{ "best_metric": null, "best_model_checkpoint": null, "epoch": 0.6001101523774555, "eval_steps": 500, "global_step": 8172, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 0.00019997062495410152, "loss": 12.3782, "step": 2 }, { "epoch": 0.0, "learning_rate": 0.000199941249908203, "loss": 16.5504, "step": 4 }, { "epoch": 0.0, "learning_rate": 0.00019991187486230447, "loss": 10.5837, "step": 6 }, { "epoch": 0.0, "learning_rate": 0.00019988249981640598, "loss": 8.2605, "step": 8 }, { "epoch": 0.0, "learning_rate": 0.00019985312477050746, "loss": 7.4006, "step": 10 }, { "epoch": 0.0, "learning_rate": 0.00019982374972460896, "loss": 6.8048, "step": 12 }, { "epoch": 0.0, "learning_rate": 0.00019979437467871044, "loss": 6.472, "step": 14 }, { "epoch": 0.0, "learning_rate": 0.00019976499963281194, "loss": 6.1023, "step": 16 }, { "epoch": 0.0, "learning_rate": 0.00019973562458691342, "loss": 5.8551, "step": 18 }, { "epoch": 0.0, "learning_rate": 0.0001997062495410149, "loss": 5.6395, "step": 20 }, { "epoch": 0.0, "learning_rate": 0.0001996768744951164, "loss": 5.5024, "step": 22 }, { "epoch": 0.0, "learning_rate": 0.0001996474994492179, "loss": 5.5148, "step": 24 }, { "epoch": 0.0, "learning_rate": 0.0001996181244033194, "loss": 5.1403, "step": 26 }, { "epoch": 0.0, "learning_rate": 0.0001995887493574209, "loss": 5.1312, "step": 28 }, { "epoch": 0.0, "learning_rate": 0.00019955937431152237, "loss": 5.1653, "step": 30 }, { "epoch": 0.0, "learning_rate": 0.00019952999926562385, "loss": 4.934, "step": 32 }, { "epoch": 0.0, "learning_rate": 0.00019950062421972536, "loss": 5.1877, "step": 34 }, { "epoch": 0.0, "learning_rate": 0.00019947124917382684, "loss": 5.1135, "step": 36 }, { "epoch": 0.0, "learning_rate": 0.00019944187412792834, "loss": 4.9425, "step": 38 }, { "epoch": 0.0, "learning_rate": 0.00019941249908202985, "loss": 4.7714, "step": 40 }, { "epoch": 0.0, "learning_rate": 0.0001993831240361313, "loss": 4.7456, "step": 42 }, { "epoch": 0.0, "learning_rate": 0.0001993537489902328, "loss": 4.6917, "step": 44 }, { "epoch": 0.0, "learning_rate": 0.0001993243739443343, "loss": 4.6662, "step": 46 }, { "epoch": 0.0, "learning_rate": 0.0001992949988984358, "loss": 4.6133, "step": 48 }, { "epoch": 0.0, "learning_rate": 0.0001992656238525373, "loss": 4.5436, "step": 50 }, { "epoch": 0.0, "learning_rate": 0.00019923624880663877, "loss": 4.4139, "step": 52 }, { "epoch": 0.0, "learning_rate": 0.00019920687376074025, "loss": 4.3963, "step": 54 }, { "epoch": 0.0, "learning_rate": 0.00019917749871484176, "loss": 4.3365, "step": 56 }, { "epoch": 0.0, "learning_rate": 0.00019914812366894323, "loss": 4.2593, "step": 58 }, { "epoch": 0.0, "learning_rate": 0.00019911874862304474, "loss": 4.2262, "step": 60 }, { "epoch": 0.0, "learning_rate": 0.00019908937357714625, "loss": 4.1038, "step": 62 }, { "epoch": 0.0, "learning_rate": 0.0001990599985312477, "loss": 4.1502, "step": 64 }, { "epoch": 0.0, "learning_rate": 0.0001990306234853492, "loss": 4.0474, "step": 66 }, { "epoch": 0.0, "learning_rate": 0.00019900124843945068, "loss": 3.9473, "step": 68 }, { "epoch": 0.01, "learning_rate": 0.00019897187339355219, "loss": 3.8854, "step": 70 }, { "epoch": 0.01, "learning_rate": 0.0001989424983476537, "loss": 3.8419, "step": 72 }, { "epoch": 0.01, "learning_rate": 0.00019891312330175517, "loss": 3.7939, "step": 74 }, { "epoch": 0.01, "learning_rate": 0.00019888374825585665, "loss": 3.9262, "step": 76 }, { "epoch": 0.01, "learning_rate": 0.00019885437320995815, "loss": 3.7726, "step": 78 }, { "epoch": 0.01, "learning_rate": 0.00019882499816405963, "loss": 3.7282, "step": 80 }, { "epoch": 0.01, "learning_rate": 0.00019879562311816114, "loss": 3.7522, "step": 82 }, { "epoch": 0.01, "learning_rate": 0.00019876624807226262, "loss": 3.5387, "step": 84 }, { "epoch": 0.01, "learning_rate": 0.00019873687302636412, "loss": 3.569, "step": 86 }, { "epoch": 0.01, "learning_rate": 0.0001987074979804656, "loss": 3.5395, "step": 88 }, { "epoch": 0.01, "learning_rate": 0.00019867812293456708, "loss": 3.5563, "step": 90 }, { "epoch": 0.01, "learning_rate": 0.00019864874788866858, "loss": 3.4672, "step": 92 }, { "epoch": 0.01, "learning_rate": 0.0001986193728427701, "loss": 3.6556, "step": 94 }, { "epoch": 0.01, "learning_rate": 0.00019858999779687157, "loss": 3.5519, "step": 96 }, { "epoch": 0.01, "learning_rate": 0.00019856062275097307, "loss": 3.4647, "step": 98 }, { "epoch": 0.01, "learning_rate": 0.00019853124770507455, "loss": 3.4406, "step": 100 }, { "epoch": 0.01, "learning_rate": 0.00019850187265917603, "loss": 3.4357, "step": 102 }, { "epoch": 0.01, "learning_rate": 0.00019847249761327754, "loss": 3.2543, "step": 104 }, { "epoch": 0.01, "learning_rate": 0.00019844312256737901, "loss": 3.3049, "step": 106 }, { "epoch": 0.01, "learning_rate": 0.00019841374752148052, "loss": 3.3703, "step": 108 }, { "epoch": 0.01, "learning_rate": 0.000198384372475582, "loss": 3.2017, "step": 110 }, { "epoch": 0.01, "learning_rate": 0.00019835499742968348, "loss": 3.1357, "step": 112 }, { "epoch": 0.01, "learning_rate": 0.00019832562238378498, "loss": 3.1462, "step": 114 }, { "epoch": 0.01, "learning_rate": 0.0001982962473378865, "loss": 3.1122, "step": 116 }, { "epoch": 0.01, "learning_rate": 0.00019826687229198797, "loss": 3.1991, "step": 118 }, { "epoch": 0.01, "learning_rate": 0.00019823749724608947, "loss": 3.1192, "step": 120 }, { "epoch": 0.01, "learning_rate": 0.00019820812220019095, "loss": 3.0166, "step": 122 }, { "epoch": 0.01, "learning_rate": 0.00019817874715429243, "loss": 3.0616, "step": 124 }, { "epoch": 0.01, "learning_rate": 0.00019814937210839393, "loss": 2.9747, "step": 126 }, { "epoch": 0.01, "learning_rate": 0.0001981199970624954, "loss": 2.8586, "step": 128 }, { "epoch": 0.01, "learning_rate": 0.00019809062201659692, "loss": 2.9149, "step": 130 }, { "epoch": 0.01, "learning_rate": 0.00019806124697069842, "loss": 2.984, "step": 132 }, { "epoch": 0.01, "learning_rate": 0.00019803187192479987, "loss": 2.9041, "step": 134 }, { "epoch": 0.01, "learning_rate": 0.00019800249687890138, "loss": 2.8702, "step": 136 }, { "epoch": 0.01, "learning_rate": 0.00019797312183300288, "loss": 2.745, "step": 138 }, { "epoch": 0.01, "learning_rate": 0.00019794374678710436, "loss": 2.7777, "step": 140 }, { "epoch": 0.01, "learning_rate": 0.00019791437174120587, "loss": 2.7682, "step": 142 }, { "epoch": 0.01, "learning_rate": 0.00019788499669530735, "loss": 2.7113, "step": 144 }, { "epoch": 0.01, "learning_rate": 0.00019785562164940882, "loss": 2.7178, "step": 146 }, { "epoch": 0.01, "learning_rate": 0.00019782624660351033, "loss": 2.6576, "step": 148 }, { "epoch": 0.01, "learning_rate": 0.0001977968715576118, "loss": 2.6961, "step": 150 }, { "epoch": 0.01, "learning_rate": 0.00019776749651171331, "loss": 2.6717, "step": 152 }, { "epoch": 0.01, "learning_rate": 0.00019773812146581482, "loss": 2.6944, "step": 154 }, { "epoch": 0.01, "learning_rate": 0.00019770874641991627, "loss": 2.4901, "step": 156 }, { "epoch": 0.01, "learning_rate": 0.00019767937137401778, "loss": 2.7044, "step": 158 }, { "epoch": 0.01, "learning_rate": 0.00019764999632811928, "loss": 2.5532, "step": 160 }, { "epoch": 0.01, "learning_rate": 0.00019762062128222076, "loss": 2.4249, "step": 162 }, { "epoch": 0.01, "learning_rate": 0.00019759124623632227, "loss": 2.5559, "step": 164 }, { "epoch": 0.01, "learning_rate": 0.00019756187119042374, "loss": 2.4405, "step": 166 }, { "epoch": 0.01, "learning_rate": 0.00019753249614452522, "loss": 2.4021, "step": 168 }, { "epoch": 0.01, "learning_rate": 0.00019750312109862673, "loss": 2.338, "step": 170 }, { "epoch": 0.01, "learning_rate": 0.0001974737460527282, "loss": 2.3567, "step": 172 }, { "epoch": 0.01, "learning_rate": 0.0001974443710068297, "loss": 2.2966, "step": 174 }, { "epoch": 0.01, "learning_rate": 0.00019741499596093122, "loss": 2.1678, "step": 176 }, { "epoch": 0.01, "learning_rate": 0.0001973856209150327, "loss": 2.2541, "step": 178 }, { "epoch": 0.01, "learning_rate": 0.00019735624586913417, "loss": 2.123, "step": 180 }, { "epoch": 0.01, "learning_rate": 0.00019732687082323565, "loss": 2.1997, "step": 182 }, { "epoch": 0.01, "learning_rate": 0.00019729749577733716, "loss": 2.1535, "step": 184 }, { "epoch": 0.01, "learning_rate": 0.00019726812073143866, "loss": 2.1728, "step": 186 }, { "epoch": 0.01, "learning_rate": 0.00019723874568554014, "loss": 2.1194, "step": 188 }, { "epoch": 0.01, "learning_rate": 0.00019720937063964165, "loss": 2.0369, "step": 190 }, { "epoch": 0.01, "learning_rate": 0.00019717999559374313, "loss": 2.1457, "step": 192 }, { "epoch": 0.01, "learning_rate": 0.0001971506205478446, "loss": 1.9228, "step": 194 }, { "epoch": 0.01, "learning_rate": 0.0001971212455019461, "loss": 2.0313, "step": 196 }, { "epoch": 0.01, "learning_rate": 0.0001970918704560476, "loss": 1.9003, "step": 198 }, { "epoch": 0.01, "learning_rate": 0.0001970624954101491, "loss": 2.0184, "step": 200 }, { "epoch": 0.01, "learning_rate": 0.0001970331203642506, "loss": 1.9061, "step": 202 }, { "epoch": 0.01, "learning_rate": 0.00019700374531835205, "loss": 1.7931, "step": 204 }, { "epoch": 0.02, "learning_rate": 0.00019697437027245356, "loss": 1.731, "step": 206 }, { "epoch": 0.02, "learning_rate": 0.00019694499522655506, "loss": 1.8405, "step": 208 }, { "epoch": 0.02, "learning_rate": 0.00019691562018065654, "loss": 1.7155, "step": 210 }, { "epoch": 0.02, "learning_rate": 0.00019688624513475804, "loss": 1.6278, "step": 212 }, { "epoch": 0.02, "learning_rate": 0.00019685687008885952, "loss": 1.6125, "step": 214 }, { "epoch": 0.02, "learning_rate": 0.000196827495042961, "loss": 1.5856, "step": 216 }, { "epoch": 0.02, "learning_rate": 0.0001967981199970625, "loss": 1.726, "step": 218 }, { "epoch": 0.02, "learning_rate": 0.00019676874495116399, "loss": 1.5797, "step": 220 }, { "epoch": 0.02, "learning_rate": 0.0001967393699052655, "loss": 1.5004, "step": 222 }, { "epoch": 0.02, "learning_rate": 0.000196709994859367, "loss": 1.4286, "step": 224 }, { "epoch": 0.02, "learning_rate": 0.00019668061981346845, "loss": 1.5036, "step": 226 }, { "epoch": 0.02, "learning_rate": 0.00019665124476756995, "loss": 1.3711, "step": 228 }, { "epoch": 0.02, "learning_rate": 0.00019662186972167146, "loss": 1.3526, "step": 230 }, { "epoch": 0.02, "learning_rate": 0.00019659249467577294, "loss": 1.2906, "step": 232 }, { "epoch": 0.02, "learning_rate": 0.00019656311962987444, "loss": 1.2697, "step": 234 }, { "epoch": 0.02, "learning_rate": 0.00019653374458397592, "loss": 1.1774, "step": 236 }, { "epoch": 0.02, "learning_rate": 0.0001965043695380774, "loss": 1.2662, "step": 238 }, { "epoch": 0.02, "learning_rate": 0.0001964749944921789, "loss": 1.2205, "step": 240 }, { "epoch": 0.02, "learning_rate": 0.00019644561944628038, "loss": 1.0326, "step": 242 }, { "epoch": 0.02, "learning_rate": 0.0001964162444003819, "loss": 1.0019, "step": 244 }, { "epoch": 0.02, "learning_rate": 0.0001963868693544834, "loss": 1.0612, "step": 246 }, { "epoch": 0.02, "learning_rate": 0.00019635749430858487, "loss": 0.9023, "step": 248 }, { "epoch": 0.02, "learning_rate": 0.00019632811926268635, "loss": 0.9539, "step": 250 }, { "epoch": 0.02, "learning_rate": 0.00019629874421678786, "loss": 0.837, "step": 252 }, { "epoch": 0.02, "learning_rate": 0.00019626936917088933, "loss": 0.8719, "step": 254 }, { "epoch": 0.02, "learning_rate": 0.00019623999412499084, "loss": 0.7567, "step": 256 }, { "epoch": 0.02, "learning_rate": 0.00019621061907909232, "loss": 0.7942, "step": 258 }, { "epoch": 0.02, "learning_rate": 0.0001961812440331938, "loss": 0.727, "step": 260 }, { "epoch": 0.02, "learning_rate": 0.0001961518689872953, "loss": 0.6441, "step": 262 }, { "epoch": 0.02, "learning_rate": 0.00019612249394139678, "loss": 0.6103, "step": 264 }, { "epoch": 0.02, "learning_rate": 0.00019609311889549829, "loss": 0.4966, "step": 266 }, { "epoch": 0.02, "learning_rate": 0.0001960637438495998, "loss": 0.5136, "step": 268 }, { "epoch": 0.02, "learning_rate": 0.00019603436880370127, "loss": 0.6294, "step": 270 }, { "epoch": 0.02, "learning_rate": 0.00019600499375780275, "loss": 0.5189, "step": 272 }, { "epoch": 0.02, "learning_rate": 0.00019597561871190425, "loss": 0.4706, "step": 274 }, { "epoch": 0.02, "learning_rate": 0.00019594624366600573, "loss": 0.5763, "step": 276 }, { "epoch": 0.02, "learning_rate": 0.00019591686862010724, "loss": 0.5302, "step": 278 }, { "epoch": 0.02, "learning_rate": 0.00019588749357420872, "loss": 0.44, "step": 280 }, { "epoch": 0.02, "learning_rate": 0.00019585811852831022, "loss": 0.4606, "step": 282 }, { "epoch": 0.02, "learning_rate": 0.0001958287434824117, "loss": 0.4732, "step": 284 }, { "epoch": 0.02, "learning_rate": 0.00019579936843651318, "loss": 0.3398, "step": 286 }, { "epoch": 0.02, "learning_rate": 0.00019576999339061468, "loss": 0.3525, "step": 288 }, { "epoch": 0.02, "learning_rate": 0.0001957406183447162, "loss": 0.3296, "step": 290 }, { "epoch": 0.02, "learning_rate": 0.00019571124329881767, "loss": 0.2613, "step": 292 }, { "epoch": 0.02, "learning_rate": 0.00019568186825291917, "loss": 0.2898, "step": 294 }, { "epoch": 0.02, "learning_rate": 0.00019565249320702062, "loss": 0.2115, "step": 296 }, { "epoch": 0.02, "learning_rate": 0.00019562311816112213, "loss": 0.2587, "step": 298 }, { "epoch": 0.02, "learning_rate": 0.00019559374311522363, "loss": 0.2094, "step": 300 }, { "epoch": 0.02, "learning_rate": 0.0001955643680693251, "loss": 0.1361, "step": 302 }, { "epoch": 0.02, "learning_rate": 0.00019553499302342662, "loss": 0.1825, "step": 304 }, { "epoch": 0.02, "learning_rate": 0.0001955056179775281, "loss": 0.1286, "step": 306 }, { "epoch": 0.02, "learning_rate": 0.00019547624293162958, "loss": 0.146, "step": 308 }, { "epoch": 0.02, "learning_rate": 0.00019544686788573108, "loss": 0.0928, "step": 310 }, { "epoch": 0.02, "learning_rate": 0.00019541749283983256, "loss": 0.1275, "step": 312 }, { "epoch": 0.02, "learning_rate": 0.00019538811779393406, "loss": 0.1373, "step": 314 }, { "epoch": 0.02, "learning_rate": 0.00019535874274803557, "loss": 0.0897, "step": 316 }, { "epoch": 0.02, "learning_rate": 0.00019532936770213702, "loss": 0.0723, "step": 318 }, { "epoch": 0.02, "learning_rate": 0.00019529999265623853, "loss": 0.0532, "step": 320 }, { "epoch": 0.02, "learning_rate": 0.00019527061761034003, "loss": 0.0657, "step": 322 }, { "epoch": 0.02, "learning_rate": 0.0001952412425644415, "loss": 0.0903, "step": 324 }, { "epoch": 0.02, "learning_rate": 0.00019521186751854302, "loss": 0.0816, "step": 326 }, { "epoch": 0.02, "learning_rate": 0.0001951824924726445, "loss": 0.041, "step": 328 }, { "epoch": 0.02, "learning_rate": 0.00019515311742674597, "loss": 0.0431, "step": 330 }, { "epoch": 0.02, "learning_rate": 0.00019512374238084748, "loss": 0.0538, "step": 332 }, { "epoch": 0.02, "learning_rate": 0.00019509436733494896, "loss": 0.0556, "step": 334 }, { "epoch": 0.02, "learning_rate": 0.00019506499228905046, "loss": 0.0544, "step": 336 }, { "epoch": 0.02, "learning_rate": 0.00019503561724315197, "loss": 0.1655, "step": 338 }, { "epoch": 0.02, "learning_rate": 0.00019500624219725345, "loss": 0.1254, "step": 340 }, { "epoch": 0.03, "learning_rate": 0.00019497686715135492, "loss": 0.1729, "step": 342 }, { "epoch": 0.03, "learning_rate": 0.00019494749210545643, "loss": 0.0798, "step": 344 }, { "epoch": 0.03, "learning_rate": 0.0001949181170595579, "loss": 0.0734, "step": 346 }, { "epoch": 0.03, "learning_rate": 0.00019488874201365941, "loss": 0.0726, "step": 348 }, { "epoch": 0.03, "learning_rate": 0.0001948593669677609, "loss": 0.0711, "step": 350 }, { "epoch": 0.03, "learning_rate": 0.0001948299919218624, "loss": 0.0416, "step": 352 }, { "epoch": 0.03, "learning_rate": 0.00019480061687596388, "loss": 0.0528, "step": 354 }, { "epoch": 0.03, "learning_rate": 0.00019477124183006535, "loss": 0.0445, "step": 356 }, { "epoch": 0.03, "learning_rate": 0.00019474186678416686, "loss": 0.0278, "step": 358 }, { "epoch": 0.03, "learning_rate": 0.00019471249173826837, "loss": 0.036, "step": 360 }, { "epoch": 0.03, "learning_rate": 0.00019468311669236984, "loss": 0.0291, "step": 362 }, { "epoch": 0.03, "learning_rate": 0.00019465374164647132, "loss": 0.0322, "step": 364 }, { "epoch": 0.03, "learning_rate": 0.00019462436660057283, "loss": 0.036, "step": 366 }, { "epoch": 0.03, "learning_rate": 0.0001945949915546743, "loss": 0.0357, "step": 368 }, { "epoch": 0.03, "learning_rate": 0.0001945656165087758, "loss": 0.0221, "step": 370 }, { "epoch": 0.03, "learning_rate": 0.0001945362414628773, "loss": 0.021, "step": 372 }, { "epoch": 0.03, "learning_rate": 0.0001945068664169788, "loss": 0.0214, "step": 374 }, { "epoch": 0.03, "learning_rate": 0.00019447749137108027, "loss": 0.0262, "step": 376 }, { "epoch": 0.03, "learning_rate": 0.00019444811632518175, "loss": 0.0179, "step": 378 }, { "epoch": 0.03, "learning_rate": 0.00019441874127928326, "loss": 0.017, "step": 380 }, { "epoch": 0.03, "learning_rate": 0.00019438936623338476, "loss": 0.0137, "step": 382 }, { "epoch": 0.03, "learning_rate": 0.00019435999118748624, "loss": 0.0158, "step": 384 }, { "epoch": 0.03, "learning_rate": 0.00019433061614158775, "loss": 0.02, "step": 386 }, { "epoch": 0.03, "learning_rate": 0.00019430124109568923, "loss": 0.0138, "step": 388 }, { "epoch": 0.03, "learning_rate": 0.0001942718660497907, "loss": 0.0134, "step": 390 }, { "epoch": 0.03, "learning_rate": 0.0001942424910038922, "loss": 0.0145, "step": 392 }, { "epoch": 0.03, "learning_rate": 0.0001942131159579937, "loss": 0.0139, "step": 394 }, { "epoch": 0.03, "learning_rate": 0.0001941837409120952, "loss": 0.012, "step": 396 }, { "epoch": 0.03, "learning_rate": 0.0001941543658661967, "loss": 0.0121, "step": 398 }, { "epoch": 0.03, "learning_rate": 0.00019412499082029815, "loss": 0.0123, "step": 400 }, { "epoch": 0.03, "learning_rate": 0.00019409561577439966, "loss": 0.0131, "step": 402 }, { "epoch": 0.03, "learning_rate": 0.00019406624072850116, "loss": 0.0107, "step": 404 }, { "epoch": 0.03, "learning_rate": 0.00019403686568260264, "loss": 0.0101, "step": 406 }, { "epoch": 0.03, "learning_rate": 0.00019400749063670414, "loss": 0.0095, "step": 408 }, { "epoch": 0.03, "learning_rate": 0.0001939781155908056, "loss": 0.0107, "step": 410 }, { "epoch": 0.03, "learning_rate": 0.0001939487405449071, "loss": 0.009, "step": 412 }, { "epoch": 0.03, "learning_rate": 0.0001939193654990086, "loss": 0.0096, "step": 414 }, { "epoch": 0.03, "learning_rate": 0.00019388999045311008, "loss": 0.0109, "step": 416 }, { "epoch": 0.03, "learning_rate": 0.0001938606154072116, "loss": 0.0102, "step": 418 }, { "epoch": 0.03, "learning_rate": 0.00019383124036131307, "loss": 0.0091, "step": 420 }, { "epoch": 0.03, "learning_rate": 0.00019380186531541455, "loss": 0.0087, "step": 422 }, { "epoch": 0.03, "learning_rate": 0.00019377249026951605, "loss": 0.0093, "step": 424 }, { "epoch": 0.03, "learning_rate": 0.00019374311522361753, "loss": 0.0246, "step": 426 }, { "epoch": 0.03, "learning_rate": 0.00019371374017771904, "loss": 0.0168, "step": 428 }, { "epoch": 0.03, "learning_rate": 0.00019368436513182054, "loss": 0.0216, "step": 430 }, { "epoch": 0.03, "learning_rate": 0.00019365499008592202, "loss": 0.0132, "step": 432 }, { "epoch": 0.03, "learning_rate": 0.0001936256150400235, "loss": 0.0135, "step": 434 }, { "epoch": 0.03, "learning_rate": 0.000193596239994125, "loss": 0.0121, "step": 436 }, { "epoch": 0.03, "learning_rate": 0.00019356686494822648, "loss": 0.0123, "step": 438 }, { "epoch": 0.03, "learning_rate": 0.000193537489902328, "loss": 0.0117, "step": 440 }, { "epoch": 0.03, "learning_rate": 0.00019350811485642947, "loss": 0.0112, "step": 442 }, { "epoch": 0.03, "learning_rate": 0.00019347873981053097, "loss": 0.0093, "step": 444 }, { "epoch": 0.03, "learning_rate": 0.00019344936476463245, "loss": 0.0326, "step": 446 }, { "epoch": 0.03, "learning_rate": 0.00019341998971873393, "loss": 0.0192, "step": 448 }, { "epoch": 0.03, "learning_rate": 0.00019339061467283543, "loss": 0.0183, "step": 450 }, { "epoch": 0.03, "learning_rate": 0.00019336123962693694, "loss": 0.0157, "step": 452 }, { "epoch": 0.03, "learning_rate": 0.00019333186458103842, "loss": 0.0127, "step": 454 }, { "epoch": 0.03, "learning_rate": 0.00019330248953513992, "loss": 0.0128, "step": 456 }, { "epoch": 0.03, "learning_rate": 0.0001932731144892414, "loss": 0.013, "step": 458 }, { "epoch": 0.03, "learning_rate": 0.00019324373944334288, "loss": 0.0112, "step": 460 }, { "epoch": 0.03, "learning_rate": 0.00019321436439744439, "loss": 0.0118, "step": 462 }, { "epoch": 0.03, "learning_rate": 0.00019318498935154586, "loss": 0.0118, "step": 464 }, { "epoch": 0.03, "learning_rate": 0.00019315561430564737, "loss": 0.0117, "step": 466 }, { "epoch": 0.03, "learning_rate": 0.00019312623925974885, "loss": 0.0111, "step": 468 }, { "epoch": 0.03, "learning_rate": 0.00019309686421385033, "loss": 0.0088, "step": 470 }, { "epoch": 0.03, "learning_rate": 0.00019306748916795183, "loss": 0.0332, "step": 472 }, { "epoch": 0.03, "learning_rate": 0.00019303811412205334, "loss": 0.0172, "step": 474 }, { "epoch": 0.03, "learning_rate": 0.00019300873907615482, "loss": 0.0127, "step": 476 }, { "epoch": 0.04, "learning_rate": 0.00019297936403025632, "loss": 0.0156, "step": 478 }, { "epoch": 0.04, "learning_rate": 0.0001929499889843578, "loss": 0.0129, "step": 480 }, { "epoch": 0.04, "learning_rate": 0.00019292061393845928, "loss": 0.0114, "step": 482 }, { "epoch": 0.04, "learning_rate": 0.00019289123889256078, "loss": 0.0124, "step": 484 }, { "epoch": 0.04, "learning_rate": 0.00019286186384666226, "loss": 0.0112, "step": 486 }, { "epoch": 0.04, "learning_rate": 0.00019283248880076377, "loss": 0.01, "step": 488 }, { "epoch": 0.04, "learning_rate": 0.00019280311375486527, "loss": 0.01, "step": 490 }, { "epoch": 0.04, "learning_rate": 0.00019277373870896672, "loss": 0.0086, "step": 492 }, { "epoch": 0.04, "learning_rate": 0.00019274436366306823, "loss": 0.0098, "step": 494 }, { "epoch": 0.04, "learning_rate": 0.00019271498861716973, "loss": 0.0092, "step": 496 }, { "epoch": 0.04, "learning_rate": 0.0001926856135712712, "loss": 0.0085, "step": 498 }, { "epoch": 0.04, "learning_rate": 0.00019265623852537272, "loss": 0.0098, "step": 500 }, { "epoch": 0.04, "learning_rate": 0.0001926268634794742, "loss": 0.0087, "step": 502 }, { "epoch": 0.04, "learning_rate": 0.00019259748843357568, "loss": 0.0088, "step": 504 }, { "epoch": 0.04, "learning_rate": 0.00019256811338767718, "loss": 0.0082, "step": 506 }, { "epoch": 0.04, "learning_rate": 0.00019253873834177866, "loss": 0.0082, "step": 508 }, { "epoch": 0.04, "learning_rate": 0.00019250936329588016, "loss": 0.009, "step": 510 }, { "epoch": 0.04, "learning_rate": 0.00019247998824998167, "loss": 0.0176, "step": 512 }, { "epoch": 0.04, "learning_rate": 0.00019245061320408312, "loss": 0.0084, "step": 514 }, { "epoch": 0.04, "learning_rate": 0.00019242123815818463, "loss": 0.009, "step": 516 }, { "epoch": 0.04, "learning_rate": 0.00019239186311228613, "loss": 0.0076, "step": 518 }, { "epoch": 0.04, "learning_rate": 0.0001923624880663876, "loss": 0.019, "step": 520 }, { "epoch": 0.04, "learning_rate": 0.00019233311302048912, "loss": 0.0087, "step": 522 }, { "epoch": 0.04, "learning_rate": 0.0001923037379745906, "loss": 0.0077, "step": 524 }, { "epoch": 0.04, "learning_rate": 0.00019227436292869207, "loss": 0.0095, "step": 526 }, { "epoch": 0.04, "learning_rate": 0.00019224498788279358, "loss": 0.0073, "step": 528 }, { "epoch": 0.04, "learning_rate": 0.00019221561283689506, "loss": 0.0078, "step": 530 }, { "epoch": 0.04, "learning_rate": 0.00019218623779099656, "loss": 0.0077, "step": 532 }, { "epoch": 0.04, "learning_rate": 0.00019215686274509807, "loss": 0.0077, "step": 534 }, { "epoch": 0.04, "learning_rate": 0.00019212748769919955, "loss": 0.0084, "step": 536 }, { "epoch": 0.04, "learning_rate": 0.00019209811265330102, "loss": 0.0071, "step": 538 }, { "epoch": 0.04, "learning_rate": 0.0001920687376074025, "loss": 0.0073, "step": 540 }, { "epoch": 0.04, "learning_rate": 0.000192039362561504, "loss": 0.0075, "step": 542 }, { "epoch": 0.04, "learning_rate": 0.00019200998751560551, "loss": 0.0076, "step": 544 }, { "epoch": 0.04, "learning_rate": 0.000191980612469707, "loss": 0.0069, "step": 546 }, { "epoch": 0.04, "learning_rate": 0.0001919512374238085, "loss": 0.0067, "step": 548 }, { "epoch": 0.04, "learning_rate": 0.00019192186237790998, "loss": 0.014, "step": 550 }, { "epoch": 0.04, "learning_rate": 0.00019189248733201145, "loss": 0.0074, "step": 552 }, { "epoch": 0.04, "learning_rate": 0.00019186311228611296, "loss": 0.0068, "step": 554 }, { "epoch": 0.04, "learning_rate": 0.00019183373724021444, "loss": 0.0068, "step": 556 }, { "epoch": 0.04, "learning_rate": 0.00019180436219431594, "loss": 0.0068, "step": 558 }, { "epoch": 0.04, "learning_rate": 0.00019177498714841742, "loss": 0.007, "step": 560 }, { "epoch": 0.04, "learning_rate": 0.0001917456121025189, "loss": 0.0071, "step": 562 }, { "epoch": 0.04, "learning_rate": 0.0001917162370566204, "loss": 0.0073, "step": 564 }, { "epoch": 0.04, "learning_rate": 0.0001916868620107219, "loss": 0.0066, "step": 566 }, { "epoch": 0.04, "learning_rate": 0.0001916574869648234, "loss": 0.0067, "step": 568 }, { "epoch": 0.04, "learning_rate": 0.0001916281119189249, "loss": 0.0073, "step": 570 }, { "epoch": 0.04, "learning_rate": 0.00019159873687302637, "loss": 0.0074, "step": 572 }, { "epoch": 0.04, "learning_rate": 0.00019156936182712785, "loss": 0.0064, "step": 574 }, { "epoch": 0.04, "learning_rate": 0.00019153998678122936, "loss": 0.0069, "step": 576 }, { "epoch": 0.04, "learning_rate": 0.00019151061173533084, "loss": 0.0071, "step": 578 }, { "epoch": 0.04, "learning_rate": 0.00019148123668943234, "loss": 0.0071, "step": 580 }, { "epoch": 0.04, "learning_rate": 0.00019145186164353385, "loss": 0.0068, "step": 582 }, { "epoch": 0.04, "learning_rate": 0.0001914224865976353, "loss": 0.0068, "step": 584 }, { "epoch": 0.04, "learning_rate": 0.0001913931115517368, "loss": 0.0092, "step": 586 }, { "epoch": 0.04, "learning_rate": 0.0001913637365058383, "loss": 0.0063, "step": 588 }, { "epoch": 0.04, "learning_rate": 0.0001913343614599398, "loss": 0.0066, "step": 590 }, { "epoch": 0.04, "learning_rate": 0.0001913049864140413, "loss": 0.0069, "step": 592 }, { "epoch": 0.04, "learning_rate": 0.00019127561136814277, "loss": 0.0068, "step": 594 }, { "epoch": 0.04, "learning_rate": 0.00019124623632224425, "loss": 0.0072, "step": 596 }, { "epoch": 0.04, "learning_rate": 0.00019121686127634575, "loss": 0.0063, "step": 598 }, { "epoch": 0.04, "learning_rate": 0.00019118748623044723, "loss": 0.0085, "step": 600 }, { "epoch": 0.04, "learning_rate": 0.00019115811118454874, "loss": 0.0066, "step": 602 }, { "epoch": 0.04, "learning_rate": 0.00019112873613865024, "loss": 0.0067, "step": 604 }, { "epoch": 0.04, "learning_rate": 0.00019109936109275172, "loss": 0.0066, "step": 606 }, { "epoch": 0.04, "learning_rate": 0.0001910699860468532, "loss": 0.0065, "step": 608 }, { "epoch": 0.04, "learning_rate": 0.0001910406110009547, "loss": 0.007, "step": 610 }, { "epoch": 0.04, "learning_rate": 0.00019101123595505618, "loss": 0.0066, "step": 612 }, { "epoch": 0.05, "learning_rate": 0.0001909818609091577, "loss": 0.0066, "step": 614 }, { "epoch": 0.05, "learning_rate": 0.00019095248586325917, "loss": 0.0066, "step": 616 }, { "epoch": 0.05, "learning_rate": 0.00019092311081736065, "loss": 0.007, "step": 618 }, { "epoch": 0.05, "learning_rate": 0.00019089373577146215, "loss": 0.0066, "step": 620 }, { "epoch": 0.05, "learning_rate": 0.00019086436072556363, "loss": 0.0066, "step": 622 }, { "epoch": 0.05, "learning_rate": 0.00019083498567966514, "loss": 0.0066, "step": 624 }, { "epoch": 0.05, "learning_rate": 0.00019080561063376664, "loss": 0.0065, "step": 626 }, { "epoch": 0.05, "learning_rate": 0.00019077623558786812, "loss": 0.0068, "step": 628 }, { "epoch": 0.05, "learning_rate": 0.0001907468605419696, "loss": 0.0065, "step": 630 }, { "epoch": 0.05, "learning_rate": 0.0001907174854960711, "loss": 0.0067, "step": 632 }, { "epoch": 0.05, "learning_rate": 0.00019068811045017258, "loss": 0.0063, "step": 634 }, { "epoch": 0.05, "learning_rate": 0.0001906587354042741, "loss": 0.0087, "step": 636 }, { "epoch": 0.05, "learning_rate": 0.00019062936035837557, "loss": 0.0069, "step": 638 }, { "epoch": 0.05, "learning_rate": 0.00019059998531247707, "loss": 0.0068, "step": 640 }, { "epoch": 0.05, "learning_rate": 0.00019057061026657855, "loss": 0.0067, "step": 642 }, { "epoch": 0.05, "learning_rate": 0.00019054123522068003, "loss": 0.0069, "step": 644 }, { "epoch": 0.05, "learning_rate": 0.00019051186017478153, "loss": 0.0067, "step": 646 }, { "epoch": 0.05, "learning_rate": 0.00019048248512888304, "loss": 0.0067, "step": 648 }, { "epoch": 0.05, "learning_rate": 0.00019045311008298452, "loss": 0.0063, "step": 650 }, { "epoch": 0.05, "learning_rate": 0.00019042373503708602, "loss": 0.0062, "step": 652 }, { "epoch": 0.05, "learning_rate": 0.00019039435999118747, "loss": 0.0065, "step": 654 }, { "epoch": 0.05, "learning_rate": 0.00019036498494528898, "loss": 0.0065, "step": 656 }, { "epoch": 0.05, "learning_rate": 0.00019033560989939049, "loss": 0.0066, "step": 658 }, { "epoch": 0.05, "learning_rate": 0.00019030623485349196, "loss": 0.0062, "step": 660 }, { "epoch": 0.05, "learning_rate": 0.00019027685980759347, "loss": 0.0063, "step": 662 }, { "epoch": 0.05, "learning_rate": 0.00019024748476169495, "loss": 0.0069, "step": 664 }, { "epoch": 0.05, "learning_rate": 0.00019021810971579643, "loss": 0.0064, "step": 666 }, { "epoch": 0.05, "learning_rate": 0.00019018873466989793, "loss": 0.0063, "step": 668 }, { "epoch": 0.05, "learning_rate": 0.0001901593596239994, "loss": 0.0065, "step": 670 }, { "epoch": 0.05, "learning_rate": 0.00019012998457810092, "loss": 0.0068, "step": 672 }, { "epoch": 0.05, "learning_rate": 0.00019010060953220242, "loss": 0.0063, "step": 674 }, { "epoch": 0.05, "learning_rate": 0.00019007123448630387, "loss": 0.0063, "step": 676 }, { "epoch": 0.05, "learning_rate": 0.00019004185944040538, "loss": 0.0064, "step": 678 }, { "epoch": 0.05, "learning_rate": 0.00019001248439450688, "loss": 0.0063, "step": 680 }, { "epoch": 0.05, "learning_rate": 0.00018998310934860836, "loss": 0.0064, "step": 682 }, { "epoch": 0.05, "learning_rate": 0.00018995373430270987, "loss": 0.0061, "step": 684 }, { "epoch": 0.05, "learning_rate": 0.00018992435925681135, "loss": 0.0062, "step": 686 }, { "epoch": 0.05, "learning_rate": 0.00018989498421091282, "loss": 0.0063, "step": 688 }, { "epoch": 0.05, "learning_rate": 0.00018986560916501433, "loss": 0.0062, "step": 690 }, { "epoch": 0.05, "learning_rate": 0.0001898362341191158, "loss": 0.0059, "step": 692 }, { "epoch": 0.05, "learning_rate": 0.0001898068590732173, "loss": 0.0067, "step": 694 }, { "epoch": 0.05, "learning_rate": 0.00018977748402731882, "loss": 0.0066, "step": 696 }, { "epoch": 0.05, "learning_rate": 0.0001897481089814203, "loss": 0.0067, "step": 698 }, { "epoch": 0.05, "learning_rate": 0.00018971873393552177, "loss": 0.0063, "step": 700 }, { "epoch": 0.05, "learning_rate": 0.00018968935888962328, "loss": 0.0065, "step": 702 }, { "epoch": 0.05, "learning_rate": 0.00018965998384372476, "loss": 0.0062, "step": 704 }, { "epoch": 0.05, "learning_rate": 0.00018963060879782626, "loss": 0.0064, "step": 706 }, { "epoch": 0.05, "learning_rate": 0.00018960123375192774, "loss": 0.0059, "step": 708 }, { "epoch": 0.05, "learning_rate": 0.00018957185870602925, "loss": 0.0059, "step": 710 }, { "epoch": 0.05, "learning_rate": 0.00018954248366013073, "loss": 0.0064, "step": 712 }, { "epoch": 0.05, "learning_rate": 0.0001895131086142322, "loss": 0.0063, "step": 714 }, { "epoch": 0.05, "learning_rate": 0.0001894837335683337, "loss": 0.0061, "step": 716 }, { "epoch": 0.05, "learning_rate": 0.00018945435852243522, "loss": 0.0062, "step": 718 }, { "epoch": 0.05, "learning_rate": 0.0001894249834765367, "loss": 0.0063, "step": 720 }, { "epoch": 0.05, "learning_rate": 0.00018939560843063817, "loss": 0.006, "step": 722 }, { "epoch": 0.05, "learning_rate": 0.00018936623338473968, "loss": 0.0068, "step": 724 }, { "epoch": 0.05, "learning_rate": 0.00018933685833884116, "loss": 0.0062, "step": 726 }, { "epoch": 0.05, "learning_rate": 0.00018930748329294266, "loss": 0.0065, "step": 728 }, { "epoch": 0.05, "learning_rate": 0.00018927810824704414, "loss": 0.0069, "step": 730 }, { "epoch": 0.05, "learning_rate": 0.00018924873320114565, "loss": 0.0064, "step": 732 }, { "epoch": 0.05, "learning_rate": 0.00018921935815524712, "loss": 0.0066, "step": 734 }, { "epoch": 0.05, "learning_rate": 0.0001891899831093486, "loss": 0.0062, "step": 736 }, { "epoch": 0.05, "learning_rate": 0.0001891606080634501, "loss": 0.0062, "step": 738 }, { "epoch": 0.05, "learning_rate": 0.0001891312330175516, "loss": 0.0063, "step": 740 }, { "epoch": 0.05, "learning_rate": 0.0001891018579716531, "loss": 0.0065, "step": 742 }, { "epoch": 0.05, "learning_rate": 0.0001890724829257546, "loss": 0.006, "step": 744 }, { "epoch": 0.05, "learning_rate": 0.00018904310787985608, "loss": 0.0063, "step": 746 }, { "epoch": 0.05, "learning_rate": 0.00018901373283395755, "loss": 0.006, "step": 748 }, { "epoch": 0.06, "learning_rate": 0.00018898435778805906, "loss": 0.006, "step": 750 }, { "epoch": 0.06, "learning_rate": 0.00018895498274216054, "loss": 0.0063, "step": 752 }, { "epoch": 0.06, "learning_rate": 0.00018892560769626204, "loss": 0.0059, "step": 754 }, { "epoch": 0.06, "learning_rate": 0.00018889623265036355, "loss": 0.0063, "step": 756 }, { "epoch": 0.06, "learning_rate": 0.000188866857604465, "loss": 0.0063, "step": 758 }, { "epoch": 0.06, "learning_rate": 0.0001888374825585665, "loss": 0.006, "step": 760 }, { "epoch": 0.06, "learning_rate": 0.000188808107512668, "loss": 0.0062, "step": 762 }, { "epoch": 0.06, "learning_rate": 0.0001887787324667695, "loss": 0.0062, "step": 764 }, { "epoch": 0.06, "learning_rate": 0.000188749357420871, "loss": 0.0063, "step": 766 }, { "epoch": 0.06, "learning_rate": 0.00018871998237497245, "loss": 0.0061, "step": 768 }, { "epoch": 0.06, "learning_rate": 0.00018869060732907395, "loss": 0.0058, "step": 770 }, { "epoch": 0.06, "learning_rate": 0.00018866123228317546, "loss": 0.0062, "step": 772 }, { "epoch": 0.06, "learning_rate": 0.00018863185723727694, "loss": 0.0063, "step": 774 }, { "epoch": 0.06, "learning_rate": 0.00018860248219137844, "loss": 0.0062, "step": 776 }, { "epoch": 0.06, "learning_rate": 0.00018857310714547992, "loss": 0.0063, "step": 778 }, { "epoch": 0.06, "learning_rate": 0.0001885437320995814, "loss": 0.0064, "step": 780 }, { "epoch": 0.06, "learning_rate": 0.0001885143570536829, "loss": 0.0063, "step": 782 }, { "epoch": 0.06, "learning_rate": 0.00018848498200778438, "loss": 0.0061, "step": 784 }, { "epoch": 0.06, "learning_rate": 0.0001884556069618859, "loss": 0.0062, "step": 786 }, { "epoch": 0.06, "learning_rate": 0.0001884262319159874, "loss": 0.0063, "step": 788 }, { "epoch": 0.06, "learning_rate": 0.00018839685687008887, "loss": 0.006, "step": 790 }, { "epoch": 0.06, "learning_rate": 0.00018836748182419035, "loss": 0.0066, "step": 792 }, { "epoch": 0.06, "learning_rate": 0.00018833810677829185, "loss": 0.0064, "step": 794 }, { "epoch": 0.06, "learning_rate": 0.00018830873173239333, "loss": 0.0062, "step": 796 }, { "epoch": 0.06, "learning_rate": 0.00018827935668649484, "loss": 0.0065, "step": 798 }, { "epoch": 0.06, "learning_rate": 0.00018824998164059632, "loss": 0.0062, "step": 800 }, { "epoch": 0.06, "learning_rate": 0.00018822060659469782, "loss": 0.0064, "step": 802 }, { "epoch": 0.06, "learning_rate": 0.0001881912315487993, "loss": 0.0063, "step": 804 }, { "epoch": 0.06, "learning_rate": 0.00018816185650290078, "loss": 0.0061, "step": 806 }, { "epoch": 0.06, "learning_rate": 0.00018813248145700228, "loss": 0.0062, "step": 808 }, { "epoch": 0.06, "learning_rate": 0.0001881031064111038, "loss": 0.0063, "step": 810 }, { "epoch": 0.06, "learning_rate": 0.00018807373136520527, "loss": 0.0061, "step": 812 }, { "epoch": 0.06, "learning_rate": 0.00018804435631930675, "loss": 0.0061, "step": 814 }, { "epoch": 0.06, "learning_rate": 0.00018801498127340825, "loss": 0.0063, "step": 816 }, { "epoch": 0.06, "learning_rate": 0.00018798560622750973, "loss": 0.0062, "step": 818 }, { "epoch": 0.06, "learning_rate": 0.00018795623118161124, "loss": 0.0062, "step": 820 }, { "epoch": 0.06, "learning_rate": 0.00018792685613571271, "loss": 0.0063, "step": 822 }, { "epoch": 0.06, "learning_rate": 0.00018789748108981422, "loss": 0.006, "step": 824 }, { "epoch": 0.06, "learning_rate": 0.0001878681060439157, "loss": 0.0064, "step": 826 }, { "epoch": 0.06, "learning_rate": 0.00018783873099801718, "loss": 0.0062, "step": 828 }, { "epoch": 0.06, "learning_rate": 0.00018780935595211868, "loss": 0.0062, "step": 830 }, { "epoch": 0.06, "learning_rate": 0.0001877799809062202, "loss": 0.006, "step": 832 }, { "epoch": 0.06, "learning_rate": 0.00018775060586032167, "loss": 0.0061, "step": 834 }, { "epoch": 0.06, "learning_rate": 0.00018772123081442317, "loss": 0.0057, "step": 836 }, { "epoch": 0.06, "learning_rate": 0.00018769185576852465, "loss": 0.0061, "step": 838 }, { "epoch": 0.06, "learning_rate": 0.00018766248072262613, "loss": 0.006, "step": 840 }, { "epoch": 0.06, "learning_rate": 0.00018763310567672763, "loss": 0.0061, "step": 842 }, { "epoch": 0.06, "learning_rate": 0.0001876037306308291, "loss": 0.0062, "step": 844 }, { "epoch": 0.06, "learning_rate": 0.00018757435558493062, "loss": 0.0059, "step": 846 }, { "epoch": 0.06, "learning_rate": 0.00018754498053903212, "loss": 0.0058, "step": 848 }, { "epoch": 0.06, "learning_rate": 0.00018751560549313357, "loss": 0.0058, "step": 850 }, { "epoch": 0.06, "learning_rate": 0.00018748623044723508, "loss": 0.0059, "step": 852 }, { "epoch": 0.06, "learning_rate": 0.00018745685540133658, "loss": 0.0062, "step": 854 }, { "epoch": 0.06, "learning_rate": 0.00018742748035543806, "loss": 0.0063, "step": 856 }, { "epoch": 0.06, "learning_rate": 0.00018739810530953957, "loss": 0.0059, "step": 858 }, { "epoch": 0.06, "learning_rate": 0.00018736873026364105, "loss": 0.0062, "step": 860 }, { "epoch": 0.06, "learning_rate": 0.00018733935521774253, "loss": 0.0062, "step": 862 }, { "epoch": 0.06, "learning_rate": 0.00018730998017184403, "loss": 0.0067, "step": 864 }, { "epoch": 0.06, "learning_rate": 0.0001872806051259455, "loss": 0.006, "step": 866 }, { "epoch": 0.06, "learning_rate": 0.00018725123008004701, "loss": 0.006, "step": 868 }, { "epoch": 0.06, "learning_rate": 0.00018722185503414852, "loss": 0.0061, "step": 870 }, { "epoch": 0.06, "learning_rate": 0.00018719247998824997, "loss": 0.0057, "step": 872 }, { "epoch": 0.06, "learning_rate": 0.00018716310494235148, "loss": 0.0062, "step": 874 }, { "epoch": 0.06, "learning_rate": 0.00018713372989645298, "loss": 0.0059, "step": 876 }, { "epoch": 0.06, "learning_rate": 0.00018710435485055446, "loss": 0.0061, "step": 878 }, { "epoch": 0.06, "learning_rate": 0.00018707497980465597, "loss": 0.0062, "step": 880 }, { "epoch": 0.06, "learning_rate": 0.00018704560475875744, "loss": 0.0058, "step": 882 }, { "epoch": 0.06, "learning_rate": 0.00018701622971285892, "loss": 0.0062, "step": 884 }, { "epoch": 0.07, "learning_rate": 0.00018698685466696043, "loss": 0.0057, "step": 886 }, { "epoch": 0.07, "learning_rate": 0.0001869574796210619, "loss": 0.0061, "step": 888 }, { "epoch": 0.07, "learning_rate": 0.0001869281045751634, "loss": 0.0061, "step": 890 }, { "epoch": 0.07, "learning_rate": 0.0001868987295292649, "loss": 0.0058, "step": 892 }, { "epoch": 0.07, "learning_rate": 0.0001868693544833664, "loss": 0.0057, "step": 894 }, { "epoch": 0.07, "learning_rate": 0.00018683997943746787, "loss": 0.0062, "step": 896 }, { "epoch": 0.07, "learning_rate": 0.00018681060439156935, "loss": 0.0061, "step": 898 }, { "epoch": 0.07, "learning_rate": 0.00018678122934567086, "loss": 0.0059, "step": 900 }, { "epoch": 0.07, "learning_rate": 0.00018675185429977236, "loss": 0.0059, "step": 902 }, { "epoch": 0.07, "learning_rate": 0.00018672247925387384, "loss": 0.006, "step": 904 }, { "epoch": 0.07, "learning_rate": 0.00018669310420797535, "loss": 0.0056, "step": 906 }, { "epoch": 0.07, "learning_rate": 0.00018666372916207683, "loss": 0.006, "step": 908 }, { "epoch": 0.07, "learning_rate": 0.0001866343541161783, "loss": 0.0059, "step": 910 }, { "epoch": 0.07, "learning_rate": 0.0001866049790702798, "loss": 0.0062, "step": 912 }, { "epoch": 0.07, "learning_rate": 0.0001865756040243813, "loss": 0.0061, "step": 914 }, { "epoch": 0.07, "learning_rate": 0.0001865462289784828, "loss": 0.0062, "step": 916 }, { "epoch": 0.07, "learning_rate": 0.00018651685393258427, "loss": 0.0061, "step": 918 }, { "epoch": 0.07, "learning_rate": 0.00018648747888668575, "loss": 0.0064, "step": 920 }, { "epoch": 0.07, "learning_rate": 0.00018645810384078726, "loss": 0.0057, "step": 922 }, { "epoch": 0.07, "learning_rate": 0.00018642872879488876, "loss": 0.0058, "step": 924 }, { "epoch": 0.07, "learning_rate": 0.00018639935374899024, "loss": 0.006, "step": 926 }, { "epoch": 0.07, "learning_rate": 0.00018636997870309175, "loss": 0.0062, "step": 928 }, { "epoch": 0.07, "learning_rate": 0.00018634060365719322, "loss": 0.0059, "step": 930 }, { "epoch": 0.07, "learning_rate": 0.0001863112286112947, "loss": 0.0061, "step": 932 }, { "epoch": 0.07, "learning_rate": 0.0001862818535653962, "loss": 0.006, "step": 934 }, { "epoch": 0.07, "learning_rate": 0.00018625247851949769, "loss": 0.0059, "step": 936 }, { "epoch": 0.07, "learning_rate": 0.0001862231034735992, "loss": 0.0058, "step": 938 }, { "epoch": 0.07, "learning_rate": 0.0001861937284277007, "loss": 0.0061, "step": 940 }, { "epoch": 0.07, "learning_rate": 0.00018616435338180215, "loss": 0.006, "step": 942 }, { "epoch": 0.07, "learning_rate": 0.00018613497833590365, "loss": 0.006, "step": 944 }, { "epoch": 0.07, "learning_rate": 0.00018610560329000516, "loss": 0.0058, "step": 946 }, { "epoch": 0.07, "learning_rate": 0.00018607622824410664, "loss": 0.0059, "step": 948 }, { "epoch": 0.07, "learning_rate": 0.00018604685319820814, "loss": 0.006, "step": 950 }, { "epoch": 0.07, "learning_rate": 0.00018601747815230962, "loss": 0.0065, "step": 952 }, { "epoch": 0.07, "learning_rate": 0.0001859881031064111, "loss": 0.0057, "step": 954 }, { "epoch": 0.07, "learning_rate": 0.0001859587280605126, "loss": 0.006, "step": 956 }, { "epoch": 0.07, "learning_rate": 0.00018592935301461408, "loss": 0.0064, "step": 958 }, { "epoch": 0.07, "learning_rate": 0.0001858999779687156, "loss": 0.0062, "step": 960 }, { "epoch": 0.07, "learning_rate": 0.0001858706029228171, "loss": 0.0061, "step": 962 }, { "epoch": 0.07, "learning_rate": 0.00018584122787691855, "loss": 0.0062, "step": 964 }, { "epoch": 0.07, "learning_rate": 0.00018581185283102005, "loss": 0.006, "step": 966 }, { "epoch": 0.07, "learning_rate": 0.00018578247778512156, "loss": 0.0059, "step": 968 }, { "epoch": 0.07, "learning_rate": 0.00018575310273922304, "loss": 0.006, "step": 970 }, { "epoch": 0.07, "learning_rate": 0.00018572372769332454, "loss": 0.006, "step": 972 }, { "epoch": 0.07, "learning_rate": 0.00018569435264742602, "loss": 0.0061, "step": 974 }, { "epoch": 0.07, "learning_rate": 0.0001856649776015275, "loss": 0.006, "step": 976 }, { "epoch": 0.07, "learning_rate": 0.000185635602555629, "loss": 0.0058, "step": 978 }, { "epoch": 0.07, "learning_rate": 0.00018560622750973048, "loss": 0.006, "step": 980 }, { "epoch": 0.07, "learning_rate": 0.000185576852463832, "loss": 0.0059, "step": 982 }, { "epoch": 0.07, "learning_rate": 0.0001855474774179335, "loss": 0.0061, "step": 984 }, { "epoch": 0.07, "learning_rate": 0.00018551810237203497, "loss": 0.0062, "step": 986 }, { "epoch": 0.07, "learning_rate": 0.00018548872732613645, "loss": 0.0061, "step": 988 }, { "epoch": 0.07, "learning_rate": 0.00018545935228023795, "loss": 0.0058, "step": 990 }, { "epoch": 0.07, "learning_rate": 0.00018542997723433943, "loss": 0.0057, "step": 992 }, { "epoch": 0.07, "learning_rate": 0.00018540060218844094, "loss": 0.0058, "step": 994 }, { "epoch": 0.07, "learning_rate": 0.00018537122714254242, "loss": 0.0061, "step": 996 }, { "epoch": 0.07, "learning_rate": 0.00018534185209664392, "loss": 0.0059, "step": 998 }, { "epoch": 0.07, "learning_rate": 0.0001853124770507454, "loss": 0.0059, "step": 1000 }, { "epoch": 0.07, "learning_rate": 0.00018528310200484688, "loss": 0.0059, "step": 1002 }, { "epoch": 0.07, "learning_rate": 0.00018525372695894838, "loss": 0.006, "step": 1004 }, { "epoch": 0.07, "learning_rate": 0.0001852243519130499, "loss": 0.0061, "step": 1006 }, { "epoch": 0.07, "learning_rate": 0.00018519497686715137, "loss": 0.0059, "step": 1008 }, { "epoch": 0.07, "learning_rate": 0.00018516560182125287, "loss": 0.0061, "step": 1010 }, { "epoch": 0.07, "learning_rate": 0.00018513622677535432, "loss": 0.0059, "step": 1012 }, { "epoch": 0.07, "learning_rate": 0.00018510685172945583, "loss": 0.006, "step": 1014 }, { "epoch": 0.07, "learning_rate": 0.00018507747668355734, "loss": 0.006, "step": 1016 }, { "epoch": 0.07, "learning_rate": 0.00018504810163765881, "loss": 0.0058, "step": 1018 }, { "epoch": 0.07, "learning_rate": 0.00018501872659176032, "loss": 0.0059, "step": 1020 }, { "epoch": 0.08, "learning_rate": 0.0001849893515458618, "loss": 0.0055, "step": 1022 }, { "epoch": 0.08, "learning_rate": 0.00018495997649996328, "loss": 0.0061, "step": 1024 }, { "epoch": 0.08, "learning_rate": 0.00018493060145406478, "loss": 0.0061, "step": 1026 }, { "epoch": 0.08, "learning_rate": 0.00018490122640816626, "loss": 0.0057, "step": 1028 }, { "epoch": 0.08, "learning_rate": 0.00018487185136226777, "loss": 0.0061, "step": 1030 }, { "epoch": 0.08, "learning_rate": 0.00018484247631636927, "loss": 0.0063, "step": 1032 }, { "epoch": 0.08, "learning_rate": 0.00018481310127047072, "loss": 0.0058, "step": 1034 }, { "epoch": 0.08, "learning_rate": 0.00018478372622457223, "loss": 0.0062, "step": 1036 }, { "epoch": 0.08, "learning_rate": 0.00018475435117867373, "loss": 0.006, "step": 1038 }, { "epoch": 0.08, "learning_rate": 0.0001847249761327752, "loss": 0.0059, "step": 1040 }, { "epoch": 0.08, "learning_rate": 0.00018469560108687672, "loss": 0.0059, "step": 1042 }, { "epoch": 0.08, "learning_rate": 0.0001846662260409782, "loss": 0.0059, "step": 1044 }, { "epoch": 0.08, "learning_rate": 0.00018463685099507967, "loss": 0.0058, "step": 1046 }, { "epoch": 0.08, "learning_rate": 0.00018460747594918118, "loss": 0.0057, "step": 1048 }, { "epoch": 0.08, "learning_rate": 0.00018457810090328266, "loss": 0.006, "step": 1050 }, { "epoch": 0.08, "learning_rate": 0.00018454872585738416, "loss": 0.0062, "step": 1052 }, { "epoch": 0.08, "learning_rate": 0.00018451935081148567, "loss": 0.0061, "step": 1054 }, { "epoch": 0.08, "learning_rate": 0.00018448997576558715, "loss": 0.0058, "step": 1056 }, { "epoch": 0.08, "learning_rate": 0.00018446060071968863, "loss": 0.0058, "step": 1058 }, { "epoch": 0.08, "learning_rate": 0.00018443122567379013, "loss": 0.006, "step": 1060 }, { "epoch": 0.08, "learning_rate": 0.0001844018506278916, "loss": 0.0059, "step": 1062 }, { "epoch": 0.08, "learning_rate": 0.00018437247558199311, "loss": 0.0058, "step": 1064 }, { "epoch": 0.08, "learning_rate": 0.0001843431005360946, "loss": 0.0058, "step": 1066 }, { "epoch": 0.08, "learning_rate": 0.00018431372549019607, "loss": 0.0061, "step": 1068 }, { "epoch": 0.08, "learning_rate": 0.00018428435044429758, "loss": 0.0064, "step": 1070 }, { "epoch": 0.08, "learning_rate": 0.00018425497539839906, "loss": 0.0057, "step": 1072 }, { "epoch": 0.08, "learning_rate": 0.00018422560035250056, "loss": 0.006, "step": 1074 }, { "epoch": 0.08, "learning_rate": 0.00018419622530660207, "loss": 0.0058, "step": 1076 }, { "epoch": 0.08, "learning_rate": 0.00018416685026070354, "loss": 0.006, "step": 1078 }, { "epoch": 0.08, "learning_rate": 0.00018413747521480502, "loss": 0.006, "step": 1080 }, { "epoch": 0.08, "learning_rate": 0.00018410810016890653, "loss": 0.0061, "step": 1082 }, { "epoch": 0.08, "learning_rate": 0.000184078725123008, "loss": 0.0058, "step": 1084 }, { "epoch": 0.08, "learning_rate": 0.0001840493500771095, "loss": 0.0057, "step": 1086 }, { "epoch": 0.08, "learning_rate": 0.000184019975031211, "loss": 0.0059, "step": 1088 }, { "epoch": 0.08, "learning_rate": 0.0001839905999853125, "loss": 0.006, "step": 1090 }, { "epoch": 0.08, "learning_rate": 0.00018396122493941397, "loss": 0.006, "step": 1092 }, { "epoch": 0.08, "learning_rate": 0.00018393184989351545, "loss": 0.006, "step": 1094 }, { "epoch": 0.08, "learning_rate": 0.00018390247484761696, "loss": 0.0061, "step": 1096 }, { "epoch": 0.08, "learning_rate": 0.00018387309980171846, "loss": 0.0063, "step": 1098 }, { "epoch": 0.08, "learning_rate": 0.00018384372475581994, "loss": 0.006, "step": 1100 }, { "epoch": 0.08, "learning_rate": 0.00018381434970992145, "loss": 0.006, "step": 1102 }, { "epoch": 0.08, "learning_rate": 0.00018378497466402293, "loss": 0.0058, "step": 1104 }, { "epoch": 0.08, "learning_rate": 0.0001837555996181244, "loss": 0.0061, "step": 1106 }, { "epoch": 0.08, "learning_rate": 0.0001837262245722259, "loss": 0.0057, "step": 1108 }, { "epoch": 0.08, "learning_rate": 0.0001836968495263274, "loss": 0.0057, "step": 1110 }, { "epoch": 0.08, "learning_rate": 0.0001836674744804289, "loss": 0.006, "step": 1112 }, { "epoch": 0.08, "learning_rate": 0.0001836380994345304, "loss": 0.0059, "step": 1114 }, { "epoch": 0.08, "learning_rate": 0.00018360872438863185, "loss": 0.0059, "step": 1116 }, { "epoch": 0.08, "learning_rate": 0.00018357934934273336, "loss": 0.0059, "step": 1118 }, { "epoch": 0.08, "learning_rate": 0.00018354997429683486, "loss": 0.0057, "step": 1120 }, { "epoch": 0.08, "learning_rate": 0.00018352059925093634, "loss": 0.0059, "step": 1122 }, { "epoch": 0.08, "learning_rate": 0.00018349122420503785, "loss": 0.006, "step": 1124 }, { "epoch": 0.08, "learning_rate": 0.0001834618491591393, "loss": 0.0056, "step": 1126 }, { "epoch": 0.08, "learning_rate": 0.0001834324741132408, "loss": 0.0056, "step": 1128 }, { "epoch": 0.08, "learning_rate": 0.0001834030990673423, "loss": 0.0061, "step": 1130 }, { "epoch": 0.08, "learning_rate": 0.00018337372402144379, "loss": 0.0057, "step": 1132 }, { "epoch": 0.08, "learning_rate": 0.0001833443489755453, "loss": 0.0062, "step": 1134 }, { "epoch": 0.08, "learning_rate": 0.00018331497392964677, "loss": 0.0058, "step": 1136 }, { "epoch": 0.08, "learning_rate": 0.00018328559888374825, "loss": 0.0061, "step": 1138 }, { "epoch": 0.08, "learning_rate": 0.00018325622383784975, "loss": 0.0066, "step": 1140 }, { "epoch": 0.08, "learning_rate": 0.00018322684879195123, "loss": 0.0058, "step": 1142 }, { "epoch": 0.08, "learning_rate": 0.00018319747374605274, "loss": 0.0057, "step": 1144 }, { "epoch": 0.08, "learning_rate": 0.00018316809870015424, "loss": 0.0061, "step": 1146 }, { "epoch": 0.08, "learning_rate": 0.00018313872365425572, "loss": 0.006, "step": 1148 }, { "epoch": 0.08, "learning_rate": 0.0001831093486083572, "loss": 0.0061, "step": 1150 }, { "epoch": 0.08, "learning_rate": 0.0001830799735624587, "loss": 0.006, "step": 1152 }, { "epoch": 0.08, "learning_rate": 0.00018305059851656018, "loss": 0.0061, "step": 1154 }, { "epoch": 0.08, "learning_rate": 0.0001830212234706617, "loss": 0.006, "step": 1156 }, { "epoch": 0.09, "learning_rate": 0.00018299184842476317, "loss": 0.0061, "step": 1158 }, { "epoch": 0.09, "learning_rate": 0.00018296247337886467, "loss": 0.006, "step": 1160 }, { "epoch": 0.09, "learning_rate": 0.00018293309833296615, "loss": 0.0061, "step": 1162 }, { "epoch": 0.09, "learning_rate": 0.00018290372328706763, "loss": 0.006, "step": 1164 }, { "epoch": 0.09, "learning_rate": 0.00018287434824116913, "loss": 0.0063, "step": 1166 }, { "epoch": 0.09, "learning_rate": 0.00018284497319527064, "loss": 0.0057, "step": 1168 }, { "epoch": 0.09, "learning_rate": 0.00018281559814937212, "loss": 0.0059, "step": 1170 }, { "epoch": 0.09, "learning_rate": 0.0001827862231034736, "loss": 0.0056, "step": 1172 }, { "epoch": 0.09, "learning_rate": 0.0001827568480575751, "loss": 0.0059, "step": 1174 }, { "epoch": 0.09, "learning_rate": 0.00018272747301167658, "loss": 0.0062, "step": 1176 }, { "epoch": 0.09, "learning_rate": 0.00018269809796577809, "loss": 0.0064, "step": 1178 }, { "epoch": 0.09, "learning_rate": 0.00018266872291987956, "loss": 0.006, "step": 1180 }, { "epoch": 0.09, "learning_rate": 0.00018263934787398107, "loss": 0.0061, "step": 1182 }, { "epoch": 0.09, "learning_rate": 0.00018260997282808255, "loss": 0.0059, "step": 1184 }, { "epoch": 0.09, "learning_rate": 0.00018258059778218403, "loss": 0.006, "step": 1186 }, { "epoch": 0.09, "learning_rate": 0.00018255122273628553, "loss": 0.006, "step": 1188 }, { "epoch": 0.09, "learning_rate": 0.00018252184769038704, "loss": 0.006, "step": 1190 }, { "epoch": 0.09, "learning_rate": 0.00018249247264448852, "loss": 0.0061, "step": 1192 }, { "epoch": 0.09, "learning_rate": 0.00018246309759859002, "loss": 0.0059, "step": 1194 }, { "epoch": 0.09, "learning_rate": 0.0001824337225526915, "loss": 0.0059, "step": 1196 }, { "epoch": 0.09, "learning_rate": 0.00018240434750679298, "loss": 0.0059, "step": 1198 }, { "epoch": 0.09, "learning_rate": 0.00018237497246089448, "loss": 0.0061, "step": 1200 }, { "epoch": 0.09, "learning_rate": 0.00018234559741499596, "loss": 0.0059, "step": 1202 }, { "epoch": 0.09, "learning_rate": 0.00018231622236909747, "loss": 0.0059, "step": 1204 }, { "epoch": 0.09, "learning_rate": 0.00018228684732319897, "loss": 0.006, "step": 1206 }, { "epoch": 0.09, "learning_rate": 0.00018225747227730042, "loss": 0.0059, "step": 1208 }, { "epoch": 0.09, "learning_rate": 0.00018222809723140193, "loss": 0.0054, "step": 1210 }, { "epoch": 0.09, "learning_rate": 0.00018219872218550344, "loss": 0.0061, "step": 1212 }, { "epoch": 0.09, "learning_rate": 0.00018216934713960491, "loss": 0.006, "step": 1214 }, { "epoch": 0.09, "learning_rate": 0.00018213997209370642, "loss": 0.0061, "step": 1216 }, { "epoch": 0.09, "learning_rate": 0.0001821105970478079, "loss": 0.0062, "step": 1218 }, { "epoch": 0.09, "learning_rate": 0.00018208122200190938, "loss": 0.0059, "step": 1220 }, { "epoch": 0.09, "learning_rate": 0.00018205184695601088, "loss": 0.0059, "step": 1222 }, { "epoch": 0.09, "learning_rate": 0.00018202247191011236, "loss": 0.006, "step": 1224 }, { "epoch": 0.09, "learning_rate": 0.00018199309686421387, "loss": 0.0058, "step": 1226 }, { "epoch": 0.09, "learning_rate": 0.00018196372181831537, "loss": 0.0057, "step": 1228 }, { "epoch": 0.09, "learning_rate": 0.00018193434677241682, "loss": 0.0056, "step": 1230 }, { "epoch": 0.09, "learning_rate": 0.00018190497172651833, "loss": 0.0057, "step": 1232 }, { "epoch": 0.09, "learning_rate": 0.00018187559668061983, "loss": 0.006, "step": 1234 }, { "epoch": 0.09, "learning_rate": 0.0001818462216347213, "loss": 0.0061, "step": 1236 }, { "epoch": 0.09, "learning_rate": 0.00018181684658882282, "loss": 0.0057, "step": 1238 }, { "epoch": 0.09, "learning_rate": 0.0001817874715429243, "loss": 0.0062, "step": 1240 }, { "epoch": 0.09, "learning_rate": 0.00018175809649702577, "loss": 0.0062, "step": 1242 }, { "epoch": 0.09, "learning_rate": 0.00018172872145112728, "loss": 0.0063, "step": 1244 }, { "epoch": 0.09, "learning_rate": 0.00018169934640522876, "loss": 0.0056, "step": 1246 }, { "epoch": 0.09, "learning_rate": 0.00018166997135933026, "loss": 0.006, "step": 1248 }, { "epoch": 0.09, "learning_rate": 0.00018164059631343174, "loss": 0.006, "step": 1250 }, { "epoch": 0.09, "learning_rate": 0.00018161122126753325, "loss": 0.0056, "step": 1252 }, { "epoch": 0.09, "learning_rate": 0.00018158184622163472, "loss": 0.0061, "step": 1254 }, { "epoch": 0.09, "learning_rate": 0.0001815524711757362, "loss": 0.0059, "step": 1256 }, { "epoch": 0.09, "learning_rate": 0.0001815230961298377, "loss": 0.0059, "step": 1258 }, { "epoch": 0.09, "learning_rate": 0.00018149372108393921, "loss": 0.006, "step": 1260 }, { "epoch": 0.09, "learning_rate": 0.0001814643460380407, "loss": 0.0058, "step": 1262 }, { "epoch": 0.09, "learning_rate": 0.0001814349709921422, "loss": 0.006, "step": 1264 }, { "epoch": 0.09, "learning_rate": 0.00018140559594624368, "loss": 0.0057, "step": 1266 }, { "epoch": 0.09, "learning_rate": 0.00018137622090034515, "loss": 0.0058, "step": 1268 }, { "epoch": 0.09, "learning_rate": 0.00018134684585444666, "loss": 0.0057, "step": 1270 }, { "epoch": 0.09, "learning_rate": 0.00018131747080854814, "loss": 0.0059, "step": 1272 }, { "epoch": 0.09, "learning_rate": 0.00018128809576264964, "loss": 0.006, "step": 1274 }, { "epoch": 0.09, "learning_rate": 0.00018125872071675112, "loss": 0.0059, "step": 1276 }, { "epoch": 0.09, "learning_rate": 0.0001812293456708526, "loss": 0.0062, "step": 1278 }, { "epoch": 0.09, "learning_rate": 0.0001811999706249541, "loss": 0.0061, "step": 1280 }, { "epoch": 0.09, "learning_rate": 0.0001811705955790556, "loss": 0.0059, "step": 1282 }, { "epoch": 0.09, "learning_rate": 0.0001811412205331571, "loss": 0.0058, "step": 1284 }, { "epoch": 0.09, "learning_rate": 0.0001811118454872586, "loss": 0.0058, "step": 1286 }, { "epoch": 0.09, "learning_rate": 0.00018108247044136007, "loss": 0.0061, "step": 1288 }, { "epoch": 0.09, "learning_rate": 0.00018105309539546155, "loss": 0.0058, "step": 1290 }, { "epoch": 0.09, "learning_rate": 0.00018102372034956306, "loss": 0.0057, "step": 1292 }, { "epoch": 0.1, "learning_rate": 0.00018099434530366454, "loss": 0.0059, "step": 1294 }, { "epoch": 0.1, "learning_rate": 0.00018096497025776604, "loss": 0.006, "step": 1296 }, { "epoch": 0.1, "learning_rate": 0.00018093559521186755, "loss": 0.0061, "step": 1298 }, { "epoch": 0.1, "learning_rate": 0.000180906220165969, "loss": 0.0061, "step": 1300 }, { "epoch": 0.1, "learning_rate": 0.0001808768451200705, "loss": 0.0059, "step": 1302 }, { "epoch": 0.1, "learning_rate": 0.000180847470074172, "loss": 0.0062, "step": 1304 }, { "epoch": 0.1, "learning_rate": 0.0001808180950282735, "loss": 0.0057, "step": 1306 }, { "epoch": 0.1, "learning_rate": 0.000180788719982375, "loss": 0.0059, "step": 1308 }, { "epoch": 0.1, "learning_rate": 0.00018075934493647647, "loss": 0.0061, "step": 1310 }, { "epoch": 0.1, "learning_rate": 0.00018072996989057795, "loss": 0.0059, "step": 1312 }, { "epoch": 0.1, "learning_rate": 0.00018070059484467946, "loss": 0.0063, "step": 1314 }, { "epoch": 0.1, "learning_rate": 0.00018067121979878093, "loss": 0.0058, "step": 1316 }, { "epoch": 0.1, "learning_rate": 0.00018064184475288244, "loss": 0.0061, "step": 1318 }, { "epoch": 0.1, "learning_rate": 0.00018061246970698394, "loss": 0.0058, "step": 1320 }, { "epoch": 0.1, "learning_rate": 0.0001805830946610854, "loss": 0.0058, "step": 1322 }, { "epoch": 0.1, "learning_rate": 0.0001805537196151869, "loss": 0.0059, "step": 1324 }, { "epoch": 0.1, "learning_rate": 0.0001805243445692884, "loss": 0.0059, "step": 1326 }, { "epoch": 0.1, "learning_rate": 0.00018049496952338989, "loss": 0.0064, "step": 1328 }, { "epoch": 0.1, "learning_rate": 0.0001804655944774914, "loss": 0.0061, "step": 1330 }, { "epoch": 0.1, "learning_rate": 0.00018043621943159287, "loss": 0.006, "step": 1332 }, { "epoch": 0.1, "learning_rate": 0.00018040684438569435, "loss": 0.006, "step": 1334 }, { "epoch": 0.1, "learning_rate": 0.00018037746933979585, "loss": 0.0062, "step": 1336 }, { "epoch": 0.1, "learning_rate": 0.00018034809429389733, "loss": 0.0062, "step": 1338 }, { "epoch": 0.1, "learning_rate": 0.00018031871924799884, "loss": 0.006, "step": 1340 }, { "epoch": 0.1, "learning_rate": 0.00018028934420210034, "loss": 0.0059, "step": 1342 }, { "epoch": 0.1, "learning_rate": 0.00018025996915620182, "loss": 0.0063, "step": 1344 }, { "epoch": 0.1, "learning_rate": 0.0001802305941103033, "loss": 0.0062, "step": 1346 }, { "epoch": 0.1, "learning_rate": 0.0001802012190644048, "loss": 0.0069, "step": 1348 }, { "epoch": 0.1, "learning_rate": 0.00018017184401850628, "loss": 0.0069, "step": 1350 }, { "epoch": 0.1, "learning_rate": 0.0001801424689726078, "loss": 0.0065, "step": 1352 }, { "epoch": 0.1, "learning_rate": 0.00018011309392670927, "loss": 0.0059, "step": 1354 }, { "epoch": 0.1, "learning_rate": 0.00018008371888081077, "loss": 0.0067, "step": 1356 }, { "epoch": 0.1, "learning_rate": 0.00018005434383491225, "loss": 0.0089, "step": 1358 }, { "epoch": 0.1, "learning_rate": 0.00018002496878901373, "loss": 0.0072, "step": 1360 }, { "epoch": 0.1, "learning_rate": 0.00017999559374311523, "loss": 0.0067, "step": 1362 }, { "epoch": 0.1, "learning_rate": 0.0001799662186972167, "loss": 0.0073, "step": 1364 }, { "epoch": 0.1, "learning_rate": 0.00017993684365131822, "loss": 0.0067, "step": 1366 }, { "epoch": 0.1, "learning_rate": 0.00017990746860541972, "loss": 0.0068, "step": 1368 }, { "epoch": 0.1, "learning_rate": 0.00017987809355952118, "loss": 0.0073, "step": 1370 }, { "epoch": 0.1, "learning_rate": 0.00017984871851362268, "loss": 0.0071, "step": 1372 }, { "epoch": 0.1, "learning_rate": 0.00017981934346772419, "loss": 0.0077, "step": 1374 }, { "epoch": 0.1, "learning_rate": 0.00017978996842182566, "loss": 0.0074, "step": 1376 }, { "epoch": 0.1, "learning_rate": 0.00017976059337592717, "loss": 0.0076, "step": 1378 }, { "epoch": 0.1, "learning_rate": 0.00017973121833002865, "loss": 0.0072, "step": 1380 }, { "epoch": 0.1, "learning_rate": 0.00017970184328413013, "loss": 0.0085, "step": 1382 }, { "epoch": 0.1, "learning_rate": 0.00017967246823823163, "loss": 0.0085, "step": 1384 }, { "epoch": 0.1, "learning_rate": 0.0001796430931923331, "loss": 0.0077, "step": 1386 }, { "epoch": 0.1, "learning_rate": 0.00017961371814643462, "loss": 0.0089, "step": 1388 }, { "epoch": 0.1, "learning_rate": 0.00017958434310053612, "loss": 0.0081, "step": 1390 }, { "epoch": 0.1, "learning_rate": 0.00017955496805463757, "loss": 0.0092, "step": 1392 }, { "epoch": 0.1, "learning_rate": 0.00017952559300873908, "loss": 0.0103, "step": 1394 }, { "epoch": 0.1, "learning_rate": 0.00017949621796284058, "loss": 0.0095, "step": 1396 }, { "epoch": 0.1, "learning_rate": 0.00017946684291694206, "loss": 0.009, "step": 1398 }, { "epoch": 0.1, "learning_rate": 0.00017943746787104357, "loss": 0.01, "step": 1400 }, { "epoch": 0.1, "learning_rate": 0.00017940809282514505, "loss": 0.0107, "step": 1402 }, { "epoch": 0.1, "learning_rate": 0.00017937871777924652, "loss": 0.0104, "step": 1404 }, { "epoch": 0.1, "learning_rate": 0.00017934934273334803, "loss": 0.01, "step": 1406 }, { "epoch": 0.1, "learning_rate": 0.0001793199676874495, "loss": 0.0112, "step": 1408 }, { "epoch": 0.1, "learning_rate": 0.000179290592641551, "loss": 0.0114, "step": 1410 }, { "epoch": 0.1, "learning_rate": 0.00017926121759565252, "loss": 0.0105, "step": 1412 }, { "epoch": 0.1, "learning_rate": 0.000179231842549754, "loss": 0.0127, "step": 1414 }, { "epoch": 0.1, "learning_rate": 0.00017920246750385548, "loss": 0.0135, "step": 1416 }, { "epoch": 0.1, "learning_rate": 0.00017917309245795698, "loss": 0.0117, "step": 1418 }, { "epoch": 0.1, "learning_rate": 0.00017914371741205846, "loss": 0.0141, "step": 1420 }, { "epoch": 0.1, "learning_rate": 0.00017911434236615996, "loss": 0.0139, "step": 1422 }, { "epoch": 0.1, "learning_rate": 0.00017908496732026144, "loss": 0.0137, "step": 1424 }, { "epoch": 0.1, "learning_rate": 0.00017905559227436292, "loss": 0.0129, "step": 1426 }, { "epoch": 0.1, "learning_rate": 0.00017902621722846443, "loss": 0.0172, "step": 1428 }, { "epoch": 0.11, "learning_rate": 0.0001789968421825659, "loss": 0.0134, "step": 1430 }, { "epoch": 0.11, "learning_rate": 0.0001789674671366674, "loss": 0.0153, "step": 1432 }, { "epoch": 0.11, "learning_rate": 0.00017893809209076892, "loss": 0.0183, "step": 1434 }, { "epoch": 0.11, "learning_rate": 0.0001789087170448704, "loss": 0.0154, "step": 1436 }, { "epoch": 0.11, "learning_rate": 0.00017887934199897187, "loss": 0.0167, "step": 1438 }, { "epoch": 0.11, "learning_rate": 0.00017884996695307338, "loss": 0.0147, "step": 1440 }, { "epoch": 0.11, "learning_rate": 0.00017882059190717486, "loss": 0.0187, "step": 1442 }, { "epoch": 0.11, "learning_rate": 0.00017879121686127636, "loss": 0.0189, "step": 1444 }, { "epoch": 0.11, "learning_rate": 0.00017876184181537784, "loss": 0.0183, "step": 1446 }, { "epoch": 0.11, "learning_rate": 0.00017873246676947935, "loss": 0.0222, "step": 1448 }, { "epoch": 0.11, "learning_rate": 0.00017870309172358082, "loss": 0.021, "step": 1450 }, { "epoch": 0.11, "learning_rate": 0.0001786737166776823, "loss": 0.024, "step": 1452 }, { "epoch": 0.11, "learning_rate": 0.0001786443416317838, "loss": 0.0202, "step": 1454 }, { "epoch": 0.11, "learning_rate": 0.00017861496658588531, "loss": 0.0236, "step": 1456 }, { "epoch": 0.11, "learning_rate": 0.0001785855915399868, "loss": 0.0242, "step": 1458 }, { "epoch": 0.11, "learning_rate": 0.0001785562164940883, "loss": 0.0221, "step": 1460 }, { "epoch": 0.11, "learning_rate": 0.00017852684144818978, "loss": 0.0259, "step": 1462 }, { "epoch": 0.11, "learning_rate": 0.00017849746640229125, "loss": 0.0206, "step": 1464 }, { "epoch": 0.11, "learning_rate": 0.00017846809135639276, "loss": 0.0232, "step": 1466 }, { "epoch": 0.11, "learning_rate": 0.00017843871631049424, "loss": 0.0246, "step": 1468 }, { "epoch": 0.11, "learning_rate": 0.00017840934126459574, "loss": 0.0288, "step": 1470 }, { "epoch": 0.11, "learning_rate": 0.00017837996621869722, "loss": 0.0326, "step": 1472 }, { "epoch": 0.11, "learning_rate": 0.0001783505911727987, "loss": 0.0315, "step": 1474 }, { "epoch": 0.11, "learning_rate": 0.0001783212161269002, "loss": 0.0401, "step": 1476 }, { "epoch": 0.11, "learning_rate": 0.0001782918410810017, "loss": 0.035, "step": 1478 }, { "epoch": 0.11, "learning_rate": 0.0001782624660351032, "loss": 0.0441, "step": 1480 }, { "epoch": 0.11, "learning_rate": 0.0001782330909892047, "loss": 0.0589, "step": 1482 }, { "epoch": 0.11, "learning_rate": 0.00017820371594330615, "loss": 0.0458, "step": 1484 }, { "epoch": 0.11, "learning_rate": 0.00017817434089740765, "loss": 0.0475, "step": 1486 }, { "epoch": 0.11, "learning_rate": 0.00017814496585150916, "loss": 0.0476, "step": 1488 }, { "epoch": 0.11, "learning_rate": 0.00017811559080561064, "loss": 0.0513, "step": 1490 }, { "epoch": 0.11, "learning_rate": 0.00017808621575971214, "loss": 0.0579, "step": 1492 }, { "epoch": 0.11, "learning_rate": 0.00017805684071381362, "loss": 0.0625, "step": 1494 }, { "epoch": 0.11, "learning_rate": 0.0001780274656679151, "loss": 0.0541, "step": 1496 }, { "epoch": 0.11, "learning_rate": 0.0001779980906220166, "loss": 0.0528, "step": 1498 }, { "epoch": 0.11, "learning_rate": 0.00017796871557611808, "loss": 0.0666, "step": 1500 }, { "epoch": 0.11, "learning_rate": 0.0001779393405302196, "loss": 0.077, "step": 1502 }, { "epoch": 0.11, "learning_rate": 0.0001779099654843211, "loss": 0.0724, "step": 1504 }, { "epoch": 0.11, "learning_rate": 0.00017788059043842257, "loss": 0.0704, "step": 1506 }, { "epoch": 0.11, "learning_rate": 0.00017785121539252405, "loss": 0.1201, "step": 1508 }, { "epoch": 0.11, "learning_rate": 0.00017782184034662556, "loss": 0.1369, "step": 1510 }, { "epoch": 0.11, "learning_rate": 0.00017779246530072703, "loss": 0.1149, "step": 1512 }, { "epoch": 0.11, "learning_rate": 0.00017776309025482854, "loss": 0.1338, "step": 1514 }, { "epoch": 0.11, "learning_rate": 0.00017773371520893002, "loss": 0.1381, "step": 1516 }, { "epoch": 0.11, "learning_rate": 0.00017770434016303152, "loss": 0.2207, "step": 1518 }, { "epoch": 0.11, "learning_rate": 0.000177674965117133, "loss": 0.1764, "step": 1520 }, { "epoch": 0.11, "learning_rate": 0.00017764559007123448, "loss": 0.177, "step": 1522 }, { "epoch": 0.11, "learning_rate": 0.00017761621502533599, "loss": 0.197, "step": 1524 }, { "epoch": 0.11, "learning_rate": 0.0001775868399794375, "loss": 0.1657, "step": 1526 }, { "epoch": 0.11, "learning_rate": 0.00017755746493353897, "loss": 0.1491, "step": 1528 }, { "epoch": 0.11, "learning_rate": 0.00017752808988764045, "loss": 0.1558, "step": 1530 }, { "epoch": 0.11, "learning_rate": 0.00017749871484174195, "loss": 0.1573, "step": 1532 }, { "epoch": 0.11, "learning_rate": 0.00017746933979584343, "loss": 0.1186, "step": 1534 }, { "epoch": 0.11, "learning_rate": 0.00017743996474994494, "loss": 0.1098, "step": 1536 }, { "epoch": 0.11, "learning_rate": 0.00017741058970404641, "loss": 0.0978, "step": 1538 }, { "epoch": 0.11, "learning_rate": 0.00017738121465814792, "loss": 0.0999, "step": 1540 }, { "epoch": 0.11, "learning_rate": 0.0001773518396122494, "loss": 0.0842, "step": 1542 }, { "epoch": 0.11, "learning_rate": 0.00017732246456635088, "loss": 0.1075, "step": 1544 }, { "epoch": 0.11, "learning_rate": 0.00017729308952045238, "loss": 0.0783, "step": 1546 }, { "epoch": 0.11, "learning_rate": 0.0001772637144745539, "loss": 0.1054, "step": 1548 }, { "epoch": 0.11, "learning_rate": 0.00017723433942865537, "loss": 0.0907, "step": 1550 }, { "epoch": 0.11, "learning_rate": 0.00017720496438275687, "loss": 0.0687, "step": 1552 }, { "epoch": 0.11, "learning_rate": 0.00017717558933685835, "loss": 0.0662, "step": 1554 }, { "epoch": 0.11, "learning_rate": 0.00017714621429095983, "loss": 0.0506, "step": 1556 }, { "epoch": 0.11, "learning_rate": 0.00017711683924506133, "loss": 0.0536, "step": 1558 }, { "epoch": 0.11, "learning_rate": 0.0001770874641991628, "loss": 0.0577, "step": 1560 }, { "epoch": 0.11, "learning_rate": 0.00017705808915326432, "loss": 0.0482, "step": 1562 }, { "epoch": 0.11, "learning_rate": 0.00017702871410736582, "loss": 0.0499, "step": 1564 }, { "epoch": 0.11, "learning_rate": 0.00017699933906146727, "loss": 0.0516, "step": 1566 }, { "epoch": 0.12, "learning_rate": 0.00017696996401556878, "loss": 0.0539, "step": 1568 }, { "epoch": 0.12, "learning_rate": 0.00017694058896967029, "loss": 0.0437, "step": 1570 }, { "epoch": 0.12, "learning_rate": 0.00017691121392377176, "loss": 0.0389, "step": 1572 }, { "epoch": 0.12, "learning_rate": 0.00017688183887787327, "loss": 0.0415, "step": 1574 }, { "epoch": 0.12, "learning_rate": 0.00017685246383197475, "loss": 0.0462, "step": 1576 }, { "epoch": 0.12, "learning_rate": 0.00017682308878607623, "loss": 0.0354, "step": 1578 }, { "epoch": 0.12, "learning_rate": 0.00017679371374017773, "loss": 0.0361, "step": 1580 }, { "epoch": 0.12, "learning_rate": 0.0001767643386942792, "loss": 0.0217, "step": 1582 }, { "epoch": 0.12, "learning_rate": 0.00017673496364838072, "loss": 0.021, "step": 1584 }, { "epoch": 0.12, "learning_rate": 0.00017670558860248222, "loss": 0.027, "step": 1586 }, { "epoch": 0.12, "learning_rate": 0.00017667621355658367, "loss": 0.0277, "step": 1588 }, { "epoch": 0.12, "learning_rate": 0.00017664683851068518, "loss": 0.0213, "step": 1590 }, { "epoch": 0.12, "learning_rate": 0.00017661746346478668, "loss": 0.0237, "step": 1592 }, { "epoch": 0.12, "learning_rate": 0.00017658808841888816, "loss": 0.0216, "step": 1594 }, { "epoch": 0.12, "learning_rate": 0.00017655871337298967, "loss": 0.0131, "step": 1596 }, { "epoch": 0.12, "learning_rate": 0.00017652933832709115, "loss": 0.0187, "step": 1598 }, { "epoch": 0.12, "learning_rate": 0.00017649996328119262, "loss": 0.0152, "step": 1600 }, { "epoch": 0.12, "learning_rate": 0.00017647058823529413, "loss": 0.015, "step": 1602 }, { "epoch": 0.12, "learning_rate": 0.0001764412131893956, "loss": 0.0123, "step": 1604 }, { "epoch": 0.12, "learning_rate": 0.0001764118381434971, "loss": 0.0157, "step": 1606 }, { "epoch": 0.12, "learning_rate": 0.0001763824630975986, "loss": 0.0129, "step": 1608 }, { "epoch": 0.12, "learning_rate": 0.0001763530880517001, "loss": 0.0134, "step": 1610 }, { "epoch": 0.12, "learning_rate": 0.00017632371300580158, "loss": 0.0118, "step": 1612 }, { "epoch": 0.12, "learning_rate": 0.00017629433795990305, "loss": 0.0161, "step": 1614 }, { "epoch": 0.12, "learning_rate": 0.00017626496291400456, "loss": 0.011, "step": 1616 }, { "epoch": 0.12, "learning_rate": 0.00017623558786810606, "loss": 0.0123, "step": 1618 }, { "epoch": 0.12, "learning_rate": 0.00017620621282220754, "loss": 0.012, "step": 1620 }, { "epoch": 0.12, "learning_rate": 0.00017617683777630905, "loss": 0.0115, "step": 1622 }, { "epoch": 0.12, "learning_rate": 0.00017614746273041053, "loss": 0.0114, "step": 1624 }, { "epoch": 0.12, "learning_rate": 0.000176118087684512, "loss": 0.0086, "step": 1626 }, { "epoch": 0.12, "learning_rate": 0.0001760887126386135, "loss": 0.0095, "step": 1628 }, { "epoch": 0.12, "learning_rate": 0.000176059337592715, "loss": 0.008, "step": 1630 }, { "epoch": 0.12, "learning_rate": 0.0001760299625468165, "loss": 0.0102, "step": 1632 }, { "epoch": 0.12, "learning_rate": 0.00017600058750091797, "loss": 0.0094, "step": 1634 }, { "epoch": 0.12, "learning_rate": 0.00017597121245501945, "loss": 0.0078, "step": 1636 }, { "epoch": 0.12, "learning_rate": 0.00017594183740912096, "loss": 0.009, "step": 1638 }, { "epoch": 0.12, "learning_rate": 0.00017591246236322246, "loss": 0.0095, "step": 1640 }, { "epoch": 0.12, "learning_rate": 0.00017588308731732394, "loss": 0.0075, "step": 1642 }, { "epoch": 0.12, "learning_rate": 0.00017585371227142545, "loss": 0.0096, "step": 1644 }, { "epoch": 0.12, "learning_rate": 0.00017582433722552692, "loss": 0.009, "step": 1646 }, { "epoch": 0.12, "learning_rate": 0.0001757949621796284, "loss": 0.0068, "step": 1648 }, { "epoch": 0.12, "learning_rate": 0.0001757655871337299, "loss": 0.007, "step": 1650 }, { "epoch": 0.12, "learning_rate": 0.0001757362120878314, "loss": 0.0073, "step": 1652 }, { "epoch": 0.12, "learning_rate": 0.0001757068370419329, "loss": 0.0078, "step": 1654 }, { "epoch": 0.12, "learning_rate": 0.0001756774619960344, "loss": 0.0079, "step": 1656 }, { "epoch": 0.12, "learning_rate": 0.00017564808695013585, "loss": 0.0067, "step": 1658 }, { "epoch": 0.12, "learning_rate": 0.00017561871190423735, "loss": 0.0063, "step": 1660 }, { "epoch": 0.12, "learning_rate": 0.00017558933685833886, "loss": 0.0071, "step": 1662 }, { "epoch": 0.12, "learning_rate": 0.00017555996181244034, "loss": 0.0071, "step": 1664 }, { "epoch": 0.12, "learning_rate": 0.00017553058676654184, "loss": 0.0063, "step": 1666 }, { "epoch": 0.12, "learning_rate": 0.00017550121172064332, "loss": 0.0067, "step": 1668 }, { "epoch": 0.12, "learning_rate": 0.0001754718366747448, "loss": 0.0072, "step": 1670 }, { "epoch": 0.12, "learning_rate": 0.0001754424616288463, "loss": 0.007, "step": 1672 }, { "epoch": 0.12, "learning_rate": 0.00017541308658294778, "loss": 0.0064, "step": 1674 }, { "epoch": 0.12, "learning_rate": 0.0001753837115370493, "loss": 0.0063, "step": 1676 }, { "epoch": 0.12, "learning_rate": 0.0001753543364911508, "loss": 0.0063, "step": 1678 }, { "epoch": 0.12, "learning_rate": 0.00017532496144525225, "loss": 0.0068, "step": 1680 }, { "epoch": 0.12, "learning_rate": 0.00017529558639935375, "loss": 0.0063, "step": 1682 }, { "epoch": 0.12, "learning_rate": 0.00017526621135345526, "loss": 0.0064, "step": 1684 }, { "epoch": 0.12, "learning_rate": 0.00017523683630755674, "loss": 0.006, "step": 1686 }, { "epoch": 0.12, "learning_rate": 0.00017520746126165824, "loss": 0.006, "step": 1688 }, { "epoch": 0.12, "learning_rate": 0.00017517808621575972, "loss": 0.0061, "step": 1690 }, { "epoch": 0.12, "learning_rate": 0.0001751487111698612, "loss": 0.0069, "step": 1692 }, { "epoch": 0.12, "learning_rate": 0.0001751193361239627, "loss": 0.0063, "step": 1694 }, { "epoch": 0.12, "learning_rate": 0.00017508996107806418, "loss": 0.0064, "step": 1696 }, { "epoch": 0.12, "learning_rate": 0.0001750605860321657, "loss": 0.0063, "step": 1698 }, { "epoch": 0.12, "learning_rate": 0.0001750312109862672, "loss": 0.0061, "step": 1700 }, { "epoch": 0.12, "learning_rate": 0.00017500183594036867, "loss": 0.0057, "step": 1702 }, { "epoch": 0.13, "learning_rate": 0.00017497246089447015, "loss": 0.0061, "step": 1704 }, { "epoch": 0.13, "learning_rate": 0.00017494308584857165, "loss": 0.0065, "step": 1706 }, { "epoch": 0.13, "learning_rate": 0.00017491371080267313, "loss": 0.0062, "step": 1708 }, { "epoch": 0.13, "learning_rate": 0.00017488433575677464, "loss": 0.0064, "step": 1710 }, { "epoch": 0.13, "learning_rate": 0.00017485496071087612, "loss": 0.0062, "step": 1712 }, { "epoch": 0.13, "learning_rate": 0.00017482558566497762, "loss": 0.0062, "step": 1714 }, { "epoch": 0.13, "learning_rate": 0.0001747962106190791, "loss": 0.0056, "step": 1716 }, { "epoch": 0.13, "learning_rate": 0.00017476683557318058, "loss": 0.0062, "step": 1718 }, { "epoch": 0.13, "learning_rate": 0.00017473746052728208, "loss": 0.0062, "step": 1720 }, { "epoch": 0.13, "learning_rate": 0.00017470808548138356, "loss": 0.006, "step": 1722 }, { "epoch": 0.13, "learning_rate": 0.00017467871043548507, "loss": 0.0057, "step": 1724 }, { "epoch": 0.13, "learning_rate": 0.00017464933538958655, "loss": 0.0062, "step": 1726 }, { "epoch": 0.13, "learning_rate": 0.00017461996034368803, "loss": 0.0059, "step": 1728 }, { "epoch": 0.13, "learning_rate": 0.00017459058529778953, "loss": 0.0059, "step": 1730 }, { "epoch": 0.13, "learning_rate": 0.00017456121025189104, "loss": 0.0061, "step": 1732 }, { "epoch": 0.13, "learning_rate": 0.00017453183520599251, "loss": 0.0062, "step": 1734 }, { "epoch": 0.13, "learning_rate": 0.00017450246016009402, "loss": 0.0062, "step": 1736 }, { "epoch": 0.13, "learning_rate": 0.0001744730851141955, "loss": 0.0058, "step": 1738 }, { "epoch": 0.13, "learning_rate": 0.00017444371006829698, "loss": 0.006, "step": 1740 }, { "epoch": 0.13, "learning_rate": 0.00017441433502239848, "loss": 0.006, "step": 1742 }, { "epoch": 0.13, "learning_rate": 0.00017438495997649996, "loss": 0.0061, "step": 1744 }, { "epoch": 0.13, "learning_rate": 0.00017435558493060147, "loss": 0.0059, "step": 1746 }, { "epoch": 0.13, "learning_rate": 0.00017432620988470297, "loss": 0.0057, "step": 1748 }, { "epoch": 0.13, "learning_rate": 0.00017429683483880442, "loss": 0.0062, "step": 1750 }, { "epoch": 0.13, "learning_rate": 0.00017426745979290593, "loss": 0.0059, "step": 1752 }, { "epoch": 0.13, "learning_rate": 0.00017423808474700743, "loss": 0.006, "step": 1754 }, { "epoch": 0.13, "learning_rate": 0.0001742087097011089, "loss": 0.006, "step": 1756 }, { "epoch": 0.13, "learning_rate": 0.00017417933465521042, "loss": 0.006, "step": 1758 }, { "epoch": 0.13, "learning_rate": 0.0001741499596093119, "loss": 0.006, "step": 1760 }, { "epoch": 0.13, "learning_rate": 0.00017412058456341337, "loss": 0.0059, "step": 1762 }, { "epoch": 0.13, "learning_rate": 0.00017409120951751488, "loss": 0.006, "step": 1764 }, { "epoch": 0.13, "learning_rate": 0.00017406183447161636, "loss": 0.0059, "step": 1766 }, { "epoch": 0.13, "learning_rate": 0.00017403245942571786, "loss": 0.0061, "step": 1768 }, { "epoch": 0.13, "learning_rate": 0.00017400308437981937, "loss": 0.0058, "step": 1770 }, { "epoch": 0.13, "learning_rate": 0.00017397370933392085, "loss": 0.0058, "step": 1772 }, { "epoch": 0.13, "learning_rate": 0.00017394433428802233, "loss": 0.0059, "step": 1774 }, { "epoch": 0.13, "learning_rate": 0.00017391495924212383, "loss": 0.0059, "step": 1776 }, { "epoch": 0.13, "learning_rate": 0.0001738855841962253, "loss": 0.0057, "step": 1778 }, { "epoch": 0.13, "learning_rate": 0.00017385620915032682, "loss": 0.0062, "step": 1780 }, { "epoch": 0.13, "learning_rate": 0.0001738268341044283, "loss": 0.0058, "step": 1782 }, { "epoch": 0.13, "learning_rate": 0.00017379745905852977, "loss": 0.0059, "step": 1784 }, { "epoch": 0.13, "learning_rate": 0.00017376808401263128, "loss": 0.006, "step": 1786 }, { "epoch": 0.13, "learning_rate": 0.00017373870896673276, "loss": 0.0058, "step": 1788 }, { "epoch": 0.13, "learning_rate": 0.00017370933392083426, "loss": 0.0057, "step": 1790 }, { "epoch": 0.13, "learning_rate": 0.00017367995887493577, "loss": 0.0061, "step": 1792 }, { "epoch": 0.13, "learning_rate": 0.00017365058382903725, "loss": 0.0059, "step": 1794 }, { "epoch": 0.13, "learning_rate": 0.00017362120878313872, "loss": 0.0061, "step": 1796 }, { "epoch": 0.13, "learning_rate": 0.00017359183373724023, "loss": 0.006, "step": 1798 }, { "epoch": 0.13, "learning_rate": 0.0001735624586913417, "loss": 0.0058, "step": 1800 }, { "epoch": 0.13, "learning_rate": 0.0001735330836454432, "loss": 0.0057, "step": 1802 }, { "epoch": 0.13, "learning_rate": 0.0001735037085995447, "loss": 0.0058, "step": 1804 }, { "epoch": 0.13, "learning_rate": 0.0001734743335536462, "loss": 0.0056, "step": 1806 }, { "epoch": 0.13, "learning_rate": 0.00017344495850774768, "loss": 0.0061, "step": 1808 }, { "epoch": 0.13, "learning_rate": 0.00017341558346184915, "loss": 0.0059, "step": 1810 }, { "epoch": 0.13, "learning_rate": 0.00017338620841595066, "loss": 0.0059, "step": 1812 }, { "epoch": 0.13, "learning_rate": 0.00017335683337005216, "loss": 0.0058, "step": 1814 }, { "epoch": 0.13, "learning_rate": 0.00017332745832415364, "loss": 0.0061, "step": 1816 }, { "epoch": 0.13, "learning_rate": 0.00017329808327825515, "loss": 0.0056, "step": 1818 }, { "epoch": 0.13, "learning_rate": 0.00017326870823235663, "loss": 0.0057, "step": 1820 }, { "epoch": 0.13, "learning_rate": 0.0001732393331864581, "loss": 0.0059, "step": 1822 }, { "epoch": 0.13, "learning_rate": 0.0001732099581405596, "loss": 0.0058, "step": 1824 }, { "epoch": 0.13, "learning_rate": 0.0001731805830946611, "loss": 0.006, "step": 1826 }, { "epoch": 0.13, "learning_rate": 0.0001731512080487626, "loss": 0.0059, "step": 1828 }, { "epoch": 0.13, "learning_rate": 0.00017312183300286407, "loss": 0.006, "step": 1830 }, { "epoch": 0.13, "learning_rate": 0.00017309245795696555, "loss": 0.0059, "step": 1832 }, { "epoch": 0.13, "learning_rate": 0.00017306308291106706, "loss": 0.0058, "step": 1834 }, { "epoch": 0.13, "learning_rate": 0.00017303370786516853, "loss": 0.0059, "step": 1836 }, { "epoch": 0.13, "learning_rate": 0.00017300433281927004, "loss": 0.0057, "step": 1838 }, { "epoch": 0.14, "learning_rate": 0.00017297495777337155, "loss": 0.0058, "step": 1840 }, { "epoch": 0.14, "learning_rate": 0.000172945582727473, "loss": 0.0058, "step": 1842 }, { "epoch": 0.14, "learning_rate": 0.0001729162076815745, "loss": 0.0058, "step": 1844 }, { "epoch": 0.14, "learning_rate": 0.000172886832635676, "loss": 0.0055, "step": 1846 }, { "epoch": 0.14, "learning_rate": 0.00017285745758977749, "loss": 0.0055, "step": 1848 }, { "epoch": 0.14, "learning_rate": 0.000172828082543879, "loss": 0.0059, "step": 1850 }, { "epoch": 0.14, "learning_rate": 0.00017279870749798047, "loss": 0.0059, "step": 1852 }, { "epoch": 0.14, "learning_rate": 0.00017276933245208195, "loss": 0.0056, "step": 1854 }, { "epoch": 0.14, "learning_rate": 0.00017273995740618345, "loss": 0.0057, "step": 1856 }, { "epoch": 0.14, "learning_rate": 0.00017271058236028493, "loss": 0.0059, "step": 1858 }, { "epoch": 0.14, "learning_rate": 0.00017268120731438644, "loss": 0.0059, "step": 1860 }, { "epoch": 0.14, "learning_rate": 0.00017265183226848794, "loss": 0.0058, "step": 1862 }, { "epoch": 0.14, "learning_rate": 0.00017262245722258942, "loss": 0.0058, "step": 1864 }, { "epoch": 0.14, "learning_rate": 0.0001725930821766909, "loss": 0.006, "step": 1866 }, { "epoch": 0.14, "learning_rate": 0.0001725637071307924, "loss": 0.0057, "step": 1868 }, { "epoch": 0.14, "learning_rate": 0.00017253433208489388, "loss": 0.0061, "step": 1870 }, { "epoch": 0.14, "learning_rate": 0.0001725049570389954, "loss": 0.0058, "step": 1872 }, { "epoch": 0.14, "learning_rate": 0.00017247558199309687, "loss": 0.0059, "step": 1874 }, { "epoch": 0.14, "learning_rate": 0.00017244620694719837, "loss": 0.0058, "step": 1876 }, { "epoch": 0.14, "learning_rate": 0.00017241683190129985, "loss": 0.006, "step": 1878 }, { "epoch": 0.14, "learning_rate": 0.00017238745685540133, "loss": 0.0063, "step": 1880 }, { "epoch": 0.14, "learning_rate": 0.00017235808180950284, "loss": 0.0058, "step": 1882 }, { "epoch": 0.14, "learning_rate": 0.00017232870676360434, "loss": 0.0057, "step": 1884 }, { "epoch": 0.14, "learning_rate": 0.00017229933171770582, "loss": 0.0058, "step": 1886 }, { "epoch": 0.14, "learning_rate": 0.0001722699566718073, "loss": 0.0059, "step": 1888 }, { "epoch": 0.14, "learning_rate": 0.0001722405816259088, "loss": 0.0055, "step": 1890 }, { "epoch": 0.14, "learning_rate": 0.00017221120658001028, "loss": 0.0059, "step": 1892 }, { "epoch": 0.14, "learning_rate": 0.0001721818315341118, "loss": 0.006, "step": 1894 }, { "epoch": 0.14, "learning_rate": 0.00017215245648821327, "loss": 0.006, "step": 1896 }, { "epoch": 0.14, "learning_rate": 0.00017212308144231477, "loss": 0.006, "step": 1898 }, { "epoch": 0.14, "learning_rate": 0.00017209370639641625, "loss": 0.0059, "step": 1900 }, { "epoch": 0.14, "learning_rate": 0.00017206433135051773, "loss": 0.006, "step": 1902 }, { "epoch": 0.14, "learning_rate": 0.00017203495630461923, "loss": 0.0057, "step": 1904 }, { "epoch": 0.14, "learning_rate": 0.00017200558125872074, "loss": 0.006, "step": 1906 }, { "epoch": 0.14, "learning_rate": 0.00017197620621282222, "loss": 0.006, "step": 1908 }, { "epoch": 0.14, "learning_rate": 0.00017194683116692372, "loss": 0.0056, "step": 1910 }, { "epoch": 0.14, "learning_rate": 0.0001719174561210252, "loss": 0.0061, "step": 1912 }, { "epoch": 0.14, "learning_rate": 0.00017188808107512668, "loss": 0.0061, "step": 1914 }, { "epoch": 0.14, "learning_rate": 0.00017185870602922818, "loss": 0.0057, "step": 1916 }, { "epoch": 0.14, "learning_rate": 0.00017182933098332966, "loss": 0.0056, "step": 1918 }, { "epoch": 0.14, "learning_rate": 0.00017179995593743117, "loss": 0.0059, "step": 1920 }, { "epoch": 0.14, "learning_rate": 0.00017177058089153267, "loss": 0.0058, "step": 1922 }, { "epoch": 0.14, "learning_rate": 0.00017174120584563413, "loss": 0.0058, "step": 1924 }, { "epoch": 0.14, "learning_rate": 0.00017171183079973563, "loss": 0.0058, "step": 1926 }, { "epoch": 0.14, "learning_rate": 0.00017168245575383714, "loss": 0.0058, "step": 1928 }, { "epoch": 0.14, "learning_rate": 0.00017165308070793861, "loss": 0.006, "step": 1930 }, { "epoch": 0.14, "learning_rate": 0.00017162370566204012, "loss": 0.0057, "step": 1932 }, { "epoch": 0.14, "learning_rate": 0.0001715943306161416, "loss": 0.006, "step": 1934 }, { "epoch": 0.14, "learning_rate": 0.00017156495557024308, "loss": 0.0057, "step": 1936 }, { "epoch": 0.14, "learning_rate": 0.00017153558052434458, "loss": 0.0058, "step": 1938 }, { "epoch": 0.14, "learning_rate": 0.00017150620547844606, "loss": 0.0058, "step": 1940 }, { "epoch": 0.14, "learning_rate": 0.00017147683043254757, "loss": 0.0057, "step": 1942 }, { "epoch": 0.14, "learning_rate": 0.00017144745538664907, "loss": 0.0058, "step": 1944 }, { "epoch": 0.14, "learning_rate": 0.00017141808034075052, "loss": 0.0054, "step": 1946 }, { "epoch": 0.14, "learning_rate": 0.00017138870529485203, "loss": 0.006, "step": 1948 }, { "epoch": 0.14, "learning_rate": 0.00017135933024895353, "loss": 0.0061, "step": 1950 }, { "epoch": 0.14, "learning_rate": 0.000171329955203055, "loss": 0.0057, "step": 1952 }, { "epoch": 0.14, "learning_rate": 0.00017130058015715652, "loss": 0.0061, "step": 1954 }, { "epoch": 0.14, "learning_rate": 0.000171271205111258, "loss": 0.0058, "step": 1956 }, { "epoch": 0.14, "learning_rate": 0.00017124183006535947, "loss": 0.0057, "step": 1958 }, { "epoch": 0.14, "learning_rate": 0.00017121245501946098, "loss": 0.0058, "step": 1960 }, { "epoch": 0.14, "learning_rate": 0.00017118307997356246, "loss": 0.0058, "step": 1962 }, { "epoch": 0.14, "learning_rate": 0.00017115370492766396, "loss": 0.0057, "step": 1964 }, { "epoch": 0.14, "learning_rate": 0.00017112432988176544, "loss": 0.0058, "step": 1966 }, { "epoch": 0.14, "learning_rate": 0.00017109495483586695, "loss": 0.0058, "step": 1968 }, { "epoch": 0.14, "learning_rate": 0.00017106557978996843, "loss": 0.0057, "step": 1970 }, { "epoch": 0.14, "learning_rate": 0.0001710362047440699, "loss": 0.0058, "step": 1972 }, { "epoch": 0.14, "learning_rate": 0.0001710068296981714, "loss": 0.0057, "step": 1974 }, { "epoch": 0.15, "learning_rate": 0.00017097745465227291, "loss": 0.0056, "step": 1976 }, { "epoch": 0.15, "learning_rate": 0.0001709480796063744, "loss": 0.0062, "step": 1978 }, { "epoch": 0.15, "learning_rate": 0.00017091870456047587, "loss": 0.0059, "step": 1980 }, { "epoch": 0.15, "learning_rate": 0.00017088932951457738, "loss": 0.0056, "step": 1982 }, { "epoch": 0.15, "learning_rate": 0.00017085995446867886, "loss": 0.0054, "step": 1984 }, { "epoch": 0.15, "learning_rate": 0.00017083057942278036, "loss": 0.0057, "step": 1986 }, { "epoch": 0.15, "learning_rate": 0.00017080120437688184, "loss": 0.0056, "step": 1988 }, { "epoch": 0.15, "learning_rate": 0.00017077182933098334, "loss": 0.0059, "step": 1990 }, { "epoch": 0.15, "learning_rate": 0.00017074245428508482, "loss": 0.0059, "step": 1992 }, { "epoch": 0.15, "learning_rate": 0.0001707130792391863, "loss": 0.0057, "step": 1994 }, { "epoch": 0.15, "learning_rate": 0.0001706837041932878, "loss": 0.0057, "step": 1996 }, { "epoch": 0.15, "learning_rate": 0.0001706543291473893, "loss": 0.0054, "step": 1998 }, { "epoch": 0.15, "learning_rate": 0.0001706249541014908, "loss": 0.0057, "step": 2000 }, { "epoch": 0.15, "learning_rate": 0.0001705955790555923, "loss": 0.0059, "step": 2002 }, { "epoch": 0.15, "learning_rate": 0.00017056620400969377, "loss": 0.006, "step": 2004 }, { "epoch": 0.15, "learning_rate": 0.00017053682896379525, "loss": 0.0058, "step": 2006 }, { "epoch": 0.15, "learning_rate": 0.00017050745391789676, "loss": 0.0058, "step": 2008 }, { "epoch": 0.15, "learning_rate": 0.00017047807887199824, "loss": 0.0056, "step": 2010 }, { "epoch": 0.15, "learning_rate": 0.00017044870382609974, "loss": 0.0058, "step": 2012 }, { "epoch": 0.15, "learning_rate": 0.00017041932878020125, "loss": 0.0057, "step": 2014 }, { "epoch": 0.15, "learning_rate": 0.0001703899537343027, "loss": 0.0061, "step": 2016 }, { "epoch": 0.15, "learning_rate": 0.0001703605786884042, "loss": 0.0059, "step": 2018 }, { "epoch": 0.15, "learning_rate": 0.0001703312036425057, "loss": 0.0059, "step": 2020 }, { "epoch": 0.15, "learning_rate": 0.0001703018285966072, "loss": 0.006, "step": 2022 }, { "epoch": 0.15, "learning_rate": 0.0001702724535507087, "loss": 0.0058, "step": 2024 }, { "epoch": 0.15, "learning_rate": 0.00017024307850481017, "loss": 0.0057, "step": 2026 }, { "epoch": 0.15, "learning_rate": 0.00017021370345891165, "loss": 0.0056, "step": 2028 }, { "epoch": 0.15, "learning_rate": 0.00017018432841301316, "loss": 0.0055, "step": 2030 }, { "epoch": 0.15, "learning_rate": 0.00017015495336711463, "loss": 0.0058, "step": 2032 }, { "epoch": 0.15, "learning_rate": 0.00017012557832121614, "loss": 0.0056, "step": 2034 }, { "epoch": 0.15, "learning_rate": 0.00017009620327531765, "loss": 0.0059, "step": 2036 }, { "epoch": 0.15, "learning_rate": 0.0001700668282294191, "loss": 0.0059, "step": 2038 }, { "epoch": 0.15, "learning_rate": 0.0001700374531835206, "loss": 0.0058, "step": 2040 }, { "epoch": 0.15, "learning_rate": 0.0001700080781376221, "loss": 0.0057, "step": 2042 }, { "epoch": 0.15, "learning_rate": 0.00016997870309172359, "loss": 0.0058, "step": 2044 }, { "epoch": 0.15, "learning_rate": 0.0001699493280458251, "loss": 0.0057, "step": 2046 }, { "epoch": 0.15, "learning_rate": 0.00016991995299992657, "loss": 0.0059, "step": 2048 }, { "epoch": 0.15, "learning_rate": 0.00016989057795402805, "loss": 0.0056, "step": 2050 }, { "epoch": 0.15, "learning_rate": 0.00016986120290812955, "loss": 0.0057, "step": 2052 }, { "epoch": 0.15, "learning_rate": 0.00016983182786223103, "loss": 0.0058, "step": 2054 }, { "epoch": 0.15, "learning_rate": 0.00016980245281633254, "loss": 0.0057, "step": 2056 }, { "epoch": 0.15, "learning_rate": 0.00016977307777043404, "loss": 0.0058, "step": 2058 }, { "epoch": 0.15, "learning_rate": 0.00016974370272453552, "loss": 0.0059, "step": 2060 }, { "epoch": 0.15, "learning_rate": 0.000169714327678637, "loss": 0.0059, "step": 2062 }, { "epoch": 0.15, "learning_rate": 0.0001696849526327385, "loss": 0.0058, "step": 2064 }, { "epoch": 0.15, "learning_rate": 0.00016965557758683998, "loss": 0.0056, "step": 2066 }, { "epoch": 0.15, "learning_rate": 0.0001696262025409415, "loss": 0.0055, "step": 2068 }, { "epoch": 0.15, "learning_rate": 0.00016959682749504297, "loss": 0.0059, "step": 2070 }, { "epoch": 0.15, "learning_rate": 0.00016956745244914447, "loss": 0.0058, "step": 2072 }, { "epoch": 0.15, "learning_rate": 0.00016953807740324595, "loss": 0.0056, "step": 2074 }, { "epoch": 0.15, "learning_rate": 0.00016950870235734743, "loss": 0.0058, "step": 2076 }, { "epoch": 0.15, "learning_rate": 0.00016947932731144894, "loss": 0.0058, "step": 2078 }, { "epoch": 0.15, "learning_rate": 0.0001694499522655504, "loss": 0.0057, "step": 2080 }, { "epoch": 0.15, "learning_rate": 0.00016942057721965192, "loss": 0.0053, "step": 2082 }, { "epoch": 0.15, "learning_rate": 0.0001693912021737534, "loss": 0.0059, "step": 2084 }, { "epoch": 0.15, "learning_rate": 0.00016936182712785488, "loss": 0.0059, "step": 2086 }, { "epoch": 0.15, "learning_rate": 0.00016933245208195638, "loss": 0.0059, "step": 2088 }, { "epoch": 0.15, "learning_rate": 0.0001693030770360579, "loss": 0.0058, "step": 2090 }, { "epoch": 0.15, "learning_rate": 0.00016927370199015937, "loss": 0.0058, "step": 2092 }, { "epoch": 0.15, "learning_rate": 0.00016924432694426087, "loss": 0.0055, "step": 2094 }, { "epoch": 0.15, "learning_rate": 0.00016921495189836235, "loss": 0.0056, "step": 2096 }, { "epoch": 0.15, "learning_rate": 0.00016918557685246383, "loss": 0.0054, "step": 2098 }, { "epoch": 0.15, "learning_rate": 0.00016915620180656533, "loss": 0.0055, "step": 2100 }, { "epoch": 0.15, "learning_rate": 0.0001691268267606668, "loss": 0.0059, "step": 2102 }, { "epoch": 0.15, "learning_rate": 0.00016909745171476832, "loss": 0.0059, "step": 2104 }, { "epoch": 0.15, "learning_rate": 0.00016906807666886982, "loss": 0.0058, "step": 2106 }, { "epoch": 0.15, "learning_rate": 0.00016903870162297127, "loss": 0.0059, "step": 2108 }, { "epoch": 0.15, "learning_rate": 0.00016900932657707278, "loss": 0.0058, "step": 2110 }, { "epoch": 0.16, "learning_rate": 0.00016897995153117428, "loss": 0.0058, "step": 2112 }, { "epoch": 0.16, "learning_rate": 0.00016895057648527576, "loss": 0.0058, "step": 2114 }, { "epoch": 0.16, "learning_rate": 0.00016892120143937727, "loss": 0.0057, "step": 2116 }, { "epoch": 0.16, "learning_rate": 0.00016889182639347875, "loss": 0.0056, "step": 2118 }, { "epoch": 0.16, "learning_rate": 0.00016886245134758022, "loss": 0.0056, "step": 2120 }, { "epoch": 0.16, "learning_rate": 0.00016883307630168173, "loss": 0.0057, "step": 2122 }, { "epoch": 0.16, "learning_rate": 0.0001688037012557832, "loss": 0.006, "step": 2124 }, { "epoch": 0.16, "learning_rate": 0.00016877432620988471, "loss": 0.0058, "step": 2126 }, { "epoch": 0.16, "learning_rate": 0.00016874495116398622, "loss": 0.0057, "step": 2128 }, { "epoch": 0.16, "learning_rate": 0.0001687155761180877, "loss": 0.0059, "step": 2130 }, { "epoch": 0.16, "learning_rate": 0.00016868620107218918, "loss": 0.0057, "step": 2132 }, { "epoch": 0.16, "learning_rate": 0.00016865682602629068, "loss": 0.0057, "step": 2134 }, { "epoch": 0.16, "learning_rate": 0.00016862745098039216, "loss": 0.0058, "step": 2136 }, { "epoch": 0.16, "learning_rate": 0.00016859807593449367, "loss": 0.0057, "step": 2138 }, { "epoch": 0.16, "learning_rate": 0.00016856870088859514, "loss": 0.0057, "step": 2140 }, { "epoch": 0.16, "learning_rate": 0.00016853932584269662, "loss": 0.0057, "step": 2142 }, { "epoch": 0.16, "learning_rate": 0.00016850995079679813, "loss": 0.0057, "step": 2144 }, { "epoch": 0.16, "learning_rate": 0.0001684805757508996, "loss": 0.0059, "step": 2146 }, { "epoch": 0.16, "learning_rate": 0.0001684512007050011, "loss": 0.0056, "step": 2148 }, { "epoch": 0.16, "learning_rate": 0.00016842182565910262, "loss": 0.006, "step": 2150 }, { "epoch": 0.16, "learning_rate": 0.0001683924506132041, "loss": 0.0056, "step": 2152 }, { "epoch": 0.16, "learning_rate": 0.00016836307556730557, "loss": 0.0056, "step": 2154 }, { "epoch": 0.16, "learning_rate": 0.00016833370052140708, "loss": 0.0056, "step": 2156 }, { "epoch": 0.16, "learning_rate": 0.00016830432547550856, "loss": 0.0054, "step": 2158 }, { "epoch": 0.16, "learning_rate": 0.00016827495042961006, "loss": 0.0057, "step": 2160 }, { "epoch": 0.16, "learning_rate": 0.00016824557538371154, "loss": 0.0057, "step": 2162 }, { "epoch": 0.16, "learning_rate": 0.00016821620033781305, "loss": 0.0057, "step": 2164 }, { "epoch": 0.16, "learning_rate": 0.00016818682529191453, "loss": 0.0056, "step": 2166 }, { "epoch": 0.16, "learning_rate": 0.000168157450246016, "loss": 0.0056, "step": 2168 }, { "epoch": 0.16, "learning_rate": 0.0001681280752001175, "loss": 0.0058, "step": 2170 }, { "epoch": 0.16, "learning_rate": 0.00016809870015421901, "loss": 0.0055, "step": 2172 }, { "epoch": 0.16, "learning_rate": 0.0001680693251083205, "loss": 0.0058, "step": 2174 }, { "epoch": 0.16, "learning_rate": 0.000168039950062422, "loss": 0.0055, "step": 2176 }, { "epoch": 0.16, "learning_rate": 0.00016801057501652348, "loss": 0.0056, "step": 2178 }, { "epoch": 0.16, "learning_rate": 0.00016798119997062496, "loss": 0.0056, "step": 2180 }, { "epoch": 0.16, "learning_rate": 0.00016795182492472646, "loss": 0.0057, "step": 2182 }, { "epoch": 0.16, "learning_rate": 0.00016792244987882794, "loss": 0.0056, "step": 2184 }, { "epoch": 0.16, "learning_rate": 0.00016789307483292944, "loss": 0.0056, "step": 2186 }, { "epoch": 0.16, "learning_rate": 0.00016786369978703092, "loss": 0.0058, "step": 2188 }, { "epoch": 0.16, "learning_rate": 0.0001678343247411324, "loss": 0.0054, "step": 2190 }, { "epoch": 0.16, "learning_rate": 0.0001678049496952339, "loss": 0.0056, "step": 2192 }, { "epoch": 0.16, "learning_rate": 0.00016777557464933539, "loss": 0.0056, "step": 2194 }, { "epoch": 0.16, "learning_rate": 0.0001677461996034369, "loss": 0.0058, "step": 2196 }, { "epoch": 0.16, "learning_rate": 0.0001677168245575384, "loss": 0.0059, "step": 2198 }, { "epoch": 0.16, "learning_rate": 0.00016768744951163985, "loss": 0.0057, "step": 2200 }, { "epoch": 0.16, "learning_rate": 0.00016765807446574135, "loss": 0.0056, "step": 2202 }, { "epoch": 0.16, "learning_rate": 0.00016762869941984286, "loss": 0.0061, "step": 2204 }, { "epoch": 0.16, "learning_rate": 0.00016759932437394434, "loss": 0.0057, "step": 2206 }, { "epoch": 0.16, "learning_rate": 0.00016756994932804584, "loss": 0.0054, "step": 2208 }, { "epoch": 0.16, "learning_rate": 0.00016754057428214732, "loss": 0.0057, "step": 2210 }, { "epoch": 0.16, "learning_rate": 0.0001675111992362488, "loss": 0.0059, "step": 2212 }, { "epoch": 0.16, "learning_rate": 0.0001674818241903503, "loss": 0.0058, "step": 2214 }, { "epoch": 0.16, "learning_rate": 0.00016745244914445178, "loss": 0.0056, "step": 2216 }, { "epoch": 0.16, "learning_rate": 0.0001674230740985533, "loss": 0.0056, "step": 2218 }, { "epoch": 0.16, "learning_rate": 0.0001673936990526548, "loss": 0.0057, "step": 2220 }, { "epoch": 0.16, "learning_rate": 0.00016736432400675627, "loss": 0.0055, "step": 2222 }, { "epoch": 0.16, "learning_rate": 0.00016733494896085775, "loss": 0.0057, "step": 2224 }, { "epoch": 0.16, "learning_rate": 0.00016730557391495926, "loss": 0.0057, "step": 2226 }, { "epoch": 0.16, "learning_rate": 0.00016727619886906073, "loss": 0.0055, "step": 2228 }, { "epoch": 0.16, "learning_rate": 0.00016724682382316224, "loss": 0.0054, "step": 2230 }, { "epoch": 0.16, "learning_rate": 0.00016721744877726372, "loss": 0.0059, "step": 2232 }, { "epoch": 0.16, "learning_rate": 0.0001671880737313652, "loss": 0.0058, "step": 2234 }, { "epoch": 0.16, "learning_rate": 0.0001671586986854667, "loss": 0.0058, "step": 2236 }, { "epoch": 0.16, "learning_rate": 0.00016712932363956818, "loss": 0.0058, "step": 2238 }, { "epoch": 0.16, "learning_rate": 0.00016709994859366969, "loss": 0.0056, "step": 2240 }, { "epoch": 0.16, "learning_rate": 0.0001670705735477712, "loss": 0.0056, "step": 2242 }, { "epoch": 0.16, "learning_rate": 0.00016704119850187267, "loss": 0.0058, "step": 2244 }, { "epoch": 0.16, "learning_rate": 0.00016701182345597415, "loss": 0.0059, "step": 2246 }, { "epoch": 0.17, "learning_rate": 0.00016698244841007565, "loss": 0.0061, "step": 2248 }, { "epoch": 0.17, "learning_rate": 0.00016695307336417713, "loss": 0.0056, "step": 2250 }, { "epoch": 0.17, "learning_rate": 0.00016692369831827864, "loss": 0.0056, "step": 2252 }, { "epoch": 0.17, "learning_rate": 0.00016689432327238012, "loss": 0.0055, "step": 2254 }, { "epoch": 0.17, "learning_rate": 0.00016686494822648162, "loss": 0.0056, "step": 2256 }, { "epoch": 0.17, "learning_rate": 0.0001668355731805831, "loss": 0.0057, "step": 2258 }, { "epoch": 0.17, "learning_rate": 0.00016680619813468458, "loss": 0.0055, "step": 2260 }, { "epoch": 0.17, "learning_rate": 0.00016677682308878608, "loss": 0.0058, "step": 2262 }, { "epoch": 0.17, "learning_rate": 0.0001667474480428876, "loss": 0.006, "step": 2264 }, { "epoch": 0.17, "learning_rate": 0.00016671807299698907, "loss": 0.0056, "step": 2266 }, { "epoch": 0.17, "learning_rate": 0.00016668869795109057, "loss": 0.0059, "step": 2268 }, { "epoch": 0.17, "learning_rate": 0.00016665932290519205, "loss": 0.0056, "step": 2270 }, { "epoch": 0.17, "learning_rate": 0.00016662994785929353, "loss": 0.0057, "step": 2272 }, { "epoch": 0.17, "learning_rate": 0.00016660057281339503, "loss": 0.0054, "step": 2274 }, { "epoch": 0.17, "learning_rate": 0.0001665711977674965, "loss": 0.0059, "step": 2276 }, { "epoch": 0.17, "learning_rate": 0.00016654182272159802, "loss": 0.0058, "step": 2278 }, { "epoch": 0.17, "learning_rate": 0.00016651244767569952, "loss": 0.0058, "step": 2280 }, { "epoch": 0.17, "learning_rate": 0.00016648307262980098, "loss": 0.0057, "step": 2282 }, { "epoch": 0.17, "learning_rate": 0.00016645369758390248, "loss": 0.0058, "step": 2284 }, { "epoch": 0.17, "learning_rate": 0.00016642432253800399, "loss": 0.0056, "step": 2286 }, { "epoch": 0.17, "learning_rate": 0.00016639494749210546, "loss": 0.0055, "step": 2288 }, { "epoch": 0.17, "learning_rate": 0.00016636557244620697, "loss": 0.0058, "step": 2290 }, { "epoch": 0.17, "learning_rate": 0.00016633619740030845, "loss": 0.0057, "step": 2292 }, { "epoch": 0.17, "learning_rate": 0.00016630682235440993, "loss": 0.0058, "step": 2294 }, { "epoch": 0.17, "learning_rate": 0.00016627744730851143, "loss": 0.0056, "step": 2296 }, { "epoch": 0.17, "learning_rate": 0.0001662480722626129, "loss": 0.0057, "step": 2298 }, { "epoch": 0.17, "learning_rate": 0.00016621869721671442, "loss": 0.0055, "step": 2300 }, { "epoch": 0.17, "learning_rate": 0.00016618932217081592, "loss": 0.0056, "step": 2302 }, { "epoch": 0.17, "learning_rate": 0.00016615994712491737, "loss": 0.0058, "step": 2304 }, { "epoch": 0.17, "learning_rate": 0.00016613057207901888, "loss": 0.0056, "step": 2306 }, { "epoch": 0.17, "learning_rate": 0.00016610119703312038, "loss": 0.0058, "step": 2308 }, { "epoch": 0.17, "learning_rate": 0.00016607182198722186, "loss": 0.0057, "step": 2310 }, { "epoch": 0.17, "learning_rate": 0.00016604244694132337, "loss": 0.0057, "step": 2312 }, { "epoch": 0.17, "learning_rate": 0.00016601307189542485, "loss": 0.0057, "step": 2314 }, { "epoch": 0.17, "learning_rate": 0.00016598369684952632, "loss": 0.0056, "step": 2316 }, { "epoch": 0.17, "learning_rate": 0.00016595432180362783, "loss": 0.006, "step": 2318 }, { "epoch": 0.17, "learning_rate": 0.0001659249467577293, "loss": 0.0059, "step": 2320 }, { "epoch": 0.17, "learning_rate": 0.00016589557171183081, "loss": 0.0056, "step": 2322 }, { "epoch": 0.17, "learning_rate": 0.0001658661966659323, "loss": 0.0058, "step": 2324 }, { "epoch": 0.17, "learning_rate": 0.0001658368216200338, "loss": 0.0056, "step": 2326 }, { "epoch": 0.17, "learning_rate": 0.00016580744657413528, "loss": 0.0057, "step": 2328 }, { "epoch": 0.17, "learning_rate": 0.00016577807152823675, "loss": 0.0055, "step": 2330 }, { "epoch": 0.17, "learning_rate": 0.00016574869648233826, "loss": 0.0056, "step": 2332 }, { "epoch": 0.17, "learning_rate": 0.00016571932143643977, "loss": 0.0057, "step": 2334 }, { "epoch": 0.17, "learning_rate": 0.00016568994639054124, "loss": 0.0056, "step": 2336 }, { "epoch": 0.17, "learning_rate": 0.00016566057134464272, "loss": 0.0055, "step": 2338 }, { "epoch": 0.17, "learning_rate": 0.00016563119629874423, "loss": 0.0055, "step": 2340 }, { "epoch": 0.17, "learning_rate": 0.0001656018212528457, "loss": 0.0055, "step": 2342 }, { "epoch": 0.17, "learning_rate": 0.0001655724462069472, "loss": 0.0059, "step": 2344 }, { "epoch": 0.17, "learning_rate": 0.0001655430711610487, "loss": 0.0055, "step": 2346 }, { "epoch": 0.17, "learning_rate": 0.0001655136961151502, "loss": 0.0059, "step": 2348 }, { "epoch": 0.17, "learning_rate": 0.00016548432106925167, "loss": 0.0054, "step": 2350 }, { "epoch": 0.17, "learning_rate": 0.00016545494602335315, "loss": 0.0058, "step": 2352 }, { "epoch": 0.17, "learning_rate": 0.00016542557097745466, "loss": 0.0058, "step": 2354 }, { "epoch": 0.17, "learning_rate": 0.00016539619593155616, "loss": 0.0056, "step": 2356 }, { "epoch": 0.17, "learning_rate": 0.00016536682088565764, "loss": 0.0057, "step": 2358 }, { "epoch": 0.17, "learning_rate": 0.00016533744583975915, "loss": 0.0058, "step": 2360 }, { "epoch": 0.17, "learning_rate": 0.00016530807079386063, "loss": 0.0058, "step": 2362 }, { "epoch": 0.17, "learning_rate": 0.0001652786957479621, "loss": 0.0056, "step": 2364 }, { "epoch": 0.17, "learning_rate": 0.0001652493207020636, "loss": 0.0057, "step": 2366 }, { "epoch": 0.17, "learning_rate": 0.0001652199456561651, "loss": 0.0056, "step": 2368 }, { "epoch": 0.17, "learning_rate": 0.0001651905706102666, "loss": 0.0056, "step": 2370 }, { "epoch": 0.17, "learning_rate": 0.0001651611955643681, "loss": 0.0056, "step": 2372 }, { "epoch": 0.17, "learning_rate": 0.00016513182051846955, "loss": 0.006, "step": 2374 }, { "epoch": 0.17, "learning_rate": 0.00016510244547257106, "loss": 0.006, "step": 2376 }, { "epoch": 0.17, "learning_rate": 0.00016507307042667256, "loss": 0.0057, "step": 2378 }, { "epoch": 0.17, "learning_rate": 0.00016504369538077404, "loss": 0.0055, "step": 2380 }, { "epoch": 0.17, "learning_rate": 0.00016501432033487554, "loss": 0.0058, "step": 2382 }, { "epoch": 0.18, "learning_rate": 0.00016498494528897702, "loss": 0.0056, "step": 2384 }, { "epoch": 0.18, "learning_rate": 0.0001649555702430785, "loss": 0.0058, "step": 2386 }, { "epoch": 0.18, "learning_rate": 0.00016492619519718, "loss": 0.0055, "step": 2388 }, { "epoch": 0.18, "learning_rate": 0.00016489682015128148, "loss": 0.0057, "step": 2390 }, { "epoch": 0.18, "learning_rate": 0.000164867445105383, "loss": 0.0056, "step": 2392 }, { "epoch": 0.18, "learning_rate": 0.0001648380700594845, "loss": 0.0057, "step": 2394 }, { "epoch": 0.18, "learning_rate": 0.00016480869501358595, "loss": 0.0053, "step": 2396 }, { "epoch": 0.18, "learning_rate": 0.00016477931996768745, "loss": 0.0056, "step": 2398 }, { "epoch": 0.18, "learning_rate": 0.00016474994492178896, "loss": 0.0058, "step": 2400 }, { "epoch": 0.18, "learning_rate": 0.00016472056987589044, "loss": 0.0059, "step": 2402 }, { "epoch": 0.18, "learning_rate": 0.00016469119482999194, "loss": 0.0058, "step": 2404 }, { "epoch": 0.18, "learning_rate": 0.00016466181978409342, "loss": 0.0057, "step": 2406 }, { "epoch": 0.18, "learning_rate": 0.0001646324447381949, "loss": 0.0058, "step": 2408 }, { "epoch": 0.18, "learning_rate": 0.0001646030696922964, "loss": 0.0057, "step": 2410 }, { "epoch": 0.18, "learning_rate": 0.00016457369464639788, "loss": 0.0059, "step": 2412 }, { "epoch": 0.18, "learning_rate": 0.0001645443196004994, "loss": 0.0057, "step": 2414 }, { "epoch": 0.18, "learning_rate": 0.0001645149445546009, "loss": 0.0055, "step": 2416 }, { "epoch": 0.18, "learning_rate": 0.00016448556950870237, "loss": 0.0055, "step": 2418 }, { "epoch": 0.18, "learning_rate": 0.00016445619446280385, "loss": 0.0056, "step": 2420 }, { "epoch": 0.18, "learning_rate": 0.00016442681941690536, "loss": 0.0056, "step": 2422 }, { "epoch": 0.18, "learning_rate": 0.00016439744437100683, "loss": 0.0056, "step": 2424 }, { "epoch": 0.18, "learning_rate": 0.00016436806932510834, "loss": 0.0057, "step": 2426 }, { "epoch": 0.18, "learning_rate": 0.00016433869427920982, "loss": 0.0056, "step": 2428 }, { "epoch": 0.18, "learning_rate": 0.00016430931923331132, "loss": 0.0059, "step": 2430 }, { "epoch": 0.18, "learning_rate": 0.0001642799441874128, "loss": 0.0056, "step": 2432 }, { "epoch": 0.18, "learning_rate": 0.00016425056914151428, "loss": 0.0059, "step": 2434 }, { "epoch": 0.18, "learning_rate": 0.00016422119409561579, "loss": 0.0055, "step": 2436 }, { "epoch": 0.18, "learning_rate": 0.00016419181904971726, "loss": 0.0055, "step": 2438 }, { "epoch": 0.18, "learning_rate": 0.00016416244400381877, "loss": 0.0055, "step": 2440 }, { "epoch": 0.18, "learning_rate": 0.00016413306895792025, "loss": 0.0055, "step": 2442 }, { "epoch": 0.18, "learning_rate": 0.00016410369391202173, "loss": 0.0056, "step": 2444 }, { "epoch": 0.18, "learning_rate": 0.00016407431886612323, "loss": 0.0056, "step": 2446 }, { "epoch": 0.18, "learning_rate": 0.00016404494382022474, "loss": 0.0056, "step": 2448 }, { "epoch": 0.18, "learning_rate": 0.00016401556877432622, "loss": 0.0056, "step": 2450 }, { "epoch": 0.18, "learning_rate": 0.00016398619372842772, "loss": 0.0059, "step": 2452 }, { "epoch": 0.18, "learning_rate": 0.0001639568186825292, "loss": 0.0058, "step": 2454 }, { "epoch": 0.18, "learning_rate": 0.00016392744363663068, "loss": 0.0056, "step": 2456 }, { "epoch": 0.18, "learning_rate": 0.00016389806859073218, "loss": 0.0056, "step": 2458 }, { "epoch": 0.18, "learning_rate": 0.00016386869354483366, "loss": 0.0055, "step": 2460 }, { "epoch": 0.18, "learning_rate": 0.00016383931849893517, "loss": 0.0056, "step": 2462 }, { "epoch": 0.18, "learning_rate": 0.00016380994345303667, "loss": 0.0056, "step": 2464 }, { "epoch": 0.18, "learning_rate": 0.00016378056840713812, "loss": 0.0056, "step": 2466 }, { "epoch": 0.18, "learning_rate": 0.00016375119336123963, "loss": 0.0058, "step": 2468 }, { "epoch": 0.18, "learning_rate": 0.00016372181831534113, "loss": 0.0058, "step": 2470 }, { "epoch": 0.18, "learning_rate": 0.0001636924432694426, "loss": 0.0056, "step": 2472 }, { "epoch": 0.18, "learning_rate": 0.00016366306822354412, "loss": 0.0058, "step": 2474 }, { "epoch": 0.18, "learning_rate": 0.0001636336931776456, "loss": 0.0056, "step": 2476 }, { "epoch": 0.18, "learning_rate": 0.00016360431813174708, "loss": 0.0056, "step": 2478 }, { "epoch": 0.18, "learning_rate": 0.00016357494308584858, "loss": 0.0059, "step": 2480 }, { "epoch": 0.18, "learning_rate": 0.00016354556803995006, "loss": 0.0055, "step": 2482 }, { "epoch": 0.18, "learning_rate": 0.00016351619299405156, "loss": 0.0058, "step": 2484 }, { "epoch": 0.18, "learning_rate": 0.00016348681794815307, "loss": 0.0057, "step": 2486 }, { "epoch": 0.18, "learning_rate": 0.00016345744290225452, "loss": 0.0057, "step": 2488 }, { "epoch": 0.18, "learning_rate": 0.00016342806785635603, "loss": 0.0057, "step": 2490 }, { "epoch": 0.18, "learning_rate": 0.00016339869281045753, "loss": 0.0056, "step": 2492 }, { "epoch": 0.18, "learning_rate": 0.000163369317764559, "loss": 0.0056, "step": 2494 }, { "epoch": 0.18, "learning_rate": 0.00016333994271866052, "loss": 0.0059, "step": 2496 }, { "epoch": 0.18, "learning_rate": 0.000163310567672762, "loss": 0.0056, "step": 2498 }, { "epoch": 0.18, "learning_rate": 0.00016328119262686347, "loss": 0.0057, "step": 2500 }, { "epoch": 0.18, "learning_rate": 0.00016325181758096498, "loss": 0.0055, "step": 2502 }, { "epoch": 0.18, "learning_rate": 0.00016322244253506646, "loss": 0.0057, "step": 2504 }, { "epoch": 0.18, "learning_rate": 0.00016319306748916796, "loss": 0.0056, "step": 2506 }, { "epoch": 0.18, "learning_rate": 0.00016316369244326947, "loss": 0.0054, "step": 2508 }, { "epoch": 0.18, "learning_rate": 0.00016313431739737095, "loss": 0.0055, "step": 2510 }, { "epoch": 0.18, "learning_rate": 0.00016310494235147242, "loss": 0.0055, "step": 2512 }, { "epoch": 0.18, "learning_rate": 0.00016307556730557393, "loss": 0.0056, "step": 2514 }, { "epoch": 0.18, "learning_rate": 0.0001630461922596754, "loss": 0.0055, "step": 2516 }, { "epoch": 0.18, "learning_rate": 0.0001630168172137769, "loss": 0.0055, "step": 2518 }, { "epoch": 0.19, "learning_rate": 0.0001629874421678784, "loss": 0.0056, "step": 2520 }, { "epoch": 0.19, "learning_rate": 0.0001629580671219799, "loss": 0.0057, "step": 2522 }, { "epoch": 0.19, "learning_rate": 0.00016292869207608138, "loss": 0.0059, "step": 2524 }, { "epoch": 0.19, "learning_rate": 0.00016289931703018285, "loss": 0.0057, "step": 2526 }, { "epoch": 0.19, "learning_rate": 0.00016286994198428436, "loss": 0.0056, "step": 2528 }, { "epoch": 0.19, "learning_rate": 0.00016284056693838587, "loss": 0.0058, "step": 2530 }, { "epoch": 0.19, "learning_rate": 0.00016281119189248734, "loss": 0.0057, "step": 2532 }, { "epoch": 0.19, "learning_rate": 0.00016278181684658885, "loss": 0.0057, "step": 2534 }, { "epoch": 0.19, "learning_rate": 0.00016275244180069033, "loss": 0.0057, "step": 2536 }, { "epoch": 0.19, "learning_rate": 0.0001627230667547918, "loss": 0.0054, "step": 2538 }, { "epoch": 0.19, "learning_rate": 0.0001626936917088933, "loss": 0.0055, "step": 2540 }, { "epoch": 0.19, "learning_rate": 0.0001626643166629948, "loss": 0.0057, "step": 2542 }, { "epoch": 0.19, "learning_rate": 0.0001626349416170963, "loss": 0.0059, "step": 2544 }, { "epoch": 0.19, "learning_rate": 0.00016260556657119777, "loss": 0.0056, "step": 2546 }, { "epoch": 0.19, "learning_rate": 0.00016257619152529925, "loss": 0.0057, "step": 2548 }, { "epoch": 0.19, "learning_rate": 0.00016254681647940076, "loss": 0.0056, "step": 2550 }, { "epoch": 0.19, "learning_rate": 0.00016251744143350224, "loss": 0.0054, "step": 2552 }, { "epoch": 0.19, "learning_rate": 0.00016248806638760374, "loss": 0.0059, "step": 2554 }, { "epoch": 0.19, "learning_rate": 0.00016245869134170525, "loss": 0.0056, "step": 2556 }, { "epoch": 0.19, "learning_rate": 0.0001624293162958067, "loss": 0.0056, "step": 2558 }, { "epoch": 0.19, "learning_rate": 0.0001623999412499082, "loss": 0.0056, "step": 2560 }, { "epoch": 0.19, "learning_rate": 0.0001623705662040097, "loss": 0.0058, "step": 2562 }, { "epoch": 0.19, "learning_rate": 0.0001623411911581112, "loss": 0.0055, "step": 2564 }, { "epoch": 0.19, "learning_rate": 0.0001623118161122127, "loss": 0.0058, "step": 2566 }, { "epoch": 0.19, "learning_rate": 0.00016228244106631417, "loss": 0.0054, "step": 2568 }, { "epoch": 0.19, "learning_rate": 0.00016225306602041565, "loss": 0.0056, "step": 2570 }, { "epoch": 0.19, "learning_rate": 0.00016222369097451715, "loss": 0.0056, "step": 2572 }, { "epoch": 0.19, "learning_rate": 0.00016219431592861863, "loss": 0.0058, "step": 2574 }, { "epoch": 0.19, "learning_rate": 0.00016216494088272014, "loss": 0.0058, "step": 2576 }, { "epoch": 0.19, "learning_rate": 0.00016213556583682164, "loss": 0.0057, "step": 2578 }, { "epoch": 0.19, "learning_rate": 0.00016210619079092312, "loss": 0.0057, "step": 2580 }, { "epoch": 0.19, "learning_rate": 0.0001620768157450246, "loss": 0.0059, "step": 2582 }, { "epoch": 0.19, "learning_rate": 0.0001620474406991261, "loss": 0.006, "step": 2584 }, { "epoch": 0.19, "learning_rate": 0.00016201806565322758, "loss": 0.0056, "step": 2586 }, { "epoch": 0.19, "learning_rate": 0.0001619886906073291, "loss": 0.0059, "step": 2588 }, { "epoch": 0.19, "learning_rate": 0.00016195931556143057, "loss": 0.0057, "step": 2590 }, { "epoch": 0.19, "learning_rate": 0.00016192994051553205, "loss": 0.0057, "step": 2592 }, { "epoch": 0.19, "learning_rate": 0.00016190056546963355, "loss": 0.0058, "step": 2594 }, { "epoch": 0.19, "learning_rate": 0.00016187119042373503, "loss": 0.0056, "step": 2596 }, { "epoch": 0.19, "learning_rate": 0.00016184181537783654, "loss": 0.0056, "step": 2598 }, { "epoch": 0.19, "learning_rate": 0.00016181244033193804, "loss": 0.0056, "step": 2600 }, { "epoch": 0.19, "learning_rate": 0.00016178306528603952, "loss": 0.0055, "step": 2602 }, { "epoch": 0.19, "learning_rate": 0.000161753690240141, "loss": 0.0056, "step": 2604 }, { "epoch": 0.19, "learning_rate": 0.0001617243151942425, "loss": 0.0057, "step": 2606 }, { "epoch": 0.19, "learning_rate": 0.00016169494014834398, "loss": 0.0056, "step": 2608 }, { "epoch": 0.19, "learning_rate": 0.0001616655651024455, "loss": 0.0058, "step": 2610 }, { "epoch": 0.19, "learning_rate": 0.00016163619005654697, "loss": 0.0058, "step": 2612 }, { "epoch": 0.19, "learning_rate": 0.00016160681501064847, "loss": 0.0054, "step": 2614 }, { "epoch": 0.19, "learning_rate": 0.00016157743996474995, "loss": 0.0055, "step": 2616 }, { "epoch": 0.19, "learning_rate": 0.00016154806491885143, "loss": 0.0057, "step": 2618 }, { "epoch": 0.19, "learning_rate": 0.00016151868987295293, "loss": 0.0054, "step": 2620 }, { "epoch": 0.19, "learning_rate": 0.00016148931482705444, "loss": 0.0057, "step": 2622 }, { "epoch": 0.19, "learning_rate": 0.00016145993978115592, "loss": 0.0057, "step": 2624 }, { "epoch": 0.19, "learning_rate": 0.00016143056473525742, "loss": 0.0056, "step": 2626 }, { "epoch": 0.19, "learning_rate": 0.0001614011896893589, "loss": 0.0054, "step": 2628 }, { "epoch": 0.19, "learning_rate": 0.00016137181464346038, "loss": 0.0058, "step": 2630 }, { "epoch": 0.19, "learning_rate": 0.00016134243959756189, "loss": 0.0057, "step": 2632 }, { "epoch": 0.19, "learning_rate": 0.00016131306455166336, "loss": 0.0059, "step": 2634 }, { "epoch": 0.19, "learning_rate": 0.00016128368950576487, "loss": 0.0058, "step": 2636 }, { "epoch": 0.19, "learning_rate": 0.00016125431445986635, "loss": 0.0055, "step": 2638 }, { "epoch": 0.19, "learning_rate": 0.00016122493941396783, "loss": 0.0057, "step": 2640 }, { "epoch": 0.19, "learning_rate": 0.00016119556436806933, "loss": 0.0058, "step": 2642 }, { "epoch": 0.19, "learning_rate": 0.00016116618932217084, "loss": 0.0057, "step": 2644 }, { "epoch": 0.19, "learning_rate": 0.00016113681427627232, "loss": 0.0056, "step": 2646 }, { "epoch": 0.19, "learning_rate": 0.00016110743923037382, "loss": 0.0055, "step": 2648 }, { "epoch": 0.19, "learning_rate": 0.0001610780641844753, "loss": 0.0057, "step": 2650 }, { "epoch": 0.19, "learning_rate": 0.00016104868913857678, "loss": 0.0058, "step": 2652 }, { "epoch": 0.19, "learning_rate": 0.00016101931409267828, "loss": 0.0055, "step": 2654 }, { "epoch": 0.2, "learning_rate": 0.00016098993904677976, "loss": 0.0057, "step": 2656 }, { "epoch": 0.2, "learning_rate": 0.00016096056400088127, "loss": 0.0056, "step": 2658 }, { "epoch": 0.2, "learning_rate": 0.00016093118895498277, "loss": 0.0058, "step": 2660 }, { "epoch": 0.2, "learning_rate": 0.00016090181390908422, "loss": 0.0058, "step": 2662 }, { "epoch": 0.2, "learning_rate": 0.00016087243886318573, "loss": 0.0055, "step": 2664 }, { "epoch": 0.2, "learning_rate": 0.0001608430638172872, "loss": 0.0056, "step": 2666 }, { "epoch": 0.2, "learning_rate": 0.0001608136887713887, "loss": 0.0058, "step": 2668 }, { "epoch": 0.2, "learning_rate": 0.00016078431372549022, "loss": 0.0057, "step": 2670 }, { "epoch": 0.2, "learning_rate": 0.0001607549386795917, "loss": 0.0056, "step": 2672 }, { "epoch": 0.2, "learning_rate": 0.00016072556363369317, "loss": 0.0059, "step": 2674 }, { "epoch": 0.2, "learning_rate": 0.00016069618858779468, "loss": 0.0056, "step": 2676 }, { "epoch": 0.2, "learning_rate": 0.00016066681354189616, "loss": 0.0055, "step": 2678 }, { "epoch": 0.2, "learning_rate": 0.00016063743849599766, "loss": 0.0056, "step": 2680 }, { "epoch": 0.2, "learning_rate": 0.00016060806345009914, "loss": 0.0058, "step": 2682 }, { "epoch": 0.2, "learning_rate": 0.00016057868840420065, "loss": 0.0057, "step": 2684 }, { "epoch": 0.2, "learning_rate": 0.00016054931335830213, "loss": 0.0055, "step": 2686 }, { "epoch": 0.2, "learning_rate": 0.0001605199383124036, "loss": 0.0054, "step": 2688 }, { "epoch": 0.2, "learning_rate": 0.0001604905632665051, "loss": 0.0055, "step": 2690 }, { "epoch": 0.2, "learning_rate": 0.00016046118822060662, "loss": 0.0056, "step": 2692 }, { "epoch": 0.2, "learning_rate": 0.0001604318131747081, "loss": 0.0058, "step": 2694 }, { "epoch": 0.2, "learning_rate": 0.00016040243812880957, "loss": 0.0057, "step": 2696 }, { "epoch": 0.2, "learning_rate": 0.00016037306308291108, "loss": 0.0055, "step": 2698 }, { "epoch": 0.2, "learning_rate": 0.00016034368803701256, "loss": 0.0058, "step": 2700 }, { "epoch": 0.2, "learning_rate": 0.00016031431299111406, "loss": 0.0054, "step": 2702 }, { "epoch": 0.2, "learning_rate": 0.00016028493794521554, "loss": 0.0057, "step": 2704 }, { "epoch": 0.2, "learning_rate": 0.00016025556289931705, "loss": 0.0056, "step": 2706 }, { "epoch": 0.2, "learning_rate": 0.00016022618785341852, "loss": 0.0055, "step": 2708 }, { "epoch": 0.2, "learning_rate": 0.00016019681280752, "loss": 0.0059, "step": 2710 }, { "epoch": 0.2, "learning_rate": 0.0001601674377616215, "loss": 0.0056, "step": 2712 }, { "epoch": 0.2, "learning_rate": 0.000160138062715723, "loss": 0.0058, "step": 2714 }, { "epoch": 0.2, "learning_rate": 0.0001601086876698245, "loss": 0.0055, "step": 2716 }, { "epoch": 0.2, "learning_rate": 0.000160079312623926, "loss": 0.0054, "step": 2718 }, { "epoch": 0.2, "learning_rate": 0.00016004993757802748, "loss": 0.0057, "step": 2720 }, { "epoch": 0.2, "learning_rate": 0.00016002056253212895, "loss": 0.0058, "step": 2722 }, { "epoch": 0.2, "learning_rate": 0.00015999118748623046, "loss": 0.0055, "step": 2724 }, { "epoch": 0.2, "learning_rate": 0.00015996181244033194, "loss": 0.0055, "step": 2726 }, { "epoch": 0.2, "learning_rate": 0.00015993243739443344, "loss": 0.0057, "step": 2728 }, { "epoch": 0.2, "learning_rate": 0.00015990306234853495, "loss": 0.0055, "step": 2730 }, { "epoch": 0.2, "learning_rate": 0.0001598736873026364, "loss": 0.006, "step": 2732 }, { "epoch": 0.2, "learning_rate": 0.0001598443122567379, "loss": 0.0055, "step": 2734 }, { "epoch": 0.2, "learning_rate": 0.0001598149372108394, "loss": 0.0055, "step": 2736 }, { "epoch": 0.2, "learning_rate": 0.0001597855621649409, "loss": 0.0056, "step": 2738 }, { "epoch": 0.2, "learning_rate": 0.0001597561871190424, "loss": 0.0056, "step": 2740 }, { "epoch": 0.2, "learning_rate": 0.00015972681207314387, "loss": 0.0056, "step": 2742 }, { "epoch": 0.2, "learning_rate": 0.00015969743702724535, "loss": 0.0053, "step": 2744 }, { "epoch": 0.2, "learning_rate": 0.00015966806198134686, "loss": 0.006, "step": 2746 }, { "epoch": 0.2, "learning_rate": 0.00015963868693544834, "loss": 0.006, "step": 2748 }, { "epoch": 0.2, "learning_rate": 0.00015960931188954984, "loss": 0.0058, "step": 2750 }, { "epoch": 0.2, "learning_rate": 0.00015957993684365135, "loss": 0.0058, "step": 2752 }, { "epoch": 0.2, "learning_rate": 0.0001595505617977528, "loss": 0.0055, "step": 2754 }, { "epoch": 0.2, "learning_rate": 0.0001595211867518543, "loss": 0.0054, "step": 2756 }, { "epoch": 0.2, "learning_rate": 0.0001594918117059558, "loss": 0.0057, "step": 2758 }, { "epoch": 0.2, "learning_rate": 0.0001594624366600573, "loss": 0.0056, "step": 2760 }, { "epoch": 0.2, "learning_rate": 0.0001594330616141588, "loss": 0.0055, "step": 2762 }, { "epoch": 0.2, "learning_rate": 0.00015940368656826027, "loss": 0.0058, "step": 2764 }, { "epoch": 0.2, "learning_rate": 0.00015937431152236175, "loss": 0.0058, "step": 2766 }, { "epoch": 0.2, "learning_rate": 0.00015934493647646325, "loss": 0.006, "step": 2768 }, { "epoch": 0.2, "learning_rate": 0.00015931556143056473, "loss": 0.0058, "step": 2770 }, { "epoch": 0.2, "learning_rate": 0.00015928618638466624, "loss": 0.0057, "step": 2772 }, { "epoch": 0.2, "learning_rate": 0.00015925681133876774, "loss": 0.0053, "step": 2774 }, { "epoch": 0.2, "learning_rate": 0.00015922743629286922, "loss": 0.0056, "step": 2776 }, { "epoch": 0.2, "learning_rate": 0.0001591980612469707, "loss": 0.0054, "step": 2778 }, { "epoch": 0.2, "learning_rate": 0.0001591686862010722, "loss": 0.0056, "step": 2780 }, { "epoch": 0.2, "learning_rate": 0.00015913931115517368, "loss": 0.0061, "step": 2782 }, { "epoch": 0.2, "learning_rate": 0.0001591099361092752, "loss": 0.0055, "step": 2784 }, { "epoch": 0.2, "learning_rate": 0.00015908056106337667, "loss": 0.0057, "step": 2786 }, { "epoch": 0.2, "learning_rate": 0.00015905118601747817, "loss": 0.0057, "step": 2788 }, { "epoch": 0.2, "learning_rate": 0.00015902181097157965, "loss": 0.0056, "step": 2790 }, { "epoch": 0.21, "learning_rate": 0.00015899243592568113, "loss": 0.0055, "step": 2792 }, { "epoch": 0.21, "learning_rate": 0.00015896306087978264, "loss": 0.0056, "step": 2794 }, { "epoch": 0.21, "learning_rate": 0.00015893368583388411, "loss": 0.0058, "step": 2796 }, { "epoch": 0.21, "learning_rate": 0.00015890431078798562, "loss": 0.0057, "step": 2798 }, { "epoch": 0.21, "learning_rate": 0.0001588749357420871, "loss": 0.0056, "step": 2800 }, { "epoch": 0.21, "learning_rate": 0.00015884556069618858, "loss": 0.0057, "step": 2802 }, { "epoch": 0.21, "learning_rate": 0.00015881618565029008, "loss": 0.0056, "step": 2804 }, { "epoch": 0.21, "learning_rate": 0.0001587868106043916, "loss": 0.0057, "step": 2806 }, { "epoch": 0.21, "learning_rate": 0.00015875743555849307, "loss": 0.0056, "step": 2808 }, { "epoch": 0.21, "learning_rate": 0.00015872806051259457, "loss": 0.0053, "step": 2810 }, { "epoch": 0.21, "learning_rate": 0.00015869868546669605, "loss": 0.0057, "step": 2812 }, { "epoch": 0.21, "learning_rate": 0.00015866931042079753, "loss": 0.0056, "step": 2814 }, { "epoch": 0.21, "learning_rate": 0.00015863993537489903, "loss": 0.0057, "step": 2816 }, { "epoch": 0.21, "learning_rate": 0.0001586105603290005, "loss": 0.0057, "step": 2818 }, { "epoch": 0.21, "learning_rate": 0.00015858118528310202, "loss": 0.0055, "step": 2820 }, { "epoch": 0.21, "learning_rate": 0.00015855181023720352, "loss": 0.0055, "step": 2822 }, { "epoch": 0.21, "learning_rate": 0.00015852243519130497, "loss": 0.0056, "step": 2824 }, { "epoch": 0.21, "learning_rate": 0.00015849306014540648, "loss": 0.0057, "step": 2826 }, { "epoch": 0.21, "learning_rate": 0.00015846368509950798, "loss": 0.0054, "step": 2828 }, { "epoch": 0.21, "learning_rate": 0.00015843431005360946, "loss": 0.0056, "step": 2830 }, { "epoch": 0.21, "learning_rate": 0.00015840493500771097, "loss": 0.0054, "step": 2832 }, { "epoch": 0.21, "learning_rate": 0.00015837555996181245, "loss": 0.0056, "step": 2834 }, { "epoch": 0.21, "learning_rate": 0.00015834618491591393, "loss": 0.0055, "step": 2836 }, { "epoch": 0.21, "learning_rate": 0.00015831680987001543, "loss": 0.0056, "step": 2838 }, { "epoch": 0.21, "learning_rate": 0.0001582874348241169, "loss": 0.0056, "step": 2840 }, { "epoch": 0.21, "learning_rate": 0.00015825805977821841, "loss": 0.0055, "step": 2842 }, { "epoch": 0.21, "learning_rate": 0.00015822868473231992, "loss": 0.0058, "step": 2844 }, { "epoch": 0.21, "learning_rate": 0.00015819930968642137, "loss": 0.0057, "step": 2846 }, { "epoch": 0.21, "learning_rate": 0.00015816993464052288, "loss": 0.0057, "step": 2848 }, { "epoch": 0.21, "learning_rate": 0.00015814055959462438, "loss": 0.0055, "step": 2850 }, { "epoch": 0.21, "learning_rate": 0.00015811118454872586, "loss": 0.0057, "step": 2852 }, { "epoch": 0.21, "learning_rate": 0.00015808180950282737, "loss": 0.0059, "step": 2854 }, { "epoch": 0.21, "learning_rate": 0.00015805243445692884, "loss": 0.0058, "step": 2856 }, { "epoch": 0.21, "learning_rate": 0.00015802305941103032, "loss": 0.0056, "step": 2858 }, { "epoch": 0.21, "learning_rate": 0.00015799368436513183, "loss": 0.0055, "step": 2860 }, { "epoch": 0.21, "learning_rate": 0.0001579643093192333, "loss": 0.0057, "step": 2862 }, { "epoch": 0.21, "learning_rate": 0.0001579349342733348, "loss": 0.0057, "step": 2864 }, { "epoch": 0.21, "learning_rate": 0.00015790555922743632, "loss": 0.0056, "step": 2866 }, { "epoch": 0.21, "learning_rate": 0.0001578761841815378, "loss": 0.0058, "step": 2868 }, { "epoch": 0.21, "learning_rate": 0.00015784680913563927, "loss": 0.0056, "step": 2870 }, { "epoch": 0.21, "learning_rate": 0.00015781743408974078, "loss": 0.0056, "step": 2872 }, { "epoch": 0.21, "learning_rate": 0.00015778805904384226, "loss": 0.0057, "step": 2874 }, { "epoch": 0.21, "learning_rate": 0.00015775868399794376, "loss": 0.0055, "step": 2876 }, { "epoch": 0.21, "learning_rate": 0.00015772930895204524, "loss": 0.0058, "step": 2878 }, { "epoch": 0.21, "learning_rate": 0.00015769993390614675, "loss": 0.0057, "step": 2880 }, { "epoch": 0.21, "learning_rate": 0.00015767055886024823, "loss": 0.0056, "step": 2882 }, { "epoch": 0.21, "learning_rate": 0.0001576411838143497, "loss": 0.0057, "step": 2884 }, { "epoch": 0.21, "learning_rate": 0.0001576118087684512, "loss": 0.0055, "step": 2886 }, { "epoch": 0.21, "learning_rate": 0.00015758243372255272, "loss": 0.0054, "step": 2888 }, { "epoch": 0.21, "learning_rate": 0.0001575530586766542, "loss": 0.0056, "step": 2890 }, { "epoch": 0.21, "learning_rate": 0.00015752368363075567, "loss": 0.0055, "step": 2892 }, { "epoch": 0.21, "learning_rate": 0.00015749430858485718, "loss": 0.0055, "step": 2894 }, { "epoch": 0.21, "learning_rate": 0.00015746493353895866, "loss": 0.0057, "step": 2896 }, { "epoch": 0.21, "learning_rate": 0.00015743555849306016, "loss": 0.0057, "step": 2898 }, { "epoch": 0.21, "learning_rate": 0.00015740618344716164, "loss": 0.0056, "step": 2900 }, { "epoch": 0.21, "learning_rate": 0.00015737680840126315, "loss": 0.0058, "step": 2902 }, { "epoch": 0.21, "learning_rate": 0.00015734743335536462, "loss": 0.0059, "step": 2904 }, { "epoch": 0.21, "learning_rate": 0.0001573180583094661, "loss": 0.0057, "step": 2906 }, { "epoch": 0.21, "learning_rate": 0.0001572886832635676, "loss": 0.0056, "step": 2908 }, { "epoch": 0.21, "learning_rate": 0.00015725930821766909, "loss": 0.0057, "step": 2910 }, { "epoch": 0.21, "learning_rate": 0.0001572299331717706, "loss": 0.0055, "step": 2912 }, { "epoch": 0.21, "learning_rate": 0.0001572005581258721, "loss": 0.0058, "step": 2914 }, { "epoch": 0.21, "learning_rate": 0.00015717118307997355, "loss": 0.0058, "step": 2916 }, { "epoch": 0.21, "learning_rate": 0.00015714180803407505, "loss": 0.0055, "step": 2918 }, { "epoch": 0.21, "learning_rate": 0.00015711243298817656, "loss": 0.0055, "step": 2920 }, { "epoch": 0.21, "learning_rate": 0.00015708305794227804, "loss": 0.006, "step": 2922 }, { "epoch": 0.21, "learning_rate": 0.00015705368289637954, "loss": 0.0057, "step": 2924 }, { "epoch": 0.21, "learning_rate": 0.00015702430785048102, "loss": 0.0055, "step": 2926 }, { "epoch": 0.22, "learning_rate": 0.0001569949328045825, "loss": 0.0056, "step": 2928 }, { "epoch": 0.22, "learning_rate": 0.000156965557758684, "loss": 0.0057, "step": 2930 }, { "epoch": 0.22, "learning_rate": 0.00015693618271278548, "loss": 0.0056, "step": 2932 }, { "epoch": 0.22, "learning_rate": 0.000156906807666887, "loss": 0.0059, "step": 2934 }, { "epoch": 0.22, "learning_rate": 0.0001568774326209885, "loss": 0.0054, "step": 2936 }, { "epoch": 0.22, "learning_rate": 0.00015684805757508997, "loss": 0.0056, "step": 2938 }, { "epoch": 0.22, "learning_rate": 0.00015681868252919145, "loss": 0.0054, "step": 2940 }, { "epoch": 0.22, "learning_rate": 0.00015678930748329296, "loss": 0.0056, "step": 2942 }, { "epoch": 0.22, "learning_rate": 0.00015675993243739443, "loss": 0.0057, "step": 2944 }, { "epoch": 0.22, "learning_rate": 0.00015673055739149594, "loss": 0.0056, "step": 2946 }, { "epoch": 0.22, "learning_rate": 0.00015670118234559742, "loss": 0.0056, "step": 2948 }, { "epoch": 0.22, "learning_rate": 0.0001566718072996989, "loss": 0.0057, "step": 2950 }, { "epoch": 0.22, "learning_rate": 0.0001566424322538004, "loss": 0.0054, "step": 2952 }, { "epoch": 0.22, "learning_rate": 0.00015661305720790188, "loss": 0.0056, "step": 2954 }, { "epoch": 0.22, "learning_rate": 0.0001565836821620034, "loss": 0.0055, "step": 2956 }, { "epoch": 0.22, "learning_rate": 0.0001565543071161049, "loss": 0.0058, "step": 2958 }, { "epoch": 0.22, "learning_rate": 0.00015652493207020637, "loss": 0.0056, "step": 2960 }, { "epoch": 0.22, "learning_rate": 0.00015649555702430785, "loss": 0.0056, "step": 2962 }, { "epoch": 0.22, "learning_rate": 0.00015646618197840935, "loss": 0.0057, "step": 2964 }, { "epoch": 0.22, "learning_rate": 0.00015643680693251083, "loss": 0.0056, "step": 2966 }, { "epoch": 0.22, "learning_rate": 0.00015640743188661234, "loss": 0.0056, "step": 2968 }, { "epoch": 0.22, "learning_rate": 0.00015637805684071382, "loss": 0.0056, "step": 2970 }, { "epoch": 0.22, "learning_rate": 0.00015634868179481532, "loss": 0.0057, "step": 2972 }, { "epoch": 0.22, "learning_rate": 0.0001563193067489168, "loss": 0.0058, "step": 2974 }, { "epoch": 0.22, "learning_rate": 0.00015628993170301828, "loss": 0.0056, "step": 2976 }, { "epoch": 0.22, "learning_rate": 0.00015626055665711978, "loss": 0.0059, "step": 2978 }, { "epoch": 0.22, "learning_rate": 0.0001562311816112213, "loss": 0.0055, "step": 2980 }, { "epoch": 0.22, "learning_rate": 0.00015620180656532277, "loss": 0.0056, "step": 2982 }, { "epoch": 0.22, "learning_rate": 0.00015617243151942427, "loss": 0.0056, "step": 2984 }, { "epoch": 0.22, "learning_rate": 0.00015614305647352575, "loss": 0.006, "step": 2986 }, { "epoch": 0.22, "learning_rate": 0.00015611368142762723, "loss": 0.0058, "step": 2988 }, { "epoch": 0.22, "learning_rate": 0.00015608430638172874, "loss": 0.0056, "step": 2990 }, { "epoch": 0.22, "learning_rate": 0.00015605493133583021, "loss": 0.0058, "step": 2992 }, { "epoch": 0.22, "learning_rate": 0.00015602555628993172, "loss": 0.0055, "step": 2994 }, { "epoch": 0.22, "learning_rate": 0.0001559961812440332, "loss": 0.0054, "step": 2996 }, { "epoch": 0.22, "learning_rate": 0.00015596680619813468, "loss": 0.0055, "step": 2998 }, { "epoch": 0.22, "learning_rate": 0.00015593743115223618, "loss": 0.0054, "step": 3000 }, { "epoch": 0.22, "learning_rate": 0.0001559080561063377, "loss": 0.0056, "step": 3002 }, { "epoch": 0.22, "learning_rate": 0.00015587868106043917, "loss": 0.0054, "step": 3004 }, { "epoch": 0.22, "learning_rate": 0.00015584930601454067, "loss": 0.0056, "step": 3006 }, { "epoch": 0.22, "learning_rate": 0.00015581993096864215, "loss": 0.0055, "step": 3008 }, { "epoch": 0.22, "learning_rate": 0.00015579055592274363, "loss": 0.0056, "step": 3010 }, { "epoch": 0.22, "learning_rate": 0.00015576118087684513, "loss": 0.0055, "step": 3012 }, { "epoch": 0.22, "learning_rate": 0.0001557318058309466, "loss": 0.0057, "step": 3014 }, { "epoch": 0.22, "learning_rate": 0.00015570243078504812, "loss": 0.0057, "step": 3016 }, { "epoch": 0.22, "learning_rate": 0.00015567305573914962, "loss": 0.0055, "step": 3018 }, { "epoch": 0.22, "learning_rate": 0.00015564368069325107, "loss": 0.0057, "step": 3020 }, { "epoch": 0.22, "learning_rate": 0.00015561430564735258, "loss": 0.0055, "step": 3022 }, { "epoch": 0.22, "learning_rate": 0.00015558493060145406, "loss": 0.0059, "step": 3024 }, { "epoch": 0.22, "learning_rate": 0.00015555555555555556, "loss": 0.0056, "step": 3026 }, { "epoch": 0.22, "learning_rate": 0.00015552618050965707, "loss": 0.0055, "step": 3028 }, { "epoch": 0.22, "learning_rate": 0.00015549680546375855, "loss": 0.0055, "step": 3030 }, { "epoch": 0.22, "learning_rate": 0.00015546743041786003, "loss": 0.0059, "step": 3032 }, { "epoch": 0.22, "learning_rate": 0.00015543805537196153, "loss": 0.0056, "step": 3034 }, { "epoch": 0.22, "learning_rate": 0.000155408680326063, "loss": 0.0057, "step": 3036 }, { "epoch": 0.22, "learning_rate": 0.00015537930528016451, "loss": 0.0057, "step": 3038 }, { "epoch": 0.22, "learning_rate": 0.000155349930234266, "loss": 0.0059, "step": 3040 }, { "epoch": 0.22, "learning_rate": 0.0001553205551883675, "loss": 0.0056, "step": 3042 }, { "epoch": 0.22, "learning_rate": 0.00015529118014246898, "loss": 0.0058, "step": 3044 }, { "epoch": 0.22, "learning_rate": 0.00015526180509657046, "loss": 0.0057, "step": 3046 }, { "epoch": 0.22, "learning_rate": 0.00015523243005067196, "loss": 0.0056, "step": 3048 }, { "epoch": 0.22, "learning_rate": 0.00015520305500477347, "loss": 0.0056, "step": 3050 }, { "epoch": 0.22, "learning_rate": 0.00015517367995887494, "loss": 0.0057, "step": 3052 }, { "epoch": 0.22, "learning_rate": 0.00015514430491297642, "loss": 0.0056, "step": 3054 }, { "epoch": 0.22, "learning_rate": 0.00015511492986707793, "loss": 0.0056, "step": 3056 }, { "epoch": 0.22, "learning_rate": 0.0001550855548211794, "loss": 0.0058, "step": 3058 }, { "epoch": 0.22, "learning_rate": 0.0001550561797752809, "loss": 0.006, "step": 3060 }, { "epoch": 0.22, "learning_rate": 0.0001550268047293824, "loss": 0.0058, "step": 3062 }, { "epoch": 0.23, "learning_rate": 0.0001549974296834839, "loss": 0.0055, "step": 3064 }, { "epoch": 0.23, "learning_rate": 0.00015496805463758537, "loss": 0.0057, "step": 3066 }, { "epoch": 0.23, "learning_rate": 0.00015493867959168685, "loss": 0.0055, "step": 3068 }, { "epoch": 0.23, "learning_rate": 0.00015490930454578836, "loss": 0.0054, "step": 3070 }, { "epoch": 0.23, "learning_rate": 0.00015487992949988986, "loss": 0.0054, "step": 3072 }, { "epoch": 0.23, "learning_rate": 0.00015485055445399134, "loss": 0.0056, "step": 3074 }, { "epoch": 0.23, "learning_rate": 0.00015482117940809285, "loss": 0.0055, "step": 3076 }, { "epoch": 0.23, "learning_rate": 0.00015479180436219433, "loss": 0.0055, "step": 3078 }, { "epoch": 0.23, "learning_rate": 0.0001547624293162958, "loss": 0.0059, "step": 3080 }, { "epoch": 0.23, "learning_rate": 0.0001547330542703973, "loss": 0.0057, "step": 3082 }, { "epoch": 0.23, "learning_rate": 0.0001547036792244988, "loss": 0.0057, "step": 3084 }, { "epoch": 0.23, "learning_rate": 0.0001546743041786003, "loss": 0.0058, "step": 3086 }, { "epoch": 0.23, "learning_rate": 0.0001546449291327018, "loss": 0.0055, "step": 3088 }, { "epoch": 0.23, "learning_rate": 0.00015461555408680325, "loss": 0.0056, "step": 3090 }, { "epoch": 0.23, "learning_rate": 0.00015458617904090476, "loss": 0.0055, "step": 3092 }, { "epoch": 0.23, "learning_rate": 0.00015455680399500626, "loss": 0.0059, "step": 3094 }, { "epoch": 0.23, "learning_rate": 0.00015452742894910774, "loss": 0.0057, "step": 3096 }, { "epoch": 0.23, "learning_rate": 0.00015449805390320925, "loss": 0.0057, "step": 3098 }, { "epoch": 0.23, "learning_rate": 0.00015446867885731072, "loss": 0.0058, "step": 3100 }, { "epoch": 0.23, "learning_rate": 0.0001544393038114122, "loss": 0.0056, "step": 3102 }, { "epoch": 0.23, "learning_rate": 0.0001544099287655137, "loss": 0.0058, "step": 3104 }, { "epoch": 0.23, "learning_rate": 0.00015438055371961519, "loss": 0.0055, "step": 3106 }, { "epoch": 0.23, "learning_rate": 0.0001543511786737167, "loss": 0.0054, "step": 3108 }, { "epoch": 0.23, "learning_rate": 0.0001543218036278182, "loss": 0.0056, "step": 3110 }, { "epoch": 0.23, "learning_rate": 0.00015429242858191965, "loss": 0.0058, "step": 3112 }, { "epoch": 0.23, "learning_rate": 0.00015426305353602115, "loss": 0.0055, "step": 3114 }, { "epoch": 0.23, "learning_rate": 0.00015423367849012266, "loss": 0.0056, "step": 3116 }, { "epoch": 0.23, "learning_rate": 0.00015420430344422414, "loss": 0.0056, "step": 3118 }, { "epoch": 0.23, "learning_rate": 0.00015417492839832564, "loss": 0.0057, "step": 3120 }, { "epoch": 0.23, "learning_rate": 0.00015414555335242712, "loss": 0.0054, "step": 3122 }, { "epoch": 0.23, "learning_rate": 0.0001541161783065286, "loss": 0.0054, "step": 3124 }, { "epoch": 0.23, "learning_rate": 0.0001540868032606301, "loss": 0.0057, "step": 3126 }, { "epoch": 0.23, "learning_rate": 0.00015405742821473158, "loss": 0.0056, "step": 3128 }, { "epoch": 0.23, "learning_rate": 0.0001540280531688331, "loss": 0.0057, "step": 3130 }, { "epoch": 0.23, "learning_rate": 0.0001539986781229346, "loss": 0.0055, "step": 3132 }, { "epoch": 0.23, "learning_rate": 0.00015396930307703607, "loss": 0.0056, "step": 3134 }, { "epoch": 0.23, "learning_rate": 0.00015393992803113755, "loss": 0.0056, "step": 3136 }, { "epoch": 0.23, "learning_rate": 0.00015391055298523903, "loss": 0.0058, "step": 3138 }, { "epoch": 0.23, "learning_rate": 0.00015388117793934053, "loss": 0.0058, "step": 3140 }, { "epoch": 0.23, "learning_rate": 0.00015385180289344204, "loss": 0.0055, "step": 3142 }, { "epoch": 0.23, "learning_rate": 0.00015382242784754352, "loss": 0.0055, "step": 3144 }, { "epoch": 0.23, "learning_rate": 0.000153793052801645, "loss": 0.0057, "step": 3146 }, { "epoch": 0.23, "learning_rate": 0.0001537636777557465, "loss": 0.0056, "step": 3148 }, { "epoch": 0.23, "learning_rate": 0.00015373430270984798, "loss": 0.0054, "step": 3150 }, { "epoch": 0.23, "learning_rate": 0.00015370492766394949, "loss": 0.0056, "step": 3152 }, { "epoch": 0.23, "learning_rate": 0.00015367555261805096, "loss": 0.0056, "step": 3154 }, { "epoch": 0.23, "learning_rate": 0.00015364617757215247, "loss": 0.0055, "step": 3156 }, { "epoch": 0.23, "learning_rate": 0.00015361680252625395, "loss": 0.0059, "step": 3158 }, { "epoch": 0.23, "learning_rate": 0.00015358742748035543, "loss": 0.0059, "step": 3160 }, { "epoch": 0.23, "learning_rate": 0.00015355805243445693, "loss": 0.0055, "step": 3162 }, { "epoch": 0.23, "learning_rate": 0.00015352867738855844, "loss": 0.0056, "step": 3164 }, { "epoch": 0.23, "learning_rate": 0.00015349930234265992, "loss": 0.0059, "step": 3166 }, { "epoch": 0.23, "learning_rate": 0.00015346992729676142, "loss": 0.0058, "step": 3168 }, { "epoch": 0.23, "learning_rate": 0.0001534405522508629, "loss": 0.0057, "step": 3170 }, { "epoch": 0.23, "learning_rate": 0.00015341117720496438, "loss": 0.0054, "step": 3172 }, { "epoch": 0.23, "learning_rate": 0.00015338180215906588, "loss": 0.0059, "step": 3174 }, { "epoch": 0.23, "learning_rate": 0.00015335242711316736, "loss": 0.0058, "step": 3176 }, { "epoch": 0.23, "learning_rate": 0.00015332305206726887, "loss": 0.0054, "step": 3178 }, { "epoch": 0.23, "learning_rate": 0.00015329367702137037, "loss": 0.0054, "step": 3180 }, { "epoch": 0.23, "learning_rate": 0.00015326430197547182, "loss": 0.0056, "step": 3182 }, { "epoch": 0.23, "learning_rate": 0.00015323492692957333, "loss": 0.0056, "step": 3184 }, { "epoch": 0.23, "learning_rate": 0.00015320555188367484, "loss": 0.0056, "step": 3186 }, { "epoch": 0.23, "learning_rate": 0.00015317617683777631, "loss": 0.0057, "step": 3188 }, { "epoch": 0.23, "learning_rate": 0.00015314680179187782, "loss": 0.0056, "step": 3190 }, { "epoch": 0.23, "learning_rate": 0.0001531174267459793, "loss": 0.0057, "step": 3192 }, { "epoch": 0.23, "learning_rate": 0.00015308805170008078, "loss": 0.0055, "step": 3194 }, { "epoch": 0.23, "learning_rate": 0.00015305867665418228, "loss": 0.0053, "step": 3196 }, { "epoch": 0.23, "learning_rate": 0.00015302930160828376, "loss": 0.0057, "step": 3198 }, { "epoch": 0.23, "learning_rate": 0.00015299992656238527, "loss": 0.0061, "step": 3200 }, { "epoch": 0.24, "learning_rate": 0.00015297055151648677, "loss": 0.0054, "step": 3202 }, { "epoch": 0.24, "learning_rate": 0.00015294117647058822, "loss": 0.0057, "step": 3204 }, { "epoch": 0.24, "learning_rate": 0.00015291180142468973, "loss": 0.0056, "step": 3206 }, { "epoch": 0.24, "learning_rate": 0.00015288242637879123, "loss": 0.0057, "step": 3208 }, { "epoch": 0.24, "learning_rate": 0.0001528530513328927, "loss": 0.0059, "step": 3210 }, { "epoch": 0.24, "learning_rate": 0.00015282367628699422, "loss": 0.0055, "step": 3212 }, { "epoch": 0.24, "learning_rate": 0.0001527943012410957, "loss": 0.0057, "step": 3214 }, { "epoch": 0.24, "learning_rate": 0.00015276492619519717, "loss": 0.0057, "step": 3216 }, { "epoch": 0.24, "learning_rate": 0.00015273555114929868, "loss": 0.0056, "step": 3218 }, { "epoch": 0.24, "learning_rate": 0.00015270617610340016, "loss": 0.0056, "step": 3220 }, { "epoch": 0.24, "learning_rate": 0.00015267680105750166, "loss": 0.0055, "step": 3222 }, { "epoch": 0.24, "learning_rate": 0.00015264742601160317, "loss": 0.0056, "step": 3224 }, { "epoch": 0.24, "learning_rate": 0.00015261805096570465, "loss": 0.0058, "step": 3226 }, { "epoch": 0.24, "learning_rate": 0.00015258867591980612, "loss": 0.0056, "step": 3228 }, { "epoch": 0.24, "learning_rate": 0.00015255930087390763, "loss": 0.0057, "step": 3230 }, { "epoch": 0.24, "learning_rate": 0.0001525299258280091, "loss": 0.0055, "step": 3232 }, { "epoch": 0.24, "learning_rate": 0.00015250055078211061, "loss": 0.0057, "step": 3234 }, { "epoch": 0.24, "learning_rate": 0.0001524711757362121, "loss": 0.0056, "step": 3236 }, { "epoch": 0.24, "learning_rate": 0.0001524418006903136, "loss": 0.0057, "step": 3238 }, { "epoch": 0.24, "learning_rate": 0.00015241242564441508, "loss": 0.0057, "step": 3240 }, { "epoch": 0.24, "learning_rate": 0.00015238305059851655, "loss": 0.0057, "step": 3242 }, { "epoch": 0.24, "learning_rate": 0.00015235367555261806, "loss": 0.0055, "step": 3244 }, { "epoch": 0.24, "learning_rate": 0.00015232430050671957, "loss": 0.0054, "step": 3246 }, { "epoch": 0.24, "learning_rate": 0.00015229492546082104, "loss": 0.0058, "step": 3248 }, { "epoch": 0.24, "learning_rate": 0.00015226555041492252, "loss": 0.0057, "step": 3250 }, { "epoch": 0.24, "learning_rate": 0.00015223617536902403, "loss": 0.0053, "step": 3252 }, { "epoch": 0.24, "learning_rate": 0.0001522068003231255, "loss": 0.0059, "step": 3254 }, { "epoch": 0.24, "learning_rate": 0.000152177425277227, "loss": 0.0056, "step": 3256 }, { "epoch": 0.24, "learning_rate": 0.0001521480502313285, "loss": 0.0056, "step": 3258 }, { "epoch": 0.24, "learning_rate": 0.00015211867518543, "loss": 0.0056, "step": 3260 }, { "epoch": 0.24, "learning_rate": 0.00015208930013953147, "loss": 0.0057, "step": 3262 }, { "epoch": 0.24, "learning_rate": 0.00015205992509363295, "loss": 0.0058, "step": 3264 }, { "epoch": 0.24, "learning_rate": 0.00015203055004773446, "loss": 0.0058, "step": 3266 }, { "epoch": 0.24, "learning_rate": 0.00015200117500183594, "loss": 0.0057, "step": 3268 }, { "epoch": 0.24, "learning_rate": 0.00015197179995593744, "loss": 0.0057, "step": 3270 }, { "epoch": 0.24, "learning_rate": 0.00015194242491003895, "loss": 0.0055, "step": 3272 }, { "epoch": 0.24, "learning_rate": 0.0001519130498641404, "loss": 0.0057, "step": 3274 }, { "epoch": 0.24, "learning_rate": 0.0001518836748182419, "loss": 0.0054, "step": 3276 }, { "epoch": 0.24, "learning_rate": 0.0001518542997723434, "loss": 0.0057, "step": 3278 }, { "epoch": 0.24, "learning_rate": 0.0001518249247264449, "loss": 0.0055, "step": 3280 }, { "epoch": 0.24, "learning_rate": 0.0001517955496805464, "loss": 0.0057, "step": 3282 }, { "epoch": 0.24, "learning_rate": 0.00015176617463464787, "loss": 0.0057, "step": 3284 }, { "epoch": 0.24, "learning_rate": 0.00015173679958874935, "loss": 0.0055, "step": 3286 }, { "epoch": 0.24, "learning_rate": 0.00015170742454285086, "loss": 0.0058, "step": 3288 }, { "epoch": 0.24, "learning_rate": 0.00015167804949695233, "loss": 0.0058, "step": 3290 }, { "epoch": 0.24, "learning_rate": 0.00015164867445105384, "loss": 0.0055, "step": 3292 }, { "epoch": 0.24, "learning_rate": 0.00015161929940515534, "loss": 0.0055, "step": 3294 }, { "epoch": 0.24, "learning_rate": 0.00015158992435925682, "loss": 0.0055, "step": 3296 }, { "epoch": 0.24, "learning_rate": 0.0001515605493133583, "loss": 0.0057, "step": 3298 }, { "epoch": 0.24, "learning_rate": 0.0001515311742674598, "loss": 0.0056, "step": 3300 }, { "epoch": 0.24, "learning_rate": 0.00015150179922156129, "loss": 0.0056, "step": 3302 }, { "epoch": 0.24, "learning_rate": 0.0001514724241756628, "loss": 0.0055, "step": 3304 }, { "epoch": 0.24, "learning_rate": 0.00015144304912976427, "loss": 0.0061, "step": 3306 }, { "epoch": 0.24, "learning_rate": 0.00015141367408386575, "loss": 0.0057, "step": 3308 }, { "epoch": 0.24, "learning_rate": 0.00015138429903796725, "loss": 0.0055, "step": 3310 }, { "epoch": 0.24, "learning_rate": 0.00015135492399206873, "loss": 0.0056, "step": 3312 }, { "epoch": 0.24, "learning_rate": 0.00015132554894617024, "loss": 0.0057, "step": 3314 }, { "epoch": 0.24, "learning_rate": 0.00015129617390027174, "loss": 0.0057, "step": 3316 }, { "epoch": 0.24, "learning_rate": 0.00015126679885437322, "loss": 0.0056, "step": 3318 }, { "epoch": 0.24, "learning_rate": 0.0001512374238084747, "loss": 0.0053, "step": 3320 }, { "epoch": 0.24, "learning_rate": 0.0001512080487625762, "loss": 0.0053, "step": 3322 }, { "epoch": 0.24, "learning_rate": 0.00015117867371667768, "loss": 0.0055, "step": 3324 }, { "epoch": 0.24, "learning_rate": 0.0001511492986707792, "loss": 0.0056, "step": 3326 }, { "epoch": 0.24, "learning_rate": 0.00015111992362488067, "loss": 0.0055, "step": 3328 }, { "epoch": 0.24, "learning_rate": 0.00015109054857898217, "loss": 0.0056, "step": 3330 }, { "epoch": 0.24, "learning_rate": 0.00015106117353308365, "loss": 0.0056, "step": 3332 }, { "epoch": 0.24, "learning_rate": 0.00015103179848718513, "loss": 0.0056, "step": 3334 }, { "epoch": 0.24, "learning_rate": 0.00015100242344128663, "loss": 0.0056, "step": 3336 }, { "epoch": 0.25, "learning_rate": 0.00015097304839538814, "loss": 0.0055, "step": 3338 }, { "epoch": 0.25, "learning_rate": 0.00015094367334948962, "loss": 0.0057, "step": 3340 }, { "epoch": 0.25, "learning_rate": 0.00015091429830359112, "loss": 0.0054, "step": 3342 }, { "epoch": 0.25, "learning_rate": 0.0001508849232576926, "loss": 0.0054, "step": 3344 }, { "epoch": 0.25, "learning_rate": 0.00015085554821179408, "loss": 0.0055, "step": 3346 }, { "epoch": 0.25, "learning_rate": 0.00015082617316589559, "loss": 0.0057, "step": 3348 }, { "epoch": 0.25, "learning_rate": 0.00015079679811999706, "loss": 0.0056, "step": 3350 }, { "epoch": 0.25, "learning_rate": 0.00015076742307409857, "loss": 0.0058, "step": 3352 }, { "epoch": 0.25, "learning_rate": 0.00015073804802820005, "loss": 0.0057, "step": 3354 }, { "epoch": 0.25, "learning_rate": 0.00015070867298230153, "loss": 0.0057, "step": 3356 }, { "epoch": 0.25, "learning_rate": 0.00015067929793640303, "loss": 0.0057, "step": 3358 }, { "epoch": 0.25, "learning_rate": 0.00015064992289050454, "loss": 0.0058, "step": 3360 }, { "epoch": 0.25, "learning_rate": 0.00015062054784460602, "loss": 0.0057, "step": 3362 }, { "epoch": 0.25, "learning_rate": 0.00015059117279870752, "loss": 0.0056, "step": 3364 }, { "epoch": 0.25, "learning_rate": 0.000150561797752809, "loss": 0.0056, "step": 3366 }, { "epoch": 0.25, "learning_rate": 0.00015053242270691048, "loss": 0.0055, "step": 3368 }, { "epoch": 0.25, "learning_rate": 0.00015050304766101198, "loss": 0.0057, "step": 3370 }, { "epoch": 0.25, "learning_rate": 0.00015047367261511346, "loss": 0.0057, "step": 3372 }, { "epoch": 0.25, "learning_rate": 0.00015044429756921497, "loss": 0.0056, "step": 3374 }, { "epoch": 0.25, "learning_rate": 0.00015041492252331647, "loss": 0.0053, "step": 3376 }, { "epoch": 0.25, "learning_rate": 0.00015038554747741792, "loss": 0.0057, "step": 3378 }, { "epoch": 0.25, "learning_rate": 0.00015035617243151943, "loss": 0.0056, "step": 3380 }, { "epoch": 0.25, "learning_rate": 0.0001503267973856209, "loss": 0.0056, "step": 3382 }, { "epoch": 0.25, "learning_rate": 0.0001502974223397224, "loss": 0.0057, "step": 3384 }, { "epoch": 0.25, "learning_rate": 0.00015026804729382392, "loss": 0.0055, "step": 3386 }, { "epoch": 0.25, "learning_rate": 0.0001502386722479254, "loss": 0.0056, "step": 3388 }, { "epoch": 0.25, "learning_rate": 0.00015020929720202688, "loss": 0.0056, "step": 3390 }, { "epoch": 0.25, "learning_rate": 0.00015017992215612838, "loss": 0.0055, "step": 3392 }, { "epoch": 0.25, "learning_rate": 0.00015015054711022986, "loss": 0.0057, "step": 3394 }, { "epoch": 0.25, "learning_rate": 0.00015012117206433136, "loss": 0.0057, "step": 3396 }, { "epoch": 0.25, "learning_rate": 0.00015009179701843284, "loss": 0.0058, "step": 3398 }, { "epoch": 0.25, "learning_rate": 0.00015006242197253432, "loss": 0.0052, "step": 3400 }, { "epoch": 0.25, "learning_rate": 0.00015003304692663583, "loss": 0.0055, "step": 3402 }, { "epoch": 0.25, "learning_rate": 0.0001500036718807373, "loss": 0.0057, "step": 3404 }, { "epoch": 0.25, "learning_rate": 0.0001499742968348388, "loss": 0.0055, "step": 3406 }, { "epoch": 0.25, "learning_rate": 0.00014994492178894032, "loss": 0.0056, "step": 3408 }, { "epoch": 0.25, "learning_rate": 0.0001499155467430418, "loss": 0.0057, "step": 3410 }, { "epoch": 0.25, "learning_rate": 0.00014988617169714327, "loss": 0.0056, "step": 3412 }, { "epoch": 0.25, "learning_rate": 0.00014985679665124478, "loss": 0.0058, "step": 3414 }, { "epoch": 0.25, "learning_rate": 0.00014982742160534626, "loss": 0.0055, "step": 3416 }, { "epoch": 0.25, "learning_rate": 0.00014979804655944776, "loss": 0.0056, "step": 3418 }, { "epoch": 0.25, "learning_rate": 0.00014976867151354924, "loss": 0.0057, "step": 3420 }, { "epoch": 0.25, "learning_rate": 0.00014973929646765075, "loss": 0.0056, "step": 3422 }, { "epoch": 0.25, "learning_rate": 0.00014970992142175222, "loss": 0.0057, "step": 3424 }, { "epoch": 0.25, "learning_rate": 0.0001496805463758537, "loss": 0.0056, "step": 3426 }, { "epoch": 0.25, "learning_rate": 0.0001496511713299552, "loss": 0.0052, "step": 3428 }, { "epoch": 0.25, "learning_rate": 0.00014962179628405671, "loss": 0.0055, "step": 3430 }, { "epoch": 0.25, "learning_rate": 0.0001495924212381582, "loss": 0.0057, "step": 3432 }, { "epoch": 0.25, "learning_rate": 0.0001495630461922597, "loss": 0.0056, "step": 3434 }, { "epoch": 0.25, "learning_rate": 0.00014953367114636118, "loss": 0.0055, "step": 3436 }, { "epoch": 0.25, "learning_rate": 0.00014950429610046265, "loss": 0.0056, "step": 3438 }, { "epoch": 0.25, "learning_rate": 0.00014947492105456416, "loss": 0.0055, "step": 3440 }, { "epoch": 0.25, "learning_rate": 0.00014944554600866564, "loss": 0.0056, "step": 3442 }, { "epoch": 0.25, "learning_rate": 0.00014941617096276714, "loss": 0.0055, "step": 3444 }, { "epoch": 0.25, "learning_rate": 0.00014938679591686865, "loss": 0.0058, "step": 3446 }, { "epoch": 0.25, "learning_rate": 0.0001493574208709701, "loss": 0.0057, "step": 3448 }, { "epoch": 0.25, "learning_rate": 0.0001493280458250716, "loss": 0.0057, "step": 3450 }, { "epoch": 0.25, "learning_rate": 0.0001492986707791731, "loss": 0.0053, "step": 3452 }, { "epoch": 0.25, "learning_rate": 0.0001492692957332746, "loss": 0.0056, "step": 3454 }, { "epoch": 0.25, "learning_rate": 0.0001492399206873761, "loss": 0.0057, "step": 3456 }, { "epoch": 0.25, "learning_rate": 0.00014921054564147757, "loss": 0.0055, "step": 3458 }, { "epoch": 0.25, "learning_rate": 0.00014918117059557905, "loss": 0.0057, "step": 3460 }, { "epoch": 0.25, "learning_rate": 0.00014915179554968056, "loss": 0.0054, "step": 3462 }, { "epoch": 0.25, "learning_rate": 0.00014912242050378204, "loss": 0.0055, "step": 3464 }, { "epoch": 0.25, "learning_rate": 0.00014909304545788354, "loss": 0.0054, "step": 3466 }, { "epoch": 0.25, "learning_rate": 0.00014906367041198505, "loss": 0.0055, "step": 3468 }, { "epoch": 0.25, "learning_rate": 0.0001490342953660865, "loss": 0.0059, "step": 3470 }, { "epoch": 0.25, "learning_rate": 0.000149004920320188, "loss": 0.0055, "step": 3472 }, { "epoch": 0.26, "learning_rate": 0.0001489755452742895, "loss": 0.0057, "step": 3474 }, { "epoch": 0.26, "learning_rate": 0.000148946170228391, "loss": 0.0058, "step": 3476 }, { "epoch": 0.26, "learning_rate": 0.0001489167951824925, "loss": 0.0056, "step": 3478 }, { "epoch": 0.26, "learning_rate": 0.00014888742013659397, "loss": 0.0056, "step": 3480 }, { "epoch": 0.26, "learning_rate": 0.00014885804509069545, "loss": 0.0057, "step": 3482 }, { "epoch": 0.26, "learning_rate": 0.00014882867004479696, "loss": 0.0056, "step": 3484 }, { "epoch": 0.26, "learning_rate": 0.00014879929499889843, "loss": 0.0056, "step": 3486 }, { "epoch": 0.26, "learning_rate": 0.00014876991995299994, "loss": 0.0058, "step": 3488 }, { "epoch": 0.26, "learning_rate": 0.00014874054490710144, "loss": 0.0055, "step": 3490 }, { "epoch": 0.26, "learning_rate": 0.00014871116986120292, "loss": 0.0056, "step": 3492 }, { "epoch": 0.26, "learning_rate": 0.0001486817948153044, "loss": 0.0056, "step": 3494 }, { "epoch": 0.26, "learning_rate": 0.00014865241976940588, "loss": 0.0056, "step": 3496 }, { "epoch": 0.26, "learning_rate": 0.00014862304472350739, "loss": 0.0056, "step": 3498 }, { "epoch": 0.26, "learning_rate": 0.0001485936696776089, "loss": 0.0055, "step": 3500 }, { "epoch": 0.26, "learning_rate": 0.00014856429463171037, "loss": 0.0054, "step": 3502 }, { "epoch": 0.26, "learning_rate": 0.00014853491958581185, "loss": 0.0056, "step": 3504 }, { "epoch": 0.26, "learning_rate": 0.00014850554453991335, "loss": 0.0055, "step": 3506 }, { "epoch": 0.26, "learning_rate": 0.00014847616949401483, "loss": 0.0056, "step": 3508 }, { "epoch": 0.26, "learning_rate": 0.00014844679444811634, "loss": 0.0055, "step": 3510 }, { "epoch": 0.26, "learning_rate": 0.00014841741940221781, "loss": 0.0057, "step": 3512 }, { "epoch": 0.26, "learning_rate": 0.00014838804435631932, "loss": 0.0057, "step": 3514 }, { "epoch": 0.26, "learning_rate": 0.0001483586693104208, "loss": 0.0058, "step": 3516 }, { "epoch": 0.26, "learning_rate": 0.00014832929426452228, "loss": 0.0057, "step": 3518 }, { "epoch": 0.26, "learning_rate": 0.00014829991921862378, "loss": 0.0053, "step": 3520 }, { "epoch": 0.26, "learning_rate": 0.0001482705441727253, "loss": 0.0055, "step": 3522 }, { "epoch": 0.26, "learning_rate": 0.00014824116912682677, "loss": 0.0056, "step": 3524 }, { "epoch": 0.26, "learning_rate": 0.00014821179408092827, "loss": 0.0057, "step": 3526 }, { "epoch": 0.26, "learning_rate": 0.00014818241903502975, "loss": 0.0055, "step": 3528 }, { "epoch": 0.26, "learning_rate": 0.00014815304398913123, "loss": 0.0056, "step": 3530 }, { "epoch": 0.26, "learning_rate": 0.00014812366894323273, "loss": 0.0053, "step": 3532 }, { "epoch": 0.26, "learning_rate": 0.0001480942938973342, "loss": 0.0057, "step": 3534 }, { "epoch": 0.26, "learning_rate": 0.00014806491885143572, "loss": 0.0056, "step": 3536 }, { "epoch": 0.26, "learning_rate": 0.00014803554380553722, "loss": 0.0055, "step": 3538 }, { "epoch": 0.26, "learning_rate": 0.00014800616875963867, "loss": 0.0055, "step": 3540 }, { "epoch": 0.26, "learning_rate": 0.00014797679371374018, "loss": 0.0057, "step": 3542 }, { "epoch": 0.26, "learning_rate": 0.00014794741866784169, "loss": 0.0054, "step": 3544 }, { "epoch": 0.26, "learning_rate": 0.00014791804362194316, "loss": 0.0054, "step": 3546 }, { "epoch": 0.26, "learning_rate": 0.00014788866857604467, "loss": 0.0058, "step": 3548 }, { "epoch": 0.26, "learning_rate": 0.00014785929353014615, "loss": 0.0055, "step": 3550 }, { "epoch": 0.26, "learning_rate": 0.00014782991848424763, "loss": 0.0058, "step": 3552 }, { "epoch": 0.26, "learning_rate": 0.00014780054343834913, "loss": 0.0054, "step": 3554 }, { "epoch": 0.26, "learning_rate": 0.0001477711683924506, "loss": 0.0055, "step": 3556 }, { "epoch": 0.26, "learning_rate": 0.00014774179334655212, "loss": 0.0057, "step": 3558 }, { "epoch": 0.26, "learning_rate": 0.00014771241830065362, "loss": 0.0054, "step": 3560 }, { "epoch": 0.26, "learning_rate": 0.00014768304325475507, "loss": 0.0057, "step": 3562 }, { "epoch": 0.26, "learning_rate": 0.00014765366820885658, "loss": 0.0057, "step": 3564 }, { "epoch": 0.26, "learning_rate": 0.00014762429316295808, "loss": 0.0056, "step": 3566 }, { "epoch": 0.26, "learning_rate": 0.00014759491811705956, "loss": 0.0054, "step": 3568 }, { "epoch": 0.26, "learning_rate": 0.00014756554307116107, "loss": 0.0057, "step": 3570 }, { "epoch": 0.26, "learning_rate": 0.00014753616802526255, "loss": 0.0056, "step": 3572 }, { "epoch": 0.26, "learning_rate": 0.00014750679297936402, "loss": 0.0056, "step": 3574 }, { "epoch": 0.26, "learning_rate": 0.00014747741793346553, "loss": 0.0056, "step": 3576 }, { "epoch": 0.26, "learning_rate": 0.000147448042887567, "loss": 0.0056, "step": 3578 }, { "epoch": 0.26, "learning_rate": 0.0001474186678416685, "loss": 0.0055, "step": 3580 }, { "epoch": 0.26, "learning_rate": 0.00014738929279577002, "loss": 0.0059, "step": 3582 }, { "epoch": 0.26, "learning_rate": 0.0001473599177498715, "loss": 0.0059, "step": 3584 }, { "epoch": 0.26, "learning_rate": 0.00014733054270397298, "loss": 0.0056, "step": 3586 }, { "epoch": 0.26, "learning_rate": 0.00014730116765807448, "loss": 0.0056, "step": 3588 }, { "epoch": 0.26, "learning_rate": 0.00014727179261217596, "loss": 0.0056, "step": 3590 }, { "epoch": 0.26, "learning_rate": 0.00014724241756627746, "loss": 0.0054, "step": 3592 }, { "epoch": 0.26, "learning_rate": 0.00014721304252037894, "loss": 0.0057, "step": 3594 }, { "epoch": 0.26, "learning_rate": 0.00014718366747448045, "loss": 0.0058, "step": 3596 }, { "epoch": 0.26, "learning_rate": 0.00014715429242858193, "loss": 0.0056, "step": 3598 }, { "epoch": 0.26, "learning_rate": 0.0001471249173826834, "loss": 0.0056, "step": 3600 }, { "epoch": 0.26, "learning_rate": 0.0001470955423367849, "loss": 0.0058, "step": 3602 }, { "epoch": 0.26, "learning_rate": 0.00014706616729088642, "loss": 0.0056, "step": 3604 }, { "epoch": 0.26, "learning_rate": 0.0001470367922449879, "loss": 0.0056, "step": 3606 }, { "epoch": 0.26, "learning_rate": 0.00014700741719908937, "loss": 0.0056, "step": 3608 }, { "epoch": 0.27, "learning_rate": 0.00014697804215319085, "loss": 0.0056, "step": 3610 }, { "epoch": 0.27, "learning_rate": 0.00014694866710729236, "loss": 0.0056, "step": 3612 }, { "epoch": 0.27, "learning_rate": 0.00014691929206139386, "loss": 0.0058, "step": 3614 }, { "epoch": 0.27, "learning_rate": 0.00014688991701549534, "loss": 0.0056, "step": 3616 }, { "epoch": 0.27, "learning_rate": 0.00014686054196959685, "loss": 0.0057, "step": 3618 }, { "epoch": 0.27, "learning_rate": 0.00014683116692369832, "loss": 0.0056, "step": 3620 }, { "epoch": 0.27, "learning_rate": 0.0001468017918777998, "loss": 0.0055, "step": 3622 }, { "epoch": 0.27, "learning_rate": 0.0001467724168319013, "loss": 0.0057, "step": 3624 }, { "epoch": 0.27, "learning_rate": 0.0001467430417860028, "loss": 0.0055, "step": 3626 }, { "epoch": 0.27, "learning_rate": 0.0001467136667401043, "loss": 0.0056, "step": 3628 }, { "epoch": 0.27, "learning_rate": 0.0001466842916942058, "loss": 0.0057, "step": 3630 }, { "epoch": 0.27, "learning_rate": 0.00014665491664830725, "loss": 0.0057, "step": 3632 }, { "epoch": 0.27, "learning_rate": 0.00014662554160240875, "loss": 0.0055, "step": 3634 }, { "epoch": 0.27, "learning_rate": 0.00014659616655651026, "loss": 0.0053, "step": 3636 }, { "epoch": 0.27, "learning_rate": 0.00014656679151061174, "loss": 0.0055, "step": 3638 }, { "epoch": 0.27, "learning_rate": 0.00014653741646471324, "loss": 0.0056, "step": 3640 }, { "epoch": 0.27, "learning_rate": 0.00014650804141881472, "loss": 0.0054, "step": 3642 }, { "epoch": 0.27, "learning_rate": 0.0001464786663729162, "loss": 0.0054, "step": 3644 }, { "epoch": 0.27, "learning_rate": 0.0001464492913270177, "loss": 0.0056, "step": 3646 }, { "epoch": 0.27, "learning_rate": 0.00014641991628111918, "loss": 0.0059, "step": 3648 }, { "epoch": 0.27, "learning_rate": 0.0001463905412352207, "loss": 0.0055, "step": 3650 }, { "epoch": 0.27, "learning_rate": 0.0001463611661893222, "loss": 0.0056, "step": 3652 }, { "epoch": 0.27, "learning_rate": 0.00014633179114342365, "loss": 0.0056, "step": 3654 }, { "epoch": 0.27, "learning_rate": 0.00014630241609752515, "loss": 0.0054, "step": 3656 }, { "epoch": 0.27, "learning_rate": 0.00014627304105162666, "loss": 0.0055, "step": 3658 }, { "epoch": 0.27, "learning_rate": 0.00014624366600572814, "loss": 0.0053, "step": 3660 }, { "epoch": 0.27, "learning_rate": 0.00014621429095982964, "loss": 0.0057, "step": 3662 }, { "epoch": 0.27, "learning_rate": 0.00014618491591393112, "loss": 0.0055, "step": 3664 }, { "epoch": 0.27, "learning_rate": 0.0001461555408680326, "loss": 0.0058, "step": 3666 }, { "epoch": 0.27, "learning_rate": 0.0001461261658221341, "loss": 0.0057, "step": 3668 }, { "epoch": 0.27, "learning_rate": 0.00014609679077623558, "loss": 0.0056, "step": 3670 }, { "epoch": 0.27, "learning_rate": 0.0001460674157303371, "loss": 0.0057, "step": 3672 }, { "epoch": 0.27, "learning_rate": 0.0001460380406844386, "loss": 0.0057, "step": 3674 }, { "epoch": 0.27, "learning_rate": 0.00014600866563854007, "loss": 0.0054, "step": 3676 }, { "epoch": 0.27, "learning_rate": 0.00014597929059264155, "loss": 0.0056, "step": 3678 }, { "epoch": 0.27, "learning_rate": 0.00014594991554674305, "loss": 0.0054, "step": 3680 }, { "epoch": 0.27, "learning_rate": 0.00014592054050084453, "loss": 0.0055, "step": 3682 }, { "epoch": 0.27, "learning_rate": 0.00014589116545494604, "loss": 0.0056, "step": 3684 }, { "epoch": 0.27, "learning_rate": 0.00014586179040904752, "loss": 0.0057, "step": 3686 }, { "epoch": 0.27, "learning_rate": 0.00014583241536314902, "loss": 0.0057, "step": 3688 }, { "epoch": 0.27, "learning_rate": 0.0001458030403172505, "loss": 0.0059, "step": 3690 }, { "epoch": 0.27, "learning_rate": 0.00014577366527135198, "loss": 0.0055, "step": 3692 }, { "epoch": 0.27, "learning_rate": 0.00014574429022545348, "loss": 0.0054, "step": 3694 }, { "epoch": 0.27, "learning_rate": 0.000145714915179555, "loss": 0.0056, "step": 3696 }, { "epoch": 0.27, "learning_rate": 0.00014568554013365647, "loss": 0.0056, "step": 3698 }, { "epoch": 0.27, "learning_rate": 0.00014565616508775797, "loss": 0.0056, "step": 3700 }, { "epoch": 0.27, "learning_rate": 0.00014562679004185945, "loss": 0.0057, "step": 3702 }, { "epoch": 0.27, "learning_rate": 0.00014559741499596093, "loss": 0.0054, "step": 3704 }, { "epoch": 0.27, "learning_rate": 0.00014556803995006244, "loss": 0.0057, "step": 3706 }, { "epoch": 0.27, "learning_rate": 0.00014553866490416391, "loss": 0.0056, "step": 3708 }, { "epoch": 0.27, "learning_rate": 0.00014550928985826542, "loss": 0.0054, "step": 3710 }, { "epoch": 0.27, "learning_rate": 0.0001454799148123669, "loss": 0.0056, "step": 3712 }, { "epoch": 0.27, "learning_rate": 0.00014545053976646838, "loss": 0.0053, "step": 3714 }, { "epoch": 0.27, "learning_rate": 0.00014542116472056988, "loss": 0.0058, "step": 3716 }, { "epoch": 0.27, "learning_rate": 0.0001453917896746714, "loss": 0.0055, "step": 3718 }, { "epoch": 0.27, "learning_rate": 0.00014536241462877287, "loss": 0.0058, "step": 3720 }, { "epoch": 0.27, "learning_rate": 0.00014533303958287437, "loss": 0.0056, "step": 3722 }, { "epoch": 0.27, "learning_rate": 0.00014530366453697585, "loss": 0.0055, "step": 3724 }, { "epoch": 0.27, "learning_rate": 0.00014527428949107733, "loss": 0.0056, "step": 3726 }, { "epoch": 0.27, "learning_rate": 0.00014524491444517883, "loss": 0.0055, "step": 3728 }, { "epoch": 0.27, "learning_rate": 0.0001452155393992803, "loss": 0.0055, "step": 3730 }, { "epoch": 0.27, "learning_rate": 0.00014518616435338182, "loss": 0.0057, "step": 3732 }, { "epoch": 0.27, "learning_rate": 0.00014515678930748332, "loss": 0.0054, "step": 3734 }, { "epoch": 0.27, "learning_rate": 0.00014512741426158477, "loss": 0.0057, "step": 3736 }, { "epoch": 0.27, "learning_rate": 0.00014509803921568628, "loss": 0.0058, "step": 3738 }, { "epoch": 0.27, "learning_rate": 0.00014506866416978776, "loss": 0.0055, "step": 3740 }, { "epoch": 0.27, "learning_rate": 0.00014503928912388926, "loss": 0.0055, "step": 3742 }, { "epoch": 0.27, "learning_rate": 0.00014500991407799077, "loss": 0.0056, "step": 3744 }, { "epoch": 0.28, "learning_rate": 0.00014498053903209225, "loss": 0.0055, "step": 3746 }, { "epoch": 0.28, "learning_rate": 0.00014495116398619373, "loss": 0.0056, "step": 3748 }, { "epoch": 0.28, "learning_rate": 0.00014492178894029523, "loss": 0.0058, "step": 3750 }, { "epoch": 0.28, "learning_rate": 0.0001448924138943967, "loss": 0.0056, "step": 3752 }, { "epoch": 0.28, "learning_rate": 0.00014486303884849822, "loss": 0.0055, "step": 3754 }, { "epoch": 0.28, "learning_rate": 0.0001448336638025997, "loss": 0.0055, "step": 3756 }, { "epoch": 0.28, "learning_rate": 0.00014480428875670117, "loss": 0.0057, "step": 3758 }, { "epoch": 0.28, "learning_rate": 0.00014477491371080268, "loss": 0.0056, "step": 3760 }, { "epoch": 0.28, "learning_rate": 0.00014474553866490416, "loss": 0.0057, "step": 3762 }, { "epoch": 0.28, "learning_rate": 0.00014471616361900566, "loss": 0.0055, "step": 3764 }, { "epoch": 0.28, "learning_rate": 0.00014468678857310717, "loss": 0.0057, "step": 3766 }, { "epoch": 0.28, "learning_rate": 0.00014465741352720865, "loss": 0.0054, "step": 3768 }, { "epoch": 0.28, "learning_rate": 0.00014462803848131012, "loss": 0.0055, "step": 3770 }, { "epoch": 0.28, "learning_rate": 0.00014459866343541163, "loss": 0.0052, "step": 3772 }, { "epoch": 0.28, "learning_rate": 0.0001445692883895131, "loss": 0.0055, "step": 3774 }, { "epoch": 0.28, "learning_rate": 0.0001445399133436146, "loss": 0.0056, "step": 3776 }, { "epoch": 0.28, "learning_rate": 0.0001445105382977161, "loss": 0.0055, "step": 3778 }, { "epoch": 0.28, "learning_rate": 0.0001444811632518176, "loss": 0.0055, "step": 3780 }, { "epoch": 0.28, "learning_rate": 0.00014445178820591907, "loss": 0.0057, "step": 3782 }, { "epoch": 0.28, "learning_rate": 0.00014442241316002055, "loss": 0.0056, "step": 3784 }, { "epoch": 0.28, "learning_rate": 0.00014439303811412206, "loss": 0.0055, "step": 3786 }, { "epoch": 0.28, "learning_rate": 0.00014436366306822356, "loss": 0.0056, "step": 3788 }, { "epoch": 0.28, "learning_rate": 0.00014433428802232504, "loss": 0.0056, "step": 3790 }, { "epoch": 0.28, "learning_rate": 0.00014430491297642655, "loss": 0.0055, "step": 3792 }, { "epoch": 0.28, "learning_rate": 0.00014427553793052803, "loss": 0.0056, "step": 3794 }, { "epoch": 0.28, "learning_rate": 0.0001442461628846295, "loss": 0.0055, "step": 3796 }, { "epoch": 0.28, "learning_rate": 0.000144216787838731, "loss": 0.0057, "step": 3798 }, { "epoch": 0.28, "learning_rate": 0.0001441874127928325, "loss": 0.0054, "step": 3800 }, { "epoch": 0.28, "learning_rate": 0.000144158037746934, "loss": 0.0055, "step": 3802 }, { "epoch": 0.28, "learning_rate": 0.00014412866270103547, "loss": 0.0057, "step": 3804 }, { "epoch": 0.28, "learning_rate": 0.00014409928765513695, "loss": 0.0056, "step": 3806 }, { "epoch": 0.28, "learning_rate": 0.00014406991260923846, "loss": 0.0055, "step": 3808 }, { "epoch": 0.28, "learning_rate": 0.00014404053756333996, "loss": 0.0059, "step": 3810 }, { "epoch": 0.28, "learning_rate": 0.00014401116251744144, "loss": 0.0056, "step": 3812 }, { "epoch": 0.28, "learning_rate": 0.00014398178747154295, "loss": 0.0056, "step": 3814 }, { "epoch": 0.28, "learning_rate": 0.00014395241242564442, "loss": 0.0055, "step": 3816 }, { "epoch": 0.28, "learning_rate": 0.0001439230373797459, "loss": 0.0056, "step": 3818 }, { "epoch": 0.28, "learning_rate": 0.0001438936623338474, "loss": 0.0056, "step": 3820 }, { "epoch": 0.28, "learning_rate": 0.00014386428728794889, "loss": 0.0058, "step": 3822 }, { "epoch": 0.28, "learning_rate": 0.0001438349122420504, "loss": 0.0053, "step": 3824 }, { "epoch": 0.28, "learning_rate": 0.0001438055371961519, "loss": 0.0054, "step": 3826 }, { "epoch": 0.28, "learning_rate": 0.00014377616215025335, "loss": 0.0058, "step": 3828 }, { "epoch": 0.28, "learning_rate": 0.00014374678710435485, "loss": 0.0057, "step": 3830 }, { "epoch": 0.28, "learning_rate": 0.00014371741205845636, "loss": 0.0057, "step": 3832 }, { "epoch": 0.28, "learning_rate": 0.00014368803701255784, "loss": 0.0055, "step": 3834 }, { "epoch": 0.28, "learning_rate": 0.00014365866196665934, "loss": 0.0056, "step": 3836 }, { "epoch": 0.28, "learning_rate": 0.00014362928692076082, "loss": 0.0056, "step": 3838 }, { "epoch": 0.28, "learning_rate": 0.0001435999118748623, "loss": 0.0056, "step": 3840 }, { "epoch": 0.28, "learning_rate": 0.0001435705368289638, "loss": 0.0058, "step": 3842 }, { "epoch": 0.28, "learning_rate": 0.00014354116178306528, "loss": 0.0056, "step": 3844 }, { "epoch": 0.28, "learning_rate": 0.0001435117867371668, "loss": 0.0055, "step": 3846 }, { "epoch": 0.28, "learning_rate": 0.0001434824116912683, "loss": 0.0054, "step": 3848 }, { "epoch": 0.28, "learning_rate": 0.00014345303664536977, "loss": 0.0055, "step": 3850 }, { "epoch": 0.28, "learning_rate": 0.00014342366159947125, "loss": 0.0057, "step": 3852 }, { "epoch": 0.28, "learning_rate": 0.00014339428655357273, "loss": 0.0056, "step": 3854 }, { "epoch": 0.28, "learning_rate": 0.00014336491150767424, "loss": 0.0056, "step": 3856 }, { "epoch": 0.28, "learning_rate": 0.00014333553646177574, "loss": 0.0054, "step": 3858 }, { "epoch": 0.28, "learning_rate": 0.00014330616141587722, "loss": 0.0056, "step": 3860 }, { "epoch": 0.28, "learning_rate": 0.0001432767863699787, "loss": 0.0057, "step": 3862 }, { "epoch": 0.28, "learning_rate": 0.0001432474113240802, "loss": 0.0054, "step": 3864 }, { "epoch": 0.28, "learning_rate": 0.00014321803627818168, "loss": 0.0055, "step": 3866 }, { "epoch": 0.28, "learning_rate": 0.0001431886612322832, "loss": 0.0056, "step": 3868 }, { "epoch": 0.28, "learning_rate": 0.00014315928618638467, "loss": 0.0057, "step": 3870 }, { "epoch": 0.28, "learning_rate": 0.00014312991114048617, "loss": 0.0056, "step": 3872 }, { "epoch": 0.28, "learning_rate": 0.00014310053609458765, "loss": 0.0058, "step": 3874 }, { "epoch": 0.28, "learning_rate": 0.00014307116104868913, "loss": 0.0058, "step": 3876 }, { "epoch": 0.28, "learning_rate": 0.00014304178600279063, "loss": 0.0056, "step": 3878 }, { "epoch": 0.28, "learning_rate": 0.00014301241095689214, "loss": 0.0057, "step": 3880 }, { "epoch": 0.29, "learning_rate": 0.00014298303591099362, "loss": 0.0056, "step": 3882 }, { "epoch": 0.29, "learning_rate": 0.00014295366086509512, "loss": 0.0057, "step": 3884 }, { "epoch": 0.29, "learning_rate": 0.0001429242858191966, "loss": 0.0055, "step": 3886 }, { "epoch": 0.29, "learning_rate": 0.00014289491077329808, "loss": 0.0057, "step": 3888 }, { "epoch": 0.29, "learning_rate": 0.00014286553572739958, "loss": 0.0056, "step": 3890 }, { "epoch": 0.29, "learning_rate": 0.00014283616068150106, "loss": 0.0058, "step": 3892 }, { "epoch": 0.29, "learning_rate": 0.00014280678563560257, "loss": 0.0055, "step": 3894 }, { "epoch": 0.29, "learning_rate": 0.00014277741058970407, "loss": 0.0055, "step": 3896 }, { "epoch": 0.29, "learning_rate": 0.00014274803554380553, "loss": 0.0057, "step": 3898 }, { "epoch": 0.29, "learning_rate": 0.00014271866049790703, "loss": 0.0056, "step": 3900 }, { "epoch": 0.29, "learning_rate": 0.00014268928545200854, "loss": 0.0055, "step": 3902 }, { "epoch": 0.29, "learning_rate": 0.00014265991040611001, "loss": 0.0054, "step": 3904 }, { "epoch": 0.29, "learning_rate": 0.00014263053536021152, "loss": 0.0055, "step": 3906 }, { "epoch": 0.29, "learning_rate": 0.000142601160314313, "loss": 0.0055, "step": 3908 }, { "epoch": 0.29, "learning_rate": 0.00014257178526841448, "loss": 0.0056, "step": 3910 }, { "epoch": 0.29, "learning_rate": 0.00014254241022251598, "loss": 0.0055, "step": 3912 }, { "epoch": 0.29, "learning_rate": 0.00014251303517661746, "loss": 0.0057, "step": 3914 }, { "epoch": 0.29, "learning_rate": 0.00014248366013071897, "loss": 0.0053, "step": 3916 }, { "epoch": 0.29, "learning_rate": 0.00014245428508482047, "loss": 0.0056, "step": 3918 }, { "epoch": 0.29, "learning_rate": 0.00014242491003892192, "loss": 0.0055, "step": 3920 }, { "epoch": 0.29, "learning_rate": 0.00014239553499302343, "loss": 0.0058, "step": 3922 }, { "epoch": 0.29, "learning_rate": 0.00014236615994712493, "loss": 0.0056, "step": 3924 }, { "epoch": 0.29, "learning_rate": 0.0001423367849012264, "loss": 0.0055, "step": 3926 }, { "epoch": 0.29, "learning_rate": 0.00014230740985532792, "loss": 0.0056, "step": 3928 }, { "epoch": 0.29, "learning_rate": 0.0001422780348094294, "loss": 0.0057, "step": 3930 }, { "epoch": 0.29, "learning_rate": 0.00014224865976353087, "loss": 0.0055, "step": 3932 }, { "epoch": 0.29, "learning_rate": 0.00014221928471763238, "loss": 0.0055, "step": 3934 }, { "epoch": 0.29, "learning_rate": 0.00014218990967173386, "loss": 0.0055, "step": 3936 }, { "epoch": 0.29, "learning_rate": 0.00014216053462583536, "loss": 0.0055, "step": 3938 }, { "epoch": 0.29, "learning_rate": 0.00014213115957993687, "loss": 0.0056, "step": 3940 }, { "epoch": 0.29, "learning_rate": 0.00014210178453403835, "loss": 0.0056, "step": 3942 }, { "epoch": 0.29, "learning_rate": 0.00014207240948813983, "loss": 0.0058, "step": 3944 }, { "epoch": 0.29, "learning_rate": 0.00014204303444224133, "loss": 0.0054, "step": 3946 }, { "epoch": 0.29, "learning_rate": 0.0001420136593963428, "loss": 0.0057, "step": 3948 }, { "epoch": 0.29, "learning_rate": 0.00014198428435044431, "loss": 0.0055, "step": 3950 }, { "epoch": 0.29, "learning_rate": 0.0001419549093045458, "loss": 0.0055, "step": 3952 }, { "epoch": 0.29, "learning_rate": 0.0001419255342586473, "loss": 0.0056, "step": 3954 }, { "epoch": 0.29, "learning_rate": 0.00014189615921274878, "loss": 0.0054, "step": 3956 }, { "epoch": 0.29, "learning_rate": 0.00014186678416685026, "loss": 0.0056, "step": 3958 }, { "epoch": 0.29, "learning_rate": 0.00014183740912095176, "loss": 0.0057, "step": 3960 }, { "epoch": 0.29, "learning_rate": 0.00014180803407505327, "loss": 0.0057, "step": 3962 }, { "epoch": 0.29, "learning_rate": 0.00014177865902915474, "loss": 0.0056, "step": 3964 }, { "epoch": 0.29, "learning_rate": 0.00014174928398325622, "loss": 0.0058, "step": 3966 }, { "epoch": 0.29, "learning_rate": 0.0001417199089373577, "loss": 0.0055, "step": 3968 }, { "epoch": 0.29, "learning_rate": 0.0001416905338914592, "loss": 0.0056, "step": 3970 }, { "epoch": 0.29, "learning_rate": 0.0001416611588455607, "loss": 0.0054, "step": 3972 }, { "epoch": 0.29, "learning_rate": 0.0001416317837996622, "loss": 0.0055, "step": 3974 }, { "epoch": 0.29, "learning_rate": 0.0001416024087537637, "loss": 0.0056, "step": 3976 }, { "epoch": 0.29, "learning_rate": 0.00014157303370786517, "loss": 0.0055, "step": 3978 }, { "epoch": 0.29, "learning_rate": 0.00014154365866196665, "loss": 0.0056, "step": 3980 }, { "epoch": 0.29, "learning_rate": 0.00014151428361606816, "loss": 0.0055, "step": 3982 }, { "epoch": 0.29, "learning_rate": 0.00014148490857016964, "loss": 0.0056, "step": 3984 }, { "epoch": 0.29, "learning_rate": 0.00014145553352427114, "loss": 0.0055, "step": 3986 }, { "epoch": 0.29, "learning_rate": 0.00014142615847837265, "loss": 0.0056, "step": 3988 }, { "epoch": 0.29, "learning_rate": 0.0001413967834324741, "loss": 0.0055, "step": 3990 }, { "epoch": 0.29, "learning_rate": 0.0001413674083865756, "loss": 0.0054, "step": 3992 }, { "epoch": 0.29, "learning_rate": 0.0001413380333406771, "loss": 0.0056, "step": 3994 }, { "epoch": 0.29, "learning_rate": 0.0001413086582947786, "loss": 0.0055, "step": 3996 }, { "epoch": 0.29, "learning_rate": 0.0001412792832488801, "loss": 0.0056, "step": 3998 }, { "epoch": 0.29, "learning_rate": 0.00014124990820298157, "loss": 0.0057, "step": 4000 }, { "epoch": 0.29, "learning_rate": 0.00014122053315708305, "loss": 0.0055, "step": 4002 }, { "epoch": 0.29, "learning_rate": 0.00014119115811118456, "loss": 0.0055, "step": 4004 }, { "epoch": 0.29, "learning_rate": 0.00014116178306528603, "loss": 0.0054, "step": 4006 }, { "epoch": 0.29, "learning_rate": 0.00014113240801938754, "loss": 0.0058, "step": 4008 }, { "epoch": 0.29, "learning_rate": 0.00014110303297348905, "loss": 0.0057, "step": 4010 }, { "epoch": 0.29, "learning_rate": 0.0001410736579275905, "loss": 0.0056, "step": 4012 }, { "epoch": 0.29, "learning_rate": 0.000141044282881692, "loss": 0.0057, "step": 4014 }, { "epoch": 0.29, "learning_rate": 0.0001410149078357935, "loss": 0.0055, "step": 4016 }, { "epoch": 0.3, "learning_rate": 0.00014098553278989499, "loss": 0.0055, "step": 4018 }, { "epoch": 0.3, "learning_rate": 0.0001409561577439965, "loss": 0.0055, "step": 4020 }, { "epoch": 0.3, "learning_rate": 0.00014092678269809797, "loss": 0.0055, "step": 4022 }, { "epoch": 0.3, "learning_rate": 0.00014089740765219945, "loss": 0.0055, "step": 4024 }, { "epoch": 0.3, "learning_rate": 0.00014086803260630095, "loss": 0.0057, "step": 4026 }, { "epoch": 0.3, "learning_rate": 0.00014083865756040243, "loss": 0.006, "step": 4028 }, { "epoch": 0.3, "learning_rate": 0.00014080928251450394, "loss": 0.0055, "step": 4030 }, { "epoch": 0.3, "learning_rate": 0.00014077990746860544, "loss": 0.0055, "step": 4032 }, { "epoch": 0.3, "learning_rate": 0.00014075053242270692, "loss": 0.0058, "step": 4034 }, { "epoch": 0.3, "learning_rate": 0.0001407211573768084, "loss": 0.0057, "step": 4036 }, { "epoch": 0.3, "learning_rate": 0.0001406917823309099, "loss": 0.0057, "step": 4038 }, { "epoch": 0.3, "learning_rate": 0.00014066240728501138, "loss": 0.0056, "step": 4040 }, { "epoch": 0.3, "learning_rate": 0.0001406330322391129, "loss": 0.0054, "step": 4042 }, { "epoch": 0.3, "learning_rate": 0.00014060365719321437, "loss": 0.0057, "step": 4044 }, { "epoch": 0.3, "learning_rate": 0.00014057428214731587, "loss": 0.0057, "step": 4046 }, { "epoch": 0.3, "learning_rate": 0.00014054490710141735, "loss": 0.0055, "step": 4048 }, { "epoch": 0.3, "learning_rate": 0.00014051553205551883, "loss": 0.0056, "step": 4050 }, { "epoch": 0.3, "learning_rate": 0.00014048615700962034, "loss": 0.0055, "step": 4052 }, { "epoch": 0.3, "learning_rate": 0.00014045678196372184, "loss": 0.0056, "step": 4054 }, { "epoch": 0.3, "learning_rate": 0.00014042740691782332, "loss": 0.0054, "step": 4056 }, { "epoch": 0.3, "learning_rate": 0.0001403980318719248, "loss": 0.0054, "step": 4058 }, { "epoch": 0.3, "learning_rate": 0.0001403686568260263, "loss": 0.0054, "step": 4060 }, { "epoch": 0.3, "learning_rate": 0.00014033928178012778, "loss": 0.0056, "step": 4062 }, { "epoch": 0.3, "learning_rate": 0.0001403099067342293, "loss": 0.0058, "step": 4064 }, { "epoch": 0.3, "learning_rate": 0.00014028053168833076, "loss": 0.0056, "step": 4066 }, { "epoch": 0.3, "learning_rate": 0.00014025115664243227, "loss": 0.0058, "step": 4068 }, { "epoch": 0.3, "learning_rate": 0.00014022178159653375, "loss": 0.0054, "step": 4070 }, { "epoch": 0.3, "learning_rate": 0.00014019240655063523, "loss": 0.0054, "step": 4072 }, { "epoch": 0.3, "learning_rate": 0.00014016303150473673, "loss": 0.0057, "step": 4074 }, { "epoch": 0.3, "learning_rate": 0.00014013365645883824, "loss": 0.0057, "step": 4076 }, { "epoch": 0.3, "learning_rate": 0.00014010428141293972, "loss": 0.0056, "step": 4078 }, { "epoch": 0.3, "learning_rate": 0.00014007490636704122, "loss": 0.0058, "step": 4080 }, { "epoch": 0.3, "learning_rate": 0.00014004553132114267, "loss": 0.0053, "step": 4082 }, { "epoch": 0.3, "learning_rate": 0.00014001615627524418, "loss": 0.0058, "step": 4084 }, { "epoch": 0.3, "learning_rate": 0.00013998678122934568, "loss": 0.0056, "step": 4086 }, { "epoch": 0.3, "learning_rate": 0.00013995740618344716, "loss": 0.0054, "step": 4088 }, { "epoch": 0.3, "learning_rate": 0.00013992803113754867, "loss": 0.0054, "step": 4090 }, { "epoch": 0.3, "learning_rate": 0.00013989865609165015, "loss": 0.0055, "step": 4092 }, { "epoch": 0.3, "learning_rate": 0.00013986928104575162, "loss": 0.0056, "step": 4094 }, { "epoch": 0.3, "learning_rate": 0.00013983990599985313, "loss": 0.0055, "step": 4096 }, { "epoch": 0.3, "learning_rate": 0.0001398105309539546, "loss": 0.0054, "step": 4098 }, { "epoch": 0.3, "learning_rate": 0.00013978115590805611, "loss": 0.0055, "step": 4100 }, { "epoch": 0.3, "learning_rate": 0.00013975178086215762, "loss": 0.0056, "step": 4102 }, { "epoch": 0.3, "learning_rate": 0.0001397224058162591, "loss": 0.0054, "step": 4104 }, { "epoch": 0.3, "learning_rate": 0.00013969303077036058, "loss": 0.0059, "step": 4106 }, { "epoch": 0.3, "learning_rate": 0.00013966365572446208, "loss": 0.0057, "step": 4108 }, { "epoch": 0.3, "learning_rate": 0.00013963428067856356, "loss": 0.0057, "step": 4110 }, { "epoch": 0.3, "learning_rate": 0.00013960490563266507, "loss": 0.0056, "step": 4112 }, { "epoch": 0.3, "learning_rate": 0.00013957553058676654, "loss": 0.0056, "step": 4114 }, { "epoch": 0.3, "learning_rate": 0.00013954615554086802, "loss": 0.0056, "step": 4116 }, { "epoch": 0.3, "learning_rate": 0.00013951678049496953, "loss": 0.0057, "step": 4118 }, { "epoch": 0.3, "learning_rate": 0.000139487405449071, "loss": 0.0054, "step": 4120 }, { "epoch": 0.3, "learning_rate": 0.0001394580304031725, "loss": 0.0057, "step": 4122 }, { "epoch": 0.3, "learning_rate": 0.00013942865535727402, "loss": 0.0054, "step": 4124 }, { "epoch": 0.3, "learning_rate": 0.0001393992803113755, "loss": 0.0058, "step": 4126 }, { "epoch": 0.3, "learning_rate": 0.00013936990526547697, "loss": 0.0055, "step": 4128 }, { "epoch": 0.3, "learning_rate": 0.00013934053021957848, "loss": 0.0057, "step": 4130 }, { "epoch": 0.3, "learning_rate": 0.00013931115517367996, "loss": 0.0056, "step": 4132 }, { "epoch": 0.3, "learning_rate": 0.00013928178012778146, "loss": 0.0056, "step": 4134 }, { "epoch": 0.3, "learning_rate": 0.00013925240508188294, "loss": 0.0055, "step": 4136 }, { "epoch": 0.3, "learning_rate": 0.00013922303003598445, "loss": 0.0058, "step": 4138 }, { "epoch": 0.3, "learning_rate": 0.00013919365499008593, "loss": 0.0058, "step": 4140 }, { "epoch": 0.3, "learning_rate": 0.0001391642799441874, "loss": 0.0054, "step": 4142 }, { "epoch": 0.3, "learning_rate": 0.0001391349048982889, "loss": 0.0056, "step": 4144 }, { "epoch": 0.3, "learning_rate": 0.00013910552985239041, "loss": 0.0056, "step": 4146 }, { "epoch": 0.3, "learning_rate": 0.0001390761548064919, "loss": 0.0055, "step": 4148 }, { "epoch": 0.3, "learning_rate": 0.0001390467797605934, "loss": 0.0055, "step": 4150 }, { "epoch": 0.3, "learning_rate": 0.00013901740471469488, "loss": 0.0055, "step": 4152 }, { "epoch": 0.31, "learning_rate": 0.00013898802966879636, "loss": 0.0056, "step": 4154 }, { "epoch": 0.31, "learning_rate": 0.00013895865462289786, "loss": 0.0057, "step": 4156 }, { "epoch": 0.31, "learning_rate": 0.00013892927957699934, "loss": 0.0056, "step": 4158 }, { "epoch": 0.31, "learning_rate": 0.00013889990453110084, "loss": 0.0057, "step": 4160 }, { "epoch": 0.31, "learning_rate": 0.00013887052948520232, "loss": 0.0054, "step": 4162 }, { "epoch": 0.31, "learning_rate": 0.0001388411544393038, "loss": 0.0057, "step": 4164 }, { "epoch": 0.31, "learning_rate": 0.0001388117793934053, "loss": 0.0055, "step": 4166 }, { "epoch": 0.31, "learning_rate": 0.0001387824043475068, "loss": 0.0057, "step": 4168 }, { "epoch": 0.31, "learning_rate": 0.0001387530293016083, "loss": 0.0054, "step": 4170 }, { "epoch": 0.31, "learning_rate": 0.0001387236542557098, "loss": 0.0056, "step": 4172 }, { "epoch": 0.31, "learning_rate": 0.00013869427920981127, "loss": 0.0057, "step": 4174 }, { "epoch": 0.31, "learning_rate": 0.00013866490416391275, "loss": 0.0054, "step": 4176 }, { "epoch": 0.31, "learning_rate": 0.00013863552911801426, "loss": 0.0055, "step": 4178 }, { "epoch": 0.31, "learning_rate": 0.00013860615407211574, "loss": 0.0058, "step": 4180 }, { "epoch": 0.31, "learning_rate": 0.00013857677902621724, "loss": 0.0053, "step": 4182 }, { "epoch": 0.31, "learning_rate": 0.00013854740398031875, "loss": 0.0053, "step": 4184 }, { "epoch": 0.31, "learning_rate": 0.0001385180289344202, "loss": 0.0057, "step": 4186 }, { "epoch": 0.31, "learning_rate": 0.0001384886538885217, "loss": 0.0057, "step": 4188 }, { "epoch": 0.31, "learning_rate": 0.0001384592788426232, "loss": 0.0056, "step": 4190 }, { "epoch": 0.31, "learning_rate": 0.0001384299037967247, "loss": 0.0056, "step": 4192 }, { "epoch": 0.31, "learning_rate": 0.0001384005287508262, "loss": 0.0058, "step": 4194 }, { "epoch": 0.31, "learning_rate": 0.00013837115370492767, "loss": 0.0055, "step": 4196 }, { "epoch": 0.31, "learning_rate": 0.00013834177865902915, "loss": 0.0056, "step": 4198 }, { "epoch": 0.31, "learning_rate": 0.00013831240361313066, "loss": 0.0058, "step": 4200 }, { "epoch": 0.31, "learning_rate": 0.00013828302856723213, "loss": 0.0055, "step": 4202 }, { "epoch": 0.31, "learning_rate": 0.00013825365352133364, "loss": 0.0057, "step": 4204 }, { "epoch": 0.31, "learning_rate": 0.00013822427847543515, "loss": 0.0055, "step": 4206 }, { "epoch": 0.31, "learning_rate": 0.00013819490342953662, "loss": 0.0058, "step": 4208 }, { "epoch": 0.31, "learning_rate": 0.0001381655283836381, "loss": 0.0056, "step": 4210 }, { "epoch": 0.31, "learning_rate": 0.00013813615333773958, "loss": 0.0057, "step": 4212 }, { "epoch": 0.31, "learning_rate": 0.00013810677829184109, "loss": 0.0058, "step": 4214 }, { "epoch": 0.31, "learning_rate": 0.0001380774032459426, "loss": 0.0057, "step": 4216 }, { "epoch": 0.31, "learning_rate": 0.00013804802820004407, "loss": 0.0054, "step": 4218 }, { "epoch": 0.31, "learning_rate": 0.00013801865315414555, "loss": 0.0056, "step": 4220 }, { "epoch": 0.31, "learning_rate": 0.00013798927810824705, "loss": 0.0057, "step": 4222 }, { "epoch": 0.31, "learning_rate": 0.00013795990306234853, "loss": 0.0055, "step": 4224 }, { "epoch": 0.31, "learning_rate": 0.00013793052801645004, "loss": 0.0056, "step": 4226 }, { "epoch": 0.31, "learning_rate": 0.00013790115297055152, "loss": 0.0054, "step": 4228 }, { "epoch": 0.31, "learning_rate": 0.00013787177792465302, "loss": 0.0056, "step": 4230 }, { "epoch": 0.31, "learning_rate": 0.0001378424028787545, "loss": 0.0057, "step": 4232 }, { "epoch": 0.31, "learning_rate": 0.00013781302783285598, "loss": 0.0055, "step": 4234 }, { "epoch": 0.31, "learning_rate": 0.00013778365278695748, "loss": 0.0055, "step": 4236 }, { "epoch": 0.31, "learning_rate": 0.000137754277741059, "loss": 0.0056, "step": 4238 }, { "epoch": 0.31, "learning_rate": 0.00013772490269516047, "loss": 0.0054, "step": 4240 }, { "epoch": 0.31, "learning_rate": 0.00013769552764926197, "loss": 0.0058, "step": 4242 }, { "epoch": 0.31, "learning_rate": 0.00013766615260336345, "loss": 0.0056, "step": 4244 }, { "epoch": 0.31, "learning_rate": 0.00013763677755746493, "loss": 0.0056, "step": 4246 }, { "epoch": 0.31, "learning_rate": 0.00013760740251156643, "loss": 0.0053, "step": 4248 }, { "epoch": 0.31, "learning_rate": 0.0001375780274656679, "loss": 0.0056, "step": 4250 }, { "epoch": 0.31, "learning_rate": 0.00013754865241976942, "loss": 0.0057, "step": 4252 }, { "epoch": 0.31, "learning_rate": 0.00013751927737387092, "loss": 0.0057, "step": 4254 }, { "epoch": 0.31, "learning_rate": 0.00013748990232797238, "loss": 0.0058, "step": 4256 }, { "epoch": 0.31, "learning_rate": 0.00013746052728207388, "loss": 0.0054, "step": 4258 }, { "epoch": 0.31, "learning_rate": 0.00013743115223617539, "loss": 0.0056, "step": 4260 }, { "epoch": 0.31, "learning_rate": 0.00013740177719027686, "loss": 0.0057, "step": 4262 }, { "epoch": 0.31, "learning_rate": 0.00013737240214437837, "loss": 0.0054, "step": 4264 }, { "epoch": 0.31, "learning_rate": 0.00013734302709847985, "loss": 0.0054, "step": 4266 }, { "epoch": 0.31, "learning_rate": 0.00013731365205258133, "loss": 0.0057, "step": 4268 }, { "epoch": 0.31, "learning_rate": 0.00013728427700668283, "loss": 0.0059, "step": 4270 }, { "epoch": 0.31, "learning_rate": 0.0001372549019607843, "loss": 0.0055, "step": 4272 }, { "epoch": 0.31, "learning_rate": 0.00013722552691488582, "loss": 0.0056, "step": 4274 }, { "epoch": 0.31, "learning_rate": 0.00013719615186898732, "loss": 0.0056, "step": 4276 }, { "epoch": 0.31, "learning_rate": 0.00013716677682308877, "loss": 0.0056, "step": 4278 }, { "epoch": 0.31, "learning_rate": 0.00013713740177719028, "loss": 0.0056, "step": 4280 }, { "epoch": 0.31, "learning_rate": 0.00013710802673129178, "loss": 0.0056, "step": 4282 }, { "epoch": 0.31, "learning_rate": 0.00013707865168539326, "loss": 0.0056, "step": 4284 }, { "epoch": 0.31, "learning_rate": 0.00013704927663949477, "loss": 0.0057, "step": 4286 }, { "epoch": 0.31, "learning_rate": 0.00013701990159359625, "loss": 0.0055, "step": 4288 }, { "epoch": 0.32, "learning_rate": 0.00013699052654769772, "loss": 0.0056, "step": 4290 }, { "epoch": 0.32, "learning_rate": 0.00013696115150179923, "loss": 0.0059, "step": 4292 }, { "epoch": 0.32, "learning_rate": 0.0001369317764559007, "loss": 0.0055, "step": 4294 }, { "epoch": 0.32, "learning_rate": 0.00013690240141000221, "loss": 0.0057, "step": 4296 }, { "epoch": 0.32, "learning_rate": 0.00013687302636410372, "loss": 0.0055, "step": 4298 }, { "epoch": 0.32, "learning_rate": 0.0001368436513182052, "loss": 0.0057, "step": 4300 }, { "epoch": 0.32, "learning_rate": 0.00013681427627230668, "loss": 0.0057, "step": 4302 }, { "epoch": 0.32, "learning_rate": 0.00013678490122640818, "loss": 0.0055, "step": 4304 }, { "epoch": 0.32, "learning_rate": 0.00013675552618050966, "loss": 0.0057, "step": 4306 }, { "epoch": 0.32, "learning_rate": 0.00013672615113461117, "loss": 0.0056, "step": 4308 }, { "epoch": 0.32, "learning_rate": 0.00013669677608871264, "loss": 0.0058, "step": 4310 }, { "epoch": 0.32, "learning_rate": 0.00013666740104281412, "loss": 0.0059, "step": 4312 }, { "epoch": 0.32, "learning_rate": 0.00013663802599691563, "loss": 0.0055, "step": 4314 }, { "epoch": 0.32, "learning_rate": 0.0001366086509510171, "loss": 0.0057, "step": 4316 }, { "epoch": 0.32, "learning_rate": 0.0001365792759051186, "loss": 0.0055, "step": 4318 }, { "epoch": 0.32, "learning_rate": 0.00013654990085922012, "loss": 0.0056, "step": 4320 }, { "epoch": 0.32, "learning_rate": 0.0001365205258133216, "loss": 0.0057, "step": 4322 }, { "epoch": 0.32, "learning_rate": 0.00013649115076742307, "loss": 0.0057, "step": 4324 }, { "epoch": 0.32, "learning_rate": 0.00013646177572152455, "loss": 0.0055, "step": 4326 }, { "epoch": 0.32, "learning_rate": 0.00013643240067562606, "loss": 0.0054, "step": 4328 }, { "epoch": 0.32, "learning_rate": 0.00013640302562972756, "loss": 0.0055, "step": 4330 }, { "epoch": 0.32, "learning_rate": 0.00013637365058382904, "loss": 0.0056, "step": 4332 }, { "epoch": 0.32, "learning_rate": 0.00013634427553793055, "loss": 0.0054, "step": 4334 }, { "epoch": 0.32, "learning_rate": 0.00013631490049203203, "loss": 0.0054, "step": 4336 }, { "epoch": 0.32, "learning_rate": 0.0001362855254461335, "loss": 0.0054, "step": 4338 }, { "epoch": 0.32, "learning_rate": 0.000136256150400235, "loss": 0.0056, "step": 4340 }, { "epoch": 0.32, "learning_rate": 0.0001362267753543365, "loss": 0.0055, "step": 4342 }, { "epoch": 0.32, "learning_rate": 0.000136197400308438, "loss": 0.0059, "step": 4344 }, { "epoch": 0.32, "learning_rate": 0.0001361680252625395, "loss": 0.0057, "step": 4346 }, { "epoch": 0.32, "learning_rate": 0.00013613865021664095, "loss": 0.0054, "step": 4348 }, { "epoch": 0.32, "learning_rate": 0.00013610927517074245, "loss": 0.0052, "step": 4350 }, { "epoch": 0.32, "learning_rate": 0.00013607990012484396, "loss": 0.0055, "step": 4352 }, { "epoch": 0.32, "learning_rate": 0.00013605052507894544, "loss": 0.0057, "step": 4354 }, { "epoch": 0.32, "learning_rate": 0.00013602115003304694, "loss": 0.0057, "step": 4356 }, { "epoch": 0.32, "learning_rate": 0.00013599177498714842, "loss": 0.0056, "step": 4358 }, { "epoch": 0.32, "learning_rate": 0.0001359623999412499, "loss": 0.0056, "step": 4360 }, { "epoch": 0.32, "learning_rate": 0.0001359330248953514, "loss": 0.0058, "step": 4362 }, { "epoch": 0.32, "learning_rate": 0.00013590364984945288, "loss": 0.0055, "step": 4364 }, { "epoch": 0.32, "learning_rate": 0.0001358742748035544, "loss": 0.0055, "step": 4366 }, { "epoch": 0.32, "learning_rate": 0.0001358448997576559, "loss": 0.0056, "step": 4368 }, { "epoch": 0.32, "learning_rate": 0.00013581552471175735, "loss": 0.0056, "step": 4370 }, { "epoch": 0.32, "learning_rate": 0.00013578614966585885, "loss": 0.0056, "step": 4372 }, { "epoch": 0.32, "learning_rate": 0.00013575677461996036, "loss": 0.0059, "step": 4374 }, { "epoch": 0.32, "learning_rate": 0.00013572739957406184, "loss": 0.0055, "step": 4376 }, { "epoch": 0.32, "learning_rate": 0.00013569802452816334, "loss": 0.0056, "step": 4378 }, { "epoch": 0.32, "learning_rate": 0.00013566864948226482, "loss": 0.0056, "step": 4380 }, { "epoch": 0.32, "learning_rate": 0.0001356392744363663, "loss": 0.0055, "step": 4382 }, { "epoch": 0.32, "learning_rate": 0.0001356098993904678, "loss": 0.0056, "step": 4384 }, { "epoch": 0.32, "learning_rate": 0.00013558052434456928, "loss": 0.0054, "step": 4386 }, { "epoch": 0.32, "learning_rate": 0.0001355511492986708, "loss": 0.0054, "step": 4388 }, { "epoch": 0.32, "learning_rate": 0.0001355217742527723, "loss": 0.0056, "step": 4390 }, { "epoch": 0.32, "learning_rate": 0.00013549239920687377, "loss": 0.0056, "step": 4392 }, { "epoch": 0.32, "learning_rate": 0.00013546302416097525, "loss": 0.0057, "step": 4394 }, { "epoch": 0.32, "learning_rate": 0.00013543364911507676, "loss": 0.0057, "step": 4396 }, { "epoch": 0.32, "learning_rate": 0.00013540427406917823, "loss": 0.0056, "step": 4398 }, { "epoch": 0.32, "learning_rate": 0.00013537489902327974, "loss": 0.0056, "step": 4400 }, { "epoch": 0.32, "learning_rate": 0.00013534552397738122, "loss": 0.0056, "step": 4402 }, { "epoch": 0.32, "learning_rate": 0.00013531614893148272, "loss": 0.0055, "step": 4404 }, { "epoch": 0.32, "learning_rate": 0.0001352867738855842, "loss": 0.0054, "step": 4406 }, { "epoch": 0.32, "learning_rate": 0.00013525739883968568, "loss": 0.0055, "step": 4408 }, { "epoch": 0.32, "learning_rate": 0.00013522802379378719, "loss": 0.0059, "step": 4410 }, { "epoch": 0.32, "learning_rate": 0.0001351986487478887, "loss": 0.0054, "step": 4412 }, { "epoch": 0.32, "learning_rate": 0.00013516927370199017, "loss": 0.0055, "step": 4414 }, { "epoch": 0.32, "learning_rate": 0.00013513989865609165, "loss": 0.0056, "step": 4416 }, { "epoch": 0.32, "learning_rate": 0.00013511052361019315, "loss": 0.0054, "step": 4418 }, { "epoch": 0.32, "learning_rate": 0.00013508114856429463, "loss": 0.0055, "step": 4420 }, { "epoch": 0.32, "learning_rate": 0.00013505177351839614, "loss": 0.0056, "step": 4422 }, { "epoch": 0.32, "learning_rate": 0.00013502239847249762, "loss": 0.0053, "step": 4424 }, { "epoch": 0.33, "learning_rate": 0.00013499302342659912, "loss": 0.0056, "step": 4426 }, { "epoch": 0.33, "learning_rate": 0.0001349636483807006, "loss": 0.0054, "step": 4428 }, { "epoch": 0.33, "learning_rate": 0.00013493427333480208, "loss": 0.0059, "step": 4430 }, { "epoch": 0.33, "learning_rate": 0.00013490489828890358, "loss": 0.0055, "step": 4432 }, { "epoch": 0.33, "learning_rate": 0.0001348755232430051, "loss": 0.0054, "step": 4434 }, { "epoch": 0.33, "learning_rate": 0.00013484614819710657, "loss": 0.0056, "step": 4436 }, { "epoch": 0.33, "learning_rate": 0.00013481677315120807, "loss": 0.0054, "step": 4438 }, { "epoch": 0.33, "learning_rate": 0.00013478739810530952, "loss": 0.0056, "step": 4440 }, { "epoch": 0.33, "learning_rate": 0.00013475802305941103, "loss": 0.0057, "step": 4442 }, { "epoch": 0.33, "learning_rate": 0.00013472864801351253, "loss": 0.0057, "step": 4444 }, { "epoch": 0.33, "learning_rate": 0.000134699272967614, "loss": 0.0055, "step": 4446 }, { "epoch": 0.33, "learning_rate": 0.00013466989792171552, "loss": 0.0057, "step": 4448 }, { "epoch": 0.33, "learning_rate": 0.000134640522875817, "loss": 0.0055, "step": 4450 }, { "epoch": 0.33, "learning_rate": 0.00013461114782991848, "loss": 0.0055, "step": 4452 }, { "epoch": 0.33, "learning_rate": 0.00013458177278401998, "loss": 0.0056, "step": 4454 }, { "epoch": 0.33, "learning_rate": 0.00013455239773812146, "loss": 0.0054, "step": 4456 }, { "epoch": 0.33, "learning_rate": 0.00013452302269222296, "loss": 0.0056, "step": 4458 }, { "epoch": 0.33, "learning_rate": 0.00013449364764632447, "loss": 0.0056, "step": 4460 }, { "epoch": 0.33, "learning_rate": 0.00013446427260042595, "loss": 0.0056, "step": 4462 }, { "epoch": 0.33, "learning_rate": 0.00013443489755452743, "loss": 0.0055, "step": 4464 }, { "epoch": 0.33, "learning_rate": 0.00013440552250862893, "loss": 0.0057, "step": 4466 }, { "epoch": 0.33, "learning_rate": 0.0001343761474627304, "loss": 0.0055, "step": 4468 }, { "epoch": 0.33, "learning_rate": 0.00013434677241683192, "loss": 0.0055, "step": 4470 }, { "epoch": 0.33, "learning_rate": 0.0001343173973709334, "loss": 0.0054, "step": 4472 }, { "epoch": 0.33, "learning_rate": 0.00013428802232503487, "loss": 0.0056, "step": 4474 }, { "epoch": 0.33, "learning_rate": 0.00013425864727913638, "loss": 0.0057, "step": 4476 }, { "epoch": 0.33, "learning_rate": 0.00013422927223323786, "loss": 0.0057, "step": 4478 }, { "epoch": 0.33, "learning_rate": 0.00013419989718733936, "loss": 0.0057, "step": 4480 }, { "epoch": 0.33, "learning_rate": 0.00013417052214144087, "loss": 0.0057, "step": 4482 }, { "epoch": 0.33, "learning_rate": 0.00013414114709554235, "loss": 0.0057, "step": 4484 }, { "epoch": 0.33, "learning_rate": 0.00013411177204964382, "loss": 0.0055, "step": 4486 }, { "epoch": 0.33, "learning_rate": 0.00013408239700374533, "loss": 0.0058, "step": 4488 }, { "epoch": 0.33, "learning_rate": 0.0001340530219578468, "loss": 0.0057, "step": 4490 }, { "epoch": 0.33, "learning_rate": 0.0001340236469119483, "loss": 0.0055, "step": 4492 }, { "epoch": 0.33, "learning_rate": 0.0001339942718660498, "loss": 0.0055, "step": 4494 }, { "epoch": 0.33, "learning_rate": 0.0001339648968201513, "loss": 0.0054, "step": 4496 }, { "epoch": 0.33, "learning_rate": 0.00013393552177425278, "loss": 0.0056, "step": 4498 }, { "epoch": 0.33, "learning_rate": 0.00013390614672835425, "loss": 0.0056, "step": 4500 }, { "epoch": 0.33, "learning_rate": 0.00013387677168245576, "loss": 0.0058, "step": 4502 }, { "epoch": 0.33, "learning_rate": 0.00013384739663655727, "loss": 0.0056, "step": 4504 }, { "epoch": 0.33, "learning_rate": 0.00013381802159065874, "loss": 0.0056, "step": 4506 }, { "epoch": 0.33, "learning_rate": 0.00013378864654476025, "loss": 0.0055, "step": 4508 }, { "epoch": 0.33, "learning_rate": 0.00013375927149886173, "loss": 0.0057, "step": 4510 }, { "epoch": 0.33, "learning_rate": 0.0001337298964529632, "loss": 0.0059, "step": 4512 }, { "epoch": 0.33, "learning_rate": 0.0001337005214070647, "loss": 0.0056, "step": 4514 }, { "epoch": 0.33, "learning_rate": 0.0001336711463611662, "loss": 0.0055, "step": 4516 }, { "epoch": 0.33, "learning_rate": 0.0001336417713152677, "loss": 0.0056, "step": 4518 }, { "epoch": 0.33, "learning_rate": 0.00013361239626936917, "loss": 0.0055, "step": 4520 }, { "epoch": 0.33, "learning_rate": 0.00013358302122347065, "loss": 0.0057, "step": 4522 }, { "epoch": 0.33, "learning_rate": 0.00013355364617757216, "loss": 0.0055, "step": 4524 }, { "epoch": 0.33, "learning_rate": 0.00013352427113167366, "loss": 0.0056, "step": 4526 }, { "epoch": 0.33, "learning_rate": 0.00013349489608577514, "loss": 0.0054, "step": 4528 }, { "epoch": 0.33, "learning_rate": 0.00013346552103987665, "loss": 0.0057, "step": 4530 }, { "epoch": 0.33, "learning_rate": 0.00013343614599397812, "loss": 0.0055, "step": 4532 }, { "epoch": 0.33, "learning_rate": 0.0001334067709480796, "loss": 0.0055, "step": 4534 }, { "epoch": 0.33, "learning_rate": 0.0001333773959021811, "loss": 0.0057, "step": 4536 }, { "epoch": 0.33, "learning_rate": 0.0001333480208562826, "loss": 0.0055, "step": 4538 }, { "epoch": 0.33, "learning_rate": 0.0001333186458103841, "loss": 0.0057, "step": 4540 }, { "epoch": 0.33, "learning_rate": 0.0001332892707644856, "loss": 0.0057, "step": 4542 }, { "epoch": 0.33, "learning_rate": 0.00013325989571858705, "loss": 0.0054, "step": 4544 }, { "epoch": 0.33, "learning_rate": 0.00013323052067268855, "loss": 0.0055, "step": 4546 }, { "epoch": 0.33, "learning_rate": 0.00013320114562679006, "loss": 0.0056, "step": 4548 }, { "epoch": 0.33, "learning_rate": 0.00013317177058089154, "loss": 0.0057, "step": 4550 }, { "epoch": 0.33, "learning_rate": 0.00013314239553499304, "loss": 0.0056, "step": 4552 }, { "epoch": 0.33, "learning_rate": 0.00013311302048909452, "loss": 0.0053, "step": 4554 }, { "epoch": 0.33, "learning_rate": 0.000133083645443196, "loss": 0.0054, "step": 4556 }, { "epoch": 0.33, "learning_rate": 0.0001330542703972975, "loss": 0.0057, "step": 4558 }, { "epoch": 0.33, "learning_rate": 0.00013302489535139898, "loss": 0.0056, "step": 4560 }, { "epoch": 0.34, "learning_rate": 0.0001329955203055005, "loss": 0.0056, "step": 4562 }, { "epoch": 0.34, "learning_rate": 0.00013296614525960197, "loss": 0.0057, "step": 4564 }, { "epoch": 0.34, "learning_rate": 0.00013293677021370345, "loss": 0.0058, "step": 4566 }, { "epoch": 0.34, "learning_rate": 0.00013290739516780495, "loss": 0.0056, "step": 4568 }, { "epoch": 0.34, "learning_rate": 0.00013287802012190643, "loss": 0.0055, "step": 4570 }, { "epoch": 0.34, "learning_rate": 0.00013284864507600794, "loss": 0.0057, "step": 4572 }, { "epoch": 0.34, "learning_rate": 0.00013281927003010944, "loss": 0.0056, "step": 4574 }, { "epoch": 0.34, "learning_rate": 0.00013278989498421092, "loss": 0.0054, "step": 4576 }, { "epoch": 0.34, "learning_rate": 0.0001327605199383124, "loss": 0.0054, "step": 4578 }, { "epoch": 0.34, "learning_rate": 0.0001327311448924139, "loss": 0.0056, "step": 4580 }, { "epoch": 0.34, "learning_rate": 0.00013270176984651538, "loss": 0.0055, "step": 4582 }, { "epoch": 0.34, "learning_rate": 0.0001326723948006169, "loss": 0.0054, "step": 4584 }, { "epoch": 0.34, "learning_rate": 0.00013264301975471837, "loss": 0.0055, "step": 4586 }, { "epoch": 0.34, "learning_rate": 0.00013261364470881987, "loss": 0.0055, "step": 4588 }, { "epoch": 0.34, "learning_rate": 0.00013258426966292135, "loss": 0.0055, "step": 4590 }, { "epoch": 0.34, "learning_rate": 0.00013255489461702283, "loss": 0.0057, "step": 4592 }, { "epoch": 0.34, "learning_rate": 0.00013252551957112433, "loss": 0.0055, "step": 4594 }, { "epoch": 0.34, "learning_rate": 0.00013249614452522584, "loss": 0.0055, "step": 4596 }, { "epoch": 0.34, "learning_rate": 0.00013246676947932732, "loss": 0.0057, "step": 4598 }, { "epoch": 0.34, "learning_rate": 0.00013243739443342882, "loss": 0.0054, "step": 4600 }, { "epoch": 0.34, "learning_rate": 0.0001324080193875303, "loss": 0.0056, "step": 4602 }, { "epoch": 0.34, "learning_rate": 0.00013237864434163178, "loss": 0.0056, "step": 4604 }, { "epoch": 0.34, "learning_rate": 0.00013234926929573329, "loss": 0.0055, "step": 4606 }, { "epoch": 0.34, "learning_rate": 0.00013231989424983476, "loss": 0.0056, "step": 4608 }, { "epoch": 0.34, "learning_rate": 0.00013229051920393627, "loss": 0.0056, "step": 4610 }, { "epoch": 0.34, "learning_rate": 0.00013226114415803777, "loss": 0.0055, "step": 4612 }, { "epoch": 0.34, "learning_rate": 0.00013223176911213923, "loss": 0.0056, "step": 4614 }, { "epoch": 0.34, "learning_rate": 0.00013220239406624073, "loss": 0.0057, "step": 4616 }, { "epoch": 0.34, "learning_rate": 0.00013217301902034224, "loss": 0.0056, "step": 4618 }, { "epoch": 0.34, "learning_rate": 0.00013214364397444372, "loss": 0.0055, "step": 4620 }, { "epoch": 0.34, "learning_rate": 0.00013211426892854522, "loss": 0.0056, "step": 4622 }, { "epoch": 0.34, "learning_rate": 0.0001320848938826467, "loss": 0.0056, "step": 4624 }, { "epoch": 0.34, "learning_rate": 0.00013205551883674818, "loss": 0.0055, "step": 4626 }, { "epoch": 0.34, "learning_rate": 0.00013202614379084968, "loss": 0.0058, "step": 4628 }, { "epoch": 0.34, "learning_rate": 0.00013199676874495116, "loss": 0.0057, "step": 4630 }, { "epoch": 0.34, "learning_rate": 0.00013196739369905267, "loss": 0.0057, "step": 4632 }, { "epoch": 0.34, "learning_rate": 0.00013193801865315417, "loss": 0.0054, "step": 4634 }, { "epoch": 0.34, "learning_rate": 0.00013190864360725562, "loss": 0.0056, "step": 4636 }, { "epoch": 0.34, "learning_rate": 0.00013187926856135713, "loss": 0.0055, "step": 4638 }, { "epoch": 0.34, "learning_rate": 0.00013184989351545863, "loss": 0.0054, "step": 4640 }, { "epoch": 0.34, "learning_rate": 0.0001318205184695601, "loss": 0.0055, "step": 4642 }, { "epoch": 0.34, "learning_rate": 0.00013179114342366162, "loss": 0.0059, "step": 4644 }, { "epoch": 0.34, "learning_rate": 0.0001317617683777631, "loss": 0.0055, "step": 4646 }, { "epoch": 0.34, "learning_rate": 0.00013173239333186457, "loss": 0.0055, "step": 4648 }, { "epoch": 0.34, "learning_rate": 0.00013170301828596608, "loss": 0.0056, "step": 4650 }, { "epoch": 0.34, "learning_rate": 0.00013167364324006756, "loss": 0.0055, "step": 4652 }, { "epoch": 0.34, "learning_rate": 0.00013164426819416906, "loss": 0.0056, "step": 4654 }, { "epoch": 0.34, "learning_rate": 0.00013161489314827057, "loss": 0.0053, "step": 4656 }, { "epoch": 0.34, "learning_rate": 0.00013158551810237205, "loss": 0.0053, "step": 4658 }, { "epoch": 0.34, "learning_rate": 0.00013155614305647353, "loss": 0.0054, "step": 4660 }, { "epoch": 0.34, "learning_rate": 0.00013152676801057503, "loss": 0.0056, "step": 4662 }, { "epoch": 0.34, "learning_rate": 0.0001314973929646765, "loss": 0.0056, "step": 4664 }, { "epoch": 0.34, "learning_rate": 0.00013146801791877802, "loss": 0.0056, "step": 4666 }, { "epoch": 0.34, "learning_rate": 0.0001314386428728795, "loss": 0.0058, "step": 4668 }, { "epoch": 0.34, "learning_rate": 0.00013140926782698097, "loss": 0.0056, "step": 4670 }, { "epoch": 0.34, "learning_rate": 0.00013137989278108248, "loss": 0.0057, "step": 4672 }, { "epoch": 0.34, "learning_rate": 0.00013135051773518396, "loss": 0.0056, "step": 4674 }, { "epoch": 0.34, "learning_rate": 0.00013132114268928546, "loss": 0.0056, "step": 4676 }, { "epoch": 0.34, "learning_rate": 0.00013129176764338697, "loss": 0.0055, "step": 4678 }, { "epoch": 0.34, "learning_rate": 0.00013126239259748845, "loss": 0.0056, "step": 4680 }, { "epoch": 0.34, "learning_rate": 0.00013123301755158992, "loss": 0.0054, "step": 4682 }, { "epoch": 0.34, "learning_rate": 0.0001312036425056914, "loss": 0.0055, "step": 4684 }, { "epoch": 0.34, "learning_rate": 0.0001311742674597929, "loss": 0.0056, "step": 4686 }, { "epoch": 0.34, "learning_rate": 0.0001311448924138944, "loss": 0.0057, "step": 4688 }, { "epoch": 0.34, "learning_rate": 0.0001311155173679959, "loss": 0.0055, "step": 4690 }, { "epoch": 0.34, "learning_rate": 0.0001310861423220974, "loss": 0.0056, "step": 4692 }, { "epoch": 0.34, "learning_rate": 0.00013105676727619888, "loss": 0.0056, "step": 4694 }, { "epoch": 0.34, "learning_rate": 0.00013102739223030035, "loss": 0.0054, "step": 4696 }, { "epoch": 0.34, "learning_rate": 0.00013099801718440186, "loss": 0.0056, "step": 4698 }, { "epoch": 0.35, "learning_rate": 0.00013096864213850334, "loss": 0.0055, "step": 4700 }, { "epoch": 0.35, "learning_rate": 0.00013093926709260484, "loss": 0.0055, "step": 4702 }, { "epoch": 0.35, "learning_rate": 0.00013090989204670635, "loss": 0.0056, "step": 4704 }, { "epoch": 0.35, "learning_rate": 0.0001308805170008078, "loss": 0.0056, "step": 4706 }, { "epoch": 0.35, "learning_rate": 0.0001308511419549093, "loss": 0.0054, "step": 4708 }, { "epoch": 0.35, "learning_rate": 0.0001308217669090108, "loss": 0.0057, "step": 4710 }, { "epoch": 0.35, "learning_rate": 0.0001307923918631123, "loss": 0.0056, "step": 4712 }, { "epoch": 0.35, "learning_rate": 0.0001307630168172138, "loss": 0.0055, "step": 4714 }, { "epoch": 0.35, "learning_rate": 0.00013073364177131527, "loss": 0.0058, "step": 4716 }, { "epoch": 0.35, "learning_rate": 0.00013070426672541675, "loss": 0.0056, "step": 4718 }, { "epoch": 0.35, "learning_rate": 0.00013067489167951826, "loss": 0.0055, "step": 4720 }, { "epoch": 0.35, "learning_rate": 0.00013064551663361974, "loss": 0.0057, "step": 4722 }, { "epoch": 0.35, "learning_rate": 0.00013061614158772124, "loss": 0.0054, "step": 4724 }, { "epoch": 0.35, "learning_rate": 0.00013058676654182275, "loss": 0.0054, "step": 4726 }, { "epoch": 0.35, "learning_rate": 0.0001305573914959242, "loss": 0.0053, "step": 4728 }, { "epoch": 0.35, "learning_rate": 0.0001305280164500257, "loss": 0.0055, "step": 4730 }, { "epoch": 0.35, "learning_rate": 0.0001304986414041272, "loss": 0.0055, "step": 4732 }, { "epoch": 0.35, "learning_rate": 0.0001304692663582287, "loss": 0.0053, "step": 4734 }, { "epoch": 0.35, "learning_rate": 0.0001304398913123302, "loss": 0.0055, "step": 4736 }, { "epoch": 0.35, "learning_rate": 0.00013041051626643167, "loss": 0.0056, "step": 4738 }, { "epoch": 0.35, "learning_rate": 0.00013038114122053315, "loss": 0.0057, "step": 4740 }, { "epoch": 0.35, "learning_rate": 0.00013035176617463465, "loss": 0.0059, "step": 4742 }, { "epoch": 0.35, "learning_rate": 0.00013032239112873613, "loss": 0.0055, "step": 4744 }, { "epoch": 0.35, "learning_rate": 0.00013029301608283764, "loss": 0.0055, "step": 4746 }, { "epoch": 0.35, "learning_rate": 0.00013026364103693914, "loss": 0.0057, "step": 4748 }, { "epoch": 0.35, "learning_rate": 0.00013023426599104062, "loss": 0.0057, "step": 4750 }, { "epoch": 0.35, "learning_rate": 0.0001302048909451421, "loss": 0.0054, "step": 4752 }, { "epoch": 0.35, "learning_rate": 0.0001301755158992436, "loss": 0.0057, "step": 4754 }, { "epoch": 0.35, "learning_rate": 0.00013014614085334508, "loss": 0.0055, "step": 4756 }, { "epoch": 0.35, "learning_rate": 0.0001301167658074466, "loss": 0.0055, "step": 4758 }, { "epoch": 0.35, "learning_rate": 0.00013008739076154807, "loss": 0.0056, "step": 4760 }, { "epoch": 0.35, "learning_rate": 0.00013005801571564957, "loss": 0.0056, "step": 4762 }, { "epoch": 0.35, "learning_rate": 0.00013002864066975105, "loss": 0.0057, "step": 4764 }, { "epoch": 0.35, "learning_rate": 0.00012999926562385253, "loss": 0.0055, "step": 4766 }, { "epoch": 0.35, "learning_rate": 0.00012996989057795404, "loss": 0.0057, "step": 4768 }, { "epoch": 0.35, "learning_rate": 0.00012994051553205554, "loss": 0.0056, "step": 4770 }, { "epoch": 0.35, "learning_rate": 0.00012991114048615702, "loss": 0.0056, "step": 4772 }, { "epoch": 0.35, "learning_rate": 0.0001298817654402585, "loss": 0.0054, "step": 4774 }, { "epoch": 0.35, "learning_rate": 0.00012985239039436, "loss": 0.0056, "step": 4776 }, { "epoch": 0.35, "learning_rate": 0.00012982301534846148, "loss": 0.0054, "step": 4778 }, { "epoch": 0.35, "learning_rate": 0.000129793640302563, "loss": 0.0055, "step": 4780 }, { "epoch": 0.35, "learning_rate": 0.00012976426525666447, "loss": 0.0057, "step": 4782 }, { "epoch": 0.35, "learning_rate": 0.00012973489021076597, "loss": 0.0057, "step": 4784 }, { "epoch": 0.35, "learning_rate": 0.00012970551516486745, "loss": 0.0057, "step": 4786 }, { "epoch": 0.35, "learning_rate": 0.00012967614011896893, "loss": 0.0056, "step": 4788 }, { "epoch": 0.35, "learning_rate": 0.00012964676507307043, "loss": 0.0057, "step": 4790 }, { "epoch": 0.35, "learning_rate": 0.00012961739002717194, "loss": 0.0055, "step": 4792 }, { "epoch": 0.35, "learning_rate": 0.00012958801498127342, "loss": 0.0056, "step": 4794 }, { "epoch": 0.35, "learning_rate": 0.00012955863993537492, "loss": 0.0055, "step": 4796 }, { "epoch": 0.35, "learning_rate": 0.00012952926488947637, "loss": 0.0057, "step": 4798 }, { "epoch": 0.35, "learning_rate": 0.00012949988984357788, "loss": 0.0056, "step": 4800 }, { "epoch": 0.35, "learning_rate": 0.00012947051479767938, "loss": 0.0057, "step": 4802 }, { "epoch": 0.35, "learning_rate": 0.00012944113975178086, "loss": 0.0056, "step": 4804 }, { "epoch": 0.35, "learning_rate": 0.00012941176470588237, "loss": 0.0055, "step": 4806 }, { "epoch": 0.35, "learning_rate": 0.00012938238965998385, "loss": 0.0055, "step": 4808 }, { "epoch": 0.35, "learning_rate": 0.00012935301461408533, "loss": 0.0056, "step": 4810 }, { "epoch": 0.35, "learning_rate": 0.00012932363956818683, "loss": 0.0057, "step": 4812 }, { "epoch": 0.35, "learning_rate": 0.0001292942645222883, "loss": 0.0057, "step": 4814 }, { "epoch": 0.35, "learning_rate": 0.00012926488947638981, "loss": 0.0057, "step": 4816 }, { "epoch": 0.35, "learning_rate": 0.00012923551443049132, "loss": 0.0054, "step": 4818 }, { "epoch": 0.35, "learning_rate": 0.00012920613938459277, "loss": 0.0055, "step": 4820 }, { "epoch": 0.35, "learning_rate": 0.00012917676433869428, "loss": 0.0054, "step": 4822 }, { "epoch": 0.35, "learning_rate": 0.00012914738929279578, "loss": 0.0056, "step": 4824 }, { "epoch": 0.35, "learning_rate": 0.00012911801424689726, "loss": 0.0055, "step": 4826 }, { "epoch": 0.35, "learning_rate": 0.00012908863920099877, "loss": 0.0056, "step": 4828 }, { "epoch": 0.35, "learning_rate": 0.00012905926415510024, "loss": 0.0054, "step": 4830 }, { "epoch": 0.35, "learning_rate": 0.00012902988910920172, "loss": 0.0055, "step": 4832 }, { "epoch": 0.35, "learning_rate": 0.00012900051406330323, "loss": 0.0056, "step": 4834 }, { "epoch": 0.36, "learning_rate": 0.0001289711390174047, "loss": 0.0056, "step": 4836 }, { "epoch": 0.36, "learning_rate": 0.0001289417639715062, "loss": 0.0055, "step": 4838 }, { "epoch": 0.36, "learning_rate": 0.00012891238892560772, "loss": 0.0054, "step": 4840 }, { "epoch": 0.36, "learning_rate": 0.0001288830138797092, "loss": 0.0055, "step": 4842 }, { "epoch": 0.36, "learning_rate": 0.00012885363883381067, "loss": 0.0058, "step": 4844 }, { "epoch": 0.36, "learning_rate": 0.00012882426378791218, "loss": 0.0055, "step": 4846 }, { "epoch": 0.36, "learning_rate": 0.00012879488874201366, "loss": 0.0056, "step": 4848 }, { "epoch": 0.36, "learning_rate": 0.00012876551369611516, "loss": 0.0064, "step": 4850 }, { "epoch": 0.36, "learning_rate": 0.00012873613865021664, "loss": 0.0055, "step": 4852 }, { "epoch": 0.36, "learning_rate": 0.00012870676360431815, "loss": 0.0056, "step": 4854 }, { "epoch": 0.36, "learning_rate": 0.00012867738855841963, "loss": 0.0055, "step": 4856 }, { "epoch": 0.36, "learning_rate": 0.0001286480135125211, "loss": 0.0058, "step": 4858 }, { "epoch": 0.36, "learning_rate": 0.0001286186384666226, "loss": 0.0058, "step": 4860 }, { "epoch": 0.36, "learning_rate": 0.00012858926342072412, "loss": 0.0056, "step": 4862 }, { "epoch": 0.36, "learning_rate": 0.0001285598883748256, "loss": 0.0057, "step": 4864 }, { "epoch": 0.36, "learning_rate": 0.0001285305133289271, "loss": 0.0067, "step": 4866 }, { "epoch": 0.36, "learning_rate": 0.00012850113828302858, "loss": 0.0063, "step": 4868 }, { "epoch": 0.36, "learning_rate": 0.00012847176323713006, "loss": 0.0061, "step": 4870 }, { "epoch": 0.36, "learning_rate": 0.00012844238819123156, "loss": 0.0061, "step": 4872 }, { "epoch": 0.36, "learning_rate": 0.00012841301314533304, "loss": 0.006, "step": 4874 }, { "epoch": 0.36, "learning_rate": 0.00012838363809943455, "loss": 0.0075, "step": 4876 }, { "epoch": 0.36, "learning_rate": 0.00012835426305353602, "loss": 0.0074, "step": 4878 }, { "epoch": 0.36, "learning_rate": 0.0001283248880076375, "loss": 0.0072, "step": 4880 }, { "epoch": 0.36, "learning_rate": 0.000128295512961739, "loss": 0.0068, "step": 4882 }, { "epoch": 0.36, "learning_rate": 0.0001282661379158405, "loss": 0.0089, "step": 4884 }, { "epoch": 0.36, "learning_rate": 0.000128236762869942, "loss": 0.0105, "step": 4886 }, { "epoch": 0.36, "learning_rate": 0.0001282073878240435, "loss": 0.0126, "step": 4888 }, { "epoch": 0.36, "learning_rate": 0.00012817801277814498, "loss": 0.011, "step": 4890 }, { "epoch": 0.36, "learning_rate": 0.00012814863773224645, "loss": 0.0124, "step": 4892 }, { "epoch": 0.36, "learning_rate": 0.00012811926268634796, "loss": 0.0154, "step": 4894 }, { "epoch": 0.36, "learning_rate": 0.00012808988764044944, "loss": 0.0205, "step": 4896 }, { "epoch": 0.36, "learning_rate": 0.00012806051259455094, "loss": 0.0245, "step": 4898 }, { "epoch": 0.36, "learning_rate": 0.00012803113754865245, "loss": 0.028, "step": 4900 }, { "epoch": 0.36, "learning_rate": 0.0001280017625027539, "loss": 0.0303, "step": 4902 }, { "epoch": 0.36, "learning_rate": 0.0001279723874568554, "loss": 0.0386, "step": 4904 }, { "epoch": 0.36, "learning_rate": 0.0001279430124109569, "loss": 0.0616, "step": 4906 }, { "epoch": 0.36, "learning_rate": 0.0001279136373650584, "loss": 0.2214, "step": 4908 }, { "epoch": 0.36, "learning_rate": 0.0001278842623191599, "loss": 0.2964, "step": 4910 }, { "epoch": 0.36, "learning_rate": 0.00012785488727326137, "loss": 0.3183, "step": 4912 }, { "epoch": 0.36, "learning_rate": 0.00012782551222736285, "loss": 0.3412, "step": 4914 }, { "epoch": 0.36, "learning_rate": 0.00012779613718146436, "loss": 0.389, "step": 4916 }, { "epoch": 0.36, "learning_rate": 0.00012776676213556583, "loss": 0.2744, "step": 4918 }, { "epoch": 0.36, "learning_rate": 0.00012773738708966734, "loss": 0.2417, "step": 4920 }, { "epoch": 0.36, "learning_rate": 0.00012770801204376882, "loss": 0.2177, "step": 4922 }, { "epoch": 0.36, "learning_rate": 0.0001276786369978703, "loss": 0.211, "step": 4924 }, { "epoch": 0.36, "learning_rate": 0.0001276492619519718, "loss": 0.2135, "step": 4926 }, { "epoch": 0.36, "learning_rate": 0.00012761988690607328, "loss": 0.1921, "step": 4928 }, { "epoch": 0.36, "learning_rate": 0.00012759051186017479, "loss": 0.1996, "step": 4930 }, { "epoch": 0.36, "learning_rate": 0.0001275611368142763, "loss": 0.1662, "step": 4932 }, { "epoch": 0.36, "learning_rate": 0.00012753176176837777, "loss": 0.1689, "step": 4934 }, { "epoch": 0.36, "learning_rate": 0.00012750238672247925, "loss": 0.1426, "step": 4936 }, { "epoch": 0.36, "learning_rate": 0.00012747301167658075, "loss": 0.1422, "step": 4938 }, { "epoch": 0.36, "learning_rate": 0.00012744363663068223, "loss": 0.1194, "step": 4940 }, { "epoch": 0.36, "learning_rate": 0.00012741426158478374, "loss": 0.104, "step": 4942 }, { "epoch": 0.36, "learning_rate": 0.00012738488653888522, "loss": 0.1357, "step": 4944 }, { "epoch": 0.36, "learning_rate": 0.00012735551149298672, "loss": 0.0774, "step": 4946 }, { "epoch": 0.36, "learning_rate": 0.0001273261364470882, "loss": 0.0894, "step": 4948 }, { "epoch": 0.36, "learning_rate": 0.00012729676140118968, "loss": 0.0769, "step": 4950 }, { "epoch": 0.36, "learning_rate": 0.00012726738635529118, "loss": 0.0793, "step": 4952 }, { "epoch": 0.36, "learning_rate": 0.0001272380113093927, "loss": 0.062, "step": 4954 }, { "epoch": 0.36, "learning_rate": 0.00012720863626349417, "loss": 0.0807, "step": 4956 }, { "epoch": 0.36, "learning_rate": 0.00012717926121759567, "loss": 0.067, "step": 4958 }, { "epoch": 0.36, "learning_rate": 0.00012714988617169715, "loss": 0.0661, "step": 4960 }, { "epoch": 0.36, "learning_rate": 0.00012712051112579863, "loss": 0.0765, "step": 4962 }, { "epoch": 0.36, "learning_rate": 0.00012709113607990014, "loss": 0.0399, "step": 4964 }, { "epoch": 0.36, "learning_rate": 0.00012706176103400161, "loss": 0.0586, "step": 4966 }, { "epoch": 0.36, "learning_rate": 0.00012703238598810312, "loss": 0.0496, "step": 4968 }, { "epoch": 0.36, "learning_rate": 0.00012700301094220462, "loss": 0.0719, "step": 4970 }, { "epoch": 0.37, "learning_rate": 0.00012697363589630608, "loss": 0.0277, "step": 4972 }, { "epoch": 0.37, "learning_rate": 0.00012694426085040758, "loss": 0.043, "step": 4974 }, { "epoch": 0.37, "learning_rate": 0.0001269148858045091, "loss": 0.0408, "step": 4976 }, { "epoch": 0.37, "learning_rate": 0.00012688551075861057, "loss": 0.0214, "step": 4978 }, { "epoch": 0.37, "learning_rate": 0.00012685613571271207, "loss": 0.0336, "step": 4980 }, { "epoch": 0.37, "learning_rate": 0.00012682676066681355, "loss": 0.0299, "step": 4982 }, { "epoch": 0.37, "learning_rate": 0.00012679738562091503, "loss": 0.0215, "step": 4984 }, { "epoch": 0.37, "learning_rate": 0.00012676801057501653, "loss": 0.0246, "step": 4986 }, { "epoch": 0.37, "learning_rate": 0.000126738635529118, "loss": 0.023, "step": 4988 }, { "epoch": 0.37, "learning_rate": 0.00012670926048321952, "loss": 0.0205, "step": 4990 }, { "epoch": 0.37, "learning_rate": 0.00012667988543732102, "loss": 0.0159, "step": 4992 }, { "epoch": 0.37, "learning_rate": 0.00012665051039142247, "loss": 0.0298, "step": 4994 }, { "epoch": 0.37, "learning_rate": 0.00012662113534552398, "loss": 0.0251, "step": 4996 }, { "epoch": 0.37, "learning_rate": 0.00012659176029962548, "loss": 0.0123, "step": 4998 }, { "epoch": 0.37, "learning_rate": 0.00012656238525372696, "loss": 0.0184, "step": 5000 }, { "epoch": 0.37, "learning_rate": 0.00012653301020782847, "loss": 0.0128, "step": 5002 }, { "epoch": 0.37, "learning_rate": 0.00012650363516192995, "loss": 0.0112, "step": 5004 }, { "epoch": 0.37, "learning_rate": 0.00012647426011603143, "loss": 0.0143, "step": 5006 }, { "epoch": 0.37, "learning_rate": 0.00012644488507013293, "loss": 0.0157, "step": 5008 }, { "epoch": 0.37, "learning_rate": 0.0001264155100242344, "loss": 0.0136, "step": 5010 }, { "epoch": 0.37, "learning_rate": 0.00012638613497833591, "loss": 0.0131, "step": 5012 }, { "epoch": 0.37, "learning_rate": 0.00012635675993243742, "loss": 0.0127, "step": 5014 }, { "epoch": 0.37, "learning_rate": 0.0001263273848865389, "loss": 0.0116, "step": 5016 }, { "epoch": 0.37, "learning_rate": 0.00012629800984064038, "loss": 0.0101, "step": 5018 }, { "epoch": 0.37, "learning_rate": 0.00012626863479474188, "loss": 0.0127, "step": 5020 }, { "epoch": 0.37, "learning_rate": 0.00012623925974884336, "loss": 0.0114, "step": 5022 }, { "epoch": 0.37, "learning_rate": 0.00012620988470294487, "loss": 0.0107, "step": 5024 }, { "epoch": 0.37, "learning_rate": 0.00012618050965704634, "loss": 0.0089, "step": 5026 }, { "epoch": 0.37, "learning_rate": 0.00012615113461114782, "loss": 0.0073, "step": 5028 }, { "epoch": 0.37, "learning_rate": 0.00012612175956524933, "loss": 0.011, "step": 5030 }, { "epoch": 0.37, "learning_rate": 0.0001260923845193508, "loss": 0.0079, "step": 5032 }, { "epoch": 0.37, "learning_rate": 0.0001260630094734523, "loss": 0.0104, "step": 5034 }, { "epoch": 0.37, "learning_rate": 0.0001260336344275538, "loss": 0.0075, "step": 5036 }, { "epoch": 0.37, "learning_rate": 0.0001260042593816553, "loss": 0.007, "step": 5038 }, { "epoch": 0.37, "learning_rate": 0.00012597488433575677, "loss": 0.0071, "step": 5040 }, { "epoch": 0.37, "learning_rate": 0.00012594550928985825, "loss": 0.0085, "step": 5042 }, { "epoch": 0.37, "learning_rate": 0.00012591613424395976, "loss": 0.009, "step": 5044 }, { "epoch": 0.37, "learning_rate": 0.00012588675919806126, "loss": 0.0083, "step": 5046 }, { "epoch": 0.37, "learning_rate": 0.00012585738415216274, "loss": 0.0076, "step": 5048 }, { "epoch": 0.37, "learning_rate": 0.00012582800910626425, "loss": 0.0073, "step": 5050 }, { "epoch": 0.37, "learning_rate": 0.00012579863406036573, "loss": 0.0066, "step": 5052 }, { "epoch": 0.37, "learning_rate": 0.0001257692590144672, "loss": 0.0068, "step": 5054 }, { "epoch": 0.37, "learning_rate": 0.0001257398839685687, "loss": 0.0064, "step": 5056 }, { "epoch": 0.37, "learning_rate": 0.0001257105089226702, "loss": 0.0068, "step": 5058 }, { "epoch": 0.37, "learning_rate": 0.0001256811338767717, "loss": 0.0065, "step": 5060 }, { "epoch": 0.37, "learning_rate": 0.0001256517588308732, "loss": 0.0084, "step": 5062 }, { "epoch": 0.37, "learning_rate": 0.00012562238378497465, "loss": 0.0068, "step": 5064 }, { "epoch": 0.37, "learning_rate": 0.00012559300873907616, "loss": 0.0075, "step": 5066 }, { "epoch": 0.37, "learning_rate": 0.00012556363369317766, "loss": 0.0061, "step": 5068 }, { "epoch": 0.37, "learning_rate": 0.00012553425864727914, "loss": 0.006, "step": 5070 }, { "epoch": 0.37, "learning_rate": 0.00012550488360138064, "loss": 0.006, "step": 5072 }, { "epoch": 0.37, "learning_rate": 0.00012547550855548212, "loss": 0.0061, "step": 5074 }, { "epoch": 0.37, "learning_rate": 0.0001254461335095836, "loss": 0.0062, "step": 5076 }, { "epoch": 0.37, "learning_rate": 0.0001254167584636851, "loss": 0.0061, "step": 5078 }, { "epoch": 0.37, "learning_rate": 0.00012538738341778659, "loss": 0.0062, "step": 5080 }, { "epoch": 0.37, "learning_rate": 0.0001253580083718881, "loss": 0.0069, "step": 5082 }, { "epoch": 0.37, "learning_rate": 0.0001253286333259896, "loss": 0.0061, "step": 5084 }, { "epoch": 0.37, "learning_rate": 0.00012529925828009105, "loss": 0.0072, "step": 5086 }, { "epoch": 0.37, "learning_rate": 0.00012526988323419255, "loss": 0.0069, "step": 5088 }, { "epoch": 0.37, "learning_rate": 0.00012524050818829406, "loss": 0.0063, "step": 5090 }, { "epoch": 0.37, "learning_rate": 0.00012521113314239554, "loss": 0.006, "step": 5092 }, { "epoch": 0.37, "learning_rate": 0.00012518175809649704, "loss": 0.0057, "step": 5094 }, { "epoch": 0.37, "learning_rate": 0.00012515238305059852, "loss": 0.0058, "step": 5096 }, { "epoch": 0.37, "learning_rate": 0.0001251230080047, "loss": 0.0059, "step": 5098 }, { "epoch": 0.37, "learning_rate": 0.0001250936329588015, "loss": 0.0061, "step": 5100 }, { "epoch": 0.37, "learning_rate": 0.00012506425791290298, "loss": 0.0061, "step": 5102 }, { "epoch": 0.37, "learning_rate": 0.0001250348828670045, "loss": 0.0057, "step": 5104 }, { "epoch": 0.37, "learning_rate": 0.000125005507821106, "loss": 0.0055, "step": 5106 }, { "epoch": 0.38, "learning_rate": 0.00012497613277520747, "loss": 0.006, "step": 5108 }, { "epoch": 0.38, "learning_rate": 0.00012494675772930895, "loss": 0.0058, "step": 5110 }, { "epoch": 0.38, "learning_rate": 0.00012491738268341046, "loss": 0.0068, "step": 5112 }, { "epoch": 0.38, "learning_rate": 0.00012488800763751193, "loss": 0.0058, "step": 5114 }, { "epoch": 0.38, "learning_rate": 0.00012485863259161344, "loss": 0.0074, "step": 5116 }, { "epoch": 0.38, "learning_rate": 0.00012482925754571492, "loss": 0.0058, "step": 5118 }, { "epoch": 0.38, "learning_rate": 0.00012479988249981642, "loss": 0.0057, "step": 5120 }, { "epoch": 0.38, "learning_rate": 0.0001247705074539179, "loss": 0.0059, "step": 5122 }, { "epoch": 0.38, "learning_rate": 0.00012474113240801938, "loss": 0.0057, "step": 5124 }, { "epoch": 0.38, "learning_rate": 0.00012471175736212089, "loss": 0.006, "step": 5126 }, { "epoch": 0.38, "learning_rate": 0.0001246823823162224, "loss": 0.0056, "step": 5128 }, { "epoch": 0.38, "learning_rate": 0.00012465300727032387, "loss": 0.0056, "step": 5130 }, { "epoch": 0.38, "learning_rate": 0.00012462363222442535, "loss": 0.0057, "step": 5132 }, { "epoch": 0.38, "learning_rate": 0.00012459425717852685, "loss": 0.0057, "step": 5134 }, { "epoch": 0.38, "learning_rate": 0.00012456488213262833, "loss": 0.0059, "step": 5136 }, { "epoch": 0.38, "learning_rate": 0.00012453550708672984, "loss": 0.0056, "step": 5138 }, { "epoch": 0.38, "learning_rate": 0.00012450613204083132, "loss": 0.0057, "step": 5140 }, { "epoch": 0.38, "learning_rate": 0.00012447675699493282, "loss": 0.0055, "step": 5142 }, { "epoch": 0.38, "learning_rate": 0.0001244473819490343, "loss": 0.0058, "step": 5144 }, { "epoch": 0.38, "learning_rate": 0.00012441800690313578, "loss": 0.0059, "step": 5146 }, { "epoch": 0.38, "learning_rate": 0.00012438863185723728, "loss": 0.0059, "step": 5148 }, { "epoch": 0.38, "learning_rate": 0.0001243592568113388, "loss": 0.0056, "step": 5150 }, { "epoch": 0.38, "learning_rate": 0.00012432988176544027, "loss": 0.0057, "step": 5152 }, { "epoch": 0.38, "learning_rate": 0.00012430050671954177, "loss": 0.0058, "step": 5154 }, { "epoch": 0.38, "learning_rate": 0.00012427113167364322, "loss": 0.0056, "step": 5156 }, { "epoch": 0.38, "learning_rate": 0.00012424175662774473, "loss": 0.0056, "step": 5158 }, { "epoch": 0.38, "learning_rate": 0.00012421238158184624, "loss": 0.0056, "step": 5160 }, { "epoch": 0.38, "learning_rate": 0.00012418300653594771, "loss": 0.0062, "step": 5162 }, { "epoch": 0.38, "learning_rate": 0.00012415363149004922, "loss": 0.0056, "step": 5164 }, { "epoch": 0.38, "learning_rate": 0.0001241242564441507, "loss": 0.0056, "step": 5166 }, { "epoch": 0.38, "learning_rate": 0.00012409488139825218, "loss": 0.0055, "step": 5168 }, { "epoch": 0.38, "learning_rate": 0.00012406550635235368, "loss": 0.0054, "step": 5170 }, { "epoch": 0.38, "learning_rate": 0.00012403613130645516, "loss": 0.0057, "step": 5172 }, { "epoch": 0.38, "learning_rate": 0.00012400675626055667, "loss": 0.0057, "step": 5174 }, { "epoch": 0.38, "learning_rate": 0.00012397738121465817, "loss": 0.0057, "step": 5176 }, { "epoch": 0.38, "learning_rate": 0.00012394800616875962, "loss": 0.0057, "step": 5178 }, { "epoch": 0.38, "learning_rate": 0.00012391863112286113, "loss": 0.0057, "step": 5180 }, { "epoch": 0.38, "learning_rate": 0.00012388925607696263, "loss": 0.0053, "step": 5182 }, { "epoch": 0.38, "learning_rate": 0.0001238598810310641, "loss": 0.0054, "step": 5184 }, { "epoch": 0.38, "learning_rate": 0.00012383050598516562, "loss": 0.0057, "step": 5186 }, { "epoch": 0.38, "learning_rate": 0.0001238011309392671, "loss": 0.0058, "step": 5188 }, { "epoch": 0.38, "learning_rate": 0.00012377175589336857, "loss": 0.0056, "step": 5190 }, { "epoch": 0.38, "learning_rate": 0.00012374238084747008, "loss": 0.0057, "step": 5192 }, { "epoch": 0.38, "learning_rate": 0.00012371300580157156, "loss": 0.0057, "step": 5194 }, { "epoch": 0.38, "learning_rate": 0.00012368363075567306, "loss": 0.0057, "step": 5196 }, { "epoch": 0.38, "learning_rate": 0.00012365425570977457, "loss": 0.0056, "step": 5198 }, { "epoch": 0.38, "learning_rate": 0.00012362488066387605, "loss": 0.0056, "step": 5200 }, { "epoch": 0.38, "learning_rate": 0.00012359550561797752, "loss": 0.0056, "step": 5202 }, { "epoch": 0.38, "learning_rate": 0.00012356613057207903, "loss": 0.0056, "step": 5204 }, { "epoch": 0.38, "learning_rate": 0.0001235367555261805, "loss": 0.0058, "step": 5206 }, { "epoch": 0.38, "learning_rate": 0.00012350738048028201, "loss": 0.0056, "step": 5208 }, { "epoch": 0.38, "learning_rate": 0.0001234780054343835, "loss": 0.0056, "step": 5210 }, { "epoch": 0.38, "learning_rate": 0.000123448630388485, "loss": 0.0056, "step": 5212 }, { "epoch": 0.38, "learning_rate": 0.00012341925534258648, "loss": 0.0056, "step": 5214 }, { "epoch": 0.38, "learning_rate": 0.00012338988029668795, "loss": 0.0056, "step": 5216 }, { "epoch": 0.38, "learning_rate": 0.00012336050525078946, "loss": 0.0058, "step": 5218 }, { "epoch": 0.38, "learning_rate": 0.00012333113020489097, "loss": 0.0058, "step": 5220 }, { "epoch": 0.38, "learning_rate": 0.00012330175515899244, "loss": 0.0054, "step": 5222 }, { "epoch": 0.38, "learning_rate": 0.00012327238011309392, "loss": 0.0057, "step": 5224 }, { "epoch": 0.38, "learning_rate": 0.00012324300506719543, "loss": 0.0058, "step": 5226 }, { "epoch": 0.38, "learning_rate": 0.0001232136300212969, "loss": 0.0057, "step": 5228 }, { "epoch": 0.38, "learning_rate": 0.0001231842549753984, "loss": 0.0056, "step": 5230 }, { "epoch": 0.38, "learning_rate": 0.0001231548799294999, "loss": 0.0056, "step": 5232 }, { "epoch": 0.38, "learning_rate": 0.0001231255048836014, "loss": 0.0057, "step": 5234 }, { "epoch": 0.38, "learning_rate": 0.00012309612983770287, "loss": 0.0055, "step": 5236 }, { "epoch": 0.38, "learning_rate": 0.00012306675479180435, "loss": 0.0058, "step": 5238 }, { "epoch": 0.38, "learning_rate": 0.00012303737974590586, "loss": 0.0055, "step": 5240 }, { "epoch": 0.38, "learning_rate": 0.00012300800470000736, "loss": 0.0056, "step": 5242 }, { "epoch": 0.39, "learning_rate": 0.00012297862965410884, "loss": 0.0056, "step": 5244 }, { "epoch": 0.39, "learning_rate": 0.00012294925460821035, "loss": 0.0056, "step": 5246 }, { "epoch": 0.39, "learning_rate": 0.00012291987956231183, "loss": 0.0055, "step": 5248 }, { "epoch": 0.39, "learning_rate": 0.0001228905045164133, "loss": 0.0058, "step": 5250 }, { "epoch": 0.39, "learning_rate": 0.0001228611294705148, "loss": 0.0058, "step": 5252 }, { "epoch": 0.39, "learning_rate": 0.0001228317544246163, "loss": 0.0054, "step": 5254 }, { "epoch": 0.39, "learning_rate": 0.0001228023793787178, "loss": 0.0056, "step": 5256 }, { "epoch": 0.39, "learning_rate": 0.0001227730043328193, "loss": 0.0057, "step": 5258 }, { "epoch": 0.39, "learning_rate": 0.00012274362928692075, "loss": 0.0056, "step": 5260 }, { "epoch": 0.39, "learning_rate": 0.00012271425424102226, "loss": 0.0056, "step": 5262 }, { "epoch": 0.39, "learning_rate": 0.00012268487919512376, "loss": 0.0058, "step": 5264 }, { "epoch": 0.39, "learning_rate": 0.00012265550414922524, "loss": 0.0057, "step": 5266 }, { "epoch": 0.39, "learning_rate": 0.00012262612910332674, "loss": 0.0055, "step": 5268 }, { "epoch": 0.39, "learning_rate": 0.00012259675405742822, "loss": 0.0056, "step": 5270 }, { "epoch": 0.39, "learning_rate": 0.0001225673790115297, "loss": 0.0056, "step": 5272 }, { "epoch": 0.39, "learning_rate": 0.0001225380039656312, "loss": 0.0056, "step": 5274 }, { "epoch": 0.39, "learning_rate": 0.00012250862891973269, "loss": 0.0055, "step": 5276 }, { "epoch": 0.39, "learning_rate": 0.0001224792538738342, "loss": 0.0064, "step": 5278 }, { "epoch": 0.39, "learning_rate": 0.00012244987882793567, "loss": 0.006, "step": 5280 }, { "epoch": 0.39, "learning_rate": 0.00012242050378203715, "loss": 0.0058, "step": 5282 }, { "epoch": 0.39, "learning_rate": 0.00012239112873613865, "loss": 0.0056, "step": 5284 }, { "epoch": 0.39, "learning_rate": 0.00012236175369024013, "loss": 0.0059, "step": 5286 }, { "epoch": 0.39, "learning_rate": 0.00012233237864434164, "loss": 0.0054, "step": 5288 }, { "epoch": 0.39, "learning_rate": 0.00012230300359844314, "loss": 0.0056, "step": 5290 }, { "epoch": 0.39, "learning_rate": 0.00012227362855254462, "loss": 0.0057, "step": 5292 }, { "epoch": 0.39, "learning_rate": 0.0001222442535066461, "loss": 0.0055, "step": 5294 }, { "epoch": 0.39, "learning_rate": 0.0001222148784607476, "loss": 0.0056, "step": 5296 }, { "epoch": 0.39, "learning_rate": 0.00012218550341484908, "loss": 0.0057, "step": 5298 }, { "epoch": 0.39, "learning_rate": 0.0001221561283689506, "loss": 0.0055, "step": 5300 }, { "epoch": 0.39, "learning_rate": 0.00012212675332305207, "loss": 0.0055, "step": 5302 }, { "epoch": 0.39, "learning_rate": 0.00012209737827715357, "loss": 0.0058, "step": 5304 }, { "epoch": 0.39, "learning_rate": 0.00012206800323125505, "loss": 0.0055, "step": 5306 }, { "epoch": 0.39, "learning_rate": 0.00012203862818535654, "loss": 0.0055, "step": 5308 }, { "epoch": 0.39, "learning_rate": 0.00012200925313945803, "loss": 0.0057, "step": 5310 }, { "epoch": 0.39, "learning_rate": 0.00012197987809355953, "loss": 0.0056, "step": 5312 }, { "epoch": 0.39, "learning_rate": 0.00012195050304766102, "loss": 0.0057, "step": 5314 }, { "epoch": 0.39, "learning_rate": 0.00012192112800176252, "loss": 0.0056, "step": 5316 }, { "epoch": 0.39, "learning_rate": 0.00012189175295586399, "loss": 0.0054, "step": 5318 }, { "epoch": 0.39, "learning_rate": 0.00012186237790996548, "loss": 0.0053, "step": 5320 }, { "epoch": 0.39, "learning_rate": 0.00012183300286406699, "loss": 0.0055, "step": 5322 }, { "epoch": 0.39, "learning_rate": 0.00012180362781816848, "loss": 0.0054, "step": 5324 }, { "epoch": 0.39, "learning_rate": 0.00012177425277226997, "loss": 0.0058, "step": 5326 }, { "epoch": 0.39, "learning_rate": 0.00012174487772637145, "loss": 0.0058, "step": 5328 }, { "epoch": 0.39, "learning_rate": 0.00012171550268047294, "loss": 0.0053, "step": 5330 }, { "epoch": 0.39, "learning_rate": 0.00012168612763457443, "loss": 0.0055, "step": 5332 }, { "epoch": 0.39, "learning_rate": 0.00012165675258867592, "loss": 0.0056, "step": 5334 }, { "epoch": 0.39, "learning_rate": 0.00012162737754277742, "loss": 0.0055, "step": 5336 }, { "epoch": 0.39, "learning_rate": 0.00012159800249687892, "loss": 0.0058, "step": 5338 }, { "epoch": 0.39, "learning_rate": 0.00012156862745098039, "loss": 0.0056, "step": 5340 }, { "epoch": 0.39, "learning_rate": 0.00012153925240508188, "loss": 0.0054, "step": 5342 }, { "epoch": 0.39, "learning_rate": 0.00012150987735918338, "loss": 0.0056, "step": 5344 }, { "epoch": 0.39, "learning_rate": 0.00012148050231328488, "loss": 0.0057, "step": 5346 }, { "epoch": 0.39, "learning_rate": 0.00012145112726738637, "loss": 0.0055, "step": 5348 }, { "epoch": 0.39, "learning_rate": 0.00012142175222148786, "loss": 0.0057, "step": 5350 }, { "epoch": 0.39, "learning_rate": 0.00012139237717558934, "loss": 0.0057, "step": 5352 }, { "epoch": 0.39, "learning_rate": 0.00012136300212969083, "loss": 0.0055, "step": 5354 }, { "epoch": 0.39, "learning_rate": 0.00012133362708379232, "loss": 0.0056, "step": 5356 }, { "epoch": 0.39, "learning_rate": 0.00012130425203789381, "loss": 0.0057, "step": 5358 }, { "epoch": 0.39, "learning_rate": 0.00012127487699199532, "loss": 0.0056, "step": 5360 }, { "epoch": 0.39, "learning_rate": 0.00012124550194609681, "loss": 0.0055, "step": 5362 }, { "epoch": 0.39, "learning_rate": 0.00012121612690019828, "loss": 0.0054, "step": 5364 }, { "epoch": 0.39, "learning_rate": 0.00012118675185429977, "loss": 0.0054, "step": 5366 }, { "epoch": 0.39, "learning_rate": 0.00012115737680840127, "loss": 0.0055, "step": 5368 }, { "epoch": 0.39, "learning_rate": 0.00012112800176250276, "loss": 0.0057, "step": 5370 }, { "epoch": 0.39, "learning_rate": 0.00012109862671660426, "loss": 0.0057, "step": 5372 }, { "epoch": 0.39, "learning_rate": 0.00012106925167070575, "loss": 0.0057, "step": 5374 }, { "epoch": 0.39, "learning_rate": 0.00012103987662480723, "loss": 0.0057, "step": 5376 }, { "epoch": 0.39, "learning_rate": 0.00012101050157890872, "loss": 0.0056, "step": 5378 }, { "epoch": 0.4, "learning_rate": 0.00012098112653301021, "loss": 0.0055, "step": 5380 }, { "epoch": 0.4, "learning_rate": 0.0001209517514871117, "loss": 0.0056, "step": 5382 }, { "epoch": 0.4, "learning_rate": 0.00012092237644121321, "loss": 0.0057, "step": 5384 }, { "epoch": 0.4, "learning_rate": 0.00012089300139531467, "loss": 0.0056, "step": 5386 }, { "epoch": 0.4, "learning_rate": 0.00012086362634941616, "loss": 0.0054, "step": 5388 }, { "epoch": 0.4, "learning_rate": 0.00012083425130351767, "loss": 0.0056, "step": 5390 }, { "epoch": 0.4, "learning_rate": 0.00012080487625761916, "loss": 0.0055, "step": 5392 }, { "epoch": 0.4, "learning_rate": 0.00012077550121172065, "loss": 0.0055, "step": 5394 }, { "epoch": 0.4, "learning_rate": 0.00012074612616582215, "loss": 0.0056, "step": 5396 }, { "epoch": 0.4, "learning_rate": 0.00012071675111992362, "loss": 0.0055, "step": 5398 }, { "epoch": 0.4, "learning_rate": 0.00012068737607402512, "loss": 0.0054, "step": 5400 }, { "epoch": 0.4, "learning_rate": 0.00012065800102812661, "loss": 0.0055, "step": 5402 }, { "epoch": 0.4, "learning_rate": 0.0001206286259822281, "loss": 0.0055, "step": 5404 }, { "epoch": 0.4, "learning_rate": 0.0001205992509363296, "loss": 0.0058, "step": 5406 }, { "epoch": 0.4, "learning_rate": 0.0001205698758904311, "loss": 0.0054, "step": 5408 }, { "epoch": 0.4, "learning_rate": 0.00012054050084453256, "loss": 0.0055, "step": 5410 }, { "epoch": 0.4, "learning_rate": 0.00012051112579863407, "loss": 0.0056, "step": 5412 }, { "epoch": 0.4, "learning_rate": 0.00012048175075273556, "loss": 0.0057, "step": 5414 }, { "epoch": 0.4, "learning_rate": 0.00012045237570683705, "loss": 0.0057, "step": 5416 }, { "epoch": 0.4, "learning_rate": 0.00012042300066093854, "loss": 0.0054, "step": 5418 }, { "epoch": 0.4, "learning_rate": 0.00012039362561504004, "loss": 0.0053, "step": 5420 }, { "epoch": 0.4, "learning_rate": 0.00012036425056914151, "loss": 0.0055, "step": 5422 }, { "epoch": 0.4, "learning_rate": 0.000120334875523243, "loss": 0.0056, "step": 5424 }, { "epoch": 0.4, "learning_rate": 0.0001203055004773445, "loss": 0.0055, "step": 5426 }, { "epoch": 0.4, "learning_rate": 0.00012027612543144599, "loss": 0.0054, "step": 5428 }, { "epoch": 0.4, "learning_rate": 0.0001202467503855475, "loss": 0.0057, "step": 5430 }, { "epoch": 0.4, "learning_rate": 0.00012021737533964896, "loss": 0.0056, "step": 5432 }, { "epoch": 0.4, "learning_rate": 0.00012018800029375045, "loss": 0.0054, "step": 5434 }, { "epoch": 0.4, "learning_rate": 0.00012015862524785196, "loss": 0.0056, "step": 5436 }, { "epoch": 0.4, "learning_rate": 0.00012012925020195345, "loss": 0.0057, "step": 5438 }, { "epoch": 0.4, "learning_rate": 0.00012009987515605494, "loss": 0.0056, "step": 5440 }, { "epoch": 0.4, "learning_rate": 0.00012007050011015643, "loss": 0.0057, "step": 5442 }, { "epoch": 0.4, "learning_rate": 0.00012004112506425791, "loss": 0.0056, "step": 5444 }, { "epoch": 0.4, "learning_rate": 0.0001200117500183594, "loss": 0.0056, "step": 5446 }, { "epoch": 0.4, "learning_rate": 0.0001199823749724609, "loss": 0.0055, "step": 5448 }, { "epoch": 0.4, "learning_rate": 0.00011995299992656239, "loss": 0.0058, "step": 5450 }, { "epoch": 0.4, "learning_rate": 0.00011992362488066389, "loss": 0.0056, "step": 5452 }, { "epoch": 0.4, "learning_rate": 0.00011989424983476538, "loss": 0.0056, "step": 5454 }, { "epoch": 0.4, "learning_rate": 0.00011986487478886685, "loss": 0.0053, "step": 5456 }, { "epoch": 0.4, "learning_rate": 0.00011983549974296836, "loss": 0.0057, "step": 5458 }, { "epoch": 0.4, "learning_rate": 0.00011980612469706985, "loss": 0.0056, "step": 5460 }, { "epoch": 0.4, "learning_rate": 0.00011977674965117134, "loss": 0.0055, "step": 5462 }, { "epoch": 0.4, "learning_rate": 0.00011974737460527283, "loss": 0.0057, "step": 5464 }, { "epoch": 0.4, "learning_rate": 0.00011971799955937432, "loss": 0.0056, "step": 5466 }, { "epoch": 0.4, "learning_rate": 0.0001196886245134758, "loss": 0.0053, "step": 5468 }, { "epoch": 0.4, "learning_rate": 0.00011965924946757729, "loss": 0.0056, "step": 5470 }, { "epoch": 0.4, "learning_rate": 0.00011962987442167878, "loss": 0.0056, "step": 5472 }, { "epoch": 0.4, "learning_rate": 0.00011960049937578029, "loss": 0.0055, "step": 5474 }, { "epoch": 0.4, "learning_rate": 0.00011957112432988178, "loss": 0.0055, "step": 5476 }, { "epoch": 0.4, "learning_rate": 0.00011954174928398325, "loss": 0.0058, "step": 5478 }, { "epoch": 0.4, "learning_rate": 0.00011951237423808474, "loss": 0.0057, "step": 5480 }, { "epoch": 0.4, "learning_rate": 0.00011948299919218624, "loss": 0.0056, "step": 5482 }, { "epoch": 0.4, "learning_rate": 0.00011945362414628774, "loss": 0.0056, "step": 5484 }, { "epoch": 0.4, "learning_rate": 0.00011942424910038923, "loss": 0.0057, "step": 5486 }, { "epoch": 0.4, "learning_rate": 0.00011939487405449072, "loss": 0.0057, "step": 5488 }, { "epoch": 0.4, "learning_rate": 0.0001193654990085922, "loss": 0.0058, "step": 5490 }, { "epoch": 0.4, "learning_rate": 0.00011933612396269369, "loss": 0.0055, "step": 5492 }, { "epoch": 0.4, "learning_rate": 0.00011930674891679518, "loss": 0.0055, "step": 5494 }, { "epoch": 0.4, "learning_rate": 0.00011927737387089667, "loss": 0.0055, "step": 5496 }, { "epoch": 0.4, "learning_rate": 0.00011924799882499818, "loss": 0.0056, "step": 5498 }, { "epoch": 0.4, "learning_rate": 0.00011921862377909967, "loss": 0.0053, "step": 5500 }, { "epoch": 0.4, "learning_rate": 0.00011918924873320114, "loss": 0.0056, "step": 5502 }, { "epoch": 0.4, "learning_rate": 0.00011915987368730264, "loss": 0.0059, "step": 5504 }, { "epoch": 0.4, "learning_rate": 0.00011913049864140413, "loss": 0.0055, "step": 5506 }, { "epoch": 0.4, "learning_rate": 0.00011910112359550563, "loss": 0.0054, "step": 5508 }, { "epoch": 0.4, "learning_rate": 0.00011907174854960712, "loss": 0.0057, "step": 5510 }, { "epoch": 0.4, "learning_rate": 0.00011904237350370861, "loss": 0.0056, "step": 5512 }, { "epoch": 0.4, "learning_rate": 0.00011901299845781009, "loss": 0.0055, "step": 5514 }, { "epoch": 0.41, "learning_rate": 0.00011898362341191158, "loss": 0.0057, "step": 5516 }, { "epoch": 0.41, "learning_rate": 0.00011895424836601307, "loss": 0.0057, "step": 5518 }, { "epoch": 0.41, "learning_rate": 0.00011892487332011458, "loss": 0.0057, "step": 5520 }, { "epoch": 0.41, "learning_rate": 0.00011889549827421607, "loss": 0.0054, "step": 5522 }, { "epoch": 0.41, "learning_rate": 0.00011886612322831756, "loss": 0.0056, "step": 5524 }, { "epoch": 0.41, "learning_rate": 0.00011883674818241904, "loss": 0.0056, "step": 5526 }, { "epoch": 0.41, "learning_rate": 0.00011880737313652053, "loss": 0.0055, "step": 5528 }, { "epoch": 0.41, "learning_rate": 0.00011877799809062202, "loss": 0.0055, "step": 5530 }, { "epoch": 0.41, "learning_rate": 0.00011874862304472352, "loss": 0.0056, "step": 5532 }, { "epoch": 0.41, "learning_rate": 0.00011871924799882501, "loss": 0.0055, "step": 5534 }, { "epoch": 0.41, "learning_rate": 0.00011868987295292649, "loss": 0.0057, "step": 5536 }, { "epoch": 0.41, "learning_rate": 0.00011866049790702798, "loss": 0.0056, "step": 5538 }, { "epoch": 0.41, "learning_rate": 0.00011863112286112947, "loss": 0.0055, "step": 5540 }, { "epoch": 0.41, "learning_rate": 0.00011860174781523096, "loss": 0.0056, "step": 5542 }, { "epoch": 0.41, "learning_rate": 0.00011857237276933247, "loss": 0.0055, "step": 5544 }, { "epoch": 0.41, "learning_rate": 0.00011854299772343396, "loss": 0.0054, "step": 5546 }, { "epoch": 0.41, "learning_rate": 0.00011851362267753542, "loss": 0.0054, "step": 5548 }, { "epoch": 0.41, "learning_rate": 0.00011848424763163693, "loss": 0.0056, "step": 5550 }, { "epoch": 0.41, "learning_rate": 0.00011845487258573842, "loss": 0.0058, "step": 5552 }, { "epoch": 0.41, "learning_rate": 0.00011842549753983991, "loss": 0.0058, "step": 5554 }, { "epoch": 0.41, "learning_rate": 0.0001183961224939414, "loss": 0.0056, "step": 5556 }, { "epoch": 0.41, "learning_rate": 0.0001183667474480429, "loss": 0.0056, "step": 5558 }, { "epoch": 0.41, "learning_rate": 0.00011833737240214438, "loss": 0.0055, "step": 5560 }, { "epoch": 0.41, "learning_rate": 0.00011830799735624587, "loss": 0.0058, "step": 5562 }, { "epoch": 0.41, "learning_rate": 0.00011827862231034736, "loss": 0.0057, "step": 5564 }, { "epoch": 0.41, "learning_rate": 0.00011824924726444886, "loss": 0.0054, "step": 5566 }, { "epoch": 0.41, "learning_rate": 0.00011821987221855036, "loss": 0.0057, "step": 5568 }, { "epoch": 0.41, "learning_rate": 0.00011819049717265185, "loss": 0.0054, "step": 5570 }, { "epoch": 0.41, "learning_rate": 0.00011816112212675333, "loss": 0.0055, "step": 5572 }, { "epoch": 0.41, "learning_rate": 0.00011813174708085482, "loss": 0.0056, "step": 5574 }, { "epoch": 0.41, "learning_rate": 0.00011810237203495631, "loss": 0.0057, "step": 5576 }, { "epoch": 0.41, "learning_rate": 0.0001180729969890578, "loss": 0.0056, "step": 5578 }, { "epoch": 0.41, "learning_rate": 0.0001180436219431593, "loss": 0.0055, "step": 5580 }, { "epoch": 0.41, "learning_rate": 0.00011801424689726077, "loss": 0.0054, "step": 5582 }, { "epoch": 0.41, "learning_rate": 0.00011798487185136226, "loss": 0.0055, "step": 5584 }, { "epoch": 0.41, "learning_rate": 0.00011795549680546376, "loss": 0.0057, "step": 5586 }, { "epoch": 0.41, "learning_rate": 0.00011792612175956526, "loss": 0.0054, "step": 5588 }, { "epoch": 0.41, "learning_rate": 0.00011789674671366675, "loss": 0.0056, "step": 5590 }, { "epoch": 0.41, "learning_rate": 0.00011786737166776825, "loss": 0.0054, "step": 5592 }, { "epoch": 0.41, "learning_rate": 0.00011783799662186971, "loss": 0.0057, "step": 5594 }, { "epoch": 0.41, "learning_rate": 0.00011780862157597122, "loss": 0.0055, "step": 5596 }, { "epoch": 0.41, "learning_rate": 0.00011777924653007271, "loss": 0.0057, "step": 5598 }, { "epoch": 0.41, "learning_rate": 0.0001177498714841742, "loss": 0.0056, "step": 5600 }, { "epoch": 0.41, "learning_rate": 0.00011772049643827569, "loss": 0.0053, "step": 5602 }, { "epoch": 0.41, "learning_rate": 0.00011769112139237718, "loss": 0.0056, "step": 5604 }, { "epoch": 0.41, "learning_rate": 0.00011766174634647866, "loss": 0.0055, "step": 5606 }, { "epoch": 0.41, "learning_rate": 0.00011763237130058015, "loss": 0.0054, "step": 5608 }, { "epoch": 0.41, "learning_rate": 0.00011760299625468165, "loss": 0.0054, "step": 5610 }, { "epoch": 0.41, "learning_rate": 0.00011757362120878315, "loss": 0.0056, "step": 5612 }, { "epoch": 0.41, "learning_rate": 0.00011754424616288464, "loss": 0.0055, "step": 5614 }, { "epoch": 0.41, "learning_rate": 0.00011751487111698614, "loss": 0.0055, "step": 5616 }, { "epoch": 0.41, "learning_rate": 0.00011748549607108761, "loss": 0.0054, "step": 5618 }, { "epoch": 0.41, "learning_rate": 0.0001174561210251891, "loss": 0.0052, "step": 5620 }, { "epoch": 0.41, "learning_rate": 0.0001174267459792906, "loss": 0.0056, "step": 5622 }, { "epoch": 0.41, "learning_rate": 0.00011739737093339209, "loss": 0.0055, "step": 5624 }, { "epoch": 0.41, "learning_rate": 0.00011736799588749358, "loss": 0.0058, "step": 5626 }, { "epoch": 0.41, "learning_rate": 0.00011733862084159509, "loss": 0.0056, "step": 5628 }, { "epoch": 0.41, "learning_rate": 0.00011730924579569655, "loss": 0.0054, "step": 5630 }, { "epoch": 0.41, "learning_rate": 0.00011727987074979804, "loss": 0.0053, "step": 5632 }, { "epoch": 0.41, "learning_rate": 0.00011725049570389955, "loss": 0.0057, "step": 5634 }, { "epoch": 0.41, "learning_rate": 0.00011722112065800104, "loss": 0.0058, "step": 5636 }, { "epoch": 0.41, "learning_rate": 0.00011719174561210253, "loss": 0.0054, "step": 5638 }, { "epoch": 0.41, "learning_rate": 0.00011716237056620401, "loss": 0.0054, "step": 5640 }, { "epoch": 0.41, "learning_rate": 0.0001171329955203055, "loss": 0.0055, "step": 5642 }, { "epoch": 0.41, "learning_rate": 0.000117103620474407, "loss": 0.0055, "step": 5644 }, { "epoch": 0.41, "learning_rate": 0.00011707424542850849, "loss": 0.0057, "step": 5646 }, { "epoch": 0.41, "learning_rate": 0.00011704487038260998, "loss": 0.0055, "step": 5648 }, { "epoch": 0.41, "learning_rate": 0.00011701549533671148, "loss": 0.0056, "step": 5650 }, { "epoch": 0.42, "learning_rate": 0.00011698612029081295, "loss": 0.0055, "step": 5652 }, { "epoch": 0.42, "learning_rate": 0.00011695674524491444, "loss": 0.0054, "step": 5654 }, { "epoch": 0.42, "learning_rate": 0.00011692737019901593, "loss": 0.0053, "step": 5656 }, { "epoch": 0.42, "learning_rate": 0.00011689799515311744, "loss": 0.0053, "step": 5658 }, { "epoch": 0.42, "learning_rate": 0.00011686862010721893, "loss": 0.0055, "step": 5660 }, { "epoch": 0.42, "learning_rate": 0.00011683924506132042, "loss": 0.0055, "step": 5662 }, { "epoch": 0.42, "learning_rate": 0.0001168098700154219, "loss": 0.0056, "step": 5664 }, { "epoch": 0.42, "learning_rate": 0.00011678049496952339, "loss": 0.0056, "step": 5666 }, { "epoch": 0.42, "learning_rate": 0.00011675111992362488, "loss": 0.0055, "step": 5668 }, { "epoch": 0.42, "learning_rate": 0.00011672174487772638, "loss": 0.0054, "step": 5670 }, { "epoch": 0.42, "learning_rate": 0.00011669236983182787, "loss": 0.0057, "step": 5672 }, { "epoch": 0.42, "learning_rate": 0.00011666299478592937, "loss": 0.0052, "step": 5674 }, { "epoch": 0.42, "learning_rate": 0.00011663361974003084, "loss": 0.0055, "step": 5676 }, { "epoch": 0.42, "learning_rate": 0.00011660424469413233, "loss": 0.0056, "step": 5678 }, { "epoch": 0.42, "learning_rate": 0.00011657486964823384, "loss": 0.0055, "step": 5680 }, { "epoch": 0.42, "learning_rate": 0.00011654549460233533, "loss": 0.0057, "step": 5682 }, { "epoch": 0.42, "learning_rate": 0.00011651611955643682, "loss": 0.0055, "step": 5684 }, { "epoch": 0.42, "learning_rate": 0.0001164867445105383, "loss": 0.0057, "step": 5686 }, { "epoch": 0.42, "learning_rate": 0.00011645736946463979, "loss": 0.0055, "step": 5688 }, { "epoch": 0.42, "learning_rate": 0.00011642799441874128, "loss": 0.0056, "step": 5690 }, { "epoch": 0.42, "learning_rate": 0.00011639861937284277, "loss": 0.0056, "step": 5692 }, { "epoch": 0.42, "learning_rate": 0.00011636924432694427, "loss": 0.0057, "step": 5694 }, { "epoch": 0.42, "learning_rate": 0.00011633986928104577, "loss": 0.0057, "step": 5696 }, { "epoch": 0.42, "learning_rate": 0.00011631049423514724, "loss": 0.0057, "step": 5698 }, { "epoch": 0.42, "learning_rate": 0.00011628111918924873, "loss": 0.0056, "step": 5700 }, { "epoch": 0.42, "learning_rate": 0.00011625174414335023, "loss": 0.0055, "step": 5702 }, { "epoch": 0.42, "learning_rate": 0.00011622236909745173, "loss": 0.0056, "step": 5704 }, { "epoch": 0.42, "learning_rate": 0.00011619299405155322, "loss": 0.0054, "step": 5706 }, { "epoch": 0.42, "learning_rate": 0.00011616361900565471, "loss": 0.0056, "step": 5708 }, { "epoch": 0.42, "learning_rate": 0.00011613424395975619, "loss": 0.0055, "step": 5710 }, { "epoch": 0.42, "learning_rate": 0.00011610486891385768, "loss": 0.0057, "step": 5712 }, { "epoch": 0.42, "learning_rate": 0.00011607549386795917, "loss": 0.0056, "step": 5714 }, { "epoch": 0.42, "learning_rate": 0.00011604611882206066, "loss": 0.0055, "step": 5716 }, { "epoch": 0.42, "learning_rate": 0.00011601674377616216, "loss": 0.0053, "step": 5718 }, { "epoch": 0.42, "learning_rate": 0.00011598736873026366, "loss": 0.0053, "step": 5720 }, { "epoch": 0.42, "learning_rate": 0.00011595799368436513, "loss": 0.0055, "step": 5722 }, { "epoch": 0.42, "learning_rate": 0.00011592861863846662, "loss": 0.0056, "step": 5724 }, { "epoch": 0.42, "learning_rate": 0.00011589924359256812, "loss": 0.0056, "step": 5726 }, { "epoch": 0.42, "learning_rate": 0.00011586986854666962, "loss": 0.0055, "step": 5728 }, { "epoch": 0.42, "learning_rate": 0.00011584049350077111, "loss": 0.0056, "step": 5730 }, { "epoch": 0.42, "learning_rate": 0.00011581111845487259, "loss": 0.0056, "step": 5732 }, { "epoch": 0.42, "learning_rate": 0.00011578174340897408, "loss": 0.0057, "step": 5734 }, { "epoch": 0.42, "learning_rate": 0.00011575236836307557, "loss": 0.0054, "step": 5736 }, { "epoch": 0.42, "learning_rate": 0.00011572299331717706, "loss": 0.0057, "step": 5738 }, { "epoch": 0.42, "learning_rate": 0.00011569361827127855, "loss": 0.0056, "step": 5740 }, { "epoch": 0.42, "learning_rate": 0.00011566424322538006, "loss": 0.0054, "step": 5742 }, { "epoch": 0.42, "learning_rate": 0.00011563486817948152, "loss": 0.0055, "step": 5744 }, { "epoch": 0.42, "learning_rate": 0.00011560549313358302, "loss": 0.0055, "step": 5746 }, { "epoch": 0.42, "learning_rate": 0.00011557611808768452, "loss": 0.0055, "step": 5748 }, { "epoch": 0.42, "learning_rate": 0.00011554674304178601, "loss": 0.0057, "step": 5750 }, { "epoch": 0.42, "learning_rate": 0.0001155173679958875, "loss": 0.0055, "step": 5752 }, { "epoch": 0.42, "learning_rate": 0.000115487992949989, "loss": 0.0057, "step": 5754 }, { "epoch": 0.42, "learning_rate": 0.00011545861790409047, "loss": 0.0055, "step": 5756 }, { "epoch": 0.42, "learning_rate": 0.00011542924285819197, "loss": 0.0056, "step": 5758 }, { "epoch": 0.42, "learning_rate": 0.00011539986781229346, "loss": 0.0054, "step": 5760 }, { "epoch": 0.42, "learning_rate": 0.00011537049276639495, "loss": 0.0058, "step": 5762 }, { "epoch": 0.42, "learning_rate": 0.00011534111772049646, "loss": 0.0055, "step": 5764 }, { "epoch": 0.42, "learning_rate": 0.00011531174267459795, "loss": 0.0055, "step": 5766 }, { "epoch": 0.42, "learning_rate": 0.00011528236762869941, "loss": 0.0055, "step": 5768 }, { "epoch": 0.42, "learning_rate": 0.00011525299258280092, "loss": 0.0056, "step": 5770 }, { "epoch": 0.42, "learning_rate": 0.00011522361753690241, "loss": 0.0056, "step": 5772 }, { "epoch": 0.42, "learning_rate": 0.0001151942424910039, "loss": 0.0055, "step": 5774 }, { "epoch": 0.42, "learning_rate": 0.0001151648674451054, "loss": 0.0054, "step": 5776 }, { "epoch": 0.42, "learning_rate": 0.00011513549239920689, "loss": 0.0054, "step": 5778 }, { "epoch": 0.42, "learning_rate": 0.00011510611735330836, "loss": 0.0056, "step": 5780 }, { "epoch": 0.42, "learning_rate": 0.00011507674230740986, "loss": 0.0054, "step": 5782 }, { "epoch": 0.42, "learning_rate": 0.00011504736726151135, "loss": 0.0055, "step": 5784 }, { "epoch": 0.42, "learning_rate": 0.00011501799221561284, "loss": 0.0056, "step": 5786 }, { "epoch": 0.43, "learning_rate": 0.00011498861716971435, "loss": 0.0056, "step": 5788 }, { "epoch": 0.43, "learning_rate": 0.00011495924212381581, "loss": 0.0055, "step": 5790 }, { "epoch": 0.43, "learning_rate": 0.0001149298670779173, "loss": 0.0055, "step": 5792 }, { "epoch": 0.43, "learning_rate": 0.00011490049203201881, "loss": 0.0054, "step": 5794 }, { "epoch": 0.43, "learning_rate": 0.0001148711169861203, "loss": 0.0056, "step": 5796 }, { "epoch": 0.43, "learning_rate": 0.00011484174194022179, "loss": 0.0056, "step": 5798 }, { "epoch": 0.43, "learning_rate": 0.00011481236689432328, "loss": 0.0056, "step": 5800 }, { "epoch": 0.43, "learning_rate": 0.00011478299184842476, "loss": 0.0055, "step": 5802 }, { "epoch": 0.43, "learning_rate": 0.00011475361680252625, "loss": 0.0058, "step": 5804 }, { "epoch": 0.43, "learning_rate": 0.00011472424175662775, "loss": 0.0057, "step": 5806 }, { "epoch": 0.43, "learning_rate": 0.00011469486671072924, "loss": 0.0055, "step": 5808 }, { "epoch": 0.43, "learning_rate": 0.00011466549166483074, "loss": 0.0056, "step": 5810 }, { "epoch": 0.43, "learning_rate": 0.00011463611661893224, "loss": 0.0057, "step": 5812 }, { "epoch": 0.43, "learning_rate": 0.0001146067415730337, "loss": 0.0056, "step": 5814 }, { "epoch": 0.43, "learning_rate": 0.0001145773665271352, "loss": 0.0056, "step": 5816 }, { "epoch": 0.43, "learning_rate": 0.0001145479914812367, "loss": 0.0055, "step": 5818 }, { "epoch": 0.43, "learning_rate": 0.00011451861643533819, "loss": 0.0055, "step": 5820 }, { "epoch": 0.43, "learning_rate": 0.00011448924138943968, "loss": 0.0055, "step": 5822 }, { "epoch": 0.43, "learning_rate": 0.00011445986634354117, "loss": 0.0054, "step": 5824 }, { "epoch": 0.43, "learning_rate": 0.00011443049129764265, "loss": 0.0055, "step": 5826 }, { "epoch": 0.43, "learning_rate": 0.00011440111625174414, "loss": 0.0056, "step": 5828 }, { "epoch": 0.43, "learning_rate": 0.00011437174120584564, "loss": 0.0056, "step": 5830 }, { "epoch": 0.43, "learning_rate": 0.00011434236615994714, "loss": 0.0054, "step": 5832 }, { "epoch": 0.43, "learning_rate": 0.00011431299111404863, "loss": 0.0055, "step": 5834 }, { "epoch": 0.43, "learning_rate": 0.0001142836160681501, "loss": 0.0054, "step": 5836 }, { "epoch": 0.43, "learning_rate": 0.00011425424102225159, "loss": 0.0055, "step": 5838 }, { "epoch": 0.43, "learning_rate": 0.0001142248659763531, "loss": 0.0057, "step": 5840 }, { "epoch": 0.43, "learning_rate": 0.00011419549093045459, "loss": 0.0057, "step": 5842 }, { "epoch": 0.43, "learning_rate": 0.00011416611588455608, "loss": 0.0056, "step": 5844 }, { "epoch": 0.43, "learning_rate": 0.00011413674083865757, "loss": 0.0057, "step": 5846 }, { "epoch": 0.43, "learning_rate": 0.00011410736579275905, "loss": 0.0057, "step": 5848 }, { "epoch": 0.43, "learning_rate": 0.00011407799074686054, "loss": 0.0054, "step": 5850 }, { "epoch": 0.43, "learning_rate": 0.00011404861570096203, "loss": 0.0055, "step": 5852 }, { "epoch": 0.43, "learning_rate": 0.00011401924065506352, "loss": 0.0055, "step": 5854 }, { "epoch": 0.43, "learning_rate": 0.00011398986560916503, "loss": 0.0055, "step": 5856 }, { "epoch": 0.43, "learning_rate": 0.00011396049056326652, "loss": 0.0055, "step": 5858 }, { "epoch": 0.43, "learning_rate": 0.00011393111551736799, "loss": 0.0055, "step": 5860 }, { "epoch": 0.43, "learning_rate": 0.00011390174047146949, "loss": 0.0055, "step": 5862 }, { "epoch": 0.43, "learning_rate": 0.00011387236542557098, "loss": 0.0056, "step": 5864 }, { "epoch": 0.43, "learning_rate": 0.00011384299037967248, "loss": 0.0056, "step": 5866 }, { "epoch": 0.43, "learning_rate": 0.00011381361533377397, "loss": 0.0055, "step": 5868 }, { "epoch": 0.43, "learning_rate": 0.00011378424028787546, "loss": 0.0054, "step": 5870 }, { "epoch": 0.43, "learning_rate": 0.00011375486524197694, "loss": 0.0054, "step": 5872 }, { "epoch": 0.43, "learning_rate": 0.00011372549019607843, "loss": 0.0058, "step": 5874 }, { "epoch": 0.43, "learning_rate": 0.00011369611515017992, "loss": 0.0053, "step": 5876 }, { "epoch": 0.43, "learning_rate": 0.00011366674010428143, "loss": 0.0055, "step": 5878 }, { "epoch": 0.43, "learning_rate": 0.00011363736505838292, "loss": 0.0054, "step": 5880 }, { "epoch": 0.43, "learning_rate": 0.00011360799001248441, "loss": 0.0055, "step": 5882 }, { "epoch": 0.43, "learning_rate": 0.00011357861496658589, "loss": 0.0055, "step": 5884 }, { "epoch": 0.43, "learning_rate": 0.00011354923992068738, "loss": 0.0055, "step": 5886 }, { "epoch": 0.43, "learning_rate": 0.00011351986487478887, "loss": 0.0056, "step": 5888 }, { "epoch": 0.43, "learning_rate": 0.00011349048982889037, "loss": 0.0055, "step": 5890 }, { "epoch": 0.43, "learning_rate": 0.00011346111478299186, "loss": 0.0056, "step": 5892 }, { "epoch": 0.43, "learning_rate": 0.00011343173973709334, "loss": 0.0057, "step": 5894 }, { "epoch": 0.43, "learning_rate": 0.00011340236469119483, "loss": 0.0055, "step": 5896 }, { "epoch": 0.43, "learning_rate": 0.00011337298964529632, "loss": 0.0055, "step": 5898 }, { "epoch": 0.43, "learning_rate": 0.00011334361459939781, "loss": 0.0054, "step": 5900 }, { "epoch": 0.43, "learning_rate": 0.00011331423955349932, "loss": 0.0056, "step": 5902 }, { "epoch": 0.43, "learning_rate": 0.00011328486450760081, "loss": 0.0056, "step": 5904 }, { "epoch": 0.43, "learning_rate": 0.00011325548946170227, "loss": 0.0054, "step": 5906 }, { "epoch": 0.43, "learning_rate": 0.00011322611441580378, "loss": 0.0055, "step": 5908 }, { "epoch": 0.43, "learning_rate": 0.00011319673936990527, "loss": 0.0053, "step": 5910 }, { "epoch": 0.43, "learning_rate": 0.00011316736432400676, "loss": 0.0054, "step": 5912 }, { "epoch": 0.43, "learning_rate": 0.00011313798927810826, "loss": 0.0055, "step": 5914 }, { "epoch": 0.43, "learning_rate": 0.00011310861423220975, "loss": 0.0056, "step": 5916 }, { "epoch": 0.43, "learning_rate": 0.00011307923918631123, "loss": 0.0054, "step": 5918 }, { "epoch": 0.43, "learning_rate": 0.00011304986414041272, "loss": 0.0055, "step": 5920 }, { "epoch": 0.43, "learning_rate": 0.00011302048909451421, "loss": 0.0054, "step": 5922 }, { "epoch": 0.44, "learning_rate": 0.00011299111404861571, "loss": 0.0057, "step": 5924 }, { "epoch": 0.44, "learning_rate": 0.00011296173900271721, "loss": 0.0056, "step": 5926 }, { "epoch": 0.44, "learning_rate": 0.0001129323639568187, "loss": 0.0054, "step": 5928 }, { "epoch": 0.44, "learning_rate": 0.00011290298891092018, "loss": 0.0055, "step": 5930 }, { "epoch": 0.44, "learning_rate": 0.00011287361386502167, "loss": 0.0057, "step": 5932 }, { "epoch": 0.44, "learning_rate": 0.00011284423881912316, "loss": 0.0055, "step": 5934 }, { "epoch": 0.44, "learning_rate": 0.00011281486377322465, "loss": 0.0055, "step": 5936 }, { "epoch": 0.44, "learning_rate": 0.00011278548872732614, "loss": 0.0055, "step": 5938 }, { "epoch": 0.44, "learning_rate": 0.00011275611368142762, "loss": 0.0055, "step": 5940 }, { "epoch": 0.44, "learning_rate": 0.00011272673863552912, "loss": 0.0054, "step": 5942 }, { "epoch": 0.44, "learning_rate": 0.00011269736358963061, "loss": 0.0052, "step": 5944 }, { "epoch": 0.44, "learning_rate": 0.00011266798854373211, "loss": 0.0055, "step": 5946 }, { "epoch": 0.44, "learning_rate": 0.0001126386134978336, "loss": 0.0054, "step": 5948 }, { "epoch": 0.44, "learning_rate": 0.0001126092384519351, "loss": 0.0055, "step": 5950 }, { "epoch": 0.44, "learning_rate": 0.00011257986340603656, "loss": 0.0058, "step": 5952 }, { "epoch": 0.44, "learning_rate": 0.00011255048836013807, "loss": 0.0055, "step": 5954 }, { "epoch": 0.44, "learning_rate": 0.00011252111331423956, "loss": 0.0055, "step": 5956 }, { "epoch": 0.44, "learning_rate": 0.00011249173826834105, "loss": 0.0056, "step": 5958 }, { "epoch": 0.44, "learning_rate": 0.00011246236322244254, "loss": 0.0055, "step": 5960 }, { "epoch": 0.44, "learning_rate": 0.00011243298817654403, "loss": 0.0055, "step": 5962 }, { "epoch": 0.44, "learning_rate": 0.00011240361313064551, "loss": 0.0057, "step": 5964 }, { "epoch": 0.44, "learning_rate": 0.000112374238084747, "loss": 0.0056, "step": 5966 }, { "epoch": 0.44, "learning_rate": 0.0001123448630388485, "loss": 0.0054, "step": 5968 }, { "epoch": 0.44, "learning_rate": 0.00011231548799295, "loss": 0.0056, "step": 5970 }, { "epoch": 0.44, "learning_rate": 0.0001122861129470515, "loss": 0.0055, "step": 5972 }, { "epoch": 0.44, "learning_rate": 0.00011225673790115299, "loss": 0.0054, "step": 5974 }, { "epoch": 0.44, "learning_rate": 0.00011222736285525446, "loss": 0.0054, "step": 5976 }, { "epoch": 0.44, "learning_rate": 0.00011219798780935596, "loss": 0.0057, "step": 5978 }, { "epoch": 0.44, "learning_rate": 0.00011216861276345745, "loss": 0.0055, "step": 5980 }, { "epoch": 0.44, "learning_rate": 0.00011213923771755894, "loss": 0.0057, "step": 5982 }, { "epoch": 0.44, "learning_rate": 0.00011210986267166043, "loss": 0.0055, "step": 5984 }, { "epoch": 0.44, "learning_rate": 0.00011208048762576191, "loss": 0.0056, "step": 5986 }, { "epoch": 0.44, "learning_rate": 0.0001120511125798634, "loss": 0.0055, "step": 5988 }, { "epoch": 0.44, "learning_rate": 0.0001120217375339649, "loss": 0.0057, "step": 5990 }, { "epoch": 0.44, "learning_rate": 0.0001119923624880664, "loss": 0.0057, "step": 5992 }, { "epoch": 0.44, "learning_rate": 0.00011196298744216789, "loss": 0.0054, "step": 5994 }, { "epoch": 0.44, "learning_rate": 0.00011193361239626938, "loss": 0.0055, "step": 5996 }, { "epoch": 0.44, "learning_rate": 0.00011190423735037086, "loss": 0.0055, "step": 5998 }, { "epoch": 0.44, "learning_rate": 0.00011187486230447235, "loss": 0.0054, "step": 6000 }, { "epoch": 0.44, "learning_rate": 0.00011184548725857385, "loss": 0.0056, "step": 6002 }, { "epoch": 0.44, "learning_rate": 0.00011181611221267534, "loss": 0.0055, "step": 6004 }, { "epoch": 0.44, "learning_rate": 0.00011178673716677683, "loss": 0.0055, "step": 6006 }, { "epoch": 0.44, "learning_rate": 0.00011175736212087833, "loss": 0.0057, "step": 6008 }, { "epoch": 0.44, "learning_rate": 0.0001117279870749798, "loss": 0.0054, "step": 6010 }, { "epoch": 0.44, "learning_rate": 0.00011169861202908129, "loss": 0.0055, "step": 6012 }, { "epoch": 0.44, "learning_rate": 0.00011166923698318278, "loss": 0.0056, "step": 6014 }, { "epoch": 0.44, "learning_rate": 0.00011163986193728429, "loss": 0.0055, "step": 6016 }, { "epoch": 0.44, "learning_rate": 0.00011161048689138578, "loss": 0.0055, "step": 6018 }, { "epoch": 0.44, "learning_rate": 0.00011158111184548727, "loss": 0.0055, "step": 6020 }, { "epoch": 0.44, "learning_rate": 0.00011155173679958875, "loss": 0.0055, "step": 6022 }, { "epoch": 0.44, "learning_rate": 0.00011152236175369024, "loss": 0.0056, "step": 6024 }, { "epoch": 0.44, "learning_rate": 0.00011149298670779174, "loss": 0.0055, "step": 6026 }, { "epoch": 0.44, "learning_rate": 0.00011146361166189323, "loss": 0.0055, "step": 6028 }, { "epoch": 0.44, "learning_rate": 0.00011143423661599472, "loss": 0.0054, "step": 6030 }, { "epoch": 0.44, "learning_rate": 0.00011140486157009622, "loss": 0.0055, "step": 6032 }, { "epoch": 0.44, "learning_rate": 0.00011137548652419769, "loss": 0.0057, "step": 6034 }, { "epoch": 0.44, "learning_rate": 0.00011134611147829918, "loss": 0.0056, "step": 6036 }, { "epoch": 0.44, "learning_rate": 0.00011131673643240069, "loss": 0.0054, "step": 6038 }, { "epoch": 0.44, "learning_rate": 0.00011128736138650218, "loss": 0.0055, "step": 6040 }, { "epoch": 0.44, "learning_rate": 0.00011125798634060367, "loss": 0.0056, "step": 6042 }, { "epoch": 0.44, "learning_rate": 0.00011122861129470515, "loss": 0.0056, "step": 6044 }, { "epoch": 0.44, "learning_rate": 0.00011119923624880664, "loss": 0.0054, "step": 6046 }, { "epoch": 0.44, "learning_rate": 0.00011116986120290813, "loss": 0.0055, "step": 6048 }, { "epoch": 0.44, "learning_rate": 0.00011114048615700962, "loss": 0.0057, "step": 6050 }, { "epoch": 0.44, "learning_rate": 0.00011111111111111112, "loss": 0.0053, "step": 6052 }, { "epoch": 0.44, "learning_rate": 0.00011108173606521262, "loss": 0.0056, "step": 6054 }, { "epoch": 0.44, "learning_rate": 0.00011105236101931409, "loss": 0.0055, "step": 6056 }, { "epoch": 0.44, "learning_rate": 0.00011102298597341558, "loss": 0.0055, "step": 6058 }, { "epoch": 0.45, "learning_rate": 0.00011099361092751708, "loss": 0.0055, "step": 6060 }, { "epoch": 0.45, "learning_rate": 0.00011096423588161858, "loss": 0.0055, "step": 6062 }, { "epoch": 0.45, "learning_rate": 0.00011093486083572007, "loss": 0.0052, "step": 6064 }, { "epoch": 0.45, "learning_rate": 0.00011090548578982156, "loss": 0.0055, "step": 6066 }, { "epoch": 0.45, "learning_rate": 0.00011087611074392304, "loss": 0.0056, "step": 6068 }, { "epoch": 0.45, "learning_rate": 0.00011084673569802453, "loss": 0.0056, "step": 6070 }, { "epoch": 0.45, "learning_rate": 0.00011081736065212602, "loss": 0.0054, "step": 6072 }, { "epoch": 0.45, "learning_rate": 0.00011078798560622751, "loss": 0.0056, "step": 6074 }, { "epoch": 0.45, "learning_rate": 0.000110758610560329, "loss": 0.0056, "step": 6076 }, { "epoch": 0.45, "learning_rate": 0.00011072923551443051, "loss": 0.0054, "step": 6078 }, { "epoch": 0.45, "learning_rate": 0.00011069986046853198, "loss": 0.0054, "step": 6080 }, { "epoch": 0.45, "learning_rate": 0.00011067048542263347, "loss": 0.0058, "step": 6082 }, { "epoch": 0.45, "learning_rate": 0.00011064111037673497, "loss": 0.0053, "step": 6084 }, { "epoch": 0.45, "learning_rate": 0.00011061173533083647, "loss": 0.0055, "step": 6086 }, { "epoch": 0.45, "learning_rate": 0.00011058236028493796, "loss": 0.0057, "step": 6088 }, { "epoch": 0.45, "learning_rate": 0.00011055298523903944, "loss": 0.0056, "step": 6090 }, { "epoch": 0.45, "learning_rate": 0.00011052361019314093, "loss": 0.0055, "step": 6092 }, { "epoch": 0.45, "learning_rate": 0.00011049423514724242, "loss": 0.0055, "step": 6094 }, { "epoch": 0.45, "learning_rate": 0.00011046486010134391, "loss": 0.0057, "step": 6096 }, { "epoch": 0.45, "learning_rate": 0.0001104354850554454, "loss": 0.0055, "step": 6098 }, { "epoch": 0.45, "learning_rate": 0.00011040611000954691, "loss": 0.0054, "step": 6100 }, { "epoch": 0.45, "learning_rate": 0.00011037673496364837, "loss": 0.0056, "step": 6102 }, { "epoch": 0.45, "learning_rate": 0.00011034735991774987, "loss": 0.0053, "step": 6104 }, { "epoch": 0.45, "learning_rate": 0.00011031798487185137, "loss": 0.0055, "step": 6106 }, { "epoch": 0.45, "learning_rate": 0.00011028860982595286, "loss": 0.0055, "step": 6108 }, { "epoch": 0.45, "learning_rate": 0.00011025923478005436, "loss": 0.0055, "step": 6110 }, { "epoch": 0.45, "learning_rate": 0.00011022985973415585, "loss": 0.0054, "step": 6112 }, { "epoch": 0.45, "learning_rate": 0.00011020048468825733, "loss": 0.0056, "step": 6114 }, { "epoch": 0.45, "learning_rate": 0.00011017110964235882, "loss": 0.0056, "step": 6116 }, { "epoch": 0.45, "learning_rate": 0.00011014173459646031, "loss": 0.0058, "step": 6118 }, { "epoch": 0.45, "learning_rate": 0.0001101123595505618, "loss": 0.0053, "step": 6120 }, { "epoch": 0.45, "learning_rate": 0.0001100829845046633, "loss": 0.0054, "step": 6122 }, { "epoch": 0.45, "learning_rate": 0.0001100536094587648, "loss": 0.0056, "step": 6124 }, { "epoch": 0.45, "learning_rate": 0.00011002423441286626, "loss": 0.0055, "step": 6126 }, { "epoch": 0.45, "learning_rate": 0.00010999485936696776, "loss": 0.0055, "step": 6128 }, { "epoch": 0.45, "learning_rate": 0.00010996548432106926, "loss": 0.0057, "step": 6130 }, { "epoch": 0.45, "learning_rate": 0.00010993610927517075, "loss": 0.0055, "step": 6132 }, { "epoch": 0.45, "learning_rate": 0.00010990673422927224, "loss": 0.0056, "step": 6134 }, { "epoch": 0.45, "learning_rate": 0.00010987735918337374, "loss": 0.0055, "step": 6136 }, { "epoch": 0.45, "learning_rate": 0.00010984798413747521, "loss": 0.0054, "step": 6138 }, { "epoch": 0.45, "learning_rate": 0.0001098186090915767, "loss": 0.0054, "step": 6140 }, { "epoch": 0.45, "learning_rate": 0.0001097892340456782, "loss": 0.0055, "step": 6142 }, { "epoch": 0.45, "learning_rate": 0.00010975985899977969, "loss": 0.0055, "step": 6144 }, { "epoch": 0.45, "learning_rate": 0.0001097304839538812, "loss": 0.0056, "step": 6146 }, { "epoch": 0.45, "learning_rate": 0.00010970110890798266, "loss": 0.0054, "step": 6148 }, { "epoch": 0.45, "learning_rate": 0.00010967173386208415, "loss": 0.0053, "step": 6150 }, { "epoch": 0.45, "learning_rate": 0.00010964235881618566, "loss": 0.0055, "step": 6152 }, { "epoch": 0.45, "learning_rate": 0.00010961298377028715, "loss": 0.0057, "step": 6154 }, { "epoch": 0.45, "learning_rate": 0.00010958360872438864, "loss": 0.0056, "step": 6156 }, { "epoch": 0.45, "learning_rate": 0.00010955423367849013, "loss": 0.0055, "step": 6158 }, { "epoch": 0.45, "learning_rate": 0.00010952485863259161, "loss": 0.0055, "step": 6160 }, { "epoch": 0.45, "learning_rate": 0.0001094954835866931, "loss": 0.0056, "step": 6162 }, { "epoch": 0.45, "learning_rate": 0.0001094661085407946, "loss": 0.0055, "step": 6164 }, { "epoch": 0.45, "learning_rate": 0.00010943673349489609, "loss": 0.0056, "step": 6166 }, { "epoch": 0.45, "learning_rate": 0.0001094073584489976, "loss": 0.0054, "step": 6168 }, { "epoch": 0.45, "learning_rate": 0.00010937798340309909, "loss": 0.0055, "step": 6170 }, { "epoch": 0.45, "learning_rate": 0.00010934860835720055, "loss": 0.0053, "step": 6172 }, { "epoch": 0.45, "learning_rate": 0.00010931923331130206, "loss": 0.0056, "step": 6174 }, { "epoch": 0.45, "learning_rate": 0.00010928985826540355, "loss": 0.0055, "step": 6176 }, { "epoch": 0.45, "learning_rate": 0.00010926048321950504, "loss": 0.0053, "step": 6178 }, { "epoch": 0.45, "learning_rate": 0.00010923110817360653, "loss": 0.0055, "step": 6180 }, { "epoch": 0.45, "learning_rate": 0.00010920173312770802, "loss": 0.0056, "step": 6182 }, { "epoch": 0.45, "learning_rate": 0.0001091723580818095, "loss": 0.0055, "step": 6184 }, { "epoch": 0.45, "learning_rate": 0.000109142983035911, "loss": 0.0057, "step": 6186 }, { "epoch": 0.45, "learning_rate": 0.00010911360799001249, "loss": 0.0057, "step": 6188 }, { "epoch": 0.45, "learning_rate": 0.00010908423294411398, "loss": 0.0058, "step": 6190 }, { "epoch": 0.45, "learning_rate": 0.00010905485789821548, "loss": 0.0057, "step": 6192 }, { "epoch": 0.45, "learning_rate": 0.00010902548285231695, "loss": 0.0055, "step": 6194 }, { "epoch": 0.46, "learning_rate": 0.00010899610780641844, "loss": 0.0055, "step": 6196 }, { "epoch": 0.46, "learning_rate": 0.00010896673276051995, "loss": 0.0054, "step": 6198 }, { "epoch": 0.46, "learning_rate": 0.00010893735771462144, "loss": 0.0053, "step": 6200 }, { "epoch": 0.46, "learning_rate": 0.00010890798266872293, "loss": 0.0054, "step": 6202 }, { "epoch": 0.46, "learning_rate": 0.00010887860762282442, "loss": 0.0054, "step": 6204 }, { "epoch": 0.46, "learning_rate": 0.0001088492325769259, "loss": 0.0057, "step": 6206 }, { "epoch": 0.46, "learning_rate": 0.00010881985753102739, "loss": 0.0056, "step": 6208 }, { "epoch": 0.46, "learning_rate": 0.00010879048248512888, "loss": 0.0055, "step": 6210 }, { "epoch": 0.46, "learning_rate": 0.00010876110743923038, "loss": 0.0055, "step": 6212 }, { "epoch": 0.46, "learning_rate": 0.00010873173239333188, "loss": 0.0058, "step": 6214 }, { "epoch": 0.46, "learning_rate": 0.00010870235734743337, "loss": 0.0057, "step": 6216 }, { "epoch": 0.46, "learning_rate": 0.00010867298230153484, "loss": 0.0054, "step": 6218 }, { "epoch": 0.46, "learning_rate": 0.00010864360725563634, "loss": 0.0055, "step": 6220 }, { "epoch": 0.46, "learning_rate": 0.00010861423220973783, "loss": 0.0057, "step": 6222 }, { "epoch": 0.46, "learning_rate": 0.00010858485716383933, "loss": 0.0056, "step": 6224 }, { "epoch": 0.46, "learning_rate": 0.00010855548211794082, "loss": 0.0054, "step": 6226 }, { "epoch": 0.46, "learning_rate": 0.00010852610707204231, "loss": 0.0058, "step": 6228 }, { "epoch": 0.46, "learning_rate": 0.00010849673202614379, "loss": 0.0055, "step": 6230 }, { "epoch": 0.46, "learning_rate": 0.00010846735698024528, "loss": 0.0056, "step": 6232 }, { "epoch": 0.46, "learning_rate": 0.00010843798193434677, "loss": 0.0057, "step": 6234 }, { "epoch": 0.46, "learning_rate": 0.00010840860688844828, "loss": 0.0055, "step": 6236 }, { "epoch": 0.46, "learning_rate": 0.00010837923184254977, "loss": 0.0055, "step": 6238 }, { "epoch": 0.46, "learning_rate": 0.00010834985679665123, "loss": 0.0056, "step": 6240 }, { "epoch": 0.46, "learning_rate": 0.00010832048175075274, "loss": 0.0053, "step": 6242 }, { "epoch": 0.46, "learning_rate": 0.00010829110670485423, "loss": 0.0053, "step": 6244 }, { "epoch": 0.46, "learning_rate": 0.00010826173165895572, "loss": 0.0054, "step": 6246 }, { "epoch": 0.46, "learning_rate": 0.00010823235661305722, "loss": 0.0055, "step": 6248 }, { "epoch": 0.46, "learning_rate": 0.00010820298156715871, "loss": 0.0056, "step": 6250 }, { "epoch": 0.46, "learning_rate": 0.00010817360652126019, "loss": 0.0056, "step": 6252 }, { "epoch": 0.46, "learning_rate": 0.00010814423147536168, "loss": 0.0055, "step": 6254 }, { "epoch": 0.46, "learning_rate": 0.00010811485642946317, "loss": 0.0056, "step": 6256 }, { "epoch": 0.46, "learning_rate": 0.00010808548138356466, "loss": 0.0055, "step": 6258 }, { "epoch": 0.46, "learning_rate": 0.00010805610633766617, "loss": 0.0054, "step": 6260 }, { "epoch": 0.46, "learning_rate": 0.00010802673129176766, "loss": 0.0056, "step": 6262 }, { "epoch": 0.46, "learning_rate": 0.00010799735624586912, "loss": 0.0056, "step": 6264 }, { "epoch": 0.46, "learning_rate": 0.00010796798119997063, "loss": 0.0055, "step": 6266 }, { "epoch": 0.46, "learning_rate": 0.00010793860615407212, "loss": 0.0055, "step": 6268 }, { "epoch": 0.46, "learning_rate": 0.00010790923110817361, "loss": 0.0055, "step": 6270 }, { "epoch": 0.46, "learning_rate": 0.0001078798560622751, "loss": 0.0056, "step": 6272 }, { "epoch": 0.46, "learning_rate": 0.0001078504810163766, "loss": 0.0055, "step": 6274 }, { "epoch": 0.46, "learning_rate": 0.00010782110597047808, "loss": 0.0057, "step": 6276 }, { "epoch": 0.46, "learning_rate": 0.00010779173092457957, "loss": 0.0056, "step": 6278 }, { "epoch": 0.46, "learning_rate": 0.00010776235587868106, "loss": 0.0054, "step": 6280 }, { "epoch": 0.46, "learning_rate": 0.00010773298083278257, "loss": 0.0056, "step": 6282 }, { "epoch": 0.46, "learning_rate": 0.00010770360578688406, "loss": 0.0056, "step": 6284 }, { "epoch": 0.46, "learning_rate": 0.00010767423074098555, "loss": 0.0056, "step": 6286 }, { "epoch": 0.46, "learning_rate": 0.00010764485569508703, "loss": 0.0056, "step": 6288 }, { "epoch": 0.46, "learning_rate": 0.00010761548064918852, "loss": 0.0057, "step": 6290 }, { "epoch": 0.46, "learning_rate": 0.00010758610560329001, "loss": 0.0054, "step": 6292 }, { "epoch": 0.46, "learning_rate": 0.0001075567305573915, "loss": 0.0054, "step": 6294 }, { "epoch": 0.46, "learning_rate": 0.000107527355511493, "loss": 0.0055, "step": 6296 }, { "epoch": 0.46, "learning_rate": 0.00010749798046559447, "loss": 0.0055, "step": 6298 }, { "epoch": 0.46, "learning_rate": 0.00010746860541969597, "loss": 0.0054, "step": 6300 }, { "epoch": 0.46, "learning_rate": 0.00010743923037379746, "loss": 0.0055, "step": 6302 }, { "epoch": 0.46, "learning_rate": 0.00010740985532789896, "loss": 0.0054, "step": 6304 }, { "epoch": 0.46, "learning_rate": 0.00010738048028200045, "loss": 0.0054, "step": 6306 }, { "epoch": 0.46, "learning_rate": 0.00010735110523610195, "loss": 0.0055, "step": 6308 }, { "epoch": 0.46, "learning_rate": 0.00010732173019020341, "loss": 0.0054, "step": 6310 }, { "epoch": 0.46, "learning_rate": 0.00010729235514430492, "loss": 0.0058, "step": 6312 }, { "epoch": 0.46, "learning_rate": 0.00010726298009840641, "loss": 0.0058, "step": 6314 }, { "epoch": 0.46, "learning_rate": 0.0001072336050525079, "loss": 0.0056, "step": 6316 }, { "epoch": 0.46, "learning_rate": 0.00010720423000660939, "loss": 0.0054, "step": 6318 }, { "epoch": 0.46, "learning_rate": 0.00010717485496071088, "loss": 0.0056, "step": 6320 }, { "epoch": 0.46, "learning_rate": 0.00010714547991481236, "loss": 0.0055, "step": 6322 }, { "epoch": 0.46, "learning_rate": 0.00010711610486891385, "loss": 0.0054, "step": 6324 }, { "epoch": 0.46, "learning_rate": 0.00010708672982301535, "loss": 0.0055, "step": 6326 }, { "epoch": 0.46, "learning_rate": 0.00010705735477711685, "loss": 0.0054, "step": 6328 }, { "epoch": 0.46, "learning_rate": 0.00010702797973121834, "loss": 0.0055, "step": 6330 }, { "epoch": 0.46, "learning_rate": 0.00010699860468531984, "loss": 0.0053, "step": 6332 }, { "epoch": 0.47, "learning_rate": 0.00010696922963942131, "loss": 0.0057, "step": 6334 }, { "epoch": 0.47, "learning_rate": 0.0001069398545935228, "loss": 0.0055, "step": 6336 }, { "epoch": 0.47, "learning_rate": 0.0001069104795476243, "loss": 0.0055, "step": 6338 }, { "epoch": 0.47, "learning_rate": 0.00010688110450172579, "loss": 0.0055, "step": 6340 }, { "epoch": 0.47, "learning_rate": 0.00010685172945582728, "loss": 0.0057, "step": 6342 }, { "epoch": 0.47, "learning_rate": 0.00010682235440992876, "loss": 0.0053, "step": 6344 }, { "epoch": 0.47, "learning_rate": 0.00010679297936403025, "loss": 0.0053, "step": 6346 }, { "epoch": 0.47, "learning_rate": 0.00010676360431813174, "loss": 0.0056, "step": 6348 }, { "epoch": 0.47, "learning_rate": 0.00010673422927223325, "loss": 0.0056, "step": 6350 }, { "epoch": 0.47, "learning_rate": 0.00010670485422633474, "loss": 0.0056, "step": 6352 }, { "epoch": 0.47, "learning_rate": 0.00010667547918043623, "loss": 0.0054, "step": 6354 }, { "epoch": 0.47, "learning_rate": 0.00010664610413453771, "loss": 0.0056, "step": 6356 }, { "epoch": 0.47, "learning_rate": 0.0001066167290886392, "loss": 0.0055, "step": 6358 }, { "epoch": 0.47, "learning_rate": 0.0001065873540427407, "loss": 0.0056, "step": 6360 }, { "epoch": 0.47, "learning_rate": 0.00010655797899684219, "loss": 0.0054, "step": 6362 }, { "epoch": 0.47, "learning_rate": 0.00010652860395094368, "loss": 0.0054, "step": 6364 }, { "epoch": 0.47, "learning_rate": 0.00010649922890504519, "loss": 0.0054, "step": 6366 }, { "epoch": 0.47, "learning_rate": 0.00010646985385914665, "loss": 0.0054, "step": 6368 }, { "epoch": 0.47, "learning_rate": 0.00010644047881324814, "loss": 0.0055, "step": 6370 }, { "epoch": 0.47, "learning_rate": 0.00010641110376734963, "loss": 0.0055, "step": 6372 }, { "epoch": 0.47, "learning_rate": 0.00010638172872145114, "loss": 0.0056, "step": 6374 }, { "epoch": 0.47, "learning_rate": 0.00010635235367555263, "loss": 0.0056, "step": 6376 }, { "epoch": 0.47, "learning_rate": 0.00010632297862965412, "loss": 0.0054, "step": 6378 }, { "epoch": 0.47, "learning_rate": 0.0001062936035837556, "loss": 0.0056, "step": 6380 }, { "epoch": 0.47, "learning_rate": 0.0001062642285378571, "loss": 0.0054, "step": 6382 }, { "epoch": 0.47, "learning_rate": 0.00010623485349195859, "loss": 0.0055, "step": 6384 }, { "epoch": 0.47, "learning_rate": 0.00010620547844606008, "loss": 0.0056, "step": 6386 }, { "epoch": 0.47, "learning_rate": 0.00010617610340016157, "loss": 0.0056, "step": 6388 }, { "epoch": 0.47, "learning_rate": 0.00010614672835426307, "loss": 0.0055, "step": 6390 }, { "epoch": 0.47, "learning_rate": 0.00010611735330836454, "loss": 0.0053, "step": 6392 }, { "epoch": 0.47, "learning_rate": 0.00010608797826246603, "loss": 0.0055, "step": 6394 }, { "epoch": 0.47, "learning_rate": 0.00010605860321656754, "loss": 0.0055, "step": 6396 }, { "epoch": 0.47, "learning_rate": 0.00010602922817066903, "loss": 0.0053, "step": 6398 }, { "epoch": 0.47, "learning_rate": 0.00010599985312477052, "loss": 0.0054, "step": 6400 }, { "epoch": 0.47, "learning_rate": 0.000105970478078872, "loss": 0.0055, "step": 6402 }, { "epoch": 0.47, "learning_rate": 0.00010594110303297349, "loss": 0.0055, "step": 6404 }, { "epoch": 0.47, "learning_rate": 0.00010591172798707498, "loss": 0.0055, "step": 6406 }, { "epoch": 0.47, "learning_rate": 0.00010588235294117647, "loss": 0.0056, "step": 6408 }, { "epoch": 0.47, "learning_rate": 0.00010585297789527797, "loss": 0.0055, "step": 6410 }, { "epoch": 0.47, "learning_rate": 0.00010582360284937947, "loss": 0.0056, "step": 6412 }, { "epoch": 0.47, "learning_rate": 0.00010579422780348094, "loss": 0.0056, "step": 6414 }, { "epoch": 0.47, "learning_rate": 0.00010576485275758243, "loss": 0.0054, "step": 6416 }, { "epoch": 0.47, "learning_rate": 0.00010573547771168393, "loss": 0.0054, "step": 6418 }, { "epoch": 0.47, "learning_rate": 0.00010570610266578543, "loss": 0.0053, "step": 6420 }, { "epoch": 0.47, "learning_rate": 0.00010567672761988692, "loss": 0.0055, "step": 6422 }, { "epoch": 0.47, "learning_rate": 0.00010564735257398841, "loss": 0.0055, "step": 6424 }, { "epoch": 0.47, "learning_rate": 0.00010561797752808989, "loss": 0.0054, "step": 6426 }, { "epoch": 0.47, "learning_rate": 0.00010558860248219138, "loss": 0.0054, "step": 6428 }, { "epoch": 0.47, "learning_rate": 0.00010555922743629287, "loss": 0.0054, "step": 6430 }, { "epoch": 0.47, "learning_rate": 0.00010552985239039436, "loss": 0.0055, "step": 6432 }, { "epoch": 0.47, "learning_rate": 0.00010550047734449586, "loss": 0.0055, "step": 6434 }, { "epoch": 0.47, "learning_rate": 0.00010547110229859736, "loss": 0.0055, "step": 6436 }, { "epoch": 0.47, "learning_rate": 0.00010544172725269883, "loss": 0.0056, "step": 6438 }, { "epoch": 0.47, "learning_rate": 0.00010541235220680032, "loss": 0.0057, "step": 6440 }, { "epoch": 0.47, "learning_rate": 0.00010538297716090182, "loss": 0.0057, "step": 6442 }, { "epoch": 0.47, "learning_rate": 0.00010535360211500332, "loss": 0.0055, "step": 6444 }, { "epoch": 0.47, "learning_rate": 0.00010532422706910481, "loss": 0.0054, "step": 6446 }, { "epoch": 0.47, "learning_rate": 0.00010529485202320629, "loss": 0.0055, "step": 6448 }, { "epoch": 0.47, "learning_rate": 0.00010526547697730778, "loss": 0.0056, "step": 6450 }, { "epoch": 0.47, "learning_rate": 0.00010523610193140927, "loss": 0.0054, "step": 6452 }, { "epoch": 0.47, "learning_rate": 0.00010520672688551076, "loss": 0.0052, "step": 6454 }, { "epoch": 0.47, "learning_rate": 0.00010517735183961225, "loss": 0.0056, "step": 6456 }, { "epoch": 0.47, "learning_rate": 0.00010514797679371376, "loss": 0.0052, "step": 6458 }, { "epoch": 0.47, "learning_rate": 0.00010511860174781522, "loss": 0.0056, "step": 6460 }, { "epoch": 0.47, "learning_rate": 0.00010508922670191672, "loss": 0.0055, "step": 6462 }, { "epoch": 0.47, "learning_rate": 0.00010505985165601822, "loss": 0.0054, "step": 6464 }, { "epoch": 0.47, "learning_rate": 0.00010503047661011971, "loss": 0.0054, "step": 6466 }, { "epoch": 0.47, "learning_rate": 0.0001050011015642212, "loss": 0.0056, "step": 6468 }, { "epoch": 0.48, "learning_rate": 0.0001049717265183227, "loss": 0.0057, "step": 6470 }, { "epoch": 0.48, "learning_rate": 0.00010494235147242418, "loss": 0.0054, "step": 6472 }, { "epoch": 0.48, "learning_rate": 0.00010491297642652567, "loss": 0.0055, "step": 6474 }, { "epoch": 0.48, "learning_rate": 0.00010488360138062716, "loss": 0.0055, "step": 6476 }, { "epoch": 0.48, "learning_rate": 0.00010485422633472865, "loss": 0.0056, "step": 6478 }, { "epoch": 0.48, "learning_rate": 0.00010482485128883016, "loss": 0.0055, "step": 6480 }, { "epoch": 0.48, "learning_rate": 0.00010479547624293165, "loss": 0.0056, "step": 6482 }, { "epoch": 0.48, "learning_rate": 0.00010476610119703311, "loss": 0.0056, "step": 6484 }, { "epoch": 0.48, "learning_rate": 0.0001047367261511346, "loss": 0.0055, "step": 6486 }, { "epoch": 0.48, "learning_rate": 0.00010470735110523611, "loss": 0.0055, "step": 6488 }, { "epoch": 0.48, "learning_rate": 0.0001046779760593376, "loss": 0.0055, "step": 6490 }, { "epoch": 0.48, "learning_rate": 0.0001046486010134391, "loss": 0.0055, "step": 6492 }, { "epoch": 0.48, "learning_rate": 0.00010461922596754057, "loss": 0.0055, "step": 6494 }, { "epoch": 0.48, "learning_rate": 0.00010458985092164207, "loss": 0.0055, "step": 6496 }, { "epoch": 0.48, "learning_rate": 0.00010456047587574356, "loss": 0.0055, "step": 6498 }, { "epoch": 0.48, "learning_rate": 0.00010453110082984505, "loss": 0.0054, "step": 6500 }, { "epoch": 0.48, "learning_rate": 0.00010450172578394654, "loss": 0.0056, "step": 6502 }, { "epoch": 0.48, "learning_rate": 0.00010447235073804805, "loss": 0.0057, "step": 6504 }, { "epoch": 0.48, "learning_rate": 0.00010444297569214951, "loss": 0.0055, "step": 6506 }, { "epoch": 0.48, "learning_rate": 0.000104413600646251, "loss": 0.0053, "step": 6508 }, { "epoch": 0.48, "learning_rate": 0.00010438422560035251, "loss": 0.0055, "step": 6510 }, { "epoch": 0.48, "learning_rate": 0.000104354850554454, "loss": 0.0056, "step": 6512 }, { "epoch": 0.48, "learning_rate": 0.00010432547550855549, "loss": 0.0054, "step": 6514 }, { "epoch": 0.48, "learning_rate": 0.00010429610046265698, "loss": 0.0056, "step": 6516 }, { "epoch": 0.48, "learning_rate": 0.00010426672541675846, "loss": 0.0056, "step": 6518 }, { "epoch": 0.48, "learning_rate": 0.00010423735037085995, "loss": 0.0055, "step": 6520 }, { "epoch": 0.48, "learning_rate": 0.00010420797532496145, "loss": 0.0053, "step": 6522 }, { "epoch": 0.48, "learning_rate": 0.00010417860027906294, "loss": 0.0055, "step": 6524 }, { "epoch": 0.48, "learning_rate": 0.00010414922523316444, "loss": 0.0055, "step": 6526 }, { "epoch": 0.48, "learning_rate": 0.00010411985018726594, "loss": 0.0055, "step": 6528 }, { "epoch": 0.48, "learning_rate": 0.0001040904751413674, "loss": 0.0056, "step": 6530 }, { "epoch": 0.48, "learning_rate": 0.0001040611000954689, "loss": 0.0053, "step": 6532 }, { "epoch": 0.48, "learning_rate": 0.0001040317250495704, "loss": 0.0054, "step": 6534 }, { "epoch": 0.48, "learning_rate": 0.00010400235000367189, "loss": 0.0054, "step": 6536 }, { "epoch": 0.48, "learning_rate": 0.00010397297495777338, "loss": 0.0056, "step": 6538 }, { "epoch": 0.48, "learning_rate": 0.00010394359991187487, "loss": 0.0056, "step": 6540 }, { "epoch": 0.48, "learning_rate": 0.00010391422486597635, "loss": 0.0055, "step": 6542 }, { "epoch": 0.48, "learning_rate": 0.00010388484982007784, "loss": 0.0054, "step": 6544 }, { "epoch": 0.48, "learning_rate": 0.00010385547477417934, "loss": 0.0056, "step": 6546 }, { "epoch": 0.48, "learning_rate": 0.00010382609972828083, "loss": 0.0055, "step": 6548 }, { "epoch": 0.48, "learning_rate": 0.00010379672468238233, "loss": 0.0055, "step": 6550 }, { "epoch": 0.48, "learning_rate": 0.0001037673496364838, "loss": 0.0054, "step": 6552 }, { "epoch": 0.48, "learning_rate": 0.00010373797459058529, "loss": 0.0053, "step": 6554 }, { "epoch": 0.48, "learning_rate": 0.0001037085995446868, "loss": 0.0056, "step": 6556 }, { "epoch": 0.48, "learning_rate": 0.00010367922449878829, "loss": 0.0054, "step": 6558 }, { "epoch": 0.48, "learning_rate": 0.00010364984945288978, "loss": 0.0055, "step": 6560 }, { "epoch": 0.48, "learning_rate": 0.00010362047440699127, "loss": 0.0056, "step": 6562 }, { "epoch": 0.48, "learning_rate": 0.00010359109936109275, "loss": 0.0054, "step": 6564 }, { "epoch": 0.48, "learning_rate": 0.00010356172431519424, "loss": 0.0054, "step": 6566 }, { "epoch": 0.48, "learning_rate": 0.00010353234926929573, "loss": 0.0057, "step": 6568 }, { "epoch": 0.48, "learning_rate": 0.00010350297422339723, "loss": 0.0056, "step": 6570 }, { "epoch": 0.48, "learning_rate": 0.00010347359917749873, "loss": 0.0055, "step": 6572 }, { "epoch": 0.48, "learning_rate": 0.00010344422413160022, "loss": 0.0054, "step": 6574 }, { "epoch": 0.48, "learning_rate": 0.00010341484908570169, "loss": 0.0057, "step": 6576 }, { "epoch": 0.48, "learning_rate": 0.00010338547403980319, "loss": 0.0053, "step": 6578 }, { "epoch": 0.48, "learning_rate": 0.00010335609899390469, "loss": 0.0055, "step": 6580 }, { "epoch": 0.48, "learning_rate": 0.00010332672394800618, "loss": 0.0055, "step": 6582 }, { "epoch": 0.48, "learning_rate": 0.00010329734890210767, "loss": 0.0055, "step": 6584 }, { "epoch": 0.48, "learning_rate": 0.00010326797385620916, "loss": 0.0056, "step": 6586 }, { "epoch": 0.48, "learning_rate": 0.00010323859881031064, "loss": 0.0054, "step": 6588 }, { "epoch": 0.48, "learning_rate": 0.00010320922376441213, "loss": 0.0053, "step": 6590 }, { "epoch": 0.48, "learning_rate": 0.00010317984871851362, "loss": 0.0056, "step": 6592 }, { "epoch": 0.48, "learning_rate": 0.00010315047367261513, "loss": 0.0054, "step": 6594 }, { "epoch": 0.48, "learning_rate": 0.00010312109862671662, "loss": 0.0055, "step": 6596 }, { "epoch": 0.48, "learning_rate": 0.00010309172358081809, "loss": 0.0054, "step": 6598 }, { "epoch": 0.48, "learning_rate": 0.00010306234853491958, "loss": 0.0054, "step": 6600 }, { "epoch": 0.48, "learning_rate": 0.00010303297348902108, "loss": 0.0057, "step": 6602 }, { "epoch": 0.48, "learning_rate": 0.00010300359844312257, "loss": 0.0057, "step": 6604 }, { "epoch": 0.49, "learning_rate": 0.00010297422339722407, "loss": 0.0054, "step": 6606 }, { "epoch": 0.49, "learning_rate": 0.00010294484835132556, "loss": 0.0056, "step": 6608 }, { "epoch": 0.49, "learning_rate": 0.00010291547330542704, "loss": 0.0054, "step": 6610 }, { "epoch": 0.49, "learning_rate": 0.00010288609825952853, "loss": 0.0056, "step": 6612 }, { "epoch": 0.49, "learning_rate": 0.00010285672321363002, "loss": 0.0056, "step": 6614 }, { "epoch": 0.49, "learning_rate": 0.00010282734816773151, "loss": 0.0056, "step": 6616 }, { "epoch": 0.49, "learning_rate": 0.00010279797312183302, "loss": 0.0056, "step": 6618 }, { "epoch": 0.49, "learning_rate": 0.00010276859807593451, "loss": 0.0053, "step": 6620 }, { "epoch": 0.49, "learning_rate": 0.00010273922303003597, "loss": 0.0057, "step": 6622 }, { "epoch": 0.49, "learning_rate": 0.00010270984798413748, "loss": 0.0054, "step": 6624 }, { "epoch": 0.49, "learning_rate": 0.00010268047293823897, "loss": 0.0055, "step": 6626 }, { "epoch": 0.49, "learning_rate": 0.00010265109789234046, "loss": 0.0056, "step": 6628 }, { "epoch": 0.49, "learning_rate": 0.00010262172284644196, "loss": 0.0054, "step": 6630 }, { "epoch": 0.49, "learning_rate": 0.00010259234780054345, "loss": 0.0055, "step": 6632 }, { "epoch": 0.49, "learning_rate": 0.00010256297275464493, "loss": 0.0055, "step": 6634 }, { "epoch": 0.49, "learning_rate": 0.00010253359770874642, "loss": 0.0055, "step": 6636 }, { "epoch": 0.49, "learning_rate": 0.00010250422266284791, "loss": 0.0056, "step": 6638 }, { "epoch": 0.49, "learning_rate": 0.00010247484761694942, "loss": 0.0054, "step": 6640 }, { "epoch": 0.49, "learning_rate": 0.00010244547257105091, "loss": 0.0054, "step": 6642 }, { "epoch": 0.49, "learning_rate": 0.00010241609752515237, "loss": 0.0055, "step": 6644 }, { "epoch": 0.49, "learning_rate": 0.00010238672247925388, "loss": 0.0055, "step": 6646 }, { "epoch": 0.49, "learning_rate": 0.00010235734743335537, "loss": 0.0054, "step": 6648 }, { "epoch": 0.49, "learning_rate": 0.00010232797238745686, "loss": 0.0055, "step": 6650 }, { "epoch": 0.49, "learning_rate": 0.00010229859734155835, "loss": 0.0054, "step": 6652 }, { "epoch": 0.49, "learning_rate": 0.00010226922229565985, "loss": 0.0055, "step": 6654 }, { "epoch": 0.49, "learning_rate": 0.00010223984724976132, "loss": 0.0056, "step": 6656 }, { "epoch": 0.49, "learning_rate": 0.00010221047220386282, "loss": 0.0054, "step": 6658 }, { "epoch": 0.49, "learning_rate": 0.00010218109715796431, "loss": 0.0059, "step": 6660 }, { "epoch": 0.49, "learning_rate": 0.0001021517221120658, "loss": 0.0055, "step": 6662 }, { "epoch": 0.49, "learning_rate": 0.0001021223470661673, "loss": 0.0055, "step": 6664 }, { "epoch": 0.49, "learning_rate": 0.0001020929720202688, "loss": 0.0056, "step": 6666 }, { "epoch": 0.49, "learning_rate": 0.00010206359697437026, "loss": 0.0054, "step": 6668 }, { "epoch": 0.49, "learning_rate": 0.00010203422192847177, "loss": 0.0055, "step": 6670 }, { "epoch": 0.49, "learning_rate": 0.00010200484688257326, "loss": 0.0054, "step": 6672 }, { "epoch": 0.49, "learning_rate": 0.00010197547183667475, "loss": 0.0056, "step": 6674 }, { "epoch": 0.49, "learning_rate": 0.00010194609679077624, "loss": 0.0055, "step": 6676 }, { "epoch": 0.49, "learning_rate": 0.00010191672174487773, "loss": 0.0055, "step": 6678 }, { "epoch": 0.49, "learning_rate": 0.00010188734669897921, "loss": 0.0053, "step": 6680 }, { "epoch": 0.49, "learning_rate": 0.0001018579716530807, "loss": 0.0053, "step": 6682 }, { "epoch": 0.49, "learning_rate": 0.0001018285966071822, "loss": 0.0054, "step": 6684 }, { "epoch": 0.49, "learning_rate": 0.0001017992215612837, "loss": 0.0055, "step": 6686 }, { "epoch": 0.49, "learning_rate": 0.0001017698465153852, "loss": 0.0053, "step": 6688 }, { "epoch": 0.49, "learning_rate": 0.00010174047146948669, "loss": 0.0055, "step": 6690 }, { "epoch": 0.49, "learning_rate": 0.00010171109642358816, "loss": 0.0056, "step": 6692 }, { "epoch": 0.49, "learning_rate": 0.00010168172137768966, "loss": 0.0056, "step": 6694 }, { "epoch": 0.49, "learning_rate": 0.00010165234633179115, "loss": 0.0055, "step": 6696 }, { "epoch": 0.49, "learning_rate": 0.00010162297128589264, "loss": 0.0056, "step": 6698 }, { "epoch": 0.49, "learning_rate": 0.00010159359623999413, "loss": 0.0056, "step": 6700 }, { "epoch": 0.49, "learning_rate": 0.00010156422119409561, "loss": 0.0055, "step": 6702 }, { "epoch": 0.49, "learning_rate": 0.0001015348461481971, "loss": 0.0055, "step": 6704 }, { "epoch": 0.49, "learning_rate": 0.0001015054711022986, "loss": 0.0053, "step": 6706 }, { "epoch": 0.49, "learning_rate": 0.0001014760960564001, "loss": 0.0055, "step": 6708 }, { "epoch": 0.49, "learning_rate": 0.00010144672101050159, "loss": 0.0056, "step": 6710 }, { "epoch": 0.49, "learning_rate": 0.00010141734596460308, "loss": 0.0055, "step": 6712 }, { "epoch": 0.49, "learning_rate": 0.00010138797091870456, "loss": 0.0054, "step": 6714 }, { "epoch": 0.49, "learning_rate": 0.00010135859587280605, "loss": 0.0056, "step": 6716 }, { "epoch": 0.49, "learning_rate": 0.00010132922082690755, "loss": 0.0054, "step": 6718 }, { "epoch": 0.49, "learning_rate": 0.00010129984578100904, "loss": 0.0053, "step": 6720 }, { "epoch": 0.49, "learning_rate": 0.00010127047073511053, "loss": 0.0056, "step": 6722 }, { "epoch": 0.49, "learning_rate": 0.00010124109568921204, "loss": 0.0056, "step": 6724 }, { "epoch": 0.49, "learning_rate": 0.0001012117206433135, "loss": 0.0055, "step": 6726 }, { "epoch": 0.49, "learning_rate": 0.00010118234559741499, "loss": 0.0054, "step": 6728 }, { "epoch": 0.49, "learning_rate": 0.00010115297055151648, "loss": 0.0054, "step": 6730 }, { "epoch": 0.49, "learning_rate": 0.00010112359550561799, "loss": 0.0056, "step": 6732 }, { "epoch": 0.49, "learning_rate": 0.00010109422045971948, "loss": 0.0054, "step": 6734 }, { "epoch": 0.49, "learning_rate": 0.00010106484541382097, "loss": 0.0055, "step": 6736 }, { "epoch": 0.49, "learning_rate": 0.00010103547036792245, "loss": 0.0054, "step": 6738 }, { "epoch": 0.49, "learning_rate": 0.00010100609532202394, "loss": 0.0055, "step": 6740 }, { "epoch": 0.5, "learning_rate": 0.00010097672027612544, "loss": 0.0054, "step": 6742 }, { "epoch": 0.5, "learning_rate": 0.00010094734523022693, "loss": 0.0054, "step": 6744 }, { "epoch": 0.5, "learning_rate": 0.00010091797018432842, "loss": 0.0054, "step": 6746 }, { "epoch": 0.5, "learning_rate": 0.0001008885951384299, "loss": 0.0055, "step": 6748 }, { "epoch": 0.5, "learning_rate": 0.00010085922009253139, "loss": 0.0054, "step": 6750 }, { "epoch": 0.5, "learning_rate": 0.00010082984504663288, "loss": 0.0058, "step": 6752 }, { "epoch": 0.5, "learning_rate": 0.00010080047000073439, "loss": 0.0055, "step": 6754 }, { "epoch": 0.5, "learning_rate": 0.00010077109495483588, "loss": 0.0054, "step": 6756 }, { "epoch": 0.5, "learning_rate": 0.00010074171990893737, "loss": 0.0054, "step": 6758 }, { "epoch": 0.5, "learning_rate": 0.00010071234486303885, "loss": 0.0054, "step": 6760 }, { "epoch": 0.5, "learning_rate": 0.00010068296981714034, "loss": 0.0054, "step": 6762 }, { "epoch": 0.5, "learning_rate": 0.00010065359477124183, "loss": 0.0055, "step": 6764 }, { "epoch": 0.5, "learning_rate": 0.00010062421972534333, "loss": 0.0056, "step": 6766 }, { "epoch": 0.5, "learning_rate": 0.00010059484467944482, "loss": 0.0056, "step": 6768 }, { "epoch": 0.5, "learning_rate": 0.00010056546963354632, "loss": 0.0055, "step": 6770 }, { "epoch": 0.5, "learning_rate": 0.00010053609458764779, "loss": 0.0055, "step": 6772 }, { "epoch": 0.5, "learning_rate": 0.00010050671954174928, "loss": 0.0056, "step": 6774 }, { "epoch": 0.5, "learning_rate": 0.00010047734449585078, "loss": 0.0055, "step": 6776 }, { "epoch": 0.5, "learning_rate": 0.00010044796944995228, "loss": 0.0055, "step": 6778 }, { "epoch": 0.5, "learning_rate": 0.00010041859440405377, "loss": 0.0055, "step": 6780 }, { "epoch": 0.5, "learning_rate": 0.00010038921935815526, "loss": 0.0055, "step": 6782 }, { "epoch": 0.5, "learning_rate": 0.00010035984431225674, "loss": 0.0054, "step": 6784 }, { "epoch": 0.5, "learning_rate": 0.00010033046926635823, "loss": 0.0055, "step": 6786 }, { "epoch": 0.5, "learning_rate": 0.00010030109422045972, "loss": 0.0055, "step": 6788 }, { "epoch": 0.5, "learning_rate": 0.00010027171917456121, "loss": 0.0057, "step": 6790 }, { "epoch": 0.5, "learning_rate": 0.0001002423441286627, "loss": 0.0056, "step": 6792 }, { "epoch": 0.5, "learning_rate": 0.00010021296908276421, "loss": 0.0054, "step": 6794 }, { "epoch": 0.5, "learning_rate": 0.00010018359403686568, "loss": 0.0056, "step": 6796 }, { "epoch": 0.5, "learning_rate": 0.00010015421899096717, "loss": 0.0054, "step": 6798 }, { "epoch": 0.5, "learning_rate": 0.00010012484394506867, "loss": 0.0055, "step": 6800 }, { "epoch": 0.5, "learning_rate": 0.00010009546889917017, "loss": 0.0055, "step": 6802 }, { "epoch": 0.5, "learning_rate": 0.00010006609385327166, "loss": 0.0055, "step": 6804 }, { "epoch": 0.5, "learning_rate": 0.00010003671880737314, "loss": 0.0054, "step": 6806 }, { "epoch": 0.5, "learning_rate": 0.00010000734376147463, "loss": 0.0055, "step": 6808 }, { "epoch": 0.5, "learning_rate": 9.997796871557612e-05, "loss": 0.0055, "step": 6810 }, { "epoch": 0.5, "learning_rate": 9.994859366967761e-05, "loss": 0.0056, "step": 6812 }, { "epoch": 0.5, "learning_rate": 9.99192186237791e-05, "loss": 0.0055, "step": 6814 }, { "epoch": 0.5, "learning_rate": 9.98898435778806e-05, "loss": 0.0054, "step": 6816 }, { "epoch": 0.5, "learning_rate": 9.986046853198209e-05, "loss": 0.0055, "step": 6818 }, { "epoch": 0.5, "learning_rate": 9.983109348608358e-05, "loss": 0.0053, "step": 6820 }, { "epoch": 0.5, "learning_rate": 9.980171844018507e-05, "loss": 0.0054, "step": 6822 }, { "epoch": 0.5, "learning_rate": 9.977234339428656e-05, "loss": 0.0057, "step": 6824 }, { "epoch": 0.5, "learning_rate": 9.974296834838804e-05, "loss": 0.0055, "step": 6826 }, { "epoch": 0.5, "learning_rate": 9.971359330248953e-05, "loss": 0.0055, "step": 6828 }, { "epoch": 0.5, "learning_rate": 9.968421825659104e-05, "loss": 0.0054, "step": 6830 }, { "epoch": 0.5, "learning_rate": 9.965484321069252e-05, "loss": 0.0053, "step": 6832 }, { "epoch": 0.5, "learning_rate": 9.962546816479401e-05, "loss": 0.0055, "step": 6834 }, { "epoch": 0.5, "learning_rate": 9.95960931188955e-05, "loss": 0.0056, "step": 6836 }, { "epoch": 0.5, "learning_rate": 9.9566718072997e-05, "loss": 0.0055, "step": 6838 }, { "epoch": 0.5, "learning_rate": 9.953734302709849e-05, "loss": 0.0058, "step": 6840 }, { "epoch": 0.5, "learning_rate": 9.950796798119998e-05, "loss": 0.0054, "step": 6842 }, { "epoch": 0.5, "learning_rate": 9.947859293530146e-05, "loss": 0.0055, "step": 6844 }, { "epoch": 0.5, "learning_rate": 9.944921788940296e-05, "loss": 0.0054, "step": 6846 }, { "epoch": 0.5, "learning_rate": 9.941984284350445e-05, "loss": 0.0053, "step": 6848 }, { "epoch": 0.5, "learning_rate": 9.939046779760593e-05, "loss": 0.0055, "step": 6850 }, { "epoch": 0.5, "learning_rate": 9.936109275170742e-05, "loss": 0.0055, "step": 6852 }, { "epoch": 0.5, "learning_rate": 9.933171770580893e-05, "loss": 0.0054, "step": 6854 }, { "epoch": 0.5, "learning_rate": 9.930234265991041e-05, "loss": 0.0055, "step": 6856 }, { "epoch": 0.5, "learning_rate": 9.92729676140119e-05, "loss": 0.0054, "step": 6858 }, { "epoch": 0.5, "learning_rate": 9.924359256811339e-05, "loss": 0.0055, "step": 6860 }, { "epoch": 0.5, "learning_rate": 9.921421752221488e-05, "loss": 0.0056, "step": 6862 }, { "epoch": 0.5, "learning_rate": 9.918484247631638e-05, "loss": 0.0054, "step": 6864 }, { "epoch": 0.5, "learning_rate": 9.915546743041787e-05, "loss": 0.0054, "step": 6866 }, { "epoch": 0.5, "learning_rate": 9.912609238451936e-05, "loss": 0.0054, "step": 6868 }, { "epoch": 0.5, "learning_rate": 9.909671733862085e-05, "loss": 0.0054, "step": 6870 }, { "epoch": 0.5, "learning_rate": 9.906734229272233e-05, "loss": 0.0057, "step": 6872 }, { "epoch": 0.5, "learning_rate": 9.903796724682382e-05, "loss": 0.0055, "step": 6874 }, { "epoch": 0.5, "learning_rate": 9.900859220092533e-05, "loss": 0.0056, "step": 6876 }, { "epoch": 0.51, "learning_rate": 9.89792171550268e-05, "loss": 0.0056, "step": 6878 }, { "epoch": 0.51, "learning_rate": 9.89498421091283e-05, "loss": 0.0056, "step": 6880 }, { "epoch": 0.51, "learning_rate": 9.892046706322979e-05, "loss": 0.0057, "step": 6882 }, { "epoch": 0.51, "learning_rate": 9.889109201733128e-05, "loss": 0.0055, "step": 6884 }, { "epoch": 0.51, "learning_rate": 9.886171697143277e-05, "loss": 0.0057, "step": 6886 }, { "epoch": 0.51, "learning_rate": 9.883234192553426e-05, "loss": 0.0054, "step": 6888 }, { "epoch": 0.51, "learning_rate": 9.880296687963576e-05, "loss": 0.0055, "step": 6890 }, { "epoch": 0.51, "learning_rate": 9.877359183373725e-05, "loss": 0.0056, "step": 6892 }, { "epoch": 0.51, "learning_rate": 9.874421678783874e-05, "loss": 0.0055, "step": 6894 }, { "epoch": 0.51, "learning_rate": 9.871484174194022e-05, "loss": 0.0054, "step": 6896 }, { "epoch": 0.51, "learning_rate": 9.868546669604171e-05, "loss": 0.0054, "step": 6898 }, { "epoch": 0.51, "learning_rate": 9.865609165014322e-05, "loss": 0.0055, "step": 6900 }, { "epoch": 0.51, "learning_rate": 9.86267166042447e-05, "loss": 0.0055, "step": 6902 }, { "epoch": 0.51, "learning_rate": 9.859734155834619e-05, "loss": 0.0055, "step": 6904 }, { "epoch": 0.51, "learning_rate": 9.856796651244768e-05, "loss": 0.0057, "step": 6906 }, { "epoch": 0.51, "learning_rate": 9.853859146654917e-05, "loss": 0.0056, "step": 6908 }, { "epoch": 0.51, "learning_rate": 9.850921642065066e-05, "loss": 0.0055, "step": 6910 }, { "epoch": 0.51, "learning_rate": 9.847984137475215e-05, "loss": 0.0054, "step": 6912 }, { "epoch": 0.51, "learning_rate": 9.845046632885365e-05, "loss": 0.0055, "step": 6914 }, { "epoch": 0.51, "learning_rate": 9.842109128295514e-05, "loss": 0.0056, "step": 6916 }, { "epoch": 0.51, "learning_rate": 9.839171623705663e-05, "loss": 0.0054, "step": 6918 }, { "epoch": 0.51, "learning_rate": 9.836234119115811e-05, "loss": 0.0055, "step": 6920 }, { "epoch": 0.51, "learning_rate": 9.833296614525961e-05, "loss": 0.0055, "step": 6922 }, { "epoch": 0.51, "learning_rate": 9.830359109936109e-05, "loss": 0.0055, "step": 6924 }, { "epoch": 0.51, "learning_rate": 9.827421605346258e-05, "loss": 0.0058, "step": 6926 }, { "epoch": 0.51, "learning_rate": 9.824484100756408e-05, "loss": 0.0055, "step": 6928 }, { "epoch": 0.51, "learning_rate": 9.821546596166557e-05, "loss": 0.0056, "step": 6930 }, { "epoch": 0.51, "learning_rate": 9.818609091576706e-05, "loss": 0.0056, "step": 6932 }, { "epoch": 0.51, "learning_rate": 9.815671586986855e-05, "loss": 0.0056, "step": 6934 }, { "epoch": 0.51, "learning_rate": 9.812734082397004e-05, "loss": 0.0055, "step": 6936 }, { "epoch": 0.51, "learning_rate": 9.809796577807154e-05, "loss": 0.0055, "step": 6938 }, { "epoch": 0.51, "learning_rate": 9.806859073217303e-05, "loss": 0.0054, "step": 6940 }, { "epoch": 0.51, "learning_rate": 9.80392156862745e-05, "loss": 0.0055, "step": 6942 }, { "epoch": 0.51, "learning_rate": 9.800984064037601e-05, "loss": 0.0054, "step": 6944 }, { "epoch": 0.51, "learning_rate": 9.79804655944775e-05, "loss": 0.0056, "step": 6946 }, { "epoch": 0.51, "learning_rate": 9.795109054857898e-05, "loss": 0.0054, "step": 6948 }, { "epoch": 0.51, "learning_rate": 9.792171550268047e-05, "loss": 0.0055, "step": 6950 }, { "epoch": 0.51, "learning_rate": 9.789234045678198e-05, "loss": 0.0055, "step": 6952 }, { "epoch": 0.51, "learning_rate": 9.786296541088346e-05, "loss": 0.0055, "step": 6954 }, { "epoch": 0.51, "learning_rate": 9.783359036498495e-05, "loss": 0.0054, "step": 6956 }, { "epoch": 0.51, "learning_rate": 9.780421531908644e-05, "loss": 0.0055, "step": 6958 }, { "epoch": 0.51, "learning_rate": 9.777484027318793e-05, "loss": 0.0053, "step": 6960 }, { "epoch": 0.51, "learning_rate": 9.774546522728942e-05, "loss": 0.0056, "step": 6962 }, { "epoch": 0.51, "learning_rate": 9.771609018139092e-05, "loss": 0.0056, "step": 6964 }, { "epoch": 0.51, "learning_rate": 9.76867151354924e-05, "loss": 0.0057, "step": 6966 }, { "epoch": 0.51, "learning_rate": 9.76573400895939e-05, "loss": 0.0055, "step": 6968 }, { "epoch": 0.51, "learning_rate": 9.762796504369539e-05, "loss": 0.0056, "step": 6970 }, { "epoch": 0.51, "learning_rate": 9.759858999779687e-05, "loss": 0.0058, "step": 6972 }, { "epoch": 0.51, "learning_rate": 9.756921495189836e-05, "loss": 0.0054, "step": 6974 }, { "epoch": 0.51, "learning_rate": 9.753983990599985e-05, "loss": 0.0052, "step": 6976 }, { "epoch": 0.51, "learning_rate": 9.751046486010135e-05, "loss": 0.0055, "step": 6978 }, { "epoch": 0.51, "learning_rate": 9.748108981420284e-05, "loss": 0.0056, "step": 6980 }, { "epoch": 0.51, "learning_rate": 9.745171476830433e-05, "loss": 0.0055, "step": 6982 }, { "epoch": 0.51, "learning_rate": 9.742233972240582e-05, "loss": 0.0055, "step": 6984 }, { "epoch": 0.51, "learning_rate": 9.739296467650731e-05, "loss": 0.0054, "step": 6986 }, { "epoch": 0.51, "learning_rate": 9.736358963060879e-05, "loss": 0.0054, "step": 6988 }, { "epoch": 0.51, "learning_rate": 9.73342145847103e-05, "loss": 0.0052, "step": 6990 }, { "epoch": 0.51, "learning_rate": 9.730483953881179e-05, "loss": 0.0054, "step": 6992 }, { "epoch": 0.51, "learning_rate": 9.727546449291327e-05, "loss": 0.0056, "step": 6994 }, { "epoch": 0.51, "learning_rate": 9.724608944701476e-05, "loss": 0.0053, "step": 6996 }, { "epoch": 0.51, "learning_rate": 9.721671440111627e-05, "loss": 0.0057, "step": 6998 }, { "epoch": 0.51, "learning_rate": 9.718733935521774e-05, "loss": 0.0054, "step": 7000 }, { "epoch": 0.51, "learning_rate": 9.715796430931924e-05, "loss": 0.0054, "step": 7002 }, { "epoch": 0.51, "learning_rate": 9.712858926342073e-05, "loss": 0.0055, "step": 7004 }, { "epoch": 0.51, "learning_rate": 9.709921421752222e-05, "loss": 0.0054, "step": 7006 }, { "epoch": 0.51, "learning_rate": 9.706983917162371e-05, "loss": 0.0055, "step": 7008 }, { "epoch": 0.51, "learning_rate": 9.70404641257252e-05, "loss": 0.0055, "step": 7010 }, { "epoch": 0.51, "learning_rate": 9.70110890798267e-05, "loss": 0.0053, "step": 7012 }, { "epoch": 0.52, "learning_rate": 9.698171403392819e-05, "loss": 0.0054, "step": 7014 }, { "epoch": 0.52, "learning_rate": 9.695233898802968e-05, "loss": 0.0058, "step": 7016 }, { "epoch": 0.52, "learning_rate": 9.692296394213116e-05, "loss": 0.0055, "step": 7018 }, { "epoch": 0.52, "learning_rate": 9.689358889623265e-05, "loss": 0.0056, "step": 7020 }, { "epoch": 0.52, "learning_rate": 9.686421385033416e-05, "loss": 0.0055, "step": 7022 }, { "epoch": 0.52, "learning_rate": 9.683483880443563e-05, "loss": 0.0054, "step": 7024 }, { "epoch": 0.52, "learning_rate": 9.680546375853713e-05, "loss": 0.0053, "step": 7026 }, { "epoch": 0.52, "learning_rate": 9.677608871263862e-05, "loss": 0.0056, "step": 7028 }, { "epoch": 0.52, "learning_rate": 9.674671366674011e-05, "loss": 0.0057, "step": 7030 }, { "epoch": 0.52, "learning_rate": 9.67173386208416e-05, "loss": 0.0056, "step": 7032 }, { "epoch": 0.52, "learning_rate": 9.668796357494308e-05, "loss": 0.0053, "step": 7034 }, { "epoch": 0.52, "learning_rate": 9.665858852904459e-05, "loss": 0.0054, "step": 7036 }, { "epoch": 0.52, "learning_rate": 9.662921348314608e-05, "loss": 0.0056, "step": 7038 }, { "epoch": 0.52, "learning_rate": 9.659983843724756e-05, "loss": 0.0053, "step": 7040 }, { "epoch": 0.52, "learning_rate": 9.657046339134905e-05, "loss": 0.0052, "step": 7042 }, { "epoch": 0.52, "learning_rate": 9.654108834545055e-05, "loss": 0.0057, "step": 7044 }, { "epoch": 0.52, "learning_rate": 9.651171329955203e-05, "loss": 0.0054, "step": 7046 }, { "epoch": 0.52, "learning_rate": 9.648233825365352e-05, "loss": 0.0056, "step": 7048 }, { "epoch": 0.52, "learning_rate": 9.645296320775502e-05, "loss": 0.0056, "step": 7050 }, { "epoch": 0.52, "learning_rate": 9.642358816185651e-05, "loss": 0.0054, "step": 7052 }, { "epoch": 0.52, "learning_rate": 9.6394213115958e-05, "loss": 0.0055, "step": 7054 }, { "epoch": 0.52, "learning_rate": 9.636483807005949e-05, "loss": 0.0056, "step": 7056 }, { "epoch": 0.52, "learning_rate": 9.633546302416098e-05, "loss": 0.0054, "step": 7058 }, { "epoch": 0.52, "learning_rate": 9.630608797826247e-05, "loss": 0.0053, "step": 7060 }, { "epoch": 0.52, "learning_rate": 9.627671293236397e-05, "loss": 0.0055, "step": 7062 }, { "epoch": 0.52, "learning_rate": 9.624733788646545e-05, "loss": 0.0052, "step": 7064 }, { "epoch": 0.52, "learning_rate": 9.621796284056695e-05, "loss": 0.0055, "step": 7066 }, { "epoch": 0.52, "learning_rate": 9.618858779466844e-05, "loss": 0.0054, "step": 7068 }, { "epoch": 0.52, "learning_rate": 9.615921274876992e-05, "loss": 0.0053, "step": 7070 }, { "epoch": 0.52, "learning_rate": 9.612983770287141e-05, "loss": 0.0056, "step": 7072 }, { "epoch": 0.52, "learning_rate": 9.610046265697292e-05, "loss": 0.0055, "step": 7074 }, { "epoch": 0.52, "learning_rate": 9.60710876110744e-05, "loss": 0.0053, "step": 7076 }, { "epoch": 0.52, "learning_rate": 9.604171256517589e-05, "loss": 0.0056, "step": 7078 }, { "epoch": 0.52, "learning_rate": 9.601233751927737e-05, "loss": 0.0055, "step": 7080 }, { "epoch": 0.52, "learning_rate": 9.598296247337887e-05, "loss": 0.0056, "step": 7082 }, { "epoch": 0.52, "learning_rate": 9.595358742748036e-05, "loss": 0.0054, "step": 7084 }, { "epoch": 0.52, "learning_rate": 9.592421238158184e-05, "loss": 0.0055, "step": 7086 }, { "epoch": 0.52, "learning_rate": 9.589483733568333e-05, "loss": 0.0056, "step": 7088 }, { "epoch": 0.52, "learning_rate": 9.586546228978484e-05, "loss": 0.0055, "step": 7090 }, { "epoch": 0.52, "learning_rate": 9.583608724388632e-05, "loss": 0.0054, "step": 7092 }, { "epoch": 0.52, "learning_rate": 9.580671219798781e-05, "loss": 0.0053, "step": 7094 }, { "epoch": 0.52, "learning_rate": 9.57773371520893e-05, "loss": 0.0054, "step": 7096 }, { "epoch": 0.52, "learning_rate": 9.57479621061908e-05, "loss": 0.0054, "step": 7098 }, { "epoch": 0.52, "learning_rate": 9.571858706029229e-05, "loss": 0.0054, "step": 7100 }, { "epoch": 0.52, "learning_rate": 9.568921201439378e-05, "loss": 0.0057, "step": 7102 }, { "epoch": 0.52, "learning_rate": 9.565983696849527e-05, "loss": 0.0056, "step": 7104 }, { "epoch": 0.52, "learning_rate": 9.563046192259676e-05, "loss": 0.0056, "step": 7106 }, { "epoch": 0.52, "learning_rate": 9.560108687669825e-05, "loss": 0.0056, "step": 7108 }, { "epoch": 0.52, "learning_rate": 9.557171183079973e-05, "loss": 0.0057, "step": 7110 }, { "epoch": 0.52, "learning_rate": 9.554233678490124e-05, "loss": 0.0055, "step": 7112 }, { "epoch": 0.52, "learning_rate": 9.551296173900273e-05, "loss": 0.0055, "step": 7114 }, { "epoch": 0.52, "learning_rate": 9.548358669310421e-05, "loss": 0.0056, "step": 7116 }, { "epoch": 0.52, "learning_rate": 9.54542116472057e-05, "loss": 0.0055, "step": 7118 }, { "epoch": 0.52, "learning_rate": 9.54248366013072e-05, "loss": 0.0053, "step": 7120 }, { "epoch": 0.52, "learning_rate": 9.539546155540868e-05, "loss": 0.0054, "step": 7122 }, { "epoch": 0.52, "learning_rate": 9.536608650951018e-05, "loss": 0.0055, "step": 7124 }, { "epoch": 0.52, "learning_rate": 9.533671146361167e-05, "loss": 0.0055, "step": 7126 }, { "epoch": 0.52, "learning_rate": 9.530733641771316e-05, "loss": 0.0053, "step": 7128 }, { "epoch": 0.52, "learning_rate": 9.527796137181465e-05, "loss": 0.0055, "step": 7130 }, { "epoch": 0.52, "learning_rate": 9.524858632591613e-05, "loss": 0.0055, "step": 7132 }, { "epoch": 0.52, "learning_rate": 9.521921128001762e-05, "loss": 0.0058, "step": 7134 }, { "epoch": 0.52, "learning_rate": 9.518983623411913e-05, "loss": 0.0057, "step": 7136 }, { "epoch": 0.52, "learning_rate": 9.51604611882206e-05, "loss": 0.0057, "step": 7138 }, { "epoch": 0.52, "learning_rate": 9.51310861423221e-05, "loss": 0.0055, "step": 7140 }, { "epoch": 0.52, "learning_rate": 9.510171109642359e-05, "loss": 0.0055, "step": 7142 }, { "epoch": 0.52, "learning_rate": 9.507233605052508e-05, "loss": 0.0055, "step": 7144 }, { "epoch": 0.52, "learning_rate": 9.504296100462657e-05, "loss": 0.0056, "step": 7146 }, { "epoch": 0.52, "learning_rate": 9.501358595872807e-05, "loss": 0.0054, "step": 7148 }, { "epoch": 0.53, "learning_rate": 9.498421091282956e-05, "loss": 0.0055, "step": 7150 }, { "epoch": 0.53, "learning_rate": 9.495483586693105e-05, "loss": 0.0056, "step": 7152 }, { "epoch": 0.53, "learning_rate": 9.492546082103254e-05, "loss": 0.0055, "step": 7154 }, { "epoch": 0.53, "learning_rate": 9.489608577513402e-05, "loss": 0.0056, "step": 7156 }, { "epoch": 0.53, "learning_rate": 9.486671072923552e-05, "loss": 0.0054, "step": 7158 }, { "epoch": 0.53, "learning_rate": 9.483733568333702e-05, "loss": 0.0054, "step": 7160 }, { "epoch": 0.53, "learning_rate": 9.48079606374385e-05, "loss": 0.0054, "step": 7162 }, { "epoch": 0.53, "learning_rate": 9.477858559153999e-05, "loss": 0.0054, "step": 7164 }, { "epoch": 0.53, "learning_rate": 9.474921054564149e-05, "loss": 0.0055, "step": 7166 }, { "epoch": 0.53, "learning_rate": 9.471983549974297e-05, "loss": 0.0055, "step": 7168 }, { "epoch": 0.53, "learning_rate": 9.469046045384446e-05, "loss": 0.0054, "step": 7170 }, { "epoch": 0.53, "learning_rate": 9.466108540794595e-05, "loss": 0.0054, "step": 7172 }, { "epoch": 0.53, "learning_rate": 9.463171036204745e-05, "loss": 0.0055, "step": 7174 }, { "epoch": 0.53, "learning_rate": 9.460233531614894e-05, "loss": 0.0055, "step": 7176 }, { "epoch": 0.53, "learning_rate": 9.457296027025042e-05, "loss": 0.0054, "step": 7178 }, { "epoch": 0.53, "learning_rate": 9.454358522435192e-05, "loss": 0.0055, "step": 7180 }, { "epoch": 0.53, "learning_rate": 9.451421017845341e-05, "loss": 0.0054, "step": 7182 }, { "epoch": 0.53, "learning_rate": 9.448483513255489e-05, "loss": 0.0054, "step": 7184 }, { "epoch": 0.53, "learning_rate": 9.445546008665638e-05, "loss": 0.0052, "step": 7186 }, { "epoch": 0.53, "learning_rate": 9.442608504075789e-05, "loss": 0.0053, "step": 7188 }, { "epoch": 0.53, "learning_rate": 9.439670999485937e-05, "loss": 0.0055, "step": 7190 }, { "epoch": 0.53, "learning_rate": 9.436733494896086e-05, "loss": 0.0054, "step": 7192 }, { "epoch": 0.53, "learning_rate": 9.433795990306235e-05, "loss": 0.0053, "step": 7194 }, { "epoch": 0.53, "learning_rate": 9.430858485716384e-05, "loss": 0.0056, "step": 7196 }, { "epoch": 0.53, "learning_rate": 9.427920981126534e-05, "loss": 0.0054, "step": 7198 }, { "epoch": 0.53, "learning_rate": 9.424983476536683e-05, "loss": 0.0057, "step": 7200 }, { "epoch": 0.53, "learning_rate": 9.42204597194683e-05, "loss": 0.0056, "step": 7202 }, { "epoch": 0.53, "learning_rate": 9.419108467356981e-05, "loss": 0.0054, "step": 7204 }, { "epoch": 0.53, "learning_rate": 9.41617096276713e-05, "loss": 0.0055, "step": 7206 }, { "epoch": 0.53, "learning_rate": 9.413233458177278e-05, "loss": 0.0054, "step": 7208 }, { "epoch": 0.53, "learning_rate": 9.410295953587427e-05, "loss": 0.0056, "step": 7210 }, { "epoch": 0.53, "learning_rate": 9.407358448997578e-05, "loss": 0.0055, "step": 7212 }, { "epoch": 0.53, "learning_rate": 9.404420944407726e-05, "loss": 0.0056, "step": 7214 }, { "epoch": 0.53, "learning_rate": 9.401483439817875e-05, "loss": 0.0054, "step": 7216 }, { "epoch": 0.53, "learning_rate": 9.398545935228024e-05, "loss": 0.0054, "step": 7218 }, { "epoch": 0.53, "learning_rate": 9.395608430638173e-05, "loss": 0.0055, "step": 7220 }, { "epoch": 0.53, "learning_rate": 9.392670926048323e-05, "loss": 0.0053, "step": 7222 }, { "epoch": 0.53, "learning_rate": 9.389733421458472e-05, "loss": 0.0056, "step": 7224 }, { "epoch": 0.53, "learning_rate": 9.386795916868621e-05, "loss": 0.0056, "step": 7226 }, { "epoch": 0.53, "learning_rate": 9.38385841227877e-05, "loss": 0.0054, "step": 7228 }, { "epoch": 0.53, "learning_rate": 9.380920907688918e-05, "loss": 0.0054, "step": 7230 }, { "epoch": 0.53, "learning_rate": 9.377983403099067e-05, "loss": 0.0055, "step": 7232 }, { "epoch": 0.53, "learning_rate": 9.375045898509218e-05, "loss": 0.0054, "step": 7234 }, { "epoch": 0.53, "learning_rate": 9.372108393919366e-05, "loss": 0.0055, "step": 7236 }, { "epoch": 0.53, "learning_rate": 9.369170889329515e-05, "loss": 0.0053, "step": 7238 }, { "epoch": 0.53, "learning_rate": 9.366233384739664e-05, "loss": 0.0054, "step": 7240 }, { "epoch": 0.53, "learning_rate": 9.363295880149813e-05, "loss": 0.0054, "step": 7242 }, { "epoch": 0.53, "learning_rate": 9.360358375559962e-05, "loss": 0.0055, "step": 7244 }, { "epoch": 0.53, "learning_rate": 9.357420870970111e-05, "loss": 0.0055, "step": 7246 }, { "epoch": 0.53, "learning_rate": 9.354483366380261e-05, "loss": 0.0054, "step": 7248 }, { "epoch": 0.53, "learning_rate": 9.35154586179041e-05, "loss": 0.0055, "step": 7250 }, { "epoch": 0.53, "learning_rate": 9.348608357200559e-05, "loss": 0.0056, "step": 7252 }, { "epoch": 0.53, "learning_rate": 9.345670852610707e-05, "loss": 0.0055, "step": 7254 }, { "epoch": 0.53, "learning_rate": 9.342733348020856e-05, "loss": 0.0057, "step": 7256 }, { "epoch": 0.53, "learning_rate": 9.339795843431007e-05, "loss": 0.0056, "step": 7258 }, { "epoch": 0.53, "learning_rate": 9.336858338841154e-05, "loss": 0.0056, "step": 7260 }, { "epoch": 0.53, "learning_rate": 9.333920834251304e-05, "loss": 0.0057, "step": 7262 }, { "epoch": 0.53, "learning_rate": 9.330983329661453e-05, "loss": 0.0055, "step": 7264 }, { "epoch": 0.53, "learning_rate": 9.328045825071602e-05, "loss": 0.0054, "step": 7266 }, { "epoch": 0.53, "learning_rate": 9.325108320481751e-05, "loss": 0.0055, "step": 7268 }, { "epoch": 0.53, "learning_rate": 9.3221708158919e-05, "loss": 0.0056, "step": 7270 }, { "epoch": 0.53, "learning_rate": 9.31923331130205e-05, "loss": 0.0054, "step": 7272 }, { "epoch": 0.53, "learning_rate": 9.316295806712199e-05, "loss": 0.0053, "step": 7274 }, { "epoch": 0.53, "learning_rate": 9.313358302122348e-05, "loss": 0.0054, "step": 7276 }, { "epoch": 0.53, "learning_rate": 9.310420797532496e-05, "loss": 0.0054, "step": 7278 }, { "epoch": 0.53, "learning_rate": 9.307483292942646e-05, "loss": 0.0056, "step": 7280 }, { "epoch": 0.53, "learning_rate": 9.304545788352794e-05, "loss": 0.0055, "step": 7282 }, { "epoch": 0.53, "learning_rate": 9.301608283762943e-05, "loss": 0.0052, "step": 7284 }, { "epoch": 0.54, "learning_rate": 9.298670779173093e-05, "loss": 0.0053, "step": 7286 }, { "epoch": 0.54, "learning_rate": 9.295733274583242e-05, "loss": 0.0055, "step": 7288 }, { "epoch": 0.54, "learning_rate": 9.292795769993391e-05, "loss": 0.0055, "step": 7290 }, { "epoch": 0.54, "learning_rate": 9.28985826540354e-05, "loss": 0.0058, "step": 7292 }, { "epoch": 0.54, "learning_rate": 9.28692076081369e-05, "loss": 0.0054, "step": 7294 }, { "epoch": 0.54, "learning_rate": 9.283983256223839e-05, "loss": 0.0055, "step": 7296 }, { "epoch": 0.54, "learning_rate": 9.281045751633988e-05, "loss": 0.0056, "step": 7298 }, { "epoch": 0.54, "learning_rate": 9.278108247044136e-05, "loss": 0.0054, "step": 7300 }, { "epoch": 0.54, "learning_rate": 9.275170742454286e-05, "loss": 0.0055, "step": 7302 }, { "epoch": 0.54, "learning_rate": 9.272233237864435e-05, "loss": 0.0054, "step": 7304 }, { "epoch": 0.54, "learning_rate": 9.269295733274583e-05, "loss": 0.0057, "step": 7306 }, { "epoch": 0.54, "learning_rate": 9.266358228684732e-05, "loss": 0.0055, "step": 7308 }, { "epoch": 0.54, "learning_rate": 9.263420724094883e-05, "loss": 0.0055, "step": 7310 }, { "epoch": 0.54, "learning_rate": 9.260483219505031e-05, "loss": 0.0057, "step": 7312 }, { "epoch": 0.54, "learning_rate": 9.25754571491518e-05, "loss": 0.0056, "step": 7314 }, { "epoch": 0.54, "learning_rate": 9.254608210325329e-05, "loss": 0.0055, "step": 7316 }, { "epoch": 0.54, "learning_rate": 9.251670705735478e-05, "loss": 0.0055, "step": 7318 }, { "epoch": 0.54, "learning_rate": 9.248733201145628e-05, "loss": 0.0055, "step": 7320 }, { "epoch": 0.54, "learning_rate": 9.245795696555777e-05, "loss": 0.0056, "step": 7322 }, { "epoch": 0.54, "learning_rate": 9.242858191965925e-05, "loss": 0.0054, "step": 7324 }, { "epoch": 0.54, "learning_rate": 9.239920687376075e-05, "loss": 0.0054, "step": 7326 }, { "epoch": 0.54, "learning_rate": 9.236983182786224e-05, "loss": 0.0054, "step": 7328 }, { "epoch": 0.54, "learning_rate": 9.234045678196372e-05, "loss": 0.0053, "step": 7330 }, { "epoch": 0.54, "learning_rate": 9.231108173606521e-05, "loss": 0.0055, "step": 7332 }, { "epoch": 0.54, "learning_rate": 9.22817066901667e-05, "loss": 0.0056, "step": 7334 }, { "epoch": 0.54, "learning_rate": 9.22523316442682e-05, "loss": 0.0056, "step": 7336 }, { "epoch": 0.54, "learning_rate": 9.222295659836969e-05, "loss": 0.0054, "step": 7338 }, { "epoch": 0.54, "learning_rate": 9.219358155247118e-05, "loss": 0.0055, "step": 7340 }, { "epoch": 0.54, "learning_rate": 9.216420650657267e-05, "loss": 0.0054, "step": 7342 }, { "epoch": 0.54, "learning_rate": 9.213483146067416e-05, "loss": 0.0054, "step": 7344 }, { "epoch": 0.54, "learning_rate": 9.210545641477564e-05, "loss": 0.0056, "step": 7346 }, { "epoch": 0.54, "learning_rate": 9.207608136887715e-05, "loss": 0.0056, "step": 7348 }, { "epoch": 0.54, "learning_rate": 9.204670632297864e-05, "loss": 0.0054, "step": 7350 }, { "epoch": 0.54, "learning_rate": 9.201733127708012e-05, "loss": 0.0054, "step": 7352 }, { "epoch": 0.54, "learning_rate": 9.198795623118161e-05, "loss": 0.0055, "step": 7354 }, { "epoch": 0.54, "learning_rate": 9.195858118528312e-05, "loss": 0.0054, "step": 7356 }, { "epoch": 0.54, "learning_rate": 9.19292061393846e-05, "loss": 0.0054, "step": 7358 }, { "epoch": 0.54, "learning_rate": 9.189983109348609e-05, "loss": 0.0054, "step": 7360 }, { "epoch": 0.54, "learning_rate": 9.187045604758758e-05, "loss": 0.0055, "step": 7362 }, { "epoch": 0.54, "learning_rate": 9.184108100168907e-05, "loss": 0.0056, "step": 7364 }, { "epoch": 0.54, "learning_rate": 9.181170595579056e-05, "loss": 0.0054, "step": 7366 }, { "epoch": 0.54, "learning_rate": 9.178233090989205e-05, "loss": 0.0054, "step": 7368 }, { "epoch": 0.54, "learning_rate": 9.175295586399353e-05, "loss": 0.0056, "step": 7370 }, { "epoch": 0.54, "learning_rate": 9.172358081809504e-05, "loss": 0.0055, "step": 7372 }, { "epoch": 0.54, "learning_rate": 9.169420577219653e-05, "loss": 0.0056, "step": 7374 }, { "epoch": 0.54, "learning_rate": 9.166483072629801e-05, "loss": 0.0055, "step": 7376 }, { "epoch": 0.54, "learning_rate": 9.16354556803995e-05, "loss": 0.0053, "step": 7378 }, { "epoch": 0.54, "learning_rate": 9.160608063450099e-05, "loss": 0.0056, "step": 7380 }, { "epoch": 0.54, "learning_rate": 9.157670558860248e-05, "loss": 0.0055, "step": 7382 }, { "epoch": 0.54, "learning_rate": 9.154733054270398e-05, "loss": 0.0055, "step": 7384 }, { "epoch": 0.54, "learning_rate": 9.151795549680547e-05, "loss": 0.0052, "step": 7386 }, { "epoch": 0.54, "learning_rate": 9.148858045090696e-05, "loss": 0.0055, "step": 7388 }, { "epoch": 0.54, "learning_rate": 9.145920540500845e-05, "loss": 0.0055, "step": 7390 }, { "epoch": 0.54, "learning_rate": 9.142983035910993e-05, "loss": 0.0055, "step": 7392 }, { "epoch": 0.54, "learning_rate": 9.140045531321144e-05, "loss": 0.0056, "step": 7394 }, { "epoch": 0.54, "learning_rate": 9.137108026731293e-05, "loss": 0.0054, "step": 7396 }, { "epoch": 0.54, "learning_rate": 9.13417052214144e-05, "loss": 0.0055, "step": 7398 }, { "epoch": 0.54, "learning_rate": 9.13123301755159e-05, "loss": 0.0056, "step": 7400 }, { "epoch": 0.54, "learning_rate": 9.12829551296174e-05, "loss": 0.0055, "step": 7402 }, { "epoch": 0.54, "learning_rate": 9.125358008371888e-05, "loss": 0.0055, "step": 7404 }, { "epoch": 0.54, "learning_rate": 9.122420503782037e-05, "loss": 0.0056, "step": 7406 }, { "epoch": 0.54, "learning_rate": 9.119482999192187e-05, "loss": 0.0055, "step": 7408 }, { "epoch": 0.54, "learning_rate": 9.116545494602336e-05, "loss": 0.0054, "step": 7410 }, { "epoch": 0.54, "learning_rate": 9.113607990012485e-05, "loss": 0.0055, "step": 7412 }, { "epoch": 0.54, "learning_rate": 9.110670485422634e-05, "loss": 0.0056, "step": 7414 }, { "epoch": 0.54, "learning_rate": 9.107732980832783e-05, "loss": 0.0054, "step": 7416 }, { "epoch": 0.54, "learning_rate": 9.104795476242933e-05, "loss": 0.0057, "step": 7418 }, { "epoch": 0.54, "learning_rate": 9.101857971653082e-05, "loss": 0.0055, "step": 7420 }, { "epoch": 0.55, "learning_rate": 9.09892046706323e-05, "loss": 0.0056, "step": 7422 }, { "epoch": 0.55, "learning_rate": 9.09598296247338e-05, "loss": 0.0053, "step": 7424 }, { "epoch": 0.55, "learning_rate": 9.093045457883529e-05, "loss": 0.0054, "step": 7426 }, { "epoch": 0.55, "learning_rate": 9.090107953293677e-05, "loss": 0.0054, "step": 7428 }, { "epoch": 0.55, "learning_rate": 9.087170448703826e-05, "loss": 0.0056, "step": 7430 }, { "epoch": 0.55, "learning_rate": 9.084232944113976e-05, "loss": 0.0055, "step": 7432 }, { "epoch": 0.55, "learning_rate": 9.081295439524125e-05, "loss": 0.0055, "step": 7434 }, { "epoch": 0.55, "learning_rate": 9.078357934934274e-05, "loss": 0.0056, "step": 7436 }, { "epoch": 0.55, "learning_rate": 9.075420430344422e-05, "loss": 0.0055, "step": 7438 }, { "epoch": 0.55, "learning_rate": 9.072482925754572e-05, "loss": 0.0054, "step": 7440 }, { "epoch": 0.55, "learning_rate": 9.069545421164721e-05, "loss": 0.0055, "step": 7442 }, { "epoch": 0.55, "learning_rate": 9.066607916574869e-05, "loss": 0.0054, "step": 7444 }, { "epoch": 0.55, "learning_rate": 9.063670411985018e-05, "loss": 0.0053, "step": 7446 }, { "epoch": 0.55, "learning_rate": 9.060732907395169e-05, "loss": 0.0057, "step": 7448 }, { "epoch": 0.55, "learning_rate": 9.057795402805317e-05, "loss": 0.0056, "step": 7450 }, { "epoch": 0.55, "learning_rate": 9.054857898215466e-05, "loss": 0.0055, "step": 7452 }, { "epoch": 0.55, "learning_rate": 9.051920393625615e-05, "loss": 0.0055, "step": 7454 }, { "epoch": 0.55, "learning_rate": 9.048982889035764e-05, "loss": 0.0054, "step": 7456 }, { "epoch": 0.55, "learning_rate": 9.046045384445914e-05, "loss": 0.0056, "step": 7458 }, { "epoch": 0.55, "learning_rate": 9.043107879856063e-05, "loss": 0.0056, "step": 7460 }, { "epoch": 0.55, "learning_rate": 9.040170375266212e-05, "loss": 0.0055, "step": 7462 }, { "epoch": 0.55, "learning_rate": 9.037232870676361e-05, "loss": 0.0055, "step": 7464 }, { "epoch": 0.55, "learning_rate": 9.03429536608651e-05, "loss": 0.0053, "step": 7466 }, { "epoch": 0.55, "learning_rate": 9.031357861496658e-05, "loss": 0.0054, "step": 7468 }, { "epoch": 0.55, "learning_rate": 9.028420356906809e-05, "loss": 0.0054, "step": 7470 }, { "epoch": 0.55, "learning_rate": 9.025482852316958e-05, "loss": 0.0054, "step": 7472 }, { "epoch": 0.55, "learning_rate": 9.022545347727106e-05, "loss": 0.0055, "step": 7474 }, { "epoch": 0.55, "learning_rate": 9.019607843137255e-05, "loss": 0.0055, "step": 7476 }, { "epoch": 0.55, "learning_rate": 9.016670338547406e-05, "loss": 0.0056, "step": 7478 }, { "epoch": 0.55, "learning_rate": 9.013732833957553e-05, "loss": 0.0056, "step": 7480 }, { "epoch": 0.55, "learning_rate": 9.010795329367703e-05, "loss": 0.0054, "step": 7482 }, { "epoch": 0.55, "learning_rate": 9.007857824777852e-05, "loss": 0.0054, "step": 7484 }, { "epoch": 0.55, "learning_rate": 9.004920320188001e-05, "loss": 0.0054, "step": 7486 }, { "epoch": 0.55, "learning_rate": 9.00198281559815e-05, "loss": 0.0055, "step": 7488 }, { "epoch": 0.55, "learning_rate": 8.999045311008298e-05, "loss": 0.0056, "step": 7490 }, { "epoch": 0.55, "learning_rate": 8.996107806418447e-05, "loss": 0.0055, "step": 7492 }, { "epoch": 0.55, "learning_rate": 8.993170301828598e-05, "loss": 0.0054, "step": 7494 }, { "epoch": 0.55, "learning_rate": 8.990232797238746e-05, "loss": 0.0056, "step": 7496 }, { "epoch": 0.55, "learning_rate": 8.987295292648895e-05, "loss": 0.0056, "step": 7498 }, { "epoch": 0.55, "learning_rate": 8.984357788059044e-05, "loss": 0.0055, "step": 7500 }, { "epoch": 0.55, "learning_rate": 8.981420283469193e-05, "loss": 0.0053, "step": 7502 }, { "epoch": 0.55, "learning_rate": 8.978482778879342e-05, "loss": 0.0053, "step": 7504 }, { "epoch": 0.55, "learning_rate": 8.975545274289492e-05, "loss": 0.0055, "step": 7506 }, { "epoch": 0.55, "learning_rate": 8.972607769699641e-05, "loss": 0.0057, "step": 7508 }, { "epoch": 0.55, "learning_rate": 8.96967026510979e-05, "loss": 0.0056, "step": 7510 }, { "epoch": 0.55, "learning_rate": 8.966732760519939e-05, "loss": 0.0054, "step": 7512 }, { "epoch": 0.55, "learning_rate": 8.963795255930087e-05, "loss": 0.0056, "step": 7514 }, { "epoch": 0.55, "learning_rate": 8.960857751340237e-05, "loss": 0.0054, "step": 7516 }, { "epoch": 0.55, "learning_rate": 8.957920246750387e-05, "loss": 0.0054, "step": 7518 }, { "epoch": 0.55, "learning_rate": 8.954982742160535e-05, "loss": 0.0055, "step": 7520 }, { "epoch": 0.55, "learning_rate": 8.952045237570684e-05, "loss": 0.0055, "step": 7522 }, { "epoch": 0.55, "learning_rate": 8.949107732980834e-05, "loss": 0.0056, "step": 7524 }, { "epoch": 0.55, "learning_rate": 8.946170228390982e-05, "loss": 0.0054, "step": 7526 }, { "epoch": 0.55, "learning_rate": 8.943232723801131e-05, "loss": 0.0053, "step": 7528 }, { "epoch": 0.55, "learning_rate": 8.94029521921128e-05, "loss": 0.0054, "step": 7530 }, { "epoch": 0.55, "learning_rate": 8.93735771462143e-05, "loss": 0.0056, "step": 7532 }, { "epoch": 0.55, "learning_rate": 8.934420210031579e-05, "loss": 0.0054, "step": 7534 }, { "epoch": 0.55, "learning_rate": 8.931482705441727e-05, "loss": 0.0058, "step": 7536 }, { "epoch": 0.55, "learning_rate": 8.928545200851877e-05, "loss": 0.0055, "step": 7538 }, { "epoch": 0.55, "learning_rate": 8.925607696262026e-05, "loss": 0.0055, "step": 7540 }, { "epoch": 0.55, "learning_rate": 8.922670191672174e-05, "loss": 0.0057, "step": 7542 }, { "epoch": 0.55, "learning_rate": 8.919732687082323e-05, "loss": 0.0055, "step": 7544 }, { "epoch": 0.55, "learning_rate": 8.916795182492474e-05, "loss": 0.0052, "step": 7546 }, { "epoch": 0.55, "learning_rate": 8.913857677902622e-05, "loss": 0.0054, "step": 7548 }, { "epoch": 0.55, "learning_rate": 8.910920173312771e-05, "loss": 0.0055, "step": 7550 }, { "epoch": 0.55, "learning_rate": 8.90798266872292e-05, "loss": 0.0054, "step": 7552 }, { "epoch": 0.55, "learning_rate": 8.90504516413307e-05, "loss": 0.0055, "step": 7554 }, { "epoch": 0.55, "learning_rate": 8.902107659543219e-05, "loss": 0.0055, "step": 7556 }, { "epoch": 0.56, "learning_rate": 8.899170154953368e-05, "loss": 0.0057, "step": 7558 }, { "epoch": 0.56, "learning_rate": 8.896232650363516e-05, "loss": 0.0054, "step": 7560 }, { "epoch": 0.56, "learning_rate": 8.893295145773666e-05, "loss": 0.0055, "step": 7562 }, { "epoch": 0.56, "learning_rate": 8.890357641183815e-05, "loss": 0.0054, "step": 7564 }, { "epoch": 0.56, "learning_rate": 8.887420136593963e-05, "loss": 0.0055, "step": 7566 }, { "epoch": 0.56, "learning_rate": 8.884482632004112e-05, "loss": 0.0055, "step": 7568 }, { "epoch": 0.56, "learning_rate": 8.881545127414263e-05, "loss": 0.0055, "step": 7570 }, { "epoch": 0.56, "learning_rate": 8.878607622824411e-05, "loss": 0.0053, "step": 7572 }, { "epoch": 0.56, "learning_rate": 8.87567011823456e-05, "loss": 0.0056, "step": 7574 }, { "epoch": 0.56, "learning_rate": 8.872732613644709e-05, "loss": 0.0055, "step": 7576 }, { "epoch": 0.56, "learning_rate": 8.869795109054858e-05, "loss": 0.0055, "step": 7578 }, { "epoch": 0.56, "learning_rate": 8.866857604465008e-05, "loss": 0.0054, "step": 7580 }, { "epoch": 0.56, "learning_rate": 8.863920099875155e-05, "loss": 0.0055, "step": 7582 }, { "epoch": 0.56, "learning_rate": 8.860982595285306e-05, "loss": 0.0055, "step": 7584 }, { "epoch": 0.56, "learning_rate": 8.858045090695455e-05, "loss": 0.0056, "step": 7586 }, { "epoch": 0.56, "learning_rate": 8.855107586105603e-05, "loss": 0.0055, "step": 7588 }, { "epoch": 0.56, "learning_rate": 8.852170081515752e-05, "loss": 0.0055, "step": 7590 }, { "epoch": 0.56, "learning_rate": 8.849232576925903e-05, "loss": 0.0054, "step": 7592 }, { "epoch": 0.56, "learning_rate": 8.84629507233605e-05, "loss": 0.0055, "step": 7594 }, { "epoch": 0.56, "learning_rate": 8.8433575677462e-05, "loss": 0.0054, "step": 7596 }, { "epoch": 0.56, "learning_rate": 8.840420063156349e-05, "loss": 0.0054, "step": 7598 }, { "epoch": 0.56, "learning_rate": 8.837482558566498e-05, "loss": 0.0054, "step": 7600 }, { "epoch": 0.56, "learning_rate": 8.834545053976647e-05, "loss": 0.0057, "step": 7602 }, { "epoch": 0.56, "learning_rate": 8.831607549386797e-05, "loss": 0.0055, "step": 7604 }, { "epoch": 0.56, "learning_rate": 8.828670044796944e-05, "loss": 0.0055, "step": 7606 }, { "epoch": 0.56, "learning_rate": 8.825732540207095e-05, "loss": 0.0057, "step": 7608 }, { "epoch": 0.56, "learning_rate": 8.822795035617244e-05, "loss": 0.0055, "step": 7610 }, { "epoch": 0.56, "learning_rate": 8.819857531027392e-05, "loss": 0.0057, "step": 7612 }, { "epoch": 0.56, "learning_rate": 8.816920026437541e-05, "loss": 0.0053, "step": 7614 }, { "epoch": 0.56, "learning_rate": 8.813982521847692e-05, "loss": 0.0055, "step": 7616 }, { "epoch": 0.56, "learning_rate": 8.81104501725784e-05, "loss": 0.0055, "step": 7618 }, { "epoch": 0.56, "learning_rate": 8.808107512667989e-05, "loss": 0.0054, "step": 7620 }, { "epoch": 0.56, "learning_rate": 8.805170008078138e-05, "loss": 0.0052, "step": 7622 }, { "epoch": 0.56, "learning_rate": 8.802232503488287e-05, "loss": 0.0055, "step": 7624 }, { "epoch": 0.56, "learning_rate": 8.799294998898436e-05, "loss": 0.0052, "step": 7626 }, { "epoch": 0.56, "learning_rate": 8.796357494308585e-05, "loss": 0.0057, "step": 7628 }, { "epoch": 0.56, "learning_rate": 8.793419989718735e-05, "loss": 0.0057, "step": 7630 }, { "epoch": 0.56, "learning_rate": 8.790482485128884e-05, "loss": 0.0054, "step": 7632 }, { "epoch": 0.56, "learning_rate": 8.787544980539032e-05, "loss": 0.0055, "step": 7634 }, { "epoch": 0.56, "learning_rate": 8.784607475949181e-05, "loss": 0.0053, "step": 7636 }, { "epoch": 0.56, "learning_rate": 8.781669971359331e-05, "loss": 0.0053, "step": 7638 }, { "epoch": 0.56, "learning_rate": 8.778732466769479e-05, "loss": 0.0054, "step": 7640 }, { "epoch": 0.56, "learning_rate": 8.775794962179628e-05, "loss": 0.0055, "step": 7642 }, { "epoch": 0.56, "learning_rate": 8.772857457589778e-05, "loss": 0.0054, "step": 7644 }, { "epoch": 0.56, "learning_rate": 8.769919952999927e-05, "loss": 0.0056, "step": 7646 }, { "epoch": 0.56, "learning_rate": 8.766982448410076e-05, "loss": 0.0056, "step": 7648 }, { "epoch": 0.56, "learning_rate": 8.764044943820225e-05, "loss": 0.0056, "step": 7650 }, { "epoch": 0.56, "learning_rate": 8.761107439230374e-05, "loss": 0.0055, "step": 7652 }, { "epoch": 0.56, "learning_rate": 8.758169934640524e-05, "loss": 0.0056, "step": 7654 }, { "epoch": 0.56, "learning_rate": 8.755232430050673e-05, "loss": 0.0054, "step": 7656 }, { "epoch": 0.56, "learning_rate": 8.75229492546082e-05, "loss": 0.0054, "step": 7658 }, { "epoch": 0.56, "learning_rate": 8.749357420870971e-05, "loss": 0.0056, "step": 7660 }, { "epoch": 0.56, "learning_rate": 8.74641991628112e-05, "loss": 0.0057, "step": 7662 }, { "epoch": 0.56, "learning_rate": 8.743482411691268e-05, "loss": 0.0056, "step": 7664 }, { "epoch": 0.56, "learning_rate": 8.740544907101417e-05, "loss": 0.0055, "step": 7666 }, { "epoch": 0.56, "learning_rate": 8.737607402511568e-05, "loss": 0.0054, "step": 7668 }, { "epoch": 0.56, "learning_rate": 8.734669897921716e-05, "loss": 0.0055, "step": 7670 }, { "epoch": 0.56, "learning_rate": 8.731732393331865e-05, "loss": 0.0054, "step": 7672 }, { "epoch": 0.56, "learning_rate": 8.728794888742014e-05, "loss": 0.0054, "step": 7674 }, { "epoch": 0.56, "learning_rate": 8.725857384152163e-05, "loss": 0.0055, "step": 7676 }, { "epoch": 0.56, "learning_rate": 8.722919879562313e-05, "loss": 0.0054, "step": 7678 }, { "epoch": 0.56, "learning_rate": 8.719982374972462e-05, "loss": 0.0055, "step": 7680 }, { "epoch": 0.56, "learning_rate": 8.71704487038261e-05, "loss": 0.0057, "step": 7682 }, { "epoch": 0.56, "learning_rate": 8.71410736579276e-05, "loss": 0.0054, "step": 7684 }, { "epoch": 0.56, "learning_rate": 8.711169861202908e-05, "loss": 0.0055, "step": 7686 }, { "epoch": 0.56, "learning_rate": 8.708232356613057e-05, "loss": 0.0054, "step": 7688 }, { "epoch": 0.56, "learning_rate": 8.705294852023206e-05, "loss": 0.0055, "step": 7690 }, { "epoch": 0.56, "learning_rate": 8.702357347433356e-05, "loss": 0.0055, "step": 7692 }, { "epoch": 0.57, "learning_rate": 8.699419842843505e-05, "loss": 0.0053, "step": 7694 }, { "epoch": 0.57, "learning_rate": 8.696482338253654e-05, "loss": 0.0053, "step": 7696 }, { "epoch": 0.57, "learning_rate": 8.693544833663803e-05, "loss": 0.0055, "step": 7698 }, { "epoch": 0.57, "learning_rate": 8.690607329073952e-05, "loss": 0.0055, "step": 7700 }, { "epoch": 0.57, "learning_rate": 8.687669824484102e-05, "loss": 0.0055, "step": 7702 }, { "epoch": 0.57, "learning_rate": 8.68473231989425e-05, "loss": 0.0053, "step": 7704 }, { "epoch": 0.57, "learning_rate": 8.6817948153044e-05, "loss": 0.0054, "step": 7706 }, { "epoch": 0.57, "learning_rate": 8.678857310714549e-05, "loss": 0.0057, "step": 7708 }, { "epoch": 0.57, "learning_rate": 8.675919806124697e-05, "loss": 0.0053, "step": 7710 }, { "epoch": 0.57, "learning_rate": 8.672982301534846e-05, "loss": 0.0054, "step": 7712 }, { "epoch": 0.57, "learning_rate": 8.670044796944997e-05, "loss": 0.0058, "step": 7714 }, { "epoch": 0.57, "learning_rate": 8.667107292355144e-05, "loss": 0.0053, "step": 7716 }, { "epoch": 0.57, "learning_rate": 8.664169787765294e-05, "loss": 0.0055, "step": 7718 }, { "epoch": 0.57, "learning_rate": 8.661232283175443e-05, "loss": 0.0055, "step": 7720 }, { "epoch": 0.57, "learning_rate": 8.658294778585592e-05, "loss": 0.0056, "step": 7722 }, { "epoch": 0.57, "learning_rate": 8.655357273995741e-05, "loss": 0.0055, "step": 7724 }, { "epoch": 0.57, "learning_rate": 8.65241976940589e-05, "loss": 0.0054, "step": 7726 }, { "epoch": 0.57, "learning_rate": 8.649482264816038e-05, "loss": 0.0054, "step": 7728 }, { "epoch": 0.57, "learning_rate": 8.646544760226189e-05, "loss": 0.0053, "step": 7730 }, { "epoch": 0.57, "learning_rate": 8.643607255636338e-05, "loss": 0.0056, "step": 7732 }, { "epoch": 0.57, "learning_rate": 8.640669751046486e-05, "loss": 0.0053, "step": 7734 }, { "epoch": 0.57, "learning_rate": 8.637732246456635e-05, "loss": 0.0055, "step": 7736 }, { "epoch": 0.57, "learning_rate": 8.634794741866784e-05, "loss": 0.0054, "step": 7738 }, { "epoch": 0.57, "learning_rate": 8.631857237276933e-05, "loss": 0.0055, "step": 7740 }, { "epoch": 0.57, "learning_rate": 8.628919732687083e-05, "loss": 0.0055, "step": 7742 }, { "epoch": 0.57, "learning_rate": 8.625982228097232e-05, "loss": 0.0055, "step": 7744 }, { "epoch": 0.57, "learning_rate": 8.623044723507381e-05, "loss": 0.0054, "step": 7746 }, { "epoch": 0.57, "learning_rate": 8.62010721891753e-05, "loss": 0.0055, "step": 7748 }, { "epoch": 0.57, "learning_rate": 8.617169714327678e-05, "loss": 0.0056, "step": 7750 }, { "epoch": 0.57, "learning_rate": 8.614232209737829e-05, "loss": 0.0056, "step": 7752 }, { "epoch": 0.57, "learning_rate": 8.611294705147978e-05, "loss": 0.0057, "step": 7754 }, { "epoch": 0.57, "learning_rate": 8.608357200558126e-05, "loss": 0.0056, "step": 7756 }, { "epoch": 0.57, "learning_rate": 8.605419695968275e-05, "loss": 0.0052, "step": 7758 }, { "epoch": 0.57, "learning_rate": 8.602482191378425e-05, "loss": 0.0055, "step": 7760 }, { "epoch": 0.57, "learning_rate": 8.599544686788573e-05, "loss": 0.0055, "step": 7762 }, { "epoch": 0.57, "learning_rate": 8.596607182198722e-05, "loss": 0.0057, "step": 7764 }, { "epoch": 0.57, "learning_rate": 8.593669677608872e-05, "loss": 0.0055, "step": 7766 }, { "epoch": 0.57, "learning_rate": 8.590732173019021e-05, "loss": 0.0055, "step": 7768 }, { "epoch": 0.57, "learning_rate": 8.58779466842917e-05, "loss": 0.0055, "step": 7770 }, { "epoch": 0.57, "learning_rate": 8.584857163839319e-05, "loss": 0.0053, "step": 7772 }, { "epoch": 0.57, "learning_rate": 8.581919659249468e-05, "loss": 0.0054, "step": 7774 }, { "epoch": 0.57, "learning_rate": 8.578982154659618e-05, "loss": 0.0054, "step": 7776 }, { "epoch": 0.57, "learning_rate": 8.576044650069767e-05, "loss": 0.0055, "step": 7778 }, { "epoch": 0.57, "learning_rate": 8.573107145479915e-05, "loss": 0.0055, "step": 7780 }, { "epoch": 0.57, "learning_rate": 8.570169640890065e-05, "loss": 0.0055, "step": 7782 }, { "epoch": 0.57, "learning_rate": 8.567232136300214e-05, "loss": 0.0055, "step": 7784 }, { "epoch": 0.57, "learning_rate": 8.564294631710362e-05, "loss": 0.0055, "step": 7786 }, { "epoch": 0.57, "learning_rate": 8.561357127120511e-05, "loss": 0.0055, "step": 7788 }, { "epoch": 0.57, "learning_rate": 8.55841962253066e-05, "loss": 0.0054, "step": 7790 }, { "epoch": 0.57, "learning_rate": 8.55548211794081e-05, "loss": 0.0056, "step": 7792 }, { "epoch": 0.57, "learning_rate": 8.552544613350959e-05, "loss": 0.0056, "step": 7794 }, { "epoch": 0.57, "learning_rate": 8.549607108761107e-05, "loss": 0.0055, "step": 7796 }, { "epoch": 0.57, "learning_rate": 8.546669604171257e-05, "loss": 0.0056, "step": 7798 }, { "epoch": 0.57, "learning_rate": 8.543732099581406e-05, "loss": 0.0055, "step": 7800 }, { "epoch": 0.57, "learning_rate": 8.540794594991554e-05, "loss": 0.0054, "step": 7802 }, { "epoch": 0.57, "learning_rate": 8.537857090401704e-05, "loss": 0.0055, "step": 7804 }, { "epoch": 0.57, "learning_rate": 8.534919585811854e-05, "loss": 0.0055, "step": 7806 }, { "epoch": 0.57, "learning_rate": 8.531982081222002e-05, "loss": 0.0054, "step": 7808 }, { "epoch": 0.57, "learning_rate": 8.529044576632151e-05, "loss": 0.0054, "step": 7810 }, { "epoch": 0.57, "learning_rate": 8.5261070720423e-05, "loss": 0.0055, "step": 7812 }, { "epoch": 0.57, "learning_rate": 8.52316956745245e-05, "loss": 0.0055, "step": 7814 }, { "epoch": 0.57, "learning_rate": 8.520232062862599e-05, "loss": 0.0054, "step": 7816 }, { "epoch": 0.57, "learning_rate": 8.517294558272748e-05, "loss": 0.0052, "step": 7818 }, { "epoch": 0.57, "learning_rate": 8.514357053682897e-05, "loss": 0.0058, "step": 7820 }, { "epoch": 0.57, "learning_rate": 8.511419549093046e-05, "loss": 0.0055, "step": 7822 }, { "epoch": 0.57, "learning_rate": 8.508482044503195e-05, "loss": 0.0053, "step": 7824 }, { "epoch": 0.57, "learning_rate": 8.505544539913343e-05, "loss": 0.0055, "step": 7826 }, { "epoch": 0.57, "learning_rate": 8.502607035323494e-05, "loss": 0.0056, "step": 7828 }, { "epoch": 0.57, "learning_rate": 8.499669530733643e-05, "loss": 0.0055, "step": 7830 }, { "epoch": 0.58, "learning_rate": 8.496732026143791e-05, "loss": 0.0055, "step": 7832 }, { "epoch": 0.58, "learning_rate": 8.49379452155394e-05, "loss": 0.0055, "step": 7834 }, { "epoch": 0.58, "learning_rate": 8.490857016964089e-05, "loss": 0.0057, "step": 7836 }, { "epoch": 0.58, "learning_rate": 8.487919512374238e-05, "loss": 0.0056, "step": 7838 }, { "epoch": 0.58, "learning_rate": 8.484982007784388e-05, "loss": 0.0056, "step": 7840 }, { "epoch": 0.58, "learning_rate": 8.482044503194535e-05, "loss": 0.0056, "step": 7842 }, { "epoch": 0.58, "learning_rate": 8.479106998604686e-05, "loss": 0.0055, "step": 7844 }, { "epoch": 0.58, "learning_rate": 8.476169494014835e-05, "loss": 0.0054, "step": 7846 }, { "epoch": 0.58, "learning_rate": 8.473231989424983e-05, "loss": 0.0054, "step": 7848 }, { "epoch": 0.58, "learning_rate": 8.470294484835132e-05, "loss": 0.0054, "step": 7850 }, { "epoch": 0.58, "learning_rate": 8.467356980245283e-05, "loss": 0.0054, "step": 7852 }, { "epoch": 0.58, "learning_rate": 8.46441947565543e-05, "loss": 0.0056, "step": 7854 }, { "epoch": 0.58, "learning_rate": 8.46148197106558e-05, "loss": 0.0055, "step": 7856 }, { "epoch": 0.58, "learning_rate": 8.458544466475729e-05, "loss": 0.0056, "step": 7858 }, { "epoch": 0.58, "learning_rate": 8.455606961885878e-05, "loss": 0.0054, "step": 7860 }, { "epoch": 0.58, "learning_rate": 8.452669457296027e-05, "loss": 0.0054, "step": 7862 }, { "epoch": 0.58, "learning_rate": 8.449731952706177e-05, "loss": 0.0054, "step": 7864 }, { "epoch": 0.58, "learning_rate": 8.446794448116326e-05, "loss": 0.0055, "step": 7866 }, { "epoch": 0.58, "learning_rate": 8.443856943526475e-05, "loss": 0.0055, "step": 7868 }, { "epoch": 0.58, "learning_rate": 8.440919438936624e-05, "loss": 0.0053, "step": 7870 }, { "epoch": 0.58, "learning_rate": 8.437981934346772e-05, "loss": 0.0055, "step": 7872 }, { "epoch": 0.58, "learning_rate": 8.435044429756923e-05, "loss": 0.0056, "step": 7874 }, { "epoch": 0.58, "learning_rate": 8.432106925167072e-05, "loss": 0.0055, "step": 7876 }, { "epoch": 0.58, "learning_rate": 8.42916942057722e-05, "loss": 0.0055, "step": 7878 }, { "epoch": 0.58, "learning_rate": 8.426231915987369e-05, "loss": 0.0053, "step": 7880 }, { "epoch": 0.58, "learning_rate": 8.423294411397519e-05, "loss": 0.0054, "step": 7882 }, { "epoch": 0.58, "learning_rate": 8.420356906807667e-05, "loss": 0.0057, "step": 7884 }, { "epoch": 0.58, "learning_rate": 8.417419402217816e-05, "loss": 0.0054, "step": 7886 }, { "epoch": 0.58, "learning_rate": 8.414481897627966e-05, "loss": 0.0053, "step": 7888 }, { "epoch": 0.58, "learning_rate": 8.411544393038115e-05, "loss": 0.0055, "step": 7890 }, { "epoch": 0.58, "learning_rate": 8.408606888448264e-05, "loss": 0.0055, "step": 7892 }, { "epoch": 0.58, "learning_rate": 8.405669383858412e-05, "loss": 0.0057, "step": 7894 }, { "epoch": 0.58, "learning_rate": 8.402731879268562e-05, "loss": 0.0056, "step": 7896 }, { "epoch": 0.58, "learning_rate": 8.399794374678711e-05, "loss": 0.0056, "step": 7898 }, { "epoch": 0.58, "learning_rate": 8.396856870088859e-05, "loss": 0.0055, "step": 7900 }, { "epoch": 0.58, "learning_rate": 8.393919365499009e-05, "loss": 0.0054, "step": 7902 }, { "epoch": 0.58, "learning_rate": 8.390981860909159e-05, "loss": 0.0054, "step": 7904 }, { "epoch": 0.58, "learning_rate": 8.388044356319307e-05, "loss": 0.0055, "step": 7906 }, { "epoch": 0.58, "learning_rate": 8.385106851729456e-05, "loss": 0.0054, "step": 7908 }, { "epoch": 0.58, "learning_rate": 8.382169347139605e-05, "loss": 0.0053, "step": 7910 }, { "epoch": 0.58, "learning_rate": 8.379231842549754e-05, "loss": 0.0055, "step": 7912 }, { "epoch": 0.58, "learning_rate": 8.376294337959904e-05, "loss": 0.0055, "step": 7914 }, { "epoch": 0.58, "learning_rate": 8.373356833370053e-05, "loss": 0.0056, "step": 7916 }, { "epoch": 0.58, "learning_rate": 8.370419328780201e-05, "loss": 0.0055, "step": 7918 }, { "epoch": 0.58, "learning_rate": 8.367481824190351e-05, "loss": 0.0055, "step": 7920 }, { "epoch": 0.58, "learning_rate": 8.3645443196005e-05, "loss": 0.0054, "step": 7922 }, { "epoch": 0.58, "learning_rate": 8.361606815010648e-05, "loss": 0.0055, "step": 7924 }, { "epoch": 0.58, "learning_rate": 8.358669310420797e-05, "loss": 0.0055, "step": 7926 }, { "epoch": 0.58, "learning_rate": 8.355731805830948e-05, "loss": 0.0056, "step": 7928 }, { "epoch": 0.58, "learning_rate": 8.352794301241096e-05, "loss": 0.0055, "step": 7930 }, { "epoch": 0.58, "learning_rate": 8.349856796651245e-05, "loss": 0.0056, "step": 7932 }, { "epoch": 0.58, "learning_rate": 8.346919292061394e-05, "loss": 0.0055, "step": 7934 }, { "epoch": 0.58, "learning_rate": 8.343981787471543e-05, "loss": 0.0054, "step": 7936 }, { "epoch": 0.58, "learning_rate": 8.341044282881693e-05, "loss": 0.0056, "step": 7938 }, { "epoch": 0.58, "learning_rate": 8.33810677829184e-05, "loss": 0.0053, "step": 7940 }, { "epoch": 0.58, "learning_rate": 8.335169273701991e-05, "loss": 0.0054, "step": 7942 }, { "epoch": 0.58, "learning_rate": 8.33223176911214e-05, "loss": 0.0054, "step": 7944 }, { "epoch": 0.58, "learning_rate": 8.329294264522288e-05, "loss": 0.0055, "step": 7946 }, { "epoch": 0.58, "learning_rate": 8.326356759932437e-05, "loss": 0.0054, "step": 7948 }, { "epoch": 0.58, "learning_rate": 8.323419255342588e-05, "loss": 0.0056, "step": 7950 }, { "epoch": 0.58, "learning_rate": 8.320481750752736e-05, "loss": 0.0055, "step": 7952 }, { "epoch": 0.58, "learning_rate": 8.317544246162885e-05, "loss": 0.0053, "step": 7954 }, { "epoch": 0.58, "learning_rate": 8.314606741573034e-05, "loss": 0.0055, "step": 7956 }, { "epoch": 0.58, "learning_rate": 8.311669236983183e-05, "loss": 0.0055, "step": 7958 }, { "epoch": 0.58, "learning_rate": 8.308731732393332e-05, "loss": 0.0055, "step": 7960 }, { "epoch": 0.58, "learning_rate": 8.305794227803482e-05, "loss": 0.0056, "step": 7962 }, { "epoch": 0.58, "learning_rate": 8.30285672321363e-05, "loss": 0.0056, "step": 7964 }, { "epoch": 0.58, "learning_rate": 8.29991921862378e-05, "loss": 0.0054, "step": 7966 }, { "epoch": 0.59, "learning_rate": 8.296981714033929e-05, "loss": 0.0054, "step": 7968 }, { "epoch": 0.59, "learning_rate": 8.294044209444077e-05, "loss": 0.0055, "step": 7970 }, { "epoch": 0.59, "learning_rate": 8.291106704854226e-05, "loss": 0.0056, "step": 7972 }, { "epoch": 0.59, "learning_rate": 8.288169200264377e-05, "loss": 0.0056, "step": 7974 }, { "epoch": 0.59, "learning_rate": 8.285231695674525e-05, "loss": 0.0055, "step": 7976 }, { "epoch": 0.59, "learning_rate": 8.282294191084674e-05, "loss": 0.0055, "step": 7978 }, { "epoch": 0.59, "learning_rate": 8.279356686494823e-05, "loss": 0.0055, "step": 7980 }, { "epoch": 0.59, "learning_rate": 8.276419181904972e-05, "loss": 0.0055, "step": 7982 }, { "epoch": 0.59, "learning_rate": 8.273481677315121e-05, "loss": 0.0055, "step": 7984 }, { "epoch": 0.59, "learning_rate": 8.27054417272527e-05, "loss": 0.0055, "step": 7986 }, { "epoch": 0.59, "learning_rate": 8.26760666813542e-05, "loss": 0.0055, "step": 7988 }, { "epoch": 0.59, "learning_rate": 8.264669163545569e-05, "loss": 0.0054, "step": 7990 }, { "epoch": 0.59, "learning_rate": 8.261731658955717e-05, "loss": 0.0055, "step": 7992 }, { "epoch": 0.59, "learning_rate": 8.258794154365866e-05, "loss": 0.0055, "step": 7994 }, { "epoch": 0.59, "learning_rate": 8.255856649776016e-05, "loss": 0.0058, "step": 7996 }, { "epoch": 0.59, "learning_rate": 8.252919145186164e-05, "loss": 0.0056, "step": 7998 }, { "epoch": 0.59, "learning_rate": 8.249981640596313e-05, "loss": 0.0055, "step": 8000 }, { "epoch": 0.59, "learning_rate": 8.247044136006463e-05, "loss": 0.0054, "step": 8002 }, { "epoch": 0.59, "learning_rate": 8.244106631416612e-05, "loss": 0.0052, "step": 8004 }, { "epoch": 0.59, "learning_rate": 8.241169126826761e-05, "loss": 0.0056, "step": 8006 }, { "epoch": 0.59, "learning_rate": 8.23823162223691e-05, "loss": 0.0056, "step": 8008 }, { "epoch": 0.59, "learning_rate": 8.23529411764706e-05, "loss": 0.0053, "step": 8010 }, { "epoch": 0.59, "learning_rate": 8.232356613057209e-05, "loss": 0.0055, "step": 8012 }, { "epoch": 0.59, "learning_rate": 8.229419108467358e-05, "loss": 0.0054, "step": 8014 }, { "epoch": 0.59, "learning_rate": 8.226481603877506e-05, "loss": 0.0054, "step": 8016 }, { "epoch": 0.59, "learning_rate": 8.223544099287656e-05, "loss": 0.0056, "step": 8018 }, { "epoch": 0.59, "learning_rate": 8.220606594697805e-05, "loss": 0.0056, "step": 8020 }, { "epoch": 0.59, "learning_rate": 8.217669090107953e-05, "loss": 0.0055, "step": 8022 }, { "epoch": 0.59, "learning_rate": 8.214731585518102e-05, "loss": 0.0057, "step": 8024 }, { "epoch": 0.59, "learning_rate": 8.211794080928252e-05, "loss": 0.0055, "step": 8026 }, { "epoch": 0.59, "learning_rate": 8.208856576338401e-05, "loss": 0.0055, "step": 8028 }, { "epoch": 0.59, "learning_rate": 8.20591907174855e-05, "loss": 0.0054, "step": 8030 }, { "epoch": 0.59, "learning_rate": 8.202981567158699e-05, "loss": 0.0054, "step": 8032 }, { "epoch": 0.59, "learning_rate": 8.200044062568848e-05, "loss": 0.0054, "step": 8034 }, { "epoch": 0.59, "learning_rate": 8.197106557978998e-05, "loss": 0.0054, "step": 8036 }, { "epoch": 0.59, "learning_rate": 8.194169053389147e-05, "loss": 0.0056, "step": 8038 }, { "epoch": 0.59, "learning_rate": 8.191231548799295e-05, "loss": 0.0054, "step": 8040 }, { "epoch": 0.59, "learning_rate": 8.188294044209445e-05, "loss": 0.0054, "step": 8042 }, { "epoch": 0.59, "learning_rate": 8.185356539619593e-05, "loss": 0.0054, "step": 8044 }, { "epoch": 0.59, "learning_rate": 8.182419035029742e-05, "loss": 0.0054, "step": 8046 }, { "epoch": 0.59, "learning_rate": 8.179481530439891e-05, "loss": 0.0056, "step": 8048 }, { "epoch": 0.59, "learning_rate": 8.17654402585004e-05, "loss": 0.0054, "step": 8050 }, { "epoch": 0.59, "learning_rate": 8.17360652126019e-05, "loss": 0.0055, "step": 8052 }, { "epoch": 0.59, "learning_rate": 8.170669016670339e-05, "loss": 0.0053, "step": 8054 }, { "epoch": 0.59, "learning_rate": 8.167731512080488e-05, "loss": 0.0054, "step": 8056 }, { "epoch": 0.59, "learning_rate": 8.164794007490637e-05, "loss": 0.0056, "step": 8058 }, { "epoch": 0.59, "learning_rate": 8.161856502900787e-05, "loss": 0.0055, "step": 8060 }, { "epoch": 0.59, "learning_rate": 8.158918998310934e-05, "loss": 0.0055, "step": 8062 }, { "epoch": 0.59, "learning_rate": 8.155981493721085e-05, "loss": 0.0055, "step": 8064 }, { "epoch": 0.59, "learning_rate": 8.153043989131234e-05, "loss": 0.0056, "step": 8066 }, { "epoch": 0.59, "learning_rate": 8.150106484541382e-05, "loss": 0.0053, "step": 8068 }, { "epoch": 0.59, "learning_rate": 8.147168979951531e-05, "loss": 0.0056, "step": 8070 }, { "epoch": 0.59, "learning_rate": 8.144231475361682e-05, "loss": 0.0055, "step": 8072 }, { "epoch": 0.59, "learning_rate": 8.14129397077183e-05, "loss": 0.0054, "step": 8074 }, { "epoch": 0.59, "learning_rate": 8.138356466181979e-05, "loss": 0.0054, "step": 8076 }, { "epoch": 0.59, "learning_rate": 8.135418961592128e-05, "loss": 0.0056, "step": 8078 }, { "epoch": 0.59, "learning_rate": 8.132481457002277e-05, "loss": 0.0056, "step": 8080 }, { "epoch": 0.59, "learning_rate": 8.129543952412426e-05, "loss": 0.0053, "step": 8082 }, { "epoch": 0.59, "learning_rate": 8.126606447822575e-05, "loss": 0.0056, "step": 8084 }, { "epoch": 0.59, "learning_rate": 8.123668943232723e-05, "loss": 0.0054, "step": 8086 }, { "epoch": 0.59, "learning_rate": 8.120731438642874e-05, "loss": 0.0055, "step": 8088 }, { "epoch": 0.59, "learning_rate": 8.117793934053022e-05, "loss": 0.0056, "step": 8090 }, { "epoch": 0.59, "learning_rate": 8.114856429463171e-05, "loss": 0.0056, "step": 8092 }, { "epoch": 0.59, "learning_rate": 8.11191892487332e-05, "loss": 0.0055, "step": 8094 }, { "epoch": 0.59, "learning_rate": 8.108981420283469e-05, "loss": 0.0055, "step": 8096 }, { "epoch": 0.59, "learning_rate": 8.106043915693618e-05, "loss": 0.0054, "step": 8098 }, { "epoch": 0.59, "learning_rate": 8.103106411103768e-05, "loss": 0.0056, "step": 8100 }, { "epoch": 0.59, "learning_rate": 8.100168906513917e-05, "loss": 0.0055, "step": 8102 }, { "epoch": 0.6, "learning_rate": 8.097231401924066e-05, "loss": 0.0055, "step": 8104 }, { "epoch": 0.6, "learning_rate": 8.094293897334215e-05, "loss": 0.0055, "step": 8106 }, { "epoch": 0.6, "learning_rate": 8.091356392744363e-05, "loss": 0.0055, "step": 8108 }, { "epoch": 0.6, "learning_rate": 8.088418888154514e-05, "loss": 0.0055, "step": 8110 }, { "epoch": 0.6, "learning_rate": 8.085481383564663e-05, "loss": 0.0055, "step": 8112 }, { "epoch": 0.6, "learning_rate": 8.08254387897481e-05, "loss": 0.0054, "step": 8114 }, { "epoch": 0.6, "learning_rate": 8.07960637438496e-05, "loss": 0.0053, "step": 8116 }, { "epoch": 0.6, "learning_rate": 8.07666886979511e-05, "loss": 0.0054, "step": 8118 }, { "epoch": 0.6, "learning_rate": 8.073731365205258e-05, "loss": 0.0055, "step": 8120 }, { "epoch": 0.6, "learning_rate": 8.070793860615407e-05, "loss": 0.0055, "step": 8122 }, { "epoch": 0.6, "learning_rate": 8.067856356025557e-05, "loss": 0.0055, "step": 8124 }, { "epoch": 0.6, "learning_rate": 8.064918851435706e-05, "loss": 0.0057, "step": 8126 }, { "epoch": 0.6, "learning_rate": 8.061981346845855e-05, "loss": 0.0053, "step": 8128 }, { "epoch": 0.6, "learning_rate": 8.059043842256004e-05, "loss": 0.0054, "step": 8130 }, { "epoch": 0.6, "learning_rate": 8.056106337666153e-05, "loss": 0.0053, "step": 8132 }, { "epoch": 0.6, "learning_rate": 8.053168833076303e-05, "loss": 0.0055, "step": 8134 }, { "epoch": 0.6, "learning_rate": 8.050231328486452e-05, "loss": 0.0056, "step": 8136 }, { "epoch": 0.6, "learning_rate": 8.0472938238966e-05, "loss": 0.0055, "step": 8138 }, { "epoch": 0.6, "learning_rate": 8.04435631930675e-05, "loss": 0.0055, "step": 8140 }, { "epoch": 0.6, "learning_rate": 8.041418814716898e-05, "loss": 0.0054, "step": 8142 }, { "epoch": 0.6, "learning_rate": 8.038481310127047e-05, "loss": 0.0055, "step": 8144 }, { "epoch": 0.6, "learning_rate": 8.035543805537196e-05, "loss": 0.0053, "step": 8146 }, { "epoch": 0.6, "learning_rate": 8.032606300947346e-05, "loss": 0.0056, "step": 8148 }, { "epoch": 0.6, "learning_rate": 8.029668796357495e-05, "loss": 0.0055, "step": 8150 }, { "epoch": 0.6, "learning_rate": 8.026731291767644e-05, "loss": 0.0054, "step": 8152 }, { "epoch": 0.6, "learning_rate": 8.023793787177792e-05, "loss": 0.0054, "step": 8154 }, { "epoch": 0.6, "learning_rate": 8.020856282587942e-05, "loss": 0.0054, "step": 8156 }, { "epoch": 0.6, "learning_rate": 8.017918777998092e-05, "loss": 0.0053, "step": 8158 }, { "epoch": 0.6, "learning_rate": 8.01498127340824e-05, "loss": 0.0055, "step": 8160 }, { "epoch": 0.6, "learning_rate": 8.012043768818389e-05, "loss": 0.0056, "step": 8162 }, { "epoch": 0.6, "learning_rate": 8.009106264228539e-05, "loss": 0.0054, "step": 8164 }, { "epoch": 0.6, "learning_rate": 8.006168759638687e-05, "loss": 0.0054, "step": 8166 }, { "epoch": 0.6, "learning_rate": 8.003231255048836e-05, "loss": 0.0056, "step": 8168 }, { "epoch": 0.6, "learning_rate": 8.000293750458985e-05, "loss": 0.0055, "step": 8170 }, { "epoch": 0.6, "learning_rate": 7.997356245869135e-05, "loss": 0.0054, "step": 8172 } ], "logging_steps": 2, "max_steps": 13617, "num_input_tokens_seen": 0, "num_train_epochs": 1, "save_steps": 2724, "total_flos": 1.081475780660822e+18, "train_batch_size": 4, "trial_name": null, "trial_params": null }