diff --git "a/trainer_state.json" "b/trainer_state.json" new file mode 100644--- /dev/null +++ "b/trainer_state.json" @@ -0,0 +1,11572 @@ +{ + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 122.99509001636662, + "global_step": 18696, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.07, + "learning_rate": 6e-06, + "loss": 27.848, + "step": 10 + }, + { + "epoch": 0.13, + "learning_rate": 1.1400000000000001e-05, + "loss": 28.8912, + "step": 20 + }, + { + "epoch": 0.2, + "learning_rate": 1.6800000000000002e-05, + "loss": 30.7944, + "step": 30 + }, + { + "epoch": 0.26, + "learning_rate": 2.2800000000000002e-05, + "loss": 24.0868, + "step": 40 + }, + { + "epoch": 0.33, + "learning_rate": 2.88e-05, + "loss": 16.5729, + "step": 50 + }, + { + "epoch": 0.39, + "learning_rate": 3.48e-05, + "loss": 14.2674, + "step": 60 + }, + { + "epoch": 0.46, + "learning_rate": 4.08e-05, + "loss": 14.6699, + "step": 70 + }, + { + "epoch": 0.52, + "learning_rate": 4.6800000000000006e-05, + "loss": 10.5936, + "step": 80 + }, + { + "epoch": 0.59, + "learning_rate": 5.28e-05, + "loss": 9.7296, + "step": 90 + }, + { + "epoch": 0.65, + "learning_rate": 5.88e-05, + "loss": 9.0511, + "step": 100 + }, + { + "epoch": 0.72, + "learning_rate": 5.9952399841332805e-05, + "loss": 9.3894, + "step": 110 + }, + { + "epoch": 0.79, + "learning_rate": 5.989289964299881e-05, + "loss": 7.1183, + "step": 120 + }, + { + "epoch": 0.85, + "learning_rate": 5.983339944466482e-05, + "loss": 6.6377, + "step": 130 + }, + { + "epoch": 0.92, + "learning_rate": 5.9773899246330826e-05, + "loss": 6.5715, + "step": 140 + }, + { + "epoch": 0.98, + "learning_rate": 5.9714399047996826e-05, + "loss": 6.2473, + "step": 150 + }, + { + "epoch": 1.05, + "learning_rate": 5.9654898849662834e-05, + "loss": 5.4192, + "step": 160 + }, + { + "epoch": 1.12, + "learning_rate": 5.959539865132884e-05, + "loss": 4.773, + "step": 170 + }, + { + "epoch": 1.18, + "learning_rate": 5.953589845299485e-05, + "loss": 4.6221, + "step": 180 + }, + { + "epoch": 1.25, + "learning_rate": 5.947639825466085e-05, + "loss": 4.446, + "step": 190 + }, + { + "epoch": 1.31, + "learning_rate": 5.9416898056326856e-05, + "loss": 3.785, + "step": 200 + }, + { + "epoch": 1.38, + "learning_rate": 5.935739785799286e-05, + "loss": 3.7092, + "step": 210 + }, + { + "epoch": 1.45, + "learning_rate": 5.929789765965886e-05, + "loss": 3.7962, + "step": 220 + }, + { + "epoch": 1.51, + "learning_rate": 5.923839746132487e-05, + "loss": 3.5789, + "step": 230 + }, + { + "epoch": 1.58, + "learning_rate": 5.917889726299088e-05, + "loss": 3.2916, + "step": 240 + }, + { + "epoch": 1.64, + "learning_rate": 5.9119397064656885e-05, + "loss": 3.2991, + "step": 250 + }, + { + "epoch": 1.71, + "learning_rate": 5.9059896866322885e-05, + "loss": 3.3718, + "step": 260 + }, + { + "epoch": 1.77, + "learning_rate": 5.900039666798889e-05, + "loss": 3.2297, + "step": 270 + }, + { + "epoch": 1.84, + "learning_rate": 5.89408964696549e-05, + "loss": 3.1159, + "step": 280 + }, + { + "epoch": 1.9, + "learning_rate": 5.888139627132091e-05, + "loss": 3.1361, + "step": 290 + }, + { + "epoch": 1.97, + "learning_rate": 5.882189607298691e-05, + "loss": 3.2319, + "step": 300 + }, + { + "epoch": 2.04, + "learning_rate": 5.8762395874652914e-05, + "loss": 3.3508, + "step": 310 + }, + { + "epoch": 2.1, + "learning_rate": 5.870289567631892e-05, + "loss": 3.0835, + "step": 320 + }, + { + "epoch": 2.17, + "learning_rate": 5.864339547798493e-05, + "loss": 3.115, + "step": 330 + }, + { + "epoch": 2.24, + "learning_rate": 5.858389527965093e-05, + "loss": 3.1754, + "step": 340 + }, + { + "epoch": 2.3, + "learning_rate": 5.8524395081316936e-05, + "loss": 3.0651, + "step": 350 + }, + { + "epoch": 2.37, + "learning_rate": 5.846489488298294e-05, + "loss": 3.0475, + "step": 360 + }, + { + "epoch": 2.43, + "learning_rate": 5.840539468464895e-05, + "loss": 3.0804, + "step": 370 + }, + { + "epoch": 2.5, + "learning_rate": 5.834589448631496e-05, + "loss": 3.1193, + "step": 380 + }, + { + "epoch": 2.56, + "learning_rate": 5.8286394287980965e-05, + "loss": 3.0054, + "step": 390 + }, + { + "epoch": 2.63, + "learning_rate": 5.822689408964697e-05, + "loss": 3.005, + "step": 400 + }, + { + "epoch": 2.69, + "learning_rate": 5.816739389131297e-05, + "loss": 3.0306, + "step": 410 + }, + { + "epoch": 2.76, + "learning_rate": 5.810789369297898e-05, + "loss": 3.0225, + "step": 420 + }, + { + "epoch": 2.82, + "learning_rate": 5.804839349464499e-05, + "loss": 2.9728, + "step": 430 + }, + { + "epoch": 2.89, + "learning_rate": 5.7988893296310994e-05, + "loss": 2.9658, + "step": 440 + }, + { + "epoch": 2.96, + "learning_rate": 5.7929393097976994e-05, + "loss": 2.9793, + "step": 450 + }, + { + "epoch": 3.03, + "learning_rate": 5.7869892899643e-05, + "loss": 3.1856, + "step": 460 + }, + { + "epoch": 3.09, + "learning_rate": 5.781039270130901e-05, + "loss": 2.9505, + "step": 470 + }, + { + "epoch": 3.16, + "learning_rate": 5.7750892502975016e-05, + "loss": 2.9522, + "step": 480 + }, + { + "epoch": 3.22, + "learning_rate": 5.7691392304641016e-05, + "loss": 2.9528, + "step": 490 + }, + { + "epoch": 3.29, + "learning_rate": 5.7631892106307024e-05, + "loss": 2.957, + "step": 500 + }, + { + "epoch": 3.29, + "eval_loss": 2.95032000541687, + "eval_runtime": 300.7309, + "eval_samples_per_second": 23.642, + "eval_steps_per_second": 1.48, + "eval_wer": 1.0, + "step": 500 + }, + { + "epoch": 3.35, + "learning_rate": 5.757239190797303e-05, + "loss": 2.9297, + "step": 510 + }, + { + "epoch": 3.42, + "learning_rate": 5.751289170963903e-05, + "loss": 2.9175, + "step": 520 + }, + { + "epoch": 3.48, + "learning_rate": 5.745339151130504e-05, + "loss": 2.9267, + "step": 530 + }, + { + "epoch": 3.55, + "learning_rate": 5.7393891312971046e-05, + "loss": 2.9391, + "step": 540 + }, + { + "epoch": 3.62, + "learning_rate": 5.733439111463705e-05, + "loss": 2.9073, + "step": 550 + }, + { + "epoch": 3.68, + "learning_rate": 5.727489091630305e-05, + "loss": 2.8946, + "step": 560 + }, + { + "epoch": 3.75, + "learning_rate": 5.721539071796906e-05, + "loss": 2.9003, + "step": 570 + }, + { + "epoch": 3.81, + "learning_rate": 5.715589051963507e-05, + "loss": 2.9124, + "step": 580 + }, + { + "epoch": 3.88, + "learning_rate": 5.7096390321301075e-05, + "loss": 2.8812, + "step": 590 + }, + { + "epoch": 3.94, + "learning_rate": 5.7036890122967075e-05, + "loss": 2.8715, + "step": 600 + }, + { + "epoch": 4.01, + "learning_rate": 5.697738992463308e-05, + "loss": 3.0957, + "step": 610 + }, + { + "epoch": 4.08, + "learning_rate": 5.691788972629909e-05, + "loss": 2.8812, + "step": 620 + }, + { + "epoch": 4.14, + "learning_rate": 5.6858389527965097e-05, + "loss": 2.8435, + "step": 630 + }, + { + "epoch": 4.21, + "learning_rate": 5.67988893296311e-05, + "loss": 2.8249, + "step": 640 + }, + { + "epoch": 4.27, + "learning_rate": 5.6739389131297104e-05, + "loss": 2.8412, + "step": 650 + }, + { + "epoch": 4.34, + "learning_rate": 5.667988893296311e-05, + "loss": 2.7992, + "step": 660 + }, + { + "epoch": 4.41, + "learning_rate": 5.662038873462911e-05, + "loss": 2.7854, + "step": 670 + }, + { + "epoch": 4.47, + "learning_rate": 5.656088853629512e-05, + "loss": 2.7678, + "step": 680 + }, + { + "epoch": 4.54, + "learning_rate": 5.6501388337961126e-05, + "loss": 2.7767, + "step": 690 + }, + { + "epoch": 4.6, + "learning_rate": 5.644188813962713e-05, + "loss": 2.6956, + "step": 700 + }, + { + "epoch": 4.67, + "learning_rate": 5.6382387941293134e-05, + "loss": 2.6563, + "step": 710 + }, + { + "epoch": 4.73, + "learning_rate": 5.632288774295914e-05, + "loss": 2.6166, + "step": 720 + }, + { + "epoch": 4.8, + "learning_rate": 5.626338754462515e-05, + "loss": 2.6088, + "step": 730 + }, + { + "epoch": 4.86, + "learning_rate": 5.6203887346291155e-05, + "loss": 2.4939, + "step": 740 + }, + { + "epoch": 4.93, + "learning_rate": 5.614438714795716e-05, + "loss": 2.4259, + "step": 750 + }, + { + "epoch": 5.0, + "learning_rate": 5.608488694962317e-05, + "loss": 2.4021, + "step": 760 + }, + { + "epoch": 5.07, + "learning_rate": 5.602538675128918e-05, + "loss": 2.4815, + "step": 770 + }, + { + "epoch": 5.13, + "learning_rate": 5.5965886552955184e-05, + "loss": 2.2209, + "step": 780 + }, + { + "epoch": 5.2, + "learning_rate": 5.5906386354621184e-05, + "loss": 2.1851, + "step": 790 + }, + { + "epoch": 5.26, + "learning_rate": 5.584688615628719e-05, + "loss": 2.2135, + "step": 800 + }, + { + "epoch": 5.33, + "learning_rate": 5.57873859579532e-05, + "loss": 2.0608, + "step": 810 + }, + { + "epoch": 5.39, + "learning_rate": 5.57278857596192e-05, + "loss": 2.0331, + "step": 820 + }, + { + "epoch": 5.46, + "learning_rate": 5.5668385561285206e-05, + "loss": 2.0301, + "step": 830 + }, + { + "epoch": 5.52, + "learning_rate": 5.5608885362951213e-05, + "loss": 2.0197, + "step": 840 + }, + { + "epoch": 5.59, + "learning_rate": 5.554938516461722e-05, + "loss": 1.9322, + "step": 850 + }, + { + "epoch": 5.65, + "learning_rate": 5.548988496628322e-05, + "loss": 1.9336, + "step": 860 + }, + { + "epoch": 5.72, + "learning_rate": 5.543038476794923e-05, + "loss": 1.9208, + "step": 870 + }, + { + "epoch": 5.79, + "learning_rate": 5.5370884569615235e-05, + "loss": 1.9261, + "step": 880 + }, + { + "epoch": 5.85, + "learning_rate": 5.531138437128124e-05, + "loss": 1.8571, + "step": 890 + }, + { + "epoch": 5.92, + "learning_rate": 5.525188417294724e-05, + "loss": 1.8512, + "step": 900 + }, + { + "epoch": 5.98, + "learning_rate": 5.519238397461325e-05, + "loss": 1.831, + "step": 910 + }, + { + "epoch": 6.05, + "learning_rate": 5.513288377627926e-05, + "loss": 1.9426, + "step": 920 + }, + { + "epoch": 6.12, + "learning_rate": 5.5073383577945265e-05, + "loss": 1.7983, + "step": 930 + }, + { + "epoch": 6.18, + "learning_rate": 5.5013883379611265e-05, + "loss": 1.7945, + "step": 940 + }, + { + "epoch": 6.25, + "learning_rate": 5.495438318127727e-05, + "loss": 1.8402, + "step": 950 + }, + { + "epoch": 6.31, + "learning_rate": 5.489488298294328e-05, + "loss": 1.7505, + "step": 960 + }, + { + "epoch": 6.38, + "learning_rate": 5.483538278460928e-05, + "loss": 1.7542, + "step": 970 + }, + { + "epoch": 6.45, + "learning_rate": 5.477588258627529e-05, + "loss": 1.7426, + "step": 980 + }, + { + "epoch": 6.51, + "learning_rate": 5.4716382387941294e-05, + "loss": 1.8086, + "step": 990 + }, + { + "epoch": 6.58, + "learning_rate": 5.46568821896073e-05, + "loss": 1.7225, + "step": 1000 + }, + { + "epoch": 6.58, + "eval_loss": 0.8859816789627075, + "eval_runtime": 314.8736, + "eval_samples_per_second": 22.58, + "eval_steps_per_second": 1.413, + "eval_wer": 0.7703087196361172, + "step": 1000 + }, + { + "epoch": 6.64, + "learning_rate": 5.45973819912733e-05, + "loss": 1.7249, + "step": 1010 + }, + { + "epoch": 6.71, + "learning_rate": 5.453788179293931e-05, + "loss": 1.7171, + "step": 1020 + }, + { + "epoch": 6.77, + "learning_rate": 5.4478381594605316e-05, + "loss": 1.7518, + "step": 1030 + }, + { + "epoch": 6.84, + "learning_rate": 5.441888139627132e-05, + "loss": 1.7065, + "step": 1040 + }, + { + "epoch": 6.9, + "learning_rate": 5.4359381197937324e-05, + "loss": 1.6863, + "step": 1050 + }, + { + "epoch": 6.97, + "learning_rate": 5.429988099960333e-05, + "loss": 1.6835, + "step": 1060 + }, + { + "epoch": 7.04, + "learning_rate": 5.424038080126934e-05, + "loss": 1.8348, + "step": 1070 + }, + { + "epoch": 7.1, + "learning_rate": 5.4180880602935345e-05, + "loss": 1.6682, + "step": 1080 + }, + { + "epoch": 7.17, + "learning_rate": 5.4121380404601346e-05, + "loss": 1.6787, + "step": 1090 + }, + { + "epoch": 7.24, + "learning_rate": 5.406188020626735e-05, + "loss": 1.6797, + "step": 1100 + }, + { + "epoch": 7.3, + "learning_rate": 5.400238000793336e-05, + "loss": 1.6773, + "step": 1110 + }, + { + "epoch": 7.37, + "learning_rate": 5.394287980959936e-05, + "loss": 1.631, + "step": 1120 + }, + { + "epoch": 7.43, + "learning_rate": 5.3883379611265374e-05, + "loss": 1.6511, + "step": 1130 + }, + { + "epoch": 7.5, + "learning_rate": 5.382387941293138e-05, + "loss": 1.6923, + "step": 1140 + }, + { + "epoch": 7.56, + "learning_rate": 5.376437921459739e-05, + "loss": 1.6204, + "step": 1150 + }, + { + "epoch": 7.63, + "learning_rate": 5.370487901626339e-05, + "loss": 1.6243, + "step": 1160 + }, + { + "epoch": 7.69, + "learning_rate": 5.3645378817929396e-05, + "loss": 1.6422, + "step": 1170 + }, + { + "epoch": 7.76, + "learning_rate": 5.35858786195954e-05, + "loss": 1.7036, + "step": 1180 + }, + { + "epoch": 7.82, + "learning_rate": 5.352637842126141e-05, + "loss": 1.5981, + "step": 1190 + }, + { + "epoch": 7.89, + "learning_rate": 5.346687822292741e-05, + "loss": 1.6271, + "step": 1200 + }, + { + "epoch": 7.96, + "learning_rate": 5.340737802459342e-05, + "loss": 1.5931, + "step": 1210 + }, + { + "epoch": 8.03, + "learning_rate": 5.3347877826259425e-05, + "loss": 1.79, + "step": 1220 + }, + { + "epoch": 8.09, + "learning_rate": 5.328837762792543e-05, + "loss": 1.5735, + "step": 1230 + }, + { + "epoch": 8.16, + "learning_rate": 5.322887742959143e-05, + "loss": 1.6044, + "step": 1240 + }, + { + "epoch": 8.22, + "learning_rate": 5.316937723125744e-05, + "loss": 1.5838, + "step": 1250 + }, + { + "epoch": 8.29, + "learning_rate": 5.310987703292345e-05, + "loss": 1.6321, + "step": 1260 + }, + { + "epoch": 8.35, + "learning_rate": 5.305037683458945e-05, + "loss": 1.58, + "step": 1270 + }, + { + "epoch": 8.42, + "learning_rate": 5.2990876636255455e-05, + "loss": 1.5729, + "step": 1280 + }, + { + "epoch": 8.48, + "learning_rate": 5.293137643792146e-05, + "loss": 1.6058, + "step": 1290 + }, + { + "epoch": 8.55, + "learning_rate": 5.287187623958747e-05, + "loss": 1.6066, + "step": 1300 + }, + { + "epoch": 8.62, + "learning_rate": 5.281237604125347e-05, + "loss": 1.5613, + "step": 1310 + }, + { + "epoch": 8.68, + "learning_rate": 5.275287584291948e-05, + "loss": 1.5781, + "step": 1320 + }, + { + "epoch": 8.75, + "learning_rate": 5.2693375644585484e-05, + "loss": 1.596, + "step": 1330 + }, + { + "epoch": 8.81, + "learning_rate": 5.263387544625149e-05, + "loss": 1.5344, + "step": 1340 + }, + { + "epoch": 8.88, + "learning_rate": 5.257437524791749e-05, + "loss": 1.5456, + "step": 1350 + }, + { + "epoch": 8.94, + "learning_rate": 5.25148750495835e-05, + "loss": 1.5413, + "step": 1360 + }, + { + "epoch": 9.01, + "learning_rate": 5.2455374851249506e-05, + "loss": 1.7312, + "step": 1370 + }, + { + "epoch": 9.08, + "learning_rate": 5.239587465291551e-05, + "loss": 1.5234, + "step": 1380 + }, + { + "epoch": 9.14, + "learning_rate": 5.2336374454581513e-05, + "loss": 1.5335, + "step": 1390 + }, + { + "epoch": 9.21, + "learning_rate": 5.227687425624752e-05, + "loss": 1.5345, + "step": 1400 + }, + { + "epoch": 9.27, + "learning_rate": 5.221737405791353e-05, + "loss": 1.6354, + "step": 1410 + }, + { + "epoch": 9.34, + "learning_rate": 5.215787385957953e-05, + "loss": 1.5181, + "step": 1420 + }, + { + "epoch": 9.41, + "learning_rate": 5.2098373661245535e-05, + "loss": 1.5537, + "step": 1430 + }, + { + "epoch": 9.47, + "learning_rate": 5.203887346291154e-05, + "loss": 1.521, + "step": 1440 + }, + { + "epoch": 9.54, + "learning_rate": 5.197937326457755e-05, + "loss": 1.5633, + "step": 1450 + }, + { + "epoch": 9.6, + "learning_rate": 5.191987306624355e-05, + "loss": 1.5192, + "step": 1460 + }, + { + "epoch": 9.67, + "learning_rate": 5.186037286790956e-05, + "loss": 1.5376, + "step": 1470 + }, + { + "epoch": 9.73, + "learning_rate": 5.1800872669575565e-05, + "loss": 1.5484, + "step": 1480 + }, + { + "epoch": 9.8, + "learning_rate": 5.174137247124157e-05, + "loss": 1.5301, + "step": 1490 + }, + { + "epoch": 9.86, + "learning_rate": 5.168187227290757e-05, + "loss": 1.4907, + "step": 1500 + }, + { + "epoch": 9.86, + "eval_loss": 0.6554940342903137, + "eval_runtime": 311.8278, + "eval_samples_per_second": 22.801, + "eval_steps_per_second": 1.427, + "eval_wer": 0.6672602342011033, + "step": 1500 + }, + { + "epoch": 9.93, + "learning_rate": 5.1622372074573586e-05, + "loss": 1.476, + "step": 1510 + }, + { + "epoch": 10.0, + "learning_rate": 5.156287187623959e-05, + "loss": 1.5746, + "step": 1520 + }, + { + "epoch": 10.07, + "learning_rate": 5.15033716779056e-05, + "loss": 1.5885, + "step": 1530 + }, + { + "epoch": 10.13, + "learning_rate": 5.14438714795716e-05, + "loss": 1.4942, + "step": 1540 + }, + { + "epoch": 10.2, + "learning_rate": 5.138437128123761e-05, + "loss": 1.5069, + "step": 1550 + }, + { + "epoch": 10.26, + "learning_rate": 5.1324871082903615e-05, + "loss": 1.5527, + "step": 1560 + }, + { + "epoch": 10.33, + "learning_rate": 5.1265370884569616e-05, + "loss": 1.4691, + "step": 1570 + }, + { + "epoch": 10.39, + "learning_rate": 5.120587068623562e-05, + "loss": 1.5139, + "step": 1580 + }, + { + "epoch": 10.46, + "learning_rate": 5.114637048790163e-05, + "loss": 1.4942, + "step": 1590 + }, + { + "epoch": 10.52, + "learning_rate": 5.108687028956764e-05, + "loss": 1.5169, + "step": 1600 + }, + { + "epoch": 10.59, + "learning_rate": 5.102737009123364e-05, + "loss": 1.4577, + "step": 1610 + }, + { + "epoch": 10.65, + "learning_rate": 5.0967869892899645e-05, + "loss": 1.4895, + "step": 1620 + }, + { + "epoch": 10.72, + "learning_rate": 5.090836969456565e-05, + "loss": 1.4558, + "step": 1630 + }, + { + "epoch": 10.79, + "learning_rate": 5.084886949623166e-05, + "loss": 1.547, + "step": 1640 + }, + { + "epoch": 10.85, + "learning_rate": 5.078936929789766e-05, + "loss": 1.4679, + "step": 1650 + }, + { + "epoch": 10.92, + "learning_rate": 5.072986909956367e-05, + "loss": 1.4676, + "step": 1660 + }, + { + "epoch": 10.98, + "learning_rate": 5.0670368901229674e-05, + "loss": 1.496, + "step": 1670 + }, + { + "epoch": 11.05, + "learning_rate": 5.061086870289568e-05, + "loss": 1.6186, + "step": 1680 + }, + { + "epoch": 11.12, + "learning_rate": 5.055136850456168e-05, + "loss": 1.4611, + "step": 1690 + }, + { + "epoch": 11.18, + "learning_rate": 5.049186830622769e-05, + "loss": 1.4517, + "step": 1700 + }, + { + "epoch": 11.25, + "learning_rate": 5.0432368107893696e-05, + "loss": 1.4987, + "step": 1710 + }, + { + "epoch": 11.31, + "learning_rate": 5.0372867909559696e-05, + "loss": 1.4539, + "step": 1720 + }, + { + "epoch": 11.38, + "learning_rate": 5.03133677112257e-05, + "loss": 1.4677, + "step": 1730 + }, + { + "epoch": 11.45, + "learning_rate": 5.025386751289171e-05, + "loss": 1.4523, + "step": 1740 + }, + { + "epoch": 11.51, + "learning_rate": 5.019436731455772e-05, + "loss": 1.5092, + "step": 1750 + }, + { + "epoch": 11.58, + "learning_rate": 5.013486711622372e-05, + "loss": 1.4423, + "step": 1760 + }, + { + "epoch": 11.64, + "learning_rate": 5.0075366917889725e-05, + "loss": 1.4498, + "step": 1770 + }, + { + "epoch": 11.71, + "learning_rate": 5.001586671955573e-05, + "loss": 1.4076, + "step": 1780 + }, + { + "epoch": 11.77, + "learning_rate": 4.995636652122174e-05, + "loss": 1.5411, + "step": 1790 + }, + { + "epoch": 11.84, + "learning_rate": 4.989686632288774e-05, + "loss": 1.4117, + "step": 1800 + }, + { + "epoch": 11.9, + "learning_rate": 4.983736612455375e-05, + "loss": 1.4459, + "step": 1810 + }, + { + "epoch": 11.97, + "learning_rate": 4.9777865926219754e-05, + "loss": 1.4549, + "step": 1820 + }, + { + "epoch": 12.04, + "learning_rate": 4.971836572788576e-05, + "loss": 1.5976, + "step": 1830 + }, + { + "epoch": 12.1, + "learning_rate": 4.965886552955176e-05, + "loss": 1.4083, + "step": 1840 + }, + { + "epoch": 12.17, + "learning_rate": 4.959936533121777e-05, + "loss": 1.4588, + "step": 1850 + }, + { + "epoch": 12.24, + "learning_rate": 4.9539865132883776e-05, + "loss": 1.4425, + "step": 1860 + }, + { + "epoch": 12.3, + "learning_rate": 4.948036493454978e-05, + "loss": 1.4567, + "step": 1870 + }, + { + "epoch": 12.37, + "learning_rate": 4.9420864736215784e-05, + "loss": 1.4363, + "step": 1880 + }, + { + "epoch": 12.43, + "learning_rate": 4.93613645378818e-05, + "loss": 1.4431, + "step": 1890 + }, + { + "epoch": 12.5, + "learning_rate": 4.9301864339547805e-05, + "loss": 1.4858, + "step": 1900 + }, + { + "epoch": 12.56, + "learning_rate": 4.9242364141213805e-05, + "loss": 1.4147, + "step": 1910 + }, + { + "epoch": 12.63, + "learning_rate": 4.918286394287981e-05, + "loss": 1.4268, + "step": 1920 + }, + { + "epoch": 12.69, + "learning_rate": 4.912336374454582e-05, + "loss": 1.4174, + "step": 1930 + }, + { + "epoch": 12.76, + "learning_rate": 4.906386354621183e-05, + "loss": 1.4528, + "step": 1940 + }, + { + "epoch": 12.82, + "learning_rate": 4.900436334787783e-05, + "loss": 1.4058, + "step": 1950 + }, + { + "epoch": 12.89, + "learning_rate": 4.8944863149543835e-05, + "loss": 1.441, + "step": 1960 + }, + { + "epoch": 12.96, + "learning_rate": 4.888536295120984e-05, + "loss": 1.3988, + "step": 1970 + }, + { + "epoch": 13.03, + "learning_rate": 4.882586275287585e-05, + "loss": 1.586, + "step": 1980 + }, + { + "epoch": 13.09, + "learning_rate": 4.876636255454185e-05, + "loss": 1.3778, + "step": 1990 + }, + { + "epoch": 13.16, + "learning_rate": 4.8706862356207857e-05, + "loss": 1.4177, + "step": 2000 + }, + { + "epoch": 13.16, + "eval_loss": 0.5784419775009155, + "eval_runtime": 314.6324, + "eval_samples_per_second": 22.598, + "eval_steps_per_second": 1.414, + "eval_wer": 0.6075679860640666, + "step": 2000 + }, + { + "epoch": 13.22, + "learning_rate": 4.8647362157873864e-05, + "loss": 1.4404, + "step": 2010 + }, + { + "epoch": 13.29, + "learning_rate": 4.8587861959539864e-05, + "loss": 1.4586, + "step": 2020 + }, + { + "epoch": 13.35, + "learning_rate": 4.852836176120587e-05, + "loss": 1.4015, + "step": 2030 + }, + { + "epoch": 13.42, + "learning_rate": 4.846886156287188e-05, + "loss": 1.4202, + "step": 2040 + }, + { + "epoch": 13.48, + "learning_rate": 4.8409361364537886e-05, + "loss": 1.4328, + "step": 2050 + }, + { + "epoch": 13.55, + "learning_rate": 4.8349861166203886e-05, + "loss": 1.4464, + "step": 2060 + }, + { + "epoch": 13.62, + "learning_rate": 4.829036096786989e-05, + "loss": 1.4048, + "step": 2070 + }, + { + "epoch": 13.68, + "learning_rate": 4.82308607695359e-05, + "loss": 1.3997, + "step": 2080 + }, + { + "epoch": 13.75, + "learning_rate": 4.817136057120191e-05, + "loss": 1.446, + "step": 2090 + }, + { + "epoch": 13.81, + "learning_rate": 4.811186037286791e-05, + "loss": 1.391, + "step": 2100 + }, + { + "epoch": 13.88, + "learning_rate": 4.8052360174533915e-05, + "loss": 1.4091, + "step": 2110 + }, + { + "epoch": 13.94, + "learning_rate": 4.799285997619992e-05, + "loss": 1.4267, + "step": 2120 + }, + { + "epoch": 14.01, + "learning_rate": 4.793335977786593e-05, + "loss": 1.5417, + "step": 2130 + }, + { + "epoch": 14.08, + "learning_rate": 4.787385957953193e-05, + "loss": 1.3711, + "step": 2140 + }, + { + "epoch": 14.14, + "learning_rate": 4.781435938119794e-05, + "loss": 1.3764, + "step": 2150 + }, + { + "epoch": 14.21, + "learning_rate": 4.7754859182863944e-05, + "loss": 1.3796, + "step": 2160 + }, + { + "epoch": 14.27, + "learning_rate": 4.7695358984529945e-05, + "loss": 1.4684, + "step": 2170 + }, + { + "epoch": 14.34, + "learning_rate": 4.763585878619595e-05, + "loss": 1.3748, + "step": 2180 + }, + { + "epoch": 14.41, + "learning_rate": 4.757635858786196e-05, + "loss": 1.3977, + "step": 2190 + }, + { + "epoch": 14.47, + "learning_rate": 4.7516858389527966e-05, + "loss": 1.4021, + "step": 2200 + }, + { + "epoch": 14.54, + "learning_rate": 4.745735819119397e-05, + "loss": 1.4422, + "step": 2210 + }, + { + "epoch": 14.6, + "learning_rate": 4.7397857992859974e-05, + "loss": 1.3541, + "step": 2220 + }, + { + "epoch": 14.67, + "learning_rate": 4.733835779452598e-05, + "loss": 1.4021, + "step": 2230 + }, + { + "epoch": 14.73, + "learning_rate": 4.727885759619199e-05, + "loss": 1.4237, + "step": 2240 + }, + { + "epoch": 14.8, + "learning_rate": 4.721935739785799e-05, + "loss": 1.3847, + "step": 2250 + }, + { + "epoch": 14.86, + "learning_rate": 4.7159857199523996e-05, + "loss": 1.365, + "step": 2260 + }, + { + "epoch": 14.93, + "learning_rate": 4.710035700119001e-05, + "loss": 1.3724, + "step": 2270 + }, + { + "epoch": 15.0, + "learning_rate": 4.704085680285602e-05, + "loss": 1.4679, + "step": 2280 + }, + { + "epoch": 15.07, + "learning_rate": 4.698135660452202e-05, + "loss": 1.4707, + "step": 2290 + }, + { + "epoch": 15.13, + "learning_rate": 4.6921856406188024e-05, + "loss": 1.3633, + "step": 2300 + }, + { + "epoch": 15.2, + "learning_rate": 4.686235620785403e-05, + "loss": 1.3463, + "step": 2310 + }, + { + "epoch": 15.26, + "learning_rate": 4.680285600952003e-05, + "loss": 1.4541, + "step": 2320 + }, + { + "epoch": 15.33, + "learning_rate": 4.674335581118604e-05, + "loss": 1.3625, + "step": 2330 + }, + { + "epoch": 15.39, + "learning_rate": 4.6683855612852046e-05, + "loss": 1.3401, + "step": 2340 + }, + { + "epoch": 15.46, + "learning_rate": 4.6624355414518054e-05, + "loss": 1.3935, + "step": 2350 + }, + { + "epoch": 15.52, + "learning_rate": 4.6564855216184054e-05, + "loss": 1.4198, + "step": 2360 + }, + { + "epoch": 15.59, + "learning_rate": 4.650535501785006e-05, + "loss": 1.3487, + "step": 2370 + }, + { + "epoch": 15.65, + "learning_rate": 4.644585481951607e-05, + "loss": 1.3766, + "step": 2380 + }, + { + "epoch": 15.72, + "learning_rate": 4.6386354621182076e-05, + "loss": 1.3678, + "step": 2390 + }, + { + "epoch": 15.79, + "learning_rate": 4.6326854422848076e-05, + "loss": 1.3904, + "step": 2400 + }, + { + "epoch": 15.85, + "learning_rate": 4.626735422451408e-05, + "loss": 1.3508, + "step": 2410 + }, + { + "epoch": 15.92, + "learning_rate": 4.620785402618009e-05, + "loss": 1.383, + "step": 2420 + }, + { + "epoch": 15.98, + "learning_rate": 4.61483538278461e-05, + "loss": 1.3906, + "step": 2430 + }, + { + "epoch": 16.05, + "learning_rate": 4.60888536295121e-05, + "loss": 1.4781, + "step": 2440 + }, + { + "epoch": 16.12, + "learning_rate": 4.6029353431178105e-05, + "loss": 1.3411, + "step": 2450 + }, + { + "epoch": 16.18, + "learning_rate": 4.596985323284411e-05, + "loss": 1.3709, + "step": 2460 + }, + { + "epoch": 16.25, + "learning_rate": 4.591035303451011e-05, + "loss": 1.395, + "step": 2470 + }, + { + "epoch": 16.31, + "learning_rate": 4.585085283617612e-05, + "loss": 1.3344, + "step": 2480 + }, + { + "epoch": 16.38, + "learning_rate": 4.579135263784213e-05, + "loss": 1.3411, + "step": 2490 + }, + { + "epoch": 16.45, + "learning_rate": 4.5731852439508134e-05, + "loss": 1.3425, + "step": 2500 + }, + { + "epoch": 16.45, + "eval_loss": 0.5378673672676086, + "eval_runtime": 315.1097, + "eval_samples_per_second": 22.564, + "eval_steps_per_second": 1.412, + "eval_wer": 0.5718184457563147, + "step": 2500 + }, + { + "epoch": 16.51, + "learning_rate": 4.5672352241174135e-05, + "loss": 1.3892, + "step": 2510 + }, + { + "epoch": 16.58, + "learning_rate": 4.561285204284014e-05, + "loss": 1.3285, + "step": 2520 + }, + { + "epoch": 16.64, + "learning_rate": 4.555335184450615e-05, + "loss": 1.3038, + "step": 2530 + }, + { + "epoch": 16.71, + "learning_rate": 4.5493851646172156e-05, + "loss": 1.3316, + "step": 2540 + }, + { + "epoch": 16.77, + "learning_rate": 4.5434351447838157e-05, + "loss": 1.3842, + "step": 2550 + }, + { + "epoch": 16.84, + "learning_rate": 4.5374851249504164e-05, + "loss": 1.3268, + "step": 2560 + }, + { + "epoch": 16.9, + "learning_rate": 4.531535105117017e-05, + "loss": 1.3091, + "step": 2570 + }, + { + "epoch": 16.97, + "learning_rate": 4.525585085283618e-05, + "loss": 1.3346, + "step": 2580 + }, + { + "epoch": 17.04, + "learning_rate": 4.519635065450218e-05, + "loss": 1.4809, + "step": 2590 + }, + { + "epoch": 17.1, + "learning_rate": 4.5136850456168186e-05, + "loss": 1.3026, + "step": 2600 + }, + { + "epoch": 17.17, + "learning_rate": 4.507735025783419e-05, + "loss": 1.332, + "step": 2610 + }, + { + "epoch": 17.24, + "learning_rate": 4.501785005950019e-05, + "loss": 1.3768, + "step": 2620 + }, + { + "epoch": 17.3, + "learning_rate": 4.49583498611662e-05, + "loss": 1.3643, + "step": 2630 + }, + { + "epoch": 17.37, + "learning_rate": 4.489884966283221e-05, + "loss": 1.3337, + "step": 2640 + }, + { + "epoch": 17.43, + "learning_rate": 4.483934946449822e-05, + "loss": 1.3713, + "step": 2650 + }, + { + "epoch": 17.5, + "learning_rate": 4.477984926616422e-05, + "loss": 1.3875, + "step": 2660 + }, + { + "epoch": 17.56, + "learning_rate": 4.472034906783023e-05, + "loss": 1.3109, + "step": 2670 + }, + { + "epoch": 17.63, + "learning_rate": 4.4660848869496236e-05, + "loss": 1.2984, + "step": 2680 + }, + { + "epoch": 17.69, + "learning_rate": 4.4601348671162243e-05, + "loss": 1.3055, + "step": 2690 + }, + { + "epoch": 17.76, + "learning_rate": 4.4541848472828244e-05, + "loss": 1.3925, + "step": 2700 + }, + { + "epoch": 17.82, + "learning_rate": 4.448234827449425e-05, + "loss": 1.2958, + "step": 2710 + }, + { + "epoch": 17.89, + "learning_rate": 4.442284807616026e-05, + "loss": 1.3094, + "step": 2720 + }, + { + "epoch": 17.96, + "learning_rate": 4.4363347877826265e-05, + "loss": 1.3485, + "step": 2730 + }, + { + "epoch": 18.03, + "learning_rate": 4.4303847679492266e-05, + "loss": 1.4891, + "step": 2740 + }, + { + "epoch": 18.09, + "learning_rate": 4.424434748115827e-05, + "loss": 1.3315, + "step": 2750 + }, + { + "epoch": 18.16, + "learning_rate": 4.418484728282428e-05, + "loss": 1.3337, + "step": 2760 + }, + { + "epoch": 18.22, + "learning_rate": 4.412534708449028e-05, + "loss": 1.3189, + "step": 2770 + }, + { + "epoch": 18.29, + "learning_rate": 4.406584688615629e-05, + "loss": 1.3324, + "step": 2780 + }, + { + "epoch": 18.35, + "learning_rate": 4.4006346687822295e-05, + "loss": 1.2899, + "step": 2790 + }, + { + "epoch": 18.42, + "learning_rate": 4.39468464894883e-05, + "loss": 1.3146, + "step": 2800 + }, + { + "epoch": 18.48, + "learning_rate": 4.38873462911543e-05, + "loss": 1.3373, + "step": 2810 + }, + { + "epoch": 18.55, + "learning_rate": 4.382784609282031e-05, + "loss": 1.3389, + "step": 2820 + }, + { + "epoch": 18.62, + "learning_rate": 4.376834589448632e-05, + "loss": 1.3078, + "step": 2830 + }, + { + "epoch": 18.68, + "learning_rate": 4.3708845696152324e-05, + "loss": 1.3183, + "step": 2840 + }, + { + "epoch": 18.75, + "learning_rate": 4.3649345497818324e-05, + "loss": 1.3632, + "step": 2850 + }, + { + "epoch": 18.81, + "learning_rate": 4.358984529948433e-05, + "loss": 1.2758, + "step": 2860 + }, + { + "epoch": 18.88, + "learning_rate": 4.353034510115034e-05, + "loss": 1.302, + "step": 2870 + }, + { + "epoch": 18.94, + "learning_rate": 4.3470844902816346e-05, + "loss": 1.3001, + "step": 2880 + }, + { + "epoch": 19.01, + "learning_rate": 4.3411344704482346e-05, + "loss": 1.4967, + "step": 2890 + }, + { + "epoch": 19.08, + "learning_rate": 4.3351844506148354e-05, + "loss": 1.2879, + "step": 2900 + }, + { + "epoch": 19.14, + "learning_rate": 4.329234430781436e-05, + "loss": 1.3083, + "step": 2910 + }, + { + "epoch": 19.21, + "learning_rate": 4.323284410948036e-05, + "loss": 1.2781, + "step": 2920 + }, + { + "epoch": 19.27, + "learning_rate": 4.317334391114637e-05, + "loss": 1.3786, + "step": 2930 + }, + { + "epoch": 19.34, + "learning_rate": 4.3113843712812376e-05, + "loss": 1.2858, + "step": 2940 + }, + { + "epoch": 19.41, + "learning_rate": 4.305434351447838e-05, + "loss": 1.3195, + "step": 2950 + }, + { + "epoch": 19.47, + "learning_rate": 4.299484331614438e-05, + "loss": 1.3097, + "step": 2960 + }, + { + "epoch": 19.54, + "learning_rate": 4.293534311781039e-05, + "loss": 1.3194, + "step": 2970 + }, + { + "epoch": 19.6, + "learning_rate": 4.28758429194764e-05, + "loss": 1.2795, + "step": 2980 + }, + { + "epoch": 19.67, + "learning_rate": 4.2816342721142405e-05, + "loss": 1.3051, + "step": 2990 + }, + { + "epoch": 19.73, + "learning_rate": 4.2756842522808405e-05, + "loss": 1.33, + "step": 3000 + }, + { + "epoch": 19.73, + "eval_loss": 0.49618762731552124, + "eval_runtime": 314.2943, + "eval_samples_per_second": 22.622, + "eval_steps_per_second": 1.416, + "eval_wer": 0.5245330494532082, + "step": 3000 + }, + { + "epoch": 19.8, + "learning_rate": 4.269734232447441e-05, + "loss": 1.3265, + "step": 3010 + }, + { + "epoch": 19.86, + "learning_rate": 4.263784212614042e-05, + "loss": 1.3148, + "step": 3020 + }, + { + "epoch": 19.93, + "learning_rate": 4.257834192780643e-05, + "loss": 1.3019, + "step": 3030 + }, + { + "epoch": 20.0, + "learning_rate": 4.2518841729472434e-05, + "loss": 1.3688, + "step": 3040 + }, + { + "epoch": 20.07, + "learning_rate": 4.245934153113844e-05, + "loss": 1.357, + "step": 3050 + }, + { + "epoch": 20.13, + "learning_rate": 4.239984133280445e-05, + "loss": 1.2812, + "step": 3060 + }, + { + "epoch": 20.2, + "learning_rate": 4.234034113447045e-05, + "loss": 1.2739, + "step": 3070 + }, + { + "epoch": 20.26, + "learning_rate": 4.2280840936136456e-05, + "loss": 1.3482, + "step": 3080 + }, + { + "epoch": 20.33, + "learning_rate": 4.222134073780246e-05, + "loss": 1.2779, + "step": 3090 + }, + { + "epoch": 20.39, + "learning_rate": 4.216184053946847e-05, + "loss": 1.3029, + "step": 3100 + }, + { + "epoch": 20.46, + "learning_rate": 4.210234034113447e-05, + "loss": 1.3525, + "step": 3110 + }, + { + "epoch": 20.52, + "learning_rate": 4.204284014280048e-05, + "loss": 1.3344, + "step": 3120 + }, + { + "epoch": 20.59, + "learning_rate": 4.1983339944466485e-05, + "loss": 1.2693, + "step": 3130 + }, + { + "epoch": 20.65, + "learning_rate": 4.192383974613249e-05, + "loss": 1.2751, + "step": 3140 + }, + { + "epoch": 20.72, + "learning_rate": 4.186433954779849e-05, + "loss": 1.2668, + "step": 3150 + }, + { + "epoch": 20.79, + "learning_rate": 4.18048393494645e-05, + "loss": 1.3379, + "step": 3160 + }, + { + "epoch": 20.85, + "learning_rate": 4.174533915113051e-05, + "loss": 1.2763, + "step": 3170 + }, + { + "epoch": 20.92, + "learning_rate": 4.1685838952796514e-05, + "loss": 1.2917, + "step": 3180 + }, + { + "epoch": 20.98, + "learning_rate": 4.1626338754462514e-05, + "loss": 1.3203, + "step": 3190 + }, + { + "epoch": 21.05, + "learning_rate": 4.156683855612852e-05, + "loss": 1.3707, + "step": 3200 + }, + { + "epoch": 21.12, + "learning_rate": 4.150733835779453e-05, + "loss": 1.265, + "step": 3210 + }, + { + "epoch": 21.18, + "learning_rate": 4.144783815946053e-05, + "loss": 1.268, + "step": 3220 + }, + { + "epoch": 21.25, + "learning_rate": 4.1388337961126536e-05, + "loss": 1.3269, + "step": 3230 + }, + { + "epoch": 21.31, + "learning_rate": 4.1328837762792543e-05, + "loss": 1.2943, + "step": 3240 + }, + { + "epoch": 21.38, + "learning_rate": 4.126933756445855e-05, + "loss": 1.269, + "step": 3250 + }, + { + "epoch": 21.45, + "learning_rate": 4.120983736612455e-05, + "loss": 1.2659, + "step": 3260 + }, + { + "epoch": 21.51, + "learning_rate": 4.115033716779056e-05, + "loss": 1.3543, + "step": 3270 + }, + { + "epoch": 21.58, + "learning_rate": 4.1090836969456565e-05, + "loss": 1.2307, + "step": 3280 + }, + { + "epoch": 21.64, + "learning_rate": 4.103133677112257e-05, + "loss": 1.2711, + "step": 3290 + }, + { + "epoch": 21.71, + "learning_rate": 4.097183657278857e-05, + "loss": 1.2897, + "step": 3300 + }, + { + "epoch": 21.77, + "learning_rate": 4.091233637445458e-05, + "loss": 1.2884, + "step": 3310 + }, + { + "epoch": 21.84, + "learning_rate": 4.085283617612059e-05, + "loss": 1.2466, + "step": 3320 + }, + { + "epoch": 21.9, + "learning_rate": 4.0793335977786595e-05, + "loss": 1.2744, + "step": 3330 + }, + { + "epoch": 21.97, + "learning_rate": 4.0733835779452595e-05, + "loss": 1.2278, + "step": 3340 + }, + { + "epoch": 22.04, + "learning_rate": 4.06743355811186e-05, + "loss": 1.416, + "step": 3350 + }, + { + "epoch": 22.1, + "learning_rate": 4.061483538278461e-05, + "loss": 1.2256, + "step": 3360 + }, + { + "epoch": 22.17, + "learning_rate": 4.055533518445061e-05, + "loss": 1.2733, + "step": 3370 + }, + { + "epoch": 22.24, + "learning_rate": 4.049583498611662e-05, + "loss": 1.3062, + "step": 3380 + }, + { + "epoch": 22.3, + "learning_rate": 4.0436334787782624e-05, + "loss": 1.2772, + "step": 3390 + }, + { + "epoch": 22.37, + "learning_rate": 4.037683458944863e-05, + "loss": 1.2667, + "step": 3400 + }, + { + "epoch": 22.43, + "learning_rate": 4.031733439111464e-05, + "loss": 1.2501, + "step": 3410 + }, + { + "epoch": 22.5, + "learning_rate": 4.0257834192780646e-05, + "loss": 1.3233, + "step": 3420 + }, + { + "epoch": 22.56, + "learning_rate": 4.019833399444665e-05, + "loss": 1.2382, + "step": 3430 + }, + { + "epoch": 22.63, + "learning_rate": 4.013883379611266e-05, + "loss": 1.2485, + "step": 3440 + }, + { + "epoch": 22.69, + "learning_rate": 4.007933359777866e-05, + "loss": 1.2442, + "step": 3450 + }, + { + "epoch": 22.76, + "learning_rate": 4.001983339944467e-05, + "loss": 1.3278, + "step": 3460 + }, + { + "epoch": 22.82, + "learning_rate": 3.9960333201110675e-05, + "loss": 1.2325, + "step": 3470 + }, + { + "epoch": 22.89, + "learning_rate": 3.990083300277668e-05, + "loss": 1.276, + "step": 3480 + }, + { + "epoch": 22.96, + "learning_rate": 3.984133280444268e-05, + "loss": 1.27, + "step": 3490 + }, + { + "epoch": 23.03, + "learning_rate": 3.978183260610869e-05, + "loss": 1.4378, + "step": 3500 + }, + { + "epoch": 23.03, + "eval_loss": 0.46987584233283997, + "eval_runtime": 315.8211, + "eval_samples_per_second": 22.513, + "eval_steps_per_second": 1.409, + "eval_wer": 0.5097841865866641, + "step": 3500 + }, + { + "epoch": 23.09, + "learning_rate": 3.97223324077747e-05, + "loss": 1.2222, + "step": 3510 + }, + { + "epoch": 23.16, + "learning_rate": 3.96628322094407e-05, + "loss": 1.2432, + "step": 3520 + }, + { + "epoch": 23.22, + "learning_rate": 3.9603332011106704e-05, + "loss": 1.236, + "step": 3530 + }, + { + "epoch": 23.29, + "learning_rate": 3.954383181277271e-05, + "loss": 1.2867, + "step": 3540 + }, + { + "epoch": 23.35, + "learning_rate": 3.948433161443872e-05, + "loss": 1.2229, + "step": 3550 + }, + { + "epoch": 23.42, + "learning_rate": 3.942483141610472e-05, + "loss": 1.2702, + "step": 3560 + }, + { + "epoch": 23.48, + "learning_rate": 3.9365331217770726e-05, + "loss": 1.273, + "step": 3570 + }, + { + "epoch": 23.55, + "learning_rate": 3.930583101943673e-05, + "loss": 1.278, + "step": 3580 + }, + { + "epoch": 23.62, + "learning_rate": 3.924633082110274e-05, + "loss": 1.2205, + "step": 3590 + }, + { + "epoch": 23.68, + "learning_rate": 3.918683062276874e-05, + "loss": 1.267, + "step": 3600 + }, + { + "epoch": 23.75, + "learning_rate": 3.912733042443475e-05, + "loss": 1.2885, + "step": 3610 + }, + { + "epoch": 23.81, + "learning_rate": 3.9067830226100755e-05, + "loss": 1.2384, + "step": 3620 + }, + { + "epoch": 23.88, + "learning_rate": 3.900833002776676e-05, + "loss": 1.2373, + "step": 3630 + }, + { + "epoch": 23.94, + "learning_rate": 3.894882982943276e-05, + "loss": 1.2617, + "step": 3640 + }, + { + "epoch": 24.01, + "learning_rate": 3.888932963109877e-05, + "loss": 1.412, + "step": 3650 + }, + { + "epoch": 24.08, + "learning_rate": 3.882982943276478e-05, + "loss": 1.1934, + "step": 3660 + }, + { + "epoch": 24.14, + "learning_rate": 3.877032923443078e-05, + "loss": 1.2418, + "step": 3670 + }, + { + "epoch": 24.21, + "learning_rate": 3.8710829036096785e-05, + "loss": 1.2606, + "step": 3680 + }, + { + "epoch": 24.27, + "learning_rate": 3.865132883776279e-05, + "loss": 1.3288, + "step": 3690 + }, + { + "epoch": 24.34, + "learning_rate": 3.85918286394288e-05, + "loss": 1.2383, + "step": 3700 + }, + { + "epoch": 24.41, + "learning_rate": 3.85323284410948e-05, + "loss": 1.2176, + "step": 3710 + }, + { + "epoch": 24.47, + "learning_rate": 3.847282824276081e-05, + "loss": 1.2255, + "step": 3720 + }, + { + "epoch": 24.54, + "learning_rate": 3.8413328044426814e-05, + "loss": 1.308, + "step": 3730 + }, + { + "epoch": 24.6, + "learning_rate": 3.835382784609282e-05, + "loss": 1.2243, + "step": 3740 + }, + { + "epoch": 24.67, + "learning_rate": 3.829432764775882e-05, + "loss": 1.2591, + "step": 3750 + }, + { + "epoch": 24.73, + "learning_rate": 3.823482744942483e-05, + "loss": 1.2523, + "step": 3760 + }, + { + "epoch": 24.8, + "learning_rate": 3.8175327251090836e-05, + "loss": 1.2499, + "step": 3770 + }, + { + "epoch": 24.86, + "learning_rate": 3.811582705275684e-05, + "loss": 1.2022, + "step": 3780 + }, + { + "epoch": 24.93, + "learning_rate": 3.805632685442285e-05, + "loss": 1.254, + "step": 3790 + }, + { + "epoch": 25.0, + "learning_rate": 3.799682665608886e-05, + "loss": 1.3321, + "step": 3800 + }, + { + "epoch": 25.07, + "learning_rate": 3.7937326457754865e-05, + "loss": 1.3216, + "step": 3810 + }, + { + "epoch": 25.13, + "learning_rate": 3.7877826259420865e-05, + "loss": 1.2439, + "step": 3820 + }, + { + "epoch": 25.2, + "learning_rate": 3.781832606108687e-05, + "loss": 1.2297, + "step": 3830 + }, + { + "epoch": 25.26, + "learning_rate": 3.775882586275288e-05, + "loss": 1.2818, + "step": 3840 + }, + { + "epoch": 25.33, + "learning_rate": 3.7699325664418887e-05, + "loss": 1.2019, + "step": 3850 + }, + { + "epoch": 25.39, + "learning_rate": 3.763982546608489e-05, + "loss": 1.2258, + "step": 3860 + }, + { + "epoch": 25.46, + "learning_rate": 3.7580325267750894e-05, + "loss": 1.2137, + "step": 3870 + }, + { + "epoch": 25.52, + "learning_rate": 3.75208250694169e-05, + "loss": 1.2879, + "step": 3880 + }, + { + "epoch": 25.59, + "learning_rate": 3.746132487108291e-05, + "loss": 1.199, + "step": 3890 + }, + { + "epoch": 25.65, + "learning_rate": 3.740182467274891e-05, + "loss": 1.2393, + "step": 3900 + }, + { + "epoch": 25.72, + "learning_rate": 3.7342324474414916e-05, + "loss": 1.2505, + "step": 3910 + }, + { + "epoch": 25.79, + "learning_rate": 3.728282427608092e-05, + "loss": 1.2483, + "step": 3920 + }, + { + "epoch": 25.85, + "learning_rate": 3.722332407774693e-05, + "loss": 1.1909, + "step": 3930 + }, + { + "epoch": 25.92, + "learning_rate": 3.716382387941293e-05, + "loss": 1.2198, + "step": 3940 + }, + { + "epoch": 25.98, + "learning_rate": 3.710432368107894e-05, + "loss": 1.2687, + "step": 3950 + }, + { + "epoch": 26.05, + "learning_rate": 3.7044823482744945e-05, + "loss": 1.3287, + "step": 3960 + }, + { + "epoch": 26.12, + "learning_rate": 3.6985323284410946e-05, + "loss": 1.206, + "step": 3970 + }, + { + "epoch": 26.18, + "learning_rate": 3.692582308607695e-05, + "loss": 1.229, + "step": 3980 + }, + { + "epoch": 26.25, + "learning_rate": 3.686632288774296e-05, + "loss": 1.2915, + "step": 3990 + }, + { + "epoch": 26.31, + "learning_rate": 3.680682268940897e-05, + "loss": 1.1894, + "step": 4000 + }, + { + "epoch": 26.31, + "eval_loss": 0.4526805579662323, + "eval_runtime": 315.4313, + "eval_samples_per_second": 22.541, + "eval_steps_per_second": 1.411, + "eval_wer": 0.48479628375108874, + "step": 4000 + }, + { + "epoch": 26.38, + "learning_rate": 4.739083673908368e-05, + "loss": 1.1884, + "step": 4010 + }, + { + "epoch": 26.45, + "learning_rate": 4.7358571735857174e-05, + "loss": 1.2414, + "step": 4020 + }, + { + "epoch": 26.51, + "learning_rate": 4.7326306732630676e-05, + "loss": 1.2956, + "step": 4030 + }, + { + "epoch": 26.58, + "learning_rate": 4.729404172940417e-05, + "loss": 1.2162, + "step": 4040 + }, + { + "epoch": 26.64, + "learning_rate": 4.726177672617767e-05, + "loss": 1.2153, + "step": 4050 + }, + { + "epoch": 26.71, + "learning_rate": 4.7229511722951175e-05, + "loss": 1.2511, + "step": 4060 + }, + { + "epoch": 26.77, + "learning_rate": 4.7197246719724676e-05, + "loss": 1.2947, + "step": 4070 + }, + { + "epoch": 26.84, + "learning_rate": 4.716498171649817e-05, + "loss": 1.1932, + "step": 4080 + }, + { + "epoch": 26.9, + "learning_rate": 4.713271671327167e-05, + "loss": 1.2163, + "step": 4090 + }, + { + "epoch": 26.97, + "learning_rate": 4.710045171004517e-05, + "loss": 1.2215, + "step": 4100 + }, + { + "epoch": 27.04, + "learning_rate": 4.706818670681867e-05, + "loss": 1.3785, + "step": 4110 + }, + { + "epoch": 27.1, + "learning_rate": 4.703592170359217e-05, + "loss": 1.2023, + "step": 4120 + }, + { + "epoch": 27.17, + "learning_rate": 4.7003656700365674e-05, + "loss": 1.2092, + "step": 4130 + }, + { + "epoch": 27.24, + "learning_rate": 4.697139169713917e-05, + "loss": 1.2561, + "step": 4140 + }, + { + "epoch": 27.3, + "learning_rate": 4.693912669391267e-05, + "loss": 1.231, + "step": 4150 + }, + { + "epoch": 27.37, + "learning_rate": 4.690686169068617e-05, + "loss": 1.2007, + "step": 4160 + }, + { + "epoch": 27.43, + "learning_rate": 4.687459668745967e-05, + "loss": 1.2063, + "step": 4170 + }, + { + "epoch": 27.5, + "learning_rate": 4.684233168423317e-05, + "loss": 1.2924, + "step": 4180 + }, + { + "epoch": 27.56, + "learning_rate": 4.681006668100667e-05, + "loss": 1.1982, + "step": 4190 + }, + { + "epoch": 27.63, + "learning_rate": 4.677780167778017e-05, + "loss": 1.1973, + "step": 4200 + }, + { + "epoch": 27.69, + "learning_rate": 4.674553667455367e-05, + "loss": 1.2303, + "step": 4210 + }, + { + "epoch": 27.76, + "learning_rate": 4.671327167132717e-05, + "loss": 1.2865, + "step": 4220 + }, + { + "epoch": 27.82, + "learning_rate": 4.6681006668100665e-05, + "loss": 1.2205, + "step": 4230 + }, + { + "epoch": 27.89, + "learning_rate": 4.6648741664874167e-05, + "loss": 1.2056, + "step": 4240 + }, + { + "epoch": 27.96, + "learning_rate": 4.661647666164767e-05, + "loss": 1.1766, + "step": 4250 + }, + { + "epoch": 28.03, + "learning_rate": 4.658421165842117e-05, + "loss": 1.376, + "step": 4260 + }, + { + "epoch": 28.09, + "learning_rate": 4.6551946655194665e-05, + "loss": 1.1793, + "step": 4270 + }, + { + "epoch": 28.16, + "learning_rate": 4.651968165196817e-05, + "loss": 1.2529, + "step": 4280 + }, + { + "epoch": 28.22, + "learning_rate": 4.648741664874166e-05, + "loss": 1.2434, + "step": 4290 + }, + { + "epoch": 28.29, + "learning_rate": 4.6455151645515164e-05, + "loss": 1.2635, + "step": 4300 + }, + { + "epoch": 28.35, + "learning_rate": 4.6422886642288666e-05, + "loss": 1.1739, + "step": 4310 + }, + { + "epoch": 28.42, + "learning_rate": 4.639062163906217e-05, + "loss": 1.2323, + "step": 4320 + }, + { + "epoch": 28.48, + "learning_rate": 4.635835663583567e-05, + "loss": 1.2578, + "step": 4330 + }, + { + "epoch": 28.55, + "learning_rate": 4.6326091632609164e-05, + "loss": 1.2275, + "step": 4340 + }, + { + "epoch": 28.62, + "learning_rate": 4.6293826629382666e-05, + "loss": 1.2, + "step": 4350 + }, + { + "epoch": 28.68, + "learning_rate": 4.626156162615616e-05, + "loss": 1.2251, + "step": 4360 + }, + { + "epoch": 28.75, + "learning_rate": 4.622929662292966e-05, + "loss": 1.2697, + "step": 4370 + }, + { + "epoch": 28.81, + "learning_rate": 4.6197031619703165e-05, + "loss": 1.1808, + "step": 4380 + }, + { + "epoch": 28.88, + "learning_rate": 4.616476661647667e-05, + "loss": 1.1927, + "step": 4390 + }, + { + "epoch": 28.94, + "learning_rate": 4.613250161325016e-05, + "loss": 1.206, + "step": 4400 + }, + { + "epoch": 29.01, + "learning_rate": 4.6100236610023664e-05, + "loss": 1.366, + "step": 4410 + }, + { + "epoch": 29.08, + "learning_rate": 4.606797160679716e-05, + "loss": 1.1786, + "step": 4420 + }, + { + "epoch": 29.14, + "learning_rate": 4.603570660357066e-05, + "loss": 1.2063, + "step": 4430 + }, + { + "epoch": 29.21, + "learning_rate": 4.600344160034416e-05, + "loss": 1.1814, + "step": 4440 + }, + { + "epoch": 29.27, + "learning_rate": 4.5971176597117664e-05, + "loss": 1.2752, + "step": 4450 + }, + { + "epoch": 29.34, + "learning_rate": 4.593891159389116e-05, + "loss": 1.1697, + "step": 4460 + }, + { + "epoch": 29.41, + "learning_rate": 4.590664659066466e-05, + "loss": 1.2066, + "step": 4470 + }, + { + "epoch": 29.47, + "learning_rate": 4.5874381587438156e-05, + "loss": 1.2298, + "step": 4480 + }, + { + "epoch": 29.54, + "learning_rate": 4.584211658421166e-05, + "loss": 1.2676, + "step": 4490 + }, + { + "epoch": 29.6, + "learning_rate": 4.580985158098516e-05, + "loss": 1.1844, + "step": 4500 + }, + { + "epoch": 29.6, + "eval_loss": 0.43086153268814087, + "eval_runtime": 323.4912, + "eval_samples_per_second": 21.979, + "eval_steps_per_second": 1.376, + "eval_wer": 0.46511177779928387, + "step": 4500 + }, + { + "epoch": 29.67, + "learning_rate": 4.577758657775866e-05, + "loss": 1.2059, + "step": 4510 + }, + { + "epoch": 29.73, + "learning_rate": 4.574532157453216e-05, + "loss": 1.2293, + "step": 4520 + }, + { + "epoch": 29.8, + "learning_rate": 4.571305657130566e-05, + "loss": 1.2226, + "step": 4530 + }, + { + "epoch": 29.86, + "learning_rate": 4.568079156807915e-05, + "loss": 1.159, + "step": 4540 + }, + { + "epoch": 29.93, + "learning_rate": 4.5648526564852655e-05, + "loss": 1.212, + "step": 4550 + }, + { + "epoch": 30.0, + "learning_rate": 4.561626156162616e-05, + "loss": 1.2842, + "step": 4560 + }, + { + "epoch": 30.07, + "learning_rate": 4.558399655839966e-05, + "loss": 1.218, + "step": 4570 + }, + { + "epoch": 30.13, + "learning_rate": 4.555173155517316e-05, + "loss": 1.1917, + "step": 4580 + }, + { + "epoch": 30.2, + "learning_rate": 4.5519466551946655e-05, + "loss": 1.1997, + "step": 4590 + }, + { + "epoch": 30.26, + "learning_rate": 4.548720154872016e-05, + "loss": 1.2586, + "step": 4600 + }, + { + "epoch": 30.33, + "learning_rate": 4.545493654549365e-05, + "loss": 1.1813, + "step": 4610 + }, + { + "epoch": 30.39, + "learning_rate": 4.5422671542267154e-05, + "loss": 1.1828, + "step": 4620 + }, + { + "epoch": 30.46, + "learning_rate": 4.5390406539040656e-05, + "loss": 1.1856, + "step": 4630 + }, + { + "epoch": 30.52, + "learning_rate": 4.535814153581416e-05, + "loss": 1.2412, + "step": 4640 + }, + { + "epoch": 30.59, + "learning_rate": 4.532587653258765e-05, + "loss": 1.1547, + "step": 4650 + }, + { + "epoch": 30.65, + "learning_rate": 4.5293611529361155e-05, + "loss": 1.179, + "step": 4660 + }, + { + "epoch": 30.72, + "learning_rate": 4.526134652613465e-05, + "loss": 1.1997, + "step": 4670 + }, + { + "epoch": 30.79, + "learning_rate": 4.522908152290815e-05, + "loss": 1.2361, + "step": 4680 + }, + { + "epoch": 30.85, + "learning_rate": 4.519681651968165e-05, + "loss": 1.1425, + "step": 4690 + }, + { + "epoch": 30.92, + "learning_rate": 4.5164551516455155e-05, + "loss": 1.1794, + "step": 4700 + }, + { + "epoch": 30.98, + "learning_rate": 4.513228651322866e-05, + "loss": 1.2061, + "step": 4710 + }, + { + "epoch": 31.05, + "learning_rate": 4.510002151000215e-05, + "loss": 1.2714, + "step": 4720 + }, + { + "epoch": 31.12, + "learning_rate": 4.506775650677565e-05, + "loss": 1.1805, + "step": 4730 + }, + { + "epoch": 31.18, + "learning_rate": 4.503549150354915e-05, + "loss": 1.1843, + "step": 4740 + }, + { + "epoch": 31.25, + "learning_rate": 4.5006453000645305e-05, + "loss": 1.2111, + "step": 4750 + }, + { + "epoch": 31.31, + "learning_rate": 4.49741879974188e-05, + "loss": 1.1613, + "step": 4760 + }, + { + "epoch": 31.38, + "learning_rate": 4.49419229941923e-05, + "loss": 1.1599, + "step": 4770 + }, + { + "epoch": 31.45, + "learning_rate": 4.49096579909658e-05, + "loss": 1.1621, + "step": 4780 + }, + { + "epoch": 31.51, + "learning_rate": 4.48773929877393e-05, + "loss": 1.2187, + "step": 4790 + }, + { + "epoch": 31.58, + "learning_rate": 4.48451279845128e-05, + "loss": 1.1546, + "step": 4800 + }, + { + "epoch": 31.64, + "learning_rate": 4.48128629812863e-05, + "loss": 1.2005, + "step": 4810 + }, + { + "epoch": 31.71, + "learning_rate": 4.47805979780598e-05, + "loss": 1.1889, + "step": 4820 + }, + { + "epoch": 31.77, + "learning_rate": 4.47483329748333e-05, + "loss": 1.2627, + "step": 4830 + }, + { + "epoch": 31.84, + "learning_rate": 4.4716067971606794e-05, + "loss": 1.1345, + "step": 4840 + }, + { + "epoch": 31.9, + "learning_rate": 4.4683802968380296e-05, + "loss": 1.1812, + "step": 4850 + }, + { + "epoch": 31.97, + "learning_rate": 4.46515379651538e-05, + "loss": 1.2073, + "step": 4860 + }, + { + "epoch": 32.04, + "learning_rate": 4.46192729619273e-05, + "loss": 1.3344, + "step": 4870 + }, + { + "epoch": 32.1, + "learning_rate": 4.45870079587008e-05, + "loss": 1.1497, + "step": 4880 + }, + { + "epoch": 32.17, + "learning_rate": 4.45547429554743e-05, + "loss": 1.1874, + "step": 4890 + }, + { + "epoch": 32.24, + "learning_rate": 4.452247795224779e-05, + "loss": 1.1958, + "step": 4900 + }, + { + "epoch": 32.3, + "learning_rate": 4.4490212949021294e-05, + "loss": 1.2074, + "step": 4910 + }, + { + "epoch": 32.37, + "learning_rate": 4.4457947945794795e-05, + "loss": 1.1549, + "step": 4920 + }, + { + "epoch": 32.43, + "learning_rate": 4.44256829425683e-05, + "loss": 1.1614, + "step": 4930 + }, + { + "epoch": 32.5, + "learning_rate": 4.43934179393418e-05, + "loss": 1.2384, + "step": 4940 + }, + { + "epoch": 32.56, + "learning_rate": 4.4361152936115294e-05, + "loss": 1.1331, + "step": 4950 + }, + { + "epoch": 32.63, + "learning_rate": 4.4328887932888796e-05, + "loss": 1.153, + "step": 4960 + }, + { + "epoch": 32.69, + "learning_rate": 4.429662292966229e-05, + "loss": 1.1782, + "step": 4970 + }, + { + "epoch": 32.76, + "learning_rate": 4.426435792643579e-05, + "loss": 1.2534, + "step": 4980 + }, + { + "epoch": 32.82, + "learning_rate": 4.4232092923209294e-05, + "loss": 1.1444, + "step": 4990 + }, + { + "epoch": 32.89, + "learning_rate": 4.4199827919982796e-05, + "loss": 1.1795, + "step": 5000 + }, + { + "epoch": 32.89, + "eval_loss": 0.41311895847320557, + "eval_runtime": 319.5742, + "eval_samples_per_second": 22.248, + "eval_steps_per_second": 1.392, + "eval_wer": 0.45235652762992357, + "step": 5000 + }, + { + "epoch": 32.96, + "learning_rate": 4.416756291675629e-05, + "loss": 1.1466, + "step": 5010 + }, + { + "epoch": 33.03, + "learning_rate": 4.413529791352979e-05, + "loss": 1.3266, + "step": 5020 + }, + { + "epoch": 33.09, + "learning_rate": 4.410303291030329e-05, + "loss": 1.1433, + "step": 5030 + }, + { + "epoch": 33.16, + "learning_rate": 4.407076790707679e-05, + "loss": 1.1581, + "step": 5040 + }, + { + "epoch": 33.22, + "learning_rate": 4.403850290385029e-05, + "loss": 1.1718, + "step": 5050 + }, + { + "epoch": 33.29, + "learning_rate": 4.4006237900623794e-05, + "loss": 1.1991, + "step": 5060 + }, + { + "epoch": 33.35, + "learning_rate": 4.3973972897397295e-05, + "loss": 1.1464, + "step": 5070 + }, + { + "epoch": 33.42, + "learning_rate": 4.394170789417079e-05, + "loss": 1.1498, + "step": 5080 + }, + { + "epoch": 33.48, + "learning_rate": 4.3909442890944285e-05, + "loss": 1.1941, + "step": 5090 + }, + { + "epoch": 33.55, + "learning_rate": 4.387717788771779e-05, + "loss": 1.2034, + "step": 5100 + }, + { + "epoch": 33.62, + "learning_rate": 4.384491288449129e-05, + "loss": 1.1222, + "step": 5110 + }, + { + "epoch": 33.68, + "learning_rate": 4.381264788126479e-05, + "loss": 1.2138, + "step": 5120 + }, + { + "epoch": 33.75, + "learning_rate": 4.378038287803829e-05, + "loss": 1.2057, + "step": 5130 + }, + { + "epoch": 33.81, + "learning_rate": 4.374811787481179e-05, + "loss": 1.1263, + "step": 5140 + }, + { + "epoch": 33.88, + "learning_rate": 4.371585287158529e-05, + "loss": 1.1765, + "step": 5150 + }, + { + "epoch": 33.94, + "learning_rate": 4.3683587868358785e-05, + "loss": 1.1571, + "step": 5160 + }, + { + "epoch": 34.01, + "learning_rate": 4.3651322865132286e-05, + "loss": 1.3424, + "step": 5170 + }, + { + "epoch": 34.08, + "learning_rate": 4.361905786190579e-05, + "loss": 1.1235, + "step": 5180 + }, + { + "epoch": 34.14, + "learning_rate": 4.358679285867929e-05, + "loss": 1.1396, + "step": 5190 + }, + { + "epoch": 34.21, + "learning_rate": 4.3554527855452785e-05, + "loss": 1.1382, + "step": 5200 + }, + { + "epoch": 34.27, + "learning_rate": 4.352226285222629e-05, + "loss": 1.2624, + "step": 5210 + }, + { + "epoch": 34.34, + "learning_rate": 4.348999784899978e-05, + "loss": 1.1128, + "step": 5220 + }, + { + "epoch": 34.41, + "learning_rate": 4.3457732845773284e-05, + "loss": 1.153, + "step": 5230 + }, + { + "epoch": 34.47, + "learning_rate": 4.3425467842546786e-05, + "loss": 1.157, + "step": 5240 + }, + { + "epoch": 34.54, + "learning_rate": 4.339320283932029e-05, + "loss": 1.2042, + "step": 5250 + }, + { + "epoch": 34.6, + "learning_rate": 4.336093783609378e-05, + "loss": 1.125, + "step": 5260 + }, + { + "epoch": 34.67, + "learning_rate": 4.3328672832867284e-05, + "loss": 1.1504, + "step": 5270 + }, + { + "epoch": 34.73, + "learning_rate": 4.3296407829640786e-05, + "loss": 1.2242, + "step": 5280 + }, + { + "epoch": 34.8, + "learning_rate": 4.326414282641428e-05, + "loss": 1.1771, + "step": 5290 + }, + { + "epoch": 34.86, + "learning_rate": 4.323187782318778e-05, + "loss": 1.1526, + "step": 5300 + }, + { + "epoch": 34.93, + "learning_rate": 4.3199612819961285e-05, + "loss": 1.1573, + "step": 5310 + }, + { + "epoch": 35.0, + "learning_rate": 4.3167347816734787e-05, + "loss": 1.1895, + "step": 5320 + }, + { + "epoch": 35.07, + "learning_rate": 4.313508281350828e-05, + "loss": 1.2058, + "step": 5330 + }, + { + "epoch": 35.13, + "learning_rate": 4.310281781028178e-05, + "loss": 1.1529, + "step": 5340 + }, + { + "epoch": 35.2, + "learning_rate": 4.307055280705528e-05, + "loss": 1.152, + "step": 5350 + }, + { + "epoch": 35.26, + "learning_rate": 4.303828780382878e-05, + "loss": 1.1925, + "step": 5360 + }, + { + "epoch": 35.33, + "learning_rate": 4.300602280060228e-05, + "loss": 1.118, + "step": 5370 + }, + { + "epoch": 35.39, + "learning_rate": 4.2973757797375784e-05, + "loss": 1.1376, + "step": 5380 + }, + { + "epoch": 35.46, + "learning_rate": 4.294149279414928e-05, + "loss": 1.1493, + "step": 5390 + }, + { + "epoch": 35.52, + "learning_rate": 4.290922779092278e-05, + "loss": 1.1937, + "step": 5400 + }, + { + "epoch": 35.59, + "learning_rate": 4.2876962787696276e-05, + "loss": 1.1275, + "step": 5410 + }, + { + "epoch": 35.65, + "learning_rate": 4.284469778446978e-05, + "loss": 1.133, + "step": 5420 + }, + { + "epoch": 35.72, + "learning_rate": 4.281243278124328e-05, + "loss": 1.1574, + "step": 5430 + }, + { + "epoch": 35.79, + "learning_rate": 4.278016777801678e-05, + "loss": 1.212, + "step": 5440 + }, + { + "epoch": 35.85, + "learning_rate": 4.2747902774790276e-05, + "loss": 1.0953, + "step": 5450 + }, + { + "epoch": 35.92, + "learning_rate": 4.271563777156378e-05, + "loss": 1.1724, + "step": 5460 + }, + { + "epoch": 35.98, + "learning_rate": 4.268337276833728e-05, + "loss": 1.2039, + "step": 5470 + }, + { + "epoch": 36.05, + "learning_rate": 4.2651107765110775e-05, + "loss": 1.273, + "step": 5480 + }, + { + "epoch": 36.12, + "learning_rate": 4.261884276188428e-05, + "loss": 1.1196, + "step": 5490 + }, + { + "epoch": 36.18, + "learning_rate": 4.258657775865778e-05, + "loss": 1.1471, + "step": 5500 + }, + { + "epoch": 36.18, + "eval_loss": 0.4052400290966034, + "eval_runtime": 317.7832, + "eval_samples_per_second": 22.374, + "eval_steps_per_second": 1.4, + "eval_wer": 0.44349172553953353, + "step": 5500 + }, + { + "epoch": 36.25, + "learning_rate": 4.255431275543128e-05, + "loss": 1.1871, + "step": 5510 + }, + { + "epoch": 36.31, + "learning_rate": 4.2522047752204775e-05, + "loss": 1.1086, + "step": 5520 + }, + { + "epoch": 36.38, + "learning_rate": 4.248978274897828e-05, + "loss": 1.1285, + "step": 5530 + }, + { + "epoch": 36.45, + "learning_rate": 4.245751774575177e-05, + "loss": 1.1219, + "step": 5540 + }, + { + "epoch": 36.51, + "learning_rate": 4.2425252742525274e-05, + "loss": 1.2096, + "step": 5550 + }, + { + "epoch": 36.58, + "learning_rate": 4.2392987739298776e-05, + "loss": 1.0957, + "step": 5560 + }, + { + "epoch": 36.64, + "learning_rate": 4.236072273607228e-05, + "loss": 1.1271, + "step": 5570 + }, + { + "epoch": 36.71, + "learning_rate": 4.232845773284577e-05, + "loss": 1.1579, + "step": 5580 + }, + { + "epoch": 36.77, + "learning_rate": 4.2296192729619274e-05, + "loss": 1.2091, + "step": 5590 + }, + { + "epoch": 36.84, + "learning_rate": 4.2263927726392776e-05, + "loss": 1.1073, + "step": 5600 + }, + { + "epoch": 36.9, + "learning_rate": 4.223166272316627e-05, + "loss": 1.1487, + "step": 5610 + }, + { + "epoch": 36.97, + "learning_rate": 4.219939771993977e-05, + "loss": 1.1298, + "step": 5620 + }, + { + "epoch": 37.04, + "learning_rate": 4.2167132716713275e-05, + "loss": 1.2871, + "step": 5630 + }, + { + "epoch": 37.1, + "learning_rate": 4.213486771348677e-05, + "loss": 1.0937, + "step": 5640 + }, + { + "epoch": 37.17, + "learning_rate": 4.210260271026027e-05, + "loss": 1.1284, + "step": 5650 + }, + { + "epoch": 37.24, + "learning_rate": 4.2070337707033774e-05, + "loss": 1.1632, + "step": 5660 + }, + { + "epoch": 37.3, + "learning_rate": 4.203807270380727e-05, + "loss": 1.1688, + "step": 5670 + }, + { + "epoch": 37.37, + "learning_rate": 4.200580770058077e-05, + "loss": 1.1007, + "step": 5680 + }, + { + "epoch": 37.43, + "learning_rate": 4.197354269735427e-05, + "loss": 1.1261, + "step": 5690 + }, + { + "epoch": 37.5, + "learning_rate": 4.1941277694127774e-05, + "loss": 1.1808, + "step": 5700 + }, + { + "epoch": 37.56, + "learning_rate": 4.190901269090127e-05, + "loss": 1.0999, + "step": 5710 + }, + { + "epoch": 37.63, + "learning_rate": 4.187674768767477e-05, + "loss": 1.1271, + "step": 5720 + }, + { + "epoch": 37.69, + "learning_rate": 4.1844482684448266e-05, + "loss": 1.1519, + "step": 5730 + }, + { + "epoch": 37.76, + "learning_rate": 4.181221768122177e-05, + "loss": 1.1955, + "step": 5740 + }, + { + "epoch": 37.82, + "learning_rate": 4.177995267799527e-05, + "loss": 1.1145, + "step": 5750 + }, + { + "epoch": 37.89, + "learning_rate": 4.174768767476877e-05, + "loss": 1.1182, + "step": 5760 + }, + { + "epoch": 37.96, + "learning_rate": 4.1715422671542266e-05, + "loss": 1.1376, + "step": 5770 + }, + { + "epoch": 38.03, + "learning_rate": 4.168315766831577e-05, + "loss": 1.3017, + "step": 5780 + }, + { + "epoch": 38.09, + "learning_rate": 4.165089266508927e-05, + "loss": 1.0855, + "step": 5790 + }, + { + "epoch": 38.16, + "learning_rate": 4.1618627661862765e-05, + "loss": 1.1092, + "step": 5800 + }, + { + "epoch": 38.22, + "learning_rate": 4.158636265863627e-05, + "loss": 1.1129, + "step": 5810 + }, + { + "epoch": 38.29, + "learning_rate": 4.155409765540977e-05, + "loss": 1.1781, + "step": 5820 + }, + { + "epoch": 38.35, + "learning_rate": 4.1521832652183264e-05, + "loss": 1.1033, + "step": 5830 + }, + { + "epoch": 38.42, + "learning_rate": 4.1489567648956766e-05, + "loss": 1.1188, + "step": 5840 + }, + { + "epoch": 38.48, + "learning_rate": 4.145730264573027e-05, + "loss": 1.1868, + "step": 5850 + }, + { + "epoch": 38.55, + "learning_rate": 4.142503764250376e-05, + "loss": 1.1578, + "step": 5860 + }, + { + "epoch": 38.62, + "learning_rate": 4.1392772639277264e-05, + "loss": 1.112, + "step": 5870 + }, + { + "epoch": 38.68, + "learning_rate": 4.1360507636050766e-05, + "loss": 1.1149, + "step": 5880 + }, + { + "epoch": 38.75, + "learning_rate": 4.132824263282427e-05, + "loss": 1.2071, + "step": 5890 + }, + { + "epoch": 38.81, + "learning_rate": 4.129597762959776e-05, + "loss": 1.0991, + "step": 5900 + }, + { + "epoch": 38.88, + "learning_rate": 4.1263712626371265e-05, + "loss": 1.1303, + "step": 5910 + }, + { + "epoch": 38.94, + "learning_rate": 4.123144762314476e-05, + "loss": 1.1012, + "step": 5920 + }, + { + "epoch": 39.01, + "learning_rate": 4.119918261991826e-05, + "loss": 1.2124, + "step": 5930 + }, + { + "epoch": 39.08, + "learning_rate": 4.116691761669176e-05, + "loss": 1.089, + "step": 5940 + }, + { + "epoch": 39.14, + "learning_rate": 4.1134652613465265e-05, + "loss": 1.0932, + "step": 5950 + }, + { + "epoch": 39.21, + "learning_rate": 4.110238761023876e-05, + "loss": 1.1506, + "step": 5960 + }, + { + "epoch": 39.27, + "learning_rate": 4.107012260701226e-05, + "loss": 1.205, + "step": 5970 + }, + { + "epoch": 39.34, + "learning_rate": 4.1037857603785764e-05, + "loss": 1.0693, + "step": 5980 + }, + { + "epoch": 39.41, + "learning_rate": 4.100559260055926e-05, + "loss": 1.1025, + "step": 5990 + }, + { + "epoch": 39.47, + "learning_rate": 4.097332759733276e-05, + "loss": 1.1337, + "step": 6000 + }, + { + "epoch": 39.47, + "eval_loss": 0.39270928502082825, + "eval_runtime": 319.378, + "eval_samples_per_second": 22.262, + "eval_steps_per_second": 1.393, + "eval_wer": 0.4363495596632149, + "step": 6000 + }, + { + "epoch": 39.54, + "learning_rate": 4.094106259410626e-05, + "loss": 1.1722, + "step": 6010 + }, + { + "epoch": 39.6, + "learning_rate": 4.090879759087976e-05, + "loss": 1.0861, + "step": 6020 + }, + { + "epoch": 39.67, + "learning_rate": 4.087653258765326e-05, + "loss": 1.1247, + "step": 6030 + }, + { + "epoch": 39.73, + "learning_rate": 4.084426758442676e-05, + "loss": 1.1784, + "step": 6040 + }, + { + "epoch": 39.8, + "learning_rate": 4.0812002581200256e-05, + "loss": 1.1134, + "step": 6050 + }, + { + "epoch": 39.86, + "learning_rate": 4.077973757797376e-05, + "loss": 1.1039, + "step": 6060 + }, + { + "epoch": 39.93, + "learning_rate": 4.074747257474726e-05, + "loss": 1.1316, + "step": 6070 + }, + { + "epoch": 40.0, + "learning_rate": 4.071520757152076e-05, + "loss": 1.17, + "step": 6080 + }, + { + "epoch": 40.07, + "learning_rate": 4.068294256829426e-05, + "loss": 1.1694, + "step": 6090 + }, + { + "epoch": 40.13, + "learning_rate": 4.065067756506776e-05, + "loss": 1.0948, + "step": 6100 + }, + { + "epoch": 40.2, + "learning_rate": 4.061841256184126e-05, + "loss": 1.1219, + "step": 6110 + }, + { + "epoch": 40.26, + "learning_rate": 4.0586147558614755e-05, + "loss": 1.1615, + "step": 6120 + }, + { + "epoch": 40.33, + "learning_rate": 4.055388255538826e-05, + "loss": 1.0687, + "step": 6130 + }, + { + "epoch": 40.39, + "learning_rate": 4.052161755216176e-05, + "loss": 1.1161, + "step": 6140 + }, + { + "epoch": 40.46, + "learning_rate": 4.0489352548935254e-05, + "loss": 1.1113, + "step": 6150 + }, + { + "epoch": 40.52, + "learning_rate": 4.0457087545708756e-05, + "loss": 1.1724, + "step": 6160 + }, + { + "epoch": 40.59, + "learning_rate": 4.042482254248226e-05, + "loss": 1.0693, + "step": 6170 + }, + { + "epoch": 40.65, + "learning_rate": 4.039255753925575e-05, + "loss": 1.1136, + "step": 6180 + }, + { + "epoch": 40.72, + "learning_rate": 4.0360292536029254e-05, + "loss": 1.13, + "step": 6190 + }, + { + "epoch": 40.79, + "learning_rate": 4.0328027532802756e-05, + "loss": 1.1549, + "step": 6200 + }, + { + "epoch": 40.85, + "learning_rate": 4.029576252957625e-05, + "loss": 1.058, + "step": 6210 + }, + { + "epoch": 40.92, + "learning_rate": 4.026349752634975e-05, + "loss": 1.1233, + "step": 6220 + }, + { + "epoch": 40.98, + "learning_rate": 4.0231232523123255e-05, + "loss": 1.157, + "step": 6230 + }, + { + "epoch": 41.05, + "learning_rate": 4.019896751989675e-05, + "loss": 1.1953, + "step": 6240 + }, + { + "epoch": 41.12, + "learning_rate": 4.016670251667025e-05, + "loss": 1.0919, + "step": 6250 + }, + { + "epoch": 41.18, + "learning_rate": 4.0134437513443754e-05, + "loss": 1.0925, + "step": 6260 + }, + { + "epoch": 41.25, + "learning_rate": 4.0102172510217255e-05, + "loss": 1.1612, + "step": 6270 + }, + { + "epoch": 41.31, + "learning_rate": 4.006990750699075e-05, + "loss": 1.0826, + "step": 6280 + }, + { + "epoch": 41.38, + "learning_rate": 4.003764250376425e-05, + "loss": 1.0994, + "step": 6290 + }, + { + "epoch": 41.45, + "learning_rate": 4.0005377500537754e-05, + "loss": 1.1424, + "step": 6300 + }, + { + "epoch": 41.51, + "learning_rate": 3.997311249731125e-05, + "loss": 1.1677, + "step": 6310 + }, + { + "epoch": 41.58, + "learning_rate": 3.994084749408475e-05, + "loss": 1.0776, + "step": 6320 + }, + { + "epoch": 41.64, + "learning_rate": 3.990858249085825e-05, + "loss": 1.0904, + "step": 6330 + }, + { + "epoch": 41.71, + "learning_rate": 3.987631748763175e-05, + "loss": 1.1126, + "step": 6340 + }, + { + "epoch": 41.77, + "learning_rate": 3.984405248440525e-05, + "loss": 1.1589, + "step": 6350 + }, + { + "epoch": 41.84, + "learning_rate": 3.981178748117875e-05, + "loss": 1.0549, + "step": 6360 + }, + { + "epoch": 41.9, + "learning_rate": 3.9779522477952246e-05, + "loss": 1.1058, + "step": 6370 + }, + { + "epoch": 41.97, + "learning_rate": 3.974725747472575e-05, + "loss": 1.1154, + "step": 6380 + }, + { + "epoch": 42.04, + "learning_rate": 3.971499247149925e-05, + "loss": 1.2687, + "step": 6390 + }, + { + "epoch": 42.1, + "learning_rate": 3.9682727468272745e-05, + "loss": 1.0773, + "step": 6400 + }, + { + "epoch": 42.17, + "learning_rate": 3.965046246504625e-05, + "loss": 1.0969, + "step": 6410 + }, + { + "epoch": 42.24, + "learning_rate": 3.961819746181975e-05, + "loss": 1.1189, + "step": 6420 + }, + { + "epoch": 42.3, + "learning_rate": 3.9585932458593244e-05, + "loss": 1.14, + "step": 6430 + }, + { + "epoch": 42.37, + "learning_rate": 3.9553667455366746e-05, + "loss": 1.0748, + "step": 6440 + }, + { + "epoch": 42.43, + "learning_rate": 3.952140245214025e-05, + "loss": 1.107, + "step": 6450 + }, + { + "epoch": 42.5, + "learning_rate": 3.948913744891374e-05, + "loss": 1.1683, + "step": 6460 + }, + { + "epoch": 42.56, + "learning_rate": 3.9456872445687244e-05, + "loss": 1.0603, + "step": 6470 + }, + { + "epoch": 42.63, + "learning_rate": 3.9424607442460746e-05, + "loss": 1.081, + "step": 6480 + }, + { + "epoch": 42.69, + "learning_rate": 3.939234243923425e-05, + "loss": 1.1066, + "step": 6490 + }, + { + "epoch": 42.76, + "learning_rate": 3.936007743600774e-05, + "loss": 1.1896, + "step": 6500 + }, + { + "epoch": 42.76, + "eval_loss": 0.38105571269989014, + "eval_runtime": 323.0014, + "eval_samples_per_second": 22.012, + "eval_steps_per_second": 1.378, + "eval_wer": 0.42543307848640277, + "step": 6500 + }, + { + "epoch": 42.82, + "learning_rate": 3.9327812432781245e-05, + "loss": 1.062, + "step": 6510 + }, + { + "epoch": 42.89, + "learning_rate": 3.9295547429554747e-05, + "loss": 1.1064, + "step": 6520 + }, + { + "epoch": 42.96, + "learning_rate": 3.926328242632824e-05, + "loss": 1.0806, + "step": 6530 + }, + { + "epoch": 43.03, + "learning_rate": 3.9231017423101743e-05, + "loss": 1.25, + "step": 6540 + }, + { + "epoch": 43.09, + "learning_rate": 3.9198752419875245e-05, + "loss": 1.034, + "step": 6550 + }, + { + "epoch": 43.16, + "learning_rate": 3.916648741664874e-05, + "loss": 1.0961, + "step": 6560 + }, + { + "epoch": 43.22, + "learning_rate": 3.913422241342224e-05, + "loss": 1.1108, + "step": 6570 + }, + { + "epoch": 43.29, + "learning_rate": 3.9101957410195744e-05, + "loss": 1.1584, + "step": 6580 + }, + { + "epoch": 43.35, + "learning_rate": 3.906969240696924e-05, + "loss": 1.1061, + "step": 6590 + }, + { + "epoch": 43.42, + "learning_rate": 3.903742740374274e-05, + "loss": 1.0788, + "step": 6600 + }, + { + "epoch": 43.48, + "learning_rate": 3.900516240051624e-05, + "loss": 1.1452, + "step": 6610 + }, + { + "epoch": 43.55, + "learning_rate": 3.8972897397289744e-05, + "loss": 1.1088, + "step": 6620 + }, + { + "epoch": 43.62, + "learning_rate": 3.894063239406324e-05, + "loss": 1.0395, + "step": 6630 + }, + { + "epoch": 43.68, + "learning_rate": 3.890836739083674e-05, + "loss": 1.104, + "step": 6640 + }, + { + "epoch": 43.75, + "learning_rate": 3.8876102387610236e-05, + "loss": 1.1578, + "step": 6650 + }, + { + "epoch": 43.81, + "learning_rate": 3.884383738438374e-05, + "loss": 1.0595, + "step": 6660 + }, + { + "epoch": 43.88, + "learning_rate": 3.881157238115724e-05, + "loss": 1.0959, + "step": 6670 + }, + { + "epoch": 43.94, + "learning_rate": 3.877930737793074e-05, + "loss": 1.0656, + "step": 6680 + }, + { + "epoch": 44.01, + "learning_rate": 3.874704237470424e-05, + "loss": 1.2342, + "step": 6690 + }, + { + "epoch": 44.08, + "learning_rate": 3.871477737147774e-05, + "loss": 1.0533, + "step": 6700 + }, + { + "epoch": 44.14, + "learning_rate": 3.868251236825124e-05, + "loss": 1.0577, + "step": 6710 + }, + { + "epoch": 44.21, + "learning_rate": 3.8650247365024735e-05, + "loss": 1.0793, + "step": 6720 + }, + { + "epoch": 44.27, + "learning_rate": 3.861798236179824e-05, + "loss": 1.1463, + "step": 6730 + }, + { + "epoch": 44.34, + "learning_rate": 3.858571735857174e-05, + "loss": 1.0532, + "step": 6740 + }, + { + "epoch": 44.41, + "learning_rate": 3.8553452355345234e-05, + "loss": 1.0896, + "step": 6750 + }, + { + "epoch": 44.47, + "learning_rate": 3.8521187352118736e-05, + "loss": 1.0941, + "step": 6760 + }, + { + "epoch": 44.54, + "learning_rate": 3.848892234889224e-05, + "loss": 1.1166, + "step": 6770 + }, + { + "epoch": 44.6, + "learning_rate": 3.845665734566573e-05, + "loss": 1.0476, + "step": 6780 + }, + { + "epoch": 44.67, + "learning_rate": 3.8424392342439234e-05, + "loss": 1.0674, + "step": 6790 + }, + { + "epoch": 44.73, + "learning_rate": 3.8392127339212736e-05, + "loss": 1.107, + "step": 6800 + }, + { + "epoch": 44.8, + "learning_rate": 3.835986233598624e-05, + "loss": 1.0972, + "step": 6810 + }, + { + "epoch": 44.86, + "learning_rate": 3.832759733275973e-05, + "loss": 1.0456, + "step": 6820 + }, + { + "epoch": 44.93, + "learning_rate": 3.8295332329533235e-05, + "loss": 1.0916, + "step": 6830 + }, + { + "epoch": 45.0, + "learning_rate": 3.8266293826629385e-05, + "loss": 1.1377, + "step": 6840 + }, + { + "epoch": 45.07, + "learning_rate": 3.8234028823402886e-05, + "loss": 1.1153, + "step": 6850 + }, + { + "epoch": 45.13, + "learning_rate": 3.820176382017639e-05, + "loss": 1.0292, + "step": 6860 + }, + { + "epoch": 45.2, + "learning_rate": 3.816949881694988e-05, + "loss": 1.0737, + "step": 6870 + }, + { + "epoch": 45.26, + "learning_rate": 3.813723381372338e-05, + "loss": 1.1493, + "step": 6880 + }, + { + "epoch": 45.33, + "learning_rate": 3.810496881049688e-05, + "loss": 1.0346, + "step": 6890 + }, + { + "epoch": 45.39, + "learning_rate": 3.807270380727038e-05, + "loss": 1.0626, + "step": 6900 + }, + { + "epoch": 45.46, + "learning_rate": 3.8040438804043884e-05, + "loss": 1.1238, + "step": 6910 + }, + { + "epoch": 45.52, + "learning_rate": 3.8008173800817386e-05, + "loss": 1.1701, + "step": 6920 + }, + { + "epoch": 45.59, + "learning_rate": 3.797590879759088e-05, + "loss": 1.0208, + "step": 6930 + }, + { + "epoch": 45.65, + "learning_rate": 3.7943643794364376e-05, + "loss": 1.0844, + "step": 6940 + }, + { + "epoch": 45.72, + "learning_rate": 3.791137879113788e-05, + "loss": 1.0585, + "step": 6950 + }, + { + "epoch": 45.79, + "learning_rate": 3.787911378791138e-05, + "loss": 1.1088, + "step": 6960 + }, + { + "epoch": 45.85, + "learning_rate": 3.784684878468488e-05, + "loss": 1.0437, + "step": 6970 + }, + { + "epoch": 45.92, + "learning_rate": 3.781458378145838e-05, + "loss": 1.0495, + "step": 6980 + }, + { + "epoch": 45.98, + "learning_rate": 3.778554527855453e-05, + "loss": 1.1288, + "step": 6990 + }, + { + "epoch": 46.05, + "learning_rate": 3.775328027532803e-05, + "loss": 1.1847, + "step": 7000 + }, + { + "epoch": 46.05, + "eval_loss": 0.3855203092098236, + "eval_runtime": 320.0475, + "eval_samples_per_second": 22.215, + "eval_steps_per_second": 1.39, + "eval_wer": 0.41294880480015483, + "step": 7000 + }, + { + "epoch": 46.12, + "learning_rate": 3.772101527210153e-05, + "loss": 1.0635, + "step": 7010 + }, + { + "epoch": 46.18, + "learning_rate": 3.7688750268875024e-05, + "loss": 1.1075, + "step": 7020 + }, + { + "epoch": 46.25, + "learning_rate": 3.7656485265648526e-05, + "loss": 1.143, + "step": 7030 + }, + { + "epoch": 46.31, + "learning_rate": 3.762422026242203e-05, + "loss": 1.0112, + "step": 7040 + }, + { + "epoch": 46.38, + "learning_rate": 3.759195525919553e-05, + "loss": 1.05, + "step": 7050 + }, + { + "epoch": 46.45, + "learning_rate": 3.7559690255969025e-05, + "loss": 1.0578, + "step": 7060 + }, + { + "epoch": 46.51, + "learning_rate": 3.7527425252742527e-05, + "loss": 1.1365, + "step": 7070 + }, + { + "epoch": 46.58, + "learning_rate": 3.749516024951603e-05, + "loss": 1.0467, + "step": 7080 + }, + { + "epoch": 46.64, + "learning_rate": 3.7462895246289523e-05, + "loss": 1.0554, + "step": 7090 + }, + { + "epoch": 46.71, + "learning_rate": 3.7430630243063025e-05, + "loss": 1.0705, + "step": 7100 + }, + { + "epoch": 46.77, + "learning_rate": 3.739836523983653e-05, + "loss": 1.1119, + "step": 7110 + }, + { + "epoch": 46.84, + "learning_rate": 3.736610023661002e-05, + "loss": 1.0608, + "step": 7120 + }, + { + "epoch": 46.9, + "learning_rate": 3.7333835233383524e-05, + "loss": 1.0459, + "step": 7130 + }, + { + "epoch": 46.97, + "learning_rate": 3.7301570230157026e-05, + "loss": 1.06, + "step": 7140 + }, + { + "epoch": 47.04, + "learning_rate": 3.726930522693052e-05, + "loss": 1.2005, + "step": 7150 + }, + { + "epoch": 47.1, + "learning_rate": 3.723704022370402e-05, + "loss": 1.0354, + "step": 7160 + }, + { + "epoch": 47.17, + "learning_rate": 3.7204775220477524e-05, + "loss": 1.0732, + "step": 7170 + }, + { + "epoch": 47.24, + "learning_rate": 3.7172510217251026e-05, + "loss": 1.1015, + "step": 7180 + }, + { + "epoch": 47.3, + "learning_rate": 3.714024521402452e-05, + "loss": 1.0637, + "step": 7190 + }, + { + "epoch": 47.37, + "learning_rate": 3.710798021079802e-05, + "loss": 1.0185, + "step": 7200 + }, + { + "epoch": 47.43, + "learning_rate": 3.707571520757152e-05, + "loss": 1.0629, + "step": 7210 + }, + { + "epoch": 47.5, + "learning_rate": 3.704345020434502e-05, + "loss": 1.1096, + "step": 7220 + }, + { + "epoch": 47.56, + "learning_rate": 3.701118520111852e-05, + "loss": 1.0534, + "step": 7230 + }, + { + "epoch": 47.63, + "learning_rate": 3.6978920197892024e-05, + "loss": 1.0467, + "step": 7240 + }, + { + "epoch": 47.69, + "learning_rate": 3.694665519466552e-05, + "loss": 1.0602, + "step": 7250 + }, + { + "epoch": 47.76, + "learning_rate": 3.691439019143902e-05, + "loss": 1.1408, + "step": 7260 + }, + { + "epoch": 47.82, + "learning_rate": 3.6882125188212515e-05, + "loss": 1.0303, + "step": 7270 + }, + { + "epoch": 47.89, + "learning_rate": 3.684986018498602e-05, + "loss": 1.0803, + "step": 7280 + }, + { + "epoch": 47.96, + "learning_rate": 3.681759518175952e-05, + "loss": 1.0774, + "step": 7290 + }, + { + "epoch": 48.03, + "learning_rate": 3.678533017853302e-05, + "loss": 1.2145, + "step": 7300 + }, + { + "epoch": 48.09, + "learning_rate": 3.675306517530652e-05, + "loss": 1.0268, + "step": 7310 + }, + { + "epoch": 48.16, + "learning_rate": 3.672080017208002e-05, + "loss": 1.0546, + "step": 7320 + }, + { + "epoch": 48.22, + "learning_rate": 3.668853516885352e-05, + "loss": 1.0875, + "step": 7330 + }, + { + "epoch": 48.29, + "learning_rate": 3.6656270165627015e-05, + "loss": 1.1135, + "step": 7340 + }, + { + "epoch": 48.35, + "learning_rate": 3.6624005162400516e-05, + "loss": 1.05, + "step": 7350 + }, + { + "epoch": 48.42, + "learning_rate": 3.659174015917402e-05, + "loss": 1.0716, + "step": 7360 + }, + { + "epoch": 48.48, + "learning_rate": 3.655947515594752e-05, + "loss": 1.0956, + "step": 7370 + }, + { + "epoch": 48.55, + "learning_rate": 3.6527210152721015e-05, + "loss": 1.0675, + "step": 7380 + }, + { + "epoch": 48.62, + "learning_rate": 3.649494514949452e-05, + "loss": 1.0263, + "step": 7390 + }, + { + "epoch": 48.68, + "learning_rate": 3.646268014626801e-05, + "loss": 1.0499, + "step": 7400 + }, + { + "epoch": 48.75, + "learning_rate": 3.6430415143041514e-05, + "loss": 1.1132, + "step": 7410 + }, + { + "epoch": 48.81, + "learning_rate": 3.6398150139815016e-05, + "loss": 1.0487, + "step": 7420 + }, + { + "epoch": 48.88, + "learning_rate": 3.636588513658852e-05, + "loss": 1.0571, + "step": 7430 + }, + { + "epoch": 48.94, + "learning_rate": 3.633362013336201e-05, + "loss": 1.0477, + "step": 7440 + }, + { + "epoch": 49.01, + "learning_rate": 3.6301355130135514e-05, + "loss": 1.2352, + "step": 7450 + }, + { + "epoch": 49.08, + "learning_rate": 3.626909012690901e-05, + "loss": 1.0383, + "step": 7460 + }, + { + "epoch": 49.14, + "learning_rate": 3.623682512368251e-05, + "loss": 1.03, + "step": 7470 + }, + { + "epoch": 49.21, + "learning_rate": 3.620456012045601e-05, + "loss": 1.0397, + "step": 7480 + }, + { + "epoch": 49.27, + "learning_rate": 3.6172295117229515e-05, + "loss": 1.1075, + "step": 7490 + }, + { + "epoch": 49.34, + "learning_rate": 3.6140030114003016e-05, + "loss": 0.9954, + "step": 7500 + }, + { + "epoch": 49.34, + "eval_loss": 0.37294888496398926, + "eval_runtime": 321.681, + "eval_samples_per_second": 22.103, + "eval_steps_per_second": 1.383, + "eval_wer": 0.3981418755443724, + "step": 7500 + }, + { + "epoch": 49.41, + "learning_rate": 3.610776511077651e-05, + "loss": 1.0484, + "step": 7510 + }, + { + "epoch": 49.47, + "learning_rate": 3.607550010755001e-05, + "loss": 1.0598, + "step": 7520 + }, + { + "epoch": 49.54, + "learning_rate": 3.604323510432351e-05, + "loss": 1.1166, + "step": 7530 + }, + { + "epoch": 49.6, + "learning_rate": 3.601097010109701e-05, + "loss": 1.0265, + "step": 7540 + }, + { + "epoch": 49.67, + "learning_rate": 3.597870509787051e-05, + "loss": 1.0442, + "step": 7550 + }, + { + "epoch": 49.73, + "learning_rate": 3.5946440094644014e-05, + "loss": 1.0912, + "step": 7560 + }, + { + "epoch": 49.8, + "learning_rate": 3.591417509141751e-05, + "loss": 1.0775, + "step": 7570 + }, + { + "epoch": 49.86, + "learning_rate": 3.588191008819101e-05, + "loss": 1.0295, + "step": 7580 + }, + { + "epoch": 49.93, + "learning_rate": 3.5849645084964506e-05, + "loss": 1.0578, + "step": 7590 + }, + { + "epoch": 50.0, + "learning_rate": 3.581738008173801e-05, + "loss": 1.1105, + "step": 7600 + }, + { + "epoch": 50.07, + "learning_rate": 3.578511507851151e-05, + "loss": 1.1128, + "step": 7610 + }, + { + "epoch": 50.13, + "learning_rate": 3.575285007528501e-05, + "loss": 1.0356, + "step": 7620 + }, + { + "epoch": 50.2, + "learning_rate": 3.572058507205851e-05, + "loss": 1.0443, + "step": 7630 + }, + { + "epoch": 50.26, + "learning_rate": 3.568832006883201e-05, + "loss": 1.1233, + "step": 7640 + }, + { + "epoch": 50.33, + "learning_rate": 3.56560550656055e-05, + "loss": 1.0156, + "step": 7650 + }, + { + "epoch": 50.39, + "learning_rate": 3.5623790062379005e-05, + "loss": 1.0593, + "step": 7660 + }, + { + "epoch": 50.46, + "learning_rate": 3.559152505915251e-05, + "loss": 1.0744, + "step": 7670 + }, + { + "epoch": 50.52, + "learning_rate": 3.555926005592601e-05, + "loss": 1.1256, + "step": 7680 + }, + { + "epoch": 50.59, + "learning_rate": 3.552699505269951e-05, + "loss": 1.0113, + "step": 7690 + }, + { + "epoch": 50.65, + "learning_rate": 3.5494730049473005e-05, + "loss": 1.0382, + "step": 7700 + }, + { + "epoch": 50.72, + "learning_rate": 3.546246504624651e-05, + "loss": 1.0535, + "step": 7710 + }, + { + "epoch": 50.79, + "learning_rate": 3.543020004302e-05, + "loss": 1.0927, + "step": 7720 + }, + { + "epoch": 50.85, + "learning_rate": 3.5397935039793504e-05, + "loss": 1.04, + "step": 7730 + }, + { + "epoch": 50.92, + "learning_rate": 3.5365670036567006e-05, + "loss": 1.0484, + "step": 7740 + }, + { + "epoch": 50.98, + "learning_rate": 3.533340503334051e-05, + "loss": 1.0908, + "step": 7750 + }, + { + "epoch": 51.05, + "learning_rate": 3.5301140030114e-05, + "loss": 1.1377, + "step": 7760 + }, + { + "epoch": 51.12, + "learning_rate": 3.5268875026887504e-05, + "loss": 1.0487, + "step": 7770 + }, + { + "epoch": 51.18, + "learning_rate": 3.5236610023661e-05, + "loss": 1.0859, + "step": 7780 + }, + { + "epoch": 51.25, + "learning_rate": 3.52043450204345e-05, + "loss": 1.1054, + "step": 7790 + }, + { + "epoch": 51.31, + "learning_rate": 3.5172080017208e-05, + "loss": 0.9997, + "step": 7800 + }, + { + "epoch": 51.38, + "learning_rate": 3.5139815013981505e-05, + "loss": 1.0613, + "step": 7810 + }, + { + "epoch": 51.45, + "learning_rate": 3.510755001075501e-05, + "loss": 1.0611, + "step": 7820 + }, + { + "epoch": 51.51, + "learning_rate": 3.50752850075285e-05, + "loss": 1.1271, + "step": 7830 + }, + { + "epoch": 51.58, + "learning_rate": 3.5043020004302e-05, + "loss": 1.027, + "step": 7840 + }, + { + "epoch": 51.64, + "learning_rate": 3.50107550010755e-05, + "loss": 1.0344, + "step": 7850 + }, + { + "epoch": 51.71, + "learning_rate": 3.4978489997849e-05, + "loss": 1.052, + "step": 7860 + }, + { + "epoch": 51.77, + "learning_rate": 3.49462249946225e-05, + "loss": 1.1153, + "step": 7870 + }, + { + "epoch": 51.84, + "learning_rate": 3.4913959991396004e-05, + "loss": 1.0125, + "step": 7880 + }, + { + "epoch": 51.9, + "learning_rate": 3.48816949881695e-05, + "loss": 1.0259, + "step": 7890 + }, + { + "epoch": 51.97, + "learning_rate": 3.4849429984943e-05, + "loss": 1.0642, + "step": 7900 + }, + { + "epoch": 52.04, + "learning_rate": 3.4817164981716496e-05, + "loss": 1.1743, + "step": 7910 + }, + { + "epoch": 52.1, + "learning_rate": 3.478489997849e-05, + "loss": 1.0113, + "step": 7920 + }, + { + "epoch": 52.17, + "learning_rate": 3.47526349752635e-05, + "loss": 1.0284, + "step": 7930 + }, + { + "epoch": 52.24, + "learning_rate": 3.4720369972037e-05, + "loss": 1.0396, + "step": 7940 + }, + { + "epoch": 52.3, + "learning_rate": 3.4688104968810496e-05, + "loss": 1.082, + "step": 7950 + }, + { + "epoch": 52.37, + "learning_rate": 3.4655839965584e-05, + "loss": 1.0178, + "step": 7960 + }, + { + "epoch": 52.43, + "learning_rate": 3.462357496235749e-05, + "loss": 1.0592, + "step": 7970 + }, + { + "epoch": 52.5, + "learning_rate": 3.4591309959130995e-05, + "loss": 1.0798, + "step": 7980 + }, + { + "epoch": 52.56, + "learning_rate": 3.45590449559045e-05, + "loss": 1.0219, + "step": 7990 + }, + { + "epoch": 52.63, + "learning_rate": 3.4526779952678e-05, + "loss": 1.0293, + "step": 8000 + }, + { + "epoch": 52.63, + "eval_loss": 0.36370599269866943, + "eval_runtime": 317.8982, + "eval_samples_per_second": 22.366, + "eval_steps_per_second": 1.4, + "eval_wer": 0.4014323042678796, + "step": 8000 + }, + { + "epoch": 52.69, + "learning_rate": 3.44945149494515e-05, + "loss": 1.0684, + "step": 8010 + }, + { + "epoch": 52.76, + "learning_rate": 3.4462249946224996e-05, + "loss": 1.1228, + "step": 8020 + }, + { + "epoch": 52.82, + "learning_rate": 3.442998494299849e-05, + "loss": 1.0044, + "step": 8030 + }, + { + "epoch": 52.89, + "learning_rate": 3.439771993977199e-05, + "loss": 1.0101, + "step": 8040 + }, + { + "epoch": 52.96, + "learning_rate": 3.4365454936545494e-05, + "loss": 1.0626, + "step": 8050 + }, + { + "epoch": 53.03, + "learning_rate": 3.4333189933318996e-05, + "loss": 1.1676, + "step": 8060 + }, + { + "epoch": 53.09, + "learning_rate": 3.43009249300925e-05, + "loss": 1.0056, + "step": 8070 + }, + { + "epoch": 53.16, + "learning_rate": 3.426865992686599e-05, + "loss": 1.0096, + "step": 8080 + }, + { + "epoch": 53.22, + "learning_rate": 3.4236394923639495e-05, + "loss": 1.061, + "step": 8090 + }, + { + "epoch": 53.29, + "learning_rate": 3.420412992041299e-05, + "loss": 1.1068, + "step": 8100 + }, + { + "epoch": 53.35, + "learning_rate": 3.417186491718649e-05, + "loss": 0.997, + "step": 8110 + }, + { + "epoch": 53.42, + "learning_rate": 3.413959991395999e-05, + "loss": 1.0424, + "step": 8120 + }, + { + "epoch": 53.48, + "learning_rate": 3.4107334910733495e-05, + "loss": 1.085, + "step": 8130 + }, + { + "epoch": 53.55, + "learning_rate": 3.4075069907507e-05, + "loss": 1.057, + "step": 8140 + }, + { + "epoch": 53.62, + "learning_rate": 3.404280490428049e-05, + "loss": 1.0096, + "step": 8150 + }, + { + "epoch": 53.68, + "learning_rate": 3.401053990105399e-05, + "loss": 1.0426, + "step": 8160 + }, + { + "epoch": 53.75, + "learning_rate": 3.397827489782749e-05, + "loss": 1.1106, + "step": 8170 + }, + { + "epoch": 53.81, + "learning_rate": 3.394600989460099e-05, + "loss": 1.0268, + "step": 8180 + }, + { + "epoch": 53.88, + "learning_rate": 3.391374489137449e-05, + "loss": 1.0431, + "step": 8190 + }, + { + "epoch": 53.94, + "learning_rate": 3.3881479888147994e-05, + "loss": 1.071, + "step": 8200 + }, + { + "epoch": 54.01, + "learning_rate": 3.384921488492149e-05, + "loss": 1.1877, + "step": 8210 + }, + { + "epoch": 54.08, + "learning_rate": 3.3816949881694984e-05, + "loss": 0.9915, + "step": 8220 + }, + { + "epoch": 54.14, + "learning_rate": 3.3784684878468486e-05, + "loss": 1.0106, + "step": 8230 + }, + { + "epoch": 54.21, + "learning_rate": 3.375241987524199e-05, + "loss": 1.0178, + "step": 8240 + }, + { + "epoch": 54.27, + "learning_rate": 3.372015487201549e-05, + "loss": 1.096, + "step": 8250 + }, + { + "epoch": 54.34, + "learning_rate": 3.368788986878899e-05, + "loss": 1.0004, + "step": 8260 + }, + { + "epoch": 54.41, + "learning_rate": 3.365562486556249e-05, + "loss": 1.0407, + "step": 8270 + }, + { + "epoch": 54.47, + "learning_rate": 3.362335986233598e-05, + "loss": 1.0405, + "step": 8280 + }, + { + "epoch": 54.54, + "learning_rate": 3.3591094859109483e-05, + "loss": 1.0556, + "step": 8290 + }, + { + "epoch": 54.6, + "learning_rate": 3.3558829855882985e-05, + "loss": 1.012, + "step": 8300 + }, + { + "epoch": 54.67, + "learning_rate": 3.352656485265649e-05, + "loss": 1.0279, + "step": 8310 + }, + { + "epoch": 54.73, + "learning_rate": 3.349429984942999e-05, + "loss": 1.0697, + "step": 8320 + }, + { + "epoch": 54.8, + "learning_rate": 3.346203484620349e-05, + "loss": 1.0495, + "step": 8330 + }, + { + "epoch": 54.86, + "learning_rate": 3.3429769842976986e-05, + "loss": 1.0157, + "step": 8340 + }, + { + "epoch": 54.93, + "learning_rate": 3.339750483975048e-05, + "loss": 1.0558, + "step": 8350 + }, + { + "epoch": 55.0, + "learning_rate": 3.336523983652398e-05, + "loss": 1.1224, + "step": 8360 + }, + { + "epoch": 55.07, + "learning_rate": 3.3332974833297484e-05, + "loss": 1.0981, + "step": 8370 + }, + { + "epoch": 55.13, + "learning_rate": 3.3300709830070986e-05, + "loss": 1.0167, + "step": 8380 + }, + { + "epoch": 55.2, + "learning_rate": 3.326844482684449e-05, + "loss": 1.016, + "step": 8390 + }, + { + "epoch": 55.26, + "learning_rate": 3.323617982361798e-05, + "loss": 1.0748, + "step": 8400 + }, + { + "epoch": 55.33, + "learning_rate": 3.320391482039148e-05, + "loss": 0.9908, + "step": 8410 + }, + { + "epoch": 55.39, + "learning_rate": 3.317164981716498e-05, + "loss": 0.9933, + "step": 8420 + }, + { + "epoch": 55.46, + "learning_rate": 3.313938481393848e-05, + "loss": 1.031, + "step": 8430 + }, + { + "epoch": 55.52, + "learning_rate": 3.3107119810711984e-05, + "loss": 1.0889, + "step": 8440 + }, + { + "epoch": 55.59, + "learning_rate": 3.3074854807485485e-05, + "loss": 0.9891, + "step": 8450 + }, + { + "epoch": 55.65, + "learning_rate": 3.304258980425898e-05, + "loss": 1.0279, + "step": 8460 + }, + { + "epoch": 55.72, + "learning_rate": 3.3010324801032475e-05, + "loss": 1.0436, + "step": 8470 + }, + { + "epoch": 55.79, + "learning_rate": 3.297805979780598e-05, + "loss": 1.097, + "step": 8480 + }, + { + "epoch": 55.85, + "learning_rate": 3.294579479457948e-05, + "loss": 1.0016, + "step": 8490 + }, + { + "epoch": 55.92, + "learning_rate": 3.291352979135298e-05, + "loss": 1.0224, + "step": 8500 + }, + { + "epoch": 55.92, + "eval_loss": 0.35778653621673584, + "eval_runtime": 322.7444, + "eval_samples_per_second": 22.03, + "eval_steps_per_second": 1.379, + "eval_wer": 0.38854156585696314, + "step": 8500 + }, + { + "epoch": 55.98, + "learning_rate": 3.288126478812648e-05, + "loss": 1.0758, + "step": 8510 + }, + { + "epoch": 56.05, + "learning_rate": 3.2848999784899985e-05, + "loss": 1.1085, + "step": 8520 + }, + { + "epoch": 56.12, + "learning_rate": 3.281673478167348e-05, + "loss": 0.9923, + "step": 8530 + }, + { + "epoch": 56.18, + "learning_rate": 3.2784469778446975e-05, + "loss": 1.0392, + "step": 8540 + }, + { + "epoch": 56.25, + "learning_rate": 3.2752204775220476e-05, + "loss": 1.0738, + "step": 8550 + }, + { + "epoch": 56.31, + "learning_rate": 3.271993977199398e-05, + "loss": 1.0161, + "step": 8560 + }, + { + "epoch": 56.38, + "learning_rate": 3.268767476876748e-05, + "loss": 0.982, + "step": 8570 + }, + { + "epoch": 56.45, + "learning_rate": 3.265540976554098e-05, + "loss": 1.0302, + "step": 8580 + }, + { + "epoch": 56.51, + "learning_rate": 3.262314476231448e-05, + "loss": 1.1226, + "step": 8590 + }, + { + "epoch": 56.58, + "learning_rate": 3.259087975908797e-05, + "loss": 1.0014, + "step": 8600 + }, + { + "epoch": 56.64, + "learning_rate": 3.2558614755861474e-05, + "loss": 1.0055, + "step": 8610 + }, + { + "epoch": 56.71, + "learning_rate": 3.2526349752634976e-05, + "loss": 1.0368, + "step": 8620 + }, + { + "epoch": 56.77, + "learning_rate": 3.249408474940848e-05, + "loss": 1.0788, + "step": 8630 + }, + { + "epoch": 56.84, + "learning_rate": 3.246181974618198e-05, + "loss": 0.9906, + "step": 8640 + }, + { + "epoch": 56.9, + "learning_rate": 3.242955474295548e-05, + "loss": 1.0247, + "step": 8650 + }, + { + "epoch": 56.97, + "learning_rate": 3.239728973972897e-05, + "loss": 1.0547, + "step": 8660 + }, + { + "epoch": 57.04, + "learning_rate": 3.236502473650247e-05, + "loss": 1.1407, + "step": 8670 + }, + { + "epoch": 57.1, + "learning_rate": 3.233275973327597e-05, + "loss": 0.9864, + "step": 8680 + }, + { + "epoch": 57.17, + "learning_rate": 3.2300494730049475e-05, + "loss": 1.0169, + "step": 8690 + }, + { + "epoch": 57.24, + "learning_rate": 3.2268229726822976e-05, + "loss": 1.0632, + "step": 8700 + }, + { + "epoch": 57.3, + "learning_rate": 3.223596472359648e-05, + "loss": 1.0245, + "step": 8710 + }, + { + "epoch": 57.37, + "learning_rate": 3.220369972036997e-05, + "loss": 0.9978, + "step": 8720 + }, + { + "epoch": 57.43, + "learning_rate": 3.217143471714347e-05, + "loss": 1.0595, + "step": 8730 + }, + { + "epoch": 57.5, + "learning_rate": 3.213916971391697e-05, + "loss": 1.0714, + "step": 8740 + }, + { + "epoch": 57.56, + "learning_rate": 3.210690471069047e-05, + "loss": 1.0067, + "step": 8750 + }, + { + "epoch": 57.63, + "learning_rate": 3.2074639707463974e-05, + "loss": 1.0159, + "step": 8760 + }, + { + "epoch": 57.69, + "learning_rate": 3.2042374704237476e-05, + "loss": 1.0272, + "step": 8770 + }, + { + "epoch": 57.76, + "learning_rate": 3.201010970101097e-05, + "loss": 1.1016, + "step": 8780 + }, + { + "epoch": 57.82, + "learning_rate": 3.1977844697784466e-05, + "loss": 0.9925, + "step": 8790 + }, + { + "epoch": 57.89, + "learning_rate": 3.194557969455797e-05, + "loss": 1.0156, + "step": 8800 + }, + { + "epoch": 57.96, + "learning_rate": 3.191331469133147e-05, + "loss": 1.016, + "step": 8810 + }, + { + "epoch": 58.03, + "learning_rate": 3.188104968810497e-05, + "loss": 1.1536, + "step": 8820 + }, + { + "epoch": 58.09, + "learning_rate": 3.184878468487847e-05, + "loss": 0.9755, + "step": 8830 + }, + { + "epoch": 58.16, + "learning_rate": 3.1816519681651975e-05, + "loss": 0.9959, + "step": 8840 + }, + { + "epoch": 58.22, + "learning_rate": 3.178425467842546e-05, + "loss": 1.0229, + "step": 8850 + }, + { + "epoch": 58.29, + "learning_rate": 3.1751989675198965e-05, + "loss": 1.0698, + "step": 8860 + }, + { + "epoch": 58.35, + "learning_rate": 3.171972467197247e-05, + "loss": 0.9774, + "step": 8870 + }, + { + "epoch": 58.42, + "learning_rate": 3.168745966874597e-05, + "loss": 1.0372, + "step": 8880 + }, + { + "epoch": 58.48, + "learning_rate": 3.165519466551947e-05, + "loss": 1.0503, + "step": 8890 + }, + { + "epoch": 58.55, + "learning_rate": 3.162292966229297e-05, + "loss": 1.0177, + "step": 8900 + }, + { + "epoch": 58.62, + "learning_rate": 3.159066465906647e-05, + "loss": 0.9865, + "step": 8910 + }, + { + "epoch": 58.68, + "learning_rate": 3.155839965583996e-05, + "loss": 1.0472, + "step": 8920 + }, + { + "epoch": 58.75, + "learning_rate": 3.1526134652613464e-05, + "loss": 1.0973, + "step": 8930 + }, + { + "epoch": 58.81, + "learning_rate": 3.1493869649386966e-05, + "loss": 0.991, + "step": 8940 + }, + { + "epoch": 58.88, + "learning_rate": 3.146160464616047e-05, + "loss": 0.9552, + "step": 8950 + }, + { + "epoch": 58.94, + "learning_rate": 3.142933964293397e-05, + "loss": 0.9968, + "step": 8960 + }, + { + "epoch": 59.01, + "learning_rate": 3.1397074639707464e-05, + "loss": 1.1483, + "step": 8970 + }, + { + "epoch": 59.08, + "learning_rate": 3.136480963648096e-05, + "loss": 1.0066, + "step": 8980 + }, + { + "epoch": 59.14, + "learning_rate": 3.133254463325446e-05, + "loss": 1.0109, + "step": 8990 + }, + { + "epoch": 59.21, + "learning_rate": 3.130027963002796e-05, + "loss": 1.012, + "step": 9000 + }, + { + "epoch": 59.21, + "eval_loss": 0.36289963126182556, + "eval_runtime": 322.3576, + "eval_samples_per_second": 22.056, + "eval_steps_per_second": 1.38, + "eval_wer": 0.3929933223652376, + "step": 9000 + }, + { + "epoch": 59.27, + "learning_rate": 3.1268014626801465e-05, + "loss": 1.06, + "step": 9010 + }, + { + "epoch": 59.34, + "learning_rate": 3.123574962357497e-05, + "loss": 0.9861, + "step": 9020 + }, + { + "epoch": 59.41, + "learning_rate": 3.120348462034847e-05, + "loss": 1.012, + "step": 9030 + }, + { + "epoch": 59.47, + "learning_rate": 3.117121961712196e-05, + "loss": 1.0072, + "step": 9040 + }, + { + "epoch": 59.54, + "learning_rate": 3.113895461389546e-05, + "loss": 1.0687, + "step": 9050 + }, + { + "epoch": 59.6, + "learning_rate": 3.110668961066896e-05, + "loss": 0.9707, + "step": 9060 + }, + { + "epoch": 59.67, + "learning_rate": 3.107442460744246e-05, + "loss": 1.0095, + "step": 9070 + }, + { + "epoch": 59.73, + "learning_rate": 3.1042159604215964e-05, + "loss": 1.0323, + "step": 9080 + }, + { + "epoch": 59.8, + "learning_rate": 3.1009894600989466e-05, + "loss": 1.0193, + "step": 9090 + }, + { + "epoch": 59.86, + "learning_rate": 3.097762959776296e-05, + "loss": 0.9957, + "step": 9100 + }, + { + "epoch": 59.93, + "learning_rate": 3.0945364594536456e-05, + "loss": 1.0012, + "step": 9110 + }, + { + "epoch": 60.0, + "learning_rate": 3.091309959130996e-05, + "loss": 1.1186, + "step": 9120 + }, + { + "epoch": 60.07, + "learning_rate": 3.088083458808346e-05, + "loss": 1.0609, + "step": 9130 + }, + { + "epoch": 60.13, + "learning_rate": 3.084856958485696e-05, + "loss": 0.987, + "step": 9140 + }, + { + "epoch": 60.2, + "learning_rate": 3.081630458163046e-05, + "loss": 0.9959, + "step": 9150 + }, + { + "epoch": 60.26, + "learning_rate": 3.0784039578403965e-05, + "loss": 1.0595, + "step": 9160 + }, + { + "epoch": 60.33, + "learning_rate": 3.075177457517745e-05, + "loss": 0.9742, + "step": 9170 + }, + { + "epoch": 60.39, + "learning_rate": 3.0719509571950955e-05, + "loss": 1.0124, + "step": 9180 + }, + { + "epoch": 60.46, + "learning_rate": 3.068724456872446e-05, + "loss": 1.0143, + "step": 9190 + }, + { + "epoch": 60.52, + "learning_rate": 3.065497956549796e-05, + "loss": 1.063, + "step": 9200 + }, + { + "epoch": 60.59, + "learning_rate": 3.062271456227146e-05, + "loss": 0.9921, + "step": 9210 + }, + { + "epoch": 60.65, + "learning_rate": 3.059044955904496e-05, + "loss": 0.9886, + "step": 9220 + }, + { + "epoch": 60.72, + "learning_rate": 3.055818455581845e-05, + "loss": 1.0083, + "step": 9230 + }, + { + "epoch": 60.79, + "learning_rate": 3.052591955259195e-05, + "loss": 1.0394, + "step": 9240 + }, + { + "epoch": 60.85, + "learning_rate": 3.0493654549365454e-05, + "loss": 0.9481, + "step": 9250 + }, + { + "epoch": 60.92, + "learning_rate": 3.0461389546138956e-05, + "loss": 1.0265, + "step": 9260 + }, + { + "epoch": 60.98, + "learning_rate": 3.0429124542912458e-05, + "loss": 1.046, + "step": 9270 + }, + { + "epoch": 61.05, + "learning_rate": 3.0396859539685956e-05, + "loss": 1.094, + "step": 9280 + }, + { + "epoch": 61.12, + "learning_rate": 3.036459453645945e-05, + "loss": 0.9846, + "step": 9290 + }, + { + "epoch": 61.18, + "learning_rate": 3.0332329533232953e-05, + "loss": 1.0055, + "step": 9300 + }, + { + "epoch": 61.25, + "learning_rate": 3.030006453000645e-05, + "loss": 1.0646, + "step": 9310 + }, + { + "epoch": 61.31, + "learning_rate": 3.0267799526779953e-05, + "loss": 0.9717, + "step": 9320 + }, + { + "epoch": 61.38, + "learning_rate": 3.0235534523553455e-05, + "loss": 0.986, + "step": 9330 + }, + { + "epoch": 61.45, + "learning_rate": 3.0203269520326954e-05, + "loss": 1.0067, + "step": 9340 + }, + { + "epoch": 61.51, + "learning_rate": 3.0171004517100455e-05, + "loss": 1.0631, + "step": 9350 + }, + { + "epoch": 61.58, + "learning_rate": 3.013873951387395e-05, + "loss": 0.9652, + "step": 9360 + }, + { + "epoch": 61.64, + "learning_rate": 3.010647451064745e-05, + "loss": 0.9762, + "step": 9370 + }, + { + "epoch": 61.71, + "learning_rate": 3.007420950742095e-05, + "loss": 1.015, + "step": 9380 + }, + { + "epoch": 61.77, + "learning_rate": 3.0041944504194452e-05, + "loss": 1.064, + "step": 9390 + }, + { + "epoch": 61.84, + "learning_rate": 3.000967950096795e-05, + "loss": 0.9729, + "step": 9400 + }, + { + "epoch": 61.9, + "learning_rate": 2.997741449774145e-05, + "loss": 1.0035, + "step": 9410 + }, + { + "epoch": 61.97, + "learning_rate": 2.994514949451495e-05, + "loss": 0.9994, + "step": 9420 + }, + { + "epoch": 62.04, + "learning_rate": 2.9912884491288453e-05, + "loss": 1.1234, + "step": 9430 + }, + { + "epoch": 62.1, + "learning_rate": 2.9880619488061948e-05, + "loss": 0.9815, + "step": 9440 + }, + { + "epoch": 62.17, + "learning_rate": 2.984835448483545e-05, + "loss": 0.9917, + "step": 9450 + }, + { + "epoch": 62.24, + "learning_rate": 2.981608948160895e-05, + "loss": 1.0241, + "step": 9460 + }, + { + "epoch": 62.3, + "learning_rate": 2.9783824478382447e-05, + "loss": 1.0205, + "step": 9470 + }, + { + "epoch": 62.37, + "learning_rate": 2.975155947515595e-05, + "loss": 0.9658, + "step": 9480 + }, + { + "epoch": 62.43, + "learning_rate": 2.971929447192945e-05, + "loss": 1.0278, + "step": 9490 + }, + { + "epoch": 62.5, + "learning_rate": 2.9687029468702945e-05, + "loss": 1.0772, + "step": 9500 + }, + { + "epoch": 62.5, + "eval_loss": 0.36351296305656433, + "eval_runtime": 322.4995, + "eval_samples_per_second": 22.047, + "eval_steps_per_second": 1.38, + "eval_wer": 0.3905738894803058, + "step": 9500 + }, + { + "epoch": 62.56, + "learning_rate": 2.9654764465476447e-05, + "loss": 0.9703, + "step": 9510 + }, + { + "epoch": 62.63, + "learning_rate": 2.962249946224995e-05, + "loss": 0.9938, + "step": 9520 + }, + { + "epoch": 62.69, + "learning_rate": 2.9590234459023444e-05, + "loss": 0.9653, + "step": 9530 + }, + { + "epoch": 62.76, + "learning_rate": 2.9557969455796946e-05, + "loss": 1.0557, + "step": 9540 + }, + { + "epoch": 62.82, + "learning_rate": 2.9525704452570448e-05, + "loss": 0.9614, + "step": 9550 + }, + { + "epoch": 62.89, + "learning_rate": 2.9493439449343943e-05, + "loss": 0.9751, + "step": 9560 + }, + { + "epoch": 62.96, + "learning_rate": 2.9461174446117444e-05, + "loss": 0.991, + "step": 9570 + }, + { + "epoch": 63.03, + "learning_rate": 2.9428909442890946e-05, + "loss": 1.1272, + "step": 9580 + }, + { + "epoch": 63.09, + "learning_rate": 2.9396644439664445e-05, + "loss": 0.9885, + "step": 9590 + }, + { + "epoch": 63.16, + "learning_rate": 2.9367605936760594e-05, + "loss": 1.0147, + "step": 9600 + }, + { + "epoch": 63.22, + "learning_rate": 2.9335340933534096e-05, + "loss": 1.0171, + "step": 9610 + }, + { + "epoch": 63.29, + "learning_rate": 2.930307593030759e-05, + "loss": 1.0517, + "step": 9620 + }, + { + "epoch": 63.35, + "learning_rate": 2.9270810927081093e-05, + "loss": 0.969, + "step": 9630 + }, + { + "epoch": 63.42, + "learning_rate": 2.9238545923854595e-05, + "loss": 0.981, + "step": 9640 + }, + { + "epoch": 63.48, + "learning_rate": 2.920628092062809e-05, + "loss": 1.0505, + "step": 9650 + }, + { + "epoch": 63.55, + "learning_rate": 2.917401591740159e-05, + "loss": 1.0056, + "step": 9660 + }, + { + "epoch": 63.62, + "learning_rate": 2.9141750914175093e-05, + "loss": 0.9903, + "step": 9670 + }, + { + "epoch": 63.68, + "learning_rate": 2.9109485910948592e-05, + "loss": 1.018, + "step": 9680 + }, + { + "epoch": 63.75, + "learning_rate": 2.907722090772209e-05, + "loss": 1.0411, + "step": 9690 + }, + { + "epoch": 63.81, + "learning_rate": 2.9044955904495592e-05, + "loss": 0.9895, + "step": 9700 + }, + { + "epoch": 63.88, + "learning_rate": 2.901269090126909e-05, + "loss": 0.9717, + "step": 9710 + }, + { + "epoch": 63.94, + "learning_rate": 2.898042589804259e-05, + "loss": 1.0232, + "step": 9720 + }, + { + "epoch": 64.01, + "learning_rate": 2.894816089481609e-05, + "loss": 1.1187, + "step": 9730 + }, + { + "epoch": 64.08, + "learning_rate": 2.8915895891589593e-05, + "loss": 0.9369, + "step": 9740 + }, + { + "epoch": 64.14, + "learning_rate": 2.8883630888363088e-05, + "loss": 1.0006, + "step": 9750 + }, + { + "epoch": 64.21, + "learning_rate": 2.885136588513659e-05, + "loss": 0.9803, + "step": 9760 + }, + { + "epoch": 64.27, + "learning_rate": 2.881910088191009e-05, + "loss": 1.0538, + "step": 9770 + }, + { + "epoch": 64.34, + "learning_rate": 2.8786835878683586e-05, + "loss": 0.9796, + "step": 9780 + }, + { + "epoch": 64.41, + "learning_rate": 2.8754570875457088e-05, + "loss": 0.9798, + "step": 9790 + }, + { + "epoch": 64.47, + "learning_rate": 2.872230587223059e-05, + "loss": 1.0311, + "step": 9800 + }, + { + "epoch": 64.54, + "learning_rate": 2.8690040869004085e-05, + "loss": 1.0271, + "step": 9810 + }, + { + "epoch": 64.6, + "learning_rate": 2.8657775865777587e-05, + "loss": 0.9556, + "step": 9820 + }, + { + "epoch": 64.67, + "learning_rate": 2.862551086255109e-05, + "loss": 0.9889, + "step": 9830 + }, + { + "epoch": 64.73, + "learning_rate": 2.8593245859324587e-05, + "loss": 1.0261, + "step": 9840 + }, + { + "epoch": 64.8, + "learning_rate": 2.8560980856098085e-05, + "loss": 0.993, + "step": 9850 + }, + { + "epoch": 64.86, + "learning_rate": 2.8528715852871587e-05, + "loss": 0.9598, + "step": 9860 + }, + { + "epoch": 64.93, + "learning_rate": 2.8496450849645086e-05, + "loss": 1.0138, + "step": 9870 + }, + { + "epoch": 65.0, + "learning_rate": 2.8464185846418584e-05, + "loss": 1.0809, + "step": 9880 + }, + { + "epoch": 65.07, + "learning_rate": 2.8431920843192086e-05, + "loss": 1.0243, + "step": 9890 + }, + { + "epoch": 65.13, + "learning_rate": 2.8399655839965584e-05, + "loss": 0.9741, + "step": 9900 + }, + { + "epoch": 65.2, + "learning_rate": 2.8367390836739083e-05, + "loss": 1.0103, + "step": 9910 + }, + { + "epoch": 65.26, + "learning_rate": 2.8335125833512585e-05, + "loss": 1.0557, + "step": 9920 + }, + { + "epoch": 65.33, + "learning_rate": 2.8302860830286086e-05, + "loss": 0.954, + "step": 9930 + }, + { + "epoch": 65.39, + "learning_rate": 2.827059582705958e-05, + "loss": 0.9687, + "step": 9940 + }, + { + "epoch": 65.46, + "learning_rate": 2.8238330823833083e-05, + "loss": 1.0184, + "step": 9950 + }, + { + "epoch": 65.52, + "learning_rate": 2.8206065820606585e-05, + "loss": 1.0414, + "step": 9960 + }, + { + "epoch": 65.59, + "learning_rate": 2.817380081738008e-05, + "loss": 0.9417, + "step": 9970 + }, + { + "epoch": 65.65, + "learning_rate": 2.8141535814153582e-05, + "loss": 0.9696, + "step": 9980 + }, + { + "epoch": 65.72, + "learning_rate": 2.8109270810927084e-05, + "loss": 0.9997, + "step": 9990 + }, + { + "epoch": 65.79, + "learning_rate": 2.8077005807700582e-05, + "loss": 1.0344, + "step": 10000 + }, + { + "epoch": 65.79, + "eval_loss": 0.3469240963459015, + "eval_runtime": 323.3838, + "eval_samples_per_second": 21.986, + "eval_steps_per_second": 1.376, + "eval_wer": 0.37708313171392627, + "step": 10000 + }, + { + "epoch": 65.85, + "learning_rate": 2.804474080447408e-05, + "loss": 0.9707, + "step": 10010 + }, + { + "epoch": 65.92, + "learning_rate": 2.8012475801247582e-05, + "loss": 1.0091, + "step": 10020 + }, + { + "epoch": 65.98, + "learning_rate": 2.798021079802108e-05, + "loss": 1.058, + "step": 10030 + }, + { + "epoch": 66.05, + "learning_rate": 2.794794579479458e-05, + "loss": 1.0737, + "step": 10040 + }, + { + "epoch": 66.12, + "learning_rate": 2.791568079156808e-05, + "loss": 0.9489, + "step": 10050 + }, + { + "epoch": 66.18, + "learning_rate": 2.788341578834158e-05, + "loss": 0.9957, + "step": 10060 + }, + { + "epoch": 66.25, + "learning_rate": 2.7851150785115078e-05, + "loss": 1.037, + "step": 10070 + }, + { + "epoch": 66.31, + "learning_rate": 2.781888578188858e-05, + "loss": 0.9564, + "step": 10080 + }, + { + "epoch": 66.38, + "learning_rate": 2.7786620778662078e-05, + "loss": 0.9497, + "step": 10090 + }, + { + "epoch": 66.45, + "learning_rate": 2.7754355775435576e-05, + "loss": 0.9675, + "step": 10100 + }, + { + "epoch": 66.51, + "learning_rate": 2.7722090772209078e-05, + "loss": 1.0325, + "step": 10110 + }, + { + "epoch": 66.58, + "learning_rate": 2.768982576898258e-05, + "loss": 0.9818, + "step": 10120 + }, + { + "epoch": 66.64, + "learning_rate": 2.7657560765756075e-05, + "loss": 0.9883, + "step": 10130 + }, + { + "epoch": 66.71, + "learning_rate": 2.7625295762529577e-05, + "loss": 1.0175, + "step": 10140 + }, + { + "epoch": 66.77, + "learning_rate": 2.759303075930308e-05, + "loss": 1.0583, + "step": 10150 + }, + { + "epoch": 66.84, + "learning_rate": 2.7560765756076574e-05, + "loss": 0.9348, + "step": 10160 + }, + { + "epoch": 66.9, + "learning_rate": 2.7528500752850076e-05, + "loss": 0.9898, + "step": 10170 + }, + { + "epoch": 66.97, + "learning_rate": 2.7496235749623577e-05, + "loss": 0.9972, + "step": 10180 + }, + { + "epoch": 67.04, + "learning_rate": 2.7463970746397076e-05, + "loss": 1.1137, + "step": 10190 + }, + { + "epoch": 67.1, + "learning_rate": 2.7431705743170574e-05, + "loss": 0.9396, + "step": 10200 + }, + { + "epoch": 67.17, + "learning_rate": 2.7399440739944076e-05, + "loss": 0.9777, + "step": 10210 + }, + { + "epoch": 67.24, + "learning_rate": 2.7367175736717575e-05, + "loss": 1.0345, + "step": 10220 + }, + { + "epoch": 67.3, + "learning_rate": 2.7334910733491073e-05, + "loss": 1.0047, + "step": 10230 + }, + { + "epoch": 67.37, + "learning_rate": 2.7302645730264575e-05, + "loss": 0.9188, + "step": 10240 + }, + { + "epoch": 67.43, + "learning_rate": 2.7270380727038073e-05, + "loss": 0.9834, + "step": 10250 + }, + { + "epoch": 67.5, + "learning_rate": 2.723811572381157e-05, + "loss": 1.038, + "step": 10260 + }, + { + "epoch": 67.56, + "learning_rate": 2.7205850720585073e-05, + "loss": 0.9701, + "step": 10270 + }, + { + "epoch": 67.63, + "learning_rate": 2.7173585717358572e-05, + "loss": 0.99, + "step": 10280 + }, + { + "epoch": 67.69, + "learning_rate": 2.714132071413207e-05, + "loss": 0.9713, + "step": 10290 + }, + { + "epoch": 67.76, + "learning_rate": 2.7109055710905572e-05, + "loss": 1.033, + "step": 10300 + }, + { + "epoch": 67.82, + "learning_rate": 2.707679070767907e-05, + "loss": 0.947, + "step": 10310 + }, + { + "epoch": 67.89, + "learning_rate": 2.704452570445257e-05, + "loss": 0.9757, + "step": 10320 + }, + { + "epoch": 67.96, + "learning_rate": 2.701226070122607e-05, + "loss": 0.9942, + "step": 10330 + }, + { + "epoch": 68.03, + "learning_rate": 2.6979995697999573e-05, + "loss": 1.108, + "step": 10340 + }, + { + "epoch": 68.09, + "learning_rate": 2.694773069477307e-05, + "loss": 0.947, + "step": 10350 + }, + { + "epoch": 68.16, + "learning_rate": 2.691546569154657e-05, + "loss": 0.974, + "step": 10360 + }, + { + "epoch": 68.22, + "learning_rate": 2.688320068832007e-05, + "loss": 0.9951, + "step": 10370 + }, + { + "epoch": 68.29, + "learning_rate": 2.685093568509357e-05, + "loss": 1.0165, + "step": 10380 + }, + { + "epoch": 68.35, + "learning_rate": 2.6818670681867068e-05, + "loss": 0.9588, + "step": 10390 + }, + { + "epoch": 68.42, + "learning_rate": 2.678640567864057e-05, + "loss": 0.9858, + "step": 10400 + }, + { + "epoch": 68.48, + "learning_rate": 2.6754140675414068e-05, + "loss": 1.0087, + "step": 10410 + }, + { + "epoch": 68.55, + "learning_rate": 2.6721875672187567e-05, + "loss": 0.9829, + "step": 10420 + }, + { + "epoch": 68.62, + "learning_rate": 2.668961066896107e-05, + "loss": 0.9589, + "step": 10430 + }, + { + "epoch": 68.68, + "learning_rate": 2.6657345665734567e-05, + "loss": 0.9937, + "step": 10440 + }, + { + "epoch": 68.75, + "learning_rate": 2.6625080662508065e-05, + "loss": 1.0606, + "step": 10450 + }, + { + "epoch": 68.81, + "learning_rate": 2.6592815659281567e-05, + "loss": 0.9556, + "step": 10460 + }, + { + "epoch": 68.88, + "learning_rate": 2.6560550656055066e-05, + "loss": 0.9569, + "step": 10470 + }, + { + "epoch": 68.94, + "learning_rate": 2.6528285652828564e-05, + "loss": 0.9859, + "step": 10480 + }, + { + "epoch": 69.01, + "learning_rate": 2.6496020649602066e-05, + "loss": 1.1064, + "step": 10490 + }, + { + "epoch": 69.08, + "learning_rate": 2.6463755646375564e-05, + "loss": 0.9457, + "step": 10500 + }, + { + "epoch": 69.08, + "eval_loss": 0.3435259163379669, + "eval_runtime": 322.1557, + "eval_samples_per_second": 22.07, + "eval_steps_per_second": 1.381, + "eval_wer": 0.37350237104422723, + "step": 10500 + }, + { + "epoch": 69.14, + "learning_rate": 2.6431490643149066e-05, + "loss": 0.9544, + "step": 10510 + }, + { + "epoch": 69.21, + "learning_rate": 2.6399225639922565e-05, + "loss": 0.9302, + "step": 10520 + }, + { + "epoch": 69.27, + "learning_rate": 2.6366960636696066e-05, + "loss": 1.0351, + "step": 10530 + }, + { + "epoch": 69.34, + "learning_rate": 2.6334695633469565e-05, + "loss": 0.9441, + "step": 10540 + }, + { + "epoch": 69.41, + "learning_rate": 2.6302430630243063e-05, + "loss": 1.0024, + "step": 10550 + }, + { + "epoch": 69.47, + "learning_rate": 2.6270165627016565e-05, + "loss": 0.9795, + "step": 10560 + }, + { + "epoch": 69.54, + "learning_rate": 2.6237900623790063e-05, + "loss": 1.0247, + "step": 10570 + }, + { + "epoch": 69.6, + "learning_rate": 2.6205635620563562e-05, + "loss": 0.9349, + "step": 10580 + }, + { + "epoch": 69.67, + "learning_rate": 2.6173370617337064e-05, + "loss": 0.9871, + "step": 10590 + }, + { + "epoch": 69.73, + "learning_rate": 2.6141105614110562e-05, + "loss": 1.0088, + "step": 10600 + }, + { + "epoch": 69.8, + "learning_rate": 2.610884061088406e-05, + "loss": 0.9886, + "step": 10610 + }, + { + "epoch": 69.86, + "learning_rate": 2.6076575607657562e-05, + "loss": 0.9544, + "step": 10620 + }, + { + "epoch": 69.93, + "learning_rate": 2.604431060443106e-05, + "loss": 0.979, + "step": 10630 + }, + { + "epoch": 70.0, + "learning_rate": 2.601204560120456e-05, + "loss": 0.9918, + "step": 10640 + }, + { + "epoch": 70.07, + "learning_rate": 2.597978059797806e-05, + "loss": 1.0281, + "step": 10650 + }, + { + "epoch": 70.13, + "learning_rate": 2.594751559475156e-05, + "loss": 0.9626, + "step": 10660 + }, + { + "epoch": 70.2, + "learning_rate": 2.5915250591525058e-05, + "loss": 1.0007, + "step": 10670 + }, + { + "epoch": 70.26, + "learning_rate": 2.588298558829856e-05, + "loss": 1.0266, + "step": 10680 + }, + { + "epoch": 70.33, + "learning_rate": 2.5850720585072058e-05, + "loss": 0.9427, + "step": 10690 + }, + { + "epoch": 70.39, + "learning_rate": 2.581845558184556e-05, + "loss": 0.9498, + "step": 10700 + }, + { + "epoch": 70.46, + "learning_rate": 2.5786190578619058e-05, + "loss": 0.9637, + "step": 10710 + }, + { + "epoch": 70.52, + "learning_rate": 2.575392557539256e-05, + "loss": 1.0638, + "step": 10720 + }, + { + "epoch": 70.59, + "learning_rate": 2.572166057216606e-05, + "loss": 0.9378, + "step": 10730 + }, + { + "epoch": 70.65, + "learning_rate": 2.5689395568939557e-05, + "loss": 0.9692, + "step": 10740 + }, + { + "epoch": 70.72, + "learning_rate": 2.565713056571306e-05, + "loss": 0.968, + "step": 10750 + }, + { + "epoch": 70.79, + "learning_rate": 2.5624865562486557e-05, + "loss": 1.0133, + "step": 10760 + }, + { + "epoch": 70.85, + "learning_rate": 2.5592600559260056e-05, + "loss": 0.9445, + "step": 10770 + }, + { + "epoch": 70.92, + "learning_rate": 2.5560335556033557e-05, + "loss": 0.9751, + "step": 10780 + }, + { + "epoch": 70.98, + "learning_rate": 2.5528070552807056e-05, + "loss": 0.9768, + "step": 10790 + }, + { + "epoch": 71.05, + "learning_rate": 2.5495805549580554e-05, + "loss": 1.058, + "step": 10800 + }, + { + "epoch": 71.12, + "learning_rate": 2.5463540546354056e-05, + "loss": 0.9296, + "step": 10810 + }, + { + "epoch": 71.18, + "learning_rate": 2.5431275543127555e-05, + "loss": 0.9902, + "step": 10820 + }, + { + "epoch": 71.25, + "learning_rate": 2.5399010539901053e-05, + "loss": 1.012, + "step": 10830 + }, + { + "epoch": 71.31, + "learning_rate": 2.5366745536674555e-05, + "loss": 0.953, + "step": 10840 + }, + { + "epoch": 71.38, + "learning_rate": 2.5334480533448053e-05, + "loss": 0.9268, + "step": 10850 + }, + { + "epoch": 71.45, + "learning_rate": 2.5302215530221555e-05, + "loss": 0.9632, + "step": 10860 + }, + { + "epoch": 71.51, + "learning_rate": 2.5269950526995053e-05, + "loss": 1.0485, + "step": 10870 + }, + { + "epoch": 71.58, + "learning_rate": 2.5237685523768552e-05, + "loss": 0.9445, + "step": 10880 + }, + { + "epoch": 71.64, + "learning_rate": 2.5205420520542054e-05, + "loss": 0.9855, + "step": 10890 + }, + { + "epoch": 71.71, + "learning_rate": 2.5173155517315552e-05, + "loss": 0.9577, + "step": 10900 + }, + { + "epoch": 71.77, + "learning_rate": 2.514089051408905e-05, + "loss": 1.0271, + "step": 10910 + }, + { + "epoch": 71.84, + "learning_rate": 2.5108625510862552e-05, + "loss": 0.9161, + "step": 10920 + }, + { + "epoch": 71.9, + "learning_rate": 2.507636050763605e-05, + "loss": 0.9792, + "step": 10930 + }, + { + "epoch": 71.97, + "learning_rate": 2.5044095504409553e-05, + "loss": 0.9772, + "step": 10940 + }, + { + "epoch": 72.04, + "learning_rate": 2.501183050118305e-05, + "loss": 1.1076, + "step": 10950 + }, + { + "epoch": 72.1, + "learning_rate": 2.497956549795655e-05, + "loss": 0.9289, + "step": 10960 + }, + { + "epoch": 72.17, + "learning_rate": 2.494730049473005e-05, + "loss": 0.9363, + "step": 10970 + }, + { + "epoch": 72.24, + "learning_rate": 2.491503549150355e-05, + "loss": 0.9921, + "step": 10980 + }, + { + "epoch": 72.3, + "learning_rate": 2.4882770488277048e-05, + "loss": 0.9726, + "step": 10990 + }, + { + "epoch": 72.37, + "learning_rate": 2.485050548505055e-05, + "loss": 0.9307, + "step": 11000 + }, + { + "epoch": 72.37, + "eval_loss": 0.3518897294998169, + "eval_runtime": 318.8359, + "eval_samples_per_second": 22.3, + "eval_steps_per_second": 1.396, + "eval_wer": 0.37615406948611246, + "step": 11000 + }, + { + "epoch": 72.43, + "learning_rate": 2.481824048182405e-05, + "loss": 0.9837, + "step": 11010 + }, + { + "epoch": 72.5, + "learning_rate": 2.478597547859755e-05, + "loss": 1.0279, + "step": 11020 + }, + { + "epoch": 72.56, + "learning_rate": 2.475371047537105e-05, + "loss": 0.934, + "step": 11030 + }, + { + "epoch": 72.63, + "learning_rate": 2.4721445472144547e-05, + "loss": 0.9445, + "step": 11040 + }, + { + "epoch": 72.69, + "learning_rate": 2.468918046891805e-05, + "loss": 0.9884, + "step": 11050 + }, + { + "epoch": 72.76, + "learning_rate": 2.4656915465691547e-05, + "loss": 0.9982, + "step": 11060 + }, + { + "epoch": 72.82, + "learning_rate": 2.4624650462465046e-05, + "loss": 0.9551, + "step": 11070 + }, + { + "epoch": 72.89, + "learning_rate": 2.4592385459238547e-05, + "loss": 0.9328, + "step": 11080 + }, + { + "epoch": 72.96, + "learning_rate": 2.4560120456012046e-05, + "loss": 0.9785, + "step": 11090 + }, + { + "epoch": 73.03, + "learning_rate": 2.4527855452785544e-05, + "loss": 1.1105, + "step": 11100 + }, + { + "epoch": 73.09, + "learning_rate": 2.4495590449559046e-05, + "loss": 0.9251, + "step": 11110 + }, + { + "epoch": 73.16, + "learning_rate": 2.4463325446332545e-05, + "loss": 0.9442, + "step": 11120 + }, + { + "epoch": 73.22, + "learning_rate": 2.4431060443106046e-05, + "loss": 0.9843, + "step": 11130 + }, + { + "epoch": 73.29, + "learning_rate": 2.4398795439879545e-05, + "loss": 1.0037, + "step": 11140 + }, + { + "epoch": 73.35, + "learning_rate": 2.4366530436653043e-05, + "loss": 0.921, + "step": 11150 + }, + { + "epoch": 73.42, + "learning_rate": 2.4334265433426545e-05, + "loss": 0.9825, + "step": 11160 + }, + { + "epoch": 73.48, + "learning_rate": 2.4302000430200043e-05, + "loss": 0.9837, + "step": 11170 + }, + { + "epoch": 73.55, + "learning_rate": 2.4269735426973545e-05, + "loss": 0.998, + "step": 11180 + }, + { + "epoch": 73.62, + "learning_rate": 2.4237470423747044e-05, + "loss": 0.9317, + "step": 11190 + }, + { + "epoch": 73.68, + "learning_rate": 2.4205205420520542e-05, + "loss": 0.958, + "step": 11200 + }, + { + "epoch": 73.75, + "learning_rate": 2.4172940417294044e-05, + "loss": 1.0169, + "step": 11210 + }, + { + "epoch": 73.81, + "learning_rate": 2.4140675414067542e-05, + "loss": 0.9379, + "step": 11220 + }, + { + "epoch": 73.88, + "learning_rate": 2.410841041084104e-05, + "loss": 0.9534, + "step": 11230 + }, + { + "epoch": 73.94, + "learning_rate": 2.4076145407614543e-05, + "loss": 0.9791, + "step": 11240 + }, + { + "epoch": 74.01, + "learning_rate": 2.404388040438804e-05, + "loss": 1.1134, + "step": 11250 + }, + { + "epoch": 74.08, + "learning_rate": 2.401161540116154e-05, + "loss": 0.9459, + "step": 11260 + }, + { + "epoch": 74.14, + "learning_rate": 2.397935039793504e-05, + "loss": 0.9346, + "step": 11270 + }, + { + "epoch": 74.21, + "learning_rate": 2.394708539470854e-05, + "loss": 0.932, + "step": 11280 + }, + { + "epoch": 74.27, + "learning_rate": 2.3914820391482038e-05, + "loss": 1.0175, + "step": 11290 + }, + { + "epoch": 74.34, + "learning_rate": 2.388255538825554e-05, + "loss": 0.9262, + "step": 11300 + }, + { + "epoch": 74.41, + "learning_rate": 2.385029038502904e-05, + "loss": 0.9554, + "step": 11310 + }, + { + "epoch": 74.47, + "learning_rate": 2.3818025381802537e-05, + "loss": 0.9507, + "step": 11320 + }, + { + "epoch": 74.54, + "learning_rate": 2.378576037857604e-05, + "loss": 0.9911, + "step": 11330 + }, + { + "epoch": 74.6, + "learning_rate": 2.3753495375349537e-05, + "loss": 0.9243, + "step": 11340 + }, + { + "epoch": 74.67, + "learning_rate": 2.372123037212304e-05, + "loss": 0.9713, + "step": 11350 + }, + { + "epoch": 74.73, + "learning_rate": 2.3688965368896537e-05, + "loss": 1.0012, + "step": 11360 + }, + { + "epoch": 74.8, + "learning_rate": 2.365670036567004e-05, + "loss": 0.9809, + "step": 11370 + }, + { + "epoch": 74.86, + "learning_rate": 2.3624435362443537e-05, + "loss": 0.9423, + "step": 11380 + }, + { + "epoch": 74.93, + "learning_rate": 2.3592170359217036e-05, + "loss": 0.9665, + "step": 11390 + }, + { + "epoch": 75.0, + "learning_rate": 2.3559905355990538e-05, + "loss": 1.0262, + "step": 11400 + }, + { + "epoch": 75.07, + "learning_rate": 2.3527640352764036e-05, + "loss": 1.027, + "step": 11410 + }, + { + "epoch": 75.13, + "learning_rate": 2.3495375349537535e-05, + "loss": 0.9059, + "step": 11420 + }, + { + "epoch": 75.2, + "learning_rate": 2.3463110346311036e-05, + "loss": 0.9484, + "step": 11430 + }, + { + "epoch": 75.26, + "learning_rate": 2.3430845343084535e-05, + "loss": 1.012, + "step": 11440 + }, + { + "epoch": 75.33, + "learning_rate": 2.3398580339858033e-05, + "loss": 0.9146, + "step": 11450 + }, + { + "epoch": 75.39, + "learning_rate": 2.3366315336631535e-05, + "loss": 0.9199, + "step": 11460 + }, + { + "epoch": 75.46, + "learning_rate": 2.3334050333405033e-05, + "loss": 0.9703, + "step": 11470 + }, + { + "epoch": 75.52, + "learning_rate": 2.3301785330178532e-05, + "loss": 1.0307, + "step": 11480 + }, + { + "epoch": 75.59, + "learning_rate": 2.3269520326952034e-05, + "loss": 0.918, + "step": 11490 + }, + { + "epoch": 75.65, + "learning_rate": 2.3237255323725532e-05, + "loss": 0.9523, + "step": 11500 + }, + { + "epoch": 75.65, + "eval_loss": 0.34434974193573, + "eval_runtime": 319.3514, + "eval_samples_per_second": 22.264, + "eval_steps_per_second": 1.393, + "eval_wer": 0.3666118261879415, + "step": 11500 + }, + { + "epoch": 75.72, + "learning_rate": 2.320499032049903e-05, + "loss": 0.9908, + "step": 11510 + }, + { + "epoch": 75.79, + "learning_rate": 2.3172725317272532e-05, + "loss": 0.9971, + "step": 11520 + }, + { + "epoch": 75.85, + "learning_rate": 2.3140460314046034e-05, + "loss": 0.9447, + "step": 11530 + }, + { + "epoch": 75.92, + "learning_rate": 2.3108195310819533e-05, + "loss": 0.961, + "step": 11540 + }, + { + "epoch": 75.98, + "learning_rate": 2.307593030759303e-05, + "loss": 0.9799, + "step": 11550 + }, + { + "epoch": 76.05, + "learning_rate": 2.3043665304366533e-05, + "loss": 1.0397, + "step": 11560 + }, + { + "epoch": 76.12, + "learning_rate": 2.301140030114003e-05, + "loss": 0.9228, + "step": 11570 + }, + { + "epoch": 76.18, + "learning_rate": 2.297913529791353e-05, + "loss": 0.9437, + "step": 11580 + }, + { + "epoch": 76.25, + "learning_rate": 2.294687029468703e-05, + "loss": 1.0054, + "step": 11590 + }, + { + "epoch": 76.31, + "learning_rate": 2.291460529146053e-05, + "loss": 0.9331, + "step": 11600 + }, + { + "epoch": 76.38, + "learning_rate": 2.288234028823403e-05, + "loss": 0.9308, + "step": 11610 + }, + { + "epoch": 76.45, + "learning_rate": 2.285007528500753e-05, + "loss": 0.9729, + "step": 11620 + }, + { + "epoch": 76.51, + "learning_rate": 2.281781028178103e-05, + "loss": 1.0308, + "step": 11630 + }, + { + "epoch": 76.58, + "learning_rate": 2.2785545278554527e-05, + "loss": 0.9252, + "step": 11640 + }, + { + "epoch": 76.64, + "learning_rate": 2.275328027532803e-05, + "loss": 0.9327, + "step": 11650 + }, + { + "epoch": 76.71, + "learning_rate": 2.2721015272101527e-05, + "loss": 0.9551, + "step": 11660 + }, + { + "epoch": 76.77, + "learning_rate": 2.2688750268875026e-05, + "loss": 1.0287, + "step": 11670 + }, + { + "epoch": 76.84, + "learning_rate": 2.2656485265648527e-05, + "loss": 0.916, + "step": 11680 + }, + { + "epoch": 76.9, + "learning_rate": 2.262422026242203e-05, + "loss": 0.9475, + "step": 11690 + }, + { + "epoch": 76.97, + "learning_rate": 2.2591955259195524e-05, + "loss": 0.9593, + "step": 11700 + }, + { + "epoch": 77.04, + "learning_rate": 2.2559690255969026e-05, + "loss": 1.0743, + "step": 11710 + }, + { + "epoch": 77.1, + "learning_rate": 2.2527425252742528e-05, + "loss": 0.8937, + "step": 11720 + }, + { + "epoch": 77.17, + "learning_rate": 2.2495160249516026e-05, + "loss": 0.9349, + "step": 11730 + }, + { + "epoch": 77.24, + "learning_rate": 2.2462895246289525e-05, + "loss": 0.961, + "step": 11740 + }, + { + "epoch": 77.3, + "learning_rate": 2.2430630243063027e-05, + "loss": 0.9697, + "step": 11750 + }, + { + "epoch": 77.37, + "learning_rate": 2.2398365239836525e-05, + "loss": 0.9366, + "step": 11760 + }, + { + "epoch": 77.43, + "learning_rate": 2.2366100236610023e-05, + "loss": 0.9625, + "step": 11770 + }, + { + "epoch": 77.5, + "learning_rate": 2.2333835233383525e-05, + "loss": 1.0514, + "step": 11780 + }, + { + "epoch": 77.56, + "learning_rate": 2.2301570230157024e-05, + "loss": 0.9365, + "step": 11790 + }, + { + "epoch": 77.63, + "learning_rate": 2.2269305226930522e-05, + "loss": 0.9364, + "step": 11800 + }, + { + "epoch": 77.69, + "learning_rate": 2.2237040223704024e-05, + "loss": 0.9634, + "step": 11810 + }, + { + "epoch": 77.76, + "learning_rate": 2.2204775220477522e-05, + "loss": 0.9954, + "step": 11820 + }, + { + "epoch": 77.82, + "learning_rate": 2.217251021725102e-05, + "loss": 0.9266, + "step": 11830 + }, + { + "epoch": 77.89, + "learning_rate": 2.2140245214024523e-05, + "loss": 0.9328, + "step": 11840 + }, + { + "epoch": 77.96, + "learning_rate": 2.210798021079802e-05, + "loss": 0.9468, + "step": 11850 + }, + { + "epoch": 78.03, + "learning_rate": 2.207571520757152e-05, + "loss": 1.0674, + "step": 11860 + }, + { + "epoch": 78.09, + "learning_rate": 2.204345020434502e-05, + "loss": 0.9287, + "step": 11870 + }, + { + "epoch": 78.16, + "learning_rate": 2.2011185201118523e-05, + "loss": 0.9528, + "step": 11880 + }, + { + "epoch": 78.22, + "learning_rate": 2.1978920197892018e-05, + "loss": 0.9728, + "step": 11890 + }, + { + "epoch": 78.29, + "learning_rate": 2.194665519466552e-05, + "loss": 1.0022, + "step": 11900 + }, + { + "epoch": 78.35, + "learning_rate": 2.1914390191439022e-05, + "loss": 0.9094, + "step": 11910 + }, + { + "epoch": 78.42, + "learning_rate": 2.1882125188212517e-05, + "loss": 0.9332, + "step": 11920 + }, + { + "epoch": 78.48, + "learning_rate": 2.184986018498602e-05, + "loss": 0.9625, + "step": 11930 + }, + { + "epoch": 78.55, + "learning_rate": 2.181759518175952e-05, + "loss": 0.9592, + "step": 11940 + }, + { + "epoch": 78.62, + "learning_rate": 2.178533017853302e-05, + "loss": 0.9387, + "step": 11950 + }, + { + "epoch": 78.68, + "learning_rate": 2.1753065175306517e-05, + "loss": 0.9453, + "step": 11960 + }, + { + "epoch": 78.75, + "learning_rate": 2.172402667240267e-05, + "loss": 1.0156, + "step": 11970 + }, + { + "epoch": 78.81, + "learning_rate": 2.1691761669176165e-05, + "loss": 0.9209, + "step": 11980 + }, + { + "epoch": 78.88, + "learning_rate": 2.1659496665949667e-05, + "loss": 0.899, + "step": 11990 + }, + { + "epoch": 78.94, + "learning_rate": 2.162723166272317e-05, + "loss": 0.9523, + "step": 12000 + }, + { + "epoch": 78.94, + "eval_loss": 0.35024702548980713, + "eval_runtime": 330.0841, + "eval_samples_per_second": 21.54, + "eval_steps_per_second": 1.348, + "eval_wer": 0.375708893835285, + "step": 12000 + }, + { + "epoch": 79.01, + "learning_rate": 2.1594966659496664e-05, + "loss": 1.0633, + "step": 12010 + }, + { + "epoch": 79.08, + "learning_rate": 2.1562701656270166e-05, + "loss": 0.9186, + "step": 12020 + }, + { + "epoch": 79.14, + "learning_rate": 2.1530436653043668e-05, + "loss": 0.9239, + "step": 12030 + }, + { + "epoch": 79.21, + "learning_rate": 2.1498171649817166e-05, + "loss": 0.9601, + "step": 12040 + }, + { + "epoch": 79.27, + "learning_rate": 2.1465906646590664e-05, + "loss": 0.9821, + "step": 12050 + }, + { + "epoch": 79.34, + "learning_rate": 2.1433641643364166e-05, + "loss": 0.9105, + "step": 12060 + }, + { + "epoch": 79.41, + "learning_rate": 2.1401376640137665e-05, + "loss": 0.935, + "step": 12070 + }, + { + "epoch": 79.47, + "learning_rate": 2.1369111636911163e-05, + "loss": 0.9417, + "step": 12080 + }, + { + "epoch": 79.54, + "learning_rate": 2.1336846633684665e-05, + "loss": 1.0025, + "step": 12090 + }, + { + "epoch": 79.6, + "learning_rate": 2.1304581630458163e-05, + "loss": 0.8884, + "step": 12100 + }, + { + "epoch": 79.67, + "learning_rate": 2.1272316627231662e-05, + "loss": 0.9733, + "step": 12110 + }, + { + "epoch": 79.73, + "learning_rate": 2.1240051624005164e-05, + "loss": 0.9558, + "step": 12120 + }, + { + "epoch": 79.8, + "learning_rate": 2.1207786620778665e-05, + "loss": 0.9564, + "step": 12130 + }, + { + "epoch": 79.86, + "learning_rate": 2.117552161755216e-05, + "loss": 0.9139, + "step": 12140 + }, + { + "epoch": 79.93, + "learning_rate": 2.1143256614325662e-05, + "loss": 0.9615, + "step": 12150 + }, + { + "epoch": 80.0, + "learning_rate": 2.1110991611099164e-05, + "loss": 0.9671, + "step": 12160 + }, + { + "epoch": 80.07, + "learning_rate": 2.107872660787266e-05, + "loss": 1.0023, + "step": 12170 + }, + { + "epoch": 80.13, + "learning_rate": 2.104646160464616e-05, + "loss": 0.9213, + "step": 12180 + }, + { + "epoch": 80.2, + "learning_rate": 2.1014196601419663e-05, + "loss": 0.9354, + "step": 12190 + }, + { + "epoch": 80.26, + "learning_rate": 2.0981931598193158e-05, + "loss": 0.9967, + "step": 12200 + }, + { + "epoch": 80.33, + "learning_rate": 2.094966659496666e-05, + "loss": 0.8934, + "step": 12210 + }, + { + "epoch": 80.39, + "learning_rate": 2.091740159174016e-05, + "loss": 0.9723, + "step": 12220 + }, + { + "epoch": 80.46, + "learning_rate": 2.0885136588513656e-05, + "loss": 0.9434, + "step": 12230 + }, + { + "epoch": 80.52, + "learning_rate": 2.0852871585287158e-05, + "loss": 1.0069, + "step": 12240 + }, + { + "epoch": 80.59, + "learning_rate": 2.082060658206066e-05, + "loss": 0.9033, + "step": 12250 + }, + { + "epoch": 80.65, + "learning_rate": 2.078834157883416e-05, + "loss": 0.9583, + "step": 12260 + }, + { + "epoch": 80.72, + "learning_rate": 2.0756076575607657e-05, + "loss": 0.9529, + "step": 12270 + }, + { + "epoch": 80.79, + "learning_rate": 2.072381157238116e-05, + "loss": 1.0007, + "step": 12280 + }, + { + "epoch": 80.85, + "learning_rate": 2.069154656915466e-05, + "loss": 0.9083, + "step": 12290 + }, + { + "epoch": 80.92, + "learning_rate": 2.0659281565928155e-05, + "loss": 0.9298, + "step": 12300 + }, + { + "epoch": 80.98, + "learning_rate": 2.0627016562701657e-05, + "loss": 0.9816, + "step": 12310 + }, + { + "epoch": 81.05, + "learning_rate": 2.059475155947516e-05, + "loss": 1.0128, + "step": 12320 + }, + { + "epoch": 81.12, + "learning_rate": 2.0562486556248654e-05, + "loss": 0.8981, + "step": 12330 + }, + { + "epoch": 81.18, + "learning_rate": 2.0530221553022156e-05, + "loss": 0.9378, + "step": 12340 + }, + { + "epoch": 81.25, + "learning_rate": 2.0497956549795658e-05, + "loss": 0.9966, + "step": 12350 + }, + { + "epoch": 81.31, + "learning_rate": 2.0465691546569153e-05, + "loss": 0.9404, + "step": 12360 + }, + { + "epoch": 81.38, + "learning_rate": 2.0433426543342655e-05, + "loss": 0.929, + "step": 12370 + }, + { + "epoch": 81.45, + "learning_rate": 2.0401161540116156e-05, + "loss": 0.9378, + "step": 12380 + }, + { + "epoch": 81.51, + "learning_rate": 2.036889653688965e-05, + "loss": 1.0067, + "step": 12390 + }, + { + "epoch": 81.58, + "learning_rate": 2.0336631533663153e-05, + "loss": 0.9035, + "step": 12400 + }, + { + "epoch": 81.64, + "learning_rate": 2.0304366530436655e-05, + "loss": 0.9394, + "step": 12410 + }, + { + "epoch": 81.71, + "learning_rate": 2.027210152721015e-05, + "loss": 0.9561, + "step": 12420 + }, + { + "epoch": 81.77, + "learning_rate": 2.0239836523983652e-05, + "loss": 0.9793, + "step": 12430 + }, + { + "epoch": 81.84, + "learning_rate": 2.0207571520757154e-05, + "loss": 0.9155, + "step": 12440 + }, + { + "epoch": 81.9, + "learning_rate": 2.0175306517530656e-05, + "loss": 0.946, + "step": 12450 + }, + { + "epoch": 81.97, + "learning_rate": 2.014304151430415e-05, + "loss": 0.956, + "step": 12460 + }, + { + "epoch": 82.04, + "learning_rate": 2.0110776511077652e-05, + "loss": 1.0618, + "step": 12470 + }, + { + "epoch": 82.1, + "learning_rate": 2.0078511507851154e-05, + "loss": 0.9279, + "step": 12480 + }, + { + "epoch": 82.17, + "learning_rate": 2.004624650462465e-05, + "loss": 0.936, + "step": 12490 + }, + { + "epoch": 82.24, + "learning_rate": 2.001398150139815e-05, + "loss": 0.9475, + "step": 12500 + }, + { + "epoch": 82.24, + "eval_loss": 0.3509426712989807, + "eval_runtime": 320.8861, + "eval_samples_per_second": 22.157, + "eval_steps_per_second": 1.387, + "eval_wer": 0.3642504596922481, + "step": 12500 + }, + { + "epoch": 82.3, + "learning_rate": 1.9981716498171653e-05, + "loss": 0.9396, + "step": 12510 + }, + { + "epoch": 82.37, + "learning_rate": 1.9949451494945148e-05, + "loss": 0.8946, + "step": 12520 + }, + { + "epoch": 82.43, + "learning_rate": 1.991718649171865e-05, + "loss": 0.9433, + "step": 12530 + }, + { + "epoch": 82.5, + "learning_rate": 1.988492148849215e-05, + "loss": 0.9939, + "step": 12540 + }, + { + "epoch": 82.56, + "learning_rate": 1.9852656485265647e-05, + "loss": 0.9078, + "step": 12550 + }, + { + "epoch": 82.63, + "learning_rate": 1.982039148203915e-05, + "loss": 0.9229, + "step": 12560 + }, + { + "epoch": 82.69, + "learning_rate": 1.978812647881265e-05, + "loss": 0.9342, + "step": 12570 + }, + { + "epoch": 82.76, + "learning_rate": 1.9755861475586145e-05, + "loss": 0.9655, + "step": 12580 + }, + { + "epoch": 82.82, + "learning_rate": 1.9723596472359647e-05, + "loss": 0.9196, + "step": 12590 + }, + { + "epoch": 82.89, + "learning_rate": 1.969133146913315e-05, + "loss": 0.922, + "step": 12600 + }, + { + "epoch": 82.96, + "learning_rate": 1.9659066465906644e-05, + "loss": 0.9361, + "step": 12610 + }, + { + "epoch": 83.03, + "learning_rate": 1.9626801462680146e-05, + "loss": 1.0438, + "step": 12620 + }, + { + "epoch": 83.09, + "learning_rate": 1.9594536459453648e-05, + "loss": 0.9, + "step": 12630 + }, + { + "epoch": 83.16, + "learning_rate": 1.9562271456227146e-05, + "loss": 0.9131, + "step": 12640 + }, + { + "epoch": 83.22, + "learning_rate": 1.9530006453000644e-05, + "loss": 0.9195, + "step": 12650 + }, + { + "epoch": 83.29, + "learning_rate": 1.9497741449774146e-05, + "loss": 0.9638, + "step": 12660 + }, + { + "epoch": 83.35, + "learning_rate": 1.9465476446547648e-05, + "loss": 0.9233, + "step": 12670 + }, + { + "epoch": 83.42, + "learning_rate": 1.9433211443321143e-05, + "loss": 0.9477, + "step": 12680 + }, + { + "epoch": 83.48, + "learning_rate": 1.9400946440094645e-05, + "loss": 0.9753, + "step": 12690 + }, + { + "epoch": 83.55, + "learning_rate": 1.9368681436868147e-05, + "loss": 0.936, + "step": 12700 + }, + { + "epoch": 83.62, + "learning_rate": 1.9336416433641642e-05, + "loss": 0.9149, + "step": 12710 + }, + { + "epoch": 83.68, + "learning_rate": 1.9304151430415144e-05, + "loss": 0.9407, + "step": 12720 + }, + { + "epoch": 83.75, + "learning_rate": 1.9271886427188645e-05, + "loss": 0.9956, + "step": 12730 + }, + { + "epoch": 83.81, + "learning_rate": 1.923962142396214e-05, + "loss": 0.9145, + "step": 12740 + }, + { + "epoch": 83.88, + "learning_rate": 1.9207356420735642e-05, + "loss": 0.932, + "step": 12750 + }, + { + "epoch": 83.94, + "learning_rate": 1.9175091417509144e-05, + "loss": 0.9436, + "step": 12760 + }, + { + "epoch": 84.01, + "learning_rate": 1.914282641428264e-05, + "loss": 1.0773, + "step": 12770 + }, + { + "epoch": 84.08, + "learning_rate": 1.911056141105614e-05, + "loss": 0.8948, + "step": 12780 + }, + { + "epoch": 84.14, + "learning_rate": 1.9078296407829643e-05, + "loss": 0.9046, + "step": 12790 + }, + { + "epoch": 84.21, + "learning_rate": 1.904603140460314e-05, + "loss": 0.9352, + "step": 12800 + }, + { + "epoch": 84.27, + "learning_rate": 1.901376640137664e-05, + "loss": 1.0043, + "step": 12810 + }, + { + "epoch": 84.34, + "learning_rate": 1.898150139815014e-05, + "loss": 0.9053, + "step": 12820 + }, + { + "epoch": 84.41, + "learning_rate": 1.894923639492364e-05, + "loss": 0.8973, + "step": 12830 + }, + { + "epoch": 84.47, + "learning_rate": 1.8916971391697138e-05, + "loss": 0.9208, + "step": 12840 + }, + { + "epoch": 84.54, + "learning_rate": 1.888470638847064e-05, + "loss": 0.9667, + "step": 12850 + }, + { + "epoch": 84.6, + "learning_rate": 1.8852441385244142e-05, + "loss": 0.9048, + "step": 12860 + }, + { + "epoch": 84.67, + "learning_rate": 1.8820176382017637e-05, + "loss": 0.9502, + "step": 12870 + }, + { + "epoch": 84.73, + "learning_rate": 1.878791137879114e-05, + "loss": 0.9538, + "step": 12880 + }, + { + "epoch": 84.8, + "learning_rate": 1.875564637556464e-05, + "loss": 0.9497, + "step": 12890 + }, + { + "epoch": 84.86, + "learning_rate": 1.8723381372338135e-05, + "loss": 0.9217, + "step": 12900 + }, + { + "epoch": 84.93, + "learning_rate": 1.8691116369111637e-05, + "loss": 0.9125, + "step": 12910 + }, + { + "epoch": 85.0, + "learning_rate": 1.865885136588514e-05, + "loss": 1.0031, + "step": 12920 + }, + { + "epoch": 85.07, + "learning_rate": 1.8626586362658634e-05, + "loss": 0.9734, + "step": 12930 + }, + { + "epoch": 85.13, + "learning_rate": 1.8594321359432136e-05, + "loss": 0.9017, + "step": 12940 + }, + { + "epoch": 85.2, + "learning_rate": 1.8562056356205638e-05, + "loss": 0.9443, + "step": 12950 + }, + { + "epoch": 85.26, + "learning_rate": 1.8529791352979136e-05, + "loss": 0.9767, + "step": 12960 + }, + { + "epoch": 85.33, + "learning_rate": 1.8497526349752635e-05, + "loss": 0.8967, + "step": 12970 + }, + { + "epoch": 85.39, + "learning_rate": 1.8465261346526136e-05, + "loss": 0.9203, + "step": 12980 + }, + { + "epoch": 85.46, + "learning_rate": 1.8432996343299635e-05, + "loss": 0.9375, + "step": 12990 + }, + { + "epoch": 85.52, + "learning_rate": 1.8400731340073133e-05, + "loss": 0.9971, + "step": 13000 + }, + { + "epoch": 85.52, + "eval_loss": 0.3501547574996948, + "eval_runtime": 319.789, + "eval_samples_per_second": 22.233, + "eval_steps_per_second": 1.392, + "eval_wer": 0.36260524533049454, + "step": 13000 + }, + { + "epoch": 85.59, + "learning_rate": 1.8368466336846635e-05, + "loss": 0.9016, + "step": 13010 + }, + { + "epoch": 85.65, + "learning_rate": 1.8336201333620134e-05, + "loss": 0.9285, + "step": 13020 + }, + { + "epoch": 85.72, + "learning_rate": 1.8303936330393632e-05, + "loss": 0.9353, + "step": 13030 + }, + { + "epoch": 85.79, + "learning_rate": 1.8271671327167134e-05, + "loss": 0.9592, + "step": 13040 + }, + { + "epoch": 85.85, + "learning_rate": 1.8239406323940636e-05, + "loss": 0.8962, + "step": 13050 + }, + { + "epoch": 85.92, + "learning_rate": 1.820714132071413e-05, + "loss": 0.9172, + "step": 13060 + }, + { + "epoch": 85.98, + "learning_rate": 1.8174876317487632e-05, + "loss": 0.955, + "step": 13070 + }, + { + "epoch": 86.05, + "learning_rate": 1.8142611314261134e-05, + "loss": 1.0116, + "step": 13080 + }, + { + "epoch": 86.12, + "learning_rate": 1.811034631103463e-05, + "loss": 0.9055, + "step": 13090 + }, + { + "epoch": 86.18, + "learning_rate": 1.807808130780813e-05, + "loss": 0.9329, + "step": 13100 + }, + { + "epoch": 86.25, + "learning_rate": 1.8045816304581633e-05, + "loss": 0.9708, + "step": 13110 + }, + { + "epoch": 86.31, + "learning_rate": 1.8013551301355128e-05, + "loss": 0.9177, + "step": 13120 + }, + { + "epoch": 86.38, + "learning_rate": 1.798128629812863e-05, + "loss": 0.9202, + "step": 13130 + }, + { + "epoch": 86.45, + "learning_rate": 1.794902129490213e-05, + "loss": 0.9273, + "step": 13140 + }, + { + "epoch": 86.51, + "learning_rate": 1.791675629167563e-05, + "loss": 1.008, + "step": 13150 + }, + { + "epoch": 86.58, + "learning_rate": 1.788449128844913e-05, + "loss": 0.8852, + "step": 13160 + }, + { + "epoch": 86.64, + "learning_rate": 1.785222628522263e-05, + "loss": 0.9258, + "step": 13170 + }, + { + "epoch": 86.71, + "learning_rate": 1.781996128199613e-05, + "loss": 0.9329, + "step": 13180 + }, + { + "epoch": 86.77, + "learning_rate": 1.7787696278769627e-05, + "loss": 0.9703, + "step": 13190 + }, + { + "epoch": 86.84, + "learning_rate": 1.775543127554313e-05, + "loss": 0.8914, + "step": 13200 + }, + { + "epoch": 86.9, + "learning_rate": 1.7723166272316627e-05, + "loss": 0.9237, + "step": 13210 + }, + { + "epoch": 86.97, + "learning_rate": 1.7690901269090126e-05, + "loss": 0.9279, + "step": 13220 + }, + { + "epoch": 87.04, + "learning_rate": 1.7658636265863628e-05, + "loss": 1.0655, + "step": 13230 + }, + { + "epoch": 87.1, + "learning_rate": 1.7626371262637126e-05, + "loss": 0.8964, + "step": 13240 + }, + { + "epoch": 87.17, + "learning_rate": 1.7594106259410624e-05, + "loss": 0.9203, + "step": 13250 + }, + { + "epoch": 87.24, + "learning_rate": 1.7561841256184126e-05, + "loss": 0.9486, + "step": 13260 + }, + { + "epoch": 87.3, + "learning_rate": 1.7529576252957628e-05, + "loss": 0.9635, + "step": 13270 + }, + { + "epoch": 87.37, + "learning_rate": 1.7497311249731123e-05, + "loss": 0.8983, + "step": 13280 + }, + { + "epoch": 87.43, + "learning_rate": 1.7465046246504625e-05, + "loss": 0.9039, + "step": 13290 + }, + { + "epoch": 87.5, + "learning_rate": 1.7432781243278127e-05, + "loss": 0.9694, + "step": 13300 + }, + { + "epoch": 87.56, + "learning_rate": 1.7400516240051625e-05, + "loss": 0.9031, + "step": 13310 + }, + { + "epoch": 87.63, + "learning_rate": 1.7368251236825124e-05, + "loss": 0.901, + "step": 13320 + }, + { + "epoch": 87.69, + "learning_rate": 1.7335986233598625e-05, + "loss": 0.9082, + "step": 13330 + }, + { + "epoch": 87.76, + "learning_rate": 1.7303721230372124e-05, + "loss": 1.013, + "step": 13340 + }, + { + "epoch": 87.82, + "learning_rate": 1.7271456227145622e-05, + "loss": 0.8796, + "step": 13350 + }, + { + "epoch": 87.89, + "learning_rate": 1.7239191223919124e-05, + "loss": 0.9077, + "step": 13360 + }, + { + "epoch": 87.96, + "learning_rate": 1.7206926220692622e-05, + "loss": 0.9101, + "step": 13370 + }, + { + "epoch": 88.03, + "learning_rate": 1.717466121746612e-05, + "loss": 1.0345, + "step": 13380 + }, + { + "epoch": 88.09, + "learning_rate": 1.7142396214239623e-05, + "loss": 0.8971, + "step": 13390 + }, + { + "epoch": 88.16, + "learning_rate": 1.711013121101312e-05, + "loss": 0.913, + "step": 13400 + }, + { + "epoch": 88.22, + "learning_rate": 1.707786620778662e-05, + "loss": 0.9057, + "step": 13410 + }, + { + "epoch": 88.29, + "learning_rate": 1.704560120456012e-05, + "loss": 0.9604, + "step": 13420 + }, + { + "epoch": 88.35, + "learning_rate": 1.701333620133362e-05, + "loss": 0.8986, + "step": 13430 + }, + { + "epoch": 88.42, + "learning_rate": 1.6981071198107118e-05, + "loss": 0.9097, + "step": 13440 + }, + { + "epoch": 88.48, + "learning_rate": 1.694880619488062e-05, + "loss": 0.9543, + "step": 13450 + }, + { + "epoch": 88.55, + "learning_rate": 1.6916541191654122e-05, + "loss": 0.9289, + "step": 13460 + }, + { + "epoch": 88.62, + "learning_rate": 1.688427618842762e-05, + "loss": 0.9003, + "step": 13470 + }, + { + "epoch": 88.68, + "learning_rate": 1.685201118520112e-05, + "loss": 0.9253, + "step": 13480 + }, + { + "epoch": 88.75, + "learning_rate": 1.681974618197462e-05, + "loss": 0.9481, + "step": 13490 + }, + { + "epoch": 88.81, + "learning_rate": 1.678748117874812e-05, + "loss": 0.9058, + "step": 13500 + }, + { + "epoch": 88.81, + "eval_loss": 0.34722036123275757, + "eval_runtime": 317.2905, + "eval_samples_per_second": 22.408, + "eval_steps_per_second": 1.403, + "eval_wer": 0.3604567889286751, + "step": 13500 + }, + { + "epoch": 88.88, + "learning_rate": 1.6755216175521617e-05, + "loss": 0.9168, + "step": 13510 + }, + { + "epoch": 88.94, + "learning_rate": 1.672295117229512e-05, + "loss": 0.9548, + "step": 13520 + }, + { + "epoch": 89.01, + "learning_rate": 1.6690686169068618e-05, + "loss": 1.082, + "step": 13530 + }, + { + "epoch": 89.08, + "learning_rate": 1.6658421165842116e-05, + "loss": 0.8845, + "step": 13540 + }, + { + "epoch": 89.14, + "learning_rate": 1.6626156162615618e-05, + "loss": 0.8901, + "step": 13550 + }, + { + "epoch": 89.21, + "learning_rate": 1.6593891159389116e-05, + "loss": 0.9123, + "step": 13560 + }, + { + "epoch": 89.27, + "learning_rate": 1.6561626156162615e-05, + "loss": 0.9849, + "step": 13570 + }, + { + "epoch": 89.34, + "learning_rate": 1.6529361152936116e-05, + "loss": 0.88, + "step": 13580 + }, + { + "epoch": 89.41, + "learning_rate": 1.6497096149709615e-05, + "loss": 0.9248, + "step": 13590 + }, + { + "epoch": 89.47, + "learning_rate": 1.6464831146483113e-05, + "loss": 0.9346, + "step": 13600 + }, + { + "epoch": 89.54, + "learning_rate": 1.6432566143256615e-05, + "loss": 0.9791, + "step": 13610 + }, + { + "epoch": 89.6, + "learning_rate": 1.6400301140030114e-05, + "loss": 0.8904, + "step": 13620 + }, + { + "epoch": 89.67, + "learning_rate": 1.6368036136803612e-05, + "loss": 0.8909, + "step": 13630 + }, + { + "epoch": 89.73, + "learning_rate": 1.6335771133577114e-05, + "loss": 0.9394, + "step": 13640 + }, + { + "epoch": 89.8, + "learning_rate": 1.6303506130350612e-05, + "loss": 0.9302, + "step": 13650 + }, + { + "epoch": 89.86, + "learning_rate": 1.6271241127124114e-05, + "loss": 0.9093, + "step": 13660 + }, + { + "epoch": 89.93, + "learning_rate": 1.6238976123897612e-05, + "loss": 0.9307, + "step": 13670 + }, + { + "epoch": 90.0, + "learning_rate": 1.6206711120671114e-05, + "loss": 0.9548, + "step": 13680 + }, + { + "epoch": 90.07, + "learning_rate": 1.6174446117444613e-05, + "loss": 0.9712, + "step": 13690 + }, + { + "epoch": 90.13, + "learning_rate": 1.614218111421811e-05, + "loss": 0.8957, + "step": 13700 + }, + { + "epoch": 90.2, + "learning_rate": 1.6109916110991613e-05, + "loss": 0.9104, + "step": 13710 + }, + { + "epoch": 90.26, + "learning_rate": 1.607765110776511e-05, + "loss": 0.9518, + "step": 13720 + }, + { + "epoch": 90.33, + "learning_rate": 1.604538610453861e-05, + "loss": 0.8965, + "step": 13730 + }, + { + "epoch": 90.39, + "learning_rate": 1.601312110131211e-05, + "loss": 0.9143, + "step": 13740 + }, + { + "epoch": 90.46, + "learning_rate": 1.598085609808561e-05, + "loss": 0.9043, + "step": 13750 + }, + { + "epoch": 90.52, + "learning_rate": 1.594859109485911e-05, + "loss": 0.9751, + "step": 13760 + }, + { + "epoch": 90.59, + "learning_rate": 1.591632609163261e-05, + "loss": 0.8719, + "step": 13770 + }, + { + "epoch": 90.65, + "learning_rate": 1.588406108840611e-05, + "loss": 0.9339, + "step": 13780 + }, + { + "epoch": 90.72, + "learning_rate": 1.5851796085179607e-05, + "loss": 0.9233, + "step": 13790 + }, + { + "epoch": 90.79, + "learning_rate": 1.581953108195311e-05, + "loss": 0.9587, + "step": 13800 + }, + { + "epoch": 90.85, + "learning_rate": 1.5787266078726607e-05, + "loss": 0.9144, + "step": 13810 + }, + { + "epoch": 90.92, + "learning_rate": 1.575500107550011e-05, + "loss": 0.9202, + "step": 13820 + }, + { + "epoch": 90.98, + "learning_rate": 1.5722736072273608e-05, + "loss": 0.9147, + "step": 13830 + }, + { + "epoch": 91.05, + "learning_rate": 1.5690471069047106e-05, + "loss": 1.0151, + "step": 13840 + }, + { + "epoch": 91.12, + "learning_rate": 1.5658206065820608e-05, + "loss": 0.893, + "step": 13850 + }, + { + "epoch": 91.18, + "learning_rate": 1.5625941062594106e-05, + "loss": 0.9064, + "step": 13860 + }, + { + "epoch": 91.25, + "learning_rate": 1.5593676059367608e-05, + "loss": 0.956, + "step": 13870 + }, + { + "epoch": 91.31, + "learning_rate": 1.5561411056141106e-05, + "loss": 0.8682, + "step": 13880 + }, + { + "epoch": 91.38, + "learning_rate": 1.5529146052914605e-05, + "loss": 0.8912, + "step": 13890 + }, + { + "epoch": 91.45, + "learning_rate": 1.5496881049688107e-05, + "loss": 0.9084, + "step": 13900 + }, + { + "epoch": 91.51, + "learning_rate": 1.5464616046461605e-05, + "loss": 0.9605, + "step": 13910 + }, + { + "epoch": 91.58, + "learning_rate": 1.5432351043235104e-05, + "loss": 0.8805, + "step": 13920 + }, + { + "epoch": 91.64, + "learning_rate": 1.5400086040008605e-05, + "loss": 0.9132, + "step": 13930 + }, + { + "epoch": 91.71, + "learning_rate": 1.5367821036782104e-05, + "loss": 0.9581, + "step": 13940 + }, + { + "epoch": 91.77, + "learning_rate": 1.5335556033555602e-05, + "loss": 0.9395, + "step": 13950 + }, + { + "epoch": 91.84, + "learning_rate": 1.5303291030329104e-05, + "loss": 0.8723, + "step": 13960 + }, + { + "epoch": 91.9, + "learning_rate": 1.5271026027102602e-05, + "loss": 0.9009, + "step": 13970 + }, + { + "epoch": 91.97, + "learning_rate": 1.5238761023876104e-05, + "loss": 0.9588, + "step": 13980 + }, + { + "epoch": 92.04, + "learning_rate": 1.5206496020649601e-05, + "loss": 1.0354, + "step": 13990 + }, + { + "epoch": 92.1, + "learning_rate": 1.5174231017423103e-05, + "loss": 0.8922, + "step": 14000 + }, + { + "epoch": 92.1, + "eval_loss": 0.3530175983905792, + "eval_runtime": 325.2566, + "eval_samples_per_second": 21.86, + "eval_steps_per_second": 1.368, + "eval_wer": 0.3618116713442369, + "step": 14000 + }, + { + "epoch": 92.17, + "learning_rate": 1.5141966014196603e-05, + "loss": 0.9222, + "step": 14010 + }, + { + "epoch": 92.24, + "learning_rate": 1.5109701010970101e-05, + "loss": 0.9484, + "step": 14020 + }, + { + "epoch": 92.3, + "learning_rate": 1.5077436007743601e-05, + "loss": 0.9421, + "step": 14030 + }, + { + "epoch": 92.37, + "learning_rate": 1.5045171004517102e-05, + "loss": 0.8909, + "step": 14040 + }, + { + "epoch": 92.43, + "learning_rate": 1.50129060012906e-05, + "loss": 0.9404, + "step": 14050 + }, + { + "epoch": 92.5, + "learning_rate": 1.49806409980641e-05, + "loss": 0.9458, + "step": 14060 + }, + { + "epoch": 92.56, + "learning_rate": 1.4948375994837599e-05, + "loss": 0.9147, + "step": 14070 + }, + { + "epoch": 92.63, + "learning_rate": 1.49161109916111e-05, + "loss": 0.9221, + "step": 14080 + }, + { + "epoch": 92.69, + "learning_rate": 1.4883845988384599e-05, + "loss": 0.8916, + "step": 14090 + }, + { + "epoch": 92.76, + "learning_rate": 1.4851580985158099e-05, + "loss": 0.9474, + "step": 14100 + }, + { + "epoch": 92.82, + "learning_rate": 1.4819315981931599e-05, + "loss": 0.8877, + "step": 14110 + }, + { + "epoch": 92.89, + "learning_rate": 1.4787050978705097e-05, + "loss": 0.8793, + "step": 14120 + }, + { + "epoch": 92.96, + "learning_rate": 1.47547859754786e-05, + "loss": 0.9027, + "step": 14130 + }, + { + "epoch": 93.03, + "learning_rate": 1.4722520972252098e-05, + "loss": 1.0347, + "step": 14140 + }, + { + "epoch": 93.09, + "learning_rate": 1.4690255969025596e-05, + "loss": 0.8792, + "step": 14150 + }, + { + "epoch": 93.16, + "learning_rate": 1.4657990965799098e-05, + "loss": 0.9272, + "step": 14160 + }, + { + "epoch": 93.22, + "learning_rate": 1.4625725962572596e-05, + "loss": 0.9519, + "step": 14170 + }, + { + "epoch": 93.29, + "learning_rate": 1.4593460959346095e-05, + "loss": 0.9434, + "step": 14180 + }, + { + "epoch": 93.35, + "learning_rate": 1.4561195956119597e-05, + "loss": 0.8696, + "step": 14190 + }, + { + "epoch": 93.42, + "learning_rate": 1.4528930952893095e-05, + "loss": 0.9093, + "step": 14200 + }, + { + "epoch": 93.48, + "learning_rate": 1.4496665949666597e-05, + "loss": 0.9385, + "step": 14210 + }, + { + "epoch": 93.55, + "learning_rate": 1.4464400946440095e-05, + "loss": 0.9052, + "step": 14220 + }, + { + "epoch": 93.62, + "learning_rate": 1.4432135943213594e-05, + "loss": 0.8826, + "step": 14230 + }, + { + "epoch": 93.68, + "learning_rate": 1.4399870939987095e-05, + "loss": 0.9008, + "step": 14240 + }, + { + "epoch": 93.75, + "learning_rate": 1.4367605936760594e-05, + "loss": 0.9739, + "step": 14250 + }, + { + "epoch": 93.81, + "learning_rate": 1.4335340933534092e-05, + "loss": 0.8942, + "step": 14260 + }, + { + "epoch": 93.88, + "learning_rate": 1.4303075930307594e-05, + "loss": 0.9117, + "step": 14270 + }, + { + "epoch": 93.94, + "learning_rate": 1.4270810927081093e-05, + "loss": 0.9197, + "step": 14280 + }, + { + "epoch": 94.01, + "learning_rate": 1.4238545923854593e-05, + "loss": 1.0093, + "step": 14290 + }, + { + "epoch": 94.08, + "learning_rate": 1.4206280920628093e-05, + "loss": 0.8751, + "step": 14300 + }, + { + "epoch": 94.14, + "learning_rate": 1.4174015917401591e-05, + "loss": 0.8825, + "step": 14310 + }, + { + "epoch": 94.21, + "learning_rate": 1.4141750914175093e-05, + "loss": 0.902, + "step": 14320 + }, + { + "epoch": 94.27, + "learning_rate": 1.4109485910948591e-05, + "loss": 0.9608, + "step": 14330 + }, + { + "epoch": 94.34, + "learning_rate": 1.407722090772209e-05, + "loss": 0.8833, + "step": 14340 + }, + { + "epoch": 94.41, + "learning_rate": 1.4044955904495592e-05, + "loss": 0.8899, + "step": 14350 + }, + { + "epoch": 94.47, + "learning_rate": 1.401269090126909e-05, + "loss": 0.8863, + "step": 14360 + }, + { + "epoch": 94.54, + "learning_rate": 1.398042589804259e-05, + "loss": 0.9543, + "step": 14370 + }, + { + "epoch": 94.6, + "learning_rate": 1.394816089481609e-05, + "loss": 0.8773, + "step": 14380 + }, + { + "epoch": 94.67, + "learning_rate": 1.3915895891589589e-05, + "loss": 0.8904, + "step": 14390 + }, + { + "epoch": 94.73, + "learning_rate": 1.3883630888363089e-05, + "loss": 0.9539, + "step": 14400 + }, + { + "epoch": 94.8, + "learning_rate": 1.3851365885136589e-05, + "loss": 0.9223, + "step": 14410 + }, + { + "epoch": 94.86, + "learning_rate": 1.3819100881910087e-05, + "loss": 0.8494, + "step": 14420 + }, + { + "epoch": 94.93, + "learning_rate": 1.378683587868359e-05, + "loss": 0.8945, + "step": 14430 + }, + { + "epoch": 95.0, + "learning_rate": 1.3754570875457088e-05, + "loss": 0.973, + "step": 14440 + }, + { + "epoch": 95.07, + "learning_rate": 1.3722305872230588e-05, + "loss": 0.9566, + "step": 14450 + }, + { + "epoch": 95.13, + "learning_rate": 1.3690040869004088e-05, + "loss": 0.9169, + "step": 14460 + }, + { + "epoch": 95.2, + "learning_rate": 1.3657775865777586e-05, + "loss": 0.8868, + "step": 14470 + }, + { + "epoch": 95.26, + "learning_rate": 1.3625510862551086e-05, + "loss": 0.9633, + "step": 14480 + }, + { + "epoch": 95.33, + "learning_rate": 1.3593245859324587e-05, + "loss": 0.8784, + "step": 14490 + }, + { + "epoch": 95.39, + "learning_rate": 1.3560980856098085e-05, + "loss": 0.9, + "step": 14500 + }, + { + "epoch": 95.39, + "eval_loss": 0.34995371103286743, + "eval_runtime": 324.3004, + "eval_samples_per_second": 21.924, + "eval_steps_per_second": 1.372, + "eval_wer": 0.35739862576212134, + "step": 14500 + }, + { + "epoch": 95.46, + "learning_rate": 1.3528715852871587e-05, + "loss": 0.9304, + "step": 14510 + }, + { + "epoch": 95.52, + "learning_rate": 1.3496450849645085e-05, + "loss": 0.945, + "step": 14520 + }, + { + "epoch": 95.59, + "learning_rate": 1.3464185846418585e-05, + "loss": 0.8791, + "step": 14530 + }, + { + "epoch": 95.65, + "learning_rate": 1.3431920843192085e-05, + "loss": 0.8943, + "step": 14540 + }, + { + "epoch": 95.72, + "learning_rate": 1.3399655839965584e-05, + "loss": 0.9172, + "step": 14550 + }, + { + "epoch": 95.79, + "learning_rate": 1.3367390836739084e-05, + "loss": 0.9405, + "step": 14560 + }, + { + "epoch": 95.85, + "learning_rate": 1.3335125833512584e-05, + "loss": 0.8616, + "step": 14570 + }, + { + "epoch": 95.92, + "learning_rate": 1.3302860830286083e-05, + "loss": 0.9063, + "step": 14580 + }, + { + "epoch": 95.98, + "learning_rate": 1.3270595827059583e-05, + "loss": 0.9359, + "step": 14590 + }, + { + "epoch": 96.05, + "learning_rate": 1.3238330823833083e-05, + "loss": 0.9663, + "step": 14600 + }, + { + "epoch": 96.12, + "learning_rate": 1.3206065820606583e-05, + "loss": 0.872, + "step": 14610 + }, + { + "epoch": 96.18, + "learning_rate": 1.3173800817380083e-05, + "loss": 0.8937, + "step": 14620 + }, + { + "epoch": 96.25, + "learning_rate": 1.3141535814153581e-05, + "loss": 0.9451, + "step": 14630 + }, + { + "epoch": 96.31, + "learning_rate": 1.3109270810927082e-05, + "loss": 0.8743, + "step": 14640 + }, + { + "epoch": 96.38, + "learning_rate": 1.3077005807700582e-05, + "loss": 0.8862, + "step": 14650 + }, + { + "epoch": 96.45, + "learning_rate": 1.304474080447408e-05, + "loss": 0.9016, + "step": 14660 + }, + { + "epoch": 96.51, + "learning_rate": 1.301247580124758e-05, + "loss": 0.9827, + "step": 14670 + }, + { + "epoch": 96.58, + "learning_rate": 1.298021079802108e-05, + "loss": 0.8827, + "step": 14680 + }, + { + "epoch": 96.64, + "learning_rate": 1.294794579479458e-05, + "loss": 0.9001, + "step": 14690 + }, + { + "epoch": 96.71, + "learning_rate": 1.2915680791568079e-05, + "loss": 0.8752, + "step": 14700 + }, + { + "epoch": 96.77, + "learning_rate": 1.2883415788341579e-05, + "loss": 0.9638, + "step": 14710 + }, + { + "epoch": 96.84, + "learning_rate": 1.2851150785115079e-05, + "loss": 0.871, + "step": 14720 + }, + { + "epoch": 96.9, + "learning_rate": 1.281888578188858e-05, + "loss": 0.9118, + "step": 14730 + }, + { + "epoch": 96.97, + "learning_rate": 1.2786620778662078e-05, + "loss": 0.9135, + "step": 14740 + }, + { + "epoch": 97.04, + "learning_rate": 1.2754355775435578e-05, + "loss": 0.991, + "step": 14750 + }, + { + "epoch": 97.1, + "learning_rate": 1.2722090772209078e-05, + "loss": 0.8699, + "step": 14760 + }, + { + "epoch": 97.17, + "learning_rate": 1.2689825768982576e-05, + "loss": 0.8852, + "step": 14770 + }, + { + "epoch": 97.24, + "learning_rate": 1.2657560765756076e-05, + "loss": 0.9257, + "step": 14780 + }, + { + "epoch": 97.3, + "learning_rate": 1.2625295762529577e-05, + "loss": 0.9189, + "step": 14790 + }, + { + "epoch": 97.37, + "learning_rate": 1.2593030759303077e-05, + "loss": 0.8538, + "step": 14800 + }, + { + "epoch": 97.43, + "learning_rate": 1.2560765756076575e-05, + "loss": 0.9018, + "step": 14810 + }, + { + "epoch": 97.5, + "learning_rate": 1.2528500752850075e-05, + "loss": 0.9465, + "step": 14820 + }, + { + "epoch": 97.56, + "learning_rate": 1.2496235749623575e-05, + "loss": 0.9016, + "step": 14830 + }, + { + "epoch": 97.63, + "learning_rate": 1.2463970746397075e-05, + "loss": 0.888, + "step": 14840 + }, + { + "epoch": 97.69, + "learning_rate": 1.2431705743170574e-05, + "loss": 0.9304, + "step": 14850 + }, + { + "epoch": 97.76, + "learning_rate": 1.2399440739944074e-05, + "loss": 0.9315, + "step": 14860 + }, + { + "epoch": 97.82, + "learning_rate": 1.2367175736717574e-05, + "loss": 0.8613, + "step": 14870 + }, + { + "epoch": 97.89, + "learning_rate": 1.2334910733491074e-05, + "loss": 0.9004, + "step": 14880 + }, + { + "epoch": 97.96, + "learning_rate": 1.2302645730264573e-05, + "loss": 0.9156, + "step": 14890 + }, + { + "epoch": 98.03, + "learning_rate": 1.2270380727038073e-05, + "loss": 1.0229, + "step": 14900 + }, + { + "epoch": 98.09, + "learning_rate": 1.2238115723811573e-05, + "loss": 0.8743, + "step": 14910 + }, + { + "epoch": 98.16, + "learning_rate": 1.2205850720585073e-05, + "loss": 0.8969, + "step": 14920 + }, + { + "epoch": 98.22, + "learning_rate": 1.2173585717358571e-05, + "loss": 0.9141, + "step": 14930 + }, + { + "epoch": 98.29, + "learning_rate": 1.2141320714132072e-05, + "loss": 0.9501, + "step": 14940 + }, + { + "epoch": 98.35, + "learning_rate": 1.2109055710905572e-05, + "loss": 0.8673, + "step": 14950 + }, + { + "epoch": 98.42, + "learning_rate": 1.2076790707679072e-05, + "loss": 0.881, + "step": 14960 + }, + { + "epoch": 98.48, + "learning_rate": 1.204452570445257e-05, + "loss": 0.9423, + "step": 14970 + }, + { + "epoch": 98.55, + "learning_rate": 1.201226070122607e-05, + "loss": 0.9134, + "step": 14980 + }, + { + "epoch": 98.62, + "learning_rate": 1.197999569799957e-05, + "loss": 0.869, + "step": 14990 + }, + { + "epoch": 98.68, + "learning_rate": 1.1947730694773069e-05, + "loss": 0.9051, + "step": 15000 + }, + { + "epoch": 98.68, + "eval_loss": 0.3455681800842285, + "eval_runtime": 323.8015, + "eval_samples_per_second": 21.958, + "eval_steps_per_second": 1.374, + "eval_wer": 0.35352753314623053, + "step": 15000 + }, + { + "epoch": 98.75, + "learning_rate": 1.1915465691546569e-05, + "loss": 0.968, + "step": 15010 + }, + { + "epoch": 98.81, + "learning_rate": 1.1883200688320069e-05, + "loss": 0.8848, + "step": 15020 + }, + { + "epoch": 98.88, + "learning_rate": 1.185093568509357e-05, + "loss": 0.8963, + "step": 15030 + }, + { + "epoch": 98.94, + "learning_rate": 1.181867068186707e-05, + "loss": 0.9089, + "step": 15040 + }, + { + "epoch": 99.01, + "learning_rate": 1.1786405678640568e-05, + "loss": 1.0464, + "step": 15050 + }, + { + "epoch": 99.08, + "learning_rate": 1.1754140675414068e-05, + "loss": 0.8532, + "step": 15060 + }, + { + "epoch": 99.14, + "learning_rate": 1.1721875672187568e-05, + "loss": 0.8836, + "step": 15070 + }, + { + "epoch": 99.21, + "learning_rate": 1.1689610668961066e-05, + "loss": 0.9005, + "step": 15080 + }, + { + "epoch": 99.27, + "learning_rate": 1.1657345665734567e-05, + "loss": 0.9232, + "step": 15090 + }, + { + "epoch": 99.34, + "learning_rate": 1.1625080662508067e-05, + "loss": 0.8787, + "step": 15100 + }, + { + "epoch": 99.41, + "learning_rate": 1.1592815659281565e-05, + "loss": 0.871, + "step": 15110 + }, + { + "epoch": 99.47, + "learning_rate": 1.1560550656055067e-05, + "loss": 0.9109, + "step": 15120 + }, + { + "epoch": 99.54, + "learning_rate": 1.1528285652828565e-05, + "loss": 0.9561, + "step": 15130 + }, + { + "epoch": 99.6, + "learning_rate": 1.1496020649602065e-05, + "loss": 0.8613, + "step": 15140 + }, + { + "epoch": 99.67, + "learning_rate": 1.1463755646375566e-05, + "loss": 0.8852, + "step": 15150 + }, + { + "epoch": 99.73, + "learning_rate": 1.1431490643149064e-05, + "loss": 0.9084, + "step": 15160 + }, + { + "epoch": 99.8, + "learning_rate": 1.1399225639922564e-05, + "loss": 0.905, + "step": 15170 + }, + { + "epoch": 99.86, + "learning_rate": 1.1366960636696064e-05, + "loss": 0.8638, + "step": 15180 + }, + { + "epoch": 99.93, + "learning_rate": 1.1334695633469563e-05, + "loss": 0.914, + "step": 15190 + }, + { + "epoch": 100.0, + "learning_rate": 1.1302430630243064e-05, + "loss": 0.9435, + "step": 15200 + }, + { + "epoch": 100.07, + "learning_rate": 1.1270165627016563e-05, + "loss": 0.9602, + "step": 15210 + }, + { + "epoch": 100.13, + "learning_rate": 1.1237900623790063e-05, + "loss": 0.8831, + "step": 15220 + }, + { + "epoch": 100.2, + "learning_rate": 1.1205635620563563e-05, + "loss": 0.8994, + "step": 15230 + }, + { + "epoch": 100.26, + "learning_rate": 1.1173370617337062e-05, + "loss": 0.9412, + "step": 15240 + }, + { + "epoch": 100.33, + "learning_rate": 1.1141105614110562e-05, + "loss": 0.842, + "step": 15250 + }, + { + "epoch": 100.39, + "learning_rate": 1.1108840610884062e-05, + "loss": 0.8825, + "step": 15260 + }, + { + "epoch": 100.46, + "learning_rate": 1.107657560765756e-05, + "loss": 0.8777, + "step": 15270 + }, + { + "epoch": 100.52, + "learning_rate": 1.1044310604431062e-05, + "loss": 0.9486, + "step": 15280 + }, + { + "epoch": 100.59, + "learning_rate": 1.101204560120456e-05, + "loss": 0.8656, + "step": 15290 + }, + { + "epoch": 100.65, + "learning_rate": 1.0979780597978059e-05, + "loss": 0.9077, + "step": 15300 + }, + { + "epoch": 100.72, + "learning_rate": 1.094751559475156e-05, + "loss": 0.9023, + "step": 15310 + }, + { + "epoch": 100.79, + "learning_rate": 1.0915250591525059e-05, + "loss": 0.9121, + "step": 15320 + }, + { + "epoch": 100.85, + "learning_rate": 1.088298558829856e-05, + "loss": 0.8787, + "step": 15330 + }, + { + "epoch": 100.92, + "learning_rate": 1.085072058507206e-05, + "loss": 0.9054, + "step": 15340 + }, + { + "epoch": 100.98, + "learning_rate": 1.0818455581845558e-05, + "loss": 0.9459, + "step": 15350 + }, + { + "epoch": 101.05, + "learning_rate": 1.0786190578619058e-05, + "loss": 0.9726, + "step": 15360 + }, + { + "epoch": 101.12, + "learning_rate": 1.0753925575392558e-05, + "loss": 0.8765, + "step": 15370 + }, + { + "epoch": 101.18, + "learning_rate": 1.0721660572166056e-05, + "loss": 0.8734, + "step": 15380 + }, + { + "epoch": 101.25, + "learning_rate": 1.0689395568939558e-05, + "loss": 0.9516, + "step": 15390 + }, + { + "epoch": 101.31, + "learning_rate": 1.0657130565713057e-05, + "loss": 0.8683, + "step": 15400 + }, + { + "epoch": 101.38, + "learning_rate": 1.0624865562486555e-05, + "loss": 0.8891, + "step": 15410 + }, + { + "epoch": 101.45, + "learning_rate": 1.0592600559260057e-05, + "loss": 0.8698, + "step": 15420 + }, + { + "epoch": 101.51, + "learning_rate": 1.0560335556033555e-05, + "loss": 0.9669, + "step": 15430 + }, + { + "epoch": 101.58, + "learning_rate": 1.0528070552807055e-05, + "loss": 0.8661, + "step": 15440 + }, + { + "epoch": 101.64, + "learning_rate": 1.0495805549580556e-05, + "loss": 0.8912, + "step": 15450 + }, + { + "epoch": 101.71, + "learning_rate": 1.0463540546354054e-05, + "loss": 0.8818, + "step": 15460 + }, + { + "epoch": 101.77, + "learning_rate": 1.0431275543127556e-05, + "loss": 0.9431, + "step": 15470 + }, + { + "epoch": 101.84, + "learning_rate": 1.0399010539901054e-05, + "loss": 0.8664, + "step": 15480 + }, + { + "epoch": 101.9, + "learning_rate": 1.0366745536674553e-05, + "loss": 0.9112, + "step": 15490 + }, + { + "epoch": 101.97, + "learning_rate": 1.0334480533448055e-05, + "loss": 0.9304, + "step": 15500 + }, + { + "epoch": 101.97, + "eval_loss": 0.34380748867988586, + "eval_runtime": 321.8146, + "eval_samples_per_second": 22.093, + "eval_steps_per_second": 1.383, + "eval_wer": 0.357766379560631, + "step": 15500 + }, + { + "epoch": 102.04, + "learning_rate": 1.0302215530221553e-05, + "loss": 1.0402, + "step": 15510 + }, + { + "epoch": 102.1, + "learning_rate": 1.0269950526995053e-05, + "loss": 0.8567, + "step": 15520 + }, + { + "epoch": 102.17, + "learning_rate": 1.0237685523768553e-05, + "loss": 0.9101, + "step": 15530 + }, + { + "epoch": 102.24, + "learning_rate": 1.0205420520542052e-05, + "loss": 0.9497, + "step": 15540 + }, + { + "epoch": 102.3, + "learning_rate": 1.0173155517315553e-05, + "loss": 0.8929, + "step": 15550 + }, + { + "epoch": 102.37, + "learning_rate": 1.0140890514089052e-05, + "loss": 0.8703, + "step": 15560 + }, + { + "epoch": 102.43, + "learning_rate": 1.010862551086255e-05, + "loss": 0.8839, + "step": 15570 + }, + { + "epoch": 102.5, + "learning_rate": 1.0076360507636052e-05, + "loss": 0.9388, + "step": 15580 + }, + { + "epoch": 102.56, + "learning_rate": 1.004409550440955e-05, + "loss": 0.8551, + "step": 15590 + }, + { + "epoch": 102.63, + "learning_rate": 1.0011830501183049e-05, + "loss": 0.8776, + "step": 15600 + }, + { + "epoch": 102.69, + "learning_rate": 9.97956549795655e-06, + "loss": 0.8986, + "step": 15610 + }, + { + "epoch": 102.76, + "learning_rate": 9.947300494730049e-06, + "loss": 0.9499, + "step": 15620 + }, + { + "epoch": 102.82, + "learning_rate": 9.915035491503551e-06, + "loss": 0.875, + "step": 15630 + }, + { + "epoch": 102.89, + "learning_rate": 9.88277048827705e-06, + "loss": 0.8834, + "step": 15640 + }, + { + "epoch": 102.96, + "learning_rate": 9.853731985373199e-06, + "loss": 0.9063, + "step": 15650 + }, + { + "epoch": 103.03, + "learning_rate": 9.821466982146697e-06, + "loss": 1.0078, + "step": 15660 + }, + { + "epoch": 103.09, + "learning_rate": 9.7892019789202e-06, + "loss": 0.8404, + "step": 15670 + }, + { + "epoch": 103.16, + "learning_rate": 9.756936975693698e-06, + "loss": 0.877, + "step": 15680 + }, + { + "epoch": 103.22, + "learning_rate": 9.724671972467196e-06, + "loss": 0.8756, + "step": 15690 + }, + { + "epoch": 103.29, + "learning_rate": 9.692406969240698e-06, + "loss": 0.9206, + "step": 15700 + }, + { + "epoch": 103.35, + "learning_rate": 9.660141966014196e-06, + "loss": 0.8646, + "step": 15710 + }, + { + "epoch": 103.42, + "learning_rate": 9.627876962787696e-06, + "loss": 0.876, + "step": 15720 + }, + { + "epoch": 103.48, + "learning_rate": 9.595611959561197e-06, + "loss": 0.9049, + "step": 15730 + }, + { + "epoch": 103.55, + "learning_rate": 9.563346956334695e-06, + "loss": 0.8997, + "step": 15740 + }, + { + "epoch": 103.62, + "learning_rate": 9.531081953108197e-06, + "loss": 0.8749, + "step": 15750 + }, + { + "epoch": 103.68, + "learning_rate": 9.498816949881695e-06, + "loss": 0.8853, + "step": 15760 + }, + { + "epoch": 103.75, + "learning_rate": 9.466551946655194e-06, + "loss": 0.9405, + "step": 15770 + }, + { + "epoch": 103.81, + "learning_rate": 9.434286943428695e-06, + "loss": 0.8564, + "step": 15780 + }, + { + "epoch": 103.88, + "learning_rate": 9.402021940202194e-06, + "loss": 0.8431, + "step": 15790 + }, + { + "epoch": 103.94, + "learning_rate": 9.369756936975694e-06, + "loss": 0.8918, + "step": 15800 + }, + { + "epoch": 104.01, + "learning_rate": 9.337491933749194e-06, + "loss": 1.0248, + "step": 15810 + }, + { + "epoch": 104.08, + "learning_rate": 9.305226930522693e-06, + "loss": 0.8441, + "step": 15820 + }, + { + "epoch": 104.14, + "learning_rate": 9.272961927296194e-06, + "loss": 0.8754, + "step": 15830 + }, + { + "epoch": 104.21, + "learning_rate": 9.240696924069693e-06, + "loss": 0.8758, + "step": 15840 + }, + { + "epoch": 104.27, + "learning_rate": 9.208431920843191e-06, + "loss": 0.9222, + "step": 15850 + }, + { + "epoch": 104.34, + "learning_rate": 9.176166917616693e-06, + "loss": 0.8714, + "step": 15860 + }, + { + "epoch": 104.41, + "learning_rate": 9.143901914390191e-06, + "loss": 0.8834, + "step": 15870 + }, + { + "epoch": 104.47, + "learning_rate": 9.111636911163692e-06, + "loss": 0.8587, + "step": 15880 + }, + { + "epoch": 104.54, + "learning_rate": 9.079371907937192e-06, + "loss": 0.9453, + "step": 15890 + }, + { + "epoch": 104.6, + "learning_rate": 9.04710690471069e-06, + "loss": 0.8731, + "step": 15900 + }, + { + "epoch": 104.67, + "learning_rate": 9.01484190148419e-06, + "loss": 0.8939, + "step": 15910 + }, + { + "epoch": 104.73, + "learning_rate": 8.98257689825769e-06, + "loss": 0.8906, + "step": 15920 + }, + { + "epoch": 104.8, + "learning_rate": 8.950311895031189e-06, + "loss": 0.8732, + "step": 15930 + }, + { + "epoch": 104.86, + "learning_rate": 8.91804689180469e-06, + "loss": 0.8841, + "step": 15940 + }, + { + "epoch": 104.93, + "learning_rate": 8.885781888578189e-06, + "loss": 0.8742, + "step": 15950 + }, + { + "epoch": 105.0, + "learning_rate": 8.853516885351689e-06, + "loss": 0.9412, + "step": 15960 + }, + { + "epoch": 105.07, + "learning_rate": 8.82125188212519e-06, + "loss": 0.9196, + "step": 15970 + }, + { + "epoch": 105.13, + "learning_rate": 8.788986878898688e-06, + "loss": 0.8791, + "step": 15980 + }, + { + "epoch": 105.2, + "learning_rate": 8.756721875672188e-06, + "loss": 0.8724, + "step": 15990 + }, + { + "epoch": 105.26, + "learning_rate": 8.724456872445688e-06, + "loss": 0.9433, + "step": 16000 + }, + { + "epoch": 105.26, + "eval_loss": 0.33956313133239746, + "eval_runtime": 325.5194, + "eval_samples_per_second": 21.842, + "eval_steps_per_second": 1.367, + "eval_wer": 0.35300493564308527, + "step": 16000 + }, + { + "epoch": 105.33, + "learning_rate": 8.692191869219186e-06, + "loss": 0.8547, + "step": 16010 + }, + { + "epoch": 105.39, + "learning_rate": 8.659926865992686e-06, + "loss": 0.8773, + "step": 16020 + }, + { + "epoch": 105.46, + "learning_rate": 8.627661862766187e-06, + "loss": 0.8786, + "step": 16030 + }, + { + "epoch": 105.52, + "learning_rate": 8.595396859539685e-06, + "loss": 0.9219, + "step": 16040 + }, + { + "epoch": 105.59, + "learning_rate": 8.563131856313187e-06, + "loss": 0.8632, + "step": 16050 + }, + { + "epoch": 105.65, + "learning_rate": 8.530866853086685e-06, + "loss": 0.8829, + "step": 16060 + }, + { + "epoch": 105.72, + "learning_rate": 8.498601849860185e-06, + "loss": 0.8814, + "step": 16070 + }, + { + "epoch": 105.79, + "learning_rate": 8.466336846633685e-06, + "loss": 0.9033, + "step": 16080 + }, + { + "epoch": 105.85, + "learning_rate": 8.434071843407184e-06, + "loss": 0.8675, + "step": 16090 + }, + { + "epoch": 105.92, + "learning_rate": 8.401806840180684e-06, + "loss": 0.8758, + "step": 16100 + }, + { + "epoch": 105.98, + "learning_rate": 8.369541836954184e-06, + "loss": 0.8747, + "step": 16110 + }, + { + "epoch": 106.05, + "learning_rate": 8.337276833727683e-06, + "loss": 0.9735, + "step": 16120 + }, + { + "epoch": 106.12, + "learning_rate": 8.305011830501184e-06, + "loss": 0.8857, + "step": 16130 + }, + { + "epoch": 106.18, + "learning_rate": 8.272746827274683e-06, + "loss": 0.87, + "step": 16140 + }, + { + "epoch": 106.25, + "learning_rate": 8.240481824048183e-06, + "loss": 0.9338, + "step": 16150 + }, + { + "epoch": 106.31, + "learning_rate": 8.208216820821683e-06, + "loss": 0.8607, + "step": 16160 + }, + { + "epoch": 106.38, + "learning_rate": 8.175951817595181e-06, + "loss": 0.8774, + "step": 16170 + }, + { + "epoch": 106.45, + "learning_rate": 8.143686814368682e-06, + "loss": 0.8379, + "step": 16180 + }, + { + "epoch": 106.51, + "learning_rate": 8.111421811142182e-06, + "loss": 0.9258, + "step": 16190 + }, + { + "epoch": 106.58, + "learning_rate": 8.07915680791568e-06, + "loss": 0.8407, + "step": 16200 + }, + { + "epoch": 106.64, + "learning_rate": 8.04689180468918e-06, + "loss": 0.8785, + "step": 16210 + }, + { + "epoch": 106.71, + "learning_rate": 8.01462680146268e-06, + "loss": 0.9169, + "step": 16220 + }, + { + "epoch": 106.77, + "learning_rate": 7.98236179823618e-06, + "loss": 0.9558, + "step": 16230 + }, + { + "epoch": 106.84, + "learning_rate": 7.95009679500968e-06, + "loss": 0.8568, + "step": 16240 + }, + { + "epoch": 106.9, + "learning_rate": 7.917831791783179e-06, + "loss": 0.8971, + "step": 16250 + }, + { + "epoch": 106.97, + "learning_rate": 7.885566788556679e-06, + "loss": 0.8865, + "step": 16260 + }, + { + "epoch": 107.04, + "learning_rate": 7.85330178533018e-06, + "loss": 1.012, + "step": 16270 + }, + { + "epoch": 107.1, + "learning_rate": 7.821036782103678e-06, + "loss": 0.871, + "step": 16280 + }, + { + "epoch": 107.17, + "learning_rate": 7.788771778877178e-06, + "loss": 0.8686, + "step": 16290 + }, + { + "epoch": 107.24, + "learning_rate": 7.756506775650678e-06, + "loss": 0.9085, + "step": 16300 + }, + { + "epoch": 107.3, + "learning_rate": 7.724241772424178e-06, + "loss": 0.905, + "step": 16310 + }, + { + "epoch": 107.37, + "learning_rate": 7.691976769197676e-06, + "loss": 0.8584, + "step": 16320 + }, + { + "epoch": 107.43, + "learning_rate": 7.659711765971177e-06, + "loss": 0.9066, + "step": 16330 + }, + { + "epoch": 107.5, + "learning_rate": 7.627446762744677e-06, + "loss": 0.9365, + "step": 16340 + }, + { + "epoch": 107.56, + "learning_rate": 7.595181759518176e-06, + "loss": 0.8756, + "step": 16350 + }, + { + "epoch": 107.63, + "learning_rate": 7.562916756291675e-06, + "loss": 0.8542, + "step": 16360 + }, + { + "epoch": 107.69, + "learning_rate": 7.530651753065176e-06, + "loss": 0.8647, + "step": 16370 + }, + { + "epoch": 107.76, + "learning_rate": 7.498386749838675e-06, + "loss": 0.9443, + "step": 16380 + }, + { + "epoch": 107.82, + "learning_rate": 7.466121746612175e-06, + "loss": 0.8396, + "step": 16390 + }, + { + "epoch": 107.89, + "learning_rate": 7.433856743385675e-06, + "loss": 0.8714, + "step": 16400 + }, + { + "epoch": 107.96, + "learning_rate": 7.401591740159175e-06, + "loss": 0.8858, + "step": 16410 + }, + { + "epoch": 108.03, + "learning_rate": 7.369326736932673e-06, + "loss": 0.9822, + "step": 16420 + }, + { + "epoch": 108.09, + "learning_rate": 7.3370617337061735e-06, + "loss": 0.8577, + "step": 16430 + }, + { + "epoch": 108.16, + "learning_rate": 7.304796730479674e-06, + "loss": 0.8978, + "step": 16440 + }, + { + "epoch": 108.22, + "learning_rate": 7.272531727253173e-06, + "loss": 0.8909, + "step": 16450 + }, + { + "epoch": 108.29, + "learning_rate": 7.240266724026672e-06, + "loss": 0.9385, + "step": 16460 + }, + { + "epoch": 108.35, + "learning_rate": 7.208001720800172e-06, + "loss": 0.8541, + "step": 16470 + }, + { + "epoch": 108.42, + "learning_rate": 7.175736717573672e-06, + "loss": 0.8644, + "step": 16480 + }, + { + "epoch": 108.48, + "learning_rate": 7.143471714347172e-06, + "loss": 0.911, + "step": 16490 + }, + { + "epoch": 108.55, + "learning_rate": 7.111206711120671e-06, + "loss": 0.8988, + "step": 16500 + }, + { + "epoch": 108.55, + "eval_loss": 0.3435661494731903, + "eval_runtime": 318.354, + "eval_samples_per_second": 22.334, + "eval_steps_per_second": 1.398, + "eval_wer": 0.3539146424078196, + "step": 16500 + }, + { + "epoch": 108.62, + "learning_rate": 7.078941707894171e-06, + "loss": 0.8563, + "step": 16510 + }, + { + "epoch": 108.68, + "learning_rate": 7.046676704667671e-06, + "loss": 0.8605, + "step": 16520 + }, + { + "epoch": 108.75, + "learning_rate": 7.017638201763821e-06, + "loss": 0.9282, + "step": 16530 + }, + { + "epoch": 108.81, + "learning_rate": 6.98537319853732e-06, + "loss": 0.8875, + "step": 16540 + }, + { + "epoch": 108.88, + "learning_rate": 6.953108195310819e-06, + "loss": 0.8576, + "step": 16550 + }, + { + "epoch": 108.94, + "learning_rate": 6.9208431920843194e-06, + "loss": 0.879, + "step": 16560 + }, + { + "epoch": 109.01, + "learning_rate": 6.8885781888578196e-06, + "loss": 0.9845, + "step": 16570 + }, + { + "epoch": 109.08, + "learning_rate": 6.856313185631319e-06, + "loss": 0.8595, + "step": 16580 + }, + { + "epoch": 109.14, + "learning_rate": 6.824048182404818e-06, + "loss": 0.8792, + "step": 16590 + }, + { + "epoch": 109.21, + "learning_rate": 6.791783179178318e-06, + "loss": 0.89, + "step": 16600 + }, + { + "epoch": 109.27, + "learning_rate": 6.759518175951818e-06, + "loss": 0.9244, + "step": 16610 + }, + { + "epoch": 109.34, + "learning_rate": 6.727253172725318e-06, + "loss": 0.8525, + "step": 16620 + }, + { + "epoch": 109.41, + "learning_rate": 6.694988169498817e-06, + "loss": 0.8619, + "step": 16630 + }, + { + "epoch": 109.47, + "learning_rate": 6.662723166272317e-06, + "loss": 0.9235, + "step": 16640 + }, + { + "epoch": 109.54, + "learning_rate": 6.630458163045816e-06, + "loss": 0.9352, + "step": 16650 + }, + { + "epoch": 109.6, + "learning_rate": 6.598193159819316e-06, + "loss": 0.8496, + "step": 16660 + }, + { + "epoch": 109.67, + "learning_rate": 6.565928156592816e-06, + "loss": 0.8729, + "step": 16670 + }, + { + "epoch": 109.73, + "learning_rate": 6.533663153366316e-06, + "loss": 0.9166, + "step": 16680 + }, + { + "epoch": 109.8, + "learning_rate": 6.501398150139815e-06, + "loss": 0.8909, + "step": 16690 + }, + { + "epoch": 109.86, + "learning_rate": 6.469133146913315e-06, + "loss": 0.882, + "step": 16700 + }, + { + "epoch": 109.93, + "learning_rate": 6.4368681436868144e-06, + "loss": 0.8945, + "step": 16710 + }, + { + "epoch": 110.0, + "learning_rate": 6.4046031404603146e-06, + "loss": 0.9299, + "step": 16720 + }, + { + "epoch": 110.07, + "learning_rate": 6.372338137233814e-06, + "loss": 0.9232, + "step": 16730 + }, + { + "epoch": 110.13, + "learning_rate": 6.340073134007314e-06, + "loss": 0.8516, + "step": 16740 + }, + { + "epoch": 110.2, + "learning_rate": 6.307808130780813e-06, + "loss": 0.8654, + "step": 16750 + }, + { + "epoch": 110.26, + "learning_rate": 6.275543127554313e-06, + "loss": 0.9617, + "step": 16760 + }, + { + "epoch": 110.33, + "learning_rate": 6.243278124327813e-06, + "loss": 0.8597, + "step": 16770 + }, + { + "epoch": 110.39, + "learning_rate": 6.211013121101312e-06, + "loss": 0.858, + "step": 16780 + }, + { + "epoch": 110.46, + "learning_rate": 6.178748117874812e-06, + "loss": 0.8777, + "step": 16790 + }, + { + "epoch": 110.52, + "learning_rate": 6.146483114648311e-06, + "loss": 0.9278, + "step": 16800 + }, + { + "epoch": 110.59, + "learning_rate": 6.114218111421811e-06, + "loss": 0.8487, + "step": 16810 + }, + { + "epoch": 110.65, + "learning_rate": 6.081953108195311e-06, + "loss": 0.8721, + "step": 16820 + }, + { + "epoch": 110.72, + "learning_rate": 6.049688104968811e-06, + "loss": 0.9058, + "step": 16830 + }, + { + "epoch": 110.79, + "learning_rate": 6.01742310174231e-06, + "loss": 0.9162, + "step": 16840 + }, + { + "epoch": 110.85, + "learning_rate": 5.98515809851581e-06, + "loss": 0.861, + "step": 16850 + }, + { + "epoch": 110.92, + "learning_rate": 5.9528930952893094e-06, + "loss": 0.8752, + "step": 16860 + }, + { + "epoch": 110.98, + "learning_rate": 5.9206280920628096e-06, + "loss": 0.8966, + "step": 16870 + }, + { + "epoch": 111.05, + "learning_rate": 5.888363088836309e-06, + "loss": 0.9639, + "step": 16880 + }, + { + "epoch": 111.12, + "learning_rate": 5.856098085609809e-06, + "loss": 0.8261, + "step": 16890 + }, + { + "epoch": 111.18, + "learning_rate": 5.823833082383308e-06, + "loss": 0.873, + "step": 16900 + }, + { + "epoch": 111.25, + "learning_rate": 5.7915680791568075e-06, + "loss": 0.9507, + "step": 16910 + }, + { + "epoch": 111.31, + "learning_rate": 5.759303075930308e-06, + "loss": 0.8536, + "step": 16920 + }, + { + "epoch": 111.38, + "learning_rate": 5.727038072703808e-06, + "loss": 0.8887, + "step": 16930 + }, + { + "epoch": 111.45, + "learning_rate": 5.694773069477307e-06, + "loss": 0.8569, + "step": 16940 + }, + { + "epoch": 111.51, + "learning_rate": 5.662508066250806e-06, + "loss": 0.9231, + "step": 16950 + }, + { + "epoch": 111.58, + "learning_rate": 5.630243063024306e-06, + "loss": 0.8588, + "step": 16960 + }, + { + "epoch": 111.64, + "learning_rate": 5.5979780597978065e-06, + "loss": 0.8674, + "step": 16970 + }, + { + "epoch": 111.71, + "learning_rate": 5.565713056571306e-06, + "loss": 0.8802, + "step": 16980 + }, + { + "epoch": 111.77, + "learning_rate": 5.533448053344805e-06, + "loss": 0.966, + "step": 16990 + }, + { + "epoch": 111.84, + "learning_rate": 5.501183050118305e-06, + "loss": 0.8789, + "step": 17000 + }, + { + "epoch": 111.84, + "eval_loss": 0.3425801992416382, + "eval_runtime": 327.0047, + "eval_samples_per_second": 21.743, + "eval_steps_per_second": 1.361, + "eval_wer": 0.3515532759121262, + "step": 17000 + }, + { + "epoch": 111.9, + "learning_rate": 5.468918046891805e-06, + "loss": 0.8766, + "step": 17010 + }, + { + "epoch": 111.97, + "learning_rate": 5.4366530436653046e-06, + "loss": 0.8643, + "step": 17020 + }, + { + "epoch": 112.04, + "learning_rate": 5.404388040438804e-06, + "loss": 0.9841, + "step": 17030 + }, + { + "epoch": 112.1, + "learning_rate": 5.372123037212304e-06, + "loss": 0.8478, + "step": 17040 + }, + { + "epoch": 112.17, + "learning_rate": 5.339858033985804e-06, + "loss": 0.8621, + "step": 17050 + }, + { + "epoch": 112.24, + "learning_rate": 5.3075930307593025e-06, + "loss": 0.8895, + "step": 17060 + }, + { + "epoch": 112.3, + "learning_rate": 5.275328027532803e-06, + "loss": 0.8931, + "step": 17070 + }, + { + "epoch": 112.37, + "learning_rate": 5.243063024306303e-06, + "loss": 0.8679, + "step": 17080 + }, + { + "epoch": 112.43, + "learning_rate": 5.210798021079803e-06, + "loss": 0.8957, + "step": 17090 + }, + { + "epoch": 112.5, + "learning_rate": 5.178533017853301e-06, + "loss": 0.8836, + "step": 17100 + }, + { + "epoch": 112.56, + "learning_rate": 5.146268014626801e-06, + "loss": 0.8735, + "step": 17110 + }, + { + "epoch": 112.63, + "learning_rate": 5.1140030114003015e-06, + "loss": 0.8746, + "step": 17120 + }, + { + "epoch": 112.69, + "learning_rate": 5.081738008173802e-06, + "loss": 0.8755, + "step": 17130 + }, + { + "epoch": 112.76, + "learning_rate": 5.0494730049473e-06, + "loss": 0.935, + "step": 17140 + }, + { + "epoch": 112.82, + "learning_rate": 5.0172080017208e-06, + "loss": 0.8438, + "step": 17150 + }, + { + "epoch": 112.89, + "learning_rate": 4.9849429984943e-06, + "loss": 0.8687, + "step": 17160 + }, + { + "epoch": 112.96, + "learning_rate": 4.9526779952678e-06, + "loss": 0.8934, + "step": 17170 + }, + { + "epoch": 113.03, + "learning_rate": 4.920412992041299e-06, + "loss": 0.9906, + "step": 17180 + }, + { + "epoch": 113.09, + "learning_rate": 4.888147988814799e-06, + "loss": 0.842, + "step": 17190 + }, + { + "epoch": 113.16, + "learning_rate": 4.855882985588299e-06, + "loss": 0.8746, + "step": 17200 + }, + { + "epoch": 113.22, + "learning_rate": 4.823617982361798e-06, + "loss": 0.8782, + "step": 17210 + }, + { + "epoch": 113.29, + "learning_rate": 4.791352979135298e-06, + "loss": 0.913, + "step": 17220 + }, + { + "epoch": 113.35, + "learning_rate": 4.759087975908798e-06, + "loss": 0.826, + "step": 17230 + }, + { + "epoch": 113.42, + "learning_rate": 4.726822972682298e-06, + "loss": 0.8792, + "step": 17240 + }, + { + "epoch": 113.48, + "learning_rate": 4.694557969455797e-06, + "loss": 0.8961, + "step": 17250 + }, + { + "epoch": 113.55, + "learning_rate": 4.662292966229296e-06, + "loss": 0.8789, + "step": 17260 + }, + { + "epoch": 113.62, + "learning_rate": 4.6300279630027965e-06, + "loss": 0.8453, + "step": 17270 + }, + { + "epoch": 113.68, + "learning_rate": 4.597762959776297e-06, + "loss": 0.8834, + "step": 17280 + }, + { + "epoch": 113.75, + "learning_rate": 4.565497956549796e-06, + "loss": 0.9247, + "step": 17290 + }, + { + "epoch": 113.81, + "learning_rate": 4.533232953323295e-06, + "loss": 0.8541, + "step": 17300 + }, + { + "epoch": 113.88, + "learning_rate": 4.500967950096795e-06, + "loss": 0.854, + "step": 17310 + }, + { + "epoch": 113.94, + "learning_rate": 4.468702946870295e-06, + "loss": 0.9033, + "step": 17320 + }, + { + "epoch": 114.01, + "learning_rate": 4.436437943643794e-06, + "loss": 1.0242, + "step": 17330 + }, + { + "epoch": 114.08, + "learning_rate": 4.404172940417294e-06, + "loss": 0.8485, + "step": 17340 + }, + { + "epoch": 114.14, + "learning_rate": 4.371907937190794e-06, + "loss": 0.8935, + "step": 17350 + }, + { + "epoch": 114.21, + "learning_rate": 4.339642933964293e-06, + "loss": 0.871, + "step": 17360 + }, + { + "epoch": 114.27, + "learning_rate": 4.307377930737793e-06, + "loss": 0.9397, + "step": 17370 + }, + { + "epoch": 114.34, + "learning_rate": 4.275112927511293e-06, + "loss": 0.8355, + "step": 17380 + }, + { + "epoch": 114.41, + "learning_rate": 4.242847924284793e-06, + "loss": 0.8424, + "step": 17390 + }, + { + "epoch": 114.47, + "learning_rate": 4.210582921058292e-06, + "loss": 0.8736, + "step": 17400 + }, + { + "epoch": 114.54, + "learning_rate": 4.178317917831791e-06, + "loss": 0.9186, + "step": 17410 + }, + { + "epoch": 114.6, + "learning_rate": 4.1460529146052915e-06, + "loss": 0.8414, + "step": 17420 + }, + { + "epoch": 114.67, + "learning_rate": 4.113787911378792e-06, + "loss": 0.8786, + "step": 17430 + }, + { + "epoch": 114.73, + "learning_rate": 4.081522908152291e-06, + "loss": 0.8891, + "step": 17440 + }, + { + "epoch": 114.8, + "learning_rate": 4.04925790492579e-06, + "loss": 0.9005, + "step": 17450 + }, + { + "epoch": 114.86, + "learning_rate": 4.01699290169929e-06, + "loss": 0.8603, + "step": 17460 + }, + { + "epoch": 114.93, + "learning_rate": 3.98472789847279e-06, + "loss": 0.869, + "step": 17470 + }, + { + "epoch": 115.0, + "learning_rate": 3.95246289524629e-06, + "loss": 0.9212, + "step": 17480 + }, + { + "epoch": 115.07, + "learning_rate": 3.920197892019789e-06, + "loss": 0.9215, + "step": 17490 + }, + { + "epoch": 115.13, + "learning_rate": 3.887932888793289e-06, + "loss": 0.8667, + "step": 17500 + }, + { + "epoch": 115.13, + "eval_loss": 0.3438130021095276, + "eval_runtime": 324.2107, + "eval_samples_per_second": 21.93, + "eval_steps_per_second": 1.373, + "eval_wer": 0.35060485822123294, + "step": 17500 + }, + { + "epoch": 115.2, + "learning_rate": 3.855667885566788e-06, + "loss": 0.8505, + "step": 17510 + }, + { + "epoch": 115.26, + "learning_rate": 3.8234028823402885e-06, + "loss": 0.935, + "step": 17520 + }, + { + "epoch": 115.33, + "learning_rate": 3.7911378791137877e-06, + "loss": 0.8515, + "step": 17530 + }, + { + "epoch": 115.39, + "learning_rate": 3.7588728758872874e-06, + "loss": 0.8767, + "step": 17540 + }, + { + "epoch": 115.46, + "learning_rate": 3.7266078726607876e-06, + "loss": 0.891, + "step": 17550 + }, + { + "epoch": 115.52, + "learning_rate": 3.694342869434287e-06, + "loss": 0.9347, + "step": 17560 + }, + { + "epoch": 115.59, + "learning_rate": 3.662077866207787e-06, + "loss": 0.8439, + "step": 17570 + }, + { + "epoch": 115.65, + "learning_rate": 3.6298128629812862e-06, + "loss": 0.8656, + "step": 17580 + }, + { + "epoch": 115.72, + "learning_rate": 3.5975478597547863e-06, + "loss": 0.8868, + "step": 17590 + }, + { + "epoch": 115.79, + "learning_rate": 3.5652828565282856e-06, + "loss": 0.919, + "step": 17600 + }, + { + "epoch": 115.85, + "learning_rate": 3.5330178533017857e-06, + "loss": 0.8582, + "step": 17610 + }, + { + "epoch": 115.92, + "learning_rate": 3.500752850075285e-06, + "loss": 0.8699, + "step": 17620 + }, + { + "epoch": 115.98, + "learning_rate": 3.4684878468487847e-06, + "loss": 0.8604, + "step": 17630 + }, + { + "epoch": 116.05, + "learning_rate": 3.4362228436222844e-06, + "loss": 0.9501, + "step": 17640 + }, + { + "epoch": 116.12, + "learning_rate": 3.403957840395784e-06, + "loss": 0.8585, + "step": 17650 + }, + { + "epoch": 116.18, + "learning_rate": 3.3716928371692838e-06, + "loss": 0.8697, + "step": 17660 + }, + { + "epoch": 116.25, + "learning_rate": 3.3394278339427835e-06, + "loss": 0.942, + "step": 17670 + }, + { + "epoch": 116.31, + "learning_rate": 3.307162830716283e-06, + "loss": 0.879, + "step": 17680 + }, + { + "epoch": 116.38, + "learning_rate": 3.274897827489783e-06, + "loss": 0.8652, + "step": 17690 + }, + { + "epoch": 116.45, + "learning_rate": 3.2426328242632826e-06, + "loss": 0.871, + "step": 17700 + }, + { + "epoch": 116.51, + "learning_rate": 3.2103678210367823e-06, + "loss": 0.9304, + "step": 17710 + }, + { + "epoch": 116.58, + "learning_rate": 3.178102817810282e-06, + "loss": 0.8397, + "step": 17720 + }, + { + "epoch": 116.64, + "learning_rate": 3.1458378145837816e-06, + "loss": 0.8661, + "step": 17730 + }, + { + "epoch": 116.71, + "learning_rate": 3.1135728113572813e-06, + "loss": 0.8686, + "step": 17740 + }, + { + "epoch": 116.77, + "learning_rate": 3.081307808130781e-06, + "loss": 0.9082, + "step": 17750 + }, + { + "epoch": 116.84, + "learning_rate": 3.0490428049042807e-06, + "loss": 0.8541, + "step": 17760 + }, + { + "epoch": 116.9, + "learning_rate": 3.0167778016777804e-06, + "loss": 0.8637, + "step": 17770 + }, + { + "epoch": 116.97, + "learning_rate": 2.9845127984512797e-06, + "loss": 0.8832, + "step": 17780 + }, + { + "epoch": 117.04, + "learning_rate": 2.95224779522478e-06, + "loss": 0.9752, + "step": 17790 + }, + { + "epoch": 117.1, + "learning_rate": 2.919982791998279e-06, + "loss": 0.8353, + "step": 17800 + }, + { + "epoch": 117.17, + "learning_rate": 2.887717788771779e-06, + "loss": 0.8827, + "step": 17810 + }, + { + "epoch": 117.24, + "learning_rate": 2.8554527855452785e-06, + "loss": 0.8747, + "step": 17820 + }, + { + "epoch": 117.3, + "learning_rate": 2.8231877823187786e-06, + "loss": 0.8857, + "step": 17830 + }, + { + "epoch": 117.37, + "learning_rate": 2.790922779092278e-06, + "loss": 0.8432, + "step": 17840 + }, + { + "epoch": 117.43, + "learning_rate": 2.758657775865778e-06, + "loss": 0.8877, + "step": 17850 + }, + { + "epoch": 117.5, + "learning_rate": 2.7263927726392773e-06, + "loss": 0.9045, + "step": 17860 + }, + { + "epoch": 117.56, + "learning_rate": 2.694127769412777e-06, + "loss": 0.8463, + "step": 17870 + }, + { + "epoch": 117.63, + "learning_rate": 2.6618627661862766e-06, + "loss": 0.8323, + "step": 17880 + }, + { + "epoch": 117.69, + "learning_rate": 2.6295977629597763e-06, + "loss": 0.8763, + "step": 17890 + }, + { + "epoch": 117.76, + "learning_rate": 2.597332759733276e-06, + "loss": 0.9353, + "step": 17900 + }, + { + "epoch": 117.82, + "learning_rate": 2.5650677565067757e-06, + "loss": 0.8474, + "step": 17910 + }, + { + "epoch": 117.89, + "learning_rate": 2.5328027532802754e-06, + "loss": 0.8529, + "step": 17920 + }, + { + "epoch": 117.96, + "learning_rate": 2.5005377500537747e-06, + "loss": 0.8589, + "step": 17930 + }, + { + "epoch": 118.03, + "learning_rate": 2.468272746827275e-06, + "loss": 1.0224, + "step": 17940 + }, + { + "epoch": 118.09, + "learning_rate": 2.436007743600774e-06, + "loss": 0.8404, + "step": 17950 + }, + { + "epoch": 118.16, + "learning_rate": 2.403742740374274e-06, + "loss": 0.8568, + "step": 17960 + }, + { + "epoch": 118.22, + "learning_rate": 2.3714777371477735e-06, + "loss": 0.8704, + "step": 17970 + }, + { + "epoch": 118.29, + "learning_rate": 2.3392127339212736e-06, + "loss": 0.9164, + "step": 17980 + }, + { + "epoch": 118.35, + "learning_rate": 2.306947730694773e-06, + "loss": 0.8434, + "step": 17990 + }, + { + "epoch": 118.42, + "learning_rate": 2.274682727468273e-06, + "loss": 0.8895, + "step": 18000 + }, + { + "epoch": 118.42, + "eval_loss": 0.34335601329803467, + "eval_runtime": 325.3466, + "eval_samples_per_second": 21.854, + "eval_steps_per_second": 1.368, + "eval_wer": 0.3502564598858028, + "step": 18000 + }, + { + "epoch": 118.48, + "learning_rate": 2.2424177242417723e-06, + "loss": 0.9049, + "step": 18010 + }, + { + "epoch": 118.55, + "learning_rate": 2.2101527210152724e-06, + "loss": 0.8719, + "step": 18020 + }, + { + "epoch": 118.62, + "learning_rate": 2.1778877177887716e-06, + "loss": 0.86, + "step": 18030 + }, + { + "epoch": 118.68, + "learning_rate": 2.1456227145622718e-06, + "loss": 0.8437, + "step": 18040 + }, + { + "epoch": 118.75, + "learning_rate": 2.113357711335771e-06, + "loss": 0.9047, + "step": 18050 + }, + { + "epoch": 118.81, + "learning_rate": 2.0810927081092707e-06, + "loss": 0.8641, + "step": 18060 + }, + { + "epoch": 118.88, + "learning_rate": 2.0488277048827704e-06, + "loss": 0.834, + "step": 18070 + }, + { + "epoch": 118.94, + "learning_rate": 2.01656270165627e-06, + "loss": 0.9071, + "step": 18080 + }, + { + "epoch": 119.01, + "learning_rate": 1.98429769842977e-06, + "loss": 0.9641, + "step": 18090 + }, + { + "epoch": 119.08, + "learning_rate": 1.9520326952032695e-06, + "loss": 0.8182, + "step": 18100 + }, + { + "epoch": 119.14, + "learning_rate": 1.919767691976769e-06, + "loss": 0.8624, + "step": 18110 + }, + { + "epoch": 119.21, + "learning_rate": 1.8875026887502691e-06, + "loss": 0.8535, + "step": 18120 + }, + { + "epoch": 119.27, + "learning_rate": 1.8552376855237686e-06, + "loss": 0.931, + "step": 18130 + }, + { + "epoch": 119.34, + "learning_rate": 1.8229726822972683e-06, + "loss": 0.8267, + "step": 18140 + }, + { + "epoch": 119.41, + "learning_rate": 1.7907076790707678e-06, + "loss": 0.8664, + "step": 18150 + }, + { + "epoch": 119.47, + "learning_rate": 1.7584426758442675e-06, + "loss": 0.8943, + "step": 18160 + }, + { + "epoch": 119.54, + "learning_rate": 1.7261776726177672e-06, + "loss": 0.9318, + "step": 18170 + }, + { + "epoch": 119.6, + "learning_rate": 1.6939126693912669e-06, + "loss": 0.8273, + "step": 18180 + }, + { + "epoch": 119.67, + "learning_rate": 1.6616476661647666e-06, + "loss": 0.8521, + "step": 18190 + }, + { + "epoch": 119.73, + "learning_rate": 1.6293826629382662e-06, + "loss": 0.8815, + "step": 18200 + }, + { + "epoch": 119.8, + "learning_rate": 1.597117659711766e-06, + "loss": 0.8892, + "step": 18210 + }, + { + "epoch": 119.86, + "learning_rate": 1.5648526564852656e-06, + "loss": 0.8461, + "step": 18220 + }, + { + "epoch": 119.93, + "learning_rate": 1.5325876532587653e-06, + "loss": 0.8724, + "step": 18230 + }, + { + "epoch": 120.0, + "learning_rate": 1.500322650032265e-06, + "loss": 0.9253, + "step": 18240 + }, + { + "epoch": 120.07, + "learning_rate": 1.4680576468057647e-06, + "loss": 0.9371, + "step": 18250 + }, + { + "epoch": 120.13, + "learning_rate": 1.4357926435792644e-06, + "loss": 0.8727, + "step": 18260 + }, + { + "epoch": 120.2, + "learning_rate": 1.4035276403527641e-06, + "loss": 0.8629, + "step": 18270 + }, + { + "epoch": 120.26, + "learning_rate": 1.3712626371262636e-06, + "loss": 0.9365, + "step": 18280 + }, + { + "epoch": 120.33, + "learning_rate": 1.3389976338997633e-06, + "loss": 0.8307, + "step": 18290 + }, + { + "epoch": 120.39, + "learning_rate": 1.306732630673263e-06, + "loss": 0.8293, + "step": 18300 + }, + { + "epoch": 120.46, + "learning_rate": 1.2744676274467627e-06, + "loss": 0.8891, + "step": 18310 + }, + { + "epoch": 120.52, + "learning_rate": 1.2422026242202624e-06, + "loss": 0.9369, + "step": 18320 + }, + { + "epoch": 120.59, + "learning_rate": 1.209937620993762e-06, + "loss": 0.8345, + "step": 18330 + }, + { + "epoch": 120.65, + "learning_rate": 1.1776726177672618e-06, + "loss": 0.863, + "step": 18340 + }, + { + "epoch": 120.72, + "learning_rate": 1.1454076145407615e-06, + "loss": 0.869, + "step": 18350 + }, + { + "epoch": 120.79, + "learning_rate": 1.1131426113142612e-06, + "loss": 0.9124, + "step": 18360 + }, + { + "epoch": 120.85, + "learning_rate": 1.0808776080877609e-06, + "loss": 0.8465, + "step": 18370 + }, + { + "epoch": 120.92, + "learning_rate": 1.0486126048612605e-06, + "loss": 0.9023, + "step": 18380 + }, + { + "epoch": 120.98, + "learning_rate": 1.0163476016347602e-06, + "loss": 0.8672, + "step": 18390 + }, + { + "epoch": 121.05, + "learning_rate": 9.8408259840826e-07, + "loss": 0.9339, + "step": 18400 + }, + { + "epoch": 121.12, + "learning_rate": 9.518175951817595e-07, + "loss": 0.8539, + "step": 18410 + }, + { + "epoch": 121.18, + "learning_rate": 9.195525919552592e-07, + "loss": 0.8677, + "step": 18420 + }, + { + "epoch": 121.25, + "learning_rate": 8.872875887287589e-07, + "loss": 0.9385, + "step": 18430 + }, + { + "epoch": 121.31, + "learning_rate": 8.550225855022586e-07, + "loss": 0.8482, + "step": 18440 + }, + { + "epoch": 121.38, + "learning_rate": 8.227575822757583e-07, + "loss": 0.8607, + "step": 18450 + }, + { + "epoch": 121.45, + "learning_rate": 7.904925790492579e-07, + "loss": 0.8591, + "step": 18460 + }, + { + "epoch": 121.51, + "learning_rate": 7.582275758227576e-07, + "loss": 0.9215, + "step": 18470 + }, + { + "epoch": 121.58, + "learning_rate": 7.259625725962573e-07, + "loss": 0.8359, + "step": 18480 + }, + { + "epoch": 121.64, + "learning_rate": 6.93697569369757e-07, + "loss": 0.8706, + "step": 18490 + }, + { + "epoch": 121.71, + "learning_rate": 6.614325661432567e-07, + "loss": 0.8888, + "step": 18500 + }, + { + "epoch": 121.71, + "eval_loss": 0.34249696135520935, + "eval_runtime": 323.0236, + "eval_samples_per_second": 22.011, + "eval_steps_per_second": 1.378, + "eval_wer": 0.3494048195103068, + "step": 18500 + }, + { + "epoch": 121.77, + "learning_rate": 6.291675629167564e-07, + "loss": 0.9699, + "step": 18510 + }, + { + "epoch": 121.84, + "learning_rate": 5.969025596902561e-07, + "loss": 0.8446, + "step": 18520 + }, + { + "epoch": 121.9, + "learning_rate": 5.646375564637557e-07, + "loss": 0.8564, + "step": 18530 + }, + { + "epoch": 121.97, + "learning_rate": 5.323725532372553e-07, + "loss": 0.8883, + "step": 18540 + }, + { + "epoch": 122.04, + "learning_rate": 5.00107550010755e-07, + "loss": 0.9933, + "step": 18550 + }, + { + "epoch": 122.1, + "learning_rate": 4.678425467842547e-07, + "loss": 0.8476, + "step": 18560 + }, + { + "epoch": 122.17, + "learning_rate": 4.355775435577544e-07, + "loss": 0.8747, + "step": 18570 + }, + { + "epoch": 122.24, + "learning_rate": 4.03312540331254e-07, + "loss": 0.8982, + "step": 18580 + }, + { + "epoch": 122.3, + "learning_rate": 3.710475371047537e-07, + "loss": 0.8518, + "step": 18590 + }, + { + "epoch": 122.37, + "learning_rate": 3.387825338782534e-07, + "loss": 0.8722, + "step": 18600 + }, + { + "epoch": 122.43, + "learning_rate": 3.065175306517531e-07, + "loss": 0.8711, + "step": 18610 + }, + { + "epoch": 122.5, + "learning_rate": 2.7425252742525275e-07, + "loss": 0.9164, + "step": 18620 + }, + { + "epoch": 122.56, + "learning_rate": 2.4198752419875245e-07, + "loss": 0.8547, + "step": 18630 + }, + { + "epoch": 122.63, + "learning_rate": 2.0972252097225211e-07, + "loss": 0.858, + "step": 18640 + }, + { + "epoch": 122.69, + "learning_rate": 1.7745751774575178e-07, + "loss": 0.8493, + "step": 18650 + }, + { + "epoch": 122.76, + "learning_rate": 1.4519251451925145e-07, + "loss": 0.9404, + "step": 18660 + }, + { + "epoch": 122.82, + "learning_rate": 1.1292751129275113e-07, + "loss": 0.8165, + "step": 18670 + }, + { + "epoch": 122.89, + "learning_rate": 8.066250806625082e-08, + "loss": 0.8654, + "step": 18680 + }, + { + "epoch": 122.96, + "learning_rate": 4.8397504839750484e-08, + "loss": 0.8826, + "step": 18690 + }, + { + "epoch": 123.0, + "step": 18696, + "total_flos": 2.5832366503240237e+20, + "train_loss": 0.7828263307311466, + "train_runtime": 82672.8006, + "train_samples_per_second": 29.058, + "train_steps_per_second": 0.226 + } + ], + "max_steps": 18696, + "num_train_epochs": 123, + "total_flos": 2.5832366503240237e+20, + "trial_name": null, + "trial_params": null +}