e-07, "loss": 1.9447, "step": 35504 }, { "epoch": 6.76, "learning_rate": 9.720686367969495e-07, "loss": 1.1127, "step": 35505 }, { "epoch": 6.76, "learning_rate": 9.720114394661583e-07, "loss": 1.6651, "step": 35506 }, { "epoch": 6.76, "learning_rate": 9.71954242135367e-07, "loss": 1.206, "step": 35507 }, { "epoch": 6.76, "learning_rate": 9.718970448045759e-07, "loss": 1.7627, "step": 35508 }, { "epoch": 6.76, "learning_rate": 9.718398474737845e-07, "loss": 1.2054, "step": 35509 }, { "epoch": 6.76, "learning_rate": 9.717826501429933e-07, "loss": 1.9086, "step": 35510 }, { "epoch": 6.76, "learning_rate": 9.71725452812202e-07, "loss": 0.3622, "step": 35511 }, { "epoch": 6.76, "learning_rate": 9.71668255481411e-07, "loss": 1.6631, "step": 35512 }, { "epoch": 6.76, "learning_rate": 9.716110581506196e-07, "loss": 1.5776, "step": 35513 }, { "epoch": 6.76, "learning_rate": 9.715538608198284e-07, "loss": 0.9585, "step": 35514 }, { "epoch": 6.76, "learning_rate": 9.714966634890372e-07, "loss": 1.4672, "step": 35515 }, { "epoch": 6.76, "learning_rate": 9.714394661582458e-07, "loss": 2.9927, "step": 35516 }, { "epoch": 6.77, "learning_rate": 9.713822688274548e-07, "loss": 3.8006, "step": 35517 }, { "epoch": 6.77, "learning_rate": 9.713250714966636e-07, "loss": 1.1279, "step": 35518 }, { "epoch": 6.77, "learning_rate": 9.712678741658722e-07, "loss": 1.4695, "step": 35519 }, { "epoch": 6.77, "learning_rate": 9.71210676835081e-07, "loss": 1.7303, "step": 35520 }, { "epoch": 6.77, "learning_rate": 9.711534795042898e-07, "loss": 1.0288, "step": 35521 }, { "epoch": 6.77, "learning_rate": 9.710962821734986e-07, "loss": 1.4767, "step": 35522 }, { "epoch": 6.77, "learning_rate": 9.710390848427074e-07, "loss": 0.9344, "step": 35523 }, { "epoch": 6.77, "learning_rate": 9.709818875119162e-07, "loss": 0.7923, "step": 35524 }, { "epoch": 6.77, "learning_rate": 9.70924690181125e-07, "loss": 2.4998, "step": 35525 }, { "epoch": 6.77, "learning_rate": 9.708674928503335e-07, "loss": 0.6066, "step": 35526 }, { "epoch": 6.77, "learning_rate": 9.708102955195423e-07, "loss": 1.0639, "step": 35527 }, { "epoch": 6.77, "learning_rate": 9.707530981887513e-07, "loss": 2.1159, "step": 35528 }, { "epoch": 6.77, "learning_rate": 9.706959008579601e-07, "loss": 2.6313, "step": 35529 }, { "epoch": 6.77, "learning_rate": 9.706387035271687e-07, "loss": 1.5761, "step": 35530 }, { "epoch": 6.77, "learning_rate": 9.705815061963775e-07, "loss": 1.9187, "step": 35531 }, { "epoch": 6.77, "learning_rate": 9.705243088655863e-07, "loss": 1.629, "step": 35532 }, { "epoch": 6.77, "learning_rate": 9.70467111534795e-07, "loss": 2.0976, "step": 35533 }, { "epoch": 6.77, "learning_rate": 9.704099142040039e-07, "loss": 1.2278, "step": 35534 }, { "epoch": 6.77, "learning_rate": 9.703527168732127e-07, "loss": 0.4747, "step": 35535 }, { "epoch": 6.77, "learning_rate": 9.702955195424215e-07, "loss": 2.9626, "step": 35536 }, { "epoch": 6.77, "learning_rate": 9.7023832221163e-07, "loss": 1.145, "step": 35537 }, { "epoch": 6.77, "learning_rate": 9.701811248808388e-07, "loss": 1.4883, "step": 35538 }, { "epoch": 6.77, "learning_rate": 9.701239275500479e-07, "loss": 1.4592, "step": 35539 }, { "epoch": 6.77, "learning_rate": 9.700667302192564e-07, "loss": 0.8306, "step": 35540 }, { "epoch": 6.77, "learning_rate": 9.700095328884652e-07, "loss": 3.3327, "step": 35541 }, { "epoch": 6.77, "learning_rate": 9.69952335557674e-07, "loss": 0.7214, "step": 35542 }, { "epoch": 6.77, "learning_rate": 9.698951382268826e-07, "loss": 2.677, "step": 35543 }, { "epoch": 6.77, "learning_rate": 9.698379408960916e-07, "loss": 1.6978, "step": 35544 }, { "epoch": 6.77, "learning_rate": 9.697807435653004e-07, "loss": 2.3972, "step": 35545 }, { "epoch": 6.77, "learning_rate": 9.697235462345092e-07, "loss": 1.0789, "step": 35546 }, { "epoch": 6.77, "learning_rate": 9.696663489037178e-07, "loss": 0.9242, "step": 35547 }, { "epoch": 6.77, "learning_rate": 9.696091515729266e-07, "loss": 2.6236, "step": 35548 }, { "epoch": 6.77, "learning_rate": 9.695519542421354e-07, "loss": 1.0032, "step": 35549 }, { "epoch": 6.77, "learning_rate": 9.694947569113442e-07, "loss": 1.365, "step": 35550 }, { "epoch": 6.77, "learning_rate": 9.69437559580553e-07, "loss": 1.0127, "step": 35551 }, { "epoch": 6.77, "learning_rate": 9.693803622497618e-07, "loss": 1.7625, "step": 35552 }, { "epoch": 6.77, "learning_rate": 9.693231649189705e-07, "loss": 2.2428, "step": 35553 }, { "epoch": 6.77, "learning_rate": 9.692659675881791e-07, "loss": 2.0261, "step": 35554 }, { "epoch": 6.77, "learning_rate": 9.69208770257388e-07, "loss": 1.3002, "step": 35555 }, { "epoch": 6.77, "learning_rate": 9.69151572926597e-07, "loss": 0.9999, "step": 35556 }, { "epoch": 6.77, "learning_rate": 9.690943755958055e-07, "loss": 1.702, "step": 35557 }, { "epoch": 6.77, "learning_rate": 9.690371782650143e-07, "loss": 2.1466, "step": 35558 }, { "epoch": 6.77, "learning_rate": 9.68979980934223e-07, "loss": 3.0206, "step": 35559 }, { "epoch": 6.77, "learning_rate": 9.689227836034317e-07, "loss": 1.1167, "step": 35560 }, { "epoch": 6.77, "learning_rate": 9.688655862726407e-07, "loss": 2.4666, "step": 35561 }, { "epoch": 6.77, "learning_rate": 9.688083889418495e-07, "loss": 1.549, "step": 35562 }, { "epoch": 6.77, "learning_rate": 9.687511916110583e-07, "loss": 1.0982, "step": 35563 }, { "epoch": 6.77, "learning_rate": 9.686939942802669e-07, "loss": 1.7897, "step": 35564 }, { "epoch": 6.77, "learning_rate": 9.686367969494756e-07, "loss": 1.0782, "step": 35565 }, { "epoch": 6.77, "learning_rate": 9.685795996186844e-07, "loss": 0.4463, "step": 35566 }, { "epoch": 6.77, "learning_rate": 9.685224022878932e-07, "loss": 2.444, "step": 35567 }, { "epoch": 6.77, "learning_rate": 9.68465204957102e-07, "loss": 1.013, "step": 35568 }, { "epoch": 6.78, "learning_rate": 9.684080076263108e-07, "loss": 1.587, "step": 35569 }, { "epoch": 6.78, "learning_rate": 9.683508102955196e-07, "loss": 1.5687, "step": 35570 }, { "epoch": 6.78, "learning_rate": 9.682936129647282e-07, "loss": 2.0746, "step": 35571 }, { "epoch": 6.78, "learning_rate": 9.682364156339372e-07, "loss": 1.437, "step": 35572 }, { "epoch": 6.78, "learning_rate": 9.68179218303146e-07, "loss": 1.9826, "step": 35573 }, { "epoch": 6.78, "learning_rate": 9.681220209723546e-07, "loss": 1.5557, "step": 35574 }, { "epoch": 6.78, "learning_rate": 9.680648236415634e-07, "loss": 2.5385, "step": 35575 }, { "epoch": 6.78, "learning_rate": 9.680076263107722e-07, "loss": 2.129, "step": 35576 }, { "epoch": 6.78, "learning_rate": 9.67950428979981e-07, "loss": 1.5898, "step": 35577 }, { "epoch": 6.78, "learning_rate": 9.678932316491898e-07, "loss": 2.5606, "step": 35578 }, { "epoch": 6.78, "learning_rate": 9.678360343183985e-07, "loss": 1.3797, "step": 35579 }, { "epoch": 6.78, "learning_rate": 9.677788369876073e-07, "loss": 0.8523, "step": 35580 }, { "epoch": 6.78, "learning_rate": 9.67721639656816e-07, "loss": 1.5337, "step": 35581 }, { "epoch": 6.78, "learning_rate": 9.676644423260247e-07, "loss": 1.7762, "step": 35582 }, { "epoch": 6.78, "learning_rate": 9.676072449952337e-07, "loss": 1.0281, "step": 35583 }, { "epoch": 6.78, "learning_rate": 9.675500476644423e-07, "loss": 0.9073, "step": 35584 }, { "epoch": 6.78, "learning_rate": 9.67492850333651e-07, "loss": 0.527, "step": 35585 }, { "epoch": 6.78, "learning_rate": 9.6743565300286e-07, "loss": 2.0486, "step": 35586 }, { "epoch": 6.78, "learning_rate": 9.673784556720687e-07, "loss": 1.2641, "step": 35587 }, { "epoch": 6.78, "learning_rate": 9.673212583412775e-07, "loss": 1.6825, "step": 35588 }, { "epoch": 6.78, "learning_rate": 9.672640610104863e-07, "loss": 2.1471, "step": 35589 }, { "epoch": 6.78, "learning_rate": 9.67206863679695e-07, "loss": 1.379, "step": 35590 }, { "epoch": 6.78, "learning_rate": 9.671496663489037e-07, "loss": 2.2556, "step": 35591 }, { "epoch": 6.78, "learning_rate": 9.670924690181124e-07, "loss": 2.313, "step": 35592 }, { "epoch": 6.78, "learning_rate": 9.670352716873212e-07, "loss": 1.2089, "step": 35593 }, { "epoch": 6.78, "learning_rate": 9.6697807435653e-07, "loss": 1.6171, "step": 35594 }, { "epoch": 6.78, "learning_rate": 9.669208770257388e-07, "loss": 3.8285, "step": 35595 }, { "epoch": 6.78, "learning_rate": 9.668636796949476e-07, "loss": 1.8121, "step": 35596 }, { "epoch": 6.78, "learning_rate": 9.668064823641564e-07, "loss": 1.5216, "step": 35597 }, { "epoch": 6.78, "learning_rate": 9.66749285033365e-07, "loss": 1.8337, "step": 35598 }, { "epoch": 6.78, "learning_rate": 9.666920877025738e-07, "loss": 1.372, "step": 35599 }, { "epoch": 6.78, "learning_rate": 9.666348903717828e-07, "loss": 2.2158, "step": 35600 }, { "epoch": 6.78, "learning_rate": 9.665776930409916e-07, "loss": 1.6284, "step": 35601 }, { "epoch": 6.78, "learning_rate": 9.665204957102002e-07, "loss": 1.7025, "step": 35602 }, { "epoch": 6.78, "learning_rate": 9.66463298379409e-07, "loss": 2.8248, "step": 35603 }, { "epoch": 6.78, "learning_rate": 9.664061010486178e-07, "loss": 2.8062, "step": 35604 }, { "epoch": 6.78, "learning_rate": 9.663489037178266e-07, "loss": 1.8156, "step": 35605 }, { "epoch": 6.78, "learning_rate": 9.662917063870353e-07, "loss": 2.2818, "step": 35606 }, { "epoch": 6.78, "learning_rate": 9.662345090562441e-07, "loss": 2.2846, "step": 35607 }, { "epoch": 6.78, "learning_rate": 9.661773117254527e-07, "loss": 3.1647, "step": 35608 }, { "epoch": 6.78, "learning_rate": 9.661201143946615e-07, "loss": 0.9542, "step": 35609 }, { "epoch": 6.78, "learning_rate": 9.660629170638703e-07, "loss": 1.8403, "step": 35610 }, { "epoch": 6.78, "learning_rate": 9.660057197330793e-07, "loss": 0.538, "step": 35611 }, { "epoch": 6.78, "learning_rate": 9.65948522402288e-07, "loss": 1.3133, "step": 35612 }, { "epoch": 6.78, "learning_rate": 9.658913250714967e-07, "loss": 2.6066, "step": 35613 }, { "epoch": 6.78, "learning_rate": 9.658341277407055e-07, "loss": 4.0878, "step": 35614 }, { "epoch": 6.78, "learning_rate": 9.65776930409914e-07, "loss": 2.3622, "step": 35615 }, { "epoch": 6.78, "learning_rate": 9.65719733079123e-07, "loss": 0.8264, "step": 35616 }, { "epoch": 6.78, "learning_rate": 9.656625357483319e-07, "loss": 1.0228, "step": 35617 }, { "epoch": 6.78, "learning_rate": 9.656053384175407e-07, "loss": 1.3893, "step": 35618 }, { "epoch": 6.78, "learning_rate": 9.655481410867492e-07, "loss": 2.0604, "step": 35619 }, { "epoch": 6.78, "learning_rate": 9.65490943755958e-07, "loss": 1.584, "step": 35620 }, { "epoch": 6.78, "learning_rate": 9.654337464251668e-07, "loss": 0.6265, "step": 35621 }, { "epoch": 6.79, "learning_rate": 9.653765490943756e-07, "loss": 0.8317, "step": 35622 }, { "epoch": 6.79, "learning_rate": 9.653193517635844e-07, "loss": 1.7852, "step": 35623 }, { "epoch": 6.79, "learning_rate": 9.652621544327932e-07, "loss": 0.9599, "step": 35624 }, { "epoch": 6.79, "learning_rate": 9.652049571020018e-07, "loss": 1.1628, "step": 35625 }, { "epoch": 6.79, "learning_rate": 9.651477597712106e-07, "loss": 1.5668, "step": 35626 }, { "epoch": 6.79, "learning_rate": 9.650905624404196e-07, "loss": 1.7086, "step": 35627 }, { "epoch": 6.79, "learning_rate": 9.650333651096284e-07, "loss": 1.5705, "step": 35628 }, { "epoch": 6.79, "learning_rate": 9.64976167778837e-07, "loss": 1.2175, "step": 35629 }, { "epoch": 6.79, "learning_rate": 9.649189704480458e-07, "loss": 2.8243, "step": 35630 }, { "epoch": 6.79, "learning_rate": 9.648617731172546e-07, "loss": 2.2528, "step": 35631 }, { "epoch": 6.79, "learning_rate": 9.648045757864634e-07, "loss": 1.4756, "step": 35632 }, { "epoch": 6.79, "learning_rate": 9.647473784556721e-07, "loss": 1.1777, "step": 35633 }, { "epoch": 6.79, "learning_rate": 9.64690181124881e-07, "loss": 2.7507, "step": 35634 }, { "epoch": 6.79, "learning_rate": 9.646329837940897e-07, "loss": 2.3832, "step": 35635 }, { "epoch": 6.79, "learning_rate": 9.645757864632983e-07, "loss": 0.7754, "step": 35636 }, { "epoch": 6.79, "learning_rate": 9.645185891325071e-07, "loss": 1.9902, "step": 35637 }, { "epoch": 6.79, "learning_rate": 9.64461391801716e-07, "loss": 0.7674, "step": 35638 }, { "epoch": 6.79, "learning_rate": 9.644041944709247e-07, "loss": 1.8513, "step": 35639 }, { "epoch": 6.79, "learning_rate": 9.643469971401335e-07, "loss": 1.7883, "step": 35640 }, { "epoch": 6.79, "learning_rate": 9.642897998093423e-07, "loss": 3.376, "step": 35641 }, { "epoch": 6.79, "learning_rate": 9.64232602478551e-07, "loss": 2.2659, "step": 35642 }, { "epoch": 6.79, "learning_rate": 9.641754051477597e-07, "loss": 1.3044, "step": 35643 }, { "epoch": 6.79, "learning_rate": 9.641182078169687e-07, "loss": 1.4333, "step": 35644 }, { "epoch": 6.79, "learning_rate": 9.640610104861775e-07, "loss": 1.6852, "step": 35645 }, { "epoch": 6.79, "learning_rate": 9.64003813155386e-07, "loss": 2.2003, "step": 35646 }, { "epoch": 6.79, "learning_rate": 9.639466158245948e-07, "loss": 1.6796, "step": 35647 }, { "epoch": 6.79, "learning_rate": 9.638894184938036e-07, "loss": 1.4564, "step": 35648 }, { "epoch": 6.79, "learning_rate": 9.638322211630124e-07, "loss": 3.2054, "step": 35649 }, { "epoch": 6.79, "learning_rate": 9.637750238322212e-07, "loss": 1.8748, "step": 35650 }, { "epoch": 6.79, "learning_rate": 9.6371782650143e-07, "loss": 2.5391, "step": 35651 }, { "epoch": 6.79, "learning_rate": 9.636606291706388e-07, "loss": 1.0838, "step": 35652 }, { "epoch": 6.79, "learning_rate": 9.636034318398474e-07, "loss": 1.3743, "step": 35653 }, { "epoch": 6.79, "learning_rate": 9.635462345090562e-07, "loss": 1.6639, "step": 35654 }, { "epoch": 6.79, "learning_rate": 9.634890371782652e-07, "loss": 0.9501, "step": 35655 }, { "epoch": 6.79, "learning_rate": 9.634318398474738e-07, "loss": 1.2565, "step": 35656 }, { "epoch": 6.79, "learning_rate": 9.633746425166826e-07, "loss": 1.3012, "step": 35657 }, { "epoch": 6.79, "learning_rate": 9.633174451858914e-07, "loss": 3.0942, "step": 35658 }, { "epoch": 6.79, "learning_rate": 9.632602478551001e-07, "loss": 1.6301, "step": 35659 }, { "epoch": 6.79, "learning_rate": 9.63203050524309e-07, "loss": 1.5765, "step": 35660 }, { "epoch": 6.79, "learning_rate": 9.631458531935177e-07, "loss": 1.1418, "step": 35661 }, { "epoch": 6.79, "learning_rate": 9.630886558627265e-07, "loss": 2.2913, "step": 35662 }, { "epoch": 6.79, "learning_rate": 9.630314585319351e-07, "loss": 2.3431, "step": 35663 }, { "epoch": 6.79, "learning_rate": 9.62974261201144e-07, "loss": 1.069, "step": 35664 }, { "epoch": 6.79, "learning_rate": 9.629170638703527e-07, "loss": 1.7905, "step": 35665 }, { "epoch": 6.79, "learning_rate": 9.628598665395615e-07, "loss": 1.4221, "step": 35666 }, { "epoch": 6.79, "learning_rate": 9.628026692087703e-07, "loss": 1.3768, "step": 35667 }, { "epoch": 6.79, "learning_rate": 9.62745471877979e-07, "loss": 1.7725, "step": 35668 }, { "epoch": 6.79, "learning_rate": 9.626882745471879e-07, "loss": 2.2433, "step": 35669 }, { "epoch": 6.79, "learning_rate": 9.626310772163965e-07, "loss": 2.732, "step": 35670 }, { "epoch": 6.79, "learning_rate": 9.625738798856053e-07, "loss": 1.2774, "step": 35671 }, { "epoch": 6.79, "learning_rate": 9.625166825548143e-07, "loss": 0.8225, "step": 35672 }, { "epoch": 6.79, "learning_rate": 9.624594852240228e-07, "loss": 1.0837, "step": 35673 }, { "epoch": 6.8, "learning_rate": 9.624022878932316e-07, "loss": 1.3673, "step": 35674 }, { "epoch": 6.8, "learning_rate": 9.623450905624404e-07, "loss": 2.3522, "step": 35675 }, { "epoch": 6.8, "learning_rate": 9.622878932316492e-07, "loss": 0.9411, "step": 35676 }, { "epoch": 6.8, "learning_rate": 9.62230695900858e-07, "loss": 0.4816, "step": 35677 }, { "epoch": 6.8, "learning_rate": 9.621734985700668e-07, "loss": 2.3201, "step": 35678 }, { "epoch": 6.8, "learning_rate": 9.621163012392756e-07, "loss": 1.0695, "step": 35679 }, { "epoch": 6.8, "learning_rate": 9.620591039084842e-07, "loss": 1.0051, "step": 35680 }, { "epoch": 6.8, "learning_rate": 9.62001906577693e-07, "loss": 1.8276, "step": 35681 }, { "epoch": 6.8, "learning_rate": 9.619447092469018e-07, "loss": 1.0476, "step": 35682 }, { "epoch": 6.8, "learning_rate": 9.618875119161108e-07, "loss": 1.493, "step": 35683 }, { "epoch": 6.8, "learning_rate": 9.618303145853194e-07, "loss": 1.1788, "step": 35684 }, { "epoch": 6.8, "learning_rate": 9.617731172545282e-07, "loss": 0.3951, "step": 35685 }, { "epoch": 6.8, "learning_rate": 9.61715919923737e-07, "loss": 2.5426, "step": 35686 }, { "epoch": 6.8, "learning_rate": 9.616587225929455e-07, "loss": 1.466, "step": 35687 }, { "epoch": 6.8, "learning_rate": 9.616015252621545e-07, "loss": 1.421, "step": 35688 }, { "epoch": 6.8, "learning_rate": 9.615443279313633e-07, "loss": 1.0382, "step": 35689 }, { "epoch": 6.8, "learning_rate": 9.61487130600572e-07, "loss": 3.3614, "step": 35690 }, { "epoch": 6.8, "learning_rate": 9.614299332697807e-07, "loss": 0.8998, "step": 35691 }, { "epoch": 6.8, "learning_rate": 9.613727359389895e-07, "loss": 0.992, "step": 35692 }, { "epoch": 6.8, "learning_rate": 9.613155386081983e-07, "loss": 1.6726, "step": 35693 }, { "epoch": 6.8, "learning_rate": 9.61258341277407e-07, "loss": 1.7282, "step": 35694 }, { "epoch": 6.8, "learning_rate": 9.612011439466159e-07, "loss": 1.7908, "step": 35695 }, { "epoch": 6.8, "learning_rate": 9.611439466158247e-07, "loss": 1.6753, "step": 35696 }, { "epoch": 6.8, "learning_rate": 9.610867492850333e-07, "loss": 3.2104, "step": 35697 }, { "epoch": 6.8, "learning_rate": 9.61029551954242e-07, "loss": 2.5253, "step": 35698 }, { "epoch": 6.8, "learning_rate": 9.60972354623451e-07, "loss": 1.5183, "step": 35699 }, { "epoch": 6.8, "learning_rate": 9.609151572926598e-07, "loss": 2.2238, "step": 35700 }, { "epoch": 6.8, "learning_rate": 9.608579599618684e-07, "loss": 2.0738, "step": 35701 }, { "epoch": 6.8, "learning_rate": 9.608007626310772e-07, "loss": 2.2664, "step": 35702 }, { "epoch": 6.8, "learning_rate": 9.60743565300286e-07, "loss": 1.3091, "step": 35703 }, { "epoch": 6.8, "learning_rate": 9.606863679694948e-07, "loss": 1.0467, "step": 35704 }, { "epoch": 6.8, "learning_rate": 9.606291706387036e-07, "loss": 3.8395, "step": 35705 }, { "epoch": 6.8, "learning_rate": 9.605719733079124e-07, "loss": 0.7803, "step": 35706 }, { "epoch": 6.8, "learning_rate": 9.605147759771212e-07, "loss": 1.7999, "step": 35707 }, { "epoch": 6.8, "learning_rate": 9.604575786463298e-07, "loss": 1.5516, "step": 35708 }, { "epoch": 6.8, "learning_rate": 9.604003813155386e-07, "loss": 2.9413, "step": 35709 }, { "epoch": 6.8, "learning_rate": 9.603431839847474e-07, "loss": 1.548, "step": 35710 }, { "epoch": 6.8, "learning_rate": 9.602859866539562e-07, "loss": 2.5006, "step": 35711 }, { "epoch": 6.8, "learning_rate": 9.60228789323165e-07, "loss": 0.5047, "step": 35712 }, { "epoch": 6.8, "learning_rate": 9.601715919923737e-07, "loss": 1.9979, "step": 35713 }, { "epoch": 6.8, "learning_rate": 9.601143946615823e-07, "loss": 3.4566, "step": 35714 }, { "epoch": 6.8, "learning_rate": 9.600571973307911e-07, "loss": 0.5432, "step": 35715 }, { "epoch": 6.8, "learning_rate": 9.600000000000001e-07, "loss": 1.1906, "step": 35716 }, { "epoch": 6.8, "learning_rate": 9.59942802669209e-07, "loss": 0.9177, "step": 35717 }, { "epoch": 6.8, "learning_rate": 9.598856053384175e-07, "loss": 0.9598, "step": 35718 }, { "epoch": 6.8, "learning_rate": 9.598284080076263e-07, "loss": 0.6598, "step": 35719 }, { "epoch": 6.8, "learning_rate": 9.59771210676835e-07, "loss": 1.7859, "step": 35720 }, { "epoch": 6.8, "learning_rate": 9.597140133460439e-07, "loss": 1.3496, "step": 35721 }, { "epoch": 6.8, "learning_rate": 9.596568160152527e-07, "loss": 0.9191, "step": 35722 }, { "epoch": 6.8, "learning_rate": 9.595996186844615e-07, "loss": 2.6266, "step": 35723 }, { "epoch": 6.8, "learning_rate": 9.595424213536703e-07, "loss": 1.8851, "step": 35724 }, { "epoch": 6.8, "learning_rate": 9.594852240228788e-07, "loss": 1.4455, "step": 35725 }, { "epoch": 6.8, "learning_rate": 9.594280266920876e-07, "loss": 1.6775, "step": 35726 }, { "epoch": 6.81, "learning_rate": 9.593708293612966e-07, "loss": 2.7502, "step": 35727 }, { "epoch": 6.81, "learning_rate": 9.593136320305052e-07, "loss": 1.6414, "step": 35728 }, { "epoch": 6.81, "learning_rate": 9.59256434699714e-07, "loss": 1.4013, "step": 35729 }, { "epoch": 6.81, "learning_rate": 9.591992373689228e-07, "loss": 1.9581, "step": 35730 }, { "epoch": 6.81, "learning_rate": 9.591420400381314e-07, "loss": 2.6022, "step": 35731 }, { "epoch": 6.81, "learning_rate": 9.590848427073404e-07, "loss": 2.2515, "step": 35732 }, { "epoch": 6.81, "learning_rate": 9.590276453765492e-07, "loss": 2.6318, "step": 35733 }, { "epoch": 6.81, "learning_rate": 9.58970448045758e-07, "loss": 1.1092, "step": 35734 }, { "epoch": 6.81, "learning_rate": 9.589132507149666e-07, "loss": 0.9809, "step": 35735 }, { "epoch": 6.81, "learning_rate": 9.588560533841754e-07, "loss": 1.3143, "step": 35736 }, { "epoch": 6.81, "learning_rate": 9.587988560533842e-07, "loss": 2.3924, "step": 35737 }, { "epoch": 6.81, "learning_rate": 9.58741658722593e-07, "loss": 1.1194, "step": 35738 }, { "epoch": 6.81, "learning_rate": 9.586844613918017e-07, "loss": 1.6502, "step": 35739 }, { "epoch": 6.81, "learning_rate": 9.586272640610105e-07, "loss": 1.852, "step": 35740 }, { "epoch": 6.81, "learning_rate": 9.585700667302193e-07, "loss": 0.5097, "step": 35741 }, { "epoch": 6.81, "learning_rate": 9.58512869399428e-07, "loss": 2.3924, "step": 35742 }, { "epoch": 6.81, "learning_rate": 9.58455672068637e-07, "loss": 2.9197, "step": 35743 }, { "epoch": 6.81, "learning_rate": 9.583984747378457e-07, "loss": 0.8013, "step": 35744 }, { "epoch": 6.81, "learning_rate": 9.583412774070543e-07, "loss": 0.9818, "step": 35745 }, { "epoch": 6.81, "learning_rate": 9.58284080076263e-07, "loss": 1.8312, "step": 35746 }, { "epoch": 6.81, "learning_rate": 9.582268827454719e-07, "loss": 0.8159, "step": 35747 }, { "epoch": 6.81, "learning_rate": 9.581696854146807e-07, "loss": 2.3828, "step": 35748 }, { "epoch": 6.81, "learning_rate": 9.581124880838895e-07, "loss": 1.7677, "step": 35749 }, { "epoch": 6.81, "learning_rate": 9.580552907530983e-07, "loss": 1.1851, "step": 35750 }, { "epoch": 6.81, "learning_rate": 9.57998093422307e-07, "loss": 2.2375, "step": 35751 }, { "epoch": 6.81, "learning_rate": 9.579408960915156e-07, "loss": 1.0628, "step": 35752 }, { "epoch": 6.81, "learning_rate": 9.578836987607244e-07, "loss": 1.9312, "step": 35753 }, { "epoch": 6.81, "learning_rate": 9.578265014299332e-07, "loss": 1.8143, "step": 35754 }, { "epoch": 6.81, "learning_rate": 9.57769304099142e-07, "loss": 1.9292, "step": 35755 }, { "epoch": 6.81, "learning_rate": 9.577121067683508e-07, "loss": 2.2758, "step": 35756 }, { "epoch": 6.81, "learning_rate": 9.576549094375596e-07, "loss": 0.6414, "step": 35757 }, { "epoch": 6.81, "learning_rate": 9.575977121067684e-07, "loss": 2.8208, "step": 35758 }, { "epoch": 6.81, "learning_rate": 9.57540514775977e-07, "loss": 1.7496, "step": 35759 }, { "epoch": 6.81, "learning_rate": 9.57483317445186e-07, "loss": 1.0709, "step": 35760 }, { "epoch": 6.81, "learning_rate": 9.574261201143948e-07, "loss": 1.817, "step": 35761 }, { "epoch": 6.81, "learning_rate": 9.573689227836034e-07, "loss": 1.7035, "step": 35762 }, { "epoch": 6.81, "learning_rate": 9.573117254528122e-07, "loss": 3.1791, "step": 35763 }, { "epoch": 6.81, "learning_rate": 9.57254528122021e-07, "loss": 2.2391, "step": 35764 }, { "epoch": 6.81, "learning_rate": 9.571973307912298e-07, "loss": 1.0274, "step": 35765 }, { "epoch": 6.81, "learning_rate": 9.571401334604385e-07, "loss": 1.5287, "step": 35766 }, { "epoch": 6.81, "learning_rate": 9.570829361296473e-07, "loss": 1.1112, "step": 35767 }, { "epoch": 6.81, "learning_rate": 9.570257387988561e-07, "loss": 2.5139, "step": 35768 }, { "epoch": 6.81, "learning_rate": 9.569685414680647e-07, "loss": 1.1071, "step": 35769 }, { "epoch": 6.81, "learning_rate": 9.569113441372735e-07, "loss": 0.8653, "step": 35770 }, { "epoch": 6.81, "learning_rate": 9.568541468064825e-07, "loss": 1.703, "step": 35771 }, { "epoch": 6.81, "learning_rate": 9.567969494756913e-07, "loss": 1.5699, "step": 35772 }, { "epoch": 6.81, "learning_rate": 9.567397521448999e-07, "loss": 1.3927, "step": 35773 }, { "epoch": 6.81, "learning_rate": 9.566825548141087e-07, "loss": 1.1568, "step": 35774 }, { "epoch": 6.81, "learning_rate": 9.566253574833175e-07, "loss": 1.5777, "step": 35775 }, { "epoch": 6.81, "learning_rate": 9.565681601525263e-07, "loss": 0.9674, "step": 35776 }, { "epoch": 6.81, "learning_rate": 9.56510962821735e-07, "loss": 1.435, "step": 35777 }, { "epoch": 6.81, "learning_rate": 9.564537654909439e-07, "loss": 3.4951, "step": 35778 }, { "epoch": 6.82, "learning_rate": 9.563965681601524e-07, "loss": 2.29, "step": 35779 }, { "epoch": 6.82, "learning_rate": 9.563393708293612e-07, "loss": 0.8516, "step": 35780 }, { "epoch": 6.82, "learning_rate": 9.5628217349857e-07, "loss": 1.5545, "step": 35781 }, { "epoch": 6.82, "learning_rate": 9.56224976167779e-07, "loss": 1.2786, "step": 35782 }, { "epoch": 6.82, "learning_rate": 9.561677788369876e-07, "loss": 1.6933, "step": 35783 }, { "epoch": 6.82, "learning_rate": 9.561105815061964e-07, "loss": 1.3827, "step": 35784 }, { "epoch": 6.82, "learning_rate": 9.560533841754052e-07, "loss": 1.6148, "step": 35785 }, { "epoch": 6.82, "learning_rate": 9.559961868446138e-07, "loss": 1.8375, "step": 35786 }, { "epoch": 6.82, "learning_rate": 9.559389895138228e-07, "loss": 1.3778, "step": 35787 }, { "epoch": 6.82, "learning_rate": 9.558817921830316e-07, "loss": 1.8738, "step": 35788 }, { "epoch": 6.82, "learning_rate": 9.558245948522404e-07, "loss": 0.8431, "step": 35789 }, { "epoch": 6.82, "learning_rate": 9.55767397521449e-07, "loss": 3.2703, "step": 35790 }, { "epoch": 6.82, "learning_rate": 9.557102001906578e-07, "loss": 2.4168, "step": 35791 }, { "epoch": 6.82, "learning_rate": 9.556530028598666e-07, "loss": 2.7591, "step": 35792 }, { "epoch": 6.82, "learning_rate": 9.555958055290753e-07, "loss": 0.5833, "step": 35793 }, { "epoch": 6.82, "learning_rate": 9.555386081982841e-07, "loss": 1.7666, "step": 35794 }, { "epoch": 6.82, "learning_rate": 9.55481410867493e-07, "loss": 2.1707, "step": 35795 }, { "epoch": 6.82, "learning_rate": 9.554242135367015e-07, "loss": 1.7317, "step": 35796 }, { "epoch": 6.82, "learning_rate": 9.553670162059103e-07, "loss": 2.1501, "step": 35797 }, { "epoch": 6.82, "learning_rate": 9.55309818875119e-07, "loss": 2.9498, "step": 35798 }, { "epoch": 6.82, "learning_rate": 9.552526215443281e-07, "loss": 1.4209, "step": 35799 }, { "epoch": 6.82, "learning_rate": 9.551954242135367e-07, "loss": 0.6106, "step": 35800 }, { "epoch": 6.82, "learning_rate": 9.551382268827455e-07, "loss": 1.2222, "step": 35801 }, { "epoch": 6.82, "learning_rate": 9.550810295519543e-07, "loss": 1.9011, "step": 35802 }, { "epoch": 6.82, "learning_rate": 9.550238322211629e-07, "loss": 2.7146, "step": 35803 }, { "epoch": 6.82, "learning_rate": 9.549666348903719e-07, "loss": 0.454, "step": 35804 }, { "epoch": 6.82, "learning_rate": 9.549094375595807e-07, "loss": 1.9376, "step": 35805 }, { "epoch": 6.82, "learning_rate": 9.548522402287895e-07, "loss": 2.6751, "step": 35806 }, { "epoch": 6.82, "learning_rate": 9.54795042897998e-07, "loss": 0.8362, "step": 35807 }, { "epoch": 6.82, "learning_rate": 9.547378455672068e-07, "loss": 2.7955, "step": 35808 }, { "epoch": 6.82, "learning_rate": 9.546806482364156e-07, "loss": 0.6305, "step": 35809 }, { "epoch": 6.82, "learning_rate": 9.546234509056244e-07, "loss": 1.139, "step": 35810 }, { "epoch": 6.82, "learning_rate": 9.545662535748332e-07, "loss": 1.571, "step": 35811 }, { "epoch": 6.82, "learning_rate": 9.54509056244042e-07, "loss": 3.5067, "step": 35812 }, { "epoch": 6.82, "learning_rate": 9.544518589132508e-07, "loss": 1.407, "step": 35813 }, { "epoch": 6.82, "learning_rate": 9.543946615824594e-07, "loss": 2.0599, "step": 35814 }, { "epoch": 6.82, "learning_rate": 9.543374642516684e-07, "loss": 1.6931, "step": 35815 }, { "epoch": 6.82, "learning_rate": 9.542802669208772e-07, "loss": 1.4037, "step": 35816 }, { "epoch": 6.82, "learning_rate": 9.542230695900858e-07, "loss": 2.2645, "step": 35817 }, { "epoch": 6.82, "learning_rate": 9.541658722592946e-07, "loss": 0.903, "step": 35818 }, { "epoch": 6.82, "learning_rate": 9.541086749285033e-07, "loss": 2.5699, "step": 35819 }, { "epoch": 6.82, "learning_rate": 9.540514775977121e-07, "loss": 3.0977, "step": 35820 }, { "epoch": 6.82, "learning_rate": 9.53994280266921e-07, "loss": 2.528, "step": 35821 }, { "epoch": 6.82, "learning_rate": 9.539370829361297e-07, "loss": 1.0861, "step": 35822 }, { "epoch": 6.82, "learning_rate": 9.538798856053385e-07, "loss": 2.0845, "step": 35823 }, { "epoch": 6.82, "learning_rate": 9.53822688274547e-07, "loss": 1.4748, "step": 35824 }, { "epoch": 6.82, "learning_rate": 9.53765490943756e-07, "loss": 2.223, "step": 35825 }, { "epoch": 6.82, "learning_rate": 9.537082936129647e-07, "loss": 1.6964, "step": 35826 }, { "epoch": 6.82, "learning_rate": 9.536510962821735e-07, "loss": 1.3507, "step": 35827 }, { "epoch": 6.82, "learning_rate": 9.535938989513823e-07, "loss": 2.3352, "step": 35828 }, { "epoch": 6.82, "learning_rate": 9.535367016205911e-07, "loss": 1.7545, "step": 35829 }, { "epoch": 6.82, "learning_rate": 9.534795042897999e-07, "loss": 1.1568, "step": 35830 }, { "epoch": 6.82, "learning_rate": 9.534223069590086e-07, "loss": 1.3924, "step": 35831 }, { "epoch": 6.83, "learning_rate": 9.533651096282174e-07, "loss": 1.8519, "step": 35832 }, { "epoch": 6.83, "learning_rate": 9.533079122974262e-07, "loss": 3.118, "step": 35833 }, { "epoch": 6.83, "learning_rate": 9.532507149666348e-07, "loss": 2.5203, "step": 35834 }, { "epoch": 6.83, "learning_rate": 9.531935176358436e-07, "loss": 1.6945, "step": 35835 }, { "epoch": 6.83, "learning_rate": 9.531363203050525e-07, "loss": 3.0704, "step": 35836 }, { "epoch": 6.83, "learning_rate": 9.530791229742613e-07, "loss": 2.4338, "step": 35837 }, { "epoch": 6.83, "learning_rate": 9.5302192564347e-07, "loss": 2.0138, "step": 35838 }, { "epoch": 6.83, "learning_rate": 9.529647283126788e-07, "loss": 2.1473, "step": 35839 }, { "epoch": 6.83, "learning_rate": 9.529075309818876e-07, "loss": 0.33, "step": 35840 }, { "epoch": 6.83, "learning_rate": 9.528503336510963e-07, "loss": 1.2935, "step": 35841 }, { "epoch": 6.83, "learning_rate": 9.527931363203051e-07, "loss": 3.4215, "step": 35842 }, { "epoch": 6.83, "learning_rate": 9.527359389895139e-07, "loss": 1.8636, "step": 35843 }, { "epoch": 6.83, "learning_rate": 9.526787416587226e-07, "loss": 1.3039, "step": 35844 }, { "epoch": 6.83, "learning_rate": 9.526215443279314e-07, "loss": 1.4969, "step": 35845 }, { "epoch": 6.83, "learning_rate": 9.525643469971401e-07, "loss": 2.5092, "step": 35846 }, { "epoch": 6.83, "learning_rate": 9.52507149666349e-07, "loss": 2.3235, "step": 35847 }, { "epoch": 6.83, "learning_rate": 9.524499523355576e-07, "loss": 2.5003, "step": 35848 }, { "epoch": 6.83, "learning_rate": 9.523927550047664e-07, "loss": 2.7455, "step": 35849 }, { "epoch": 6.83, "learning_rate": 9.523355576739753e-07, "loss": 1.417, "step": 35850 }, { "epoch": 6.83, "learning_rate": 9.522783603431839e-07, "loss": 1.3029, "step": 35851 }, { "epoch": 6.83, "learning_rate": 9.522211630123928e-07, "loss": 1.6022, "step": 35852 }, { "epoch": 6.83, "learning_rate": 9.521639656816016e-07, "loss": 1.3354, "step": 35853 }, { "epoch": 6.83, "learning_rate": 9.521067683508104e-07, "loss": 1.6917, "step": 35854 }, { "epoch": 6.83, "learning_rate": 9.520495710200191e-07, "loss": 2.29, "step": 35855 }, { "epoch": 6.83, "learning_rate": 9.519923736892279e-07, "loss": 2.7079, "step": 35856 }, { "epoch": 6.83, "learning_rate": 9.519351763584367e-07, "loss": 1.0052, "step": 35857 }, { "epoch": 6.83, "learning_rate": 9.518779790276454e-07, "loss": 1.3886, "step": 35858 }, { "epoch": 6.83, "learning_rate": 9.518207816968541e-07, "loss": 0.7925, "step": 35859 }, { "epoch": 6.83, "learning_rate": 9.517635843660629e-07, "loss": 0.6529, "step": 35860 }, { "epoch": 6.83, "learning_rate": 9.517063870352716e-07, "loss": 0.7093, "step": 35861 }, { "epoch": 6.83, "learning_rate": 9.516491897044804e-07, "loss": 1.9272, "step": 35862 }, { "epoch": 6.83, "learning_rate": 9.515919923736892e-07, "loss": 3.9466, "step": 35863 }, { "epoch": 6.83, "learning_rate": 9.515347950428981e-07, "loss": 2.3765, "step": 35864 }, { "epoch": 6.83, "learning_rate": 9.514775977121067e-07, "loss": 2.8724, "step": 35865 }, { "epoch": 6.83, "learning_rate": 9.514204003813156e-07, "loss": 2.1354, "step": 35866 }, { "epoch": 6.83, "learning_rate": 9.513632030505244e-07, "loss": 1.7119, "step": 35867 }, { "epoch": 6.83, "learning_rate": 9.513060057197331e-07, "loss": 0.4867, "step": 35868 }, { "epoch": 6.83, "learning_rate": 9.512488083889419e-07, "loss": 3.1066, "step": 35869 }, { "epoch": 6.83, "learning_rate": 9.511916110581507e-07, "loss": 1.3104, "step": 35870 }, { "epoch": 6.83, "learning_rate": 9.511344137273595e-07, "loss": 1.5572, "step": 35871 }, { "epoch": 6.83, "learning_rate": 9.510772163965681e-07, "loss": 2.4522, "step": 35872 }, { "epoch": 6.83, "learning_rate": 9.510200190657769e-07, "loss": 1.6694, "step": 35873 }, { "epoch": 6.83, "learning_rate": 9.509628217349857e-07, "loss": 1.8847, "step": 35874 }, { "epoch": 6.83, "learning_rate": 9.509056244041944e-07, "loss": 1.9089, "step": 35875 }, { "epoch": 6.83, "learning_rate": 9.508484270734032e-07, "loss": 0.884, "step": 35876 }, { "epoch": 6.83, "learning_rate": 9.507912297426121e-07, "loss": 1.7538, "step": 35877 }, { "epoch": 6.83, "learning_rate": 9.507340324118209e-07, "loss": 2.0544, "step": 35878 }, { "epoch": 6.83, "learning_rate": 9.506768350810295e-07, "loss": 1.8799, "step": 35879 }, { "epoch": 6.83, "learning_rate": 9.506196377502384e-07, "loss": 2.6753, "step": 35880 }, { "epoch": 6.83, "learning_rate": 9.505624404194472e-07, "loss": 1.6081, "step": 35881 }, { "epoch": 6.83, "learning_rate": 9.505052430886559e-07, "loss": 1.5326, "step": 35882 }, { "epoch": 6.83, "learning_rate": 9.504480457578647e-07, "loss": 1.6479, "step": 35883 }, { "epoch": 6.84, "learning_rate": 9.503908484270735e-07, "loss": 2.472, "step": 35884 }, { "epoch": 6.84, "learning_rate": 9.503336510962822e-07, "loss": 2.0765, "step": 35885 }, { "epoch": 6.84, "learning_rate": 9.502764537654909e-07, "loss": 1.4062, "step": 35886 }, { "epoch": 6.84, "learning_rate": 9.502192564346997e-07, "loss": 0.4401, "step": 35887 }, { "epoch": 6.84, "learning_rate": 9.501620591039085e-07, "loss": 1.8096, "step": 35888 }, { "epoch": 6.84, "learning_rate": 9.501048617731172e-07, "loss": 1.5563, "step": 35889 }, { "epoch": 6.84, "learning_rate": 9.50047664442326e-07, "loss": 0.7041, "step": 35890 }, { "epoch": 6.84, "learning_rate": 9.499904671115349e-07, "loss": 2.0311, "step": 35891 }, { "epoch": 6.84, "learning_rate": 9.499332697807435e-07, "loss": 0.8114, "step": 35892 }, { "epoch": 6.84, "learning_rate": 9.498760724499523e-07, "loss": 2.2217, "step": 35893 }, { "epoch": 6.84, "learning_rate": 9.498188751191612e-07, "loss": 2.1237, "step": 35894 }, { "epoch": 6.84, "learning_rate": 9.4976167778837e-07, "loss": 1.7531, "step": 35895 }, { "epoch": 6.84, "learning_rate": 9.497044804575787e-07, "loss": 1.8943, "step": 35896 }, { "epoch": 6.84, "learning_rate": 9.496472831267875e-07, "loss": 1.7671, "step": 35897 }, { "epoch": 6.84, "learning_rate": 9.495900857959963e-07, "loss": 1.6697, "step": 35898 }, { "epoch": 6.84, "learning_rate": 9.49532888465205e-07, "loss": 1.9343, "step": 35899 }, { "epoch": 6.84, "learning_rate": 9.494756911344137e-07, "loss": 1.7528, "step": 35900 }, { "epoch": 6.84, "learning_rate": 9.494184938036225e-07, "loss": 1.3357, "step": 35901 }, { "epoch": 6.84, "learning_rate": 9.493612964728313e-07, "loss": 1.1008, "step": 35902 }, { "epoch": 6.84, "learning_rate": 9.4930409914204e-07, "loss": 1.911, "step": 35903 }, { "epoch": 6.84, "learning_rate": 9.492469018112488e-07, "loss": 1.6907, "step": 35904 }, { "epoch": 6.84, "learning_rate": 9.491897044804577e-07, "loss": 1.0286, "step": 35905 }, { "epoch": 6.84, "learning_rate": 9.491325071496663e-07, "loss": 1.4357, "step": 35906 }, { "epoch": 6.84, "learning_rate": 9.490753098188751e-07, "loss": 2.5454, "step": 35907 }, { "epoch": 6.84, "learning_rate": 9.49018112488084e-07, "loss": 2.4511, "step": 35908 }, { "epoch": 6.84, "learning_rate": 9.489609151572926e-07, "loss": 1.8345, "step": 35909 }, { "epoch": 6.84, "learning_rate": 9.489037178265015e-07, "loss": 1.5649, "step": 35910 }, { "epoch": 6.84, "learning_rate": 9.488465204957103e-07, "loss": 1.696, "step": 35911 }, { "epoch": 6.84, "learning_rate": 9.487893231649191e-07, "loss": 1.7527, "step": 35912 }, { "epoch": 6.84, "learning_rate": 9.487321258341277e-07, "loss": 1.3535, "step": 35913 }, { "epoch": 6.84, "learning_rate": 9.486749285033365e-07, "loss": 1.7173, "step": 35914 }, { "epoch": 6.84, "learning_rate": 9.486177311725453e-07, "loss": 1.9137, "step": 35915 }, { "epoch": 6.84, "learning_rate": 9.48560533841754e-07, "loss": 2.56, "step": 35916 }, { "epoch": 6.84, "learning_rate": 9.485033365109628e-07, "loss": 0.5813, "step": 35917 }, { "epoch": 6.84, "learning_rate": 9.484461391801716e-07, "loss": 0.958, "step": 35918 }, { "epoch": 6.84, "learning_rate": 9.483889418493805e-07, "loss": 1.1124, "step": 35919 }, { "epoch": 6.84, "learning_rate": 9.483317445185891e-07, "loss": 2.053, "step": 35920 }, { "epoch": 6.84, "learning_rate": 9.48274547187798e-07, "loss": 1.277, "step": 35921 }, { "epoch": 6.84, "learning_rate": 9.482173498570068e-07, "loss": 2.7811, "step": 35922 }, { "epoch": 6.84, "learning_rate": 9.481601525262154e-07, "loss": 1.3576, "step": 35923 }, { "epoch": 6.84, "learning_rate": 9.481029551954243e-07, "loss": 1.6264, "step": 35924 }, { "epoch": 6.84, "learning_rate": 9.480457578646331e-07, "loss": 2.5719, "step": 35925 }, { "epoch": 6.84, "learning_rate": 9.479885605338417e-07, "loss": 2.0721, "step": 35926 }, { "epoch": 6.84, "learning_rate": 9.479313632030505e-07, "loss": 0.4531, "step": 35927 }, { "epoch": 6.84, "learning_rate": 9.478741658722593e-07, "loss": 1.7175, "step": 35928 }, { "epoch": 6.84, "learning_rate": 9.478169685414681e-07, "loss": 1.5964, "step": 35929 }, { "epoch": 6.84, "learning_rate": 9.477597712106768e-07, "loss": 2.397, "step": 35930 }, { "epoch": 6.84, "learning_rate": 9.477025738798856e-07, "loss": 1.1154, "step": 35931 }, { "epoch": 6.84, "learning_rate": 9.476453765490944e-07, "loss": 1.9188, "step": 35932 }, { "epoch": 6.84, "learning_rate": 9.475881792183031e-07, "loss": 0.8114, "step": 35933 }, { "epoch": 6.84, "learning_rate": 9.475309818875119e-07, "loss": 1.3036, "step": 35934 }, { "epoch": 6.84, "learning_rate": 9.474737845567208e-07, "loss": 2.2149, "step": 35935 }, { "epoch": 6.84, "learning_rate": 9.474165872259296e-07, "loss": 1.0465, "step": 35936 }, { "epoch": 6.85, "learning_rate": 9.473593898951382e-07, "loss": 1.8516, "step": 35937 }, { "epoch": 6.85, "learning_rate": 9.473021925643471e-07, "loss": 2.9036, "step": 35938 }, { "epoch": 6.85, "learning_rate": 9.472449952335559e-07, "loss": 0.752, "step": 35939 }, { "epoch": 6.85, "learning_rate": 9.471877979027645e-07, "loss": 2.9728, "step": 35940 }, { "epoch": 6.85, "learning_rate": 9.471306005719733e-07, "loss": 1.1658, "step": 35941 }, { "epoch": 6.85, "learning_rate": 9.470734032411821e-07, "loss": 1.4919, "step": 35942 }, { "epoch": 6.85, "learning_rate": 9.470162059103909e-07, "loss": 1.6186, "step": 35943 }, { "epoch": 6.85, "learning_rate": 9.469590085795996e-07, "loss": 2.5326, "step": 35944 }, { "epoch": 6.85, "learning_rate": 9.469018112488084e-07, "loss": 3.1404, "step": 35945 }, { "epoch": 6.85, "learning_rate": 9.468446139180172e-07, "loss": 1.2713, "step": 35946 }, { "epoch": 6.85, "learning_rate": 9.467874165872259e-07, "loss": 1.6025, "step": 35947 }, { "epoch": 6.85, "learning_rate": 9.467302192564347e-07, "loss": 3.2164, "step": 35948 }, { "epoch": 6.85, "learning_rate": 9.466730219256436e-07, "loss": 3.2725, "step": 35949 }, { "epoch": 6.85, "learning_rate": 9.466158245948522e-07, "loss": 2.1154, "step": 35950 }, { "epoch": 6.85, "learning_rate": 9.46558627264061e-07, "loss": 1.0366, "step": 35951 }, { "epoch": 6.85, "learning_rate": 9.465014299332699e-07, "loss": 2.0194, "step": 35952 }, { "epoch": 6.85, "learning_rate": 9.464442326024786e-07, "loss": 0.467, "step": 35953 }, { "epoch": 6.85, "learning_rate": 9.463870352716873e-07, "loss": 1.4649, "step": 35954 }, { "epoch": 6.85, "learning_rate": 9.463298379408961e-07, "loss": 3.2946, "step": 35955 }, { "epoch": 6.85, "learning_rate": 9.462726406101049e-07, "loss": 1.9595, "step": 35956 }, { "epoch": 6.85, "learning_rate": 9.462154432793136e-07, "loss": 3.1595, "step": 35957 }, { "epoch": 6.85, "learning_rate": 9.461582459485224e-07, "loss": 3.1556, "step": 35958 }, { "epoch": 6.85, "learning_rate": 9.461010486177312e-07, "loss": 1.9759, "step": 35959 }, { "epoch": 6.85, "learning_rate": 9.4604385128694e-07, "loss": 2.6214, "step": 35960 }, { "epoch": 6.85, "learning_rate": 9.459866539561487e-07, "loss": 2.5975, "step": 35961 }, { "epoch": 6.85, "learning_rate": 9.459294566253575e-07, "loss": 1.0489, "step": 35962 }, { "epoch": 6.85, "learning_rate": 9.458722592945664e-07, "loss": 2.3688, "step": 35963 }, { "epoch": 6.85, "learning_rate": 9.45815061963775e-07, "loss": 2.4117, "step": 35964 }, { "epoch": 6.85, "learning_rate": 9.457578646329838e-07, "loss": 1.4918, "step": 35965 }, { "epoch": 6.85, "learning_rate": 9.457006673021927e-07, "loss": 1.6103, "step": 35966 }, { "epoch": 6.85, "learning_rate": 9.456434699714012e-07, "loss": 1.2483, "step": 35967 }, { "epoch": 6.85, "learning_rate": 9.455862726406101e-07, "loss": 2.4257, "step": 35968 }, { "epoch": 6.85, "learning_rate": 9.455290753098189e-07, "loss": 2.061, "step": 35969 }, { "epoch": 6.85, "learning_rate": 9.454718779790277e-07, "loss": 1.6797, "step": 35970 }, { "epoch": 6.85, "learning_rate": 9.454146806482364e-07, "loss": 0.8127, "step": 35971 }, { "epoch": 6.85, "learning_rate": 9.453574833174452e-07, "loss": 1.7038, "step": 35972 }, { "epoch": 6.85, "learning_rate": 9.45300285986654e-07, "loss": 1.8899, "step": 35973 }, { "epoch": 6.85, "learning_rate": 9.452430886558627e-07, "loss": 1.4636, "step": 35974 }, { "epoch": 6.85, "learning_rate": 9.451858913250715e-07, "loss": 1.9546, "step": 35975 }, { "epoch": 6.85, "learning_rate": 9.451286939942803e-07, "loss": 1.8152, "step": 35976 }, { "epoch": 6.85, "learning_rate": 9.450714966634892e-07, "loss": 1.7559, "step": 35977 }, { "epoch": 6.85, "learning_rate": 9.450142993326978e-07, "loss": 2.6744, "step": 35978 }, { "epoch": 6.85, "learning_rate": 9.449571020019067e-07, "loss": 1.4296, "step": 35979 }, { "epoch": 6.85, "learning_rate": 9.448999046711154e-07, "loss": 2.3459, "step": 35980 }, { "epoch": 6.85, "learning_rate": 9.44842707340324e-07, "loss": 1.6178, "step": 35981 }, { "epoch": 6.85, "learning_rate": 9.447855100095329e-07, "loss": 1.351, "step": 35982 }, { "epoch": 6.85, "learning_rate": 9.447283126787417e-07, "loss": 1.8026, "step": 35983 }, { "epoch": 6.85, "learning_rate": 9.446711153479505e-07, "loss": 1.6855, "step": 35984 }, { "epoch": 6.85, "learning_rate": 9.446139180171592e-07, "loss": 2.8236, "step": 35985 }, { "epoch": 6.85, "learning_rate": 9.44556720686368e-07, "loss": 2.5062, "step": 35986 }, { "epoch": 6.85, "learning_rate": 9.444995233555768e-07, "loss": 3.7011, "step": 35987 }, { "epoch": 6.85, "learning_rate": 9.444423260247855e-07, "loss": 2.247, "step": 35988 }, { "epoch": 6.86, "learning_rate": 9.443851286939943e-07, "loss": 1.5721, "step": 35989 }, { "epoch": 6.86, "learning_rate": 9.443279313632031e-07, "loss": 1.245, "step": 35990 }, { "epoch": 6.86, "learning_rate": 9.442707340324118e-07, "loss": 2.7975, "step": 35991 }, { "epoch": 6.86, "learning_rate": 9.442135367016205e-07, "loss": 1.4673, "step": 35992 }, { "epoch": 6.86, "learning_rate": 9.441563393708294e-07, "loss": 1.9111, "step": 35993 }, { "epoch": 6.86, "learning_rate": 9.440991420400382e-07, "loss": 0.6297, "step": 35994 }, { "epoch": 6.86, "learning_rate": 9.440419447092468e-07, "loss": 1.6098, "step": 35995 }, { "epoch": 6.86, "learning_rate": 9.439847473784557e-07, "loss": 2.6113, "step": 35996 }, { "epoch": 6.86, "learning_rate": 9.439275500476645e-07, "loss": 1.5768, "step": 35997 }, { "epoch": 6.86, "learning_rate": 9.438703527168732e-07, "loss": 3.1763, "step": 35998 }, { "epoch": 6.86, "learning_rate": 9.43813155386082e-07, "loss": 1.3649, "step": 35999 }, { "epoch": 6.86, "learning_rate": 9.437559580552908e-07, "loss": 1.007, "step": 36000 }, { "epoch": 6.86, "learning_rate": 9.436987607244996e-07, "loss": 1.3662, "step": 36001 }, { "epoch": 6.86, "learning_rate": 9.436415633937083e-07, "loss": 1.0948, "step": 36002 }, { "epoch": 6.86, "learning_rate": 9.435843660629171e-07, "loss": 2.9967, "step": 36003 }, { "epoch": 6.86, "learning_rate": 9.435271687321259e-07, "loss": 2.2138, "step": 36004 }, { "epoch": 6.86, "learning_rate": 9.434699714013346e-07, "loss": 1.8232, "step": 36005 }, { "epoch": 6.86, "learning_rate": 9.434127740705433e-07, "loss": 2.1463, "step": 36006 }, { "epoch": 6.86, "learning_rate": 9.433555767397522e-07, "loss": 3.7916, "step": 36007 }, { "epoch": 6.86, "learning_rate": 9.43298379408961e-07, "loss": 1.4603, "step": 36008 }, { "epoch": 6.86, "learning_rate": 9.432411820781696e-07, "loss": 2.8458, "step": 36009 }, { "epoch": 6.86, "learning_rate": 9.431839847473785e-07, "loss": 1.9099, "step": 36010 }, { "epoch": 6.86, "learning_rate": 9.431267874165873e-07, "loss": 2.5505, "step": 36011 }, { "epoch": 6.86, "learning_rate": 9.43069590085796e-07, "loss": 2.2623, "step": 36012 }, { "epoch": 6.86, "learning_rate": 9.430123927550048e-07, "loss": 0.6788, "step": 36013 }, { "epoch": 6.86, "learning_rate": 9.429551954242136e-07, "loss": 1.6942, "step": 36014 }, { "epoch": 6.86, "learning_rate": 9.428979980934223e-07, "loss": 2.4115, "step": 36015 }, { "epoch": 6.86, "learning_rate": 9.428408007626311e-07, "loss": 1.0979, "step": 36016 }, { "epoch": 6.86, "learning_rate": 9.427836034318399e-07, "loss": 1.6039, "step": 36017 }, { "epoch": 6.86, "learning_rate": 9.427264061010487e-07, "loss": 1.6362, "step": 36018 }, { "epoch": 6.86, "learning_rate": 9.426692087702573e-07, "loss": 0.8698, "step": 36019 }, { "epoch": 6.86, "learning_rate": 9.426120114394661e-07, "loss": 0.7776, "step": 36020 }, { "epoch": 6.86, "learning_rate": 9.42554814108675e-07, "loss": 1.8443, "step": 36021 }, { "epoch": 6.86, "learning_rate": 9.424976167778836e-07, "loss": 1.6517, "step": 36022 }, { "epoch": 6.86, "learning_rate": 9.424404194470925e-07, "loss": 1.3541, "step": 36023 }, { "epoch": 6.86, "learning_rate": 9.423832221163013e-07, "loss": 0.9759, "step": 36024 }, { "epoch": 6.86, "learning_rate": 9.423260247855101e-07, "loss": 1.0386, "step": 36025 }, { "epoch": 6.86, "learning_rate": 9.422688274547188e-07, "loss": 1.8498, "step": 36026 }, { "epoch": 6.86, "learning_rate": 9.422116301239276e-07, "loss": 2.1023, "step": 36027 }, { "epoch": 6.86, "learning_rate": 9.421544327931364e-07, "loss": 1.2448, "step": 36028 }, { "epoch": 6.86, "learning_rate": 9.420972354623451e-07, "loss": 1.8764, "step": 36029 }, { "epoch": 6.86, "learning_rate": 9.420400381315539e-07, "loss": 0.9411, "step": 36030 }, { "epoch": 6.86, "learning_rate": 9.419828408007627e-07, "loss": 2.9082, "step": 36031 }, { "epoch": 6.86, "learning_rate": 9.419256434699713e-07, "loss": 1.9303, "step": 36032 }, { "epoch": 6.86, "learning_rate": 9.418684461391801e-07, "loss": 2.4083, "step": 36033 }, { "epoch": 6.86, "learning_rate": 9.418112488083889e-07, "loss": 1.0413, "step": 36034 }, { "epoch": 6.86, "learning_rate": 9.417540514775978e-07, "loss": 2.7024, "step": 36035 }, { "epoch": 6.86, "learning_rate": 9.416968541468064e-07, "loss": 1.9601, "step": 36036 }, { "epoch": 6.86, "learning_rate": 9.416396568160153e-07, "loss": 0.3754, "step": 36037 }, { "epoch": 6.86, "learning_rate": 9.415824594852241e-07, "loss": 2.5444, "step": 36038 }, { "epoch": 6.86, "learning_rate": 9.415252621544327e-07, "loss": 1.3878, "step": 36039 }, { "epoch": 6.86, "learning_rate": 9.414680648236416e-07, "loss": 1.4939, "step": 36040 }, { "epoch": 6.86, "learning_rate": 9.414108674928504e-07, "loss": 3.7558, "step": 36041 }, { "epoch": 6.87, "learning_rate": 9.413536701620592e-07, "loss": 0.7569, "step": 36042 }, { "epoch": 6.87, "learning_rate": 9.412964728312679e-07, "loss": 1.7729, "step": 36043 }, { "epoch": 6.87, "learning_rate": 9.412392755004767e-07, "loss": 1.4366, "step": 36044 }, { "epoch": 6.87, "learning_rate": 9.411820781696855e-07, "loss": 1.8554, "step": 36045 }, { "epoch": 6.87, "learning_rate": 9.411248808388941e-07, "loss": 1.2744, "step": 36046 }, { "epoch": 6.87, "learning_rate": 9.410676835081029e-07, "loss": 1.4854, "step": 36047 }, { "epoch": 6.87, "learning_rate": 9.410104861773117e-07, "loss": 1.6813, "step": 36048 }, { "epoch": 6.87, "learning_rate": 9.409532888465206e-07, "loss": 3.6772, "step": 36049 }, { "epoch": 6.87, "learning_rate": 9.408960915157292e-07, "loss": 2.1405, "step": 36050 }, { "epoch": 6.87, "learning_rate": 9.408388941849381e-07, "loss": 2.2535, "step": 36051 }, { "epoch": 6.87, "learning_rate": 9.407816968541469e-07, "loss": 2.5859, "step": 36052 }, { "epoch": 6.87, "learning_rate": 9.407244995233555e-07, "loss": 1.4933, "step": 36053 }, { "epoch": 6.87, "learning_rate": 9.406673021925644e-07, "loss": 1.206, "step": 36054 }, { "epoch": 6.87, "learning_rate": 9.406101048617732e-07, "loss": 2.0272, "step": 36055 }, { "epoch": 6.87, "learning_rate": 9.405529075309819e-07, "loss": 1.3796, "step": 36056 }, { "epoch": 6.87, "learning_rate": 9.404957102001907e-07, "loss": 1.4894, "step": 36057 }, { "epoch": 6.87, "learning_rate": 9.404385128693995e-07, "loss": 1.6938, "step": 36058 }, { "epoch": 6.87, "learning_rate": 9.403813155386083e-07, "loss": 2.1359, "step": 36059 }, { "epoch": 6.87, "learning_rate": 9.403241182078169e-07, "loss": 0.8377, "step": 36060 }, { "epoch": 6.87, "learning_rate": 9.402669208770257e-07, "loss": 2.4003, "step": 36061 }, { "epoch": 6.87, "learning_rate": 9.402097235462345e-07, "loss": 1.899, "step": 36062 }, { "epoch": 6.87, "learning_rate": 9.401525262154432e-07, "loss": 1.257, "step": 36063 }, { "epoch": 6.87, "learning_rate": 9.40095328884652e-07, "loss": 1.3992, "step": 36064 }, { "epoch": 6.87, "learning_rate": 9.400381315538609e-07, "loss": 2.1548, "step": 36065 }, { "epoch": 6.87, "learning_rate": 9.399809342230697e-07, "loss": 0.9653, "step": 36066 }, { "epoch": 6.87, "learning_rate": 9.399237368922784e-07, "loss": 1.3987, "step": 36067 }, { "epoch": 6.87, "learning_rate": 9.398665395614872e-07, "loss": 1.4174, "step": 36068 }, { "epoch": 6.87, "learning_rate": 9.39809342230696e-07, "loss": 2.6473, "step": 36069 }, { "epoch": 6.87, "learning_rate": 9.397521448999047e-07, "loss": 1.3692, "step": 36070 }, { "epoch": 6.87, "learning_rate": 9.396949475691135e-07, "loss": 1.5884, "step": 36071 }, { "epoch": 6.87, "learning_rate": 9.396377502383223e-07, "loss": 1.4603, "step": 36072 }, { "epoch": 6.87, "learning_rate": 9.39580552907531e-07, "loss": 1.6466, "step": 36073 }, { "epoch": 6.87, "learning_rate": 9.395233555767397e-07, "loss": 1.8563, "step": 36074 }, { "epoch": 6.87, "learning_rate": 9.394661582459485e-07, "loss": 0.7218, "step": 36075 }, { "epoch": 6.87, "learning_rate": 9.394089609151574e-07, "loss": 1.4687, "step": 36076 }, { "epoch": 6.87, "learning_rate": 9.39351763584366e-07, "loss": 1.0735, "step": 36077 }, { "epoch": 6.87, "learning_rate": 9.392945662535748e-07, "loss": 2.3935, "step": 36078 }, { "epoch": 6.87, "learning_rate": 9.392373689227837e-07, "loss": 1.4927, "step": 36079 }, { "epoch": 6.87, "learning_rate": 9.391801715919923e-07, "loss": 2.5579, "step": 36080 }, { "epoch": 6.87, "learning_rate": 9.391229742612012e-07, "loss": 1.6902, "step": 36081 }, { "epoch": 6.87, "learning_rate": 9.3906577693041e-07, "loss": 1.0754, "step": 36082 }, { "epoch": 6.87, "learning_rate": 9.390085795996188e-07, "loss": 0.9619, "step": 36083 }, { "epoch": 6.87, "learning_rate": 9.389513822688275e-07, "loss": 3.0845, "step": 36084 }, { "epoch": 6.87, "learning_rate": 9.388941849380363e-07, "loss": 2.2429, "step": 36085 }, { "epoch": 6.87, "learning_rate": 9.38836987607245e-07, "loss": 0.5184, "step": 36086 }, { "epoch": 6.87, "learning_rate": 9.387797902764537e-07, "loss": 1.7839, "step": 36087 }, { "epoch": 6.87, "learning_rate": 9.387225929456625e-07, "loss": 1.3009, "step": 36088 }, { "epoch": 6.87, "learning_rate": 9.386653956148713e-07, "loss": 2.193, "step": 36089 }, { "epoch": 6.87, "learning_rate": 9.386081982840802e-07, "loss": 2.2701, "step": 36090 }, { "epoch": 6.87, "learning_rate": 9.385510009532888e-07, "loss": 1.5903, "step": 36091 }, { "epoch": 6.87, "learning_rate": 9.384938036224976e-07, "loss": 1.5687, "step": 36092 }, { "epoch": 6.87, "learning_rate": 9.384366062917065e-07, "loss": 2.0121, "step": 36093 }, { "epoch": 6.88, "learning_rate": 9.383794089609151e-07, "loss": 1.4029, "step": 36094 }, { "epoch": 6.88, "learning_rate": 9.38322211630124e-07, "loss": 1.6768, "step": 36095 }, { "epoch": 6.88, "learning_rate": 9.382650142993328e-07, "loss": 2.0326, "step": 36096 }, { "epoch": 6.88, "learning_rate": 9.382078169685414e-07, "loss": 0.7086, "step": 36097 }, { "epoch": 6.88, "learning_rate": 9.381506196377503e-07, "loss": 1.7137, "step": 36098 }, { "epoch": 6.88, "learning_rate": 9.38093422306959e-07, "loss": 2.1998, "step": 36099 }, { "epoch": 6.88, "learning_rate": 9.380362249761678e-07, "loss": 2.3262, "step": 36100 }, { "epoch": 6.88, "learning_rate": 9.379790276453765e-07, "loss": 1.8414, "step": 36101 }, { "epoch": 6.88, "learning_rate": 9.379218303145853e-07, "loss": 3.2945, "step": 36102 }, { "epoch": 6.88, "learning_rate": 9.378646329837941e-07, "loss": 2.9669, "step": 36103 }, { "epoch": 6.88, "learning_rate": 9.378074356530028e-07, "loss": 2.3755, "step": 36104 }, { "epoch": 6.88, "learning_rate": 9.377502383222116e-07, "loss": 2.7809, "step": 36105 }, { "epoch": 6.88, "learning_rate": 9.376930409914204e-07, "loss": 1.457, "step": 36106 }, { "epoch": 6.88, "learning_rate": 9.376358436606293e-07, "loss": 1.8209, "step": 36107 }, { "epoch": 6.88, "learning_rate": 9.375786463298379e-07, "loss": 1.9585, "step": 36108 }, { "epoch": 6.88, "learning_rate": 9.375214489990468e-07, "loss": 0.9198, "step": 36109 }, { "epoch": 6.88, "learning_rate": 9.374642516682556e-07, "loss": 0.7966, "step": 36110 }, { "epoch": 6.88, "learning_rate": 9.374070543374643e-07, "loss": 1.718, "step": 36111 }, { "epoch": 6.88, "learning_rate": 9.373498570066731e-07, "loss": 2.0702, "step": 36112 }, { "epoch": 6.88, "learning_rate": 9.372926596758818e-07, "loss": 1.5443, "step": 36113 }, { "epoch": 6.88, "learning_rate": 9.372354623450906e-07, "loss": 1.4657, "step": 36114 }, { "epoch": 6.88, "learning_rate": 9.371782650142993e-07, "loss": 1.7716, "step": 36115 }, { "epoch": 6.88, "learning_rate": 9.371210676835081e-07, "loss": 2.4942, "step": 36116 }, { "epoch": 6.88, "learning_rate": 9.370638703527169e-07, "loss": 2.045, "step": 36117 }, { "epoch": 6.88, "learning_rate": 9.370066730219256e-07, "loss": 2.1726, "step": 36118 }, { "epoch": 6.88, "learning_rate": 9.369494756911344e-07, "loss": 1.0498, "step": 36119 }, { "epoch": 6.88, "learning_rate": 9.368922783603432e-07, "loss": 0.3364, "step": 36120 }, { "epoch": 6.88, "learning_rate": 9.368350810295519e-07, "loss": 1.3526, "step": 36121 }, { "epoch": 6.88, "learning_rate": 9.367778836987607e-07, "loss": 1.6291, "step": 36122 }, { "epoch": 6.88, "learning_rate": 9.367206863679696e-07, "loss": 3.5771, "step": 36123 }, { "epoch": 6.88, "learning_rate": 9.366634890371784e-07, "loss": 2.0222, "step": 36124 }, { "epoch": 6.88, "learning_rate": 9.366062917063871e-07, "loss": 3.8335, "step": 36125 }, { "epoch": 6.88, "learning_rate": 9.365490943755959e-07, "loss": 1.988, "step": 36126 }, { "epoch": 6.88, "learning_rate": 9.364918970448046e-07, "loss": 1.3665, "step": 36127 }, { "epoch": 6.88, "learning_rate": 9.364346997140133e-07, "loss": 2.1172, "step": 36128 }, { "epoch": 6.88, "learning_rate": 9.363775023832221e-07, "loss": 1.5853, "step": 36129 }, { "epoch": 6.88, "learning_rate": 9.363203050524309e-07, "loss": 1.9994, "step": 36130 }, { "epoch": 6.88, "learning_rate": 9.362631077216397e-07, "loss": 1.9088, "step": 36131 }, { "epoch": 6.88, "learning_rate": 9.362059103908484e-07, "loss": 2.0163, "step": 36132 }, { "epoch": 6.88, "learning_rate": 9.361487130600572e-07, "loss": 1.9526, "step": 36133 }, { "epoch": 6.88, "learning_rate": 9.360915157292661e-07, "loss": 1.4162, "step": 36134 }, { "epoch": 6.88, "learning_rate": 9.360343183984747e-07, "loss": 0.8445, "step": 36135 }, { "epoch": 6.88, "learning_rate": 9.359771210676835e-07, "loss": 1.3311, "step": 36136 }, { "epoch": 6.88, "learning_rate": 9.359199237368924e-07, "loss": 1.791, "step": 36137 }, { "epoch": 6.88, "learning_rate": 9.358627264061012e-07, "loss": 1.5605, "step": 36138 }, { "epoch": 6.88, "learning_rate": 9.358055290753099e-07, "loss": 1.0251, "step": 36139 }, { "epoch": 6.88, "learning_rate": 9.357483317445186e-07, "loss": 1.9547, "step": 36140 }, { "epoch": 6.88, "learning_rate": 9.356911344137274e-07, "loss": 1.4278, "step": 36141 }, { "epoch": 6.88, "learning_rate": 9.356339370829361e-07, "loss": 1.6461, "step": 36142 }, { "epoch": 6.88, "learning_rate": 9.355767397521449e-07, "loss": 0.8889, "step": 36143 }, { "epoch": 6.88, "learning_rate": 9.355195424213537e-07, "loss": 1.4587, "step": 36144 }, { "epoch": 6.88, "learning_rate": 9.354623450905624e-07, "loss": 1.5874, "step": 36145 }, { "epoch": 6.88, "learning_rate": 9.354051477597712e-07, "loss": 1.2296, "step": 36146 }, { "epoch": 6.89, "learning_rate": 9.3534795042898e-07, "loss": 2.7734, "step": 36147 }, { "epoch": 6.89, "learning_rate": 9.352907530981889e-07, "loss": 1.2508, "step": 36148 }, { "epoch": 6.89, "learning_rate": 9.352335557673975e-07, "loss": 1.7159, "step": 36149 }, { "epoch": 6.89, "learning_rate": 9.351763584366063e-07, "loss": 1.3109, "step": 36150 }, { "epoch": 6.89, "learning_rate": 9.351191611058152e-07, "loss": 2.2676, "step": 36151 }, { "epoch": 6.89, "learning_rate": 9.350619637750237e-07, "loss": 1.5621, "step": 36152 }, { "epoch": 6.89, "learning_rate": 9.350047664442326e-07, "loss": 1.4825, "step": 36153 }, { "epoch": 6.89, "learning_rate": 9.349475691134414e-07, "loss": 1.8642, "step": 36154 }, { "epoch": 6.89, "learning_rate": 9.348903717826502e-07, "loss": 2.3624, "step": 36155 }, { "epoch": 6.89, "learning_rate": 9.348331744518589e-07, "loss": 1.0733, "step": 36156 }, { "epoch": 6.89, "learning_rate": 9.347759771210677e-07, "loss": 1.5746, "step": 36157 }, { "epoch": 6.89, "learning_rate": 9.347187797902765e-07, "loss": 1.4682, "step": 36158 }, { "epoch": 6.89, "learning_rate": 9.346615824594852e-07, "loss": 1.7165, "step": 36159 }, { "epoch": 6.89, "learning_rate": 9.34604385128694e-07, "loss": 0.828, "step": 36160 }, { "epoch": 6.89, "learning_rate": 9.345471877979028e-07, "loss": 2.1828, "step": 36161 }, { "epoch": 6.89, "learning_rate": 9.344899904671115e-07, "loss": 1.5019, "step": 36162 }, { "epoch": 6.89, "learning_rate": 9.344327931363203e-07, "loss": 3.1195, "step": 36163 }, { "epoch": 6.89, "learning_rate": 9.343755958055291e-07, "loss": 1.9912, "step": 36164 }, { "epoch": 6.89, "learning_rate": 9.34318398474738e-07, "loss": 1.1325, "step": 36165 }, { "epoch": 6.89, "learning_rate": 9.342612011439465e-07, "loss": 1.2548, "step": 36166 }, { "epoch": 6.89, "learning_rate": 9.342040038131554e-07, "loss": 2.0987, "step": 36167 }, { "epoch": 6.89, "learning_rate": 9.341468064823642e-07, "loss": 1.0895, "step": 36168 }, { "epoch": 6.89, "learning_rate": 9.340896091515729e-07, "loss": 1.6988, "step": 36169 }, { "epoch": 6.89, "learning_rate": 9.340324118207817e-07, "loss": 1.5422, "step": 36170 }, { "epoch": 6.89, "learning_rate": 9.339752144899905e-07, "loss": 1.9872, "step": 36171 }, { "epoch": 6.89, "learning_rate": 9.339180171591993e-07, "loss": 1.5267, "step": 36172 }, { "epoch": 6.89, "learning_rate": 9.33860819828408e-07, "loss": 1.7533, "step": 36173 }, { "epoch": 6.89, "learning_rate": 9.338036224976168e-07, "loss": 0.6028, "step": 36174 }, { "epoch": 6.89, "learning_rate": 9.337464251668256e-07, "loss": 1.7853, "step": 36175 }, { "epoch": 6.89, "learning_rate": 9.336892278360343e-07, "loss": 2.8094, "step": 36176 }, { "epoch": 6.89, "learning_rate": 9.336320305052431e-07, "loss": 1.751, "step": 36177 }, { "epoch": 6.89, "learning_rate": 9.33574833174452e-07, "loss": 2.4022, "step": 36178 }, { "epoch": 6.89, "learning_rate": 9.335176358436608e-07, "loss": 0.918, "step": 36179 }, { "epoch": 6.89, "learning_rate": 9.334604385128693e-07, "loss": 0.4723, "step": 36180 }, { "epoch": 6.89, "learning_rate": 9.334032411820782e-07, "loss": 1.7212, "step": 36181 }, { "epoch": 6.89, "learning_rate": 9.33346043851287e-07, "loss": 1.5381, "step": 36182 }, { "epoch": 6.89, "learning_rate": 9.332888465204957e-07, "loss": 1.8865, "step": 36183 }, { "epoch": 6.89, "learning_rate": 9.332316491897045e-07, "loss": 1.0634, "step": 36184 }, { "epoch": 6.89, "learning_rate": 9.331744518589133e-07, "loss": 1.1789, "step": 36185 }, { "epoch": 6.89, "learning_rate": 9.33117254528122e-07, "loss": 1.7512, "step": 36186 }, { "epoch": 6.89, "learning_rate": 9.330600571973308e-07, "loss": 1.14, "step": 36187 }, { "epoch": 6.89, "learning_rate": 9.330028598665396e-07, "loss": 2.6822, "step": 36188 }, { "epoch": 6.89, "learning_rate": 9.329456625357484e-07, "loss": 1.5933, "step": 36189 }, { "epoch": 6.89, "learning_rate": 9.328884652049571e-07, "loss": 2.0096, "step": 36190 }, { "epoch": 6.89, "learning_rate": 9.328312678741659e-07, "loss": 1.3083, "step": 36191 }, { "epoch": 6.89, "learning_rate": 9.327740705433748e-07, "loss": 1.353, "step": 36192 }, { "epoch": 6.89, "learning_rate": 9.327168732125833e-07, "loss": 1.8262, "step": 36193 }, { "epoch": 6.89, "learning_rate": 9.326596758817921e-07, "loss": 2.2136, "step": 36194 }, { "epoch": 6.89, "learning_rate": 9.32602478551001e-07, "loss": 3.1754, "step": 36195 }, { "epoch": 6.89, "learning_rate": 9.325452812202098e-07, "loss": 2.8017, "step": 36196 }, { "epoch": 6.89, "learning_rate": 9.324880838894185e-07, "loss": 1.6374, "step": 36197 }, { "epoch": 6.89, "learning_rate": 9.324308865586273e-07, "loss": 1.2492, "step": 36198 }, { "epoch": 6.9, "learning_rate": 9.323736892278361e-07, "loss": 1.0704, "step": 36199 }, { "epoch": 6.9, "learning_rate": 9.323164918970448e-07, "loss": 1.4004, "step": 36200 }, { "epoch": 6.9, "learning_rate": 9.322592945662536e-07, "loss": 1.3051, "step": 36201 }, { "epoch": 6.9, "learning_rate": 9.322020972354624e-07, "loss": 1.2476, "step": 36202 }, { "epoch": 6.9, "learning_rate": 9.321448999046712e-07, "loss": 1.6377, "step": 36203 }, { "epoch": 6.9, "learning_rate": 9.320877025738799e-07, "loss": 0.5438, "step": 36204 }, { "epoch": 6.9, "learning_rate": 9.320305052430887e-07, "loss": 2.0283, "step": 36205 }, { "epoch": 6.9, "learning_rate": 9.319733079122976e-07, "loss": 1.453, "step": 36206 }, { "epoch": 6.9, "learning_rate": 9.319161105815061e-07, "loss": 2.3426, "step": 36207 }, { "epoch": 6.9, "learning_rate": 9.318589132507149e-07, "loss": 2.353, "step": 36208 }, { "epoch": 6.9, "learning_rate": 9.318017159199238e-07, "loss": 1.4137, "step": 36209 }, { "epoch": 6.9, "learning_rate": 9.317445185891324e-07, "loss": 1.9282, "step": 36210 }, { "epoch": 6.9, "learning_rate": 9.316873212583413e-07, "loss": 2.6274, "step": 36211 }, { "epoch": 6.9, "learning_rate": 9.316301239275501e-07, "loss": 0.3303, "step": 36212 }, { "epoch": 6.9, "learning_rate": 9.315729265967589e-07, "loss": 1.5019, "step": 36213 }, { "epoch": 6.9, "learning_rate": 9.315157292659676e-07, "loss": 1.3818, "step": 36214 }, { "epoch": 6.9, "learning_rate": 9.314585319351764e-07, "loss": 1.0408, "step": 36215 }, { "epoch": 6.9, "learning_rate": 9.314013346043852e-07, "loss": 1.8086, "step": 36216 }, { "epoch": 6.9, "learning_rate": 9.313441372735939e-07, "loss": 2.4799, "step": 36217 }, { "epoch": 6.9, "learning_rate": 9.312869399428027e-07, "loss": 1.637, "step": 36218 }, { "epoch": 6.9, "learning_rate": 9.312297426120115e-07, "loss": 1.663, "step": 36219 }, { "epoch": 6.9, "learning_rate": 9.311725452812204e-07, "loss": 1.6543, "step": 36220 }, { "epoch": 6.9, "learning_rate": 9.311153479504289e-07, "loss": 2.5077, "step": 36221 }, { "epoch": 6.9, "learning_rate": 9.310581506196378e-07, "loss": 1.1906, "step": 36222 }, { "epoch": 6.9, "learning_rate": 9.310009532888466e-07, "loss": 1.4853, "step": 36223 }, { "epoch": 6.9, "learning_rate": 9.309437559580552e-07, "loss": 1.4572, "step": 36224 }, { "epoch": 6.9, "learning_rate": 9.308865586272641e-07, "loss": 1.7197, "step": 36225 }, { "epoch": 6.9, "learning_rate": 9.308293612964729e-07, "loss": 0.4332, "step": 36226 }, { "epoch": 6.9, "learning_rate": 9.307721639656816e-07, "loss": 1.8425, "step": 36227 }, { "epoch": 6.9, "learning_rate": 9.307149666348904e-07, "loss": 1.5534, "step": 36228 }, { "epoch": 6.9, "learning_rate": 9.306577693040992e-07, "loss": 2.2371, "step": 36229 }, { "epoch": 6.9, "learning_rate": 9.30600571973308e-07, "loss": 2.156, "step": 36230 }, { "epoch": 6.9, "learning_rate": 9.305433746425167e-07, "loss": 1.4857, "step": 36231 }, { "epoch": 6.9, "learning_rate": 9.304861773117255e-07, "loss": 1.2601, "step": 36232 }, { "epoch": 6.9, "learning_rate": 9.304289799809342e-07, "loss": 1.2017, "step": 36233 }, { "epoch": 6.9, "learning_rate": 9.303717826501429e-07, "loss": 1.0446, "step": 36234 }, { "epoch": 6.9, "learning_rate": 9.303145853193517e-07, "loss": 2.1787, "step": 36235 }, { "epoch": 6.9, "learning_rate": 9.302573879885606e-07, "loss": 1.2538, "step": 36236 }, { "epoch": 6.9, "learning_rate": 9.302001906577694e-07, "loss": 3.5553, "step": 36237 }, { "epoch": 6.9, "learning_rate": 9.30142993326978e-07, "loss": 0.5906, "step": 36238 }, { "epoch": 6.9, "learning_rate": 9.300857959961869e-07, "loss": 0.8913, "step": 36239 }, { "epoch": 6.9, "learning_rate": 9.300285986653957e-07, "loss": 1.1907, "step": 36240 }, { "epoch": 6.9, "learning_rate": 9.299714013346044e-07, "loss": 1.6347, "step": 36241 }, { "epoch": 6.9, "learning_rate": 9.299142040038132e-07, "loss": 1.3156, "step": 36242 }, { "epoch": 6.9, "learning_rate": 9.29857006673022e-07, "loss": 1.4118, "step": 36243 }, { "epoch": 6.9, "learning_rate": 9.297998093422308e-07, "loss": 2.1574, "step": 36244 }, { "epoch": 6.9, "learning_rate": 9.297426120114395e-07, "loss": 2.7728, "step": 36245 }, { "epoch": 6.9, "learning_rate": 9.296854146806482e-07, "loss": 1.5764, "step": 36246 }, { "epoch": 6.9, "learning_rate": 9.29628217349857e-07, "loss": 1.6346, "step": 36247 }, { "epoch": 6.9, "learning_rate": 9.295710200190657e-07, "loss": 1.5356, "step": 36248 }, { "epoch": 6.9, "learning_rate": 9.295138226882745e-07, "loss": 2.0682, "step": 36249 }, { "epoch": 6.9, "learning_rate": 9.294566253574834e-07, "loss": 1.2371, "step": 36250 }, { "epoch": 6.9, "learning_rate": 9.29399428026692e-07, "loss": 1.9133, "step": 36251 }, { "epoch": 6.91, "learning_rate": 9.293422306959008e-07, "loss": 1.3645, "step": 36252 }, { "epoch": 6.91, "learning_rate": 9.292850333651097e-07, "loss": 2.2844, "step": 36253 }, { "epoch": 6.91, "learning_rate": 9.292278360343185e-07, "loss": 1.1818, "step": 36254 }, { "epoch": 6.91, "learning_rate": 9.291706387035272e-07, "loss": 1.2736, "step": 36255 }, { "epoch": 6.91, "learning_rate": 9.29113441372736e-07, "loss": 1.1194, "step": 36256 }, { "epoch": 6.91, "learning_rate": 9.290562440419448e-07, "loss": 1.5855, "step": 36257 }, { "epoch": 6.91, "learning_rate": 9.289990467111535e-07, "loss": 0.7615, "step": 36258 }, { "epoch": 6.91, "learning_rate": 9.289418493803623e-07, "loss": 2.489, "step": 36259 }, { "epoch": 6.91, "learning_rate": 9.28884652049571e-07, "loss": 2.7533, "step": 36260 }, { "epoch": 6.91, "learning_rate": 9.288274547187798e-07, "loss": 1.1963, "step": 36261 }, { "epoch": 6.91, "learning_rate": 9.287702573879885e-07, "loss": 2.1828, "step": 36262 }, { "epoch": 6.91, "learning_rate": 9.287130600571973e-07, "loss": 1.232, "step": 36263 }, { "epoch": 6.91, "learning_rate": 9.286558627264062e-07, "loss": 2.351, "step": 36264 }, { "epoch": 6.91, "learning_rate": 9.285986653956148e-07, "loss": 2.5479, "step": 36265 }, { "epoch": 6.91, "learning_rate": 9.285414680648237e-07, "loss": 0.9223, "step": 36266 }, { "epoch": 6.91, "learning_rate": 9.284842707340325e-07, "loss": 1.2383, "step": 36267 }, { "epoch": 6.91, "learning_rate": 9.284270734032413e-07, "loss": 0.7466, "step": 36268 }, { "epoch": 6.91, "learning_rate": 9.2836987607245e-07, "loss": 2.8364, "step": 36269 }, { "epoch": 6.91, "learning_rate": 9.283126787416588e-07, "loss": 0.9788, "step": 36270 }, { "epoch": 6.91, "learning_rate": 9.282554814108676e-07, "loss": 0.4613, "step": 36271 }, { "epoch": 6.91, "learning_rate": 9.281982840800763e-07, "loss": 1.1321, "step": 36272 }, { "epoch": 6.91, "learning_rate": 9.28141086749285e-07, "loss": 1.985, "step": 36273 }, { "epoch": 6.91, "learning_rate": 9.280838894184938e-07, "loss": 1.2328, "step": 36274 }, { "epoch": 6.91, "learning_rate": 9.280266920877025e-07, "loss": 1.282, "step": 36275 }, { "epoch": 6.91, "learning_rate": 9.279694947569113e-07, "loss": 0.7218, "step": 36276 }, { "epoch": 6.91, "learning_rate": 9.279122974261201e-07, "loss": 2.5829, "step": 36277 }, { "epoch": 6.91, "learning_rate": 9.27855100095329e-07, "loss": 2.3415, "step": 36278 }, { "epoch": 6.91, "learning_rate": 9.277979027645376e-07, "loss": 1.9994, "step": 36279 }, { "epoch": 6.91, "learning_rate": 9.277407054337465e-07, "loss": 1.8659, "step": 36280 }, { "epoch": 6.91, "learning_rate": 9.276835081029553e-07, "loss": 1.8833, "step": 36281 }, { "epoch": 6.91, "learning_rate": 9.276263107721639e-07, "loss": 2.8675, "step": 36282 }, { "epoch": 6.91, "learning_rate": 9.275691134413728e-07, "loss": 2.0922, "step": 36283 }, { "epoch": 6.91, "learning_rate": 9.275119161105816e-07, "loss": 1.9048, "step": 36284 }, { "epoch": 6.91, "learning_rate": 9.274547187797904e-07, "loss": 1.2276, "step": 36285 }, { "epoch": 6.91, "learning_rate": 9.27397521448999e-07, "loss": 1.6208, "step": 36286 }, { "epoch": 6.91, "learning_rate": 9.273403241182078e-07, "loss": 0.4735, "step": 36287 }, { "epoch": 6.91, "learning_rate": 9.272831267874166e-07, "loss": 2.0763, "step": 36288 }, { "epoch": 6.91, "learning_rate": 9.272259294566253e-07, "loss": 0.6464, "step": 36289 }, { "epoch": 6.91, "learning_rate": 9.271687321258341e-07, "loss": 2.4988, "step": 36290 }, { "epoch": 6.91, "learning_rate": 9.271115347950429e-07, "loss": 2.0487, "step": 36291 }, { "epoch": 6.91, "learning_rate": 9.270543374642516e-07, "loss": 2.2647, "step": 36292 }, { "epoch": 6.91, "learning_rate": 9.269971401334604e-07, "loss": 1.2031, "step": 36293 }, { "epoch": 6.91, "learning_rate": 9.269399428026693e-07, "loss": 3.552, "step": 36294 }, { "epoch": 6.91, "learning_rate": 9.268827454718781e-07, "loss": 1.7791, "step": 36295 }, { "epoch": 6.91, "learning_rate": 9.268255481410867e-07, "loss": 2.3125, "step": 36296 }, { "epoch": 6.91, "learning_rate": 9.267683508102956e-07, "loss": 1.5097, "step": 36297 }, { "epoch": 6.91, "learning_rate": 9.267111534795044e-07, "loss": 3.7909, "step": 36298 }, { "epoch": 6.91, "learning_rate": 9.26653956148713e-07, "loss": 2.1662, "step": 36299 }, { "epoch": 6.91, "learning_rate": 9.265967588179218e-07, "loss": 0.5129, "step": 36300 }, { "epoch": 6.91, "learning_rate": 9.265395614871306e-07, "loss": 1.5878, "step": 36301 }, { "epoch": 6.91, "learning_rate": 9.264823641563394e-07, "loss": 2.7116, "step": 36302 }, { "epoch": 6.91, "learning_rate": 9.264251668255481e-07, "loss": 1.9409, "step": 36303 }, { "epoch": 6.92, "learning_rate": 9.263679694947569e-07, "loss": 1.9686, "step": 36304 }, { "epoch": 6.92, "learning_rate": 9.263107721639657e-07, "loss": 2.0561, "step": 36305 }, { "epoch": 6.92, "learning_rate": 9.262535748331744e-07, "loss": 1.3356, "step": 36306 }, { "epoch": 6.92, "learning_rate": 9.261963775023832e-07, "loss": 1.6339, "step": 36307 }, { "epoch": 6.92, "learning_rate": 9.261391801715921e-07, "loss": 1.2373, "step": 36308 }, { "epoch": 6.92, "learning_rate": 9.260819828408009e-07, "loss": 1.5514, "step": 36309 }, { "epoch": 6.92, "learning_rate": 9.260247855100096e-07, "loss": 0.8844, "step": 36310 }, { "epoch": 6.92, "learning_rate": 9.259675881792184e-07, "loss": 1.985, "step": 36311 }, { "epoch": 6.92, "learning_rate": 9.259103908484272e-07, "loss": 1.7699, "step": 36312 }, { "epoch": 6.92, "learning_rate": 9.258531935176358e-07, "loss": 0.9732, "step": 36313 }, { "epoch": 6.92, "learning_rate": 9.257959961868446e-07, "loss": 1.6571, "step": 36314 }, { "epoch": 6.92, "learning_rate": 9.257387988560534e-07, "loss": 2.5284, "step": 36315 }, { "epoch": 6.92, "learning_rate": 9.256816015252621e-07, "loss": 1.2313, "step": 36316 }, { "epoch": 6.92, "learning_rate": 9.256244041944709e-07, "loss": 2.3694, "step": 36317 }, { "epoch": 6.92, "learning_rate": 9.255672068636797e-07, "loss": 1.4385, "step": 36318 }, { "epoch": 6.92, "learning_rate": 9.255100095328885e-07, "loss": 2.5526, "step": 36319 }, { "epoch": 6.92, "learning_rate": 9.254528122020972e-07, "loss": 2.2605, "step": 36320 }, { "epoch": 6.92, "learning_rate": 9.25395614871306e-07, "loss": 1.2624, "step": 36321 }, { "epoch": 6.92, "learning_rate": 9.253384175405149e-07, "loss": 0.3209, "step": 36322 }, { "epoch": 6.92, "learning_rate": 9.252812202097235e-07, "loss": 1.7864, "step": 36323 }, { "epoch": 6.92, "learning_rate": 9.252240228789324e-07, "loss": 1.8894, "step": 36324 }, { "epoch": 6.92, "learning_rate": 9.251668255481412e-07, "loss": 3.1987, "step": 36325 }, { "epoch": 6.92, "learning_rate": 9.2510962821735e-07, "loss": 1.4683, "step": 36326 }, { "epoch": 6.92, "learning_rate": 9.250524308865586e-07, "loss": 1.9378, "step": 36327 }, { "epoch": 6.92, "learning_rate": 9.249952335557674e-07, "loss": 1.1613, "step": 36328 }, { "epoch": 6.92, "learning_rate": 9.249380362249762e-07, "loss": 1.0031, "step": 36329 }, { "epoch": 6.92, "learning_rate": 9.248808388941849e-07, "loss": 0.7517, "step": 36330 }, { "epoch": 6.92, "learning_rate": 9.248236415633937e-07, "loss": 0.8605, "step": 36331 }, { "epoch": 6.92, "learning_rate": 9.247664442326025e-07, "loss": 2.251, "step": 36332 }, { "epoch": 6.92, "learning_rate": 9.247092469018112e-07, "loss": 1.4575, "step": 36333 }, { "epoch": 6.92, "learning_rate": 9.2465204957102e-07, "loss": 1.4069, "step": 36334 }, { "epoch": 6.92, "learning_rate": 9.245948522402288e-07, "loss": 1.187, "step": 36335 }, { "epoch": 6.92, "learning_rate": 9.245376549094377e-07, "loss": 1.1163, "step": 36336 }, { "epoch": 6.92, "learning_rate": 9.244804575786463e-07, "loss": 3.4424, "step": 36337 }, { "epoch": 6.92, "learning_rate": 9.244232602478552e-07, "loss": 1.9025, "step": 36338 }, { "epoch": 6.92, "learning_rate": 9.24366062917064e-07, "loss": 1.8234, "step": 36339 }, { "epoch": 6.92, "learning_rate": 9.243088655862725e-07, "loss": 0.3901, "step": 36340 }, { "epoch": 6.92, "learning_rate": 9.242516682554814e-07, "loss": 1.665, "step": 36341 }, { "epoch": 6.92, "learning_rate": 9.241944709246902e-07, "loss": 2.0716, "step": 36342 }, { "epoch": 6.92, "learning_rate": 9.24137273593899e-07, "loss": 1.4516, "step": 36343 }, { "epoch": 6.92, "learning_rate": 9.240800762631077e-07, "loss": 1.5412, "step": 36344 }, { "epoch": 6.92, "learning_rate": 9.240228789323165e-07, "loss": 2.1983, "step": 36345 }, { "epoch": 6.92, "learning_rate": 9.239656816015253e-07, "loss": 1.3049, "step": 36346 }, { "epoch": 6.92, "learning_rate": 9.23908484270734e-07, "loss": 2.5706, "step": 36347 }, { "epoch": 6.92, "learning_rate": 9.238512869399428e-07, "loss": 2.5959, "step": 36348 }, { "epoch": 6.92, "learning_rate": 9.237940896091516e-07, "loss": 1.3465, "step": 36349 }, { "epoch": 6.92, "learning_rate": 9.237368922783605e-07, "loss": 1.8144, "step": 36350 }, { "epoch": 6.92, "learning_rate": 9.236796949475691e-07, "loss": 1.1358, "step": 36351 }, { "epoch": 6.92, "learning_rate": 9.23622497616778e-07, "loss": 1.3064, "step": 36352 }, { "epoch": 6.92, "learning_rate": 9.235653002859868e-07, "loss": 1.1605, "step": 36353 }, { "epoch": 6.92, "learning_rate": 9.235081029551953e-07, "loss": 1.2468, "step": 36354 }, { "epoch": 6.92, "learning_rate": 9.234509056244042e-07, "loss": 1.1327, "step": 36355 }, { "epoch": 6.92, "learning_rate": 9.23393708293613e-07, "loss": 1.9315, "step": 36356 }, { "epoch": 6.93, "learning_rate": 9.233365109628217e-07, "loss": 1.7491, "step": 36357 }, { "epoch": 6.93, "learning_rate": 9.232793136320305e-07, "loss": 1.102, "step": 36358 }, { "epoch": 6.93, "learning_rate": 9.232221163012393e-07, "loss": 2.3819, "step": 36359 }, { "epoch": 6.93, "learning_rate": 9.231649189704481e-07, "loss": 2.202, "step": 36360 }, { "epoch": 6.93, "learning_rate": 9.231077216396568e-07, "loss": 1.0862, "step": 36361 }, { "epoch": 6.93, "learning_rate": 9.230505243088656e-07, "loss": 0.911, "step": 36362 }, { "epoch": 6.93, "learning_rate": 9.229933269780744e-07, "loss": 1.1133, "step": 36363 }, { "epoch": 6.93, "learning_rate": 9.229361296472831e-07, "loss": 3.3245, "step": 36364 }, { "epoch": 6.93, "learning_rate": 9.228789323164919e-07, "loss": 1.782, "step": 36365 }, { "epoch": 6.93, "learning_rate": 9.228217349857008e-07, "loss": 2.0804, "step": 36366 }, { "epoch": 6.93, "learning_rate": 9.227645376549095e-07, "loss": 1.3806, "step": 36367 }, { "epoch": 6.93, "learning_rate": 9.227073403241182e-07, "loss": 2.5867, "step": 36368 }, { "epoch": 6.93, "learning_rate": 9.22650142993327e-07, "loss": 2.2262, "step": 36369 }, { "epoch": 6.93, "learning_rate": 9.225929456625358e-07, "loss": 1.111, "step": 36370 }, { "epoch": 6.93, "learning_rate": 9.225357483317445e-07, "loss": 2.0861, "step": 36371 }, { "epoch": 6.93, "learning_rate": 9.224785510009533e-07, "loss": 1.0571, "step": 36372 }, { "epoch": 6.93, "learning_rate": 9.224213536701621e-07, "loss": 1.1905, "step": 36373 }, { "epoch": 6.93, "learning_rate": 9.223641563393709e-07, "loss": 0.9486, "step": 36374 }, { "epoch": 6.93, "learning_rate": 9.223069590085796e-07, "loss": 2.0566, "step": 36375 }, { "epoch": 6.93, "learning_rate": 9.222497616777884e-07, "loss": 0.4086, "step": 36376 }, { "epoch": 6.93, "learning_rate": 9.221925643469973e-07, "loss": 2.3823, "step": 36377 }, { "epoch": 6.93, "learning_rate": 9.221353670162059e-07, "loss": 2.865, "step": 36378 }, { "epoch": 6.93, "learning_rate": 9.220781696854147e-07, "loss": 1.2995, "step": 36379 }, { "epoch": 6.93, "learning_rate": 9.220209723546236e-07, "loss": 1.1422, "step": 36380 }, { "epoch": 6.93, "learning_rate": 9.219637750238321e-07, "loss": 1.6859, "step": 36381 }, { "epoch": 6.93, "learning_rate": 9.21906577693041e-07, "loss": 0.7677, "step": 36382 }, { "epoch": 6.93, "learning_rate": 9.218493803622498e-07, "loss": 1.0012, "step": 36383 }, { "epoch": 6.93, "learning_rate": 9.217921830314586e-07, "loss": 0.8854, "step": 36384 }, { "epoch": 6.93, "learning_rate": 9.217349857006673e-07, "loss": 1.6629, "step": 36385 }, { "epoch": 6.93, "learning_rate": 9.216777883698761e-07, "loss": 1.2228, "step": 36386 }, { "epoch": 6.93, "learning_rate": 9.216205910390849e-07, "loss": 2.0414, "step": 36387 }, { "epoch": 6.93, "learning_rate": 9.215633937082936e-07, "loss": 1.3767, "step": 36388 }, { "epoch": 6.93, "learning_rate": 9.215061963775024e-07, "loss": 1.9636, "step": 36389 }, { "epoch": 6.93, "learning_rate": 9.214489990467112e-07, "loss": 0.9795, "step": 36390 }, { "epoch": 6.93, "learning_rate": 9.213918017159201e-07, "loss": 2.1812, "step": 36391 }, { "epoch": 6.93, "learning_rate": 9.213346043851287e-07, "loss": 2.2484, "step": 36392 }, { "epoch": 6.93, "learning_rate": 9.212774070543374e-07, "loss": 1.0166, "step": 36393 }, { "epoch": 6.93, "learning_rate": 9.212202097235463e-07, "loss": 1.7056, "step": 36394 }, { "epoch": 6.93, "learning_rate": 9.211630123927549e-07, "loss": 1.578, "step": 36395 }, { "epoch": 6.93, "learning_rate": 9.211058150619638e-07, "loss": 1.5045, "step": 36396 }, { "epoch": 6.93, "learning_rate": 9.210486177311726e-07, "loss": 1.6673, "step": 36397 }, { "epoch": 6.93, "learning_rate": 9.209914204003812e-07, "loss": 2.1746, "step": 36398 }, { "epoch": 6.93, "learning_rate": 9.209342230695901e-07, "loss": 2.5264, "step": 36399 }, { "epoch": 6.93, "learning_rate": 9.208770257387989e-07, "loss": 0.6587, "step": 36400 }, { "epoch": 6.93, "learning_rate": 9.208198284080077e-07, "loss": 1.4765, "step": 36401 }, { "epoch": 6.93, "learning_rate": 9.207626310772164e-07, "loss": 2.0195, "step": 36402 }, { "epoch": 6.93, "learning_rate": 9.207054337464252e-07, "loss": 0.6169, "step": 36403 }, { "epoch": 6.93, "learning_rate": 9.20648236415634e-07, "loss": 1.2282, "step": 36404 }, { "epoch": 6.93, "learning_rate": 9.205910390848427e-07, "loss": 1.0888, "step": 36405 }, { "epoch": 6.93, "learning_rate": 9.205338417540514e-07, "loss": 1.8622, "step": 36406 }, { "epoch": 6.93, "learning_rate": 9.204766444232602e-07, "loss": 2.0725, "step": 36407 }, { "epoch": 6.93, "learning_rate": 9.204194470924691e-07, "loss": 0.8332, "step": 36408 }, { "epoch": 6.94, "learning_rate": 9.203622497616777e-07, "loss": 1.5473, "step": 36409 }, { "epoch": 6.94, "learning_rate": 9.203050524308866e-07, "loss": 1.7707, "step": 36410 }, { "epoch": 6.94, "learning_rate": 9.202478551000954e-07, "loss": 2.4009, "step": 36411 }, { "epoch": 6.94, "learning_rate": 9.201906577693041e-07, "loss": 2.9815, "step": 36412 }, { "epoch": 6.94, "learning_rate": 9.201334604385129e-07, "loss": 0.3347, "step": 36413 }, { "epoch": 6.94, "learning_rate": 9.200762631077217e-07, "loss": 1.2872, "step": 36414 }, { "epoch": 6.94, "learning_rate": 9.200190657769305e-07, "loss": 1.638, "step": 36415 }, { "epoch": 6.94, "learning_rate": 9.199618684461392e-07, "loss": 1.755, "step": 36416 }, { "epoch": 6.94, "learning_rate": 9.19904671115348e-07, "loss": 2.0479, "step": 36417 }, { "epoch": 6.94, "learning_rate": 9.198474737845568e-07, "loss": 2.1556, "step": 36418 }, { "epoch": 6.94, "learning_rate": 9.197902764537655e-07, "loss": 1.1228, "step": 36419 }, { "epoch": 6.94, "learning_rate": 9.197330791229742e-07, "loss": 1.1818, "step": 36420 }, { "epoch": 6.94, "learning_rate": 9.196758817921831e-07, "loss": 2.6521, "step": 36421 }, { "epoch": 6.94, "learning_rate": 9.196186844613917e-07, "loss": 2.1125, "step": 36422 }, { "epoch": 6.94, "learning_rate": 9.195614871306005e-07, "loss": 0.9297, "step": 36423 }, { "epoch": 6.94, "learning_rate": 9.195042897998094e-07, "loss": 1.887, "step": 36424 }, { "epoch": 6.94, "learning_rate": 9.194470924690182e-07, "loss": 3.8823, "step": 36425 }, { "epoch": 6.94, "learning_rate": 9.193898951382269e-07, "loss": 1.1134, "step": 36426 }, { "epoch": 6.94, "learning_rate": 9.193326978074357e-07, "loss": 2.6533, "step": 36427 }, { "epoch": 6.94, "learning_rate": 9.192755004766445e-07, "loss": 0.8327, "step": 36428 }, { "epoch": 6.94, "learning_rate": 9.192183031458532e-07, "loss": 0.8055, "step": 36429 }, { "epoch": 6.94, "learning_rate": 9.19161105815062e-07, "loss": 1.799, "step": 36430 }, { "epoch": 6.94, "learning_rate": 9.191039084842708e-07, "loss": 0.5148, "step": 36431 }, { "epoch": 6.94, "learning_rate": 9.190467111534796e-07, "loss": 1.5698, "step": 36432 }, { "epoch": 6.94, "learning_rate": 9.189895138226882e-07, "loss": 1.3428, "step": 36433 }, { "epoch": 6.94, "learning_rate": 9.18932316491897e-07, "loss": 1.2609, "step": 36434 }, { "epoch": 6.94, "learning_rate": 9.188751191611059e-07, "loss": 1.4671, "step": 36435 }, { "epoch": 6.94, "learning_rate": 9.188179218303145e-07, "loss": 1.7854, "step": 36436 }, { "epoch": 6.94, "learning_rate": 9.187607244995233e-07, "loss": 1.8765, "step": 36437 }, { "epoch": 6.94, "learning_rate": 9.187035271687322e-07, "loss": 1.455, "step": 36438 }, { "epoch": 6.94, "learning_rate": 9.18646329837941e-07, "loss": 1.3483, "step": 36439 }, { "epoch": 6.94, "learning_rate": 9.185891325071497e-07, "loss": 2.0046, "step": 36440 }, { "epoch": 6.94, "learning_rate": 9.185319351763585e-07, "loss": 1.2879, "step": 36441 }, { "epoch": 6.94, "learning_rate": 9.184747378455673e-07, "loss": 1.2597, "step": 36442 }, { "epoch": 6.94, "learning_rate": 9.18417540514776e-07, "loss": 1.1376, "step": 36443 }, { "epoch": 6.94, "learning_rate": 9.183603431839848e-07, "loss": 0.7125, "step": 36444 }, { "epoch": 6.94, "learning_rate": 9.183031458531936e-07, "loss": 2.3103, "step": 36445 }, { "epoch": 6.94, "learning_rate": 9.182459485224022e-07, "loss": 0.9749, "step": 36446 }, { "epoch": 6.94, "learning_rate": 9.18188751191611e-07, "loss": 1.7309, "step": 36447 }, { "epoch": 6.94, "learning_rate": 9.181315538608198e-07, "loss": 1.9174, "step": 36448 }, { "epoch": 6.94, "learning_rate": 9.180743565300287e-07, "loss": 2.0125, "step": 36449 }, { "epoch": 6.94, "learning_rate": 9.180171591992373e-07, "loss": 1.7966, "step": 36450 }, { "epoch": 6.94, "learning_rate": 9.179599618684461e-07, "loss": 1.1798, "step": 36451 }, { "epoch": 6.94, "learning_rate": 9.17902764537655e-07, "loss": 1.2924, "step": 36452 }, { "epoch": 6.94, "learning_rate": 9.178455672068636e-07, "loss": 2.016, "step": 36453 }, { "epoch": 6.94, "learning_rate": 9.177883698760725e-07, "loss": 1.5589, "step": 36454 }, { "epoch": 6.94, "learning_rate": 9.177311725452813e-07, "loss": 0.9523, "step": 36455 }, { "epoch": 6.94, "learning_rate": 9.176739752144901e-07, "loss": 2.0728, "step": 36456 }, { "epoch": 6.94, "learning_rate": 9.176167778836988e-07, "loss": 2.2378, "step": 36457 }, { "epoch": 6.94, "learning_rate": 9.175595805529076e-07, "loss": 2.4762, "step": 36458 }, { "epoch": 6.94, "learning_rate": 9.175023832221164e-07, "loss": 2.6632, "step": 36459 }, { "epoch": 6.94, "learning_rate": 9.17445185891325e-07, "loss": 1.2463, "step": 36460 }, { "epoch": 6.94, "learning_rate": 9.173879885605338e-07, "loss": 0.9564, "step": 36461 }, { "epoch": 6.95, "learning_rate": 9.173307912297426e-07, "loss": 0.9078, "step": 36462 }, { "epoch": 6.95, "learning_rate": 9.172735938989513e-07, "loss": 1.6265, "step": 36463 }, { "epoch": 6.95, "learning_rate": 9.172163965681601e-07, "loss": 2.359, "step": 36464 }, { "epoch": 6.95, "learning_rate": 9.17159199237369e-07, "loss": 1.491, "step": 36465 }, { "epoch": 6.95, "learning_rate": 9.171020019065778e-07, "loss": 1.2881, "step": 36466 }, { "epoch": 6.95, "learning_rate": 9.170448045757864e-07, "loss": 1.1311, "step": 36467 }, { "epoch": 6.95, "learning_rate": 9.169876072449953e-07, "loss": 1.6482, "step": 36468 }, { "epoch": 6.95, "learning_rate": 9.169304099142041e-07, "loss": 1.4715, "step": 36469 }, { "epoch": 6.95, "learning_rate": 9.168732125834128e-07, "loss": 1.8732, "step": 36470 }, { "epoch": 6.95, "learning_rate": 9.168160152526216e-07, "loss": 1.4528, "step": 36471 }, { "epoch": 6.95, "learning_rate": 9.167588179218304e-07, "loss": 1.3947, "step": 36472 }, { "epoch": 6.95, "learning_rate": 9.167016205910392e-07, "loss": 2.2744, "step": 36473 }, { "epoch": 6.95, "learning_rate": 9.166444232602478e-07, "loss": 2.1488, "step": 36474 }, { "epoch": 6.95, "learning_rate": 9.165872259294566e-07, "loss": 1.8145, "step": 36475 }, { "epoch": 6.95, "learning_rate": 9.165300285986654e-07, "loss": 0.8479, "step": 36476 }, { "epoch": 6.95, "learning_rate": 9.164728312678741e-07, "loss": 2.0903, "step": 36477 }, { "epoch": 6.95, "learning_rate": 9.164156339370829e-07, "loss": 1.6538, "step": 36478 }, { "epoch": 6.95, "learning_rate": 9.163584366062918e-07, "loss": 2.4432, "step": 36479 }, { "epoch": 6.95, "learning_rate": 9.163012392755006e-07, "loss": 1.9874, "step": 36480 }, { "epoch": 6.95, "learning_rate": 9.162440419447092e-07, "loss": 0.9068, "step": 36481 }, { "epoch": 6.95, "learning_rate": 9.161868446139181e-07, "loss": 1.1844, "step": 36482 }, { "epoch": 6.95, "learning_rate": 9.161296472831269e-07, "loss": 1.5521, "step": 36483 }, { "epoch": 6.95, "learning_rate": 9.160724499523356e-07, "loss": 1.7176, "step": 36484 }, { "epoch": 6.95, "learning_rate": 9.160152526215444e-07, "loss": 0.6681, "step": 36485 }, { "epoch": 6.95, "learning_rate": 9.159580552907532e-07, "loss": 2.1274, "step": 36486 }, { "epoch": 6.95, "learning_rate": 9.159008579599618e-07, "loss": 1.5544, "step": 36487 }, { "epoch": 6.95, "learning_rate": 9.158436606291706e-07, "loss": 1.5354, "step": 36488 }, { "epoch": 6.95, "learning_rate": 9.157864632983794e-07, "loss": 1.3897, "step": 36489 }, { "epoch": 6.95, "learning_rate": 9.157292659675882e-07, "loss": 1.6806, "step": 36490 }, { "epoch": 6.95, "learning_rate": 9.156720686367969e-07, "loss": 1.3604, "step": 36491 }, { "epoch": 6.95, "learning_rate": 9.156148713060057e-07, "loss": 1.0981, "step": 36492 }, { "epoch": 6.95, "learning_rate": 9.155576739752146e-07, "loss": 1.7383, "step": 36493 }, { "epoch": 6.95, "learning_rate": 9.155004766444232e-07, "loss": 0.8549, "step": 36494 }, { "epoch": 6.95, "learning_rate": 9.15443279313632e-07, "loss": 1.7225, "step": 36495 }, { "epoch": 6.95, "learning_rate": 9.153860819828409e-07, "loss": 2.3797, "step": 36496 }, { "epoch": 6.95, "learning_rate": 9.153288846520497e-07, "loss": 1.8569, "step": 36497 }, { "epoch": 6.95, "learning_rate": 9.152716873212584e-07, "loss": 0.6503, "step": 36498 }, { "epoch": 6.95, "learning_rate": 9.152144899904672e-07, "loss": 1.6605, "step": 36499 }, { "epoch": 6.95, "learning_rate": 9.15157292659676e-07, "loss": 1.4905, "step": 36500 }, { "epoch": 6.95, "learning_rate": 9.151000953288846e-07, "loss": 1.0697, "step": 36501 }, { "epoch": 6.95, "learning_rate": 9.150428979980934e-07, "loss": 1.8727, "step": 36502 }, { "epoch": 6.95, "learning_rate": 9.149857006673022e-07, "loss": 2.9288, "step": 36503 }, { "epoch": 6.95, "learning_rate": 9.14928503336511e-07, "loss": 1.6553, "step": 36504 }, { "epoch": 6.95, "learning_rate": 9.148713060057197e-07, "loss": 1.3325, "step": 36505 }, { "epoch": 6.95, "learning_rate": 9.148141086749285e-07, "loss": 1.6875, "step": 36506 }, { "epoch": 6.95, "learning_rate": 9.147569113441374e-07, "loss": 1.1965, "step": 36507 }, { "epoch": 6.95, "learning_rate": 9.14699714013346e-07, "loss": 2.4389, "step": 36508 }, { "epoch": 6.95, "learning_rate": 9.146425166825549e-07, "loss": 1.1056, "step": 36509 }, { "epoch": 6.95, "learning_rate": 9.145853193517637e-07, "loss": 2.9927, "step": 36510 }, { "epoch": 6.95, "learning_rate": 9.145281220209723e-07, "loss": 1.5874, "step": 36511 }, { "epoch": 6.95, "learning_rate": 9.144709246901812e-07, "loss": 1.4867, "step": 36512 }, { "epoch": 6.95, "learning_rate": 9.1441372735939e-07, "loss": 0.5008, "step": 36513 }, { "epoch": 6.96, "learning_rate": 9.143565300285987e-07, "loss": 2.6771, "step": 36514 }, { "epoch": 6.96, "learning_rate": 9.142993326978074e-07, "loss": 0.9744, "step": 36515 }, { "epoch": 6.96, "learning_rate": 9.142421353670162e-07, "loss": 2.6652, "step": 36516 }, { "epoch": 6.96, "learning_rate": 9.14184938036225e-07, "loss": 2.2196, "step": 36517 }, { "epoch": 6.96, "learning_rate": 9.141277407054337e-07, "loss": 2.076, "step": 36518 }, { "epoch": 6.96, "learning_rate": 9.140705433746425e-07, "loss": 1.5044, "step": 36519 }, { "epoch": 6.96, "learning_rate": 9.140133460438513e-07, "loss": 3.3662, "step": 36520 }, { "epoch": 6.96, "learning_rate": 9.139561487130602e-07, "loss": 0.9757, "step": 36521 }, { "epoch": 6.96, "learning_rate": 9.138989513822688e-07, "loss": 2.3447, "step": 36522 }, { "epoch": 6.96, "learning_rate": 9.138417540514777e-07, "loss": 1.9259, "step": 36523 }, { "epoch": 6.96, "learning_rate": 9.137845567206865e-07, "loss": 1.0709, "step": 36524 }, { "epoch": 6.96, "learning_rate": 9.137273593898951e-07, "loss": 1.9933, "step": 36525 }, { "epoch": 6.96, "learning_rate": 9.13670162059104e-07, "loss": 1.4788, "step": 36526 }, { "epoch": 6.96, "learning_rate": 9.136129647283127e-07, "loss": 1.9765, "step": 36527 }, { "epoch": 6.96, "learning_rate": 9.135557673975214e-07, "loss": 2.2351, "step": 36528 }, { "epoch": 6.96, "learning_rate": 9.134985700667302e-07, "loss": 1.2866, "step": 36529 }, { "epoch": 6.96, "learning_rate": 9.13441372735939e-07, "loss": 1.4745, "step": 36530 }, { "epoch": 6.96, "learning_rate": 9.133841754051478e-07, "loss": 1.7874, "step": 36531 }, { "epoch": 6.96, "learning_rate": 9.133269780743565e-07, "loss": 2.4913, "step": 36532 }, { "epoch": 6.96, "learning_rate": 9.132697807435653e-07, "loss": 2.5778, "step": 36533 }, { "epoch": 6.96, "learning_rate": 9.132125834127741e-07, "loss": 0.3538, "step": 36534 }, { "epoch": 6.96, "learning_rate": 9.131553860819828e-07, "loss": 2.6208, "step": 36535 }, { "epoch": 6.96, "learning_rate": 9.130981887511916e-07, "loss": 2.0976, "step": 36536 }, { "epoch": 6.96, "learning_rate": 9.130409914204005e-07, "loss": 1.6643, "step": 36537 }, { "epoch": 6.96, "learning_rate": 9.129837940896093e-07, "loss": 2.968, "step": 36538 }, { "epoch": 6.96, "learning_rate": 9.129265967588179e-07, "loss": 2.7739, "step": 36539 }, { "epoch": 6.96, "learning_rate": 9.128693994280268e-07, "loss": 1.3007, "step": 36540 }, { "epoch": 6.96, "learning_rate": 9.128122020972355e-07, "loss": 1.8444, "step": 36541 }, { "epoch": 6.96, "learning_rate": 9.127550047664442e-07, "loss": 0.8033, "step": 36542 }, { "epoch": 6.96, "learning_rate": 9.12697807435653e-07, "loss": 1.973, "step": 36543 }, { "epoch": 6.96, "learning_rate": 9.126406101048618e-07, "loss": 1.9926, "step": 36544 }, { "epoch": 6.96, "learning_rate": 9.125834127740706e-07, "loss": 2.622, "step": 36545 }, { "epoch": 6.96, "learning_rate": 9.125262154432793e-07, "loss": 1.9933, "step": 36546 }, { "epoch": 6.96, "learning_rate": 9.124690181124881e-07, "loss": 2.0586, "step": 36547 }, { "epoch": 6.96, "learning_rate": 9.124118207816969e-07, "loss": 1.2282, "step": 36548 }, { "epoch": 6.96, "learning_rate": 9.123546234509056e-07, "loss": 1.6083, "step": 36549 }, { "epoch": 6.96, "learning_rate": 9.122974261201144e-07, "loss": 1.0132, "step": 36550 }, { "epoch": 6.96, "learning_rate": 9.122402287893233e-07, "loss": 0.835, "step": 36551 }, { "epoch": 6.96, "learning_rate": 9.121830314585319e-07, "loss": 1.582, "step": 36552 }, { "epoch": 6.96, "learning_rate": 9.121258341277406e-07, "loss": 2.0516, "step": 36553 }, { "epoch": 6.96, "learning_rate": 9.120686367969495e-07, "loss": 0.7833, "step": 36554 }, { "epoch": 6.96, "learning_rate": 9.120114394661583e-07, "loss": 0.6914, "step": 36555 }, { "epoch": 6.96, "learning_rate": 9.11954242135367e-07, "loss": 1.3068, "step": 36556 }, { "epoch": 6.96, "learning_rate": 9.118970448045758e-07, "loss": 1.7284, "step": 36557 }, { "epoch": 6.96, "learning_rate": 9.118398474737846e-07, "loss": 1.4462, "step": 36558 }, { "epoch": 6.96, "learning_rate": 9.117826501429933e-07, "loss": 1.978, "step": 36559 }, { "epoch": 6.96, "learning_rate": 9.117254528122021e-07, "loss": 0.9145, "step": 36560 }, { "epoch": 6.96, "learning_rate": 9.116682554814109e-07, "loss": 2.1101, "step": 36561 }, { "epoch": 6.96, "learning_rate": 9.116110581506197e-07, "loss": 1.8478, "step": 36562 }, { "epoch": 6.96, "learning_rate": 9.115538608198284e-07, "loss": 0.7891, "step": 36563 }, { "epoch": 6.96, "learning_rate": 9.114966634890372e-07, "loss": 1.995, "step": 36564 }, { "epoch": 6.96, "learning_rate": 9.114394661582461e-07, "loss": 1.9566, "step": 36565 }, { "epoch": 6.96, "learning_rate": 9.113822688274546e-07, "loss": 1.6615, "step": 36566 }, { "epoch": 6.97, "learning_rate": 9.113250714966635e-07, "loss": 1.212, "step": 36567 }, { "epoch": 6.97, "learning_rate": 9.112678741658723e-07, "loss": 1.3071, "step": 36568 }, { "epoch": 6.97, "learning_rate": 9.112106768350811e-07, "loss": 0.4474, "step": 36569 }, { "epoch": 6.97, "learning_rate": 9.111534795042898e-07, "loss": 1.317, "step": 36570 }, { "epoch": 6.97, "learning_rate": 9.110962821734986e-07, "loss": 1.0887, "step": 36571 }, { "epoch": 6.97, "learning_rate": 9.110390848427074e-07, "loss": 2.3682, "step": 36572 }, { "epoch": 6.97, "learning_rate": 9.109818875119161e-07, "loss": 1.5233, "step": 36573 }, { "epoch": 6.97, "learning_rate": 9.109246901811249e-07, "loss": 1.2486, "step": 36574 }, { "epoch": 6.97, "learning_rate": 9.108674928503337e-07, "loss": 2.2032, "step": 36575 }, { "epoch": 6.97, "learning_rate": 9.108102955195424e-07, "loss": 1.6342, "step": 36576 }, { "epoch": 6.97, "learning_rate": 9.107530981887512e-07, "loss": 1.5127, "step": 36577 }, { "epoch": 6.97, "learning_rate": 9.1069590085796e-07, "loss": 0.9162, "step": 36578 }, { "epoch": 6.97, "learning_rate": 9.106387035271689e-07, "loss": 2.8307, "step": 36579 }, { "epoch": 6.97, "learning_rate": 9.105815061963774e-07, "loss": 0.8011, "step": 36580 }, { "epoch": 6.97, "learning_rate": 9.105243088655863e-07, "loss": 2.1622, "step": 36581 }, { "epoch": 6.97, "learning_rate": 9.104671115347951e-07, "loss": 1.9, "step": 36582 }, { "epoch": 6.97, "learning_rate": 9.104099142040037e-07, "loss": 1.2873, "step": 36583 }, { "epoch": 6.97, "learning_rate": 9.103527168732126e-07, "loss": 1.5723, "step": 36584 }, { "epoch": 6.97, "learning_rate": 9.102955195424214e-07, "loss": 1.6812, "step": 36585 }, { "epoch": 6.97, "learning_rate": 9.102383222116302e-07, "loss": 3.6835, "step": 36586 }, { "epoch": 6.97, "learning_rate": 9.101811248808389e-07, "loss": 2.2528, "step": 36587 }, { "epoch": 6.97, "learning_rate": 9.101239275500477e-07, "loss": 1.0877, "step": 36588 }, { "epoch": 6.97, "learning_rate": 9.100667302192565e-07, "loss": 1.1396, "step": 36589 }, { "epoch": 6.97, "learning_rate": 9.100095328884652e-07, "loss": 1.5596, "step": 36590 }, { "epoch": 6.97, "learning_rate": 9.09952335557674e-07, "loss": 1.5527, "step": 36591 }, { "epoch": 6.97, "learning_rate": 9.098951382268828e-07, "loss": 1.7214, "step": 36592 }, { "epoch": 6.97, "learning_rate": 9.098379408960914e-07, "loss": 1.9134, "step": 36593 }, { "epoch": 6.97, "learning_rate": 9.097807435653002e-07, "loss": 1.9531, "step": 36594 }, { "epoch": 6.97, "learning_rate": 9.097235462345091e-07, "loss": 0.7647, "step": 36595 }, { "epoch": 6.97, "learning_rate": 9.096663489037179e-07, "loss": 2.2522, "step": 36596 }, { "epoch": 6.97, "learning_rate": 9.096091515729265e-07, "loss": 2.2785, "step": 36597 }, { "epoch": 6.97, "learning_rate": 9.095519542421354e-07, "loss": 0.3038, "step": 36598 }, { "epoch": 6.97, "learning_rate": 9.094947569113442e-07, "loss": 2.808, "step": 36599 }, { "epoch": 6.97, "learning_rate": 9.094375595805529e-07, "loss": 1.792, "step": 36600 }, { "epoch": 6.97, "learning_rate": 9.093803622497617e-07, "loss": 2.8963, "step": 36601 }, { "epoch": 6.97, "learning_rate": 9.093231649189705e-07, "loss": 2.4576, "step": 36602 }, { "epoch": 6.97, "learning_rate": 9.092659675881793e-07, "loss": 2.091, "step": 36603 }, { "epoch": 6.97, "learning_rate": 9.09208770257388e-07, "loss": 2.7508, "step": 36604 }, { "epoch": 6.97, "learning_rate": 9.091515729265968e-07, "loss": 1.0061, "step": 36605 }, { "epoch": 6.97, "learning_rate": 9.090943755958056e-07, "loss": 0.4779, "step": 36606 }, { "epoch": 6.97, "learning_rate": 9.090371782650142e-07, "loss": 1.7152, "step": 36607 }, { "epoch": 6.97, "learning_rate": 9.08979980934223e-07, "loss": 0.2786, "step": 36608 }, { "epoch": 6.97, "learning_rate": 9.089227836034319e-07, "loss": 2.0006, "step": 36609 }, { "epoch": 6.97, "learning_rate": 9.088655862726407e-07, "loss": 1.2591, "step": 36610 }, { "epoch": 6.97, "learning_rate": 9.088083889418494e-07, "loss": 1.6753, "step": 36611 }, { "epoch": 6.97, "learning_rate": 9.087511916110582e-07, "loss": 0.5983, "step": 36612 }, { "epoch": 6.97, "learning_rate": 9.08693994280267e-07, "loss": 1.9406, "step": 36613 }, { "epoch": 6.97, "learning_rate": 9.086367969494757e-07, "loss": 1.8671, "step": 36614 }, { "epoch": 6.97, "learning_rate": 9.085795996186845e-07, "loss": 0.6976, "step": 36615 }, { "epoch": 6.97, "learning_rate": 9.085224022878933e-07, "loss": 2.6034, "step": 36616 }, { "epoch": 6.97, "learning_rate": 9.08465204957102e-07, "loss": 1.1974, "step": 36617 }, { "epoch": 6.97, "learning_rate": 9.084080076263108e-07, "loss": 2.8422, "step": 36618 }, { "epoch": 6.98, "learning_rate": 9.083508102955196e-07, "loss": 0.9307, "step": 36619 }, { "epoch": 6.98, "learning_rate": 9.082936129647285e-07, "loss": 1.8718, "step": 36620 }, { "epoch": 6.98, "learning_rate": 9.08236415633937e-07, "loss": 1.6734, "step": 36621 }, { "epoch": 6.98, "learning_rate": 9.081792183031458e-07, "loss": 1.4418, "step": 36622 }, { "epoch": 6.98, "learning_rate": 9.081220209723547e-07, "loss": 1.9846, "step": 36623 }, { "epoch": 6.98, "learning_rate": 9.080648236415633e-07, "loss": 0.7751, "step": 36624 }, { "epoch": 6.98, "learning_rate": 9.080076263107722e-07, "loss": 0.9042, "step": 36625 }, { "epoch": 6.98, "learning_rate": 9.07950428979981e-07, "loss": 1.2327, "step": 36626 }, { "epoch": 6.98, "learning_rate": 9.078932316491898e-07, "loss": 2.3908, "step": 36627 }, { "epoch": 6.98, "learning_rate": 9.078360343183985e-07, "loss": 1.6989, "step": 36628 }, { "epoch": 6.98, "learning_rate": 9.077788369876073e-07, "loss": 1.2895, "step": 36629 }, { "epoch": 6.98, "learning_rate": 9.077216396568161e-07, "loss": 1.7886, "step": 36630 }, { "epoch": 6.98, "learning_rate": 9.076644423260248e-07, "loss": 1.8481, "step": 36631 }, { "epoch": 6.98, "learning_rate": 9.076072449952336e-07, "loss": 1.9316, "step": 36632 }, { "epoch": 6.98, "learning_rate": 9.075500476644424e-07, "loss": 3.2403, "step": 36633 }, { "epoch": 6.98, "learning_rate": 9.07492850333651e-07, "loss": 0.7287, "step": 36634 }, { "epoch": 6.98, "learning_rate": 9.074356530028598e-07, "loss": 2.2064, "step": 36635 }, { "epoch": 6.98, "learning_rate": 9.073784556720686e-07, "loss": 1.1643, "step": 36636 }, { "epoch": 6.98, "learning_rate": 9.073212583412775e-07, "loss": 1.7929, "step": 36637 }, { "epoch": 6.98, "learning_rate": 9.072640610104861e-07, "loss": 1.5565, "step": 36638 }, { "epoch": 6.98, "learning_rate": 9.07206863679695e-07, "loss": 1.5045, "step": 36639 }, { "epoch": 6.98, "learning_rate": 9.071496663489038e-07, "loss": 1.6748, "step": 36640 }, { "epoch": 6.98, "learning_rate": 9.070924690181124e-07, "loss": 1.438, "step": 36641 }, { "epoch": 6.98, "learning_rate": 9.070352716873213e-07, "loss": 0.9889, "step": 36642 }, { "epoch": 6.98, "learning_rate": 9.069780743565301e-07, "loss": 1.5873, "step": 36643 }, { "epoch": 6.98, "learning_rate": 9.069208770257389e-07, "loss": 3.3731, "step": 36644 }, { "epoch": 6.98, "learning_rate": 9.068636796949476e-07, "loss": 2.3432, "step": 36645 }, { "epoch": 6.98, "learning_rate": 9.068064823641564e-07, "loss": 2.4995, "step": 36646 }, { "epoch": 6.98, "learning_rate": 9.067492850333651e-07, "loss": 1.3826, "step": 36647 }, { "epoch": 6.98, "learning_rate": 9.066920877025738e-07, "loss": 0.6219, "step": 36648 }, { "epoch": 6.98, "learning_rate": 9.066348903717826e-07, "loss": 1.4388, "step": 36649 }, { "epoch": 6.98, "learning_rate": 9.065776930409914e-07, "loss": 1.477, "step": 36650 }, { "epoch": 6.98, "learning_rate": 9.065204957102003e-07, "loss": 1.4977, "step": 36651 }, { "epoch": 6.98, "learning_rate": 9.064632983794089e-07, "loss": 1.4637, "step": 36652 }, { "epoch": 6.98, "learning_rate": 9.064061010486178e-07, "loss": 2.3778, "step": 36653 }, { "epoch": 6.98, "learning_rate": 9.063489037178266e-07, "loss": 1.6118, "step": 36654 }, { "epoch": 6.98, "learning_rate": 9.062917063870353e-07, "loss": 1.336, "step": 36655 }, { "epoch": 6.98, "learning_rate": 9.062345090562441e-07, "loss": 1.3102, "step": 36656 }, { "epoch": 6.98, "learning_rate": 9.061773117254529e-07, "loss": 2.3595, "step": 36657 }, { "epoch": 6.98, "learning_rate": 9.061201143946616e-07, "loss": 2.7859, "step": 36658 }, { "epoch": 6.98, "learning_rate": 9.060629170638704e-07, "loss": 1.6305, "step": 36659 }, { "epoch": 6.98, "learning_rate": 9.060057197330791e-07, "loss": 1.4079, "step": 36660 }, { "epoch": 6.98, "learning_rate": 9.059485224022879e-07, "loss": 2.3497, "step": 36661 }, { "epoch": 6.98, "learning_rate": 9.058913250714966e-07, "loss": 1.6396, "step": 36662 }, { "epoch": 6.98, "learning_rate": 9.058341277407054e-07, "loss": 1.8185, "step": 36663 }, { "epoch": 6.98, "learning_rate": 9.057769304099143e-07, "loss": 1.884, "step": 36664 }, { "epoch": 6.98, "learning_rate": 9.057197330791229e-07, "loss": 0.3749, "step": 36665 }, { "epoch": 6.98, "learning_rate": 9.056625357483317e-07, "loss": 1.5094, "step": 36666 }, { "epoch": 6.98, "learning_rate": 9.056053384175406e-07, "loss": 0.7796, "step": 36667 }, { "epoch": 6.98, "learning_rate": 9.055481410867494e-07, "loss": 1.2072, "step": 36668 }, { "epoch": 6.98, "learning_rate": 9.054909437559581e-07, "loss": 1.9084, "step": 36669 }, { "epoch": 6.98, "learning_rate": 9.054337464251669e-07, "loss": 2.3684, "step": 36670 }, { "epoch": 6.98, "learning_rate": 9.053765490943757e-07, "loss": 1.5728, "step": 36671 }, { "epoch": 6.99, "learning_rate": 9.053193517635844e-07, "loss": 1.824, "step": 36672 }, { "epoch": 6.99, "learning_rate": 9.052621544327932e-07, "loss": 0.8637, "step": 36673 }, { "epoch": 6.99, "learning_rate": 9.052049571020019e-07, "loss": 1.145, "step": 36674 }, { "epoch": 6.99, "learning_rate": 9.051477597712107e-07, "loss": 1.6572, "step": 36675 }, { "epoch": 6.99, "learning_rate": 9.050905624404194e-07, "loss": 1.6295, "step": 36676 }, { "epoch": 6.99, "learning_rate": 9.050333651096282e-07, "loss": 0.9784, "step": 36677 }, { "epoch": 6.99, "learning_rate": 9.049761677788371e-07, "loss": 1.3715, "step": 36678 }, { "epoch": 6.99, "learning_rate": 9.049189704480457e-07, "loss": 1.9432, "step": 36679 }, { "epoch": 6.99, "learning_rate": 9.048617731172545e-07, "loss": 2.1084, "step": 36680 }, { "epoch": 6.99, "learning_rate": 9.048045757864634e-07, "loss": 0.9867, "step": 36681 }, { "epoch": 6.99, "learning_rate": 9.04747378455672e-07, "loss": 1.6369, "step": 36682 }, { "epoch": 6.99, "learning_rate": 9.046901811248809e-07, "loss": 2.5055, "step": 36683 }, { "epoch": 6.99, "learning_rate": 9.046329837940897e-07, "loss": 1.8369, "step": 36684 }, { "epoch": 6.99, "learning_rate": 9.045757864632985e-07, "loss": 2.1219, "step": 36685 }, { "epoch": 6.99, "learning_rate": 9.045185891325072e-07, "loss": 1.1695, "step": 36686 }, { "epoch": 6.99, "learning_rate": 9.04461391801716e-07, "loss": 2.997, "step": 36687 }, { "epoch": 6.99, "learning_rate": 9.044041944709247e-07, "loss": 2.9529, "step": 36688 }, { "epoch": 6.99, "learning_rate": 9.043469971401334e-07, "loss": 2.3946, "step": 36689 }, { "epoch": 6.99, "learning_rate": 9.042897998093422e-07, "loss": 1.5593, "step": 36690 }, { "epoch": 6.99, "learning_rate": 9.04232602478551e-07, "loss": 0.9115, "step": 36691 }, { "epoch": 6.99, "learning_rate": 9.041754051477599e-07, "loss": 0.9327, "step": 36692 }, { "epoch": 6.99, "learning_rate": 9.041182078169685e-07, "loss": 1.1547, "step": 36693 }, { "epoch": 6.99, "learning_rate": 9.040610104861773e-07, "loss": 1.0463, "step": 36694 }, { "epoch": 6.99, "learning_rate": 9.040038131553862e-07, "loss": 1.5193, "step": 36695 }, { "epoch": 6.99, "learning_rate": 9.039466158245948e-07, "loss": 2.084, "step": 36696 }, { "epoch": 6.99, "learning_rate": 9.038894184938037e-07, "loss": 0.7055, "step": 36697 }, { "epoch": 6.99, "learning_rate": 9.038322211630125e-07, "loss": 2.2993, "step": 36698 }, { "epoch": 6.99, "learning_rate": 9.037750238322212e-07, "loss": 1.856, "step": 36699 }, { "epoch": 6.99, "learning_rate": 9.0371782650143e-07, "loss": 1.973, "step": 36700 }, { "epoch": 6.99, "learning_rate": 9.036606291706387e-07, "loss": 1.1415, "step": 36701 }, { "epoch": 6.99, "learning_rate": 9.036034318398475e-07, "loss": 3.0765, "step": 36702 }, { "epoch": 6.99, "learning_rate": 9.035462345090562e-07, "loss": 0.9255, "step": 36703 }, { "epoch": 6.99, "learning_rate": 9.03489037178265e-07, "loss": 3.7047, "step": 36704 }, { "epoch": 6.99, "learning_rate": 9.034318398474738e-07, "loss": 0.5188, "step": 36705 }, { "epoch": 6.99, "learning_rate": 9.033746425166825e-07, "loss": 2.2178, "step": 36706 }, { "epoch": 6.99, "learning_rate": 9.033174451858913e-07, "loss": 1.5192, "step": 36707 }, { "epoch": 6.99, "learning_rate": 9.032602478551001e-07, "loss": 1.3882, "step": 36708 }, { "epoch": 6.99, "learning_rate": 9.03203050524309e-07, "loss": 1.6224, "step": 36709 }, { "epoch": 6.99, "learning_rate": 9.031458531935176e-07, "loss": 2.4543, "step": 36710 }, { "epoch": 6.99, "learning_rate": 9.030886558627265e-07, "loss": 1.1971, "step": 36711 }, { "epoch": 6.99, "learning_rate": 9.030314585319353e-07, "loss": 1.7889, "step": 36712 }, { "epoch": 6.99, "learning_rate": 9.02974261201144e-07, "loss": 1.7076, "step": 36713 }, { "epoch": 6.99, "learning_rate": 9.029170638703527e-07, "loss": 1.6503, "step": 36714 }, { "epoch": 6.99, "learning_rate": 9.028598665395615e-07, "loss": 2.0513, "step": 36715 }, { "epoch": 6.99, "learning_rate": 9.028026692087703e-07, "loss": 0.224, "step": 36716 }, { "epoch": 6.99, "learning_rate": 9.02745471877979e-07, "loss": 2.6884, "step": 36717 }, { "epoch": 6.99, "learning_rate": 9.026882745471878e-07, "loss": 2.2515, "step": 36718 }, { "epoch": 6.99, "learning_rate": 9.026310772163966e-07, "loss": 3.0244, "step": 36719 }, { "epoch": 6.99, "learning_rate": 9.025738798856053e-07, "loss": 2.774, "step": 36720 }, { "epoch": 6.99, "learning_rate": 9.025166825548141e-07, "loss": 1.7596, "step": 36721 }, { "epoch": 6.99, "learning_rate": 9.02459485224023e-07, "loss": 2.4581, "step": 36722 }, { "epoch": 6.99, "learning_rate": 9.024022878932316e-07, "loss": 2.3756, "step": 36723 }, { "epoch": 7.0, "learning_rate": 9.023450905624404e-07, "loss": 0.5662, "step": 36724 }, { "epoch": 7.0, "learning_rate": 9.022878932316493e-07, "loss": 2.6845, "step": 36725 }, { "epoch": 7.0, "learning_rate": 9.022306959008581e-07, "loss": 2.1083, "step": 36726 }, { "epoch": 7.0, "learning_rate": 9.021734985700667e-07, "loss": 1.0833, "step": 36727 }, { "epoch": 7.0, "learning_rate": 9.021163012392755e-07, "loss": 0.974, "step": 36728 }, { "epoch": 7.0, "learning_rate": 9.020591039084843e-07, "loss": 1.0315, "step": 36729 }, { "epoch": 7.0, "learning_rate": 9.02001906577693e-07, "loss": 1.5212, "step": 36730 }, { "epoch": 7.0, "learning_rate": 9.019447092469018e-07, "loss": 1.5434, "step": 36731 }, { "epoch": 7.0, "learning_rate": 9.018875119161106e-07, "loss": 0.6872, "step": 36732 }, { "epoch": 7.0, "learning_rate": 9.018303145853194e-07, "loss": 2.0647, "step": 36733 }, { "epoch": 7.0, "learning_rate": 9.017731172545281e-07, "loss": 2.4043, "step": 36734 }, { "epoch": 7.0, "learning_rate": 9.017159199237369e-07, "loss": 2.376, "step": 36735 }, { "epoch": 7.0, "learning_rate": 9.016587225929458e-07, "loss": 1.614, "step": 36736 }, { "epoch": 7.0, "learning_rate": 9.016015252621544e-07, "loss": 1.4017, "step": 36737 }, { "epoch": 7.0, "learning_rate": 9.015443279313632e-07, "loss": 1.135, "step": 36738 }, { "epoch": 7.0, "learning_rate": 9.014871306005721e-07, "loss": 1.8979, "step": 36739 }, { "epoch": 7.0, "learning_rate": 9.014299332697809e-07, "loss": 1.4441, "step": 36740 }, { "epoch": 7.0, "learning_rate": 9.013727359389895e-07, "loss": 0.8043, "step": 36741 }, { "epoch": 7.0, "learning_rate": 9.013155386081983e-07, "loss": 2.8699, "step": 36742 }, { "epoch": 7.0, "learning_rate": 9.012583412774071e-07, "loss": 2.7379, "step": 36743 }, { "epoch": 7.0, "learning_rate": 9.012011439466158e-07, "loss": 1.8004, "step": 36744 }, { "epoch": 7.0, "learning_rate": 9.011439466158246e-07, "loss": 1.5861, "step": 36745 }, { "epoch": 7.0, "learning_rate": 9.010867492850334e-07, "loss": 2.7842, "step": 36746 }, { "epoch": 7.0, "learning_rate": 9.010295519542421e-07, "loss": 2.6532, "step": 36747 }, { "epoch": 7.0, "learning_rate": 9.009723546234509e-07, "loss": 0.6829, "step": 36748 }, { "epoch": 7.0, "learning_rate": 9.009151572926597e-07, "loss": 2.534, "step": 36749 }, { "epoch": 7.0, "learning_rate": 9.008579599618686e-07, "loss": 1.9252, "step": 36750 }, { "epoch": 7.0, "eval_accuracy": 0.5829404200600086, "eval_loss": 1.626317024230957, "eval_runtime": 127.127, "eval_samples_per_second": 18.352, "eval_steps_per_second": 4.594, "step": 36750 }, { "epoch": 7.0, "learning_rate": 9.008007626310772e-07, "loss": 1.8911, "step": 36751 }, { "epoch": 7.0, "learning_rate": 9.00743565300286e-07, "loss": 0.7689, "step": 36752 }, { "epoch": 7.0, "learning_rate": 9.006863679694949e-07, "loss": 2.1096, "step": 36753 }, { "epoch": 7.0, "learning_rate": 9.006291706387034e-07, "loss": 1.0633, "step": 36754 }, { "epoch": 7.0, "learning_rate": 9.005719733079123e-07, "loss": 1.0188, "step": 36755 }, { "epoch": 7.0, "learning_rate": 9.005147759771211e-07, "loss": 2.3898, "step": 36756 }, { "epoch": 7.0, "learning_rate": 9.004575786463299e-07, "loss": 1.4637, "step": 36757 }, { "epoch": 7.0, "learning_rate": 9.004003813155386e-07, "loss": 1.077, "step": 36758 }, { "epoch": 7.0, "learning_rate": 9.003431839847474e-07, "loss": 1.2721, "step": 36759 }, { "epoch": 7.0, "learning_rate": 9.002859866539562e-07, "loss": 1.969, "step": 36760 }, { "epoch": 7.0, "learning_rate": 9.002287893231649e-07, "loss": 1.9028, "step": 36761 }, { "epoch": 7.0, "learning_rate": 9.001715919923737e-07, "loss": 1.3189, "step": 36762 }, { "epoch": 7.0, "learning_rate": 9.001143946615825e-07, "loss": 1.9153, "step": 36763 }, { "epoch": 7.0, "learning_rate": 9.000571973307912e-07, "loss": 2.8014, "step": 36764 }, { "epoch": 7.0, "learning_rate": 9e-07, "loss": 1.8007, "step": 36765 }, { "epoch": 7.0, "learning_rate": 8.999428026692089e-07, "loss": 4.071, "step": 36766 }, { "epoch": 7.0, "learning_rate": 8.998856053384177e-07, "loss": 1.0384, "step": 36767 }, { "epoch": 7.0, "learning_rate": 8.998284080076262e-07, "loss": 2.0046, "step": 36768 }, { "epoch": 7.0, "learning_rate": 8.997712106768351e-07, "loss": 1.9146, "step": 36769 }, { "epoch": 7.0, "learning_rate": 8.997140133460439e-07, "loss": 1.2275, "step": 36770 }, { "epoch": 7.0, "learning_rate": 8.996568160152526e-07, "loss": 2.5746, "step": 36771 }, { "epoch": 7.0, "learning_rate": 8.995996186844614e-07, "loss": 1.7763, "step": 36772 }, { "epoch": 7.0, "learning_rate": 8.995424213536702e-07, "loss": 3.1239, "step": 36773 }, { "epoch": 7.0, "learning_rate": 8.99485224022879e-07, "loss": 2.3121, "step": 36774 }, { "epoch": 7.0, "learning_rate": 8.994280266920877e-07, "loss": 0.6564, "step": 36775 }, { "epoch": 7.0, "learning_rate": 8.993708293612965e-07, "loss": 0.758, "step": 36776 }, { "epoch": 7.01, "learning_rate": 8.993136320305053e-07, "loss": 1.7144, "step": 36777 }, { "epoch": 7.01, "learning_rate": 8.99256434699714e-07, "loss": 1.3893, "step": 36778 }, { "epoch": 7.01, "learning_rate": 8.991992373689228e-07, "loss": 1.8158, "step": 36779 }, { "epoch": 7.01, "learning_rate": 8.991420400381317e-07, "loss": 0.5684, "step": 36780 }, { "epoch": 7.01, "learning_rate": 8.990848427073404e-07, "loss": 1.2336, "step": 36781 }, { "epoch": 7.01, "learning_rate": 8.99027645376549e-07, "loss": 0.8484, "step": 36782 }, { "epoch": 7.01, "learning_rate": 8.989704480457579e-07, "loss": 1.2604, "step": 36783 }, { "epoch": 7.01, "learning_rate": 8.989132507149667e-07, "loss": 0.9781, "step": 36784 }, { "epoch": 7.01, "learning_rate": 8.988560533841754e-07, "loss": 0.7549, "step": 36785 }, { "epoch": 7.01, "learning_rate": 8.987988560533842e-07, "loss": 0.8489, "step": 36786 }, { "epoch": 7.01, "learning_rate": 8.98741658722593e-07, "loss": 1.7135, "step": 36787 }, { "epoch": 7.01, "learning_rate": 8.986844613918017e-07, "loss": 2.2909, "step": 36788 }, { "epoch": 7.01, "learning_rate": 8.986272640610105e-07, "loss": 0.3679, "step": 36789 }, { "epoch": 7.01, "learning_rate": 8.985700667302193e-07, "loss": 2.6582, "step": 36790 }, { "epoch": 7.01, "learning_rate": 8.985128693994281e-07, "loss": 2.1033, "step": 36791 }, { "epoch": 7.01, "learning_rate": 8.984556720686368e-07, "loss": 1.3347, "step": 36792 }, { "epoch": 7.01, "learning_rate": 8.983984747378456e-07, "loss": 1.718, "step": 36793 }, { "epoch": 7.01, "learning_rate": 8.983412774070545e-07, "loss": 0.9608, "step": 36794 }, { "epoch": 7.01, "learning_rate": 8.98284080076263e-07, "loss": 1.3201, "step": 36795 }, { "epoch": 7.01, "learning_rate": 8.982268827454718e-07, "loss": 0.8847, "step": 36796 }, { "epoch": 7.01, "learning_rate": 8.981696854146807e-07, "loss": 1.5177, "step": 36797 }, { "epoch": 7.01, "learning_rate": 8.981124880838895e-07, "loss": 1.8019, "step": 36798 }, { "epoch": 7.01, "learning_rate": 8.980552907530982e-07, "loss": 0.8051, "step": 36799 }, { "epoch": 7.01, "learning_rate": 8.97998093422307e-07, "loss": 0.998, "step": 36800 }, { "epoch": 7.01, "learning_rate": 8.979408960915158e-07, "loss": 1.3733, "step": 36801 }, { "epoch": 7.01, "learning_rate": 8.978836987607245e-07, "loss": 1.2909, "step": 36802 }, { "epoch": 7.01, "learning_rate": 8.978265014299333e-07, "loss": 0.9644, "step": 36803 }, { "epoch": 7.01, "learning_rate": 8.977693040991421e-07, "loss": 1.6561, "step": 36804 }, { "epoch": 7.01, "learning_rate": 8.977121067683509e-07, "loss": 1.5534, "step": 36805 }, { "epoch": 7.01, "learning_rate": 8.976549094375596e-07, "loss": 1.3003, "step": 36806 }, { "epoch": 7.01, "learning_rate": 8.975977121067683e-07, "loss": 1.3406, "step": 36807 }, { "epoch": 7.01, "learning_rate": 8.975405147759772e-07, "loss": 1.6151, "step": 36808 }, { "epoch": 7.01, "learning_rate": 8.974833174451858e-07, "loss": 1.7981, "step": 36809 }, { "epoch": 7.01, "learning_rate": 8.974261201143947e-07, "loss": 1.9012, "step": 36810 }, { "epoch": 7.01, "learning_rate": 8.973689227836035e-07, "loss": 1.9065, "step": 36811 }, { "epoch": 7.01, "learning_rate": 8.973117254528121e-07, "loss": 1.9267, "step": 36812 }, { "epoch": 7.01, "learning_rate": 8.97254528122021e-07, "loss": 1.3407, "step": 36813 }, { "epoch": 7.01, "learning_rate": 8.971973307912298e-07, "loss": 1.8726, "step": 36814 }, { "epoch": 7.01, "learning_rate": 8.971401334604386e-07, "loss": 1.9709, "step": 36815 }, { "epoch": 7.01, "learning_rate": 8.970829361296473e-07, "loss": 2.1433, "step": 36816 }, { "epoch": 7.01, "learning_rate": 8.970257387988561e-07, "loss": 1.6732, "step": 36817 }, { "epoch": 7.01, "learning_rate": 8.969685414680649e-07, "loss": 0.871, "step": 36818 }, { "epoch": 7.01, "learning_rate": 8.969113441372736e-07, "loss": 1.9107, "step": 36819 }, { "epoch": 7.01, "learning_rate": 8.968541468064823e-07, "loss": 1.79, "step": 36820 }, { "epoch": 7.01, "learning_rate": 8.967969494756911e-07, "loss": 1.5771, "step": 36821 }, { "epoch": 7.01, "learning_rate": 8.967397521449e-07, "loss": 1.0437, "step": 36822 }, { "epoch": 7.01, "learning_rate": 8.966825548141086e-07, "loss": 1.8922, "step": 36823 }, { "epoch": 7.01, "learning_rate": 8.966253574833175e-07, "loss": 2.559, "step": 36824 }, { "epoch": 7.01, "learning_rate": 8.965681601525263e-07, "loss": 1.7945, "step": 36825 }, { "epoch": 7.01, "learning_rate": 8.965109628217349e-07, "loss": 1.7081, "step": 36826 }, { "epoch": 7.01, "learning_rate": 8.964537654909438e-07, "loss": 0.7867, "step": 36827 }, { "epoch": 7.01, "learning_rate": 8.963965681601526e-07, "loss": 2.0062, "step": 36828 }, { "epoch": 7.02, "learning_rate": 8.963393708293613e-07, "loss": 1.4302, "step": 36829 }, { "epoch": 7.02, "learning_rate": 8.962821734985701e-07, "loss": 1.7459, "step": 36830 }, { "epoch": 7.02, "learning_rate": 8.962249761677789e-07, "loss": 0.9324, "step": 36831 }, { "epoch": 7.02, "learning_rate": 8.961677788369877e-07, "loss": 1.4842, "step": 36832 }, { "epoch": 7.02, "learning_rate": 8.961105815061964e-07, "loss": 1.1988, "step": 36833 }, { "epoch": 7.02, "learning_rate": 8.960533841754051e-07, "loss": 0.7933, "step": 36834 }, { "epoch": 7.02, "learning_rate": 8.959961868446139e-07, "loss": 1.6893, "step": 36835 }, { "epoch": 7.02, "learning_rate": 8.959389895138226e-07, "loss": 1.358, "step": 36836 }, { "epoch": 7.02, "learning_rate": 8.958817921830314e-07, "loss": 0.4247, "step": 36837 }, { "epoch": 7.02, "learning_rate": 8.958245948522403e-07, "loss": 1.7188, "step": 36838 }, { "epoch": 7.02, "learning_rate": 8.957673975214491e-07, "loss": 2.492, "step": 36839 }, { "epoch": 7.02, "learning_rate": 8.957102001906577e-07, "loss": 1.1283, "step": 36840 }, { "epoch": 7.02, "learning_rate": 8.956530028598666e-07, "loss": 0.2376, "step": 36841 }, { "epoch": 7.02, "learning_rate": 8.955958055290754e-07, "loss": 0.7078, "step": 36842 }, { "epoch": 7.02, "learning_rate": 8.955386081982841e-07, "loss": 1.816, "step": 36843 }, { "epoch": 7.02, "learning_rate": 8.954814108674929e-07, "loss": 1.7057, "step": 36844 }, { "epoch": 7.02, "learning_rate": 8.954242135367017e-07, "loss": 3.2767, "step": 36845 }, { "epoch": 7.02, "learning_rate": 8.953670162059105e-07, "loss": 3.0045, "step": 36846 }, { "epoch": 7.02, "learning_rate": 8.953098188751191e-07, "loss": 1.622, "step": 36847 }, { "epoch": 7.02, "learning_rate": 8.952526215443279e-07, "loss": 1.2798, "step": 36848 }, { "epoch": 7.02, "learning_rate": 8.951954242135367e-07, "loss": 2.7048, "step": 36849 }, { "epoch": 7.02, "learning_rate": 8.951382268827454e-07, "loss": 1.0054, "step": 36850 }, { "epoch": 7.02, "learning_rate": 8.950810295519542e-07, "loss": 0.7835, "step": 36851 }, { "epoch": 7.02, "learning_rate": 8.950238322211631e-07, "loss": 1.6966, "step": 36852 }, { "epoch": 7.02, "learning_rate": 8.949666348903717e-07, "loss": 2.0645, "step": 36853 }, { "epoch": 7.02, "learning_rate": 8.949094375595806e-07, "loss": 0.7366, "step": 36854 }, { "epoch": 7.02, "learning_rate": 8.948522402287894e-07, "loss": 1.0502, "step": 36855 }, { "epoch": 7.02, "learning_rate": 8.947950428979982e-07, "loss": 2.3656, "step": 36856 }, { "epoch": 7.02, "learning_rate": 8.947378455672069e-07, "loss": 1.2317, "step": 36857 }, { "epoch": 7.02, "learning_rate": 8.946806482364157e-07, "loss": 2.0515, "step": 36858 }, { "epoch": 7.02, "learning_rate": 8.946234509056245e-07, "loss": 2.1002, "step": 36859 }, { "epoch": 7.02, "learning_rate": 8.945662535748331e-07, "loss": 2.4538, "step": 36860 }, { "epoch": 7.02, "learning_rate": 8.945090562440419e-07, "loss": 1.8872, "step": 36861 }, { "epoch": 7.02, "learning_rate": 8.944518589132507e-07, "loss": 1.3748, "step": 36862 }, { "epoch": 7.02, "learning_rate": 8.943946615824596e-07, "loss": 1.6287, "step": 36863 }, { "epoch": 7.02, "learning_rate": 8.943374642516682e-07, "loss": 1.9175, "step": 36864 }, { "epoch": 7.02, "learning_rate": 8.94280266920877e-07, "loss": 1.5454, "step": 36865 }, { "epoch": 7.02, "learning_rate": 8.942230695900859e-07, "loss": 1.5538, "step": 36866 }, { "epoch": 7.02, "learning_rate": 8.941658722592945e-07, "loss": 1.6044, "step": 36867 }, { "epoch": 7.02, "learning_rate": 8.941086749285034e-07, "loss": 1.4427, "step": 36868 }, { "epoch": 7.02, "learning_rate": 8.940514775977122e-07, "loss": 0.9411, "step": 36869 }, { "epoch": 7.02, "learning_rate": 8.93994280266921e-07, "loss": 3.3497, "step": 36870 }, { "epoch": 7.02, "learning_rate": 8.939370829361297e-07, "loss": 1.7376, "step": 36871 }, { "epoch": 7.02, "learning_rate": 8.938798856053385e-07, "loss": 1.7985, "step": 36872 }, { "epoch": 7.02, "learning_rate": 8.938226882745473e-07, "loss": 1.7187, "step": 36873 }, { "epoch": 7.02, "learning_rate": 8.937654909437559e-07, "loss": 0.7404, "step": 36874 }, { "epoch": 7.02, "learning_rate": 8.937082936129647e-07, "loss": 1.203, "step": 36875 }, { "epoch": 7.02, "learning_rate": 8.936510962821735e-07, "loss": 1.9136, "step": 36876 }, { "epoch": 7.02, "learning_rate": 8.935938989513822e-07, "loss": 1.6952, "step": 36877 }, { "epoch": 7.02, "learning_rate": 8.93536701620591e-07, "loss": 1.409, "step": 36878 }, { "epoch": 7.02, "learning_rate": 8.934795042897998e-07, "loss": 0.7839, "step": 36879 }, { "epoch": 7.02, "learning_rate": 8.934223069590087e-07, "loss": 0.5947, "step": 36880 }, { "epoch": 7.02, "learning_rate": 8.933651096282173e-07, "loss": 1.3937, "step": 36881 }, { "epoch": 7.03, "learning_rate": 8.933079122974262e-07, "loss": 0.861, "step": 36882 }, { "epoch": 7.03, "learning_rate": 8.93250714966635e-07, "loss": 1.7913, "step": 36883 }, { "epoch": 7.03, "learning_rate": 8.931935176358436e-07, "loss": 1.4437, "step": 36884 }, { "epoch": 7.03, "learning_rate": 8.931363203050525e-07, "loss": 1.918, "step": 36885 }, { "epoch": 7.03, "learning_rate": 8.930791229742613e-07, "loss": 1.3169, "step": 36886 }, { "epoch": 7.03, "learning_rate": 8.9302192564347e-07, "loss": 2.209, "step": 36887 }, { "epoch": 7.03, "learning_rate": 8.929647283126787e-07, "loss": 1.0219, "step": 36888 }, { "epoch": 7.03, "learning_rate": 8.929075309818875e-07, "loss": 1.1989, "step": 36889 }, { "epoch": 7.03, "learning_rate": 8.928503336510963e-07, "loss": 2.8953, "step": 36890 }, { "epoch": 7.03, "learning_rate": 8.92793136320305e-07, "loss": 1.9325, "step": 36891 }, { "epoch": 7.03, "learning_rate": 8.927359389895138e-07, "loss": 1.6948, "step": 36892 }, { "epoch": 7.03, "learning_rate": 8.926787416587226e-07, "loss": 1.6879, "step": 36893 }, { "epoch": 7.03, "learning_rate": 8.926215443279313e-07, "loss": 2.0251, "step": 36894 }, { "epoch": 7.03, "learning_rate": 8.925643469971401e-07, "loss": 1.5984, "step": 36895 }, { "epoch": 7.03, "learning_rate": 8.92507149666349e-07, "loss": 1.7805, "step": 36896 }, { "epoch": 7.03, "learning_rate": 8.924499523355578e-07, "loss": 1.0116, "step": 36897 }, { "epoch": 7.03, "learning_rate": 8.923927550047665e-07, "loss": 2.0972, "step": 36898 }, { "epoch": 7.03, "learning_rate": 8.923355576739753e-07, "loss": 0.8439, "step": 36899 }, { "epoch": 7.03, "learning_rate": 8.922783603431841e-07, "loss": 2.7905, "step": 36900 }, { "epoch": 7.03, "learning_rate": 8.922211630123927e-07, "loss": 1.5065, "step": 36901 }, { "epoch": 7.03, "learning_rate": 8.921639656816015e-07, "loss": 1.5954, "step": 36902 }, { "epoch": 7.03, "learning_rate": 8.921067683508103e-07, "loss": 1.5628, "step": 36903 }, { "epoch": 7.03, "learning_rate": 8.920495710200191e-07, "loss": 2.6338, "step": 36904 }, { "epoch": 7.03, "learning_rate": 8.919923736892278e-07, "loss": 2.6591, "step": 36905 }, { "epoch": 7.03, "learning_rate": 8.919351763584366e-07, "loss": 1.4775, "step": 36906 }, { "epoch": 7.03, "learning_rate": 8.918779790276454e-07, "loss": 1.1289, "step": 36907 }, { "epoch": 7.03, "learning_rate": 8.918207816968541e-07, "loss": 1.9293, "step": 36908 }, { "epoch": 7.03, "learning_rate": 8.917635843660629e-07, "loss": 2.558, "step": 36909 }, { "epoch": 7.03, "learning_rate": 8.917063870352718e-07, "loss": 1.7181, "step": 36910 }, { "epoch": 7.03, "learning_rate": 8.916491897044806e-07, "loss": 0.9281, "step": 36911 }, { "epoch": 7.03, "learning_rate": 8.915919923736893e-07, "loss": 2.0076, "step": 36912 }, { "epoch": 7.03, "learning_rate": 8.915347950428981e-07, "loss": 1.5298, "step": 36913 }, { "epoch": 7.03, "learning_rate": 8.914775977121068e-07, "loss": 1.471, "step": 36914 }, { "epoch": 7.03, "learning_rate": 8.914204003813155e-07, "loss": 0.6994, "step": 36915 }, { "epoch": 7.03, "learning_rate": 8.913632030505243e-07, "loss": 2.2963, "step": 36916 }, { "epoch": 7.03, "learning_rate": 8.913060057197331e-07, "loss": 2.6633, "step": 36917 }, { "epoch": 7.03, "learning_rate": 8.912488083889418e-07, "loss": 1.4839, "step": 36918 }, { "epoch": 7.03, "learning_rate": 8.911916110581506e-07, "loss": 1.7769, "step": 36919 }, { "epoch": 7.03, "learning_rate": 8.911344137273594e-07, "loss": 1.0501, "step": 36920 }, { "epoch": 7.03, "learning_rate": 8.910772163965683e-07, "loss": 3.3346, "step": 36921 }, { "epoch": 7.03, "learning_rate": 8.910200190657769e-07, "loss": 1.7008, "step": 36922 }, { "epoch": 7.03, "learning_rate": 8.909628217349857e-07, "loss": 1.5173, "step": 36923 }, { "epoch": 7.03, "learning_rate": 8.909056244041946e-07, "loss": 1.4363, "step": 36924 }, { "epoch": 7.03, "learning_rate": 8.908484270734032e-07, "loss": 1.7455, "step": 36925 }, { "epoch": 7.03, "learning_rate": 8.907912297426121e-07, "loss": 1.0036, "step": 36926 }, { "epoch": 7.03, "learning_rate": 8.907340324118209e-07, "loss": 2.9717, "step": 36927 }, { "epoch": 7.03, "learning_rate": 8.906768350810296e-07, "loss": 1.4279, "step": 36928 }, { "epoch": 7.03, "learning_rate": 8.906196377502383e-07, "loss": 1.2837, "step": 36929 }, { "epoch": 7.03, "learning_rate": 8.905624404194471e-07, "loss": 1.5101, "step": 36930 }, { "epoch": 7.03, "learning_rate": 8.905052430886559e-07, "loss": 1.7814, "step": 36931 }, { "epoch": 7.03, "learning_rate": 8.904480457578646e-07, "loss": 2.4934, "step": 36932 }, { "epoch": 7.03, "learning_rate": 8.903908484270734e-07, "loss": 1.8032, "step": 36933 }, { "epoch": 7.04, "learning_rate": 8.903336510962822e-07, "loss": 0.2915, "step": 36934 }, { "epoch": 7.04, "learning_rate": 8.902764537654909e-07, "loss": 1.9406, "step": 36935 }, { "epoch": 7.04, "learning_rate": 8.902192564346997e-07, "loss": 1.7287, "step": 36936 }, { "epoch": 7.04, "learning_rate": 8.901620591039085e-07, "loss": 0.865, "step": 36937 }, { "epoch": 7.04, "learning_rate": 8.901048617731174e-07, "loss": 0.8669, "step": 36938 }, { "epoch": 7.04, "learning_rate": 8.90047664442326e-07, "loss": 1.5788, "step": 36939 }, { "epoch": 7.04, "learning_rate": 8.899904671115349e-07, "loss": 1.6267, "step": 36940 }, { "epoch": 7.04, "learning_rate": 8.899332697807436e-07, "loss": 0.924, "step": 36941 }, { "epoch": 7.04, "learning_rate": 8.898760724499522e-07, "loss": 1.4483, "step": 36942 }, { "epoch": 7.04, "learning_rate": 8.898188751191611e-07, "loss": 1.8981, "step": 36943 }, { "epoch": 7.04, "learning_rate": 8.897616777883699e-07, "loss": 0.3625, "step": 36944 }, { "epoch": 7.04, "learning_rate": 8.897044804575787e-07, "loss": 1.502, "step": 36945 }, { "epoch": 7.04, "learning_rate": 8.896472831267874e-07, "loss": 1.0757, "step": 36946 }, { "epoch": 7.04, "learning_rate": 8.895900857959962e-07, "loss": 0.5888, "step": 36947 }, { "epoch": 7.04, "learning_rate": 8.89532888465205e-07, "loss": 0.8882, "step": 36948 }, { "epoch": 7.04, "learning_rate": 8.894756911344137e-07, "loss": 2.8148, "step": 36949 }, { "epoch": 7.04, "learning_rate": 8.894184938036225e-07, "loss": 0.9448, "step": 36950 }, { "epoch": 7.04, "learning_rate": 8.893612964728313e-07, "loss": 1.5104, "step": 36951 }, { "epoch": 7.04, "learning_rate": 8.893040991420402e-07, "loss": 2.0548, "step": 36952 }, { "epoch": 7.04, "learning_rate": 8.892469018112488e-07, "loss": 1.8939, "step": 36953 }, { "epoch": 7.04, "learning_rate": 8.891897044804576e-07, "loss": 1.6387, "step": 36954 }, { "epoch": 7.04, "learning_rate": 8.891325071496664e-07, "loss": 1.5521, "step": 36955 }, { "epoch": 7.04, "learning_rate": 8.890753098188751e-07, "loss": 3.254, "step": 36956 }, { "epoch": 7.04, "learning_rate": 8.890181124880839e-07, "loss": 1.5989, "step": 36957 }, { "epoch": 7.04, "learning_rate": 8.889609151572927e-07, "loss": 2.4643, "step": 36958 }, { "epoch": 7.04, "learning_rate": 8.889037178265014e-07, "loss": 2.6168, "step": 36959 }, { "epoch": 7.04, "learning_rate": 8.888465204957102e-07, "loss": 3.2671, "step": 36960 }, { "epoch": 7.04, "learning_rate": 8.88789323164919e-07, "loss": 0.8305, "step": 36961 }, { "epoch": 7.04, "learning_rate": 8.887321258341278e-07, "loss": 1.2687, "step": 36962 }, { "epoch": 7.04, "learning_rate": 8.886749285033365e-07, "loss": 1.7324, "step": 36963 }, { "epoch": 7.04, "learning_rate": 8.886177311725453e-07, "loss": 1.6801, "step": 36964 }, { "epoch": 7.04, "learning_rate": 8.885605338417542e-07, "loss": 0.6027, "step": 36965 }, { "epoch": 7.04, "learning_rate": 8.885033365109628e-07, "loss": 0.5025, "step": 36966 }, { "epoch": 7.04, "learning_rate": 8.884461391801715e-07, "loss": 0.549, "step": 36967 }, { "epoch": 7.04, "learning_rate": 8.883889418493804e-07, "loss": 3.6972, "step": 36968 }, { "epoch": 7.04, "learning_rate": 8.883317445185892e-07, "loss": 1.8452, "step": 36969 }, { "epoch": 7.04, "learning_rate": 8.882745471877979e-07, "loss": 1.958, "step": 36970 }, { "epoch": 7.04, "learning_rate": 8.882173498570067e-07, "loss": 2.074, "step": 36971 }, { "epoch": 7.04, "learning_rate": 8.881601525262155e-07, "loss": 2.6028, "step": 36972 }, { "epoch": 7.04, "learning_rate": 8.881029551954242e-07, "loss": 3.1637, "step": 36973 }, { "epoch": 7.04, "learning_rate": 8.88045757864633e-07, "loss": 0.8531, "step": 36974 }, { "epoch": 7.04, "learning_rate": 8.879885605338418e-07, "loss": 1.1885, "step": 36975 }, { "epoch": 7.04, "learning_rate": 8.879313632030506e-07, "loss": 1.7638, "step": 36976 }, { "epoch": 7.04, "learning_rate": 8.878741658722593e-07, "loss": 3.1991, "step": 36977 }, { "epoch": 7.04, "learning_rate": 8.878169685414681e-07, "loss": 1.5825, "step": 36978 }, { "epoch": 7.04, "learning_rate": 8.87759771210677e-07, "loss": 2.5616, "step": 36979 }, { "epoch": 7.04, "learning_rate": 8.877025738798855e-07, "loss": 2.4732, "step": 36980 }, { "epoch": 7.04, "learning_rate": 8.876453765490943e-07, "loss": 3.0028, "step": 36981 }, { "epoch": 7.04, "learning_rate": 8.875881792183032e-07, "loss": 2.1685, "step": 36982 }, { "epoch": 7.04, "learning_rate": 8.875309818875118e-07, "loss": 1.3008, "step": 36983 }, { "epoch": 7.04, "learning_rate": 8.874737845567207e-07, "loss": 0.435, "step": 36984 }, { "epoch": 7.04, "learning_rate": 8.874165872259295e-07, "loss": 1.0486, "step": 36985 }, { "epoch": 7.04, "learning_rate": 8.873593898951383e-07, "loss": 1.0165, "step": 36986 }, { "epoch": 7.05, "learning_rate": 8.87302192564347e-07, "loss": 1.4194, "step": 36987 }, { "epoch": 7.05, "learning_rate": 8.872449952335558e-07, "loss": 0.8076, "step": 36988 }, { "epoch": 7.05, "learning_rate": 8.871877979027646e-07, "loss": 1.753, "step": 36989 }, { "epoch": 7.05, "learning_rate": 8.871306005719733e-07, "loss": 1.2076, "step": 36990 }, { "epoch": 7.05, "learning_rate": 8.870734032411821e-07, "loss": 2.2738, "step": 36991 }, { "epoch": 7.05, "learning_rate": 8.870162059103909e-07, "loss": 1.3708, "step": 36992 }, { "epoch": 7.05, "learning_rate": 8.869590085795998e-07, "loss": 2.1618, "step": 36993 }, { "epoch": 7.05, "learning_rate": 8.869018112488083e-07, "loss": 1.0145, "step": 36994 }, { "epoch": 7.05, "learning_rate": 8.868446139180171e-07, "loss": 0.5507, "step": 36995 }, { "epoch": 7.05, "learning_rate": 8.86787416587226e-07, "loss": 2.4807, "step": 36996 }, { "epoch": 7.05, "learning_rate": 8.867302192564346e-07, "loss": 0.8819, "step": 36997 }, { "epoch": 7.05, "learning_rate": 8.866730219256435e-07, "loss": 1.116, "step": 36998 }, { "epoch": 7.05, "learning_rate": 8.866158245948523e-07, "loss": 1.7699, "step": 36999 }, { "epoch": 7.05, "learning_rate": 8.86558627264061e-07, "loss": 3.3387, "step": 37000 }, { "epoch": 7.05, "learning_rate": 8.865014299332698e-07, "loss": 2.1761, "step": 37001 }, { "epoch": 7.05, "learning_rate": 8.864442326024786e-07, "loss": 1.0252, "step": 37002 }, { "epoch": 7.05, "learning_rate": 8.863870352716874e-07, "loss": 1.3058, "step": 37003 }, { "epoch": 7.05, "learning_rate": 8.863298379408961e-07, "loss": 1.1583, "step": 37004 }, { "epoch": 7.05, "learning_rate": 8.862726406101049e-07, "loss": 1.9015, "step": 37005 }, { "epoch": 7.05, "learning_rate": 8.862154432793137e-07, "loss": 1.7482, "step": 37006 }, { "epoch": 7.05, "learning_rate": 8.861582459485223e-07, "loss": 2.7031, "step": 37007 }, { "epoch": 7.05, "learning_rate": 8.861010486177311e-07, "loss": 1.0547, "step": 37008 }, { "epoch": 7.05, "learning_rate": 8.8604385128694e-07, "loss": 0.5176, "step": 37009 }, { "epoch": 7.05, "learning_rate": 8.859866539561488e-07, "loss": 1.5589, "step": 37010 }, { "epoch": 7.05, "learning_rate": 8.859294566253574e-07, "loss": 2.4941, "step": 37011 }, { "epoch": 7.05, "learning_rate": 8.858722592945663e-07, "loss": 2.8115, "step": 37012 }, { "epoch": 7.05, "learning_rate": 8.858150619637751e-07, "loss": 2.873, "step": 37013 }, { "epoch": 7.05, "learning_rate": 8.857578646329838e-07, "loss": 1.1603, "step": 37014 }, { "epoch": 7.05, "learning_rate": 8.857006673021926e-07, "loss": 1.0926, "step": 37015 }, { "epoch": 7.05, "learning_rate": 8.856434699714014e-07, "loss": 0.7978, "step": 37016 }, { "epoch": 7.05, "learning_rate": 8.855862726406102e-07, "loss": 0.9275, "step": 37017 }, { "epoch": 7.05, "learning_rate": 8.855290753098189e-07, "loss": 1.3248, "step": 37018 }, { "epoch": 7.05, "learning_rate": 8.854718779790277e-07, "loss": 1.3456, "step": 37019 }, { "epoch": 7.05, "learning_rate": 8.854146806482365e-07, "loss": 1.2443, "step": 37020 }, { "epoch": 7.05, "learning_rate": 8.853574833174451e-07, "loss": 0.8543, "step": 37021 }, { "epoch": 7.05, "learning_rate": 8.853002859866539e-07, "loss": 1.5069, "step": 37022 }, { "epoch": 7.05, "learning_rate": 8.852430886558628e-07, "loss": 2.2088, "step": 37023 }, { "epoch": 7.05, "learning_rate": 8.851858913250714e-07, "loss": 0.8448, "step": 37024 }, { "epoch": 7.05, "learning_rate": 8.851286939942802e-07, "loss": 0.8321, "step": 37025 }, { "epoch": 7.05, "learning_rate": 8.850714966634891e-07, "loss": 1.4264, "step": 37026 }, { "epoch": 7.05, "learning_rate": 8.850142993326979e-07, "loss": 2.2585, "step": 37027 }, { "epoch": 7.05, "learning_rate": 8.849571020019066e-07, "loss": 1.0675, "step": 37028 }, { "epoch": 7.05, "learning_rate": 8.848999046711154e-07, "loss": 1.4331, "step": 37029 }, { "epoch": 7.05, "learning_rate": 8.848427073403242e-07, "loss": 2.0147, "step": 37030 }, { "epoch": 7.05, "learning_rate": 8.847855100095329e-07, "loss": 1.3595, "step": 37031 }, { "epoch": 7.05, "learning_rate": 8.847283126787417e-07, "loss": 0.4476, "step": 37032 }, { "epoch": 7.05, "learning_rate": 8.846711153479505e-07, "loss": 1.6419, "step": 37033 }, { "epoch": 7.05, "learning_rate": 8.846139180171592e-07, "loss": 2.0857, "step": 37034 }, { "epoch": 7.05, "learning_rate": 8.845567206863679e-07, "loss": 2.1248, "step": 37035 }, { "epoch": 7.05, "learning_rate": 8.844995233555767e-07, "loss": 2.921, "step": 37036 }, { "epoch": 7.05, "learning_rate": 8.844423260247856e-07, "loss": 2.9144, "step": 37037 }, { "epoch": 7.05, "learning_rate": 8.843851286939942e-07, "loss": 2.1065, "step": 37038 }, { "epoch": 7.06, "learning_rate": 8.84327931363203e-07, "loss": 0.5575, "step": 37039 }, { "epoch": 7.06, "learning_rate": 8.842707340324119e-07, "loss": 2.8487, "step": 37040 }, { "epoch": 7.06, "learning_rate": 8.842135367016207e-07, "loss": 2.003, "step": 37041 }, { "epoch": 7.06, "learning_rate": 8.841563393708294e-07, "loss": 1.8117, "step": 37042 }, { "epoch": 7.06, "learning_rate": 8.840991420400382e-07, "loss": 2.4196, "step": 37043 }, { "epoch": 7.06, "learning_rate": 8.84041944709247e-07, "loss": 1.925, "step": 37044 }, { "epoch": 7.06, "learning_rate": 8.839847473784557e-07, "loss": 1.2935, "step": 37045 }, { "epoch": 7.06, "learning_rate": 8.839275500476645e-07, "loss": 2.4595, "step": 37046 }, { "epoch": 7.06, "learning_rate": 8.838703527168733e-07, "loss": 1.5226, "step": 37047 }, { "epoch": 7.06, "learning_rate": 8.838131553860819e-07, "loss": 2.4835, "step": 37048 }, { "epoch": 7.06, "learning_rate": 8.837559580552907e-07, "loss": 1.4442, "step": 37049 }, { "epoch": 7.06, "learning_rate": 8.836987607244995e-07, "loss": 1.5041, "step": 37050 }, { "epoch": 7.06, "learning_rate": 8.836415633937084e-07, "loss": 1.3223, "step": 37051 }, { "epoch": 7.06, "learning_rate": 8.83584366062917e-07, "loss": 1.432, "step": 37052 }, { "epoch": 7.06, "learning_rate": 8.835271687321259e-07, "loss": 1.1155, "step": 37053 }, { "epoch": 7.06, "learning_rate": 8.834699714013347e-07, "loss": 1.7083, "step": 37054 }, { "epoch": 7.06, "learning_rate": 8.834127740705433e-07, "loss": 1.182, "step": 37055 }, { "epoch": 7.06, "learning_rate": 8.833555767397522e-07, "loss": 2.2086, "step": 37056 }, { "epoch": 7.06, "learning_rate": 8.83298379408961e-07, "loss": 1.8108, "step": 37057 }, { "epoch": 7.06, "learning_rate": 8.832411820781698e-07, "loss": 0.9362, "step": 37058 }, { "epoch": 7.06, "learning_rate": 8.831839847473785e-07, "loss": 1.0, "step": 37059 }, { "epoch": 7.06, "learning_rate": 8.831267874165873e-07, "loss": 2.2652, "step": 37060 }, { "epoch": 7.06, "learning_rate": 8.83069590085796e-07, "loss": 2.5476, "step": 37061 }, { "epoch": 7.06, "learning_rate": 8.830123927550047e-07, "loss": 1.901, "step": 37062 }, { "epoch": 7.06, "learning_rate": 8.829551954242135e-07, "loss": 1.3736, "step": 37063 }, { "epoch": 7.06, "learning_rate": 8.828979980934223e-07, "loss": 3.3763, "step": 37064 }, { "epoch": 7.06, "learning_rate": 8.82840800762631e-07, "loss": 1.9967, "step": 37065 }, { "epoch": 7.06, "learning_rate": 8.827836034318398e-07, "loss": 1.6407, "step": 37066 }, { "epoch": 7.06, "learning_rate": 8.827264061010487e-07, "loss": 1.8359, "step": 37067 }, { "epoch": 7.06, "learning_rate": 8.826692087702575e-07, "loss": 2.2069, "step": 37068 }, { "epoch": 7.06, "learning_rate": 8.826120114394661e-07, "loss": 2.93, "step": 37069 }, { "epoch": 7.06, "learning_rate": 8.82554814108675e-07, "loss": 2.5345, "step": 37070 }, { "epoch": 7.06, "learning_rate": 8.824976167778838e-07, "loss": 1.3605, "step": 37071 }, { "epoch": 7.06, "learning_rate": 8.824404194470925e-07, "loss": 1.096, "step": 37072 }, { "epoch": 7.06, "learning_rate": 8.823832221163013e-07, "loss": 1.5757, "step": 37073 }, { "epoch": 7.06, "learning_rate": 8.8232602478551e-07, "loss": 2.3748, "step": 37074 }, { "epoch": 7.06, "learning_rate": 8.822688274547188e-07, "loss": 1.7144, "step": 37075 }, { "epoch": 7.06, "learning_rate": 8.822116301239275e-07, "loss": 0.6107, "step": 37076 }, { "epoch": 7.06, "learning_rate": 8.821544327931363e-07, "loss": 0.9627, "step": 37077 }, { "epoch": 7.06, "learning_rate": 8.820972354623451e-07, "loss": 1.803, "step": 37078 }, { "epoch": 7.06, "learning_rate": 8.820400381315538e-07, "loss": 1.3048, "step": 37079 }, { "epoch": 7.06, "learning_rate": 8.819828408007626e-07, "loss": 3.2021, "step": 37080 }, { "epoch": 7.06, "learning_rate": 8.819256434699715e-07, "loss": 1.2815, "step": 37081 }, { "epoch": 7.06, "learning_rate": 8.818684461391803e-07, "loss": 2.1927, "step": 37082 }, { "epoch": 7.06, "learning_rate": 8.818112488083889e-07, "loss": 1.9497, "step": 37083 }, { "epoch": 7.06, "learning_rate": 8.817540514775978e-07, "loss": 1.7748, "step": 37084 }, { "epoch": 7.06, "learning_rate": 8.816968541468066e-07, "loss": 1.4483, "step": 37085 }, { "epoch": 7.06, "learning_rate": 8.816396568160153e-07, "loss": 1.2284, "step": 37086 }, { "epoch": 7.06, "learning_rate": 8.81582459485224e-07, "loss": 0.4731, "step": 37087 }, { "epoch": 7.06, "learning_rate": 8.815252621544328e-07, "loss": 1.4422, "step": 37088 }, { "epoch": 7.06, "learning_rate": 8.814680648236415e-07, "loss": 1.1764, "step": 37089 }, { "epoch": 7.06, "learning_rate": 8.814108674928503e-07, "loss": 2.3617, "step": 37090 }, { "epoch": 7.06, "learning_rate": 8.813536701620591e-07, "loss": 2.5718, "step": 37091 }, { "epoch": 7.07, "learning_rate": 8.812964728312679e-07, "loss": 1.5149, "step": 37092 }, { "epoch": 7.07, "learning_rate": 8.812392755004766e-07, "loss": 0.4632, "step": 37093 }, { "epoch": 7.07, "learning_rate": 8.811820781696854e-07, "loss": 2.3091, "step": 37094 }, { "epoch": 7.07, "learning_rate": 8.811248808388943e-07, "loss": 2.0255, "step": 37095 }, { "epoch": 7.07, "learning_rate": 8.810676835081029e-07, "loss": 1.0053, "step": 37096 }, { "epoch": 7.07, "learning_rate": 8.810104861773118e-07, "loss": 1.7688, "step": 37097 }, { "epoch": 7.07, "learning_rate": 8.809532888465206e-07, "loss": 2.4283, "step": 37098 }, { "epoch": 7.07, "learning_rate": 8.808960915157294e-07, "loss": 1.8887, "step": 37099 }, { "epoch": 7.07, "learning_rate": 8.808388941849381e-07, "loss": 2.1501, "step": 37100 }, { "epoch": 7.07, "learning_rate": 8.807816968541468e-07, "loss": 2.3586, "step": 37101 }, { "epoch": 7.07, "learning_rate": 8.807244995233556e-07, "loss": 2.3318, "step": 37102 }, { "epoch": 7.07, "learning_rate": 8.806673021925643e-07, "loss": 0.7292, "step": 37103 }, { "epoch": 7.07, "learning_rate": 8.806101048617731e-07, "loss": 1.8629, "step": 37104 }, { "epoch": 7.07, "learning_rate": 8.805529075309819e-07, "loss": 1.774, "step": 37105 }, { "epoch": 7.07, "learning_rate": 8.804957102001907e-07, "loss": 0.4966, "step": 37106 }, { "epoch": 7.07, "learning_rate": 8.804385128693994e-07, "loss": 1.4934, "step": 37107 }, { "epoch": 7.07, "learning_rate": 8.803813155386082e-07, "loss": 1.9583, "step": 37108 }, { "epoch": 7.07, "learning_rate": 8.803241182078171e-07, "loss": 1.5026, "step": 37109 }, { "epoch": 7.07, "learning_rate": 8.802669208770257e-07, "loss": 1.3869, "step": 37110 }, { "epoch": 7.07, "learning_rate": 8.802097235462346e-07, "loss": 1.5044, "step": 37111 }, { "epoch": 7.07, "learning_rate": 8.801525262154434e-07, "loss": 1.1661, "step": 37112 }, { "epoch": 7.07, "learning_rate": 8.80095328884652e-07, "loss": 0.4307, "step": 37113 }, { "epoch": 7.07, "learning_rate": 8.800381315538608e-07, "loss": 1.329, "step": 37114 }, { "epoch": 7.07, "learning_rate": 8.799809342230696e-07, "loss": 1.2921, "step": 37115 }, { "epoch": 7.07, "learning_rate": 8.799237368922784e-07, "loss": 0.9385, "step": 37116 }, { "epoch": 7.07, "learning_rate": 8.798665395614871e-07, "loss": 0.6668, "step": 37117 }, { "epoch": 7.07, "learning_rate": 8.798093422306959e-07, "loss": 0.3863, "step": 37118 }, { "epoch": 7.07, "learning_rate": 8.797521448999047e-07, "loss": 0.7889, "step": 37119 }, { "epoch": 7.07, "learning_rate": 8.796949475691134e-07, "loss": 1.7591, "step": 37120 }, { "epoch": 7.07, "learning_rate": 8.796377502383222e-07, "loss": 1.0967, "step": 37121 }, { "epoch": 7.07, "learning_rate": 8.79580552907531e-07, "loss": 1.8117, "step": 37122 }, { "epoch": 7.07, "learning_rate": 8.795233555767399e-07, "loss": 1.7625, "step": 37123 }, { "epoch": 7.07, "learning_rate": 8.794661582459485e-07, "loss": 1.7149, "step": 37124 }, { "epoch": 7.07, "learning_rate": 8.794089609151574e-07, "loss": 1.8992, "step": 37125 }, { "epoch": 7.07, "learning_rate": 8.793517635843662e-07, "loss": 2.4925, "step": 37126 }, { "epoch": 7.07, "learning_rate": 8.792945662535747e-07, "loss": 2.336, "step": 37127 }, { "epoch": 7.07, "learning_rate": 8.792373689227836e-07, "loss": 1.0004, "step": 37128 }, { "epoch": 7.07, "learning_rate": 8.791801715919924e-07, "loss": 1.7323, "step": 37129 }, { "epoch": 7.07, "learning_rate": 8.791229742612011e-07, "loss": 1.1384, "step": 37130 }, { "epoch": 7.07, "learning_rate": 8.790657769304099e-07, "loss": 2.0524, "step": 37131 }, { "epoch": 7.07, "learning_rate": 8.790085795996187e-07, "loss": 1.8049, "step": 37132 }, { "epoch": 7.07, "learning_rate": 8.789513822688275e-07, "loss": 1.3012, "step": 37133 }, { "epoch": 7.07, "learning_rate": 8.788941849380362e-07, "loss": 3.7769, "step": 37134 }, { "epoch": 7.07, "learning_rate": 8.78836987607245e-07, "loss": 1.8273, "step": 37135 }, { "epoch": 7.07, "learning_rate": 8.787797902764538e-07, "loss": 0.974, "step": 37136 }, { "epoch": 7.07, "learning_rate": 8.787225929456625e-07, "loss": 1.727, "step": 37137 }, { "epoch": 7.07, "learning_rate": 8.786653956148713e-07, "loss": 2.6568, "step": 37138 }, { "epoch": 7.07, "learning_rate": 8.786081982840802e-07, "loss": 1.6238, "step": 37139 }, { "epoch": 7.07, "learning_rate": 8.78551000953289e-07, "loss": 0.8486, "step": 37140 }, { "epoch": 7.07, "learning_rate": 8.784938036224975e-07, "loss": 1.6791, "step": 37141 }, { "epoch": 7.07, "learning_rate": 8.784366062917064e-07, "loss": 1.5404, "step": 37142 }, { "epoch": 7.07, "learning_rate": 8.783794089609152e-07, "loss": 2.321, "step": 37143 }, { "epoch": 7.08, "learning_rate": 8.783222116301239e-07, "loss": 1.9835, "step": 37144 }, { "epoch": 7.08, "learning_rate": 8.782650142993327e-07, "loss": 0.6369, "step": 37145 }, { "epoch": 7.08, "learning_rate": 8.782078169685415e-07, "loss": 1.6301, "step": 37146 }, { "epoch": 7.08, "learning_rate": 8.781506196377503e-07, "loss": 1.2352, "step": 37147 }, { "epoch": 7.08, "learning_rate": 8.78093422306959e-07, "loss": 1.3788, "step": 37148 }, { "epoch": 7.08, "learning_rate": 8.780362249761678e-07, "loss": 1.7807, "step": 37149 }, { "epoch": 7.08, "learning_rate": 8.779790276453766e-07, "loss": 2.1782, "step": 37150 }, { "epoch": 7.08, "learning_rate": 8.779218303145853e-07, "loss": 0.8936, "step": 37151 }, { "epoch": 7.08, "learning_rate": 8.778646329837941e-07, "loss": 0.6011, "step": 37152 }, { "epoch": 7.08, "learning_rate": 8.77807435653003e-07, "loss": 2.0285, "step": 37153 }, { "epoch": 7.08, "learning_rate": 8.777502383222115e-07, "loss": 0.5171, "step": 37154 }, { "epoch": 7.08, "learning_rate": 8.776930409914204e-07, "loss": 0.7027, "step": 37155 }, { "epoch": 7.08, "learning_rate": 8.776358436606292e-07, "loss": 1.8306, "step": 37156 }, { "epoch": 7.08, "learning_rate": 8.77578646329838e-07, "loss": 2.0359, "step": 37157 }, { "epoch": 7.08, "learning_rate": 8.775214489990467e-07, "loss": 3.3623, "step": 37158 }, { "epoch": 7.08, "learning_rate": 8.774642516682555e-07, "loss": 2.3723, "step": 37159 }, { "epoch": 7.08, "learning_rate": 8.774070543374643e-07, "loss": 1.8554, "step": 37160 }, { "epoch": 7.08, "learning_rate": 8.77349857006673e-07, "loss": 0.742, "step": 37161 }, { "epoch": 7.08, "learning_rate": 8.772926596758818e-07, "loss": 1.0208, "step": 37162 }, { "epoch": 7.08, "learning_rate": 8.772354623450906e-07, "loss": 2.7572, "step": 37163 }, { "epoch": 7.08, "learning_rate": 8.771782650142995e-07, "loss": 0.803, "step": 37164 }, { "epoch": 7.08, "learning_rate": 8.771210676835081e-07, "loss": 1.355, "step": 37165 }, { "epoch": 7.08, "learning_rate": 8.770638703527169e-07, "loss": 0.5691, "step": 37166 }, { "epoch": 7.08, "learning_rate": 8.770066730219258e-07, "loss": 0.9467, "step": 37167 }, { "epoch": 7.08, "learning_rate": 8.769494756911343e-07, "loss": 1.1937, "step": 37168 }, { "epoch": 7.08, "learning_rate": 8.768922783603432e-07, "loss": 1.3377, "step": 37169 }, { "epoch": 7.08, "learning_rate": 8.76835081029552e-07, "loss": 0.4269, "step": 37170 }, { "epoch": 7.08, "learning_rate": 8.767778836987608e-07, "loss": 0.5344, "step": 37171 }, { "epoch": 7.08, "learning_rate": 8.767206863679695e-07, "loss": 2.1732, "step": 37172 }, { "epoch": 7.08, "learning_rate": 8.766634890371783e-07, "loss": 0.9705, "step": 37173 }, { "epoch": 7.08, "learning_rate": 8.766062917063871e-07, "loss": 1.4834, "step": 37174 }, { "epoch": 7.08, "learning_rate": 8.765490943755958e-07, "loss": 1.9116, "step": 37175 }, { "epoch": 7.08, "learning_rate": 8.764918970448046e-07, "loss": 2.177, "step": 37176 }, { "epoch": 7.08, "learning_rate": 8.764346997140134e-07, "loss": 2.5294, "step": 37177 }, { "epoch": 7.08, "learning_rate": 8.763775023832221e-07, "loss": 1.5646, "step": 37178 }, { "epoch": 7.08, "learning_rate": 8.763203050524309e-07, "loss": 0.4226, "step": 37179 }, { "epoch": 7.08, "learning_rate": 8.762631077216397e-07, "loss": 2.5211, "step": 37180 }, { "epoch": 7.08, "learning_rate": 8.762059103908486e-07, "loss": 2.5367, "step": 37181 }, { "epoch": 7.08, "learning_rate": 8.761487130600571e-07, "loss": 1.6779, "step": 37182 }, { "epoch": 7.08, "learning_rate": 8.76091515729266e-07, "loss": 1.4948, "step": 37183 }, { "epoch": 7.08, "learning_rate": 8.760343183984748e-07, "loss": 1.5671, "step": 37184 }, { "epoch": 7.08, "learning_rate": 8.759771210676834e-07, "loss": 1.719, "step": 37185 }, { "epoch": 7.08, "learning_rate": 8.759199237368923e-07, "loss": 1.8513, "step": 37186 }, { "epoch": 7.08, "learning_rate": 8.758627264061011e-07, "loss": 1.8718, "step": 37187 }, { "epoch": 7.08, "learning_rate": 8.758055290753099e-07, "loss": 1.065, "step": 37188 }, { "epoch": 7.08, "learning_rate": 8.757483317445186e-07, "loss": 1.7169, "step": 37189 }, { "epoch": 7.08, "learning_rate": 8.756911344137274e-07, "loss": 2.1588, "step": 37190 }, { "epoch": 7.08, "learning_rate": 8.756339370829362e-07, "loss": 1.0172, "step": 37191 }, { "epoch": 7.08, "learning_rate": 8.755767397521449e-07, "loss": 2.0089, "step": 37192 }, { "epoch": 7.08, "learning_rate": 8.755195424213537e-07, "loss": 0.9688, "step": 37193 }, { "epoch": 7.08, "learning_rate": 8.754623450905624e-07, "loss": 1.6687, "step": 37194 }, { "epoch": 7.08, "learning_rate": 8.754051477597711e-07, "loss": 1.6308, "step": 37195 }, { "epoch": 7.08, "learning_rate": 8.753479504289799e-07, "loss": 0.8548, "step": 37196 }, { "epoch": 7.09, "learning_rate": 8.752907530981888e-07, "loss": 2.9041, "step": 37197 }, { "epoch": 7.09, "learning_rate": 8.752335557673976e-07, "loss": 0.9904, "step": 37198 }, { "epoch": 7.09, "learning_rate": 8.751763584366063e-07, "loss": 1.1954, "step": 37199 }, { "epoch": 7.09, "learning_rate": 8.751191611058151e-07, "loss": 1.3095, "step": 37200 }, { "epoch": 7.09, "learning_rate": 8.750619637750239e-07, "loss": 1.4047, "step": 37201 }, { "epoch": 7.09, "learning_rate": 8.750047664442326e-07, "loss": 3.044, "step": 37202 }, { "epoch": 7.09, "learning_rate": 8.749475691134414e-07, "loss": 1.0623, "step": 37203 }, { "epoch": 7.09, "learning_rate": 8.748903717826502e-07, "loss": 2.5638, "step": 37204 }, { "epoch": 7.09, "learning_rate": 8.74833174451859e-07, "loss": 2.2635, "step": 37205 }, { "epoch": 7.09, "learning_rate": 8.747759771210677e-07, "loss": 2.8259, "step": 37206 }, { "epoch": 7.09, "learning_rate": 8.747187797902765e-07, "loss": 4.6047, "step": 37207 }, { "epoch": 7.09, "learning_rate": 8.746615824594854e-07, "loss": 1.5643, "step": 37208 }, { "epoch": 7.09, "learning_rate": 8.746043851286939e-07, "loss": 1.1679, "step": 37209 }, { "epoch": 7.09, "learning_rate": 8.745471877979027e-07, "loss": 0.9602, "step": 37210 }, { "epoch": 7.09, "learning_rate": 8.744899904671116e-07, "loss": 0.898, "step": 37211 }, { "epoch": 7.09, "learning_rate": 8.744327931363204e-07, "loss": 2.3599, "step": 37212 }, { "epoch": 7.09, "learning_rate": 8.743755958055291e-07, "loss": 1.2715, "step": 37213 }, { "epoch": 7.09, "learning_rate": 8.743183984747379e-07, "loss": 1.9629, "step": 37214 }, { "epoch": 7.09, "learning_rate": 8.742612011439467e-07, "loss": 2.3173, "step": 37215 }, { "epoch": 7.09, "learning_rate": 8.742040038131554e-07, "loss": 1.9958, "step": 37216 }, { "epoch": 7.09, "learning_rate": 8.741468064823642e-07, "loss": 2.5972, "step": 37217 }, { "epoch": 7.09, "learning_rate": 8.74089609151573e-07, "loss": 2.1466, "step": 37218 }, { "epoch": 7.09, "learning_rate": 8.740324118207817e-07, "loss": 1.756, "step": 37219 }, { "epoch": 7.09, "learning_rate": 8.739752144899905e-07, "loss": 1.957, "step": 37220 }, { "epoch": 7.09, "learning_rate": 8.739180171591992e-07, "loss": 2.7503, "step": 37221 }, { "epoch": 7.09, "learning_rate": 8.738608198284081e-07, "loss": 2.2376, "step": 37222 }, { "epoch": 7.09, "learning_rate": 8.738036224976167e-07, "loss": 1.7811, "step": 37223 }, { "epoch": 7.09, "learning_rate": 8.737464251668255e-07, "loss": 0.7659, "step": 37224 }, { "epoch": 7.09, "learning_rate": 8.736892278360344e-07, "loss": 1.599, "step": 37225 }, { "epoch": 7.09, "learning_rate": 8.73632030505243e-07, "loss": 2.3323, "step": 37226 }, { "epoch": 7.09, "learning_rate": 8.735748331744519e-07, "loss": 1.7655, "step": 37227 }, { "epoch": 7.09, "learning_rate": 8.735176358436607e-07, "loss": 1.8671, "step": 37228 }, { "epoch": 7.09, "learning_rate": 8.734604385128695e-07, "loss": 0.7918, "step": 37229 }, { "epoch": 7.09, "learning_rate": 8.734032411820782e-07, "loss": 1.9869, "step": 37230 }, { "epoch": 7.09, "learning_rate": 8.73346043851287e-07, "loss": 3.3347, "step": 37231 }, { "epoch": 7.09, "learning_rate": 8.732888465204958e-07, "loss": 3.1, "step": 37232 }, { "epoch": 7.09, "learning_rate": 8.732316491897045e-07, "loss": 1.6822, "step": 37233 }, { "epoch": 7.09, "learning_rate": 8.731744518589132e-07, "loss": 1.6151, "step": 37234 }, { "epoch": 7.09, "learning_rate": 8.73117254528122e-07, "loss": 1.2878, "step": 37235 }, { "epoch": 7.09, "learning_rate": 8.730600571973309e-07, "loss": 1.9924, "step": 37236 }, { "epoch": 7.09, "learning_rate": 8.730028598665395e-07, "loss": 2.2315, "step": 37237 }, { "epoch": 7.09, "learning_rate": 8.729456625357483e-07, "loss": 1.2122, "step": 37238 }, { "epoch": 7.09, "learning_rate": 8.728884652049572e-07, "loss": 3.5353, "step": 37239 }, { "epoch": 7.09, "learning_rate": 8.728312678741658e-07, "loss": 1.0722, "step": 37240 }, { "epoch": 7.09, "learning_rate": 8.727740705433747e-07, "loss": 1.4567, "step": 37241 }, { "epoch": 7.09, "learning_rate": 8.727168732125835e-07, "loss": 2.1172, "step": 37242 }, { "epoch": 7.09, "learning_rate": 8.726596758817922e-07, "loss": 1.0616, "step": 37243 }, { "epoch": 7.09, "learning_rate": 8.72602478551001e-07, "loss": 1.6628, "step": 37244 }, { "epoch": 7.09, "learning_rate": 8.725452812202098e-07, "loss": 1.6319, "step": 37245 }, { "epoch": 7.09, "learning_rate": 8.724880838894186e-07, "loss": 0.9863, "step": 37246 }, { "epoch": 7.09, "learning_rate": 8.724308865586273e-07, "loss": 1.5158, "step": 37247 }, { "epoch": 7.09, "learning_rate": 8.72373689227836e-07, "loss": 0.5417, "step": 37248 }, { "epoch": 7.1, "learning_rate": 8.723164918970448e-07, "loss": 2.5129, "step": 37249 }, { "epoch": 7.1, "learning_rate": 8.722592945662535e-07, "loss": 0.7645, "step": 37250 }, { "epoch": 7.1, "learning_rate": 8.722020972354623e-07, "loss": 0.7925, "step": 37251 }, { "epoch": 7.1, "learning_rate": 8.721448999046712e-07, "loss": 0.9197, "step": 37252 }, { "epoch": 7.1, "learning_rate": 8.7208770257388e-07, "loss": 2.2609, "step": 37253 }, { "epoch": 7.1, "learning_rate": 8.720305052430886e-07, "loss": 1.5013, "step": 37254 }, { "epoch": 7.1, "learning_rate": 8.719733079122975e-07, "loss": 1.5747, "step": 37255 }, { "epoch": 7.1, "learning_rate": 8.719161105815063e-07, "loss": 1.448, "step": 37256 }, { "epoch": 7.1, "learning_rate": 8.71858913250715e-07, "loss": 1.7665, "step": 37257 }, { "epoch": 7.1, "learning_rate": 8.718017159199238e-07, "loss": 1.9674, "step": 37258 }, { "epoch": 7.1, "learning_rate": 8.717445185891326e-07, "loss": 2.1007, "step": 37259 }, { "epoch": 7.1, "learning_rate": 8.716873212583413e-07, "loss": 1.1526, "step": 37260 }, { "epoch": 7.1, "learning_rate": 8.7163012392755e-07, "loss": 1.5723, "step": 37261 }, { "epoch": 7.1, "learning_rate": 8.715729265967588e-07, "loss": 0.9792, "step": 37262 }, { "epoch": 7.1, "learning_rate": 8.715157292659676e-07, "loss": 1.7149, "step": 37263 }, { "epoch": 7.1, "learning_rate": 8.714585319351763e-07, "loss": 1.8643, "step": 37264 }, { "epoch": 7.1, "learning_rate": 8.714013346043851e-07, "loss": 1.6092, "step": 37265 }, { "epoch": 7.1, "learning_rate": 8.71344137273594e-07, "loss": 1.74, "step": 37266 }, { "epoch": 7.1, "learning_rate": 8.712869399428026e-07, "loss": 1.7742, "step": 37267 }, { "epoch": 7.1, "learning_rate": 8.712297426120114e-07, "loss": 1.0412, "step": 37268 }, { "epoch": 7.1, "learning_rate": 8.711725452812203e-07, "loss": 1.6977, "step": 37269 }, { "epoch": 7.1, "learning_rate": 8.711153479504291e-07, "loss": 1.5213, "step": 37270 }, { "epoch": 7.1, "learning_rate": 8.710581506196378e-07, "loss": 0.3255, "step": 37271 }, { "epoch": 7.1, "learning_rate": 8.710009532888466e-07, "loss": 0.9737, "step": 37272 }, { "epoch": 7.1, "learning_rate": 8.709437559580554e-07, "loss": 0.9471, "step": 37273 }, { "epoch": 7.1, "learning_rate": 8.70886558627264e-07, "loss": 0.9808, "step": 37274 }, { "epoch": 7.1, "learning_rate": 8.708293612964728e-07, "loss": 1.7257, "step": 37275 }, { "epoch": 7.1, "learning_rate": 8.707721639656816e-07, "loss": 1.1946, "step": 37276 }, { "epoch": 7.1, "learning_rate": 8.707149666348904e-07, "loss": 2.6779, "step": 37277 }, { "epoch": 7.1, "learning_rate": 8.706577693040991e-07, "loss": 1.0505, "step": 37278 }, { "epoch": 7.1, "learning_rate": 8.706005719733079e-07, "loss": 0.6536, "step": 37279 }, { "epoch": 7.1, "learning_rate": 8.705433746425168e-07, "loss": 2.3074, "step": 37280 }, { "epoch": 7.1, "learning_rate": 8.704861773117254e-07, "loss": 0.7799, "step": 37281 }, { "epoch": 7.1, "learning_rate": 8.704289799809342e-07, "loss": 1.1623, "step": 37282 }, { "epoch": 7.1, "learning_rate": 8.703717826501431e-07, "loss": 2.4632, "step": 37283 }, { "epoch": 7.1, "learning_rate": 8.703145853193517e-07, "loss": 0.5056, "step": 37284 }, { "epoch": 7.1, "learning_rate": 8.702573879885606e-07, "loss": 0.6879, "step": 37285 }, { "epoch": 7.1, "learning_rate": 8.702001906577694e-07, "loss": 1.6617, "step": 37286 }, { "epoch": 7.1, "learning_rate": 8.701429933269782e-07, "loss": 1.7606, "step": 37287 }, { "epoch": 7.1, "learning_rate": 8.700857959961868e-07, "loss": 1.2486, "step": 37288 }, { "epoch": 7.1, "learning_rate": 8.700285986653956e-07, "loss": 2.1781, "step": 37289 }, { "epoch": 7.1, "learning_rate": 8.699714013346044e-07, "loss": 1.273, "step": 37290 }, { "epoch": 7.1, "learning_rate": 8.699142040038131e-07, "loss": 1.9002, "step": 37291 }, { "epoch": 7.1, "learning_rate": 8.698570066730219e-07, "loss": 0.8228, "step": 37292 }, { "epoch": 7.1, "learning_rate": 8.697998093422307e-07, "loss": 2.2829, "step": 37293 }, { "epoch": 7.1, "learning_rate": 8.697426120114396e-07, "loss": 0.8078, "step": 37294 }, { "epoch": 7.1, "learning_rate": 8.696854146806482e-07, "loss": 0.7894, "step": 37295 }, { "epoch": 7.1, "learning_rate": 8.69628217349857e-07, "loss": 1.1871, "step": 37296 }, { "epoch": 7.1, "learning_rate": 8.695710200190659e-07, "loss": 0.906, "step": 37297 }, { "epoch": 7.1, "learning_rate": 8.695138226882745e-07, "loss": 1.4807, "step": 37298 }, { "epoch": 7.1, "learning_rate": 8.694566253574834e-07, "loss": 1.9311, "step": 37299 }, { "epoch": 7.1, "learning_rate": 8.693994280266922e-07, "loss": 1.2006, "step": 37300 }, { "epoch": 7.1, "learning_rate": 8.693422306959008e-07, "loss": 1.4907, "step": 37301 }, { "epoch": 7.11, "learning_rate": 8.692850333651096e-07, "loss": 1.0078, "step": 37302 }, { "epoch": 7.11, "learning_rate": 8.692278360343184e-07, "loss": 2.7812, "step": 37303 }, { "epoch": 7.11, "learning_rate": 8.691706387035272e-07, "loss": 1.4534, "step": 37304 }, { "epoch": 7.11, "learning_rate": 8.691134413727359e-07, "loss": 0.6932, "step": 37305 }, { "epoch": 7.11, "learning_rate": 8.690562440419447e-07, "loss": 0.9928, "step": 37306 }, { "epoch": 7.11, "learning_rate": 8.689990467111535e-07, "loss": 1.8253, "step": 37307 }, { "epoch": 7.11, "learning_rate": 8.689418493803622e-07, "loss": 1.1135, "step": 37308 }, { "epoch": 7.11, "learning_rate": 8.68884652049571e-07, "loss": 2.0758, "step": 37309 }, { "epoch": 7.11, "learning_rate": 8.688274547187799e-07, "loss": 1.027, "step": 37310 }, { "epoch": 7.11, "learning_rate": 8.687702573879887e-07, "loss": 1.7026, "step": 37311 }, { "epoch": 7.11, "learning_rate": 8.687130600571973e-07, "loss": 2.4034, "step": 37312 }, { "epoch": 7.11, "learning_rate": 8.686558627264062e-07, "loss": 0.3521, "step": 37313 }, { "epoch": 7.11, "learning_rate": 8.68598665395615e-07, "loss": 1.9049, "step": 37314 }, { "epoch": 7.11, "learning_rate": 8.685414680648236e-07, "loss": 1.6522, "step": 37315 }, { "epoch": 7.11, "learning_rate": 8.684842707340324e-07, "loss": 2.1147, "step": 37316 }, { "epoch": 7.11, "learning_rate": 8.684270734032412e-07, "loss": 1.3991, "step": 37317 }, { "epoch": 7.11, "learning_rate": 8.6836987607245e-07, "loss": 2.5639, "step": 37318 }, { "epoch": 7.11, "learning_rate": 8.683126787416587e-07, "loss": 2.814, "step": 37319 }, { "epoch": 7.11, "learning_rate": 8.682554814108675e-07, "loss": 1.205, "step": 37320 }, { "epoch": 7.11, "learning_rate": 8.681982840800763e-07, "loss": 0.8258, "step": 37321 }, { "epoch": 7.11, "learning_rate": 8.68141086749285e-07, "loss": 1.47, "step": 37322 }, { "epoch": 7.11, "learning_rate": 8.680838894184938e-07, "loss": 2.2036, "step": 37323 }, { "epoch": 7.11, "learning_rate": 8.680266920877027e-07, "loss": 1.1931, "step": 37324 }, { "epoch": 7.11, "learning_rate": 8.679694947569113e-07, "loss": 0.7617, "step": 37325 }, { "epoch": 7.11, "learning_rate": 8.679122974261201e-07, "loss": 1.5705, "step": 37326 }, { "epoch": 7.11, "learning_rate": 8.67855100095329e-07, "loss": 1.5692, "step": 37327 }, { "epoch": 7.11, "learning_rate": 8.677979027645377e-07, "loss": 3.2636, "step": 37328 }, { "epoch": 7.11, "learning_rate": 8.677407054337464e-07, "loss": 4.0518, "step": 37329 }, { "epoch": 7.11, "learning_rate": 8.676835081029552e-07, "loss": 2.1062, "step": 37330 }, { "epoch": 7.11, "learning_rate": 8.67626310772164e-07, "loss": 0.7368, "step": 37331 }, { "epoch": 7.11, "learning_rate": 8.675691134413727e-07, "loss": 3.3849, "step": 37332 }, { "epoch": 7.11, "learning_rate": 8.675119161105815e-07, "loss": 0.9519, "step": 37333 }, { "epoch": 7.11, "learning_rate": 8.674547187797903e-07, "loss": 1.8426, "step": 37334 }, { "epoch": 7.11, "learning_rate": 8.673975214489991e-07, "loss": 0.6176, "step": 37335 }, { "epoch": 7.11, "learning_rate": 8.673403241182078e-07, "loss": 1.6625, "step": 37336 }, { "epoch": 7.11, "learning_rate": 8.672831267874166e-07, "loss": 1.4771, "step": 37337 }, { "epoch": 7.11, "learning_rate": 8.672259294566255e-07, "loss": 1.8257, "step": 37338 }, { "epoch": 7.11, "learning_rate": 8.671687321258341e-07, "loss": 1.0956, "step": 37339 }, { "epoch": 7.11, "learning_rate": 8.671115347950429e-07, "loss": 1.8821, "step": 37340 }, { "epoch": 7.11, "learning_rate": 8.670543374642518e-07, "loss": 1.4146, "step": 37341 }, { "epoch": 7.11, "learning_rate": 8.669971401334605e-07, "loss": 2.5794, "step": 37342 }, { "epoch": 7.11, "learning_rate": 8.669399428026692e-07, "loss": 2.8992, "step": 37343 }, { "epoch": 7.11, "learning_rate": 8.66882745471878e-07, "loss": 2.2598, "step": 37344 }, { "epoch": 7.11, "learning_rate": 8.668255481410868e-07, "loss": 0.6678, "step": 37345 }, { "epoch": 7.11, "learning_rate": 8.667683508102955e-07, "loss": 1.4422, "step": 37346 }, { "epoch": 7.11, "learning_rate": 8.667111534795043e-07, "loss": 1.0739, "step": 37347 }, { "epoch": 7.11, "learning_rate": 8.666539561487131e-07, "loss": 2.4385, "step": 37348 }, { "epoch": 7.11, "learning_rate": 8.665967588179218e-07, "loss": 1.2824, "step": 37349 }, { "epoch": 7.11, "learning_rate": 8.665395614871306e-07, "loss": 1.8053, "step": 37350 }, { "epoch": 7.11, "learning_rate": 8.664823641563394e-07, "loss": 1.6638, "step": 37351 }, { "epoch": 7.11, "learning_rate": 8.664251668255483e-07, "loss": 1.8334, "step": 37352 }, { "epoch": 7.11, "learning_rate": 8.663679694947569e-07, "loss": 1.0229, "step": 37353 }, { "epoch": 7.12, "learning_rate": 8.663107721639658e-07, "loss": 0.6361, "step": 37354 }, { "epoch": 7.12, "learning_rate": 8.662535748331745e-07, "loss": 2.3212, "step": 37355 }, { "epoch": 7.12, "learning_rate": 8.661963775023831e-07, "loss": 0.7012, "step": 37356 }, { "epoch": 7.12, "learning_rate": 8.66139180171592e-07, "loss": 0.9036, "step": 37357 }, { "epoch": 7.12, "learning_rate": 8.660819828408008e-07, "loss": 2.4449, "step": 37358 }, { "epoch": 7.12, "learning_rate": 8.660247855100096e-07, "loss": 2.0446, "step": 37359 }, { "epoch": 7.12, "learning_rate": 8.659675881792183e-07, "loss": 1.7779, "step": 37360 }, { "epoch": 7.12, "learning_rate": 8.659103908484271e-07, "loss": 0.4992, "step": 37361 }, { "epoch": 7.12, "learning_rate": 8.658531935176359e-07, "loss": 1.547, "step": 37362 }, { "epoch": 7.12, "learning_rate": 8.657959961868446e-07, "loss": 1.0566, "step": 37363 }, { "epoch": 7.12, "learning_rate": 8.657387988560534e-07, "loss": 1.1098, "step": 37364 }, { "epoch": 7.12, "learning_rate": 8.656816015252622e-07, "loss": 1.599, "step": 37365 }, { "epoch": 7.12, "learning_rate": 8.656244041944709e-07, "loss": 2.3152, "step": 37366 }, { "epoch": 7.12, "learning_rate": 8.655672068636796e-07, "loss": 1.3616, "step": 37367 }, { "epoch": 7.12, "learning_rate": 8.655100095328885e-07, "loss": 2.0792, "step": 37368 }, { "epoch": 7.12, "learning_rate": 8.654528122020973e-07, "loss": 1.3322, "step": 37369 }, { "epoch": 7.12, "learning_rate": 8.653956148713059e-07, "loss": 2.4874, "step": 37370 }, { "epoch": 7.12, "learning_rate": 8.653384175405148e-07, "loss": 1.309, "step": 37371 }, { "epoch": 7.12, "learning_rate": 8.652812202097236e-07, "loss": 3.8306, "step": 37372 }, { "epoch": 7.12, "learning_rate": 8.652240228789323e-07, "loss": 3.5651, "step": 37373 }, { "epoch": 7.12, "learning_rate": 8.651668255481411e-07, "loss": 1.7784, "step": 37374 }, { "epoch": 7.12, "learning_rate": 8.651096282173499e-07, "loss": 2.1493, "step": 37375 }, { "epoch": 7.12, "learning_rate": 8.650524308865587e-07, "loss": 1.9147, "step": 37376 }, { "epoch": 7.12, "learning_rate": 8.649952335557674e-07, "loss": 1.0393, "step": 37377 }, { "epoch": 7.12, "learning_rate": 8.649380362249762e-07, "loss": 2.9565, "step": 37378 }, { "epoch": 7.12, "learning_rate": 8.64880838894185e-07, "loss": 1.9774, "step": 37379 }, { "epoch": 7.12, "learning_rate": 8.648236415633937e-07, "loss": 1.3588, "step": 37380 }, { "epoch": 7.12, "learning_rate": 8.647664442326024e-07, "loss": 2.2995, "step": 37381 }, { "epoch": 7.12, "learning_rate": 8.647092469018113e-07, "loss": 2.3914, "step": 37382 }, { "epoch": 7.12, "learning_rate": 8.646520495710201e-07, "loss": 2.2565, "step": 37383 }, { "epoch": 7.12, "learning_rate": 8.645948522402287e-07, "loss": 1.328, "step": 37384 }, { "epoch": 7.12, "learning_rate": 8.645376549094376e-07, "loss": 2.3048, "step": 37385 }, { "epoch": 7.12, "learning_rate": 8.644804575786464e-07, "loss": 1.5955, "step": 37386 }, { "epoch": 7.12, "learning_rate": 8.644232602478551e-07, "loss": 1.1731, "step": 37387 }, { "epoch": 7.12, "learning_rate": 8.643660629170639e-07, "loss": 2.1641, "step": 37388 }, { "epoch": 7.12, "learning_rate": 8.643088655862727e-07, "loss": 1.9724, "step": 37389 }, { "epoch": 7.12, "learning_rate": 8.642516682554814e-07, "loss": 1.6192, "step": 37390 }, { "epoch": 7.12, "learning_rate": 8.641944709246902e-07, "loss": 2.719, "step": 37391 }, { "epoch": 7.12, "learning_rate": 8.64137273593899e-07, "loss": 2.2153, "step": 37392 }, { "epoch": 7.12, "learning_rate": 8.640800762631078e-07, "loss": 1.9932, "step": 37393 }, { "epoch": 7.12, "learning_rate": 8.640228789323164e-07, "loss": 2.2175, "step": 37394 }, { "epoch": 7.12, "learning_rate": 8.639656816015252e-07, "loss": 2.9418, "step": 37395 }, { "epoch": 7.12, "learning_rate": 8.639084842707341e-07, "loss": 2.2631, "step": 37396 }, { "epoch": 7.12, "learning_rate": 8.638512869399427e-07, "loss": 2.1683, "step": 37397 }, { "epoch": 7.12, "learning_rate": 8.637940896091516e-07, "loss": 1.8739, "step": 37398 }, { "epoch": 7.12, "learning_rate": 8.637368922783604e-07, "loss": 2.0283, "step": 37399 }, { "epoch": 7.12, "learning_rate": 8.636796949475692e-07, "loss": 0.9066, "step": 37400 }, { "epoch": 7.12, "learning_rate": 8.636224976167779e-07, "loss": 1.447, "step": 37401 }, { "epoch": 7.12, "learning_rate": 8.635653002859867e-07, "loss": 0.8718, "step": 37402 }, { "epoch": 7.12, "learning_rate": 8.635081029551955e-07, "loss": 1.4673, "step": 37403 }, { "epoch": 7.12, "learning_rate": 8.634509056244042e-07, "loss": 0.9577, "step": 37404 }, { "epoch": 7.12, "learning_rate": 8.63393708293613e-07, "loss": 1.9508, "step": 37405 }, { "epoch": 7.12, "learning_rate": 8.633365109628218e-07, "loss": 2.4448, "step": 37406 }, { "epoch": 7.13, "learning_rate": 8.632793136320307e-07, "loss": 1.4958, "step": 37407 }, { "epoch": 7.13, "learning_rate": 8.632221163012392e-07, "loss": 2.492, "step": 37408 }, { "epoch": 7.13, "learning_rate": 8.63164918970448e-07, "loss": 1.5028, "step": 37409 }, { "epoch": 7.13, "learning_rate": 8.631077216396569e-07, "loss": 1.5947, "step": 37410 }, { "epoch": 7.13, "learning_rate": 8.630505243088655e-07, "loss": 2.5942, "step": 37411 }, { "epoch": 7.13, "learning_rate": 8.629933269780744e-07, "loss": 0.2397, "step": 37412 }, { "epoch": 7.13, "learning_rate": 8.629361296472832e-07, "loss": 3.2627, "step": 37413 }, { "epoch": 7.13, "learning_rate": 8.628789323164918e-07, "loss": 4.1202, "step": 37414 }, { "epoch": 7.13, "learning_rate": 8.628217349857007e-07, "loss": 1.8957, "step": 37415 }, { "epoch": 7.13, "learning_rate": 8.627645376549095e-07, "loss": 0.8588, "step": 37416 }, { "epoch": 7.13, "learning_rate": 8.627073403241183e-07, "loss": 3.3657, "step": 37417 }, { "epoch": 7.13, "learning_rate": 8.62650142993327e-07, "loss": 2.3141, "step": 37418 }, { "epoch": 7.13, "learning_rate": 8.625929456625358e-07, "loss": 0.8725, "step": 37419 }, { "epoch": 7.13, "learning_rate": 8.625357483317446e-07, "loss": 0.2925, "step": 37420 }, { "epoch": 7.13, "learning_rate": 8.624785510009532e-07, "loss": 0.5443, "step": 37421 }, { "epoch": 7.13, "learning_rate": 8.62421353670162e-07, "loss": 1.7238, "step": 37422 }, { "epoch": 7.13, "learning_rate": 8.623641563393708e-07, "loss": 1.7329, "step": 37423 }, { "epoch": 7.13, "learning_rate": 8.623069590085797e-07, "loss": 0.4349, "step": 37424 }, { "epoch": 7.13, "learning_rate": 8.622497616777883e-07, "loss": 1.0549, "step": 37425 }, { "epoch": 7.13, "learning_rate": 8.621925643469972e-07, "loss": 2.0325, "step": 37426 }, { "epoch": 7.13, "learning_rate": 8.62135367016206e-07, "loss": 2.7773, "step": 37427 }, { "epoch": 7.13, "learning_rate": 8.620781696854146e-07, "loss": 0.8929, "step": 37428 }, { "epoch": 7.13, "learning_rate": 8.620209723546235e-07, "loss": 0.4753, "step": 37429 }, { "epoch": 7.13, "learning_rate": 8.619637750238323e-07, "loss": 2.4594, "step": 37430 }, { "epoch": 7.13, "learning_rate": 8.61906577693041e-07, "loss": 1.8029, "step": 37431 }, { "epoch": 7.13, "learning_rate": 8.618493803622498e-07, "loss": 1.5743, "step": 37432 }, { "epoch": 7.13, "learning_rate": 8.617921830314586e-07, "loss": 2.9649, "step": 37433 }, { "epoch": 7.13, "learning_rate": 8.617349857006674e-07, "loss": 2.5367, "step": 37434 }, { "epoch": 7.13, "learning_rate": 8.61677788369876e-07, "loss": 1.1765, "step": 37435 }, { "epoch": 7.13, "learning_rate": 8.616205910390848e-07, "loss": 2.5145, "step": 37436 }, { "epoch": 7.13, "learning_rate": 8.615633937082936e-07, "loss": 1.1604, "step": 37437 }, { "epoch": 7.13, "learning_rate": 8.615061963775023e-07, "loss": 2.5587, "step": 37438 }, { "epoch": 7.13, "learning_rate": 8.614489990467111e-07, "loss": 1.4862, "step": 37439 }, { "epoch": 7.13, "learning_rate": 8.6139180171592e-07, "loss": 3.1554, "step": 37440 }, { "epoch": 7.13, "learning_rate": 8.613346043851288e-07, "loss": 0.9451, "step": 37441 }, { "epoch": 7.13, "learning_rate": 8.612774070543375e-07, "loss": 0.5239, "step": 37442 }, { "epoch": 7.13, "learning_rate": 8.612202097235463e-07, "loss": 1.879, "step": 37443 }, { "epoch": 7.13, "learning_rate": 8.611630123927551e-07, "loss": 1.4602, "step": 37444 }, { "epoch": 7.13, "learning_rate": 8.611058150619638e-07, "loss": 0.4789, "step": 37445 }, { "epoch": 7.13, "learning_rate": 8.610486177311726e-07, "loss": 2.4287, "step": 37446 }, { "epoch": 7.13, "learning_rate": 8.609914204003814e-07, "loss": 0.812, "step": 37447 }, { "epoch": 7.13, "learning_rate": 8.609342230695901e-07, "loss": 2.1535, "step": 37448 }, { "epoch": 7.13, "learning_rate": 8.608770257387988e-07, "loss": 2.3727, "step": 37449 }, { "epoch": 7.13, "learning_rate": 8.608198284080076e-07, "loss": 0.4322, "step": 37450 }, { "epoch": 7.13, "learning_rate": 8.607626310772164e-07, "loss": 2.2995, "step": 37451 }, { "epoch": 7.13, "learning_rate": 8.607054337464251e-07, "loss": 1.3599, "step": 37452 }, { "epoch": 7.13, "learning_rate": 8.606482364156339e-07, "loss": 1.424, "step": 37453 }, { "epoch": 7.13, "learning_rate": 8.605910390848428e-07, "loss": 1.9055, "step": 37454 }, { "epoch": 7.13, "learning_rate": 8.605338417540514e-07, "loss": 1.6798, "step": 37455 }, { "epoch": 7.13, "learning_rate": 8.604766444232603e-07, "loss": 2.1444, "step": 37456 }, { "epoch": 7.13, "learning_rate": 8.604194470924691e-07, "loss": 0.4138, "step": 37457 }, { "epoch": 7.13, "learning_rate": 8.603622497616779e-07, "loss": 2.6286, "step": 37458 }, { "epoch": 7.14, "learning_rate": 8.603050524308866e-07, "loss": 2.8148, "step": 37459 }, { "epoch": 7.14, "learning_rate": 8.602478551000954e-07, "loss": 1.4061, "step": 37460 }, { "epoch": 7.14, "learning_rate": 8.601906577693042e-07, "loss": 1.8648, "step": 37461 }, { "epoch": 7.14, "learning_rate": 8.601334604385128e-07, "loss": 0.6626, "step": 37462 }, { "epoch": 7.14, "learning_rate": 8.600762631077216e-07, "loss": 1.9698, "step": 37463 }, { "epoch": 7.14, "learning_rate": 8.600190657769304e-07, "loss": 1.4573, "step": 37464 }, { "epoch": 7.14, "learning_rate": 8.599618684461393e-07, "loss": 1.3397, "step": 37465 }, { "epoch": 7.14, "learning_rate": 8.599046711153479e-07, "loss": 0.7094, "step": 37466 }, { "epoch": 7.14, "learning_rate": 8.598474737845567e-07, "loss": 1.0775, "step": 37467 }, { "epoch": 7.14, "learning_rate": 8.597902764537656e-07, "loss": 1.635, "step": 37468 }, { "epoch": 7.14, "learning_rate": 8.597330791229742e-07, "loss": 1.8109, "step": 37469 }, { "epoch": 7.14, "learning_rate": 8.596758817921831e-07, "loss": 1.0263, "step": 37470 }, { "epoch": 7.14, "learning_rate": 8.596186844613919e-07, "loss": 2.8622, "step": 37471 }, { "epoch": 7.14, "learning_rate": 8.595614871306007e-07, "loss": 2.2059, "step": 37472 }, { "epoch": 7.14, "learning_rate": 8.595042897998094e-07, "loss": 1.4655, "step": 37473 }, { "epoch": 7.14, "learning_rate": 8.594470924690182e-07, "loss": 2.496, "step": 37474 }, { "epoch": 7.14, "learning_rate": 8.593898951382269e-07, "loss": 1.5779, "step": 37475 }, { "epoch": 7.14, "learning_rate": 8.593326978074356e-07, "loss": 1.5015, "step": 37476 }, { "epoch": 7.14, "learning_rate": 8.592755004766444e-07, "loss": 1.2263, "step": 37477 }, { "epoch": 7.14, "learning_rate": 8.592183031458532e-07, "loss": 1.0441, "step": 37478 }, { "epoch": 7.14, "learning_rate": 8.591611058150619e-07, "loss": 1.7945, "step": 37479 }, { "epoch": 7.14, "learning_rate": 8.591039084842707e-07, "loss": 2.7651, "step": 37480 }, { "epoch": 7.14, "learning_rate": 8.590467111534795e-07, "loss": 1.9859, "step": 37481 }, { "epoch": 7.14, "learning_rate": 8.589895138226884e-07, "loss": 1.5468, "step": 37482 }, { "epoch": 7.14, "learning_rate": 8.58932316491897e-07, "loss": 1.7596, "step": 37483 }, { "epoch": 7.14, "learning_rate": 8.588751191611059e-07, "loss": 2.3423, "step": 37484 }, { "epoch": 7.14, "learning_rate": 8.588179218303147e-07, "loss": 2.2884, "step": 37485 }, { "epoch": 7.14, "learning_rate": 8.587607244995234e-07, "loss": 1.8825, "step": 37486 }, { "epoch": 7.14, "learning_rate": 8.587035271687322e-07, "loss": 2.4364, "step": 37487 }, { "epoch": 7.14, "learning_rate": 8.58646329837941e-07, "loss": 0.8503, "step": 37488 }, { "epoch": 7.14, "learning_rate": 8.585891325071497e-07, "loss": 2.1147, "step": 37489 }, { "epoch": 7.14, "learning_rate": 8.585319351763584e-07, "loss": 2.9332, "step": 37490 }, { "epoch": 7.14, "learning_rate": 8.584747378455672e-07, "loss": 1.5619, "step": 37491 }, { "epoch": 7.14, "learning_rate": 8.58417540514776e-07, "loss": 1.6381, "step": 37492 }, { "epoch": 7.14, "learning_rate": 8.583603431839847e-07, "loss": 2.222, "step": 37493 }, { "epoch": 7.14, "learning_rate": 8.583031458531935e-07, "loss": 1.5883, "step": 37494 }, { "epoch": 7.14, "learning_rate": 8.582459485224023e-07, "loss": 1.9038, "step": 37495 }, { "epoch": 7.14, "learning_rate": 8.58188751191611e-07, "loss": 1.9963, "step": 37496 }, { "epoch": 7.14, "learning_rate": 8.581315538608198e-07, "loss": 2.1882, "step": 37497 }, { "epoch": 7.14, "learning_rate": 8.580743565300287e-07, "loss": 1.8388, "step": 37498 }, { "epoch": 7.14, "learning_rate": 8.580171591992375e-07, "loss": 0.4893, "step": 37499 }, { "epoch": 7.14, "learning_rate": 8.579599618684462e-07, "loss": 1.6548, "step": 37500 }, { "epoch": 7.14, "learning_rate": 8.57902764537655e-07, "loss": 1.8365, "step": 37501 }, { "epoch": 7.14, "learning_rate": 8.578455672068637e-07, "loss": 1.8725, "step": 37502 }, { "epoch": 7.14, "learning_rate": 8.577883698760724e-07, "loss": 1.6617, "step": 37503 }, { "epoch": 7.14, "learning_rate": 8.577311725452812e-07, "loss": 1.1231, "step": 37504 }, { "epoch": 7.14, "learning_rate": 8.5767397521449e-07, "loss": 2.0445, "step": 37505 }, { "epoch": 7.14, "learning_rate": 8.576167778836988e-07, "loss": 2.4306, "step": 37506 }, { "epoch": 7.14, "learning_rate": 8.575595805529075e-07, "loss": 2.6477, "step": 37507 }, { "epoch": 7.14, "learning_rate": 8.575023832221163e-07, "loss": 2.4811, "step": 37508 }, { "epoch": 7.14, "learning_rate": 8.574451858913252e-07, "loss": 2.1239, "step": 37509 }, { "epoch": 7.14, "learning_rate": 8.573879885605338e-07, "loss": 1.8451, "step": 37510 }, { "epoch": 7.14, "learning_rate": 8.573307912297426e-07, "loss": 1.0403, "step": 37511 }, { "epoch": 7.15, "learning_rate": 8.572735938989515e-07, "loss": 2.8683, "step": 37512 }, { "epoch": 7.15, "learning_rate": 8.572163965681603e-07, "loss": 0.8396, "step": 37513 }, { "epoch": 7.15, "learning_rate": 8.57159199237369e-07, "loss": 0.3612, "step": 37514 }, { "epoch": 7.15, "learning_rate": 8.571020019065777e-07, "loss": 1.9027, "step": 37515 }, { "epoch": 7.15, "learning_rate": 8.570448045757865e-07, "loss": 1.435, "step": 37516 }, { "epoch": 7.15, "learning_rate": 8.569876072449952e-07, "loss": 0.3643, "step": 37517 }, { "epoch": 7.15, "learning_rate": 8.56930409914204e-07, "loss": 0.527, "step": 37518 }, { "epoch": 7.15, "learning_rate": 8.568732125834128e-07, "loss": 1.3797, "step": 37519 }, { "epoch": 7.15, "learning_rate": 8.568160152526215e-07, "loss": 1.7532, "step": 37520 }, { "epoch": 7.15, "learning_rate": 8.567588179218303e-07, "loss": 1.1818, "step": 37521 }, { "epoch": 7.15, "learning_rate": 8.567016205910391e-07, "loss": 2.4852, "step": 37522 }, { "epoch": 7.15, "learning_rate": 8.56644423260248e-07, "loss": 1.4095, "step": 37523 }, { "epoch": 7.15, "learning_rate": 8.565872259294566e-07, "loss": 1.8742, "step": 37524 }, { "epoch": 7.15, "learning_rate": 8.565300285986654e-07, "loss": 0.9448, "step": 37525 }, { "epoch": 7.15, "learning_rate": 8.564728312678743e-07, "loss": 1.2516, "step": 37526 }, { "epoch": 7.15, "learning_rate": 8.564156339370828e-07, "loss": 1.8789, "step": 37527 }, { "epoch": 7.15, "learning_rate": 8.563584366062917e-07, "loss": 2.3891, "step": 37528 }, { "epoch": 7.15, "learning_rate": 8.563012392755005e-07, "loss": 1.7287, "step": 37529 }, { "epoch": 7.15, "learning_rate": 8.562440419447093e-07, "loss": 1.8629, "step": 37530 }, { "epoch": 7.15, "learning_rate": 8.56186844613918e-07, "loss": 1.5354, "step": 37531 }, { "epoch": 7.15, "learning_rate": 8.561296472831268e-07, "loss": 0.9721, "step": 37532 }, { "epoch": 7.15, "learning_rate": 8.560724499523356e-07, "loss": 2.5226, "step": 37533 }, { "epoch": 7.15, "learning_rate": 8.560152526215443e-07, "loss": 1.298, "step": 37534 }, { "epoch": 7.15, "learning_rate": 8.559580552907531e-07, "loss": 2.1895, "step": 37535 }, { "epoch": 7.15, "learning_rate": 8.559008579599619e-07, "loss": 2.6117, "step": 37536 }, { "epoch": 7.15, "learning_rate": 8.558436606291708e-07, "loss": 1.9784, "step": 37537 }, { "epoch": 7.15, "learning_rate": 8.557864632983794e-07, "loss": 1.6006, "step": 37538 }, { "epoch": 7.15, "learning_rate": 8.557292659675882e-07, "loss": 2.6661, "step": 37539 }, { "epoch": 7.15, "learning_rate": 8.556720686367971e-07, "loss": 1.5237, "step": 37540 }, { "epoch": 7.15, "learning_rate": 8.556148713060056e-07, "loss": 2.1878, "step": 37541 }, { "epoch": 7.15, "learning_rate": 8.555576739752145e-07, "loss": 1.7464, "step": 37542 }, { "epoch": 7.15, "learning_rate": 8.555004766444233e-07, "loss": 1.5536, "step": 37543 }, { "epoch": 7.15, "learning_rate": 8.55443279313632e-07, "loss": 0.6579, "step": 37544 }, { "epoch": 7.15, "learning_rate": 8.553860819828408e-07, "loss": 1.3377, "step": 37545 }, { "epoch": 7.15, "learning_rate": 8.553288846520496e-07, "loss": 0.8607, "step": 37546 }, { "epoch": 7.15, "learning_rate": 8.552716873212584e-07, "loss": 1.3585, "step": 37547 }, { "epoch": 7.15, "learning_rate": 8.552144899904671e-07, "loss": 2.4612, "step": 37548 }, { "epoch": 7.15, "learning_rate": 8.551572926596759e-07, "loss": 2.2643, "step": 37549 }, { "epoch": 7.15, "learning_rate": 8.551000953288847e-07, "loss": 1.6145, "step": 37550 }, { "epoch": 7.15, "learning_rate": 8.550428979980934e-07, "loss": 1.6951, "step": 37551 }, { "epoch": 7.15, "learning_rate": 8.549857006673022e-07, "loss": 0.4228, "step": 37552 }, { "epoch": 7.15, "learning_rate": 8.549285033365111e-07, "loss": 1.6113, "step": 37553 }, { "epoch": 7.15, "learning_rate": 8.548713060057199e-07, "loss": 1.2065, "step": 37554 }, { "epoch": 7.15, "learning_rate": 8.548141086749284e-07, "loss": 0.8683, "step": 37555 }, { "epoch": 7.15, "learning_rate": 8.547569113441373e-07, "loss": 1.3576, "step": 37556 }, { "epoch": 7.15, "learning_rate": 8.546997140133461e-07, "loss": 3.2598, "step": 37557 }, { "epoch": 7.15, "learning_rate": 8.546425166825548e-07, "loss": 1.4186, "step": 37558 }, { "epoch": 7.15, "learning_rate": 8.545853193517636e-07, "loss": 0.962, "step": 37559 }, { "epoch": 7.15, "learning_rate": 8.545281220209724e-07, "loss": 1.6381, "step": 37560 }, { "epoch": 7.15, "learning_rate": 8.544709246901811e-07, "loss": 1.3862, "step": 37561 }, { "epoch": 7.15, "learning_rate": 8.544137273593899e-07, "loss": 0.9225, "step": 37562 }, { "epoch": 7.15, "learning_rate": 8.543565300285987e-07, "loss": 1.0252, "step": 37563 }, { "epoch": 7.16, "learning_rate": 8.542993326978075e-07, "loss": 2.1389, "step": 37564 }, { "epoch": 7.16, "learning_rate": 8.542421353670162e-07, "loss": 2.146, "step": 37565 }, { "epoch": 7.16, "learning_rate": 8.54184938036225e-07, "loss": 2.7825, "step": 37566 }, { "epoch": 7.16, "learning_rate": 8.541277407054339e-07, "loss": 1.1856, "step": 37567 }, { "epoch": 7.16, "learning_rate": 8.540705433746424e-07, "loss": 0.8971, "step": 37568 }, { "epoch": 7.16, "learning_rate": 8.540133460438512e-07, "loss": 0.7514, "step": 37569 }, { "epoch": 7.16, "learning_rate": 8.539561487130601e-07, "loss": 1.0903, "step": 37570 }, { "epoch": 7.16, "learning_rate": 8.538989513822689e-07, "loss": 1.5508, "step": 37571 }, { "epoch": 7.16, "learning_rate": 8.538417540514776e-07, "loss": 1.6207, "step": 37572 }, { "epoch": 7.16, "learning_rate": 8.537845567206864e-07, "loss": 2.3003, "step": 37573 }, { "epoch": 7.16, "learning_rate": 8.537273593898952e-07, "loss": 2.7545, "step": 37574 }, { "epoch": 7.16, "learning_rate": 8.536701620591039e-07, "loss": 1.8418, "step": 37575 }, { "epoch": 7.16, "learning_rate": 8.536129647283127e-07, "loss": 1.4341, "step": 37576 }, { "epoch": 7.16, "learning_rate": 8.535557673975215e-07, "loss": 1.3541, "step": 37577 }, { "epoch": 7.16, "learning_rate": 8.534985700667303e-07, "loss": 1.5596, "step": 37578 }, { "epoch": 7.16, "learning_rate": 8.53441372735939e-07, "loss": 2.219, "step": 37579 }, { "epoch": 7.16, "learning_rate": 8.533841754051478e-07, "loss": 2.009, "step": 37580 }, { "epoch": 7.16, "learning_rate": 8.533269780743567e-07, "loss": 0.6538, "step": 37581 }, { "epoch": 7.16, "learning_rate": 8.532697807435652e-07, "loss": 2.2876, "step": 37582 }, { "epoch": 7.16, "learning_rate": 8.53212583412774e-07, "loss": 2.4665, "step": 37583 }, { "epoch": 7.16, "learning_rate": 8.531553860819829e-07, "loss": 1.7845, "step": 37584 }, { "epoch": 7.16, "learning_rate": 8.530981887511915e-07, "loss": 2.829, "step": 37585 }, { "epoch": 7.16, "learning_rate": 8.530409914204004e-07, "loss": 2.2278, "step": 37586 }, { "epoch": 7.16, "learning_rate": 8.529837940896092e-07, "loss": 0.6627, "step": 37587 }, { "epoch": 7.16, "learning_rate": 8.52926596758818e-07, "loss": 0.4649, "step": 37588 }, { "epoch": 7.16, "learning_rate": 8.528693994280267e-07, "loss": 1.054, "step": 37589 }, { "epoch": 7.16, "learning_rate": 8.528122020972355e-07, "loss": 2.175, "step": 37590 }, { "epoch": 7.16, "learning_rate": 8.527550047664443e-07, "loss": 1.5636, "step": 37591 }, { "epoch": 7.16, "learning_rate": 8.52697807435653e-07, "loss": 1.6955, "step": 37592 }, { "epoch": 7.16, "learning_rate": 8.526406101048618e-07, "loss": 1.0562, "step": 37593 }, { "epoch": 7.16, "learning_rate": 8.525834127740706e-07, "loss": 0.5738, "step": 37594 }, { "epoch": 7.16, "learning_rate": 8.525262154432795e-07, "loss": 1.7474, "step": 37595 }, { "epoch": 7.16, "learning_rate": 8.52469018112488e-07, "loss": 1.1963, "step": 37596 }, { "epoch": 7.16, "learning_rate": 8.524118207816969e-07, "loss": 1.1779, "step": 37597 }, { "epoch": 7.16, "learning_rate": 8.523546234509057e-07, "loss": 2.965, "step": 37598 }, { "epoch": 7.16, "learning_rate": 8.522974261201143e-07, "loss": 2.5143, "step": 37599 }, { "epoch": 7.16, "learning_rate": 8.522402287893232e-07, "loss": 2.1605, "step": 37600 }, { "epoch": 7.16, "learning_rate": 8.52183031458532e-07, "loss": 2.1149, "step": 37601 }, { "epoch": 7.16, "learning_rate": 8.521258341277407e-07, "loss": 1.6672, "step": 37602 }, { "epoch": 7.16, "learning_rate": 8.520686367969495e-07, "loss": 2.3513, "step": 37603 }, { "epoch": 7.16, "learning_rate": 8.520114394661583e-07, "loss": 1.8613, "step": 37604 }, { "epoch": 7.16, "learning_rate": 8.519542421353671e-07, "loss": 1.5229, "step": 37605 }, { "epoch": 7.16, "learning_rate": 8.518970448045758e-07, "loss": 1.9671, "step": 37606 }, { "epoch": 7.16, "learning_rate": 8.518398474737846e-07, "loss": 2.3977, "step": 37607 }, { "epoch": 7.16, "learning_rate": 8.517826501429933e-07, "loss": 1.6915, "step": 37608 }, { "epoch": 7.16, "learning_rate": 8.51725452812202e-07, "loss": 2.6899, "step": 37609 }, { "epoch": 7.16, "learning_rate": 8.516682554814108e-07, "loss": 1.3208, "step": 37610 }, { "epoch": 7.16, "learning_rate": 8.516110581506197e-07, "loss": 1.5612, "step": 37611 }, { "epoch": 7.16, "learning_rate": 8.515538608198285e-07, "loss": 2.3016, "step": 37612 }, { "epoch": 7.16, "learning_rate": 8.514966634890371e-07, "loss": 1.0652, "step": 37613 }, { "epoch": 7.16, "learning_rate": 8.51439466158246e-07, "loss": 1.8699, "step": 37614 }, { "epoch": 7.16, "learning_rate": 8.513822688274548e-07, "loss": 1.6548, "step": 37615 }, { "epoch": 7.16, "learning_rate": 8.513250714966635e-07, "loss": 1.3983, "step": 37616 }, { "epoch": 7.17, "learning_rate": 8.512678741658723e-07, "loss": 1.3165, "step": 37617 }, { "epoch": 7.17, "learning_rate": 8.512106768350811e-07, "loss": 2.0703, "step": 37618 }, { "epoch": 7.17, "learning_rate": 8.511534795042899e-07, "loss": 1.8087, "step": 37619 }, { "epoch": 7.17, "learning_rate": 8.510962821734986e-07, "loss": 2.572, "step": 37620 }, { "epoch": 7.17, "learning_rate": 8.510390848427074e-07, "loss": 2.0138, "step": 37621 }, { "epoch": 7.17, "learning_rate": 8.509818875119161e-07, "loss": 1.2665, "step": 37622 }, { "epoch": 7.17, "learning_rate": 8.509246901811248e-07, "loss": 1.0143, "step": 37623 }, { "epoch": 7.17, "learning_rate": 8.508674928503336e-07, "loss": 0.8523, "step": 37624 }, { "epoch": 7.17, "learning_rate": 8.508102955195425e-07, "loss": 1.4028, "step": 37625 }, { "epoch": 7.17, "learning_rate": 8.507530981887511e-07, "loss": 1.5994, "step": 37626 }, { "epoch": 7.17, "learning_rate": 8.506959008579599e-07, "loss": 0.835, "step": 37627 }, { "epoch": 7.17, "learning_rate": 8.506387035271688e-07, "loss": 1.5774, "step": 37628 }, { "epoch": 7.17, "learning_rate": 8.505815061963776e-07, "loss": 0.2814, "step": 37629 }, { "epoch": 7.17, "learning_rate": 8.505243088655863e-07, "loss": 1.0718, "step": 37630 }, { "epoch": 7.17, "learning_rate": 8.504671115347951e-07, "loss": 1.248, "step": 37631 }, { "epoch": 7.17, "learning_rate": 8.504099142040039e-07, "loss": 1.5894, "step": 37632 }, { "epoch": 7.17, "learning_rate": 8.503527168732126e-07, "loss": 1.6053, "step": 37633 }, { "epoch": 7.17, "learning_rate": 8.502955195424214e-07, "loss": 1.6025, "step": 37634 }, { "epoch": 7.17, "learning_rate": 8.502383222116301e-07, "loss": 2.6061, "step": 37635 }, { "epoch": 7.17, "learning_rate": 8.501811248808389e-07, "loss": 0.4908, "step": 37636 }, { "epoch": 7.17, "learning_rate": 8.501239275500476e-07, "loss": 1.5369, "step": 37637 }, { "epoch": 7.17, "learning_rate": 8.500667302192564e-07, "loss": 2.0264, "step": 37638 }, { "epoch": 7.17, "learning_rate": 8.500095328884653e-07, "loss": 2.3836, "step": 37639 }, { "epoch": 7.17, "learning_rate": 8.499523355576739e-07, "loss": 1.5976, "step": 37640 }, { "epoch": 7.17, "learning_rate": 8.498951382268828e-07, "loss": 0.9214, "step": 37641 }, { "epoch": 7.17, "learning_rate": 8.498379408960916e-07, "loss": 1.0952, "step": 37642 }, { "epoch": 7.17, "learning_rate": 8.497807435653004e-07, "loss": 0.686, "step": 37643 }, { "epoch": 7.17, "learning_rate": 8.497235462345091e-07, "loss": 1.8928, "step": 37644 }, { "epoch": 7.17, "learning_rate": 8.496663489037179e-07, "loss": 2.5669, "step": 37645 }, { "epoch": 7.17, "learning_rate": 8.496091515729267e-07, "loss": 2.8469, "step": 37646 }, { "epoch": 7.17, "learning_rate": 8.495519542421354e-07, "loss": 1.7853, "step": 37647 }, { "epoch": 7.17, "learning_rate": 8.494947569113441e-07, "loss": 1.8275, "step": 37648 }, { "epoch": 7.17, "learning_rate": 8.494375595805529e-07, "loss": 1.6221, "step": 37649 }, { "epoch": 7.17, "learning_rate": 8.493803622497616e-07, "loss": 3.5796, "step": 37650 }, { "epoch": 7.17, "learning_rate": 8.493231649189704e-07, "loss": 1.9804, "step": 37651 }, { "epoch": 7.17, "learning_rate": 8.492659675881792e-07, "loss": 0.4405, "step": 37652 }, { "epoch": 7.17, "learning_rate": 8.492087702573881e-07, "loss": 1.5885, "step": 37653 }, { "epoch": 7.17, "learning_rate": 8.491515729265967e-07, "loss": 1.6916, "step": 37654 }, { "epoch": 7.17, "learning_rate": 8.490943755958056e-07, "loss": 1.3241, "step": 37655 }, { "epoch": 7.17, "learning_rate": 8.490371782650144e-07, "loss": 1.8394, "step": 37656 }, { "epoch": 7.17, "learning_rate": 8.48979980934223e-07, "loss": 0.8454, "step": 37657 }, { "epoch": 7.17, "learning_rate": 8.489227836034319e-07, "loss": 3.2601, "step": 37658 }, { "epoch": 7.17, "learning_rate": 8.488655862726407e-07, "loss": 1.2655, "step": 37659 }, { "epoch": 7.17, "learning_rate": 8.488083889418495e-07, "loss": 0.7727, "step": 37660 }, { "epoch": 7.17, "learning_rate": 8.487511916110582e-07, "loss": 2.0357, "step": 37661 }, { "epoch": 7.17, "learning_rate": 8.486939942802669e-07, "loss": 2.0595, "step": 37662 }, { "epoch": 7.17, "learning_rate": 8.486367969494757e-07, "loss": 2.0546, "step": 37663 }, { "epoch": 7.17, "learning_rate": 8.485795996186844e-07, "loss": 2.2303, "step": 37664 }, { "epoch": 7.17, "learning_rate": 8.485224022878932e-07, "loss": 1.2681, "step": 37665 }, { "epoch": 7.17, "learning_rate": 8.48465204957102e-07, "loss": 1.983, "step": 37666 }, { "epoch": 7.17, "learning_rate": 8.484080076263107e-07, "loss": 1.8895, "step": 37667 }, { "epoch": 7.17, "learning_rate": 8.483508102955195e-07, "loss": 1.2341, "step": 37668 }, { "epoch": 7.18, "learning_rate": 8.482936129647284e-07, "loss": 1.9696, "step": 37669 }, { "epoch": 7.18, "learning_rate": 8.482364156339372e-07, "loss": 2.0942, "step": 37670 }, { "epoch": 7.18, "learning_rate": 8.481792183031458e-07, "loss": 1.2895, "step": 37671 }, { "epoch": 7.18, "learning_rate": 8.481220209723547e-07, "loss": 0.8309, "step": 37672 }, { "epoch": 7.18, "learning_rate": 8.480648236415635e-07, "loss": 2.0334, "step": 37673 }, { "epoch": 7.18, "learning_rate": 8.480076263107722e-07, "loss": 0.7073, "step": 37674 }, { "epoch": 7.18, "learning_rate": 8.479504289799809e-07, "loss": 0.9782, "step": 37675 }, { "epoch": 7.18, "learning_rate": 8.478932316491897e-07, "loss": 1.2476, "step": 37676 }, { "epoch": 7.18, "learning_rate": 8.478360343183985e-07, "loss": 2.6749, "step": 37677 }, { "epoch": 7.18, "learning_rate": 8.477788369876072e-07, "loss": 2.1057, "step": 37678 }, { "epoch": 7.18, "learning_rate": 8.47721639656816e-07, "loss": 1.084, "step": 37679 }, { "epoch": 7.18, "learning_rate": 8.476644423260248e-07, "loss": 1.1093, "step": 37680 }, { "epoch": 7.18, "learning_rate": 8.476072449952335e-07, "loss": 1.7414, "step": 37681 }, { "epoch": 7.18, "learning_rate": 8.475500476644423e-07, "loss": 1.9878, "step": 37682 }, { "epoch": 7.18, "learning_rate": 8.474928503336512e-07, "loss": 1.2836, "step": 37683 }, { "epoch": 7.18, "learning_rate": 8.4743565300286e-07, "loss": 1.9076, "step": 37684 }, { "epoch": 7.18, "learning_rate": 8.473784556720686e-07, "loss": 0.7839, "step": 37685 }, { "epoch": 7.18, "learning_rate": 8.473212583412775e-07, "loss": 1.5817, "step": 37686 }, { "epoch": 7.18, "learning_rate": 8.472640610104863e-07, "loss": 1.4191, "step": 37687 }, { "epoch": 7.18, "learning_rate": 8.47206863679695e-07, "loss": 1.4612, "step": 37688 }, { "epoch": 7.18, "learning_rate": 8.471496663489037e-07, "loss": 1.4858, "step": 37689 }, { "epoch": 7.18, "learning_rate": 8.470924690181125e-07, "loss": 2.2306, "step": 37690 }, { "epoch": 7.18, "learning_rate": 8.470352716873212e-07, "loss": 2.0731, "step": 37691 }, { "epoch": 7.18, "learning_rate": 8.4697807435653e-07, "loss": 2.1147, "step": 37692 }, { "epoch": 7.18, "learning_rate": 8.469208770257388e-07, "loss": 1.3506, "step": 37693 }, { "epoch": 7.18, "learning_rate": 8.468636796949476e-07, "loss": 0.9073, "step": 37694 }, { "epoch": 7.18, "learning_rate": 8.468064823641563e-07, "loss": 0.7099, "step": 37695 }, { "epoch": 7.18, "learning_rate": 8.467492850333651e-07, "loss": 1.7821, "step": 37696 }, { "epoch": 7.18, "learning_rate": 8.46692087702574e-07, "loss": 2.7499, "step": 37697 }, { "epoch": 7.18, "learning_rate": 8.466348903717826e-07, "loss": 1.3678, "step": 37698 }, { "epoch": 7.18, "learning_rate": 8.465776930409915e-07, "loss": 1.7612, "step": 37699 }, { "epoch": 7.18, "learning_rate": 8.465204957102003e-07, "loss": 1.3492, "step": 37700 }, { "epoch": 7.18, "learning_rate": 8.464632983794091e-07, "loss": 2.6828, "step": 37701 }, { "epoch": 7.18, "learning_rate": 8.464061010486177e-07, "loss": 1.2355, "step": 37702 }, { "epoch": 7.18, "learning_rate": 8.463489037178265e-07, "loss": 1.4229, "step": 37703 }, { "epoch": 7.18, "learning_rate": 8.462917063870353e-07, "loss": 2.8679, "step": 37704 }, { "epoch": 7.18, "learning_rate": 8.46234509056244e-07, "loss": 4.0769, "step": 37705 }, { "epoch": 7.18, "learning_rate": 8.461773117254528e-07, "loss": 1.2011, "step": 37706 }, { "epoch": 7.18, "learning_rate": 8.461201143946616e-07, "loss": 1.4406, "step": 37707 }, { "epoch": 7.18, "learning_rate": 8.460629170638705e-07, "loss": 1.2712, "step": 37708 }, { "epoch": 7.18, "learning_rate": 8.460057197330791e-07, "loss": 1.0509, "step": 37709 }, { "epoch": 7.18, "learning_rate": 8.459485224022879e-07, "loss": 1.6731, "step": 37710 }, { "epoch": 7.18, "learning_rate": 8.458913250714968e-07, "loss": 1.0409, "step": 37711 }, { "epoch": 7.18, "learning_rate": 8.458341277407054e-07, "loss": 0.6065, "step": 37712 }, { "epoch": 7.18, "learning_rate": 8.457769304099143e-07, "loss": 2.2388, "step": 37713 }, { "epoch": 7.18, "learning_rate": 8.457197330791231e-07, "loss": 1.9067, "step": 37714 }, { "epoch": 7.18, "learning_rate": 8.456625357483316e-07, "loss": 1.2496, "step": 37715 }, { "epoch": 7.18, "learning_rate": 8.456053384175405e-07, "loss": 1.5326, "step": 37716 }, { "epoch": 7.18, "learning_rate": 8.455481410867493e-07, "loss": 0.8706, "step": 37717 }, { "epoch": 7.18, "learning_rate": 8.454909437559581e-07, "loss": 2.7131, "step": 37718 }, { "epoch": 7.18, "learning_rate": 8.454337464251668e-07, "loss": 1.6477, "step": 37719 }, { "epoch": 7.18, "learning_rate": 8.453765490943756e-07, "loss": 1.1746, "step": 37720 }, { "epoch": 7.18, "learning_rate": 8.453193517635844e-07, "loss": 1.7793, "step": 37721 }, { "epoch": 7.19, "learning_rate": 8.452621544327931e-07, "loss": 2.1696, "step": 37722 }, { "epoch": 7.19, "learning_rate": 8.452049571020019e-07, "loss": 1.3806, "step": 37723 }, { "epoch": 7.19, "learning_rate": 8.451477597712107e-07, "loss": 1.8971, "step": 37724 }, { "epoch": 7.19, "learning_rate": 8.450905624404196e-07, "loss": 2.2415, "step": 37725 }, { "epoch": 7.19, "learning_rate": 8.450333651096282e-07, "loss": 1.3078, "step": 37726 }, { "epoch": 7.19, "learning_rate": 8.449761677788371e-07, "loss": 2.9659, "step": 37727 }, { "epoch": 7.19, "learning_rate": 8.449189704480459e-07, "loss": 2.1822, "step": 37728 }, { "epoch": 7.19, "learning_rate": 8.448617731172544e-07, "loss": 1.4517, "step": 37729 }, { "epoch": 7.19, "learning_rate": 8.448045757864633e-07, "loss": 1.7432, "step": 37730 }, { "epoch": 7.19, "learning_rate": 8.447473784556721e-07, "loss": 0.9258, "step": 37731 }, { "epoch": 7.19, "learning_rate": 8.446901811248808e-07, "loss": 3.2808, "step": 37732 }, { "epoch": 7.19, "learning_rate": 8.446329837940896e-07, "loss": 1.6911, "step": 37733 }, { "epoch": 7.19, "learning_rate": 8.445757864632984e-07, "loss": 1.9025, "step": 37734 }, { "epoch": 7.19, "learning_rate": 8.445185891325072e-07, "loss": 1.8475, "step": 37735 }, { "epoch": 7.19, "learning_rate": 8.444613918017159e-07, "loss": 2.7804, "step": 37736 }, { "epoch": 7.19, "learning_rate": 8.444041944709247e-07, "loss": 1.7126, "step": 37737 }, { "epoch": 7.19, "learning_rate": 8.443469971401335e-07, "loss": 1.9114, "step": 37738 }, { "epoch": 7.19, "learning_rate": 8.442897998093422e-07, "loss": 1.1737, "step": 37739 }, { "epoch": 7.19, "learning_rate": 8.44232602478551e-07, "loss": 1.2519, "step": 37740 }, { "epoch": 7.19, "learning_rate": 8.441754051477599e-07, "loss": 1.6376, "step": 37741 }, { "epoch": 7.19, "learning_rate": 8.441182078169686e-07, "loss": 1.3608, "step": 37742 }, { "epoch": 7.19, "learning_rate": 8.440610104861773e-07, "loss": 1.4655, "step": 37743 }, { "epoch": 7.19, "learning_rate": 8.440038131553861e-07, "loss": 0.9798, "step": 37744 }, { "epoch": 7.19, "learning_rate": 8.439466158245949e-07, "loss": 2.7672, "step": 37745 }, { "epoch": 7.19, "learning_rate": 8.438894184938036e-07, "loss": 1.6055, "step": 37746 }, { "epoch": 7.19, "learning_rate": 8.438322211630124e-07, "loss": 2.1547, "step": 37747 }, { "epoch": 7.19, "learning_rate": 8.437750238322212e-07, "loss": 2.5595, "step": 37748 }, { "epoch": 7.19, "learning_rate": 8.4371782650143e-07, "loss": 1.5321, "step": 37749 }, { "epoch": 7.19, "learning_rate": 8.436606291706387e-07, "loss": 0.673, "step": 37750 }, { "epoch": 7.19, "learning_rate": 8.436034318398475e-07, "loss": 2.2058, "step": 37751 }, { "epoch": 7.19, "learning_rate": 8.435462345090564e-07, "loss": 2.8445, "step": 37752 }, { "epoch": 7.19, "learning_rate": 8.43489037178265e-07, "loss": 1.3596, "step": 37753 }, { "epoch": 7.19, "learning_rate": 8.434318398474738e-07, "loss": 1.2289, "step": 37754 }, { "epoch": 7.19, "learning_rate": 8.433746425166827e-07, "loss": 2.3246, "step": 37755 }, { "epoch": 7.19, "learning_rate": 8.433174451858912e-07, "loss": 1.6799, "step": 37756 }, { "epoch": 7.19, "learning_rate": 8.432602478551001e-07, "loss": 2.6639, "step": 37757 }, { "epoch": 7.19, "learning_rate": 8.432030505243089e-07, "loss": 1.8867, "step": 37758 }, { "epoch": 7.19, "learning_rate": 8.431458531935177e-07, "loss": 0.9726, "step": 37759 }, { "epoch": 7.19, "learning_rate": 8.430886558627264e-07, "loss": 1.3927, "step": 37760 }, { "epoch": 7.19, "learning_rate": 8.430314585319352e-07, "loss": 1.0678, "step": 37761 }, { "epoch": 7.19, "learning_rate": 8.42974261201144e-07, "loss": 2.7575, "step": 37762 }, { "epoch": 7.19, "learning_rate": 8.429170638703527e-07, "loss": 1.9742, "step": 37763 }, { "epoch": 7.19, "learning_rate": 8.428598665395615e-07, "loss": 1.0389, "step": 37764 }, { "epoch": 7.19, "learning_rate": 8.428026692087703e-07, "loss": 1.2916, "step": 37765 }, { "epoch": 7.19, "learning_rate": 8.427454718779792e-07, "loss": 1.3953, "step": 37766 }, { "epoch": 7.19, "learning_rate": 8.426882745471878e-07, "loss": 2.104, "step": 37767 }, { "epoch": 7.19, "learning_rate": 8.426310772163965e-07, "loss": 1.4641, "step": 37768 }, { "epoch": 7.19, "learning_rate": 8.425738798856054e-07, "loss": 1.4631, "step": 37769 }, { "epoch": 7.19, "learning_rate": 8.42516682554814e-07, "loss": 1.1656, "step": 37770 }, { "epoch": 7.19, "learning_rate": 8.424594852240229e-07, "loss": 2.7106, "step": 37771 }, { "epoch": 7.19, "learning_rate": 8.424022878932317e-07, "loss": 2.822, "step": 37772 }, { "epoch": 7.19, "learning_rate": 8.423450905624405e-07, "loss": 0.9725, "step": 37773 }, { "epoch": 7.2, "learning_rate": 8.422878932316492e-07, "loss": 1.5945, "step": 37774 }, { "epoch": 7.2, "learning_rate": 8.42230695900858e-07, "loss": 1.1099, "step": 37775 }, { "epoch": 7.2, "learning_rate": 8.421734985700668e-07, "loss": 1.5173, "step": 37776 }, { "epoch": 7.2, "learning_rate": 8.421163012392755e-07, "loss": 1.7872, "step": 37777 }, { "epoch": 7.2, "learning_rate": 8.420591039084843e-07, "loss": 1.3704, "step": 37778 }, { "epoch": 7.2, "learning_rate": 8.420019065776931e-07, "loss": 2.039, "step": 37779 }, { "epoch": 7.2, "learning_rate": 8.419447092469018e-07, "loss": 0.7917, "step": 37780 }, { "epoch": 7.2, "learning_rate": 8.418875119161105e-07, "loss": 2.8304, "step": 37781 }, { "epoch": 7.2, "learning_rate": 8.418303145853193e-07, "loss": 2.3565, "step": 37782 }, { "epoch": 7.2, "learning_rate": 8.417731172545282e-07, "loss": 2.9604, "step": 37783 }, { "epoch": 7.2, "learning_rate": 8.417159199237368e-07, "loss": 1.379, "step": 37784 }, { "epoch": 7.2, "learning_rate": 8.416587225929457e-07, "loss": 1.2975, "step": 37785 }, { "epoch": 7.2, "learning_rate": 8.416015252621545e-07, "loss": 1.2857, "step": 37786 }, { "epoch": 7.2, "learning_rate": 8.415443279313632e-07, "loss": 1.2624, "step": 37787 }, { "epoch": 7.2, "learning_rate": 8.41487130600572e-07, "loss": 2.8894, "step": 37788 }, { "epoch": 7.2, "learning_rate": 8.414299332697808e-07, "loss": 1.3205, "step": 37789 }, { "epoch": 7.2, "learning_rate": 8.413727359389896e-07, "loss": 1.773, "step": 37790 }, { "epoch": 7.2, "learning_rate": 8.413155386081983e-07, "loss": 1.0315, "step": 37791 }, { "epoch": 7.2, "learning_rate": 8.412583412774071e-07, "loss": 1.9329, "step": 37792 }, { "epoch": 7.2, "learning_rate": 8.412011439466159e-07, "loss": 2.2275, "step": 37793 }, { "epoch": 7.2, "learning_rate": 8.411439466158246e-07, "loss": 1.6272, "step": 37794 }, { "epoch": 7.2, "learning_rate": 8.410867492850333e-07, "loss": 0.9957, "step": 37795 }, { "epoch": 7.2, "learning_rate": 8.410295519542422e-07, "loss": 2.6905, "step": 37796 }, { "epoch": 7.2, "learning_rate": 8.409723546234508e-07, "loss": 1.9438, "step": 37797 }, { "epoch": 7.2, "learning_rate": 8.409151572926596e-07, "loss": 1.9985, "step": 37798 }, { "epoch": 7.2, "learning_rate": 8.408579599618685e-07, "loss": 2.2782, "step": 37799 }, { "epoch": 7.2, "learning_rate": 8.408007626310773e-07, "loss": 2.3612, "step": 37800 }, { "epoch": 7.2, "learning_rate": 8.40743565300286e-07, "loss": 2.1306, "step": 37801 }, { "epoch": 7.2, "learning_rate": 8.406863679694948e-07, "loss": 2.5028, "step": 37802 }, { "epoch": 7.2, "learning_rate": 8.406291706387036e-07, "loss": 1.2005, "step": 37803 }, { "epoch": 7.2, "learning_rate": 8.405719733079123e-07, "loss": 0.7822, "step": 37804 }, { "epoch": 7.2, "learning_rate": 8.405147759771211e-07, "loss": 3.0082, "step": 37805 }, { "epoch": 7.2, "learning_rate": 8.404575786463299e-07, "loss": 1.4772, "step": 37806 }, { "epoch": 7.2, "learning_rate": 8.404003813155387e-07, "loss": 2.2901, "step": 37807 }, { "epoch": 7.2, "learning_rate": 8.403431839847473e-07, "loss": 2.1319, "step": 37808 }, { "epoch": 7.2, "learning_rate": 8.402859866539561e-07, "loss": 1.405, "step": 37809 }, { "epoch": 7.2, "learning_rate": 8.40228789323165e-07, "loss": 1.9922, "step": 37810 }, { "epoch": 7.2, "learning_rate": 8.401715919923736e-07, "loss": 1.6231, "step": 37811 }, { "epoch": 7.2, "learning_rate": 8.401143946615824e-07, "loss": 1.321, "step": 37812 }, { "epoch": 7.2, "learning_rate": 8.400571973307913e-07, "loss": 3.0005, "step": 37813 }, { "epoch": 7.2, "learning_rate": 8.400000000000001e-07, "loss": 1.0466, "step": 37814 }, { "epoch": 7.2, "learning_rate": 8.399428026692088e-07, "loss": 2.8394, "step": 37815 }, { "epoch": 7.2, "learning_rate": 8.398856053384176e-07, "loss": 1.8432, "step": 37816 }, { "epoch": 7.2, "learning_rate": 8.398284080076264e-07, "loss": 1.3605, "step": 37817 }, { "epoch": 7.2, "learning_rate": 8.397712106768351e-07, "loss": 3.0916, "step": 37818 }, { "epoch": 7.2, "learning_rate": 8.397140133460439e-07, "loss": 3.432, "step": 37819 }, { "epoch": 7.2, "learning_rate": 8.396568160152527e-07, "loss": 0.4905, "step": 37820 }, { "epoch": 7.2, "learning_rate": 8.395996186844613e-07, "loss": 0.2617, "step": 37821 }, { "epoch": 7.2, "learning_rate": 8.395424213536701e-07, "loss": 1.7799, "step": 37822 }, { "epoch": 7.2, "learning_rate": 8.394852240228789e-07, "loss": 2.5372, "step": 37823 }, { "epoch": 7.2, "learning_rate": 8.394280266920878e-07, "loss": 1.5466, "step": 37824 }, { "epoch": 7.2, "learning_rate": 8.393708293612964e-07, "loss": 2.0771, "step": 37825 }, { "epoch": 7.2, "learning_rate": 8.393136320305052e-07, "loss": 1.8585, "step": 37826 }, { "epoch": 7.21, "learning_rate": 8.392564346997141e-07, "loss": 2.8253, "step": 37827 }, { "epoch": 7.21, "learning_rate": 8.391992373689227e-07, "loss": 2.0959, "step": 37828 }, { "epoch": 7.21, "learning_rate": 8.391420400381316e-07, "loss": 1.7812, "step": 37829 }, { "epoch": 7.21, "learning_rate": 8.390848427073404e-07, "loss": 1.2986, "step": 37830 }, { "epoch": 7.21, "learning_rate": 8.390276453765492e-07, "loss": 1.2538, "step": 37831 }, { "epoch": 7.21, "learning_rate": 8.389704480457579e-07, "loss": 3.3574, "step": 37832 }, { "epoch": 7.21, "learning_rate": 8.389132507149667e-07, "loss": 1.2865, "step": 37833 }, { "epoch": 7.21, "learning_rate": 8.388560533841755e-07, "loss": 2.9748, "step": 37834 }, { "epoch": 7.21, "learning_rate": 8.387988560533841e-07, "loss": 1.3142, "step": 37835 }, { "epoch": 7.21, "learning_rate": 8.387416587225929e-07, "loss": 2.8438, "step": 37836 }, { "epoch": 7.21, "learning_rate": 8.386844613918017e-07, "loss": 2.4757, "step": 37837 }, { "epoch": 7.21, "learning_rate": 8.386272640610106e-07, "loss": 1.2714, "step": 37838 }, { "epoch": 7.21, "learning_rate": 8.385700667302192e-07, "loss": 1.024, "step": 37839 }, { "epoch": 7.21, "learning_rate": 8.385128693994281e-07, "loss": 0.8408, "step": 37840 }, { "epoch": 7.21, "learning_rate": 8.384556720686369e-07, "loss": 2.7199, "step": 37841 }, { "epoch": 7.21, "learning_rate": 8.383984747378455e-07, "loss": 1.0852, "step": 37842 }, { "epoch": 7.21, "learning_rate": 8.383412774070544e-07, "loss": 2.0585, "step": 37843 }, { "epoch": 7.21, "learning_rate": 8.382840800762632e-07, "loss": 0.94, "step": 37844 }, { "epoch": 7.21, "learning_rate": 8.382268827454719e-07, "loss": 3.2702, "step": 37845 }, { "epoch": 7.21, "learning_rate": 8.381696854146807e-07, "loss": 0.4722, "step": 37846 }, { "epoch": 7.21, "learning_rate": 8.381124880838895e-07, "loss": 0.831, "step": 37847 }, { "epoch": 7.21, "learning_rate": 8.380552907530983e-07, "loss": 1.5246, "step": 37848 }, { "epoch": 7.21, "learning_rate": 8.379980934223069e-07, "loss": 1.9002, "step": 37849 }, { "epoch": 7.21, "learning_rate": 8.379408960915157e-07, "loss": 1.5366, "step": 37850 }, { "epoch": 7.21, "learning_rate": 8.378836987607245e-07, "loss": 2.9153, "step": 37851 }, { "epoch": 7.21, "learning_rate": 8.378265014299332e-07, "loss": 2.3645, "step": 37852 }, { "epoch": 7.21, "learning_rate": 8.37769304099142e-07, "loss": 0.9805, "step": 37853 }, { "epoch": 7.21, "learning_rate": 8.377121067683509e-07, "loss": 1.8756, "step": 37854 }, { "epoch": 7.21, "learning_rate": 8.376549094375597e-07, "loss": 1.0855, "step": 37855 }, { "epoch": 7.21, "learning_rate": 8.375977121067683e-07, "loss": 2.3622, "step": 37856 }, { "epoch": 7.21, "learning_rate": 8.375405147759772e-07, "loss": 1.9574, "step": 37857 }, { "epoch": 7.21, "learning_rate": 8.37483317445186e-07, "loss": 2.0102, "step": 37858 }, { "epoch": 7.21, "learning_rate": 8.374261201143947e-07, "loss": 2.1809, "step": 37859 }, { "epoch": 7.21, "learning_rate": 8.373689227836035e-07, "loss": 1.1192, "step": 37860 }, { "epoch": 7.21, "learning_rate": 8.373117254528123e-07, "loss": 2.1297, "step": 37861 }, { "epoch": 7.21, "learning_rate": 8.372545281220209e-07, "loss": 1.5478, "step": 37862 }, { "epoch": 7.21, "learning_rate": 8.371973307912297e-07, "loss": 0.7277, "step": 37863 }, { "epoch": 7.21, "learning_rate": 8.371401334604385e-07, "loss": 1.7789, "step": 37864 }, { "epoch": 7.21, "learning_rate": 8.370829361296473e-07, "loss": 0.7375, "step": 37865 }, { "epoch": 7.21, "learning_rate": 8.37025738798856e-07, "loss": 1.7942, "step": 37866 }, { "epoch": 7.21, "learning_rate": 8.369685414680648e-07, "loss": 1.0969, "step": 37867 }, { "epoch": 7.21, "learning_rate": 8.369113441372737e-07, "loss": 0.8677, "step": 37868 }, { "epoch": 7.21, "learning_rate": 8.368541468064823e-07, "loss": 1.0453, "step": 37869 }, { "epoch": 7.21, "learning_rate": 8.367969494756911e-07, "loss": 0.9411, "step": 37870 }, { "epoch": 7.21, "learning_rate": 8.367397521449e-07, "loss": 0.6654, "step": 37871 }, { "epoch": 7.21, "learning_rate": 8.366825548141088e-07, "loss": 2.1441, "step": 37872 }, { "epoch": 7.21, "learning_rate": 8.366253574833175e-07, "loss": 2.0264, "step": 37873 }, { "epoch": 7.21, "learning_rate": 8.365681601525263e-07, "loss": 2.045, "step": 37874 }, { "epoch": 7.21, "learning_rate": 8.36510962821735e-07, "loss": 2.3088, "step": 37875 }, { "epoch": 7.21, "learning_rate": 8.364537654909437e-07, "loss": 1.7455, "step": 37876 }, { "epoch": 7.21, "learning_rate": 8.363965681601525e-07, "loss": 2.1218, "step": 37877 }, { "epoch": 7.21, "learning_rate": 8.363393708293613e-07, "loss": 1.8825, "step": 37878 }, { "epoch": 7.22, "learning_rate": 8.362821734985701e-07, "loss": 1.9497, "step": 37879 }, { "epoch": 7.22, "learning_rate": 8.362249761677788e-07, "loss": 1.8835, "step": 37880 }, { "epoch": 7.22, "learning_rate": 8.361677788369876e-07, "loss": 3.2429, "step": 37881 }, { "epoch": 7.22, "learning_rate": 8.361105815061965e-07, "loss": 2.6285, "step": 37882 }, { "epoch": 7.22, "learning_rate": 8.360533841754051e-07, "loss": 2.8349, "step": 37883 }, { "epoch": 7.22, "learning_rate": 8.359961868446139e-07, "loss": 0.7694, "step": 37884 }, { "epoch": 7.22, "learning_rate": 8.359389895138228e-07, "loss": 2.8032, "step": 37885 }, { "epoch": 7.22, "learning_rate": 8.358817921830314e-07, "loss": 0.5717, "step": 37886 }, { "epoch": 7.22, "learning_rate": 8.358245948522403e-07, "loss": 2.9553, "step": 37887 }, { "epoch": 7.22, "learning_rate": 8.35767397521449e-07, "loss": 0.4581, "step": 37888 }, { "epoch": 7.22, "learning_rate": 8.357102001906578e-07, "loss": 2.5696, "step": 37889 }, { "epoch": 7.22, "learning_rate": 8.356530028598665e-07, "loss": 1.2003, "step": 37890 }, { "epoch": 7.22, "learning_rate": 8.355958055290753e-07, "loss": 3.6296, "step": 37891 }, { "epoch": 7.22, "learning_rate": 8.355386081982841e-07, "loss": 1.8868, "step": 37892 }, { "epoch": 7.22, "learning_rate": 8.354814108674928e-07, "loss": 2.1316, "step": 37893 }, { "epoch": 7.22, "learning_rate": 8.354242135367016e-07, "loss": 1.9738, "step": 37894 }, { "epoch": 7.22, "learning_rate": 8.353670162059104e-07, "loss": 3.3929, "step": 37895 }, { "epoch": 7.22, "learning_rate": 8.353098188751193e-07, "loss": 2.1034, "step": 37896 }, { "epoch": 7.22, "learning_rate": 8.352526215443279e-07, "loss": 1.7282, "step": 37897 }, { "epoch": 7.22, "learning_rate": 8.351954242135368e-07, "loss": 0.8585, "step": 37898 }, { "epoch": 7.22, "learning_rate": 8.351382268827456e-07, "loss": 3.2406, "step": 37899 }, { "epoch": 7.22, "learning_rate": 8.350810295519542e-07, "loss": 1.4617, "step": 37900 }, { "epoch": 7.22, "learning_rate": 8.350238322211631e-07, "loss": 0.9834, "step": 37901 }, { "epoch": 7.22, "learning_rate": 8.349666348903718e-07, "loss": 1.5365, "step": 37902 }, { "epoch": 7.22, "learning_rate": 8.349094375595806e-07, "loss": 1.7325, "step": 37903 }, { "epoch": 7.22, "learning_rate": 8.348522402287893e-07, "loss": 2.1596, "step": 37904 }, { "epoch": 7.22, "learning_rate": 8.347950428979981e-07, "loss": 0.9869, "step": 37905 }, { "epoch": 7.22, "learning_rate": 8.347378455672069e-07, "loss": 1.4164, "step": 37906 }, { "epoch": 7.22, "learning_rate": 8.346806482364156e-07, "loss": 1.4399, "step": 37907 }, { "epoch": 7.22, "learning_rate": 8.346234509056244e-07, "loss": 1.4355, "step": 37908 }, { "epoch": 7.22, "learning_rate": 8.345662535748332e-07, "loss": 1.8038, "step": 37909 }, { "epoch": 7.22, "learning_rate": 8.345090562440419e-07, "loss": 2.8896, "step": 37910 }, { "epoch": 7.22, "learning_rate": 8.344518589132507e-07, "loss": 2.2146, "step": 37911 }, { "epoch": 7.22, "learning_rate": 8.343946615824596e-07, "loss": 3.0784, "step": 37912 }, { "epoch": 7.22, "learning_rate": 8.343374642516684e-07, "loss": 1.2239, "step": 37913 }, { "epoch": 7.22, "learning_rate": 8.34280266920877e-07, "loss": 0.7847, "step": 37914 }, { "epoch": 7.22, "learning_rate": 8.342230695900859e-07, "loss": 2.0309, "step": 37915 }, { "epoch": 7.22, "learning_rate": 8.341658722592946e-07, "loss": 1.5146, "step": 37916 }, { "epoch": 7.22, "learning_rate": 8.341086749285033e-07, "loss": 1.2222, "step": 37917 }, { "epoch": 7.22, "learning_rate": 8.340514775977121e-07, "loss": 1.8564, "step": 37918 }, { "epoch": 7.22, "learning_rate": 8.339942802669209e-07, "loss": 1.1486, "step": 37919 }, { "epoch": 7.22, "learning_rate": 8.339370829361297e-07, "loss": 1.504, "step": 37920 }, { "epoch": 7.22, "learning_rate": 8.338798856053384e-07, "loss": 1.302, "step": 37921 }, { "epoch": 7.22, "learning_rate": 8.338226882745472e-07, "loss": 2.0888, "step": 37922 }, { "epoch": 7.22, "learning_rate": 8.33765490943756e-07, "loss": 1.8991, "step": 37923 }, { "epoch": 7.22, "learning_rate": 8.337082936129647e-07, "loss": 3.5011, "step": 37924 }, { "epoch": 7.22, "learning_rate": 8.336510962821735e-07, "loss": 2.8723, "step": 37925 }, { "epoch": 7.22, "learning_rate": 8.335938989513824e-07, "loss": 1.7218, "step": 37926 }, { "epoch": 7.22, "learning_rate": 8.33536701620591e-07, "loss": 1.1613, "step": 37927 }, { "epoch": 7.22, "learning_rate": 8.334795042897997e-07, "loss": 2.8805, "step": 37928 }, { "epoch": 7.22, "learning_rate": 8.334223069590086e-07, "loss": 1.6347, "step": 37929 }, { "epoch": 7.22, "learning_rate": 8.333651096282174e-07, "loss": 2.2814, "step": 37930 }, { "epoch": 7.22, "learning_rate": 8.333079122974261e-07, "loss": 2.9034, "step": 37931 }, { "epoch": 7.23, "learning_rate": 8.332507149666349e-07, "loss": 1.4157, "step": 37932 }, { "epoch": 7.23, "learning_rate": 8.331935176358437e-07, "loss": 3.1969, "step": 37933 }, { "epoch": 7.23, "learning_rate": 8.331363203050524e-07, "loss": 2.2979, "step": 37934 }, { "epoch": 7.23, "learning_rate": 8.330791229742612e-07, "loss": 2.0377, "step": 37935 }, { "epoch": 7.23, "learning_rate": 8.3302192564347e-07, "loss": 1.9104, "step": 37936 }, { "epoch": 7.23, "learning_rate": 8.329647283126788e-07, "loss": 1.3769, "step": 37937 }, { "epoch": 7.23, "learning_rate": 8.329075309818875e-07, "loss": 1.4467, "step": 37938 }, { "epoch": 7.23, "learning_rate": 8.328503336510963e-07, "loss": 1.3549, "step": 37939 }, { "epoch": 7.23, "learning_rate": 8.327931363203052e-07, "loss": 1.7306, "step": 37940 }, { "epoch": 7.23, "learning_rate": 8.327359389895137e-07, "loss": 0.3737, "step": 37941 }, { "epoch": 7.23, "learning_rate": 8.326787416587226e-07, "loss": 2.3244, "step": 37942 }, { "epoch": 7.23, "learning_rate": 8.326215443279314e-07, "loss": 1.8488, "step": 37943 }, { "epoch": 7.23, "learning_rate": 8.325643469971402e-07, "loss": 2.6231, "step": 37944 }, { "epoch": 7.23, "learning_rate": 8.325071496663489e-07, "loss": 0.8966, "step": 37945 }, { "epoch": 7.23, "learning_rate": 8.324499523355577e-07, "loss": 1.659, "step": 37946 }, { "epoch": 7.23, "learning_rate": 8.323927550047665e-07, "loss": 1.3189, "step": 37947 }, { "epoch": 7.23, "learning_rate": 8.323355576739752e-07, "loss": 2.4336, "step": 37948 }, { "epoch": 7.23, "learning_rate": 8.32278360343184e-07, "loss": 1.4124, "step": 37949 }, { "epoch": 7.23, "learning_rate": 8.322211630123928e-07, "loss": 0.5763, "step": 37950 }, { "epoch": 7.23, "learning_rate": 8.321639656816015e-07, "loss": 2.8355, "step": 37951 }, { "epoch": 7.23, "learning_rate": 8.321067683508103e-07, "loss": 4.0602, "step": 37952 }, { "epoch": 7.23, "learning_rate": 8.320495710200191e-07, "loss": 1.9188, "step": 37953 }, { "epoch": 7.23, "learning_rate": 8.31992373689228e-07, "loss": 1.8635, "step": 37954 }, { "epoch": 7.23, "learning_rate": 8.319351763584365e-07, "loss": 1.8998, "step": 37955 }, { "epoch": 7.23, "learning_rate": 8.318779790276454e-07, "loss": 2.2206, "step": 37956 }, { "epoch": 7.23, "learning_rate": 8.318207816968542e-07, "loss": 2.2151, "step": 37957 }, { "epoch": 7.23, "learning_rate": 8.317635843660628e-07, "loss": 1.7401, "step": 37958 }, { "epoch": 7.23, "learning_rate": 8.317063870352717e-07, "loss": 2.1611, "step": 37959 }, { "epoch": 7.23, "learning_rate": 8.316491897044805e-07, "loss": 1.657, "step": 37960 }, { "epoch": 7.23, "learning_rate": 8.315919923736893e-07, "loss": 3.1739, "step": 37961 }, { "epoch": 7.23, "learning_rate": 8.31534795042898e-07, "loss": 1.4523, "step": 37962 }, { "epoch": 7.23, "learning_rate": 8.314775977121068e-07, "loss": 1.6767, "step": 37963 }, { "epoch": 7.23, "learning_rate": 8.314204003813156e-07, "loss": 2.7681, "step": 37964 }, { "epoch": 7.23, "learning_rate": 8.313632030505243e-07, "loss": 1.8755, "step": 37965 }, { "epoch": 7.23, "learning_rate": 8.313060057197331e-07, "loss": 1.6087, "step": 37966 }, { "epoch": 7.23, "learning_rate": 8.312488083889419e-07, "loss": 3.0563, "step": 37967 }, { "epoch": 7.23, "learning_rate": 8.311916110581505e-07, "loss": 1.142, "step": 37968 }, { "epoch": 7.23, "learning_rate": 8.311344137273593e-07, "loss": 2.1856, "step": 37969 }, { "epoch": 7.23, "learning_rate": 8.310772163965682e-07, "loss": 1.0192, "step": 37970 }, { "epoch": 7.23, "learning_rate": 8.31020019065777e-07, "loss": 2.4122, "step": 37971 }, { "epoch": 7.23, "learning_rate": 8.309628217349856e-07, "loss": 2.2152, "step": 37972 }, { "epoch": 7.23, "learning_rate": 8.309056244041945e-07, "loss": 1.9036, "step": 37973 }, { "epoch": 7.23, "learning_rate": 8.308484270734033e-07, "loss": 2.0672, "step": 37974 }, { "epoch": 7.23, "learning_rate": 8.30791229742612e-07, "loss": 1.8995, "step": 37975 }, { "epoch": 7.23, "learning_rate": 8.307340324118208e-07, "loss": 1.3915, "step": 37976 }, { "epoch": 7.23, "learning_rate": 8.306768350810296e-07, "loss": 1.1165, "step": 37977 }, { "epoch": 7.23, "learning_rate": 8.306196377502384e-07, "loss": 2.012, "step": 37978 }, { "epoch": 7.23, "learning_rate": 8.305624404194471e-07, "loss": 1.6461, "step": 37979 }, { "epoch": 7.23, "learning_rate": 8.305052430886559e-07, "loss": 2.2375, "step": 37980 }, { "epoch": 7.23, "learning_rate": 8.304480457578647e-07, "loss": 1.9004, "step": 37981 }, { "epoch": 7.23, "learning_rate": 8.303908484270733e-07, "loss": 2.0202, "step": 37982 }, { "epoch": 7.23, "learning_rate": 8.303336510962821e-07, "loss": 2.6574, "step": 37983 }, { "epoch": 7.24, "learning_rate": 8.30276453765491e-07, "loss": 1.2185, "step": 37984 }, { "epoch": 7.24, "learning_rate": 8.302192564346998e-07, "loss": 2.2269, "step": 37985 }, { "epoch": 7.24, "learning_rate": 8.301620591039085e-07, "loss": 2.4275, "step": 37986 }, { "epoch": 7.24, "learning_rate": 8.301048617731173e-07, "loss": 3.2756, "step": 37987 }, { "epoch": 7.24, "learning_rate": 8.300476644423261e-07, "loss": 2.7699, "step": 37988 }, { "epoch": 7.24, "learning_rate": 8.299904671115348e-07, "loss": 1.6699, "step": 37989 }, { "epoch": 7.24, "learning_rate": 8.299332697807436e-07, "loss": 0.5254, "step": 37990 }, { "epoch": 7.24, "learning_rate": 8.298760724499524e-07, "loss": 1.4846, "step": 37991 }, { "epoch": 7.24, "learning_rate": 8.298188751191611e-07, "loss": 1.1132, "step": 37992 }, { "epoch": 7.24, "learning_rate": 8.297616777883699e-07, "loss": 1.1963, "step": 37993 }, { "epoch": 7.24, "learning_rate": 8.297044804575787e-07, "loss": 1.5209, "step": 37994 }, { "epoch": 7.24, "learning_rate": 8.296472831267876e-07, "loss": 2.5618, "step": 37995 }, { "epoch": 7.24, "learning_rate": 8.295900857959961e-07, "loss": 2.3724, "step": 37996 }, { "epoch": 7.24, "learning_rate": 8.295328884652049e-07, "loss": 3.2013, "step": 37997 }, { "epoch": 7.24, "learning_rate": 8.294756911344138e-07, "loss": 1.7424, "step": 37998 }, { "epoch": 7.24, "learning_rate": 8.294184938036224e-07, "loss": 1.6477, "step": 37999 }, { "epoch": 7.24, "learning_rate": 8.293612964728313e-07, "loss": 1.0375, "step": 38000 }, { "epoch": 7.24, "learning_rate": 8.293040991420401e-07, "loss": 0.8768, "step": 38001 }, { "epoch": 7.24, "learning_rate": 8.292469018112489e-07, "loss": 0.7891, "step": 38002 }, { "epoch": 7.24, "learning_rate": 8.291897044804576e-07, "loss": 2.1662, "step": 38003 }, { "epoch": 7.24, "learning_rate": 8.291325071496664e-07, "loss": 0.6219, "step": 38004 }, { "epoch": 7.24, "learning_rate": 8.290753098188752e-07, "loss": 1.4937, "step": 38005 }, { "epoch": 7.24, "learning_rate": 8.290181124880839e-07, "loss": 1.0675, "step": 38006 }, { "epoch": 7.24, "learning_rate": 8.289609151572927e-07, "loss": 1.9098, "step": 38007 }, { "epoch": 7.24, "learning_rate": 8.289037178265015e-07, "loss": 1.0736, "step": 38008 }, { "epoch": 7.24, "learning_rate": 8.288465204957104e-07, "loss": 1.1994, "step": 38009 }, { "epoch": 7.24, "learning_rate": 8.287893231649189e-07, "loss": 0.8348, "step": 38010 }, { "epoch": 7.24, "learning_rate": 8.287321258341277e-07, "loss": 1.1629, "step": 38011 }, { "epoch": 7.24, "learning_rate": 8.286749285033366e-07, "loss": 1.3042, "step": 38012 }, { "epoch": 7.24, "learning_rate": 8.286177311725452e-07, "loss": 2.3439, "step": 38013 }, { "epoch": 7.24, "learning_rate": 8.285605338417541e-07, "loss": 0.7502, "step": 38014 }, { "epoch": 7.24, "learning_rate": 8.285033365109629e-07, "loss": 2.8064, "step": 38015 }, { "epoch": 7.24, "learning_rate": 8.284461391801715e-07, "loss": 1.5793, "step": 38016 }, { "epoch": 7.24, "learning_rate": 8.283889418493804e-07, "loss": 3.0053, "step": 38017 }, { "epoch": 7.24, "learning_rate": 8.283317445185892e-07, "loss": 2.0036, "step": 38018 }, { "epoch": 7.24, "learning_rate": 8.28274547187798e-07, "loss": 1.4585, "step": 38019 }, { "epoch": 7.24, "learning_rate": 8.282173498570067e-07, "loss": 2.0084, "step": 38020 }, { "epoch": 7.24, "learning_rate": 8.281601525262155e-07, "loss": 1.7214, "step": 38021 }, { "epoch": 7.24, "learning_rate": 8.281029551954242e-07, "loss": 0.6176, "step": 38022 }, { "epoch": 7.24, "learning_rate": 8.280457578646329e-07, "loss": 2.614, "step": 38023 }, { "epoch": 7.24, "learning_rate": 8.279885605338417e-07, "loss": 2.2982, "step": 38024 }, { "epoch": 7.24, "learning_rate": 8.279313632030505e-07, "loss": 2.1367, "step": 38025 }, { "epoch": 7.24, "learning_rate": 8.278741658722594e-07, "loss": 2.3263, "step": 38026 }, { "epoch": 7.24, "learning_rate": 8.27816968541468e-07, "loss": 1.5363, "step": 38027 }, { "epoch": 7.24, "learning_rate": 8.277597712106769e-07, "loss": 1.2451, "step": 38028 }, { "epoch": 7.24, "learning_rate": 8.277025738798857e-07, "loss": 0.7904, "step": 38029 }, { "epoch": 7.24, "learning_rate": 8.276453765490944e-07, "loss": 1.4235, "step": 38030 }, { "epoch": 7.24, "learning_rate": 8.275881792183032e-07, "loss": 1.4543, "step": 38031 }, { "epoch": 7.24, "learning_rate": 8.27530981887512e-07, "loss": 1.2059, "step": 38032 }, { "epoch": 7.24, "learning_rate": 8.274737845567207e-07, "loss": 1.7831, "step": 38033 }, { "epoch": 7.24, "learning_rate": 8.274165872259295e-07, "loss": 1.6813, "step": 38034 }, { "epoch": 7.24, "learning_rate": 8.273593898951383e-07, "loss": 1.3593, "step": 38035 }, { "epoch": 7.24, "learning_rate": 8.27302192564347e-07, "loss": 0.5843, "step": 38036 }, { "epoch": 7.25, "learning_rate": 8.272449952335557e-07, "loss": 0.7625, "step": 38037 }, { "epoch": 7.25, "learning_rate": 8.271877979027645e-07, "loss": 2.552, "step": 38038 }, { "epoch": 7.25, "learning_rate": 8.271306005719733e-07, "loss": 2.0161, "step": 38039 }, { "epoch": 7.25, "learning_rate": 8.27073403241182e-07, "loss": 2.4913, "step": 38040 }, { "epoch": 7.25, "learning_rate": 8.270162059103908e-07, "loss": 1.3214, "step": 38041 }, { "epoch": 7.25, "learning_rate": 8.269590085795997e-07, "loss": 1.416, "step": 38042 }, { "epoch": 7.25, "learning_rate": 8.269018112488085e-07, "loss": 1.9801, "step": 38043 }, { "epoch": 7.25, "learning_rate": 8.268446139180172e-07, "loss": 1.9483, "step": 38044 }, { "epoch": 7.25, "learning_rate": 8.26787416587226e-07, "loss": 2.3138, "step": 38045 }, { "epoch": 7.25, "learning_rate": 8.267302192564348e-07, "loss": 2.0885, "step": 38046 }, { "epoch": 7.25, "learning_rate": 8.266730219256435e-07, "loss": 1.3159, "step": 38047 }, { "epoch": 7.25, "learning_rate": 8.266158245948523e-07, "loss": 1.8732, "step": 38048 }, { "epoch": 7.25, "learning_rate": 8.26558627264061e-07, "loss": 1.5474, "step": 38049 }, { "epoch": 7.25, "learning_rate": 8.265014299332698e-07, "loss": 1.215, "step": 38050 }, { "epoch": 7.25, "learning_rate": 8.264442326024785e-07, "loss": 1.7186, "step": 38051 }, { "epoch": 7.25, "learning_rate": 8.263870352716873e-07, "loss": 1.7059, "step": 38052 }, { "epoch": 7.25, "learning_rate": 8.263298379408962e-07, "loss": 1.2913, "step": 38053 }, { "epoch": 7.25, "learning_rate": 8.262726406101048e-07, "loss": 0.9445, "step": 38054 }, { "epoch": 7.25, "learning_rate": 8.262154432793136e-07, "loss": 1.3124, "step": 38055 }, { "epoch": 7.25, "learning_rate": 8.261582459485225e-07, "loss": 0.6536, "step": 38056 }, { "epoch": 7.25, "learning_rate": 8.261010486177311e-07, "loss": 1.0231, "step": 38057 }, { "epoch": 7.25, "learning_rate": 8.2604385128694e-07, "loss": 1.1435, "step": 38058 }, { "epoch": 7.25, "learning_rate": 8.259866539561488e-07, "loss": 1.8145, "step": 38059 }, { "epoch": 7.25, "learning_rate": 8.259294566253576e-07, "loss": 2.502, "step": 38060 }, { "epoch": 7.25, "learning_rate": 8.258722592945663e-07, "loss": 3.7887, "step": 38061 }, { "epoch": 7.25, "learning_rate": 8.25815061963775e-07, "loss": 1.3439, "step": 38062 }, { "epoch": 7.25, "learning_rate": 8.257578646329838e-07, "loss": 2.4289, "step": 38063 }, { "epoch": 7.25, "learning_rate": 8.257006673021925e-07, "loss": 0.4622, "step": 38064 }, { "epoch": 7.25, "learning_rate": 8.256434699714013e-07, "loss": 0.9935, "step": 38065 }, { "epoch": 7.25, "learning_rate": 8.255862726406101e-07, "loss": 1.5134, "step": 38066 }, { "epoch": 7.25, "learning_rate": 8.25529075309819e-07, "loss": 2.3598, "step": 38067 }, { "epoch": 7.25, "learning_rate": 8.254718779790276e-07, "loss": 0.6317, "step": 38068 }, { "epoch": 7.25, "learning_rate": 8.254146806482364e-07, "loss": 0.7487, "step": 38069 }, { "epoch": 7.25, "learning_rate": 8.253574833174453e-07, "loss": 2.7011, "step": 38070 }, { "epoch": 7.25, "learning_rate": 8.253002859866539e-07, "loss": 1.6293, "step": 38071 }, { "epoch": 7.25, "learning_rate": 8.252430886558628e-07, "loss": 1.514, "step": 38072 }, { "epoch": 7.25, "learning_rate": 8.251858913250716e-07, "loss": 2.3768, "step": 38073 }, { "epoch": 7.25, "learning_rate": 8.251286939942804e-07, "loss": 2.1429, "step": 38074 }, { "epoch": 7.25, "learning_rate": 8.25071496663489e-07, "loss": 4.4024, "step": 38075 }, { "epoch": 7.25, "learning_rate": 8.250142993326978e-07, "loss": 1.6018, "step": 38076 }, { "epoch": 7.25, "learning_rate": 8.249571020019066e-07, "loss": 1.73, "step": 38077 }, { "epoch": 7.25, "learning_rate": 8.248999046711153e-07, "loss": 1.1292, "step": 38078 }, { "epoch": 7.25, "learning_rate": 8.248427073403241e-07, "loss": 2.0017, "step": 38079 }, { "epoch": 7.25, "learning_rate": 8.247855100095329e-07, "loss": 2.0799, "step": 38080 }, { "epoch": 7.25, "learning_rate": 8.247283126787416e-07, "loss": 1.9069, "step": 38081 }, { "epoch": 7.25, "learning_rate": 8.246711153479504e-07, "loss": 2.4938, "step": 38082 }, { "epoch": 7.25, "learning_rate": 8.246139180171592e-07, "loss": 1.3299, "step": 38083 }, { "epoch": 7.25, "learning_rate": 8.245567206863681e-07, "loss": 1.7985, "step": 38084 }, { "epoch": 7.25, "learning_rate": 8.244995233555767e-07, "loss": 1.252, "step": 38085 }, { "epoch": 7.25, "learning_rate": 8.244423260247856e-07, "loss": 2.1156, "step": 38086 }, { "epoch": 7.25, "learning_rate": 8.243851286939944e-07, "loss": 2.4429, "step": 38087 }, { "epoch": 7.25, "learning_rate": 8.24327931363203e-07, "loss": 1.8602, "step": 38088 }, { "epoch": 7.26, "learning_rate": 8.242707340324118e-07, "loss": 1.5, "step": 38089 }, { "epoch": 7.26, "learning_rate": 8.242135367016206e-07, "loss": 1.0333, "step": 38090 }, { "epoch": 7.26, "learning_rate": 8.241563393708294e-07, "loss": 1.0381, "step": 38091 }, { "epoch": 7.26, "learning_rate": 8.240991420400381e-07, "loss": 0.8497, "step": 38092 }, { "epoch": 7.26, "learning_rate": 8.240419447092469e-07, "loss": 0.842, "step": 38093 }, { "epoch": 7.26, "learning_rate": 8.239847473784557e-07, "loss": 2.0755, "step": 38094 }, { "epoch": 7.26, "learning_rate": 8.239275500476644e-07, "loss": 1.2972, "step": 38095 }, { "epoch": 7.26, "learning_rate": 8.238703527168732e-07, "loss": 2.5268, "step": 38096 }, { "epoch": 7.26, "learning_rate": 8.238131553860821e-07, "loss": 2.4711, "step": 38097 }, { "epoch": 7.26, "learning_rate": 8.237559580552907e-07, "loss": 3.6108, "step": 38098 }, { "epoch": 7.26, "learning_rate": 8.236987607244995e-07, "loss": 1.9892, "step": 38099 }, { "epoch": 7.26, "learning_rate": 8.236415633937084e-07, "loss": 1.7482, "step": 38100 }, { "epoch": 7.26, "learning_rate": 8.235843660629172e-07, "loss": 1.7108, "step": 38101 }, { "epoch": 7.26, "learning_rate": 8.235271687321258e-07, "loss": 2.9134, "step": 38102 }, { "epoch": 7.26, "learning_rate": 8.234699714013346e-07, "loss": 1.5532, "step": 38103 }, { "epoch": 7.26, "learning_rate": 8.234127740705434e-07, "loss": 1.2713, "step": 38104 }, { "epoch": 7.26, "learning_rate": 8.233555767397521e-07, "loss": 1.9605, "step": 38105 }, { "epoch": 7.26, "learning_rate": 8.232983794089609e-07, "loss": 2.5062, "step": 38106 }, { "epoch": 7.26, "learning_rate": 8.232411820781697e-07, "loss": 1.7358, "step": 38107 }, { "epoch": 7.26, "learning_rate": 8.231839847473785e-07, "loss": 0.7634, "step": 38108 }, { "epoch": 7.26, "learning_rate": 8.231267874165872e-07, "loss": 1.261, "step": 38109 }, { "epoch": 7.26, "learning_rate": 8.23069590085796e-07, "loss": 1.4428, "step": 38110 }, { "epoch": 7.26, "learning_rate": 8.230123927550049e-07, "loss": 2.8442, "step": 38111 }, { "epoch": 7.26, "learning_rate": 8.229551954242135e-07, "loss": 1.5217, "step": 38112 }, { "epoch": 7.26, "learning_rate": 8.228979980934223e-07, "loss": 2.3477, "step": 38113 }, { "epoch": 7.26, "learning_rate": 8.228408007626312e-07, "loss": 2.5569, "step": 38114 }, { "epoch": 7.26, "learning_rate": 8.2278360343184e-07, "loss": 2.0225, "step": 38115 }, { "epoch": 7.26, "learning_rate": 8.227264061010486e-07, "loss": 2.2497, "step": 38116 }, { "epoch": 7.26, "learning_rate": 8.226692087702574e-07, "loss": 2.2937, "step": 38117 }, { "epoch": 7.26, "learning_rate": 8.226120114394662e-07, "loss": 1.0197, "step": 38118 }, { "epoch": 7.26, "learning_rate": 8.225548141086749e-07, "loss": 2.4504, "step": 38119 }, { "epoch": 7.26, "learning_rate": 8.224976167778837e-07, "loss": 1.5241, "step": 38120 }, { "epoch": 7.26, "learning_rate": 8.224404194470925e-07, "loss": 1.2898, "step": 38121 }, { "epoch": 7.26, "learning_rate": 8.223832221163012e-07, "loss": 1.6639, "step": 38122 }, { "epoch": 7.26, "learning_rate": 8.2232602478551e-07, "loss": 0.9571, "step": 38123 }, { "epoch": 7.26, "learning_rate": 8.222688274547188e-07, "loss": 1.7575, "step": 38124 }, { "epoch": 7.26, "learning_rate": 8.222116301239277e-07, "loss": 3.1966, "step": 38125 }, { "epoch": 7.26, "learning_rate": 8.221544327931363e-07, "loss": 1.9332, "step": 38126 }, { "epoch": 7.26, "learning_rate": 8.220972354623451e-07, "loss": 2.4053, "step": 38127 }, { "epoch": 7.26, "learning_rate": 8.22040038131554e-07, "loss": 2.6411, "step": 38128 }, { "epoch": 7.26, "learning_rate": 8.219828408007625e-07, "loss": 2.6266, "step": 38129 }, { "epoch": 7.26, "learning_rate": 8.219256434699714e-07, "loss": 4.3116, "step": 38130 }, { "epoch": 7.26, "learning_rate": 8.218684461391802e-07, "loss": 0.9114, "step": 38131 }, { "epoch": 7.26, "learning_rate": 8.21811248808389e-07, "loss": 1.8585, "step": 38132 }, { "epoch": 7.26, "learning_rate": 8.217540514775977e-07, "loss": 2.7925, "step": 38133 }, { "epoch": 7.26, "learning_rate": 8.216968541468065e-07, "loss": 1.0499, "step": 38134 }, { "epoch": 7.26, "learning_rate": 8.216396568160153e-07, "loss": 1.7766, "step": 38135 }, { "epoch": 7.26, "learning_rate": 8.21582459485224e-07, "loss": 1.5071, "step": 38136 }, { "epoch": 7.26, "learning_rate": 8.215252621544328e-07, "loss": 1.4174, "step": 38137 }, { "epoch": 7.26, "learning_rate": 8.214680648236416e-07, "loss": 1.9555, "step": 38138 }, { "epoch": 7.26, "learning_rate": 8.214108674928505e-07, "loss": 1.701, "step": 38139 }, { "epoch": 7.26, "learning_rate": 8.213536701620591e-07, "loss": 1.868, "step": 38140 }, { "epoch": 7.26, "learning_rate": 8.21296472831268e-07, "loss": 0.8667, "step": 38141 }, { "epoch": 7.27, "learning_rate": 8.212392755004768e-07, "loss": 1.6888, "step": 38142 }, { "epoch": 7.27, "learning_rate": 8.211820781696853e-07, "loss": 1.3478, "step": 38143 }, { "epoch": 7.27, "learning_rate": 8.211248808388942e-07, "loss": 1.6526, "step": 38144 }, { "epoch": 7.27, "learning_rate": 8.21067683508103e-07, "loss": 2.1201, "step": 38145 }, { "epoch": 7.27, "learning_rate": 8.210104861773117e-07, "loss": 3.2893, "step": 38146 }, { "epoch": 7.27, "learning_rate": 8.209532888465205e-07, "loss": 2.4936, "step": 38147 }, { "epoch": 7.27, "learning_rate": 8.208960915157293e-07, "loss": 1.5535, "step": 38148 }, { "epoch": 7.27, "learning_rate": 8.208388941849381e-07, "loss": 0.8643, "step": 38149 }, { "epoch": 7.27, "learning_rate": 8.207816968541468e-07, "loss": 2.7608, "step": 38150 }, { "epoch": 7.27, "learning_rate": 8.207244995233556e-07, "loss": 2.1615, "step": 38151 }, { "epoch": 7.27, "learning_rate": 8.206673021925644e-07, "loss": 0.9481, "step": 38152 }, { "epoch": 7.27, "learning_rate": 8.206101048617731e-07, "loss": 1.6952, "step": 38153 }, { "epoch": 7.27, "learning_rate": 8.205529075309819e-07, "loss": 1.8072, "step": 38154 }, { "epoch": 7.27, "learning_rate": 8.204957102001908e-07, "loss": 2.6112, "step": 38155 }, { "epoch": 7.27, "learning_rate": 8.204385128693995e-07, "loss": 2.6459, "step": 38156 }, { "epoch": 7.27, "learning_rate": 8.203813155386081e-07, "loss": 1.8384, "step": 38157 }, { "epoch": 7.27, "learning_rate": 8.20324118207817e-07, "loss": 1.4162, "step": 38158 }, { "epoch": 7.27, "learning_rate": 8.202669208770258e-07, "loss": 1.9166, "step": 38159 }, { "epoch": 7.27, "learning_rate": 8.202097235462345e-07, "loss": 2.3423, "step": 38160 }, { "epoch": 7.27, "learning_rate": 8.201525262154433e-07, "loss": 1.9292, "step": 38161 }, { "epoch": 7.27, "learning_rate": 8.200953288846521e-07, "loss": 0.7942, "step": 38162 }, { "epoch": 7.27, "learning_rate": 8.200381315538608e-07, "loss": 0.7465, "step": 38163 }, { "epoch": 7.27, "learning_rate": 8.199809342230696e-07, "loss": 1.6443, "step": 38164 }, { "epoch": 7.27, "learning_rate": 8.199237368922784e-07, "loss": 1.5047, "step": 38165 }, { "epoch": 7.27, "learning_rate": 8.198665395614872e-07, "loss": 2.1232, "step": 38166 }, { "epoch": 7.27, "learning_rate": 8.198093422306959e-07, "loss": 1.7517, "step": 38167 }, { "epoch": 7.27, "learning_rate": 8.197521448999047e-07, "loss": 2.8047, "step": 38168 }, { "epoch": 7.27, "learning_rate": 8.196949475691136e-07, "loss": 2.1351, "step": 38169 }, { "epoch": 7.27, "learning_rate": 8.196377502383221e-07, "loss": 0.9857, "step": 38170 }, { "epoch": 7.27, "learning_rate": 8.195805529075309e-07, "loss": 1.0598, "step": 38171 }, { "epoch": 7.27, "learning_rate": 8.195233555767398e-07, "loss": 1.5329, "step": 38172 }, { "epoch": 7.27, "learning_rate": 8.194661582459486e-07, "loss": 1.1763, "step": 38173 }, { "epoch": 7.27, "learning_rate": 8.194089609151573e-07, "loss": 1.3102, "step": 38174 }, { "epoch": 7.27, "learning_rate": 8.193517635843661e-07, "loss": 2.3451, "step": 38175 }, { "epoch": 7.27, "learning_rate": 8.192945662535749e-07, "loss": 1.1724, "step": 38176 }, { "epoch": 7.27, "learning_rate": 8.192373689227836e-07, "loss": 1.4056, "step": 38177 }, { "epoch": 7.27, "learning_rate": 8.191801715919924e-07, "loss": 1.2206, "step": 38178 }, { "epoch": 7.27, "learning_rate": 8.191229742612012e-07, "loss": 0.4143, "step": 38179 }, { "epoch": 7.27, "learning_rate": 8.1906577693041e-07, "loss": 1.9103, "step": 38180 }, { "epoch": 7.27, "learning_rate": 8.190085795996187e-07, "loss": 0.752, "step": 38181 }, { "epoch": 7.27, "learning_rate": 8.189513822688274e-07, "loss": 1.6598, "step": 38182 }, { "epoch": 7.27, "learning_rate": 8.188941849380363e-07, "loss": 0.9689, "step": 38183 }, { "epoch": 7.27, "learning_rate": 8.188369876072449e-07, "loss": 2.5294, "step": 38184 }, { "epoch": 7.27, "learning_rate": 8.187797902764538e-07, "loss": 0.9511, "step": 38185 }, { "epoch": 7.27, "learning_rate": 8.187225929456626e-07, "loss": 2.7219, "step": 38186 }, { "epoch": 7.27, "learning_rate": 8.186653956148712e-07, "loss": 2.2115, "step": 38187 }, { "epoch": 7.27, "learning_rate": 8.186081982840801e-07, "loss": 1.6082, "step": 38188 }, { "epoch": 7.27, "learning_rate": 8.185510009532889e-07, "loss": 2.1611, "step": 38189 }, { "epoch": 7.27, "learning_rate": 8.184938036224977e-07, "loss": 1.8083, "step": 38190 }, { "epoch": 7.27, "learning_rate": 8.184366062917064e-07, "loss": 2.5879, "step": 38191 }, { "epoch": 7.27, "learning_rate": 8.183794089609152e-07, "loss": 0.7832, "step": 38192 }, { "epoch": 7.27, "learning_rate": 8.18322211630124e-07, "loss": 1.3719, "step": 38193 }, { "epoch": 7.28, "learning_rate": 8.182650142993327e-07, "loss": 2.2406, "step": 38194 }, { "epoch": 7.28, "learning_rate": 8.182078169685414e-07, "loss": 0.8033, "step": 38195 }, { "epoch": 7.28, "learning_rate": 8.181506196377502e-07, "loss": 1.8177, "step": 38196 }, { "epoch": 7.28, "learning_rate": 8.180934223069591e-07, "loss": 1.4522, "step": 38197 }, { "epoch": 7.28, "learning_rate": 8.180362249761677e-07, "loss": 2.0059, "step": 38198 }, { "epoch": 7.28, "learning_rate": 8.179790276453766e-07, "loss": 1.0487, "step": 38199 }, { "epoch": 7.28, "learning_rate": 8.179218303145854e-07, "loss": 1.5843, "step": 38200 }, { "epoch": 7.28, "learning_rate": 8.17864632983794e-07, "loss": 1.9289, "step": 38201 }, { "epoch": 7.28, "learning_rate": 8.178074356530029e-07, "loss": 1.0635, "step": 38202 }, { "epoch": 7.28, "learning_rate": 8.177502383222117e-07, "loss": 3.3791, "step": 38203 }, { "epoch": 7.28, "learning_rate": 8.176930409914205e-07, "loss": 0.383, "step": 38204 }, { "epoch": 7.28, "learning_rate": 8.176358436606292e-07, "loss": 2.4376, "step": 38205 }, { "epoch": 7.28, "learning_rate": 8.17578646329838e-07, "loss": 0.3013, "step": 38206 }, { "epoch": 7.28, "learning_rate": 8.175214489990468e-07, "loss": 1.689, "step": 38207 }, { "epoch": 7.28, "learning_rate": 8.174642516682555e-07, "loss": 0.8931, "step": 38208 }, { "epoch": 7.28, "learning_rate": 8.174070543374642e-07, "loss": 1.8512, "step": 38209 }, { "epoch": 7.28, "learning_rate": 8.17349857006673e-07, "loss": 2.7234, "step": 38210 }, { "epoch": 7.28, "learning_rate": 8.172926596758817e-07, "loss": 1.6519, "step": 38211 }, { "epoch": 7.28, "learning_rate": 8.172354623450905e-07, "loss": 3.3204, "step": 38212 }, { "epoch": 7.28, "learning_rate": 8.171782650142994e-07, "loss": 0.6318, "step": 38213 }, { "epoch": 7.28, "learning_rate": 8.171210676835082e-07, "loss": 1.745, "step": 38214 }, { "epoch": 7.28, "learning_rate": 8.170638703527168e-07, "loss": 1.1259, "step": 38215 }, { "epoch": 7.28, "learning_rate": 8.170066730219257e-07, "loss": 0.4932, "step": 38216 }, { "epoch": 7.28, "learning_rate": 8.169494756911345e-07, "loss": 1.7944, "step": 38217 }, { "epoch": 7.28, "learning_rate": 8.168922783603432e-07, "loss": 0.9207, "step": 38218 }, { "epoch": 7.28, "learning_rate": 8.16835081029552e-07, "loss": 2.5795, "step": 38219 }, { "epoch": 7.28, "learning_rate": 8.167778836987608e-07, "loss": 1.084, "step": 38220 }, { "epoch": 7.28, "learning_rate": 8.167206863679696e-07, "loss": 2.9778, "step": 38221 }, { "epoch": 7.28, "learning_rate": 8.166634890371782e-07, "loss": 1.0173, "step": 38222 }, { "epoch": 7.28, "learning_rate": 8.16606291706387e-07, "loss": 2.233, "step": 38223 }, { "epoch": 7.28, "learning_rate": 8.165490943755958e-07, "loss": 0.8804, "step": 38224 }, { "epoch": 7.28, "learning_rate": 8.164918970448045e-07, "loss": 1.4234, "step": 38225 }, { "epoch": 7.28, "learning_rate": 8.164346997140133e-07, "loss": 2.41, "step": 38226 }, { "epoch": 7.28, "learning_rate": 8.163775023832222e-07, "loss": 2.6554, "step": 38227 }, { "epoch": 7.28, "learning_rate": 8.163203050524308e-07, "loss": 1.8583, "step": 38228 }, { "epoch": 7.28, "learning_rate": 8.162631077216397e-07, "loss": 0.6928, "step": 38229 }, { "epoch": 7.28, "learning_rate": 8.162059103908485e-07, "loss": 2.2102, "step": 38230 }, { "epoch": 7.28, "learning_rate": 8.161487130600573e-07, "loss": 1.5456, "step": 38231 }, { "epoch": 7.28, "learning_rate": 8.16091515729266e-07, "loss": 1.3984, "step": 38232 }, { "epoch": 7.28, "learning_rate": 8.160343183984748e-07, "loss": 1.5392, "step": 38233 }, { "epoch": 7.28, "learning_rate": 8.159771210676836e-07, "loss": 1.333, "step": 38234 }, { "epoch": 7.28, "learning_rate": 8.159199237368922e-07, "loss": 1.389, "step": 38235 }, { "epoch": 7.28, "learning_rate": 8.15862726406101e-07, "loss": 1.0316, "step": 38236 }, { "epoch": 7.28, "learning_rate": 8.158055290753098e-07, "loss": 1.5111, "step": 38237 }, { "epoch": 7.28, "learning_rate": 8.157483317445186e-07, "loss": 0.9655, "step": 38238 }, { "epoch": 7.28, "learning_rate": 8.156911344137273e-07, "loss": 2.2102, "step": 38239 }, { "epoch": 7.28, "learning_rate": 8.156339370829361e-07, "loss": 1.5144, "step": 38240 }, { "epoch": 7.28, "learning_rate": 8.15576739752145e-07, "loss": 1.8742, "step": 38241 }, { "epoch": 7.28, "learning_rate": 8.155195424213536e-07, "loss": 1.0741, "step": 38242 }, { "epoch": 7.28, "learning_rate": 8.154623450905625e-07, "loss": 1.7488, "step": 38243 }, { "epoch": 7.28, "learning_rate": 8.154051477597713e-07, "loss": 1.8972, "step": 38244 }, { "epoch": 7.28, "learning_rate": 8.153479504289801e-07, "loss": 1.0373, "step": 38245 }, { "epoch": 7.28, "learning_rate": 8.152907530981888e-07, "loss": 1.7498, "step": 38246 }, { "epoch": 7.29, "learning_rate": 8.152335557673976e-07, "loss": 1.3293, "step": 38247 }, { "epoch": 7.29, "learning_rate": 8.151763584366064e-07, "loss": 1.1645, "step": 38248 }, { "epoch": 7.29, "learning_rate": 8.15119161105815e-07, "loss": 1.8795, "step": 38249 }, { "epoch": 7.29, "learning_rate": 8.150619637750238e-07, "loss": 1.1067, "step": 38250 }, { "epoch": 7.29, "learning_rate": 8.150047664442326e-07, "loss": 1.5451, "step": 38251 }, { "epoch": 7.29, "learning_rate": 8.149475691134413e-07, "loss": 1.0614, "step": 38252 }, { "epoch": 7.29, "learning_rate": 8.148903717826501e-07, "loss": 2.3544, "step": 38253 }, { "epoch": 7.29, "learning_rate": 8.148331744518589e-07, "loss": 0.8579, "step": 38254 }, { "epoch": 7.29, "learning_rate": 8.147759771210678e-07, "loss": 1.6086, "step": 38255 }, { "epoch": 7.29, "learning_rate": 8.147187797902764e-07, "loss": 1.9535, "step": 38256 }, { "epoch": 7.29, "learning_rate": 8.146615824594853e-07, "loss": 1.9999, "step": 38257 }, { "epoch": 7.29, "learning_rate": 8.146043851286941e-07, "loss": 2.0505, "step": 38258 }, { "epoch": 7.29, "learning_rate": 8.145471877979027e-07, "loss": 1.9155, "step": 38259 }, { "epoch": 7.29, "learning_rate": 8.144899904671116e-07, "loss": 3.3904, "step": 38260 }, { "epoch": 7.29, "learning_rate": 8.144327931363204e-07, "loss": 2.4309, "step": 38261 }, { "epoch": 7.29, "learning_rate": 8.143755958055292e-07, "loss": 1.3884, "step": 38262 }, { "epoch": 7.29, "learning_rate": 8.143183984747378e-07, "loss": 0.8257, "step": 38263 }, { "epoch": 7.29, "learning_rate": 8.142612011439466e-07, "loss": 2.6251, "step": 38264 }, { "epoch": 7.29, "learning_rate": 8.142040038131554e-07, "loss": 2.4682, "step": 38265 }, { "epoch": 7.29, "learning_rate": 8.141468064823641e-07, "loss": 1.7216, "step": 38266 }, { "epoch": 7.29, "learning_rate": 8.140896091515729e-07, "loss": 1.2966, "step": 38267 }, { "epoch": 7.29, "learning_rate": 8.140324118207817e-07, "loss": 0.9998, "step": 38268 }, { "epoch": 7.29, "learning_rate": 8.139752144899904e-07, "loss": 2.0413, "step": 38269 }, { "epoch": 7.29, "learning_rate": 8.139180171591992e-07, "loss": 0.9459, "step": 38270 }, { "epoch": 7.29, "learning_rate": 8.138608198284081e-07, "loss": 1.2657, "step": 38271 }, { "epoch": 7.29, "learning_rate": 8.138036224976169e-07, "loss": 2.7241, "step": 38272 }, { "epoch": 7.29, "learning_rate": 8.137464251668255e-07, "loss": 1.6533, "step": 38273 }, { "epoch": 7.29, "learning_rate": 8.136892278360344e-07, "loss": 1.0471, "step": 38274 }, { "epoch": 7.29, "learning_rate": 8.136320305052432e-07, "loss": 2.5005, "step": 38275 }, { "epoch": 7.29, "learning_rate": 8.135748331744518e-07, "loss": 1.0779, "step": 38276 }, { "epoch": 7.29, "learning_rate": 8.135176358436606e-07, "loss": 1.1999, "step": 38277 }, { "epoch": 7.29, "learning_rate": 8.134604385128694e-07, "loss": 1.6835, "step": 38278 }, { "epoch": 7.29, "learning_rate": 8.134032411820782e-07, "loss": 1.7545, "step": 38279 }, { "epoch": 7.29, "learning_rate": 8.133460438512869e-07, "loss": 2.0857, "step": 38280 }, { "epoch": 7.29, "learning_rate": 8.132888465204957e-07, "loss": 1.3517, "step": 38281 }, { "epoch": 7.29, "learning_rate": 8.132316491897045e-07, "loss": 1.8685, "step": 38282 }, { "epoch": 7.29, "learning_rate": 8.131744518589132e-07, "loss": 1.4904, "step": 38283 }, { "epoch": 7.29, "learning_rate": 8.13117254528122e-07, "loss": 1.9396, "step": 38284 }, { "epoch": 7.29, "learning_rate": 8.130600571973309e-07, "loss": 1.1268, "step": 38285 }, { "epoch": 7.29, "learning_rate": 8.130028598665397e-07, "loss": 1.9124, "step": 38286 }, { "epoch": 7.29, "learning_rate": 8.129456625357484e-07, "loss": 0.6302, "step": 38287 }, { "epoch": 7.29, "learning_rate": 8.128884652049572e-07, "loss": 1.0667, "step": 38288 }, { "epoch": 7.29, "learning_rate": 8.12831267874166e-07, "loss": 0.853, "step": 38289 }, { "epoch": 7.29, "learning_rate": 8.127740705433746e-07, "loss": 2.9119, "step": 38290 }, { "epoch": 7.29, "learning_rate": 8.127168732125834e-07, "loss": 2.8141, "step": 38291 }, { "epoch": 7.29, "learning_rate": 8.126596758817922e-07, "loss": 1.9148, "step": 38292 }, { "epoch": 7.29, "learning_rate": 8.126024785510009e-07, "loss": 2.4769, "step": 38293 }, { "epoch": 7.29, "learning_rate": 8.125452812202097e-07, "loss": 2.7797, "step": 38294 }, { "epoch": 7.29, "learning_rate": 8.124880838894185e-07, "loss": 0.5484, "step": 38295 }, { "epoch": 7.29, "learning_rate": 8.124308865586274e-07, "loss": 2.1657, "step": 38296 }, { "epoch": 7.29, "learning_rate": 8.12373689227836e-07, "loss": 2.1738, "step": 38297 }, { "epoch": 7.29, "learning_rate": 8.123164918970448e-07, "loss": 1.1065, "step": 38298 }, { "epoch": 7.3, "learning_rate": 8.122592945662537e-07, "loss": 2.2752, "step": 38299 }, { "epoch": 7.3, "learning_rate": 8.122020972354623e-07, "loss": 1.4819, "step": 38300 }, { "epoch": 7.3, "learning_rate": 8.121448999046712e-07, "loss": 0.7929, "step": 38301 }, { "epoch": 7.3, "learning_rate": 8.1208770257388e-07, "loss": 1.2331, "step": 38302 }, { "epoch": 7.3, "learning_rate": 8.120305052430887e-07, "loss": 2.342, "step": 38303 }, { "epoch": 7.3, "learning_rate": 8.119733079122974e-07, "loss": 0.8689, "step": 38304 }, { "epoch": 7.3, "learning_rate": 8.119161105815062e-07, "loss": 1.4543, "step": 38305 }, { "epoch": 7.3, "learning_rate": 8.11858913250715e-07, "loss": 0.96, "step": 38306 }, { "epoch": 7.3, "learning_rate": 8.118017159199237e-07, "loss": 1.6565, "step": 38307 }, { "epoch": 7.3, "learning_rate": 8.117445185891325e-07, "loss": 1.8122, "step": 38308 }, { "epoch": 7.3, "learning_rate": 8.116873212583413e-07, "loss": 1.2797, "step": 38309 }, { "epoch": 7.3, "learning_rate": 8.116301239275502e-07, "loss": 1.1681, "step": 38310 }, { "epoch": 7.3, "learning_rate": 8.115729265967588e-07, "loss": 0.8982, "step": 38311 }, { "epoch": 7.3, "learning_rate": 8.115157292659676e-07, "loss": 1.0302, "step": 38312 }, { "epoch": 7.3, "learning_rate": 8.114585319351765e-07, "loss": 1.7407, "step": 38313 }, { "epoch": 7.3, "learning_rate": 8.114013346043851e-07, "loss": 2.6747, "step": 38314 }, { "epoch": 7.3, "learning_rate": 8.11344137273594e-07, "loss": 2.0913, "step": 38315 }, { "epoch": 7.3, "learning_rate": 8.112869399428027e-07, "loss": 1.319, "step": 38316 }, { "epoch": 7.3, "learning_rate": 8.112297426120113e-07, "loss": 1.9152, "step": 38317 }, { "epoch": 7.3, "learning_rate": 8.111725452812202e-07, "loss": 1.8958, "step": 38318 }, { "epoch": 7.3, "learning_rate": 8.11115347950429e-07, "loss": 1.6013, "step": 38319 }, { "epoch": 7.3, "learning_rate": 8.110581506196378e-07, "loss": 0.409, "step": 38320 }, { "epoch": 7.3, "learning_rate": 8.110009532888465e-07, "loss": 2.066, "step": 38321 }, { "epoch": 7.3, "learning_rate": 8.109437559580553e-07, "loss": 2.7648, "step": 38322 }, { "epoch": 7.3, "learning_rate": 8.108865586272641e-07, "loss": 2.0734, "step": 38323 }, { "epoch": 7.3, "learning_rate": 8.108293612964728e-07, "loss": 1.1994, "step": 38324 }, { "epoch": 7.3, "learning_rate": 8.107721639656816e-07, "loss": 1.5728, "step": 38325 }, { "epoch": 7.3, "learning_rate": 8.107149666348904e-07, "loss": 1.5998, "step": 38326 }, { "epoch": 7.3, "learning_rate": 8.106577693040993e-07, "loss": 1.0038, "step": 38327 }, { "epoch": 7.3, "learning_rate": 8.106005719733079e-07, "loss": 2.6855, "step": 38328 }, { "epoch": 7.3, "learning_rate": 8.105433746425168e-07, "loss": 0.5535, "step": 38329 }, { "epoch": 7.3, "learning_rate": 8.104861773117255e-07, "loss": 0.9331, "step": 38330 }, { "epoch": 7.3, "learning_rate": 8.104289799809342e-07, "loss": 0.8709, "step": 38331 }, { "epoch": 7.3, "learning_rate": 8.10371782650143e-07, "loss": 1.0416, "step": 38332 }, { "epoch": 7.3, "learning_rate": 8.103145853193518e-07, "loss": 2.0282, "step": 38333 }, { "epoch": 7.3, "learning_rate": 8.102573879885605e-07, "loss": 1.7541, "step": 38334 }, { "epoch": 7.3, "learning_rate": 8.102001906577693e-07, "loss": 0.4629, "step": 38335 }, { "epoch": 7.3, "learning_rate": 8.101429933269781e-07, "loss": 0.3626, "step": 38336 }, { "epoch": 7.3, "learning_rate": 8.100857959961869e-07, "loss": 1.9202, "step": 38337 }, { "epoch": 7.3, "learning_rate": 8.100285986653956e-07, "loss": 1.5914, "step": 38338 }, { "epoch": 7.3, "learning_rate": 8.099714013346044e-07, "loss": 2.7201, "step": 38339 }, { "epoch": 7.3, "learning_rate": 8.099142040038133e-07, "loss": 0.7092, "step": 38340 }, { "epoch": 7.3, "learning_rate": 8.098570066730219e-07, "loss": 3.89, "step": 38341 }, { "epoch": 7.3, "learning_rate": 8.097998093422306e-07, "loss": 2.2565, "step": 38342 }, { "epoch": 7.3, "learning_rate": 8.097426120114395e-07, "loss": 2.0885, "step": 38343 }, { "epoch": 7.3, "learning_rate": 8.096854146806483e-07, "loss": 1.6007, "step": 38344 }, { "epoch": 7.3, "learning_rate": 8.09628217349857e-07, "loss": 1.9651, "step": 38345 }, { "epoch": 7.3, "learning_rate": 8.095710200190658e-07, "loss": 2.4424, "step": 38346 }, { "epoch": 7.3, "learning_rate": 8.095138226882746e-07, "loss": 0.978, "step": 38347 }, { "epoch": 7.3, "learning_rate": 8.094566253574833e-07, "loss": 3.0874, "step": 38348 }, { "epoch": 7.3, "learning_rate": 8.093994280266921e-07, "loss": 1.5312, "step": 38349 }, { "epoch": 7.3, "learning_rate": 8.093422306959009e-07, "loss": 0.8809, "step": 38350 }, { "epoch": 7.3, "learning_rate": 8.092850333651097e-07, "loss": 2.2818, "step": 38351 }, { "epoch": 7.31, "learning_rate": 8.092278360343184e-07, "loss": 1.4576, "step": 38352 }, { "epoch": 7.31, "learning_rate": 8.091706387035272e-07, "loss": 1.8222, "step": 38353 }, { "epoch": 7.31, "learning_rate": 8.091134413727361e-07, "loss": 1.7605, "step": 38354 }, { "epoch": 7.31, "learning_rate": 8.090562440419446e-07, "loss": 1.9921, "step": 38355 }, { "epoch": 7.31, "learning_rate": 8.089990467111534e-07, "loss": 1.5967, "step": 38356 }, { "epoch": 7.31, "learning_rate": 8.089418493803623e-07, "loss": 2.4793, "step": 38357 }, { "epoch": 7.31, "learning_rate": 8.088846520495709e-07, "loss": 0.8681, "step": 38358 }, { "epoch": 7.31, "learning_rate": 8.088274547187798e-07, "loss": 2.8493, "step": 38359 }, { "epoch": 7.31, "learning_rate": 8.087702573879886e-07, "loss": 2.3297, "step": 38360 }, { "epoch": 7.31, "learning_rate": 8.087130600571974e-07, "loss": 1.2604, "step": 38361 }, { "epoch": 7.31, "learning_rate": 8.086558627264061e-07, "loss": 1.1659, "step": 38362 }, { "epoch": 7.31, "learning_rate": 8.085986653956149e-07, "loss": 1.205, "step": 38363 }, { "epoch": 7.31, "learning_rate": 8.085414680648237e-07, "loss": 2.1089, "step": 38364 }, { "epoch": 7.31, "learning_rate": 8.084842707340324e-07, "loss": 0.8956, "step": 38365 }, { "epoch": 7.31, "learning_rate": 8.084270734032412e-07, "loss": 1.1751, "step": 38366 }, { "epoch": 7.31, "learning_rate": 8.0836987607245e-07, "loss": 3.8533, "step": 38367 }, { "epoch": 7.31, "learning_rate": 8.083126787416589e-07, "loss": 1.8582, "step": 38368 }, { "epoch": 7.31, "learning_rate": 8.082554814108674e-07, "loss": 1.505, "step": 38369 }, { "epoch": 7.31, "learning_rate": 8.081982840800762e-07, "loss": 2.7175, "step": 38370 }, { "epoch": 7.31, "learning_rate": 8.081410867492851e-07, "loss": 2.2108, "step": 38371 }, { "epoch": 7.31, "learning_rate": 8.080838894184937e-07, "loss": 1.2662, "step": 38372 }, { "epoch": 7.31, "learning_rate": 8.080266920877026e-07, "loss": 1.6347, "step": 38373 }, { "epoch": 7.31, "learning_rate": 8.079694947569114e-07, "loss": 2.5658, "step": 38374 }, { "epoch": 7.31, "learning_rate": 8.079122974261202e-07, "loss": 1.0007, "step": 38375 }, { "epoch": 7.31, "learning_rate": 8.078551000953289e-07, "loss": 2.0879, "step": 38376 }, { "epoch": 7.31, "learning_rate": 8.077979027645377e-07, "loss": 1.18, "step": 38377 }, { "epoch": 7.31, "learning_rate": 8.077407054337465e-07, "loss": 1.9491, "step": 38378 }, { "epoch": 7.31, "learning_rate": 8.076835081029552e-07, "loss": 1.5862, "step": 38379 }, { "epoch": 7.31, "learning_rate": 8.07626310772164e-07, "loss": 1.0969, "step": 38380 }, { "epoch": 7.31, "learning_rate": 8.075691134413728e-07, "loss": 2.7064, "step": 38381 }, { "epoch": 7.31, "learning_rate": 8.075119161105814e-07, "loss": 1.7664, "step": 38382 }, { "epoch": 7.31, "learning_rate": 8.074547187797902e-07, "loss": 0.842, "step": 38383 }, { "epoch": 7.31, "learning_rate": 8.073975214489991e-07, "loss": 1.7725, "step": 38384 }, { "epoch": 7.31, "learning_rate": 8.073403241182079e-07, "loss": 1.1677, "step": 38385 }, { "epoch": 7.31, "learning_rate": 8.072831267874165e-07, "loss": 2.8932, "step": 38386 }, { "epoch": 7.31, "learning_rate": 8.072259294566254e-07, "loss": 2.7922, "step": 38387 }, { "epoch": 7.31, "learning_rate": 8.071687321258342e-07, "loss": 1.5572, "step": 38388 }, { "epoch": 7.31, "learning_rate": 8.071115347950429e-07, "loss": 1.2265, "step": 38389 }, { "epoch": 7.31, "learning_rate": 8.070543374642517e-07, "loss": 0.7994, "step": 38390 }, { "epoch": 7.31, "learning_rate": 8.069971401334605e-07, "loss": 0.8431, "step": 38391 }, { "epoch": 7.31, "learning_rate": 8.069399428026693e-07, "loss": 2.2599, "step": 38392 }, { "epoch": 7.31, "learning_rate": 8.06882745471878e-07, "loss": 1.1062, "step": 38393 }, { "epoch": 7.31, "learning_rate": 8.068255481410868e-07, "loss": 1.3361, "step": 38394 }, { "epoch": 7.31, "learning_rate": 8.067683508102956e-07, "loss": 1.9791, "step": 38395 }, { "epoch": 7.31, "learning_rate": 8.067111534795042e-07, "loss": 1.6513, "step": 38396 }, { "epoch": 7.31, "learning_rate": 8.06653956148713e-07, "loss": 2.4503, "step": 38397 }, { "epoch": 7.31, "learning_rate": 8.065967588179219e-07, "loss": 1.2916, "step": 38398 }, { "epoch": 7.31, "learning_rate": 8.065395614871305e-07, "loss": 0.7778, "step": 38399 }, { "epoch": 7.31, "learning_rate": 8.064823641563393e-07, "loss": 1.0185, "step": 38400 }, { "epoch": 7.31, "learning_rate": 8.064251668255482e-07, "loss": 1.6104, "step": 38401 }, { "epoch": 7.31, "learning_rate": 8.06367969494757e-07, "loss": 1.4183, "step": 38402 }, { "epoch": 7.31, "learning_rate": 8.063107721639657e-07, "loss": 1.3809, "step": 38403 }, { "epoch": 7.32, "learning_rate": 8.062535748331745e-07, "loss": 2.9615, "step": 38404 }, { "epoch": 7.32, "learning_rate": 8.061963775023833e-07, "loss": 2.822, "step": 38405 }, { "epoch": 7.32, "learning_rate": 8.06139180171592e-07, "loss": 1.0831, "step": 38406 }, { "epoch": 7.32, "learning_rate": 8.060819828408008e-07, "loss": 1.9756, "step": 38407 }, { "epoch": 7.32, "learning_rate": 8.060247855100096e-07, "loss": 2.0807, "step": 38408 }, { "epoch": 7.32, "learning_rate": 8.059675881792183e-07, "loss": 0.8285, "step": 38409 }, { "epoch": 7.32, "learning_rate": 8.05910390848427e-07, "loss": 1.3843, "step": 38410 }, { "epoch": 7.32, "learning_rate": 8.058531935176358e-07, "loss": 1.3326, "step": 38411 }, { "epoch": 7.32, "learning_rate": 8.057959961868447e-07, "loss": 2.3492, "step": 38412 }, { "epoch": 7.32, "learning_rate": 8.057387988560533e-07, "loss": 2.394, "step": 38413 }, { "epoch": 7.32, "learning_rate": 8.056816015252621e-07, "loss": 1.3805, "step": 38414 }, { "epoch": 7.32, "learning_rate": 8.05624404194471e-07, "loss": 1.2267, "step": 38415 }, { "epoch": 7.32, "learning_rate": 8.055672068636798e-07, "loss": 1.9217, "step": 38416 }, { "epoch": 7.32, "learning_rate": 8.055100095328885e-07, "loss": 1.1121, "step": 38417 }, { "epoch": 7.32, "learning_rate": 8.054528122020973e-07, "loss": 2.7879, "step": 38418 }, { "epoch": 7.32, "learning_rate": 8.053956148713061e-07, "loss": 2.1507, "step": 38419 }, { "epoch": 7.32, "learning_rate": 8.053384175405148e-07, "loss": 2.1458, "step": 38420 }, { "epoch": 7.32, "learning_rate": 8.052812202097236e-07, "loss": 2.1214, "step": 38421 }, { "epoch": 7.32, "learning_rate": 8.052240228789324e-07, "loss": 1.0457, "step": 38422 }, { "epoch": 7.32, "learning_rate": 8.05166825548141e-07, "loss": 1.8943, "step": 38423 }, { "epoch": 7.32, "learning_rate": 8.051096282173498e-07, "loss": 2.7735, "step": 38424 }, { "epoch": 7.32, "learning_rate": 8.050524308865586e-07, "loss": 1.5975, "step": 38425 }, { "epoch": 7.32, "learning_rate": 8.049952335557675e-07, "loss": 1.0095, "step": 38426 }, { "epoch": 7.32, "learning_rate": 8.049380362249761e-07, "loss": 1.8406, "step": 38427 }, { "epoch": 7.32, "learning_rate": 8.04880838894185e-07, "loss": 0.9671, "step": 38428 }, { "epoch": 7.32, "learning_rate": 8.048236415633938e-07, "loss": 1.6373, "step": 38429 }, { "epoch": 7.32, "learning_rate": 8.047664442326024e-07, "loss": 1.7817, "step": 38430 }, { "epoch": 7.32, "learning_rate": 8.047092469018113e-07, "loss": 1.7738, "step": 38431 }, { "epoch": 7.32, "learning_rate": 8.046520495710201e-07, "loss": 1.6109, "step": 38432 }, { "epoch": 7.32, "learning_rate": 8.045948522402289e-07, "loss": 1.2531, "step": 38433 }, { "epoch": 7.32, "learning_rate": 8.045376549094376e-07, "loss": 3.1473, "step": 38434 }, { "epoch": 7.32, "learning_rate": 8.044804575786464e-07, "loss": 1.8317, "step": 38435 }, { "epoch": 7.32, "learning_rate": 8.044232602478551e-07, "loss": 1.6671, "step": 38436 }, { "epoch": 7.32, "learning_rate": 8.043660629170638e-07, "loss": 1.4856, "step": 38437 }, { "epoch": 7.32, "learning_rate": 8.043088655862726e-07, "loss": 0.8923, "step": 38438 }, { "epoch": 7.32, "learning_rate": 8.042516682554814e-07, "loss": 1.8917, "step": 38439 }, { "epoch": 7.32, "learning_rate": 8.041944709246903e-07, "loss": 1.7003, "step": 38440 }, { "epoch": 7.32, "learning_rate": 8.041372735938989e-07, "loss": 1.7249, "step": 38441 }, { "epoch": 7.32, "learning_rate": 8.040800762631078e-07, "loss": 1.4295, "step": 38442 }, { "epoch": 7.32, "learning_rate": 8.040228789323166e-07, "loss": 1.5887, "step": 38443 }, { "epoch": 7.32, "learning_rate": 8.039656816015252e-07, "loss": 2.4401, "step": 38444 }, { "epoch": 7.32, "learning_rate": 8.039084842707341e-07, "loss": 3.0914, "step": 38445 }, { "epoch": 7.32, "learning_rate": 8.038512869399429e-07, "loss": 2.0419, "step": 38446 }, { "epoch": 7.32, "learning_rate": 8.037940896091516e-07, "loss": 2.112, "step": 38447 }, { "epoch": 7.32, "learning_rate": 8.037368922783604e-07, "loss": 0.7563, "step": 38448 }, { "epoch": 7.32, "learning_rate": 8.036796949475691e-07, "loss": 0.9116, "step": 38449 }, { "epoch": 7.32, "learning_rate": 8.036224976167779e-07, "loss": 1.9175, "step": 38450 }, { "epoch": 7.32, "learning_rate": 8.035653002859866e-07, "loss": 1.881, "step": 38451 }, { "epoch": 7.32, "learning_rate": 8.035081029551954e-07, "loss": 1.1812, "step": 38452 }, { "epoch": 7.32, "learning_rate": 8.034509056244042e-07, "loss": 1.6023, "step": 38453 }, { "epoch": 7.32, "learning_rate": 8.033937082936129e-07, "loss": 1.2904, "step": 38454 }, { "epoch": 7.32, "learning_rate": 8.033365109628217e-07, "loss": 2.143, "step": 38455 }, { "epoch": 7.32, "learning_rate": 8.032793136320306e-07, "loss": 3.3515, "step": 38456 }, { "epoch": 7.33, "learning_rate": 8.032221163012394e-07, "loss": 2.0773, "step": 38457 }, { "epoch": 7.33, "learning_rate": 8.03164918970448e-07, "loss": 1.5676, "step": 38458 }, { "epoch": 7.33, "learning_rate": 8.031077216396569e-07, "loss": 1.6635, "step": 38459 }, { "epoch": 7.33, "learning_rate": 8.030505243088657e-07, "loss": 1.8678, "step": 38460 }, { "epoch": 7.33, "learning_rate": 8.029933269780744e-07, "loss": 0.873, "step": 38461 }, { "epoch": 7.33, "learning_rate": 8.029361296472832e-07, "loss": 1.2534, "step": 38462 }, { "epoch": 7.33, "learning_rate": 8.028789323164919e-07, "loss": 1.4542, "step": 38463 }, { "epoch": 7.33, "learning_rate": 8.028217349857006e-07, "loss": 2.3972, "step": 38464 }, { "epoch": 7.33, "learning_rate": 8.027645376549094e-07, "loss": 2.7515, "step": 38465 }, { "epoch": 7.33, "learning_rate": 8.027073403241182e-07, "loss": 2.0743, "step": 38466 }, { "epoch": 7.33, "learning_rate": 8.02650142993327e-07, "loss": 2.9108, "step": 38467 }, { "epoch": 7.33, "learning_rate": 8.025929456625357e-07, "loss": 1.4574, "step": 38468 }, { "epoch": 7.33, "learning_rate": 8.025357483317445e-07, "loss": 1.1799, "step": 38469 }, { "epoch": 7.33, "learning_rate": 8.024785510009534e-07, "loss": 1.9951, "step": 38470 }, { "epoch": 7.33, "learning_rate": 8.02421353670162e-07, "loss": 2.4686, "step": 38471 }, { "epoch": 7.33, "learning_rate": 8.023641563393708e-07, "loss": 1.2962, "step": 38472 }, { "epoch": 7.33, "learning_rate": 8.023069590085797e-07, "loss": 1.4474, "step": 38473 }, { "epoch": 7.33, "learning_rate": 8.022497616777885e-07, "loss": 2.8086, "step": 38474 }, { "epoch": 7.33, "learning_rate": 8.021925643469972e-07, "loss": 1.9053, "step": 38475 }, { "epoch": 7.33, "learning_rate": 8.02135367016206e-07, "loss": 2.2589, "step": 38476 }, { "epoch": 7.33, "learning_rate": 8.020781696854147e-07, "loss": 1.6588, "step": 38477 }, { "epoch": 7.33, "learning_rate": 8.020209723546234e-07, "loss": 0.5153, "step": 38478 }, { "epoch": 7.33, "learning_rate": 8.019637750238322e-07, "loss": 1.5401, "step": 38479 }, { "epoch": 7.33, "learning_rate": 8.01906577693041e-07, "loss": 1.1539, "step": 38480 }, { "epoch": 7.33, "learning_rate": 8.018493803622498e-07, "loss": 1.0779, "step": 38481 }, { "epoch": 7.33, "learning_rate": 8.017921830314585e-07, "loss": 2.176, "step": 38482 }, { "epoch": 7.33, "learning_rate": 8.017349857006673e-07, "loss": 2.1611, "step": 38483 }, { "epoch": 7.33, "learning_rate": 8.016777883698762e-07, "loss": 0.8245, "step": 38484 }, { "epoch": 7.33, "learning_rate": 8.016205910390848e-07, "loss": 1.8693, "step": 38485 }, { "epoch": 7.33, "learning_rate": 8.015633937082937e-07, "loss": 2.0649, "step": 38486 }, { "epoch": 7.33, "learning_rate": 8.015061963775025e-07, "loss": 2.6905, "step": 38487 }, { "epoch": 7.33, "learning_rate": 8.01448999046711e-07, "loss": 2.0395, "step": 38488 }, { "epoch": 7.33, "learning_rate": 8.0139180171592e-07, "loss": 0.899, "step": 38489 }, { "epoch": 7.33, "learning_rate": 8.013346043851287e-07, "loss": 1.7551, "step": 38490 }, { "epoch": 7.33, "learning_rate": 8.012774070543375e-07, "loss": 2.7924, "step": 38491 }, { "epoch": 7.33, "learning_rate": 8.012202097235462e-07, "loss": 0.6029, "step": 38492 }, { "epoch": 7.33, "learning_rate": 8.01163012392755e-07, "loss": 2.7173, "step": 38493 }, { "epoch": 7.33, "learning_rate": 8.011058150619638e-07, "loss": 1.186, "step": 38494 }, { "epoch": 7.33, "learning_rate": 8.010486177311725e-07, "loss": 2.63, "step": 38495 }, { "epoch": 7.33, "learning_rate": 8.009914204003813e-07, "loss": 1.1561, "step": 38496 }, { "epoch": 7.33, "learning_rate": 8.009342230695901e-07, "loss": 1.3071, "step": 38497 }, { "epoch": 7.33, "learning_rate": 8.00877025738799e-07, "loss": 0.4038, "step": 38498 }, { "epoch": 7.33, "learning_rate": 8.008198284080076e-07, "loss": 2.349, "step": 38499 }, { "epoch": 7.33, "learning_rate": 8.007626310772165e-07, "loss": 0.9842, "step": 38500 }, { "epoch": 7.33, "learning_rate": 8.007054337464253e-07, "loss": 2.1054, "step": 38501 }, { "epoch": 7.33, "learning_rate": 8.006482364156338e-07, "loss": 1.5571, "step": 38502 }, { "epoch": 7.33, "learning_rate": 8.005910390848427e-07, "loss": 0.417, "step": 38503 }, { "epoch": 7.33, "learning_rate": 8.005338417540515e-07, "loss": 1.602, "step": 38504 }, { "epoch": 7.33, "learning_rate": 8.004766444232603e-07, "loss": 2.192, "step": 38505 }, { "epoch": 7.33, "learning_rate": 8.00419447092469e-07, "loss": 2.6235, "step": 38506 }, { "epoch": 7.33, "learning_rate": 8.003622497616778e-07, "loss": 1.3285, "step": 38507 }, { "epoch": 7.33, "learning_rate": 8.003050524308866e-07, "loss": 1.9402, "step": 38508 }, { "epoch": 7.34, "learning_rate": 8.002478551000953e-07, "loss": 0.8124, "step": 38509 }, { "epoch": 7.34, "learning_rate": 8.001906577693041e-07, "loss": 1.3213, "step": 38510 }, { "epoch": 7.34, "learning_rate": 8.001334604385129e-07, "loss": 0.5642, "step": 38511 }, { "epoch": 7.34, "learning_rate": 8.000762631077216e-07, "loss": 0.9625, "step": 38512 }, { "epoch": 7.34, "learning_rate": 8.000190657769304e-07, "loss": 2.0453, "step": 38513 }, { "epoch": 7.34, "learning_rate": 7.999618684461393e-07, "loss": 1.5519, "step": 38514 }, { "epoch": 7.34, "learning_rate": 7.999046711153481e-07, "loss": 2.1317, "step": 38515 }, { "epoch": 7.34, "learning_rate": 7.998474737845566e-07, "loss": 0.4733, "step": 38516 }, { "epoch": 7.34, "learning_rate": 7.997902764537655e-07, "loss": 2.0896, "step": 38517 }, { "epoch": 7.34, "learning_rate": 7.997330791229743e-07, "loss": 1.9884, "step": 38518 }, { "epoch": 7.34, "learning_rate": 7.99675881792183e-07, "loss": 1.7929, "step": 38519 }, { "epoch": 7.34, "learning_rate": 7.996186844613918e-07, "loss": 2.28, "step": 38520 }, { "epoch": 7.34, "learning_rate": 7.995614871306006e-07, "loss": 0.7134, "step": 38521 }, { "epoch": 7.34, "learning_rate": 7.995042897998094e-07, "loss": 2.6666, "step": 38522 }, { "epoch": 7.34, "learning_rate": 7.994470924690181e-07, "loss": 1.2683, "step": 38523 }, { "epoch": 7.34, "learning_rate": 7.993898951382269e-07, "loss": 2.0929, "step": 38524 }, { "epoch": 7.34, "learning_rate": 7.993326978074357e-07, "loss": 1.6422, "step": 38525 }, { "epoch": 7.34, "learning_rate": 7.992755004766444e-07, "loss": 1.521, "step": 38526 }, { "epoch": 7.34, "learning_rate": 7.992183031458532e-07, "loss": 1.1847, "step": 38527 }, { "epoch": 7.34, "learning_rate": 7.991611058150621e-07, "loss": 1.9799, "step": 38528 }, { "epoch": 7.34, "learning_rate": 7.991039084842706e-07, "loss": 2.553, "step": 38529 }, { "epoch": 7.34, "learning_rate": 7.990467111534795e-07, "loss": 0.5264, "step": 38530 }, { "epoch": 7.34, "learning_rate": 7.989895138226883e-07, "loss": 1.7858, "step": 38531 }, { "epoch": 7.34, "learning_rate": 7.989323164918971e-07, "loss": 2.1206, "step": 38532 }, { "epoch": 7.34, "learning_rate": 7.988751191611058e-07, "loss": 2.2702, "step": 38533 }, { "epoch": 7.34, "learning_rate": 7.988179218303146e-07, "loss": 1.0311, "step": 38534 }, { "epoch": 7.34, "learning_rate": 7.987607244995234e-07, "loss": 1.5023, "step": 38535 }, { "epoch": 7.34, "learning_rate": 7.987035271687321e-07, "loss": 1.4993, "step": 38536 }, { "epoch": 7.34, "learning_rate": 7.986463298379409e-07, "loss": 0.6259, "step": 38537 }, { "epoch": 7.34, "learning_rate": 7.985891325071497e-07, "loss": 2.058, "step": 38538 }, { "epoch": 7.34, "learning_rate": 7.985319351763586e-07, "loss": 1.2906, "step": 38539 }, { "epoch": 7.34, "learning_rate": 7.984747378455672e-07, "loss": 0.424, "step": 38540 }, { "epoch": 7.34, "learning_rate": 7.98417540514776e-07, "loss": 0.5521, "step": 38541 }, { "epoch": 7.34, "learning_rate": 7.983603431839849e-07, "loss": 0.9401, "step": 38542 }, { "epoch": 7.34, "learning_rate": 7.983031458531934e-07, "loss": 0.7975, "step": 38543 }, { "epoch": 7.34, "learning_rate": 7.982459485224023e-07, "loss": 1.9113, "step": 38544 }, { "epoch": 7.34, "learning_rate": 7.981887511916111e-07, "loss": 0.5951, "step": 38545 }, { "epoch": 7.34, "learning_rate": 7.981315538608199e-07, "loss": 1.6271, "step": 38546 }, { "epoch": 7.34, "learning_rate": 7.980743565300286e-07, "loss": 2.8989, "step": 38547 }, { "epoch": 7.34, "learning_rate": 7.980171591992374e-07, "loss": 3.9352, "step": 38548 }, { "epoch": 7.34, "learning_rate": 7.979599618684462e-07, "loss": 1.5959, "step": 38549 }, { "epoch": 7.34, "learning_rate": 7.979027645376549e-07, "loss": 2.2837, "step": 38550 }, { "epoch": 7.34, "learning_rate": 7.978455672068637e-07, "loss": 2.5182, "step": 38551 }, { "epoch": 7.34, "learning_rate": 7.977883698760725e-07, "loss": 2.0435, "step": 38552 }, { "epoch": 7.34, "learning_rate": 7.977311725452812e-07, "loss": 2.2017, "step": 38553 }, { "epoch": 7.34, "learning_rate": 7.9767397521449e-07, "loss": 2.3615, "step": 38554 }, { "epoch": 7.34, "learning_rate": 7.976167778836988e-07, "loss": 2.369, "step": 38555 }, { "epoch": 7.34, "learning_rate": 7.975595805529077e-07, "loss": 2.0162, "step": 38556 }, { "epoch": 7.34, "learning_rate": 7.975023832221162e-07, "loss": 1.0495, "step": 38557 }, { "epoch": 7.34, "learning_rate": 7.974451858913251e-07, "loss": 2.3892, "step": 38558 }, { "epoch": 7.34, "learning_rate": 7.973879885605339e-07, "loss": 2.8137, "step": 38559 }, { "epoch": 7.34, "learning_rate": 7.973307912297425e-07, "loss": 2.3326, "step": 38560 }, { "epoch": 7.34, "learning_rate": 7.972735938989514e-07, "loss": 1.1221, "step": 38561 }, { "epoch": 7.35, "learning_rate": 7.972163965681602e-07, "loss": 1.46, "step": 38562 }, { "epoch": 7.35, "learning_rate": 7.97159199237369e-07, "loss": 1.8, "step": 38563 }, { "epoch": 7.35, "learning_rate": 7.971020019065777e-07, "loss": 0.8572, "step": 38564 }, { "epoch": 7.35, "learning_rate": 7.970448045757865e-07, "loss": 2.6401, "step": 38565 }, { "epoch": 7.35, "learning_rate": 7.969876072449953e-07, "loss": 1.0393, "step": 38566 }, { "epoch": 7.35, "learning_rate": 7.96930409914204e-07, "loss": 2.4406, "step": 38567 }, { "epoch": 7.35, "learning_rate": 7.968732125834128e-07, "loss": 1.3856, "step": 38568 }, { "epoch": 7.35, "learning_rate": 7.968160152526215e-07, "loss": 1.8646, "step": 38569 }, { "epoch": 7.35, "learning_rate": 7.967588179218302e-07, "loss": 1.362, "step": 38570 }, { "epoch": 7.35, "learning_rate": 7.96701620591039e-07, "loss": 2.6979, "step": 38571 }, { "epoch": 7.35, "learning_rate": 7.966444232602479e-07, "loss": 1.822, "step": 38572 }, { "epoch": 7.35, "learning_rate": 7.965872259294567e-07, "loss": 1.0438, "step": 38573 }, { "epoch": 7.35, "learning_rate": 7.965300285986654e-07, "loss": 0.3953, "step": 38574 }, { "epoch": 7.35, "learning_rate": 7.964728312678742e-07, "loss": 0.4598, "step": 38575 }, { "epoch": 7.35, "learning_rate": 7.96415633937083e-07, "loss": 1.3595, "step": 38576 }, { "epoch": 7.35, "learning_rate": 7.963584366062917e-07, "loss": 1.1867, "step": 38577 }, { "epoch": 7.35, "learning_rate": 7.963012392755005e-07, "loss": 1.5327, "step": 38578 }, { "epoch": 7.35, "learning_rate": 7.962440419447093e-07, "loss": 0.5328, "step": 38579 }, { "epoch": 7.35, "learning_rate": 7.961868446139181e-07, "loss": 1.7157, "step": 38580 }, { "epoch": 7.35, "learning_rate": 7.961296472831268e-07, "loss": 1.0047, "step": 38581 }, { "epoch": 7.35, "learning_rate": 7.960724499523356e-07, "loss": 0.9222, "step": 38582 }, { "epoch": 7.35, "learning_rate": 7.960152526215445e-07, "loss": 2.0472, "step": 38583 }, { "epoch": 7.35, "learning_rate": 7.95958055290753e-07, "loss": 2.5572, "step": 38584 }, { "epoch": 7.35, "learning_rate": 7.959008579599618e-07, "loss": 1.1536, "step": 38585 }, { "epoch": 7.35, "learning_rate": 7.958436606291707e-07, "loss": 1.309, "step": 38586 }, { "epoch": 7.35, "learning_rate": 7.957864632983795e-07, "loss": 1.5849, "step": 38587 }, { "epoch": 7.35, "learning_rate": 7.957292659675882e-07, "loss": 2.1672, "step": 38588 }, { "epoch": 7.35, "learning_rate": 7.95672068636797e-07, "loss": 1.7831, "step": 38589 }, { "epoch": 7.35, "learning_rate": 7.956148713060058e-07, "loss": 2.2268, "step": 38590 }, { "epoch": 7.35, "learning_rate": 7.955576739752145e-07, "loss": 1.2645, "step": 38591 }, { "epoch": 7.35, "learning_rate": 7.955004766444233e-07, "loss": 1.3712, "step": 38592 }, { "epoch": 7.35, "learning_rate": 7.954432793136321e-07, "loss": 1.7844, "step": 38593 }, { "epoch": 7.35, "learning_rate": 7.953860819828408e-07, "loss": 0.8826, "step": 38594 }, { "epoch": 7.35, "learning_rate": 7.953288846520496e-07, "loss": 0.4576, "step": 38595 }, { "epoch": 7.35, "learning_rate": 7.952716873212583e-07, "loss": 1.5655, "step": 38596 }, { "epoch": 7.35, "learning_rate": 7.952144899904672e-07, "loss": 1.7515, "step": 38597 }, { "epoch": 7.35, "learning_rate": 7.951572926596758e-07, "loss": 1.3885, "step": 38598 }, { "epoch": 7.35, "learning_rate": 7.951000953288846e-07, "loss": 2.2685, "step": 38599 }, { "epoch": 7.35, "learning_rate": 7.950428979980935e-07, "loss": 2.6235, "step": 38600 }, { "epoch": 7.35, "learning_rate": 7.949857006673021e-07, "loss": 2.185, "step": 38601 }, { "epoch": 7.35, "learning_rate": 7.94928503336511e-07, "loss": 1.4769, "step": 38602 }, { "epoch": 7.35, "learning_rate": 7.948713060057198e-07, "loss": 1.7682, "step": 38603 }, { "epoch": 7.35, "learning_rate": 7.948141086749286e-07, "loss": 1.7743, "step": 38604 }, { "epoch": 7.35, "learning_rate": 7.947569113441373e-07, "loss": 2.7945, "step": 38605 }, { "epoch": 7.35, "learning_rate": 7.946997140133461e-07, "loss": 0.5581, "step": 38606 }, { "epoch": 7.35, "learning_rate": 7.946425166825549e-07, "loss": 1.3498, "step": 38607 }, { "epoch": 7.35, "learning_rate": 7.945853193517636e-07, "loss": 2.5672, "step": 38608 }, { "epoch": 7.35, "learning_rate": 7.945281220209723e-07, "loss": 1.5759, "step": 38609 }, { "epoch": 7.35, "learning_rate": 7.944709246901811e-07, "loss": 1.2324, "step": 38610 }, { "epoch": 7.35, "learning_rate": 7.9441372735939e-07, "loss": 2.2406, "step": 38611 }, { "epoch": 7.35, "learning_rate": 7.943565300285986e-07, "loss": 1.0853, "step": 38612 }, { "epoch": 7.35, "learning_rate": 7.942993326978074e-07, "loss": 1.4702, "step": 38613 }, { "epoch": 7.36, "learning_rate": 7.942421353670163e-07, "loss": 2.1444, "step": 38614 }, { "epoch": 7.36, "learning_rate": 7.941849380362249e-07, "loss": 1.0856, "step": 38615 }, { "epoch": 7.36, "learning_rate": 7.941277407054338e-07, "loss": 1.9062, "step": 38616 }, { "epoch": 7.36, "learning_rate": 7.940705433746426e-07, "loss": 1.5254, "step": 38617 }, { "epoch": 7.36, "learning_rate": 7.940133460438513e-07, "loss": 2.4852, "step": 38618 }, { "epoch": 7.36, "learning_rate": 7.939561487130601e-07, "loss": 1.2192, "step": 38619 }, { "epoch": 7.36, "learning_rate": 7.938989513822689e-07, "loss": 2.5789, "step": 38620 }, { "epoch": 7.36, "learning_rate": 7.938417540514777e-07, "loss": 1.9542, "step": 38621 }, { "epoch": 7.36, "learning_rate": 7.937845567206864e-07, "loss": 1.8199, "step": 38622 }, { "epoch": 7.36, "learning_rate": 7.937273593898951e-07, "loss": 2.5684, "step": 38623 }, { "epoch": 7.36, "learning_rate": 7.936701620591039e-07, "loss": 0.9279, "step": 38624 }, { "epoch": 7.36, "learning_rate": 7.936129647283126e-07, "loss": 0.8923, "step": 38625 }, { "epoch": 7.36, "learning_rate": 7.935557673975214e-07, "loss": 2.3296, "step": 38626 }, { "epoch": 7.36, "learning_rate": 7.934985700667302e-07, "loss": 1.7856, "step": 38627 }, { "epoch": 7.36, "learning_rate": 7.934413727359391e-07, "loss": 1.1584, "step": 38628 }, { "epoch": 7.36, "learning_rate": 7.933841754051477e-07, "loss": 1.6187, "step": 38629 }, { "epoch": 7.36, "learning_rate": 7.933269780743566e-07, "loss": 0.8881, "step": 38630 }, { "epoch": 7.36, "learning_rate": 7.932697807435654e-07, "loss": 1.3576, "step": 38631 }, { "epoch": 7.36, "learning_rate": 7.932125834127741e-07, "loss": 0.7062, "step": 38632 }, { "epoch": 7.36, "learning_rate": 7.931553860819829e-07, "loss": 0.9294, "step": 38633 }, { "epoch": 7.36, "learning_rate": 7.930981887511917e-07, "loss": 1.5653, "step": 38634 }, { "epoch": 7.36, "learning_rate": 7.930409914204004e-07, "loss": 1.4023, "step": 38635 }, { "epoch": 7.36, "learning_rate": 7.929837940896091e-07, "loss": 0.8935, "step": 38636 }, { "epoch": 7.36, "learning_rate": 7.929265967588179e-07, "loss": 1.6322, "step": 38637 }, { "epoch": 7.36, "learning_rate": 7.928693994280267e-07, "loss": 1.1601, "step": 38638 }, { "epoch": 7.36, "learning_rate": 7.928122020972354e-07, "loss": 1.6375, "step": 38639 }, { "epoch": 7.36, "learning_rate": 7.927550047664442e-07, "loss": 1.2873, "step": 38640 }, { "epoch": 7.36, "learning_rate": 7.926978074356531e-07, "loss": 2.6, "step": 38641 }, { "epoch": 7.36, "learning_rate": 7.926406101048617e-07, "loss": 2.2735, "step": 38642 }, { "epoch": 7.36, "learning_rate": 7.925834127740705e-07, "loss": 1.2829, "step": 38643 }, { "epoch": 7.36, "learning_rate": 7.925262154432794e-07, "loss": 1.0512, "step": 38644 }, { "epoch": 7.36, "learning_rate": 7.924690181124882e-07, "loss": 1.6489, "step": 38645 }, { "epoch": 7.36, "learning_rate": 7.924118207816969e-07, "loss": 1.6331, "step": 38646 }, { "epoch": 7.36, "learning_rate": 7.923546234509057e-07, "loss": 1.3503, "step": 38647 }, { "epoch": 7.36, "learning_rate": 7.922974261201145e-07, "loss": 2.0472, "step": 38648 }, { "epoch": 7.36, "learning_rate": 7.922402287893231e-07, "loss": 1.8138, "step": 38649 }, { "epoch": 7.36, "learning_rate": 7.921830314585319e-07, "loss": 2.0201, "step": 38650 }, { "epoch": 7.36, "learning_rate": 7.921258341277407e-07, "loss": 1.5395, "step": 38651 }, { "epoch": 7.36, "learning_rate": 7.920686367969495e-07, "loss": 0.7248, "step": 38652 }, { "epoch": 7.36, "learning_rate": 7.920114394661582e-07, "loss": 0.6266, "step": 38653 }, { "epoch": 7.36, "learning_rate": 7.91954242135367e-07, "loss": 0.8013, "step": 38654 }, { "epoch": 7.36, "learning_rate": 7.918970448045759e-07, "loss": 1.1238, "step": 38655 }, { "epoch": 7.36, "learning_rate": 7.918398474737845e-07, "loss": 0.8991, "step": 38656 }, { "epoch": 7.36, "learning_rate": 7.917826501429933e-07, "loss": 1.401, "step": 38657 }, { "epoch": 7.36, "learning_rate": 7.917254528122022e-07, "loss": 1.0978, "step": 38658 }, { "epoch": 7.36, "learning_rate": 7.916682554814108e-07, "loss": 1.1372, "step": 38659 }, { "epoch": 7.36, "learning_rate": 7.916110581506197e-07, "loss": 2.3902, "step": 38660 }, { "epoch": 7.36, "learning_rate": 7.915538608198285e-07, "loss": 1.0219, "step": 38661 }, { "epoch": 7.36, "learning_rate": 7.914966634890373e-07, "loss": 1.1248, "step": 38662 }, { "epoch": 7.36, "learning_rate": 7.914394661582459e-07, "loss": 1.5318, "step": 38663 }, { "epoch": 7.36, "learning_rate": 7.913822688274547e-07, "loss": 0.6962, "step": 38664 }, { "epoch": 7.36, "learning_rate": 7.913250714966635e-07, "loss": 1.2737, "step": 38665 }, { "epoch": 7.36, "learning_rate": 7.912678741658722e-07, "loss": 2.5176, "step": 38666 }, { "epoch": 7.37, "learning_rate": 7.91210676835081e-07, "loss": 2.9225, "step": 38667 }, { "epoch": 7.37, "learning_rate": 7.911534795042898e-07, "loss": 1.4266, "step": 38668 }, { "epoch": 7.37, "learning_rate": 7.910962821734987e-07, "loss": 1.883, "step": 38669 }, { "epoch": 7.37, "learning_rate": 7.910390848427073e-07, "loss": 1.7537, "step": 38670 }, { "epoch": 7.37, "learning_rate": 7.909818875119161e-07, "loss": 1.1781, "step": 38671 }, { "epoch": 7.37, "learning_rate": 7.90924690181125e-07, "loss": 1.8715, "step": 38672 }, { "epoch": 7.37, "learning_rate": 7.908674928503336e-07, "loss": 1.7245, "step": 38673 }, { "epoch": 7.37, "learning_rate": 7.908102955195425e-07, "loss": 1.3825, "step": 38674 }, { "epoch": 7.37, "learning_rate": 7.907530981887513e-07, "loss": 0.6188, "step": 38675 }, { "epoch": 7.37, "learning_rate": 7.9069590085796e-07, "loss": 2.0362, "step": 38676 }, { "epoch": 7.37, "learning_rate": 7.906387035271687e-07, "loss": 1.74, "step": 38677 }, { "epoch": 7.37, "learning_rate": 7.905815061963775e-07, "loss": 1.573, "step": 38678 }, { "epoch": 7.37, "learning_rate": 7.905243088655863e-07, "loss": 1.9346, "step": 38679 }, { "epoch": 7.37, "learning_rate": 7.90467111534795e-07, "loss": 2.9508, "step": 38680 }, { "epoch": 7.37, "learning_rate": 7.904099142040038e-07, "loss": 3.3376, "step": 38681 }, { "epoch": 7.37, "learning_rate": 7.903527168732126e-07, "loss": 2.754, "step": 38682 }, { "epoch": 7.37, "learning_rate": 7.902955195424213e-07, "loss": 1.4737, "step": 38683 }, { "epoch": 7.37, "learning_rate": 7.902383222116301e-07, "loss": 2.653, "step": 38684 }, { "epoch": 7.37, "learning_rate": 7.90181124880839e-07, "loss": 1.6464, "step": 38685 }, { "epoch": 7.37, "learning_rate": 7.901239275500478e-07, "loss": 2.3913, "step": 38686 }, { "epoch": 7.37, "learning_rate": 7.900667302192564e-07, "loss": 2.6148, "step": 38687 }, { "epoch": 7.37, "learning_rate": 7.900095328884653e-07, "loss": 0.9069, "step": 38688 }, { "epoch": 7.37, "learning_rate": 7.899523355576741e-07, "loss": 2.5088, "step": 38689 }, { "epoch": 7.37, "learning_rate": 7.898951382268827e-07, "loss": 3.2629, "step": 38690 }, { "epoch": 7.37, "learning_rate": 7.898379408960915e-07, "loss": 2.2074, "step": 38691 }, { "epoch": 7.37, "learning_rate": 7.897807435653003e-07, "loss": 2.5699, "step": 38692 }, { "epoch": 7.37, "learning_rate": 7.897235462345091e-07, "loss": 1.2329, "step": 38693 }, { "epoch": 7.37, "learning_rate": 7.896663489037178e-07, "loss": 4.5294, "step": 38694 }, { "epoch": 7.37, "learning_rate": 7.896091515729266e-07, "loss": 1.2223, "step": 38695 }, { "epoch": 7.37, "learning_rate": 7.895519542421354e-07, "loss": 1.777, "step": 38696 }, { "epoch": 7.37, "learning_rate": 7.894947569113441e-07, "loss": 1.3752, "step": 38697 }, { "epoch": 7.37, "learning_rate": 7.894375595805529e-07, "loss": 2.1911, "step": 38698 }, { "epoch": 7.37, "learning_rate": 7.893803622497618e-07, "loss": 2.4023, "step": 38699 }, { "epoch": 7.37, "learning_rate": 7.893231649189704e-07, "loss": 1.5673, "step": 38700 }, { "epoch": 7.37, "learning_rate": 7.892659675881792e-07, "loss": 2.3804, "step": 38701 }, { "epoch": 7.37, "learning_rate": 7.892087702573881e-07, "loss": 2.343, "step": 38702 }, { "epoch": 7.37, "learning_rate": 7.891515729265969e-07, "loss": 1.1959, "step": 38703 }, { "epoch": 7.37, "learning_rate": 7.890943755958055e-07, "loss": 1.5867, "step": 38704 }, { "epoch": 7.37, "learning_rate": 7.890371782650143e-07, "loss": 2.2605, "step": 38705 }, { "epoch": 7.37, "learning_rate": 7.889799809342231e-07, "loss": 0.9328, "step": 38706 }, { "epoch": 7.37, "learning_rate": 7.889227836034318e-07, "loss": 1.3248, "step": 38707 }, { "epoch": 7.37, "learning_rate": 7.888655862726406e-07, "loss": 0.9667, "step": 38708 }, { "epoch": 7.37, "learning_rate": 7.888083889418494e-07, "loss": 2.0404, "step": 38709 }, { "epoch": 7.37, "learning_rate": 7.887511916110582e-07, "loss": 1.0061, "step": 38710 }, { "epoch": 7.37, "learning_rate": 7.886939942802669e-07, "loss": 1.5644, "step": 38711 }, { "epoch": 7.37, "learning_rate": 7.886367969494757e-07, "loss": 2.5416, "step": 38712 }, { "epoch": 7.37, "learning_rate": 7.885795996186846e-07, "loss": 1.4436, "step": 38713 }, { "epoch": 7.37, "learning_rate": 7.885224022878932e-07, "loss": 1.8873, "step": 38714 }, { "epoch": 7.37, "learning_rate": 7.88465204957102e-07, "loss": 1.8127, "step": 38715 }, { "epoch": 7.37, "learning_rate": 7.884080076263109e-07, "loss": 2.7868, "step": 38716 }, { "epoch": 7.37, "learning_rate": 7.883508102955196e-07, "loss": 1.6708, "step": 38717 }, { "epoch": 7.37, "learning_rate": 7.882936129647283e-07, "loss": 1.4116, "step": 38718 }, { "epoch": 7.38, "learning_rate": 7.882364156339371e-07, "loss": 0.569, "step": 38719 }, { "epoch": 7.38, "learning_rate": 7.881792183031459e-07, "loss": 1.2707, "step": 38720 }, { "epoch": 7.38, "learning_rate": 7.881220209723546e-07, "loss": 1.0992, "step": 38721 }, { "epoch": 7.38, "learning_rate": 7.880648236415634e-07, "loss": 2.5521, "step": 38722 }, { "epoch": 7.38, "learning_rate": 7.880076263107722e-07, "loss": 1.6841, "step": 38723 }, { "epoch": 7.38, "learning_rate": 7.879504289799809e-07, "loss": 1.5408, "step": 38724 }, { "epoch": 7.38, "learning_rate": 7.878932316491897e-07, "loss": 0.96, "step": 38725 }, { "epoch": 7.38, "learning_rate": 7.878360343183985e-07, "loss": 1.8555, "step": 38726 }, { "epoch": 7.38, "learning_rate": 7.877788369876074e-07, "loss": 1.1046, "step": 38727 }, { "epoch": 7.38, "learning_rate": 7.87721639656816e-07, "loss": 1.2238, "step": 38728 }, { "epoch": 7.38, "learning_rate": 7.876644423260249e-07, "loss": 1.555, "step": 38729 }, { "epoch": 7.38, "learning_rate": 7.876072449952336e-07, "loss": 3.4265, "step": 38730 }, { "epoch": 7.38, "learning_rate": 7.875500476644422e-07, "loss": 0.9559, "step": 38731 }, { "epoch": 7.38, "learning_rate": 7.874928503336511e-07, "loss": 2.1353, "step": 38732 }, { "epoch": 7.38, "learning_rate": 7.874356530028599e-07, "loss": 2.08, "step": 38733 }, { "epoch": 7.38, "learning_rate": 7.873784556720687e-07, "loss": 1.7505, "step": 38734 }, { "epoch": 7.38, "learning_rate": 7.873212583412774e-07, "loss": 1.4314, "step": 38735 }, { "epoch": 7.38, "learning_rate": 7.872640610104862e-07, "loss": 1.186, "step": 38736 }, { "epoch": 7.38, "learning_rate": 7.87206863679695e-07, "loss": 3.1226, "step": 38737 }, { "epoch": 7.38, "learning_rate": 7.871496663489037e-07, "loss": 2.0792, "step": 38738 }, { "epoch": 7.38, "learning_rate": 7.870924690181125e-07, "loss": 2.3008, "step": 38739 }, { "epoch": 7.38, "learning_rate": 7.870352716873213e-07, "loss": 1.9569, "step": 38740 }, { "epoch": 7.38, "learning_rate": 7.869780743565302e-07, "loss": 1.4106, "step": 38741 }, { "epoch": 7.38, "learning_rate": 7.869208770257388e-07, "loss": 1.862, "step": 38742 }, { "epoch": 7.38, "learning_rate": 7.868636796949477e-07, "loss": 1.8137, "step": 38743 }, { "epoch": 7.38, "learning_rate": 7.868064823641564e-07, "loss": 2.0883, "step": 38744 }, { "epoch": 7.38, "learning_rate": 7.86749285033365e-07, "loss": 1.0991, "step": 38745 }, { "epoch": 7.38, "learning_rate": 7.866920877025739e-07, "loss": 1.2308, "step": 38746 }, { "epoch": 7.38, "learning_rate": 7.866348903717827e-07, "loss": 2.2208, "step": 38747 }, { "epoch": 7.38, "learning_rate": 7.865776930409914e-07, "loss": 1.8242, "step": 38748 }, { "epoch": 7.38, "learning_rate": 7.865204957102002e-07, "loss": 0.72, "step": 38749 }, { "epoch": 7.38, "learning_rate": 7.86463298379409e-07, "loss": 1.5695, "step": 38750 }, { "epoch": 7.38, "learning_rate": 7.864061010486178e-07, "loss": 1.5459, "step": 38751 }, { "epoch": 7.38, "learning_rate": 7.863489037178265e-07, "loss": 1.9476, "step": 38752 }, { "epoch": 7.38, "learning_rate": 7.862917063870353e-07, "loss": 2.46, "step": 38753 }, { "epoch": 7.38, "learning_rate": 7.862345090562441e-07, "loss": 1.8971, "step": 38754 }, { "epoch": 7.38, "learning_rate": 7.861773117254528e-07, "loss": 1.1706, "step": 38755 }, { "epoch": 7.38, "learning_rate": 7.861201143946615e-07, "loss": 2.2905, "step": 38756 }, { "epoch": 7.38, "learning_rate": 7.860629170638704e-07, "loss": 1.0179, "step": 38757 }, { "epoch": 7.38, "learning_rate": 7.860057197330792e-07, "loss": 1.1545, "step": 38758 }, { "epoch": 7.38, "learning_rate": 7.859485224022878e-07, "loss": 1.8812, "step": 38759 }, { "epoch": 7.38, "learning_rate": 7.858913250714967e-07, "loss": 2.2515, "step": 38760 }, { "epoch": 7.38, "learning_rate": 7.858341277407055e-07, "loss": 1.6354, "step": 38761 }, { "epoch": 7.38, "learning_rate": 7.857769304099142e-07, "loss": 1.5692, "step": 38762 }, { "epoch": 7.38, "learning_rate": 7.85719733079123e-07, "loss": 2.8723, "step": 38763 }, { "epoch": 7.38, "learning_rate": 7.856625357483318e-07, "loss": 1.2087, "step": 38764 }, { "epoch": 7.38, "learning_rate": 7.856053384175405e-07, "loss": 1.1159, "step": 38765 }, { "epoch": 7.38, "learning_rate": 7.855481410867493e-07, "loss": 1.1185, "step": 38766 }, { "epoch": 7.38, "learning_rate": 7.854909437559581e-07, "loss": 0.8789, "step": 38767 }, { "epoch": 7.38, "learning_rate": 7.854337464251669e-07, "loss": 1.7171, "step": 38768 }, { "epoch": 7.38, "learning_rate": 7.853765490943755e-07, "loss": 1.8897, "step": 38769 }, { "epoch": 7.38, "learning_rate": 7.853193517635843e-07, "loss": 2.6202, "step": 38770 }, { "epoch": 7.38, "learning_rate": 7.852621544327932e-07, "loss": 1.614, "step": 38771 }, { "epoch": 7.39, "learning_rate": 7.852049571020018e-07, "loss": 1.4267, "step": 38772 }, { "epoch": 7.39, "learning_rate": 7.851477597712107e-07, "loss": 0.208, "step": 38773 }, { "epoch": 7.39, "learning_rate": 7.850905624404195e-07, "loss": 2.4941, "step": 38774 }, { "epoch": 7.39, "learning_rate": 7.850333651096283e-07, "loss": 1.72, "step": 38775 }, { "epoch": 7.39, "learning_rate": 7.84976167778837e-07, "loss": 1.7393, "step": 38776 }, { "epoch": 7.39, "learning_rate": 7.849189704480458e-07, "loss": 1.1399, "step": 38777 }, { "epoch": 7.39, "learning_rate": 7.848617731172546e-07, "loss": 1.3318, "step": 38778 }, { "epoch": 7.39, "learning_rate": 7.848045757864633e-07, "loss": 0.4117, "step": 38779 }, { "epoch": 7.39, "learning_rate": 7.847473784556721e-07, "loss": 1.5893, "step": 38780 }, { "epoch": 7.39, "learning_rate": 7.846901811248809e-07, "loss": 2.4342, "step": 38781 }, { "epoch": 7.39, "learning_rate": 7.846329837940897e-07, "loss": 2.6607, "step": 38782 }, { "epoch": 7.39, "learning_rate": 7.845757864632983e-07, "loss": 0.5988, "step": 38783 }, { "epoch": 7.39, "learning_rate": 7.845185891325071e-07, "loss": 1.3068, "step": 38784 }, { "epoch": 7.39, "learning_rate": 7.84461391801716e-07, "loss": 1.4758, "step": 38785 }, { "epoch": 7.39, "learning_rate": 7.844041944709246e-07, "loss": 1.7887, "step": 38786 }, { "epoch": 7.39, "learning_rate": 7.843469971401335e-07, "loss": 1.8304, "step": 38787 }, { "epoch": 7.39, "learning_rate": 7.842897998093423e-07, "loss": 1.6989, "step": 38788 }, { "epoch": 7.39, "learning_rate": 7.842326024785509e-07, "loss": 1.8132, "step": 38789 }, { "epoch": 7.39, "learning_rate": 7.841754051477598e-07, "loss": 2.3115, "step": 38790 }, { "epoch": 7.39, "learning_rate": 7.841182078169686e-07, "loss": 1.5532, "step": 38791 }, { "epoch": 7.39, "learning_rate": 7.840610104861774e-07, "loss": 1.7186, "step": 38792 }, { "epoch": 7.39, "learning_rate": 7.840038131553861e-07, "loss": 1.4827, "step": 38793 }, { "epoch": 7.39, "learning_rate": 7.839466158245949e-07, "loss": 1.1713, "step": 38794 }, { "epoch": 7.39, "learning_rate": 7.838894184938037e-07, "loss": 2.8114, "step": 38795 }, { "epoch": 7.39, "learning_rate": 7.838322211630123e-07, "loss": 2.255, "step": 38796 }, { "epoch": 7.39, "learning_rate": 7.837750238322211e-07, "loss": 1.8701, "step": 38797 }, { "epoch": 7.39, "learning_rate": 7.837178265014299e-07, "loss": 1.7204, "step": 38798 }, { "epoch": 7.39, "learning_rate": 7.836606291706388e-07, "loss": 2.2732, "step": 38799 }, { "epoch": 7.39, "learning_rate": 7.836034318398474e-07, "loss": 0.8943, "step": 38800 }, { "epoch": 7.39, "learning_rate": 7.835462345090563e-07, "loss": 2.4131, "step": 38801 }, { "epoch": 7.39, "learning_rate": 7.834890371782651e-07, "loss": 1.1719, "step": 38802 }, { "epoch": 7.39, "learning_rate": 7.834318398474737e-07, "loss": 1.5729, "step": 38803 }, { "epoch": 7.39, "learning_rate": 7.833746425166826e-07, "loss": 2.0556, "step": 38804 }, { "epoch": 7.39, "learning_rate": 7.833174451858914e-07, "loss": 1.9728, "step": 38805 }, { "epoch": 7.39, "learning_rate": 7.832602478551002e-07, "loss": 1.2944, "step": 38806 }, { "epoch": 7.39, "learning_rate": 7.832030505243089e-07, "loss": 1.5918, "step": 38807 }, { "epoch": 7.39, "learning_rate": 7.831458531935177e-07, "loss": 1.4163, "step": 38808 }, { "epoch": 7.39, "learning_rate": 7.830886558627265e-07, "loss": 2.4036, "step": 38809 }, { "epoch": 7.39, "learning_rate": 7.830314585319351e-07, "loss": 1.9369, "step": 38810 }, { "epoch": 7.39, "learning_rate": 7.829742612011439e-07, "loss": 1.5677, "step": 38811 }, { "epoch": 7.39, "learning_rate": 7.829170638703527e-07, "loss": 1.616, "step": 38812 }, { "epoch": 7.39, "learning_rate": 7.828598665395614e-07, "loss": 2.0832, "step": 38813 }, { "epoch": 7.39, "learning_rate": 7.828026692087702e-07, "loss": 0.5905, "step": 38814 }, { "epoch": 7.39, "learning_rate": 7.827454718779791e-07, "loss": 2.5331, "step": 38815 }, { "epoch": 7.39, "learning_rate": 7.826882745471879e-07, "loss": 1.002, "step": 38816 }, { "epoch": 7.39, "learning_rate": 7.826310772163966e-07, "loss": 2.0098, "step": 38817 }, { "epoch": 7.39, "learning_rate": 7.825738798856054e-07, "loss": 2.0389, "step": 38818 }, { "epoch": 7.39, "learning_rate": 7.825166825548142e-07, "loss": 1.6365, "step": 38819 }, { "epoch": 7.39, "learning_rate": 7.824594852240229e-07, "loss": 1.2464, "step": 38820 }, { "epoch": 7.39, "learning_rate": 7.824022878932317e-07, "loss": 1.78, "step": 38821 }, { "epoch": 7.39, "learning_rate": 7.823450905624405e-07, "loss": 2.5892, "step": 38822 }, { "epoch": 7.39, "learning_rate": 7.822878932316492e-07, "loss": 1.8904, "step": 38823 }, { "epoch": 7.4, "learning_rate": 7.822306959008579e-07, "loss": 0.9635, "step": 38824 }, { "epoch": 7.4, "learning_rate": 7.821734985700667e-07, "loss": 1.1879, "step": 38825 }, { "epoch": 7.4, "learning_rate": 7.821163012392755e-07, "loss": 0.841, "step": 38826 }, { "epoch": 7.4, "learning_rate": 7.820591039084842e-07, "loss": 2.1796, "step": 38827 }, { "epoch": 7.4, "learning_rate": 7.82001906577693e-07, "loss": 1.7545, "step": 38828 }, { "epoch": 7.4, "learning_rate": 7.819447092469019e-07, "loss": 1.0991, "step": 38829 }, { "epoch": 7.4, "learning_rate": 7.818875119161105e-07, "loss": 2.3016, "step": 38830 }, { "epoch": 7.4, "learning_rate": 7.818303145853194e-07, "loss": 2.2681, "step": 38831 }, { "epoch": 7.4, "learning_rate": 7.817731172545282e-07, "loss": 1.1203, "step": 38832 }, { "epoch": 7.4, "learning_rate": 7.81715919923737e-07, "loss": 1.361, "step": 38833 }, { "epoch": 7.4, "learning_rate": 7.816587225929457e-07, "loss": 1.5028, "step": 38834 }, { "epoch": 7.4, "learning_rate": 7.816015252621545e-07, "loss": 1.3826, "step": 38835 }, { "epoch": 7.4, "learning_rate": 7.815443279313633e-07, "loss": 1.325, "step": 38836 }, { "epoch": 7.4, "learning_rate": 7.814871306005719e-07, "loss": 0.2801, "step": 38837 }, { "epoch": 7.4, "learning_rate": 7.814299332697807e-07, "loss": 1.3702, "step": 38838 }, { "epoch": 7.4, "learning_rate": 7.813727359389895e-07, "loss": 2.2951, "step": 38839 }, { "epoch": 7.4, "learning_rate": 7.813155386081984e-07, "loss": 1.5916, "step": 38840 }, { "epoch": 7.4, "learning_rate": 7.81258341277407e-07, "loss": 1.7763, "step": 38841 }, { "epoch": 7.4, "learning_rate": 7.812011439466158e-07, "loss": 0.8296, "step": 38842 }, { "epoch": 7.4, "learning_rate": 7.811439466158247e-07, "loss": 3.2477, "step": 38843 }, { "epoch": 7.4, "learning_rate": 7.810867492850333e-07, "loss": 1.6079, "step": 38844 }, { "epoch": 7.4, "learning_rate": 7.810295519542422e-07, "loss": 0.6868, "step": 38845 }, { "epoch": 7.4, "learning_rate": 7.80972354623451e-07, "loss": 0.9799, "step": 38846 }, { "epoch": 7.4, "learning_rate": 7.809151572926598e-07, "loss": 2.261, "step": 38847 }, { "epoch": 7.4, "learning_rate": 7.808579599618685e-07, "loss": 0.9728, "step": 38848 }, { "epoch": 7.4, "learning_rate": 7.808007626310773e-07, "loss": 1.8775, "step": 38849 }, { "epoch": 7.4, "learning_rate": 7.80743565300286e-07, "loss": 0.9592, "step": 38850 }, { "epoch": 7.4, "learning_rate": 7.806863679694947e-07, "loss": 0.7085, "step": 38851 }, { "epoch": 7.4, "learning_rate": 7.806291706387035e-07, "loss": 1.2456, "step": 38852 }, { "epoch": 7.4, "learning_rate": 7.805719733079123e-07, "loss": 2.6452, "step": 38853 }, { "epoch": 7.4, "learning_rate": 7.80514775977121e-07, "loss": 0.9252, "step": 38854 }, { "epoch": 7.4, "learning_rate": 7.804575786463298e-07, "loss": 1.5395, "step": 38855 }, { "epoch": 7.4, "learning_rate": 7.804003813155386e-07, "loss": 1.1348, "step": 38856 }, { "epoch": 7.4, "learning_rate": 7.803431839847475e-07, "loss": 1.6746, "step": 38857 }, { "epoch": 7.4, "learning_rate": 7.802859866539561e-07, "loss": 2.1735, "step": 38858 }, { "epoch": 7.4, "learning_rate": 7.80228789323165e-07, "loss": 1.8562, "step": 38859 }, { "epoch": 7.4, "learning_rate": 7.801715919923738e-07, "loss": 2.1402, "step": 38860 }, { "epoch": 7.4, "learning_rate": 7.801143946615824e-07, "loss": 0.952, "step": 38861 }, { "epoch": 7.4, "learning_rate": 7.800571973307913e-07, "loss": 1.554, "step": 38862 }, { "epoch": 7.4, "learning_rate": 7.8e-07, "loss": 0.6848, "step": 38863 }, { "epoch": 7.4, "learning_rate": 7.799428026692088e-07, "loss": 2.1244, "step": 38864 }, { "epoch": 7.4, "learning_rate": 7.798856053384175e-07, "loss": 0.6394, "step": 38865 }, { "epoch": 7.4, "learning_rate": 7.798284080076263e-07, "loss": 2.2064, "step": 38866 }, { "epoch": 7.4, "learning_rate": 7.797712106768351e-07, "loss": 0.2852, "step": 38867 }, { "epoch": 7.4, "learning_rate": 7.797140133460438e-07, "loss": 1.0372, "step": 38868 }, { "epoch": 7.4, "learning_rate": 7.796568160152526e-07, "loss": 1.1302, "step": 38869 }, { "epoch": 7.4, "learning_rate": 7.795996186844614e-07, "loss": 2.9579, "step": 38870 }, { "epoch": 7.4, "learning_rate": 7.795424213536703e-07, "loss": 1.784, "step": 38871 }, { "epoch": 7.4, "learning_rate": 7.794852240228789e-07, "loss": 0.9742, "step": 38872 }, { "epoch": 7.4, "learning_rate": 7.794280266920878e-07, "loss": 1.1358, "step": 38873 }, { "epoch": 7.4, "learning_rate": 7.793708293612966e-07, "loss": 1.5493, "step": 38874 }, { "epoch": 7.4, "learning_rate": 7.793136320305053e-07, "loss": 0.8413, "step": 38875 }, { "epoch": 7.4, "learning_rate": 7.79256434699714e-07, "loss": 1.2161, "step": 38876 }, { "epoch": 7.41, "learning_rate": 7.791992373689228e-07, "loss": 1.5154, "step": 38877 }, { "epoch": 7.41, "learning_rate": 7.791420400381315e-07, "loss": 1.8279, "step": 38878 }, { "epoch": 7.41, "learning_rate": 7.790848427073403e-07, "loss": 0.9271, "step": 38879 }, { "epoch": 7.41, "learning_rate": 7.790276453765491e-07, "loss": 2.5677, "step": 38880 }, { "epoch": 7.41, "learning_rate": 7.789704480457579e-07, "loss": 1.4392, "step": 38881 }, { "epoch": 7.41, "learning_rate": 7.789132507149666e-07, "loss": 2.1394, "step": 38882 }, { "epoch": 7.41, "learning_rate": 7.788560533841754e-07, "loss": 0.9814, "step": 38883 }, { "epoch": 7.41, "learning_rate": 7.787988560533843e-07, "loss": 2.2969, "step": 38884 }, { "epoch": 7.41, "learning_rate": 7.787416587225929e-07, "loss": 1.2384, "step": 38885 }, { "epoch": 7.41, "learning_rate": 7.786844613918017e-07, "loss": 2.0535, "step": 38886 }, { "epoch": 7.41, "learning_rate": 7.786272640610106e-07, "loss": 1.9247, "step": 38887 }, { "epoch": 7.41, "learning_rate": 7.785700667302194e-07, "loss": 1.3474, "step": 38888 }, { "epoch": 7.41, "learning_rate": 7.785128693994281e-07, "loss": 1.194, "step": 38889 }, { "epoch": 7.41, "learning_rate": 7.784556720686368e-07, "loss": 1.0665, "step": 38890 }, { "epoch": 7.41, "learning_rate": 7.783984747378456e-07, "loss": 0.9232, "step": 38891 }, { "epoch": 7.41, "learning_rate": 7.783412774070543e-07, "loss": 1.0896, "step": 38892 }, { "epoch": 7.41, "learning_rate": 7.782840800762631e-07, "loss": 1.9935, "step": 38893 }, { "epoch": 7.41, "learning_rate": 7.782268827454719e-07, "loss": 2.3148, "step": 38894 }, { "epoch": 7.41, "learning_rate": 7.781696854146806e-07, "loss": 1.2326, "step": 38895 }, { "epoch": 7.41, "learning_rate": 7.781124880838894e-07, "loss": 2.0146, "step": 38896 }, { "epoch": 7.41, "learning_rate": 7.780552907530982e-07, "loss": 2.2754, "step": 38897 }, { "epoch": 7.41, "learning_rate": 7.779980934223071e-07, "loss": 1.0138, "step": 38898 }, { "epoch": 7.41, "learning_rate": 7.779408960915157e-07, "loss": 2.6898, "step": 38899 }, { "epoch": 7.41, "learning_rate": 7.778836987607245e-07, "loss": 1.2602, "step": 38900 }, { "epoch": 7.41, "learning_rate": 7.778265014299334e-07, "loss": 1.3233, "step": 38901 }, { "epoch": 7.41, "learning_rate": 7.77769304099142e-07, "loss": 1.0415, "step": 38902 }, { "epoch": 7.41, "learning_rate": 7.777121067683508e-07, "loss": 0.9923, "step": 38903 }, { "epoch": 7.41, "learning_rate": 7.776549094375596e-07, "loss": 1.1571, "step": 38904 }, { "epoch": 7.41, "learning_rate": 7.775977121067684e-07, "loss": 1.4828, "step": 38905 }, { "epoch": 7.41, "learning_rate": 7.775405147759771e-07, "loss": 2.7259, "step": 38906 }, { "epoch": 7.41, "learning_rate": 7.774833174451859e-07, "loss": 0.7233, "step": 38907 }, { "epoch": 7.41, "learning_rate": 7.774261201143947e-07, "loss": 1.6406, "step": 38908 }, { "epoch": 7.41, "learning_rate": 7.773689227836034e-07, "loss": 0.554, "step": 38909 }, { "epoch": 7.41, "learning_rate": 7.773117254528122e-07, "loss": 0.7869, "step": 38910 }, { "epoch": 7.41, "learning_rate": 7.77254528122021e-07, "loss": 0.7757, "step": 38911 }, { "epoch": 7.41, "learning_rate": 7.771973307912299e-07, "loss": 0.9252, "step": 38912 }, { "epoch": 7.41, "learning_rate": 7.771401334604385e-07, "loss": 1.0482, "step": 38913 }, { "epoch": 7.41, "learning_rate": 7.770829361296473e-07, "loss": 2.4402, "step": 38914 }, { "epoch": 7.41, "learning_rate": 7.770257387988562e-07, "loss": 2.0488, "step": 38915 }, { "epoch": 7.41, "learning_rate": 7.769685414680647e-07, "loss": 0.7626, "step": 38916 }, { "epoch": 7.41, "learning_rate": 7.769113441372736e-07, "loss": 2.4754, "step": 38917 }, { "epoch": 7.41, "learning_rate": 7.768541468064824e-07, "loss": 1.8644, "step": 38918 }, { "epoch": 7.41, "learning_rate": 7.767969494756911e-07, "loss": 3.5005, "step": 38919 }, { "epoch": 7.41, "learning_rate": 7.767397521448999e-07, "loss": 3.0946, "step": 38920 }, { "epoch": 7.41, "learning_rate": 7.766825548141087e-07, "loss": 1.2479, "step": 38921 }, { "epoch": 7.41, "learning_rate": 7.766253574833175e-07, "loss": 1.3269, "step": 38922 }, { "epoch": 7.41, "learning_rate": 7.765681601525262e-07, "loss": 1.5627, "step": 38923 }, { "epoch": 7.41, "learning_rate": 7.76510962821735e-07, "loss": 1.7423, "step": 38924 }, { "epoch": 7.41, "learning_rate": 7.764537654909438e-07, "loss": 1.1149, "step": 38925 }, { "epoch": 7.41, "learning_rate": 7.763965681601525e-07, "loss": 1.0865, "step": 38926 }, { "epoch": 7.41, "learning_rate": 7.763393708293613e-07, "loss": 1.1673, "step": 38927 }, { "epoch": 7.41, "learning_rate": 7.762821734985702e-07, "loss": 1.1931, "step": 38928 }, { "epoch": 7.42, "learning_rate": 7.76224976167779e-07, "loss": 2.0594, "step": 38929 }, { "epoch": 7.42, "learning_rate": 7.761677788369875e-07, "loss": 2.8212, "step": 38930 }, { "epoch": 7.42, "learning_rate": 7.761105815061964e-07, "loss": 1.8724, "step": 38931 }, { "epoch": 7.42, "learning_rate": 7.760533841754052e-07, "loss": 1.0349, "step": 38932 }, { "epoch": 7.42, "learning_rate": 7.759961868446139e-07, "loss": 0.9366, "step": 38933 }, { "epoch": 7.42, "learning_rate": 7.759389895138227e-07, "loss": 1.5882, "step": 38934 }, { "epoch": 7.42, "learning_rate": 7.758817921830315e-07, "loss": 3.6167, "step": 38935 }, { "epoch": 7.42, "learning_rate": 7.758245948522402e-07, "loss": 1.4098, "step": 38936 }, { "epoch": 7.42, "learning_rate": 7.75767397521449e-07, "loss": 2.3221, "step": 38937 }, { "epoch": 7.42, "learning_rate": 7.757102001906578e-07, "loss": 2.1945, "step": 38938 }, { "epoch": 7.42, "learning_rate": 7.756530028598666e-07, "loss": 0.9303, "step": 38939 }, { "epoch": 7.42, "learning_rate": 7.755958055290753e-07, "loss": 1.293, "step": 38940 }, { "epoch": 7.42, "learning_rate": 7.755386081982841e-07, "loss": 0.4207, "step": 38941 }, { "epoch": 7.42, "learning_rate": 7.75481410867493e-07, "loss": 1.6748, "step": 38942 }, { "epoch": 7.42, "learning_rate": 7.754242135367015e-07, "loss": 1.989, "step": 38943 }, { "epoch": 7.42, "learning_rate": 7.753670162059103e-07, "loss": 2.4502, "step": 38944 }, { "epoch": 7.42, "learning_rate": 7.753098188751192e-07, "loss": 1.0848, "step": 38945 }, { "epoch": 7.42, "learning_rate": 7.75252621544328e-07, "loss": 1.0216, "step": 38946 }, { "epoch": 7.42, "learning_rate": 7.751954242135367e-07, "loss": 2.5819, "step": 38947 }, { "epoch": 7.42, "learning_rate": 7.751382268827455e-07, "loss": 1.1015, "step": 38948 }, { "epoch": 7.42, "learning_rate": 7.750810295519543e-07, "loss": 1.0443, "step": 38949 }, { "epoch": 7.42, "learning_rate": 7.75023832221163e-07, "loss": 1.7869, "step": 38950 }, { "epoch": 7.42, "learning_rate": 7.749666348903718e-07, "loss": 1.568, "step": 38951 }, { "epoch": 7.42, "learning_rate": 7.749094375595806e-07, "loss": 1.928, "step": 38952 }, { "epoch": 7.42, "learning_rate": 7.748522402287894e-07, "loss": 1.8664, "step": 38953 }, { "epoch": 7.42, "learning_rate": 7.747950428979981e-07, "loss": 1.3328, "step": 38954 }, { "epoch": 7.42, "learning_rate": 7.747378455672069e-07, "loss": 1.588, "step": 38955 }, { "epoch": 7.42, "learning_rate": 7.746806482364158e-07, "loss": 2.2901, "step": 38956 }, { "epoch": 7.42, "learning_rate": 7.746234509056243e-07, "loss": 1.2049, "step": 38957 }, { "epoch": 7.42, "learning_rate": 7.745662535748331e-07, "loss": 1.683, "step": 38958 }, { "epoch": 7.42, "learning_rate": 7.74509056244042e-07, "loss": 2.1575, "step": 38959 }, { "epoch": 7.42, "learning_rate": 7.744518589132506e-07, "loss": 0.8674, "step": 38960 }, { "epoch": 7.42, "learning_rate": 7.743946615824595e-07, "loss": 1.1259, "step": 38961 }, { "epoch": 7.42, "learning_rate": 7.743374642516683e-07, "loss": 1.8949, "step": 38962 }, { "epoch": 7.42, "learning_rate": 7.742802669208771e-07, "loss": 1.9844, "step": 38963 }, { "epoch": 7.42, "learning_rate": 7.742230695900858e-07, "loss": 1.8066, "step": 38964 }, { "epoch": 7.42, "learning_rate": 7.741658722592946e-07, "loss": 0.8693, "step": 38965 }, { "epoch": 7.42, "learning_rate": 7.741086749285034e-07, "loss": 2.376, "step": 38966 }, { "epoch": 7.42, "learning_rate": 7.740514775977121e-07, "loss": 2.0513, "step": 38967 }, { "epoch": 7.42, "learning_rate": 7.739942802669209e-07, "loss": 1.273, "step": 38968 }, { "epoch": 7.42, "learning_rate": 7.739370829361297e-07, "loss": 2.3387, "step": 38969 }, { "epoch": 7.42, "learning_rate": 7.738798856053386e-07, "loss": 1.0996, "step": 38970 }, { "epoch": 7.42, "learning_rate": 7.738226882745471e-07, "loss": 2.772, "step": 38971 }, { "epoch": 7.42, "learning_rate": 7.73765490943756e-07, "loss": 2.2584, "step": 38972 }, { "epoch": 7.42, "learning_rate": 7.737082936129648e-07, "loss": 1.0785, "step": 38973 }, { "epoch": 7.42, "learning_rate": 7.736510962821734e-07, "loss": 1.2467, "step": 38974 }, { "epoch": 7.42, "learning_rate": 7.735938989513823e-07, "loss": 1.5138, "step": 38975 }, { "epoch": 7.42, "learning_rate": 7.735367016205911e-07, "loss": 1.8748, "step": 38976 }, { "epoch": 7.42, "learning_rate": 7.734795042897999e-07, "loss": 2.0042, "step": 38977 }, { "epoch": 7.42, "learning_rate": 7.734223069590086e-07, "loss": 1.1281, "step": 38978 }, { "epoch": 7.42, "learning_rate": 7.733651096282174e-07, "loss": 1.9347, "step": 38979 }, { "epoch": 7.42, "learning_rate": 7.733079122974262e-07, "loss": 2.0547, "step": 38980 }, { "epoch": 7.42, "learning_rate": 7.732507149666349e-07, "loss": 2.0941, "step": 38981 }, { "epoch": 7.43, "learning_rate": 7.731935176358437e-07, "loss": 1.4122, "step": 38982 }, { "epoch": 7.43, "learning_rate": 7.731363203050524e-07, "loss": 2.5835, "step": 38983 }, { "epoch": 7.43, "learning_rate": 7.730791229742611e-07, "loss": 1.5337, "step": 38984 }, { "epoch": 7.43, "learning_rate": 7.730219256434699e-07, "loss": 1.5277, "step": 38985 }, { "epoch": 7.43, "learning_rate": 7.729647283126788e-07, "loss": 1.2714, "step": 38986 }, { "epoch": 7.43, "learning_rate": 7.729075309818876e-07, "loss": 0.8428, "step": 38987 }, { "epoch": 7.43, "learning_rate": 7.728503336510962e-07, "loss": 1.6615, "step": 38988 }, { "epoch": 7.43, "learning_rate": 7.727931363203051e-07, "loss": 1.3604, "step": 38989 }, { "epoch": 7.43, "learning_rate": 7.727359389895139e-07, "loss": 1.4331, "step": 38990 }, { "epoch": 7.43, "learning_rate": 7.726787416587226e-07, "loss": 1.7812, "step": 38991 }, { "epoch": 7.43, "learning_rate": 7.726215443279314e-07, "loss": 1.0653, "step": 38992 }, { "epoch": 7.43, "learning_rate": 7.725643469971402e-07, "loss": 2.3258, "step": 38993 }, { "epoch": 7.43, "learning_rate": 7.72507149666349e-07, "loss": 1.7085, "step": 38994 }, { "epoch": 7.43, "learning_rate": 7.724499523355577e-07, "loss": 1.2482, "step": 38995 }, { "epoch": 7.43, "learning_rate": 7.723927550047665e-07, "loss": 2.1044, "step": 38996 }, { "epoch": 7.43, "learning_rate": 7.723355576739752e-07, "loss": 1.5822, "step": 38997 }, { "epoch": 7.43, "learning_rate": 7.722783603431839e-07, "loss": 1.3411, "step": 38998 }, { "epoch": 7.43, "learning_rate": 7.722211630123927e-07, "loss": 1.8415, "step": 38999 }, { "epoch": 7.43, "learning_rate": 7.721639656816016e-07, "loss": 0.949, "step": 39000 }, { "epoch": 7.43, "learning_rate": 7.721067683508102e-07, "loss": 1.1351, "step": 39001 }, { "epoch": 7.43, "learning_rate": 7.72049571020019e-07, "loss": 1.2275, "step": 39002 }, { "epoch": 7.43, "learning_rate": 7.719923736892279e-07, "loss": 0.5211, "step": 39003 }, { "epoch": 7.43, "learning_rate": 7.719351763584367e-07, "loss": 1.609, "step": 39004 }, { "epoch": 7.43, "learning_rate": 7.718779790276454e-07, "loss": 1.9311, "step": 39005 }, { "epoch": 7.43, "learning_rate": 7.718207816968542e-07, "loss": 1.8547, "step": 39006 }, { "epoch": 7.43, "learning_rate": 7.71763584366063e-07, "loss": 2.6031, "step": 39007 }, { "epoch": 7.43, "learning_rate": 7.717063870352717e-07, "loss": 1.6794, "step": 39008 }, { "epoch": 7.43, "learning_rate": 7.716491897044805e-07, "loss": 1.426, "step": 39009 }, { "epoch": 7.43, "learning_rate": 7.715919923736892e-07, "loss": 3.0163, "step": 39010 }, { "epoch": 7.43, "learning_rate": 7.71534795042898e-07, "loss": 1.2347, "step": 39011 }, { "epoch": 7.43, "learning_rate": 7.714775977121067e-07, "loss": 1.2588, "step": 39012 }, { "epoch": 7.43, "learning_rate": 7.714204003813155e-07, "loss": 1.8342, "step": 39013 }, { "epoch": 7.43, "learning_rate": 7.713632030505244e-07, "loss": 1.3491, "step": 39014 }, { "epoch": 7.43, "learning_rate": 7.71306005719733e-07, "loss": 2.407, "step": 39015 }, { "epoch": 7.43, "learning_rate": 7.712488083889418e-07, "loss": 0.4292, "step": 39016 }, { "epoch": 7.43, "learning_rate": 7.711916110581507e-07, "loss": 1.75, "step": 39017 }, { "epoch": 7.43, "learning_rate": 7.711344137273595e-07, "loss": 1.2956, "step": 39018 }, { "epoch": 7.43, "learning_rate": 7.710772163965682e-07, "loss": 2.0817, "step": 39019 }, { "epoch": 7.43, "learning_rate": 7.71020019065777e-07, "loss": 1.9339, "step": 39020 }, { "epoch": 7.43, "learning_rate": 7.709628217349858e-07, "loss": 1.0519, "step": 39021 }, { "epoch": 7.43, "learning_rate": 7.709056244041945e-07, "loss": 1.0682, "step": 39022 }, { "epoch": 7.43, "learning_rate": 7.708484270734032e-07, "loss": 1.0405, "step": 39023 }, { "epoch": 7.43, "learning_rate": 7.70791229742612e-07, "loss": 1.4002, "step": 39024 }, { "epoch": 7.43, "learning_rate": 7.707340324118207e-07, "loss": 2.5715, "step": 39025 }, { "epoch": 7.43, "learning_rate": 7.706768350810295e-07, "loss": 1.5466, "step": 39026 }, { "epoch": 7.43, "learning_rate": 7.706196377502383e-07, "loss": 1.8117, "step": 39027 }, { "epoch": 7.43, "learning_rate": 7.705624404194472e-07, "loss": 1.7285, "step": 39028 }, { "epoch": 7.43, "learning_rate": 7.705052430886558e-07, "loss": 1.1967, "step": 39029 }, { "epoch": 7.43, "learning_rate": 7.704480457578647e-07, "loss": 1.4769, "step": 39030 }, { "epoch": 7.43, "learning_rate": 7.703908484270735e-07, "loss": 0.9854, "step": 39031 }, { "epoch": 7.43, "learning_rate": 7.703336510962821e-07, "loss": 1.2841, "step": 39032 }, { "epoch": 7.43, "learning_rate": 7.70276453765491e-07, "loss": 2.1638, "step": 39033 }, { "epoch": 7.44, "learning_rate": 7.702192564346998e-07, "loss": 1.9912, "step": 39034 }, { "epoch": 7.44, "learning_rate": 7.701620591039086e-07, "loss": 1.681, "step": 39035 }, { "epoch": 7.44, "learning_rate": 7.701048617731173e-07, "loss": 2.7987, "step": 39036 }, { "epoch": 7.44, "learning_rate": 7.70047664442326e-07, "loss": 0.5487, "step": 39037 }, { "epoch": 7.44, "learning_rate": 7.699904671115348e-07, "loss": 1.2779, "step": 39038 }, { "epoch": 7.44, "learning_rate": 7.699332697807435e-07, "loss": 1.3925, "step": 39039 }, { "epoch": 7.44, "learning_rate": 7.698760724499523e-07, "loss": 0.585, "step": 39040 }, { "epoch": 7.44, "learning_rate": 7.698188751191611e-07, "loss": 1.8818, "step": 39041 }, { "epoch": 7.44, "learning_rate": 7.6976167778837e-07, "loss": 0.5887, "step": 39042 }, { "epoch": 7.44, "learning_rate": 7.697044804575786e-07, "loss": 1.3311, "step": 39043 }, { "epoch": 7.44, "learning_rate": 7.696472831267875e-07, "loss": 0.8909, "step": 39044 }, { "epoch": 7.44, "learning_rate": 7.695900857959963e-07, "loss": 1.2059, "step": 39045 }, { "epoch": 7.44, "learning_rate": 7.695328884652049e-07, "loss": 3.3514, "step": 39046 }, { "epoch": 7.44, "learning_rate": 7.694756911344138e-07, "loss": 1.2676, "step": 39047 }, { "epoch": 7.44, "learning_rate": 7.694184938036226e-07, "loss": 2.2163, "step": 39048 }, { "epoch": 7.44, "learning_rate": 7.693612964728313e-07, "loss": 1.826, "step": 39049 }, { "epoch": 7.44, "learning_rate": 7.6930409914204e-07, "loss": 2.6062, "step": 39050 }, { "epoch": 7.44, "learning_rate": 7.692469018112488e-07, "loss": 2.0379, "step": 39051 }, { "epoch": 7.44, "learning_rate": 7.691897044804576e-07, "loss": 1.2247, "step": 39052 }, { "epoch": 7.44, "learning_rate": 7.691325071496663e-07, "loss": 2.2506, "step": 39053 }, { "epoch": 7.44, "learning_rate": 7.690753098188751e-07, "loss": 2.4319, "step": 39054 }, { "epoch": 7.44, "learning_rate": 7.690181124880839e-07, "loss": 1.0656, "step": 39055 }, { "epoch": 7.44, "learning_rate": 7.689609151572926e-07, "loss": 1.4016, "step": 39056 }, { "epoch": 7.44, "learning_rate": 7.689037178265014e-07, "loss": 0.7919, "step": 39057 }, { "epoch": 7.44, "learning_rate": 7.688465204957103e-07, "loss": 0.9118, "step": 39058 }, { "epoch": 7.44, "learning_rate": 7.687893231649191e-07, "loss": 2.2216, "step": 39059 }, { "epoch": 7.44, "learning_rate": 7.687321258341277e-07, "loss": 1.6555, "step": 39060 }, { "epoch": 7.44, "learning_rate": 7.686749285033366e-07, "loss": 1.2467, "step": 39061 }, { "epoch": 7.44, "learning_rate": 7.686177311725454e-07, "loss": 1.2212, "step": 39062 }, { "epoch": 7.44, "learning_rate": 7.68560533841754e-07, "loss": 1.3827, "step": 39063 }, { "epoch": 7.44, "learning_rate": 7.685033365109628e-07, "loss": 1.6423, "step": 39064 }, { "epoch": 7.44, "learning_rate": 7.684461391801716e-07, "loss": 1.6059, "step": 39065 }, { "epoch": 7.44, "learning_rate": 7.683889418493803e-07, "loss": 1.2792, "step": 39066 }, { "epoch": 7.44, "learning_rate": 7.683317445185891e-07, "loss": 1.4128, "step": 39067 }, { "epoch": 7.44, "learning_rate": 7.682745471877979e-07, "loss": 0.9469, "step": 39068 }, { "epoch": 7.44, "learning_rate": 7.682173498570067e-07, "loss": 0.9032, "step": 39069 }, { "epoch": 7.44, "learning_rate": 7.681601525262154e-07, "loss": 1.1335, "step": 39070 }, { "epoch": 7.44, "learning_rate": 7.681029551954242e-07, "loss": 0.9867, "step": 39071 }, { "epoch": 7.44, "learning_rate": 7.680457578646331e-07, "loss": 1.8025, "step": 39072 }, { "epoch": 7.44, "learning_rate": 7.679885605338417e-07, "loss": 1.3391, "step": 39073 }, { "epoch": 7.44, "learning_rate": 7.679313632030506e-07, "loss": 0.6777, "step": 39074 }, { "epoch": 7.44, "learning_rate": 7.678741658722594e-07, "loss": 0.8256, "step": 39075 }, { "epoch": 7.44, "learning_rate": 7.678169685414682e-07, "loss": 1.4975, "step": 39076 }, { "epoch": 7.44, "learning_rate": 7.677597712106768e-07, "loss": 2.781, "step": 39077 }, { "epoch": 7.44, "learning_rate": 7.677025738798856e-07, "loss": 0.759, "step": 39078 }, { "epoch": 7.44, "learning_rate": 7.676453765490944e-07, "loss": 1.496, "step": 39079 }, { "epoch": 7.44, "learning_rate": 7.675881792183031e-07, "loss": 1.6784, "step": 39080 }, { "epoch": 7.44, "learning_rate": 7.675309818875119e-07, "loss": 0.5042, "step": 39081 }, { "epoch": 7.44, "learning_rate": 7.674737845567207e-07, "loss": 1.496, "step": 39082 }, { "epoch": 7.44, "learning_rate": 7.674165872259296e-07, "loss": 1.8202, "step": 39083 }, { "epoch": 7.44, "learning_rate": 7.673593898951382e-07, "loss": 2.5484, "step": 39084 }, { "epoch": 7.44, "learning_rate": 7.67302192564347e-07, "loss": 1.554, "step": 39085 }, { "epoch": 7.44, "learning_rate": 7.672449952335559e-07, "loss": 1.6941, "step": 39086 }, { "epoch": 7.45, "learning_rate": 7.671877979027645e-07, "loss": 2.4224, "step": 39087 }, { "epoch": 7.45, "learning_rate": 7.671306005719734e-07, "loss": 0.2797, "step": 39088 }, { "epoch": 7.45, "learning_rate": 7.670734032411822e-07, "loss": 1.5054, "step": 39089 }, { "epoch": 7.45, "learning_rate": 7.670162059103907e-07, "loss": 1.3752, "step": 39090 }, { "epoch": 7.45, "learning_rate": 7.669590085795996e-07, "loss": 0.5752, "step": 39091 }, { "epoch": 7.45, "learning_rate": 7.669018112488084e-07, "loss": 2.5679, "step": 39092 }, { "epoch": 7.45, "learning_rate": 7.668446139180172e-07, "loss": 1.4552, "step": 39093 }, { "epoch": 7.45, "learning_rate": 7.667874165872259e-07, "loss": 1.5992, "step": 39094 }, { "epoch": 7.45, "learning_rate": 7.667302192564347e-07, "loss": 1.9812, "step": 39095 }, { "epoch": 7.45, "learning_rate": 7.666730219256435e-07, "loss": 2.0736, "step": 39096 }, { "epoch": 7.45, "learning_rate": 7.666158245948522e-07, "loss": 1.6972, "step": 39097 }, { "epoch": 7.45, "learning_rate": 7.66558627264061e-07, "loss": 0.9457, "step": 39098 }, { "epoch": 7.45, "learning_rate": 7.665014299332698e-07, "loss": 2.6277, "step": 39099 }, { "epoch": 7.45, "learning_rate": 7.664442326024787e-07, "loss": 2.298, "step": 39100 }, { "epoch": 7.45, "learning_rate": 7.663870352716873e-07, "loss": 0.9803, "step": 39101 }, { "epoch": 7.45, "learning_rate": 7.663298379408962e-07, "loss": 1.7913, "step": 39102 }, { "epoch": 7.45, "learning_rate": 7.66272640610105e-07, "loss": 3.0805, "step": 39103 }, { "epoch": 7.45, "learning_rate": 7.662154432793135e-07, "loss": 0.8587, "step": 39104 }, { "epoch": 7.45, "learning_rate": 7.661582459485224e-07, "loss": 0.7172, "step": 39105 }, { "epoch": 7.45, "learning_rate": 7.661010486177312e-07, "loss": 0.733, "step": 39106 }, { "epoch": 7.45, "learning_rate": 7.6604385128694e-07, "loss": 1.3559, "step": 39107 }, { "epoch": 7.45, "learning_rate": 7.659866539561487e-07, "loss": 1.2977, "step": 39108 }, { "epoch": 7.45, "learning_rate": 7.659294566253575e-07, "loss": 2.1146, "step": 39109 }, { "epoch": 7.45, "learning_rate": 7.658722592945663e-07, "loss": 1.5906, "step": 39110 }, { "epoch": 7.45, "learning_rate": 7.65815061963775e-07, "loss": 0.4466, "step": 39111 }, { "epoch": 7.45, "learning_rate": 7.657578646329838e-07, "loss": 0.6781, "step": 39112 }, { "epoch": 7.45, "learning_rate": 7.657006673021926e-07, "loss": 0.7777, "step": 39113 }, { "epoch": 7.45, "learning_rate": 7.656434699714013e-07, "loss": 1.0628, "step": 39114 }, { "epoch": 7.45, "learning_rate": 7.655862726406101e-07, "loss": 1.3221, "step": 39115 }, { "epoch": 7.45, "learning_rate": 7.65529075309819e-07, "loss": 0.8635, "step": 39116 }, { "epoch": 7.45, "learning_rate": 7.654718779790277e-07, "loss": 1.5323, "step": 39117 }, { "epoch": 7.45, "learning_rate": 7.654146806482364e-07, "loss": 2.2684, "step": 39118 }, { "epoch": 7.45, "learning_rate": 7.653574833174452e-07, "loss": 1.7285, "step": 39119 }, { "epoch": 7.45, "learning_rate": 7.65300285986654e-07, "loss": 1.0321, "step": 39120 }, { "epoch": 7.45, "learning_rate": 7.652430886558627e-07, "loss": 2.5718, "step": 39121 }, { "epoch": 7.45, "learning_rate": 7.651858913250715e-07, "loss": 1.374, "step": 39122 }, { "epoch": 7.45, "learning_rate": 7.651286939942803e-07, "loss": 0.9214, "step": 39123 }, { "epoch": 7.45, "learning_rate": 7.650714966634891e-07, "loss": 1.4703, "step": 39124 }, { "epoch": 7.45, "learning_rate": 7.650142993326978e-07, "loss": 2.1571, "step": 39125 }, { "epoch": 7.45, "learning_rate": 7.649571020019066e-07, "loss": 1.9114, "step": 39126 }, { "epoch": 7.45, "learning_rate": 7.648999046711155e-07, "loss": 3.555, "step": 39127 }, { "epoch": 7.45, "learning_rate": 7.648427073403241e-07, "loss": 2.4285, "step": 39128 }, { "epoch": 7.45, "learning_rate": 7.647855100095329e-07, "loss": 2.0621, "step": 39129 }, { "epoch": 7.45, "learning_rate": 7.647283126787418e-07, "loss": 0.9722, "step": 39130 }, { "epoch": 7.45, "learning_rate": 7.646711153479503e-07, "loss": 1.7729, "step": 39131 }, { "epoch": 7.45, "learning_rate": 7.646139180171592e-07, "loss": 1.7807, "step": 39132 }, { "epoch": 7.45, "learning_rate": 7.64556720686368e-07, "loss": 1.8483, "step": 39133 }, { "epoch": 7.45, "learning_rate": 7.644995233555768e-07, "loss": 1.4972, "step": 39134 }, { "epoch": 7.45, "learning_rate": 7.644423260247855e-07, "loss": 1.0426, "step": 39135 }, { "epoch": 7.45, "learning_rate": 7.643851286939943e-07, "loss": 1.2833, "step": 39136 }, { "epoch": 7.45, "learning_rate": 7.643279313632031e-07, "loss": 1.2501, "step": 39137 }, { "epoch": 7.45, "learning_rate": 7.642707340324118e-07, "loss": 0.8742, "step": 39138 }, { "epoch": 7.46, "learning_rate": 7.642135367016206e-07, "loss": 1.6828, "step": 39139 }, { "epoch": 7.46, "learning_rate": 7.641563393708294e-07, "loss": 2.6088, "step": 39140 }, { "epoch": 7.46, "learning_rate": 7.640991420400383e-07, "loss": 1.7743, "step": 39141 }, { "epoch": 7.46, "learning_rate": 7.640419447092469e-07, "loss": 1.8097, "step": 39142 }, { "epoch": 7.46, "learning_rate": 7.639847473784556e-07, "loss": 0.2108, "step": 39143 }, { "epoch": 7.46, "learning_rate": 7.639275500476645e-07, "loss": 3.021, "step": 39144 }, { "epoch": 7.46, "learning_rate": 7.638703527168731e-07, "loss": 0.5725, "step": 39145 }, { "epoch": 7.46, "learning_rate": 7.63813155386082e-07, "loss": 1.2452, "step": 39146 }, { "epoch": 7.46, "learning_rate": 7.637559580552908e-07, "loss": 0.8247, "step": 39147 }, { "epoch": 7.46, "learning_rate": 7.636987607244996e-07, "loss": 1.714, "step": 39148 }, { "epoch": 7.46, "learning_rate": 7.636415633937083e-07, "loss": 1.2015, "step": 39149 }, { "epoch": 7.46, "learning_rate": 7.635843660629171e-07, "loss": 2.3811, "step": 39150 }, { "epoch": 7.46, "learning_rate": 7.635271687321259e-07, "loss": 1.5518, "step": 39151 }, { "epoch": 7.46, "learning_rate": 7.634699714013346e-07, "loss": 2.2295, "step": 39152 }, { "epoch": 7.46, "learning_rate": 7.634127740705434e-07, "loss": 2.3852, "step": 39153 }, { "epoch": 7.46, "learning_rate": 7.633555767397522e-07, "loss": 1.3846, "step": 39154 }, { "epoch": 7.46, "learning_rate": 7.632983794089609e-07, "loss": 1.9358, "step": 39155 }, { "epoch": 7.46, "learning_rate": 7.632411820781697e-07, "loss": 1.2441, "step": 39156 }, { "epoch": 7.46, "learning_rate": 7.631839847473784e-07, "loss": 1.5196, "step": 39157 }, { "epoch": 7.46, "learning_rate": 7.631267874165873e-07, "loss": 3.273, "step": 39158 }, { "epoch": 7.46, "learning_rate": 7.630695900857959e-07, "loss": 2.0961, "step": 39159 }, { "epoch": 7.46, "learning_rate": 7.630123927550048e-07, "loss": 1.9476, "step": 39160 }, { "epoch": 7.46, "learning_rate": 7.629551954242136e-07, "loss": 3.5578, "step": 39161 }, { "epoch": 7.46, "learning_rate": 7.628979980934223e-07, "loss": 0.7284, "step": 39162 }, { "epoch": 7.46, "learning_rate": 7.628408007626311e-07, "loss": 1.6932, "step": 39163 }, { "epoch": 7.46, "learning_rate": 7.627836034318399e-07, "loss": 1.566, "step": 39164 }, { "epoch": 7.46, "learning_rate": 7.627264061010487e-07, "loss": 0.7183, "step": 39165 }, { "epoch": 7.46, "learning_rate": 7.626692087702574e-07, "loss": 0.9755, "step": 39166 }, { "epoch": 7.46, "learning_rate": 7.626120114394662e-07, "loss": 2.9244, "step": 39167 }, { "epoch": 7.46, "learning_rate": 7.62554814108675e-07, "loss": 1.0469, "step": 39168 }, { "epoch": 7.46, "learning_rate": 7.624976167778837e-07, "loss": 1.854, "step": 39169 }, { "epoch": 7.46, "learning_rate": 7.624404194470924e-07, "loss": 1.6396, "step": 39170 }, { "epoch": 7.46, "learning_rate": 7.623832221163013e-07, "loss": 2.6054, "step": 39171 }, { "epoch": 7.46, "learning_rate": 7.623260247855101e-07, "loss": 1.6054, "step": 39172 }, { "epoch": 7.46, "learning_rate": 7.622688274547187e-07, "loss": 2.6867, "step": 39173 }, { "epoch": 7.46, "learning_rate": 7.622116301239276e-07, "loss": 2.1814, "step": 39174 }, { "epoch": 7.46, "learning_rate": 7.621544327931364e-07, "loss": 1.6747, "step": 39175 }, { "epoch": 7.46, "learning_rate": 7.620972354623451e-07, "loss": 0.6076, "step": 39176 }, { "epoch": 7.46, "learning_rate": 7.620400381315539e-07, "loss": 1.0086, "step": 39177 }, { "epoch": 7.46, "learning_rate": 7.619828408007627e-07, "loss": 1.6488, "step": 39178 }, { "epoch": 7.46, "learning_rate": 7.619256434699714e-07, "loss": 1.477, "step": 39179 }, { "epoch": 7.46, "learning_rate": 7.618684461391802e-07, "loss": 1.5994, "step": 39180 }, { "epoch": 7.46, "learning_rate": 7.61811248808389e-07, "loss": 0.2686, "step": 39181 }, { "epoch": 7.46, "learning_rate": 7.617540514775978e-07, "loss": 2.5588, "step": 39182 }, { "epoch": 7.46, "learning_rate": 7.616968541468064e-07, "loss": 1.4839, "step": 39183 }, { "epoch": 7.46, "learning_rate": 7.616396568160152e-07, "loss": 1.0327, "step": 39184 }, { "epoch": 7.46, "learning_rate": 7.615824594852241e-07, "loss": 3.6607, "step": 39185 }, { "epoch": 7.46, "learning_rate": 7.615252621544327e-07, "loss": 1.4876, "step": 39186 }, { "epoch": 7.46, "learning_rate": 7.614680648236415e-07, "loss": 2.6619, "step": 39187 }, { "epoch": 7.46, "learning_rate": 7.614108674928504e-07, "loss": 1.9489, "step": 39188 }, { "epoch": 7.46, "learning_rate": 7.613536701620592e-07, "loss": 0.9312, "step": 39189 }, { "epoch": 7.46, "learning_rate": 7.612964728312679e-07, "loss": 1.8066, "step": 39190 }, { "epoch": 7.46, "learning_rate": 7.612392755004767e-07, "loss": 1.6662, "step": 39191 }, { "epoch": 7.47, "learning_rate": 7.611820781696855e-07, "loss": 1.7694, "step": 39192 }, { "epoch": 7.47, "learning_rate": 7.611248808388942e-07, "loss": 1.3379, "step": 39193 }, { "epoch": 7.47, "learning_rate": 7.61067683508103e-07, "loss": 2.7911, "step": 39194 }, { "epoch": 7.47, "learning_rate": 7.610104861773118e-07, "loss": 2.2719, "step": 39195 }, { "epoch": 7.47, "learning_rate": 7.609532888465204e-07, "loss": 2.3447, "step": 39196 }, { "epoch": 7.47, "learning_rate": 7.608960915157292e-07, "loss": 1.7587, "step": 39197 }, { "epoch": 7.47, "learning_rate": 7.60838894184938e-07, "loss": 0.8857, "step": 39198 }, { "epoch": 7.47, "learning_rate": 7.607816968541469e-07, "loss": 1.094, "step": 39199 }, { "epoch": 7.47, "learning_rate": 7.607244995233555e-07, "loss": 0.7296, "step": 39200 }, { "epoch": 7.47, "learning_rate": 7.606673021925643e-07, "loss": 1.4988, "step": 39201 }, { "epoch": 7.47, "learning_rate": 7.606101048617732e-07, "loss": 2.063, "step": 39202 }, { "epoch": 7.47, "learning_rate": 7.605529075309818e-07, "loss": 1.9738, "step": 39203 }, { "epoch": 7.47, "learning_rate": 7.604957102001907e-07, "loss": 4.4152, "step": 39204 }, { "epoch": 7.47, "learning_rate": 7.604385128693995e-07, "loss": 2.293, "step": 39205 }, { "epoch": 7.47, "learning_rate": 7.603813155386083e-07, "loss": 0.5304, "step": 39206 }, { "epoch": 7.47, "learning_rate": 7.60324118207817e-07, "loss": 3.5876, "step": 39207 }, { "epoch": 7.47, "learning_rate": 7.602669208770258e-07, "loss": 1.986, "step": 39208 }, { "epoch": 7.47, "learning_rate": 7.602097235462346e-07, "loss": 1.9877, "step": 39209 }, { "epoch": 7.47, "learning_rate": 7.601525262154432e-07, "loss": 0.9408, "step": 39210 }, { "epoch": 7.47, "learning_rate": 7.60095328884652e-07, "loss": 1.3304, "step": 39211 }, { "epoch": 7.47, "learning_rate": 7.600381315538608e-07, "loss": 1.5264, "step": 39212 }, { "epoch": 7.47, "learning_rate": 7.599809342230697e-07, "loss": 0.8657, "step": 39213 }, { "epoch": 7.47, "learning_rate": 7.599237368922783e-07, "loss": 1.4905, "step": 39214 }, { "epoch": 7.47, "learning_rate": 7.598665395614871e-07, "loss": 2.2491, "step": 39215 }, { "epoch": 7.47, "learning_rate": 7.59809342230696e-07, "loss": 2.2956, "step": 39216 }, { "epoch": 7.47, "learning_rate": 7.597521448999046e-07, "loss": 0.8846, "step": 39217 }, { "epoch": 7.47, "learning_rate": 7.596949475691135e-07, "loss": 2.6668, "step": 39218 }, { "epoch": 7.47, "learning_rate": 7.596377502383223e-07, "loss": 2.904, "step": 39219 }, { "epoch": 7.47, "learning_rate": 7.59580552907531e-07, "loss": 1.3133, "step": 39220 }, { "epoch": 7.47, "learning_rate": 7.595233555767398e-07, "loss": 0.8635, "step": 39221 }, { "epoch": 7.47, "learning_rate": 7.594661582459486e-07, "loss": 3.3417, "step": 39222 }, { "epoch": 7.47, "learning_rate": 7.594089609151574e-07, "loss": 0.6309, "step": 39223 }, { "epoch": 7.47, "learning_rate": 7.59351763584366e-07, "loss": 1.1572, "step": 39224 }, { "epoch": 7.47, "learning_rate": 7.592945662535748e-07, "loss": 1.6828, "step": 39225 }, { "epoch": 7.47, "learning_rate": 7.592373689227836e-07, "loss": 2.3851, "step": 39226 }, { "epoch": 7.47, "learning_rate": 7.591801715919923e-07, "loss": 2.0214, "step": 39227 }, { "epoch": 7.47, "learning_rate": 7.591229742612011e-07, "loss": 0.8642, "step": 39228 }, { "epoch": 7.47, "learning_rate": 7.5906577693041e-07, "loss": 3.0354, "step": 39229 }, { "epoch": 7.47, "learning_rate": 7.590085795996188e-07, "loss": 0.485, "step": 39230 }, { "epoch": 7.47, "learning_rate": 7.589513822688274e-07, "loss": 0.8876, "step": 39231 }, { "epoch": 7.47, "learning_rate": 7.588941849380363e-07, "loss": 0.882, "step": 39232 }, { "epoch": 7.47, "learning_rate": 7.588369876072451e-07, "loss": 1.2232, "step": 39233 }, { "epoch": 7.47, "learning_rate": 7.587797902764538e-07, "loss": 0.4828, "step": 39234 }, { "epoch": 7.47, "learning_rate": 7.587225929456626e-07, "loss": 2.1046, "step": 39235 }, { "epoch": 7.47, "learning_rate": 7.586653956148714e-07, "loss": 1.1378, "step": 39236 }, { "epoch": 7.47, "learning_rate": 7.5860819828408e-07, "loss": 1.8539, "step": 39237 }, { "epoch": 7.47, "learning_rate": 7.585510009532888e-07, "loss": 2.1277, "step": 39238 }, { "epoch": 7.47, "learning_rate": 7.584938036224976e-07, "loss": 0.3772, "step": 39239 }, { "epoch": 7.47, "learning_rate": 7.584366062917064e-07, "loss": 2.0977, "step": 39240 }, { "epoch": 7.47, "learning_rate": 7.583794089609151e-07, "loss": 0.6022, "step": 39241 }, { "epoch": 7.47, "learning_rate": 7.583222116301239e-07, "loss": 1.5539, "step": 39242 }, { "epoch": 7.47, "learning_rate": 7.582650142993328e-07, "loss": 1.0313, "step": 39243 }, { "epoch": 7.48, "learning_rate": 7.582078169685414e-07, "loss": 2.7862, "step": 39244 }, { "epoch": 7.48, "learning_rate": 7.581506196377502e-07, "loss": 1.1235, "step": 39245 }, { "epoch": 7.48, "learning_rate": 7.580934223069591e-07, "loss": 2.7456, "step": 39246 }, { "epoch": 7.48, "learning_rate": 7.580362249761679e-07, "loss": 1.5445, "step": 39247 }, { "epoch": 7.48, "learning_rate": 7.579790276453766e-07, "loss": 2.8691, "step": 39248 }, { "epoch": 7.48, "learning_rate": 7.579218303145854e-07, "loss": 1.3287, "step": 39249 }, { "epoch": 7.48, "learning_rate": 7.578646329837942e-07, "loss": 1.0112, "step": 39250 }, { "epoch": 7.48, "learning_rate": 7.578074356530028e-07, "loss": 2.9403, "step": 39251 }, { "epoch": 7.48, "learning_rate": 7.577502383222116e-07, "loss": 1.4019, "step": 39252 }, { "epoch": 7.48, "learning_rate": 7.576930409914204e-07, "loss": 1.608, "step": 39253 }, { "epoch": 7.48, "learning_rate": 7.576358436606292e-07, "loss": 2.3262, "step": 39254 }, { "epoch": 7.48, "learning_rate": 7.575786463298379e-07, "loss": 0.6686, "step": 39255 }, { "epoch": 7.48, "learning_rate": 7.575214489990467e-07, "loss": 2.3539, "step": 39256 }, { "epoch": 7.48, "learning_rate": 7.574642516682556e-07, "loss": 1.0854, "step": 39257 }, { "epoch": 7.48, "learning_rate": 7.574070543374642e-07, "loss": 1.1486, "step": 39258 }, { "epoch": 7.48, "learning_rate": 7.57349857006673e-07, "loss": 3.0384, "step": 39259 }, { "epoch": 7.48, "learning_rate": 7.572926596758819e-07, "loss": 3.1078, "step": 39260 }, { "epoch": 7.48, "learning_rate": 7.572354623450905e-07, "loss": 1.6105, "step": 39261 }, { "epoch": 7.48, "learning_rate": 7.571782650142994e-07, "loss": 2.8088, "step": 39262 }, { "epoch": 7.48, "learning_rate": 7.571210676835082e-07, "loss": 2.3917, "step": 39263 }, { "epoch": 7.48, "learning_rate": 7.57063870352717e-07, "loss": 2.1391, "step": 39264 }, { "epoch": 7.48, "learning_rate": 7.570066730219256e-07, "loss": 2.0536, "step": 39265 }, { "epoch": 7.48, "learning_rate": 7.569494756911344e-07, "loss": 1.6055, "step": 39266 }, { "epoch": 7.48, "learning_rate": 7.568922783603432e-07, "loss": 2.1943, "step": 39267 }, { "epoch": 7.48, "learning_rate": 7.568350810295519e-07, "loss": 0.9807, "step": 39268 }, { "epoch": 7.48, "learning_rate": 7.567778836987607e-07, "loss": 2.2929, "step": 39269 }, { "epoch": 7.48, "learning_rate": 7.567206863679695e-07, "loss": 0.6838, "step": 39270 }, { "epoch": 7.48, "learning_rate": 7.566634890371784e-07, "loss": 1.592, "step": 39271 }, { "epoch": 7.48, "learning_rate": 7.56606291706387e-07, "loss": 1.0794, "step": 39272 }, { "epoch": 7.48, "learning_rate": 7.565490943755959e-07, "loss": 1.5717, "step": 39273 }, { "epoch": 7.48, "learning_rate": 7.564918970448047e-07, "loss": 1.211, "step": 39274 }, { "epoch": 7.48, "learning_rate": 7.564346997140133e-07, "loss": 0.7443, "step": 39275 }, { "epoch": 7.48, "learning_rate": 7.563775023832222e-07, "loss": 2.0336, "step": 39276 }, { "epoch": 7.48, "learning_rate": 7.56320305052431e-07, "loss": 1.4516, "step": 39277 }, { "epoch": 7.48, "learning_rate": 7.562631077216397e-07, "loss": 1.9692, "step": 39278 }, { "epoch": 7.48, "learning_rate": 7.562059103908484e-07, "loss": 3.2062, "step": 39279 }, { "epoch": 7.48, "learning_rate": 7.561487130600572e-07, "loss": 2.1949, "step": 39280 }, { "epoch": 7.48, "learning_rate": 7.56091515729266e-07, "loss": 1.472, "step": 39281 }, { "epoch": 7.48, "learning_rate": 7.560343183984747e-07, "loss": 1.1468, "step": 39282 }, { "epoch": 7.48, "learning_rate": 7.559771210676835e-07, "loss": 2.3651, "step": 39283 }, { "epoch": 7.48, "learning_rate": 7.559199237368923e-07, "loss": 1.4141, "step": 39284 }, { "epoch": 7.48, "learning_rate": 7.55862726406101e-07, "loss": 2.4082, "step": 39285 }, { "epoch": 7.48, "learning_rate": 7.558055290753098e-07, "loss": 2.5101, "step": 39286 }, { "epoch": 7.48, "learning_rate": 7.557483317445187e-07, "loss": 2.692, "step": 39287 }, { "epoch": 7.48, "learning_rate": 7.556911344137275e-07, "loss": 2.1779, "step": 39288 }, { "epoch": 7.48, "learning_rate": 7.55633937082936e-07, "loss": 1.5617, "step": 39289 }, { "epoch": 7.48, "learning_rate": 7.55576739752145e-07, "loss": 1.4282, "step": 39290 }, { "epoch": 7.48, "learning_rate": 7.555195424213537e-07, "loss": 1.9759, "step": 39291 }, { "epoch": 7.48, "learning_rate": 7.554623450905624e-07, "loss": 2.807, "step": 39292 }, { "epoch": 7.48, "learning_rate": 7.554051477597712e-07, "loss": 2.9354, "step": 39293 }, { "epoch": 7.48, "learning_rate": 7.5534795042898e-07, "loss": 1.4157, "step": 39294 }, { "epoch": 7.48, "learning_rate": 7.552907530981888e-07, "loss": 2.3083, "step": 39295 }, { "epoch": 7.48, "learning_rate": 7.552335557673975e-07, "loss": 2.1517, "step": 39296 }, { "epoch": 7.49, "learning_rate": 7.551763584366063e-07, "loss": 1.1003, "step": 39297 }, { "epoch": 7.49, "learning_rate": 7.551191611058151e-07, "loss": 1.5777, "step": 39298 }, { "epoch": 7.49, "learning_rate": 7.550619637750238e-07, "loss": 0.6139, "step": 39299 }, { "epoch": 7.49, "learning_rate": 7.550047664442326e-07, "loss": 1.8615, "step": 39300 }, { "epoch": 7.49, "learning_rate": 7.549475691134415e-07, "loss": 0.9453, "step": 39301 }, { "epoch": 7.49, "learning_rate": 7.548903717826501e-07, "loss": 1.3728, "step": 39302 }, { "epoch": 7.49, "learning_rate": 7.548331744518588e-07, "loss": 1.4316, "step": 39303 }, { "epoch": 7.49, "learning_rate": 7.547759771210677e-07, "loss": 2.2617, "step": 39304 }, { "epoch": 7.49, "learning_rate": 7.547187797902765e-07, "loss": 1.4823, "step": 39305 }, { "epoch": 7.49, "learning_rate": 7.546615824594852e-07, "loss": 2.1588, "step": 39306 }, { "epoch": 7.49, "learning_rate": 7.54604385128694e-07, "loss": 0.8534, "step": 39307 }, { "epoch": 7.49, "learning_rate": 7.545471877979028e-07, "loss": 1.8612, "step": 39308 }, { "epoch": 7.49, "learning_rate": 7.544899904671115e-07, "loss": 0.6264, "step": 39309 }, { "epoch": 7.49, "learning_rate": 7.544327931363203e-07, "loss": 1.1653, "step": 39310 }, { "epoch": 7.49, "learning_rate": 7.543755958055291e-07, "loss": 1.3437, "step": 39311 }, { "epoch": 7.49, "learning_rate": 7.543183984747379e-07, "loss": 1.722, "step": 39312 }, { "epoch": 7.49, "learning_rate": 7.542612011439466e-07, "loss": 3.926, "step": 39313 }, { "epoch": 7.49, "learning_rate": 7.542040038131554e-07, "loss": 1.4508, "step": 39314 }, { "epoch": 7.49, "learning_rate": 7.541468064823643e-07, "loss": 0.7463, "step": 39315 }, { "epoch": 7.49, "learning_rate": 7.540896091515728e-07, "loss": 0.9569, "step": 39316 }, { "epoch": 7.49, "learning_rate": 7.540324118207817e-07, "loss": 1.3148, "step": 39317 }, { "epoch": 7.49, "learning_rate": 7.539752144899905e-07, "loss": 3.4108, "step": 39318 }, { "epoch": 7.49, "learning_rate": 7.539180171591993e-07, "loss": 1.9081, "step": 39319 }, { "epoch": 7.49, "learning_rate": 7.53860819828408e-07, "loss": 1.0197, "step": 39320 }, { "epoch": 7.49, "learning_rate": 7.538036224976168e-07, "loss": 1.1741, "step": 39321 }, { "epoch": 7.49, "learning_rate": 7.537464251668256e-07, "loss": 0.5102, "step": 39322 }, { "epoch": 7.49, "learning_rate": 7.536892278360343e-07, "loss": 1.8899, "step": 39323 }, { "epoch": 7.49, "learning_rate": 7.536320305052431e-07, "loss": 2.0684, "step": 39324 }, { "epoch": 7.49, "learning_rate": 7.535748331744519e-07, "loss": 2.151, "step": 39325 }, { "epoch": 7.49, "learning_rate": 7.535176358436606e-07, "loss": 1.3997, "step": 39326 }, { "epoch": 7.49, "learning_rate": 7.534604385128694e-07, "loss": 2.5261, "step": 39327 }, { "epoch": 7.49, "learning_rate": 7.534032411820782e-07, "loss": 0.7962, "step": 39328 }, { "epoch": 7.49, "learning_rate": 7.533460438512871e-07, "loss": 1.9249, "step": 39329 }, { "epoch": 7.49, "learning_rate": 7.532888465204956e-07, "loss": 1.8718, "step": 39330 }, { "epoch": 7.49, "learning_rate": 7.532316491897045e-07, "loss": 1.8963, "step": 39331 }, { "epoch": 7.49, "learning_rate": 7.531744518589133e-07, "loss": 2.0551, "step": 39332 }, { "epoch": 7.49, "learning_rate": 7.531172545281219e-07, "loss": 1.3316, "step": 39333 }, { "epoch": 7.49, "learning_rate": 7.530600571973308e-07, "loss": 2.1541, "step": 39334 }, { "epoch": 7.49, "learning_rate": 7.530028598665396e-07, "loss": 1.6765, "step": 39335 }, { "epoch": 7.49, "learning_rate": 7.529456625357484e-07, "loss": 0.9264, "step": 39336 }, { "epoch": 7.49, "learning_rate": 7.528884652049571e-07, "loss": 1.0027, "step": 39337 }, { "epoch": 7.49, "learning_rate": 7.528312678741659e-07, "loss": 1.7569, "step": 39338 }, { "epoch": 7.49, "learning_rate": 7.527740705433747e-07, "loss": 0.6079, "step": 39339 }, { "epoch": 7.49, "learning_rate": 7.527168732125834e-07, "loss": 1.1249, "step": 39340 }, { "epoch": 7.49, "learning_rate": 7.526596758817922e-07, "loss": 1.6493, "step": 39341 }, { "epoch": 7.49, "learning_rate": 7.52602478551001e-07, "loss": 1.4415, "step": 39342 }, { "epoch": 7.49, "learning_rate": 7.525452812202099e-07, "loss": 1.5485, "step": 39343 }, { "epoch": 7.49, "learning_rate": 7.524880838894184e-07, "loss": 1.1828, "step": 39344 }, { "epoch": 7.49, "learning_rate": 7.524308865586273e-07, "loss": 1.4632, "step": 39345 }, { "epoch": 7.49, "learning_rate": 7.523736892278361e-07, "loss": 1.5253, "step": 39346 }, { "epoch": 7.49, "learning_rate": 7.523164918970447e-07, "loss": 2.6459, "step": 39347 }, { "epoch": 7.49, "learning_rate": 7.522592945662536e-07, "loss": 2.111, "step": 39348 }, { "epoch": 7.5, "learning_rate": 7.522020972354624e-07, "loss": 2.7414, "step": 39349 }, { "epoch": 7.5, "learning_rate": 7.521448999046711e-07, "loss": 1.8234, "step": 39350 }, { "epoch": 7.5, "learning_rate": 7.520877025738799e-07, "loss": 3.6262, "step": 39351 }, { "epoch": 7.5, "learning_rate": 7.520305052430887e-07, "loss": 1.297, "step": 39352 }, { "epoch": 7.5, "learning_rate": 7.519733079122975e-07, "loss": 1.4067, "step": 39353 }, { "epoch": 7.5, "learning_rate": 7.519161105815062e-07, "loss": 2.0481, "step": 39354 }, { "epoch": 7.5, "learning_rate": 7.51858913250715e-07, "loss": 1.7564, "step": 39355 }, { "epoch": 7.5, "learning_rate": 7.518017159199238e-07, "loss": 1.8785, "step": 39356 }, { "epoch": 7.5, "learning_rate": 7.517445185891324e-07, "loss": 1.7587, "step": 39357 }, { "epoch": 7.5, "learning_rate": 7.516873212583412e-07, "loss": 1.5193, "step": 39358 }, { "epoch": 7.5, "learning_rate": 7.516301239275501e-07, "loss": 1.1612, "step": 39359 }, { "epoch": 7.5, "learning_rate": 7.515729265967589e-07, "loss": 1.1206, "step": 39360 }, { "epoch": 7.5, "learning_rate": 7.515157292659676e-07, "loss": 2.1694, "step": 39361 }, { "epoch": 7.5, "learning_rate": 7.514585319351764e-07, "loss": 0.6544, "step": 39362 }, { "epoch": 7.5, "learning_rate": 7.514013346043852e-07, "loss": 1.2925, "step": 39363 }, { "epoch": 7.5, "learning_rate": 7.513441372735939e-07, "loss": 1.3506, "step": 39364 }, { "epoch": 7.5, "learning_rate": 7.512869399428027e-07, "loss": 1.5419, "step": 39365 }, { "epoch": 7.5, "learning_rate": 7.512297426120115e-07, "loss": 2.2867, "step": 39366 }, { "epoch": 7.5, "learning_rate": 7.511725452812202e-07, "loss": 1.3742, "step": 39367 }, { "epoch": 7.5, "learning_rate": 7.51115347950429e-07, "loss": 0.5442, "step": 39368 }, { "epoch": 7.5, "learning_rate": 7.510581506196378e-07, "loss": 1.9779, "step": 39369 }, { "epoch": 7.5, "learning_rate": 7.510009532888466e-07, "loss": 1.018, "step": 39370 }, { "epoch": 7.5, "learning_rate": 7.509437559580552e-07, "loss": 2.5749, "step": 39371 }, { "epoch": 7.5, "learning_rate": 7.50886558627264e-07, "loss": 2.56, "step": 39372 }, { "epoch": 7.5, "learning_rate": 7.508293612964729e-07, "loss": 2.0498, "step": 39373 }, { "epoch": 7.5, "learning_rate": 7.507721639656815e-07, "loss": 1.9917, "step": 39374 }, { "epoch": 7.5, "learning_rate": 7.507149666348904e-07, "loss": 0.9469, "step": 39375 }, { "epoch": 7.5, "learning_rate": 7.506577693040992e-07, "loss": 2.1126, "step": 39376 }, { "epoch": 7.5, "learning_rate": 7.50600571973308e-07, "loss": 1.3268, "step": 39377 }, { "epoch": 7.5, "learning_rate": 7.505433746425167e-07, "loss": 2.2594, "step": 39378 }, { "epoch": 7.5, "learning_rate": 7.504861773117255e-07, "loss": 1.1464, "step": 39379 }, { "epoch": 7.5, "learning_rate": 7.504289799809343e-07, "loss": 1.8322, "step": 39380 }, { "epoch": 7.5, "learning_rate": 7.50371782650143e-07, "loss": 2.5594, "step": 39381 }, { "epoch": 7.5, "learning_rate": 7.503145853193518e-07, "loss": 2.1723, "step": 39382 }, { "epoch": 7.5, "learning_rate": 7.502573879885606e-07, "loss": 1.9639, "step": 39383 }, { "epoch": 7.5, "learning_rate": 7.502001906577695e-07, "loss": 2.7949, "step": 39384 }, { "epoch": 7.5, "learning_rate": 7.50142993326978e-07, "loss": 1.3662, "step": 39385 }, { "epoch": 7.5, "learning_rate": 7.500857959961868e-07, "loss": 1.6744, "step": 39386 }, { "epoch": 7.5, "learning_rate": 7.500285986653957e-07, "loss": 1.7978, "step": 39387 }, { "epoch": 7.5, "learning_rate": 7.499714013346044e-07, "loss": 1.6379, "step": 39388 }, { "epoch": 7.5, "learning_rate": 7.499142040038132e-07, "loss": 2.5182, "step": 39389 }, { "epoch": 7.5, "learning_rate": 7.49857006673022e-07, "loss": 0.835, "step": 39390 }, { "epoch": 7.5, "learning_rate": 7.497998093422307e-07, "loss": 1.8645, "step": 39391 }, { "epoch": 7.5, "learning_rate": 7.497426120114395e-07, "loss": 2.4095, "step": 39392 }, { "epoch": 7.5, "learning_rate": 7.496854146806483e-07, "loss": 1.2144, "step": 39393 }, { "epoch": 7.5, "learning_rate": 7.49628217349857e-07, "loss": 1.8063, "step": 39394 }, { "epoch": 7.5, "learning_rate": 7.495710200190659e-07, "loss": 1.1448, "step": 39395 }, { "epoch": 7.5, "learning_rate": 7.495138226882746e-07, "loss": 3.0503, "step": 39396 }, { "epoch": 7.5, "learning_rate": 7.494566253574833e-07, "loss": 2.0849, "step": 39397 }, { "epoch": 7.5, "learning_rate": 7.493994280266921e-07, "loss": 0.6669, "step": 39398 }, { "epoch": 7.5, "learning_rate": 7.493422306959008e-07, "loss": 0.9779, "step": 39399 }, { "epoch": 7.5, "learning_rate": 7.492850333651096e-07, "loss": 1.8894, "step": 39400 }, { "epoch": 7.5, "learning_rate": 7.492278360343184e-07, "loss": 0.425, "step": 39401 }, { "epoch": 7.51, "learning_rate": 7.491706387035272e-07, "loss": 0.9178, "step": 39402 }, { "epoch": 7.51, "learning_rate": 7.49113441372736e-07, "loss": 1.595, "step": 39403 }, { "epoch": 7.51, "learning_rate": 7.490562440419447e-07, "loss": 1.9261, "step": 39404 }, { "epoch": 7.51, "learning_rate": 7.489990467111535e-07, "loss": 1.3749, "step": 39405 }, { "epoch": 7.51, "learning_rate": 7.489418493803623e-07, "loss": 2.4883, "step": 39406 }, { "epoch": 7.51, "learning_rate": 7.488846520495711e-07, "loss": 1.8572, "step": 39407 }, { "epoch": 7.51, "learning_rate": 7.488274547187798e-07, "loss": 1.9209, "step": 39408 }, { "epoch": 7.51, "learning_rate": 7.487702573879886e-07, "loss": 0.2852, "step": 39409 }, { "epoch": 7.51, "learning_rate": 7.487130600571974e-07, "loss": 1.6901, "step": 39410 }, { "epoch": 7.51, "learning_rate": 7.486558627264061e-07, "loss": 0.9052, "step": 39411 }, { "epoch": 7.51, "learning_rate": 7.485986653956149e-07, "loss": 0.9115, "step": 39412 }, { "epoch": 7.51, "learning_rate": 7.485414680648236e-07, "loss": 1.8643, "step": 39413 }, { "epoch": 7.51, "learning_rate": 7.484842707340324e-07, "loss": 2.3642, "step": 39414 }, { "epoch": 7.51, "learning_rate": 7.484270734032412e-07, "loss": 2.4218, "step": 39415 }, { "epoch": 7.51, "learning_rate": 7.483698760724499e-07, "loss": 0.8065, "step": 39416 }, { "epoch": 7.51, "learning_rate": 7.483126787416588e-07, "loss": 1.6654, "step": 39417 }, { "epoch": 7.51, "learning_rate": 7.482554814108675e-07, "loss": 0.7894, "step": 39418 }, { "epoch": 7.51, "learning_rate": 7.481982840800763e-07, "loss": 2.0239, "step": 39419 }, { "epoch": 7.51, "learning_rate": 7.481410867492851e-07, "loss": 1.2345, "step": 39420 }, { "epoch": 7.51, "learning_rate": 7.480838894184938e-07, "loss": 1.1118, "step": 39421 }, { "epoch": 7.51, "learning_rate": 7.480266920877026e-07, "loss": 2.5963, "step": 39422 }, { "epoch": 7.51, "learning_rate": 7.479694947569114e-07, "loss": 3.1611, "step": 39423 }, { "epoch": 7.51, "learning_rate": 7.479122974261201e-07, "loss": 1.3756, "step": 39424 }, { "epoch": 7.51, "learning_rate": 7.478551000953289e-07, "loss": 0.8616, "step": 39425 }, { "epoch": 7.51, "learning_rate": 7.477979027645376e-07, "loss": 2.7196, "step": 39426 }, { "epoch": 7.51, "learning_rate": 7.477407054337464e-07, "loss": 2.7065, "step": 39427 }, { "epoch": 7.51, "learning_rate": 7.476835081029552e-07, "loss": 1.9271, "step": 39428 }, { "epoch": 7.51, "learning_rate": 7.47626310772164e-07, "loss": 1.9969, "step": 39429 }, { "epoch": 7.51, "learning_rate": 7.475691134413727e-07, "loss": 1.7664, "step": 39430 }, { "epoch": 7.51, "learning_rate": 7.475119161105816e-07, "loss": 2.1782, "step": 39431 }, { "epoch": 7.51, "learning_rate": 7.474547187797903e-07, "loss": 2.5629, "step": 39432 }, { "epoch": 7.51, "learning_rate": 7.473975214489991e-07, "loss": 1.0789, "step": 39433 }, { "epoch": 7.51, "learning_rate": 7.473403241182079e-07, "loss": 3.711, "step": 39434 }, { "epoch": 7.51, "learning_rate": 7.472831267874166e-07, "loss": 0.5489, "step": 39435 }, { "epoch": 7.51, "learning_rate": 7.472259294566254e-07, "loss": 4.3083, "step": 39436 }, { "epoch": 7.51, "learning_rate": 7.471687321258341e-07, "loss": 3.2353, "step": 39437 }, { "epoch": 7.51, "learning_rate": 7.471115347950428e-07, "loss": 2.7618, "step": 39438 }, { "epoch": 7.51, "learning_rate": 7.470543374642517e-07, "loss": 1.1642, "step": 39439 }, { "epoch": 7.51, "learning_rate": 7.469971401334604e-07, "loss": 1.1065, "step": 39440 }, { "epoch": 7.51, "learning_rate": 7.469399428026692e-07, "loss": 1.8352, "step": 39441 }, { "epoch": 7.51, "learning_rate": 7.46882745471878e-07, "loss": 1.2206, "step": 39442 }, { "epoch": 7.51, "learning_rate": 7.468255481410868e-07, "loss": 1.8347, "step": 39443 }, { "epoch": 7.51, "learning_rate": 7.467683508102955e-07, "loss": 2.3023, "step": 39444 }, { "epoch": 7.51, "learning_rate": 7.467111534795043e-07, "loss": 2.5001, "step": 39445 }, { "epoch": 7.51, "learning_rate": 7.466539561487131e-07, "loss": 1.1913, "step": 39446 }, { "epoch": 7.51, "learning_rate": 7.465967588179219e-07, "loss": 1.2923, "step": 39447 }, { "epoch": 7.51, "learning_rate": 7.465395614871307e-07, "loss": 3.0087, "step": 39448 }, { "epoch": 7.51, "learning_rate": 7.464823641563394e-07, "loss": 1.8262, "step": 39449 }, { "epoch": 7.51, "learning_rate": 7.464251668255482e-07, "loss": 1.8424, "step": 39450 }, { "epoch": 7.51, "learning_rate": 7.463679694947569e-07, "loss": 1.7437, "step": 39451 }, { "epoch": 7.51, "learning_rate": 7.463107721639656e-07, "loss": 2.0342, "step": 39452 }, { "epoch": 7.51, "learning_rate": 7.462535748331745e-07, "loss": 0.9072, "step": 39453 }, { "epoch": 7.52, "learning_rate": 7.461963775023832e-07, "loss": 2.3906, "step": 39454 }, { "epoch": 7.52, "learning_rate": 7.46139180171592e-07, "loss": 1.8765, "step": 39455 }, { "epoch": 7.52, "learning_rate": 7.460819828408008e-07, "loss": 1.4582, "step": 39456 }, { "epoch": 7.52, "learning_rate": 7.460247855100095e-07, "loss": 1.3451, "step": 39457 }, { "epoch": 7.52, "learning_rate": 7.459675881792183e-07, "loss": 2.4035, "step": 39458 }, { "epoch": 7.52, "learning_rate": 7.459103908484271e-07, "loss": 2.1939, "step": 39459 }, { "epoch": 7.52, "learning_rate": 7.458531935176359e-07, "loss": 1.8952, "step": 39460 }, { "epoch": 7.52, "learning_rate": 7.457959961868447e-07, "loss": 0.3054, "step": 39461 }, { "epoch": 7.52, "learning_rate": 7.457387988560534e-07, "loss": 1.2641, "step": 39462 }, { "epoch": 7.52, "learning_rate": 7.456816015252622e-07, "loss": 2.7841, "step": 39463 }, { "epoch": 7.52, "learning_rate": 7.456244041944709e-07, "loss": 1.0092, "step": 39464 }, { "epoch": 7.52, "learning_rate": 7.455672068636797e-07, "loss": 0.6564, "step": 39465 }, { "epoch": 7.52, "learning_rate": 7.455100095328884e-07, "loss": 2.3361, "step": 39466 }, { "epoch": 7.52, "learning_rate": 7.454528122020973e-07, "loss": 1.0308, "step": 39467 }, { "epoch": 7.52, "learning_rate": 7.45395614871306e-07, "loss": 2.2659, "step": 39468 }, { "epoch": 7.52, "learning_rate": 7.453384175405148e-07, "loss": 1.2168, "step": 39469 }, { "epoch": 7.52, "learning_rate": 7.452812202097236e-07, "loss": 1.5348, "step": 39470 }, { "epoch": 7.52, "learning_rate": 7.452240228789323e-07, "loss": 2.0558, "step": 39471 }, { "epoch": 7.52, "learning_rate": 7.451668255481412e-07, "loss": 2.8725, "step": 39472 }, { "epoch": 7.52, "learning_rate": 7.451096282173499e-07, "loss": 3.0053, "step": 39473 }, { "epoch": 7.52, "learning_rate": 7.450524308865586e-07, "loss": 1.787, "step": 39474 }, { "epoch": 7.52, "learning_rate": 7.449952335557675e-07, "loss": 0.68, "step": 39475 }, { "epoch": 7.52, "learning_rate": 7.449380362249762e-07, "loss": 0.484, "step": 39476 }, { "epoch": 7.52, "learning_rate": 7.44880838894185e-07, "loss": 1.63, "step": 39477 }, { "epoch": 7.52, "learning_rate": 7.448236415633937e-07, "loss": 1.5482, "step": 39478 }, { "epoch": 7.52, "learning_rate": 7.447664442326024e-07, "loss": 3.3657, "step": 39479 }, { "epoch": 7.52, "learning_rate": 7.447092469018112e-07, "loss": 1.6502, "step": 39480 }, { "epoch": 7.52, "learning_rate": 7.4465204957102e-07, "loss": 0.6815, "step": 39481 }, { "epoch": 7.52, "learning_rate": 7.445948522402288e-07, "loss": 1.4777, "step": 39482 }, { "epoch": 7.52, "learning_rate": 7.445376549094376e-07, "loss": 0.6141, "step": 39483 }, { "epoch": 7.52, "learning_rate": 7.444804575786464e-07, "loss": 1.4845, "step": 39484 }, { "epoch": 7.52, "learning_rate": 7.444232602478551e-07, "loss": 1.3416, "step": 39485 }, { "epoch": 7.52, "learning_rate": 7.443660629170639e-07, "loss": 1.9119, "step": 39486 }, { "epoch": 7.52, "learning_rate": 7.443088655862727e-07, "loss": 1.9982, "step": 39487 }, { "epoch": 7.52, "learning_rate": 7.442516682554814e-07, "loss": 2.7087, "step": 39488 }, { "epoch": 7.52, "learning_rate": 7.441944709246903e-07, "loss": 1.0043, "step": 39489 }, { "epoch": 7.52, "learning_rate": 7.44137273593899e-07, "loss": 0.8066, "step": 39490 }, { "epoch": 7.52, "learning_rate": 7.440800762631077e-07, "loss": 1.5241, "step": 39491 }, { "epoch": 7.52, "learning_rate": 7.440228789323165e-07, "loss": 1.4252, "step": 39492 }, { "epoch": 7.52, "learning_rate": 7.439656816015252e-07, "loss": 2.5674, "step": 39493 }, { "epoch": 7.52, "learning_rate": 7.439084842707341e-07, "loss": 3.7313, "step": 39494 }, { "epoch": 7.52, "learning_rate": 7.438512869399428e-07, "loss": 2.1048, "step": 39495 }, { "epoch": 7.52, "learning_rate": 7.437940896091516e-07, "loss": 2.4228, "step": 39496 }, { "epoch": 7.52, "learning_rate": 7.437368922783604e-07, "loss": 1.5953, "step": 39497 }, { "epoch": 7.52, "learning_rate": 7.436796949475691e-07, "loss": 1.4031, "step": 39498 }, { "epoch": 7.52, "learning_rate": 7.436224976167779e-07, "loss": 1.6511, "step": 39499 }, { "epoch": 7.52, "learning_rate": 7.435653002859867e-07, "loss": 1.4593, "step": 39500 }, { "epoch": 7.52, "learning_rate": 7.435081029551955e-07, "loss": 1.8888, "step": 39501 }, { "epoch": 7.52, "learning_rate": 7.434509056244042e-07, "loss": 0.5189, "step": 39502 }, { "epoch": 7.52, "learning_rate": 7.43393708293613e-07, "loss": 1.9475, "step": 39503 }, { "epoch": 7.52, "learning_rate": 7.433365109628217e-07, "loss": 1.5648, "step": 39504 }, { "epoch": 7.52, "learning_rate": 7.432793136320305e-07, "loss": 2.4894, "step": 39505 }, { "epoch": 7.52, "learning_rate": 7.432221163012393e-07, "loss": 1.6241, "step": 39506 }, { "epoch": 7.53, "learning_rate": 7.43164918970448e-07, "loss": 1.2141, "step": 39507 }, { "epoch": 7.53, "learning_rate": 7.431077216396569e-07, "loss": 0.9956, "step": 39508 }, { "epoch": 7.53, "learning_rate": 7.430505243088656e-07, "loss": 1.9915, "step": 39509 }, { "epoch": 7.53, "learning_rate": 7.429933269780743e-07, "loss": 0.58, "step": 39510 }, { "epoch": 7.53, "learning_rate": 7.429361296472832e-07, "loss": 2.099, "step": 39511 }, { "epoch": 7.53, "learning_rate": 7.428789323164919e-07, "loss": 1.3443, "step": 39512 }, { "epoch": 7.53, "learning_rate": 7.428217349857007e-07, "loss": 1.9549, "step": 39513 }, { "epoch": 7.53, "learning_rate": 7.427645376549095e-07, "loss": 1.0214, "step": 39514 }, { "epoch": 7.53, "learning_rate": 7.427073403241182e-07, "loss": 1.5012, "step": 39515 }, { "epoch": 7.53, "learning_rate": 7.426501429933271e-07, "loss": 3.1741, "step": 39516 }, { "epoch": 7.53, "learning_rate": 7.425929456625357e-07, "loss": 1.1537, "step": 39517 }, { "epoch": 7.53, "learning_rate": 7.425357483317445e-07, "loss": 1.2264, "step": 39518 }, { "epoch": 7.53, "learning_rate": 7.424785510009533e-07, "loss": 0.9847, "step": 39519 }, { "epoch": 7.53, "learning_rate": 7.424213536701621e-07, "loss": 1.3645, "step": 39520 }, { "epoch": 7.53, "learning_rate": 7.423641563393708e-07, "loss": 0.394, "step": 39521 }, { "epoch": 7.53, "learning_rate": 7.423069590085796e-07, "loss": 0.9178, "step": 39522 }, { "epoch": 7.53, "learning_rate": 7.422497616777884e-07, "loss": 1.356, "step": 39523 }, { "epoch": 7.53, "learning_rate": 7.421925643469971e-07, "loss": 1.85, "step": 39524 }, { "epoch": 7.53, "learning_rate": 7.42135367016206e-07, "loss": 1.4134, "step": 39525 }, { "epoch": 7.53, "learning_rate": 7.420781696854147e-07, "loss": 2.5655, "step": 39526 }, { "epoch": 7.53, "learning_rate": 7.420209723546235e-07, "loss": 0.972, "step": 39527 }, { "epoch": 7.53, "learning_rate": 7.419637750238323e-07, "loss": 1.2789, "step": 39528 }, { "epoch": 7.53, "learning_rate": 7.41906577693041e-07, "loss": 2.0323, "step": 39529 }, { "epoch": 7.53, "learning_rate": 7.418493803622499e-07, "loss": 0.6859, "step": 39530 }, { "epoch": 7.53, "learning_rate": 7.417921830314585e-07, "loss": 2.0669, "step": 39531 }, { "epoch": 7.53, "learning_rate": 7.417349857006673e-07, "loss": 1.2909, "step": 39532 }, { "epoch": 7.53, "learning_rate": 7.416777883698761e-07, "loss": 0.1662, "step": 39533 }, { "epoch": 7.53, "learning_rate": 7.416205910390848e-07, "loss": 2.0307, "step": 39534 }, { "epoch": 7.53, "learning_rate": 7.415633937082936e-07, "loss": 1.4648, "step": 39535 }, { "epoch": 7.53, "learning_rate": 7.415061963775024e-07, "loss": 0.7674, "step": 39536 }, { "epoch": 7.53, "learning_rate": 7.414489990467112e-07, "loss": 0.7504, "step": 39537 }, { "epoch": 7.53, "learning_rate": 7.4139180171592e-07, "loss": 1.2999, "step": 39538 }, { "epoch": 7.53, "learning_rate": 7.413346043851287e-07, "loss": 1.7683, "step": 39539 }, { "epoch": 7.53, "learning_rate": 7.412774070543375e-07, "loss": 2.5309, "step": 39540 }, { "epoch": 7.53, "learning_rate": 7.412202097235463e-07, "loss": 1.7516, "step": 39541 }, { "epoch": 7.53, "learning_rate": 7.411630123927551e-07, "loss": 1.5116, "step": 39542 }, { "epoch": 7.53, "learning_rate": 7.411058150619638e-07, "loss": 2.0819, "step": 39543 }, { "epoch": 7.53, "learning_rate": 7.410486177311725e-07, "loss": 1.8571, "step": 39544 }, { "epoch": 7.53, "learning_rate": 7.409914204003813e-07, "loss": 1.1914, "step": 39545 }, { "epoch": 7.53, "learning_rate": 7.4093422306959e-07, "loss": 1.7085, "step": 39546 }, { "epoch": 7.53, "learning_rate": 7.408770257387989e-07, "loss": 2.3495, "step": 39547 }, { "epoch": 7.53, "learning_rate": 7.408198284080076e-07, "loss": 1.7602, "step": 39548 }, { "epoch": 7.53, "learning_rate": 7.407626310772164e-07, "loss": 1.5067, "step": 39549 }, { "epoch": 7.53, "learning_rate": 7.407054337464252e-07, "loss": 2.2669, "step": 39550 }, { "epoch": 7.53, "learning_rate": 7.406482364156339e-07, "loss": 1.7719, "step": 39551 }, { "epoch": 7.53, "learning_rate": 7.405910390848428e-07, "loss": 1.5193, "step": 39552 }, { "epoch": 7.53, "learning_rate": 7.405338417540515e-07, "loss": 1.5214, "step": 39553 }, { "epoch": 7.53, "learning_rate": 7.404766444232603e-07, "loss": 1.0356, "step": 39554 }, { "epoch": 7.53, "learning_rate": 7.404194470924691e-07, "loss": 1.7226, "step": 39555 }, { "epoch": 7.53, "learning_rate": 7.403622497616778e-07, "loss": 1.7254, "step": 39556 }, { "epoch": 7.53, "learning_rate": 7.403050524308865e-07, "loss": 1.906, "step": 39557 }, { "epoch": 7.53, "learning_rate": 7.402478551000953e-07, "loss": 0.803, "step": 39558 }, { "epoch": 7.54, "learning_rate": 7.401906577693041e-07, "loss": 1.6037, "step": 39559 }, { "epoch": 7.54, "learning_rate": 7.401334604385129e-07, "loss": 2.2763, "step": 39560 }, { "epoch": 7.54, "learning_rate": 7.400762631077217e-07, "loss": 2.0175, "step": 39561 }, { "epoch": 7.54, "learning_rate": 7.400190657769304e-07, "loss": 1.2219, "step": 39562 }, { "epoch": 7.54, "learning_rate": 7.399618684461392e-07, "loss": 2.6408, "step": 39563 }, { "epoch": 7.54, "learning_rate": 7.39904671115348e-07, "loss": 2.1888, "step": 39564 }, { "epoch": 7.54, "learning_rate": 7.398474737845567e-07, "loss": 0.9086, "step": 39565 }, { "epoch": 7.54, "learning_rate": 7.397902764537656e-07, "loss": 1.4465, "step": 39566 }, { "epoch": 7.54, "learning_rate": 7.397330791229743e-07, "loss": 0.8204, "step": 39567 }, { "epoch": 7.54, "learning_rate": 7.39675881792183e-07, "loss": 1.7943, "step": 39568 }, { "epoch": 7.54, "learning_rate": 7.396186844613919e-07, "loss": 2.0778, "step": 39569 }, { "epoch": 7.54, "learning_rate": 7.395614871306005e-07, "loss": 2.8069, "step": 39570 }, { "epoch": 7.54, "learning_rate": 7.395042897998093e-07, "loss": 1.3413, "step": 39571 }, { "epoch": 7.54, "learning_rate": 7.394470924690181e-07, "loss": 1.8543, "step": 39572 }, { "epoch": 7.54, "learning_rate": 7.393898951382269e-07, "loss": 2.6421, "step": 39573 }, { "epoch": 7.54, "learning_rate": 7.393326978074357e-07, "loss": 1.2159, "step": 39574 }, { "epoch": 7.54, "learning_rate": 7.392755004766444e-07, "loss": 0.6047, "step": 39575 }, { "epoch": 7.54, "learning_rate": 7.392183031458532e-07, "loss": 1.1216, "step": 39576 }, { "epoch": 7.54, "learning_rate": 7.39161105815062e-07, "loss": 1.5792, "step": 39577 }, { "epoch": 7.54, "learning_rate": 7.391039084842708e-07, "loss": 2.6953, "step": 39578 }, { "epoch": 7.54, "learning_rate": 7.390467111534795e-07, "loss": 1.0548, "step": 39579 }, { "epoch": 7.54, "learning_rate": 7.389895138226883e-07, "loss": 1.9298, "step": 39580 }, { "epoch": 7.54, "learning_rate": 7.389323164918971e-07, "loss": 1.7165, "step": 39581 }, { "epoch": 7.54, "learning_rate": 7.388751191611059e-07, "loss": 2.1122, "step": 39582 }, { "epoch": 7.54, "learning_rate": 7.388179218303147e-07, "loss": 2.2528, "step": 39583 }, { "epoch": 7.54, "learning_rate": 7.387607244995233e-07, "loss": 2.1994, "step": 39584 }, { "epoch": 7.54, "learning_rate": 7.387035271687321e-07, "loss": 0.6352, "step": 39585 }, { "epoch": 7.54, "learning_rate": 7.386463298379409e-07, "loss": 2.1722, "step": 39586 }, { "epoch": 7.54, "learning_rate": 7.385891325071496e-07, "loss": 2.1876, "step": 39587 }, { "epoch": 7.54, "learning_rate": 7.385319351763585e-07, "loss": 1.6792, "step": 39588 }, { "epoch": 7.54, "learning_rate": 7.384747378455672e-07, "loss": 1.8128, "step": 39589 }, { "epoch": 7.54, "learning_rate": 7.38417540514776e-07, "loss": 1.8266, "step": 39590 }, { "epoch": 7.54, "learning_rate": 7.383603431839848e-07, "loss": 2.2369, "step": 39591 }, { "epoch": 7.54, "learning_rate": 7.383031458531935e-07, "loss": 1.1358, "step": 39592 }, { "epoch": 7.54, "learning_rate": 7.382459485224023e-07, "loss": 2.0985, "step": 39593 }, { "epoch": 7.54, "learning_rate": 7.381887511916111e-07, "loss": 0.8625, "step": 39594 }, { "epoch": 7.54, "learning_rate": 7.381315538608199e-07, "loss": 1.023, "step": 39595 }, { "epoch": 7.54, "learning_rate": 7.380743565300287e-07, "loss": 2.1413, "step": 39596 }, { "epoch": 7.54, "learning_rate": 7.380171591992375e-07, "loss": 2.698, "step": 39597 }, { "epoch": 7.54, "learning_rate": 7.379599618684461e-07, "loss": 2.1041, "step": 39598 }, { "epoch": 7.54, "learning_rate": 7.379027645376549e-07, "loss": 1.1308, "step": 39599 }, { "epoch": 7.54, "learning_rate": 7.378455672068637e-07, "loss": 2.1822, "step": 39600 }, { "epoch": 7.54, "learning_rate": 7.377883698760724e-07, "loss": 0.9285, "step": 39601 }, { "epoch": 7.54, "learning_rate": 7.377311725452813e-07, "loss": 1.1009, "step": 39602 }, { "epoch": 7.54, "learning_rate": 7.3767397521449e-07, "loss": 1.3263, "step": 39603 }, { "epoch": 7.54, "learning_rate": 7.376167778836987e-07, "loss": 2.509, "step": 39604 }, { "epoch": 7.54, "learning_rate": 7.375595805529076e-07, "loss": 1.5906, "step": 39605 }, { "epoch": 7.54, "learning_rate": 7.375023832221163e-07, "loss": 2.4352, "step": 39606 }, { "epoch": 7.54, "learning_rate": 7.374451858913251e-07, "loss": 1.3806, "step": 39607 }, { "epoch": 7.54, "learning_rate": 7.373879885605339e-07, "loss": 1.0749, "step": 39608 }, { "epoch": 7.54, "learning_rate": 7.373307912297426e-07, "loss": 2.7747, "step": 39609 }, { "epoch": 7.54, "learning_rate": 7.372735938989515e-07, "loss": 2.2854, "step": 39610 }, { "epoch": 7.54, "learning_rate": 7.372163965681601e-07, "loss": 1.3965, "step": 39611 }, { "epoch": 7.55, "learning_rate": 7.371591992373689e-07, "loss": 0.3167, "step": 39612 }, { "epoch": 7.55, "learning_rate": 7.371020019065777e-07, "loss": 1.1691, "step": 39613 }, { "epoch": 7.55, "learning_rate": 7.370448045757865e-07, "loss": 1.4643, "step": 39614 }, { "epoch": 7.55, "learning_rate": 7.369876072449952e-07, "loss": 1.4355, "step": 39615 }, { "epoch": 7.55, "learning_rate": 7.36930409914204e-07, "loss": 1.0212, "step": 39616 }, { "epoch": 7.55, "learning_rate": 7.368732125834128e-07, "loss": 0.7772, "step": 39617 }, { "epoch": 7.55, "learning_rate": 7.368160152526216e-07, "loss": 1.0292, "step": 39618 }, { "epoch": 7.55, "learning_rate": 7.367588179218304e-07, "loss": 1.6996, "step": 39619 }, { "epoch": 7.55, "learning_rate": 7.367016205910391e-07, "loss": 0.6518, "step": 39620 }, { "epoch": 7.55, "learning_rate": 7.366444232602479e-07, "loss": 1.4488, "step": 39621 }, { "epoch": 7.55, "learning_rate": 7.365872259294567e-07, "loss": 2.6582, "step": 39622 }, { "epoch": 7.55, "learning_rate": 7.365300285986654e-07, "loss": 2.8271, "step": 39623 }, { "epoch": 7.55, "learning_rate": 7.364728312678743e-07, "loss": 1.1276, "step": 39624 }, { "epoch": 7.55, "learning_rate": 7.364156339370829e-07, "loss": 1.871, "step": 39625 }, { "epoch": 7.55, "learning_rate": 7.363584366062917e-07, "loss": 3.0555, "step": 39626 }, { "epoch": 7.55, "learning_rate": 7.363012392755005e-07, "loss": 3.416, "step": 39627 }, { "epoch": 7.55, "learning_rate": 7.362440419447092e-07, "loss": 0.2708, "step": 39628 }, { "epoch": 7.55, "learning_rate": 7.36186844613918e-07, "loss": 2.6374, "step": 39629 }, { "epoch": 7.55, "learning_rate": 7.361296472831268e-07, "loss": 2.2331, "step": 39630 }, { "epoch": 7.55, "learning_rate": 7.360724499523356e-07, "loss": 1.0015, "step": 39631 }, { "epoch": 7.55, "learning_rate": 7.360152526215444e-07, "loss": 1.5129, "step": 39632 }, { "epoch": 7.55, "learning_rate": 7.359580552907531e-07, "loss": 0.9998, "step": 39633 }, { "epoch": 7.55, "learning_rate": 7.359008579599619e-07, "loss": 1.702, "step": 39634 }, { "epoch": 7.55, "learning_rate": 7.358436606291707e-07, "loss": 0.9229, "step": 39635 }, { "epoch": 7.55, "learning_rate": 7.357864632983795e-07, "loss": 2.6588, "step": 39636 }, { "epoch": 7.55, "learning_rate": 7.357292659675881e-07, "loss": 2.0587, "step": 39637 }, { "epoch": 7.55, "learning_rate": 7.35672068636797e-07, "loss": 2.3091, "step": 39638 }, { "epoch": 7.55, "learning_rate": 7.356148713060057e-07, "loss": 2.3892, "step": 39639 }, { "epoch": 7.55, "learning_rate": 7.355576739752145e-07, "loss": 2.1095, "step": 39640 }, { "epoch": 7.55, "learning_rate": 7.355004766444233e-07, "loss": 1.3838, "step": 39641 }, { "epoch": 7.55, "learning_rate": 7.35443279313632e-07, "loss": 0.9106, "step": 39642 }, { "epoch": 7.55, "learning_rate": 7.353860819828408e-07, "loss": 2.0209, "step": 39643 }, { "epoch": 7.55, "learning_rate": 7.353288846520496e-07, "loss": 1.9627, "step": 39644 }, { "epoch": 7.55, "learning_rate": 7.352716873212583e-07, "loss": 2.7879, "step": 39645 }, { "epoch": 7.55, "learning_rate": 7.352144899904672e-07, "loss": 0.9368, "step": 39646 }, { "epoch": 7.55, "learning_rate": 7.351572926596759e-07, "loss": 1.25, "step": 39647 }, { "epoch": 7.55, "learning_rate": 7.351000953288847e-07, "loss": 2.8948, "step": 39648 }, { "epoch": 7.55, "learning_rate": 7.350428979980935e-07, "loss": 1.9729, "step": 39649 }, { "epoch": 7.55, "learning_rate": 7.349857006673023e-07, "loss": 2.1854, "step": 39650 }, { "epoch": 7.55, "learning_rate": 7.349285033365109e-07, "loss": 0.9252, "step": 39651 }, { "epoch": 7.55, "learning_rate": 7.348713060057197e-07, "loss": 0.5546, "step": 39652 }, { "epoch": 7.55, "learning_rate": 7.348141086749285e-07, "loss": 1.401, "step": 39653 }, { "epoch": 7.55, "learning_rate": 7.347569113441373e-07, "loss": 0.6442, "step": 39654 }, { "epoch": 7.55, "learning_rate": 7.346997140133461e-07, "loss": 1.853, "step": 39655 }, { "epoch": 7.55, "learning_rate": 7.346425166825548e-07, "loss": 1.8935, "step": 39656 }, { "epoch": 7.55, "learning_rate": 7.345853193517636e-07, "loss": 1.2175, "step": 39657 }, { "epoch": 7.55, "learning_rate": 7.345281220209724e-07, "loss": 1.6037, "step": 39658 }, { "epoch": 7.55, "learning_rate": 7.344709246901811e-07, "loss": 1.0253, "step": 39659 }, { "epoch": 7.55, "learning_rate": 7.3441372735939e-07, "loss": 1.3289, "step": 39660 }, { "epoch": 7.55, "learning_rate": 7.343565300285987e-07, "loss": 1.3717, "step": 39661 }, { "epoch": 7.55, "learning_rate": 7.342993326978075e-07, "loss": 0.6279, "step": 39662 }, { "epoch": 7.55, "learning_rate": 7.342421353670163e-07, "loss": 2.6818, "step": 39663 }, { "epoch": 7.56, "learning_rate": 7.341849380362249e-07, "loss": 1.93, "step": 39664 }, { "epoch": 7.56, "learning_rate": 7.341277407054337e-07, "loss": 1.3135, "step": 39665 }, { "epoch": 7.56, "learning_rate": 7.340705433746425e-07, "loss": 0.4475, "step": 39666 }, { "epoch": 7.56, "learning_rate": 7.340133460438513e-07, "loss": 2.0211, "step": 39667 }, { "epoch": 7.56, "learning_rate": 7.339561487130601e-07, "loss": 2.2263, "step": 39668 }, { "epoch": 7.56, "learning_rate": 7.338989513822688e-07, "loss": 3.4597, "step": 39669 }, { "epoch": 7.56, "learning_rate": 7.338417540514776e-07, "loss": 0.7242, "step": 39670 }, { "epoch": 7.56, "learning_rate": 7.337845567206864e-07, "loss": 2.3202, "step": 39671 }, { "epoch": 7.56, "learning_rate": 7.337273593898952e-07, "loss": 1.9707, "step": 39672 }, { "epoch": 7.56, "learning_rate": 7.336701620591039e-07, "loss": 2.5043, "step": 39673 }, { "epoch": 7.56, "learning_rate": 7.336129647283127e-07, "loss": 2.4191, "step": 39674 }, { "epoch": 7.56, "learning_rate": 7.335557673975215e-07, "loss": 1.0743, "step": 39675 }, { "epoch": 7.56, "learning_rate": 7.334985700667303e-07, "loss": 1.9366, "step": 39676 }, { "epoch": 7.56, "learning_rate": 7.334413727359391e-07, "loss": 1.8892, "step": 39677 }, { "epoch": 7.56, "learning_rate": 7.333841754051477e-07, "loss": 1.0378, "step": 39678 }, { "epoch": 7.56, "learning_rate": 7.333269780743565e-07, "loss": 1.1423, "step": 39679 }, { "epoch": 7.56, "learning_rate": 7.332697807435653e-07, "loss": 1.1115, "step": 39680 }, { "epoch": 7.56, "learning_rate": 7.33212583412774e-07, "loss": 2.0631, "step": 39681 }, { "epoch": 7.56, "learning_rate": 7.331553860819829e-07, "loss": 2.0898, "step": 39682 }, { "epoch": 7.56, "learning_rate": 7.330981887511916e-07, "loss": 1.9809, "step": 39683 }, { "epoch": 7.56, "learning_rate": 7.330409914204004e-07, "loss": 1.6791, "step": 39684 }, { "epoch": 7.56, "learning_rate": 7.329837940896092e-07, "loss": 1.8616, "step": 39685 }, { "epoch": 7.56, "learning_rate": 7.329265967588179e-07, "loss": 1.1905, "step": 39686 }, { "epoch": 7.56, "learning_rate": 7.328693994280267e-07, "loss": 1.5384, "step": 39687 }, { "epoch": 7.56, "learning_rate": 7.328122020972355e-07, "loss": 0.9504, "step": 39688 }, { "epoch": 7.56, "learning_rate": 7.327550047664443e-07, "loss": 0.7128, "step": 39689 }, { "epoch": 7.56, "learning_rate": 7.326978074356531e-07, "loss": 0.4654, "step": 39690 }, { "epoch": 7.56, "learning_rate": 7.326406101048618e-07, "loss": 0.9461, "step": 39691 }, { "epoch": 7.56, "learning_rate": 7.325834127740705e-07, "loss": 1.2576, "step": 39692 }, { "epoch": 7.56, "learning_rate": 7.325262154432793e-07, "loss": 0.9561, "step": 39693 }, { "epoch": 7.56, "learning_rate": 7.324690181124881e-07, "loss": 2.0493, "step": 39694 }, { "epoch": 7.56, "learning_rate": 7.324118207816968e-07, "loss": 2.2741, "step": 39695 }, { "epoch": 7.56, "learning_rate": 7.323546234509057e-07, "loss": 1.4267, "step": 39696 }, { "epoch": 7.56, "learning_rate": 7.322974261201144e-07, "loss": 3.1205, "step": 39697 }, { "epoch": 7.56, "learning_rate": 7.322402287893232e-07, "loss": 2.6406, "step": 39698 }, { "epoch": 7.56, "learning_rate": 7.32183031458532e-07, "loss": 1.4086, "step": 39699 }, { "epoch": 7.56, "learning_rate": 7.321258341277407e-07, "loss": 4.0165, "step": 39700 }, { "epoch": 7.56, "learning_rate": 7.320686367969495e-07, "loss": 4.6265, "step": 39701 }, { "epoch": 7.56, "learning_rate": 7.320114394661583e-07, "loss": 1.9784, "step": 39702 }, { "epoch": 7.56, "learning_rate": 7.319542421353671e-07, "loss": 1.2622, "step": 39703 }, { "epoch": 7.56, "learning_rate": 7.318970448045759e-07, "loss": 0.2496, "step": 39704 }, { "epoch": 7.56, "learning_rate": 7.318398474737845e-07, "loss": 2.0576, "step": 39705 }, { "epoch": 7.56, "learning_rate": 7.317826501429933e-07, "loss": 1.0781, "step": 39706 }, { "epoch": 7.56, "learning_rate": 7.317254528122021e-07, "loss": 1.5401, "step": 39707 }, { "epoch": 7.56, "learning_rate": 7.316682554814109e-07, "loss": 0.8972, "step": 39708 }, { "epoch": 7.56, "learning_rate": 7.316110581506196e-07, "loss": 1.2733, "step": 39709 }, { "epoch": 7.56, "learning_rate": 7.315538608198284e-07, "loss": 1.8624, "step": 39710 }, { "epoch": 7.56, "learning_rate": 7.314966634890372e-07, "loss": 1.5938, "step": 39711 }, { "epoch": 7.56, "learning_rate": 7.31439466158246e-07, "loss": 1.6474, "step": 39712 }, { "epoch": 7.56, "learning_rate": 7.313822688274548e-07, "loss": 1.7419, "step": 39713 }, { "epoch": 7.56, "learning_rate": 7.313250714966635e-07, "loss": 1.4995, "step": 39714 }, { "epoch": 7.56, "learning_rate": 7.312678741658724e-07, "loss": 2.7475, "step": 39715 }, { "epoch": 7.56, "learning_rate": 7.312106768350811e-07, "loss": 1.4953, "step": 39716 }, { "epoch": 7.57, "learning_rate": 7.311534795042897e-07, "loss": 2.644, "step": 39717 }, { "epoch": 7.57, "learning_rate": 7.310962821734986e-07, "loss": 1.8814, "step": 39718 }, { "epoch": 7.57, "learning_rate": 7.310390848427073e-07, "loss": 1.8359, "step": 39719 }, { "epoch": 7.57, "learning_rate": 7.309818875119161e-07, "loss": 2.1618, "step": 39720 }, { "epoch": 7.57, "learning_rate": 7.309246901811249e-07, "loss": 2.4013, "step": 39721 }, { "epoch": 7.57, "learning_rate": 7.308674928503336e-07, "loss": 2.0264, "step": 39722 }, { "epoch": 7.57, "learning_rate": 7.308102955195424e-07, "loss": 1.7704, "step": 39723 }, { "epoch": 7.57, "learning_rate": 7.307530981887512e-07, "loss": 0.8997, "step": 39724 }, { "epoch": 7.57, "learning_rate": 7.3069590085796e-07, "loss": 2.5437, "step": 39725 }, { "epoch": 7.57, "learning_rate": 7.306387035271688e-07, "loss": 0.384, "step": 39726 }, { "epoch": 7.57, "learning_rate": 7.305815061963775e-07, "loss": 2.5555, "step": 39727 }, { "epoch": 7.57, "learning_rate": 7.305243088655863e-07, "loss": 2.124, "step": 39728 }, { "epoch": 7.57, "learning_rate": 7.304671115347951e-07, "loss": 2.4896, "step": 39729 }, { "epoch": 7.57, "learning_rate": 7.304099142040039e-07, "loss": 1.6767, "step": 39730 }, { "epoch": 7.57, "learning_rate": 7.303527168732125e-07, "loss": 0.6608, "step": 39731 }, { "epoch": 7.57, "learning_rate": 7.302955195424214e-07, "loss": 1.5103, "step": 39732 }, { "epoch": 7.57, "learning_rate": 7.302383222116301e-07, "loss": 1.0737, "step": 39733 }, { "epoch": 7.57, "learning_rate": 7.301811248808389e-07, "loss": 1.3792, "step": 39734 }, { "epoch": 7.57, "learning_rate": 7.301239275500477e-07, "loss": 1.9183, "step": 39735 }, { "epoch": 7.57, "learning_rate": 7.300667302192564e-07, "loss": 1.5683, "step": 39736 }, { "epoch": 7.57, "learning_rate": 7.300095328884653e-07, "loss": 0.5494, "step": 39737 }, { "epoch": 7.57, "learning_rate": 7.29952335557674e-07, "loss": 2.2024, "step": 39738 }, { "epoch": 7.57, "learning_rate": 7.298951382268827e-07, "loss": 0.9189, "step": 39739 }, { "epoch": 7.57, "learning_rate": 7.298379408960916e-07, "loss": 1.0856, "step": 39740 }, { "epoch": 7.57, "learning_rate": 7.297807435653003e-07, "loss": 1.8576, "step": 39741 }, { "epoch": 7.57, "learning_rate": 7.297235462345091e-07, "loss": 2.0792, "step": 39742 }, { "epoch": 7.57, "learning_rate": 7.296663489037179e-07, "loss": 1.771, "step": 39743 }, { "epoch": 7.57, "learning_rate": 7.296091515729267e-07, "loss": 2.5666, "step": 39744 }, { "epoch": 7.57, "learning_rate": 7.295519542421353e-07, "loss": 0.715, "step": 39745 }, { "epoch": 7.57, "learning_rate": 7.294947569113441e-07, "loss": 0.5917, "step": 39746 }, { "epoch": 7.57, "learning_rate": 7.294375595805529e-07, "loss": 1.6043, "step": 39747 }, { "epoch": 7.57, "learning_rate": 7.293803622497617e-07, "loss": 1.0013, "step": 39748 }, { "epoch": 7.57, "learning_rate": 7.293231649189705e-07, "loss": 0.8041, "step": 39749 }, { "epoch": 7.57, "learning_rate": 7.292659675881792e-07, "loss": 2.6421, "step": 39750 }, { "epoch": 7.57, "learning_rate": 7.29208770257388e-07, "loss": 2.4517, "step": 39751 }, { "epoch": 7.57, "learning_rate": 7.291515729265968e-07, "loss": 1.0763, "step": 39752 }, { "epoch": 7.57, "learning_rate": 7.290943755958055e-07, "loss": 1.5729, "step": 39753 }, { "epoch": 7.57, "learning_rate": 7.290371782650144e-07, "loss": 1.3519, "step": 39754 }, { "epoch": 7.57, "learning_rate": 7.289799809342231e-07, "loss": 2.7628, "step": 39755 }, { "epoch": 7.57, "learning_rate": 7.289227836034319e-07, "loss": 0.3871, "step": 39756 }, { "epoch": 7.57, "learning_rate": 7.288655862726407e-07, "loss": 2.8686, "step": 39757 }, { "epoch": 7.57, "learning_rate": 7.288083889418493e-07, "loss": 1.8734, "step": 39758 }, { "epoch": 7.57, "learning_rate": 7.287511916110582e-07, "loss": 1.6707, "step": 39759 }, { "epoch": 7.57, "learning_rate": 7.286939942802669e-07, "loss": 2.0565, "step": 39760 }, { "epoch": 7.57, "learning_rate": 7.286367969494757e-07, "loss": 0.8626, "step": 39761 }, { "epoch": 7.57, "learning_rate": 7.285795996186845e-07, "loss": 1.5115, "step": 39762 }, { "epoch": 7.57, "learning_rate": 7.285224022878932e-07, "loss": 3.7595, "step": 39763 }, { "epoch": 7.57, "learning_rate": 7.28465204957102e-07, "loss": 2.1154, "step": 39764 }, { "epoch": 7.57, "learning_rate": 7.284080076263108e-07, "loss": 1.5333, "step": 39765 }, { "epoch": 7.57, "learning_rate": 7.283508102955196e-07, "loss": 0.7504, "step": 39766 }, { "epoch": 7.57, "learning_rate": 7.282936129647283e-07, "loss": 2.3356, "step": 39767 }, { "epoch": 7.57, "learning_rate": 7.282364156339372e-07, "loss": 1.9947, "step": 39768 }, { "epoch": 7.58, "learning_rate": 7.281792183031459e-07, "loss": 1.8823, "step": 39769 }, { "epoch": 7.58, "learning_rate": 7.281220209723547e-07, "loss": 3.476, "step": 39770 }, { "epoch": 7.58, "learning_rate": 7.280648236415634e-07, "loss": 1.7266, "step": 39771 }, { "epoch": 7.58, "learning_rate": 7.280076263107721e-07, "loss": 1.2469, "step": 39772 }, { "epoch": 7.58, "learning_rate": 7.27950428979981e-07, "loss": 1.1822, "step": 39773 }, { "epoch": 7.58, "learning_rate": 7.278932316491897e-07, "loss": 0.9862, "step": 39774 }, { "epoch": 7.58, "learning_rate": 7.278360343183984e-07, "loss": 3.2874, "step": 39775 }, { "epoch": 7.58, "learning_rate": 7.277788369876073e-07, "loss": 1.6531, "step": 39776 }, { "epoch": 7.58, "learning_rate": 7.27721639656816e-07, "loss": 1.3764, "step": 39777 }, { "epoch": 7.58, "learning_rate": 7.276644423260248e-07, "loss": 1.0437, "step": 39778 }, { "epoch": 7.58, "learning_rate": 7.276072449952336e-07, "loss": 1.2269, "step": 39779 }, { "epoch": 7.58, "learning_rate": 7.275500476644424e-07, "loss": 1.7539, "step": 39780 }, { "epoch": 7.58, "learning_rate": 7.274928503336511e-07, "loss": 1.484, "step": 39781 }, { "epoch": 7.58, "learning_rate": 7.274356530028599e-07, "loss": 0.7809, "step": 39782 }, { "epoch": 7.58, "learning_rate": 7.273784556720687e-07, "loss": 1.3799, "step": 39783 }, { "epoch": 7.58, "learning_rate": 7.273212583412775e-07, "loss": 2.9864, "step": 39784 }, { "epoch": 7.58, "learning_rate": 7.272640610104862e-07, "loss": 1.4639, "step": 39785 }, { "epoch": 7.58, "learning_rate": 7.272068636796949e-07, "loss": 1.561, "step": 39786 }, { "epoch": 7.58, "learning_rate": 7.271496663489037e-07, "loss": 1.4486, "step": 39787 }, { "epoch": 7.58, "learning_rate": 7.270924690181125e-07, "loss": 1.5457, "step": 39788 }, { "epoch": 7.58, "learning_rate": 7.270352716873212e-07, "loss": 1.7762, "step": 39789 }, { "epoch": 7.58, "learning_rate": 7.269780743565301e-07, "loss": 1.8457, "step": 39790 }, { "epoch": 7.58, "learning_rate": 7.269208770257388e-07, "loss": 0.4347, "step": 39791 }, { "epoch": 7.58, "learning_rate": 7.268636796949476e-07, "loss": 1.9908, "step": 39792 }, { "epoch": 7.58, "learning_rate": 7.268064823641564e-07, "loss": 3.597, "step": 39793 }, { "epoch": 7.58, "learning_rate": 7.267492850333651e-07, "loss": 2.5416, "step": 39794 }, { "epoch": 7.58, "learning_rate": 7.26692087702574e-07, "loss": 1.653, "step": 39795 }, { "epoch": 7.58, "learning_rate": 7.266348903717827e-07, "loss": 1.0196, "step": 39796 }, { "epoch": 7.58, "learning_rate": 7.265776930409915e-07, "loss": 0.948, "step": 39797 }, { "epoch": 7.58, "learning_rate": 7.265204957102002e-07, "loss": 1.9187, "step": 39798 }, { "epoch": 7.58, "learning_rate": 7.264632983794089e-07, "loss": 3.1929, "step": 39799 }, { "epoch": 7.58, "learning_rate": 7.264061010486177e-07, "loss": 1.5566, "step": 39800 }, { "epoch": 7.58, "learning_rate": 7.263489037178265e-07, "loss": 1.4384, "step": 39801 }, { "epoch": 7.58, "learning_rate": 7.262917063870353e-07, "loss": 0.9185, "step": 39802 }, { "epoch": 7.58, "learning_rate": 7.26234509056244e-07, "loss": 1.3628, "step": 39803 }, { "epoch": 7.58, "learning_rate": 7.261773117254528e-07, "loss": 1.6257, "step": 39804 }, { "epoch": 7.58, "learning_rate": 7.261201143946616e-07, "loss": 2.5409, "step": 39805 }, { "epoch": 7.58, "learning_rate": 7.260629170638704e-07, "loss": 0.7987, "step": 39806 }, { "epoch": 7.58, "learning_rate": 7.260057197330792e-07, "loss": 1.866, "step": 39807 }, { "epoch": 7.58, "learning_rate": 7.259485224022879e-07, "loss": 0.4096, "step": 39808 }, { "epoch": 7.58, "learning_rate": 7.258913250714968e-07, "loss": 1.1592, "step": 39809 }, { "epoch": 7.58, "learning_rate": 7.258341277407055e-07, "loss": 0.8417, "step": 39810 }, { "epoch": 7.58, "learning_rate": 7.257769304099141e-07, "loss": 2.178, "step": 39811 }, { "epoch": 7.58, "learning_rate": 7.25719733079123e-07, "loss": 1.7296, "step": 39812 }, { "epoch": 7.58, "learning_rate": 7.256625357483317e-07, "loss": 2.2143, "step": 39813 }, { "epoch": 7.58, "learning_rate": 7.256053384175405e-07, "loss": 0.6327, "step": 39814 }, { "epoch": 7.58, "learning_rate": 7.255481410867493e-07, "loss": 2.0719, "step": 39815 }, { "epoch": 7.58, "learning_rate": 7.25490943755958e-07, "loss": 1.053, "step": 39816 }, { "epoch": 7.58, "learning_rate": 7.254337464251669e-07, "loss": 0.4542, "step": 39817 }, { "epoch": 7.58, "learning_rate": 7.253765490943756e-07, "loss": 2.2558, "step": 39818 }, { "epoch": 7.58, "learning_rate": 7.253193517635844e-07, "loss": 0.6418, "step": 39819 }, { "epoch": 7.58, "learning_rate": 7.252621544327932e-07, "loss": 1.5704, "step": 39820 }, { "epoch": 7.58, "learning_rate": 7.25204957102002e-07, "loss": 0.4704, "step": 39821 }, { "epoch": 7.59, "learning_rate": 7.251477597712107e-07, "loss": 1.7339, "step": 39822 }, { "epoch": 7.59, "learning_rate": 7.250905624404195e-07, "loss": 1.7115, "step": 39823 }, { "epoch": 7.59, "learning_rate": 7.250333651096283e-07, "loss": 1.5966, "step": 39824 }, { "epoch": 7.59, "learning_rate": 7.249761677788369e-07, "loss": 1.0379, "step": 39825 }, { "epoch": 7.59, "learning_rate": 7.249189704480458e-07, "loss": 1.0985, "step": 39826 }, { "epoch": 7.59, "learning_rate": 7.248617731172545e-07, "loss": 1.4201, "step": 39827 }, { "epoch": 7.59, "learning_rate": 7.248045757864633e-07, "loss": 1.265, "step": 39828 }, { "epoch": 7.59, "learning_rate": 7.247473784556721e-07, "loss": 1.658, "step": 39829 }, { "epoch": 7.59, "learning_rate": 7.246901811248808e-07, "loss": 1.4444, "step": 39830 }, { "epoch": 7.59, "learning_rate": 7.246329837940897e-07, "loss": 1.3075, "step": 39831 }, { "epoch": 7.59, "learning_rate": 7.245757864632984e-07, "loss": 1.9986, "step": 39832 }, { "epoch": 7.59, "learning_rate": 7.245185891325072e-07, "loss": 1.6193, "step": 39833 }, { "epoch": 7.59, "learning_rate": 7.24461391801716e-07, "loss": 2.2896, "step": 39834 }, { "epoch": 7.59, "learning_rate": 7.244041944709247e-07, "loss": 0.6812, "step": 39835 }, { "epoch": 7.59, "learning_rate": 7.243469971401335e-07, "loss": 1.0591, "step": 39836 }, { "epoch": 7.59, "learning_rate": 7.242897998093423e-07, "loss": 2.4146, "step": 39837 }, { "epoch": 7.59, "learning_rate": 7.24232602478551e-07, "loss": 1.2524, "step": 39838 }, { "epoch": 7.59, "learning_rate": 7.241754051477598e-07, "loss": 2.2672, "step": 39839 }, { "epoch": 7.59, "learning_rate": 7.241182078169685e-07, "loss": 2.58, "step": 39840 }, { "epoch": 7.59, "learning_rate": 7.240610104861773e-07, "loss": 1.6805, "step": 39841 }, { "epoch": 7.59, "learning_rate": 7.240038131553861e-07, "loss": 1.7345, "step": 39842 }, { "epoch": 7.59, "learning_rate": 7.239466158245949e-07, "loss": 0.4422, "step": 39843 }, { "epoch": 7.59, "learning_rate": 7.238894184938036e-07, "loss": 1.5047, "step": 39844 }, { "epoch": 7.59, "learning_rate": 7.238322211630124e-07, "loss": 0.5894, "step": 39845 }, { "epoch": 7.59, "learning_rate": 7.237750238322212e-07, "loss": 3.3884, "step": 39846 }, { "epoch": 7.59, "learning_rate": 7.237178265014299e-07, "loss": 1.8112, "step": 39847 }, { "epoch": 7.59, "learning_rate": 7.236606291706388e-07, "loss": 1.7376, "step": 39848 }, { "epoch": 7.59, "learning_rate": 7.236034318398475e-07, "loss": 1.4176, "step": 39849 }, { "epoch": 7.59, "learning_rate": 7.235462345090563e-07, "loss": 2.7854, "step": 39850 }, { "epoch": 7.59, "learning_rate": 7.23489037178265e-07, "loss": 2.4533, "step": 39851 }, { "epoch": 7.59, "learning_rate": 7.234318398474737e-07, "loss": 1.5922, "step": 39852 }, { "epoch": 7.59, "learning_rate": 7.233746425166826e-07, "loss": 1.5498, "step": 39853 }, { "epoch": 7.59, "learning_rate": 7.233174451858913e-07, "loss": 0.8325, "step": 39854 }, { "epoch": 7.59, "learning_rate": 7.232602478551001e-07, "loss": 1.1512, "step": 39855 }, { "epoch": 7.59, "learning_rate": 7.232030505243089e-07, "loss": 1.8134, "step": 39856 }, { "epoch": 7.59, "learning_rate": 7.231458531935176e-07, "loss": 1.4426, "step": 39857 }, { "epoch": 7.59, "learning_rate": 7.230886558627264e-07, "loss": 1.1493, "step": 39858 }, { "epoch": 7.59, "learning_rate": 7.230314585319352e-07, "loss": 1.585, "step": 39859 }, { "epoch": 7.59, "learning_rate": 7.22974261201144e-07, "loss": 2.9062, "step": 39860 }, { "epoch": 7.59, "learning_rate": 7.229170638703528e-07, "loss": 2.083, "step": 39861 }, { "epoch": 7.59, "learning_rate": 7.228598665395616e-07, "loss": 2.0365, "step": 39862 }, { "epoch": 7.59, "learning_rate": 7.228026692087703e-07, "loss": 0.5334, "step": 39863 }, { "epoch": 7.59, "learning_rate": 7.22745471877979e-07, "loss": 0.9759, "step": 39864 }, { "epoch": 7.59, "learning_rate": 7.226882745471878e-07, "loss": 1.3152, "step": 39865 }, { "epoch": 7.59, "learning_rate": 7.226310772163965e-07, "loss": 2.0457, "step": 39866 }, { "epoch": 7.59, "learning_rate": 7.225738798856054e-07, "loss": 2.0077, "step": 39867 }, { "epoch": 7.59, "learning_rate": 7.225166825548141e-07, "loss": 1.0436, "step": 39868 }, { "epoch": 7.59, "learning_rate": 7.224594852240228e-07, "loss": 1.521, "step": 39869 }, { "epoch": 7.59, "learning_rate": 7.224022878932317e-07, "loss": 0.8351, "step": 39870 }, { "epoch": 7.59, "learning_rate": 7.223450905624404e-07, "loss": 1.7269, "step": 39871 }, { "epoch": 7.59, "learning_rate": 7.222878932316492e-07, "loss": 1.3882, "step": 39872 }, { "epoch": 7.59, "learning_rate": 7.22230695900858e-07, "loss": 1.8872, "step": 39873 }, { "epoch": 7.6, "learning_rate": 7.221734985700668e-07, "loss": 1.4733, "step": 39874 }, { "epoch": 7.6, "learning_rate": 7.221163012392756e-07, "loss": 0.7546, "step": 39875 }, { "epoch": 7.6, "learning_rate": 7.220591039084843e-07, "loss": 2.4639, "step": 39876 }, { "epoch": 7.6, "learning_rate": 7.22001906577693e-07, "loss": 0.6696, "step": 39877 }, { "epoch": 7.6, "learning_rate": 7.219447092469018e-07, "loss": 4.3592, "step": 39878 }, { "epoch": 7.6, "learning_rate": 7.218875119161106e-07, "loss": 1.5823, "step": 39879 }, { "epoch": 7.6, "learning_rate": 7.218303145853193e-07, "loss": 3.2739, "step": 39880 }, { "epoch": 7.6, "learning_rate": 7.217731172545281e-07, "loss": 1.1615, "step": 39881 }, { "epoch": 7.6, "learning_rate": 7.217159199237369e-07, "loss": 2.6471, "step": 39882 }, { "epoch": 7.6, "learning_rate": 7.216587225929457e-07, "loss": 0.8441, "step": 39883 }, { "epoch": 7.6, "learning_rate": 7.216015252621545e-07, "loss": 1.2774, "step": 39884 }, { "epoch": 7.6, "learning_rate": 7.215443279313632e-07, "loss": 1.099, "step": 39885 }, { "epoch": 7.6, "learning_rate": 7.21487130600572e-07, "loss": 1.197, "step": 39886 }, { "epoch": 7.6, "learning_rate": 7.214299332697808e-07, "loss": 2.4651, "step": 39887 }, { "epoch": 7.6, "learning_rate": 7.213727359389895e-07, "loss": 2.5763, "step": 39888 }, { "epoch": 7.6, "learning_rate": 7.213155386081984e-07, "loss": 1.9768, "step": 39889 }, { "epoch": 7.6, "learning_rate": 7.212583412774071e-07, "loss": 0.6892, "step": 39890 }, { "epoch": 7.6, "learning_rate": 7.212011439466158e-07, "loss": 1.1972, "step": 39891 }, { "epoch": 7.6, "learning_rate": 7.211439466158246e-07, "loss": 1.4055, "step": 39892 }, { "epoch": 7.6, "learning_rate": 7.210867492850333e-07, "loss": 0.7973, "step": 39893 }, { "epoch": 7.6, "learning_rate": 7.210295519542421e-07, "loss": 2.5841, "step": 39894 }, { "epoch": 7.6, "learning_rate": 7.209723546234509e-07, "loss": 0.5091, "step": 39895 }, { "epoch": 7.6, "learning_rate": 7.209151572926597e-07, "loss": 0.6794, "step": 39896 }, { "epoch": 7.6, "learning_rate": 7.208579599618685e-07, "loss": 2.7646, "step": 39897 }, { "epoch": 7.6, "learning_rate": 7.208007626310773e-07, "loss": 2.0325, "step": 39898 }, { "epoch": 7.6, "learning_rate": 7.20743565300286e-07, "loss": 1.2048, "step": 39899 }, { "epoch": 7.6, "learning_rate": 7.206863679694948e-07, "loss": 1.0642, "step": 39900 }, { "epoch": 7.6, "learning_rate": 7.206291706387036e-07, "loss": 0.8597, "step": 39901 }, { "epoch": 7.6, "learning_rate": 7.205719733079123e-07, "loss": 1.986, "step": 39902 }, { "epoch": 7.6, "learning_rate": 7.205147759771212e-07, "loss": 1.6856, "step": 39903 }, { "epoch": 7.6, "learning_rate": 7.204575786463298e-07, "loss": 2.3115, "step": 39904 }, { "epoch": 7.6, "learning_rate": 7.204003813155386e-07, "loss": 0.6817, "step": 39905 }, { "epoch": 7.6, "learning_rate": 7.203431839847474e-07, "loss": 1.9261, "step": 39906 }, { "epoch": 7.6, "learning_rate": 7.202859866539561e-07, "loss": 2.3405, "step": 39907 }, { "epoch": 7.6, "learning_rate": 7.202287893231649e-07, "loss": 0.742, "step": 39908 }, { "epoch": 7.6, "learning_rate": 7.201715919923737e-07, "loss": 1.6397, "step": 39909 }, { "epoch": 7.6, "learning_rate": 7.201143946615824e-07, "loss": 0.6643, "step": 39910 }, { "epoch": 7.6, "learning_rate": 7.200571973307913e-07, "loss": 2.2405, "step": 39911 }, { "epoch": 7.6, "learning_rate": 7.2e-07, "loss": 1.8379, "step": 39912 }, { "epoch": 7.6, "learning_rate": 7.199428026692088e-07, "loss": 1.5046, "step": 39913 }, { "epoch": 7.6, "learning_rate": 7.198856053384176e-07, "loss": 2.6665, "step": 39914 }, { "epoch": 7.6, "learning_rate": 7.198284080076264e-07, "loss": 1.6163, "step": 39915 }, { "epoch": 7.6, "learning_rate": 7.197712106768351e-07, "loss": 1.5082, "step": 39916 }, { "epoch": 7.6, "learning_rate": 7.197140133460439e-07, "loss": 1.6181, "step": 39917 }, { "epoch": 7.6, "learning_rate": 7.196568160152526e-07, "loss": 0.9994, "step": 39918 }, { "epoch": 7.6, "learning_rate": 7.195996186844614e-07, "loss": 1.8404, "step": 39919 }, { "epoch": 7.6, "learning_rate": 7.195424213536702e-07, "loss": 0.8025, "step": 39920 }, { "epoch": 7.6, "learning_rate": 7.194852240228789e-07, "loss": 0.5488, "step": 39921 }, { "epoch": 7.6, "learning_rate": 7.194280266920877e-07, "loss": 2.0784, "step": 39922 }, { "epoch": 7.6, "learning_rate": 7.193708293612965e-07, "loss": 1.8195, "step": 39923 }, { "epoch": 7.6, "learning_rate": 7.193136320305052e-07, "loss": 1.8303, "step": 39924 }, { "epoch": 7.6, "learning_rate": 7.192564346997141e-07, "loss": 1.0428, "step": 39925 }, { "epoch": 7.6, "learning_rate": 7.191992373689228e-07, "loss": 0.7605, "step": 39926 }, { "epoch": 7.61, "learning_rate": 7.191420400381316e-07, "loss": 1.3377, "step": 39927 }, { "epoch": 7.61, "learning_rate": 7.190848427073404e-07, "loss": 0.7115, "step": 39928 }, { "epoch": 7.61, "learning_rate": 7.190276453765491e-07, "loss": 1.7997, "step": 39929 }, { "epoch": 7.61, "learning_rate": 7.189704480457579e-07, "loss": 1.7182, "step": 39930 }, { "epoch": 7.61, "learning_rate": 7.189132507149666e-07, "loss": 1.139, "step": 39931 }, { "epoch": 7.61, "learning_rate": 7.188560533841754e-07, "loss": 0.4031, "step": 39932 }, { "epoch": 7.61, "learning_rate": 7.187988560533842e-07, "loss": 0.9501, "step": 39933 }, { "epoch": 7.61, "learning_rate": 7.187416587225929e-07, "loss": 1.6819, "step": 39934 }, { "epoch": 7.61, "learning_rate": 7.186844613918017e-07, "loss": 0.7766, "step": 39935 }, { "epoch": 7.61, "learning_rate": 7.186272640610105e-07, "loss": 0.9972, "step": 39936 }, { "epoch": 7.61, "learning_rate": 7.185700667302193e-07, "loss": 1.3739, "step": 39937 }, { "epoch": 7.61, "learning_rate": 7.18512869399428e-07, "loss": 0.9762, "step": 39938 }, { "epoch": 7.61, "learning_rate": 7.184556720686369e-07, "loss": 1.028, "step": 39939 }, { "epoch": 7.61, "learning_rate": 7.183984747378456e-07, "loss": 0.9909, "step": 39940 }, { "epoch": 7.61, "learning_rate": 7.183412774070544e-07, "loss": 0.9733, "step": 39941 }, { "epoch": 7.61, "learning_rate": 7.182840800762632e-07, "loss": 2.4611, "step": 39942 }, { "epoch": 7.61, "learning_rate": 7.182268827454719e-07, "loss": 1.5222, "step": 39943 }, { "epoch": 7.61, "learning_rate": 7.181696854146806e-07, "loss": 1.5565, "step": 39944 }, { "epoch": 7.61, "learning_rate": 7.181124880838894e-07, "loss": 3.6654, "step": 39945 }, { "epoch": 7.61, "learning_rate": 7.180552907530981e-07, "loss": 2.1521, "step": 39946 }, { "epoch": 7.61, "learning_rate": 7.17998093422307e-07, "loss": 2.0779, "step": 39947 }, { "epoch": 7.61, "learning_rate": 7.179408960915157e-07, "loss": 1.8635, "step": 39948 }, { "epoch": 7.61, "learning_rate": 7.178836987607245e-07, "loss": 1.9239, "step": 39949 }, { "epoch": 7.61, "learning_rate": 7.178265014299333e-07, "loss": 0.4952, "step": 39950 }, { "epoch": 7.61, "learning_rate": 7.177693040991421e-07, "loss": 2.0247, "step": 39951 }, { "epoch": 7.61, "learning_rate": 7.177121067683508e-07, "loss": 1.7404, "step": 39952 }, { "epoch": 7.61, "learning_rate": 7.176549094375596e-07, "loss": 1.4464, "step": 39953 }, { "epoch": 7.61, "learning_rate": 7.175977121067684e-07, "loss": 0.816, "step": 39954 }, { "epoch": 7.61, "learning_rate": 7.175405147759772e-07, "loss": 1.3617, "step": 39955 }, { "epoch": 7.61, "learning_rate": 7.17483317445186e-07, "loss": 1.1765, "step": 39956 }, { "epoch": 7.61, "learning_rate": 7.174261201143947e-07, "loss": 1.2245, "step": 39957 }, { "epoch": 7.61, "learning_rate": 7.173689227836034e-07, "loss": 0.5876, "step": 39958 }, { "epoch": 7.61, "learning_rate": 7.173117254528122e-07, "loss": 0.959, "step": 39959 }, { "epoch": 7.61, "learning_rate": 7.172545281220209e-07, "loss": 1.3332, "step": 39960 }, { "epoch": 7.61, "learning_rate": 7.171973307912298e-07, "loss": 1.255, "step": 39961 }, { "epoch": 7.61, "learning_rate": 7.171401334604385e-07, "loss": 0.7475, "step": 39962 }, { "epoch": 7.61, "learning_rate": 7.170829361296473e-07, "loss": 1.3332, "step": 39963 }, { "epoch": 7.61, "learning_rate": 7.170257387988561e-07, "loss": 1.5365, "step": 39964 }, { "epoch": 7.61, "learning_rate": 7.169685414680648e-07, "loss": 2.1937, "step": 39965 }, { "epoch": 7.61, "learning_rate": 7.169113441372736e-07, "loss": 1.585, "step": 39966 }, { "epoch": 7.61, "learning_rate": 7.168541468064824e-07, "loss": 1.5933, "step": 39967 }, { "epoch": 7.61, "learning_rate": 7.167969494756912e-07, "loss": 2.185, "step": 39968 }, { "epoch": 7.61, "learning_rate": 7.167397521449e-07, "loss": 2.4492, "step": 39969 }, { "epoch": 7.61, "learning_rate": 7.166825548141087e-07, "loss": 1.0959, "step": 39970 }, { "epoch": 7.61, "learning_rate": 7.166253574833174e-07, "loss": 0.9822, "step": 39971 }, { "epoch": 7.61, "learning_rate": 7.165681601525262e-07, "loss": 2.5612, "step": 39972 }, { "epoch": 7.61, "learning_rate": 7.16510962821735e-07, "loss": 2.1929, "step": 39973 }, { "epoch": 7.61, "learning_rate": 7.164537654909437e-07, "loss": 0.4529, "step": 39974 }, { "epoch": 7.61, "learning_rate": 7.163965681601525e-07, "loss": 3.3991, "step": 39975 }, { "epoch": 7.61, "learning_rate": 7.163393708293613e-07, "loss": 0.7973, "step": 39976 }, { "epoch": 7.61, "learning_rate": 7.162821734985701e-07, "loss": 1.6503, "step": 39977 }, { "epoch": 7.61, "learning_rate": 7.162249761677789e-07, "loss": 1.0189, "step": 39978 }, { "epoch": 7.62, "learning_rate": 7.161677788369876e-07, "loss": 0.2717, "step": 39979 }, { "epoch": 7.62, "learning_rate": 7.161105815061964e-07, "loss": 2.8446, "step": 39980 }, { "epoch": 7.62, "learning_rate": 7.160533841754052e-07, "loss": 1.0179, "step": 39981 }, { "epoch": 7.62, "learning_rate": 7.159961868446139e-07, "loss": 1.4607, "step": 39982 }, { "epoch": 7.62, "learning_rate": 7.159389895138228e-07, "loss": 2.0804, "step": 39983 }, { "epoch": 7.62, "learning_rate": 7.158817921830314e-07, "loss": 1.0011, "step": 39984 }, { "epoch": 7.62, "learning_rate": 7.158245948522402e-07, "loss": 0.8589, "step": 39985 }, { "epoch": 7.62, "learning_rate": 7.15767397521449e-07, "loss": 2.1233, "step": 39986 }, { "epoch": 7.62, "learning_rate": 7.157102001906577e-07, "loss": 1.3587, "step": 39987 }, { "epoch": 7.62, "learning_rate": 7.156530028598665e-07, "loss": 2.699, "step": 39988 }, { "epoch": 7.62, "learning_rate": 7.155958055290753e-07, "loss": 1.5261, "step": 39989 }, { "epoch": 7.62, "learning_rate": 7.155386081982841e-07, "loss": 1.2889, "step": 39990 }, { "epoch": 7.62, "learning_rate": 7.154814108674929e-07, "loss": 1.2606, "step": 39991 }, { "epoch": 7.62, "learning_rate": 7.154242135367017e-07, "loss": 2.3976, "step": 39992 }, { "epoch": 7.62, "learning_rate": 7.153670162059104e-07, "loss": 1.6408, "step": 39993 }, { "epoch": 7.62, "learning_rate": 7.153098188751192e-07, "loss": 1.6249, "step": 39994 }, { "epoch": 7.62, "learning_rate": 7.15252621544328e-07, "loss": 2.3099, "step": 39995 }, { "epoch": 7.62, "learning_rate": 7.151954242135367e-07, "loss": 0.3446, "step": 39996 }, { "epoch": 7.62, "learning_rate": 7.151382268827456e-07, "loss": 1.588, "step": 39997 }, { "epoch": 7.62, "learning_rate": 7.150810295519542e-07, "loss": 1.6179, "step": 39998 }, { "epoch": 7.62, "learning_rate": 7.15023832221163e-07, "loss": 1.2581, "step": 39999 }, { "epoch": 7.62, "learning_rate": 7.149666348903718e-07, "loss": 0.3594, "step": 40000 }, { "epoch": 7.62, "learning_rate": 7.149094375595805e-07, "loss": 0.814, "step": 40001 }, { "epoch": 7.62, "learning_rate": 7.148522402287893e-07, "loss": 0.7113, "step": 40002 }, { "epoch": 7.62, "learning_rate": 7.147950428979981e-07, "loss": 1.4044, "step": 40003 }, { "epoch": 7.62, "learning_rate": 7.147378455672069e-07, "loss": 1.5928, "step": 40004 }, { "epoch": 7.62, "learning_rate": 7.146806482364157e-07, "loss": 1.2894, "step": 40005 }, { "epoch": 7.62, "learning_rate": 7.146234509056244e-07, "loss": 2.4081, "step": 40006 }, { "epoch": 7.62, "learning_rate": 7.145662535748332e-07, "loss": 2.3772, "step": 40007 }, { "epoch": 7.62, "learning_rate": 7.14509056244042e-07, "loss": 1.2258, "step": 40008 }, { "epoch": 7.62, "learning_rate": 7.144518589132508e-07, "loss": 1.9184, "step": 40009 }, { "epoch": 7.62, "learning_rate": 7.143946615824595e-07, "loss": 2.7341, "step": 40010 }, { "epoch": 7.62, "learning_rate": 7.143374642516682e-07, "loss": 2.2679, "step": 40011 }, { "epoch": 7.62, "learning_rate": 7.14280266920877e-07, "loss": 0.9445, "step": 40012 }, { "epoch": 7.62, "learning_rate": 7.142230695900858e-07, "loss": 1.4013, "step": 40013 }, { "epoch": 7.62, "learning_rate": 7.141658722592946e-07, "loss": 2.0835, "step": 40014 }, { "epoch": 7.62, "learning_rate": 7.141086749285033e-07, "loss": 1.718, "step": 40015 }, { "epoch": 7.62, "learning_rate": 7.140514775977122e-07, "loss": 0.713, "step": 40016 }, { "epoch": 7.62, "learning_rate": 7.139942802669209e-07, "loss": 1.1832, "step": 40017 }, { "epoch": 7.62, "learning_rate": 7.139370829361296e-07, "loss": 0.831, "step": 40018 }, { "epoch": 7.62, "learning_rate": 7.138798856053385e-07, "loss": 1.8787, "step": 40019 }, { "epoch": 7.62, "learning_rate": 7.138226882745472e-07, "loss": 0.2538, "step": 40020 }, { "epoch": 7.62, "learning_rate": 7.13765490943756e-07, "loss": 2.4197, "step": 40021 }, { "epoch": 7.62, "learning_rate": 7.137082936129648e-07, "loss": 1.2685, "step": 40022 }, { "epoch": 7.62, "learning_rate": 7.136510962821735e-07, "loss": 2.2582, "step": 40023 }, { "epoch": 7.62, "learning_rate": 7.135938989513822e-07, "loss": 1.5489, "step": 40024 }, { "epoch": 7.62, "learning_rate": 7.13536701620591e-07, "loss": 3.8814, "step": 40025 }, { "epoch": 7.62, "learning_rate": 7.134795042897998e-07, "loss": 2.6025, "step": 40026 }, { "epoch": 7.62, "learning_rate": 7.134223069590086e-07, "loss": 1.7871, "step": 40027 }, { "epoch": 7.62, "learning_rate": 7.133651096282173e-07, "loss": 0.6365, "step": 40028 }, { "epoch": 7.62, "learning_rate": 7.133079122974261e-07, "loss": 1.0098, "step": 40029 }, { "epoch": 7.62, "learning_rate": 7.132507149666349e-07, "loss": 0.3126, "step": 40030 }, { "epoch": 7.62, "learning_rate": 7.131935176358437e-07, "loss": 1.4416, "step": 40031 }, { "epoch": 7.63, "learning_rate": 7.131363203050524e-07, "loss": 1.3076, "step": 40032 }, { "epoch": 7.63, "learning_rate": 7.130791229742613e-07, "loss": 0.7136, "step": 40033 }, { "epoch": 7.63, "learning_rate": 7.1302192564347e-07, "loss": 0.7708, "step": 40034 }, { "epoch": 7.63, "learning_rate": 7.129647283126788e-07, "loss": 1.082, "step": 40035 }, { "epoch": 7.63, "learning_rate": 7.129075309818876e-07, "loss": 1.739, "step": 40036 }, { "epoch": 7.63, "learning_rate": 7.128503336510963e-07, "loss": 2.8235, "step": 40037 }, { "epoch": 7.63, "learning_rate": 7.127931363203052e-07, "loss": 1.2938, "step": 40038 }, { "epoch": 7.63, "learning_rate": 7.127359389895138e-07, "loss": 1.5045, "step": 40039 }, { "epoch": 7.63, "learning_rate": 7.126787416587225e-07, "loss": 0.797, "step": 40040 }, { "epoch": 7.63, "learning_rate": 7.126215443279314e-07, "loss": 0.6396, "step": 40041 }, { "epoch": 7.63, "learning_rate": 7.125643469971401e-07, "loss": 2.0137, "step": 40042 }, { "epoch": 7.63, "learning_rate": 7.125071496663489e-07, "loss": 2.1536, "step": 40043 }, { "epoch": 7.63, "learning_rate": 7.124499523355577e-07, "loss": 1.1751, "step": 40044 }, { "epoch": 7.63, "learning_rate": 7.123927550047665e-07, "loss": 2.961, "step": 40045 }, { "epoch": 7.63, "learning_rate": 7.123355576739752e-07, "loss": 2.1418, "step": 40046 }, { "epoch": 7.63, "learning_rate": 7.12278360343184e-07, "loss": 2.4849, "step": 40047 }, { "epoch": 7.63, "learning_rate": 7.122211630123928e-07, "loss": 2.6219, "step": 40048 }, { "epoch": 7.63, "learning_rate": 7.121639656816016e-07, "loss": 1.9724, "step": 40049 }, { "epoch": 7.63, "learning_rate": 7.121067683508104e-07, "loss": 1.1931, "step": 40050 }, { "epoch": 7.63, "learning_rate": 7.12049571020019e-07, "loss": 0.7869, "step": 40051 }, { "epoch": 7.63, "learning_rate": 7.119923736892278e-07, "loss": 2.0297, "step": 40052 }, { "epoch": 7.63, "learning_rate": 7.119351763584366e-07, "loss": 0.7017, "step": 40053 }, { "epoch": 7.63, "learning_rate": 7.118779790276453e-07, "loss": 1.4248, "step": 40054 }, { "epoch": 7.63, "learning_rate": 7.118207816968542e-07, "loss": 2.7835, "step": 40055 }, { "epoch": 7.63, "learning_rate": 7.117635843660629e-07, "loss": 1.5303, "step": 40056 }, { "epoch": 7.63, "learning_rate": 7.117063870352717e-07, "loss": 1.6574, "step": 40057 }, { "epoch": 7.63, "learning_rate": 7.116491897044805e-07, "loss": 1.0298, "step": 40058 }, { "epoch": 7.63, "learning_rate": 7.115919923736892e-07, "loss": 1.4479, "step": 40059 }, { "epoch": 7.63, "learning_rate": 7.115347950428981e-07, "loss": 3.4176, "step": 40060 }, { "epoch": 7.63, "learning_rate": 7.114775977121068e-07, "loss": 2.1467, "step": 40061 }, { "epoch": 7.63, "learning_rate": 7.114204003813156e-07, "loss": 2.2868, "step": 40062 }, { "epoch": 7.63, "learning_rate": 7.113632030505244e-07, "loss": 0.468, "step": 40063 }, { "epoch": 7.63, "learning_rate": 7.11306005719733e-07, "loss": 1.5944, "step": 40064 }, { "epoch": 7.63, "learning_rate": 7.112488083889418e-07, "loss": 2.6316, "step": 40065 }, { "epoch": 7.63, "learning_rate": 7.111916110581506e-07, "loss": 1.293, "step": 40066 }, { "epoch": 7.63, "learning_rate": 7.111344137273594e-07, "loss": 0.637, "step": 40067 }, { "epoch": 7.63, "learning_rate": 7.110772163965681e-07, "loss": 1.1183, "step": 40068 }, { "epoch": 7.63, "learning_rate": 7.11020019065777e-07, "loss": 2.1339, "step": 40069 }, { "epoch": 7.63, "learning_rate": 7.109628217349857e-07, "loss": 2.3755, "step": 40070 }, { "epoch": 7.63, "learning_rate": 7.109056244041945e-07, "loss": 2.1814, "step": 40071 }, { "epoch": 7.63, "learning_rate": 7.108484270734033e-07, "loss": 2.1483, "step": 40072 }, { "epoch": 7.63, "learning_rate": 7.10791229742612e-07, "loss": 1.4825, "step": 40073 }, { "epoch": 7.63, "learning_rate": 7.107340324118209e-07, "loss": 2.1701, "step": 40074 }, { "epoch": 7.63, "learning_rate": 7.106768350810296e-07, "loss": 2.1105, "step": 40075 }, { "epoch": 7.63, "learning_rate": 7.106196377502383e-07, "loss": 1.1187, "step": 40076 }, { "epoch": 7.63, "learning_rate": 7.105624404194472e-07, "loss": 2.3607, "step": 40077 }, { "epoch": 7.63, "learning_rate": 7.105052430886558e-07, "loss": 3.0975, "step": 40078 }, { "epoch": 7.63, "learning_rate": 7.104480457578646e-07, "loss": 1.3207, "step": 40079 }, { "epoch": 7.63, "learning_rate": 7.103908484270734e-07, "loss": 1.0407, "step": 40080 }, { "epoch": 7.63, "learning_rate": 7.103336510962822e-07, "loss": 1.9736, "step": 40081 }, { "epoch": 7.63, "learning_rate": 7.10276453765491e-07, "loss": 2.5465, "step": 40082 }, { "epoch": 7.63, "learning_rate": 7.102192564346997e-07, "loss": 1.8664, "step": 40083 }, { "epoch": 7.64, "learning_rate": 7.101620591039085e-07, "loss": 1.7357, "step": 40084 }, { "epoch": 7.64, "learning_rate": 7.101048617731173e-07, "loss": 1.4112, "step": 40085 }, { "epoch": 7.64, "learning_rate": 7.100476644423261e-07, "loss": 2.1038, "step": 40086 }, { "epoch": 7.64, "learning_rate": 7.099904671115348e-07, "loss": 1.5148, "step": 40087 }, { "epoch": 7.64, "learning_rate": 7.099332697807436e-07, "loss": 1.244, "step": 40088 }, { "epoch": 7.64, "learning_rate": 7.098760724499524e-07, "loss": 1.9534, "step": 40089 }, { "epoch": 7.64, "learning_rate": 7.098188751191611e-07, "loss": 1.1023, "step": 40090 }, { "epoch": 7.64, "learning_rate": 7.0976167778837e-07, "loss": 2.2219, "step": 40091 }, { "epoch": 7.64, "learning_rate": 7.097044804575786e-07, "loss": 2.245, "step": 40092 }, { "epoch": 7.64, "learning_rate": 7.096472831267874e-07, "loss": 2.1808, "step": 40093 }, { "epoch": 7.64, "learning_rate": 7.095900857959962e-07, "loss": 1.5331, "step": 40094 }, { "epoch": 7.64, "learning_rate": 7.095328884652049e-07, "loss": 2.2222, "step": 40095 }, { "epoch": 7.64, "learning_rate": 7.094756911344138e-07, "loss": 1.7224, "step": 40096 }, { "epoch": 7.64, "learning_rate": 7.094184938036225e-07, "loss": 1.5878, "step": 40097 }, { "epoch": 7.64, "learning_rate": 7.093612964728313e-07, "loss": 1.783, "step": 40098 }, { "epoch": 7.64, "learning_rate": 7.093040991420401e-07, "loss": 1.8575, "step": 40099 }, { "epoch": 7.64, "learning_rate": 7.092469018112488e-07, "loss": 1.811, "step": 40100 }, { "epoch": 7.64, "learning_rate": 7.091897044804576e-07, "loss": 1.8977, "step": 40101 }, { "epoch": 7.64, "learning_rate": 7.091325071496664e-07, "loss": 0.9667, "step": 40102 }, { "epoch": 7.64, "learning_rate": 7.090753098188752e-07, "loss": 2.2931, "step": 40103 }, { "epoch": 7.64, "learning_rate": 7.09018112488084e-07, "loss": 1.6799, "step": 40104 }, { "epoch": 7.64, "learning_rate": 7.089609151572926e-07, "loss": 2.2752, "step": 40105 }, { "epoch": 7.64, "learning_rate": 7.089037178265014e-07, "loss": 2.6255, "step": 40106 }, { "epoch": 7.64, "learning_rate": 7.088465204957102e-07, "loss": 2.234, "step": 40107 }, { "epoch": 7.64, "learning_rate": 7.08789323164919e-07, "loss": 1.1664, "step": 40108 }, { "epoch": 7.64, "learning_rate": 7.087321258341277e-07, "loss": 2.5099, "step": 40109 }, { "epoch": 7.64, "learning_rate": 7.086749285033366e-07, "loss": 1.6068, "step": 40110 }, { "epoch": 7.64, "learning_rate": 7.086177311725453e-07, "loss": 1.1808, "step": 40111 }, { "epoch": 7.64, "learning_rate": 7.08560533841754e-07, "loss": 1.5607, "step": 40112 }, { "epoch": 7.64, "learning_rate": 7.085033365109629e-07, "loss": 1.458, "step": 40113 }, { "epoch": 7.64, "learning_rate": 7.084461391801716e-07, "loss": 2.1111, "step": 40114 }, { "epoch": 7.64, "learning_rate": 7.083889418493804e-07, "loss": 2.0314, "step": 40115 }, { "epoch": 7.64, "learning_rate": 7.083317445185892e-07, "loss": 0.9999, "step": 40116 }, { "epoch": 7.64, "learning_rate": 7.082745471877979e-07, "loss": 1.9022, "step": 40117 }, { "epoch": 7.64, "learning_rate": 7.082173498570068e-07, "loss": 1.1861, "step": 40118 }, { "epoch": 7.64, "learning_rate": 7.081601525262154e-07, "loss": 3.3676, "step": 40119 }, { "epoch": 7.64, "learning_rate": 7.081029551954242e-07, "loss": 2.5542, "step": 40120 }, { "epoch": 7.64, "learning_rate": 7.08045757864633e-07, "loss": 2.1408, "step": 40121 }, { "epoch": 7.64, "learning_rate": 7.079885605338418e-07, "loss": 1.187, "step": 40122 }, { "epoch": 7.64, "learning_rate": 7.079313632030505e-07, "loss": 1.4567, "step": 40123 }, { "epoch": 7.64, "learning_rate": 7.078741658722593e-07, "loss": 1.9575, "step": 40124 }, { "epoch": 7.64, "learning_rate": 7.078169685414681e-07, "loss": 1.1635, "step": 40125 }, { "epoch": 7.64, "learning_rate": 7.077597712106769e-07, "loss": 0.7472, "step": 40126 }, { "epoch": 7.64, "learning_rate": 7.077025738798857e-07, "loss": 3.2175, "step": 40127 }, { "epoch": 7.64, "learning_rate": 7.076453765490944e-07, "loss": 1.6555, "step": 40128 }, { "epoch": 7.64, "learning_rate": 7.075881792183032e-07, "loss": 1.1867, "step": 40129 }, { "epoch": 7.64, "learning_rate": 7.07530981887512e-07, "loss": 1.1574, "step": 40130 }, { "epoch": 7.64, "learning_rate": 7.074737845567206e-07, "loss": 2.0042, "step": 40131 }, { "epoch": 7.64, "learning_rate": 7.074165872259295e-07, "loss": 0.5692, "step": 40132 }, { "epoch": 7.64, "learning_rate": 7.073593898951382e-07, "loss": 1.5681, "step": 40133 }, { "epoch": 7.64, "learning_rate": 7.07302192564347e-07, "loss": 1.0204, "step": 40134 }, { "epoch": 7.64, "learning_rate": 7.072449952335558e-07, "loss": 2.1345, "step": 40135 }, { "epoch": 7.64, "learning_rate": 7.071877979027645e-07, "loss": 2.1622, "step": 40136 }, { "epoch": 7.65, "learning_rate": 7.071306005719733e-07, "loss": 1.2062, "step": 40137 }, { "epoch": 7.65, "learning_rate": 7.070734032411821e-07, "loss": 2.8354, "step": 40138 }, { "epoch": 7.65, "learning_rate": 7.070162059103909e-07, "loss": 2.2667, "step": 40139 }, { "epoch": 7.65, "learning_rate": 7.069590085795997e-07, "loss": 1.2078, "step": 40140 }, { "epoch": 7.65, "learning_rate": 7.069018112488084e-07, "loss": 3.1417, "step": 40141 }, { "epoch": 7.65, "learning_rate": 7.068446139180172e-07, "loss": 0.7495, "step": 40142 }, { "epoch": 7.65, "learning_rate": 7.06787416587226e-07, "loss": 2.3602, "step": 40143 }, { "epoch": 7.65, "learning_rate": 7.067302192564348e-07, "loss": 1.259, "step": 40144 }, { "epoch": 7.65, "learning_rate": 7.066730219256434e-07, "loss": 2.0841, "step": 40145 }, { "epoch": 7.65, "learning_rate": 7.066158245948522e-07, "loss": 2.2386, "step": 40146 }, { "epoch": 7.65, "learning_rate": 7.06558627264061e-07, "loss": 0.8214, "step": 40147 }, { "epoch": 7.65, "learning_rate": 7.065014299332698e-07, "loss": 1.4878, "step": 40148 }, { "epoch": 7.65, "learning_rate": 7.064442326024786e-07, "loss": 1.7783, "step": 40149 }, { "epoch": 7.65, "learning_rate": 7.063870352716873e-07, "loss": 2.0966, "step": 40150 }, { "epoch": 7.65, "learning_rate": 7.063298379408961e-07, "loss": 1.5795, "step": 40151 }, { "epoch": 7.65, "learning_rate": 7.062726406101049e-07, "loss": 1.3852, "step": 40152 }, { "epoch": 7.65, "learning_rate": 7.062154432793136e-07, "loss": 0.496, "step": 40153 }, { "epoch": 7.65, "learning_rate": 7.061582459485225e-07, "loss": 0.9428, "step": 40154 }, { "epoch": 7.65, "learning_rate": 7.061010486177312e-07, "loss": 2.283, "step": 40155 }, { "epoch": 7.65, "learning_rate": 7.0604385128694e-07, "loss": 2.0816, "step": 40156 }, { "epoch": 7.65, "learning_rate": 7.059866539561488e-07, "loss": 1.6754, "step": 40157 }, { "epoch": 7.65, "learning_rate": 7.059294566253574e-07, "loss": 0.6267, "step": 40158 }, { "epoch": 7.65, "learning_rate": 7.058722592945662e-07, "loss": 1.8344, "step": 40159 }, { "epoch": 7.65, "learning_rate": 7.05815061963775e-07, "loss": 2.262, "step": 40160 }, { "epoch": 7.65, "learning_rate": 7.057578646329838e-07, "loss": 1.8848, "step": 40161 }, { "epoch": 7.65, "learning_rate": 7.057006673021926e-07, "loss": 1.5226, "step": 40162 }, { "epoch": 7.65, "learning_rate": 7.056434699714014e-07, "loss": 1.2564, "step": 40163 }, { "epoch": 7.65, "learning_rate": 7.055862726406101e-07, "loss": 2.5645, "step": 40164 }, { "epoch": 7.65, "learning_rate": 7.055290753098189e-07, "loss": 1.461, "step": 40165 }, { "epoch": 7.65, "learning_rate": 7.054718779790277e-07, "loss": 1.1956, "step": 40166 }, { "epoch": 7.65, "learning_rate": 7.054146806482364e-07, "loss": 2.1445, "step": 40167 }, { "epoch": 7.65, "learning_rate": 7.053574833174453e-07, "loss": 1.9182, "step": 40168 }, { "epoch": 7.65, "learning_rate": 7.05300285986654e-07, "loss": 1.9925, "step": 40169 }, { "epoch": 7.65, "learning_rate": 7.052430886558628e-07, "loss": 1.3305, "step": 40170 }, { "epoch": 7.65, "learning_rate": 7.051858913250716e-07, "loss": 0.9444, "step": 40171 }, { "epoch": 7.65, "learning_rate": 7.051286939942802e-07, "loss": 1.0355, "step": 40172 }, { "epoch": 7.65, "learning_rate": 7.05071496663489e-07, "loss": 2.4008, "step": 40173 }, { "epoch": 7.65, "learning_rate": 7.050142993326978e-07, "loss": 3.1853, "step": 40174 }, { "epoch": 7.65, "learning_rate": 7.049571020019066e-07, "loss": 1.5505, "step": 40175 }, { "epoch": 7.65, "learning_rate": 7.048999046711154e-07, "loss": 1.9821, "step": 40176 }, { "epoch": 7.65, "learning_rate": 7.048427073403241e-07, "loss": 1.0095, "step": 40177 }, { "epoch": 7.65, "learning_rate": 7.047855100095329e-07, "loss": 1.9987, "step": 40178 }, { "epoch": 7.65, "learning_rate": 7.047283126787417e-07, "loss": 0.5909, "step": 40179 }, { "epoch": 7.65, "learning_rate": 7.046711153479505e-07, "loss": 2.4184, "step": 40180 }, { "epoch": 7.65, "learning_rate": 7.046139180171592e-07, "loss": 1.2797, "step": 40181 }, { "epoch": 7.65, "learning_rate": 7.04556720686368e-07, "loss": 3.4535, "step": 40182 }, { "epoch": 7.65, "learning_rate": 7.044995233555768e-07, "loss": 1.4307, "step": 40183 }, { "epoch": 7.65, "learning_rate": 7.044423260247856e-07, "loss": 0.9992, "step": 40184 }, { "epoch": 7.65, "learning_rate": 7.043851286939943e-07, "loss": 0.8943, "step": 40185 }, { "epoch": 7.65, "learning_rate": 7.04327931363203e-07, "loss": 0.7935, "step": 40186 }, { "epoch": 7.65, "learning_rate": 7.042707340324118e-07, "loss": 0.9508, "step": 40187 }, { "epoch": 7.65, "learning_rate": 7.042135367016206e-07, "loss": 2.17, "step": 40188 }, { "epoch": 7.66, "learning_rate": 7.041563393708293e-07, "loss": 1.9977, "step": 40189 }, { "epoch": 7.66, "learning_rate": 7.040991420400382e-07, "loss": 2.7241, "step": 40190 }, { "epoch": 7.66, "learning_rate": 7.040419447092469e-07, "loss": 1.6963, "step": 40191 }, { "epoch": 7.66, "learning_rate": 7.039847473784557e-07, "loss": 1.5419, "step": 40192 }, { "epoch": 7.66, "learning_rate": 7.039275500476645e-07, "loss": 1.6432, "step": 40193 }, { "epoch": 7.66, "learning_rate": 7.038703527168732e-07, "loss": 0.8231, "step": 40194 }, { "epoch": 7.66, "learning_rate": 7.03813155386082e-07, "loss": 2.4725, "step": 40195 }, { "epoch": 7.66, "learning_rate": 7.037559580552908e-07, "loss": 1.3881, "step": 40196 }, { "epoch": 7.66, "learning_rate": 7.036987607244996e-07, "loss": 1.8606, "step": 40197 }, { "epoch": 7.66, "learning_rate": 7.036415633937084e-07, "loss": 0.4828, "step": 40198 }, { "epoch": 7.66, "learning_rate": 7.035843660629171e-07, "loss": 1.5031, "step": 40199 }, { "epoch": 7.66, "learning_rate": 7.035271687321258e-07, "loss": 1.8921, "step": 40200 }, { "epoch": 7.66, "learning_rate": 7.034699714013346e-07, "loss": 2.8283, "step": 40201 }, { "epoch": 7.66, "learning_rate": 7.034127740705434e-07, "loss": 1.7679, "step": 40202 }, { "epoch": 7.66, "learning_rate": 7.033555767397521e-07, "loss": 1.5969, "step": 40203 }, { "epoch": 7.66, "learning_rate": 7.03298379408961e-07, "loss": 0.4565, "step": 40204 }, { "epoch": 7.66, "learning_rate": 7.032411820781697e-07, "loss": 0.5574, "step": 40205 }, { "epoch": 7.66, "learning_rate": 7.031839847473785e-07, "loss": 1.6568, "step": 40206 }, { "epoch": 7.66, "learning_rate": 7.031267874165873e-07, "loss": 2.8094, "step": 40207 }, { "epoch": 7.66, "learning_rate": 7.03069590085796e-07, "loss": 1.5138, "step": 40208 }, { "epoch": 7.66, "learning_rate": 7.030123927550048e-07, "loss": 1.6577, "step": 40209 }, { "epoch": 7.66, "learning_rate": 7.029551954242136e-07, "loss": 1.3844, "step": 40210 }, { "epoch": 7.66, "learning_rate": 7.028979980934222e-07, "loss": 1.8164, "step": 40211 }, { "epoch": 7.66, "learning_rate": 7.028408007626311e-07, "loss": 0.4311, "step": 40212 }, { "epoch": 7.66, "learning_rate": 7.027836034318398e-07, "loss": 1.6612, "step": 40213 }, { "epoch": 7.66, "learning_rate": 7.027264061010486e-07, "loss": 1.3286, "step": 40214 }, { "epoch": 7.66, "learning_rate": 7.026692087702574e-07, "loss": 1.2762, "step": 40215 }, { "epoch": 7.66, "learning_rate": 7.026120114394662e-07, "loss": 1.714, "step": 40216 }, { "epoch": 7.66, "learning_rate": 7.025548141086749e-07, "loss": 1.4685, "step": 40217 }, { "epoch": 7.66, "learning_rate": 7.024976167778837e-07, "loss": 1.7523, "step": 40218 }, { "epoch": 7.66, "learning_rate": 7.024404194470925e-07, "loss": 2.2531, "step": 40219 }, { "epoch": 7.66, "learning_rate": 7.023832221163013e-07, "loss": 0.8117, "step": 40220 }, { "epoch": 7.66, "learning_rate": 7.023260247855101e-07, "loss": 1.6581, "step": 40221 }, { "epoch": 7.66, "learning_rate": 7.022688274547188e-07, "loss": 1.4005, "step": 40222 }, { "epoch": 7.66, "learning_rate": 7.022116301239276e-07, "loss": 1.5321, "step": 40223 }, { "epoch": 7.66, "learning_rate": 7.021544327931364e-07, "loss": 1.4853, "step": 40224 }, { "epoch": 7.66, "learning_rate": 7.02097235462345e-07, "loss": 2.3741, "step": 40225 }, { "epoch": 7.66, "learning_rate": 7.020400381315539e-07, "loss": 1.1045, "step": 40226 }, { "epoch": 7.66, "learning_rate": 7.019828408007626e-07, "loss": 1.4523, "step": 40227 }, { "epoch": 7.66, "learning_rate": 7.019256434699714e-07, "loss": 1.3688, "step": 40228 }, { "epoch": 7.66, "learning_rate": 7.018684461391802e-07, "loss": 0.8039, "step": 40229 }, { "epoch": 7.66, "learning_rate": 7.018112488083889e-07, "loss": 1.4007, "step": 40230 }, { "epoch": 7.66, "learning_rate": 7.017540514775977e-07, "loss": 1.3716, "step": 40231 }, { "epoch": 7.66, "learning_rate": 7.016968541468065e-07, "loss": 0.4325, "step": 40232 }, { "epoch": 7.66, "learning_rate": 7.016396568160153e-07, "loss": 0.9507, "step": 40233 }, { "epoch": 7.66, "learning_rate": 7.015824594852241e-07, "loss": 1.8904, "step": 40234 }, { "epoch": 7.66, "learning_rate": 7.015252621544328e-07, "loss": 2.0154, "step": 40235 }, { "epoch": 7.66, "learning_rate": 7.014680648236416e-07, "loss": 0.8, "step": 40236 }, { "epoch": 7.66, "learning_rate": 7.014108674928504e-07, "loss": 0.7123, "step": 40237 }, { "epoch": 7.66, "learning_rate": 7.013536701620591e-07, "loss": 1.5965, "step": 40238 }, { "epoch": 7.66, "learning_rate": 7.012964728312678e-07, "loss": 1.4636, "step": 40239 }, { "epoch": 7.66, "learning_rate": 7.012392755004767e-07, "loss": 1.1692, "step": 40240 }, { "epoch": 7.66, "learning_rate": 7.011820781696854e-07, "loss": 1.7889, "step": 40241 }, { "epoch": 7.67, "learning_rate": 7.011248808388942e-07, "loss": 0.7703, "step": 40242 }, { "epoch": 7.67, "learning_rate": 7.01067683508103e-07, "loss": 1.7383, "step": 40243 }, { "epoch": 7.67, "learning_rate": 7.010104861773117e-07, "loss": 2.774, "step": 40244 }, { "epoch": 7.67, "learning_rate": 7.009532888465205e-07, "loss": 1.745, "step": 40245 }, { "epoch": 7.67, "learning_rate": 7.008960915157293e-07, "loss": 1.2282, "step": 40246 }, { "epoch": 7.67, "learning_rate": 7.00838894184938e-07, "loss": 1.8055, "step": 40247 }, { "epoch": 7.67, "learning_rate": 7.007816968541469e-07, "loss": 1.1634, "step": 40248 }, { "epoch": 7.67, "learning_rate": 7.007244995233556e-07, "loss": 0.7273, "step": 40249 }, { "epoch": 7.67, "learning_rate": 7.006673021925644e-07, "loss": 2.45, "step": 40250 }, { "epoch": 7.67, "learning_rate": 7.006101048617732e-07, "loss": 1.2768, "step": 40251 }, { "epoch": 7.67, "learning_rate": 7.005529075309819e-07, "loss": 1.8977, "step": 40252 }, { "epoch": 7.67, "learning_rate": 7.004957102001906e-07, "loss": 0.8557, "step": 40253 }, { "epoch": 7.67, "learning_rate": 7.004385128693994e-07, "loss": 1.868, "step": 40254 }, { "epoch": 7.67, "learning_rate": 7.003813155386082e-07, "loss": 1.5747, "step": 40255 }, { "epoch": 7.67, "learning_rate": 7.00324118207817e-07, "loss": 2.5924, "step": 40256 }, { "epoch": 7.67, "learning_rate": 7.002669208770258e-07, "loss": 1.9526, "step": 40257 }, { "epoch": 7.67, "learning_rate": 7.002097235462345e-07, "loss": 2.1069, "step": 40258 }, { "epoch": 7.67, "learning_rate": 7.001525262154433e-07, "loss": 1.4948, "step": 40259 }, { "epoch": 7.67, "learning_rate": 7.000953288846521e-07, "loss": 1.2019, "step": 40260 }, { "epoch": 7.67, "learning_rate": 7.000381315538608e-07, "loss": 0.7992, "step": 40261 }, { "epoch": 7.67, "learning_rate": 6.999809342230697e-07, "loss": 1.6756, "step": 40262 }, { "epoch": 7.67, "learning_rate": 6.999237368922784e-07, "loss": 1.6756, "step": 40263 }, { "epoch": 7.67, "learning_rate": 6.998665395614872e-07, "loss": 0.7715, "step": 40264 }, { "epoch": 7.67, "learning_rate": 6.99809342230696e-07, "loss": 1.2598, "step": 40265 }, { "epoch": 7.67, "learning_rate": 6.997521448999046e-07, "loss": 1.8532, "step": 40266 }, { "epoch": 7.67, "learning_rate": 6.996949475691134e-07, "loss": 1.6172, "step": 40267 }, { "epoch": 7.67, "learning_rate": 6.996377502383222e-07, "loss": 1.5285, "step": 40268 }, { "epoch": 7.67, "learning_rate": 6.99580552907531e-07, "loss": 0.9036, "step": 40269 }, { "epoch": 7.67, "learning_rate": 6.995233555767398e-07, "loss": 2.4307, "step": 40270 }, { "epoch": 7.67, "learning_rate": 6.994661582459485e-07, "loss": 1.6008, "step": 40271 }, { "epoch": 7.67, "learning_rate": 6.994089609151573e-07, "loss": 1.6477, "step": 40272 }, { "epoch": 7.67, "learning_rate": 6.993517635843661e-07, "loss": 2.2703, "step": 40273 }, { "epoch": 7.67, "learning_rate": 6.992945662535749e-07, "loss": 2.1809, "step": 40274 }, { "epoch": 7.67, "learning_rate": 6.992373689227836e-07, "loss": 2.3549, "step": 40275 }, { "epoch": 7.67, "learning_rate": 6.991801715919924e-07, "loss": 1.7448, "step": 40276 }, { "epoch": 7.67, "learning_rate": 6.991229742612012e-07, "loss": 1.0941, "step": 40277 }, { "epoch": 7.67, "learning_rate": 6.9906577693041e-07, "loss": 2.8429, "step": 40278 }, { "epoch": 7.67, "learning_rate": 6.990085795996187e-07, "loss": 2.5212, "step": 40279 }, { "epoch": 7.67, "learning_rate": 6.989513822688274e-07, "loss": 1.9382, "step": 40280 }, { "epoch": 7.67, "learning_rate": 6.988941849380363e-07, "loss": 1.2908, "step": 40281 }, { "epoch": 7.67, "learning_rate": 6.98836987607245e-07, "loss": 1.1983, "step": 40282 }, { "epoch": 7.67, "learning_rate": 6.987797902764537e-07, "loss": 2.0448, "step": 40283 }, { "epoch": 7.67, "learning_rate": 6.987225929456626e-07, "loss": 1.6434, "step": 40284 }, { "epoch": 7.67, "learning_rate": 6.986653956148713e-07, "loss": 2.3918, "step": 40285 }, { "epoch": 7.67, "learning_rate": 6.986081982840801e-07, "loss": 1.6469, "step": 40286 }, { "epoch": 7.67, "learning_rate": 6.985510009532889e-07, "loss": 1.2871, "step": 40287 }, { "epoch": 7.67, "learning_rate": 6.984938036224976e-07, "loss": 0.5808, "step": 40288 }, { "epoch": 7.67, "learning_rate": 6.984366062917064e-07, "loss": 1.4401, "step": 40289 }, { "epoch": 7.67, "learning_rate": 6.983794089609152e-07, "loss": 1.1397, "step": 40290 }, { "epoch": 7.67, "learning_rate": 6.98322211630124e-07, "loss": 1.6226, "step": 40291 }, { "epoch": 7.67, "learning_rate": 6.982650142993327e-07, "loss": 1.2806, "step": 40292 }, { "epoch": 7.67, "learning_rate": 6.982078169685415e-07, "loss": 0.4864, "step": 40293 }, { "epoch": 7.68, "learning_rate": 6.981506196377502e-07, "loss": 2.533, "step": 40294 }, { "epoch": 7.68, "learning_rate": 6.98093422306959e-07, "loss": 1.2375, "step": 40295 }, { "epoch": 7.68, "learning_rate": 6.980362249761678e-07, "loss": 1.2684, "step": 40296 }, { "epoch": 7.68, "learning_rate": 6.979790276453765e-07, "loss": 1.4833, "step": 40297 }, { "epoch": 7.68, "learning_rate": 6.979218303145854e-07, "loss": 1.4862, "step": 40298 }, { "epoch": 7.68, "learning_rate": 6.978646329837941e-07, "loss": 3.0565, "step": 40299 }, { "epoch": 7.68, "learning_rate": 6.978074356530029e-07, "loss": 1.7886, "step": 40300 }, { "epoch": 7.68, "learning_rate": 6.977502383222117e-07, "loss": 2.6995, "step": 40301 }, { "epoch": 7.68, "learning_rate": 6.976930409914204e-07, "loss": 0.4983, "step": 40302 }, { "epoch": 7.68, "learning_rate": 6.976358436606293e-07, "loss": 2.8307, "step": 40303 }, { "epoch": 7.68, "learning_rate": 6.97578646329838e-07, "loss": 1.4486, "step": 40304 }, { "epoch": 7.68, "learning_rate": 6.975214489990467e-07, "loss": 0.7029, "step": 40305 }, { "epoch": 7.68, "learning_rate": 6.974642516682555e-07, "loss": 1.4234, "step": 40306 }, { "epoch": 7.68, "learning_rate": 6.974070543374642e-07, "loss": 1.4693, "step": 40307 }, { "epoch": 7.68, "learning_rate": 6.97349857006673e-07, "loss": 1.6264, "step": 40308 }, { "epoch": 7.68, "learning_rate": 6.972926596758818e-07, "loss": 1.5828, "step": 40309 }, { "epoch": 7.68, "learning_rate": 6.972354623450906e-07, "loss": 0.9697, "step": 40310 }, { "epoch": 7.68, "learning_rate": 6.971782650142993e-07, "loss": 1.2868, "step": 40311 }, { "epoch": 7.68, "learning_rate": 6.971210676835081e-07, "loss": 1.3074, "step": 40312 }, { "epoch": 7.68, "learning_rate": 6.970638703527169e-07, "loss": 2.2189, "step": 40313 }, { "epoch": 7.68, "learning_rate": 6.970066730219257e-07, "loss": 3.0462, "step": 40314 }, { "epoch": 7.68, "learning_rate": 6.969494756911345e-07, "loss": 1.3535, "step": 40315 }, { "epoch": 7.68, "learning_rate": 6.968922783603432e-07, "loss": 1.8286, "step": 40316 }, { "epoch": 7.68, "learning_rate": 6.968350810295521e-07, "loss": 2.239, "step": 40317 }, { "epoch": 7.68, "learning_rate": 6.967778836987607e-07, "loss": 0.801, "step": 40318 }, { "epoch": 7.68, "learning_rate": 6.967206863679694e-07, "loss": 2.1331, "step": 40319 }, { "epoch": 7.68, "learning_rate": 6.966634890371783e-07, "loss": 0.5159, "step": 40320 }, { "epoch": 7.68, "learning_rate": 6.96606291706387e-07, "loss": 1.6843, "step": 40321 }, { "epoch": 7.68, "learning_rate": 6.965490943755958e-07, "loss": 2.7233, "step": 40322 }, { "epoch": 7.68, "learning_rate": 6.964918970448046e-07, "loss": 2.3406, "step": 40323 }, { "epoch": 7.68, "learning_rate": 6.964346997140133e-07, "loss": 1.2223, "step": 40324 }, { "epoch": 7.68, "learning_rate": 6.963775023832222e-07, "loss": 1.6446, "step": 40325 }, { "epoch": 7.68, "learning_rate": 6.963203050524309e-07, "loss": 1.8309, "step": 40326 }, { "epoch": 7.68, "learning_rate": 6.962631077216397e-07, "loss": 1.554, "step": 40327 }, { "epoch": 7.68, "learning_rate": 6.962059103908485e-07, "loss": 1.2386, "step": 40328 }, { "epoch": 7.68, "learning_rate": 6.961487130600572e-07, "loss": 1.1151, "step": 40329 }, { "epoch": 7.68, "learning_rate": 6.96091515729266e-07, "loss": 1.1816, "step": 40330 }, { "epoch": 7.68, "learning_rate": 6.960343183984748e-07, "loss": 2.3071, "step": 40331 }, { "epoch": 7.68, "learning_rate": 6.959771210676835e-07, "loss": 1.9986, "step": 40332 }, { "epoch": 7.68, "learning_rate": 6.959199237368922e-07, "loss": 2.415, "step": 40333 }, { "epoch": 7.68, "learning_rate": 6.958627264061011e-07, "loss": 2.3138, "step": 40334 }, { "epoch": 7.68, "learning_rate": 6.958055290753098e-07, "loss": 1.1034, "step": 40335 }, { "epoch": 7.68, "learning_rate": 6.957483317445186e-07, "loss": 1.8208, "step": 40336 }, { "epoch": 7.68, "learning_rate": 6.956911344137274e-07, "loss": 2.7236, "step": 40337 }, { "epoch": 7.68, "learning_rate": 6.956339370829361e-07, "loss": 0.8644, "step": 40338 }, { "epoch": 7.68, "learning_rate": 6.95576739752145e-07, "loss": 1.9316, "step": 40339 }, { "epoch": 7.68, "learning_rate": 6.955195424213537e-07, "loss": 0.3655, "step": 40340 }, { "epoch": 7.68, "learning_rate": 6.954623450905624e-07, "loss": 0.9568, "step": 40341 }, { "epoch": 7.68, "learning_rate": 6.954051477597713e-07, "loss": 2.4491, "step": 40342 }, { "epoch": 7.68, "learning_rate": 6.9534795042898e-07, "loss": 1.9663, "step": 40343 }, { "epoch": 7.68, "learning_rate": 6.952907530981888e-07, "loss": 1.5659, "step": 40344 }, { "epoch": 7.68, "learning_rate": 6.952335557673975e-07, "loss": 1.4228, "step": 40345 }, { "epoch": 7.68, "learning_rate": 6.951763584366063e-07, "loss": 0.9342, "step": 40346 }, { "epoch": 7.69, "learning_rate": 6.95119161105815e-07, "loss": 1.478, "step": 40347 }, { "epoch": 7.69, "learning_rate": 6.950619637750238e-07, "loss": 1.4268, "step": 40348 }, { "epoch": 7.69, "learning_rate": 6.950047664442326e-07, "loss": 1.2143, "step": 40349 }, { "epoch": 7.69, "learning_rate": 6.949475691134414e-07, "loss": 2.3148, "step": 40350 }, { "epoch": 7.69, "learning_rate": 6.948903717826502e-07, "loss": 1.2642, "step": 40351 }, { "epoch": 7.69, "learning_rate": 6.948331744518589e-07, "loss": 2.0629, "step": 40352 }, { "epoch": 7.69, "learning_rate": 6.947759771210677e-07, "loss": 2.5873, "step": 40353 }, { "epoch": 7.69, "learning_rate": 6.947187797902765e-07, "loss": 2.3537, "step": 40354 }, { "epoch": 7.69, "learning_rate": 6.946615824594852e-07, "loss": 2.2612, "step": 40355 }, { "epoch": 7.69, "learning_rate": 6.946043851286941e-07, "loss": 0.8166, "step": 40356 }, { "epoch": 7.69, "learning_rate": 6.945471877979028e-07, "loss": 1.9841, "step": 40357 }, { "epoch": 7.69, "learning_rate": 6.944899904671115e-07, "loss": 1.7646, "step": 40358 }, { "epoch": 7.69, "learning_rate": 6.944327931363203e-07, "loss": 4.4758, "step": 40359 }, { "epoch": 7.69, "learning_rate": 6.94375595805529e-07, "loss": 1.9271, "step": 40360 }, { "epoch": 7.69, "learning_rate": 6.943183984747379e-07, "loss": 2.0707, "step": 40361 }, { "epoch": 7.69, "learning_rate": 6.942612011439466e-07, "loss": 1.5549, "step": 40362 }, { "epoch": 7.69, "learning_rate": 6.942040038131554e-07, "loss": 1.6876, "step": 40363 }, { "epoch": 7.69, "learning_rate": 6.941468064823642e-07, "loss": 2.4275, "step": 40364 }, { "epoch": 7.69, "learning_rate": 6.940896091515729e-07, "loss": 1.6634, "step": 40365 }, { "epoch": 7.69, "learning_rate": 6.940324118207817e-07, "loss": 1.141, "step": 40366 }, { "epoch": 7.69, "learning_rate": 6.939752144899905e-07, "loss": 0.4671, "step": 40367 }, { "epoch": 7.69, "learning_rate": 6.939180171591993e-07, "loss": 1.7656, "step": 40368 }, { "epoch": 7.69, "learning_rate": 6.93860819828408e-07, "loss": 1.6542, "step": 40369 }, { "epoch": 7.69, "learning_rate": 6.938036224976169e-07, "loss": 2.5604, "step": 40370 }, { "epoch": 7.69, "learning_rate": 6.937464251668256e-07, "loss": 1.4086, "step": 40371 }, { "epoch": 7.69, "learning_rate": 6.936892278360343e-07, "loss": 1.3465, "step": 40372 }, { "epoch": 7.69, "learning_rate": 6.936320305052431e-07, "loss": 1.8089, "step": 40373 }, { "epoch": 7.69, "learning_rate": 6.935748331744518e-07, "loss": 0.756, "step": 40374 }, { "epoch": 7.69, "learning_rate": 6.935176358436607e-07, "loss": 1.26, "step": 40375 }, { "epoch": 7.69, "learning_rate": 6.934604385128694e-07, "loss": 2.2025, "step": 40376 }, { "epoch": 7.69, "learning_rate": 6.934032411820781e-07, "loss": 2.0967, "step": 40377 }, { "epoch": 7.69, "learning_rate": 6.93346043851287e-07, "loss": 1.0643, "step": 40378 }, { "epoch": 7.69, "learning_rate": 6.932888465204957e-07, "loss": 2.5658, "step": 40379 }, { "epoch": 7.69, "learning_rate": 6.932316491897045e-07, "loss": 1.8895, "step": 40380 }, { "epoch": 7.69, "learning_rate": 6.931744518589133e-07, "loss": 2.1054, "step": 40381 }, { "epoch": 7.69, "learning_rate": 6.931172545281221e-07, "loss": 2.5775, "step": 40382 }, { "epoch": 7.69, "learning_rate": 6.930600571973309e-07, "loss": 2.3239, "step": 40383 }, { "epoch": 7.69, "learning_rate": 6.930028598665396e-07, "loss": 1.7307, "step": 40384 }, { "epoch": 7.69, "learning_rate": 6.929456625357483e-07, "loss": 0.2732, "step": 40385 }, { "epoch": 7.69, "learning_rate": 6.928884652049571e-07, "loss": 1.9178, "step": 40386 }, { "epoch": 7.69, "learning_rate": 6.928312678741659e-07, "loss": 1.94, "step": 40387 }, { "epoch": 7.69, "learning_rate": 6.927740705433746e-07, "loss": 2.0295, "step": 40388 }, { "epoch": 7.69, "learning_rate": 6.927168732125834e-07, "loss": 1.8956, "step": 40389 }, { "epoch": 7.69, "learning_rate": 6.926596758817922e-07, "loss": 1.3826, "step": 40390 }, { "epoch": 7.69, "learning_rate": 6.926024785510009e-07, "loss": 1.3807, "step": 40391 }, { "epoch": 7.69, "learning_rate": 6.925452812202098e-07, "loss": 1.9511, "step": 40392 }, { "epoch": 7.69, "learning_rate": 6.924880838894185e-07, "loss": 1.7483, "step": 40393 }, { "epoch": 7.69, "learning_rate": 6.924308865586273e-07, "loss": 1.3617, "step": 40394 }, { "epoch": 7.69, "learning_rate": 6.923736892278361e-07, "loss": 1.7757, "step": 40395 }, { "epoch": 7.69, "learning_rate": 6.923164918970448e-07, "loss": 1.8966, "step": 40396 }, { "epoch": 7.69, "learning_rate": 6.922592945662537e-07, "loss": 0.8746, "step": 40397 }, { "epoch": 7.69, "learning_rate": 6.922020972354623e-07, "loss": 0.7895, "step": 40398 }, { "epoch": 7.7, "learning_rate": 6.921448999046711e-07, "loss": 2.0753, "step": 40399 }, { "epoch": 7.7, "learning_rate": 6.920877025738799e-07, "loss": 1.9556, "step": 40400 }, { "epoch": 7.7, "learning_rate": 6.920305052430886e-07, "loss": 2.6935, "step": 40401 }, { "epoch": 7.7, "learning_rate": 6.919733079122974e-07, "loss": 1.5226, "step": 40402 }, { "epoch": 7.7, "learning_rate": 6.919161105815062e-07, "loss": 0.6883, "step": 40403 }, { "epoch": 7.7, "learning_rate": 6.91858913250715e-07, "loss": 2.0182, "step": 40404 }, { "epoch": 7.7, "learning_rate": 6.918017159199238e-07, "loss": 1.9538, "step": 40405 }, { "epoch": 7.7, "learning_rate": 6.917445185891325e-07, "loss": 1.5531, "step": 40406 }, { "epoch": 7.7, "learning_rate": 6.916873212583413e-07, "loss": 1.3206, "step": 40407 }, { "epoch": 7.7, "learning_rate": 6.916301239275501e-07, "loss": 2.0138, "step": 40408 }, { "epoch": 7.7, "learning_rate": 6.915729265967589e-07, "loss": 1.1002, "step": 40409 }, { "epoch": 7.7, "learning_rate": 6.915157292659676e-07, "loss": 1.5073, "step": 40410 }, { "epoch": 7.7, "learning_rate": 6.914585319351765e-07, "loss": 2.1723, "step": 40411 }, { "epoch": 7.7, "learning_rate": 6.914013346043851e-07, "loss": 1.4911, "step": 40412 }, { "epoch": 7.7, "learning_rate": 6.913441372735938e-07, "loss": 1.0, "step": 40413 }, { "epoch": 7.7, "learning_rate": 6.912869399428027e-07, "loss": 0.2574, "step": 40414 }, { "epoch": 7.7, "learning_rate": 6.912297426120114e-07, "loss": 2.2233, "step": 40415 }, { "epoch": 7.7, "learning_rate": 6.911725452812202e-07, "loss": 1.5861, "step": 40416 }, { "epoch": 7.7, "learning_rate": 6.91115347950429e-07, "loss": 0.5596, "step": 40417 }, { "epoch": 7.7, "learning_rate": 6.910581506196377e-07, "loss": 1.0436, "step": 40418 }, { "epoch": 7.7, "learning_rate": 6.910009532888466e-07, "loss": 0.886, "step": 40419 }, { "epoch": 7.7, "learning_rate": 6.909437559580553e-07, "loss": 1.7391, "step": 40420 }, { "epoch": 7.7, "learning_rate": 6.908865586272641e-07, "loss": 2.3448, "step": 40421 }, { "epoch": 7.7, "learning_rate": 6.908293612964729e-07, "loss": 2.1101, "step": 40422 }, { "epoch": 7.7, "learning_rate": 6.907721639656817e-07, "loss": 1.2905, "step": 40423 }, { "epoch": 7.7, "learning_rate": 6.907149666348904e-07, "loss": 1.3097, "step": 40424 }, { "epoch": 7.7, "learning_rate": 6.906577693040991e-07, "loss": 0.5507, "step": 40425 }, { "epoch": 7.7, "learning_rate": 6.906005719733079e-07, "loss": 0.9581, "step": 40426 }, { "epoch": 7.7, "learning_rate": 6.905433746425167e-07, "loss": 1.2135, "step": 40427 }, { "epoch": 7.7, "learning_rate": 6.904861773117255e-07, "loss": 2.9133, "step": 40428 }, { "epoch": 7.7, "learning_rate": 6.904289799809342e-07, "loss": 2.0522, "step": 40429 }, { "epoch": 7.7, "learning_rate": 6.90371782650143e-07, "loss": 1.8247, "step": 40430 }, { "epoch": 7.7, "learning_rate": 6.903145853193518e-07, "loss": 1.9278, "step": 40431 }, { "epoch": 7.7, "learning_rate": 6.902573879885605e-07, "loss": 1.0494, "step": 40432 }, { "epoch": 7.7, "learning_rate": 6.902001906577694e-07, "loss": 2.0887, "step": 40433 }, { "epoch": 7.7, "learning_rate": 6.901429933269781e-07, "loss": 0.7317, "step": 40434 }, { "epoch": 7.7, "learning_rate": 6.900857959961869e-07, "loss": 1.7452, "step": 40435 }, { "epoch": 7.7, "learning_rate": 6.900285986653957e-07, "loss": 0.71, "step": 40436 }, { "epoch": 7.7, "learning_rate": 6.899714013346044e-07, "loss": 1.6794, "step": 40437 }, { "epoch": 7.7, "learning_rate": 6.899142040038131e-07, "loss": 1.3427, "step": 40438 }, { "epoch": 7.7, "learning_rate": 6.898570066730219e-07, "loss": 1.3387, "step": 40439 }, { "epoch": 7.7, "learning_rate": 6.897998093422307e-07, "loss": 3.516, "step": 40440 }, { "epoch": 7.7, "learning_rate": 6.897426120114395e-07, "loss": 1.5273, "step": 40441 }, { "epoch": 7.7, "learning_rate": 6.896854146806482e-07, "loss": 1.4684, "step": 40442 }, { "epoch": 7.7, "learning_rate": 6.89628217349857e-07, "loss": 0.994, "step": 40443 }, { "epoch": 7.7, "learning_rate": 6.895710200190658e-07, "loss": 1.6216, "step": 40444 }, { "epoch": 7.7, "learning_rate": 6.895138226882746e-07, "loss": 1.6763, "step": 40445 }, { "epoch": 7.7, "learning_rate": 6.894566253574833e-07, "loss": 1.5847, "step": 40446 }, { "epoch": 7.7, "learning_rate": 6.893994280266922e-07, "loss": 1.6777, "step": 40447 }, { "epoch": 7.7, "learning_rate": 6.893422306959009e-07, "loss": 1.967, "step": 40448 }, { "epoch": 7.7, "learning_rate": 6.892850333651097e-07, "loss": 1.5222, "step": 40449 }, { "epoch": 7.7, "learning_rate": 6.892278360343185e-07, "loss": 1.2954, "step": 40450 }, { "epoch": 7.7, "learning_rate": 6.891706387035272e-07, "loss": 1.7204, "step": 40451 }, { "epoch": 7.71, "learning_rate": 6.891134413727359e-07, "loss": 1.3074, "step": 40452 }, { "epoch": 7.71, "learning_rate": 6.890562440419447e-07, "loss": 1.4343, "step": 40453 }, { "epoch": 7.71, "learning_rate": 6.889990467111534e-07, "loss": 1.6984, "step": 40454 }, { "epoch": 7.71, "learning_rate": 6.889418493803623e-07, "loss": 1.8627, "step": 40455 }, { "epoch": 7.71, "learning_rate": 6.88884652049571e-07, "loss": 1.0452, "step": 40456 }, { "epoch": 7.71, "learning_rate": 6.888274547187798e-07, "loss": 1.5397, "step": 40457 }, { "epoch": 7.71, "learning_rate": 6.887702573879886e-07, "loss": 3.2354, "step": 40458 }, { "epoch": 7.71, "learning_rate": 6.887130600571973e-07, "loss": 2.2835, "step": 40459 }, { "epoch": 7.71, "learning_rate": 6.886558627264061e-07, "loss": 1.5575, "step": 40460 }, { "epoch": 7.71, "learning_rate": 6.885986653956149e-07, "loss": 2.853, "step": 40461 }, { "epoch": 7.71, "learning_rate": 6.885414680648237e-07, "loss": 1.886, "step": 40462 }, { "epoch": 7.71, "learning_rate": 6.884842707340325e-07, "loss": 0.8337, "step": 40463 }, { "epoch": 7.71, "learning_rate": 6.884270734032413e-07, "loss": 0.807, "step": 40464 }, { "epoch": 7.71, "learning_rate": 6.883698760724499e-07, "loss": 0.9019, "step": 40465 }, { "epoch": 7.71, "learning_rate": 6.883126787416587e-07, "loss": 1.8669, "step": 40466 }, { "epoch": 7.71, "learning_rate": 6.882554814108675e-07, "loss": 0.8927, "step": 40467 }, { "epoch": 7.71, "learning_rate": 6.881982840800762e-07, "loss": 1.9384, "step": 40468 }, { "epoch": 7.71, "learning_rate": 6.881410867492851e-07, "loss": 1.0718, "step": 40469 }, { "epoch": 7.71, "learning_rate": 6.880838894184938e-07, "loss": 0.8536, "step": 40470 }, { "epoch": 7.71, "learning_rate": 6.880266920877026e-07, "loss": 2.1739, "step": 40471 }, { "epoch": 7.71, "learning_rate": 6.879694947569114e-07, "loss": 0.6862, "step": 40472 }, { "epoch": 7.71, "learning_rate": 6.879122974261201e-07, "loss": 3.6594, "step": 40473 }, { "epoch": 7.71, "learning_rate": 6.878551000953289e-07, "loss": 2.0759, "step": 40474 }, { "epoch": 7.71, "learning_rate": 6.877979027645377e-07, "loss": 1.7265, "step": 40475 }, { "epoch": 7.71, "learning_rate": 6.877407054337465e-07, "loss": 1.2466, "step": 40476 }, { "epoch": 7.71, "learning_rate": 6.876835081029553e-07, "loss": 1.8987, "step": 40477 }, { "epoch": 7.71, "learning_rate": 6.87626310772164e-07, "loss": 1.6368, "step": 40478 }, { "epoch": 7.71, "learning_rate": 6.875691134413727e-07, "loss": 0.8628, "step": 40479 }, { "epoch": 7.71, "learning_rate": 6.875119161105815e-07, "loss": 1.7086, "step": 40480 }, { "epoch": 7.71, "learning_rate": 6.874547187797903e-07, "loss": 0.5805, "step": 40481 }, { "epoch": 7.71, "learning_rate": 6.87397521448999e-07, "loss": 2.3687, "step": 40482 }, { "epoch": 7.71, "learning_rate": 6.873403241182078e-07, "loss": 2.9621, "step": 40483 }, { "epoch": 7.71, "learning_rate": 6.872831267874166e-07, "loss": 0.5501, "step": 40484 }, { "epoch": 7.71, "learning_rate": 6.872259294566254e-07, "loss": 1.9896, "step": 40485 }, { "epoch": 7.71, "learning_rate": 6.871687321258342e-07, "loss": 1.1263, "step": 40486 }, { "epoch": 7.71, "learning_rate": 6.871115347950429e-07, "loss": 2.8668, "step": 40487 }, { "epoch": 7.71, "learning_rate": 6.870543374642517e-07, "loss": 2.5648, "step": 40488 }, { "epoch": 7.71, "learning_rate": 6.869971401334605e-07, "loss": 0.9928, "step": 40489 }, { "epoch": 7.71, "learning_rate": 6.869399428026692e-07, "loss": 1.4039, "step": 40490 }, { "epoch": 7.71, "learning_rate": 6.868827454718781e-07, "loss": 1.0577, "step": 40491 }, { "epoch": 7.71, "learning_rate": 6.868255481410867e-07, "loss": 1.782, "step": 40492 }, { "epoch": 7.71, "learning_rate": 6.867683508102955e-07, "loss": 2.7153, "step": 40493 }, { "epoch": 7.71, "learning_rate": 6.867111534795043e-07, "loss": 1.3167, "step": 40494 }, { "epoch": 7.71, "learning_rate": 6.86653956148713e-07, "loss": 1.5043, "step": 40495 }, { "epoch": 7.71, "learning_rate": 6.865967588179218e-07, "loss": 0.3182, "step": 40496 }, { "epoch": 7.71, "learning_rate": 6.865395614871306e-07, "loss": 2.0613, "step": 40497 }, { "epoch": 7.71, "learning_rate": 6.864823641563394e-07, "loss": 1.6634, "step": 40498 }, { "epoch": 7.71, "learning_rate": 6.864251668255482e-07, "loss": 2.1545, "step": 40499 }, { "epoch": 7.71, "learning_rate": 6.86367969494757e-07, "loss": 1.5291, "step": 40500 }, { "epoch": 7.71, "learning_rate": 6.863107721639657e-07, "loss": 0.6768, "step": 40501 }, { "epoch": 7.71, "learning_rate": 6.862535748331745e-07, "loss": 0.596, "step": 40502 }, { "epoch": 7.71, "learning_rate": 6.861963775023833e-07, "loss": 1.3736, "step": 40503 }, { "epoch": 7.72, "learning_rate": 6.86139180171592e-07, "loss": 2.3177, "step": 40504 }, { "epoch": 7.72, "learning_rate": 6.860819828408009e-07, "loss": 0.6167, "step": 40505 }, { "epoch": 7.72, "learning_rate": 6.860247855100095e-07, "loss": 0.8746, "step": 40506 }, { "epoch": 7.72, "learning_rate": 6.859675881792183e-07, "loss": 1.0727, "step": 40507 }, { "epoch": 7.72, "learning_rate": 6.859103908484271e-07, "loss": 1.9197, "step": 40508 }, { "epoch": 7.72, "learning_rate": 6.858531935176358e-07, "loss": 2.3291, "step": 40509 }, { "epoch": 7.72, "learning_rate": 6.857959961868446e-07, "loss": 1.8771, "step": 40510 }, { "epoch": 7.72, "learning_rate": 6.857387988560534e-07, "loss": 2.1916, "step": 40511 }, { "epoch": 7.72, "learning_rate": 6.856816015252621e-07, "loss": 1.1461, "step": 40512 }, { "epoch": 7.72, "learning_rate": 6.85624404194471e-07, "loss": 3.2948, "step": 40513 }, { "epoch": 7.72, "learning_rate": 6.855672068636797e-07, "loss": 2.7892, "step": 40514 }, { "epoch": 7.72, "learning_rate": 6.855100095328885e-07, "loss": 2.1209, "step": 40515 }, { "epoch": 7.72, "learning_rate": 6.854528122020973e-07, "loss": 0.7292, "step": 40516 }, { "epoch": 7.72, "learning_rate": 6.853956148713061e-07, "loss": 2.7993, "step": 40517 }, { "epoch": 7.72, "learning_rate": 6.853384175405147e-07, "loss": 0.6319, "step": 40518 }, { "epoch": 7.72, "learning_rate": 6.852812202097235e-07, "loss": 2.8745, "step": 40519 }, { "epoch": 7.72, "learning_rate": 6.852240228789323e-07, "loss": 1.9394, "step": 40520 }, { "epoch": 7.72, "learning_rate": 6.851668255481411e-07, "loss": 2.0287, "step": 40521 }, { "epoch": 7.72, "learning_rate": 6.851096282173499e-07, "loss": 4.0997, "step": 40522 }, { "epoch": 7.72, "learning_rate": 6.850524308865586e-07, "loss": 1.08, "step": 40523 }, { "epoch": 7.72, "learning_rate": 6.849952335557674e-07, "loss": 2.5601, "step": 40524 }, { "epoch": 7.72, "learning_rate": 6.849380362249762e-07, "loss": 1.4267, "step": 40525 }, { "epoch": 7.72, "learning_rate": 6.848808388941849e-07, "loss": 1.3592, "step": 40526 }, { "epoch": 7.72, "learning_rate": 6.848236415633938e-07, "loss": 0.2495, "step": 40527 }, { "epoch": 7.72, "learning_rate": 6.847664442326025e-07, "loss": 1.3483, "step": 40528 }, { "epoch": 7.72, "learning_rate": 6.847092469018113e-07, "loss": 2.59, "step": 40529 }, { "epoch": 7.72, "learning_rate": 6.846520495710201e-07, "loss": 2.694, "step": 40530 }, { "epoch": 7.72, "learning_rate": 6.845948522402288e-07, "loss": 1.446, "step": 40531 }, { "epoch": 7.72, "learning_rate": 6.845376549094375e-07, "loss": 1.6226, "step": 40532 }, { "epoch": 7.72, "learning_rate": 6.844804575786463e-07, "loss": 2.4616, "step": 40533 }, { "epoch": 7.72, "learning_rate": 6.844232602478551e-07, "loss": 1.6377, "step": 40534 }, { "epoch": 7.72, "learning_rate": 6.843660629170639e-07, "loss": 1.6006, "step": 40535 }, { "epoch": 7.72, "learning_rate": 6.843088655862726e-07, "loss": 0.7372, "step": 40536 }, { "epoch": 7.72, "learning_rate": 6.842516682554814e-07, "loss": 1.7871, "step": 40537 }, { "epoch": 7.72, "learning_rate": 6.841944709246902e-07, "loss": 2.6174, "step": 40538 }, { "epoch": 7.72, "learning_rate": 6.84137273593899e-07, "loss": 1.6377, "step": 40539 }, { "epoch": 7.72, "learning_rate": 6.840800762631077e-07, "loss": 2.2124, "step": 40540 }, { "epoch": 7.72, "learning_rate": 6.840228789323166e-07, "loss": 2.287, "step": 40541 }, { "epoch": 7.72, "learning_rate": 6.839656816015253e-07, "loss": 1.341, "step": 40542 }, { "epoch": 7.72, "learning_rate": 6.839084842707341e-07, "loss": 0.62, "step": 40543 }, { "epoch": 7.72, "learning_rate": 6.838512869399429e-07, "loss": 2.5241, "step": 40544 }, { "epoch": 7.72, "learning_rate": 6.837940896091515e-07, "loss": 1.9192, "step": 40545 }, { "epoch": 7.72, "learning_rate": 6.837368922783603e-07, "loss": 1.5231, "step": 40546 }, { "epoch": 7.72, "learning_rate": 6.836796949475691e-07, "loss": 2.2225, "step": 40547 }, { "epoch": 7.72, "learning_rate": 6.836224976167778e-07, "loss": 1.851, "step": 40548 }, { "epoch": 7.72, "learning_rate": 6.835653002859867e-07, "loss": 1.1115, "step": 40549 }, { "epoch": 7.72, "learning_rate": 6.835081029551954e-07, "loss": 2.3132, "step": 40550 }, { "epoch": 7.72, "learning_rate": 6.834509056244042e-07, "loss": 1.7465, "step": 40551 }, { "epoch": 7.72, "learning_rate": 6.83393708293613e-07, "loss": 1.4558, "step": 40552 }, { "epoch": 7.72, "learning_rate": 6.833365109628218e-07, "loss": 1.7487, "step": 40553 }, { "epoch": 7.72, "learning_rate": 6.832793136320305e-07, "loss": 1.5448, "step": 40554 }, { "epoch": 7.72, "learning_rate": 6.832221163012393e-07, "loss": 0.5884, "step": 40555 }, { "epoch": 7.72, "learning_rate": 6.831649189704481e-07, "loss": 1.7264, "step": 40556 }, { "epoch": 7.73, "learning_rate": 6.831077216396569e-07, "loss": 1.9644, "step": 40557 }, { "epoch": 7.73, "learning_rate": 6.830505243088657e-07, "loss": 2.1319, "step": 40558 }, { "epoch": 7.73, "learning_rate": 6.829933269780743e-07, "loss": 1.4132, "step": 40559 }, { "epoch": 7.73, "learning_rate": 6.829361296472831e-07, "loss": 1.6106, "step": 40560 }, { "epoch": 7.73, "learning_rate": 6.828789323164919e-07, "loss": 2.5484, "step": 40561 }, { "epoch": 7.73, "learning_rate": 6.828217349857006e-07, "loss": 0.437, "step": 40562 }, { "epoch": 7.73, "learning_rate": 6.827645376549095e-07, "loss": 1.853, "step": 40563 }, { "epoch": 7.73, "learning_rate": 6.827073403241182e-07, "loss": 2.151, "step": 40564 }, { "epoch": 7.73, "learning_rate": 6.82650142993327e-07, "loss": 1.3008, "step": 40565 }, { "epoch": 7.73, "learning_rate": 6.825929456625358e-07, "loss": 1.386, "step": 40566 }, { "epoch": 7.73, "learning_rate": 6.825357483317445e-07, "loss": 0.84, "step": 40567 }, { "epoch": 7.73, "learning_rate": 6.824785510009533e-07, "loss": 1.9618, "step": 40568 }, { "epoch": 7.73, "learning_rate": 6.824213536701621e-07, "loss": 0.6408, "step": 40569 }, { "epoch": 7.73, "learning_rate": 6.823641563393709e-07, "loss": 1.9029, "step": 40570 }, { "epoch": 7.73, "learning_rate": 6.823069590085797e-07, "loss": 1.1923, "step": 40571 }, { "epoch": 7.73, "learning_rate": 6.822497616777883e-07, "loss": 2.1853, "step": 40572 }, { "epoch": 7.73, "learning_rate": 6.821925643469971e-07, "loss": 2.1056, "step": 40573 }, { "epoch": 7.73, "learning_rate": 6.821353670162059e-07, "loss": 3.058, "step": 40574 }, { "epoch": 7.73, "learning_rate": 6.820781696854147e-07, "loss": 1.0107, "step": 40575 }, { "epoch": 7.73, "learning_rate": 6.820209723546234e-07, "loss": 1.3557, "step": 40576 }, { "epoch": 7.73, "learning_rate": 6.819637750238322e-07, "loss": 1.5934, "step": 40577 }, { "epoch": 7.73, "learning_rate": 6.81906577693041e-07, "loss": 0.3469, "step": 40578 }, { "epoch": 7.73, "learning_rate": 6.818493803622498e-07, "loss": 1.0702, "step": 40579 }, { "epoch": 7.73, "learning_rate": 6.817921830314586e-07, "loss": 1.3457, "step": 40580 }, { "epoch": 7.73, "learning_rate": 6.817349857006673e-07, "loss": 1.9547, "step": 40581 }, { "epoch": 7.73, "learning_rate": 6.816777883698762e-07, "loss": 1.1748, "step": 40582 }, { "epoch": 7.73, "learning_rate": 6.816205910390849e-07, "loss": 1.8412, "step": 40583 }, { "epoch": 7.73, "learning_rate": 6.815633937082936e-07, "loss": 1.9866, "step": 40584 }, { "epoch": 7.73, "learning_rate": 6.815061963775025e-07, "loss": 2.4121, "step": 40585 }, { "epoch": 7.73, "learning_rate": 6.814489990467111e-07, "loss": 0.6484, "step": 40586 }, { "epoch": 7.73, "learning_rate": 6.813918017159199e-07, "loss": 1.92, "step": 40587 }, { "epoch": 7.73, "learning_rate": 6.813346043851287e-07, "loss": 1.3082, "step": 40588 }, { "epoch": 7.73, "learning_rate": 6.812774070543374e-07, "loss": 0.6301, "step": 40589 }, { "epoch": 7.73, "learning_rate": 6.812202097235462e-07, "loss": 1.3723, "step": 40590 }, { "epoch": 7.73, "learning_rate": 6.81163012392755e-07, "loss": 1.8647, "step": 40591 }, { "epoch": 7.73, "learning_rate": 6.811058150619638e-07, "loss": 0.5748, "step": 40592 }, { "epoch": 7.73, "learning_rate": 6.810486177311726e-07, "loss": 1.4082, "step": 40593 }, { "epoch": 7.73, "learning_rate": 6.809914204003814e-07, "loss": 2.092, "step": 40594 }, { "epoch": 7.73, "learning_rate": 6.809342230695901e-07, "loss": 2.5928, "step": 40595 }, { "epoch": 7.73, "learning_rate": 6.808770257387989e-07, "loss": 2.7755, "step": 40596 }, { "epoch": 7.73, "learning_rate": 6.808198284080077e-07, "loss": 0.9615, "step": 40597 }, { "epoch": 7.73, "learning_rate": 6.807626310772163e-07, "loss": 0.911, "step": 40598 }, { "epoch": 7.73, "learning_rate": 6.807054337464252e-07, "loss": 1.1277, "step": 40599 }, { "epoch": 7.73, "learning_rate": 6.806482364156339e-07, "loss": 2.1423, "step": 40600 }, { "epoch": 7.73, "learning_rate": 6.805910390848427e-07, "loss": 2.8763, "step": 40601 }, { "epoch": 7.73, "learning_rate": 6.805338417540515e-07, "loss": 1.8449, "step": 40602 }, { "epoch": 7.73, "learning_rate": 6.804766444232602e-07, "loss": 2.1373, "step": 40603 }, { "epoch": 7.73, "learning_rate": 6.804194470924691e-07, "loss": 1.2426, "step": 40604 }, { "epoch": 7.73, "learning_rate": 6.803622497616778e-07, "loss": 2.7634, "step": 40605 }, { "epoch": 7.73, "learning_rate": 6.803050524308866e-07, "loss": 0.7405, "step": 40606 }, { "epoch": 7.73, "learning_rate": 6.802478551000954e-07, "loss": 1.0296, "step": 40607 }, { "epoch": 7.73, "learning_rate": 6.801906577693041e-07, "loss": 1.7265, "step": 40608 }, { "epoch": 7.74, "learning_rate": 6.801334604385129e-07, "loss": 1.6026, "step": 40609 }, { "epoch": 7.74, "learning_rate": 6.800762631077217e-07, "loss": 2.1088, "step": 40610 }, { "epoch": 7.74, "learning_rate": 6.800190657769305e-07, "loss": 2.4939, "step": 40611 }, { "epoch": 7.74, "learning_rate": 6.799618684461391e-07, "loss": 1.9146, "step": 40612 }, { "epoch": 7.74, "learning_rate": 6.799046711153479e-07, "loss": 2.8431, "step": 40613 }, { "epoch": 7.74, "learning_rate": 6.798474737845567e-07, "loss": 2.1308, "step": 40614 }, { "epoch": 7.74, "learning_rate": 6.797902764537655e-07, "loss": 0.6947, "step": 40615 }, { "epoch": 7.74, "learning_rate": 6.797330791229743e-07, "loss": 0.9528, "step": 40616 }, { "epoch": 7.74, "learning_rate": 6.79675881792183e-07, "loss": 2.394, "step": 40617 }, { "epoch": 7.74, "learning_rate": 6.796186844613919e-07, "loss": 3.3569, "step": 40618 }, { "epoch": 7.74, "learning_rate": 6.795614871306006e-07, "loss": 1.8119, "step": 40619 }, { "epoch": 7.74, "learning_rate": 6.795042897998093e-07, "loss": 1.6233, "step": 40620 }, { "epoch": 7.74, "learning_rate": 6.794470924690182e-07, "loss": 0.8168, "step": 40621 }, { "epoch": 7.74, "learning_rate": 6.793898951382269e-07, "loss": 2.755, "step": 40622 }, { "epoch": 7.74, "learning_rate": 6.793326978074357e-07, "loss": 1.9327, "step": 40623 }, { "epoch": 7.74, "learning_rate": 6.792755004766445e-07, "loss": 1.2878, "step": 40624 }, { "epoch": 7.74, "learning_rate": 6.792183031458531e-07, "loss": 0.8068, "step": 40625 }, { "epoch": 7.74, "learning_rate": 6.79161105815062e-07, "loss": 1.9847, "step": 40626 }, { "epoch": 7.74, "learning_rate": 6.791039084842707e-07, "loss": 2.6861, "step": 40627 }, { "epoch": 7.74, "learning_rate": 6.790467111534795e-07, "loss": 3.1067, "step": 40628 }, { "epoch": 7.74, "learning_rate": 6.789895138226883e-07, "loss": 0.8799, "step": 40629 }, { "epoch": 7.74, "learning_rate": 6.78932316491897e-07, "loss": 2.7299, "step": 40630 }, { "epoch": 7.74, "learning_rate": 6.788751191611058e-07, "loss": 1.2228, "step": 40631 }, { "epoch": 7.74, "learning_rate": 6.788179218303146e-07, "loss": 1.5201, "step": 40632 }, { "epoch": 7.74, "learning_rate": 6.787607244995234e-07, "loss": 0.5707, "step": 40633 }, { "epoch": 7.74, "learning_rate": 6.787035271687321e-07, "loss": 0.3962, "step": 40634 }, { "epoch": 7.74, "learning_rate": 6.78646329837941e-07, "loss": 1.7466, "step": 40635 }, { "epoch": 7.74, "learning_rate": 6.785891325071497e-07, "loss": 0.5532, "step": 40636 }, { "epoch": 7.74, "learning_rate": 6.785319351763585e-07, "loss": 4.1628, "step": 40637 }, { "epoch": 7.74, "learning_rate": 6.784747378455673e-07, "loss": 1.2681, "step": 40638 }, { "epoch": 7.74, "learning_rate": 6.784175405147759e-07, "loss": 2.1784, "step": 40639 }, { "epoch": 7.74, "learning_rate": 6.783603431839848e-07, "loss": 1.6293, "step": 40640 }, { "epoch": 7.74, "learning_rate": 6.783031458531935e-07, "loss": 0.6196, "step": 40641 }, { "epoch": 7.74, "learning_rate": 6.782459485224022e-07, "loss": 2.3316, "step": 40642 }, { "epoch": 7.74, "learning_rate": 6.781887511916111e-07, "loss": 0.2557, "step": 40643 }, { "epoch": 7.74, "learning_rate": 6.781315538608198e-07, "loss": 2.6429, "step": 40644 }, { "epoch": 7.74, "learning_rate": 6.780743565300286e-07, "loss": 1.451, "step": 40645 }, { "epoch": 7.74, "learning_rate": 6.780171591992374e-07, "loss": 1.5948, "step": 40646 }, { "epoch": 7.74, "learning_rate": 6.779599618684462e-07, "loss": 1.3607, "step": 40647 }, { "epoch": 7.74, "learning_rate": 6.77902764537655e-07, "loss": 1.3586, "step": 40648 }, { "epoch": 7.74, "learning_rate": 6.778455672068637e-07, "loss": 2.0388, "step": 40649 }, { "epoch": 7.74, "learning_rate": 6.777883698760725e-07, "loss": 1.8675, "step": 40650 }, { "epoch": 7.74, "learning_rate": 6.777311725452813e-07, "loss": 1.7633, "step": 40651 }, { "epoch": 7.74, "learning_rate": 6.7767397521449e-07, "loss": 0.4959, "step": 40652 }, { "epoch": 7.74, "learning_rate": 6.776167778836987e-07, "loss": 0.627, "step": 40653 }, { "epoch": 7.74, "learning_rate": 6.775595805529075e-07, "loss": 1.3466, "step": 40654 }, { "epoch": 7.74, "learning_rate": 6.775023832221163e-07, "loss": 1.5625, "step": 40655 }, { "epoch": 7.74, "learning_rate": 6.77445185891325e-07, "loss": 2.1616, "step": 40656 }, { "epoch": 7.74, "learning_rate": 6.773879885605339e-07, "loss": 0.7777, "step": 40657 }, { "epoch": 7.74, "learning_rate": 6.773307912297426e-07, "loss": 1.7546, "step": 40658 }, { "epoch": 7.74, "learning_rate": 6.772735938989514e-07, "loss": 1.2771, "step": 40659 }, { "epoch": 7.74, "learning_rate": 6.772163965681602e-07, "loss": 1.842, "step": 40660 }, { "epoch": 7.74, "learning_rate": 6.771591992373689e-07, "loss": 0.8051, "step": 40661 }, { "epoch": 7.75, "learning_rate": 6.771020019065778e-07, "loss": 0.907, "step": 40662 }, { "epoch": 7.75, "learning_rate": 6.770448045757865e-07, "loss": 2.6167, "step": 40663 }, { "epoch": 7.75, "learning_rate": 6.769876072449953e-07, "loss": 0.6782, "step": 40664 }, { "epoch": 7.75, "learning_rate": 6.76930409914204e-07, "loss": 1.4268, "step": 40665 }, { "epoch": 7.75, "learning_rate": 6.768732125834127e-07, "loss": 1.4868, "step": 40666 }, { "epoch": 7.75, "learning_rate": 6.768160152526215e-07, "loss": 0.3768, "step": 40667 }, { "epoch": 7.75, "learning_rate": 6.767588179218303e-07, "loss": 0.9321, "step": 40668 }, { "epoch": 7.75, "learning_rate": 6.767016205910391e-07, "loss": 1.1455, "step": 40669 }, { "epoch": 7.75, "learning_rate": 6.766444232602479e-07, "loss": 1.9548, "step": 40670 }, { "epoch": 7.75, "learning_rate": 6.765872259294567e-07, "loss": 0.7325, "step": 40671 }, { "epoch": 7.75, "learning_rate": 6.765300285986654e-07, "loss": 1.4803, "step": 40672 }, { "epoch": 7.75, "learning_rate": 6.764728312678742e-07, "loss": 1.9516, "step": 40673 }, { "epoch": 7.75, "learning_rate": 6.76415633937083e-07, "loss": 1.7251, "step": 40674 }, { "epoch": 7.75, "learning_rate": 6.763584366062917e-07, "loss": 1.3027, "step": 40675 }, { "epoch": 7.75, "learning_rate": 6.763012392755006e-07, "loss": 1.0252, "step": 40676 }, { "epoch": 7.75, "learning_rate": 6.762440419447093e-07, "loss": 2.0089, "step": 40677 }, { "epoch": 7.75, "learning_rate": 6.76186844613918e-07, "loss": 1.258, "step": 40678 }, { "epoch": 7.75, "learning_rate": 6.761296472831268e-07, "loss": 1.5286, "step": 40679 }, { "epoch": 7.75, "learning_rate": 6.760724499523355e-07, "loss": 1.2327, "step": 40680 }, { "epoch": 7.75, "learning_rate": 6.760152526215443e-07, "loss": 1.6969, "step": 40681 }, { "epoch": 7.75, "learning_rate": 6.759580552907531e-07, "loss": 2.1414, "step": 40682 }, { "epoch": 7.75, "learning_rate": 6.759008579599619e-07, "loss": 2.0902, "step": 40683 }, { "epoch": 7.75, "learning_rate": 6.758436606291707e-07, "loss": 1.3025, "step": 40684 }, { "epoch": 7.75, "learning_rate": 6.757864632983794e-07, "loss": 2.1836, "step": 40685 }, { "epoch": 7.75, "learning_rate": 6.757292659675882e-07, "loss": 1.3983, "step": 40686 }, { "epoch": 7.75, "learning_rate": 6.75672068636797e-07, "loss": 1.9029, "step": 40687 }, { "epoch": 7.75, "learning_rate": 6.756148713060058e-07, "loss": 0.8486, "step": 40688 }, { "epoch": 7.75, "learning_rate": 6.755576739752145e-07, "loss": 1.9059, "step": 40689 }, { "epoch": 7.75, "learning_rate": 6.755004766444233e-07, "loss": 2.1954, "step": 40690 }, { "epoch": 7.75, "learning_rate": 6.754432793136321e-07, "loss": 1.4076, "step": 40691 }, { "epoch": 7.75, "learning_rate": 6.753860819828408e-07, "loss": 1.399, "step": 40692 }, { "epoch": 7.75, "learning_rate": 6.753288846520496e-07, "loss": 1.7953, "step": 40693 }, { "epoch": 7.75, "learning_rate": 6.752716873212583e-07, "loss": 2.728, "step": 40694 }, { "epoch": 7.75, "learning_rate": 6.752144899904671e-07, "loss": 0.9687, "step": 40695 }, { "epoch": 7.75, "learning_rate": 6.751572926596759e-07, "loss": 3.4369, "step": 40696 }, { "epoch": 7.75, "learning_rate": 6.751000953288846e-07, "loss": 1.2255, "step": 40697 }, { "epoch": 7.75, "learning_rate": 6.750428979980935e-07, "loss": 0.6468, "step": 40698 }, { "epoch": 7.75, "learning_rate": 6.749857006673022e-07, "loss": 1.2022, "step": 40699 }, { "epoch": 7.75, "learning_rate": 6.74928503336511e-07, "loss": 1.5411, "step": 40700 }, { "epoch": 7.75, "learning_rate": 6.748713060057198e-07, "loss": 3.2399, "step": 40701 }, { "epoch": 7.75, "learning_rate": 6.748141086749285e-07, "loss": 1.9449, "step": 40702 }, { "epoch": 7.75, "learning_rate": 6.747569113441373e-07, "loss": 1.1888, "step": 40703 }, { "epoch": 7.75, "learning_rate": 6.746997140133461e-07, "loss": 1.9022, "step": 40704 }, { "epoch": 7.75, "learning_rate": 6.746425166825549e-07, "loss": 2.0296, "step": 40705 }, { "epoch": 7.75, "learning_rate": 6.745853193517636e-07, "loss": 1.5144, "step": 40706 }, { "epoch": 7.75, "learning_rate": 6.745281220209723e-07, "loss": 1.2134, "step": 40707 }, { "epoch": 7.75, "learning_rate": 6.744709246901811e-07, "loss": 1.2461, "step": 40708 }, { "epoch": 7.75, "learning_rate": 6.744137273593899e-07, "loss": 1.1338, "step": 40709 }, { "epoch": 7.75, "learning_rate": 6.743565300285987e-07, "loss": 1.7871, "step": 40710 }, { "epoch": 7.75, "learning_rate": 6.742993326978074e-07, "loss": 1.3809, "step": 40711 }, { "epoch": 7.75, "learning_rate": 6.742421353670163e-07, "loss": 1.3421, "step": 40712 }, { "epoch": 7.75, "learning_rate": 6.74184938036225e-07, "loss": 1.3014, "step": 40713 }, { "epoch": 7.76, "learning_rate": 6.741277407054338e-07, "loss": 1.2396, "step": 40714 }, { "epoch": 7.76, "learning_rate": 6.740705433746426e-07, "loss": 2.3831, "step": 40715 }, { "epoch": 7.76, "learning_rate": 6.740133460438513e-07, "loss": 0.209, "step": 40716 }, { "epoch": 7.76, "learning_rate": 6.739561487130601e-07, "loss": 2.2412, "step": 40717 }, { "epoch": 7.76, "learning_rate": 6.738989513822689e-07, "loss": 0.8662, "step": 40718 }, { "epoch": 7.76, "learning_rate": 6.738417540514775e-07, "loss": 1.9517, "step": 40719 }, { "epoch": 7.76, "learning_rate": 6.737845567206864e-07, "loss": 1.2611, "step": 40720 }, { "epoch": 7.76, "learning_rate": 6.737273593898951e-07, "loss": 1.9643, "step": 40721 }, { "epoch": 7.76, "learning_rate": 6.736701620591039e-07, "loss": 0.7529, "step": 40722 }, { "epoch": 7.76, "learning_rate": 6.736129647283127e-07, "loss": 2.3156, "step": 40723 }, { "epoch": 7.76, "learning_rate": 6.735557673975215e-07, "loss": 1.6123, "step": 40724 }, { "epoch": 7.76, "learning_rate": 6.734985700667302e-07, "loss": 2.0646, "step": 40725 }, { "epoch": 7.76, "learning_rate": 6.73441372735939e-07, "loss": 2.3594, "step": 40726 }, { "epoch": 7.76, "learning_rate": 6.733841754051478e-07, "loss": 1.267, "step": 40727 }, { "epoch": 7.76, "learning_rate": 6.733269780743566e-07, "loss": 0.4906, "step": 40728 }, { "epoch": 7.76, "learning_rate": 6.732697807435654e-07, "loss": 1.2624, "step": 40729 }, { "epoch": 7.76, "learning_rate": 6.732125834127741e-07, "loss": 2.034, "step": 40730 }, { "epoch": 7.76, "learning_rate": 6.731553860819829e-07, "loss": 0.8022, "step": 40731 }, { "epoch": 7.76, "learning_rate": 6.730981887511916e-07, "loss": 1.8669, "step": 40732 }, { "epoch": 7.76, "learning_rate": 6.730409914204003e-07, "loss": 1.083, "step": 40733 }, { "epoch": 7.76, "learning_rate": 6.729837940896092e-07, "loss": 1.614, "step": 40734 }, { "epoch": 7.76, "learning_rate": 6.729265967588179e-07, "loss": 2.1266, "step": 40735 }, { "epoch": 7.76, "learning_rate": 6.728693994280267e-07, "loss": 1.9209, "step": 40736 }, { "epoch": 7.76, "learning_rate": 6.728122020972355e-07, "loss": 0.7416, "step": 40737 }, { "epoch": 7.76, "learning_rate": 6.727550047664442e-07, "loss": 1.3866, "step": 40738 }, { "epoch": 7.76, "learning_rate": 6.72697807435653e-07, "loss": 0.7351, "step": 40739 }, { "epoch": 7.76, "learning_rate": 6.726406101048618e-07, "loss": 1.0062, "step": 40740 }, { "epoch": 7.76, "learning_rate": 6.725834127740706e-07, "loss": 1.9551, "step": 40741 }, { "epoch": 7.76, "learning_rate": 6.725262154432794e-07, "loss": 1.5591, "step": 40742 }, { "epoch": 7.76, "learning_rate": 6.724690181124881e-07, "loss": 0.8468, "step": 40743 }, { "epoch": 7.76, "learning_rate": 6.724118207816969e-07, "loss": 1.232, "step": 40744 }, { "epoch": 7.76, "learning_rate": 6.723546234509057e-07, "loss": 1.2728, "step": 40745 }, { "epoch": 7.76, "learning_rate": 6.722974261201144e-07, "loss": 0.8968, "step": 40746 }, { "epoch": 7.76, "learning_rate": 6.722402287893231e-07, "loss": 1.1067, "step": 40747 }, { "epoch": 7.76, "learning_rate": 6.72183031458532e-07, "loss": 1.1059, "step": 40748 }, { "epoch": 7.76, "learning_rate": 6.721258341277407e-07, "loss": 1.6572, "step": 40749 }, { "epoch": 7.76, "learning_rate": 6.720686367969495e-07, "loss": 2.8778, "step": 40750 }, { "epoch": 7.76, "learning_rate": 6.720114394661583e-07, "loss": 1.3737, "step": 40751 }, { "epoch": 7.76, "learning_rate": 6.71954242135367e-07, "loss": 2.5697, "step": 40752 }, { "epoch": 7.76, "learning_rate": 6.718970448045758e-07, "loss": 2.1721, "step": 40753 }, { "epoch": 7.76, "learning_rate": 6.718398474737846e-07, "loss": 0.5951, "step": 40754 }, { "epoch": 7.76, "learning_rate": 6.717826501429933e-07, "loss": 0.7986, "step": 40755 }, { "epoch": 7.76, "learning_rate": 6.717254528122022e-07, "loss": 1.0295, "step": 40756 }, { "epoch": 7.76, "learning_rate": 6.716682554814109e-07, "loss": 0.9651, "step": 40757 }, { "epoch": 7.76, "learning_rate": 6.716110581506197e-07, "loss": 0.8621, "step": 40758 }, { "epoch": 7.76, "learning_rate": 6.715538608198284e-07, "loss": 1.5684, "step": 40759 }, { "epoch": 7.76, "learning_rate": 6.714966634890371e-07, "loss": 1.2514, "step": 40760 }, { "epoch": 7.76, "learning_rate": 6.714394661582459e-07, "loss": 0.78, "step": 40761 }, { "epoch": 7.76, "learning_rate": 6.713822688274547e-07, "loss": 0.8409, "step": 40762 }, { "epoch": 7.76, "learning_rate": 6.713250714966635e-07, "loss": 2.849, "step": 40763 }, { "epoch": 7.76, "learning_rate": 6.712678741658723e-07, "loss": 1.1748, "step": 40764 }, { "epoch": 7.76, "learning_rate": 6.712106768350811e-07, "loss": 2.5483, "step": 40765 }, { "epoch": 7.76, "learning_rate": 6.711534795042898e-07, "loss": 0.4562, "step": 40766 }, { "epoch": 7.77, "learning_rate": 6.710962821734986e-07, "loss": 1.4212, "step": 40767 }, { "epoch": 7.77, "learning_rate": 6.710390848427074e-07, "loss": 2.0911, "step": 40768 }, { "epoch": 7.77, "learning_rate": 6.709818875119161e-07, "loss": 0.8914, "step": 40769 }, { "epoch": 7.77, "learning_rate": 6.70924690181125e-07, "loss": 1.3901, "step": 40770 }, { "epoch": 7.77, "learning_rate": 6.708674928503337e-07, "loss": 0.7731, "step": 40771 }, { "epoch": 7.77, "learning_rate": 6.708102955195424e-07, "loss": 1.2586, "step": 40772 }, { "epoch": 7.77, "learning_rate": 6.707530981887512e-07, "loss": 2.3389, "step": 40773 }, { "epoch": 7.77, "learning_rate": 6.706959008579599e-07, "loss": 0.9658, "step": 40774 }, { "epoch": 7.77, "learning_rate": 6.706387035271687e-07, "loss": 1.7137, "step": 40775 }, { "epoch": 7.77, "learning_rate": 6.705815061963775e-07, "loss": 0.9412, "step": 40776 }, { "epoch": 7.77, "learning_rate": 6.705243088655863e-07, "loss": 2.8929, "step": 40777 }, { "epoch": 7.77, "learning_rate": 6.704671115347951e-07, "loss": 1.529, "step": 40778 }, { "epoch": 7.77, "learning_rate": 6.704099142040038e-07, "loss": 1.8856, "step": 40779 }, { "epoch": 7.77, "learning_rate": 6.703527168732126e-07, "loss": 1.8205, "step": 40780 }, { "epoch": 7.77, "learning_rate": 6.702955195424214e-07, "loss": 1.4885, "step": 40781 }, { "epoch": 7.77, "learning_rate": 6.702383222116302e-07, "loss": 1.673, "step": 40782 }, { "epoch": 7.77, "learning_rate": 6.701811248808389e-07, "loss": 0.781, "step": 40783 }, { "epoch": 7.77, "learning_rate": 6.701239275500477e-07, "loss": 1.1502, "step": 40784 }, { "epoch": 7.77, "learning_rate": 6.700667302192565e-07, "loss": 1.6474, "step": 40785 }, { "epoch": 7.77, "learning_rate": 6.700095328884652e-07, "loss": 1.2582, "step": 40786 }, { "epoch": 7.77, "learning_rate": 6.69952335557674e-07, "loss": 0.9374, "step": 40787 }, { "epoch": 7.77, "learning_rate": 6.698951382268827e-07, "loss": 1.746, "step": 40788 }, { "epoch": 7.77, "learning_rate": 6.698379408960915e-07, "loss": 1.6086, "step": 40789 }, { "epoch": 7.77, "learning_rate": 6.697807435653003e-07, "loss": 1.5949, "step": 40790 }, { "epoch": 7.77, "learning_rate": 6.69723546234509e-07, "loss": 1.9334, "step": 40791 }, { "epoch": 7.77, "learning_rate": 6.696663489037179e-07, "loss": 1.4466, "step": 40792 }, { "epoch": 7.77, "learning_rate": 6.696091515729266e-07, "loss": 2.2855, "step": 40793 }, { "epoch": 7.77, "learning_rate": 6.695519542421354e-07, "loss": 0.2636, "step": 40794 }, { "epoch": 7.77, "learning_rate": 6.694947569113442e-07, "loss": 1.9237, "step": 40795 }, { "epoch": 7.77, "learning_rate": 6.694375595805529e-07, "loss": 1.7527, "step": 40796 }, { "epoch": 7.77, "learning_rate": 6.693803622497617e-07, "loss": 2.0264, "step": 40797 }, { "epoch": 7.77, "learning_rate": 6.693231649189705e-07, "loss": 2.44, "step": 40798 }, { "epoch": 7.77, "learning_rate": 6.692659675881792e-07, "loss": 1.7948, "step": 40799 }, { "epoch": 7.77, "learning_rate": 6.69208770257388e-07, "loss": 2.1787, "step": 40800 }, { "epoch": 7.77, "learning_rate": 6.691515729265968e-07, "loss": 2.1301, "step": 40801 }, { "epoch": 7.77, "learning_rate": 6.690943755958055e-07, "loss": 0.4749, "step": 40802 }, { "epoch": 7.77, "learning_rate": 6.690371782650143e-07, "loss": 1.5226, "step": 40803 }, { "epoch": 7.77, "learning_rate": 6.689799809342231e-07, "loss": 1.406, "step": 40804 }, { "epoch": 7.77, "learning_rate": 6.689227836034318e-07, "loss": 2.0722, "step": 40805 }, { "epoch": 7.77, "learning_rate": 6.688655862726407e-07, "loss": 0.8618, "step": 40806 }, { "epoch": 7.77, "learning_rate": 6.688083889418494e-07, "loss": 2.011, "step": 40807 }, { "epoch": 7.77, "learning_rate": 6.687511916110582e-07, "loss": 1.8041, "step": 40808 }, { "epoch": 7.77, "learning_rate": 6.68693994280267e-07, "loss": 0.9021, "step": 40809 }, { "epoch": 7.77, "learning_rate": 6.686367969494757e-07, "loss": 0.4896, "step": 40810 }, { "epoch": 7.77, "learning_rate": 6.685795996186845e-07, "loss": 2.0247, "step": 40811 }, { "epoch": 7.77, "learning_rate": 6.685224022878932e-07, "loss": 1.5169, "step": 40812 }, { "epoch": 7.77, "learning_rate": 6.684652049571019e-07, "loss": 1.5577, "step": 40813 }, { "epoch": 7.77, "learning_rate": 6.684080076263108e-07, "loss": 1.5897, "step": 40814 }, { "epoch": 7.77, "learning_rate": 6.683508102955195e-07, "loss": 1.3268, "step": 40815 }, { "epoch": 7.77, "learning_rate": 6.682936129647283e-07, "loss": 1.9481, "step": 40816 }, { "epoch": 7.77, "learning_rate": 6.682364156339371e-07, "loss": 1.4564, "step": 40817 }, { "epoch": 7.77, "learning_rate": 6.681792183031459e-07, "loss": 1.1269, "step": 40818 }, { "epoch": 7.78, "learning_rate": 6.681220209723546e-07, "loss": 1.9837, "step": 40819 }, { "epoch": 7.78, "learning_rate": 6.680648236415634e-07, "loss": 1.5417, "step": 40820 }, { "epoch": 7.78, "learning_rate": 6.680076263107722e-07, "loss": 0.9064, "step": 40821 }, { "epoch": 7.78, "learning_rate": 6.67950428979981e-07, "loss": 2.1284, "step": 40822 }, { "epoch": 7.78, "learning_rate": 6.678932316491898e-07, "loss": 1.9323, "step": 40823 }, { "epoch": 7.78, "learning_rate": 6.678360343183985e-07, "loss": 0.5761, "step": 40824 }, { "epoch": 7.78, "learning_rate": 6.677788369876073e-07, "loss": 0.4566, "step": 40825 }, { "epoch": 7.78, "learning_rate": 6.67721639656816e-07, "loss": 1.1978, "step": 40826 }, { "epoch": 7.78, "learning_rate": 6.676644423260247e-07, "loss": 1.4216, "step": 40827 }, { "epoch": 7.78, "learning_rate": 6.676072449952336e-07, "loss": 1.2513, "step": 40828 }, { "epoch": 7.78, "learning_rate": 6.675500476644423e-07, "loss": 1.6349, "step": 40829 }, { "epoch": 7.78, "learning_rate": 6.674928503336511e-07, "loss": 0.8182, "step": 40830 }, { "epoch": 7.78, "learning_rate": 6.674356530028599e-07, "loss": 2.2017, "step": 40831 }, { "epoch": 7.78, "learning_rate": 6.673784556720686e-07, "loss": 1.0302, "step": 40832 }, { "epoch": 7.78, "learning_rate": 6.673212583412774e-07, "loss": 1.6206, "step": 40833 }, { "epoch": 7.78, "learning_rate": 6.672640610104862e-07, "loss": 1.6885, "step": 40834 }, { "epoch": 7.78, "learning_rate": 6.67206863679695e-07, "loss": 2.3235, "step": 40835 }, { "epoch": 7.78, "learning_rate": 6.671496663489038e-07, "loss": 1.2583, "step": 40836 }, { "epoch": 7.78, "learning_rate": 6.670924690181125e-07, "loss": 1.0888, "step": 40837 }, { "epoch": 7.78, "learning_rate": 6.670352716873213e-07, "loss": 0.958, "step": 40838 }, { "epoch": 7.78, "learning_rate": 6.6697807435653e-07, "loss": 0.5295, "step": 40839 }, { "epoch": 7.78, "learning_rate": 6.669208770257388e-07, "loss": 1.8588, "step": 40840 }, { "epoch": 7.78, "learning_rate": 6.668636796949475e-07, "loss": 2.0294, "step": 40841 }, { "epoch": 7.78, "learning_rate": 6.668064823641564e-07, "loss": 1.264, "step": 40842 }, { "epoch": 7.78, "learning_rate": 6.667492850333651e-07, "loss": 1.1777, "step": 40843 }, { "epoch": 7.78, "learning_rate": 6.666920877025739e-07, "loss": 2.8872, "step": 40844 }, { "epoch": 7.78, "learning_rate": 6.666348903717827e-07, "loss": 1.1921, "step": 40845 }, { "epoch": 7.78, "learning_rate": 6.665776930409914e-07, "loss": 0.7075, "step": 40846 }, { "epoch": 7.78, "learning_rate": 6.665204957102003e-07, "loss": 1.7647, "step": 40847 }, { "epoch": 7.78, "learning_rate": 6.66463298379409e-07, "loss": 1.6251, "step": 40848 }, { "epoch": 7.78, "learning_rate": 6.664061010486177e-07, "loss": 0.8491, "step": 40849 }, { "epoch": 7.78, "learning_rate": 6.663489037178266e-07, "loss": 0.5674, "step": 40850 }, { "epoch": 7.78, "learning_rate": 6.662917063870353e-07, "loss": 1.2756, "step": 40851 }, { "epoch": 7.78, "learning_rate": 6.66234509056244e-07, "loss": 0.2994, "step": 40852 }, { "epoch": 7.78, "learning_rate": 6.661773117254528e-07, "loss": 2.3296, "step": 40853 }, { "epoch": 7.78, "learning_rate": 6.661201143946616e-07, "loss": 1.1978, "step": 40854 }, { "epoch": 7.78, "learning_rate": 6.660629170638703e-07, "loss": 1.6228, "step": 40855 }, { "epoch": 7.78, "learning_rate": 6.660057197330791e-07, "loss": 1.947, "step": 40856 }, { "epoch": 7.78, "learning_rate": 6.659485224022879e-07, "loss": 2.6458, "step": 40857 }, { "epoch": 7.78, "learning_rate": 6.658913250714967e-07, "loss": 3.3935, "step": 40858 }, { "epoch": 7.78, "learning_rate": 6.658341277407055e-07, "loss": 2.8288, "step": 40859 }, { "epoch": 7.78, "learning_rate": 6.657769304099142e-07, "loss": 3.2554, "step": 40860 }, { "epoch": 7.78, "learning_rate": 6.65719733079123e-07, "loss": 2.4255, "step": 40861 }, { "epoch": 7.78, "learning_rate": 6.656625357483318e-07, "loss": 1.1627, "step": 40862 }, { "epoch": 7.78, "learning_rate": 6.656053384175405e-07, "loss": 1.9963, "step": 40863 }, { "epoch": 7.78, "learning_rate": 6.655481410867494e-07, "loss": 1.4013, "step": 40864 }, { "epoch": 7.78, "learning_rate": 6.65490943755958e-07, "loss": 0.7209, "step": 40865 }, { "epoch": 7.78, "learning_rate": 6.654337464251668e-07, "loss": 1.9777, "step": 40866 }, { "epoch": 7.78, "learning_rate": 6.653765490943756e-07, "loss": 2.8328, "step": 40867 }, { "epoch": 7.78, "learning_rate": 6.653193517635843e-07, "loss": 1.2849, "step": 40868 }, { "epoch": 7.78, "learning_rate": 6.652621544327932e-07, "loss": 2.6605, "step": 40869 }, { "epoch": 7.78, "learning_rate": 6.652049571020019e-07, "loss": 1.5718, "step": 40870 }, { "epoch": 7.78, "learning_rate": 6.651477597712107e-07, "loss": 1.2487, "step": 40871 }, { "epoch": 7.79, "learning_rate": 6.650905624404195e-07, "loss": 1.0285, "step": 40872 }, { "epoch": 7.79, "learning_rate": 6.650333651096282e-07, "loss": 2.9773, "step": 40873 }, { "epoch": 7.79, "learning_rate": 6.64976167778837e-07, "loss": 2.1156, "step": 40874 }, { "epoch": 7.79, "learning_rate": 6.649189704480458e-07, "loss": 1.422, "step": 40875 }, { "epoch": 7.79, "learning_rate": 6.648617731172546e-07, "loss": 1.1133, "step": 40876 }, { "epoch": 7.79, "learning_rate": 6.648045757864633e-07, "loss": 1.9288, "step": 40877 }, { "epoch": 7.79, "learning_rate": 6.64747378455672e-07, "loss": 1.5383, "step": 40878 }, { "epoch": 7.79, "learning_rate": 6.646901811248808e-07, "loss": 1.822, "step": 40879 }, { "epoch": 7.79, "learning_rate": 6.646329837940896e-07, "loss": 2.1989, "step": 40880 }, { "epoch": 7.79, "learning_rate": 6.645757864632984e-07, "loss": 1.0408, "step": 40881 }, { "epoch": 7.79, "learning_rate": 6.645185891325071e-07, "loss": 1.5237, "step": 40882 }, { "epoch": 7.79, "learning_rate": 6.64461391801716e-07, "loss": 2.2057, "step": 40883 }, { "epoch": 7.79, "learning_rate": 6.644041944709247e-07, "loss": 1.621, "step": 40884 }, { "epoch": 7.79, "learning_rate": 6.643469971401334e-07, "loss": 0.9811, "step": 40885 }, { "epoch": 7.79, "learning_rate": 6.642897998093423e-07, "loss": 0.3686, "step": 40886 }, { "epoch": 7.79, "learning_rate": 6.64232602478551e-07, "loss": 2.3264, "step": 40887 }, { "epoch": 7.79, "learning_rate": 6.641754051477598e-07, "loss": 1.5566, "step": 40888 }, { "epoch": 7.79, "learning_rate": 6.641182078169686e-07, "loss": 1.2462, "step": 40889 }, { "epoch": 7.79, "learning_rate": 6.640610104861773e-07, "loss": 0.8814, "step": 40890 }, { "epoch": 7.79, "learning_rate": 6.640038131553862e-07, "loss": 3.3367, "step": 40891 }, { "epoch": 7.79, "learning_rate": 6.639466158245948e-07, "loss": 1.2813, "step": 40892 }, { "epoch": 7.79, "learning_rate": 6.638894184938036e-07, "loss": 1.1008, "step": 40893 }, { "epoch": 7.79, "learning_rate": 6.638322211630124e-07, "loss": 2.2906, "step": 40894 }, { "epoch": 7.79, "learning_rate": 6.637750238322212e-07, "loss": 0.808, "step": 40895 }, { "epoch": 7.79, "learning_rate": 6.637178265014299e-07, "loss": 3.8827, "step": 40896 }, { "epoch": 7.79, "learning_rate": 6.636606291706387e-07, "loss": 1.2508, "step": 40897 }, { "epoch": 7.79, "learning_rate": 6.636034318398475e-07, "loss": 0.8526, "step": 40898 }, { "epoch": 7.79, "learning_rate": 6.635462345090562e-07, "loss": 2.6997, "step": 40899 }, { "epoch": 7.79, "learning_rate": 6.634890371782651e-07, "loss": 1.4739, "step": 40900 }, { "epoch": 7.79, "learning_rate": 6.634318398474738e-07, "loss": 1.1869, "step": 40901 }, { "epoch": 7.79, "learning_rate": 6.633746425166826e-07, "loss": 2.2349, "step": 40902 }, { "epoch": 7.79, "learning_rate": 6.633174451858914e-07, "loss": 1.2485, "step": 40903 }, { "epoch": 7.79, "learning_rate": 6.632602478551001e-07, "loss": 1.4211, "step": 40904 }, { "epoch": 7.79, "learning_rate": 6.63203050524309e-07, "loss": 2.0772, "step": 40905 }, { "epoch": 7.79, "learning_rate": 6.631458531935176e-07, "loss": 1.729, "step": 40906 }, { "epoch": 7.79, "learning_rate": 6.630886558627264e-07, "loss": 0.8113, "step": 40907 }, { "epoch": 7.79, "learning_rate": 6.630314585319352e-07, "loss": 0.8365, "step": 40908 }, { "epoch": 7.79, "learning_rate": 6.629742612011439e-07, "loss": 0.5358, "step": 40909 }, { "epoch": 7.79, "learning_rate": 6.629170638703527e-07, "loss": 2.2077, "step": 40910 }, { "epoch": 7.79, "learning_rate": 6.628598665395615e-07, "loss": 1.6972, "step": 40911 }, { "epoch": 7.79, "learning_rate": 6.628026692087703e-07, "loss": 1.2858, "step": 40912 }, { "epoch": 7.79, "learning_rate": 6.627454718779791e-07, "loss": 2.489, "step": 40913 }, { "epoch": 7.79, "learning_rate": 6.626882745471878e-07, "loss": 1.4996, "step": 40914 }, { "epoch": 7.79, "learning_rate": 6.626310772163966e-07, "loss": 2.1241, "step": 40915 }, { "epoch": 7.79, "learning_rate": 6.625738798856054e-07, "loss": 2.0269, "step": 40916 }, { "epoch": 7.79, "learning_rate": 6.625166825548142e-07, "loss": 2.3628, "step": 40917 }, { "epoch": 7.79, "learning_rate": 6.624594852240229e-07, "loss": 1.9037, "step": 40918 }, { "epoch": 7.79, "learning_rate": 6.624022878932318e-07, "loss": 1.9675, "step": 40919 }, { "epoch": 7.79, "learning_rate": 6.623450905624404e-07, "loss": 1.0833, "step": 40920 }, { "epoch": 7.79, "learning_rate": 6.622878932316491e-07, "loss": 1.4225, "step": 40921 }, { "epoch": 7.79, "learning_rate": 6.62230695900858e-07, "loss": 1.435, "step": 40922 }, { "epoch": 7.79, "learning_rate": 6.621734985700667e-07, "loss": 1.5581, "step": 40923 }, { "epoch": 7.8, "learning_rate": 6.621163012392755e-07, "loss": 1.438, "step": 40924 }, { "epoch": 7.8, "learning_rate": 6.620591039084843e-07, "loss": 0.6465, "step": 40925 }, { "epoch": 7.8, "learning_rate": 6.62001906577693e-07, "loss": 2.289, "step": 40926 }, { "epoch": 7.8, "learning_rate": 6.619447092469019e-07, "loss": 2.1115, "step": 40927 }, { "epoch": 7.8, "learning_rate": 6.618875119161106e-07, "loss": 3.0296, "step": 40928 }, { "epoch": 7.8, "learning_rate": 6.618303145853194e-07, "loss": 1.6306, "step": 40929 }, { "epoch": 7.8, "learning_rate": 6.617731172545282e-07, "loss": 1.6169, "step": 40930 }, { "epoch": 7.8, "learning_rate": 6.61715919923737e-07, "loss": 0.8674, "step": 40931 }, { "epoch": 7.8, "learning_rate": 6.616587225929456e-07, "loss": 1.0504, "step": 40932 }, { "epoch": 7.8, "learning_rate": 6.616015252621544e-07, "loss": 1.3103, "step": 40933 }, { "epoch": 7.8, "learning_rate": 6.615443279313632e-07, "loss": 2.2605, "step": 40934 }, { "epoch": 7.8, "learning_rate": 6.614871306005719e-07, "loss": 0.9771, "step": 40935 }, { "epoch": 7.8, "learning_rate": 6.614299332697808e-07, "loss": 2.8458, "step": 40936 }, { "epoch": 7.8, "learning_rate": 6.613727359389895e-07, "loss": 1.595, "step": 40937 }, { "epoch": 7.8, "learning_rate": 6.613155386081983e-07, "loss": 0.6793, "step": 40938 }, { "epoch": 7.8, "learning_rate": 6.612583412774071e-07, "loss": 2.681, "step": 40939 }, { "epoch": 7.8, "learning_rate": 6.612011439466158e-07, "loss": 0.6777, "step": 40940 }, { "epoch": 7.8, "learning_rate": 6.611439466158247e-07, "loss": 0.5186, "step": 40941 }, { "epoch": 7.8, "learning_rate": 6.610867492850334e-07, "loss": 2.5015, "step": 40942 }, { "epoch": 7.8, "learning_rate": 6.610295519542421e-07, "loss": 1.4823, "step": 40943 }, { "epoch": 7.8, "learning_rate": 6.60972354623451e-07, "loss": 2.4586, "step": 40944 }, { "epoch": 7.8, "learning_rate": 6.609151572926597e-07, "loss": 0.6032, "step": 40945 }, { "epoch": 7.8, "learning_rate": 6.608579599618684e-07, "loss": 1.8313, "step": 40946 }, { "epoch": 7.8, "learning_rate": 6.608007626310772e-07, "loss": 1.3396, "step": 40947 }, { "epoch": 7.8, "learning_rate": 6.60743565300286e-07, "loss": 0.6788, "step": 40948 }, { "epoch": 7.8, "learning_rate": 6.606863679694948e-07, "loss": 1.2646, "step": 40949 }, { "epoch": 7.8, "learning_rate": 6.606291706387035e-07, "loss": 4.0101, "step": 40950 }, { "epoch": 7.8, "learning_rate": 6.605719733079123e-07, "loss": 0.507, "step": 40951 }, { "epoch": 7.8, "learning_rate": 6.605147759771211e-07, "loss": 0.931, "step": 40952 }, { "epoch": 7.8, "learning_rate": 6.604575786463299e-07, "loss": 3.4548, "step": 40953 }, { "epoch": 7.8, "learning_rate": 6.604003813155386e-07, "loss": 1.3752, "step": 40954 }, { "epoch": 7.8, "learning_rate": 6.603431839847474e-07, "loss": 3.0221, "step": 40955 }, { "epoch": 7.8, "learning_rate": 6.602859866539562e-07, "loss": 1.856, "step": 40956 }, { "epoch": 7.8, "learning_rate": 6.602287893231649e-07, "loss": 1.9399, "step": 40957 }, { "epoch": 7.8, "learning_rate": 6.601715919923738e-07, "loss": 1.588, "step": 40958 }, { "epoch": 7.8, "learning_rate": 6.601143946615824e-07, "loss": 0.5375, "step": 40959 }, { "epoch": 7.8, "learning_rate": 6.600571973307912e-07, "loss": 1.7815, "step": 40960 }, { "epoch": 7.8, "learning_rate": 6.6e-07, "loss": 1.9328, "step": 40961 }, { "epoch": 7.8, "learning_rate": 6.599428026692087e-07, "loss": 1.6971, "step": 40962 }, { "epoch": 7.8, "learning_rate": 6.598856053384176e-07, "loss": 1.6952, "step": 40963 }, { "epoch": 7.8, "learning_rate": 6.598284080076263e-07, "loss": 2.6362, "step": 40964 }, { "epoch": 7.8, "learning_rate": 6.597712106768351e-07, "loss": 1.3089, "step": 40965 }, { "epoch": 7.8, "learning_rate": 6.597140133460439e-07, "loss": 0.9485, "step": 40966 }, { "epoch": 7.8, "learning_rate": 6.596568160152526e-07, "loss": 1.7486, "step": 40967 }, { "epoch": 7.8, "learning_rate": 6.595996186844614e-07, "loss": 1.1318, "step": 40968 }, { "epoch": 7.8, "learning_rate": 6.595424213536702e-07, "loss": 0.77, "step": 40969 }, { "epoch": 7.8, "learning_rate": 6.59485224022879e-07, "loss": 2.1834, "step": 40970 }, { "epoch": 7.8, "learning_rate": 6.594280266920878e-07, "loss": 1.468, "step": 40971 }, { "epoch": 7.8, "learning_rate": 6.593708293612966e-07, "loss": 2.6818, "step": 40972 }, { "epoch": 7.8, "learning_rate": 6.593136320305052e-07, "loss": 1.1121, "step": 40973 }, { "epoch": 7.8, "learning_rate": 6.59256434699714e-07, "loss": 1.9271, "step": 40974 }, { "epoch": 7.8, "learning_rate": 6.591992373689228e-07, "loss": 1.7867, "step": 40975 }, { "epoch": 7.8, "learning_rate": 6.591420400381315e-07, "loss": 0.7771, "step": 40976 }, { "epoch": 7.81, "learning_rate": 6.590848427073404e-07, "loss": 1.7476, "step": 40977 }, { "epoch": 7.81, "learning_rate": 6.590276453765491e-07, "loss": 1.7769, "step": 40978 }, { "epoch": 7.81, "learning_rate": 6.589704480457578e-07, "loss": 0.9692, "step": 40979 }, { "epoch": 7.81, "learning_rate": 6.589132507149667e-07, "loss": 2.5816, "step": 40980 }, { "epoch": 7.81, "learning_rate": 6.588560533841754e-07, "loss": 0.849, "step": 40981 }, { "epoch": 7.81, "learning_rate": 6.587988560533842e-07, "loss": 1.9276, "step": 40982 }, { "epoch": 7.81, "learning_rate": 6.58741658722593e-07, "loss": 0.8518, "step": 40983 }, { "epoch": 7.81, "learning_rate": 6.586844613918018e-07, "loss": 1.9084, "step": 40984 }, { "epoch": 7.81, "learning_rate": 6.586272640610106e-07, "loss": 1.8985, "step": 40985 }, { "epoch": 7.81, "learning_rate": 6.585700667302192e-07, "loss": 1.5479, "step": 40986 }, { "epoch": 7.81, "learning_rate": 6.58512869399428e-07, "loss": 1.724, "step": 40987 }, { "epoch": 7.81, "learning_rate": 6.584556720686368e-07, "loss": 2.0283, "step": 40988 }, { "epoch": 7.81, "learning_rate": 6.583984747378456e-07, "loss": 1.6306, "step": 40989 }, { "epoch": 7.81, "learning_rate": 6.583412774070543e-07, "loss": 1.7387, "step": 40990 }, { "epoch": 7.81, "learning_rate": 6.582840800762631e-07, "loss": 1.3692, "step": 40991 }, { "epoch": 7.81, "learning_rate": 6.582268827454719e-07, "loss": 0.9626, "step": 40992 }, { "epoch": 7.81, "learning_rate": 6.581696854146807e-07, "loss": 1.246, "step": 40993 }, { "epoch": 7.81, "learning_rate": 6.581124880838895e-07, "loss": 2.5673, "step": 40994 }, { "epoch": 7.81, "learning_rate": 6.580552907530982e-07, "loss": 1.3052, "step": 40995 }, { "epoch": 7.81, "learning_rate": 6.57998093422307e-07, "loss": 1.6861, "step": 40996 }, { "epoch": 7.81, "learning_rate": 6.579408960915158e-07, "loss": 2.3002, "step": 40997 }, { "epoch": 7.81, "learning_rate": 6.578836987607245e-07, "loss": 0.9445, "step": 40998 }, { "epoch": 7.81, "learning_rate": 6.578265014299334e-07, "loss": 0.8114, "step": 40999 }, { "epoch": 7.81, "learning_rate": 6.57769304099142e-07, "loss": 0.5449, "step": 41000 }, { "epoch": 7.81, "learning_rate": 6.577121067683508e-07, "loss": 1.4895, "step": 41001 }, { "epoch": 7.81, "learning_rate": 6.576549094375596e-07, "loss": 1.0576, "step": 41002 }, { "epoch": 7.81, "learning_rate": 6.575977121067683e-07, "loss": 1.4608, "step": 41003 }, { "epoch": 7.81, "learning_rate": 6.575405147759771e-07, "loss": 2.3459, "step": 41004 }, { "epoch": 7.81, "learning_rate": 6.574833174451859e-07, "loss": 1.1365, "step": 41005 }, { "epoch": 7.81, "learning_rate": 6.574261201143947e-07, "loss": 0.7132, "step": 41006 }, { "epoch": 7.81, "learning_rate": 6.573689227836035e-07, "loss": 1.2476, "step": 41007 }, { "epoch": 7.81, "learning_rate": 6.573117254528122e-07, "loss": 1.3986, "step": 41008 }, { "epoch": 7.81, "learning_rate": 6.57254528122021e-07, "loss": 0.5913, "step": 41009 }, { "epoch": 7.81, "learning_rate": 6.571973307912298e-07, "loss": 2.3202, "step": 41010 }, { "epoch": 7.81, "learning_rate": 6.571401334604386e-07, "loss": 0.7494, "step": 41011 }, { "epoch": 7.81, "learning_rate": 6.570829361296472e-07, "loss": 1.2369, "step": 41012 }, { "epoch": 7.81, "learning_rate": 6.570257387988561e-07, "loss": 1.4141, "step": 41013 }, { "epoch": 7.81, "learning_rate": 6.569685414680648e-07, "loss": 3.0677, "step": 41014 }, { "epoch": 7.81, "learning_rate": 6.569113441372736e-07, "loss": 1.4335, "step": 41015 }, { "epoch": 7.81, "learning_rate": 6.568541468064824e-07, "loss": 0.3597, "step": 41016 }, { "epoch": 7.81, "learning_rate": 6.567969494756911e-07, "loss": 1.3761, "step": 41017 }, { "epoch": 7.81, "learning_rate": 6.567397521448999e-07, "loss": 1.1861, "step": 41018 }, { "epoch": 7.81, "learning_rate": 6.566825548141087e-07, "loss": 2.3917, "step": 41019 }, { "epoch": 7.81, "learning_rate": 6.566253574833174e-07, "loss": 1.4284, "step": 41020 }, { "epoch": 7.81, "learning_rate": 6.565681601525263e-07, "loss": 3.813, "step": 41021 }, { "epoch": 7.81, "learning_rate": 6.56510962821735e-07, "loss": 0.9965, "step": 41022 }, { "epoch": 7.81, "learning_rate": 6.564537654909438e-07, "loss": 1.7808, "step": 41023 }, { "epoch": 7.81, "learning_rate": 6.563965681601526e-07, "loss": 0.6628, "step": 41024 }, { "epoch": 7.81, "learning_rate": 6.563393708293614e-07, "loss": 2.0299, "step": 41025 }, { "epoch": 7.81, "learning_rate": 6.5628217349857e-07, "loss": 2.3947, "step": 41026 }, { "epoch": 7.81, "learning_rate": 6.562249761677788e-07, "loss": 2.0738, "step": 41027 }, { "epoch": 7.81, "learning_rate": 6.561677788369876e-07, "loss": 2.8588, "step": 41028 }, { "epoch": 7.82, "learning_rate": 6.561105815061964e-07, "loss": 1.4029, "step": 41029 }, { "epoch": 7.82, "learning_rate": 6.560533841754052e-07, "loss": 2.2211, "step": 41030 }, { "epoch": 7.82, "learning_rate": 6.559961868446139e-07, "loss": 2.1627, "step": 41031 }, { "epoch": 7.82, "learning_rate": 6.559389895138227e-07, "loss": 0.3738, "step": 41032 }, { "epoch": 7.82, "learning_rate": 6.558817921830315e-07, "loss": 2.1965, "step": 41033 }, { "epoch": 7.82, "learning_rate": 6.558245948522402e-07, "loss": 2.1945, "step": 41034 }, { "epoch": 7.82, "learning_rate": 6.557673975214491e-07, "loss": 1.7857, "step": 41035 }, { "epoch": 7.82, "learning_rate": 6.557102001906578e-07, "loss": 2.7572, "step": 41036 }, { "epoch": 7.82, "learning_rate": 6.556530028598666e-07, "loss": 1.5425, "step": 41037 }, { "epoch": 7.82, "learning_rate": 6.555958055290754e-07, "loss": 1.9167, "step": 41038 }, { "epoch": 7.82, "learning_rate": 6.55538608198284e-07, "loss": 1.4125, "step": 41039 }, { "epoch": 7.82, "learning_rate": 6.554814108674928e-07, "loss": 1.6104, "step": 41040 }, { "epoch": 7.82, "learning_rate": 6.554242135367016e-07, "loss": 2.6269, "step": 41041 }, { "epoch": 7.82, "learning_rate": 6.553670162059104e-07, "loss": 0.9151, "step": 41042 }, { "epoch": 7.82, "learning_rate": 6.553098188751192e-07, "loss": 1.7883, "step": 41043 }, { "epoch": 7.82, "learning_rate": 6.552526215443279e-07, "loss": 1.3028, "step": 41044 }, { "epoch": 7.82, "learning_rate": 6.551954242135367e-07, "loss": 1.0636, "step": 41045 }, { "epoch": 7.82, "learning_rate": 6.551382268827455e-07, "loss": 1.8673, "step": 41046 }, { "epoch": 7.82, "learning_rate": 6.550810295519543e-07, "loss": 2.0812, "step": 41047 }, { "epoch": 7.82, "learning_rate": 6.55023832221163e-07, "loss": 1.1124, "step": 41048 }, { "epoch": 7.82, "learning_rate": 6.549666348903719e-07, "loss": 2.3573, "step": 41049 }, { "epoch": 7.82, "learning_rate": 6.549094375595806e-07, "loss": 1.9471, "step": 41050 }, { "epoch": 7.82, "learning_rate": 6.548522402287894e-07, "loss": 0.9956, "step": 41051 }, { "epoch": 7.82, "learning_rate": 6.547950428979982e-07, "loss": 2.3808, "step": 41052 }, { "epoch": 7.82, "learning_rate": 6.547378455672068e-07, "loss": 1.6309, "step": 41053 }, { "epoch": 7.82, "learning_rate": 6.546806482364156e-07, "loss": 1.7203, "step": 41054 }, { "epoch": 7.82, "learning_rate": 6.546234509056244e-07, "loss": 0.8392, "step": 41055 }, { "epoch": 7.82, "learning_rate": 6.545662535748331e-07, "loss": 1.9958, "step": 41056 }, { "epoch": 7.82, "learning_rate": 6.54509056244042e-07, "loss": 3.6593, "step": 41057 }, { "epoch": 7.82, "learning_rate": 6.544518589132507e-07, "loss": 1.4587, "step": 41058 }, { "epoch": 7.82, "learning_rate": 6.543946615824595e-07, "loss": 1.1013, "step": 41059 }, { "epoch": 7.82, "learning_rate": 6.543374642516683e-07, "loss": 2.8659, "step": 41060 }, { "epoch": 7.82, "learning_rate": 6.54280266920877e-07, "loss": 1.2782, "step": 41061 }, { "epoch": 7.82, "learning_rate": 6.542230695900858e-07, "loss": 1.048, "step": 41062 }, { "epoch": 7.82, "learning_rate": 6.541658722592946e-07, "loss": 1.6597, "step": 41063 }, { "epoch": 7.82, "learning_rate": 6.541086749285034e-07, "loss": 1.4153, "step": 41064 }, { "epoch": 7.82, "learning_rate": 6.540514775977122e-07, "loss": 1.1136, "step": 41065 }, { "epoch": 7.82, "learning_rate": 6.53994280266921e-07, "loss": 1.0344, "step": 41066 }, { "epoch": 7.82, "learning_rate": 6.539370829361296e-07, "loss": 0.9741, "step": 41067 }, { "epoch": 7.82, "learning_rate": 6.538798856053384e-07, "loss": 3.0651, "step": 41068 }, { "epoch": 7.82, "learning_rate": 6.538226882745472e-07, "loss": 1.6375, "step": 41069 }, { "epoch": 7.82, "learning_rate": 6.537654909437559e-07, "loss": 0.7671, "step": 41070 }, { "epoch": 7.82, "learning_rate": 6.537082936129648e-07, "loss": 1.5107, "step": 41071 }, { "epoch": 7.82, "learning_rate": 6.536510962821735e-07, "loss": 1.2482, "step": 41072 }, { "epoch": 7.82, "learning_rate": 6.535938989513823e-07, "loss": 1.3034, "step": 41073 }, { "epoch": 7.82, "learning_rate": 6.535367016205911e-07, "loss": 1.4152, "step": 41074 }, { "epoch": 7.82, "learning_rate": 6.534795042897998e-07, "loss": 2.7864, "step": 41075 }, { "epoch": 7.82, "learning_rate": 6.534223069590086e-07, "loss": 2.8589, "step": 41076 }, { "epoch": 7.82, "learning_rate": 6.533651096282174e-07, "loss": 1.144, "step": 41077 }, { "epoch": 7.82, "learning_rate": 6.533079122974262e-07, "loss": 1.4058, "step": 41078 }, { "epoch": 7.82, "learning_rate": 6.53250714966635e-07, "loss": 3.0106, "step": 41079 }, { "epoch": 7.82, "learning_rate": 6.531935176358436e-07, "loss": 1.31, "step": 41080 }, { "epoch": 7.82, "learning_rate": 6.531363203050524e-07, "loss": 0.7164, "step": 41081 }, { "epoch": 7.83, "learning_rate": 6.530791229742612e-07, "loss": 1.5155, "step": 41082 }, { "epoch": 7.83, "learning_rate": 6.5302192564347e-07, "loss": 2.3552, "step": 41083 }, { "epoch": 7.83, "learning_rate": 6.529647283126787e-07, "loss": 1.8259, "step": 41084 }, { "epoch": 7.83, "learning_rate": 6.529075309818875e-07, "loss": 1.8519, "step": 41085 }, { "epoch": 7.83, "learning_rate": 6.528503336510963e-07, "loss": 2.6507, "step": 41086 }, { "epoch": 7.83, "learning_rate": 6.527931363203051e-07, "loss": 1.4041, "step": 41087 }, { "epoch": 7.83, "learning_rate": 6.527359389895139e-07, "loss": 1.3795, "step": 41088 }, { "epoch": 7.83, "learning_rate": 6.526787416587226e-07, "loss": 1.4766, "step": 41089 }, { "epoch": 7.83, "learning_rate": 6.526215443279315e-07, "loss": 1.432, "step": 41090 }, { "epoch": 7.83, "learning_rate": 6.525643469971402e-07, "loss": 3.1202, "step": 41091 }, { "epoch": 7.83, "learning_rate": 6.525071496663488e-07, "loss": 0.9338, "step": 41092 }, { "epoch": 7.83, "learning_rate": 6.524499523355577e-07, "loss": 1.2812, "step": 41093 }, { "epoch": 7.83, "learning_rate": 6.523927550047664e-07, "loss": 2.6802, "step": 41094 }, { "epoch": 7.83, "learning_rate": 6.523355576739752e-07, "loss": 1.7104, "step": 41095 }, { "epoch": 7.83, "learning_rate": 6.52278360343184e-07, "loss": 1.4219, "step": 41096 }, { "epoch": 7.83, "learning_rate": 6.522211630123927e-07, "loss": 0.4556, "step": 41097 }, { "epoch": 7.83, "learning_rate": 6.521639656816015e-07, "loss": 1.3795, "step": 41098 }, { "epoch": 7.83, "learning_rate": 6.521067683508103e-07, "loss": 1.2712, "step": 41099 }, { "epoch": 7.83, "learning_rate": 6.520495710200191e-07, "loss": 2.6641, "step": 41100 }, { "epoch": 7.83, "learning_rate": 6.519923736892279e-07, "loss": 1.7378, "step": 41101 }, { "epoch": 7.83, "learning_rate": 6.519351763584367e-07, "loss": 0.7918, "step": 41102 }, { "epoch": 7.83, "learning_rate": 6.518779790276454e-07, "loss": 1.6686, "step": 41103 }, { "epoch": 7.83, "learning_rate": 6.518207816968542e-07, "loss": 1.5106, "step": 41104 }, { "epoch": 7.83, "learning_rate": 6.51763584366063e-07, "loss": 1.6696, "step": 41105 }, { "epoch": 7.83, "learning_rate": 6.517063870352716e-07, "loss": 3.5768, "step": 41106 }, { "epoch": 7.83, "learning_rate": 6.516491897044805e-07, "loss": 1.1624, "step": 41107 }, { "epoch": 7.83, "learning_rate": 6.515919923736892e-07, "loss": 0.8356, "step": 41108 }, { "epoch": 7.83, "learning_rate": 6.51534795042898e-07, "loss": 2.5295, "step": 41109 }, { "epoch": 7.83, "learning_rate": 6.514775977121068e-07, "loss": 2.4569, "step": 41110 }, { "epoch": 7.83, "learning_rate": 6.514204003813155e-07, "loss": 2.2698, "step": 41111 }, { "epoch": 7.83, "learning_rate": 6.513632030505243e-07, "loss": 1.7712, "step": 41112 }, { "epoch": 7.83, "learning_rate": 6.513060057197331e-07, "loss": 1.7294, "step": 41113 }, { "epoch": 7.83, "learning_rate": 6.512488083889418e-07, "loss": 1.767, "step": 41114 }, { "epoch": 7.83, "learning_rate": 6.511916110581507e-07, "loss": 0.6477, "step": 41115 }, { "epoch": 7.83, "learning_rate": 6.511344137273594e-07, "loss": 1.9896, "step": 41116 }, { "epoch": 7.83, "learning_rate": 6.510772163965682e-07, "loss": 0.6818, "step": 41117 }, { "epoch": 7.83, "learning_rate": 6.51020019065777e-07, "loss": 0.8495, "step": 41118 }, { "epoch": 7.83, "learning_rate": 6.509628217349858e-07, "loss": 0.2825, "step": 41119 }, { "epoch": 7.83, "learning_rate": 6.509056244041944e-07, "loss": 2.0644, "step": 41120 }, { "epoch": 7.83, "learning_rate": 6.508484270734032e-07, "loss": 0.5801, "step": 41121 }, { "epoch": 7.83, "learning_rate": 6.50791229742612e-07, "loss": 2.1801, "step": 41122 }, { "epoch": 7.83, "learning_rate": 6.507340324118208e-07, "loss": 3.4755, "step": 41123 }, { "epoch": 7.83, "learning_rate": 6.506768350810296e-07, "loss": 0.6347, "step": 41124 }, { "epoch": 7.83, "learning_rate": 6.506196377502383e-07, "loss": 0.9076, "step": 41125 }, { "epoch": 7.83, "learning_rate": 6.505624404194471e-07, "loss": 2.0119, "step": 41126 }, { "epoch": 7.83, "learning_rate": 6.505052430886559e-07, "loss": 2.9057, "step": 41127 }, { "epoch": 7.83, "learning_rate": 6.504480457578646e-07, "loss": 1.1894, "step": 41128 }, { "epoch": 7.83, "learning_rate": 6.503908484270735e-07, "loss": 1.8929, "step": 41129 }, { "epoch": 7.83, "learning_rate": 6.503336510962822e-07, "loss": 0.9939, "step": 41130 }, { "epoch": 7.83, "learning_rate": 6.50276453765491e-07, "loss": 1.6712, "step": 41131 }, { "epoch": 7.83, "learning_rate": 6.502192564346998e-07, "loss": 3.827, "step": 41132 }, { "epoch": 7.83, "learning_rate": 6.501620591039084e-07, "loss": 0.4719, "step": 41133 }, { "epoch": 7.84, "learning_rate": 6.501048617731172e-07, "loss": 1.9573, "step": 41134 }, { "epoch": 7.84, "learning_rate": 6.50047664442326e-07, "loss": 1.9828, "step": 41135 }, { "epoch": 7.84, "learning_rate": 6.499904671115348e-07, "loss": 2.7052, "step": 41136 }, { "epoch": 7.84, "learning_rate": 6.499332697807436e-07, "loss": 1.667, "step": 41137 }, { "epoch": 7.84, "learning_rate": 6.498760724499523e-07, "loss": 2.838, "step": 41138 }, { "epoch": 7.84, "learning_rate": 6.498188751191611e-07, "loss": 1.947, "step": 41139 }, { "epoch": 7.84, "learning_rate": 6.497616777883699e-07, "loss": 1.1889, "step": 41140 }, { "epoch": 7.84, "learning_rate": 6.497044804575787e-07, "loss": 1.7107, "step": 41141 }, { "epoch": 7.84, "learning_rate": 6.496472831267874e-07, "loss": 1.7929, "step": 41142 }, { "epoch": 7.84, "learning_rate": 6.495900857959963e-07, "loss": 1.8263, "step": 41143 }, { "epoch": 7.84, "learning_rate": 6.49532888465205e-07, "loss": 1.8329, "step": 41144 }, { "epoch": 7.84, "learning_rate": 6.494756911344138e-07, "loss": 0.688, "step": 41145 }, { "epoch": 7.84, "learning_rate": 6.494184938036225e-07, "loss": 1.6221, "step": 41146 }, { "epoch": 7.84, "learning_rate": 6.493612964728312e-07, "loss": 1.9883, "step": 41147 }, { "epoch": 7.84, "learning_rate": 6.493040991420401e-07, "loss": 0.8241, "step": 41148 }, { "epoch": 7.84, "learning_rate": 6.492469018112488e-07, "loss": 1.4424, "step": 41149 }, { "epoch": 7.84, "learning_rate": 6.491897044804575e-07, "loss": 1.3164, "step": 41150 }, { "epoch": 7.84, "learning_rate": 6.491325071496664e-07, "loss": 1.0734, "step": 41151 }, { "epoch": 7.84, "learning_rate": 6.490753098188751e-07, "loss": 2.0926, "step": 41152 }, { "epoch": 7.84, "learning_rate": 6.490181124880839e-07, "loss": 1.1335, "step": 41153 }, { "epoch": 7.84, "learning_rate": 6.489609151572927e-07, "loss": 0.663, "step": 41154 }, { "epoch": 7.84, "learning_rate": 6.489037178265015e-07, "loss": 0.7633, "step": 41155 }, { "epoch": 7.84, "learning_rate": 6.488465204957102e-07, "loss": 1.3918, "step": 41156 }, { "epoch": 7.84, "learning_rate": 6.48789323164919e-07, "loss": 1.2836, "step": 41157 }, { "epoch": 7.84, "learning_rate": 6.487321258341278e-07, "loss": 1.2899, "step": 41158 }, { "epoch": 7.84, "learning_rate": 6.486749285033366e-07, "loss": 2.1436, "step": 41159 }, { "epoch": 7.84, "learning_rate": 6.486177311725453e-07, "loss": 1.6016, "step": 41160 }, { "epoch": 7.84, "learning_rate": 6.48560533841754e-07, "loss": 1.3886, "step": 41161 }, { "epoch": 7.84, "learning_rate": 6.485033365109628e-07, "loss": 1.2436, "step": 41162 }, { "epoch": 7.84, "learning_rate": 6.484461391801716e-07, "loss": 1.9992, "step": 41163 }, { "epoch": 7.84, "learning_rate": 6.483889418493803e-07, "loss": 0.8126, "step": 41164 }, { "epoch": 7.84, "learning_rate": 6.483317445185892e-07, "loss": 1.7056, "step": 41165 }, { "epoch": 7.84, "learning_rate": 6.482745471877979e-07, "loss": 1.9969, "step": 41166 }, { "epoch": 7.84, "learning_rate": 6.482173498570067e-07, "loss": 1.4003, "step": 41167 }, { "epoch": 7.84, "learning_rate": 6.481601525262155e-07, "loss": 2.0543, "step": 41168 }, { "epoch": 7.84, "learning_rate": 6.481029551954242e-07, "loss": 1.488, "step": 41169 }, { "epoch": 7.84, "learning_rate": 6.480457578646331e-07, "loss": 1.3743, "step": 41170 }, { "epoch": 7.84, "learning_rate": 6.479885605338418e-07, "loss": 2.6623, "step": 41171 }, { "epoch": 7.84, "learning_rate": 6.479313632030506e-07, "loss": 1.8222, "step": 41172 }, { "epoch": 7.84, "learning_rate": 6.478741658722593e-07, "loss": 0.5545, "step": 41173 }, { "epoch": 7.84, "learning_rate": 6.47816968541468e-07, "loss": 0.7544, "step": 41174 }, { "epoch": 7.84, "learning_rate": 6.477597712106768e-07, "loss": 2.2667, "step": 41175 }, { "epoch": 7.84, "learning_rate": 6.477025738798856e-07, "loss": 2.0144, "step": 41176 }, { "epoch": 7.84, "learning_rate": 6.476453765490944e-07, "loss": 2.3757, "step": 41177 }, { "epoch": 7.84, "learning_rate": 6.475881792183031e-07, "loss": 0.9914, "step": 41178 }, { "epoch": 7.84, "learning_rate": 6.475309818875119e-07, "loss": 2.501, "step": 41179 }, { "epoch": 7.84, "learning_rate": 6.474737845567207e-07, "loss": 2.1509, "step": 41180 }, { "epoch": 7.84, "learning_rate": 6.474165872259295e-07, "loss": 1.7263, "step": 41181 }, { "epoch": 7.84, "learning_rate": 6.473593898951383e-07, "loss": 1.6477, "step": 41182 }, { "epoch": 7.84, "learning_rate": 6.47302192564347e-07, "loss": 1.0119, "step": 41183 }, { "epoch": 7.84, "learning_rate": 6.472449952335559e-07, "loss": 2.5402, "step": 41184 }, { "epoch": 7.84, "learning_rate": 6.471877979027646e-07, "loss": 1.4855, "step": 41185 }, { "epoch": 7.84, "learning_rate": 6.471306005719732e-07, "loss": 1.2979, "step": 41186 }, { "epoch": 7.85, "learning_rate": 6.470734032411821e-07, "loss": 0.3466, "step": 41187 }, { "epoch": 7.85, "learning_rate": 6.470162059103908e-07, "loss": 1.4461, "step": 41188 }, { "epoch": 7.85, "learning_rate": 6.469590085795996e-07, "loss": 3.8913, "step": 41189 }, { "epoch": 7.85, "learning_rate": 6.469018112488084e-07, "loss": 1.573, "step": 41190 }, { "epoch": 7.85, "learning_rate": 6.468446139180171e-07, "loss": 1.1237, "step": 41191 }, { "epoch": 7.85, "learning_rate": 6.46787416587226e-07, "loss": 0.8713, "step": 41192 }, { "epoch": 7.85, "learning_rate": 6.467302192564347e-07, "loss": 1.2167, "step": 41193 }, { "epoch": 7.85, "learning_rate": 6.466730219256435e-07, "loss": 1.5815, "step": 41194 }, { "epoch": 7.85, "learning_rate": 6.466158245948523e-07, "loss": 0.3032, "step": 41195 }, { "epoch": 7.85, "learning_rate": 6.465586272640611e-07, "loss": 1.6716, "step": 41196 }, { "epoch": 7.85, "learning_rate": 6.465014299332698e-07, "loss": 2.2981, "step": 41197 }, { "epoch": 7.85, "learning_rate": 6.464442326024786e-07, "loss": 0.8969, "step": 41198 }, { "epoch": 7.85, "learning_rate": 6.463870352716874e-07, "loss": 2.0127, "step": 41199 }, { "epoch": 7.85, "learning_rate": 6.46329837940896e-07, "loss": 0.8666, "step": 41200 }, { "epoch": 7.85, "learning_rate": 6.462726406101049e-07, "loss": 0.6188, "step": 41201 }, { "epoch": 7.85, "learning_rate": 6.462154432793136e-07, "loss": 1.0638, "step": 41202 }, { "epoch": 7.85, "learning_rate": 6.461582459485224e-07, "loss": 1.8856, "step": 41203 }, { "epoch": 7.85, "learning_rate": 6.461010486177312e-07, "loss": 1.4345, "step": 41204 }, { "epoch": 7.85, "learning_rate": 6.460438512869399e-07, "loss": 1.4077, "step": 41205 }, { "epoch": 7.85, "learning_rate": 6.459866539561488e-07, "loss": 1.9295, "step": 41206 }, { "epoch": 7.85, "learning_rate": 6.459294566253575e-07, "loss": 3.0313, "step": 41207 }, { "epoch": 7.85, "learning_rate": 6.458722592945663e-07, "loss": 1.8182, "step": 41208 }, { "epoch": 7.85, "learning_rate": 6.458150619637751e-07, "loss": 2.284, "step": 41209 }, { "epoch": 7.85, "learning_rate": 6.457578646329838e-07, "loss": 1.6251, "step": 41210 }, { "epoch": 7.85, "learning_rate": 6.457006673021926e-07, "loss": 1.7495, "step": 41211 }, { "epoch": 7.85, "learning_rate": 6.456434699714014e-07, "loss": 1.5348, "step": 41212 }, { "epoch": 7.85, "learning_rate": 6.455862726406101e-07, "loss": 1.2877, "step": 41213 }, { "epoch": 7.85, "learning_rate": 6.455290753098189e-07, "loss": 1.597, "step": 41214 }, { "epoch": 7.85, "learning_rate": 6.454718779790276e-07, "loss": 1.5076, "step": 41215 }, { "epoch": 7.85, "learning_rate": 6.454146806482364e-07, "loss": 1.6318, "step": 41216 }, { "epoch": 7.85, "learning_rate": 6.453574833174452e-07, "loss": 0.9352, "step": 41217 }, { "epoch": 7.85, "learning_rate": 6.45300285986654e-07, "loss": 1.5142, "step": 41218 }, { "epoch": 7.85, "learning_rate": 6.452430886558627e-07, "loss": 1.6661, "step": 41219 }, { "epoch": 7.85, "learning_rate": 6.451858913250716e-07, "loss": 1.3689, "step": 41220 }, { "epoch": 7.85, "learning_rate": 6.451286939942803e-07, "loss": 1.3368, "step": 41221 }, { "epoch": 7.85, "learning_rate": 6.45071496663489e-07, "loss": 1.8463, "step": 41222 }, { "epoch": 7.85, "learning_rate": 6.450142993326979e-07, "loss": 1.8732, "step": 41223 }, { "epoch": 7.85, "learning_rate": 6.449571020019066e-07, "loss": 1.3802, "step": 41224 }, { "epoch": 7.85, "learning_rate": 6.448999046711154e-07, "loss": 1.9352, "step": 41225 }, { "epoch": 7.85, "learning_rate": 6.448427073403241e-07, "loss": 1.6506, "step": 41226 }, { "epoch": 7.85, "learning_rate": 6.447855100095328e-07, "loss": 0.4552, "step": 41227 }, { "epoch": 7.85, "learning_rate": 6.447283126787417e-07, "loss": 0.8569, "step": 41228 }, { "epoch": 7.85, "learning_rate": 6.446711153479504e-07, "loss": 0.7377, "step": 41229 }, { "epoch": 7.85, "learning_rate": 6.446139180171592e-07, "loss": 1.3399, "step": 41230 }, { "epoch": 7.85, "learning_rate": 6.44556720686368e-07, "loss": 1.6071, "step": 41231 }, { "epoch": 7.85, "learning_rate": 6.444995233555768e-07, "loss": 2.3171, "step": 41232 }, { "epoch": 7.85, "learning_rate": 6.444423260247855e-07, "loss": 0.7522, "step": 41233 }, { "epoch": 7.85, "learning_rate": 6.443851286939943e-07, "loss": 1.2036, "step": 41234 }, { "epoch": 7.85, "learning_rate": 6.443279313632031e-07, "loss": 1.0547, "step": 41235 }, { "epoch": 7.85, "learning_rate": 6.442707340324119e-07, "loss": 1.2074, "step": 41236 }, { "epoch": 7.85, "learning_rate": 6.442135367016207e-07, "loss": 3.6382, "step": 41237 }, { "epoch": 7.85, "learning_rate": 6.441563393708294e-07, "loss": 1.5215, "step": 41238 }, { "epoch": 7.86, "learning_rate": 6.440991420400382e-07, "loss": 1.4918, "step": 41239 }, { "epoch": 7.86, "learning_rate": 6.440419447092469e-07, "loss": 1.6608, "step": 41240 }, { "epoch": 7.86, "learning_rate": 6.439847473784556e-07, "loss": 1.5634, "step": 41241 }, { "epoch": 7.86, "learning_rate": 6.439275500476645e-07, "loss": 0.7019, "step": 41242 }, { "epoch": 7.86, "learning_rate": 6.438703527168732e-07, "loss": 1.0376, "step": 41243 }, { "epoch": 7.86, "learning_rate": 6.438131553860819e-07, "loss": 1.2139, "step": 41244 }, { "epoch": 7.86, "learning_rate": 6.437559580552908e-07, "loss": 2.9932, "step": 41245 }, { "epoch": 7.86, "learning_rate": 6.436987607244995e-07, "loss": 3.9803, "step": 41246 }, { "epoch": 7.86, "learning_rate": 6.436415633937083e-07, "loss": 0.4215, "step": 41247 }, { "epoch": 7.86, "learning_rate": 6.435843660629171e-07, "loss": 1.0119, "step": 41248 }, { "epoch": 7.86, "learning_rate": 6.435271687321259e-07, "loss": 1.8128, "step": 41249 }, { "epoch": 7.86, "learning_rate": 6.434699714013347e-07, "loss": 1.0753, "step": 41250 }, { "epoch": 7.86, "learning_rate": 6.434127740705434e-07, "loss": 1.8388, "step": 41251 }, { "epoch": 7.86, "learning_rate": 6.433555767397522e-07, "loss": 1.2696, "step": 41252 }, { "epoch": 7.86, "learning_rate": 6.432983794089609e-07, "loss": 1.0955, "step": 41253 }, { "epoch": 7.86, "learning_rate": 6.432411820781697e-07, "loss": 0.9676, "step": 41254 }, { "epoch": 7.86, "learning_rate": 6.431839847473784e-07, "loss": 0.5431, "step": 41255 }, { "epoch": 7.86, "learning_rate": 6.431267874165872e-07, "loss": 1.2025, "step": 41256 }, { "epoch": 7.86, "learning_rate": 6.43069590085796e-07, "loss": 1.9407, "step": 41257 }, { "epoch": 7.86, "learning_rate": 6.430123927550048e-07, "loss": 0.8047, "step": 41258 }, { "epoch": 7.86, "learning_rate": 6.429551954242136e-07, "loss": 0.9313, "step": 41259 }, { "epoch": 7.86, "learning_rate": 6.428979980934223e-07, "loss": 1.1223, "step": 41260 }, { "epoch": 7.86, "learning_rate": 6.428408007626311e-07, "loss": 1.8605, "step": 41261 }, { "epoch": 7.86, "learning_rate": 6.427836034318399e-07, "loss": 1.6735, "step": 41262 }, { "epoch": 7.86, "learning_rate": 6.427264061010486e-07, "loss": 0.9338, "step": 41263 }, { "epoch": 7.86, "learning_rate": 6.426692087702575e-07, "loss": 1.87, "step": 41264 }, { "epoch": 7.86, "learning_rate": 6.426120114394662e-07, "loss": 1.6962, "step": 41265 }, { "epoch": 7.86, "learning_rate": 6.42554814108675e-07, "loss": 0.3826, "step": 41266 }, { "epoch": 7.86, "learning_rate": 6.424976167778837e-07, "loss": 2.1102, "step": 41267 }, { "epoch": 7.86, "learning_rate": 6.424404194470924e-07, "loss": 0.8858, "step": 41268 }, { "epoch": 7.86, "learning_rate": 6.423832221163012e-07, "loss": 1.2331, "step": 41269 }, { "epoch": 7.86, "learning_rate": 6.4232602478551e-07, "loss": 2.2341, "step": 41270 }, { "epoch": 7.86, "learning_rate": 6.422688274547188e-07, "loss": 1.2419, "step": 41271 }, { "epoch": 7.86, "learning_rate": 6.422116301239276e-07, "loss": 1.3472, "step": 41272 }, { "epoch": 7.86, "learning_rate": 6.421544327931364e-07, "loss": 1.8189, "step": 41273 }, { "epoch": 7.86, "learning_rate": 6.420972354623451e-07, "loss": 1.8986, "step": 41274 }, { "epoch": 7.86, "learning_rate": 6.420400381315539e-07, "loss": 3.2468, "step": 41275 }, { "epoch": 7.86, "learning_rate": 6.419828408007627e-07, "loss": 1.1408, "step": 41276 }, { "epoch": 7.86, "learning_rate": 6.419256434699714e-07, "loss": 1.3365, "step": 41277 }, { "epoch": 7.86, "learning_rate": 6.418684461391803e-07, "loss": 0.2564, "step": 41278 }, { "epoch": 7.86, "learning_rate": 6.41811248808389e-07, "loss": 1.403, "step": 41279 }, { "epoch": 7.86, "learning_rate": 6.417540514775977e-07, "loss": 1.4047, "step": 41280 }, { "epoch": 7.86, "learning_rate": 6.416968541468065e-07, "loss": 0.6739, "step": 41281 }, { "epoch": 7.86, "learning_rate": 6.416396568160152e-07, "loss": 1.2373, "step": 41282 }, { "epoch": 7.86, "learning_rate": 6.41582459485224e-07, "loss": 1.0544, "step": 41283 }, { "epoch": 7.86, "learning_rate": 6.415252621544328e-07, "loss": 1.3431, "step": 41284 }, { "epoch": 7.86, "learning_rate": 6.414680648236416e-07, "loss": 1.9961, "step": 41285 }, { "epoch": 7.86, "learning_rate": 6.414108674928504e-07, "loss": 1.0469, "step": 41286 }, { "epoch": 7.86, "learning_rate": 6.413536701620591e-07, "loss": 2.8204, "step": 41287 }, { "epoch": 7.86, "learning_rate": 6.412964728312679e-07, "loss": 2.8746, "step": 41288 }, { "epoch": 7.86, "learning_rate": 6.412392755004767e-07, "loss": 2.9006, "step": 41289 }, { "epoch": 7.86, "learning_rate": 6.411820781696855e-07, "loss": 0.9849, "step": 41290 }, { "epoch": 7.86, "learning_rate": 6.411248808388942e-07, "loss": 2.4918, "step": 41291 }, { "epoch": 7.87, "learning_rate": 6.41067683508103e-07, "loss": 1.3288, "step": 41292 }, { "epoch": 7.87, "learning_rate": 6.410104861773117e-07, "loss": 0.8182, "step": 41293 }, { "epoch": 7.87, "learning_rate": 6.409532888465205e-07, "loss": 0.6746, "step": 41294 }, { "epoch": 7.87, "learning_rate": 6.408960915157293e-07, "loss": 3.0112, "step": 41295 }, { "epoch": 7.87, "learning_rate": 6.40838894184938e-07, "loss": 1.1607, "step": 41296 }, { "epoch": 7.87, "learning_rate": 6.407816968541468e-07, "loss": 2.16, "step": 41297 }, { "epoch": 7.87, "learning_rate": 6.407244995233556e-07, "loss": 1.0574, "step": 41298 }, { "epoch": 7.87, "learning_rate": 6.406673021925643e-07, "loss": 2.6673, "step": 41299 }, { "epoch": 7.87, "learning_rate": 6.406101048617732e-07, "loss": 2.5253, "step": 41300 }, { "epoch": 7.87, "learning_rate": 6.405529075309819e-07, "loss": 2.5204, "step": 41301 }, { "epoch": 7.87, "learning_rate": 6.404957102001907e-07, "loss": 0.8653, "step": 41302 }, { "epoch": 7.87, "learning_rate": 6.404385128693995e-07, "loss": 1.0398, "step": 41303 }, { "epoch": 7.87, "learning_rate": 6.403813155386082e-07, "loss": 1.4587, "step": 41304 }, { "epoch": 7.87, "learning_rate": 6.40324118207817e-07, "loss": 2.415, "step": 41305 }, { "epoch": 7.87, "learning_rate": 6.402669208770257e-07, "loss": 1.6877, "step": 41306 }, { "epoch": 7.87, "learning_rate": 6.402097235462345e-07, "loss": 1.5743, "step": 41307 }, { "epoch": 7.87, "learning_rate": 6.401525262154433e-07, "loss": 1.0072, "step": 41308 }, { "epoch": 7.87, "learning_rate": 6.40095328884652e-07, "loss": 1.0709, "step": 41309 }, { "epoch": 7.87, "learning_rate": 6.400381315538608e-07, "loss": 1.6509, "step": 41310 }, { "epoch": 7.87, "learning_rate": 6.399809342230696e-07, "loss": 1.0841, "step": 41311 }, { "epoch": 7.87, "learning_rate": 6.399237368922784e-07, "loss": 1.1267, "step": 41312 }, { "epoch": 7.87, "learning_rate": 6.398665395614871e-07, "loss": 1.95, "step": 41313 }, { "epoch": 7.87, "learning_rate": 6.39809342230696e-07, "loss": 2.0577, "step": 41314 }, { "epoch": 7.87, "learning_rate": 6.397521448999047e-07, "loss": 1.2547, "step": 41315 }, { "epoch": 7.87, "learning_rate": 6.396949475691135e-07, "loss": 1.2143, "step": 41316 }, { "epoch": 7.87, "learning_rate": 6.396377502383223e-07, "loss": 2.3522, "step": 41317 }, { "epoch": 7.87, "learning_rate": 6.39580552907531e-07, "loss": 1.6178, "step": 41318 }, { "epoch": 7.87, "learning_rate": 6.395233555767398e-07, "loss": 1.1925, "step": 41319 }, { "epoch": 7.87, "learning_rate": 6.394661582459485e-07, "loss": 4.4244, "step": 41320 }, { "epoch": 7.87, "learning_rate": 6.394089609151572e-07, "loss": 2.1725, "step": 41321 }, { "epoch": 7.87, "learning_rate": 6.393517635843661e-07, "loss": 1.9011, "step": 41322 }, { "epoch": 7.87, "learning_rate": 6.392945662535748e-07, "loss": 1.7311, "step": 41323 }, { "epoch": 7.87, "learning_rate": 6.392373689227836e-07, "loss": 1.2573, "step": 41324 }, { "epoch": 7.87, "learning_rate": 6.391801715919924e-07, "loss": 2.4894, "step": 41325 }, { "epoch": 7.87, "learning_rate": 6.391229742612012e-07, "loss": 2.8096, "step": 41326 }, { "epoch": 7.87, "learning_rate": 6.390657769304099e-07, "loss": 1.5345, "step": 41327 }, { "epoch": 7.87, "learning_rate": 6.390085795996187e-07, "loss": 1.7485, "step": 41328 }, { "epoch": 7.87, "learning_rate": 6.389513822688275e-07, "loss": 0.9284, "step": 41329 }, { "epoch": 7.87, "learning_rate": 6.388941849380363e-07, "loss": 1.4779, "step": 41330 }, { "epoch": 7.87, "learning_rate": 6.388369876072451e-07, "loss": 0.9084, "step": 41331 }, { "epoch": 7.87, "learning_rate": 6.387797902764538e-07, "loss": 0.2697, "step": 41332 }, { "epoch": 7.87, "learning_rate": 6.387225929456625e-07, "loss": 2.4775, "step": 41333 }, { "epoch": 7.87, "learning_rate": 6.386653956148713e-07, "loss": 0.6624, "step": 41334 }, { "epoch": 7.87, "learning_rate": 6.3860819828408e-07, "loss": 1.4723, "step": 41335 }, { "epoch": 7.87, "learning_rate": 6.385510009532889e-07, "loss": 1.3117, "step": 41336 }, { "epoch": 7.87, "learning_rate": 6.384938036224976e-07, "loss": 1.1193, "step": 41337 }, { "epoch": 7.87, "learning_rate": 6.384366062917064e-07, "loss": 0.9425, "step": 41338 }, { "epoch": 7.87, "learning_rate": 6.383794089609152e-07, "loss": 1.8769, "step": 41339 }, { "epoch": 7.87, "learning_rate": 6.383222116301239e-07, "loss": 1.1556, "step": 41340 }, { "epoch": 7.87, "learning_rate": 6.382650142993327e-07, "loss": 0.4584, "step": 41341 }, { "epoch": 7.87, "learning_rate": 6.382078169685415e-07, "loss": 0.6107, "step": 41342 }, { "epoch": 7.87, "learning_rate": 6.381506196377503e-07, "loss": 2.4985, "step": 41343 }, { "epoch": 7.88, "learning_rate": 6.380934223069591e-07, "loss": 2.5056, "step": 41344 }, { "epoch": 7.88, "learning_rate": 6.380362249761678e-07, "loss": 2.2025, "step": 41345 }, { "epoch": 7.88, "learning_rate": 6.379790276453765e-07, "loss": 1.8804, "step": 41346 }, { "epoch": 7.88, "learning_rate": 6.379218303145853e-07, "loss": 1.3531, "step": 41347 }, { "epoch": 7.88, "learning_rate": 6.378646329837941e-07, "loss": 2.456, "step": 41348 }, { "epoch": 7.88, "learning_rate": 6.378074356530028e-07, "loss": 1.7323, "step": 41349 }, { "epoch": 7.88, "learning_rate": 6.377502383222117e-07, "loss": 2.3338, "step": 41350 }, { "epoch": 7.88, "learning_rate": 6.376930409914204e-07, "loss": 1.969, "step": 41351 }, { "epoch": 7.88, "learning_rate": 6.376358436606292e-07, "loss": 1.1245, "step": 41352 }, { "epoch": 7.88, "learning_rate": 6.37578646329838e-07, "loss": 1.8446, "step": 41353 }, { "epoch": 7.88, "learning_rate": 6.375214489990467e-07, "loss": 1.4807, "step": 41354 }, { "epoch": 7.88, "learning_rate": 6.374642516682555e-07, "loss": 1.7419, "step": 41355 }, { "epoch": 7.88, "learning_rate": 6.374070543374643e-07, "loss": 2.2315, "step": 41356 }, { "epoch": 7.88, "learning_rate": 6.37349857006673e-07, "loss": 1.2039, "step": 41357 }, { "epoch": 7.88, "learning_rate": 6.372926596758819e-07, "loss": 1.7372, "step": 41358 }, { "epoch": 7.88, "learning_rate": 6.372354623450905e-07, "loss": 1.8936, "step": 41359 }, { "epoch": 7.88, "learning_rate": 6.371782650142993e-07, "loss": 1.6365, "step": 41360 }, { "epoch": 7.88, "learning_rate": 6.371210676835081e-07, "loss": 1.7988, "step": 41361 }, { "epoch": 7.88, "learning_rate": 6.370638703527168e-07, "loss": 0.3114, "step": 41362 }, { "epoch": 7.88, "learning_rate": 6.370066730219256e-07, "loss": 2.0883, "step": 41363 }, { "epoch": 7.88, "learning_rate": 6.369494756911344e-07, "loss": 2.9023, "step": 41364 }, { "epoch": 7.88, "learning_rate": 6.368922783603432e-07, "loss": 1.6336, "step": 41365 }, { "epoch": 7.88, "learning_rate": 6.36835081029552e-07, "loss": 0.5452, "step": 41366 }, { "epoch": 7.88, "learning_rate": 6.367778836987608e-07, "loss": 1.5195, "step": 41367 }, { "epoch": 7.88, "learning_rate": 6.367206863679695e-07, "loss": 2.0772, "step": 41368 }, { "epoch": 7.88, "learning_rate": 6.366634890371783e-07, "loss": 1.4262, "step": 41369 }, { "epoch": 7.88, "learning_rate": 6.366062917063871e-07, "loss": 2.347, "step": 41370 }, { "epoch": 7.88, "learning_rate": 6.365490943755958e-07, "loss": 3.1237, "step": 41371 }, { "epoch": 7.88, "learning_rate": 6.364918970448047e-07, "loss": 2.009, "step": 41372 }, { "epoch": 7.88, "learning_rate": 6.364346997140133e-07, "loss": 2.0484, "step": 41373 }, { "epoch": 7.88, "learning_rate": 6.363775023832221e-07, "loss": 0.9014, "step": 41374 }, { "epoch": 7.88, "learning_rate": 6.363203050524309e-07, "loss": 1.5416, "step": 41375 }, { "epoch": 7.88, "learning_rate": 6.362631077216396e-07, "loss": 3.1793, "step": 41376 }, { "epoch": 7.88, "learning_rate": 6.362059103908484e-07, "loss": 0.9778, "step": 41377 }, { "epoch": 7.88, "learning_rate": 6.361487130600572e-07, "loss": 2.8931, "step": 41378 }, { "epoch": 7.88, "learning_rate": 6.36091515729266e-07, "loss": 1.5639, "step": 41379 }, { "epoch": 7.88, "learning_rate": 6.360343183984748e-07, "loss": 1.6783, "step": 41380 }, { "epoch": 7.88, "learning_rate": 6.359771210676835e-07, "loss": 1.5518, "step": 41381 }, { "epoch": 7.88, "learning_rate": 6.359199237368923e-07, "loss": 2.3856, "step": 41382 }, { "epoch": 7.88, "learning_rate": 6.358627264061011e-07, "loss": 1.0932, "step": 41383 }, { "epoch": 7.88, "learning_rate": 6.358055290753099e-07, "loss": 2.3558, "step": 41384 }, { "epoch": 7.88, "learning_rate": 6.357483317445186e-07, "loss": 2.145, "step": 41385 }, { "epoch": 7.88, "learning_rate": 6.356911344137273e-07, "loss": 1.6409, "step": 41386 }, { "epoch": 7.88, "learning_rate": 6.356339370829361e-07, "loss": 1.4737, "step": 41387 }, { "epoch": 7.88, "learning_rate": 6.355767397521449e-07, "loss": 1.1074, "step": 41388 }, { "epoch": 7.88, "learning_rate": 6.355195424213537e-07, "loss": 1.5222, "step": 41389 }, { "epoch": 7.88, "learning_rate": 6.354623450905624e-07, "loss": 1.1219, "step": 41390 }, { "epoch": 7.88, "learning_rate": 6.354051477597713e-07, "loss": 0.3595, "step": 41391 }, { "epoch": 7.88, "learning_rate": 6.3534795042898e-07, "loss": 1.4574, "step": 41392 }, { "epoch": 7.88, "learning_rate": 6.352907530981887e-07, "loss": 0.8748, "step": 41393 }, { "epoch": 7.88, "learning_rate": 6.352335557673976e-07, "loss": 1.7707, "step": 41394 }, { "epoch": 7.88, "learning_rate": 6.351763584366063e-07, "loss": 0.7303, "step": 41395 }, { "epoch": 7.88, "learning_rate": 6.351191611058151e-07, "loss": 0.8012, "step": 41396 }, { "epoch": 7.89, "learning_rate": 6.350619637750239e-07, "loss": 2.3424, "step": 41397 }, { "epoch": 7.89, "learning_rate": 6.350047664442326e-07, "loss": 1.9987, "step": 41398 }, { "epoch": 7.89, "learning_rate": 6.349475691134413e-07, "loss": 2.3691, "step": 41399 }, { "epoch": 7.89, "learning_rate": 6.348903717826501e-07, "loss": 2.1361, "step": 41400 }, { "epoch": 7.89, "learning_rate": 6.348331744518589e-07, "loss": 1.4291, "step": 41401 }, { "epoch": 7.89, "learning_rate": 6.347759771210677e-07, "loss": 2.0701, "step": 41402 }, { "epoch": 7.89, "learning_rate": 6.347187797902765e-07, "loss": 1.679, "step": 41403 }, { "epoch": 7.89, "learning_rate": 6.346615824594852e-07, "loss": 0.7301, "step": 41404 }, { "epoch": 7.89, "learning_rate": 6.34604385128694e-07, "loss": 1.4921, "step": 41405 }, { "epoch": 7.89, "learning_rate": 6.345471877979028e-07, "loss": 0.4755, "step": 41406 }, { "epoch": 7.89, "learning_rate": 6.344899904671115e-07, "loss": 1.3722, "step": 41407 }, { "epoch": 7.89, "learning_rate": 6.344327931363204e-07, "loss": 1.8808, "step": 41408 }, { "epoch": 7.89, "learning_rate": 6.343755958055291e-07, "loss": 2.5715, "step": 41409 }, { "epoch": 7.89, "learning_rate": 6.343183984747379e-07, "loss": 1.0092, "step": 41410 }, { "epoch": 7.89, "learning_rate": 6.342612011439467e-07, "loss": 3.1514, "step": 41411 }, { "epoch": 7.89, "learning_rate": 6.342040038131554e-07, "loss": 1.0592, "step": 41412 }, { "epoch": 7.89, "learning_rate": 6.341468064823643e-07, "loss": 1.498, "step": 41413 }, { "epoch": 7.89, "learning_rate": 6.340896091515729e-07, "loss": 0.1939, "step": 41414 }, { "epoch": 7.89, "learning_rate": 6.340324118207817e-07, "loss": 2.1148, "step": 41415 }, { "epoch": 7.89, "learning_rate": 6.339752144899905e-07, "loss": 0.8958, "step": 41416 }, { "epoch": 7.89, "learning_rate": 6.339180171591992e-07, "loss": 1.5347, "step": 41417 }, { "epoch": 7.89, "learning_rate": 6.33860819828408e-07, "loss": 1.0412, "step": 41418 }, { "epoch": 7.89, "learning_rate": 6.338036224976168e-07, "loss": 1.9634, "step": 41419 }, { "epoch": 7.89, "learning_rate": 6.337464251668256e-07, "loss": 1.1458, "step": 41420 }, { "epoch": 7.89, "learning_rate": 6.336892278360343e-07, "loss": 2.0735, "step": 41421 }, { "epoch": 7.89, "learning_rate": 6.336320305052431e-07, "loss": 1.5055, "step": 41422 }, { "epoch": 7.89, "learning_rate": 6.335748331744519e-07, "loss": 1.2161, "step": 41423 }, { "epoch": 7.89, "learning_rate": 6.335176358436607e-07, "loss": 1.8556, "step": 41424 }, { "epoch": 7.89, "learning_rate": 6.334604385128695e-07, "loss": 1.5523, "step": 41425 }, { "epoch": 7.89, "learning_rate": 6.334032411820781e-07, "loss": 1.5823, "step": 41426 }, { "epoch": 7.89, "learning_rate": 6.333460438512869e-07, "loss": 2.1811, "step": 41427 }, { "epoch": 7.89, "learning_rate": 6.332888465204957e-07, "loss": 2.6061, "step": 41428 }, { "epoch": 7.89, "learning_rate": 6.332316491897044e-07, "loss": 2.1314, "step": 41429 }, { "epoch": 7.89, "learning_rate": 6.331744518589133e-07, "loss": 1.9342, "step": 41430 }, { "epoch": 7.89, "learning_rate": 6.33117254528122e-07, "loss": 1.0008, "step": 41431 }, { "epoch": 7.89, "learning_rate": 6.330600571973308e-07, "loss": 2.6954, "step": 41432 }, { "epoch": 7.89, "learning_rate": 6.330028598665396e-07, "loss": 1.9527, "step": 41433 }, { "epoch": 7.89, "learning_rate": 6.329456625357483e-07, "loss": 1.3525, "step": 41434 }, { "epoch": 7.89, "learning_rate": 6.328884652049572e-07, "loss": 0.8911, "step": 41435 }, { "epoch": 7.89, "learning_rate": 6.328312678741659e-07, "loss": 0.7033, "step": 41436 }, { "epoch": 7.89, "learning_rate": 6.327740705433747e-07, "loss": 2.4177, "step": 41437 }, { "epoch": 7.89, "learning_rate": 6.327168732125835e-07, "loss": 1.4071, "step": 41438 }, { "epoch": 7.89, "learning_rate": 6.326596758817921e-07, "loss": 1.2514, "step": 41439 }, { "epoch": 7.89, "learning_rate": 6.326024785510009e-07, "loss": 3.1825, "step": 41440 }, { "epoch": 7.89, "learning_rate": 6.325452812202097e-07, "loss": 0.9542, "step": 41441 }, { "epoch": 7.89, "learning_rate": 6.324880838894185e-07, "loss": 1.8382, "step": 41442 }, { "epoch": 7.89, "learning_rate": 6.324308865586272e-07, "loss": 0.967, "step": 41443 }, { "epoch": 7.89, "learning_rate": 6.323736892278361e-07, "loss": 2.0737, "step": 41444 }, { "epoch": 7.89, "learning_rate": 6.323164918970448e-07, "loss": 1.1172, "step": 41445 }, { "epoch": 7.89, "learning_rate": 6.322592945662536e-07, "loss": 0.594, "step": 41446 }, { "epoch": 7.89, "learning_rate": 6.322020972354624e-07, "loss": 1.5484, "step": 41447 }, { "epoch": 7.89, "learning_rate": 6.321448999046711e-07, "loss": 2.4842, "step": 41448 }, { "epoch": 7.9, "learning_rate": 6.3208770257388e-07, "loss": 2.2821, "step": 41449 }, { "epoch": 7.9, "learning_rate": 6.320305052430887e-07, "loss": 1.7754, "step": 41450 }, { "epoch": 7.9, "learning_rate": 6.319733079122974e-07, "loss": 1.3934, "step": 41451 }, { "epoch": 7.9, "learning_rate": 6.319161105815063e-07, "loss": 0.8285, "step": 41452 }, { "epoch": 7.9, "learning_rate": 6.318589132507149e-07, "loss": 2.7415, "step": 41453 }, { "epoch": 7.9, "learning_rate": 6.318017159199237e-07, "loss": 1.7223, "step": 41454 }, { "epoch": 7.9, "learning_rate": 6.317445185891325e-07, "loss": 0.9587, "step": 41455 }, { "epoch": 7.9, "learning_rate": 6.316873212583413e-07, "loss": 0.7708, "step": 41456 }, { "epoch": 7.9, "learning_rate": 6.316301239275501e-07, "loss": 0.6313, "step": 41457 }, { "epoch": 7.9, "learning_rate": 6.315729265967588e-07, "loss": 1.3953, "step": 41458 }, { "epoch": 7.9, "learning_rate": 6.315157292659676e-07, "loss": 1.2308, "step": 41459 }, { "epoch": 7.9, "learning_rate": 6.314585319351764e-07, "loss": 2.0297, "step": 41460 }, { "epoch": 7.9, "learning_rate": 6.314013346043852e-07, "loss": 1.9359, "step": 41461 }, { "epoch": 7.9, "learning_rate": 6.313441372735939e-07, "loss": 2.6165, "step": 41462 }, { "epoch": 7.9, "learning_rate": 6.312869399428027e-07, "loss": 2.2262, "step": 41463 }, { "epoch": 7.9, "learning_rate": 6.312297426120115e-07, "loss": 2.1658, "step": 41464 }, { "epoch": 7.9, "learning_rate": 6.311725452812202e-07, "loss": 1.0692, "step": 41465 }, { "epoch": 7.9, "learning_rate": 6.311153479504291e-07, "loss": 0.5798, "step": 41466 }, { "epoch": 7.9, "learning_rate": 6.310581506196377e-07, "loss": 0.7074, "step": 41467 }, { "epoch": 7.9, "learning_rate": 6.310009532888465e-07, "loss": 1.2107, "step": 41468 }, { "epoch": 7.9, "learning_rate": 6.309437559580553e-07, "loss": 2.702, "step": 41469 }, { "epoch": 7.9, "learning_rate": 6.30886558627264e-07, "loss": 1.2873, "step": 41470 }, { "epoch": 7.9, "learning_rate": 6.308293612964729e-07, "loss": 1.4623, "step": 41471 }, { "epoch": 7.9, "learning_rate": 6.307721639656816e-07, "loss": 3.4769, "step": 41472 }, { "epoch": 7.9, "learning_rate": 6.307149666348904e-07, "loss": 1.8511, "step": 41473 }, { "epoch": 7.9, "learning_rate": 6.306577693040992e-07, "loss": 0.8091, "step": 41474 }, { "epoch": 7.9, "learning_rate": 6.306005719733079e-07, "loss": 1.7436, "step": 41475 }, { "epoch": 7.9, "learning_rate": 6.305433746425167e-07, "loss": 1.9976, "step": 41476 }, { "epoch": 7.9, "learning_rate": 6.304861773117255e-07, "loss": 1.3264, "step": 41477 }, { "epoch": 7.9, "learning_rate": 6.304289799809343e-07, "loss": 1.3199, "step": 41478 }, { "epoch": 7.9, "learning_rate": 6.303717826501431e-07, "loss": 2.0688, "step": 41479 }, { "epoch": 7.9, "learning_rate": 6.303145853193517e-07, "loss": 2.2528, "step": 41480 }, { "epoch": 7.9, "learning_rate": 6.302573879885605e-07, "loss": 1.8194, "step": 41481 }, { "epoch": 7.9, "learning_rate": 6.302001906577693e-07, "loss": 0.6147, "step": 41482 }, { "epoch": 7.9, "learning_rate": 6.301429933269781e-07, "loss": 0.678, "step": 41483 }, { "epoch": 7.9, "learning_rate": 6.300857959961868e-07, "loss": 1.608, "step": 41484 }, { "epoch": 7.9, "learning_rate": 6.300285986653957e-07, "loss": 1.9872, "step": 41485 }, { "epoch": 7.9, "learning_rate": 6.299714013346044e-07, "loss": 1.6303, "step": 41486 }, { "epoch": 7.9, "learning_rate": 6.299142040038131e-07, "loss": 1.856, "step": 41487 }, { "epoch": 7.9, "learning_rate": 6.29857006673022e-07, "loss": 2.5357, "step": 41488 }, { "epoch": 7.9, "learning_rate": 6.297998093422307e-07, "loss": 1.7514, "step": 41489 }, { "epoch": 7.9, "learning_rate": 6.297426120114395e-07, "loss": 1.8104, "step": 41490 }, { "epoch": 7.9, "learning_rate": 6.296854146806483e-07, "loss": 2.6043, "step": 41491 }, { "epoch": 7.9, "learning_rate": 6.29628217349857e-07, "loss": 1.7103, "step": 41492 }, { "epoch": 7.9, "learning_rate": 6.295710200190659e-07, "loss": 1.5097, "step": 41493 }, { "epoch": 7.9, "learning_rate": 6.295138226882745e-07, "loss": 0.6203, "step": 41494 }, { "epoch": 7.9, "learning_rate": 6.294566253574833e-07, "loss": 1.7658, "step": 41495 }, { "epoch": 7.9, "learning_rate": 6.293994280266921e-07, "loss": 1.695, "step": 41496 }, { "epoch": 7.9, "learning_rate": 6.293422306959009e-07, "loss": 2.5763, "step": 41497 }, { "epoch": 7.9, "learning_rate": 6.292850333651096e-07, "loss": 0.9018, "step": 41498 }, { "epoch": 7.9, "learning_rate": 6.292278360343184e-07, "loss": 1.6915, "step": 41499 }, { "epoch": 7.9, "learning_rate": 6.291706387035272e-07, "loss": 1.0005, "step": 41500 }, { "epoch": 7.9, "learning_rate": 6.29113441372736e-07, "loss": 1.0272, "step": 41501 }, { "epoch": 7.91, "learning_rate": 6.290562440419448e-07, "loss": 2.7865, "step": 41502 }, { "epoch": 7.91, "learning_rate": 6.289990467111535e-07, "loss": 2.0491, "step": 41503 }, { "epoch": 7.91, "learning_rate": 6.289418493803623e-07, "loss": 1.6023, "step": 41504 }, { "epoch": 7.91, "learning_rate": 6.288846520495711e-07, "loss": 3.0613, "step": 41505 }, { "epoch": 7.91, "learning_rate": 6.288274547187797e-07, "loss": 0.692, "step": 41506 }, { "epoch": 7.91, "learning_rate": 6.287702573879886e-07, "loss": 2.1606, "step": 41507 }, { "epoch": 7.91, "learning_rate": 6.287130600571973e-07, "loss": 1.0606, "step": 41508 }, { "epoch": 7.91, "learning_rate": 6.286558627264061e-07, "loss": 0.8, "step": 41509 }, { "epoch": 7.91, "learning_rate": 6.285986653956149e-07, "loss": 2.4281, "step": 41510 }, { "epoch": 7.91, "learning_rate": 6.285414680648236e-07, "loss": 2.1407, "step": 41511 }, { "epoch": 7.91, "learning_rate": 6.284842707340324e-07, "loss": 1.5255, "step": 41512 }, { "epoch": 7.91, "learning_rate": 6.284270734032412e-07, "loss": 1.9178, "step": 41513 }, { "epoch": 7.91, "learning_rate": 6.2836987607245e-07, "loss": 1.0171, "step": 41514 }, { "epoch": 7.91, "learning_rate": 6.283126787416588e-07, "loss": 2.4362, "step": 41515 }, { "epoch": 7.91, "learning_rate": 6.282554814108675e-07, "loss": 0.7624, "step": 41516 }, { "epoch": 7.91, "learning_rate": 6.281982840800763e-07, "loss": 2.9015, "step": 41517 }, { "epoch": 7.91, "learning_rate": 6.281410867492851e-07, "loss": 1.1837, "step": 41518 }, { "epoch": 7.91, "learning_rate": 6.280838894184939e-07, "loss": 2.0111, "step": 41519 }, { "epoch": 7.91, "learning_rate": 6.280266920877025e-07, "loss": 2.4099, "step": 41520 }, { "epoch": 7.91, "learning_rate": 6.279694947569114e-07, "loss": 1.5908, "step": 41521 }, { "epoch": 7.91, "learning_rate": 6.279122974261201e-07, "loss": 0.9542, "step": 41522 }, { "epoch": 7.91, "learning_rate": 6.278551000953288e-07, "loss": 2.2262, "step": 41523 }, { "epoch": 7.91, "learning_rate": 6.277979027645377e-07, "loss": 2.0069, "step": 41524 }, { "epoch": 7.91, "learning_rate": 6.277407054337464e-07, "loss": 1.9462, "step": 41525 }, { "epoch": 7.91, "learning_rate": 6.276835081029552e-07, "loss": 1.5609, "step": 41526 }, { "epoch": 7.91, "learning_rate": 6.27626310772164e-07, "loss": 0.9764, "step": 41527 }, { "epoch": 7.91, "learning_rate": 6.275691134413727e-07, "loss": 0.8159, "step": 41528 }, { "epoch": 7.91, "learning_rate": 6.275119161105816e-07, "loss": 1.3169, "step": 41529 }, { "epoch": 7.91, "learning_rate": 6.274547187797903e-07, "loss": 1.0332, "step": 41530 }, { "epoch": 7.91, "learning_rate": 6.273975214489991e-07, "loss": 2.1903, "step": 41531 }, { "epoch": 7.91, "learning_rate": 6.273403241182079e-07, "loss": 1.5404, "step": 41532 }, { "epoch": 7.91, "learning_rate": 6.272831267874167e-07, "loss": 2.0439, "step": 41533 }, { "epoch": 7.91, "learning_rate": 6.272259294566253e-07, "loss": 2.3376, "step": 41534 }, { "epoch": 7.91, "learning_rate": 6.271687321258341e-07, "loss": 1.3189, "step": 41535 }, { "epoch": 7.91, "learning_rate": 6.271115347950429e-07, "loss": 1.3909, "step": 41536 }, { "epoch": 7.91, "learning_rate": 6.270543374642517e-07, "loss": 1.6527, "step": 41537 }, { "epoch": 7.91, "learning_rate": 6.269971401334605e-07, "loss": 1.8684, "step": 41538 }, { "epoch": 7.91, "learning_rate": 6.269399428026692e-07, "loss": 1.2884, "step": 41539 }, { "epoch": 7.91, "learning_rate": 6.26882745471878e-07, "loss": 2.0155, "step": 41540 }, { "epoch": 7.91, "learning_rate": 6.268255481410868e-07, "loss": 2.9009, "step": 41541 }, { "epoch": 7.91, "learning_rate": 6.267683508102955e-07, "loss": 2.4339, "step": 41542 }, { "epoch": 7.91, "learning_rate": 6.267111534795044e-07, "loss": 0.625, "step": 41543 }, { "epoch": 7.91, "learning_rate": 6.266539561487131e-07, "loss": 1.5675, "step": 41544 }, { "epoch": 7.91, "learning_rate": 6.265967588179218e-07, "loss": 1.825, "step": 41545 }, { "epoch": 7.91, "learning_rate": 6.265395614871307e-07, "loss": 1.4799, "step": 41546 }, { "epoch": 7.91, "learning_rate": 6.264823641563393e-07, "loss": 0.8508, "step": 41547 }, { "epoch": 7.91, "learning_rate": 6.264251668255481e-07, "loss": 2.6423, "step": 41548 }, { "epoch": 7.91, "learning_rate": 6.263679694947569e-07, "loss": 0.8899, "step": 41549 }, { "epoch": 7.91, "learning_rate": 6.263107721639657e-07, "loss": 0.7295, "step": 41550 }, { "epoch": 7.91, "learning_rate": 6.262535748331745e-07, "loss": 1.9572, "step": 41551 }, { "epoch": 7.91, "learning_rate": 6.261963775023832e-07, "loss": 1.1842, "step": 41552 }, { "epoch": 7.91, "learning_rate": 6.26139180171592e-07, "loss": 1.0856, "step": 41553 }, { "epoch": 7.92, "learning_rate": 6.260819828408008e-07, "loss": 1.3767, "step": 41554 }, { "epoch": 7.92, "learning_rate": 6.260247855100096e-07, "loss": 1.9489, "step": 41555 }, { "epoch": 7.92, "learning_rate": 6.259675881792183e-07, "loss": 2.436, "step": 41556 }, { "epoch": 7.92, "learning_rate": 6.259103908484271e-07, "loss": 2.4527, "step": 41557 }, { "epoch": 7.92, "learning_rate": 6.258531935176359e-07, "loss": 0.926, "step": 41558 }, { "epoch": 7.92, "learning_rate": 6.257959961868447e-07, "loss": 2.0838, "step": 41559 }, { "epoch": 7.92, "learning_rate": 6.257387988560534e-07, "loss": 1.3462, "step": 41560 }, { "epoch": 7.92, "learning_rate": 6.256816015252621e-07, "loss": 0.826, "step": 41561 }, { "epoch": 7.92, "learning_rate": 6.256244041944709e-07, "loss": 0.7318, "step": 41562 }, { "epoch": 7.92, "learning_rate": 6.255672068636797e-07, "loss": 2.5713, "step": 41563 }, { "epoch": 7.92, "learning_rate": 6.255100095328884e-07, "loss": 2.5153, "step": 41564 }, { "epoch": 7.92, "learning_rate": 6.254528122020973e-07, "loss": 1.2028, "step": 41565 }, { "epoch": 7.92, "learning_rate": 6.25395614871306e-07, "loss": 1.4426, "step": 41566 }, { "epoch": 7.92, "learning_rate": 6.253384175405148e-07, "loss": 1.3749, "step": 41567 }, { "epoch": 7.92, "learning_rate": 6.252812202097236e-07, "loss": 1.6473, "step": 41568 }, { "epoch": 7.92, "learning_rate": 6.252240228789323e-07, "loss": 2.1363, "step": 41569 }, { "epoch": 7.92, "learning_rate": 6.251668255481411e-07, "loss": 1.8214, "step": 41570 }, { "epoch": 7.92, "learning_rate": 6.251096282173499e-07, "loss": 1.5432, "step": 41571 }, { "epoch": 7.92, "learning_rate": 6.250524308865587e-07, "loss": 1.416, "step": 41572 }, { "epoch": 7.92, "learning_rate": 6.249952335557675e-07, "loss": 1.0846, "step": 41573 }, { "epoch": 7.92, "learning_rate": 6.249380362249762e-07, "loss": 2.7176, "step": 41574 }, { "epoch": 7.92, "learning_rate": 6.248808388941849e-07, "loss": 2.0843, "step": 41575 }, { "epoch": 7.92, "learning_rate": 6.248236415633937e-07, "loss": 1.722, "step": 41576 }, { "epoch": 7.92, "learning_rate": 6.247664442326025e-07, "loss": 0.9764, "step": 41577 }, { "epoch": 7.92, "learning_rate": 6.247092469018112e-07, "loss": 0.9094, "step": 41578 }, { "epoch": 7.92, "learning_rate": 6.246520495710201e-07, "loss": 2.3036, "step": 41579 }, { "epoch": 7.92, "learning_rate": 6.245948522402288e-07, "loss": 2.4324, "step": 41580 }, { "epoch": 7.92, "learning_rate": 6.245376549094376e-07, "loss": 1.8647, "step": 41581 }, { "epoch": 7.92, "learning_rate": 6.244804575786464e-07, "loss": 2.4594, "step": 41582 }, { "epoch": 7.92, "learning_rate": 6.244232602478551e-07, "loss": 2.8373, "step": 41583 }, { "epoch": 7.92, "learning_rate": 6.243660629170639e-07, "loss": 0.9864, "step": 41584 }, { "epoch": 7.92, "learning_rate": 6.243088655862727e-07, "loss": 1.4704, "step": 41585 }, { "epoch": 7.92, "learning_rate": 6.242516682554815e-07, "loss": 1.599, "step": 41586 }, { "epoch": 7.92, "learning_rate": 6.241944709246902e-07, "loss": 1.272, "step": 41587 }, { "epoch": 7.92, "learning_rate": 6.241372735938989e-07, "loss": 1.6794, "step": 41588 }, { "epoch": 7.92, "learning_rate": 6.240800762631077e-07, "loss": 1.0968, "step": 41589 }, { "epoch": 7.92, "learning_rate": 6.240228789323165e-07, "loss": 1.52, "step": 41590 }, { "epoch": 7.92, "learning_rate": 6.239656816015253e-07, "loss": 3.4548, "step": 41591 }, { "epoch": 7.92, "learning_rate": 6.23908484270734e-07, "loss": 1.32, "step": 41592 }, { "epoch": 7.92, "learning_rate": 6.238512869399428e-07, "loss": 0.9918, "step": 41593 }, { "epoch": 7.92, "learning_rate": 6.237940896091516e-07, "loss": 2.0387, "step": 41594 }, { "epoch": 7.92, "learning_rate": 6.237368922783604e-07, "loss": 1.0107, "step": 41595 }, { "epoch": 7.92, "learning_rate": 6.236796949475692e-07, "loss": 1.6563, "step": 41596 }, { "epoch": 7.92, "learning_rate": 6.236224976167779e-07, "loss": 1.1372, "step": 41597 }, { "epoch": 7.92, "learning_rate": 6.235653002859867e-07, "loss": 2.4547, "step": 41598 }, { "epoch": 7.92, "learning_rate": 6.235081029551955e-07, "loss": 2.3258, "step": 41599 }, { "epoch": 7.92, "learning_rate": 6.234509056244041e-07, "loss": 2.0509, "step": 41600 }, { "epoch": 7.92, "learning_rate": 6.23393708293613e-07, "loss": 1.5716, "step": 41601 }, { "epoch": 7.92, "learning_rate": 6.233365109628217e-07, "loss": 0.6735, "step": 41602 }, { "epoch": 7.92, "learning_rate": 6.232793136320305e-07, "loss": 1.6934, "step": 41603 }, { "epoch": 7.92, "learning_rate": 6.232221163012393e-07, "loss": 1.7092, "step": 41604 }, { "epoch": 7.92, "learning_rate": 6.23164918970448e-07, "loss": 2.1968, "step": 41605 }, { "epoch": 7.92, "learning_rate": 6.231077216396568e-07, "loss": 0.3561, "step": 41606 }, { "epoch": 7.93, "learning_rate": 6.230505243088656e-07, "loss": 0.8685, "step": 41607 }, { "epoch": 7.93, "learning_rate": 6.229933269780744e-07, "loss": 1.6301, "step": 41608 }, { "epoch": 7.93, "learning_rate": 6.229361296472832e-07, "loss": 1.679, "step": 41609 }, { "epoch": 7.93, "learning_rate": 6.228789323164919e-07, "loss": 1.1068, "step": 41610 }, { "epoch": 7.93, "learning_rate": 6.228217349857007e-07, "loss": 1.9895, "step": 41611 }, { "epoch": 7.93, "learning_rate": 6.227645376549095e-07, "loss": 2.0275, "step": 41612 }, { "epoch": 7.93, "learning_rate": 6.227073403241183e-07, "loss": 1.427, "step": 41613 }, { "epoch": 7.93, "learning_rate": 6.226501429933269e-07, "loss": 1.6171, "step": 41614 }, { "epoch": 7.93, "learning_rate": 6.225929456625358e-07, "loss": 0.6404, "step": 41615 }, { "epoch": 7.93, "learning_rate": 6.225357483317445e-07, "loss": 0.4273, "step": 41616 }, { "epoch": 7.93, "learning_rate": 6.224785510009533e-07, "loss": 1.6945, "step": 41617 }, { "epoch": 7.93, "learning_rate": 6.224213536701621e-07, "loss": 3.0275, "step": 41618 }, { "epoch": 7.93, "learning_rate": 6.223641563393708e-07, "loss": 1.0459, "step": 41619 }, { "epoch": 7.93, "learning_rate": 6.223069590085796e-07, "loss": 1.2444, "step": 41620 }, { "epoch": 7.93, "learning_rate": 6.222497616777884e-07, "loss": 1.2833, "step": 41621 }, { "epoch": 7.93, "learning_rate": 6.221925643469971e-07, "loss": 1.606, "step": 41622 }, { "epoch": 7.93, "learning_rate": 6.22135367016206e-07, "loss": 2.3831, "step": 41623 }, { "epoch": 7.93, "learning_rate": 6.220781696854147e-07, "loss": 1.4594, "step": 41624 }, { "epoch": 7.93, "learning_rate": 6.220209723546235e-07, "loss": 1.1194, "step": 41625 }, { "epoch": 7.93, "learning_rate": 6.219637750238323e-07, "loss": 0.2049, "step": 41626 }, { "epoch": 7.93, "learning_rate": 6.21906577693041e-07, "loss": 0.3542, "step": 41627 }, { "epoch": 7.93, "learning_rate": 6.218493803622497e-07, "loss": 1.5354, "step": 41628 }, { "epoch": 7.93, "learning_rate": 6.217921830314585e-07, "loss": 2.4615, "step": 41629 }, { "epoch": 7.93, "learning_rate": 6.217349857006673e-07, "loss": 1.1559, "step": 41630 }, { "epoch": 7.93, "learning_rate": 6.216777883698761e-07, "loss": 1.0106, "step": 41631 }, { "epoch": 7.93, "learning_rate": 6.216205910390849e-07, "loss": 2.0252, "step": 41632 }, { "epoch": 7.93, "learning_rate": 6.215633937082936e-07, "loss": 0.8762, "step": 41633 }, { "epoch": 7.93, "learning_rate": 6.215061963775024e-07, "loss": 1.4014, "step": 41634 }, { "epoch": 7.93, "learning_rate": 6.214489990467112e-07, "loss": 1.3505, "step": 41635 }, { "epoch": 7.93, "learning_rate": 6.213918017159199e-07, "loss": 0.733, "step": 41636 }, { "epoch": 7.93, "learning_rate": 6.213346043851288e-07, "loss": 1.3234, "step": 41637 }, { "epoch": 7.93, "learning_rate": 6.212774070543375e-07, "loss": 3.5067, "step": 41638 }, { "epoch": 7.93, "learning_rate": 6.212202097235463e-07, "loss": 1.3564, "step": 41639 }, { "epoch": 7.93, "learning_rate": 6.21163012392755e-07, "loss": 2.5184, "step": 41640 }, { "epoch": 7.93, "learning_rate": 6.211058150619637e-07, "loss": 2.9344, "step": 41641 }, { "epoch": 7.93, "learning_rate": 6.210486177311725e-07, "loss": 0.9738, "step": 41642 }, { "epoch": 7.93, "learning_rate": 6.209914204003813e-07, "loss": 2.6333, "step": 41643 }, { "epoch": 7.93, "learning_rate": 6.209342230695901e-07, "loss": 1.376, "step": 41644 }, { "epoch": 7.93, "learning_rate": 6.208770257387989e-07, "loss": 0.4174, "step": 41645 }, { "epoch": 7.93, "learning_rate": 6.208198284080076e-07, "loss": 1.876, "step": 41646 }, { "epoch": 7.93, "learning_rate": 6.207626310772164e-07, "loss": 2.2836, "step": 41647 }, { "epoch": 7.93, "learning_rate": 6.207054337464252e-07, "loss": 1.0666, "step": 41648 }, { "epoch": 7.93, "learning_rate": 6.20648236415634e-07, "loss": 1.6532, "step": 41649 }, { "epoch": 7.93, "learning_rate": 6.205910390848427e-07, "loss": 2.1346, "step": 41650 }, { "epoch": 7.93, "learning_rate": 6.205338417540516e-07, "loss": 2.2392, "step": 41651 }, { "epoch": 7.93, "learning_rate": 6.204766444232603e-07, "loss": 1.8302, "step": 41652 }, { "epoch": 7.93, "learning_rate": 6.20419447092469e-07, "loss": 1.0416, "step": 41653 }, { "epoch": 7.93, "learning_rate": 6.203622497616778e-07, "loss": 2.8937, "step": 41654 }, { "epoch": 7.93, "learning_rate": 6.203050524308865e-07, "loss": 0.7758, "step": 41655 }, { "epoch": 7.93, "learning_rate": 6.202478551000954e-07, "loss": 0.8928, "step": 41656 }, { "epoch": 7.93, "learning_rate": 6.201906577693041e-07, "loss": 0.4553, "step": 41657 }, { "epoch": 7.93, "learning_rate": 6.201334604385128e-07, "loss": 1.3785, "step": 41658 }, { "epoch": 7.94, "learning_rate": 6.200762631077217e-07, "loss": 2.2628, "step": 41659 }, { "epoch": 7.94, "learning_rate": 6.200190657769304e-07, "loss": 1.5106, "step": 41660 }, { "epoch": 7.94, "learning_rate": 6.199618684461392e-07, "loss": 0.7049, "step": 41661 }, { "epoch": 7.94, "learning_rate": 6.19904671115348e-07, "loss": 1.1629, "step": 41662 }, { "epoch": 7.94, "learning_rate": 6.198474737845567e-07, "loss": 1.6817, "step": 41663 }, { "epoch": 7.94, "learning_rate": 6.197902764537655e-07, "loss": 1.2471, "step": 41664 }, { "epoch": 7.94, "learning_rate": 6.197330791229743e-07, "loss": 1.1348, "step": 41665 }, { "epoch": 7.94, "learning_rate": 6.19675881792183e-07, "loss": 1.0682, "step": 41666 }, { "epoch": 7.94, "learning_rate": 6.196186844613918e-07, "loss": 1.7142, "step": 41667 }, { "epoch": 7.94, "learning_rate": 6.195614871306006e-07, "loss": 0.9933, "step": 41668 }, { "epoch": 7.94, "learning_rate": 6.195042897998093e-07, "loss": 1.4309, "step": 41669 }, { "epoch": 7.94, "learning_rate": 6.194470924690181e-07, "loss": 2.1805, "step": 41670 }, { "epoch": 7.94, "learning_rate": 6.193898951382269e-07, "loss": 1.0273, "step": 41671 }, { "epoch": 7.94, "learning_rate": 6.193326978074356e-07, "loss": 0.3579, "step": 41672 }, { "epoch": 7.94, "learning_rate": 6.192755004766445e-07, "loss": 1.3229, "step": 41673 }, { "epoch": 7.94, "learning_rate": 6.192183031458532e-07, "loss": 1.9286, "step": 41674 }, { "epoch": 7.94, "learning_rate": 6.19161105815062e-07, "loss": 2.0448, "step": 41675 }, { "epoch": 7.94, "learning_rate": 6.191039084842708e-07, "loss": 2.4257, "step": 41676 }, { "epoch": 7.94, "learning_rate": 6.190467111534795e-07, "loss": 1.2727, "step": 41677 }, { "epoch": 7.94, "learning_rate": 6.189895138226883e-07, "loss": 1.934, "step": 41678 }, { "epoch": 7.94, "learning_rate": 6.189323164918971e-07, "loss": 2.3898, "step": 41679 }, { "epoch": 7.94, "learning_rate": 6.188751191611058e-07, "loss": 1.4251, "step": 41680 }, { "epoch": 7.94, "learning_rate": 6.188179218303146e-07, "loss": 1.7468, "step": 41681 }, { "epoch": 7.94, "learning_rate": 6.187607244995233e-07, "loss": 2.1999, "step": 41682 }, { "epoch": 7.94, "learning_rate": 6.187035271687321e-07, "loss": 1.5467, "step": 41683 }, { "epoch": 7.94, "learning_rate": 6.186463298379409e-07, "loss": 2.5643, "step": 41684 }, { "epoch": 7.94, "learning_rate": 6.185891325071497e-07, "loss": 1.7473, "step": 41685 }, { "epoch": 7.94, "learning_rate": 6.185319351763584e-07, "loss": 3.0583, "step": 41686 }, { "epoch": 7.94, "learning_rate": 6.184747378455672e-07, "loss": 2.1768, "step": 41687 }, { "epoch": 7.94, "learning_rate": 6.18417540514776e-07, "loss": 0.4458, "step": 41688 }, { "epoch": 7.94, "learning_rate": 6.183603431839848e-07, "loss": 1.3313, "step": 41689 }, { "epoch": 7.94, "learning_rate": 6.183031458531936e-07, "loss": 1.1999, "step": 41690 }, { "epoch": 7.94, "learning_rate": 6.182459485224023e-07, "loss": 2.6196, "step": 41691 }, { "epoch": 7.94, "learning_rate": 6.181887511916112e-07, "loss": 1.1197, "step": 41692 }, { "epoch": 7.94, "learning_rate": 6.181315538608199e-07, "loss": 1.5401, "step": 41693 }, { "epoch": 7.94, "learning_rate": 6.180743565300285e-07, "loss": 1.1626, "step": 41694 }, { "epoch": 7.94, "learning_rate": 6.180171591992374e-07, "loss": 2.582, "step": 41695 }, { "epoch": 7.94, "learning_rate": 6.179599618684461e-07, "loss": 0.6439, "step": 41696 }, { "epoch": 7.94, "learning_rate": 6.179027645376549e-07, "loss": 2.0634, "step": 41697 }, { "epoch": 7.94, "learning_rate": 6.178455672068637e-07, "loss": 2.3058, "step": 41698 }, { "epoch": 7.94, "learning_rate": 6.177883698760724e-07, "loss": 3.0123, "step": 41699 }, { "epoch": 7.94, "learning_rate": 6.177311725452812e-07, "loss": 1.6028, "step": 41700 }, { "epoch": 7.94, "learning_rate": 6.1767397521449e-07, "loss": 2.0915, "step": 41701 }, { "epoch": 7.94, "learning_rate": 6.176167778836988e-07, "loss": 1.39, "step": 41702 }, { "epoch": 7.94, "learning_rate": 6.175595805529076e-07, "loss": 0.9719, "step": 41703 }, { "epoch": 7.94, "learning_rate": 6.175023832221164e-07, "loss": 1.6061, "step": 41704 }, { "epoch": 7.94, "learning_rate": 6.174451858913251e-07, "loss": 0.8761, "step": 41705 }, { "epoch": 7.94, "learning_rate": 6.173879885605339e-07, "loss": 1.0357, "step": 41706 }, { "epoch": 7.94, "learning_rate": 6.173307912297426e-07, "loss": 1.8187, "step": 41707 }, { "epoch": 7.94, "learning_rate": 6.172735938989513e-07, "loss": 0.7694, "step": 41708 }, { "epoch": 7.94, "learning_rate": 6.172163965681602e-07, "loss": 2.7715, "step": 41709 }, { "epoch": 7.94, "learning_rate": 6.171591992373689e-07, "loss": 1.2943, "step": 41710 }, { "epoch": 7.94, "learning_rate": 6.171020019065777e-07, "loss": 2.5653, "step": 41711 }, { "epoch": 7.95, "learning_rate": 6.170448045757865e-07, "loss": 1.8139, "step": 41712 }, { "epoch": 7.95, "learning_rate": 6.169876072449952e-07, "loss": 1.4067, "step": 41713 }, { "epoch": 7.95, "learning_rate": 6.169304099142041e-07, "loss": 1.561, "step": 41714 }, { "epoch": 7.95, "learning_rate": 6.168732125834128e-07, "loss": 1.7518, "step": 41715 }, { "epoch": 7.95, "learning_rate": 6.168160152526216e-07, "loss": 0.4709, "step": 41716 }, { "epoch": 7.95, "learning_rate": 6.167588179218304e-07, "loss": 0.9169, "step": 41717 }, { "epoch": 7.95, "learning_rate": 6.167016205910391e-07, "loss": 1.087, "step": 41718 }, { "epoch": 7.95, "learning_rate": 6.166444232602479e-07, "loss": 2.7815, "step": 41719 }, { "epoch": 7.95, "learning_rate": 6.165872259294566e-07, "loss": 1.847, "step": 41720 }, { "epoch": 7.95, "learning_rate": 6.165300285986654e-07, "loss": 1.6433, "step": 41721 }, { "epoch": 7.95, "learning_rate": 6.164728312678741e-07, "loss": 1.945, "step": 41722 }, { "epoch": 7.95, "learning_rate": 6.164156339370829e-07, "loss": 2.5898, "step": 41723 }, { "epoch": 7.95, "learning_rate": 6.163584366062917e-07, "loss": 1.0276, "step": 41724 }, { "epoch": 7.95, "learning_rate": 6.163012392755005e-07, "loss": 1.2131, "step": 41725 }, { "epoch": 7.95, "learning_rate": 6.162440419447093e-07, "loss": 0.7013, "step": 41726 }, { "epoch": 7.95, "learning_rate": 6.16186844613918e-07, "loss": 2.0994, "step": 41727 }, { "epoch": 7.95, "learning_rate": 6.161296472831268e-07, "loss": 2.2096, "step": 41728 }, { "epoch": 7.95, "learning_rate": 6.160724499523356e-07, "loss": 1.0609, "step": 41729 }, { "epoch": 7.95, "learning_rate": 6.160152526215443e-07, "loss": 1.0438, "step": 41730 }, { "epoch": 7.95, "learning_rate": 6.159580552907532e-07, "loss": 2.3363, "step": 41731 }, { "epoch": 7.95, "learning_rate": 6.159008579599619e-07, "loss": 0.2908, "step": 41732 }, { "epoch": 7.95, "learning_rate": 6.158436606291706e-07, "loss": 2.2517, "step": 41733 }, { "epoch": 7.95, "learning_rate": 6.157864632983794e-07, "loss": 1.7245, "step": 41734 }, { "epoch": 7.95, "learning_rate": 6.157292659675881e-07, "loss": 1.0628, "step": 41735 }, { "epoch": 7.95, "learning_rate": 6.15672068636797e-07, "loss": 1.5224, "step": 41736 }, { "epoch": 7.95, "learning_rate": 6.156148713060057e-07, "loss": 2.069, "step": 41737 }, { "epoch": 7.95, "learning_rate": 6.155576739752145e-07, "loss": 2.504, "step": 41738 }, { "epoch": 7.95, "learning_rate": 6.155004766444233e-07, "loss": 1.1928, "step": 41739 }, { "epoch": 7.95, "learning_rate": 6.15443279313632e-07, "loss": 2.0948, "step": 41740 }, { "epoch": 7.95, "learning_rate": 6.153860819828408e-07, "loss": 3.0666, "step": 41741 }, { "epoch": 7.95, "learning_rate": 6.153288846520496e-07, "loss": 1.8859, "step": 41742 }, { "epoch": 7.95, "learning_rate": 6.152716873212584e-07, "loss": 2.0208, "step": 41743 }, { "epoch": 7.95, "learning_rate": 6.152144899904671e-07, "loss": 2.187, "step": 41744 }, { "epoch": 7.95, "learning_rate": 6.15157292659676e-07, "loss": 1.8328, "step": 41745 }, { "epoch": 7.95, "learning_rate": 6.151000953288847e-07, "loss": 2.1815, "step": 41746 }, { "epoch": 7.95, "learning_rate": 6.150428979980934e-07, "loss": 1.3079, "step": 41747 }, { "epoch": 7.95, "learning_rate": 6.149857006673022e-07, "loss": 1.6843, "step": 41748 }, { "epoch": 7.95, "learning_rate": 6.149285033365109e-07, "loss": 1.0563, "step": 41749 }, { "epoch": 7.95, "learning_rate": 6.148713060057198e-07, "loss": 2.8466, "step": 41750 }, { "epoch": 7.95, "learning_rate": 6.148141086749285e-07, "loss": 1.2658, "step": 41751 }, { "epoch": 7.95, "learning_rate": 6.147569113441372e-07, "loss": 2.4655, "step": 41752 }, { "epoch": 7.95, "learning_rate": 6.146997140133461e-07, "loss": 1.3577, "step": 41753 }, { "epoch": 7.95, "learning_rate": 6.146425166825548e-07, "loss": 1.4352, "step": 41754 }, { "epoch": 7.95, "learning_rate": 6.145853193517636e-07, "loss": 0.8239, "step": 41755 }, { "epoch": 7.95, "learning_rate": 6.145281220209724e-07, "loss": 1.3354, "step": 41756 }, { "epoch": 7.95, "learning_rate": 6.144709246901812e-07, "loss": 1.1737, "step": 41757 }, { "epoch": 7.95, "learning_rate": 6.1441372735939e-07, "loss": 2.2637, "step": 41758 }, { "epoch": 7.95, "learning_rate": 6.143565300285987e-07, "loss": 1.0287, "step": 41759 }, { "epoch": 7.95, "learning_rate": 6.142993326978074e-07, "loss": 1.5849, "step": 41760 }, { "epoch": 7.95, "learning_rate": 6.142421353670162e-07, "loss": 0.919, "step": 41761 }, { "epoch": 7.95, "learning_rate": 6.14184938036225e-07, "loss": 1.7669, "step": 41762 }, { "epoch": 7.95, "learning_rate": 6.141277407054337e-07, "loss": 2.3267, "step": 41763 }, { "epoch": 7.96, "learning_rate": 6.140705433746425e-07, "loss": 1.2672, "step": 41764 }, { "epoch": 7.96, "learning_rate": 6.140133460438513e-07, "loss": 1.6771, "step": 41765 }, { "epoch": 7.96, "learning_rate": 6.1395614871306e-07, "loss": 1.5295, "step": 41766 }, { "epoch": 7.96, "learning_rate": 6.138989513822689e-07, "loss": 1.1813, "step": 41767 }, { "epoch": 7.96, "learning_rate": 6.138417540514776e-07, "loss": 0.2751, "step": 41768 }, { "epoch": 7.96, "learning_rate": 6.137845567206864e-07, "loss": 1.4462, "step": 41769 }, { "epoch": 7.96, "learning_rate": 6.137273593898952e-07, "loss": 1.1316, "step": 41770 }, { "epoch": 7.96, "learning_rate": 6.136701620591039e-07, "loss": 1.4977, "step": 41771 }, { "epoch": 7.96, "learning_rate": 6.136129647283128e-07, "loss": 2.4081, "step": 41772 }, { "epoch": 7.96, "learning_rate": 6.135557673975214e-07, "loss": 0.8633, "step": 41773 }, { "epoch": 7.96, "learning_rate": 6.134985700667302e-07, "loss": 1.2702, "step": 41774 }, { "epoch": 7.96, "learning_rate": 6.13441372735939e-07, "loss": 1.2572, "step": 41775 }, { "epoch": 7.96, "learning_rate": 6.133841754051477e-07, "loss": 1.1487, "step": 41776 }, { "epoch": 7.96, "learning_rate": 6.133269780743565e-07, "loss": 1.7489, "step": 41777 }, { "epoch": 7.96, "learning_rate": 6.132697807435653e-07, "loss": 2.6386, "step": 41778 }, { "epoch": 7.96, "learning_rate": 6.132125834127741e-07, "loss": 0.4524, "step": 41779 }, { "epoch": 7.96, "learning_rate": 6.131553860819829e-07, "loss": 0.6494, "step": 41780 }, { "epoch": 7.96, "learning_rate": 6.130981887511916e-07, "loss": 1.6517, "step": 41781 }, { "epoch": 7.96, "learning_rate": 6.130409914204004e-07, "loss": 2.2819, "step": 41782 }, { "epoch": 7.96, "learning_rate": 6.129837940896092e-07, "loss": 0.9366, "step": 41783 }, { "epoch": 7.96, "learning_rate": 6.12926596758818e-07, "loss": 1.5338, "step": 41784 }, { "epoch": 7.96, "learning_rate": 6.128693994280267e-07, "loss": 0.9925, "step": 41785 }, { "epoch": 7.96, "learning_rate": 6.128122020972356e-07, "loss": 1.5005, "step": 41786 }, { "epoch": 7.96, "learning_rate": 6.127550047664442e-07, "loss": 0.9149, "step": 41787 }, { "epoch": 7.96, "learning_rate": 6.126978074356529e-07, "loss": 1.8806, "step": 41788 }, { "epoch": 7.96, "learning_rate": 6.126406101048618e-07, "loss": 0.6583, "step": 41789 }, { "epoch": 7.96, "learning_rate": 6.125834127740705e-07, "loss": 1.6716, "step": 41790 }, { "epoch": 7.96, "learning_rate": 6.125262154432793e-07, "loss": 1.6006, "step": 41791 }, { "epoch": 7.96, "learning_rate": 6.124690181124881e-07, "loss": 0.522, "step": 41792 }, { "epoch": 7.96, "learning_rate": 6.124118207816968e-07, "loss": 0.8842, "step": 41793 }, { "epoch": 7.96, "learning_rate": 6.123546234509057e-07, "loss": 1.8608, "step": 41794 }, { "epoch": 7.96, "learning_rate": 6.122974261201144e-07, "loss": 1.5473, "step": 41795 }, { "epoch": 7.96, "learning_rate": 6.122402287893232e-07, "loss": 1.1372, "step": 41796 }, { "epoch": 7.96, "learning_rate": 6.12183031458532e-07, "loss": 1.6194, "step": 41797 }, { "epoch": 7.96, "learning_rate": 6.121258341277408e-07, "loss": 2.1377, "step": 41798 }, { "epoch": 7.96, "learning_rate": 6.120686367969495e-07, "loss": 0.2196, "step": 41799 }, { "epoch": 7.96, "learning_rate": 6.120114394661582e-07, "loss": 1.0031, "step": 41800 }, { "epoch": 7.96, "learning_rate": 6.11954242135367e-07, "loss": 1.7772, "step": 41801 }, { "epoch": 7.96, "learning_rate": 6.118970448045758e-07, "loss": 1.1397, "step": 41802 }, { "epoch": 7.96, "learning_rate": 6.118398474737846e-07, "loss": 1.5245, "step": 41803 }, { "epoch": 7.96, "learning_rate": 6.117826501429933e-07, "loss": 0.7546, "step": 41804 }, { "epoch": 7.96, "learning_rate": 6.117254528122021e-07, "loss": 0.9842, "step": 41805 }, { "epoch": 7.96, "learning_rate": 6.116682554814109e-07, "loss": 1.0136, "step": 41806 }, { "epoch": 7.96, "learning_rate": 6.116110581506196e-07, "loss": 2.2857, "step": 41807 }, { "epoch": 7.96, "learning_rate": 6.115538608198285e-07, "loss": 1.47, "step": 41808 }, { "epoch": 7.96, "learning_rate": 6.114966634890372e-07, "loss": 0.5548, "step": 41809 }, { "epoch": 7.96, "learning_rate": 6.11439466158246e-07, "loss": 1.7981, "step": 41810 }, { "epoch": 7.96, "learning_rate": 6.113822688274548e-07, "loss": 1.1379, "step": 41811 }, { "epoch": 7.96, "learning_rate": 6.113250714966635e-07, "loss": 3.1359, "step": 41812 }, { "epoch": 7.96, "learning_rate": 6.112678741658722e-07, "loss": 1.4979, "step": 41813 }, { "epoch": 7.96, "learning_rate": 6.11210676835081e-07, "loss": 1.8915, "step": 41814 }, { "epoch": 7.96, "learning_rate": 6.111534795042898e-07, "loss": 2.0954, "step": 41815 }, { "epoch": 7.96, "learning_rate": 6.110962821734986e-07, "loss": 1.5518, "step": 41816 }, { "epoch": 7.97, "learning_rate": 6.110390848427073e-07, "loss": 1.723, "step": 41817 }, { "epoch": 7.97, "learning_rate": 6.109818875119161e-07, "loss": 1.4641, "step": 41818 }, { "epoch": 7.97, "learning_rate": 6.109246901811249e-07, "loss": 1.5573, "step": 41819 }, { "epoch": 7.97, "learning_rate": 6.108674928503337e-07, "loss": 0.9244, "step": 41820 }, { "epoch": 7.97, "learning_rate": 6.108102955195424e-07, "loss": 2.5585, "step": 41821 }, { "epoch": 7.97, "learning_rate": 6.107530981887513e-07, "loss": 1.6244, "step": 41822 }, { "epoch": 7.97, "learning_rate": 6.1069590085796e-07, "loss": 1.6719, "step": 41823 }, { "epoch": 7.97, "learning_rate": 6.106387035271688e-07, "loss": 1.2711, "step": 41824 }, { "epoch": 7.97, "learning_rate": 6.105815061963776e-07, "loss": 0.8583, "step": 41825 }, { "epoch": 7.97, "learning_rate": 6.105243088655863e-07, "loss": 2.0139, "step": 41826 }, { "epoch": 7.97, "learning_rate": 6.10467111534795e-07, "loss": 0.829, "step": 41827 }, { "epoch": 7.97, "learning_rate": 6.104099142040038e-07, "loss": 0.9991, "step": 41828 }, { "epoch": 7.97, "learning_rate": 6.103527168732125e-07, "loss": 2.648, "step": 41829 }, { "epoch": 7.97, "learning_rate": 6.102955195424214e-07, "loss": 2.3417, "step": 41830 }, { "epoch": 7.97, "learning_rate": 6.102383222116301e-07, "loss": 1.0733, "step": 41831 }, { "epoch": 7.97, "learning_rate": 6.101811248808389e-07, "loss": 1.3384, "step": 41832 }, { "epoch": 7.97, "learning_rate": 6.101239275500477e-07, "loss": 2.2208, "step": 41833 }, { "epoch": 7.97, "learning_rate": 6.100667302192565e-07, "loss": 1.5784, "step": 41834 }, { "epoch": 7.97, "learning_rate": 6.100095328884652e-07, "loss": 2.758, "step": 41835 }, { "epoch": 7.97, "learning_rate": 6.09952335557674e-07, "loss": 1.2874, "step": 41836 }, { "epoch": 7.97, "learning_rate": 6.098951382268828e-07, "loss": 0.9471, "step": 41837 }, { "epoch": 7.97, "learning_rate": 6.098379408960916e-07, "loss": 2.6177, "step": 41838 }, { "epoch": 7.97, "learning_rate": 6.097807435653004e-07, "loss": 0.9481, "step": 41839 }, { "epoch": 7.97, "learning_rate": 6.09723546234509e-07, "loss": 0.975, "step": 41840 }, { "epoch": 7.97, "learning_rate": 6.096663489037178e-07, "loss": 0.7727, "step": 41841 }, { "epoch": 7.97, "learning_rate": 6.096091515729266e-07, "loss": 1.6409, "step": 41842 }, { "epoch": 7.97, "learning_rate": 6.095519542421353e-07, "loss": 1.6123, "step": 41843 }, { "epoch": 7.97, "learning_rate": 6.094947569113442e-07, "loss": 2.5152, "step": 41844 }, { "epoch": 7.97, "learning_rate": 6.094375595805529e-07, "loss": 1.253, "step": 41845 }, { "epoch": 7.97, "learning_rate": 6.093803622497617e-07, "loss": 1.7803, "step": 41846 }, { "epoch": 7.97, "learning_rate": 6.093231649189705e-07, "loss": 2.9797, "step": 41847 }, { "epoch": 7.97, "learning_rate": 6.092659675881792e-07, "loss": 2.1356, "step": 41848 }, { "epoch": 7.97, "learning_rate": 6.09208770257388e-07, "loss": 1.2232, "step": 41849 }, { "epoch": 7.97, "learning_rate": 6.091515729265968e-07, "loss": 0.8816, "step": 41850 }, { "epoch": 7.97, "learning_rate": 6.090943755958056e-07, "loss": 1.7105, "step": 41851 }, { "epoch": 7.97, "learning_rate": 6.090371782650144e-07, "loss": 1.9906, "step": 41852 }, { "epoch": 7.97, "learning_rate": 6.08979980934223e-07, "loss": 1.9497, "step": 41853 }, { "epoch": 7.97, "learning_rate": 6.089227836034318e-07, "loss": 1.5076, "step": 41854 }, { "epoch": 7.97, "learning_rate": 6.088655862726406e-07, "loss": 2.664, "step": 41855 }, { "epoch": 7.97, "learning_rate": 6.088083889418494e-07, "loss": 1.3749, "step": 41856 }, { "epoch": 7.97, "learning_rate": 6.087511916110581e-07, "loss": 2.1633, "step": 41857 }, { "epoch": 7.97, "learning_rate": 6.086939942802669e-07, "loss": 0.8866, "step": 41858 }, { "epoch": 7.97, "learning_rate": 6.086367969494757e-07, "loss": 2.0616, "step": 41859 }, { "epoch": 7.97, "learning_rate": 6.085795996186845e-07, "loss": 0.6717, "step": 41860 }, { "epoch": 7.97, "learning_rate": 6.085224022878933e-07, "loss": 1.6157, "step": 41861 }, { "epoch": 7.97, "learning_rate": 6.08465204957102e-07, "loss": 0.291, "step": 41862 }, { "epoch": 7.97, "learning_rate": 6.084080076263108e-07, "loss": 1.741, "step": 41863 }, { "epoch": 7.97, "learning_rate": 6.083508102955196e-07, "loss": 1.9194, "step": 41864 }, { "epoch": 7.97, "learning_rate": 6.082936129647283e-07, "loss": 0.6013, "step": 41865 }, { "epoch": 7.97, "learning_rate": 6.082364156339372e-07, "loss": 1.6373, "step": 41866 }, { "epoch": 7.97, "learning_rate": 6.081792183031458e-07, "loss": 2.7525, "step": 41867 }, { "epoch": 7.97, "learning_rate": 6.081220209723546e-07, "loss": 2.4281, "step": 41868 }, { "epoch": 7.98, "learning_rate": 6.080648236415634e-07, "loss": 2.2645, "step": 41869 }, { "epoch": 7.98, "learning_rate": 6.080076263107721e-07, "loss": 2.5657, "step": 41870 }, { "epoch": 7.98, "learning_rate": 6.079504289799809e-07, "loss": 1.6433, "step": 41871 }, { "epoch": 7.98, "learning_rate": 6.078932316491897e-07, "loss": 0.6045, "step": 41872 }, { "epoch": 7.98, "learning_rate": 6.078360343183985e-07, "loss": 1.8545, "step": 41873 }, { "epoch": 7.98, "learning_rate": 6.077788369876073e-07, "loss": 1.2178, "step": 41874 }, { "epoch": 7.98, "learning_rate": 6.077216396568161e-07, "loss": 0.4617, "step": 41875 }, { "epoch": 7.98, "learning_rate": 6.076644423260248e-07, "loss": 0.8592, "step": 41876 }, { "epoch": 7.98, "learning_rate": 6.076072449952336e-07, "loss": 0.7004, "step": 41877 }, { "epoch": 7.98, "learning_rate": 6.075500476644424e-07, "loss": 1.8934, "step": 41878 }, { "epoch": 7.98, "learning_rate": 6.074928503336511e-07, "loss": 0.7302, "step": 41879 }, { "epoch": 7.98, "learning_rate": 6.0743565300286e-07, "loss": 1.0289, "step": 41880 }, { "epoch": 7.98, "learning_rate": 6.073784556720686e-07, "loss": 1.4084, "step": 41881 }, { "epoch": 7.98, "learning_rate": 6.073212583412774e-07, "loss": 3.2739, "step": 41882 }, { "epoch": 7.98, "learning_rate": 6.072640610104862e-07, "loss": 1.0954, "step": 41883 }, { "epoch": 7.98, "learning_rate": 6.072068636796949e-07, "loss": 0.9649, "step": 41884 }, { "epoch": 7.98, "learning_rate": 6.071496663489037e-07, "loss": 1.7409, "step": 41885 }, { "epoch": 7.98, "learning_rate": 6.070924690181125e-07, "loss": 1.0823, "step": 41886 }, { "epoch": 7.98, "learning_rate": 6.070352716873213e-07, "loss": 2.0699, "step": 41887 }, { "epoch": 7.98, "learning_rate": 6.069780743565301e-07, "loss": 0.725, "step": 41888 }, { "epoch": 7.98, "learning_rate": 6.069208770257388e-07, "loss": 1.5116, "step": 41889 }, { "epoch": 7.98, "learning_rate": 6.068636796949476e-07, "loss": 0.6403, "step": 41890 }, { "epoch": 7.98, "learning_rate": 6.068064823641564e-07, "loss": 0.6296, "step": 41891 }, { "epoch": 7.98, "learning_rate": 6.067492850333652e-07, "loss": 0.9537, "step": 41892 }, { "epoch": 7.98, "learning_rate": 6.066920877025738e-07, "loss": 1.5138, "step": 41893 }, { "epoch": 7.98, "learning_rate": 6.066348903717826e-07, "loss": 2.5496, "step": 41894 }, { "epoch": 7.98, "learning_rate": 6.065776930409914e-07, "loss": 2.8939, "step": 41895 }, { "epoch": 7.98, "learning_rate": 6.065204957102002e-07, "loss": 0.5642, "step": 41896 }, { "epoch": 7.98, "learning_rate": 6.06463298379409e-07, "loss": 0.9022, "step": 41897 }, { "epoch": 7.98, "learning_rate": 6.064061010486177e-07, "loss": 1.1752, "step": 41898 }, { "epoch": 7.98, "learning_rate": 6.063489037178265e-07, "loss": 1.4967, "step": 41899 }, { "epoch": 7.98, "learning_rate": 6.062917063870353e-07, "loss": 0.7811, "step": 41900 }, { "epoch": 7.98, "learning_rate": 6.06234509056244e-07, "loss": 1.2038, "step": 41901 }, { "epoch": 7.98, "learning_rate": 6.061773117254529e-07, "loss": 2.2904, "step": 41902 }, { "epoch": 7.98, "learning_rate": 6.061201143946616e-07, "loss": 1.7052, "step": 41903 }, { "epoch": 7.98, "learning_rate": 6.060629170638704e-07, "loss": 3.2123, "step": 41904 }, { "epoch": 7.98, "learning_rate": 6.060057197330792e-07, "loss": 2.5763, "step": 41905 }, { "epoch": 7.98, "learning_rate": 6.059485224022879e-07, "loss": 2.6048, "step": 41906 }, { "epoch": 7.98, "learning_rate": 6.058913250714966e-07, "loss": 0.3368, "step": 41907 }, { "epoch": 7.98, "learning_rate": 6.058341277407054e-07, "loss": 0.8689, "step": 41908 }, { "epoch": 7.98, "learning_rate": 6.057769304099142e-07, "loss": 1.9861, "step": 41909 }, { "epoch": 7.98, "learning_rate": 6.05719733079123e-07, "loss": 0.541, "step": 41910 }, { "epoch": 7.98, "learning_rate": 6.056625357483317e-07, "loss": 1.9111, "step": 41911 }, { "epoch": 7.98, "learning_rate": 6.056053384175405e-07, "loss": 1.2775, "step": 41912 }, { "epoch": 7.98, "learning_rate": 6.055481410867493e-07, "loss": 0.8371, "step": 41913 }, { "epoch": 7.98, "learning_rate": 6.054909437559581e-07, "loss": 0.9123, "step": 41914 }, { "epoch": 7.98, "learning_rate": 6.054337464251668e-07, "loss": 0.8764, "step": 41915 }, { "epoch": 7.98, "learning_rate": 6.053765490943757e-07, "loss": 2.0359, "step": 41916 }, { "epoch": 7.98, "learning_rate": 6.053193517635844e-07, "loss": 2.0244, "step": 41917 }, { "epoch": 7.98, "learning_rate": 6.052621544327932e-07, "loss": 1.8587, "step": 41918 }, { "epoch": 7.98, "learning_rate": 6.05204957102002e-07, "loss": 1.8508, "step": 41919 }, { "epoch": 7.98, "learning_rate": 6.051477597712106e-07, "loss": 1.9069, "step": 41920 }, { "epoch": 7.98, "learning_rate": 6.050905624404194e-07, "loss": 1.1751, "step": 41921 }, { "epoch": 7.99, "learning_rate": 6.050333651096282e-07, "loss": 2.0632, "step": 41922 }, { "epoch": 7.99, "learning_rate": 6.049761677788369e-07, "loss": 2.0242, "step": 41923 }, { "epoch": 7.99, "learning_rate": 6.049189704480458e-07, "loss": 1.5861, "step": 41924 }, { "epoch": 7.99, "learning_rate": 6.048617731172545e-07, "loss": 0.818, "step": 41925 }, { "epoch": 7.99, "learning_rate": 6.048045757864633e-07, "loss": 1.05, "step": 41926 }, { "epoch": 7.99, "learning_rate": 6.047473784556721e-07, "loss": 1.2601, "step": 41927 }, { "epoch": 7.99, "learning_rate": 6.046901811248809e-07, "loss": 2.0838, "step": 41928 }, { "epoch": 7.99, "learning_rate": 6.046329837940896e-07, "loss": 0.6365, "step": 41929 }, { "epoch": 7.99, "learning_rate": 6.045757864632984e-07, "loss": 0.8131, "step": 41930 }, { "epoch": 7.99, "learning_rate": 6.045185891325072e-07, "loss": 0.3429, "step": 41931 }, { "epoch": 7.99, "learning_rate": 6.04461391801716e-07, "loss": 0.8766, "step": 41932 }, { "epoch": 7.99, "learning_rate": 6.044041944709248e-07, "loss": 1.4974, "step": 41933 }, { "epoch": 7.99, "learning_rate": 6.043469971401334e-07, "loss": 1.5851, "step": 41934 }, { "epoch": 7.99, "learning_rate": 6.042897998093422e-07, "loss": 0.9886, "step": 41935 }, { "epoch": 7.99, "learning_rate": 6.04232602478551e-07, "loss": 0.8424, "step": 41936 }, { "epoch": 7.99, "learning_rate": 6.041754051477597e-07, "loss": 1.2333, "step": 41937 }, { "epoch": 7.99, "learning_rate": 6.041182078169686e-07, "loss": 0.7793, "step": 41938 }, { "epoch": 7.99, "learning_rate": 6.040610104861773e-07, "loss": 0.7411, "step": 41939 }, { "epoch": 7.99, "learning_rate": 6.040038131553861e-07, "loss": 1.2617, "step": 41940 }, { "epoch": 7.99, "learning_rate": 6.039466158245949e-07, "loss": 2.8117, "step": 41941 }, { "epoch": 7.99, "learning_rate": 6.038894184938036e-07, "loss": 1.166, "step": 41942 }, { "epoch": 7.99, "learning_rate": 6.038322211630124e-07, "loss": 3.897, "step": 41943 }, { "epoch": 7.99, "learning_rate": 6.037750238322212e-07, "loss": 1.4885, "step": 41944 }, { "epoch": 7.99, "learning_rate": 6.0371782650143e-07, "loss": 1.4918, "step": 41945 }, { "epoch": 7.99, "learning_rate": 6.036606291706388e-07, "loss": 2.2767, "step": 41946 }, { "epoch": 7.99, "learning_rate": 6.036034318398474e-07, "loss": 2.0107, "step": 41947 }, { "epoch": 7.99, "learning_rate": 6.035462345090562e-07, "loss": 0.7889, "step": 41948 }, { "epoch": 7.99, "learning_rate": 6.03489037178265e-07, "loss": 1.6475, "step": 41949 }, { "epoch": 7.99, "learning_rate": 6.034318398474738e-07, "loss": 1.9791, "step": 41950 }, { "epoch": 7.99, "learning_rate": 6.033746425166825e-07, "loss": 1.5377, "step": 41951 }, { "epoch": 7.99, "learning_rate": 6.033174451858914e-07, "loss": 1.3129, "step": 41952 }, { "epoch": 7.99, "learning_rate": 6.032602478551001e-07, "loss": 2.0691, "step": 41953 }, { "epoch": 7.99, "learning_rate": 6.032030505243089e-07, "loss": 0.9207, "step": 41954 }, { "epoch": 7.99, "learning_rate": 6.031458531935177e-07, "loss": 2.724, "step": 41955 }, { "epoch": 7.99, "learning_rate": 6.030886558627264e-07, "loss": 1.952, "step": 41956 }, { "epoch": 7.99, "learning_rate": 6.030314585319353e-07, "loss": 1.3386, "step": 41957 }, { "epoch": 7.99, "learning_rate": 6.02974261201144e-07, "loss": 1.4283, "step": 41958 }, { "epoch": 7.99, "learning_rate": 6.029170638703527e-07, "loss": 2.295, "step": 41959 }, { "epoch": 7.99, "learning_rate": 6.028598665395616e-07, "loss": 1.6701, "step": 41960 }, { "epoch": 7.99, "learning_rate": 6.028026692087702e-07, "loss": 1.6079, "step": 41961 }, { "epoch": 7.99, "learning_rate": 6.02745471877979e-07, "loss": 1.8817, "step": 41962 }, { "epoch": 7.99, "learning_rate": 6.026882745471878e-07, "loss": 3.1371, "step": 41963 }, { "epoch": 7.99, "learning_rate": 6.026310772163965e-07, "loss": 0.6321, "step": 41964 }, { "epoch": 7.99, "learning_rate": 6.025738798856053e-07, "loss": 1.5353, "step": 41965 }, { "epoch": 7.99, "learning_rate": 6.025166825548141e-07, "loss": 1.4824, "step": 41966 }, { "epoch": 7.99, "learning_rate": 6.024594852240229e-07, "loss": 1.4474, "step": 41967 }, { "epoch": 7.99, "learning_rate": 6.024022878932317e-07, "loss": 1.6718, "step": 41968 }, { "epoch": 7.99, "learning_rate": 6.023450905624405e-07, "loss": 2.1781, "step": 41969 }, { "epoch": 7.99, "learning_rate": 6.022878932316492e-07, "loss": 1.3169, "step": 41970 }, { "epoch": 7.99, "learning_rate": 6.02230695900858e-07, "loss": 0.6089, "step": 41971 }, { "epoch": 7.99, "learning_rate": 6.021734985700668e-07, "loss": 1.6107, "step": 41972 }, { "epoch": 7.99, "learning_rate": 6.021163012392754e-07, "loss": 2.1336, "step": 41973 }, { "epoch": 8.0, "learning_rate": 6.020591039084843e-07, "loss": 2.7853, "step": 41974 }, { "epoch": 8.0, "learning_rate": 6.02001906577693e-07, "loss": 2.0397, "step": 41975 }, { "epoch": 8.0, "learning_rate": 6.019447092469018e-07, "loss": 2.5169, "step": 41976 }, { "epoch": 8.0, "learning_rate": 6.018875119161106e-07, "loss": 0.7082, "step": 41977 }, { "epoch": 8.0, "learning_rate": 6.018303145853193e-07, "loss": 2.08, "step": 41978 }, { "epoch": 8.0, "learning_rate": 6.017731172545282e-07, "loss": 1.2403, "step": 41979 }, { "epoch": 8.0, "learning_rate": 6.017159199237369e-07, "loss": 1.2542, "step": 41980 }, { "epoch": 8.0, "learning_rate": 6.016587225929457e-07, "loss": 1.5751, "step": 41981 }, { "epoch": 8.0, "learning_rate": 6.016015252621545e-07, "loss": 1.0558, "step": 41982 }, { "epoch": 8.0, "learning_rate": 6.015443279313632e-07, "loss": 2.0648, "step": 41983 }, { "epoch": 8.0, "learning_rate": 6.01487130600572e-07, "loss": 2.5629, "step": 41984 }, { "epoch": 8.0, "learning_rate": 6.014299332697808e-07, "loss": 1.3232, "step": 41985 }, { "epoch": 8.0, "learning_rate": 6.013727359389896e-07, "loss": 1.9577, "step": 41986 }, { "epoch": 8.0, "learning_rate": 6.013155386081982e-07, "loss": 2.3408, "step": 41987 }, { "epoch": 8.0, "learning_rate": 6.01258341277407e-07, "loss": 1.5815, "step": 41988 }, { "epoch": 8.0, "learning_rate": 6.012011439466158e-07, "loss": 1.653, "step": 41989 }, { "epoch": 8.0, "learning_rate": 6.011439466158246e-07, "loss": 1.5324, "step": 41990 }, { "epoch": 8.0, "learning_rate": 6.010867492850334e-07, "loss": 1.7431, "step": 41991 }, { "epoch": 8.0, "learning_rate": 6.010295519542421e-07, "loss": 2.2229, "step": 41992 }, { "epoch": 8.0, "learning_rate": 6.00972354623451e-07, "loss": 2.3869, "step": 41993 }, { "epoch": 8.0, "learning_rate": 6.009151572926597e-07, "loss": 1.5111, "step": 41994 }, { "epoch": 8.0, "learning_rate": 6.008579599618684e-07, "loss": 1.1717, "step": 41995 }, { "epoch": 8.0, "learning_rate": 6.008007626310773e-07, "loss": 2.8538, "step": 41996 }, { "epoch": 8.0, "learning_rate": 6.00743565300286e-07, "loss": 3.109, "step": 41997 }, { "epoch": 8.0, "learning_rate": 6.006863679694948e-07, "loss": 2.1437, "step": 41998 }, { "epoch": 8.0, "learning_rate": 6.006291706387036e-07, "loss": 0.9398, "step": 41999 }, { "epoch": 8.0, "learning_rate": 6.005719733079122e-07, "loss": 1.1662, "step": 42000 }, { "epoch": 8.0, "eval_accuracy": 0.6078011144449207, "eval_loss": 1.5368942022323608, "eval_runtime": 128.8164, "eval_samples_per_second": 18.111, "eval_steps_per_second": 4.534, "step": 42000 }, { "epoch": 8.0, "learning_rate": 6.005147759771211e-07, "loss": 1.8529, "step": 42001 }, { "epoch": 8.0, "learning_rate": 6.004575786463298e-07, "loss": 1.0994, "step": 42002 }, { "epoch": 8.0, "learning_rate": 6.004003813155386e-07, "loss": 0.4974, "step": 42003 }, { "epoch": 8.0, "learning_rate": 6.003431839847474e-07, "loss": 0.8213, "step": 42004 }, { "epoch": 8.0, "learning_rate": 6.002859866539562e-07, "loss": 0.3384, "step": 42005 }, { "epoch": 8.0, "learning_rate": 6.002287893231649e-07, "loss": 1.2379, "step": 42006 }, { "epoch": 8.0, "learning_rate": 6.001715919923737e-07, "loss": 2.5201, "step": 42007 }, { "epoch": 8.0, "learning_rate": 6.001143946615825e-07, "loss": 0.4377, "step": 42008 }, { "epoch": 8.0, "learning_rate": 6.000571973307912e-07, "loss": 2.61, "step": 42009 }, { "epoch": 8.0, "learning_rate": 6.000000000000001e-07, "loss": 1.7327, "step": 42010 }, { "epoch": 8.0, "learning_rate": 5.999428026692088e-07, "loss": 0.5242, "step": 42011 }, { "epoch": 8.0, "learning_rate": 5.998856053384176e-07, "loss": 1.2039, "step": 42012 }, { "epoch": 8.0, "learning_rate": 5.998284080076264e-07, "loss": 1.9269, "step": 42013 }, { "epoch": 8.0, "learning_rate": 5.99771210676835e-07, "loss": 1.6217, "step": 42014 }, { "epoch": 8.0, "learning_rate": 5.997140133460439e-07, "loss": 0.3126, "step": 42015 }, { "epoch": 8.0, "learning_rate": 5.996568160152526e-07, "loss": 3.3609, "step": 42016 }, { "epoch": 8.0, "learning_rate": 5.995996186844614e-07, "loss": 1.0014, "step": 42017 }, { "epoch": 8.0, "learning_rate": 5.995424213536702e-07, "loss": 1.0941, "step": 42018 }, { "epoch": 8.0, "learning_rate": 5.994852240228789e-07, "loss": 1.9461, "step": 42019 }, { "epoch": 8.0, "learning_rate": 5.994280266920877e-07, "loss": 1.9329, "step": 42020 }, { "epoch": 8.0, "learning_rate": 5.993708293612965e-07, "loss": 0.9057, "step": 42021 }, { "epoch": 8.0, "learning_rate": 5.993136320305053e-07, "loss": 0.9101, "step": 42022 }, { "epoch": 8.0, "learning_rate": 5.992564346997141e-07, "loss": 0.6238, "step": 42023 }, { "epoch": 8.0, "learning_rate": 5.991992373689228e-07, "loss": 1.1084, "step": 42024 }, { "epoch": 8.0, "learning_rate": 5.991420400381316e-07, "loss": 1.7407, "step": 42025 }, { "epoch": 8.0, "learning_rate": 5.990848427073404e-07, "loss": 3.7747, "step": 42026 }, { "epoch": 8.01, "learning_rate": 5.990276453765492e-07, "loss": 1.9973, "step": 42027 }, { "epoch": 8.01, "learning_rate": 5.989704480457578e-07, "loss": 1.5281, "step": 42028 }, { "epoch": 8.01, "learning_rate": 5.989132507149666e-07, "loss": 1.3945, "step": 42029 }, { "epoch": 8.01, "learning_rate": 5.988560533841754e-07, "loss": 1.3607, "step": 42030 }, { "epoch": 8.01, "learning_rate": 5.987988560533841e-07, "loss": 1.0115, "step": 42031 }, { "epoch": 8.01, "learning_rate": 5.98741658722593e-07, "loss": 1.2649, "step": 42032 }, { "epoch": 8.01, "learning_rate": 5.986844613918017e-07, "loss": 1.5957, "step": 42033 }, { "epoch": 8.01, "learning_rate": 5.986272640610105e-07, "loss": 2.4621, "step": 42034 }, { "epoch": 8.01, "learning_rate": 5.985700667302193e-07, "loss": 0.9524, "step": 42035 }, { "epoch": 8.01, "learning_rate": 5.98512869399428e-07, "loss": 1.2167, "step": 42036 }, { "epoch": 8.01, "learning_rate": 5.984556720686369e-07, "loss": 1.3812, "step": 42037 }, { "epoch": 8.01, "learning_rate": 5.983984747378456e-07, "loss": 1.615, "step": 42038 }, { "epoch": 8.01, "learning_rate": 5.983412774070544e-07, "loss": 0.7789, "step": 42039 }, { "epoch": 8.01, "learning_rate": 5.982840800762632e-07, "loss": 1.3763, "step": 42040 }, { "epoch": 8.01, "learning_rate": 5.982268827454718e-07, "loss": 1.2088, "step": 42041 }, { "epoch": 8.01, "learning_rate": 5.981696854146806e-07, "loss": 3.3332, "step": 42042 }, { "epoch": 8.01, "learning_rate": 5.981124880838894e-07, "loss": 1.8932, "step": 42043 }, { "epoch": 8.01, "learning_rate": 5.980552907530982e-07, "loss": 2.4154, "step": 42044 }, { "epoch": 8.01, "learning_rate": 5.97998093422307e-07, "loss": 0.3509, "step": 42045 }, { "epoch": 8.01, "learning_rate": 5.979408960915158e-07, "loss": 1.4771, "step": 42046 }, { "epoch": 8.01, "learning_rate": 5.978836987607245e-07, "loss": 2.118, "step": 42047 }, { "epoch": 8.01, "learning_rate": 5.978265014299333e-07, "loss": 1.5474, "step": 42048 }, { "epoch": 8.01, "learning_rate": 5.977693040991421e-07, "loss": 0.8565, "step": 42049 }, { "epoch": 8.01, "learning_rate": 5.977121067683508e-07, "loss": 1.4723, "step": 42050 }, { "epoch": 8.01, "learning_rate": 5.976549094375597e-07, "loss": 0.689, "step": 42051 }, { "epoch": 8.01, "learning_rate": 5.975977121067684e-07, "loss": 1.4811, "step": 42052 }, { "epoch": 8.01, "learning_rate": 5.97540514775977e-07, "loss": 1.6678, "step": 42053 }, { "epoch": 8.01, "learning_rate": 5.97483317445186e-07, "loss": 0.8487, "step": 42054 }, { "epoch": 8.01, "learning_rate": 5.974261201143946e-07, "loss": 0.567, "step": 42055 }, { "epoch": 8.01, "learning_rate": 5.973689227836034e-07, "loss": 1.2573, "step": 42056 }, { "epoch": 8.01, "learning_rate": 5.973117254528122e-07, "loss": 0.5654, "step": 42057 }, { "epoch": 8.01, "learning_rate": 5.97254528122021e-07, "loss": 1.2517, "step": 42058 }, { "epoch": 8.01, "learning_rate": 5.971973307912298e-07, "loss": 2.7039, "step": 42059 }, { "epoch": 8.01, "learning_rate": 5.971401334604385e-07, "loss": 1.5022, "step": 42060 }, { "epoch": 8.01, "learning_rate": 5.970829361296473e-07, "loss": 1.5808, "step": 42061 }, { "epoch": 8.01, "learning_rate": 5.970257387988561e-07, "loss": 1.3706, "step": 42062 }, { "epoch": 8.01, "learning_rate": 5.969685414680649e-07, "loss": 2.1921, "step": 42063 }, { "epoch": 8.01, "learning_rate": 5.969113441372736e-07, "loss": 2.6473, "step": 42064 }, { "epoch": 8.01, "learning_rate": 5.968541468064824e-07, "loss": 3.3189, "step": 42065 }, { "epoch": 8.01, "learning_rate": 5.967969494756912e-07, "loss": 2.451, "step": 42066 }, { "epoch": 8.01, "learning_rate": 5.967397521449e-07, "loss": 1.5637, "step": 42067 }, { "epoch": 8.01, "learning_rate": 5.966825548141087e-07, "loss": 1.32, "step": 42068 }, { "epoch": 8.01, "learning_rate": 5.966253574833174e-07, "loss": 2.0327, "step": 42069 }, { "epoch": 8.01, "learning_rate": 5.965681601525262e-07, "loss": 1.5024, "step": 42070 }, { "epoch": 8.01, "learning_rate": 5.96510962821735e-07, "loss": 2.3525, "step": 42071 }, { "epoch": 8.01, "learning_rate": 5.964537654909437e-07, "loss": 2.3577, "step": 42072 }, { "epoch": 8.01, "learning_rate": 5.963965681601526e-07, "loss": 1.3015, "step": 42073 }, { "epoch": 8.01, "learning_rate": 5.963393708293613e-07, "loss": 1.3484, "step": 42074 }, { "epoch": 8.01, "learning_rate": 5.962821734985701e-07, "loss": 0.6327, "step": 42075 }, { "epoch": 8.01, "learning_rate": 5.962249761677789e-07, "loss": 1.9789, "step": 42076 }, { "epoch": 8.01, "learning_rate": 5.961677788369876e-07, "loss": 1.5488, "step": 42077 }, { "epoch": 8.01, "learning_rate": 5.961105815061964e-07, "loss": 1.4898, "step": 42078 }, { "epoch": 8.02, "learning_rate": 5.960533841754052e-07, "loss": 1.9618, "step": 42079 }, { "epoch": 8.02, "learning_rate": 5.95996186844614e-07, "loss": 2.2087, "step": 42080 }, { "epoch": 8.02, "learning_rate": 5.959389895138227e-07, "loss": 1.0539, "step": 42081 }, { "epoch": 8.02, "learning_rate": 5.958817921830315e-07, "loss": 1.4183, "step": 42082 }, { "epoch": 8.02, "learning_rate": 5.958245948522402e-07, "loss": 1.5381, "step": 42083 }, { "epoch": 8.02, "learning_rate": 5.95767397521449e-07, "loss": 0.4675, "step": 42084 }, { "epoch": 8.02, "learning_rate": 5.957102001906578e-07, "loss": 1.042, "step": 42085 }, { "epoch": 8.02, "learning_rate": 5.956530028598665e-07, "loss": 0.7628, "step": 42086 }, { "epoch": 8.02, "learning_rate": 5.955958055290754e-07, "loss": 1.0376, "step": 42087 }, { "epoch": 8.02, "learning_rate": 5.955386081982841e-07, "loss": 0.9072, "step": 42088 }, { "epoch": 8.02, "learning_rate": 5.954814108674928e-07, "loss": 2.6908, "step": 42089 }, { "epoch": 8.02, "learning_rate": 5.954242135367017e-07, "loss": 0.7596, "step": 42090 }, { "epoch": 8.02, "learning_rate": 5.953670162059104e-07, "loss": 1.4653, "step": 42091 }, { "epoch": 8.02, "learning_rate": 5.953098188751192e-07, "loss": 2.3951, "step": 42092 }, { "epoch": 8.02, "learning_rate": 5.95252621544328e-07, "loss": 0.6025, "step": 42093 }, { "epoch": 8.02, "learning_rate": 5.951954242135366e-07, "loss": 0.3418, "step": 42094 }, { "epoch": 8.02, "learning_rate": 5.951382268827455e-07, "loss": 0.7932, "step": 42095 }, { "epoch": 8.02, "learning_rate": 5.950810295519542e-07, "loss": 1.5908, "step": 42096 }, { "epoch": 8.02, "learning_rate": 5.95023832221163e-07, "loss": 1.5076, "step": 42097 }, { "epoch": 8.02, "learning_rate": 5.949666348903718e-07, "loss": 1.4082, "step": 42098 }, { "epoch": 8.02, "learning_rate": 5.949094375595806e-07, "loss": 0.3304, "step": 42099 }, { "epoch": 8.02, "learning_rate": 5.948522402287893e-07, "loss": 2.5503, "step": 42100 }, { "epoch": 8.02, "learning_rate": 5.947950428979981e-07, "loss": 1.3823, "step": 42101 }, { "epoch": 8.02, "learning_rate": 5.947378455672069e-07, "loss": 1.6882, "step": 42102 }, { "epoch": 8.02, "learning_rate": 5.946806482364157e-07, "loss": 1.8492, "step": 42103 }, { "epoch": 8.02, "learning_rate": 5.946234509056245e-07, "loss": 2.3922, "step": 42104 }, { "epoch": 8.02, "learning_rate": 5.945662535748332e-07, "loss": 1.6113, "step": 42105 }, { "epoch": 8.02, "learning_rate": 5.94509056244042e-07, "loss": 1.499, "step": 42106 }, { "epoch": 8.02, "learning_rate": 5.944518589132507e-07, "loss": 1.7106, "step": 42107 }, { "epoch": 8.02, "learning_rate": 5.943946615824594e-07, "loss": 1.6975, "step": 42108 }, { "epoch": 8.02, "learning_rate": 5.943374642516683e-07, "loss": 0.8028, "step": 42109 }, { "epoch": 8.02, "learning_rate": 5.94280266920877e-07, "loss": 0.827, "step": 42110 }, { "epoch": 8.02, "learning_rate": 5.942230695900858e-07, "loss": 2.3815, "step": 42111 }, { "epoch": 8.02, "learning_rate": 5.941658722592946e-07, "loss": 1.5174, "step": 42112 }, { "epoch": 8.02, "learning_rate": 5.941086749285033e-07, "loss": 1.5503, "step": 42113 }, { "epoch": 8.02, "learning_rate": 5.940514775977121e-07, "loss": 2.4878, "step": 42114 }, { "epoch": 8.02, "learning_rate": 5.939942802669209e-07, "loss": 2.1587, "step": 42115 }, { "epoch": 8.02, "learning_rate": 5.939370829361297e-07, "loss": 1.601, "step": 42116 }, { "epoch": 8.02, "learning_rate": 5.938798856053385e-07, "loss": 1.6554, "step": 42117 }, { "epoch": 8.02, "learning_rate": 5.938226882745472e-07, "loss": 2.0948, "step": 42118 }, { "epoch": 8.02, "learning_rate": 5.93765490943756e-07, "loss": 0.7448, "step": 42119 }, { "epoch": 8.02, "learning_rate": 5.937082936129648e-07, "loss": 0.3221, "step": 42120 }, { "epoch": 8.02, "learning_rate": 5.936510962821735e-07, "loss": 2.0849, "step": 42121 }, { "epoch": 8.02, "learning_rate": 5.935938989513822e-07, "loss": 1.4101, "step": 42122 }, { "epoch": 8.02, "learning_rate": 5.935367016205911e-07, "loss": 1.0706, "step": 42123 }, { "epoch": 8.02, "learning_rate": 5.934795042897998e-07, "loss": 0.898, "step": 42124 }, { "epoch": 8.02, "learning_rate": 5.934223069590086e-07, "loss": 2.0988, "step": 42125 }, { "epoch": 8.02, "learning_rate": 5.933651096282174e-07, "loss": 1.105, "step": 42126 }, { "epoch": 8.02, "learning_rate": 5.933079122974261e-07, "loss": 1.6588, "step": 42127 }, { "epoch": 8.02, "learning_rate": 5.932507149666349e-07, "loss": 0.924, "step": 42128 }, { "epoch": 8.02, "learning_rate": 5.931935176358437e-07, "loss": 1.6381, "step": 42129 }, { "epoch": 8.02, "learning_rate": 5.931363203050524e-07, "loss": 1.1289, "step": 42130 }, { "epoch": 8.02, "learning_rate": 5.930791229742613e-07, "loss": 1.5305, "step": 42131 }, { "epoch": 8.03, "learning_rate": 5.9302192564347e-07, "loss": 1.2, "step": 42132 }, { "epoch": 8.03, "learning_rate": 5.929647283126788e-07, "loss": 2.3381, "step": 42133 }, { "epoch": 8.03, "learning_rate": 5.929075309818875e-07, "loss": 1.0677, "step": 42134 }, { "epoch": 8.03, "learning_rate": 5.928503336510963e-07, "loss": 2.4276, "step": 42135 }, { "epoch": 8.03, "learning_rate": 5.92793136320305e-07, "loss": 0.5743, "step": 42136 }, { "epoch": 8.03, "learning_rate": 5.927359389895138e-07, "loss": 1.7958, "step": 42137 }, { "epoch": 8.03, "learning_rate": 5.926787416587226e-07, "loss": 0.472, "step": 42138 }, { "epoch": 8.03, "learning_rate": 5.926215443279314e-07, "loss": 2.8295, "step": 42139 }, { "epoch": 8.03, "learning_rate": 5.925643469971402e-07, "loss": 1.7751, "step": 42140 }, { "epoch": 8.03, "learning_rate": 5.925071496663489e-07, "loss": 2.003, "step": 42141 }, { "epoch": 8.03, "learning_rate": 5.924499523355577e-07, "loss": 0.9716, "step": 42142 }, { "epoch": 8.03, "learning_rate": 5.923927550047665e-07, "loss": 2.7346, "step": 42143 }, { "epoch": 8.03, "learning_rate": 5.923355576739752e-07, "loss": 1.248, "step": 42144 }, { "epoch": 8.03, "learning_rate": 5.922783603431841e-07, "loss": 1.2076, "step": 42145 }, { "epoch": 8.03, "learning_rate": 5.922211630123928e-07, "loss": 1.7381, "step": 42146 }, { "epoch": 8.03, "learning_rate": 5.921639656816015e-07, "loss": 2.0904, "step": 42147 }, { "epoch": 8.03, "learning_rate": 5.921067683508103e-07, "loss": 1.4512, "step": 42148 }, { "epoch": 8.03, "learning_rate": 5.92049571020019e-07, "loss": 1.2225, "step": 42149 }, { "epoch": 8.03, "learning_rate": 5.919923736892278e-07, "loss": 3.1624, "step": 42150 }, { "epoch": 8.03, "learning_rate": 5.919351763584366e-07, "loss": 1.2372, "step": 42151 }, { "epoch": 8.03, "learning_rate": 5.918779790276454e-07, "loss": 3.117, "step": 42152 }, { "epoch": 8.03, "learning_rate": 5.918207816968542e-07, "loss": 1.5421, "step": 42153 }, { "epoch": 8.03, "learning_rate": 5.917635843660629e-07, "loss": 2.2276, "step": 42154 }, { "epoch": 8.03, "learning_rate": 5.917063870352717e-07, "loss": 0.6272, "step": 42155 }, { "epoch": 8.03, "learning_rate": 5.916491897044805e-07, "loss": 4.3044, "step": 42156 }, { "epoch": 8.03, "learning_rate": 5.915919923736893e-07, "loss": 1.0351, "step": 42157 }, { "epoch": 8.03, "learning_rate": 5.91534795042898e-07, "loss": 2.0039, "step": 42158 }, { "epoch": 8.03, "learning_rate": 5.914775977121068e-07, "loss": 0.7613, "step": 42159 }, { "epoch": 8.03, "learning_rate": 5.914204003813156e-07, "loss": 0.8315, "step": 42160 }, { "epoch": 8.03, "learning_rate": 5.913632030505243e-07, "loss": 1.8832, "step": 42161 }, { "epoch": 8.03, "learning_rate": 5.913060057197331e-07, "loss": 1.5053, "step": 42162 }, { "epoch": 8.03, "learning_rate": 5.912488083889418e-07, "loss": 1.0323, "step": 42163 }, { "epoch": 8.03, "learning_rate": 5.911916110581506e-07, "loss": 1.5153, "step": 42164 }, { "epoch": 8.03, "learning_rate": 5.911344137273594e-07, "loss": 2.1968, "step": 42165 }, { "epoch": 8.03, "learning_rate": 5.910772163965681e-07, "loss": 2.7812, "step": 42166 }, { "epoch": 8.03, "learning_rate": 5.91020019065777e-07, "loss": 1.4942, "step": 42167 }, { "epoch": 8.03, "learning_rate": 5.909628217349857e-07, "loss": 1.6207, "step": 42168 }, { "epoch": 8.03, "learning_rate": 5.909056244041945e-07, "loss": 1.4197, "step": 42169 }, { "epoch": 8.03, "learning_rate": 5.908484270734033e-07, "loss": 1.301, "step": 42170 }, { "epoch": 8.03, "learning_rate": 5.90791229742612e-07, "loss": 1.6771, "step": 42171 }, { "epoch": 8.03, "learning_rate": 5.907340324118208e-07, "loss": 2.2128, "step": 42172 }, { "epoch": 8.03, "learning_rate": 5.906768350810296e-07, "loss": 0.9637, "step": 42173 }, { "epoch": 8.03, "learning_rate": 5.906196377502383e-07, "loss": 1.5047, "step": 42174 }, { "epoch": 8.03, "learning_rate": 5.905624404194471e-07, "loss": 2.3013, "step": 42175 }, { "epoch": 8.03, "learning_rate": 5.905052430886559e-07, "loss": 1.5396, "step": 42176 }, { "epoch": 8.03, "learning_rate": 5.904480457578646e-07, "loss": 1.7753, "step": 42177 }, { "epoch": 8.03, "learning_rate": 5.903908484270734e-07, "loss": 1.4922, "step": 42178 }, { "epoch": 8.03, "learning_rate": 5.903336510962822e-07, "loss": 1.3413, "step": 42179 }, { "epoch": 8.03, "learning_rate": 5.902764537654909e-07, "loss": 1.9597, "step": 42180 }, { "epoch": 8.03, "learning_rate": 5.902192564346998e-07, "loss": 1.5912, "step": 42181 }, { "epoch": 8.03, "learning_rate": 5.901620591039085e-07, "loss": 1.6215, "step": 42182 }, { "epoch": 8.03, "learning_rate": 5.901048617731173e-07, "loss": 0.7714, "step": 42183 }, { "epoch": 8.04, "learning_rate": 5.900476644423261e-07, "loss": 1.2408, "step": 42184 }, { "epoch": 8.04, "learning_rate": 5.899904671115348e-07, "loss": 0.8081, "step": 42185 }, { "epoch": 8.04, "learning_rate": 5.899332697807436e-07, "loss": 1.4235, "step": 42186 }, { "epoch": 8.04, "learning_rate": 5.898760724499523e-07, "loss": 1.6682, "step": 42187 }, { "epoch": 8.04, "learning_rate": 5.898188751191611e-07, "loss": 0.2516, "step": 42188 }, { "epoch": 8.04, "learning_rate": 5.897616777883699e-07, "loss": 0.7433, "step": 42189 }, { "epoch": 8.04, "learning_rate": 5.897044804575786e-07, "loss": 2.1329, "step": 42190 }, { "epoch": 8.04, "learning_rate": 5.896472831267874e-07, "loss": 1.4912, "step": 42191 }, { "epoch": 8.04, "learning_rate": 5.895900857959962e-07, "loss": 1.4111, "step": 42192 }, { "epoch": 8.04, "learning_rate": 5.89532888465205e-07, "loss": 0.2626, "step": 42193 }, { "epoch": 8.04, "learning_rate": 5.894756911344137e-07, "loss": 1.0978, "step": 42194 }, { "epoch": 8.04, "learning_rate": 5.894184938036225e-07, "loss": 1.8026, "step": 42195 }, { "epoch": 8.04, "learning_rate": 5.893612964728313e-07, "loss": 1.6706, "step": 42196 }, { "epoch": 8.04, "learning_rate": 5.893040991420401e-07, "loss": 1.4142, "step": 42197 }, { "epoch": 8.04, "learning_rate": 5.892469018112489e-07, "loss": 3.0117, "step": 42198 }, { "epoch": 8.04, "learning_rate": 5.891897044804576e-07, "loss": 1.9649, "step": 42199 }, { "epoch": 8.04, "learning_rate": 5.891325071496665e-07, "loss": 1.1296, "step": 42200 }, { "epoch": 8.04, "learning_rate": 5.890753098188751e-07, "loss": 1.5799, "step": 42201 }, { "epoch": 8.04, "learning_rate": 5.890181124880838e-07, "loss": 1.443, "step": 42202 }, { "epoch": 8.04, "learning_rate": 5.889609151572927e-07, "loss": 1.4265, "step": 42203 }, { "epoch": 8.04, "learning_rate": 5.889037178265014e-07, "loss": 2.344, "step": 42204 }, { "epoch": 8.04, "learning_rate": 5.888465204957102e-07, "loss": 2.9516, "step": 42205 }, { "epoch": 8.04, "learning_rate": 5.88789323164919e-07, "loss": 1.1987, "step": 42206 }, { "epoch": 8.04, "learning_rate": 5.887321258341277e-07, "loss": 0.5342, "step": 42207 }, { "epoch": 8.04, "learning_rate": 5.886749285033365e-07, "loss": 1.5182, "step": 42208 }, { "epoch": 8.04, "learning_rate": 5.886177311725453e-07, "loss": 2.377, "step": 42209 }, { "epoch": 8.04, "learning_rate": 5.885605338417541e-07, "loss": 1.311, "step": 42210 }, { "epoch": 8.04, "learning_rate": 5.885033365109629e-07, "loss": 1.0397, "step": 42211 }, { "epoch": 8.04, "learning_rate": 5.884461391801716e-07, "loss": 3.4203, "step": 42212 }, { "epoch": 8.04, "learning_rate": 5.883889418493804e-07, "loss": 0.4576, "step": 42213 }, { "epoch": 8.04, "learning_rate": 5.883317445185891e-07, "loss": 3.3428, "step": 42214 }, { "epoch": 8.04, "learning_rate": 5.882745471877979e-07, "loss": 1.7797, "step": 42215 }, { "epoch": 8.04, "learning_rate": 5.882173498570066e-07, "loss": 1.9145, "step": 42216 }, { "epoch": 8.04, "learning_rate": 5.881601525262155e-07, "loss": 2.1133, "step": 42217 }, { "epoch": 8.04, "learning_rate": 5.881029551954242e-07, "loss": 1.8283, "step": 42218 }, { "epoch": 8.04, "learning_rate": 5.88045757864633e-07, "loss": 1.1424, "step": 42219 }, { "epoch": 8.04, "learning_rate": 5.879885605338418e-07, "loss": 2.4819, "step": 42220 }, { "epoch": 8.04, "learning_rate": 5.879313632030505e-07, "loss": 1.2658, "step": 42221 }, { "epoch": 8.04, "learning_rate": 5.878741658722594e-07, "loss": 0.9944, "step": 42222 }, { "epoch": 8.04, "learning_rate": 5.878169685414681e-07, "loss": 2.0192, "step": 42223 }, { "epoch": 8.04, "learning_rate": 5.877597712106768e-07, "loss": 1.6084, "step": 42224 }, { "epoch": 8.04, "learning_rate": 5.877025738798857e-07, "loss": 0.8855, "step": 42225 }, { "epoch": 8.04, "learning_rate": 5.876453765490944e-07, "loss": 0.2997, "step": 42226 }, { "epoch": 8.04, "learning_rate": 5.875881792183031e-07, "loss": 2.4398, "step": 42227 }, { "epoch": 8.04, "learning_rate": 5.875309818875119e-07, "loss": 1.5981, "step": 42228 }, { "epoch": 8.04, "learning_rate": 5.874737845567207e-07, "loss": 0.8242, "step": 42229 }, { "epoch": 8.04, "learning_rate": 5.874165872259294e-07, "loss": 2.3031, "step": 42230 }, { "epoch": 8.04, "learning_rate": 5.873593898951382e-07, "loss": 1.4926, "step": 42231 }, { "epoch": 8.04, "learning_rate": 5.87302192564347e-07, "loss": 2.3276, "step": 42232 }, { "epoch": 8.04, "learning_rate": 5.872449952335558e-07, "loss": 0.542, "step": 42233 }, { "epoch": 8.04, "learning_rate": 5.871877979027646e-07, "loss": 1.7616, "step": 42234 }, { "epoch": 8.04, "learning_rate": 5.871306005719733e-07, "loss": 1.5588, "step": 42235 }, { "epoch": 8.04, "learning_rate": 5.870734032411821e-07, "loss": 1.081, "step": 42236 }, { "epoch": 8.05, "learning_rate": 5.870162059103909e-07, "loss": 0.9683, "step": 42237 }, { "epoch": 8.05, "learning_rate": 5.869590085795996e-07, "loss": 2.6763, "step": 42238 }, { "epoch": 8.05, "learning_rate": 5.869018112488085e-07, "loss": 3.2016, "step": 42239 }, { "epoch": 8.05, "learning_rate": 5.868446139180172e-07, "loss": 2.3908, "step": 42240 }, { "epoch": 8.05, "learning_rate": 5.867874165872259e-07, "loss": 1.4765, "step": 42241 }, { "epoch": 8.05, "learning_rate": 5.867302192564347e-07, "loss": 0.664, "step": 42242 }, { "epoch": 8.05, "learning_rate": 5.866730219256434e-07, "loss": 1.0948, "step": 42243 }, { "epoch": 8.05, "learning_rate": 5.866158245948523e-07, "loss": 1.2867, "step": 42244 }, { "epoch": 8.05, "learning_rate": 5.86558627264061e-07, "loss": 0.4921, "step": 42245 }, { "epoch": 8.05, "learning_rate": 5.865014299332698e-07, "loss": 1.7301, "step": 42246 }, { "epoch": 8.05, "learning_rate": 5.864442326024786e-07, "loss": 1.8993, "step": 42247 }, { "epoch": 8.05, "learning_rate": 5.863870352716873e-07, "loss": 1.2675, "step": 42248 }, { "epoch": 8.05, "learning_rate": 5.863298379408961e-07, "loss": 1.1552, "step": 42249 }, { "epoch": 8.05, "learning_rate": 5.862726406101049e-07, "loss": 1.8095, "step": 42250 }, { "epoch": 8.05, "learning_rate": 5.862154432793137e-07, "loss": 2.0054, "step": 42251 }, { "epoch": 8.05, "learning_rate": 5.861582459485224e-07, "loss": 0.2487, "step": 42252 }, { "epoch": 8.05, "learning_rate": 5.861010486177313e-07, "loss": 1.9976, "step": 42253 }, { "epoch": 8.05, "learning_rate": 5.860438512869399e-07, "loss": 2.1749, "step": 42254 }, { "epoch": 8.05, "learning_rate": 5.859866539561487e-07, "loss": 0.4951, "step": 42255 }, { "epoch": 8.05, "learning_rate": 5.859294566253575e-07, "loss": 1.875, "step": 42256 }, { "epoch": 8.05, "learning_rate": 5.858722592945662e-07, "loss": 2.4053, "step": 42257 }, { "epoch": 8.05, "learning_rate": 5.858150619637751e-07, "loss": 0.9886, "step": 42258 }, { "epoch": 8.05, "learning_rate": 5.857578646329838e-07, "loss": 1.784, "step": 42259 }, { "epoch": 8.05, "learning_rate": 5.857006673021925e-07, "loss": 1.6818, "step": 42260 }, { "epoch": 8.05, "learning_rate": 5.856434699714014e-07, "loss": 2.8296, "step": 42261 }, { "epoch": 8.05, "learning_rate": 5.855862726406101e-07, "loss": 2.08, "step": 42262 }, { "epoch": 8.05, "learning_rate": 5.855290753098189e-07, "loss": 1.8636, "step": 42263 }, { "epoch": 8.05, "learning_rate": 5.854718779790277e-07, "loss": 3.0759, "step": 42264 }, { "epoch": 8.05, "learning_rate": 5.854146806482364e-07, "loss": 3.0556, "step": 42265 }, { "epoch": 8.05, "learning_rate": 5.853574833174452e-07, "loss": 1.104, "step": 42266 }, { "epoch": 8.05, "learning_rate": 5.85300285986654e-07, "loss": 0.8034, "step": 42267 }, { "epoch": 8.05, "learning_rate": 5.852430886558627e-07, "loss": 1.4235, "step": 42268 }, { "epoch": 8.05, "learning_rate": 5.851858913250715e-07, "loss": 1.5422, "step": 42269 }, { "epoch": 8.05, "learning_rate": 5.851286939942803e-07, "loss": 2.1178, "step": 42270 }, { "epoch": 8.05, "learning_rate": 5.85071496663489e-07, "loss": 0.776, "step": 42271 }, { "epoch": 8.05, "learning_rate": 5.850142993326978e-07, "loss": 1.3955, "step": 42272 }, { "epoch": 8.05, "learning_rate": 5.849571020019066e-07, "loss": 2.9029, "step": 42273 }, { "epoch": 8.05, "learning_rate": 5.848999046711153e-07, "loss": 0.343, "step": 42274 }, { "epoch": 8.05, "learning_rate": 5.848427073403242e-07, "loss": 1.0028, "step": 42275 }, { "epoch": 8.05, "learning_rate": 5.847855100095329e-07, "loss": 3.6762, "step": 42276 }, { "epoch": 8.05, "learning_rate": 5.847283126787417e-07, "loss": 1.5985, "step": 42277 }, { "epoch": 8.05, "learning_rate": 5.846711153479505e-07, "loss": 1.3418, "step": 42278 }, { "epoch": 8.05, "learning_rate": 5.846139180171592e-07, "loss": 0.9476, "step": 42279 }, { "epoch": 8.05, "learning_rate": 5.845567206863681e-07, "loss": 1.5882, "step": 42280 }, { "epoch": 8.05, "learning_rate": 5.844995233555767e-07, "loss": 1.3993, "step": 42281 }, { "epoch": 8.05, "learning_rate": 5.844423260247855e-07, "loss": 1.9127, "step": 42282 }, { "epoch": 8.05, "learning_rate": 5.843851286939943e-07, "loss": 3.7033, "step": 42283 }, { "epoch": 8.05, "learning_rate": 5.84327931363203e-07, "loss": 2.2089, "step": 42284 }, { "epoch": 8.05, "learning_rate": 5.842707340324118e-07, "loss": 1.0765, "step": 42285 }, { "epoch": 8.05, "learning_rate": 5.842135367016206e-07, "loss": 2.509, "step": 42286 }, { "epoch": 8.05, "learning_rate": 5.841563393708294e-07, "loss": 1.5649, "step": 42287 }, { "epoch": 8.05, "learning_rate": 5.840991420400381e-07, "loss": 1.8681, "step": 42288 }, { "epoch": 8.06, "learning_rate": 5.840419447092469e-07, "loss": 1.8484, "step": 42289 }, { "epoch": 8.06, "learning_rate": 5.839847473784557e-07, "loss": 3.13, "step": 42290 }, { "epoch": 8.06, "learning_rate": 5.839275500476645e-07, "loss": 0.5632, "step": 42291 }, { "epoch": 8.06, "learning_rate": 5.838703527168733e-07, "loss": 1.9735, "step": 42292 }, { "epoch": 8.06, "learning_rate": 5.83813155386082e-07, "loss": 1.9759, "step": 42293 }, { "epoch": 8.06, "learning_rate": 5.837559580552909e-07, "loss": 2.0025, "step": 42294 }, { "epoch": 8.06, "learning_rate": 5.836987607244995e-07, "loss": 1.9379, "step": 42295 }, { "epoch": 8.06, "learning_rate": 5.836415633937082e-07, "loss": 1.6357, "step": 42296 }, { "epoch": 8.06, "learning_rate": 5.835843660629171e-07, "loss": 0.5033, "step": 42297 }, { "epoch": 8.06, "learning_rate": 5.835271687321258e-07, "loss": 1.2416, "step": 42298 }, { "epoch": 8.06, "learning_rate": 5.834699714013346e-07, "loss": 1.3638, "step": 42299 }, { "epoch": 8.06, "learning_rate": 5.834127740705434e-07, "loss": 0.8486, "step": 42300 }, { "epoch": 8.06, "learning_rate": 5.833555767397521e-07, "loss": 2.2485, "step": 42301 }, { "epoch": 8.06, "learning_rate": 5.83298379408961e-07, "loss": 3.6554, "step": 42302 }, { "epoch": 8.06, "learning_rate": 5.832411820781697e-07, "loss": 1.7026, "step": 42303 }, { "epoch": 8.06, "learning_rate": 5.831839847473785e-07, "loss": 0.6662, "step": 42304 }, { "epoch": 8.06, "learning_rate": 5.831267874165873e-07, "loss": 2.5505, "step": 42305 }, { "epoch": 8.06, "learning_rate": 5.830695900857961e-07, "loss": 2.3777, "step": 42306 }, { "epoch": 8.06, "learning_rate": 5.830123927550047e-07, "loss": 1.1437, "step": 42307 }, { "epoch": 8.06, "learning_rate": 5.829551954242135e-07, "loss": 1.2078, "step": 42308 }, { "epoch": 8.06, "learning_rate": 5.828979980934223e-07, "loss": 1.7753, "step": 42309 }, { "epoch": 8.06, "learning_rate": 5.82840800762631e-07, "loss": 1.9897, "step": 42310 }, { "epoch": 8.06, "learning_rate": 5.827836034318399e-07, "loss": 2.301, "step": 42311 }, { "epoch": 8.06, "learning_rate": 5.827264061010486e-07, "loss": 2.5795, "step": 42312 }, { "epoch": 8.06, "learning_rate": 5.826692087702574e-07, "loss": 1.4062, "step": 42313 }, { "epoch": 8.06, "learning_rate": 5.826120114394662e-07, "loss": 1.4978, "step": 42314 }, { "epoch": 8.06, "learning_rate": 5.825548141086749e-07, "loss": 2.2757, "step": 42315 }, { "epoch": 8.06, "learning_rate": 5.824976167778838e-07, "loss": 1.1031, "step": 42316 }, { "epoch": 8.06, "learning_rate": 5.824404194470925e-07, "loss": 1.5378, "step": 42317 }, { "epoch": 8.06, "learning_rate": 5.823832221163013e-07, "loss": 1.2578, "step": 42318 }, { "epoch": 8.06, "learning_rate": 5.823260247855101e-07, "loss": 2.18, "step": 42319 }, { "epoch": 8.06, "learning_rate": 5.822688274547188e-07, "loss": 1.7728, "step": 42320 }, { "epoch": 8.06, "learning_rate": 5.822116301239275e-07, "loss": 0.8427, "step": 42321 }, { "epoch": 8.06, "learning_rate": 5.821544327931363e-07, "loss": 1.7988, "step": 42322 }, { "epoch": 8.06, "learning_rate": 5.820972354623451e-07, "loss": 0.8148, "step": 42323 }, { "epoch": 8.06, "learning_rate": 5.820400381315539e-07, "loss": 2.364, "step": 42324 }, { "epoch": 8.06, "learning_rate": 5.819828408007626e-07, "loss": 1.5436, "step": 42325 }, { "epoch": 8.06, "learning_rate": 5.819256434699714e-07, "loss": 1.6678, "step": 42326 }, { "epoch": 8.06, "learning_rate": 5.818684461391802e-07, "loss": 2.1619, "step": 42327 }, { "epoch": 8.06, "learning_rate": 5.81811248808389e-07, "loss": 0.497, "step": 42328 }, { "epoch": 8.06, "learning_rate": 5.817540514775977e-07, "loss": 0.4413, "step": 42329 }, { "epoch": 8.06, "learning_rate": 5.816968541468065e-07, "loss": 0.36, "step": 42330 }, { "epoch": 8.06, "learning_rate": 5.816396568160153e-07, "loss": 1.9723, "step": 42331 }, { "epoch": 8.06, "learning_rate": 5.81582459485224e-07, "loss": 1.254, "step": 42332 }, { "epoch": 8.06, "learning_rate": 5.815252621544329e-07, "loss": 1.5915, "step": 42333 }, { "epoch": 8.06, "learning_rate": 5.814680648236415e-07, "loss": 2.1958, "step": 42334 }, { "epoch": 8.06, "learning_rate": 5.814108674928503e-07, "loss": 1.2787, "step": 42335 }, { "epoch": 8.06, "learning_rate": 5.813536701620591e-07, "loss": 1.6741, "step": 42336 }, { "epoch": 8.06, "learning_rate": 5.812964728312678e-07, "loss": 0.7592, "step": 42337 }, { "epoch": 8.06, "learning_rate": 5.812392755004767e-07, "loss": 1.8129, "step": 42338 }, { "epoch": 8.06, "learning_rate": 5.811820781696854e-07, "loss": 1.23, "step": 42339 }, { "epoch": 8.06, "learning_rate": 5.811248808388942e-07, "loss": 3.176, "step": 42340 }, { "epoch": 8.06, "learning_rate": 5.81067683508103e-07, "loss": 1.0075, "step": 42341 }, { "epoch": 8.07, "learning_rate": 5.810104861773117e-07, "loss": 2.0383, "step": 42342 }, { "epoch": 8.07, "learning_rate": 5.809532888465205e-07, "loss": 2.3994, "step": 42343 }, { "epoch": 8.07, "learning_rate": 5.808960915157293e-07, "loss": 3.3034, "step": 42344 }, { "epoch": 8.07, "learning_rate": 5.808388941849381e-07, "loss": 1.1733, "step": 42345 }, { "epoch": 8.07, "learning_rate": 5.807816968541469e-07, "loss": 0.6113, "step": 42346 }, { "epoch": 8.07, "learning_rate": 5.807244995233557e-07, "loss": 2.4218, "step": 42347 }, { "epoch": 8.07, "learning_rate": 5.806673021925643e-07, "loss": 2.1138, "step": 42348 }, { "epoch": 8.07, "learning_rate": 5.806101048617731e-07, "loss": 1.4487, "step": 42349 }, { "epoch": 8.07, "learning_rate": 5.805529075309819e-07, "loss": 1.4763, "step": 42350 }, { "epoch": 8.07, "learning_rate": 5.804957102001906e-07, "loss": 1.0713, "step": 42351 }, { "epoch": 8.07, "learning_rate": 5.804385128693995e-07, "loss": 1.2889, "step": 42352 }, { "epoch": 8.07, "learning_rate": 5.803813155386082e-07, "loss": 1.3983, "step": 42353 }, { "epoch": 8.07, "learning_rate": 5.803241182078169e-07, "loss": 0.503, "step": 42354 }, { "epoch": 8.07, "learning_rate": 5.802669208770258e-07, "loss": 1.7976, "step": 42355 }, { "epoch": 8.07, "learning_rate": 5.802097235462345e-07, "loss": 0.6879, "step": 42356 }, { "epoch": 8.07, "learning_rate": 5.801525262154433e-07, "loss": 0.6249, "step": 42357 }, { "epoch": 8.07, "learning_rate": 5.800953288846521e-07, "loss": 0.1895, "step": 42358 }, { "epoch": 8.07, "learning_rate": 5.800381315538609e-07, "loss": 1.2904, "step": 42359 }, { "epoch": 8.07, "learning_rate": 5.799809342230697e-07, "loss": 1.1768, "step": 42360 }, { "epoch": 8.07, "learning_rate": 5.799237368922783e-07, "loss": 0.8993, "step": 42361 }, { "epoch": 8.07, "learning_rate": 5.798665395614871e-07, "loss": 1.9421, "step": 42362 }, { "epoch": 8.07, "learning_rate": 5.798093422306959e-07, "loss": 1.4281, "step": 42363 }, { "epoch": 8.07, "learning_rate": 5.797521448999047e-07, "loss": 1.585, "step": 42364 }, { "epoch": 8.07, "learning_rate": 5.796949475691134e-07, "loss": 1.3138, "step": 42365 }, { "epoch": 8.07, "learning_rate": 5.796377502383222e-07, "loss": 0.8386, "step": 42366 }, { "epoch": 8.07, "learning_rate": 5.79580552907531e-07, "loss": 2.3393, "step": 42367 }, { "epoch": 8.07, "learning_rate": 5.795233555767398e-07, "loss": 1.3568, "step": 42368 }, { "epoch": 8.07, "learning_rate": 5.794661582459486e-07, "loss": 1.7824, "step": 42369 }, { "epoch": 8.07, "learning_rate": 5.794089609151573e-07, "loss": 2.1108, "step": 42370 }, { "epoch": 8.07, "learning_rate": 5.793517635843661e-07, "loss": 1.6843, "step": 42371 }, { "epoch": 8.07, "learning_rate": 5.792945662535749e-07, "loss": 1.1334, "step": 42372 }, { "epoch": 8.07, "learning_rate": 5.792373689227836e-07, "loss": 2.8843, "step": 42373 }, { "epoch": 8.07, "learning_rate": 5.791801715919925e-07, "loss": 1.5662, "step": 42374 }, { "epoch": 8.07, "learning_rate": 5.791229742612011e-07, "loss": 1.592, "step": 42375 }, { "epoch": 8.07, "learning_rate": 5.790657769304099e-07, "loss": 0.8221, "step": 42376 }, { "epoch": 8.07, "learning_rate": 5.790085795996187e-07, "loss": 2.404, "step": 42377 }, { "epoch": 8.07, "learning_rate": 5.789513822688274e-07, "loss": 2.1238, "step": 42378 }, { "epoch": 8.07, "learning_rate": 5.788941849380362e-07, "loss": 2.1619, "step": 42379 }, { "epoch": 8.07, "learning_rate": 5.78836987607245e-07, "loss": 1.4872, "step": 42380 }, { "epoch": 8.07, "learning_rate": 5.787797902764538e-07, "loss": 1.5701, "step": 42381 }, { "epoch": 8.07, "learning_rate": 5.787225929456626e-07, "loss": 1.4397, "step": 42382 }, { "epoch": 8.07, "learning_rate": 5.786653956148714e-07, "loss": 1.0817, "step": 42383 }, { "epoch": 8.07, "learning_rate": 5.786081982840801e-07, "loss": 0.9814, "step": 42384 }, { "epoch": 8.07, "learning_rate": 5.785510009532889e-07, "loss": 1.402, "step": 42385 }, { "epoch": 8.07, "learning_rate": 5.784938036224977e-07, "loss": 1.6132, "step": 42386 }, { "epoch": 8.07, "learning_rate": 5.784366062917063e-07, "loss": 1.3474, "step": 42387 }, { "epoch": 8.07, "learning_rate": 5.783794089609152e-07, "loss": 2.0368, "step": 42388 }, { "epoch": 8.07, "learning_rate": 5.783222116301239e-07, "loss": 3.6729, "step": 42389 }, { "epoch": 8.07, "learning_rate": 5.782650142993327e-07, "loss": 0.8466, "step": 42390 }, { "epoch": 8.07, "learning_rate": 5.782078169685415e-07, "loss": 0.8855, "step": 42391 }, { "epoch": 8.07, "learning_rate": 5.781506196377502e-07, "loss": 1.3491, "step": 42392 }, { "epoch": 8.07, "learning_rate": 5.78093422306959e-07, "loss": 1.1615, "step": 42393 }, { "epoch": 8.08, "learning_rate": 5.780362249761678e-07, "loss": 2.1266, "step": 42394 }, { "epoch": 8.08, "learning_rate": 5.779790276453765e-07, "loss": 2.011, "step": 42395 }, { "epoch": 8.08, "learning_rate": 5.779218303145854e-07, "loss": 2.0029, "step": 42396 }, { "epoch": 8.08, "learning_rate": 5.778646329837941e-07, "loss": 1.7742, "step": 42397 }, { "epoch": 8.08, "learning_rate": 5.778074356530029e-07, "loss": 1.7387, "step": 42398 }, { "epoch": 8.08, "learning_rate": 5.777502383222117e-07, "loss": 1.3124, "step": 42399 }, { "epoch": 8.08, "learning_rate": 5.776930409914205e-07, "loss": 0.472, "step": 42400 }, { "epoch": 8.08, "learning_rate": 5.776358436606291e-07, "loss": 1.4271, "step": 42401 }, { "epoch": 8.08, "learning_rate": 5.775786463298379e-07, "loss": 1.7995, "step": 42402 }, { "epoch": 8.08, "learning_rate": 5.775214489990467e-07, "loss": 0.9238, "step": 42403 }, { "epoch": 8.08, "learning_rate": 5.774642516682555e-07, "loss": 1.1802, "step": 42404 }, { "epoch": 8.08, "learning_rate": 5.774070543374643e-07, "loss": 0.538, "step": 42405 }, { "epoch": 8.08, "learning_rate": 5.77349857006673e-07, "loss": 2.277, "step": 42406 }, { "epoch": 8.08, "learning_rate": 5.772926596758818e-07, "loss": 1.7754, "step": 42407 }, { "epoch": 8.08, "learning_rate": 5.772354623450906e-07, "loss": 0.8713, "step": 42408 }, { "epoch": 8.08, "learning_rate": 5.771782650142993e-07, "loss": 2.2951, "step": 42409 }, { "epoch": 8.08, "learning_rate": 5.771210676835082e-07, "loss": 0.7533, "step": 42410 }, { "epoch": 8.08, "learning_rate": 5.770638703527169e-07, "loss": 1.8291, "step": 42411 }, { "epoch": 8.08, "learning_rate": 5.770066730219257e-07, "loss": 1.5465, "step": 42412 }, { "epoch": 8.08, "learning_rate": 5.769494756911345e-07, "loss": 0.9745, "step": 42413 }, { "epoch": 8.08, "learning_rate": 5.768922783603431e-07, "loss": 1.4087, "step": 42414 }, { "epoch": 8.08, "learning_rate": 5.768350810295519e-07, "loss": 2.3459, "step": 42415 }, { "epoch": 8.08, "learning_rate": 5.767778836987607e-07, "loss": 0.4928, "step": 42416 }, { "epoch": 8.08, "learning_rate": 5.767206863679695e-07, "loss": 2.0878, "step": 42417 }, { "epoch": 8.08, "learning_rate": 5.766634890371783e-07, "loss": 1.7828, "step": 42418 }, { "epoch": 8.08, "learning_rate": 5.76606291706387e-07, "loss": 2.6188, "step": 42419 }, { "epoch": 8.08, "learning_rate": 5.765490943755958e-07, "loss": 1.2323, "step": 42420 }, { "epoch": 8.08, "learning_rate": 5.764918970448046e-07, "loss": 0.7028, "step": 42421 }, { "epoch": 8.08, "learning_rate": 5.764346997140134e-07, "loss": 2.9888, "step": 42422 }, { "epoch": 8.08, "learning_rate": 5.763775023832221e-07, "loss": 1.8762, "step": 42423 }, { "epoch": 8.08, "learning_rate": 5.76320305052431e-07, "loss": 1.0944, "step": 42424 }, { "epoch": 8.08, "learning_rate": 5.762631077216397e-07, "loss": 2.3417, "step": 42425 }, { "epoch": 8.08, "learning_rate": 5.762059103908485e-07, "loss": 0.7448, "step": 42426 }, { "epoch": 8.08, "learning_rate": 5.761487130600573e-07, "loss": 1.8531, "step": 42427 }, { "epoch": 8.08, "learning_rate": 5.760915157292659e-07, "loss": 1.6176, "step": 42428 }, { "epoch": 8.08, "learning_rate": 5.760343183984747e-07, "loss": 1.595, "step": 42429 }, { "epoch": 8.08, "learning_rate": 5.759771210676835e-07, "loss": 0.6411, "step": 42430 }, { "epoch": 8.08, "learning_rate": 5.759199237368922e-07, "loss": 0.5694, "step": 42431 }, { "epoch": 8.08, "learning_rate": 5.758627264061011e-07, "loss": 1.1579, "step": 42432 }, { "epoch": 8.08, "learning_rate": 5.758055290753098e-07, "loss": 2.3912, "step": 42433 }, { "epoch": 8.08, "learning_rate": 5.757483317445186e-07, "loss": 0.839, "step": 42434 }, { "epoch": 8.08, "learning_rate": 5.756911344137274e-07, "loss": 1.8478, "step": 42435 }, { "epoch": 8.08, "learning_rate": 5.756339370829362e-07, "loss": 2.6717, "step": 42436 }, { "epoch": 8.08, "learning_rate": 5.755767397521449e-07, "loss": 1.8168, "step": 42437 }, { "epoch": 8.08, "learning_rate": 5.755195424213537e-07, "loss": 0.9928, "step": 42438 }, { "epoch": 8.08, "learning_rate": 5.754623450905625e-07, "loss": 1.0499, "step": 42439 }, { "epoch": 8.08, "learning_rate": 5.754051477597713e-07, "loss": 3.1198, "step": 42440 }, { "epoch": 8.08, "learning_rate": 5.7534795042898e-07, "loss": 1.4098, "step": 42441 }, { "epoch": 8.08, "learning_rate": 5.752907530981887e-07, "loss": 1.5836, "step": 42442 }, { "epoch": 8.08, "learning_rate": 5.752335557673975e-07, "loss": 1.7425, "step": 42443 }, { "epoch": 8.08, "learning_rate": 5.751763584366063e-07, "loss": 2.2824, "step": 42444 }, { "epoch": 8.08, "learning_rate": 5.75119161105815e-07, "loss": 0.3008, "step": 42445 }, { "epoch": 8.08, "learning_rate": 5.750619637750239e-07, "loss": 3.2042, "step": 42446 }, { "epoch": 8.09, "learning_rate": 5.750047664442326e-07, "loss": 2.1807, "step": 42447 }, { "epoch": 8.09, "learning_rate": 5.749475691134414e-07, "loss": 1.6071, "step": 42448 }, { "epoch": 8.09, "learning_rate": 5.748903717826502e-07, "loss": 1.053, "step": 42449 }, { "epoch": 8.09, "learning_rate": 5.748331744518589e-07, "loss": 0.3593, "step": 42450 }, { "epoch": 8.09, "learning_rate": 5.747759771210677e-07, "loss": 1.6358, "step": 42451 }, { "epoch": 8.09, "learning_rate": 5.747187797902765e-07, "loss": 0.4306, "step": 42452 }, { "epoch": 8.09, "learning_rate": 5.746615824594853e-07, "loss": 0.4893, "step": 42453 }, { "epoch": 8.09, "learning_rate": 5.74604385128694e-07, "loss": 1.8957, "step": 42454 }, { "epoch": 8.09, "learning_rate": 5.745471877979027e-07, "loss": 0.6396, "step": 42455 }, { "epoch": 8.09, "learning_rate": 5.744899904671115e-07, "loss": 3.0562, "step": 42456 }, { "epoch": 8.09, "learning_rate": 5.744327931363203e-07, "loss": 1.6422, "step": 42457 }, { "epoch": 8.09, "learning_rate": 5.743755958055291e-07, "loss": 0.55, "step": 42458 }, { "epoch": 8.09, "learning_rate": 5.743183984747378e-07, "loss": 1.8752, "step": 42459 }, { "epoch": 8.09, "learning_rate": 5.742612011439466e-07, "loss": 1.5846, "step": 42460 }, { "epoch": 8.09, "learning_rate": 5.742040038131554e-07, "loss": 1.0857, "step": 42461 }, { "epoch": 8.09, "learning_rate": 5.741468064823642e-07, "loss": 2.3234, "step": 42462 }, { "epoch": 8.09, "learning_rate": 5.74089609151573e-07, "loss": 1.1232, "step": 42463 }, { "epoch": 8.09, "learning_rate": 5.740324118207817e-07, "loss": 1.5844, "step": 42464 }, { "epoch": 8.09, "learning_rate": 5.739752144899905e-07, "loss": 0.8811, "step": 42465 }, { "epoch": 8.09, "learning_rate": 5.739180171591993e-07, "loss": 2.1483, "step": 42466 }, { "epoch": 8.09, "learning_rate": 5.73860819828408e-07, "loss": 1.2661, "step": 42467 }, { "epoch": 8.09, "learning_rate": 5.738036224976168e-07, "loss": 1.542, "step": 42468 }, { "epoch": 8.09, "learning_rate": 5.737464251668255e-07, "loss": 2.3376, "step": 42469 }, { "epoch": 8.09, "learning_rate": 5.736892278360343e-07, "loss": 2.4984, "step": 42470 }, { "epoch": 8.09, "learning_rate": 5.736320305052431e-07, "loss": 1.6313, "step": 42471 }, { "epoch": 8.09, "learning_rate": 5.735748331744518e-07, "loss": 1.6721, "step": 42472 }, { "epoch": 8.09, "learning_rate": 5.735176358436606e-07, "loss": 1.3736, "step": 42473 }, { "epoch": 8.09, "learning_rate": 5.734604385128694e-07, "loss": 1.8489, "step": 42474 }, { "epoch": 8.09, "learning_rate": 5.734032411820782e-07, "loss": 1.322, "step": 42475 }, { "epoch": 8.09, "learning_rate": 5.73346043851287e-07, "loss": 1.206, "step": 42476 }, { "epoch": 8.09, "learning_rate": 5.732888465204958e-07, "loss": 1.98, "step": 42477 }, { "epoch": 8.09, "learning_rate": 5.732316491897045e-07, "loss": 1.2802, "step": 42478 }, { "epoch": 8.09, "learning_rate": 5.731744518589133e-07, "loss": 0.9935, "step": 42479 }, { "epoch": 8.09, "learning_rate": 5.731172545281221e-07, "loss": 1.9155, "step": 42480 }, { "epoch": 8.09, "learning_rate": 5.730600571973307e-07, "loss": 1.4423, "step": 42481 }, { "epoch": 8.09, "learning_rate": 5.730028598665396e-07, "loss": 1.4416, "step": 42482 }, { "epoch": 8.09, "learning_rate": 5.729456625357483e-07, "loss": 1.4141, "step": 42483 }, { "epoch": 8.09, "learning_rate": 5.728884652049571e-07, "loss": 1.0079, "step": 42484 }, { "epoch": 8.09, "learning_rate": 5.728312678741659e-07, "loss": 0.7066, "step": 42485 }, { "epoch": 8.09, "learning_rate": 5.727740705433746e-07, "loss": 1.651, "step": 42486 }, { "epoch": 8.09, "learning_rate": 5.727168732125834e-07, "loss": 1.0403, "step": 42487 }, { "epoch": 8.09, "learning_rate": 5.726596758817922e-07, "loss": 0.8167, "step": 42488 }, { "epoch": 8.09, "learning_rate": 5.72602478551001e-07, "loss": 0.8435, "step": 42489 }, { "epoch": 8.09, "learning_rate": 5.725452812202098e-07, "loss": 1.7193, "step": 42490 }, { "epoch": 8.09, "learning_rate": 5.724880838894185e-07, "loss": 1.8341, "step": 42491 }, { "epoch": 8.09, "learning_rate": 5.724308865586273e-07, "loss": 1.6684, "step": 42492 }, { "epoch": 8.09, "learning_rate": 5.723736892278361e-07, "loss": 1.1797, "step": 42493 }, { "epoch": 8.09, "learning_rate": 5.723164918970449e-07, "loss": 1.3854, "step": 42494 }, { "epoch": 8.09, "learning_rate": 5.722592945662535e-07, "loss": 2.1087, "step": 42495 }, { "epoch": 8.09, "learning_rate": 5.722020972354623e-07, "loss": 2.5549, "step": 42496 }, { "epoch": 8.09, "learning_rate": 5.721448999046711e-07, "loss": 1.6165, "step": 42497 }, { "epoch": 8.09, "learning_rate": 5.720877025738799e-07, "loss": 1.6193, "step": 42498 }, { "epoch": 8.1, "learning_rate": 5.720305052430887e-07, "loss": 1.7024, "step": 42499 }, { "epoch": 8.1, "learning_rate": 5.719733079122974e-07, "loss": 2.6313, "step": 42500 }, { "epoch": 8.1, "learning_rate": 5.719161105815063e-07, "loss": 1.1229, "step": 42501 }, { "epoch": 8.1, "learning_rate": 5.71858913250715e-07, "loss": 1.2769, "step": 42502 }, { "epoch": 8.1, "learning_rate": 5.718017159199237e-07, "loss": 3.1028, "step": 42503 }, { "epoch": 8.1, "learning_rate": 5.717445185891326e-07, "loss": 2.9942, "step": 42504 }, { "epoch": 8.1, "learning_rate": 5.716873212583413e-07, "loss": 2.0327, "step": 42505 }, { "epoch": 8.1, "learning_rate": 5.716301239275501e-07, "loss": 1.2572, "step": 42506 }, { "epoch": 8.1, "learning_rate": 5.715729265967589e-07, "loss": 1.3787, "step": 42507 }, { "epoch": 8.1, "learning_rate": 5.715157292659675e-07, "loss": 2.8551, "step": 42508 }, { "epoch": 8.1, "learning_rate": 5.714585319351763e-07, "loss": 1.4758, "step": 42509 }, { "epoch": 8.1, "learning_rate": 5.714013346043851e-07, "loss": 1.9634, "step": 42510 }, { "epoch": 8.1, "learning_rate": 5.713441372735939e-07, "loss": 1.4479, "step": 42511 }, { "epoch": 8.1, "learning_rate": 5.712869399428027e-07, "loss": 1.1087, "step": 42512 }, { "epoch": 8.1, "learning_rate": 5.712297426120114e-07, "loss": 1.9432, "step": 42513 }, { "epoch": 8.1, "learning_rate": 5.711725452812202e-07, "loss": 1.6573, "step": 42514 }, { "epoch": 8.1, "learning_rate": 5.71115347950429e-07, "loss": 1.2984, "step": 42515 }, { "epoch": 8.1, "learning_rate": 5.710581506196378e-07, "loss": 1.3675, "step": 42516 }, { "epoch": 8.1, "learning_rate": 5.710009532888465e-07, "loss": 2.6151, "step": 42517 }, { "epoch": 8.1, "learning_rate": 5.709437559580554e-07, "loss": 0.888, "step": 42518 }, { "epoch": 8.1, "learning_rate": 5.708865586272641e-07, "loss": 1.852, "step": 42519 }, { "epoch": 8.1, "learning_rate": 5.708293612964729e-07, "loss": 1.198, "step": 42520 }, { "epoch": 8.1, "learning_rate": 5.707721639656816e-07, "loss": 3.0655, "step": 42521 }, { "epoch": 8.1, "learning_rate": 5.707149666348903e-07, "loss": 1.0114, "step": 42522 }, { "epoch": 8.1, "learning_rate": 5.706577693040992e-07, "loss": 2.0857, "step": 42523 }, { "epoch": 8.1, "learning_rate": 5.706005719733079e-07, "loss": 0.4786, "step": 42524 }, { "epoch": 8.1, "learning_rate": 5.705433746425166e-07, "loss": 0.5398, "step": 42525 }, { "epoch": 8.1, "learning_rate": 5.704861773117255e-07, "loss": 1.7961, "step": 42526 }, { "epoch": 8.1, "learning_rate": 5.704289799809342e-07, "loss": 2.3427, "step": 42527 }, { "epoch": 8.1, "learning_rate": 5.70371782650143e-07, "loss": 0.8184, "step": 42528 }, { "epoch": 8.1, "learning_rate": 5.703145853193518e-07, "loss": 1.0839, "step": 42529 }, { "epoch": 8.1, "learning_rate": 5.702573879885606e-07, "loss": 1.6582, "step": 42530 }, { "epoch": 8.1, "learning_rate": 5.702001906577693e-07, "loss": 1.2009, "step": 42531 }, { "epoch": 8.1, "learning_rate": 5.701429933269781e-07, "loss": 1.7258, "step": 42532 }, { "epoch": 8.1, "learning_rate": 5.700857959961869e-07, "loss": 2.027, "step": 42533 }, { "epoch": 8.1, "learning_rate": 5.700285986653957e-07, "loss": 1.7543, "step": 42534 }, { "epoch": 8.1, "learning_rate": 5.699714013346044e-07, "loss": 1.685, "step": 42535 }, { "epoch": 8.1, "learning_rate": 5.699142040038131e-07, "loss": 1.5942, "step": 42536 }, { "epoch": 8.1, "learning_rate": 5.698570066730219e-07, "loss": 0.6159, "step": 42537 }, { "epoch": 8.1, "learning_rate": 5.697998093422307e-07, "loss": 0.8814, "step": 42538 }, { "epoch": 8.1, "learning_rate": 5.697426120114394e-07, "loss": 1.9144, "step": 42539 }, { "epoch": 8.1, "learning_rate": 5.696854146806483e-07, "loss": 1.1645, "step": 42540 }, { "epoch": 8.1, "learning_rate": 5.69628217349857e-07, "loss": 2.077, "step": 42541 }, { "epoch": 8.1, "learning_rate": 5.695710200190658e-07, "loss": 1.4937, "step": 42542 }, { "epoch": 8.1, "learning_rate": 5.695138226882746e-07, "loss": 0.9158, "step": 42543 }, { "epoch": 8.1, "learning_rate": 5.694566253574833e-07, "loss": 1.2511, "step": 42544 }, { "epoch": 8.1, "learning_rate": 5.693994280266922e-07, "loss": 1.4763, "step": 42545 }, { "epoch": 8.1, "learning_rate": 5.693422306959009e-07, "loss": 1.5292, "step": 42546 }, { "epoch": 8.1, "learning_rate": 5.692850333651097e-07, "loss": 1.0987, "step": 42547 }, { "epoch": 8.1, "learning_rate": 5.692278360343184e-07, "loss": 1.3853, "step": 42548 }, { "epoch": 8.1, "learning_rate": 5.691706387035271e-07, "loss": 1.2275, "step": 42549 }, { "epoch": 8.1, "learning_rate": 5.691134413727359e-07, "loss": 0.9462, "step": 42550 }, { "epoch": 8.1, "learning_rate": 5.690562440419447e-07, "loss": 2.161, "step": 42551 }, { "epoch": 8.11, "learning_rate": 5.689990467111535e-07, "loss": 1.7382, "step": 42552 }, { "epoch": 8.11, "learning_rate": 5.689418493803622e-07, "loss": 1.9521, "step": 42553 }, { "epoch": 8.11, "learning_rate": 5.688846520495711e-07, "loss": 0.9099, "step": 42554 }, { "epoch": 8.11, "learning_rate": 5.688274547187798e-07, "loss": 0.3658, "step": 42555 }, { "epoch": 8.11, "learning_rate": 5.687702573879886e-07, "loss": 2.4225, "step": 42556 }, { "epoch": 8.11, "learning_rate": 5.687130600571974e-07, "loss": 1.1391, "step": 42557 }, { "epoch": 8.11, "learning_rate": 5.686558627264061e-07, "loss": 1.5646, "step": 42558 }, { "epoch": 8.11, "learning_rate": 5.68598665395615e-07, "loss": 0.6389, "step": 42559 }, { "epoch": 8.11, "learning_rate": 5.685414680648237e-07, "loss": 1.8545, "step": 42560 }, { "epoch": 8.11, "learning_rate": 5.684842707340323e-07, "loss": 1.2841, "step": 42561 }, { "epoch": 8.11, "learning_rate": 5.684270734032412e-07, "loss": 1.7981, "step": 42562 }, { "epoch": 8.11, "learning_rate": 5.683698760724499e-07, "loss": 1.3186, "step": 42563 }, { "epoch": 8.11, "learning_rate": 5.683126787416587e-07, "loss": 2.1271, "step": 42564 }, { "epoch": 8.11, "learning_rate": 5.682554814108675e-07, "loss": 1.6019, "step": 42565 }, { "epoch": 8.11, "learning_rate": 5.681982840800763e-07, "loss": 1.7067, "step": 42566 }, { "epoch": 8.11, "learning_rate": 5.681410867492851e-07, "loss": 0.6007, "step": 42567 }, { "epoch": 8.11, "learning_rate": 5.680838894184938e-07, "loss": 1.8424, "step": 42568 }, { "epoch": 8.11, "learning_rate": 5.680266920877026e-07, "loss": 0.953, "step": 42569 }, { "epoch": 8.11, "learning_rate": 5.679694947569114e-07, "loss": 1.9167, "step": 42570 }, { "epoch": 8.11, "learning_rate": 5.679122974261202e-07, "loss": 1.3935, "step": 42571 }, { "epoch": 8.11, "learning_rate": 5.678551000953289e-07, "loss": 0.6473, "step": 42572 }, { "epoch": 8.11, "learning_rate": 5.677979027645377e-07, "loss": 1.0703, "step": 42573 }, { "epoch": 8.11, "learning_rate": 5.677407054337465e-07, "loss": 1.6775, "step": 42574 }, { "epoch": 8.11, "learning_rate": 5.676835081029551e-07, "loss": 1.1712, "step": 42575 }, { "epoch": 8.11, "learning_rate": 5.67626310772164e-07, "loss": 3.1485, "step": 42576 }, { "epoch": 8.11, "learning_rate": 5.675691134413727e-07, "loss": 1.8945, "step": 42577 }, { "epoch": 8.11, "learning_rate": 5.675119161105815e-07, "loss": 1.472, "step": 42578 }, { "epoch": 8.11, "learning_rate": 5.674547187797903e-07, "loss": 1.4626, "step": 42579 }, { "epoch": 8.11, "learning_rate": 5.67397521448999e-07, "loss": 2.613, "step": 42580 }, { "epoch": 8.11, "learning_rate": 5.673403241182079e-07, "loss": 1.6208, "step": 42581 }, { "epoch": 8.11, "learning_rate": 5.672831267874166e-07, "loss": 0.9231, "step": 42582 }, { "epoch": 8.11, "learning_rate": 5.672259294566254e-07, "loss": 0.4915, "step": 42583 }, { "epoch": 8.11, "learning_rate": 5.671687321258342e-07, "loss": 1.7562, "step": 42584 }, { "epoch": 8.11, "learning_rate": 5.671115347950429e-07, "loss": 1.2115, "step": 42585 }, { "epoch": 8.11, "learning_rate": 5.670543374642517e-07, "loss": 1.5627, "step": 42586 }, { "epoch": 8.11, "learning_rate": 5.669971401334605e-07, "loss": 0.9946, "step": 42587 }, { "epoch": 8.11, "learning_rate": 5.669399428026692e-07, "loss": 1.282, "step": 42588 }, { "epoch": 8.11, "learning_rate": 5.66882745471878e-07, "loss": 1.5983, "step": 42589 }, { "epoch": 8.11, "learning_rate": 5.668255481410867e-07, "loss": 0.2606, "step": 42590 }, { "epoch": 8.11, "learning_rate": 5.667683508102955e-07, "loss": 1.7875, "step": 42591 }, { "epoch": 8.11, "learning_rate": 5.667111534795043e-07, "loss": 1.0963, "step": 42592 }, { "epoch": 8.11, "learning_rate": 5.666539561487131e-07, "loss": 2.5282, "step": 42593 }, { "epoch": 8.11, "learning_rate": 5.665967588179218e-07, "loss": 1.4678, "step": 42594 }, { "epoch": 8.11, "learning_rate": 5.665395614871307e-07, "loss": 1.4928, "step": 42595 }, { "epoch": 8.11, "learning_rate": 5.664823641563394e-07, "loss": 1.7201, "step": 42596 }, { "epoch": 8.11, "learning_rate": 5.664251668255481e-07, "loss": 1.2721, "step": 42597 }, { "epoch": 8.11, "learning_rate": 5.66367969494757e-07, "loss": 2.1514, "step": 42598 }, { "epoch": 8.11, "learning_rate": 5.663107721639657e-07, "loss": 1.6282, "step": 42599 }, { "epoch": 8.11, "learning_rate": 5.662535748331745e-07, "loss": 3.1939, "step": 42600 }, { "epoch": 8.11, "learning_rate": 5.661963775023832e-07, "loss": 1.7084, "step": 42601 }, { "epoch": 8.11, "learning_rate": 5.661391801715919e-07, "loss": 0.6402, "step": 42602 }, { "epoch": 8.11, "learning_rate": 5.660819828408008e-07, "loss": 1.0526, "step": 42603 }, { "epoch": 8.12, "learning_rate": 5.660247855100095e-07, "loss": 0.9532, "step": 42604 }, { "epoch": 8.12, "learning_rate": 5.659675881792183e-07, "loss": 1.456, "step": 42605 }, { "epoch": 8.12, "learning_rate": 5.659103908484271e-07, "loss": 1.4224, "step": 42606 }, { "epoch": 8.12, "learning_rate": 5.658531935176359e-07, "loss": 1.2997, "step": 42607 }, { "epoch": 8.12, "learning_rate": 5.657959961868446e-07, "loss": 1.0111, "step": 42608 }, { "epoch": 8.12, "learning_rate": 5.657387988560534e-07, "loss": 2.5704, "step": 42609 }, { "epoch": 8.12, "learning_rate": 5.656816015252622e-07, "loss": 1.889, "step": 42610 }, { "epoch": 8.12, "learning_rate": 5.65624404194471e-07, "loss": 1.2157, "step": 42611 }, { "epoch": 8.12, "learning_rate": 5.655672068636798e-07, "loss": 1.8518, "step": 42612 }, { "epoch": 8.12, "learning_rate": 5.655100095328885e-07, "loss": 0.4341, "step": 42613 }, { "epoch": 8.12, "learning_rate": 5.654528122020973e-07, "loss": 0.6328, "step": 42614 }, { "epoch": 8.12, "learning_rate": 5.65395614871306e-07, "loss": 1.5281, "step": 42615 }, { "epoch": 8.12, "learning_rate": 5.653384175405147e-07, "loss": 0.6469, "step": 42616 }, { "epoch": 8.12, "learning_rate": 5.652812202097236e-07, "loss": 2.4403, "step": 42617 }, { "epoch": 8.12, "learning_rate": 5.652240228789323e-07, "loss": 1.9388, "step": 42618 }, { "epoch": 8.12, "learning_rate": 5.651668255481411e-07, "loss": 1.0078, "step": 42619 }, { "epoch": 8.12, "learning_rate": 5.651096282173499e-07, "loss": 1.9701, "step": 42620 }, { "epoch": 8.12, "learning_rate": 5.650524308865586e-07, "loss": 1.62, "step": 42621 }, { "epoch": 8.12, "learning_rate": 5.649952335557674e-07, "loss": 1.7116, "step": 42622 }, { "epoch": 8.12, "learning_rate": 5.649380362249762e-07, "loss": 0.8699, "step": 42623 }, { "epoch": 8.12, "learning_rate": 5.64880838894185e-07, "loss": 1.6651, "step": 42624 }, { "epoch": 8.12, "learning_rate": 5.648236415633938e-07, "loss": 0.6295, "step": 42625 }, { "epoch": 8.12, "learning_rate": 5.647664442326025e-07, "loss": 1.4332, "step": 42626 }, { "epoch": 8.12, "learning_rate": 5.647092469018113e-07, "loss": 1.6125, "step": 42627 }, { "epoch": 8.12, "learning_rate": 5.6465204957102e-07, "loss": 2.1494, "step": 42628 }, { "epoch": 8.12, "learning_rate": 5.645948522402288e-07, "loss": 1.1518, "step": 42629 }, { "epoch": 8.12, "learning_rate": 5.645376549094375e-07, "loss": 1.6561, "step": 42630 }, { "epoch": 8.12, "learning_rate": 5.644804575786463e-07, "loss": 0.85, "step": 42631 }, { "epoch": 8.12, "learning_rate": 5.644232602478551e-07, "loss": 2.3298, "step": 42632 }, { "epoch": 8.12, "learning_rate": 5.643660629170639e-07, "loss": 2.1716, "step": 42633 }, { "epoch": 8.12, "learning_rate": 5.643088655862727e-07, "loss": 1.239, "step": 42634 }, { "epoch": 8.12, "learning_rate": 5.642516682554814e-07, "loss": 2.0665, "step": 42635 }, { "epoch": 8.12, "learning_rate": 5.641944709246902e-07, "loss": 1.717, "step": 42636 }, { "epoch": 8.12, "learning_rate": 5.64137273593899e-07, "loss": 2.6327, "step": 42637 }, { "epoch": 8.12, "learning_rate": 5.640800762631077e-07, "loss": 2.7069, "step": 42638 }, { "epoch": 8.12, "learning_rate": 5.640228789323166e-07, "loss": 1.7666, "step": 42639 }, { "epoch": 8.12, "learning_rate": 5.639656816015253e-07, "loss": 1.613, "step": 42640 }, { "epoch": 8.12, "learning_rate": 5.63908484270734e-07, "loss": 1.7281, "step": 42641 }, { "epoch": 8.12, "learning_rate": 5.638512869399428e-07, "loss": 1.8406, "step": 42642 }, { "epoch": 8.12, "learning_rate": 5.637940896091515e-07, "loss": 1.5881, "step": 42643 }, { "epoch": 8.12, "learning_rate": 5.637368922783603e-07, "loss": 1.2361, "step": 42644 }, { "epoch": 8.12, "learning_rate": 5.636796949475691e-07, "loss": 1.3271, "step": 42645 }, { "epoch": 8.12, "learning_rate": 5.636224976167779e-07, "loss": 2.034, "step": 42646 }, { "epoch": 8.12, "learning_rate": 5.635653002859867e-07, "loss": 1.8478, "step": 42647 }, { "epoch": 8.12, "learning_rate": 5.635081029551955e-07, "loss": 1.0157, "step": 42648 }, { "epoch": 8.12, "learning_rate": 5.634509056244042e-07, "loss": 2.3682, "step": 42649 }, { "epoch": 8.12, "learning_rate": 5.63393708293613e-07, "loss": 1.7399, "step": 42650 }, { "epoch": 8.12, "learning_rate": 5.633365109628218e-07, "loss": 1.9887, "step": 42651 }, { "epoch": 8.12, "learning_rate": 5.632793136320305e-07, "loss": 1.774, "step": 42652 }, { "epoch": 8.12, "learning_rate": 5.632221163012394e-07, "loss": 2.4796, "step": 42653 }, { "epoch": 8.12, "learning_rate": 5.63164918970448e-07, "loss": 1.614, "step": 42654 }, { "epoch": 8.12, "learning_rate": 5.631077216396568e-07, "loss": 0.2416, "step": 42655 }, { "epoch": 8.12, "learning_rate": 5.630505243088656e-07, "loss": 1.5037, "step": 42656 }, { "epoch": 8.13, "learning_rate": 5.629933269780743e-07, "loss": 0.6647, "step": 42657 }, { "epoch": 8.13, "learning_rate": 5.629361296472831e-07, "loss": 1.742, "step": 42658 }, { "epoch": 8.13, "learning_rate": 5.628789323164919e-07, "loss": 1.4585, "step": 42659 }, { "epoch": 8.13, "learning_rate": 5.628217349857007e-07, "loss": 1.2183, "step": 42660 }, { "epoch": 8.13, "learning_rate": 5.627645376549095e-07, "loss": 1.5459, "step": 42661 }, { "epoch": 8.13, "learning_rate": 5.627073403241182e-07, "loss": 1.5089, "step": 42662 }, { "epoch": 8.13, "learning_rate": 5.62650142993327e-07, "loss": 0.9117, "step": 42663 }, { "epoch": 8.13, "learning_rate": 5.625929456625358e-07, "loss": 1.9631, "step": 42664 }, { "epoch": 8.13, "learning_rate": 5.625357483317446e-07, "loss": 2.551, "step": 42665 }, { "epoch": 8.13, "learning_rate": 5.624785510009533e-07, "loss": 1.6424, "step": 42666 }, { "epoch": 8.13, "learning_rate": 5.62421353670162e-07, "loss": 1.9751, "step": 42667 }, { "epoch": 8.13, "learning_rate": 5.623641563393708e-07, "loss": 0.5053, "step": 42668 }, { "epoch": 8.13, "learning_rate": 5.623069590085796e-07, "loss": 1.5119, "step": 42669 }, { "epoch": 8.13, "learning_rate": 5.622497616777884e-07, "loss": 2.1832, "step": 42670 }, { "epoch": 8.13, "learning_rate": 5.621925643469971e-07, "loss": 0.7399, "step": 42671 }, { "epoch": 8.13, "learning_rate": 5.621353670162059e-07, "loss": 1.0449, "step": 42672 }, { "epoch": 8.13, "learning_rate": 5.620781696854147e-07, "loss": 0.933, "step": 42673 }, { "epoch": 8.13, "learning_rate": 5.620209723546234e-07, "loss": 0.7635, "step": 42674 }, { "epoch": 8.13, "learning_rate": 5.619637750238323e-07, "loss": 2.0938, "step": 42675 }, { "epoch": 8.13, "learning_rate": 5.61906577693041e-07, "loss": 1.2854, "step": 42676 }, { "epoch": 8.13, "learning_rate": 5.618493803622498e-07, "loss": 0.8822, "step": 42677 }, { "epoch": 8.13, "learning_rate": 5.617921830314586e-07, "loss": 1.5744, "step": 42678 }, { "epoch": 8.13, "learning_rate": 5.617349857006673e-07, "loss": 1.3514, "step": 42679 }, { "epoch": 8.13, "learning_rate": 5.616777883698761e-07, "loss": 1.3186, "step": 42680 }, { "epoch": 8.13, "learning_rate": 5.616205910390848e-07, "loss": 2.6013, "step": 42681 }, { "epoch": 8.13, "learning_rate": 5.615633937082936e-07, "loss": 1.2725, "step": 42682 }, { "epoch": 8.13, "learning_rate": 5.615061963775024e-07, "loss": 2.5088, "step": 42683 }, { "epoch": 8.13, "learning_rate": 5.614489990467112e-07, "loss": 2.3981, "step": 42684 }, { "epoch": 8.13, "learning_rate": 5.613918017159199e-07, "loss": 1.6586, "step": 42685 }, { "epoch": 8.13, "learning_rate": 5.613346043851287e-07, "loss": 2.7515, "step": 42686 }, { "epoch": 8.13, "learning_rate": 5.612774070543375e-07, "loss": 1.2127, "step": 42687 }, { "epoch": 8.13, "learning_rate": 5.612202097235462e-07, "loss": 1.3971, "step": 42688 }, { "epoch": 8.13, "learning_rate": 5.611630123927551e-07, "loss": 1.0653, "step": 42689 }, { "epoch": 8.13, "learning_rate": 5.611058150619638e-07, "loss": 1.3862, "step": 42690 }, { "epoch": 8.13, "learning_rate": 5.610486177311726e-07, "loss": 3.0968, "step": 42691 }, { "epoch": 8.13, "learning_rate": 5.609914204003814e-07, "loss": 1.058, "step": 42692 }, { "epoch": 8.13, "learning_rate": 5.609342230695901e-07, "loss": 3.5094, "step": 42693 }, { "epoch": 8.13, "learning_rate": 5.608770257387989e-07, "loss": 1.7294, "step": 42694 }, { "epoch": 8.13, "learning_rate": 5.608198284080076e-07, "loss": 0.7295, "step": 42695 }, { "epoch": 8.13, "learning_rate": 5.607626310772163e-07, "loss": 0.8721, "step": 42696 }, { "epoch": 8.13, "learning_rate": 5.607054337464252e-07, "loss": 1.2441, "step": 42697 }, { "epoch": 8.13, "learning_rate": 5.606482364156339e-07, "loss": 1.3895, "step": 42698 }, { "epoch": 8.13, "learning_rate": 5.605910390848427e-07, "loss": 0.5766, "step": 42699 }, { "epoch": 8.13, "learning_rate": 5.605338417540515e-07, "loss": 1.5381, "step": 42700 }, { "epoch": 8.13, "learning_rate": 5.604766444232603e-07, "loss": 1.193, "step": 42701 }, { "epoch": 8.13, "learning_rate": 5.60419447092469e-07, "loss": 0.7879, "step": 42702 }, { "epoch": 8.13, "learning_rate": 5.603622497616778e-07, "loss": 2.1946, "step": 42703 }, { "epoch": 8.13, "learning_rate": 5.603050524308866e-07, "loss": 1.0189, "step": 42704 }, { "epoch": 8.13, "learning_rate": 5.602478551000954e-07, "loss": 2.8168, "step": 42705 }, { "epoch": 8.13, "learning_rate": 5.601906577693042e-07, "loss": 0.8196, "step": 42706 }, { "epoch": 8.13, "learning_rate": 5.601334604385129e-07, "loss": 1.239, "step": 42707 }, { "epoch": 8.13, "learning_rate": 5.600762631077216e-07, "loss": 1.7195, "step": 42708 }, { "epoch": 8.14, "learning_rate": 5.600190657769304e-07, "loss": 2.256, "step": 42709 }, { "epoch": 8.14, "learning_rate": 5.599618684461391e-07, "loss": 0.7189, "step": 42710 }, { "epoch": 8.14, "learning_rate": 5.59904671115348e-07, "loss": 1.2994, "step": 42711 }, { "epoch": 8.14, "learning_rate": 5.598474737845567e-07, "loss": 1.4535, "step": 42712 }, { "epoch": 8.14, "learning_rate": 5.597902764537655e-07, "loss": 0.5464, "step": 42713 }, { "epoch": 8.14, "learning_rate": 5.597330791229743e-07, "loss": 1.2229, "step": 42714 }, { "epoch": 8.14, "learning_rate": 5.59675881792183e-07, "loss": 0.8727, "step": 42715 }, { "epoch": 8.14, "learning_rate": 5.596186844613918e-07, "loss": 1.1821, "step": 42716 }, { "epoch": 8.14, "learning_rate": 5.595614871306006e-07, "loss": 2.9995, "step": 42717 }, { "epoch": 8.14, "learning_rate": 5.595042897998094e-07, "loss": 0.7446, "step": 42718 }, { "epoch": 8.14, "learning_rate": 5.594470924690182e-07, "loss": 0.5484, "step": 42719 }, { "epoch": 8.14, "learning_rate": 5.593898951382269e-07, "loss": 2.8995, "step": 42720 }, { "epoch": 8.14, "learning_rate": 5.593326978074356e-07, "loss": 1.4719, "step": 42721 }, { "epoch": 8.14, "learning_rate": 5.592755004766444e-07, "loss": 1.6072, "step": 42722 }, { "epoch": 8.14, "learning_rate": 5.592183031458532e-07, "loss": 2.7161, "step": 42723 }, { "epoch": 8.14, "learning_rate": 5.591611058150619e-07, "loss": 2.6883, "step": 42724 }, { "epoch": 8.14, "learning_rate": 5.591039084842708e-07, "loss": 0.7958, "step": 42725 }, { "epoch": 8.14, "learning_rate": 5.590467111534795e-07, "loss": 0.9167, "step": 42726 }, { "epoch": 8.14, "learning_rate": 5.589895138226883e-07, "loss": 1.1518, "step": 42727 }, { "epoch": 8.14, "learning_rate": 5.589323164918971e-07, "loss": 1.8308, "step": 42728 }, { "epoch": 8.14, "learning_rate": 5.588751191611058e-07, "loss": 1.4933, "step": 42729 }, { "epoch": 8.14, "learning_rate": 5.588179218303146e-07, "loss": 1.6482, "step": 42730 }, { "epoch": 8.14, "learning_rate": 5.587607244995234e-07, "loss": 0.7244, "step": 42731 }, { "epoch": 8.14, "learning_rate": 5.587035271687321e-07, "loss": 1.7249, "step": 42732 }, { "epoch": 8.14, "learning_rate": 5.58646329837941e-07, "loss": 2.2088, "step": 42733 }, { "epoch": 8.14, "learning_rate": 5.585891325071497e-07, "loss": 1.4407, "step": 42734 }, { "epoch": 8.14, "learning_rate": 5.585319351763584e-07, "loss": 2.4866, "step": 42735 }, { "epoch": 8.14, "learning_rate": 5.584747378455672e-07, "loss": 0.2766, "step": 42736 }, { "epoch": 8.14, "learning_rate": 5.58417540514776e-07, "loss": 0.7202, "step": 42737 }, { "epoch": 8.14, "learning_rate": 5.583603431839847e-07, "loss": 1.322, "step": 42738 }, { "epoch": 8.14, "learning_rate": 5.583031458531935e-07, "loss": 1.1792, "step": 42739 }, { "epoch": 8.14, "learning_rate": 5.582459485224023e-07, "loss": 1.684, "step": 42740 }, { "epoch": 8.14, "learning_rate": 5.581887511916111e-07, "loss": 2.7152, "step": 42741 }, { "epoch": 8.14, "learning_rate": 5.581315538608199e-07, "loss": 0.7796, "step": 42742 }, { "epoch": 8.14, "learning_rate": 5.580743565300286e-07, "loss": 1.5626, "step": 42743 }, { "epoch": 8.14, "learning_rate": 5.580171591992374e-07, "loss": 1.7952, "step": 42744 }, { "epoch": 8.14, "learning_rate": 5.579599618684462e-07, "loss": 0.464, "step": 42745 }, { "epoch": 8.14, "learning_rate": 5.579027645376549e-07, "loss": 1.6311, "step": 42746 }, { "epoch": 8.14, "learning_rate": 5.578455672068638e-07, "loss": 2.1967, "step": 42747 }, { "epoch": 8.14, "learning_rate": 5.577883698760724e-07, "loss": 0.4597, "step": 42748 }, { "epoch": 8.14, "learning_rate": 5.577311725452812e-07, "loss": 0.5504, "step": 42749 }, { "epoch": 8.14, "learning_rate": 5.5767397521449e-07, "loss": 1.235, "step": 42750 }, { "epoch": 8.14, "learning_rate": 5.576167778836987e-07, "loss": 1.0983, "step": 42751 }, { "epoch": 8.14, "learning_rate": 5.575595805529075e-07, "loss": 1.0774, "step": 42752 }, { "epoch": 8.14, "learning_rate": 5.575023832221163e-07, "loss": 2.4397, "step": 42753 }, { "epoch": 8.14, "learning_rate": 5.574451858913251e-07, "loss": 1.1854, "step": 42754 }, { "epoch": 8.14, "learning_rate": 5.573879885605339e-07, "loss": 1.1414, "step": 42755 }, { "epoch": 8.14, "learning_rate": 5.573307912297426e-07, "loss": 1.1535, "step": 42756 }, { "epoch": 8.14, "learning_rate": 5.572735938989514e-07, "loss": 1.1269, "step": 42757 }, { "epoch": 8.14, "learning_rate": 5.572163965681602e-07, "loss": 2.9086, "step": 42758 }, { "epoch": 8.14, "learning_rate": 5.57159199237369e-07, "loss": 2.325, "step": 42759 }, { "epoch": 8.14, "learning_rate": 5.571020019065777e-07, "loss": 1.2207, "step": 42760 }, { "epoch": 8.14, "learning_rate": 5.570448045757864e-07, "loss": 2.7788, "step": 42761 }, { "epoch": 8.15, "learning_rate": 5.569876072449952e-07, "loss": 2.5071, "step": 42762 }, { "epoch": 8.15, "learning_rate": 5.56930409914204e-07, "loss": 1.5836, "step": 42763 }, { "epoch": 8.15, "learning_rate": 5.568732125834128e-07, "loss": 1.8385, "step": 42764 }, { "epoch": 8.15, "learning_rate": 5.568160152526215e-07, "loss": 1.0574, "step": 42765 }, { "epoch": 8.15, "learning_rate": 5.567588179218304e-07, "loss": 1.2041, "step": 42766 }, { "epoch": 8.15, "learning_rate": 5.567016205910391e-07, "loss": 1.115, "step": 42767 }, { "epoch": 8.15, "learning_rate": 5.566444232602478e-07, "loss": 2.0781, "step": 42768 }, { "epoch": 8.15, "learning_rate": 5.565872259294567e-07, "loss": 2.3426, "step": 42769 }, { "epoch": 8.15, "learning_rate": 5.565300285986654e-07, "loss": 2.2907, "step": 42770 }, { "epoch": 8.15, "learning_rate": 5.564728312678742e-07, "loss": 2.2318, "step": 42771 }, { "epoch": 8.15, "learning_rate": 5.56415633937083e-07, "loss": 0.9201, "step": 42772 }, { "epoch": 8.15, "learning_rate": 5.563584366062917e-07, "loss": 2.9878, "step": 42773 }, { "epoch": 8.15, "learning_rate": 5.563012392755005e-07, "loss": 1.9153, "step": 42774 }, { "epoch": 8.15, "learning_rate": 5.562440419447092e-07, "loss": 1.9941, "step": 42775 }, { "epoch": 8.15, "learning_rate": 5.56186844613918e-07, "loss": 0.7079, "step": 42776 }, { "epoch": 8.15, "learning_rate": 5.561296472831268e-07, "loss": 2.5926, "step": 42777 }, { "epoch": 8.15, "learning_rate": 5.560724499523356e-07, "loss": 2.7535, "step": 42778 }, { "epoch": 8.15, "learning_rate": 5.560152526215443e-07, "loss": 0.6327, "step": 42779 }, { "epoch": 8.15, "learning_rate": 5.559580552907531e-07, "loss": 1.1719, "step": 42780 }, { "epoch": 8.15, "learning_rate": 5.559008579599619e-07, "loss": 0.894, "step": 42781 }, { "epoch": 8.15, "learning_rate": 5.558436606291706e-07, "loss": 1.2904, "step": 42782 }, { "epoch": 8.15, "learning_rate": 5.557864632983795e-07, "loss": 2.0517, "step": 42783 }, { "epoch": 8.15, "learning_rate": 5.557292659675882e-07, "loss": 2.2676, "step": 42784 }, { "epoch": 8.15, "learning_rate": 5.55672068636797e-07, "loss": 1.8962, "step": 42785 }, { "epoch": 8.15, "learning_rate": 5.556148713060058e-07, "loss": 2.4316, "step": 42786 }, { "epoch": 8.15, "learning_rate": 5.555576739752145e-07, "loss": 0.9831, "step": 42787 }, { "epoch": 8.15, "learning_rate": 5.555004766444234e-07, "loss": 1.6139, "step": 42788 }, { "epoch": 8.15, "learning_rate": 5.55443279313632e-07, "loss": 1.136, "step": 42789 }, { "epoch": 8.15, "learning_rate": 5.553860819828408e-07, "loss": 0.9066, "step": 42790 }, { "epoch": 8.15, "learning_rate": 5.553288846520496e-07, "loss": 0.436, "step": 42791 }, { "epoch": 8.15, "learning_rate": 5.552716873212583e-07, "loss": 1.218, "step": 42792 }, { "epoch": 8.15, "learning_rate": 5.552144899904671e-07, "loss": 2.175, "step": 42793 }, { "epoch": 8.15, "learning_rate": 5.551572926596759e-07, "loss": 1.92, "step": 42794 }, { "epoch": 8.15, "learning_rate": 5.551000953288847e-07, "loss": 1.7652, "step": 42795 }, { "epoch": 8.15, "learning_rate": 5.550428979980934e-07, "loss": 1.5589, "step": 42796 }, { "epoch": 8.15, "learning_rate": 5.549857006673022e-07, "loss": 2.1881, "step": 42797 }, { "epoch": 8.15, "learning_rate": 5.54928503336511e-07, "loss": 1.5193, "step": 42798 }, { "epoch": 8.15, "learning_rate": 5.548713060057198e-07, "loss": 1.1637, "step": 42799 }, { "epoch": 8.15, "learning_rate": 5.548141086749286e-07, "loss": 2.3945, "step": 42800 }, { "epoch": 8.15, "learning_rate": 5.547569113441372e-07, "loss": 1.8711, "step": 42801 }, { "epoch": 8.15, "learning_rate": 5.546997140133461e-07, "loss": 2.0037, "step": 42802 }, { "epoch": 8.15, "learning_rate": 5.546425166825548e-07, "loss": 0.7383, "step": 42803 }, { "epoch": 8.15, "learning_rate": 5.545853193517635e-07, "loss": 0.9863, "step": 42804 }, { "epoch": 8.15, "learning_rate": 5.545281220209724e-07, "loss": 1.7535, "step": 42805 }, { "epoch": 8.15, "learning_rate": 5.544709246901811e-07, "loss": 0.7082, "step": 42806 }, { "epoch": 8.15, "learning_rate": 5.544137273593899e-07, "loss": 0.9997, "step": 42807 }, { "epoch": 8.15, "learning_rate": 5.543565300285987e-07, "loss": 0.9137, "step": 42808 }, { "epoch": 8.15, "learning_rate": 5.542993326978074e-07, "loss": 3.4805, "step": 42809 }, { "epoch": 8.15, "learning_rate": 5.542421353670163e-07, "loss": 0.733, "step": 42810 }, { "epoch": 8.15, "learning_rate": 5.54184938036225e-07, "loss": 1.4277, "step": 42811 }, { "epoch": 8.15, "learning_rate": 5.541277407054338e-07, "loss": 2.4166, "step": 42812 }, { "epoch": 8.15, "learning_rate": 5.540705433746426e-07, "loss": 1.4359, "step": 42813 }, { "epoch": 8.16, "learning_rate": 5.540133460438513e-07, "loss": 2.1501, "step": 42814 }, { "epoch": 8.16, "learning_rate": 5.5395614871306e-07, "loss": 1.0868, "step": 42815 }, { "epoch": 8.16, "learning_rate": 5.538989513822688e-07, "loss": 2.3542, "step": 42816 }, { "epoch": 8.16, "learning_rate": 5.538417540514776e-07, "loss": 0.9916, "step": 42817 }, { "epoch": 8.16, "learning_rate": 5.537845567206863e-07, "loss": 1.01, "step": 42818 }, { "epoch": 8.16, "learning_rate": 5.537273593898952e-07, "loss": 0.4189, "step": 42819 }, { "epoch": 8.16, "learning_rate": 5.536701620591039e-07, "loss": 1.0229, "step": 42820 }, { "epoch": 8.16, "learning_rate": 5.536129647283127e-07, "loss": 2.1439, "step": 42821 }, { "epoch": 8.16, "learning_rate": 5.535557673975215e-07, "loss": 2.6958, "step": 42822 }, { "epoch": 8.16, "learning_rate": 5.534985700667302e-07, "loss": 1.3941, "step": 42823 }, { "epoch": 8.16, "learning_rate": 5.534413727359391e-07, "loss": 1.7085, "step": 42824 }, { "epoch": 8.16, "learning_rate": 5.533841754051478e-07, "loss": 1.5993, "step": 42825 }, { "epoch": 8.16, "learning_rate": 5.533269780743565e-07, "loss": 1.3213, "step": 42826 }, { "epoch": 8.16, "learning_rate": 5.532697807435654e-07, "loss": 1.3548, "step": 42827 }, { "epoch": 8.16, "learning_rate": 5.53212583412774e-07, "loss": 1.4784, "step": 42828 }, { "epoch": 8.16, "learning_rate": 5.531553860819828e-07, "loss": 2.4068, "step": 42829 }, { "epoch": 8.16, "learning_rate": 5.530981887511916e-07, "loss": 2.3264, "step": 42830 }, { "epoch": 8.16, "learning_rate": 5.530409914204004e-07, "loss": 1.7167, "step": 42831 }, { "epoch": 8.16, "learning_rate": 5.529837940896091e-07, "loss": 1.6366, "step": 42832 }, { "epoch": 8.16, "learning_rate": 5.529265967588179e-07, "loss": 0.7905, "step": 42833 }, { "epoch": 8.16, "learning_rate": 5.528693994280267e-07, "loss": 1.5245, "step": 42834 }, { "epoch": 8.16, "learning_rate": 5.528122020972355e-07, "loss": 1.7028, "step": 42835 }, { "epoch": 8.16, "learning_rate": 5.527550047664443e-07, "loss": 3.1511, "step": 42836 }, { "epoch": 8.16, "learning_rate": 5.52697807435653e-07, "loss": 1.5518, "step": 42837 }, { "epoch": 8.16, "learning_rate": 5.526406101048618e-07, "loss": 1.8999, "step": 42838 }, { "epoch": 8.16, "learning_rate": 5.525834127740706e-07, "loss": 0.8645, "step": 42839 }, { "epoch": 8.16, "learning_rate": 5.525262154432793e-07, "loss": 1.1058, "step": 42840 }, { "epoch": 8.16, "learning_rate": 5.524690181124882e-07, "loss": 1.5708, "step": 42841 }, { "epoch": 8.16, "learning_rate": 5.524118207816968e-07, "loss": 0.742, "step": 42842 }, { "epoch": 8.16, "learning_rate": 5.523546234509056e-07, "loss": 1.2318, "step": 42843 }, { "epoch": 8.16, "learning_rate": 5.522974261201144e-07, "loss": 2.9093, "step": 42844 }, { "epoch": 8.16, "learning_rate": 5.522402287893231e-07, "loss": 2.3414, "step": 42845 }, { "epoch": 8.16, "learning_rate": 5.52183031458532e-07, "loss": 1.7557, "step": 42846 }, { "epoch": 8.16, "learning_rate": 5.521258341277407e-07, "loss": 2.2235, "step": 42847 }, { "epoch": 8.16, "learning_rate": 5.520686367969495e-07, "loss": 1.3306, "step": 42848 }, { "epoch": 8.16, "learning_rate": 5.520114394661583e-07, "loss": 1.7568, "step": 42849 }, { "epoch": 8.16, "learning_rate": 5.51954242135367e-07, "loss": 3.5754, "step": 42850 }, { "epoch": 8.16, "learning_rate": 5.518970448045758e-07, "loss": 2.2607, "step": 42851 }, { "epoch": 8.16, "learning_rate": 5.518398474737846e-07, "loss": 1.7733, "step": 42852 }, { "epoch": 8.16, "learning_rate": 5.517826501429934e-07, "loss": 1.9384, "step": 42853 }, { "epoch": 8.16, "learning_rate": 5.51725452812202e-07, "loss": 1.4317, "step": 42854 }, { "epoch": 8.16, "learning_rate": 5.51668255481411e-07, "loss": 1.0345, "step": 42855 }, { "epoch": 8.16, "learning_rate": 5.516110581506196e-07, "loss": 0.6003, "step": 42856 }, { "epoch": 8.16, "learning_rate": 5.515538608198284e-07, "loss": 0.9331, "step": 42857 }, { "epoch": 8.16, "learning_rate": 5.514966634890372e-07, "loss": 0.5503, "step": 42858 }, { "epoch": 8.16, "learning_rate": 5.514394661582459e-07, "loss": 1.3928, "step": 42859 }, { "epoch": 8.16, "learning_rate": 5.513822688274548e-07, "loss": 2.6631, "step": 42860 }, { "epoch": 8.16, "learning_rate": 5.513250714966635e-07, "loss": 2.1149, "step": 42861 }, { "epoch": 8.16, "learning_rate": 5.512678741658722e-07, "loss": 1.2409, "step": 42862 }, { "epoch": 8.16, "learning_rate": 5.512106768350811e-07, "loss": 2.9596, "step": 42863 }, { "epoch": 8.16, "learning_rate": 5.511534795042898e-07, "loss": 1.3943, "step": 42864 }, { "epoch": 8.16, "learning_rate": 5.510962821734986e-07, "loss": 2.6076, "step": 42865 }, { "epoch": 8.16, "learning_rate": 5.510390848427074e-07, "loss": 0.8231, "step": 42866 }, { "epoch": 8.17, "learning_rate": 5.509818875119162e-07, "loss": 2.8209, "step": 42867 }, { "epoch": 8.17, "learning_rate": 5.50924690181125e-07, "loss": 1.3215, "step": 42868 }, { "epoch": 8.17, "learning_rate": 5.508674928503336e-07, "loss": 0.9697, "step": 42869 }, { "epoch": 8.17, "learning_rate": 5.508102955195424e-07, "loss": 1.1668, "step": 42870 }, { "epoch": 8.17, "learning_rate": 5.507530981887512e-07, "loss": 1.9214, "step": 42871 }, { "epoch": 8.17, "learning_rate": 5.5069590085796e-07, "loss": 2.9888, "step": 42872 }, { "epoch": 8.17, "learning_rate": 5.506387035271687e-07, "loss": 1.1725, "step": 42873 }, { "epoch": 8.17, "learning_rate": 5.505815061963775e-07, "loss": 2.1981, "step": 42874 }, { "epoch": 8.17, "learning_rate": 5.505243088655863e-07, "loss": 1.155, "step": 42875 }, { "epoch": 8.17, "learning_rate": 5.50467111534795e-07, "loss": 1.6244, "step": 42876 }, { "epoch": 8.17, "learning_rate": 5.504099142040039e-07, "loss": 2.0005, "step": 42877 }, { "epoch": 8.17, "learning_rate": 5.503527168732126e-07, "loss": 1.3297, "step": 42878 }, { "epoch": 8.17, "learning_rate": 5.502955195424214e-07, "loss": 1.3913, "step": 42879 }, { "epoch": 8.17, "learning_rate": 5.502383222116302e-07, "loss": 0.7918, "step": 42880 }, { "epoch": 8.17, "learning_rate": 5.501811248808388e-07, "loss": 1.9113, "step": 42881 }, { "epoch": 8.17, "learning_rate": 5.501239275500477e-07, "loss": 1.754, "step": 42882 }, { "epoch": 8.17, "learning_rate": 5.500667302192564e-07, "loss": 0.86, "step": 42883 }, { "epoch": 8.17, "learning_rate": 5.500095328884652e-07, "loss": 0.5163, "step": 42884 }, { "epoch": 8.17, "learning_rate": 5.49952335557674e-07, "loss": 0.7391, "step": 42885 }, { "epoch": 8.17, "learning_rate": 5.498951382268827e-07, "loss": 0.9787, "step": 42886 }, { "epoch": 8.17, "learning_rate": 5.498379408960915e-07, "loss": 3.2076, "step": 42887 }, { "epoch": 8.17, "learning_rate": 5.497807435653003e-07, "loss": 1.1175, "step": 42888 }, { "epoch": 8.17, "learning_rate": 5.497235462345091e-07, "loss": 0.4311, "step": 42889 }, { "epoch": 8.17, "learning_rate": 5.496663489037179e-07, "loss": 0.3124, "step": 42890 }, { "epoch": 8.17, "learning_rate": 5.496091515729266e-07, "loss": 2.0842, "step": 42891 }, { "epoch": 8.17, "learning_rate": 5.495519542421354e-07, "loss": 1.9535, "step": 42892 }, { "epoch": 8.17, "learning_rate": 5.494947569113442e-07, "loss": 0.7274, "step": 42893 }, { "epoch": 8.17, "learning_rate": 5.49437559580553e-07, "loss": 1.28, "step": 42894 }, { "epoch": 8.17, "learning_rate": 5.493803622497616e-07, "loss": 2.1022, "step": 42895 }, { "epoch": 8.17, "learning_rate": 5.493231649189705e-07, "loss": 2.7038, "step": 42896 }, { "epoch": 8.17, "learning_rate": 5.492659675881792e-07, "loss": 2.0784, "step": 42897 }, { "epoch": 8.17, "learning_rate": 5.492087702573879e-07, "loss": 1.4301, "step": 42898 }, { "epoch": 8.17, "learning_rate": 5.491515729265968e-07, "loss": 0.8101, "step": 42899 }, { "epoch": 8.17, "learning_rate": 5.490943755958055e-07, "loss": 2.6085, "step": 42900 }, { "epoch": 8.17, "learning_rate": 5.490371782650143e-07, "loss": 2.3055, "step": 42901 }, { "epoch": 8.17, "learning_rate": 5.489799809342231e-07, "loss": 0.8937, "step": 42902 }, { "epoch": 8.17, "learning_rate": 5.489227836034318e-07, "loss": 1.8771, "step": 42903 }, { "epoch": 8.17, "learning_rate": 5.488655862726407e-07, "loss": 2.659, "step": 42904 }, { "epoch": 8.17, "learning_rate": 5.488083889418494e-07, "loss": 2.7195, "step": 42905 }, { "epoch": 8.17, "learning_rate": 5.487511916110582e-07, "loss": 0.7919, "step": 42906 }, { "epoch": 8.17, "learning_rate": 5.48693994280267e-07, "loss": 2.2914, "step": 42907 }, { "epoch": 8.17, "learning_rate": 5.486367969494758e-07, "loss": 2.4319, "step": 42908 }, { "epoch": 8.17, "learning_rate": 5.485795996186844e-07, "loss": 0.7238, "step": 42909 }, { "epoch": 8.17, "learning_rate": 5.485224022878932e-07, "loss": 1.6307, "step": 42910 }, { "epoch": 8.17, "learning_rate": 5.48465204957102e-07, "loss": 1.58, "step": 42911 }, { "epoch": 8.17, "learning_rate": 5.484080076263108e-07, "loss": 1.6738, "step": 42912 }, { "epoch": 8.17, "learning_rate": 5.483508102955196e-07, "loss": 0.8227, "step": 42913 }, { "epoch": 8.17, "learning_rate": 5.482936129647283e-07, "loss": 1.4839, "step": 42914 }, { "epoch": 8.17, "learning_rate": 5.482364156339371e-07, "loss": 2.7947, "step": 42915 }, { "epoch": 8.17, "learning_rate": 5.481792183031459e-07, "loss": 0.6101, "step": 42916 }, { "epoch": 8.17, "learning_rate": 5.481220209723546e-07, "loss": 0.7295, "step": 42917 }, { "epoch": 8.17, "learning_rate": 5.480648236415635e-07, "loss": 2.3229, "step": 42918 }, { "epoch": 8.18, "learning_rate": 5.480076263107722e-07, "loss": 1.8338, "step": 42919 }, { "epoch": 8.18, "learning_rate": 5.47950428979981e-07, "loss": 1.7172, "step": 42920 }, { "epoch": 8.18, "learning_rate": 5.478932316491898e-07, "loss": 2.7627, "step": 42921 }, { "epoch": 8.18, "learning_rate": 5.478360343183984e-07, "loss": 1.8016, "step": 42922 }, { "epoch": 8.18, "learning_rate": 5.477788369876072e-07, "loss": 1.8202, "step": 42923 }, { "epoch": 8.18, "learning_rate": 5.47721639656816e-07, "loss": 0.7066, "step": 42924 }, { "epoch": 8.18, "learning_rate": 5.476644423260248e-07, "loss": 0.5597, "step": 42925 }, { "epoch": 8.18, "learning_rate": 5.476072449952336e-07, "loss": 1.0885, "step": 42926 }, { "epoch": 8.18, "learning_rate": 5.475500476644423e-07, "loss": 1.3738, "step": 42927 }, { "epoch": 8.18, "learning_rate": 5.474928503336511e-07, "loss": 1.06, "step": 42928 }, { "epoch": 8.18, "learning_rate": 5.474356530028599e-07, "loss": 3.7601, "step": 42929 }, { "epoch": 8.18, "learning_rate": 5.473784556720687e-07, "loss": 3.442, "step": 42930 }, { "epoch": 8.18, "learning_rate": 5.473212583412774e-07, "loss": 1.691, "step": 42931 }, { "epoch": 8.18, "learning_rate": 5.472640610104862e-07, "loss": 1.6871, "step": 42932 }, { "epoch": 8.18, "learning_rate": 5.47206863679695e-07, "loss": 2.8606, "step": 42933 }, { "epoch": 8.18, "learning_rate": 5.471496663489038e-07, "loss": 0.4182, "step": 42934 }, { "epoch": 8.18, "learning_rate": 5.470924690181125e-07, "loss": 2.367, "step": 42935 }, { "epoch": 8.18, "learning_rate": 5.470352716873212e-07, "loss": 2.83, "step": 42936 }, { "epoch": 8.18, "learning_rate": 5.4697807435653e-07, "loss": 1.7973, "step": 42937 }, { "epoch": 8.18, "learning_rate": 5.469208770257388e-07, "loss": 2.4828, "step": 42938 }, { "epoch": 8.18, "learning_rate": 5.468636796949475e-07, "loss": 1.6626, "step": 42939 }, { "epoch": 8.18, "learning_rate": 5.468064823641564e-07, "loss": 1.4755, "step": 42940 }, { "epoch": 8.18, "learning_rate": 5.467492850333651e-07, "loss": 2.4573, "step": 42941 }, { "epoch": 8.18, "learning_rate": 5.466920877025739e-07, "loss": 0.7564, "step": 42942 }, { "epoch": 8.18, "learning_rate": 5.466348903717827e-07, "loss": 1.4191, "step": 42943 }, { "epoch": 8.18, "learning_rate": 5.465776930409914e-07, "loss": 3.0555, "step": 42944 }, { "epoch": 8.18, "learning_rate": 5.465204957102002e-07, "loss": 2.8513, "step": 42945 }, { "epoch": 8.18, "learning_rate": 5.46463298379409e-07, "loss": 2.3106, "step": 42946 }, { "epoch": 8.18, "learning_rate": 5.464061010486178e-07, "loss": 2.7288, "step": 42947 }, { "epoch": 8.18, "learning_rate": 5.463489037178266e-07, "loss": 1.7282, "step": 42948 }, { "epoch": 8.18, "learning_rate": 5.462917063870353e-07, "loss": 0.9918, "step": 42949 }, { "epoch": 8.18, "learning_rate": 5.46234509056244e-07, "loss": 1.6192, "step": 42950 }, { "epoch": 8.18, "learning_rate": 5.461773117254528e-07, "loss": 1.5454, "step": 42951 }, { "epoch": 8.18, "learning_rate": 5.461201143946616e-07, "loss": 2.462, "step": 42952 }, { "epoch": 8.18, "learning_rate": 5.460629170638703e-07, "loss": 1.395, "step": 42953 }, { "epoch": 8.18, "learning_rate": 5.460057197330792e-07, "loss": 2.0131, "step": 42954 }, { "epoch": 8.18, "learning_rate": 5.459485224022879e-07, "loss": 1.7004, "step": 42955 }, { "epoch": 8.18, "learning_rate": 5.458913250714967e-07, "loss": 1.2771, "step": 42956 }, { "epoch": 8.18, "learning_rate": 5.458341277407055e-07, "loss": 0.8562, "step": 42957 }, { "epoch": 8.18, "learning_rate": 5.457769304099142e-07, "loss": 2.1466, "step": 42958 }, { "epoch": 8.18, "learning_rate": 5.45719733079123e-07, "loss": 0.6045, "step": 42959 }, { "epoch": 8.18, "learning_rate": 5.456625357483318e-07, "loss": 1.3854, "step": 42960 }, { "epoch": 8.18, "learning_rate": 5.456053384175406e-07, "loss": 1.4707, "step": 42961 }, { "epoch": 8.18, "learning_rate": 5.455481410867493e-07, "loss": 1.0726, "step": 42962 }, { "epoch": 8.18, "learning_rate": 5.45490943755958e-07, "loss": 2.2714, "step": 42963 }, { "epoch": 8.18, "learning_rate": 5.454337464251668e-07, "loss": 1.6249, "step": 42964 }, { "epoch": 8.18, "learning_rate": 5.453765490943756e-07, "loss": 3.1993, "step": 42965 }, { "epoch": 8.18, "learning_rate": 5.453193517635844e-07, "loss": 1.1428, "step": 42966 }, { "epoch": 8.18, "learning_rate": 5.452621544327931e-07, "loss": 0.8539, "step": 42967 }, { "epoch": 8.18, "learning_rate": 5.452049571020019e-07, "loss": 2.2255, "step": 42968 }, { "epoch": 8.18, "learning_rate": 5.451477597712107e-07, "loss": 2.5279, "step": 42969 }, { "epoch": 8.18, "learning_rate": 5.450905624404195e-07, "loss": 1.7376, "step": 42970 }, { "epoch": 8.18, "learning_rate": 5.450333651096283e-07, "loss": 1.4982, "step": 42971 }, { "epoch": 8.19, "learning_rate": 5.44976167778837e-07, "loss": 0.8424, "step": 42972 }, { "epoch": 8.19, "learning_rate": 5.449189704480458e-07, "loss": 2.076, "step": 42973 }, { "epoch": 8.19, "learning_rate": 5.448617731172546e-07, "loss": 2.3715, "step": 42974 }, { "epoch": 8.19, "learning_rate": 5.448045757864632e-07, "loss": 1.0796, "step": 42975 }, { "epoch": 8.19, "learning_rate": 5.447473784556721e-07, "loss": 0.6811, "step": 42976 }, { "epoch": 8.19, "learning_rate": 5.446901811248808e-07, "loss": 1.0295, "step": 42977 }, { "epoch": 8.19, "learning_rate": 5.446329837940896e-07, "loss": 1.6491, "step": 42978 }, { "epoch": 8.19, "learning_rate": 5.445757864632984e-07, "loss": 0.4554, "step": 42979 }, { "epoch": 8.19, "learning_rate": 5.445185891325071e-07, "loss": 2.0694, "step": 42980 }, { "epoch": 8.19, "learning_rate": 5.444613918017159e-07, "loss": 2.9628, "step": 42981 }, { "epoch": 8.19, "learning_rate": 5.444041944709247e-07, "loss": 1.0472, "step": 42982 }, { "epoch": 8.19, "learning_rate": 5.443469971401335e-07, "loss": 0.6608, "step": 42983 }, { "epoch": 8.19, "learning_rate": 5.442897998093423e-07, "loss": 1.1512, "step": 42984 }, { "epoch": 8.19, "learning_rate": 5.442326024785511e-07, "loss": 0.9881, "step": 42985 }, { "epoch": 8.19, "learning_rate": 5.441754051477598e-07, "loss": 0.9104, "step": 42986 }, { "epoch": 8.19, "learning_rate": 5.441182078169686e-07, "loss": 0.8928, "step": 42987 }, { "epoch": 8.19, "learning_rate": 5.440610104861774e-07, "loss": 2.3044, "step": 42988 }, { "epoch": 8.19, "learning_rate": 5.44003813155386e-07, "loss": 0.7805, "step": 42989 }, { "epoch": 8.19, "learning_rate": 5.439466158245949e-07, "loss": 1.7906, "step": 42990 }, { "epoch": 8.19, "learning_rate": 5.438894184938036e-07, "loss": 3.0155, "step": 42991 }, { "epoch": 8.19, "learning_rate": 5.438322211630124e-07, "loss": 1.2654, "step": 42992 }, { "epoch": 8.19, "learning_rate": 5.437750238322212e-07, "loss": 2.0316, "step": 42993 }, { "epoch": 8.19, "learning_rate": 5.437178265014299e-07, "loss": 1.2333, "step": 42994 }, { "epoch": 8.19, "learning_rate": 5.436606291706387e-07, "loss": 0.8892, "step": 42995 }, { "epoch": 8.19, "learning_rate": 5.436034318398475e-07, "loss": 0.8115, "step": 42996 }, { "epoch": 8.19, "learning_rate": 5.435462345090562e-07, "loss": 3.2267, "step": 42997 }, { "epoch": 8.19, "learning_rate": 5.434890371782651e-07, "loss": 1.1339, "step": 42998 }, { "epoch": 8.19, "learning_rate": 5.434318398474738e-07, "loss": 2.4141, "step": 42999 }, { "epoch": 8.19, "learning_rate": 5.433746425166826e-07, "loss": 2.0034, "step": 43000 }, { "epoch": 8.19, "learning_rate": 5.433174451858914e-07, "loss": 0.759, "step": 43001 }, { "epoch": 8.19, "learning_rate": 5.432602478551001e-07, "loss": 1.5237, "step": 43002 }, { "epoch": 8.19, "learning_rate": 5.432030505243088e-07, "loss": 0.365, "step": 43003 }, { "epoch": 8.19, "learning_rate": 5.431458531935176e-07, "loss": 1.6203, "step": 43004 }, { "epoch": 8.19, "learning_rate": 5.430886558627264e-07, "loss": 0.3351, "step": 43005 }, { "epoch": 8.19, "learning_rate": 5.430314585319352e-07, "loss": 1.5389, "step": 43006 }, { "epoch": 8.19, "learning_rate": 5.42974261201144e-07, "loss": 0.8631, "step": 43007 }, { "epoch": 8.19, "learning_rate": 5.429170638703527e-07, "loss": 1.8143, "step": 43008 }, { "epoch": 8.19, "learning_rate": 5.428598665395615e-07, "loss": 1.6815, "step": 43009 }, { "epoch": 8.19, "learning_rate": 5.428026692087703e-07, "loss": 2.099, "step": 43010 }, { "epoch": 8.19, "learning_rate": 5.42745471877979e-07, "loss": 2.2475, "step": 43011 }, { "epoch": 8.19, "learning_rate": 5.426882745471879e-07, "loss": 1.3709, "step": 43012 }, { "epoch": 8.19, "learning_rate": 5.426310772163966e-07, "loss": 2.1813, "step": 43013 }, { "epoch": 8.19, "learning_rate": 5.425738798856054e-07, "loss": 1.0481, "step": 43014 }, { "epoch": 8.19, "learning_rate": 5.425166825548141e-07, "loss": 1.0812, "step": 43015 }, { "epoch": 8.19, "learning_rate": 5.424594852240228e-07, "loss": 1.0146, "step": 43016 }, { "epoch": 8.19, "learning_rate": 5.424022878932316e-07, "loss": 1.7148, "step": 43017 }, { "epoch": 8.19, "learning_rate": 5.423450905624404e-07, "loss": 2.2335, "step": 43018 }, { "epoch": 8.19, "learning_rate": 5.422878932316492e-07, "loss": 1.8419, "step": 43019 }, { "epoch": 8.19, "learning_rate": 5.42230695900858e-07, "loss": 1.5974, "step": 43020 }, { "epoch": 8.19, "learning_rate": 5.421734985700667e-07, "loss": 0.2437, "step": 43021 }, { "epoch": 8.19, "learning_rate": 5.421163012392755e-07, "loss": 0.9131, "step": 43022 }, { "epoch": 8.19, "learning_rate": 5.420591039084843e-07, "loss": 1.6184, "step": 43023 }, { "epoch": 8.2, "learning_rate": 5.420019065776931e-07, "loss": 1.3436, "step": 43024 }, { "epoch": 8.2, "learning_rate": 5.419447092469018e-07, "loss": 1.3778, "step": 43025 }, { "epoch": 8.2, "learning_rate": 5.418875119161107e-07, "loss": 1.5976, "step": 43026 }, { "epoch": 8.2, "learning_rate": 5.418303145853194e-07, "loss": 1.6842, "step": 43027 }, { "epoch": 8.2, "learning_rate": 5.417731172545282e-07, "loss": 2.9705, "step": 43028 }, { "epoch": 8.2, "learning_rate": 5.417159199237369e-07, "loss": 0.9056, "step": 43029 }, { "epoch": 8.2, "learning_rate": 5.416587225929456e-07, "loss": 2.4797, "step": 43030 }, { "epoch": 8.2, "learning_rate": 5.416015252621544e-07, "loss": 1.7023, "step": 43031 }, { "epoch": 8.2, "learning_rate": 5.415443279313632e-07, "loss": 0.942, "step": 43032 }, { "epoch": 8.2, "learning_rate": 5.414871306005719e-07, "loss": 1.5095, "step": 43033 }, { "epoch": 8.2, "learning_rate": 5.414299332697808e-07, "loss": 2.174, "step": 43034 }, { "epoch": 8.2, "learning_rate": 5.413727359389895e-07, "loss": 0.9519, "step": 43035 }, { "epoch": 8.2, "learning_rate": 5.413155386081983e-07, "loss": 3.0115, "step": 43036 }, { "epoch": 8.2, "learning_rate": 5.412583412774071e-07, "loss": 1.1913, "step": 43037 }, { "epoch": 8.2, "learning_rate": 5.412011439466159e-07, "loss": 0.9138, "step": 43038 }, { "epoch": 8.2, "learning_rate": 5.411439466158246e-07, "loss": 1.3084, "step": 43039 }, { "epoch": 8.2, "learning_rate": 5.410867492850334e-07, "loss": 2.08, "step": 43040 }, { "epoch": 8.2, "learning_rate": 5.410295519542422e-07, "loss": 3.1608, "step": 43041 }, { "epoch": 8.2, "learning_rate": 5.409723546234509e-07, "loss": 1.5487, "step": 43042 }, { "epoch": 8.2, "learning_rate": 5.409151572926597e-07, "loss": 1.8304, "step": 43043 }, { "epoch": 8.2, "learning_rate": 5.408579599618684e-07, "loss": 1.2011, "step": 43044 }, { "epoch": 8.2, "learning_rate": 5.408007626310772e-07, "loss": 1.2072, "step": 43045 }, { "epoch": 8.2, "learning_rate": 5.40743565300286e-07, "loss": 2.0404, "step": 43046 }, { "epoch": 8.2, "learning_rate": 5.406863679694947e-07, "loss": 0.482, "step": 43047 }, { "epoch": 8.2, "learning_rate": 5.406291706387036e-07, "loss": 1.7271, "step": 43048 }, { "epoch": 8.2, "learning_rate": 5.405719733079123e-07, "loss": 1.4803, "step": 43049 }, { "epoch": 8.2, "learning_rate": 5.405147759771211e-07, "loss": 0.2159, "step": 43050 }, { "epoch": 8.2, "learning_rate": 5.404575786463299e-07, "loss": 1.5159, "step": 43051 }, { "epoch": 8.2, "learning_rate": 5.404003813155386e-07, "loss": 2.4697, "step": 43052 }, { "epoch": 8.2, "learning_rate": 5.403431839847474e-07, "loss": 0.6977, "step": 43053 }, { "epoch": 8.2, "learning_rate": 5.402859866539562e-07, "loss": 1.5158, "step": 43054 }, { "epoch": 8.2, "learning_rate": 5.40228789323165e-07, "loss": 1.1666, "step": 43055 }, { "epoch": 8.2, "learning_rate": 5.401715919923737e-07, "loss": 1.1472, "step": 43056 }, { "epoch": 8.2, "learning_rate": 5.401143946615824e-07, "loss": 1.5626, "step": 43057 }, { "epoch": 8.2, "learning_rate": 5.400571973307912e-07, "loss": 1.5314, "step": 43058 }, { "epoch": 8.2, "learning_rate": 5.4e-07, "loss": 1.6658, "step": 43059 }, { "epoch": 8.2, "learning_rate": 5.399428026692088e-07, "loss": 0.9572, "step": 43060 }, { "epoch": 8.2, "learning_rate": 5.398856053384175e-07, "loss": 3.1356, "step": 43061 }, { "epoch": 8.2, "learning_rate": 5.398284080076263e-07, "loss": 1.9203, "step": 43062 }, { "epoch": 8.2, "learning_rate": 5.397712106768351e-07, "loss": 1.2709, "step": 43063 }, { "epoch": 8.2, "learning_rate": 5.397140133460439e-07, "loss": 1.293, "step": 43064 }, { "epoch": 8.2, "learning_rate": 5.396568160152527e-07, "loss": 0.892, "step": 43065 }, { "epoch": 8.2, "learning_rate": 5.395996186844614e-07, "loss": 0.5165, "step": 43066 }, { "epoch": 8.2, "learning_rate": 5.395424213536703e-07, "loss": 0.4632, "step": 43067 }, { "epoch": 8.2, "learning_rate": 5.39485224022879e-07, "loss": 1.7563, "step": 43068 }, { "epoch": 8.2, "learning_rate": 5.394280266920876e-07, "loss": 2.3084, "step": 43069 }, { "epoch": 8.2, "learning_rate": 5.393708293612965e-07, "loss": 2.2123, "step": 43070 }, { "epoch": 8.2, "learning_rate": 5.393136320305052e-07, "loss": 2.927, "step": 43071 }, { "epoch": 8.2, "learning_rate": 5.39256434699714e-07, "loss": 2.1932, "step": 43072 }, { "epoch": 8.2, "learning_rate": 5.391992373689228e-07, "loss": 1.8539, "step": 43073 }, { "epoch": 8.2, "learning_rate": 5.391420400381315e-07, "loss": 0.9252, "step": 43074 }, { "epoch": 8.2, "learning_rate": 5.390848427073403e-07, "loss": 2.1515, "step": 43075 }, { "epoch": 8.2, "learning_rate": 5.390276453765491e-07, "loss": 1.2894, "step": 43076 }, { "epoch": 8.21, "learning_rate": 5.389704480457579e-07, "loss": 1.942, "step": 43077 }, { "epoch": 8.21, "learning_rate": 5.389132507149667e-07, "loss": 1.0076, "step": 43078 }, { "epoch": 8.21, "learning_rate": 5.388560533841755e-07, "loss": 0.9292, "step": 43079 }, { "epoch": 8.21, "learning_rate": 5.387988560533842e-07, "loss": 1.2887, "step": 43080 }, { "epoch": 8.21, "learning_rate": 5.38741658722593e-07, "loss": 0.8301, "step": 43081 }, { "epoch": 8.21, "learning_rate": 5.386844613918017e-07, "loss": 2.2577, "step": 43082 }, { "epoch": 8.21, "learning_rate": 5.386272640610104e-07, "loss": 1.2389, "step": 43083 }, { "epoch": 8.21, "learning_rate": 5.385700667302193e-07, "loss": 1.0052, "step": 43084 }, { "epoch": 8.21, "learning_rate": 5.38512869399428e-07, "loss": 0.8854, "step": 43085 }, { "epoch": 8.21, "learning_rate": 5.384556720686368e-07, "loss": 2.389, "step": 43086 }, { "epoch": 8.21, "learning_rate": 5.383984747378456e-07, "loss": 1.8612, "step": 43087 }, { "epoch": 8.21, "learning_rate": 5.383412774070543e-07, "loss": 1.1015, "step": 43088 }, { "epoch": 8.21, "learning_rate": 5.382840800762632e-07, "loss": 1.2488, "step": 43089 }, { "epoch": 8.21, "learning_rate": 5.382268827454719e-07, "loss": 2.0919, "step": 43090 }, { "epoch": 8.21, "learning_rate": 5.381696854146807e-07, "loss": 1.5622, "step": 43091 }, { "epoch": 8.21, "learning_rate": 5.381124880838895e-07, "loss": 2.2214, "step": 43092 }, { "epoch": 8.21, "learning_rate": 5.380552907530982e-07, "loss": 2.761, "step": 43093 }, { "epoch": 8.21, "learning_rate": 5.37998093422307e-07, "loss": 1.4335, "step": 43094 }, { "epoch": 8.21, "learning_rate": 5.379408960915157e-07, "loss": 2.3348, "step": 43095 }, { "epoch": 8.21, "learning_rate": 5.378836987607245e-07, "loss": 1.5115, "step": 43096 }, { "epoch": 8.21, "learning_rate": 5.378265014299332e-07, "loss": 1.7271, "step": 43097 }, { "epoch": 8.21, "learning_rate": 5.37769304099142e-07, "loss": 1.4529, "step": 43098 }, { "epoch": 8.21, "learning_rate": 5.377121067683508e-07, "loss": 2.1783, "step": 43099 }, { "epoch": 8.21, "learning_rate": 5.376549094375596e-07, "loss": 0.579, "step": 43100 }, { "epoch": 8.21, "learning_rate": 5.375977121067684e-07, "loss": 1.3684, "step": 43101 }, { "epoch": 8.21, "learning_rate": 5.375405147759771e-07, "loss": 2.0942, "step": 43102 }, { "epoch": 8.21, "learning_rate": 5.37483317445186e-07, "loss": 2.4923, "step": 43103 }, { "epoch": 8.21, "learning_rate": 5.374261201143947e-07, "loss": 2.1108, "step": 43104 }, { "epoch": 8.21, "learning_rate": 5.373689227836034e-07, "loss": 1.4965, "step": 43105 }, { "epoch": 8.21, "learning_rate": 5.373117254528123e-07, "loss": 2.4334, "step": 43106 }, { "epoch": 8.21, "learning_rate": 5.37254528122021e-07, "loss": 1.8732, "step": 43107 }, { "epoch": 8.21, "learning_rate": 5.371973307912298e-07, "loss": 1.2642, "step": 43108 }, { "epoch": 8.21, "learning_rate": 5.371401334604385e-07, "loss": 3.0049, "step": 43109 }, { "epoch": 8.21, "learning_rate": 5.370829361296472e-07, "loss": 0.8165, "step": 43110 }, { "epoch": 8.21, "learning_rate": 5.370257387988561e-07, "loss": 2.0724, "step": 43111 }, { "epoch": 8.21, "learning_rate": 5.369685414680648e-07, "loss": 1.1315, "step": 43112 }, { "epoch": 8.21, "learning_rate": 5.369113441372736e-07, "loss": 0.7231, "step": 43113 }, { "epoch": 8.21, "learning_rate": 5.368541468064824e-07, "loss": 0.3322, "step": 43114 }, { "epoch": 8.21, "learning_rate": 5.367969494756911e-07, "loss": 1.2759, "step": 43115 }, { "epoch": 8.21, "learning_rate": 5.367397521448999e-07, "loss": 0.8861, "step": 43116 }, { "epoch": 8.21, "learning_rate": 5.366825548141087e-07, "loss": 1.4659, "step": 43117 }, { "epoch": 8.21, "learning_rate": 5.366253574833175e-07, "loss": 1.6142, "step": 43118 }, { "epoch": 8.21, "learning_rate": 5.365681601525262e-07, "loss": 1.1549, "step": 43119 }, { "epoch": 8.21, "learning_rate": 5.365109628217351e-07, "loss": 2.8663, "step": 43120 }, { "epoch": 8.21, "learning_rate": 5.364537654909438e-07, "loss": 1.4035, "step": 43121 }, { "epoch": 8.21, "learning_rate": 5.363965681601525e-07, "loss": 1.8328, "step": 43122 }, { "epoch": 8.21, "learning_rate": 5.363393708293613e-07, "loss": 2.2033, "step": 43123 }, { "epoch": 8.21, "learning_rate": 5.3628217349857e-07, "loss": 1.5866, "step": 43124 }, { "epoch": 8.21, "learning_rate": 5.362249761677789e-07, "loss": 1.7042, "step": 43125 }, { "epoch": 8.21, "learning_rate": 5.361677788369876e-07, "loss": 1.0352, "step": 43126 }, { "epoch": 8.21, "learning_rate": 5.361105815061963e-07, "loss": 3.2435, "step": 43127 }, { "epoch": 8.21, "learning_rate": 5.360533841754052e-07, "loss": 1.1814, "step": 43128 }, { "epoch": 8.22, "learning_rate": 5.359961868446139e-07, "loss": 1.489, "step": 43129 }, { "epoch": 8.22, "learning_rate": 5.359389895138227e-07, "loss": 1.9852, "step": 43130 }, { "epoch": 8.22, "learning_rate": 5.358817921830315e-07, "loss": 0.5922, "step": 43131 }, { "epoch": 8.22, "learning_rate": 5.358245948522403e-07, "loss": 2.177, "step": 43132 }, { "epoch": 8.22, "learning_rate": 5.357673975214491e-07, "loss": 3.3197, "step": 43133 }, { "epoch": 8.22, "learning_rate": 5.357102001906578e-07, "loss": 0.9227, "step": 43134 }, { "epoch": 8.22, "learning_rate": 5.356530028598665e-07, "loss": 2.4177, "step": 43135 }, { "epoch": 8.22, "learning_rate": 5.355958055290753e-07, "loss": 0.5945, "step": 43136 }, { "epoch": 8.22, "learning_rate": 5.355386081982841e-07, "loss": 1.1218, "step": 43137 }, { "epoch": 8.22, "learning_rate": 5.354814108674928e-07, "loss": 2.1233, "step": 43138 }, { "epoch": 8.22, "learning_rate": 5.354242135367016e-07, "loss": 1.334, "step": 43139 }, { "epoch": 8.22, "learning_rate": 5.353670162059104e-07, "loss": 2.003, "step": 43140 }, { "epoch": 8.22, "learning_rate": 5.353098188751191e-07, "loss": 2.4832, "step": 43141 }, { "epoch": 8.22, "learning_rate": 5.35252621544328e-07, "loss": 1.7489, "step": 43142 }, { "epoch": 8.22, "learning_rate": 5.351954242135367e-07, "loss": 1.6204, "step": 43143 }, { "epoch": 8.22, "learning_rate": 5.351382268827455e-07, "loss": 2.3608, "step": 43144 }, { "epoch": 8.22, "learning_rate": 5.350810295519543e-07, "loss": 0.7505, "step": 43145 }, { "epoch": 8.22, "learning_rate": 5.35023832221163e-07, "loss": 2.4301, "step": 43146 }, { "epoch": 8.22, "learning_rate": 5.349666348903719e-07, "loss": 1.1214, "step": 43147 }, { "epoch": 8.22, "learning_rate": 5.349094375595806e-07, "loss": 1.6861, "step": 43148 }, { "epoch": 8.22, "learning_rate": 5.348522402287893e-07, "loss": 2.5733, "step": 43149 }, { "epoch": 8.22, "learning_rate": 5.347950428979981e-07, "loss": 1.4285, "step": 43150 }, { "epoch": 8.22, "learning_rate": 5.347378455672068e-07, "loss": 1.0382, "step": 43151 }, { "epoch": 8.22, "learning_rate": 5.346806482364156e-07, "loss": 0.4756, "step": 43152 }, { "epoch": 8.22, "learning_rate": 5.346234509056244e-07, "loss": 1.5893, "step": 43153 }, { "epoch": 8.22, "learning_rate": 5.345662535748332e-07, "loss": 1.1332, "step": 43154 }, { "epoch": 8.22, "learning_rate": 5.34509056244042e-07, "loss": 0.8713, "step": 43155 }, { "epoch": 8.22, "learning_rate": 5.344518589132508e-07, "loss": 0.6952, "step": 43156 }, { "epoch": 8.22, "learning_rate": 5.343946615824595e-07, "loss": 1.3186, "step": 43157 }, { "epoch": 8.22, "learning_rate": 5.343374642516683e-07, "loss": 0.785, "step": 43158 }, { "epoch": 8.22, "learning_rate": 5.342802669208771e-07, "loss": 2.3328, "step": 43159 }, { "epoch": 8.22, "learning_rate": 5.342230695900858e-07, "loss": 0.8088, "step": 43160 }, { "epoch": 8.22, "learning_rate": 5.341658722592947e-07, "loss": 1.7443, "step": 43161 }, { "epoch": 8.22, "learning_rate": 5.341086749285033e-07, "loss": 1.6662, "step": 43162 }, { "epoch": 8.22, "learning_rate": 5.34051477597712e-07, "loss": 2.0804, "step": 43163 }, { "epoch": 8.22, "learning_rate": 5.339942802669209e-07, "loss": 1.9111, "step": 43164 }, { "epoch": 8.22, "learning_rate": 5.339370829361296e-07, "loss": 1.0437, "step": 43165 }, { "epoch": 8.22, "learning_rate": 5.338798856053384e-07, "loss": 2.2775, "step": 43166 }, { "epoch": 8.22, "learning_rate": 5.338226882745472e-07, "loss": 2.8237, "step": 43167 }, { "epoch": 8.22, "learning_rate": 5.33765490943756e-07, "loss": 2.5346, "step": 43168 }, { "epoch": 8.22, "learning_rate": 5.337082936129648e-07, "loss": 1.5564, "step": 43169 }, { "epoch": 8.22, "learning_rate": 5.336510962821735e-07, "loss": 1.3962, "step": 43170 }, { "epoch": 8.22, "learning_rate": 5.335938989513823e-07, "loss": 0.97, "step": 43171 }, { "epoch": 8.22, "learning_rate": 5.335367016205911e-07, "loss": 0.8549, "step": 43172 }, { "epoch": 8.22, "learning_rate": 5.334795042897999e-07, "loss": 1.6547, "step": 43173 }, { "epoch": 8.22, "learning_rate": 5.334223069590086e-07, "loss": 0.685, "step": 43174 }, { "epoch": 8.22, "learning_rate": 5.333651096282173e-07, "loss": 2.3667, "step": 43175 }, { "epoch": 8.22, "learning_rate": 5.333079122974261e-07, "loss": 0.8312, "step": 43176 }, { "epoch": 8.22, "learning_rate": 5.332507149666349e-07, "loss": 2.2409, "step": 43177 }, { "epoch": 8.22, "learning_rate": 5.331935176358437e-07, "loss": 1.4568, "step": 43178 }, { "epoch": 8.22, "learning_rate": 5.331363203050524e-07, "loss": 2.5042, "step": 43179 }, { "epoch": 8.22, "learning_rate": 5.330791229742612e-07, "loss": 2.195, "step": 43180 }, { "epoch": 8.22, "learning_rate": 5.3302192564347e-07, "loss": 0.4387, "step": 43181 }, { "epoch": 8.23, "learning_rate": 5.329647283126787e-07, "loss": 0.9499, "step": 43182 }, { "epoch": 8.23, "learning_rate": 5.329075309818876e-07, "loss": 1.5747, "step": 43183 }, { "epoch": 8.23, "learning_rate": 5.328503336510963e-07, "loss": 3.6388, "step": 43184 }, { "epoch": 8.23, "learning_rate": 5.327931363203051e-07, "loss": 1.4887, "step": 43185 }, { "epoch": 8.23, "learning_rate": 5.327359389895139e-07, "loss": 1.8836, "step": 43186 }, { "epoch": 8.23, "learning_rate": 5.326787416587226e-07, "loss": 2.1145, "step": 43187 }, { "epoch": 8.23, "learning_rate": 5.326215443279313e-07, "loss": 1.327, "step": 43188 }, { "epoch": 8.23, "learning_rate": 5.325643469971401e-07, "loss": 1.4948, "step": 43189 }, { "epoch": 8.23, "learning_rate": 5.325071496663489e-07, "loss": 1.6996, "step": 43190 }, { "epoch": 8.23, "learning_rate": 5.324499523355577e-07, "loss": 0.7315, "step": 43191 }, { "epoch": 8.23, "learning_rate": 5.323927550047664e-07, "loss": 1.1844, "step": 43192 }, { "epoch": 8.23, "learning_rate": 5.323355576739752e-07, "loss": 1.2675, "step": 43193 }, { "epoch": 8.23, "learning_rate": 5.32278360343184e-07, "loss": 0.4773, "step": 43194 }, { "epoch": 8.23, "learning_rate": 5.322211630123928e-07, "loss": 1.0747, "step": 43195 }, { "epoch": 8.23, "learning_rate": 5.321639656816015e-07, "loss": 2.8206, "step": 43196 }, { "epoch": 8.23, "learning_rate": 5.321067683508104e-07, "loss": 1.8188, "step": 43197 }, { "epoch": 8.23, "learning_rate": 5.320495710200191e-07, "loss": 2.1667, "step": 43198 }, { "epoch": 8.23, "learning_rate": 5.319923736892279e-07, "loss": 1.7578, "step": 43199 }, { "epoch": 8.23, "learning_rate": 5.319351763584367e-07, "loss": 3.6163, "step": 43200 }, { "epoch": 8.23, "learning_rate": 5.318779790276454e-07, "loss": 2.7069, "step": 43201 }, { "epoch": 8.23, "learning_rate": 5.318207816968541e-07, "loss": 1.8878, "step": 43202 }, { "epoch": 8.23, "learning_rate": 5.317635843660629e-07, "loss": 1.1839, "step": 43203 }, { "epoch": 8.23, "learning_rate": 5.317063870352716e-07, "loss": 1.8526, "step": 43204 }, { "epoch": 8.23, "learning_rate": 5.316491897044805e-07, "loss": 1.7042, "step": 43205 }, { "epoch": 8.23, "learning_rate": 5.315919923736892e-07, "loss": 0.8079, "step": 43206 }, { "epoch": 8.23, "learning_rate": 5.31534795042898e-07, "loss": 1.8632, "step": 43207 }, { "epoch": 8.23, "learning_rate": 5.314775977121068e-07, "loss": 1.838, "step": 43208 }, { "epoch": 8.23, "learning_rate": 5.314204003813156e-07, "loss": 0.8897, "step": 43209 }, { "epoch": 8.23, "learning_rate": 5.313632030505243e-07, "loss": 1.823, "step": 43210 }, { "epoch": 8.23, "learning_rate": 5.313060057197331e-07, "loss": 2.2536, "step": 43211 }, { "epoch": 8.23, "learning_rate": 5.312488083889419e-07, "loss": 2.6771, "step": 43212 }, { "epoch": 8.23, "learning_rate": 5.311916110581507e-07, "loss": 1.2311, "step": 43213 }, { "epoch": 8.23, "learning_rate": 5.311344137273595e-07, "loss": 0.7656, "step": 43214 }, { "epoch": 8.23, "learning_rate": 5.310772163965681e-07, "loss": 2.504, "step": 43215 }, { "epoch": 8.23, "learning_rate": 5.310200190657769e-07, "loss": 2.1488, "step": 43216 }, { "epoch": 8.23, "learning_rate": 5.309628217349857e-07, "loss": 1.54, "step": 43217 }, { "epoch": 8.23, "learning_rate": 5.309056244041944e-07, "loss": 2.2043, "step": 43218 }, { "epoch": 8.23, "learning_rate": 5.308484270734033e-07, "loss": 1.9276, "step": 43219 }, { "epoch": 8.23, "learning_rate": 5.30791229742612e-07, "loss": 1.5363, "step": 43220 }, { "epoch": 8.23, "learning_rate": 5.307340324118208e-07, "loss": 0.6814, "step": 43221 }, { "epoch": 8.23, "learning_rate": 5.306768350810296e-07, "loss": 2.4132, "step": 43222 }, { "epoch": 8.23, "learning_rate": 5.306196377502383e-07, "loss": 0.689, "step": 43223 }, { "epoch": 8.23, "learning_rate": 5.305624404194471e-07, "loss": 1.1974, "step": 43224 }, { "epoch": 8.23, "learning_rate": 5.305052430886559e-07, "loss": 2.9714, "step": 43225 }, { "epoch": 8.23, "learning_rate": 5.304480457578647e-07, "loss": 1.1187, "step": 43226 }, { "epoch": 8.23, "learning_rate": 5.303908484270735e-07, "loss": 0.9253, "step": 43227 }, { "epoch": 8.23, "learning_rate": 5.303336510962821e-07, "loss": 2.6629, "step": 43228 }, { "epoch": 8.23, "learning_rate": 5.302764537654909e-07, "loss": 0.3781, "step": 43229 }, { "epoch": 8.23, "learning_rate": 5.302192564346997e-07, "loss": 1.0591, "step": 43230 }, { "epoch": 8.23, "learning_rate": 5.301620591039085e-07, "loss": 1.5182, "step": 43231 }, { "epoch": 8.23, "learning_rate": 5.301048617731172e-07, "loss": 0.6601, "step": 43232 }, { "epoch": 8.23, "learning_rate": 5.300476644423261e-07, "loss": 0.6434, "step": 43233 }, { "epoch": 8.24, "learning_rate": 5.299904671115348e-07, "loss": 1.8659, "step": 43234 }, { "epoch": 8.24, "learning_rate": 5.299332697807436e-07, "loss": 1.5435, "step": 43235 }, { "epoch": 8.24, "learning_rate": 5.298760724499524e-07, "loss": 1.2572, "step": 43236 }, { "epoch": 8.24, "learning_rate": 5.298188751191611e-07, "loss": 1.9842, "step": 43237 }, { "epoch": 8.24, "learning_rate": 5.297616777883699e-07, "loss": 0.6164, "step": 43238 }, { "epoch": 8.24, "learning_rate": 5.297044804575787e-07, "loss": 1.0508, "step": 43239 }, { "epoch": 8.24, "learning_rate": 5.296472831267874e-07, "loss": 1.9439, "step": 43240 }, { "epoch": 8.24, "learning_rate": 5.295900857959963e-07, "loss": 2.5666, "step": 43241 }, { "epoch": 8.24, "learning_rate": 5.295328884652049e-07, "loss": 1.5481, "step": 43242 }, { "epoch": 8.24, "learning_rate": 5.294756911344137e-07, "loss": 1.5439, "step": 43243 }, { "epoch": 8.24, "learning_rate": 5.294184938036225e-07, "loss": 1.7877, "step": 43244 }, { "epoch": 8.24, "learning_rate": 5.293612964728312e-07, "loss": 1.6216, "step": 43245 }, { "epoch": 8.24, "learning_rate": 5.2930409914204e-07, "loss": 2.1814, "step": 43246 }, { "epoch": 8.24, "learning_rate": 5.292469018112488e-07, "loss": 1.836, "step": 43247 }, { "epoch": 8.24, "learning_rate": 5.291897044804576e-07, "loss": 2.0183, "step": 43248 }, { "epoch": 8.24, "learning_rate": 5.291325071496664e-07, "loss": 2.0732, "step": 43249 }, { "epoch": 8.24, "learning_rate": 5.290753098188752e-07, "loss": 1.6208, "step": 43250 }, { "epoch": 8.24, "learning_rate": 5.290181124880839e-07, "loss": 1.1573, "step": 43251 }, { "epoch": 8.24, "learning_rate": 5.289609151572927e-07, "loss": 1.7368, "step": 43252 }, { "epoch": 8.24, "learning_rate": 5.289037178265015e-07, "loss": 3.1805, "step": 43253 }, { "epoch": 8.24, "learning_rate": 5.288465204957102e-07, "loss": 1.0754, "step": 43254 }, { "epoch": 8.24, "learning_rate": 5.287893231649191e-07, "loss": 3.4582, "step": 43255 }, { "epoch": 8.24, "learning_rate": 5.287321258341277e-07, "loss": 0.5822, "step": 43256 }, { "epoch": 8.24, "learning_rate": 5.286749285033365e-07, "loss": 1.6724, "step": 43257 }, { "epoch": 8.24, "learning_rate": 5.286177311725453e-07, "loss": 2.9915, "step": 43258 }, { "epoch": 8.24, "learning_rate": 5.28560533841754e-07, "loss": 1.7545, "step": 43259 }, { "epoch": 8.24, "learning_rate": 5.285033365109628e-07, "loss": 1.7728, "step": 43260 }, { "epoch": 8.24, "learning_rate": 5.284461391801716e-07, "loss": 1.7011, "step": 43261 }, { "epoch": 8.24, "learning_rate": 5.283889418493804e-07, "loss": 2.9017, "step": 43262 }, { "epoch": 8.24, "learning_rate": 5.283317445185892e-07, "loss": 1.2827, "step": 43263 }, { "epoch": 8.24, "learning_rate": 5.282745471877979e-07, "loss": 1.0842, "step": 43264 }, { "epoch": 8.24, "learning_rate": 5.282173498570067e-07, "loss": 2.1405, "step": 43265 }, { "epoch": 8.24, "learning_rate": 5.281601525262155e-07, "loss": 1.4538, "step": 43266 }, { "epoch": 8.24, "learning_rate": 5.281029551954243e-07, "loss": 0.7576, "step": 43267 }, { "epoch": 8.24, "learning_rate": 5.28045757864633e-07, "loss": 1.0971, "step": 43268 }, { "epoch": 8.24, "learning_rate": 5.279885605338417e-07, "loss": 1.2402, "step": 43269 }, { "epoch": 8.24, "learning_rate": 5.279313632030505e-07, "loss": 1.7687, "step": 43270 }, { "epoch": 8.24, "learning_rate": 5.278741658722593e-07, "loss": 2.1892, "step": 43271 }, { "epoch": 8.24, "learning_rate": 5.278169685414681e-07, "loss": 1.0119, "step": 43272 }, { "epoch": 8.24, "learning_rate": 5.277597712106768e-07, "loss": 0.5367, "step": 43273 }, { "epoch": 8.24, "learning_rate": 5.277025738798856e-07, "loss": 1.5681, "step": 43274 }, { "epoch": 8.24, "learning_rate": 5.276453765490944e-07, "loss": 0.807, "step": 43275 }, { "epoch": 8.24, "learning_rate": 5.275881792183031e-07, "loss": 2.395, "step": 43276 }, { "epoch": 8.24, "learning_rate": 5.27530981887512e-07, "loss": 2.1526, "step": 43277 }, { "epoch": 8.24, "learning_rate": 5.274737845567207e-07, "loss": 2.1311, "step": 43278 }, { "epoch": 8.24, "learning_rate": 5.274165872259295e-07, "loss": 1.9883, "step": 43279 }, { "epoch": 8.24, "learning_rate": 5.273593898951383e-07, "loss": 0.7999, "step": 43280 }, { "epoch": 8.24, "learning_rate": 5.27302192564347e-07, "loss": 1.7865, "step": 43281 }, { "epoch": 8.24, "learning_rate": 5.272449952335557e-07, "loss": 1.2745, "step": 43282 }, { "epoch": 8.24, "learning_rate": 5.271877979027645e-07, "loss": 1.9823, "step": 43283 }, { "epoch": 8.24, "learning_rate": 5.271306005719733e-07, "loss": 1.6027, "step": 43284 }, { "epoch": 8.24, "learning_rate": 5.270734032411821e-07, "loss": 1.3223, "step": 43285 }, { "epoch": 8.24, "learning_rate": 5.270162059103909e-07, "loss": 0.881, "step": 43286 }, { "epoch": 8.25, "learning_rate": 5.269590085795996e-07, "loss": 2.2805, "step": 43287 }, { "epoch": 8.25, "learning_rate": 5.269018112488084e-07, "loss": 1.5339, "step": 43288 }, { "epoch": 8.25, "learning_rate": 5.268446139180172e-07, "loss": 1.873, "step": 43289 }, { "epoch": 8.25, "learning_rate": 5.267874165872259e-07, "loss": 2.0027, "step": 43290 }, { "epoch": 8.25, "learning_rate": 5.267302192564348e-07, "loss": 0.9932, "step": 43291 }, { "epoch": 8.25, "learning_rate": 5.266730219256435e-07, "loss": 2.1038, "step": 43292 }, { "epoch": 8.25, "learning_rate": 5.266158245948523e-07, "loss": 0.5501, "step": 43293 }, { "epoch": 8.25, "learning_rate": 5.265586272640611e-07, "loss": 1.8886, "step": 43294 }, { "epoch": 8.25, "learning_rate": 5.265014299332697e-07, "loss": 2.4628, "step": 43295 }, { "epoch": 8.25, "learning_rate": 5.264442326024785e-07, "loss": 0.4491, "step": 43296 }, { "epoch": 8.25, "learning_rate": 5.263870352716873e-07, "loss": 0.8571, "step": 43297 }, { "epoch": 8.25, "learning_rate": 5.26329837940896e-07, "loss": 0.3293, "step": 43298 }, { "epoch": 8.25, "learning_rate": 5.262726406101049e-07, "loss": 0.5111, "step": 43299 }, { "epoch": 8.25, "learning_rate": 5.262154432793136e-07, "loss": 1.4723, "step": 43300 }, { "epoch": 8.25, "learning_rate": 5.261582459485224e-07, "loss": 0.5209, "step": 43301 }, { "epoch": 8.25, "learning_rate": 5.261010486177312e-07, "loss": 2.2924, "step": 43302 }, { "epoch": 8.25, "learning_rate": 5.2604385128694e-07, "loss": 1.0196, "step": 43303 }, { "epoch": 8.25, "learning_rate": 5.259866539561487e-07, "loss": 0.8173, "step": 43304 }, { "epoch": 8.25, "learning_rate": 5.259294566253575e-07, "loss": 2.9156, "step": 43305 }, { "epoch": 8.25, "learning_rate": 5.258722592945663e-07, "loss": 1.1772, "step": 43306 }, { "epoch": 8.25, "learning_rate": 5.258150619637751e-07, "loss": 1.9279, "step": 43307 }, { "epoch": 8.25, "learning_rate": 5.257578646329839e-07, "loss": 1.2391, "step": 43308 }, { "epoch": 8.25, "learning_rate": 5.257006673021925e-07, "loss": 2.0377, "step": 43309 }, { "epoch": 8.25, "learning_rate": 5.256434699714013e-07, "loss": 1.5299, "step": 43310 }, { "epoch": 8.25, "learning_rate": 5.255862726406101e-07, "loss": 1.6286, "step": 43311 }, { "epoch": 8.25, "learning_rate": 5.255290753098188e-07, "loss": 0.2633, "step": 43312 }, { "epoch": 8.25, "learning_rate": 5.254718779790277e-07, "loss": 2.0749, "step": 43313 }, { "epoch": 8.25, "learning_rate": 5.254146806482364e-07, "loss": 0.3659, "step": 43314 }, { "epoch": 8.25, "learning_rate": 5.253574833174452e-07, "loss": 1.7654, "step": 43315 }, { "epoch": 8.25, "learning_rate": 5.25300285986654e-07, "loss": 2.0963, "step": 43316 }, { "epoch": 8.25, "learning_rate": 5.252430886558627e-07, "loss": 0.464, "step": 43317 }, { "epoch": 8.25, "learning_rate": 5.251858913250715e-07, "loss": 2.0227, "step": 43318 }, { "epoch": 8.25, "learning_rate": 5.251286939942803e-07, "loss": 1.6397, "step": 43319 }, { "epoch": 8.25, "learning_rate": 5.250714966634891e-07, "loss": 2.0773, "step": 43320 }, { "epoch": 8.25, "learning_rate": 5.250142993326979e-07, "loss": 1.6822, "step": 43321 }, { "epoch": 8.25, "learning_rate": 5.249571020019065e-07, "loss": 1.8235, "step": 43322 }, { "epoch": 8.25, "learning_rate": 5.248999046711153e-07, "loss": 1.2814, "step": 43323 }, { "epoch": 8.25, "learning_rate": 5.248427073403241e-07, "loss": 0.671, "step": 43324 }, { "epoch": 8.25, "learning_rate": 5.247855100095329e-07, "loss": 0.3915, "step": 43325 }, { "epoch": 8.25, "learning_rate": 5.247283126787416e-07, "loss": 1.9818, "step": 43326 }, { "epoch": 8.25, "learning_rate": 5.246711153479505e-07, "loss": 2.1297, "step": 43327 }, { "epoch": 8.25, "learning_rate": 5.246139180171592e-07, "loss": 3.0761, "step": 43328 }, { "epoch": 8.25, "learning_rate": 5.24556720686368e-07, "loss": 2.5905, "step": 43329 }, { "epoch": 8.25, "learning_rate": 5.244995233555768e-07, "loss": 2.8059, "step": 43330 }, { "epoch": 8.25, "learning_rate": 5.244423260247855e-07, "loss": 2.004, "step": 43331 }, { "epoch": 8.25, "learning_rate": 5.243851286939944e-07, "loss": 0.1147, "step": 43332 }, { "epoch": 8.25, "learning_rate": 5.243279313632031e-07, "loss": 1.6291, "step": 43333 }, { "epoch": 8.25, "learning_rate": 5.242707340324118e-07, "loss": 2.0672, "step": 43334 }, { "epoch": 8.25, "learning_rate": 5.242135367016207e-07, "loss": 1.11, "step": 43335 }, { "epoch": 8.25, "learning_rate": 5.241563393708293e-07, "loss": 1.6328, "step": 43336 }, { "epoch": 8.25, "learning_rate": 5.240991420400381e-07, "loss": 1.7185, "step": 43337 }, { "epoch": 8.25, "learning_rate": 5.240419447092469e-07, "loss": 1.0945, "step": 43338 }, { "epoch": 8.26, "learning_rate": 5.239847473784557e-07, "loss": 1.0291, "step": 43339 }, { "epoch": 8.26, "learning_rate": 5.239275500476644e-07, "loss": 1.4376, "step": 43340 }, { "epoch": 8.26, "learning_rate": 5.238703527168732e-07, "loss": 2.1889, "step": 43341 }, { "epoch": 8.26, "learning_rate": 5.23813155386082e-07, "loss": 2.2631, "step": 43342 }, { "epoch": 8.26, "learning_rate": 5.237559580552908e-07, "loss": 1.655, "step": 43343 }, { "epoch": 8.26, "learning_rate": 5.236987607244996e-07, "loss": 1.9311, "step": 43344 }, { "epoch": 8.26, "learning_rate": 5.236415633937083e-07, "loss": 1.2502, "step": 43345 }, { "epoch": 8.26, "learning_rate": 5.235843660629171e-07, "loss": 0.9713, "step": 43346 }, { "epoch": 8.26, "learning_rate": 5.235271687321259e-07, "loss": 0.3558, "step": 43347 }, { "epoch": 8.26, "learning_rate": 5.234699714013345e-07, "loss": 1.9393, "step": 43348 }, { "epoch": 8.26, "learning_rate": 5.234127740705434e-07, "loss": 2.4434, "step": 43349 }, { "epoch": 8.26, "learning_rate": 5.233555767397521e-07, "loss": 0.9518, "step": 43350 }, { "epoch": 8.26, "learning_rate": 5.232983794089609e-07, "loss": 1.7612, "step": 43351 }, { "epoch": 8.26, "learning_rate": 5.232411820781697e-07, "loss": 1.7627, "step": 43352 }, { "epoch": 8.26, "learning_rate": 5.231839847473784e-07, "loss": 2.047, "step": 43353 }, { "epoch": 8.26, "learning_rate": 5.231267874165873e-07, "loss": 2.5148, "step": 43354 }, { "epoch": 8.26, "learning_rate": 5.23069590085796e-07, "loss": 1.6336, "step": 43355 }, { "epoch": 8.26, "learning_rate": 5.230123927550048e-07, "loss": 1.1694, "step": 43356 }, { "epoch": 8.26, "learning_rate": 5.229551954242136e-07, "loss": 2.8759, "step": 43357 }, { "epoch": 8.26, "learning_rate": 5.228979980934223e-07, "loss": 1.1624, "step": 43358 }, { "epoch": 8.26, "learning_rate": 5.228408007626311e-07, "loss": 1.9118, "step": 43359 }, { "epoch": 8.26, "learning_rate": 5.227836034318399e-07, "loss": 1.1368, "step": 43360 }, { "epoch": 8.26, "learning_rate": 5.227264061010487e-07, "loss": 2.3743, "step": 43361 }, { "epoch": 8.26, "learning_rate": 5.226692087702573e-07, "loss": 1.0187, "step": 43362 }, { "epoch": 8.26, "learning_rate": 5.226120114394661e-07, "loss": 0.8594, "step": 43363 }, { "epoch": 8.26, "learning_rate": 5.225548141086749e-07, "loss": 1.8222, "step": 43364 }, { "epoch": 8.26, "learning_rate": 5.224976167778837e-07, "loss": 0.8471, "step": 43365 }, { "epoch": 8.26, "learning_rate": 5.224404194470925e-07, "loss": 0.5742, "step": 43366 }, { "epoch": 8.26, "learning_rate": 5.223832221163012e-07, "loss": 1.3838, "step": 43367 }, { "epoch": 8.26, "learning_rate": 5.223260247855101e-07, "loss": 0.6141, "step": 43368 }, { "epoch": 8.26, "learning_rate": 5.222688274547188e-07, "loss": 1.4156, "step": 43369 }, { "epoch": 8.26, "learning_rate": 5.222116301239275e-07, "loss": 1.3083, "step": 43370 }, { "epoch": 8.26, "learning_rate": 5.221544327931364e-07, "loss": 2.4809, "step": 43371 }, { "epoch": 8.26, "learning_rate": 5.220972354623451e-07, "loss": 1.2314, "step": 43372 }, { "epoch": 8.26, "learning_rate": 5.220400381315539e-07, "loss": 1.3777, "step": 43373 }, { "epoch": 8.26, "learning_rate": 5.219828408007627e-07, "loss": 1.921, "step": 43374 }, { "epoch": 8.26, "learning_rate": 5.219256434699713e-07, "loss": 1.3592, "step": 43375 }, { "epoch": 8.26, "learning_rate": 5.218684461391802e-07, "loss": 2.5792, "step": 43376 }, { "epoch": 8.26, "learning_rate": 5.218112488083889e-07, "loss": 1.6278, "step": 43377 }, { "epoch": 8.26, "learning_rate": 5.217540514775977e-07, "loss": 1.1441, "step": 43378 }, { "epoch": 8.26, "learning_rate": 5.216968541468065e-07, "loss": 1.1029, "step": 43379 }, { "epoch": 8.26, "learning_rate": 5.216396568160153e-07, "loss": 2.8935, "step": 43380 }, { "epoch": 8.26, "learning_rate": 5.21582459485224e-07, "loss": 1.451, "step": 43381 }, { "epoch": 8.26, "learning_rate": 5.215252621544328e-07, "loss": 1.2703, "step": 43382 }, { "epoch": 8.26, "learning_rate": 5.214680648236416e-07, "loss": 2.0409, "step": 43383 }, { "epoch": 8.26, "learning_rate": 5.214108674928503e-07, "loss": 1.4195, "step": 43384 }, { "epoch": 8.26, "learning_rate": 5.213536701620592e-07, "loss": 2.8432, "step": 43385 }, { "epoch": 8.26, "learning_rate": 5.212964728312679e-07, "loss": 2.0633, "step": 43386 }, { "epoch": 8.26, "learning_rate": 5.212392755004767e-07, "loss": 1.4226, "step": 43387 }, { "epoch": 8.26, "learning_rate": 5.211820781696855e-07, "loss": 1.7101, "step": 43388 }, { "epoch": 8.26, "learning_rate": 5.211248808388941e-07, "loss": 1.4637, "step": 43389 }, { "epoch": 8.26, "learning_rate": 5.21067683508103e-07, "loss": 1.4538, "step": 43390 }, { "epoch": 8.26, "learning_rate": 5.210104861773117e-07, "loss": 1.8765, "step": 43391 }, { "epoch": 8.27, "learning_rate": 5.209532888465205e-07, "loss": 1.3667, "step": 43392 }, { "epoch": 8.27, "learning_rate": 5.208960915157293e-07, "loss": 0.9936, "step": 43393 }, { "epoch": 8.27, "learning_rate": 5.20838894184938e-07, "loss": 2.1545, "step": 43394 }, { "epoch": 8.27, "learning_rate": 5.207816968541468e-07, "loss": 1.1735, "step": 43395 }, { "epoch": 8.27, "learning_rate": 5.207244995233556e-07, "loss": 1.5024, "step": 43396 }, { "epoch": 8.27, "learning_rate": 5.206673021925644e-07, "loss": 1.6444, "step": 43397 }, { "epoch": 8.27, "learning_rate": 5.206101048617732e-07, "loss": 1.1896, "step": 43398 }, { "epoch": 8.27, "learning_rate": 5.205529075309819e-07, "loss": 0.7758, "step": 43399 }, { "epoch": 8.27, "learning_rate": 5.204957102001907e-07, "loss": 1.7782, "step": 43400 }, { "epoch": 8.27, "learning_rate": 5.204385128693995e-07, "loss": 1.9879, "step": 43401 }, { "epoch": 8.27, "learning_rate": 5.203813155386083e-07, "loss": 0.7495, "step": 43402 }, { "epoch": 8.27, "learning_rate": 5.203241182078169e-07, "loss": 0.8699, "step": 43403 }, { "epoch": 8.27, "learning_rate": 5.202669208770258e-07, "loss": 2.0817, "step": 43404 }, { "epoch": 8.27, "learning_rate": 5.202097235462345e-07, "loss": 1.6705, "step": 43405 }, { "epoch": 8.27, "learning_rate": 5.201525262154432e-07, "loss": 0.5602, "step": 43406 }, { "epoch": 8.27, "learning_rate": 5.200953288846521e-07, "loss": 2.0993, "step": 43407 }, { "epoch": 8.27, "learning_rate": 5.200381315538608e-07, "loss": 0.5104, "step": 43408 }, { "epoch": 8.27, "learning_rate": 5.199809342230696e-07, "loss": 0.8953, "step": 43409 }, { "epoch": 8.27, "learning_rate": 5.199237368922784e-07, "loss": 1.2857, "step": 43410 }, { "epoch": 8.27, "learning_rate": 5.198665395614871e-07, "loss": 1.6964, "step": 43411 }, { "epoch": 8.27, "learning_rate": 5.19809342230696e-07, "loss": 1.3967, "step": 43412 }, { "epoch": 8.27, "learning_rate": 5.197521448999047e-07, "loss": 2.5328, "step": 43413 }, { "epoch": 8.27, "learning_rate": 5.196949475691135e-07, "loss": 1.4764, "step": 43414 }, { "epoch": 8.27, "learning_rate": 5.196377502383223e-07, "loss": 1.4359, "step": 43415 }, { "epoch": 8.27, "learning_rate": 5.195805529075309e-07, "loss": 2.0274, "step": 43416 }, { "epoch": 8.27, "learning_rate": 5.195233555767397e-07, "loss": 1.4078, "step": 43417 }, { "epoch": 8.27, "learning_rate": 5.194661582459485e-07, "loss": 2.1108, "step": 43418 }, { "epoch": 8.27, "learning_rate": 5.194089609151573e-07, "loss": 1.1981, "step": 43419 }, { "epoch": 8.27, "learning_rate": 5.19351763584366e-07, "loss": 3.951, "step": 43420 }, { "epoch": 8.27, "learning_rate": 5.192945662535749e-07, "loss": 1.4861, "step": 43421 }, { "epoch": 8.27, "learning_rate": 5.192373689227836e-07, "loss": 0.8584, "step": 43422 }, { "epoch": 8.27, "learning_rate": 5.191801715919924e-07, "loss": 1.932, "step": 43423 }, { "epoch": 8.27, "learning_rate": 5.191229742612012e-07, "loss": 2.0079, "step": 43424 }, { "epoch": 8.27, "learning_rate": 5.190657769304099e-07, "loss": 0.9746, "step": 43425 }, { "epoch": 8.27, "learning_rate": 5.190085795996188e-07, "loss": 1.58, "step": 43426 }, { "epoch": 8.27, "learning_rate": 5.189513822688275e-07, "loss": 0.89, "step": 43427 }, { "epoch": 8.27, "learning_rate": 5.188941849380361e-07, "loss": 2.5518, "step": 43428 }, { "epoch": 8.27, "learning_rate": 5.18836987607245e-07, "loss": 2.0251, "step": 43429 }, { "epoch": 8.27, "learning_rate": 5.187797902764537e-07, "loss": 0.6975, "step": 43430 }, { "epoch": 8.27, "learning_rate": 5.187225929456625e-07, "loss": 2.6865, "step": 43431 }, { "epoch": 8.27, "learning_rate": 5.186653956148713e-07, "loss": 1.3931, "step": 43432 }, { "epoch": 8.27, "learning_rate": 5.186081982840801e-07, "loss": 0.6787, "step": 43433 }, { "epoch": 8.27, "learning_rate": 5.185510009532889e-07, "loss": 2.307, "step": 43434 }, { "epoch": 8.27, "learning_rate": 5.184938036224976e-07, "loss": 1.4496, "step": 43435 }, { "epoch": 8.27, "learning_rate": 5.184366062917064e-07, "loss": 1.5983, "step": 43436 }, { "epoch": 8.27, "learning_rate": 5.183794089609152e-07, "loss": 2.5038, "step": 43437 }, { "epoch": 8.27, "learning_rate": 5.18322211630124e-07, "loss": 1.4671, "step": 43438 }, { "epoch": 8.27, "learning_rate": 5.182650142993327e-07, "loss": 1.2064, "step": 43439 }, { "epoch": 8.27, "learning_rate": 5.182078169685415e-07, "loss": 0.9898, "step": 43440 }, { "epoch": 8.27, "learning_rate": 5.181506196377503e-07, "loss": 1.1674, "step": 43441 }, { "epoch": 8.27, "learning_rate": 5.180934223069589e-07, "loss": 0.6626, "step": 43442 }, { "epoch": 8.27, "learning_rate": 5.180362249761678e-07, "loss": 1.076, "step": 43443 }, { "epoch": 8.28, "learning_rate": 5.179790276453765e-07, "loss": 0.9965, "step": 43444 }, { "epoch": 8.28, "learning_rate": 5.179218303145853e-07, "loss": 1.1067, "step": 43445 }, { "epoch": 8.28, "learning_rate": 5.178646329837941e-07, "loss": 0.6461, "step": 43446 }, { "epoch": 8.28, "learning_rate": 5.178074356530028e-07, "loss": 1.4187, "step": 43447 }, { "epoch": 8.28, "learning_rate": 5.177502383222117e-07, "loss": 1.9727, "step": 43448 }, { "epoch": 8.28, "learning_rate": 5.176930409914204e-07, "loss": 2.7136, "step": 43449 }, { "epoch": 8.28, "learning_rate": 5.176358436606292e-07, "loss": 1.3454, "step": 43450 }, { "epoch": 8.28, "learning_rate": 5.17578646329838e-07, "loss": 1.1882, "step": 43451 }, { "epoch": 8.28, "learning_rate": 5.175214489990467e-07, "loss": 2.008, "step": 43452 }, { "epoch": 8.28, "learning_rate": 5.174642516682555e-07, "loss": 1.7927, "step": 43453 }, { "epoch": 8.28, "learning_rate": 5.174070543374643e-07, "loss": 1.5399, "step": 43454 }, { "epoch": 8.28, "learning_rate": 5.17349857006673e-07, "loss": 4.0302, "step": 43455 }, { "epoch": 8.28, "learning_rate": 5.172926596758818e-07, "loss": 2.2349, "step": 43456 }, { "epoch": 8.28, "learning_rate": 5.172354623450906e-07, "loss": 1.0861, "step": 43457 }, { "epoch": 8.28, "learning_rate": 5.171782650142993e-07, "loss": 1.9213, "step": 43458 }, { "epoch": 8.28, "learning_rate": 5.171210676835081e-07, "loss": 1.1372, "step": 43459 }, { "epoch": 8.28, "learning_rate": 5.170638703527169e-07, "loss": 1.7275, "step": 43460 }, { "epoch": 8.28, "learning_rate": 5.170066730219256e-07, "loss": 1.7664, "step": 43461 }, { "epoch": 8.28, "learning_rate": 5.169494756911345e-07, "loss": 1.0406, "step": 43462 }, { "epoch": 8.28, "learning_rate": 5.168922783603432e-07, "loss": 1.4568, "step": 43463 }, { "epoch": 8.28, "learning_rate": 5.168350810295519e-07, "loss": 1.0758, "step": 43464 }, { "epoch": 8.28, "learning_rate": 5.167778836987608e-07, "loss": 0.2308, "step": 43465 }, { "epoch": 8.28, "learning_rate": 5.167206863679695e-07, "loss": 2.6763, "step": 43466 }, { "epoch": 8.28, "learning_rate": 5.166634890371783e-07, "loss": 0.9925, "step": 43467 }, { "epoch": 8.28, "learning_rate": 5.166062917063871e-07, "loss": 0.9048, "step": 43468 }, { "epoch": 8.28, "learning_rate": 5.165490943755958e-07, "loss": 0.4874, "step": 43469 }, { "epoch": 8.28, "learning_rate": 5.164918970448046e-07, "loss": 1.5647, "step": 43470 }, { "epoch": 8.28, "learning_rate": 5.164346997140133e-07, "loss": 2.2971, "step": 43471 }, { "epoch": 8.28, "learning_rate": 5.163775023832221e-07, "loss": 2.2466, "step": 43472 }, { "epoch": 8.28, "learning_rate": 5.163203050524309e-07, "loss": 2.2713, "step": 43473 }, { "epoch": 8.28, "learning_rate": 5.162631077216397e-07, "loss": 1.8829, "step": 43474 }, { "epoch": 8.28, "learning_rate": 5.162059103908484e-07, "loss": 1.6067, "step": 43475 }, { "epoch": 8.28, "learning_rate": 5.161487130600572e-07, "loss": 0.97, "step": 43476 }, { "epoch": 8.28, "learning_rate": 5.16091515729266e-07, "loss": 1.3047, "step": 43477 }, { "epoch": 8.28, "learning_rate": 5.160343183984748e-07, "loss": 2.0063, "step": 43478 }, { "epoch": 8.28, "learning_rate": 5.159771210676836e-07, "loss": 1.1572, "step": 43479 }, { "epoch": 8.28, "learning_rate": 5.159199237368923e-07, "loss": 0.9906, "step": 43480 }, { "epoch": 8.28, "learning_rate": 5.158627264061011e-07, "loss": 3.9369, "step": 43481 }, { "epoch": 8.28, "learning_rate": 5.158055290753099e-07, "loss": 0.6903, "step": 43482 }, { "epoch": 8.28, "learning_rate": 5.157483317445185e-07, "loss": 2.2521, "step": 43483 }, { "epoch": 8.28, "learning_rate": 5.156911344137274e-07, "loss": 1.0382, "step": 43484 }, { "epoch": 8.28, "learning_rate": 5.156339370829361e-07, "loss": 2.6004, "step": 43485 }, { "epoch": 8.28, "learning_rate": 5.155767397521449e-07, "loss": 2.3645, "step": 43486 }, { "epoch": 8.28, "learning_rate": 5.155195424213537e-07, "loss": 1.449, "step": 43487 }, { "epoch": 8.28, "learning_rate": 5.154623450905624e-07, "loss": 0.5709, "step": 43488 }, { "epoch": 8.28, "learning_rate": 5.154051477597712e-07, "loss": 2.9701, "step": 43489 }, { "epoch": 8.28, "learning_rate": 5.1534795042898e-07, "loss": 2.2181, "step": 43490 }, { "epoch": 8.28, "learning_rate": 5.152907530981888e-07, "loss": 1.3541, "step": 43491 }, { "epoch": 8.28, "learning_rate": 5.152335557673976e-07, "loss": 2.0617, "step": 43492 }, { "epoch": 8.28, "learning_rate": 5.151763584366063e-07, "loss": 2.8523, "step": 43493 }, { "epoch": 8.28, "learning_rate": 5.151191611058151e-07, "loss": 1.4403, "step": 43494 }, { "epoch": 8.28, "learning_rate": 5.150619637750239e-07, "loss": 0.2919, "step": 43495 }, { "epoch": 8.28, "learning_rate": 5.150047664442326e-07, "loss": 0.9491, "step": 43496 }, { "epoch": 8.29, "learning_rate": 5.149475691134413e-07, "loss": 1.0166, "step": 43497 }, { "epoch": 8.29, "learning_rate": 5.148903717826502e-07, "loss": 0.4817, "step": 43498 }, { "epoch": 8.29, "learning_rate": 5.148331744518589e-07, "loss": 2.7194, "step": 43499 }, { "epoch": 8.29, "learning_rate": 5.147759771210677e-07, "loss": 1.6633, "step": 43500 }, { "epoch": 8.29, "learning_rate": 5.147187797902765e-07, "loss": 0.8564, "step": 43501 }, { "epoch": 8.29, "learning_rate": 5.146615824594852e-07, "loss": 1.6297, "step": 43502 }, { "epoch": 8.29, "learning_rate": 5.14604385128694e-07, "loss": 1.6011, "step": 43503 }, { "epoch": 8.29, "learning_rate": 5.145471877979028e-07, "loss": 0.7575, "step": 43504 }, { "epoch": 8.29, "learning_rate": 5.144899904671115e-07, "loss": 1.2168, "step": 43505 }, { "epoch": 8.29, "learning_rate": 5.144327931363204e-07, "loss": 1.3511, "step": 43506 }, { "epoch": 8.29, "learning_rate": 5.143755958055291e-07, "loss": 1.2007, "step": 43507 }, { "epoch": 8.29, "learning_rate": 5.143183984747379e-07, "loss": 1.3042, "step": 43508 }, { "epoch": 8.29, "learning_rate": 5.142612011439466e-07, "loss": 1.9766, "step": 43509 }, { "epoch": 8.29, "learning_rate": 5.142040038131554e-07, "loss": 2.1057, "step": 43510 }, { "epoch": 8.29, "learning_rate": 5.141468064823641e-07, "loss": 1.1245, "step": 43511 }, { "epoch": 8.29, "learning_rate": 5.140896091515729e-07, "loss": 1.2534, "step": 43512 }, { "epoch": 8.29, "learning_rate": 5.140324118207817e-07, "loss": 1.1765, "step": 43513 }, { "epoch": 8.29, "learning_rate": 5.139752144899905e-07, "loss": 1.3033, "step": 43514 }, { "epoch": 8.29, "learning_rate": 5.139180171591993e-07, "loss": 1.548, "step": 43515 }, { "epoch": 8.29, "learning_rate": 5.13860819828408e-07, "loss": 3.131, "step": 43516 }, { "epoch": 8.29, "learning_rate": 5.138036224976168e-07, "loss": 1.9114, "step": 43517 }, { "epoch": 8.29, "learning_rate": 5.137464251668256e-07, "loss": 1.0089, "step": 43518 }, { "epoch": 8.29, "learning_rate": 5.136892278360343e-07, "loss": 1.1647, "step": 43519 }, { "epoch": 8.29, "learning_rate": 5.136320305052432e-07, "loss": 1.9376, "step": 43520 }, { "epoch": 8.29, "learning_rate": 5.135748331744519e-07, "loss": 3.2235, "step": 43521 }, { "epoch": 8.29, "learning_rate": 5.135176358436607e-07, "loss": 2.4175, "step": 43522 }, { "epoch": 8.29, "learning_rate": 5.134604385128694e-07, "loss": 1.0826, "step": 43523 }, { "epoch": 8.29, "learning_rate": 5.134032411820781e-07, "loss": 1.1326, "step": 43524 }, { "epoch": 8.29, "learning_rate": 5.133460438512869e-07, "loss": 2.0731, "step": 43525 }, { "epoch": 8.29, "learning_rate": 5.132888465204957e-07, "loss": 1.2524, "step": 43526 }, { "epoch": 8.29, "learning_rate": 5.132316491897045e-07, "loss": 0.6909, "step": 43527 }, { "epoch": 8.29, "learning_rate": 5.131744518589133e-07, "loss": 1.3279, "step": 43528 }, { "epoch": 8.29, "learning_rate": 5.13117254528122e-07, "loss": 1.4141, "step": 43529 }, { "epoch": 8.29, "learning_rate": 5.130600571973308e-07, "loss": 1.1575, "step": 43530 }, { "epoch": 8.29, "learning_rate": 5.130028598665396e-07, "loss": 1.1744, "step": 43531 }, { "epoch": 8.29, "learning_rate": 5.129456625357484e-07, "loss": 2.8652, "step": 43532 }, { "epoch": 8.29, "learning_rate": 5.128884652049571e-07, "loss": 2.9035, "step": 43533 }, { "epoch": 8.29, "learning_rate": 5.12831267874166e-07, "loss": 0.6859, "step": 43534 }, { "epoch": 8.29, "learning_rate": 5.127740705433747e-07, "loss": 2.1861, "step": 43535 }, { "epoch": 8.29, "learning_rate": 5.127168732125834e-07, "loss": 0.9242, "step": 43536 }, { "epoch": 8.29, "learning_rate": 5.126596758817922e-07, "loss": 2.9241, "step": 43537 }, { "epoch": 8.29, "learning_rate": 5.126024785510009e-07, "loss": 2.3249, "step": 43538 }, { "epoch": 8.29, "learning_rate": 5.125452812202097e-07, "loss": 1.9795, "step": 43539 }, { "epoch": 8.29, "learning_rate": 5.124880838894185e-07, "loss": 2.5615, "step": 43540 }, { "epoch": 8.29, "learning_rate": 5.124308865586272e-07, "loss": 1.4933, "step": 43541 }, { "epoch": 8.29, "learning_rate": 5.123736892278361e-07, "loss": 1.1881, "step": 43542 }, { "epoch": 8.29, "learning_rate": 5.123164918970448e-07, "loss": 1.1556, "step": 43543 }, { "epoch": 8.29, "learning_rate": 5.122592945662536e-07, "loss": 1.3622, "step": 43544 }, { "epoch": 8.29, "learning_rate": 5.122020972354624e-07, "loss": 1.346, "step": 43545 }, { "epoch": 8.29, "learning_rate": 5.121448999046711e-07, "loss": 1.54, "step": 43546 }, { "epoch": 8.29, "learning_rate": 5.120877025738799e-07, "loss": 1.944, "step": 43547 }, { "epoch": 8.29, "learning_rate": 5.120305052430887e-07, "loss": 2.5533, "step": 43548 }, { "epoch": 8.3, "learning_rate": 5.119733079122974e-07, "loss": 1.174, "step": 43549 }, { "epoch": 8.3, "learning_rate": 5.119161105815062e-07, "loss": 1.6239, "step": 43550 }, { "epoch": 8.3, "learning_rate": 5.11858913250715e-07, "loss": 1.209, "step": 43551 }, { "epoch": 8.3, "learning_rate": 5.118017159199237e-07, "loss": 0.9023, "step": 43552 }, { "epoch": 8.3, "learning_rate": 5.117445185891325e-07, "loss": 1.8266, "step": 43553 }, { "epoch": 8.3, "learning_rate": 5.116873212583413e-07, "loss": 2.2791, "step": 43554 }, { "epoch": 8.3, "learning_rate": 5.1163012392755e-07, "loss": 2.4717, "step": 43555 }, { "epoch": 8.3, "learning_rate": 5.115729265967589e-07, "loss": 1.9513, "step": 43556 }, { "epoch": 8.3, "learning_rate": 5.115157292659676e-07, "loss": 1.347, "step": 43557 }, { "epoch": 8.3, "learning_rate": 5.114585319351764e-07, "loss": 1.174, "step": 43558 }, { "epoch": 8.3, "learning_rate": 5.114013346043852e-07, "loss": 3.3075, "step": 43559 }, { "epoch": 8.3, "learning_rate": 5.113441372735939e-07, "loss": 2.4945, "step": 43560 }, { "epoch": 8.3, "learning_rate": 5.112869399428027e-07, "loss": 0.4718, "step": 43561 }, { "epoch": 8.3, "learning_rate": 5.112297426120114e-07, "loss": 1.086, "step": 43562 }, { "epoch": 8.3, "learning_rate": 5.111725452812202e-07, "loss": 1.1487, "step": 43563 }, { "epoch": 8.3, "learning_rate": 5.11115347950429e-07, "loss": 0.6754, "step": 43564 }, { "epoch": 8.3, "learning_rate": 5.110581506196377e-07, "loss": 2.8338, "step": 43565 }, { "epoch": 8.3, "learning_rate": 5.110009532888465e-07, "loss": 1.6978, "step": 43566 }, { "epoch": 8.3, "learning_rate": 5.109437559580553e-07, "loss": 2.3984, "step": 43567 }, { "epoch": 8.3, "learning_rate": 5.108865586272641e-07, "loss": 1.2889, "step": 43568 }, { "epoch": 8.3, "learning_rate": 5.108293612964728e-07, "loss": 1.0379, "step": 43569 }, { "epoch": 8.3, "learning_rate": 5.107721639656816e-07, "loss": 1.5031, "step": 43570 }, { "epoch": 8.3, "learning_rate": 5.107149666348904e-07, "loss": 3.0228, "step": 43571 }, { "epoch": 8.3, "learning_rate": 5.106577693040992e-07, "loss": 1.5357, "step": 43572 }, { "epoch": 8.3, "learning_rate": 5.10600571973308e-07, "loss": 2.5794, "step": 43573 }, { "epoch": 8.3, "learning_rate": 5.105433746425167e-07, "loss": 0.673, "step": 43574 }, { "epoch": 8.3, "learning_rate": 5.104861773117256e-07, "loss": 1.8463, "step": 43575 }, { "epoch": 8.3, "learning_rate": 5.104289799809342e-07, "loss": 2.881, "step": 43576 }, { "epoch": 8.3, "learning_rate": 5.103717826501429e-07, "loss": 3.0104, "step": 43577 }, { "epoch": 8.3, "learning_rate": 5.103145853193518e-07, "loss": 1.6119, "step": 43578 }, { "epoch": 8.3, "learning_rate": 5.102573879885605e-07, "loss": 0.3455, "step": 43579 }, { "epoch": 8.3, "learning_rate": 5.102001906577693e-07, "loss": 0.8622, "step": 43580 }, { "epoch": 8.3, "learning_rate": 5.101429933269781e-07, "loss": 1.3206, "step": 43581 }, { "epoch": 8.3, "learning_rate": 5.100857959961868e-07, "loss": 1.3292, "step": 43582 }, { "epoch": 8.3, "learning_rate": 5.100285986653956e-07, "loss": 2.2935, "step": 43583 }, { "epoch": 8.3, "learning_rate": 5.099714013346044e-07, "loss": 1.9368, "step": 43584 }, { "epoch": 8.3, "learning_rate": 5.099142040038132e-07, "loss": 1.0115, "step": 43585 }, { "epoch": 8.3, "learning_rate": 5.09857006673022e-07, "loss": 1.2444, "step": 43586 }, { "epoch": 8.3, "learning_rate": 5.097998093422308e-07, "loss": 0.7195, "step": 43587 }, { "epoch": 8.3, "learning_rate": 5.097426120114395e-07, "loss": 0.697, "step": 43588 }, { "epoch": 8.3, "learning_rate": 5.096854146806482e-07, "loss": 0.9562, "step": 43589 }, { "epoch": 8.3, "learning_rate": 5.09628217349857e-07, "loss": 0.9687, "step": 43590 }, { "epoch": 8.3, "learning_rate": 5.095710200190657e-07, "loss": 1.8318, "step": 43591 }, { "epoch": 8.3, "learning_rate": 5.095138226882746e-07, "loss": 2.2777, "step": 43592 }, { "epoch": 8.3, "learning_rate": 5.094566253574833e-07, "loss": 1.7643, "step": 43593 }, { "epoch": 8.3, "learning_rate": 5.093994280266921e-07, "loss": 1.379, "step": 43594 }, { "epoch": 8.3, "learning_rate": 5.093422306959009e-07, "loss": 0.7766, "step": 43595 }, { "epoch": 8.3, "learning_rate": 5.092850333651096e-07, "loss": 2.2595, "step": 43596 }, { "epoch": 8.3, "learning_rate": 5.092278360343184e-07, "loss": 1.5342, "step": 43597 }, { "epoch": 8.3, "learning_rate": 5.091706387035272e-07, "loss": 0.3626, "step": 43598 }, { "epoch": 8.3, "learning_rate": 5.091134413727359e-07, "loss": 1.0589, "step": 43599 }, { "epoch": 8.3, "learning_rate": 5.090562440419448e-07, "loss": 2.4206, "step": 43600 }, { "epoch": 8.3, "learning_rate": 5.089990467111535e-07, "loss": 2.4877, "step": 43601 }, { "epoch": 8.31, "learning_rate": 5.089418493803622e-07, "loss": 1.0389, "step": 43602 }, { "epoch": 8.31, "learning_rate": 5.08884652049571e-07, "loss": 0.7086, "step": 43603 }, { "epoch": 8.31, "learning_rate": 5.088274547187798e-07, "loss": 1.9982, "step": 43604 }, { "epoch": 8.31, "learning_rate": 5.087702573879885e-07, "loss": 2.7341, "step": 43605 }, { "epoch": 8.31, "learning_rate": 5.087130600571973e-07, "loss": 1.4153, "step": 43606 }, { "epoch": 8.31, "learning_rate": 5.086558627264061e-07, "loss": 2.3259, "step": 43607 }, { "epoch": 8.31, "learning_rate": 5.085986653956149e-07, "loss": 1.2302, "step": 43608 }, { "epoch": 8.31, "learning_rate": 5.085414680648237e-07, "loss": 2.7673, "step": 43609 }, { "epoch": 8.31, "learning_rate": 5.084842707340324e-07, "loss": 2.3531, "step": 43610 }, { "epoch": 8.31, "learning_rate": 5.084270734032412e-07, "loss": 1.8325, "step": 43611 }, { "epoch": 8.31, "learning_rate": 5.0836987607245e-07, "loss": 1.7479, "step": 43612 }, { "epoch": 8.31, "learning_rate": 5.083126787416587e-07, "loss": 2.1303, "step": 43613 }, { "epoch": 8.31, "learning_rate": 5.082554814108676e-07, "loss": 1.1213, "step": 43614 }, { "epoch": 8.31, "learning_rate": 5.081982840800763e-07, "loss": 1.7484, "step": 43615 }, { "epoch": 8.31, "learning_rate": 5.08141086749285e-07, "loss": 2.399, "step": 43616 }, { "epoch": 8.31, "learning_rate": 5.080838894184938e-07, "loss": 1.4721, "step": 43617 }, { "epoch": 8.31, "learning_rate": 5.080266920877025e-07, "loss": 1.6198, "step": 43618 }, { "epoch": 8.31, "learning_rate": 5.079694947569113e-07, "loss": 1.4269, "step": 43619 }, { "epoch": 8.31, "learning_rate": 5.079122974261201e-07, "loss": 0.9003, "step": 43620 }, { "epoch": 8.31, "learning_rate": 5.078551000953289e-07, "loss": 2.1955, "step": 43621 }, { "epoch": 8.31, "learning_rate": 5.077979027645377e-07, "loss": 3.139, "step": 43622 }, { "epoch": 8.31, "learning_rate": 5.077407054337464e-07, "loss": 2.5241, "step": 43623 }, { "epoch": 8.31, "learning_rate": 5.076835081029552e-07, "loss": 1.9998, "step": 43624 }, { "epoch": 8.31, "learning_rate": 5.07626310772164e-07, "loss": 1.8235, "step": 43625 }, { "epoch": 8.31, "learning_rate": 5.075691134413728e-07, "loss": 1.8033, "step": 43626 }, { "epoch": 8.31, "learning_rate": 5.075119161105815e-07, "loss": 2.1182, "step": 43627 }, { "epoch": 8.31, "learning_rate": 5.074547187797904e-07, "loss": 1.4982, "step": 43628 }, { "epoch": 8.31, "learning_rate": 5.07397521448999e-07, "loss": 2.3092, "step": 43629 }, { "epoch": 8.31, "learning_rate": 5.073403241182078e-07, "loss": 1.2304, "step": 43630 }, { "epoch": 8.31, "learning_rate": 5.072831267874166e-07, "loss": 1.4677, "step": 43631 }, { "epoch": 8.31, "learning_rate": 5.072259294566253e-07, "loss": 0.9955, "step": 43632 }, { "epoch": 8.31, "learning_rate": 5.071687321258342e-07, "loss": 1.715, "step": 43633 }, { "epoch": 8.31, "learning_rate": 5.071115347950429e-07, "loss": 1.327, "step": 43634 }, { "epoch": 8.31, "learning_rate": 5.070543374642516e-07, "loss": 1.7689, "step": 43635 }, { "epoch": 8.31, "learning_rate": 5.069971401334605e-07, "loss": 1.8491, "step": 43636 }, { "epoch": 8.31, "learning_rate": 5.069399428026692e-07, "loss": 2.1485, "step": 43637 }, { "epoch": 8.31, "learning_rate": 5.06882745471878e-07, "loss": 2.3746, "step": 43638 }, { "epoch": 8.31, "learning_rate": 5.068255481410868e-07, "loss": 0.9594, "step": 43639 }, { "epoch": 8.31, "learning_rate": 5.067683508102956e-07, "loss": 1.3528, "step": 43640 }, { "epoch": 8.31, "learning_rate": 5.067111534795043e-07, "loss": 2.5651, "step": 43641 }, { "epoch": 8.31, "learning_rate": 5.06653956148713e-07, "loss": 1.741, "step": 43642 }, { "epoch": 8.31, "learning_rate": 5.065967588179218e-07, "loss": 1.6486, "step": 43643 }, { "epoch": 8.31, "learning_rate": 5.065395614871306e-07, "loss": 1.2643, "step": 43644 }, { "epoch": 8.31, "learning_rate": 5.064823641563394e-07, "loss": 2.4659, "step": 43645 }, { "epoch": 8.31, "learning_rate": 5.064251668255481e-07, "loss": 1.5126, "step": 43646 }, { "epoch": 8.31, "learning_rate": 5.063679694947569e-07, "loss": 0.9742, "step": 43647 }, { "epoch": 8.31, "learning_rate": 5.063107721639657e-07, "loss": 2.2295, "step": 43648 }, { "epoch": 8.31, "learning_rate": 5.062535748331744e-07, "loss": 1.5977, "step": 43649 }, { "epoch": 8.31, "learning_rate": 5.061963775023833e-07, "loss": 1.7137, "step": 43650 }, { "epoch": 8.31, "learning_rate": 5.06139180171592e-07, "loss": 0.5834, "step": 43651 }, { "epoch": 8.31, "learning_rate": 5.060819828408008e-07, "loss": 1.6138, "step": 43652 }, { "epoch": 8.31, "learning_rate": 5.060247855100096e-07, "loss": 1.743, "step": 43653 }, { "epoch": 8.32, "learning_rate": 5.059675881792183e-07, "loss": 1.078, "step": 43654 }, { "epoch": 8.32, "learning_rate": 5.059103908484272e-07, "loss": 0.4662, "step": 43655 }, { "epoch": 8.32, "learning_rate": 5.058531935176358e-07, "loss": 1.208, "step": 43656 }, { "epoch": 8.32, "learning_rate": 5.057959961868446e-07, "loss": 0.8873, "step": 43657 }, { "epoch": 8.32, "learning_rate": 5.057387988560534e-07, "loss": 1.9756, "step": 43658 }, { "epoch": 8.32, "learning_rate": 5.056816015252621e-07, "loss": 1.7517, "step": 43659 }, { "epoch": 8.32, "learning_rate": 5.056244041944709e-07, "loss": 1.0113, "step": 43660 }, { "epoch": 8.32, "learning_rate": 5.055672068636797e-07, "loss": 0.9705, "step": 43661 }, { "epoch": 8.32, "learning_rate": 5.055100095328885e-07, "loss": 1.9913, "step": 43662 }, { "epoch": 8.32, "learning_rate": 5.054528122020972e-07, "loss": 1.3329, "step": 43663 }, { "epoch": 8.32, "learning_rate": 5.05395614871306e-07, "loss": 1.2708, "step": 43664 }, { "epoch": 8.32, "learning_rate": 5.053384175405148e-07, "loss": 1.836, "step": 43665 }, { "epoch": 8.32, "learning_rate": 5.052812202097236e-07, "loss": 2.0595, "step": 43666 }, { "epoch": 8.32, "learning_rate": 5.052240228789324e-07, "loss": 1.7035, "step": 43667 }, { "epoch": 8.32, "learning_rate": 5.051668255481411e-07, "loss": 2.1002, "step": 43668 }, { "epoch": 8.32, "learning_rate": 5.0510962821735e-07, "loss": 1.7243, "step": 43669 }, { "epoch": 8.32, "learning_rate": 5.050524308865586e-07, "loss": 2.1073, "step": 43670 }, { "epoch": 8.32, "learning_rate": 5.049952335557673e-07, "loss": 1.3147, "step": 43671 }, { "epoch": 8.32, "learning_rate": 5.049380362249762e-07, "loss": 1.0512, "step": 43672 }, { "epoch": 8.32, "learning_rate": 5.048808388941849e-07, "loss": 1.1641, "step": 43673 }, { "epoch": 8.32, "learning_rate": 5.048236415633937e-07, "loss": 1.5324, "step": 43674 }, { "epoch": 8.32, "learning_rate": 5.047664442326025e-07, "loss": 1.8392, "step": 43675 }, { "epoch": 8.32, "learning_rate": 5.047092469018112e-07, "loss": 0.4945, "step": 43676 }, { "epoch": 8.32, "learning_rate": 5.046520495710201e-07, "loss": 0.8187, "step": 43677 }, { "epoch": 8.32, "learning_rate": 5.045948522402288e-07, "loss": 1.3846, "step": 43678 }, { "epoch": 8.32, "learning_rate": 5.045376549094376e-07, "loss": 1.941, "step": 43679 }, { "epoch": 8.32, "learning_rate": 5.044804575786464e-07, "loss": 1.3946, "step": 43680 }, { "epoch": 8.32, "learning_rate": 5.044232602478552e-07, "loss": 2.1252, "step": 43681 }, { "epoch": 8.32, "learning_rate": 5.043660629170638e-07, "loss": 2.8232, "step": 43682 }, { "epoch": 8.32, "learning_rate": 5.043088655862726e-07, "loss": 1.0417, "step": 43683 }, { "epoch": 8.32, "learning_rate": 5.042516682554814e-07, "loss": 1.0215, "step": 43684 }, { "epoch": 8.32, "learning_rate": 5.041944709246901e-07, "loss": 1.5385, "step": 43685 }, { "epoch": 8.32, "learning_rate": 5.04137273593899e-07, "loss": 3.7081, "step": 43686 }, { "epoch": 8.32, "learning_rate": 5.040800762631077e-07, "loss": 1.5003, "step": 43687 }, { "epoch": 8.32, "learning_rate": 5.040228789323165e-07, "loss": 3.1199, "step": 43688 }, { "epoch": 8.32, "learning_rate": 5.039656816015253e-07, "loss": 0.9812, "step": 43689 }, { "epoch": 8.32, "learning_rate": 5.03908484270734e-07, "loss": 1.8385, "step": 43690 }, { "epoch": 8.32, "learning_rate": 5.038512869399429e-07, "loss": 0.9444, "step": 43691 }, { "epoch": 8.32, "learning_rate": 5.037940896091516e-07, "loss": 2.1729, "step": 43692 }, { "epoch": 8.32, "learning_rate": 5.037368922783604e-07, "loss": 0.5535, "step": 43693 }, { "epoch": 8.32, "learning_rate": 5.036796949475692e-07, "loss": 1.2464, "step": 43694 }, { "epoch": 8.32, "learning_rate": 5.036224976167779e-07, "loss": 1.0136, "step": 43695 }, { "epoch": 8.32, "learning_rate": 5.035653002859866e-07, "loss": 1.7025, "step": 43696 }, { "epoch": 8.32, "learning_rate": 5.035081029551954e-07, "loss": 2.1952, "step": 43697 }, { "epoch": 8.32, "learning_rate": 5.034509056244042e-07, "loss": 0.2054, "step": 43698 }, { "epoch": 8.32, "learning_rate": 5.03393708293613e-07, "loss": 1.5602, "step": 43699 }, { "epoch": 8.32, "learning_rate": 5.033365109628217e-07, "loss": 1.3324, "step": 43700 }, { "epoch": 8.32, "learning_rate": 5.032793136320305e-07, "loss": 1.9309, "step": 43701 }, { "epoch": 8.32, "learning_rate": 5.032221163012393e-07, "loss": 1.4246, "step": 43702 }, { "epoch": 8.32, "learning_rate": 5.031649189704481e-07, "loss": 0.2862, "step": 43703 }, { "epoch": 8.32, "learning_rate": 5.031077216396568e-07, "loss": 2.1514, "step": 43704 }, { "epoch": 8.32, "learning_rate": 5.030505243088657e-07, "loss": 0.3204, "step": 43705 }, { "epoch": 8.32, "learning_rate": 5.029933269780744e-07, "loss": 1.7514, "step": 43706 }, { "epoch": 8.33, "learning_rate": 5.029361296472831e-07, "loss": 1.147, "step": 43707 }, { "epoch": 8.33, "learning_rate": 5.02878932316492e-07, "loss": 1.215, "step": 43708 }, { "epoch": 8.33, "learning_rate": 5.028217349857006e-07, "loss": 0.5188, "step": 43709 }, { "epoch": 8.33, "learning_rate": 5.027645376549094e-07, "loss": 1.5512, "step": 43710 }, { "epoch": 8.33, "learning_rate": 5.027073403241182e-07, "loss": 1.7271, "step": 43711 }, { "epoch": 8.33, "learning_rate": 5.026501429933269e-07, "loss": 1.4609, "step": 43712 }, { "epoch": 8.33, "learning_rate": 5.025929456625358e-07, "loss": 1.034, "step": 43713 }, { "epoch": 8.33, "learning_rate": 5.025357483317445e-07, "loss": 2.5822, "step": 43714 }, { "epoch": 8.33, "learning_rate": 5.024785510009533e-07, "loss": 1.6964, "step": 43715 }, { "epoch": 8.33, "learning_rate": 5.024213536701621e-07, "loss": 2.4316, "step": 43716 }, { "epoch": 8.33, "learning_rate": 5.023641563393709e-07, "loss": 1.613, "step": 43717 }, { "epoch": 8.33, "learning_rate": 5.023069590085796e-07, "loss": 1.2665, "step": 43718 }, { "epoch": 8.33, "learning_rate": 5.022497616777884e-07, "loss": 1.4853, "step": 43719 }, { "epoch": 8.33, "learning_rate": 5.021925643469972e-07, "loss": 1.3571, "step": 43720 }, { "epoch": 8.33, "learning_rate": 5.02135367016206e-07, "loss": 2.5369, "step": 43721 }, { "epoch": 8.33, "learning_rate": 5.020781696854148e-07, "loss": 0.9168, "step": 43722 }, { "epoch": 8.33, "learning_rate": 5.020209723546234e-07, "loss": 2.2551, "step": 43723 }, { "epoch": 8.33, "learning_rate": 5.019637750238322e-07, "loss": 4.6796, "step": 43724 }, { "epoch": 8.33, "learning_rate": 5.01906577693041e-07, "loss": 0.7887, "step": 43725 }, { "epoch": 8.33, "learning_rate": 5.018493803622497e-07, "loss": 0.9372, "step": 43726 }, { "epoch": 8.33, "learning_rate": 5.017921830314586e-07, "loss": 2.9941, "step": 43727 }, { "epoch": 8.33, "learning_rate": 5.017349857006673e-07, "loss": 1.2001, "step": 43728 }, { "epoch": 8.33, "learning_rate": 5.01677788369876e-07, "loss": 1.2393, "step": 43729 }, { "epoch": 8.33, "learning_rate": 5.016205910390849e-07, "loss": 1.0896, "step": 43730 }, { "epoch": 8.33, "learning_rate": 5.015633937082936e-07, "loss": 1.3326, "step": 43731 }, { "epoch": 8.33, "learning_rate": 5.015061963775024e-07, "loss": 1.6275, "step": 43732 }, { "epoch": 8.33, "learning_rate": 5.014489990467112e-07, "loss": 0.736, "step": 43733 }, { "epoch": 8.33, "learning_rate": 5.0139180171592e-07, "loss": 1.6071, "step": 43734 }, { "epoch": 8.33, "learning_rate": 5.013346043851288e-07, "loss": 2.129, "step": 43735 }, { "epoch": 8.33, "learning_rate": 5.012774070543374e-07, "loss": 0.7042, "step": 43736 }, { "epoch": 8.33, "learning_rate": 5.012202097235462e-07, "loss": 2.2788, "step": 43737 }, { "epoch": 8.33, "learning_rate": 5.01163012392755e-07, "loss": 1.3883, "step": 43738 }, { "epoch": 8.33, "learning_rate": 5.011058150619638e-07, "loss": 1.1953, "step": 43739 }, { "epoch": 8.33, "learning_rate": 5.010486177311725e-07, "loss": 0.8939, "step": 43740 }, { "epoch": 8.33, "learning_rate": 5.009914204003813e-07, "loss": 2.339, "step": 43741 }, { "epoch": 8.33, "learning_rate": 5.009342230695901e-07, "loss": 0.5746, "step": 43742 }, { "epoch": 8.33, "learning_rate": 5.008770257387989e-07, "loss": 1.1192, "step": 43743 }, { "epoch": 8.33, "learning_rate": 5.008198284080077e-07, "loss": 0.5315, "step": 43744 }, { "epoch": 8.33, "learning_rate": 5.007626310772164e-07, "loss": 2.2033, "step": 43745 }, { "epoch": 8.33, "learning_rate": 5.007054337464252e-07, "loss": 1.5509, "step": 43746 }, { "epoch": 8.33, "learning_rate": 5.00648236415634e-07, "loss": 1.3266, "step": 43747 }, { "epoch": 8.33, "learning_rate": 5.005910390848427e-07, "loss": 1.1153, "step": 43748 }, { "epoch": 8.33, "learning_rate": 5.005338417540516e-07, "loss": 1.0213, "step": 43749 }, { "epoch": 8.33, "learning_rate": 5.004766444232602e-07, "loss": 1.6702, "step": 43750 }, { "epoch": 8.33, "learning_rate": 5.00419447092469e-07, "loss": 1.2152, "step": 43751 }, { "epoch": 8.33, "learning_rate": 5.003622497616778e-07, "loss": 1.8666, "step": 43752 }, { "epoch": 8.33, "learning_rate": 5.003050524308865e-07, "loss": 2.3098, "step": 43753 }, { "epoch": 8.33, "learning_rate": 5.002478551000953e-07, "loss": 1.6725, "step": 43754 }, { "epoch": 8.33, "learning_rate": 5.001906577693041e-07, "loss": 2.4542, "step": 43755 }, { "epoch": 8.33, "learning_rate": 5.001334604385129e-07, "loss": 1.6257, "step": 43756 }, { "epoch": 8.33, "learning_rate": 5.000762631077217e-07, "loss": 1.1738, "step": 43757 }, { "epoch": 8.33, "learning_rate": 5.000190657769305e-07, "loss": 0.4111, "step": 43758 }, { "epoch": 8.34, "learning_rate": 4.999618684461392e-07, "loss": 1.6742, "step": 43759 }, { "epoch": 8.34, "learning_rate": 4.99904671115348e-07, "loss": 1.1194, "step": 43760 }, { "epoch": 8.34, "learning_rate": 4.998474737845568e-07, "loss": 2.3205, "step": 43761 }, { "epoch": 8.34, "learning_rate": 4.997902764537654e-07, "loss": 1.8446, "step": 43762 }, { "epoch": 8.34, "learning_rate": 4.997330791229743e-07, "loss": 1.0958, "step": 43763 }, { "epoch": 8.34, "learning_rate": 4.99675881792183e-07, "loss": 1.4047, "step": 43764 }, { "epoch": 8.34, "learning_rate": 4.996186844613918e-07, "loss": 1.9385, "step": 43765 }, { "epoch": 8.34, "learning_rate": 4.995614871306006e-07, "loss": 1.1022, "step": 43766 }, { "epoch": 8.34, "learning_rate": 4.995042897998093e-07, "loss": 1.6146, "step": 43767 }, { "epoch": 8.34, "learning_rate": 4.994470924690181e-07, "loss": 0.7495, "step": 43768 }, { "epoch": 8.34, "learning_rate": 4.993898951382269e-07, "loss": 3.6158, "step": 43769 }, { "epoch": 8.34, "learning_rate": 4.993326978074357e-07, "loss": 1.3577, "step": 43770 }, { "epoch": 8.34, "learning_rate": 4.992755004766445e-07, "loss": 2.3292, "step": 43771 }, { "epoch": 8.34, "learning_rate": 4.992183031458532e-07, "loss": 1.0128, "step": 43772 }, { "epoch": 8.34, "learning_rate": 4.99161105815062e-07, "loss": 1.5538, "step": 43773 }, { "epoch": 8.34, "learning_rate": 4.991039084842708e-07, "loss": 1.458, "step": 43774 }, { "epoch": 8.34, "learning_rate": 4.990467111534796e-07, "loss": 1.3797, "step": 43775 }, { "epoch": 8.34, "learning_rate": 4.989895138226882e-07, "loss": 1.0655, "step": 43776 }, { "epoch": 8.34, "learning_rate": 4.98932316491897e-07, "loss": 0.6746, "step": 43777 }, { "epoch": 8.34, "learning_rate": 4.988751191611058e-07, "loss": 1.0118, "step": 43778 }, { "epoch": 8.34, "learning_rate": 4.988179218303146e-07, "loss": 2.3846, "step": 43779 }, { "epoch": 8.34, "learning_rate": 4.987607244995234e-07, "loss": 3.0064, "step": 43780 }, { "epoch": 8.34, "learning_rate": 4.987035271687321e-07, "loss": 1.8586, "step": 43781 }, { "epoch": 8.34, "learning_rate": 4.986463298379409e-07, "loss": 1.4341, "step": 43782 }, { "epoch": 8.34, "learning_rate": 4.985891325071497e-07, "loss": 1.6216, "step": 43783 }, { "epoch": 8.34, "learning_rate": 4.985319351763584e-07, "loss": 2.4303, "step": 43784 }, { "epoch": 8.34, "learning_rate": 4.984747378455673e-07, "loss": 3.0041, "step": 43785 }, { "epoch": 8.34, "learning_rate": 4.98417540514776e-07, "loss": 2.8826, "step": 43786 }, { "epoch": 8.34, "learning_rate": 4.983603431839848e-07, "loss": 0.6549, "step": 43787 }, { "epoch": 8.34, "learning_rate": 4.983031458531936e-07, "loss": 1.6411, "step": 43788 }, { "epoch": 8.34, "learning_rate": 4.982459485224022e-07, "loss": 0.8711, "step": 43789 }, { "epoch": 8.34, "learning_rate": 4.98188751191611e-07, "loss": 2.5831, "step": 43790 }, { "epoch": 8.34, "learning_rate": 4.981315538608198e-07, "loss": 2.606, "step": 43791 }, { "epoch": 8.34, "learning_rate": 4.980743565300286e-07, "loss": 1.7398, "step": 43792 }, { "epoch": 8.34, "learning_rate": 4.980171591992374e-07, "loss": 1.4017, "step": 43793 }, { "epoch": 8.34, "learning_rate": 4.979599618684461e-07, "loss": 1.3372, "step": 43794 }, { "epoch": 8.34, "learning_rate": 4.979027645376549e-07, "loss": 1.2486, "step": 43795 }, { "epoch": 8.34, "learning_rate": 4.978455672068637e-07, "loss": 0.8193, "step": 43796 }, { "epoch": 8.34, "learning_rate": 4.977883698760725e-07, "loss": 2.0778, "step": 43797 }, { "epoch": 8.34, "learning_rate": 4.977311725452812e-07, "loss": 1.8849, "step": 43798 }, { "epoch": 8.34, "learning_rate": 4.976739752144901e-07, "loss": 1.7881, "step": 43799 }, { "epoch": 8.34, "learning_rate": 4.976167778836988e-07, "loss": 2.7694, "step": 43800 }, { "epoch": 8.34, "learning_rate": 4.975595805529076e-07, "loss": 1.7957, "step": 43801 }, { "epoch": 8.34, "learning_rate": 4.975023832221164e-07, "loss": 2.5746, "step": 43802 }, { "epoch": 8.34, "learning_rate": 4.97445185891325e-07, "loss": 1.2515, "step": 43803 }, { "epoch": 8.34, "learning_rate": 4.973879885605338e-07, "loss": 1.4582, "step": 43804 }, { "epoch": 8.34, "learning_rate": 4.973307912297426e-07, "loss": 1.7538, "step": 43805 }, { "epoch": 8.34, "learning_rate": 4.972735938989513e-07, "loss": 2.0144, "step": 43806 }, { "epoch": 8.34, "learning_rate": 4.972163965681602e-07, "loss": 0.6944, "step": 43807 }, { "epoch": 8.34, "learning_rate": 4.971591992373689e-07, "loss": 1.8807, "step": 43808 }, { "epoch": 8.34, "learning_rate": 4.971020019065777e-07, "loss": 1.4196, "step": 43809 }, { "epoch": 8.34, "learning_rate": 4.970448045757865e-07, "loss": 1.6591, "step": 43810 }, { "epoch": 8.34, "learning_rate": 4.969876072449953e-07, "loss": 0.384, "step": 43811 }, { "epoch": 8.35, "learning_rate": 4.96930409914204e-07, "loss": 1.7446, "step": 43812 }, { "epoch": 8.35, "learning_rate": 4.968732125834128e-07, "loss": 0.9571, "step": 43813 }, { "epoch": 8.35, "learning_rate": 4.968160152526216e-07, "loss": 0.9098, "step": 43814 }, { "epoch": 8.35, "learning_rate": 4.967588179218304e-07, "loss": 1.3218, "step": 43815 }, { "epoch": 8.35, "learning_rate": 4.967016205910392e-07, "loss": 0.5468, "step": 43816 }, { "epoch": 8.35, "learning_rate": 4.966444232602478e-07, "loss": 2.7753, "step": 43817 }, { "epoch": 8.35, "learning_rate": 4.965872259294566e-07, "loss": 0.5887, "step": 43818 }, { "epoch": 8.35, "learning_rate": 4.965300285986654e-07, "loss": 0.8618, "step": 43819 }, { "epoch": 8.35, "learning_rate": 4.964728312678741e-07, "loss": 1.5674, "step": 43820 }, { "epoch": 8.35, "learning_rate": 4.96415633937083e-07, "loss": 1.5894, "step": 43821 }, { "epoch": 8.35, "learning_rate": 4.963584366062917e-07, "loss": 1.9912, "step": 43822 }, { "epoch": 8.35, "learning_rate": 4.963012392755005e-07, "loss": 2.2071, "step": 43823 }, { "epoch": 8.35, "learning_rate": 4.962440419447093e-07, "loss": 1.3236, "step": 43824 }, { "epoch": 8.35, "learning_rate": 4.96186844613918e-07, "loss": 2.4229, "step": 43825 }, { "epoch": 8.35, "learning_rate": 4.961296472831268e-07, "loss": 1.5125, "step": 43826 }, { "epoch": 8.35, "learning_rate": 4.960724499523356e-07, "loss": 1.3701, "step": 43827 }, { "epoch": 8.35, "learning_rate": 4.960152526215444e-07, "loss": 0.3837, "step": 43828 }, { "epoch": 8.35, "learning_rate": 4.959580552907532e-07, "loss": 2.4914, "step": 43829 }, { "epoch": 8.35, "learning_rate": 4.959008579599618e-07, "loss": 0.9947, "step": 43830 }, { "epoch": 8.35, "learning_rate": 4.958436606291706e-07, "loss": 1.2222, "step": 43831 }, { "epoch": 8.35, "learning_rate": 4.957864632983794e-07, "loss": 0.8325, "step": 43832 }, { "epoch": 8.35, "learning_rate": 4.957292659675882e-07, "loss": 2.8543, "step": 43833 }, { "epoch": 8.35, "learning_rate": 4.956720686367969e-07, "loss": 2.7297, "step": 43834 }, { "epoch": 8.35, "learning_rate": 4.956148713060058e-07, "loss": 1.0325, "step": 43835 }, { "epoch": 8.35, "learning_rate": 4.955576739752145e-07, "loss": 0.3481, "step": 43836 }, { "epoch": 8.35, "learning_rate": 4.955004766444233e-07, "loss": 2.0553, "step": 43837 }, { "epoch": 8.35, "learning_rate": 4.954432793136321e-07, "loss": 1.1792, "step": 43838 }, { "epoch": 8.35, "learning_rate": 4.953860819828408e-07, "loss": 1.4079, "step": 43839 }, { "epoch": 8.35, "learning_rate": 4.953288846520496e-07, "loss": 1.1366, "step": 43840 }, { "epoch": 8.35, "learning_rate": 4.952716873212584e-07, "loss": 1.653, "step": 43841 }, { "epoch": 8.35, "learning_rate": 4.95214489990467e-07, "loss": 1.6466, "step": 43842 }, { "epoch": 8.35, "learning_rate": 4.95157292659676e-07, "loss": 1.288, "step": 43843 }, { "epoch": 8.35, "learning_rate": 4.951000953288846e-07, "loss": 1.3923, "step": 43844 }, { "epoch": 8.35, "learning_rate": 4.950428979980934e-07, "loss": 1.1633, "step": 43845 }, { "epoch": 8.35, "learning_rate": 4.949857006673022e-07, "loss": 1.6693, "step": 43846 }, { "epoch": 8.35, "learning_rate": 4.949285033365109e-07, "loss": 1.2558, "step": 43847 }, { "epoch": 8.35, "learning_rate": 4.948713060057197e-07, "loss": 0.9988, "step": 43848 }, { "epoch": 8.35, "learning_rate": 4.948141086749285e-07, "loss": 2.2321, "step": 43849 }, { "epoch": 8.35, "learning_rate": 4.947569113441373e-07, "loss": 1.7478, "step": 43850 }, { "epoch": 8.35, "learning_rate": 4.946997140133461e-07, "loss": 2.6883, "step": 43851 }, { "epoch": 8.35, "learning_rate": 4.946425166825549e-07, "loss": 2.1878, "step": 43852 }, { "epoch": 8.35, "learning_rate": 4.945853193517636e-07, "loss": 1.8343, "step": 43853 }, { "epoch": 8.35, "learning_rate": 4.945281220209724e-07, "loss": 1.1929, "step": 43854 }, { "epoch": 8.35, "learning_rate": 4.944709246901812e-07, "loss": 1.1942, "step": 43855 }, { "epoch": 8.35, "learning_rate": 4.944137273593898e-07, "loss": 0.6702, "step": 43856 }, { "epoch": 8.35, "learning_rate": 4.943565300285987e-07, "loss": 1.531, "step": 43857 }, { "epoch": 8.35, "learning_rate": 4.942993326978074e-07, "loss": 2.5552, "step": 43858 }, { "epoch": 8.35, "learning_rate": 4.942421353670162e-07, "loss": 1.6259, "step": 43859 }, { "epoch": 8.35, "learning_rate": 4.94184938036225e-07, "loss": 2.3824, "step": 43860 }, { "epoch": 8.35, "learning_rate": 4.941277407054337e-07, "loss": 1.1903, "step": 43861 }, { "epoch": 8.35, "learning_rate": 4.940705433746425e-07, "loss": 2.0876, "step": 43862 }, { "epoch": 8.35, "learning_rate": 4.940133460438513e-07, "loss": 0.213, "step": 43863 }, { "epoch": 8.36, "learning_rate": 4.939561487130601e-07, "loss": 0.4007, "step": 43864 }, { "epoch": 8.36, "learning_rate": 4.938989513822689e-07, "loss": 0.8824, "step": 43865 }, { "epoch": 8.36, "learning_rate": 4.938417540514776e-07, "loss": 1.5236, "step": 43866 }, { "epoch": 8.36, "learning_rate": 4.937845567206864e-07, "loss": 1.9685, "step": 43867 }, { "epoch": 8.36, "learning_rate": 4.937273593898952e-07, "loss": 2.088, "step": 43868 }, { "epoch": 8.36, "learning_rate": 4.93670162059104e-07, "loss": 1.6603, "step": 43869 }, { "epoch": 8.36, "learning_rate": 4.936129647283126e-07, "loss": 1.7665, "step": 43870 }, { "epoch": 8.36, "learning_rate": 4.935557673975214e-07, "loss": 1.7173, "step": 43871 }, { "epoch": 8.36, "learning_rate": 4.934985700667302e-07, "loss": 1.1047, "step": 43872 }, { "epoch": 8.36, "learning_rate": 4.93441372735939e-07, "loss": 1.6711, "step": 43873 }, { "epoch": 8.36, "learning_rate": 4.933841754051478e-07, "loss": 2.5214, "step": 43874 }, { "epoch": 8.36, "learning_rate": 4.933269780743565e-07, "loss": 1.7291, "step": 43875 }, { "epoch": 8.36, "learning_rate": 4.932697807435654e-07, "loss": 1.6443, "step": 43876 }, { "epoch": 8.36, "learning_rate": 4.932125834127741e-07, "loss": 1.31, "step": 43877 }, { "epoch": 8.36, "learning_rate": 4.931553860819828e-07, "loss": 1.0299, "step": 43878 }, { "epoch": 8.36, "learning_rate": 4.930981887511917e-07, "loss": 1.2429, "step": 43879 }, { "epoch": 8.36, "learning_rate": 4.930409914204004e-07, "loss": 1.4264, "step": 43880 }, { "epoch": 8.36, "learning_rate": 4.929837940896092e-07, "loss": 2.0011, "step": 43881 }, { "epoch": 8.36, "learning_rate": 4.92926596758818e-07, "loss": 1.4497, "step": 43882 }, { "epoch": 8.36, "learning_rate": 4.928693994280266e-07, "loss": 1.5689, "step": 43883 }, { "epoch": 8.36, "learning_rate": 4.928122020972354e-07, "loss": 2.2886, "step": 43884 }, { "epoch": 8.36, "learning_rate": 4.927550047664442e-07, "loss": 1.8563, "step": 43885 }, { "epoch": 8.36, "learning_rate": 4.92697807435653e-07, "loss": 1.4371, "step": 43886 }, { "epoch": 8.36, "learning_rate": 4.926406101048618e-07, "loss": 1.7408, "step": 43887 }, { "epoch": 8.36, "learning_rate": 4.925834127740706e-07, "loss": 1.0761, "step": 43888 }, { "epoch": 8.36, "learning_rate": 4.925262154432793e-07, "loss": 0.4056, "step": 43889 }, { "epoch": 8.36, "learning_rate": 4.924690181124881e-07, "loss": 1.0509, "step": 43890 }, { "epoch": 8.36, "learning_rate": 4.924118207816969e-07, "loss": 1.7235, "step": 43891 }, { "epoch": 8.36, "learning_rate": 4.923546234509056e-07, "loss": 1.8707, "step": 43892 }, { "epoch": 8.36, "learning_rate": 4.922974261201145e-07, "loss": 1.4715, "step": 43893 }, { "epoch": 8.36, "learning_rate": 4.922402287893232e-07, "loss": 1.7178, "step": 43894 }, { "epoch": 8.36, "learning_rate": 4.92183031458532e-07, "loss": 1.5211, "step": 43895 }, { "epoch": 8.36, "learning_rate": 4.921258341277407e-07, "loss": 3.0273, "step": 43896 }, { "epoch": 8.36, "learning_rate": 4.920686367969494e-07, "loss": 2.0493, "step": 43897 }, { "epoch": 8.36, "learning_rate": 4.920114394661583e-07, "loss": 1.792, "step": 43898 }, { "epoch": 8.36, "learning_rate": 4.91954242135367e-07, "loss": 1.0167, "step": 43899 }, { "epoch": 8.36, "learning_rate": 4.918970448045757e-07, "loss": 2.5766, "step": 43900 }, { "epoch": 8.36, "learning_rate": 4.918398474737846e-07, "loss": 1.3662, "step": 43901 }, { "epoch": 8.36, "learning_rate": 4.917826501429933e-07, "loss": 0.7809, "step": 43902 }, { "epoch": 8.36, "learning_rate": 4.917254528122021e-07, "loss": 1.1511, "step": 43903 }, { "epoch": 8.36, "learning_rate": 4.916682554814109e-07, "loss": 1.3173, "step": 43904 }, { "epoch": 8.36, "learning_rate": 4.916110581506197e-07, "loss": 1.6098, "step": 43905 }, { "epoch": 8.36, "learning_rate": 4.915538608198284e-07, "loss": 2.7685, "step": 43906 }, { "epoch": 8.36, "learning_rate": 4.914966634890372e-07, "loss": 2.4148, "step": 43907 }, { "epoch": 8.36, "learning_rate": 4.91439466158246e-07, "loss": 1.0104, "step": 43908 }, { "epoch": 8.36, "learning_rate": 4.913822688274548e-07, "loss": 1.0355, "step": 43909 }, { "epoch": 8.36, "learning_rate": 4.913250714966635e-07, "loss": 1.1635, "step": 43910 }, { "epoch": 8.36, "learning_rate": 4.912678741658722e-07, "loss": 2.264, "step": 43911 }, { "epoch": 8.36, "learning_rate": 4.91210676835081e-07, "loss": 0.4906, "step": 43912 }, { "epoch": 8.36, "learning_rate": 4.911534795042898e-07, "loss": 1.3561, "step": 43913 }, { "epoch": 8.36, "learning_rate": 4.910962821734985e-07, "loss": 1.6352, "step": 43914 }, { "epoch": 8.36, "learning_rate": 4.910390848427074e-07, "loss": 2.7525, "step": 43915 }, { "epoch": 8.36, "learning_rate": 4.909818875119161e-07, "loss": 1.0726, "step": 43916 }, { "epoch": 8.37, "learning_rate": 4.909246901811249e-07, "loss": 1.8686, "step": 43917 }, { "epoch": 8.37, "learning_rate": 4.908674928503337e-07, "loss": 1.6815, "step": 43918 }, { "epoch": 8.37, "learning_rate": 4.908102955195424e-07, "loss": 1.6822, "step": 43919 }, { "epoch": 8.37, "learning_rate": 4.907530981887513e-07, "loss": 2.1945, "step": 43920 }, { "epoch": 8.37, "learning_rate": 4.9069590085796e-07, "loss": 0.6776, "step": 43921 }, { "epoch": 8.37, "learning_rate": 4.906387035271688e-07, "loss": 2.4539, "step": 43922 }, { "epoch": 8.37, "learning_rate": 4.905815061963775e-07, "loss": 1.2077, "step": 43923 }, { "epoch": 8.37, "learning_rate": 4.905243088655862e-07, "loss": 3.6672, "step": 43924 }, { "epoch": 8.37, "learning_rate": 4.90467111534795e-07, "loss": 2.0722, "step": 43925 }, { "epoch": 8.37, "learning_rate": 4.904099142040038e-07, "loss": 1.1404, "step": 43926 }, { "epoch": 8.37, "learning_rate": 4.903527168732126e-07, "loss": 0.4353, "step": 43927 }, { "epoch": 8.37, "learning_rate": 4.902955195424213e-07, "loss": 1.2914, "step": 43928 }, { "epoch": 8.37, "learning_rate": 4.902383222116302e-07, "loss": 2.5028, "step": 43929 }, { "epoch": 8.37, "learning_rate": 4.901811248808389e-07, "loss": 1.098, "step": 43930 }, { "epoch": 8.37, "learning_rate": 4.901239275500477e-07, "loss": 1.632, "step": 43931 }, { "epoch": 8.37, "learning_rate": 4.900667302192565e-07, "loss": 2.433, "step": 43932 }, { "epoch": 8.37, "learning_rate": 4.900095328884652e-07, "loss": 1.5243, "step": 43933 }, { "epoch": 8.37, "learning_rate": 4.899523355576741e-07, "loss": 1.0366, "step": 43934 }, { "epoch": 8.37, "learning_rate": 4.898951382268828e-07, "loss": 1.2809, "step": 43935 }, { "epoch": 8.37, "learning_rate": 4.898379408960914e-07, "loss": 1.5448, "step": 43936 }, { "epoch": 8.37, "learning_rate": 4.897807435653003e-07, "loss": 2.0976, "step": 43937 }, { "epoch": 8.37, "learning_rate": 4.89723546234509e-07, "loss": 1.9013, "step": 43938 }, { "epoch": 8.37, "learning_rate": 4.896663489037178e-07, "loss": 1.4463, "step": 43939 }, { "epoch": 8.37, "learning_rate": 4.896091515729266e-07, "loss": 1.7933, "step": 43940 }, { "epoch": 8.37, "learning_rate": 4.895519542421354e-07, "loss": 1.6521, "step": 43941 }, { "epoch": 8.37, "learning_rate": 4.894947569113442e-07, "loss": 1.9527, "step": 43942 }, { "epoch": 8.37, "learning_rate": 4.894375595805529e-07, "loss": 1.7568, "step": 43943 }, { "epoch": 8.37, "learning_rate": 4.893803622497617e-07, "loss": 1.8433, "step": 43944 }, { "epoch": 8.37, "learning_rate": 4.893231649189705e-07, "loss": 1.4446, "step": 43945 }, { "epoch": 8.37, "learning_rate": 4.892659675881793e-07, "loss": 0.324, "step": 43946 }, { "epoch": 8.37, "learning_rate": 4.89208770257388e-07, "loss": 0.9415, "step": 43947 }, { "epoch": 8.37, "learning_rate": 4.891515729265968e-07, "loss": 2.3003, "step": 43948 }, { "epoch": 8.37, "learning_rate": 4.890943755958056e-07, "loss": 1.7642, "step": 43949 }, { "epoch": 8.37, "learning_rate": 4.890371782650142e-07, "loss": 1.443, "step": 43950 }, { "epoch": 8.37, "learning_rate": 4.889799809342231e-07, "loss": 1.2618, "step": 43951 }, { "epoch": 8.37, "learning_rate": 4.889227836034318e-07, "loss": 2.4986, "step": 43952 }, { "epoch": 8.37, "learning_rate": 4.888655862726406e-07, "loss": 1.1733, "step": 43953 }, { "epoch": 8.37, "learning_rate": 4.888083889418494e-07, "loss": 1.1971, "step": 43954 }, { "epoch": 8.37, "learning_rate": 4.887511916110581e-07, "loss": 2.0047, "step": 43955 }, { "epoch": 8.37, "learning_rate": 4.88693994280267e-07, "loss": 2.4556, "step": 43956 }, { "epoch": 8.37, "learning_rate": 4.886367969494757e-07, "loss": 1.6764, "step": 43957 }, { "epoch": 8.37, "learning_rate": 4.885795996186845e-07, "loss": 3.0806, "step": 43958 }, { "epoch": 8.37, "learning_rate": 4.885224022878933e-07, "loss": 2.1861, "step": 43959 }, { "epoch": 8.37, "learning_rate": 4.88465204957102e-07, "loss": 1.3224, "step": 43960 }, { "epoch": 8.37, "learning_rate": 4.884080076263108e-07, "loss": 2.6315, "step": 43961 }, { "epoch": 8.37, "learning_rate": 4.883508102955196e-07, "loss": 0.2922, "step": 43962 }, { "epoch": 8.37, "learning_rate": 4.882936129647283e-07, "loss": 1.1322, "step": 43963 }, { "epoch": 8.37, "learning_rate": 4.882364156339371e-07, "loss": 1.411, "step": 43964 }, { "epoch": 8.37, "learning_rate": 4.881792183031458e-07, "loss": 1.8087, "step": 43965 }, { "epoch": 8.37, "learning_rate": 4.881220209723546e-07, "loss": 1.8026, "step": 43966 }, { "epoch": 8.37, "learning_rate": 4.880648236415634e-07, "loss": 0.718, "step": 43967 }, { "epoch": 8.37, "learning_rate": 4.880076263107722e-07, "loss": 0.9526, "step": 43968 }, { "epoch": 8.38, "learning_rate": 4.879504289799809e-07, "loss": 1.6495, "step": 43969 }, { "epoch": 8.38, "learning_rate": 4.878932316491898e-07, "loss": 1.6052, "step": 43970 }, { "epoch": 8.38, "learning_rate": 4.878360343183985e-07, "loss": 2.2947, "step": 43971 }, { "epoch": 8.38, "learning_rate": 4.877788369876072e-07, "loss": 2.1557, "step": 43972 }, { "epoch": 8.38, "learning_rate": 4.877216396568161e-07, "loss": 0.7354, "step": 43973 }, { "epoch": 8.38, "learning_rate": 4.876644423260248e-07, "loss": 1.4802, "step": 43974 }, { "epoch": 8.38, "learning_rate": 4.876072449952336e-07, "loss": 0.6345, "step": 43975 }, { "epoch": 8.38, "learning_rate": 4.875500476644423e-07, "loss": 1.0895, "step": 43976 }, { "epoch": 8.38, "learning_rate": 4.87492850333651e-07, "loss": 1.2479, "step": 43977 }, { "epoch": 8.38, "learning_rate": 4.874356530028599e-07, "loss": 1.0819, "step": 43978 }, { "epoch": 8.38, "learning_rate": 4.873784556720686e-07, "loss": 1.5377, "step": 43979 }, { "epoch": 8.38, "learning_rate": 4.873212583412774e-07, "loss": 1.6731, "step": 43980 }, { "epoch": 8.38, "learning_rate": 4.872640610104862e-07, "loss": 0.2117, "step": 43981 }, { "epoch": 8.38, "learning_rate": 4.87206863679695e-07, "loss": 1.3199, "step": 43982 }, { "epoch": 8.38, "learning_rate": 4.871496663489037e-07, "loss": 1.1359, "step": 43983 }, { "epoch": 8.38, "learning_rate": 4.870924690181125e-07, "loss": 2.0337, "step": 43984 }, { "epoch": 8.38, "learning_rate": 4.870352716873213e-07, "loss": 2.3203, "step": 43985 }, { "epoch": 8.38, "learning_rate": 4.869780743565301e-07, "loss": 0.9865, "step": 43986 }, { "epoch": 8.38, "learning_rate": 4.869208770257389e-07, "loss": 1.8091, "step": 43987 }, { "epoch": 8.38, "learning_rate": 4.868636796949476e-07, "loss": 1.5934, "step": 43988 }, { "epoch": 8.38, "learning_rate": 4.868064823641564e-07, "loss": 2.4606, "step": 43989 }, { "epoch": 8.38, "learning_rate": 4.867492850333651e-07, "loss": 1.5984, "step": 43990 }, { "epoch": 8.38, "learning_rate": 4.866920877025738e-07, "loss": 2.313, "step": 43991 }, { "epoch": 8.38, "learning_rate": 4.866348903717827e-07, "loss": 0.5764, "step": 43992 }, { "epoch": 8.38, "learning_rate": 4.865776930409914e-07, "loss": 1.5862, "step": 43993 }, { "epoch": 8.38, "learning_rate": 4.865204957102002e-07, "loss": 2.3667, "step": 43994 }, { "epoch": 8.38, "learning_rate": 4.86463298379409e-07, "loss": 1.2127, "step": 43995 }, { "epoch": 8.38, "learning_rate": 4.864061010486177e-07, "loss": 1.5579, "step": 43996 }, { "epoch": 8.38, "learning_rate": 4.863489037178265e-07, "loss": 0.2738, "step": 43997 }, { "epoch": 8.38, "learning_rate": 4.862917063870353e-07, "loss": 0.9826, "step": 43998 }, { "epoch": 8.38, "learning_rate": 4.862345090562441e-07, "loss": 2.0825, "step": 43999 }, { "epoch": 8.38, "learning_rate": 4.861773117254529e-07, "loss": 2.3051, "step": 44000 }, { "epoch": 8.38, "learning_rate": 4.861201143946616e-07, "loss": 2.7674, "step": 44001 }, { "epoch": 8.38, "learning_rate": 4.860629170638704e-07, "loss": 2.0563, "step": 44002 }, { "epoch": 8.38, "learning_rate": 4.860057197330791e-07, "loss": 1.7661, "step": 44003 }, { "epoch": 8.38, "learning_rate": 4.859485224022879e-07, "loss": 1.8114, "step": 44004 }, { "epoch": 8.38, "learning_rate": 4.858913250714966e-07, "loss": 1.0751, "step": 44005 }, { "epoch": 8.38, "learning_rate": 4.858341277407055e-07, "loss": 1.2281, "step": 44006 }, { "epoch": 8.38, "learning_rate": 4.857769304099142e-07, "loss": 0.8976, "step": 44007 }, { "epoch": 8.38, "learning_rate": 4.857197330791229e-07, "loss": 1.0972, "step": 44008 }, { "epoch": 8.38, "learning_rate": 4.856625357483318e-07, "loss": 1.3079, "step": 44009 }, { "epoch": 8.38, "learning_rate": 4.856053384175405e-07, "loss": 0.7364, "step": 44010 }, { "epoch": 8.38, "learning_rate": 4.855481410867493e-07, "loss": 1.9107, "step": 44011 }, { "epoch": 8.38, "learning_rate": 4.854909437559581e-07, "loss": 1.6878, "step": 44012 }, { "epoch": 8.38, "learning_rate": 4.854337464251668e-07, "loss": 0.8608, "step": 44013 }, { "epoch": 8.38, "learning_rate": 4.853765490943757e-07, "loss": 1.456, "step": 44014 }, { "epoch": 8.38, "learning_rate": 4.853193517635844e-07, "loss": 1.7888, "step": 44015 }, { "epoch": 8.38, "learning_rate": 4.852621544327931e-07, "loss": 2.0656, "step": 44016 }, { "epoch": 8.38, "learning_rate": 4.852049571020019e-07, "loss": 0.6645, "step": 44017 }, { "epoch": 8.38, "learning_rate": 4.851477597712107e-07, "loss": 1.8909, "step": 44018 }, { "epoch": 8.38, "learning_rate": 4.850905624404194e-07, "loss": 1.4493, "step": 44019 }, { "epoch": 8.38, "learning_rate": 4.850333651096282e-07, "loss": 1.0925, "step": 44020 }, { "epoch": 8.38, "learning_rate": 4.84976167778837e-07, "loss": 0.9256, "step": 44021 }, { "epoch": 8.39, "learning_rate": 4.849189704480458e-07, "loss": 2.509, "step": 44022 }, { "epoch": 8.39, "learning_rate": 4.848617731172546e-07, "loss": 0.6867, "step": 44023 }, { "epoch": 8.39, "learning_rate": 4.848045757864633e-07, "loss": 2.5567, "step": 44024 }, { "epoch": 8.39, "learning_rate": 4.847473784556721e-07, "loss": 1.4187, "step": 44025 }, { "epoch": 8.39, "learning_rate": 4.846901811248809e-07, "loss": 1.1511, "step": 44026 }, { "epoch": 8.39, "learning_rate": 4.846329837940896e-07, "loss": 2.8815, "step": 44027 }, { "epoch": 8.39, "learning_rate": 4.845757864632985e-07, "loss": 1.7222, "step": 44028 }, { "epoch": 8.39, "learning_rate": 4.845185891325072e-07, "loss": 2.8456, "step": 44029 }, { "epoch": 8.39, "learning_rate": 4.844613918017158e-07, "loss": 1.5524, "step": 44030 }, { "epoch": 8.39, "learning_rate": 4.844041944709247e-07, "loss": 0.7217, "step": 44031 }, { "epoch": 8.39, "learning_rate": 4.843469971401334e-07, "loss": 0.2475, "step": 44032 }, { "epoch": 8.39, "learning_rate": 4.842897998093422e-07, "loss": 1.3231, "step": 44033 }, { "epoch": 8.39, "learning_rate": 4.84232602478551e-07, "loss": 2.123, "step": 44034 }, { "epoch": 8.39, "learning_rate": 4.841754051477598e-07, "loss": 2.7794, "step": 44035 }, { "epoch": 8.39, "learning_rate": 4.841182078169686e-07, "loss": 2.3216, "step": 44036 }, { "epoch": 8.39, "learning_rate": 4.840610104861773e-07, "loss": 1.5016, "step": 44037 }, { "epoch": 8.39, "learning_rate": 4.840038131553861e-07, "loss": 2.0574, "step": 44038 }, { "epoch": 8.39, "learning_rate": 4.839466158245949e-07, "loss": 1.4277, "step": 44039 }, { "epoch": 8.39, "learning_rate": 4.838894184938037e-07, "loss": 0.8576, "step": 44040 }, { "epoch": 8.39, "learning_rate": 4.838322211630124e-07, "loss": 0.557, "step": 44041 }, { "epoch": 8.39, "learning_rate": 4.837750238322212e-07, "loss": 1.2864, "step": 44042 }, { "epoch": 8.39, "learning_rate": 4.8371782650143e-07, "loss": 3.4225, "step": 44043 }, { "epoch": 8.39, "learning_rate": 4.836606291706387e-07, "loss": 0.9333, "step": 44044 }, { "epoch": 8.39, "learning_rate": 4.836034318398475e-07, "loss": 1.0286, "step": 44045 }, { "epoch": 8.39, "learning_rate": 4.835462345090562e-07, "loss": 1.0459, "step": 44046 }, { "epoch": 8.39, "learning_rate": 4.83489037178265e-07, "loss": 1.9134, "step": 44047 }, { "epoch": 8.39, "learning_rate": 4.834318398474738e-07, "loss": 1.9034, "step": 44048 }, { "epoch": 8.39, "learning_rate": 4.833746425166825e-07, "loss": 1.2185, "step": 44049 }, { "epoch": 8.39, "learning_rate": 4.833174451858914e-07, "loss": 0.9414, "step": 44050 }, { "epoch": 8.39, "learning_rate": 4.832602478551001e-07, "loss": 1.752, "step": 44051 }, { "epoch": 8.39, "learning_rate": 4.832030505243089e-07, "loss": 1.533, "step": 44052 }, { "epoch": 8.39, "learning_rate": 4.831458531935177e-07, "loss": 2.1012, "step": 44053 }, { "epoch": 8.39, "learning_rate": 4.830886558627264e-07, "loss": 0.8591, "step": 44054 }, { "epoch": 8.39, "learning_rate": 4.830314585319352e-07, "loss": 0.2734, "step": 44055 }, { "epoch": 8.39, "learning_rate": 4.82974261201144e-07, "loss": 0.7904, "step": 44056 }, { "epoch": 8.39, "learning_rate": 4.829170638703527e-07, "loss": 2.1201, "step": 44057 }, { "epoch": 8.39, "learning_rate": 4.828598665395615e-07, "loss": 0.9742, "step": 44058 }, { "epoch": 8.39, "learning_rate": 4.828026692087703e-07, "loss": 1.8016, "step": 44059 }, { "epoch": 8.39, "learning_rate": 4.82745471877979e-07, "loss": 0.6288, "step": 44060 }, { "epoch": 8.39, "learning_rate": 4.826882745471878e-07, "loss": 1.605, "step": 44061 }, { "epoch": 8.39, "learning_rate": 4.826310772163966e-07, "loss": 1.504, "step": 44062 }, { "epoch": 8.39, "learning_rate": 4.825738798856053e-07, "loss": 0.4372, "step": 44063 }, { "epoch": 8.39, "learning_rate": 4.825166825548142e-07, "loss": 2.3283, "step": 44064 }, { "epoch": 8.39, "learning_rate": 4.824594852240229e-07, "loss": 1.0033, "step": 44065 }, { "epoch": 8.39, "learning_rate": 4.824022878932317e-07, "loss": 1.9074, "step": 44066 }, { "epoch": 8.39, "learning_rate": 4.823450905624405e-07, "loss": 1.041, "step": 44067 }, { "epoch": 8.39, "learning_rate": 4.822878932316492e-07, "loss": 1.5352, "step": 44068 }, { "epoch": 8.39, "learning_rate": 4.82230695900858e-07, "loss": 1.4923, "step": 44069 }, { "epoch": 8.39, "learning_rate": 4.821734985700667e-07, "loss": 2.806, "step": 44070 }, { "epoch": 8.39, "learning_rate": 4.821163012392755e-07, "loss": 1.7881, "step": 44071 }, { "epoch": 8.39, "learning_rate": 4.820591039084843e-07, "loss": 1.3761, "step": 44072 }, { "epoch": 8.39, "learning_rate": 4.82001906577693e-07, "loss": 1.7391, "step": 44073 }, { "epoch": 8.4, "learning_rate": 4.819447092469018e-07, "loss": 2.6277, "step": 44074 }, { "epoch": 8.4, "learning_rate": 4.818875119161106e-07, "loss": 2.3696, "step": 44075 }, { "epoch": 8.4, "learning_rate": 4.818303145853194e-07, "loss": 0.515, "step": 44076 }, { "epoch": 8.4, "learning_rate": 4.817731172545281e-07, "loss": 1.4879, "step": 44077 }, { "epoch": 8.4, "learning_rate": 4.817159199237369e-07, "loss": 1.4151, "step": 44078 }, { "epoch": 8.4, "learning_rate": 4.816587225929457e-07, "loss": 2.2653, "step": 44079 }, { "epoch": 8.4, "learning_rate": 4.816015252621545e-07, "loss": 1.294, "step": 44080 }, { "epoch": 8.4, "learning_rate": 4.815443279313633e-07, "loss": 1.8792, "step": 44081 }, { "epoch": 8.4, "learning_rate": 4.81487130600572e-07, "loss": 1.1326, "step": 44082 }, { "epoch": 8.4, "learning_rate": 4.814299332697807e-07, "loss": 1.4026, "step": 44083 }, { "epoch": 8.4, "learning_rate": 4.813727359389895e-07, "loss": 0.4033, "step": 44084 }, { "epoch": 8.4, "learning_rate": 4.813155386081982e-07, "loss": 1.445, "step": 44085 }, { "epoch": 8.4, "learning_rate": 4.812583412774071e-07, "loss": 2.3718, "step": 44086 }, { "epoch": 8.4, "learning_rate": 4.812011439466158e-07, "loss": 1.9385, "step": 44087 }, { "epoch": 8.4, "learning_rate": 4.811439466158246e-07, "loss": 1.7376, "step": 44088 }, { "epoch": 8.4, "learning_rate": 4.810867492850334e-07, "loss": 1.6805, "step": 44089 }, { "epoch": 8.4, "learning_rate": 4.810295519542421e-07, "loss": 2.9771, "step": 44090 }, { "epoch": 8.4, "learning_rate": 4.809723546234509e-07, "loss": 2.4317, "step": 44091 }, { "epoch": 8.4, "learning_rate": 4.809151572926597e-07, "loss": 1.5631, "step": 44092 }, { "epoch": 8.4, "learning_rate": 4.808579599618685e-07, "loss": 2.1607, "step": 44093 }, { "epoch": 8.4, "learning_rate": 4.808007626310773e-07, "loss": 2.1752, "step": 44094 }, { "epoch": 8.4, "learning_rate": 4.80743565300286e-07, "loss": 1.5366, "step": 44095 }, { "epoch": 8.4, "learning_rate": 4.806863679694947e-07, "loss": 1.6129, "step": 44096 }, { "epoch": 8.4, "learning_rate": 4.806291706387035e-07, "loss": 2.05, "step": 44097 }, { "epoch": 8.4, "learning_rate": 4.805719733079123e-07, "loss": 0.8011, "step": 44098 }, { "epoch": 8.4, "learning_rate": 4.80514775977121e-07, "loss": 0.6304, "step": 44099 }, { "epoch": 8.4, "learning_rate": 4.804575786463299e-07, "loss": 0.1571, "step": 44100 }, { "epoch": 8.4, "learning_rate": 4.804003813155386e-07, "loss": 0.5005, "step": 44101 }, { "epoch": 8.4, "learning_rate": 4.803431839847474e-07, "loss": 1.4827, "step": 44102 }, { "epoch": 8.4, "learning_rate": 4.802859866539562e-07, "loss": 1.2465, "step": 44103 }, { "epoch": 8.4, "learning_rate": 4.802287893231649e-07, "loss": 1.0876, "step": 44104 }, { "epoch": 8.4, "learning_rate": 4.801715919923737e-07, "loss": 0.6289, "step": 44105 }, { "epoch": 8.4, "learning_rate": 4.801143946615825e-07, "loss": 1.7974, "step": 44106 }, { "epoch": 8.4, "learning_rate": 4.800571973307912e-07, "loss": 1.6013, "step": 44107 }, { "epoch": 8.4, "learning_rate": 4.800000000000001e-07, "loss": 1.9302, "step": 44108 }, { "epoch": 8.4, "learning_rate": 4.799428026692088e-07, "loss": 1.1406, "step": 44109 }, { "epoch": 8.4, "learning_rate": 4.798856053384175e-07, "loss": 1.0095, "step": 44110 }, { "epoch": 8.4, "learning_rate": 4.798284080076263e-07, "loss": 2.0795, "step": 44111 }, { "epoch": 8.4, "learning_rate": 4.797712106768351e-07, "loss": 0.9424, "step": 44112 }, { "epoch": 8.4, "learning_rate": 4.797140133460438e-07, "loss": 1.689, "step": 44113 }, { "epoch": 8.4, "learning_rate": 4.796568160152526e-07, "loss": 0.7093, "step": 44114 }, { "epoch": 8.4, "learning_rate": 4.795996186844614e-07, "loss": 2.741, "step": 44115 }, { "epoch": 8.4, "learning_rate": 4.795424213536702e-07, "loss": 1.7667, "step": 44116 }, { "epoch": 8.4, "learning_rate": 4.79485224022879e-07, "loss": 0.8484, "step": 44117 }, { "epoch": 8.4, "learning_rate": 4.794280266920877e-07, "loss": 1.9018, "step": 44118 }, { "epoch": 8.4, "learning_rate": 4.793708293612965e-07, "loss": 2.7544, "step": 44119 }, { "epoch": 8.4, "learning_rate": 4.793136320305053e-07, "loss": 1.8817, "step": 44120 }, { "epoch": 8.4, "learning_rate": 4.79256434699714e-07, "loss": 1.0976, "step": 44121 }, { "epoch": 8.4, "learning_rate": 4.791992373689229e-07, "loss": 2.6933, "step": 44122 }, { "epoch": 8.4, "learning_rate": 4.791420400381315e-07, "loss": 1.269, "step": 44123 }, { "epoch": 8.4, "learning_rate": 4.790848427073403e-07, "loss": 2.2112, "step": 44124 }, { "epoch": 8.4, "learning_rate": 4.790276453765491e-07, "loss": 1.1697, "step": 44125 }, { "epoch": 8.4, "learning_rate": 4.789704480457578e-07, "loss": 1.8054, "step": 44126 }, { "epoch": 8.41, "learning_rate": 4.789132507149666e-07, "loss": 1.136, "step": 44127 }, { "epoch": 8.41, "learning_rate": 4.788560533841754e-07, "loss": 2.0315, "step": 44128 }, { "epoch": 8.41, "learning_rate": 4.787988560533842e-07, "loss": 0.5929, "step": 44129 }, { "epoch": 8.41, "learning_rate": 4.78741658722593e-07, "loss": 2.7068, "step": 44130 }, { "epoch": 8.41, "learning_rate": 4.786844613918017e-07, "loss": 0.3089, "step": 44131 }, { "epoch": 8.41, "learning_rate": 4.786272640610105e-07, "loss": 2.3757, "step": 44132 }, { "epoch": 8.41, "learning_rate": 4.785700667302193e-07, "loss": 0.8965, "step": 44133 }, { "epoch": 8.41, "learning_rate": 4.785128693994281e-07, "loss": 2.0245, "step": 44134 }, { "epoch": 8.41, "learning_rate": 4.784556720686368e-07, "loss": 1.459, "step": 44135 }, { "epoch": 8.41, "learning_rate": 4.783984747378457e-07, "loss": 2.2331, "step": 44136 }, { "epoch": 8.41, "learning_rate": 4.783412774070543e-07, "loss": 1.2986, "step": 44137 }, { "epoch": 8.41, "learning_rate": 4.782840800762631e-07, "loss": 2.2113, "step": 44138 }, { "epoch": 8.41, "learning_rate": 4.782268827454719e-07, "loss": 1.9925, "step": 44139 }, { "epoch": 8.41, "learning_rate": 4.781696854146806e-07, "loss": 2.6614, "step": 44140 }, { "epoch": 8.41, "learning_rate": 4.781124880838895e-07, "loss": 0.3272, "step": 44141 }, { "epoch": 8.41, "learning_rate": 4.780552907530982e-07, "loss": 1.2023, "step": 44142 }, { "epoch": 8.41, "learning_rate": 4.779980934223069e-07, "loss": 1.278, "step": 44143 }, { "epoch": 8.41, "learning_rate": 4.779408960915158e-07, "loss": 0.9963, "step": 44144 }, { "epoch": 8.41, "learning_rate": 4.778836987607245e-07, "loss": 1.6743, "step": 44145 }, { "epoch": 8.41, "learning_rate": 4.778265014299333e-07, "loss": 0.9888, "step": 44146 }, { "epoch": 8.41, "learning_rate": 4.777693040991421e-07, "loss": 0.4843, "step": 44147 }, { "epoch": 8.41, "learning_rate": 4.777121067683508e-07, "loss": 1.4982, "step": 44148 }, { "epoch": 8.41, "learning_rate": 4.776549094375596e-07, "loss": 3.6956, "step": 44149 }, { "epoch": 8.41, "learning_rate": 4.775977121067683e-07, "loss": 1.2504, "step": 44150 }, { "epoch": 8.41, "learning_rate": 4.775405147759771e-07, "loss": 1.9823, "step": 44151 }, { "epoch": 8.41, "learning_rate": 4.774833174451859e-07, "loss": 1.3349, "step": 44152 }, { "epoch": 8.41, "learning_rate": 4.774261201143947e-07, "loss": 1.9945, "step": 44153 }, { "epoch": 8.41, "learning_rate": 4.773689227836034e-07, "loss": 2.9186, "step": 44154 }, { "epoch": 8.41, "learning_rate": 4.773117254528122e-07, "loss": 1.7447, "step": 44155 }, { "epoch": 8.41, "learning_rate": 4.77254528122021e-07, "loss": 0.916, "step": 44156 }, { "epoch": 8.41, "learning_rate": 4.771973307912297e-07, "loss": 2.6657, "step": 44157 }, { "epoch": 8.41, "learning_rate": 4.771401334604386e-07, "loss": 2.9041, "step": 44158 }, { "epoch": 8.41, "learning_rate": 4.770829361296473e-07, "loss": 0.9089, "step": 44159 }, { "epoch": 8.41, "learning_rate": 4.770257387988561e-07, "loss": 1.6853, "step": 44160 }, { "epoch": 8.41, "learning_rate": 4.769685414680649e-07, "loss": 0.3089, "step": 44161 }, { "epoch": 8.41, "learning_rate": 4.769113441372736e-07, "loss": 1.607, "step": 44162 }, { "epoch": 8.41, "learning_rate": 4.768541468064823e-07, "loss": 0.8801, "step": 44163 }, { "epoch": 8.41, "learning_rate": 4.7679694947569114e-07, "loss": 0.7935, "step": 44164 }, { "epoch": 8.41, "learning_rate": 4.7673975214489993e-07, "loss": 1.124, "step": 44165 }, { "epoch": 8.41, "learning_rate": 4.766825548141087e-07, "loss": 1.6264, "step": 44166 }, { "epoch": 8.41, "learning_rate": 4.766253574833174e-07, "loss": 2.3521, "step": 44167 }, { "epoch": 8.41, "learning_rate": 4.7656816015252626e-07, "loss": 1.8198, "step": 44168 }, { "epoch": 8.41, "learning_rate": 4.76510962821735e-07, "loss": 2.2268, "step": 44169 }, { "epoch": 8.41, "learning_rate": 4.764537654909438e-07, "loss": 1.3413, "step": 44170 }, { "epoch": 8.41, "learning_rate": 4.7639656816015254e-07, "loss": 2.4222, "step": 44171 }, { "epoch": 8.41, "learning_rate": 4.763393708293613e-07, "loss": 0.3871, "step": 44172 }, { "epoch": 8.41, "learning_rate": 4.7628217349857007e-07, "loss": 1.2621, "step": 44173 }, { "epoch": 8.41, "learning_rate": 4.762249761677788e-07, "loss": 2.3383, "step": 44174 }, { "epoch": 8.41, "learning_rate": 4.7616777883698766e-07, "loss": 2.5593, "step": 44175 }, { "epoch": 8.41, "learning_rate": 4.761105815061964e-07, "loss": 1.9888, "step": 44176 }, { "epoch": 8.41, "learning_rate": 4.760533841754052e-07, "loss": 1.4048, "step": 44177 }, { "epoch": 8.41, "learning_rate": 4.7599618684461394e-07, "loss": 1.8343, "step": 44178 }, { "epoch": 8.42, "learning_rate": 4.759389895138227e-07, "loss": 0.9951, "step": 44179 }, { "epoch": 8.42, "learning_rate": 4.7588179218303147e-07, "loss": 1.2301, "step": 44180 }, { "epoch": 8.42, "learning_rate": 4.758245948522402e-07, "loss": 1.7176, "step": 44181 }, { "epoch": 8.42, "learning_rate": 4.7576739752144906e-07, "loss": 1.8809, "step": 44182 }, { "epoch": 8.42, "learning_rate": 4.757102001906578e-07, "loss": 1.8511, "step": 44183 }, { "epoch": 8.42, "learning_rate": 4.7565300285986654e-07, "loss": 0.4114, "step": 44184 }, { "epoch": 8.42, "learning_rate": 4.7559580552907533e-07, "loss": 0.402, "step": 44185 }, { "epoch": 8.42, "learning_rate": 4.755386081982841e-07, "loss": 3.0522, "step": 44186 }, { "epoch": 8.42, "learning_rate": 4.7548141086749287e-07, "loss": 1.0854, "step": 44187 }, { "epoch": 8.42, "learning_rate": 4.754242135367016e-07, "loss": 1.6379, "step": 44188 }, { "epoch": 8.42, "learning_rate": 4.7536701620591046e-07, "loss": 1.7016, "step": 44189 }, { "epoch": 8.42, "learning_rate": 4.753098188751192e-07, "loss": 1.5002, "step": 44190 }, { "epoch": 8.42, "learning_rate": 4.7525262154432794e-07, "loss": 1.3908, "step": 44191 }, { "epoch": 8.42, "learning_rate": 4.7519542421353673e-07, "loss": 2.2942, "step": 44192 }, { "epoch": 8.42, "learning_rate": 4.7513822688274547e-07, "loss": 1.8712, "step": 44193 }, { "epoch": 8.42, "learning_rate": 4.7508102955195427e-07, "loss": 1.072, "step": 44194 }, { "epoch": 8.42, "learning_rate": 4.75023832221163e-07, "loss": 1.4967, "step": 44195 }, { "epoch": 8.42, "learning_rate": 4.7496663489037175e-07, "loss": 0.9811, "step": 44196 }, { "epoch": 8.42, "learning_rate": 4.749094375595806e-07, "loss": 0.369, "step": 44197 }, { "epoch": 8.42, "learning_rate": 4.7485224022878934e-07, "loss": 1.1542, "step": 44198 }, { "epoch": 8.42, "learning_rate": 4.7479504289799813e-07, "loss": 1.1122, "step": 44199 }, { "epoch": 8.42, "learning_rate": 4.7473784556720687e-07, "loss": 2.3716, "step": 44200 }, { "epoch": 8.42, "learning_rate": 4.7468064823641566e-07, "loss": 1.6588, "step": 44201 }, { "epoch": 8.42, "learning_rate": 4.746234509056244e-07, "loss": 0.8572, "step": 44202 }, { "epoch": 8.42, "learning_rate": 4.7456625357483315e-07, "loss": 1.2351, "step": 44203 }, { "epoch": 8.42, "learning_rate": 4.74509056244042e-07, "loss": 1.932, "step": 44204 }, { "epoch": 8.42, "learning_rate": 4.7445185891325073e-07, "loss": 1.8474, "step": 44205 }, { "epoch": 8.42, "learning_rate": 4.7439466158245953e-07, "loss": 2.375, "step": 44206 }, { "epoch": 8.42, "learning_rate": 4.7433746425166827e-07, "loss": 3.0705, "step": 44207 }, { "epoch": 8.42, "learning_rate": 4.74280266920877e-07, "loss": 1.4699, "step": 44208 }, { "epoch": 8.42, "learning_rate": 4.742230695900858e-07, "loss": 1.3916, "step": 44209 }, { "epoch": 8.42, "learning_rate": 4.7416587225929454e-07, "loss": 0.7151, "step": 44210 }, { "epoch": 8.42, "learning_rate": 4.741086749285034e-07, "loss": 1.6786, "step": 44211 }, { "epoch": 8.42, "learning_rate": 4.7405147759771213e-07, "loss": 2.2115, "step": 44212 }, { "epoch": 8.42, "learning_rate": 4.7399428026692087e-07, "loss": 1.3524, "step": 44213 }, { "epoch": 8.42, "learning_rate": 4.7393708293612967e-07, "loss": 0.4308, "step": 44214 }, { "epoch": 8.42, "learning_rate": 4.738798856053384e-07, "loss": 2.2673, "step": 44215 }, { "epoch": 8.42, "learning_rate": 4.738226882745472e-07, "loss": 1.5065, "step": 44216 }, { "epoch": 8.42, "learning_rate": 4.7376549094375594e-07, "loss": 2.2296, "step": 44217 }, { "epoch": 8.42, "learning_rate": 4.737082936129648e-07, "loss": 1.6274, "step": 44218 }, { "epoch": 8.42, "learning_rate": 4.7365109628217353e-07, "loss": 1.6401, "step": 44219 }, { "epoch": 8.42, "learning_rate": 4.7359389895138227e-07, "loss": 2.17, "step": 44220 }, { "epoch": 8.42, "learning_rate": 4.7353670162059106e-07, "loss": 0.6873, "step": 44221 }, { "epoch": 8.42, "learning_rate": 4.734795042897998e-07, "loss": 1.2224, "step": 44222 }, { "epoch": 8.42, "learning_rate": 4.734223069590086e-07, "loss": 2.3341, "step": 44223 }, { "epoch": 8.42, "learning_rate": 4.7336510962821734e-07, "loss": 2.0737, "step": 44224 }, { "epoch": 8.42, "learning_rate": 4.733079122974261e-07, "loss": 1.0598, "step": 44225 }, { "epoch": 8.42, "learning_rate": 4.7325071496663493e-07, "loss": 2.2779, "step": 44226 }, { "epoch": 8.42, "learning_rate": 4.7319351763584367e-07, "loss": 1.4483, "step": 44227 }, { "epoch": 8.42, "learning_rate": 4.7313632030505246e-07, "loss": 1.6685, "step": 44228 }, { "epoch": 8.42, "learning_rate": 4.730791229742612e-07, "loss": 2.2018, "step": 44229 }, { "epoch": 8.42, "learning_rate": 4.7302192564347e-07, "loss": 1.7183, "step": 44230 }, { "epoch": 8.42, "learning_rate": 4.7296472831267874e-07, "loss": 1.8078, "step": 44231 }, { "epoch": 8.43, "learning_rate": 4.729075309818875e-07, "loss": 0.8048, "step": 44232 }, { "epoch": 8.43, "learning_rate": 4.728503336510963e-07, "loss": 2.5306, "step": 44233 }, { "epoch": 8.43, "learning_rate": 4.7279313632030507e-07, "loss": 2.2589, "step": 44234 }, { "epoch": 8.43, "learning_rate": 4.7273593898951386e-07, "loss": 1.395, "step": 44235 }, { "epoch": 8.43, "learning_rate": 4.726787416587226e-07, "loss": 1.3428, "step": 44236 }, { "epoch": 8.43, "learning_rate": 4.7262154432793134e-07, "loss": 1.474, "step": 44237 }, { "epoch": 8.43, "learning_rate": 4.7256434699714014e-07, "loss": 3.5969, "step": 44238 }, { "epoch": 8.43, "learning_rate": 4.725071496663489e-07, "loss": 1.856, "step": 44239 }, { "epoch": 8.43, "learning_rate": 4.724499523355577e-07, "loss": 3.7093, "step": 44240 }, { "epoch": 8.43, "learning_rate": 4.7239275500476646e-07, "loss": 1.2629, "step": 44241 }, { "epoch": 8.43, "learning_rate": 4.7233555767397526e-07, "loss": 0.5254, "step": 44242 }, { "epoch": 8.43, "learning_rate": 4.72278360343184e-07, "loss": 1.9485, "step": 44243 }, { "epoch": 8.43, "learning_rate": 4.7222116301239274e-07, "loss": 1.2373, "step": 44244 }, { "epoch": 8.43, "learning_rate": 4.7216396568160153e-07, "loss": 0.3373, "step": 44245 }, { "epoch": 8.43, "learning_rate": 4.721067683508103e-07, "loss": 1.3465, "step": 44246 }, { "epoch": 8.43, "learning_rate": 4.720495710200191e-07, "loss": 2.4601, "step": 44247 }, { "epoch": 8.43, "learning_rate": 4.7199237368922786e-07, "loss": 1.023, "step": 44248 }, { "epoch": 8.43, "learning_rate": 4.719351763584366e-07, "loss": 1.4604, "step": 44249 }, { "epoch": 8.43, "learning_rate": 4.718779790276454e-07, "loss": 3.2338, "step": 44250 }, { "epoch": 8.43, "learning_rate": 4.7182078169685414e-07, "loss": 1.3203, "step": 44251 }, { "epoch": 8.43, "learning_rate": 4.7176358436606293e-07, "loss": 1.033, "step": 44252 }, { "epoch": 8.43, "learning_rate": 4.7170638703527167e-07, "loss": 1.5814, "step": 44253 }, { "epoch": 8.43, "learning_rate": 4.716491897044805e-07, "loss": 0.3829, "step": 44254 }, { "epoch": 8.43, "learning_rate": 4.7159199237368926e-07, "loss": 1.5959, "step": 44255 }, { "epoch": 8.43, "learning_rate": 4.71534795042898e-07, "loss": 1.9534, "step": 44256 }, { "epoch": 8.43, "learning_rate": 4.714775977121068e-07, "loss": 3.4138, "step": 44257 }, { "epoch": 8.43, "learning_rate": 4.7142040038131554e-07, "loss": 1.4996, "step": 44258 }, { "epoch": 8.43, "learning_rate": 4.7136320305052433e-07, "loss": 1.7753, "step": 44259 }, { "epoch": 8.43, "learning_rate": 4.7130600571973307e-07, "loss": 1.9806, "step": 44260 }, { "epoch": 8.43, "learning_rate": 4.712488083889418e-07, "loss": 1.2786, "step": 44261 }, { "epoch": 8.43, "learning_rate": 4.7119161105815066e-07, "loss": 2.9478, "step": 44262 }, { "epoch": 8.43, "learning_rate": 4.711344137273594e-07, "loss": 2.0667, "step": 44263 }, { "epoch": 8.43, "learning_rate": 4.710772163965682e-07, "loss": 1.3055, "step": 44264 }, { "epoch": 8.43, "learning_rate": 4.7102001906577693e-07, "loss": 1.7846, "step": 44265 }, { "epoch": 8.43, "learning_rate": 4.709628217349857e-07, "loss": 1.827, "step": 44266 }, { "epoch": 8.43, "learning_rate": 4.7090562440419447e-07, "loss": 1.0944, "step": 44267 }, { "epoch": 8.43, "learning_rate": 4.708484270734032e-07, "loss": 1.5885, "step": 44268 }, { "epoch": 8.43, "learning_rate": 4.7079122974261206e-07, "loss": 2.4938, "step": 44269 }, { "epoch": 8.43, "learning_rate": 4.707340324118208e-07, "loss": 1.5626, "step": 44270 }, { "epoch": 8.43, "learning_rate": 4.706768350810296e-07, "loss": 0.5756, "step": 44271 }, { "epoch": 8.43, "learning_rate": 4.7061963775023833e-07, "loss": 2.9526, "step": 44272 }, { "epoch": 8.43, "learning_rate": 4.7056244041944707e-07, "loss": 1.5683, "step": 44273 }, { "epoch": 8.43, "learning_rate": 4.7050524308865587e-07, "loss": 2.4167, "step": 44274 }, { "epoch": 8.43, "learning_rate": 4.704480457578646e-07, "loss": 1.4338, "step": 44275 }, { "epoch": 8.43, "learning_rate": 4.7039084842707345e-07, "loss": 0.9465, "step": 44276 }, { "epoch": 8.43, "learning_rate": 4.703336510962822e-07, "loss": 1.7223, "step": 44277 }, { "epoch": 8.43, "learning_rate": 4.7027645376549094e-07, "loss": 1.3017, "step": 44278 }, { "epoch": 8.43, "learning_rate": 4.7021925643469973e-07, "loss": 1.7196, "step": 44279 }, { "epoch": 8.43, "learning_rate": 4.7016205910390847e-07, "loss": 1.5615, "step": 44280 }, { "epoch": 8.43, "learning_rate": 4.7010486177311726e-07, "loss": 0.8976, "step": 44281 }, { "epoch": 8.43, "learning_rate": 4.70047664442326e-07, "loss": 0.8187, "step": 44282 }, { "epoch": 8.43, "learning_rate": 4.6999046711153485e-07, "loss": 2.2991, "step": 44283 }, { "epoch": 8.44, "learning_rate": 4.699332697807436e-07, "loss": 0.9783, "step": 44284 }, { "epoch": 8.44, "learning_rate": 4.6987607244995233e-07, "loss": 0.9725, "step": 44285 }, { "epoch": 8.44, "learning_rate": 4.6981887511916113e-07, "loss": 1.2502, "step": 44286 }, { "epoch": 8.44, "learning_rate": 4.6976167778836987e-07, "loss": 1.1147, "step": 44287 }, { "epoch": 8.44, "learning_rate": 4.697044804575787e-07, "loss": 0.7375, "step": 44288 }, { "epoch": 8.44, "learning_rate": 4.696472831267874e-07, "loss": 0.8409, "step": 44289 }, { "epoch": 8.44, "learning_rate": 4.6959008579599614e-07, "loss": 1.4725, "step": 44290 }, { "epoch": 8.44, "learning_rate": 4.69532888465205e-07, "loss": 1.4548, "step": 44291 }, { "epoch": 8.44, "learning_rate": 4.6947569113441373e-07, "loss": 1.9664, "step": 44292 }, { "epoch": 8.44, "learning_rate": 4.694184938036225e-07, "loss": 2.8788, "step": 44293 }, { "epoch": 8.44, "learning_rate": 4.6936129647283127e-07, "loss": 1.6871, "step": 44294 }, { "epoch": 8.44, "learning_rate": 4.693040991420401e-07, "loss": 2.1192, "step": 44295 }, { "epoch": 8.44, "learning_rate": 4.692469018112488e-07, "loss": 0.9004, "step": 44296 }, { "epoch": 8.44, "learning_rate": 4.6918970448045754e-07, "loss": 1.4153, "step": 44297 }, { "epoch": 8.44, "learning_rate": 4.691325071496664e-07, "loss": 0.8316, "step": 44298 }, { "epoch": 8.44, "learning_rate": 4.6907530981887513e-07, "loss": 1.052, "step": 44299 }, { "epoch": 8.44, "learning_rate": 4.690181124880839e-07, "loss": 2.0338, "step": 44300 }, { "epoch": 8.44, "learning_rate": 4.6896091515729266e-07, "loss": 2.0281, "step": 44301 }, { "epoch": 8.44, "learning_rate": 4.689037178265014e-07, "loss": 1.6912, "step": 44302 }, { "epoch": 8.44, "learning_rate": 4.688465204957102e-07, "loss": 1.7302, "step": 44303 }, { "epoch": 8.44, "learning_rate": 4.6878932316491894e-07, "loss": 2.547, "step": 44304 }, { "epoch": 8.44, "learning_rate": 4.687321258341278e-07, "loss": 0.894, "step": 44305 }, { "epoch": 8.44, "learning_rate": 4.6867492850333653e-07, "loss": 0.4458, "step": 44306 }, { "epoch": 8.44, "learning_rate": 4.686177311725453e-07, "loss": 1.584, "step": 44307 }, { "epoch": 8.44, "learning_rate": 4.6856053384175406e-07, "loss": 0.8221, "step": 44308 }, { "epoch": 8.44, "learning_rate": 4.685033365109628e-07, "loss": 1.1566, "step": 44309 }, { "epoch": 8.44, "learning_rate": 4.684461391801716e-07, "loss": 1.0331, "step": 44310 }, { "epoch": 8.44, "learning_rate": 4.6838894184938034e-07, "loss": 1.8366, "step": 44311 }, { "epoch": 8.44, "learning_rate": 4.683317445185892e-07, "loss": 2.2638, "step": 44312 }, { "epoch": 8.44, "learning_rate": 4.682745471877979e-07, "loss": 1.1202, "step": 44313 }, { "epoch": 8.44, "learning_rate": 4.6821734985700667e-07, "loss": 2.3037, "step": 44314 }, { "epoch": 8.44, "learning_rate": 4.6816015252621546e-07, "loss": 2.373, "step": 44315 }, { "epoch": 8.44, "learning_rate": 4.681029551954242e-07, "loss": 1.6103, "step": 44316 }, { "epoch": 8.44, "learning_rate": 4.6804575786463305e-07, "loss": 1.8876, "step": 44317 }, { "epoch": 8.44, "learning_rate": 4.6798856053384174e-07, "loss": 2.2729, "step": 44318 }, { "epoch": 8.44, "learning_rate": 4.679313632030506e-07, "loss": 1.5086, "step": 44319 }, { "epoch": 8.44, "learning_rate": 4.678741658722593e-07, "loss": 0.2684, "step": 44320 }, { "epoch": 8.44, "learning_rate": 4.6781696854146806e-07, "loss": 1.5017, "step": 44321 }, { "epoch": 8.44, "learning_rate": 4.6775977121067686e-07, "loss": 1.0833, "step": 44322 }, { "epoch": 8.44, "learning_rate": 4.677025738798856e-07, "loss": 2.0936, "step": 44323 }, { "epoch": 8.44, "learning_rate": 4.6764537654909445e-07, "loss": 1.8627, "step": 44324 }, { "epoch": 8.44, "learning_rate": 4.6758817921830313e-07, "loss": 1.6864, "step": 44325 }, { "epoch": 8.44, "learning_rate": 4.675309818875119e-07, "loss": 1.1723, "step": 44326 }, { "epoch": 8.44, "learning_rate": 4.674737845567207e-07, "loss": 2.0798, "step": 44327 }, { "epoch": 8.44, "learning_rate": 4.6741658722592946e-07, "loss": 3.9816, "step": 44328 }, { "epoch": 8.44, "learning_rate": 4.6735938989513826e-07, "loss": 1.7313, "step": 44329 }, { "epoch": 8.44, "learning_rate": 4.67302192564347e-07, "loss": 1.7695, "step": 44330 }, { "epoch": 8.44, "learning_rate": 4.6724499523355574e-07, "loss": 2.5812, "step": 44331 }, { "epoch": 8.44, "learning_rate": 4.6718779790276453e-07, "loss": 3.8044, "step": 44332 }, { "epoch": 8.44, "learning_rate": 4.6713060057197327e-07, "loss": 1.3971, "step": 44333 }, { "epoch": 8.44, "learning_rate": 4.670734032411821e-07, "loss": 2.2279, "step": 44334 }, { "epoch": 8.44, "learning_rate": 4.6701620591039086e-07, "loss": 1.4959, "step": 44335 }, { "epoch": 8.44, "learning_rate": 4.6695900857959965e-07, "loss": 1.4035, "step": 44336 }, { "epoch": 8.45, "learning_rate": 4.669018112488084e-07, "loss": 2.1945, "step": 44337 }, { "epoch": 8.45, "learning_rate": 4.6684461391801714e-07, "loss": 1.547, "step": 44338 }, { "epoch": 8.45, "learning_rate": 4.66787416587226e-07, "loss": 2.4001, "step": 44339 }, { "epoch": 8.45, "learning_rate": 4.6673021925643467e-07, "loss": 1.7127, "step": 44340 }, { "epoch": 8.45, "learning_rate": 4.666730219256435e-07, "loss": 0.8979, "step": 44341 }, { "epoch": 8.45, "learning_rate": 4.6661582459485226e-07, "loss": 0.8316, "step": 44342 }, { "epoch": 8.45, "learning_rate": 4.66558627264061e-07, "loss": 1.284, "step": 44343 }, { "epoch": 8.45, "learning_rate": 4.665014299332698e-07, "loss": 0.7148, "step": 44344 }, { "epoch": 8.45, "learning_rate": 4.6644423260247853e-07, "loss": 2.0822, "step": 44345 }, { "epoch": 8.45, "learning_rate": 4.663870352716874e-07, "loss": 1.0025, "step": 44346 }, { "epoch": 8.45, "learning_rate": 4.6632983794089607e-07, "loss": 1.683, "step": 44347 }, { "epoch": 8.45, "learning_rate": 4.662726406101049e-07, "loss": 1.7699, "step": 44348 }, { "epoch": 8.45, "learning_rate": 4.6621544327931366e-07, "loss": 1.7381, "step": 44349 }, { "epoch": 8.45, "learning_rate": 4.661582459485224e-07, "loss": 0.3979, "step": 44350 }, { "epoch": 8.45, "learning_rate": 4.661010486177312e-07, "loss": 1.6899, "step": 44351 }, { "epoch": 8.45, "learning_rate": 4.6604385128693993e-07, "loss": 1.3729, "step": 44352 }, { "epoch": 8.45, "learning_rate": 4.659866539561488e-07, "loss": 0.751, "step": 44353 }, { "epoch": 8.45, "learning_rate": 4.6592945662535747e-07, "loss": 0.9015, "step": 44354 }, { "epoch": 8.45, "learning_rate": 4.658722592945662e-07, "loss": 1.1175, "step": 44355 }, { "epoch": 8.45, "learning_rate": 4.6581506196377505e-07, "loss": 1.2734, "step": 44356 }, { "epoch": 8.45, "learning_rate": 4.657578646329838e-07, "loss": 1.106, "step": 44357 }, { "epoch": 8.45, "learning_rate": 4.657006673021926e-07, "loss": 0.5685, "step": 44358 }, { "epoch": 8.45, "learning_rate": 4.6564346997140133e-07, "loss": 0.4569, "step": 44359 }, { "epoch": 8.45, "learning_rate": 4.655862726406102e-07, "loss": 2.1496, "step": 44360 }, { "epoch": 8.45, "learning_rate": 4.655290753098189e-07, "loss": 1.2027, "step": 44361 }, { "epoch": 8.45, "learning_rate": 4.654718779790276e-07, "loss": 1.1249, "step": 44362 }, { "epoch": 8.45, "learning_rate": 4.6541468064823645e-07, "loss": 1.6744, "step": 44363 }, { "epoch": 8.45, "learning_rate": 4.653574833174452e-07, "loss": 3.0811, "step": 44364 }, { "epoch": 8.45, "learning_rate": 4.65300285986654e-07, "loss": 1.5899, "step": 44365 }, { "epoch": 8.45, "learning_rate": 4.6524308865586273e-07, "loss": 1.0775, "step": 44366 }, { "epoch": 8.45, "learning_rate": 4.6518589132507147e-07, "loss": 1.6233, "step": 44367 }, { "epoch": 8.45, "learning_rate": 4.651286939942803e-07, "loss": 2.7644, "step": 44368 }, { "epoch": 8.45, "learning_rate": 4.65071496663489e-07, "loss": 1.2606, "step": 44369 }, { "epoch": 8.45, "learning_rate": 4.6501429933269785e-07, "loss": 1.759, "step": 44370 }, { "epoch": 8.45, "learning_rate": 4.649571020019066e-07, "loss": 0.3968, "step": 44371 }, { "epoch": 8.45, "learning_rate": 4.648999046711154e-07, "loss": 1.2537, "step": 44372 }, { "epoch": 8.45, "learning_rate": 4.648427073403241e-07, "loss": 1.7856, "step": 44373 }, { "epoch": 8.45, "learning_rate": 4.6478551000953287e-07, "loss": 2.9813, "step": 44374 }, { "epoch": 8.45, "learning_rate": 4.647283126787417e-07, "loss": 0.5998, "step": 44375 }, { "epoch": 8.45, "learning_rate": 4.646711153479504e-07, "loss": 2.5643, "step": 44376 }, { "epoch": 8.45, "learning_rate": 4.6461391801715925e-07, "loss": 1.551, "step": 44377 }, { "epoch": 8.45, "learning_rate": 4.64556720686368e-07, "loss": 0.9328, "step": 44378 }, { "epoch": 8.45, "learning_rate": 4.6449952335557673e-07, "loss": 2.0817, "step": 44379 }, { "epoch": 8.45, "learning_rate": 4.644423260247855e-07, "loss": 0.9194, "step": 44380 }, { "epoch": 8.45, "learning_rate": 4.6438512869399426e-07, "loss": 2.002, "step": 44381 }, { "epoch": 8.45, "learning_rate": 4.643279313632031e-07, "loss": 1.787, "step": 44382 }, { "epoch": 8.45, "learning_rate": 4.6427073403241185e-07, "loss": 1.1623, "step": 44383 }, { "epoch": 8.45, "learning_rate": 4.6421353670162065e-07, "loss": 2.3985, "step": 44384 }, { "epoch": 8.45, "learning_rate": 4.641563393708294e-07, "loss": 1.4728, "step": 44385 }, { "epoch": 8.45, "learning_rate": 4.6409914204003813e-07, "loss": 1.7038, "step": 44386 }, { "epoch": 8.45, "learning_rate": 4.640419447092469e-07, "loss": 1.2448, "step": 44387 }, { "epoch": 8.45, "learning_rate": 4.6398474737845566e-07, "loss": 1.8239, "step": 44388 }, { "epoch": 8.46, "learning_rate": 4.639275500476645e-07, "loss": 2.1994, "step": 44389 }, { "epoch": 8.46, "learning_rate": 4.6387035271687325e-07, "loss": 1.5569, "step": 44390 }, { "epoch": 8.46, "learning_rate": 4.6381315538608194e-07, "loss": 0.686, "step": 44391 }, { "epoch": 8.46, "learning_rate": 4.637559580552908e-07, "loss": 2.1649, "step": 44392 }, { "epoch": 8.46, "learning_rate": 4.636987607244995e-07, "loss": 0.8886, "step": 44393 }, { "epoch": 8.46, "learning_rate": 4.636415633937083e-07, "loss": 0.4082, "step": 44394 }, { "epoch": 8.46, "learning_rate": 4.6358436606291706e-07, "loss": 1.9935, "step": 44395 }, { "epoch": 8.46, "learning_rate": 4.635271687321258e-07, "loss": 1.6904, "step": 44396 }, { "epoch": 8.46, "learning_rate": 4.6346997140133465e-07, "loss": 0.4932, "step": 44397 }, { "epoch": 8.46, "learning_rate": 4.6341277407054334e-07, "loss": 2.7568, "step": 44398 }, { "epoch": 8.46, "learning_rate": 4.633555767397522e-07, "loss": 1.1043, "step": 44399 }, { "epoch": 8.46, "learning_rate": 4.632983794089609e-07, "loss": 0.5773, "step": 44400 }, { "epoch": 8.46, "learning_rate": 4.632411820781697e-07, "loss": 0.9905, "step": 44401 }, { "epoch": 8.46, "learning_rate": 4.6318398474737846e-07, "loss": 1.2503, "step": 44402 }, { "epoch": 8.46, "learning_rate": 4.631267874165872e-07, "loss": 2.019, "step": 44403 }, { "epoch": 8.46, "learning_rate": 4.6306959008579605e-07, "loss": 1.0692, "step": 44404 }, { "epoch": 8.46, "learning_rate": 4.630123927550048e-07, "loss": 0.9895, "step": 44405 }, { "epoch": 8.46, "learning_rate": 4.629551954242136e-07, "loss": 0.1852, "step": 44406 }, { "epoch": 8.46, "learning_rate": 4.628979980934223e-07, "loss": 1.3281, "step": 44407 }, { "epoch": 8.46, "learning_rate": 4.6284080076263106e-07, "loss": 2.7044, "step": 44408 }, { "epoch": 8.46, "learning_rate": 4.6278360343183986e-07, "loss": 1.7108, "step": 44409 }, { "epoch": 8.46, "learning_rate": 4.627264061010486e-07, "loss": 1.8969, "step": 44410 }, { "epoch": 8.46, "learning_rate": 4.6266920877025744e-07, "loss": 0.964, "step": 44411 }, { "epoch": 8.46, "learning_rate": 4.626120114394662e-07, "loss": 1.3281, "step": 44412 }, { "epoch": 8.46, "learning_rate": 4.62554814108675e-07, "loss": 2.1165, "step": 44413 }, { "epoch": 8.46, "learning_rate": 4.624976167778837e-07, "loss": 2.4235, "step": 44414 }, { "epoch": 8.46, "learning_rate": 4.6244041944709246e-07, "loss": 1.9117, "step": 44415 }, { "epoch": 8.46, "learning_rate": 4.6238322211630125e-07, "loss": 1.3163, "step": 44416 }, { "epoch": 8.46, "learning_rate": 4.6232602478551e-07, "loss": 0.9168, "step": 44417 }, { "epoch": 8.46, "learning_rate": 4.6226882745471884e-07, "loss": 0.7008, "step": 44418 }, { "epoch": 8.46, "learning_rate": 4.622116301239276e-07, "loss": 1.7586, "step": 44419 }, { "epoch": 8.46, "learning_rate": 4.6215443279313627e-07, "loss": 2.9073, "step": 44420 }, { "epoch": 8.46, "learning_rate": 4.620972354623451e-07, "loss": 1.0407, "step": 44421 }, { "epoch": 8.46, "learning_rate": 4.6204003813155386e-07, "loss": 1.7001, "step": 44422 }, { "epoch": 8.46, "learning_rate": 4.6198284080076265e-07, "loss": 2.8636, "step": 44423 }, { "epoch": 8.46, "learning_rate": 4.619256434699714e-07, "loss": 1.3329, "step": 44424 }, { "epoch": 8.46, "learning_rate": 4.6186844613918024e-07, "loss": 0.7335, "step": 44425 }, { "epoch": 8.46, "learning_rate": 4.61811248808389e-07, "loss": 2.5132, "step": 44426 }, { "epoch": 8.46, "learning_rate": 4.6175405147759767e-07, "loss": 2.3917, "step": 44427 }, { "epoch": 8.46, "learning_rate": 4.616968541468065e-07, "loss": 2.1503, "step": 44428 }, { "epoch": 8.46, "learning_rate": 4.6163965681601526e-07, "loss": 1.1474, "step": 44429 }, { "epoch": 8.46, "learning_rate": 4.6158245948522405e-07, "loss": 1.1901, "step": 44430 }, { "epoch": 8.46, "learning_rate": 4.615252621544328e-07, "loss": 0.6574, "step": 44431 }, { "epoch": 8.46, "learning_rate": 4.6146806482364153e-07, "loss": 1.0047, "step": 44432 }, { "epoch": 8.46, "learning_rate": 4.614108674928504e-07, "loss": 2.5726, "step": 44433 }, { "epoch": 8.46, "learning_rate": 4.613536701620591e-07, "loss": 0.994, "step": 44434 }, { "epoch": 8.46, "learning_rate": 4.612964728312679e-07, "loss": 2.9452, "step": 44435 }, { "epoch": 8.46, "learning_rate": 4.6123927550047665e-07, "loss": 0.9484, "step": 44436 }, { "epoch": 8.46, "learning_rate": 4.6118207816968545e-07, "loss": 1.3186, "step": 44437 }, { "epoch": 8.46, "learning_rate": 4.611248808388942e-07, "loss": 0.3849, "step": 44438 }, { "epoch": 8.46, "learning_rate": 4.6106768350810293e-07, "loss": 2.4852, "step": 44439 }, { "epoch": 8.46, "learning_rate": 4.610104861773118e-07, "loss": 1.8212, "step": 44440 }, { "epoch": 8.46, "learning_rate": 4.609532888465205e-07, "loss": 2.9042, "step": 44441 }, { "epoch": 8.47, "learning_rate": 4.608960915157293e-07, "loss": 1.4145, "step": 44442 }, { "epoch": 8.47, "learning_rate": 4.6083889418493805e-07, "loss": 1.7022, "step": 44443 }, { "epoch": 8.47, "learning_rate": 4.607816968541468e-07, "loss": 1.2991, "step": 44444 }, { "epoch": 8.47, "learning_rate": 4.607244995233556e-07, "loss": 2.3846, "step": 44445 }, { "epoch": 8.47, "learning_rate": 4.6066730219256433e-07, "loss": 1.1295, "step": 44446 }, { "epoch": 8.47, "learning_rate": 4.6061010486177317e-07, "loss": 1.1201, "step": 44447 }, { "epoch": 8.47, "learning_rate": 4.605529075309819e-07, "loss": 0.8264, "step": 44448 }, { "epoch": 8.47, "learning_rate": 4.604957102001906e-07, "loss": 2.2804, "step": 44449 }, { "epoch": 8.47, "learning_rate": 4.6043851286939945e-07, "loss": 2.0942, "step": 44450 }, { "epoch": 8.47, "learning_rate": 4.603813155386082e-07, "loss": 1.1441, "step": 44451 }, { "epoch": 8.47, "learning_rate": 4.60324118207817e-07, "loss": 1.1454, "step": 44452 }, { "epoch": 8.47, "learning_rate": 4.602669208770257e-07, "loss": 1.8218, "step": 44453 }, { "epoch": 8.47, "learning_rate": 4.6020972354623457e-07, "loss": 2.4613, "step": 44454 }, { "epoch": 8.47, "learning_rate": 4.601525262154433e-07, "loss": 1.2981, "step": 44455 }, { "epoch": 8.47, "learning_rate": 4.6009532888465205e-07, "loss": 2.017, "step": 44456 }, { "epoch": 8.47, "learning_rate": 4.6003813155386085e-07, "loss": 0.9296, "step": 44457 }, { "epoch": 8.47, "learning_rate": 4.599809342230696e-07, "loss": 0.8643, "step": 44458 }, { "epoch": 8.47, "learning_rate": 4.599237368922784e-07, "loss": 1.3037, "step": 44459 }, { "epoch": 8.47, "learning_rate": 4.598665395614871e-07, "loss": 2.2911, "step": 44460 }, { "epoch": 8.47, "learning_rate": 4.5980934223069586e-07, "loss": 0.5972, "step": 44461 }, { "epoch": 8.47, "learning_rate": 4.597521448999047e-07, "loss": 1.3658, "step": 44462 }, { "epoch": 8.47, "learning_rate": 4.5969494756911345e-07, "loss": 1.7146, "step": 44463 }, { "epoch": 8.47, "learning_rate": 4.5963775023832224e-07, "loss": 1.5834, "step": 44464 }, { "epoch": 8.47, "learning_rate": 4.59580552907531e-07, "loss": 0.8542, "step": 44465 }, { "epoch": 8.47, "learning_rate": 4.595233555767398e-07, "loss": 1.6646, "step": 44466 }, { "epoch": 8.47, "learning_rate": 4.594661582459485e-07, "loss": 2.019, "step": 44467 }, { "epoch": 8.47, "learning_rate": 4.5940896091515726e-07, "loss": 1.6733, "step": 44468 }, { "epoch": 8.47, "learning_rate": 4.593517635843661e-07, "loss": 3.054, "step": 44469 }, { "epoch": 8.47, "learning_rate": 4.5929456625357485e-07, "loss": 1.0658, "step": 44470 }, { "epoch": 8.47, "learning_rate": 4.5923736892278364e-07, "loss": 1.2669, "step": 44471 }, { "epoch": 8.47, "learning_rate": 4.591801715919924e-07, "loss": 2.2101, "step": 44472 }, { "epoch": 8.47, "learning_rate": 4.591229742612011e-07, "loss": 1.9205, "step": 44473 }, { "epoch": 8.47, "learning_rate": 4.590657769304099e-07, "loss": 0.677, "step": 44474 }, { "epoch": 8.47, "learning_rate": 4.5900857959961866e-07, "loss": 1.2215, "step": 44475 }, { "epoch": 8.47, "learning_rate": 4.589513822688275e-07, "loss": 0.5178, "step": 44476 }, { "epoch": 8.47, "learning_rate": 4.5889418493803625e-07, "loss": 1.3671, "step": 44477 }, { "epoch": 8.47, "learning_rate": 4.5883698760724504e-07, "loss": 1.5703, "step": 44478 }, { "epoch": 8.47, "learning_rate": 4.587797902764538e-07, "loss": 1.2985, "step": 44479 }, { "epoch": 8.47, "learning_rate": 4.587225929456625e-07, "loss": 1.8417, "step": 44480 }, { "epoch": 8.47, "learning_rate": 4.586653956148713e-07, "loss": 1.4339, "step": 44481 }, { "epoch": 8.47, "learning_rate": 4.5860819828408006e-07, "loss": 1.1793, "step": 44482 }, { "epoch": 8.47, "learning_rate": 4.585510009532889e-07, "loss": 1.024, "step": 44483 }, { "epoch": 8.47, "learning_rate": 4.5849380362249764e-07, "loss": 3.3917, "step": 44484 }, { "epoch": 8.47, "learning_rate": 4.584366062917064e-07, "loss": 2.5451, "step": 44485 }, { "epoch": 8.47, "learning_rate": 4.583794089609152e-07, "loss": 2.2361, "step": 44486 }, { "epoch": 8.47, "learning_rate": 4.583222116301239e-07, "loss": 1.4121, "step": 44487 }, { "epoch": 8.47, "learning_rate": 4.582650142993327e-07, "loss": 3.4618, "step": 44488 }, { "epoch": 8.47, "learning_rate": 4.5820781696854145e-07, "loss": 1.7525, "step": 44489 }, { "epoch": 8.47, "learning_rate": 4.581506196377503e-07, "loss": 1.0887, "step": 44490 }, { "epoch": 8.47, "learning_rate": 4.5809342230695904e-07, "loss": 1.1875, "step": 44491 }, { "epoch": 8.47, "learning_rate": 4.580362249761678e-07, "loss": 1.3683, "step": 44492 }, { "epoch": 8.47, "learning_rate": 4.579790276453766e-07, "loss": 1.8943, "step": 44493 }, { "epoch": 8.48, "learning_rate": 4.579218303145853e-07, "loss": 1.8164, "step": 44494 }, { "epoch": 8.48, "learning_rate": 4.578646329837941e-07, "loss": 1.0096, "step": 44495 }, { "epoch": 8.48, "learning_rate": 4.5780743565300285e-07, "loss": 2.1644, "step": 44496 }, { "epoch": 8.48, "learning_rate": 4.577502383222116e-07, "loss": 0.9638, "step": 44497 }, { "epoch": 8.48, "learning_rate": 4.5769304099142044e-07, "loss": 1.4248, "step": 44498 }, { "epoch": 8.48, "learning_rate": 4.576358436606292e-07, "loss": 1.8872, "step": 44499 }, { "epoch": 8.48, "learning_rate": 4.57578646329838e-07, "loss": 1.4732, "step": 44500 }, { "epoch": 8.48, "learning_rate": 4.575214489990467e-07, "loss": 2.2537, "step": 44501 }, { "epoch": 8.48, "learning_rate": 4.574642516682555e-07, "loss": 1.2248, "step": 44502 }, { "epoch": 8.48, "learning_rate": 4.5740705433746425e-07, "loss": 1.3832, "step": 44503 }, { "epoch": 8.48, "learning_rate": 4.57349857006673e-07, "loss": 1.3389, "step": 44504 }, { "epoch": 8.48, "learning_rate": 4.5729265967588184e-07, "loss": 2.0047, "step": 44505 }, { "epoch": 8.48, "learning_rate": 4.572354623450906e-07, "loss": 2.3449, "step": 44506 }, { "epoch": 8.48, "learning_rate": 4.5717826501429937e-07, "loss": 1.6979, "step": 44507 }, { "epoch": 8.48, "learning_rate": 4.571210676835081e-07, "loss": 1.1158, "step": 44508 }, { "epoch": 8.48, "learning_rate": 4.5706387035271685e-07, "loss": 1.1873, "step": 44509 }, { "epoch": 8.48, "learning_rate": 4.5700667302192565e-07, "loss": 1.6352, "step": 44510 }, { "epoch": 8.48, "learning_rate": 4.569494756911344e-07, "loss": 1.6758, "step": 44511 }, { "epoch": 8.48, "learning_rate": 4.5689227836034324e-07, "loss": 1.837, "step": 44512 }, { "epoch": 8.48, "learning_rate": 4.56835081029552e-07, "loss": 1.4765, "step": 44513 }, { "epoch": 8.48, "learning_rate": 4.567778836987607e-07, "loss": 2.1757, "step": 44514 }, { "epoch": 8.48, "learning_rate": 4.567206863679695e-07, "loss": 1.5757, "step": 44515 }, { "epoch": 8.48, "learning_rate": 4.5666348903717825e-07, "loss": 1.8436, "step": 44516 }, { "epoch": 8.48, "learning_rate": 4.5660629170638705e-07, "loss": 0.8686, "step": 44517 }, { "epoch": 8.48, "learning_rate": 4.565490943755958e-07, "loss": 0.889, "step": 44518 }, { "epoch": 8.48, "learning_rate": 4.5649189704480463e-07, "loss": 1.1943, "step": 44519 }, { "epoch": 8.48, "learning_rate": 4.564346997140134e-07, "loss": 0.7989, "step": 44520 }, { "epoch": 8.48, "learning_rate": 4.563775023832221e-07, "loss": 2.0949, "step": 44521 }, { "epoch": 8.48, "learning_rate": 4.563203050524309e-07, "loss": 2.6883, "step": 44522 }, { "epoch": 8.48, "learning_rate": 4.5626310772163965e-07, "loss": 0.556, "step": 44523 }, { "epoch": 8.48, "learning_rate": 4.5620591039084844e-07, "loss": 2.0299, "step": 44524 }, { "epoch": 8.48, "learning_rate": 4.561487130600572e-07, "loss": 2.6348, "step": 44525 }, { "epoch": 8.48, "learning_rate": 4.560915157292659e-07, "loss": 2.3618, "step": 44526 }, { "epoch": 8.48, "learning_rate": 4.5603431839847477e-07, "loss": 1.0053, "step": 44527 }, { "epoch": 8.48, "learning_rate": 4.559771210676835e-07, "loss": 2.0813, "step": 44528 }, { "epoch": 8.48, "learning_rate": 4.559199237368923e-07, "loss": 0.7617, "step": 44529 }, { "epoch": 8.48, "learning_rate": 4.5586272640610105e-07, "loss": 1.4554, "step": 44530 }, { "epoch": 8.48, "learning_rate": 4.5580552907530984e-07, "loss": 0.5796, "step": 44531 }, { "epoch": 8.48, "learning_rate": 4.557483317445186e-07, "loss": 1.3317, "step": 44532 }, { "epoch": 8.48, "learning_rate": 4.556911344137273e-07, "loss": 1.4834, "step": 44533 }, { "epoch": 8.48, "learning_rate": 4.5563393708293617e-07, "loss": 1.0942, "step": 44534 }, { "epoch": 8.48, "learning_rate": 4.555767397521449e-07, "loss": 1.7214, "step": 44535 }, { "epoch": 8.48, "learning_rate": 4.555195424213537e-07, "loss": 1.7057, "step": 44536 }, { "epoch": 8.48, "learning_rate": 4.5546234509056245e-07, "loss": 1.0205, "step": 44537 }, { "epoch": 8.48, "learning_rate": 4.554051477597712e-07, "loss": 1.3472, "step": 44538 }, { "epoch": 8.48, "learning_rate": 4.5534795042898e-07, "loss": 1.7732, "step": 44539 }, { "epoch": 8.48, "learning_rate": 4.552907530981887e-07, "loss": 1.6017, "step": 44540 }, { "epoch": 8.48, "learning_rate": 4.5523355576739757e-07, "loss": 3.7499, "step": 44541 }, { "epoch": 8.48, "learning_rate": 4.551763584366063e-07, "loss": 0.999, "step": 44542 }, { "epoch": 8.48, "learning_rate": 4.551191611058151e-07, "loss": 1.4937, "step": 44543 }, { "epoch": 8.48, "learning_rate": 4.5506196377502384e-07, "loss": 0.6004, "step": 44544 }, { "epoch": 8.48, "learning_rate": 4.550047664442326e-07, "loss": 1.6421, "step": 44545 }, { "epoch": 8.48, "learning_rate": 4.549475691134414e-07, "loss": 1.3059, "step": 44546 }, { "epoch": 8.49, "learning_rate": 4.548903717826501e-07, "loss": 2.6363, "step": 44547 }, { "epoch": 8.49, "learning_rate": 4.5483317445185897e-07, "loss": 0.4257, "step": 44548 }, { "epoch": 8.49, "learning_rate": 4.547759771210677e-07, "loss": 1.5676, "step": 44549 }, { "epoch": 8.49, "learning_rate": 4.5471877979027645e-07, "loss": 1.6676, "step": 44550 }, { "epoch": 8.49, "learning_rate": 4.5466158245948524e-07, "loss": 2.1232, "step": 44551 }, { "epoch": 8.49, "learning_rate": 4.54604385128694e-07, "loss": 1.1315, "step": 44552 }, { "epoch": 8.49, "learning_rate": 4.545471877979028e-07, "loss": 1.0815, "step": 44553 }, { "epoch": 8.49, "learning_rate": 4.544899904671115e-07, "loss": 2.5175, "step": 44554 }, { "epoch": 8.49, "learning_rate": 4.5443279313632036e-07, "loss": 1.4434, "step": 44555 }, { "epoch": 8.49, "learning_rate": 4.543755958055291e-07, "loss": 1.6091, "step": 44556 }, { "epoch": 8.49, "learning_rate": 4.5431839847473785e-07, "loss": 1.5919, "step": 44557 }, { "epoch": 8.49, "learning_rate": 4.5426120114394664e-07, "loss": 2.7682, "step": 44558 }, { "epoch": 8.49, "learning_rate": 4.542040038131554e-07, "loss": 1.7936, "step": 44559 }, { "epoch": 8.49, "learning_rate": 4.5414680648236423e-07, "loss": 2.2321, "step": 44560 }, { "epoch": 8.49, "learning_rate": 4.540896091515729e-07, "loss": 0.5176, "step": 44561 }, { "epoch": 8.49, "learning_rate": 4.5403241182078166e-07, "loss": 0.6267, "step": 44562 }, { "epoch": 8.49, "learning_rate": 4.539752144899905e-07, "loss": 0.9718, "step": 44563 }, { "epoch": 8.49, "learning_rate": 4.5391801715919924e-07, "loss": 2.0258, "step": 44564 }, { "epoch": 8.49, "learning_rate": 4.5386081982840804e-07, "loss": 1.3701, "step": 44565 }, { "epoch": 8.49, "learning_rate": 4.538036224976168e-07, "loss": 2.1241, "step": 44566 }, { "epoch": 8.49, "learning_rate": 4.537464251668255e-07, "loss": 0.3652, "step": 44567 }, { "epoch": 8.49, "learning_rate": 4.536892278360343e-07, "loss": 1.5229, "step": 44568 }, { "epoch": 8.49, "learning_rate": 4.5363203050524305e-07, "loss": 1.0936, "step": 44569 }, { "epoch": 8.49, "learning_rate": 4.535748331744519e-07, "loss": 2.112, "step": 44570 }, { "epoch": 8.49, "learning_rate": 4.5351763584366064e-07, "loss": 2.0502, "step": 44571 }, { "epoch": 8.49, "learning_rate": 4.5346043851286944e-07, "loss": 0.9422, "step": 44572 }, { "epoch": 8.49, "learning_rate": 4.534032411820782e-07, "loss": 2.0268, "step": 44573 }, { "epoch": 8.49, "learning_rate": 4.533460438512869e-07, "loss": 3.5744, "step": 44574 }, { "epoch": 8.49, "learning_rate": 4.532888465204957e-07, "loss": 1.6922, "step": 44575 }, { "epoch": 8.49, "learning_rate": 4.5323164918970445e-07, "loss": 1.9868, "step": 44576 }, { "epoch": 8.49, "learning_rate": 4.531744518589133e-07, "loss": 0.8949, "step": 44577 }, { "epoch": 8.49, "learning_rate": 4.5311725452812204e-07, "loss": 1.4208, "step": 44578 }, { "epoch": 8.49, "learning_rate": 4.530600571973308e-07, "loss": 1.6769, "step": 44579 }, { "epoch": 8.49, "learning_rate": 4.530028598665396e-07, "loss": 3.3099, "step": 44580 }, { "epoch": 8.49, "learning_rate": 4.529456625357483e-07, "loss": 1.1552, "step": 44581 }, { "epoch": 8.49, "learning_rate": 4.5288846520495716e-07, "loss": 2.0008, "step": 44582 }, { "epoch": 8.49, "learning_rate": 4.5283126787416585e-07, "loss": 1.1233, "step": 44583 }, { "epoch": 8.49, "learning_rate": 4.527740705433747e-07, "loss": 2.2153, "step": 44584 }, { "epoch": 8.49, "learning_rate": 4.5271687321258344e-07, "loss": 0.2343, "step": 44585 }, { "epoch": 8.49, "learning_rate": 4.526596758817922e-07, "loss": 1.2918, "step": 44586 }, { "epoch": 8.49, "learning_rate": 4.5260247855100097e-07, "loss": 1.6398, "step": 44587 }, { "epoch": 8.49, "learning_rate": 4.525452812202097e-07, "loss": 1.8621, "step": 44588 }, { "epoch": 8.49, "learning_rate": 4.5248808388941856e-07, "loss": 2.5777, "step": 44589 }, { "epoch": 8.49, "learning_rate": 4.5243088655862725e-07, "loss": 1.5803, "step": 44590 }, { "epoch": 8.49, "learning_rate": 4.52373689227836e-07, "loss": 1.8913, "step": 44591 }, { "epoch": 8.49, "learning_rate": 4.5231649189704484e-07, "loss": 2.4827, "step": 44592 }, { "epoch": 8.49, "learning_rate": 4.522592945662536e-07, "loss": 2.0549, "step": 44593 }, { "epoch": 8.49, "learning_rate": 4.5220209723546237e-07, "loss": 1.9342, "step": 44594 }, { "epoch": 8.49, "learning_rate": 4.521448999046711e-07, "loss": 0.8509, "step": 44595 }, { "epoch": 8.49, "learning_rate": 4.5208770257387996e-07, "loss": 1.7837, "step": 44596 }, { "epoch": 8.49, "learning_rate": 4.5203050524308865e-07, "loss": 1.3188, "step": 44597 }, { "epoch": 8.49, "learning_rate": 4.519733079122974e-07, "loss": 1.2429, "step": 44598 }, { "epoch": 8.5, "learning_rate": 4.5191611058150623e-07, "loss": 1.5207, "step": 44599 }, { "epoch": 8.5, "learning_rate": 4.51858913250715e-07, "loss": 1.1099, "step": 44600 }, { "epoch": 8.5, "learning_rate": 4.5180171591992377e-07, "loss": 1.4416, "step": 44601 }, { "epoch": 8.5, "learning_rate": 4.517445185891325e-07, "loss": 0.2968, "step": 44602 }, { "epoch": 8.5, "learning_rate": 4.5168732125834125e-07, "loss": 1.1382, "step": 44603 }, { "epoch": 8.5, "learning_rate": 4.5163012392755004e-07, "loss": 0.9212, "step": 44604 }, { "epoch": 8.5, "learning_rate": 4.515729265967588e-07, "loss": 1.14, "step": 44605 }, { "epoch": 8.5, "learning_rate": 4.5151572926596763e-07, "loss": 1.9055, "step": 44606 }, { "epoch": 8.5, "learning_rate": 4.5145853193517637e-07, "loss": 2.6419, "step": 44607 }, { "epoch": 8.5, "learning_rate": 4.5140133460438517e-07, "loss": 0.9344, "step": 44608 }, { "epoch": 8.5, "learning_rate": 4.513441372735939e-07, "loss": 3.1704, "step": 44609 }, { "epoch": 8.5, "learning_rate": 4.5128693994280265e-07, "loss": 1.0628, "step": 44610 }, { "epoch": 8.5, "learning_rate": 4.512297426120115e-07, "loss": 0.1232, "step": 44611 }, { "epoch": 8.5, "learning_rate": 4.511725452812202e-07, "loss": 1.0216, "step": 44612 }, { "epoch": 8.5, "learning_rate": 4.5111534795042903e-07, "loss": 1.1065, "step": 44613 }, { "epoch": 8.5, "learning_rate": 4.5105815061963777e-07, "loss": 1.9179, "step": 44614 }, { "epoch": 8.5, "learning_rate": 4.510009532888465e-07, "loss": 1.4492, "step": 44615 }, { "epoch": 8.5, "learning_rate": 4.509437559580553e-07, "loss": 1.4356, "step": 44616 }, { "epoch": 8.5, "learning_rate": 4.5088655862726405e-07, "loss": 0.859, "step": 44617 }, { "epoch": 8.5, "learning_rate": 4.508293612964729e-07, "loss": 1.7386, "step": 44618 }, { "epoch": 8.5, "learning_rate": 4.507721639656816e-07, "loss": 1.5163, "step": 44619 }, { "epoch": 8.5, "learning_rate": 4.5071496663489043e-07, "loss": 1.4271, "step": 44620 }, { "epoch": 8.5, "learning_rate": 4.5065776930409917e-07, "loss": 2.76, "step": 44621 }, { "epoch": 8.5, "learning_rate": 4.506005719733079e-07, "loss": 1.5227, "step": 44622 }, { "epoch": 8.5, "learning_rate": 4.505433746425167e-07, "loss": 2.524, "step": 44623 }, { "epoch": 8.5, "learning_rate": 4.5048617731172544e-07, "loss": 1.3609, "step": 44624 }, { "epoch": 8.5, "learning_rate": 4.504289799809343e-07, "loss": 1.6373, "step": 44625 }, { "epoch": 8.5, "learning_rate": 4.50371782650143e-07, "loss": 2.6406, "step": 44626 }, { "epoch": 8.5, "learning_rate": 4.503145853193517e-07, "loss": 1.2565, "step": 44627 }, { "epoch": 8.5, "learning_rate": 4.5025738798856057e-07, "loss": 1.8064, "step": 44628 }, { "epoch": 8.5, "learning_rate": 4.502001906577693e-07, "loss": 2.4389, "step": 44629 }, { "epoch": 8.5, "learning_rate": 4.501429933269781e-07, "loss": 1.5284, "step": 44630 }, { "epoch": 8.5, "learning_rate": 4.5008579599618684e-07, "loss": 1.4734, "step": 44631 }, { "epoch": 8.5, "learning_rate": 4.500285986653956e-07, "loss": 2.3167, "step": 44632 }, { "epoch": 8.5, "learning_rate": 4.4997140133460443e-07, "loss": 2.3374, "step": 44633 }, { "epoch": 8.5, "learning_rate": 4.499142040038131e-07, "loss": 0.5226, "step": 44634 }, { "epoch": 8.5, "learning_rate": 4.4985700667302196e-07, "loss": 2.513, "step": 44635 }, { "epoch": 8.5, "learning_rate": 4.497998093422307e-07, "loss": 1.0435, "step": 44636 }, { "epoch": 8.5, "learning_rate": 4.497426120114395e-07, "loss": 1.4085, "step": 44637 }, { "epoch": 8.5, "learning_rate": 4.4968541468064824e-07, "loss": 1.9435, "step": 44638 }, { "epoch": 8.5, "learning_rate": 4.49628217349857e-07, "loss": 1.5762, "step": 44639 }, { "epoch": 8.5, "learning_rate": 4.4957102001906583e-07, "loss": 1.3829, "step": 44640 }, { "epoch": 8.5, "learning_rate": 4.495138226882745e-07, "loss": 2.1993, "step": 44641 }, { "epoch": 8.5, "learning_rate": 4.4945662535748336e-07, "loss": 1.8926, "step": 44642 }, { "epoch": 8.5, "learning_rate": 4.493994280266921e-07, "loss": 1.5501, "step": 44643 }, { "epoch": 8.5, "learning_rate": 4.4934223069590084e-07, "loss": 2.9031, "step": 44644 }, { "epoch": 8.5, "learning_rate": 4.4928503336510964e-07, "loss": 1.4608, "step": 44645 }, { "epoch": 8.5, "learning_rate": 4.492278360343184e-07, "loss": 0.8061, "step": 44646 }, { "epoch": 8.5, "learning_rate": 4.491706387035272e-07, "loss": 1.4084, "step": 44647 }, { "epoch": 8.5, "learning_rate": 4.491134413727359e-07, "loss": 2.7826, "step": 44648 }, { "epoch": 8.5, "learning_rate": 4.4905624404194476e-07, "loss": 0.8638, "step": 44649 }, { "epoch": 8.5, "learning_rate": 4.489990467111535e-07, "loss": 1.3003, "step": 44650 }, { "epoch": 8.5, "learning_rate": 4.4894184938036224e-07, "loss": 2.2103, "step": 44651 }, { "epoch": 8.51, "learning_rate": 4.4888465204957104e-07, "loss": 1.1729, "step": 44652 }, { "epoch": 8.51, "learning_rate": 4.488274547187798e-07, "loss": 1.1931, "step": 44653 }, { "epoch": 8.51, "learning_rate": 4.487702573879886e-07, "loss": 2.0644, "step": 44654 }, { "epoch": 8.51, "learning_rate": 4.4871306005719736e-07, "loss": 2.024, "step": 44655 }, { "epoch": 8.51, "learning_rate": 4.4865586272640605e-07, "loss": 2.7361, "step": 44656 }, { "epoch": 8.51, "learning_rate": 4.485986653956149e-07, "loss": 1.6245, "step": 44657 }, { "epoch": 8.51, "learning_rate": 4.4854146806482364e-07, "loss": 1.4631, "step": 44658 }, { "epoch": 8.51, "learning_rate": 4.4848427073403243e-07, "loss": 1.7586, "step": 44659 }, { "epoch": 8.51, "learning_rate": 4.484270734032412e-07, "loss": 0.9014, "step": 44660 }, { "epoch": 8.51, "learning_rate": 4.4836987607245e-07, "loss": 1.6318, "step": 44661 }, { "epoch": 8.51, "learning_rate": 4.4831267874165876e-07, "loss": 1.6462, "step": 44662 }, { "epoch": 8.51, "learning_rate": 4.4825548141086745e-07, "loss": 0.5019, "step": 44663 }, { "epoch": 8.51, "learning_rate": 4.481982840800763e-07, "loss": 1.5616, "step": 44664 }, { "epoch": 8.51, "learning_rate": 4.4814108674928504e-07, "loss": 1.0123, "step": 44665 }, { "epoch": 8.51, "learning_rate": 4.4808388941849383e-07, "loss": 1.8044, "step": 44666 }, { "epoch": 8.51, "learning_rate": 4.4802669208770257e-07, "loss": 1.0025, "step": 44667 }, { "epoch": 8.51, "learning_rate": 4.479694947569113e-07, "loss": 0.411, "step": 44668 }, { "epoch": 8.51, "learning_rate": 4.4791229742612016e-07, "loss": 0.6915, "step": 44669 }, { "epoch": 8.51, "learning_rate": 4.4785510009532885e-07, "loss": 1.295, "step": 44670 }, { "epoch": 8.51, "learning_rate": 4.477979027645377e-07, "loss": 0.2357, "step": 44671 }, { "epoch": 8.51, "learning_rate": 4.4774070543374644e-07, "loss": 2.4256, "step": 44672 }, { "epoch": 8.51, "learning_rate": 4.4768350810295523e-07, "loss": 0.8318, "step": 44673 }, { "epoch": 8.51, "learning_rate": 4.4762631077216397e-07, "loss": 2.5319, "step": 44674 }, { "epoch": 8.51, "learning_rate": 4.475691134413727e-07, "loss": 1.5721, "step": 44675 }, { "epoch": 8.51, "learning_rate": 4.4751191611058156e-07, "loss": 2.3795, "step": 44676 }, { "epoch": 8.51, "learning_rate": 4.474547187797903e-07, "loss": 1.3231, "step": 44677 }, { "epoch": 8.51, "learning_rate": 4.473975214489991e-07, "loss": 1.6146, "step": 44678 }, { "epoch": 8.51, "learning_rate": 4.4734032411820783e-07, "loss": 1.2297, "step": 44679 }, { "epoch": 8.51, "learning_rate": 4.472831267874166e-07, "loss": 3.2836, "step": 44680 }, { "epoch": 8.51, "learning_rate": 4.4722592945662537e-07, "loss": 1.6419, "step": 44681 }, { "epoch": 8.51, "learning_rate": 4.471687321258341e-07, "loss": 1.143, "step": 44682 }, { "epoch": 8.51, "learning_rate": 4.4711153479504296e-07, "loss": 1.9672, "step": 44683 }, { "epoch": 8.51, "learning_rate": 4.470543374642517e-07, "loss": 2.8881, "step": 44684 }, { "epoch": 8.51, "learning_rate": 4.469971401334605e-07, "loss": 1.8127, "step": 44685 }, { "epoch": 8.51, "learning_rate": 4.4693994280266923e-07, "loss": 1.8244, "step": 44686 }, { "epoch": 8.51, "learning_rate": 4.4688274547187797e-07, "loss": 0.5019, "step": 44687 }, { "epoch": 8.51, "learning_rate": 4.4682554814108677e-07, "loss": 1.8915, "step": 44688 }, { "epoch": 8.51, "learning_rate": 4.467683508102955e-07, "loss": 2.0779, "step": 44689 }, { "epoch": 8.51, "learning_rate": 4.4671115347950435e-07, "loss": 2.0679, "step": 44690 }, { "epoch": 8.51, "learning_rate": 4.466539561487131e-07, "loss": 1.5224, "step": 44691 }, { "epoch": 8.51, "learning_rate": 4.465967588179218e-07, "loss": 1.1854, "step": 44692 }, { "epoch": 8.51, "learning_rate": 4.4653956148713063e-07, "loss": 1.3765, "step": 44693 }, { "epoch": 8.51, "learning_rate": 4.4648236415633937e-07, "loss": 0.5991, "step": 44694 }, { "epoch": 8.51, "learning_rate": 4.4642516682554816e-07, "loss": 1.4898, "step": 44695 }, { "epoch": 8.51, "learning_rate": 4.463679694947569e-07, "loss": 0.6193, "step": 44696 }, { "epoch": 8.51, "learning_rate": 4.4631077216396565e-07, "loss": 1.6521, "step": 44697 }, { "epoch": 8.51, "learning_rate": 4.462535748331745e-07, "loss": 1.1039, "step": 44698 }, { "epoch": 8.51, "learning_rate": 4.4619637750238323e-07, "loss": 0.8434, "step": 44699 }, { "epoch": 8.51, "learning_rate": 4.4613918017159203e-07, "loss": 1.1538, "step": 44700 }, { "epoch": 8.51, "learning_rate": 4.4608198284080077e-07, "loss": 2.5006, "step": 44701 }, { "epoch": 8.51, "learning_rate": 4.4602478551000956e-07, "loss": 1.7351, "step": 44702 }, { "epoch": 8.51, "learning_rate": 4.459675881792183e-07, "loss": 0.6572, "step": 44703 }, { "epoch": 8.52, "learning_rate": 4.4591039084842704e-07, "loss": 0.3973, "step": 44704 }, { "epoch": 8.52, "learning_rate": 4.458531935176359e-07, "loss": 1.8348, "step": 44705 }, { "epoch": 8.52, "learning_rate": 4.4579599618684463e-07, "loss": 1.3319, "step": 44706 }, { "epoch": 8.52, "learning_rate": 4.457387988560534e-07, "loss": 1.816, "step": 44707 }, { "epoch": 8.52, "learning_rate": 4.4568160152526217e-07, "loss": 1.1098, "step": 44708 }, { "epoch": 8.52, "learning_rate": 4.456244041944709e-07, "loss": 1.4534, "step": 44709 }, { "epoch": 8.52, "learning_rate": 4.455672068636797e-07, "loss": 1.343, "step": 44710 }, { "epoch": 8.52, "learning_rate": 4.4551000953288844e-07, "loss": 1.2049, "step": 44711 }, { "epoch": 8.52, "learning_rate": 4.454528122020973e-07, "loss": 1.7925, "step": 44712 }, { "epoch": 8.52, "learning_rate": 4.4539561487130603e-07, "loss": 3.4666, "step": 44713 }, { "epoch": 8.52, "learning_rate": 4.453384175405148e-07, "loss": 1.7266, "step": 44714 }, { "epoch": 8.52, "learning_rate": 4.4528122020972356e-07, "loss": 2.0186, "step": 44715 }, { "epoch": 8.52, "learning_rate": 4.452240228789323e-07, "loss": 1.8378, "step": 44716 }, { "epoch": 8.52, "learning_rate": 4.451668255481411e-07, "loss": 1.3136, "step": 44717 }, { "epoch": 8.52, "learning_rate": 4.4510962821734984e-07, "loss": 1.0675, "step": 44718 }, { "epoch": 8.52, "learning_rate": 4.450524308865587e-07, "loss": 1.5946, "step": 44719 }, { "epoch": 8.52, "learning_rate": 4.4499523355576743e-07, "loss": 2.0043, "step": 44720 }, { "epoch": 8.52, "learning_rate": 4.449380362249761e-07, "loss": 1.6056, "step": 44721 }, { "epoch": 8.52, "learning_rate": 4.4488083889418496e-07, "loss": 0.9344, "step": 44722 }, { "epoch": 8.52, "learning_rate": 4.448236415633937e-07, "loss": 2.0319, "step": 44723 }, { "epoch": 8.52, "learning_rate": 4.447664442326025e-07, "loss": 1.8738, "step": 44724 }, { "epoch": 8.52, "learning_rate": 4.4470924690181124e-07, "loss": 0.6527, "step": 44725 }, { "epoch": 8.52, "learning_rate": 4.446520495710201e-07, "loss": 2.5123, "step": 44726 }, { "epoch": 8.52, "learning_rate": 4.445948522402288e-07, "loss": 1.3016, "step": 44727 }, { "epoch": 8.52, "learning_rate": 4.4453765490943757e-07, "loss": 2.5118, "step": 44728 }, { "epoch": 8.52, "learning_rate": 4.4448045757864636e-07, "loss": 3.893, "step": 44729 }, { "epoch": 8.52, "learning_rate": 4.444232602478551e-07, "loss": 1.6925, "step": 44730 }, { "epoch": 8.52, "learning_rate": 4.443660629170639e-07, "loss": 1.8484, "step": 44731 }, { "epoch": 8.52, "learning_rate": 4.4430886558627264e-07, "loss": 2.1907, "step": 44732 }, { "epoch": 8.52, "learning_rate": 4.442516682554814e-07, "loss": 1.8873, "step": 44733 }, { "epoch": 8.52, "learning_rate": 4.441944709246902e-07, "loss": 2.3214, "step": 44734 }, { "epoch": 8.52, "learning_rate": 4.4413727359389896e-07, "loss": 2.6797, "step": 44735 }, { "epoch": 8.52, "learning_rate": 4.4408007626310776e-07, "loss": 1.4195, "step": 44736 }, { "epoch": 8.52, "learning_rate": 4.440228789323165e-07, "loss": 1.3237, "step": 44737 }, { "epoch": 8.52, "learning_rate": 4.439656816015253e-07, "loss": 0.6507, "step": 44738 }, { "epoch": 8.52, "learning_rate": 4.4390848427073403e-07, "loss": 1.2239, "step": 44739 }, { "epoch": 8.52, "learning_rate": 4.438512869399428e-07, "loss": 1.0149, "step": 44740 }, { "epoch": 8.52, "learning_rate": 4.437940896091516e-07, "loss": 2.5491, "step": 44741 }, { "epoch": 8.52, "learning_rate": 4.4373689227836036e-07, "loss": 1.3152, "step": 44742 }, { "epoch": 8.52, "learning_rate": 4.4367969494756916e-07, "loss": 1.464, "step": 44743 }, { "epoch": 8.52, "learning_rate": 4.436224976167779e-07, "loss": 1.6482, "step": 44744 }, { "epoch": 8.52, "learning_rate": 4.4356530028598664e-07, "loss": 0.6614, "step": 44745 }, { "epoch": 8.52, "learning_rate": 4.4350810295519543e-07, "loss": 0.6405, "step": 44746 }, { "epoch": 8.52, "learning_rate": 4.4345090562440417e-07, "loss": 1.917, "step": 44747 }, { "epoch": 8.52, "learning_rate": 4.43393708293613e-07, "loss": 0.7649, "step": 44748 }, { "epoch": 8.52, "learning_rate": 4.4333651096282176e-07, "loss": 2.8918, "step": 44749 }, { "epoch": 8.52, "learning_rate": 4.432793136320305e-07, "loss": 1.5071, "step": 44750 }, { "epoch": 8.52, "learning_rate": 4.432221163012393e-07, "loss": 2.3533, "step": 44751 }, { "epoch": 8.52, "learning_rate": 4.4316491897044804e-07, "loss": 1.6432, "step": 44752 }, { "epoch": 8.52, "learning_rate": 4.4310772163965683e-07, "loss": 2.9701, "step": 44753 }, { "epoch": 8.52, "learning_rate": 4.4305052430886557e-07, "loss": 3.0364, "step": 44754 }, { "epoch": 8.52, "learning_rate": 4.429933269780744e-07, "loss": 1.1915, "step": 44755 }, { "epoch": 8.52, "learning_rate": 4.4293612964728316e-07, "loss": 1.2021, "step": 44756 }, { "epoch": 8.53, "learning_rate": 4.428789323164919e-07, "loss": 2.0632, "step": 44757 }, { "epoch": 8.53, "learning_rate": 4.428217349857007e-07, "loss": 0.9056, "step": 44758 }, { "epoch": 8.53, "learning_rate": 4.4276453765490943e-07, "loss": 1.5443, "step": 44759 }, { "epoch": 8.53, "learning_rate": 4.4270734032411823e-07, "loss": 2.424, "step": 44760 }, { "epoch": 8.53, "learning_rate": 4.4265014299332697e-07, "loss": 1.2583, "step": 44761 }, { "epoch": 8.53, "learning_rate": 4.425929456625357e-07, "loss": 1.4471, "step": 44762 }, { "epoch": 8.53, "learning_rate": 4.4253574833174456e-07, "loss": 1.8087, "step": 44763 }, { "epoch": 8.53, "learning_rate": 4.424785510009533e-07, "loss": 1.5713, "step": 44764 }, { "epoch": 8.53, "learning_rate": 4.424213536701621e-07, "loss": 0.7015, "step": 44765 }, { "epoch": 8.53, "learning_rate": 4.4236415633937083e-07, "loss": 1.1868, "step": 44766 }, { "epoch": 8.53, "learning_rate": 4.423069590085796e-07, "loss": 0.8168, "step": 44767 }, { "epoch": 8.53, "learning_rate": 4.4224976167778837e-07, "loss": 2.5626, "step": 44768 }, { "epoch": 8.53, "learning_rate": 4.421925643469971e-07, "loss": 1.9889, "step": 44769 }, { "epoch": 8.53, "learning_rate": 4.4213536701620595e-07, "loss": 1.1504, "step": 44770 }, { "epoch": 8.53, "learning_rate": 4.420781696854147e-07, "loss": 0.999, "step": 44771 }, { "epoch": 8.53, "learning_rate": 4.420209723546235e-07, "loss": 0.976, "step": 44772 }, { "epoch": 8.53, "learning_rate": 4.4196377502383223e-07, "loss": 1.2353, "step": 44773 }, { "epoch": 8.53, "learning_rate": 4.4190657769304097e-07, "loss": 1.963, "step": 44774 }, { "epoch": 8.53, "learning_rate": 4.4184938036224976e-07, "loss": 2.5393, "step": 44775 }, { "epoch": 8.53, "learning_rate": 4.417921830314585e-07, "loss": 2.0642, "step": 44776 }, { "epoch": 8.53, "learning_rate": 4.4173498570066735e-07, "loss": 1.0314, "step": 44777 }, { "epoch": 8.53, "learning_rate": 4.416777883698761e-07, "loss": 1.2892, "step": 44778 }, { "epoch": 8.53, "learning_rate": 4.416205910390849e-07, "loss": 1.1191, "step": 44779 }, { "epoch": 8.53, "learning_rate": 4.4156339370829363e-07, "loss": 2.6891, "step": 44780 }, { "epoch": 8.53, "learning_rate": 4.4150619637750237e-07, "loss": 0.2938, "step": 44781 }, { "epoch": 8.53, "learning_rate": 4.4144899904671116e-07, "loss": 1.4086, "step": 44782 }, { "epoch": 8.53, "learning_rate": 4.413918017159199e-07, "loss": 2.3287, "step": 44783 }, { "epoch": 8.53, "learning_rate": 4.4133460438512875e-07, "loss": 1.2646, "step": 44784 }, { "epoch": 8.53, "learning_rate": 4.412774070543375e-07, "loss": 0.9439, "step": 44785 }, { "epoch": 8.53, "learning_rate": 4.4122020972354623e-07, "loss": 1.8413, "step": 44786 }, { "epoch": 8.53, "learning_rate": 4.41163012392755e-07, "loss": 0.7194, "step": 44787 }, { "epoch": 8.53, "learning_rate": 4.4110581506196377e-07, "loss": 1.8296, "step": 44788 }, { "epoch": 8.53, "learning_rate": 4.4104861773117256e-07, "loss": 1.3679, "step": 44789 }, { "epoch": 8.53, "learning_rate": 4.409914204003813e-07, "loss": 1.2672, "step": 44790 }, { "epoch": 8.53, "learning_rate": 4.4093422306959015e-07, "loss": 2.3549, "step": 44791 }, { "epoch": 8.53, "learning_rate": 4.408770257387989e-07, "loss": 0.8749, "step": 44792 }, { "epoch": 8.53, "learning_rate": 4.4081982840800763e-07, "loss": 1.4926, "step": 44793 }, { "epoch": 8.53, "learning_rate": 4.407626310772164e-07, "loss": 1.4239, "step": 44794 }, { "epoch": 8.53, "learning_rate": 4.4070543374642516e-07, "loss": 0.4482, "step": 44795 }, { "epoch": 8.53, "learning_rate": 4.4064823641563396e-07, "loss": 0.2772, "step": 44796 }, { "epoch": 8.53, "learning_rate": 4.405910390848427e-07, "loss": 2.4598, "step": 44797 }, { "epoch": 8.53, "learning_rate": 4.4053384175405144e-07, "loss": 1.8382, "step": 44798 }, { "epoch": 8.53, "learning_rate": 4.404766444232603e-07, "loss": 1.968, "step": 44799 }, { "epoch": 8.53, "learning_rate": 4.4041944709246903e-07, "loss": 2.1592, "step": 44800 }, { "epoch": 8.53, "learning_rate": 4.403622497616778e-07, "loss": 0.9604, "step": 44801 }, { "epoch": 8.53, "learning_rate": 4.4030505243088656e-07, "loss": 1.5798, "step": 44802 }, { "epoch": 8.53, "learning_rate": 4.4024785510009536e-07, "loss": 1.3623, "step": 44803 }, { "epoch": 8.53, "learning_rate": 4.401906577693041e-07, "loss": 1.2908, "step": 44804 }, { "epoch": 8.53, "learning_rate": 4.4013346043851284e-07, "loss": 1.1765, "step": 44805 }, { "epoch": 8.53, "learning_rate": 4.400762631077217e-07, "loss": 0.481, "step": 44806 }, { "epoch": 8.53, "learning_rate": 4.400190657769304e-07, "loss": 1.5818, "step": 44807 }, { "epoch": 8.53, "learning_rate": 4.399618684461392e-07, "loss": 0.9834, "step": 44808 }, { "epoch": 8.54, "learning_rate": 4.3990467111534796e-07, "loss": 1.4136, "step": 44809 }, { "epoch": 8.54, "learning_rate": 4.398474737845567e-07, "loss": 1.4357, "step": 44810 }, { "epoch": 8.54, "learning_rate": 4.397902764537655e-07, "loss": 2.1755, "step": 44811 }, { "epoch": 8.54, "learning_rate": 4.3973307912297423e-07, "loss": 1.1858, "step": 44812 }, { "epoch": 8.54, "learning_rate": 4.396758817921831e-07, "loss": 1.6482, "step": 44813 }, { "epoch": 8.54, "learning_rate": 4.396186844613918e-07, "loss": 0.9545, "step": 44814 }, { "epoch": 8.54, "learning_rate": 4.3956148713060056e-07, "loss": 0.6669, "step": 44815 }, { "epoch": 8.54, "learning_rate": 4.3950428979980936e-07, "loss": 2.148, "step": 44816 }, { "epoch": 8.54, "learning_rate": 4.394470924690181e-07, "loss": 1.951, "step": 44817 }, { "epoch": 8.54, "learning_rate": 4.393898951382269e-07, "loss": 1.1927, "step": 44818 }, { "epoch": 8.54, "learning_rate": 4.3933269780743563e-07, "loss": 1.6505, "step": 44819 }, { "epoch": 8.54, "learning_rate": 4.392755004766445e-07, "loss": 1.8076, "step": 44820 }, { "epoch": 8.54, "learning_rate": 4.392183031458532e-07, "loss": 2.3979, "step": 44821 }, { "epoch": 8.54, "learning_rate": 4.3916110581506196e-07, "loss": 1.8988, "step": 44822 }, { "epoch": 8.54, "learning_rate": 4.3910390848427076e-07, "loss": 2.5579, "step": 44823 }, { "epoch": 8.54, "learning_rate": 4.390467111534795e-07, "loss": 2.6705, "step": 44824 }, { "epoch": 8.54, "learning_rate": 4.389895138226883e-07, "loss": 0.6045, "step": 44825 }, { "epoch": 8.54, "learning_rate": 4.3893231649189703e-07, "loss": 0.9264, "step": 44826 }, { "epoch": 8.54, "learning_rate": 4.3887511916110577e-07, "loss": 1.6878, "step": 44827 }, { "epoch": 8.54, "learning_rate": 4.388179218303146e-07, "loss": 2.6018, "step": 44828 }, { "epoch": 8.54, "learning_rate": 4.3876072449952336e-07, "loss": 0.2093, "step": 44829 }, { "epoch": 8.54, "learning_rate": 4.3870352716873215e-07, "loss": 1.4349, "step": 44830 }, { "epoch": 8.54, "learning_rate": 4.386463298379409e-07, "loss": 1.9756, "step": 44831 }, { "epoch": 8.54, "learning_rate": 4.3858913250714974e-07, "loss": 1.8442, "step": 44832 }, { "epoch": 8.54, "learning_rate": 4.3853193517635843e-07, "loss": 0.9541, "step": 44833 }, { "epoch": 8.54, "learning_rate": 4.3847473784556717e-07, "loss": 2.2657, "step": 44834 }, { "epoch": 8.54, "learning_rate": 4.38417540514776e-07, "loss": 0.5287, "step": 44835 }, { "epoch": 8.54, "learning_rate": 4.3836034318398476e-07, "loss": 1.0383, "step": 44836 }, { "epoch": 8.54, "learning_rate": 4.3830314585319355e-07, "loss": 0.2847, "step": 44837 }, { "epoch": 8.54, "learning_rate": 4.382459485224023e-07, "loss": 1.4794, "step": 44838 }, { "epoch": 8.54, "learning_rate": 4.3818875119161103e-07, "loss": 0.2989, "step": 44839 }, { "epoch": 8.54, "learning_rate": 4.381315538608198e-07, "loss": 1.8206, "step": 44840 }, { "epoch": 8.54, "learning_rate": 4.3807435653002857e-07, "loss": 1.8847, "step": 44841 }, { "epoch": 8.54, "learning_rate": 4.380171591992374e-07, "loss": 0.9467, "step": 44842 }, { "epoch": 8.54, "learning_rate": 4.3795996186844615e-07, "loss": 1.4877, "step": 44843 }, { "epoch": 8.54, "learning_rate": 4.3790276453765495e-07, "loss": 1.2274, "step": 44844 }, { "epoch": 8.54, "learning_rate": 4.378455672068637e-07, "loss": 0.8844, "step": 44845 }, { "epoch": 8.54, "learning_rate": 4.3778836987607243e-07, "loss": 2.8201, "step": 44846 }, { "epoch": 8.54, "learning_rate": 4.377311725452812e-07, "loss": 1.2816, "step": 44847 }, { "epoch": 8.54, "learning_rate": 4.3767397521448997e-07, "loss": 1.9695, "step": 44848 }, { "epoch": 8.54, "learning_rate": 4.376167778836988e-07, "loss": 1.3039, "step": 44849 }, { "epoch": 8.54, "learning_rate": 4.3755958055290755e-07, "loss": 2.5632, "step": 44850 }, { "epoch": 8.54, "learning_rate": 4.375023832221163e-07, "loss": 1.7291, "step": 44851 }, { "epoch": 8.54, "learning_rate": 4.374451858913251e-07, "loss": 2.9074, "step": 44852 }, { "epoch": 8.54, "learning_rate": 4.3738798856053383e-07, "loss": 1.9115, "step": 44853 }, { "epoch": 8.54, "learning_rate": 4.373307912297427e-07, "loss": 0.9766, "step": 44854 }, { "epoch": 8.54, "learning_rate": 4.3727359389895136e-07, "loss": 0.8429, "step": 44855 }, { "epoch": 8.54, "learning_rate": 4.372163965681602e-07, "loss": 1.5708, "step": 44856 }, { "epoch": 8.54, "learning_rate": 4.3715919923736895e-07, "loss": 2.2556, "step": 44857 }, { "epoch": 8.54, "learning_rate": 4.371020019065777e-07, "loss": 3.0116, "step": 44858 }, { "epoch": 8.54, "learning_rate": 4.370448045757865e-07, "loss": 1.0459, "step": 44859 }, { "epoch": 8.54, "learning_rate": 4.369876072449952e-07, "loss": 2.7392, "step": 44860 }, { "epoch": 8.54, "learning_rate": 4.3693040991420407e-07, "loss": 1.6114, "step": 44861 }, { "epoch": 8.55, "learning_rate": 4.3687321258341276e-07, "loss": 1.2908, "step": 44862 }, { "epoch": 8.55, "learning_rate": 4.368160152526215e-07, "loss": 3.0125, "step": 44863 }, { "epoch": 8.55, "learning_rate": 4.3675881792183035e-07, "loss": 1.3362, "step": 44864 }, { "epoch": 8.55, "learning_rate": 4.367016205910391e-07, "loss": 2.1331, "step": 44865 }, { "epoch": 8.55, "learning_rate": 4.366444232602479e-07, "loss": 1.477, "step": 44866 }, { "epoch": 8.55, "learning_rate": 4.365872259294566e-07, "loss": 1.6267, "step": 44867 }, { "epoch": 8.55, "learning_rate": 4.3653002859866547e-07, "loss": 0.6998, "step": 44868 }, { "epoch": 8.55, "learning_rate": 4.3647283126787416e-07, "loss": 2.9409, "step": 44869 }, { "epoch": 8.55, "learning_rate": 4.364156339370829e-07, "loss": 1.3165, "step": 44870 }, { "epoch": 8.55, "learning_rate": 4.3635843660629175e-07, "loss": 0.4155, "step": 44871 }, { "epoch": 8.55, "learning_rate": 4.363012392755005e-07, "loss": 1.0831, "step": 44872 }, { "epoch": 8.55, "learning_rate": 4.362440419447093e-07, "loss": 1.4696, "step": 44873 }, { "epoch": 8.55, "learning_rate": 4.36186844613918e-07, "loss": 1.4925, "step": 44874 }, { "epoch": 8.55, "learning_rate": 4.3612964728312676e-07, "loss": 0.4766, "step": 44875 }, { "epoch": 8.55, "learning_rate": 4.360724499523356e-07, "loss": 3.2466, "step": 44876 }, { "epoch": 8.55, "learning_rate": 4.360152526215443e-07, "loss": 1.0589, "step": 44877 }, { "epoch": 8.55, "learning_rate": 4.3595805529075314e-07, "loss": 2.3424, "step": 44878 }, { "epoch": 8.55, "learning_rate": 4.359008579599619e-07, "loss": 1.2425, "step": 44879 }, { "epoch": 8.55, "learning_rate": 4.358436606291706e-07, "loss": 1.3623, "step": 44880 }, { "epoch": 8.55, "learning_rate": 4.357864632983794e-07, "loss": 0.4696, "step": 44881 }, { "epoch": 8.55, "learning_rate": 4.3572926596758816e-07, "loss": 1.5002, "step": 44882 }, { "epoch": 8.55, "learning_rate": 4.35672068636797e-07, "loss": 0.7554, "step": 44883 }, { "epoch": 8.55, "learning_rate": 4.356148713060057e-07, "loss": 0.7217, "step": 44884 }, { "epoch": 8.55, "learning_rate": 4.3555767397521454e-07, "loss": 1.7295, "step": 44885 }, { "epoch": 8.55, "learning_rate": 4.355004766444233e-07, "loss": 1.6576, "step": 44886 }, { "epoch": 8.55, "learning_rate": 4.35443279313632e-07, "loss": 0.8201, "step": 44887 }, { "epoch": 8.55, "learning_rate": 4.353860819828408e-07, "loss": 1.6765, "step": 44888 }, { "epoch": 8.55, "learning_rate": 4.3532888465204956e-07, "loss": 0.7605, "step": 44889 }, { "epoch": 8.55, "learning_rate": 4.352716873212584e-07, "loss": 1.6073, "step": 44890 }, { "epoch": 8.55, "learning_rate": 4.352144899904671e-07, "loss": 0.3874, "step": 44891 }, { "epoch": 8.55, "learning_rate": 4.3515729265967583e-07, "loss": 1.1805, "step": 44892 }, { "epoch": 8.55, "learning_rate": 4.351000953288847e-07, "loss": 2.1588, "step": 44893 }, { "epoch": 8.55, "learning_rate": 4.350428979980934e-07, "loss": 1.6689, "step": 44894 }, { "epoch": 8.55, "learning_rate": 4.349857006673022e-07, "loss": 0.7121, "step": 44895 }, { "epoch": 8.55, "learning_rate": 4.3492850333651096e-07, "loss": 2.9921, "step": 44896 }, { "epoch": 8.55, "learning_rate": 4.348713060057198e-07, "loss": 3.7865, "step": 44897 }, { "epoch": 8.55, "learning_rate": 4.348141086749285e-07, "loss": 2.5727, "step": 44898 }, { "epoch": 8.55, "learning_rate": 4.3475691134413723e-07, "loss": 1.5909, "step": 44899 }, { "epoch": 8.55, "learning_rate": 4.346997140133461e-07, "loss": 0.6341, "step": 44900 }, { "epoch": 8.55, "learning_rate": 4.346425166825548e-07, "loss": 0.896, "step": 44901 }, { "epoch": 8.55, "learning_rate": 4.345853193517636e-07, "loss": 1.1427, "step": 44902 }, { "epoch": 8.55, "learning_rate": 4.3452812202097235e-07, "loss": 1.1381, "step": 44903 }, { "epoch": 8.55, "learning_rate": 4.344709246901811e-07, "loss": 0.8421, "step": 44904 }, { "epoch": 8.55, "learning_rate": 4.3441372735938994e-07, "loss": 1.8883, "step": 44905 }, { "epoch": 8.55, "learning_rate": 4.3435653002859863e-07, "loss": 1.9076, "step": 44906 }, { "epoch": 8.55, "learning_rate": 4.342993326978075e-07, "loss": 1.2403, "step": 44907 }, { "epoch": 8.55, "learning_rate": 4.342421353670162e-07, "loss": 1.9409, "step": 44908 }, { "epoch": 8.55, "learning_rate": 4.34184938036225e-07, "loss": 1.3257, "step": 44909 }, { "epoch": 8.55, "learning_rate": 4.3412774070543375e-07, "loss": 1.5658, "step": 44910 }, { "epoch": 8.55, "learning_rate": 4.340705433746425e-07, "loss": 0.9535, "step": 44911 }, { "epoch": 8.55, "learning_rate": 4.3401334604385134e-07, "loss": 1.34, "step": 44912 }, { "epoch": 8.55, "learning_rate": 4.3395614871306003e-07, "loss": 0.3889, "step": 44913 }, { "epoch": 8.56, "learning_rate": 4.338989513822689e-07, "loss": 2.4672, "step": 44914 }, { "epoch": 8.56, "learning_rate": 4.338417540514776e-07, "loss": 1.4665, "step": 44915 }, { "epoch": 8.56, "learning_rate": 4.3378455672068636e-07, "loss": 1.4086, "step": 44916 }, { "epoch": 8.56, "learning_rate": 4.3372735938989515e-07, "loss": 4.3568, "step": 44917 }, { "epoch": 8.56, "learning_rate": 4.336701620591039e-07, "loss": 1.8375, "step": 44918 }, { "epoch": 8.56, "learning_rate": 4.3361296472831274e-07, "loss": 0.9309, "step": 44919 }, { "epoch": 8.56, "learning_rate": 4.335557673975214e-07, "loss": 1.7731, "step": 44920 }, { "epoch": 8.56, "learning_rate": 4.3349857006673027e-07, "loss": 1.0711, "step": 44921 }, { "epoch": 8.56, "learning_rate": 4.33441372735939e-07, "loss": 2.5083, "step": 44922 }, { "epoch": 8.56, "learning_rate": 4.3338417540514775e-07, "loss": 0.8885, "step": 44923 }, { "epoch": 8.56, "learning_rate": 4.3332697807435655e-07, "loss": 2.6312, "step": 44924 }, { "epoch": 8.56, "learning_rate": 4.332697807435653e-07, "loss": 1.4221, "step": 44925 }, { "epoch": 8.56, "learning_rate": 4.3321258341277414e-07, "loss": 0.9921, "step": 44926 }, { "epoch": 8.56, "learning_rate": 4.331553860819829e-07, "loss": 2.096, "step": 44927 }, { "epoch": 8.56, "learning_rate": 4.3309818875119156e-07, "loss": 1.3966, "step": 44928 }, { "epoch": 8.56, "learning_rate": 4.330409914204004e-07, "loss": 1.3595, "step": 44929 }, { "epoch": 8.56, "learning_rate": 4.3298379408960915e-07, "loss": 0.9184, "step": 44930 }, { "epoch": 8.56, "learning_rate": 4.3292659675881795e-07, "loss": 1.5008, "step": 44931 }, { "epoch": 8.56, "learning_rate": 4.328693994280267e-07, "loss": 1.156, "step": 44932 }, { "epoch": 8.56, "learning_rate": 4.3281220209723543e-07, "loss": 2.8234, "step": 44933 }, { "epoch": 8.56, "learning_rate": 4.327550047664443e-07, "loss": 1.4512, "step": 44934 }, { "epoch": 8.56, "learning_rate": 4.3269780743565296e-07, "loss": 1.3465, "step": 44935 }, { "epoch": 8.56, "learning_rate": 4.326406101048618e-07, "loss": 2.552, "step": 44936 }, { "epoch": 8.56, "learning_rate": 4.3258341277407055e-07, "loss": 1.5525, "step": 44937 }, { "epoch": 8.56, "learning_rate": 4.3252621544327934e-07, "loss": 1.525, "step": 44938 }, { "epoch": 8.56, "learning_rate": 4.324690181124881e-07, "loss": 0.5486, "step": 44939 }, { "epoch": 8.56, "learning_rate": 4.324118207816968e-07, "loss": 2.8169, "step": 44940 }, { "epoch": 8.56, "learning_rate": 4.3235462345090567e-07, "loss": 1.0494, "step": 44941 }, { "epoch": 8.56, "learning_rate": 4.3229742612011436e-07, "loss": 2.2802, "step": 44942 }, { "epoch": 8.56, "learning_rate": 4.322402287893232e-07, "loss": 0.8313, "step": 44943 }, { "epoch": 8.56, "learning_rate": 4.3218303145853195e-07, "loss": 2.427, "step": 44944 }, { "epoch": 8.56, "learning_rate": 4.321258341277407e-07, "loss": 2.0088, "step": 44945 }, { "epoch": 8.56, "learning_rate": 4.320686367969495e-07, "loss": 1.4572, "step": 44946 }, { "epoch": 8.56, "learning_rate": 4.320114394661582e-07, "loss": 2.1117, "step": 44947 }, { "epoch": 8.56, "learning_rate": 4.3195424213536707e-07, "loss": 1.4015, "step": 44948 }, { "epoch": 8.56, "learning_rate": 4.318970448045758e-07, "loss": 1.8562, "step": 44949 }, { "epoch": 8.56, "learning_rate": 4.318398474737846e-07, "loss": 0.998, "step": 44950 }, { "epoch": 8.56, "learning_rate": 4.3178265014299335e-07, "loss": 2.0079, "step": 44951 }, { "epoch": 8.56, "learning_rate": 4.317254528122021e-07, "loss": 0.3032, "step": 44952 }, { "epoch": 8.56, "learning_rate": 4.316682554814109e-07, "loss": 2.7901, "step": 44953 }, { "epoch": 8.56, "learning_rate": 4.316110581506196e-07, "loss": 0.8264, "step": 44954 }, { "epoch": 8.56, "learning_rate": 4.3155386081982847e-07, "loss": 2.9654, "step": 44955 }, { "epoch": 8.56, "learning_rate": 4.314966634890372e-07, "loss": 0.5177, "step": 44956 }, { "epoch": 8.56, "learning_rate": 4.314394661582459e-07, "loss": 1.305, "step": 44957 }, { "epoch": 8.56, "learning_rate": 4.3138226882745474e-07, "loss": 2.4928, "step": 44958 }, { "epoch": 8.56, "learning_rate": 4.313250714966635e-07, "loss": 2.5271, "step": 44959 }, { "epoch": 8.56, "learning_rate": 4.312678741658723e-07, "loss": 2.8466, "step": 44960 }, { "epoch": 8.56, "learning_rate": 4.31210676835081e-07, "loss": 2.3892, "step": 44961 }, { "epoch": 8.56, "learning_rate": 4.3115347950428987e-07, "loss": 0.9505, "step": 44962 }, { "epoch": 8.56, "learning_rate": 4.310962821734986e-07, "loss": 1.6735, "step": 44963 }, { "epoch": 8.56, "learning_rate": 4.310390848427073e-07, "loss": 0.6583, "step": 44964 }, { "epoch": 8.56, "learning_rate": 4.3098188751191614e-07, "loss": 0.3351, "step": 44965 }, { "epoch": 8.56, "learning_rate": 4.309246901811249e-07, "loss": 2.1211, "step": 44966 }, { "epoch": 8.57, "learning_rate": 4.308674928503337e-07, "loss": 1.15, "step": 44967 }, { "epoch": 8.57, "learning_rate": 4.308102955195424e-07, "loss": 1.214, "step": 44968 }, { "epoch": 8.57, "learning_rate": 4.3075309818875116e-07, "loss": 2.3616, "step": 44969 }, { "epoch": 8.57, "learning_rate": 4.3069590085796e-07, "loss": 1.9014, "step": 44970 }, { "epoch": 8.57, "learning_rate": 4.3063870352716875e-07, "loss": 0.9019, "step": 44971 }, { "epoch": 8.57, "learning_rate": 4.3058150619637754e-07, "loss": 1.0522, "step": 44972 }, { "epoch": 8.57, "learning_rate": 4.305243088655863e-07, "loss": 1.4915, "step": 44973 }, { "epoch": 8.57, "learning_rate": 4.304671115347951e-07, "loss": 2.1436, "step": 44974 }, { "epoch": 8.57, "learning_rate": 4.304099142040038e-07, "loss": 1.7722, "step": 44975 }, { "epoch": 8.57, "learning_rate": 4.3035271687321256e-07, "loss": 0.9211, "step": 44976 }, { "epoch": 8.57, "learning_rate": 4.302955195424214e-07, "loss": 1.8155, "step": 44977 }, { "epoch": 8.57, "learning_rate": 4.3023832221163014e-07, "loss": 1.8051, "step": 44978 }, { "epoch": 8.57, "learning_rate": 4.3018112488083894e-07, "loss": 1.4639, "step": 44979 }, { "epoch": 8.57, "learning_rate": 4.301239275500477e-07, "loss": 1.8835, "step": 44980 }, { "epoch": 8.57, "learning_rate": 4.300667302192564e-07, "loss": 1.6595, "step": 44981 }, { "epoch": 8.57, "learning_rate": 4.300095328884652e-07, "loss": 2.227, "step": 44982 }, { "epoch": 8.57, "learning_rate": 4.2995233555767395e-07, "loss": 1.4439, "step": 44983 }, { "epoch": 8.57, "learning_rate": 4.298951382268828e-07, "loss": 1.7067, "step": 44984 }, { "epoch": 8.57, "learning_rate": 4.2983794089609154e-07, "loss": 2.3423, "step": 44985 }, { "epoch": 8.57, "learning_rate": 4.2978074356530034e-07, "loss": 1.2588, "step": 44986 }, { "epoch": 8.57, "learning_rate": 4.297235462345091e-07, "loss": 0.4615, "step": 44987 }, { "epoch": 8.57, "learning_rate": 4.296663489037178e-07, "loss": 2.3697, "step": 44988 }, { "epoch": 8.57, "learning_rate": 4.296091515729266e-07, "loss": 3.2387, "step": 44989 }, { "epoch": 8.57, "learning_rate": 4.2955195424213535e-07, "loss": 2.5348, "step": 44990 }, { "epoch": 8.57, "learning_rate": 4.294947569113442e-07, "loss": 1.1658, "step": 44991 }, { "epoch": 8.57, "learning_rate": 4.2943755958055294e-07, "loss": 0.7276, "step": 44992 }, { "epoch": 8.57, "learning_rate": 4.293803622497617e-07, "loss": 2.1159, "step": 44993 }, { "epoch": 8.57, "learning_rate": 4.293231649189705e-07, "loss": 0.8473, "step": 44994 }, { "epoch": 8.57, "learning_rate": 4.292659675881792e-07, "loss": 3.0002, "step": 44995 }, { "epoch": 8.57, "learning_rate": 4.29208770257388e-07, "loss": 0.6515, "step": 44996 }, { "epoch": 8.57, "learning_rate": 4.2915157292659675e-07, "loss": 0.8657, "step": 44997 }, { "epoch": 8.57, "learning_rate": 4.290943755958055e-07, "loss": 0.8589, "step": 44998 }, { "epoch": 8.57, "learning_rate": 4.2903717826501434e-07, "loss": 1.5985, "step": 44999 }, { "epoch": 8.57, "learning_rate": 4.289799809342231e-07, "loss": 0.8779, "step": 45000 }, { "epoch": 8.57, "learning_rate": 4.2892278360343187e-07, "loss": 1.2623, "step": 45001 }, { "epoch": 8.57, "learning_rate": 4.288655862726406e-07, "loss": 1.4137, "step": 45002 }, { "epoch": 8.57, "learning_rate": 4.288083889418494e-07, "loss": 3.8401, "step": 45003 }, { "epoch": 8.57, "learning_rate": 4.2875119161105815e-07, "loss": 3.3452, "step": 45004 }, { "epoch": 8.57, "learning_rate": 4.286939942802669e-07, "loss": 1.4394, "step": 45005 }, { "epoch": 8.57, "learning_rate": 4.2863679694947574e-07, "loss": 1.6142, "step": 45006 }, { "epoch": 8.57, "learning_rate": 4.285795996186845e-07, "loss": 0.2302, "step": 45007 }, { "epoch": 8.57, "learning_rate": 4.2852240228789327e-07, "loss": 1.8867, "step": 45008 }, { "epoch": 8.57, "learning_rate": 4.28465204957102e-07, "loss": 0.7116, "step": 45009 }, { "epoch": 8.57, "learning_rate": 4.2840800762631075e-07, "loss": 2.4667, "step": 45010 }, { "epoch": 8.57, "learning_rate": 4.2835081029551955e-07, "loss": 1.7958, "step": 45011 }, { "epoch": 8.57, "learning_rate": 4.282936129647283e-07, "loss": 1.6634, "step": 45012 }, { "epoch": 8.57, "learning_rate": 4.2823641563393713e-07, "loss": 1.3536, "step": 45013 }, { "epoch": 8.57, "learning_rate": 4.281792183031459e-07, "loss": 0.4215, "step": 45014 }, { "epoch": 8.57, "learning_rate": 4.2812202097235467e-07, "loss": 1.0973, "step": 45015 }, { "epoch": 8.57, "learning_rate": 4.280648236415634e-07, "loss": 3.3479, "step": 45016 }, { "epoch": 8.57, "learning_rate": 4.2800762631077215e-07, "loss": 1.6281, "step": 45017 }, { "epoch": 8.57, "learning_rate": 4.2795042897998094e-07, "loss": 1.3106, "step": 45018 }, { "epoch": 8.58, "learning_rate": 4.278932316491897e-07, "loss": 1.9319, "step": 45019 }, { "epoch": 8.58, "learning_rate": 4.2783603431839853e-07, "loss": 3.4217, "step": 45020 }, { "epoch": 8.58, "learning_rate": 4.2777883698760727e-07, "loss": 0.5134, "step": 45021 }, { "epoch": 8.58, "learning_rate": 4.27721639656816e-07, "loss": 1.4438, "step": 45022 }, { "epoch": 8.58, "learning_rate": 4.276644423260248e-07, "loss": 1.3209, "step": 45023 }, { "epoch": 8.58, "learning_rate": 4.2760724499523355e-07, "loss": 1.2559, "step": 45024 }, { "epoch": 8.58, "learning_rate": 4.2755004766444234e-07, "loss": 0.7681, "step": 45025 }, { "epoch": 8.58, "learning_rate": 4.274928503336511e-07, "loss": 2.2505, "step": 45026 }, { "epoch": 8.58, "learning_rate": 4.2743565300285993e-07, "loss": 1.9368, "step": 45027 }, { "epoch": 8.58, "learning_rate": 4.2737845567206867e-07, "loss": 1.8429, "step": 45028 }, { "epoch": 8.58, "learning_rate": 4.273212583412774e-07, "loss": 0.685, "step": 45029 }, { "epoch": 8.58, "learning_rate": 4.272640610104862e-07, "loss": 1.0398, "step": 45030 }, { "epoch": 8.58, "learning_rate": 4.2720686367969495e-07, "loss": 0.855, "step": 45031 }, { "epoch": 8.58, "learning_rate": 4.2714966634890374e-07, "loss": 1.851, "step": 45032 }, { "epoch": 8.58, "learning_rate": 4.270924690181125e-07, "loss": 1.2919, "step": 45033 }, { "epoch": 8.58, "learning_rate": 4.270352716873212e-07, "loss": 0.3787, "step": 45034 }, { "epoch": 8.58, "learning_rate": 4.2697807435653007e-07, "loss": 2.3932, "step": 45035 }, { "epoch": 8.58, "learning_rate": 4.269208770257388e-07, "loss": 0.4855, "step": 45036 }, { "epoch": 8.58, "learning_rate": 4.268636796949476e-07, "loss": 2.1911, "step": 45037 }, { "epoch": 8.58, "learning_rate": 4.2680648236415634e-07, "loss": 1.5309, "step": 45038 }, { "epoch": 8.58, "learning_rate": 4.2674928503336514e-07, "loss": 1.7331, "step": 45039 }, { "epoch": 8.58, "learning_rate": 4.266920877025739e-07, "loss": 0.6787, "step": 45040 }, { "epoch": 8.58, "learning_rate": 4.266348903717826e-07, "loss": 3.0483, "step": 45041 }, { "epoch": 8.58, "learning_rate": 4.2657769304099147e-07, "loss": 1.5401, "step": 45042 }, { "epoch": 8.58, "learning_rate": 4.265204957102002e-07, "loss": 1.5004, "step": 45043 }, { "epoch": 8.58, "learning_rate": 4.26463298379409e-07, "loss": 1.1986, "step": 45044 }, { "epoch": 8.58, "learning_rate": 4.2640610104861774e-07, "loss": 1.9979, "step": 45045 }, { "epoch": 8.58, "learning_rate": 4.263489037178265e-07, "loss": 1.0338, "step": 45046 }, { "epoch": 8.58, "learning_rate": 4.262917063870353e-07, "loss": 1.2019, "step": 45047 }, { "epoch": 8.58, "learning_rate": 4.26234509056244e-07, "loss": 2.7848, "step": 45048 }, { "epoch": 8.58, "learning_rate": 4.2617731172545286e-07, "loss": 1.3811, "step": 45049 }, { "epoch": 8.58, "learning_rate": 4.261201143946616e-07, "loss": 1.3956, "step": 45050 }, { "epoch": 8.58, "learning_rate": 4.2606291706387035e-07, "loss": 2.069, "step": 45051 }, { "epoch": 8.58, "learning_rate": 4.2600571973307914e-07, "loss": 0.3531, "step": 45052 }, { "epoch": 8.58, "learning_rate": 4.259485224022879e-07, "loss": 1.6037, "step": 45053 }, { "epoch": 8.58, "learning_rate": 4.258913250714967e-07, "loss": 0.9441, "step": 45054 }, { "epoch": 8.58, "learning_rate": 4.258341277407054e-07, "loss": 0.4661, "step": 45055 }, { "epoch": 8.58, "learning_rate": 4.2577693040991426e-07, "loss": 1.8624, "step": 45056 }, { "epoch": 8.58, "learning_rate": 4.25719733079123e-07, "loss": 1.3545, "step": 45057 }, { "epoch": 8.58, "learning_rate": 4.2566253574833174e-07, "loss": 2.0017, "step": 45058 }, { "epoch": 8.58, "learning_rate": 4.2560533841754054e-07, "loss": 2.3371, "step": 45059 }, { "epoch": 8.58, "learning_rate": 4.255481410867493e-07, "loss": 1.6632, "step": 45060 }, { "epoch": 8.58, "learning_rate": 4.2549094375595807e-07, "loss": 2.9521, "step": 45061 }, { "epoch": 8.58, "learning_rate": 4.254337464251668e-07, "loss": 2.5018, "step": 45062 }, { "epoch": 8.58, "learning_rate": 4.2537654909437555e-07, "loss": 1.9886, "step": 45063 }, { "epoch": 8.58, "learning_rate": 4.253193517635844e-07, "loss": 1.1508, "step": 45064 }, { "epoch": 8.58, "learning_rate": 4.2526215443279314e-07, "loss": 0.9526, "step": 45065 }, { "epoch": 8.58, "learning_rate": 4.2520495710200194e-07, "loss": 1.3358, "step": 45066 }, { "epoch": 8.58, "learning_rate": 4.251477597712107e-07, "loss": 1.407, "step": 45067 }, { "epoch": 8.58, "learning_rate": 4.2509056244041947e-07, "loss": 0.5838, "step": 45068 }, { "epoch": 8.58, "learning_rate": 4.250333651096282e-07, "loss": 2.939, "step": 45069 }, { "epoch": 8.58, "learning_rate": 4.2497616777883695e-07, "loss": 1.5009, "step": 45070 }, { "epoch": 8.58, "learning_rate": 4.249189704480458e-07, "loss": 0.6666, "step": 45071 }, { "epoch": 8.59, "learning_rate": 4.2486177311725454e-07, "loss": 0.8156, "step": 45072 }, { "epoch": 8.59, "learning_rate": 4.2480457578646333e-07, "loss": 3.3483, "step": 45073 }, { "epoch": 8.59, "learning_rate": 4.247473784556721e-07, "loss": 1.8293, "step": 45074 }, { "epoch": 8.59, "learning_rate": 4.246901811248808e-07, "loss": 1.5872, "step": 45075 }, { "epoch": 8.59, "learning_rate": 4.246329837940896e-07, "loss": 1.1402, "step": 45076 }, { "epoch": 8.59, "learning_rate": 4.2457578646329835e-07, "loss": 1.693, "step": 45077 }, { "epoch": 8.59, "learning_rate": 4.245185891325072e-07, "loss": 0.8629, "step": 45078 }, { "epoch": 8.59, "learning_rate": 4.2446139180171594e-07, "loss": 1.3361, "step": 45079 }, { "epoch": 8.59, "learning_rate": 4.2440419447092473e-07, "loss": 1.0294, "step": 45080 }, { "epoch": 8.59, "learning_rate": 4.2434699714013347e-07, "loss": 0.8352, "step": 45081 }, { "epoch": 8.59, "learning_rate": 4.242897998093422e-07, "loss": 1.2768, "step": 45082 }, { "epoch": 8.59, "learning_rate": 4.24232602478551e-07, "loss": 1.3317, "step": 45083 }, { "epoch": 8.59, "learning_rate": 4.2417540514775975e-07, "loss": 1.6279, "step": 45084 }, { "epoch": 8.59, "learning_rate": 4.241182078169686e-07, "loss": 2.0315, "step": 45085 }, { "epoch": 8.59, "learning_rate": 4.2406101048617734e-07, "loss": 0.9759, "step": 45086 }, { "epoch": 8.59, "learning_rate": 4.240038131553861e-07, "loss": 1.3723, "step": 45087 }, { "epoch": 8.59, "learning_rate": 4.2394661582459487e-07, "loss": 1.4788, "step": 45088 }, { "epoch": 8.59, "learning_rate": 4.238894184938036e-07, "loss": 1.3126, "step": 45089 }, { "epoch": 8.59, "learning_rate": 4.238322211630124e-07, "loss": 2.587, "step": 45090 }, { "epoch": 8.59, "learning_rate": 4.2377502383222115e-07, "loss": 1.4216, "step": 45091 }, { "epoch": 8.59, "learning_rate": 4.2371782650143e-07, "loss": 2.3933, "step": 45092 }, { "epoch": 8.59, "learning_rate": 4.2366062917063873e-07, "loss": 0.962, "step": 45093 }, { "epoch": 8.59, "learning_rate": 4.236034318398475e-07, "loss": 2.3953, "step": 45094 }, { "epoch": 8.59, "learning_rate": 4.2354623450905627e-07, "loss": 1.0163, "step": 45095 }, { "epoch": 8.59, "learning_rate": 4.23489037178265e-07, "loss": 2.0222, "step": 45096 }, { "epoch": 8.59, "learning_rate": 4.234318398474738e-07, "loss": 1.9226, "step": 45097 }, { "epoch": 8.59, "learning_rate": 4.2337464251668254e-07, "loss": 2.3881, "step": 45098 }, { "epoch": 8.59, "learning_rate": 4.233174451858913e-07, "loss": 0.9382, "step": 45099 }, { "epoch": 8.59, "learning_rate": 4.2326024785510013e-07, "loss": 1.8897, "step": 45100 }, { "epoch": 8.59, "learning_rate": 4.2320305052430887e-07, "loss": 0.8012, "step": 45101 }, { "epoch": 8.59, "learning_rate": 4.2314585319351767e-07, "loss": 2.834, "step": 45102 }, { "epoch": 8.59, "learning_rate": 4.230886558627264e-07, "loss": 0.7109, "step": 45103 }, { "epoch": 8.59, "learning_rate": 4.2303145853193525e-07, "loss": 0.6719, "step": 45104 }, { "epoch": 8.59, "learning_rate": 4.2297426120114394e-07, "loss": 0.9905, "step": 45105 }, { "epoch": 8.59, "learning_rate": 4.229170638703527e-07, "loss": 2.9646, "step": 45106 }, { "epoch": 8.59, "learning_rate": 4.2285986653956153e-07, "loss": 0.6406, "step": 45107 }, { "epoch": 8.59, "learning_rate": 4.2280266920877027e-07, "loss": 0.8592, "step": 45108 }, { "epoch": 8.59, "learning_rate": 4.2274547187797906e-07, "loss": 2.9354, "step": 45109 }, { "epoch": 8.59, "learning_rate": 4.226882745471878e-07, "loss": 1.4968, "step": 45110 }, { "epoch": 8.59, "learning_rate": 4.2263107721639655e-07, "loss": 0.4637, "step": 45111 }, { "epoch": 8.59, "learning_rate": 4.2257387988560534e-07, "loss": 1.0282, "step": 45112 }, { "epoch": 8.59, "learning_rate": 4.225166825548141e-07, "loss": 2.1289, "step": 45113 }, { "epoch": 8.59, "learning_rate": 4.2245948522402293e-07, "loss": 1.3526, "step": 45114 }, { "epoch": 8.59, "learning_rate": 4.2240228789323167e-07, "loss": 0.6533, "step": 45115 }, { "epoch": 8.59, "learning_rate": 4.223450905624404e-07, "loss": 0.7866, "step": 45116 }, { "epoch": 8.59, "learning_rate": 4.222878932316492e-07, "loss": 1.7231, "step": 45117 }, { "epoch": 8.59, "learning_rate": 4.2223069590085794e-07, "loss": 1.2768, "step": 45118 }, { "epoch": 8.59, "learning_rate": 4.2217349857006674e-07, "loss": 0.3317, "step": 45119 }, { "epoch": 8.59, "learning_rate": 4.221163012392755e-07, "loss": 2.9899, "step": 45120 }, { "epoch": 8.59, "learning_rate": 4.220591039084843e-07, "loss": 1.6207, "step": 45121 }, { "epoch": 8.59, "learning_rate": 4.2200190657769307e-07, "loss": 2.258, "step": 45122 }, { "epoch": 8.59, "learning_rate": 4.219447092469018e-07, "loss": 1.6586, "step": 45123 }, { "epoch": 8.6, "learning_rate": 4.218875119161106e-07, "loss": 1.2697, "step": 45124 }, { "epoch": 8.6, "learning_rate": 4.2183031458531934e-07, "loss": 1.17, "step": 45125 }, { "epoch": 8.6, "learning_rate": 4.217731172545282e-07, "loss": 3.3925, "step": 45126 }, { "epoch": 8.6, "learning_rate": 4.217159199237369e-07, "loss": 2.4178, "step": 45127 }, { "epoch": 8.6, "learning_rate": 4.216587225929456e-07, "loss": 0.653, "step": 45128 }, { "epoch": 8.6, "learning_rate": 4.2160152526215446e-07, "loss": 2.7295, "step": 45129 }, { "epoch": 8.6, "learning_rate": 4.215443279313632e-07, "loss": 1.5275, "step": 45130 }, { "epoch": 8.6, "learning_rate": 4.21487130600572e-07, "loss": 1.9398, "step": 45131 }, { "epoch": 8.6, "learning_rate": 4.2142993326978074e-07, "loss": 0.4085, "step": 45132 }, { "epoch": 8.6, "learning_rate": 4.213727359389896e-07, "loss": 2.0671, "step": 45133 }, { "epoch": 8.6, "learning_rate": 4.213155386081983e-07, "loss": 1.272, "step": 45134 }, { "epoch": 8.6, "learning_rate": 4.21258341277407e-07, "loss": 2.7, "step": 45135 }, { "epoch": 8.6, "learning_rate": 4.2120114394661586e-07, "loss": 1.9403, "step": 45136 }, { "epoch": 8.6, "learning_rate": 4.211439466158246e-07, "loss": 2.8146, "step": 45137 }, { "epoch": 8.6, "learning_rate": 4.210867492850334e-07, "loss": 1.9687, "step": 45138 }, { "epoch": 8.6, "learning_rate": 4.2102955195424214e-07, "loss": 0.7697, "step": 45139 }, { "epoch": 8.6, "learning_rate": 4.209723546234509e-07, "loss": 1.9766, "step": 45140 }, { "epoch": 8.6, "learning_rate": 4.2091515729265967e-07, "loss": 1.5472, "step": 45141 }, { "epoch": 8.6, "learning_rate": 4.208579599618684e-07, "loss": 0.3733, "step": 45142 }, { "epoch": 8.6, "learning_rate": 4.2080076263107726e-07, "loss": 1.4441, "step": 45143 }, { "epoch": 8.6, "learning_rate": 4.20743565300286e-07, "loss": 1.1326, "step": 45144 }, { "epoch": 8.6, "learning_rate": 4.206863679694948e-07, "loss": 2.5562, "step": 45145 }, { "epoch": 8.6, "learning_rate": 4.2062917063870353e-07, "loss": 1.5821, "step": 45146 }, { "epoch": 8.6, "learning_rate": 4.205719733079123e-07, "loss": 1.0388, "step": 45147 }, { "epoch": 8.6, "learning_rate": 4.205147759771211e-07, "loss": 2.333, "step": 45148 }, { "epoch": 8.6, "learning_rate": 4.204575786463298e-07, "loss": 1.6333, "step": 45149 }, { "epoch": 8.6, "learning_rate": 4.2040038131553866e-07, "loss": 0.7245, "step": 45150 }, { "epoch": 8.6, "learning_rate": 4.203431839847474e-07, "loss": 1.6089, "step": 45151 }, { "epoch": 8.6, "learning_rate": 4.2028598665395614e-07, "loss": 1.3348, "step": 45152 }, { "epoch": 8.6, "learning_rate": 4.2022878932316493e-07, "loss": 1.7181, "step": 45153 }, { "epoch": 8.6, "learning_rate": 4.201715919923737e-07, "loss": 0.5616, "step": 45154 }, { "epoch": 8.6, "learning_rate": 4.201143946615825e-07, "loss": 2.4509, "step": 45155 }, { "epoch": 8.6, "learning_rate": 4.200571973307912e-07, "loss": 1.0324, "step": 45156 }, { "epoch": 8.6, "learning_rate": 4.2000000000000006e-07, "loss": 1.3037, "step": 45157 }, { "epoch": 8.6, "learning_rate": 4.199428026692088e-07, "loss": 1.478, "step": 45158 }, { "epoch": 8.6, "learning_rate": 4.1988560533841754e-07, "loss": 1.1476, "step": 45159 }, { "epoch": 8.6, "learning_rate": 4.1982840800762633e-07, "loss": 2.8087, "step": 45160 }, { "epoch": 8.6, "learning_rate": 4.1977121067683507e-07, "loss": 1.2852, "step": 45161 }, { "epoch": 8.6, "learning_rate": 4.197140133460439e-07, "loss": 1.7822, "step": 45162 }, { "epoch": 8.6, "learning_rate": 4.196568160152526e-07, "loss": 0.4847, "step": 45163 }, { "epoch": 8.6, "learning_rate": 4.1959961868446135e-07, "loss": 1.0901, "step": 45164 }, { "epoch": 8.6, "learning_rate": 4.195424213536702e-07, "loss": 1.0117, "step": 45165 }, { "epoch": 8.6, "learning_rate": 4.1948522402287893e-07, "loss": 1.3245, "step": 45166 }, { "epoch": 8.6, "learning_rate": 4.1942802669208773e-07, "loss": 1.4048, "step": 45167 }, { "epoch": 8.6, "learning_rate": 4.1937082936129647e-07, "loss": 1.3928, "step": 45168 }, { "epoch": 8.6, "learning_rate": 4.193136320305053e-07, "loss": 0.5338, "step": 45169 }, { "epoch": 8.6, "learning_rate": 4.1925643469971406e-07, "loss": 2.295, "step": 45170 }, { "epoch": 8.6, "learning_rate": 4.1919923736892275e-07, "loss": 2.1953, "step": 45171 }, { "epoch": 8.6, "learning_rate": 4.191420400381316e-07, "loss": 1.6683, "step": 45172 }, { "epoch": 8.6, "learning_rate": 4.1908484270734033e-07, "loss": 1.3909, "step": 45173 }, { "epoch": 8.6, "learning_rate": 4.1902764537654913e-07, "loss": 0.8993, "step": 45174 }, { "epoch": 8.6, "learning_rate": 4.1897044804575787e-07, "loss": 0.7662, "step": 45175 }, { "epoch": 8.6, "learning_rate": 4.189132507149666e-07, "loss": 2.5116, "step": 45176 }, { "epoch": 8.61, "learning_rate": 4.1885605338417546e-07, "loss": 1.4016, "step": 45177 }, { "epoch": 8.61, "learning_rate": 4.1879885605338414e-07, "loss": 1.4018, "step": 45178 }, { "epoch": 8.61, "learning_rate": 4.18741658722593e-07, "loss": 0.7027, "step": 45179 }, { "epoch": 8.61, "learning_rate": 4.1868446139180173e-07, "loss": 1.3912, "step": 45180 }, { "epoch": 8.61, "learning_rate": 4.1862726406101047e-07, "loss": 0.9504, "step": 45181 }, { "epoch": 8.61, "learning_rate": 4.1857006673021927e-07, "loss": 2.0403, "step": 45182 }, { "epoch": 8.61, "learning_rate": 4.18512869399428e-07, "loss": 1.3369, "step": 45183 }, { "epoch": 8.61, "learning_rate": 4.1845567206863685e-07, "loss": 1.7637, "step": 45184 }, { "epoch": 8.61, "learning_rate": 4.1839847473784554e-07, "loss": 1.0255, "step": 45185 }, { "epoch": 8.61, "learning_rate": 4.183412774070544e-07, "loss": 0.3799, "step": 45186 }, { "epoch": 8.61, "learning_rate": 4.1828408007626313e-07, "loss": 0.9128, "step": 45187 }, { "epoch": 8.61, "learning_rate": 4.1822688274547187e-07, "loss": 1.4141, "step": 45188 }, { "epoch": 8.61, "learning_rate": 4.1816968541468066e-07, "loss": 2.7661, "step": 45189 }, { "epoch": 8.61, "learning_rate": 4.181124880838894e-07, "loss": 1.3258, "step": 45190 }, { "epoch": 8.61, "learning_rate": 4.1805529075309825e-07, "loss": 2.1908, "step": 45191 }, { "epoch": 8.61, "learning_rate": 4.1799809342230694e-07, "loss": 1.7542, "step": 45192 }, { "epoch": 8.61, "learning_rate": 4.179408960915157e-07, "loss": 1.5112, "step": 45193 }, { "epoch": 8.61, "learning_rate": 4.178836987607245e-07, "loss": 1.098, "step": 45194 }, { "epoch": 8.61, "learning_rate": 4.1782650142993327e-07, "loss": 2.2086, "step": 45195 }, { "epoch": 8.61, "learning_rate": 4.1776930409914206e-07, "loss": 2.6748, "step": 45196 }, { "epoch": 8.61, "learning_rate": 4.177121067683508e-07, "loss": 1.5784, "step": 45197 }, { "epoch": 8.61, "learning_rate": 4.1765490943755965e-07, "loss": 2.0249, "step": 45198 }, { "epoch": 8.61, "learning_rate": 4.175977121067684e-07, "loss": 0.3199, "step": 45199 }, { "epoch": 8.61, "learning_rate": 4.175405147759771e-07, "loss": 1.9853, "step": 45200 }, { "epoch": 8.61, "learning_rate": 4.174833174451859e-07, "loss": 1.2152, "step": 45201 }, { "epoch": 8.61, "learning_rate": 4.1742612011439467e-07, "loss": 2.1334, "step": 45202 }, { "epoch": 8.61, "learning_rate": 4.1736892278360346e-07, "loss": 1.0301, "step": 45203 }, { "epoch": 8.61, "learning_rate": 4.173117254528122e-07, "loss": 2.3351, "step": 45204 }, { "epoch": 8.61, "learning_rate": 4.1725452812202094e-07, "loss": 1.5624, "step": 45205 }, { "epoch": 8.61, "learning_rate": 4.171973307912298e-07, "loss": 1.9907, "step": 45206 }, { "epoch": 8.61, "learning_rate": 4.171401334604385e-07, "loss": 1.0191, "step": 45207 }, { "epoch": 8.61, "learning_rate": 4.170829361296473e-07, "loss": 1.3716, "step": 45208 }, { "epoch": 8.61, "learning_rate": 4.1702573879885606e-07, "loss": 1.4563, "step": 45209 }, { "epoch": 8.61, "learning_rate": 4.1696854146806486e-07, "loss": 1.9637, "step": 45210 }, { "epoch": 8.61, "learning_rate": 4.169113441372736e-07, "loss": 1.1574, "step": 45211 }, { "epoch": 8.61, "learning_rate": 4.1685414680648234e-07, "loss": 0.9144, "step": 45212 }, { "epoch": 8.61, "learning_rate": 4.167969494756912e-07, "loss": 1.1967, "step": 45213 }, { "epoch": 8.61, "learning_rate": 4.1673975214489987e-07, "loss": 0.3573, "step": 45214 }, { "epoch": 8.61, "learning_rate": 4.166825548141087e-07, "loss": 0.921, "step": 45215 }, { "epoch": 8.61, "learning_rate": 4.1662535748331746e-07, "loss": 2.0295, "step": 45216 }, { "epoch": 8.61, "learning_rate": 4.165681601525262e-07, "loss": 1.2309, "step": 45217 }, { "epoch": 8.61, "learning_rate": 4.16510962821735e-07, "loss": 0.7406, "step": 45218 }, { "epoch": 8.61, "learning_rate": 4.1645376549094374e-07, "loss": 2.0206, "step": 45219 }, { "epoch": 8.61, "learning_rate": 4.163965681601526e-07, "loss": 1.3934, "step": 45220 }, { "epoch": 8.61, "learning_rate": 4.163393708293613e-07, "loss": 1.8478, "step": 45221 }, { "epoch": 8.61, "learning_rate": 4.162821734985701e-07, "loss": 1.4761, "step": 45222 }, { "epoch": 8.61, "learning_rate": 4.1622497616777886e-07, "loss": 1.8741, "step": 45223 }, { "epoch": 8.61, "learning_rate": 4.161677788369876e-07, "loss": 2.5445, "step": 45224 }, { "epoch": 8.61, "learning_rate": 4.161105815061964e-07, "loss": 1.0099, "step": 45225 }, { "epoch": 8.61, "learning_rate": 4.1605338417540513e-07, "loss": 1.2377, "step": 45226 }, { "epoch": 8.61, "learning_rate": 4.15996186844614e-07, "loss": 2.009, "step": 45227 }, { "epoch": 8.61, "learning_rate": 4.159389895138227e-07, "loss": 1.0581, "step": 45228 }, { "epoch": 8.62, "learning_rate": 4.158817921830314e-07, "loss": 2.6694, "step": 45229 }, { "epoch": 8.62, "learning_rate": 4.1582459485224026e-07, "loss": 1.4635, "step": 45230 }, { "epoch": 8.62, "learning_rate": 4.15767397521449e-07, "loss": 2.1943, "step": 45231 }, { "epoch": 8.62, "learning_rate": 4.157102001906578e-07, "loss": 2.3504, "step": 45232 }, { "epoch": 8.62, "learning_rate": 4.1565300285986653e-07, "loss": 0.8217, "step": 45233 }, { "epoch": 8.62, "learning_rate": 4.1559580552907527e-07, "loss": 1.6267, "step": 45234 }, { "epoch": 8.62, "learning_rate": 4.155386081982841e-07, "loss": 3.319, "step": 45235 }, { "epoch": 8.62, "learning_rate": 4.154814108674928e-07, "loss": 1.3546, "step": 45236 }, { "epoch": 8.62, "learning_rate": 4.1542421353670165e-07, "loss": 3.0524, "step": 45237 }, { "epoch": 8.62, "learning_rate": 4.153670162059104e-07, "loss": 2.4322, "step": 45238 }, { "epoch": 8.62, "learning_rate": 4.153098188751192e-07, "loss": 2.9219, "step": 45239 }, { "epoch": 8.62, "learning_rate": 4.1525262154432793e-07, "loss": 1.3415, "step": 45240 }, { "epoch": 8.62, "learning_rate": 4.1519542421353667e-07, "loss": 2.0973, "step": 45241 }, { "epoch": 8.62, "learning_rate": 4.151382268827455e-07, "loss": 1.7633, "step": 45242 }, { "epoch": 8.62, "learning_rate": 4.1508102955195426e-07, "loss": 2.4227, "step": 45243 }, { "epoch": 8.62, "learning_rate": 4.1502383222116305e-07, "loss": 1.2612, "step": 45244 }, { "epoch": 8.62, "learning_rate": 4.149666348903718e-07, "loss": 1.1788, "step": 45245 }, { "epoch": 8.62, "learning_rate": 4.1490943755958053e-07, "loss": 1.801, "step": 45246 }, { "epoch": 8.62, "learning_rate": 4.1485224022878933e-07, "loss": 1.6786, "step": 45247 }, { "epoch": 8.62, "learning_rate": 4.1479504289799807e-07, "loss": 1.1493, "step": 45248 }, { "epoch": 8.62, "learning_rate": 4.147378455672069e-07, "loss": 2.8802, "step": 45249 }, { "epoch": 8.62, "learning_rate": 4.1468064823641566e-07, "loss": 1.0911, "step": 45250 }, { "epoch": 8.62, "learning_rate": 4.1462345090562445e-07, "loss": 2.3205, "step": 45251 }, { "epoch": 8.62, "learning_rate": 4.145662535748332e-07, "loss": 1.9727, "step": 45252 }, { "epoch": 8.62, "learning_rate": 4.1450905624404193e-07, "loss": 1.4834, "step": 45253 }, { "epoch": 8.62, "learning_rate": 4.144518589132507e-07, "loss": 1.4114, "step": 45254 }, { "epoch": 8.62, "learning_rate": 4.1439466158245947e-07, "loss": 0.863, "step": 45255 }, { "epoch": 8.62, "learning_rate": 4.143374642516683e-07, "loss": 1.9147, "step": 45256 }, { "epoch": 8.62, "learning_rate": 4.1428026692087705e-07, "loss": 3.2063, "step": 45257 }, { "epoch": 8.62, "learning_rate": 4.1422306959008574e-07, "loss": 0.8385, "step": 45258 }, { "epoch": 8.62, "learning_rate": 4.141658722592946e-07, "loss": 0.3126, "step": 45259 }, { "epoch": 8.62, "learning_rate": 4.1410867492850333e-07, "loss": 1.92, "step": 45260 }, { "epoch": 8.62, "learning_rate": 4.140514775977121e-07, "loss": 2.8943, "step": 45261 }, { "epoch": 8.62, "learning_rate": 4.1399428026692086e-07, "loss": 0.964, "step": 45262 }, { "epoch": 8.62, "learning_rate": 4.139370829361297e-07, "loss": 2.4833, "step": 45263 }, { "epoch": 8.62, "learning_rate": 4.1387988560533845e-07, "loss": 1.2991, "step": 45264 }, { "epoch": 8.62, "learning_rate": 4.138226882745472e-07, "loss": 1.4727, "step": 45265 }, { "epoch": 8.62, "learning_rate": 4.13765490943756e-07, "loss": 0.9015, "step": 45266 }, { "epoch": 8.62, "learning_rate": 4.1370829361296473e-07, "loss": 0.5707, "step": 45267 }, { "epoch": 8.62, "learning_rate": 4.136510962821735e-07, "loss": 1.1164, "step": 45268 }, { "epoch": 8.62, "learning_rate": 4.1359389895138226e-07, "loss": 1.4217, "step": 45269 }, { "epoch": 8.62, "learning_rate": 4.13536701620591e-07, "loss": 0.9411, "step": 45270 }, { "epoch": 8.62, "learning_rate": 4.1347950428979985e-07, "loss": 1.2105, "step": 45271 }, { "epoch": 8.62, "learning_rate": 4.134223069590086e-07, "loss": 2.0859, "step": 45272 }, { "epoch": 8.62, "learning_rate": 4.133651096282174e-07, "loss": 1.6586, "step": 45273 }, { "epoch": 8.62, "learning_rate": 4.133079122974261e-07, "loss": 1.4644, "step": 45274 }, { "epoch": 8.62, "learning_rate": 4.132507149666349e-07, "loss": 2.0895, "step": 45275 }, { "epoch": 8.62, "learning_rate": 4.1319351763584366e-07, "loss": 1.9882, "step": 45276 }, { "epoch": 8.62, "learning_rate": 4.131363203050524e-07, "loss": 1.2415, "step": 45277 }, { "epoch": 8.62, "learning_rate": 4.1307912297426125e-07, "loss": 0.959, "step": 45278 }, { "epoch": 8.62, "learning_rate": 4.1302192564347e-07, "loss": 0.475, "step": 45279 }, { "epoch": 8.62, "learning_rate": 4.129647283126788e-07, "loss": 1.5397, "step": 45280 }, { "epoch": 8.62, "learning_rate": 4.129075309818875e-07, "loss": 0.8331, "step": 45281 }, { "epoch": 8.63, "learning_rate": 4.1285033365109626e-07, "loss": 1.3847, "step": 45282 }, { "epoch": 8.63, "learning_rate": 4.1279313632030506e-07, "loss": 2.5594, "step": 45283 }, { "epoch": 8.63, "learning_rate": 4.127359389895138e-07, "loss": 3.1056, "step": 45284 }, { "epoch": 8.63, "learning_rate": 4.1267874165872265e-07, "loss": 1.118, "step": 45285 }, { "epoch": 8.63, "learning_rate": 4.126215443279314e-07, "loss": 1.3145, "step": 45286 }, { "epoch": 8.63, "learning_rate": 4.125643469971402e-07, "loss": 1.3374, "step": 45287 }, { "epoch": 8.63, "learning_rate": 4.125071496663489e-07, "loss": 1.9544, "step": 45288 }, { "epoch": 8.63, "learning_rate": 4.1244995233555766e-07, "loss": 1.2738, "step": 45289 }, { "epoch": 8.63, "learning_rate": 4.1239275500476646e-07, "loss": 1.514, "step": 45290 }, { "epoch": 8.63, "learning_rate": 4.123355576739752e-07, "loss": 0.466, "step": 45291 }, { "epoch": 8.63, "learning_rate": 4.1227836034318404e-07, "loss": 1.6786, "step": 45292 }, { "epoch": 8.63, "learning_rate": 4.122211630123928e-07, "loss": 0.9894, "step": 45293 }, { "epoch": 8.63, "learning_rate": 4.121639656816015e-07, "loss": 0.5657, "step": 45294 }, { "epoch": 8.63, "learning_rate": 4.121067683508103e-07, "loss": 1.2595, "step": 45295 }, { "epoch": 8.63, "learning_rate": 4.1204957102001906e-07, "loss": 1.3132, "step": 45296 }, { "epoch": 8.63, "learning_rate": 4.1199237368922785e-07, "loss": 1.7814, "step": 45297 }, { "epoch": 8.63, "learning_rate": 4.119351763584366e-07, "loss": 0.8087, "step": 45298 }, { "epoch": 8.63, "learning_rate": 4.1187797902764534e-07, "loss": 1.0804, "step": 45299 }, { "epoch": 8.63, "learning_rate": 4.118207816968542e-07, "loss": 1.5276, "step": 45300 }, { "epoch": 8.63, "learning_rate": 4.117635843660629e-07, "loss": 2.5731, "step": 45301 }, { "epoch": 8.63, "learning_rate": 4.117063870352717e-07, "loss": 1.3575, "step": 45302 }, { "epoch": 8.63, "learning_rate": 4.1164918970448046e-07, "loss": 1.4772, "step": 45303 }, { "epoch": 8.63, "learning_rate": 4.1159199237368925e-07, "loss": 1.5623, "step": 45304 }, { "epoch": 8.63, "learning_rate": 4.11534795042898e-07, "loss": 2.0738, "step": 45305 }, { "epoch": 8.63, "learning_rate": 4.1147759771210673e-07, "loss": 2.2783, "step": 45306 }, { "epoch": 8.63, "learning_rate": 4.114204003813156e-07, "loss": 1.4384, "step": 45307 }, { "epoch": 8.63, "learning_rate": 4.113632030505243e-07, "loss": 1.3202, "step": 45308 }, { "epoch": 8.63, "learning_rate": 4.113060057197331e-07, "loss": 1.5245, "step": 45309 }, { "epoch": 8.63, "learning_rate": 4.1124880838894186e-07, "loss": 0.8001, "step": 45310 }, { "epoch": 8.63, "learning_rate": 4.111916110581506e-07, "loss": 1.6432, "step": 45311 }, { "epoch": 8.63, "learning_rate": 4.111344137273594e-07, "loss": 2.2295, "step": 45312 }, { "epoch": 8.63, "learning_rate": 4.1107721639656813e-07, "loss": 1.0947, "step": 45313 }, { "epoch": 8.63, "learning_rate": 4.11020019065777e-07, "loss": 1.2453, "step": 45314 }, { "epoch": 8.63, "learning_rate": 4.109628217349857e-07, "loss": 1.1046, "step": 45315 }, { "epoch": 8.63, "learning_rate": 4.109056244041945e-07, "loss": 1.7221, "step": 45316 }, { "epoch": 8.63, "learning_rate": 4.1084842707340325e-07, "loss": 1.8678, "step": 45317 }, { "epoch": 8.63, "learning_rate": 4.10791229742612e-07, "loss": 0.8158, "step": 45318 }, { "epoch": 8.63, "learning_rate": 4.107340324118208e-07, "loss": 0.9765, "step": 45319 }, { "epoch": 8.63, "learning_rate": 4.1067683508102953e-07, "loss": 2.0921, "step": 45320 }, { "epoch": 8.63, "learning_rate": 4.106196377502384e-07, "loss": 1.3978, "step": 45321 }, { "epoch": 8.63, "learning_rate": 4.105624404194471e-07, "loss": 1.8598, "step": 45322 }, { "epoch": 8.63, "learning_rate": 4.1050524308865586e-07, "loss": 2.1084, "step": 45323 }, { "epoch": 8.63, "learning_rate": 4.1044804575786465e-07, "loss": 2.2565, "step": 45324 }, { "epoch": 8.63, "learning_rate": 4.103908484270734e-07, "loss": 1.3792, "step": 45325 }, { "epoch": 8.63, "learning_rate": 4.103336510962822e-07, "loss": 1.2172, "step": 45326 }, { "epoch": 8.63, "learning_rate": 4.1027645376549093e-07, "loss": 1.1298, "step": 45327 }, { "epoch": 8.63, "learning_rate": 4.102192564346998e-07, "loss": 2.915, "step": 45328 }, { "epoch": 8.63, "learning_rate": 4.101620591039085e-07, "loss": 2.7151, "step": 45329 }, { "epoch": 8.63, "learning_rate": 4.1010486177311726e-07, "loss": 0.6228, "step": 45330 }, { "epoch": 8.63, "learning_rate": 4.1004766444232605e-07, "loss": 1.3164, "step": 45331 }, { "epoch": 8.63, "learning_rate": 4.099904671115348e-07, "loss": 1.7447, "step": 45332 }, { "epoch": 8.63, "learning_rate": 4.099332697807436e-07, "loss": 2.1816, "step": 45333 }, { "epoch": 8.64, "learning_rate": 4.098760724499523e-07, "loss": 1.2291, "step": 45334 }, { "epoch": 8.64, "learning_rate": 4.0981887511916107e-07, "loss": 0.9846, "step": 45335 }, { "epoch": 8.64, "learning_rate": 4.097616777883699e-07, "loss": 1.1658, "step": 45336 }, { "epoch": 8.64, "learning_rate": 4.0970448045757865e-07, "loss": 0.7725, "step": 45337 }, { "epoch": 8.64, "learning_rate": 4.0964728312678745e-07, "loss": 1.2891, "step": 45338 }, { "epoch": 8.64, "learning_rate": 4.095900857959962e-07, "loss": 1.4693, "step": 45339 }, { "epoch": 8.64, "learning_rate": 4.09532888465205e-07, "loss": 1.7351, "step": 45340 }, { "epoch": 8.64, "learning_rate": 4.094756911344137e-07, "loss": 1.8524, "step": 45341 }, { "epoch": 8.64, "learning_rate": 4.0941849380362246e-07, "loss": 1.8151, "step": 45342 }, { "epoch": 8.64, "learning_rate": 4.093612964728313e-07, "loss": 0.351, "step": 45343 }, { "epoch": 8.64, "learning_rate": 4.0930409914204005e-07, "loss": 1.1072, "step": 45344 }, { "epoch": 8.64, "learning_rate": 4.0924690181124885e-07, "loss": 2.0114, "step": 45345 }, { "epoch": 8.64, "learning_rate": 4.091897044804576e-07, "loss": 1.3616, "step": 45346 }, { "epoch": 8.64, "learning_rate": 4.0913250714966633e-07, "loss": 1.5744, "step": 45347 }, { "epoch": 8.64, "learning_rate": 4.090753098188751e-07, "loss": 1.1899, "step": 45348 }, { "epoch": 8.64, "learning_rate": 4.0901811248808386e-07, "loss": 1.2994, "step": 45349 }, { "epoch": 8.64, "learning_rate": 4.089609151572927e-07, "loss": 2.1847, "step": 45350 }, { "epoch": 8.64, "learning_rate": 4.0890371782650145e-07, "loss": 0.8683, "step": 45351 }, { "epoch": 8.64, "learning_rate": 4.0884652049571024e-07, "loss": 1.5629, "step": 45352 }, { "epoch": 8.64, "learning_rate": 4.08789323164919e-07, "loss": 1.6814, "step": 45353 }, { "epoch": 8.64, "learning_rate": 4.087321258341277e-07, "loss": 0.7588, "step": 45354 }, { "epoch": 8.64, "learning_rate": 4.086749285033365e-07, "loss": 1.5607, "step": 45355 }, { "epoch": 8.64, "learning_rate": 4.0861773117254526e-07, "loss": 0.9785, "step": 45356 }, { "epoch": 8.64, "learning_rate": 4.085605338417541e-07, "loss": 1.6694, "step": 45357 }, { "epoch": 8.64, "learning_rate": 4.0850333651096285e-07, "loss": 0.5131, "step": 45358 }, { "epoch": 8.64, "learning_rate": 4.084461391801716e-07, "loss": 3.1901, "step": 45359 }, { "epoch": 8.64, "learning_rate": 4.083889418493804e-07, "loss": 1.9252, "step": 45360 }, { "epoch": 8.64, "learning_rate": 4.083317445185891e-07, "loss": 1.6437, "step": 45361 }, { "epoch": 8.64, "learning_rate": 4.082745471877979e-07, "loss": 2.6555, "step": 45362 }, { "epoch": 8.64, "learning_rate": 4.0821734985700666e-07, "loss": 1.1597, "step": 45363 }, { "epoch": 8.64, "learning_rate": 4.081601525262154e-07, "loss": 0.3081, "step": 45364 }, { "epoch": 8.64, "learning_rate": 4.0810295519542425e-07, "loss": 1.3185, "step": 45365 }, { "epoch": 8.64, "learning_rate": 4.08045757864633e-07, "loss": 0.5458, "step": 45366 }, { "epoch": 8.64, "learning_rate": 4.079885605338418e-07, "loss": 1.2552, "step": 45367 }, { "epoch": 8.64, "learning_rate": 4.079313632030505e-07, "loss": 1.8673, "step": 45368 }, { "epoch": 8.64, "learning_rate": 4.078741658722593e-07, "loss": 2.2997, "step": 45369 }, { "epoch": 8.64, "learning_rate": 4.0781696854146806e-07, "loss": 1.635, "step": 45370 }, { "epoch": 8.64, "learning_rate": 4.077597712106768e-07, "loss": 2.885, "step": 45371 }, { "epoch": 8.64, "learning_rate": 4.0770257387988564e-07, "loss": 0.9147, "step": 45372 }, { "epoch": 8.64, "learning_rate": 4.076453765490944e-07, "loss": 0.9204, "step": 45373 }, { "epoch": 8.64, "learning_rate": 4.075881792183032e-07, "loss": 1.2317, "step": 45374 }, { "epoch": 8.64, "learning_rate": 4.075309818875119e-07, "loss": 1.5218, "step": 45375 }, { "epoch": 8.64, "learning_rate": 4.0747378455672066e-07, "loss": 1.1489, "step": 45376 }, { "epoch": 8.64, "learning_rate": 4.0741658722592945e-07, "loss": 2.0176, "step": 45377 }, { "epoch": 8.64, "learning_rate": 4.073593898951382e-07, "loss": 0.4744, "step": 45378 }, { "epoch": 8.64, "learning_rate": 4.0730219256434704e-07, "loss": 2.6747, "step": 45379 }, { "epoch": 8.64, "learning_rate": 4.072449952335558e-07, "loss": 1.794, "step": 45380 }, { "epoch": 8.64, "learning_rate": 4.071877979027646e-07, "loss": 1.6156, "step": 45381 }, { "epoch": 8.64, "learning_rate": 4.071306005719733e-07, "loss": 1.4537, "step": 45382 }, { "epoch": 8.64, "learning_rate": 4.0707340324118206e-07, "loss": 1.6588, "step": 45383 }, { "epoch": 8.64, "learning_rate": 4.0701620591039085e-07, "loss": 0.5615, "step": 45384 }, { "epoch": 8.64, "learning_rate": 4.069590085795996e-07, "loss": 1.1546, "step": 45385 }, { "epoch": 8.64, "learning_rate": 4.0690181124880844e-07, "loss": 1.8326, "step": 45386 }, { "epoch": 8.65, "learning_rate": 4.068446139180172e-07, "loss": 2.7803, "step": 45387 }, { "epoch": 8.65, "learning_rate": 4.067874165872259e-07, "loss": 1.4617, "step": 45388 }, { "epoch": 8.65, "learning_rate": 4.067302192564347e-07, "loss": 1.8246, "step": 45389 }, { "epoch": 8.65, "learning_rate": 4.0667302192564346e-07, "loss": 2.4861, "step": 45390 }, { "epoch": 8.65, "learning_rate": 4.0661582459485225e-07, "loss": 1.1873, "step": 45391 }, { "epoch": 8.65, "learning_rate": 4.06558627264061e-07, "loss": 2.1445, "step": 45392 }, { "epoch": 8.65, "learning_rate": 4.0650142993326984e-07, "loss": 2.0638, "step": 45393 }, { "epoch": 8.65, "learning_rate": 4.064442326024786e-07, "loss": 1.9639, "step": 45394 }, { "epoch": 8.65, "learning_rate": 4.063870352716873e-07, "loss": 1.4565, "step": 45395 }, { "epoch": 8.65, "learning_rate": 4.063298379408961e-07, "loss": 1.2907, "step": 45396 }, { "epoch": 8.65, "learning_rate": 4.0627264061010485e-07, "loss": 2.2401, "step": 45397 }, { "epoch": 8.65, "learning_rate": 4.062154432793137e-07, "loss": 0.7428, "step": 45398 }, { "epoch": 8.65, "learning_rate": 4.061582459485224e-07, "loss": 0.1849, "step": 45399 }, { "epoch": 8.65, "learning_rate": 4.0610104861773113e-07, "loss": 0.5193, "step": 45400 }, { "epoch": 8.65, "learning_rate": 4.0604385128694e-07, "loss": 1.5433, "step": 45401 }, { "epoch": 8.65, "learning_rate": 4.059866539561487e-07, "loss": 1.2782, "step": 45402 }, { "epoch": 8.65, "learning_rate": 4.059294566253575e-07, "loss": 1.5518, "step": 45403 }, { "epoch": 8.65, "learning_rate": 4.0587225929456625e-07, "loss": 2.4894, "step": 45404 }, { "epoch": 8.65, "learning_rate": 4.058150619637751e-07, "loss": 0.6419, "step": 45405 }, { "epoch": 8.65, "learning_rate": 4.057578646329838e-07, "loss": 0.9092, "step": 45406 }, { "epoch": 8.65, "learning_rate": 4.0570066730219253e-07, "loss": 1.662, "step": 45407 }, { "epoch": 8.65, "learning_rate": 4.056434699714014e-07, "loss": 2.3052, "step": 45408 }, { "epoch": 8.65, "learning_rate": 4.055862726406101e-07, "loss": 0.6588, "step": 45409 }, { "epoch": 8.65, "learning_rate": 4.055290753098189e-07, "loss": 1.6538, "step": 45410 }, { "epoch": 8.65, "learning_rate": 4.0547187797902765e-07, "loss": 0.8712, "step": 45411 }, { "epoch": 8.65, "learning_rate": 4.054146806482364e-07, "loss": 1.6061, "step": 45412 }, { "epoch": 8.65, "learning_rate": 4.053574833174452e-07, "loss": 1.4268, "step": 45413 }, { "epoch": 8.65, "learning_rate": 4.053002859866539e-07, "loss": 1.0522, "step": 45414 }, { "epoch": 8.65, "learning_rate": 4.0524308865586277e-07, "loss": 0.8329, "step": 45415 }, { "epoch": 8.65, "learning_rate": 4.051858913250715e-07, "loss": 0.9516, "step": 45416 }, { "epoch": 8.65, "learning_rate": 4.0512869399428025e-07, "loss": 2.3632, "step": 45417 }, { "epoch": 8.65, "learning_rate": 4.0507149666348905e-07, "loss": 0.9127, "step": 45418 }, { "epoch": 8.65, "learning_rate": 4.050142993326978e-07, "loss": 2.3172, "step": 45419 }, { "epoch": 8.65, "learning_rate": 4.0495710200190664e-07, "loss": 1.0845, "step": 45420 }, { "epoch": 8.65, "learning_rate": 4.048999046711153e-07, "loss": 0.7053, "step": 45421 }, { "epoch": 8.65, "learning_rate": 4.0484270734032417e-07, "loss": 1.2563, "step": 45422 }, { "epoch": 8.65, "learning_rate": 4.047855100095329e-07, "loss": 1.3192, "step": 45423 }, { "epoch": 8.65, "learning_rate": 4.0472831267874165e-07, "loss": 0.8673, "step": 45424 }, { "epoch": 8.65, "learning_rate": 4.0467111534795045e-07, "loss": 0.3801, "step": 45425 }, { "epoch": 8.65, "learning_rate": 4.046139180171592e-07, "loss": 1.4629, "step": 45426 }, { "epoch": 8.65, "learning_rate": 4.0455672068636803e-07, "loss": 0.8402, "step": 45427 }, { "epoch": 8.65, "learning_rate": 4.044995233555767e-07, "loss": 1.4055, "step": 45428 }, { "epoch": 8.65, "learning_rate": 4.0444232602478546e-07, "loss": 1.2954, "step": 45429 }, { "epoch": 8.65, "learning_rate": 4.043851286939943e-07, "loss": 1.1048, "step": 45430 }, { "epoch": 8.65, "learning_rate": 4.0432793136320305e-07, "loss": 2.4575, "step": 45431 }, { "epoch": 8.65, "learning_rate": 4.0427073403241184e-07, "loss": 1.2632, "step": 45432 }, { "epoch": 8.65, "learning_rate": 4.042135367016206e-07, "loss": 1.6932, "step": 45433 }, { "epoch": 8.65, "learning_rate": 4.0415633937082943e-07, "loss": 1.8076, "step": 45434 }, { "epoch": 8.65, "learning_rate": 4.040991420400381e-07, "loss": 3.2492, "step": 45435 }, { "epoch": 8.65, "learning_rate": 4.0404194470924686e-07, "loss": 0.8049, "step": 45436 }, { "epoch": 8.65, "learning_rate": 4.039847473784557e-07, "loss": 1.5512, "step": 45437 }, { "epoch": 8.65, "learning_rate": 4.0392755004766445e-07, "loss": 1.2052, "step": 45438 }, { "epoch": 8.66, "learning_rate": 4.0387035271687324e-07, "loss": 2.6922, "step": 45439 }, { "epoch": 8.66, "learning_rate": 4.03813155386082e-07, "loss": 0.8879, "step": 45440 }, { "epoch": 8.66, "learning_rate": 4.037559580552907e-07, "loss": 2.5112, "step": 45441 }, { "epoch": 8.66, "learning_rate": 4.0369876072449957e-07, "loss": 1.6772, "step": 45442 }, { "epoch": 8.66, "learning_rate": 4.0364156339370826e-07, "loss": 2.0473, "step": 45443 }, { "epoch": 8.66, "learning_rate": 4.035843660629171e-07, "loss": 0.6868, "step": 45444 }, { "epoch": 8.66, "learning_rate": 4.0352716873212585e-07, "loss": 1.5324, "step": 45445 }, { "epoch": 8.66, "learning_rate": 4.0346997140133464e-07, "loss": 2.9768, "step": 45446 }, { "epoch": 8.66, "learning_rate": 4.034127740705434e-07, "loss": 1.0345, "step": 45447 }, { "epoch": 8.66, "learning_rate": 4.033555767397521e-07, "loss": 0.8225, "step": 45448 }, { "epoch": 8.66, "learning_rate": 4.0329837940896097e-07, "loss": 1.8195, "step": 45449 }, { "epoch": 8.66, "learning_rate": 4.0324118207816966e-07, "loss": 2.4219, "step": 45450 }, { "epoch": 8.66, "learning_rate": 4.031839847473785e-07, "loss": 1.0805, "step": 45451 }, { "epoch": 8.66, "learning_rate": 4.0312678741658724e-07, "loss": 1.71, "step": 45452 }, { "epoch": 8.66, "learning_rate": 4.03069590085796e-07, "loss": 0.6429, "step": 45453 }, { "epoch": 8.66, "learning_rate": 4.030123927550048e-07, "loss": 0.4714, "step": 45454 }, { "epoch": 8.66, "learning_rate": 4.029551954242135e-07, "loss": 1.6608, "step": 45455 }, { "epoch": 8.66, "learning_rate": 4.0289799809342237e-07, "loss": 2.6001, "step": 45456 }, { "epoch": 8.66, "learning_rate": 4.0284080076263105e-07, "loss": 1.8102, "step": 45457 }, { "epoch": 8.66, "learning_rate": 4.027836034318399e-07, "loss": 2.2236, "step": 45458 }, { "epoch": 8.66, "learning_rate": 4.0272640610104864e-07, "loss": 0.966, "step": 45459 }, { "epoch": 8.66, "learning_rate": 4.026692087702574e-07, "loss": 1.5534, "step": 45460 }, { "epoch": 8.66, "learning_rate": 4.026120114394662e-07, "loss": 1.5881, "step": 45461 }, { "epoch": 8.66, "learning_rate": 4.025548141086749e-07, "loss": 1.9538, "step": 45462 }, { "epoch": 8.66, "learning_rate": 4.0249761677788376e-07, "loss": 2.1344, "step": 45463 }, { "epoch": 8.66, "learning_rate": 4.024404194470925e-07, "loss": 1.9383, "step": 45464 }, { "epoch": 8.66, "learning_rate": 4.023832221163012e-07, "loss": 0.8205, "step": 45465 }, { "epoch": 8.66, "learning_rate": 4.0232602478551004e-07, "loss": 1.3229, "step": 45466 }, { "epoch": 8.66, "learning_rate": 4.022688274547188e-07, "loss": 1.828, "step": 45467 }, { "epoch": 8.66, "learning_rate": 4.022116301239276e-07, "loss": 2.4121, "step": 45468 }, { "epoch": 8.66, "learning_rate": 4.021544327931363e-07, "loss": 2.4506, "step": 45469 }, { "epoch": 8.66, "learning_rate": 4.0209723546234516e-07, "loss": 2.9073, "step": 45470 }, { "epoch": 8.66, "learning_rate": 4.020400381315539e-07, "loss": 3.1052, "step": 45471 }, { "epoch": 8.66, "learning_rate": 4.019828408007626e-07, "loss": 1.3862, "step": 45472 }, { "epoch": 8.66, "learning_rate": 4.0192564346997144e-07, "loss": 2.1204, "step": 45473 }, { "epoch": 8.66, "learning_rate": 4.018684461391802e-07, "loss": 1.7107, "step": 45474 }, { "epoch": 8.66, "learning_rate": 4.0181124880838897e-07, "loss": 1.7854, "step": 45475 }, { "epoch": 8.66, "learning_rate": 4.017540514775977e-07, "loss": 2.5906, "step": 45476 }, { "epoch": 8.66, "learning_rate": 4.0169685414680645e-07, "loss": 1.0505, "step": 45477 }, { "epoch": 8.66, "learning_rate": 4.016396568160153e-07, "loss": 2.4519, "step": 45478 }, { "epoch": 8.66, "learning_rate": 4.01582459485224e-07, "loss": 1.8491, "step": 45479 }, { "epoch": 8.66, "learning_rate": 4.0152526215443283e-07, "loss": 0.9311, "step": 45480 }, { "epoch": 8.66, "learning_rate": 4.014680648236416e-07, "loss": 1.7418, "step": 45481 }, { "epoch": 8.66, "learning_rate": 4.014108674928503e-07, "loss": 1.7837, "step": 45482 }, { "epoch": 8.66, "learning_rate": 4.013536701620591e-07, "loss": 1.2164, "step": 45483 }, { "epoch": 8.66, "learning_rate": 4.0129647283126785e-07, "loss": 2.7963, "step": 45484 }, { "epoch": 8.66, "learning_rate": 4.012392755004767e-07, "loss": 2.2808, "step": 45485 }, { "epoch": 8.66, "learning_rate": 4.011820781696854e-07, "loss": 3.7796, "step": 45486 }, { "epoch": 8.66, "learning_rate": 4.0112488083889423e-07, "loss": 2.3877, "step": 45487 }, { "epoch": 8.66, "learning_rate": 4.01067683508103e-07, "loss": 0.8555, "step": 45488 }, { "epoch": 8.66, "learning_rate": 4.010104861773117e-07, "loss": 0.6464, "step": 45489 }, { "epoch": 8.66, "learning_rate": 4.009532888465205e-07, "loss": 1.3937, "step": 45490 }, { "epoch": 8.66, "learning_rate": 4.0089609151572925e-07, "loss": 1.3383, "step": 45491 }, { "epoch": 8.67, "learning_rate": 4.008388941849381e-07, "loss": 1.6049, "step": 45492 }, { "epoch": 8.67, "learning_rate": 4.0078169685414684e-07, "loss": 2.8892, "step": 45493 }, { "epoch": 8.67, "learning_rate": 4.007244995233555e-07, "loss": 0.4822, "step": 45494 }, { "epoch": 8.67, "learning_rate": 4.0066730219256437e-07, "loss": 2.1096, "step": 45495 }, { "epoch": 8.67, "learning_rate": 4.006101048617731e-07, "loss": 0.4083, "step": 45496 }, { "epoch": 8.67, "learning_rate": 4.005529075309819e-07, "loss": 2.001, "step": 45497 }, { "epoch": 8.67, "learning_rate": 4.0049571020019065e-07, "loss": 0.9336, "step": 45498 }, { "epoch": 8.67, "learning_rate": 4.004385128693995e-07, "loss": 1.8562, "step": 45499 }, { "epoch": 8.67, "learning_rate": 4.0038131553860823e-07, "loss": 0.9371, "step": 45500 }, { "epoch": 8.67, "learning_rate": 4.003241182078169e-07, "loss": 1.3366, "step": 45501 }, { "epoch": 8.67, "learning_rate": 4.0026692087702577e-07, "loss": 1.3875, "step": 45502 }, { "epoch": 8.67, "learning_rate": 4.002097235462345e-07, "loss": 1.0222, "step": 45503 }, { "epoch": 8.67, "learning_rate": 4.001525262154433e-07, "loss": 0.3952, "step": 45504 }, { "epoch": 8.67, "learning_rate": 4.0009532888465205e-07, "loss": 1.0555, "step": 45505 }, { "epoch": 8.67, "learning_rate": 4.000381315538608e-07, "loss": 0.9731, "step": 45506 }, { "epoch": 8.67, "learning_rate": 3.9998093422306963e-07, "loss": 1.1773, "step": 45507 }, { "epoch": 8.67, "learning_rate": 3.999237368922783e-07, "loss": 2.1363, "step": 45508 }, { "epoch": 8.67, "learning_rate": 3.9986653956148717e-07, "loss": 0.5381, "step": 45509 }, { "epoch": 8.67, "learning_rate": 3.998093422306959e-07, "loss": 2.5548, "step": 45510 }, { "epoch": 8.67, "learning_rate": 3.997521448999047e-07, "loss": 0.6604, "step": 45511 }, { "epoch": 8.67, "learning_rate": 3.9969494756911344e-07, "loss": 1.207, "step": 45512 }, { "epoch": 8.67, "learning_rate": 3.996377502383222e-07, "loss": 0.9737, "step": 45513 }, { "epoch": 8.67, "learning_rate": 3.9958055290753103e-07, "loss": 0.8799, "step": 45514 }, { "epoch": 8.67, "learning_rate": 3.9952335557673977e-07, "loss": 0.72, "step": 45515 }, { "epoch": 8.67, "learning_rate": 3.9946615824594857e-07, "loss": 2.2139, "step": 45516 }, { "epoch": 8.67, "learning_rate": 3.994089609151573e-07, "loss": 0.4866, "step": 45517 }, { "epoch": 8.67, "learning_rate": 3.9935176358436605e-07, "loss": 1.3245, "step": 45518 }, { "epoch": 8.67, "learning_rate": 3.9929456625357484e-07, "loss": 1.7607, "step": 45519 }, { "epoch": 8.67, "learning_rate": 3.992373689227836e-07, "loss": 0.8009, "step": 45520 }, { "epoch": 8.67, "learning_rate": 3.9918017159199243e-07, "loss": 1.6452, "step": 45521 }, { "epoch": 8.67, "learning_rate": 3.9912297426120117e-07, "loss": 0.831, "step": 45522 }, { "epoch": 8.67, "learning_rate": 3.9906577693040996e-07, "loss": 0.4084, "step": 45523 }, { "epoch": 8.67, "learning_rate": 3.990085795996187e-07, "loss": 2.5222, "step": 45524 }, { "epoch": 8.67, "learning_rate": 3.9895138226882745e-07, "loss": 0.3103, "step": 45525 }, { "epoch": 8.67, "learning_rate": 3.9889418493803624e-07, "loss": 1.5902, "step": 45526 }, { "epoch": 8.67, "learning_rate": 3.98836987607245e-07, "loss": 1.8316, "step": 45527 }, { "epoch": 8.67, "learning_rate": 3.9877979027645383e-07, "loss": 1.4287, "step": 45528 }, { "epoch": 8.67, "learning_rate": 3.9872259294566257e-07, "loss": 2.6304, "step": 45529 }, { "epoch": 8.67, "learning_rate": 3.9866539561487126e-07, "loss": 2.4136, "step": 45530 }, { "epoch": 8.67, "learning_rate": 3.986081982840801e-07, "loss": 2.1964, "step": 45531 }, { "epoch": 8.67, "learning_rate": 3.9855100095328884e-07, "loss": 0.7491, "step": 45532 }, { "epoch": 8.67, "learning_rate": 3.9849380362249764e-07, "loss": 3.1142, "step": 45533 }, { "epoch": 8.67, "learning_rate": 3.984366062917064e-07, "loss": 1.3678, "step": 45534 }, { "epoch": 8.67, "learning_rate": 3.983794089609151e-07, "loss": 1.6119, "step": 45535 }, { "epoch": 8.67, "learning_rate": 3.9832221163012397e-07, "loss": 1.9157, "step": 45536 }, { "epoch": 8.67, "learning_rate": 3.982650142993327e-07, "loss": 1.3206, "step": 45537 }, { "epoch": 8.67, "learning_rate": 3.982078169685415e-07, "loss": 1.0737, "step": 45538 }, { "epoch": 8.67, "learning_rate": 3.9815061963775024e-07, "loss": 1.2205, "step": 45539 }, { "epoch": 8.67, "learning_rate": 3.9809342230695903e-07, "loss": 1.9743, "step": 45540 }, { "epoch": 8.67, "learning_rate": 3.980362249761678e-07, "loss": 2.7539, "step": 45541 }, { "epoch": 8.67, "learning_rate": 3.979790276453765e-07, "loss": 1.3023, "step": 45542 }, { "epoch": 8.67, "learning_rate": 3.9792183031458536e-07, "loss": 1.9879, "step": 45543 }, { "epoch": 8.68, "learning_rate": 3.978646329837941e-07, "loss": 1.2726, "step": 45544 }, { "epoch": 8.68, "learning_rate": 3.978074356530029e-07, "loss": 0.4297, "step": 45545 }, { "epoch": 8.68, "learning_rate": 3.9775023832221164e-07, "loss": 1.2266, "step": 45546 }, { "epoch": 8.68, "learning_rate": 3.976930409914204e-07, "loss": 1.9806, "step": 45547 }, { "epoch": 8.68, "learning_rate": 3.9763584366062917e-07, "loss": 1.6101, "step": 45548 }, { "epoch": 8.68, "learning_rate": 3.975786463298379e-07, "loss": 2.4155, "step": 45549 }, { "epoch": 8.68, "learning_rate": 3.9752144899904676e-07, "loss": 3.0623, "step": 45550 }, { "epoch": 8.68, "learning_rate": 3.974642516682555e-07, "loss": 3.0188, "step": 45551 }, { "epoch": 8.68, "learning_rate": 3.974070543374643e-07, "loss": 1.4381, "step": 45552 }, { "epoch": 8.68, "learning_rate": 3.9734985700667304e-07, "loss": 0.3274, "step": 45553 }, { "epoch": 8.68, "learning_rate": 3.972926596758818e-07, "loss": 1.7798, "step": 45554 }, { "epoch": 8.68, "learning_rate": 3.9723546234509057e-07, "loss": 1.4601, "step": 45555 }, { "epoch": 8.68, "learning_rate": 3.971782650142993e-07, "loss": 1.2556, "step": 45556 }, { "epoch": 8.68, "learning_rate": 3.9712106768350816e-07, "loss": 1.7007, "step": 45557 }, { "epoch": 8.68, "learning_rate": 3.970638703527169e-07, "loss": 0.8888, "step": 45558 }, { "epoch": 8.68, "learning_rate": 3.9700667302192564e-07, "loss": 2.3071, "step": 45559 }, { "epoch": 8.68, "learning_rate": 3.9694947569113443e-07, "loss": 0.5927, "step": 45560 }, { "epoch": 8.68, "learning_rate": 3.968922783603432e-07, "loss": 0.8822, "step": 45561 }, { "epoch": 8.68, "learning_rate": 3.9683508102955197e-07, "loss": 1.2837, "step": 45562 }, { "epoch": 8.68, "learning_rate": 3.967778836987607e-07, "loss": 1.239, "step": 45563 }, { "epoch": 8.68, "learning_rate": 3.9672068636796956e-07, "loss": 1.417, "step": 45564 }, { "epoch": 8.68, "learning_rate": 3.966634890371783e-07, "loss": 2.1215, "step": 45565 }, { "epoch": 8.68, "learning_rate": 3.9660629170638704e-07, "loss": 1.9281, "step": 45566 }, { "epoch": 8.68, "learning_rate": 3.9654909437559583e-07, "loss": 1.8911, "step": 45567 }, { "epoch": 8.68, "learning_rate": 3.9649189704480457e-07, "loss": 2.2421, "step": 45568 }, { "epoch": 8.68, "learning_rate": 3.9643469971401337e-07, "loss": 1.194, "step": 45569 }, { "epoch": 8.68, "learning_rate": 3.963775023832221e-07, "loss": 0.6642, "step": 45570 }, { "epoch": 8.68, "learning_rate": 3.9632030505243085e-07, "loss": 2.427, "step": 45571 }, { "epoch": 8.68, "learning_rate": 3.962631077216397e-07, "loss": 0.989, "step": 45572 }, { "epoch": 8.68, "learning_rate": 3.9620591039084844e-07, "loss": 2.4938, "step": 45573 }, { "epoch": 8.68, "learning_rate": 3.9614871306005723e-07, "loss": 2.1737, "step": 45574 }, { "epoch": 8.68, "learning_rate": 3.9609151572926597e-07, "loss": 1.4443, "step": 45575 }, { "epoch": 8.68, "learning_rate": 3.9603431839847477e-07, "loss": 1.8697, "step": 45576 }, { "epoch": 8.68, "learning_rate": 3.959771210676835e-07, "loss": 1.2488, "step": 45577 }, { "epoch": 8.68, "learning_rate": 3.9591992373689225e-07, "loss": 2.1985, "step": 45578 }, { "epoch": 8.68, "learning_rate": 3.958627264061011e-07, "loss": 0.9161, "step": 45579 }, { "epoch": 8.68, "learning_rate": 3.9580552907530983e-07, "loss": 1.1536, "step": 45580 }, { "epoch": 8.68, "learning_rate": 3.9574833174451863e-07, "loss": 1.0079, "step": 45581 }, { "epoch": 8.68, "learning_rate": 3.9569113441372737e-07, "loss": 1.0064, "step": 45582 }, { "epoch": 8.68, "learning_rate": 3.956339370829361e-07, "loss": 3.0916, "step": 45583 }, { "epoch": 8.68, "learning_rate": 3.955767397521449e-07, "loss": 1.6852, "step": 45584 }, { "epoch": 8.68, "learning_rate": 3.9551954242135364e-07, "loss": 1.3333, "step": 45585 }, { "epoch": 8.68, "learning_rate": 3.954623450905625e-07, "loss": 2.4052, "step": 45586 }, { "epoch": 8.68, "learning_rate": 3.9540514775977123e-07, "loss": 1.2459, "step": 45587 }, { "epoch": 8.68, "learning_rate": 3.9534795042898e-07, "loss": 1.4389, "step": 45588 }, { "epoch": 8.68, "learning_rate": 3.9529075309818877e-07, "loss": 2.2197, "step": 45589 }, { "epoch": 8.68, "learning_rate": 3.952335557673975e-07, "loss": 1.2263, "step": 45590 }, { "epoch": 8.68, "learning_rate": 3.951763584366063e-07, "loss": 1.3147, "step": 45591 }, { "epoch": 8.68, "learning_rate": 3.9511916110581504e-07, "loss": 2.1028, "step": 45592 }, { "epoch": 8.68, "learning_rate": 3.950619637750239e-07, "loss": 1.3301, "step": 45593 }, { "epoch": 8.68, "learning_rate": 3.9500476644423263e-07, "loss": 1.5857, "step": 45594 }, { "epoch": 8.68, "learning_rate": 3.9494756911344137e-07, "loss": 1.2024, "step": 45595 }, { "epoch": 8.68, "learning_rate": 3.9489037178265016e-07, "loss": 1.4174, "step": 45596 }, { "epoch": 8.69, "learning_rate": 3.948331744518589e-07, "loss": 1.7356, "step": 45597 }, { "epoch": 8.69, "learning_rate": 3.947759771210677e-07, "loss": 2.9612, "step": 45598 }, { "epoch": 8.69, "learning_rate": 3.9471877979027644e-07, "loss": 2.0814, "step": 45599 }, { "epoch": 8.69, "learning_rate": 3.946615824594852e-07, "loss": 1.5424, "step": 45600 }, { "epoch": 8.69, "learning_rate": 3.9460438512869403e-07, "loss": 2.323, "step": 45601 }, { "epoch": 8.69, "learning_rate": 3.9454718779790277e-07, "loss": 1.1413, "step": 45602 }, { "epoch": 8.69, "learning_rate": 3.9448999046711156e-07, "loss": 1.3317, "step": 45603 }, { "epoch": 8.69, "learning_rate": 3.944327931363203e-07, "loss": 1.7629, "step": 45604 }, { "epoch": 8.69, "learning_rate": 3.943755958055291e-07, "loss": 1.7226, "step": 45605 }, { "epoch": 8.69, "learning_rate": 3.9431839847473784e-07, "loss": 2.7402, "step": 45606 }, { "epoch": 8.69, "learning_rate": 3.942612011439466e-07, "loss": 0.9442, "step": 45607 }, { "epoch": 8.69, "learning_rate": 3.942040038131554e-07, "loss": 1.1418, "step": 45608 }, { "epoch": 8.69, "learning_rate": 3.9414680648236417e-07, "loss": 2.7241, "step": 45609 }, { "epoch": 8.69, "learning_rate": 3.9408960915157296e-07, "loss": 1.6851, "step": 45610 }, { "epoch": 8.69, "learning_rate": 3.940324118207817e-07, "loss": 0.7338, "step": 45611 }, { "epoch": 8.69, "learning_rate": 3.9397521448999044e-07, "loss": 2.3046, "step": 45612 }, { "epoch": 8.69, "learning_rate": 3.9391801715919924e-07, "loss": 0.65, "step": 45613 }, { "epoch": 8.69, "learning_rate": 3.93860819828408e-07, "loss": 1.1427, "step": 45614 }, { "epoch": 8.69, "learning_rate": 3.938036224976168e-07, "loss": 2.3173, "step": 45615 }, { "epoch": 8.69, "learning_rate": 3.9374642516682556e-07, "loss": 2.1617, "step": 45616 }, { "epoch": 8.69, "learning_rate": 3.9368922783603436e-07, "loss": 0.9699, "step": 45617 }, { "epoch": 8.69, "learning_rate": 3.936320305052431e-07, "loss": 0.8387, "step": 45618 }, { "epoch": 8.69, "learning_rate": 3.9357483317445184e-07, "loss": 0.4792, "step": 45619 }, { "epoch": 8.69, "learning_rate": 3.9351763584366063e-07, "loss": 2.4374, "step": 45620 }, { "epoch": 8.69, "learning_rate": 3.934604385128694e-07, "loss": 0.8953, "step": 45621 }, { "epoch": 8.69, "learning_rate": 3.934032411820782e-07, "loss": 0.7445, "step": 45622 }, { "epoch": 8.69, "learning_rate": 3.9334604385128696e-07, "loss": 1.1287, "step": 45623 }, { "epoch": 8.69, "learning_rate": 3.932888465204957e-07, "loss": 0.7547, "step": 45624 }, { "epoch": 8.69, "learning_rate": 3.932316491897045e-07, "loss": 1.1073, "step": 45625 }, { "epoch": 8.69, "learning_rate": 3.9317445185891324e-07, "loss": 1.6813, "step": 45626 }, { "epoch": 8.69, "learning_rate": 3.9311725452812203e-07, "loss": 1.5483, "step": 45627 }, { "epoch": 8.69, "learning_rate": 3.9306005719733077e-07, "loss": 3.6626, "step": 45628 }, { "epoch": 8.69, "learning_rate": 3.930028598665396e-07, "loss": 1.3012, "step": 45629 }, { "epoch": 8.69, "learning_rate": 3.9294566253574836e-07, "loss": 1.5353, "step": 45630 }, { "epoch": 8.69, "learning_rate": 3.928884652049571e-07, "loss": 1.9595, "step": 45631 }, { "epoch": 8.69, "learning_rate": 3.928312678741659e-07, "loss": 2.4511, "step": 45632 }, { "epoch": 8.69, "learning_rate": 3.9277407054337464e-07, "loss": 3.2929, "step": 45633 }, { "epoch": 8.69, "learning_rate": 3.9271687321258343e-07, "loss": 1.3158, "step": 45634 }, { "epoch": 8.69, "learning_rate": 3.9265967588179217e-07, "loss": 1.3468, "step": 45635 }, { "epoch": 8.69, "learning_rate": 3.926024785510009e-07, "loss": 1.4271, "step": 45636 }, { "epoch": 8.69, "learning_rate": 3.9254528122020976e-07, "loss": 0.3785, "step": 45637 }, { "epoch": 8.69, "learning_rate": 3.924880838894185e-07, "loss": 1.0613, "step": 45638 }, { "epoch": 8.69, "learning_rate": 3.924308865586273e-07, "loss": 1.5834, "step": 45639 }, { "epoch": 8.69, "learning_rate": 3.9237368922783603e-07, "loss": 0.9582, "step": 45640 }, { "epoch": 8.69, "learning_rate": 3.9231649189704483e-07, "loss": 1.019, "step": 45641 }, { "epoch": 8.69, "learning_rate": 3.9225929456625357e-07, "loss": 0.6009, "step": 45642 }, { "epoch": 8.69, "learning_rate": 3.922020972354623e-07, "loss": 1.2325, "step": 45643 }, { "epoch": 8.69, "learning_rate": 3.9214489990467116e-07, "loss": 2.5134, "step": 45644 }, { "epoch": 8.69, "learning_rate": 3.920877025738799e-07, "loss": 1.1027, "step": 45645 }, { "epoch": 8.69, "learning_rate": 3.920305052430887e-07, "loss": 0.3017, "step": 45646 }, { "epoch": 8.69, "learning_rate": 3.9197330791229743e-07, "loss": 0.7618, "step": 45647 }, { "epoch": 8.69, "learning_rate": 3.9191611058150617e-07, "loss": 0.5485, "step": 45648 }, { "epoch": 8.7, "learning_rate": 3.9185891325071497e-07, "loss": 0.8334, "step": 45649 }, { "epoch": 8.7, "learning_rate": 3.918017159199237e-07, "loss": 0.8649, "step": 45650 }, { "epoch": 8.7, "learning_rate": 3.9174451858913255e-07, "loss": 1.5495, "step": 45651 }, { "epoch": 8.7, "learning_rate": 3.916873212583413e-07, "loss": 2.1934, "step": 45652 }, { "epoch": 8.7, "learning_rate": 3.916301239275501e-07, "loss": 1.7933, "step": 45653 }, { "epoch": 8.7, "learning_rate": 3.9157292659675883e-07, "loss": 2.3315, "step": 45654 }, { "epoch": 8.7, "learning_rate": 3.9151572926596757e-07, "loss": 2.8741, "step": 45655 }, { "epoch": 8.7, "learning_rate": 3.9145853193517636e-07, "loss": 1.8162, "step": 45656 }, { "epoch": 8.7, "learning_rate": 3.914013346043851e-07, "loss": 1.0001, "step": 45657 }, { "epoch": 8.7, "learning_rate": 3.9134413727359395e-07, "loss": 0.9081, "step": 45658 }, { "epoch": 8.7, "learning_rate": 3.912869399428027e-07, "loss": 0.8776, "step": 45659 }, { "epoch": 8.7, "learning_rate": 3.9122974261201143e-07, "loss": 0.7919, "step": 45660 }, { "epoch": 8.7, "learning_rate": 3.9117254528122023e-07, "loss": 1.1471, "step": 45661 }, { "epoch": 8.7, "learning_rate": 3.9111534795042897e-07, "loss": 1.6424, "step": 45662 }, { "epoch": 8.7, "learning_rate": 3.9105815061963776e-07, "loss": 1.7561, "step": 45663 }, { "epoch": 8.7, "learning_rate": 3.910009532888465e-07, "loss": 1.7762, "step": 45664 }, { "epoch": 8.7, "learning_rate": 3.9094375595805524e-07, "loss": 0.532, "step": 45665 }, { "epoch": 8.7, "learning_rate": 3.908865586272641e-07, "loss": 2.0737, "step": 45666 }, { "epoch": 8.7, "learning_rate": 3.9082936129647283e-07, "loss": 1.0847, "step": 45667 }, { "epoch": 8.7, "learning_rate": 3.907721639656816e-07, "loss": 1.2257, "step": 45668 }, { "epoch": 8.7, "learning_rate": 3.9071496663489037e-07, "loss": 0.6476, "step": 45669 }, { "epoch": 8.7, "learning_rate": 3.906577693040992e-07, "loss": 2.5525, "step": 45670 }, { "epoch": 8.7, "learning_rate": 3.906005719733079e-07, "loss": 0.8702, "step": 45671 }, { "epoch": 8.7, "learning_rate": 3.9054337464251664e-07, "loss": 1.1451, "step": 45672 }, { "epoch": 8.7, "learning_rate": 3.904861773117255e-07, "loss": 0.903, "step": 45673 }, { "epoch": 8.7, "learning_rate": 3.9042897998093423e-07, "loss": 0.6405, "step": 45674 }, { "epoch": 8.7, "learning_rate": 3.90371782650143e-07, "loss": 0.4505, "step": 45675 }, { "epoch": 8.7, "learning_rate": 3.9031458531935176e-07, "loss": 1.8377, "step": 45676 }, { "epoch": 8.7, "learning_rate": 3.902573879885605e-07, "loss": 1.1821, "step": 45677 }, { "epoch": 8.7, "learning_rate": 3.902001906577693e-07, "loss": 3.0742, "step": 45678 }, { "epoch": 8.7, "learning_rate": 3.9014299332697804e-07, "loss": 2.2574, "step": 45679 }, { "epoch": 8.7, "learning_rate": 3.900857959961869e-07, "loss": 1.2238, "step": 45680 }, { "epoch": 8.7, "learning_rate": 3.9002859866539563e-07, "loss": 1.2366, "step": 45681 }, { "epoch": 8.7, "learning_rate": 3.899714013346044e-07, "loss": 1.689, "step": 45682 }, { "epoch": 8.7, "learning_rate": 3.8991420400381316e-07, "loss": 1.6125, "step": 45683 }, { "epoch": 8.7, "learning_rate": 3.898570066730219e-07, "loss": 1.4778, "step": 45684 }, { "epoch": 8.7, "learning_rate": 3.897998093422307e-07, "loss": 2.6845, "step": 45685 }, { "epoch": 8.7, "learning_rate": 3.8974261201143944e-07, "loss": 2.1657, "step": 45686 }, { "epoch": 8.7, "learning_rate": 3.896854146806483e-07, "loss": 1.7889, "step": 45687 }, { "epoch": 8.7, "learning_rate": 3.89628217349857e-07, "loss": 1.6647, "step": 45688 }, { "epoch": 8.7, "learning_rate": 3.8957102001906577e-07, "loss": 1.7019, "step": 45689 }, { "epoch": 8.7, "learning_rate": 3.8951382268827456e-07, "loss": 0.8726, "step": 45690 }, { "epoch": 8.7, "learning_rate": 3.894566253574833e-07, "loss": 1.2593, "step": 45691 }, { "epoch": 8.7, "learning_rate": 3.8939942802669215e-07, "loss": 1.3722, "step": 45692 }, { "epoch": 8.7, "learning_rate": 3.8934223069590084e-07, "loss": 2.0817, "step": 45693 }, { "epoch": 8.7, "learning_rate": 3.892850333651097e-07, "loss": 2.4562, "step": 45694 }, { "epoch": 8.7, "learning_rate": 3.892278360343184e-07, "loss": 2.4971, "step": 45695 }, { "epoch": 8.7, "learning_rate": 3.8917063870352716e-07, "loss": 2.6998, "step": 45696 }, { "epoch": 8.7, "learning_rate": 3.8911344137273596e-07, "loss": 1.8496, "step": 45697 }, { "epoch": 8.7, "learning_rate": 3.890562440419447e-07, "loss": 2.8396, "step": 45698 }, { "epoch": 8.7, "learning_rate": 3.8899904671115355e-07, "loss": 0.2288, "step": 45699 }, { "epoch": 8.7, "learning_rate": 3.8894184938036223e-07, "loss": 2.5713, "step": 45700 }, { "epoch": 8.7, "learning_rate": 3.88884652049571e-07, "loss": 1.3042, "step": 45701 }, { "epoch": 8.71, "learning_rate": 3.888274547187798e-07, "loss": 0.7051, "step": 45702 }, { "epoch": 8.71, "learning_rate": 3.8877025738798856e-07, "loss": 2.279, "step": 45703 }, { "epoch": 8.71, "learning_rate": 3.8871306005719736e-07, "loss": 2.4122, "step": 45704 }, { "epoch": 8.71, "learning_rate": 3.886558627264061e-07, "loss": 1.7954, "step": 45705 }, { "epoch": 8.71, "learning_rate": 3.8859866539561494e-07, "loss": 2.1926, "step": 45706 }, { "epoch": 8.71, "learning_rate": 3.8854146806482363e-07, "loss": 1.2381, "step": 45707 }, { "epoch": 8.71, "learning_rate": 3.8848427073403237e-07, "loss": 1.3435, "step": 45708 }, { "epoch": 8.71, "learning_rate": 3.884270734032412e-07, "loss": 0.9206, "step": 45709 }, { "epoch": 8.71, "learning_rate": 3.8836987607244996e-07, "loss": 0.3726, "step": 45710 }, { "epoch": 8.71, "learning_rate": 3.8831267874165875e-07, "loss": 1.2525, "step": 45711 }, { "epoch": 8.71, "learning_rate": 3.882554814108675e-07, "loss": 1.5058, "step": 45712 }, { "epoch": 8.71, "learning_rate": 3.8819828408007624e-07, "loss": 1.3716, "step": 45713 }, { "epoch": 8.71, "learning_rate": 3.881410867492851e-07, "loss": 1.4324, "step": 45714 }, { "epoch": 8.71, "learning_rate": 3.8808388941849377e-07, "loss": 2.1548, "step": 45715 }, { "epoch": 8.71, "learning_rate": 3.880266920877026e-07, "loss": 1.848, "step": 45716 }, { "epoch": 8.71, "learning_rate": 3.8796949475691136e-07, "loss": 1.09, "step": 45717 }, { "epoch": 8.71, "learning_rate": 3.879122974261201e-07, "loss": 2.3883, "step": 45718 }, { "epoch": 8.71, "learning_rate": 3.878551000953289e-07, "loss": 1.249, "step": 45719 }, { "epoch": 8.71, "learning_rate": 3.8779790276453763e-07, "loss": 1.3572, "step": 45720 }, { "epoch": 8.71, "learning_rate": 3.877407054337465e-07, "loss": 1.4284, "step": 45721 }, { "epoch": 8.71, "learning_rate": 3.8768350810295517e-07, "loss": 0.872, "step": 45722 }, { "epoch": 8.71, "learning_rate": 3.87626310772164e-07, "loss": 0.9128, "step": 45723 }, { "epoch": 8.71, "learning_rate": 3.8756911344137276e-07, "loss": 2.4882, "step": 45724 }, { "epoch": 8.71, "learning_rate": 3.875119161105815e-07, "loss": 0.9864, "step": 45725 }, { "epoch": 8.71, "learning_rate": 3.874547187797903e-07, "loss": 2.104, "step": 45726 }, { "epoch": 8.71, "learning_rate": 3.8739752144899903e-07, "loss": 2.5375, "step": 45727 }, { "epoch": 8.71, "learning_rate": 3.873403241182079e-07, "loss": 0.6522, "step": 45728 }, { "epoch": 8.71, "learning_rate": 3.8728312678741657e-07, "loss": 0.2234, "step": 45729 }, { "epoch": 8.71, "learning_rate": 3.872259294566253e-07, "loss": 1.7713, "step": 45730 }, { "epoch": 8.71, "learning_rate": 3.8716873212583415e-07, "loss": 1.0742, "step": 45731 }, { "epoch": 8.71, "learning_rate": 3.871115347950429e-07, "loss": 0.363, "step": 45732 }, { "epoch": 8.71, "learning_rate": 3.870543374642517e-07, "loss": 0.8937, "step": 45733 }, { "epoch": 8.71, "learning_rate": 3.8699714013346043e-07, "loss": 2.123, "step": 45734 }, { "epoch": 8.71, "learning_rate": 3.869399428026693e-07, "loss": 1.2892, "step": 45735 }, { "epoch": 8.71, "learning_rate": 3.86882745471878e-07, "loss": 1.2372, "step": 45736 }, { "epoch": 8.71, "learning_rate": 3.868255481410867e-07, "loss": 1.7274, "step": 45737 }, { "epoch": 8.71, "learning_rate": 3.8676835081029555e-07, "loss": 3.3777, "step": 45738 }, { "epoch": 8.71, "learning_rate": 3.867111534795043e-07, "loss": 0.3354, "step": 45739 }, { "epoch": 8.71, "learning_rate": 3.866539561487131e-07, "loss": 0.492, "step": 45740 }, { "epoch": 8.71, "learning_rate": 3.8659675881792183e-07, "loss": 1.7879, "step": 45741 }, { "epoch": 8.71, "learning_rate": 3.8653956148713057e-07, "loss": 2.1202, "step": 45742 }, { "epoch": 8.71, "learning_rate": 3.864823641563394e-07, "loss": 2.6441, "step": 45743 }, { "epoch": 8.71, "learning_rate": 3.864251668255481e-07, "loss": 2.6501, "step": 45744 }, { "epoch": 8.71, "learning_rate": 3.8636796949475695e-07, "loss": 2.1994, "step": 45745 }, { "epoch": 8.71, "learning_rate": 3.863107721639657e-07, "loss": 1.3928, "step": 45746 }, { "epoch": 8.71, "learning_rate": 3.862535748331745e-07, "loss": 1.2435, "step": 45747 }, { "epoch": 8.71, "learning_rate": 3.861963775023832e-07, "loss": 0.7452, "step": 45748 }, { "epoch": 8.71, "learning_rate": 3.8613918017159197e-07, "loss": 2.4022, "step": 45749 }, { "epoch": 8.71, "learning_rate": 3.860819828408008e-07, "loss": 1.1244, "step": 45750 }, { "epoch": 8.71, "learning_rate": 3.860247855100095e-07, "loss": 0.8859, "step": 45751 }, { "epoch": 8.71, "learning_rate": 3.8596758817921835e-07, "loss": 1.0611, "step": 45752 }, { "epoch": 8.71, "learning_rate": 3.859103908484271e-07, "loss": 2.3767, "step": 45753 }, { "epoch": 8.72, "learning_rate": 3.8585319351763583e-07, "loss": 2.0107, "step": 45754 }, { "epoch": 8.72, "learning_rate": 3.857959961868446e-07, "loss": 1.1861, "step": 45755 }, { "epoch": 8.72, "learning_rate": 3.8573879885605336e-07, "loss": 1.3676, "step": 45756 }, { "epoch": 8.72, "learning_rate": 3.856816015252622e-07, "loss": 0.7255, "step": 45757 }, { "epoch": 8.72, "learning_rate": 3.856244041944709e-07, "loss": 0.7491, "step": 45758 }, { "epoch": 8.72, "learning_rate": 3.8556720686367975e-07, "loss": 0.7491, "step": 45759 }, { "epoch": 8.72, "learning_rate": 3.855100095328885e-07, "loss": 1.9682, "step": 45760 }, { "epoch": 8.72, "learning_rate": 3.8545281220209723e-07, "loss": 1.5673, "step": 45761 }, { "epoch": 8.72, "learning_rate": 3.85395614871306e-07, "loss": 1.8295, "step": 45762 }, { "epoch": 8.72, "learning_rate": 3.8533841754051476e-07, "loss": 1.4692, "step": 45763 }, { "epoch": 8.72, "learning_rate": 3.852812202097236e-07, "loss": 1.4959, "step": 45764 }, { "epoch": 8.72, "learning_rate": 3.8522402287893235e-07, "loss": 1.5824, "step": 45765 }, { "epoch": 8.72, "learning_rate": 3.8516682554814104e-07, "loss": 2.0844, "step": 45766 }, { "epoch": 8.72, "learning_rate": 3.851096282173499e-07, "loss": 0.6029, "step": 45767 }, { "epoch": 8.72, "learning_rate": 3.850524308865586e-07, "loss": 3.3207, "step": 45768 }, { "epoch": 8.72, "learning_rate": 3.849952335557674e-07, "loss": 2.0681, "step": 45769 }, { "epoch": 8.72, "learning_rate": 3.8493803622497616e-07, "loss": 1.8765, "step": 45770 }, { "epoch": 8.72, "learning_rate": 3.84880838894185e-07, "loss": 1.0361, "step": 45771 }, { "epoch": 8.72, "learning_rate": 3.8482364156339375e-07, "loss": 1.2981, "step": 45772 }, { "epoch": 8.72, "learning_rate": 3.8476644423260244e-07, "loss": 1.4867, "step": 45773 }, { "epoch": 8.72, "learning_rate": 3.847092469018113e-07, "loss": 1.1398, "step": 45774 }, { "epoch": 8.72, "learning_rate": 3.8465204957102e-07, "loss": 1.4338, "step": 45775 }, { "epoch": 8.72, "learning_rate": 3.845948522402288e-07, "loss": 2.1231, "step": 45776 }, { "epoch": 8.72, "learning_rate": 3.8453765490943756e-07, "loss": 1.0122, "step": 45777 }, { "epoch": 8.72, "learning_rate": 3.844804575786463e-07, "loss": 1.3138, "step": 45778 }, { "epoch": 8.72, "learning_rate": 3.8442326024785515e-07, "loss": 2.242, "step": 45779 }, { "epoch": 8.72, "learning_rate": 3.8436606291706383e-07, "loss": 0.8302, "step": 45780 }, { "epoch": 8.72, "learning_rate": 3.843088655862727e-07, "loss": 2.1463, "step": 45781 }, { "epoch": 8.72, "learning_rate": 3.842516682554814e-07, "loss": 1.6893, "step": 45782 }, { "epoch": 8.72, "learning_rate": 3.8419447092469016e-07, "loss": 1.2775, "step": 45783 }, { "epoch": 8.72, "learning_rate": 3.8413727359389896e-07, "loss": 0.9565, "step": 45784 }, { "epoch": 8.72, "learning_rate": 3.840800762631077e-07, "loss": 1.8869, "step": 45785 }, { "epoch": 8.72, "learning_rate": 3.8402287893231654e-07, "loss": 0.5569, "step": 45786 }, { "epoch": 8.72, "learning_rate": 3.839656816015253e-07, "loss": 2.3387, "step": 45787 }, { "epoch": 8.72, "learning_rate": 3.839084842707341e-07, "loss": 2.5786, "step": 45788 }, { "epoch": 8.72, "learning_rate": 3.838512869399428e-07, "loss": 2.0031, "step": 45789 }, { "epoch": 8.72, "learning_rate": 3.8379408960915156e-07, "loss": 1.8942, "step": 45790 }, { "epoch": 8.72, "learning_rate": 3.8373689227836035e-07, "loss": 0.9777, "step": 45791 }, { "epoch": 8.72, "learning_rate": 3.836796949475691e-07, "loss": 0.9885, "step": 45792 }, { "epoch": 8.72, "learning_rate": 3.8362249761677794e-07, "loss": 1.2113, "step": 45793 }, { "epoch": 8.72, "learning_rate": 3.835653002859867e-07, "loss": 1.1817, "step": 45794 }, { "epoch": 8.72, "learning_rate": 3.8350810295519537e-07, "loss": 1.2868, "step": 45795 }, { "epoch": 8.72, "learning_rate": 3.834509056244042e-07, "loss": 0.5862, "step": 45796 }, { "epoch": 8.72, "learning_rate": 3.8339370829361296e-07, "loss": 1.192, "step": 45797 }, { "epoch": 8.72, "learning_rate": 3.8333651096282175e-07, "loss": 2.7458, "step": 45798 }, { "epoch": 8.72, "learning_rate": 3.832793136320305e-07, "loss": 2.3787, "step": 45799 }, { "epoch": 8.72, "learning_rate": 3.8322211630123934e-07, "loss": 1.7061, "step": 45800 }, { "epoch": 8.72, "learning_rate": 3.831649189704481e-07, "loss": 1.2758, "step": 45801 }, { "epoch": 8.72, "learning_rate": 3.8310772163965677e-07, "loss": 0.7249, "step": 45802 }, { "epoch": 8.72, "learning_rate": 3.830505243088656e-07, "loss": 0.9553, "step": 45803 }, { "epoch": 8.72, "learning_rate": 3.8299332697807436e-07, "loss": 1.1621, "step": 45804 }, { "epoch": 8.72, "learning_rate": 3.8293612964728315e-07, "loss": 0.7987, "step": 45805 }, { "epoch": 8.72, "learning_rate": 3.828789323164919e-07, "loss": 1.8962, "step": 45806 }, { "epoch": 8.73, "learning_rate": 3.8282173498570063e-07, "loss": 2.4209, "step": 45807 }, { "epoch": 8.73, "learning_rate": 3.827645376549095e-07, "loss": 2.0766, "step": 45808 }, { "epoch": 8.73, "learning_rate": 3.827073403241182e-07, "loss": 1.3903, "step": 45809 }, { "epoch": 8.73, "learning_rate": 3.82650142993327e-07, "loss": 0.9984, "step": 45810 }, { "epoch": 8.73, "learning_rate": 3.8259294566253575e-07, "loss": 2.5514, "step": 45811 }, { "epoch": 8.73, "learning_rate": 3.8253574833174455e-07, "loss": 1.8597, "step": 45812 }, { "epoch": 8.73, "learning_rate": 3.824785510009533e-07, "loss": 1.4966, "step": 45813 }, { "epoch": 8.73, "learning_rate": 3.8242135367016203e-07, "loss": 1.9451, "step": 45814 }, { "epoch": 8.73, "learning_rate": 3.823641563393709e-07, "loss": 1.1795, "step": 45815 }, { "epoch": 8.73, "learning_rate": 3.823069590085796e-07, "loss": 2.2379, "step": 45816 }, { "epoch": 8.73, "learning_rate": 3.822497616777884e-07, "loss": 2.2073, "step": 45817 }, { "epoch": 8.73, "learning_rate": 3.8219256434699715e-07, "loss": 1.8055, "step": 45818 }, { "epoch": 8.73, "learning_rate": 3.821353670162059e-07, "loss": 1.6975, "step": 45819 }, { "epoch": 8.73, "learning_rate": 3.820781696854147e-07, "loss": 2.2583, "step": 45820 }, { "epoch": 8.73, "learning_rate": 3.8202097235462343e-07, "loss": 1.8473, "step": 45821 }, { "epoch": 8.73, "learning_rate": 3.819637750238323e-07, "loss": 3.6739, "step": 45822 }, { "epoch": 8.73, "learning_rate": 3.81906577693041e-07, "loss": 1.8793, "step": 45823 }, { "epoch": 8.73, "learning_rate": 3.818493803622498e-07, "loss": 2.0823, "step": 45824 }, { "epoch": 8.73, "learning_rate": 3.8179218303145855e-07, "loss": 0.3801, "step": 45825 }, { "epoch": 8.73, "learning_rate": 3.817349857006673e-07, "loss": 1.0075, "step": 45826 }, { "epoch": 8.73, "learning_rate": 3.816777883698761e-07, "loss": 0.3143, "step": 45827 }, { "epoch": 8.73, "learning_rate": 3.816205910390848e-07, "loss": 1.4852, "step": 45828 }, { "epoch": 8.73, "learning_rate": 3.8156339370829367e-07, "loss": 1.6925, "step": 45829 }, { "epoch": 8.73, "learning_rate": 3.815061963775024e-07, "loss": 1.8442, "step": 45830 }, { "epoch": 8.73, "learning_rate": 3.8144899904671115e-07, "loss": 0.6844, "step": 45831 }, { "epoch": 8.73, "learning_rate": 3.8139180171591995e-07, "loss": 1.8984, "step": 45832 }, { "epoch": 8.73, "learning_rate": 3.813346043851287e-07, "loss": 0.2404, "step": 45833 }, { "epoch": 8.73, "learning_rate": 3.812774070543375e-07, "loss": 1.6275, "step": 45834 }, { "epoch": 8.73, "learning_rate": 3.812202097235462e-07, "loss": 1.0898, "step": 45835 }, { "epoch": 8.73, "learning_rate": 3.8116301239275507e-07, "loss": 1.0396, "step": 45836 }, { "epoch": 8.73, "learning_rate": 3.811058150619638e-07, "loss": 1.5953, "step": 45837 }, { "epoch": 8.73, "learning_rate": 3.8104861773117255e-07, "loss": 1.25, "step": 45838 }, { "epoch": 8.73, "learning_rate": 3.8099142040038135e-07, "loss": 2.8293, "step": 45839 }, { "epoch": 8.73, "learning_rate": 3.809342230695901e-07, "loss": 0.7551, "step": 45840 }, { "epoch": 8.73, "learning_rate": 3.808770257387989e-07, "loss": 2.7962, "step": 45841 }, { "epoch": 8.73, "learning_rate": 3.808198284080076e-07, "loss": 1.864, "step": 45842 }, { "epoch": 8.73, "learning_rate": 3.8076263107721636e-07, "loss": 1.7674, "step": 45843 }, { "epoch": 8.73, "learning_rate": 3.807054337464252e-07, "loss": 0.8103, "step": 45844 }, { "epoch": 8.73, "learning_rate": 3.8064823641563395e-07, "loss": 0.8441, "step": 45845 }, { "epoch": 8.73, "learning_rate": 3.8059103908484274e-07, "loss": 2.238, "step": 45846 }, { "epoch": 8.73, "learning_rate": 3.805338417540515e-07, "loss": 0.4595, "step": 45847 }, { "epoch": 8.73, "learning_rate": 3.804766444232602e-07, "loss": 1.7078, "step": 45848 }, { "epoch": 8.73, "learning_rate": 3.80419447092469e-07, "loss": 1.5602, "step": 45849 }, { "epoch": 8.73, "learning_rate": 3.8036224976167776e-07, "loss": 2.9518, "step": 45850 }, { "epoch": 8.73, "learning_rate": 3.803050524308866e-07, "loss": 2.193, "step": 45851 }, { "epoch": 8.73, "learning_rate": 3.8024785510009535e-07, "loss": 1.4975, "step": 45852 }, { "epoch": 8.73, "learning_rate": 3.8019065776930414e-07, "loss": 1.3969, "step": 45853 }, { "epoch": 8.73, "learning_rate": 3.801334604385129e-07, "loss": 0.9202, "step": 45854 }, { "epoch": 8.73, "learning_rate": 3.800762631077216e-07, "loss": 1.6168, "step": 45855 }, { "epoch": 8.73, "learning_rate": 3.800190657769304e-07, "loss": 1.0119, "step": 45856 }, { "epoch": 8.73, "learning_rate": 3.7996186844613916e-07, "loss": 1.3915, "step": 45857 }, { "epoch": 8.73, "learning_rate": 3.79904671115348e-07, "loss": 0.7981, "step": 45858 }, { "epoch": 8.74, "learning_rate": 3.7984747378455675e-07, "loss": 2.1696, "step": 45859 }, { "epoch": 8.74, "learning_rate": 3.797902764537655e-07, "loss": 2.9866, "step": 45860 }, { "epoch": 8.74, "learning_rate": 3.797330791229743e-07, "loss": 1.7976, "step": 45861 }, { "epoch": 8.74, "learning_rate": 3.79675881792183e-07, "loss": 0.5933, "step": 45862 }, { "epoch": 8.74, "learning_rate": 3.796186844613918e-07, "loss": 1.5081, "step": 45863 }, { "epoch": 8.74, "learning_rate": 3.7956148713060056e-07, "loss": 1.249, "step": 45864 }, { "epoch": 8.74, "learning_rate": 3.795042897998094e-07, "loss": 2.4794, "step": 45865 }, { "epoch": 8.74, "learning_rate": 3.7944709246901814e-07, "loss": 2.0304, "step": 45866 }, { "epoch": 8.74, "learning_rate": 3.793898951382269e-07, "loss": 1.2085, "step": 45867 }, { "epoch": 8.74, "learning_rate": 3.793326978074357e-07, "loss": 1.23, "step": 45868 }, { "epoch": 8.74, "learning_rate": 3.792755004766444e-07, "loss": 1.3781, "step": 45869 }, { "epoch": 8.74, "learning_rate": 3.792183031458532e-07, "loss": 2.1331, "step": 45870 }, { "epoch": 8.74, "learning_rate": 3.7916110581506195e-07, "loss": 2.6574, "step": 45871 }, { "epoch": 8.74, "learning_rate": 3.791039084842707e-07, "loss": 0.3333, "step": 45872 }, { "epoch": 8.74, "learning_rate": 3.7904671115347954e-07, "loss": 2.2268, "step": 45873 }, { "epoch": 8.74, "learning_rate": 3.789895138226883e-07, "loss": 0.8998, "step": 45874 }, { "epoch": 8.74, "learning_rate": 3.789323164918971e-07, "loss": 2.0317, "step": 45875 }, { "epoch": 8.74, "learning_rate": 3.788751191611058e-07, "loss": 0.9787, "step": 45876 }, { "epoch": 8.74, "learning_rate": 3.788179218303146e-07, "loss": 0.8461, "step": 45877 }, { "epoch": 8.74, "learning_rate": 3.7876072449952335e-07, "loss": 0.8781, "step": 45878 }, { "epoch": 8.74, "learning_rate": 3.787035271687321e-07, "loss": 1.8012, "step": 45879 }, { "epoch": 8.74, "learning_rate": 3.7864632983794094e-07, "loss": 0.9385, "step": 45880 }, { "epoch": 8.74, "learning_rate": 3.785891325071497e-07, "loss": 1.3536, "step": 45881 }, { "epoch": 8.74, "learning_rate": 3.785319351763585e-07, "loss": 3.4556, "step": 45882 }, { "epoch": 8.74, "learning_rate": 3.784747378455672e-07, "loss": 1.8352, "step": 45883 }, { "epoch": 8.74, "learning_rate": 3.7841754051477596e-07, "loss": 1.5178, "step": 45884 }, { "epoch": 8.74, "learning_rate": 3.7836034318398475e-07, "loss": 1.2836, "step": 45885 }, { "epoch": 8.74, "learning_rate": 3.783031458531935e-07, "loss": 0.5977, "step": 45886 }, { "epoch": 8.74, "learning_rate": 3.7824594852240234e-07, "loss": 0.5248, "step": 45887 }, { "epoch": 8.74, "learning_rate": 3.781887511916111e-07, "loss": 0.9202, "step": 45888 }, { "epoch": 8.74, "learning_rate": 3.7813155386081987e-07, "loss": 2.5317, "step": 45889 }, { "epoch": 8.74, "learning_rate": 3.780743565300286e-07, "loss": 1.6636, "step": 45890 }, { "epoch": 8.74, "learning_rate": 3.7801715919923735e-07, "loss": 1.3274, "step": 45891 }, { "epoch": 8.74, "learning_rate": 3.7795996186844615e-07, "loss": 1.9945, "step": 45892 }, { "epoch": 8.74, "learning_rate": 3.779027645376549e-07, "loss": 0.4846, "step": 45893 }, { "epoch": 8.74, "learning_rate": 3.7784556720686373e-07, "loss": 1.8443, "step": 45894 }, { "epoch": 8.74, "learning_rate": 3.777883698760725e-07, "loss": 2.6379, "step": 45895 }, { "epoch": 8.74, "learning_rate": 3.777311725452812e-07, "loss": 2.3531, "step": 45896 }, { "epoch": 8.74, "learning_rate": 3.7767397521449e-07, "loss": 0.7583, "step": 45897 }, { "epoch": 8.74, "learning_rate": 3.7761677788369875e-07, "loss": 1.3965, "step": 45898 }, { "epoch": 8.74, "learning_rate": 3.7755958055290754e-07, "loss": 2.2202, "step": 45899 }, { "epoch": 8.74, "learning_rate": 3.775023832221163e-07, "loss": 0.9678, "step": 45900 }, { "epoch": 8.74, "learning_rate": 3.7744518589132503e-07, "loss": 2.1639, "step": 45901 }, { "epoch": 8.74, "learning_rate": 3.7738798856053387e-07, "loss": 1.483, "step": 45902 }, { "epoch": 8.74, "learning_rate": 3.773307912297426e-07, "loss": 1.075, "step": 45903 }, { "epoch": 8.74, "learning_rate": 3.772735938989514e-07, "loss": 0.9126, "step": 45904 }, { "epoch": 8.74, "learning_rate": 3.7721639656816015e-07, "loss": 1.9887, "step": 45905 }, { "epoch": 8.74, "learning_rate": 3.7715919923736894e-07, "loss": 2.9803, "step": 45906 }, { "epoch": 8.74, "learning_rate": 3.771020019065777e-07, "loss": 2.1536, "step": 45907 }, { "epoch": 8.74, "learning_rate": 3.770448045757864e-07, "loss": 1.7756, "step": 45908 }, { "epoch": 8.74, "learning_rate": 3.7698760724499527e-07, "loss": 0.7573, "step": 45909 }, { "epoch": 8.74, "learning_rate": 3.76930409914204e-07, "loss": 2.4482, "step": 45910 }, { "epoch": 8.74, "learning_rate": 3.768732125834128e-07, "loss": 1.2389, "step": 45911 }, { "epoch": 8.75, "learning_rate": 3.7681601525262155e-07, "loss": 0.9181, "step": 45912 }, { "epoch": 8.75, "learning_rate": 3.767588179218303e-07, "loss": 0.8065, "step": 45913 }, { "epoch": 8.75, "learning_rate": 3.767016205910391e-07, "loss": 1.4292, "step": 45914 }, { "epoch": 8.75, "learning_rate": 3.766444232602478e-07, "loss": 2.1393, "step": 45915 }, { "epoch": 8.75, "learning_rate": 3.7658722592945667e-07, "loss": 2.0386, "step": 45916 }, { "epoch": 8.75, "learning_rate": 3.765300285986654e-07, "loss": 1.4236, "step": 45917 }, { "epoch": 8.75, "learning_rate": 3.764728312678742e-07, "loss": 2.1444, "step": 45918 }, { "epoch": 8.75, "learning_rate": 3.7641563393708294e-07, "loss": 1.2772, "step": 45919 }, { "epoch": 8.75, "learning_rate": 3.763584366062917e-07, "loss": 0.8402, "step": 45920 }, { "epoch": 8.75, "learning_rate": 3.763012392755005e-07, "loss": 0.7649, "step": 45921 }, { "epoch": 8.75, "learning_rate": 3.762440419447092e-07, "loss": 2.1665, "step": 45922 }, { "epoch": 8.75, "learning_rate": 3.7618684461391807e-07, "loss": 1.6204, "step": 45923 }, { "epoch": 8.75, "learning_rate": 3.761296472831268e-07, "loss": 2.5425, "step": 45924 }, { "epoch": 8.75, "learning_rate": 3.7607244995233555e-07, "loss": 3.7484, "step": 45925 }, { "epoch": 8.75, "learning_rate": 3.7601525262154434e-07, "loss": 1.2499, "step": 45926 }, { "epoch": 8.75, "learning_rate": 3.759580552907531e-07, "loss": 1.4872, "step": 45927 }, { "epoch": 8.75, "learning_rate": 3.759008579599619e-07, "loss": 0.7228, "step": 45928 }, { "epoch": 8.75, "learning_rate": 3.758436606291706e-07, "loss": 0.7471, "step": 45929 }, { "epoch": 8.75, "learning_rate": 3.7578646329837947e-07, "loss": 3.0979, "step": 45930 }, { "epoch": 8.75, "learning_rate": 3.757292659675882e-07, "loss": 1.937, "step": 45931 }, { "epoch": 8.75, "learning_rate": 3.7567206863679695e-07, "loss": 1.6972, "step": 45932 }, { "epoch": 8.75, "learning_rate": 3.7561487130600574e-07, "loss": 1.842, "step": 45933 }, { "epoch": 8.75, "learning_rate": 3.755576739752145e-07, "loss": 3.7778, "step": 45934 }, { "epoch": 8.75, "learning_rate": 3.755004766444233e-07, "loss": 1.2829, "step": 45935 }, { "epoch": 8.75, "learning_rate": 3.75443279313632e-07, "loss": 2.2945, "step": 45936 }, { "epoch": 8.75, "learning_rate": 3.7538608198284076e-07, "loss": 3.0479, "step": 45937 }, { "epoch": 8.75, "learning_rate": 3.753288846520496e-07, "loss": 0.9516, "step": 45938 }, { "epoch": 8.75, "learning_rate": 3.7527168732125834e-07, "loss": 1.8794, "step": 45939 }, { "epoch": 8.75, "learning_rate": 3.7521448999046714e-07, "loss": 1.2515, "step": 45940 }, { "epoch": 8.75, "learning_rate": 3.751572926596759e-07, "loss": 2.137, "step": 45941 }, { "epoch": 8.75, "learning_rate": 3.751000953288847e-07, "loss": 1.5903, "step": 45942 }, { "epoch": 8.75, "learning_rate": 3.750428979980934e-07, "loss": 1.5359, "step": 45943 }, { "epoch": 8.75, "learning_rate": 3.749857006673022e-07, "loss": 0.2446, "step": 45944 }, { "epoch": 8.75, "learning_rate": 3.74928503336511e-07, "loss": 1.3143, "step": 45945 }, { "epoch": 8.75, "learning_rate": 3.7487130600571974e-07, "loss": 2.0188, "step": 45946 }, { "epoch": 8.75, "learning_rate": 3.748141086749285e-07, "loss": 2.3014, "step": 45947 }, { "epoch": 8.75, "learning_rate": 3.747569113441373e-07, "loss": 2.206, "step": 45948 }, { "epoch": 8.75, "learning_rate": 3.7469971401334607e-07, "loss": 1.2099, "step": 45949 }, { "epoch": 8.75, "learning_rate": 3.746425166825548e-07, "loss": 0.7023, "step": 45950 }, { "epoch": 8.75, "learning_rate": 3.745853193517636e-07, "loss": 2.0756, "step": 45951 }, { "epoch": 8.75, "learning_rate": 3.7452812202097235e-07, "loss": 2.2671, "step": 45952 }, { "epoch": 8.75, "learning_rate": 3.7447092469018114e-07, "loss": 2.0746, "step": 45953 }, { "epoch": 8.75, "learning_rate": 3.744137273593899e-07, "loss": 1.0353, "step": 45954 }, { "epoch": 8.75, "learning_rate": 3.743565300285987e-07, "loss": 0.6546, "step": 45955 }, { "epoch": 8.75, "learning_rate": 3.7429933269780747e-07, "loss": 2.0049, "step": 45956 }, { "epoch": 8.75, "learning_rate": 3.742421353670162e-07, "loss": 1.83, "step": 45957 }, { "epoch": 8.75, "learning_rate": 3.7418493803622495e-07, "loss": 1.6147, "step": 45958 }, { "epoch": 8.75, "learning_rate": 3.7412774070543374e-07, "loss": 2.5097, "step": 45959 }, { "epoch": 8.75, "learning_rate": 3.7407054337464254e-07, "loss": 0.4922, "step": 45960 }, { "epoch": 8.75, "learning_rate": 3.740133460438513e-07, "loss": 1.6469, "step": 45961 }, { "epoch": 8.75, "learning_rate": 3.7395614871306007e-07, "loss": 1.8601, "step": 45962 }, { "epoch": 8.75, "learning_rate": 3.738989513822688e-07, "loss": 1.8398, "step": 45963 }, { "epoch": 8.76, "learning_rate": 3.738417540514776e-07, "loss": 0.9677, "step": 45964 }, { "epoch": 8.76, "learning_rate": 3.7378455672068635e-07, "loss": 1.9713, "step": 45965 }, { "epoch": 8.76, "learning_rate": 3.7372735938989514e-07, "loss": 2.0577, "step": 45966 }, { "epoch": 8.76, "learning_rate": 3.7367016205910394e-07, "loss": 0.9275, "step": 45967 }, { "epoch": 8.76, "learning_rate": 3.736129647283127e-07, "loss": 1.8283, "step": 45968 }, { "epoch": 8.76, "learning_rate": 3.735557673975214e-07, "loss": 0.5911, "step": 45969 }, { "epoch": 8.76, "learning_rate": 3.734985700667302e-07, "loss": 0.9452, "step": 45970 }, { "epoch": 8.76, "learning_rate": 3.73441372735939e-07, "loss": 2.0776, "step": 45971 }, { "epoch": 8.76, "learning_rate": 3.7338417540514775e-07, "loss": 1.4978, "step": 45972 }, { "epoch": 8.76, "learning_rate": 3.7332697807435654e-07, "loss": 1.8078, "step": 45973 }, { "epoch": 8.76, "learning_rate": 3.7326978074356533e-07, "loss": 1.0849, "step": 45974 }, { "epoch": 8.76, "learning_rate": 3.732125834127741e-07, "loss": 3.0662, "step": 45975 }, { "epoch": 8.76, "learning_rate": 3.731553860819828e-07, "loss": 1.2755, "step": 45976 }, { "epoch": 8.76, "learning_rate": 3.730981887511916e-07, "loss": 1.0555, "step": 45977 }, { "epoch": 8.76, "learning_rate": 3.730409914204004e-07, "loss": 2.9919, "step": 45978 }, { "epoch": 8.76, "learning_rate": 3.7298379408960914e-07, "loss": 0.9607, "step": 45979 }, { "epoch": 8.76, "learning_rate": 3.7292659675881794e-07, "loss": 2.26, "step": 45980 }, { "epoch": 8.76, "learning_rate": 3.728693994280267e-07, "loss": 2.1955, "step": 45981 }, { "epoch": 8.76, "learning_rate": 3.7281220209723547e-07, "loss": 1.3809, "step": 45982 }, { "epoch": 8.76, "learning_rate": 3.727550047664442e-07, "loss": 1.8773, "step": 45983 }, { "epoch": 8.76, "learning_rate": 3.72697807435653e-07, "loss": 1.8647, "step": 45984 }, { "epoch": 8.76, "learning_rate": 3.726406101048618e-07, "loss": 1.6213, "step": 45985 }, { "epoch": 8.76, "learning_rate": 3.725834127740706e-07, "loss": 0.9023, "step": 45986 }, { "epoch": 8.76, "learning_rate": 3.725262154432793e-07, "loss": 1.6234, "step": 45987 }, { "epoch": 8.76, "learning_rate": 3.724690181124881e-07, "loss": 2.6992, "step": 45988 }, { "epoch": 8.76, "learning_rate": 3.7241182078169687e-07, "loss": 2.4303, "step": 45989 }, { "epoch": 8.76, "learning_rate": 3.723546234509056e-07, "loss": 1.5973, "step": 45990 }, { "epoch": 8.76, "learning_rate": 3.722974261201144e-07, "loss": 0.6826, "step": 45991 }, { "epoch": 8.76, "learning_rate": 3.722402287893232e-07, "loss": 2.1676, "step": 45992 }, { "epoch": 8.76, "learning_rate": 3.7218303145853194e-07, "loss": 1.9703, "step": 45993 }, { "epoch": 8.76, "learning_rate": 3.721258341277407e-07, "loss": 1.138, "step": 45994 }, { "epoch": 8.76, "learning_rate": 3.720686367969495e-07, "loss": 2.7879, "step": 45995 }, { "epoch": 8.76, "learning_rate": 3.7201143946615827e-07, "loss": 1.8641, "step": 45996 }, { "epoch": 8.76, "learning_rate": 3.7195424213536706e-07, "loss": 2.8181, "step": 45997 }, { "epoch": 8.76, "learning_rate": 3.718970448045758e-07, "loss": 0.3377, "step": 45998 }, { "epoch": 8.76, "learning_rate": 3.7183984747378454e-07, "loss": 2.426, "step": 45999 }, { "epoch": 8.76, "learning_rate": 3.7178265014299334e-07, "loss": 1.7134, "step": 46000 }, { "epoch": 8.76, "learning_rate": 3.717254528122021e-07, "loss": 0.763, "step": 46001 }, { "epoch": 8.76, "learning_rate": 3.7166825548141087e-07, "loss": 0.9451, "step": 46002 }, { "epoch": 8.76, "learning_rate": 3.7161105815061967e-07, "loss": 0.8399, "step": 46003 }, { "epoch": 8.76, "learning_rate": 3.7155386081982846e-07, "loss": 0.631, "step": 46004 }, { "epoch": 8.76, "learning_rate": 3.7149666348903715e-07, "loss": 0.9322, "step": 46005 }, { "epoch": 8.76, "learning_rate": 3.7143946615824594e-07, "loss": 2.0663, "step": 46006 }, { "epoch": 8.76, "learning_rate": 3.7138226882745474e-07, "loss": 4.0867, "step": 46007 }, { "epoch": 8.76, "learning_rate": 3.7132507149666353e-07, "loss": 1.3273, "step": 46008 }, { "epoch": 8.76, "learning_rate": 3.7126787416587227e-07, "loss": 0.6433, "step": 46009 }, { "epoch": 8.76, "learning_rate": 3.7121067683508106e-07, "loss": 0.5639, "step": 46010 }, { "epoch": 8.76, "learning_rate": 3.711534795042898e-07, "loss": 1.4984, "step": 46011 }, { "epoch": 8.76, "learning_rate": 3.7109628217349855e-07, "loss": 1.8504, "step": 46012 }, { "epoch": 8.76, "learning_rate": 3.7103908484270734e-07, "loss": 1.5014, "step": 46013 }, { "epoch": 8.76, "learning_rate": 3.7098188751191613e-07, "loss": 1.4407, "step": 46014 }, { "epoch": 8.76, "learning_rate": 3.7092469018112493e-07, "loss": 0.3424, "step": 46015 }, { "epoch": 8.76, "learning_rate": 3.7086749285033367e-07, "loss": 2.5424, "step": 46016 }, { "epoch": 8.77, "learning_rate": 3.708102955195424e-07, "loss": 1.0805, "step": 46017 }, { "epoch": 8.77, "learning_rate": 3.707530981887512e-07, "loss": 1.8383, "step": 46018 }, { "epoch": 8.77, "learning_rate": 3.7069590085796e-07, "loss": 0.8467, "step": 46019 }, { "epoch": 8.77, "learning_rate": 3.7063870352716874e-07, "loss": 1.4385, "step": 46020 }, { "epoch": 8.77, "learning_rate": 3.7058150619637753e-07, "loss": 1.4435, "step": 46021 }, { "epoch": 8.77, "learning_rate": 3.7052430886558627e-07, "loss": 2.5829, "step": 46022 }, { "epoch": 8.77, "learning_rate": 3.70467111534795e-07, "loss": 0.7997, "step": 46023 }, { "epoch": 8.77, "learning_rate": 3.704099142040038e-07, "loss": 1.196, "step": 46024 }, { "epoch": 8.77, "learning_rate": 3.703527168732126e-07, "loss": 1.0245, "step": 46025 }, { "epoch": 8.77, "learning_rate": 3.702955195424214e-07, "loss": 1.5177, "step": 46026 }, { "epoch": 8.77, "learning_rate": 3.7023832221163014e-07, "loss": 0.493, "step": 46027 }, { "epoch": 8.77, "learning_rate": 3.701811248808389e-07, "loss": 1.6913, "step": 46028 }, { "epoch": 8.77, "learning_rate": 3.7012392755004767e-07, "loss": 1.7482, "step": 46029 }, { "epoch": 8.77, "learning_rate": 3.7006673021925646e-07, "loss": 0.7944, "step": 46030 }, { "epoch": 8.77, "learning_rate": 3.700095328884652e-07, "loss": 0.601, "step": 46031 }, { "epoch": 8.77, "learning_rate": 3.69952335557674e-07, "loss": 0.8095, "step": 46032 }, { "epoch": 8.77, "learning_rate": 3.698951382268828e-07, "loss": 0.8718, "step": 46033 }, { "epoch": 8.77, "learning_rate": 3.698379408960915e-07, "loss": 1.2017, "step": 46034 }, { "epoch": 8.77, "learning_rate": 3.697807435653003e-07, "loss": 1.7647, "step": 46035 }, { "epoch": 8.77, "learning_rate": 3.6972354623450907e-07, "loss": 1.9171, "step": 46036 }, { "epoch": 8.77, "learning_rate": 3.6966634890371786e-07, "loss": 2.0212, "step": 46037 }, { "epoch": 8.77, "learning_rate": 3.696091515729266e-07, "loss": 2.3437, "step": 46038 }, { "epoch": 8.77, "learning_rate": 3.695519542421354e-07, "loss": 1.9282, "step": 46039 }, { "epoch": 8.77, "learning_rate": 3.6949475691134414e-07, "loss": 1.6574, "step": 46040 }, { "epoch": 8.77, "learning_rate": 3.6943755958055293e-07, "loss": 1.6669, "step": 46041 }, { "epoch": 8.77, "learning_rate": 3.6938036224976167e-07, "loss": 1.5603, "step": 46042 }, { "epoch": 8.77, "learning_rate": 3.6932316491897047e-07, "loss": 1.307, "step": 46043 }, { "epoch": 8.77, "learning_rate": 3.6926596758817926e-07, "loss": 1.7032, "step": 46044 }, { "epoch": 8.77, "learning_rate": 3.69208770257388e-07, "loss": 2.1572, "step": 46045 }, { "epoch": 8.77, "learning_rate": 3.6915157292659674e-07, "loss": 1.3407, "step": 46046 }, { "epoch": 8.77, "learning_rate": 3.6909437559580554e-07, "loss": 1.8021, "step": 46047 }, { "epoch": 8.77, "learning_rate": 3.6903717826501433e-07, "loss": 1.7771, "step": 46048 }, { "epoch": 8.77, "learning_rate": 3.6897998093422307e-07, "loss": 1.4161, "step": 46049 }, { "epoch": 8.77, "learning_rate": 3.6892278360343186e-07, "loss": 1.5434, "step": 46050 }, { "epoch": 8.77, "learning_rate": 3.6886558627264066e-07, "loss": 1.7786, "step": 46051 }, { "epoch": 8.77, "learning_rate": 3.6880838894184935e-07, "loss": 2.3795, "step": 46052 }, { "epoch": 8.77, "learning_rate": 3.6875119161105814e-07, "loss": 0.7562, "step": 46053 }, { "epoch": 8.77, "learning_rate": 3.6869399428026693e-07, "loss": 1.0156, "step": 46054 }, { "epoch": 8.77, "learning_rate": 3.6863679694947573e-07, "loss": 1.3323, "step": 46055 }, { "epoch": 8.77, "learning_rate": 3.6857959961868447e-07, "loss": 1.5673, "step": 46056 }, { "epoch": 8.77, "learning_rate": 3.6852240228789326e-07, "loss": 2.3236, "step": 46057 }, { "epoch": 8.77, "learning_rate": 3.68465204957102e-07, "loss": 1.8777, "step": 46058 }, { "epoch": 8.77, "learning_rate": 3.684080076263108e-07, "loss": 1.0237, "step": 46059 }, { "epoch": 8.77, "learning_rate": 3.6835081029551954e-07, "loss": 0.9509, "step": 46060 }, { "epoch": 8.77, "learning_rate": 3.6829361296472833e-07, "loss": 1.289, "step": 46061 }, { "epoch": 8.77, "learning_rate": 3.682364156339371e-07, "loss": 1.2921, "step": 46062 }, { "epoch": 8.77, "learning_rate": 3.6817921830314587e-07, "loss": 1.5902, "step": 46063 }, { "epoch": 8.77, "learning_rate": 3.681220209723546e-07, "loss": 1.9193, "step": 46064 }, { "epoch": 8.77, "learning_rate": 3.680648236415634e-07, "loss": 1.8881, "step": 46065 }, { "epoch": 8.77, "learning_rate": 3.680076263107722e-07, "loss": 0.9769, "step": 46066 }, { "epoch": 8.77, "learning_rate": 3.6795042897998094e-07, "loss": 1.1126, "step": 46067 }, { "epoch": 8.77, "learning_rate": 3.6789323164918973e-07, "loss": 2.9157, "step": 46068 }, { "epoch": 8.78, "learning_rate": 3.678360343183985e-07, "loss": 1.6722, "step": 46069 }, { "epoch": 8.78, "learning_rate": 3.6777883698760726e-07, "loss": 1.0259, "step": 46070 }, { "epoch": 8.78, "learning_rate": 3.67721639656816e-07, "loss": 1.3859, "step": 46071 }, { "epoch": 8.78, "learning_rate": 3.676644423260248e-07, "loss": 1.3496, "step": 46072 }, { "epoch": 8.78, "learning_rate": 3.676072449952336e-07, "loss": 1.9418, "step": 46073 }, { "epoch": 8.78, "learning_rate": 3.6755004766444233e-07, "loss": 1.2508, "step": 46074 }, { "epoch": 8.78, "learning_rate": 3.6749285033365113e-07, "loss": 1.7904, "step": 46075 }, { "epoch": 8.78, "learning_rate": 3.6743565300285987e-07, "loss": 1.161, "step": 46076 }, { "epoch": 8.78, "learning_rate": 3.6737845567206866e-07, "loss": 2.944, "step": 46077 }, { "epoch": 8.78, "learning_rate": 3.673212583412774e-07, "loss": 1.0388, "step": 46078 }, { "epoch": 8.78, "learning_rate": 3.672640610104862e-07, "loss": 2.2483, "step": 46079 }, { "epoch": 8.78, "learning_rate": 3.67206863679695e-07, "loss": 2.0642, "step": 46080 }, { "epoch": 8.78, "learning_rate": 3.6714966634890373e-07, "loss": 1.676, "step": 46081 }, { "epoch": 8.78, "learning_rate": 3.6709246901811247e-07, "loss": 1.6743, "step": 46082 }, { "epoch": 8.78, "learning_rate": 3.6703527168732127e-07, "loss": 2.7665, "step": 46083 }, { "epoch": 8.78, "learning_rate": 3.6697807435653006e-07, "loss": 0.746, "step": 46084 }, { "epoch": 8.78, "learning_rate": 3.669208770257388e-07, "loss": 1.3195, "step": 46085 }, { "epoch": 8.78, "learning_rate": 3.668636796949476e-07, "loss": 0.6942, "step": 46086 }, { "epoch": 8.78, "learning_rate": 3.6680648236415634e-07, "loss": 2.8412, "step": 46087 }, { "epoch": 8.78, "learning_rate": 3.6674928503336513e-07, "loss": 1.9422, "step": 46088 }, { "epoch": 8.78, "learning_rate": 3.6669208770257387e-07, "loss": 1.5306, "step": 46089 }, { "epoch": 8.78, "learning_rate": 3.6663489037178266e-07, "loss": 1.4872, "step": 46090 }, { "epoch": 8.78, "learning_rate": 3.6657769304099146e-07, "loss": 2.4112, "step": 46091 }, { "epoch": 8.78, "learning_rate": 3.665204957102002e-07, "loss": 0.8509, "step": 46092 }, { "epoch": 8.78, "learning_rate": 3.6646329837940894e-07, "loss": 1.8841, "step": 46093 }, { "epoch": 8.78, "learning_rate": 3.6640610104861773e-07, "loss": 1.6125, "step": 46094 }, { "epoch": 8.78, "learning_rate": 3.6634890371782653e-07, "loss": 1.1177, "step": 46095 }, { "epoch": 8.78, "learning_rate": 3.6629170638703527e-07, "loss": 1.907, "step": 46096 }, { "epoch": 8.78, "learning_rate": 3.6623450905624406e-07, "loss": 1.4404, "step": 46097 }, { "epoch": 8.78, "learning_rate": 3.6617731172545286e-07, "loss": 1.4176, "step": 46098 }, { "epoch": 8.78, "learning_rate": 3.661201143946616e-07, "loss": 1.34, "step": 46099 }, { "epoch": 8.78, "learning_rate": 3.6606291706387034e-07, "loss": 0.9597, "step": 46100 }, { "epoch": 8.78, "learning_rate": 3.6600571973307913e-07, "loss": 2.1218, "step": 46101 }, { "epoch": 8.78, "learning_rate": 3.659485224022879e-07, "loss": 1.5962, "step": 46102 }, { "epoch": 8.78, "learning_rate": 3.6589132507149667e-07, "loss": 1.8867, "step": 46103 }, { "epoch": 8.78, "learning_rate": 3.6583412774070546e-07, "loss": 3.2138, "step": 46104 }, { "epoch": 8.78, "learning_rate": 3.657769304099142e-07, "loss": 1.3903, "step": 46105 }, { "epoch": 8.78, "learning_rate": 3.65719733079123e-07, "loss": 1.2546, "step": 46106 }, { "epoch": 8.78, "learning_rate": 3.6566253574833174e-07, "loss": 0.9403, "step": 46107 }, { "epoch": 8.78, "learning_rate": 3.6560533841754053e-07, "loss": 0.7508, "step": 46108 }, { "epoch": 8.78, "learning_rate": 3.655481410867493e-07, "loss": 1.5876, "step": 46109 }, { "epoch": 8.78, "learning_rate": 3.6549094375595806e-07, "loss": 1.9678, "step": 46110 }, { "epoch": 8.78, "learning_rate": 3.654337464251668e-07, "loss": 3.0793, "step": 46111 }, { "epoch": 8.78, "learning_rate": 3.653765490943756e-07, "loss": 2.1453, "step": 46112 }, { "epoch": 8.78, "learning_rate": 3.653193517635844e-07, "loss": 0.8686, "step": 46113 }, { "epoch": 8.78, "learning_rate": 3.6526215443279313e-07, "loss": 2.1059, "step": 46114 }, { "epoch": 8.78, "learning_rate": 3.6520495710200193e-07, "loss": 1.0259, "step": 46115 }, { "epoch": 8.78, "learning_rate": 3.651477597712107e-07, "loss": 3.7489, "step": 46116 }, { "epoch": 8.78, "learning_rate": 3.6509056244041946e-07, "loss": 1.8974, "step": 46117 }, { "epoch": 8.78, "learning_rate": 3.650333651096282e-07, "loss": 1.1083, "step": 46118 }, { "epoch": 8.78, "learning_rate": 3.64976167778837e-07, "loss": 1.9716, "step": 46119 }, { "epoch": 8.78, "learning_rate": 3.649189704480458e-07, "loss": 2.3681, "step": 46120 }, { "epoch": 8.78, "learning_rate": 3.6486177311725453e-07, "loss": 0.2134, "step": 46121 }, { "epoch": 8.79, "learning_rate": 3.648045757864633e-07, "loss": 2.1835, "step": 46122 }, { "epoch": 8.79, "learning_rate": 3.6474737845567207e-07, "loss": 0.9306, "step": 46123 }, { "epoch": 8.79, "learning_rate": 3.6469018112488086e-07, "loss": 1.3358, "step": 46124 }, { "epoch": 8.79, "learning_rate": 3.646329837940896e-07, "loss": 2.3718, "step": 46125 }, { "epoch": 8.79, "learning_rate": 3.645757864632984e-07, "loss": 1.3414, "step": 46126 }, { "epoch": 8.79, "learning_rate": 3.645185891325072e-07, "loss": 1.4363, "step": 46127 }, { "epoch": 8.79, "learning_rate": 3.6446139180171593e-07, "loss": 1.4065, "step": 46128 }, { "epoch": 8.79, "learning_rate": 3.6440419447092467e-07, "loss": 2.5799, "step": 46129 }, { "epoch": 8.79, "learning_rate": 3.6434699714013346e-07, "loss": 1.5705, "step": 46130 }, { "epoch": 8.79, "learning_rate": 3.6428979980934226e-07, "loss": 1.595, "step": 46131 }, { "epoch": 8.79, "learning_rate": 3.64232602478551e-07, "loss": 1.3452, "step": 46132 }, { "epoch": 8.79, "learning_rate": 3.641754051477598e-07, "loss": 0.7316, "step": 46133 }, { "epoch": 8.79, "learning_rate": 3.641182078169686e-07, "loss": 1.0678, "step": 46134 }, { "epoch": 8.79, "learning_rate": 3.6406101048617733e-07, "loss": 0.4027, "step": 46135 }, { "epoch": 8.79, "learning_rate": 3.6400381315538607e-07, "loss": 0.7436, "step": 46136 }, { "epoch": 8.79, "learning_rate": 3.6394661582459486e-07, "loss": 1.1909, "step": 46137 }, { "epoch": 8.79, "learning_rate": 3.6388941849380366e-07, "loss": 1.6832, "step": 46138 }, { "epoch": 8.79, "learning_rate": 3.638322211630124e-07, "loss": 2.0598, "step": 46139 }, { "epoch": 8.79, "learning_rate": 3.637750238322212e-07, "loss": 1.4677, "step": 46140 }, { "epoch": 8.79, "learning_rate": 3.6371782650142993e-07, "loss": 2.4259, "step": 46141 }, { "epoch": 8.79, "learning_rate": 3.636606291706387e-07, "loss": 1.6203, "step": 46142 }, { "epoch": 8.79, "learning_rate": 3.6360343183984747e-07, "loss": 1.3059, "step": 46143 }, { "epoch": 8.79, "learning_rate": 3.6354623450905626e-07, "loss": 0.7658, "step": 46144 }, { "epoch": 8.79, "learning_rate": 3.6348903717826505e-07, "loss": 2.15, "step": 46145 }, { "epoch": 8.79, "learning_rate": 3.634318398474738e-07, "loss": 2.0987, "step": 46146 }, { "epoch": 8.79, "learning_rate": 3.6337464251668254e-07, "loss": 1.1908, "step": 46147 }, { "epoch": 8.79, "learning_rate": 3.6331744518589133e-07, "loss": 2.0826, "step": 46148 }, { "epoch": 8.79, "learning_rate": 3.632602478551001e-07, "loss": 2.5007, "step": 46149 }, { "epoch": 8.79, "learning_rate": 3.6320305052430886e-07, "loss": 1.9062, "step": 46150 }, { "epoch": 8.79, "learning_rate": 3.6314585319351766e-07, "loss": 2.6839, "step": 46151 }, { "epoch": 8.79, "learning_rate": 3.630886558627264e-07, "loss": 1.9101, "step": 46152 }, { "epoch": 8.79, "learning_rate": 3.630314585319352e-07, "loss": 1.5877, "step": 46153 }, { "epoch": 8.79, "learning_rate": 3.6297426120114393e-07, "loss": 1.7344, "step": 46154 }, { "epoch": 8.79, "learning_rate": 3.6291706387035273e-07, "loss": 2.7679, "step": 46155 }, { "epoch": 8.79, "learning_rate": 3.628598665395615e-07, "loss": 3.4501, "step": 46156 }, { "epoch": 8.79, "learning_rate": 3.6280266920877026e-07, "loss": 2.2377, "step": 46157 }, { "epoch": 8.79, "learning_rate": 3.62745471877979e-07, "loss": 2.0626, "step": 46158 }, { "epoch": 8.79, "learning_rate": 3.626882745471878e-07, "loss": 1.944, "step": 46159 }, { "epoch": 8.79, "learning_rate": 3.626310772163966e-07, "loss": 1.8355, "step": 46160 }, { "epoch": 8.79, "learning_rate": 3.6257387988560533e-07, "loss": 2.002, "step": 46161 }, { "epoch": 8.79, "learning_rate": 3.625166825548141e-07, "loss": 0.9826, "step": 46162 }, { "epoch": 8.79, "learning_rate": 3.624594852240229e-07, "loss": 0.9858, "step": 46163 }, { "epoch": 8.79, "learning_rate": 3.6240228789323166e-07, "loss": 2.2975, "step": 46164 }, { "epoch": 8.79, "learning_rate": 3.623450905624404e-07, "loss": 1.0357, "step": 46165 }, { "epoch": 8.79, "learning_rate": 3.622878932316492e-07, "loss": 0.5551, "step": 46166 }, { "epoch": 8.79, "learning_rate": 3.62230695900858e-07, "loss": 2.9768, "step": 46167 }, { "epoch": 8.79, "learning_rate": 3.6217349857006673e-07, "loss": 1.0051, "step": 46168 }, { "epoch": 8.79, "learning_rate": 3.621163012392755e-07, "loss": 1.8917, "step": 46169 }, { "epoch": 8.79, "learning_rate": 3.6205910390848426e-07, "loss": 2.5385, "step": 46170 }, { "epoch": 8.79, "learning_rate": 3.6200190657769306e-07, "loss": 1.2157, "step": 46171 }, { "epoch": 8.79, "learning_rate": 3.619447092469018e-07, "loss": 1.0056, "step": 46172 }, { "epoch": 8.79, "learning_rate": 3.618875119161106e-07, "loss": 2.4228, "step": 46173 }, { "epoch": 8.8, "learning_rate": 3.618303145853194e-07, "loss": 1.0087, "step": 46174 }, { "epoch": 8.8, "learning_rate": 3.6177311725452813e-07, "loss": 1.8298, "step": 46175 }, { "epoch": 8.8, "learning_rate": 3.6171591992373687e-07, "loss": 2.2648, "step": 46176 }, { "epoch": 8.8, "learning_rate": 3.6165872259294566e-07, "loss": 2.3643, "step": 46177 }, { "epoch": 8.8, "learning_rate": 3.6160152526215446e-07, "loss": 1.402, "step": 46178 }, { "epoch": 8.8, "learning_rate": 3.615443279313632e-07, "loss": 3.7679, "step": 46179 }, { "epoch": 8.8, "learning_rate": 3.61487130600572e-07, "loss": 2.1462, "step": 46180 }, { "epoch": 8.8, "learning_rate": 3.614299332697808e-07, "loss": 1.0311, "step": 46181 }, { "epoch": 8.8, "learning_rate": 3.613727359389895e-07, "loss": 1.4637, "step": 46182 }, { "epoch": 8.8, "learning_rate": 3.6131553860819827e-07, "loss": 0.9988, "step": 46183 }, { "epoch": 8.8, "learning_rate": 3.6125834127740706e-07, "loss": 0.8602, "step": 46184 }, { "epoch": 8.8, "learning_rate": 3.6120114394661585e-07, "loss": 1.2863, "step": 46185 }, { "epoch": 8.8, "learning_rate": 3.611439466158246e-07, "loss": 0.3264, "step": 46186 }, { "epoch": 8.8, "learning_rate": 3.610867492850334e-07, "loss": 1.5121, "step": 46187 }, { "epoch": 8.8, "learning_rate": 3.6102955195424213e-07, "loss": 0.9516, "step": 46188 }, { "epoch": 8.8, "learning_rate": 3.609723546234509e-07, "loss": 1.3902, "step": 46189 }, { "epoch": 8.8, "learning_rate": 3.6091515729265966e-07, "loss": 1.5536, "step": 46190 }, { "epoch": 8.8, "learning_rate": 3.6085795996186846e-07, "loss": 1.7858, "step": 46191 }, { "epoch": 8.8, "learning_rate": 3.6080076263107725e-07, "loss": 0.4653, "step": 46192 }, { "epoch": 8.8, "learning_rate": 3.60743565300286e-07, "loss": 1.7122, "step": 46193 }, { "epoch": 8.8, "learning_rate": 3.6068636796949473e-07, "loss": 1.7491, "step": 46194 }, { "epoch": 8.8, "learning_rate": 3.6062917063870353e-07, "loss": 1.737, "step": 46195 }, { "epoch": 8.8, "learning_rate": 3.605719733079123e-07, "loss": 1.3079, "step": 46196 }, { "epoch": 8.8, "learning_rate": 3.6051477597712106e-07, "loss": 1.8077, "step": 46197 }, { "epoch": 8.8, "learning_rate": 3.6045757864632986e-07, "loss": 0.8379, "step": 46198 }, { "epoch": 8.8, "learning_rate": 3.6040038131553865e-07, "loss": 0.5346, "step": 46199 }, { "epoch": 8.8, "learning_rate": 3.603431839847474e-07, "loss": 0.4323, "step": 46200 }, { "epoch": 8.8, "learning_rate": 3.6028598665395613e-07, "loss": 1.2259, "step": 46201 }, { "epoch": 8.8, "learning_rate": 3.602287893231649e-07, "loss": 2.2851, "step": 46202 }, { "epoch": 8.8, "learning_rate": 3.601715919923737e-07, "loss": 1.8841, "step": 46203 }, { "epoch": 8.8, "learning_rate": 3.6011439466158246e-07, "loss": 1.6434, "step": 46204 }, { "epoch": 8.8, "learning_rate": 3.600571973307912e-07, "loss": 0.9288, "step": 46205 }, { "epoch": 8.8, "learning_rate": 3.6e-07, "loss": 1.1648, "step": 46206 }, { "epoch": 8.8, "learning_rate": 3.599428026692088e-07, "loss": 1.441, "step": 46207 }, { "epoch": 8.8, "learning_rate": 3.5988560533841753e-07, "loss": 0.8569, "step": 46208 }, { "epoch": 8.8, "learning_rate": 3.598284080076263e-07, "loss": 1.1948, "step": 46209 }, { "epoch": 8.8, "learning_rate": 3.597712106768351e-07, "loss": 1.0179, "step": 46210 }, { "epoch": 8.8, "learning_rate": 3.5971401334604386e-07, "loss": 1.4201, "step": 46211 }, { "epoch": 8.8, "learning_rate": 3.596568160152526e-07, "loss": 1.7443, "step": 46212 }, { "epoch": 8.8, "learning_rate": 3.595996186844614e-07, "loss": 1.9505, "step": 46213 }, { "epoch": 8.8, "learning_rate": 3.595424213536702e-07, "loss": 1.8208, "step": 46214 }, { "epoch": 8.8, "learning_rate": 3.5948522402287893e-07, "loss": 0.5488, "step": 46215 }, { "epoch": 8.8, "learning_rate": 3.594280266920877e-07, "loss": 0.5547, "step": 46216 }, { "epoch": 8.8, "learning_rate": 3.5937082936129646e-07, "loss": 1.6973, "step": 46217 }, { "epoch": 8.8, "learning_rate": 3.5931363203050526e-07, "loss": 1.5979, "step": 46218 }, { "epoch": 8.8, "learning_rate": 3.59256434699714e-07, "loss": 2.3884, "step": 46219 }, { "epoch": 8.8, "learning_rate": 3.591992373689228e-07, "loss": 2.2486, "step": 46220 }, { "epoch": 8.8, "learning_rate": 3.591420400381316e-07, "loss": 1.7981, "step": 46221 }, { "epoch": 8.8, "learning_rate": 3.590848427073403e-07, "loss": 1.4283, "step": 46222 }, { "epoch": 8.8, "learning_rate": 3.5902764537654907e-07, "loss": 0.7407, "step": 46223 }, { "epoch": 8.8, "learning_rate": 3.5897044804575786e-07, "loss": 1.1045, "step": 46224 }, { "epoch": 8.8, "learning_rate": 3.5891325071496665e-07, "loss": 0.962, "step": 46225 }, { "epoch": 8.8, "learning_rate": 3.588560533841754e-07, "loss": 3.9755, "step": 46226 }, { "epoch": 8.81, "learning_rate": 3.587988560533842e-07, "loss": 3.2067, "step": 46227 }, { "epoch": 8.81, "learning_rate": 3.58741658722593e-07, "loss": 1.394, "step": 46228 }, { "epoch": 8.81, "learning_rate": 3.586844613918017e-07, "loss": 0.5476, "step": 46229 }, { "epoch": 8.81, "learning_rate": 3.5862726406101046e-07, "loss": 2.0205, "step": 46230 }, { "epoch": 8.81, "learning_rate": 3.5857006673021926e-07, "loss": 2.4358, "step": 46231 }, { "epoch": 8.81, "learning_rate": 3.5851286939942805e-07, "loss": 2.7927, "step": 46232 }, { "epoch": 8.81, "learning_rate": 3.584556720686368e-07, "loss": 1.3931, "step": 46233 }, { "epoch": 8.81, "learning_rate": 3.583984747378456e-07, "loss": 0.7551, "step": 46234 }, { "epoch": 8.81, "learning_rate": 3.5834127740705433e-07, "loss": 3.2078, "step": 46235 }, { "epoch": 8.81, "learning_rate": 3.582840800762631e-07, "loss": 1.4689, "step": 46236 }, { "epoch": 8.81, "learning_rate": 3.5822688274547186e-07, "loss": 1.6797, "step": 46237 }, { "epoch": 8.81, "learning_rate": 3.5816968541468066e-07, "loss": 2.8472, "step": 46238 }, { "epoch": 8.81, "learning_rate": 3.5811248808388945e-07, "loss": 2.5814, "step": 46239 }, { "epoch": 8.81, "learning_rate": 3.580552907530982e-07, "loss": 1.0879, "step": 46240 }, { "epoch": 8.81, "learning_rate": 3.5799809342230693e-07, "loss": 1.3932, "step": 46241 }, { "epoch": 8.81, "learning_rate": 3.579408960915157e-07, "loss": 0.9843, "step": 46242 }, { "epoch": 8.81, "learning_rate": 3.578836987607245e-07, "loss": 1.7121, "step": 46243 }, { "epoch": 8.81, "learning_rate": 3.5782650142993326e-07, "loss": 1.7093, "step": 46244 }, { "epoch": 8.81, "learning_rate": 3.5776930409914205e-07, "loss": 1.8986, "step": 46245 }, { "epoch": 8.81, "learning_rate": 3.5771210676835085e-07, "loss": 1.2098, "step": 46246 }, { "epoch": 8.81, "learning_rate": 3.576549094375596e-07, "loss": 3.053, "step": 46247 }, { "epoch": 8.81, "learning_rate": 3.5759771210676833e-07, "loss": 1.9323, "step": 46248 }, { "epoch": 8.81, "learning_rate": 3.575405147759771e-07, "loss": 1.4707, "step": 46249 }, { "epoch": 8.81, "learning_rate": 3.574833174451859e-07, "loss": 1.5991, "step": 46250 }, { "epoch": 8.81, "learning_rate": 3.5742612011439466e-07, "loss": 1.0056, "step": 46251 }, { "epoch": 8.81, "learning_rate": 3.5736892278360345e-07, "loss": 0.6963, "step": 46252 }, { "epoch": 8.81, "learning_rate": 3.573117254528122e-07, "loss": 0.4271, "step": 46253 }, { "epoch": 8.81, "learning_rate": 3.57254528122021e-07, "loss": 1.3384, "step": 46254 }, { "epoch": 8.81, "learning_rate": 3.5719733079122973e-07, "loss": 1.1592, "step": 46255 }, { "epoch": 8.81, "learning_rate": 3.571401334604385e-07, "loss": 1.6113, "step": 46256 }, { "epoch": 8.81, "learning_rate": 3.570829361296473e-07, "loss": 1.44, "step": 46257 }, { "epoch": 8.81, "learning_rate": 3.570257387988561e-07, "loss": 2.0457, "step": 46258 }, { "epoch": 8.81, "learning_rate": 3.569685414680648e-07, "loss": 1.2928, "step": 46259 }, { "epoch": 8.81, "learning_rate": 3.569113441372736e-07, "loss": 1.1373, "step": 46260 }, { "epoch": 8.81, "learning_rate": 3.568541468064824e-07, "loss": 1.7339, "step": 46261 }, { "epoch": 8.81, "learning_rate": 3.567969494756911e-07, "loss": 1.9464, "step": 46262 }, { "epoch": 8.81, "learning_rate": 3.567397521448999e-07, "loss": 1.2268, "step": 46263 }, { "epoch": 8.81, "learning_rate": 3.5668255481410866e-07, "loss": 1.9584, "step": 46264 }, { "epoch": 8.81, "learning_rate": 3.5662535748331745e-07, "loss": 1.3706, "step": 46265 }, { "epoch": 8.81, "learning_rate": 3.565681601525262e-07, "loss": 1.3434, "step": 46266 }, { "epoch": 8.81, "learning_rate": 3.56510962821735e-07, "loss": 1.1694, "step": 46267 }, { "epoch": 8.81, "learning_rate": 3.564537654909438e-07, "loss": 1.3725, "step": 46268 }, { "epoch": 8.81, "learning_rate": 3.563965681601526e-07, "loss": 2.5646, "step": 46269 }, { "epoch": 8.81, "learning_rate": 3.5633937082936126e-07, "loss": 0.9258, "step": 46270 }, { "epoch": 8.81, "learning_rate": 3.5628217349857006e-07, "loss": 0.7974, "step": 46271 }, { "epoch": 8.81, "learning_rate": 3.5622497616777885e-07, "loss": 1.4519, "step": 46272 }, { "epoch": 8.81, "learning_rate": 3.561677788369876e-07, "loss": 0.9484, "step": 46273 }, { "epoch": 8.81, "learning_rate": 3.561105815061964e-07, "loss": 1.8333, "step": 46274 }, { "epoch": 8.81, "learning_rate": 3.560533841754052e-07, "loss": 0.21, "step": 46275 }, { "epoch": 8.81, "learning_rate": 3.559961868446139e-07, "loss": 1.5732, "step": 46276 }, { "epoch": 8.81, "learning_rate": 3.5593898951382266e-07, "loss": 2.7768, "step": 46277 }, { "epoch": 8.81, "learning_rate": 3.5588179218303146e-07, "loss": 1.5683, "step": 46278 }, { "epoch": 8.82, "learning_rate": 3.5582459485224025e-07, "loss": 0.5818, "step": 46279 }, { "epoch": 8.82, "learning_rate": 3.5576739752144904e-07, "loss": 2.8839, "step": 46280 }, { "epoch": 8.82, "learning_rate": 3.557102001906578e-07, "loss": 0.7013, "step": 46281 }, { "epoch": 8.82, "learning_rate": 3.556530028598665e-07, "loss": 1.663, "step": 46282 }, { "epoch": 8.82, "learning_rate": 3.555958055290753e-07, "loss": 1.3562, "step": 46283 }, { "epoch": 8.82, "learning_rate": 3.5553860819828406e-07, "loss": 2.7406, "step": 46284 }, { "epoch": 8.82, "learning_rate": 3.5548141086749285e-07, "loss": 0.3763, "step": 46285 }, { "epoch": 8.82, "learning_rate": 3.5542421353670165e-07, "loss": 1.0985, "step": 46286 }, { "epoch": 8.82, "learning_rate": 3.5536701620591044e-07, "loss": 1.1243, "step": 46287 }, { "epoch": 8.82, "learning_rate": 3.5530981887511913e-07, "loss": 2.0106, "step": 46288 }, { "epoch": 8.82, "learning_rate": 3.552526215443279e-07, "loss": 1.6933, "step": 46289 }, { "epoch": 8.82, "learning_rate": 3.551954242135367e-07, "loss": 1.191, "step": 46290 }, { "epoch": 8.82, "learning_rate": 3.551382268827455e-07, "loss": 2.1129, "step": 46291 }, { "epoch": 8.82, "learning_rate": 3.5508102955195425e-07, "loss": 0.4613, "step": 46292 }, { "epoch": 8.82, "learning_rate": 3.5502383222116304e-07, "loss": 0.8375, "step": 46293 }, { "epoch": 8.82, "learning_rate": 3.549666348903718e-07, "loss": 2.0137, "step": 46294 }, { "epoch": 8.82, "learning_rate": 3.5490943755958053e-07, "loss": 2.3843, "step": 46295 }, { "epoch": 8.82, "learning_rate": 3.548522402287893e-07, "loss": 1.0371, "step": 46296 }, { "epoch": 8.82, "learning_rate": 3.547950428979981e-07, "loss": 1.5029, "step": 46297 }, { "epoch": 8.82, "learning_rate": 3.547378455672069e-07, "loss": 0.5251, "step": 46298 }, { "epoch": 8.82, "learning_rate": 3.5468064823641565e-07, "loss": 0.6934, "step": 46299 }, { "epoch": 8.82, "learning_rate": 3.546234509056244e-07, "loss": 1.2058, "step": 46300 }, { "epoch": 8.82, "learning_rate": 3.545662535748332e-07, "loss": 1.7215, "step": 46301 }, { "epoch": 8.82, "learning_rate": 3.54509056244042e-07, "loss": 0.5284, "step": 46302 }, { "epoch": 8.82, "learning_rate": 3.544518589132507e-07, "loss": 0.2836, "step": 46303 }, { "epoch": 8.82, "learning_rate": 3.543946615824595e-07, "loss": 1.0201, "step": 46304 }, { "epoch": 8.82, "learning_rate": 3.543374642516683e-07, "loss": 1.8913, "step": 46305 }, { "epoch": 8.82, "learning_rate": 3.54280266920877e-07, "loss": 1.0843, "step": 46306 }, { "epoch": 8.82, "learning_rate": 3.542230695900858e-07, "loss": 2.2836, "step": 46307 }, { "epoch": 8.82, "learning_rate": 3.541658722592946e-07, "loss": 0.8128, "step": 46308 }, { "epoch": 8.82, "learning_rate": 3.541086749285034e-07, "loss": 0.9784, "step": 46309 }, { "epoch": 8.82, "learning_rate": 3.540514775977121e-07, "loss": 1.342, "step": 46310 }, { "epoch": 8.82, "learning_rate": 3.539942802669209e-07, "loss": 0.3532, "step": 46311 }, { "epoch": 8.82, "learning_rate": 3.5393708293612965e-07, "loss": 0.7551, "step": 46312 }, { "epoch": 8.82, "learning_rate": 3.5387988560533844e-07, "loss": 1.5192, "step": 46313 }, { "epoch": 8.82, "learning_rate": 3.538226882745472e-07, "loss": 0.5818, "step": 46314 }, { "epoch": 8.82, "learning_rate": 3.53765490943756e-07, "loss": 2.2664, "step": 46315 }, { "epoch": 8.82, "learning_rate": 3.5370829361296477e-07, "loss": 0.4073, "step": 46316 }, { "epoch": 8.82, "learning_rate": 3.536510962821735e-07, "loss": 0.5246, "step": 46317 }, { "epoch": 8.82, "learning_rate": 3.5359389895138225e-07, "loss": 2.0439, "step": 46318 }, { "epoch": 8.82, "learning_rate": 3.5353670162059105e-07, "loss": 1.8209, "step": 46319 }, { "epoch": 8.82, "learning_rate": 3.5347950428979984e-07, "loss": 0.7678, "step": 46320 }, { "epoch": 8.82, "learning_rate": 3.534223069590086e-07, "loss": 3.3985, "step": 46321 }, { "epoch": 8.82, "learning_rate": 3.533651096282174e-07, "loss": 2.2038, "step": 46322 }, { "epoch": 8.82, "learning_rate": 3.533079122974261e-07, "loss": 2.667, "step": 46323 }, { "epoch": 8.82, "learning_rate": 3.532507149666349e-07, "loss": 1.7877, "step": 46324 }, { "epoch": 8.82, "learning_rate": 3.5319351763584365e-07, "loss": 1.756, "step": 46325 }, { "epoch": 8.82, "learning_rate": 3.5313632030505245e-07, "loss": 0.9852, "step": 46326 }, { "epoch": 8.82, "learning_rate": 3.5307912297426124e-07, "loss": 0.8146, "step": 46327 }, { "epoch": 8.82, "learning_rate": 3.5302192564347e-07, "loss": 2.0584, "step": 46328 }, { "epoch": 8.82, "learning_rate": 3.529647283126787e-07, "loss": 1.8902, "step": 46329 }, { "epoch": 8.82, "learning_rate": 3.529075309818875e-07, "loss": 1.737, "step": 46330 }, { "epoch": 8.82, "learning_rate": 3.528503336510963e-07, "loss": 0.8334, "step": 46331 }, { "epoch": 8.83, "learning_rate": 3.5279313632030505e-07, "loss": 2.1669, "step": 46332 }, { "epoch": 8.83, "learning_rate": 3.5273593898951384e-07, "loss": 2.1925, "step": 46333 }, { "epoch": 8.83, "learning_rate": 3.5267874165872264e-07, "loss": 1.2522, "step": 46334 }, { "epoch": 8.83, "learning_rate": 3.526215443279314e-07, "loss": 1.0356, "step": 46335 }, { "epoch": 8.83, "learning_rate": 3.525643469971401e-07, "loss": 3.0523, "step": 46336 }, { "epoch": 8.83, "learning_rate": 3.525071496663489e-07, "loss": 0.4114, "step": 46337 }, { "epoch": 8.83, "learning_rate": 3.524499523355577e-07, "loss": 1.0376, "step": 46338 }, { "epoch": 8.83, "learning_rate": 3.5239275500476645e-07, "loss": 1.8699, "step": 46339 }, { "epoch": 8.83, "learning_rate": 3.5233555767397524e-07, "loss": 1.2347, "step": 46340 }, { "epoch": 8.83, "learning_rate": 3.52278360343184e-07, "loss": 1.4699, "step": 46341 }, { "epoch": 8.83, "learning_rate": 3.522211630123928e-07, "loss": 0.8628, "step": 46342 }, { "epoch": 8.83, "learning_rate": 3.521639656816015e-07, "loss": 1.1231, "step": 46343 }, { "epoch": 8.83, "learning_rate": 3.521067683508103e-07, "loss": 2.4844, "step": 46344 }, { "epoch": 8.83, "learning_rate": 3.520495710200191e-07, "loss": 2.1752, "step": 46345 }, { "epoch": 8.83, "learning_rate": 3.5199237368922785e-07, "loss": 1.2568, "step": 46346 }, { "epoch": 8.83, "learning_rate": 3.519351763584366e-07, "loss": 1.0511, "step": 46347 }, { "epoch": 8.83, "learning_rate": 3.518779790276454e-07, "loss": 1.8543, "step": 46348 }, { "epoch": 8.83, "learning_rate": 3.518207816968542e-07, "loss": 1.0403, "step": 46349 }, { "epoch": 8.83, "learning_rate": 3.517635843660629e-07, "loss": 1.1131, "step": 46350 }, { "epoch": 8.83, "learning_rate": 3.517063870352717e-07, "loss": 1.1809, "step": 46351 }, { "epoch": 8.83, "learning_rate": 3.516491897044805e-07, "loss": 2.0143, "step": 46352 }, { "epoch": 8.83, "learning_rate": 3.5159199237368924e-07, "loss": 1.3617, "step": 46353 }, { "epoch": 8.83, "learning_rate": 3.51534795042898e-07, "loss": 1.5399, "step": 46354 }, { "epoch": 8.83, "learning_rate": 3.514775977121068e-07, "loss": 2.0416, "step": 46355 }, { "epoch": 8.83, "learning_rate": 3.5142040038131557e-07, "loss": 2.2849, "step": 46356 }, { "epoch": 8.83, "learning_rate": 3.513632030505243e-07, "loss": 2.5646, "step": 46357 }, { "epoch": 8.83, "learning_rate": 3.513060057197331e-07, "loss": 0.7892, "step": 46358 }, { "epoch": 8.83, "learning_rate": 3.5124880838894185e-07, "loss": 1.3397, "step": 46359 }, { "epoch": 8.83, "learning_rate": 3.5119161105815064e-07, "loss": 2.3425, "step": 46360 }, { "epoch": 8.83, "learning_rate": 3.511344137273594e-07, "loss": 0.5998, "step": 46361 }, { "epoch": 8.83, "learning_rate": 3.510772163965682e-07, "loss": 1.689, "step": 46362 }, { "epoch": 8.83, "learning_rate": 3.5102001906577697e-07, "loss": 2.2362, "step": 46363 }, { "epoch": 8.83, "learning_rate": 3.509628217349857e-07, "loss": 1.0037, "step": 46364 }, { "epoch": 8.83, "learning_rate": 3.5090562440419445e-07, "loss": 0.9867, "step": 46365 }, { "epoch": 8.83, "learning_rate": 3.5084842707340325e-07, "loss": 0.3482, "step": 46366 }, { "epoch": 8.83, "learning_rate": 3.5079122974261204e-07, "loss": 1.34, "step": 46367 }, { "epoch": 8.83, "learning_rate": 3.507340324118208e-07, "loss": 1.1389, "step": 46368 }, { "epoch": 8.83, "learning_rate": 3.506768350810296e-07, "loss": 1.8989, "step": 46369 }, { "epoch": 8.83, "learning_rate": 3.5061963775023837e-07, "loss": 1.8543, "step": 46370 }, { "epoch": 8.83, "learning_rate": 3.505624404194471e-07, "loss": 0.3779, "step": 46371 }, { "epoch": 8.83, "learning_rate": 3.5050524308865585e-07, "loss": 1.3429, "step": 46372 }, { "epoch": 8.83, "learning_rate": 3.5044804575786464e-07, "loss": 0.5093, "step": 46373 }, { "epoch": 8.83, "learning_rate": 3.5039084842707344e-07, "loss": 1.9239, "step": 46374 }, { "epoch": 8.83, "learning_rate": 3.503336510962822e-07, "loss": 2.4353, "step": 46375 }, { "epoch": 8.83, "learning_rate": 3.5027645376549097e-07, "loss": 2.3725, "step": 46376 }, { "epoch": 8.83, "learning_rate": 3.502192564346997e-07, "loss": 0.8859, "step": 46377 }, { "epoch": 8.83, "learning_rate": 3.501620591039085e-07, "loss": 2.7077, "step": 46378 }, { "epoch": 8.83, "learning_rate": 3.5010486177311725e-07, "loss": 1.1165, "step": 46379 }, { "epoch": 8.83, "learning_rate": 3.5004766444232604e-07, "loss": 1.0662, "step": 46380 }, { "epoch": 8.83, "learning_rate": 3.4999046711153484e-07, "loss": 0.3979, "step": 46381 }, { "epoch": 8.83, "learning_rate": 3.499332697807436e-07, "loss": 1.9504, "step": 46382 }, { "epoch": 8.83, "learning_rate": 3.498760724499523e-07, "loss": 1.3134, "step": 46383 }, { "epoch": 8.84, "learning_rate": 3.498188751191611e-07, "loss": 2.1458, "step": 46384 }, { "epoch": 8.84, "learning_rate": 3.497616777883699e-07, "loss": 0.499, "step": 46385 }, { "epoch": 8.84, "learning_rate": 3.4970448045757865e-07, "loss": 1.1016, "step": 46386 }, { "epoch": 8.84, "learning_rate": 3.4964728312678744e-07, "loss": 2.4691, "step": 46387 }, { "epoch": 8.84, "learning_rate": 3.495900857959962e-07, "loss": 0.794, "step": 46388 }, { "epoch": 8.84, "learning_rate": 3.49532888465205e-07, "loss": 1.5575, "step": 46389 }, { "epoch": 8.84, "learning_rate": 3.494756911344137e-07, "loss": 0.9438, "step": 46390 }, { "epoch": 8.84, "learning_rate": 3.494184938036225e-07, "loss": 2.1942, "step": 46391 }, { "epoch": 8.84, "learning_rate": 3.493612964728313e-07, "loss": 1.689, "step": 46392 }, { "epoch": 8.84, "learning_rate": 3.4930409914204004e-07, "loss": 0.8705, "step": 46393 }, { "epoch": 8.84, "learning_rate": 3.492469018112488e-07, "loss": 1.7666, "step": 46394 }, { "epoch": 8.84, "learning_rate": 3.491897044804576e-07, "loss": 0.4046, "step": 46395 }, { "epoch": 8.84, "learning_rate": 3.4913250714966637e-07, "loss": 1.3097, "step": 46396 }, { "epoch": 8.84, "learning_rate": 3.490753098188751e-07, "loss": 1.3534, "step": 46397 }, { "epoch": 8.84, "learning_rate": 3.490181124880839e-07, "loss": 1.4014, "step": 46398 }, { "epoch": 8.84, "learning_rate": 3.489609151572927e-07, "loss": 1.4822, "step": 46399 }, { "epoch": 8.84, "learning_rate": 3.4890371782650144e-07, "loss": 2.2043, "step": 46400 }, { "epoch": 8.84, "learning_rate": 3.488465204957102e-07, "loss": 2.4886, "step": 46401 }, { "epoch": 8.84, "learning_rate": 3.48789323164919e-07, "loss": 1.0121, "step": 46402 }, { "epoch": 8.84, "learning_rate": 3.4873212583412777e-07, "loss": 2.038, "step": 46403 }, { "epoch": 8.84, "learning_rate": 3.486749285033365e-07, "loss": 2.6624, "step": 46404 }, { "epoch": 8.84, "learning_rate": 3.486177311725453e-07, "loss": 1.1871, "step": 46405 }, { "epoch": 8.84, "learning_rate": 3.4856053384175405e-07, "loss": 0.3021, "step": 46406 }, { "epoch": 8.84, "learning_rate": 3.4850333651096284e-07, "loss": 0.8544, "step": 46407 }, { "epoch": 8.84, "learning_rate": 3.484461391801716e-07, "loss": 1.479, "step": 46408 }, { "epoch": 8.84, "learning_rate": 3.483889418493804e-07, "loss": 2.1567, "step": 46409 }, { "epoch": 8.84, "learning_rate": 3.4833174451858917e-07, "loss": 2.2269, "step": 46410 }, { "epoch": 8.84, "learning_rate": 3.482745471877979e-07, "loss": 1.3872, "step": 46411 }, { "epoch": 8.84, "learning_rate": 3.4821734985700665e-07, "loss": 2.2271, "step": 46412 }, { "epoch": 8.84, "learning_rate": 3.4816015252621544e-07, "loss": 1.2853, "step": 46413 }, { "epoch": 8.84, "learning_rate": 3.4810295519542424e-07, "loss": 2.8295, "step": 46414 }, { "epoch": 8.84, "learning_rate": 3.48045757864633e-07, "loss": 0.5181, "step": 46415 }, { "epoch": 8.84, "learning_rate": 3.4798856053384177e-07, "loss": 2.3219, "step": 46416 }, { "epoch": 8.84, "learning_rate": 3.4793136320305057e-07, "loss": 0.9077, "step": 46417 }, { "epoch": 8.84, "learning_rate": 3.478741658722593e-07, "loss": 1.7666, "step": 46418 }, { "epoch": 8.84, "learning_rate": 3.4781696854146805e-07, "loss": 1.6272, "step": 46419 }, { "epoch": 8.84, "learning_rate": 3.4775977121067684e-07, "loss": 1.6712, "step": 46420 }, { "epoch": 8.84, "learning_rate": 3.4770257387988564e-07, "loss": 1.4801, "step": 46421 }, { "epoch": 8.84, "learning_rate": 3.476453765490944e-07, "loss": 0.7847, "step": 46422 }, { "epoch": 8.84, "learning_rate": 3.4758817921830317e-07, "loss": 0.9676, "step": 46423 }, { "epoch": 8.84, "learning_rate": 3.475309818875119e-07, "loss": 2.0746, "step": 46424 }, { "epoch": 8.84, "learning_rate": 3.474737845567207e-07, "loss": 0.4938, "step": 46425 }, { "epoch": 8.84, "learning_rate": 3.4741658722592945e-07, "loss": 2.2443, "step": 46426 }, { "epoch": 8.84, "learning_rate": 3.4735938989513824e-07, "loss": 1.7763, "step": 46427 }, { "epoch": 8.84, "learning_rate": 3.4730219256434703e-07, "loss": 1.8186, "step": 46428 }, { "epoch": 8.84, "learning_rate": 3.472449952335558e-07, "loss": 1.2611, "step": 46429 }, { "epoch": 8.84, "learning_rate": 3.471877979027645e-07, "loss": 1.6199, "step": 46430 }, { "epoch": 8.84, "learning_rate": 3.471306005719733e-07, "loss": 1.2129, "step": 46431 }, { "epoch": 8.84, "learning_rate": 3.470734032411821e-07, "loss": 0.8208, "step": 46432 }, { "epoch": 8.84, "learning_rate": 3.4701620591039084e-07, "loss": 1.5014, "step": 46433 }, { "epoch": 8.84, "learning_rate": 3.4695900857959964e-07, "loss": 2.5125, "step": 46434 }, { "epoch": 8.84, "learning_rate": 3.4690181124880843e-07, "loss": 1.3741, "step": 46435 }, { "epoch": 8.84, "learning_rate": 3.4684461391801717e-07, "loss": 1.3866, "step": 46436 }, { "epoch": 8.85, "learning_rate": 3.467874165872259e-07, "loss": 2.1226, "step": 46437 }, { "epoch": 8.85, "learning_rate": 3.467302192564347e-07, "loss": 2.1354, "step": 46438 }, { "epoch": 8.85, "learning_rate": 3.466730219256435e-07, "loss": 2.2895, "step": 46439 }, { "epoch": 8.85, "learning_rate": 3.4661582459485224e-07, "loss": 1.0893, "step": 46440 }, { "epoch": 8.85, "learning_rate": 3.4655862726406104e-07, "loss": 1.9076, "step": 46441 }, { "epoch": 8.85, "learning_rate": 3.465014299332698e-07, "loss": 1.7277, "step": 46442 }, { "epoch": 8.85, "learning_rate": 3.4644423260247857e-07, "loss": 2.3344, "step": 46443 }, { "epoch": 8.85, "learning_rate": 3.463870352716873e-07, "loss": 1.313, "step": 46444 }, { "epoch": 8.85, "learning_rate": 3.463298379408961e-07, "loss": 2.552, "step": 46445 }, { "epoch": 8.85, "learning_rate": 3.462726406101049e-07, "loss": 1.4454, "step": 46446 }, { "epoch": 8.85, "learning_rate": 3.4621544327931364e-07, "loss": 1.5787, "step": 46447 }, { "epoch": 8.85, "learning_rate": 3.461582459485224e-07, "loss": 1.9631, "step": 46448 }, { "epoch": 8.85, "learning_rate": 3.461010486177312e-07, "loss": 2.6218, "step": 46449 }, { "epoch": 8.85, "learning_rate": 3.4604385128693997e-07, "loss": 2.1675, "step": 46450 }, { "epoch": 8.85, "learning_rate": 3.459866539561487e-07, "loss": 3.0956, "step": 46451 }, { "epoch": 8.85, "learning_rate": 3.459294566253575e-07, "loss": 1.6644, "step": 46452 }, { "epoch": 8.85, "learning_rate": 3.4587225929456624e-07, "loss": 1.5065, "step": 46453 }, { "epoch": 8.85, "learning_rate": 3.4581506196377504e-07, "loss": 1.1219, "step": 46454 }, { "epoch": 8.85, "learning_rate": 3.457578646329838e-07, "loss": 0.8656, "step": 46455 }, { "epoch": 8.85, "learning_rate": 3.4570066730219257e-07, "loss": 2.0082, "step": 46456 }, { "epoch": 8.85, "learning_rate": 3.4564346997140137e-07, "loss": 1.6771, "step": 46457 }, { "epoch": 8.85, "learning_rate": 3.455862726406101e-07, "loss": 0.9739, "step": 46458 }, { "epoch": 8.85, "learning_rate": 3.4552907530981885e-07, "loss": 0.5967, "step": 46459 }, { "epoch": 8.85, "learning_rate": 3.4547187797902764e-07, "loss": 1.1025, "step": 46460 }, { "epoch": 8.85, "learning_rate": 3.4541468064823644e-07, "loss": 2.8022, "step": 46461 }, { "epoch": 8.85, "learning_rate": 3.453574833174452e-07, "loss": 1.0611, "step": 46462 }, { "epoch": 8.85, "learning_rate": 3.4530028598665397e-07, "loss": 0.9305, "step": 46463 }, { "epoch": 8.85, "learning_rate": 3.4524308865586276e-07, "loss": 1.9748, "step": 46464 }, { "epoch": 8.85, "learning_rate": 3.451858913250715e-07, "loss": 1.3294, "step": 46465 }, { "epoch": 8.85, "learning_rate": 3.4512869399428025e-07, "loss": 1.4212, "step": 46466 }, { "epoch": 8.85, "learning_rate": 3.4507149666348904e-07, "loss": 1.4313, "step": 46467 }, { "epoch": 8.85, "learning_rate": 3.4501429933269783e-07, "loss": 1.8137, "step": 46468 }, { "epoch": 8.85, "learning_rate": 3.449571020019066e-07, "loss": 2.3737, "step": 46469 }, { "epoch": 8.85, "learning_rate": 3.4489990467111537e-07, "loss": 0.4887, "step": 46470 }, { "epoch": 8.85, "learning_rate": 3.448427073403241e-07, "loss": 0.8489, "step": 46471 }, { "epoch": 8.85, "learning_rate": 3.447855100095329e-07, "loss": 0.5894, "step": 46472 }, { "epoch": 8.85, "learning_rate": 3.4472831267874164e-07, "loss": 0.6833, "step": 46473 }, { "epoch": 8.85, "learning_rate": 3.4467111534795044e-07, "loss": 0.4845, "step": 46474 }, { "epoch": 8.85, "learning_rate": 3.4461391801715923e-07, "loss": 1.8918, "step": 46475 }, { "epoch": 8.85, "learning_rate": 3.4455672068636797e-07, "loss": 1.3351, "step": 46476 }, { "epoch": 8.85, "learning_rate": 3.444995233555767e-07, "loss": 1.3647, "step": 46477 }, { "epoch": 8.85, "learning_rate": 3.444423260247855e-07, "loss": 1.7005, "step": 46478 }, { "epoch": 8.85, "learning_rate": 3.443851286939943e-07, "loss": 1.2576, "step": 46479 }, { "epoch": 8.85, "learning_rate": 3.4432793136320304e-07, "loss": 1.4839, "step": 46480 }, { "epoch": 8.85, "learning_rate": 3.4427073403241184e-07, "loss": 1.6498, "step": 46481 }, { "epoch": 8.85, "learning_rate": 3.4421353670162063e-07, "loss": 1.7675, "step": 46482 }, { "epoch": 8.85, "learning_rate": 3.4415633937082937e-07, "loss": 2.3577, "step": 46483 }, { "epoch": 8.85, "learning_rate": 3.440991420400381e-07, "loss": 1.267, "step": 46484 }, { "epoch": 8.85, "learning_rate": 3.440419447092469e-07, "loss": 0.3604, "step": 46485 }, { "epoch": 8.85, "learning_rate": 3.439847473784557e-07, "loss": 3.0989, "step": 46486 }, { "epoch": 8.85, "learning_rate": 3.4392755004766444e-07, "loss": 1.9265, "step": 46487 }, { "epoch": 8.85, "learning_rate": 3.4387035271687323e-07, "loss": 2.8147, "step": 46488 }, { "epoch": 8.86, "learning_rate": 3.43813155386082e-07, "loss": 1.6902, "step": 46489 }, { "epoch": 8.86, "learning_rate": 3.4375595805529077e-07, "loss": 2.2487, "step": 46490 }, { "epoch": 8.86, "learning_rate": 3.436987607244995e-07, "loss": 2.8692, "step": 46491 }, { "epoch": 8.86, "learning_rate": 3.436415633937083e-07, "loss": 1.5589, "step": 46492 }, { "epoch": 8.86, "learning_rate": 3.435843660629171e-07, "loss": 3.044, "step": 46493 }, { "epoch": 8.86, "learning_rate": 3.4352716873212584e-07, "loss": 1.825, "step": 46494 }, { "epoch": 8.86, "learning_rate": 3.434699714013346e-07, "loss": 2.3691, "step": 46495 }, { "epoch": 8.86, "learning_rate": 3.4341277407054337e-07, "loss": 1.8164, "step": 46496 }, { "epoch": 8.86, "learning_rate": 3.4335557673975217e-07, "loss": 2.2626, "step": 46497 }, { "epoch": 8.86, "learning_rate": 3.432983794089609e-07, "loss": 2.0626, "step": 46498 }, { "epoch": 8.86, "learning_rate": 3.432411820781697e-07, "loss": 0.7498, "step": 46499 }, { "epoch": 8.86, "learning_rate": 3.431839847473785e-07, "loss": 2.8029, "step": 46500 }, { "epoch": 8.86, "learning_rate": 3.4312678741658724e-07, "loss": 1.7577, "step": 46501 }, { "epoch": 8.86, "learning_rate": 3.43069590085796e-07, "loss": 1.9311, "step": 46502 }, { "epoch": 8.86, "learning_rate": 3.4301239275500477e-07, "loss": 1.5337, "step": 46503 }, { "epoch": 8.86, "learning_rate": 3.4295519542421356e-07, "loss": 1.2961, "step": 46504 }, { "epoch": 8.86, "learning_rate": 3.428979980934223e-07, "loss": 1.9032, "step": 46505 }, { "epoch": 8.86, "learning_rate": 3.4284080076263105e-07, "loss": 1.3725, "step": 46506 }, { "epoch": 8.86, "learning_rate": 3.4278360343183984e-07, "loss": 1.2476, "step": 46507 }, { "epoch": 8.86, "learning_rate": 3.4272640610104863e-07, "loss": 1.5285, "step": 46508 }, { "epoch": 8.86, "learning_rate": 3.426692087702574e-07, "loss": 1.8209, "step": 46509 }, { "epoch": 8.86, "learning_rate": 3.4261201143946617e-07, "loss": 1.5823, "step": 46510 }, { "epoch": 8.86, "learning_rate": 3.4255481410867496e-07, "loss": 0.8314, "step": 46511 }, { "epoch": 8.86, "learning_rate": 3.424976167778837e-07, "loss": 1.3113, "step": 46512 }, { "epoch": 8.86, "learning_rate": 3.4244041944709244e-07, "loss": 1.9938, "step": 46513 }, { "epoch": 8.86, "learning_rate": 3.4238322211630124e-07, "loss": 0.6206, "step": 46514 }, { "epoch": 8.86, "learning_rate": 3.4232602478551003e-07, "loss": 1.9818, "step": 46515 }, { "epoch": 8.86, "learning_rate": 3.4226882745471877e-07, "loss": 1.2776, "step": 46516 }, { "epoch": 8.86, "learning_rate": 3.4221163012392757e-07, "loss": 2.0167, "step": 46517 }, { "epoch": 8.86, "learning_rate": 3.421544327931363e-07, "loss": 0.2041, "step": 46518 }, { "epoch": 8.86, "learning_rate": 3.420972354623451e-07, "loss": 1.2296, "step": 46519 }, { "epoch": 8.86, "learning_rate": 3.4204003813155384e-07, "loss": 1.1703, "step": 46520 }, { "epoch": 8.86, "learning_rate": 3.4198284080076264e-07, "loss": 2.3135, "step": 46521 }, { "epoch": 8.86, "learning_rate": 3.4192564346997143e-07, "loss": 1.2884, "step": 46522 }, { "epoch": 8.86, "learning_rate": 3.4186844613918017e-07, "loss": 1.0721, "step": 46523 }, { "epoch": 8.86, "learning_rate": 3.418112488083889e-07, "loss": 0.9963, "step": 46524 }, { "epoch": 8.86, "learning_rate": 3.417540514775977e-07, "loss": 1.9868, "step": 46525 }, { "epoch": 8.86, "learning_rate": 3.416968541468065e-07, "loss": 2.1007, "step": 46526 }, { "epoch": 8.86, "learning_rate": 3.4163965681601524e-07, "loss": 1.2368, "step": 46527 }, { "epoch": 8.86, "learning_rate": 3.4158245948522403e-07, "loss": 2.4152, "step": 46528 }, { "epoch": 8.86, "learning_rate": 3.4152526215443283e-07, "loss": 1.7673, "step": 46529 }, { "epoch": 8.86, "learning_rate": 3.4146806482364157e-07, "loss": 1.3862, "step": 46530 }, { "epoch": 8.86, "learning_rate": 3.414108674928503e-07, "loss": 0.9951, "step": 46531 }, { "epoch": 8.86, "learning_rate": 3.413536701620591e-07, "loss": 1.7818, "step": 46532 }, { "epoch": 8.86, "learning_rate": 3.412964728312679e-07, "loss": 1.8469, "step": 46533 }, { "epoch": 8.86, "learning_rate": 3.4123927550047664e-07, "loss": 1.701, "step": 46534 }, { "epoch": 8.86, "learning_rate": 3.4118207816968543e-07, "loss": 1.7139, "step": 46535 }, { "epoch": 8.86, "learning_rate": 3.4112488083889417e-07, "loss": 1.4225, "step": 46536 }, { "epoch": 8.86, "learning_rate": 3.4106768350810297e-07, "loss": 0.867, "step": 46537 }, { "epoch": 8.86, "learning_rate": 3.410104861773117e-07, "loss": 1.1757, "step": 46538 }, { "epoch": 8.86, "learning_rate": 3.409532888465205e-07, "loss": 2.2777, "step": 46539 }, { "epoch": 8.86, "learning_rate": 3.408960915157293e-07, "loss": 1.4885, "step": 46540 }, { "epoch": 8.86, "learning_rate": 3.408388941849381e-07, "loss": 0.8777, "step": 46541 }, { "epoch": 8.87, "learning_rate": 3.407816968541468e-07, "loss": 1.8203, "step": 46542 }, { "epoch": 8.87, "learning_rate": 3.4072449952335557e-07, "loss": 1.7915, "step": 46543 }, { "epoch": 8.87, "learning_rate": 3.4066730219256436e-07, "loss": 0.3412, "step": 46544 }, { "epoch": 8.87, "learning_rate": 3.406101048617731e-07, "loss": 1.4682, "step": 46545 }, { "epoch": 8.87, "learning_rate": 3.405529075309819e-07, "loss": 1.0691, "step": 46546 }, { "epoch": 8.87, "learning_rate": 3.404957102001907e-07, "loss": 1.632, "step": 46547 }, { "epoch": 8.87, "learning_rate": 3.4043851286939943e-07, "loss": 0.9118, "step": 46548 }, { "epoch": 8.87, "learning_rate": 3.403813155386082e-07, "loss": 2.0651, "step": 46549 }, { "epoch": 8.87, "learning_rate": 3.4032411820781697e-07, "loss": 2.3381, "step": 46550 }, { "epoch": 8.87, "learning_rate": 3.4026692087702576e-07, "loss": 0.5994, "step": 46551 }, { "epoch": 8.87, "learning_rate": 3.4020972354623456e-07, "loss": 1.4096, "step": 46552 }, { "epoch": 8.87, "learning_rate": 3.401525262154433e-07, "loss": 0.7035, "step": 46553 }, { "epoch": 8.87, "learning_rate": 3.4009532888465204e-07, "loss": 1.0684, "step": 46554 }, { "epoch": 8.87, "learning_rate": 3.4003813155386083e-07, "loss": 1.6863, "step": 46555 }, { "epoch": 8.87, "learning_rate": 3.3998093422306957e-07, "loss": 2.6028, "step": 46556 }, { "epoch": 8.87, "learning_rate": 3.3992373689227837e-07, "loss": 1.774, "step": 46557 }, { "epoch": 8.87, "learning_rate": 3.3986653956148716e-07, "loss": 1.0635, "step": 46558 }, { "epoch": 8.87, "learning_rate": 3.3980934223069595e-07, "loss": 1.054, "step": 46559 }, { "epoch": 8.87, "learning_rate": 3.3975214489990464e-07, "loss": 2.0264, "step": 46560 }, { "epoch": 8.87, "learning_rate": 3.3969494756911344e-07, "loss": 0.5712, "step": 46561 }, { "epoch": 8.87, "learning_rate": 3.3963775023832223e-07, "loss": 0.6186, "step": 46562 }, { "epoch": 8.87, "learning_rate": 3.39580552907531e-07, "loss": 1.7111, "step": 46563 }, { "epoch": 8.87, "learning_rate": 3.3952335557673976e-07, "loss": 1.4974, "step": 46564 }, { "epoch": 8.87, "learning_rate": 3.394661582459485e-07, "loss": 1.4846, "step": 46565 }, { "epoch": 8.87, "learning_rate": 3.394089609151573e-07, "loss": 0.7563, "step": 46566 }, { "epoch": 8.87, "learning_rate": 3.3935176358436604e-07, "loss": 2.0885, "step": 46567 }, { "epoch": 8.87, "learning_rate": 3.3929456625357483e-07, "loss": 1.2453, "step": 46568 }, { "epoch": 8.87, "learning_rate": 3.3923736892278363e-07, "loss": 1.4232, "step": 46569 }, { "epoch": 8.87, "learning_rate": 3.391801715919924e-07, "loss": 1.3889, "step": 46570 }, { "epoch": 8.87, "learning_rate": 3.391229742612011e-07, "loss": 1.4237, "step": 46571 }, { "epoch": 8.87, "learning_rate": 3.390657769304099e-07, "loss": 1.7121, "step": 46572 }, { "epoch": 8.87, "learning_rate": 3.390085795996187e-07, "loss": 1.6573, "step": 46573 }, { "epoch": 8.87, "learning_rate": 3.389513822688275e-07, "loss": 1.3625, "step": 46574 }, { "epoch": 8.87, "learning_rate": 3.3889418493803623e-07, "loss": 1.792, "step": 46575 }, { "epoch": 8.87, "learning_rate": 3.38836987607245e-07, "loss": 1.2702, "step": 46576 }, { "epoch": 8.87, "learning_rate": 3.3877979027645377e-07, "loss": 2.3377, "step": 46577 }, { "epoch": 8.87, "learning_rate": 3.387225929456625e-07, "loss": 3.026, "step": 46578 }, { "epoch": 8.87, "learning_rate": 3.386653956148713e-07, "loss": 1.544, "step": 46579 }, { "epoch": 8.87, "learning_rate": 3.386081982840801e-07, "loss": 1.3241, "step": 46580 }, { "epoch": 8.87, "learning_rate": 3.385510009532889e-07, "loss": 1.3524, "step": 46581 }, { "epoch": 8.87, "learning_rate": 3.3849380362249763e-07, "loss": 0.8041, "step": 46582 }, { "epoch": 8.87, "learning_rate": 3.3843660629170637e-07, "loss": 1.4517, "step": 46583 }, { "epoch": 8.87, "learning_rate": 3.3837940896091516e-07, "loss": 0.7516, "step": 46584 }, { "epoch": 8.87, "learning_rate": 3.3832221163012396e-07, "loss": 3.2356, "step": 46585 }, { "epoch": 8.87, "learning_rate": 3.382650142993327e-07, "loss": 1.0991, "step": 46586 }, { "epoch": 8.87, "learning_rate": 3.382078169685415e-07, "loss": 0.8108, "step": 46587 }, { "epoch": 8.87, "learning_rate": 3.381506196377503e-07, "loss": 2.4176, "step": 46588 }, { "epoch": 8.87, "learning_rate": 3.38093422306959e-07, "loss": 0.9889, "step": 46589 }, { "epoch": 8.87, "learning_rate": 3.3803622497616777e-07, "loss": 0.7085, "step": 46590 }, { "epoch": 8.87, "learning_rate": 3.3797902764537656e-07, "loss": 1.9737, "step": 46591 }, { "epoch": 8.87, "learning_rate": 3.3792183031458536e-07, "loss": 0.6761, "step": 46592 }, { "epoch": 8.87, "learning_rate": 3.378646329837941e-07, "loss": 0.7902, "step": 46593 }, { "epoch": 8.88, "learning_rate": 3.378074356530029e-07, "loss": 1.0336, "step": 46594 }, { "epoch": 8.88, "learning_rate": 3.3775023832221163e-07, "loss": 2.8762, "step": 46595 }, { "epoch": 8.88, "learning_rate": 3.376930409914204e-07, "loss": 0.8693, "step": 46596 }, { "epoch": 8.88, "learning_rate": 3.3763584366062917e-07, "loss": 1.4206, "step": 46597 }, { "epoch": 8.88, "learning_rate": 3.3757864632983796e-07, "loss": 2.6845, "step": 46598 }, { "epoch": 8.88, "learning_rate": 3.3752144899904675e-07, "loss": 1.7232, "step": 46599 }, { "epoch": 8.88, "learning_rate": 3.374642516682555e-07, "loss": 2.0965, "step": 46600 }, { "epoch": 8.88, "learning_rate": 3.3740705433746423e-07, "loss": 0.6174, "step": 46601 }, { "epoch": 8.88, "learning_rate": 3.3734985700667303e-07, "loss": 1.008, "step": 46602 }, { "epoch": 8.88, "learning_rate": 3.372926596758818e-07, "loss": 0.8333, "step": 46603 }, { "epoch": 8.88, "learning_rate": 3.3723546234509056e-07, "loss": 1.5037, "step": 46604 }, { "epoch": 8.88, "learning_rate": 3.3717826501429936e-07, "loss": 3.5716, "step": 46605 }, { "epoch": 8.88, "learning_rate": 3.3712106768350815e-07, "loss": 1.5949, "step": 46606 }, { "epoch": 8.88, "learning_rate": 3.370638703527169e-07, "loss": 1.1578, "step": 46607 }, { "epoch": 8.88, "learning_rate": 3.3700667302192563e-07, "loss": 0.9535, "step": 46608 }, { "epoch": 8.88, "learning_rate": 3.3694947569113443e-07, "loss": 1.8423, "step": 46609 }, { "epoch": 8.88, "learning_rate": 3.368922783603432e-07, "loss": 0.2948, "step": 46610 }, { "epoch": 8.88, "learning_rate": 3.3683508102955196e-07, "loss": 1.1729, "step": 46611 }, { "epoch": 8.88, "learning_rate": 3.3677788369876076e-07, "loss": 1.0441, "step": 46612 }, { "epoch": 8.88, "learning_rate": 3.367206863679695e-07, "loss": 1.9383, "step": 46613 }, { "epoch": 8.88, "learning_rate": 3.366634890371783e-07, "loss": 1.8552, "step": 46614 }, { "epoch": 8.88, "learning_rate": 3.3660629170638703e-07, "loss": 1.3691, "step": 46615 }, { "epoch": 8.88, "learning_rate": 3.365490943755958e-07, "loss": 2.4562, "step": 46616 }, { "epoch": 8.88, "learning_rate": 3.364918970448046e-07, "loss": 2.2128, "step": 46617 }, { "epoch": 8.88, "learning_rate": 3.3643469971401336e-07, "loss": 1.0856, "step": 46618 }, { "epoch": 8.88, "learning_rate": 3.363775023832221e-07, "loss": 1.4095, "step": 46619 }, { "epoch": 8.88, "learning_rate": 3.363203050524309e-07, "loss": 1.3281, "step": 46620 }, { "epoch": 8.88, "learning_rate": 3.362631077216397e-07, "loss": 1.2017, "step": 46621 }, { "epoch": 8.88, "learning_rate": 3.3620591039084843e-07, "loss": 1.8743, "step": 46622 }, { "epoch": 8.88, "learning_rate": 3.361487130600572e-07, "loss": 1.6928, "step": 46623 }, { "epoch": 8.88, "learning_rate": 3.36091515729266e-07, "loss": 1.3655, "step": 46624 }, { "epoch": 8.88, "learning_rate": 3.3603431839847476e-07, "loss": 1.0414, "step": 46625 }, { "epoch": 8.88, "learning_rate": 3.359771210676835e-07, "loss": 0.7844, "step": 46626 }, { "epoch": 8.88, "learning_rate": 3.359199237368923e-07, "loss": 1.0479, "step": 46627 }, { "epoch": 8.88, "learning_rate": 3.358627264061011e-07, "loss": 1.6531, "step": 46628 }, { "epoch": 8.88, "learning_rate": 3.3580552907530983e-07, "loss": 1.3128, "step": 46629 }, { "epoch": 8.88, "learning_rate": 3.3574833174451857e-07, "loss": 2.1722, "step": 46630 }, { "epoch": 8.88, "learning_rate": 3.3569113441372736e-07, "loss": 1.2104, "step": 46631 }, { "epoch": 8.88, "learning_rate": 3.3563393708293616e-07, "loss": 2.227, "step": 46632 }, { "epoch": 8.88, "learning_rate": 3.355767397521449e-07, "loss": 2.8105, "step": 46633 }, { "epoch": 8.88, "learning_rate": 3.355195424213537e-07, "loss": 1.7106, "step": 46634 }, { "epoch": 8.88, "learning_rate": 3.354623450905625e-07, "loss": 1.4088, "step": 46635 }, { "epoch": 8.88, "learning_rate": 3.354051477597712e-07, "loss": 0.4735, "step": 46636 }, { "epoch": 8.88, "learning_rate": 3.3534795042897997e-07, "loss": 1.2454, "step": 46637 }, { "epoch": 8.88, "learning_rate": 3.3529075309818876e-07, "loss": 1.6177, "step": 46638 }, { "epoch": 8.88, "learning_rate": 3.3523355576739755e-07, "loss": 0.891, "step": 46639 }, { "epoch": 8.88, "learning_rate": 3.351763584366063e-07, "loss": 2.1206, "step": 46640 }, { "epoch": 8.88, "learning_rate": 3.351191611058151e-07, "loss": 1.6464, "step": 46641 }, { "epoch": 8.88, "learning_rate": 3.3506196377502383e-07, "loss": 1.2228, "step": 46642 }, { "epoch": 8.88, "learning_rate": 3.350047664442326e-07, "loss": 1.7079, "step": 46643 }, { "epoch": 8.88, "learning_rate": 3.3494756911344136e-07, "loss": 1.5262, "step": 46644 }, { "epoch": 8.88, "learning_rate": 3.3489037178265016e-07, "loss": 1.5995, "step": 46645 }, { "epoch": 8.88, "learning_rate": 3.3483317445185895e-07, "loss": 0.777, "step": 46646 }, { "epoch": 8.89, "learning_rate": 3.347759771210677e-07, "loss": 2.2717, "step": 46647 }, { "epoch": 8.89, "learning_rate": 3.3471877979027643e-07, "loss": 0.479, "step": 46648 }, { "epoch": 8.89, "learning_rate": 3.3466158245948523e-07, "loss": 2.1176, "step": 46649 }, { "epoch": 8.89, "learning_rate": 3.34604385128694e-07, "loss": 2.8348, "step": 46650 }, { "epoch": 8.89, "learning_rate": 3.3454718779790276e-07, "loss": 1.1449, "step": 46651 }, { "epoch": 8.89, "learning_rate": 3.3448999046711156e-07, "loss": 1.0257, "step": 46652 }, { "epoch": 8.89, "learning_rate": 3.3443279313632035e-07, "loss": 0.8976, "step": 46653 }, { "epoch": 8.89, "learning_rate": 3.343755958055291e-07, "loss": 0.768, "step": 46654 }, { "epoch": 8.89, "learning_rate": 3.3431839847473783e-07, "loss": 0.7749, "step": 46655 }, { "epoch": 8.89, "learning_rate": 3.342612011439466e-07, "loss": 3.0685, "step": 46656 }, { "epoch": 8.89, "learning_rate": 3.342040038131554e-07, "loss": 1.0151, "step": 46657 }, { "epoch": 8.89, "learning_rate": 3.3414680648236416e-07, "loss": 1.2132, "step": 46658 }, { "epoch": 8.89, "learning_rate": 3.3408960915157295e-07, "loss": 1.2551, "step": 46659 }, { "epoch": 8.89, "learning_rate": 3.340324118207817e-07, "loss": 1.1338, "step": 46660 }, { "epoch": 8.89, "learning_rate": 3.339752144899905e-07, "loss": 2.0091, "step": 46661 }, { "epoch": 8.89, "learning_rate": 3.3391801715919923e-07, "loss": 1.2168, "step": 46662 }, { "epoch": 8.89, "learning_rate": 3.33860819828408e-07, "loss": 1.4768, "step": 46663 }, { "epoch": 8.89, "learning_rate": 3.338036224976168e-07, "loss": 1.3272, "step": 46664 }, { "epoch": 8.89, "learning_rate": 3.3374642516682556e-07, "loss": 2.3693, "step": 46665 }, { "epoch": 8.89, "learning_rate": 3.336892278360343e-07, "loss": 1.6536, "step": 46666 }, { "epoch": 8.89, "learning_rate": 3.336320305052431e-07, "loss": 1.1984, "step": 46667 }, { "epoch": 8.89, "learning_rate": 3.335748331744519e-07, "loss": 0.7793, "step": 46668 }, { "epoch": 8.89, "learning_rate": 3.335176358436606e-07, "loss": 2.0189, "step": 46669 }, { "epoch": 8.89, "learning_rate": 3.334604385128694e-07, "loss": 0.8006, "step": 46670 }, { "epoch": 8.89, "learning_rate": 3.334032411820782e-07, "loss": 1.9133, "step": 46671 }, { "epoch": 8.89, "learning_rate": 3.3334604385128695e-07, "loss": 1.2625, "step": 46672 }, { "epoch": 8.89, "learning_rate": 3.332888465204957e-07, "loss": 2.5759, "step": 46673 }, { "epoch": 8.89, "learning_rate": 3.332316491897045e-07, "loss": 1.6628, "step": 46674 }, { "epoch": 8.89, "learning_rate": 3.331744518589133e-07, "loss": 1.2695, "step": 46675 }, { "epoch": 8.89, "learning_rate": 3.33117254528122e-07, "loss": 1.9949, "step": 46676 }, { "epoch": 8.89, "learning_rate": 3.330600571973308e-07, "loss": 0.6695, "step": 46677 }, { "epoch": 8.89, "learning_rate": 3.3300285986653956e-07, "loss": 0.2541, "step": 46678 }, { "epoch": 8.89, "learning_rate": 3.3294566253574835e-07, "loss": 1.2745, "step": 46679 }, { "epoch": 8.89, "learning_rate": 3.328884652049571e-07, "loss": 1.2873, "step": 46680 }, { "epoch": 8.89, "learning_rate": 3.328312678741659e-07, "loss": 1.8205, "step": 46681 }, { "epoch": 8.89, "learning_rate": 3.327740705433747e-07, "loss": 1.3318, "step": 46682 }, { "epoch": 8.89, "learning_rate": 3.327168732125834e-07, "loss": 1.5493, "step": 46683 }, { "epoch": 8.89, "learning_rate": 3.3265967588179216e-07, "loss": 0.7348, "step": 46684 }, { "epoch": 8.89, "learning_rate": 3.3260247855100096e-07, "loss": 1.3205, "step": 46685 }, { "epoch": 8.89, "learning_rate": 3.3254528122020975e-07, "loss": 1.1297, "step": 46686 }, { "epoch": 8.89, "learning_rate": 3.324880838894185e-07, "loss": 0.9582, "step": 46687 }, { "epoch": 8.89, "learning_rate": 3.324308865586273e-07, "loss": 1.4905, "step": 46688 }, { "epoch": 8.89, "learning_rate": 3.32373689227836e-07, "loss": 1.937, "step": 46689 }, { "epoch": 8.89, "learning_rate": 3.323164918970448e-07, "loss": 0.7248, "step": 46690 }, { "epoch": 8.89, "learning_rate": 3.3225929456625356e-07, "loss": 1.3301, "step": 46691 }, { "epoch": 8.89, "learning_rate": 3.3220209723546235e-07, "loss": 1.8834, "step": 46692 }, { "epoch": 8.89, "learning_rate": 3.3214489990467115e-07, "loss": 1.1686, "step": 46693 }, { "epoch": 8.89, "learning_rate": 3.320877025738799e-07, "loss": 2.688, "step": 46694 }, { "epoch": 8.89, "learning_rate": 3.3203050524308863e-07, "loss": 1.8459, "step": 46695 }, { "epoch": 8.89, "learning_rate": 3.319733079122974e-07, "loss": 3.6775, "step": 46696 }, { "epoch": 8.89, "learning_rate": 3.319161105815062e-07, "loss": 1.5477, "step": 46697 }, { "epoch": 8.89, "learning_rate": 3.3185891325071496e-07, "loss": 1.6855, "step": 46698 }, { "epoch": 8.9, "learning_rate": 3.3180171591992375e-07, "loss": 1.0579, "step": 46699 }, { "epoch": 8.9, "learning_rate": 3.3174451858913255e-07, "loss": 1.2574, "step": 46700 }, { "epoch": 8.9, "learning_rate": 3.316873212583413e-07, "loss": 1.3208, "step": 46701 }, { "epoch": 8.9, "learning_rate": 3.3163012392755003e-07, "loss": 1.5823, "step": 46702 }, { "epoch": 8.9, "learning_rate": 3.315729265967588e-07, "loss": 2.6674, "step": 46703 }, { "epoch": 8.9, "learning_rate": 3.315157292659676e-07, "loss": 0.4141, "step": 46704 }, { "epoch": 8.9, "learning_rate": 3.3145853193517636e-07, "loss": 1.9068, "step": 46705 }, { "epoch": 8.9, "learning_rate": 3.3140133460438515e-07, "loss": 0.2555, "step": 46706 }, { "epoch": 8.9, "learning_rate": 3.313441372735939e-07, "loss": 1.4546, "step": 46707 }, { "epoch": 8.9, "learning_rate": 3.312869399428027e-07, "loss": 0.8082, "step": 46708 }, { "epoch": 8.9, "learning_rate": 3.312297426120114e-07, "loss": 0.4241, "step": 46709 }, { "epoch": 8.9, "learning_rate": 3.311725452812202e-07, "loss": 0.8201, "step": 46710 }, { "epoch": 8.9, "learning_rate": 3.31115347950429e-07, "loss": 1.168, "step": 46711 }, { "epoch": 8.9, "learning_rate": 3.3105815061963775e-07, "loss": 2.5091, "step": 46712 }, { "epoch": 8.9, "learning_rate": 3.310009532888465e-07, "loss": 1.6539, "step": 46713 }, { "epoch": 8.9, "learning_rate": 3.309437559580553e-07, "loss": 2.2777, "step": 46714 }, { "epoch": 8.9, "learning_rate": 3.308865586272641e-07, "loss": 1.3977, "step": 46715 }, { "epoch": 8.9, "learning_rate": 3.308293612964728e-07, "loss": 1.9066, "step": 46716 }, { "epoch": 8.9, "learning_rate": 3.307721639656816e-07, "loss": 2.1261, "step": 46717 }, { "epoch": 8.9, "learning_rate": 3.307149666348904e-07, "loss": 1.3541, "step": 46718 }, { "epoch": 8.9, "learning_rate": 3.3065776930409915e-07, "loss": 2.1384, "step": 46719 }, { "epoch": 8.9, "learning_rate": 3.306005719733079e-07, "loss": 0.8133, "step": 46720 }, { "epoch": 8.9, "learning_rate": 3.305433746425167e-07, "loss": 1.7271, "step": 46721 }, { "epoch": 8.9, "learning_rate": 3.304861773117255e-07, "loss": 1.7933, "step": 46722 }, { "epoch": 8.9, "learning_rate": 3.304289799809342e-07, "loss": 1.7143, "step": 46723 }, { "epoch": 8.9, "learning_rate": 3.30371782650143e-07, "loss": 1.1167, "step": 46724 }, { "epoch": 8.9, "learning_rate": 3.3031458531935176e-07, "loss": 1.2326, "step": 46725 }, { "epoch": 8.9, "learning_rate": 3.3025738798856055e-07, "loss": 1.4515, "step": 46726 }, { "epoch": 8.9, "learning_rate": 3.302001906577693e-07, "loss": 1.1556, "step": 46727 }, { "epoch": 8.9, "learning_rate": 3.301429933269781e-07, "loss": 1.7035, "step": 46728 }, { "epoch": 8.9, "learning_rate": 3.300857959961869e-07, "loss": 1.0105, "step": 46729 }, { "epoch": 8.9, "learning_rate": 3.300285986653956e-07, "loss": 1.4683, "step": 46730 }, { "epoch": 8.9, "learning_rate": 3.2997140133460436e-07, "loss": 1.6283, "step": 46731 }, { "epoch": 8.9, "learning_rate": 3.2991420400381315e-07, "loss": 2.6367, "step": 46732 }, { "epoch": 8.9, "learning_rate": 3.2985700667302195e-07, "loss": 1.5753, "step": 46733 }, { "epoch": 8.9, "learning_rate": 3.297998093422307e-07, "loss": 1.0746, "step": 46734 }, { "epoch": 8.9, "learning_rate": 3.297426120114395e-07, "loss": 1.2177, "step": 46735 }, { "epoch": 8.9, "learning_rate": 3.296854146806483e-07, "loss": 1.4769, "step": 46736 }, { "epoch": 8.9, "learning_rate": 3.29628217349857e-07, "loss": 3.1489, "step": 46737 }, { "epoch": 8.9, "learning_rate": 3.2957102001906576e-07, "loss": 2.0813, "step": 46738 }, { "epoch": 8.9, "learning_rate": 3.2951382268827455e-07, "loss": 2.9404, "step": 46739 }, { "epoch": 8.9, "learning_rate": 3.2945662535748335e-07, "loss": 2.0618, "step": 46740 }, { "epoch": 8.9, "learning_rate": 3.293994280266921e-07, "loss": 0.4702, "step": 46741 }, { "epoch": 8.9, "learning_rate": 3.293422306959009e-07, "loss": 2.3372, "step": 46742 }, { "epoch": 8.9, "learning_rate": 3.292850333651096e-07, "loss": 1.3153, "step": 46743 }, { "epoch": 8.9, "learning_rate": 3.292278360343184e-07, "loss": 2.3334, "step": 46744 }, { "epoch": 8.9, "learning_rate": 3.2917063870352716e-07, "loss": 2.6487, "step": 46745 }, { "epoch": 8.9, "learning_rate": 3.2911344137273595e-07, "loss": 2.0587, "step": 46746 }, { "epoch": 8.9, "learning_rate": 3.2905624404194474e-07, "loss": 1.6231, "step": 46747 }, { "epoch": 8.9, "learning_rate": 3.289990467111535e-07, "loss": 0.306, "step": 46748 }, { "epoch": 8.9, "learning_rate": 3.289418493803622e-07, "loss": 0.6015, "step": 46749 }, { "epoch": 8.9, "learning_rate": 3.28884652049571e-07, "loss": 1.6808, "step": 46750 }, { "epoch": 8.9, "learning_rate": 3.288274547187798e-07, "loss": 1.9302, "step": 46751 }, { "epoch": 8.91, "learning_rate": 3.2877025738798855e-07, "loss": 1.5069, "step": 46752 }, { "epoch": 8.91, "learning_rate": 3.2871306005719735e-07, "loss": 0.5393, "step": 46753 }, { "epoch": 8.91, "learning_rate": 3.286558627264061e-07, "loss": 1.6727, "step": 46754 }, { "epoch": 8.91, "learning_rate": 3.285986653956149e-07, "loss": 1.3313, "step": 46755 }, { "epoch": 8.91, "learning_rate": 3.285414680648236e-07, "loss": 1.0885, "step": 46756 }, { "epoch": 8.91, "learning_rate": 3.284842707340324e-07, "loss": 0.4495, "step": 46757 }, { "epoch": 8.91, "learning_rate": 3.284270734032412e-07, "loss": 1.3265, "step": 46758 }, { "epoch": 8.91, "learning_rate": 3.2836987607244995e-07, "loss": 1.3755, "step": 46759 }, { "epoch": 8.91, "learning_rate": 3.283126787416587e-07, "loss": 1.6819, "step": 46760 }, { "epoch": 8.91, "learning_rate": 3.282554814108675e-07, "loss": 1.5036, "step": 46761 }, { "epoch": 8.91, "learning_rate": 3.281982840800763e-07, "loss": 1.0808, "step": 46762 }, { "epoch": 8.91, "learning_rate": 3.28141086749285e-07, "loss": 0.8996, "step": 46763 }, { "epoch": 8.91, "learning_rate": 3.280838894184938e-07, "loss": 1.6217, "step": 46764 }, { "epoch": 8.91, "learning_rate": 3.280266920877026e-07, "loss": 3.2214, "step": 46765 }, { "epoch": 8.91, "learning_rate": 3.2796949475691135e-07, "loss": 0.4679, "step": 46766 }, { "epoch": 8.91, "learning_rate": 3.279122974261201e-07, "loss": 2.2715, "step": 46767 }, { "epoch": 8.91, "learning_rate": 3.278551000953289e-07, "loss": 1.8891, "step": 46768 }, { "epoch": 8.91, "learning_rate": 3.277979027645377e-07, "loss": 0.5224, "step": 46769 }, { "epoch": 8.91, "learning_rate": 3.277407054337464e-07, "loss": 1.472, "step": 46770 }, { "epoch": 8.91, "learning_rate": 3.276835081029552e-07, "loss": 1.6244, "step": 46771 }, { "epoch": 8.91, "learning_rate": 3.2762631077216395e-07, "loss": 2.319, "step": 46772 }, { "epoch": 8.91, "learning_rate": 3.2756911344137275e-07, "loss": 2.0906, "step": 46773 }, { "epoch": 8.91, "learning_rate": 3.275119161105815e-07, "loss": 1.3471, "step": 46774 }, { "epoch": 8.91, "learning_rate": 3.274547187797903e-07, "loss": 2.0044, "step": 46775 }, { "epoch": 8.91, "learning_rate": 3.273975214489991e-07, "loss": 2.3451, "step": 46776 }, { "epoch": 8.91, "learning_rate": 3.273403241182078e-07, "loss": 0.5864, "step": 46777 }, { "epoch": 8.91, "learning_rate": 3.2728312678741656e-07, "loss": 1.2487, "step": 46778 }, { "epoch": 8.91, "learning_rate": 3.2722592945662535e-07, "loss": 1.6116, "step": 46779 }, { "epoch": 8.91, "learning_rate": 3.2716873212583415e-07, "loss": 0.8179, "step": 46780 }, { "epoch": 8.91, "learning_rate": 3.271115347950429e-07, "loss": 0.4647, "step": 46781 }, { "epoch": 8.91, "learning_rate": 3.270543374642517e-07, "loss": 0.8742, "step": 46782 }, { "epoch": 8.91, "learning_rate": 3.269971401334605e-07, "loss": 3.6032, "step": 46783 }, { "epoch": 8.91, "learning_rate": 3.269399428026692e-07, "loss": 1.378, "step": 46784 }, { "epoch": 8.91, "learning_rate": 3.2688274547187796e-07, "loss": 2.7998, "step": 46785 }, { "epoch": 8.91, "learning_rate": 3.2682554814108675e-07, "loss": 2.3479, "step": 46786 }, { "epoch": 8.91, "learning_rate": 3.2676835081029554e-07, "loss": 2.3302, "step": 46787 }, { "epoch": 8.91, "learning_rate": 3.267111534795043e-07, "loss": 0.3624, "step": 46788 }, { "epoch": 8.91, "learning_rate": 3.266539561487131e-07, "loss": 1.2257, "step": 46789 }, { "epoch": 8.91, "learning_rate": 3.265967588179218e-07, "loss": 2.4703, "step": 46790 }, { "epoch": 8.91, "learning_rate": 3.265395614871306e-07, "loss": 1.9971, "step": 46791 }, { "epoch": 8.91, "learning_rate": 3.2648236415633935e-07, "loss": 0.6951, "step": 46792 }, { "epoch": 8.91, "learning_rate": 3.2642516682554815e-07, "loss": 0.7474, "step": 46793 }, { "epoch": 8.91, "learning_rate": 3.2636796949475694e-07, "loss": 0.7264, "step": 46794 }, { "epoch": 8.91, "learning_rate": 3.2631077216396574e-07, "loss": 1.6401, "step": 46795 }, { "epoch": 8.91, "learning_rate": 3.262535748331744e-07, "loss": 2.8328, "step": 46796 }, { "epoch": 8.91, "learning_rate": 3.261963775023832e-07, "loss": 1.6861, "step": 46797 }, { "epoch": 8.91, "learning_rate": 3.26139180171592e-07, "loss": 1.0298, "step": 46798 }, { "epoch": 8.91, "learning_rate": 3.2608198284080075e-07, "loss": 0.7615, "step": 46799 }, { "epoch": 8.91, "learning_rate": 3.2602478551000955e-07, "loss": 1.6753, "step": 46800 }, { "epoch": 8.91, "learning_rate": 3.2596758817921834e-07, "loss": 1.3231, "step": 46801 }, { "epoch": 8.91, "learning_rate": 3.259103908484271e-07, "loss": 1.9117, "step": 46802 }, { "epoch": 8.91, "learning_rate": 3.258531935176358e-07, "loss": 0.9549, "step": 46803 }, { "epoch": 8.92, "learning_rate": 3.257959961868446e-07, "loss": 0.7583, "step": 46804 }, { "epoch": 8.92, "learning_rate": 3.257387988560534e-07, "loss": 1.9737, "step": 46805 }, { "epoch": 8.92, "learning_rate": 3.2568160152526215e-07, "loss": 2.8349, "step": 46806 }, { "epoch": 8.92, "learning_rate": 3.256244041944709e-07, "loss": 1.5617, "step": 46807 }, { "epoch": 8.92, "learning_rate": 3.255672068636797e-07, "loss": 1.4755, "step": 46808 }, { "epoch": 8.92, "learning_rate": 3.255100095328885e-07, "loss": 1.3436, "step": 46809 }, { "epoch": 8.92, "learning_rate": 3.254528122020972e-07, "loss": 1.3328, "step": 46810 }, { "epoch": 8.92, "learning_rate": 3.25395614871306e-07, "loss": 0.9786, "step": 46811 }, { "epoch": 8.92, "learning_rate": 3.253384175405148e-07, "loss": 1.6901, "step": 46812 }, { "epoch": 8.92, "learning_rate": 3.2528122020972355e-07, "loss": 0.7775, "step": 46813 }, { "epoch": 8.92, "learning_rate": 3.252240228789323e-07, "loss": 3.8607, "step": 46814 }, { "epoch": 8.92, "learning_rate": 3.251668255481411e-07, "loss": 1.1935, "step": 46815 }, { "epoch": 8.92, "learning_rate": 3.251096282173499e-07, "loss": 3.042, "step": 46816 }, { "epoch": 8.92, "learning_rate": 3.250524308865586e-07, "loss": 1.9271, "step": 46817 }, { "epoch": 8.92, "learning_rate": 3.249952335557674e-07, "loss": 1.6903, "step": 46818 }, { "epoch": 8.92, "learning_rate": 3.2493803622497615e-07, "loss": 1.3574, "step": 46819 }, { "epoch": 8.92, "learning_rate": 3.2488083889418495e-07, "loss": 1.7564, "step": 46820 }, { "epoch": 8.92, "learning_rate": 3.248236415633937e-07, "loss": 1.1569, "step": 46821 }, { "epoch": 8.92, "learning_rate": 3.247664442326025e-07, "loss": 1.5852, "step": 46822 }, { "epoch": 8.92, "learning_rate": 3.247092469018113e-07, "loss": 1.7192, "step": 46823 }, { "epoch": 8.92, "learning_rate": 3.2465204957102007e-07, "loss": 0.965, "step": 46824 }, { "epoch": 8.92, "learning_rate": 3.2459485224022876e-07, "loss": 3.1355, "step": 46825 }, { "epoch": 8.92, "learning_rate": 3.2453765490943755e-07, "loss": 1.4358, "step": 46826 }, { "epoch": 8.92, "learning_rate": 3.2448045757864634e-07, "loss": 0.5735, "step": 46827 }, { "epoch": 8.92, "learning_rate": 3.244232602478551e-07, "loss": 0.9624, "step": 46828 }, { "epoch": 8.92, "learning_rate": 3.243660629170639e-07, "loss": 1.6428, "step": 46829 }, { "epoch": 8.92, "learning_rate": 3.2430886558627267e-07, "loss": 2.2411, "step": 46830 }, { "epoch": 8.92, "learning_rate": 3.242516682554814e-07, "loss": 3.1997, "step": 46831 }, { "epoch": 8.92, "learning_rate": 3.2419447092469015e-07, "loss": 1.6755, "step": 46832 }, { "epoch": 8.92, "learning_rate": 3.2413727359389895e-07, "loss": 1.2982, "step": 46833 }, { "epoch": 8.92, "learning_rate": 3.2408007626310774e-07, "loss": 0.7104, "step": 46834 }, { "epoch": 8.92, "learning_rate": 3.2402287893231654e-07, "loss": 1.9732, "step": 46835 }, { "epoch": 8.92, "learning_rate": 3.239656816015253e-07, "loss": 2.4934, "step": 46836 }, { "epoch": 8.92, "learning_rate": 3.23908484270734e-07, "loss": 2.2278, "step": 46837 }, { "epoch": 8.92, "learning_rate": 3.238512869399428e-07, "loss": 3.2022, "step": 46838 }, { "epoch": 8.92, "learning_rate": 3.2379408960915155e-07, "loss": 1.7823, "step": 46839 }, { "epoch": 8.92, "learning_rate": 3.2373689227836035e-07, "loss": 0.8256, "step": 46840 }, { "epoch": 8.92, "learning_rate": 3.2367969494756914e-07, "loss": 0.3902, "step": 46841 }, { "epoch": 8.92, "learning_rate": 3.2362249761677793e-07, "loss": 0.438, "step": 46842 }, { "epoch": 8.92, "learning_rate": 3.235653002859866e-07, "loss": 0.955, "step": 46843 }, { "epoch": 8.92, "learning_rate": 3.235081029551954e-07, "loss": 1.9062, "step": 46844 }, { "epoch": 8.92, "learning_rate": 3.234509056244042e-07, "loss": 2.1897, "step": 46845 }, { "epoch": 8.92, "learning_rate": 3.23393708293613e-07, "loss": 0.9656, "step": 46846 }, { "epoch": 8.92, "learning_rate": 3.2333651096282174e-07, "loss": 1.9489, "step": 46847 }, { "epoch": 8.92, "learning_rate": 3.2327931363203054e-07, "loss": 2.0314, "step": 46848 }, { "epoch": 8.92, "learning_rate": 3.232221163012393e-07, "loss": 3.4178, "step": 46849 }, { "epoch": 8.92, "learning_rate": 3.23164918970448e-07, "loss": 3.2267, "step": 46850 }, { "epoch": 8.92, "learning_rate": 3.231077216396568e-07, "loss": 3.5818, "step": 46851 }, { "epoch": 8.92, "learning_rate": 3.230505243088656e-07, "loss": 3.0785, "step": 46852 }, { "epoch": 8.92, "learning_rate": 3.229933269780744e-07, "loss": 2.0092, "step": 46853 }, { "epoch": 8.92, "learning_rate": 3.2293612964728314e-07, "loss": 0.8663, "step": 46854 }, { "epoch": 8.92, "learning_rate": 3.228789323164919e-07, "loss": 1.843, "step": 46855 }, { "epoch": 8.92, "learning_rate": 3.228217349857007e-07, "loss": 1.4364, "step": 46856 }, { "epoch": 8.93, "learning_rate": 3.2276453765490947e-07, "loss": 1.9383, "step": 46857 }, { "epoch": 8.93, "learning_rate": 3.227073403241182e-07, "loss": 2.551, "step": 46858 }, { "epoch": 8.93, "learning_rate": 3.22650142993327e-07, "loss": 1.7534, "step": 46859 }, { "epoch": 8.93, "learning_rate": 3.225929456625358e-07, "loss": 2.655, "step": 46860 }, { "epoch": 8.93, "learning_rate": 3.225357483317445e-07, "loss": 1.0226, "step": 46861 }, { "epoch": 8.93, "learning_rate": 3.224785510009533e-07, "loss": 1.0932, "step": 46862 }, { "epoch": 8.93, "learning_rate": 3.224213536701621e-07, "loss": 2.0522, "step": 46863 }, { "epoch": 8.93, "learning_rate": 3.2236415633937087e-07, "loss": 1.122, "step": 46864 }, { "epoch": 8.93, "learning_rate": 3.223069590085796e-07, "loss": 1.8637, "step": 46865 }, { "epoch": 8.93, "learning_rate": 3.222497616777884e-07, "loss": 1.3571, "step": 46866 }, { "epoch": 8.93, "learning_rate": 3.2219256434699714e-07, "loss": 0.5354, "step": 46867 }, { "epoch": 8.93, "learning_rate": 3.2213536701620594e-07, "loss": 2.1203, "step": 46868 }, { "epoch": 8.93, "learning_rate": 3.220781696854147e-07, "loss": 2.0877, "step": 46869 }, { "epoch": 8.93, "learning_rate": 3.2202097235462347e-07, "loss": 1.0188, "step": 46870 }, { "epoch": 8.93, "learning_rate": 3.2196377502383227e-07, "loss": 1.7846, "step": 46871 }, { "epoch": 8.93, "learning_rate": 3.2190657769304095e-07, "loss": 1.8319, "step": 46872 }, { "epoch": 8.93, "learning_rate": 3.2184938036224975e-07, "loss": 1.3478, "step": 46873 }, { "epoch": 8.93, "learning_rate": 3.2179218303145854e-07, "loss": 1.2513, "step": 46874 }, { "epoch": 8.93, "learning_rate": 3.2173498570066734e-07, "loss": 1.7888, "step": 46875 }, { "epoch": 8.93, "learning_rate": 3.216777883698761e-07, "loss": 1.2989, "step": 46876 }, { "epoch": 8.93, "learning_rate": 3.2162059103908487e-07, "loss": 1.0845, "step": 46877 }, { "epoch": 8.93, "learning_rate": 3.215633937082936e-07, "loss": 0.6873, "step": 46878 }, { "epoch": 8.93, "learning_rate": 3.215061963775024e-07, "loss": 0.6032, "step": 46879 }, { "epoch": 8.93, "learning_rate": 3.2144899904671115e-07, "loss": 1.3619, "step": 46880 }, { "epoch": 8.93, "learning_rate": 3.2139180171591994e-07, "loss": 1.9802, "step": 46881 }, { "epoch": 8.93, "learning_rate": 3.2133460438512873e-07, "loss": 1.1969, "step": 46882 }, { "epoch": 8.93, "learning_rate": 3.212774070543375e-07, "loss": 1.3531, "step": 46883 }, { "epoch": 8.93, "learning_rate": 3.212202097235462e-07, "loss": 1.0047, "step": 46884 }, { "epoch": 8.93, "learning_rate": 3.21163012392755e-07, "loss": 1.5695, "step": 46885 }, { "epoch": 8.93, "learning_rate": 3.211058150619638e-07, "loss": 3.1175, "step": 46886 }, { "epoch": 8.93, "learning_rate": 3.2104861773117254e-07, "loss": 1.4478, "step": 46887 }, { "epoch": 8.93, "learning_rate": 3.2099142040038134e-07, "loss": 1.5076, "step": 46888 }, { "epoch": 8.93, "learning_rate": 3.2093422306959013e-07, "loss": 1.3814, "step": 46889 }, { "epoch": 8.93, "learning_rate": 3.2087702573879887e-07, "loss": 2.0244, "step": 46890 }, { "epoch": 8.93, "learning_rate": 3.208198284080076e-07, "loss": 0.8466, "step": 46891 }, { "epoch": 8.93, "learning_rate": 3.207626310772164e-07, "loss": 2.4657, "step": 46892 }, { "epoch": 8.93, "learning_rate": 3.207054337464252e-07, "loss": 1.0604, "step": 46893 }, { "epoch": 8.93, "learning_rate": 3.2064823641563394e-07, "loss": 1.8442, "step": 46894 }, { "epoch": 8.93, "learning_rate": 3.2059103908484274e-07, "loss": 2.3872, "step": 46895 }, { "epoch": 8.93, "learning_rate": 3.205338417540515e-07, "loss": 2.2118, "step": 46896 }, { "epoch": 8.93, "learning_rate": 3.2047664442326027e-07, "loss": 1.2598, "step": 46897 }, { "epoch": 8.93, "learning_rate": 3.20419447092469e-07, "loss": 2.0444, "step": 46898 }, { "epoch": 8.93, "learning_rate": 3.203622497616778e-07, "loss": 1.6019, "step": 46899 }, { "epoch": 8.93, "learning_rate": 3.203050524308866e-07, "loss": 1.2228, "step": 46900 }, { "epoch": 8.93, "learning_rate": 3.2024785510009534e-07, "loss": 0.788, "step": 46901 }, { "epoch": 8.93, "learning_rate": 3.201906577693041e-07, "loss": 1.0119, "step": 46902 }, { "epoch": 8.93, "learning_rate": 3.201334604385129e-07, "loss": 1.1684, "step": 46903 }, { "epoch": 8.93, "learning_rate": 3.2007626310772167e-07, "loss": 2.2562, "step": 46904 }, { "epoch": 8.93, "learning_rate": 3.200190657769304e-07, "loss": 0.7452, "step": 46905 }, { "epoch": 8.93, "learning_rate": 3.199618684461392e-07, "loss": 1.1959, "step": 46906 }, { "epoch": 8.93, "learning_rate": 3.19904671115348e-07, "loss": 2.1176, "step": 46907 }, { "epoch": 8.93, "learning_rate": 3.1984747378455674e-07, "loss": 2.3899, "step": 46908 }, { "epoch": 8.94, "learning_rate": 3.197902764537655e-07, "loss": 1.3802, "step": 46909 }, { "epoch": 8.94, "learning_rate": 3.1973307912297427e-07, "loss": 1.3766, "step": 46910 }, { "epoch": 8.94, "learning_rate": 3.1967588179218307e-07, "loss": 1.0504, "step": 46911 }, { "epoch": 8.94, "learning_rate": 3.196186844613918e-07, "loss": 0.8975, "step": 46912 }, { "epoch": 8.94, "learning_rate": 3.195614871306006e-07, "loss": 2.3677, "step": 46913 }, { "epoch": 8.94, "learning_rate": 3.1950428979980934e-07, "loss": 1.75, "step": 46914 }, { "epoch": 8.94, "learning_rate": 3.1944709246901814e-07, "loss": 0.4995, "step": 46915 }, { "epoch": 8.94, "learning_rate": 3.193898951382269e-07, "loss": 2.5371, "step": 46916 }, { "epoch": 8.94, "learning_rate": 3.1933269780743567e-07, "loss": 1.6411, "step": 46917 }, { "epoch": 8.94, "learning_rate": 3.1927550047664446e-07, "loss": 1.2873, "step": 46918 }, { "epoch": 8.94, "learning_rate": 3.192183031458532e-07, "loss": 1.6847, "step": 46919 }, { "epoch": 8.94, "learning_rate": 3.1916110581506195e-07, "loss": 3.8353, "step": 46920 }, { "epoch": 8.94, "learning_rate": 3.1910390848427074e-07, "loss": 1.8272, "step": 46921 }, { "epoch": 8.94, "learning_rate": 3.1904671115347953e-07, "loss": 2.5348, "step": 46922 }, { "epoch": 8.94, "learning_rate": 3.189895138226883e-07, "loss": 1.1471, "step": 46923 }, { "epoch": 8.94, "learning_rate": 3.1893231649189707e-07, "loss": 0.9216, "step": 46924 }, { "epoch": 8.94, "learning_rate": 3.1887511916110586e-07, "loss": 1.601, "step": 46925 }, { "epoch": 8.94, "learning_rate": 3.188179218303146e-07, "loss": 1.9169, "step": 46926 }, { "epoch": 8.94, "learning_rate": 3.1876072449952334e-07, "loss": 0.4092, "step": 46927 }, { "epoch": 8.94, "learning_rate": 3.1870352716873214e-07, "loss": 1.3099, "step": 46928 }, { "epoch": 8.94, "learning_rate": 3.1864632983794093e-07, "loss": 1.7092, "step": 46929 }, { "epoch": 8.94, "learning_rate": 3.1858913250714967e-07, "loss": 1.036, "step": 46930 }, { "epoch": 8.94, "learning_rate": 3.185319351763584e-07, "loss": 0.713, "step": 46931 }, { "epoch": 8.94, "learning_rate": 3.184747378455672e-07, "loss": 1.367, "step": 46932 }, { "epoch": 8.94, "learning_rate": 3.18417540514776e-07, "loss": 1.9542, "step": 46933 }, { "epoch": 8.94, "learning_rate": 3.1836034318398474e-07, "loss": 0.8799, "step": 46934 }, { "epoch": 8.94, "learning_rate": 3.1830314585319354e-07, "loss": 1.406, "step": 46935 }, { "epoch": 8.94, "learning_rate": 3.1824594852240233e-07, "loss": 1.1319, "step": 46936 }, { "epoch": 8.94, "learning_rate": 3.1818875119161107e-07, "loss": 1.3685, "step": 46937 }, { "epoch": 8.94, "learning_rate": 3.181315538608198e-07, "loss": 2.3055, "step": 46938 }, { "epoch": 8.94, "learning_rate": 3.180743565300286e-07, "loss": 1.6229, "step": 46939 }, { "epoch": 8.94, "learning_rate": 3.180171591992374e-07, "loss": 0.9851, "step": 46940 }, { "epoch": 8.94, "learning_rate": 3.1795996186844614e-07, "loss": 1.5357, "step": 46941 }, { "epoch": 8.94, "learning_rate": 3.1790276453765493e-07, "loss": 3.005, "step": 46942 }, { "epoch": 8.94, "learning_rate": 3.178455672068637e-07, "loss": 1.57, "step": 46943 }, { "epoch": 8.94, "learning_rate": 3.1778836987607247e-07, "loss": 0.7045, "step": 46944 }, { "epoch": 8.94, "learning_rate": 3.177311725452812e-07, "loss": 2.4502, "step": 46945 }, { "epoch": 8.94, "learning_rate": 3.1767397521449e-07, "loss": 1.6984, "step": 46946 }, { "epoch": 8.94, "learning_rate": 3.176167778836988e-07, "loss": 0.9432, "step": 46947 }, { "epoch": 8.94, "learning_rate": 3.1755958055290754e-07, "loss": 1.2844, "step": 46948 }, { "epoch": 8.94, "learning_rate": 3.175023832221163e-07, "loss": 0.9404, "step": 46949 }, { "epoch": 8.94, "learning_rate": 3.1744518589132507e-07, "loss": 0.8894, "step": 46950 }, { "epoch": 8.94, "learning_rate": 3.1738798856053387e-07, "loss": 0.2186, "step": 46951 }, { "epoch": 8.94, "learning_rate": 3.173307912297426e-07, "loss": 1.7984, "step": 46952 }, { "epoch": 8.94, "learning_rate": 3.172735938989514e-07, "loss": 1.9516, "step": 46953 }, { "epoch": 8.94, "learning_rate": 3.172163965681602e-07, "loss": 2.2106, "step": 46954 }, { "epoch": 8.94, "learning_rate": 3.1715919923736893e-07, "loss": 2.103, "step": 46955 }, { "epoch": 8.94, "learning_rate": 3.171020019065777e-07, "loss": 1.416, "step": 46956 }, { "epoch": 8.94, "learning_rate": 3.1704480457578647e-07, "loss": 1.2103, "step": 46957 }, { "epoch": 8.94, "learning_rate": 3.1698760724499526e-07, "loss": 1.6306, "step": 46958 }, { "epoch": 8.94, "learning_rate": 3.16930409914204e-07, "loss": 1.543, "step": 46959 }, { "epoch": 8.94, "learning_rate": 3.168732125834128e-07, "loss": 0.9701, "step": 46960 }, { "epoch": 8.94, "learning_rate": 3.1681601525262154e-07, "loss": 2.3873, "step": 46961 }, { "epoch": 8.95, "learning_rate": 3.1675881792183033e-07, "loss": 0.3117, "step": 46962 }, { "epoch": 8.95, "learning_rate": 3.167016205910391e-07, "loss": 1.4559, "step": 46963 }, { "epoch": 8.95, "learning_rate": 3.1664442326024787e-07, "loss": 2.2625, "step": 46964 }, { "epoch": 8.95, "learning_rate": 3.1658722592945666e-07, "loss": 1.6021, "step": 46965 }, { "epoch": 8.95, "learning_rate": 3.165300285986654e-07, "loss": 2.1422, "step": 46966 }, { "epoch": 8.95, "learning_rate": 3.1647283126787414e-07, "loss": 1.0107, "step": 46967 }, { "epoch": 8.95, "learning_rate": 3.1641563393708294e-07, "loss": 0.9347, "step": 46968 }, { "epoch": 8.95, "learning_rate": 3.1635843660629173e-07, "loss": 1.647, "step": 46969 }, { "epoch": 8.95, "learning_rate": 3.1630123927550047e-07, "loss": 2.5457, "step": 46970 }, { "epoch": 8.95, "learning_rate": 3.1624404194470927e-07, "loss": 0.4609, "step": 46971 }, { "epoch": 8.95, "learning_rate": 3.1618684461391806e-07, "loss": 1.0062, "step": 46972 }, { "epoch": 8.95, "learning_rate": 3.161296472831268e-07, "loss": 1.5697, "step": 46973 }, { "epoch": 8.95, "learning_rate": 3.1607244995233554e-07, "loss": 1.4557, "step": 46974 }, { "epoch": 8.95, "learning_rate": 3.1601525262154433e-07, "loss": 0.3704, "step": 46975 }, { "epoch": 8.95, "learning_rate": 3.1595805529075313e-07, "loss": 0.3092, "step": 46976 }, { "epoch": 8.95, "learning_rate": 3.1590085795996187e-07, "loss": 1.0479, "step": 46977 }, { "epoch": 8.95, "learning_rate": 3.1584366062917066e-07, "loss": 0.9093, "step": 46978 }, { "epoch": 8.95, "learning_rate": 3.157864632983794e-07, "loss": 1.8449, "step": 46979 }, { "epoch": 8.95, "learning_rate": 3.157292659675882e-07, "loss": 1.796, "step": 46980 }, { "epoch": 8.95, "learning_rate": 3.1567206863679694e-07, "loss": 1.6252, "step": 46981 }, { "epoch": 8.95, "learning_rate": 3.1561487130600573e-07, "loss": 1.619, "step": 46982 }, { "epoch": 8.95, "learning_rate": 3.1555767397521453e-07, "loss": 1.7131, "step": 46983 }, { "epoch": 8.95, "learning_rate": 3.1550047664442327e-07, "loss": 0.8228, "step": 46984 }, { "epoch": 8.95, "learning_rate": 3.15443279313632e-07, "loss": 1.5687, "step": 46985 }, { "epoch": 8.95, "learning_rate": 3.153860819828408e-07, "loss": 1.1294, "step": 46986 }, { "epoch": 8.95, "learning_rate": 3.153288846520496e-07, "loss": 1.5616, "step": 46987 }, { "epoch": 8.95, "learning_rate": 3.1527168732125834e-07, "loss": 1.7465, "step": 46988 }, { "epoch": 8.95, "learning_rate": 3.1521448999046713e-07, "loss": 0.4226, "step": 46989 }, { "epoch": 8.95, "learning_rate": 3.1515729265967587e-07, "loss": 1.379, "step": 46990 }, { "epoch": 8.95, "learning_rate": 3.1510009532888467e-07, "loss": 1.7642, "step": 46991 }, { "epoch": 8.95, "learning_rate": 3.150428979980934e-07, "loss": 1.115, "step": 46992 }, { "epoch": 8.95, "learning_rate": 3.149857006673022e-07, "loss": 2.0432, "step": 46993 }, { "epoch": 8.95, "learning_rate": 3.14928503336511e-07, "loss": 0.3788, "step": 46994 }, { "epoch": 8.95, "learning_rate": 3.1487130600571973e-07, "loss": 2.9264, "step": 46995 }, { "epoch": 8.95, "learning_rate": 3.148141086749285e-07, "loss": 1.1389, "step": 46996 }, { "epoch": 8.95, "learning_rate": 3.1475691134413727e-07, "loss": 0.8439, "step": 46997 }, { "epoch": 8.95, "learning_rate": 3.1469971401334606e-07, "loss": 1.5364, "step": 46998 }, { "epoch": 8.95, "learning_rate": 3.146425166825548e-07, "loss": 1.5796, "step": 46999 }, { "epoch": 8.95, "learning_rate": 3.145853193517636e-07, "loss": 1.6312, "step": 47000 }, { "epoch": 8.95, "learning_rate": 3.145281220209724e-07, "loss": 2.9984, "step": 47001 }, { "epoch": 8.95, "learning_rate": 3.1447092469018113e-07, "loss": 0.6778, "step": 47002 }, { "epoch": 8.95, "learning_rate": 3.144137273593899e-07, "loss": 1.3189, "step": 47003 }, { "epoch": 8.95, "learning_rate": 3.1435653002859867e-07, "loss": 1.1654, "step": 47004 }, { "epoch": 8.95, "learning_rate": 3.1429933269780746e-07, "loss": 1.1704, "step": 47005 }, { "epoch": 8.95, "learning_rate": 3.142421353670162e-07, "loss": 0.5956, "step": 47006 }, { "epoch": 8.95, "learning_rate": 3.14184938036225e-07, "loss": 1.9638, "step": 47007 }, { "epoch": 8.95, "learning_rate": 3.1412774070543374e-07, "loss": 0.2492, "step": 47008 }, { "epoch": 8.95, "learning_rate": 3.1407054337464253e-07, "loss": 2.4321, "step": 47009 }, { "epoch": 8.95, "learning_rate": 3.1401334604385127e-07, "loss": 0.5683, "step": 47010 }, { "epoch": 8.95, "learning_rate": 3.1395614871306007e-07, "loss": 0.6759, "step": 47011 }, { "epoch": 8.95, "learning_rate": 3.1389895138226886e-07, "loss": 1.5168, "step": 47012 }, { "epoch": 8.95, "learning_rate": 3.138417540514776e-07, "loss": 2.653, "step": 47013 }, { "epoch": 8.96, "learning_rate": 3.1378455672068634e-07, "loss": 3.2879, "step": 47014 }, { "epoch": 8.96, "learning_rate": 3.1372735938989513e-07, "loss": 1.0839, "step": 47015 }, { "epoch": 8.96, "learning_rate": 3.1367016205910393e-07, "loss": 1.1376, "step": 47016 }, { "epoch": 8.96, "learning_rate": 3.1361296472831267e-07, "loss": 3.5795, "step": 47017 }, { "epoch": 8.96, "learning_rate": 3.1355576739752146e-07, "loss": 1.5563, "step": 47018 }, { "epoch": 8.96, "learning_rate": 3.1349857006673026e-07, "loss": 0.6899, "step": 47019 }, { "epoch": 8.96, "learning_rate": 3.13441372735939e-07, "loss": 2.0775, "step": 47020 }, { "epoch": 8.96, "learning_rate": 3.1338417540514774e-07, "loss": 2.1622, "step": 47021 }, { "epoch": 8.96, "learning_rate": 3.1332697807435653e-07, "loss": 1.8382, "step": 47022 }, { "epoch": 8.96, "learning_rate": 3.132697807435653e-07, "loss": 0.7163, "step": 47023 }, { "epoch": 8.96, "learning_rate": 3.1321258341277407e-07, "loss": 1.4012, "step": 47024 }, { "epoch": 8.96, "learning_rate": 3.1315538608198286e-07, "loss": 1.3506, "step": 47025 }, { "epoch": 8.96, "learning_rate": 3.130981887511916e-07, "loss": 0.613, "step": 47026 }, { "epoch": 8.96, "learning_rate": 3.130409914204004e-07, "loss": 1.7186, "step": 47027 }, { "epoch": 8.96, "learning_rate": 3.1298379408960914e-07, "loss": 0.9839, "step": 47028 }, { "epoch": 8.96, "learning_rate": 3.1292659675881793e-07, "loss": 1.6536, "step": 47029 }, { "epoch": 8.96, "learning_rate": 3.128693994280267e-07, "loss": 1.1506, "step": 47030 }, { "epoch": 8.96, "learning_rate": 3.1281220209723547e-07, "loss": 1.7952, "step": 47031 }, { "epoch": 8.96, "learning_rate": 3.127550047664442e-07, "loss": 1.1362, "step": 47032 }, { "epoch": 8.96, "learning_rate": 3.12697807435653e-07, "loss": 1.3901, "step": 47033 }, { "epoch": 8.96, "learning_rate": 3.126406101048618e-07, "loss": 1.8818, "step": 47034 }, { "epoch": 8.96, "learning_rate": 3.1258341277407053e-07, "loss": 0.944, "step": 47035 }, { "epoch": 8.96, "learning_rate": 3.1252621544327933e-07, "loss": 1.5819, "step": 47036 }, { "epoch": 8.96, "learning_rate": 3.124690181124881e-07, "loss": 1.8006, "step": 47037 }, { "epoch": 8.96, "learning_rate": 3.1241182078169686e-07, "loss": 1.2842, "step": 47038 }, { "epoch": 8.96, "learning_rate": 3.123546234509056e-07, "loss": 1.0556, "step": 47039 }, { "epoch": 8.96, "learning_rate": 3.122974261201144e-07, "loss": 1.1148, "step": 47040 }, { "epoch": 8.96, "learning_rate": 3.122402287893232e-07, "loss": 1.416, "step": 47041 }, { "epoch": 8.96, "learning_rate": 3.1218303145853193e-07, "loss": 1.3246, "step": 47042 }, { "epoch": 8.96, "learning_rate": 3.121258341277407e-07, "loss": 2.6177, "step": 47043 }, { "epoch": 8.96, "learning_rate": 3.1206863679694947e-07, "loss": 1.5604, "step": 47044 }, { "epoch": 8.96, "learning_rate": 3.1201143946615826e-07, "loss": 1.0328, "step": 47045 }, { "epoch": 8.96, "learning_rate": 3.11954242135367e-07, "loss": 1.531, "step": 47046 }, { "epoch": 8.96, "learning_rate": 3.118970448045758e-07, "loss": 3.0719, "step": 47047 }, { "epoch": 8.96, "learning_rate": 3.118398474737846e-07, "loss": 2.126, "step": 47048 }, { "epoch": 8.96, "learning_rate": 3.1178265014299333e-07, "loss": 2.6891, "step": 47049 }, { "epoch": 8.96, "learning_rate": 3.1172545281220207e-07, "loss": 0.94, "step": 47050 }, { "epoch": 8.96, "learning_rate": 3.1166825548141087e-07, "loss": 1.8527, "step": 47051 }, { "epoch": 8.96, "learning_rate": 3.1161105815061966e-07, "loss": 1.057, "step": 47052 }, { "epoch": 8.96, "learning_rate": 3.115538608198284e-07, "loss": 2.0686, "step": 47053 }, { "epoch": 8.96, "learning_rate": 3.114966634890372e-07, "loss": 3.1101, "step": 47054 }, { "epoch": 8.96, "learning_rate": 3.1143946615824593e-07, "loss": 0.9873, "step": 47055 }, { "epoch": 8.96, "learning_rate": 3.1138226882745473e-07, "loss": 0.5366, "step": 47056 }, { "epoch": 8.96, "learning_rate": 3.1132507149666347e-07, "loss": 1.1038, "step": 47057 }, { "epoch": 8.96, "learning_rate": 3.1126787416587226e-07, "loss": 1.4024, "step": 47058 }, { "epoch": 8.96, "learning_rate": 3.1121067683508106e-07, "loss": 1.0424, "step": 47059 }, { "epoch": 8.96, "learning_rate": 3.111534795042898e-07, "loss": 1.6281, "step": 47060 }, { "epoch": 8.96, "learning_rate": 3.1109628217349854e-07, "loss": 1.7748, "step": 47061 }, { "epoch": 8.96, "learning_rate": 3.1103908484270733e-07, "loss": 1.171, "step": 47062 }, { "epoch": 8.96, "learning_rate": 3.109818875119161e-07, "loss": 1.8181, "step": 47063 }, { "epoch": 8.96, "learning_rate": 3.1092469018112487e-07, "loss": 1.5245, "step": 47064 }, { "epoch": 8.96, "learning_rate": 3.1086749285033366e-07, "loss": 1.3868, "step": 47065 }, { "epoch": 8.96, "learning_rate": 3.1081029551954245e-07, "loss": 2.0994, "step": 47066 }, { "epoch": 8.97, "learning_rate": 3.107530981887512e-07, "loss": 0.9723, "step": 47067 }, { "epoch": 8.97, "learning_rate": 3.1069590085795994e-07, "loss": 2.4096, "step": 47068 }, { "epoch": 8.97, "learning_rate": 3.1063870352716873e-07, "loss": 1.9341, "step": 47069 }, { "epoch": 8.97, "learning_rate": 3.105815061963775e-07, "loss": 1.3216, "step": 47070 }, { "epoch": 8.97, "learning_rate": 3.1052430886558626e-07, "loss": 1.22, "step": 47071 }, { "epoch": 8.97, "learning_rate": 3.1046711153479506e-07, "loss": 1.7063, "step": 47072 }, { "epoch": 8.97, "learning_rate": 3.104099142040038e-07, "loss": 1.3332, "step": 47073 }, { "epoch": 8.97, "learning_rate": 3.103527168732126e-07, "loss": 1.2615, "step": 47074 }, { "epoch": 8.97, "learning_rate": 3.1029551954242133e-07, "loss": 0.6168, "step": 47075 }, { "epoch": 8.97, "learning_rate": 3.1023832221163013e-07, "loss": 3.0828, "step": 47076 }, { "epoch": 8.97, "learning_rate": 3.101811248808389e-07, "loss": 0.8776, "step": 47077 }, { "epoch": 8.97, "learning_rate": 3.101239275500477e-07, "loss": 1.2883, "step": 47078 }, { "epoch": 8.97, "learning_rate": 3.100667302192564e-07, "loss": 2.5548, "step": 47079 }, { "epoch": 8.97, "learning_rate": 3.100095328884652e-07, "loss": 2.0348, "step": 47080 }, { "epoch": 8.97, "learning_rate": 3.09952335557674e-07, "loss": 2.7318, "step": 47081 }, { "epoch": 8.97, "learning_rate": 3.0989513822688273e-07, "loss": 0.5916, "step": 47082 }, { "epoch": 8.97, "learning_rate": 3.098379408960915e-07, "loss": 1.8076, "step": 47083 }, { "epoch": 8.97, "learning_rate": 3.097807435653003e-07, "loss": 2.1509, "step": 47084 }, { "epoch": 8.97, "learning_rate": 3.0972354623450906e-07, "loss": 0.3955, "step": 47085 }, { "epoch": 8.97, "learning_rate": 3.096663489037178e-07, "loss": 2.1908, "step": 47086 }, { "epoch": 8.97, "learning_rate": 3.096091515729266e-07, "loss": 1.7156, "step": 47087 }, { "epoch": 8.97, "learning_rate": 3.095519542421354e-07, "loss": 0.7455, "step": 47088 }, { "epoch": 8.97, "learning_rate": 3.0949475691134413e-07, "loss": 1.6292, "step": 47089 }, { "epoch": 8.97, "learning_rate": 3.094375595805529e-07, "loss": 2.3682, "step": 47090 }, { "epoch": 8.97, "learning_rate": 3.0938036224976166e-07, "loss": 3.8637, "step": 47091 }, { "epoch": 8.97, "learning_rate": 3.0932316491897046e-07, "loss": 3.095, "step": 47092 }, { "epoch": 8.97, "learning_rate": 3.092659675881792e-07, "loss": 0.6423, "step": 47093 }, { "epoch": 8.97, "learning_rate": 3.09208770257388e-07, "loss": 3.0824, "step": 47094 }, { "epoch": 8.97, "learning_rate": 3.091515729265968e-07, "loss": 2.9459, "step": 47095 }, { "epoch": 8.97, "learning_rate": 3.090943755958056e-07, "loss": 1.6125, "step": 47096 }, { "epoch": 8.97, "learning_rate": 3.0903717826501427e-07, "loss": 0.8089, "step": 47097 }, { "epoch": 8.97, "learning_rate": 3.0897998093422306e-07, "loss": 1.6226, "step": 47098 }, { "epoch": 8.97, "learning_rate": 3.0892278360343186e-07, "loss": 0.2762, "step": 47099 }, { "epoch": 8.97, "learning_rate": 3.088655862726406e-07, "loss": 0.6954, "step": 47100 }, { "epoch": 8.97, "learning_rate": 3.088083889418494e-07, "loss": 1.8783, "step": 47101 }, { "epoch": 8.97, "learning_rate": 3.087511916110582e-07, "loss": 1.6257, "step": 47102 }, { "epoch": 8.97, "learning_rate": 3.086939942802669e-07, "loss": 1.734, "step": 47103 }, { "epoch": 8.97, "learning_rate": 3.0863679694947567e-07, "loss": 0.3451, "step": 47104 }, { "epoch": 8.97, "learning_rate": 3.0857959961868446e-07, "loss": 1.4857, "step": 47105 }, { "epoch": 8.97, "learning_rate": 3.0852240228789325e-07, "loss": 0.9124, "step": 47106 }, { "epoch": 8.97, "learning_rate": 3.0846520495710205e-07, "loss": 1.4974, "step": 47107 }, { "epoch": 8.97, "learning_rate": 3.084080076263108e-07, "loss": 0.8546, "step": 47108 }, { "epoch": 8.97, "learning_rate": 3.0835081029551953e-07, "loss": 1.528, "step": 47109 }, { "epoch": 8.97, "learning_rate": 3.082936129647283e-07, "loss": 1.9436, "step": 47110 }, { "epoch": 8.97, "learning_rate": 3.0823641563393706e-07, "loss": 1.417, "step": 47111 }, { "epoch": 8.97, "learning_rate": 3.0817921830314586e-07, "loss": 1.1458, "step": 47112 }, { "epoch": 8.97, "learning_rate": 3.0812202097235465e-07, "loss": 1.9355, "step": 47113 }, { "epoch": 8.97, "learning_rate": 3.080648236415634e-07, "loss": 1.16, "step": 47114 }, { "epoch": 8.97, "learning_rate": 3.0800762631077213e-07, "loss": 1.4852, "step": 47115 }, { "epoch": 8.97, "learning_rate": 3.0795042897998093e-07, "loss": 1.5398, "step": 47116 }, { "epoch": 8.97, "learning_rate": 3.078932316491897e-07, "loss": 1.7785, "step": 47117 }, { "epoch": 8.97, "learning_rate": 3.078360343183985e-07, "loss": 2.0784, "step": 47118 }, { "epoch": 8.98, "learning_rate": 3.0777883698760726e-07, "loss": 2.1823, "step": 47119 }, { "epoch": 8.98, "learning_rate": 3.07721639656816e-07, "loss": 0.8073, "step": 47120 }, { "epoch": 8.98, "learning_rate": 3.076644423260248e-07, "loss": 2.393, "step": 47121 }, { "epoch": 8.98, "learning_rate": 3.0760724499523353e-07, "loss": 1.9211, "step": 47122 }, { "epoch": 8.98, "learning_rate": 3.075500476644423e-07, "loss": 1.8449, "step": 47123 }, { "epoch": 8.98, "learning_rate": 3.074928503336511e-07, "loss": 1.3973, "step": 47124 }, { "epoch": 8.98, "learning_rate": 3.074356530028599e-07, "loss": 1.7525, "step": 47125 }, { "epoch": 8.98, "learning_rate": 3.073784556720686e-07, "loss": 1.8951, "step": 47126 }, { "epoch": 8.98, "learning_rate": 3.073212583412774e-07, "loss": 1.1138, "step": 47127 }, { "epoch": 8.98, "learning_rate": 3.072640610104862e-07, "loss": 1.056, "step": 47128 }, { "epoch": 8.98, "learning_rate": 3.07206863679695e-07, "loss": 1.4082, "step": 47129 }, { "epoch": 8.98, "learning_rate": 3.071496663489037e-07, "loss": 1.6029, "step": 47130 }, { "epoch": 8.98, "learning_rate": 3.070924690181125e-07, "loss": 1.7794, "step": 47131 }, { "epoch": 8.98, "learning_rate": 3.0703527168732126e-07, "loss": 0.8818, "step": 47132 }, { "epoch": 8.98, "learning_rate": 3.0697807435653e-07, "loss": 1.6465, "step": 47133 }, { "epoch": 8.98, "learning_rate": 3.069208770257388e-07, "loss": 0.2963, "step": 47134 }, { "epoch": 8.98, "learning_rate": 3.068636796949476e-07, "loss": 1.0467, "step": 47135 }, { "epoch": 8.98, "learning_rate": 3.068064823641564e-07, "loss": 1.3805, "step": 47136 }, { "epoch": 8.98, "learning_rate": 3.067492850333651e-07, "loss": 1.7996, "step": 47137 }, { "epoch": 8.98, "learning_rate": 3.0669208770257386e-07, "loss": 2.3407, "step": 47138 }, { "epoch": 8.98, "learning_rate": 3.0663489037178266e-07, "loss": 1.0983, "step": 47139 }, { "epoch": 8.98, "learning_rate": 3.0657769304099145e-07, "loss": 1.4154, "step": 47140 }, { "epoch": 8.98, "learning_rate": 3.065204957102002e-07, "loss": 1.4389, "step": 47141 }, { "epoch": 8.98, "learning_rate": 3.06463298379409e-07, "loss": 1.0148, "step": 47142 }, { "epoch": 8.98, "learning_rate": 3.064061010486178e-07, "loss": 1.5496, "step": 47143 }, { "epoch": 8.98, "learning_rate": 3.0634890371782647e-07, "loss": 1.5708, "step": 47144 }, { "epoch": 8.98, "learning_rate": 3.0629170638703526e-07, "loss": 0.3403, "step": 47145 }, { "epoch": 8.98, "learning_rate": 3.0623450905624405e-07, "loss": 1.1598, "step": 47146 }, { "epoch": 8.98, "learning_rate": 3.0617731172545285e-07, "loss": 1.0907, "step": 47147 }, { "epoch": 8.98, "learning_rate": 3.061201143946616e-07, "loss": 1.4846, "step": 47148 }, { "epoch": 8.98, "learning_rate": 3.060629170638704e-07, "loss": 0.9874, "step": 47149 }, { "epoch": 8.98, "learning_rate": 3.060057197330791e-07, "loss": 2.0085, "step": 47150 }, { "epoch": 8.98, "learning_rate": 3.059485224022879e-07, "loss": 2.6988, "step": 47151 }, { "epoch": 8.98, "learning_rate": 3.0589132507149666e-07, "loss": 1.1159, "step": 47152 }, { "epoch": 8.98, "learning_rate": 3.0583412774070545e-07, "loss": 0.9114, "step": 47153 }, { "epoch": 8.98, "learning_rate": 3.0577693040991425e-07, "loss": 2.1371, "step": 47154 }, { "epoch": 8.98, "learning_rate": 3.05719733079123e-07, "loss": 2.9274, "step": 47155 }, { "epoch": 8.98, "learning_rate": 3.0566253574833173e-07, "loss": 3.4145, "step": 47156 }, { "epoch": 8.98, "learning_rate": 3.056053384175405e-07, "loss": 2.1583, "step": 47157 }, { "epoch": 8.98, "learning_rate": 3.055481410867493e-07, "loss": 3.0261, "step": 47158 }, { "epoch": 8.98, "learning_rate": 3.0549094375595806e-07, "loss": 0.6717, "step": 47159 }, { "epoch": 8.98, "learning_rate": 3.0543374642516685e-07, "loss": 0.8816, "step": 47160 }, { "epoch": 8.98, "learning_rate": 3.0537654909437564e-07, "loss": 3.5323, "step": 47161 }, { "epoch": 8.98, "learning_rate": 3.053193517635844e-07, "loss": 1.7257, "step": 47162 }, { "epoch": 8.98, "learning_rate": 3.052621544327931e-07, "loss": 2.7989, "step": 47163 }, { "epoch": 8.98, "learning_rate": 3.052049571020019e-07, "loss": 1.4914, "step": 47164 }, { "epoch": 8.98, "learning_rate": 3.051477597712107e-07, "loss": 0.5801, "step": 47165 }, { "epoch": 8.98, "learning_rate": 3.0509056244041945e-07, "loss": 0.6645, "step": 47166 }, { "epoch": 8.98, "learning_rate": 3.0503336510962825e-07, "loss": 0.7843, "step": 47167 }, { "epoch": 8.98, "learning_rate": 3.04976167778837e-07, "loss": 1.3103, "step": 47168 }, { "epoch": 8.98, "learning_rate": 3.049189704480458e-07, "loss": 2.2884, "step": 47169 }, { "epoch": 8.98, "learning_rate": 3.048617731172545e-07, "loss": 1.5283, "step": 47170 }, { "epoch": 8.98, "learning_rate": 3.048045757864633e-07, "loss": 0.8419, "step": 47171 }, { "epoch": 8.99, "learning_rate": 3.047473784556721e-07, "loss": 2.6287, "step": 47172 }, { "epoch": 8.99, "learning_rate": 3.0469018112488085e-07, "loss": 0.9889, "step": 47173 }, { "epoch": 8.99, "learning_rate": 3.046329837940896e-07, "loss": 0.9043, "step": 47174 }, { "epoch": 8.99, "learning_rate": 3.045757864632984e-07, "loss": 1.1945, "step": 47175 }, { "epoch": 8.99, "learning_rate": 3.045185891325072e-07, "loss": 1.1508, "step": 47176 }, { "epoch": 8.99, "learning_rate": 3.044613918017159e-07, "loss": 2.6009, "step": 47177 }, { "epoch": 8.99, "learning_rate": 3.044041944709247e-07, "loss": 2.4504, "step": 47178 }, { "epoch": 8.99, "learning_rate": 3.0434699714013346e-07, "loss": 2.3534, "step": 47179 }, { "epoch": 8.99, "learning_rate": 3.0428979980934225e-07, "loss": 2.2877, "step": 47180 }, { "epoch": 8.99, "learning_rate": 3.04232602478551e-07, "loss": 1.0004, "step": 47181 }, { "epoch": 8.99, "learning_rate": 3.041754051477598e-07, "loss": 1.9727, "step": 47182 }, { "epoch": 8.99, "learning_rate": 3.041182078169686e-07, "loss": 1.4798, "step": 47183 }, { "epoch": 8.99, "learning_rate": 3.040610104861773e-07, "loss": 1.7483, "step": 47184 }, { "epoch": 8.99, "learning_rate": 3.0400381315538606e-07, "loss": 0.3481, "step": 47185 }, { "epoch": 8.99, "learning_rate": 3.0394661582459485e-07, "loss": 1.8417, "step": 47186 }, { "epoch": 8.99, "learning_rate": 3.0388941849380365e-07, "loss": 1.2617, "step": 47187 }, { "epoch": 8.99, "learning_rate": 3.038322211630124e-07, "loss": 1.1581, "step": 47188 }, { "epoch": 8.99, "learning_rate": 3.037750238322212e-07, "loss": 2.3725, "step": 47189 }, { "epoch": 8.99, "learning_rate": 3.0371782650143e-07, "loss": 0.4586, "step": 47190 }, { "epoch": 8.99, "learning_rate": 3.036606291706387e-07, "loss": 2.2857, "step": 47191 }, { "epoch": 8.99, "learning_rate": 3.0360343183984746e-07, "loss": 1.0856, "step": 47192 }, { "epoch": 8.99, "learning_rate": 3.0354623450905625e-07, "loss": 1.1134, "step": 47193 }, { "epoch": 8.99, "learning_rate": 3.0348903717826505e-07, "loss": 0.8624, "step": 47194 }, { "epoch": 8.99, "learning_rate": 3.034318398474738e-07, "loss": 1.9603, "step": 47195 }, { "epoch": 8.99, "learning_rate": 3.033746425166826e-07, "loss": 2.3831, "step": 47196 }, { "epoch": 8.99, "learning_rate": 3.033174451858913e-07, "loss": 1.5611, "step": 47197 }, { "epoch": 8.99, "learning_rate": 3.032602478551001e-07, "loss": 1.654, "step": 47198 }, { "epoch": 8.99, "learning_rate": 3.0320305052430886e-07, "loss": 1.8808, "step": 47199 }, { "epoch": 8.99, "learning_rate": 3.0314585319351765e-07, "loss": 1.7111, "step": 47200 }, { "epoch": 8.99, "learning_rate": 3.0308865586272644e-07, "loss": 2.7175, "step": 47201 }, { "epoch": 8.99, "learning_rate": 3.030314585319352e-07, "loss": 0.8752, "step": 47202 }, { "epoch": 8.99, "learning_rate": 3.029742612011439e-07, "loss": 0.8249, "step": 47203 }, { "epoch": 8.99, "learning_rate": 3.029170638703527e-07, "loss": 0.9281, "step": 47204 }, { "epoch": 8.99, "learning_rate": 3.028598665395615e-07, "loss": 1.008, "step": 47205 }, { "epoch": 8.99, "learning_rate": 3.0280266920877025e-07, "loss": 1.8519, "step": 47206 }, { "epoch": 8.99, "learning_rate": 3.0274547187797905e-07, "loss": 2.3032, "step": 47207 }, { "epoch": 8.99, "learning_rate": 3.0268827454718784e-07, "loss": 1.8414, "step": 47208 }, { "epoch": 8.99, "learning_rate": 3.026310772163966e-07, "loss": 2.1692, "step": 47209 }, { "epoch": 8.99, "learning_rate": 3.025738798856053e-07, "loss": 1.8714, "step": 47210 }, { "epoch": 8.99, "learning_rate": 3.025166825548141e-07, "loss": 1.1179, "step": 47211 }, { "epoch": 8.99, "learning_rate": 3.024594852240229e-07, "loss": 1.2326, "step": 47212 }, { "epoch": 8.99, "learning_rate": 3.0240228789323165e-07, "loss": 0.9393, "step": 47213 }, { "epoch": 8.99, "learning_rate": 3.0234509056244045e-07, "loss": 1.7062, "step": 47214 }, { "epoch": 8.99, "learning_rate": 3.022878932316492e-07, "loss": 1.7608, "step": 47215 }, { "epoch": 8.99, "learning_rate": 3.02230695900858e-07, "loss": 3.0913, "step": 47216 }, { "epoch": 8.99, "learning_rate": 3.021734985700667e-07, "loss": 2.1359, "step": 47217 }, { "epoch": 8.99, "learning_rate": 3.021163012392755e-07, "loss": 1.4974, "step": 47218 }, { "epoch": 8.99, "learning_rate": 3.020591039084843e-07, "loss": 1.9558, "step": 47219 }, { "epoch": 8.99, "learning_rate": 3.0200190657769305e-07, "loss": 1.3833, "step": 47220 }, { "epoch": 8.99, "learning_rate": 3.019447092469018e-07, "loss": 1.3767, "step": 47221 }, { "epoch": 8.99, "learning_rate": 3.018875119161106e-07, "loss": 2.0599, "step": 47222 }, { "epoch": 8.99, "learning_rate": 3.018303145853194e-07, "loss": 1.3957, "step": 47223 }, { "epoch": 9.0, "learning_rate": 3.017731172545281e-07, "loss": 1.0587, "step": 47224 }, { "epoch": 9.0, "learning_rate": 3.017159199237369e-07, "loss": 1.4143, "step": 47225 }, { "epoch": 9.0, "learning_rate": 3.016587225929457e-07, "loss": 0.8141, "step": 47226 }, { "epoch": 9.0, "learning_rate": 3.0160152526215445e-07, "loss": 1.2773, "step": 47227 }, { "epoch": 9.0, "learning_rate": 3.015443279313632e-07, "loss": 1.0755, "step": 47228 }, { "epoch": 9.0, "learning_rate": 3.01487130600572e-07, "loss": 0.4485, "step": 47229 }, { "epoch": 9.0, "learning_rate": 3.014299332697808e-07, "loss": 2.5362, "step": 47230 }, { "epoch": 9.0, "learning_rate": 3.013727359389895e-07, "loss": 0.8283, "step": 47231 }, { "epoch": 9.0, "learning_rate": 3.0131553860819826e-07, "loss": 1.1969, "step": 47232 }, { "epoch": 9.0, "learning_rate": 3.0125834127740705e-07, "loss": 1.7963, "step": 47233 }, { "epoch": 9.0, "learning_rate": 3.0120114394661585e-07, "loss": 0.5682, "step": 47234 }, { "epoch": 9.0, "learning_rate": 3.011439466158246e-07, "loss": 0.7263, "step": 47235 }, { "epoch": 9.0, "learning_rate": 3.010867492850334e-07, "loss": 1.9127, "step": 47236 }, { "epoch": 9.0, "learning_rate": 3.010295519542422e-07, "loss": 1.7652, "step": 47237 }, { "epoch": 9.0, "learning_rate": 3.009723546234509e-07, "loss": 1.5226, "step": 47238 }, { "epoch": 9.0, "learning_rate": 3.0091515729265966e-07, "loss": 1.2341, "step": 47239 }, { "epoch": 9.0, "learning_rate": 3.0085795996186845e-07, "loss": 1.7707, "step": 47240 }, { "epoch": 9.0, "learning_rate": 3.0080076263107724e-07, "loss": 1.8071, "step": 47241 }, { "epoch": 9.0, "learning_rate": 3.00743565300286e-07, "loss": 1.621, "step": 47242 }, { "epoch": 9.0, "learning_rate": 3.006863679694948e-07, "loss": 0.5189, "step": 47243 }, { "epoch": 9.0, "learning_rate": 3.006291706387035e-07, "loss": 1.5236, "step": 47244 }, { "epoch": 9.0, "learning_rate": 3.005719733079123e-07, "loss": 1.3213, "step": 47245 }, { "epoch": 9.0, "learning_rate": 3.0051477597712105e-07, "loss": 1.1162, "step": 47246 }, { "epoch": 9.0, "learning_rate": 3.0045757864632985e-07, "loss": 0.6187, "step": 47247 }, { "epoch": 9.0, "learning_rate": 3.0040038131553864e-07, "loss": 1.3959, "step": 47248 }, { "epoch": 9.0, "learning_rate": 3.003431839847474e-07, "loss": 2.7575, "step": 47249 }, { "epoch": 9.0, "learning_rate": 3.002859866539561e-07, "loss": 0.4078, "step": 47250 }, { "epoch": 9.0, "eval_accuracy": 0.6210887269609944, "eval_loss": 1.459395170211792, "eval_runtime": 127.3525, "eval_samples_per_second": 18.319, "eval_steps_per_second": 4.586, "step": 47250 }, { "epoch": 9.0, "learning_rate": 3.002287893231649e-07, "loss": 1.9603, "step": 47251 }, { "epoch": 9.0, "learning_rate": 3.001715919923737e-07, "loss": 1.5044, "step": 47252 }, { "epoch": 9.0, "learning_rate": 3.0011439466158245e-07, "loss": 1.4493, "step": 47253 }, { "epoch": 9.0, "learning_rate": 3.0005719733079125e-07, "loss": 1.0657, "step": 47254 }, { "epoch": 9.0, "learning_rate": 3.0000000000000004e-07, "loss": 1.7197, "step": 47255 }, { "epoch": 9.0, "learning_rate": 2.999428026692088e-07, "loss": 1.2455, "step": 47256 }, { "epoch": 9.0, "learning_rate": 2.998856053384175e-07, "loss": 0.9777, "step": 47257 }, { "epoch": 9.0, "learning_rate": 2.998284080076263e-07, "loss": 1.3524, "step": 47258 }, { "epoch": 9.0, "learning_rate": 2.997712106768351e-07, "loss": 2.0358, "step": 47259 }, { "epoch": 9.0, "learning_rate": 2.9971401334604385e-07, "loss": 2.1427, "step": 47260 }, { "epoch": 9.0, "learning_rate": 2.9965681601525264e-07, "loss": 0.5138, "step": 47261 }, { "epoch": 9.0, "learning_rate": 2.995996186844614e-07, "loss": 1.5001, "step": 47262 }, { "epoch": 9.0, "learning_rate": 2.995424213536702e-07, "loss": 1.3175, "step": 47263 }, { "epoch": 9.0, "learning_rate": 2.994852240228789e-07, "loss": 1.3809, "step": 47264 }, { "epoch": 9.0, "learning_rate": 2.994280266920877e-07, "loss": 1.4496, "step": 47265 }, { "epoch": 9.0, "learning_rate": 2.993708293612965e-07, "loss": 1.9497, "step": 47266 }, { "epoch": 9.0, "learning_rate": 2.9931363203050525e-07, "loss": 1.661, "step": 47267 }, { "epoch": 9.0, "learning_rate": 2.99256434699714e-07, "loss": 1.8293, "step": 47268 }, { "epoch": 9.0, "learning_rate": 2.991992373689228e-07, "loss": 1.8483, "step": 47269 }, { "epoch": 9.0, "learning_rate": 2.991420400381316e-07, "loss": 1.7086, "step": 47270 }, { "epoch": 9.0, "learning_rate": 2.990848427073403e-07, "loss": 1.5104, "step": 47271 }, { "epoch": 9.0, "learning_rate": 2.990276453765491e-07, "loss": 1.4959, "step": 47272 }, { "epoch": 9.0, "learning_rate": 2.989704480457579e-07, "loss": 1.9959, "step": 47273 }, { "epoch": 9.0, "learning_rate": 2.9891325071496665e-07, "loss": 1.4573, "step": 47274 }, { "epoch": 9.0, "learning_rate": 2.988560533841754e-07, "loss": 0.4774, "step": 47275 }, { "epoch": 9.0, "learning_rate": 2.987988560533842e-07, "loss": 2.2378, "step": 47276 }, { "epoch": 9.01, "learning_rate": 2.98741658722593e-07, "loss": 1.4955, "step": 47277 }, { "epoch": 9.01, "learning_rate": 2.986844613918017e-07, "loss": 1.5799, "step": 47278 }, { "epoch": 9.01, "learning_rate": 2.986272640610105e-07, "loss": 1.1126, "step": 47279 }, { "epoch": 9.01, "learning_rate": 2.9857006673021925e-07, "loss": 2.4087, "step": 47280 }, { "epoch": 9.01, "learning_rate": 2.9851286939942804e-07, "loss": 1.2267, "step": 47281 }, { "epoch": 9.01, "learning_rate": 2.984556720686368e-07, "loss": 1.2298, "step": 47282 }, { "epoch": 9.01, "learning_rate": 2.983984747378456e-07, "loss": 1.6476, "step": 47283 }, { "epoch": 9.01, "learning_rate": 2.9834127740705437e-07, "loss": 2.2369, "step": 47284 }, { "epoch": 9.01, "learning_rate": 2.982840800762631e-07, "loss": 1.774, "step": 47285 }, { "epoch": 9.01, "learning_rate": 2.9822688274547185e-07, "loss": 1.4894, "step": 47286 }, { "epoch": 9.01, "learning_rate": 2.9816968541468065e-07, "loss": 0.6428, "step": 47287 }, { "epoch": 9.01, "learning_rate": 2.9811248808388944e-07, "loss": 1.7982, "step": 47288 }, { "epoch": 9.01, "learning_rate": 2.980552907530982e-07, "loss": 2.3599, "step": 47289 }, { "epoch": 9.01, "learning_rate": 2.97998093422307e-07, "loss": 1.8246, "step": 47290 }, { "epoch": 9.01, "learning_rate": 2.9794089609151577e-07, "loss": 1.0907, "step": 47291 }, { "epoch": 9.01, "learning_rate": 2.978836987607245e-07, "loss": 2.5944, "step": 47292 }, { "epoch": 9.01, "learning_rate": 2.9782650142993325e-07, "loss": 2.5884, "step": 47293 }, { "epoch": 9.01, "learning_rate": 2.9776930409914205e-07, "loss": 0.3937, "step": 47294 }, { "epoch": 9.01, "learning_rate": 2.9771210676835084e-07, "loss": 2.7145, "step": 47295 }, { "epoch": 9.01, "learning_rate": 2.976549094375596e-07, "loss": 2.0424, "step": 47296 }, { "epoch": 9.01, "learning_rate": 2.975977121067683e-07, "loss": 2.2099, "step": 47297 }, { "epoch": 9.01, "learning_rate": 2.975405147759771e-07, "loss": 1.933, "step": 47298 }, { "epoch": 9.01, "learning_rate": 2.974833174451859e-07, "loss": 2.2837, "step": 47299 }, { "epoch": 9.01, "learning_rate": 2.9742612011439465e-07, "loss": 0.9216, "step": 47300 }, { "epoch": 9.01, "learning_rate": 2.9736892278360344e-07, "loss": 2.603, "step": 47301 }, { "epoch": 9.01, "learning_rate": 2.9731172545281224e-07, "loss": 1.9293, "step": 47302 }, { "epoch": 9.01, "learning_rate": 2.97254528122021e-07, "loss": 1.292, "step": 47303 }, { "epoch": 9.01, "learning_rate": 2.971973307912297e-07, "loss": 0.2635, "step": 47304 }, { "epoch": 9.01, "learning_rate": 2.971401334604385e-07, "loss": 0.6583, "step": 47305 }, { "epoch": 9.01, "learning_rate": 2.970829361296473e-07, "loss": 1.4285, "step": 47306 }, { "epoch": 9.01, "learning_rate": 2.9702573879885605e-07, "loss": 3.5569, "step": 47307 }, { "epoch": 9.01, "learning_rate": 2.9696854146806484e-07, "loss": 1.1597, "step": 47308 }, { "epoch": 9.01, "learning_rate": 2.969113441372736e-07, "loss": 1.2506, "step": 47309 }, { "epoch": 9.01, "learning_rate": 2.968541468064824e-07, "loss": 1.0492, "step": 47310 }, { "epoch": 9.01, "learning_rate": 2.967969494756911e-07, "loss": 1.6726, "step": 47311 }, { "epoch": 9.01, "learning_rate": 2.967397521448999e-07, "loss": 1.0908, "step": 47312 }, { "epoch": 9.01, "learning_rate": 2.966825548141087e-07, "loss": 0.4281, "step": 47313 }, { "epoch": 9.01, "learning_rate": 2.9662535748331745e-07, "loss": 1.0077, "step": 47314 }, { "epoch": 9.01, "learning_rate": 2.965681601525262e-07, "loss": 1.0605, "step": 47315 }, { "epoch": 9.01, "learning_rate": 2.96510962821735e-07, "loss": 1.868, "step": 47316 }, { "epoch": 9.01, "learning_rate": 2.964537654909438e-07, "loss": 1.4973, "step": 47317 }, { "epoch": 9.01, "learning_rate": 2.963965681601525e-07, "loss": 1.4363, "step": 47318 }, { "epoch": 9.01, "learning_rate": 2.963393708293613e-07, "loss": 2.2979, "step": 47319 }, { "epoch": 9.01, "learning_rate": 2.962821734985701e-07, "loss": 0.7044, "step": 47320 }, { "epoch": 9.01, "learning_rate": 2.9622497616777884e-07, "loss": 1.4403, "step": 47321 }, { "epoch": 9.01, "learning_rate": 2.961677788369876e-07, "loss": 1.5629, "step": 47322 }, { "epoch": 9.01, "learning_rate": 2.961105815061964e-07, "loss": 3.5833, "step": 47323 }, { "epoch": 9.01, "learning_rate": 2.9605338417540517e-07, "loss": 2.0422, "step": 47324 }, { "epoch": 9.01, "learning_rate": 2.959961868446139e-07, "loss": 2.854, "step": 47325 }, { "epoch": 9.01, "learning_rate": 2.959389895138227e-07, "loss": 1.3414, "step": 47326 }, { "epoch": 9.01, "learning_rate": 2.9588179218303145e-07, "loss": 1.8568, "step": 47327 }, { "epoch": 9.01, "learning_rate": 2.9582459485224024e-07, "loss": 4.0646, "step": 47328 }, { "epoch": 9.02, "learning_rate": 2.95767397521449e-07, "loss": 1.4823, "step": 47329 }, { "epoch": 9.02, "learning_rate": 2.957102001906578e-07, "loss": 2.5352, "step": 47330 }, { "epoch": 9.02, "learning_rate": 2.9565300285986657e-07, "loss": 2.0962, "step": 47331 }, { "epoch": 9.02, "learning_rate": 2.955958055290753e-07, "loss": 1.8544, "step": 47332 }, { "epoch": 9.02, "learning_rate": 2.9553860819828405e-07, "loss": 2.6307, "step": 47333 }, { "epoch": 9.02, "learning_rate": 2.9548141086749285e-07, "loss": 1.3711, "step": 47334 }, { "epoch": 9.02, "learning_rate": 2.9542421353670164e-07, "loss": 1.2532, "step": 47335 }, { "epoch": 9.02, "learning_rate": 2.953670162059104e-07, "loss": 0.9437, "step": 47336 }, { "epoch": 9.02, "learning_rate": 2.953098188751192e-07, "loss": 0.1829, "step": 47337 }, { "epoch": 9.02, "learning_rate": 2.9525262154432797e-07, "loss": 2.0382, "step": 47338 }, { "epoch": 9.02, "learning_rate": 2.951954242135367e-07, "loss": 1.6913, "step": 47339 }, { "epoch": 9.02, "learning_rate": 2.9513822688274545e-07, "loss": 2.2268, "step": 47340 }, { "epoch": 9.02, "learning_rate": 2.9508102955195424e-07, "loss": 0.7274, "step": 47341 }, { "epoch": 9.02, "learning_rate": 2.9502383222116304e-07, "loss": 1.4338, "step": 47342 }, { "epoch": 9.02, "learning_rate": 2.949666348903718e-07, "loss": 1.3632, "step": 47343 }, { "epoch": 9.02, "learning_rate": 2.9490943755958057e-07, "loss": 2.1887, "step": 47344 }, { "epoch": 9.02, "learning_rate": 2.948522402287893e-07, "loss": 1.5273, "step": 47345 }, { "epoch": 9.02, "learning_rate": 2.947950428979981e-07, "loss": 0.9673, "step": 47346 }, { "epoch": 9.02, "learning_rate": 2.9473784556720685e-07, "loss": 1.1957, "step": 47347 }, { "epoch": 9.02, "learning_rate": 2.9468064823641564e-07, "loss": 0.74, "step": 47348 }, { "epoch": 9.02, "learning_rate": 2.9462345090562443e-07, "loss": 2.5872, "step": 47349 }, { "epoch": 9.02, "learning_rate": 2.9456625357483323e-07, "loss": 1.0942, "step": 47350 }, { "epoch": 9.02, "learning_rate": 2.945090562440419e-07, "loss": 1.9822, "step": 47351 }, { "epoch": 9.02, "learning_rate": 2.944518589132507e-07, "loss": 1.2308, "step": 47352 }, { "epoch": 9.02, "learning_rate": 2.943946615824595e-07, "loss": 1.4878, "step": 47353 }, { "epoch": 9.02, "learning_rate": 2.9433746425166825e-07, "loss": 1.4955, "step": 47354 }, { "epoch": 9.02, "learning_rate": 2.9428026692087704e-07, "loss": 1.4835, "step": 47355 }, { "epoch": 9.02, "learning_rate": 2.942230695900858e-07, "loss": 2.0326, "step": 47356 }, { "epoch": 9.02, "learning_rate": 2.941658722592946e-07, "loss": 1.3041, "step": 47357 }, { "epoch": 9.02, "learning_rate": 2.941086749285033e-07, "loss": 2.3314, "step": 47358 }, { "epoch": 9.02, "learning_rate": 2.940514775977121e-07, "loss": 2.0374, "step": 47359 }, { "epoch": 9.02, "learning_rate": 2.939942802669209e-07, "loss": 1.0089, "step": 47360 }, { "epoch": 9.02, "learning_rate": 2.939370829361297e-07, "loss": 2.1215, "step": 47361 }, { "epoch": 9.02, "learning_rate": 2.938798856053384e-07, "loss": 1.0908, "step": 47362 }, { "epoch": 9.02, "learning_rate": 2.938226882745472e-07, "loss": 2.0168, "step": 47363 }, { "epoch": 9.02, "learning_rate": 2.9376549094375597e-07, "loss": 0.7187, "step": 47364 }, { "epoch": 9.02, "learning_rate": 2.937082936129647e-07, "loss": 1.5256, "step": 47365 }, { "epoch": 9.02, "learning_rate": 2.936510962821735e-07, "loss": 2.001, "step": 47366 }, { "epoch": 9.02, "learning_rate": 2.935938989513823e-07, "loss": 0.398, "step": 47367 }, { "epoch": 9.02, "learning_rate": 2.9353670162059104e-07, "loss": 1.9966, "step": 47368 }, { "epoch": 9.02, "learning_rate": 2.934795042897998e-07, "loss": 1.5627, "step": 47369 }, { "epoch": 9.02, "learning_rate": 2.934223069590086e-07, "loss": 2.6205, "step": 47370 }, { "epoch": 9.02, "learning_rate": 2.9336510962821737e-07, "loss": 2.0605, "step": 47371 }, { "epoch": 9.02, "learning_rate": 2.9330791229742616e-07, "loss": 1.0521, "step": 47372 }, { "epoch": 9.02, "learning_rate": 2.932507149666349e-07, "loss": 1.3982, "step": 47373 }, { "epoch": 9.02, "learning_rate": 2.9319351763584364e-07, "loss": 0.5092, "step": 47374 }, { "epoch": 9.02, "learning_rate": 2.9313632030505244e-07, "loss": 3.0808, "step": 47375 }, { "epoch": 9.02, "learning_rate": 2.930791229742612e-07, "loss": 0.4503, "step": 47376 }, { "epoch": 9.02, "learning_rate": 2.9302192564346997e-07, "loss": 1.7733, "step": 47377 }, { "epoch": 9.02, "learning_rate": 2.9296472831267877e-07, "loss": 1.2442, "step": 47378 }, { "epoch": 9.02, "learning_rate": 2.9290753098188756e-07, "loss": 1.2191, "step": 47379 }, { "epoch": 9.02, "learning_rate": 2.9285033365109625e-07, "loss": 1.6188, "step": 47380 }, { "epoch": 9.02, "learning_rate": 2.9279313632030504e-07, "loss": 1.1433, "step": 47381 }, { "epoch": 9.03, "learning_rate": 2.9273593898951384e-07, "loss": 0.8568, "step": 47382 }, { "epoch": 9.03, "learning_rate": 2.926787416587226e-07, "loss": 1.667, "step": 47383 }, { "epoch": 9.03, "learning_rate": 2.9262154432793137e-07, "loss": 1.632, "step": 47384 }, { "epoch": 9.03, "learning_rate": 2.9256434699714017e-07, "loss": 1.9065, "step": 47385 }, { "epoch": 9.03, "learning_rate": 2.925071496663489e-07, "loss": 0.7656, "step": 47386 }, { "epoch": 9.03, "learning_rate": 2.9244995233555765e-07, "loss": 1.0784, "step": 47387 }, { "epoch": 9.03, "learning_rate": 2.9239275500476644e-07, "loss": 1.4287, "step": 47388 }, { "epoch": 9.03, "learning_rate": 2.9233555767397523e-07, "loss": 1.6891, "step": 47389 }, { "epoch": 9.03, "learning_rate": 2.9227836034318403e-07, "loss": 2.0123, "step": 47390 }, { "epoch": 9.03, "learning_rate": 2.9222116301239277e-07, "loss": 1.9159, "step": 47391 }, { "epoch": 9.03, "learning_rate": 2.921639656816015e-07, "loss": 2.1406, "step": 47392 }, { "epoch": 9.03, "learning_rate": 2.921067683508103e-07, "loss": 2.287, "step": 47393 }, { "epoch": 9.03, "learning_rate": 2.9204957102001904e-07, "loss": 2.0329, "step": 47394 }, { "epoch": 9.03, "learning_rate": 2.9199237368922784e-07, "loss": 1.8742, "step": 47395 }, { "epoch": 9.03, "learning_rate": 2.9193517635843663e-07, "loss": 0.9671, "step": 47396 }, { "epoch": 9.03, "learning_rate": 2.918779790276454e-07, "loss": 1.0914, "step": 47397 }, { "epoch": 9.03, "learning_rate": 2.918207816968541e-07, "loss": 2.4631, "step": 47398 }, { "epoch": 9.03, "learning_rate": 2.917635843660629e-07, "loss": 1.3939, "step": 47399 }, { "epoch": 9.03, "learning_rate": 2.917063870352717e-07, "loss": 1.0246, "step": 47400 }, { "epoch": 9.03, "learning_rate": 2.916491897044805e-07, "loss": 1.9421, "step": 47401 }, { "epoch": 9.03, "learning_rate": 2.9159199237368924e-07, "loss": 0.5655, "step": 47402 }, { "epoch": 9.03, "learning_rate": 2.9153479504289803e-07, "loss": 1.7741, "step": 47403 }, { "epoch": 9.03, "learning_rate": 2.9147759771210677e-07, "loss": 1.0728, "step": 47404 }, { "epoch": 9.03, "learning_rate": 2.914204003813155e-07, "loss": 0.484, "step": 47405 }, { "epoch": 9.03, "learning_rate": 2.913632030505243e-07, "loss": 1.2473, "step": 47406 }, { "epoch": 9.03, "learning_rate": 2.913060057197331e-07, "loss": 1.8467, "step": 47407 }, { "epoch": 9.03, "learning_rate": 2.912488083889419e-07, "loss": 2.0156, "step": 47408 }, { "epoch": 9.03, "learning_rate": 2.9119161105815063e-07, "loss": 0.8065, "step": 47409 }, { "epoch": 9.03, "learning_rate": 2.911344137273594e-07, "loss": 1.6638, "step": 47410 }, { "epoch": 9.03, "learning_rate": 2.9107721639656817e-07, "loss": 1.598, "step": 47411 }, { "epoch": 9.03, "learning_rate": 2.9102001906577696e-07, "loss": 1.2119, "step": 47412 }, { "epoch": 9.03, "learning_rate": 2.909628217349857e-07, "loss": 0.4437, "step": 47413 }, { "epoch": 9.03, "learning_rate": 2.909056244041945e-07, "loss": 1.3581, "step": 47414 }, { "epoch": 9.03, "learning_rate": 2.9084842707340324e-07, "loss": 0.9522, "step": 47415 }, { "epoch": 9.03, "learning_rate": 2.90791229742612e-07, "loss": 1.719, "step": 47416 }, { "epoch": 9.03, "learning_rate": 2.9073403241182077e-07, "loss": 2.152, "step": 47417 }, { "epoch": 9.03, "learning_rate": 2.9067683508102957e-07, "loss": 1.7104, "step": 47418 }, { "epoch": 9.03, "learning_rate": 2.9061963775023836e-07, "loss": 1.8852, "step": 47419 }, { "epoch": 9.03, "learning_rate": 2.905624404194471e-07, "loss": 2.3108, "step": 47420 }, { "epoch": 9.03, "learning_rate": 2.9050524308865584e-07, "loss": 1.8357, "step": 47421 }, { "epoch": 9.03, "learning_rate": 2.9044804575786464e-07, "loss": 1.9166, "step": 47422 }, { "epoch": 9.03, "learning_rate": 2.9039084842707343e-07, "loss": 3.598, "step": 47423 }, { "epoch": 9.03, "learning_rate": 2.9033365109628217e-07, "loss": 1.7139, "step": 47424 }, { "epoch": 9.03, "learning_rate": 2.9027645376549096e-07, "loss": 2.096, "step": 47425 }, { "epoch": 9.03, "learning_rate": 2.9021925643469976e-07, "loss": 1.3627, "step": 47426 }, { "epoch": 9.03, "learning_rate": 2.9016205910390845e-07, "loss": 1.3453, "step": 47427 }, { "epoch": 9.03, "learning_rate": 2.9010486177311724e-07, "loss": 1.3327, "step": 47428 }, { "epoch": 9.03, "learning_rate": 2.9004766444232603e-07, "loss": 1.0796, "step": 47429 }, { "epoch": 9.03, "learning_rate": 2.8999046711153483e-07, "loss": 0.9922, "step": 47430 }, { "epoch": 9.03, "learning_rate": 2.8993326978074357e-07, "loss": 1.41, "step": 47431 }, { "epoch": 9.03, "learning_rate": 2.8987607244995236e-07, "loss": 3.3814, "step": 47432 }, { "epoch": 9.03, "learning_rate": 2.898188751191611e-07, "loss": 0.3578, "step": 47433 }, { "epoch": 9.04, "learning_rate": 2.897616777883699e-07, "loss": 1.6064, "step": 47434 }, { "epoch": 9.04, "learning_rate": 2.8970448045757864e-07, "loss": 0.5781, "step": 47435 }, { "epoch": 9.04, "learning_rate": 2.8964728312678743e-07, "loss": 2.2371, "step": 47436 }, { "epoch": 9.04, "learning_rate": 2.895900857959962e-07, "loss": 1.6914, "step": 47437 }, { "epoch": 9.04, "learning_rate": 2.8953288846520497e-07, "loss": 1.1415, "step": 47438 }, { "epoch": 9.04, "learning_rate": 2.894756911344137e-07, "loss": 2.0878, "step": 47439 }, { "epoch": 9.04, "learning_rate": 2.894184938036225e-07, "loss": 1.2631, "step": 47440 }, { "epoch": 9.04, "learning_rate": 2.893612964728313e-07, "loss": 2.3937, "step": 47441 }, { "epoch": 9.04, "learning_rate": 2.8930409914204004e-07, "loss": 1.5232, "step": 47442 }, { "epoch": 9.04, "learning_rate": 2.8924690181124883e-07, "loss": 1.571, "step": 47443 }, { "epoch": 9.04, "learning_rate": 2.891897044804576e-07, "loss": 1.0954, "step": 47444 }, { "epoch": 9.04, "learning_rate": 2.8913250714966636e-07, "loss": 1.534, "step": 47445 }, { "epoch": 9.04, "learning_rate": 2.890753098188751e-07, "loss": 0.5719, "step": 47446 }, { "epoch": 9.04, "learning_rate": 2.890181124880839e-07, "loss": 1.4298, "step": 47447 }, { "epoch": 9.04, "learning_rate": 2.889609151572927e-07, "loss": 2.4759, "step": 47448 }, { "epoch": 9.04, "learning_rate": 2.8890371782650143e-07, "loss": 1.7578, "step": 47449 }, { "epoch": 9.04, "learning_rate": 2.8884652049571023e-07, "loss": 2.808, "step": 47450 }, { "epoch": 9.04, "learning_rate": 2.8878932316491897e-07, "loss": 2.8125, "step": 47451 }, { "epoch": 9.04, "learning_rate": 2.8873212583412776e-07, "loss": 1.4488, "step": 47452 }, { "epoch": 9.04, "learning_rate": 2.886749285033365e-07, "loss": 2.0869, "step": 47453 }, { "epoch": 9.04, "learning_rate": 2.886177311725453e-07, "loss": 1.17, "step": 47454 }, { "epoch": 9.04, "learning_rate": 2.885605338417541e-07, "loss": 2.3044, "step": 47455 }, { "epoch": 9.04, "learning_rate": 2.8850333651096283e-07, "loss": 1.6283, "step": 47456 }, { "epoch": 9.04, "learning_rate": 2.8844613918017157e-07, "loss": 1.2502, "step": 47457 }, { "epoch": 9.04, "learning_rate": 2.8838894184938037e-07, "loss": 1.0573, "step": 47458 }, { "epoch": 9.04, "learning_rate": 2.8833174451858916e-07, "loss": 0.731, "step": 47459 }, { "epoch": 9.04, "learning_rate": 2.882745471877979e-07, "loss": 1.5223, "step": 47460 }, { "epoch": 9.04, "learning_rate": 2.882173498570067e-07, "loss": 1.397, "step": 47461 }, { "epoch": 9.04, "learning_rate": 2.881601525262155e-07, "loss": 1.4025, "step": 47462 }, { "epoch": 9.04, "learning_rate": 2.8810295519542423e-07, "loss": 0.9788, "step": 47463 }, { "epoch": 9.04, "learning_rate": 2.8804575786463297e-07, "loss": 2.2323, "step": 47464 }, { "epoch": 9.04, "learning_rate": 2.8798856053384176e-07, "loss": 2.3107, "step": 47465 }, { "epoch": 9.04, "learning_rate": 2.8793136320305056e-07, "loss": 0.6459, "step": 47466 }, { "epoch": 9.04, "learning_rate": 2.878741658722593e-07, "loss": 1.5062, "step": 47467 }, { "epoch": 9.04, "learning_rate": 2.878169685414681e-07, "loss": 1.0604, "step": 47468 }, { "epoch": 9.04, "learning_rate": 2.8775977121067683e-07, "loss": 0.6209, "step": 47469 }, { "epoch": 9.04, "learning_rate": 2.8770257387988563e-07, "loss": 1.2426, "step": 47470 }, { "epoch": 9.04, "learning_rate": 2.8764537654909437e-07, "loss": 1.2338, "step": 47471 }, { "epoch": 9.04, "learning_rate": 2.8758817921830316e-07, "loss": 0.923, "step": 47472 }, { "epoch": 9.04, "learning_rate": 2.8753098188751196e-07, "loss": 2.085, "step": 47473 }, { "epoch": 9.04, "learning_rate": 2.874737845567207e-07, "loss": 0.8712, "step": 47474 }, { "epoch": 9.04, "learning_rate": 2.8741658722592944e-07, "loss": 0.5951, "step": 47475 }, { "epoch": 9.04, "learning_rate": 2.8735938989513823e-07, "loss": 0.7245, "step": 47476 }, { "epoch": 9.04, "learning_rate": 2.87302192564347e-07, "loss": 1.0202, "step": 47477 }, { "epoch": 9.04, "learning_rate": 2.8724499523355577e-07, "loss": 2.0839, "step": 47478 }, { "epoch": 9.04, "learning_rate": 2.8718779790276456e-07, "loss": 1.2447, "step": 47479 }, { "epoch": 9.04, "learning_rate": 2.871306005719733e-07, "loss": 1.206, "step": 47480 }, { "epoch": 9.04, "learning_rate": 2.870734032411821e-07, "loss": 1.6535, "step": 47481 }, { "epoch": 9.04, "learning_rate": 2.8701620591039084e-07, "loss": 0.946, "step": 47482 }, { "epoch": 9.04, "learning_rate": 2.8695900857959963e-07, "loss": 2.7042, "step": 47483 }, { "epoch": 9.04, "learning_rate": 2.869018112488084e-07, "loss": 1.0688, "step": 47484 }, { "epoch": 9.04, "learning_rate": 2.8684461391801716e-07, "loss": 1.1801, "step": 47485 }, { "epoch": 9.04, "learning_rate": 2.867874165872259e-07, "loss": 1.6836, "step": 47486 }, { "epoch": 9.05, "learning_rate": 2.867302192564347e-07, "loss": 1.9574, "step": 47487 }, { "epoch": 9.05, "learning_rate": 2.866730219256435e-07, "loss": 1.4013, "step": 47488 }, { "epoch": 9.05, "learning_rate": 2.8661582459485223e-07, "loss": 1.9218, "step": 47489 }, { "epoch": 9.05, "learning_rate": 2.8655862726406103e-07, "loss": 0.9703, "step": 47490 }, { "epoch": 9.05, "learning_rate": 2.865014299332698e-07, "loss": 1.6946, "step": 47491 }, { "epoch": 9.05, "learning_rate": 2.8644423260247856e-07, "loss": 1.9332, "step": 47492 }, { "epoch": 9.05, "learning_rate": 2.863870352716873e-07, "loss": 1.4061, "step": 47493 }, { "epoch": 9.05, "learning_rate": 2.863298379408961e-07, "loss": 0.5001, "step": 47494 }, { "epoch": 9.05, "learning_rate": 2.862726406101049e-07, "loss": 1.0744, "step": 47495 }, { "epoch": 9.05, "learning_rate": 2.8621544327931363e-07, "loss": 2.4507, "step": 47496 }, { "epoch": 9.05, "learning_rate": 2.861582459485224e-07, "loss": 0.564, "step": 47497 }, { "epoch": 9.05, "learning_rate": 2.8610104861773117e-07, "loss": 1.8598, "step": 47498 }, { "epoch": 9.05, "learning_rate": 2.8604385128693996e-07, "loss": 1.998, "step": 47499 }, { "epoch": 9.05, "learning_rate": 2.859866539561487e-07, "loss": 1.3817, "step": 47500 }, { "epoch": 9.05, "learning_rate": 2.859294566253575e-07, "loss": 3.1092, "step": 47501 }, { "epoch": 9.05, "learning_rate": 2.858722592945663e-07, "loss": 1.253, "step": 47502 }, { "epoch": 9.05, "learning_rate": 2.8581506196377503e-07, "loss": 2.1159, "step": 47503 }, { "epoch": 9.05, "learning_rate": 2.8575786463298377e-07, "loss": 2.4581, "step": 47504 }, { "epoch": 9.05, "learning_rate": 2.8570066730219256e-07, "loss": 2.2763, "step": 47505 }, { "epoch": 9.05, "learning_rate": 2.8564346997140136e-07, "loss": 2.598, "step": 47506 }, { "epoch": 9.05, "learning_rate": 2.855862726406101e-07, "loss": 1.8391, "step": 47507 }, { "epoch": 9.05, "learning_rate": 2.855290753098189e-07, "loss": 0.7821, "step": 47508 }, { "epoch": 9.05, "learning_rate": 2.854718779790277e-07, "loss": 1.7055, "step": 47509 }, { "epoch": 9.05, "learning_rate": 2.8541468064823643e-07, "loss": 1.6046, "step": 47510 }, { "epoch": 9.05, "learning_rate": 2.8535748331744517e-07, "loss": 2.2174, "step": 47511 }, { "epoch": 9.05, "learning_rate": 2.8530028598665396e-07, "loss": 3.2901, "step": 47512 }, { "epoch": 9.05, "learning_rate": 2.8524308865586276e-07, "loss": 1.0462, "step": 47513 }, { "epoch": 9.05, "learning_rate": 2.851858913250715e-07, "loss": 1.389, "step": 47514 }, { "epoch": 9.05, "learning_rate": 2.851286939942803e-07, "loss": 0.8377, "step": 47515 }, { "epoch": 9.05, "learning_rate": 2.8507149666348903e-07, "loss": 0.2314, "step": 47516 }, { "epoch": 9.05, "learning_rate": 2.850142993326978e-07, "loss": 0.8387, "step": 47517 }, { "epoch": 9.05, "learning_rate": 2.8495710200190657e-07, "loss": 1.1449, "step": 47518 }, { "epoch": 9.05, "learning_rate": 2.8489990467111536e-07, "loss": 1.8018, "step": 47519 }, { "epoch": 9.05, "learning_rate": 2.8484270734032415e-07, "loss": 0.8804, "step": 47520 }, { "epoch": 9.05, "learning_rate": 2.847855100095329e-07, "loss": 3.0672, "step": 47521 }, { "epoch": 9.05, "learning_rate": 2.8472831267874164e-07, "loss": 1.9956, "step": 47522 }, { "epoch": 9.05, "learning_rate": 2.8467111534795043e-07, "loss": 2.4001, "step": 47523 }, { "epoch": 9.05, "learning_rate": 2.846139180171592e-07, "loss": 0.4246, "step": 47524 }, { "epoch": 9.05, "learning_rate": 2.8455672068636796e-07, "loss": 2.7412, "step": 47525 }, { "epoch": 9.05, "learning_rate": 2.8449952335557676e-07, "loss": 1.1154, "step": 47526 }, { "epoch": 9.05, "learning_rate": 2.8444232602478555e-07, "loss": 2.8526, "step": 47527 }, { "epoch": 9.05, "learning_rate": 2.843851286939943e-07, "loss": 2.1204, "step": 47528 }, { "epoch": 9.05, "learning_rate": 2.8432793136320303e-07, "loss": 1.9021, "step": 47529 }, { "epoch": 9.05, "learning_rate": 2.8427073403241183e-07, "loss": 0.6907, "step": 47530 }, { "epoch": 9.05, "learning_rate": 2.842135367016206e-07, "loss": 1.1024, "step": 47531 }, { "epoch": 9.05, "learning_rate": 2.8415633937082936e-07, "loss": 1.2287, "step": 47532 }, { "epoch": 9.05, "learning_rate": 2.8409914204003816e-07, "loss": 0.652, "step": 47533 }, { "epoch": 9.05, "learning_rate": 2.840419447092469e-07, "loss": 1.6163, "step": 47534 }, { "epoch": 9.05, "learning_rate": 2.839847473784557e-07, "loss": 3.1799, "step": 47535 }, { "epoch": 9.05, "learning_rate": 2.8392755004766443e-07, "loss": 1.7115, "step": 47536 }, { "epoch": 9.05, "learning_rate": 2.838703527168732e-07, "loss": 0.5317, "step": 47537 }, { "epoch": 9.05, "learning_rate": 2.83813155386082e-07, "loss": 2.3743, "step": 47538 }, { "epoch": 9.06, "learning_rate": 2.8375595805529076e-07, "loss": 1.9226, "step": 47539 }, { "epoch": 9.06, "learning_rate": 2.836987607244995e-07, "loss": 1.3559, "step": 47540 }, { "epoch": 9.06, "learning_rate": 2.836415633937083e-07, "loss": 1.3077, "step": 47541 }, { "epoch": 9.06, "learning_rate": 2.835843660629171e-07, "loss": 1.9268, "step": 47542 }, { "epoch": 9.06, "learning_rate": 2.8352716873212583e-07, "loss": 2.8144, "step": 47543 }, { "epoch": 9.06, "learning_rate": 2.834699714013346e-07, "loss": 1.8044, "step": 47544 }, { "epoch": 9.06, "learning_rate": 2.8341277407054336e-07, "loss": 0.7407, "step": 47545 }, { "epoch": 9.06, "learning_rate": 2.8335557673975216e-07, "loss": 2.7578, "step": 47546 }, { "epoch": 9.06, "learning_rate": 2.832983794089609e-07, "loss": 2.6948, "step": 47547 }, { "epoch": 9.06, "learning_rate": 2.832411820781697e-07, "loss": 0.6098, "step": 47548 }, { "epoch": 9.06, "learning_rate": 2.831839847473785e-07, "loss": 1.5528, "step": 47549 }, { "epoch": 9.06, "learning_rate": 2.8312678741658723e-07, "loss": 0.7622, "step": 47550 }, { "epoch": 9.06, "learning_rate": 2.8306959008579597e-07, "loss": 1.0576, "step": 47551 }, { "epoch": 9.06, "learning_rate": 2.8301239275500476e-07, "loss": 1.8697, "step": 47552 }, { "epoch": 9.06, "learning_rate": 2.8295519542421356e-07, "loss": 1.7365, "step": 47553 }, { "epoch": 9.06, "learning_rate": 2.828979980934223e-07, "loss": 1.3385, "step": 47554 }, { "epoch": 9.06, "learning_rate": 2.828408007626311e-07, "loss": 1.2038, "step": 47555 }, { "epoch": 9.06, "learning_rate": 2.827836034318399e-07, "loss": 1.603, "step": 47556 }, { "epoch": 9.06, "learning_rate": 2.827264061010486e-07, "loss": 1.0752, "step": 47557 }, { "epoch": 9.06, "learning_rate": 2.8266920877025737e-07, "loss": 1.5976, "step": 47558 }, { "epoch": 9.06, "learning_rate": 2.8261201143946616e-07, "loss": 1.043, "step": 47559 }, { "epoch": 9.06, "learning_rate": 2.8255481410867495e-07, "loss": 1.6997, "step": 47560 }, { "epoch": 9.06, "learning_rate": 2.824976167778837e-07, "loss": 1.6889, "step": 47561 }, { "epoch": 9.06, "learning_rate": 2.824404194470925e-07, "loss": 1.4283, "step": 47562 }, { "epoch": 9.06, "learning_rate": 2.8238322211630123e-07, "loss": 2.8229, "step": 47563 }, { "epoch": 9.06, "learning_rate": 2.8232602478551e-07, "loss": 1.2007, "step": 47564 }, { "epoch": 9.06, "learning_rate": 2.8226882745471876e-07, "loss": 1.079, "step": 47565 }, { "epoch": 9.06, "learning_rate": 2.8221163012392756e-07, "loss": 1.5445, "step": 47566 }, { "epoch": 9.06, "learning_rate": 2.8215443279313635e-07, "loss": 2.2186, "step": 47567 }, { "epoch": 9.06, "learning_rate": 2.820972354623451e-07, "loss": 1.9766, "step": 47568 }, { "epoch": 9.06, "learning_rate": 2.8204003813155383e-07, "loss": 2.2241, "step": 47569 }, { "epoch": 9.06, "learning_rate": 2.8198284080076263e-07, "loss": 2.2123, "step": 47570 }, { "epoch": 9.06, "learning_rate": 2.819256434699714e-07, "loss": 1.6558, "step": 47571 }, { "epoch": 9.06, "learning_rate": 2.8186844613918016e-07, "loss": 1.564, "step": 47572 }, { "epoch": 9.06, "learning_rate": 2.8181124880838896e-07, "loss": 2.1537, "step": 47573 }, { "epoch": 9.06, "learning_rate": 2.8175405147759775e-07, "loss": 0.794, "step": 47574 }, { "epoch": 9.06, "learning_rate": 2.816968541468065e-07, "loss": 0.6385, "step": 47575 }, { "epoch": 9.06, "learning_rate": 2.8163965681601523e-07, "loss": 1.3718, "step": 47576 }, { "epoch": 9.06, "learning_rate": 2.81582459485224e-07, "loss": 1.866, "step": 47577 }, { "epoch": 9.06, "learning_rate": 2.815252621544328e-07, "loss": 1.3475, "step": 47578 }, { "epoch": 9.06, "learning_rate": 2.8146806482364156e-07, "loss": 0.5691, "step": 47579 }, { "epoch": 9.06, "learning_rate": 2.8141086749285035e-07, "loss": 1.5487, "step": 47580 }, { "epoch": 9.06, "learning_rate": 2.813536701620591e-07, "loss": 2.097, "step": 47581 }, { "epoch": 9.06, "learning_rate": 2.812964728312679e-07, "loss": 1.6434, "step": 47582 }, { "epoch": 9.06, "learning_rate": 2.8123927550047663e-07, "loss": 0.7062, "step": 47583 }, { "epoch": 9.06, "learning_rate": 2.811820781696854e-07, "loss": 1.6845, "step": 47584 }, { "epoch": 9.06, "learning_rate": 2.811248808388942e-07, "loss": 1.4427, "step": 47585 }, { "epoch": 9.06, "learning_rate": 2.8106768350810296e-07, "loss": 2.04, "step": 47586 }, { "epoch": 9.06, "learning_rate": 2.810104861773117e-07, "loss": 0.5853, "step": 47587 }, { "epoch": 9.06, "learning_rate": 2.809532888465205e-07, "loss": 2.9272, "step": 47588 }, { "epoch": 9.06, "learning_rate": 2.808960915157293e-07, "loss": 1.1587, "step": 47589 }, { "epoch": 9.06, "learning_rate": 2.8083889418493803e-07, "loss": 1.6085, "step": 47590 }, { "epoch": 9.06, "learning_rate": 2.807816968541468e-07, "loss": 2.0766, "step": 47591 }, { "epoch": 9.07, "learning_rate": 2.807244995233556e-07, "loss": 1.2232, "step": 47592 }, { "epoch": 9.07, "learning_rate": 2.8066730219256436e-07, "loss": 1.6962, "step": 47593 }, { "epoch": 9.07, "learning_rate": 2.806101048617731e-07, "loss": 2.1199, "step": 47594 }, { "epoch": 9.07, "learning_rate": 2.805529075309819e-07, "loss": 0.929, "step": 47595 }, { "epoch": 9.07, "learning_rate": 2.804957102001907e-07, "loss": 0.9101, "step": 47596 }, { "epoch": 9.07, "learning_rate": 2.804385128693994e-07, "loss": 2.9005, "step": 47597 }, { "epoch": 9.07, "learning_rate": 2.8038131553860817e-07, "loss": 1.3098, "step": 47598 }, { "epoch": 9.07, "learning_rate": 2.8032411820781696e-07, "loss": 1.7955, "step": 47599 }, { "epoch": 9.07, "learning_rate": 2.8026692087702575e-07, "loss": 0.9681, "step": 47600 }, { "epoch": 9.07, "learning_rate": 2.802097235462345e-07, "loss": 1.6959, "step": 47601 }, { "epoch": 9.07, "learning_rate": 2.801525262154433e-07, "loss": 1.5418, "step": 47602 }, { "epoch": 9.07, "learning_rate": 2.800953288846521e-07, "loss": 2.3558, "step": 47603 }, { "epoch": 9.07, "learning_rate": 2.800381315538608e-07, "loss": 2.0408, "step": 47604 }, { "epoch": 9.07, "learning_rate": 2.7998093422306956e-07, "loss": 2.3905, "step": 47605 }, { "epoch": 9.07, "learning_rate": 2.7992373689227836e-07, "loss": 2.3878, "step": 47606 }, { "epoch": 9.07, "learning_rate": 2.7986653956148715e-07, "loss": 0.5667, "step": 47607 }, { "epoch": 9.07, "learning_rate": 2.798093422306959e-07, "loss": 1.1219, "step": 47608 }, { "epoch": 9.07, "learning_rate": 2.797521448999047e-07, "loss": 1.8626, "step": 47609 }, { "epoch": 9.07, "learning_rate": 2.7969494756911343e-07, "loss": 1.4773, "step": 47610 }, { "epoch": 9.07, "learning_rate": 2.796377502383222e-07, "loss": 0.9593, "step": 47611 }, { "epoch": 9.07, "learning_rate": 2.7958055290753096e-07, "loss": 2.1719, "step": 47612 }, { "epoch": 9.07, "learning_rate": 2.7952335557673976e-07, "loss": 1.7458, "step": 47613 }, { "epoch": 9.07, "learning_rate": 2.7946615824594855e-07, "loss": 1.4661, "step": 47614 }, { "epoch": 9.07, "learning_rate": 2.794089609151573e-07, "loss": 1.481, "step": 47615 }, { "epoch": 9.07, "learning_rate": 2.7935176358436603e-07, "loss": 0.5616, "step": 47616 }, { "epoch": 9.07, "learning_rate": 2.792945662535748e-07, "loss": 0.7796, "step": 47617 }, { "epoch": 9.07, "learning_rate": 2.792373689227836e-07, "loss": 1.9286, "step": 47618 }, { "epoch": 9.07, "learning_rate": 2.7918017159199236e-07, "loss": 2.6269, "step": 47619 }, { "epoch": 9.07, "learning_rate": 2.7912297426120115e-07, "loss": 2.1284, "step": 47620 }, { "epoch": 9.07, "learning_rate": 2.7906577693040995e-07, "loss": 1.3383, "step": 47621 }, { "epoch": 9.07, "learning_rate": 2.790085795996187e-07, "loss": 1.2543, "step": 47622 }, { "epoch": 9.07, "learning_rate": 2.7895138226882743e-07, "loss": 0.5924, "step": 47623 }, { "epoch": 9.07, "learning_rate": 2.788941849380362e-07, "loss": 0.8975, "step": 47624 }, { "epoch": 9.07, "learning_rate": 2.78836987607245e-07, "loss": 1.9844, "step": 47625 }, { "epoch": 9.07, "learning_rate": 2.7877979027645376e-07, "loss": 0.9969, "step": 47626 }, { "epoch": 9.07, "learning_rate": 2.7872259294566255e-07, "loss": 1.8748, "step": 47627 }, { "epoch": 9.07, "learning_rate": 2.786653956148713e-07, "loss": 0.9349, "step": 47628 }, { "epoch": 9.07, "learning_rate": 2.786081982840801e-07, "loss": 1.0302, "step": 47629 }, { "epoch": 9.07, "learning_rate": 2.7855100095328883e-07, "loss": 1.1889, "step": 47630 }, { "epoch": 9.07, "learning_rate": 2.784938036224976e-07, "loss": 3.1418, "step": 47631 }, { "epoch": 9.07, "learning_rate": 2.784366062917064e-07, "loss": 2.0434, "step": 47632 }, { "epoch": 9.07, "learning_rate": 2.783794089609152e-07, "loss": 1.7462, "step": 47633 }, { "epoch": 9.07, "learning_rate": 2.783222116301239e-07, "loss": 1.3524, "step": 47634 }, { "epoch": 9.07, "learning_rate": 2.782650142993327e-07, "loss": 2.0693, "step": 47635 }, { "epoch": 9.07, "learning_rate": 2.782078169685415e-07, "loss": 1.1303, "step": 47636 }, { "epoch": 9.07, "learning_rate": 2.781506196377502e-07, "loss": 1.7183, "step": 47637 }, { "epoch": 9.07, "learning_rate": 2.78093422306959e-07, "loss": 1.3871, "step": 47638 }, { "epoch": 9.07, "learning_rate": 2.780362249761678e-07, "loss": 1.9513, "step": 47639 }, { "epoch": 9.07, "learning_rate": 2.7797902764537655e-07, "loss": 1.1601, "step": 47640 }, { "epoch": 9.07, "learning_rate": 2.779218303145853e-07, "loss": 3.1195, "step": 47641 }, { "epoch": 9.07, "learning_rate": 2.778646329837941e-07, "loss": 1.9828, "step": 47642 }, { "epoch": 9.07, "learning_rate": 2.778074356530029e-07, "loss": 0.8787, "step": 47643 }, { "epoch": 9.08, "learning_rate": 2.777502383222117e-07, "loss": 1.0462, "step": 47644 }, { "epoch": 9.08, "learning_rate": 2.776930409914204e-07, "loss": 2.9669, "step": 47645 }, { "epoch": 9.08, "learning_rate": 2.7763584366062916e-07, "loss": 1.549, "step": 47646 }, { "epoch": 9.08, "learning_rate": 2.7757864632983795e-07, "loss": 2.5239, "step": 47647 }, { "epoch": 9.08, "learning_rate": 2.775214489990467e-07, "loss": 2.3336, "step": 47648 }, { "epoch": 9.08, "learning_rate": 2.774642516682555e-07, "loss": 1.7029, "step": 47649 }, { "epoch": 9.08, "learning_rate": 2.774070543374643e-07, "loss": 2.0181, "step": 47650 }, { "epoch": 9.08, "learning_rate": 2.773498570066731e-07, "loss": 2.3672, "step": 47651 }, { "epoch": 9.08, "learning_rate": 2.7729265967588176e-07, "loss": 0.3263, "step": 47652 }, { "epoch": 9.08, "learning_rate": 2.7723546234509056e-07, "loss": 3.1312, "step": 47653 }, { "epoch": 9.08, "learning_rate": 2.7717826501429935e-07, "loss": 1.6269, "step": 47654 }, { "epoch": 9.08, "learning_rate": 2.7712106768350814e-07, "loss": 1.8558, "step": 47655 }, { "epoch": 9.08, "learning_rate": 2.770638703527169e-07, "loss": 1.2281, "step": 47656 }, { "epoch": 9.08, "learning_rate": 2.770066730219256e-07, "loss": 1.7775, "step": 47657 }, { "epoch": 9.08, "learning_rate": 2.769494756911344e-07, "loss": 1.7039, "step": 47658 }, { "epoch": 9.08, "learning_rate": 2.7689227836034316e-07, "loss": 2.6978, "step": 47659 }, { "epoch": 9.08, "learning_rate": 2.7683508102955195e-07, "loss": 1.306, "step": 47660 }, { "epoch": 9.08, "learning_rate": 2.7677788369876075e-07, "loss": 0.5762, "step": 47661 }, { "epoch": 9.08, "learning_rate": 2.7672068636796954e-07, "loss": 2.1313, "step": 47662 }, { "epoch": 9.08, "learning_rate": 2.7666348903717823e-07, "loss": 1.2964, "step": 47663 }, { "epoch": 9.08, "learning_rate": 2.76606291706387e-07, "loss": 2.0321, "step": 47664 }, { "epoch": 9.08, "learning_rate": 2.765490943755958e-07, "loss": 0.3134, "step": 47665 }, { "epoch": 9.08, "learning_rate": 2.7649189704480456e-07, "loss": 1.6298, "step": 47666 }, { "epoch": 9.08, "learning_rate": 2.7643469971401335e-07, "loss": 1.0826, "step": 47667 }, { "epoch": 9.08, "learning_rate": 2.7637750238322215e-07, "loss": 1.5724, "step": 47668 }, { "epoch": 9.08, "learning_rate": 2.763203050524309e-07, "loss": 1.3565, "step": 47669 }, { "epoch": 9.08, "learning_rate": 2.7626310772163963e-07, "loss": 2.4397, "step": 47670 }, { "epoch": 9.08, "learning_rate": 2.762059103908484e-07, "loss": 1.0193, "step": 47671 }, { "epoch": 9.08, "learning_rate": 2.761487130600572e-07, "loss": 1.5811, "step": 47672 }, { "epoch": 9.08, "learning_rate": 2.76091515729266e-07, "loss": 2.2506, "step": 47673 }, { "epoch": 9.08, "learning_rate": 2.7603431839847475e-07, "loss": 1.0054, "step": 47674 }, { "epoch": 9.08, "learning_rate": 2.759771210676835e-07, "loss": 1.4223, "step": 47675 }, { "epoch": 9.08, "learning_rate": 2.759199237368923e-07, "loss": 1.5839, "step": 47676 }, { "epoch": 9.08, "learning_rate": 2.75862726406101e-07, "loss": 2.0338, "step": 47677 }, { "epoch": 9.08, "learning_rate": 2.758055290753098e-07, "loss": 2.5712, "step": 47678 }, { "epoch": 9.08, "learning_rate": 2.757483317445186e-07, "loss": 0.7139, "step": 47679 }, { "epoch": 9.08, "learning_rate": 2.756911344137274e-07, "loss": 1.3632, "step": 47680 }, { "epoch": 9.08, "learning_rate": 2.756339370829361e-07, "loss": 0.8578, "step": 47681 }, { "epoch": 9.08, "learning_rate": 2.755767397521449e-07, "loss": 1.913, "step": 47682 }, { "epoch": 9.08, "learning_rate": 2.755195424213537e-07, "loss": 2.2, "step": 47683 }, { "epoch": 9.08, "learning_rate": 2.754623450905625e-07, "loss": 1.9163, "step": 47684 }, { "epoch": 9.08, "learning_rate": 2.754051477597712e-07, "loss": 1.7016, "step": 47685 }, { "epoch": 9.08, "learning_rate": 2.7534795042898e-07, "loss": 1.93, "step": 47686 }, { "epoch": 9.08, "learning_rate": 2.7529075309818875e-07, "loss": 2.3648, "step": 47687 }, { "epoch": 9.08, "learning_rate": 2.752335557673975e-07, "loss": 0.3055, "step": 47688 }, { "epoch": 9.08, "learning_rate": 2.751763584366063e-07, "loss": 1.8926, "step": 47689 }, { "epoch": 9.08, "learning_rate": 2.751191611058151e-07, "loss": 0.9244, "step": 47690 }, { "epoch": 9.08, "learning_rate": 2.750619637750239e-07, "loss": 2.334, "step": 47691 }, { "epoch": 9.08, "learning_rate": 2.750047664442326e-07, "loss": 1.2855, "step": 47692 }, { "epoch": 9.08, "learning_rate": 2.7494756911344136e-07, "loss": 3.1192, "step": 47693 }, { "epoch": 9.08, "learning_rate": 2.7489037178265015e-07, "loss": 1.4226, "step": 47694 }, { "epoch": 9.08, "learning_rate": 2.7483317445185894e-07, "loss": 1.6512, "step": 47695 }, { "epoch": 9.08, "learning_rate": 2.747759771210677e-07, "loss": 1.3076, "step": 47696 }, { "epoch": 9.09, "learning_rate": 2.747187797902765e-07, "loss": 2.0886, "step": 47697 }, { "epoch": 9.09, "learning_rate": 2.7466158245948527e-07, "loss": 2.5855, "step": 47698 }, { "epoch": 9.09, "learning_rate": 2.7460438512869396e-07, "loss": 1.6174, "step": 47699 }, { "epoch": 9.09, "learning_rate": 2.7454718779790275e-07, "loss": 2.0929, "step": 47700 }, { "epoch": 9.09, "learning_rate": 2.7448999046711155e-07, "loss": 1.4825, "step": 47701 }, { "epoch": 9.09, "learning_rate": 2.7443279313632034e-07, "loss": 1.8121, "step": 47702 }, { "epoch": 9.09, "learning_rate": 2.743755958055291e-07, "loss": 2.3506, "step": 47703 }, { "epoch": 9.09, "learning_rate": 2.743183984747379e-07, "loss": 1.0078, "step": 47704 }, { "epoch": 9.09, "learning_rate": 2.742612011439466e-07, "loss": 1.5767, "step": 47705 }, { "epoch": 9.09, "learning_rate": 2.742040038131554e-07, "loss": 2.8139, "step": 47706 }, { "epoch": 9.09, "learning_rate": 2.7414680648236415e-07, "loss": 0.4164, "step": 47707 }, { "epoch": 9.09, "learning_rate": 2.7408960915157295e-07, "loss": 1.1883, "step": 47708 }, { "epoch": 9.09, "learning_rate": 2.7403241182078174e-07, "loss": 2.6827, "step": 47709 }, { "epoch": 9.09, "learning_rate": 2.739752144899905e-07, "loss": 2.6002, "step": 47710 }, { "epoch": 9.09, "learning_rate": 2.739180171591992e-07, "loss": 2.2058, "step": 47711 }, { "epoch": 9.09, "learning_rate": 2.73860819828408e-07, "loss": 2.0141, "step": 47712 }, { "epoch": 9.09, "learning_rate": 2.738036224976168e-07, "loss": 1.0667, "step": 47713 }, { "epoch": 9.09, "learning_rate": 2.7374642516682555e-07, "loss": 1.6621, "step": 47714 }, { "epoch": 9.09, "learning_rate": 2.7368922783603434e-07, "loss": 1.7687, "step": 47715 }, { "epoch": 9.09, "learning_rate": 2.736320305052431e-07, "loss": 0.6031, "step": 47716 }, { "epoch": 9.09, "learning_rate": 2.735748331744519e-07, "loss": 1.5265, "step": 47717 }, { "epoch": 9.09, "learning_rate": 2.735176358436606e-07, "loss": 1.886, "step": 47718 }, { "epoch": 9.09, "learning_rate": 2.734604385128694e-07, "loss": 1.2919, "step": 47719 }, { "epoch": 9.09, "learning_rate": 2.734032411820782e-07, "loss": 1.8251, "step": 47720 }, { "epoch": 9.09, "learning_rate": 2.7334604385128695e-07, "loss": 0.7906, "step": 47721 }, { "epoch": 9.09, "learning_rate": 2.732888465204957e-07, "loss": 1.2266, "step": 47722 }, { "epoch": 9.09, "learning_rate": 2.732316491897045e-07, "loss": 1.0732, "step": 47723 }, { "epoch": 9.09, "learning_rate": 2.731744518589133e-07, "loss": 0.5098, "step": 47724 }, { "epoch": 9.09, "learning_rate": 2.73117254528122e-07, "loss": 1.6141, "step": 47725 }, { "epoch": 9.09, "learning_rate": 2.730600571973308e-07, "loss": 0.471, "step": 47726 }, { "epoch": 9.09, "learning_rate": 2.730028598665396e-07, "loss": 1.3813, "step": 47727 }, { "epoch": 9.09, "learning_rate": 2.7294566253574834e-07, "loss": 2.8902, "step": 47728 }, { "epoch": 9.09, "learning_rate": 2.728884652049571e-07, "loss": 2.4095, "step": 47729 }, { "epoch": 9.09, "learning_rate": 2.728312678741659e-07, "loss": 0.8934, "step": 47730 }, { "epoch": 9.09, "learning_rate": 2.7277407054337467e-07, "loss": 1.5688, "step": 47731 }, { "epoch": 9.09, "learning_rate": 2.727168732125834e-07, "loss": 0.7548, "step": 47732 }, { "epoch": 9.09, "learning_rate": 2.726596758817922e-07, "loss": 2.1847, "step": 47733 }, { "epoch": 9.09, "learning_rate": 2.7260247855100095e-07, "loss": 1.1389, "step": 47734 }, { "epoch": 9.09, "learning_rate": 2.7254528122020974e-07, "loss": 2.4901, "step": 47735 }, { "epoch": 9.09, "learning_rate": 2.724880838894185e-07, "loss": 1.2394, "step": 47736 }, { "epoch": 9.09, "learning_rate": 2.724308865586273e-07, "loss": 1.9372, "step": 47737 }, { "epoch": 9.09, "learning_rate": 2.7237368922783607e-07, "loss": 3.1061, "step": 47738 }, { "epoch": 9.09, "learning_rate": 2.723164918970448e-07, "loss": 1.5527, "step": 47739 }, { "epoch": 9.09, "learning_rate": 2.7225929456625355e-07, "loss": 2.5292, "step": 47740 }, { "epoch": 9.09, "learning_rate": 2.7220209723546235e-07, "loss": 0.7627, "step": 47741 }, { "epoch": 9.09, "learning_rate": 2.7214489990467114e-07, "loss": 2.2379, "step": 47742 }, { "epoch": 9.09, "learning_rate": 2.720877025738799e-07, "loss": 2.1099, "step": 47743 }, { "epoch": 9.09, "learning_rate": 2.720305052430887e-07, "loss": 2.2852, "step": 47744 }, { "epoch": 9.09, "learning_rate": 2.7197330791229747e-07, "loss": 0.2576, "step": 47745 }, { "epoch": 9.09, "learning_rate": 2.719161105815062e-07, "loss": 2.2445, "step": 47746 }, { "epoch": 9.09, "learning_rate": 2.7185891325071495e-07, "loss": 1.0346, "step": 47747 }, { "epoch": 9.09, "learning_rate": 2.7180171591992374e-07, "loss": 0.9902, "step": 47748 }, { "epoch": 9.1, "learning_rate": 2.7174451858913254e-07, "loss": 0.53, "step": 47749 }, { "epoch": 9.1, "learning_rate": 2.716873212583413e-07, "loss": 1.4851, "step": 47750 }, { "epoch": 9.1, "learning_rate": 2.7163012392755007e-07, "loss": 1.7563, "step": 47751 }, { "epoch": 9.1, "learning_rate": 2.715729265967588e-07, "loss": 0.669, "step": 47752 }, { "epoch": 9.1, "learning_rate": 2.715157292659676e-07, "loss": 0.9615, "step": 47753 }, { "epoch": 9.1, "learning_rate": 2.7145853193517635e-07, "loss": 1.49, "step": 47754 }, { "epoch": 9.1, "learning_rate": 2.7140133460438514e-07, "loss": 1.2933, "step": 47755 }, { "epoch": 9.1, "learning_rate": 2.7134413727359394e-07, "loss": 0.942, "step": 47756 }, { "epoch": 9.1, "learning_rate": 2.712869399428027e-07, "loss": 1.0657, "step": 47757 }, { "epoch": 9.1, "learning_rate": 2.712297426120114e-07, "loss": 1.5293, "step": 47758 }, { "epoch": 9.1, "learning_rate": 2.711725452812202e-07, "loss": 0.8165, "step": 47759 }, { "epoch": 9.1, "learning_rate": 2.71115347950429e-07, "loss": 2.2686, "step": 47760 }, { "epoch": 9.1, "learning_rate": 2.7105815061963775e-07, "loss": 0.4907, "step": 47761 }, { "epoch": 9.1, "learning_rate": 2.7100095328884654e-07, "loss": 3.066, "step": 47762 }, { "epoch": 9.1, "learning_rate": 2.7094375595805533e-07, "loss": 1.6484, "step": 47763 }, { "epoch": 9.1, "learning_rate": 2.708865586272641e-07, "loss": 1.2669, "step": 47764 }, { "epoch": 9.1, "learning_rate": 2.708293612964728e-07, "loss": 2.3262, "step": 47765 }, { "epoch": 9.1, "learning_rate": 2.707721639656816e-07, "loss": 1.2936, "step": 47766 }, { "epoch": 9.1, "learning_rate": 2.707149666348904e-07, "loss": 1.0366, "step": 47767 }, { "epoch": 9.1, "learning_rate": 2.7065776930409914e-07, "loss": 1.8439, "step": 47768 }, { "epoch": 9.1, "learning_rate": 2.7060057197330794e-07, "loss": 2.0549, "step": 47769 }, { "epoch": 9.1, "learning_rate": 2.705433746425167e-07, "loss": 2.6055, "step": 47770 }, { "epoch": 9.1, "learning_rate": 2.7048617731172547e-07, "loss": 1.4195, "step": 47771 }, { "epoch": 9.1, "learning_rate": 2.704289799809342e-07, "loss": 2.2614, "step": 47772 }, { "epoch": 9.1, "learning_rate": 2.70371782650143e-07, "loss": 0.2193, "step": 47773 }, { "epoch": 9.1, "learning_rate": 2.703145853193518e-07, "loss": 0.2225, "step": 47774 }, { "epoch": 9.1, "learning_rate": 2.7025738798856054e-07, "loss": 2.0887, "step": 47775 }, { "epoch": 9.1, "learning_rate": 2.702001906577693e-07, "loss": 1.2564, "step": 47776 }, { "epoch": 9.1, "learning_rate": 2.701429933269781e-07, "loss": 2.0741, "step": 47777 }, { "epoch": 9.1, "learning_rate": 2.7008579599618687e-07, "loss": 1.3607, "step": 47778 }, { "epoch": 9.1, "learning_rate": 2.700285986653956e-07, "loss": 2.4816, "step": 47779 }, { "epoch": 9.1, "learning_rate": 2.699714013346044e-07, "loss": 1.7435, "step": 47780 }, { "epoch": 9.1, "learning_rate": 2.6991420400381315e-07, "loss": 1.629, "step": 47781 }, { "epoch": 9.1, "learning_rate": 2.6985700667302194e-07, "loss": 0.7391, "step": 47782 }, { "epoch": 9.1, "learning_rate": 2.697998093422307e-07, "loss": 2.8805, "step": 47783 }, { "epoch": 9.1, "learning_rate": 2.697426120114395e-07, "loss": 1.7677, "step": 47784 }, { "epoch": 9.1, "learning_rate": 2.6968541468064827e-07, "loss": 2.0016, "step": 47785 }, { "epoch": 9.1, "learning_rate": 2.69628217349857e-07, "loss": 1.9925, "step": 47786 }, { "epoch": 9.1, "learning_rate": 2.6957102001906575e-07, "loss": 2.0517, "step": 47787 }, { "epoch": 9.1, "learning_rate": 2.6951382268827454e-07, "loss": 2.0056, "step": 47788 }, { "epoch": 9.1, "learning_rate": 2.6945662535748334e-07, "loss": 1.1083, "step": 47789 }, { "epoch": 9.1, "learning_rate": 2.693994280266921e-07, "loss": 1.3733, "step": 47790 }, { "epoch": 9.1, "learning_rate": 2.6934223069590087e-07, "loss": 0.2984, "step": 47791 }, { "epoch": 9.1, "learning_rate": 2.6928503336510967e-07, "loss": 1.5558, "step": 47792 }, { "epoch": 9.1, "learning_rate": 2.692278360343184e-07, "loss": 0.8547, "step": 47793 }, { "epoch": 9.1, "learning_rate": 2.6917063870352715e-07, "loss": 2.3052, "step": 47794 }, { "epoch": 9.1, "learning_rate": 2.6911344137273594e-07, "loss": 1.5219, "step": 47795 }, { "epoch": 9.1, "learning_rate": 2.6905624404194474e-07, "loss": 2.2598, "step": 47796 }, { "epoch": 9.1, "learning_rate": 2.689990467111535e-07, "loss": 0.3631, "step": 47797 }, { "epoch": 9.1, "learning_rate": 2.6894184938036227e-07, "loss": 1.7787, "step": 47798 }, { "epoch": 9.1, "learning_rate": 2.68884652049571e-07, "loss": 0.4422, "step": 47799 }, { "epoch": 9.1, "learning_rate": 2.688274547187798e-07, "loss": 2.1006, "step": 47800 }, { "epoch": 9.1, "learning_rate": 2.6877025738798855e-07, "loss": 1.9592, "step": 47801 }, { "epoch": 9.11, "learning_rate": 2.6871306005719734e-07, "loss": 0.9976, "step": 47802 }, { "epoch": 9.11, "learning_rate": 2.6865586272640613e-07, "loss": 1.6638, "step": 47803 }, { "epoch": 9.11, "learning_rate": 2.685986653956149e-07, "loss": 2.4969, "step": 47804 }, { "epoch": 9.11, "learning_rate": 2.685414680648236e-07, "loss": 1.7851, "step": 47805 }, { "epoch": 9.11, "learning_rate": 2.684842707340324e-07, "loss": 2.3246, "step": 47806 }, { "epoch": 9.11, "learning_rate": 2.684270734032412e-07, "loss": 2.0192, "step": 47807 }, { "epoch": 9.11, "learning_rate": 2.6836987607244994e-07, "loss": 2.0383, "step": 47808 }, { "epoch": 9.11, "learning_rate": 2.6831267874165874e-07, "loss": 1.5351, "step": 47809 }, { "epoch": 9.11, "learning_rate": 2.6825548141086753e-07, "loss": 1.1747, "step": 47810 }, { "epoch": 9.11, "learning_rate": 2.6819828408007627e-07, "loss": 2.0794, "step": 47811 }, { "epoch": 9.11, "learning_rate": 2.68141086749285e-07, "loss": 0.3246, "step": 47812 }, { "epoch": 9.11, "learning_rate": 2.680838894184938e-07, "loss": 2.0743, "step": 47813 }, { "epoch": 9.11, "learning_rate": 2.680266920877026e-07, "loss": 1.0269, "step": 47814 }, { "epoch": 9.11, "learning_rate": 2.6796949475691134e-07, "loss": 1.8298, "step": 47815 }, { "epoch": 9.11, "learning_rate": 2.6791229742612014e-07, "loss": 1.1341, "step": 47816 }, { "epoch": 9.11, "learning_rate": 2.678551000953289e-07, "loss": 0.9548, "step": 47817 }, { "epoch": 9.11, "learning_rate": 2.6779790276453767e-07, "loss": 0.9083, "step": 47818 }, { "epoch": 9.11, "learning_rate": 2.677407054337464e-07, "loss": 2.6638, "step": 47819 }, { "epoch": 9.11, "learning_rate": 2.676835081029552e-07, "loss": 1.1956, "step": 47820 }, { "epoch": 9.11, "learning_rate": 2.67626310772164e-07, "loss": 0.7833, "step": 47821 }, { "epoch": 9.11, "learning_rate": 2.6756911344137274e-07, "loss": 2.1328, "step": 47822 }, { "epoch": 9.11, "learning_rate": 2.675119161105815e-07, "loss": 0.691, "step": 47823 }, { "epoch": 9.11, "learning_rate": 2.674547187797903e-07, "loss": 2.5262, "step": 47824 }, { "epoch": 9.11, "learning_rate": 2.6739752144899907e-07, "loss": 2.4629, "step": 47825 }, { "epoch": 9.11, "learning_rate": 2.673403241182078e-07, "loss": 0.4472, "step": 47826 }, { "epoch": 9.11, "learning_rate": 2.672831267874166e-07, "loss": 1.4983, "step": 47827 }, { "epoch": 9.11, "learning_rate": 2.672259294566254e-07, "loss": 0.7595, "step": 47828 }, { "epoch": 9.11, "learning_rate": 2.6716873212583414e-07, "loss": 1.9759, "step": 47829 }, { "epoch": 9.11, "learning_rate": 2.671115347950429e-07, "loss": 1.1342, "step": 47830 }, { "epoch": 9.11, "learning_rate": 2.6705433746425167e-07, "loss": 0.4604, "step": 47831 }, { "epoch": 9.11, "learning_rate": 2.6699714013346047e-07, "loss": 1.2324, "step": 47832 }, { "epoch": 9.11, "learning_rate": 2.669399428026692e-07, "loss": 3.7763, "step": 47833 }, { "epoch": 9.11, "learning_rate": 2.66882745471878e-07, "loss": 2.251, "step": 47834 }, { "epoch": 9.11, "learning_rate": 2.6682554814108674e-07, "loss": 1.1274, "step": 47835 }, { "epoch": 9.11, "learning_rate": 2.6676835081029554e-07, "loss": 0.3306, "step": 47836 }, { "epoch": 9.11, "learning_rate": 2.667111534795043e-07, "loss": 1.3154, "step": 47837 }, { "epoch": 9.11, "learning_rate": 2.6665395614871307e-07, "loss": 2.0169, "step": 47838 }, { "epoch": 9.11, "learning_rate": 2.6659675881792186e-07, "loss": 1.4715, "step": 47839 }, { "epoch": 9.11, "learning_rate": 2.665395614871306e-07, "loss": 0.7852, "step": 47840 }, { "epoch": 9.11, "learning_rate": 2.6648236415633935e-07, "loss": 1.6629, "step": 47841 }, { "epoch": 9.11, "learning_rate": 2.6642516682554814e-07, "loss": 1.055, "step": 47842 }, { "epoch": 9.11, "learning_rate": 2.6636796949475693e-07, "loss": 2.9714, "step": 47843 }, { "epoch": 9.11, "learning_rate": 2.663107721639657e-07, "loss": 1.9169, "step": 47844 }, { "epoch": 9.11, "learning_rate": 2.6625357483317447e-07, "loss": 1.0281, "step": 47845 }, { "epoch": 9.11, "learning_rate": 2.661963775023832e-07, "loss": 1.7159, "step": 47846 }, { "epoch": 9.11, "learning_rate": 2.66139180171592e-07, "loss": 1.2589, "step": 47847 }, { "epoch": 9.11, "learning_rate": 2.6608198284080074e-07, "loss": 1.3609, "step": 47848 }, { "epoch": 9.11, "learning_rate": 2.6602478551000954e-07, "loss": 0.5129, "step": 47849 }, { "epoch": 9.11, "learning_rate": 2.6596758817921833e-07, "loss": 0.6972, "step": 47850 }, { "epoch": 9.11, "learning_rate": 2.6591039084842707e-07, "loss": 2.4321, "step": 47851 }, { "epoch": 9.11, "learning_rate": 2.658531935176358e-07, "loss": 0.9263, "step": 47852 }, { "epoch": 9.11, "learning_rate": 2.657959961868446e-07, "loss": 1.9757, "step": 47853 }, { "epoch": 9.12, "learning_rate": 2.657387988560534e-07, "loss": 0.334, "step": 47854 }, { "epoch": 9.12, "learning_rate": 2.6568160152526214e-07, "loss": 1.9894, "step": 47855 }, { "epoch": 9.12, "learning_rate": 2.6562440419447094e-07, "loss": 1.8919, "step": 47856 }, { "epoch": 9.12, "learning_rate": 2.6556720686367973e-07, "loss": 2.7061, "step": 47857 }, { "epoch": 9.12, "learning_rate": 2.6551000953288847e-07, "loss": 1.2959, "step": 47858 }, { "epoch": 9.12, "learning_rate": 2.654528122020972e-07, "loss": 1.6106, "step": 47859 }, { "epoch": 9.12, "learning_rate": 2.65395614871306e-07, "loss": 2.2842, "step": 47860 }, { "epoch": 9.12, "learning_rate": 2.653384175405148e-07, "loss": 1.236, "step": 47861 }, { "epoch": 9.12, "learning_rate": 2.6528122020972354e-07, "loss": 1.6687, "step": 47862 }, { "epoch": 9.12, "learning_rate": 2.6522402287893233e-07, "loss": 0.4191, "step": 47863 }, { "epoch": 9.12, "learning_rate": 2.651668255481411e-07, "loss": 1.1662, "step": 47864 }, { "epoch": 9.12, "learning_rate": 2.6510962821734987e-07, "loss": 0.4252, "step": 47865 }, { "epoch": 9.12, "learning_rate": 2.650524308865586e-07, "loss": 0.98, "step": 47866 }, { "epoch": 9.12, "learning_rate": 2.649952335557674e-07, "loss": 1.5931, "step": 47867 }, { "epoch": 9.12, "learning_rate": 2.649380362249762e-07, "loss": 1.4131, "step": 47868 }, { "epoch": 9.12, "learning_rate": 2.6488083889418494e-07, "loss": 2.2565, "step": 47869 }, { "epoch": 9.12, "learning_rate": 2.648236415633937e-07, "loss": 1.6847, "step": 47870 }, { "epoch": 9.12, "learning_rate": 2.6476644423260247e-07, "loss": 1.6276, "step": 47871 }, { "epoch": 9.12, "learning_rate": 2.6470924690181127e-07, "loss": 1.5847, "step": 47872 }, { "epoch": 9.12, "learning_rate": 2.6465204957102e-07, "loss": 1.2709, "step": 47873 }, { "epoch": 9.12, "learning_rate": 2.645948522402288e-07, "loss": 1.105, "step": 47874 }, { "epoch": 9.12, "learning_rate": 2.645376549094376e-07, "loss": 1.1969, "step": 47875 }, { "epoch": 9.12, "learning_rate": 2.6448045757864634e-07, "loss": 0.9176, "step": 47876 }, { "epoch": 9.12, "learning_rate": 2.644232602478551e-07, "loss": 0.7278, "step": 47877 }, { "epoch": 9.12, "learning_rate": 2.6436606291706387e-07, "loss": 1.6505, "step": 47878 }, { "epoch": 9.12, "learning_rate": 2.6430886558627266e-07, "loss": 1.2354, "step": 47879 }, { "epoch": 9.12, "learning_rate": 2.642516682554814e-07, "loss": 1.0552, "step": 47880 }, { "epoch": 9.12, "learning_rate": 2.641944709246902e-07, "loss": 1.5034, "step": 47881 }, { "epoch": 9.12, "learning_rate": 2.6413727359389894e-07, "loss": 0.9103, "step": 47882 }, { "epoch": 9.12, "learning_rate": 2.6408007626310773e-07, "loss": 2.7875, "step": 47883 }, { "epoch": 9.12, "learning_rate": 2.640228789323165e-07, "loss": 1.7579, "step": 47884 }, { "epoch": 9.12, "learning_rate": 2.6396568160152527e-07, "loss": 1.8352, "step": 47885 }, { "epoch": 9.12, "learning_rate": 2.6390848427073406e-07, "loss": 0.8171, "step": 47886 }, { "epoch": 9.12, "learning_rate": 2.638512869399428e-07, "loss": 1.7331, "step": 47887 }, { "epoch": 9.12, "learning_rate": 2.6379408960915154e-07, "loss": 2.1727, "step": 47888 }, { "epoch": 9.12, "learning_rate": 2.6373689227836034e-07, "loss": 1.7004, "step": 47889 }, { "epoch": 9.12, "learning_rate": 2.6367969494756913e-07, "loss": 2.0877, "step": 47890 }, { "epoch": 9.12, "learning_rate": 2.6362249761677787e-07, "loss": 1.2057, "step": 47891 }, { "epoch": 9.12, "learning_rate": 2.6356530028598667e-07, "loss": 0.9162, "step": 47892 }, { "epoch": 9.12, "learning_rate": 2.6350810295519546e-07, "loss": 0.9993, "step": 47893 }, { "epoch": 9.12, "learning_rate": 2.634509056244042e-07, "loss": 0.6062, "step": 47894 }, { "epoch": 9.12, "learning_rate": 2.6339370829361294e-07, "loss": 0.4285, "step": 47895 }, { "epoch": 9.12, "learning_rate": 2.6333651096282174e-07, "loss": 1.3326, "step": 47896 }, { "epoch": 9.12, "learning_rate": 2.6327931363203053e-07, "loss": 2.3322, "step": 47897 }, { "epoch": 9.12, "learning_rate": 2.6322211630123927e-07, "loss": 1.4628, "step": 47898 }, { "epoch": 9.12, "learning_rate": 2.63164918970448e-07, "loss": 0.887, "step": 47899 }, { "epoch": 9.12, "learning_rate": 2.631077216396568e-07, "loss": 0.4025, "step": 47900 }, { "epoch": 9.12, "learning_rate": 2.630505243088656e-07, "loss": 1.6288, "step": 47901 }, { "epoch": 9.12, "learning_rate": 2.6299332697807434e-07, "loss": 1.64, "step": 47902 }, { "epoch": 9.12, "learning_rate": 2.6293612964728313e-07, "loss": 1.7619, "step": 47903 }, { "epoch": 9.12, "learning_rate": 2.6287893231649193e-07, "loss": 1.4669, "step": 47904 }, { "epoch": 9.12, "learning_rate": 2.6282173498570067e-07, "loss": 1.07, "step": 47905 }, { "epoch": 9.12, "learning_rate": 2.627645376549094e-07, "loss": 1.3111, "step": 47906 }, { "epoch": 9.13, "learning_rate": 2.627073403241182e-07, "loss": 2.08, "step": 47907 }, { "epoch": 9.13, "learning_rate": 2.62650142993327e-07, "loss": 0.8695, "step": 47908 }, { "epoch": 9.13, "learning_rate": 2.6259294566253574e-07, "loss": 1.1634, "step": 47909 }, { "epoch": 9.13, "learning_rate": 2.6253574833174453e-07, "loss": 1.1079, "step": 47910 }, { "epoch": 9.13, "learning_rate": 2.6247855100095327e-07, "loss": 1.365, "step": 47911 }, { "epoch": 9.13, "learning_rate": 2.6242135367016207e-07, "loss": 0.8214, "step": 47912 }, { "epoch": 9.13, "learning_rate": 2.623641563393708e-07, "loss": 0.7751, "step": 47913 }, { "epoch": 9.13, "learning_rate": 2.623069590085796e-07, "loss": 2.0763, "step": 47914 }, { "epoch": 9.13, "learning_rate": 2.622497616777884e-07, "loss": 1.2341, "step": 47915 }, { "epoch": 9.13, "learning_rate": 2.621925643469972e-07, "loss": 1.3872, "step": 47916 }, { "epoch": 9.13, "learning_rate": 2.621353670162059e-07, "loss": 1.4821, "step": 47917 }, { "epoch": 9.13, "learning_rate": 2.6207816968541467e-07, "loss": 1.6179, "step": 47918 }, { "epoch": 9.13, "learning_rate": 2.6202097235462346e-07, "loss": 1.2809, "step": 47919 }, { "epoch": 9.13, "learning_rate": 2.619637750238322e-07, "loss": 1.1962, "step": 47920 }, { "epoch": 9.13, "learning_rate": 2.61906577693041e-07, "loss": 0.7918, "step": 47921 }, { "epoch": 9.13, "learning_rate": 2.618493803622498e-07, "loss": 1.0024, "step": 47922 }, { "epoch": 9.13, "learning_rate": 2.6179218303145853e-07, "loss": 1.3718, "step": 47923 }, { "epoch": 9.13, "learning_rate": 2.617349857006673e-07, "loss": 2.5933, "step": 47924 }, { "epoch": 9.13, "learning_rate": 2.6167778836987607e-07, "loss": 1.5644, "step": 47925 }, { "epoch": 9.13, "learning_rate": 2.6162059103908486e-07, "loss": 2.2196, "step": 47926 }, { "epoch": 9.13, "learning_rate": 2.6156339370829366e-07, "loss": 1.673, "step": 47927 }, { "epoch": 9.13, "learning_rate": 2.615061963775024e-07, "loss": 2.8209, "step": 47928 }, { "epoch": 9.13, "learning_rate": 2.6144899904671114e-07, "loss": 1.5045, "step": 47929 }, { "epoch": 9.13, "learning_rate": 2.6139180171591993e-07, "loss": 0.8091, "step": 47930 }, { "epoch": 9.13, "learning_rate": 2.6133460438512867e-07, "loss": 3.569, "step": 47931 }, { "epoch": 9.13, "learning_rate": 2.6127740705433747e-07, "loss": 1.6768, "step": 47932 }, { "epoch": 9.13, "learning_rate": 2.6122020972354626e-07, "loss": 1.7, "step": 47933 }, { "epoch": 9.13, "learning_rate": 2.6116301239275505e-07, "loss": 2.2719, "step": 47934 }, { "epoch": 9.13, "learning_rate": 2.6110581506196374e-07, "loss": 0.581, "step": 47935 }, { "epoch": 9.13, "learning_rate": 2.6104861773117254e-07, "loss": 1.6949, "step": 47936 }, { "epoch": 9.13, "learning_rate": 2.6099142040038133e-07, "loss": 1.0774, "step": 47937 }, { "epoch": 9.13, "learning_rate": 2.609342230695901e-07, "loss": 1.0891, "step": 47938 }, { "epoch": 9.13, "learning_rate": 2.6087702573879886e-07, "loss": 1.3622, "step": 47939 }, { "epoch": 9.13, "learning_rate": 2.6081982840800766e-07, "loss": 1.543, "step": 47940 }, { "epoch": 9.13, "learning_rate": 2.607626310772164e-07, "loss": 1.0872, "step": 47941 }, { "epoch": 9.13, "learning_rate": 2.6070543374642514e-07, "loss": 2.0834, "step": 47942 }, { "epoch": 9.13, "learning_rate": 2.6064823641563393e-07, "loss": 1.8025, "step": 47943 }, { "epoch": 9.13, "learning_rate": 2.6059103908484273e-07, "loss": 2.3632, "step": 47944 }, { "epoch": 9.13, "learning_rate": 2.605338417540515e-07, "loss": 2.5733, "step": 47945 }, { "epoch": 9.13, "learning_rate": 2.6047664442326026e-07, "loss": 1.7455, "step": 47946 }, { "epoch": 9.13, "learning_rate": 2.60419447092469e-07, "loss": 2.7921, "step": 47947 }, { "epoch": 9.13, "learning_rate": 2.603622497616778e-07, "loss": 1.8108, "step": 47948 }, { "epoch": 9.13, "learning_rate": 2.603050524308866e-07, "loss": 2.2112, "step": 47949 }, { "epoch": 9.13, "learning_rate": 2.6024785510009533e-07, "loss": 3.9613, "step": 47950 }, { "epoch": 9.13, "learning_rate": 2.601906577693041e-07, "loss": 2.2838, "step": 47951 }, { "epoch": 9.13, "learning_rate": 2.601334604385129e-07, "loss": 0.8555, "step": 47952 }, { "epoch": 9.13, "learning_rate": 2.600762631077216e-07, "loss": 1.7695, "step": 47953 }, { "epoch": 9.13, "learning_rate": 2.600190657769304e-07, "loss": 2.4633, "step": 47954 }, { "epoch": 9.13, "learning_rate": 2.599618684461392e-07, "loss": 1.5832, "step": 47955 }, { "epoch": 9.13, "learning_rate": 2.59904671115348e-07, "loss": 2.0672, "step": 47956 }, { "epoch": 9.13, "learning_rate": 2.5984747378455673e-07, "loss": 2.374, "step": 47957 }, { "epoch": 9.13, "learning_rate": 2.5979027645376547e-07, "loss": 1.8449, "step": 47958 }, { "epoch": 9.14, "learning_rate": 2.5973307912297426e-07, "loss": 1.7863, "step": 47959 }, { "epoch": 9.14, "learning_rate": 2.59675881792183e-07, "loss": 1.3009, "step": 47960 }, { "epoch": 9.14, "learning_rate": 2.596186844613918e-07, "loss": 2.0533, "step": 47961 }, { "epoch": 9.14, "learning_rate": 2.595614871306006e-07, "loss": 2.1273, "step": 47962 }, { "epoch": 9.14, "learning_rate": 2.595042897998094e-07, "loss": 0.9351, "step": 47963 }, { "epoch": 9.14, "learning_rate": 2.594470924690181e-07, "loss": 1.4275, "step": 47964 }, { "epoch": 9.14, "learning_rate": 2.5938989513822687e-07, "loss": 2.1413, "step": 47965 }, { "epoch": 9.14, "learning_rate": 2.5933269780743566e-07, "loss": 2.3097, "step": 47966 }, { "epoch": 9.14, "learning_rate": 2.5927550047664446e-07, "loss": 0.7797, "step": 47967 }, { "epoch": 9.14, "learning_rate": 2.592183031458532e-07, "loss": 1.8767, "step": 47968 }, { "epoch": 9.14, "learning_rate": 2.59161105815062e-07, "loss": 1.15, "step": 47969 }, { "epoch": 9.14, "learning_rate": 2.5910390848427073e-07, "loss": 1.3651, "step": 47970 }, { "epoch": 9.14, "learning_rate": 2.5904671115347947e-07, "loss": 1.8062, "step": 47971 }, { "epoch": 9.14, "learning_rate": 2.5898951382268827e-07, "loss": 2.103, "step": 47972 }, { "epoch": 9.14, "learning_rate": 2.5893231649189706e-07, "loss": 2.4217, "step": 47973 }, { "epoch": 9.14, "learning_rate": 2.5887511916110585e-07, "loss": 0.5723, "step": 47974 }, { "epoch": 9.14, "learning_rate": 2.588179218303146e-07, "loss": 1.4106, "step": 47975 }, { "epoch": 9.14, "learning_rate": 2.5876072449952334e-07, "loss": 1.3799, "step": 47976 }, { "epoch": 9.14, "learning_rate": 2.5870352716873213e-07, "loss": 2.0246, "step": 47977 }, { "epoch": 9.14, "learning_rate": 2.586463298379409e-07, "loss": 3.1836, "step": 47978 }, { "epoch": 9.14, "learning_rate": 2.5858913250714966e-07, "loss": 1.9998, "step": 47979 }, { "epoch": 9.14, "learning_rate": 2.5853193517635846e-07, "loss": 1.7503, "step": 47980 }, { "epoch": 9.14, "learning_rate": 2.5847473784556725e-07, "loss": 1.3301, "step": 47981 }, { "epoch": 9.14, "learning_rate": 2.5841754051477594e-07, "loss": 1.8115, "step": 47982 }, { "epoch": 9.14, "learning_rate": 2.5836034318398473e-07, "loss": 1.9715, "step": 47983 }, { "epoch": 9.14, "learning_rate": 2.5830314585319353e-07, "loss": 1.3811, "step": 47984 }, { "epoch": 9.14, "learning_rate": 2.582459485224023e-07, "loss": 1.1378, "step": 47985 }, { "epoch": 9.14, "learning_rate": 2.5818875119161106e-07, "loss": 1.9143, "step": 47986 }, { "epoch": 9.14, "learning_rate": 2.5813155386081986e-07, "loss": 1.3447, "step": 47987 }, { "epoch": 9.14, "learning_rate": 2.580743565300286e-07, "loss": 1.3506, "step": 47988 }, { "epoch": 9.14, "learning_rate": 2.580171591992374e-07, "loss": 0.6292, "step": 47989 }, { "epoch": 9.14, "learning_rate": 2.5795996186844613e-07, "loss": 1.9944, "step": 47990 }, { "epoch": 9.14, "learning_rate": 2.579027645376549e-07, "loss": 1.501, "step": 47991 }, { "epoch": 9.14, "learning_rate": 2.578455672068637e-07, "loss": 2.3924, "step": 47992 }, { "epoch": 9.14, "learning_rate": 2.5778836987607246e-07, "loss": 2.5927, "step": 47993 }, { "epoch": 9.14, "learning_rate": 2.577311725452812e-07, "loss": 2.3155, "step": 47994 }, { "epoch": 9.14, "learning_rate": 2.5767397521449e-07, "loss": 1.4558, "step": 47995 }, { "epoch": 9.14, "learning_rate": 2.576167778836988e-07, "loss": 1.1085, "step": 47996 }, { "epoch": 9.14, "learning_rate": 2.5755958055290753e-07, "loss": 0.4325, "step": 47997 }, { "epoch": 9.14, "learning_rate": 2.575023832221163e-07, "loss": 3.1849, "step": 47998 }, { "epoch": 9.14, "learning_rate": 2.574451858913251e-07, "loss": 2.0457, "step": 47999 }, { "epoch": 9.14, "learning_rate": 2.5738798856053386e-07, "loss": 3.0997, "step": 48000 }, { "epoch": 9.14, "learning_rate": 2.573307912297426e-07, "loss": 1.5046, "step": 48001 }, { "epoch": 9.14, "learning_rate": 2.572735938989514e-07, "loss": 1.4409, "step": 48002 }, { "epoch": 9.14, "learning_rate": 2.572163965681602e-07, "loss": 2.3237, "step": 48003 }, { "epoch": 9.14, "learning_rate": 2.5715919923736893e-07, "loss": 0.988, "step": 48004 }, { "epoch": 9.14, "learning_rate": 2.571020019065777e-07, "loss": 1.4373, "step": 48005 }, { "epoch": 9.14, "learning_rate": 2.5704480457578646e-07, "loss": 2.1496, "step": 48006 }, { "epoch": 9.14, "learning_rate": 2.5698760724499526e-07, "loss": 2.0491, "step": 48007 }, { "epoch": 9.14, "learning_rate": 2.56930409914204e-07, "loss": 1.6786, "step": 48008 }, { "epoch": 9.14, "learning_rate": 2.568732125834128e-07, "loss": 1.0529, "step": 48009 }, { "epoch": 9.14, "learning_rate": 2.568160152526216e-07, "loss": 1.1272, "step": 48010 }, { "epoch": 9.14, "learning_rate": 2.567588179218303e-07, "loss": 1.6475, "step": 48011 }, { "epoch": 9.15, "learning_rate": 2.5670162059103907e-07, "loss": 1.5578, "step": 48012 }, { "epoch": 9.15, "learning_rate": 2.5664442326024786e-07, "loss": 0.6937, "step": 48013 }, { "epoch": 9.15, "learning_rate": 2.5658722592945665e-07, "loss": 1.8967, "step": 48014 }, { "epoch": 9.15, "learning_rate": 2.565300285986654e-07, "loss": 2.4134, "step": 48015 }, { "epoch": 9.15, "learning_rate": 2.564728312678742e-07, "loss": 2.7552, "step": 48016 }, { "epoch": 9.15, "learning_rate": 2.56415633937083e-07, "loss": 1.5404, "step": 48017 }, { "epoch": 9.15, "learning_rate": 2.563584366062917e-07, "loss": 1.6416, "step": 48018 }, { "epoch": 9.15, "learning_rate": 2.5630123927550046e-07, "loss": 2.4081, "step": 48019 }, { "epoch": 9.15, "learning_rate": 2.5624404194470926e-07, "loss": 0.3089, "step": 48020 }, { "epoch": 9.15, "learning_rate": 2.5618684461391805e-07, "loss": 1.7118, "step": 48021 }, { "epoch": 9.15, "learning_rate": 2.561296472831268e-07, "loss": 2.2315, "step": 48022 }, { "epoch": 9.15, "learning_rate": 2.5607244995233553e-07, "loss": 1.4628, "step": 48023 }, { "epoch": 9.15, "learning_rate": 2.5601525262154433e-07, "loss": 0.6038, "step": 48024 }, { "epoch": 9.15, "learning_rate": 2.559580552907531e-07, "loss": 1.4738, "step": 48025 }, { "epoch": 9.15, "learning_rate": 2.5590085795996186e-07, "loss": 2.073, "step": 48026 }, { "epoch": 9.15, "learning_rate": 2.5584366062917066e-07, "loss": 2.7179, "step": 48027 }, { "epoch": 9.15, "learning_rate": 2.5578646329837945e-07, "loss": 0.9229, "step": 48028 }, { "epoch": 9.15, "learning_rate": 2.557292659675882e-07, "loss": 1.7123, "step": 48029 }, { "epoch": 9.15, "learning_rate": 2.5567206863679693e-07, "loss": 2.0028, "step": 48030 }, { "epoch": 9.15, "learning_rate": 2.556148713060057e-07, "loss": 2.1524, "step": 48031 }, { "epoch": 9.15, "learning_rate": 2.555576739752145e-07, "loss": 1.7519, "step": 48032 }, { "epoch": 9.15, "learning_rate": 2.5550047664442326e-07, "loss": 2.2762, "step": 48033 }, { "epoch": 9.15, "learning_rate": 2.5544327931363205e-07, "loss": 1.0466, "step": 48034 }, { "epoch": 9.15, "learning_rate": 2.553860819828408e-07, "loss": 0.9348, "step": 48035 }, { "epoch": 9.15, "learning_rate": 2.553288846520496e-07, "loss": 1.4035, "step": 48036 }, { "epoch": 9.15, "learning_rate": 2.5527168732125833e-07, "loss": 1.0337, "step": 48037 }, { "epoch": 9.15, "learning_rate": 2.552144899904671e-07, "loss": 1.1427, "step": 48038 }, { "epoch": 9.15, "learning_rate": 2.551572926596759e-07, "loss": 2.6755, "step": 48039 }, { "epoch": 9.15, "learning_rate": 2.5510009532888466e-07, "loss": 1.7006, "step": 48040 }, { "epoch": 9.15, "learning_rate": 2.550428979980934e-07, "loss": 1.2371, "step": 48041 }, { "epoch": 9.15, "learning_rate": 2.549857006673022e-07, "loss": 2.5212, "step": 48042 }, { "epoch": 9.15, "learning_rate": 2.54928503336511e-07, "loss": 1.1587, "step": 48043 }, { "epoch": 9.15, "learning_rate": 2.5487130600571973e-07, "loss": 2.2624, "step": 48044 }, { "epoch": 9.15, "learning_rate": 2.548141086749285e-07, "loss": 0.66, "step": 48045 }, { "epoch": 9.15, "learning_rate": 2.547569113441373e-07, "loss": 1.1961, "step": 48046 }, { "epoch": 9.15, "learning_rate": 2.5469971401334606e-07, "loss": 1.4118, "step": 48047 }, { "epoch": 9.15, "learning_rate": 2.546425166825548e-07, "loss": 1.0603, "step": 48048 }, { "epoch": 9.15, "learning_rate": 2.545853193517636e-07, "loss": 1.3171, "step": 48049 }, { "epoch": 9.15, "learning_rate": 2.545281220209724e-07, "loss": 1.4575, "step": 48050 }, { "epoch": 9.15, "learning_rate": 2.544709246901811e-07, "loss": 1.5367, "step": 48051 }, { "epoch": 9.15, "learning_rate": 2.544137273593899e-07, "loss": 1.5719, "step": 48052 }, { "epoch": 9.15, "learning_rate": 2.5435653002859866e-07, "loss": 1.4306, "step": 48053 }, { "epoch": 9.15, "learning_rate": 2.5429933269780745e-07, "loss": 1.366, "step": 48054 }, { "epoch": 9.15, "learning_rate": 2.542421353670162e-07, "loss": 0.9793, "step": 48055 }, { "epoch": 9.15, "learning_rate": 2.54184938036225e-07, "loss": 1.3427, "step": 48056 }, { "epoch": 9.15, "learning_rate": 2.541277407054338e-07, "loss": 1.1713, "step": 48057 }, { "epoch": 9.15, "learning_rate": 2.540705433746425e-07, "loss": 1.6695, "step": 48058 }, { "epoch": 9.15, "learning_rate": 2.5401334604385126e-07, "loss": 4.6883, "step": 48059 }, { "epoch": 9.15, "learning_rate": 2.5395614871306006e-07, "loss": 2.1481, "step": 48060 }, { "epoch": 9.15, "learning_rate": 2.5389895138226885e-07, "loss": 1.6727, "step": 48061 }, { "epoch": 9.15, "learning_rate": 2.538417540514776e-07, "loss": 0.9969, "step": 48062 }, { "epoch": 9.15, "learning_rate": 2.537845567206864e-07, "loss": 0.9875, "step": 48063 }, { "epoch": 9.16, "learning_rate": 2.537273593898952e-07, "loss": 2.4322, "step": 48064 }, { "epoch": 9.16, "learning_rate": 2.536701620591039e-07, "loss": 1.2714, "step": 48065 }, { "epoch": 9.16, "learning_rate": 2.5361296472831266e-07, "loss": 1.3919, "step": 48066 }, { "epoch": 9.16, "learning_rate": 2.5355576739752146e-07, "loss": 1.2636, "step": 48067 }, { "epoch": 9.16, "learning_rate": 2.5349857006673025e-07, "loss": 2.5409, "step": 48068 }, { "epoch": 9.16, "learning_rate": 2.53441372735939e-07, "loss": 1.1141, "step": 48069 }, { "epoch": 9.16, "learning_rate": 2.533841754051478e-07, "loss": 1.4178, "step": 48070 }, { "epoch": 9.16, "learning_rate": 2.533269780743565e-07, "loss": 1.8308, "step": 48071 }, { "epoch": 9.16, "learning_rate": 2.532697807435653e-07, "loss": 1.4176, "step": 48072 }, { "epoch": 9.16, "learning_rate": 2.5321258341277406e-07, "loss": 2.5541, "step": 48073 }, { "epoch": 9.16, "learning_rate": 2.5315538608198285e-07, "loss": 1.2223, "step": 48074 }, { "epoch": 9.16, "learning_rate": 2.5309818875119165e-07, "loss": 0.772, "step": 48075 }, { "epoch": 9.16, "learning_rate": 2.530409914204004e-07, "loss": 1.4399, "step": 48076 }, { "epoch": 9.16, "learning_rate": 2.5298379408960913e-07, "loss": 1.8584, "step": 48077 }, { "epoch": 9.16, "learning_rate": 2.529265967588179e-07, "loss": 1.3686, "step": 48078 }, { "epoch": 9.16, "learning_rate": 2.528693994280267e-07, "loss": 2.1352, "step": 48079 }, { "epoch": 9.16, "learning_rate": 2.5281220209723546e-07, "loss": 2.6055, "step": 48080 }, { "epoch": 9.16, "learning_rate": 2.5275500476644425e-07, "loss": 1.1126, "step": 48081 }, { "epoch": 9.16, "learning_rate": 2.52697807435653e-07, "loss": 1.5068, "step": 48082 }, { "epoch": 9.16, "learning_rate": 2.526406101048618e-07, "loss": 1.1189, "step": 48083 }, { "epoch": 9.16, "learning_rate": 2.5258341277407053e-07, "loss": 0.7377, "step": 48084 }, { "epoch": 9.16, "learning_rate": 2.525262154432793e-07, "loss": 1.8916, "step": 48085 }, { "epoch": 9.16, "learning_rate": 2.524690181124881e-07, "loss": 1.557, "step": 48086 }, { "epoch": 9.16, "learning_rate": 2.5241182078169686e-07, "loss": 1.9901, "step": 48087 }, { "epoch": 9.16, "learning_rate": 2.523546234509056e-07, "loss": 1.4356, "step": 48088 }, { "epoch": 9.16, "learning_rate": 2.522974261201144e-07, "loss": 1.0622, "step": 48089 }, { "epoch": 9.16, "learning_rate": 2.522402287893232e-07, "loss": 2.735, "step": 48090 }, { "epoch": 9.16, "learning_rate": 2.521830314585319e-07, "loss": 2.2064, "step": 48091 }, { "epoch": 9.16, "learning_rate": 2.521258341277407e-07, "loss": 1.6016, "step": 48092 }, { "epoch": 9.16, "learning_rate": 2.520686367969495e-07, "loss": 0.9334, "step": 48093 }, { "epoch": 9.16, "learning_rate": 2.5201143946615825e-07, "loss": 1.9297, "step": 48094 }, { "epoch": 9.16, "learning_rate": 2.51954242135367e-07, "loss": 1.1097, "step": 48095 }, { "epoch": 9.16, "learning_rate": 2.518970448045758e-07, "loss": 1.216, "step": 48096 }, { "epoch": 9.16, "learning_rate": 2.518398474737846e-07, "loss": 1.6184, "step": 48097 }, { "epoch": 9.16, "learning_rate": 2.517826501429933e-07, "loss": 1.1591, "step": 48098 }, { "epoch": 9.16, "learning_rate": 2.517254528122021e-07, "loss": 2.1578, "step": 48099 }, { "epoch": 9.16, "learning_rate": 2.5166825548141086e-07, "loss": 2.0649, "step": 48100 }, { "epoch": 9.16, "learning_rate": 2.5161105815061965e-07, "loss": 0.236, "step": 48101 }, { "epoch": 9.16, "learning_rate": 2.515538608198284e-07, "loss": 1.7429, "step": 48102 }, { "epoch": 9.16, "learning_rate": 2.514966634890372e-07, "loss": 0.7434, "step": 48103 }, { "epoch": 9.16, "learning_rate": 2.51439466158246e-07, "loss": 0.956, "step": 48104 }, { "epoch": 9.16, "learning_rate": 2.513822688274547e-07, "loss": 1.8318, "step": 48105 }, { "epoch": 9.16, "learning_rate": 2.5132507149666346e-07, "loss": 0.5634, "step": 48106 }, { "epoch": 9.16, "learning_rate": 2.5126787416587226e-07, "loss": 1.0901, "step": 48107 }, { "epoch": 9.16, "learning_rate": 2.5121067683508105e-07, "loss": 1.2444, "step": 48108 }, { "epoch": 9.16, "learning_rate": 2.511534795042898e-07, "loss": 2.5374, "step": 48109 }, { "epoch": 9.16, "learning_rate": 2.510962821734986e-07, "loss": 2.3071, "step": 48110 }, { "epoch": 9.16, "learning_rate": 2.510390848427074e-07, "loss": 1.8461, "step": 48111 }, { "epoch": 9.16, "learning_rate": 2.509818875119161e-07, "loss": 2.6447, "step": 48112 }, { "epoch": 9.16, "learning_rate": 2.5092469018112486e-07, "loss": 1.1298, "step": 48113 }, { "epoch": 9.16, "learning_rate": 2.5086749285033365e-07, "loss": 2.5052, "step": 48114 }, { "epoch": 9.16, "learning_rate": 2.5081029551954245e-07, "loss": 1.2368, "step": 48115 }, { "epoch": 9.16, "learning_rate": 2.507530981887512e-07, "loss": 2.353, "step": 48116 }, { "epoch": 9.17, "learning_rate": 2.5069590085796e-07, "loss": 1.9901, "step": 48117 }, { "epoch": 9.17, "learning_rate": 2.506387035271687e-07, "loss": 2.4361, "step": 48118 }, { "epoch": 9.17, "learning_rate": 2.505815061963775e-07, "loss": 0.4975, "step": 48119 }, { "epoch": 9.17, "learning_rate": 2.5052430886558626e-07, "loss": 1.6539, "step": 48120 }, { "epoch": 9.17, "learning_rate": 2.5046711153479505e-07, "loss": 1.8714, "step": 48121 }, { "epoch": 9.17, "learning_rate": 2.5040991420400384e-07, "loss": 2.2343, "step": 48122 }, { "epoch": 9.17, "learning_rate": 2.503527168732126e-07, "loss": 0.7059, "step": 48123 }, { "epoch": 9.17, "learning_rate": 2.5029551954242133e-07, "loss": 2.008, "step": 48124 }, { "epoch": 9.17, "learning_rate": 2.502383222116301e-07, "loss": 1.5338, "step": 48125 }, { "epoch": 9.17, "learning_rate": 2.501811248808389e-07, "loss": 1.7992, "step": 48126 }, { "epoch": 9.17, "learning_rate": 2.5012392755004766e-07, "loss": 1.0223, "step": 48127 }, { "epoch": 9.17, "learning_rate": 2.5006673021925645e-07, "loss": 2.0582, "step": 48128 }, { "epoch": 9.17, "learning_rate": 2.5000953288846524e-07, "loss": 0.6912, "step": 48129 }, { "epoch": 9.17, "learning_rate": 2.49952335557674e-07, "loss": 2.9068, "step": 48130 }, { "epoch": 9.17, "learning_rate": 2.498951382268827e-07, "loss": 2.384, "step": 48131 }, { "epoch": 9.17, "learning_rate": 2.498379408960915e-07, "loss": 1.7617, "step": 48132 }, { "epoch": 9.17, "learning_rate": 2.497807435653003e-07, "loss": 0.6439, "step": 48133 }, { "epoch": 9.17, "learning_rate": 2.4972354623450905e-07, "loss": 1.301, "step": 48134 }, { "epoch": 9.17, "learning_rate": 2.4966634890371785e-07, "loss": 1.7973, "step": 48135 }, { "epoch": 9.17, "learning_rate": 2.496091515729266e-07, "loss": 0.6184, "step": 48136 }, { "epoch": 9.17, "learning_rate": 2.495519542421354e-07, "loss": 2.3422, "step": 48137 }, { "epoch": 9.17, "learning_rate": 2.494947569113441e-07, "loss": 1.8212, "step": 48138 }, { "epoch": 9.17, "learning_rate": 2.494375595805529e-07, "loss": 2.4878, "step": 48139 }, { "epoch": 9.17, "learning_rate": 2.493803622497617e-07, "loss": 1.6692, "step": 48140 }, { "epoch": 9.17, "learning_rate": 2.4932316491897045e-07, "loss": 2.3823, "step": 48141 }, { "epoch": 9.17, "learning_rate": 2.492659675881792e-07, "loss": 1.2186, "step": 48142 }, { "epoch": 9.17, "learning_rate": 2.49208770257388e-07, "loss": 0.9334, "step": 48143 }, { "epoch": 9.17, "learning_rate": 2.491515729265968e-07, "loss": 0.1708, "step": 48144 }, { "epoch": 9.17, "learning_rate": 2.490943755958055e-07, "loss": 1.6763, "step": 48145 }, { "epoch": 9.17, "learning_rate": 2.490371782650143e-07, "loss": 3.7263, "step": 48146 }, { "epoch": 9.17, "learning_rate": 2.4897998093422305e-07, "loss": 2.6072, "step": 48147 }, { "epoch": 9.17, "learning_rate": 2.4892278360343185e-07, "loss": 1.5882, "step": 48148 }, { "epoch": 9.17, "learning_rate": 2.488655862726406e-07, "loss": 1.7259, "step": 48149 }, { "epoch": 9.17, "learning_rate": 2.488083889418494e-07, "loss": 1.377, "step": 48150 }, { "epoch": 9.17, "learning_rate": 2.487511916110582e-07, "loss": 1.4538, "step": 48151 }, { "epoch": 9.17, "learning_rate": 2.486939942802669e-07, "loss": 1.3161, "step": 48152 }, { "epoch": 9.17, "learning_rate": 2.4863679694947566e-07, "loss": 0.8726, "step": 48153 }, { "epoch": 9.17, "learning_rate": 2.4857959961868445e-07, "loss": 1.3523, "step": 48154 }, { "epoch": 9.17, "learning_rate": 2.4852240228789325e-07, "loss": 2.101, "step": 48155 }, { "epoch": 9.17, "learning_rate": 2.48465204957102e-07, "loss": 1.1803, "step": 48156 }, { "epoch": 9.17, "learning_rate": 2.484080076263108e-07, "loss": 1.3588, "step": 48157 }, { "epoch": 9.17, "learning_rate": 2.483508102955196e-07, "loss": 1.1628, "step": 48158 }, { "epoch": 9.17, "learning_rate": 2.482936129647283e-07, "loss": 3.0213, "step": 48159 }, { "epoch": 9.17, "learning_rate": 2.4823641563393706e-07, "loss": 1.395, "step": 48160 }, { "epoch": 9.17, "learning_rate": 2.4817921830314585e-07, "loss": 0.903, "step": 48161 }, { "epoch": 9.17, "learning_rate": 2.4812202097235464e-07, "loss": 0.4327, "step": 48162 }, { "epoch": 9.17, "learning_rate": 2.480648236415634e-07, "loss": 1.542, "step": 48163 }, { "epoch": 9.17, "learning_rate": 2.480076263107722e-07, "loss": 0.892, "step": 48164 }, { "epoch": 9.17, "learning_rate": 2.479504289799809e-07, "loss": 1.9449, "step": 48165 }, { "epoch": 9.17, "learning_rate": 2.478932316491897e-07, "loss": 2.3962, "step": 48166 }, { "epoch": 9.17, "learning_rate": 2.4783603431839845e-07, "loss": 2.72, "step": 48167 }, { "epoch": 9.17, "learning_rate": 2.4777883698760725e-07, "loss": 1.2946, "step": 48168 }, { "epoch": 9.18, "learning_rate": 2.4772163965681604e-07, "loss": 2.9212, "step": 48169 }, { "epoch": 9.18, "learning_rate": 2.476644423260248e-07, "loss": 1.3183, "step": 48170 }, { "epoch": 9.18, "learning_rate": 2.476072449952335e-07, "loss": 0.7461, "step": 48171 }, { "epoch": 9.18, "learning_rate": 2.475500476644423e-07, "loss": 1.5414, "step": 48172 }, { "epoch": 9.18, "learning_rate": 2.474928503336511e-07, "loss": 1.1061, "step": 48173 }, { "epoch": 9.18, "learning_rate": 2.4743565300285985e-07, "loss": 1.3586, "step": 48174 }, { "epoch": 9.18, "learning_rate": 2.4737845567206865e-07, "loss": 0.8764, "step": 48175 }, { "epoch": 9.18, "learning_rate": 2.4732125834127744e-07, "loss": 2.2863, "step": 48176 }, { "epoch": 9.18, "learning_rate": 2.472640610104862e-07, "loss": 1.1583, "step": 48177 }, { "epoch": 9.18, "learning_rate": 2.472068636796949e-07, "loss": 1.006, "step": 48178 }, { "epoch": 9.18, "learning_rate": 2.471496663489037e-07, "loss": 2.1042, "step": 48179 }, { "epoch": 9.18, "learning_rate": 2.470924690181125e-07, "loss": 2.0713, "step": 48180 }, { "epoch": 9.18, "learning_rate": 2.4703527168732125e-07, "loss": 1.6807, "step": 48181 }, { "epoch": 9.18, "learning_rate": 2.4697807435653004e-07, "loss": 1.6, "step": 48182 }, { "epoch": 9.18, "learning_rate": 2.469208770257388e-07, "loss": 0.7235, "step": 48183 }, { "epoch": 9.18, "learning_rate": 2.468636796949476e-07, "loss": 2.2481, "step": 48184 }, { "epoch": 9.18, "learning_rate": 2.468064823641563e-07, "loss": 2.0079, "step": 48185 }, { "epoch": 9.18, "learning_rate": 2.467492850333651e-07, "loss": 2.9438, "step": 48186 }, { "epoch": 9.18, "learning_rate": 2.466920877025739e-07, "loss": 2.9073, "step": 48187 }, { "epoch": 9.18, "learning_rate": 2.466348903717827e-07, "loss": 2.534, "step": 48188 }, { "epoch": 9.18, "learning_rate": 2.465776930409914e-07, "loss": 1.7792, "step": 48189 }, { "epoch": 9.18, "learning_rate": 2.465204957102002e-07, "loss": 1.3029, "step": 48190 }, { "epoch": 9.18, "learning_rate": 2.46463298379409e-07, "loss": 1.9942, "step": 48191 }, { "epoch": 9.18, "learning_rate": 2.464061010486177e-07, "loss": 0.6207, "step": 48192 }, { "epoch": 9.18, "learning_rate": 2.463489037178265e-07, "loss": 0.9585, "step": 48193 }, { "epoch": 9.18, "learning_rate": 2.462917063870353e-07, "loss": 0.7469, "step": 48194 }, { "epoch": 9.18, "learning_rate": 2.4623450905624405e-07, "loss": 2.4648, "step": 48195 }, { "epoch": 9.18, "learning_rate": 2.461773117254528e-07, "loss": 3.2065, "step": 48196 }, { "epoch": 9.18, "learning_rate": 2.461201143946616e-07, "loss": 0.991, "step": 48197 }, { "epoch": 9.18, "learning_rate": 2.460629170638704e-07, "loss": 2.5706, "step": 48198 }, { "epoch": 9.18, "learning_rate": 2.4600571973307917e-07, "loss": 1.5828, "step": 48199 }, { "epoch": 9.18, "learning_rate": 2.4594852240228786e-07, "loss": 1.1145, "step": 48200 }, { "epoch": 9.18, "learning_rate": 2.4589132507149665e-07, "loss": 1.2952, "step": 48201 }, { "epoch": 9.18, "learning_rate": 2.4583412774070544e-07, "loss": 2.5325, "step": 48202 }, { "epoch": 9.18, "learning_rate": 2.457769304099142e-07, "loss": 2.0954, "step": 48203 }, { "epoch": 9.18, "learning_rate": 2.45719733079123e-07, "loss": 1.3703, "step": 48204 }, { "epoch": 9.18, "learning_rate": 2.4566253574833177e-07, "loss": 1.6752, "step": 48205 }, { "epoch": 9.18, "learning_rate": 2.456053384175405e-07, "loss": 3.9523, "step": 48206 }, { "epoch": 9.18, "learning_rate": 2.4554814108674925e-07, "loss": 1.166, "step": 48207 }, { "epoch": 9.18, "learning_rate": 2.4549094375595805e-07, "loss": 1.42, "step": 48208 }, { "epoch": 9.18, "learning_rate": 2.4543374642516684e-07, "loss": 0.4346, "step": 48209 }, { "epoch": 9.18, "learning_rate": 2.4537654909437564e-07, "loss": 2.2277, "step": 48210 }, { "epoch": 9.18, "learning_rate": 2.453193517635844e-07, "loss": 0.4033, "step": 48211 }, { "epoch": 9.18, "learning_rate": 2.452621544327931e-07, "loss": 0.7573, "step": 48212 }, { "epoch": 9.18, "learning_rate": 2.452049571020019e-07, "loss": 1.1088, "step": 48213 }, { "epoch": 9.18, "learning_rate": 2.4514775977121065e-07, "loss": 1.1716, "step": 48214 }, { "epoch": 9.18, "learning_rate": 2.4509056244041945e-07, "loss": 2.0554, "step": 48215 }, { "epoch": 9.18, "learning_rate": 2.4503336510962824e-07, "loss": 2.2108, "step": 48216 }, { "epoch": 9.18, "learning_rate": 2.4497616777883703e-07, "loss": 2.9475, "step": 48217 }, { "epoch": 9.18, "learning_rate": 2.449189704480457e-07, "loss": 1.3842, "step": 48218 }, { "epoch": 9.18, "learning_rate": 2.448617731172545e-07, "loss": 1.0447, "step": 48219 }, { "epoch": 9.18, "learning_rate": 2.448045757864633e-07, "loss": 1.693, "step": 48220 }, { "epoch": 9.18, "learning_rate": 2.447473784556721e-07, "loss": 2.8915, "step": 48221 }, { "epoch": 9.19, "learning_rate": 2.4469018112488084e-07, "loss": 0.3932, "step": 48222 }, { "epoch": 9.19, "learning_rate": 2.4463298379408964e-07, "loss": 0.4126, "step": 48223 }, { "epoch": 9.19, "learning_rate": 2.445757864632984e-07, "loss": 2.069, "step": 48224 }, { "epoch": 9.19, "learning_rate": 2.445185891325071e-07, "loss": 1.3679, "step": 48225 }, { "epoch": 9.19, "learning_rate": 2.444613918017159e-07, "loss": 1.4624, "step": 48226 }, { "epoch": 9.19, "learning_rate": 2.444041944709247e-07, "loss": 0.7747, "step": 48227 }, { "epoch": 9.19, "learning_rate": 2.443469971401335e-07, "loss": 2.6844, "step": 48228 }, { "epoch": 9.19, "learning_rate": 2.4428979980934224e-07, "loss": 1.7432, "step": 48229 }, { "epoch": 9.19, "learning_rate": 2.44232602478551e-07, "loss": 0.8254, "step": 48230 }, { "epoch": 9.19, "learning_rate": 2.441754051477598e-07, "loss": 1.9934, "step": 48231 }, { "epoch": 9.19, "learning_rate": 2.4411820781696857e-07, "loss": 0.3574, "step": 48232 }, { "epoch": 9.19, "learning_rate": 2.440610104861773e-07, "loss": 2.3125, "step": 48233 }, { "epoch": 9.19, "learning_rate": 2.440038131553861e-07, "loss": 0.6387, "step": 48234 }, { "epoch": 9.19, "learning_rate": 2.439466158245949e-07, "loss": 0.7999, "step": 48235 }, { "epoch": 9.19, "learning_rate": 2.438894184938036e-07, "loss": 0.7571, "step": 48236 }, { "epoch": 9.19, "learning_rate": 2.438322211630124e-07, "loss": 0.8201, "step": 48237 }, { "epoch": 9.19, "learning_rate": 2.437750238322212e-07, "loss": 3.394, "step": 48238 }, { "epoch": 9.19, "learning_rate": 2.4371782650142997e-07, "loss": 0.7233, "step": 48239 }, { "epoch": 9.19, "learning_rate": 2.436606291706387e-07, "loss": 1.4825, "step": 48240 }, { "epoch": 9.19, "learning_rate": 2.436034318398475e-07, "loss": 1.379, "step": 48241 }, { "epoch": 9.19, "learning_rate": 2.4354623450905624e-07, "loss": 1.8805, "step": 48242 }, { "epoch": 9.19, "learning_rate": 2.4348903717826504e-07, "loss": 2.3832, "step": 48243 }, { "epoch": 9.19, "learning_rate": 2.434318398474738e-07, "loss": 2.8278, "step": 48244 }, { "epoch": 9.19, "learning_rate": 2.4337464251668257e-07, "loss": 0.5503, "step": 48245 }, { "epoch": 9.19, "learning_rate": 2.4331744518589137e-07, "loss": 1.9914, "step": 48246 }, { "epoch": 9.19, "learning_rate": 2.432602478551001e-07, "loss": 1.2114, "step": 48247 }, { "epoch": 9.19, "learning_rate": 2.4320305052430885e-07, "loss": 0.5971, "step": 48248 }, { "epoch": 9.19, "learning_rate": 2.4314585319351764e-07, "loss": 2.4033, "step": 48249 }, { "epoch": 9.19, "learning_rate": 2.4308865586272644e-07, "loss": 1.1013, "step": 48250 }, { "epoch": 9.19, "learning_rate": 2.430314585319352e-07, "loss": 0.7717, "step": 48251 }, { "epoch": 9.19, "learning_rate": 2.4297426120114397e-07, "loss": 1.2724, "step": 48252 }, { "epoch": 9.19, "learning_rate": 2.4291706387035276e-07, "loss": 0.8485, "step": 48253 }, { "epoch": 9.19, "learning_rate": 2.4285986653956145e-07, "loss": 2.0858, "step": 48254 }, { "epoch": 9.19, "learning_rate": 2.4280266920877025e-07, "loss": 2.5622, "step": 48255 }, { "epoch": 9.19, "learning_rate": 2.4274547187797904e-07, "loss": 2.1843, "step": 48256 }, { "epoch": 9.19, "learning_rate": 2.4268827454718783e-07, "loss": 3.087, "step": 48257 }, { "epoch": 9.19, "learning_rate": 2.426310772163966e-07, "loss": 0.8105, "step": 48258 }, { "epoch": 9.19, "learning_rate": 2.4257387988560537e-07, "loss": 0.6405, "step": 48259 }, { "epoch": 9.19, "learning_rate": 2.425166825548141e-07, "loss": 1.3032, "step": 48260 }, { "epoch": 9.19, "learning_rate": 2.424594852240229e-07, "loss": 1.5416, "step": 48261 }, { "epoch": 9.19, "learning_rate": 2.4240228789323164e-07, "loss": 1.5017, "step": 48262 }, { "epoch": 9.19, "learning_rate": 2.4234509056244044e-07, "loss": 2.2599, "step": 48263 }, { "epoch": 9.19, "learning_rate": 2.4228789323164923e-07, "loss": 1.4694, "step": 48264 }, { "epoch": 9.19, "learning_rate": 2.422306959008579e-07, "loss": 1.982, "step": 48265 }, { "epoch": 9.19, "learning_rate": 2.421734985700667e-07, "loss": 0.572, "step": 48266 }, { "epoch": 9.19, "learning_rate": 2.421163012392755e-07, "loss": 0.5855, "step": 48267 }, { "epoch": 9.19, "learning_rate": 2.420591039084843e-07, "loss": 1.3978, "step": 48268 }, { "epoch": 9.19, "learning_rate": 2.4200190657769304e-07, "loss": 1.5167, "step": 48269 }, { "epoch": 9.19, "learning_rate": 2.4194470924690184e-07, "loss": 1.215, "step": 48270 }, { "epoch": 9.19, "learning_rate": 2.418875119161106e-07, "loss": 0.5697, "step": 48271 }, { "epoch": 9.19, "learning_rate": 2.4183031458531937e-07, "loss": 0.896, "step": 48272 }, { "epoch": 9.19, "learning_rate": 2.417731172545281e-07, "loss": 0.8707, "step": 48273 }, { "epoch": 9.2, "learning_rate": 2.417159199237369e-07, "loss": 1.267, "step": 48274 }, { "epoch": 9.2, "learning_rate": 2.416587225929457e-07, "loss": 1.1367, "step": 48275 }, { "epoch": 9.2, "learning_rate": 2.4160152526215444e-07, "loss": 2.8762, "step": 48276 }, { "epoch": 9.2, "learning_rate": 2.415443279313632e-07, "loss": 1.0351, "step": 48277 }, { "epoch": 9.2, "learning_rate": 2.41487130600572e-07, "loss": 0.7273, "step": 48278 }, { "epoch": 9.2, "learning_rate": 2.4142993326978077e-07, "loss": 1.6795, "step": 48279 }, { "epoch": 9.2, "learning_rate": 2.413727359389895e-07, "loss": 0.9128, "step": 48280 }, { "epoch": 9.2, "learning_rate": 2.413155386081983e-07, "loss": 0.6364, "step": 48281 }, { "epoch": 9.2, "learning_rate": 2.412583412774071e-07, "loss": 1.5125, "step": 48282 }, { "epoch": 9.2, "learning_rate": 2.4120114394661584e-07, "loss": 2.235, "step": 48283 }, { "epoch": 9.2, "learning_rate": 2.411439466158246e-07, "loss": 3.0382, "step": 48284 }, { "epoch": 9.2, "learning_rate": 2.4108674928503337e-07, "loss": 1.3985, "step": 48285 }, { "epoch": 9.2, "learning_rate": 2.4102955195424217e-07, "loss": 1.4591, "step": 48286 }, { "epoch": 9.2, "learning_rate": 2.409723546234509e-07, "loss": 1.5772, "step": 48287 }, { "epoch": 9.2, "learning_rate": 2.409151572926597e-07, "loss": 2.0254, "step": 48288 }, { "epoch": 9.2, "learning_rate": 2.4085795996186844e-07, "loss": 1.93, "step": 48289 }, { "epoch": 9.2, "learning_rate": 2.4080076263107724e-07, "loss": 2.5702, "step": 48290 }, { "epoch": 9.2, "learning_rate": 2.40743565300286e-07, "loss": 1.1838, "step": 48291 }, { "epoch": 9.2, "learning_rate": 2.4068636796949477e-07, "loss": 1.5196, "step": 48292 }, { "epoch": 9.2, "learning_rate": 2.4062917063870356e-07, "loss": 1.5726, "step": 48293 }, { "epoch": 9.2, "learning_rate": 2.405719733079123e-07, "loss": 1.6586, "step": 48294 }, { "epoch": 9.2, "learning_rate": 2.4051477597712105e-07, "loss": 1.1194, "step": 48295 }, { "epoch": 9.2, "learning_rate": 2.4045757864632984e-07, "loss": 1.3678, "step": 48296 }, { "epoch": 9.2, "learning_rate": 2.4040038131553863e-07, "loss": 1.6708, "step": 48297 }, { "epoch": 9.2, "learning_rate": 2.403431839847474e-07, "loss": 2.4552, "step": 48298 }, { "epoch": 9.2, "learning_rate": 2.4028598665395617e-07, "loss": 2.0115, "step": 48299 }, { "epoch": 9.2, "learning_rate": 2.4022878932316496e-07, "loss": 0.734, "step": 48300 }, { "epoch": 9.2, "learning_rate": 2.401715919923737e-07, "loss": 1.1814, "step": 48301 }, { "epoch": 9.2, "learning_rate": 2.4011439466158244e-07, "loss": 2.1642, "step": 48302 }, { "epoch": 9.2, "learning_rate": 2.4005719733079124e-07, "loss": 2.7567, "step": 48303 }, { "epoch": 9.2, "learning_rate": 2.4000000000000003e-07, "loss": 0.5696, "step": 48304 }, { "epoch": 9.2, "learning_rate": 2.3994280266920877e-07, "loss": 0.9421, "step": 48305 }, { "epoch": 9.2, "learning_rate": 2.3988560533841757e-07, "loss": 1.3196, "step": 48306 }, { "epoch": 9.2, "learning_rate": 2.398284080076263e-07, "loss": 0.9919, "step": 48307 }, { "epoch": 9.2, "learning_rate": 2.397712106768351e-07, "loss": 1.4063, "step": 48308 }, { "epoch": 9.2, "learning_rate": 2.3971401334604384e-07, "loss": 0.357, "step": 48309 }, { "epoch": 9.2, "learning_rate": 2.3965681601525264e-07, "loss": 2.1661, "step": 48310 }, { "epoch": 9.2, "learning_rate": 2.3959961868446143e-07, "loss": 1.8648, "step": 48311 }, { "epoch": 9.2, "learning_rate": 2.3954242135367017e-07, "loss": 0.9541, "step": 48312 }, { "epoch": 9.2, "learning_rate": 2.394852240228789e-07, "loss": 2.356, "step": 48313 }, { "epoch": 9.2, "learning_rate": 2.394280266920877e-07, "loss": 1.0508, "step": 48314 }, { "epoch": 9.2, "learning_rate": 2.393708293612965e-07, "loss": 3.2281, "step": 48315 }, { "epoch": 9.2, "learning_rate": 2.3931363203050524e-07, "loss": 0.8116, "step": 48316 }, { "epoch": 9.2, "learning_rate": 2.3925643469971403e-07, "loss": 1.7572, "step": 48317 }, { "epoch": 9.2, "learning_rate": 2.3919923736892283e-07, "loss": 2.805, "step": 48318 }, { "epoch": 9.2, "learning_rate": 2.3914204003813157e-07, "loss": 1.5749, "step": 48319 }, { "epoch": 9.2, "learning_rate": 2.390848427073403e-07, "loss": 0.7841, "step": 48320 }, { "epoch": 9.2, "learning_rate": 2.390276453765491e-07, "loss": 1.7752, "step": 48321 }, { "epoch": 9.2, "learning_rate": 2.389704480457579e-07, "loss": 1.3027, "step": 48322 }, { "epoch": 9.2, "learning_rate": 2.3891325071496664e-07, "loss": 0.4096, "step": 48323 }, { "epoch": 9.2, "learning_rate": 2.388560533841754e-07, "loss": 0.9249, "step": 48324 }, { "epoch": 9.2, "learning_rate": 2.3879885605338417e-07, "loss": 1.3972, "step": 48325 }, { "epoch": 9.2, "learning_rate": 2.3874165872259297e-07, "loss": 0.8259, "step": 48326 }, { "epoch": 9.21, "learning_rate": 2.386844613918017e-07, "loss": 0.7644, "step": 48327 }, { "epoch": 9.21, "learning_rate": 2.386272640610105e-07, "loss": 1.2303, "step": 48328 }, { "epoch": 9.21, "learning_rate": 2.385700667302193e-07, "loss": 2.0447, "step": 48329 }, { "epoch": 9.21, "learning_rate": 2.3851286939942804e-07, "loss": 1.5537, "step": 48330 }, { "epoch": 9.21, "learning_rate": 2.384556720686368e-07, "loss": 0.9432, "step": 48331 }, { "epoch": 9.21, "learning_rate": 2.3839847473784557e-07, "loss": 1.6512, "step": 48332 }, { "epoch": 9.21, "learning_rate": 2.3834127740705434e-07, "loss": 1.5943, "step": 48333 }, { "epoch": 9.21, "learning_rate": 2.3828408007626313e-07, "loss": 1.7674, "step": 48334 }, { "epoch": 9.21, "learning_rate": 2.382268827454719e-07, "loss": 0.9621, "step": 48335 }, { "epoch": 9.21, "learning_rate": 2.3816968541468064e-07, "loss": 0.7404, "step": 48336 }, { "epoch": 9.21, "learning_rate": 2.381124880838894e-07, "loss": 2.967, "step": 48337 }, { "epoch": 9.21, "learning_rate": 2.380552907530982e-07, "loss": 1.3142, "step": 48338 }, { "epoch": 9.21, "learning_rate": 2.3799809342230697e-07, "loss": 1.1438, "step": 48339 }, { "epoch": 9.21, "learning_rate": 2.3794089609151574e-07, "loss": 2.1979, "step": 48340 }, { "epoch": 9.21, "learning_rate": 2.3788369876072453e-07, "loss": 0.9438, "step": 48341 }, { "epoch": 9.21, "learning_rate": 2.3782650142993327e-07, "loss": 1.1519, "step": 48342 }, { "epoch": 9.21, "learning_rate": 2.3776930409914204e-07, "loss": 2.695, "step": 48343 }, { "epoch": 9.21, "learning_rate": 2.377121067683508e-07, "loss": 1.8378, "step": 48344 }, { "epoch": 9.21, "learning_rate": 2.376549094375596e-07, "loss": 1.7097, "step": 48345 }, { "epoch": 9.21, "learning_rate": 2.3759771210676837e-07, "loss": 1.332, "step": 48346 }, { "epoch": 9.21, "learning_rate": 2.3754051477597713e-07, "loss": 1.0936, "step": 48347 }, { "epoch": 9.21, "learning_rate": 2.3748331744518587e-07, "loss": 1.7452, "step": 48348 }, { "epoch": 9.21, "learning_rate": 2.3742612011439467e-07, "loss": 1.9308, "step": 48349 }, { "epoch": 9.21, "learning_rate": 2.3736892278360344e-07, "loss": 0.9748, "step": 48350 }, { "epoch": 9.21, "learning_rate": 2.373117254528122e-07, "loss": 1.032, "step": 48351 }, { "epoch": 9.21, "learning_rate": 2.37254528122021e-07, "loss": 0.4572, "step": 48352 }, { "epoch": 9.21, "learning_rate": 2.3719733079122976e-07, "loss": 1.7738, "step": 48353 }, { "epoch": 9.21, "learning_rate": 2.371401334604385e-07, "loss": 1.2732, "step": 48354 }, { "epoch": 9.21, "learning_rate": 2.3708293612964727e-07, "loss": 1.5766, "step": 48355 }, { "epoch": 9.21, "learning_rate": 2.3702573879885607e-07, "loss": 1.8376, "step": 48356 }, { "epoch": 9.21, "learning_rate": 2.3696854146806483e-07, "loss": 1.4173, "step": 48357 }, { "epoch": 9.21, "learning_rate": 2.369113441372736e-07, "loss": 1.5722, "step": 48358 }, { "epoch": 9.21, "learning_rate": 2.368541468064824e-07, "loss": 1.6647, "step": 48359 }, { "epoch": 9.21, "learning_rate": 2.3679694947569114e-07, "loss": 1.2869, "step": 48360 }, { "epoch": 9.21, "learning_rate": 2.367397521448999e-07, "loss": 2.0771, "step": 48361 }, { "epoch": 9.21, "learning_rate": 2.3668255481410867e-07, "loss": 3.4797, "step": 48362 }, { "epoch": 9.21, "learning_rate": 2.3662535748331746e-07, "loss": 2.0067, "step": 48363 }, { "epoch": 9.21, "learning_rate": 2.3656816015252623e-07, "loss": 1.4688, "step": 48364 }, { "epoch": 9.21, "learning_rate": 2.36510962821735e-07, "loss": 1.7862, "step": 48365 }, { "epoch": 9.21, "learning_rate": 2.3645376549094374e-07, "loss": 1.233, "step": 48366 }, { "epoch": 9.21, "learning_rate": 2.3639656816015253e-07, "loss": 1.1676, "step": 48367 }, { "epoch": 9.21, "learning_rate": 2.363393708293613e-07, "loss": 1.2942, "step": 48368 }, { "epoch": 9.21, "learning_rate": 2.3628217349857007e-07, "loss": 1.7656, "step": 48369 }, { "epoch": 9.21, "learning_rate": 2.3622497616777886e-07, "loss": 0.6644, "step": 48370 }, { "epoch": 9.21, "learning_rate": 2.3616777883698763e-07, "loss": 2.3248, "step": 48371 }, { "epoch": 9.21, "learning_rate": 2.3611058150619637e-07, "loss": 2.3658, "step": 48372 }, { "epoch": 9.21, "learning_rate": 2.3605338417540514e-07, "loss": 0.766, "step": 48373 }, { "epoch": 9.21, "learning_rate": 2.3599618684461393e-07, "loss": 1.2928, "step": 48374 }, { "epoch": 9.21, "learning_rate": 2.359389895138227e-07, "loss": 0.8894, "step": 48375 }, { "epoch": 9.21, "learning_rate": 2.3588179218303147e-07, "loss": 1.3597, "step": 48376 }, { "epoch": 9.21, "learning_rate": 2.3582459485224026e-07, "loss": 2.4259, "step": 48377 }, { "epoch": 9.21, "learning_rate": 2.35767397521449e-07, "loss": 1.491, "step": 48378 }, { "epoch": 9.22, "learning_rate": 2.3571020019065777e-07, "loss": 1.6144, "step": 48379 }, { "epoch": 9.22, "learning_rate": 2.3565300285986654e-07, "loss": 2.6895, "step": 48380 }, { "epoch": 9.22, "learning_rate": 2.3559580552907533e-07, "loss": 1.2374, "step": 48381 }, { "epoch": 9.22, "learning_rate": 2.355386081982841e-07, "loss": 1.7629, "step": 48382 }, { "epoch": 9.22, "learning_rate": 2.3548141086749284e-07, "loss": 0.6431, "step": 48383 }, { "epoch": 9.22, "learning_rate": 2.354242135367016e-07, "loss": 2.3829, "step": 48384 }, { "epoch": 9.22, "learning_rate": 2.353670162059104e-07, "loss": 1.7832, "step": 48385 }, { "epoch": 9.22, "learning_rate": 2.3530981887511917e-07, "loss": 0.9271, "step": 48386 }, { "epoch": 9.22, "learning_rate": 2.3525262154432793e-07, "loss": 0.6841, "step": 48387 }, { "epoch": 9.22, "learning_rate": 2.3519542421353673e-07, "loss": 1.6802, "step": 48388 }, { "epoch": 9.22, "learning_rate": 2.3513822688274547e-07, "loss": 1.7584, "step": 48389 }, { "epoch": 9.22, "learning_rate": 2.3508102955195424e-07, "loss": 1.6599, "step": 48390 }, { "epoch": 9.22, "learning_rate": 2.35023832221163e-07, "loss": 1.3509, "step": 48391 }, { "epoch": 9.22, "learning_rate": 2.349666348903718e-07, "loss": 1.6094, "step": 48392 }, { "epoch": 9.22, "learning_rate": 2.3490943755958056e-07, "loss": 0.5611, "step": 48393 }, { "epoch": 9.22, "learning_rate": 2.3485224022878936e-07, "loss": 2.4641, "step": 48394 }, { "epoch": 9.22, "learning_rate": 2.3479504289799807e-07, "loss": 2.4324, "step": 48395 }, { "epoch": 9.22, "learning_rate": 2.3473784556720687e-07, "loss": 1.2118, "step": 48396 }, { "epoch": 9.22, "learning_rate": 2.3468064823641563e-07, "loss": 1.0205, "step": 48397 }, { "epoch": 9.22, "learning_rate": 2.346234509056244e-07, "loss": 1.9933, "step": 48398 }, { "epoch": 9.22, "learning_rate": 2.345662535748332e-07, "loss": 1.982, "step": 48399 }, { "epoch": 9.22, "learning_rate": 2.3450905624404196e-07, "loss": 1.6029, "step": 48400 }, { "epoch": 9.22, "learning_rate": 2.344518589132507e-07, "loss": 3.1566, "step": 48401 }, { "epoch": 9.22, "learning_rate": 2.3439466158245947e-07, "loss": 1.5026, "step": 48402 }, { "epoch": 9.22, "learning_rate": 2.3433746425166826e-07, "loss": 2.8973, "step": 48403 }, { "epoch": 9.22, "learning_rate": 2.3428026692087703e-07, "loss": 1.8748, "step": 48404 }, { "epoch": 9.22, "learning_rate": 2.342230695900858e-07, "loss": 2.7269, "step": 48405 }, { "epoch": 9.22, "learning_rate": 2.341658722592946e-07, "loss": 2.5001, "step": 48406 }, { "epoch": 9.22, "learning_rate": 2.3410867492850333e-07, "loss": 0.9417, "step": 48407 }, { "epoch": 9.22, "learning_rate": 2.340514775977121e-07, "loss": 2.1389, "step": 48408 }, { "epoch": 9.22, "learning_rate": 2.3399428026692087e-07, "loss": 0.746, "step": 48409 }, { "epoch": 9.22, "learning_rate": 2.3393708293612966e-07, "loss": 2.4067, "step": 48410 }, { "epoch": 9.22, "learning_rate": 2.3387988560533843e-07, "loss": 1.1237, "step": 48411 }, { "epoch": 9.22, "learning_rate": 2.3382268827454722e-07, "loss": 1.4426, "step": 48412 }, { "epoch": 9.22, "learning_rate": 2.3376549094375594e-07, "loss": 0.8872, "step": 48413 }, { "epoch": 9.22, "learning_rate": 2.3370829361296473e-07, "loss": 2.5482, "step": 48414 }, { "epoch": 9.22, "learning_rate": 2.336510962821735e-07, "loss": 0.6359, "step": 48415 }, { "epoch": 9.22, "learning_rate": 2.3359389895138227e-07, "loss": 1.5022, "step": 48416 }, { "epoch": 9.22, "learning_rate": 2.3353670162059106e-07, "loss": 2.068, "step": 48417 }, { "epoch": 9.22, "learning_rate": 2.3347950428979983e-07, "loss": 1.6804, "step": 48418 }, { "epoch": 9.22, "learning_rate": 2.3342230695900857e-07, "loss": 1.6504, "step": 48419 }, { "epoch": 9.22, "learning_rate": 2.3336510962821734e-07, "loss": 0.8576, "step": 48420 }, { "epoch": 9.22, "learning_rate": 2.3330791229742613e-07, "loss": 0.8294, "step": 48421 }, { "epoch": 9.22, "learning_rate": 2.332507149666349e-07, "loss": 0.6521, "step": 48422 }, { "epoch": 9.22, "learning_rate": 2.331935176358437e-07, "loss": 0.9348, "step": 48423 }, { "epoch": 9.22, "learning_rate": 2.3313632030505246e-07, "loss": 1.9109, "step": 48424 }, { "epoch": 9.22, "learning_rate": 2.330791229742612e-07, "loss": 1.2309, "step": 48425 }, { "epoch": 9.22, "learning_rate": 2.3302192564346997e-07, "loss": 1.0101, "step": 48426 }, { "epoch": 9.22, "learning_rate": 2.3296472831267873e-07, "loss": 2.5436, "step": 48427 }, { "epoch": 9.22, "learning_rate": 2.3290753098188753e-07, "loss": 1.1201, "step": 48428 }, { "epoch": 9.22, "learning_rate": 2.328503336510963e-07, "loss": 1.4861, "step": 48429 }, { "epoch": 9.22, "learning_rate": 2.327931363203051e-07, "loss": 2.448, "step": 48430 }, { "epoch": 9.22, "learning_rate": 2.327359389895138e-07, "loss": 3.0207, "step": 48431 }, { "epoch": 9.23, "learning_rate": 2.326787416587226e-07, "loss": 1.0881, "step": 48432 }, { "epoch": 9.23, "learning_rate": 2.3262154432793136e-07, "loss": 2.1994, "step": 48433 }, { "epoch": 9.23, "learning_rate": 2.3256434699714016e-07, "loss": 2.309, "step": 48434 }, { "epoch": 9.23, "learning_rate": 2.3250714966634892e-07, "loss": 2.2686, "step": 48435 }, { "epoch": 9.23, "learning_rate": 2.324499523355577e-07, "loss": 0.9583, "step": 48436 }, { "epoch": 9.23, "learning_rate": 2.3239275500476643e-07, "loss": 1.5904, "step": 48437 }, { "epoch": 9.23, "learning_rate": 2.323355576739752e-07, "loss": 2.5077, "step": 48438 }, { "epoch": 9.23, "learning_rate": 2.32278360343184e-07, "loss": 1.7701, "step": 48439 }, { "epoch": 9.23, "learning_rate": 2.3222116301239276e-07, "loss": 1.7144, "step": 48440 }, { "epoch": 9.23, "learning_rate": 2.3216396568160156e-07, "loss": 1.9181, "step": 48441 }, { "epoch": 9.23, "learning_rate": 2.3210676835081032e-07, "loss": 1.9367, "step": 48442 }, { "epoch": 9.23, "learning_rate": 2.3204957102001906e-07, "loss": 1.0049, "step": 48443 }, { "epoch": 9.23, "learning_rate": 2.3199237368922783e-07, "loss": 2.0586, "step": 48444 }, { "epoch": 9.23, "learning_rate": 2.3193517635843662e-07, "loss": 1.7351, "step": 48445 }, { "epoch": 9.23, "learning_rate": 2.318779790276454e-07, "loss": 1.7845, "step": 48446 }, { "epoch": 9.23, "learning_rate": 2.3182078169685416e-07, "loss": 1.0056, "step": 48447 }, { "epoch": 9.23, "learning_rate": 2.317635843660629e-07, "loss": 0.9843, "step": 48448 }, { "epoch": 9.23, "learning_rate": 2.3170638703527167e-07, "loss": 0.693, "step": 48449 }, { "epoch": 9.23, "learning_rate": 2.3164918970448046e-07, "loss": 1.0513, "step": 48450 }, { "epoch": 9.23, "learning_rate": 2.3159199237368923e-07, "loss": 0.3146, "step": 48451 }, { "epoch": 9.23, "learning_rate": 2.3153479504289802e-07, "loss": 1.3106, "step": 48452 }, { "epoch": 9.23, "learning_rate": 2.314775977121068e-07, "loss": 0.6961, "step": 48453 }, { "epoch": 9.23, "learning_rate": 2.3142040038131553e-07, "loss": 1.0242, "step": 48454 }, { "epoch": 9.23, "learning_rate": 2.313632030505243e-07, "loss": 2.2286, "step": 48455 }, { "epoch": 9.23, "learning_rate": 2.313060057197331e-07, "loss": 0.7068, "step": 48456 }, { "epoch": 9.23, "learning_rate": 2.3124880838894186e-07, "loss": 1.2386, "step": 48457 }, { "epoch": 9.23, "learning_rate": 2.3119161105815063e-07, "loss": 0.5989, "step": 48458 }, { "epoch": 9.23, "learning_rate": 2.3113441372735942e-07, "loss": 1.1444, "step": 48459 }, { "epoch": 9.23, "learning_rate": 2.3107721639656813e-07, "loss": 1.4569, "step": 48460 }, { "epoch": 9.23, "learning_rate": 2.3102001906577693e-07, "loss": 1.3438, "step": 48461 }, { "epoch": 9.23, "learning_rate": 2.309628217349857e-07, "loss": 2.4385, "step": 48462 }, { "epoch": 9.23, "learning_rate": 2.309056244041945e-07, "loss": 2.1573, "step": 48463 }, { "epoch": 9.23, "learning_rate": 2.3084842707340326e-07, "loss": 0.9038, "step": 48464 }, { "epoch": 9.23, "learning_rate": 2.3079122974261202e-07, "loss": 1.7524, "step": 48465 }, { "epoch": 9.23, "learning_rate": 2.3073403241182077e-07, "loss": 1.6671, "step": 48466 }, { "epoch": 9.23, "learning_rate": 2.3067683508102956e-07, "loss": 2.1829, "step": 48467 }, { "epoch": 9.23, "learning_rate": 2.3061963775023833e-07, "loss": 1.8864, "step": 48468 }, { "epoch": 9.23, "learning_rate": 2.305624404194471e-07, "loss": 1.479, "step": 48469 }, { "epoch": 9.23, "learning_rate": 2.305052430886559e-07, "loss": 1.2819, "step": 48470 }, { "epoch": 9.23, "learning_rate": 2.3044804575786466e-07, "loss": 1.5336, "step": 48471 }, { "epoch": 9.23, "learning_rate": 2.303908484270734e-07, "loss": 0.317, "step": 48472 }, { "epoch": 9.23, "learning_rate": 2.3033365109628216e-07, "loss": 0.6767, "step": 48473 }, { "epoch": 9.23, "learning_rate": 2.3027645376549096e-07, "loss": 2.0683, "step": 48474 }, { "epoch": 9.23, "learning_rate": 2.3021925643469972e-07, "loss": 1.6198, "step": 48475 }, { "epoch": 9.23, "learning_rate": 2.301620591039085e-07, "loss": 2.1094, "step": 48476 }, { "epoch": 9.23, "learning_rate": 2.3010486177311729e-07, "loss": 1.8316, "step": 48477 }, { "epoch": 9.23, "learning_rate": 2.3004766444232603e-07, "loss": 2.1545, "step": 48478 }, { "epoch": 9.23, "learning_rate": 2.299904671115348e-07, "loss": 0.8273, "step": 48479 }, { "epoch": 9.23, "learning_rate": 2.2993326978074356e-07, "loss": 1.2088, "step": 48480 }, { "epoch": 9.23, "learning_rate": 2.2987607244995236e-07, "loss": 0.7849, "step": 48481 }, { "epoch": 9.23, "learning_rate": 2.2981887511916112e-07, "loss": 1.4804, "step": 48482 }, { "epoch": 9.23, "learning_rate": 2.297616777883699e-07, "loss": 1.6432, "step": 48483 }, { "epoch": 9.24, "learning_rate": 2.2970448045757863e-07, "loss": 1.2612, "step": 48484 }, { "epoch": 9.24, "learning_rate": 2.2964728312678742e-07, "loss": 1.9584, "step": 48485 }, { "epoch": 9.24, "learning_rate": 2.295900857959962e-07, "loss": 2.774, "step": 48486 }, { "epoch": 9.24, "learning_rate": 2.2953288846520496e-07, "loss": 0.7174, "step": 48487 }, { "epoch": 9.24, "learning_rate": 2.2947569113441375e-07, "loss": 2.4033, "step": 48488 }, { "epoch": 9.24, "learning_rate": 2.2941849380362252e-07, "loss": 1.7536, "step": 48489 }, { "epoch": 9.24, "learning_rate": 2.2936129647283126e-07, "loss": 0.9908, "step": 48490 }, { "epoch": 9.24, "learning_rate": 2.2930409914204003e-07, "loss": 1.6605, "step": 48491 }, { "epoch": 9.24, "learning_rate": 2.2924690181124882e-07, "loss": 2.1265, "step": 48492 }, { "epoch": 9.24, "learning_rate": 2.291897044804576e-07, "loss": 1.587, "step": 48493 }, { "epoch": 9.24, "learning_rate": 2.2913250714966636e-07, "loss": 1.6278, "step": 48494 }, { "epoch": 9.24, "learning_rate": 2.2907530981887515e-07, "loss": 1.4361, "step": 48495 }, { "epoch": 9.24, "learning_rate": 2.290181124880839e-07, "loss": 2.4024, "step": 48496 }, { "epoch": 9.24, "learning_rate": 2.2896091515729266e-07, "loss": 1.2619, "step": 48497 }, { "epoch": 9.24, "learning_rate": 2.2890371782650143e-07, "loss": 1.581, "step": 48498 }, { "epoch": 9.24, "learning_rate": 2.2884652049571022e-07, "loss": 2.8122, "step": 48499 }, { "epoch": 9.24, "learning_rate": 2.28789323164919e-07, "loss": 0.7432, "step": 48500 }, { "epoch": 9.24, "learning_rate": 2.2873212583412775e-07, "loss": 1.709, "step": 48501 }, { "epoch": 9.24, "learning_rate": 2.286749285033365e-07, "loss": 1.4984, "step": 48502 }, { "epoch": 9.24, "learning_rate": 2.286177311725453e-07, "loss": 0.3923, "step": 48503 }, { "epoch": 9.24, "learning_rate": 2.2856053384175406e-07, "loss": 1.309, "step": 48504 }, { "epoch": 9.24, "learning_rate": 2.2850333651096282e-07, "loss": 1.0239, "step": 48505 }, { "epoch": 9.24, "learning_rate": 2.2844613918017162e-07, "loss": 1.5066, "step": 48506 }, { "epoch": 9.24, "learning_rate": 2.2838894184938036e-07, "loss": 0.9456, "step": 48507 }, { "epoch": 9.24, "learning_rate": 2.2833174451858913e-07, "loss": 2.931, "step": 48508 }, { "epoch": 9.24, "learning_rate": 2.282745471877979e-07, "loss": 1.4968, "step": 48509 }, { "epoch": 9.24, "learning_rate": 2.282173498570067e-07, "loss": 1.6784, "step": 48510 }, { "epoch": 9.24, "learning_rate": 2.2816015252621545e-07, "loss": 0.7687, "step": 48511 }, { "epoch": 9.24, "learning_rate": 2.2810295519542422e-07, "loss": 1.33, "step": 48512 }, { "epoch": 9.24, "learning_rate": 2.2804575786463296e-07, "loss": 1.7646, "step": 48513 }, { "epoch": 9.24, "learning_rate": 2.2798856053384176e-07, "loss": 0.9012, "step": 48514 }, { "epoch": 9.24, "learning_rate": 2.2793136320305052e-07, "loss": 2.3214, "step": 48515 }, { "epoch": 9.24, "learning_rate": 2.278741658722593e-07, "loss": 2.823, "step": 48516 }, { "epoch": 9.24, "learning_rate": 2.2781696854146809e-07, "loss": 1.2925, "step": 48517 }, { "epoch": 9.24, "learning_rate": 2.2775977121067685e-07, "loss": 1.6134, "step": 48518 }, { "epoch": 9.24, "learning_rate": 2.277025738798856e-07, "loss": 2.2261, "step": 48519 }, { "epoch": 9.24, "learning_rate": 2.2764537654909436e-07, "loss": 1.4245, "step": 48520 }, { "epoch": 9.24, "learning_rate": 2.2758817921830315e-07, "loss": 2.4781, "step": 48521 }, { "epoch": 9.24, "learning_rate": 2.2753098188751192e-07, "loss": 0.7328, "step": 48522 }, { "epoch": 9.24, "learning_rate": 2.274737845567207e-07, "loss": 2.4058, "step": 48523 }, { "epoch": 9.24, "learning_rate": 2.2741658722592948e-07, "loss": 2.7005, "step": 48524 }, { "epoch": 9.24, "learning_rate": 2.2735938989513822e-07, "loss": 1.0123, "step": 48525 }, { "epoch": 9.24, "learning_rate": 2.27302192564347e-07, "loss": 0.8692, "step": 48526 }, { "epoch": 9.24, "learning_rate": 2.2724499523355576e-07, "loss": 2.1115, "step": 48527 }, { "epoch": 9.24, "learning_rate": 2.2718779790276455e-07, "loss": 1.7558, "step": 48528 }, { "epoch": 9.24, "learning_rate": 2.2713060057197332e-07, "loss": 1.6097, "step": 48529 }, { "epoch": 9.24, "learning_rate": 2.2707340324118211e-07, "loss": 1.9568, "step": 48530 }, { "epoch": 9.24, "learning_rate": 2.2701620591039083e-07, "loss": 1.6044, "step": 48531 }, { "epoch": 9.24, "learning_rate": 2.2695900857959962e-07, "loss": 0.3125, "step": 48532 }, { "epoch": 9.24, "learning_rate": 2.269018112488084e-07, "loss": 1.1034, "step": 48533 }, { "epoch": 9.24, "learning_rate": 2.2684461391801716e-07, "loss": 1.5532, "step": 48534 }, { "epoch": 9.24, "learning_rate": 2.2678741658722595e-07, "loss": 2.2071, "step": 48535 }, { "epoch": 9.24, "learning_rate": 2.2673021925643472e-07, "loss": 0.6874, "step": 48536 }, { "epoch": 9.25, "learning_rate": 2.2667302192564346e-07, "loss": 1.7166, "step": 48537 }, { "epoch": 9.25, "learning_rate": 2.2661582459485223e-07, "loss": 1.7636, "step": 48538 }, { "epoch": 9.25, "learning_rate": 2.2655862726406102e-07, "loss": 0.2248, "step": 48539 }, { "epoch": 9.25, "learning_rate": 2.265014299332698e-07, "loss": 0.8469, "step": 48540 }, { "epoch": 9.25, "learning_rate": 2.2644423260247858e-07, "loss": 0.8859, "step": 48541 }, { "epoch": 9.25, "learning_rate": 2.2638703527168735e-07, "loss": 1.6747, "step": 48542 }, { "epoch": 9.25, "learning_rate": 2.263298379408961e-07, "loss": 2.0253, "step": 48543 }, { "epoch": 9.25, "learning_rate": 2.2627264061010486e-07, "loss": 2.3966, "step": 48544 }, { "epoch": 9.25, "learning_rate": 2.2621544327931362e-07, "loss": 1.7543, "step": 48545 }, { "epoch": 9.25, "learning_rate": 2.2615824594852242e-07, "loss": 2.5047, "step": 48546 }, { "epoch": 9.25, "learning_rate": 2.2610104861773119e-07, "loss": 0.9686, "step": 48547 }, { "epoch": 9.25, "learning_rate": 2.2604385128693998e-07, "loss": 0.7911, "step": 48548 }, { "epoch": 9.25, "learning_rate": 2.259866539561487e-07, "loss": 1.3434, "step": 48549 }, { "epoch": 9.25, "learning_rate": 2.259294566253575e-07, "loss": 1.2035, "step": 48550 }, { "epoch": 9.25, "learning_rate": 2.2587225929456625e-07, "loss": 2.5363, "step": 48551 }, { "epoch": 9.25, "learning_rate": 2.2581506196377502e-07, "loss": 1.4978, "step": 48552 }, { "epoch": 9.25, "learning_rate": 2.2575786463298382e-07, "loss": 0.7363, "step": 48553 }, { "epoch": 9.25, "learning_rate": 2.2570066730219258e-07, "loss": 2.6847, "step": 48554 }, { "epoch": 9.25, "learning_rate": 2.2564346997140132e-07, "loss": 1.6727, "step": 48555 }, { "epoch": 9.25, "learning_rate": 2.255862726406101e-07, "loss": 0.6126, "step": 48556 }, { "epoch": 9.25, "learning_rate": 2.2552907530981889e-07, "loss": 1.5896, "step": 48557 }, { "epoch": 9.25, "learning_rate": 2.2547187797902765e-07, "loss": 2.9292, "step": 48558 }, { "epoch": 9.25, "learning_rate": 2.2541468064823645e-07, "loss": 0.92, "step": 48559 }, { "epoch": 9.25, "learning_rate": 2.2535748331744521e-07, "loss": 1.2126, "step": 48560 }, { "epoch": 9.25, "learning_rate": 2.2530028598665395e-07, "loss": 1.1693, "step": 48561 }, { "epoch": 9.25, "learning_rate": 2.2524308865586272e-07, "loss": 2.0596, "step": 48562 }, { "epoch": 9.25, "learning_rate": 2.251858913250715e-07, "loss": 0.9737, "step": 48563 }, { "epoch": 9.25, "learning_rate": 2.2512869399428028e-07, "loss": 2.7104, "step": 48564 }, { "epoch": 9.25, "learning_rate": 2.2507149666348905e-07, "loss": 1.7487, "step": 48565 }, { "epoch": 9.25, "learning_rate": 2.250142993326978e-07, "loss": 1.8825, "step": 48566 }, { "epoch": 9.25, "learning_rate": 2.2495710200190656e-07, "loss": 1.1012, "step": 48567 }, { "epoch": 9.25, "learning_rate": 2.2489990467111535e-07, "loss": 2.3276, "step": 48568 }, { "epoch": 9.25, "learning_rate": 2.2484270734032412e-07, "loss": 1.1595, "step": 48569 }, { "epoch": 9.25, "learning_rate": 2.2478551000953291e-07, "loss": 1.7125, "step": 48570 }, { "epoch": 9.25, "learning_rate": 2.2472831267874168e-07, "loss": 1.5552, "step": 48571 }, { "epoch": 9.25, "learning_rate": 2.2467111534795042e-07, "loss": 2.4474, "step": 48572 }, { "epoch": 9.25, "learning_rate": 2.246139180171592e-07, "loss": 1.4545, "step": 48573 }, { "epoch": 9.25, "learning_rate": 2.2455672068636796e-07, "loss": 3.2519, "step": 48574 }, { "epoch": 9.25, "learning_rate": 2.2449952335557675e-07, "loss": 2.1439, "step": 48575 }, { "epoch": 9.25, "learning_rate": 2.2444232602478552e-07, "loss": 0.765, "step": 48576 }, { "epoch": 9.25, "learning_rate": 2.243851286939943e-07, "loss": 0.4911, "step": 48577 }, { "epoch": 9.25, "learning_rate": 2.2432793136320303e-07, "loss": 1.1825, "step": 48578 }, { "epoch": 9.25, "learning_rate": 2.2427073403241182e-07, "loss": 0.8872, "step": 48579 }, { "epoch": 9.25, "learning_rate": 2.242135367016206e-07, "loss": 1.2803, "step": 48580 }, { "epoch": 9.25, "learning_rate": 2.2415633937082938e-07, "loss": 1.623, "step": 48581 }, { "epoch": 9.25, "learning_rate": 2.2409914204003815e-07, "loss": 1.9503, "step": 48582 }, { "epoch": 9.25, "learning_rate": 2.2404194470924692e-07, "loss": 2.4254, "step": 48583 }, { "epoch": 9.25, "learning_rate": 2.2398474737845566e-07, "loss": 0.8764, "step": 48584 }, { "epoch": 9.25, "learning_rate": 2.2392755004766442e-07, "loss": 2.3008, "step": 48585 }, { "epoch": 9.25, "learning_rate": 2.2387035271687322e-07, "loss": 0.5623, "step": 48586 }, { "epoch": 9.25, "learning_rate": 2.2381315538608199e-07, "loss": 0.7995, "step": 48587 }, { "epoch": 9.25, "learning_rate": 2.2375595805529078e-07, "loss": 0.8363, "step": 48588 }, { "epoch": 9.26, "learning_rate": 2.2369876072449955e-07, "loss": 2.7208, "step": 48589 }, { "epoch": 9.26, "learning_rate": 2.236415633937083e-07, "loss": 2.7029, "step": 48590 }, { "epoch": 9.26, "learning_rate": 2.2358436606291705e-07, "loss": 2.5295, "step": 48591 }, { "epoch": 9.26, "learning_rate": 2.2352716873212585e-07, "loss": 0.9957, "step": 48592 }, { "epoch": 9.26, "learning_rate": 2.2346997140133462e-07, "loss": 0.9915, "step": 48593 }, { "epoch": 9.26, "learning_rate": 2.2341277407054338e-07, "loss": 0.5336, "step": 48594 }, { "epoch": 9.26, "learning_rate": 2.2335557673975218e-07, "loss": 1.9565, "step": 48595 }, { "epoch": 9.26, "learning_rate": 2.232983794089609e-07, "loss": 1.0375, "step": 48596 }, { "epoch": 9.26, "learning_rate": 2.2324118207816969e-07, "loss": 1.0184, "step": 48597 }, { "epoch": 9.26, "learning_rate": 2.2318398474737845e-07, "loss": 1.2121, "step": 48598 }, { "epoch": 9.26, "learning_rate": 2.2312678741658725e-07, "loss": 1.3816, "step": 48599 }, { "epoch": 9.26, "learning_rate": 2.2306959008579601e-07, "loss": 2.362, "step": 48600 }, { "epoch": 9.26, "learning_rate": 2.2301239275500478e-07, "loss": 0.9669, "step": 48601 }, { "epoch": 9.26, "learning_rate": 2.2295519542421352e-07, "loss": 0.744, "step": 48602 }, { "epoch": 9.26, "learning_rate": 2.2289799809342232e-07, "loss": 1.3838, "step": 48603 }, { "epoch": 9.26, "learning_rate": 2.2284080076263108e-07, "loss": 0.8337, "step": 48604 }, { "epoch": 9.26, "learning_rate": 2.2278360343183985e-07, "loss": 0.3541, "step": 48605 }, { "epoch": 9.26, "learning_rate": 2.2272640610104864e-07, "loss": 1.4123, "step": 48606 }, { "epoch": 9.26, "learning_rate": 2.226692087702574e-07, "loss": 1.9171, "step": 48607 }, { "epoch": 9.26, "learning_rate": 2.2261201143946615e-07, "loss": 0.8524, "step": 48608 }, { "epoch": 9.26, "learning_rate": 2.2255481410867492e-07, "loss": 1.9058, "step": 48609 }, { "epoch": 9.26, "learning_rate": 2.2249761677788371e-07, "loss": 2.2877, "step": 48610 }, { "epoch": 9.26, "learning_rate": 2.2244041944709248e-07, "loss": 1.3731, "step": 48611 }, { "epoch": 9.26, "learning_rate": 2.2238322211630125e-07, "loss": 2.6546, "step": 48612 }, { "epoch": 9.26, "learning_rate": 2.2232602478551004e-07, "loss": 1.5265, "step": 48613 }, { "epoch": 9.26, "learning_rate": 2.2226882745471878e-07, "loss": 3.1946, "step": 48614 }, { "epoch": 9.26, "learning_rate": 2.2221163012392755e-07, "loss": 2.2714, "step": 48615 }, { "epoch": 9.26, "learning_rate": 2.2215443279313632e-07, "loss": 1.2766, "step": 48616 }, { "epoch": 9.26, "learning_rate": 2.220972354623451e-07, "loss": 4.0083, "step": 48617 }, { "epoch": 9.26, "learning_rate": 2.2204003813155388e-07, "loss": 1.5172, "step": 48618 }, { "epoch": 9.26, "learning_rate": 2.2198284080076265e-07, "loss": 1.9484, "step": 48619 }, { "epoch": 9.26, "learning_rate": 2.219256434699714e-07, "loss": 1.484, "step": 48620 }, { "epoch": 9.26, "learning_rate": 2.2186844613918018e-07, "loss": 3.0538, "step": 48621 }, { "epoch": 9.26, "learning_rate": 2.2181124880838895e-07, "loss": 0.9103, "step": 48622 }, { "epoch": 9.26, "learning_rate": 2.2175405147759772e-07, "loss": 2.4942, "step": 48623 }, { "epoch": 9.26, "learning_rate": 2.216968541468065e-07, "loss": 2.1685, "step": 48624 }, { "epoch": 9.26, "learning_rate": 2.2163965681601525e-07, "loss": 0.4173, "step": 48625 }, { "epoch": 9.26, "learning_rate": 2.2158245948522402e-07, "loss": 1.7247, "step": 48626 }, { "epoch": 9.26, "learning_rate": 2.2152526215443278e-07, "loss": 2.3451, "step": 48627 }, { "epoch": 9.26, "learning_rate": 2.2146806482364158e-07, "loss": 2.7274, "step": 48628 }, { "epoch": 9.26, "learning_rate": 2.2141086749285035e-07, "loss": 0.9036, "step": 48629 }, { "epoch": 9.26, "learning_rate": 2.2135367016205911e-07, "loss": 2.8257, "step": 48630 }, { "epoch": 9.26, "learning_rate": 2.2129647283126785e-07, "loss": 1.7569, "step": 48631 }, { "epoch": 9.26, "learning_rate": 2.2123927550047665e-07, "loss": 0.807, "step": 48632 }, { "epoch": 9.26, "learning_rate": 2.2118207816968542e-07, "loss": 2.0381, "step": 48633 }, { "epoch": 9.26, "learning_rate": 2.2112488083889418e-07, "loss": 1.2915, "step": 48634 }, { "epoch": 9.26, "learning_rate": 2.2106768350810298e-07, "loss": 2.2458, "step": 48635 }, { "epoch": 9.26, "learning_rate": 2.2101048617731174e-07, "loss": 0.9577, "step": 48636 }, { "epoch": 9.26, "learning_rate": 2.2095328884652048e-07, "loss": 0.4356, "step": 48637 }, { "epoch": 9.26, "learning_rate": 2.2089609151572925e-07, "loss": 2.1946, "step": 48638 }, { "epoch": 9.26, "learning_rate": 2.2083889418493805e-07, "loss": 3.4413, "step": 48639 }, { "epoch": 9.26, "learning_rate": 2.2078169685414681e-07, "loss": 1.8016, "step": 48640 }, { "epoch": 9.26, "learning_rate": 2.2072449952335558e-07, "loss": 1.3989, "step": 48641 }, { "epoch": 9.27, "learning_rate": 2.2066730219256437e-07, "loss": 0.2231, "step": 48642 }, { "epoch": 9.27, "learning_rate": 2.2061010486177312e-07, "loss": 2.0869, "step": 48643 }, { "epoch": 9.27, "learning_rate": 2.2055290753098188e-07, "loss": 1.8028, "step": 48644 }, { "epoch": 9.27, "learning_rate": 2.2049571020019065e-07, "loss": 1.3733, "step": 48645 }, { "epoch": 9.27, "learning_rate": 2.2043851286939944e-07, "loss": 1.1419, "step": 48646 }, { "epoch": 9.27, "learning_rate": 2.203813155386082e-07, "loss": 1.6444, "step": 48647 }, { "epoch": 9.27, "learning_rate": 2.2032411820781698e-07, "loss": 0.8325, "step": 48648 }, { "epoch": 9.27, "learning_rate": 2.2026692087702572e-07, "loss": 0.4218, "step": 48649 }, { "epoch": 9.27, "learning_rate": 2.2020972354623451e-07, "loss": 1.6249, "step": 48650 }, { "epoch": 9.27, "learning_rate": 2.2015252621544328e-07, "loss": 0.6786, "step": 48651 }, { "epoch": 9.27, "learning_rate": 2.2009532888465205e-07, "loss": 1.9154, "step": 48652 }, { "epoch": 9.27, "learning_rate": 2.2003813155386084e-07, "loss": 1.2452, "step": 48653 }, { "epoch": 9.27, "learning_rate": 2.199809342230696e-07, "loss": 1.1889, "step": 48654 }, { "epoch": 9.27, "learning_rate": 2.1992373689227835e-07, "loss": 2.1422, "step": 48655 }, { "epoch": 9.27, "learning_rate": 2.1986653956148712e-07, "loss": 0.8236, "step": 48656 }, { "epoch": 9.27, "learning_rate": 2.198093422306959e-07, "loss": 1.8904, "step": 48657 }, { "epoch": 9.27, "learning_rate": 2.1975214489990468e-07, "loss": 2.0738, "step": 48658 }, { "epoch": 9.27, "learning_rate": 2.1969494756911345e-07, "loss": 1.0956, "step": 48659 }, { "epoch": 9.27, "learning_rate": 2.1963775023832224e-07, "loss": 1.724, "step": 48660 }, { "epoch": 9.27, "learning_rate": 2.1958055290753098e-07, "loss": 1.0373, "step": 48661 }, { "epoch": 9.27, "learning_rate": 2.1952335557673975e-07, "loss": 1.3086, "step": 48662 }, { "epoch": 9.27, "learning_rate": 2.1946615824594852e-07, "loss": 0.7046, "step": 48663 }, { "epoch": 9.27, "learning_rate": 2.194089609151573e-07, "loss": 1.8758, "step": 48664 }, { "epoch": 9.27, "learning_rate": 2.1935176358436608e-07, "loss": 1.9707, "step": 48665 }, { "epoch": 9.27, "learning_rate": 2.1929456625357487e-07, "loss": 2.1946, "step": 48666 }, { "epoch": 9.27, "learning_rate": 2.1923736892278358e-07, "loss": 1.3731, "step": 48667 }, { "epoch": 9.27, "learning_rate": 2.1918017159199238e-07, "loss": 2.2265, "step": 48668 }, { "epoch": 9.27, "learning_rate": 2.1912297426120115e-07, "loss": 2.1782, "step": 48669 }, { "epoch": 9.27, "learning_rate": 2.190657769304099e-07, "loss": 1.0003, "step": 48670 }, { "epoch": 9.27, "learning_rate": 2.190085795996187e-07, "loss": 1.5388, "step": 48671 }, { "epoch": 9.27, "learning_rate": 2.1895138226882747e-07, "loss": 2.0871, "step": 48672 }, { "epoch": 9.27, "learning_rate": 2.1889418493803622e-07, "loss": 1.5944, "step": 48673 }, { "epoch": 9.27, "learning_rate": 2.1883698760724498e-07, "loss": 1.2249, "step": 48674 }, { "epoch": 9.27, "learning_rate": 2.1877979027645378e-07, "loss": 1.4576, "step": 48675 }, { "epoch": 9.27, "learning_rate": 2.1872259294566254e-07, "loss": 1.7294, "step": 48676 }, { "epoch": 9.27, "learning_rate": 2.1866539561487134e-07, "loss": 1.7808, "step": 48677 }, { "epoch": 9.27, "learning_rate": 2.186081982840801e-07, "loss": 1.2737, "step": 48678 }, { "epoch": 9.27, "learning_rate": 2.1855100095328885e-07, "loss": 1.1927, "step": 48679 }, { "epoch": 9.27, "learning_rate": 2.184938036224976e-07, "loss": 2.4651, "step": 48680 }, { "epoch": 9.27, "learning_rate": 2.1843660629170638e-07, "loss": 2.0557, "step": 48681 }, { "epoch": 9.27, "learning_rate": 2.1837940896091517e-07, "loss": 0.446, "step": 48682 }, { "epoch": 9.27, "learning_rate": 2.1832221163012394e-07, "loss": 0.6602, "step": 48683 }, { "epoch": 9.27, "learning_rate": 2.1826501429933274e-07, "loss": 1.132, "step": 48684 }, { "epoch": 9.27, "learning_rate": 2.1820781696854145e-07, "loss": 2.3671, "step": 48685 }, { "epoch": 9.27, "learning_rate": 2.1815061963775024e-07, "loss": 1.7527, "step": 48686 }, { "epoch": 9.27, "learning_rate": 2.18093422306959e-07, "loss": 0.8812, "step": 48687 }, { "epoch": 9.27, "learning_rate": 2.180362249761678e-07, "loss": 0.4844, "step": 48688 }, { "epoch": 9.27, "learning_rate": 2.1797902764537657e-07, "loss": 1.2286, "step": 48689 }, { "epoch": 9.27, "learning_rate": 2.179218303145853e-07, "loss": 0.8138, "step": 48690 }, { "epoch": 9.27, "learning_rate": 2.1786463298379408e-07, "loss": 2.061, "step": 48691 }, { "epoch": 9.27, "learning_rate": 2.1780743565300285e-07, "loss": 1.9262, "step": 48692 }, { "epoch": 9.27, "learning_rate": 2.1775023832221164e-07, "loss": 0.6078, "step": 48693 }, { "epoch": 9.28, "learning_rate": 2.176930409914204e-07, "loss": 1.0925, "step": 48694 }, { "epoch": 9.28, "learning_rate": 2.176358436606292e-07, "loss": 1.9441, "step": 48695 }, { "epoch": 9.28, "learning_rate": 2.1757864632983792e-07, "loss": 0.6736, "step": 48696 }, { "epoch": 9.28, "learning_rate": 2.175214489990467e-07, "loss": 0.2503, "step": 48697 }, { "epoch": 9.28, "learning_rate": 2.1746425166825548e-07, "loss": 0.4667, "step": 48698 }, { "epoch": 9.28, "learning_rate": 2.1740705433746425e-07, "loss": 1.3078, "step": 48699 }, { "epoch": 9.28, "learning_rate": 2.1734985700667304e-07, "loss": 0.8599, "step": 48700 }, { "epoch": 9.28, "learning_rate": 2.172926596758818e-07, "loss": 1.7279, "step": 48701 }, { "epoch": 9.28, "learning_rate": 2.1723546234509055e-07, "loss": 1.3946, "step": 48702 }, { "epoch": 9.28, "learning_rate": 2.1717826501429932e-07, "loss": 1.4752, "step": 48703 }, { "epoch": 9.28, "learning_rate": 2.171210676835081e-07, "loss": 2.2615, "step": 48704 }, { "epoch": 9.28, "learning_rate": 2.1706387035271688e-07, "loss": 2.4682, "step": 48705 }, { "epoch": 9.28, "learning_rate": 2.1700667302192567e-07, "loss": 3.2168, "step": 48706 }, { "epoch": 9.28, "learning_rate": 2.1694947569113444e-07, "loss": 1.537, "step": 48707 }, { "epoch": 9.28, "learning_rate": 2.1689227836034318e-07, "loss": 1.9094, "step": 48708 }, { "epoch": 9.28, "learning_rate": 2.1683508102955195e-07, "loss": 2.3506, "step": 48709 }, { "epoch": 9.28, "learning_rate": 2.167778836987607e-07, "loss": 1.9675, "step": 48710 }, { "epoch": 9.28, "learning_rate": 2.167206863679695e-07, "loss": 2.3437, "step": 48711 }, { "epoch": 9.28, "learning_rate": 2.1666348903717827e-07, "loss": 0.6746, "step": 48712 }, { "epoch": 9.28, "learning_rate": 2.1660629170638707e-07, "loss": 1.8595, "step": 48713 }, { "epoch": 9.28, "learning_rate": 2.1654909437559578e-07, "loss": 2.5253, "step": 48714 }, { "epoch": 9.28, "learning_rate": 2.1649189704480458e-07, "loss": 2.1203, "step": 48715 }, { "epoch": 9.28, "learning_rate": 2.1643469971401334e-07, "loss": 1.5659, "step": 48716 }, { "epoch": 9.28, "learning_rate": 2.1637750238322214e-07, "loss": 2.2156, "step": 48717 }, { "epoch": 9.28, "learning_rate": 2.163203050524309e-07, "loss": 1.6757, "step": 48718 }, { "epoch": 9.28, "learning_rate": 2.1626310772163967e-07, "loss": 2.2407, "step": 48719 }, { "epoch": 9.28, "learning_rate": 2.162059103908484e-07, "loss": 2.5465, "step": 48720 }, { "epoch": 9.28, "learning_rate": 2.1614871306005718e-07, "loss": 1.2496, "step": 48721 }, { "epoch": 9.28, "learning_rate": 2.1609151572926597e-07, "loss": 1.0135, "step": 48722 }, { "epoch": 9.28, "learning_rate": 2.1603431839847474e-07, "loss": 1.4392, "step": 48723 }, { "epoch": 9.28, "learning_rate": 2.1597712106768354e-07, "loss": 0.595, "step": 48724 }, { "epoch": 9.28, "learning_rate": 2.159199237368923e-07, "loss": 0.5504, "step": 48725 }, { "epoch": 9.28, "learning_rate": 2.1586272640610104e-07, "loss": 0.5576, "step": 48726 }, { "epoch": 9.28, "learning_rate": 2.158055290753098e-07, "loss": 1.0482, "step": 48727 }, { "epoch": 9.28, "learning_rate": 2.157483317445186e-07, "loss": 1.0098, "step": 48728 }, { "epoch": 9.28, "learning_rate": 2.1569113441372737e-07, "loss": 1.8944, "step": 48729 }, { "epoch": 9.28, "learning_rate": 2.1563393708293614e-07, "loss": 1.0121, "step": 48730 }, { "epoch": 9.28, "learning_rate": 2.1557673975214493e-07, "loss": 1.2091, "step": 48731 }, { "epoch": 9.28, "learning_rate": 2.1551954242135365e-07, "loss": 1.4081, "step": 48732 }, { "epoch": 9.28, "learning_rate": 2.1546234509056244e-07, "loss": 0.6447, "step": 48733 }, { "epoch": 9.28, "learning_rate": 2.154051477597712e-07, "loss": 2.056, "step": 48734 }, { "epoch": 9.28, "learning_rate": 2.1534795042898e-07, "loss": 2.9251, "step": 48735 }, { "epoch": 9.28, "learning_rate": 2.1529075309818877e-07, "loss": 1.307, "step": 48736 }, { "epoch": 9.28, "learning_rate": 2.1523355576739754e-07, "loss": 1.3224, "step": 48737 }, { "epoch": 9.28, "learning_rate": 2.1517635843660628e-07, "loss": 0.4656, "step": 48738 }, { "epoch": 9.28, "learning_rate": 2.1511916110581507e-07, "loss": 0.9463, "step": 48739 }, { "epoch": 9.28, "learning_rate": 2.1506196377502384e-07, "loss": 2.9947, "step": 48740 }, { "epoch": 9.28, "learning_rate": 2.150047664442326e-07, "loss": 1.4714, "step": 48741 }, { "epoch": 9.28, "learning_rate": 2.149475691134414e-07, "loss": 0.6978, "step": 48742 }, { "epoch": 9.28, "learning_rate": 2.1489037178265017e-07, "loss": 0.2265, "step": 48743 }, { "epoch": 9.28, "learning_rate": 2.148331744518589e-07, "loss": 0.618, "step": 48744 }, { "epoch": 9.28, "learning_rate": 2.1477597712106768e-07, "loss": 2.6504, "step": 48745 }, { "epoch": 9.28, "learning_rate": 2.1471877979027647e-07, "loss": 1.2138, "step": 48746 }, { "epoch": 9.29, "learning_rate": 2.1466158245948524e-07, "loss": 1.4392, "step": 48747 }, { "epoch": 9.29, "learning_rate": 2.14604385128694e-07, "loss": 1.4722, "step": 48748 }, { "epoch": 9.29, "learning_rate": 2.1454718779790275e-07, "loss": 1.2996, "step": 48749 }, { "epoch": 9.29, "learning_rate": 2.1448999046711154e-07, "loss": 0.5219, "step": 48750 }, { "epoch": 9.29, "learning_rate": 2.144327931363203e-07, "loss": 1.9642, "step": 48751 }, { "epoch": 9.29, "learning_rate": 2.1437559580552907e-07, "loss": 0.778, "step": 48752 }, { "epoch": 9.29, "learning_rate": 2.1431839847473787e-07, "loss": 2.6739, "step": 48753 }, { "epoch": 9.29, "learning_rate": 2.1426120114394664e-07, "loss": 0.9928, "step": 48754 }, { "epoch": 9.29, "learning_rate": 2.1420400381315538e-07, "loss": 0.9118, "step": 48755 }, { "epoch": 9.29, "learning_rate": 2.1414680648236414e-07, "loss": 2.5111, "step": 48756 }, { "epoch": 9.29, "learning_rate": 2.1408960915157294e-07, "loss": 0.8045, "step": 48757 }, { "epoch": 9.29, "learning_rate": 2.140324118207817e-07, "loss": 1.3774, "step": 48758 }, { "epoch": 9.29, "learning_rate": 2.1397521448999047e-07, "loss": 1.6532, "step": 48759 }, { "epoch": 9.29, "learning_rate": 2.1391801715919927e-07, "loss": 1.3478, "step": 48760 }, { "epoch": 9.29, "learning_rate": 2.13860819828408e-07, "loss": 1.5585, "step": 48761 }, { "epoch": 9.29, "learning_rate": 2.1380362249761677e-07, "loss": 2.6252, "step": 48762 }, { "epoch": 9.29, "learning_rate": 2.1374642516682554e-07, "loss": 2.4363, "step": 48763 }, { "epoch": 9.29, "learning_rate": 2.1368922783603434e-07, "loss": 2.3895, "step": 48764 }, { "epoch": 9.29, "learning_rate": 2.136320305052431e-07, "loss": 0.7579, "step": 48765 }, { "epoch": 9.29, "learning_rate": 2.1357483317445187e-07, "loss": 3.3643, "step": 48766 }, { "epoch": 9.29, "learning_rate": 2.135176358436606e-07, "loss": 0.8895, "step": 48767 }, { "epoch": 9.29, "learning_rate": 2.134604385128694e-07, "loss": 0.563, "step": 48768 }, { "epoch": 9.29, "learning_rate": 2.1340324118207817e-07, "loss": 1.3695, "step": 48769 }, { "epoch": 9.29, "learning_rate": 2.1334604385128694e-07, "loss": 2.3757, "step": 48770 }, { "epoch": 9.29, "learning_rate": 2.1328884652049573e-07, "loss": 2.7833, "step": 48771 }, { "epoch": 9.29, "learning_rate": 2.132316491897045e-07, "loss": 1.9327, "step": 48772 }, { "epoch": 9.29, "learning_rate": 2.1317445185891324e-07, "loss": 1.4175, "step": 48773 }, { "epoch": 9.29, "learning_rate": 2.13117254528122e-07, "loss": 1.2316, "step": 48774 }, { "epoch": 9.29, "learning_rate": 2.130600571973308e-07, "loss": 1.7977, "step": 48775 }, { "epoch": 9.29, "learning_rate": 2.1300285986653957e-07, "loss": 0.6574, "step": 48776 }, { "epoch": 9.29, "learning_rate": 2.1294566253574834e-07, "loss": 1.1831, "step": 48777 }, { "epoch": 9.29, "learning_rate": 2.1288846520495713e-07, "loss": 1.2389, "step": 48778 }, { "epoch": 9.29, "learning_rate": 2.1283126787416587e-07, "loss": 2.7072, "step": 48779 }, { "epoch": 9.29, "learning_rate": 2.1277407054337464e-07, "loss": 2.1821, "step": 48780 }, { "epoch": 9.29, "learning_rate": 2.127168732125834e-07, "loss": 0.6634, "step": 48781 }, { "epoch": 9.29, "learning_rate": 2.126596758817922e-07, "loss": 1.8627, "step": 48782 }, { "epoch": 9.29, "learning_rate": 2.1260247855100097e-07, "loss": 0.2623, "step": 48783 }, { "epoch": 9.29, "learning_rate": 2.1254528122020973e-07, "loss": 0.9771, "step": 48784 }, { "epoch": 9.29, "learning_rate": 2.1248808388941848e-07, "loss": 2.371, "step": 48785 }, { "epoch": 9.29, "learning_rate": 2.1243088655862727e-07, "loss": 0.8923, "step": 48786 }, { "epoch": 9.29, "learning_rate": 2.1237368922783604e-07, "loss": 1.1332, "step": 48787 }, { "epoch": 9.29, "learning_rate": 2.123164918970448e-07, "loss": 1.3689, "step": 48788 }, { "epoch": 9.29, "learning_rate": 2.122592945662536e-07, "loss": 1.1782, "step": 48789 }, { "epoch": 9.29, "learning_rate": 2.1220209723546237e-07, "loss": 3.6739, "step": 48790 }, { "epoch": 9.29, "learning_rate": 2.121448999046711e-07, "loss": 0.9694, "step": 48791 }, { "epoch": 9.29, "learning_rate": 2.1208770257387987e-07, "loss": 1.1853, "step": 48792 }, { "epoch": 9.29, "learning_rate": 2.1203050524308867e-07, "loss": 2.1073, "step": 48793 }, { "epoch": 9.29, "learning_rate": 2.1197330791229743e-07, "loss": 2.133, "step": 48794 }, { "epoch": 9.29, "learning_rate": 2.119161105815062e-07, "loss": 0.9645, "step": 48795 }, { "epoch": 9.29, "learning_rate": 2.11858913250715e-07, "loss": 0.26, "step": 48796 }, { "epoch": 9.29, "learning_rate": 2.1180171591992374e-07, "loss": 2.381, "step": 48797 }, { "epoch": 9.29, "learning_rate": 2.117445185891325e-07, "loss": 1.7025, "step": 48798 }, { "epoch": 9.3, "learning_rate": 2.1168732125834127e-07, "loss": 1.4351, "step": 48799 }, { "epoch": 9.3, "learning_rate": 2.1163012392755007e-07, "loss": 2.2912, "step": 48800 }, { "epoch": 9.3, "learning_rate": 2.1157292659675883e-07, "loss": 1.2889, "step": 48801 }, { "epoch": 9.3, "learning_rate": 2.1151572926596763e-07, "loss": 0.54, "step": 48802 }, { "epoch": 9.3, "learning_rate": 2.1145853193517634e-07, "loss": 0.9048, "step": 48803 }, { "epoch": 9.3, "learning_rate": 2.1140133460438513e-07, "loss": 1.2507, "step": 48804 }, { "epoch": 9.3, "learning_rate": 2.113441372735939e-07, "loss": 1.01, "step": 48805 }, { "epoch": 9.3, "learning_rate": 2.1128693994280267e-07, "loss": 1.9241, "step": 48806 }, { "epoch": 9.3, "learning_rate": 2.1122974261201146e-07, "loss": 1.0346, "step": 48807 }, { "epoch": 9.3, "learning_rate": 2.111725452812202e-07, "loss": 2.2855, "step": 48808 }, { "epoch": 9.3, "learning_rate": 2.1111534795042897e-07, "loss": 1.0884, "step": 48809 }, { "epoch": 9.3, "learning_rate": 2.1105815061963774e-07, "loss": 0.967, "step": 48810 }, { "epoch": 9.3, "learning_rate": 2.1100095328884653e-07, "loss": 1.9569, "step": 48811 }, { "epoch": 9.3, "learning_rate": 2.109437559580553e-07, "loss": 0.9254, "step": 48812 }, { "epoch": 9.3, "learning_rate": 2.108865586272641e-07, "loss": 0.9986, "step": 48813 }, { "epoch": 9.3, "learning_rate": 2.108293612964728e-07, "loss": 1.6127, "step": 48814 }, { "epoch": 9.3, "learning_rate": 2.107721639656816e-07, "loss": 1.0869, "step": 48815 }, { "epoch": 9.3, "learning_rate": 2.1071496663489037e-07, "loss": 3.2372, "step": 48816 }, { "epoch": 9.3, "learning_rate": 2.1065776930409914e-07, "loss": 2.5344, "step": 48817 }, { "epoch": 9.3, "learning_rate": 2.1060057197330793e-07, "loss": 1.584, "step": 48818 }, { "epoch": 9.3, "learning_rate": 2.105433746425167e-07, "loss": 1.4821, "step": 48819 }, { "epoch": 9.3, "learning_rate": 2.1048617731172544e-07, "loss": 2.3786, "step": 48820 }, { "epoch": 9.3, "learning_rate": 2.104289799809342e-07, "loss": 2.3168, "step": 48821 }, { "epoch": 9.3, "learning_rate": 2.10371782650143e-07, "loss": 2.1897, "step": 48822 }, { "epoch": 9.3, "learning_rate": 2.1031458531935177e-07, "loss": 1.4153, "step": 48823 }, { "epoch": 9.3, "learning_rate": 2.1025738798856056e-07, "loss": 1.3648, "step": 48824 }, { "epoch": 9.3, "learning_rate": 2.1020019065776933e-07, "loss": 0.2746, "step": 48825 }, { "epoch": 9.3, "learning_rate": 2.1014299332697807e-07, "loss": 1.2998, "step": 48826 }, { "epoch": 9.3, "learning_rate": 2.1008579599618684e-07, "loss": 3.063, "step": 48827 }, { "epoch": 9.3, "learning_rate": 2.100285986653956e-07, "loss": 1.8776, "step": 48828 }, { "epoch": 9.3, "learning_rate": 2.099714013346044e-07, "loss": 1.7616, "step": 48829 }, { "epoch": 9.3, "learning_rate": 2.0991420400381317e-07, "loss": 2.411, "step": 48830 }, { "epoch": 9.3, "learning_rate": 2.0985700667302196e-07, "loss": 1.5196, "step": 48831 }, { "epoch": 9.3, "learning_rate": 2.0979980934223067e-07, "loss": 2.2742, "step": 48832 }, { "epoch": 9.3, "learning_rate": 2.0974261201143947e-07, "loss": 1.0869, "step": 48833 }, { "epoch": 9.3, "learning_rate": 2.0968541468064823e-07, "loss": 0.8901, "step": 48834 }, { "epoch": 9.3, "learning_rate": 2.0962821734985703e-07, "loss": 1.6666, "step": 48835 }, { "epoch": 9.3, "learning_rate": 2.095710200190658e-07, "loss": 1.2175, "step": 48836 }, { "epoch": 9.3, "learning_rate": 2.0951382268827456e-07, "loss": 1.0785, "step": 48837 }, { "epoch": 9.3, "learning_rate": 2.094566253574833e-07, "loss": 1.7572, "step": 48838 }, { "epoch": 9.3, "learning_rate": 2.0939942802669207e-07, "loss": 4.2543, "step": 48839 }, { "epoch": 9.3, "learning_rate": 2.0934223069590087e-07, "loss": 1.126, "step": 48840 }, { "epoch": 9.3, "learning_rate": 2.0928503336510963e-07, "loss": 1.572, "step": 48841 }, { "epoch": 9.3, "learning_rate": 2.0922783603431843e-07, "loss": 1.6006, "step": 48842 }, { "epoch": 9.3, "learning_rate": 2.091706387035272e-07, "loss": 2.3219, "step": 48843 }, { "epoch": 9.3, "learning_rate": 2.0911344137273593e-07, "loss": 1.2477, "step": 48844 }, { "epoch": 9.3, "learning_rate": 2.090562440419447e-07, "loss": 0.7709, "step": 48845 }, { "epoch": 9.3, "learning_rate": 2.0899904671115347e-07, "loss": 0.7808, "step": 48846 }, { "epoch": 9.3, "learning_rate": 2.0894184938036226e-07, "loss": 1.2614, "step": 48847 }, { "epoch": 9.3, "learning_rate": 2.0888465204957103e-07, "loss": 1.1069, "step": 48848 }, { "epoch": 9.3, "learning_rate": 2.0882745471877982e-07, "loss": 1.3727, "step": 48849 }, { "epoch": 9.3, "learning_rate": 2.0877025738798854e-07, "loss": 0.7673, "step": 48850 }, { "epoch": 9.3, "learning_rate": 2.0871306005719733e-07, "loss": 0.8556, "step": 48851 }, { "epoch": 9.31, "learning_rate": 2.086558627264061e-07, "loss": 1.8554, "step": 48852 }, { "epoch": 9.31, "learning_rate": 2.085986653956149e-07, "loss": 0.505, "step": 48853 }, { "epoch": 9.31, "learning_rate": 2.0854146806482366e-07, "loss": 1.9304, "step": 48854 }, { "epoch": 9.31, "learning_rate": 2.0848427073403243e-07, "loss": 2.974, "step": 48855 }, { "epoch": 9.31, "learning_rate": 2.0842707340324117e-07, "loss": 1.6617, "step": 48856 }, { "epoch": 9.31, "learning_rate": 2.0836987607244994e-07, "loss": 1.5612, "step": 48857 }, { "epoch": 9.31, "learning_rate": 2.0831267874165873e-07, "loss": 0.421, "step": 48858 }, { "epoch": 9.31, "learning_rate": 2.082554814108675e-07, "loss": 1.1992, "step": 48859 }, { "epoch": 9.31, "learning_rate": 2.081982840800763e-07, "loss": 0.8855, "step": 48860 }, { "epoch": 9.31, "learning_rate": 2.0814108674928506e-07, "loss": 0.9896, "step": 48861 }, { "epoch": 9.31, "learning_rate": 2.080838894184938e-07, "loss": 1.9301, "step": 48862 }, { "epoch": 9.31, "learning_rate": 2.0802669208770257e-07, "loss": 0.9363, "step": 48863 }, { "epoch": 9.31, "learning_rate": 2.0796949475691136e-07, "loss": 2.7803, "step": 48864 }, { "epoch": 9.31, "learning_rate": 2.0791229742612013e-07, "loss": 1.4816, "step": 48865 }, { "epoch": 9.31, "learning_rate": 2.078551000953289e-07, "loss": 2.3111, "step": 48866 }, { "epoch": 9.31, "learning_rate": 2.0779790276453764e-07, "loss": 1.761, "step": 48867 }, { "epoch": 9.31, "learning_rate": 2.077407054337464e-07, "loss": 1.0173, "step": 48868 }, { "epoch": 9.31, "learning_rate": 2.076835081029552e-07, "loss": 1.243, "step": 48869 }, { "epoch": 9.31, "learning_rate": 2.0762631077216397e-07, "loss": 1.5697, "step": 48870 }, { "epoch": 9.31, "learning_rate": 2.0756911344137276e-07, "loss": 2.4564, "step": 48871 }, { "epoch": 9.31, "learning_rate": 2.0751191611058153e-07, "loss": 1.9025, "step": 48872 }, { "epoch": 9.31, "learning_rate": 2.0745471877979027e-07, "loss": 1.4659, "step": 48873 }, { "epoch": 9.31, "learning_rate": 2.0739752144899903e-07, "loss": 0.921, "step": 48874 }, { "epoch": 9.31, "learning_rate": 2.0734032411820783e-07, "loss": 1.5406, "step": 48875 }, { "epoch": 9.31, "learning_rate": 2.072831267874166e-07, "loss": 1.8982, "step": 48876 }, { "epoch": 9.31, "learning_rate": 2.0722592945662536e-07, "loss": 1.9192, "step": 48877 }, { "epoch": 9.31, "learning_rate": 2.0716873212583416e-07, "loss": 1.5242, "step": 48878 }, { "epoch": 9.31, "learning_rate": 2.0711153479504287e-07, "loss": 1.7805, "step": 48879 }, { "epoch": 9.31, "learning_rate": 2.0705433746425167e-07, "loss": 1.4883, "step": 48880 }, { "epoch": 9.31, "learning_rate": 2.0699714013346043e-07, "loss": 1.9072, "step": 48881 }, { "epoch": 9.31, "learning_rate": 2.0693994280266923e-07, "loss": 1.9216, "step": 48882 }, { "epoch": 9.31, "learning_rate": 2.06882745471878e-07, "loss": 0.5948, "step": 48883 }, { "epoch": 9.31, "learning_rate": 2.0682554814108676e-07, "loss": 2.543, "step": 48884 }, { "epoch": 9.31, "learning_rate": 2.067683508102955e-07, "loss": 1.7137, "step": 48885 }, { "epoch": 9.31, "learning_rate": 2.067111534795043e-07, "loss": 2.6409, "step": 48886 }, { "epoch": 9.31, "learning_rate": 2.0665395614871306e-07, "loss": 1.9281, "step": 48887 }, { "epoch": 9.31, "learning_rate": 2.0659675881792183e-07, "loss": 1.4883, "step": 48888 }, { "epoch": 9.31, "learning_rate": 2.0653956148713062e-07, "loss": 2.4979, "step": 48889 }, { "epoch": 9.31, "learning_rate": 2.064823641563394e-07, "loss": 2.2161, "step": 48890 }, { "epoch": 9.31, "learning_rate": 2.0642516682554813e-07, "loss": 2.3346, "step": 48891 }, { "epoch": 9.31, "learning_rate": 2.063679694947569e-07, "loss": 1.4752, "step": 48892 }, { "epoch": 9.31, "learning_rate": 2.063107721639657e-07, "loss": 2.5833, "step": 48893 }, { "epoch": 9.31, "learning_rate": 2.0625357483317446e-07, "loss": 2.5844, "step": 48894 }, { "epoch": 9.31, "learning_rate": 2.0619637750238323e-07, "loss": 1.5112, "step": 48895 }, { "epoch": 9.31, "learning_rate": 2.0613918017159202e-07, "loss": 1.0334, "step": 48896 }, { "epoch": 9.31, "learning_rate": 2.0608198284080076e-07, "loss": 0.898, "step": 48897 }, { "epoch": 9.31, "learning_rate": 2.0602478551000953e-07, "loss": 1.0925, "step": 48898 }, { "epoch": 9.31, "learning_rate": 2.059675881792183e-07, "loss": 1.3946, "step": 48899 }, { "epoch": 9.31, "learning_rate": 2.059103908484271e-07, "loss": 1.1394, "step": 48900 }, { "epoch": 9.31, "learning_rate": 2.0585319351763586e-07, "loss": 2.4158, "step": 48901 }, { "epoch": 9.31, "learning_rate": 2.0579599618684463e-07, "loss": 1.7129, "step": 48902 }, { "epoch": 9.31, "learning_rate": 2.0573879885605337e-07, "loss": 1.2654, "step": 48903 }, { "epoch": 9.32, "learning_rate": 2.0568160152526216e-07, "loss": 2.2387, "step": 48904 }, { "epoch": 9.32, "learning_rate": 2.0562440419447093e-07, "loss": 1.077, "step": 48905 }, { "epoch": 9.32, "learning_rate": 2.055672068636797e-07, "loss": 2.6202, "step": 48906 }, { "epoch": 9.32, "learning_rate": 2.055100095328885e-07, "loss": 0.9917, "step": 48907 }, { "epoch": 9.32, "learning_rate": 2.0545281220209726e-07, "loss": 3.4145, "step": 48908 }, { "epoch": 9.32, "learning_rate": 2.05395614871306e-07, "loss": 0.2977, "step": 48909 }, { "epoch": 9.32, "learning_rate": 2.0533841754051476e-07, "loss": 1.7572, "step": 48910 }, { "epoch": 9.32, "learning_rate": 2.0528122020972356e-07, "loss": 1.9939, "step": 48911 }, { "epoch": 9.32, "learning_rate": 2.0522402287893233e-07, "loss": 2.8123, "step": 48912 }, { "epoch": 9.32, "learning_rate": 2.051668255481411e-07, "loss": 1.4438, "step": 48913 }, { "epoch": 9.32, "learning_rate": 2.051096282173499e-07, "loss": 0.4469, "step": 48914 }, { "epoch": 9.32, "learning_rate": 2.0505243088655863e-07, "loss": 2.9511, "step": 48915 }, { "epoch": 9.32, "learning_rate": 2.049952335557674e-07, "loss": 1.1848, "step": 48916 }, { "epoch": 9.32, "learning_rate": 2.0493803622497616e-07, "loss": 0.5154, "step": 48917 }, { "epoch": 9.32, "learning_rate": 2.0488083889418496e-07, "loss": 2.4219, "step": 48918 }, { "epoch": 9.32, "learning_rate": 2.0482364156339372e-07, "loss": 1.4244, "step": 48919 }, { "epoch": 9.32, "learning_rate": 2.047664442326025e-07, "loss": 0.9031, "step": 48920 }, { "epoch": 9.32, "learning_rate": 2.0470924690181123e-07, "loss": 0.9081, "step": 48921 }, { "epoch": 9.32, "learning_rate": 2.0465204957102003e-07, "loss": 3.0207, "step": 48922 }, { "epoch": 9.32, "learning_rate": 2.045948522402288e-07, "loss": 4.8617, "step": 48923 }, { "epoch": 9.32, "learning_rate": 2.0453765490943756e-07, "loss": 0.5486, "step": 48924 }, { "epoch": 9.32, "learning_rate": 2.0448045757864635e-07, "loss": 1.4352, "step": 48925 }, { "epoch": 9.32, "learning_rate": 2.0442326024785512e-07, "loss": 1.5981, "step": 48926 }, { "epoch": 9.32, "learning_rate": 2.0436606291706386e-07, "loss": 0.9395, "step": 48927 }, { "epoch": 9.32, "learning_rate": 2.0430886558627263e-07, "loss": 1.3132, "step": 48928 }, { "epoch": 9.32, "learning_rate": 2.0425166825548142e-07, "loss": 2.8085, "step": 48929 }, { "epoch": 9.32, "learning_rate": 2.041944709246902e-07, "loss": 0.5907, "step": 48930 }, { "epoch": 9.32, "learning_rate": 2.0413727359389896e-07, "loss": 2.1041, "step": 48931 }, { "epoch": 9.32, "learning_rate": 2.040800762631077e-07, "loss": 1.3936, "step": 48932 }, { "epoch": 9.32, "learning_rate": 2.040228789323165e-07, "loss": 0.6793, "step": 48933 }, { "epoch": 9.32, "learning_rate": 2.0396568160152526e-07, "loss": 1.3656, "step": 48934 }, { "epoch": 9.32, "learning_rate": 2.0390848427073403e-07, "loss": 0.828, "step": 48935 }, { "epoch": 9.32, "learning_rate": 2.0385128693994282e-07, "loss": 0.4604, "step": 48936 }, { "epoch": 9.32, "learning_rate": 2.037940896091516e-07, "loss": 1.5229, "step": 48937 }, { "epoch": 9.32, "learning_rate": 2.0373689227836033e-07, "loss": 2.3507, "step": 48938 }, { "epoch": 9.32, "learning_rate": 2.036796949475691e-07, "loss": 3.5792, "step": 48939 }, { "epoch": 9.32, "learning_rate": 2.036224976167779e-07, "loss": 1.5826, "step": 48940 }, { "epoch": 9.32, "learning_rate": 2.0356530028598666e-07, "loss": 2.3747, "step": 48941 }, { "epoch": 9.32, "learning_rate": 2.0350810295519543e-07, "loss": 1.1592, "step": 48942 }, { "epoch": 9.32, "learning_rate": 2.0345090562440422e-07, "loss": 3.1942, "step": 48943 }, { "epoch": 9.32, "learning_rate": 2.0339370829361296e-07, "loss": 0.3572, "step": 48944 }, { "epoch": 9.32, "learning_rate": 2.0333651096282173e-07, "loss": 0.7454, "step": 48945 }, { "epoch": 9.32, "learning_rate": 2.032793136320305e-07, "loss": 1.9569, "step": 48946 }, { "epoch": 9.32, "learning_rate": 2.032221163012393e-07, "loss": 1.1605, "step": 48947 }, { "epoch": 9.32, "learning_rate": 2.0316491897044806e-07, "loss": 2.1906, "step": 48948 }, { "epoch": 9.32, "learning_rate": 2.0310772163965685e-07, "loss": 1.6727, "step": 48949 }, { "epoch": 9.32, "learning_rate": 2.0305052430886556e-07, "loss": 2.0338, "step": 48950 }, { "epoch": 9.32, "learning_rate": 2.0299332697807436e-07, "loss": 0.9484, "step": 48951 }, { "epoch": 9.32, "learning_rate": 2.0293612964728313e-07, "loss": 1.166, "step": 48952 }, { "epoch": 9.32, "learning_rate": 2.028789323164919e-07, "loss": 1.3194, "step": 48953 }, { "epoch": 9.32, "learning_rate": 2.028217349857007e-07, "loss": 1.5933, "step": 48954 }, { "epoch": 9.32, "learning_rate": 2.0276453765490945e-07, "loss": 2.1454, "step": 48955 }, { "epoch": 9.32, "learning_rate": 2.027073403241182e-07, "loss": 0.8079, "step": 48956 }, { "epoch": 9.33, "learning_rate": 2.0265014299332696e-07, "loss": 1.6124, "step": 48957 }, { "epoch": 9.33, "learning_rate": 2.0259294566253576e-07, "loss": 0.5037, "step": 48958 }, { "epoch": 9.33, "learning_rate": 2.0253574833174452e-07, "loss": 2.0277, "step": 48959 }, { "epoch": 9.33, "learning_rate": 2.0247855100095332e-07, "loss": 0.8431, "step": 48960 }, { "epoch": 9.33, "learning_rate": 2.0242135367016208e-07, "loss": 1.1835, "step": 48961 }, { "epoch": 9.33, "learning_rate": 2.0236415633937083e-07, "loss": 1.6698, "step": 48962 }, { "epoch": 9.33, "learning_rate": 2.023069590085796e-07, "loss": 1.8292, "step": 48963 }, { "epoch": 9.33, "learning_rate": 2.0224976167778836e-07, "loss": 1.1014, "step": 48964 }, { "epoch": 9.33, "learning_rate": 2.0219256434699715e-07, "loss": 1.0598, "step": 48965 }, { "epoch": 9.33, "learning_rate": 2.0213536701620592e-07, "loss": 0.9185, "step": 48966 }, { "epoch": 9.33, "learning_rate": 2.0207816968541472e-07, "loss": 3.2915, "step": 48967 }, { "epoch": 9.33, "learning_rate": 2.0202097235462343e-07, "loss": 0.478, "step": 48968 }, { "epoch": 9.33, "learning_rate": 2.0196377502383222e-07, "loss": 0.9736, "step": 48969 }, { "epoch": 9.33, "learning_rate": 2.01906577693041e-07, "loss": 2.2524, "step": 48970 }, { "epoch": 9.33, "learning_rate": 2.0184938036224978e-07, "loss": 2.919, "step": 48971 }, { "epoch": 9.33, "learning_rate": 2.0179218303145855e-07, "loss": 0.5021, "step": 48972 }, { "epoch": 9.33, "learning_rate": 2.0173498570066732e-07, "loss": 1.3829, "step": 48973 }, { "epoch": 9.33, "learning_rate": 2.0167778836987606e-07, "loss": 1.6964, "step": 48974 }, { "epoch": 9.33, "learning_rate": 2.0162059103908483e-07, "loss": 1.2665, "step": 48975 }, { "epoch": 9.33, "learning_rate": 2.0156339370829362e-07, "loss": 1.0934, "step": 48976 }, { "epoch": 9.33, "learning_rate": 2.015061963775024e-07, "loss": 2.2917, "step": 48977 }, { "epoch": 9.33, "learning_rate": 2.0144899904671118e-07, "loss": 0.8881, "step": 48978 }, { "epoch": 9.33, "learning_rate": 2.0139180171591995e-07, "loss": 1.2521, "step": 48979 }, { "epoch": 9.33, "learning_rate": 2.013346043851287e-07, "loss": 2.5469, "step": 48980 }, { "epoch": 9.33, "learning_rate": 2.0127740705433746e-07, "loss": 0.2007, "step": 48981 }, { "epoch": 9.33, "learning_rate": 2.0122020972354625e-07, "loss": 1.4627, "step": 48982 }, { "epoch": 9.33, "learning_rate": 2.0116301239275502e-07, "loss": 0.5123, "step": 48983 }, { "epoch": 9.33, "learning_rate": 2.011058150619638e-07, "loss": 1.4379, "step": 48984 }, { "epoch": 9.33, "learning_rate": 2.0104861773117258e-07, "loss": 1.0777, "step": 48985 }, { "epoch": 9.33, "learning_rate": 2.009914204003813e-07, "loss": 0.7272, "step": 48986 }, { "epoch": 9.33, "learning_rate": 2.009342230695901e-07, "loss": 1.3528, "step": 48987 }, { "epoch": 9.33, "learning_rate": 2.0087702573879886e-07, "loss": 2.0138, "step": 48988 }, { "epoch": 9.33, "learning_rate": 2.0081982840800765e-07, "loss": 2.2443, "step": 48989 }, { "epoch": 9.33, "learning_rate": 2.0076263107721642e-07, "loss": 0.7085, "step": 48990 }, { "epoch": 9.33, "learning_rate": 2.0070543374642516e-07, "loss": 1.3724, "step": 48991 }, { "epoch": 9.33, "learning_rate": 2.0064823641563393e-07, "loss": 2.0722, "step": 48992 }, { "epoch": 9.33, "learning_rate": 2.005910390848427e-07, "loss": 0.7303, "step": 48993 }, { "epoch": 9.33, "learning_rate": 2.005338417540515e-07, "loss": 3.0176, "step": 48994 }, { "epoch": 9.33, "learning_rate": 2.0047664442326025e-07, "loss": 1.6307, "step": 48995 }, { "epoch": 9.33, "learning_rate": 2.0041944709246905e-07, "loss": 0.8105, "step": 48996 }, { "epoch": 9.33, "learning_rate": 2.0036224976167776e-07, "loss": 1.7562, "step": 48997 }, { "epoch": 9.33, "learning_rate": 2.0030505243088656e-07, "loss": 1.4729, "step": 48998 }, { "epoch": 9.33, "learning_rate": 2.0024785510009532e-07, "loss": 1.6086, "step": 48999 }, { "epoch": 9.33, "learning_rate": 2.0019065776930412e-07, "loss": 1.7463, "step": 49000 }, { "epoch": 9.33, "learning_rate": 2.0013346043851288e-07, "loss": 1.3218, "step": 49001 }, { "epoch": 9.33, "learning_rate": 2.0007626310772165e-07, "loss": 1.3697, "step": 49002 }, { "epoch": 9.33, "learning_rate": 2.000190657769304e-07, "loss": 1.9425, "step": 49003 }, { "epoch": 9.33, "learning_rate": 1.9996186844613916e-07, "loss": 0.8025, "step": 49004 }, { "epoch": 9.33, "learning_rate": 1.9990467111534795e-07, "loss": 0.7167, "step": 49005 }, { "epoch": 9.33, "learning_rate": 1.9984747378455672e-07, "loss": 2.0327, "step": 49006 }, { "epoch": 9.33, "learning_rate": 1.9979027645376552e-07, "loss": 1.3912, "step": 49007 }, { "epoch": 9.33, "learning_rate": 1.9973307912297428e-07, "loss": 1.4004, "step": 49008 }, { "epoch": 9.34, "learning_rate": 1.9967588179218302e-07, "loss": 1.3979, "step": 49009 }, { "epoch": 9.34, "learning_rate": 1.996186844613918e-07, "loss": 2.1645, "step": 49010 }, { "epoch": 9.34, "learning_rate": 1.9956148713060058e-07, "loss": 0.3656, "step": 49011 }, { "epoch": 9.34, "learning_rate": 1.9950428979980935e-07, "loss": 1.1111, "step": 49012 }, { "epoch": 9.34, "learning_rate": 1.9944709246901812e-07, "loss": 2.2511, "step": 49013 }, { "epoch": 9.34, "learning_rate": 1.9938989513822691e-07, "loss": 1.8626, "step": 49014 }, { "epoch": 9.34, "learning_rate": 1.9933269780743563e-07, "loss": 1.4653, "step": 49015 }, { "epoch": 9.34, "learning_rate": 1.9927550047664442e-07, "loss": 1.6549, "step": 49016 }, { "epoch": 9.34, "learning_rate": 1.992183031458532e-07, "loss": 1.9092, "step": 49017 }, { "epoch": 9.34, "learning_rate": 1.9916110581506198e-07, "loss": 1.0261, "step": 49018 }, { "epoch": 9.34, "learning_rate": 1.9910390848427075e-07, "loss": 2.1375, "step": 49019 }, { "epoch": 9.34, "learning_rate": 1.9904671115347952e-07, "loss": 0.8669, "step": 49020 }, { "epoch": 9.34, "learning_rate": 1.9898951382268826e-07, "loss": 1.1774, "step": 49021 }, { "epoch": 9.34, "learning_rate": 1.9893231649189705e-07, "loss": 2.5544, "step": 49022 }, { "epoch": 9.34, "learning_rate": 1.9887511916110582e-07, "loss": 0.9399, "step": 49023 }, { "epoch": 9.34, "learning_rate": 1.9881792183031459e-07, "loss": 0.8806, "step": 49024 }, { "epoch": 9.34, "learning_rate": 1.9876072449952338e-07, "loss": 0.2785, "step": 49025 }, { "epoch": 9.34, "learning_rate": 1.9870352716873215e-07, "loss": 2.6635, "step": 49026 }, { "epoch": 9.34, "learning_rate": 1.986463298379409e-07, "loss": 2.0327, "step": 49027 }, { "epoch": 9.34, "learning_rate": 1.9858913250714966e-07, "loss": 1.6827, "step": 49028 }, { "epoch": 9.34, "learning_rate": 1.9853193517635845e-07, "loss": 1.3235, "step": 49029 }, { "epoch": 9.34, "learning_rate": 1.9847473784556722e-07, "loss": 2.288, "step": 49030 }, { "epoch": 9.34, "learning_rate": 1.9841754051477598e-07, "loss": 1.9852, "step": 49031 }, { "epoch": 9.34, "learning_rate": 1.9836034318398478e-07, "loss": 1.0268, "step": 49032 }, { "epoch": 9.34, "learning_rate": 1.9830314585319352e-07, "loss": 1.4184, "step": 49033 }, { "epoch": 9.34, "learning_rate": 1.9824594852240229e-07, "loss": 1.9857, "step": 49034 }, { "epoch": 9.34, "learning_rate": 1.9818875119161105e-07, "loss": 1.5811, "step": 49035 }, { "epoch": 9.34, "learning_rate": 1.9813155386081985e-07, "loss": 2.1444, "step": 49036 }, { "epoch": 9.34, "learning_rate": 1.9807435653002862e-07, "loss": 1.53, "step": 49037 }, { "epoch": 9.34, "learning_rate": 1.9801715919923738e-07, "loss": 0.9236, "step": 49038 }, { "epoch": 9.34, "learning_rate": 1.9795996186844612e-07, "loss": 3.2414, "step": 49039 }, { "epoch": 9.34, "learning_rate": 1.9790276453765492e-07, "loss": 2.1091, "step": 49040 }, { "epoch": 9.34, "learning_rate": 1.9784556720686368e-07, "loss": 1.8137, "step": 49041 }, { "epoch": 9.34, "learning_rate": 1.9778836987607245e-07, "loss": 1.7975, "step": 49042 }, { "epoch": 9.34, "learning_rate": 1.9773117254528125e-07, "loss": 2.4392, "step": 49043 }, { "epoch": 9.34, "learning_rate": 1.9767397521449e-07, "loss": 2.2373, "step": 49044 }, { "epoch": 9.34, "learning_rate": 1.9761677788369875e-07, "loss": 1.2787, "step": 49045 }, { "epoch": 9.34, "learning_rate": 1.9755958055290752e-07, "loss": 2.4955, "step": 49046 }, { "epoch": 9.34, "learning_rate": 1.9750238322211632e-07, "loss": 0.9468, "step": 49047 }, { "epoch": 9.34, "learning_rate": 1.9744518589132508e-07, "loss": 1.0713, "step": 49048 }, { "epoch": 9.34, "learning_rate": 1.9738798856053385e-07, "loss": 1.481, "step": 49049 }, { "epoch": 9.34, "learning_rate": 1.973307912297426e-07, "loss": 1.6311, "step": 49050 }, { "epoch": 9.34, "learning_rate": 1.9727359389895138e-07, "loss": 0.2323, "step": 49051 }, { "epoch": 9.34, "learning_rate": 1.9721639656816015e-07, "loss": 0.6048, "step": 49052 }, { "epoch": 9.34, "learning_rate": 1.9715919923736892e-07, "loss": 2.2447, "step": 49053 }, { "epoch": 9.34, "learning_rate": 1.971020019065777e-07, "loss": 1.5603, "step": 49054 }, { "epoch": 9.34, "learning_rate": 1.9704480457578648e-07, "loss": 1.7811, "step": 49055 }, { "epoch": 9.34, "learning_rate": 1.9698760724499522e-07, "loss": 1.9935, "step": 49056 }, { "epoch": 9.34, "learning_rate": 1.96930409914204e-07, "loss": 1.1479, "step": 49057 }, { "epoch": 9.34, "learning_rate": 1.9687321258341278e-07, "loss": 1.038, "step": 49058 }, { "epoch": 9.34, "learning_rate": 1.9681601525262155e-07, "loss": 2.0404, "step": 49059 }, { "epoch": 9.34, "learning_rate": 1.9675881792183032e-07, "loss": 1.8472, "step": 49060 }, { "epoch": 9.34, "learning_rate": 1.967016205910391e-07, "loss": 1.3902, "step": 49061 }, { "epoch": 9.35, "learning_rate": 1.9664442326024785e-07, "loss": 0.9107, "step": 49062 }, { "epoch": 9.35, "learning_rate": 1.9658722592945662e-07, "loss": 0.7021, "step": 49063 }, { "epoch": 9.35, "learning_rate": 1.9653002859866539e-07, "loss": 3.633, "step": 49064 }, { "epoch": 9.35, "learning_rate": 1.9647283126787418e-07, "loss": 0.9027, "step": 49065 }, { "epoch": 9.35, "learning_rate": 1.9641563393708295e-07, "loss": 3.0487, "step": 49066 }, { "epoch": 9.35, "learning_rate": 1.9635843660629172e-07, "loss": 0.2686, "step": 49067 }, { "epoch": 9.35, "learning_rate": 1.9630123927550046e-07, "loss": 2.3919, "step": 49068 }, { "epoch": 9.35, "learning_rate": 1.9624404194470925e-07, "loss": 2.3197, "step": 49069 }, { "epoch": 9.35, "learning_rate": 1.9618684461391802e-07, "loss": 1.3366, "step": 49070 }, { "epoch": 9.35, "learning_rate": 1.9612964728312678e-07, "loss": 2.3851, "step": 49071 }, { "epoch": 9.35, "learning_rate": 1.9607244995233558e-07, "loss": 2.0731, "step": 49072 }, { "epoch": 9.35, "learning_rate": 1.9601525262154435e-07, "loss": 3.4493, "step": 49073 }, { "epoch": 9.35, "learning_rate": 1.9595805529075309e-07, "loss": 1.738, "step": 49074 }, { "epoch": 9.35, "learning_rate": 1.9590085795996185e-07, "loss": 1.5378, "step": 49075 }, { "epoch": 9.35, "learning_rate": 1.9584366062917065e-07, "loss": 1.3878, "step": 49076 }, { "epoch": 9.35, "learning_rate": 1.9578646329837941e-07, "loss": 0.4154, "step": 49077 }, { "epoch": 9.35, "learning_rate": 1.9572926596758818e-07, "loss": 1.6469, "step": 49078 }, { "epoch": 9.35, "learning_rate": 1.9567206863679698e-07, "loss": 1.1226, "step": 49079 }, { "epoch": 9.35, "learning_rate": 1.9561487130600572e-07, "loss": 1.4767, "step": 49080 }, { "epoch": 9.35, "learning_rate": 1.9555767397521448e-07, "loss": 0.9614, "step": 49081 }, { "epoch": 9.35, "learning_rate": 1.9550047664442325e-07, "loss": 2.835, "step": 49082 }, { "epoch": 9.35, "learning_rate": 1.9544327931363205e-07, "loss": 0.6454, "step": 49083 }, { "epoch": 9.35, "learning_rate": 1.953860819828408e-07, "loss": 0.7669, "step": 49084 }, { "epoch": 9.35, "learning_rate": 1.953288846520496e-07, "loss": 2.7175, "step": 49085 }, { "epoch": 9.35, "learning_rate": 1.9527168732125832e-07, "loss": 0.9878, "step": 49086 }, { "epoch": 9.35, "learning_rate": 1.9521448999046711e-07, "loss": 1.4482, "step": 49087 }, { "epoch": 9.35, "learning_rate": 1.9515729265967588e-07, "loss": 1.5078, "step": 49088 }, { "epoch": 9.35, "learning_rate": 1.9510009532888465e-07, "loss": 2.4457, "step": 49089 }, { "epoch": 9.35, "learning_rate": 1.9504289799809344e-07, "loss": 1.1394, "step": 49090 }, { "epoch": 9.35, "learning_rate": 1.949857006673022e-07, "loss": 1.1034, "step": 49091 }, { "epoch": 9.35, "learning_rate": 1.9492850333651095e-07, "loss": 1.9472, "step": 49092 }, { "epoch": 9.35, "learning_rate": 1.9487130600571972e-07, "loss": 1.4428, "step": 49093 }, { "epoch": 9.35, "learning_rate": 1.948141086749285e-07, "loss": 0.1793, "step": 49094 }, { "epoch": 9.35, "learning_rate": 1.9475691134413728e-07, "loss": 2.1619, "step": 49095 }, { "epoch": 9.35, "learning_rate": 1.9469971401334607e-07, "loss": 1.3623, "step": 49096 }, { "epoch": 9.35, "learning_rate": 1.9464251668255484e-07, "loss": 1.7004, "step": 49097 }, { "epoch": 9.35, "learning_rate": 1.9458531935176358e-07, "loss": 2.8521, "step": 49098 }, { "epoch": 9.35, "learning_rate": 1.9452812202097235e-07, "loss": 1.9263, "step": 49099 }, { "epoch": 9.35, "learning_rate": 1.9447092469018112e-07, "loss": 1.3256, "step": 49100 }, { "epoch": 9.35, "learning_rate": 1.944137273593899e-07, "loss": 2.5069, "step": 49101 }, { "epoch": 9.35, "learning_rate": 1.9435653002859868e-07, "loss": 1.4305, "step": 49102 }, { "epoch": 9.35, "learning_rate": 1.9429933269780747e-07, "loss": 2.4985, "step": 49103 }, { "epoch": 9.35, "learning_rate": 1.9424213536701619e-07, "loss": 1.4431, "step": 49104 }, { "epoch": 9.35, "learning_rate": 1.9418493803622498e-07, "loss": 1.0598, "step": 49105 }, { "epoch": 9.35, "learning_rate": 1.9412774070543375e-07, "loss": 0.8725, "step": 49106 }, { "epoch": 9.35, "learning_rate": 1.9407054337464254e-07, "loss": 1.6309, "step": 49107 }, { "epoch": 9.35, "learning_rate": 1.940133460438513e-07, "loss": 1.4273, "step": 49108 }, { "epoch": 9.35, "learning_rate": 1.9395614871306005e-07, "loss": 1.6452, "step": 49109 }, { "epoch": 9.35, "learning_rate": 1.9389895138226882e-07, "loss": 0.7666, "step": 49110 }, { "epoch": 9.35, "learning_rate": 1.9384175405147758e-07, "loss": 1.0506, "step": 49111 }, { "epoch": 9.35, "learning_rate": 1.9378455672068638e-07, "loss": 0.2695, "step": 49112 }, { "epoch": 9.35, "learning_rate": 1.9372735938989515e-07, "loss": 0.9007, "step": 49113 }, { "epoch": 9.36, "learning_rate": 1.9367016205910394e-07, "loss": 2.2123, "step": 49114 }, { "epoch": 9.36, "learning_rate": 1.9361296472831265e-07, "loss": 1.2193, "step": 49115 }, { "epoch": 9.36, "learning_rate": 1.9355576739752145e-07, "loss": 2.0046, "step": 49116 }, { "epoch": 9.36, "learning_rate": 1.9349857006673021e-07, "loss": 0.3605, "step": 49117 }, { "epoch": 9.36, "learning_rate": 1.93441372735939e-07, "loss": 2.1118, "step": 49118 }, { "epoch": 9.36, "learning_rate": 1.9338417540514778e-07, "loss": 1.6369, "step": 49119 }, { "epoch": 9.36, "learning_rate": 1.9332697807435654e-07, "loss": 1.4822, "step": 49120 }, { "epoch": 9.36, "learning_rate": 1.9326978074356528e-07, "loss": 2.3641, "step": 49121 }, { "epoch": 9.36, "learning_rate": 1.9321258341277405e-07, "loss": 0.461, "step": 49122 }, { "epoch": 9.36, "learning_rate": 1.9315538608198285e-07, "loss": 2.3294, "step": 49123 }, { "epoch": 9.36, "learning_rate": 1.930981887511916e-07, "loss": 1.569, "step": 49124 }, { "epoch": 9.36, "learning_rate": 1.930409914204004e-07, "loss": 1.0566, "step": 49125 }, { "epoch": 9.36, "learning_rate": 1.9298379408960917e-07, "loss": 2.4805, "step": 49126 }, { "epoch": 9.36, "learning_rate": 1.9292659675881791e-07, "loss": 1.2536, "step": 49127 }, { "epoch": 9.36, "learning_rate": 1.9286939942802668e-07, "loss": 1.8441, "step": 49128 }, { "epoch": 9.36, "learning_rate": 1.9281220209723545e-07, "loss": 1.4768, "step": 49129 }, { "epoch": 9.36, "learning_rate": 1.9275500476644424e-07, "loss": 2.9803, "step": 49130 }, { "epoch": 9.36, "learning_rate": 1.92697807435653e-07, "loss": 1.1578, "step": 49131 }, { "epoch": 9.36, "learning_rate": 1.926406101048618e-07, "loss": 2.0934, "step": 49132 }, { "epoch": 9.36, "learning_rate": 1.9258341277407052e-07, "loss": 1.1558, "step": 49133 }, { "epoch": 9.36, "learning_rate": 1.925262154432793e-07, "loss": 3.1925, "step": 49134 }, { "epoch": 9.36, "learning_rate": 1.9246901811248808e-07, "loss": 1.2384, "step": 49135 }, { "epoch": 9.36, "learning_rate": 1.9241182078169687e-07, "loss": 1.082, "step": 49136 }, { "epoch": 9.36, "learning_rate": 1.9235462345090564e-07, "loss": 1.1089, "step": 49137 }, { "epoch": 9.36, "learning_rate": 1.922974261201144e-07, "loss": 0.804, "step": 49138 }, { "epoch": 9.36, "learning_rate": 1.9224022878932315e-07, "loss": 0.2867, "step": 49139 }, { "epoch": 9.36, "learning_rate": 1.9218303145853192e-07, "loss": 2.4711, "step": 49140 }, { "epoch": 9.36, "learning_rate": 1.921258341277407e-07, "loss": 0.6406, "step": 49141 }, { "epoch": 9.36, "learning_rate": 1.9206863679694948e-07, "loss": 1.5327, "step": 49142 }, { "epoch": 9.36, "learning_rate": 1.9201143946615827e-07, "loss": 2.03, "step": 49143 }, { "epoch": 9.36, "learning_rate": 1.9195424213536704e-07, "loss": 2.0173, "step": 49144 }, { "epoch": 9.36, "learning_rate": 1.9189704480457578e-07, "loss": 1.5485, "step": 49145 }, { "epoch": 9.36, "learning_rate": 1.9183984747378455e-07, "loss": 0.808, "step": 49146 }, { "epoch": 9.36, "learning_rate": 1.9178265014299334e-07, "loss": 2.1312, "step": 49147 }, { "epoch": 9.36, "learning_rate": 1.917254528122021e-07, "loss": 1.1788, "step": 49148 }, { "epoch": 9.36, "learning_rate": 1.9166825548141088e-07, "loss": 1.1592, "step": 49149 }, { "epoch": 9.36, "learning_rate": 1.9161105815061967e-07, "loss": 1.2647, "step": 49150 }, { "epoch": 9.36, "learning_rate": 1.9155386081982838e-07, "loss": 0.8397, "step": 49151 }, { "epoch": 9.36, "learning_rate": 1.9149666348903718e-07, "loss": 2.4523, "step": 49152 }, { "epoch": 9.36, "learning_rate": 1.9143946615824595e-07, "loss": 1.6586, "step": 49153 }, { "epoch": 9.36, "learning_rate": 1.9138226882745474e-07, "loss": 3.0399, "step": 49154 }, { "epoch": 9.36, "learning_rate": 1.913250714966635e-07, "loss": 0.6255, "step": 49155 }, { "epoch": 9.36, "learning_rate": 1.9126787416587227e-07, "loss": 2.8696, "step": 49156 }, { "epoch": 9.36, "learning_rate": 1.9121067683508101e-07, "loss": 3.3121, "step": 49157 }, { "epoch": 9.36, "learning_rate": 1.911534795042898e-07, "loss": 3.5029, "step": 49158 }, { "epoch": 9.36, "learning_rate": 1.9109628217349858e-07, "loss": 2.0801, "step": 49159 }, { "epoch": 9.36, "learning_rate": 1.9103908484270734e-07, "loss": 1.2846, "step": 49160 }, { "epoch": 9.36, "learning_rate": 1.9098188751191614e-07, "loss": 1.4352, "step": 49161 }, { "epoch": 9.36, "learning_rate": 1.909246901811249e-07, "loss": 2.3832, "step": 49162 }, { "epoch": 9.36, "learning_rate": 1.9086749285033365e-07, "loss": 1.0139, "step": 49163 }, { "epoch": 9.36, "learning_rate": 1.908102955195424e-07, "loss": 0.691, "step": 49164 }, { "epoch": 9.36, "learning_rate": 1.907530981887512e-07, "loss": 1.6349, "step": 49165 }, { "epoch": 9.36, "learning_rate": 1.9069590085795997e-07, "loss": 1.4905, "step": 49166 }, { "epoch": 9.37, "learning_rate": 1.9063870352716874e-07, "loss": 1.122, "step": 49167 }, { "epoch": 9.37, "learning_rate": 1.9058150619637753e-07, "loss": 0.8537, "step": 49168 }, { "epoch": 9.37, "learning_rate": 1.9052430886558628e-07, "loss": 2.1016, "step": 49169 }, { "epoch": 9.37, "learning_rate": 1.9046711153479504e-07, "loss": 0.6031, "step": 49170 }, { "epoch": 9.37, "learning_rate": 1.904099142040038e-07, "loss": 2.023, "step": 49171 }, { "epoch": 9.37, "learning_rate": 1.903527168732126e-07, "loss": 1.7371, "step": 49172 }, { "epoch": 9.37, "learning_rate": 1.9029551954242137e-07, "loss": 0.6567, "step": 49173 }, { "epoch": 9.37, "learning_rate": 1.902383222116301e-07, "loss": 1.4111, "step": 49174 }, { "epoch": 9.37, "learning_rate": 1.9018112488083888e-07, "loss": 1.1609, "step": 49175 }, { "epoch": 9.37, "learning_rate": 1.9012392755004767e-07, "loss": 0.2927, "step": 49176 }, { "epoch": 9.37, "learning_rate": 1.9006673021925644e-07, "loss": 1.8066, "step": 49177 }, { "epoch": 9.37, "learning_rate": 1.900095328884652e-07, "loss": 1.7856, "step": 49178 }, { "epoch": 9.37, "learning_rate": 1.89952335557674e-07, "loss": 1.7015, "step": 49179 }, { "epoch": 9.37, "learning_rate": 1.8989513822688274e-07, "loss": 1.6613, "step": 49180 }, { "epoch": 9.37, "learning_rate": 1.898379408960915e-07, "loss": 2.2422, "step": 49181 }, { "epoch": 9.37, "learning_rate": 1.8978074356530028e-07, "loss": 0.4299, "step": 49182 }, { "epoch": 9.37, "learning_rate": 1.8972354623450907e-07, "loss": 1.1274, "step": 49183 }, { "epoch": 9.37, "learning_rate": 1.8966634890371784e-07, "loss": 0.8692, "step": 49184 }, { "epoch": 9.37, "learning_rate": 1.896091515729266e-07, "loss": 1.9892, "step": 49185 }, { "epoch": 9.37, "learning_rate": 1.8955195424213535e-07, "loss": 1.298, "step": 49186 }, { "epoch": 9.37, "learning_rate": 1.8949475691134414e-07, "loss": 0.8806, "step": 49187 }, { "epoch": 9.37, "learning_rate": 1.894375595805529e-07, "loss": 0.5084, "step": 49188 }, { "epoch": 9.37, "learning_rate": 1.8938036224976168e-07, "loss": 0.6653, "step": 49189 }, { "epoch": 9.37, "learning_rate": 1.8932316491897047e-07, "loss": 0.2699, "step": 49190 }, { "epoch": 9.37, "learning_rate": 1.8926596758817924e-07, "loss": 1.9143, "step": 49191 }, { "epoch": 9.37, "learning_rate": 1.8920877025738798e-07, "loss": 1.021, "step": 49192 }, { "epoch": 9.37, "learning_rate": 1.8915157292659675e-07, "loss": 1.1419, "step": 49193 }, { "epoch": 9.37, "learning_rate": 1.8909437559580554e-07, "loss": 0.8825, "step": 49194 }, { "epoch": 9.37, "learning_rate": 1.890371782650143e-07, "loss": 1.1848, "step": 49195 }, { "epoch": 9.37, "learning_rate": 1.8897998093422307e-07, "loss": 1.6212, "step": 49196 }, { "epoch": 9.37, "learning_rate": 1.8892278360343187e-07, "loss": 1.5567, "step": 49197 }, { "epoch": 9.37, "learning_rate": 1.888655862726406e-07, "loss": 0.6332, "step": 49198 }, { "epoch": 9.37, "learning_rate": 1.8880838894184938e-07, "loss": 0.3649, "step": 49199 }, { "epoch": 9.37, "learning_rate": 1.8875119161105814e-07, "loss": 1.5373, "step": 49200 }, { "epoch": 9.37, "learning_rate": 1.8869399428026694e-07, "loss": 0.9495, "step": 49201 }, { "epoch": 9.37, "learning_rate": 1.886367969494757e-07, "loss": 1.9414, "step": 49202 }, { "epoch": 9.37, "learning_rate": 1.8857959961868447e-07, "loss": 1.845, "step": 49203 }, { "epoch": 9.37, "learning_rate": 1.885224022878932e-07, "loss": 0.3086, "step": 49204 }, { "epoch": 9.37, "learning_rate": 1.88465204957102e-07, "loss": 0.776, "step": 49205 }, { "epoch": 9.37, "learning_rate": 1.8840800762631077e-07, "loss": 1.4582, "step": 49206 }, { "epoch": 9.37, "learning_rate": 1.8835081029551954e-07, "loss": 2.1632, "step": 49207 }, { "epoch": 9.37, "learning_rate": 1.8829361296472833e-07, "loss": 2.0442, "step": 49208 }, { "epoch": 9.37, "learning_rate": 1.882364156339371e-07, "loss": 2.2327, "step": 49209 }, { "epoch": 9.37, "learning_rate": 1.8817921830314584e-07, "loss": 1.8867, "step": 49210 }, { "epoch": 9.37, "learning_rate": 1.881220209723546e-07, "loss": 1.4368, "step": 49211 }, { "epoch": 9.37, "learning_rate": 1.880648236415634e-07, "loss": 0.8158, "step": 49212 }, { "epoch": 9.37, "learning_rate": 1.8800762631077217e-07, "loss": 0.834, "step": 49213 }, { "epoch": 9.37, "learning_rate": 1.8795042897998094e-07, "loss": 0.9797, "step": 49214 }, { "epoch": 9.37, "learning_rate": 1.8789323164918973e-07, "loss": 1.9832, "step": 49215 }, { "epoch": 9.37, "learning_rate": 1.8783603431839847e-07, "loss": 1.3065, "step": 49216 }, { "epoch": 9.37, "learning_rate": 1.8777883698760724e-07, "loss": 1.9437, "step": 49217 }, { "epoch": 9.37, "learning_rate": 1.87721639656816e-07, "loss": 1.3285, "step": 49218 }, { "epoch": 9.38, "learning_rate": 1.876644423260248e-07, "loss": 1.7865, "step": 49219 }, { "epoch": 9.38, "learning_rate": 1.8760724499523357e-07, "loss": 0.5906, "step": 49220 }, { "epoch": 9.38, "learning_rate": 1.8755004766444236e-07, "loss": 2.5882, "step": 49221 }, { "epoch": 9.38, "learning_rate": 1.874928503336511e-07, "loss": 0.9919, "step": 49222 }, { "epoch": 9.38, "learning_rate": 1.8743565300285987e-07, "loss": 0.6669, "step": 49223 }, { "epoch": 9.38, "learning_rate": 1.8737845567206864e-07, "loss": 1.3816, "step": 49224 }, { "epoch": 9.38, "learning_rate": 1.873212583412774e-07, "loss": 1.5403, "step": 49225 }, { "epoch": 9.38, "learning_rate": 1.8726406101048617e-07, "loss": 2.15, "step": 49226 }, { "epoch": 9.38, "learning_rate": 1.8720686367969494e-07, "loss": 1.9517, "step": 49227 }, { "epoch": 9.38, "learning_rate": 1.8714966634890373e-07, "loss": 0.343, "step": 49228 }, { "epoch": 9.38, "learning_rate": 1.8709246901811248e-07, "loss": 0.3956, "step": 49229 }, { "epoch": 9.38, "learning_rate": 1.8703527168732127e-07, "loss": 1.7745, "step": 49230 }, { "epoch": 9.38, "learning_rate": 1.8697807435653004e-07, "loss": 0.4638, "step": 49231 }, { "epoch": 9.38, "learning_rate": 1.869208770257388e-07, "loss": 1.3838, "step": 49232 }, { "epoch": 9.38, "learning_rate": 1.8686367969494757e-07, "loss": 3.6253, "step": 49233 }, { "epoch": 9.38, "learning_rate": 1.8680648236415634e-07, "loss": 2.0539, "step": 49234 }, { "epoch": 9.38, "learning_rate": 1.867492850333651e-07, "loss": 3.1804, "step": 49235 }, { "epoch": 9.38, "learning_rate": 1.8669208770257387e-07, "loss": 1.0777, "step": 49236 }, { "epoch": 9.38, "learning_rate": 1.8663489037178267e-07, "loss": 1.9009, "step": 49237 }, { "epoch": 9.38, "learning_rate": 1.865776930409914e-07, "loss": 1.7433, "step": 49238 }, { "epoch": 9.38, "learning_rate": 1.865204957102002e-07, "loss": 1.2947, "step": 49239 }, { "epoch": 9.38, "learning_rate": 1.8646329837940897e-07, "loss": 1.7313, "step": 49240 }, { "epoch": 9.38, "learning_rate": 1.8640610104861774e-07, "loss": 0.8949, "step": 49241 }, { "epoch": 9.38, "learning_rate": 1.863489037178265e-07, "loss": 0.4371, "step": 49242 }, { "epoch": 9.38, "learning_rate": 1.862917063870353e-07, "loss": 1.0877, "step": 49243 }, { "epoch": 9.38, "learning_rate": 1.8623450905624404e-07, "loss": 0.3859, "step": 49244 }, { "epoch": 9.38, "learning_rate": 1.861773117254528e-07, "loss": 1.2522, "step": 49245 }, { "epoch": 9.38, "learning_rate": 1.861201143946616e-07, "loss": 0.3659, "step": 49246 }, { "epoch": 9.38, "learning_rate": 1.8606291706387034e-07, "loss": 2.2673, "step": 49247 }, { "epoch": 9.38, "learning_rate": 1.8600571973307913e-07, "loss": 2.0589, "step": 49248 }, { "epoch": 9.38, "learning_rate": 1.859485224022879e-07, "loss": 2.5419, "step": 49249 }, { "epoch": 9.38, "learning_rate": 1.8589132507149667e-07, "loss": 1.1013, "step": 49250 }, { "epoch": 9.38, "learning_rate": 1.8583412774070544e-07, "loss": 2.0984, "step": 49251 }, { "epoch": 9.38, "learning_rate": 1.8577693040991423e-07, "loss": 1.3998, "step": 49252 }, { "epoch": 9.38, "learning_rate": 1.8571973307912297e-07, "loss": 1.4682, "step": 49253 }, { "epoch": 9.38, "learning_rate": 1.8566253574833176e-07, "loss": 1.1743, "step": 49254 }, { "epoch": 9.38, "learning_rate": 1.8560533841754053e-07, "loss": 1.1174, "step": 49255 }, { "epoch": 9.38, "learning_rate": 1.8554814108674927e-07, "loss": 1.1683, "step": 49256 }, { "epoch": 9.38, "learning_rate": 1.8549094375595807e-07, "loss": 0.6479, "step": 49257 }, { "epoch": 9.38, "learning_rate": 1.8543374642516683e-07, "loss": 1.1906, "step": 49258 }, { "epoch": 9.38, "learning_rate": 1.853765490943756e-07, "loss": 1.0285, "step": 49259 }, { "epoch": 9.38, "learning_rate": 1.8531935176358437e-07, "loss": 1.7211, "step": 49260 }, { "epoch": 9.38, "learning_rate": 1.8526215443279314e-07, "loss": 1.3007, "step": 49261 }, { "epoch": 9.38, "learning_rate": 1.852049571020019e-07, "loss": 0.9876, "step": 49262 }, { "epoch": 9.38, "learning_rate": 1.851477597712107e-07, "loss": 2.282, "step": 49263 }, { "epoch": 9.38, "learning_rate": 1.8509056244041944e-07, "loss": 1.3124, "step": 49264 }, { "epoch": 9.38, "learning_rate": 1.8503336510962823e-07, "loss": 1.1857, "step": 49265 }, { "epoch": 9.38, "learning_rate": 1.84976167778837e-07, "loss": 1.868, "step": 49266 }, { "epoch": 9.38, "learning_rate": 1.8491897044804574e-07, "loss": 1.8771, "step": 49267 }, { "epoch": 9.38, "learning_rate": 1.8486177311725453e-07, "loss": 1.7798, "step": 49268 }, { "epoch": 9.38, "learning_rate": 1.848045757864633e-07, "loss": 0.8334, "step": 49269 }, { "epoch": 9.38, "learning_rate": 1.8474737845567207e-07, "loss": 1.0121, "step": 49270 }, { "epoch": 9.38, "learning_rate": 1.8469018112488084e-07, "loss": 1.811, "step": 49271 }, { "epoch": 9.39, "learning_rate": 1.8463298379408963e-07, "loss": 1.295, "step": 49272 }, { "epoch": 9.39, "learning_rate": 1.8457578646329837e-07, "loss": 2.4542, "step": 49273 }, { "epoch": 9.39, "learning_rate": 1.8451858913250716e-07, "loss": 1.5518, "step": 49274 }, { "epoch": 9.39, "learning_rate": 1.8446139180171593e-07, "loss": 1.6937, "step": 49275 }, { "epoch": 9.39, "learning_rate": 1.8440419447092467e-07, "loss": 1.0655, "step": 49276 }, { "epoch": 9.39, "learning_rate": 1.8434699714013347e-07, "loss": 1.4493, "step": 49277 }, { "epoch": 9.39, "learning_rate": 1.8428979980934223e-07, "loss": 2.6496, "step": 49278 }, { "epoch": 9.39, "learning_rate": 1.84232602478551e-07, "loss": 1.1522, "step": 49279 }, { "epoch": 9.39, "learning_rate": 1.8417540514775977e-07, "loss": 1.5789, "step": 49280 }, { "epoch": 9.39, "learning_rate": 1.8411820781696856e-07, "loss": 0.6724, "step": 49281 }, { "epoch": 9.39, "learning_rate": 1.840610104861773e-07, "loss": 1.0184, "step": 49282 }, { "epoch": 9.39, "learning_rate": 1.840038131553861e-07, "loss": 1.973, "step": 49283 }, { "epoch": 9.39, "learning_rate": 1.8394661582459486e-07, "loss": 1.6797, "step": 49284 }, { "epoch": 9.39, "learning_rate": 1.8388941849380363e-07, "loss": 1.2512, "step": 49285 }, { "epoch": 9.39, "learning_rate": 1.838322211630124e-07, "loss": 1.9174, "step": 49286 }, { "epoch": 9.39, "learning_rate": 1.8377502383222117e-07, "loss": 0.5574, "step": 49287 }, { "epoch": 9.39, "learning_rate": 1.8371782650142993e-07, "loss": 2.0876, "step": 49288 }, { "epoch": 9.39, "learning_rate": 1.836606291706387e-07, "loss": 1.2004, "step": 49289 }, { "epoch": 9.39, "learning_rate": 1.836034318398475e-07, "loss": 1.6546, "step": 49290 }, { "epoch": 9.39, "learning_rate": 1.8354623450905624e-07, "loss": 1.4046, "step": 49291 }, { "epoch": 9.39, "learning_rate": 1.8348903717826503e-07, "loss": 0.842, "step": 49292 }, { "epoch": 9.39, "learning_rate": 1.834318398474738e-07, "loss": 1.8676, "step": 49293 }, { "epoch": 9.39, "learning_rate": 1.8337464251668256e-07, "loss": 0.8167, "step": 49294 }, { "epoch": 9.39, "learning_rate": 1.8331744518589133e-07, "loss": 3.0844, "step": 49295 }, { "epoch": 9.39, "learning_rate": 1.832602478551001e-07, "loss": 1.5226, "step": 49296 }, { "epoch": 9.39, "learning_rate": 1.8320305052430887e-07, "loss": 2.4602, "step": 49297 }, { "epoch": 9.39, "learning_rate": 1.8314585319351763e-07, "loss": 1.5818, "step": 49298 }, { "epoch": 9.39, "learning_rate": 1.8308865586272643e-07, "loss": 1.8433, "step": 49299 }, { "epoch": 9.39, "learning_rate": 1.8303145853193517e-07, "loss": 1.1378, "step": 49300 }, { "epoch": 9.39, "learning_rate": 1.8297426120114396e-07, "loss": 1.6975, "step": 49301 }, { "epoch": 9.39, "learning_rate": 1.8291706387035273e-07, "loss": 1.3162, "step": 49302 }, { "epoch": 9.39, "learning_rate": 1.828598665395615e-07, "loss": 1.5648, "step": 49303 }, { "epoch": 9.39, "learning_rate": 1.8280266920877026e-07, "loss": 1.1652, "step": 49304 }, { "epoch": 9.39, "learning_rate": 1.8274547187797903e-07, "loss": 1.5764, "step": 49305 }, { "epoch": 9.39, "learning_rate": 1.826882745471878e-07, "loss": 2.4025, "step": 49306 }, { "epoch": 9.39, "learning_rate": 1.8263107721639657e-07, "loss": 1.1067, "step": 49307 }, { "epoch": 9.39, "learning_rate": 1.8257387988560536e-07, "loss": 0.3754, "step": 49308 }, { "epoch": 9.39, "learning_rate": 1.825166825548141e-07, "loss": 2.9463, "step": 49309 }, { "epoch": 9.39, "learning_rate": 1.824594852240229e-07, "loss": 0.3458, "step": 49310 }, { "epoch": 9.39, "learning_rate": 1.8240228789323166e-07, "loss": 0.5578, "step": 49311 }, { "epoch": 9.39, "learning_rate": 1.8234509056244043e-07, "loss": 1.7718, "step": 49312 }, { "epoch": 9.39, "learning_rate": 1.822878932316492e-07, "loss": 1.227, "step": 49313 }, { "epoch": 9.39, "learning_rate": 1.8223069590085796e-07, "loss": 2.7603, "step": 49314 }, { "epoch": 9.39, "learning_rate": 1.8217349857006673e-07, "loss": 1.7741, "step": 49315 }, { "epoch": 9.39, "learning_rate": 1.821163012392755e-07, "loss": 1.8355, "step": 49316 }, { "epoch": 9.39, "learning_rate": 1.820591039084843e-07, "loss": 1.2568, "step": 49317 }, { "epoch": 9.39, "learning_rate": 1.8200190657769303e-07, "loss": 1.1316, "step": 49318 }, { "epoch": 9.39, "learning_rate": 1.8194470924690183e-07, "loss": 2.6941, "step": 49319 }, { "epoch": 9.39, "learning_rate": 1.818875119161106e-07, "loss": 0.2551, "step": 49320 }, { "epoch": 9.39, "learning_rate": 1.8183031458531936e-07, "loss": 0.9749, "step": 49321 }, { "epoch": 9.39, "learning_rate": 1.8177311725452813e-07, "loss": 0.8, "step": 49322 }, { "epoch": 9.39, "learning_rate": 1.817159199237369e-07, "loss": 1.2775, "step": 49323 }, { "epoch": 9.4, "learning_rate": 1.8165872259294566e-07, "loss": 0.3177, "step": 49324 }, { "epoch": 9.4, "learning_rate": 1.8160152526215443e-07, "loss": 2.6396, "step": 49325 }, { "epoch": 9.4, "learning_rate": 1.815443279313632e-07, "loss": 1.3963, "step": 49326 }, { "epoch": 9.4, "learning_rate": 1.8148713060057197e-07, "loss": 1.3118, "step": 49327 }, { "epoch": 9.4, "learning_rate": 1.8142993326978076e-07, "loss": 2.4801, "step": 49328 }, { "epoch": 9.4, "learning_rate": 1.813727359389895e-07, "loss": 2.8372, "step": 49329 }, { "epoch": 9.4, "learning_rate": 1.813155386081983e-07, "loss": 2.5642, "step": 49330 }, { "epoch": 9.4, "learning_rate": 1.8125834127740706e-07, "loss": 1.1269, "step": 49331 }, { "epoch": 9.4, "learning_rate": 1.8120114394661583e-07, "loss": 0.3942, "step": 49332 }, { "epoch": 9.4, "learning_rate": 1.811439466158246e-07, "loss": 0.9004, "step": 49333 }, { "epoch": 9.4, "learning_rate": 1.8108674928503336e-07, "loss": 0.5984, "step": 49334 }, { "epoch": 9.4, "learning_rate": 1.8102955195424213e-07, "loss": 0.7602, "step": 49335 }, { "epoch": 9.4, "learning_rate": 1.809723546234509e-07, "loss": 1.6558, "step": 49336 }, { "epoch": 9.4, "learning_rate": 1.809151572926597e-07, "loss": 1.3, "step": 49337 }, { "epoch": 9.4, "learning_rate": 1.8085795996186843e-07, "loss": 2.2074, "step": 49338 }, { "epoch": 9.4, "learning_rate": 1.8080076263107723e-07, "loss": 1.0107, "step": 49339 }, { "epoch": 9.4, "learning_rate": 1.80743565300286e-07, "loss": 0.9796, "step": 49340 }, { "epoch": 9.4, "learning_rate": 1.8068636796949476e-07, "loss": 1.115, "step": 49341 }, { "epoch": 9.4, "learning_rate": 1.8062917063870353e-07, "loss": 1.3577, "step": 49342 }, { "epoch": 9.4, "learning_rate": 1.805719733079123e-07, "loss": 0.9937, "step": 49343 }, { "epoch": 9.4, "learning_rate": 1.8051477597712106e-07, "loss": 1.6128, "step": 49344 }, { "epoch": 9.4, "learning_rate": 1.8045757864632983e-07, "loss": 1.4765, "step": 49345 }, { "epoch": 9.4, "learning_rate": 1.8040038131553863e-07, "loss": 0.9798, "step": 49346 }, { "epoch": 9.4, "learning_rate": 1.8034318398474737e-07, "loss": 0.8579, "step": 49347 }, { "epoch": 9.4, "learning_rate": 1.8028598665395616e-07, "loss": 1.8167, "step": 49348 }, { "epoch": 9.4, "learning_rate": 1.8022878932316493e-07, "loss": 1.0349, "step": 49349 }, { "epoch": 9.4, "learning_rate": 1.801715919923737e-07, "loss": 0.7132, "step": 49350 }, { "epoch": 9.4, "learning_rate": 1.8011439466158246e-07, "loss": 2.0311, "step": 49351 }, { "epoch": 9.4, "learning_rate": 1.8005719733079123e-07, "loss": 2.0253, "step": 49352 }, { "epoch": 9.4, "learning_rate": 1.8e-07, "loss": 1.0826, "step": 49353 }, { "epoch": 9.4, "learning_rate": 1.7994280266920876e-07, "loss": 1.1668, "step": 49354 }, { "epoch": 9.4, "learning_rate": 1.7988560533841756e-07, "loss": 0.5225, "step": 49355 }, { "epoch": 9.4, "learning_rate": 1.798284080076263e-07, "loss": 0.3702, "step": 49356 }, { "epoch": 9.4, "learning_rate": 1.797712106768351e-07, "loss": 1.9741, "step": 49357 }, { "epoch": 9.4, "learning_rate": 1.7971401334604386e-07, "loss": 1.9573, "step": 49358 }, { "epoch": 9.4, "learning_rate": 1.7965681601525263e-07, "loss": 2.2184, "step": 49359 }, { "epoch": 9.4, "learning_rate": 1.795996186844614e-07, "loss": 1.3147, "step": 49360 }, { "epoch": 9.4, "learning_rate": 1.7954242135367016e-07, "loss": 1.6745, "step": 49361 }, { "epoch": 9.4, "learning_rate": 1.7948522402287893e-07, "loss": 1.3508, "step": 49362 }, { "epoch": 9.4, "learning_rate": 1.794280266920877e-07, "loss": 0.5335, "step": 49363 }, { "epoch": 9.4, "learning_rate": 1.793708293612965e-07, "loss": 1.2769, "step": 49364 }, { "epoch": 9.4, "learning_rate": 1.7931363203050523e-07, "loss": 2.5561, "step": 49365 }, { "epoch": 9.4, "learning_rate": 1.7925643469971403e-07, "loss": 2.442, "step": 49366 }, { "epoch": 9.4, "learning_rate": 1.791992373689228e-07, "loss": 1.1012, "step": 49367 }, { "epoch": 9.4, "learning_rate": 1.7914204003813156e-07, "loss": 1.396, "step": 49368 }, { "epoch": 9.4, "learning_rate": 1.7908484270734033e-07, "loss": 1.65, "step": 49369 }, { "epoch": 9.4, "learning_rate": 1.790276453765491e-07, "loss": 0.6973, "step": 49370 }, { "epoch": 9.4, "learning_rate": 1.7897044804575786e-07, "loss": 2.5578, "step": 49371 }, { "epoch": 9.4, "learning_rate": 1.7891325071496663e-07, "loss": 3.2884, "step": 49372 }, { "epoch": 9.4, "learning_rate": 1.7885605338417542e-07, "loss": 0.9736, "step": 49373 }, { "epoch": 9.4, "learning_rate": 1.7879885605338416e-07, "loss": 1.7127, "step": 49374 }, { "epoch": 9.4, "learning_rate": 1.7874165872259296e-07, "loss": 3.0969, "step": 49375 }, { "epoch": 9.4, "learning_rate": 1.7868446139180173e-07, "loss": 0.588, "step": 49376 }, { "epoch": 9.41, "learning_rate": 1.786272640610105e-07, "loss": 1.4091, "step": 49377 }, { "epoch": 9.41, "learning_rate": 1.7857006673021926e-07, "loss": 2.233, "step": 49378 }, { "epoch": 9.41, "learning_rate": 1.7851286939942805e-07, "loss": 2.7259, "step": 49379 }, { "epoch": 9.41, "learning_rate": 1.784556720686368e-07, "loss": 1.5294, "step": 49380 }, { "epoch": 9.41, "learning_rate": 1.7839847473784556e-07, "loss": 0.7567, "step": 49381 }, { "epoch": 9.41, "learning_rate": 1.7834127740705433e-07, "loss": 1.1258, "step": 49382 }, { "epoch": 9.41, "learning_rate": 1.782840800762631e-07, "loss": 2.4753, "step": 49383 }, { "epoch": 9.41, "learning_rate": 1.782268827454719e-07, "loss": 0.7557, "step": 49384 }, { "epoch": 9.41, "learning_rate": 1.7816968541468063e-07, "loss": 1.592, "step": 49385 }, { "epoch": 9.41, "learning_rate": 1.7811248808388943e-07, "loss": 1.5488, "step": 49386 }, { "epoch": 9.41, "learning_rate": 1.780552907530982e-07, "loss": 1.1143, "step": 49387 }, { "epoch": 9.41, "learning_rate": 1.7799809342230696e-07, "loss": 1.286, "step": 49388 }, { "epoch": 9.41, "learning_rate": 1.7794089609151573e-07, "loss": 1.6589, "step": 49389 }, { "epoch": 9.41, "learning_rate": 1.7788369876072452e-07, "loss": 2.2292, "step": 49390 }, { "epoch": 9.41, "learning_rate": 1.7782650142993326e-07, "loss": 1.3599, "step": 49391 }, { "epoch": 9.41, "learning_rate": 1.7776930409914203e-07, "loss": 2.4698, "step": 49392 }, { "epoch": 9.41, "learning_rate": 1.7771210676835082e-07, "loss": 1.0089, "step": 49393 }, { "epoch": 9.41, "learning_rate": 1.7765490943755956e-07, "loss": 1.3421, "step": 49394 }, { "epoch": 9.41, "learning_rate": 1.7759771210676836e-07, "loss": 2.2132, "step": 49395 }, { "epoch": 9.41, "learning_rate": 1.7754051477597713e-07, "loss": 0.8467, "step": 49396 }, { "epoch": 9.41, "learning_rate": 1.774833174451859e-07, "loss": 2.5186, "step": 49397 }, { "epoch": 9.41, "learning_rate": 1.7742612011439466e-07, "loss": 1.1331, "step": 49398 }, { "epoch": 9.41, "learning_rate": 1.7736892278360345e-07, "loss": 1.3836, "step": 49399 }, { "epoch": 9.41, "learning_rate": 1.773117254528122e-07, "loss": 1.8349, "step": 49400 }, { "epoch": 9.41, "learning_rate": 1.77254528122021e-07, "loss": 1.1721, "step": 49401 }, { "epoch": 9.41, "learning_rate": 1.7719733079122976e-07, "loss": 0.9824, "step": 49402 }, { "epoch": 9.41, "learning_rate": 1.771401334604385e-07, "loss": 0.5338, "step": 49403 }, { "epoch": 9.41, "learning_rate": 1.770829361296473e-07, "loss": 1.2092, "step": 49404 }, { "epoch": 9.41, "learning_rate": 1.7702573879885606e-07, "loss": 1.1258, "step": 49405 }, { "epoch": 9.41, "learning_rate": 1.7696854146806483e-07, "loss": 1.7063, "step": 49406 }, { "epoch": 9.41, "learning_rate": 1.769113441372736e-07, "loss": 1.5892, "step": 49407 }, { "epoch": 9.41, "learning_rate": 1.7685414680648239e-07, "loss": 1.7637, "step": 49408 }, { "epoch": 9.41, "learning_rate": 1.7679694947569113e-07, "loss": 1.0967, "step": 49409 }, { "epoch": 9.41, "learning_rate": 1.7673975214489992e-07, "loss": 1.6984, "step": 49410 }, { "epoch": 9.41, "learning_rate": 1.766825548141087e-07, "loss": 1.6682, "step": 49411 }, { "epoch": 9.41, "learning_rate": 1.7662535748331746e-07, "loss": 1.2631, "step": 49412 }, { "epoch": 9.41, "learning_rate": 1.7656816015252622e-07, "loss": 1.33, "step": 49413 }, { "epoch": 9.41, "learning_rate": 1.76510962821735e-07, "loss": 1.4664, "step": 49414 }, { "epoch": 9.41, "learning_rate": 1.7645376549094376e-07, "loss": 0.6962, "step": 49415 }, { "epoch": 9.41, "learning_rate": 1.7639656816015253e-07, "loss": 0.5403, "step": 49416 }, { "epoch": 9.41, "learning_rate": 1.7633937082936132e-07, "loss": 0.8718, "step": 49417 }, { "epoch": 9.41, "learning_rate": 1.7628217349857006e-07, "loss": 1.1526, "step": 49418 }, { "epoch": 9.41, "learning_rate": 1.7622497616777885e-07, "loss": 2.7268, "step": 49419 }, { "epoch": 9.41, "learning_rate": 1.7616777883698762e-07, "loss": 0.5927, "step": 49420 }, { "epoch": 9.41, "learning_rate": 1.761105815061964e-07, "loss": 0.8402, "step": 49421 }, { "epoch": 9.41, "learning_rate": 1.7605338417540516e-07, "loss": 0.8862, "step": 49422 }, { "epoch": 9.41, "learning_rate": 1.7599618684461392e-07, "loss": 2.2881, "step": 49423 }, { "epoch": 9.41, "learning_rate": 1.759389895138227e-07, "loss": 1.5209, "step": 49424 }, { "epoch": 9.41, "learning_rate": 1.7588179218303146e-07, "loss": 1.6642, "step": 49425 }, { "epoch": 9.41, "learning_rate": 1.7582459485224025e-07, "loss": 1.4913, "step": 49426 }, { "epoch": 9.41, "learning_rate": 1.75767397521449e-07, "loss": 0.75, "step": 49427 }, { "epoch": 9.41, "learning_rate": 1.7571020019065779e-07, "loss": 1.4519, "step": 49428 }, { "epoch": 9.42, "learning_rate": 1.7565300285986655e-07, "loss": 3.5461, "step": 49429 }, { "epoch": 9.42, "learning_rate": 1.7559580552907532e-07, "loss": 1.6646, "step": 49430 }, { "epoch": 9.42, "learning_rate": 1.755386081982841e-07, "loss": 0.8233, "step": 49431 }, { "epoch": 9.42, "learning_rate": 1.7548141086749286e-07, "loss": 2.8645, "step": 49432 }, { "epoch": 9.42, "learning_rate": 1.7542421353670162e-07, "loss": 1.4078, "step": 49433 }, { "epoch": 9.42, "learning_rate": 1.753670162059104e-07, "loss": 1.4445, "step": 49434 }, { "epoch": 9.42, "learning_rate": 1.7530981887511918e-07, "loss": 0.8202, "step": 49435 }, { "epoch": 9.42, "learning_rate": 1.7525262154432793e-07, "loss": 0.9898, "step": 49436 }, { "epoch": 9.42, "learning_rate": 1.7519542421353672e-07, "loss": 1.6421, "step": 49437 }, { "epoch": 9.42, "learning_rate": 1.7513822688274549e-07, "loss": 1.2902, "step": 49438 }, { "epoch": 9.42, "learning_rate": 1.7508102955195425e-07, "loss": 1.1235, "step": 49439 }, { "epoch": 9.42, "learning_rate": 1.7502383222116302e-07, "loss": 1.1563, "step": 49440 }, { "epoch": 9.42, "learning_rate": 1.749666348903718e-07, "loss": 0.9595, "step": 49441 }, { "epoch": 9.42, "learning_rate": 1.7490943755958056e-07, "loss": 0.6343, "step": 49442 }, { "epoch": 9.42, "learning_rate": 1.7485224022878932e-07, "loss": 1.4019, "step": 49443 }, { "epoch": 9.42, "learning_rate": 1.747950428979981e-07, "loss": 2.2893, "step": 49444 }, { "epoch": 9.42, "learning_rate": 1.7473784556720686e-07, "loss": 0.8338, "step": 49445 }, { "epoch": 9.42, "learning_rate": 1.7468064823641565e-07, "loss": 2.368, "step": 49446 }, { "epoch": 9.42, "learning_rate": 1.746234509056244e-07, "loss": 1.3543, "step": 49447 }, { "epoch": 9.42, "learning_rate": 1.7456625357483319e-07, "loss": 1.8453, "step": 49448 }, { "epoch": 9.42, "learning_rate": 1.7450905624404195e-07, "loss": 0.2163, "step": 49449 }, { "epoch": 9.42, "learning_rate": 1.7445185891325072e-07, "loss": 0.5199, "step": 49450 }, { "epoch": 9.42, "learning_rate": 1.743946615824595e-07, "loss": 1.405, "step": 49451 }, { "epoch": 9.42, "learning_rate": 1.7433746425166826e-07, "loss": 1.036, "step": 49452 }, { "epoch": 9.42, "learning_rate": 1.7428026692087702e-07, "loss": 1.3702, "step": 49453 }, { "epoch": 9.42, "learning_rate": 1.742230695900858e-07, "loss": 1.5279, "step": 49454 }, { "epoch": 9.42, "learning_rate": 1.7416587225929458e-07, "loss": 2.2035, "step": 49455 }, { "epoch": 9.42, "learning_rate": 1.7410867492850333e-07, "loss": 2.0364, "step": 49456 }, { "epoch": 9.42, "learning_rate": 1.7405147759771212e-07, "loss": 0.8996, "step": 49457 }, { "epoch": 9.42, "learning_rate": 1.7399428026692089e-07, "loss": 1.3268, "step": 49458 }, { "epoch": 9.42, "learning_rate": 1.7393708293612965e-07, "loss": 0.5439, "step": 49459 }, { "epoch": 9.42, "learning_rate": 1.7387988560533842e-07, "loss": 1.2764, "step": 49460 }, { "epoch": 9.42, "learning_rate": 1.738226882745472e-07, "loss": 2.9415, "step": 49461 }, { "epoch": 9.42, "learning_rate": 1.7376549094375596e-07, "loss": 1.1802, "step": 49462 }, { "epoch": 9.42, "learning_rate": 1.7370829361296472e-07, "loss": 1.7033, "step": 49463 }, { "epoch": 9.42, "learning_rate": 1.7365109628217352e-07, "loss": 1.4112, "step": 49464 }, { "epoch": 9.42, "learning_rate": 1.7359389895138226e-07, "loss": 0.6088, "step": 49465 }, { "epoch": 9.42, "learning_rate": 1.7353670162059105e-07, "loss": 0.9099, "step": 49466 }, { "epoch": 9.42, "learning_rate": 1.7347950428979982e-07, "loss": 0.484, "step": 49467 }, { "epoch": 9.42, "learning_rate": 1.7342230695900859e-07, "loss": 0.6276, "step": 49468 }, { "epoch": 9.42, "learning_rate": 1.7336510962821735e-07, "loss": 2.3283, "step": 49469 }, { "epoch": 9.42, "learning_rate": 1.7330791229742612e-07, "loss": 1.4611, "step": 49470 }, { "epoch": 9.42, "learning_rate": 1.732507149666349e-07, "loss": 2.7165, "step": 49471 }, { "epoch": 9.42, "learning_rate": 1.7319351763584366e-07, "loss": 1.0125, "step": 49472 }, { "epoch": 9.42, "learning_rate": 1.7313632030505245e-07, "loss": 1.8759, "step": 49473 }, { "epoch": 9.42, "learning_rate": 1.730791229742612e-07, "loss": 0.9661, "step": 49474 }, { "epoch": 9.42, "learning_rate": 1.7302192564346998e-07, "loss": 2.4324, "step": 49475 }, { "epoch": 9.42, "learning_rate": 1.7296472831267875e-07, "loss": 1.7659, "step": 49476 }, { "epoch": 9.42, "learning_rate": 1.7290753098188752e-07, "loss": 1.367, "step": 49477 }, { "epoch": 9.42, "learning_rate": 1.7285033365109629e-07, "loss": 0.8231, "step": 49478 }, { "epoch": 9.42, "learning_rate": 1.7279313632030505e-07, "loss": 1.7996, "step": 49479 }, { "epoch": 9.42, "learning_rate": 1.7273593898951382e-07, "loss": 2.0301, "step": 49480 }, { "epoch": 9.42, "learning_rate": 1.726787416587226e-07, "loss": 1.5141, "step": 49481 }, { "epoch": 9.43, "learning_rate": 1.7262154432793138e-07, "loss": 1.3276, "step": 49482 }, { "epoch": 9.43, "learning_rate": 1.7256434699714012e-07, "loss": 1.9489, "step": 49483 }, { "epoch": 9.43, "learning_rate": 1.7250714966634892e-07, "loss": 0.4339, "step": 49484 }, { "epoch": 9.43, "learning_rate": 1.7244995233555768e-07, "loss": 0.9236, "step": 49485 }, { "epoch": 9.43, "learning_rate": 1.7239275500476645e-07, "loss": 2.6233, "step": 49486 }, { "epoch": 9.43, "learning_rate": 1.7233555767397522e-07, "loss": 2.1542, "step": 49487 }, { "epoch": 9.43, "learning_rate": 1.7227836034318399e-07, "loss": 2.0037, "step": 49488 }, { "epoch": 9.43, "learning_rate": 1.7222116301239275e-07, "loss": 0.7454, "step": 49489 }, { "epoch": 9.43, "learning_rate": 1.7216396568160152e-07, "loss": 0.8483, "step": 49490 }, { "epoch": 9.43, "learning_rate": 1.7210676835081031e-07, "loss": 2.4014, "step": 49491 }, { "epoch": 9.43, "learning_rate": 1.7204957102001906e-07, "loss": 0.9873, "step": 49492 }, { "epoch": 9.43, "learning_rate": 1.7199237368922785e-07, "loss": 1.772, "step": 49493 }, { "epoch": 9.43, "learning_rate": 1.7193517635843662e-07, "loss": 0.8911, "step": 49494 }, { "epoch": 9.43, "learning_rate": 1.7187797902764538e-07, "loss": 0.464, "step": 49495 }, { "epoch": 9.43, "learning_rate": 1.7182078169685415e-07, "loss": 2.506, "step": 49496 }, { "epoch": 9.43, "learning_rate": 1.7176358436606292e-07, "loss": 0.4345, "step": 49497 }, { "epoch": 9.43, "learning_rate": 1.7170638703527169e-07, "loss": 1.1734, "step": 49498 }, { "epoch": 9.43, "learning_rate": 1.7164918970448045e-07, "loss": 0.8927, "step": 49499 }, { "epoch": 9.43, "learning_rate": 1.7159199237368925e-07, "loss": 0.9744, "step": 49500 }, { "epoch": 9.43, "learning_rate": 1.71534795042898e-07, "loss": 0.6193, "step": 49501 }, { "epoch": 9.43, "learning_rate": 1.7147759771210678e-07, "loss": 1.2461, "step": 49502 }, { "epoch": 9.43, "learning_rate": 1.7142040038131552e-07, "loss": 1.261, "step": 49503 }, { "epoch": 9.43, "learning_rate": 1.7136320305052432e-07, "loss": 0.9089, "step": 49504 }, { "epoch": 9.43, "learning_rate": 1.7130600571973308e-07, "loss": 1.3561, "step": 49505 }, { "epoch": 9.43, "learning_rate": 1.7124880838894185e-07, "loss": 1.8023, "step": 49506 }, { "epoch": 9.43, "learning_rate": 1.7119161105815062e-07, "loss": 0.9052, "step": 49507 }, { "epoch": 9.43, "learning_rate": 1.7113441372735939e-07, "loss": 2.0467, "step": 49508 }, { "epoch": 9.43, "learning_rate": 1.7107721639656815e-07, "loss": 1.3166, "step": 49509 }, { "epoch": 9.43, "learning_rate": 1.7102001906577692e-07, "loss": 0.611, "step": 49510 }, { "epoch": 9.43, "learning_rate": 1.7096282173498571e-07, "loss": 0.6588, "step": 49511 }, { "epoch": 9.43, "learning_rate": 1.7090562440419446e-07, "loss": 1.4198, "step": 49512 }, { "epoch": 9.43, "learning_rate": 1.7084842707340325e-07, "loss": 1.5498, "step": 49513 }, { "epoch": 9.43, "learning_rate": 1.7079122974261202e-07, "loss": 1.2001, "step": 49514 }, { "epoch": 9.43, "learning_rate": 1.7073403241182078e-07, "loss": 1.8105, "step": 49515 }, { "epoch": 9.43, "learning_rate": 1.7067683508102955e-07, "loss": 1.0937, "step": 49516 }, { "epoch": 9.43, "learning_rate": 1.7061963775023832e-07, "loss": 1.4794, "step": 49517 }, { "epoch": 9.43, "learning_rate": 1.7056244041944709e-07, "loss": 2.2753, "step": 49518 }, { "epoch": 9.43, "learning_rate": 1.7050524308865585e-07, "loss": 2.0054, "step": 49519 }, { "epoch": 9.43, "learning_rate": 1.7044804575786465e-07, "loss": 2.0557, "step": 49520 }, { "epoch": 9.43, "learning_rate": 1.703908484270734e-07, "loss": 0.7859, "step": 49521 }, { "epoch": 9.43, "learning_rate": 1.7033365109628218e-07, "loss": 1.8637, "step": 49522 }, { "epoch": 9.43, "learning_rate": 1.7027645376549095e-07, "loss": 0.8129, "step": 49523 }, { "epoch": 9.43, "learning_rate": 1.7021925643469972e-07, "loss": 0.5817, "step": 49524 }, { "epoch": 9.43, "learning_rate": 1.7016205910390848e-07, "loss": 2.7087, "step": 49525 }, { "epoch": 9.43, "learning_rate": 1.7010486177311728e-07, "loss": 0.4808, "step": 49526 }, { "epoch": 9.43, "learning_rate": 1.7004766444232602e-07, "loss": 1.0902, "step": 49527 }, { "epoch": 9.43, "learning_rate": 1.6999046711153479e-07, "loss": 1.8039, "step": 49528 }, { "epoch": 9.43, "learning_rate": 1.6993326978074358e-07, "loss": 2.2721, "step": 49529 }, { "epoch": 9.43, "learning_rate": 1.6987607244995232e-07, "loss": 1.9468, "step": 49530 }, { "epoch": 9.43, "learning_rate": 1.6981887511916111e-07, "loss": 1.3372, "step": 49531 }, { "epoch": 9.43, "learning_rate": 1.6976167778836988e-07, "loss": 1.2643, "step": 49532 }, { "epoch": 9.43, "learning_rate": 1.6970448045757865e-07, "loss": 1.6747, "step": 49533 }, { "epoch": 9.44, "learning_rate": 1.6964728312678742e-07, "loss": 1.0026, "step": 49534 }, { "epoch": 9.44, "learning_rate": 1.695900857959962e-07, "loss": 0.657, "step": 49535 }, { "epoch": 9.44, "learning_rate": 1.6953288846520495e-07, "loss": 1.4022, "step": 49536 }, { "epoch": 9.44, "learning_rate": 1.6947569113441375e-07, "loss": 2.1956, "step": 49537 }, { "epoch": 9.44, "learning_rate": 1.694184938036225e-07, "loss": 1.2629, "step": 49538 }, { "epoch": 9.44, "learning_rate": 1.6936129647283125e-07, "loss": 2.5017, "step": 49539 }, { "epoch": 9.44, "learning_rate": 1.6930409914204005e-07, "loss": 0.5097, "step": 49540 }, { "epoch": 9.44, "learning_rate": 1.6924690181124881e-07, "loss": 2.2438, "step": 49541 }, { "epoch": 9.44, "learning_rate": 1.6918970448045758e-07, "loss": 0.8651, "step": 49542 }, { "epoch": 9.44, "learning_rate": 1.6913250714966635e-07, "loss": 3.0355, "step": 49543 }, { "epoch": 9.44, "learning_rate": 1.6907530981887514e-07, "loss": 1.5972, "step": 49544 }, { "epoch": 9.44, "learning_rate": 1.6901811248808388e-07, "loss": 0.862, "step": 49545 }, { "epoch": 9.44, "learning_rate": 1.6896091515729268e-07, "loss": 1.2433, "step": 49546 }, { "epoch": 9.44, "learning_rate": 1.6890371782650145e-07, "loss": 0.7846, "step": 49547 }, { "epoch": 9.44, "learning_rate": 1.688465204957102e-07, "loss": 0.7337, "step": 49548 }, { "epoch": 9.44, "learning_rate": 1.6878932316491898e-07, "loss": 1.6815, "step": 49549 }, { "epoch": 9.44, "learning_rate": 1.6873212583412775e-07, "loss": 1.6653, "step": 49550 }, { "epoch": 9.44, "learning_rate": 1.6867492850333651e-07, "loss": 1.2753, "step": 49551 }, { "epoch": 9.44, "learning_rate": 1.6861773117254528e-07, "loss": 2.0453, "step": 49552 }, { "epoch": 9.44, "learning_rate": 1.6856053384175408e-07, "loss": 1.5567, "step": 49553 }, { "epoch": 9.44, "learning_rate": 1.6850333651096282e-07, "loss": 0.994, "step": 49554 }, { "epoch": 9.44, "learning_rate": 1.684461391801716e-07, "loss": 0.3918, "step": 49555 }, { "epoch": 9.44, "learning_rate": 1.6838894184938038e-07, "loss": 1.7606, "step": 49556 }, { "epoch": 9.44, "learning_rate": 1.6833174451858914e-07, "loss": 2.2521, "step": 49557 }, { "epoch": 9.44, "learning_rate": 1.682745471877979e-07, "loss": 1.7361, "step": 49558 }, { "epoch": 9.44, "learning_rate": 1.6821734985700668e-07, "loss": 1.1535, "step": 49559 }, { "epoch": 9.44, "learning_rate": 1.6816015252621545e-07, "loss": 0.5878, "step": 49560 }, { "epoch": 9.44, "learning_rate": 1.6810295519542421e-07, "loss": 1.6676, "step": 49561 }, { "epoch": 9.44, "learning_rate": 1.68045757864633e-07, "loss": 2.1125, "step": 49562 }, { "epoch": 9.44, "learning_rate": 1.6798856053384175e-07, "loss": 1.8667, "step": 49563 }, { "epoch": 9.44, "learning_rate": 1.6793136320305054e-07, "loss": 1.7602, "step": 49564 }, { "epoch": 9.44, "learning_rate": 1.6787416587225928e-07, "loss": 0.3813, "step": 49565 }, { "epoch": 9.44, "learning_rate": 1.6781696854146808e-07, "loss": 0.5682, "step": 49566 }, { "epoch": 9.44, "learning_rate": 1.6775977121067684e-07, "loss": 2.006, "step": 49567 }, { "epoch": 9.44, "learning_rate": 1.677025738798856e-07, "loss": 1.5131, "step": 49568 }, { "epoch": 9.44, "learning_rate": 1.6764537654909438e-07, "loss": 2.2303, "step": 49569 }, { "epoch": 9.44, "learning_rate": 1.6758817921830315e-07, "loss": 1.8928, "step": 49570 }, { "epoch": 9.44, "learning_rate": 1.6753098188751191e-07, "loss": 2.0188, "step": 49571 }, { "epoch": 9.44, "learning_rate": 1.6747378455672068e-07, "loss": 0.5538, "step": 49572 }, { "epoch": 9.44, "learning_rate": 1.6741658722592948e-07, "loss": 0.5684, "step": 49573 }, { "epoch": 9.44, "learning_rate": 1.6735938989513822e-07, "loss": 0.6727, "step": 49574 }, { "epoch": 9.44, "learning_rate": 1.67302192564347e-07, "loss": 2.9225, "step": 49575 }, { "epoch": 9.44, "learning_rate": 1.6724499523355578e-07, "loss": 1.8095, "step": 49576 }, { "epoch": 9.44, "learning_rate": 1.6718779790276454e-07, "loss": 2.2783, "step": 49577 }, { "epoch": 9.44, "learning_rate": 1.671306005719733e-07, "loss": 0.2936, "step": 49578 }, { "epoch": 9.44, "learning_rate": 1.6707340324118208e-07, "loss": 2.8417, "step": 49579 }, { "epoch": 9.44, "learning_rate": 1.6701620591039085e-07, "loss": 1.5451, "step": 49580 }, { "epoch": 9.44, "learning_rate": 1.6695900857959961e-07, "loss": 1.7171, "step": 49581 }, { "epoch": 9.44, "learning_rate": 1.669018112488084e-07, "loss": 1.8051, "step": 49582 }, { "epoch": 9.44, "learning_rate": 1.6684461391801715e-07, "loss": 1.1529, "step": 49583 }, { "epoch": 9.44, "learning_rate": 1.6678741658722594e-07, "loss": 1.623, "step": 49584 }, { "epoch": 9.44, "learning_rate": 1.667302192564347e-07, "loss": 1.2544, "step": 49585 }, { "epoch": 9.44, "learning_rate": 1.6667302192564348e-07, "loss": 1.7496, "step": 49586 }, { "epoch": 9.45, "learning_rate": 1.6661582459485224e-07, "loss": 2.044, "step": 49587 }, { "epoch": 9.45, "learning_rate": 1.66558627264061e-07, "loss": 1.7192, "step": 49588 }, { "epoch": 9.45, "learning_rate": 1.6650142993326978e-07, "loss": 1.5147, "step": 49589 }, { "epoch": 9.45, "learning_rate": 1.6644423260247855e-07, "loss": 2.1421, "step": 49590 }, { "epoch": 9.45, "learning_rate": 1.6638703527168734e-07, "loss": 1.2354, "step": 49591 }, { "epoch": 9.45, "learning_rate": 1.6632983794089608e-07, "loss": 1.7553, "step": 49592 }, { "epoch": 9.45, "learning_rate": 1.6627264061010488e-07, "loss": 1.909, "step": 49593 }, { "epoch": 9.45, "learning_rate": 1.6621544327931364e-07, "loss": 2.6906, "step": 49594 }, { "epoch": 9.45, "learning_rate": 1.661582459485224e-07, "loss": 0.775, "step": 49595 }, { "epoch": 9.45, "learning_rate": 1.6610104861773118e-07, "loss": 0.88, "step": 49596 }, { "epoch": 9.45, "learning_rate": 1.6604385128693994e-07, "loss": 1.5956, "step": 49597 }, { "epoch": 9.45, "learning_rate": 1.659866539561487e-07, "loss": 1.7458, "step": 49598 }, { "epoch": 9.45, "learning_rate": 1.6592945662535748e-07, "loss": 2.4359, "step": 49599 }, { "epoch": 9.45, "learning_rate": 1.6587225929456627e-07, "loss": 2.0269, "step": 49600 }, { "epoch": 9.45, "learning_rate": 1.6581506196377501e-07, "loss": 1.8016, "step": 49601 }, { "epoch": 9.45, "learning_rate": 1.657578646329838e-07, "loss": 1.0056, "step": 49602 }, { "epoch": 9.45, "learning_rate": 1.6570066730219258e-07, "loss": 1.2794, "step": 49603 }, { "epoch": 9.45, "learning_rate": 1.6564346997140134e-07, "loss": 1.3353, "step": 49604 }, { "epoch": 9.45, "learning_rate": 1.655862726406101e-07, "loss": 0.3241, "step": 49605 }, { "epoch": 9.45, "learning_rate": 1.6552907530981888e-07, "loss": 1.0562, "step": 49606 }, { "epoch": 9.45, "learning_rate": 1.6547187797902764e-07, "loss": 1.8033, "step": 49607 }, { "epoch": 9.45, "learning_rate": 1.654146806482364e-07, "loss": 0.7536, "step": 49608 }, { "epoch": 9.45, "learning_rate": 1.653574833174452e-07, "loss": 1.2749, "step": 49609 }, { "epoch": 9.45, "learning_rate": 1.6530028598665395e-07, "loss": 1.6731, "step": 49610 }, { "epoch": 9.45, "learning_rate": 1.6524308865586274e-07, "loss": 1.0893, "step": 49611 }, { "epoch": 9.45, "learning_rate": 1.651858913250715e-07, "loss": 1.7362, "step": 49612 }, { "epoch": 9.45, "learning_rate": 1.6512869399428028e-07, "loss": 2.939, "step": 49613 }, { "epoch": 9.45, "learning_rate": 1.6507149666348904e-07, "loss": 0.9922, "step": 49614 }, { "epoch": 9.45, "learning_rate": 1.650142993326978e-07, "loss": 1.782, "step": 49615 }, { "epoch": 9.45, "learning_rate": 1.6495710200190658e-07, "loss": 2.071, "step": 49616 }, { "epoch": 9.45, "learning_rate": 1.6489990467111534e-07, "loss": 1.9804, "step": 49617 }, { "epoch": 9.45, "learning_rate": 1.6484270734032414e-07, "loss": 1.8414, "step": 49618 }, { "epoch": 9.45, "learning_rate": 1.6478551000953288e-07, "loss": 1.8713, "step": 49619 }, { "epoch": 9.45, "learning_rate": 1.6472831267874167e-07, "loss": 1.2303, "step": 49620 }, { "epoch": 9.45, "learning_rate": 1.6467111534795044e-07, "loss": 1.7938, "step": 49621 }, { "epoch": 9.45, "learning_rate": 1.646139180171592e-07, "loss": 1.2009, "step": 49622 }, { "epoch": 9.45, "learning_rate": 1.6455672068636798e-07, "loss": 1.4643, "step": 49623 }, { "epoch": 9.45, "learning_rate": 1.6449952335557674e-07, "loss": 3.2388, "step": 49624 }, { "epoch": 9.45, "learning_rate": 1.644423260247855e-07, "loss": 0.7371, "step": 49625 }, { "epoch": 9.45, "learning_rate": 1.6438512869399428e-07, "loss": 1.3113, "step": 49626 }, { "epoch": 9.45, "learning_rate": 1.6432793136320304e-07, "loss": 1.4943, "step": 49627 }, { "epoch": 9.45, "learning_rate": 1.642707340324118e-07, "loss": 1.5402, "step": 49628 }, { "epoch": 9.45, "learning_rate": 1.642135367016206e-07, "loss": 0.9736, "step": 49629 }, { "epoch": 9.45, "learning_rate": 1.6415633937082935e-07, "loss": 1.2496, "step": 49630 }, { "epoch": 9.45, "learning_rate": 1.6409914204003814e-07, "loss": 1.4315, "step": 49631 }, { "epoch": 9.45, "learning_rate": 1.640419447092469e-07, "loss": 1.0992, "step": 49632 }, { "epoch": 9.45, "learning_rate": 1.6398474737845568e-07, "loss": 1.5437, "step": 49633 }, { "epoch": 9.45, "learning_rate": 1.6392755004766444e-07, "loss": 1.5248, "step": 49634 }, { "epoch": 9.45, "learning_rate": 1.638703527168732e-07, "loss": 1.8644, "step": 49635 }, { "epoch": 9.45, "learning_rate": 1.6381315538608198e-07, "loss": 0.6968, "step": 49636 }, { "epoch": 9.45, "learning_rate": 1.6375595805529074e-07, "loss": 1.4106, "step": 49637 }, { "epoch": 9.45, "learning_rate": 1.6369876072449954e-07, "loss": 1.129, "step": 49638 }, { "epoch": 9.46, "learning_rate": 1.6364156339370828e-07, "loss": 0.7215, "step": 49639 }, { "epoch": 9.46, "learning_rate": 1.6358436606291707e-07, "loss": 1.1752, "step": 49640 }, { "epoch": 9.46, "learning_rate": 1.6352716873212584e-07, "loss": 1.8242, "step": 49641 }, { "epoch": 9.46, "learning_rate": 1.634699714013346e-07, "loss": 1.2771, "step": 49642 }, { "epoch": 9.46, "learning_rate": 1.6341277407054338e-07, "loss": 1.2576, "step": 49643 }, { "epoch": 9.46, "learning_rate": 1.6335557673975214e-07, "loss": 1.3457, "step": 49644 }, { "epoch": 9.46, "learning_rate": 1.632983794089609e-07, "loss": 1.0331, "step": 49645 }, { "epoch": 9.46, "learning_rate": 1.6324118207816968e-07, "loss": 1.6229, "step": 49646 }, { "epoch": 9.46, "learning_rate": 1.6318398474737847e-07, "loss": 1.2824, "step": 49647 }, { "epoch": 9.46, "learning_rate": 1.631267874165872e-07, "loss": 1.7293, "step": 49648 }, { "epoch": 9.46, "learning_rate": 1.63069590085796e-07, "loss": 1.9634, "step": 49649 }, { "epoch": 9.46, "learning_rate": 1.6301239275500477e-07, "loss": 1.5976, "step": 49650 }, { "epoch": 9.46, "learning_rate": 1.6295519542421354e-07, "loss": 0.4196, "step": 49651 }, { "epoch": 9.46, "learning_rate": 1.628979980934223e-07, "loss": 3.2336, "step": 49652 }, { "epoch": 9.46, "learning_rate": 1.6284080076263108e-07, "loss": 1.8354, "step": 49653 }, { "epoch": 9.46, "learning_rate": 1.6278360343183984e-07, "loss": 1.0466, "step": 49654 }, { "epoch": 9.46, "learning_rate": 1.627264061010486e-07, "loss": 1.2458, "step": 49655 }, { "epoch": 9.46, "learning_rate": 1.626692087702574e-07, "loss": 1.608, "step": 49656 }, { "epoch": 9.46, "learning_rate": 1.6261201143946614e-07, "loss": 1.2284, "step": 49657 }, { "epoch": 9.46, "learning_rate": 1.6255481410867494e-07, "loss": 1.1194, "step": 49658 }, { "epoch": 9.46, "learning_rate": 1.624976167778837e-07, "loss": 0.7552, "step": 49659 }, { "epoch": 9.46, "learning_rate": 1.6244041944709247e-07, "loss": 0.3746, "step": 49660 }, { "epoch": 9.46, "learning_rate": 1.6238322211630124e-07, "loss": 2.0095, "step": 49661 }, { "epoch": 9.46, "learning_rate": 1.6232602478551003e-07, "loss": 2.6849, "step": 49662 }, { "epoch": 9.46, "learning_rate": 1.6226882745471878e-07, "loss": 1.9653, "step": 49663 }, { "epoch": 9.46, "learning_rate": 1.6221163012392754e-07, "loss": 1.6574, "step": 49664 }, { "epoch": 9.46, "learning_rate": 1.6215443279313634e-07, "loss": 1.2237, "step": 49665 }, { "epoch": 9.46, "learning_rate": 1.6209723546234508e-07, "loss": 1.9803, "step": 49666 }, { "epoch": 9.46, "learning_rate": 1.6204003813155387e-07, "loss": 1.4574, "step": 49667 }, { "epoch": 9.46, "learning_rate": 1.6198284080076264e-07, "loss": 1.0703, "step": 49668 }, { "epoch": 9.46, "learning_rate": 1.619256434699714e-07, "loss": 1.898, "step": 49669 }, { "epoch": 9.46, "learning_rate": 1.6186844613918017e-07, "loss": 1.4298, "step": 49670 }, { "epoch": 9.46, "learning_rate": 1.6181124880838897e-07, "loss": 0.3368, "step": 49671 }, { "epoch": 9.46, "learning_rate": 1.617540514775977e-07, "loss": 1.5295, "step": 49672 }, { "epoch": 9.46, "learning_rate": 1.616968541468065e-07, "loss": 1.4054, "step": 49673 }, { "epoch": 9.46, "learning_rate": 1.6163965681601527e-07, "loss": 2.5394, "step": 49674 }, { "epoch": 9.46, "learning_rate": 1.61582459485224e-07, "loss": 2.7495, "step": 49675 }, { "epoch": 9.46, "learning_rate": 1.615252621544328e-07, "loss": 1.8523, "step": 49676 }, { "epoch": 9.46, "learning_rate": 1.6146806482364157e-07, "loss": 1.4757, "step": 49677 }, { "epoch": 9.46, "learning_rate": 1.6141086749285034e-07, "loss": 2.0991, "step": 49678 }, { "epoch": 9.46, "learning_rate": 1.613536701620591e-07, "loss": 1.8883, "step": 49679 }, { "epoch": 9.46, "learning_rate": 1.612964728312679e-07, "loss": 0.7188, "step": 49680 }, { "epoch": 9.46, "learning_rate": 1.6123927550047664e-07, "loss": 2.2479, "step": 49681 }, { "epoch": 9.46, "learning_rate": 1.6118207816968543e-07, "loss": 1.7848, "step": 49682 }, { "epoch": 9.46, "learning_rate": 1.611248808388942e-07, "loss": 1.5486, "step": 49683 }, { "epoch": 9.46, "learning_rate": 1.6106768350810297e-07, "loss": 2.4476, "step": 49684 }, { "epoch": 9.46, "learning_rate": 1.6101048617731174e-07, "loss": 2.423, "step": 49685 }, { "epoch": 9.46, "learning_rate": 1.6095328884652048e-07, "loss": 1.5861, "step": 49686 }, { "epoch": 9.46, "learning_rate": 1.6089609151572927e-07, "loss": 1.6247, "step": 49687 }, { "epoch": 9.46, "learning_rate": 1.6083889418493804e-07, "loss": 0.4331, "step": 49688 }, { "epoch": 9.46, "learning_rate": 1.607816968541468e-07, "loss": 2.9995, "step": 49689 }, { "epoch": 9.46, "learning_rate": 1.6072449952335557e-07, "loss": 1.6776, "step": 49690 }, { "epoch": 9.46, "learning_rate": 1.6066730219256437e-07, "loss": 1.125, "step": 49691 }, { "epoch": 9.47, "learning_rate": 1.606101048617731e-07, "loss": 1.8648, "step": 49692 }, { "epoch": 9.47, "learning_rate": 1.605529075309819e-07, "loss": 0.5111, "step": 49693 }, { "epoch": 9.47, "learning_rate": 1.6049571020019067e-07, "loss": 1.4878, "step": 49694 }, { "epoch": 9.47, "learning_rate": 1.6043851286939944e-07, "loss": 1.0155, "step": 49695 }, { "epoch": 9.47, "learning_rate": 1.603813155386082e-07, "loss": 1.4472, "step": 49696 }, { "epoch": 9.47, "learning_rate": 1.6032411820781697e-07, "loss": 3.395, "step": 49697 }, { "epoch": 9.47, "learning_rate": 1.6026692087702574e-07, "loss": 1.5147, "step": 49698 }, { "epoch": 9.47, "learning_rate": 1.602097235462345e-07, "loss": 1.827, "step": 49699 }, { "epoch": 9.47, "learning_rate": 1.601525262154433e-07, "loss": 2.9307, "step": 49700 }, { "epoch": 9.47, "learning_rate": 1.6009532888465204e-07, "loss": 0.9249, "step": 49701 }, { "epoch": 9.47, "learning_rate": 1.6003813155386083e-07, "loss": 0.6316, "step": 49702 }, { "epoch": 9.47, "learning_rate": 1.599809342230696e-07, "loss": 1.4335, "step": 49703 }, { "epoch": 9.47, "learning_rate": 1.5992373689227837e-07, "loss": 2.2097, "step": 49704 }, { "epoch": 9.47, "learning_rate": 1.5986653956148714e-07, "loss": 2.0638, "step": 49705 }, { "epoch": 9.47, "learning_rate": 1.598093422306959e-07, "loss": 0.9475, "step": 49706 }, { "epoch": 9.47, "learning_rate": 1.5975214489990467e-07, "loss": 2.0338, "step": 49707 }, { "epoch": 9.47, "learning_rate": 1.5969494756911344e-07, "loss": 1.4476, "step": 49708 }, { "epoch": 9.47, "learning_rate": 1.5963775023832223e-07, "loss": 0.9114, "step": 49709 }, { "epoch": 9.47, "learning_rate": 1.5958055290753097e-07, "loss": 3.359, "step": 49710 }, { "epoch": 9.47, "learning_rate": 1.5952335557673977e-07, "loss": 1.0813, "step": 49711 }, { "epoch": 9.47, "learning_rate": 1.5946615824594853e-07, "loss": 1.4555, "step": 49712 }, { "epoch": 9.47, "learning_rate": 1.594089609151573e-07, "loss": 1.4163, "step": 49713 }, { "epoch": 9.47, "learning_rate": 1.5935176358436607e-07, "loss": 1.3247, "step": 49714 }, { "epoch": 9.47, "learning_rate": 1.5929456625357484e-07, "loss": 1.8143, "step": 49715 }, { "epoch": 9.47, "learning_rate": 1.592373689227836e-07, "loss": 1.7138, "step": 49716 }, { "epoch": 9.47, "learning_rate": 1.5918017159199237e-07, "loss": 1.3418, "step": 49717 }, { "epoch": 9.47, "learning_rate": 1.5912297426120116e-07, "loss": 1.6013, "step": 49718 }, { "epoch": 9.47, "learning_rate": 1.590657769304099e-07, "loss": 1.3228, "step": 49719 }, { "epoch": 9.47, "learning_rate": 1.590085795996187e-07, "loss": 3.2814, "step": 49720 }, { "epoch": 9.47, "learning_rate": 1.5895138226882747e-07, "loss": 2.7074, "step": 49721 }, { "epoch": 9.47, "learning_rate": 1.5889418493803623e-07, "loss": 1.1675, "step": 49722 }, { "epoch": 9.47, "learning_rate": 1.58836987607245e-07, "loss": 1.4252, "step": 49723 }, { "epoch": 9.47, "learning_rate": 1.5877979027645377e-07, "loss": 0.5841, "step": 49724 }, { "epoch": 9.47, "learning_rate": 1.5872259294566254e-07, "loss": 2.0041, "step": 49725 }, { "epoch": 9.47, "learning_rate": 1.586653956148713e-07, "loss": 1.285, "step": 49726 }, { "epoch": 9.47, "learning_rate": 1.586081982840801e-07, "loss": 1.845, "step": 49727 }, { "epoch": 9.47, "learning_rate": 1.5855100095328884e-07, "loss": 2.3022, "step": 49728 }, { "epoch": 9.47, "learning_rate": 1.5849380362249763e-07, "loss": 2.2786, "step": 49729 }, { "epoch": 9.47, "learning_rate": 1.584366062917064e-07, "loss": 1.3579, "step": 49730 }, { "epoch": 9.47, "learning_rate": 1.5837940896091517e-07, "loss": 2.0195, "step": 49731 }, { "epoch": 9.47, "learning_rate": 1.5832221163012393e-07, "loss": 2.4444, "step": 49732 }, { "epoch": 9.47, "learning_rate": 1.582650142993327e-07, "loss": 0.2676, "step": 49733 }, { "epoch": 9.47, "learning_rate": 1.5820781696854147e-07, "loss": 1.3065, "step": 49734 }, { "epoch": 9.47, "learning_rate": 1.5815061963775024e-07, "loss": 2.2175, "step": 49735 }, { "epoch": 9.47, "learning_rate": 1.5809342230695903e-07, "loss": 1.2324, "step": 49736 }, { "epoch": 9.47, "learning_rate": 1.5803622497616777e-07, "loss": 0.93, "step": 49737 }, { "epoch": 9.47, "learning_rate": 1.5797902764537656e-07, "loss": 1.1185, "step": 49738 }, { "epoch": 9.47, "learning_rate": 1.5792183031458533e-07, "loss": 1.7678, "step": 49739 }, { "epoch": 9.47, "learning_rate": 1.578646329837941e-07, "loss": 0.8507, "step": 49740 }, { "epoch": 9.47, "learning_rate": 1.5780743565300287e-07, "loss": 1.8463, "step": 49741 }, { "epoch": 9.47, "learning_rate": 1.5775023832221163e-07, "loss": 0.4836, "step": 49742 }, { "epoch": 9.47, "learning_rate": 1.576930409914204e-07, "loss": 1.2632, "step": 49743 }, { "epoch": 9.48, "learning_rate": 1.5763584366062917e-07, "loss": 2.1243, "step": 49744 }, { "epoch": 9.48, "learning_rate": 1.5757864632983794e-07, "loss": 0.4878, "step": 49745 }, { "epoch": 9.48, "learning_rate": 1.575214489990467e-07, "loss": 0.8578, "step": 49746 }, { "epoch": 9.48, "learning_rate": 1.574642516682555e-07, "loss": 1.2283, "step": 49747 }, { "epoch": 9.48, "learning_rate": 1.5740705433746424e-07, "loss": 1.7861, "step": 49748 }, { "epoch": 9.48, "learning_rate": 1.5734985700667303e-07, "loss": 2.0115, "step": 49749 }, { "epoch": 9.48, "learning_rate": 1.572926596758818e-07, "loss": 1.7751, "step": 49750 }, { "epoch": 9.48, "learning_rate": 1.5723546234509057e-07, "loss": 0.9628, "step": 49751 }, { "epoch": 9.48, "learning_rate": 1.5717826501429933e-07, "loss": 0.2498, "step": 49752 }, { "epoch": 9.48, "learning_rate": 1.571210676835081e-07, "loss": 0.6916, "step": 49753 }, { "epoch": 9.48, "learning_rate": 1.5706387035271687e-07, "loss": 1.0989, "step": 49754 }, { "epoch": 9.48, "learning_rate": 1.5700667302192564e-07, "loss": 1.0185, "step": 49755 }, { "epoch": 9.48, "learning_rate": 1.5694947569113443e-07, "loss": 1.6213, "step": 49756 }, { "epoch": 9.48, "learning_rate": 1.5689227836034317e-07, "loss": 2.7627, "step": 49757 }, { "epoch": 9.48, "learning_rate": 1.5683508102955196e-07, "loss": 0.8789, "step": 49758 }, { "epoch": 9.48, "learning_rate": 1.5677788369876073e-07, "loss": 1.1604, "step": 49759 }, { "epoch": 9.48, "learning_rate": 1.567206863679695e-07, "loss": 1.3235, "step": 49760 }, { "epoch": 9.48, "learning_rate": 1.5666348903717827e-07, "loss": 1.6031, "step": 49761 }, { "epoch": 9.48, "learning_rate": 1.5660629170638703e-07, "loss": 1.0068, "step": 49762 }, { "epoch": 9.48, "learning_rate": 1.565490943755958e-07, "loss": 1.5719, "step": 49763 }, { "epoch": 9.48, "learning_rate": 1.5649189704480457e-07, "loss": 2.343, "step": 49764 }, { "epoch": 9.48, "learning_rate": 1.5643469971401336e-07, "loss": 1.2381, "step": 49765 }, { "epoch": 9.48, "learning_rate": 1.563775023832221e-07, "loss": 0.9017, "step": 49766 }, { "epoch": 9.48, "learning_rate": 1.563203050524309e-07, "loss": 1.01, "step": 49767 }, { "epoch": 9.48, "learning_rate": 1.5626310772163966e-07, "loss": 1.4018, "step": 49768 }, { "epoch": 9.48, "learning_rate": 1.5620591039084843e-07, "loss": 0.9632, "step": 49769 }, { "epoch": 9.48, "learning_rate": 1.561487130600572e-07, "loss": 0.2283, "step": 49770 }, { "epoch": 9.48, "learning_rate": 1.5609151572926597e-07, "loss": 1.6584, "step": 49771 }, { "epoch": 9.48, "learning_rate": 1.5603431839847473e-07, "loss": 1.0076, "step": 49772 }, { "epoch": 9.48, "learning_rate": 1.559771210676835e-07, "loss": 1.6857, "step": 49773 }, { "epoch": 9.48, "learning_rate": 1.559199237368923e-07, "loss": 1.4517, "step": 49774 }, { "epoch": 9.48, "learning_rate": 1.5586272640610104e-07, "loss": 1.6765, "step": 49775 }, { "epoch": 9.48, "learning_rate": 1.5580552907530983e-07, "loss": 0.8773, "step": 49776 }, { "epoch": 9.48, "learning_rate": 1.557483317445186e-07, "loss": 1.4922, "step": 49777 }, { "epoch": 9.48, "learning_rate": 1.5569113441372736e-07, "loss": 2.2279, "step": 49778 }, { "epoch": 9.48, "learning_rate": 1.5563393708293613e-07, "loss": 2.0708, "step": 49779 }, { "epoch": 9.48, "learning_rate": 1.555767397521449e-07, "loss": 0.9445, "step": 49780 }, { "epoch": 9.48, "learning_rate": 1.5551954242135367e-07, "loss": 1.5065, "step": 49781 }, { "epoch": 9.48, "learning_rate": 1.5546234509056243e-07, "loss": 1.2453, "step": 49782 }, { "epoch": 9.48, "learning_rate": 1.5540514775977123e-07, "loss": 2.9816, "step": 49783 }, { "epoch": 9.48, "learning_rate": 1.5534795042897997e-07, "loss": 2.3409, "step": 49784 }, { "epoch": 9.48, "learning_rate": 1.5529075309818876e-07, "loss": 2.4708, "step": 49785 }, { "epoch": 9.48, "learning_rate": 1.5523355576739753e-07, "loss": 1.0898, "step": 49786 }, { "epoch": 9.48, "learning_rate": 1.551763584366063e-07, "loss": 0.9139, "step": 49787 }, { "epoch": 9.48, "learning_rate": 1.5511916110581506e-07, "loss": 1.8732, "step": 49788 }, { "epoch": 9.48, "learning_rate": 1.5506196377502386e-07, "loss": 2.1149, "step": 49789 }, { "epoch": 9.48, "learning_rate": 1.550047664442326e-07, "loss": 1.1547, "step": 49790 }, { "epoch": 9.48, "learning_rate": 1.5494756911344137e-07, "loss": 1.9743, "step": 49791 }, { "epoch": 9.48, "learning_rate": 1.5489037178265016e-07, "loss": 1.767, "step": 49792 }, { "epoch": 9.48, "learning_rate": 1.548331744518589e-07, "loss": 0.9221, "step": 49793 }, { "epoch": 9.48, "learning_rate": 1.547759771210677e-07, "loss": 1.8717, "step": 49794 }, { "epoch": 9.48, "learning_rate": 1.5471877979027646e-07, "loss": 1.0161, "step": 49795 }, { "epoch": 9.48, "learning_rate": 1.5466158245948523e-07, "loss": 2.5301, "step": 49796 }, { "epoch": 9.49, "learning_rate": 1.54604385128694e-07, "loss": 0.9224, "step": 49797 }, { "epoch": 9.49, "learning_rate": 1.545471877979028e-07, "loss": 1.083, "step": 49798 }, { "epoch": 9.49, "learning_rate": 1.5448999046711153e-07, "loss": 1.445, "step": 49799 }, { "epoch": 9.49, "learning_rate": 1.544327931363203e-07, "loss": 2.1976, "step": 49800 }, { "epoch": 9.49, "learning_rate": 1.543755958055291e-07, "loss": 0.7837, "step": 49801 }, { "epoch": 9.49, "learning_rate": 1.5431839847473783e-07, "loss": 2.1847, "step": 49802 }, { "epoch": 9.49, "learning_rate": 1.5426120114394663e-07, "loss": 1.3701, "step": 49803 }, { "epoch": 9.49, "learning_rate": 1.542040038131554e-07, "loss": 2.6064, "step": 49804 }, { "epoch": 9.49, "learning_rate": 1.5414680648236416e-07, "loss": 1.9965, "step": 49805 }, { "epoch": 9.49, "learning_rate": 1.5408960915157293e-07, "loss": 0.8882, "step": 49806 }, { "epoch": 9.49, "learning_rate": 1.540324118207817e-07, "loss": 0.9015, "step": 49807 }, { "epoch": 9.49, "learning_rate": 1.5397521448999046e-07, "loss": 0.8255, "step": 49808 }, { "epoch": 9.49, "learning_rate": 1.5391801715919926e-07, "loss": 0.9804, "step": 49809 }, { "epoch": 9.49, "learning_rate": 1.53860819828408e-07, "loss": 2.0712, "step": 49810 }, { "epoch": 9.49, "learning_rate": 1.5380362249761677e-07, "loss": 1.8273, "step": 49811 }, { "epoch": 9.49, "learning_rate": 1.5374642516682556e-07, "loss": 1.6098, "step": 49812 }, { "epoch": 9.49, "learning_rate": 1.536892278360343e-07, "loss": 1.7672, "step": 49813 }, { "epoch": 9.49, "learning_rate": 1.536320305052431e-07, "loss": 1.0708, "step": 49814 }, { "epoch": 9.49, "learning_rate": 1.5357483317445186e-07, "loss": 1.1475, "step": 49815 }, { "epoch": 9.49, "learning_rate": 1.5351763584366063e-07, "loss": 0.8977, "step": 49816 }, { "epoch": 9.49, "learning_rate": 1.534604385128694e-07, "loss": 2.0099, "step": 49817 }, { "epoch": 9.49, "learning_rate": 1.534032411820782e-07, "loss": 2.4683, "step": 49818 }, { "epoch": 9.49, "learning_rate": 1.5334604385128693e-07, "loss": 1.3112, "step": 49819 }, { "epoch": 9.49, "learning_rate": 1.5328884652049573e-07, "loss": 2.2655, "step": 49820 }, { "epoch": 9.49, "learning_rate": 1.532316491897045e-07, "loss": 1.2401, "step": 49821 }, { "epoch": 9.49, "learning_rate": 1.5317445185891323e-07, "loss": 1.2978, "step": 49822 }, { "epoch": 9.49, "learning_rate": 1.5311725452812203e-07, "loss": 1.5422, "step": 49823 }, { "epoch": 9.49, "learning_rate": 1.530600571973308e-07, "loss": 1.6993, "step": 49824 }, { "epoch": 9.49, "learning_rate": 1.5300285986653956e-07, "loss": 1.4, "step": 49825 }, { "epoch": 9.49, "learning_rate": 1.5294566253574833e-07, "loss": 1.0954, "step": 49826 }, { "epoch": 9.49, "learning_rate": 1.5288846520495712e-07, "loss": 1.7916, "step": 49827 }, { "epoch": 9.49, "learning_rate": 1.5283126787416586e-07, "loss": 2.3518, "step": 49828 }, { "epoch": 9.49, "learning_rate": 1.5277407054337466e-07, "loss": 2.7315, "step": 49829 }, { "epoch": 9.49, "learning_rate": 1.5271687321258343e-07, "loss": 1.2948, "step": 49830 }, { "epoch": 9.49, "learning_rate": 1.526596758817922e-07, "loss": 1.3778, "step": 49831 }, { "epoch": 9.49, "learning_rate": 1.5260247855100096e-07, "loss": 2.4378, "step": 49832 }, { "epoch": 9.49, "learning_rate": 1.5254528122020973e-07, "loss": 1.874, "step": 49833 }, { "epoch": 9.49, "learning_rate": 1.524880838894185e-07, "loss": 3.6171, "step": 49834 }, { "epoch": 9.49, "learning_rate": 1.5243088655862726e-07, "loss": 2.074, "step": 49835 }, { "epoch": 9.49, "learning_rate": 1.5237368922783606e-07, "loss": 1.8137, "step": 49836 }, { "epoch": 9.49, "learning_rate": 1.523164918970448e-07, "loss": 1.9921, "step": 49837 }, { "epoch": 9.49, "learning_rate": 1.522592945662536e-07, "loss": 1.6913, "step": 49838 }, { "epoch": 9.49, "learning_rate": 1.5220209723546236e-07, "loss": 3.2991, "step": 49839 }, { "epoch": 9.49, "learning_rate": 1.5214489990467113e-07, "loss": 1.5559, "step": 49840 }, { "epoch": 9.49, "learning_rate": 1.520877025738799e-07, "loss": 1.6202, "step": 49841 }, { "epoch": 9.49, "learning_rate": 1.5203050524308866e-07, "loss": 0.9108, "step": 49842 }, { "epoch": 9.49, "learning_rate": 1.5197330791229743e-07, "loss": 1.5424, "step": 49843 }, { "epoch": 9.49, "learning_rate": 1.519161105815062e-07, "loss": 1.1548, "step": 49844 }, { "epoch": 9.49, "learning_rate": 1.51858913250715e-07, "loss": 2.1778, "step": 49845 }, { "epoch": 9.49, "learning_rate": 1.5180171591992373e-07, "loss": 1.9959, "step": 49846 }, { "epoch": 9.49, "learning_rate": 1.5174451858913252e-07, "loss": 1.5349, "step": 49847 }, { "epoch": 9.49, "learning_rate": 1.516873212583413e-07, "loss": 0.8832, "step": 49848 }, { "epoch": 9.5, "learning_rate": 1.5163012392755006e-07, "loss": 1.1272, "step": 49849 }, { "epoch": 9.5, "learning_rate": 1.5157292659675882e-07, "loss": 0.942, "step": 49850 }, { "epoch": 9.5, "learning_rate": 1.515157292659676e-07, "loss": 0.7822, "step": 49851 }, { "epoch": 9.5, "learning_rate": 1.5145853193517636e-07, "loss": 2.1795, "step": 49852 }, { "epoch": 9.5, "learning_rate": 1.5140133460438513e-07, "loss": 0.3616, "step": 49853 }, { "epoch": 9.5, "learning_rate": 1.5134413727359392e-07, "loss": 3.3421, "step": 49854 }, { "epoch": 9.5, "learning_rate": 1.5128693994280266e-07, "loss": 1.2707, "step": 49855 }, { "epoch": 9.5, "learning_rate": 1.5122974261201146e-07, "loss": 1.1515, "step": 49856 }, { "epoch": 9.5, "learning_rate": 1.5117254528122022e-07, "loss": 2.2138, "step": 49857 }, { "epoch": 9.5, "learning_rate": 1.51115347950429e-07, "loss": 0.6123, "step": 49858 }, { "epoch": 9.5, "learning_rate": 1.5105815061963776e-07, "loss": 2.1202, "step": 49859 }, { "epoch": 9.5, "learning_rate": 1.5100095328884652e-07, "loss": 2.022, "step": 49860 }, { "epoch": 9.5, "learning_rate": 1.509437559580553e-07, "loss": 3.0016, "step": 49861 }, { "epoch": 9.5, "learning_rate": 1.5088655862726406e-07, "loss": 1.38, "step": 49862 }, { "epoch": 9.5, "learning_rate": 1.5082936129647285e-07, "loss": 2.2708, "step": 49863 }, { "epoch": 9.5, "learning_rate": 1.507721639656816e-07, "loss": 2.7968, "step": 49864 }, { "epoch": 9.5, "learning_rate": 1.507149666348904e-07, "loss": 3.8648, "step": 49865 }, { "epoch": 9.5, "learning_rate": 1.5065776930409913e-07, "loss": 2.0733, "step": 49866 }, { "epoch": 9.5, "learning_rate": 1.5060057197330792e-07, "loss": 1.0114, "step": 49867 }, { "epoch": 9.5, "learning_rate": 1.505433746425167e-07, "loss": 0.8994, "step": 49868 }, { "epoch": 9.5, "learning_rate": 1.5048617731172546e-07, "loss": 2.3004, "step": 49869 }, { "epoch": 9.5, "learning_rate": 1.5042897998093422e-07, "loss": 0.9002, "step": 49870 }, { "epoch": 9.5, "learning_rate": 1.50371782650143e-07, "loss": 0.6467, "step": 49871 }, { "epoch": 9.5, "learning_rate": 1.5031458531935176e-07, "loss": 2.3621, "step": 49872 }, { "epoch": 9.5, "learning_rate": 1.5025738798856053e-07, "loss": 0.334, "step": 49873 }, { "epoch": 9.5, "learning_rate": 1.5020019065776932e-07, "loss": 1.9699, "step": 49874 }, { "epoch": 9.5, "learning_rate": 1.5014299332697806e-07, "loss": 2.4374, "step": 49875 }, { "epoch": 9.5, "learning_rate": 1.5008579599618686e-07, "loss": 1.6513, "step": 49876 }, { "epoch": 9.5, "learning_rate": 1.5002859866539562e-07, "loss": 1.1728, "step": 49877 }, { "epoch": 9.5, "learning_rate": 1.499714013346044e-07, "loss": 1.141, "step": 49878 }, { "epoch": 9.5, "learning_rate": 1.4991420400381316e-07, "loss": 1.7749, "step": 49879 }, { "epoch": 9.5, "learning_rate": 1.4985700667302192e-07, "loss": 1.1536, "step": 49880 }, { "epoch": 9.5, "learning_rate": 1.497998093422307e-07, "loss": 0.8374, "step": 49881 }, { "epoch": 9.5, "learning_rate": 1.4974261201143946e-07, "loss": 1.3022, "step": 49882 }, { "epoch": 9.5, "learning_rate": 1.4968541468064825e-07, "loss": 1.3732, "step": 49883 }, { "epoch": 9.5, "learning_rate": 1.49628217349857e-07, "loss": 0.5723, "step": 49884 }, { "epoch": 9.5, "learning_rate": 1.495710200190658e-07, "loss": 1.6234, "step": 49885 }, { "epoch": 9.5, "learning_rate": 1.4951382268827456e-07, "loss": 2.211, "step": 49886 }, { "epoch": 9.5, "learning_rate": 1.4945662535748332e-07, "loss": 1.6985, "step": 49887 }, { "epoch": 9.5, "learning_rate": 1.493994280266921e-07, "loss": 1.5137, "step": 49888 }, { "epoch": 9.5, "learning_rate": 1.4934223069590086e-07, "loss": 1.5577, "step": 49889 }, { "epoch": 9.5, "learning_rate": 1.4928503336510962e-07, "loss": 1.4991, "step": 49890 }, { "epoch": 9.5, "learning_rate": 1.492278360343184e-07, "loss": 2.073, "step": 49891 }, { "epoch": 9.5, "learning_rate": 1.4917063870352719e-07, "loss": 2.4592, "step": 49892 }, { "epoch": 9.5, "learning_rate": 1.4911344137273593e-07, "loss": 0.9584, "step": 49893 }, { "epoch": 9.5, "learning_rate": 1.4905624404194472e-07, "loss": 2.0412, "step": 49894 }, { "epoch": 9.5, "learning_rate": 1.489990467111535e-07, "loss": 2.3492, "step": 49895 }, { "epoch": 9.5, "learning_rate": 1.4894184938036226e-07, "loss": 0.4219, "step": 49896 }, { "epoch": 9.5, "learning_rate": 1.4888465204957102e-07, "loss": 2.3337, "step": 49897 }, { "epoch": 9.5, "learning_rate": 1.488274547187798e-07, "loss": 2.8626, "step": 49898 }, { "epoch": 9.5, "learning_rate": 1.4877025738798856e-07, "loss": 1.3359, "step": 49899 }, { "epoch": 9.5, "learning_rate": 1.4871306005719732e-07, "loss": 1.6723, "step": 49900 }, { "epoch": 9.5, "learning_rate": 1.4865586272640612e-07, "loss": 1.3845, "step": 49901 }, { "epoch": 9.51, "learning_rate": 1.4859866539561486e-07, "loss": 1.5632, "step": 49902 }, { "epoch": 9.51, "learning_rate": 1.4854146806482365e-07, "loss": 0.6567, "step": 49903 }, { "epoch": 9.51, "learning_rate": 1.4848427073403242e-07, "loss": 2.2392, "step": 49904 }, { "epoch": 9.51, "learning_rate": 1.484270734032412e-07, "loss": 0.2301, "step": 49905 }, { "epoch": 9.51, "learning_rate": 1.4836987607244996e-07, "loss": 2.2663, "step": 49906 }, { "epoch": 9.51, "learning_rate": 1.4831267874165872e-07, "loss": 1.2977, "step": 49907 }, { "epoch": 9.51, "learning_rate": 1.482554814108675e-07, "loss": 1.7948, "step": 49908 }, { "epoch": 9.51, "learning_rate": 1.4819828408007626e-07, "loss": 0.294, "step": 49909 }, { "epoch": 9.51, "learning_rate": 1.4814108674928505e-07, "loss": 1.0059, "step": 49910 }, { "epoch": 9.51, "learning_rate": 1.480838894184938e-07, "loss": 1.5289, "step": 49911 }, { "epoch": 9.51, "learning_rate": 1.4802669208770259e-07, "loss": 1.8263, "step": 49912 }, { "epoch": 9.51, "learning_rate": 1.4796949475691135e-07, "loss": 1.5399, "step": 49913 }, { "epoch": 9.51, "learning_rate": 1.4791229742612012e-07, "loss": 0.5116, "step": 49914 }, { "epoch": 9.51, "learning_rate": 1.478551000953289e-07, "loss": 1.0803, "step": 49915 }, { "epoch": 9.51, "learning_rate": 1.4779790276453766e-07, "loss": 2.1212, "step": 49916 }, { "epoch": 9.51, "learning_rate": 1.4774070543374642e-07, "loss": 1.4285, "step": 49917 }, { "epoch": 9.51, "learning_rate": 1.476835081029552e-07, "loss": 0.9223, "step": 49918 }, { "epoch": 9.51, "learning_rate": 1.4762631077216398e-07, "loss": 1.9915, "step": 49919 }, { "epoch": 9.51, "learning_rate": 1.4756911344137272e-07, "loss": 1.0554, "step": 49920 }, { "epoch": 9.51, "learning_rate": 1.4751191611058152e-07, "loss": 1.033, "step": 49921 }, { "epoch": 9.51, "learning_rate": 1.4745471877979029e-07, "loss": 2.6247, "step": 49922 }, { "epoch": 9.51, "learning_rate": 1.4739752144899905e-07, "loss": 0.8299, "step": 49923 }, { "epoch": 9.51, "learning_rate": 1.4734032411820782e-07, "loss": 1.6086, "step": 49924 }, { "epoch": 9.51, "learning_rate": 1.4728312678741661e-07, "loss": 1.4506, "step": 49925 }, { "epoch": 9.51, "learning_rate": 1.4722592945662536e-07, "loss": 1.9846, "step": 49926 }, { "epoch": 9.51, "learning_rate": 1.4716873212583412e-07, "loss": 1.9056, "step": 49927 }, { "epoch": 9.51, "learning_rate": 1.471115347950429e-07, "loss": 0.7319, "step": 49928 }, { "epoch": 9.51, "learning_rate": 1.4705433746425166e-07, "loss": 1.9286, "step": 49929 }, { "epoch": 9.51, "learning_rate": 1.4699714013346045e-07, "loss": 0.79, "step": 49930 }, { "epoch": 9.51, "learning_rate": 1.469399428026692e-07, "loss": 2.4256, "step": 49931 }, { "epoch": 9.51, "learning_rate": 1.4688274547187799e-07, "loss": 2.1771, "step": 49932 }, { "epoch": 9.51, "learning_rate": 1.4682554814108675e-07, "loss": 1.238, "step": 49933 }, { "epoch": 9.51, "learning_rate": 1.4676835081029552e-07, "loss": 2.5004, "step": 49934 }, { "epoch": 9.51, "learning_rate": 1.467111534795043e-07, "loss": 1.9464, "step": 49935 }, { "epoch": 9.51, "learning_rate": 1.4665395614871308e-07, "loss": 0.9338, "step": 49936 }, { "epoch": 9.51, "learning_rate": 1.4659675881792182e-07, "loss": 1.76, "step": 49937 }, { "epoch": 9.51, "learning_rate": 1.465395614871306e-07, "loss": 1.2929, "step": 49938 }, { "epoch": 9.51, "learning_rate": 1.4648236415633938e-07, "loss": 1.1475, "step": 49939 }, { "epoch": 9.51, "learning_rate": 1.4642516682554812e-07, "loss": 2.8186, "step": 49940 }, { "epoch": 9.51, "learning_rate": 1.4636796949475692e-07, "loss": 0.728, "step": 49941 }, { "epoch": 9.51, "learning_rate": 1.4631077216396569e-07, "loss": 1.5509, "step": 49942 }, { "epoch": 9.51, "learning_rate": 1.4625357483317445e-07, "loss": 1.0883, "step": 49943 }, { "epoch": 9.51, "learning_rate": 1.4619637750238322e-07, "loss": 2.8245, "step": 49944 }, { "epoch": 9.51, "learning_rate": 1.4613918017159201e-07, "loss": 1.2388, "step": 49945 }, { "epoch": 9.51, "learning_rate": 1.4608198284080076e-07, "loss": 0.8752, "step": 49946 }, { "epoch": 9.51, "learning_rate": 1.4602478551000952e-07, "loss": 1.243, "step": 49947 }, { "epoch": 9.51, "learning_rate": 1.4596758817921832e-07, "loss": 1.5247, "step": 49948 }, { "epoch": 9.51, "learning_rate": 1.4591039084842706e-07, "loss": 1.0073, "step": 49949 }, { "epoch": 9.51, "learning_rate": 1.4585319351763585e-07, "loss": 3.2864, "step": 49950 }, { "epoch": 9.51, "learning_rate": 1.4579599618684462e-07, "loss": 1.04, "step": 49951 }, { "epoch": 9.51, "learning_rate": 1.4573879885605339e-07, "loss": 2.4513, "step": 49952 }, { "epoch": 9.51, "learning_rate": 1.4568160152526215e-07, "loss": 1.6807, "step": 49953 }, { "epoch": 9.52, "learning_rate": 1.4562440419447095e-07, "loss": 1.0602, "step": 49954 }, { "epoch": 9.52, "learning_rate": 1.455672068636797e-07, "loss": 1.8996, "step": 49955 }, { "epoch": 9.52, "learning_rate": 1.4551000953288848e-07, "loss": 1.9902, "step": 49956 }, { "epoch": 9.52, "learning_rate": 1.4545281220209725e-07, "loss": 2.593, "step": 49957 }, { "epoch": 9.52, "learning_rate": 1.45395614871306e-07, "loss": 0.2404, "step": 49958 }, { "epoch": 9.52, "learning_rate": 1.4533841754051478e-07, "loss": 1.054, "step": 49959 }, { "epoch": 9.52, "learning_rate": 1.4528122020972355e-07, "loss": 1.3796, "step": 49960 }, { "epoch": 9.52, "learning_rate": 1.4522402287893232e-07, "loss": 1.38, "step": 49961 }, { "epoch": 9.52, "learning_rate": 1.4516682554814109e-07, "loss": 0.5094, "step": 49962 }, { "epoch": 9.52, "learning_rate": 1.4510962821734988e-07, "loss": 1.5415, "step": 49963 }, { "epoch": 9.52, "learning_rate": 1.4505243088655862e-07, "loss": 1.215, "step": 49964 }, { "epoch": 9.52, "learning_rate": 1.4499523355576741e-07, "loss": 2.1076, "step": 49965 }, { "epoch": 9.52, "learning_rate": 1.4493803622497618e-07, "loss": 0.9405, "step": 49966 }, { "epoch": 9.52, "learning_rate": 1.4488083889418495e-07, "loss": 1.4148, "step": 49967 }, { "epoch": 9.52, "learning_rate": 1.4482364156339372e-07, "loss": 0.8469, "step": 49968 }, { "epoch": 9.52, "learning_rate": 1.4476644423260248e-07, "loss": 0.8293, "step": 49969 }, { "epoch": 9.52, "learning_rate": 1.4470924690181125e-07, "loss": 0.9634, "step": 49970 }, { "epoch": 9.52, "learning_rate": 1.4465204957102002e-07, "loss": 1.6962, "step": 49971 }, { "epoch": 9.52, "learning_rate": 1.445948522402288e-07, "loss": 1.676, "step": 49972 }, { "epoch": 9.52, "learning_rate": 1.4453765490943755e-07, "loss": 1.0483, "step": 49973 }, { "epoch": 9.52, "learning_rate": 1.4448045757864635e-07, "loss": 3.0609, "step": 49974 }, { "epoch": 9.52, "learning_rate": 1.4442326024785511e-07, "loss": 1.7931, "step": 49975 }, { "epoch": 9.52, "learning_rate": 1.4436606291706388e-07, "loss": 0.4112, "step": 49976 }, { "epoch": 9.52, "learning_rate": 1.4430886558627265e-07, "loss": 2.0358, "step": 49977 }, { "epoch": 9.52, "learning_rate": 1.4425166825548142e-07, "loss": 1.0704, "step": 49978 }, { "epoch": 9.52, "learning_rate": 1.4419447092469018e-07, "loss": 0.8937, "step": 49979 }, { "epoch": 9.52, "learning_rate": 1.4413727359389895e-07, "loss": 1.443, "step": 49980 }, { "epoch": 9.52, "learning_rate": 1.4408007626310774e-07, "loss": 1.9237, "step": 49981 }, { "epoch": 9.52, "learning_rate": 1.4402287893231649e-07, "loss": 2.8521, "step": 49982 }, { "epoch": 9.52, "learning_rate": 1.4396568160152528e-07, "loss": 1.9777, "step": 49983 }, { "epoch": 9.52, "learning_rate": 1.4390848427073405e-07, "loss": 0.5786, "step": 49984 }, { "epoch": 9.52, "learning_rate": 1.4385128693994281e-07, "loss": 1.6746, "step": 49985 }, { "epoch": 9.52, "learning_rate": 1.4379408960915158e-07, "loss": 1.0822, "step": 49986 }, { "epoch": 9.52, "learning_rate": 1.4373689227836035e-07, "loss": 2.5819, "step": 49987 }, { "epoch": 9.52, "learning_rate": 1.4367969494756912e-07, "loss": 1.4186, "step": 49988 }, { "epoch": 9.52, "learning_rate": 1.4362249761677788e-07, "loss": 2.3092, "step": 49989 }, { "epoch": 9.52, "learning_rate": 1.4356530028598665e-07, "loss": 2.1898, "step": 49990 }, { "epoch": 9.52, "learning_rate": 1.4350810295519542e-07, "loss": 1.8228, "step": 49991 }, { "epoch": 9.52, "learning_rate": 1.434509056244042e-07, "loss": 1.2992, "step": 49992 }, { "epoch": 9.52, "learning_rate": 1.4339370829361295e-07, "loss": 0.2256, "step": 49993 }, { "epoch": 9.52, "learning_rate": 1.4333651096282175e-07, "loss": 1.9075, "step": 49994 }, { "epoch": 9.52, "learning_rate": 1.4327931363203051e-07, "loss": 2.5943, "step": 49995 }, { "epoch": 9.52, "learning_rate": 1.4322211630123928e-07, "loss": 2.9885, "step": 49996 }, { "epoch": 9.52, "learning_rate": 1.4316491897044805e-07, "loss": 1.4674, "step": 49997 }, { "epoch": 9.52, "learning_rate": 1.4310772163965682e-07, "loss": 1.1888, "step": 49998 }, { "epoch": 9.52, "learning_rate": 1.4305052430886558e-07, "loss": 0.5082, "step": 49999 }, { "epoch": 9.52, "learning_rate": 1.4299332697807435e-07, "loss": 1.7951, "step": 50000 }, { "epoch": 9.52, "learning_rate": 1.4293612964728314e-07, "loss": 1.5277, "step": 50001 }, { "epoch": 9.52, "learning_rate": 1.4287893231649189e-07, "loss": 1.7049, "step": 50002 }, { "epoch": 9.52, "learning_rate": 1.4282173498570068e-07, "loss": 0.8866, "step": 50003 }, { "epoch": 9.52, "learning_rate": 1.4276453765490945e-07, "loss": 1.7929, "step": 50004 }, { "epoch": 9.52, "learning_rate": 1.4270734032411821e-07, "loss": 0.8513, "step": 50005 }, { "epoch": 9.52, "learning_rate": 1.4265014299332698e-07, "loss": 1.3184, "step": 50006 }, { "epoch": 9.53, "learning_rate": 1.4259294566253575e-07, "loss": 1.8985, "step": 50007 }, { "epoch": 9.53, "learning_rate": 1.4253574833174452e-07, "loss": 1.3851, "step": 50008 }, { "epoch": 9.53, "learning_rate": 1.4247855100095328e-07, "loss": 0.6725, "step": 50009 }, { "epoch": 9.53, "learning_rate": 1.4242135367016208e-07, "loss": 1.125, "step": 50010 }, { "epoch": 9.53, "learning_rate": 1.4236415633937082e-07, "loss": 1.2888, "step": 50011 }, { "epoch": 9.53, "learning_rate": 1.423069590085796e-07, "loss": 1.1135, "step": 50012 }, { "epoch": 9.53, "learning_rate": 1.4224976167778838e-07, "loss": 0.7633, "step": 50013 }, { "epoch": 9.53, "learning_rate": 1.4219256434699715e-07, "loss": 2.0017, "step": 50014 }, { "epoch": 9.53, "learning_rate": 1.4213536701620591e-07, "loss": 1.7989, "step": 50015 }, { "epoch": 9.53, "learning_rate": 1.4207816968541468e-07, "loss": 1.7588, "step": 50016 }, { "epoch": 9.53, "learning_rate": 1.4202097235462345e-07, "loss": 0.7618, "step": 50017 }, { "epoch": 9.53, "learning_rate": 1.4196377502383222e-07, "loss": 3.1769, "step": 50018 }, { "epoch": 9.53, "learning_rate": 1.41906577693041e-07, "loss": 1.3134, "step": 50019 }, { "epoch": 9.53, "learning_rate": 1.4184938036224975e-07, "loss": 0.5137, "step": 50020 }, { "epoch": 9.53, "learning_rate": 1.4179218303145854e-07, "loss": 1.1746, "step": 50021 }, { "epoch": 9.53, "learning_rate": 1.417349857006673e-07, "loss": 1.598, "step": 50022 }, { "epoch": 9.53, "learning_rate": 1.4167778836987608e-07, "loss": 0.7788, "step": 50023 }, { "epoch": 9.53, "learning_rate": 1.4162059103908485e-07, "loss": 1.1076, "step": 50024 }, { "epoch": 9.53, "learning_rate": 1.4156339370829361e-07, "loss": 2.0391, "step": 50025 }, { "epoch": 9.53, "learning_rate": 1.4150619637750238e-07, "loss": 1.5677, "step": 50026 }, { "epoch": 9.53, "learning_rate": 1.4144899904671115e-07, "loss": 0.4131, "step": 50027 }, { "epoch": 9.53, "learning_rate": 1.4139180171591994e-07, "loss": 2.2642, "step": 50028 }, { "epoch": 9.53, "learning_rate": 1.4133460438512868e-07, "loss": 2.5259, "step": 50029 }, { "epoch": 9.53, "learning_rate": 1.4127740705433748e-07, "loss": 1.5313, "step": 50030 }, { "epoch": 9.53, "learning_rate": 1.4122020972354624e-07, "loss": 1.2482, "step": 50031 }, { "epoch": 9.53, "learning_rate": 1.41163012392755e-07, "loss": 1.7717, "step": 50032 }, { "epoch": 9.53, "learning_rate": 1.4110581506196378e-07, "loss": 1.4009, "step": 50033 }, { "epoch": 9.53, "learning_rate": 1.4104861773117255e-07, "loss": 0.9775, "step": 50034 }, { "epoch": 9.53, "learning_rate": 1.4099142040038131e-07, "loss": 2.2217, "step": 50035 }, { "epoch": 9.53, "learning_rate": 1.4093422306959008e-07, "loss": 0.6009, "step": 50036 }, { "epoch": 9.53, "learning_rate": 1.4087702573879887e-07, "loss": 1.8553, "step": 50037 }, { "epoch": 9.53, "learning_rate": 1.4081982840800762e-07, "loss": 1.7703, "step": 50038 }, { "epoch": 9.53, "learning_rate": 1.407626310772164e-07, "loss": 0.3179, "step": 50039 }, { "epoch": 9.53, "learning_rate": 1.4070543374642518e-07, "loss": 2.3426, "step": 50040 }, { "epoch": 9.53, "learning_rate": 1.4064823641563394e-07, "loss": 0.8999, "step": 50041 }, { "epoch": 9.53, "learning_rate": 1.405910390848427e-07, "loss": 1.4112, "step": 50042 }, { "epoch": 9.53, "learning_rate": 1.4053384175405148e-07, "loss": 1.228, "step": 50043 }, { "epoch": 9.53, "learning_rate": 1.4047664442326025e-07, "loss": 1.1821, "step": 50044 }, { "epoch": 9.53, "learning_rate": 1.4041944709246901e-07, "loss": 1.6171, "step": 50045 }, { "epoch": 9.53, "learning_rate": 1.403622497616778e-07, "loss": 1.9526, "step": 50046 }, { "epoch": 9.53, "learning_rate": 1.4030505243088655e-07, "loss": 1.4129, "step": 50047 }, { "epoch": 9.53, "learning_rate": 1.4024785510009534e-07, "loss": 1.7849, "step": 50048 }, { "epoch": 9.53, "learning_rate": 1.4019065776930408e-07, "loss": 0.8322, "step": 50049 }, { "epoch": 9.53, "learning_rate": 1.4013346043851288e-07, "loss": 2.5799, "step": 50050 }, { "epoch": 9.53, "learning_rate": 1.4007626310772164e-07, "loss": 1.2677, "step": 50051 }, { "epoch": 9.53, "learning_rate": 1.400190657769304e-07, "loss": 1.9746, "step": 50052 }, { "epoch": 9.53, "learning_rate": 1.3996186844613918e-07, "loss": 1.1029, "step": 50053 }, { "epoch": 9.53, "learning_rate": 1.3990467111534795e-07, "loss": 0.7482, "step": 50054 }, { "epoch": 9.53, "learning_rate": 1.3984747378455671e-07, "loss": 1.7503, "step": 50055 }, { "epoch": 9.53, "learning_rate": 1.3979027645376548e-07, "loss": 1.9631, "step": 50056 }, { "epoch": 9.53, "learning_rate": 1.3973307912297427e-07, "loss": 2.1094, "step": 50057 }, { "epoch": 9.53, "learning_rate": 1.3967588179218302e-07, "loss": 2.6705, "step": 50058 }, { "epoch": 9.54, "learning_rate": 1.396186844613918e-07, "loss": 1.3245, "step": 50059 }, { "epoch": 9.54, "learning_rate": 1.3956148713060058e-07, "loss": 0.7718, "step": 50060 }, { "epoch": 9.54, "learning_rate": 1.3950428979980934e-07, "loss": 1.9166, "step": 50061 }, { "epoch": 9.54, "learning_rate": 1.394470924690181e-07, "loss": 2.07, "step": 50062 }, { "epoch": 9.54, "learning_rate": 1.3938989513822688e-07, "loss": 1.1043, "step": 50063 }, { "epoch": 9.54, "learning_rate": 1.3933269780743565e-07, "loss": 1.3511, "step": 50064 }, { "epoch": 9.54, "learning_rate": 1.3927550047664441e-07, "loss": 1.5218, "step": 50065 }, { "epoch": 9.54, "learning_rate": 1.392183031458532e-07, "loss": 1.039, "step": 50066 }, { "epoch": 9.54, "learning_rate": 1.3916110581506195e-07, "loss": 0.6138, "step": 50067 }, { "epoch": 9.54, "learning_rate": 1.3910390848427074e-07, "loss": 1.5405, "step": 50068 }, { "epoch": 9.54, "learning_rate": 1.390467111534795e-07, "loss": 0.6121, "step": 50069 }, { "epoch": 9.54, "learning_rate": 1.3898951382268828e-07, "loss": 1.0522, "step": 50070 }, { "epoch": 9.54, "learning_rate": 1.3893231649189704e-07, "loss": 2.6608, "step": 50071 }, { "epoch": 9.54, "learning_rate": 1.3887511916110584e-07, "loss": 0.8422, "step": 50072 }, { "epoch": 9.54, "learning_rate": 1.3881792183031458e-07, "loss": 2.2145, "step": 50073 }, { "epoch": 9.54, "learning_rate": 1.3876072449952335e-07, "loss": 1.1727, "step": 50074 }, { "epoch": 9.54, "learning_rate": 1.3870352716873214e-07, "loss": 0.8147, "step": 50075 }, { "epoch": 9.54, "learning_rate": 1.3864632983794088e-07, "loss": 1.8178, "step": 50076 }, { "epoch": 9.54, "learning_rate": 1.3858913250714967e-07, "loss": 1.1949, "step": 50077 }, { "epoch": 9.54, "learning_rate": 1.3853193517635844e-07, "loss": 1.1103, "step": 50078 }, { "epoch": 9.54, "learning_rate": 1.384747378455672e-07, "loss": 0.9904, "step": 50079 }, { "epoch": 9.54, "learning_rate": 1.3841754051477598e-07, "loss": 0.7396, "step": 50080 }, { "epoch": 9.54, "learning_rate": 1.3836034318398477e-07, "loss": 0.6891, "step": 50081 }, { "epoch": 9.54, "learning_rate": 1.383031458531935e-07, "loss": 0.8806, "step": 50082 }, { "epoch": 9.54, "learning_rate": 1.3824594852240228e-07, "loss": 1.2312, "step": 50083 }, { "epoch": 9.54, "learning_rate": 1.3818875119161107e-07, "loss": 0.2734, "step": 50084 }, { "epoch": 9.54, "learning_rate": 1.3813155386081981e-07, "loss": 2.2246, "step": 50085 }, { "epoch": 9.54, "learning_rate": 1.380743565300286e-07, "loss": 1.0598, "step": 50086 }, { "epoch": 9.54, "learning_rate": 1.3801715919923737e-07, "loss": 2.344, "step": 50087 }, { "epoch": 9.54, "learning_rate": 1.3795996186844614e-07, "loss": 0.7227, "step": 50088 }, { "epoch": 9.54, "learning_rate": 1.379027645376549e-07, "loss": 0.6976, "step": 50089 }, { "epoch": 9.54, "learning_rate": 1.378455672068637e-07, "loss": 2.4798, "step": 50090 }, { "epoch": 9.54, "learning_rate": 1.3778836987607244e-07, "loss": 0.9225, "step": 50091 }, { "epoch": 9.54, "learning_rate": 1.3773117254528124e-07, "loss": 2.998, "step": 50092 }, { "epoch": 9.54, "learning_rate": 1.3767397521449e-07, "loss": 2.0468, "step": 50093 }, { "epoch": 9.54, "learning_rate": 1.3761677788369875e-07, "loss": 3.0664, "step": 50094 }, { "epoch": 9.54, "learning_rate": 1.3755958055290754e-07, "loss": 0.7892, "step": 50095 }, { "epoch": 9.54, "learning_rate": 1.375023832221163e-07, "loss": 1.0263, "step": 50096 }, { "epoch": 9.54, "learning_rate": 1.3744518589132507e-07, "loss": 2.0366, "step": 50097 }, { "epoch": 9.54, "learning_rate": 1.3738798856053384e-07, "loss": 0.8871, "step": 50098 }, { "epoch": 9.54, "learning_rate": 1.3733079122974264e-07, "loss": 1.4616, "step": 50099 }, { "epoch": 9.54, "learning_rate": 1.3727359389895138e-07, "loss": 1.4709, "step": 50100 }, { "epoch": 9.54, "learning_rate": 1.3721639656816017e-07, "loss": 0.7933, "step": 50101 }, { "epoch": 9.54, "learning_rate": 1.3715919923736894e-07, "loss": 0.68, "step": 50102 }, { "epoch": 9.54, "learning_rate": 1.371020019065777e-07, "loss": 2.3724, "step": 50103 }, { "epoch": 9.54, "learning_rate": 1.3704480457578647e-07, "loss": 0.8615, "step": 50104 }, { "epoch": 9.54, "learning_rate": 1.3698760724499524e-07, "loss": 2.0864, "step": 50105 }, { "epoch": 9.54, "learning_rate": 1.36930409914204e-07, "loss": 1.4696, "step": 50106 }, { "epoch": 9.54, "learning_rate": 1.3687321258341277e-07, "loss": 1.1061, "step": 50107 }, { "epoch": 9.54, "learning_rate": 1.3681601525262154e-07, "loss": 1.7728, "step": 50108 }, { "epoch": 9.54, "learning_rate": 1.367588179218303e-07, "loss": 1.238, "step": 50109 }, { "epoch": 9.54, "learning_rate": 1.367016205910391e-07, "loss": 2.1856, "step": 50110 }, { "epoch": 9.54, "learning_rate": 1.3664442326024784e-07, "loss": 1.287, "step": 50111 }, { "epoch": 9.55, "learning_rate": 1.3658722592945664e-07, "loss": 1.5319, "step": 50112 }, { "epoch": 9.55, "learning_rate": 1.365300285986654e-07, "loss": 0.5771, "step": 50113 }, { "epoch": 9.55, "learning_rate": 1.3647283126787417e-07, "loss": 1.875, "step": 50114 }, { "epoch": 9.55, "learning_rate": 1.3641563393708294e-07, "loss": 1.9935, "step": 50115 }, { "epoch": 9.55, "learning_rate": 1.363584366062917e-07, "loss": 2.6648, "step": 50116 }, { "epoch": 9.55, "learning_rate": 1.3630123927550047e-07, "loss": 1.4912, "step": 50117 }, { "epoch": 9.55, "learning_rate": 1.3624404194470924e-07, "loss": 0.9935, "step": 50118 }, { "epoch": 9.55, "learning_rate": 1.3618684461391804e-07, "loss": 3.3118, "step": 50119 }, { "epoch": 9.55, "learning_rate": 1.3612964728312678e-07, "loss": 1.7026, "step": 50120 }, { "epoch": 9.55, "learning_rate": 1.3607244995233557e-07, "loss": 1.0515, "step": 50121 }, { "epoch": 9.55, "learning_rate": 1.3601525262154434e-07, "loss": 1.6542, "step": 50122 }, { "epoch": 9.55, "learning_rate": 1.359580552907531e-07, "loss": 1.3442, "step": 50123 }, { "epoch": 9.55, "learning_rate": 1.3590085795996187e-07, "loss": 1.217, "step": 50124 }, { "epoch": 9.55, "learning_rate": 1.3584366062917064e-07, "loss": 2.1124, "step": 50125 }, { "epoch": 9.55, "learning_rate": 1.357864632983794e-07, "loss": 1.8471, "step": 50126 }, { "epoch": 9.55, "learning_rate": 1.3572926596758817e-07, "loss": 0.8907, "step": 50127 }, { "epoch": 9.55, "learning_rate": 1.3567206863679697e-07, "loss": 0.3704, "step": 50128 }, { "epoch": 9.55, "learning_rate": 1.356148713060057e-07, "loss": 2.3126, "step": 50129 }, { "epoch": 9.55, "learning_rate": 1.355576739752145e-07, "loss": 1.3652, "step": 50130 }, { "epoch": 9.55, "learning_rate": 1.3550047664442327e-07, "loss": 1.8246, "step": 50131 }, { "epoch": 9.55, "learning_rate": 1.3544327931363204e-07, "loss": 1.1393, "step": 50132 }, { "epoch": 9.55, "learning_rate": 1.353860819828408e-07, "loss": 1.1144, "step": 50133 }, { "epoch": 9.55, "learning_rate": 1.3532888465204957e-07, "loss": 1.2258, "step": 50134 }, { "epoch": 9.55, "learning_rate": 1.3527168732125834e-07, "loss": 1.5934, "step": 50135 }, { "epoch": 9.55, "learning_rate": 1.352144899904671e-07, "loss": 1.0767, "step": 50136 }, { "epoch": 9.55, "learning_rate": 1.351572926596759e-07, "loss": 2.5479, "step": 50137 }, { "epoch": 9.55, "learning_rate": 1.3510009532888464e-07, "loss": 1.6585, "step": 50138 }, { "epoch": 9.55, "learning_rate": 1.3504289799809344e-07, "loss": 2.1423, "step": 50139 }, { "epoch": 9.55, "learning_rate": 1.349857006673022e-07, "loss": 1.3802, "step": 50140 }, { "epoch": 9.55, "learning_rate": 1.3492850333651097e-07, "loss": 1.4298, "step": 50141 }, { "epoch": 9.55, "learning_rate": 1.3487130600571974e-07, "loss": 2.2631, "step": 50142 }, { "epoch": 9.55, "learning_rate": 1.348141086749285e-07, "loss": 0.4153, "step": 50143 }, { "epoch": 9.55, "learning_rate": 1.3475691134413727e-07, "loss": 3.0173, "step": 50144 }, { "epoch": 9.55, "learning_rate": 1.3469971401334604e-07, "loss": 1.6499, "step": 50145 }, { "epoch": 9.55, "learning_rate": 1.3464251668255483e-07, "loss": 1.4568, "step": 50146 }, { "epoch": 9.55, "learning_rate": 1.3458531935176357e-07, "loss": 1.5055, "step": 50147 }, { "epoch": 9.55, "learning_rate": 1.3452812202097237e-07, "loss": 1.7167, "step": 50148 }, { "epoch": 9.55, "learning_rate": 1.3447092469018114e-07, "loss": 0.3651, "step": 50149 }, { "epoch": 9.55, "learning_rate": 1.344137273593899e-07, "loss": 1.4416, "step": 50150 }, { "epoch": 9.55, "learning_rate": 1.3435653002859867e-07, "loss": 1.5051, "step": 50151 }, { "epoch": 9.55, "learning_rate": 1.3429933269780744e-07, "loss": 1.1153, "step": 50152 }, { "epoch": 9.55, "learning_rate": 1.342421353670162e-07, "loss": 1.3737, "step": 50153 }, { "epoch": 9.55, "learning_rate": 1.3418493803622497e-07, "loss": 2.8896, "step": 50154 }, { "epoch": 9.55, "learning_rate": 1.3412774070543377e-07, "loss": 1.1789, "step": 50155 }, { "epoch": 9.55, "learning_rate": 1.340705433746425e-07, "loss": 1.9181, "step": 50156 }, { "epoch": 9.55, "learning_rate": 1.340133460438513e-07, "loss": 1.0078, "step": 50157 }, { "epoch": 9.55, "learning_rate": 1.3395614871306007e-07, "loss": 1.3824, "step": 50158 }, { "epoch": 9.55, "learning_rate": 1.3389895138226884e-07, "loss": 1.6084, "step": 50159 }, { "epoch": 9.55, "learning_rate": 1.338417540514776e-07, "loss": 1.1538, "step": 50160 }, { "epoch": 9.55, "learning_rate": 1.3378455672068637e-07, "loss": 1.572, "step": 50161 }, { "epoch": 9.55, "learning_rate": 1.3372735938989514e-07, "loss": 1.0606, "step": 50162 }, { "epoch": 9.55, "learning_rate": 1.336701620591039e-07, "loss": 2.292, "step": 50163 }, { "epoch": 9.56, "learning_rate": 1.336129647283127e-07, "loss": 1.8354, "step": 50164 }, { "epoch": 9.56, "learning_rate": 1.3355576739752144e-07, "loss": 2.1231, "step": 50165 }, { "epoch": 9.56, "learning_rate": 1.3349857006673023e-07, "loss": 1.5842, "step": 50166 }, { "epoch": 9.56, "learning_rate": 1.33441372735939e-07, "loss": 1.6386, "step": 50167 }, { "epoch": 9.56, "learning_rate": 1.3338417540514777e-07, "loss": 2.3093, "step": 50168 }, { "epoch": 9.56, "learning_rate": 1.3332697807435654e-07, "loss": 0.3294, "step": 50169 }, { "epoch": 9.56, "learning_rate": 1.332697807435653e-07, "loss": 2.0174, "step": 50170 }, { "epoch": 9.56, "learning_rate": 1.3321258341277407e-07, "loss": 1.789, "step": 50171 }, { "epoch": 9.56, "learning_rate": 1.3315538608198284e-07, "loss": 0.9395, "step": 50172 }, { "epoch": 9.56, "learning_rate": 1.330981887511916e-07, "loss": 1.3188, "step": 50173 }, { "epoch": 9.56, "learning_rate": 1.3304099142040037e-07, "loss": 3.6838, "step": 50174 }, { "epoch": 9.56, "learning_rate": 1.3298379408960917e-07, "loss": 0.8021, "step": 50175 }, { "epoch": 9.56, "learning_rate": 1.329265967588179e-07, "loss": 0.8837, "step": 50176 }, { "epoch": 9.56, "learning_rate": 1.328693994280267e-07, "loss": 1.2943, "step": 50177 }, { "epoch": 9.56, "learning_rate": 1.3281220209723547e-07, "loss": 1.1574, "step": 50178 }, { "epoch": 9.56, "learning_rate": 1.3275500476644424e-07, "loss": 0.8907, "step": 50179 }, { "epoch": 9.56, "learning_rate": 1.32697807435653e-07, "loss": 2.1935, "step": 50180 }, { "epoch": 9.56, "learning_rate": 1.3264061010486177e-07, "loss": 2.7188, "step": 50181 }, { "epoch": 9.56, "learning_rate": 1.3258341277407054e-07, "loss": 1.1654, "step": 50182 }, { "epoch": 9.56, "learning_rate": 1.325262154432793e-07, "loss": 0.6675, "step": 50183 }, { "epoch": 9.56, "learning_rate": 1.324690181124881e-07, "loss": 0.7592, "step": 50184 }, { "epoch": 9.56, "learning_rate": 1.3241182078169684e-07, "loss": 1.4529, "step": 50185 }, { "epoch": 9.56, "learning_rate": 1.3235462345090563e-07, "loss": 1.0547, "step": 50186 }, { "epoch": 9.56, "learning_rate": 1.322974261201144e-07, "loss": 1.5937, "step": 50187 }, { "epoch": 9.56, "learning_rate": 1.3224022878932317e-07, "loss": 1.719, "step": 50188 }, { "epoch": 9.56, "learning_rate": 1.3218303145853194e-07, "loss": 1.8024, "step": 50189 }, { "epoch": 9.56, "learning_rate": 1.321258341277407e-07, "loss": 1.6829, "step": 50190 }, { "epoch": 9.56, "learning_rate": 1.3206863679694947e-07, "loss": 0.7705, "step": 50191 }, { "epoch": 9.56, "learning_rate": 1.3201143946615824e-07, "loss": 2.0863, "step": 50192 }, { "epoch": 9.56, "learning_rate": 1.3195424213536703e-07, "loss": 2.8892, "step": 50193 }, { "epoch": 9.56, "learning_rate": 1.3189704480457577e-07, "loss": 0.8069, "step": 50194 }, { "epoch": 9.56, "learning_rate": 1.3183984747378457e-07, "loss": 1.8451, "step": 50195 }, { "epoch": 9.56, "learning_rate": 1.3178265014299333e-07, "loss": 1.0601, "step": 50196 }, { "epoch": 9.56, "learning_rate": 1.317254528122021e-07, "loss": 0.8079, "step": 50197 }, { "epoch": 9.56, "learning_rate": 1.3166825548141087e-07, "loss": 1.1708, "step": 50198 }, { "epoch": 9.56, "learning_rate": 1.3161105815061964e-07, "loss": 1.9233, "step": 50199 }, { "epoch": 9.56, "learning_rate": 1.315538608198284e-07, "loss": 2.2708, "step": 50200 }, { "epoch": 9.56, "learning_rate": 1.3149666348903717e-07, "loss": 1.5621, "step": 50201 }, { "epoch": 9.56, "learning_rate": 1.3143946615824596e-07, "loss": 0.8674, "step": 50202 }, { "epoch": 9.56, "learning_rate": 1.313822688274547e-07, "loss": 1.6152, "step": 50203 }, { "epoch": 9.56, "learning_rate": 1.313250714966635e-07, "loss": 2.1756, "step": 50204 }, { "epoch": 9.56, "learning_rate": 1.3126787416587227e-07, "loss": 2.1373, "step": 50205 }, { "epoch": 9.56, "learning_rate": 1.3121067683508103e-07, "loss": 1.9178, "step": 50206 }, { "epoch": 9.56, "learning_rate": 1.311534795042898e-07, "loss": 2.9478, "step": 50207 }, { "epoch": 9.56, "learning_rate": 1.310962821734986e-07, "loss": 2.28, "step": 50208 }, { "epoch": 9.56, "learning_rate": 1.3103908484270734e-07, "loss": 2.5076, "step": 50209 }, { "epoch": 9.56, "learning_rate": 1.309818875119161e-07, "loss": 0.4544, "step": 50210 }, { "epoch": 9.56, "learning_rate": 1.309246901811249e-07, "loss": 1.9428, "step": 50211 }, { "epoch": 9.56, "learning_rate": 1.3086749285033364e-07, "loss": 1.1246, "step": 50212 }, { "epoch": 9.56, "learning_rate": 1.3081029551954243e-07, "loss": 1.8264, "step": 50213 }, { "epoch": 9.56, "learning_rate": 1.307530981887512e-07, "loss": 2.919, "step": 50214 }, { "epoch": 9.56, "learning_rate": 1.3069590085795997e-07, "loss": 1.2953, "step": 50215 }, { "epoch": 9.56, "learning_rate": 1.3063870352716873e-07, "loss": 1.114, "step": 50216 }, { "epoch": 9.57, "learning_rate": 1.3058150619637753e-07, "loss": 1.5378, "step": 50217 }, { "epoch": 9.57, "learning_rate": 1.3052430886558627e-07, "loss": 0.8659, "step": 50218 }, { "epoch": 9.57, "learning_rate": 1.3046711153479506e-07, "loss": 2.7437, "step": 50219 }, { "epoch": 9.57, "learning_rate": 1.3040991420400383e-07, "loss": 1.3526, "step": 50220 }, { "epoch": 9.57, "learning_rate": 1.3035271687321257e-07, "loss": 1.8243, "step": 50221 }, { "epoch": 9.57, "learning_rate": 1.3029551954242136e-07, "loss": 0.496, "step": 50222 }, { "epoch": 9.57, "learning_rate": 1.3023832221163013e-07, "loss": 2.0591, "step": 50223 }, { "epoch": 9.57, "learning_rate": 1.301811248808389e-07, "loss": 0.5625, "step": 50224 }, { "epoch": 9.57, "learning_rate": 1.3012392755004767e-07, "loss": 0.7358, "step": 50225 }, { "epoch": 9.57, "learning_rate": 1.3006673021925646e-07, "loss": 1.4538, "step": 50226 }, { "epoch": 9.57, "learning_rate": 1.300095328884652e-07, "loss": 2.1724, "step": 50227 }, { "epoch": 9.57, "learning_rate": 1.29952335557674e-07, "loss": 1.7934, "step": 50228 }, { "epoch": 9.57, "learning_rate": 1.2989513822688274e-07, "loss": 1.1527, "step": 50229 }, { "epoch": 9.57, "learning_rate": 1.298379408960915e-07, "loss": 2.3483, "step": 50230 }, { "epoch": 9.57, "learning_rate": 1.297807435653003e-07, "loss": 2.1095, "step": 50231 }, { "epoch": 9.57, "learning_rate": 1.2972354623450904e-07, "loss": 1.5211, "step": 50232 }, { "epoch": 9.57, "learning_rate": 1.2966634890371783e-07, "loss": 0.8802, "step": 50233 }, { "epoch": 9.57, "learning_rate": 1.296091515729266e-07, "loss": 2.8106, "step": 50234 }, { "epoch": 9.57, "learning_rate": 1.2955195424213537e-07, "loss": 1.0485, "step": 50235 }, { "epoch": 9.57, "learning_rate": 1.2949475691134413e-07, "loss": 1.7323, "step": 50236 }, { "epoch": 9.57, "learning_rate": 1.2943755958055293e-07, "loss": 1.0332, "step": 50237 }, { "epoch": 9.57, "learning_rate": 1.2938036224976167e-07, "loss": 1.087, "step": 50238 }, { "epoch": 9.57, "learning_rate": 1.2932316491897046e-07, "loss": 0.2458, "step": 50239 }, { "epoch": 9.57, "learning_rate": 1.2926596758817923e-07, "loss": 1.5954, "step": 50240 }, { "epoch": 9.57, "learning_rate": 1.2920877025738797e-07, "loss": 2.7196, "step": 50241 }, { "epoch": 9.57, "learning_rate": 1.2915157292659676e-07, "loss": 1.1959, "step": 50242 }, { "epoch": 9.57, "learning_rate": 1.2909437559580553e-07, "loss": 1.9598, "step": 50243 }, { "epoch": 9.57, "learning_rate": 1.290371782650143e-07, "loss": 0.6287, "step": 50244 }, { "epoch": 9.57, "learning_rate": 1.2897998093422307e-07, "loss": 1.4605, "step": 50245 }, { "epoch": 9.57, "learning_rate": 1.2892278360343186e-07, "loss": 1.769, "step": 50246 }, { "epoch": 9.57, "learning_rate": 1.288655862726406e-07, "loss": 2.1045, "step": 50247 }, { "epoch": 9.57, "learning_rate": 1.288083889418494e-07, "loss": 1.5665, "step": 50248 }, { "epoch": 9.57, "learning_rate": 1.2875119161105816e-07, "loss": 1.3435, "step": 50249 }, { "epoch": 9.57, "learning_rate": 1.2869399428026693e-07, "loss": 1.638, "step": 50250 }, { "epoch": 9.57, "learning_rate": 1.286367969494757e-07, "loss": 1.7416, "step": 50251 }, { "epoch": 9.57, "learning_rate": 1.2857959961868446e-07, "loss": 2.7691, "step": 50252 }, { "epoch": 9.57, "learning_rate": 1.2852240228789323e-07, "loss": 0.3972, "step": 50253 }, { "epoch": 9.57, "learning_rate": 1.28465204957102e-07, "loss": 1.4956, "step": 50254 }, { "epoch": 9.57, "learning_rate": 1.284080076263108e-07, "loss": 0.9372, "step": 50255 }, { "epoch": 9.57, "learning_rate": 1.2835081029551953e-07, "loss": 0.306, "step": 50256 }, { "epoch": 9.57, "learning_rate": 1.2829361296472833e-07, "loss": 1.4025, "step": 50257 }, { "epoch": 9.57, "learning_rate": 1.282364156339371e-07, "loss": 0.5148, "step": 50258 }, { "epoch": 9.57, "learning_rate": 1.2817921830314586e-07, "loss": 0.866, "step": 50259 }, { "epoch": 9.57, "learning_rate": 1.2812202097235463e-07, "loss": 1.3189, "step": 50260 }, { "epoch": 9.57, "learning_rate": 1.280648236415634e-07, "loss": 0.3516, "step": 50261 }, { "epoch": 9.57, "learning_rate": 1.2800762631077216e-07, "loss": 1.7478, "step": 50262 }, { "epoch": 9.57, "learning_rate": 1.2795042897998093e-07, "loss": 1.3995, "step": 50263 }, { "epoch": 9.57, "learning_rate": 1.2789323164918972e-07, "loss": 2.8366, "step": 50264 }, { "epoch": 9.57, "learning_rate": 1.2783603431839847e-07, "loss": 2.0208, "step": 50265 }, { "epoch": 9.57, "learning_rate": 1.2777883698760726e-07, "loss": 1.3871, "step": 50266 }, { "epoch": 9.57, "learning_rate": 1.2772163965681603e-07, "loss": 2.1267, "step": 50267 }, { "epoch": 9.57, "learning_rate": 1.276644423260248e-07, "loss": 1.0736, "step": 50268 }, { "epoch": 9.58, "learning_rate": 1.2760724499523356e-07, "loss": 1.8623, "step": 50269 }, { "epoch": 9.58, "learning_rate": 1.2755004766444233e-07, "loss": 1.2674, "step": 50270 }, { "epoch": 9.58, "learning_rate": 1.274928503336511e-07, "loss": 0.9063, "step": 50271 }, { "epoch": 9.58, "learning_rate": 1.2743565300285986e-07, "loss": 2.6942, "step": 50272 }, { "epoch": 9.58, "learning_rate": 1.2737845567206866e-07, "loss": 2.0101, "step": 50273 }, { "epoch": 9.58, "learning_rate": 1.273212583412774e-07, "loss": 2.4455, "step": 50274 }, { "epoch": 9.58, "learning_rate": 1.272640610104862e-07, "loss": 1.411, "step": 50275 }, { "epoch": 9.58, "learning_rate": 1.2720686367969496e-07, "loss": 1.1762, "step": 50276 }, { "epoch": 9.58, "learning_rate": 1.2714966634890373e-07, "loss": 1.7491, "step": 50277 }, { "epoch": 9.58, "learning_rate": 1.270924690181125e-07, "loss": 0.3158, "step": 50278 }, { "epoch": 9.58, "learning_rate": 1.2703527168732126e-07, "loss": 0.9999, "step": 50279 }, { "epoch": 9.58, "learning_rate": 1.2697807435653003e-07, "loss": 1.5584, "step": 50280 }, { "epoch": 9.58, "learning_rate": 1.269208770257388e-07, "loss": 1.3119, "step": 50281 }, { "epoch": 9.58, "learning_rate": 1.268636796949476e-07, "loss": 2.9874, "step": 50282 }, { "epoch": 9.58, "learning_rate": 1.2680648236415633e-07, "loss": 1.0513, "step": 50283 }, { "epoch": 9.58, "learning_rate": 1.2674928503336512e-07, "loss": 1.2448, "step": 50284 }, { "epoch": 9.58, "learning_rate": 1.266920877025739e-07, "loss": 1.9976, "step": 50285 }, { "epoch": 9.58, "learning_rate": 1.2663489037178266e-07, "loss": 1.2557, "step": 50286 }, { "epoch": 9.58, "learning_rate": 1.2657769304099143e-07, "loss": 2.5376, "step": 50287 }, { "epoch": 9.58, "learning_rate": 1.265204957102002e-07, "loss": 0.7619, "step": 50288 }, { "epoch": 9.58, "learning_rate": 1.2646329837940896e-07, "loss": 1.7346, "step": 50289 }, { "epoch": 9.58, "learning_rate": 1.2640610104861773e-07, "loss": 1.8589, "step": 50290 }, { "epoch": 9.58, "learning_rate": 1.263489037178265e-07, "loss": 1.6517, "step": 50291 }, { "epoch": 9.58, "learning_rate": 1.2629170638703526e-07, "loss": 1.0642, "step": 50292 }, { "epoch": 9.58, "learning_rate": 1.2623450905624406e-07, "loss": 0.9053, "step": 50293 }, { "epoch": 9.58, "learning_rate": 1.261773117254528e-07, "loss": 1.4424, "step": 50294 }, { "epoch": 9.58, "learning_rate": 1.261201143946616e-07, "loss": 2.8587, "step": 50295 }, { "epoch": 9.58, "learning_rate": 1.2606291706387036e-07, "loss": 2.17, "step": 50296 }, { "epoch": 9.58, "learning_rate": 1.2600571973307913e-07, "loss": 0.7598, "step": 50297 }, { "epoch": 9.58, "learning_rate": 1.259485224022879e-07, "loss": 1.5658, "step": 50298 }, { "epoch": 9.58, "learning_rate": 1.2589132507149666e-07, "loss": 2.1455, "step": 50299 }, { "epoch": 9.58, "learning_rate": 1.2583412774070543e-07, "loss": 1.3325, "step": 50300 }, { "epoch": 9.58, "learning_rate": 1.257769304099142e-07, "loss": 1.9463, "step": 50301 }, { "epoch": 9.58, "learning_rate": 1.25719733079123e-07, "loss": 1.6674, "step": 50302 }, { "epoch": 9.58, "learning_rate": 1.2566253574833173e-07, "loss": 1.3359, "step": 50303 }, { "epoch": 9.58, "learning_rate": 1.2560533841754052e-07, "loss": 1.9856, "step": 50304 }, { "epoch": 9.58, "learning_rate": 1.255481410867493e-07, "loss": 1.9146, "step": 50305 }, { "epoch": 9.58, "learning_rate": 1.2549094375595806e-07, "loss": 2.09, "step": 50306 }, { "epoch": 9.58, "learning_rate": 1.2543374642516683e-07, "loss": 0.7736, "step": 50307 }, { "epoch": 9.58, "learning_rate": 1.253765490943756e-07, "loss": 0.6801, "step": 50308 }, { "epoch": 9.58, "learning_rate": 1.2531935176358436e-07, "loss": 1.5788, "step": 50309 }, { "epoch": 9.58, "learning_rate": 1.2526215443279313e-07, "loss": 1.3748, "step": 50310 }, { "epoch": 9.58, "learning_rate": 1.2520495710200192e-07, "loss": 1.6156, "step": 50311 }, { "epoch": 9.58, "learning_rate": 1.2514775977121066e-07, "loss": 2.2972, "step": 50312 }, { "epoch": 9.58, "learning_rate": 1.2509056244041946e-07, "loss": 1.7465, "step": 50313 }, { "epoch": 9.58, "learning_rate": 1.2503336510962822e-07, "loss": 2.0374, "step": 50314 }, { "epoch": 9.58, "learning_rate": 1.24976167778837e-07, "loss": 0.8927, "step": 50315 }, { "epoch": 9.58, "learning_rate": 1.2491897044804576e-07, "loss": 1.4983, "step": 50316 }, { "epoch": 9.58, "learning_rate": 1.2486177311725453e-07, "loss": 2.9909, "step": 50317 }, { "epoch": 9.58, "learning_rate": 1.248045757864633e-07, "loss": 1.5711, "step": 50318 }, { "epoch": 9.58, "learning_rate": 1.2474737845567206e-07, "loss": 2.5584, "step": 50319 }, { "epoch": 9.58, "learning_rate": 1.2469018112488085e-07, "loss": 0.9451, "step": 50320 }, { "epoch": 9.58, "learning_rate": 1.246329837940896e-07, "loss": 2.4899, "step": 50321 }, { "epoch": 9.59, "learning_rate": 1.245757864632984e-07, "loss": 1.5897, "step": 50322 }, { "epoch": 9.59, "learning_rate": 1.2451858913250716e-07, "loss": 1.4663, "step": 50323 }, { "epoch": 9.59, "learning_rate": 1.2446139180171592e-07, "loss": 2.2301, "step": 50324 }, { "epoch": 9.59, "learning_rate": 1.244041944709247e-07, "loss": 2.7269, "step": 50325 }, { "epoch": 9.59, "learning_rate": 1.2434699714013346e-07, "loss": 1.1629, "step": 50326 }, { "epoch": 9.59, "learning_rate": 1.2428979980934223e-07, "loss": 1.8254, "step": 50327 }, { "epoch": 9.59, "learning_rate": 1.24232602478551e-07, "loss": 1.9188, "step": 50328 }, { "epoch": 9.59, "learning_rate": 1.241754051477598e-07, "loss": 1.5894, "step": 50329 }, { "epoch": 9.59, "learning_rate": 1.2411820781696853e-07, "loss": 0.8521, "step": 50330 }, { "epoch": 9.59, "learning_rate": 1.2406101048617732e-07, "loss": 2.3121, "step": 50331 }, { "epoch": 9.59, "learning_rate": 1.240038131553861e-07, "loss": 1.2386, "step": 50332 }, { "epoch": 9.59, "learning_rate": 1.2394661582459486e-07, "loss": 1.4943, "step": 50333 }, { "epoch": 9.59, "learning_rate": 1.2388941849380362e-07, "loss": 1.4651, "step": 50334 }, { "epoch": 9.59, "learning_rate": 1.238322211630124e-07, "loss": 0.5515, "step": 50335 }, { "epoch": 9.59, "learning_rate": 1.2377502383222116e-07, "loss": 1.8117, "step": 50336 }, { "epoch": 9.59, "learning_rate": 1.2371782650142993e-07, "loss": 1.4389, "step": 50337 }, { "epoch": 9.59, "learning_rate": 1.2366062917063872e-07, "loss": 3.3628, "step": 50338 }, { "epoch": 9.59, "learning_rate": 1.2360343183984746e-07, "loss": 1.3038, "step": 50339 }, { "epoch": 9.59, "learning_rate": 1.2354623450905625e-07, "loss": 1.3738, "step": 50340 }, { "epoch": 9.59, "learning_rate": 1.2348903717826502e-07, "loss": 2.2868, "step": 50341 }, { "epoch": 9.59, "learning_rate": 1.234318398474738e-07, "loss": 1.9423, "step": 50342 }, { "epoch": 9.59, "learning_rate": 1.2337464251668256e-07, "loss": 1.1372, "step": 50343 }, { "epoch": 9.59, "learning_rate": 1.2331744518589135e-07, "loss": 1.5962, "step": 50344 }, { "epoch": 9.59, "learning_rate": 1.232602478551001e-07, "loss": 0.6327, "step": 50345 }, { "epoch": 9.59, "learning_rate": 1.2320305052430886e-07, "loss": 1.3865, "step": 50346 }, { "epoch": 9.59, "learning_rate": 1.2314585319351765e-07, "loss": 0.3701, "step": 50347 }, { "epoch": 9.59, "learning_rate": 1.230886558627264e-07, "loss": 0.7323, "step": 50348 }, { "epoch": 9.59, "learning_rate": 1.230314585319352e-07, "loss": 1.2392, "step": 50349 }, { "epoch": 9.59, "learning_rate": 1.2297426120114393e-07, "loss": 0.5751, "step": 50350 }, { "epoch": 9.59, "learning_rate": 1.2291706387035272e-07, "loss": 0.8867, "step": 50351 }, { "epoch": 9.59, "learning_rate": 1.228598665395615e-07, "loss": 1.5524, "step": 50352 }, { "epoch": 9.59, "learning_rate": 1.2280266920877026e-07, "loss": 1.585, "step": 50353 }, { "epoch": 9.59, "learning_rate": 1.2274547187797902e-07, "loss": 1.682, "step": 50354 }, { "epoch": 9.59, "learning_rate": 1.2268827454718782e-07, "loss": 1.0793, "step": 50355 }, { "epoch": 9.59, "learning_rate": 1.2263107721639656e-07, "loss": 1.5375, "step": 50356 }, { "epoch": 9.59, "learning_rate": 1.2257387988560533e-07, "loss": 2.3973, "step": 50357 }, { "epoch": 9.59, "learning_rate": 1.2251668255481412e-07, "loss": 2.3952, "step": 50358 }, { "epoch": 9.59, "learning_rate": 1.2245948522402286e-07, "loss": 2.1467, "step": 50359 }, { "epoch": 9.59, "learning_rate": 1.2240228789323165e-07, "loss": 2.0879, "step": 50360 }, { "epoch": 9.59, "learning_rate": 1.2234509056244042e-07, "loss": 0.9395, "step": 50361 }, { "epoch": 9.59, "learning_rate": 1.222878932316492e-07, "loss": 2.4502, "step": 50362 }, { "epoch": 9.59, "learning_rate": 1.2223069590085796e-07, "loss": 0.3754, "step": 50363 }, { "epoch": 9.59, "learning_rate": 1.2217349857006675e-07, "loss": 2.1421, "step": 50364 }, { "epoch": 9.59, "learning_rate": 1.221163012392755e-07, "loss": 0.9559, "step": 50365 }, { "epoch": 9.59, "learning_rate": 1.2205910390848429e-07, "loss": 0.3982, "step": 50366 }, { "epoch": 9.59, "learning_rate": 1.2200190657769305e-07, "loss": 3.3049, "step": 50367 }, { "epoch": 9.59, "learning_rate": 1.219447092469018e-07, "loss": 1.4761, "step": 50368 }, { "epoch": 9.59, "learning_rate": 1.218875119161106e-07, "loss": 1.4887, "step": 50369 }, { "epoch": 9.59, "learning_rate": 1.2183031458531935e-07, "loss": 1.3605, "step": 50370 }, { "epoch": 9.59, "learning_rate": 1.2177311725452812e-07, "loss": 1.7035, "step": 50371 }, { "epoch": 9.59, "learning_rate": 1.217159199237369e-07, "loss": 1.7341, "step": 50372 }, { "epoch": 9.59, "learning_rate": 1.2165872259294568e-07, "loss": 1.749, "step": 50373 }, { "epoch": 9.6, "learning_rate": 1.2160152526215442e-07, "loss": 2.5486, "step": 50374 }, { "epoch": 9.6, "learning_rate": 1.2154432793136322e-07, "loss": 0.8126, "step": 50375 }, { "epoch": 9.6, "learning_rate": 1.2148713060057199e-07, "loss": 2.5072, "step": 50376 }, { "epoch": 9.6, "learning_rate": 1.2142993326978073e-07, "loss": 1.1664, "step": 50377 }, { "epoch": 9.6, "learning_rate": 1.2137273593898952e-07, "loss": 0.9938, "step": 50378 }, { "epoch": 9.6, "learning_rate": 1.213155386081983e-07, "loss": 1.3364, "step": 50379 }, { "epoch": 9.6, "learning_rate": 1.2125834127740705e-07, "loss": 2.5441, "step": 50380 }, { "epoch": 9.6, "learning_rate": 1.2120114394661582e-07, "loss": 1.6777, "step": 50381 }, { "epoch": 9.6, "learning_rate": 1.2114394661582462e-07, "loss": 1.2286, "step": 50382 }, { "epoch": 9.6, "learning_rate": 1.2108674928503336e-07, "loss": 2.4462, "step": 50383 }, { "epoch": 9.6, "learning_rate": 1.2102955195424215e-07, "loss": 0.8357, "step": 50384 }, { "epoch": 9.6, "learning_rate": 1.2097235462345092e-07, "loss": 2.7882, "step": 50385 }, { "epoch": 9.6, "learning_rate": 1.2091515729265969e-07, "loss": 0.9582, "step": 50386 }, { "epoch": 9.6, "learning_rate": 1.2085795996186845e-07, "loss": 0.3319, "step": 50387 }, { "epoch": 9.6, "learning_rate": 1.2080076263107722e-07, "loss": 0.6946, "step": 50388 }, { "epoch": 9.6, "learning_rate": 1.20743565300286e-07, "loss": 1.1986, "step": 50389 }, { "epoch": 9.6, "learning_rate": 1.2068636796949475e-07, "loss": 1.3562, "step": 50390 }, { "epoch": 9.6, "learning_rate": 1.2062917063870355e-07, "loss": 1.8216, "step": 50391 }, { "epoch": 9.6, "learning_rate": 1.205719733079123e-07, "loss": 0.6161, "step": 50392 }, { "epoch": 9.6, "learning_rate": 1.2051477597712108e-07, "loss": 2.4522, "step": 50393 }, { "epoch": 9.6, "learning_rate": 1.2045757864632985e-07, "loss": 1.859, "step": 50394 }, { "epoch": 9.6, "learning_rate": 1.2040038131553862e-07, "loss": 1.6686, "step": 50395 }, { "epoch": 9.6, "learning_rate": 1.2034318398474739e-07, "loss": 3.0959, "step": 50396 }, { "epoch": 9.6, "learning_rate": 1.2028598665395615e-07, "loss": 0.7995, "step": 50397 }, { "epoch": 9.6, "learning_rate": 1.2022878932316492e-07, "loss": 0.9373, "step": 50398 }, { "epoch": 9.6, "learning_rate": 1.201715919923737e-07, "loss": 1.0241, "step": 50399 }, { "epoch": 9.6, "learning_rate": 1.2011439466158248e-07, "loss": 1.726, "step": 50400 }, { "epoch": 9.6, "learning_rate": 1.2005719733079122e-07, "loss": 1.9246, "step": 50401 }, { "epoch": 9.6, "learning_rate": 1.2000000000000002e-07, "loss": 1.0105, "step": 50402 }, { "epoch": 9.6, "learning_rate": 1.1994280266920878e-07, "loss": 2.6025, "step": 50403 }, { "epoch": 9.6, "learning_rate": 1.1988560533841755e-07, "loss": 3.131, "step": 50404 }, { "epoch": 9.6, "learning_rate": 1.1982840800762632e-07, "loss": 2.0771, "step": 50405 }, { "epoch": 9.6, "learning_rate": 1.1977121067683509e-07, "loss": 1.6819, "step": 50406 }, { "epoch": 9.6, "learning_rate": 1.1971401334604385e-07, "loss": 1.6731, "step": 50407 }, { "epoch": 9.6, "learning_rate": 1.1965681601525262e-07, "loss": 1.7594, "step": 50408 }, { "epoch": 9.6, "learning_rate": 1.1959961868446141e-07, "loss": 1.4479, "step": 50409 }, { "epoch": 9.6, "learning_rate": 1.1954242135367015e-07, "loss": 2.0496, "step": 50410 }, { "epoch": 9.6, "learning_rate": 1.1948522402287895e-07, "loss": 0.5904, "step": 50411 }, { "epoch": 9.6, "learning_rate": 1.194280266920877e-07, "loss": 1.0499, "step": 50412 }, { "epoch": 9.6, "learning_rate": 1.1937082936129648e-07, "loss": 1.6472, "step": 50413 }, { "epoch": 9.6, "learning_rate": 1.1931363203050525e-07, "loss": 1.0625, "step": 50414 }, { "epoch": 9.6, "learning_rate": 1.1925643469971402e-07, "loss": 1.3611, "step": 50415 }, { "epoch": 9.6, "learning_rate": 1.1919923736892279e-07, "loss": 1.0778, "step": 50416 }, { "epoch": 9.6, "learning_rate": 1.1914204003813157e-07, "loss": 2.6113, "step": 50417 }, { "epoch": 9.6, "learning_rate": 1.1908484270734032e-07, "loss": 0.4972, "step": 50418 }, { "epoch": 9.6, "learning_rate": 1.190276453765491e-07, "loss": 2.8378, "step": 50419 }, { "epoch": 9.6, "learning_rate": 1.1897044804575787e-07, "loss": 1.9998, "step": 50420 }, { "epoch": 9.6, "learning_rate": 1.1891325071496664e-07, "loss": 2.1944, "step": 50421 }, { "epoch": 9.6, "learning_rate": 1.188560533841754e-07, "loss": 1.0643, "step": 50422 }, { "epoch": 9.6, "learning_rate": 1.1879885605338418e-07, "loss": 0.8511, "step": 50423 }, { "epoch": 9.6, "learning_rate": 1.1874165872259294e-07, "loss": 0.3084, "step": 50424 }, { "epoch": 9.6, "learning_rate": 1.1868446139180172e-07, "loss": 1.1348, "step": 50425 }, { "epoch": 9.6, "learning_rate": 1.186272640610105e-07, "loss": 0.842, "step": 50426 }, { "epoch": 9.61, "learning_rate": 1.1857006673021925e-07, "loss": 1.4727, "step": 50427 }, { "epoch": 9.61, "learning_rate": 1.1851286939942803e-07, "loss": 0.8283, "step": 50428 }, { "epoch": 9.61, "learning_rate": 1.184556720686368e-07, "loss": 0.7943, "step": 50429 }, { "epoch": 9.61, "learning_rate": 1.1839847473784557e-07, "loss": 3.2253, "step": 50430 }, { "epoch": 9.61, "learning_rate": 1.1834127740705434e-07, "loss": 1.6046, "step": 50431 }, { "epoch": 9.61, "learning_rate": 1.1828408007626312e-07, "loss": 2.9791, "step": 50432 }, { "epoch": 9.61, "learning_rate": 1.1822688274547187e-07, "loss": 2.0902, "step": 50433 }, { "epoch": 9.61, "learning_rate": 1.1816968541468065e-07, "loss": 1.8419, "step": 50434 }, { "epoch": 9.61, "learning_rate": 1.1811248808388943e-07, "loss": 3.2425, "step": 50435 }, { "epoch": 9.61, "learning_rate": 1.1805529075309818e-07, "loss": 0.548, "step": 50436 }, { "epoch": 9.61, "learning_rate": 1.1799809342230697e-07, "loss": 1.8304, "step": 50437 }, { "epoch": 9.61, "learning_rate": 1.1794089609151573e-07, "loss": 1.6493, "step": 50438 }, { "epoch": 9.61, "learning_rate": 1.178836987607245e-07, "loss": 0.3541, "step": 50439 }, { "epoch": 9.61, "learning_rate": 1.1782650142993327e-07, "loss": 0.1543, "step": 50440 }, { "epoch": 9.61, "learning_rate": 1.1776930409914205e-07, "loss": 1.3745, "step": 50441 }, { "epoch": 9.61, "learning_rate": 1.177121067683508e-07, "loss": 1.1897, "step": 50442 }, { "epoch": 9.61, "learning_rate": 1.1765490943755958e-07, "loss": 2.51, "step": 50443 }, { "epoch": 9.61, "learning_rate": 1.1759771210676836e-07, "loss": 1.5683, "step": 50444 }, { "epoch": 9.61, "learning_rate": 1.1754051477597712e-07, "loss": 1.9926, "step": 50445 }, { "epoch": 9.61, "learning_rate": 1.174833174451859e-07, "loss": 0.4107, "step": 50446 }, { "epoch": 9.61, "learning_rate": 1.1742612011439468e-07, "loss": 1.7528, "step": 50447 }, { "epoch": 9.61, "learning_rate": 1.1736892278360343e-07, "loss": 1.6608, "step": 50448 }, { "epoch": 9.61, "learning_rate": 1.173117254528122e-07, "loss": 0.8823, "step": 50449 }, { "epoch": 9.61, "learning_rate": 1.1725452812202098e-07, "loss": 1.6579, "step": 50450 }, { "epoch": 9.61, "learning_rate": 1.1719733079122973e-07, "loss": 1.675, "step": 50451 }, { "epoch": 9.61, "learning_rate": 1.1714013346043852e-07, "loss": 2.3075, "step": 50452 }, { "epoch": 9.61, "learning_rate": 1.170829361296473e-07, "loss": 0.958, "step": 50453 }, { "epoch": 9.61, "learning_rate": 1.1702573879885605e-07, "loss": 1.3598, "step": 50454 }, { "epoch": 9.61, "learning_rate": 1.1696854146806483e-07, "loss": 0.7414, "step": 50455 }, { "epoch": 9.61, "learning_rate": 1.1691134413727361e-07, "loss": 1.827, "step": 50456 }, { "epoch": 9.61, "learning_rate": 1.1685414680648237e-07, "loss": 0.3938, "step": 50457 }, { "epoch": 9.61, "learning_rate": 1.1679694947569113e-07, "loss": 0.4839, "step": 50458 }, { "epoch": 9.61, "learning_rate": 1.1673975214489991e-07, "loss": 2.5105, "step": 50459 }, { "epoch": 9.61, "learning_rate": 1.1668255481410867e-07, "loss": 0.8906, "step": 50460 }, { "epoch": 9.61, "learning_rate": 1.1662535748331745e-07, "loss": 1.8712, "step": 50461 }, { "epoch": 9.61, "learning_rate": 1.1656816015252623e-07, "loss": 1.677, "step": 50462 }, { "epoch": 9.61, "learning_rate": 1.1651096282173498e-07, "loss": 2.7073, "step": 50463 }, { "epoch": 9.61, "learning_rate": 1.1645376549094376e-07, "loss": 2.0935, "step": 50464 }, { "epoch": 9.61, "learning_rate": 1.1639656816015254e-07, "loss": 1.754, "step": 50465 }, { "epoch": 9.61, "learning_rate": 1.163393708293613e-07, "loss": 0.3325, "step": 50466 }, { "epoch": 9.61, "learning_rate": 1.1628217349857008e-07, "loss": 0.6657, "step": 50467 }, { "epoch": 9.61, "learning_rate": 1.1622497616777885e-07, "loss": 2.4926, "step": 50468 }, { "epoch": 9.61, "learning_rate": 1.161677788369876e-07, "loss": 2.1048, "step": 50469 }, { "epoch": 9.61, "learning_rate": 1.1611058150619638e-07, "loss": 1.3857, "step": 50470 }, { "epoch": 9.61, "learning_rate": 1.1605338417540516e-07, "loss": 1.4578, "step": 50471 }, { "epoch": 9.61, "learning_rate": 1.1599618684461392e-07, "loss": 3.2617, "step": 50472 }, { "epoch": 9.61, "learning_rate": 1.159389895138227e-07, "loss": 1.9163, "step": 50473 }, { "epoch": 9.61, "learning_rate": 1.1588179218303145e-07, "loss": 1.3053, "step": 50474 }, { "epoch": 9.61, "learning_rate": 1.1582459485224023e-07, "loss": 0.6681, "step": 50475 }, { "epoch": 9.61, "learning_rate": 1.1576739752144901e-07, "loss": 1.169, "step": 50476 }, { "epoch": 9.61, "learning_rate": 1.1571020019065777e-07, "loss": 1.3754, "step": 50477 }, { "epoch": 9.61, "learning_rate": 1.1565300285986655e-07, "loss": 0.2973, "step": 50478 }, { "epoch": 9.62, "learning_rate": 1.1559580552907531e-07, "loss": 1.3806, "step": 50479 }, { "epoch": 9.62, "learning_rate": 1.1553860819828407e-07, "loss": 0.5702, "step": 50480 }, { "epoch": 9.62, "learning_rate": 1.1548141086749285e-07, "loss": 1.0433, "step": 50481 }, { "epoch": 9.62, "learning_rate": 1.1542421353670163e-07, "loss": 1.511, "step": 50482 }, { "epoch": 9.62, "learning_rate": 1.1536701620591038e-07, "loss": 1.606, "step": 50483 }, { "epoch": 9.62, "learning_rate": 1.1530981887511916e-07, "loss": 2.0379, "step": 50484 }, { "epoch": 9.62, "learning_rate": 1.1525262154432794e-07, "loss": 1.2348, "step": 50485 }, { "epoch": 9.62, "learning_rate": 1.151954242135367e-07, "loss": 1.4281, "step": 50486 }, { "epoch": 9.62, "learning_rate": 1.1513822688274548e-07, "loss": 1.5617, "step": 50487 }, { "epoch": 9.62, "learning_rate": 1.1508102955195425e-07, "loss": 0.7851, "step": 50488 }, { "epoch": 9.62, "learning_rate": 1.1502383222116301e-07, "loss": 1.2353, "step": 50489 }, { "epoch": 9.62, "learning_rate": 1.1496663489037178e-07, "loss": 2.2779, "step": 50490 }, { "epoch": 9.62, "learning_rate": 1.1490943755958056e-07, "loss": 1.1589, "step": 50491 }, { "epoch": 9.62, "learning_rate": 1.1485224022878932e-07, "loss": 1.5595, "step": 50492 }, { "epoch": 9.62, "learning_rate": 1.147950428979981e-07, "loss": 2.0911, "step": 50493 }, { "epoch": 9.62, "learning_rate": 1.1473784556720688e-07, "loss": 2.4169, "step": 50494 }, { "epoch": 9.62, "learning_rate": 1.1468064823641563e-07, "loss": 4.1819, "step": 50495 }, { "epoch": 9.62, "learning_rate": 1.1462345090562441e-07, "loss": 1.7669, "step": 50496 }, { "epoch": 9.62, "learning_rate": 1.1456625357483318e-07, "loss": 2.4538, "step": 50497 }, { "epoch": 9.62, "learning_rate": 1.1450905624404195e-07, "loss": 1.3837, "step": 50498 }, { "epoch": 9.62, "learning_rate": 1.1445185891325071e-07, "loss": 1.1806, "step": 50499 }, { "epoch": 9.62, "learning_rate": 1.143946615824595e-07, "loss": 0.9082, "step": 50500 }, { "epoch": 9.62, "learning_rate": 1.1433746425166825e-07, "loss": 1.5185, "step": 50501 }, { "epoch": 9.62, "learning_rate": 1.1428026692087703e-07, "loss": 1.6055, "step": 50502 }, { "epoch": 9.62, "learning_rate": 1.1422306959008581e-07, "loss": 0.8036, "step": 50503 }, { "epoch": 9.62, "learning_rate": 1.1416587225929456e-07, "loss": 0.8419, "step": 50504 }, { "epoch": 9.62, "learning_rate": 1.1410867492850334e-07, "loss": 1.4829, "step": 50505 }, { "epoch": 9.62, "learning_rate": 1.1405147759771211e-07, "loss": 1.509, "step": 50506 }, { "epoch": 9.62, "learning_rate": 1.1399428026692088e-07, "loss": 1.3431, "step": 50507 }, { "epoch": 9.62, "learning_rate": 1.1393708293612965e-07, "loss": 0.8364, "step": 50508 }, { "epoch": 9.62, "learning_rate": 1.1387988560533843e-07, "loss": 1.6681, "step": 50509 }, { "epoch": 9.62, "learning_rate": 1.1382268827454718e-07, "loss": 2.2184, "step": 50510 }, { "epoch": 9.62, "learning_rate": 1.1376549094375596e-07, "loss": 1.1305, "step": 50511 }, { "epoch": 9.62, "learning_rate": 1.1370829361296474e-07, "loss": 1.6082, "step": 50512 }, { "epoch": 9.62, "learning_rate": 1.136510962821735e-07, "loss": 1.2926, "step": 50513 }, { "epoch": 9.62, "learning_rate": 1.1359389895138228e-07, "loss": 2.2362, "step": 50514 }, { "epoch": 9.62, "learning_rate": 1.1353670162059106e-07, "loss": 0.6752, "step": 50515 }, { "epoch": 9.62, "learning_rate": 1.1347950428979981e-07, "loss": 1.6671, "step": 50516 }, { "epoch": 9.62, "learning_rate": 1.1342230695900858e-07, "loss": 0.9497, "step": 50517 }, { "epoch": 9.62, "learning_rate": 1.1336510962821736e-07, "loss": 3.0264, "step": 50518 }, { "epoch": 9.62, "learning_rate": 1.1330791229742611e-07, "loss": 2.6393, "step": 50519 }, { "epoch": 9.62, "learning_rate": 1.132507149666349e-07, "loss": 2.0709, "step": 50520 }, { "epoch": 9.62, "learning_rate": 1.1319351763584367e-07, "loss": 1.2409, "step": 50521 }, { "epoch": 9.62, "learning_rate": 1.1313632030505243e-07, "loss": 1.5292, "step": 50522 }, { "epoch": 9.62, "learning_rate": 1.1307912297426121e-07, "loss": 1.9838, "step": 50523 }, { "epoch": 9.62, "learning_rate": 1.1302192564346999e-07, "loss": 1.5383, "step": 50524 }, { "epoch": 9.62, "learning_rate": 1.1296472831267874e-07, "loss": 1.5313, "step": 50525 }, { "epoch": 9.62, "learning_rate": 1.1290753098188751e-07, "loss": 1.7469, "step": 50526 }, { "epoch": 9.62, "learning_rate": 1.1285033365109629e-07, "loss": 2.7786, "step": 50527 }, { "epoch": 9.62, "learning_rate": 1.1279313632030505e-07, "loss": 1.9398, "step": 50528 }, { "epoch": 9.62, "learning_rate": 1.1273593898951383e-07, "loss": 1.0821, "step": 50529 }, { "epoch": 9.62, "learning_rate": 1.1267874165872261e-07, "loss": 3.7311, "step": 50530 }, { "epoch": 9.62, "learning_rate": 1.1262154432793136e-07, "loss": 1.2192, "step": 50531 }, { "epoch": 9.63, "learning_rate": 1.1256434699714014e-07, "loss": 1.0368, "step": 50532 }, { "epoch": 9.63, "learning_rate": 1.125071496663489e-07, "loss": 1.7971, "step": 50533 }, { "epoch": 9.63, "learning_rate": 1.1244995233555768e-07, "loss": 0.7862, "step": 50534 }, { "epoch": 9.63, "learning_rate": 1.1239275500476646e-07, "loss": 1.064, "step": 50535 }, { "epoch": 9.63, "learning_rate": 1.1233555767397521e-07, "loss": 1.7325, "step": 50536 }, { "epoch": 9.63, "learning_rate": 1.1227836034318398e-07, "loss": 1.8314, "step": 50537 }, { "epoch": 9.63, "learning_rate": 1.1222116301239276e-07, "loss": 1.6367, "step": 50538 }, { "epoch": 9.63, "learning_rate": 1.1216396568160151e-07, "loss": 1.978, "step": 50539 }, { "epoch": 9.63, "learning_rate": 1.121067683508103e-07, "loss": 1.1073, "step": 50540 }, { "epoch": 9.63, "learning_rate": 1.1204957102001907e-07, "loss": 1.1716, "step": 50541 }, { "epoch": 9.63, "learning_rate": 1.1199237368922783e-07, "loss": 1.779, "step": 50542 }, { "epoch": 9.63, "learning_rate": 1.1193517635843661e-07, "loss": 2.8274, "step": 50543 }, { "epoch": 9.63, "learning_rate": 1.1187797902764539e-07, "loss": 1.6016, "step": 50544 }, { "epoch": 9.63, "learning_rate": 1.1182078169685414e-07, "loss": 1.2326, "step": 50545 }, { "epoch": 9.63, "learning_rate": 1.1176358436606292e-07, "loss": 0.6785, "step": 50546 }, { "epoch": 9.63, "learning_rate": 1.1170638703527169e-07, "loss": 1.133, "step": 50547 }, { "epoch": 9.63, "learning_rate": 1.1164918970448045e-07, "loss": 1.7657, "step": 50548 }, { "epoch": 9.63, "learning_rate": 1.1159199237368923e-07, "loss": 0.277, "step": 50549 }, { "epoch": 9.63, "learning_rate": 1.1153479504289801e-07, "loss": 0.2579, "step": 50550 }, { "epoch": 9.63, "learning_rate": 1.1147759771210676e-07, "loss": 2.0404, "step": 50551 }, { "epoch": 9.63, "learning_rate": 1.1142040038131554e-07, "loss": 0.6875, "step": 50552 }, { "epoch": 9.63, "learning_rate": 1.1136320305052432e-07, "loss": 3.6483, "step": 50553 }, { "epoch": 9.63, "learning_rate": 1.1130600571973308e-07, "loss": 2.7711, "step": 50554 }, { "epoch": 9.63, "learning_rate": 1.1124880838894186e-07, "loss": 1.2695, "step": 50555 }, { "epoch": 9.63, "learning_rate": 1.1119161105815062e-07, "loss": 1.4615, "step": 50556 }, { "epoch": 9.63, "learning_rate": 1.1113441372735939e-07, "loss": 1.456, "step": 50557 }, { "epoch": 9.63, "learning_rate": 1.1107721639656816e-07, "loss": 0.5665, "step": 50558 }, { "epoch": 9.63, "learning_rate": 1.1102001906577694e-07, "loss": 0.5848, "step": 50559 }, { "epoch": 9.63, "learning_rate": 1.109628217349857e-07, "loss": 1.1403, "step": 50560 }, { "epoch": 9.63, "learning_rate": 1.1090562440419447e-07, "loss": 0.923, "step": 50561 }, { "epoch": 9.63, "learning_rate": 1.1084842707340325e-07, "loss": 2.9427, "step": 50562 }, { "epoch": 9.63, "learning_rate": 1.1079122974261201e-07, "loss": 1.8724, "step": 50563 }, { "epoch": 9.63, "learning_rate": 1.1073403241182079e-07, "loss": 2.5779, "step": 50564 }, { "epoch": 9.63, "learning_rate": 1.1067683508102956e-07, "loss": 2.1843, "step": 50565 }, { "epoch": 9.63, "learning_rate": 1.1061963775023832e-07, "loss": 1.0733, "step": 50566 }, { "epoch": 9.63, "learning_rate": 1.1056244041944709e-07, "loss": 0.3078, "step": 50567 }, { "epoch": 9.63, "learning_rate": 1.1050524308865587e-07, "loss": 2.939, "step": 50568 }, { "epoch": 9.63, "learning_rate": 1.1044804575786463e-07, "loss": 1.8233, "step": 50569 }, { "epoch": 9.63, "learning_rate": 1.1039084842707341e-07, "loss": 1.3479, "step": 50570 }, { "epoch": 9.63, "learning_rate": 1.1033365109628219e-07, "loss": 3.4307, "step": 50571 }, { "epoch": 9.63, "learning_rate": 1.1027645376549094e-07, "loss": 1.2631, "step": 50572 }, { "epoch": 9.63, "learning_rate": 1.1021925643469972e-07, "loss": 1.4263, "step": 50573 }, { "epoch": 9.63, "learning_rate": 1.1016205910390849e-07, "loss": 1.0277, "step": 50574 }, { "epoch": 9.63, "learning_rate": 1.1010486177311726e-07, "loss": 0.6679, "step": 50575 }, { "epoch": 9.63, "learning_rate": 1.1004766444232602e-07, "loss": 1.9769, "step": 50576 }, { "epoch": 9.63, "learning_rate": 1.099904671115348e-07, "loss": 0.7629, "step": 50577 }, { "epoch": 9.63, "learning_rate": 1.0993326978074356e-07, "loss": 1.9683, "step": 50578 }, { "epoch": 9.63, "learning_rate": 1.0987607244995234e-07, "loss": 1.3776, "step": 50579 }, { "epoch": 9.63, "learning_rate": 1.0981887511916112e-07, "loss": 2.1796, "step": 50580 }, { "epoch": 9.63, "learning_rate": 1.0976167778836987e-07, "loss": 0.3917, "step": 50581 }, { "epoch": 9.63, "learning_rate": 1.0970448045757865e-07, "loss": 1.8542, "step": 50582 }, { "epoch": 9.63, "learning_rate": 1.0964728312678744e-07, "loss": 2.1177, "step": 50583 }, { "epoch": 9.64, "learning_rate": 1.0959008579599619e-07, "loss": 0.65, "step": 50584 }, { "epoch": 9.64, "learning_rate": 1.0953288846520496e-07, "loss": 1.367, "step": 50585 }, { "epoch": 9.64, "learning_rate": 1.0947569113441374e-07, "loss": 1.6127, "step": 50586 }, { "epoch": 9.64, "learning_rate": 1.0941849380362249e-07, "loss": 0.2681, "step": 50587 }, { "epoch": 9.64, "learning_rate": 1.0936129647283127e-07, "loss": 2.2349, "step": 50588 }, { "epoch": 9.64, "learning_rate": 1.0930409914204005e-07, "loss": 2.4866, "step": 50589 }, { "epoch": 9.64, "learning_rate": 1.092469018112488e-07, "loss": 1.6124, "step": 50590 }, { "epoch": 9.64, "learning_rate": 1.0918970448045759e-07, "loss": 1.3772, "step": 50591 }, { "epoch": 9.64, "learning_rate": 1.0913250714966637e-07, "loss": 1.7138, "step": 50592 }, { "epoch": 9.64, "learning_rate": 1.0907530981887512e-07, "loss": 0.909, "step": 50593 }, { "epoch": 9.64, "learning_rate": 1.090181124880839e-07, "loss": 1.1331, "step": 50594 }, { "epoch": 9.64, "learning_rate": 1.0896091515729266e-07, "loss": 1.7662, "step": 50595 }, { "epoch": 9.64, "learning_rate": 1.0890371782650142e-07, "loss": 2.0432, "step": 50596 }, { "epoch": 9.64, "learning_rate": 1.088465204957102e-07, "loss": 1.9652, "step": 50597 }, { "epoch": 9.64, "learning_rate": 1.0878932316491896e-07, "loss": 1.396, "step": 50598 }, { "epoch": 9.64, "learning_rate": 1.0873212583412774e-07, "loss": 0.2051, "step": 50599 }, { "epoch": 9.64, "learning_rate": 1.0867492850333652e-07, "loss": 1.8223, "step": 50600 }, { "epoch": 9.64, "learning_rate": 1.0861773117254527e-07, "loss": 1.4228, "step": 50601 }, { "epoch": 9.64, "learning_rate": 1.0856053384175405e-07, "loss": 1.3526, "step": 50602 }, { "epoch": 9.64, "learning_rate": 1.0850333651096284e-07, "loss": 1.7136, "step": 50603 }, { "epoch": 9.64, "learning_rate": 1.0844613918017159e-07, "loss": 1.063, "step": 50604 }, { "epoch": 9.64, "learning_rate": 1.0838894184938036e-07, "loss": 1.9898, "step": 50605 }, { "epoch": 9.64, "learning_rate": 1.0833174451858914e-07, "loss": 0.8759, "step": 50606 }, { "epoch": 9.64, "learning_rate": 1.0827454718779789e-07, "loss": 1.1094, "step": 50607 }, { "epoch": 9.64, "learning_rate": 1.0821734985700667e-07, "loss": 2.5883, "step": 50608 }, { "epoch": 9.64, "learning_rate": 1.0816015252621545e-07, "loss": 1.2769, "step": 50609 }, { "epoch": 9.64, "learning_rate": 1.081029551954242e-07, "loss": 0.6698, "step": 50610 }, { "epoch": 9.64, "learning_rate": 1.0804575786463299e-07, "loss": 1.1393, "step": 50611 }, { "epoch": 9.64, "learning_rate": 1.0798856053384177e-07, "loss": 1.0523, "step": 50612 }, { "epoch": 9.64, "learning_rate": 1.0793136320305052e-07, "loss": 1.5142, "step": 50613 }, { "epoch": 9.64, "learning_rate": 1.078741658722593e-07, "loss": 1.4806, "step": 50614 }, { "epoch": 9.64, "learning_rate": 1.0781696854146807e-07, "loss": 0.8939, "step": 50615 }, { "epoch": 9.64, "learning_rate": 1.0775977121067682e-07, "loss": 2.1264, "step": 50616 }, { "epoch": 9.64, "learning_rate": 1.077025738798856e-07, "loss": 0.4574, "step": 50617 }, { "epoch": 9.64, "learning_rate": 1.0764537654909438e-07, "loss": 1.1063, "step": 50618 }, { "epoch": 9.64, "learning_rate": 1.0758817921830314e-07, "loss": 2.3096, "step": 50619 }, { "epoch": 9.64, "learning_rate": 1.0753098188751192e-07, "loss": 0.9941, "step": 50620 }, { "epoch": 9.64, "learning_rate": 1.074737845567207e-07, "loss": 2.6103, "step": 50621 }, { "epoch": 9.64, "learning_rate": 1.0741658722592945e-07, "loss": 1.9594, "step": 50622 }, { "epoch": 9.64, "learning_rate": 1.0735938989513823e-07, "loss": 1.5895, "step": 50623 }, { "epoch": 9.64, "learning_rate": 1.07302192564347e-07, "loss": 2.2097, "step": 50624 }, { "epoch": 9.64, "learning_rate": 1.0724499523355577e-07, "loss": 1.8093, "step": 50625 }, { "epoch": 9.64, "learning_rate": 1.0718779790276454e-07, "loss": 2.1506, "step": 50626 }, { "epoch": 9.64, "learning_rate": 1.0713060057197332e-07, "loss": 0.9788, "step": 50627 }, { "epoch": 9.64, "learning_rate": 1.0707340324118207e-07, "loss": 0.9004, "step": 50628 }, { "epoch": 9.64, "learning_rate": 1.0701620591039085e-07, "loss": 0.7832, "step": 50629 }, { "epoch": 9.64, "learning_rate": 1.0695900857959963e-07, "loss": 1.2623, "step": 50630 }, { "epoch": 9.64, "learning_rate": 1.0690181124880839e-07, "loss": 1.6586, "step": 50631 }, { "epoch": 9.64, "learning_rate": 1.0684461391801717e-07, "loss": 1.2742, "step": 50632 }, { "epoch": 9.64, "learning_rate": 1.0678741658722593e-07, "loss": 3.1338, "step": 50633 }, { "epoch": 9.64, "learning_rate": 1.067302192564347e-07, "loss": 2.0779, "step": 50634 }, { "epoch": 9.64, "learning_rate": 1.0667302192564347e-07, "loss": 0.8832, "step": 50635 }, { "epoch": 9.64, "learning_rate": 1.0661582459485225e-07, "loss": 3.4032, "step": 50636 }, { "epoch": 9.65, "learning_rate": 1.06558627264061e-07, "loss": 2.382, "step": 50637 }, { "epoch": 9.65, "learning_rate": 1.0650142993326978e-07, "loss": 0.5186, "step": 50638 }, { "epoch": 9.65, "learning_rate": 1.0644423260247857e-07, "loss": 2.3077, "step": 50639 }, { "epoch": 9.65, "learning_rate": 1.0638703527168732e-07, "loss": 2.785, "step": 50640 }, { "epoch": 9.65, "learning_rate": 1.063298379408961e-07, "loss": 2.2834, "step": 50641 }, { "epoch": 9.65, "learning_rate": 1.0627264061010487e-07, "loss": 1.2912, "step": 50642 }, { "epoch": 9.65, "learning_rate": 1.0621544327931363e-07, "loss": 1.8231, "step": 50643 }, { "epoch": 9.65, "learning_rate": 1.061582459485224e-07, "loss": 2.0711, "step": 50644 }, { "epoch": 9.65, "learning_rate": 1.0610104861773118e-07, "loss": 0.5238, "step": 50645 }, { "epoch": 9.65, "learning_rate": 1.0604385128693994e-07, "loss": 0.353, "step": 50646 }, { "epoch": 9.65, "learning_rate": 1.0598665395614872e-07, "loss": 1.8089, "step": 50647 }, { "epoch": 9.65, "learning_rate": 1.059294566253575e-07, "loss": 1.5425, "step": 50648 }, { "epoch": 9.65, "learning_rate": 1.0587225929456625e-07, "loss": 1.6385, "step": 50649 }, { "epoch": 9.65, "learning_rate": 1.0581506196377503e-07, "loss": 1.4235, "step": 50650 }, { "epoch": 9.65, "learning_rate": 1.0575786463298381e-07, "loss": 1.6845, "step": 50651 }, { "epoch": 9.65, "learning_rate": 1.0570066730219257e-07, "loss": 1.8306, "step": 50652 }, { "epoch": 9.65, "learning_rate": 1.0564346997140133e-07, "loss": 1.3966, "step": 50653 }, { "epoch": 9.65, "learning_rate": 1.055862726406101e-07, "loss": 1.0572, "step": 50654 }, { "epoch": 9.65, "learning_rate": 1.0552907530981887e-07, "loss": 0.6832, "step": 50655 }, { "epoch": 9.65, "learning_rate": 1.0547187797902765e-07, "loss": 0.5731, "step": 50656 }, { "epoch": 9.65, "learning_rate": 1.054146806482364e-07, "loss": 1.1742, "step": 50657 }, { "epoch": 9.65, "learning_rate": 1.0535748331744518e-07, "loss": 0.8916, "step": 50658 }, { "epoch": 9.65, "learning_rate": 1.0530028598665397e-07, "loss": 1.8309, "step": 50659 }, { "epoch": 9.65, "learning_rate": 1.0524308865586272e-07, "loss": 1.6487, "step": 50660 }, { "epoch": 9.65, "learning_rate": 1.051858913250715e-07, "loss": 2.8128, "step": 50661 }, { "epoch": 9.65, "learning_rate": 1.0512869399428028e-07, "loss": 2.6691, "step": 50662 }, { "epoch": 9.65, "learning_rate": 1.0507149666348903e-07, "loss": 1.4163, "step": 50663 }, { "epoch": 9.65, "learning_rate": 1.050142993326978e-07, "loss": 1.7891, "step": 50664 }, { "epoch": 9.65, "learning_rate": 1.0495710200190658e-07, "loss": 1.939, "step": 50665 }, { "epoch": 9.65, "learning_rate": 1.0489990467111534e-07, "loss": 2.7201, "step": 50666 }, { "epoch": 9.65, "learning_rate": 1.0484270734032412e-07, "loss": 1.6473, "step": 50667 }, { "epoch": 9.65, "learning_rate": 1.047855100095329e-07, "loss": 2.6372, "step": 50668 }, { "epoch": 9.65, "learning_rate": 1.0472831267874165e-07, "loss": 1.2186, "step": 50669 }, { "epoch": 9.65, "learning_rate": 1.0467111534795043e-07, "loss": 0.7006, "step": 50670 }, { "epoch": 9.65, "learning_rate": 1.0461391801715921e-07, "loss": 2.8429, "step": 50671 }, { "epoch": 9.65, "learning_rate": 1.0455672068636797e-07, "loss": 1.111, "step": 50672 }, { "epoch": 9.65, "learning_rate": 1.0449952335557673e-07, "loss": 2.1524, "step": 50673 }, { "epoch": 9.65, "learning_rate": 1.0444232602478552e-07, "loss": 2.4121, "step": 50674 }, { "epoch": 9.65, "learning_rate": 1.0438512869399427e-07, "loss": 0.8562, "step": 50675 }, { "epoch": 9.65, "learning_rate": 1.0432793136320305e-07, "loss": 0.577, "step": 50676 }, { "epoch": 9.65, "learning_rate": 1.0427073403241183e-07, "loss": 1.8586, "step": 50677 }, { "epoch": 9.65, "learning_rate": 1.0421353670162058e-07, "loss": 2.392, "step": 50678 }, { "epoch": 9.65, "learning_rate": 1.0415633937082937e-07, "loss": 0.8317, "step": 50679 }, { "epoch": 9.65, "learning_rate": 1.0409914204003815e-07, "loss": 1.1567, "step": 50680 }, { "epoch": 9.65, "learning_rate": 1.040419447092469e-07, "loss": 2.2625, "step": 50681 }, { "epoch": 9.65, "learning_rate": 1.0398474737845568e-07, "loss": 2.338, "step": 50682 }, { "epoch": 9.65, "learning_rate": 1.0392755004766445e-07, "loss": 3.5887, "step": 50683 }, { "epoch": 9.65, "learning_rate": 1.038703527168732e-07, "loss": 0.8912, "step": 50684 }, { "epoch": 9.65, "learning_rate": 1.0381315538608198e-07, "loss": 3.7546, "step": 50685 }, { "epoch": 9.65, "learning_rate": 1.0375595805529076e-07, "loss": 3.2086, "step": 50686 }, { "epoch": 9.65, "learning_rate": 1.0369876072449952e-07, "loss": 1.2556, "step": 50687 }, { "epoch": 9.65, "learning_rate": 1.036415633937083e-07, "loss": 1.7549, "step": 50688 }, { "epoch": 9.66, "learning_rate": 1.0358436606291708e-07, "loss": 0.9176, "step": 50689 }, { "epoch": 9.66, "learning_rate": 1.0352716873212583e-07, "loss": 2.8453, "step": 50690 }, { "epoch": 9.66, "learning_rate": 1.0346997140133461e-07, "loss": 1.9005, "step": 50691 }, { "epoch": 9.66, "learning_rate": 1.0341277407054338e-07, "loss": 0.3376, "step": 50692 }, { "epoch": 9.66, "learning_rate": 1.0335557673975215e-07, "loss": 0.9537, "step": 50693 }, { "epoch": 9.66, "learning_rate": 1.0329837940896092e-07, "loss": 0.6762, "step": 50694 }, { "epoch": 9.66, "learning_rate": 1.032411820781697e-07, "loss": 1.2158, "step": 50695 }, { "epoch": 9.66, "learning_rate": 1.0318398474737845e-07, "loss": 2.382, "step": 50696 }, { "epoch": 9.66, "learning_rate": 1.0312678741658723e-07, "loss": 2.2535, "step": 50697 }, { "epoch": 9.66, "learning_rate": 1.0306959008579601e-07, "loss": 2.4834, "step": 50698 }, { "epoch": 9.66, "learning_rate": 1.0301239275500477e-07, "loss": 0.7908, "step": 50699 }, { "epoch": 9.66, "learning_rate": 1.0295519542421355e-07, "loss": 2.1594, "step": 50700 }, { "epoch": 9.66, "learning_rate": 1.0289799809342231e-07, "loss": 1.5984, "step": 50701 }, { "epoch": 9.66, "learning_rate": 1.0284080076263108e-07, "loss": 1.3174, "step": 50702 }, { "epoch": 9.66, "learning_rate": 1.0278360343183985e-07, "loss": 0.4622, "step": 50703 }, { "epoch": 9.66, "learning_rate": 1.0272640610104863e-07, "loss": 1.9877, "step": 50704 }, { "epoch": 9.66, "learning_rate": 1.0266920877025738e-07, "loss": 0.7899, "step": 50705 }, { "epoch": 9.66, "learning_rate": 1.0261201143946616e-07, "loss": 0.77, "step": 50706 }, { "epoch": 9.66, "learning_rate": 1.0255481410867494e-07, "loss": 0.9498, "step": 50707 }, { "epoch": 9.66, "learning_rate": 1.024976167778837e-07, "loss": 1.6732, "step": 50708 }, { "epoch": 9.66, "learning_rate": 1.0244041944709248e-07, "loss": 1.7128, "step": 50709 }, { "epoch": 9.66, "learning_rate": 1.0238322211630125e-07, "loss": 1.0886, "step": 50710 }, { "epoch": 9.66, "learning_rate": 1.0232602478551001e-07, "loss": 1.7055, "step": 50711 }, { "epoch": 9.66, "learning_rate": 1.0226882745471878e-07, "loss": 0.8339, "step": 50712 }, { "epoch": 9.66, "learning_rate": 1.0221163012392756e-07, "loss": 1.1141, "step": 50713 }, { "epoch": 9.66, "learning_rate": 1.0215443279313632e-07, "loss": 1.4256, "step": 50714 }, { "epoch": 9.66, "learning_rate": 1.020972354623451e-07, "loss": 1.9613, "step": 50715 }, { "epoch": 9.66, "learning_rate": 1.0204003813155385e-07, "loss": 1.253, "step": 50716 }, { "epoch": 9.66, "learning_rate": 1.0198284080076263e-07, "loss": 0.6497, "step": 50717 }, { "epoch": 9.66, "learning_rate": 1.0192564346997141e-07, "loss": 1.6167, "step": 50718 }, { "epoch": 9.66, "learning_rate": 1.0186844613918017e-07, "loss": 0.9691, "step": 50719 }, { "epoch": 9.66, "learning_rate": 1.0181124880838895e-07, "loss": 0.4491, "step": 50720 }, { "epoch": 9.66, "learning_rate": 1.0175405147759771e-07, "loss": 0.2214, "step": 50721 }, { "epoch": 9.66, "learning_rate": 1.0169685414680648e-07, "loss": 0.5521, "step": 50722 }, { "epoch": 9.66, "learning_rate": 1.0163965681601525e-07, "loss": 2.9136, "step": 50723 }, { "epoch": 9.66, "learning_rate": 1.0158245948522403e-07, "loss": 2.1379, "step": 50724 }, { "epoch": 9.66, "learning_rate": 1.0152526215443278e-07, "loss": 1.2268, "step": 50725 }, { "epoch": 9.66, "learning_rate": 1.0146806482364156e-07, "loss": 1.3415, "step": 50726 }, { "epoch": 9.66, "learning_rate": 1.0141086749285034e-07, "loss": 1.1306, "step": 50727 }, { "epoch": 9.66, "learning_rate": 1.013536701620591e-07, "loss": 1.9566, "step": 50728 }, { "epoch": 9.66, "learning_rate": 1.0129647283126788e-07, "loss": 0.957, "step": 50729 }, { "epoch": 9.66, "learning_rate": 1.0123927550047666e-07, "loss": 0.3705, "step": 50730 }, { "epoch": 9.66, "learning_rate": 1.0118207816968541e-07, "loss": 0.9566, "step": 50731 }, { "epoch": 9.66, "learning_rate": 1.0112488083889418e-07, "loss": 1.5085, "step": 50732 }, { "epoch": 9.66, "learning_rate": 1.0106768350810296e-07, "loss": 1.5261, "step": 50733 }, { "epoch": 9.66, "learning_rate": 1.0101048617731171e-07, "loss": 1.571, "step": 50734 }, { "epoch": 9.66, "learning_rate": 1.009532888465205e-07, "loss": 3.1356, "step": 50735 }, { "epoch": 9.66, "learning_rate": 1.0089609151572928e-07, "loss": 0.7771, "step": 50736 }, { "epoch": 9.66, "learning_rate": 1.0083889418493803e-07, "loss": 1.9242, "step": 50737 }, { "epoch": 9.66, "learning_rate": 1.0078169685414681e-07, "loss": 1.3496, "step": 50738 }, { "epoch": 9.66, "learning_rate": 1.0072449952335559e-07, "loss": 2.3465, "step": 50739 }, { "epoch": 9.66, "learning_rate": 1.0066730219256435e-07, "loss": 2.2142, "step": 50740 }, { "epoch": 9.66, "learning_rate": 1.0061010486177313e-07, "loss": 0.9989, "step": 50741 }, { "epoch": 9.67, "learning_rate": 1.005529075309819e-07, "loss": 1.1599, "step": 50742 }, { "epoch": 9.67, "learning_rate": 1.0049571020019065e-07, "loss": 0.9291, "step": 50743 }, { "epoch": 9.67, "learning_rate": 1.0043851286939943e-07, "loss": 1.4116, "step": 50744 }, { "epoch": 9.67, "learning_rate": 1.0038131553860821e-07, "loss": 2.2693, "step": 50745 }, { "epoch": 9.67, "learning_rate": 1.0032411820781696e-07, "loss": 0.2987, "step": 50746 }, { "epoch": 9.67, "learning_rate": 1.0026692087702574e-07, "loss": 0.5838, "step": 50747 }, { "epoch": 9.67, "learning_rate": 1.0020972354623452e-07, "loss": 2.4108, "step": 50748 }, { "epoch": 9.67, "learning_rate": 1.0015252621544328e-07, "loss": 0.8999, "step": 50749 }, { "epoch": 9.67, "learning_rate": 1.0009532888465206e-07, "loss": 2.4134, "step": 50750 }, { "epoch": 9.67, "learning_rate": 1.0003813155386083e-07, "loss": 0.5986, "step": 50751 }, { "epoch": 9.67, "learning_rate": 9.998093422306958e-08, "loss": 1.7095, "step": 50752 }, { "epoch": 9.67, "learning_rate": 9.992373689227836e-08, "loss": 1.2494, "step": 50753 }, { "epoch": 9.67, "learning_rate": 9.986653956148714e-08, "loss": 1.4497, "step": 50754 }, { "epoch": 9.67, "learning_rate": 9.98093422306959e-08, "loss": 2.2648, "step": 50755 }, { "epoch": 9.67, "learning_rate": 9.975214489990468e-08, "loss": 1.9292, "step": 50756 }, { "epoch": 9.67, "learning_rate": 9.969494756911346e-08, "loss": 1.1457, "step": 50757 }, { "epoch": 9.67, "learning_rate": 9.963775023832221e-08, "loss": 1.5043, "step": 50758 }, { "epoch": 9.67, "learning_rate": 9.958055290753099e-08, "loss": 2.6217, "step": 50759 }, { "epoch": 9.67, "learning_rate": 9.952335557673976e-08, "loss": 0.4036, "step": 50760 }, { "epoch": 9.67, "learning_rate": 9.946615824594853e-08, "loss": 0.8488, "step": 50761 }, { "epoch": 9.67, "learning_rate": 9.940896091515729e-08, "loss": 1.3567, "step": 50762 }, { "epoch": 9.67, "learning_rate": 9.935176358436607e-08, "loss": 1.1468, "step": 50763 }, { "epoch": 9.67, "learning_rate": 9.929456625357483e-08, "loss": 0.6624, "step": 50764 }, { "epoch": 9.67, "learning_rate": 9.923736892278361e-08, "loss": 2.6186, "step": 50765 }, { "epoch": 9.67, "learning_rate": 9.918017159199239e-08, "loss": 2.6807, "step": 50766 }, { "epoch": 9.67, "learning_rate": 9.912297426120114e-08, "loss": 2.2401, "step": 50767 }, { "epoch": 9.67, "learning_rate": 9.906577693040992e-08, "loss": 1.2067, "step": 50768 }, { "epoch": 9.67, "learning_rate": 9.900857959961869e-08, "loss": 0.6919, "step": 50769 }, { "epoch": 9.67, "learning_rate": 9.895138226882746e-08, "loss": 2.2596, "step": 50770 }, { "epoch": 9.67, "learning_rate": 9.889418493803623e-08, "loss": 2.8375, "step": 50771 }, { "epoch": 9.67, "learning_rate": 9.8836987607245e-08, "loss": 1.3621, "step": 50772 }, { "epoch": 9.67, "learning_rate": 9.877979027645376e-08, "loss": 1.2238, "step": 50773 }, { "epoch": 9.67, "learning_rate": 9.872259294566254e-08, "loss": 1.4048, "step": 50774 }, { "epoch": 9.67, "learning_rate": 9.86653956148713e-08, "loss": 1.1764, "step": 50775 }, { "epoch": 9.67, "learning_rate": 9.860819828408008e-08, "loss": 1.4254, "step": 50776 }, { "epoch": 9.67, "learning_rate": 9.855100095328886e-08, "loss": 0.4288, "step": 50777 }, { "epoch": 9.67, "learning_rate": 9.849380362249761e-08, "loss": 0.9096, "step": 50778 }, { "epoch": 9.67, "learning_rate": 9.843660629170639e-08, "loss": 0.223, "step": 50779 }, { "epoch": 9.67, "learning_rate": 9.837940896091516e-08, "loss": 1.3954, "step": 50780 }, { "epoch": 9.67, "learning_rate": 9.832221163012393e-08, "loss": 0.7099, "step": 50781 }, { "epoch": 9.67, "learning_rate": 9.826501429933269e-08, "loss": 1.6845, "step": 50782 }, { "epoch": 9.67, "learning_rate": 9.820781696854147e-08, "loss": 1.3299, "step": 50783 }, { "epoch": 9.67, "learning_rate": 9.815061963775023e-08, "loss": 1.5827, "step": 50784 }, { "epoch": 9.67, "learning_rate": 9.809342230695901e-08, "loss": 0.3746, "step": 50785 }, { "epoch": 9.67, "learning_rate": 9.803622497616779e-08, "loss": 0.626, "step": 50786 }, { "epoch": 9.67, "learning_rate": 9.797902764537654e-08, "loss": 2.4941, "step": 50787 }, { "epoch": 9.67, "learning_rate": 9.792183031458532e-08, "loss": 0.3805, "step": 50788 }, { "epoch": 9.67, "learning_rate": 9.786463298379409e-08, "loss": 1.2165, "step": 50789 }, { "epoch": 9.67, "learning_rate": 9.780743565300286e-08, "loss": 1.1313, "step": 50790 }, { "epoch": 9.67, "learning_rate": 9.775023832221163e-08, "loss": 2.4759, "step": 50791 }, { "epoch": 9.67, "learning_rate": 9.76930409914204e-08, "loss": 0.9556, "step": 50792 }, { "epoch": 9.67, "learning_rate": 9.763584366062916e-08, "loss": 0.5433, "step": 50793 }, { "epoch": 9.68, "learning_rate": 9.757864632983794e-08, "loss": 1.4153, "step": 50794 }, { "epoch": 9.68, "learning_rate": 9.752144899904672e-08, "loss": 0.805, "step": 50795 }, { "epoch": 9.68, "learning_rate": 9.746425166825548e-08, "loss": 1.9771, "step": 50796 }, { "epoch": 9.68, "learning_rate": 9.740705433746426e-08, "loss": 2.0195, "step": 50797 }, { "epoch": 9.68, "learning_rate": 9.734985700667304e-08, "loss": 2.4296, "step": 50798 }, { "epoch": 9.68, "learning_rate": 9.729265967588179e-08, "loss": 1.5881, "step": 50799 }, { "epoch": 9.68, "learning_rate": 9.723546234509056e-08, "loss": 1.8548, "step": 50800 }, { "epoch": 9.68, "learning_rate": 9.717826501429934e-08, "loss": 0.4526, "step": 50801 }, { "epoch": 9.68, "learning_rate": 9.712106768350809e-08, "loss": 1.5611, "step": 50802 }, { "epoch": 9.68, "learning_rate": 9.706387035271687e-08, "loss": 1.4076, "step": 50803 }, { "epoch": 9.68, "learning_rate": 9.700667302192565e-08, "loss": 1.2642, "step": 50804 }, { "epoch": 9.68, "learning_rate": 9.694947569113441e-08, "loss": 0.8719, "step": 50805 }, { "epoch": 9.68, "learning_rate": 9.689227836034319e-08, "loss": 1.9088, "step": 50806 }, { "epoch": 9.68, "learning_rate": 9.683508102955197e-08, "loss": 1.1554, "step": 50807 }, { "epoch": 9.68, "learning_rate": 9.677788369876072e-08, "loss": 3.1631, "step": 50808 }, { "epoch": 9.68, "learning_rate": 9.67206863679695e-08, "loss": 2.2059, "step": 50809 }, { "epoch": 9.68, "learning_rate": 9.666348903717827e-08, "loss": 0.7181, "step": 50810 }, { "epoch": 9.68, "learning_rate": 9.660629170638703e-08, "loss": 0.2932, "step": 50811 }, { "epoch": 9.68, "learning_rate": 9.65490943755958e-08, "loss": 1.7183, "step": 50812 }, { "epoch": 9.68, "learning_rate": 9.649189704480459e-08, "loss": 1.5458, "step": 50813 }, { "epoch": 9.68, "learning_rate": 9.643469971401334e-08, "loss": 1.7771, "step": 50814 }, { "epoch": 9.68, "learning_rate": 9.637750238322212e-08, "loss": 1.6144, "step": 50815 }, { "epoch": 9.68, "learning_rate": 9.63203050524309e-08, "loss": 1.4265, "step": 50816 }, { "epoch": 9.68, "learning_rate": 9.626310772163966e-08, "loss": 1.1867, "step": 50817 }, { "epoch": 9.68, "learning_rate": 9.620591039084844e-08, "loss": 1.4387, "step": 50818 }, { "epoch": 9.68, "learning_rate": 9.61487130600572e-08, "loss": 1.6922, "step": 50819 }, { "epoch": 9.68, "learning_rate": 9.609151572926596e-08, "loss": 1.0644, "step": 50820 }, { "epoch": 9.68, "learning_rate": 9.603431839847474e-08, "loss": 2.0653, "step": 50821 }, { "epoch": 9.68, "learning_rate": 9.597712106768352e-08, "loss": 2.2294, "step": 50822 }, { "epoch": 9.68, "learning_rate": 9.591992373689227e-08, "loss": 1.2847, "step": 50823 }, { "epoch": 9.68, "learning_rate": 9.586272640610105e-08, "loss": 2.7415, "step": 50824 }, { "epoch": 9.68, "learning_rate": 9.580552907530983e-08, "loss": 1.2815, "step": 50825 }, { "epoch": 9.68, "learning_rate": 9.574833174451859e-08, "loss": 2.2755, "step": 50826 }, { "epoch": 9.68, "learning_rate": 9.569113441372737e-08, "loss": 1.9644, "step": 50827 }, { "epoch": 9.68, "learning_rate": 9.563393708293614e-08, "loss": 2.0756, "step": 50828 }, { "epoch": 9.68, "learning_rate": 9.55767397521449e-08, "loss": 1.3579, "step": 50829 }, { "epoch": 9.68, "learning_rate": 9.551954242135367e-08, "loss": 1.4218, "step": 50830 }, { "epoch": 9.68, "learning_rate": 9.546234509056245e-08, "loss": 1.3805, "step": 50831 }, { "epoch": 9.68, "learning_rate": 9.54051477597712e-08, "loss": 0.7414, "step": 50832 }, { "epoch": 9.68, "learning_rate": 9.534795042897999e-08, "loss": 2.1946, "step": 50833 }, { "epoch": 9.68, "learning_rate": 9.529075309818877e-08, "loss": 1.9435, "step": 50834 }, { "epoch": 9.68, "learning_rate": 9.523355576739752e-08, "loss": 1.3006, "step": 50835 }, { "epoch": 9.68, "learning_rate": 9.51763584366063e-08, "loss": 1.9904, "step": 50836 }, { "epoch": 9.68, "learning_rate": 9.511916110581506e-08, "loss": 1.7856, "step": 50837 }, { "epoch": 9.68, "learning_rate": 9.506196377502384e-08, "loss": 1.1723, "step": 50838 }, { "epoch": 9.68, "learning_rate": 9.50047664442326e-08, "loss": 2.0578, "step": 50839 }, { "epoch": 9.68, "learning_rate": 9.494756911344137e-08, "loss": 0.8176, "step": 50840 }, { "epoch": 9.68, "learning_rate": 9.489037178265014e-08, "loss": 1.9973, "step": 50841 }, { "epoch": 9.68, "learning_rate": 9.483317445185892e-08, "loss": 1.3723, "step": 50842 }, { "epoch": 9.68, "learning_rate": 9.477597712106767e-08, "loss": 1.0559, "step": 50843 }, { "epoch": 9.68, "learning_rate": 9.471877979027645e-08, "loss": 1.9706, "step": 50844 }, { "epoch": 9.68, "learning_rate": 9.466158245948523e-08, "loss": 1.133, "step": 50845 }, { "epoch": 9.68, "learning_rate": 9.460438512869399e-08, "loss": 2.1789, "step": 50846 }, { "epoch": 9.69, "learning_rate": 9.454718779790277e-08, "loss": 1.4918, "step": 50847 }, { "epoch": 9.69, "learning_rate": 9.448999046711154e-08, "loss": 0.5566, "step": 50848 }, { "epoch": 9.69, "learning_rate": 9.44327931363203e-08, "loss": 0.9806, "step": 50849 }, { "epoch": 9.69, "learning_rate": 9.437559580552907e-08, "loss": 1.8944, "step": 50850 }, { "epoch": 9.69, "learning_rate": 9.431839847473785e-08, "loss": 2.0109, "step": 50851 }, { "epoch": 9.69, "learning_rate": 9.42612011439466e-08, "loss": 1.2156, "step": 50852 }, { "epoch": 9.69, "learning_rate": 9.420400381315539e-08, "loss": 1.2272, "step": 50853 }, { "epoch": 9.69, "learning_rate": 9.414680648236417e-08, "loss": 1.0022, "step": 50854 }, { "epoch": 9.69, "learning_rate": 9.408960915157292e-08, "loss": 1.2133, "step": 50855 }, { "epoch": 9.69, "learning_rate": 9.40324118207817e-08, "loss": 1.0849, "step": 50856 }, { "epoch": 9.69, "learning_rate": 9.397521448999047e-08, "loss": 1.4261, "step": 50857 }, { "epoch": 9.69, "learning_rate": 9.391801715919924e-08, "loss": 1.056, "step": 50858 }, { "epoch": 9.69, "learning_rate": 9.3860819828408e-08, "loss": 1.6893, "step": 50859 }, { "epoch": 9.69, "learning_rate": 9.380362249761678e-08, "loss": 0.8188, "step": 50860 }, { "epoch": 9.69, "learning_rate": 9.374642516682555e-08, "loss": 2.0351, "step": 50861 }, { "epoch": 9.69, "learning_rate": 9.368922783603432e-08, "loss": 1.2269, "step": 50862 }, { "epoch": 9.69, "learning_rate": 9.363203050524309e-08, "loss": 2.2529, "step": 50863 }, { "epoch": 9.69, "learning_rate": 9.357483317445187e-08, "loss": 2.0927, "step": 50864 }, { "epoch": 9.69, "learning_rate": 9.351763584366063e-08, "loss": 1.6004, "step": 50865 }, { "epoch": 9.69, "learning_rate": 9.34604385128694e-08, "loss": 1.3698, "step": 50866 }, { "epoch": 9.69, "learning_rate": 9.340324118207817e-08, "loss": 2.0645, "step": 50867 }, { "epoch": 9.69, "learning_rate": 9.334604385128694e-08, "loss": 2.738, "step": 50868 }, { "epoch": 9.69, "learning_rate": 9.32888465204957e-08, "loss": 1.7003, "step": 50869 }, { "epoch": 9.69, "learning_rate": 9.323164918970448e-08, "loss": 1.7768, "step": 50870 }, { "epoch": 9.69, "learning_rate": 9.317445185891325e-08, "loss": 1.1154, "step": 50871 }, { "epoch": 9.69, "learning_rate": 9.311725452812202e-08, "loss": 1.0566, "step": 50872 }, { "epoch": 9.69, "learning_rate": 9.30600571973308e-08, "loss": 0.7658, "step": 50873 }, { "epoch": 9.69, "learning_rate": 9.300285986653957e-08, "loss": 0.953, "step": 50874 }, { "epoch": 9.69, "learning_rate": 9.294566253574833e-08, "loss": 2.2905, "step": 50875 }, { "epoch": 9.69, "learning_rate": 9.288846520495712e-08, "loss": 2.0521, "step": 50876 }, { "epoch": 9.69, "learning_rate": 9.283126787416588e-08, "loss": 1.2942, "step": 50877 }, { "epoch": 9.69, "learning_rate": 9.277407054337464e-08, "loss": 1.6514, "step": 50878 }, { "epoch": 9.69, "learning_rate": 9.271687321258342e-08, "loss": 2.2399, "step": 50879 }, { "epoch": 9.69, "learning_rate": 9.265967588179218e-08, "loss": 1.1689, "step": 50880 }, { "epoch": 9.69, "learning_rate": 9.260247855100095e-08, "loss": 0.5596, "step": 50881 }, { "epoch": 9.69, "learning_rate": 9.254528122020972e-08, "loss": 0.7825, "step": 50882 }, { "epoch": 9.69, "learning_rate": 9.24880838894185e-08, "loss": 2.4215, "step": 50883 }, { "epoch": 9.69, "learning_rate": 9.243088655862727e-08, "loss": 1.223, "step": 50884 }, { "epoch": 9.69, "learning_rate": 9.237368922783603e-08, "loss": 1.9418, "step": 50885 }, { "epoch": 9.69, "learning_rate": 9.231649189704482e-08, "loss": 0.9127, "step": 50886 }, { "epoch": 9.69, "learning_rate": 9.225929456625358e-08, "loss": 0.9635, "step": 50887 }, { "epoch": 9.69, "learning_rate": 9.220209723546234e-08, "loss": 1.8215, "step": 50888 }, { "epoch": 9.69, "learning_rate": 9.214489990467112e-08, "loss": 0.9367, "step": 50889 }, { "epoch": 9.69, "learning_rate": 9.208770257387988e-08, "loss": 0.5741, "step": 50890 }, { "epoch": 9.69, "learning_rate": 9.203050524308865e-08, "loss": 0.8288, "step": 50891 }, { "epoch": 9.69, "learning_rate": 9.197330791229743e-08, "loss": 0.7242, "step": 50892 }, { "epoch": 9.69, "learning_rate": 9.19161105815062e-08, "loss": 1.8527, "step": 50893 }, { "epoch": 9.69, "learning_rate": 9.185891325071497e-08, "loss": 0.7001, "step": 50894 }, { "epoch": 9.69, "learning_rate": 9.180171591992375e-08, "loss": 0.3359, "step": 50895 }, { "epoch": 9.69, "learning_rate": 9.174451858913252e-08, "loss": 1.6819, "step": 50896 }, { "epoch": 9.69, "learning_rate": 9.168732125834128e-08, "loss": 1.15, "step": 50897 }, { "epoch": 9.69, "learning_rate": 9.163012392755005e-08, "loss": 2.344, "step": 50898 }, { "epoch": 9.7, "learning_rate": 9.157292659675882e-08, "loss": 0.7565, "step": 50899 }, { "epoch": 9.7, "learning_rate": 9.151572926596758e-08, "loss": 0.7401, "step": 50900 }, { "epoch": 9.7, "learning_rate": 9.145853193517637e-08, "loss": 1.2567, "step": 50901 }, { "epoch": 9.7, "learning_rate": 9.140133460438513e-08, "loss": 2.5298, "step": 50902 }, { "epoch": 9.7, "learning_rate": 9.13441372735939e-08, "loss": 1.9757, "step": 50903 }, { "epoch": 9.7, "learning_rate": 9.128693994280268e-08, "loss": 1.87, "step": 50904 }, { "epoch": 9.7, "learning_rate": 9.122974261201145e-08, "loss": 1.245, "step": 50905 }, { "epoch": 9.7, "learning_rate": 9.117254528122021e-08, "loss": 2.9455, "step": 50906 }, { "epoch": 9.7, "learning_rate": 9.111534795042898e-08, "loss": 0.6068, "step": 50907 }, { "epoch": 9.7, "learning_rate": 9.105815061963775e-08, "loss": 1.2092, "step": 50908 }, { "epoch": 9.7, "learning_rate": 9.100095328884652e-08, "loss": 2.4835, "step": 50909 }, { "epoch": 9.7, "learning_rate": 9.09437559580553e-08, "loss": 1.2811, "step": 50910 }, { "epoch": 9.7, "learning_rate": 9.088655862726406e-08, "loss": 1.7321, "step": 50911 }, { "epoch": 9.7, "learning_rate": 9.082936129647283e-08, "loss": 0.6146, "step": 50912 }, { "epoch": 9.7, "learning_rate": 9.07721639656816e-08, "loss": 1.0976, "step": 50913 }, { "epoch": 9.7, "learning_rate": 9.071496663489038e-08, "loss": 2.0063, "step": 50914 }, { "epoch": 9.7, "learning_rate": 9.065776930409915e-08, "loss": 0.6686, "step": 50915 }, { "epoch": 9.7, "learning_rate": 9.060057197330791e-08, "loss": 1.0674, "step": 50916 }, { "epoch": 9.7, "learning_rate": 9.054337464251668e-08, "loss": 2.3824, "step": 50917 }, { "epoch": 9.7, "learning_rate": 9.048617731172545e-08, "loss": 2.0489, "step": 50918 }, { "epoch": 9.7, "learning_rate": 9.042897998093422e-08, "loss": 1.5966, "step": 50919 }, { "epoch": 9.7, "learning_rate": 9.0371782650143e-08, "loss": 1.2824, "step": 50920 }, { "epoch": 9.7, "learning_rate": 9.031458531935176e-08, "loss": 0.8755, "step": 50921 }, { "epoch": 9.7, "learning_rate": 9.025738798856053e-08, "loss": 1.8891, "step": 50922 }, { "epoch": 9.7, "learning_rate": 9.020019065776931e-08, "loss": 2.078, "step": 50923 }, { "epoch": 9.7, "learning_rate": 9.014299332697808e-08, "loss": 1.9525, "step": 50924 }, { "epoch": 9.7, "learning_rate": 9.008579599618685e-08, "loss": 0.856, "step": 50925 }, { "epoch": 9.7, "learning_rate": 9.002859866539561e-08, "loss": 1.1876, "step": 50926 }, { "epoch": 9.7, "learning_rate": 8.997140133460438e-08, "loss": 1.7512, "step": 50927 }, { "epoch": 9.7, "learning_rate": 8.991420400381315e-08, "loss": 0.5073, "step": 50928 }, { "epoch": 9.7, "learning_rate": 8.985700667302193e-08, "loss": 1.1356, "step": 50929 }, { "epoch": 9.7, "learning_rate": 8.97998093422307e-08, "loss": 1.357, "step": 50930 }, { "epoch": 9.7, "learning_rate": 8.974261201143946e-08, "loss": 1.1966, "step": 50931 }, { "epoch": 9.7, "learning_rate": 8.968541468064825e-08, "loss": 2.3898, "step": 50932 }, { "epoch": 9.7, "learning_rate": 8.962821734985701e-08, "loss": 2.2407, "step": 50933 }, { "epoch": 9.7, "learning_rate": 8.957102001906578e-08, "loss": 3.3456, "step": 50934 }, { "epoch": 9.7, "learning_rate": 8.951382268827455e-08, "loss": 0.7489, "step": 50935 }, { "epoch": 9.7, "learning_rate": 8.945662535748331e-08, "loss": 0.5721, "step": 50936 }, { "epoch": 9.7, "learning_rate": 8.939942802669208e-08, "loss": 2.2202, "step": 50937 }, { "epoch": 9.7, "learning_rate": 8.934223069590086e-08, "loss": 0.9352, "step": 50938 }, { "epoch": 9.7, "learning_rate": 8.928503336510963e-08, "loss": 0.914, "step": 50939 }, { "epoch": 9.7, "learning_rate": 8.92278360343184e-08, "loss": 1.416, "step": 50940 }, { "epoch": 9.7, "learning_rate": 8.917063870352716e-08, "loss": 1.3278, "step": 50941 }, { "epoch": 9.7, "learning_rate": 8.911344137273595e-08, "loss": 1.4665, "step": 50942 }, { "epoch": 9.7, "learning_rate": 8.905624404194471e-08, "loss": 1.336, "step": 50943 }, { "epoch": 9.7, "learning_rate": 8.899904671115348e-08, "loss": 1.4679, "step": 50944 }, { "epoch": 9.7, "learning_rate": 8.894184938036226e-08, "loss": 1.6491, "step": 50945 }, { "epoch": 9.7, "learning_rate": 8.888465204957101e-08, "loss": 2.276, "step": 50946 }, { "epoch": 9.7, "learning_rate": 8.882745471877978e-08, "loss": 0.6538, "step": 50947 }, { "epoch": 9.7, "learning_rate": 8.877025738798856e-08, "loss": 0.8368, "step": 50948 }, { "epoch": 9.7, "learning_rate": 8.871306005719733e-08, "loss": 1.6976, "step": 50949 }, { "epoch": 9.7, "learning_rate": 8.86558627264061e-08, "loss": 0.8103, "step": 50950 }, { "epoch": 9.7, "learning_rate": 8.859866539561488e-08, "loss": 1.8112, "step": 50951 }, { "epoch": 9.71, "learning_rate": 8.854146806482365e-08, "loss": 1.8907, "step": 50952 }, { "epoch": 9.71, "learning_rate": 8.848427073403241e-08, "loss": 2.0766, "step": 50953 }, { "epoch": 9.71, "learning_rate": 8.842707340324119e-08, "loss": 1.6008, "step": 50954 }, { "epoch": 9.71, "learning_rate": 8.836987607244996e-08, "loss": 0.5337, "step": 50955 }, { "epoch": 9.71, "learning_rate": 8.831267874165873e-08, "loss": 1.751, "step": 50956 }, { "epoch": 9.71, "learning_rate": 8.82554814108675e-08, "loss": 2.1909, "step": 50957 }, { "epoch": 9.71, "learning_rate": 8.819828408007626e-08, "loss": 0.8559, "step": 50958 }, { "epoch": 9.71, "learning_rate": 8.814108674928503e-08, "loss": 1.6013, "step": 50959 }, { "epoch": 9.71, "learning_rate": 8.808388941849381e-08, "loss": 0.7128, "step": 50960 }, { "epoch": 9.71, "learning_rate": 8.802669208770258e-08, "loss": 1.6305, "step": 50961 }, { "epoch": 9.71, "learning_rate": 8.796949475691135e-08, "loss": 0.848, "step": 50962 }, { "epoch": 9.71, "learning_rate": 8.791229742612013e-08, "loss": 2.1161, "step": 50963 }, { "epoch": 9.71, "learning_rate": 8.785510009532889e-08, "loss": 2.4981, "step": 50964 }, { "epoch": 9.71, "learning_rate": 8.779790276453766e-08, "loss": 0.4508, "step": 50965 }, { "epoch": 9.71, "learning_rate": 8.774070543374643e-08, "loss": 1.4539, "step": 50966 }, { "epoch": 9.71, "learning_rate": 8.76835081029552e-08, "loss": 1.1585, "step": 50967 }, { "epoch": 9.71, "learning_rate": 8.762631077216396e-08, "loss": 1.6866, "step": 50968 }, { "epoch": 9.71, "learning_rate": 8.756911344137274e-08, "loss": 2.1857, "step": 50969 }, { "epoch": 9.71, "learning_rate": 8.751191611058151e-08, "loss": 2.5223, "step": 50970 }, { "epoch": 9.71, "learning_rate": 8.745471877979028e-08, "loss": 1.3859, "step": 50971 }, { "epoch": 9.71, "learning_rate": 8.739752144899905e-08, "loss": 1.5975, "step": 50972 }, { "epoch": 9.71, "learning_rate": 8.734032411820783e-08, "loss": 1.1416, "step": 50973 }, { "epoch": 9.71, "learning_rate": 8.728312678741659e-08, "loss": 1.4408, "step": 50974 }, { "epoch": 9.71, "learning_rate": 8.722592945662536e-08, "loss": 0.8081, "step": 50975 }, { "epoch": 9.71, "learning_rate": 8.716873212583413e-08, "loss": 2.714, "step": 50976 }, { "epoch": 9.71, "learning_rate": 8.71115347950429e-08, "loss": 2.1376, "step": 50977 }, { "epoch": 9.71, "learning_rate": 8.705433746425166e-08, "loss": 1.4647, "step": 50978 }, { "epoch": 9.71, "learning_rate": 8.699714013346044e-08, "loss": 1.4154, "step": 50979 }, { "epoch": 9.71, "learning_rate": 8.693994280266921e-08, "loss": 1.8071, "step": 50980 }, { "epoch": 9.71, "learning_rate": 8.688274547187798e-08, "loss": 2.3121, "step": 50981 }, { "epoch": 9.71, "learning_rate": 8.682554814108676e-08, "loss": 1.0081, "step": 50982 }, { "epoch": 9.71, "learning_rate": 8.676835081029553e-08, "loss": 1.7909, "step": 50983 }, { "epoch": 9.71, "learning_rate": 8.671115347950429e-08, "loss": 1.1709, "step": 50984 }, { "epoch": 9.71, "learning_rate": 8.665395614871306e-08, "loss": 1.3573, "step": 50985 }, { "epoch": 9.71, "learning_rate": 8.659675881792183e-08, "loss": 2.1991, "step": 50986 }, { "epoch": 9.71, "learning_rate": 8.65395614871306e-08, "loss": 2.1865, "step": 50987 }, { "epoch": 9.71, "learning_rate": 8.648236415633938e-08, "loss": 3.2978, "step": 50988 }, { "epoch": 9.71, "learning_rate": 8.642516682554814e-08, "loss": 1.477, "step": 50989 }, { "epoch": 9.71, "learning_rate": 8.636796949475691e-08, "loss": 1.1835, "step": 50990 }, { "epoch": 9.71, "learning_rate": 8.631077216396569e-08, "loss": 1.5386, "step": 50991 }, { "epoch": 9.71, "learning_rate": 8.625357483317446e-08, "loss": 1.3193, "step": 50992 }, { "epoch": 9.71, "learning_rate": 8.619637750238323e-08, "loss": 2.3674, "step": 50993 }, { "epoch": 9.71, "learning_rate": 8.613918017159199e-08, "loss": 0.9227, "step": 50994 }, { "epoch": 9.71, "learning_rate": 8.608198284080076e-08, "loss": 1.7238, "step": 50995 }, { "epoch": 9.71, "learning_rate": 8.602478551000953e-08, "loss": 1.5016, "step": 50996 }, { "epoch": 9.71, "learning_rate": 8.596758817921831e-08, "loss": 1.13, "step": 50997 }, { "epoch": 9.71, "learning_rate": 8.591039084842708e-08, "loss": 0.677, "step": 50998 }, { "epoch": 9.71, "learning_rate": 8.585319351763584e-08, "loss": 0.2113, "step": 50999 }, { "epoch": 9.71, "learning_rate": 8.579599618684462e-08, "loss": 0.7897, "step": 51000 }, { "epoch": 9.71, "learning_rate": 8.573879885605339e-08, "loss": 0.5898, "step": 51001 }, { "epoch": 9.71, "learning_rate": 8.568160152526216e-08, "loss": 1.6105, "step": 51002 }, { "epoch": 9.71, "learning_rate": 8.562440419447093e-08, "loss": 1.2573, "step": 51003 }, { "epoch": 9.72, "learning_rate": 8.556720686367969e-08, "loss": 1.7614, "step": 51004 }, { "epoch": 9.72, "learning_rate": 8.551000953288846e-08, "loss": 0.7659, "step": 51005 }, { "epoch": 9.72, "learning_rate": 8.545281220209723e-08, "loss": 1.028, "step": 51006 }, { "epoch": 9.72, "learning_rate": 8.539561487130601e-08, "loss": 0.3193, "step": 51007 }, { "epoch": 9.72, "learning_rate": 8.533841754051478e-08, "loss": 2.0679, "step": 51008 }, { "epoch": 9.72, "learning_rate": 8.528122020972354e-08, "loss": 2.1172, "step": 51009 }, { "epoch": 9.72, "learning_rate": 8.522402287893232e-08, "loss": 1.0856, "step": 51010 }, { "epoch": 9.72, "learning_rate": 8.516682554814109e-08, "loss": 1.8652, "step": 51011 }, { "epoch": 9.72, "learning_rate": 8.510962821734986e-08, "loss": 2.698, "step": 51012 }, { "epoch": 9.72, "learning_rate": 8.505243088655864e-08, "loss": 0.9762, "step": 51013 }, { "epoch": 9.72, "learning_rate": 8.499523355576739e-08, "loss": 1.1119, "step": 51014 }, { "epoch": 9.72, "learning_rate": 8.493803622497616e-08, "loss": 0.7304, "step": 51015 }, { "epoch": 9.72, "learning_rate": 8.488083889418494e-08, "loss": 0.6748, "step": 51016 }, { "epoch": 9.72, "learning_rate": 8.482364156339371e-08, "loss": 3.4732, "step": 51017 }, { "epoch": 9.72, "learning_rate": 8.476644423260248e-08, "loss": 1.6842, "step": 51018 }, { "epoch": 9.72, "learning_rate": 8.470924690181126e-08, "loss": 0.4101, "step": 51019 }, { "epoch": 9.72, "learning_rate": 8.465204957102002e-08, "loss": 0.971, "step": 51020 }, { "epoch": 9.72, "learning_rate": 8.459485224022879e-08, "loss": 1.4901, "step": 51021 }, { "epoch": 9.72, "learning_rate": 8.453765490943757e-08, "loss": 1.0327, "step": 51022 }, { "epoch": 9.72, "learning_rate": 8.448045757864634e-08, "loss": 1.3507, "step": 51023 }, { "epoch": 9.72, "learning_rate": 8.44232602478551e-08, "loss": 1.6984, "step": 51024 }, { "epoch": 9.72, "learning_rate": 8.436606291706387e-08, "loss": 0.3027, "step": 51025 }, { "epoch": 9.72, "learning_rate": 8.430886558627264e-08, "loss": 1.43, "step": 51026 }, { "epoch": 9.72, "learning_rate": 8.425166825548141e-08, "loss": 0.9624, "step": 51027 }, { "epoch": 9.72, "learning_rate": 8.419447092469019e-08, "loss": 1.9736, "step": 51028 }, { "epoch": 9.72, "learning_rate": 8.413727359389896e-08, "loss": 0.9775, "step": 51029 }, { "epoch": 9.72, "learning_rate": 8.408007626310772e-08, "loss": 2.0237, "step": 51030 }, { "epoch": 9.72, "learning_rate": 8.40228789323165e-08, "loss": 1.5659, "step": 51031 }, { "epoch": 9.72, "learning_rate": 8.396568160152527e-08, "loss": 3.1718, "step": 51032 }, { "epoch": 9.72, "learning_rate": 8.390848427073404e-08, "loss": 2.3765, "step": 51033 }, { "epoch": 9.72, "learning_rate": 8.38512869399428e-08, "loss": 1.228, "step": 51034 }, { "epoch": 9.72, "learning_rate": 8.379408960915157e-08, "loss": 1.7691, "step": 51035 }, { "epoch": 9.72, "learning_rate": 8.373689227836034e-08, "loss": 0.6377, "step": 51036 }, { "epoch": 9.72, "learning_rate": 8.367969494756911e-08, "loss": 1.5785, "step": 51037 }, { "epoch": 9.72, "learning_rate": 8.362249761677789e-08, "loss": 1.3302, "step": 51038 }, { "epoch": 9.72, "learning_rate": 8.356530028598666e-08, "loss": 2.1737, "step": 51039 }, { "epoch": 9.72, "learning_rate": 8.350810295519542e-08, "loss": 2.638, "step": 51040 }, { "epoch": 9.72, "learning_rate": 8.34509056244042e-08, "loss": 3.3177, "step": 51041 }, { "epoch": 9.72, "learning_rate": 8.339370829361297e-08, "loss": 1.1861, "step": 51042 }, { "epoch": 9.72, "learning_rate": 8.333651096282174e-08, "loss": 1.6539, "step": 51043 }, { "epoch": 9.72, "learning_rate": 8.32793136320305e-08, "loss": 1.3282, "step": 51044 }, { "epoch": 9.72, "learning_rate": 8.322211630123927e-08, "loss": 0.5638, "step": 51045 }, { "epoch": 9.72, "learning_rate": 8.316491897044804e-08, "loss": 0.9547, "step": 51046 }, { "epoch": 9.72, "learning_rate": 8.310772163965682e-08, "loss": 1.4401, "step": 51047 }, { "epoch": 9.72, "learning_rate": 8.305052430886559e-08, "loss": 1.1347, "step": 51048 }, { "epoch": 9.72, "learning_rate": 8.299332697807436e-08, "loss": 1.7404, "step": 51049 }, { "epoch": 9.72, "learning_rate": 8.293612964728314e-08, "loss": 2.0763, "step": 51050 }, { "epoch": 9.72, "learning_rate": 8.28789323164919e-08, "loss": 1.0865, "step": 51051 }, { "epoch": 9.72, "learning_rate": 8.282173498570067e-08, "loss": 2.2834, "step": 51052 }, { "epoch": 9.72, "learning_rate": 8.276453765490944e-08, "loss": 2.016, "step": 51053 }, { "epoch": 9.72, "learning_rate": 8.27073403241182e-08, "loss": 0.9245, "step": 51054 }, { "epoch": 9.72, "learning_rate": 8.265014299332697e-08, "loss": 0.6084, "step": 51055 }, { "epoch": 9.72, "learning_rate": 8.259294566253575e-08, "loss": 0.8995, "step": 51056 }, { "epoch": 9.73, "learning_rate": 8.253574833174452e-08, "loss": 1.9154, "step": 51057 }, { "epoch": 9.73, "learning_rate": 8.247855100095329e-08, "loss": 0.4603, "step": 51058 }, { "epoch": 9.73, "learning_rate": 8.242135367016207e-08, "loss": 1.0921, "step": 51059 }, { "epoch": 9.73, "learning_rate": 8.236415633937084e-08, "loss": 0.8745, "step": 51060 }, { "epoch": 9.73, "learning_rate": 8.23069590085796e-08, "loss": 1.569, "step": 51061 }, { "epoch": 9.73, "learning_rate": 8.224976167778837e-08, "loss": 1.4476, "step": 51062 }, { "epoch": 9.73, "learning_rate": 8.219256434699714e-08, "loss": 1.5234, "step": 51063 }, { "epoch": 9.73, "learning_rate": 8.21353670162059e-08, "loss": 0.7689, "step": 51064 }, { "epoch": 9.73, "learning_rate": 8.207816968541467e-08, "loss": 1.1787, "step": 51065 }, { "epoch": 9.73, "learning_rate": 8.202097235462345e-08, "loss": 1.5562, "step": 51066 }, { "epoch": 9.73, "learning_rate": 8.196377502383222e-08, "loss": 1.6525, "step": 51067 }, { "epoch": 9.73, "learning_rate": 8.190657769304099e-08, "loss": 0.9163, "step": 51068 }, { "epoch": 9.73, "learning_rate": 8.184938036224977e-08, "loss": 1.6641, "step": 51069 }, { "epoch": 9.73, "learning_rate": 8.179218303145854e-08, "loss": 1.8733, "step": 51070 }, { "epoch": 9.73, "learning_rate": 8.17349857006673e-08, "loss": 0.6293, "step": 51071 }, { "epoch": 9.73, "learning_rate": 8.167778836987607e-08, "loss": 1.6119, "step": 51072 }, { "epoch": 9.73, "learning_rate": 8.162059103908484e-08, "loss": 1.9628, "step": 51073 }, { "epoch": 9.73, "learning_rate": 8.15633937082936e-08, "loss": 1.4485, "step": 51074 }, { "epoch": 9.73, "learning_rate": 8.150619637750239e-08, "loss": 1.6553, "step": 51075 }, { "epoch": 9.73, "learning_rate": 8.144899904671115e-08, "loss": 2.3793, "step": 51076 }, { "epoch": 9.73, "learning_rate": 8.139180171591992e-08, "loss": 0.217, "step": 51077 }, { "epoch": 9.73, "learning_rate": 8.13346043851287e-08, "loss": 2.0908, "step": 51078 }, { "epoch": 9.73, "learning_rate": 8.127740705433747e-08, "loss": 0.4871, "step": 51079 }, { "epoch": 9.73, "learning_rate": 8.122020972354624e-08, "loss": 0.3351, "step": 51080 }, { "epoch": 9.73, "learning_rate": 8.116301239275502e-08, "loss": 1.1438, "step": 51081 }, { "epoch": 9.73, "learning_rate": 8.110581506196377e-08, "loss": 3.0883, "step": 51082 }, { "epoch": 9.73, "learning_rate": 8.104861773117254e-08, "loss": 1.5598, "step": 51083 }, { "epoch": 9.73, "learning_rate": 8.099142040038132e-08, "loss": 1.8553, "step": 51084 }, { "epoch": 9.73, "learning_rate": 8.093422306959009e-08, "loss": 1.6176, "step": 51085 }, { "epoch": 9.73, "learning_rate": 8.087702573879885e-08, "loss": 1.0101, "step": 51086 }, { "epoch": 9.73, "learning_rate": 8.081982840800763e-08, "loss": 2.3321, "step": 51087 }, { "epoch": 9.73, "learning_rate": 8.07626310772164e-08, "loss": 0.9762, "step": 51088 }, { "epoch": 9.73, "learning_rate": 8.070543374642517e-08, "loss": 0.6769, "step": 51089 }, { "epoch": 9.73, "learning_rate": 8.064823641563395e-08, "loss": 1.6278, "step": 51090 }, { "epoch": 9.73, "learning_rate": 8.059103908484272e-08, "loss": 0.9976, "step": 51091 }, { "epoch": 9.73, "learning_rate": 8.053384175405148e-08, "loss": 1.1178, "step": 51092 }, { "epoch": 9.73, "learning_rate": 8.047664442326024e-08, "loss": 1.3211, "step": 51093 }, { "epoch": 9.73, "learning_rate": 8.041944709246902e-08, "loss": 0.9841, "step": 51094 }, { "epoch": 9.73, "learning_rate": 8.036224976167779e-08, "loss": 0.6888, "step": 51095 }, { "epoch": 9.73, "learning_rate": 8.030505243088655e-08, "loss": 1.3806, "step": 51096 }, { "epoch": 9.73, "learning_rate": 8.024785510009533e-08, "loss": 1.3954, "step": 51097 }, { "epoch": 9.73, "learning_rate": 8.01906577693041e-08, "loss": 1.736, "step": 51098 }, { "epoch": 9.73, "learning_rate": 8.013346043851287e-08, "loss": 1.5187, "step": 51099 }, { "epoch": 9.73, "learning_rate": 8.007626310772165e-08, "loss": 1.8587, "step": 51100 }, { "epoch": 9.73, "learning_rate": 8.001906577693042e-08, "loss": 2.1319, "step": 51101 }, { "epoch": 9.73, "learning_rate": 7.996186844613918e-08, "loss": 0.3228, "step": 51102 }, { "epoch": 9.73, "learning_rate": 7.990467111534795e-08, "loss": 0.3275, "step": 51103 }, { "epoch": 9.73, "learning_rate": 7.984747378455672e-08, "loss": 1.6459, "step": 51104 }, { "epoch": 9.73, "learning_rate": 7.979027645376549e-08, "loss": 1.3151, "step": 51105 }, { "epoch": 9.73, "learning_rate": 7.973307912297427e-08, "loss": 1.7309, "step": 51106 }, { "epoch": 9.73, "learning_rate": 7.967588179218303e-08, "loss": 0.9268, "step": 51107 }, { "epoch": 9.73, "learning_rate": 7.96186844613918e-08, "loss": 2.1209, "step": 51108 }, { "epoch": 9.74, "learning_rate": 7.956148713060058e-08, "loss": 2.0425, "step": 51109 }, { "epoch": 9.74, "learning_rate": 7.950428979980935e-08, "loss": 2.0359, "step": 51110 }, { "epoch": 9.74, "learning_rate": 7.944709246901812e-08, "loss": 1.2797, "step": 51111 }, { "epoch": 9.74, "learning_rate": 7.938989513822688e-08, "loss": 0.8541, "step": 51112 }, { "epoch": 9.74, "learning_rate": 7.933269780743565e-08, "loss": 1.3781, "step": 51113 }, { "epoch": 9.74, "learning_rate": 7.927550047664442e-08, "loss": 1.6956, "step": 51114 }, { "epoch": 9.74, "learning_rate": 7.92183031458532e-08, "loss": 0.8513, "step": 51115 }, { "epoch": 9.74, "learning_rate": 7.916110581506197e-08, "loss": 0.5972, "step": 51116 }, { "epoch": 9.74, "learning_rate": 7.910390848427073e-08, "loss": 1.7354, "step": 51117 }, { "epoch": 9.74, "learning_rate": 7.904671115347951e-08, "loss": 2.8023, "step": 51118 }, { "epoch": 9.74, "learning_rate": 7.898951382268828e-08, "loss": 1.914, "step": 51119 }, { "epoch": 9.74, "learning_rate": 7.893231649189705e-08, "loss": 1.7156, "step": 51120 }, { "epoch": 9.74, "learning_rate": 7.887511916110582e-08, "loss": 1.7389, "step": 51121 }, { "epoch": 9.74, "learning_rate": 7.881792183031458e-08, "loss": 0.8623, "step": 51122 }, { "epoch": 9.74, "learning_rate": 7.876072449952335e-08, "loss": 0.7253, "step": 51123 }, { "epoch": 9.74, "learning_rate": 7.870352716873212e-08, "loss": 1.6374, "step": 51124 }, { "epoch": 9.74, "learning_rate": 7.86463298379409e-08, "loss": 0.3526, "step": 51125 }, { "epoch": 9.74, "learning_rate": 7.858913250714967e-08, "loss": 1.0965, "step": 51126 }, { "epoch": 9.74, "learning_rate": 7.853193517635843e-08, "loss": 1.8299, "step": 51127 }, { "epoch": 9.74, "learning_rate": 7.847473784556721e-08, "loss": 0.631, "step": 51128 }, { "epoch": 9.74, "learning_rate": 7.841754051477598e-08, "loss": 2.7741, "step": 51129 }, { "epoch": 9.74, "learning_rate": 7.836034318398475e-08, "loss": 1.4564, "step": 51130 }, { "epoch": 9.74, "learning_rate": 7.830314585319352e-08, "loss": 1.2697, "step": 51131 }, { "epoch": 9.74, "learning_rate": 7.824594852240228e-08, "loss": 1.6595, "step": 51132 }, { "epoch": 9.74, "learning_rate": 7.818875119161105e-08, "loss": 1.2634, "step": 51133 }, { "epoch": 9.74, "learning_rate": 7.813155386081983e-08, "loss": 1.4342, "step": 51134 }, { "epoch": 9.74, "learning_rate": 7.80743565300286e-08, "loss": 2.4387, "step": 51135 }, { "epoch": 9.74, "learning_rate": 7.801715919923737e-08, "loss": 0.942, "step": 51136 }, { "epoch": 9.74, "learning_rate": 7.795996186844615e-08, "loss": 1.1653, "step": 51137 }, { "epoch": 9.74, "learning_rate": 7.790276453765491e-08, "loss": 0.8865, "step": 51138 }, { "epoch": 9.74, "learning_rate": 7.784556720686368e-08, "loss": 1.7736, "step": 51139 }, { "epoch": 9.74, "learning_rate": 7.778836987607245e-08, "loss": 0.6752, "step": 51140 }, { "epoch": 9.74, "learning_rate": 7.773117254528122e-08, "loss": 0.8731, "step": 51141 }, { "epoch": 9.74, "learning_rate": 7.767397521448998e-08, "loss": 2.4551, "step": 51142 }, { "epoch": 9.74, "learning_rate": 7.761677788369876e-08, "loss": 1.5746, "step": 51143 }, { "epoch": 9.74, "learning_rate": 7.755958055290753e-08, "loss": 3.3475, "step": 51144 }, { "epoch": 9.74, "learning_rate": 7.75023832221163e-08, "loss": 2.4454, "step": 51145 }, { "epoch": 9.74, "learning_rate": 7.744518589132508e-08, "loss": 1.0071, "step": 51146 }, { "epoch": 9.74, "learning_rate": 7.738798856053385e-08, "loss": 2.0997, "step": 51147 }, { "epoch": 9.74, "learning_rate": 7.733079122974261e-08, "loss": 0.7817, "step": 51148 }, { "epoch": 9.74, "learning_rate": 7.72735938989514e-08, "loss": 2.2065, "step": 51149 }, { "epoch": 9.74, "learning_rate": 7.721639656816015e-08, "loss": 1.499, "step": 51150 }, { "epoch": 9.74, "learning_rate": 7.715919923736892e-08, "loss": 2.2832, "step": 51151 }, { "epoch": 9.74, "learning_rate": 7.71020019065777e-08, "loss": 1.8876, "step": 51152 }, { "epoch": 9.74, "learning_rate": 7.704480457578646e-08, "loss": 1.0845, "step": 51153 }, { "epoch": 9.74, "learning_rate": 7.698760724499523e-08, "loss": 1.319, "step": 51154 }, { "epoch": 9.74, "learning_rate": 7.6930409914204e-08, "loss": 0.71, "step": 51155 }, { "epoch": 9.74, "learning_rate": 7.687321258341278e-08, "loss": 0.8735, "step": 51156 }, { "epoch": 9.74, "learning_rate": 7.681601525262155e-08, "loss": 2.5548, "step": 51157 }, { "epoch": 9.74, "learning_rate": 7.675881792183031e-08, "loss": 1.2493, "step": 51158 }, { "epoch": 9.74, "learning_rate": 7.67016205910391e-08, "loss": 1.2239, "step": 51159 }, { "epoch": 9.74, "learning_rate": 7.664442326024786e-08, "loss": 1.6258, "step": 51160 }, { "epoch": 9.74, "learning_rate": 7.658722592945662e-08, "loss": 2.1453, "step": 51161 }, { "epoch": 9.75, "learning_rate": 7.65300285986654e-08, "loss": 1.7696, "step": 51162 }, { "epoch": 9.75, "learning_rate": 7.647283126787416e-08, "loss": 1.769, "step": 51163 }, { "epoch": 9.75, "learning_rate": 7.641563393708293e-08, "loss": 1.7798, "step": 51164 }, { "epoch": 9.75, "learning_rate": 7.635843660629171e-08, "loss": 1.1989, "step": 51165 }, { "epoch": 9.75, "learning_rate": 7.630123927550048e-08, "loss": 2.8291, "step": 51166 }, { "epoch": 9.75, "learning_rate": 7.624404194470925e-08, "loss": 1.0998, "step": 51167 }, { "epoch": 9.75, "learning_rate": 7.618684461391803e-08, "loss": 2.0889, "step": 51168 }, { "epoch": 9.75, "learning_rate": 7.61296472831268e-08, "loss": 1.6986, "step": 51169 }, { "epoch": 9.75, "learning_rate": 7.607244995233556e-08, "loss": 1.0353, "step": 51170 }, { "epoch": 9.75, "learning_rate": 7.601525262154433e-08, "loss": 2.2053, "step": 51171 }, { "epoch": 9.75, "learning_rate": 7.59580552907531e-08, "loss": 0.6323, "step": 51172 }, { "epoch": 9.75, "learning_rate": 7.590085795996186e-08, "loss": 2.1078, "step": 51173 }, { "epoch": 9.75, "learning_rate": 7.584366062917065e-08, "loss": 3.4724, "step": 51174 }, { "epoch": 9.75, "learning_rate": 7.578646329837941e-08, "loss": 3.104, "step": 51175 }, { "epoch": 9.75, "learning_rate": 7.572926596758818e-08, "loss": 1.6509, "step": 51176 }, { "epoch": 9.75, "learning_rate": 7.567206863679696e-08, "loss": 0.7748, "step": 51177 }, { "epoch": 9.75, "learning_rate": 7.561487130600573e-08, "loss": 1.0777, "step": 51178 }, { "epoch": 9.75, "learning_rate": 7.55576739752145e-08, "loss": 2.2713, "step": 51179 }, { "epoch": 9.75, "learning_rate": 7.550047664442326e-08, "loss": 2.2779, "step": 51180 }, { "epoch": 9.75, "learning_rate": 7.544327931363203e-08, "loss": 2.4252, "step": 51181 }, { "epoch": 9.75, "learning_rate": 7.53860819828408e-08, "loss": 2.6979, "step": 51182 }, { "epoch": 9.75, "learning_rate": 7.532888465204956e-08, "loss": 1.4213, "step": 51183 }, { "epoch": 9.75, "learning_rate": 7.527168732125835e-08, "loss": 0.7217, "step": 51184 }, { "epoch": 9.75, "learning_rate": 7.521448999046711e-08, "loss": 1.9743, "step": 51185 }, { "epoch": 9.75, "learning_rate": 7.515729265967588e-08, "loss": 1.1764, "step": 51186 }, { "epoch": 9.75, "learning_rate": 7.510009532888466e-08, "loss": 1.9532, "step": 51187 }, { "epoch": 9.75, "learning_rate": 7.504289799809343e-08, "loss": 1.9395, "step": 51188 }, { "epoch": 9.75, "learning_rate": 7.49857006673022e-08, "loss": 2.8892, "step": 51189 }, { "epoch": 9.75, "learning_rate": 7.492850333651096e-08, "loss": 0.8117, "step": 51190 }, { "epoch": 9.75, "learning_rate": 7.487130600571973e-08, "loss": 1.019, "step": 51191 }, { "epoch": 9.75, "learning_rate": 7.48141086749285e-08, "loss": 0.9767, "step": 51192 }, { "epoch": 9.75, "learning_rate": 7.475691134413728e-08, "loss": 1.9282, "step": 51193 }, { "epoch": 9.75, "learning_rate": 7.469971401334605e-08, "loss": 1.4561, "step": 51194 }, { "epoch": 9.75, "learning_rate": 7.464251668255481e-08, "loss": 2.9037, "step": 51195 }, { "epoch": 9.75, "learning_rate": 7.458531935176359e-08, "loss": 2.2227, "step": 51196 }, { "epoch": 9.75, "learning_rate": 7.452812202097236e-08, "loss": 1.2778, "step": 51197 }, { "epoch": 9.75, "learning_rate": 7.447092469018113e-08, "loss": 2.1775, "step": 51198 }, { "epoch": 9.75, "learning_rate": 7.44137273593899e-08, "loss": 1.339, "step": 51199 }, { "epoch": 9.75, "learning_rate": 7.435653002859866e-08, "loss": 0.8771, "step": 51200 }, { "epoch": 9.75, "learning_rate": 7.429933269780743e-08, "loss": 1.3641, "step": 51201 }, { "epoch": 9.75, "learning_rate": 7.424213536701621e-08, "loss": 1.2062, "step": 51202 }, { "epoch": 9.75, "learning_rate": 7.418493803622498e-08, "loss": 1.3685, "step": 51203 }, { "epoch": 9.75, "learning_rate": 7.412774070543374e-08, "loss": 2.4868, "step": 51204 }, { "epoch": 9.75, "learning_rate": 7.407054337464253e-08, "loss": 2.0289, "step": 51205 }, { "epoch": 9.75, "learning_rate": 7.401334604385129e-08, "loss": 3.3235, "step": 51206 }, { "epoch": 9.75, "learning_rate": 7.395614871306006e-08, "loss": 0.3474, "step": 51207 }, { "epoch": 9.75, "learning_rate": 7.389895138226883e-08, "loss": 0.9354, "step": 51208 }, { "epoch": 9.75, "learning_rate": 7.38417540514776e-08, "loss": 0.49, "step": 51209 }, { "epoch": 9.75, "learning_rate": 7.378455672068636e-08, "loss": 2.3924, "step": 51210 }, { "epoch": 9.75, "learning_rate": 7.372735938989514e-08, "loss": 1.1546, "step": 51211 }, { "epoch": 9.75, "learning_rate": 7.367016205910391e-08, "loss": 2.2972, "step": 51212 }, { "epoch": 9.75, "learning_rate": 7.361296472831268e-08, "loss": 1.3932, "step": 51213 }, { "epoch": 9.76, "learning_rate": 7.355576739752144e-08, "loss": 1.2068, "step": 51214 }, { "epoch": 9.76, "learning_rate": 7.349857006673023e-08, "loss": 1.2101, "step": 51215 }, { "epoch": 9.76, "learning_rate": 7.344137273593899e-08, "loss": 0.6133, "step": 51216 }, { "epoch": 9.76, "learning_rate": 7.338417540514776e-08, "loss": 2.3532, "step": 51217 }, { "epoch": 9.76, "learning_rate": 7.332697807435654e-08, "loss": 1.9003, "step": 51218 }, { "epoch": 9.76, "learning_rate": 7.32697807435653e-08, "loss": 1.5818, "step": 51219 }, { "epoch": 9.76, "learning_rate": 7.321258341277406e-08, "loss": 1.0413, "step": 51220 }, { "epoch": 9.76, "learning_rate": 7.315538608198284e-08, "loss": 1.2861, "step": 51221 }, { "epoch": 9.76, "learning_rate": 7.309818875119161e-08, "loss": 0.709, "step": 51222 }, { "epoch": 9.76, "learning_rate": 7.304099142040038e-08, "loss": 0.7641, "step": 51223 }, { "epoch": 9.76, "learning_rate": 7.298379408960916e-08, "loss": 2.1747, "step": 51224 }, { "epoch": 9.76, "learning_rate": 7.292659675881793e-08, "loss": 2.0244, "step": 51225 }, { "epoch": 9.76, "learning_rate": 7.286939942802669e-08, "loss": 2.0677, "step": 51226 }, { "epoch": 9.76, "learning_rate": 7.281220209723547e-08, "loss": 1.3225, "step": 51227 }, { "epoch": 9.76, "learning_rate": 7.275500476644424e-08, "loss": 0.7378, "step": 51228 }, { "epoch": 9.76, "learning_rate": 7.2697807435653e-08, "loss": 2.433, "step": 51229 }, { "epoch": 9.76, "learning_rate": 7.264061010486178e-08, "loss": 1.2758, "step": 51230 }, { "epoch": 9.76, "learning_rate": 7.258341277407054e-08, "loss": 1.9308, "step": 51231 }, { "epoch": 9.76, "learning_rate": 7.252621544327931e-08, "loss": 1.5069, "step": 51232 }, { "epoch": 9.76, "learning_rate": 7.246901811248809e-08, "loss": 1.3184, "step": 51233 }, { "epoch": 9.76, "learning_rate": 7.241182078169686e-08, "loss": 2.2322, "step": 51234 }, { "epoch": 9.76, "learning_rate": 7.235462345090563e-08, "loss": 2.4023, "step": 51235 }, { "epoch": 9.76, "learning_rate": 7.22974261201144e-08, "loss": 1.1575, "step": 51236 }, { "epoch": 9.76, "learning_rate": 7.224022878932317e-08, "loss": 2.195, "step": 51237 }, { "epoch": 9.76, "learning_rate": 7.218303145853194e-08, "loss": 1.127, "step": 51238 }, { "epoch": 9.76, "learning_rate": 7.212583412774071e-08, "loss": 4.0548, "step": 51239 }, { "epoch": 9.76, "learning_rate": 7.206863679694948e-08, "loss": 2.0438, "step": 51240 }, { "epoch": 9.76, "learning_rate": 7.201143946615824e-08, "loss": 1.2155, "step": 51241 }, { "epoch": 9.76, "learning_rate": 7.195424213536702e-08, "loss": 2.1661, "step": 51242 }, { "epoch": 9.76, "learning_rate": 7.189704480457579e-08, "loss": 0.7635, "step": 51243 }, { "epoch": 9.76, "learning_rate": 7.183984747378456e-08, "loss": 1.1973, "step": 51244 }, { "epoch": 9.76, "learning_rate": 7.178265014299333e-08, "loss": 2.9905, "step": 51245 }, { "epoch": 9.76, "learning_rate": 7.17254528122021e-08, "loss": 1.2584, "step": 51246 }, { "epoch": 9.76, "learning_rate": 7.166825548141087e-08, "loss": 2.1986, "step": 51247 }, { "epoch": 9.76, "learning_rate": 7.161105815061964e-08, "loss": 0.8788, "step": 51248 }, { "epoch": 9.76, "learning_rate": 7.155386081982841e-08, "loss": 2.0656, "step": 51249 }, { "epoch": 9.76, "learning_rate": 7.149666348903718e-08, "loss": 1.7609, "step": 51250 }, { "epoch": 9.76, "learning_rate": 7.143946615824594e-08, "loss": 2.6424, "step": 51251 }, { "epoch": 9.76, "learning_rate": 7.138226882745472e-08, "loss": 1.3547, "step": 51252 }, { "epoch": 9.76, "learning_rate": 7.132507149666349e-08, "loss": 0.57, "step": 51253 }, { "epoch": 9.76, "learning_rate": 7.126787416587226e-08, "loss": 0.7464, "step": 51254 }, { "epoch": 9.76, "learning_rate": 7.121067683508104e-08, "loss": 1.2525, "step": 51255 }, { "epoch": 9.76, "learning_rate": 7.11534795042898e-08, "loss": 0.3494, "step": 51256 }, { "epoch": 9.76, "learning_rate": 7.109628217349857e-08, "loss": 1.8341, "step": 51257 }, { "epoch": 9.76, "learning_rate": 7.103908484270734e-08, "loss": 1.6724, "step": 51258 }, { "epoch": 9.76, "learning_rate": 7.098188751191611e-08, "loss": 2.0308, "step": 51259 }, { "epoch": 9.76, "learning_rate": 7.092469018112488e-08, "loss": 1.9787, "step": 51260 }, { "epoch": 9.76, "learning_rate": 7.086749285033366e-08, "loss": 1.2125, "step": 51261 }, { "epoch": 9.76, "learning_rate": 7.081029551954242e-08, "loss": 3.1506, "step": 51262 }, { "epoch": 9.76, "learning_rate": 7.075309818875119e-08, "loss": 1.4594, "step": 51263 }, { "epoch": 9.76, "learning_rate": 7.069590085795997e-08, "loss": 0.5899, "step": 51264 }, { "epoch": 9.76, "learning_rate": 7.063870352716874e-08, "loss": 1.4934, "step": 51265 }, { "epoch": 9.76, "learning_rate": 7.05815061963775e-08, "loss": 1.6467, "step": 51266 }, { "epoch": 9.77, "learning_rate": 7.052430886558627e-08, "loss": 0.5446, "step": 51267 }, { "epoch": 9.77, "learning_rate": 7.046711153479504e-08, "loss": 0.8022, "step": 51268 }, { "epoch": 9.77, "learning_rate": 7.040991420400381e-08, "loss": 0.9241, "step": 51269 }, { "epoch": 9.77, "learning_rate": 7.035271687321259e-08, "loss": 1.0217, "step": 51270 }, { "epoch": 9.77, "learning_rate": 7.029551954242136e-08, "loss": 1.0565, "step": 51271 }, { "epoch": 9.77, "learning_rate": 7.023832221163012e-08, "loss": 1.276, "step": 51272 }, { "epoch": 9.77, "learning_rate": 7.01811248808389e-08, "loss": 2.1359, "step": 51273 }, { "epoch": 9.77, "learning_rate": 7.012392755004767e-08, "loss": 1.0899, "step": 51274 }, { "epoch": 9.77, "learning_rate": 7.006673021925644e-08, "loss": 1.9053, "step": 51275 }, { "epoch": 9.77, "learning_rate": 7.00095328884652e-08, "loss": 1.5839, "step": 51276 }, { "epoch": 9.77, "learning_rate": 6.995233555767397e-08, "loss": 1.2797, "step": 51277 }, { "epoch": 9.77, "learning_rate": 6.989513822688274e-08, "loss": 0.6845, "step": 51278 }, { "epoch": 9.77, "learning_rate": 6.983794089609151e-08, "loss": 2.0237, "step": 51279 }, { "epoch": 9.77, "learning_rate": 6.978074356530029e-08, "loss": 0.8639, "step": 51280 }, { "epoch": 9.77, "learning_rate": 6.972354623450906e-08, "loss": 0.5584, "step": 51281 }, { "epoch": 9.77, "learning_rate": 6.966634890371782e-08, "loss": 1.6811, "step": 51282 }, { "epoch": 9.77, "learning_rate": 6.96091515729266e-08, "loss": 1.977, "step": 51283 }, { "epoch": 9.77, "learning_rate": 6.955195424213537e-08, "loss": 1.3631, "step": 51284 }, { "epoch": 9.77, "learning_rate": 6.949475691134414e-08, "loss": 0.4688, "step": 51285 }, { "epoch": 9.77, "learning_rate": 6.943755958055292e-08, "loss": 2.0223, "step": 51286 }, { "epoch": 9.77, "learning_rate": 6.938036224976167e-08, "loss": 1.9283, "step": 51287 }, { "epoch": 9.77, "learning_rate": 6.932316491897044e-08, "loss": 1.1372, "step": 51288 }, { "epoch": 9.77, "learning_rate": 6.926596758817922e-08, "loss": 2.6619, "step": 51289 }, { "epoch": 9.77, "learning_rate": 6.920877025738799e-08, "loss": 0.5879, "step": 51290 }, { "epoch": 9.77, "learning_rate": 6.915157292659676e-08, "loss": 1.8676, "step": 51291 }, { "epoch": 9.77, "learning_rate": 6.909437559580554e-08, "loss": 1.3308, "step": 51292 }, { "epoch": 9.77, "learning_rate": 6.90371782650143e-08, "loss": 1.4426, "step": 51293 }, { "epoch": 9.77, "learning_rate": 6.897998093422307e-08, "loss": 1.6708, "step": 51294 }, { "epoch": 9.77, "learning_rate": 6.892278360343185e-08, "loss": 1.6505, "step": 51295 }, { "epoch": 9.77, "learning_rate": 6.886558627264062e-08, "loss": 0.5287, "step": 51296 }, { "epoch": 9.77, "learning_rate": 6.880838894184937e-08, "loss": 1.181, "step": 51297 }, { "epoch": 9.77, "learning_rate": 6.875119161105815e-08, "loss": 0.8816, "step": 51298 }, { "epoch": 9.77, "learning_rate": 6.869399428026692e-08, "loss": 2.6663, "step": 51299 }, { "epoch": 9.77, "learning_rate": 6.863679694947569e-08, "loss": 2.9737, "step": 51300 }, { "epoch": 9.77, "learning_rate": 6.857959961868447e-08, "loss": 1.6804, "step": 51301 }, { "epoch": 9.77, "learning_rate": 6.852240228789324e-08, "loss": 1.5743, "step": 51302 }, { "epoch": 9.77, "learning_rate": 6.8465204957102e-08, "loss": 1.6742, "step": 51303 }, { "epoch": 9.77, "learning_rate": 6.840800762631077e-08, "loss": 2.4122, "step": 51304 }, { "epoch": 9.77, "learning_rate": 6.835081029551955e-08, "loss": 2.996, "step": 51305 }, { "epoch": 9.77, "learning_rate": 6.829361296472832e-08, "loss": 1.1876, "step": 51306 }, { "epoch": 9.77, "learning_rate": 6.823641563393709e-08, "loss": 2.9739, "step": 51307 }, { "epoch": 9.77, "learning_rate": 6.817921830314585e-08, "loss": 2.0748, "step": 51308 }, { "epoch": 9.77, "learning_rate": 6.812202097235462e-08, "loss": 2.0776, "step": 51309 }, { "epoch": 9.77, "learning_rate": 6.806482364156339e-08, "loss": 0.2068, "step": 51310 }, { "epoch": 9.77, "learning_rate": 6.800762631077217e-08, "loss": 1.1921, "step": 51311 }, { "epoch": 9.77, "learning_rate": 6.795042897998094e-08, "loss": 1.8169, "step": 51312 }, { "epoch": 9.77, "learning_rate": 6.78932316491897e-08, "loss": 1.5796, "step": 51313 }, { "epoch": 9.77, "learning_rate": 6.783603431839848e-08, "loss": 1.4497, "step": 51314 }, { "epoch": 9.77, "learning_rate": 6.777883698760725e-08, "loss": 1.4834, "step": 51315 }, { "epoch": 9.77, "learning_rate": 6.772163965681602e-08, "loss": 1.4985, "step": 51316 }, { "epoch": 9.77, "learning_rate": 6.766444232602479e-08, "loss": 0.8037, "step": 51317 }, { "epoch": 9.77, "learning_rate": 6.760724499523355e-08, "loss": 1.4958, "step": 51318 }, { "epoch": 9.78, "learning_rate": 6.755004766444232e-08, "loss": 1.7111, "step": 51319 }, { "epoch": 9.78, "learning_rate": 6.74928503336511e-08, "loss": 2.0129, "step": 51320 }, { "epoch": 9.78, "learning_rate": 6.743565300285987e-08, "loss": 1.9144, "step": 51321 }, { "epoch": 9.78, "learning_rate": 6.737845567206864e-08, "loss": 1.7005, "step": 51322 }, { "epoch": 9.78, "learning_rate": 6.732125834127742e-08, "loss": 1.2592, "step": 51323 }, { "epoch": 9.78, "learning_rate": 6.726406101048618e-08, "loss": 0.7144, "step": 51324 }, { "epoch": 9.78, "learning_rate": 6.720686367969495e-08, "loss": 1.6936, "step": 51325 }, { "epoch": 9.78, "learning_rate": 6.714966634890372e-08, "loss": 1.447, "step": 51326 }, { "epoch": 9.78, "learning_rate": 6.709246901811249e-08, "loss": 1.2212, "step": 51327 }, { "epoch": 9.78, "learning_rate": 6.703527168732125e-08, "loss": 2.4036, "step": 51328 }, { "epoch": 9.78, "learning_rate": 6.697807435653003e-08, "loss": 1.1249, "step": 51329 }, { "epoch": 9.78, "learning_rate": 6.69208770257388e-08, "loss": 0.917, "step": 51330 }, { "epoch": 9.78, "learning_rate": 6.686367969494757e-08, "loss": 1.8304, "step": 51331 }, { "epoch": 9.78, "learning_rate": 6.680648236415635e-08, "loss": 1.9839, "step": 51332 }, { "epoch": 9.78, "learning_rate": 6.674928503336512e-08, "loss": 1.6376, "step": 51333 }, { "epoch": 9.78, "learning_rate": 6.669208770257388e-08, "loss": 1.4124, "step": 51334 }, { "epoch": 9.78, "learning_rate": 6.663489037178265e-08, "loss": 0.7315, "step": 51335 }, { "epoch": 9.78, "learning_rate": 6.657769304099142e-08, "loss": 1.2548, "step": 51336 }, { "epoch": 9.78, "learning_rate": 6.652049571020019e-08, "loss": 1.5373, "step": 51337 }, { "epoch": 9.78, "learning_rate": 6.646329837940895e-08, "loss": 1.3713, "step": 51338 }, { "epoch": 9.78, "learning_rate": 6.640610104861773e-08, "loss": 2.413, "step": 51339 }, { "epoch": 9.78, "learning_rate": 6.63489037178265e-08, "loss": 3.2819, "step": 51340 }, { "epoch": 9.78, "learning_rate": 6.629170638703527e-08, "loss": 2.1154, "step": 51341 }, { "epoch": 9.78, "learning_rate": 6.623450905624405e-08, "loss": 2.0199, "step": 51342 }, { "epoch": 9.78, "learning_rate": 6.617731172545282e-08, "loss": 1.3271, "step": 51343 }, { "epoch": 9.78, "learning_rate": 6.612011439466158e-08, "loss": 2.0193, "step": 51344 }, { "epoch": 9.78, "learning_rate": 6.606291706387035e-08, "loss": 0.9185, "step": 51345 }, { "epoch": 9.78, "learning_rate": 6.600571973307912e-08, "loss": 1.3973, "step": 51346 }, { "epoch": 9.78, "learning_rate": 6.594852240228789e-08, "loss": 1.7697, "step": 51347 }, { "epoch": 9.78, "learning_rate": 6.589132507149667e-08, "loss": 1.5909, "step": 51348 }, { "epoch": 9.78, "learning_rate": 6.583412774070543e-08, "loss": 1.4015, "step": 51349 }, { "epoch": 9.78, "learning_rate": 6.57769304099142e-08, "loss": 2.5259, "step": 51350 }, { "epoch": 9.78, "learning_rate": 6.571973307912298e-08, "loss": 2.5378, "step": 51351 }, { "epoch": 9.78, "learning_rate": 6.566253574833175e-08, "loss": 1.222, "step": 51352 }, { "epoch": 9.78, "learning_rate": 6.560533841754052e-08, "loss": 1.1818, "step": 51353 }, { "epoch": 9.78, "learning_rate": 6.55481410867493e-08, "loss": 1.3519, "step": 51354 }, { "epoch": 9.78, "learning_rate": 6.549094375595805e-08, "loss": 1.0227, "step": 51355 }, { "epoch": 9.78, "learning_rate": 6.543374642516682e-08, "loss": 1.6564, "step": 51356 }, { "epoch": 9.78, "learning_rate": 6.53765490943756e-08, "loss": 1.0247, "step": 51357 }, { "epoch": 9.78, "learning_rate": 6.531935176358437e-08, "loss": 1.4986, "step": 51358 }, { "epoch": 9.78, "learning_rate": 6.526215443279313e-08, "loss": 0.6384, "step": 51359 }, { "epoch": 9.78, "learning_rate": 6.520495710200191e-08, "loss": 1.8324, "step": 51360 }, { "epoch": 9.78, "learning_rate": 6.514775977121068e-08, "loss": 0.8361, "step": 51361 }, { "epoch": 9.78, "learning_rate": 6.509056244041945e-08, "loss": 0.3349, "step": 51362 }, { "epoch": 9.78, "learning_rate": 6.503336510962823e-08, "loss": 0.787, "step": 51363 }, { "epoch": 9.78, "learning_rate": 6.4976167778837e-08, "loss": 2.8343, "step": 51364 }, { "epoch": 9.78, "learning_rate": 6.491897044804575e-08, "loss": 1.0487, "step": 51365 }, { "epoch": 9.78, "learning_rate": 6.486177311725452e-08, "loss": 1.0942, "step": 51366 }, { "epoch": 9.78, "learning_rate": 6.48045757864633e-08, "loss": 0.6618, "step": 51367 }, { "epoch": 9.78, "learning_rate": 6.474737845567207e-08, "loss": 2.0164, "step": 51368 }, { "epoch": 9.78, "learning_rate": 6.469018112488083e-08, "loss": 1.5514, "step": 51369 }, { "epoch": 9.78, "learning_rate": 6.463298379408961e-08, "loss": 1.0419, "step": 51370 }, { "epoch": 9.78, "learning_rate": 6.457578646329838e-08, "loss": 1.9424, "step": 51371 }, { "epoch": 9.79, "learning_rate": 6.451858913250715e-08, "loss": 2.4544, "step": 51372 }, { "epoch": 9.79, "learning_rate": 6.446139180171593e-08, "loss": 1.2536, "step": 51373 }, { "epoch": 9.79, "learning_rate": 6.44041944709247e-08, "loss": 1.1065, "step": 51374 }, { "epoch": 9.79, "learning_rate": 6.434699714013346e-08, "loss": 0.8889, "step": 51375 }, { "epoch": 9.79, "learning_rate": 6.428979980934223e-08, "loss": 1.5669, "step": 51376 }, { "epoch": 9.79, "learning_rate": 6.4232602478551e-08, "loss": 0.5499, "step": 51377 }, { "epoch": 9.79, "learning_rate": 6.417540514775977e-08, "loss": 1.148, "step": 51378 }, { "epoch": 9.79, "learning_rate": 6.411820781696855e-08, "loss": 1.6772, "step": 51379 }, { "epoch": 9.79, "learning_rate": 6.406101048617731e-08, "loss": 1.5852, "step": 51380 }, { "epoch": 9.79, "learning_rate": 6.400381315538608e-08, "loss": 2.7483, "step": 51381 }, { "epoch": 9.79, "learning_rate": 6.394661582459486e-08, "loss": 0.8132, "step": 51382 }, { "epoch": 9.79, "learning_rate": 6.388941849380363e-08, "loss": 1.5175, "step": 51383 }, { "epoch": 9.79, "learning_rate": 6.38322211630124e-08, "loss": 2.3319, "step": 51384 }, { "epoch": 9.79, "learning_rate": 6.377502383222116e-08, "loss": 1.9422, "step": 51385 }, { "epoch": 9.79, "learning_rate": 6.371782650142993e-08, "loss": 1.5356, "step": 51386 }, { "epoch": 9.79, "learning_rate": 6.36606291706387e-08, "loss": 0.4094, "step": 51387 }, { "epoch": 9.79, "learning_rate": 6.360343183984748e-08, "loss": 1.2651, "step": 51388 }, { "epoch": 9.79, "learning_rate": 6.354623450905625e-08, "loss": 1.3783, "step": 51389 }, { "epoch": 9.79, "learning_rate": 6.348903717826501e-08, "loss": 1.6784, "step": 51390 }, { "epoch": 9.79, "learning_rate": 6.34318398474738e-08, "loss": 1.7685, "step": 51391 }, { "epoch": 9.79, "learning_rate": 6.337464251668256e-08, "loss": 3.3089, "step": 51392 }, { "epoch": 9.79, "learning_rate": 6.331744518589133e-08, "loss": 1.5958, "step": 51393 }, { "epoch": 9.79, "learning_rate": 6.32602478551001e-08, "loss": 1.5237, "step": 51394 }, { "epoch": 9.79, "learning_rate": 6.320305052430886e-08, "loss": 1.8066, "step": 51395 }, { "epoch": 9.79, "learning_rate": 6.314585319351763e-08, "loss": 2.548, "step": 51396 }, { "epoch": 9.79, "learning_rate": 6.30886558627264e-08, "loss": 1.0634, "step": 51397 }, { "epoch": 9.79, "learning_rate": 6.303145853193518e-08, "loss": 2.0354, "step": 51398 }, { "epoch": 9.79, "learning_rate": 6.297426120114395e-08, "loss": 0.8766, "step": 51399 }, { "epoch": 9.79, "learning_rate": 6.291706387035271e-08, "loss": 1.5449, "step": 51400 }, { "epoch": 9.79, "learning_rate": 6.28598665395615e-08, "loss": 1.3745, "step": 51401 }, { "epoch": 9.79, "learning_rate": 6.280266920877026e-08, "loss": 0.8223, "step": 51402 }, { "epoch": 9.79, "learning_rate": 6.274547187797903e-08, "loss": 1.9075, "step": 51403 }, { "epoch": 9.79, "learning_rate": 6.26882745471878e-08, "loss": 1.5107, "step": 51404 }, { "epoch": 9.79, "learning_rate": 6.263107721639656e-08, "loss": 1.5095, "step": 51405 }, { "epoch": 9.79, "learning_rate": 6.257387988560533e-08, "loss": 1.1805, "step": 51406 }, { "epoch": 9.79, "learning_rate": 6.251668255481411e-08, "loss": 2.0787, "step": 51407 }, { "epoch": 9.79, "learning_rate": 6.245948522402288e-08, "loss": 0.4719, "step": 51408 }, { "epoch": 9.79, "learning_rate": 6.240228789323165e-08, "loss": 1.8386, "step": 51409 }, { "epoch": 9.79, "learning_rate": 6.234509056244043e-08, "loss": 1.0264, "step": 51410 }, { "epoch": 9.79, "learning_rate": 6.22878932316492e-08, "loss": 1.4774, "step": 51411 }, { "epoch": 9.79, "learning_rate": 6.223069590085796e-08, "loss": 1.0916, "step": 51412 }, { "epoch": 9.79, "learning_rate": 6.217349857006673e-08, "loss": 1.1748, "step": 51413 }, { "epoch": 9.79, "learning_rate": 6.21163012392755e-08, "loss": 2.0826, "step": 51414 }, { "epoch": 9.79, "learning_rate": 6.205910390848426e-08, "loss": 1.3793, "step": 51415 }, { "epoch": 9.79, "learning_rate": 6.200190657769304e-08, "loss": 1.005, "step": 51416 }, { "epoch": 9.79, "learning_rate": 6.194470924690181e-08, "loss": 1.0185, "step": 51417 }, { "epoch": 9.79, "learning_rate": 6.188751191611058e-08, "loss": 0.428, "step": 51418 }, { "epoch": 9.79, "learning_rate": 6.183031458531936e-08, "loss": 1.8903, "step": 51419 }, { "epoch": 9.79, "learning_rate": 6.177311725452813e-08, "loss": 1.6544, "step": 51420 }, { "epoch": 9.79, "learning_rate": 6.17159199237369e-08, "loss": 1.1666, "step": 51421 }, { "epoch": 9.79, "learning_rate": 6.165872259294568e-08, "loss": 0.9022, "step": 51422 }, { "epoch": 9.79, "learning_rate": 6.160152526215443e-08, "loss": 2.6693, "step": 51423 }, { "epoch": 9.8, "learning_rate": 6.15443279313632e-08, "loss": 1.3743, "step": 51424 }, { "epoch": 9.8, "learning_rate": 6.148713060057196e-08, "loss": 2.4105, "step": 51425 }, { "epoch": 9.8, "learning_rate": 6.142993326978074e-08, "loss": 1.1297, "step": 51426 }, { "epoch": 9.8, "learning_rate": 6.137273593898951e-08, "loss": 2.201, "step": 51427 }, { "epoch": 9.8, "learning_rate": 6.131553860819828e-08, "loss": 2.1025, "step": 51428 }, { "epoch": 9.8, "learning_rate": 6.125834127740706e-08, "loss": 0.7854, "step": 51429 }, { "epoch": 9.8, "learning_rate": 6.120114394661583e-08, "loss": 1.0626, "step": 51430 }, { "epoch": 9.8, "learning_rate": 6.11439466158246e-08, "loss": 0.9472, "step": 51431 }, { "epoch": 9.8, "learning_rate": 6.108674928503338e-08, "loss": 1.5371, "step": 51432 }, { "epoch": 9.8, "learning_rate": 6.102955195424214e-08, "loss": 1.212, "step": 51433 }, { "epoch": 9.8, "learning_rate": 6.09723546234509e-08, "loss": 0.751, "step": 51434 }, { "epoch": 9.8, "learning_rate": 6.091515729265968e-08, "loss": 1.622, "step": 51435 }, { "epoch": 9.8, "learning_rate": 6.085795996186844e-08, "loss": 1.4796, "step": 51436 }, { "epoch": 9.8, "learning_rate": 6.080076263107721e-08, "loss": 3.0524, "step": 51437 }, { "epoch": 9.8, "learning_rate": 6.074356530028599e-08, "loss": 0.974, "step": 51438 }, { "epoch": 9.8, "learning_rate": 6.068636796949476e-08, "loss": 1.7582, "step": 51439 }, { "epoch": 9.8, "learning_rate": 6.062917063870353e-08, "loss": 0.8902, "step": 51440 }, { "epoch": 9.8, "learning_rate": 6.057197330791231e-08, "loss": 0.6879, "step": 51441 }, { "epoch": 9.8, "learning_rate": 6.051477597712108e-08, "loss": 1.3847, "step": 51442 }, { "epoch": 9.8, "learning_rate": 6.045757864632984e-08, "loss": 0.7912, "step": 51443 }, { "epoch": 9.8, "learning_rate": 6.040038131553861e-08, "loss": 1.5525, "step": 51444 }, { "epoch": 9.8, "learning_rate": 6.034318398474738e-08, "loss": 2.3947, "step": 51445 }, { "epoch": 9.8, "learning_rate": 6.028598665395614e-08, "loss": 0.6006, "step": 51446 }, { "epoch": 9.8, "learning_rate": 6.022878932316493e-08, "loss": 0.2958, "step": 51447 }, { "epoch": 9.8, "learning_rate": 6.017159199237369e-08, "loss": 1.2964, "step": 51448 }, { "epoch": 9.8, "learning_rate": 6.011439466158246e-08, "loss": 1.9753, "step": 51449 }, { "epoch": 9.8, "learning_rate": 6.005719733079124e-08, "loss": 1.8071, "step": 51450 }, { "epoch": 9.8, "learning_rate": 6.000000000000001e-08, "loss": 2.5828, "step": 51451 }, { "epoch": 9.8, "learning_rate": 5.994280266920878e-08, "loss": 0.9651, "step": 51452 }, { "epoch": 9.8, "learning_rate": 5.988560533841754e-08, "loss": 0.4825, "step": 51453 }, { "epoch": 9.8, "learning_rate": 5.982840800762631e-08, "loss": 1.4904, "step": 51454 }, { "epoch": 9.8, "learning_rate": 5.977121067683508e-08, "loss": 1.1157, "step": 51455 }, { "epoch": 9.8, "learning_rate": 5.971401334604384e-08, "loss": 2.4033, "step": 51456 }, { "epoch": 9.8, "learning_rate": 5.965681601525263e-08, "loss": 0.8614, "step": 51457 }, { "epoch": 9.8, "learning_rate": 5.959961868446139e-08, "loss": 1.9467, "step": 51458 }, { "epoch": 9.8, "learning_rate": 5.954242135367016e-08, "loss": 1.0758, "step": 51459 }, { "epoch": 9.8, "learning_rate": 5.9485224022878934e-08, "loss": 2.9643, "step": 51460 }, { "epoch": 9.8, "learning_rate": 5.94280266920877e-08, "loss": 0.8527, "step": 51461 }, { "epoch": 9.8, "learning_rate": 5.937082936129647e-08, "loss": 0.9989, "step": 51462 }, { "epoch": 9.8, "learning_rate": 5.931363203050525e-08, "loss": 0.2918, "step": 51463 }, { "epoch": 9.8, "learning_rate": 5.9256434699714016e-08, "loss": 0.7078, "step": 51464 }, { "epoch": 9.8, "learning_rate": 5.9199237368922784e-08, "loss": 1.6532, "step": 51465 }, { "epoch": 9.8, "learning_rate": 5.914204003813156e-08, "loss": 1.8395, "step": 51466 }, { "epoch": 9.8, "learning_rate": 5.9084842707340325e-08, "loss": 0.9175, "step": 51467 }, { "epoch": 9.8, "learning_rate": 5.902764537654909e-08, "loss": 1.4512, "step": 51468 }, { "epoch": 9.8, "learning_rate": 5.8970448045757866e-08, "loss": 1.0217, "step": 51469 }, { "epoch": 9.8, "learning_rate": 5.8913250714966634e-08, "loss": 0.9783, "step": 51470 }, { "epoch": 9.8, "learning_rate": 5.88560533841754e-08, "loss": 1.3743, "step": 51471 }, { "epoch": 9.8, "learning_rate": 5.879885605338418e-08, "loss": 1.4564, "step": 51472 }, { "epoch": 9.8, "learning_rate": 5.874165872259295e-08, "loss": 1.1215, "step": 51473 }, { "epoch": 9.8, "learning_rate": 5.8684461391801716e-08, "loss": 1.5081, "step": 51474 }, { "epoch": 9.8, "learning_rate": 5.862726406101049e-08, "loss": 1.7689, "step": 51475 }, { "epoch": 9.8, "learning_rate": 5.857006673021926e-08, "loss": 1.5748, "step": 51476 }, { "epoch": 9.81, "learning_rate": 5.8512869399428025e-08, "loss": 0.5984, "step": 51477 }, { "epoch": 9.81, "learning_rate": 5.8455672068636806e-08, "loss": 1.8612, "step": 51478 }, { "epoch": 9.81, "learning_rate": 5.8398474737845566e-08, "loss": 1.4353, "step": 51479 }, { "epoch": 9.81, "learning_rate": 5.8341277407054334e-08, "loss": 1.6065, "step": 51480 }, { "epoch": 9.81, "learning_rate": 5.8284080076263114e-08, "loss": 2.0907, "step": 51481 }, { "epoch": 9.81, "learning_rate": 5.822688274547188e-08, "loss": 1.8679, "step": 51482 }, { "epoch": 9.81, "learning_rate": 5.816968541468065e-08, "loss": 0.9623, "step": 51483 }, { "epoch": 9.81, "learning_rate": 5.811248808388942e-08, "loss": 1.1763, "step": 51484 }, { "epoch": 9.81, "learning_rate": 5.805529075309819e-08, "loss": 1.0441, "step": 51485 }, { "epoch": 9.81, "learning_rate": 5.799809342230696e-08, "loss": 2.5635, "step": 51486 }, { "epoch": 9.81, "learning_rate": 5.7940896091515725e-08, "loss": 1.8393, "step": 51487 }, { "epoch": 9.81, "learning_rate": 5.7883698760724506e-08, "loss": 1.217, "step": 51488 }, { "epoch": 9.81, "learning_rate": 5.782650142993327e-08, "loss": 1.9087, "step": 51489 }, { "epoch": 9.81, "learning_rate": 5.7769304099142034e-08, "loss": 1.8022, "step": 51490 }, { "epoch": 9.81, "learning_rate": 5.7712106768350814e-08, "loss": 0.9441, "step": 51491 }, { "epoch": 9.81, "learning_rate": 5.765490943755958e-08, "loss": 1.0721, "step": 51492 }, { "epoch": 9.81, "learning_rate": 5.759771210676835e-08, "loss": 0.4917, "step": 51493 }, { "epoch": 9.81, "learning_rate": 5.754051477597712e-08, "loss": 1.6768, "step": 51494 }, { "epoch": 9.81, "learning_rate": 5.748331744518589e-08, "loss": 0.4527, "step": 51495 }, { "epoch": 9.81, "learning_rate": 5.742612011439466e-08, "loss": 1.1139, "step": 51496 }, { "epoch": 9.81, "learning_rate": 5.736892278360344e-08, "loss": 2.7279, "step": 51497 }, { "epoch": 9.81, "learning_rate": 5.7311725452812206e-08, "loss": 2.2228, "step": 51498 }, { "epoch": 9.81, "learning_rate": 5.725452812202097e-08, "loss": 1.1909, "step": 51499 }, { "epoch": 9.81, "learning_rate": 5.719733079122975e-08, "loss": 1.0921, "step": 51500 }, { "epoch": 9.81, "learning_rate": 5.7140133460438514e-08, "loss": 0.5139, "step": 51501 }, { "epoch": 9.81, "learning_rate": 5.708293612964728e-08, "loss": 1.7009, "step": 51502 }, { "epoch": 9.81, "learning_rate": 5.7025738798856056e-08, "loss": 1.788, "step": 51503 }, { "epoch": 9.81, "learning_rate": 5.696854146806482e-08, "loss": 1.3683, "step": 51504 }, { "epoch": 9.81, "learning_rate": 5.691134413727359e-08, "loss": 4.115, "step": 51505 }, { "epoch": 9.81, "learning_rate": 5.685414680648237e-08, "loss": 1.0057, "step": 51506 }, { "epoch": 9.81, "learning_rate": 5.679694947569114e-08, "loss": 0.4387, "step": 51507 }, { "epoch": 9.81, "learning_rate": 5.6739752144899906e-08, "loss": 0.4377, "step": 51508 }, { "epoch": 9.81, "learning_rate": 5.668255481410868e-08, "loss": 1.208, "step": 51509 }, { "epoch": 9.81, "learning_rate": 5.662535748331745e-08, "loss": 0.6647, "step": 51510 }, { "epoch": 9.81, "learning_rate": 5.6568160152526214e-08, "loss": 1.6084, "step": 51511 }, { "epoch": 9.81, "learning_rate": 5.6510962821734995e-08, "loss": 0.9326, "step": 51512 }, { "epoch": 9.81, "learning_rate": 5.6453765490943756e-08, "loss": 2.7945, "step": 51513 }, { "epoch": 9.81, "learning_rate": 5.639656816015252e-08, "loss": 2.6593, "step": 51514 }, { "epoch": 9.81, "learning_rate": 5.6339370829361303e-08, "loss": 0.6775, "step": 51515 }, { "epoch": 9.81, "learning_rate": 5.628217349857007e-08, "loss": 2.0949, "step": 51516 }, { "epoch": 9.81, "learning_rate": 5.622497616777884e-08, "loss": 1.067, "step": 51517 }, { "epoch": 9.81, "learning_rate": 5.6167778836987605e-08, "loss": 2.7866, "step": 51518 }, { "epoch": 9.81, "learning_rate": 5.611058150619638e-08, "loss": 2.2863, "step": 51519 }, { "epoch": 9.81, "learning_rate": 5.605338417540515e-08, "loss": 1.4165, "step": 51520 }, { "epoch": 9.81, "learning_rate": 5.5996186844613914e-08, "loss": 1.3159, "step": 51521 }, { "epoch": 9.81, "learning_rate": 5.5938989513822695e-08, "loss": 2.2571, "step": 51522 }, { "epoch": 9.81, "learning_rate": 5.588179218303146e-08, "loss": 0.7318, "step": 51523 }, { "epoch": 9.81, "learning_rate": 5.582459485224022e-08, "loss": 2.6251, "step": 51524 }, { "epoch": 9.81, "learning_rate": 5.5767397521449003e-08, "loss": 1.78, "step": 51525 }, { "epoch": 9.81, "learning_rate": 5.571020019065777e-08, "loss": 0.7298, "step": 51526 }, { "epoch": 9.81, "learning_rate": 5.565300285986654e-08, "loss": 2.8581, "step": 51527 }, { "epoch": 9.81, "learning_rate": 5.559580552907531e-08, "loss": 2.6496, "step": 51528 }, { "epoch": 9.82, "learning_rate": 5.553860819828408e-08, "loss": 1.8608, "step": 51529 }, { "epoch": 9.82, "learning_rate": 5.548141086749285e-08, "loss": 1.1178, "step": 51530 }, { "epoch": 9.82, "learning_rate": 5.542421353670163e-08, "loss": 1.2399, "step": 51531 }, { "epoch": 9.82, "learning_rate": 5.5367016205910395e-08, "loss": 1.8521, "step": 51532 }, { "epoch": 9.82, "learning_rate": 5.530981887511916e-08, "loss": 1.6505, "step": 51533 }, { "epoch": 9.82, "learning_rate": 5.5252621544327936e-08, "loss": 3.049, "step": 51534 }, { "epoch": 9.82, "learning_rate": 5.5195424213536703e-08, "loss": 3.4997, "step": 51535 }, { "epoch": 9.82, "learning_rate": 5.513822688274547e-08, "loss": 1.3411, "step": 51536 }, { "epoch": 9.82, "learning_rate": 5.5081029551954245e-08, "loss": 1.3345, "step": 51537 }, { "epoch": 9.82, "learning_rate": 5.502383222116301e-08, "loss": 1.1186, "step": 51538 }, { "epoch": 9.82, "learning_rate": 5.496663489037178e-08, "loss": 0.6075, "step": 51539 }, { "epoch": 9.82, "learning_rate": 5.490943755958056e-08, "loss": 1.7796, "step": 51540 }, { "epoch": 9.82, "learning_rate": 5.485224022878933e-08, "loss": 0.4801, "step": 51541 }, { "epoch": 9.82, "learning_rate": 5.4795042897998095e-08, "loss": 2.0911, "step": 51542 }, { "epoch": 9.82, "learning_rate": 5.473784556720687e-08, "loss": 2.2735, "step": 51543 }, { "epoch": 9.82, "learning_rate": 5.4680648236415636e-08, "loss": 1.3145, "step": 51544 }, { "epoch": 9.82, "learning_rate": 5.46234509056244e-08, "loss": 1.1723, "step": 51545 }, { "epoch": 9.82, "learning_rate": 5.4566253574833184e-08, "loss": 1.2282, "step": 51546 }, { "epoch": 9.82, "learning_rate": 5.450905624404195e-08, "loss": 1.6807, "step": 51547 }, { "epoch": 9.82, "learning_rate": 5.445185891325071e-08, "loss": 1.8858, "step": 51548 }, { "epoch": 9.82, "learning_rate": 5.439466158245948e-08, "loss": 0.8859, "step": 51549 }, { "epoch": 9.82, "learning_rate": 5.433746425166826e-08, "loss": 1.8485, "step": 51550 }, { "epoch": 9.82, "learning_rate": 5.428026692087703e-08, "loss": 2.1198, "step": 51551 }, { "epoch": 9.82, "learning_rate": 5.4223069590085795e-08, "loss": 0.5028, "step": 51552 }, { "epoch": 9.82, "learning_rate": 5.416587225929457e-08, "loss": 0.8022, "step": 51553 }, { "epoch": 9.82, "learning_rate": 5.4108674928503336e-08, "loss": 0.7397, "step": 51554 }, { "epoch": 9.82, "learning_rate": 5.40514775977121e-08, "loss": 1.119, "step": 51555 }, { "epoch": 9.82, "learning_rate": 5.3994280266920884e-08, "loss": 0.6008, "step": 51556 }, { "epoch": 9.82, "learning_rate": 5.393708293612965e-08, "loss": 0.8244, "step": 51557 }, { "epoch": 9.82, "learning_rate": 5.387988560533841e-08, "loss": 1.0738, "step": 51558 }, { "epoch": 9.82, "learning_rate": 5.382268827454719e-08, "loss": 2.4871, "step": 51559 }, { "epoch": 9.82, "learning_rate": 5.376549094375596e-08, "loss": 2.4344, "step": 51560 }, { "epoch": 9.82, "learning_rate": 5.370829361296473e-08, "loss": 1.245, "step": 51561 }, { "epoch": 9.82, "learning_rate": 5.36510962821735e-08, "loss": 2.6164, "step": 51562 }, { "epoch": 9.82, "learning_rate": 5.359389895138227e-08, "loss": 1.1649, "step": 51563 }, { "epoch": 9.82, "learning_rate": 5.3536701620591036e-08, "loss": 1.181, "step": 51564 }, { "epoch": 9.82, "learning_rate": 5.3479504289799816e-08, "loss": 0.4938, "step": 51565 }, { "epoch": 9.82, "learning_rate": 5.3422306959008584e-08, "loss": 1.6414, "step": 51566 }, { "epoch": 9.82, "learning_rate": 5.336510962821735e-08, "loss": 2.1325, "step": 51567 }, { "epoch": 9.82, "learning_rate": 5.3307912297426125e-08, "loss": 1.9323, "step": 51568 }, { "epoch": 9.82, "learning_rate": 5.325071496663489e-08, "loss": 1.489, "step": 51569 }, { "epoch": 9.82, "learning_rate": 5.319351763584366e-08, "loss": 1.4235, "step": 51570 }, { "epoch": 9.82, "learning_rate": 5.3136320305052434e-08, "loss": 2.1424, "step": 51571 }, { "epoch": 9.82, "learning_rate": 5.30791229742612e-08, "loss": 2.3483, "step": 51572 }, { "epoch": 9.82, "learning_rate": 5.302192564346997e-08, "loss": 1.4235, "step": 51573 }, { "epoch": 9.82, "learning_rate": 5.296472831267875e-08, "loss": 0.6603, "step": 51574 }, { "epoch": 9.82, "learning_rate": 5.2907530981887516e-08, "loss": 0.3953, "step": 51575 }, { "epoch": 9.82, "learning_rate": 5.2850333651096284e-08, "loss": 1.9047, "step": 51576 }, { "epoch": 9.82, "learning_rate": 5.279313632030505e-08, "loss": 1.5018, "step": 51577 }, { "epoch": 9.82, "learning_rate": 5.2735938989513825e-08, "loss": 0.755, "step": 51578 }, { "epoch": 9.82, "learning_rate": 5.267874165872259e-08, "loss": 1.0648, "step": 51579 }, { "epoch": 9.82, "learning_rate": 5.262154432793136e-08, "loss": 1.2571, "step": 51580 }, { "epoch": 9.82, "learning_rate": 5.256434699714014e-08, "loss": 1.7666, "step": 51581 }, { "epoch": 9.83, "learning_rate": 5.25071496663489e-08, "loss": 1.2373, "step": 51582 }, { "epoch": 9.83, "learning_rate": 5.244995233555767e-08, "loss": 2.5729, "step": 51583 }, { "epoch": 9.83, "learning_rate": 5.239275500476645e-08, "loss": 0.5816, "step": 51584 }, { "epoch": 9.83, "learning_rate": 5.2335557673975216e-08, "loss": 1.004, "step": 51585 }, { "epoch": 9.83, "learning_rate": 5.2278360343183984e-08, "loss": 0.9532, "step": 51586 }, { "epoch": 9.83, "learning_rate": 5.222116301239276e-08, "loss": 0.8321, "step": 51587 }, { "epoch": 9.83, "learning_rate": 5.2163965681601525e-08, "loss": 0.7241, "step": 51588 }, { "epoch": 9.83, "learning_rate": 5.210676835081029e-08, "loss": 0.4754, "step": 51589 }, { "epoch": 9.83, "learning_rate": 5.204957102001907e-08, "loss": 2.2531, "step": 51590 }, { "epoch": 9.83, "learning_rate": 5.199237368922784e-08, "loss": 1.0559, "step": 51591 }, { "epoch": 9.83, "learning_rate": 5.19351763584366e-08, "loss": 3.5915, "step": 51592 }, { "epoch": 9.83, "learning_rate": 5.187797902764538e-08, "loss": 0.9514, "step": 51593 }, { "epoch": 9.83, "learning_rate": 5.182078169685415e-08, "loss": 0.6778, "step": 51594 }, { "epoch": 9.83, "learning_rate": 5.1763584366062916e-08, "loss": 2.038, "step": 51595 }, { "epoch": 9.83, "learning_rate": 5.170638703527169e-08, "loss": 0.3685, "step": 51596 }, { "epoch": 9.83, "learning_rate": 5.164918970448046e-08, "loss": 0.9393, "step": 51597 }, { "epoch": 9.83, "learning_rate": 5.1591992373689225e-08, "loss": 1.6423, "step": 51598 }, { "epoch": 9.83, "learning_rate": 5.1534795042898006e-08, "loss": 0.5367, "step": 51599 }, { "epoch": 9.83, "learning_rate": 5.147759771210677e-08, "loss": 0.6548, "step": 51600 }, { "epoch": 9.83, "learning_rate": 5.142040038131554e-08, "loss": 2.4697, "step": 51601 }, { "epoch": 9.83, "learning_rate": 5.1363203050524314e-08, "loss": 1.4048, "step": 51602 }, { "epoch": 9.83, "learning_rate": 5.130600571973308e-08, "loss": 1.3665, "step": 51603 }, { "epoch": 9.83, "learning_rate": 5.124880838894185e-08, "loss": 0.8822, "step": 51604 }, { "epoch": 9.83, "learning_rate": 5.119161105815062e-08, "loss": 0.3518, "step": 51605 }, { "epoch": 9.83, "learning_rate": 5.113441372735939e-08, "loss": 0.9105, "step": 51606 }, { "epoch": 9.83, "learning_rate": 5.107721639656816e-08, "loss": 1.1205, "step": 51607 }, { "epoch": 9.83, "learning_rate": 5.1020019065776925e-08, "loss": 0.8134, "step": 51608 }, { "epoch": 9.83, "learning_rate": 5.0962821734985705e-08, "loss": 1.1568, "step": 51609 }, { "epoch": 9.83, "learning_rate": 5.090562440419447e-08, "loss": 0.7685, "step": 51610 }, { "epoch": 9.83, "learning_rate": 5.084842707340324e-08, "loss": 0.8692, "step": 51611 }, { "epoch": 9.83, "learning_rate": 5.0791229742612014e-08, "loss": 2.9372, "step": 51612 }, { "epoch": 9.83, "learning_rate": 5.073403241182078e-08, "loss": 0.8344, "step": 51613 }, { "epoch": 9.83, "learning_rate": 5.067683508102955e-08, "loss": 1.8481, "step": 51614 }, { "epoch": 9.83, "learning_rate": 5.061963775023833e-08, "loss": 1.4831, "step": 51615 }, { "epoch": 9.83, "learning_rate": 5.056244041944709e-08, "loss": 1.9575, "step": 51616 }, { "epoch": 9.83, "learning_rate": 5.050524308865586e-08, "loss": 1.5018, "step": 51617 }, { "epoch": 9.83, "learning_rate": 5.044804575786464e-08, "loss": 1.2798, "step": 51618 }, { "epoch": 9.83, "learning_rate": 5.0390848427073405e-08, "loss": 0.7183, "step": 51619 }, { "epoch": 9.83, "learning_rate": 5.033365109628217e-08, "loss": 0.8563, "step": 51620 }, { "epoch": 9.83, "learning_rate": 5.027645376549095e-08, "loss": 0.6699, "step": 51621 }, { "epoch": 9.83, "learning_rate": 5.0219256434699714e-08, "loss": 1.2604, "step": 51622 }, { "epoch": 9.83, "learning_rate": 5.016205910390848e-08, "loss": 2.1594, "step": 51623 }, { "epoch": 9.83, "learning_rate": 5.010486177311726e-08, "loss": 3.3013, "step": 51624 }, { "epoch": 9.83, "learning_rate": 5.004766444232603e-08, "loss": 1.3228, "step": 51625 }, { "epoch": 9.83, "learning_rate": 4.999046711153479e-08, "loss": 2.5661, "step": 51626 }, { "epoch": 9.83, "learning_rate": 4.993326978074357e-08, "loss": 1.2152, "step": 51627 }, { "epoch": 9.83, "learning_rate": 4.987607244995234e-08, "loss": 2.1498, "step": 51628 }, { "epoch": 9.83, "learning_rate": 4.9818875119161105e-08, "loss": 1.8025, "step": 51629 }, { "epoch": 9.83, "learning_rate": 4.976167778836988e-08, "loss": 0.5165, "step": 51630 }, { "epoch": 9.83, "learning_rate": 4.9704480457578647e-08, "loss": 2.8571, "step": 51631 }, { "epoch": 9.83, "learning_rate": 4.9647283126787414e-08, "loss": 0.991, "step": 51632 }, { "epoch": 9.83, "learning_rate": 4.9590085795996195e-08, "loss": 3.4316, "step": 51633 }, { "epoch": 9.84, "learning_rate": 4.953288846520496e-08, "loss": 2.4353, "step": 51634 }, { "epoch": 9.84, "learning_rate": 4.947569113441373e-08, "loss": 1.4166, "step": 51635 }, { "epoch": 9.84, "learning_rate": 4.94184938036225e-08, "loss": 1.4582, "step": 51636 }, { "epoch": 9.84, "learning_rate": 4.936129647283127e-08, "loss": 1.6511, "step": 51637 }, { "epoch": 9.84, "learning_rate": 4.930409914204004e-08, "loss": 1.6153, "step": 51638 }, { "epoch": 9.84, "learning_rate": 4.9246901811248805e-08, "loss": 2.6787, "step": 51639 }, { "epoch": 9.84, "learning_rate": 4.918970448045758e-08, "loss": 1.821, "step": 51640 }, { "epoch": 9.84, "learning_rate": 4.9132507149666347e-08, "loss": 1.5927, "step": 51641 }, { "epoch": 9.84, "learning_rate": 4.9075309818875114e-08, "loss": 1.7637, "step": 51642 }, { "epoch": 9.84, "learning_rate": 4.9018112488083895e-08, "loss": 1.3126, "step": 51643 }, { "epoch": 9.84, "learning_rate": 4.896091515729266e-08, "loss": 1.2732, "step": 51644 }, { "epoch": 9.84, "learning_rate": 4.890371782650143e-08, "loss": 2.7485, "step": 51645 }, { "epoch": 9.84, "learning_rate": 4.88465204957102e-08, "loss": 2.8461, "step": 51646 }, { "epoch": 9.84, "learning_rate": 4.878932316491897e-08, "loss": 0.3167, "step": 51647 }, { "epoch": 9.84, "learning_rate": 4.873212583412774e-08, "loss": 1.5876, "step": 51648 }, { "epoch": 9.84, "learning_rate": 4.867492850333652e-08, "loss": 0.6311, "step": 51649 }, { "epoch": 9.84, "learning_rate": 4.861773117254528e-08, "loss": 0.7658, "step": 51650 }, { "epoch": 9.84, "learning_rate": 4.8560533841754047e-08, "loss": 0.6566, "step": 51651 }, { "epoch": 9.84, "learning_rate": 4.850333651096283e-08, "loss": 1.9353, "step": 51652 }, { "epoch": 9.84, "learning_rate": 4.8446139180171595e-08, "loss": 1.3127, "step": 51653 }, { "epoch": 9.84, "learning_rate": 4.838894184938036e-08, "loss": 1.1385, "step": 51654 }, { "epoch": 9.84, "learning_rate": 4.8331744518589136e-08, "loss": 2.8335, "step": 51655 }, { "epoch": 9.84, "learning_rate": 4.82745471877979e-08, "loss": 1.5355, "step": 51656 }, { "epoch": 9.84, "learning_rate": 4.821734985700667e-08, "loss": 0.6122, "step": 51657 }, { "epoch": 9.84, "learning_rate": 4.816015252621545e-08, "loss": 0.8719, "step": 51658 }, { "epoch": 9.84, "learning_rate": 4.810295519542422e-08, "loss": 2.3395, "step": 51659 }, { "epoch": 9.84, "learning_rate": 4.804575786463298e-08, "loss": 0.9559, "step": 51660 }, { "epoch": 9.84, "learning_rate": 4.798856053384176e-08, "loss": 1.1176, "step": 51661 }, { "epoch": 9.84, "learning_rate": 4.793136320305053e-08, "loss": 0.8954, "step": 51662 }, { "epoch": 9.84, "learning_rate": 4.7874165872259294e-08, "loss": 1.5226, "step": 51663 }, { "epoch": 9.84, "learning_rate": 4.781696854146807e-08, "loss": 1.0251, "step": 51664 }, { "epoch": 9.84, "learning_rate": 4.7759771210676836e-08, "loss": 0.9618, "step": 51665 }, { "epoch": 9.84, "learning_rate": 4.77025738798856e-08, "loss": 1.2805, "step": 51666 }, { "epoch": 9.84, "learning_rate": 4.7645376549094384e-08, "loss": 2.1129, "step": 51667 }, { "epoch": 9.84, "learning_rate": 4.758817921830315e-08, "loss": 1.1962, "step": 51668 }, { "epoch": 9.84, "learning_rate": 4.753098188751192e-08, "loss": 1.7094, "step": 51669 }, { "epoch": 9.84, "learning_rate": 4.7473784556720686e-08, "loss": 0.9807, "step": 51670 }, { "epoch": 9.84, "learning_rate": 4.741658722592946e-08, "loss": 1.5169, "step": 51671 }, { "epoch": 9.84, "learning_rate": 4.735938989513823e-08, "loss": 0.3373, "step": 51672 }, { "epoch": 9.84, "learning_rate": 4.7302192564346994e-08, "loss": 2.8543, "step": 51673 }, { "epoch": 9.84, "learning_rate": 4.724499523355577e-08, "loss": 1.7803, "step": 51674 }, { "epoch": 9.84, "learning_rate": 4.7187797902764536e-08, "loss": 1.3105, "step": 51675 }, { "epoch": 9.84, "learning_rate": 4.71306005719733e-08, "loss": 1.4548, "step": 51676 }, { "epoch": 9.84, "learning_rate": 4.7073403241182084e-08, "loss": 2.0552, "step": 51677 }, { "epoch": 9.84, "learning_rate": 4.701620591039085e-08, "loss": 1.385, "step": 51678 }, { "epoch": 9.84, "learning_rate": 4.695900857959962e-08, "loss": 2.3265, "step": 51679 }, { "epoch": 9.84, "learning_rate": 4.690181124880839e-08, "loss": 0.4414, "step": 51680 }, { "epoch": 9.84, "learning_rate": 4.684461391801716e-08, "loss": 3.0095, "step": 51681 }, { "epoch": 9.84, "learning_rate": 4.6787416587225934e-08, "loss": 1.8146, "step": 51682 }, { "epoch": 9.84, "learning_rate": 4.67302192564347e-08, "loss": 1.8452, "step": 51683 }, { "epoch": 9.84, "learning_rate": 4.667302192564347e-08, "loss": 0.9947, "step": 51684 }, { "epoch": 9.84, "learning_rate": 4.661582459485224e-08, "loss": 1.5824, "step": 51685 }, { "epoch": 9.84, "learning_rate": 4.655862726406101e-08, "loss": 1.3663, "step": 51686 }, { "epoch": 9.85, "learning_rate": 4.6501429933269784e-08, "loss": 1.3818, "step": 51687 }, { "epoch": 9.85, "learning_rate": 4.644423260247856e-08, "loss": 0.8571, "step": 51688 }, { "epoch": 9.85, "learning_rate": 4.638703527168732e-08, "loss": 1.9345, "step": 51689 }, { "epoch": 9.85, "learning_rate": 4.632983794089609e-08, "loss": 1.4189, "step": 51690 }, { "epoch": 9.85, "learning_rate": 4.627264061010486e-08, "loss": 0.8178, "step": 51691 }, { "epoch": 9.85, "learning_rate": 4.6215443279313634e-08, "loss": 1.1409, "step": 51692 }, { "epoch": 9.85, "learning_rate": 4.615824594852241e-08, "loss": 1.5201, "step": 51693 }, { "epoch": 9.85, "learning_rate": 4.610104861773117e-08, "loss": 2.6444, "step": 51694 }, { "epoch": 9.85, "learning_rate": 4.604385128693994e-08, "loss": 1.7854, "step": 51695 }, { "epoch": 9.85, "learning_rate": 4.5986653956148716e-08, "loss": 2.1731, "step": 51696 }, { "epoch": 9.85, "learning_rate": 4.5929456625357484e-08, "loss": 1.0282, "step": 51697 }, { "epoch": 9.85, "learning_rate": 4.587225929456626e-08, "loss": 0.3402, "step": 51698 }, { "epoch": 9.85, "learning_rate": 4.5815061963775025e-08, "loss": 1.3148, "step": 51699 }, { "epoch": 9.85, "learning_rate": 4.575786463298379e-08, "loss": 1.5337, "step": 51700 }, { "epoch": 9.85, "learning_rate": 4.5700667302192566e-08, "loss": 0.4918, "step": 51701 }, { "epoch": 9.85, "learning_rate": 4.564346997140134e-08, "loss": 1.3096, "step": 51702 }, { "epoch": 9.85, "learning_rate": 4.558627264061011e-08, "loss": 1.0205, "step": 51703 }, { "epoch": 9.85, "learning_rate": 4.5529075309818875e-08, "loss": 1.3379, "step": 51704 }, { "epoch": 9.85, "learning_rate": 4.547187797902765e-08, "loss": 2.3227, "step": 51705 }, { "epoch": 9.85, "learning_rate": 4.5414680648236416e-08, "loss": 2.3639, "step": 51706 }, { "epoch": 9.85, "learning_rate": 4.535748331744519e-08, "loss": 2.2537, "step": 51707 }, { "epoch": 9.85, "learning_rate": 4.530028598665396e-08, "loss": 2.2605, "step": 51708 }, { "epoch": 9.85, "learning_rate": 4.5243088655862725e-08, "loss": 1.6238, "step": 51709 }, { "epoch": 9.85, "learning_rate": 4.51858913250715e-08, "loss": 1.5887, "step": 51710 }, { "epoch": 9.85, "learning_rate": 4.5128693994280266e-08, "loss": 1.2882, "step": 51711 }, { "epoch": 9.85, "learning_rate": 4.507149666348904e-08, "loss": 1.2804, "step": 51712 }, { "epoch": 9.85, "learning_rate": 4.501429933269781e-08, "loss": 1.328, "step": 51713 }, { "epoch": 9.85, "learning_rate": 4.4957102001906575e-08, "loss": 1.02, "step": 51714 }, { "epoch": 9.85, "learning_rate": 4.489990467111535e-08, "loss": 1.7496, "step": 51715 }, { "epoch": 9.85, "learning_rate": 4.484270734032412e-08, "loss": 1.5681, "step": 51716 }, { "epoch": 9.85, "learning_rate": 4.478551000953289e-08, "loss": 2.0744, "step": 51717 }, { "epoch": 9.85, "learning_rate": 4.472831267874166e-08, "loss": 0.9152, "step": 51718 }, { "epoch": 9.85, "learning_rate": 4.467111534795043e-08, "loss": 1.0294, "step": 51719 }, { "epoch": 9.85, "learning_rate": 4.46139180171592e-08, "loss": 0.979, "step": 51720 }, { "epoch": 9.85, "learning_rate": 4.455672068636797e-08, "loss": 0.8845, "step": 51721 }, { "epoch": 9.85, "learning_rate": 4.449952335557674e-08, "loss": 1.1507, "step": 51722 }, { "epoch": 9.85, "learning_rate": 4.444232602478551e-08, "loss": 1.6675, "step": 51723 }, { "epoch": 9.85, "learning_rate": 4.438512869399428e-08, "loss": 1.0892, "step": 51724 }, { "epoch": 9.85, "learning_rate": 4.432793136320305e-08, "loss": 0.876, "step": 51725 }, { "epoch": 9.85, "learning_rate": 4.427073403241182e-08, "loss": 1.7772, "step": 51726 }, { "epoch": 9.85, "learning_rate": 4.4213536701620597e-08, "loss": 1.4705, "step": 51727 }, { "epoch": 9.85, "learning_rate": 4.4156339370829364e-08, "loss": 1.9811, "step": 51728 }, { "epoch": 9.85, "learning_rate": 4.409914204003813e-08, "loss": 1.6536, "step": 51729 }, { "epoch": 9.85, "learning_rate": 4.4041944709246905e-08, "loss": 1.1422, "step": 51730 }, { "epoch": 9.85, "learning_rate": 4.398474737845567e-08, "loss": 4.0568, "step": 51731 }, { "epoch": 9.85, "learning_rate": 4.3927550047664447e-08, "loss": 1.3146, "step": 51732 }, { "epoch": 9.85, "learning_rate": 4.3870352716873214e-08, "loss": 1.221, "step": 51733 }, { "epoch": 9.85, "learning_rate": 4.381315538608198e-08, "loss": 1.648, "step": 51734 }, { "epoch": 9.85, "learning_rate": 4.3755958055290755e-08, "loss": 1.5334, "step": 51735 }, { "epoch": 9.85, "learning_rate": 4.369876072449952e-08, "loss": 1.487, "step": 51736 }, { "epoch": 9.85, "learning_rate": 4.3641563393708297e-08, "loss": 0.9297, "step": 51737 }, { "epoch": 9.85, "learning_rate": 4.3584366062917064e-08, "loss": 2.3517, "step": 51738 }, { "epoch": 9.86, "learning_rate": 4.352716873212583e-08, "loss": 1.719, "step": 51739 }, { "epoch": 9.86, "learning_rate": 4.3469971401334605e-08, "loss": 1.0378, "step": 51740 }, { "epoch": 9.86, "learning_rate": 4.341277407054338e-08, "loss": 0.9921, "step": 51741 }, { "epoch": 9.86, "learning_rate": 4.3355576739752147e-08, "loss": 1.8219, "step": 51742 }, { "epoch": 9.86, "learning_rate": 4.3298379408960914e-08, "loss": 1.1002, "step": 51743 }, { "epoch": 9.86, "learning_rate": 4.324118207816969e-08, "loss": 1.5777, "step": 51744 }, { "epoch": 9.86, "learning_rate": 4.3183984747378455e-08, "loss": 0.5041, "step": 51745 }, { "epoch": 9.86, "learning_rate": 4.312678741658723e-08, "loss": 0.5455, "step": 51746 }, { "epoch": 9.86, "learning_rate": 4.3069590085795997e-08, "loss": 2.0762, "step": 51747 }, { "epoch": 9.86, "learning_rate": 4.3012392755004764e-08, "loss": 0.4929, "step": 51748 }, { "epoch": 9.86, "learning_rate": 4.295519542421354e-08, "loss": 1.1172, "step": 51749 }, { "epoch": 9.86, "learning_rate": 4.289799809342231e-08, "loss": 1.8309, "step": 51750 }, { "epoch": 9.86, "learning_rate": 4.284080076263108e-08, "loss": 1.3818, "step": 51751 }, { "epoch": 9.86, "learning_rate": 4.2783603431839847e-08, "loss": 1.9662, "step": 51752 }, { "epoch": 9.86, "learning_rate": 4.2726406101048614e-08, "loss": 0.8991, "step": 51753 }, { "epoch": 9.86, "learning_rate": 4.266920877025739e-08, "loss": 2.375, "step": 51754 }, { "epoch": 9.86, "learning_rate": 4.261201143946616e-08, "loss": 2.0316, "step": 51755 }, { "epoch": 9.86, "learning_rate": 4.255481410867493e-08, "loss": 1.7458, "step": 51756 }, { "epoch": 9.86, "learning_rate": 4.2497616777883696e-08, "loss": 2.7336, "step": 51757 }, { "epoch": 9.86, "learning_rate": 4.244041944709247e-08, "loss": 0.7026, "step": 51758 }, { "epoch": 9.86, "learning_rate": 4.238322211630124e-08, "loss": 1.3561, "step": 51759 }, { "epoch": 9.86, "learning_rate": 4.232602478551001e-08, "loss": 2.2428, "step": 51760 }, { "epoch": 9.86, "learning_rate": 4.2268827454718786e-08, "loss": 1.9709, "step": 51761 }, { "epoch": 9.86, "learning_rate": 4.221163012392755e-08, "loss": 0.9356, "step": 51762 }, { "epoch": 9.86, "learning_rate": 4.215443279313632e-08, "loss": 2.9409, "step": 51763 }, { "epoch": 9.86, "learning_rate": 4.2097235462345094e-08, "loss": 1.4138, "step": 51764 }, { "epoch": 9.86, "learning_rate": 4.204003813155386e-08, "loss": 0.4755, "step": 51765 }, { "epoch": 9.86, "learning_rate": 4.1982840800762636e-08, "loss": 0.7107, "step": 51766 }, { "epoch": 9.86, "learning_rate": 4.19256434699714e-08, "loss": 1.2927, "step": 51767 }, { "epoch": 9.86, "learning_rate": 4.186844613918017e-08, "loss": 1.0071, "step": 51768 }, { "epoch": 9.86, "learning_rate": 4.1811248808388944e-08, "loss": 1.3046, "step": 51769 }, { "epoch": 9.86, "learning_rate": 4.175405147759771e-08, "loss": 0.9818, "step": 51770 }, { "epoch": 9.86, "learning_rate": 4.1696854146806486e-08, "loss": 2.415, "step": 51771 }, { "epoch": 9.86, "learning_rate": 4.163965681601525e-08, "loss": 1.021, "step": 51772 }, { "epoch": 9.86, "learning_rate": 4.158245948522402e-08, "loss": 2.3971, "step": 51773 }, { "epoch": 9.86, "learning_rate": 4.1525262154432794e-08, "loss": 1.3511, "step": 51774 }, { "epoch": 9.86, "learning_rate": 4.146806482364157e-08, "loss": 1.0001, "step": 51775 }, { "epoch": 9.86, "learning_rate": 4.1410867492850336e-08, "loss": 1.6442, "step": 51776 }, { "epoch": 9.86, "learning_rate": 4.13536701620591e-08, "loss": 0.5903, "step": 51777 }, { "epoch": 9.86, "learning_rate": 4.129647283126788e-08, "loss": 2.0336, "step": 51778 }, { "epoch": 9.86, "learning_rate": 4.1239275500476644e-08, "loss": 0.5989, "step": 51779 }, { "epoch": 9.86, "learning_rate": 4.118207816968542e-08, "loss": 2.9101, "step": 51780 }, { "epoch": 9.86, "learning_rate": 4.1124880838894186e-08, "loss": 0.9165, "step": 51781 }, { "epoch": 9.86, "learning_rate": 4.106768350810295e-08, "loss": 1.8217, "step": 51782 }, { "epoch": 9.86, "learning_rate": 4.101048617731173e-08, "loss": 1.3187, "step": 51783 }, { "epoch": 9.86, "learning_rate": 4.0953288846520494e-08, "loss": 1.8932, "step": 51784 }, { "epoch": 9.86, "learning_rate": 4.089609151572927e-08, "loss": 2.4207, "step": 51785 }, { "epoch": 9.86, "learning_rate": 4.0838894184938036e-08, "loss": 2.2168, "step": 51786 }, { "epoch": 9.86, "learning_rate": 4.07816968541468e-08, "loss": 1.7674, "step": 51787 }, { "epoch": 9.86, "learning_rate": 4.072449952335558e-08, "loss": 0.9706, "step": 51788 }, { "epoch": 9.86, "learning_rate": 4.066730219256435e-08, "loss": 1.8388, "step": 51789 }, { "epoch": 9.86, "learning_rate": 4.061010486177312e-08, "loss": 1.0821, "step": 51790 }, { "epoch": 9.86, "learning_rate": 4.0552907530981886e-08, "loss": 3.4515, "step": 51791 }, { "epoch": 9.87, "learning_rate": 4.049571020019066e-08, "loss": 0.8276, "step": 51792 }, { "epoch": 9.87, "learning_rate": 4.043851286939943e-08, "loss": 1.4688, "step": 51793 }, { "epoch": 9.87, "learning_rate": 4.03813155386082e-08, "loss": 1.833, "step": 51794 }, { "epoch": 9.87, "learning_rate": 4.0324118207816975e-08, "loss": 1.1136, "step": 51795 }, { "epoch": 9.87, "learning_rate": 4.026692087702574e-08, "loss": 1.3049, "step": 51796 }, { "epoch": 9.87, "learning_rate": 4.020972354623451e-08, "loss": 1.8657, "step": 51797 }, { "epoch": 9.87, "learning_rate": 4.015252621544328e-08, "loss": 1.0903, "step": 51798 }, { "epoch": 9.87, "learning_rate": 4.009532888465205e-08, "loss": 1.3243, "step": 51799 }, { "epoch": 9.87, "learning_rate": 4.0038131553860825e-08, "loss": 1.6866, "step": 51800 }, { "epoch": 9.87, "learning_rate": 3.998093422306959e-08, "loss": 1.748, "step": 51801 }, { "epoch": 9.87, "learning_rate": 3.992373689227836e-08, "loss": 1.408, "step": 51802 }, { "epoch": 9.87, "learning_rate": 3.9866539561487133e-08, "loss": 2.4444, "step": 51803 }, { "epoch": 9.87, "learning_rate": 3.98093422306959e-08, "loss": 1.6595, "step": 51804 }, { "epoch": 9.87, "learning_rate": 3.9752144899904675e-08, "loss": 1.902, "step": 51805 }, { "epoch": 9.87, "learning_rate": 3.969494756911344e-08, "loss": 1.4122, "step": 51806 }, { "epoch": 9.87, "learning_rate": 3.963775023832221e-08, "loss": 1.1533, "step": 51807 }, { "epoch": 9.87, "learning_rate": 3.9580552907530983e-08, "loss": 1.1553, "step": 51808 }, { "epoch": 9.87, "learning_rate": 3.952335557673976e-08, "loss": 1.1949, "step": 51809 }, { "epoch": 9.87, "learning_rate": 3.9466158245948525e-08, "loss": 1.6222, "step": 51810 }, { "epoch": 9.87, "learning_rate": 3.940896091515729e-08, "loss": 2.9296, "step": 51811 }, { "epoch": 9.87, "learning_rate": 3.935176358436606e-08, "loss": 3.3129, "step": 51812 }, { "epoch": 9.87, "learning_rate": 3.9294566253574833e-08, "loss": 1.4663, "step": 51813 }, { "epoch": 9.87, "learning_rate": 3.923736892278361e-08, "loss": 2.0209, "step": 51814 }, { "epoch": 9.87, "learning_rate": 3.9180171591992375e-08, "loss": 1.6277, "step": 51815 }, { "epoch": 9.87, "learning_rate": 3.912297426120114e-08, "loss": 2.9048, "step": 51816 }, { "epoch": 9.87, "learning_rate": 3.9065776930409916e-08, "loss": 2.7342, "step": 51817 }, { "epoch": 9.87, "learning_rate": 3.9008579599618683e-08, "loss": 1.0948, "step": 51818 }, { "epoch": 9.87, "learning_rate": 3.895138226882746e-08, "loss": 1.7899, "step": 51819 }, { "epoch": 9.87, "learning_rate": 3.8894184938036225e-08, "loss": 0.8665, "step": 51820 }, { "epoch": 9.87, "learning_rate": 3.883698760724499e-08, "loss": 1.5666, "step": 51821 }, { "epoch": 9.87, "learning_rate": 3.8779790276453766e-08, "loss": 1.4935, "step": 51822 }, { "epoch": 9.87, "learning_rate": 3.872259294566254e-08, "loss": 0.7876, "step": 51823 }, { "epoch": 9.87, "learning_rate": 3.866539561487131e-08, "loss": 1.6489, "step": 51824 }, { "epoch": 9.87, "learning_rate": 3.8608198284080075e-08, "loss": 1.2748, "step": 51825 }, { "epoch": 9.87, "learning_rate": 3.855100095328885e-08, "loss": 3.0712, "step": 51826 }, { "epoch": 9.87, "learning_rate": 3.8493803622497616e-08, "loss": 1.0601, "step": 51827 }, { "epoch": 9.87, "learning_rate": 3.843660629170639e-08, "loss": 1.8167, "step": 51828 }, { "epoch": 9.87, "learning_rate": 3.837940896091516e-08, "loss": 1.1026, "step": 51829 }, { "epoch": 9.87, "learning_rate": 3.832221163012393e-08, "loss": 2.5811, "step": 51830 }, { "epoch": 9.87, "learning_rate": 3.82650142993327e-08, "loss": 2.3433, "step": 51831 }, { "epoch": 9.87, "learning_rate": 3.8207816968541466e-08, "loss": 1.3459, "step": 51832 }, { "epoch": 9.87, "learning_rate": 3.815061963775024e-08, "loss": 2.1444, "step": 51833 }, { "epoch": 9.87, "learning_rate": 3.8093422306959014e-08, "loss": 2.0635, "step": 51834 }, { "epoch": 9.87, "learning_rate": 3.803622497616778e-08, "loss": 1.6932, "step": 51835 }, { "epoch": 9.87, "learning_rate": 3.797902764537655e-08, "loss": 2.2694, "step": 51836 }, { "epoch": 9.87, "learning_rate": 3.792183031458532e-08, "loss": 2.132, "step": 51837 }, { "epoch": 9.87, "learning_rate": 3.786463298379409e-08, "loss": 2.1032, "step": 51838 }, { "epoch": 9.87, "learning_rate": 3.7807435653002864e-08, "loss": 1.6166, "step": 51839 }, { "epoch": 9.87, "learning_rate": 3.775023832221163e-08, "loss": 1.3869, "step": 51840 }, { "epoch": 9.87, "learning_rate": 3.76930409914204e-08, "loss": 2.8531, "step": 51841 }, { "epoch": 9.87, "learning_rate": 3.763584366062917e-08, "loss": 0.638, "step": 51842 }, { "epoch": 9.87, "learning_rate": 3.757864632983794e-08, "loss": 0.3672, "step": 51843 }, { "epoch": 9.88, "learning_rate": 3.7521448999046714e-08, "loss": 1.8198, "step": 51844 }, { "epoch": 9.88, "learning_rate": 3.746425166825548e-08, "loss": 1.2818, "step": 51845 }, { "epoch": 9.88, "learning_rate": 3.740705433746425e-08, "loss": 1.9258, "step": 51846 }, { "epoch": 9.88, "learning_rate": 3.734985700667302e-08, "loss": 0.9937, "step": 51847 }, { "epoch": 9.88, "learning_rate": 3.7292659675881796e-08, "loss": 1.3506, "step": 51848 }, { "epoch": 9.88, "learning_rate": 3.7235462345090564e-08, "loss": 1.2135, "step": 51849 }, { "epoch": 9.88, "learning_rate": 3.717826501429933e-08, "loss": 1.0035, "step": 51850 }, { "epoch": 9.88, "learning_rate": 3.7121067683508105e-08, "loss": 2.5925, "step": 51851 }, { "epoch": 9.88, "learning_rate": 3.706387035271687e-08, "loss": 1.6258, "step": 51852 }, { "epoch": 9.88, "learning_rate": 3.7006673021925646e-08, "loss": 1.7116, "step": 51853 }, { "epoch": 9.88, "learning_rate": 3.6949475691134414e-08, "loss": 3.0417, "step": 51854 }, { "epoch": 9.88, "learning_rate": 3.689227836034318e-08, "loss": 1.8455, "step": 51855 }, { "epoch": 9.88, "learning_rate": 3.6835081029551955e-08, "loss": 1.0629, "step": 51856 }, { "epoch": 9.88, "learning_rate": 3.677788369876072e-08, "loss": 2.1158, "step": 51857 }, { "epoch": 9.88, "learning_rate": 3.6720686367969496e-08, "loss": 0.4367, "step": 51858 }, { "epoch": 9.88, "learning_rate": 3.666348903717827e-08, "loss": 0.9302, "step": 51859 }, { "epoch": 9.88, "learning_rate": 3.660629170638703e-08, "loss": 1.2128, "step": 51860 }, { "epoch": 9.88, "learning_rate": 3.6549094375595805e-08, "loss": 2.2238, "step": 51861 }, { "epoch": 9.88, "learning_rate": 3.649189704480458e-08, "loss": 0.7392, "step": 51862 }, { "epoch": 9.88, "learning_rate": 3.6434699714013346e-08, "loss": 2.568, "step": 51863 }, { "epoch": 9.88, "learning_rate": 3.637750238322212e-08, "loss": 0.7669, "step": 51864 }, { "epoch": 9.88, "learning_rate": 3.632030505243089e-08, "loss": 0.3599, "step": 51865 }, { "epoch": 9.88, "learning_rate": 3.6263107721639655e-08, "loss": 0.2282, "step": 51866 }, { "epoch": 9.88, "learning_rate": 3.620591039084843e-08, "loss": 1.1641, "step": 51867 }, { "epoch": 9.88, "learning_rate": 3.61487130600572e-08, "loss": 1.2184, "step": 51868 }, { "epoch": 9.88, "learning_rate": 3.609151572926597e-08, "loss": 1.3767, "step": 51869 }, { "epoch": 9.88, "learning_rate": 3.603431839847474e-08, "loss": 2.2224, "step": 51870 }, { "epoch": 9.88, "learning_rate": 3.597712106768351e-08, "loss": 0.2184, "step": 51871 }, { "epoch": 9.88, "learning_rate": 3.591992373689228e-08, "loss": 2.0653, "step": 51872 }, { "epoch": 9.88, "learning_rate": 3.586272640610105e-08, "loss": 1.6776, "step": 51873 }, { "epoch": 9.88, "learning_rate": 3.580552907530982e-08, "loss": 0.7126, "step": 51874 }, { "epoch": 9.88, "learning_rate": 3.574833174451859e-08, "loss": 0.6956, "step": 51875 }, { "epoch": 9.88, "learning_rate": 3.569113441372736e-08, "loss": 1.7402, "step": 51876 }, { "epoch": 9.88, "learning_rate": 3.563393708293613e-08, "loss": 1.4238, "step": 51877 }, { "epoch": 9.88, "learning_rate": 3.55767397521449e-08, "loss": 1.268, "step": 51878 }, { "epoch": 9.88, "learning_rate": 3.551954242135367e-08, "loss": 1.9688, "step": 51879 }, { "epoch": 9.88, "learning_rate": 3.546234509056244e-08, "loss": 1.142, "step": 51880 }, { "epoch": 9.88, "learning_rate": 3.540514775977121e-08, "loss": 2.7009, "step": 51881 }, { "epoch": 9.88, "learning_rate": 3.5347950428979986e-08, "loss": 2.6246, "step": 51882 }, { "epoch": 9.88, "learning_rate": 3.529075309818875e-08, "loss": 2.2228, "step": 51883 }, { "epoch": 9.88, "learning_rate": 3.523355576739752e-08, "loss": 1.9802, "step": 51884 }, { "epoch": 9.88, "learning_rate": 3.5176358436606294e-08, "loss": 0.375, "step": 51885 }, { "epoch": 9.88, "learning_rate": 3.511916110581506e-08, "loss": 0.901, "step": 51886 }, { "epoch": 9.88, "learning_rate": 3.5061963775023836e-08, "loss": 1.8416, "step": 51887 }, { "epoch": 9.88, "learning_rate": 3.50047664442326e-08, "loss": 1.979, "step": 51888 }, { "epoch": 9.88, "learning_rate": 3.494756911344137e-08, "loss": 1.7251, "step": 51889 }, { "epoch": 9.88, "learning_rate": 3.4890371782650144e-08, "loss": 2.014, "step": 51890 }, { "epoch": 9.88, "learning_rate": 3.483317445185891e-08, "loss": 0.7131, "step": 51891 }, { "epoch": 9.88, "learning_rate": 3.4775977121067686e-08, "loss": 1.2129, "step": 51892 }, { "epoch": 9.88, "learning_rate": 3.471877979027646e-08, "loss": 2.2236, "step": 51893 }, { "epoch": 9.88, "learning_rate": 3.466158245948522e-08, "loss": 1.3467, "step": 51894 }, { "epoch": 9.88, "learning_rate": 3.4604385128693994e-08, "loss": 0.7679, "step": 51895 }, { "epoch": 9.88, "learning_rate": 3.454718779790277e-08, "loss": 2.086, "step": 51896 }, { "epoch": 9.89, "learning_rate": 3.4489990467111536e-08, "loss": 1.8759, "step": 51897 }, { "epoch": 9.89, "learning_rate": 3.443279313632031e-08, "loss": 1.4811, "step": 51898 }, { "epoch": 9.89, "learning_rate": 3.437559580552908e-08, "loss": 1.0651, "step": 51899 }, { "epoch": 9.89, "learning_rate": 3.4318398474737844e-08, "loss": 0.7818, "step": 51900 }, { "epoch": 9.89, "learning_rate": 3.426120114394662e-08, "loss": 1.0275, "step": 51901 }, { "epoch": 9.89, "learning_rate": 3.4204003813155385e-08, "loss": 2.0268, "step": 51902 }, { "epoch": 9.89, "learning_rate": 3.414680648236416e-08, "loss": 1.1758, "step": 51903 }, { "epoch": 9.89, "learning_rate": 3.408960915157293e-08, "loss": 1.5909, "step": 51904 }, { "epoch": 9.89, "learning_rate": 3.4032411820781694e-08, "loss": 1.6585, "step": 51905 }, { "epoch": 9.89, "learning_rate": 3.397521448999047e-08, "loss": 1.0252, "step": 51906 }, { "epoch": 9.89, "learning_rate": 3.391801715919924e-08, "loss": 0.2265, "step": 51907 }, { "epoch": 9.89, "learning_rate": 3.386081982840801e-08, "loss": 1.1931, "step": 51908 }, { "epoch": 9.89, "learning_rate": 3.380362249761678e-08, "loss": 2.4211, "step": 51909 }, { "epoch": 9.89, "learning_rate": 3.374642516682555e-08, "loss": 0.5468, "step": 51910 }, { "epoch": 9.89, "learning_rate": 3.368922783603432e-08, "loss": 1.2119, "step": 51911 }, { "epoch": 9.89, "learning_rate": 3.363203050524309e-08, "loss": 3.2318, "step": 51912 }, { "epoch": 9.89, "learning_rate": 3.357483317445186e-08, "loss": 1.8418, "step": 51913 }, { "epoch": 9.89, "learning_rate": 3.351763584366063e-08, "loss": 1.1683, "step": 51914 }, { "epoch": 9.89, "learning_rate": 3.34604385128694e-08, "loss": 1.6681, "step": 51915 }, { "epoch": 9.89, "learning_rate": 3.3403241182078175e-08, "loss": 1.1029, "step": 51916 }, { "epoch": 9.89, "learning_rate": 3.334604385128694e-08, "loss": 0.9411, "step": 51917 }, { "epoch": 9.89, "learning_rate": 3.328884652049571e-08, "loss": 0.6843, "step": 51918 }, { "epoch": 9.89, "learning_rate": 3.323164918970448e-08, "loss": 1.4122, "step": 51919 }, { "epoch": 9.89, "learning_rate": 3.317445185891325e-08, "loss": 0.6318, "step": 51920 }, { "epoch": 9.89, "learning_rate": 3.3117254528122025e-08, "loss": 3.519, "step": 51921 }, { "epoch": 9.89, "learning_rate": 3.306005719733079e-08, "loss": 0.3007, "step": 51922 }, { "epoch": 9.89, "learning_rate": 3.300285986653956e-08, "loss": 1.7842, "step": 51923 }, { "epoch": 9.89, "learning_rate": 3.2945662535748333e-08, "loss": 0.8155, "step": 51924 }, { "epoch": 9.89, "learning_rate": 3.28884652049571e-08, "loss": 1.3888, "step": 51925 }, { "epoch": 9.89, "learning_rate": 3.2831267874165875e-08, "loss": 1.1273, "step": 51926 }, { "epoch": 9.89, "learning_rate": 3.277407054337465e-08, "loss": 2.0425, "step": 51927 }, { "epoch": 9.89, "learning_rate": 3.271687321258341e-08, "loss": 2.1187, "step": 51928 }, { "epoch": 9.89, "learning_rate": 3.265967588179218e-08, "loss": 1.252, "step": 51929 }, { "epoch": 9.89, "learning_rate": 3.260247855100096e-08, "loss": 1.5074, "step": 51930 }, { "epoch": 9.89, "learning_rate": 3.2545281220209725e-08, "loss": 1.2174, "step": 51931 }, { "epoch": 9.89, "learning_rate": 3.24880838894185e-08, "loss": 2.1112, "step": 51932 }, { "epoch": 9.89, "learning_rate": 3.243088655862726e-08, "loss": 0.6617, "step": 51933 }, { "epoch": 9.89, "learning_rate": 3.237368922783603e-08, "loss": 1.329, "step": 51934 }, { "epoch": 9.89, "learning_rate": 3.231649189704481e-08, "loss": 1.1476, "step": 51935 }, { "epoch": 9.89, "learning_rate": 3.2259294566253575e-08, "loss": 1.9501, "step": 51936 }, { "epoch": 9.89, "learning_rate": 3.220209723546235e-08, "loss": 2.402, "step": 51937 }, { "epoch": 9.89, "learning_rate": 3.2144899904671116e-08, "loss": 0.8007, "step": 51938 }, { "epoch": 9.89, "learning_rate": 3.208770257387988e-08, "loss": 2.3855, "step": 51939 }, { "epoch": 9.89, "learning_rate": 3.203050524308866e-08, "loss": 1.2271, "step": 51940 }, { "epoch": 9.89, "learning_rate": 3.197330791229743e-08, "loss": 0.9663, "step": 51941 }, { "epoch": 9.89, "learning_rate": 3.19161105815062e-08, "loss": 1.3084, "step": 51942 }, { "epoch": 9.89, "learning_rate": 3.1858913250714966e-08, "loss": 1.645, "step": 51943 }, { "epoch": 9.89, "learning_rate": 3.180171591992374e-08, "loss": 0.6358, "step": 51944 }, { "epoch": 9.89, "learning_rate": 3.174451858913251e-08, "loss": 1.2602, "step": 51945 }, { "epoch": 9.89, "learning_rate": 3.168732125834128e-08, "loss": 0.8751, "step": 51946 }, { "epoch": 9.89, "learning_rate": 3.163012392755005e-08, "loss": 2.1405, "step": 51947 }, { "epoch": 9.89, "learning_rate": 3.1572926596758816e-08, "loss": 0.3025, "step": 51948 }, { "epoch": 9.9, "learning_rate": 3.151572926596759e-08, "loss": 1.0826, "step": 51949 }, { "epoch": 9.9, "learning_rate": 3.145853193517636e-08, "loss": 1.5893, "step": 51950 }, { "epoch": 9.9, "learning_rate": 3.140133460438513e-08, "loss": 2.5027, "step": 51951 }, { "epoch": 9.9, "learning_rate": 3.13441372735939e-08, "loss": 1.8405, "step": 51952 }, { "epoch": 9.9, "learning_rate": 3.1286939942802666e-08, "loss": 1.6054, "step": 51953 }, { "epoch": 9.9, "learning_rate": 3.122974261201144e-08, "loss": 1.2028, "step": 51954 }, { "epoch": 9.9, "learning_rate": 3.1172545281220214e-08, "loss": 0.8246, "step": 51955 }, { "epoch": 9.9, "learning_rate": 3.111534795042898e-08, "loss": 0.939, "step": 51956 }, { "epoch": 9.9, "learning_rate": 3.105815061963775e-08, "loss": 1.7382, "step": 51957 }, { "epoch": 9.9, "learning_rate": 3.100095328884652e-08, "loss": 1.1429, "step": 51958 }, { "epoch": 9.9, "learning_rate": 3.094375595805529e-08, "loss": 1.5328, "step": 51959 }, { "epoch": 9.9, "learning_rate": 3.0886558627264064e-08, "loss": 2.1362, "step": 51960 }, { "epoch": 9.9, "learning_rate": 3.082936129647284e-08, "loss": 0.9563, "step": 51961 }, { "epoch": 9.9, "learning_rate": 3.07721639656816e-08, "loss": 1.3049, "step": 51962 }, { "epoch": 9.9, "learning_rate": 3.071496663489037e-08, "loss": 1.7857, "step": 51963 }, { "epoch": 9.9, "learning_rate": 3.065776930409914e-08, "loss": 1.981, "step": 51964 }, { "epoch": 9.9, "learning_rate": 3.0600571973307914e-08, "loss": 1.7386, "step": 51965 }, { "epoch": 9.9, "learning_rate": 3.054337464251669e-08, "loss": 1.8256, "step": 51966 }, { "epoch": 9.9, "learning_rate": 3.048617731172545e-08, "loss": 1.1565, "step": 51967 }, { "epoch": 9.9, "learning_rate": 3.042897998093422e-08, "loss": 1.7823, "step": 51968 }, { "epoch": 9.9, "learning_rate": 3.0371782650142996e-08, "loss": 1.9834, "step": 51969 }, { "epoch": 9.9, "learning_rate": 3.0314585319351764e-08, "loss": 1.1327, "step": 51970 }, { "epoch": 9.9, "learning_rate": 3.025738798856054e-08, "loss": 1.7473, "step": 51971 }, { "epoch": 9.9, "learning_rate": 3.0200190657769305e-08, "loss": 1.6031, "step": 51972 }, { "epoch": 9.9, "learning_rate": 3.014299332697807e-08, "loss": 0.4577, "step": 51973 }, { "epoch": 9.9, "learning_rate": 3.0085795996186846e-08, "loss": 2.5989, "step": 51974 }, { "epoch": 9.9, "learning_rate": 3.002859866539562e-08, "loss": 1.3785, "step": 51975 }, { "epoch": 9.9, "learning_rate": 2.997140133460439e-08, "loss": 1.6767, "step": 51976 }, { "epoch": 9.9, "learning_rate": 2.9914204003813155e-08, "loss": 1.1347, "step": 51977 }, { "epoch": 9.9, "learning_rate": 2.985700667302192e-08, "loss": 0.8532, "step": 51978 }, { "epoch": 9.9, "learning_rate": 2.9799809342230696e-08, "loss": 1.16, "step": 51979 }, { "epoch": 9.9, "learning_rate": 2.9742612011439467e-08, "loss": 1.8669, "step": 51980 }, { "epoch": 9.9, "learning_rate": 2.9685414680648234e-08, "loss": 1.4568, "step": 51981 }, { "epoch": 9.9, "learning_rate": 2.9628217349857008e-08, "loss": 0.632, "step": 51982 }, { "epoch": 9.9, "learning_rate": 2.957102001906578e-08, "loss": 0.3649, "step": 51983 }, { "epoch": 9.9, "learning_rate": 2.9513822688274546e-08, "loss": 1.0602, "step": 51984 }, { "epoch": 9.9, "learning_rate": 2.9456625357483317e-08, "loss": 1.7228, "step": 51985 }, { "epoch": 9.9, "learning_rate": 2.939942802669209e-08, "loss": 1.6739, "step": 51986 }, { "epoch": 9.9, "learning_rate": 2.9342230695900858e-08, "loss": 0.7441, "step": 51987 }, { "epoch": 9.9, "learning_rate": 2.928503336510963e-08, "loss": 1.7296, "step": 51988 }, { "epoch": 9.9, "learning_rate": 2.9227836034318403e-08, "loss": 0.5008, "step": 51989 }, { "epoch": 9.9, "learning_rate": 2.9170638703527167e-08, "loss": 0.9077, "step": 51990 }, { "epoch": 9.9, "learning_rate": 2.911344137273594e-08, "loss": 1.5184, "step": 51991 }, { "epoch": 9.9, "learning_rate": 2.905624404194471e-08, "loss": 2.0003, "step": 51992 }, { "epoch": 9.9, "learning_rate": 2.899904671115348e-08, "loss": 1.8127, "step": 51993 }, { "epoch": 9.9, "learning_rate": 2.8941849380362253e-08, "loss": 1.1011, "step": 51994 }, { "epoch": 9.9, "learning_rate": 2.8884652049571017e-08, "loss": 1.7208, "step": 51995 }, { "epoch": 9.9, "learning_rate": 2.882745471877979e-08, "loss": 1.7005, "step": 51996 }, { "epoch": 9.9, "learning_rate": 2.877025738798856e-08, "loss": 1.4613, "step": 51997 }, { "epoch": 9.9, "learning_rate": 2.871306005719733e-08, "loss": 0.6409, "step": 51998 }, { "epoch": 9.9, "learning_rate": 2.8655862726406103e-08, "loss": 1.08, "step": 51999 }, { "epoch": 9.9, "learning_rate": 2.8598665395614873e-08, "loss": 1.7318, "step": 52000 }, { "epoch": 9.9, "learning_rate": 2.854146806482364e-08, "loss": 2.6514, "step": 52001 }, { "epoch": 9.91, "learning_rate": 2.848427073403241e-08, "loss": 0.8031, "step": 52002 }, { "epoch": 9.91, "learning_rate": 2.8427073403241185e-08, "loss": 0.3871, "step": 52003 }, { "epoch": 9.91, "learning_rate": 2.8369876072449953e-08, "loss": 1.8233, "step": 52004 }, { "epoch": 9.91, "learning_rate": 2.8312678741658723e-08, "loss": 2.0202, "step": 52005 }, { "epoch": 9.91, "learning_rate": 2.8255481410867497e-08, "loss": 1.9812, "step": 52006 }, { "epoch": 9.91, "learning_rate": 2.819828408007626e-08, "loss": 1.8161, "step": 52007 }, { "epoch": 9.91, "learning_rate": 2.8141086749285035e-08, "loss": 0.4395, "step": 52008 }, { "epoch": 9.91, "learning_rate": 2.8083889418493803e-08, "loss": 1.896, "step": 52009 }, { "epoch": 9.91, "learning_rate": 2.8026692087702573e-08, "loss": 0.9797, "step": 52010 }, { "epoch": 9.91, "learning_rate": 2.7969494756911347e-08, "loss": 1.8633, "step": 52011 }, { "epoch": 9.91, "learning_rate": 2.791229742612011e-08, "loss": 1.4096, "step": 52012 }, { "epoch": 9.91, "learning_rate": 2.7855100095328885e-08, "loss": 0.4448, "step": 52013 }, { "epoch": 9.91, "learning_rate": 2.7797902764537656e-08, "loss": 1.7441, "step": 52014 }, { "epoch": 9.91, "learning_rate": 2.7740705433746423e-08, "loss": 1.2231, "step": 52015 }, { "epoch": 9.91, "learning_rate": 2.7683508102955197e-08, "loss": 1.6318, "step": 52016 }, { "epoch": 9.91, "learning_rate": 2.7626310772163968e-08, "loss": 1.7842, "step": 52017 }, { "epoch": 9.91, "learning_rate": 2.7569113441372735e-08, "loss": 1.8106, "step": 52018 }, { "epoch": 9.91, "learning_rate": 2.7511916110581506e-08, "loss": 1.9204, "step": 52019 }, { "epoch": 9.91, "learning_rate": 2.745471877979028e-08, "loss": 1.7276, "step": 52020 }, { "epoch": 9.91, "learning_rate": 2.7397521448999047e-08, "loss": 1.8145, "step": 52021 }, { "epoch": 9.91, "learning_rate": 2.7340324118207818e-08, "loss": 1.3287, "step": 52022 }, { "epoch": 9.91, "learning_rate": 2.7283126787416592e-08, "loss": 2.5562, "step": 52023 }, { "epoch": 9.91, "learning_rate": 2.7225929456625356e-08, "loss": 1.1771, "step": 52024 }, { "epoch": 9.91, "learning_rate": 2.716873212583413e-08, "loss": 2.3929, "step": 52025 }, { "epoch": 9.91, "learning_rate": 2.7111534795042897e-08, "loss": 0.9471, "step": 52026 }, { "epoch": 9.91, "learning_rate": 2.7054337464251668e-08, "loss": 1.3156, "step": 52027 }, { "epoch": 9.91, "learning_rate": 2.6997140133460442e-08, "loss": 3.4965, "step": 52028 }, { "epoch": 9.91, "learning_rate": 2.6939942802669206e-08, "loss": 2.4414, "step": 52029 }, { "epoch": 9.91, "learning_rate": 2.688274547187798e-08, "loss": 2.3931, "step": 52030 }, { "epoch": 9.91, "learning_rate": 2.682554814108675e-08, "loss": 1.5671, "step": 52031 }, { "epoch": 9.91, "learning_rate": 2.6768350810295518e-08, "loss": 1.3539, "step": 52032 }, { "epoch": 9.91, "learning_rate": 2.6711153479504292e-08, "loss": 3.1368, "step": 52033 }, { "epoch": 9.91, "learning_rate": 2.6653956148713063e-08, "loss": 1.6965, "step": 52034 }, { "epoch": 9.91, "learning_rate": 2.659675881792183e-08, "loss": 1.1263, "step": 52035 }, { "epoch": 9.91, "learning_rate": 2.65395614871306e-08, "loss": 2.3845, "step": 52036 }, { "epoch": 9.91, "learning_rate": 2.6482364156339375e-08, "loss": 0.864, "step": 52037 }, { "epoch": 9.91, "learning_rate": 2.6425166825548142e-08, "loss": 0.9567, "step": 52038 }, { "epoch": 9.91, "learning_rate": 2.6367969494756913e-08, "loss": 0.8147, "step": 52039 }, { "epoch": 9.91, "learning_rate": 2.631077216396568e-08, "loss": 2.5095, "step": 52040 }, { "epoch": 9.91, "learning_rate": 2.625357483317445e-08, "loss": 0.9012, "step": 52041 }, { "epoch": 9.91, "learning_rate": 2.6196377502383224e-08, "loss": 2.598, "step": 52042 }, { "epoch": 9.91, "learning_rate": 2.6139180171591992e-08, "loss": 0.8048, "step": 52043 }, { "epoch": 9.91, "learning_rate": 2.6081982840800762e-08, "loss": 2.6846, "step": 52044 }, { "epoch": 9.91, "learning_rate": 2.6024785510009536e-08, "loss": 1.8835, "step": 52045 }, { "epoch": 9.91, "learning_rate": 2.59675881792183e-08, "loss": 2.1147, "step": 52046 }, { "epoch": 9.91, "learning_rate": 2.5910390848427074e-08, "loss": 0.9625, "step": 52047 }, { "epoch": 9.91, "learning_rate": 2.5853193517635845e-08, "loss": 1.2628, "step": 52048 }, { "epoch": 9.91, "learning_rate": 2.5795996186844612e-08, "loss": 0.6056, "step": 52049 }, { "epoch": 9.91, "learning_rate": 2.5738798856053386e-08, "loss": 3.3752, "step": 52050 }, { "epoch": 9.91, "learning_rate": 2.5681601525262157e-08, "loss": 1.8251, "step": 52051 }, { "epoch": 9.91, "learning_rate": 2.5624404194470924e-08, "loss": 1.0269, "step": 52052 }, { "epoch": 9.91, "learning_rate": 2.5567206863679695e-08, "loss": 0.432, "step": 52053 }, { "epoch": 9.92, "learning_rate": 2.5510009532888462e-08, "loss": 0.8709, "step": 52054 }, { "epoch": 9.92, "learning_rate": 2.5452812202097236e-08, "loss": 2.9195, "step": 52055 }, { "epoch": 9.92, "learning_rate": 2.5395614871306007e-08, "loss": 1.482, "step": 52056 }, { "epoch": 9.92, "learning_rate": 2.5338417540514774e-08, "loss": 0.9371, "step": 52057 }, { "epoch": 9.92, "learning_rate": 2.5281220209723545e-08, "loss": 1.9566, "step": 52058 }, { "epoch": 9.92, "learning_rate": 2.522402287893232e-08, "loss": 1.8835, "step": 52059 }, { "epoch": 9.92, "learning_rate": 2.5166825548141086e-08, "loss": 3.0264, "step": 52060 }, { "epoch": 9.92, "learning_rate": 2.5109628217349857e-08, "loss": 1.315, "step": 52061 }, { "epoch": 9.92, "learning_rate": 2.505243088655863e-08, "loss": 1.6353, "step": 52062 }, { "epoch": 9.92, "learning_rate": 2.4995233555767395e-08, "loss": 1.1717, "step": 52063 }, { "epoch": 9.92, "learning_rate": 2.493803622497617e-08, "loss": 2.9474, "step": 52064 }, { "epoch": 9.92, "learning_rate": 2.488083889418494e-08, "loss": 2.2506, "step": 52065 }, { "epoch": 9.92, "learning_rate": 2.4823641563393707e-08, "loss": 1.2628, "step": 52066 }, { "epoch": 9.92, "learning_rate": 2.476644423260248e-08, "loss": 1.5548, "step": 52067 }, { "epoch": 9.92, "learning_rate": 2.470924690181125e-08, "loss": 1.6868, "step": 52068 }, { "epoch": 9.92, "learning_rate": 2.465204957102002e-08, "loss": 1.6956, "step": 52069 }, { "epoch": 9.92, "learning_rate": 2.459485224022879e-08, "loss": 1.3876, "step": 52070 }, { "epoch": 9.92, "learning_rate": 2.4537654909437557e-08, "loss": 0.3653, "step": 52071 }, { "epoch": 9.92, "learning_rate": 2.448045757864633e-08, "loss": 2.2268, "step": 52072 }, { "epoch": 9.92, "learning_rate": 2.44232602478551e-08, "loss": 1.2329, "step": 52073 }, { "epoch": 9.92, "learning_rate": 2.436606291706387e-08, "loss": 1.6707, "step": 52074 }, { "epoch": 9.92, "learning_rate": 2.430886558627264e-08, "loss": 2.562, "step": 52075 }, { "epoch": 9.92, "learning_rate": 2.4251668255481414e-08, "loss": 1.67, "step": 52076 }, { "epoch": 9.92, "learning_rate": 2.419447092469018e-08, "loss": 1.7095, "step": 52077 }, { "epoch": 9.92, "learning_rate": 2.413727359389895e-08, "loss": 1.6019, "step": 52078 }, { "epoch": 9.92, "learning_rate": 2.4080076263107726e-08, "loss": 1.497, "step": 52079 }, { "epoch": 9.92, "learning_rate": 2.402287893231649e-08, "loss": 3.0485, "step": 52080 }, { "epoch": 9.92, "learning_rate": 2.3965681601525264e-08, "loss": 2.0289, "step": 52081 }, { "epoch": 9.92, "learning_rate": 2.3908484270734034e-08, "loss": 0.7072, "step": 52082 }, { "epoch": 9.92, "learning_rate": 2.38512869399428e-08, "loss": 1.0716, "step": 52083 }, { "epoch": 9.92, "learning_rate": 2.3794089609151576e-08, "loss": 2.6605, "step": 52084 }, { "epoch": 9.92, "learning_rate": 2.3736892278360343e-08, "loss": 0.8386, "step": 52085 }, { "epoch": 9.92, "learning_rate": 2.3679694947569114e-08, "loss": 1.1774, "step": 52086 }, { "epoch": 9.92, "learning_rate": 2.3622497616777884e-08, "loss": 0.5524, "step": 52087 }, { "epoch": 9.92, "learning_rate": 2.356530028598665e-08, "loss": 1.7115, "step": 52088 }, { "epoch": 9.92, "learning_rate": 2.3508102955195426e-08, "loss": 0.6731, "step": 52089 }, { "epoch": 9.92, "learning_rate": 2.3450905624404196e-08, "loss": 1.0548, "step": 52090 }, { "epoch": 9.92, "learning_rate": 2.3393708293612967e-08, "loss": 2.3322, "step": 52091 }, { "epoch": 9.92, "learning_rate": 2.3336510962821734e-08, "loss": 1.2596, "step": 52092 }, { "epoch": 9.92, "learning_rate": 2.3279313632030505e-08, "loss": 3.1212, "step": 52093 }, { "epoch": 9.92, "learning_rate": 2.322211630123928e-08, "loss": 2.1613, "step": 52094 }, { "epoch": 9.92, "learning_rate": 2.3164918970448046e-08, "loss": 2.5667, "step": 52095 }, { "epoch": 9.92, "learning_rate": 2.3107721639656817e-08, "loss": 1.11, "step": 52096 }, { "epoch": 9.92, "learning_rate": 2.3050524308865584e-08, "loss": 1.0797, "step": 52097 }, { "epoch": 9.92, "learning_rate": 2.2993326978074358e-08, "loss": 1.5449, "step": 52098 }, { "epoch": 9.92, "learning_rate": 2.293612964728313e-08, "loss": 0.1389, "step": 52099 }, { "epoch": 9.92, "learning_rate": 2.2878932316491896e-08, "loss": 0.8603, "step": 52100 }, { "epoch": 9.92, "learning_rate": 2.282173498570067e-08, "loss": 1.2094, "step": 52101 }, { "epoch": 9.92, "learning_rate": 2.2764537654909437e-08, "loss": 1.0348, "step": 52102 }, { "epoch": 9.92, "learning_rate": 2.2707340324118208e-08, "loss": 2.2736, "step": 52103 }, { "epoch": 9.92, "learning_rate": 2.265014299332698e-08, "loss": 1.6961, "step": 52104 }, { "epoch": 9.92, "learning_rate": 2.259294566253575e-08, "loss": 1.6484, "step": 52105 }, { "epoch": 9.92, "learning_rate": 2.253574833174452e-08, "loss": 1.7707, "step": 52106 }, { "epoch": 9.93, "learning_rate": 2.2478551000953287e-08, "loss": 1.7681, "step": 52107 }, { "epoch": 9.93, "learning_rate": 2.242135367016206e-08, "loss": 2.1913, "step": 52108 }, { "epoch": 9.93, "learning_rate": 2.236415633937083e-08, "loss": 0.8506, "step": 52109 }, { "epoch": 9.93, "learning_rate": 2.23069590085796e-08, "loss": 2.7317, "step": 52110 }, { "epoch": 9.93, "learning_rate": 2.224976167778837e-08, "loss": 1.2414, "step": 52111 }, { "epoch": 9.93, "learning_rate": 2.219256434699714e-08, "loss": 1.4934, "step": 52112 }, { "epoch": 9.93, "learning_rate": 2.213536701620591e-08, "loss": 1.6594, "step": 52113 }, { "epoch": 9.93, "learning_rate": 2.2078169685414682e-08, "loss": 1.385, "step": 52114 }, { "epoch": 9.93, "learning_rate": 2.2020972354623453e-08, "loss": 1.534, "step": 52115 }, { "epoch": 9.93, "learning_rate": 2.1963775023832223e-08, "loss": 1.4853, "step": 52116 }, { "epoch": 9.93, "learning_rate": 2.190657769304099e-08, "loss": 1.2966, "step": 52117 }, { "epoch": 9.93, "learning_rate": 2.184938036224976e-08, "loss": 1.5104, "step": 52118 }, { "epoch": 9.93, "learning_rate": 2.1792183031458532e-08, "loss": 0.9841, "step": 52119 }, { "epoch": 9.93, "learning_rate": 2.1734985700667303e-08, "loss": 0.625, "step": 52120 }, { "epoch": 9.93, "learning_rate": 2.1677788369876073e-08, "loss": 2.065, "step": 52121 }, { "epoch": 9.93, "learning_rate": 2.1620591039084844e-08, "loss": 1.1453, "step": 52122 }, { "epoch": 9.93, "learning_rate": 2.1563393708293615e-08, "loss": 2.0297, "step": 52123 }, { "epoch": 9.93, "learning_rate": 2.1506196377502382e-08, "loss": 2.5083, "step": 52124 }, { "epoch": 9.93, "learning_rate": 2.1448999046711156e-08, "loss": 1.8016, "step": 52125 }, { "epoch": 9.93, "learning_rate": 2.1391801715919923e-08, "loss": 1.1119, "step": 52126 }, { "epoch": 9.93, "learning_rate": 2.1334604385128694e-08, "loss": 1.5134, "step": 52127 }, { "epoch": 9.93, "learning_rate": 2.1277407054337465e-08, "loss": 1.2936, "step": 52128 }, { "epoch": 9.93, "learning_rate": 2.1220209723546235e-08, "loss": 1.3036, "step": 52129 }, { "epoch": 9.93, "learning_rate": 2.1163012392755006e-08, "loss": 1.888, "step": 52130 }, { "epoch": 9.93, "learning_rate": 2.1105815061963777e-08, "loss": 2.7797, "step": 52131 }, { "epoch": 9.93, "learning_rate": 2.1048617731172547e-08, "loss": 2.0292, "step": 52132 }, { "epoch": 9.93, "learning_rate": 2.0991420400381318e-08, "loss": 0.9611, "step": 52133 }, { "epoch": 9.93, "learning_rate": 2.0934223069590085e-08, "loss": 1.0303, "step": 52134 }, { "epoch": 9.93, "learning_rate": 2.0877025738798856e-08, "loss": 0.9273, "step": 52135 }, { "epoch": 9.93, "learning_rate": 2.0819828408007627e-08, "loss": 1.6453, "step": 52136 }, { "epoch": 9.93, "learning_rate": 2.0762631077216397e-08, "loss": 2.1085, "step": 52137 }, { "epoch": 9.93, "learning_rate": 2.0705433746425168e-08, "loss": 2.5215, "step": 52138 }, { "epoch": 9.93, "learning_rate": 2.064823641563394e-08, "loss": 0.7483, "step": 52139 }, { "epoch": 9.93, "learning_rate": 2.059103908484271e-08, "loss": 2.1003, "step": 52140 }, { "epoch": 9.93, "learning_rate": 2.0533841754051476e-08, "loss": 2.05, "step": 52141 }, { "epoch": 9.93, "learning_rate": 2.0476644423260247e-08, "loss": 1.4602, "step": 52142 }, { "epoch": 9.93, "learning_rate": 2.0419447092469018e-08, "loss": 2.0835, "step": 52143 }, { "epoch": 9.93, "learning_rate": 2.036224976167779e-08, "loss": 1.4675, "step": 52144 }, { "epoch": 9.93, "learning_rate": 2.030505243088656e-08, "loss": 0.7027, "step": 52145 }, { "epoch": 9.93, "learning_rate": 2.024785510009533e-08, "loss": 0.7635, "step": 52146 }, { "epoch": 9.93, "learning_rate": 2.01906577693041e-08, "loss": 2.0134, "step": 52147 }, { "epoch": 9.93, "learning_rate": 2.013346043851287e-08, "loss": 0.4952, "step": 52148 }, { "epoch": 9.93, "learning_rate": 2.007626310772164e-08, "loss": 1.232, "step": 52149 }, { "epoch": 9.93, "learning_rate": 2.0019065776930412e-08, "loss": 3.1717, "step": 52150 }, { "epoch": 9.93, "learning_rate": 1.996186844613918e-08, "loss": 1.9041, "step": 52151 }, { "epoch": 9.93, "learning_rate": 1.990467111534795e-08, "loss": 0.9855, "step": 52152 }, { "epoch": 9.93, "learning_rate": 1.984747378455672e-08, "loss": 1.8506, "step": 52153 }, { "epoch": 9.93, "learning_rate": 1.9790276453765492e-08, "loss": 0.8343, "step": 52154 }, { "epoch": 9.93, "learning_rate": 1.9733079122974262e-08, "loss": 0.3859, "step": 52155 }, { "epoch": 9.93, "learning_rate": 1.967588179218303e-08, "loss": 0.6184, "step": 52156 }, { "epoch": 9.93, "learning_rate": 1.9618684461391804e-08, "loss": 0.8777, "step": 52157 }, { "epoch": 9.93, "learning_rate": 1.956148713060057e-08, "loss": 0.6066, "step": 52158 }, { "epoch": 9.94, "learning_rate": 1.9504289799809342e-08, "loss": 0.4742, "step": 52159 }, { "epoch": 9.94, "learning_rate": 1.9447092469018112e-08, "loss": 2.7727, "step": 52160 }, { "epoch": 9.94, "learning_rate": 1.9389895138226883e-08, "loss": 0.2275, "step": 52161 }, { "epoch": 9.94, "learning_rate": 1.9332697807435654e-08, "loss": 2.1556, "step": 52162 }, { "epoch": 9.94, "learning_rate": 1.9275500476644424e-08, "loss": 1.7379, "step": 52163 }, { "epoch": 9.94, "learning_rate": 1.9218303145853195e-08, "loss": 1.9117, "step": 52164 }, { "epoch": 9.94, "learning_rate": 1.9161105815061966e-08, "loss": 1.9758, "step": 52165 }, { "epoch": 9.94, "learning_rate": 1.9103908484270733e-08, "loss": 1.5569, "step": 52166 }, { "epoch": 9.94, "learning_rate": 1.9046711153479507e-08, "loss": 1.0314, "step": 52167 }, { "epoch": 9.94, "learning_rate": 1.8989513822688274e-08, "loss": 2.3847, "step": 52168 }, { "epoch": 9.94, "learning_rate": 1.8932316491897045e-08, "loss": 0.6326, "step": 52169 }, { "epoch": 9.94, "learning_rate": 1.8875119161105816e-08, "loss": 1.9758, "step": 52170 }, { "epoch": 9.94, "learning_rate": 1.8817921830314586e-08, "loss": 0.8734, "step": 52171 }, { "epoch": 9.94, "learning_rate": 1.8760724499523357e-08, "loss": 0.9405, "step": 52172 }, { "epoch": 9.94, "learning_rate": 1.8703527168732124e-08, "loss": 1.6641, "step": 52173 }, { "epoch": 9.94, "learning_rate": 1.8646329837940898e-08, "loss": 1.5318, "step": 52174 }, { "epoch": 9.94, "learning_rate": 1.8589132507149666e-08, "loss": 0.9134, "step": 52175 }, { "epoch": 9.94, "learning_rate": 1.8531935176358436e-08, "loss": 1.8311, "step": 52176 }, { "epoch": 9.94, "learning_rate": 1.8474737845567207e-08, "loss": 0.8587, "step": 52177 }, { "epoch": 9.94, "learning_rate": 1.8417540514775978e-08, "loss": 1.9399, "step": 52178 }, { "epoch": 9.94, "learning_rate": 1.8360343183984748e-08, "loss": 1.8181, "step": 52179 }, { "epoch": 9.94, "learning_rate": 1.8303145853193516e-08, "loss": 0.7401, "step": 52180 }, { "epoch": 9.94, "learning_rate": 1.824594852240229e-08, "loss": 0.9277, "step": 52181 }, { "epoch": 9.94, "learning_rate": 1.818875119161106e-08, "loss": 0.7868, "step": 52182 }, { "epoch": 9.94, "learning_rate": 1.8131553860819828e-08, "loss": 0.985, "step": 52183 }, { "epoch": 9.94, "learning_rate": 1.80743565300286e-08, "loss": 1.2411, "step": 52184 }, { "epoch": 9.94, "learning_rate": 1.801715919923737e-08, "loss": 1.3266, "step": 52185 }, { "epoch": 9.94, "learning_rate": 1.795996186844614e-08, "loss": 1.4885, "step": 52186 }, { "epoch": 9.94, "learning_rate": 1.790276453765491e-08, "loss": 1.3985, "step": 52187 }, { "epoch": 9.94, "learning_rate": 1.784556720686368e-08, "loss": 2.2761, "step": 52188 }, { "epoch": 9.94, "learning_rate": 1.778836987607245e-08, "loss": 1.1143, "step": 52189 }, { "epoch": 9.94, "learning_rate": 1.773117254528122e-08, "loss": 1.1714, "step": 52190 }, { "epoch": 9.94, "learning_rate": 1.7673975214489993e-08, "loss": 1.605, "step": 52191 }, { "epoch": 9.94, "learning_rate": 1.761677788369876e-08, "loss": 0.8689, "step": 52192 }, { "epoch": 9.94, "learning_rate": 1.755958055290753e-08, "loss": 2.4294, "step": 52193 }, { "epoch": 9.94, "learning_rate": 1.75023832221163e-08, "loss": 1.4176, "step": 52194 }, { "epoch": 9.94, "learning_rate": 1.7445185891325072e-08, "loss": 1.7695, "step": 52195 }, { "epoch": 9.94, "learning_rate": 1.7387988560533843e-08, "loss": 2.1919, "step": 52196 }, { "epoch": 9.94, "learning_rate": 1.733079122974261e-08, "loss": 2.051, "step": 52197 }, { "epoch": 9.94, "learning_rate": 1.7273593898951384e-08, "loss": 1.5872, "step": 52198 }, { "epoch": 9.94, "learning_rate": 1.7216396568160155e-08, "loss": 2.6683, "step": 52199 }, { "epoch": 9.94, "learning_rate": 1.7159199237368922e-08, "loss": 1.8689, "step": 52200 }, { "epoch": 9.94, "learning_rate": 1.7102001906577693e-08, "loss": 1.6799, "step": 52201 }, { "epoch": 9.94, "learning_rate": 1.7044804575786463e-08, "loss": 0.7915, "step": 52202 }, { "epoch": 9.94, "learning_rate": 1.6987607244995234e-08, "loss": 0.7717, "step": 52203 }, { "epoch": 9.94, "learning_rate": 1.6930409914204005e-08, "loss": 1.9912, "step": 52204 }, { "epoch": 9.94, "learning_rate": 1.6873212583412775e-08, "loss": 1.1776, "step": 52205 }, { "epoch": 9.94, "learning_rate": 1.6816015252621546e-08, "loss": 2.4191, "step": 52206 }, { "epoch": 9.94, "learning_rate": 1.6758817921830313e-08, "loss": 1.1403, "step": 52207 }, { "epoch": 9.94, "learning_rate": 1.6701620591039087e-08, "loss": 2.4479, "step": 52208 }, { "epoch": 9.94, "learning_rate": 1.6644423260247855e-08, "loss": 0.5578, "step": 52209 }, { "epoch": 9.94, "learning_rate": 1.6587225929456625e-08, "loss": 1.3175, "step": 52210 }, { "epoch": 9.94, "learning_rate": 1.6530028598665396e-08, "loss": 1.1471, "step": 52211 }, { "epoch": 9.95, "learning_rate": 1.6472831267874167e-08, "loss": 3.5477, "step": 52212 }, { "epoch": 9.95, "learning_rate": 1.6415633937082937e-08, "loss": 1.5504, "step": 52213 }, { "epoch": 9.95, "learning_rate": 1.6358436606291705e-08, "loss": 1.2332, "step": 52214 }, { "epoch": 9.95, "learning_rate": 1.630123927550048e-08, "loss": 0.8841, "step": 52215 }, { "epoch": 9.95, "learning_rate": 1.624404194470925e-08, "loss": 0.927, "step": 52216 }, { "epoch": 9.95, "learning_rate": 1.6186844613918017e-08, "loss": 0.8487, "step": 52217 }, { "epoch": 9.95, "learning_rate": 1.6129647283126787e-08, "loss": 1.1372, "step": 52218 }, { "epoch": 9.95, "learning_rate": 1.6072449952335558e-08, "loss": 1.1346, "step": 52219 }, { "epoch": 9.95, "learning_rate": 1.601525262154433e-08, "loss": 1.2785, "step": 52220 }, { "epoch": 9.95, "learning_rate": 1.59580552907531e-08, "loss": 1.2323, "step": 52221 }, { "epoch": 9.95, "learning_rate": 1.590085795996187e-08, "loss": 2.4712, "step": 52222 }, { "epoch": 9.95, "learning_rate": 1.584366062917064e-08, "loss": 1.2984, "step": 52223 }, { "epoch": 9.95, "learning_rate": 1.5786463298379408e-08, "loss": 2.6383, "step": 52224 }, { "epoch": 9.95, "learning_rate": 1.572926596758818e-08, "loss": 0.6478, "step": 52225 }, { "epoch": 9.95, "learning_rate": 1.567206863679695e-08, "loss": 1.6368, "step": 52226 }, { "epoch": 9.95, "learning_rate": 1.561487130600572e-08, "loss": 1.4027, "step": 52227 }, { "epoch": 9.95, "learning_rate": 1.555767397521449e-08, "loss": 1.6632, "step": 52228 }, { "epoch": 9.95, "learning_rate": 1.550047664442326e-08, "loss": 1.0921, "step": 52229 }, { "epoch": 9.95, "learning_rate": 1.5443279313632032e-08, "loss": 1.593, "step": 52230 }, { "epoch": 9.95, "learning_rate": 1.53860819828408e-08, "loss": 2.5629, "step": 52231 }, { "epoch": 9.95, "learning_rate": 1.532888465204957e-08, "loss": 0.6048, "step": 52232 }, { "epoch": 9.95, "learning_rate": 1.5271687321258344e-08, "loss": 1.5203, "step": 52233 }, { "epoch": 9.95, "learning_rate": 1.521448999046711e-08, "loss": 1.4532, "step": 52234 }, { "epoch": 9.95, "learning_rate": 1.5157292659675882e-08, "loss": 1.4307, "step": 52235 }, { "epoch": 9.95, "learning_rate": 1.5100095328884652e-08, "loss": 1.8241, "step": 52236 }, { "epoch": 9.95, "learning_rate": 1.5042897998093423e-08, "loss": 1.992, "step": 52237 }, { "epoch": 9.95, "learning_rate": 1.4985700667302194e-08, "loss": 1.1224, "step": 52238 }, { "epoch": 9.95, "learning_rate": 1.492850333651096e-08, "loss": 1.1811, "step": 52239 }, { "epoch": 9.95, "learning_rate": 1.4871306005719733e-08, "loss": 0.964, "step": 52240 }, { "epoch": 9.95, "learning_rate": 1.4814108674928504e-08, "loss": 2.5361, "step": 52241 }, { "epoch": 9.95, "learning_rate": 1.4756911344137273e-08, "loss": 1.2788, "step": 52242 }, { "epoch": 9.95, "learning_rate": 1.4699714013346045e-08, "loss": 0.3045, "step": 52243 }, { "epoch": 9.95, "learning_rate": 1.4642516682554814e-08, "loss": 2.4956, "step": 52244 }, { "epoch": 9.95, "learning_rate": 1.4585319351763583e-08, "loss": 1.4882, "step": 52245 }, { "epoch": 9.95, "learning_rate": 1.4528122020972356e-08, "loss": 1.6937, "step": 52246 }, { "epoch": 9.95, "learning_rate": 1.4470924690181126e-08, "loss": 1.2847, "step": 52247 }, { "epoch": 9.95, "learning_rate": 1.4413727359389895e-08, "loss": 2.4515, "step": 52248 }, { "epoch": 9.95, "learning_rate": 1.4356530028598664e-08, "loss": 1.3707, "step": 52249 }, { "epoch": 9.95, "learning_rate": 1.4299332697807437e-08, "loss": 1.2341, "step": 52250 }, { "epoch": 9.95, "learning_rate": 1.4242135367016206e-08, "loss": 2.3258, "step": 52251 }, { "epoch": 9.95, "learning_rate": 1.4184938036224976e-08, "loss": 0.4292, "step": 52252 }, { "epoch": 9.95, "learning_rate": 1.4127740705433749e-08, "loss": 0.4694, "step": 52253 }, { "epoch": 9.95, "learning_rate": 1.4070543374642518e-08, "loss": 1.8488, "step": 52254 }, { "epoch": 9.95, "learning_rate": 1.4013346043851287e-08, "loss": 0.3459, "step": 52255 }, { "epoch": 9.95, "learning_rate": 1.3956148713060056e-08, "loss": 1.3394, "step": 52256 }, { "epoch": 9.95, "learning_rate": 1.3898951382268828e-08, "loss": 0.4518, "step": 52257 }, { "epoch": 9.95, "learning_rate": 1.3841754051477599e-08, "loss": 1.3021, "step": 52258 }, { "epoch": 9.95, "learning_rate": 1.3784556720686368e-08, "loss": 1.7129, "step": 52259 }, { "epoch": 9.95, "learning_rate": 1.372735938989514e-08, "loss": 1.6806, "step": 52260 }, { "epoch": 9.95, "learning_rate": 1.3670162059103909e-08, "loss": 0.8424, "step": 52261 }, { "epoch": 9.95, "learning_rate": 1.3612964728312678e-08, "loss": 1.0454, "step": 52262 }, { "epoch": 9.95, "learning_rate": 1.3555767397521449e-08, "loss": 2.8272, "step": 52263 }, { "epoch": 9.96, "learning_rate": 1.3498570066730221e-08, "loss": 1.9105, "step": 52264 }, { "epoch": 9.96, "learning_rate": 1.344137273593899e-08, "loss": 2.2903, "step": 52265 }, { "epoch": 9.96, "learning_rate": 1.3384175405147759e-08, "loss": 1.9904, "step": 52266 }, { "epoch": 9.96, "learning_rate": 1.3326978074356531e-08, "loss": 1.4223, "step": 52267 }, { "epoch": 9.96, "learning_rate": 1.32697807435653e-08, "loss": 1.0428, "step": 52268 }, { "epoch": 9.96, "learning_rate": 1.3212583412774071e-08, "loss": 1.4737, "step": 52269 }, { "epoch": 9.96, "learning_rate": 1.315538608198284e-08, "loss": 1.3775, "step": 52270 }, { "epoch": 9.96, "learning_rate": 1.3098188751191612e-08, "loss": 0.4901, "step": 52271 }, { "epoch": 9.96, "learning_rate": 1.3040991420400381e-08, "loss": 0.5388, "step": 52272 }, { "epoch": 9.96, "learning_rate": 1.298379408960915e-08, "loss": 1.6508, "step": 52273 }, { "epoch": 9.96, "learning_rate": 1.2926596758817923e-08, "loss": 2.846, "step": 52274 }, { "epoch": 9.96, "learning_rate": 1.2869399428026693e-08, "loss": 1.768, "step": 52275 }, { "epoch": 9.96, "learning_rate": 1.2812202097235462e-08, "loss": 2.4289, "step": 52276 }, { "epoch": 9.96, "learning_rate": 1.2755004766444231e-08, "loss": 1.52, "step": 52277 }, { "epoch": 9.96, "learning_rate": 1.2697807435653004e-08, "loss": 1.9225, "step": 52278 }, { "epoch": 9.96, "learning_rate": 1.2640610104861773e-08, "loss": 1.0906, "step": 52279 }, { "epoch": 9.96, "learning_rate": 1.2583412774070543e-08, "loss": 0.8356, "step": 52280 }, { "epoch": 9.96, "learning_rate": 1.2526215443279316e-08, "loss": 0.9242, "step": 52281 }, { "epoch": 9.96, "learning_rate": 1.2469018112488085e-08, "loss": 1.6558, "step": 52282 }, { "epoch": 9.96, "learning_rate": 1.2411820781696854e-08, "loss": 1.9078, "step": 52283 }, { "epoch": 9.96, "learning_rate": 1.2354623450905626e-08, "loss": 3.4115, "step": 52284 }, { "epoch": 9.96, "learning_rate": 1.2297426120114395e-08, "loss": 2.6025, "step": 52285 }, { "epoch": 9.96, "learning_rate": 1.2240228789323165e-08, "loss": 1.0519, "step": 52286 }, { "epoch": 9.96, "learning_rate": 1.2183031458531934e-08, "loss": 1.9465, "step": 52287 }, { "epoch": 9.96, "learning_rate": 1.2125834127740707e-08, "loss": 1.6583, "step": 52288 }, { "epoch": 9.96, "learning_rate": 1.2068636796949476e-08, "loss": 2.5855, "step": 52289 }, { "epoch": 9.96, "learning_rate": 1.2011439466158245e-08, "loss": 1.3995, "step": 52290 }, { "epoch": 9.96, "learning_rate": 1.1954242135367017e-08, "loss": 1.6795, "step": 52291 }, { "epoch": 9.96, "learning_rate": 1.1897044804575788e-08, "loss": 1.7165, "step": 52292 }, { "epoch": 9.96, "learning_rate": 1.1839847473784557e-08, "loss": 1.2547, "step": 52293 }, { "epoch": 9.96, "learning_rate": 1.1782650142993326e-08, "loss": 1.7067, "step": 52294 }, { "epoch": 9.96, "learning_rate": 1.1725452812202098e-08, "loss": 1.1143, "step": 52295 }, { "epoch": 9.96, "learning_rate": 1.1668255481410867e-08, "loss": 1.6214, "step": 52296 }, { "epoch": 9.96, "learning_rate": 1.161105815061964e-08, "loss": 1.1174, "step": 52297 }, { "epoch": 9.96, "learning_rate": 1.1553860819828408e-08, "loss": 1.1343, "step": 52298 }, { "epoch": 9.96, "learning_rate": 1.1496663489037179e-08, "loss": 0.1928, "step": 52299 }, { "epoch": 9.96, "learning_rate": 1.1439466158245948e-08, "loss": 1.7704, "step": 52300 }, { "epoch": 9.96, "learning_rate": 1.1382268827454719e-08, "loss": 1.9707, "step": 52301 }, { "epoch": 9.96, "learning_rate": 1.132507149666349e-08, "loss": 0.872, "step": 52302 }, { "epoch": 9.96, "learning_rate": 1.126787416587226e-08, "loss": 1.4836, "step": 52303 }, { "epoch": 9.96, "learning_rate": 1.121067683508103e-08, "loss": 1.3251, "step": 52304 }, { "epoch": 9.96, "learning_rate": 1.11534795042898e-08, "loss": 1.1926, "step": 52305 }, { "epoch": 9.96, "learning_rate": 1.109628217349857e-08, "loss": 1.7418, "step": 52306 }, { "epoch": 9.96, "learning_rate": 1.1039084842707341e-08, "loss": 1.1711, "step": 52307 }, { "epoch": 9.96, "learning_rate": 1.0981887511916112e-08, "loss": 1.5703, "step": 52308 }, { "epoch": 9.96, "learning_rate": 1.092469018112488e-08, "loss": 2.8251, "step": 52309 }, { "epoch": 9.96, "learning_rate": 1.0867492850333651e-08, "loss": 1.3514, "step": 52310 }, { "epoch": 9.96, "learning_rate": 1.0810295519542422e-08, "loss": 1.3984, "step": 52311 }, { "epoch": 9.96, "learning_rate": 1.0753098188751191e-08, "loss": 1.2658, "step": 52312 }, { "epoch": 9.96, "learning_rate": 1.0695900857959962e-08, "loss": 1.0578, "step": 52313 }, { "epoch": 9.96, "learning_rate": 1.0638703527168732e-08, "loss": 1.1707, "step": 52314 }, { "epoch": 9.96, "learning_rate": 1.0581506196377503e-08, "loss": 1.4119, "step": 52315 }, { "epoch": 9.96, "learning_rate": 1.0524308865586274e-08, "loss": 1.1746, "step": 52316 }, { "epoch": 9.97, "learning_rate": 1.0467111534795043e-08, "loss": 0.5024, "step": 52317 }, { "epoch": 9.97, "learning_rate": 1.0409914204003813e-08, "loss": 1.7594, "step": 52318 }, { "epoch": 9.97, "learning_rate": 1.0352716873212584e-08, "loss": 1.152, "step": 52319 }, { "epoch": 9.97, "learning_rate": 1.0295519542421355e-08, "loss": 1.4829, "step": 52320 }, { "epoch": 9.97, "learning_rate": 1.0238322211630124e-08, "loss": 1.6424, "step": 52321 }, { "epoch": 9.97, "learning_rate": 1.0181124880838894e-08, "loss": 1.367, "step": 52322 }, { "epoch": 9.97, "learning_rate": 1.0123927550047665e-08, "loss": 1.3632, "step": 52323 }, { "epoch": 9.97, "learning_rate": 1.0066730219256436e-08, "loss": 1.3121, "step": 52324 }, { "epoch": 9.97, "learning_rate": 1.0009532888465206e-08, "loss": 1.6734, "step": 52325 }, { "epoch": 9.97, "learning_rate": 9.952335557673975e-09, "loss": 0.3563, "step": 52326 }, { "epoch": 9.97, "learning_rate": 9.895138226882746e-09, "loss": 2.4842, "step": 52327 }, { "epoch": 9.97, "learning_rate": 9.837940896091515e-09, "loss": 1.7576, "step": 52328 }, { "epoch": 9.97, "learning_rate": 9.780743565300286e-09, "loss": 1.871, "step": 52329 }, { "epoch": 9.97, "learning_rate": 9.723546234509056e-09, "loss": 0.8097, "step": 52330 }, { "epoch": 9.97, "learning_rate": 9.666348903717827e-09, "loss": 1.5946, "step": 52331 }, { "epoch": 9.97, "learning_rate": 9.609151572926597e-09, "loss": 2.5681, "step": 52332 }, { "epoch": 9.97, "learning_rate": 9.551954242135366e-09, "loss": 1.5969, "step": 52333 }, { "epoch": 9.97, "learning_rate": 9.494756911344137e-09, "loss": 1.521, "step": 52334 }, { "epoch": 9.97, "learning_rate": 9.437559580552908e-09, "loss": 1.818, "step": 52335 }, { "epoch": 9.97, "learning_rate": 9.380362249761678e-09, "loss": 2.1266, "step": 52336 }, { "epoch": 9.97, "learning_rate": 9.323164918970449e-09, "loss": 0.5349, "step": 52337 }, { "epoch": 9.97, "learning_rate": 9.265967588179218e-09, "loss": 1.7139, "step": 52338 }, { "epoch": 9.97, "learning_rate": 9.208770257387989e-09, "loss": 2.3621, "step": 52339 }, { "epoch": 9.97, "learning_rate": 9.151572926596758e-09, "loss": 1.2058, "step": 52340 }, { "epoch": 9.97, "learning_rate": 9.09437559580553e-09, "loss": 2.7063, "step": 52341 }, { "epoch": 9.97, "learning_rate": 9.0371782650143e-09, "loss": 2.0254, "step": 52342 }, { "epoch": 9.97, "learning_rate": 8.97998093422307e-09, "loss": 1.5672, "step": 52343 }, { "epoch": 9.97, "learning_rate": 8.92278360343184e-09, "loss": 1.1511, "step": 52344 }, { "epoch": 9.97, "learning_rate": 8.86558627264061e-09, "loss": 1.2297, "step": 52345 }, { "epoch": 9.97, "learning_rate": 8.80838894184938e-09, "loss": 3.1755, "step": 52346 }, { "epoch": 9.97, "learning_rate": 8.75119161105815e-09, "loss": 1.3947, "step": 52347 }, { "epoch": 9.97, "learning_rate": 8.693994280266921e-09, "loss": 1.4228, "step": 52348 }, { "epoch": 9.97, "learning_rate": 8.636796949475692e-09, "loss": 1.0107, "step": 52349 }, { "epoch": 9.97, "learning_rate": 8.579599618684461e-09, "loss": 1.5731, "step": 52350 }, { "epoch": 9.97, "learning_rate": 8.522402287893232e-09, "loss": 1.1868, "step": 52351 }, { "epoch": 9.97, "learning_rate": 8.465204957102002e-09, "loss": 1.1971, "step": 52352 }, { "epoch": 9.97, "learning_rate": 8.408007626310773e-09, "loss": 2.2062, "step": 52353 }, { "epoch": 9.97, "learning_rate": 8.350810295519544e-09, "loss": 1.9649, "step": 52354 }, { "epoch": 9.97, "learning_rate": 8.293612964728313e-09, "loss": 0.8507, "step": 52355 }, { "epoch": 9.97, "learning_rate": 8.236415633937083e-09, "loss": 0.5233, "step": 52356 }, { "epoch": 9.97, "learning_rate": 8.179218303145852e-09, "loss": 1.2528, "step": 52357 }, { "epoch": 9.97, "learning_rate": 8.122020972354625e-09, "loss": 1.3919, "step": 52358 }, { "epoch": 9.97, "learning_rate": 8.064823641563394e-09, "loss": 1.823, "step": 52359 }, { "epoch": 9.97, "learning_rate": 8.007626310772164e-09, "loss": 2.3407, "step": 52360 }, { "epoch": 9.97, "learning_rate": 7.950428979980935e-09, "loss": 2.7713, "step": 52361 }, { "epoch": 9.97, "learning_rate": 7.893231649189704e-09, "loss": 1.5465, "step": 52362 }, { "epoch": 9.97, "learning_rate": 7.836034318398475e-09, "loss": 1.9174, "step": 52363 }, { "epoch": 9.97, "learning_rate": 7.778836987607245e-09, "loss": 1.062, "step": 52364 }, { "epoch": 9.97, "learning_rate": 7.721639656816016e-09, "loss": 0.7179, "step": 52365 }, { "epoch": 9.97, "learning_rate": 7.664442326024785e-09, "loss": 1.3856, "step": 52366 }, { "epoch": 9.97, "learning_rate": 7.607244995233556e-09, "loss": 1.1437, "step": 52367 }, { "epoch": 9.97, "learning_rate": 7.550047664442326e-09, "loss": 0.8239, "step": 52368 }, { "epoch": 9.98, "learning_rate": 7.492850333651097e-09, "loss": 0.7251, "step": 52369 }, { "epoch": 9.98, "learning_rate": 7.435653002859867e-09, "loss": 1.9065, "step": 52370 }, { "epoch": 9.98, "learning_rate": 7.3784556720686366e-09, "loss": 2.2963, "step": 52371 }, { "epoch": 9.98, "learning_rate": 7.321258341277407e-09, "loss": 1.2661, "step": 52372 }, { "epoch": 9.98, "learning_rate": 7.264061010486178e-09, "loss": 1.8799, "step": 52373 }, { "epoch": 9.98, "learning_rate": 7.206863679694948e-09, "loss": 2.0337, "step": 52374 }, { "epoch": 9.98, "learning_rate": 7.149666348903718e-09, "loss": 0.8101, "step": 52375 }, { "epoch": 9.98, "learning_rate": 7.092469018112488e-09, "loss": 0.9202, "step": 52376 }, { "epoch": 9.98, "learning_rate": 7.035271687321259e-09, "loss": 1.2864, "step": 52377 }, { "epoch": 9.98, "learning_rate": 6.978074356530028e-09, "loss": 1.8933, "step": 52378 }, { "epoch": 9.98, "learning_rate": 6.920877025738799e-09, "loss": 1.3257, "step": 52379 }, { "epoch": 9.98, "learning_rate": 6.86367969494757e-09, "loss": 2.3132, "step": 52380 }, { "epoch": 9.98, "learning_rate": 6.806482364156339e-09, "loss": 1.3275, "step": 52381 }, { "epoch": 9.98, "learning_rate": 6.7492850333651105e-09, "loss": 2.5942, "step": 52382 }, { "epoch": 9.98, "learning_rate": 6.6920877025738795e-09, "loss": 0.937, "step": 52383 }, { "epoch": 9.98, "learning_rate": 6.63489037178265e-09, "loss": 1.8378, "step": 52384 }, { "epoch": 9.98, "learning_rate": 6.57769304099142e-09, "loss": 1.3803, "step": 52385 }, { "epoch": 9.98, "learning_rate": 6.520495710200191e-09, "loss": 1.5098, "step": 52386 }, { "epoch": 9.98, "learning_rate": 6.463298379408961e-09, "loss": 0.4092, "step": 52387 }, { "epoch": 9.98, "learning_rate": 6.406101048617731e-09, "loss": 1.7601, "step": 52388 }, { "epoch": 9.98, "learning_rate": 6.348903717826502e-09, "loss": 2.6273, "step": 52389 }, { "epoch": 9.98, "learning_rate": 6.291706387035272e-09, "loss": 1.6605, "step": 52390 }, { "epoch": 9.98, "learning_rate": 6.234509056244042e-09, "loss": 1.4939, "step": 52391 }, { "epoch": 9.98, "learning_rate": 6.177311725452813e-09, "loss": 0.956, "step": 52392 }, { "epoch": 9.98, "learning_rate": 6.120114394661583e-09, "loss": 1.6476, "step": 52393 }, { "epoch": 9.98, "learning_rate": 6.062917063870353e-09, "loss": 1.9616, "step": 52394 }, { "epoch": 9.98, "learning_rate": 6.005719733079122e-09, "loss": 3.3432, "step": 52395 }, { "epoch": 9.98, "learning_rate": 5.948522402287894e-09, "loss": 1.7452, "step": 52396 }, { "epoch": 9.98, "learning_rate": 5.891325071496663e-09, "loss": 1.0399, "step": 52397 }, { "epoch": 9.98, "learning_rate": 5.8341277407054335e-09, "loss": 1.6851, "step": 52398 }, { "epoch": 9.98, "learning_rate": 5.776930409914204e-09, "loss": 1.8689, "step": 52399 }, { "epoch": 9.98, "learning_rate": 5.719733079122974e-09, "loss": 1.9304, "step": 52400 }, { "epoch": 9.98, "learning_rate": 5.662535748331745e-09, "loss": 1.6234, "step": 52401 }, { "epoch": 9.98, "learning_rate": 5.605338417540515e-09, "loss": 0.7978, "step": 52402 }, { "epoch": 9.98, "learning_rate": 5.548141086749285e-09, "loss": 1.4228, "step": 52403 }, { "epoch": 9.98, "learning_rate": 5.490943755958056e-09, "loss": 1.7285, "step": 52404 }, { "epoch": 9.98, "learning_rate": 5.433746425166826e-09, "loss": 1.6344, "step": 52405 }, { "epoch": 9.98, "learning_rate": 5.3765490943755955e-09, "loss": 1.4586, "step": 52406 }, { "epoch": 9.98, "learning_rate": 5.319351763584366e-09, "loss": 1.2588, "step": 52407 }, { "epoch": 9.98, "learning_rate": 5.262154432793137e-09, "loss": 1.4164, "step": 52408 }, { "epoch": 9.98, "learning_rate": 5.204957102001907e-09, "loss": 2.6543, "step": 52409 }, { "epoch": 9.98, "learning_rate": 5.147759771210677e-09, "loss": 2.3715, "step": 52410 }, { "epoch": 9.98, "learning_rate": 5.090562440419447e-09, "loss": 0.9664, "step": 52411 }, { "epoch": 9.98, "learning_rate": 5.033365109628218e-09, "loss": 2.1546, "step": 52412 }, { "epoch": 9.98, "learning_rate": 4.976167778836988e-09, "loss": 1.4015, "step": 52413 }, { "epoch": 9.98, "learning_rate": 4.9189704480457574e-09, "loss": 1.5497, "step": 52414 }, { "epoch": 9.98, "learning_rate": 4.861773117254528e-09, "loss": 2.094, "step": 52415 }, { "epoch": 9.98, "learning_rate": 4.804575786463299e-09, "loss": 1.2334, "step": 52416 }, { "epoch": 9.98, "learning_rate": 4.7473784556720686e-09, "loss": 0.6947, "step": 52417 }, { "epoch": 9.98, "learning_rate": 4.690181124880839e-09, "loss": 2.0869, "step": 52418 }, { "epoch": 9.98, "learning_rate": 4.632983794089609e-09, "loss": 2.3911, "step": 52419 }, { "epoch": 9.98, "learning_rate": 4.575786463298379e-09, "loss": 1.7618, "step": 52420 }, { "epoch": 9.98, "learning_rate": 4.51858913250715e-09, "loss": 1.8797, "step": 52421 }, { "epoch": 9.99, "learning_rate": 4.46139180171592e-09, "loss": 2.1944, "step": 52422 }, { "epoch": 9.99, "learning_rate": 4.40419447092469e-09, "loss": 1.5849, "step": 52423 }, { "epoch": 9.99, "learning_rate": 4.346997140133461e-09, "loss": 1.3419, "step": 52424 }, { "epoch": 9.99, "learning_rate": 4.2897998093422305e-09, "loss": 2.724, "step": 52425 }, { "epoch": 9.99, "learning_rate": 4.232602478551001e-09, "loss": 1.4045, "step": 52426 }, { "epoch": 9.99, "learning_rate": 4.175405147759772e-09, "loss": 0.6644, "step": 52427 }, { "epoch": 9.99, "learning_rate": 4.118207816968542e-09, "loss": 1.8746, "step": 52428 }, { "epoch": 9.99, "learning_rate": 4.061010486177312e-09, "loss": 1.1105, "step": 52429 }, { "epoch": 9.99, "learning_rate": 4.003813155386082e-09, "loss": 2.0631, "step": 52430 }, { "epoch": 9.99, "learning_rate": 3.946615824594852e-09, "loss": 1.0428, "step": 52431 }, { "epoch": 9.99, "learning_rate": 3.889418493803623e-09, "loss": 1.3437, "step": 52432 }, { "epoch": 9.99, "learning_rate": 3.8322211630123925e-09, "loss": 0.7418, "step": 52433 }, { "epoch": 9.99, "learning_rate": 3.775023832221163e-09, "loss": 0.6054, "step": 52434 }, { "epoch": 9.99, "learning_rate": 3.7178265014299334e-09, "loss": 0.4506, "step": 52435 }, { "epoch": 9.99, "learning_rate": 3.6606291706387036e-09, "loss": 1.9446, "step": 52436 }, { "epoch": 9.99, "learning_rate": 3.603431839847474e-09, "loss": 0.4051, "step": 52437 }, { "epoch": 9.99, "learning_rate": 3.546234509056244e-09, "loss": 0.376, "step": 52438 }, { "epoch": 9.99, "learning_rate": 3.489037178265014e-09, "loss": 0.8953, "step": 52439 }, { "epoch": 9.99, "learning_rate": 3.431839847473785e-09, "loss": 1.537, "step": 52440 }, { "epoch": 9.99, "learning_rate": 3.3746425166825552e-09, "loss": 1.3591, "step": 52441 }, { "epoch": 9.99, "learning_rate": 3.317445185891325e-09, "loss": 1.0836, "step": 52442 }, { "epoch": 9.99, "learning_rate": 3.2602478551000953e-09, "loss": 2.2872, "step": 52443 }, { "epoch": 9.99, "learning_rate": 3.2030505243088656e-09, "loss": 1.5353, "step": 52444 }, { "epoch": 9.99, "learning_rate": 3.145853193517636e-09, "loss": 0.9428, "step": 52445 }, { "epoch": 9.99, "learning_rate": 3.0886558627264065e-09, "loss": 0.9048, "step": 52446 }, { "epoch": 9.99, "learning_rate": 3.0314585319351767e-09, "loss": 0.718, "step": 52447 }, { "epoch": 9.99, "learning_rate": 2.974261201143947e-09, "loss": 1.4824, "step": 52448 }, { "epoch": 9.99, "learning_rate": 2.9170638703527168e-09, "loss": 2.163, "step": 52449 }, { "epoch": 9.99, "learning_rate": 2.859866539561487e-09, "loss": 1.7132, "step": 52450 }, { "epoch": 9.99, "learning_rate": 2.8026692087702577e-09, "loss": 2.0899, "step": 52451 }, { "epoch": 9.99, "learning_rate": 2.745471877979028e-09, "loss": 1.3651, "step": 52452 }, { "epoch": 9.99, "learning_rate": 2.6882745471877977e-09, "loss": 1.2585, "step": 52453 }, { "epoch": 9.99, "learning_rate": 2.6310772163965684e-09, "loss": 2.8554, "step": 52454 }, { "epoch": 9.99, "learning_rate": 2.5738798856053386e-09, "loss": 1.4049, "step": 52455 }, { "epoch": 9.99, "learning_rate": 2.516682554814109e-09, "loss": 2.1843, "step": 52456 }, { "epoch": 9.99, "learning_rate": 2.4594852240228787e-09, "loss": 1.4084, "step": 52457 }, { "epoch": 9.99, "learning_rate": 2.4022878932316494e-09, "loss": 1.5947, "step": 52458 }, { "epoch": 9.99, "learning_rate": 2.3450905624404196e-09, "loss": 4.2022, "step": 52459 }, { "epoch": 9.99, "learning_rate": 2.2878932316491894e-09, "loss": 2.0246, "step": 52460 }, { "epoch": 9.99, "learning_rate": 2.23069590085796e-09, "loss": 1.0674, "step": 52461 }, { "epoch": 9.99, "learning_rate": 2.1734985700667303e-09, "loss": 1.2815, "step": 52462 }, { "epoch": 9.99, "learning_rate": 2.1163012392755006e-09, "loss": 1.4626, "step": 52463 }, { "epoch": 9.99, "learning_rate": 2.059103908484271e-09, "loss": 0.5963, "step": 52464 }, { "epoch": 9.99, "learning_rate": 2.001906577693041e-09, "loss": 1.7039, "step": 52465 }, { "epoch": 9.99, "learning_rate": 1.9447092469018113e-09, "loss": 0.9534, "step": 52466 }, { "epoch": 9.99, "learning_rate": 1.8875119161105816e-09, "loss": 1.8521, "step": 52467 }, { "epoch": 9.99, "learning_rate": 1.8303145853193518e-09, "loss": 2.7043, "step": 52468 }, { "epoch": 9.99, "learning_rate": 1.773117254528122e-09, "loss": 3.526, "step": 52469 }, { "epoch": 9.99, "learning_rate": 1.7159199237368925e-09, "loss": 0.6938, "step": 52470 }, { "epoch": 9.99, "learning_rate": 1.6587225929456625e-09, "loss": 1.9371, "step": 52471 }, { "epoch": 9.99, "learning_rate": 1.6015252621544328e-09, "loss": 1.0234, "step": 52472 }, { "epoch": 9.99, "learning_rate": 1.5443279313632032e-09, "loss": 1.005, "step": 52473 }, { "epoch": 10.0, "learning_rate": 1.4871306005719735e-09, "loss": 1.3462, "step": 52474 }, { "epoch": 10.0, "learning_rate": 1.4299332697807435e-09, "loss": 2.4987, "step": 52475 }, { "epoch": 10.0, "learning_rate": 1.372735938989514e-09, "loss": 1.1007, "step": 52476 }, { "epoch": 10.0, "learning_rate": 1.3155386081982842e-09, "loss": 2.6454, "step": 52477 }, { "epoch": 10.0, "learning_rate": 1.2583412774070544e-09, "loss": 2.3447, "step": 52478 }, { "epoch": 10.0, "learning_rate": 1.2011439466158247e-09, "loss": 0.5097, "step": 52479 }, { "epoch": 10.0, "learning_rate": 1.1439466158245947e-09, "loss": 1.5485, "step": 52480 }, { "epoch": 10.0, "learning_rate": 1.0867492850333652e-09, "loss": 3.5813, "step": 52481 }, { "epoch": 10.0, "learning_rate": 1.0295519542421354e-09, "loss": 1.3935, "step": 52482 }, { "epoch": 10.0, "learning_rate": 9.723546234509057e-10, "loss": 1.4222, "step": 52483 }, { "epoch": 10.0, "learning_rate": 9.151572926596759e-10, "loss": 1.8258, "step": 52484 }, { "epoch": 10.0, "learning_rate": 8.579599618684462e-10, "loss": 1.7961, "step": 52485 }, { "epoch": 10.0, "learning_rate": 8.007626310772164e-10, "loss": 1.7656, "step": 52486 }, { "epoch": 10.0, "learning_rate": 7.435653002859867e-10, "loss": 1.4396, "step": 52487 }, { "epoch": 10.0, "learning_rate": 6.86367969494757e-10, "loss": 1.7745, "step": 52488 }, { "epoch": 10.0, "learning_rate": 6.291706387035272e-10, "loss": 2.7577, "step": 52489 }, { "epoch": 10.0, "learning_rate": 5.719733079122974e-10, "loss": 0.876, "step": 52490 }, { "epoch": 10.0, "learning_rate": 5.147759771210677e-10, "loss": 1.5727, "step": 52491 }, { "epoch": 10.0, "learning_rate": 4.5757864632983795e-10, "loss": 0.9004, "step": 52492 }, { "epoch": 10.0, "learning_rate": 4.003813155386082e-10, "loss": 0.8644, "step": 52493 }, { "epoch": 10.0, "learning_rate": 3.431839847473785e-10, "loss": 0.901, "step": 52494 }, { "epoch": 10.0, "learning_rate": 2.859866539561487e-10, "loss": 0.5157, "step": 52495 }, { "epoch": 10.0, "learning_rate": 2.2878932316491898e-10, "loss": 0.5836, "step": 52496 }, { "epoch": 10.0, "learning_rate": 1.7159199237368924e-10, "loss": 3.0521, "step": 52497 }, { "epoch": 10.0, "learning_rate": 1.1439466158245949e-10, "loss": 2.7232, "step": 52498 }, { "epoch": 10.0, "learning_rate": 5.7197330791229744e-11, "loss": 1.0875, "step": 52499 }, { "epoch": 10.0, "learning_rate": 0.0, "loss": 2.6584, "step": 52500 }, { "epoch": 10.0, "eval_accuracy": 0.6258036862408916, "eval_loss": 1.4374229907989502, "eval_runtime": 127.4644, "eval_samples_per_second": 18.303, "eval_steps_per_second": 4.582, "step": 52500 } ], "max_steps": 52500, "num_train_epochs": 10, "total_flos": 7.261913476124655e+18, "trial_name": null, "trial_params": null }