| { |
| "best_global_step": null, |
| "best_metric": null, |
| "best_model_checkpoint": null, |
| "epoch": 1.8251875887243967, |
| "eval_steps": 1500, |
| "global_step": 90000, |
| "is_hyper_param_search": false, |
| "is_local_process_zero": true, |
| "is_world_process_zero": true, |
| "log_history": [ |
| { |
| "epoch": 0.001323276432446738, |
| "grad_norm": 48.5, |
| "learning_rate": 1.5560495395363608e-06, |
| "loss": 4.0059, |
| "step": 50 |
| }, |
| { |
| "epoch": 0.002646552864893476, |
| "grad_norm": 5.46875, |
| "learning_rate": 3.1438551921244835e-06, |
| "loss": 2.8343, |
| "step": 100 |
| }, |
| { |
| "epoch": 0.003969829297340214, |
| "grad_norm": 4.8125, |
| "learning_rate": 4.731660844712607e-06, |
| "loss": 2.5854, |
| "step": 150 |
| }, |
| { |
| "epoch": 0.005293105729786952, |
| "grad_norm": 6.1875, |
| "learning_rate": 6.31946649730073e-06, |
| "loss": 2.5068, |
| "step": 200 |
| }, |
| { |
| "epoch": 0.0066163821622336905, |
| "grad_norm": 8.25, |
| "learning_rate": 7.907272149888853e-06, |
| "loss": 2.4497, |
| "step": 250 |
| }, |
| { |
| "epoch": 0.007939658594680429, |
| "grad_norm": 6.90625, |
| "learning_rate": 9.495077802476977e-06, |
| "loss": 2.374, |
| "step": 300 |
| }, |
| { |
| "epoch": 0.009262935027127167, |
| "grad_norm": 7.15625, |
| "learning_rate": 1.1082883455065099e-05, |
| "loss": 2.3558, |
| "step": 350 |
| }, |
| { |
| "epoch": 0.010586211459573905, |
| "grad_norm": 14.25, |
| "learning_rate": 1.2670689107653221e-05, |
| "loss": 2.3181, |
| "step": 400 |
| }, |
| { |
| "epoch": 0.011909487892020643, |
| "grad_norm": 11.0625, |
| "learning_rate": 1.4258494760241345e-05, |
| "loss": 2.2872, |
| "step": 450 |
| }, |
| { |
| "epoch": 0.013232764324467381, |
| "grad_norm": 13.0, |
| "learning_rate": 1.584630041282947e-05, |
| "loss": 2.2729, |
| "step": 500 |
| }, |
| { |
| "epoch": 0.01455604075691412, |
| "grad_norm": 16.375, |
| "learning_rate": 1.743410606541759e-05, |
| "loss": 2.2402, |
| "step": 550 |
| }, |
| { |
| "epoch": 0.015879317189360857, |
| "grad_norm": 8.4375, |
| "learning_rate": 1.902191171800571e-05, |
| "loss": 2.196, |
| "step": 600 |
| }, |
| { |
| "epoch": 0.017202593621807597, |
| "grad_norm": 8.75, |
| "learning_rate": 2.0609717370593838e-05, |
| "loss": 2.1501, |
| "step": 650 |
| }, |
| { |
| "epoch": 0.018525870054254334, |
| "grad_norm": 7.5625, |
| "learning_rate": 2.2197523023181962e-05, |
| "loss": 2.1772, |
| "step": 700 |
| }, |
| { |
| "epoch": 0.019849146486701073, |
| "grad_norm": 7.96875, |
| "learning_rate": 2.3785328675770086e-05, |
| "loss": 2.1476, |
| "step": 750 |
| }, |
| { |
| "epoch": 0.02117242291914781, |
| "grad_norm": 7.78125, |
| "learning_rate": 2.5373134328358206e-05, |
| "loss": 2.1264, |
| "step": 800 |
| }, |
| { |
| "epoch": 0.02249569935159455, |
| "grad_norm": 10.4375, |
| "learning_rate": 2.696093998094633e-05, |
| "loss": 2.0611, |
| "step": 850 |
| }, |
| { |
| "epoch": 0.023818975784041286, |
| "grad_norm": 13.0625, |
| "learning_rate": 2.8548745633534454e-05, |
| "loss": 2.0682, |
| "step": 900 |
| }, |
| { |
| "epoch": 0.025142252216488026, |
| "grad_norm": 8.25, |
| "learning_rate": 3.0136551286122578e-05, |
| "loss": 2.0182, |
| "step": 950 |
| }, |
| { |
| "epoch": 0.026465528648934762, |
| "grad_norm": 8.75, |
| "learning_rate": 3.17243569387107e-05, |
| "loss": 1.9976, |
| "step": 1000 |
| }, |
| { |
| "epoch": 0.027788805081381502, |
| "grad_norm": 6.5625, |
| "learning_rate": 3.331216259129882e-05, |
| "loss": 1.9901, |
| "step": 1050 |
| }, |
| { |
| "epoch": 0.02911208151382824, |
| "grad_norm": 8.4375, |
| "learning_rate": 3.489996824388694e-05, |
| "loss": 1.9705, |
| "step": 1100 |
| }, |
| { |
| "epoch": 0.030435357946274978, |
| "grad_norm": 9.25, |
| "learning_rate": 3.6487773896475066e-05, |
| "loss": 1.9057, |
| "step": 1150 |
| }, |
| { |
| "epoch": 0.031758634378721715, |
| "grad_norm": 7.5, |
| "learning_rate": 3.807557954906319e-05, |
| "loss": 1.9104, |
| "step": 1200 |
| }, |
| { |
| "epoch": 0.03308191081116845, |
| "grad_norm": 12.6875, |
| "learning_rate": 3.9663385201651314e-05, |
| "loss": 1.8885, |
| "step": 1250 |
| }, |
| { |
| "epoch": 0.034405187243615194, |
| "grad_norm": 7.34375, |
| "learning_rate": 4.125119085423944e-05, |
| "loss": 1.785, |
| "step": 1300 |
| }, |
| { |
| "epoch": 0.03572846367606193, |
| "grad_norm": 7.21875, |
| "learning_rate": 4.283899650682756e-05, |
| "loss": 1.2491, |
| "step": 1350 |
| }, |
| { |
| "epoch": 0.03705174010850867, |
| "grad_norm": 4.03125, |
| "learning_rate": 4.442680215941568e-05, |
| "loss": 0.543, |
| "step": 1400 |
| }, |
| { |
| "epoch": 0.0383750165409554, |
| "grad_norm": 3.359375, |
| "learning_rate": 4.60146078120038e-05, |
| "loss": 0.3884, |
| "step": 1450 |
| }, |
| { |
| "epoch": 0.03969829297340215, |
| "grad_norm": 2.921875, |
| "learning_rate": 4.7602413464591926e-05, |
| "loss": 0.3543, |
| "step": 1500 |
| }, |
| { |
| "epoch": 0.03969829297340215, |
| "eval_cer": 0.08528059032633051, |
| "eval_loss": 0.4342532753944397, |
| "eval_runtime": 157.0758, |
| "eval_samples_per_second": 16.101, |
| "eval_steps_per_second": 0.127, |
| "eval_wer": 0.21550466744457408, |
| "step": 1500 |
| }, |
| { |
| "epoch": 0.04102156940584888, |
| "grad_norm": 2.6875, |
| "learning_rate": 4.919021911718005e-05, |
| "loss": 0.3145, |
| "step": 1550 |
| }, |
| { |
| "epoch": 0.04234484583829562, |
| "grad_norm": 2.734375, |
| "learning_rate": 5.0778024769768174e-05, |
| "loss": 0.3196, |
| "step": 1600 |
| }, |
| { |
| "epoch": 0.043668122270742356, |
| "grad_norm": 2.609375, |
| "learning_rate": 5.2365830422356304e-05, |
| "loss": 0.2973, |
| "step": 1650 |
| }, |
| { |
| "epoch": 0.0449913987031891, |
| "grad_norm": 2.0625, |
| "learning_rate": 5.395363607494443e-05, |
| "loss": 0.3009, |
| "step": 1700 |
| }, |
| { |
| "epoch": 0.046314675135635835, |
| "grad_norm": 2.0625, |
| "learning_rate": 5.554144172753255e-05, |
| "loss": 0.2681, |
| "step": 1750 |
| }, |
| { |
| "epoch": 0.04763795156808257, |
| "grad_norm": 1.9375, |
| "learning_rate": 5.712924738012067e-05, |
| "loss": 0.2769, |
| "step": 1800 |
| }, |
| { |
| "epoch": 0.04896122800052931, |
| "grad_norm": 12.625, |
| "learning_rate": 5.871705303270879e-05, |
| "loss": 0.2833, |
| "step": 1850 |
| }, |
| { |
| "epoch": 0.05028450443297605, |
| "grad_norm": 2.0, |
| "learning_rate": 6.0304858685296916e-05, |
| "loss": 0.2482, |
| "step": 1900 |
| }, |
| { |
| "epoch": 0.05160778086542279, |
| "grad_norm": 2.125, |
| "learning_rate": 6.189266433788503e-05, |
| "loss": 0.2641, |
| "step": 1950 |
| }, |
| { |
| "epoch": 0.052931057297869524, |
| "grad_norm": 1.609375, |
| "learning_rate": 6.348046999047316e-05, |
| "loss": 0.2513, |
| "step": 2000 |
| }, |
| { |
| "epoch": 0.05425433373031626, |
| "grad_norm": 1.859375, |
| "learning_rate": 6.506827564306128e-05, |
| "loss": 0.2599, |
| "step": 2050 |
| }, |
| { |
| "epoch": 0.055577610162763004, |
| "grad_norm": 1.6796875, |
| "learning_rate": 6.66560812956494e-05, |
| "loss": 0.2556, |
| "step": 2100 |
| }, |
| { |
| "epoch": 0.05690088659520974, |
| "grad_norm": 1.5390625, |
| "learning_rate": 6.824388694823753e-05, |
| "loss": 0.2412, |
| "step": 2150 |
| }, |
| { |
| "epoch": 0.05822416302765648, |
| "grad_norm": 1.6875, |
| "learning_rate": 6.983169260082565e-05, |
| "loss": 0.2564, |
| "step": 2200 |
| }, |
| { |
| "epoch": 0.05954743946010321, |
| "grad_norm": 1.6796875, |
| "learning_rate": 7.141949825341378e-05, |
| "loss": 0.2599, |
| "step": 2250 |
| }, |
| { |
| "epoch": 0.060870715892549956, |
| "grad_norm": 1.8515625, |
| "learning_rate": 7.30073039060019e-05, |
| "loss": 0.249, |
| "step": 2300 |
| }, |
| { |
| "epoch": 0.06219399232499669, |
| "grad_norm": 1.5, |
| "learning_rate": 7.459510955859002e-05, |
| "loss": 0.2474, |
| "step": 2350 |
| }, |
| { |
| "epoch": 0.06351726875744343, |
| "grad_norm": 1.5859375, |
| "learning_rate": 7.618291521117815e-05, |
| "loss": 0.2364, |
| "step": 2400 |
| }, |
| { |
| "epoch": 0.06484054518989017, |
| "grad_norm": 1.21875, |
| "learning_rate": 7.777072086376627e-05, |
| "loss": 0.2814, |
| "step": 2450 |
| }, |
| { |
| "epoch": 0.0661638216223369, |
| "grad_norm": 1.5859375, |
| "learning_rate": 7.93585265163544e-05, |
| "loss": 0.2427, |
| "step": 2500 |
| }, |
| { |
| "epoch": 0.06748709805478365, |
| "grad_norm": 1.6484375, |
| "learning_rate": 8.094633216894252e-05, |
| "loss": 0.2416, |
| "step": 2550 |
| }, |
| { |
| "epoch": 0.06881037448723039, |
| "grad_norm": 1.3984375, |
| "learning_rate": 8.253413782153064e-05, |
| "loss": 0.2402, |
| "step": 2600 |
| }, |
| { |
| "epoch": 0.07013365091967712, |
| "grad_norm": 1.5234375, |
| "learning_rate": 8.412194347411875e-05, |
| "loss": 0.2467, |
| "step": 2650 |
| }, |
| { |
| "epoch": 0.07145692735212386, |
| "grad_norm": 1.40625, |
| "learning_rate": 8.570974912670689e-05, |
| "loss": 0.232, |
| "step": 2700 |
| }, |
| { |
| "epoch": 0.07278020378457059, |
| "grad_norm": 1.359375, |
| "learning_rate": 8.7297554779295e-05, |
| "loss": 0.2449, |
| "step": 2750 |
| }, |
| { |
| "epoch": 0.07410348021701733, |
| "grad_norm": 1.28125, |
| "learning_rate": 8.888536043188314e-05, |
| "loss": 0.2173, |
| "step": 2800 |
| }, |
| { |
| "epoch": 0.07542675664946408, |
| "grad_norm": 1.3203125, |
| "learning_rate": 9.047316608447125e-05, |
| "loss": 0.2315, |
| "step": 2850 |
| }, |
| { |
| "epoch": 0.0767500330819108, |
| "grad_norm": 1.3984375, |
| "learning_rate": 9.206097173705939e-05, |
| "loss": 0.235, |
| "step": 2900 |
| }, |
| { |
| "epoch": 0.07807330951435755, |
| "grad_norm": 1.3359375, |
| "learning_rate": 9.36487773896475e-05, |
| "loss": 0.237, |
| "step": 2950 |
| }, |
| { |
| "epoch": 0.0793965859468043, |
| "grad_norm": 1.109375, |
| "learning_rate": 9.523658304223562e-05, |
| "loss": 0.232, |
| "step": 3000 |
| }, |
| { |
| "epoch": 0.0793965859468043, |
| "eval_cer": 0.05592713018964346, |
| "eval_loss": 0.3169604539871216, |
| "eval_runtime": 121.8922, |
| "eval_samples_per_second": 20.748, |
| "eval_steps_per_second": 0.164, |
| "eval_wer": 0.1594588681446908, |
| "step": 3000 |
| }, |
| { |
| "epoch": 0.08071986237925102, |
| "grad_norm": 1.2578125, |
| "learning_rate": 9.682438869482374e-05, |
| "loss": 0.2229, |
| "step": 3050 |
| }, |
| { |
| "epoch": 0.08204313881169777, |
| "grad_norm": 1.28125, |
| "learning_rate": 9.841219434741187e-05, |
| "loss": 0.2255, |
| "step": 3100 |
| }, |
| { |
| "epoch": 0.0833664152441445, |
| "grad_norm": 1.296875, |
| "learning_rate": 9.999999999999999e-05, |
| "loss": 0.2313, |
| "step": 3150 |
| }, |
| { |
| "epoch": 0.08468969167659124, |
| "grad_norm": 1.328125, |
| "learning_rate": 0.00010158780565258812, |
| "loss": 0.2704, |
| "step": 3200 |
| }, |
| { |
| "epoch": 0.08601296810903798, |
| "grad_norm": 1.3125, |
| "learning_rate": 0.00010317561130517623, |
| "loss": 0.228, |
| "step": 3250 |
| }, |
| { |
| "epoch": 0.08733624454148471, |
| "grad_norm": 1.421875, |
| "learning_rate": 0.00010476341695776436, |
| "loss": 0.2261, |
| "step": 3300 |
| }, |
| { |
| "epoch": 0.08865952097393145, |
| "grad_norm": 1.15625, |
| "learning_rate": 0.00010635122261035247, |
| "loss": 0.2303, |
| "step": 3350 |
| }, |
| { |
| "epoch": 0.0899827974063782, |
| "grad_norm": 1.296875, |
| "learning_rate": 0.00010793902826294061, |
| "loss": 0.231, |
| "step": 3400 |
| }, |
| { |
| "epoch": 0.09130607383882493, |
| "grad_norm": 1.25, |
| "learning_rate": 0.00010952683391552872, |
| "loss": 0.2282, |
| "step": 3450 |
| }, |
| { |
| "epoch": 0.09262935027127167, |
| "grad_norm": 1.109375, |
| "learning_rate": 0.00011111463956811686, |
| "loss": 0.2599, |
| "step": 3500 |
| }, |
| { |
| "epoch": 0.0939526267037184, |
| "grad_norm": 1.1015625, |
| "learning_rate": 0.00011270244522070497, |
| "loss": 0.2263, |
| "step": 3550 |
| }, |
| { |
| "epoch": 0.09527590313616514, |
| "grad_norm": 1.125, |
| "learning_rate": 0.0001142902508732931, |
| "loss": 0.2176, |
| "step": 3600 |
| }, |
| { |
| "epoch": 0.09659917956861189, |
| "grad_norm": 1.25, |
| "learning_rate": 0.00011587805652588123, |
| "loss": 0.2458, |
| "step": 3650 |
| }, |
| { |
| "epoch": 0.09792245600105862, |
| "grad_norm": 1.2265625, |
| "learning_rate": 0.00011746586217846934, |
| "loss": 0.229, |
| "step": 3700 |
| }, |
| { |
| "epoch": 0.09924573243350536, |
| "grad_norm": 1.0078125, |
| "learning_rate": 0.00011905366783105748, |
| "loss": 0.2172, |
| "step": 3750 |
| }, |
| { |
| "epoch": 0.1005690088659521, |
| "grad_norm": 1.2265625, |
| "learning_rate": 0.00012064147348364559, |
| "loss": 0.2444, |
| "step": 3800 |
| }, |
| { |
| "epoch": 0.10189228529839883, |
| "grad_norm": 1.046875, |
| "learning_rate": 0.00012222927913623372, |
| "loss": 0.2339, |
| "step": 3850 |
| }, |
| { |
| "epoch": 0.10321556173084558, |
| "grad_norm": 1.0390625, |
| "learning_rate": 0.00012381708478882185, |
| "loss": 0.2192, |
| "step": 3900 |
| }, |
| { |
| "epoch": 0.1045388381632923, |
| "grad_norm": 0.98046875, |
| "learning_rate": 0.00012540489044140997, |
| "loss": 0.222, |
| "step": 3950 |
| }, |
| { |
| "epoch": 0.10586211459573905, |
| "grad_norm": 1.046875, |
| "learning_rate": 0.00012699269609399807, |
| "loss": 0.2504, |
| "step": 4000 |
| }, |
| { |
| "epoch": 0.10718539102818579, |
| "grad_norm": 1.09375, |
| "learning_rate": 0.00012858050174658622, |
| "loss": 0.2471, |
| "step": 4050 |
| }, |
| { |
| "epoch": 0.10850866746063252, |
| "grad_norm": 1.4921875, |
| "learning_rate": 0.00013016830739917432, |
| "loss": 0.2208, |
| "step": 4100 |
| }, |
| { |
| "epoch": 0.10983194389307926, |
| "grad_norm": 1.1171875, |
| "learning_rate": 0.00013175611305176247, |
| "loss": 0.2317, |
| "step": 4150 |
| }, |
| { |
| "epoch": 0.11115522032552601, |
| "grad_norm": 1.15625, |
| "learning_rate": 0.00013334391870435056, |
| "loss": 0.2151, |
| "step": 4200 |
| }, |
| { |
| "epoch": 0.11247849675797274, |
| "grad_norm": 1.234375, |
| "learning_rate": 0.00013493172435693871, |
| "loss": 0.2224, |
| "step": 4250 |
| }, |
| { |
| "epoch": 0.11380177319041948, |
| "grad_norm": 1.453125, |
| "learning_rate": 0.0001365195300095268, |
| "loss": 0.2208, |
| "step": 4300 |
| }, |
| { |
| "epoch": 0.11512504962286621, |
| "grad_norm": 1.015625, |
| "learning_rate": 0.00013810733566211494, |
| "loss": 0.2482, |
| "step": 4350 |
| }, |
| { |
| "epoch": 0.11644832605531295, |
| "grad_norm": 0.94140625, |
| "learning_rate": 0.00013969514131470306, |
| "loss": 0.2133, |
| "step": 4400 |
| }, |
| { |
| "epoch": 0.1177716024877597, |
| "grad_norm": 1.0390625, |
| "learning_rate": 0.00014128294696729118, |
| "loss": 0.2217, |
| "step": 4450 |
| }, |
| { |
| "epoch": 0.11909487892020643, |
| "grad_norm": 1.046875, |
| "learning_rate": 0.0001428707526198793, |
| "loss": 0.2714, |
| "step": 4500 |
| }, |
| { |
| "epoch": 0.11909487892020643, |
| "eval_cer": 0.059526817954636074, |
| "eval_loss": 0.30962076783180237, |
| "eval_runtime": 110.9375, |
| "eval_samples_per_second": 22.797, |
| "eval_steps_per_second": 0.18, |
| "eval_wer": 0.15847432905484246, |
| "step": 4500 |
| }, |
| { |
| "epoch": 0.12041815535265317, |
| "grad_norm": 0.9453125, |
| "learning_rate": 0.00014445855827246743, |
| "loss": 0.2175, |
| "step": 4550 |
| }, |
| { |
| "epoch": 0.12174143178509991, |
| "grad_norm": 1.1171875, |
| "learning_rate": 0.00014604636392505555, |
| "loss": 0.2319, |
| "step": 4600 |
| }, |
| { |
| "epoch": 0.12306470821754664, |
| "grad_norm": 1.09375, |
| "learning_rate": 0.00014763416957764368, |
| "loss": 0.2312, |
| "step": 4650 |
| }, |
| { |
| "epoch": 0.12438798464999339, |
| "grad_norm": 1.234375, |
| "learning_rate": 0.0001492219752302318, |
| "loss": 0.2293, |
| "step": 4700 |
| }, |
| { |
| "epoch": 0.12571126108244013, |
| "grad_norm": 0.9453125, |
| "learning_rate": 0.00015080978088281993, |
| "loss": 0.2221, |
| "step": 4750 |
| }, |
| { |
| "epoch": 0.12703453751488686, |
| "grad_norm": 1.0078125, |
| "learning_rate": 0.00015239758653540805, |
| "loss": 0.2238, |
| "step": 4800 |
| }, |
| { |
| "epoch": 0.1283578139473336, |
| "grad_norm": 1.03125, |
| "learning_rate": 0.00015398539218799615, |
| "loss": 0.2354, |
| "step": 4850 |
| }, |
| { |
| "epoch": 0.12968109037978034, |
| "grad_norm": 0.90234375, |
| "learning_rate": 0.0001555731978405843, |
| "loss": 0.2111, |
| "step": 4900 |
| }, |
| { |
| "epoch": 0.13100436681222707, |
| "grad_norm": 0.95703125, |
| "learning_rate": 0.00015716100349317242, |
| "loss": 0.2164, |
| "step": 4950 |
| }, |
| { |
| "epoch": 0.1323276432446738, |
| "grad_norm": 1.046875, |
| "learning_rate": 0.00015874880914576054, |
| "loss": 0.2217, |
| "step": 5000 |
| }, |
| { |
| "epoch": 0.13365091967712056, |
| "grad_norm": 0.9375, |
| "learning_rate": 0.0001603366147983487, |
| "loss": 0.2148, |
| "step": 5050 |
| }, |
| { |
| "epoch": 0.1349741961095673, |
| "grad_norm": 0.96875, |
| "learning_rate": 0.0001619244204509368, |
| "loss": 0.2289, |
| "step": 5100 |
| }, |
| { |
| "epoch": 0.13629747254201402, |
| "grad_norm": 0.91796875, |
| "learning_rate": 0.00016351222610352492, |
| "loss": 0.2314, |
| "step": 5150 |
| }, |
| { |
| "epoch": 0.13762074897446078, |
| "grad_norm": 0.984375, |
| "learning_rate": 0.00016510003175611304, |
| "loss": 0.2169, |
| "step": 5200 |
| }, |
| { |
| "epoch": 0.1389440254069075, |
| "grad_norm": 0.90625, |
| "learning_rate": 0.00016668783740870116, |
| "loss": 0.2268, |
| "step": 5250 |
| }, |
| { |
| "epoch": 0.14026730183935424, |
| "grad_norm": 0.95703125, |
| "learning_rate": 0.0001682756430612893, |
| "loss": 0.2184, |
| "step": 5300 |
| }, |
| { |
| "epoch": 0.141590578271801, |
| "grad_norm": 0.828125, |
| "learning_rate": 0.0001698634487138774, |
| "loss": 0.2333, |
| "step": 5350 |
| }, |
| { |
| "epoch": 0.14291385470424772, |
| "grad_norm": 0.8203125, |
| "learning_rate": 0.0001714512543664655, |
| "loss": 0.2297, |
| "step": 5400 |
| }, |
| { |
| "epoch": 0.14423713113669445, |
| "grad_norm": 0.90625, |
| "learning_rate": 0.00017303906001905366, |
| "loss": 0.2156, |
| "step": 5450 |
| }, |
| { |
| "epoch": 0.14556040756914118, |
| "grad_norm": 0.80078125, |
| "learning_rate": 0.00017462686567164178, |
| "loss": 0.2253, |
| "step": 5500 |
| }, |
| { |
| "epoch": 0.14688368400158794, |
| "grad_norm": 0.921875, |
| "learning_rate": 0.0001762146713242299, |
| "loss": 0.2328, |
| "step": 5550 |
| }, |
| { |
| "epoch": 0.14820696043403467, |
| "grad_norm": 1.0078125, |
| "learning_rate": 0.000177802476976818, |
| "loss": 0.2157, |
| "step": 5600 |
| }, |
| { |
| "epoch": 0.1495302368664814, |
| "grad_norm": 1.0078125, |
| "learning_rate": 0.00017939028262940615, |
| "loss": 0.2118, |
| "step": 5650 |
| }, |
| { |
| "epoch": 0.15085351329892815, |
| "grad_norm": 0.8359375, |
| "learning_rate": 0.00018097808828199428, |
| "loss": 0.2223, |
| "step": 5700 |
| }, |
| { |
| "epoch": 0.15217678973137488, |
| "grad_norm": 0.83984375, |
| "learning_rate": 0.00018256589393458237, |
| "loss": 0.2301, |
| "step": 5750 |
| }, |
| { |
| "epoch": 0.1535000661638216, |
| "grad_norm": 0.70703125, |
| "learning_rate": 0.0001841536995871705, |
| "loss": 0.2152, |
| "step": 5800 |
| }, |
| { |
| "epoch": 0.15482334259626837, |
| "grad_norm": 1.03125, |
| "learning_rate": 0.00018574150523975865, |
| "loss": 0.2105, |
| "step": 5850 |
| }, |
| { |
| "epoch": 0.1561466190287151, |
| "grad_norm": 1.03125, |
| "learning_rate": 0.00018732931089234677, |
| "loss": 0.2686, |
| "step": 5900 |
| }, |
| { |
| "epoch": 0.15746989546116183, |
| "grad_norm": 0.859375, |
| "learning_rate": 0.00018891711654493487, |
| "loss": 0.2262, |
| "step": 5950 |
| }, |
| { |
| "epoch": 0.1587931718936086, |
| "grad_norm": 0.8046875, |
| "learning_rate": 0.000190504922197523, |
| "loss": 0.2176, |
| "step": 6000 |
| }, |
| { |
| "epoch": 0.1587931718936086, |
| "eval_cer": 0.05191187784647697, |
| "eval_loss": 0.3073718547821045, |
| "eval_runtime": 114.513, |
| "eval_samples_per_second": 22.085, |
| "eval_steps_per_second": 0.175, |
| "eval_wer": 0.1504521586931155, |
| "step": 6000 |
| }, |
| { |
| "epoch": 0.16011644832605532, |
| "grad_norm": 0.94140625, |
| "learning_rate": 0.00019209272785011114, |
| "loss": 0.2144, |
| "step": 6050 |
| }, |
| { |
| "epoch": 0.16143972475850205, |
| "grad_norm": 0.76171875, |
| "learning_rate": 0.00019368053350269924, |
| "loss": 0.2096, |
| "step": 6100 |
| }, |
| { |
| "epoch": 0.1627630011909488, |
| "grad_norm": 0.8359375, |
| "learning_rate": 0.00019526833915528737, |
| "loss": 0.2307, |
| "step": 6150 |
| }, |
| { |
| "epoch": 0.16408627762339553, |
| "grad_norm": 0.7734375, |
| "learning_rate": 0.0001968561448078755, |
| "loss": 0.2096, |
| "step": 6200 |
| }, |
| { |
| "epoch": 0.16540955405584226, |
| "grad_norm": 1.4140625, |
| "learning_rate": 0.00019844395046046364, |
| "loss": 0.2407, |
| "step": 6250 |
| }, |
| { |
| "epoch": 0.166732830488289, |
| "grad_norm": 0.98046875, |
| "learning_rate": 0.00020003175611305174, |
| "loss": 0.3033, |
| "step": 6300 |
| }, |
| { |
| "epoch": 0.16805610692073575, |
| "grad_norm": 0.91796875, |
| "learning_rate": 0.00020161956176563986, |
| "loss": 0.2261, |
| "step": 6350 |
| }, |
| { |
| "epoch": 0.16937938335318248, |
| "grad_norm": 0.796875, |
| "learning_rate": 0.000203207367418228, |
| "loss": 0.234, |
| "step": 6400 |
| }, |
| { |
| "epoch": 0.1707026597856292, |
| "grad_norm": 0.76171875, |
| "learning_rate": 0.0002047951730708161, |
| "loss": 0.2131, |
| "step": 6450 |
| }, |
| { |
| "epoch": 0.17202593621807596, |
| "grad_norm": 0.8984375, |
| "learning_rate": 0.00020638297872340423, |
| "loss": 0.2275, |
| "step": 6500 |
| }, |
| { |
| "epoch": 0.1733492126505227, |
| "grad_norm": 0.75, |
| "learning_rate": 0.00020797078437599236, |
| "loss": 0.2231, |
| "step": 6550 |
| }, |
| { |
| "epoch": 0.17467248908296942, |
| "grad_norm": 0.6796875, |
| "learning_rate": 0.0002095585900285805, |
| "loss": 0.2131, |
| "step": 6600 |
| }, |
| { |
| "epoch": 0.17599576551541618, |
| "grad_norm": 0.7421875, |
| "learning_rate": 0.0002111463956811686, |
| "loss": 0.2196, |
| "step": 6650 |
| }, |
| { |
| "epoch": 0.1773190419478629, |
| "grad_norm": 0.76171875, |
| "learning_rate": 0.00021273420133375673, |
| "loss": 0.2097, |
| "step": 6700 |
| }, |
| { |
| "epoch": 0.17864231838030964, |
| "grad_norm": 0.77734375, |
| "learning_rate": 0.00021432200698634485, |
| "loss": 0.2114, |
| "step": 6750 |
| }, |
| { |
| "epoch": 0.1799655948127564, |
| "grad_norm": 0.8125, |
| "learning_rate": 0.000215909812638933, |
| "loss": 0.2283, |
| "step": 6800 |
| }, |
| { |
| "epoch": 0.18128887124520313, |
| "grad_norm": 0.76953125, |
| "learning_rate": 0.0002174976182915211, |
| "loss": 0.2285, |
| "step": 6850 |
| }, |
| { |
| "epoch": 0.18261214767764986, |
| "grad_norm": 0.734375, |
| "learning_rate": 0.00021908542394410922, |
| "loss": 0.2159, |
| "step": 6900 |
| }, |
| { |
| "epoch": 0.1839354241100966, |
| "grad_norm": 0.80078125, |
| "learning_rate": 0.00022067322959669732, |
| "loss": 0.2125, |
| "step": 6950 |
| }, |
| { |
| "epoch": 0.18525870054254334, |
| "grad_norm": 0.75390625, |
| "learning_rate": 0.00022226103524928547, |
| "loss": 0.2028, |
| "step": 7000 |
| }, |
| { |
| "epoch": 0.18658197697499007, |
| "grad_norm": 0.9453125, |
| "learning_rate": 0.0002238488409018736, |
| "loss": 0.2058, |
| "step": 7050 |
| }, |
| { |
| "epoch": 0.1879052534074368, |
| "grad_norm": 0.67578125, |
| "learning_rate": 0.00022543664655446172, |
| "loss": 0.2004, |
| "step": 7100 |
| }, |
| { |
| "epoch": 0.18922852983988356, |
| "grad_norm": 0.671875, |
| "learning_rate": 0.00022702445220704981, |
| "loss": 0.2159, |
| "step": 7150 |
| }, |
| { |
| "epoch": 0.1905518062723303, |
| "grad_norm": 0.76953125, |
| "learning_rate": 0.00022861225785963797, |
| "loss": 0.2102, |
| "step": 7200 |
| }, |
| { |
| "epoch": 0.19187508270477702, |
| "grad_norm": 0.75, |
| "learning_rate": 0.0002302000635122261, |
| "loss": 0.2061, |
| "step": 7250 |
| }, |
| { |
| "epoch": 0.19319835913722377, |
| "grad_norm": 0.84765625, |
| "learning_rate": 0.0002317878691648142, |
| "loss": 0.2354, |
| "step": 7300 |
| }, |
| { |
| "epoch": 0.1945216355696705, |
| "grad_norm": 0.78125, |
| "learning_rate": 0.0002333756748174023, |
| "loss": 0.2471, |
| "step": 7350 |
| }, |
| { |
| "epoch": 0.19584491200211723, |
| "grad_norm": 0.80859375, |
| "learning_rate": 0.00023496348046999046, |
| "loss": 0.2141, |
| "step": 7400 |
| }, |
| { |
| "epoch": 0.197168188434564, |
| "grad_norm": 0.75390625, |
| "learning_rate": 0.00023655128612257858, |
| "loss": 0.2106, |
| "step": 7450 |
| }, |
| { |
| "epoch": 0.19849146486701072, |
| "grad_norm": 0.8125, |
| "learning_rate": 0.00023813909177516668, |
| "loss": 0.2073, |
| "step": 7500 |
| }, |
| { |
| "epoch": 0.19849146486701072, |
| "eval_cer": 0.05027208248395303, |
| "eval_loss": 0.2880328893661499, |
| "eval_runtime": 114.539, |
| "eval_samples_per_second": 22.08, |
| "eval_steps_per_second": 0.175, |
| "eval_wer": 0.1450554259043174, |
| "step": 7500 |
| }, |
| { |
| "epoch": 0.19981474129945745, |
| "grad_norm": 0.734375, |
| "learning_rate": 0.00023972689742775483, |
| "loss": 0.1975, |
| "step": 7550 |
| }, |
| { |
| "epoch": 0.2011380177319042, |
| "grad_norm": 0.8125, |
| "learning_rate": 0.00024131470308034296, |
| "loss": 0.2048, |
| "step": 7600 |
| }, |
| { |
| "epoch": 0.20246129416435094, |
| "grad_norm": 0.76953125, |
| "learning_rate": 0.00024290250873293108, |
| "loss": 0.2097, |
| "step": 7650 |
| }, |
| { |
| "epoch": 0.20378457059679767, |
| "grad_norm": 0.7890625, |
| "learning_rate": 0.0002444903143855192, |
| "loss": 0.2029, |
| "step": 7700 |
| }, |
| { |
| "epoch": 0.20510784702924442, |
| "grad_norm": 0.8046875, |
| "learning_rate": 0.00024607812003810733, |
| "loss": 0.2114, |
| "step": 7750 |
| }, |
| { |
| "epoch": 0.20643112346169115, |
| "grad_norm": 0.80859375, |
| "learning_rate": 0.0002476659256906954, |
| "loss": 0.2223, |
| "step": 7800 |
| }, |
| { |
| "epoch": 0.20775439989413788, |
| "grad_norm": 0.828125, |
| "learning_rate": 0.0002492537313432836, |
| "loss": 0.2074, |
| "step": 7850 |
| }, |
| { |
| "epoch": 0.2090776763265846, |
| "grad_norm": 0.671875, |
| "learning_rate": 0.00025084153699587167, |
| "loss": 0.2023, |
| "step": 7900 |
| }, |
| { |
| "epoch": 0.21040095275903137, |
| "grad_norm": 1.078125, |
| "learning_rate": 0.0002524293426484598, |
| "loss": 0.2184, |
| "step": 7950 |
| }, |
| { |
| "epoch": 0.2117242291914781, |
| "grad_norm": 0.70703125, |
| "learning_rate": 0.0002540171483010479, |
| "loss": 0.2055, |
| "step": 8000 |
| }, |
| { |
| "epoch": 0.21304750562392483, |
| "grad_norm": 0.921875, |
| "learning_rate": 0.00025560495395363607, |
| "loss": 0.2169, |
| "step": 8050 |
| }, |
| { |
| "epoch": 0.21437078205637158, |
| "grad_norm": 0.8046875, |
| "learning_rate": 0.00025719275960622417, |
| "loss": 0.2325, |
| "step": 8100 |
| }, |
| { |
| "epoch": 0.2156940584888183, |
| "grad_norm": 0.82421875, |
| "learning_rate": 0.0002587805652588123, |
| "loss": 0.1956, |
| "step": 8150 |
| }, |
| { |
| "epoch": 0.21701733492126504, |
| "grad_norm": 0.7578125, |
| "learning_rate": 0.0002603683709114004, |
| "loss": 0.2345, |
| "step": 8200 |
| }, |
| { |
| "epoch": 0.2183406113537118, |
| "grad_norm": 0.80859375, |
| "learning_rate": 0.00026195617656398857, |
| "loss": 0.203, |
| "step": 8250 |
| }, |
| { |
| "epoch": 0.21966388778615853, |
| "grad_norm": 0.67578125, |
| "learning_rate": 0.00026354398221657666, |
| "loss": 0.2008, |
| "step": 8300 |
| }, |
| { |
| "epoch": 0.22098716421860526, |
| "grad_norm": 0.77734375, |
| "learning_rate": 0.0002651317878691648, |
| "loss": 0.2265, |
| "step": 8350 |
| }, |
| { |
| "epoch": 0.22231044065105202, |
| "grad_norm": 0.72265625, |
| "learning_rate": 0.0002667195935217529, |
| "loss": 0.2012, |
| "step": 8400 |
| }, |
| { |
| "epoch": 0.22363371708349875, |
| "grad_norm": 0.76953125, |
| "learning_rate": 0.000268307399174341, |
| "loss": 0.2045, |
| "step": 8450 |
| }, |
| { |
| "epoch": 0.22495699351594547, |
| "grad_norm": 0.64453125, |
| "learning_rate": 0.00026989520482692916, |
| "loss": 0.2057, |
| "step": 8500 |
| }, |
| { |
| "epoch": 0.22628026994839223, |
| "grad_norm": 0.828125, |
| "learning_rate": 0.0002714830104795173, |
| "loss": 0.2172, |
| "step": 8550 |
| }, |
| { |
| "epoch": 0.22760354638083896, |
| "grad_norm": 0.88671875, |
| "learning_rate": 0.0002730708161321054, |
| "loss": 0.2137, |
| "step": 8600 |
| }, |
| { |
| "epoch": 0.2289268228132857, |
| "grad_norm": 0.6640625, |
| "learning_rate": 0.0002746586217846935, |
| "loss": 0.2021, |
| "step": 8650 |
| }, |
| { |
| "epoch": 0.23025009924573242, |
| "grad_norm": 1.0390625, |
| "learning_rate": 0.00027624642743728165, |
| "loss": 0.2064, |
| "step": 8700 |
| }, |
| { |
| "epoch": 0.23157337567817918, |
| "grad_norm": 0.74609375, |
| "learning_rate": 0.0002778342330898698, |
| "loss": 0.2082, |
| "step": 8750 |
| }, |
| { |
| "epoch": 0.2328966521106259, |
| "grad_norm": 1.3046875, |
| "learning_rate": 0.0002794220387424579, |
| "loss": 0.2139, |
| "step": 8800 |
| }, |
| { |
| "epoch": 0.23421992854307264, |
| "grad_norm": 0.7890625, |
| "learning_rate": 0.000281009844395046, |
| "loss": 0.2009, |
| "step": 8850 |
| }, |
| { |
| "epoch": 0.2355432049755194, |
| "grad_norm": 0.6875, |
| "learning_rate": 0.00028259765004763415, |
| "loss": 0.2015, |
| "step": 8900 |
| }, |
| { |
| "epoch": 0.23686648140796612, |
| "grad_norm": 0.69140625, |
| "learning_rate": 0.0002841854557002223, |
| "loss": 0.2006, |
| "step": 8950 |
| }, |
| { |
| "epoch": 0.23818975784041285, |
| "grad_norm": 0.6640625, |
| "learning_rate": 0.0002857732613528104, |
| "loss": 0.2276, |
| "step": 9000 |
| }, |
| { |
| "epoch": 0.23818975784041285, |
| "eval_cer": 0.05210842866047812, |
| "eval_loss": 0.2827185392379761, |
| "eval_runtime": 111.4286, |
| "eval_samples_per_second": 22.696, |
| "eval_steps_per_second": 0.179, |
| "eval_wer": 0.1450554259043174, |
| "step": 9000 |
| }, |
| { |
| "epoch": 0.2395130342728596, |
| "grad_norm": 0.6484375, |
| "learning_rate": 0.0002873610670053985, |
| "loss": 0.192, |
| "step": 9050 |
| }, |
| { |
| "epoch": 0.24083631070530634, |
| "grad_norm": 0.6328125, |
| "learning_rate": 0.00028894887265798664, |
| "loss": 0.2112, |
| "step": 9100 |
| }, |
| { |
| "epoch": 0.24215958713775307, |
| "grad_norm": 0.671875, |
| "learning_rate": 0.00029053667831057474, |
| "loss": 0.2005, |
| "step": 9150 |
| }, |
| { |
| "epoch": 0.24348286357019983, |
| "grad_norm": 0.71875, |
| "learning_rate": 0.0002921244839631629, |
| "loss": 0.2092, |
| "step": 9200 |
| }, |
| { |
| "epoch": 0.24480614000264656, |
| "grad_norm": 0.66015625, |
| "learning_rate": 0.000293712289615751, |
| "loss": 0.2021, |
| "step": 9250 |
| }, |
| { |
| "epoch": 0.24612941643509328, |
| "grad_norm": 0.6875, |
| "learning_rate": 0.00029530009526833914, |
| "loss": 0.2614, |
| "step": 9300 |
| }, |
| { |
| "epoch": 0.24745269286754004, |
| "grad_norm": 0.671875, |
| "learning_rate": 0.00029688790092092723, |
| "loss": 0.2146, |
| "step": 9350 |
| }, |
| { |
| "epoch": 0.24877596929998677, |
| "grad_norm": 0.62109375, |
| "learning_rate": 0.0002984757065735154, |
| "loss": 0.2077, |
| "step": 9400 |
| }, |
| { |
| "epoch": 0.25009924573243353, |
| "grad_norm": 0.96484375, |
| "learning_rate": 0.0002999999999080824, |
| "loss": 0.2342, |
| "step": 9450 |
| }, |
| { |
| "epoch": 0.25142252216488026, |
| "grad_norm": 0.65234375, |
| "learning_rate": 0.00029999993786372365, |
| "loss": 0.221, |
| "step": 9500 |
| }, |
| { |
| "epoch": 0.252745798597327, |
| "grad_norm": 0.62890625, |
| "learning_rate": 0.0002999997609224513, |
| "loss": 0.1937, |
| "step": 9550 |
| }, |
| { |
| "epoch": 0.2540690750297737, |
| "grad_norm": 1.1484375, |
| "learning_rate": 0.0002999994690844009, |
| "loss": 0.2232, |
| "step": 9600 |
| }, |
| { |
| "epoch": 0.25539235146222045, |
| "grad_norm": 0.6953125, |
| "learning_rate": 0.0002999990623497959, |
| "loss": 0.2037, |
| "step": 9650 |
| }, |
| { |
| "epoch": 0.2567156278946672, |
| "grad_norm": 0.67578125, |
| "learning_rate": 0.00029999854071894786, |
| "loss": 0.1957, |
| "step": 9700 |
| }, |
| { |
| "epoch": 0.25803890432711396, |
| "grad_norm": 0.8046875, |
| "learning_rate": 0.0002999979041922564, |
| "loss": 0.2061, |
| "step": 9750 |
| }, |
| { |
| "epoch": 0.2593621807595607, |
| "grad_norm": 0.7734375, |
| "learning_rate": 0.0002999971527702091, |
| "loss": 0.2019, |
| "step": 9800 |
| }, |
| { |
| "epoch": 0.2606854571920074, |
| "grad_norm": 0.7265625, |
| "learning_rate": 0.00029999628645338154, |
| "loss": 0.2129, |
| "step": 9850 |
| }, |
| { |
| "epoch": 0.26200873362445415, |
| "grad_norm": 0.625, |
| "learning_rate": 0.0002999953052424372, |
| "loss": 0.1945, |
| "step": 9900 |
| }, |
| { |
| "epoch": 0.2633320100569009, |
| "grad_norm": 0.6875, |
| "learning_rate": 0.00029999420913812776, |
| "loss": 0.2055, |
| "step": 9950 |
| }, |
| { |
| "epoch": 0.2646552864893476, |
| "grad_norm": 0.71484375, |
| "learning_rate": 0.0002999929981412927, |
| "loss": 0.1968, |
| "step": 10000 |
| }, |
| { |
| "epoch": 0.26597856292179434, |
| "grad_norm": 0.6015625, |
| "learning_rate": 0.0002999916722528598, |
| "loss": 0.2005, |
| "step": 10050 |
| }, |
| { |
| "epoch": 0.2673018393542411, |
| "grad_norm": 0.70703125, |
| "learning_rate": 0.0002999902314738445, |
| "loss": 0.2064, |
| "step": 10100 |
| }, |
| { |
| "epoch": 0.26862511578668785, |
| "grad_norm": 0.69921875, |
| "learning_rate": 0.0002999886758053506, |
| "loss": 0.2014, |
| "step": 10150 |
| }, |
| { |
| "epoch": 0.2699483922191346, |
| "grad_norm": 0.59375, |
| "learning_rate": 0.00029998700524856947, |
| "loss": 0.1872, |
| "step": 10200 |
| }, |
| { |
| "epoch": 0.2712716686515813, |
| "grad_norm": 0.65234375, |
| "learning_rate": 0.00029998521980478083, |
| "loss": 0.1846, |
| "step": 10250 |
| }, |
| { |
| "epoch": 0.27259494508402804, |
| "grad_norm": 0.67578125, |
| "learning_rate": 0.0002999833194753523, |
| "loss": 0.1961, |
| "step": 10300 |
| }, |
| { |
| "epoch": 0.27391822151647477, |
| "grad_norm": 0.65625, |
| "learning_rate": 0.00029998130426173956, |
| "loss": 0.1887, |
| "step": 10350 |
| }, |
| { |
| "epoch": 0.27524149794892155, |
| "grad_norm": 0.58203125, |
| "learning_rate": 0.0002999791741654861, |
| "loss": 0.1912, |
| "step": 10400 |
| }, |
| { |
| "epoch": 0.2765647743813683, |
| "grad_norm": 0.76171875, |
| "learning_rate": 0.0002999769291882236, |
| "loss": 0.2197, |
| "step": 10450 |
| }, |
| { |
| "epoch": 0.277888050813815, |
| "grad_norm": 0.68359375, |
| "learning_rate": 0.0002999745693316717, |
| "loss": 0.2021, |
| "step": 10500 |
| }, |
| { |
| "epoch": 0.277888050813815, |
| "eval_cer": 0.06062750251304255, |
| "eval_loss": 0.2784423530101776, |
| "eval_runtime": 172.9503, |
| "eval_samples_per_second": 14.623, |
| "eval_steps_per_second": 0.116, |
| "eval_wer": 0.1528588098016336, |
| "step": 10500 |
| }, |
| { |
| "epoch": 0.27921132724626174, |
| "grad_norm": 0.62109375, |
| "learning_rate": 0.0002999720945976379, |
| "loss": 0.2076, |
| "step": 10550 |
| }, |
| { |
| "epoch": 0.28053460367870847, |
| "grad_norm": 0.5546875, |
| "learning_rate": 0.00029996950498801785, |
| "loss": 0.2289, |
| "step": 10600 |
| }, |
| { |
| "epoch": 0.2818578801111552, |
| "grad_norm": 0.8125, |
| "learning_rate": 0.00029996680050479514, |
| "loss": 0.1823, |
| "step": 10650 |
| }, |
| { |
| "epoch": 0.283181156543602, |
| "grad_norm": 0.609375, |
| "learning_rate": 0.00029996398115004135, |
| "loss": 0.2035, |
| "step": 10700 |
| }, |
| { |
| "epoch": 0.2845044329760487, |
| "grad_norm": 0.55078125, |
| "learning_rate": 0.0002999610469259161, |
| "loss": 0.1845, |
| "step": 10750 |
| }, |
| { |
| "epoch": 0.28582770940849545, |
| "grad_norm": 0.62890625, |
| "learning_rate": 0.00029995799783466684, |
| "loss": 0.1911, |
| "step": 10800 |
| }, |
| { |
| "epoch": 0.2871509858409422, |
| "grad_norm": 0.5546875, |
| "learning_rate": 0.00029995483387862924, |
| "loss": 0.1996, |
| "step": 10850 |
| }, |
| { |
| "epoch": 0.2884742622733889, |
| "grad_norm": 0.765625, |
| "learning_rate": 0.0002999515550602267, |
| "loss": 0.1826, |
| "step": 10900 |
| }, |
| { |
| "epoch": 0.28979753870583563, |
| "grad_norm": 0.76171875, |
| "learning_rate": 0.00029994816138197074, |
| "loss": 0.1857, |
| "step": 10950 |
| }, |
| { |
| "epoch": 0.29112081513828236, |
| "grad_norm": 0.625, |
| "learning_rate": 0.000299944652846461, |
| "loss": 0.1872, |
| "step": 11000 |
| }, |
| { |
| "epoch": 0.29244409157072915, |
| "grad_norm": 0.66796875, |
| "learning_rate": 0.0002999410294563847, |
| "loss": 0.1938, |
| "step": 11050 |
| }, |
| { |
| "epoch": 0.2937673680031759, |
| "grad_norm": 0.73828125, |
| "learning_rate": 0.0002999372912145175, |
| "loss": 0.1905, |
| "step": 11100 |
| }, |
| { |
| "epoch": 0.2950906444356226, |
| "grad_norm": 0.8671875, |
| "learning_rate": 0.00029993343812372274, |
| "loss": 0.1967, |
| "step": 11150 |
| }, |
| { |
| "epoch": 0.29641392086806934, |
| "grad_norm": 0.59765625, |
| "learning_rate": 0.00029992947018695187, |
| "loss": 0.1769, |
| "step": 11200 |
| }, |
| { |
| "epoch": 0.29773719730051607, |
| "grad_norm": 0.61328125, |
| "learning_rate": 0.0002999253874072441, |
| "loss": 0.1993, |
| "step": 11250 |
| }, |
| { |
| "epoch": 0.2990604737329628, |
| "grad_norm": 0.578125, |
| "learning_rate": 0.0002999211897877269, |
| "loss": 0.1774, |
| "step": 11300 |
| }, |
| { |
| "epoch": 0.3003837501654096, |
| "grad_norm": 0.59765625, |
| "learning_rate": 0.00029991687733161546, |
| "loss": 0.1969, |
| "step": 11350 |
| }, |
| { |
| "epoch": 0.3017070265978563, |
| "grad_norm": 0.6171875, |
| "learning_rate": 0.0002999124500422131, |
| "loss": 0.1819, |
| "step": 11400 |
| }, |
| { |
| "epoch": 0.30303030303030304, |
| "grad_norm": 0.671875, |
| "learning_rate": 0.00029990790792291103, |
| "loss": 0.1726, |
| "step": 11450 |
| }, |
| { |
| "epoch": 0.30435357946274977, |
| "grad_norm": 0.6015625, |
| "learning_rate": 0.0002999032509771884, |
| "loss": 0.1836, |
| "step": 11500 |
| }, |
| { |
| "epoch": 0.3056768558951965, |
| "grad_norm": 0.55859375, |
| "learning_rate": 0.0002998984792086123, |
| "loss": 0.1972, |
| "step": 11550 |
| }, |
| { |
| "epoch": 0.3070001323276432, |
| "grad_norm": 0.71484375, |
| "learning_rate": 0.0002998935926208379, |
| "loss": 0.1854, |
| "step": 11600 |
| }, |
| { |
| "epoch": 0.30832340876008996, |
| "grad_norm": 0.7265625, |
| "learning_rate": 0.0002998885912176081, |
| "loss": 0.1767, |
| "step": 11650 |
| }, |
| { |
| "epoch": 0.30964668519253674, |
| "grad_norm": 0.578125, |
| "learning_rate": 0.000299883475002754, |
| "loss": 0.1745, |
| "step": 11700 |
| }, |
| { |
| "epoch": 0.31096996162498347, |
| "grad_norm": 0.62890625, |
| "learning_rate": 0.0002998782439801945, |
| "loss": 0.1822, |
| "step": 11750 |
| }, |
| { |
| "epoch": 0.3122932380574302, |
| "grad_norm": 0.63671875, |
| "learning_rate": 0.00029987289815393633, |
| "loss": 0.1706, |
| "step": 11800 |
| }, |
| { |
| "epoch": 0.31361651448987693, |
| "grad_norm": 0.546875, |
| "learning_rate": 0.0002998674375280745, |
| "loss": 0.179, |
| "step": 11850 |
| }, |
| { |
| "epoch": 0.31493979092232366, |
| "grad_norm": 0.486328125, |
| "learning_rate": 0.00029986186210679156, |
| "loss": 0.174, |
| "step": 11900 |
| }, |
| { |
| "epoch": 0.3162630673547704, |
| "grad_norm": 0.66796875, |
| "learning_rate": 0.00029985617189435827, |
| "loss": 0.1769, |
| "step": 11950 |
| }, |
| { |
| "epoch": 0.3175863437872172, |
| "grad_norm": 0.53515625, |
| "learning_rate": 0.0002998503668951331, |
| "loss": 0.1696, |
| "step": 12000 |
| }, |
| { |
| "epoch": 0.3175863437872172, |
| "eval_cer": 0.049974448394179846, |
| "eval_loss": 0.25553205609321594, |
| "eval_runtime": 121.3958, |
| "eval_samples_per_second": 20.833, |
| "eval_steps_per_second": 0.165, |
| "eval_wer": 0.13353267211201866, |
| "step": 12000 |
| }, |
| { |
| "epoch": 0.3189096202196639, |
| "grad_norm": 0.625, |
| "learning_rate": 0.00029984444711356266, |
| "loss": 0.1806, |
| "step": 12050 |
| }, |
| { |
| "epoch": 0.32023289665211063, |
| "grad_norm": 0.640625, |
| "learning_rate": 0.0002998384125541814, |
| "loss": 0.1838, |
| "step": 12100 |
| }, |
| { |
| "epoch": 0.32155617308455736, |
| "grad_norm": 0.47265625, |
| "learning_rate": 0.0002998322632216116, |
| "loss": 0.1843, |
| "step": 12150 |
| }, |
| { |
| "epoch": 0.3228794495170041, |
| "grad_norm": 0.6953125, |
| "learning_rate": 0.0002998259991205635, |
| "loss": 0.1808, |
| "step": 12200 |
| }, |
| { |
| "epoch": 0.3242027259494508, |
| "grad_norm": 0.51953125, |
| "learning_rate": 0.0002998196202558354, |
| "loss": 0.1796, |
| "step": 12250 |
| }, |
| { |
| "epoch": 0.3255260023818976, |
| "grad_norm": 0.53125, |
| "learning_rate": 0.0002998131266323133, |
| "loss": 0.1744, |
| "step": 12300 |
| }, |
| { |
| "epoch": 0.32684927881434434, |
| "grad_norm": 0.6015625, |
| "learning_rate": 0.0002998065182549712, |
| "loss": 0.1725, |
| "step": 12350 |
| }, |
| { |
| "epoch": 0.32817255524679106, |
| "grad_norm": 0.59765625, |
| "learning_rate": 0.00029979979512887096, |
| "loss": 0.179, |
| "step": 12400 |
| }, |
| { |
| "epoch": 0.3294958316792378, |
| "grad_norm": 0.578125, |
| "learning_rate": 0.0002997929572591624, |
| "loss": 0.1667, |
| "step": 12450 |
| }, |
| { |
| "epoch": 0.3308191081116845, |
| "grad_norm": 0.6015625, |
| "learning_rate": 0.00029978600465108314, |
| "loss": 0.1695, |
| "step": 12500 |
| }, |
| { |
| "epoch": 0.33214238454413125, |
| "grad_norm": 0.6953125, |
| "learning_rate": 0.00029977893730995876, |
| "loss": 0.1863, |
| "step": 12550 |
| }, |
| { |
| "epoch": 0.333465660976578, |
| "grad_norm": 0.5078125, |
| "learning_rate": 0.0002997717552412027, |
| "loss": 0.1763, |
| "step": 12600 |
| }, |
| { |
| "epoch": 0.33478893740902477, |
| "grad_norm": 0.6015625, |
| "learning_rate": 0.0002997644584503163, |
| "loss": 0.1765, |
| "step": 12650 |
| }, |
| { |
| "epoch": 0.3361122138414715, |
| "grad_norm": 0.62890625, |
| "learning_rate": 0.0002997570469428887, |
| "loss": 0.1882, |
| "step": 12700 |
| }, |
| { |
| "epoch": 0.3374354902739182, |
| "grad_norm": 0.52734375, |
| "learning_rate": 0.00029974952072459705, |
| "loss": 0.1783, |
| "step": 12750 |
| }, |
| { |
| "epoch": 0.33875876670636496, |
| "grad_norm": 0.5546875, |
| "learning_rate": 0.0002997418798012062, |
| "loss": 0.1825, |
| "step": 12800 |
| }, |
| { |
| "epoch": 0.3400820431388117, |
| "grad_norm": 0.65625, |
| "learning_rate": 0.00029973412417856897, |
| "loss": 0.1803, |
| "step": 12850 |
| }, |
| { |
| "epoch": 0.3414053195712584, |
| "grad_norm": 0.671875, |
| "learning_rate": 0.000299726253862626, |
| "loss": 0.1644, |
| "step": 12900 |
| }, |
| { |
| "epoch": 0.3427285960037052, |
| "grad_norm": 0.6015625, |
| "learning_rate": 0.00029971826885940585, |
| "loss": 0.1705, |
| "step": 12950 |
| }, |
| { |
| "epoch": 0.34405187243615193, |
| "grad_norm": 0.55078125, |
| "learning_rate": 0.0002997101691750248, |
| "loss": 0.1676, |
| "step": 13000 |
| }, |
| { |
| "epoch": 0.34537514886859866, |
| "grad_norm": 0.51171875, |
| "learning_rate": 0.00029970195481568706, |
| "loss": 0.1611, |
| "step": 13050 |
| }, |
| { |
| "epoch": 0.3466984253010454, |
| "grad_norm": 0.6328125, |
| "learning_rate": 0.0002996936257876847, |
| "loss": 0.1686, |
| "step": 13100 |
| }, |
| { |
| "epoch": 0.3480217017334921, |
| "grad_norm": 0.55078125, |
| "learning_rate": 0.00029968518209739755, |
| "loss": 0.1608, |
| "step": 13150 |
| }, |
| { |
| "epoch": 0.34934497816593885, |
| "grad_norm": 0.625, |
| "learning_rate": 0.0002996766237512933, |
| "loss": 0.1686, |
| "step": 13200 |
| }, |
| { |
| "epoch": 0.3506682545983856, |
| "grad_norm": 0.58203125, |
| "learning_rate": 0.00029966795075592753, |
| "loss": 0.1846, |
| "step": 13250 |
| }, |
| { |
| "epoch": 0.35199153103083236, |
| "grad_norm": 0.59765625, |
| "learning_rate": 0.0002996591631179435, |
| "loss": 0.17, |
| "step": 13300 |
| }, |
| { |
| "epoch": 0.3533148074632791, |
| "grad_norm": 0.52734375, |
| "learning_rate": 0.0002996502608440724, |
| "loss": 0.1615, |
| "step": 13350 |
| }, |
| { |
| "epoch": 0.3546380838957258, |
| "grad_norm": 0.52734375, |
| "learning_rate": 0.0002996412439411332, |
| "loss": 0.1666, |
| "step": 13400 |
| }, |
| { |
| "epoch": 0.35596136032817255, |
| "grad_norm": 1.0234375, |
| "learning_rate": 0.0002996321124160326, |
| "loss": 0.1908, |
| "step": 13450 |
| }, |
| { |
| "epoch": 0.3572846367606193, |
| "grad_norm": 0.49609375, |
| "learning_rate": 0.0002996228662757653, |
| "loss": 0.1604, |
| "step": 13500 |
| }, |
| { |
| "epoch": 0.3572846367606193, |
| "eval_cer": 0.052984483717169, |
| "eval_loss": 0.2499123215675354, |
| "eval_runtime": 129.0848, |
| "eval_samples_per_second": 19.592, |
| "eval_steps_per_second": 0.155, |
| "eval_wer": 0.13105309218203035, |
| "step": 13500 |
| }, |
| { |
| "epoch": 0.358607913193066, |
| "grad_norm": 0.55078125, |
| "learning_rate": 0.0002996135055274135, |
| "loss": 0.1627, |
| "step": 13550 |
| }, |
| { |
| "epoch": 0.3599311896255128, |
| "grad_norm": 0.51953125, |
| "learning_rate": 0.00029960403017814744, |
| "loss": 0.192, |
| "step": 13600 |
| }, |
| { |
| "epoch": 0.3612544660579595, |
| "grad_norm": 0.6484375, |
| "learning_rate": 0.00029959444023522504, |
| "loss": 0.1543, |
| "step": 13650 |
| }, |
| { |
| "epoch": 0.36257774249040625, |
| "grad_norm": 0.52734375, |
| "learning_rate": 0.000299584735705992, |
| "loss": 0.1754, |
| "step": 13700 |
| }, |
| { |
| "epoch": 0.363901018922853, |
| "grad_norm": 0.62890625, |
| "learning_rate": 0.00029957491659788173, |
| "loss": 0.175, |
| "step": 13750 |
| }, |
| { |
| "epoch": 0.3652242953552997, |
| "grad_norm": 0.490234375, |
| "learning_rate": 0.0002995649829184155, |
| "loss": 0.1673, |
| "step": 13800 |
| }, |
| { |
| "epoch": 0.36654757178774644, |
| "grad_norm": 0.515625, |
| "learning_rate": 0.00029955493467520236, |
| "loss": 0.162, |
| "step": 13850 |
| }, |
| { |
| "epoch": 0.3678708482201932, |
| "grad_norm": 0.5859375, |
| "learning_rate": 0.00029954477187593904, |
| "loss": 0.1672, |
| "step": 13900 |
| }, |
| { |
| "epoch": 0.36919412465263995, |
| "grad_norm": 0.53125, |
| "learning_rate": 0.00029953449452840993, |
| "loss": 0.1821, |
| "step": 13950 |
| }, |
| { |
| "epoch": 0.3705174010850867, |
| "grad_norm": 0.57421875, |
| "learning_rate": 0.0002995241026404875, |
| "loss": 0.1619, |
| "step": 14000 |
| }, |
| { |
| "epoch": 0.3718406775175334, |
| "grad_norm": 0.498046875, |
| "learning_rate": 0.0002995135962201315, |
| "loss": 0.1708, |
| "step": 14050 |
| }, |
| { |
| "epoch": 0.37316395394998014, |
| "grad_norm": 0.53125, |
| "learning_rate": 0.0002995029752753897, |
| "loss": 0.1717, |
| "step": 14100 |
| }, |
| { |
| "epoch": 0.37448723038242687, |
| "grad_norm": 0.58203125, |
| "learning_rate": 0.00029949223981439756, |
| "loss": 0.1787, |
| "step": 14150 |
| }, |
| { |
| "epoch": 0.3758105068148736, |
| "grad_norm": 0.52734375, |
| "learning_rate": 0.0002994813898453783, |
| "loss": 0.158, |
| "step": 14200 |
| }, |
| { |
| "epoch": 0.3771337832473204, |
| "grad_norm": 0.46484375, |
| "learning_rate": 0.00029947042537664263, |
| "loss": 0.1773, |
| "step": 14250 |
| }, |
| { |
| "epoch": 0.3784570596797671, |
| "grad_norm": 0.4921875, |
| "learning_rate": 0.0002994593464165891, |
| "loss": 0.1711, |
| "step": 14300 |
| }, |
| { |
| "epoch": 0.37978033611221385, |
| "grad_norm": 0.53515625, |
| "learning_rate": 0.00029944815297370414, |
| "loss": 0.1647, |
| "step": 14350 |
| }, |
| { |
| "epoch": 0.3811036125446606, |
| "grad_norm": 0.640625, |
| "learning_rate": 0.0002994368450565616, |
| "loss": 0.1713, |
| "step": 14400 |
| }, |
| { |
| "epoch": 0.3824268889771073, |
| "grad_norm": 0.515625, |
| "learning_rate": 0.00029942542267382305, |
| "loss": 0.1574, |
| "step": 14450 |
| }, |
| { |
| "epoch": 0.38375016540955403, |
| "grad_norm": 0.69140625, |
| "learning_rate": 0.00029941388583423794, |
| "loss": 0.1604, |
| "step": 14500 |
| }, |
| { |
| "epoch": 0.3850734418420008, |
| "grad_norm": 0.50390625, |
| "learning_rate": 0.0002994022345466432, |
| "loss": 0.166, |
| "step": 14550 |
| }, |
| { |
| "epoch": 0.38639671827444755, |
| "grad_norm": 0.455078125, |
| "learning_rate": 0.00029939046881996345, |
| "loss": 0.1752, |
| "step": 14600 |
| }, |
| { |
| "epoch": 0.3877199947068943, |
| "grad_norm": 0.515625, |
| "learning_rate": 0.000299378588663211, |
| "loss": 0.1578, |
| "step": 14650 |
| }, |
| { |
| "epoch": 0.389043271139341, |
| "grad_norm": 0.48828125, |
| "learning_rate": 0.00029936659408548585, |
| "loss": 0.1617, |
| "step": 14700 |
| }, |
| { |
| "epoch": 0.39036654757178774, |
| "grad_norm": 0.412109375, |
| "learning_rate": 0.00029935448509597554, |
| "loss": 0.1744, |
| "step": 14750 |
| }, |
| { |
| "epoch": 0.39168982400423447, |
| "grad_norm": 0.50390625, |
| "learning_rate": 0.00029934226170395535, |
| "loss": 0.1676, |
| "step": 14800 |
| }, |
| { |
| "epoch": 0.3930131004366812, |
| "grad_norm": 0.52734375, |
| "learning_rate": 0.00029932992391878814, |
| "loss": 0.1588, |
| "step": 14850 |
| }, |
| { |
| "epoch": 0.394336376869128, |
| "grad_norm": 0.57421875, |
| "learning_rate": 0.0002993174717499244, |
| "loss": 0.1616, |
| "step": 14900 |
| }, |
| { |
| "epoch": 0.3956596533015747, |
| "grad_norm": 0.4921875, |
| "learning_rate": 0.0002993049052069023, |
| "loss": 0.169, |
| "step": 14950 |
| }, |
| { |
| "epoch": 0.39698292973402144, |
| "grad_norm": 0.79296875, |
| "learning_rate": 0.00029929222429934753, |
| "loss": 0.1694, |
| "step": 15000 |
| }, |
| { |
| "epoch": 0.39698292973402144, |
| "eval_cer": 0.055393635123068885, |
| "eval_loss": 0.2413298487663269, |
| "eval_runtime": 128.1516, |
| "eval_samples_per_second": 19.734, |
| "eval_steps_per_second": 0.156, |
| "eval_wer": 0.13995040840140024, |
| "step": 15000 |
| }, |
| { |
| "epoch": 0.39830620616646817, |
| "grad_norm": 0.5078125, |
| "learning_rate": 0.0002992794290369733, |
| "loss": 0.1575, |
| "step": 15050 |
| }, |
| { |
| "epoch": 0.3996294825989149, |
| "grad_norm": 0.498046875, |
| "learning_rate": 0.0002992665194295807, |
| "loss": 0.1689, |
| "step": 15100 |
| }, |
| { |
| "epoch": 0.4009527590313616, |
| "grad_norm": 0.56640625, |
| "learning_rate": 0.00029925349548705815, |
| "loss": 0.1648, |
| "step": 15150 |
| }, |
| { |
| "epoch": 0.4022760354638084, |
| "grad_norm": 0.61328125, |
| "learning_rate": 0.0002992403572193816, |
| "loss": 0.1529, |
| "step": 15200 |
| }, |
| { |
| "epoch": 0.40359931189625514, |
| "grad_norm": 1.03125, |
| "learning_rate": 0.0002992271046366149, |
| "loss": 0.1612, |
| "step": 15250 |
| }, |
| { |
| "epoch": 0.40492258832870187, |
| "grad_norm": 0.498046875, |
| "learning_rate": 0.00029921373774890916, |
| "loss": 0.1633, |
| "step": 15300 |
| }, |
| { |
| "epoch": 0.4062458647611486, |
| "grad_norm": 0.5546875, |
| "learning_rate": 0.00029920025656650313, |
| "loss": 0.1653, |
| "step": 15350 |
| }, |
| { |
| "epoch": 0.40756914119359533, |
| "grad_norm": 0.609375, |
| "learning_rate": 0.00029918666109972315, |
| "loss": 0.1697, |
| "step": 15400 |
| }, |
| { |
| "epoch": 0.40889241762604206, |
| "grad_norm": 0.5390625, |
| "learning_rate": 0.00029917295135898313, |
| "loss": 0.1736, |
| "step": 15450 |
| }, |
| { |
| "epoch": 0.41021569405848884, |
| "grad_norm": 0.54296875, |
| "learning_rate": 0.00029915912735478434, |
| "loss": 0.1663, |
| "step": 15500 |
| }, |
| { |
| "epoch": 0.4115389704909356, |
| "grad_norm": 0.5, |
| "learning_rate": 0.00029914518909771574, |
| "loss": 0.1557, |
| "step": 15550 |
| }, |
| { |
| "epoch": 0.4128622469233823, |
| "grad_norm": 0.51171875, |
| "learning_rate": 0.00029913113659845375, |
| "loss": 0.147, |
| "step": 15600 |
| }, |
| { |
| "epoch": 0.41418552335582903, |
| "grad_norm": 0.546875, |
| "learning_rate": 0.00029911696986776237, |
| "loss": 0.1564, |
| "step": 15650 |
| }, |
| { |
| "epoch": 0.41550879978827576, |
| "grad_norm": 0.56640625, |
| "learning_rate": 0.00029910268891649285, |
| "loss": 0.1541, |
| "step": 15700 |
| }, |
| { |
| "epoch": 0.4168320762207225, |
| "grad_norm": 0.55078125, |
| "learning_rate": 0.0002990882937555843, |
| "loss": 0.17, |
| "step": 15750 |
| }, |
| { |
| "epoch": 0.4181553526531692, |
| "grad_norm": 0.5625, |
| "learning_rate": 0.000299073784396063, |
| "loss": 0.161, |
| "step": 15800 |
| }, |
| { |
| "epoch": 0.419478629085616, |
| "grad_norm": 0.56640625, |
| "learning_rate": 0.00029905916084904286, |
| "loss": 0.162, |
| "step": 15850 |
| }, |
| { |
| "epoch": 0.42080190551806274, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.0002990444231257253, |
| "loss": 0.15, |
| "step": 15900 |
| }, |
| { |
| "epoch": 0.42212518195050946, |
| "grad_norm": 0.55859375, |
| "learning_rate": 0.000299029571237399, |
| "loss": 0.1605, |
| "step": 15950 |
| }, |
| { |
| "epoch": 0.4234484583829562, |
| "grad_norm": 0.52734375, |
| "learning_rate": 0.00029901460519544026, |
| "loss": 0.156, |
| "step": 16000 |
| }, |
| { |
| "epoch": 0.4247717348154029, |
| "grad_norm": 0.5859375, |
| "learning_rate": 0.0002989995250113128, |
| "loss": 0.1642, |
| "step": 16050 |
| }, |
| { |
| "epoch": 0.42609501124784965, |
| "grad_norm": 0.52734375, |
| "learning_rate": 0.00029898433069656765, |
| "loss": 0.1515, |
| "step": 16100 |
| }, |
| { |
| "epoch": 0.42741828768029644, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.0002989690222628434, |
| "loss": 0.1491, |
| "step": 16150 |
| }, |
| { |
| "epoch": 0.42874156411274317, |
| "grad_norm": 0.87109375, |
| "learning_rate": 0.00029895359972186594, |
| "loss": 0.1596, |
| "step": 16200 |
| }, |
| { |
| "epoch": 0.4300648405451899, |
| "grad_norm": 0.57421875, |
| "learning_rate": 0.00029893806308544876, |
| "loss": 0.1605, |
| "step": 16250 |
| }, |
| { |
| "epoch": 0.4313881169776366, |
| "grad_norm": 0.44921875, |
| "learning_rate": 0.00029892241236549247, |
| "loss": 0.1636, |
| "step": 16300 |
| }, |
| { |
| "epoch": 0.43271139341008336, |
| "grad_norm": 0.494140625, |
| "learning_rate": 0.00029890664757398525, |
| "loss": 0.156, |
| "step": 16350 |
| }, |
| { |
| "epoch": 0.4340346698425301, |
| "grad_norm": 0.6171875, |
| "learning_rate": 0.00029889076872300263, |
| "loss": 0.1624, |
| "step": 16400 |
| }, |
| { |
| "epoch": 0.43535794627497687, |
| "grad_norm": 0.53125, |
| "learning_rate": 0.00029887477582470745, |
| "loss": 0.1548, |
| "step": 16450 |
| }, |
| { |
| "epoch": 0.4366812227074236, |
| "grad_norm": 0.56640625, |
| "learning_rate": 0.00029885866889135007, |
| "loss": 0.1643, |
| "step": 16500 |
| }, |
| { |
| "epoch": 0.4366812227074236, |
| "eval_cer": 0.043830831522257974, |
| "eval_loss": 0.23458540439605713, |
| "eval_runtime": 111.6706, |
| "eval_samples_per_second": 22.647, |
| "eval_steps_per_second": 0.179, |
| "eval_wer": 0.12106184364060676, |
| "step": 16500 |
| }, |
| { |
| "epoch": 0.43800449913987033, |
| "grad_norm": 0.5, |
| "learning_rate": 0.0002988424479352679, |
| "loss": 0.1565, |
| "step": 16550 |
| }, |
| { |
| "epoch": 0.43932777557231706, |
| "grad_norm": 0.46484375, |
| "learning_rate": 0.0002988261129688859, |
| "loss": 0.1727, |
| "step": 16600 |
| }, |
| { |
| "epoch": 0.4406510520047638, |
| "grad_norm": 0.55078125, |
| "learning_rate": 0.0002988096640047164, |
| "loss": 0.1578, |
| "step": 16650 |
| }, |
| { |
| "epoch": 0.4419743284372105, |
| "grad_norm": 0.46484375, |
| "learning_rate": 0.000298793101055359, |
| "loss": 0.1491, |
| "step": 16700 |
| }, |
| { |
| "epoch": 0.44329760486965725, |
| "grad_norm": 0.515625, |
| "learning_rate": 0.0002987764241335005, |
| "loss": 0.1616, |
| "step": 16750 |
| }, |
| { |
| "epoch": 0.44462088130210403, |
| "grad_norm": 0.51953125, |
| "learning_rate": 0.0002987596332519151, |
| "loss": 0.1539, |
| "step": 16800 |
| }, |
| { |
| "epoch": 0.44594415773455076, |
| "grad_norm": 0.51953125, |
| "learning_rate": 0.0002987427284234643, |
| "loss": 0.1558, |
| "step": 16850 |
| }, |
| { |
| "epoch": 0.4472674341669975, |
| "grad_norm": 0.68359375, |
| "learning_rate": 0.00029872570966109677, |
| "loss": 0.1489, |
| "step": 16900 |
| }, |
| { |
| "epoch": 0.4485907105994442, |
| "grad_norm": 0.51953125, |
| "learning_rate": 0.0002987085769778486, |
| "loss": 0.1629, |
| "step": 16950 |
| }, |
| { |
| "epoch": 0.44991398703189095, |
| "grad_norm": 0.6640625, |
| "learning_rate": 0.0002986913303868431, |
| "loss": 0.3016, |
| "step": 17000 |
| }, |
| { |
| "epoch": 0.4512372634643377, |
| "grad_norm": 0.55078125, |
| "learning_rate": 0.00029867396990129085, |
| "loss": 0.1673, |
| "step": 17050 |
| }, |
| { |
| "epoch": 0.45256053989678446, |
| "grad_norm": 0.58984375, |
| "learning_rate": 0.00029865649553448954, |
| "loss": 0.1649, |
| "step": 17100 |
| }, |
| { |
| "epoch": 0.4538838163292312, |
| "grad_norm": 0.65234375, |
| "learning_rate": 0.0002986389072998242, |
| "loss": 0.1577, |
| "step": 17150 |
| }, |
| { |
| "epoch": 0.4552070927616779, |
| "grad_norm": 0.5703125, |
| "learning_rate": 0.00029862120521076705, |
| "loss": 0.1677, |
| "step": 17200 |
| }, |
| { |
| "epoch": 0.45653036919412465, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.0002986033892808776, |
| "loss": 0.151, |
| "step": 17250 |
| }, |
| { |
| "epoch": 0.4578536456265714, |
| "grad_norm": 0.92578125, |
| "learning_rate": 0.00029858545952380245, |
| "loss": 0.1642, |
| "step": 17300 |
| }, |
| { |
| "epoch": 0.4591769220590181, |
| "grad_norm": 0.49609375, |
| "learning_rate": 0.0002985674159532754, |
| "loss": 0.1544, |
| "step": 17350 |
| }, |
| { |
| "epoch": 0.46050019849146484, |
| "grad_norm": 0.5, |
| "learning_rate": 0.0002985492585831175, |
| "loss": 0.1686, |
| "step": 17400 |
| }, |
| { |
| "epoch": 0.4618234749239116, |
| "grad_norm": 0.58203125, |
| "learning_rate": 0.0002985309874272369, |
| "loss": 0.1577, |
| "step": 17450 |
| }, |
| { |
| "epoch": 0.46314675135635835, |
| "grad_norm": 0.58984375, |
| "learning_rate": 0.000298512602499629, |
| "loss": 0.1566, |
| "step": 17500 |
| }, |
| { |
| "epoch": 0.4644700277888051, |
| "grad_norm": 0.57421875, |
| "learning_rate": 0.0002984941038143762, |
| "loss": 0.1552, |
| "step": 17550 |
| }, |
| { |
| "epoch": 0.4657933042212518, |
| "grad_norm": 0.51171875, |
| "learning_rate": 0.00029847549138564813, |
| "loss": 0.1583, |
| "step": 17600 |
| }, |
| { |
| "epoch": 0.46711658065369854, |
| "grad_norm": 0.50390625, |
| "learning_rate": 0.0002984567652277016, |
| "loss": 0.1421, |
| "step": 17650 |
| }, |
| { |
| "epoch": 0.4684398570861453, |
| "grad_norm": 0.67578125, |
| "learning_rate": 0.00029843792535488037, |
| "loss": 0.1549, |
| "step": 17700 |
| }, |
| { |
| "epoch": 0.46976313351859206, |
| "grad_norm": 0.56640625, |
| "learning_rate": 0.0002984189717816155, |
| "loss": 0.159, |
| "step": 17750 |
| }, |
| { |
| "epoch": 0.4710864099510388, |
| "grad_norm": 0.609375, |
| "learning_rate": 0.000298399904522425, |
| "loss": 0.1471, |
| "step": 17800 |
| }, |
| { |
| "epoch": 0.4724096863834855, |
| "grad_norm": 0.498046875, |
| "learning_rate": 0.0002983807235919139, |
| "loss": 0.153, |
| "step": 17850 |
| }, |
| { |
| "epoch": 0.47373296281593225, |
| "grad_norm": 0.53125, |
| "learning_rate": 0.00029836142900477466, |
| "loss": 0.1473, |
| "step": 17900 |
| }, |
| { |
| "epoch": 0.475056239248379, |
| "grad_norm": 0.546875, |
| "learning_rate": 0.00029834202077578625, |
| "loss": 0.155, |
| "step": 17950 |
| }, |
| { |
| "epoch": 0.4763795156808257, |
| "grad_norm": 0.5078125, |
| "learning_rate": 0.0002983224989198152, |
| "loss": 0.146, |
| "step": 18000 |
| }, |
| { |
| "epoch": 0.4763795156808257, |
| "eval_cer": 0.043162558754654044, |
| "eval_loss": 0.22735626995563507, |
| "eval_runtime": 134.6423, |
| "eval_samples_per_second": 18.783, |
| "eval_steps_per_second": 0.149, |
| "eval_wer": 0.11829054842473746, |
| "step": 18000 |
| }, |
| { |
| "epoch": 0.4777027921132725, |
| "grad_norm": 0.447265625, |
| "learning_rate": 0.00029830286345181473, |
| "loss": 0.1641, |
| "step": 18050 |
| }, |
| { |
| "epoch": 0.4790260685457192, |
| "grad_norm": 0.5078125, |
| "learning_rate": 0.0002982831143868253, |
| "loss": 0.1573, |
| "step": 18100 |
| }, |
| { |
| "epoch": 0.48034934497816595, |
| "grad_norm": 0.53125, |
| "learning_rate": 0.00029826325173997424, |
| "loss": 0.149, |
| "step": 18150 |
| }, |
| { |
| "epoch": 0.4816726214106127, |
| "grad_norm": 0.47265625, |
| "learning_rate": 0.0002982432755264759, |
| "loss": 0.1431, |
| "step": 18200 |
| }, |
| { |
| "epoch": 0.4829958978430594, |
| "grad_norm": 0.515625, |
| "learning_rate": 0.00029822318576163174, |
| "loss": 0.1417, |
| "step": 18250 |
| }, |
| { |
| "epoch": 0.48431917427550614, |
| "grad_norm": 0.546875, |
| "learning_rate": 0.00029820298246083, |
| "loss": 0.1555, |
| "step": 18300 |
| }, |
| { |
| "epoch": 0.48564245070795287, |
| "grad_norm": 0.52734375, |
| "learning_rate": 0.0002981826656395461, |
| "loss": 0.1489, |
| "step": 18350 |
| }, |
| { |
| "epoch": 0.48696572714039965, |
| "grad_norm": 0.5390625, |
| "learning_rate": 0.00029816223531334236, |
| "loss": 0.1438, |
| "step": 18400 |
| }, |
| { |
| "epoch": 0.4882890035728464, |
| "grad_norm": 0.5703125, |
| "learning_rate": 0.0002981416914978679, |
| "loss": 0.1751, |
| "step": 18450 |
| }, |
| { |
| "epoch": 0.4896122800052931, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.0002981210342088588, |
| "loss": 0.1499, |
| "step": 18500 |
| }, |
| { |
| "epoch": 0.49093555643773984, |
| "grad_norm": 0.55859375, |
| "learning_rate": 0.00029810026346213826, |
| "loss": 0.1436, |
| "step": 18550 |
| }, |
| { |
| "epoch": 0.49225883287018657, |
| "grad_norm": 0.5703125, |
| "learning_rate": 0.00029807937927361623, |
| "loss": 0.1462, |
| "step": 18600 |
| }, |
| { |
| "epoch": 0.4935821093026333, |
| "grad_norm": 0.546875, |
| "learning_rate": 0.00029805838165928954, |
| "loss": 0.1518, |
| "step": 18650 |
| }, |
| { |
| "epoch": 0.4949053857350801, |
| "grad_norm": 0.4140625, |
| "learning_rate": 0.00029803727063524194, |
| "loss": 0.1433, |
| "step": 18700 |
| }, |
| { |
| "epoch": 0.4962286621675268, |
| "grad_norm": 0.466796875, |
| "learning_rate": 0.000298016046217644, |
| "loss": 0.1457, |
| "step": 18750 |
| }, |
| { |
| "epoch": 0.49755193859997354, |
| "grad_norm": 0.55078125, |
| "learning_rate": 0.0002979947084227533, |
| "loss": 0.1422, |
| "step": 18800 |
| }, |
| { |
| "epoch": 0.49887521503242027, |
| "grad_norm": 0.7734375, |
| "learning_rate": 0.00029797325726691403, |
| "loss": 0.1656, |
| "step": 18850 |
| }, |
| { |
| "epoch": 0.5001984914648671, |
| "grad_norm": 0.46875, |
| "learning_rate": 0.0002979516927665575, |
| "loss": 0.1438, |
| "step": 18900 |
| }, |
| { |
| "epoch": 0.5015217678973137, |
| "grad_norm": 0.44140625, |
| "learning_rate": 0.00029793001493820157, |
| "loss": 0.1491, |
| "step": 18950 |
| }, |
| { |
| "epoch": 0.5028450443297605, |
| "grad_norm": 0.515625, |
| "learning_rate": 0.000297908223798451, |
| "loss": 0.163, |
| "step": 19000 |
| }, |
| { |
| "epoch": 0.5041683207622072, |
| "grad_norm": 0.416015625, |
| "learning_rate": 0.0002978863193639975, |
| "loss": 0.152, |
| "step": 19050 |
| }, |
| { |
| "epoch": 0.505491597194654, |
| "grad_norm": 0.46875, |
| "learning_rate": 0.00029786430165161927, |
| "loss": 0.1483, |
| "step": 19100 |
| }, |
| { |
| "epoch": 0.5068148736271006, |
| "grad_norm": 0.50390625, |
| "learning_rate": 0.0002978421706781815, |
| "loss": 0.156, |
| "step": 19150 |
| }, |
| { |
| "epoch": 0.5081381500595474, |
| "grad_norm": 0.46875, |
| "learning_rate": 0.000297819926460636, |
| "loss": 0.1438, |
| "step": 19200 |
| }, |
| { |
| "epoch": 0.5094614264919942, |
| "grad_norm": 0.58203125, |
| "learning_rate": 0.0002977975690160215, |
| "loss": 0.1462, |
| "step": 19250 |
| }, |
| { |
| "epoch": 0.5107847029244409, |
| "grad_norm": 0.50390625, |
| "learning_rate": 0.0002977750983614633, |
| "loss": 0.16, |
| "step": 19300 |
| }, |
| { |
| "epoch": 0.5121079793568877, |
| "grad_norm": 0.484375, |
| "learning_rate": 0.00029775251451417343, |
| "loss": 0.1568, |
| "step": 19350 |
| }, |
| { |
| "epoch": 0.5134312557893344, |
| "grad_norm": 0.51171875, |
| "learning_rate": 0.0002977298174914507, |
| "loss": 0.161, |
| "step": 19400 |
| }, |
| { |
| "epoch": 0.5147545322217811, |
| "grad_norm": 1.2421875, |
| "learning_rate": 0.0002977070073106806, |
| "loss": 0.1666, |
| "step": 19450 |
| }, |
| { |
| "epoch": 0.5160778086542279, |
| "grad_norm": 0.4375, |
| "learning_rate": 0.0002976840839893352, |
| "loss": 0.1483, |
| "step": 19500 |
| }, |
| { |
| "epoch": 0.5160778086542279, |
| "eval_cer": 0.04413408134957405, |
| "eval_loss": 0.22323232889175415, |
| "eval_runtime": 111.5308, |
| "eval_samples_per_second": 22.675, |
| "eval_steps_per_second": 0.179, |
| "eval_wer": 0.11942094515752626, |
| "step": 19500 |
| }, |
| { |
| "epoch": 0.5174010850866746, |
| "grad_norm": 0.5546875, |
| "learning_rate": 0.00029766104754497334, |
| "loss": 0.1511, |
| "step": 19550 |
| }, |
| { |
| "epoch": 0.5187243615191214, |
| "grad_norm": 0.51171875, |
| "learning_rate": 0.0002976378979952404, |
| "loss": 0.1402, |
| "step": 19600 |
| }, |
| { |
| "epoch": 0.520047637951568, |
| "grad_norm": 0.546875, |
| "learning_rate": 0.00029761463535786856, |
| "loss": 0.1436, |
| "step": 19650 |
| }, |
| { |
| "epoch": 0.5213709143840148, |
| "grad_norm": 0.48828125, |
| "learning_rate": 0.00029759125965067655, |
| "loss": 0.152, |
| "step": 19700 |
| }, |
| { |
| "epoch": 0.5226941908164615, |
| "grad_norm": 0.453125, |
| "learning_rate": 0.00029756777089156954, |
| "loss": 0.1526, |
| "step": 19750 |
| }, |
| { |
| "epoch": 0.5240174672489083, |
| "grad_norm": 0.470703125, |
| "learning_rate": 0.0002975441690985396, |
| "loss": 0.1408, |
| "step": 19800 |
| }, |
| { |
| "epoch": 0.5253407436813551, |
| "grad_norm": 0.59375, |
| "learning_rate": 0.0002975204542896651, |
| "loss": 0.1391, |
| "step": 19850 |
| }, |
| { |
| "epoch": 0.5266640201138018, |
| "grad_norm": 0.53125, |
| "learning_rate": 0.0002974966264831112, |
| "loss": 0.1611, |
| "step": 19900 |
| }, |
| { |
| "epoch": 0.5279872965462485, |
| "grad_norm": 0.5546875, |
| "learning_rate": 0.00029747268569712946, |
| "loss": 0.1466, |
| "step": 19950 |
| }, |
| { |
| "epoch": 0.5293105729786952, |
| "grad_norm": 0.47265625, |
| "learning_rate": 0.00029744863195005805, |
| "loss": 0.1471, |
| "step": 20000 |
| }, |
| { |
| "epoch": 0.530633849411142, |
| "grad_norm": 0.5390625, |
| "learning_rate": 0.00029742446526032166, |
| "loss": 0.1517, |
| "step": 20050 |
| }, |
| { |
| "epoch": 0.5319571258435887, |
| "grad_norm": 0.5078125, |
| "learning_rate": 0.00029740018564643157, |
| "loss": 0.1489, |
| "step": 20100 |
| }, |
| { |
| "epoch": 0.5332804022760355, |
| "grad_norm": 0.48046875, |
| "learning_rate": 0.00029737579312698526, |
| "loss": 0.1572, |
| "step": 20150 |
| }, |
| { |
| "epoch": 0.5346036787084822, |
| "grad_norm": 0.474609375, |
| "learning_rate": 0.0002973512877206671, |
| "loss": 0.143, |
| "step": 20200 |
| }, |
| { |
| "epoch": 0.5359269551409289, |
| "grad_norm": 0.498046875, |
| "learning_rate": 0.00029732666944624767, |
| "loss": 0.1455, |
| "step": 20250 |
| }, |
| { |
| "epoch": 0.5372502315733757, |
| "grad_norm": 0.462890625, |
| "learning_rate": 0.00029730193832258405, |
| "loss": 0.1581, |
| "step": 20300 |
| }, |
| { |
| "epoch": 0.5385735080058224, |
| "grad_norm": 0.546875, |
| "learning_rate": 0.0002972770943686198, |
| "loss": 0.1361, |
| "step": 20350 |
| }, |
| { |
| "epoch": 0.5398967844382692, |
| "grad_norm": 0.53125, |
| "learning_rate": 0.00029725213760338487, |
| "loss": 0.1553, |
| "step": 20400 |
| }, |
| { |
| "epoch": 0.541220060870716, |
| "grad_norm": 0.4765625, |
| "learning_rate": 0.00029722706804599566, |
| "loss": 0.1556, |
| "step": 20450 |
| }, |
| { |
| "epoch": 0.5425433373031626, |
| "grad_norm": 0.498046875, |
| "learning_rate": 0.00029720188571565493, |
| "loss": 0.1612, |
| "step": 20500 |
| }, |
| { |
| "epoch": 0.5438666137356094, |
| "grad_norm": 0.5, |
| "learning_rate": 0.00029717659063165184, |
| "loss": 0.1431, |
| "step": 20550 |
| }, |
| { |
| "epoch": 0.5451898901680561, |
| "grad_norm": 0.515625, |
| "learning_rate": 0.0002971511828133619, |
| "loss": 0.1526, |
| "step": 20600 |
| }, |
| { |
| "epoch": 0.5465131666005029, |
| "grad_norm": 0.4375, |
| "learning_rate": 0.00029712566228024695, |
| "loss": 0.1467, |
| "step": 20650 |
| }, |
| { |
| "epoch": 0.5478364430329495, |
| "grad_norm": 0.474609375, |
| "learning_rate": 0.00029710002905185533, |
| "loss": 0.1445, |
| "step": 20700 |
| }, |
| { |
| "epoch": 0.5491597194653963, |
| "grad_norm": 0.5703125, |
| "learning_rate": 0.00029707428314782147, |
| "loss": 0.1477, |
| "step": 20750 |
| }, |
| { |
| "epoch": 0.5504829958978431, |
| "grad_norm": 0.49609375, |
| "learning_rate": 0.0002970484245878662, |
| "loss": 0.155, |
| "step": 20800 |
| }, |
| { |
| "epoch": 0.5518062723302898, |
| "grad_norm": 0.515625, |
| "learning_rate": 0.00029702245339179663, |
| "loss": 0.146, |
| "step": 20850 |
| }, |
| { |
| "epoch": 0.5531295487627366, |
| "grad_norm": 0.5625, |
| "learning_rate": 0.00029699636957950627, |
| "loss": 0.1406, |
| "step": 20900 |
| }, |
| { |
| "epoch": 0.5544528251951832, |
| "grad_norm": 0.72265625, |
| "learning_rate": 0.00029697017317097476, |
| "loss": 0.1482, |
| "step": 20950 |
| }, |
| { |
| "epoch": 0.55577610162763, |
| "grad_norm": 0.416015625, |
| "learning_rate": 0.000296943864186268, |
| "loss": 0.15, |
| "step": 21000 |
| }, |
| { |
| "epoch": 0.55577610162763, |
| "eval_cer": 0.07418389294157948, |
| "eval_loss": 0.21479840576648712, |
| "eval_runtime": 175.2228, |
| "eval_samples_per_second": 14.433, |
| "eval_steps_per_second": 0.114, |
| "eval_wer": 0.1202960910151692, |
| "step": 21000 |
| }, |
| { |
| "epoch": 0.5570993780600767, |
| "grad_norm": 0.453125, |
| "learning_rate": 0.0002969174426455381, |
| "loss": 0.1374, |
| "step": 21050 |
| }, |
| { |
| "epoch": 0.5584226544925235, |
| "grad_norm": 0.50390625, |
| "learning_rate": 0.0002968909085690235, |
| "loss": 0.1431, |
| "step": 21100 |
| }, |
| { |
| "epoch": 0.5597459309249703, |
| "grad_norm": 0.484375, |
| "learning_rate": 0.00029686426197704876, |
| "loss": 0.1466, |
| "step": 21150 |
| }, |
| { |
| "epoch": 0.5610692073574169, |
| "grad_norm": 0.466796875, |
| "learning_rate": 0.00029683750289002454, |
| "loss": 0.1447, |
| "step": 21200 |
| }, |
| { |
| "epoch": 0.5623924837898637, |
| "grad_norm": 0.478515625, |
| "learning_rate": 0.00029681063132844785, |
| "loss": 0.1361, |
| "step": 21250 |
| }, |
| { |
| "epoch": 0.5637157602223104, |
| "grad_norm": 0.484375, |
| "learning_rate": 0.00029678364731290175, |
| "loss": 0.143, |
| "step": 21300 |
| }, |
| { |
| "epoch": 0.5650390366547572, |
| "grad_norm": 0.515625, |
| "learning_rate": 0.0002967565508640554, |
| "loss": 0.1401, |
| "step": 21350 |
| }, |
| { |
| "epoch": 0.566362313087204, |
| "grad_norm": 0.44140625, |
| "learning_rate": 0.00029672934200266417, |
| "loss": 0.1404, |
| "step": 21400 |
| }, |
| { |
| "epoch": 0.5676855895196506, |
| "grad_norm": 0.53125, |
| "learning_rate": 0.00029670202074956947, |
| "loss": 0.1378, |
| "step": 21450 |
| }, |
| { |
| "epoch": 0.5690088659520974, |
| "grad_norm": 0.390625, |
| "learning_rate": 0.00029667458712569883, |
| "loss": 0.1569, |
| "step": 21500 |
| }, |
| { |
| "epoch": 0.5703321423845441, |
| "grad_norm": 0.54296875, |
| "learning_rate": 0.0002966470411520659, |
| "loss": 0.1547, |
| "step": 21550 |
| }, |
| { |
| "epoch": 0.5716554188169909, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.0002966193828497703, |
| "loss": 0.1372, |
| "step": 21600 |
| }, |
| { |
| "epoch": 0.5729786952494376, |
| "grad_norm": 0.92578125, |
| "learning_rate": 0.0002965916122399977, |
| "loss": 0.1566, |
| "step": 21650 |
| }, |
| { |
| "epoch": 0.5743019716818843, |
| "grad_norm": 0.5, |
| "learning_rate": 0.00029656372934401987, |
| "loss": 0.1493, |
| "step": 21700 |
| }, |
| { |
| "epoch": 0.5756252481143311, |
| "grad_norm": 0.57421875, |
| "learning_rate": 0.00029653573418319455, |
| "loss": 0.139, |
| "step": 21750 |
| }, |
| { |
| "epoch": 0.5769485245467778, |
| "grad_norm": 0.546875, |
| "learning_rate": 0.0002965076267789654, |
| "loss": 0.1449, |
| "step": 21800 |
| }, |
| { |
| "epoch": 0.5782718009792246, |
| "grad_norm": 0.54296875, |
| "learning_rate": 0.00029647940715286225, |
| "loss": 0.145, |
| "step": 21850 |
| }, |
| { |
| "epoch": 0.5795950774116713, |
| "grad_norm": 0.431640625, |
| "learning_rate": 0.0002964510753265006, |
| "loss": 0.1548, |
| "step": 21900 |
| }, |
| { |
| "epoch": 0.580918353844118, |
| "grad_norm": 0.490234375, |
| "learning_rate": 0.00029642263132158216, |
| "loss": 0.1496, |
| "step": 21950 |
| }, |
| { |
| "epoch": 0.5822416302765647, |
| "grad_norm": 0.5078125, |
| "learning_rate": 0.00029639407515989443, |
| "loss": 0.138, |
| "step": 22000 |
| }, |
| { |
| "epoch": 0.5835649067090115, |
| "grad_norm": 0.41796875, |
| "learning_rate": 0.0002963654068633109, |
| "loss": 0.139, |
| "step": 22050 |
| }, |
| { |
| "epoch": 0.5848881831414583, |
| "grad_norm": 0.4453125, |
| "learning_rate": 0.00029633662645379077, |
| "loss": 0.143, |
| "step": 22100 |
| }, |
| { |
| "epoch": 0.586211459573905, |
| "grad_norm": 0.546875, |
| "learning_rate": 0.00029630773395337946, |
| "loss": 0.154, |
| "step": 22150 |
| }, |
| { |
| "epoch": 0.5875347360063518, |
| "grad_norm": 0.4375, |
| "learning_rate": 0.0002962787293842078, |
| "loss": 0.1394, |
| "step": 22200 |
| }, |
| { |
| "epoch": 0.5888580124387984, |
| "grad_norm": 0.56640625, |
| "learning_rate": 0.0002962496127684929, |
| "loss": 0.1426, |
| "step": 22250 |
| }, |
| { |
| "epoch": 0.5901812888712452, |
| "grad_norm": 0.7109375, |
| "learning_rate": 0.0002962203841285374, |
| "loss": 0.1391, |
| "step": 22300 |
| }, |
| { |
| "epoch": 0.5915045653036919, |
| "grad_norm": 0.498046875, |
| "learning_rate": 0.00029619104348672985, |
| "loss": 0.1356, |
| "step": 22350 |
| }, |
| { |
| "epoch": 0.5928278417361387, |
| "grad_norm": 0.4765625, |
| "learning_rate": 0.0002961615908655446, |
| "loss": 0.1336, |
| "step": 22400 |
| }, |
| { |
| "epoch": 0.5941511181685855, |
| "grad_norm": 0.50390625, |
| "learning_rate": 0.0002961320262875418, |
| "loss": 0.1366, |
| "step": 22450 |
| }, |
| { |
| "epoch": 0.5954743946010321, |
| "grad_norm": 0.48828125, |
| "learning_rate": 0.00029610234977536727, |
| "loss": 0.1455, |
| "step": 22500 |
| }, |
| { |
| "epoch": 0.5954743946010321, |
| "eval_cer": 0.036670766155072976, |
| "eval_loss": 0.22347043454647064, |
| "eval_runtime": 125.5065, |
| "eval_samples_per_second": 20.15, |
| "eval_steps_per_second": 0.159, |
| "eval_wer": 0.11056009334889148, |
| "step": 22500 |
| }, |
| { |
| "epoch": 0.5967976710334789, |
| "grad_norm": 0.466796875, |
| "learning_rate": 0.00029607256135175266, |
| "loss": 0.1498, |
| "step": 22550 |
| }, |
| { |
| "epoch": 0.5981209474659256, |
| "grad_norm": 0.498046875, |
| "learning_rate": 0.0002960426610395153, |
| "loss": 0.1375, |
| "step": 22600 |
| }, |
| { |
| "epoch": 0.5994442238983724, |
| "grad_norm": 0.55078125, |
| "learning_rate": 0.00029601264886155814, |
| "loss": 0.1413, |
| "step": 22650 |
| }, |
| { |
| "epoch": 0.6007675003308192, |
| "grad_norm": 0.51953125, |
| "learning_rate": 0.0002959825248408699, |
| "loss": 0.145, |
| "step": 22700 |
| }, |
| { |
| "epoch": 0.6020907767632658, |
| "grad_norm": 0.466796875, |
| "learning_rate": 0.0002959522890005251, |
| "loss": 0.1416, |
| "step": 22750 |
| }, |
| { |
| "epoch": 0.6034140531957126, |
| "grad_norm": 0.46875, |
| "learning_rate": 0.00029592194136368366, |
| "loss": 0.134, |
| "step": 22800 |
| }, |
| { |
| "epoch": 0.6047373296281593, |
| "grad_norm": 0.453125, |
| "learning_rate": 0.00029589148195359135, |
| "loss": 0.1442, |
| "step": 22850 |
| }, |
| { |
| "epoch": 0.6060606060606061, |
| "grad_norm": 0.46484375, |
| "learning_rate": 0.0002958609107935794, |
| "loss": 0.152, |
| "step": 22900 |
| }, |
| { |
| "epoch": 0.6073838824930528, |
| "grad_norm": 0.47265625, |
| "learning_rate": 0.00029583022790706466, |
| "loss": 0.1338, |
| "step": 22950 |
| }, |
| { |
| "epoch": 0.6087071589254995, |
| "grad_norm": 0.546875, |
| "learning_rate": 0.00029579943331754967, |
| "loss": 0.1446, |
| "step": 23000 |
| }, |
| { |
| "epoch": 0.6100304353579463, |
| "grad_norm": 0.44140625, |
| "learning_rate": 0.0002957685270486225, |
| "loss": 0.1435, |
| "step": 23050 |
| }, |
| { |
| "epoch": 0.611353711790393, |
| "grad_norm": 0.439453125, |
| "learning_rate": 0.0002957375091239566, |
| "loss": 0.1292, |
| "step": 23100 |
| }, |
| { |
| "epoch": 0.6126769882228398, |
| "grad_norm": 0.53515625, |
| "learning_rate": 0.00029570637956731114, |
| "loss": 0.1556, |
| "step": 23150 |
| }, |
| { |
| "epoch": 0.6140002646552865, |
| "grad_norm": 0.5625, |
| "learning_rate": 0.0002956751384025308, |
| "loss": 0.1523, |
| "step": 23200 |
| }, |
| { |
| "epoch": 0.6153235410877332, |
| "grad_norm": 0.45703125, |
| "learning_rate": 0.00029564378565354553, |
| "loss": 0.1533, |
| "step": 23250 |
| }, |
| { |
| "epoch": 0.6166468175201799, |
| "grad_norm": 0.53125, |
| "learning_rate": 0.000295612321344371, |
| "loss": 0.1399, |
| "step": 23300 |
| }, |
| { |
| "epoch": 0.6179700939526267, |
| "grad_norm": 0.48046875, |
| "learning_rate": 0.00029558074549910823, |
| "loss": 0.1396, |
| "step": 23350 |
| }, |
| { |
| "epoch": 0.6192933703850735, |
| "grad_norm": 0.46484375, |
| "learning_rate": 0.00029554905814194366, |
| "loss": 0.1468, |
| "step": 23400 |
| }, |
| { |
| "epoch": 0.6206166468175202, |
| "grad_norm": 0.6796875, |
| "learning_rate": 0.00029551725929714915, |
| "loss": 0.1447, |
| "step": 23450 |
| }, |
| { |
| "epoch": 0.6219399232499669, |
| "grad_norm": 0.515625, |
| "learning_rate": 0.000295485348989082, |
| "loss": 0.1527, |
| "step": 23500 |
| }, |
| { |
| "epoch": 0.6232631996824136, |
| "grad_norm": 0.44921875, |
| "learning_rate": 0.00029545332724218484, |
| "loss": 0.1515, |
| "step": 23550 |
| }, |
| { |
| "epoch": 0.6245864761148604, |
| "grad_norm": 0.5625, |
| "learning_rate": 0.00029542119408098563, |
| "loss": 0.1417, |
| "step": 23600 |
| }, |
| { |
| "epoch": 0.6259097525473072, |
| "grad_norm": 0.46875, |
| "learning_rate": 0.00029538894953009783, |
| "loss": 0.1379, |
| "step": 23650 |
| }, |
| { |
| "epoch": 0.6272330289797539, |
| "grad_norm": 0.43359375, |
| "learning_rate": 0.00029535659361422004, |
| "loss": 0.1314, |
| "step": 23700 |
| }, |
| { |
| "epoch": 0.6285563054122006, |
| "grad_norm": 0.5, |
| "learning_rate": 0.00029532412635813627, |
| "loss": 0.1349, |
| "step": 23750 |
| }, |
| { |
| "epoch": 0.6298795818446473, |
| "grad_norm": 0.396484375, |
| "learning_rate": 0.00029529154778671577, |
| "loss": 0.1515, |
| "step": 23800 |
| }, |
| { |
| "epoch": 0.6312028582770941, |
| "grad_norm": 0.443359375, |
| "learning_rate": 0.000295258857924913, |
| "loss": 0.1452, |
| "step": 23850 |
| }, |
| { |
| "epoch": 0.6325261347095408, |
| "grad_norm": 0.546875, |
| "learning_rate": 0.0002952260567977678, |
| "loss": 0.1431, |
| "step": 23900 |
| }, |
| { |
| "epoch": 0.6338494111419876, |
| "grad_norm": 0.41015625, |
| "learning_rate": 0.0002951931444304052, |
| "loss": 0.1405, |
| "step": 23950 |
| }, |
| { |
| "epoch": 0.6351726875744343, |
| "grad_norm": 0.4765625, |
| "learning_rate": 0.0002951601208480354, |
| "loss": 0.1367, |
| "step": 24000 |
| }, |
| { |
| "epoch": 0.6351726875744343, |
| "eval_cer": 0.04273576270139439, |
| "eval_loss": 0.2082057148218155, |
| "eval_runtime": 120.662, |
| "eval_samples_per_second": 20.959, |
| "eval_steps_per_second": 0.166, |
| "eval_wer": 0.11376896149358226, |
| "step": 24000 |
| }, |
| { |
| "epoch": 0.636495964006881, |
| "grad_norm": 0.58203125, |
| "learning_rate": 0.0002951269860759537, |
| "loss": 0.1605, |
| "step": 24050 |
| }, |
| { |
| "epoch": 0.6378192404393278, |
| "grad_norm": 0.470703125, |
| "learning_rate": 0.0002950937401395408, |
| "loss": 0.1345, |
| "step": 24100 |
| }, |
| { |
| "epoch": 0.6391425168717745, |
| "grad_norm": 0.435546875, |
| "learning_rate": 0.0002950603830642622, |
| "loss": 0.1339, |
| "step": 24150 |
| }, |
| { |
| "epoch": 0.6404657933042213, |
| "grad_norm": 0.404296875, |
| "learning_rate": 0.00029502691487566903, |
| "loss": 0.1347, |
| "step": 24200 |
| }, |
| { |
| "epoch": 0.6417890697366679, |
| "grad_norm": 0.5, |
| "learning_rate": 0.000294993335599397, |
| "loss": 0.1482, |
| "step": 24250 |
| }, |
| { |
| "epoch": 0.6431123461691147, |
| "grad_norm": 0.453125, |
| "learning_rate": 0.0002949596452611673, |
| "loss": 0.1414, |
| "step": 24300 |
| }, |
| { |
| "epoch": 0.6444356226015615, |
| "grad_norm": 0.431640625, |
| "learning_rate": 0.000294925843886786, |
| "loss": 0.1357, |
| "step": 24350 |
| }, |
| { |
| "epoch": 0.6457588990340082, |
| "grad_norm": 0.671875, |
| "learning_rate": 0.0002948919315021443, |
| "loss": 0.1431, |
| "step": 24400 |
| }, |
| { |
| "epoch": 0.647082175466455, |
| "grad_norm": 0.447265625, |
| "learning_rate": 0.0002948579081332183, |
| "loss": 0.131, |
| "step": 24450 |
| }, |
| { |
| "epoch": 0.6484054518989016, |
| "grad_norm": 0.46875, |
| "learning_rate": 0.0002948237738060693, |
| "loss": 0.1476, |
| "step": 24500 |
| }, |
| { |
| "epoch": 0.6497287283313484, |
| "grad_norm": 0.4296875, |
| "learning_rate": 0.00029478952854684344, |
| "loss": 0.1305, |
| "step": 24550 |
| }, |
| { |
| "epoch": 0.6510520047637952, |
| "grad_norm": 0.46875, |
| "learning_rate": 0.000294755172381772, |
| "loss": 0.1459, |
| "step": 24600 |
| }, |
| { |
| "epoch": 0.6523752811962419, |
| "grad_norm": 0.470703125, |
| "learning_rate": 0.00029472070533717105, |
| "loss": 0.1484, |
| "step": 24650 |
| }, |
| { |
| "epoch": 0.6536985576286887, |
| "grad_norm": 0.51171875, |
| "learning_rate": 0.0002946861274394416, |
| "loss": 0.1533, |
| "step": 24700 |
| }, |
| { |
| "epoch": 0.6550218340611353, |
| "grad_norm": 0.490234375, |
| "learning_rate": 0.0002946514387150697, |
| "loss": 0.1465, |
| "step": 24750 |
| }, |
| { |
| "epoch": 0.6563451104935821, |
| "grad_norm": 0.470703125, |
| "learning_rate": 0.0002946166391906261, |
| "loss": 0.1529, |
| "step": 24800 |
| }, |
| { |
| "epoch": 0.6576683869260288, |
| "grad_norm": 0.462890625, |
| "learning_rate": 0.00029458172889276665, |
| "loss": 0.136, |
| "step": 24850 |
| }, |
| { |
| "epoch": 0.6589916633584756, |
| "grad_norm": 0.498046875, |
| "learning_rate": 0.00029454670784823193, |
| "loss": 0.1409, |
| "step": 24900 |
| }, |
| { |
| "epoch": 0.6603149397909224, |
| "grad_norm": 0.51953125, |
| "learning_rate": 0.0002945115760838472, |
| "loss": 0.1424, |
| "step": 24950 |
| }, |
| { |
| "epoch": 0.661638216223369, |
| "grad_norm": 0.427734375, |
| "learning_rate": 0.00029447633362652284, |
| "loss": 0.1382, |
| "step": 25000 |
| }, |
| { |
| "epoch": 0.6629614926558158, |
| "grad_norm": 0.54296875, |
| "learning_rate": 0.0002944409805032538, |
| "loss": 0.1477, |
| "step": 25050 |
| }, |
| { |
| "epoch": 0.6642847690882625, |
| "grad_norm": 0.47265625, |
| "learning_rate": 0.00029440551674111986, |
| "loss": 0.1273, |
| "step": 25100 |
| }, |
| { |
| "epoch": 0.6656080455207093, |
| "grad_norm": 0.5, |
| "learning_rate": 0.00029436994236728555, |
| "loss": 0.134, |
| "step": 25150 |
| }, |
| { |
| "epoch": 0.666931321953156, |
| "grad_norm": 0.57421875, |
| "learning_rate": 0.0002943342574090001, |
| "loss": 0.1328, |
| "step": 25200 |
| }, |
| { |
| "epoch": 0.6682545983856027, |
| "grad_norm": 0.578125, |
| "learning_rate": 0.00029429846189359745, |
| "loss": 0.1326, |
| "step": 25250 |
| }, |
| { |
| "epoch": 0.6695778748180495, |
| "grad_norm": 0.54296875, |
| "learning_rate": 0.00029426255584849624, |
| "loss": 0.141, |
| "step": 25300 |
| }, |
| { |
| "epoch": 0.6709011512504962, |
| "grad_norm": 0.4609375, |
| "learning_rate": 0.00029422653930119986, |
| "loss": 0.1269, |
| "step": 25350 |
| }, |
| { |
| "epoch": 0.672224427682943, |
| "grad_norm": 0.4609375, |
| "learning_rate": 0.0002941904122792962, |
| "loss": 0.1463, |
| "step": 25400 |
| }, |
| { |
| "epoch": 0.6735477041153897, |
| "grad_norm": 0.58203125, |
| "learning_rate": 0.00029415417481045775, |
| "loss": 0.1468, |
| "step": 25450 |
| }, |
| { |
| "epoch": 0.6748709805478365, |
| "grad_norm": 0.470703125, |
| "learning_rate": 0.0002941178269224418, |
| "loss": 0.1389, |
| "step": 25500 |
| }, |
| { |
| "epoch": 0.6748709805478365, |
| "eval_cer": 0.038456570693712055, |
| "eval_loss": 0.20660458505153656, |
| "eval_runtime": 118.1521, |
| "eval_samples_per_second": 21.405, |
| "eval_steps_per_second": 0.169, |
| "eval_wer": 0.10811697782963828, |
| "step": 25500 |
| }, |
| { |
| "epoch": 0.6761942569802831, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.00029408136864309003, |
| "loss": 0.1431, |
| "step": 25550 |
| }, |
| { |
| "epoch": 0.6775175334127299, |
| "grad_norm": 0.443359375, |
| "learning_rate": 0.00029404480000032876, |
| "loss": 0.1378, |
| "step": 25600 |
| }, |
| { |
| "epoch": 0.6788408098451767, |
| "grad_norm": 0.4609375, |
| "learning_rate": 0.00029400812102216886, |
| "loss": 0.1453, |
| "step": 25650 |
| }, |
| { |
| "epoch": 0.6801640862776234, |
| "grad_norm": 0.423828125, |
| "learning_rate": 0.00029397133173670556, |
| "loss": 0.1377, |
| "step": 25700 |
| }, |
| { |
| "epoch": 0.6814873627100702, |
| "grad_norm": 0.4453125, |
| "learning_rate": 0.0002939344321721189, |
| "loss": 0.1581, |
| "step": 25750 |
| }, |
| { |
| "epoch": 0.6828106391425168, |
| "grad_norm": 0.51171875, |
| "learning_rate": 0.000293897422356673, |
| "loss": 0.1432, |
| "step": 25800 |
| }, |
| { |
| "epoch": 0.6841339155749636, |
| "grad_norm": 0.455078125, |
| "learning_rate": 0.0002938603023187168, |
| "loss": 0.134, |
| "step": 25850 |
| }, |
| { |
| "epoch": 0.6854571920074104, |
| "grad_norm": 0.515625, |
| "learning_rate": 0.0002938230720866834, |
| "loss": 0.134, |
| "step": 25900 |
| }, |
| { |
| "epoch": 0.6867804684398571, |
| "grad_norm": 0.375, |
| "learning_rate": 0.0002937857316890904, |
| "loss": 0.1342, |
| "step": 25950 |
| }, |
| { |
| "epoch": 0.6881037448723039, |
| "grad_norm": 0.431640625, |
| "learning_rate": 0.00029374828115453984, |
| "loss": 0.1308, |
| "step": 26000 |
| }, |
| { |
| "epoch": 0.6894270213047505, |
| "grad_norm": 0.46875, |
| "learning_rate": 0.00029371072051171803, |
| "loss": 0.1352, |
| "step": 26050 |
| }, |
| { |
| "epoch": 0.6907502977371973, |
| "grad_norm": 0.41015625, |
| "learning_rate": 0.0002936730497893957, |
| "loss": 0.1275, |
| "step": 26100 |
| }, |
| { |
| "epoch": 0.692073574169644, |
| "grad_norm": 0.5078125, |
| "learning_rate": 0.00029363526901642783, |
| "loss": 0.1436, |
| "step": 26150 |
| }, |
| { |
| "epoch": 0.6933968506020908, |
| "grad_norm": 0.474609375, |
| "learning_rate": 0.00029359737822175376, |
| "loss": 0.1345, |
| "step": 26200 |
| }, |
| { |
| "epoch": 0.6947201270345376, |
| "grad_norm": 0.447265625, |
| "learning_rate": 0.00029355937743439704, |
| "loss": 0.1342, |
| "step": 26250 |
| }, |
| { |
| "epoch": 0.6960434034669842, |
| "grad_norm": 0.412109375, |
| "learning_rate": 0.00029352126668346544, |
| "loss": 0.1356, |
| "step": 26300 |
| }, |
| { |
| "epoch": 0.697366679899431, |
| "grad_norm": 0.46484375, |
| "learning_rate": 0.0002934830459981511, |
| "loss": 0.1397, |
| "step": 26350 |
| }, |
| { |
| "epoch": 0.6986899563318777, |
| "grad_norm": 0.46484375, |
| "learning_rate": 0.00029344471540773036, |
| "loss": 0.1475, |
| "step": 26400 |
| }, |
| { |
| "epoch": 0.7000132327643245, |
| "grad_norm": 0.45703125, |
| "learning_rate": 0.0002934062749415635, |
| "loss": 0.1326, |
| "step": 26450 |
| }, |
| { |
| "epoch": 0.7013365091967712, |
| "grad_norm": 0.41015625, |
| "learning_rate": 0.00029336772462909533, |
| "loss": 0.1494, |
| "step": 26500 |
| }, |
| { |
| "epoch": 0.7026597856292179, |
| "grad_norm": 0.51171875, |
| "learning_rate": 0.0002933290644998544, |
| "loss": 0.1326, |
| "step": 26550 |
| }, |
| { |
| "epoch": 0.7039830620616647, |
| "grad_norm": 0.45703125, |
| "learning_rate": 0.0002932902945834538, |
| "loss": 0.1393, |
| "step": 26600 |
| }, |
| { |
| "epoch": 0.7053063384941114, |
| "grad_norm": 0.609375, |
| "learning_rate": 0.0002932514149095904, |
| "loss": 0.1376, |
| "step": 26650 |
| }, |
| { |
| "epoch": 0.7066296149265582, |
| "grad_norm": 0.470703125, |
| "learning_rate": 0.0002932124255080452, |
| "loss": 0.1345, |
| "step": 26700 |
| }, |
| { |
| "epoch": 0.7079528913590049, |
| "grad_norm": 0.51953125, |
| "learning_rate": 0.0002931733264086834, |
| "loss": 0.1355, |
| "step": 26750 |
| }, |
| { |
| "epoch": 0.7092761677914516, |
| "grad_norm": 0.490234375, |
| "learning_rate": 0.00029313411764145394, |
| "loss": 0.1358, |
| "step": 26800 |
| }, |
| { |
| "epoch": 0.7105994442238984, |
| "grad_norm": 0.486328125, |
| "learning_rate": 0.0002930947992363901, |
| "loss": 0.133, |
| "step": 26850 |
| }, |
| { |
| "epoch": 0.7119227206563451, |
| "grad_norm": 0.486328125, |
| "learning_rate": 0.00029305537122360895, |
| "loss": 0.1293, |
| "step": 26900 |
| }, |
| { |
| "epoch": 0.7132459970887919, |
| "grad_norm": 0.51953125, |
| "learning_rate": 0.00029301583363331156, |
| "loss": 0.1333, |
| "step": 26950 |
| }, |
| { |
| "epoch": 0.7145692735212386, |
| "grad_norm": 0.67578125, |
| "learning_rate": 0.00029297618649578287, |
| "loss": 0.1349, |
| "step": 27000 |
| }, |
| { |
| "epoch": 0.7145692735212386, |
| "eval_cer": 0.038091547753424194, |
| "eval_loss": 0.20265592634677887, |
| "eval_runtime": 122.8325, |
| "eval_samples_per_second": 20.589, |
| "eval_steps_per_second": 0.163, |
| "eval_wer": 0.10917444574095683, |
| "step": 27000 |
| }, |
| { |
| "epoch": 0.7158925499536853, |
| "grad_norm": 0.4921875, |
| "learning_rate": 0.0002929364298413918, |
| "loss": 0.1419, |
| "step": 27050 |
| }, |
| { |
| "epoch": 0.717215826386132, |
| "grad_norm": 0.5078125, |
| "learning_rate": 0.00029289656370059115, |
| "loss": 0.1369, |
| "step": 27100 |
| }, |
| { |
| "epoch": 0.7185391028185788, |
| "grad_norm": 0.4609375, |
| "learning_rate": 0.00029285658810391765, |
| "loss": 0.1345, |
| "step": 27150 |
| }, |
| { |
| "epoch": 0.7198623792510256, |
| "grad_norm": 0.466796875, |
| "learning_rate": 0.00029281650308199164, |
| "loss": 0.1357, |
| "step": 27200 |
| }, |
| { |
| "epoch": 0.7211856556834723, |
| "grad_norm": 0.390625, |
| "learning_rate": 0.00029277630866551753, |
| "loss": 0.1405, |
| "step": 27250 |
| }, |
| { |
| "epoch": 0.722508932115919, |
| "grad_norm": 0.46484375, |
| "learning_rate": 0.00029273600488528346, |
| "loss": 0.1423, |
| "step": 27300 |
| }, |
| { |
| "epoch": 0.7238322085483657, |
| "grad_norm": 0.482421875, |
| "learning_rate": 0.00029269559177216126, |
| "loss": 0.1449, |
| "step": 27350 |
| }, |
| { |
| "epoch": 0.7251554849808125, |
| "grad_norm": 0.466796875, |
| "learning_rate": 0.0002926550693571066, |
| "loss": 0.143, |
| "step": 27400 |
| }, |
| { |
| "epoch": 0.7264787614132592, |
| "grad_norm": 0.46875, |
| "learning_rate": 0.0002926144376711587, |
| "loss": 0.1349, |
| "step": 27450 |
| }, |
| { |
| "epoch": 0.727802037845706, |
| "grad_norm": 0.478515625, |
| "learning_rate": 0.00029257369674544077, |
| "loss": 0.1281, |
| "step": 27500 |
| }, |
| { |
| "epoch": 0.7291253142781527, |
| "grad_norm": 0.50390625, |
| "learning_rate": 0.00029253284661115946, |
| "loss": 0.1504, |
| "step": 27550 |
| }, |
| { |
| "epoch": 0.7304485907105994, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.0002924918872996052, |
| "loss": 0.1454, |
| "step": 27600 |
| }, |
| { |
| "epoch": 0.7317718671430462, |
| "grad_norm": 0.5625, |
| "learning_rate": 0.00029245081884215183, |
| "loss": 0.1291, |
| "step": 27650 |
| }, |
| { |
| "epoch": 0.7330951435754929, |
| "grad_norm": 0.478515625, |
| "learning_rate": 0.00029240964127025715, |
| "loss": 0.1334, |
| "step": 27700 |
| }, |
| { |
| "epoch": 0.7344184200079397, |
| "grad_norm": 0.427734375, |
| "learning_rate": 0.00029236835461546226, |
| "loss": 0.1351, |
| "step": 27750 |
| }, |
| { |
| "epoch": 0.7357416964403865, |
| "grad_norm": 0.462890625, |
| "learning_rate": 0.0002923269589093919, |
| "loss": 0.1372, |
| "step": 27800 |
| }, |
| { |
| "epoch": 0.7370649728728331, |
| "grad_norm": 0.4921875, |
| "learning_rate": 0.00029228545418375434, |
| "loss": 0.1306, |
| "step": 27850 |
| }, |
| { |
| "epoch": 0.7383882493052799, |
| "grad_norm": 0.53125, |
| "learning_rate": 0.0002922438404703414, |
| "loss": 0.1327, |
| "step": 27900 |
| }, |
| { |
| "epoch": 0.7397115257377266, |
| "grad_norm": 0.486328125, |
| "learning_rate": 0.0002922021178010283, |
| "loss": 0.1334, |
| "step": 27950 |
| }, |
| { |
| "epoch": 0.7410348021701734, |
| "grad_norm": 0.46484375, |
| "learning_rate": 0.0002921602862077738, |
| "loss": 0.1239, |
| "step": 28000 |
| }, |
| { |
| "epoch": 0.74235807860262, |
| "grad_norm": 0.4765625, |
| "learning_rate": 0.00029211834572262, |
| "loss": 0.1456, |
| "step": 28050 |
| }, |
| { |
| "epoch": 0.7436813550350668, |
| "grad_norm": 0.50390625, |
| "learning_rate": 0.0002920762963776924, |
| "loss": 0.1368, |
| "step": 28100 |
| }, |
| { |
| "epoch": 0.7450046314675136, |
| "grad_norm": 0.447265625, |
| "learning_rate": 0.00029203413820520015, |
| "loss": 0.1421, |
| "step": 28150 |
| }, |
| { |
| "epoch": 0.7463279078999603, |
| "grad_norm": 0.474609375, |
| "learning_rate": 0.0002919918712374354, |
| "loss": 0.1339, |
| "step": 28200 |
| }, |
| { |
| "epoch": 0.7476511843324071, |
| "grad_norm": 0.4375, |
| "learning_rate": 0.0002919494955067738, |
| "loss": 0.1368, |
| "step": 28250 |
| }, |
| { |
| "epoch": 0.7489744607648537, |
| "grad_norm": 0.7734375, |
| "learning_rate": 0.00029190701104567435, |
| "loss": 0.137, |
| "step": 28300 |
| }, |
| { |
| "epoch": 0.7502977371973005, |
| "grad_norm": 0.49609375, |
| "learning_rate": 0.0002918644178866793, |
| "loss": 0.1391, |
| "step": 28350 |
| }, |
| { |
| "epoch": 0.7516210136297472, |
| "grad_norm": 0.5, |
| "learning_rate": 0.00029182171606241406, |
| "loss": 0.131, |
| "step": 28400 |
| }, |
| { |
| "epoch": 0.752944290062194, |
| "grad_norm": 0.44921875, |
| "learning_rate": 0.00029177890560558745, |
| "loss": 0.1301, |
| "step": 28450 |
| }, |
| { |
| "epoch": 0.7542675664946408, |
| "grad_norm": 0.39453125, |
| "learning_rate": 0.00029173598654899137, |
| "loss": 0.131, |
| "step": 28500 |
| }, |
| { |
| "epoch": 0.7542675664946408, |
| "eval_cer": 0.04311201711676803, |
| "eval_loss": 0.20271959900856018, |
| "eval_runtime": 112.6571, |
| "eval_samples_per_second": 22.449, |
| "eval_steps_per_second": 0.178, |
| "eval_wer": 0.11314906651108518, |
| "step": 28500 |
| }, |
| { |
| "epoch": 0.7555908429270874, |
| "grad_norm": 0.412109375, |
| "learning_rate": 0.000291692958925501, |
| "loss": 0.1268, |
| "step": 28550 |
| }, |
| { |
| "epoch": 0.7569141193595342, |
| "grad_norm": 0.431640625, |
| "learning_rate": 0.0002916498227680745, |
| "loss": 0.1309, |
| "step": 28600 |
| }, |
| { |
| "epoch": 0.7582373957919809, |
| "grad_norm": 0.4921875, |
| "learning_rate": 0.0002916065781097534, |
| "loss": 0.1373, |
| "step": 28650 |
| }, |
| { |
| "epoch": 0.7595606722244277, |
| "grad_norm": 0.42578125, |
| "learning_rate": 0.00029156322498366223, |
| "loss": 0.1306, |
| "step": 28700 |
| }, |
| { |
| "epoch": 0.7608839486568745, |
| "grad_norm": 0.49609375, |
| "learning_rate": 0.0002915197634230086, |
| "loss": 0.1433, |
| "step": 28750 |
| }, |
| { |
| "epoch": 0.7622072250893211, |
| "grad_norm": 0.39453125, |
| "learning_rate": 0.00029147619346108313, |
| "loss": 0.1342, |
| "step": 28800 |
| }, |
| { |
| "epoch": 0.7635305015217679, |
| "grad_norm": 0.54296875, |
| "learning_rate": 0.00029143251513125964, |
| "loss": 0.1373, |
| "step": 28850 |
| }, |
| { |
| "epoch": 0.7648537779542146, |
| "grad_norm": 0.51953125, |
| "learning_rate": 0.0002913887284669947, |
| "loss": 0.1289, |
| "step": 28900 |
| }, |
| { |
| "epoch": 0.7661770543866614, |
| "grad_norm": 0.5078125, |
| "learning_rate": 0.0002913448335018281, |
| "loss": 0.136, |
| "step": 28950 |
| }, |
| { |
| "epoch": 0.7675003308191081, |
| "grad_norm": 0.48046875, |
| "learning_rate": 0.0002913008302693825, |
| "loss": 0.136, |
| "step": 29000 |
| }, |
| { |
| "epoch": 0.7688236072515549, |
| "grad_norm": 0.451171875, |
| "learning_rate": 0.00029125671880336345, |
| "loss": 0.1337, |
| "step": 29050 |
| }, |
| { |
| "epoch": 0.7701468836840016, |
| "grad_norm": 0.408203125, |
| "learning_rate": 0.00029121249913755947, |
| "loss": 0.1303, |
| "step": 29100 |
| }, |
| { |
| "epoch": 0.7714701601164483, |
| "grad_norm": 0.453125, |
| "learning_rate": 0.00029116817130584187, |
| "loss": 0.1353, |
| "step": 29150 |
| }, |
| { |
| "epoch": 0.7727934365488951, |
| "grad_norm": 0.494140625, |
| "learning_rate": 0.00029112373534216494, |
| "loss": 0.1254, |
| "step": 29200 |
| }, |
| { |
| "epoch": 0.7741167129813418, |
| "grad_norm": 0.54296875, |
| "learning_rate": 0.0002910791912805657, |
| "loss": 0.1278, |
| "step": 29250 |
| }, |
| { |
| "epoch": 0.7754399894137886, |
| "grad_norm": 0.439453125, |
| "learning_rate": 0.00029103453915516395, |
| "loss": 0.1466, |
| "step": 29300 |
| }, |
| { |
| "epoch": 0.7767632658462352, |
| "grad_norm": 1.109375, |
| "learning_rate": 0.00029098977900016235, |
| "loss": 0.1372, |
| "step": 29350 |
| }, |
| { |
| "epoch": 0.778086542278682, |
| "grad_norm": 0.478515625, |
| "learning_rate": 0.00029094491084984635, |
| "loss": 0.1476, |
| "step": 29400 |
| }, |
| { |
| "epoch": 0.7794098187111288, |
| "grad_norm": 0.50390625, |
| "learning_rate": 0.00029089993473858396, |
| "loss": 0.1412, |
| "step": 29450 |
| }, |
| { |
| "epoch": 0.7807330951435755, |
| "grad_norm": 0.39453125, |
| "learning_rate": 0.000290854850700826, |
| "loss": 0.1296, |
| "step": 29500 |
| }, |
| { |
| "epoch": 0.7820563715760223, |
| "grad_norm": 0.447265625, |
| "learning_rate": 0.00029080965877110586, |
| "loss": 0.1395, |
| "step": 29550 |
| }, |
| { |
| "epoch": 0.7833796480084689, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.0002907643589840397, |
| "loss": 0.1372, |
| "step": 29600 |
| }, |
| { |
| "epoch": 0.7847029244409157, |
| "grad_norm": 0.51171875, |
| "learning_rate": 0.0002907189513743263, |
| "loss": 0.1438, |
| "step": 29650 |
| }, |
| { |
| "epoch": 0.7860262008733624, |
| "grad_norm": 0.80078125, |
| "learning_rate": 0.00029067343597674686, |
| "loss": 0.1471, |
| "step": 29700 |
| }, |
| { |
| "epoch": 0.7873494773058092, |
| "grad_norm": 0.79296875, |
| "learning_rate": 0.00029062781282616527, |
| "loss": 0.1402, |
| "step": 29750 |
| }, |
| { |
| "epoch": 0.788672753738256, |
| "grad_norm": 0.490234375, |
| "learning_rate": 0.00029058208195752806, |
| "loss": 0.1309, |
| "step": 29800 |
| }, |
| { |
| "epoch": 0.7899960301707026, |
| "grad_norm": 0.37890625, |
| "learning_rate": 0.00029053624340586397, |
| "loss": 0.1306, |
| "step": 29850 |
| }, |
| { |
| "epoch": 0.7913193066031494, |
| "grad_norm": 0.5390625, |
| "learning_rate": 0.00029049029720628456, |
| "loss": 0.1362, |
| "step": 29900 |
| }, |
| { |
| "epoch": 0.7926425830355961, |
| "grad_norm": 0.419921875, |
| "learning_rate": 0.00029044424339398357, |
| "loss": 0.1369, |
| "step": 29950 |
| }, |
| { |
| "epoch": 0.7939658594680429, |
| "grad_norm": 0.41015625, |
| "learning_rate": 0.00029039808200423736, |
| "loss": 0.1336, |
| "step": 30000 |
| }, |
| { |
| "epoch": 0.7939658594680429, |
| "eval_cer": 0.037906228414508816, |
| "eval_loss": 0.2005736529827118, |
| "eval_runtime": 140.39, |
| "eval_samples_per_second": 18.014, |
| "eval_steps_per_second": 0.142, |
| "eval_wer": 0.11282088681446908, |
| "step": 30000 |
| }, |
| { |
| "epoch": 0.6094098560129791, |
| "grad_norm": 0.7734375, |
| "learning_rate": 0.00029578268099337873, |
| "loss": 0.3185, |
| "step": 30050 |
| }, |
| { |
| "epoch": 0.610423849117826, |
| "grad_norm": 0.57421875, |
| "learning_rate": 0.0002957589619350814, |
| "loss": 0.2563, |
| "step": 30100 |
| }, |
| { |
| "epoch": 0.6114378422226728, |
| "grad_norm": 0.76953125, |
| "learning_rate": 0.00029573517731969906, |
| "loss": 0.223, |
| "step": 30150 |
| }, |
| { |
| "epoch": 0.6124518353275198, |
| "grad_norm": 0.59375, |
| "learning_rate": 0.0002957113271579292, |
| "loss": 0.2166, |
| "step": 30200 |
| }, |
| { |
| "epoch": 0.6134658284323666, |
| "grad_norm": 0.65234375, |
| "learning_rate": 0.00029568741146049875, |
| "loss": 0.2127, |
| "step": 30250 |
| }, |
| { |
| "epoch": 0.6144798215372136, |
| "grad_norm": 0.6484375, |
| "learning_rate": 0.0002956634302381641, |
| "loss": 0.2099, |
| "step": 30300 |
| }, |
| { |
| "epoch": 0.6154938146420604, |
| "grad_norm": 0.6015625, |
| "learning_rate": 0.0002956393835017111, |
| "loss": 0.197, |
| "step": 30350 |
| }, |
| { |
| "epoch": 0.6165078077469073, |
| "grad_norm": 0.71875, |
| "learning_rate": 0.0002956152712619552, |
| "loss": 0.2084, |
| "step": 30400 |
| }, |
| { |
| "epoch": 0.6175218008517542, |
| "grad_norm": 0.53125, |
| "learning_rate": 0.00029559109352974113, |
| "loss": 0.1977, |
| "step": 30450 |
| }, |
| { |
| "epoch": 0.6185357939566011, |
| "grad_norm": 0.55078125, |
| "learning_rate": 0.0002955668503159432, |
| "loss": 0.2103, |
| "step": 30500 |
| }, |
| { |
| "epoch": 0.619549787061448, |
| "grad_norm": 0.60546875, |
| "learning_rate": 0.00029554254163146503, |
| "loss": 0.2018, |
| "step": 30550 |
| }, |
| { |
| "epoch": 0.6205637801662949, |
| "grad_norm": 0.6953125, |
| "learning_rate": 0.00029551816748723994, |
| "loss": 0.2006, |
| "step": 30600 |
| }, |
| { |
| "epoch": 0.6215777732711417, |
| "grad_norm": 0.61328125, |
| "learning_rate": 0.00029549372789423043, |
| "loss": 0.1851, |
| "step": 30650 |
| }, |
| { |
| "epoch": 0.6225917663759887, |
| "grad_norm": 0.578125, |
| "learning_rate": 0.00029546922286342853, |
| "loss": 0.2051, |
| "step": 30700 |
| }, |
| { |
| "epoch": 0.6236057594808355, |
| "grad_norm": 0.671875, |
| "learning_rate": 0.00029544465240585575, |
| "loss": 0.2019, |
| "step": 30750 |
| }, |
| { |
| "epoch": 0.6246197525856824, |
| "grad_norm": 0.60546875, |
| "learning_rate": 0.000295420016532563, |
| "loss": 0.1788, |
| "step": 30800 |
| }, |
| { |
| "epoch": 0.6256337456905293, |
| "grad_norm": 0.65234375, |
| "learning_rate": 0.00029539531525463056, |
| "loss": 0.1929, |
| "step": 30850 |
| }, |
| { |
| "epoch": 0.6266477387953762, |
| "grad_norm": 0.625, |
| "learning_rate": 0.0002953705485831682, |
| "loss": 0.1908, |
| "step": 30900 |
| }, |
| { |
| "epoch": 0.6276617319002231, |
| "grad_norm": 0.62109375, |
| "learning_rate": 0.0002953457165293151, |
| "loss": 0.1935, |
| "step": 30950 |
| }, |
| { |
| "epoch": 0.62867572500507, |
| "grad_norm": 0.61328125, |
| "learning_rate": 0.0002953208191042397, |
| "loss": 0.1895, |
| "step": 31000 |
| }, |
| { |
| "epoch": 0.6296897181099168, |
| "grad_norm": 0.5546875, |
| "learning_rate": 0.00029529585631914005, |
| "loss": 0.1893, |
| "step": 31050 |
| }, |
| { |
| "epoch": 0.6307037112147638, |
| "grad_norm": 0.63671875, |
| "learning_rate": 0.0002952708281852435, |
| "loss": 0.1987, |
| "step": 31100 |
| }, |
| { |
| "epoch": 0.6317177043196106, |
| "grad_norm": 0.5546875, |
| "learning_rate": 0.0002952457347138068, |
| "loss": 0.1742, |
| "step": 31150 |
| }, |
| { |
| "epoch": 0.6327316974244576, |
| "grad_norm": 0.6796875, |
| "learning_rate": 0.000295220575916116, |
| "loss": 0.1945, |
| "step": 31200 |
| }, |
| { |
| "epoch": 0.6337456905293044, |
| "grad_norm": 0.640625, |
| "learning_rate": 0.00029519535180348673, |
| "loss": 0.1767, |
| "step": 31250 |
| }, |
| { |
| "epoch": 0.6347596836341513, |
| "grad_norm": 0.59765625, |
| "learning_rate": 0.00029517006238726385, |
| "loss": 0.1791, |
| "step": 31300 |
| }, |
| { |
| "epoch": 0.6357736767389982, |
| "grad_norm": 0.64453125, |
| "learning_rate": 0.00029514470767882153, |
| "loss": 0.1845, |
| "step": 31350 |
| }, |
| { |
| "epoch": 0.636787669843845, |
| "grad_norm": 0.57421875, |
| "learning_rate": 0.0002951192876895635, |
| "loss": 0.1796, |
| "step": 31400 |
| }, |
| { |
| "epoch": 0.6378016629486919, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.0002950938024309227, |
| "loss": 0.1785, |
| "step": 31450 |
| }, |
| { |
| "epoch": 0.6388156560535388, |
| "grad_norm": 0.5859375, |
| "learning_rate": 0.0002950682519143615, |
| "loss": 0.1928, |
| "step": 31500 |
| }, |
| { |
| "epoch": 0.6388156560535388, |
| "eval_cer": 0.03631643837754322, |
| "eval_loss": 0.20944587886333466, |
| "eval_runtime": 136.9355, |
| "eval_samples_per_second": 20.506, |
| "eval_steps_per_second": 0.161, |
| "eval_wer": 0.11327781163247343, |
| "step": 31500 |
| }, |
| { |
| "epoch": 0.6398296491583857, |
| "grad_norm": 0.51171875, |
| "learning_rate": 0.00029504263615137157, |
| "loss": 0.1756, |
| "step": 31550 |
| }, |
| { |
| "epoch": 0.6408436422632326, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.00029501695515347394, |
| "loss": 0.1728, |
| "step": 31600 |
| }, |
| { |
| "epoch": 0.6418576353680795, |
| "grad_norm": 0.58984375, |
| "learning_rate": 0.000294991208932219, |
| "loss": 0.1881, |
| "step": 31650 |
| }, |
| { |
| "epoch": 0.6428716284729263, |
| "grad_norm": 0.5859375, |
| "learning_rate": 0.0002949653974991865, |
| "loss": 0.1809, |
| "step": 31700 |
| }, |
| { |
| "epoch": 0.6438856215777733, |
| "grad_norm": 0.58203125, |
| "learning_rate": 0.00029493952086598546, |
| "loss": 0.1769, |
| "step": 31750 |
| }, |
| { |
| "epoch": 0.6448996146826201, |
| "grad_norm": 0.68359375, |
| "learning_rate": 0.0002949135790442542, |
| "loss": 0.1814, |
| "step": 31800 |
| }, |
| { |
| "epoch": 0.6459136077874671, |
| "grad_norm": 0.63671875, |
| "learning_rate": 0.0002948875720456605, |
| "loss": 0.178, |
| "step": 31850 |
| }, |
| { |
| "epoch": 0.6469276008923139, |
| "grad_norm": 0.5390625, |
| "learning_rate": 0.0002948614998819013, |
| "loss": 0.1779, |
| "step": 31900 |
| }, |
| { |
| "epoch": 0.6479415939971608, |
| "grad_norm": 0.51171875, |
| "learning_rate": 0.0002948353625647029, |
| "loss": 0.1719, |
| "step": 31950 |
| }, |
| { |
| "epoch": 0.6489555871020077, |
| "grad_norm": 0.67578125, |
| "learning_rate": 0.000294809160105821, |
| "loss": 0.168, |
| "step": 32000 |
| }, |
| { |
| "epoch": 0.6499695802068546, |
| "grad_norm": 0.53515625, |
| "learning_rate": 0.0002947828925170404, |
| "loss": 0.1754, |
| "step": 32050 |
| }, |
| { |
| "epoch": 0.6509835733117015, |
| "grad_norm": 0.5859375, |
| "learning_rate": 0.0002947565598101754, |
| "loss": 0.1804, |
| "step": 32100 |
| }, |
| { |
| "epoch": 0.6519975664165484, |
| "grad_norm": 0.5859375, |
| "learning_rate": 0.0002947301619970695, |
| "loss": 0.1624, |
| "step": 32150 |
| }, |
| { |
| "epoch": 0.6530115595213952, |
| "grad_norm": 0.56640625, |
| "learning_rate": 0.0002947036990895954, |
| "loss": 0.1862, |
| "step": 32200 |
| }, |
| { |
| "epoch": 0.6540255526262422, |
| "grad_norm": 0.51171875, |
| "learning_rate": 0.00029467717109965514, |
| "loss": 0.1796, |
| "step": 32250 |
| }, |
| { |
| "epoch": 0.655039545731089, |
| "grad_norm": 0.65234375, |
| "learning_rate": 0.00029465057803918014, |
| "loss": 0.1796, |
| "step": 32300 |
| }, |
| { |
| "epoch": 0.6560535388359359, |
| "grad_norm": 0.5703125, |
| "learning_rate": 0.000294623919920131, |
| "loss": 0.1871, |
| "step": 32350 |
| }, |
| { |
| "epoch": 0.6570675319407828, |
| "grad_norm": 0.6015625, |
| "learning_rate": 0.0002945971967544974, |
| "loss": 0.1733, |
| "step": 32400 |
| }, |
| { |
| "epoch": 0.6580815250456297, |
| "grad_norm": 0.62109375, |
| "learning_rate": 0.0002945704085542987, |
| "loss": 0.1643, |
| "step": 32450 |
| }, |
| { |
| "epoch": 0.6590955181504766, |
| "grad_norm": 0.60546875, |
| "learning_rate": 0.0002945435553315831, |
| "loss": 0.175, |
| "step": 32500 |
| }, |
| { |
| "epoch": 0.6601095112553235, |
| "grad_norm": 0.5703125, |
| "learning_rate": 0.0002945166370984282, |
| "loss": 0.1808, |
| "step": 32550 |
| }, |
| { |
| "epoch": 0.6611235043601703, |
| "grad_norm": 0.625, |
| "learning_rate": 0.0002944896538669409, |
| "loss": 0.1634, |
| "step": 32600 |
| }, |
| { |
| "epoch": 0.6621374974650173, |
| "grad_norm": 0.63671875, |
| "learning_rate": 0.0002944626056492572, |
| "loss": 0.1746, |
| "step": 32650 |
| }, |
| { |
| "epoch": 0.6631514905698641, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.0002944354924575426, |
| "loss": 0.1633, |
| "step": 32700 |
| }, |
| { |
| "epoch": 0.6641654836747111, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.0002944083143039913, |
| "loss": 0.1617, |
| "step": 32750 |
| }, |
| { |
| "epoch": 0.6651794767795579, |
| "grad_norm": 0.53515625, |
| "learning_rate": 0.0002943810712008274, |
| "loss": 0.1649, |
| "step": 32800 |
| }, |
| { |
| "epoch": 0.6661934698844048, |
| "grad_norm": 0.703125, |
| "learning_rate": 0.0002943537631603036, |
| "loss": 0.1709, |
| "step": 32850 |
| }, |
| { |
| "epoch": 0.6672074629892517, |
| "grad_norm": 0.58984375, |
| "learning_rate": 0.0002943263901947022, |
| "loss": 0.1606, |
| "step": 32900 |
| }, |
| { |
| "epoch": 0.6682214560940986, |
| "grad_norm": 0.58984375, |
| "learning_rate": 0.00029429895231633453, |
| "loss": 0.1711, |
| "step": 32950 |
| }, |
| { |
| "epoch": 0.6692354491989454, |
| "grad_norm": 0.65234375, |
| "learning_rate": 0.0002942714495375411, |
| "loss": 0.1683, |
| "step": 33000 |
| }, |
| { |
| "epoch": 0.6692354491989454, |
| "eval_cer": 0.03620719253798894, |
| "eval_loss": 0.19835558533668518, |
| "eval_runtime": 121.7996, |
| "eval_samples_per_second": 23.054, |
| "eval_steps_per_second": 0.181, |
| "eval_wer": 0.10728553050308078, |
| "step": 33000 |
| }, |
| { |
| "epoch": 0.6702494423037924, |
| "grad_norm": 0.60546875, |
| "learning_rate": 0.0002942438818706917, |
| "loss": 0.1757, |
| "step": 33050 |
| }, |
| { |
| "epoch": 0.6712634354086392, |
| "grad_norm": 0.578125, |
| "learning_rate": 0.0002942162493281853, |
| "loss": 0.1617, |
| "step": 33100 |
| }, |
| { |
| "epoch": 0.6722774285134862, |
| "grad_norm": 0.70703125, |
| "learning_rate": 0.00029418855192245, |
| "loss": 0.171, |
| "step": 33150 |
| }, |
| { |
| "epoch": 0.673291421618333, |
| "grad_norm": 0.59375, |
| "learning_rate": 0.00029416078966594306, |
| "loss": 0.1655, |
| "step": 33200 |
| }, |
| { |
| "epoch": 0.6743054147231798, |
| "grad_norm": 0.5625, |
| "learning_rate": 0.00029413296257115094, |
| "loss": 0.1581, |
| "step": 33250 |
| }, |
| { |
| "epoch": 0.6753194078280268, |
| "grad_norm": 0.5703125, |
| "learning_rate": 0.00029410507065058924, |
| "loss": 0.1796, |
| "step": 33300 |
| }, |
| { |
| "epoch": 0.6763334009328736, |
| "grad_norm": 0.6015625, |
| "learning_rate": 0.00029407711391680276, |
| "loss": 0.1639, |
| "step": 33350 |
| }, |
| { |
| "epoch": 0.6773473940377206, |
| "grad_norm": 0.5859375, |
| "learning_rate": 0.00029404909238236543, |
| "loss": 0.1702, |
| "step": 33400 |
| }, |
| { |
| "epoch": 0.6783613871425674, |
| "grad_norm": 0.58203125, |
| "learning_rate": 0.0002940210060598803, |
| "loss": 0.1678, |
| "step": 33450 |
| }, |
| { |
| "epoch": 0.6793753802474143, |
| "grad_norm": 0.55859375, |
| "learning_rate": 0.0002939928549619796, |
| "loss": 0.1799, |
| "step": 33500 |
| }, |
| { |
| "epoch": 0.6803893733522612, |
| "grad_norm": 0.5625, |
| "learning_rate": 0.0002939646391013246, |
| "loss": 0.1712, |
| "step": 33550 |
| }, |
| { |
| "epoch": 0.6814033664571081, |
| "grad_norm": 0.59765625, |
| "learning_rate": 0.00029393635849060587, |
| "loss": 0.1714, |
| "step": 33600 |
| }, |
| { |
| "epoch": 0.6824173595619549, |
| "grad_norm": 0.5625, |
| "learning_rate": 0.00029390801314254294, |
| "loss": 0.1745, |
| "step": 33650 |
| }, |
| { |
| "epoch": 0.6834313526668019, |
| "grad_norm": 0.48828125, |
| "learning_rate": 0.0002938796030698845, |
| "loss": 0.1775, |
| "step": 33700 |
| }, |
| { |
| "epoch": 0.6844453457716487, |
| "grad_norm": 0.62890625, |
| "learning_rate": 0.0002938511282854084, |
| "loss": 0.1827, |
| "step": 33750 |
| }, |
| { |
| "epoch": 0.6854593388764957, |
| "grad_norm": 0.54296875, |
| "learning_rate": 0.0002938225888019216, |
| "loss": 0.1627, |
| "step": 33800 |
| }, |
| { |
| "epoch": 0.6864733319813425, |
| "grad_norm": 0.60546875, |
| "learning_rate": 0.0002937939846322601, |
| "loss": 0.1854, |
| "step": 33850 |
| }, |
| { |
| "epoch": 0.6874873250861894, |
| "grad_norm": 0.5, |
| "learning_rate": 0.00029376531578928895, |
| "loss": 0.1605, |
| "step": 33900 |
| }, |
| { |
| "epoch": 0.6885013181910363, |
| "grad_norm": 0.55859375, |
| "learning_rate": 0.0002937365822859024, |
| "loss": 0.1625, |
| "step": 33950 |
| }, |
| { |
| "epoch": 0.6895153112958832, |
| "grad_norm": 0.59375, |
| "learning_rate": 0.00029370778413502377, |
| "loss": 0.1674, |
| "step": 34000 |
| }, |
| { |
| "epoch": 0.6905293044007301, |
| "grad_norm": 0.64453125, |
| "learning_rate": 0.0002936789213496054, |
| "loss": 0.1667, |
| "step": 34050 |
| }, |
| { |
| "epoch": 0.691543297505577, |
| "grad_norm": 0.54296875, |
| "learning_rate": 0.0002936499939426287, |
| "loss": 0.1615, |
| "step": 34100 |
| }, |
| { |
| "epoch": 0.6925572906104238, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.0002936210019271042, |
| "loss": 0.1538, |
| "step": 34150 |
| }, |
| { |
| "epoch": 0.6935712837152708, |
| "grad_norm": 0.625, |
| "learning_rate": 0.00029359194531607135, |
| "loss": 0.1781, |
| "step": 34200 |
| }, |
| { |
| "epoch": 0.6945852768201176, |
| "grad_norm": 0.515625, |
| "learning_rate": 0.0002935628241225989, |
| "loss": 0.164, |
| "step": 34250 |
| }, |
| { |
| "epoch": 0.6955992699249645, |
| "grad_norm": 0.5859375, |
| "learning_rate": 0.00029353363835978453, |
| "loss": 0.1556, |
| "step": 34300 |
| }, |
| { |
| "epoch": 0.6966132630298114, |
| "grad_norm": 0.5546875, |
| "learning_rate": 0.0002935043880407548, |
| "loss": 0.1611, |
| "step": 34350 |
| }, |
| { |
| "epoch": 0.6976272561346583, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.00029347507317866546, |
| "loss": 0.1599, |
| "step": 34400 |
| }, |
| { |
| "epoch": 0.6986412492395052, |
| "grad_norm": 0.51953125, |
| "learning_rate": 0.0002934456937867013, |
| "loss": 0.1635, |
| "step": 34450 |
| }, |
| { |
| "epoch": 0.6996552423443521, |
| "grad_norm": 0.6015625, |
| "learning_rate": 0.00029341624987807614, |
| "loss": 0.153, |
| "step": 34500 |
| }, |
| { |
| "epoch": 0.6996552423443521, |
| "eval_cer": 0.03746091860144517, |
| "eval_loss": 0.1974794715642929, |
| "eval_runtime": 130.0224, |
| "eval_samples_per_second": 21.596, |
| "eval_steps_per_second": 0.169, |
| "eval_wer": 0.11148351276322026, |
| "step": 34500 |
| }, |
| { |
| "epoch": 0.7006692354491989, |
| "grad_norm": 0.57421875, |
| "learning_rate": 0.0002933867414660327, |
| "loss": 0.16, |
| "step": 34550 |
| }, |
| { |
| "epoch": 0.7016832285540459, |
| "grad_norm": 0.5859375, |
| "learning_rate": 0.0002933571685638429, |
| "loss": 0.1684, |
| "step": 34600 |
| }, |
| { |
| "epoch": 0.7026972216588927, |
| "grad_norm": 0.59375, |
| "learning_rate": 0.0002933275311848075, |
| "loss": 0.1581, |
| "step": 34650 |
| }, |
| { |
| "epoch": 0.7037112147637397, |
| "grad_norm": 0.59765625, |
| "learning_rate": 0.00029329782934225625, |
| "loss": 0.1547, |
| "step": 34700 |
| }, |
| { |
| "epoch": 0.7047252078685865, |
| "grad_norm": 0.64453125, |
| "learning_rate": 0.00029326806304954807, |
| "loss": 0.1759, |
| "step": 34750 |
| }, |
| { |
| "epoch": 0.7057392009734333, |
| "grad_norm": 0.546875, |
| "learning_rate": 0.0002932382323200707, |
| "loss": 0.1586, |
| "step": 34800 |
| }, |
| { |
| "epoch": 0.7067531940782803, |
| "grad_norm": 0.45703125, |
| "learning_rate": 0.00029320833716724095, |
| "loss": 0.1551, |
| "step": 34850 |
| }, |
| { |
| "epoch": 0.7077671871831271, |
| "grad_norm": 0.5390625, |
| "learning_rate": 0.0002931783776045045, |
| "loss": 0.1769, |
| "step": 34900 |
| }, |
| { |
| "epoch": 0.708781180287974, |
| "grad_norm": 0.53125, |
| "learning_rate": 0.0002931483536453362, |
| "loss": 0.1638, |
| "step": 34950 |
| }, |
| { |
| "epoch": 0.7097951733928209, |
| "grad_norm": 0.50390625, |
| "learning_rate": 0.0002931182653032396, |
| "loss": 0.1739, |
| "step": 35000 |
| }, |
| { |
| "epoch": 0.7108091664976678, |
| "grad_norm": 0.58984375, |
| "learning_rate": 0.0002930881125917474, |
| "loss": 0.1675, |
| "step": 35050 |
| }, |
| { |
| "epoch": 0.7118231596025147, |
| "grad_norm": 0.54296875, |
| "learning_rate": 0.0002930578955244212, |
| "loss": 0.1586, |
| "step": 35100 |
| }, |
| { |
| "epoch": 0.7128371527073616, |
| "grad_norm": 0.640625, |
| "learning_rate": 0.00029302761411485154, |
| "loss": 0.1505, |
| "step": 35150 |
| }, |
| { |
| "epoch": 0.7138511458122084, |
| "grad_norm": 0.53125, |
| "learning_rate": 0.0002929972683766579, |
| "loss": 0.1574, |
| "step": 35200 |
| }, |
| { |
| "epoch": 0.7148651389170554, |
| "grad_norm": 0.56640625, |
| "learning_rate": 0.0002929668583234887, |
| "loss": 0.156, |
| "step": 35250 |
| }, |
| { |
| "epoch": 0.7158791320219022, |
| "grad_norm": 0.52734375, |
| "learning_rate": 0.0002929363839690212, |
| "loss": 0.1539, |
| "step": 35300 |
| }, |
| { |
| "epoch": 0.7168931251267492, |
| "grad_norm": 0.5625, |
| "learning_rate": 0.0002929058453269618, |
| "loss": 0.1543, |
| "step": 35350 |
| }, |
| { |
| "epoch": 0.717907118231596, |
| "grad_norm": 0.515625, |
| "learning_rate": 0.00029287524241104556, |
| "loss": 0.159, |
| "step": 35400 |
| }, |
| { |
| "epoch": 0.7189211113364429, |
| "grad_norm": 0.55078125, |
| "learning_rate": 0.0002928445752350366, |
| "loss": 0.152, |
| "step": 35450 |
| }, |
| { |
| "epoch": 0.7199351044412898, |
| "grad_norm": 0.53125, |
| "learning_rate": 0.00029281384381272786, |
| "loss": 0.1588, |
| "step": 35500 |
| }, |
| { |
| "epoch": 0.7209490975461367, |
| "grad_norm": 0.5390625, |
| "learning_rate": 0.00029278304815794125, |
| "loss": 0.1541, |
| "step": 35550 |
| }, |
| { |
| "epoch": 0.7219630906509835, |
| "grad_norm": 0.48046875, |
| "learning_rate": 0.0002927521882845276, |
| "loss": 0.149, |
| "step": 35600 |
| }, |
| { |
| "epoch": 0.7229770837558305, |
| "grad_norm": 0.5078125, |
| "learning_rate": 0.0002927212642063664, |
| "loss": 0.1593, |
| "step": 35650 |
| }, |
| { |
| "epoch": 0.7239910768606773, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.0002926902759373663, |
| "loss": 0.1538, |
| "step": 35700 |
| }, |
| { |
| "epoch": 0.7250050699655243, |
| "grad_norm": 0.6328125, |
| "learning_rate": 0.0002926592234914647, |
| "loss": 0.1649, |
| "step": 35750 |
| }, |
| { |
| "epoch": 0.7260190630703711, |
| "grad_norm": 0.5078125, |
| "learning_rate": 0.00029262810688262777, |
| "loss": 0.1607, |
| "step": 35800 |
| }, |
| { |
| "epoch": 0.727033056175218, |
| "grad_norm": 0.5546875, |
| "learning_rate": 0.00029259692612485075, |
| "loss": 0.1637, |
| "step": 35850 |
| }, |
| { |
| "epoch": 0.7280470492800649, |
| "grad_norm": 0.53125, |
| "learning_rate": 0.00029256568123215755, |
| "loss": 0.1569, |
| "step": 35900 |
| }, |
| { |
| "epoch": 0.7290610423849118, |
| "grad_norm": 0.58203125, |
| "learning_rate": 0.00029253437221860094, |
| "loss": 0.1652, |
| "step": 35950 |
| }, |
| { |
| "epoch": 0.7300750354897587, |
| "grad_norm": 0.7890625, |
| "learning_rate": 0.0002925029990982627, |
| "loss": 0.1693, |
| "step": 36000 |
| }, |
| { |
| "epoch": 0.7300750354897587, |
| "eval_cer": 0.0336737294968969, |
| "eval_loss": 0.1916583925485611, |
| "eval_runtime": 128.7475, |
| "eval_samples_per_second": 21.81, |
| "eval_steps_per_second": 0.171, |
| "eval_wer": 0.10501726589477961, |
| "step": 36000 |
| }, |
| { |
| "epoch": 0.7310890285946056, |
| "grad_norm": 0.6328125, |
| "learning_rate": 0.0002924715618852532, |
| "loss": 0.1507, |
| "step": 36050 |
| }, |
| { |
| "epoch": 0.7321030216994524, |
| "grad_norm": 0.4921875, |
| "learning_rate": 0.00029244006059371184, |
| "loss": 0.1471, |
| "step": 36100 |
| }, |
| { |
| "epoch": 0.7331170148042994, |
| "grad_norm": 0.546875, |
| "learning_rate": 0.0002924084952378067, |
| "loss": 0.1509, |
| "step": 36150 |
| }, |
| { |
| "epoch": 0.7341310079091462, |
| "grad_norm": 0.50390625, |
| "learning_rate": 0.0002923768658317347, |
| "loss": 0.1499, |
| "step": 36200 |
| }, |
| { |
| "epoch": 0.7351450010139932, |
| "grad_norm": 0.6015625, |
| "learning_rate": 0.0002923451723897217, |
| "loss": 0.1729, |
| "step": 36250 |
| }, |
| { |
| "epoch": 0.73615899411884, |
| "grad_norm": 0.54296875, |
| "learning_rate": 0.0002923134149260222, |
| "loss": 0.1592, |
| "step": 36300 |
| }, |
| { |
| "epoch": 0.7371729872236868, |
| "grad_norm": 0.478515625, |
| "learning_rate": 0.0002922815934549195, |
| "loss": 0.1567, |
| "step": 36350 |
| }, |
| { |
| "epoch": 0.7381869803285338, |
| "grad_norm": 0.53125, |
| "learning_rate": 0.00029224970799072585, |
| "loss": 0.1517, |
| "step": 36400 |
| }, |
| { |
| "epoch": 0.7392009734333806, |
| "grad_norm": 0.640625, |
| "learning_rate": 0.00029221775854778215, |
| "loss": 0.154, |
| "step": 36450 |
| }, |
| { |
| "epoch": 0.7402149665382275, |
| "grad_norm": 0.5390625, |
| "learning_rate": 0.00029218574514045805, |
| "loss": 0.1487, |
| "step": 36500 |
| }, |
| { |
| "epoch": 0.7412289596430744, |
| "grad_norm": 0.8359375, |
| "learning_rate": 0.00029215366778315204, |
| "loss": 0.1556, |
| "step": 36550 |
| }, |
| { |
| "epoch": 0.7422429527479213, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.00029212152649029136, |
| "loss": 0.1538, |
| "step": 36600 |
| }, |
| { |
| "epoch": 0.7432569458527682, |
| "grad_norm": 0.51953125, |
| "learning_rate": 0.00029208932127633197, |
| "loss": 0.1568, |
| "step": 36650 |
| }, |
| { |
| "epoch": 0.7442709389576151, |
| "grad_norm": 0.5, |
| "learning_rate": 0.00029205705215575863, |
| "loss": 0.1575, |
| "step": 36700 |
| }, |
| { |
| "epoch": 0.7452849320624619, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.00029202471914308476, |
| "loss": 0.1544, |
| "step": 36750 |
| }, |
| { |
| "epoch": 0.7462989251673089, |
| "grad_norm": 0.55859375, |
| "learning_rate": 0.0002919923222528527, |
| "loss": 0.1535, |
| "step": 36800 |
| }, |
| { |
| "epoch": 0.7473129182721557, |
| "grad_norm": 0.50390625, |
| "learning_rate": 0.00029195986149963324, |
| "loss": 0.1521, |
| "step": 36850 |
| }, |
| { |
| "epoch": 0.7483269113770027, |
| "grad_norm": 0.6796875, |
| "learning_rate": 0.00029192733689802617, |
| "loss": 0.1489, |
| "step": 36900 |
| }, |
| { |
| "epoch": 0.7493409044818495, |
| "grad_norm": 0.51171875, |
| "learning_rate": 0.0002918947484626598, |
| "loss": 0.1578, |
| "step": 36950 |
| }, |
| { |
| "epoch": 0.7503548975866964, |
| "grad_norm": 0.5, |
| "learning_rate": 0.00029186209620819125, |
| "loss": 0.1551, |
| "step": 37000 |
| }, |
| { |
| "epoch": 0.7513688906915433, |
| "grad_norm": 0.6640625, |
| "learning_rate": 0.00029182938014930636, |
| "loss": 0.1525, |
| "step": 37050 |
| }, |
| { |
| "epoch": 0.7523828837963902, |
| "grad_norm": 0.5546875, |
| "learning_rate": 0.00029179660030071954, |
| "loss": 0.1497, |
| "step": 37100 |
| }, |
| { |
| "epoch": 0.753396876901237, |
| "grad_norm": 0.515625, |
| "learning_rate": 0.000291763756677174, |
| "loss": 0.1476, |
| "step": 37150 |
| }, |
| { |
| "epoch": 0.754410870006084, |
| "grad_norm": 0.462890625, |
| "learning_rate": 0.0002917308492934417, |
| "loss": 0.1577, |
| "step": 37200 |
| }, |
| { |
| "epoch": 0.7554248631109308, |
| "grad_norm": 0.52734375, |
| "learning_rate": 0.0002916978781643231, |
| "loss": 0.1644, |
| "step": 37250 |
| }, |
| { |
| "epoch": 0.7564388562157778, |
| "grad_norm": 0.5625, |
| "learning_rate": 0.0002916648433046474, |
| "loss": 0.1537, |
| "step": 37300 |
| }, |
| { |
| "epoch": 0.7574528493206246, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.00029163174472927254, |
| "loss": 0.1487, |
| "step": 37350 |
| }, |
| { |
| "epoch": 0.7584668424254715, |
| "grad_norm": 0.51171875, |
| "learning_rate": 0.00029159858245308496, |
| "loss": 0.159, |
| "step": 37400 |
| }, |
| { |
| "epoch": 0.7594808355303184, |
| "grad_norm": 0.546875, |
| "learning_rate": 0.000291565356491, |
| "loss": 0.1509, |
| "step": 37450 |
| }, |
| { |
| "epoch": 0.7604948286351653, |
| "grad_norm": 0.5625, |
| "learning_rate": 0.00029153206685796133, |
| "loss": 0.1477, |
| "step": 37500 |
| }, |
| { |
| "epoch": 0.7604948286351653, |
| "eval_cer": 0.036925093769345614, |
| "eval_loss": 0.19335635006427765, |
| "eval_runtime": 124.8684, |
| "eval_samples_per_second": 22.488, |
| "eval_steps_per_second": 0.176, |
| "eval_wer": 0.10775949624212879, |
| "step": 37500 |
| }, |
| { |
| "epoch": 0.7615088217400122, |
| "grad_norm": 0.515625, |
| "learning_rate": 0.0002914987135689415, |
| "loss": 0.1557, |
| "step": 37550 |
| }, |
| { |
| "epoch": 0.7625228148448591, |
| "grad_norm": 0.58203125, |
| "learning_rate": 0.00029146529663894163, |
| "loss": 0.1481, |
| "step": 37600 |
| }, |
| { |
| "epoch": 0.7635368079497059, |
| "grad_norm": 0.5859375, |
| "learning_rate": 0.00029143181608299137, |
| "loss": 0.1521, |
| "step": 37650 |
| }, |
| { |
| "epoch": 0.7645508010545529, |
| "grad_norm": 0.55859375, |
| "learning_rate": 0.0002913982719161491, |
| "loss": 0.1447, |
| "step": 37700 |
| }, |
| { |
| "epoch": 0.7655647941593997, |
| "grad_norm": 0.5859375, |
| "learning_rate": 0.0002913646641535018, |
| "loss": 0.1512, |
| "step": 37750 |
| }, |
| { |
| "epoch": 0.7665787872642466, |
| "grad_norm": 0.4921875, |
| "learning_rate": 0.00029133099281016493, |
| "loss": 0.1573, |
| "step": 37800 |
| }, |
| { |
| "epoch": 0.7675927803690935, |
| "grad_norm": 0.490234375, |
| "learning_rate": 0.0002912972579012828, |
| "loss": 0.1612, |
| "step": 37850 |
| }, |
| { |
| "epoch": 0.7686067734739404, |
| "grad_norm": 0.5703125, |
| "learning_rate": 0.0002912634594420279, |
| "loss": 0.153, |
| "step": 37900 |
| }, |
| { |
| "epoch": 0.7696207665787873, |
| "grad_norm": 0.51953125, |
| "learning_rate": 0.00029122959744760175, |
| "loss": 0.1449, |
| "step": 37950 |
| }, |
| { |
| "epoch": 0.7706347596836342, |
| "grad_norm": 0.58984375, |
| "learning_rate": 0.00029119567193323416, |
| "loss": 0.1518, |
| "step": 38000 |
| }, |
| { |
| "epoch": 0.771648752788481, |
| "grad_norm": 0.4921875, |
| "learning_rate": 0.00029116168291418366, |
| "loss": 0.1431, |
| "step": 38050 |
| }, |
| { |
| "epoch": 0.772662745893328, |
| "grad_norm": 0.58203125, |
| "learning_rate": 0.00029112763040573723, |
| "loss": 0.1513, |
| "step": 38100 |
| }, |
| { |
| "epoch": 0.7736767389981748, |
| "grad_norm": 0.484375, |
| "learning_rate": 0.0002910935144232104, |
| "loss": 0.1471, |
| "step": 38150 |
| }, |
| { |
| "epoch": 0.7746907321030218, |
| "grad_norm": 0.470703125, |
| "learning_rate": 0.0002910593349819473, |
| "loss": 0.1581, |
| "step": 38200 |
| }, |
| { |
| "epoch": 0.7757047252078686, |
| "grad_norm": 0.546875, |
| "learning_rate": 0.00029102509209732075, |
| "loss": 0.1545, |
| "step": 38250 |
| }, |
| { |
| "epoch": 0.7767187183127154, |
| "grad_norm": 0.54296875, |
| "learning_rate": 0.00029099078578473186, |
| "loss": 0.156, |
| "step": 38300 |
| }, |
| { |
| "epoch": 0.7777327114175624, |
| "grad_norm": 0.51953125, |
| "learning_rate": 0.0002909564160596103, |
| "loss": 0.1599, |
| "step": 38350 |
| }, |
| { |
| "epoch": 0.7787467045224092, |
| "grad_norm": 0.63671875, |
| "learning_rate": 0.00029092198293741443, |
| "loss": 0.147, |
| "step": 38400 |
| }, |
| { |
| "epoch": 0.7797606976272561, |
| "grad_norm": 0.52734375, |
| "learning_rate": 0.00029088748643363093, |
| "loss": 0.1582, |
| "step": 38450 |
| }, |
| { |
| "epoch": 0.780774690732103, |
| "grad_norm": 0.51953125, |
| "learning_rate": 0.00029085292656377517, |
| "loss": 0.1536, |
| "step": 38500 |
| }, |
| { |
| "epoch": 0.7817886838369499, |
| "grad_norm": 0.498046875, |
| "learning_rate": 0.0002908183033433908, |
| "loss": 0.1633, |
| "step": 38550 |
| }, |
| { |
| "epoch": 0.7828026769417968, |
| "grad_norm": 0.5390625, |
| "learning_rate": 0.00029078361678805024, |
| "loss": 0.1404, |
| "step": 38600 |
| }, |
| { |
| "epoch": 0.7838166700466437, |
| "grad_norm": 0.55078125, |
| "learning_rate": 0.0002907488669133541, |
| "loss": 0.1606, |
| "step": 38650 |
| }, |
| { |
| "epoch": 0.7848306631514905, |
| "grad_norm": 0.53125, |
| "learning_rate": 0.0002907140537349317, |
| "loss": 0.1494, |
| "step": 38700 |
| }, |
| { |
| "epoch": 0.7858446562563375, |
| "grad_norm": 0.4609375, |
| "learning_rate": 0.0002906791772684406, |
| "loss": 0.1413, |
| "step": 38750 |
| }, |
| { |
| "epoch": 0.7868586493611843, |
| "grad_norm": 0.53125, |
| "learning_rate": 0.0002906442375295672, |
| "loss": 0.152, |
| "step": 38800 |
| }, |
| { |
| "epoch": 0.7878726424660313, |
| "grad_norm": 0.546875, |
| "learning_rate": 0.00029060923453402593, |
| "loss": 0.1441, |
| "step": 38850 |
| }, |
| { |
| "epoch": 0.7888866355708781, |
| "grad_norm": 0.53125, |
| "learning_rate": 0.00029057416829756, |
| "loss": 0.1444, |
| "step": 38900 |
| }, |
| { |
| "epoch": 0.789900628675725, |
| "grad_norm": 0.5, |
| "learning_rate": 0.0002905390388359408, |
| "loss": 0.1516, |
| "step": 38950 |
| }, |
| { |
| "epoch": 0.7909146217805719, |
| "grad_norm": 0.46484375, |
| "learning_rate": 0.0002905038461649684, |
| "loss": 0.1629, |
| "step": 39000 |
| }, |
| { |
| "epoch": 0.7909146217805719, |
| "eval_cer": 0.033736155690927916, |
| "eval_loss": 0.1899910569190979, |
| "eval_runtime": 125.9975, |
| "eval_samples_per_second": 22.286, |
| "eval_steps_per_second": 0.175, |
| "eval_wer": 0.10474642832960931, |
| "step": 39000 |
| }, |
| { |
| "epoch": 0.7919286148854188, |
| "grad_norm": 0.45703125, |
| "learning_rate": 0.0002904685903004711, |
| "loss": 0.1465, |
| "step": 39050 |
| }, |
| { |
| "epoch": 0.7929426079902656, |
| "grad_norm": 0.54296875, |
| "learning_rate": 0.0002904332712583058, |
| "loss": 0.1409, |
| "step": 39100 |
| }, |
| { |
| "epoch": 0.7939566010951126, |
| "grad_norm": 0.546875, |
| "learning_rate": 0.0002903978890543576, |
| "loss": 0.1453, |
| "step": 39150 |
| }, |
| { |
| "epoch": 0.7949705941999594, |
| "grad_norm": 0.462890625, |
| "learning_rate": 0.00029036244370454024, |
| "loss": 0.1445, |
| "step": 39200 |
| }, |
| { |
| "epoch": 0.7959845873048064, |
| "grad_norm": 0.61328125, |
| "learning_rate": 0.0002903269352247957, |
| "loss": 0.1488, |
| "step": 39250 |
| }, |
| { |
| "epoch": 0.7969985804096532, |
| "grad_norm": 0.58984375, |
| "learning_rate": 0.0002902913636310944, |
| "loss": 0.1614, |
| "step": 39300 |
| }, |
| { |
| "epoch": 0.7980125735145001, |
| "grad_norm": 0.490234375, |
| "learning_rate": 0.00029025572893943524, |
| "loss": 0.1469, |
| "step": 39350 |
| }, |
| { |
| "epoch": 0.799026566619347, |
| "grad_norm": 0.486328125, |
| "learning_rate": 0.00029022003116584524, |
| "loss": 0.1457, |
| "step": 39400 |
| }, |
| { |
| "epoch": 0.8000405597241939, |
| "grad_norm": 0.55859375, |
| "learning_rate": 0.00029018427032638007, |
| "loss": 0.1575, |
| "step": 39450 |
| }, |
| { |
| "epoch": 0.8010545528290408, |
| "grad_norm": 0.54296875, |
| "learning_rate": 0.00029014844643712365, |
| "loss": 0.1588, |
| "step": 39500 |
| }, |
| { |
| "epoch": 0.8020685459338877, |
| "grad_norm": 0.447265625, |
| "learning_rate": 0.0002901125595141882, |
| "loss": 0.1606, |
| "step": 39550 |
| }, |
| { |
| "epoch": 0.8030825390387345, |
| "grad_norm": 0.5703125, |
| "learning_rate": 0.0002900766095737145, |
| "loss": 0.1488, |
| "step": 39600 |
| }, |
| { |
| "epoch": 0.8040965321435815, |
| "grad_norm": 0.54296875, |
| "learning_rate": 0.00029004059663187133, |
| "loss": 0.1525, |
| "step": 39650 |
| }, |
| { |
| "epoch": 0.8051105252484283, |
| "grad_norm": 0.82421875, |
| "learning_rate": 0.00029000452070485614, |
| "loss": 0.1462, |
| "step": 39700 |
| }, |
| { |
| "epoch": 0.8061245183532751, |
| "grad_norm": 0.4609375, |
| "learning_rate": 0.00028996838180889454, |
| "loss": 0.14, |
| "step": 39750 |
| }, |
| { |
| "epoch": 0.8071385114581221, |
| "grad_norm": 0.61328125, |
| "learning_rate": 0.0002899321799602405, |
| "loss": 0.1386, |
| "step": 39800 |
| }, |
| { |
| "epoch": 0.8081525045629689, |
| "grad_norm": 0.494140625, |
| "learning_rate": 0.00028989591517517634, |
| "loss": 0.1514, |
| "step": 39850 |
| }, |
| { |
| "epoch": 0.8091664976678159, |
| "grad_norm": 0.6640625, |
| "learning_rate": 0.00028985958747001253, |
| "loss": 0.1427, |
| "step": 39900 |
| }, |
| { |
| "epoch": 0.8101804907726627, |
| "grad_norm": 0.4921875, |
| "learning_rate": 0.000289823196861088, |
| "loss": 0.1426, |
| "step": 39950 |
| }, |
| { |
| "epoch": 0.8111944838775096, |
| "grad_norm": 0.5, |
| "learning_rate": 0.00028978674336476996, |
| "loss": 0.1632, |
| "step": 40000 |
| }, |
| { |
| "epoch": 0.8122084769823565, |
| "grad_norm": 0.5078125, |
| "learning_rate": 0.0002897502269974538, |
| "loss": 0.1505, |
| "step": 40050 |
| }, |
| { |
| "epoch": 0.8132224700872034, |
| "grad_norm": 0.515625, |
| "learning_rate": 0.00028971364777556337, |
| "loss": 0.1429, |
| "step": 40100 |
| }, |
| { |
| "epoch": 0.8142364631920503, |
| "grad_norm": 0.53515625, |
| "learning_rate": 0.00028967700571555055, |
| "loss": 0.1628, |
| "step": 40150 |
| }, |
| { |
| "epoch": 0.8152504562968972, |
| "grad_norm": 0.54296875, |
| "learning_rate": 0.00028964030083389563, |
| "loss": 0.1501, |
| "step": 40200 |
| }, |
| { |
| "epoch": 0.816264449401744, |
| "grad_norm": 0.57421875, |
| "learning_rate": 0.0002896035331471072, |
| "loss": 0.1397, |
| "step": 40250 |
| }, |
| { |
| "epoch": 0.817278442506591, |
| "grad_norm": 0.5078125, |
| "learning_rate": 0.000289566702671722, |
| "loss": 0.1568, |
| "step": 40300 |
| }, |
| { |
| "epoch": 0.8182924356114378, |
| "grad_norm": 0.515625, |
| "learning_rate": 0.00028952980942430505, |
| "loss": 0.1536, |
| "step": 40350 |
| }, |
| { |
| "epoch": 0.8193064287162848, |
| "grad_norm": 0.55859375, |
| "learning_rate": 0.00028949285342144953, |
| "loss": 0.142, |
| "step": 40400 |
| }, |
| { |
| "epoch": 0.8203204218211316, |
| "grad_norm": 0.462890625, |
| "learning_rate": 0.00028945583467977694, |
| "loss": 0.1432, |
| "step": 40450 |
| }, |
| { |
| "epoch": 0.8213344149259785, |
| "grad_norm": 0.5546875, |
| "learning_rate": 0.000289418753215937, |
| "loss": 0.1536, |
| "step": 40500 |
| }, |
| { |
| "epoch": 0.8213344149259785, |
| "eval_cer": 0.03131714067222607, |
| "eval_loss": 0.1901644915342331, |
| "eval_runtime": 128.9949, |
| "eval_samples_per_second": 21.768, |
| "eval_steps_per_second": 0.171, |
| "eval_wer": 0.10068386485205498, |
| "step": 40500 |
| }, |
| { |
| "epoch": 0.8223484080308254, |
| "grad_norm": 0.51953125, |
| "learning_rate": 0.00028938160904660754, |
| "loss": 0.1552, |
| "step": 40550 |
| }, |
| { |
| "epoch": 0.8233624011356723, |
| "grad_norm": 0.5625, |
| "learning_rate": 0.0002893444021884948, |
| "loss": 0.1527, |
| "step": 40600 |
| }, |
| { |
| "epoch": 0.8243763942405191, |
| "grad_norm": 0.53515625, |
| "learning_rate": 0.00028930713265833287, |
| "loss": 0.1427, |
| "step": 40650 |
| }, |
| { |
| "epoch": 0.8253903873453661, |
| "grad_norm": 0.53125, |
| "learning_rate": 0.0002892698004728843, |
| "loss": 0.1537, |
| "step": 40700 |
| }, |
| { |
| "epoch": 0.8264043804502129, |
| "grad_norm": 0.5078125, |
| "learning_rate": 0.0002892324056489399, |
| "loss": 0.1366, |
| "step": 40750 |
| }, |
| { |
| "epoch": 0.8274183735550599, |
| "grad_norm": 0.59765625, |
| "learning_rate": 0.0002891949482033183, |
| "loss": 0.155, |
| "step": 40800 |
| }, |
| { |
| "epoch": 0.8284323666599067, |
| "grad_norm": 0.57421875, |
| "learning_rate": 0.00028915742815286663, |
| "loss": 0.1468, |
| "step": 40850 |
| }, |
| { |
| "epoch": 0.8294463597647536, |
| "grad_norm": 0.56640625, |
| "learning_rate": 0.00028911984551445994, |
| "loss": 0.1511, |
| "step": 40900 |
| }, |
| { |
| "epoch": 0.8304603528696005, |
| "grad_norm": 0.53515625, |
| "learning_rate": 0.00028908220030500167, |
| "loss": 0.1611, |
| "step": 40950 |
| }, |
| { |
| "epoch": 0.8314743459744474, |
| "grad_norm": 0.5703125, |
| "learning_rate": 0.00028904449254142314, |
| "loss": 0.1406, |
| "step": 41000 |
| }, |
| { |
| "epoch": 0.8324883390792943, |
| "grad_norm": 0.49609375, |
| "learning_rate": 0.000289006722240684, |
| "loss": 0.1424, |
| "step": 41050 |
| }, |
| { |
| "epoch": 0.8335023321841412, |
| "grad_norm": 0.50390625, |
| "learning_rate": 0.00028896888941977195, |
| "loss": 0.1418, |
| "step": 41100 |
| }, |
| { |
| "epoch": 0.834516325288988, |
| "grad_norm": 0.404296875, |
| "learning_rate": 0.00028893099409570274, |
| "loss": 0.1542, |
| "step": 41150 |
| }, |
| { |
| "epoch": 0.835530318393835, |
| "grad_norm": 0.494140625, |
| "learning_rate": 0.00028889303628552046, |
| "loss": 0.1377, |
| "step": 41200 |
| }, |
| { |
| "epoch": 0.8365443114986818, |
| "grad_norm": 0.5390625, |
| "learning_rate": 0.00028885501600629706, |
| "loss": 0.1413, |
| "step": 41250 |
| }, |
| { |
| "epoch": 0.8375583046035286, |
| "grad_norm": 0.50390625, |
| "learning_rate": 0.00028881693327513264, |
| "loss": 0.1439, |
| "step": 41300 |
| }, |
| { |
| "epoch": 0.8385722977083756, |
| "grad_norm": 0.50390625, |
| "learning_rate": 0.0002887787881091555, |
| "loss": 0.1466, |
| "step": 41350 |
| }, |
| { |
| "epoch": 0.8395862908132224, |
| "grad_norm": 0.58203125, |
| "learning_rate": 0.000288740580525522, |
| "loss": 0.1419, |
| "step": 41400 |
| }, |
| { |
| "epoch": 0.8406002839180694, |
| "grad_norm": 0.46484375, |
| "learning_rate": 0.0002887023105414164, |
| "loss": 0.1413, |
| "step": 41450 |
| }, |
| { |
| "epoch": 0.8416142770229162, |
| "grad_norm": 0.578125, |
| "learning_rate": 0.0002886639781740512, |
| "loss": 0.1376, |
| "step": 41500 |
| }, |
| { |
| "epoch": 0.8426282701277631, |
| "grad_norm": 0.51953125, |
| "learning_rate": 0.0002886255834406669, |
| "loss": 0.1384, |
| "step": 41550 |
| }, |
| { |
| "epoch": 0.84364226323261, |
| "grad_norm": 0.5546875, |
| "learning_rate": 0.0002885871263585321, |
| "loss": 0.1469, |
| "step": 41600 |
| }, |
| { |
| "epoch": 0.8446562563374569, |
| "grad_norm": 0.48828125, |
| "learning_rate": 0.0002885486069449434, |
| "loss": 0.148, |
| "step": 41650 |
| }, |
| { |
| "epoch": 0.8456702494423038, |
| "grad_norm": 0.5859375, |
| "learning_rate": 0.0002885100252172253, |
| "loss": 0.1373, |
| "step": 41700 |
| }, |
| { |
| "epoch": 0.8466842425471507, |
| "grad_norm": 0.484375, |
| "learning_rate": 0.0002884713811927306, |
| "loss": 0.1521, |
| "step": 41750 |
| }, |
| { |
| "epoch": 0.8476982356519975, |
| "grad_norm": 0.56640625, |
| "learning_rate": 0.00028843267488883994, |
| "loss": 0.1516, |
| "step": 41800 |
| }, |
| { |
| "epoch": 0.8487122287568445, |
| "grad_norm": 0.609375, |
| "learning_rate": 0.000288393906322962, |
| "loss": 0.1398, |
| "step": 41850 |
| }, |
| { |
| "epoch": 0.8497262218616913, |
| "grad_norm": 0.5390625, |
| "learning_rate": 0.00028835507551253347, |
| "loss": 0.1478, |
| "step": 41900 |
| }, |
| { |
| "epoch": 0.8507402149665382, |
| "grad_norm": 0.447265625, |
| "learning_rate": 0.00028831618247501903, |
| "loss": 0.151, |
| "step": 41950 |
| }, |
| { |
| "epoch": 0.8517542080713851, |
| "grad_norm": 0.65625, |
| "learning_rate": 0.0002882772272279114, |
| "loss": 0.1449, |
| "step": 42000 |
| }, |
| { |
| "epoch": 0.8517542080713851, |
| "eval_cer": 0.037216416008157026, |
| "eval_loss": 0.18569107353687286, |
| "eval_runtime": 126.3528, |
| "eval_samples_per_second": 22.223, |
| "eval_steps_per_second": 0.174, |
| "eval_wer": 0.10640530841627734, |
| "step": 42000 |
| }, |
| { |
| "epoch": 0.852768201176232, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.00028823820978873116, |
| "loss": 0.1369, |
| "step": 42050 |
| }, |
| { |
| "epoch": 0.8537821942810789, |
| "grad_norm": 0.453125, |
| "learning_rate": 0.000288199130175027, |
| "loss": 0.1542, |
| "step": 42100 |
| }, |
| { |
| "epoch": 0.8547961873859258, |
| "grad_norm": 0.50390625, |
| "learning_rate": 0.0002881599884043755, |
| "loss": 0.1407, |
| "step": 42150 |
| }, |
| { |
| "epoch": 0.8558101804907726, |
| "grad_norm": 0.52734375, |
| "learning_rate": 0.0002881207844943812, |
| "loss": 0.1549, |
| "step": 42200 |
| }, |
| { |
| "epoch": 0.8568241735956196, |
| "grad_norm": 0.7890625, |
| "learning_rate": 0.0002880815184626765, |
| "loss": 0.1517, |
| "step": 42250 |
| }, |
| { |
| "epoch": 0.8578381667004664, |
| "grad_norm": 0.482421875, |
| "learning_rate": 0.00028804219032692205, |
| "loss": 0.1339, |
| "step": 42300 |
| }, |
| { |
| "epoch": 0.8588521598053134, |
| "grad_norm": 0.453125, |
| "learning_rate": 0.00028800280010480607, |
| "loss": 0.1433, |
| "step": 42350 |
| }, |
| { |
| "epoch": 0.8598661529101602, |
| "grad_norm": 0.56640625, |
| "learning_rate": 0.00028796334781404484, |
| "loss": 0.1443, |
| "step": 42400 |
| }, |
| { |
| "epoch": 0.8608801460150071, |
| "grad_norm": 0.50390625, |
| "learning_rate": 0.00028792383347238257, |
| "loss": 0.1509, |
| "step": 42450 |
| }, |
| { |
| "epoch": 0.861894139119854, |
| "grad_norm": 0.46875, |
| "learning_rate": 0.0002878842570975914, |
| "loss": 0.1417, |
| "step": 42500 |
| }, |
| { |
| "epoch": 0.8629081322247009, |
| "grad_norm": 0.55859375, |
| "learning_rate": 0.0002878446187074714, |
| "loss": 0.1432, |
| "step": 42550 |
| }, |
| { |
| "epoch": 0.8639221253295477, |
| "grad_norm": 0.49609375, |
| "learning_rate": 0.00028780491831985035, |
| "loss": 0.1363, |
| "step": 42600 |
| }, |
| { |
| "epoch": 0.8649361184343947, |
| "grad_norm": 0.5625, |
| "learning_rate": 0.00028776515595258423, |
| "loss": 0.1327, |
| "step": 42650 |
| }, |
| { |
| "epoch": 0.8659501115392415, |
| "grad_norm": 0.53125, |
| "learning_rate": 0.0002877253316235565, |
| "loss": 0.1462, |
| "step": 42700 |
| }, |
| { |
| "epoch": 0.8669641046440885, |
| "grad_norm": 0.4609375, |
| "learning_rate": 0.00028768544535067884, |
| "loss": 0.1597, |
| "step": 42750 |
| }, |
| { |
| "epoch": 0.8679780977489353, |
| "grad_norm": 0.490234375, |
| "learning_rate": 0.00028764549715189057, |
| "loss": 0.1422, |
| "step": 42800 |
| }, |
| { |
| "epoch": 0.8689920908537822, |
| "grad_norm": 0.46484375, |
| "learning_rate": 0.000287605487045159, |
| "loss": 0.1407, |
| "step": 42850 |
| }, |
| { |
| "epoch": 0.8700060839586291, |
| "grad_norm": 0.490234375, |
| "learning_rate": 0.0002875654150484792, |
| "loss": 0.1473, |
| "step": 42900 |
| }, |
| { |
| "epoch": 0.871020077063476, |
| "grad_norm": 0.55078125, |
| "learning_rate": 0.00028752528117987404, |
| "loss": 0.1529, |
| "step": 42950 |
| }, |
| { |
| "epoch": 0.8720340701683229, |
| "grad_norm": 0.47265625, |
| "learning_rate": 0.00028748508545739434, |
| "loss": 0.1508, |
| "step": 43000 |
| }, |
| { |
| "epoch": 0.8730480632731697, |
| "grad_norm": 0.4765625, |
| "learning_rate": 0.00028744482789911865, |
| "loss": 0.1389, |
| "step": 43050 |
| }, |
| { |
| "epoch": 0.8740620563780166, |
| "grad_norm": 0.53125, |
| "learning_rate": 0.0002874045085231534, |
| "loss": 0.1504, |
| "step": 43100 |
| }, |
| { |
| "epoch": 0.8750760494828635, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.0002873641273476327, |
| "loss": 0.1356, |
| "step": 43150 |
| }, |
| { |
| "epoch": 0.8760900425877104, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.0002873236843907186, |
| "loss": 0.1497, |
| "step": 43200 |
| }, |
| { |
| "epoch": 0.8771040356925572, |
| "grad_norm": 0.46875, |
| "learning_rate": 0.0002872831796706009, |
| "loss": 0.1367, |
| "step": 43250 |
| }, |
| { |
| "epoch": 0.8781180287974042, |
| "grad_norm": 0.51171875, |
| "learning_rate": 0.00028724261320549707, |
| "loss": 0.154, |
| "step": 43300 |
| }, |
| { |
| "epoch": 0.879132021902251, |
| "grad_norm": 0.5, |
| "learning_rate": 0.00028720198501365246, |
| "loss": 0.1381, |
| "step": 43350 |
| }, |
| { |
| "epoch": 0.880146015007098, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.00028716129511334015, |
| "loss": 0.1431, |
| "step": 43400 |
| }, |
| { |
| "epoch": 0.8811600081119448, |
| "grad_norm": 0.44140625, |
| "learning_rate": 0.00028712054352286097, |
| "loss": 0.1467, |
| "step": 43450 |
| }, |
| { |
| "epoch": 0.8821740012167917, |
| "grad_norm": 0.490234375, |
| "learning_rate": 0.0002870797302605436, |
| "loss": 0.1482, |
| "step": 43500 |
| }, |
| { |
| "epoch": 0.8821740012167917, |
| "eval_cer": 0.030463982687135523, |
| "eval_loss": 0.18255223333835602, |
| "eval_runtime": 128.3696, |
| "eval_samples_per_second": 21.874, |
| "eval_steps_per_second": 0.171, |
| "eval_wer": 0.09966822398266639, |
| "step": 43500 |
| }, |
| { |
| "epoch": 0.8831879943216386, |
| "grad_norm": 0.9609375, |
| "learning_rate": 0.00028703885534474424, |
| "loss": 0.1583, |
| "step": 43550 |
| }, |
| { |
| "epoch": 0.8842019874264855, |
| "grad_norm": 0.5859375, |
| "learning_rate": 0.000286997918793847, |
| "loss": 0.1471, |
| "step": 43600 |
| }, |
| { |
| "epoch": 0.8852159805313324, |
| "grad_norm": 0.59375, |
| "learning_rate": 0.0002869569206262636, |
| "loss": 0.1374, |
| "step": 43650 |
| }, |
| { |
| "epoch": 0.8862299736361793, |
| "grad_norm": 0.51171875, |
| "learning_rate": 0.0002869158608604336, |
| "loss": 0.1507, |
| "step": 43700 |
| }, |
| { |
| "epoch": 0.8872439667410261, |
| "grad_norm": 0.56640625, |
| "learning_rate": 0.00028687473951482417, |
| "loss": 0.1357, |
| "step": 43750 |
| }, |
| { |
| "epoch": 0.8882579598458731, |
| "grad_norm": 0.416015625, |
| "learning_rate": 0.00028683355660793017, |
| "loss": 0.1379, |
| "step": 43800 |
| }, |
| { |
| "epoch": 0.8892719529507199, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.00028679231215827423, |
| "loss": 0.1396, |
| "step": 43850 |
| }, |
| { |
| "epoch": 0.8902859460555668, |
| "grad_norm": 0.515625, |
| "learning_rate": 0.0002867510061844066, |
| "loss": 0.1507, |
| "step": 43900 |
| }, |
| { |
| "epoch": 0.8912999391604137, |
| "grad_norm": 0.5703125, |
| "learning_rate": 0.00028670963870490505, |
| "loss": 0.1629, |
| "step": 43950 |
| }, |
| { |
| "epoch": 0.8923139322652606, |
| "grad_norm": 0.453125, |
| "learning_rate": 0.0002866682097383754, |
| "loss": 0.1376, |
| "step": 44000 |
| }, |
| { |
| "epoch": 0.8933279253701075, |
| "grad_norm": 0.494140625, |
| "learning_rate": 0.0002866267193034507, |
| "loss": 0.1442, |
| "step": 44050 |
| }, |
| { |
| "epoch": 0.8943419184749544, |
| "grad_norm": 0.51171875, |
| "learning_rate": 0.000286585167418792, |
| "loss": 0.1448, |
| "step": 44100 |
| }, |
| { |
| "epoch": 0.8953559115798012, |
| "grad_norm": 0.455078125, |
| "learning_rate": 0.0002865435541030877, |
| "loss": 0.1503, |
| "step": 44150 |
| }, |
| { |
| "epoch": 0.8963699046846482, |
| "grad_norm": 0.490234375, |
| "learning_rate": 0.00028650187937505404, |
| "loss": 0.1459, |
| "step": 44200 |
| }, |
| { |
| "epoch": 0.897383897789495, |
| "grad_norm": 0.490234375, |
| "learning_rate": 0.00028646014325343477, |
| "loss": 0.1412, |
| "step": 44250 |
| }, |
| { |
| "epoch": 0.898397890894342, |
| "grad_norm": 0.47265625, |
| "learning_rate": 0.0002864183457570013, |
| "loss": 0.141, |
| "step": 44300 |
| }, |
| { |
| "epoch": 0.8994118839991888, |
| "grad_norm": 0.53515625, |
| "learning_rate": 0.0002863764869045525, |
| "loss": 0.1498, |
| "step": 44350 |
| }, |
| { |
| "epoch": 0.9004258771040357, |
| "grad_norm": 0.462890625, |
| "learning_rate": 0.0002863345667149152, |
| "loss": 0.1488, |
| "step": 44400 |
| }, |
| { |
| "epoch": 0.9014398702088826, |
| "grad_norm": 0.478515625, |
| "learning_rate": 0.0002862925852069434, |
| "loss": 0.1432, |
| "step": 44450 |
| }, |
| { |
| "epoch": 0.9024538633137295, |
| "grad_norm": 0.443359375, |
| "learning_rate": 0.0002862505423995189, |
| "loss": 0.1402, |
| "step": 44500 |
| }, |
| { |
| "epoch": 0.9034678564185763, |
| "grad_norm": 0.478515625, |
| "learning_rate": 0.000286208438311551, |
| "loss": 0.1542, |
| "step": 44550 |
| }, |
| { |
| "epoch": 0.9044818495234233, |
| "grad_norm": 0.44921875, |
| "learning_rate": 0.00028616627296197667, |
| "loss": 0.1386, |
| "step": 44600 |
| }, |
| { |
| "epoch": 0.9054958426282701, |
| "grad_norm": 0.55078125, |
| "learning_rate": 0.00028612404636976026, |
| "loss": 0.1499, |
| "step": 44650 |
| }, |
| { |
| "epoch": 0.906509835733117, |
| "grad_norm": 0.470703125, |
| "learning_rate": 0.00028608175855389384, |
| "loss": 0.1374, |
| "step": 44700 |
| }, |
| { |
| "epoch": 0.9075238288379639, |
| "grad_norm": 0.470703125, |
| "learning_rate": 0.0002860394095333968, |
| "loss": 0.1417, |
| "step": 44750 |
| }, |
| { |
| "epoch": 0.9085378219428107, |
| "grad_norm": 0.498046875, |
| "learning_rate": 0.0002859969993273164, |
| "loss": 0.136, |
| "step": 44800 |
| }, |
| { |
| "epoch": 0.9095518150476577, |
| "grad_norm": 0.546875, |
| "learning_rate": 0.000285954527954727, |
| "loss": 0.1446, |
| "step": 44850 |
| }, |
| { |
| "epoch": 0.9105658081525045, |
| "grad_norm": 0.53125, |
| "learning_rate": 0.0002859119954347308, |
| "loss": 0.1469, |
| "step": 44900 |
| }, |
| { |
| "epoch": 0.9115798012573515, |
| "grad_norm": 0.6015625, |
| "learning_rate": 0.00028586940178645736, |
| "loss": 0.1435, |
| "step": 44950 |
| }, |
| { |
| "epoch": 0.9125937943621983, |
| "grad_norm": 0.498046875, |
| "learning_rate": 0.00028582674702906375, |
| "loss": 0.1398, |
| "step": 45000 |
| }, |
| { |
| "epoch": 0.9125937943621983, |
| "eval_cer": 0.03009462770578535, |
| "eval_loss": 0.18028628826141357, |
| "eval_runtime": 128.6346, |
| "eval_samples_per_second": 21.829, |
| "eval_steps_per_second": 0.171, |
| "eval_wer": 0.09557180580946577, |
| "step": 45000 |
| }, |
| { |
| "epoch": 0.9136077874670452, |
| "grad_norm": 0.478515625, |
| "learning_rate": 0.0002857840311817345, |
| "loss": 0.1439, |
| "step": 45050 |
| }, |
| { |
| "epoch": 0.9146217805718921, |
| "grad_norm": 0.5078125, |
| "learning_rate": 0.0002857412542636817, |
| "loss": 0.132, |
| "step": 45100 |
| }, |
| { |
| "epoch": 0.915635773676739, |
| "grad_norm": 0.51171875, |
| "learning_rate": 0.00028569841629414485, |
| "loss": 0.1418, |
| "step": 45150 |
| }, |
| { |
| "epoch": 0.9166497667815859, |
| "grad_norm": 0.453125, |
| "learning_rate": 0.0002856555172923909, |
| "loss": 0.135, |
| "step": 45200 |
| }, |
| { |
| "epoch": 0.9176637598864328, |
| "grad_norm": 0.404296875, |
| "learning_rate": 0.00028561255727771424, |
| "loss": 0.1546, |
| "step": 45250 |
| }, |
| { |
| "epoch": 0.9186777529912796, |
| "grad_norm": 0.443359375, |
| "learning_rate": 0.0002855695362694367, |
| "loss": 0.1422, |
| "step": 45300 |
| }, |
| { |
| "epoch": 0.9196917460961266, |
| "grad_norm": 0.5625, |
| "learning_rate": 0.00028552645428690755, |
| "loss": 0.152, |
| "step": 45350 |
| }, |
| { |
| "epoch": 0.9207057392009734, |
| "grad_norm": 0.515625, |
| "learning_rate": 0.00028548331134950364, |
| "loss": 0.1614, |
| "step": 45400 |
| }, |
| { |
| "epoch": 0.9217197323058203, |
| "grad_norm": 0.59375, |
| "learning_rate": 0.00028544010747662903, |
| "loss": 0.1406, |
| "step": 45450 |
| }, |
| { |
| "epoch": 0.9227337254106672, |
| "grad_norm": 0.4921875, |
| "learning_rate": 0.00028539684268771517, |
| "loss": 0.1474, |
| "step": 45500 |
| }, |
| { |
| "epoch": 0.9237477185155141, |
| "grad_norm": 0.466796875, |
| "learning_rate": 0.00028535351700222105, |
| "loss": 0.1485, |
| "step": 45550 |
| }, |
| { |
| "epoch": 0.924761711620361, |
| "grad_norm": 0.458984375, |
| "learning_rate": 0.00028531013043963304, |
| "loss": 0.1368, |
| "step": 45600 |
| }, |
| { |
| "epoch": 0.9257757047252079, |
| "grad_norm": 0.474609375, |
| "learning_rate": 0.0002852666830194648, |
| "loss": 0.1446, |
| "step": 45650 |
| }, |
| { |
| "epoch": 0.9267896978300547, |
| "grad_norm": 0.427734375, |
| "learning_rate": 0.0002852231747612573, |
| "loss": 0.1343, |
| "step": 45700 |
| }, |
| { |
| "epoch": 0.9278036909349017, |
| "grad_norm": 0.58203125, |
| "learning_rate": 0.00028517960568457914, |
| "loss": 0.1394, |
| "step": 45750 |
| }, |
| { |
| "epoch": 0.9288176840397485, |
| "grad_norm": 0.54296875, |
| "learning_rate": 0.000285135975809026, |
| "loss": 0.1463, |
| "step": 45800 |
| }, |
| { |
| "epoch": 0.9298316771445955, |
| "grad_norm": 0.484375, |
| "learning_rate": 0.00028509228515422106, |
| "loss": 0.1455, |
| "step": 45850 |
| }, |
| { |
| "epoch": 0.9308456702494423, |
| "grad_norm": 0.48046875, |
| "learning_rate": 0.00028504853373981475, |
| "loss": 0.1413, |
| "step": 45900 |
| }, |
| { |
| "epoch": 0.9318596633542892, |
| "grad_norm": 0.57421875, |
| "learning_rate": 0.0002850047215854849, |
| "loss": 0.1314, |
| "step": 45950 |
| }, |
| { |
| "epoch": 0.9328736564591361, |
| "grad_norm": 0.4765625, |
| "learning_rate": 0.0002849608487109366, |
| "loss": 0.1345, |
| "step": 46000 |
| }, |
| { |
| "epoch": 0.933887649563983, |
| "grad_norm": 0.46484375, |
| "learning_rate": 0.00028491691513590235, |
| "loss": 0.1321, |
| "step": 46050 |
| }, |
| { |
| "epoch": 0.9349016426688298, |
| "grad_norm": 0.546875, |
| "learning_rate": 0.00028487292088014176, |
| "loss": 0.1289, |
| "step": 46100 |
| }, |
| { |
| "epoch": 0.9359156357736768, |
| "grad_norm": 0.490234375, |
| "learning_rate": 0.0002848288659634419, |
| "loss": 0.1321, |
| "step": 46150 |
| }, |
| { |
| "epoch": 0.9369296288785236, |
| "grad_norm": 0.51171875, |
| "learning_rate": 0.00028478475040561715, |
| "loss": 0.1545, |
| "step": 46200 |
| }, |
| { |
| "epoch": 0.9379436219833706, |
| "grad_norm": 0.466796875, |
| "learning_rate": 0.0002847405742265089, |
| "loss": 0.1353, |
| "step": 46250 |
| }, |
| { |
| "epoch": 0.9389576150882174, |
| "grad_norm": 0.470703125, |
| "learning_rate": 0.00028469633744598617, |
| "loss": 0.1407, |
| "step": 46300 |
| }, |
| { |
| "epoch": 0.9399716081930642, |
| "grad_norm": 0.5078125, |
| "learning_rate": 0.000284652040083945, |
| "loss": 0.1375, |
| "step": 46350 |
| }, |
| { |
| "epoch": 0.9409856012979112, |
| "grad_norm": 0.486328125, |
| "learning_rate": 0.0002846076821603087, |
| "loss": 0.1499, |
| "step": 46400 |
| }, |
| { |
| "epoch": 0.941999594402758, |
| "grad_norm": 0.6171875, |
| "learning_rate": 0.00028456326369502783, |
| "loss": 0.1354, |
| "step": 46450 |
| }, |
| { |
| "epoch": 0.943013587507605, |
| "grad_norm": 0.48828125, |
| "learning_rate": 0.0002845187847080803, |
| "loss": 0.1431, |
| "step": 46500 |
| }, |
| { |
| "epoch": 0.943013587507605, |
| "eval_cer": 0.03134315158640565, |
| "eval_loss": 0.18719163537025452, |
| "eval_runtime": 129.4245, |
| "eval_samples_per_second": 21.696, |
| "eval_steps_per_second": 0.17, |
| "eval_wer": 0.09970207867831268, |
| "step": 46500 |
| }, |
| { |
| "epoch": 0.9440275806124518, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.00028447424521947103, |
| "loss": 0.1401, |
| "step": 46550 |
| }, |
| { |
| "epoch": 0.9450415737172987, |
| "grad_norm": 0.4609375, |
| "learning_rate": 0.00028442964524923235, |
| "loss": 0.1431, |
| "step": 46600 |
| }, |
| { |
| "epoch": 0.9460555668221456, |
| "grad_norm": 0.546875, |
| "learning_rate": 0.0002843849848174237, |
| "loss": 0.1404, |
| "step": 46650 |
| }, |
| { |
| "epoch": 0.9470695599269925, |
| "grad_norm": 0.47265625, |
| "learning_rate": 0.00028434026394413164, |
| "loss": 0.1339, |
| "step": 46700 |
| }, |
| { |
| "epoch": 0.9480835530318393, |
| "grad_norm": 0.451171875, |
| "learning_rate": 0.00028429548264947014, |
| "loss": 0.1372, |
| "step": 46750 |
| }, |
| { |
| "epoch": 0.9490975461366863, |
| "grad_norm": 0.474609375, |
| "learning_rate": 0.00028425064095358003, |
| "loss": 0.14, |
| "step": 46800 |
| }, |
| { |
| "epoch": 0.9501115392415331, |
| "grad_norm": 0.46484375, |
| "learning_rate": 0.0002842057388766296, |
| "loss": 0.1387, |
| "step": 46850 |
| }, |
| { |
| "epoch": 0.9511255323463801, |
| "grad_norm": 0.478515625, |
| "learning_rate": 0.00028416077643881406, |
| "loss": 0.1414, |
| "step": 46900 |
| }, |
| { |
| "epoch": 0.9521395254512269, |
| "grad_norm": 0.55859375, |
| "learning_rate": 0.000284115753660356, |
| "loss": 0.132, |
| "step": 46950 |
| }, |
| { |
| "epoch": 0.9531535185560738, |
| "grad_norm": 0.47265625, |
| "learning_rate": 0.000284070670561505, |
| "loss": 0.1377, |
| "step": 47000 |
| }, |
| { |
| "epoch": 0.9541675116609207, |
| "grad_norm": 0.5390625, |
| "learning_rate": 0.0002840255271625377, |
| "loss": 0.1322, |
| "step": 47050 |
| }, |
| { |
| "epoch": 0.9551815047657676, |
| "grad_norm": 0.5, |
| "learning_rate": 0.00028398032348375815, |
| "loss": 0.1324, |
| "step": 47100 |
| }, |
| { |
| "epoch": 0.9561954978706145, |
| "grad_norm": 0.474609375, |
| "learning_rate": 0.00028393505954549714, |
| "loss": 0.1445, |
| "step": 47150 |
| }, |
| { |
| "epoch": 0.9572094909754614, |
| "grad_norm": 0.4609375, |
| "learning_rate": 0.00028388973536811276, |
| "loss": 0.1346, |
| "step": 47200 |
| }, |
| { |
| "epoch": 0.9582234840803082, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.0002838443509719903, |
| "loss": 0.1389, |
| "step": 47250 |
| }, |
| { |
| "epoch": 0.9592374771851552, |
| "grad_norm": 0.46484375, |
| "learning_rate": 0.00028379890637754196, |
| "loss": 0.1317, |
| "step": 47300 |
| }, |
| { |
| "epoch": 0.960251470290002, |
| "grad_norm": 0.48046875, |
| "learning_rate": 0.000283753401605207, |
| "loss": 0.1411, |
| "step": 47350 |
| }, |
| { |
| "epoch": 0.9612654633948489, |
| "grad_norm": 0.466796875, |
| "learning_rate": 0.0002837078366754519, |
| "loss": 0.1434, |
| "step": 47400 |
| }, |
| { |
| "epoch": 0.9622794564996958, |
| "grad_norm": 0.46484375, |
| "learning_rate": 0.0002836622116087701, |
| "loss": 0.1378, |
| "step": 47450 |
| }, |
| { |
| "epoch": 0.9632934496045427, |
| "grad_norm": 0.53515625, |
| "learning_rate": 0.000283616526425682, |
| "loss": 0.1385, |
| "step": 47500 |
| }, |
| { |
| "epoch": 0.9643074427093896, |
| "grad_norm": 0.478515625, |
| "learning_rate": 0.00028357078114673527, |
| "loss": 0.1385, |
| "step": 47550 |
| }, |
| { |
| "epoch": 0.9653214358142365, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.0002835249757925044, |
| "loss": 0.1381, |
| "step": 47600 |
| }, |
| { |
| "epoch": 0.9663354289190833, |
| "grad_norm": 0.6171875, |
| "learning_rate": 0.000283479110383591, |
| "loss": 0.142, |
| "step": 47650 |
| }, |
| { |
| "epoch": 0.9673494220239303, |
| "grad_norm": 0.478515625, |
| "learning_rate": 0.0002834331849406236, |
| "loss": 0.1537, |
| "step": 47700 |
| }, |
| { |
| "epoch": 0.9683634151287771, |
| "grad_norm": 0.490234375, |
| "learning_rate": 0.0002833871994842579, |
| "loss": 0.1318, |
| "step": 47750 |
| }, |
| { |
| "epoch": 0.9693774082336241, |
| "grad_norm": 0.494140625, |
| "learning_rate": 0.00028334115403517643, |
| "loss": 0.1472, |
| "step": 47800 |
| }, |
| { |
| "epoch": 0.9703914013384709, |
| "grad_norm": 0.50390625, |
| "learning_rate": 0.00028329504861408875, |
| "loss": 0.1366, |
| "step": 47850 |
| }, |
| { |
| "epoch": 0.9714053944433177, |
| "grad_norm": 0.546875, |
| "learning_rate": 0.00028324888324173143, |
| "loss": 0.1294, |
| "step": 47900 |
| }, |
| { |
| "epoch": 0.9724193875481647, |
| "grad_norm": 0.478515625, |
| "learning_rate": 0.00028320265793886797, |
| "loss": 0.1353, |
| "step": 47950 |
| }, |
| { |
| "epoch": 0.9734333806530115, |
| "grad_norm": 0.6484375, |
| "learning_rate": 0.00028315637272628875, |
| "loss": 0.1313, |
| "step": 48000 |
| }, |
| { |
| "epoch": 0.9734333806530115, |
| "eval_cer": 0.030838539851321616, |
| "eval_loss": 0.18022538721561432, |
| "eval_runtime": 129.1875, |
| "eval_samples_per_second": 21.736, |
| "eval_steps_per_second": 0.17, |
| "eval_wer": 0.0979416345047058, |
| "step": 48000 |
| }, |
| { |
| "epoch": 0.9744473737578584, |
| "grad_norm": 0.51171875, |
| "learning_rate": 0.00028311002762481134, |
| "loss": 0.1304, |
| "step": 48050 |
| }, |
| { |
| "epoch": 0.9754613668627053, |
| "grad_norm": 0.498046875, |
| "learning_rate": 0.00028306362265528, |
| "loss": 0.127, |
| "step": 48100 |
| }, |
| { |
| "epoch": 0.9764753599675522, |
| "grad_norm": 0.5546875, |
| "learning_rate": 0.00028301715783856594, |
| "loss": 0.1405, |
| "step": 48150 |
| }, |
| { |
| "epoch": 0.9774893530723991, |
| "grad_norm": 0.51171875, |
| "learning_rate": 0.00028297063319556744, |
| "loss": 0.1317, |
| "step": 48200 |
| }, |
| { |
| "epoch": 0.978503346177246, |
| "grad_norm": 0.498046875, |
| "learning_rate": 0.0002829240487472095, |
| "loss": 0.134, |
| "step": 48250 |
| }, |
| { |
| "epoch": 0.9795173392820928, |
| "grad_norm": 0.408203125, |
| "learning_rate": 0.00028287740451444425, |
| "loss": 0.131, |
| "step": 48300 |
| }, |
| { |
| "epoch": 0.9805313323869398, |
| "grad_norm": 0.578125, |
| "learning_rate": 0.0002828307005182504, |
| "loss": 0.1492, |
| "step": 48350 |
| }, |
| { |
| "epoch": 0.9815453254917866, |
| "grad_norm": 0.55859375, |
| "learning_rate": 0.0002827839367796339, |
| "loss": 0.1392, |
| "step": 48400 |
| }, |
| { |
| "epoch": 0.9825593185966336, |
| "grad_norm": 0.5078125, |
| "learning_rate": 0.0002827371133196272, |
| "loss": 0.1289, |
| "step": 48450 |
| }, |
| { |
| "epoch": 0.9835733117014804, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.0002826902301592899, |
| "loss": 0.1473, |
| "step": 48500 |
| }, |
| { |
| "epoch": 0.9845873048063273, |
| "grad_norm": 0.416015625, |
| "learning_rate": 0.0002826432873197083, |
| "loss": 0.1496, |
| "step": 48550 |
| }, |
| { |
| "epoch": 0.9856012979111742, |
| "grad_norm": 0.609375, |
| "learning_rate": 0.0002825962848219956, |
| "loss": 0.1412, |
| "step": 48600 |
| }, |
| { |
| "epoch": 0.9866152910160211, |
| "grad_norm": 0.4375, |
| "learning_rate": 0.00028254922268729183, |
| "loss": 0.1349, |
| "step": 48650 |
| }, |
| { |
| "epoch": 0.9876292841208679, |
| "grad_norm": 0.466796875, |
| "learning_rate": 0.0002825021009367638, |
| "loss": 0.1315, |
| "step": 48700 |
| }, |
| { |
| "epoch": 0.9886432772257149, |
| "grad_norm": 0.58203125, |
| "learning_rate": 0.0002824549195916052, |
| "loss": 0.1331, |
| "step": 48750 |
| }, |
| { |
| "epoch": 0.9896572703305617, |
| "grad_norm": 0.53125, |
| "learning_rate": 0.00028240767867303637, |
| "loss": 0.1371, |
| "step": 48800 |
| }, |
| { |
| "epoch": 0.9906712634354087, |
| "grad_norm": 0.462890625, |
| "learning_rate": 0.0002823603782023047, |
| "loss": 0.1465, |
| "step": 48850 |
| }, |
| { |
| "epoch": 0.9916852565402555, |
| "grad_norm": 0.42578125, |
| "learning_rate": 0.0002823130182006842, |
| "loss": 0.1319, |
| "step": 48900 |
| }, |
| { |
| "epoch": 0.9926992496451024, |
| "grad_norm": 0.490234375, |
| "learning_rate": 0.00028226559868947557, |
| "loss": 0.1367, |
| "step": 48950 |
| }, |
| { |
| "epoch": 0.9937132427499493, |
| "grad_norm": 0.53125, |
| "learning_rate": 0.00028221811969000654, |
| "loss": 0.1442, |
| "step": 49000 |
| }, |
| { |
| "epoch": 0.9947272358547962, |
| "grad_norm": 0.486328125, |
| "learning_rate": 0.0002821705812236313, |
| "loss": 0.127, |
| "step": 49050 |
| }, |
| { |
| "epoch": 0.9957412289596431, |
| "grad_norm": 0.53515625, |
| "learning_rate": 0.000282122983311731, |
| "loss": 0.1355, |
| "step": 49100 |
| }, |
| { |
| "epoch": 0.99675522206449, |
| "grad_norm": 0.466796875, |
| "learning_rate": 0.00028207532597571345, |
| "loss": 0.1403, |
| "step": 49150 |
| }, |
| { |
| "epoch": 0.9977692151693368, |
| "grad_norm": 0.515625, |
| "learning_rate": 0.0002820276092370132, |
| "loss": 0.1313, |
| "step": 49200 |
| }, |
| { |
| "epoch": 0.9987832082741838, |
| "grad_norm": 0.453125, |
| "learning_rate": 0.0002819798331170914, |
| "loss": 0.1301, |
| "step": 49250 |
| }, |
| { |
| "epoch": 0.9997972013790306, |
| "grad_norm": 0.49609375, |
| "learning_rate": 0.0002819319976374361, |
| "loss": 0.1437, |
| "step": 49300 |
| }, |
| { |
| "epoch": 1.0008111944838776, |
| "grad_norm": 0.51953125, |
| "learning_rate": 0.000281884102819562, |
| "loss": 0.1182, |
| "step": 49350 |
| }, |
| { |
| "epoch": 1.0018251875887243, |
| "grad_norm": 0.439453125, |
| "learning_rate": 0.00028183614868501037, |
| "loss": 0.1136, |
| "step": 49400 |
| }, |
| { |
| "epoch": 1.0028391806935713, |
| "grad_norm": 0.4140625, |
| "learning_rate": 0.0002817881352553493, |
| "loss": 0.1285, |
| "step": 49450 |
| }, |
| { |
| "epoch": 1.0038531737984182, |
| "grad_norm": 0.5390625, |
| "learning_rate": 0.00028174006255217345, |
| "loss": 0.1181, |
| "step": 49500 |
| }, |
| { |
| "epoch": 1.0038531737984182, |
| "eval_cer": 0.03714358544845417, |
| "eval_loss": 0.17951156198978424, |
| "eval_runtime": 123.1021, |
| "eval_samples_per_second": 22.81, |
| "eval_steps_per_second": 0.179, |
| "eval_wer": 0.10264743719953957, |
| "step": 49500 |
| }, |
| { |
| "epoch": 1.0048671669032652, |
| "grad_norm": 0.56640625, |
| "learning_rate": 0.00028169193059710413, |
| "loss": 0.1281, |
| "step": 49550 |
| }, |
| { |
| "epoch": 1.005881160008112, |
| "grad_norm": 0.443359375, |
| "learning_rate": 0.00028164373941178943, |
| "loss": 0.1142, |
| "step": 49600 |
| }, |
| { |
| "epoch": 1.0068951531129589, |
| "grad_norm": 0.53515625, |
| "learning_rate": 0.00028159548901790396, |
| "loss": 0.12, |
| "step": 49650 |
| }, |
| { |
| "epoch": 1.0079091462178058, |
| "grad_norm": 0.54296875, |
| "learning_rate": 0.00028154717943714894, |
| "loss": 0.1161, |
| "step": 49700 |
| }, |
| { |
| "epoch": 1.0089231393226525, |
| "grad_norm": 0.4765625, |
| "learning_rate": 0.00028149881069125234, |
| "loss": 0.125, |
| "step": 49750 |
| }, |
| { |
| "epoch": 1.0099371324274995, |
| "grad_norm": 0.52734375, |
| "learning_rate": 0.00028145038280196866, |
| "loss": 0.1093, |
| "step": 49800 |
| }, |
| { |
| "epoch": 1.0109511255323465, |
| "grad_norm": 0.423828125, |
| "learning_rate": 0.000281401895791079, |
| "loss": 0.1106, |
| "step": 49850 |
| }, |
| { |
| "epoch": 1.0119651186371932, |
| "grad_norm": 0.439453125, |
| "learning_rate": 0.000281353349680391, |
| "loss": 0.1161, |
| "step": 49900 |
| }, |
| { |
| "epoch": 1.0129791117420401, |
| "grad_norm": 0.486328125, |
| "learning_rate": 0.000281304744491739, |
| "loss": 0.1113, |
| "step": 49950 |
| }, |
| { |
| "epoch": 1.013993104846887, |
| "grad_norm": 0.53515625, |
| "learning_rate": 0.00028125608024698383, |
| "loss": 0.1167, |
| "step": 50000 |
| }, |
| { |
| "epoch": 1.0150070979517338, |
| "grad_norm": 0.57421875, |
| "learning_rate": 0.0002812073569680128, |
| "loss": 0.1201, |
| "step": 50050 |
| }, |
| { |
| "epoch": 1.0160210910565808, |
| "grad_norm": 0.421875, |
| "learning_rate": 0.0002811585746767401, |
| "loss": 0.1118, |
| "step": 50100 |
| }, |
| { |
| "epoch": 1.0170350841614277, |
| "grad_norm": 0.498046875, |
| "learning_rate": 0.000281109733395106, |
| "loss": 0.125, |
| "step": 50150 |
| }, |
| { |
| "epoch": 1.0180490772662747, |
| "grad_norm": 0.53125, |
| "learning_rate": 0.00028106083314507774, |
| "loss": 0.1188, |
| "step": 50200 |
| }, |
| { |
| "epoch": 1.0190630703711214, |
| "grad_norm": 0.451171875, |
| "learning_rate": 0.0002810118739486487, |
| "loss": 0.1068, |
| "step": 50250 |
| }, |
| { |
| "epoch": 1.0200770634759684, |
| "grad_norm": 0.51953125, |
| "learning_rate": 0.00028096285582783904, |
| "loss": 0.1147, |
| "step": 50300 |
| }, |
| { |
| "epoch": 1.0210910565808153, |
| "grad_norm": 0.515625, |
| "learning_rate": 0.00028091377880469525, |
| "loss": 0.119, |
| "step": 50350 |
| }, |
| { |
| "epoch": 1.022105049685662, |
| "grad_norm": 0.5078125, |
| "learning_rate": 0.00028086464290129044, |
| "loss": 0.1162, |
| "step": 50400 |
| }, |
| { |
| "epoch": 1.023119042790509, |
| "grad_norm": 0.41796875, |
| "learning_rate": 0.00028081544813972424, |
| "loss": 0.1157, |
| "step": 50450 |
| }, |
| { |
| "epoch": 1.024133035895356, |
| "grad_norm": 0.47265625, |
| "learning_rate": 0.00028076619454212254, |
| "loss": 0.1114, |
| "step": 50500 |
| }, |
| { |
| "epoch": 1.0251470290002027, |
| "grad_norm": 0.423828125, |
| "learning_rate": 0.0002807168821306378, |
| "loss": 0.1102, |
| "step": 50550 |
| }, |
| { |
| "epoch": 1.0261610221050497, |
| "grad_norm": 0.515625, |
| "learning_rate": 0.0002806675109274491, |
| "loss": 0.1135, |
| "step": 50600 |
| }, |
| { |
| "epoch": 1.0271750152098966, |
| "grad_norm": 0.4609375, |
| "learning_rate": 0.0002806180809547617, |
| "loss": 0.1136, |
| "step": 50650 |
| }, |
| { |
| "epoch": 1.0281890083147434, |
| "grad_norm": 0.59375, |
| "learning_rate": 0.0002805685922348075, |
| "loss": 0.1217, |
| "step": 50700 |
| }, |
| { |
| "epoch": 1.0292030014195903, |
| "grad_norm": 0.5078125, |
| "learning_rate": 0.0002805190447898446, |
| "loss": 0.1211, |
| "step": 50750 |
| }, |
| { |
| "epoch": 1.0302169945244373, |
| "grad_norm": 0.55859375, |
| "learning_rate": 0.0002804694386421577, |
| "loss": 0.131, |
| "step": 50800 |
| }, |
| { |
| "epoch": 1.0312309876292842, |
| "grad_norm": 0.466796875, |
| "learning_rate": 0.0002804197738140578, |
| "loss": 0.1207, |
| "step": 50850 |
| }, |
| { |
| "epoch": 1.032244980734131, |
| "grad_norm": 0.5625, |
| "learning_rate": 0.0002803700503278825, |
| "loss": 0.1171, |
| "step": 50900 |
| }, |
| { |
| "epoch": 1.033258973838978, |
| "grad_norm": 0.4453125, |
| "learning_rate": 0.0002803202682059954, |
| "loss": 0.1112, |
| "step": 50950 |
| }, |
| { |
| "epoch": 1.0342729669438249, |
| "grad_norm": 0.41015625, |
| "learning_rate": 0.00028027042747078685, |
| "loss": 0.1101, |
| "step": 51000 |
| }, |
| { |
| "epoch": 1.0342729669438249, |
| "eval_cer": 0.030817731119977942, |
| "eval_loss": 0.18131904304027557, |
| "eval_runtime": 126.3098, |
| "eval_samples_per_second": 22.231, |
| "eval_steps_per_second": 0.174, |
| "eval_wer": 0.09763694224388923, |
| "step": 51000 |
| }, |
| { |
| "epoch": 1.0352869600486716, |
| "grad_norm": 0.470703125, |
| "learning_rate": 0.0002802205281446733, |
| "loss": 0.1144, |
| "step": 51050 |
| }, |
| { |
| "epoch": 1.0363009531535186, |
| "grad_norm": 0.50390625, |
| "learning_rate": 0.0002801705702500977, |
| "loss": 0.1174, |
| "step": 51100 |
| }, |
| { |
| "epoch": 1.0373149462583655, |
| "grad_norm": 0.515625, |
| "learning_rate": 0.0002801205538095293, |
| "loss": 0.1229, |
| "step": 51150 |
| }, |
| { |
| "epoch": 1.0383289393632122, |
| "grad_norm": 0.49609375, |
| "learning_rate": 0.00028007047884546365, |
| "loss": 0.1153, |
| "step": 51200 |
| }, |
| { |
| "epoch": 1.0393429324680592, |
| "grad_norm": 0.458984375, |
| "learning_rate": 0.00028002034538042263, |
| "loss": 0.1263, |
| "step": 51250 |
| }, |
| { |
| "epoch": 1.0403569255729062, |
| "grad_norm": 0.48046875, |
| "learning_rate": 0.00027997015343695454, |
| "loss": 0.1095, |
| "step": 51300 |
| }, |
| { |
| "epoch": 1.0413709186777529, |
| "grad_norm": 0.46875, |
| "learning_rate": 0.0002799199030376339, |
| "loss": 0.1166, |
| "step": 51350 |
| }, |
| { |
| "epoch": 1.0423849117825998, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.0002798695942050613, |
| "loss": 0.1277, |
| "step": 51400 |
| }, |
| { |
| "epoch": 1.0433989048874468, |
| "grad_norm": 0.498046875, |
| "learning_rate": 0.00027981922696186407, |
| "loss": 0.1101, |
| "step": 51450 |
| }, |
| { |
| "epoch": 1.0444128979922938, |
| "grad_norm": 0.515625, |
| "learning_rate": 0.00027976880133069555, |
| "loss": 0.1212, |
| "step": 51500 |
| }, |
| { |
| "epoch": 1.0454268910971405, |
| "grad_norm": 0.4921875, |
| "learning_rate": 0.00027971831733423515, |
| "loss": 0.1169, |
| "step": 51550 |
| }, |
| { |
| "epoch": 1.0464408842019874, |
| "grad_norm": 0.47265625, |
| "learning_rate": 0.00027966777499518895, |
| "loss": 0.1097, |
| "step": 51600 |
| }, |
| { |
| "epoch": 1.0474548773068344, |
| "grad_norm": 0.51171875, |
| "learning_rate": 0.0002796171743362889, |
| "loss": 0.1184, |
| "step": 51650 |
| }, |
| { |
| "epoch": 1.0484688704116811, |
| "grad_norm": 0.494140625, |
| "learning_rate": 0.00027956651538029343, |
| "loss": 0.1165, |
| "step": 51700 |
| }, |
| { |
| "epoch": 1.049482863516528, |
| "grad_norm": 0.490234375, |
| "learning_rate": 0.00027951579814998717, |
| "loss": 0.1223, |
| "step": 51750 |
| }, |
| { |
| "epoch": 1.050496856621375, |
| "grad_norm": 0.50390625, |
| "learning_rate": 0.00027946502266818076, |
| "loss": 0.1151, |
| "step": 51800 |
| }, |
| { |
| "epoch": 1.0515108497262218, |
| "grad_norm": 0.478515625, |
| "learning_rate": 0.00027941418895771115, |
| "loss": 0.1116, |
| "step": 51850 |
| }, |
| { |
| "epoch": 1.0525248428310687, |
| "grad_norm": 0.462890625, |
| "learning_rate": 0.0002793632970414416, |
| "loss": 0.1182, |
| "step": 51900 |
| }, |
| { |
| "epoch": 1.0535388359359157, |
| "grad_norm": 0.51953125, |
| "learning_rate": 0.00027931234694226136, |
| "loss": 0.117, |
| "step": 51950 |
| }, |
| { |
| "epoch": 1.0545528290407624, |
| "grad_norm": 0.486328125, |
| "learning_rate": 0.00027926133868308597, |
| "loss": 0.1206, |
| "step": 52000 |
| }, |
| { |
| "epoch": 1.0555668221456094, |
| "grad_norm": 0.73046875, |
| "learning_rate": 0.00027921027228685713, |
| "loss": 0.1191, |
| "step": 52050 |
| }, |
| { |
| "epoch": 1.0565808152504563, |
| "grad_norm": 0.48046875, |
| "learning_rate": 0.0002791591477765426, |
| "loss": 0.1166, |
| "step": 52100 |
| }, |
| { |
| "epoch": 1.0575948083553033, |
| "grad_norm": 0.4609375, |
| "learning_rate": 0.00027910796517513633, |
| "loss": 0.122, |
| "step": 52150 |
| }, |
| { |
| "epoch": 1.05860880146015, |
| "grad_norm": 0.50390625, |
| "learning_rate": 0.0002790567245056584, |
| "loss": 0.1101, |
| "step": 52200 |
| }, |
| { |
| "epoch": 1.059622794564997, |
| "grad_norm": 0.47265625, |
| "learning_rate": 0.00027900542579115497, |
| "loss": 0.1227, |
| "step": 52250 |
| }, |
| { |
| "epoch": 1.060636787669844, |
| "grad_norm": 0.63671875, |
| "learning_rate": 0.00027895406905469847, |
| "loss": 0.1106, |
| "step": 52300 |
| }, |
| { |
| "epoch": 1.0616507807746907, |
| "grad_norm": 0.8828125, |
| "learning_rate": 0.0002789026543193871, |
| "loss": 0.1257, |
| "step": 52350 |
| }, |
| { |
| "epoch": 1.0626647738795376, |
| "grad_norm": 0.49609375, |
| "learning_rate": 0.0002788511816083455, |
| "loss": 0.1105, |
| "step": 52400 |
| }, |
| { |
| "epoch": 1.0636787669843846, |
| "grad_norm": 0.435546875, |
| "learning_rate": 0.0002787996509447241, |
| "loss": 0.1144, |
| "step": 52450 |
| }, |
| { |
| "epoch": 1.0646927600892313, |
| "grad_norm": 0.498046875, |
| "learning_rate": 0.00027874806235169964, |
| "loss": 0.1158, |
| "step": 52500 |
| }, |
| { |
| "epoch": 1.0646927600892313, |
| "eval_cer": 0.031348353769241576, |
| "eval_loss": 0.18024787306785583, |
| "eval_runtime": 126.6593, |
| "eval_samples_per_second": 22.17, |
| "eval_steps_per_second": 0.174, |
| "eval_wer": 0.09773850633082809, |
| "step": 52500 |
| }, |
| { |
| "epoch": 1.0657067531940783, |
| "grad_norm": 0.443359375, |
| "learning_rate": 0.0002786964158524747, |
| "loss": 0.1189, |
| "step": 52550 |
| }, |
| { |
| "epoch": 1.0667207462989252, |
| "grad_norm": 0.52734375, |
| "learning_rate": 0.0002786447114702781, |
| "loss": 0.1121, |
| "step": 52600 |
| }, |
| { |
| "epoch": 1.0677347394037722, |
| "grad_norm": 0.484375, |
| "learning_rate": 0.0002785929492283644, |
| "loss": 0.1214, |
| "step": 52650 |
| }, |
| { |
| "epoch": 1.068748732508619, |
| "grad_norm": 0.50390625, |
| "learning_rate": 0.00027854112915001456, |
| "loss": 0.1259, |
| "step": 52700 |
| }, |
| { |
| "epoch": 1.0697627256134659, |
| "grad_norm": 0.58203125, |
| "learning_rate": 0.00027848925125853534, |
| "loss": 0.1195, |
| "step": 52750 |
| }, |
| { |
| "epoch": 1.0707767187183128, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.0002784373155772595, |
| "loss": 0.1177, |
| "step": 52800 |
| }, |
| { |
| "epoch": 1.0717907118231595, |
| "grad_norm": 0.427734375, |
| "learning_rate": 0.0002783853221295458, |
| "loss": 0.1259, |
| "step": 52850 |
| }, |
| { |
| "epoch": 1.0728047049280065, |
| "grad_norm": 0.5078125, |
| "learning_rate": 0.00027833327093877903, |
| "loss": 0.1113, |
| "step": 52900 |
| }, |
| { |
| "epoch": 1.0738186980328535, |
| "grad_norm": 0.451171875, |
| "learning_rate": 0.00027828116202836996, |
| "loss": 0.12, |
| "step": 52950 |
| }, |
| { |
| "epoch": 1.0748326911377002, |
| "grad_norm": 0.439453125, |
| "learning_rate": 0.00027822899542175516, |
| "loss": 0.1148, |
| "step": 53000 |
| }, |
| { |
| "epoch": 1.0758466842425471, |
| "grad_norm": 0.5, |
| "learning_rate": 0.00027817677114239743, |
| "loss": 0.1225, |
| "step": 53050 |
| }, |
| { |
| "epoch": 1.076860677347394, |
| "grad_norm": 0.50390625, |
| "learning_rate": 0.0002781244892137853, |
| "loss": 0.1148, |
| "step": 53100 |
| }, |
| { |
| "epoch": 1.0778746704522408, |
| "grad_norm": 0.515625, |
| "learning_rate": 0.0002780721496594332, |
| "loss": 0.1228, |
| "step": 53150 |
| }, |
| { |
| "epoch": 1.0788886635570878, |
| "grad_norm": 0.48046875, |
| "learning_rate": 0.0002780197525028817, |
| "loss": 0.1193, |
| "step": 53200 |
| }, |
| { |
| "epoch": 1.0799026566619347, |
| "grad_norm": 0.435546875, |
| "learning_rate": 0.000277967297767697, |
| "loss": 0.119, |
| "step": 53250 |
| }, |
| { |
| "epoch": 1.0809166497667815, |
| "grad_norm": 0.5625, |
| "learning_rate": 0.00027791478547747146, |
| "loss": 0.1126, |
| "step": 53300 |
| }, |
| { |
| "epoch": 1.0819306428716284, |
| "grad_norm": 0.5625, |
| "learning_rate": 0.00027786221565582304, |
| "loss": 0.1292, |
| "step": 53350 |
| }, |
| { |
| "epoch": 1.0829446359764754, |
| "grad_norm": 0.5703125, |
| "learning_rate": 0.0002778095883263959, |
| "loss": 0.1131, |
| "step": 53400 |
| }, |
| { |
| "epoch": 1.0839586290813223, |
| "grad_norm": 0.478515625, |
| "learning_rate": 0.00027775690351285984, |
| "loss": 0.1225, |
| "step": 53450 |
| }, |
| { |
| "epoch": 1.084972622186169, |
| "grad_norm": 0.47265625, |
| "learning_rate": 0.00027770416123891055, |
| "loss": 0.1154, |
| "step": 53500 |
| }, |
| { |
| "epoch": 1.085986615291016, |
| "grad_norm": 0.490234375, |
| "learning_rate": 0.00027765136152826963, |
| "loss": 0.1089, |
| "step": 53550 |
| }, |
| { |
| "epoch": 1.087000608395863, |
| "grad_norm": 0.515625, |
| "learning_rate": 0.0002775985044046844, |
| "loss": 0.1189, |
| "step": 53600 |
| }, |
| { |
| "epoch": 1.0880146015007097, |
| "grad_norm": 0.484375, |
| "learning_rate": 0.00027754558989192814, |
| "loss": 0.1189, |
| "step": 53650 |
| }, |
| { |
| "epoch": 1.0890285946055567, |
| "grad_norm": 0.5, |
| "learning_rate": 0.00027749261801379986, |
| "loss": 0.1354, |
| "step": 53700 |
| }, |
| { |
| "epoch": 1.0900425877104036, |
| "grad_norm": 0.5546875, |
| "learning_rate": 0.0002774395887941244, |
| "loss": 0.1146, |
| "step": 53750 |
| }, |
| { |
| "epoch": 1.0910565808152504, |
| "grad_norm": 0.453125, |
| "learning_rate": 0.0002773865022567524, |
| "loss": 0.1173, |
| "step": 53800 |
| }, |
| { |
| "epoch": 1.0920705739200973, |
| "grad_norm": 0.5, |
| "learning_rate": 0.0002773333584255601, |
| "loss": 0.1121, |
| "step": 53850 |
| }, |
| { |
| "epoch": 1.0930845670249443, |
| "grad_norm": 0.55859375, |
| "learning_rate": 0.0002772801573244499, |
| "loss": 0.1281, |
| "step": 53900 |
| }, |
| { |
| "epoch": 1.0940985601297912, |
| "grad_norm": 0.486328125, |
| "learning_rate": 0.0002772268989773495, |
| "loss": 0.1216, |
| "step": 53950 |
| }, |
| { |
| "epoch": 1.095112553234638, |
| "grad_norm": 0.49609375, |
| "learning_rate": 0.0002771735834082127, |
| "loss": 0.1193, |
| "step": 54000 |
| }, |
| { |
| "epoch": 1.095112553234638, |
| "eval_cer": 0.031686495653576244, |
| "eval_loss": 0.18309062719345093, |
| "eval_runtime": 129.5836, |
| "eval_samples_per_second": 21.669, |
| "eval_steps_per_second": 0.17, |
| "eval_wer": 0.09777236102647437, |
| "step": 54000 |
| }, |
| { |
| "epoch": 1.096126546339485, |
| "grad_norm": 0.47265625, |
| "learning_rate": 0.0002771202106410189, |
| "loss": 0.1112, |
| "step": 54050 |
| }, |
| { |
| "epoch": 1.0971405394443319, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.0002770667806997732, |
| "loss": 0.1232, |
| "step": 54100 |
| }, |
| { |
| "epoch": 1.0981545325491786, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.00027701329360850643, |
| "loss": 0.1179, |
| "step": 54150 |
| }, |
| { |
| "epoch": 1.0991685256540256, |
| "grad_norm": 0.435546875, |
| "learning_rate": 0.00027695974939127523, |
| "loss": 0.114, |
| "step": 54200 |
| }, |
| { |
| "epoch": 1.1001825187588725, |
| "grad_norm": 0.482421875, |
| "learning_rate": 0.0002769061480721617, |
| "loss": 0.1169, |
| "step": 54250 |
| }, |
| { |
| "epoch": 1.1011965118637193, |
| "grad_norm": 0.51953125, |
| "learning_rate": 0.00027685248967527387, |
| "loss": 0.124, |
| "step": 54300 |
| }, |
| { |
| "epoch": 1.1022105049685662, |
| "grad_norm": 0.56640625, |
| "learning_rate": 0.0002767987742247453, |
| "loss": 0.1214, |
| "step": 54350 |
| }, |
| { |
| "epoch": 1.1032244980734132, |
| "grad_norm": 0.43359375, |
| "learning_rate": 0.00027674500174473533, |
| "loss": 0.124, |
| "step": 54400 |
| }, |
| { |
| "epoch": 1.10423849117826, |
| "grad_norm": 0.490234375, |
| "learning_rate": 0.00027669117225942875, |
| "loss": 0.1162, |
| "step": 54450 |
| }, |
| { |
| "epoch": 1.1052524842831069, |
| "grad_norm": 0.58203125, |
| "learning_rate": 0.0002766372857930361, |
| "loss": 0.1246, |
| "step": 54500 |
| }, |
| { |
| "epoch": 1.1062664773879538, |
| "grad_norm": 0.498046875, |
| "learning_rate": 0.0002765833423697937, |
| "loss": 0.1143, |
| "step": 54550 |
| }, |
| { |
| "epoch": 1.1072804704928005, |
| "grad_norm": 0.435546875, |
| "learning_rate": 0.0002765293420139632, |
| "loss": 0.1126, |
| "step": 54600 |
| }, |
| { |
| "epoch": 1.1082944635976475, |
| "grad_norm": 0.578125, |
| "learning_rate": 0.00027647528474983204, |
| "loss": 0.1118, |
| "step": 54650 |
| }, |
| { |
| "epoch": 1.1093084567024944, |
| "grad_norm": 0.51171875, |
| "learning_rate": 0.0002764211706017132, |
| "loss": 0.1158, |
| "step": 54700 |
| }, |
| { |
| "epoch": 1.1103224498073414, |
| "grad_norm": 0.466796875, |
| "learning_rate": 0.0002763669995939453, |
| "loss": 0.1147, |
| "step": 54750 |
| }, |
| { |
| "epoch": 1.1113364429121881, |
| "grad_norm": 0.423828125, |
| "learning_rate": 0.00027631277175089254, |
| "loss": 0.1136, |
| "step": 54800 |
| }, |
| { |
| "epoch": 1.112350436017035, |
| "grad_norm": 0.51953125, |
| "learning_rate": 0.00027625848709694447, |
| "loss": 0.1218, |
| "step": 54850 |
| }, |
| { |
| "epoch": 1.113364429121882, |
| "grad_norm": 0.466796875, |
| "learning_rate": 0.00027620414565651647, |
| "loss": 0.115, |
| "step": 54900 |
| }, |
| { |
| "epoch": 1.1143784222267288, |
| "grad_norm": 0.484375, |
| "learning_rate": 0.00027614974745404934, |
| "loss": 0.1181, |
| "step": 54950 |
| }, |
| { |
| "epoch": 1.1153924153315757, |
| "grad_norm": 0.515625, |
| "learning_rate": 0.0002760952925140094, |
| "loss": 0.1186, |
| "step": 55000 |
| }, |
| { |
| "epoch": 1.1164064084364227, |
| "grad_norm": 0.51953125, |
| "learning_rate": 0.00027604078086088845, |
| "loss": 0.1163, |
| "step": 55050 |
| }, |
| { |
| "epoch": 1.1174204015412694, |
| "grad_norm": 0.53125, |
| "learning_rate": 0.000275986212519204, |
| "loss": 0.1153, |
| "step": 55100 |
| }, |
| { |
| "epoch": 1.1184343946461164, |
| "grad_norm": 0.431640625, |
| "learning_rate": 0.00027593158751349877, |
| "loss": 0.1123, |
| "step": 55150 |
| }, |
| { |
| "epoch": 1.1194483877509633, |
| "grad_norm": 0.5078125, |
| "learning_rate": 0.0002758769058683411, |
| "loss": 0.1121, |
| "step": 55200 |
| }, |
| { |
| "epoch": 1.1204623808558103, |
| "grad_norm": 0.80078125, |
| "learning_rate": 0.0002758221676083249, |
| "loss": 0.1217, |
| "step": 55250 |
| }, |
| { |
| "epoch": 1.121476373960657, |
| "grad_norm": 0.484375, |
| "learning_rate": 0.0002757673727580693, |
| "loss": 0.1168, |
| "step": 55300 |
| }, |
| { |
| "epoch": 1.122490367065504, |
| "grad_norm": 0.51171875, |
| "learning_rate": 0.00027571252134221924, |
| "loss": 0.1136, |
| "step": 55350 |
| }, |
| { |
| "epoch": 1.123504360170351, |
| "grad_norm": 0.451171875, |
| "learning_rate": 0.00027565761338544477, |
| "loss": 0.1174, |
| "step": 55400 |
| }, |
| { |
| "epoch": 1.1245183532751977, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.0002756026489124415, |
| "loss": 0.1198, |
| "step": 55450 |
| }, |
| { |
| "epoch": 1.1255323463800446, |
| "grad_norm": 0.4609375, |
| "learning_rate": 0.0002755476279479304, |
| "loss": 0.1197, |
| "step": 55500 |
| }, |
| { |
| "epoch": 1.1255323463800446, |
| "eval_cer": 0.03313270248196143, |
| "eval_loss": 0.17757254838943481, |
| "eval_runtime": 123.3499, |
| "eval_samples_per_second": 22.765, |
| "eval_steps_per_second": 0.178, |
| "eval_wer": 0.09929582233055725, |
| "step": 55500 |
| }, |
| { |
| "epoch": 1.1265463394848916, |
| "grad_norm": 0.4921875, |
| "learning_rate": 0.0002754925505166581, |
| "loss": 0.1192, |
| "step": 55550 |
| }, |
| { |
| "epoch": 1.1275603325897383, |
| "grad_norm": 0.478515625, |
| "learning_rate": 0.0002754374166433962, |
| "loss": 0.1162, |
| "step": 55600 |
| }, |
| { |
| "epoch": 1.1285743256945853, |
| "grad_norm": 0.455078125, |
| "learning_rate": 0.00027538222635294205, |
| "loss": 0.1115, |
| "step": 55650 |
| }, |
| { |
| "epoch": 1.1295883187994322, |
| "grad_norm": 0.48046875, |
| "learning_rate": 0.0002753269796701181, |
| "loss": 0.1084, |
| "step": 55700 |
| }, |
| { |
| "epoch": 1.130602311904279, |
| "grad_norm": 0.57421875, |
| "learning_rate": 0.00027527167661977247, |
| "loss": 0.1251, |
| "step": 55750 |
| }, |
| { |
| "epoch": 1.131616305009126, |
| "grad_norm": 0.494140625, |
| "learning_rate": 0.0002752163172267784, |
| "loss": 0.1282, |
| "step": 55800 |
| }, |
| { |
| "epoch": 1.1326302981139729, |
| "grad_norm": 0.51171875, |
| "learning_rate": 0.00027516090151603445, |
| "loss": 0.1097, |
| "step": 55850 |
| }, |
| { |
| "epoch": 1.1336442912188196, |
| "grad_norm": 0.47265625, |
| "learning_rate": 0.0002751054295124647, |
| "loss": 0.1191, |
| "step": 55900 |
| }, |
| { |
| "epoch": 1.1346582843236666, |
| "grad_norm": 0.55078125, |
| "learning_rate": 0.0002750499012410184, |
| "loss": 0.1082, |
| "step": 55950 |
| }, |
| { |
| "epoch": 1.1356722774285135, |
| "grad_norm": 0.462890625, |
| "learning_rate": 0.0002749943167266702, |
| "loss": 0.1229, |
| "step": 56000 |
| }, |
| { |
| "epoch": 1.1366862705333605, |
| "grad_norm": 0.4921875, |
| "learning_rate": 0.00027493867599441985, |
| "loss": 0.1107, |
| "step": 56050 |
| }, |
| { |
| "epoch": 1.1377002636382072, |
| "grad_norm": 0.56640625, |
| "learning_rate": 0.0002748829790692927, |
| "loss": 0.1143, |
| "step": 56100 |
| }, |
| { |
| "epoch": 1.1387142567430542, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.0002748272259763391, |
| "loss": 0.1139, |
| "step": 56150 |
| }, |
| { |
| "epoch": 1.139728249847901, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.00027477141674063483, |
| "loss": 0.1206, |
| "step": 56200 |
| }, |
| { |
| "epoch": 1.1407422429527478, |
| "grad_norm": 0.58203125, |
| "learning_rate": 0.00027471555138728075, |
| "loss": 0.1178, |
| "step": 56250 |
| }, |
| { |
| "epoch": 1.1417562360575948, |
| "grad_norm": 0.44921875, |
| "learning_rate": 0.0002746596299414032, |
| "loss": 0.1115, |
| "step": 56300 |
| }, |
| { |
| "epoch": 1.1427702291624418, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.00027460365242815355, |
| "loss": 0.1184, |
| "step": 56350 |
| }, |
| { |
| "epoch": 1.1437842222672885, |
| "grad_norm": 0.64453125, |
| "learning_rate": 0.00027454761887270844, |
| "loss": 0.1124, |
| "step": 56400 |
| }, |
| { |
| "epoch": 1.1447982153721354, |
| "grad_norm": 0.46484375, |
| "learning_rate": 0.00027449152930026976, |
| "loss": 0.1177, |
| "step": 56450 |
| }, |
| { |
| "epoch": 1.1458122084769824, |
| "grad_norm": 0.498046875, |
| "learning_rate": 0.00027443538373606453, |
| "loss": 0.1235, |
| "step": 56500 |
| }, |
| { |
| "epoch": 1.1468262015818294, |
| "grad_norm": 0.484375, |
| "learning_rate": 0.000274379182205345, |
| "loss": 0.1216, |
| "step": 56550 |
| }, |
| { |
| "epoch": 1.147840194686676, |
| "grad_norm": 0.55078125, |
| "learning_rate": 0.00027432292473338857, |
| "loss": 0.1152, |
| "step": 56600 |
| }, |
| { |
| "epoch": 1.148854187791523, |
| "grad_norm": 0.66796875, |
| "learning_rate": 0.0002742666113454978, |
| "loss": 0.1182, |
| "step": 56650 |
| }, |
| { |
| "epoch": 1.14986818089637, |
| "grad_norm": 0.515625, |
| "learning_rate": 0.0002742102420670004, |
| "loss": 0.1146, |
| "step": 56700 |
| }, |
| { |
| "epoch": 1.1508821740012167, |
| "grad_norm": 0.4609375, |
| "learning_rate": 0.0002741538169232493, |
| "loss": 0.1144, |
| "step": 56750 |
| }, |
| { |
| "epoch": 1.1518961671060637, |
| "grad_norm": 0.439453125, |
| "learning_rate": 0.00027409733593962237, |
| "loss": 0.1145, |
| "step": 56800 |
| }, |
| { |
| "epoch": 1.1529101602109106, |
| "grad_norm": 0.416015625, |
| "learning_rate": 0.0002740407991415228, |
| "loss": 0.1088, |
| "step": 56850 |
| }, |
| { |
| "epoch": 1.1539241533157574, |
| "grad_norm": 0.5390625, |
| "learning_rate": 0.00027398420655437865, |
| "loss": 0.114, |
| "step": 56900 |
| }, |
| { |
| "epoch": 1.1549381464206043, |
| "grad_norm": 0.484375, |
| "learning_rate": 0.0002739275582036434, |
| "loss": 0.1179, |
| "step": 56950 |
| }, |
| { |
| "epoch": 1.1559521395254513, |
| "grad_norm": 0.50390625, |
| "learning_rate": 0.0002738708541147953, |
| "loss": 0.1235, |
| "step": 57000 |
| }, |
| { |
| "epoch": 1.1559521395254513, |
| "eval_cer": 0.03642048203426158, |
| "eval_loss": 0.17667804658412933, |
| "eval_runtime": 121.0428, |
| "eval_samples_per_second": 23.198, |
| "eval_steps_per_second": 0.182, |
| "eval_wer": 0.10159794163450471, |
| "step": 57000 |
| }, |
| { |
| "epoch": 1.156966132630298, |
| "grad_norm": 0.46875, |
| "learning_rate": 0.0002738140943133378, |
| "loss": 0.1207, |
| "step": 57050 |
| }, |
| { |
| "epoch": 1.157980125735145, |
| "grad_norm": 0.55078125, |
| "learning_rate": 0.00027375727882479936, |
| "loss": 0.1203, |
| "step": 57100 |
| }, |
| { |
| "epoch": 1.158994118839992, |
| "grad_norm": 0.51171875, |
| "learning_rate": 0.0002737004076747336, |
| "loss": 0.1161, |
| "step": 57150 |
| }, |
| { |
| "epoch": 1.1600081119448387, |
| "grad_norm": 0.412109375, |
| "learning_rate": 0.0002736434808887191, |
| "loss": 0.1128, |
| "step": 57200 |
| }, |
| { |
| "epoch": 1.1610221050496856, |
| "grad_norm": 0.70703125, |
| "learning_rate": 0.0002735864984923594, |
| "loss": 0.1128, |
| "step": 57250 |
| }, |
| { |
| "epoch": 1.1620360981545326, |
| "grad_norm": 0.53515625, |
| "learning_rate": 0.0002735294605112831, |
| "loss": 0.1191, |
| "step": 57300 |
| }, |
| { |
| "epoch": 1.1630500912593795, |
| "grad_norm": 0.53515625, |
| "learning_rate": 0.0002734723669711439, |
| "loss": 0.1236, |
| "step": 57350 |
| }, |
| { |
| "epoch": 1.1640640843642263, |
| "grad_norm": 0.54296875, |
| "learning_rate": 0.0002734152178976203, |
| "loss": 0.1138, |
| "step": 57400 |
| }, |
| { |
| "epoch": 1.1650780774690732, |
| "grad_norm": 0.5546875, |
| "learning_rate": 0.0002733580133164159, |
| "loss": 0.136, |
| "step": 57450 |
| }, |
| { |
| "epoch": 1.1660920705739202, |
| "grad_norm": 0.49609375, |
| "learning_rate": 0.0002733007532532594, |
| "loss": 0.1121, |
| "step": 57500 |
| }, |
| { |
| "epoch": 1.167106063678767, |
| "grad_norm": 0.5, |
| "learning_rate": 0.0002732434377339041, |
| "loss": 0.1138, |
| "step": 57550 |
| }, |
| { |
| "epoch": 1.1681200567836139, |
| "grad_norm": 0.45703125, |
| "learning_rate": 0.0002731860667841285, |
| "loss": 0.1103, |
| "step": 57600 |
| }, |
| { |
| "epoch": 1.1691340498884608, |
| "grad_norm": 0.49609375, |
| "learning_rate": 0.0002731286404297359, |
| "loss": 0.1127, |
| "step": 57650 |
| }, |
| { |
| "epoch": 1.1701480429933078, |
| "grad_norm": 0.515625, |
| "learning_rate": 0.0002730711586965547, |
| "loss": 0.1166, |
| "step": 57700 |
| }, |
| { |
| "epoch": 1.1711620360981545, |
| "grad_norm": 0.56640625, |
| "learning_rate": 0.0002730136216104381, |
| "loss": 0.1127, |
| "step": 57750 |
| }, |
| { |
| "epoch": 1.1721760292030015, |
| "grad_norm": 0.59765625, |
| "learning_rate": 0.00027295602919726405, |
| "loss": 0.1094, |
| "step": 57800 |
| }, |
| { |
| "epoch": 1.1731900223078484, |
| "grad_norm": 0.482421875, |
| "learning_rate": 0.0002728983814829357, |
| "loss": 0.1173, |
| "step": 57850 |
| }, |
| { |
| "epoch": 1.1742040154126951, |
| "grad_norm": 0.50390625, |
| "learning_rate": 0.0002728406784933808, |
| "loss": 0.1172, |
| "step": 57900 |
| }, |
| { |
| "epoch": 1.175218008517542, |
| "grad_norm": 0.5, |
| "learning_rate": 0.00027278292025455204, |
| "loss": 0.1112, |
| "step": 57950 |
| }, |
| { |
| "epoch": 1.176232001622389, |
| "grad_norm": 0.515625, |
| "learning_rate": 0.00027272510679242704, |
| "loss": 0.1084, |
| "step": 58000 |
| }, |
| { |
| "epoch": 1.1772459947272358, |
| "grad_norm": 0.44140625, |
| "learning_rate": 0.00027266723813300814, |
| "loss": 0.1137, |
| "step": 58050 |
| }, |
| { |
| "epoch": 1.1782599878320827, |
| "grad_norm": 0.53515625, |
| "learning_rate": 0.0002726093143023226, |
| "loss": 0.1219, |
| "step": 58100 |
| }, |
| { |
| "epoch": 1.1792739809369297, |
| "grad_norm": 0.56640625, |
| "learning_rate": 0.00027255133532642245, |
| "loss": 0.1134, |
| "step": 58150 |
| }, |
| { |
| "epoch": 1.1802879740417764, |
| "grad_norm": 0.52734375, |
| "learning_rate": 0.0002724933012313845, |
| "loss": 0.1149, |
| "step": 58200 |
| }, |
| { |
| "epoch": 1.1813019671466234, |
| "grad_norm": 0.51171875, |
| "learning_rate": 0.00027243521204331044, |
| "loss": 0.1182, |
| "step": 58250 |
| }, |
| { |
| "epoch": 1.1823159602514703, |
| "grad_norm": 0.49609375, |
| "learning_rate": 0.00027237706778832657, |
| "loss": 0.1107, |
| "step": 58300 |
| }, |
| { |
| "epoch": 1.183329953356317, |
| "grad_norm": 0.51171875, |
| "learning_rate": 0.00027231886849258413, |
| "loss": 0.1106, |
| "step": 58350 |
| }, |
| { |
| "epoch": 1.184343946461164, |
| "grad_norm": 0.51171875, |
| "learning_rate": 0.00027226061418225906, |
| "loss": 0.1178, |
| "step": 58400 |
| }, |
| { |
| "epoch": 1.185357939566011, |
| "grad_norm": 0.474609375, |
| "learning_rate": 0.000272202304883552, |
| "loss": 0.1082, |
| "step": 58450 |
| }, |
| { |
| "epoch": 1.1863719326708577, |
| "grad_norm": 0.45703125, |
| "learning_rate": 0.0002721439406226884, |
| "loss": 0.1078, |
| "step": 58500 |
| }, |
| { |
| "epoch": 1.1863719326708577, |
| "eval_cer": 0.03631643837754322, |
| "eval_loss": 0.1754067838191986, |
| "eval_runtime": 118.1992, |
| "eval_samples_per_second": 23.757, |
| "eval_steps_per_second": 0.186, |
| "eval_wer": 0.09960051459137383, |
| "step": 58500 |
| }, |
| { |
| "epoch": 1.1873859257757047, |
| "grad_norm": 0.515625, |
| "learning_rate": 0.0002720855214259183, |
| "loss": 0.1299, |
| "step": 58550 |
| }, |
| { |
| "epoch": 1.1883999188805516, |
| "grad_norm": 0.45703125, |
| "learning_rate": 0.0002720270473195166, |
| "loss": 0.1295, |
| "step": 58600 |
| }, |
| { |
| "epoch": 1.1894139119853986, |
| "grad_norm": 0.490234375, |
| "learning_rate": 0.0002719685183297829, |
| "loss": 0.1291, |
| "step": 58650 |
| }, |
| { |
| "epoch": 1.1904279050902453, |
| "grad_norm": 0.5078125, |
| "learning_rate": 0.0002719099344830412, |
| "loss": 0.1129, |
| "step": 58700 |
| }, |
| { |
| "epoch": 1.1914418981950923, |
| "grad_norm": 0.48046875, |
| "learning_rate": 0.00027185129580564054, |
| "loss": 0.1265, |
| "step": 58750 |
| }, |
| { |
| "epoch": 1.1924558912999392, |
| "grad_norm": 0.458984375, |
| "learning_rate": 0.00027179260232395447, |
| "loss": 0.1219, |
| "step": 58800 |
| }, |
| { |
| "epoch": 1.193469884404786, |
| "grad_norm": 0.498046875, |
| "learning_rate": 0.00027173385406438114, |
| "loss": 0.1215, |
| "step": 58850 |
| }, |
| { |
| "epoch": 1.194483877509633, |
| "grad_norm": 0.47265625, |
| "learning_rate": 0.00027167505105334344, |
| "loss": 0.1101, |
| "step": 58900 |
| }, |
| { |
| "epoch": 1.1954978706144799, |
| "grad_norm": 0.5078125, |
| "learning_rate": 0.00027161619331728874, |
| "loss": 0.1064, |
| "step": 58950 |
| }, |
| { |
| "epoch": 1.1965118637193268, |
| "grad_norm": 0.4765625, |
| "learning_rate": 0.00027155728088268914, |
| "loss": 0.112, |
| "step": 59000 |
| }, |
| { |
| "epoch": 1.1975258568241736, |
| "grad_norm": 0.474609375, |
| "learning_rate": 0.00027149831377604135, |
| "loss": 0.1149, |
| "step": 59050 |
| }, |
| { |
| "epoch": 1.1985398499290205, |
| "grad_norm": 0.458984375, |
| "learning_rate": 0.00027143929202386657, |
| "loss": 0.107, |
| "step": 59100 |
| }, |
| { |
| "epoch": 1.1995538430338675, |
| "grad_norm": 0.4609375, |
| "learning_rate": 0.00027138021565271074, |
| "loss": 0.1205, |
| "step": 59150 |
| }, |
| { |
| "epoch": 1.2005678361387142, |
| "grad_norm": 0.484375, |
| "learning_rate": 0.00027132108468914414, |
| "loss": 0.1107, |
| "step": 59200 |
| }, |
| { |
| "epoch": 1.2015818292435612, |
| "grad_norm": 0.3984375, |
| "learning_rate": 0.0002712618991597618, |
| "loss": 0.1086, |
| "step": 59250 |
| }, |
| { |
| "epoch": 1.2025958223484081, |
| "grad_norm": 0.44140625, |
| "learning_rate": 0.0002712026590911833, |
| "loss": 0.1091, |
| "step": 59300 |
| }, |
| { |
| "epoch": 1.2036098154532548, |
| "grad_norm": 0.703125, |
| "learning_rate": 0.00027114336451005246, |
| "loss": 0.1192, |
| "step": 59350 |
| }, |
| { |
| "epoch": 1.2046238085581018, |
| "grad_norm": 0.486328125, |
| "learning_rate": 0.00027108401544303807, |
| "loss": 0.1189, |
| "step": 59400 |
| }, |
| { |
| "epoch": 1.2056378016629488, |
| "grad_norm": 0.453125, |
| "learning_rate": 0.000271024611916833, |
| "loss": 0.1178, |
| "step": 59450 |
| }, |
| { |
| "epoch": 1.2066517947677955, |
| "grad_norm": 0.474609375, |
| "learning_rate": 0.00027096515395815484, |
| "loss": 0.1118, |
| "step": 59500 |
| }, |
| { |
| "epoch": 1.2076657878726424, |
| "grad_norm": 0.52734375, |
| "learning_rate": 0.0002709056415937457, |
| "loss": 0.1199, |
| "step": 59550 |
| }, |
| { |
| "epoch": 1.2086797809774894, |
| "grad_norm": 0.470703125, |
| "learning_rate": 0.000270846074850372, |
| "loss": 0.1243, |
| "step": 59600 |
| }, |
| { |
| "epoch": 1.2096937740823361, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.0002707864537548247, |
| "loss": 0.1071, |
| "step": 59650 |
| }, |
| { |
| "epoch": 1.210707767187183, |
| "grad_norm": 0.46875, |
| "learning_rate": 0.00027072677833391923, |
| "loss": 0.1227, |
| "step": 59700 |
| }, |
| { |
| "epoch": 1.21172176029203, |
| "grad_norm": 0.4453125, |
| "learning_rate": 0.0002706670486144954, |
| "loss": 0.1176, |
| "step": 59750 |
| }, |
| { |
| "epoch": 1.2127357533968768, |
| "grad_norm": 0.50390625, |
| "learning_rate": 0.0002706072646234175, |
| "loss": 0.1136, |
| "step": 59800 |
| }, |
| { |
| "epoch": 1.2137497465017237, |
| "grad_norm": 0.72265625, |
| "learning_rate": 0.00027054742638757417, |
| "loss": 0.1143, |
| "step": 59850 |
| }, |
| { |
| "epoch": 1.2147637396065707, |
| "grad_norm": 0.38671875, |
| "learning_rate": 0.00027048753393387853, |
| "loss": 0.1238, |
| "step": 59900 |
| }, |
| { |
| "epoch": 1.2157777327114176, |
| "grad_norm": 0.453125, |
| "learning_rate": 0.00027042758728926795, |
| "loss": 0.1077, |
| "step": 59950 |
| }, |
| { |
| "epoch": 1.2167917258162644, |
| "grad_norm": 0.423828125, |
| "learning_rate": 0.0002703675864807043, |
| "loss": 0.1045, |
| "step": 60000 |
| }, |
| { |
| "epoch": 1.2167917258162644, |
| "eval_cer": 0.03392863645585688, |
| "eval_loss": 0.17490266263484955, |
| "eval_runtime": 118.1431, |
| "eval_samples_per_second": 23.768, |
| "eval_steps_per_second": 0.186, |
| "eval_wer": 0.09658744667885436, |
| "step": 60000 |
| }, |
| { |
| "epoch": 1.2178057189211113, |
| "grad_norm": 0.427734375, |
| "learning_rate": 0.0002703075315351738, |
| "loss": 0.1222, |
| "step": 60050 |
| }, |
| { |
| "epoch": 1.2188197120259583, |
| "grad_norm": 0.60546875, |
| "learning_rate": 0.00027024742247968686, |
| "loss": 0.1133, |
| "step": 60100 |
| }, |
| { |
| "epoch": 1.219833705130805, |
| "grad_norm": 0.5078125, |
| "learning_rate": 0.00027018725934127845, |
| "loss": 0.1125, |
| "step": 60150 |
| }, |
| { |
| "epoch": 1.220847698235652, |
| "grad_norm": 0.455078125, |
| "learning_rate": 0.0002701270421470077, |
| "loss": 0.1083, |
| "step": 60200 |
| }, |
| { |
| "epoch": 1.221861691340499, |
| "grad_norm": 0.447265625, |
| "learning_rate": 0.0002700667709239582, |
| "loss": 0.1113, |
| "step": 60250 |
| }, |
| { |
| "epoch": 1.222875684445346, |
| "grad_norm": 0.439453125, |
| "learning_rate": 0.0002700064456992377, |
| "loss": 0.1194, |
| "step": 60300 |
| }, |
| { |
| "epoch": 1.2238896775501926, |
| "grad_norm": 0.53125, |
| "learning_rate": 0.00026994606649997834, |
| "loss": 0.118, |
| "step": 60350 |
| }, |
| { |
| "epoch": 1.2249036706550396, |
| "grad_norm": 0.458984375, |
| "learning_rate": 0.0002698856333533364, |
| "loss": 0.1087, |
| "step": 60400 |
| }, |
| { |
| "epoch": 1.2259176637598865, |
| "grad_norm": 0.447265625, |
| "learning_rate": 0.00026982514628649254, |
| "loss": 0.1124, |
| "step": 60450 |
| }, |
| { |
| "epoch": 1.2269316568647333, |
| "grad_norm": 0.4765625, |
| "learning_rate": 0.00026976460532665173, |
| "loss": 0.1128, |
| "step": 60500 |
| }, |
| { |
| "epoch": 1.2279456499695802, |
| "grad_norm": 0.47265625, |
| "learning_rate": 0.000269704010501043, |
| "loss": 0.1093, |
| "step": 60550 |
| }, |
| { |
| "epoch": 1.2289596430744272, |
| "grad_norm": 0.5, |
| "learning_rate": 0.00026964336183691977, |
| "loss": 0.1239, |
| "step": 60600 |
| }, |
| { |
| "epoch": 1.229973636179274, |
| "grad_norm": 0.58203125, |
| "learning_rate": 0.0002695826593615596, |
| "loss": 0.1243, |
| "step": 60650 |
| }, |
| { |
| "epoch": 1.2309876292841209, |
| "grad_norm": 0.515625, |
| "learning_rate": 0.00026952190310226414, |
| "loss": 0.1211, |
| "step": 60700 |
| }, |
| { |
| "epoch": 1.2320016223889678, |
| "grad_norm": 0.50390625, |
| "learning_rate": 0.0002694610930863595, |
| "loss": 0.1103, |
| "step": 60750 |
| }, |
| { |
| "epoch": 1.2330156154938146, |
| "grad_norm": 0.470703125, |
| "learning_rate": 0.0002694002293411957, |
| "loss": 0.1077, |
| "step": 60800 |
| }, |
| { |
| "epoch": 1.2340296085986615, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.0002693393118941471, |
| "loss": 0.1185, |
| "step": 60850 |
| }, |
| { |
| "epoch": 1.2350436017035085, |
| "grad_norm": 0.46484375, |
| "learning_rate": 0.0002692783407726122, |
| "loss": 0.1284, |
| "step": 60900 |
| }, |
| { |
| "epoch": 1.2360575948083552, |
| "grad_norm": 0.453125, |
| "learning_rate": 0.00026921731600401357, |
| "loss": 0.1129, |
| "step": 60950 |
| }, |
| { |
| "epoch": 1.2370715879132022, |
| "grad_norm": 0.466796875, |
| "learning_rate": 0.00026915623761579786, |
| "loss": 0.1149, |
| "step": 61000 |
| }, |
| { |
| "epoch": 1.238085581018049, |
| "grad_norm": 1.1875, |
| "learning_rate": 0.00026909510563543596, |
| "loss": 0.1254, |
| "step": 61050 |
| }, |
| { |
| "epoch": 1.2390995741228958, |
| "grad_norm": 0.392578125, |
| "learning_rate": 0.0002690339200904228, |
| "loss": 0.1163, |
| "step": 61100 |
| }, |
| { |
| "epoch": 1.2401135672277428, |
| "grad_norm": 0.578125, |
| "learning_rate": 0.00026897268100827747, |
| "loss": 0.1157, |
| "step": 61150 |
| }, |
| { |
| "epoch": 1.2411275603325898, |
| "grad_norm": 0.80078125, |
| "learning_rate": 0.00026891138841654295, |
| "loss": 0.1151, |
| "step": 61200 |
| }, |
| { |
| "epoch": 1.2421415534374367, |
| "grad_norm": 0.54296875, |
| "learning_rate": 0.00026885004234278656, |
| "loss": 0.1227, |
| "step": 61250 |
| }, |
| { |
| "epoch": 1.2431555465422834, |
| "grad_norm": 0.443359375, |
| "learning_rate": 0.0002687886428145995, |
| "loss": 0.1091, |
| "step": 61300 |
| }, |
| { |
| "epoch": 1.2441695396471304, |
| "grad_norm": 0.490234375, |
| "learning_rate": 0.00026872718985959703, |
| "loss": 0.1171, |
| "step": 61350 |
| }, |
| { |
| "epoch": 1.2451835327519774, |
| "grad_norm": 0.56640625, |
| "learning_rate": 0.00026866568350541846, |
| "loss": 0.1073, |
| "step": 61400 |
| }, |
| { |
| "epoch": 1.246197525856824, |
| "grad_norm": 0.5390625, |
| "learning_rate": 0.0002686041237797271, |
| "loss": 0.1129, |
| "step": 61450 |
| }, |
| { |
| "epoch": 1.247211518961671, |
| "grad_norm": 0.498046875, |
| "learning_rate": 0.0002685425107102103, |
| "loss": 0.1057, |
| "step": 61500 |
| }, |
| { |
| "epoch": 1.247211518961671, |
| "eval_cer": 0.03003740369459025, |
| "eval_loss": 0.17543725669384003, |
| "eval_runtime": 129.7983, |
| "eval_samples_per_second": 21.634, |
| "eval_steps_per_second": 0.169, |
| "eval_wer": 0.09222019094048345, |
| "step": 61500 |
| }, |
| { |
| "epoch": 1.248225512066518, |
| "grad_norm": 0.44921875, |
| "learning_rate": 0.00026848084432457936, |
| "loss": 0.1111, |
| "step": 61550 |
| }, |
| { |
| "epoch": 1.249239505171365, |
| "grad_norm": 0.447265625, |
| "learning_rate": 0.00026841912465056964, |
| "loss": 0.1184, |
| "step": 61600 |
| }, |
| { |
| "epoch": 1.2502534982762117, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.00026835735171594027, |
| "loss": 0.1064, |
| "step": 61650 |
| }, |
| { |
| "epoch": 1.2512674913810586, |
| "grad_norm": 0.486328125, |
| "learning_rate": 0.00026829552554847463, |
| "loss": 0.1176, |
| "step": 61700 |
| }, |
| { |
| "epoch": 1.2522814844859056, |
| "grad_norm": 0.52734375, |
| "learning_rate": 0.00026823364617597974, |
| "loss": 0.1141, |
| "step": 61750 |
| }, |
| { |
| "epoch": 1.2532954775907523, |
| "grad_norm": 1.4375, |
| "learning_rate": 0.0002681717136262868, |
| "loss": 0.1298, |
| "step": 61800 |
| }, |
| { |
| "epoch": 1.2543094706955993, |
| "grad_norm": 0.46875, |
| "learning_rate": 0.0002681097279272508, |
| "loss": 0.1078, |
| "step": 61850 |
| }, |
| { |
| "epoch": 1.2553234638004462, |
| "grad_norm": 0.84765625, |
| "learning_rate": 0.00026804768910675066, |
| "loss": 0.1121, |
| "step": 61900 |
| }, |
| { |
| "epoch": 1.256337456905293, |
| "grad_norm": 0.482421875, |
| "learning_rate": 0.0002679855971926891, |
| "loss": 0.1138, |
| "step": 61950 |
| }, |
| { |
| "epoch": 1.25735145001014, |
| "grad_norm": 0.51953125, |
| "learning_rate": 0.00026792345221299295, |
| "loss": 0.1086, |
| "step": 62000 |
| }, |
| { |
| "epoch": 1.2583654431149869, |
| "grad_norm": 0.494140625, |
| "learning_rate": 0.0002678612541956127, |
| "loss": 0.1085, |
| "step": 62050 |
| }, |
| { |
| "epoch": 1.2593794362198336, |
| "grad_norm": 0.51171875, |
| "learning_rate": 0.00026779900316852267, |
| "loss": 0.1159, |
| "step": 62100 |
| }, |
| { |
| "epoch": 1.2603934293246806, |
| "grad_norm": 0.5390625, |
| "learning_rate": 0.00026773669915972125, |
| "loss": 0.1088, |
| "step": 62150 |
| }, |
| { |
| "epoch": 1.2614074224295275, |
| "grad_norm": 0.4375, |
| "learning_rate": 0.0002676743421972304, |
| "loss": 0.1102, |
| "step": 62200 |
| }, |
| { |
| "epoch": 1.2624214155343743, |
| "grad_norm": 0.482421875, |
| "learning_rate": 0.0002676119323090961, |
| "loss": 0.114, |
| "step": 62250 |
| }, |
| { |
| "epoch": 1.2634354086392212, |
| "grad_norm": 0.58203125, |
| "learning_rate": 0.000267549469523388, |
| "loss": 0.1156, |
| "step": 62300 |
| }, |
| { |
| "epoch": 1.2644494017440682, |
| "grad_norm": 0.515625, |
| "learning_rate": 0.00026748695386819964, |
| "loss": 0.1079, |
| "step": 62350 |
| }, |
| { |
| "epoch": 1.265463394848915, |
| "grad_norm": 0.44921875, |
| "learning_rate": 0.00026742438537164827, |
| "loss": 0.1083, |
| "step": 62400 |
| }, |
| { |
| "epoch": 1.2664773879537619, |
| "grad_norm": 0.73046875, |
| "learning_rate": 0.0002673617640618749, |
| "loss": 0.1233, |
| "step": 62450 |
| }, |
| { |
| "epoch": 1.2674913810586088, |
| "grad_norm": 0.56640625, |
| "learning_rate": 0.00026729908996704437, |
| "loss": 0.1105, |
| "step": 62500 |
| }, |
| { |
| "epoch": 1.2685053741634558, |
| "grad_norm": 0.51171875, |
| "learning_rate": 0.00026723636311534515, |
| "loss": 0.1129, |
| "step": 62550 |
| }, |
| { |
| "epoch": 1.2695193672683025, |
| "grad_norm": 0.458984375, |
| "learning_rate": 0.00026717358353498953, |
| "loss": 0.1131, |
| "step": 62600 |
| }, |
| { |
| "epoch": 1.2705333603731495, |
| "grad_norm": 0.55859375, |
| "learning_rate": 0.0002671107512542135, |
| "loss": 0.1277, |
| "step": 62650 |
| }, |
| { |
| "epoch": 1.2715473534779964, |
| "grad_norm": 0.474609375, |
| "learning_rate": 0.0002670478663012767, |
| "loss": 0.1089, |
| "step": 62700 |
| }, |
| { |
| "epoch": 1.2725613465828434, |
| "grad_norm": 0.4453125, |
| "learning_rate": 0.00026698492870446257, |
| "loss": 0.1256, |
| "step": 62750 |
| }, |
| { |
| "epoch": 1.27357533968769, |
| "grad_norm": 0.48828125, |
| "learning_rate": 0.0002669219384920781, |
| "loss": 0.1067, |
| "step": 62800 |
| }, |
| { |
| "epoch": 1.274589332792537, |
| "grad_norm": 0.5546875, |
| "learning_rate": 0.00026685889569245393, |
| "loss": 0.1146, |
| "step": 62850 |
| }, |
| { |
| "epoch": 1.275603325897384, |
| "grad_norm": 0.515625, |
| "learning_rate": 0.0002667958003339445, |
| "loss": 0.1066, |
| "step": 62900 |
| }, |
| { |
| "epoch": 1.2766173190022307, |
| "grad_norm": 0.48828125, |
| "learning_rate": 0.0002667326524449279, |
| "loss": 0.1214, |
| "step": 62950 |
| }, |
| { |
| "epoch": 1.2776313121070777, |
| "grad_norm": 0.466796875, |
| "learning_rate": 0.00026666945205380555, |
| "loss": 0.1071, |
| "step": 63000 |
| }, |
| { |
| "epoch": 1.2776313121070777, |
| "eval_cer": 0.03263329292971331, |
| "eval_loss": 0.1732511967420578, |
| "eval_runtime": 127.7415, |
| "eval_samples_per_second": 21.982, |
| "eval_steps_per_second": 0.172, |
| "eval_wer": 0.09442074615749205, |
| "step": 63000 |
| }, |
| { |
| "epoch": 1.2786453052119247, |
| "grad_norm": 0.474609375, |
| "learning_rate": 0.00026660619918900287, |
| "loss": 0.1134, |
| "step": 63050 |
| }, |
| { |
| "epoch": 1.2796592983167714, |
| "grad_norm": 0.458984375, |
| "learning_rate": 0.00026654289387896855, |
| "loss": 0.1119, |
| "step": 63100 |
| }, |
| { |
| "epoch": 1.2806732914216183, |
| "grad_norm": 0.51171875, |
| "learning_rate": 0.00026647953615217514, |
| "loss": 0.1172, |
| "step": 63150 |
| }, |
| { |
| "epoch": 1.2816872845264653, |
| "grad_norm": 0.4375, |
| "learning_rate": 0.00026641612603711857, |
| "loss": 0.1076, |
| "step": 63200 |
| }, |
| { |
| "epoch": 1.282701277631312, |
| "grad_norm": 0.458984375, |
| "learning_rate": 0.0002663526635623185, |
| "loss": 0.112, |
| "step": 63250 |
| }, |
| { |
| "epoch": 1.283715270736159, |
| "grad_norm": 0.6875, |
| "learning_rate": 0.00026628914875631793, |
| "loss": 0.1193, |
| "step": 63300 |
| }, |
| { |
| "epoch": 1.284729263841006, |
| "grad_norm": 0.5625, |
| "learning_rate": 0.0002662255816476836, |
| "loss": 0.1232, |
| "step": 63350 |
| }, |
| { |
| "epoch": 1.2857432569458527, |
| "grad_norm": 0.44140625, |
| "learning_rate": 0.0002661619622650056, |
| "loss": 0.1134, |
| "step": 63400 |
| }, |
| { |
| "epoch": 1.2867572500506996, |
| "grad_norm": 0.515625, |
| "learning_rate": 0.00026609829063689767, |
| "loss": 0.1161, |
| "step": 63450 |
| }, |
| { |
| "epoch": 1.2877712431555466, |
| "grad_norm": 0.44140625, |
| "learning_rate": 0.000266034566791997, |
| "loss": 0.1152, |
| "step": 63500 |
| }, |
| { |
| "epoch": 1.2887852362603933, |
| "grad_norm": 0.455078125, |
| "learning_rate": 0.0002659707907589643, |
| "loss": 0.1139, |
| "step": 63550 |
| }, |
| { |
| "epoch": 1.2897992293652403, |
| "grad_norm": 0.47265625, |
| "learning_rate": 0.0002659069625664836, |
| "loss": 0.113, |
| "step": 63600 |
| }, |
| { |
| "epoch": 1.2908132224700872, |
| "grad_norm": 0.50390625, |
| "learning_rate": 0.0002658430822432626, |
| "loss": 0.1143, |
| "step": 63650 |
| }, |
| { |
| "epoch": 1.291827215574934, |
| "grad_norm": 0.482421875, |
| "learning_rate": 0.0002657791498180323, |
| "loss": 0.1104, |
| "step": 63700 |
| }, |
| { |
| "epoch": 1.292841208679781, |
| "grad_norm": 0.4765625, |
| "learning_rate": 0.0002657151653195472, |
| "loss": 0.1082, |
| "step": 63750 |
| }, |
| { |
| "epoch": 1.2938552017846279, |
| "grad_norm": 0.439453125, |
| "learning_rate": 0.00026565112877658515, |
| "loss": 0.1084, |
| "step": 63800 |
| }, |
| { |
| "epoch": 1.2948691948894748, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.00026558704021794753, |
| "loss": 0.112, |
| "step": 63850 |
| }, |
| { |
| "epoch": 1.2958831879943216, |
| "grad_norm": 0.482421875, |
| "learning_rate": 0.00026552289967245906, |
| "loss": 0.1244, |
| "step": 63900 |
| }, |
| { |
| "epoch": 1.2968971810991685, |
| "grad_norm": 0.53125, |
| "learning_rate": 0.00026545870716896774, |
| "loss": 0.1187, |
| "step": 63950 |
| }, |
| { |
| "epoch": 1.2979111742040155, |
| "grad_norm": 0.53125, |
| "learning_rate": 0.0002653944627363451, |
| "loss": 0.1134, |
| "step": 64000 |
| }, |
| { |
| "epoch": 1.2989251673088624, |
| "grad_norm": 0.7109375, |
| "learning_rate": 0.0002653301664034859, |
| "loss": 0.1151, |
| "step": 64050 |
| }, |
| { |
| "epoch": 1.2999391604137092, |
| "grad_norm": 0.515625, |
| "learning_rate": 0.00026526581819930837, |
| "loss": 0.1206, |
| "step": 64100 |
| }, |
| { |
| "epoch": 1.3009531535185561, |
| "grad_norm": 0.423828125, |
| "learning_rate": 0.0002652014181527539, |
| "loss": 0.1081, |
| "step": 64150 |
| }, |
| { |
| "epoch": 1.301967146623403, |
| "grad_norm": 0.455078125, |
| "learning_rate": 0.00026513696629278734, |
| "loss": 0.105, |
| "step": 64200 |
| }, |
| { |
| "epoch": 1.3029811397282498, |
| "grad_norm": 0.66015625, |
| "learning_rate": 0.0002650724626483968, |
| "loss": 0.1237, |
| "step": 64250 |
| }, |
| { |
| "epoch": 1.3039951328330968, |
| "grad_norm": 0.5078125, |
| "learning_rate": 0.0002650079072485937, |
| "loss": 0.1159, |
| "step": 64300 |
| }, |
| { |
| "epoch": 1.3050091259379437, |
| "grad_norm": 0.482421875, |
| "learning_rate": 0.0002649433001224126, |
| "loss": 0.1145, |
| "step": 64350 |
| }, |
| { |
| "epoch": 1.3060231190427904, |
| "grad_norm": 0.6015625, |
| "learning_rate": 0.00026487864129891156, |
| "loss": 0.1206, |
| "step": 64400 |
| }, |
| { |
| "epoch": 1.3070371121476374, |
| "grad_norm": 0.466796875, |
| "learning_rate": 0.00026481393080717176, |
| "loss": 0.1213, |
| "step": 64450 |
| }, |
| { |
| "epoch": 1.3080511052524844, |
| "grad_norm": 0.609375, |
| "learning_rate": 0.00026474916867629753, |
| "loss": 0.1216, |
| "step": 64500 |
| }, |
| { |
| "epoch": 1.3080511052524844, |
| "eval_cer": 0.030984200970727318, |
| "eval_loss": 0.1738138198852539, |
| "eval_runtime": 127.5722, |
| "eval_samples_per_second": 22.011, |
| "eval_steps_per_second": 0.172, |
| "eval_wer": 0.09448845554878461, |
| "step": 64500 |
| }, |
| { |
| "epoch": 1.309065098357331, |
| "grad_norm": 0.458984375, |
| "learning_rate": 0.0002646843549354166, |
| "loss": 0.1128, |
| "step": 64550 |
| }, |
| { |
| "epoch": 1.310079091462178, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.00026461948961367985, |
| "loss": 0.1133, |
| "step": 64600 |
| }, |
| { |
| "epoch": 1.311093084567025, |
| "grad_norm": 0.4921875, |
| "learning_rate": 0.0002645545727402613, |
| "loss": 0.1129, |
| "step": 64650 |
| }, |
| { |
| "epoch": 1.3121070776718717, |
| "grad_norm": 0.474609375, |
| "learning_rate": 0.0002644896043443582, |
| "loss": 0.1043, |
| "step": 64700 |
| }, |
| { |
| "epoch": 1.3131210707767187, |
| "grad_norm": 0.4921875, |
| "learning_rate": 0.0002644245844551911, |
| "loss": 0.1278, |
| "step": 64750 |
| }, |
| { |
| "epoch": 1.3141350638815656, |
| "grad_norm": 0.55078125, |
| "learning_rate": 0.00026435951310200336, |
| "loss": 0.1126, |
| "step": 64800 |
| }, |
| { |
| "epoch": 1.3151490569864124, |
| "grad_norm": 0.4453125, |
| "learning_rate": 0.0002642943903140619, |
| "loss": 0.1189, |
| "step": 64850 |
| }, |
| { |
| "epoch": 1.3161630500912593, |
| "grad_norm": 0.46875, |
| "learning_rate": 0.00026422921612065647, |
| "loss": 0.1065, |
| "step": 64900 |
| }, |
| { |
| "epoch": 1.3171770431961063, |
| "grad_norm": 0.5546875, |
| "learning_rate": 0.0002641639905511001, |
| "loss": 0.1217, |
| "step": 64950 |
| }, |
| { |
| "epoch": 1.318191036300953, |
| "grad_norm": 0.431640625, |
| "learning_rate": 0.0002640987136347289, |
| "loss": 0.1134, |
| "step": 65000 |
| }, |
| { |
| "epoch": 1.3192050294058, |
| "grad_norm": 0.5078125, |
| "learning_rate": 0.00026403338540090195, |
| "loss": 0.1124, |
| "step": 65050 |
| }, |
| { |
| "epoch": 1.320219022510647, |
| "grad_norm": 0.490234375, |
| "learning_rate": 0.00026396800587900166, |
| "loss": 0.1141, |
| "step": 65100 |
| }, |
| { |
| "epoch": 1.3212330156154939, |
| "grad_norm": 0.478515625, |
| "learning_rate": 0.00026390257509843333, |
| "loss": 0.1168, |
| "step": 65150 |
| }, |
| { |
| "epoch": 1.3222470087203406, |
| "grad_norm": 0.51953125, |
| "learning_rate": 0.0002638370930886253, |
| "loss": 0.1317, |
| "step": 65200 |
| }, |
| { |
| "epoch": 1.3232610018251876, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.000263771559879029, |
| "loss": 0.1137, |
| "step": 65250 |
| }, |
| { |
| "epoch": 1.3242749949300345, |
| "grad_norm": 0.6796875, |
| "learning_rate": 0.0002637059754991189, |
| "loss": 0.1166, |
| "step": 65300 |
| }, |
| { |
| "epoch": 1.3252889880348815, |
| "grad_norm": 0.470703125, |
| "learning_rate": 0.0002636403399783925, |
| "loss": 0.105, |
| "step": 65350 |
| }, |
| { |
| "epoch": 1.3263029811397282, |
| "grad_norm": 0.494140625, |
| "learning_rate": 0.0002635746533463702, |
| "loss": 0.1048, |
| "step": 65400 |
| }, |
| { |
| "epoch": 1.3273169742445752, |
| "grad_norm": 0.470703125, |
| "learning_rate": 0.00026350891563259554, |
| "loss": 0.1124, |
| "step": 65450 |
| }, |
| { |
| "epoch": 1.3283309673494221, |
| "grad_norm": 0.478515625, |
| "learning_rate": 0.0002634431268666349, |
| "loss": 0.1062, |
| "step": 65500 |
| }, |
| { |
| "epoch": 1.3293449604542689, |
| "grad_norm": 0.4453125, |
| "learning_rate": 0.0002633772870780777, |
| "loss": 0.1224, |
| "step": 65550 |
| }, |
| { |
| "epoch": 1.3303589535591158, |
| "grad_norm": 0.466796875, |
| "learning_rate": 0.00026331139629653635, |
| "loss": 0.1156, |
| "step": 65600 |
| }, |
| { |
| "epoch": 1.3313729466639628, |
| "grad_norm": 0.4921875, |
| "learning_rate": 0.000263245454551646, |
| "loss": 0.1185, |
| "step": 65650 |
| }, |
| { |
| "epoch": 1.3323869397688095, |
| "grad_norm": 0.40234375, |
| "learning_rate": 0.0002631794618730649, |
| "loss": 0.112, |
| "step": 65700 |
| }, |
| { |
| "epoch": 1.3334009328736565, |
| "grad_norm": 0.4140625, |
| "learning_rate": 0.0002631134182904742, |
| "loss": 0.1193, |
| "step": 65750 |
| }, |
| { |
| "epoch": 1.3344149259785034, |
| "grad_norm": 0.51171875, |
| "learning_rate": 0.0002630473238335779, |
| "loss": 0.1049, |
| "step": 65800 |
| }, |
| { |
| "epoch": 1.3354289190833502, |
| "grad_norm": 0.48828125, |
| "learning_rate": 0.0002629811785321028, |
| "loss": 0.1129, |
| "step": 65850 |
| }, |
| { |
| "epoch": 1.336442912188197, |
| "grad_norm": 0.48046875, |
| "learning_rate": 0.00026291498241579883, |
| "loss": 0.1185, |
| "step": 65900 |
| }, |
| { |
| "epoch": 1.337456905293044, |
| "grad_norm": 0.458984375, |
| "learning_rate": 0.00026284873551443835, |
| "loss": 0.1077, |
| "step": 65950 |
| }, |
| { |
| "epoch": 1.3384708983978908, |
| "grad_norm": 0.4375, |
| "learning_rate": 0.000262782437857817, |
| "loss": 0.1071, |
| "step": 66000 |
| }, |
| { |
| "epoch": 1.3384708983978908, |
| "eval_cer": 0.02826345934754223, |
| "eval_loss": 0.17099952697753906, |
| "eval_runtime": 126.6833, |
| "eval_samples_per_second": 22.166, |
| "eval_steps_per_second": 0.174, |
| "eval_wer": 0.0911368406798023, |
| "step": 66000 |
| }, |
| { |
| "epoch": 1.3394848915027378, |
| "grad_norm": 0.46484375, |
| "learning_rate": 0.000262716089475753, |
| "loss": 0.116, |
| "step": 66050 |
| }, |
| { |
| "epoch": 1.3404988846075847, |
| "grad_norm": 0.5078125, |
| "learning_rate": 0.00026264969039808744, |
| "loss": 0.1075, |
| "step": 66100 |
| }, |
| { |
| "epoch": 1.3415128777124314, |
| "grad_norm": 0.478515625, |
| "learning_rate": 0.0002625832406546842, |
| "loss": 0.1159, |
| "step": 66150 |
| }, |
| { |
| "epoch": 1.3425268708172784, |
| "grad_norm": 0.486328125, |
| "learning_rate": 0.00026251674027543, |
| "loss": 0.1205, |
| "step": 66200 |
| }, |
| { |
| "epoch": 1.3435408639221254, |
| "grad_norm": 0.427734375, |
| "learning_rate": 0.00026245018929023427, |
| "loss": 0.1104, |
| "step": 66250 |
| }, |
| { |
| "epoch": 1.344554857026972, |
| "grad_norm": 0.52734375, |
| "learning_rate": 0.00026238358772902917, |
| "loss": 0.1065, |
| "step": 66300 |
| }, |
| { |
| "epoch": 1.345568850131819, |
| "grad_norm": 0.458984375, |
| "learning_rate": 0.00026231693562176974, |
| "loss": 0.1178, |
| "step": 66350 |
| }, |
| { |
| "epoch": 1.346582843236666, |
| "grad_norm": 0.5, |
| "learning_rate": 0.0002622502329984337, |
| "loss": 0.1141, |
| "step": 66400 |
| }, |
| { |
| "epoch": 1.347596836341513, |
| "grad_norm": 0.435546875, |
| "learning_rate": 0.0002621834798890214, |
| "loss": 0.1033, |
| "step": 66450 |
| }, |
| { |
| "epoch": 1.3486108294463597, |
| "grad_norm": 0.46875, |
| "learning_rate": 0.0002621166763235559, |
| "loss": 0.1063, |
| "step": 66500 |
| }, |
| { |
| "epoch": 1.3496248225512066, |
| "grad_norm": 0.53515625, |
| "learning_rate": 0.0002620498223320832, |
| "loss": 0.1157, |
| "step": 66550 |
| }, |
| { |
| "epoch": 1.3506388156560536, |
| "grad_norm": 0.462890625, |
| "learning_rate": 0.0002619829179446717, |
| "loss": 0.111, |
| "step": 66600 |
| }, |
| { |
| "epoch": 1.3516528087609005, |
| "grad_norm": 0.44140625, |
| "learning_rate": 0.00026191596319141247, |
| "loss": 0.109, |
| "step": 66650 |
| }, |
| { |
| "epoch": 1.3526668018657473, |
| "grad_norm": 0.482421875, |
| "learning_rate": 0.0002618489581024194, |
| "loss": 0.1123, |
| "step": 66700 |
| }, |
| { |
| "epoch": 1.3536807949705942, |
| "grad_norm": 0.435546875, |
| "learning_rate": 0.0002617819027078291, |
| "loss": 0.107, |
| "step": 66750 |
| }, |
| { |
| "epoch": 1.3546947880754412, |
| "grad_norm": 0.51953125, |
| "learning_rate": 0.00026171479703780037, |
| "loss": 0.1096, |
| "step": 66800 |
| }, |
| { |
| "epoch": 1.355708781180288, |
| "grad_norm": 0.578125, |
| "learning_rate": 0.00026164764112251505, |
| "loss": 0.1121, |
| "step": 66850 |
| }, |
| { |
| "epoch": 1.3567227742851349, |
| "grad_norm": 0.69921875, |
| "learning_rate": 0.0002615804349921775, |
| "loss": 0.1072, |
| "step": 66900 |
| }, |
| { |
| "epoch": 1.3577367673899818, |
| "grad_norm": 0.44140625, |
| "learning_rate": 0.0002615131786770145, |
| "loss": 0.1115, |
| "step": 66950 |
| }, |
| { |
| "epoch": 1.3587507604948286, |
| "grad_norm": 0.474609375, |
| "learning_rate": 0.00026144587220727544, |
| "loss": 0.1058, |
| "step": 67000 |
| }, |
| { |
| "epoch": 1.3597647535996755, |
| "grad_norm": 0.484375, |
| "learning_rate": 0.0002613785156132324, |
| "loss": 0.1184, |
| "step": 67050 |
| }, |
| { |
| "epoch": 1.3607787467045225, |
| "grad_norm": 0.3984375, |
| "learning_rate": 0.00026131110892518, |
| "loss": 0.1115, |
| "step": 67100 |
| }, |
| { |
| "epoch": 1.3617927398093692, |
| "grad_norm": 0.5078125, |
| "learning_rate": 0.0002612436521734352, |
| "loss": 0.1063, |
| "step": 67150 |
| }, |
| { |
| "epoch": 1.3628067329142162, |
| "grad_norm": 0.51953125, |
| "learning_rate": 0.0002611761453883376, |
| "loss": 0.1118, |
| "step": 67200 |
| }, |
| { |
| "epoch": 1.3638207260190631, |
| "grad_norm": 0.412109375, |
| "learning_rate": 0.0002611085886002494, |
| "loss": 0.1089, |
| "step": 67250 |
| }, |
| { |
| "epoch": 1.3648347191239099, |
| "grad_norm": 0.50390625, |
| "learning_rate": 0.00026104098183955506, |
| "loss": 0.1059, |
| "step": 67300 |
| }, |
| { |
| "epoch": 1.3658487122287568, |
| "grad_norm": 0.42578125, |
| "learning_rate": 0.00026097332513666174, |
| "loss": 0.1196, |
| "step": 67350 |
| }, |
| { |
| "epoch": 1.3668627053336038, |
| "grad_norm": 0.494140625, |
| "learning_rate": 0.000260905618521999, |
| "loss": 0.1123, |
| "step": 67400 |
| }, |
| { |
| "epoch": 1.3678766984384505, |
| "grad_norm": 0.462890625, |
| "learning_rate": 0.0002608378620260187, |
| "loss": 0.1106, |
| "step": 67450 |
| }, |
| { |
| "epoch": 1.3688906915432975, |
| "grad_norm": 0.56640625, |
| "learning_rate": 0.00026077005567919526, |
| "loss": 0.109, |
| "step": 67500 |
| }, |
| { |
| "epoch": 1.3688906915432975, |
| "eval_cer": 0.03368413386256874, |
| "eval_loss": 0.17319071292877197, |
| "eval_runtime": 129.0671, |
| "eval_samples_per_second": 21.756, |
| "eval_steps_per_second": 0.17, |
| "eval_wer": 0.09628275441803778, |
| "step": 67500 |
| }, |
| { |
| "epoch": 1.3699046846481444, |
| "grad_norm": 0.431640625, |
| "learning_rate": 0.00026070219951202563, |
| "loss": 0.1146, |
| "step": 67550 |
| }, |
| { |
| "epoch": 1.3709186777529911, |
| "grad_norm": 0.451171875, |
| "learning_rate": 0.00026063429355502896, |
| "loss": 0.1129, |
| "step": 67600 |
| }, |
| { |
| "epoch": 1.371932670857838, |
| "grad_norm": 0.49609375, |
| "learning_rate": 0.0002605663378387469, |
| "loss": 0.1059, |
| "step": 67650 |
| }, |
| { |
| "epoch": 1.372946663962685, |
| "grad_norm": 0.482421875, |
| "learning_rate": 0.00026049833239374346, |
| "loss": 0.1052, |
| "step": 67700 |
| }, |
| { |
| "epoch": 1.373960657067532, |
| "grad_norm": 0.4921875, |
| "learning_rate": 0.00026043027725060493, |
| "loss": 0.1215, |
| "step": 67750 |
| }, |
| { |
| "epoch": 1.374974650172379, |
| "grad_norm": 0.451171875, |
| "learning_rate": 0.00026036217243994024, |
| "loss": 0.123, |
| "step": 67800 |
| }, |
| { |
| "epoch": 1.3759886432772257, |
| "grad_norm": 0.50390625, |
| "learning_rate": 0.00026029401799238026, |
| "loss": 0.12, |
| "step": 67850 |
| }, |
| { |
| "epoch": 1.3770026363820727, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.00026022581393857853, |
| "loss": 0.1261, |
| "step": 67900 |
| }, |
| { |
| "epoch": 1.3780166294869196, |
| "grad_norm": 0.50390625, |
| "learning_rate": 0.00026015756030921066, |
| "loss": 0.1167, |
| "step": 67950 |
| }, |
| { |
| "epoch": 1.3790306225917663, |
| "grad_norm": 0.466796875, |
| "learning_rate": 0.0002600892571349747, |
| "loss": 0.115, |
| "step": 68000 |
| }, |
| { |
| "epoch": 1.3800446156966133, |
| "grad_norm": 0.48828125, |
| "learning_rate": 0.0002600209044465909, |
| "loss": 0.1067, |
| "step": 68050 |
| }, |
| { |
| "epoch": 1.3810586088014603, |
| "grad_norm": 0.5859375, |
| "learning_rate": 0.00025995250227480186, |
| "loss": 0.1135, |
| "step": 68100 |
| }, |
| { |
| "epoch": 1.382072601906307, |
| "grad_norm": 0.51953125, |
| "learning_rate": 0.0002598840506503724, |
| "loss": 0.1133, |
| "step": 68150 |
| }, |
| { |
| "epoch": 1.383086595011154, |
| "grad_norm": 0.52734375, |
| "learning_rate": 0.0002598155496040895, |
| "loss": 0.1121, |
| "step": 68200 |
| }, |
| { |
| "epoch": 1.384100588116001, |
| "grad_norm": 0.478515625, |
| "learning_rate": 0.00025974699916676245, |
| "loss": 0.1075, |
| "step": 68250 |
| }, |
| { |
| "epoch": 1.3851145812208476, |
| "grad_norm": 0.486328125, |
| "learning_rate": 0.0002596783993692228, |
| "loss": 0.1109, |
| "step": 68300 |
| }, |
| { |
| "epoch": 1.3861285743256946, |
| "grad_norm": 0.435546875, |
| "learning_rate": 0.00025960975024232427, |
| "loss": 0.1139, |
| "step": 68350 |
| }, |
| { |
| "epoch": 1.3871425674305415, |
| "grad_norm": 0.494140625, |
| "learning_rate": 0.00025954105181694273, |
| "loss": 0.1045, |
| "step": 68400 |
| }, |
| { |
| "epoch": 1.3881565605353883, |
| "grad_norm": 0.57421875, |
| "learning_rate": 0.0002594723041239762, |
| "loss": 0.1155, |
| "step": 68450 |
| }, |
| { |
| "epoch": 1.3891705536402352, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.00025940350719434486, |
| "loss": 0.1115, |
| "step": 68500 |
| }, |
| { |
| "epoch": 1.3901845467450822, |
| "grad_norm": 0.51171875, |
| "learning_rate": 0.0002593346610589911, |
| "loss": 0.112, |
| "step": 68550 |
| }, |
| { |
| "epoch": 1.391198539849929, |
| "grad_norm": 0.48828125, |
| "learning_rate": 0.0002592657657488794, |
| "loss": 0.124, |
| "step": 68600 |
| }, |
| { |
| "epoch": 1.3922125329547759, |
| "grad_norm": 0.48046875, |
| "learning_rate": 0.0002591968212949964, |
| "loss": 0.1315, |
| "step": 68650 |
| }, |
| { |
| "epoch": 1.3932265260596228, |
| "grad_norm": 0.51171875, |
| "learning_rate": 0.00025912782772835087, |
| "loss": 0.1083, |
| "step": 68700 |
| }, |
| { |
| "epoch": 1.3942405191644696, |
| "grad_norm": 0.43359375, |
| "learning_rate": 0.0002590587850799734, |
| "loss": 0.1172, |
| "step": 68750 |
| }, |
| { |
| "epoch": 1.3952545122693165, |
| "grad_norm": 0.470703125, |
| "learning_rate": 0.0002589896933809172, |
| "loss": 0.1109, |
| "step": 68800 |
| }, |
| { |
| "epoch": 1.3962685053741635, |
| "grad_norm": 0.478515625, |
| "learning_rate": 0.0002589205526622569, |
| "loss": 0.1145, |
| "step": 68850 |
| }, |
| { |
| "epoch": 1.3972824984790102, |
| "grad_norm": 0.45703125, |
| "learning_rate": 0.0002588513629550895, |
| "loss": 0.1126, |
| "step": 68900 |
| }, |
| { |
| "epoch": 1.3982964915838572, |
| "grad_norm": 0.515625, |
| "learning_rate": 0.00025878212429053423, |
| "loss": 0.1229, |
| "step": 68950 |
| }, |
| { |
| "epoch": 1.3993104846887041, |
| "grad_norm": 0.5078125, |
| "learning_rate": 0.000258712836699732, |
| "loss": 0.1069, |
| "step": 69000 |
| }, |
| { |
| "epoch": 1.3993104846887041, |
| "eval_cer": 0.031514823619990945, |
| "eval_loss": 0.17294800281524658, |
| "eval_runtime": 126.3316, |
| "eval_samples_per_second": 22.227, |
| "eval_steps_per_second": 0.174, |
| "eval_wer": 0.09462387433136976, |
| "step": 69000 |
| }, |
| { |
| "epoch": 1.400324477793551, |
| "grad_norm": 0.54296875, |
| "learning_rate": 0.0002586435002138458, |
| "loss": 0.111, |
| "step": 69050 |
| }, |
| { |
| "epoch": 1.401338470898398, |
| "grad_norm": 0.404296875, |
| "learning_rate": 0.0002585741148640607, |
| "loss": 0.1053, |
| "step": 69100 |
| }, |
| { |
| "epoch": 1.4023524640032448, |
| "grad_norm": 0.5390625, |
| "learning_rate": 0.0002585046806815838, |
| "loss": 0.1124, |
| "step": 69150 |
| }, |
| { |
| "epoch": 1.4033664571080917, |
| "grad_norm": 0.44921875, |
| "learning_rate": 0.0002584351976976439, |
| "loss": 0.1064, |
| "step": 69200 |
| }, |
| { |
| "epoch": 1.4043804502129387, |
| "grad_norm": 0.447265625, |
| "learning_rate": 0.00025836566594349207, |
| "loss": 0.111, |
| "step": 69250 |
| }, |
| { |
| "epoch": 1.4053944433177854, |
| "grad_norm": 0.484375, |
| "learning_rate": 0.00025829608545040113, |
| "loss": 0.1121, |
| "step": 69300 |
| }, |
| { |
| "epoch": 1.4064084364226324, |
| "grad_norm": 0.41015625, |
| "learning_rate": 0.00025822645624966583, |
| "loss": 0.1094, |
| "step": 69350 |
| }, |
| { |
| "epoch": 1.4074224295274793, |
| "grad_norm": 0.7265625, |
| "learning_rate": 0.00025815677837260286, |
| "loss": 0.1129, |
| "step": 69400 |
| }, |
| { |
| "epoch": 1.408436422632326, |
| "grad_norm": 0.5, |
| "learning_rate": 0.0002580870518505508, |
| "loss": 0.1138, |
| "step": 69450 |
| }, |
| { |
| "epoch": 1.409450415737173, |
| "grad_norm": 0.5390625, |
| "learning_rate": 0.0002580172767148701, |
| "loss": 0.1224, |
| "step": 69500 |
| }, |
| { |
| "epoch": 1.41046440884202, |
| "grad_norm": 0.5, |
| "learning_rate": 0.00025794745299694314, |
| "loss": 0.1087, |
| "step": 69550 |
| }, |
| { |
| "epoch": 1.4114784019468667, |
| "grad_norm": 0.451171875, |
| "learning_rate": 0.0002578775807281741, |
| "loss": 0.1081, |
| "step": 69600 |
| }, |
| { |
| "epoch": 1.4124923950517136, |
| "grad_norm": 0.62109375, |
| "learning_rate": 0.0002578076599399888, |
| "loss": 0.1153, |
| "step": 69650 |
| }, |
| { |
| "epoch": 1.4135063881565606, |
| "grad_norm": 0.5546875, |
| "learning_rate": 0.0002577376906638352, |
| "loss": 0.1224, |
| "step": 69700 |
| }, |
| { |
| "epoch": 1.4145203812614073, |
| "grad_norm": 0.50390625, |
| "learning_rate": 0.00025766767293118305, |
| "loss": 0.114, |
| "step": 69750 |
| }, |
| { |
| "epoch": 1.4155343743662543, |
| "grad_norm": 0.42578125, |
| "learning_rate": 0.0002575976067735236, |
| "loss": 0.1222, |
| "step": 69800 |
| }, |
| { |
| "epoch": 1.4165483674711012, |
| "grad_norm": 0.47265625, |
| "learning_rate": 0.00025752749222237014, |
| "loss": 0.1123, |
| "step": 69850 |
| }, |
| { |
| "epoch": 1.417562360575948, |
| "grad_norm": 0.4921875, |
| "learning_rate": 0.00025745732930925765, |
| "loss": 0.1068, |
| "step": 69900 |
| }, |
| { |
| "epoch": 1.418576353680795, |
| "grad_norm": 0.47265625, |
| "learning_rate": 0.0002573871180657428, |
| "loss": 0.1054, |
| "step": 69950 |
| }, |
| { |
| "epoch": 1.4195903467856419, |
| "grad_norm": 0.474609375, |
| "learning_rate": 0.00025731685852340417, |
| "loss": 0.1149, |
| "step": 70000 |
| }, |
| { |
| "epoch": 1.4206043398904886, |
| "grad_norm": 0.4609375, |
| "learning_rate": 0.00025724655071384186, |
| "loss": 0.1088, |
| "step": 70050 |
| }, |
| { |
| "epoch": 1.4216183329953356, |
| "grad_norm": 0.43359375, |
| "learning_rate": 0.0002571761946686778, |
| "loss": 0.1075, |
| "step": 70100 |
| }, |
| { |
| "epoch": 1.4226323261001825, |
| "grad_norm": 0.46484375, |
| "learning_rate": 0.0002571057904195556, |
| "loss": 0.1133, |
| "step": 70150 |
| }, |
| { |
| "epoch": 1.4236463192050295, |
| "grad_norm": 0.50390625, |
| "learning_rate": 0.00025703533799814046, |
| "loss": 0.1191, |
| "step": 70200 |
| }, |
| { |
| "epoch": 1.4246603123098762, |
| "grad_norm": 0.96875, |
| "learning_rate": 0.0002569648374361194, |
| "loss": 0.1081, |
| "step": 70250 |
| }, |
| { |
| "epoch": 1.4256743054147232, |
| "grad_norm": 0.458984375, |
| "learning_rate": 0.00025689428876520096, |
| "loss": 0.1192, |
| "step": 70300 |
| }, |
| { |
| "epoch": 1.4266882985195701, |
| "grad_norm": 0.486328125, |
| "learning_rate": 0.00025682369201711537, |
| "loss": 0.124, |
| "step": 70350 |
| }, |
| { |
| "epoch": 1.427702291624417, |
| "grad_norm": 0.44921875, |
| "learning_rate": 0.00025675304722361453, |
| "loss": 0.1125, |
| "step": 70400 |
| }, |
| { |
| "epoch": 1.4287162847292638, |
| "grad_norm": 0.56640625, |
| "learning_rate": 0.00025668235441647186, |
| "loss": 0.1115, |
| "step": 70450 |
| }, |
| { |
| "epoch": 1.4297302778341108, |
| "grad_norm": 0.484375, |
| "learning_rate": 0.00025661161362748244, |
| "loss": 0.1083, |
| "step": 70500 |
| }, |
| { |
| "epoch": 1.4297302778341108, |
| "eval_cer": 0.03154083453417054, |
| "eval_loss": 0.1698320358991623, |
| "eval_runtime": 128.2715, |
| "eval_samples_per_second": 21.891, |
| "eval_steps_per_second": 0.172, |
| "eval_wer": 0.09438689146184576, |
| "step": 70500 |
| }, |
| { |
| "epoch": 1.4307442709389577, |
| "grad_norm": 0.46484375, |
| "learning_rate": 0.0002565408248884629, |
| "loss": 0.1153, |
| "step": 70550 |
| }, |
| { |
| "epoch": 1.4317582640438045, |
| "grad_norm": 0.54296875, |
| "learning_rate": 0.0002564699882312514, |
| "loss": 0.1121, |
| "step": 70600 |
| }, |
| { |
| "epoch": 1.4327722571486514, |
| "grad_norm": 0.54296875, |
| "learning_rate": 0.00025639910368770776, |
| "loss": 0.1116, |
| "step": 70650 |
| }, |
| { |
| "epoch": 1.4337862502534984, |
| "grad_norm": 0.56640625, |
| "learning_rate": 0.00025632817128971326, |
| "loss": 0.1163, |
| "step": 70700 |
| }, |
| { |
| "epoch": 1.434800243358345, |
| "grad_norm": 0.50390625, |
| "learning_rate": 0.00025625719106917065, |
| "loss": 0.1139, |
| "step": 70750 |
| }, |
| { |
| "epoch": 1.435814236463192, |
| "grad_norm": 0.44140625, |
| "learning_rate": 0.00025618616305800437, |
| "loss": 0.1053, |
| "step": 70800 |
| }, |
| { |
| "epoch": 1.436828229568039, |
| "grad_norm": 0.4765625, |
| "learning_rate": 0.0002561150872881601, |
| "loss": 0.1142, |
| "step": 70850 |
| }, |
| { |
| "epoch": 1.4378422226728858, |
| "grad_norm": 0.443359375, |
| "learning_rate": 0.00025604396379160524, |
| "loss": 0.1098, |
| "step": 70900 |
| }, |
| { |
| "epoch": 1.4388562157777327, |
| "grad_norm": 0.4375, |
| "learning_rate": 0.00025597279260032856, |
| "loss": 0.1152, |
| "step": 70950 |
| }, |
| { |
| "epoch": 1.4398702088825797, |
| "grad_norm": 0.46484375, |
| "learning_rate": 0.0002559015737463402, |
| "loss": 0.1055, |
| "step": 71000 |
| }, |
| { |
| "epoch": 1.4408842019874264, |
| "grad_norm": 0.490234375, |
| "learning_rate": 0.00025583030726167186, |
| "loss": 0.1115, |
| "step": 71050 |
| }, |
| { |
| "epoch": 1.4418981950922733, |
| "grad_norm": 0.46875, |
| "learning_rate": 0.0002557589931783766, |
| "loss": 0.1056, |
| "step": 71100 |
| }, |
| { |
| "epoch": 1.4429121881971203, |
| "grad_norm": 0.5546875, |
| "learning_rate": 0.0002556876315285289, |
| "loss": 0.1037, |
| "step": 71150 |
| }, |
| { |
| "epoch": 1.443926181301967, |
| "grad_norm": 0.455078125, |
| "learning_rate": 0.00025561622234422475, |
| "loss": 0.1278, |
| "step": 71200 |
| }, |
| { |
| "epoch": 1.444940174406814, |
| "grad_norm": 0.546875, |
| "learning_rate": 0.0002555447656575812, |
| "loss": 0.1075, |
| "step": 71250 |
| }, |
| { |
| "epoch": 1.445954167511661, |
| "grad_norm": 0.5078125, |
| "learning_rate": 0.00025547326150073696, |
| "loss": 0.1063, |
| "step": 71300 |
| }, |
| { |
| "epoch": 1.4469681606165077, |
| "grad_norm": 0.482421875, |
| "learning_rate": 0.00025540170990585214, |
| "loss": 0.1095, |
| "step": 71350 |
| }, |
| { |
| "epoch": 1.4479821537213546, |
| "grad_norm": 0.431640625, |
| "learning_rate": 0.0002553301109051078, |
| "loss": 0.1155, |
| "step": 71400 |
| }, |
| { |
| "epoch": 1.4489961468262016, |
| "grad_norm": 0.431640625, |
| "learning_rate": 0.0002552584645307068, |
| "loss": 0.1011, |
| "step": 71450 |
| }, |
| { |
| "epoch": 1.4500101399310485, |
| "grad_norm": 0.498046875, |
| "learning_rate": 0.0002551867708148729, |
| "loss": 0.1079, |
| "step": 71500 |
| }, |
| { |
| "epoch": 1.4510241330358953, |
| "grad_norm": 0.69921875, |
| "learning_rate": 0.0002551150297898514, |
| "loss": 0.1103, |
| "step": 71550 |
| }, |
| { |
| "epoch": 1.4520381261407422, |
| "grad_norm": 0.443359375, |
| "learning_rate": 0.00025504324148790887, |
| "loss": 0.1136, |
| "step": 71600 |
| }, |
| { |
| "epoch": 1.4530521192455892, |
| "grad_norm": 0.486328125, |
| "learning_rate": 0.00025497140594133295, |
| "loss": 0.109, |
| "step": 71650 |
| }, |
| { |
| "epoch": 1.4540661123504361, |
| "grad_norm": 0.5625, |
| "learning_rate": 0.00025489952318243276, |
| "loss": 0.1051, |
| "step": 71700 |
| }, |
| { |
| "epoch": 1.4550801054552829, |
| "grad_norm": 0.5078125, |
| "learning_rate": 0.0002548275932435385, |
| "loss": 0.1064, |
| "step": 71750 |
| }, |
| { |
| "epoch": 1.4560940985601298, |
| "grad_norm": 0.486328125, |
| "learning_rate": 0.0002547556161570017, |
| "loss": 0.1107, |
| "step": 71800 |
| }, |
| { |
| "epoch": 1.4571080916649768, |
| "grad_norm": 0.49609375, |
| "learning_rate": 0.00025468359195519495, |
| "loss": 0.106, |
| "step": 71850 |
| }, |
| { |
| "epoch": 1.4581220847698235, |
| "grad_norm": 0.39453125, |
| "learning_rate": 0.0002546115206705122, |
| "loss": 0.1016, |
| "step": 71900 |
| }, |
| { |
| "epoch": 1.4591360778746705, |
| "grad_norm": 0.482421875, |
| "learning_rate": 0.0002545394023353685, |
| "loss": 0.1131, |
| "step": 71950 |
| }, |
| { |
| "epoch": 1.4601500709795174, |
| "grad_norm": 0.546875, |
| "learning_rate": 0.0002544672369821999, |
| "loss": 0.1045, |
| "step": 72000 |
| }, |
| { |
| "epoch": 1.4601500709795174, |
| "eval_cer": 0.030864550765501205, |
| "eval_loss": 0.16758893430233002, |
| "eval_runtime": 128.1097, |
| "eval_samples_per_second": 21.919, |
| "eval_steps_per_second": 0.172, |
| "eval_wer": 0.09211862685354459, |
| "step": 72000 |
| }, |
| { |
| "epoch": 1.4611640640843642, |
| "grad_norm": 0.4609375, |
| "learning_rate": 0.00025439502464346395, |
| "loss": 0.107, |
| "step": 72050 |
| }, |
| { |
| "epoch": 1.4621780571892111, |
| "grad_norm": 0.42578125, |
| "learning_rate": 0.0002543227653516391, |
| "loss": 0.1042, |
| "step": 72100 |
| }, |
| { |
| "epoch": 1.463192050294058, |
| "grad_norm": 0.5078125, |
| "learning_rate": 0.0002542504591392249, |
| "loss": 0.1144, |
| "step": 72150 |
| }, |
| { |
| "epoch": 1.4642060433989048, |
| "grad_norm": 0.421875, |
| "learning_rate": 0.000254178106038742, |
| "loss": 0.1123, |
| "step": 72200 |
| }, |
| { |
| "epoch": 1.4652200365037518, |
| "grad_norm": 0.5, |
| "learning_rate": 0.0002541057060827323, |
| "loss": 0.1156, |
| "step": 72250 |
| }, |
| { |
| "epoch": 1.4662340296085987, |
| "grad_norm": 0.55078125, |
| "learning_rate": 0.00025403325930375867, |
| "loss": 0.1076, |
| "step": 72300 |
| }, |
| { |
| "epoch": 1.4672480227134455, |
| "grad_norm": 0.515625, |
| "learning_rate": 0.000253960765734405, |
| "loss": 0.1211, |
| "step": 72350 |
| }, |
| { |
| "epoch": 1.4682620158182924, |
| "grad_norm": 0.443359375, |
| "learning_rate": 0.00025388822540727625, |
| "loss": 0.1082, |
| "step": 72400 |
| }, |
| { |
| "epoch": 1.4692760089231394, |
| "grad_norm": 0.55859375, |
| "learning_rate": 0.0002538156383549984, |
| "loss": 0.1055, |
| "step": 72450 |
| }, |
| { |
| "epoch": 1.470290002027986, |
| "grad_norm": 0.490234375, |
| "learning_rate": 0.0002537430046102186, |
| "loss": 0.1095, |
| "step": 72500 |
| }, |
| { |
| "epoch": 1.471303995132833, |
| "grad_norm": 0.53515625, |
| "learning_rate": 0.0002536703242056047, |
| "loss": 0.1101, |
| "step": 72550 |
| }, |
| { |
| "epoch": 1.47231798823768, |
| "grad_norm": 0.490234375, |
| "learning_rate": 0.00025359759717384585, |
| "loss": 0.1079, |
| "step": 72600 |
| }, |
| { |
| "epoch": 1.4733319813425267, |
| "grad_norm": 0.447265625, |
| "learning_rate": 0.00025352482354765193, |
| "loss": 0.106, |
| "step": 72650 |
| }, |
| { |
| "epoch": 1.4743459744473737, |
| "grad_norm": 0.53515625, |
| "learning_rate": 0.0002534520033597539, |
| "loss": 0.1101, |
| "step": 72700 |
| }, |
| { |
| "epoch": 1.4753599675522207, |
| "grad_norm": 0.455078125, |
| "learning_rate": 0.0002533791366429036, |
| "loss": 0.1158, |
| "step": 72750 |
| }, |
| { |
| "epoch": 1.4763739606570676, |
| "grad_norm": 0.474609375, |
| "learning_rate": 0.00025330622342987396, |
| "loss": 0.104, |
| "step": 72800 |
| }, |
| { |
| "epoch": 1.4773879537619143, |
| "grad_norm": 0.59375, |
| "learning_rate": 0.00025323326375345853, |
| "loss": 0.1024, |
| "step": 72850 |
| }, |
| { |
| "epoch": 1.4784019468667613, |
| "grad_norm": 0.44140625, |
| "learning_rate": 0.0002531602576464721, |
| "loss": 0.1044, |
| "step": 72900 |
| }, |
| { |
| "epoch": 1.4794159399716083, |
| "grad_norm": 0.53125, |
| "learning_rate": 0.00025308720514175, |
| "loss": 0.1053, |
| "step": 72950 |
| }, |
| { |
| "epoch": 1.4804299330764552, |
| "grad_norm": 0.455078125, |
| "learning_rate": 0.0002530141062721486, |
| "loss": 0.1079, |
| "step": 73000 |
| }, |
| { |
| "epoch": 1.481443926181302, |
| "grad_norm": 0.4765625, |
| "learning_rate": 0.00025294096107054525, |
| "loss": 0.1059, |
| "step": 73050 |
| }, |
| { |
| "epoch": 1.482457919286149, |
| "grad_norm": 0.435546875, |
| "learning_rate": 0.00025286776956983786, |
| "loss": 0.1017, |
| "step": 73100 |
| }, |
| { |
| "epoch": 1.4834719123909959, |
| "grad_norm": 0.55078125, |
| "learning_rate": 0.0002527945318029454, |
| "loss": 0.1073, |
| "step": 73150 |
| }, |
| { |
| "epoch": 1.4844859054958426, |
| "grad_norm": 0.4921875, |
| "learning_rate": 0.00025272124780280754, |
| "loss": 0.104, |
| "step": 73200 |
| }, |
| { |
| "epoch": 1.4854998986006895, |
| "grad_norm": 0.474609375, |
| "learning_rate": 0.0002526479176023847, |
| "loss": 0.1118, |
| "step": 73250 |
| }, |
| { |
| "epoch": 1.4865138917055365, |
| "grad_norm": 0.427734375, |
| "learning_rate": 0.00025257454123465826, |
| "loss": 0.1031, |
| "step": 73300 |
| }, |
| { |
| "epoch": 1.4875278848103832, |
| "grad_norm": 0.52734375, |
| "learning_rate": 0.0002525011187326301, |
| "loss": 0.1084, |
| "step": 73350 |
| }, |
| { |
| "epoch": 1.4885418779152302, |
| "grad_norm": 0.4609375, |
| "learning_rate": 0.0002524276501293231, |
| "loss": 0.1104, |
| "step": 73400 |
| }, |
| { |
| "epoch": 1.4895558710200771, |
| "grad_norm": 0.453125, |
| "learning_rate": 0.0002523541354577807, |
| "loss": 0.108, |
| "step": 73450 |
| }, |
| { |
| "epoch": 1.4905698641249239, |
| "grad_norm": 0.53125, |
| "learning_rate": 0.00025228057475106723, |
| "loss": 0.1221, |
| "step": 73500 |
| }, |
| { |
| "epoch": 1.4905698641249239, |
| "eval_cer": 0.03172291093342767, |
| "eval_loss": 0.16636300086975098, |
| "eval_runtime": 127.4384, |
| "eval_samples_per_second": 22.034, |
| "eval_steps_per_second": 0.173, |
| "eval_wer": 0.09469158372266233, |
| "step": 73500 |
| }, |
| { |
| "epoch": 1.4915838572297708, |
| "grad_norm": 0.466796875, |
| "learning_rate": 0.00025220696804226754, |
| "loss": 0.1057, |
| "step": 73550 |
| }, |
| { |
| "epoch": 1.4925978503346178, |
| "grad_norm": 0.484375, |
| "learning_rate": 0.0002521333153644872, |
| "loss": 0.1189, |
| "step": 73600 |
| }, |
| { |
| "epoch": 1.4936118434394645, |
| "grad_norm": 0.443359375, |
| "learning_rate": 0.00025205961675085265, |
| "loss": 0.11, |
| "step": 73650 |
| }, |
| { |
| "epoch": 1.4946258365443115, |
| "grad_norm": 0.4921875, |
| "learning_rate": 0.0002519858722345108, |
| "loss": 0.1086, |
| "step": 73700 |
| }, |
| { |
| "epoch": 1.4956398296491584, |
| "grad_norm": 0.43359375, |
| "learning_rate": 0.0002519120818486291, |
| "loss": 0.102, |
| "step": 73750 |
| }, |
| { |
| "epoch": 1.4966538227540052, |
| "grad_norm": 0.578125, |
| "learning_rate": 0.000251838245626396, |
| "loss": 0.0992, |
| "step": 73800 |
| }, |
| { |
| "epoch": 1.4976678158588521, |
| "grad_norm": 0.4921875, |
| "learning_rate": 0.0002517643636010203, |
| "loss": 0.1055, |
| "step": 73850 |
| }, |
| { |
| "epoch": 1.498681808963699, |
| "grad_norm": 0.458984375, |
| "learning_rate": 0.0002516904358057313, |
| "loss": 0.1112, |
| "step": 73900 |
| }, |
| { |
| "epoch": 1.4996958020685458, |
| "grad_norm": 0.52734375, |
| "learning_rate": 0.00025161646227377923, |
| "loss": 0.1111, |
| "step": 73950 |
| }, |
| { |
| "epoch": 1.500709795173393, |
| "grad_norm": 0.490234375, |
| "learning_rate": 0.0002515424430384346, |
| "loss": 0.1123, |
| "step": 74000 |
| }, |
| { |
| "epoch": 1.5017237882782397, |
| "grad_norm": 0.435546875, |
| "learning_rate": 0.00025146837813298854, |
| "loss": 0.1148, |
| "step": 74050 |
| }, |
| { |
| "epoch": 1.5027377813830864, |
| "grad_norm": 0.51171875, |
| "learning_rate": 0.00025139426759075294, |
| "loss": 0.1003, |
| "step": 74100 |
| }, |
| { |
| "epoch": 1.5037517744879336, |
| "grad_norm": 0.578125, |
| "learning_rate": 0.0002513201114450598, |
| "loss": 0.1031, |
| "step": 74150 |
| }, |
| { |
| "epoch": 1.5047657675927804, |
| "grad_norm": 0.447265625, |
| "learning_rate": 0.0002512459097292619, |
| "loss": 0.1107, |
| "step": 74200 |
| }, |
| { |
| "epoch": 1.505779760697627, |
| "grad_norm": 0.478515625, |
| "learning_rate": 0.00025117166247673255, |
| "loss": 0.1146, |
| "step": 74250 |
| }, |
| { |
| "epoch": 1.5067937538024743, |
| "grad_norm": 0.47265625, |
| "learning_rate": 0.00025109736972086556, |
| "loss": 0.1065, |
| "step": 74300 |
| }, |
| { |
| "epoch": 1.507807746907321, |
| "grad_norm": 0.486328125, |
| "learning_rate": 0.0002510230314950749, |
| "loss": 0.1019, |
| "step": 74350 |
| }, |
| { |
| "epoch": 1.508821740012168, |
| "grad_norm": 0.494140625, |
| "learning_rate": 0.0002509486478327953, |
| "loss": 0.1144, |
| "step": 74400 |
| }, |
| { |
| "epoch": 1.509835733117015, |
| "grad_norm": 0.458984375, |
| "learning_rate": 0.00025087421876748183, |
| "loss": 0.1062, |
| "step": 74450 |
| }, |
| { |
| "epoch": 1.5108497262218616, |
| "grad_norm": 0.55078125, |
| "learning_rate": 0.00025079974433260993, |
| "loss": 0.1161, |
| "step": 74500 |
| }, |
| { |
| "epoch": 1.5118637193267086, |
| "grad_norm": 0.54296875, |
| "learning_rate": 0.0002507252245616756, |
| "loss": 0.11, |
| "step": 74550 |
| }, |
| { |
| "epoch": 1.5128777124315556, |
| "grad_norm": 0.41796875, |
| "learning_rate": 0.00025065065948819506, |
| "loss": 0.101, |
| "step": 74600 |
| }, |
| { |
| "epoch": 1.5138917055364023, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.000250576049145705, |
| "loss": 0.1134, |
| "step": 74650 |
| }, |
| { |
| "epoch": 1.5149056986412492, |
| "grad_norm": 0.55859375, |
| "learning_rate": 0.0002505013935677624, |
| "loss": 0.1045, |
| "step": 74700 |
| }, |
| { |
| "epoch": 1.5159196917460962, |
| "grad_norm": 0.427734375, |
| "learning_rate": 0.0002504266927879446, |
| "loss": 0.1068, |
| "step": 74750 |
| }, |
| { |
| "epoch": 1.516933684850943, |
| "grad_norm": 0.443359375, |
| "learning_rate": 0.0002503519468398495, |
| "loss": 0.1057, |
| "step": 74800 |
| }, |
| { |
| "epoch": 1.5179476779557899, |
| "grad_norm": 0.50390625, |
| "learning_rate": 0.00025027715575709475, |
| "loss": 0.1029, |
| "step": 74850 |
| }, |
| { |
| "epoch": 1.5189616710606368, |
| "grad_norm": 0.56640625, |
| "learning_rate": 0.00025020231957331897, |
| "loss": 0.1096, |
| "step": 74900 |
| }, |
| { |
| "epoch": 1.5199756641654836, |
| "grad_norm": 0.44140625, |
| "learning_rate": 0.0002501274383221806, |
| "loss": 0.1098, |
| "step": 74950 |
| }, |
| { |
| "epoch": 1.5209896572703305, |
| "grad_norm": 0.609375, |
| "learning_rate": 0.0002500525120373586, |
| "loss": 0.1069, |
| "step": 75000 |
| }, |
| { |
| "epoch": 1.5209896572703305, |
| "eval_cer": 0.031192288284164035, |
| "eval_loss": 0.16423167288303375, |
| "eval_runtime": 132.6118, |
| "eval_samples_per_second": 21.175, |
| "eval_steps_per_second": 0.166, |
| "eval_wer": 0.09225404563612973, |
| "step": 75000 |
| }, |
| { |
| "epoch": 1.5220036503751775, |
| "grad_norm": 0.4609375, |
| "learning_rate": 0.0002499775407525521, |
| "loss": 0.1113, |
| "step": 75050 |
| }, |
| { |
| "epoch": 1.5230176434800242, |
| "grad_norm": 0.56640625, |
| "learning_rate": 0.0002499025245014803, |
| "loss": 0.103, |
| "step": 75100 |
| }, |
| { |
| "epoch": 1.5240316365848712, |
| "grad_norm": 0.466796875, |
| "learning_rate": 0.00024982746331788297, |
| "loss": 0.1089, |
| "step": 75150 |
| }, |
| { |
| "epoch": 1.5250456296897181, |
| "grad_norm": 0.45703125, |
| "learning_rate": 0.00024975235723551976, |
| "loss": 0.1048, |
| "step": 75200 |
| }, |
| { |
| "epoch": 1.5260596227945649, |
| "grad_norm": 0.44140625, |
| "learning_rate": 0.00024967720628817075, |
| "loss": 0.1116, |
| "step": 75250 |
| }, |
| { |
| "epoch": 1.527073615899412, |
| "grad_norm": 0.484375, |
| "learning_rate": 0.0002496020105096361, |
| "loss": 0.1109, |
| "step": 75300 |
| }, |
| { |
| "epoch": 1.5280876090042588, |
| "grad_norm": 0.498046875, |
| "learning_rate": 0.0002495267699337361, |
| "loss": 0.1092, |
| "step": 75350 |
| }, |
| { |
| "epoch": 1.5291016021091055, |
| "grad_norm": 0.5390625, |
| "learning_rate": 0.0002494514845943112, |
| "loss": 0.1077, |
| "step": 75400 |
| }, |
| { |
| "epoch": 1.5301155952139527, |
| "grad_norm": 0.451171875, |
| "learning_rate": 0.0002493761545252221, |
| "loss": 0.1145, |
| "step": 75450 |
| }, |
| { |
| "epoch": 1.5311295883187994, |
| "grad_norm": 0.56640625, |
| "learning_rate": 0.00024930077976034943, |
| "loss": 0.1142, |
| "step": 75500 |
| }, |
| { |
| "epoch": 1.5321435814236464, |
| "grad_norm": 0.490234375, |
| "learning_rate": 0.0002492253603335942, |
| "loss": 0.1109, |
| "step": 75550 |
| }, |
| { |
| "epoch": 1.5331575745284933, |
| "grad_norm": 0.5390625, |
| "learning_rate": 0.00024914989627887707, |
| "loss": 0.1086, |
| "step": 75600 |
| }, |
| { |
| "epoch": 1.53417156763334, |
| "grad_norm": 0.49609375, |
| "learning_rate": 0.0002490743876301392, |
| "loss": 0.1249, |
| "step": 75650 |
| }, |
| { |
| "epoch": 1.535185560738187, |
| "grad_norm": 0.486328125, |
| "learning_rate": 0.0002489988344213416, |
| "loss": 0.1147, |
| "step": 75700 |
| }, |
| { |
| "epoch": 1.536199553843034, |
| "grad_norm": 0.482421875, |
| "learning_rate": 0.00024892323668646536, |
| "loss": 0.1007, |
| "step": 75750 |
| }, |
| { |
| "epoch": 1.5372135469478807, |
| "grad_norm": 0.451171875, |
| "learning_rate": 0.0002488475944595116, |
| "loss": 0.1163, |
| "step": 75800 |
| }, |
| { |
| "epoch": 1.5382275400527277, |
| "grad_norm": 0.447265625, |
| "learning_rate": 0.0002487719077745014, |
| "loss": 0.1034, |
| "step": 75850 |
| }, |
| { |
| "epoch": 1.5392415331575746, |
| "grad_norm": 0.4453125, |
| "learning_rate": 0.00024869617666547594, |
| "loss": 0.1084, |
| "step": 75900 |
| }, |
| { |
| "epoch": 1.5402555262624213, |
| "grad_norm": 0.55859375, |
| "learning_rate": 0.00024862040116649625, |
| "loss": 0.1069, |
| "step": 75950 |
| }, |
| { |
| "epoch": 1.5412695193672683, |
| "grad_norm": 0.51953125, |
| "learning_rate": 0.00024854458131164347, |
| "loss": 0.108, |
| "step": 76000 |
| }, |
| { |
| "epoch": 1.5422835124721153, |
| "grad_norm": 0.47265625, |
| "learning_rate": 0.00024846871713501864, |
| "loss": 0.1068, |
| "step": 76050 |
| }, |
| { |
| "epoch": 1.543297505576962, |
| "grad_norm": 0.50390625, |
| "learning_rate": 0.00024839280867074257, |
| "loss": 0.1014, |
| "step": 76100 |
| }, |
| { |
| "epoch": 1.544311498681809, |
| "grad_norm": 0.55078125, |
| "learning_rate": 0.0002483168559529563, |
| "loss": 0.1049, |
| "step": 76150 |
| }, |
| { |
| "epoch": 1.545325491786656, |
| "grad_norm": 0.53515625, |
| "learning_rate": 0.0002482408590158205, |
| "loss": 0.1043, |
| "step": 76200 |
| }, |
| { |
| "epoch": 1.5463394848915026, |
| "grad_norm": 0.474609375, |
| "learning_rate": 0.0002481648178935158, |
| "loss": 0.1031, |
| "step": 76250 |
| }, |
| { |
| "epoch": 1.5473534779963496, |
| "grad_norm": 0.455078125, |
| "learning_rate": 0.00024808873262024283, |
| "loss": 0.1089, |
| "step": 76300 |
| }, |
| { |
| "epoch": 1.5483674711011965, |
| "grad_norm": 0.5, |
| "learning_rate": 0.0002480126032302219, |
| "loss": 0.1165, |
| "step": 76350 |
| }, |
| { |
| "epoch": 1.5493814642060433, |
| "grad_norm": 0.41796875, |
| "learning_rate": 0.00024793642975769333, |
| "loss": 0.112, |
| "step": 76400 |
| }, |
| { |
| "epoch": 1.5503954573108902, |
| "grad_norm": 0.51171875, |
| "learning_rate": 0.0002478602122369171, |
| "loss": 0.1002, |
| "step": 76450 |
| }, |
| { |
| "epoch": 1.5514094504157372, |
| "grad_norm": 0.4609375, |
| "learning_rate": 0.0002477839507021731, |
| "loss": 0.1097, |
| "step": 76500 |
| }, |
| { |
| "epoch": 1.5514094504157372, |
| "eval_cer": 0.02821143751918305, |
| "eval_loss": 0.16545580327510834, |
| "eval_runtime": 130.6901, |
| "eval_samples_per_second": 21.486, |
| "eval_steps_per_second": 0.168, |
| "eval_wer": 0.08995192633218227, |
| "step": 76500 |
| }, |
| { |
| "epoch": 1.552423443520584, |
| "grad_norm": 0.44140625, |
| "learning_rate": 0.000247707645187761, |
| "loss": 0.1089, |
| "step": 76550 |
| }, |
| { |
| "epoch": 1.553437436625431, |
| "grad_norm": 0.515625, |
| "learning_rate": 0.0002476312957280003, |
| "loss": 0.1065, |
| "step": 76600 |
| }, |
| { |
| "epoch": 1.5544514297302778, |
| "grad_norm": 0.486328125, |
| "learning_rate": 0.00024755490235723015, |
| "loss": 0.1152, |
| "step": 76650 |
| }, |
| { |
| "epoch": 1.5554654228351246, |
| "grad_norm": 0.578125, |
| "learning_rate": 0.00024747846510980953, |
| "loss": 0.1117, |
| "step": 76700 |
| }, |
| { |
| "epoch": 1.5564794159399717, |
| "grad_norm": 0.53125, |
| "learning_rate": 0.0002474019840201172, |
| "loss": 0.1045, |
| "step": 76750 |
| }, |
| { |
| "epoch": 1.5574934090448185, |
| "grad_norm": 0.7421875, |
| "learning_rate": 0.0002473254591225514, |
| "loss": 0.1048, |
| "step": 76800 |
| }, |
| { |
| "epoch": 1.5585074021496654, |
| "grad_norm": 0.45703125, |
| "learning_rate": 0.0002472488904515304, |
| "loss": 0.111, |
| "step": 76850 |
| }, |
| { |
| "epoch": 1.5595213952545124, |
| "grad_norm": 0.494140625, |
| "learning_rate": 0.00024717227804149205, |
| "loss": 0.1055, |
| "step": 76900 |
| }, |
| { |
| "epoch": 1.5605353883593591, |
| "grad_norm": 0.466796875, |
| "learning_rate": 0.00024709562192689363, |
| "loss": 0.1107, |
| "step": 76950 |
| }, |
| { |
| "epoch": 1.561549381464206, |
| "grad_norm": 0.56640625, |
| "learning_rate": 0.00024701892214221247, |
| "loss": 0.1081, |
| "step": 77000 |
| }, |
| { |
| "epoch": 1.562563374569053, |
| "grad_norm": 0.5, |
| "learning_rate": 0.0002469421787219452, |
| "loss": 0.1189, |
| "step": 77050 |
| }, |
| { |
| "epoch": 1.5635773676738998, |
| "grad_norm": 0.51953125, |
| "learning_rate": 0.00024686539170060834, |
| "loss": 0.1131, |
| "step": 77100 |
| }, |
| { |
| "epoch": 1.5645913607787467, |
| "grad_norm": 0.478515625, |
| "learning_rate": 0.0002467885611127379, |
| "loss": 0.1194, |
| "step": 77150 |
| }, |
| { |
| "epoch": 1.5656053538835937, |
| "grad_norm": 0.380859375, |
| "learning_rate": 0.0002467116869928894, |
| "loss": 0.1119, |
| "step": 77200 |
| }, |
| { |
| "epoch": 1.5666193469884404, |
| "grad_norm": 0.5, |
| "learning_rate": 0.00024663476937563804, |
| "loss": 0.1053, |
| "step": 77250 |
| }, |
| { |
| "epoch": 1.5676333400932874, |
| "grad_norm": 0.51171875, |
| "learning_rate": 0.0002465578082955786, |
| "loss": 0.1016, |
| "step": 77300 |
| }, |
| { |
| "epoch": 1.5686473331981343, |
| "grad_norm": 0.4296875, |
| "learning_rate": 0.00024648080378732537, |
| "loss": 0.1087, |
| "step": 77350 |
| }, |
| { |
| "epoch": 1.569661326302981, |
| "grad_norm": 0.515625, |
| "learning_rate": 0.00024640375588551213, |
| "loss": 0.1133, |
| "step": 77400 |
| }, |
| { |
| "epoch": 1.570675319407828, |
| "grad_norm": 0.52734375, |
| "learning_rate": 0.00024632666462479243, |
| "loss": 0.1123, |
| "step": 77450 |
| }, |
| { |
| "epoch": 1.571689312512675, |
| "grad_norm": 0.421875, |
| "learning_rate": 0.0002462495300398388, |
| "loss": 0.1154, |
| "step": 77500 |
| }, |
| { |
| "epoch": 1.5727033056175217, |
| "grad_norm": 0.435546875, |
| "learning_rate": 0.00024617235216534383, |
| "loss": 0.1068, |
| "step": 77550 |
| }, |
| { |
| "epoch": 1.5737172987223687, |
| "grad_norm": 0.51171875, |
| "learning_rate": 0.00024609513103601927, |
| "loss": 0.1102, |
| "step": 77600 |
| }, |
| { |
| "epoch": 1.5747312918272156, |
| "grad_norm": 0.498046875, |
| "learning_rate": 0.00024601786668659626, |
| "loss": 0.1027, |
| "step": 77650 |
| }, |
| { |
| "epoch": 1.5757452849320623, |
| "grad_norm": 0.470703125, |
| "learning_rate": 0.0002459405591518256, |
| "loss": 0.103, |
| "step": 77700 |
| }, |
| { |
| "epoch": 1.5767592780369093, |
| "grad_norm": 0.55078125, |
| "learning_rate": 0.0002458632084664774, |
| "loss": 0.0998, |
| "step": 77750 |
| }, |
| { |
| "epoch": 1.5777732711417563, |
| "grad_norm": 0.4609375, |
| "learning_rate": 0.0002457858146653412, |
| "loss": 0.1073, |
| "step": 77800 |
| }, |
| { |
| "epoch": 1.578787264246603, |
| "grad_norm": 0.458984375, |
| "learning_rate": 0.00024570837778322584, |
| "loss": 0.1014, |
| "step": 77850 |
| }, |
| { |
| "epoch": 1.5798012573514502, |
| "grad_norm": 0.451171875, |
| "learning_rate": 0.0002456308978549597, |
| "loss": 0.1076, |
| "step": 77900 |
| }, |
| { |
| "epoch": 1.580815250456297, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.0002455533749153904, |
| "loss": 0.1071, |
| "step": 77950 |
| }, |
| { |
| "epoch": 1.5818292435611436, |
| "grad_norm": 0.453125, |
| "learning_rate": 0.000245475808999385, |
| "loss": 0.1008, |
| "step": 78000 |
| }, |
| { |
| "epoch": 1.5818292435611436, |
| "eval_cer": 0.027639197407232075, |
| "eval_loss": 0.16488178074359894, |
| "eval_runtime": 127.6157, |
| "eval_samples_per_second": 22.004, |
| "eval_steps_per_second": 0.172, |
| "eval_wer": 0.08788678989775882, |
| "step": 78000 |
| }, |
| { |
| "epoch": 1.5828432366659908, |
| "grad_norm": 0.470703125, |
| "learning_rate": 0.0002453982001418298, |
| "loss": 0.1016, |
| "step": 78050 |
| }, |
| { |
| "epoch": 1.5838572297708375, |
| "grad_norm": 0.451171875, |
| "learning_rate": 0.0002453205483776303, |
| "loss": 0.1195, |
| "step": 78100 |
| }, |
| { |
| "epoch": 1.5848712228756845, |
| "grad_norm": 0.57421875, |
| "learning_rate": 0.00024524285374171167, |
| "loss": 0.112, |
| "step": 78150 |
| }, |
| { |
| "epoch": 1.5858852159805314, |
| "grad_norm": 0.5078125, |
| "learning_rate": 0.000245165116269018, |
| "loss": 0.1045, |
| "step": 78200 |
| }, |
| { |
| "epoch": 1.5868992090853782, |
| "grad_norm": 0.4375, |
| "learning_rate": 0.0002450873359945128, |
| "loss": 0.1025, |
| "step": 78250 |
| }, |
| { |
| "epoch": 1.5879132021902251, |
| "grad_norm": 0.5078125, |
| "learning_rate": 0.0002450095129531788, |
| "loss": 0.105, |
| "step": 78300 |
| }, |
| { |
| "epoch": 1.588927195295072, |
| "grad_norm": 0.609375, |
| "learning_rate": 0.0002449316471800181, |
| "loss": 0.105, |
| "step": 78350 |
| }, |
| { |
| "epoch": 1.5899411883999188, |
| "grad_norm": 0.41796875, |
| "learning_rate": 0.00024485373871005174, |
| "loss": 0.1076, |
| "step": 78400 |
| }, |
| { |
| "epoch": 1.5909551815047658, |
| "grad_norm": 0.546875, |
| "learning_rate": 0.00024477578757832013, |
| "loss": 0.1097, |
| "step": 78450 |
| }, |
| { |
| "epoch": 1.5919691746096127, |
| "grad_norm": 0.474609375, |
| "learning_rate": 0.000244697793819883, |
| "loss": 0.0967, |
| "step": 78500 |
| }, |
| { |
| "epoch": 1.5929831677144595, |
| "grad_norm": 0.50390625, |
| "learning_rate": 0.0002446197574698189, |
| "loss": 0.1034, |
| "step": 78550 |
| }, |
| { |
| "epoch": 1.5939971608193064, |
| "grad_norm": 0.478515625, |
| "learning_rate": 0.000244541678563226, |
| "loss": 0.1228, |
| "step": 78600 |
| }, |
| { |
| "epoch": 1.5950111539241534, |
| "grad_norm": 0.55078125, |
| "learning_rate": 0.0002444635571352211, |
| "loss": 0.1063, |
| "step": 78650 |
| }, |
| { |
| "epoch": 1.5960251470290001, |
| "grad_norm": 0.46875, |
| "learning_rate": 0.00024438539322094047, |
| "loss": 0.1098, |
| "step": 78700 |
| }, |
| { |
| "epoch": 1.597039140133847, |
| "grad_norm": 0.4921875, |
| "learning_rate": 0.0002443071868555395, |
| "loss": 0.1023, |
| "step": 78750 |
| }, |
| { |
| "epoch": 1.598053133238694, |
| "grad_norm": 0.5, |
| "learning_rate": 0.0002442289380741924, |
| "loss": 0.1072, |
| "step": 78800 |
| }, |
| { |
| "epoch": 1.5990671263435408, |
| "grad_norm": 0.5078125, |
| "learning_rate": 0.00024415064691209274, |
| "loss": 0.1055, |
| "step": 78850 |
| }, |
| { |
| "epoch": 1.6000811194483877, |
| "grad_norm": 0.45703125, |
| "learning_rate": 0.00024407231340445307, |
| "loss": 0.1113, |
| "step": 78900 |
| }, |
| { |
| "epoch": 1.6010951125532347, |
| "grad_norm": 0.52734375, |
| "learning_rate": 0.00024399393758650493, |
| "loss": 0.1024, |
| "step": 78950 |
| }, |
| { |
| "epoch": 1.6021091056580814, |
| "grad_norm": 0.48828125, |
| "learning_rate": 0.00024391551949349882, |
| "loss": 0.1129, |
| "step": 79000 |
| }, |
| { |
| "epoch": 1.6031230987629286, |
| "grad_norm": 0.474609375, |
| "learning_rate": 0.00024383705916070444, |
| "loss": 0.0993, |
| "step": 79050 |
| }, |
| { |
| "epoch": 1.6041370918677753, |
| "grad_norm": 0.50390625, |
| "learning_rate": 0.0002437585566234104, |
| "loss": 0.1073, |
| "step": 79100 |
| }, |
| { |
| "epoch": 1.605151084972622, |
| "grad_norm": 0.439453125, |
| "learning_rate": 0.00024368001191692423, |
| "loss": 0.1026, |
| "step": 79150 |
| }, |
| { |
| "epoch": 1.6061650780774692, |
| "grad_norm": 0.412109375, |
| "learning_rate": 0.00024360142507657255, |
| "loss": 0.1002, |
| "step": 79200 |
| }, |
| { |
| "epoch": 1.607179071182316, |
| "grad_norm": 0.515625, |
| "learning_rate": 0.00024352279613770083, |
| "loss": 0.1042, |
| "step": 79250 |
| }, |
| { |
| "epoch": 1.6081930642871627, |
| "grad_norm": 0.53125, |
| "learning_rate": 0.00024344412513567347, |
| "loss": 0.1088, |
| "step": 79300 |
| }, |
| { |
| "epoch": 1.6092070573920099, |
| "grad_norm": 0.486328125, |
| "learning_rate": 0.00024336541210587392, |
| "loss": 0.105, |
| "step": 79350 |
| }, |
| { |
| "epoch": 1.6102210504968566, |
| "grad_norm": 0.56640625, |
| "learning_rate": 0.00024328665708370437, |
| "loss": 0.1089, |
| "step": 79400 |
| }, |
| { |
| "epoch": 1.6112350436017036, |
| "grad_norm": 0.4140625, |
| "learning_rate": 0.000243207860104586, |
| "loss": 0.1117, |
| "step": 79450 |
| }, |
| { |
| "epoch": 1.6122490367065505, |
| "grad_norm": 0.466796875, |
| "learning_rate": 0.00024312902120395883, |
| "loss": 0.1057, |
| "step": 79500 |
| }, |
| { |
| "epoch": 1.6122490367065505, |
| "eval_cer": 0.031676091287904405, |
| "eval_loss": 0.16220040619373322, |
| "eval_runtime": 128.8141, |
| "eval_samples_per_second": 21.799, |
| "eval_steps_per_second": 0.171, |
| "eval_wer": 0.09296499424470174, |
| "step": 79500 |
| }, |
| { |
| "epoch": 1.6132630298113972, |
| "grad_norm": 0.515625, |
| "learning_rate": 0.00024305014041728172, |
| "loss": 0.105, |
| "step": 79550 |
| }, |
| { |
| "epoch": 1.6142770229162442, |
| "grad_norm": 0.4296875, |
| "learning_rate": 0.00024297121778003237, |
| "loss": 0.1134, |
| "step": 79600 |
| }, |
| { |
| "epoch": 1.6152910160210912, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.00024289225332770737, |
| "loss": 0.112, |
| "step": 79650 |
| }, |
| { |
| "epoch": 1.6163050091259379, |
| "grad_norm": 0.451171875, |
| "learning_rate": 0.00024281324709582195, |
| "loss": 0.1129, |
| "step": 79700 |
| }, |
| { |
| "epoch": 1.6173190022307848, |
| "grad_norm": 0.400390625, |
| "learning_rate": 0.00024273419911991035, |
| "loss": 0.1032, |
| "step": 79750 |
| }, |
| { |
| "epoch": 1.6183329953356318, |
| "grad_norm": 0.5703125, |
| "learning_rate": 0.00024265510943552543, |
| "loss": 0.1057, |
| "step": 79800 |
| }, |
| { |
| "epoch": 1.6193469884404785, |
| "grad_norm": 0.462890625, |
| "learning_rate": 0.00024257597807823887, |
| "loss": 0.1124, |
| "step": 79850 |
| }, |
| { |
| "epoch": 1.6203609815453255, |
| "grad_norm": 0.482421875, |
| "learning_rate": 0.00024249680508364111, |
| "loss": 0.1129, |
| "step": 79900 |
| }, |
| { |
| "epoch": 1.6213749746501724, |
| "grad_norm": 0.419921875, |
| "learning_rate": 0.00024241759048734124, |
| "loss": 0.1016, |
| "step": 79950 |
| }, |
| { |
| "epoch": 1.6223889677550192, |
| "grad_norm": 0.486328125, |
| "learning_rate": 0.0002423383343249671, |
| "loss": 0.1117, |
| "step": 80000 |
| }, |
| { |
| "epoch": 1.6234029608598661, |
| "grad_norm": 0.431640625, |
| "learning_rate": 0.00024225903663216525, |
| "loss": 0.1155, |
| "step": 80050 |
| }, |
| { |
| "epoch": 1.624416953964713, |
| "grad_norm": 0.53515625, |
| "learning_rate": 0.0002421796974446009, |
| "loss": 0.1136, |
| "step": 80100 |
| }, |
| { |
| "epoch": 1.6254309470695598, |
| "grad_norm": 0.45703125, |
| "learning_rate": 0.00024210031679795796, |
| "loss": 0.0956, |
| "step": 80150 |
| }, |
| { |
| "epoch": 1.6264449401744068, |
| "grad_norm": 0.5859375, |
| "learning_rate": 0.00024202089472793897, |
| "loss": 0.1087, |
| "step": 80200 |
| }, |
| { |
| "epoch": 1.6274589332792537, |
| "grad_norm": 0.421875, |
| "learning_rate": 0.00024194143127026504, |
| "loss": 0.1037, |
| "step": 80250 |
| }, |
| { |
| "epoch": 1.6284729263841005, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.00024186192646067598, |
| "loss": 0.1133, |
| "step": 80300 |
| }, |
| { |
| "epoch": 1.6294869194889476, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.00024178238033493015, |
| "loss": 0.1023, |
| "step": 80350 |
| }, |
| { |
| "epoch": 1.6305009125937944, |
| "grad_norm": 0.46875, |
| "learning_rate": 0.00024170279292880456, |
| "loss": 0.1014, |
| "step": 80400 |
| }, |
| { |
| "epoch": 1.631514905698641, |
| "grad_norm": 0.4296875, |
| "learning_rate": 0.00024162316427809463, |
| "loss": 0.1064, |
| "step": 80450 |
| }, |
| { |
| "epoch": 1.6325288988034883, |
| "grad_norm": 0.47265625, |
| "learning_rate": 0.00024154349441861457, |
| "loss": 0.1039, |
| "step": 80500 |
| }, |
| { |
| "epoch": 1.633542891908335, |
| "grad_norm": 0.466796875, |
| "learning_rate": 0.00024146378338619684, |
| "loss": 0.1013, |
| "step": 80550 |
| }, |
| { |
| "epoch": 1.6345568850131817, |
| "grad_norm": 0.4453125, |
| "learning_rate": 0.00024138403121669262, |
| "loss": 0.1025, |
| "step": 80600 |
| }, |
| { |
| "epoch": 1.635570878118029, |
| "grad_norm": 0.49609375, |
| "learning_rate": 0.0002413042379459716, |
| "loss": 0.1095, |
| "step": 80650 |
| }, |
| { |
| "epoch": 1.6365848712228757, |
| "grad_norm": 0.51171875, |
| "learning_rate": 0.00024122440360992186, |
| "loss": 0.0999, |
| "step": 80700 |
| }, |
| { |
| "epoch": 1.6375988643277226, |
| "grad_norm": 0.431640625, |
| "learning_rate": 0.00024114452824444994, |
| "loss": 0.0965, |
| "step": 80750 |
| }, |
| { |
| "epoch": 1.6386128574325696, |
| "grad_norm": 0.44921875, |
| "learning_rate": 0.00024106461188548096, |
| "loss": 0.1026, |
| "step": 80800 |
| }, |
| { |
| "epoch": 1.6396268505374163, |
| "grad_norm": 0.5078125, |
| "learning_rate": 0.0002409846545689583, |
| "loss": 0.1043, |
| "step": 80850 |
| }, |
| { |
| "epoch": 1.6406408436422633, |
| "grad_norm": 0.482421875, |
| "learning_rate": 0.00024090465633084393, |
| "loss": 0.1068, |
| "step": 80900 |
| }, |
| { |
| "epoch": 1.6416548367471102, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.00024082461720711807, |
| "loss": 0.1043, |
| "step": 80950 |
| }, |
| { |
| "epoch": 1.642668829851957, |
| "grad_norm": 0.478515625, |
| "learning_rate": 0.00024074453723377953, |
| "loss": 0.0995, |
| "step": 81000 |
| }, |
| { |
| "epoch": 1.642668829851957, |
| "eval_cer": 0.030604441623705306, |
| "eval_loss": 0.16193068027496338, |
| "eval_runtime": 127.4447, |
| "eval_samples_per_second": 22.033, |
| "eval_steps_per_second": 0.173, |
| "eval_wer": 0.09025661859299886, |
| "step": 81000 |
| }, |
| { |
| "epoch": 1.643682822956804, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.00024066441644684521, |
| "loss": 0.1061, |
| "step": 81050 |
| }, |
| { |
| "epoch": 1.6446968160616509, |
| "grad_norm": 0.484375, |
| "learning_rate": 0.00024058425488235073, |
| "loss": 0.1031, |
| "step": 81100 |
| }, |
| { |
| "epoch": 1.6457108091664976, |
| "grad_norm": 0.53125, |
| "learning_rate": 0.00024050405257634963, |
| "loss": 0.1019, |
| "step": 81150 |
| }, |
| { |
| "epoch": 1.6467248022713445, |
| "grad_norm": 0.4609375, |
| "learning_rate": 0.0002404238095649142, |
| "loss": 0.1021, |
| "step": 81200 |
| }, |
| { |
| "epoch": 1.6477387953761915, |
| "grad_norm": 0.51953125, |
| "learning_rate": 0.00024034352588413467, |
| "loss": 0.1127, |
| "step": 81250 |
| }, |
| { |
| "epoch": 1.6487527884810382, |
| "grad_norm": 0.486328125, |
| "learning_rate": 0.00024026320157011983, |
| "loss": 0.1041, |
| "step": 81300 |
| }, |
| { |
| "epoch": 1.6497667815858852, |
| "grad_norm": 0.42578125, |
| "learning_rate": 0.0002401828366589966, |
| "loss": 0.0965, |
| "step": 81350 |
| }, |
| { |
| "epoch": 1.6507807746907321, |
| "grad_norm": 0.46484375, |
| "learning_rate": 0.00024010243118691016, |
| "loss": 0.1128, |
| "step": 81400 |
| }, |
| { |
| "epoch": 1.6517947677955789, |
| "grad_norm": 0.40234375, |
| "learning_rate": 0.00024002198519002403, |
| "loss": 0.1085, |
| "step": 81450 |
| }, |
| { |
| "epoch": 1.6528087609004258, |
| "grad_norm": 0.482421875, |
| "learning_rate": 0.00023994149870451984, |
| "loss": 0.0986, |
| "step": 81500 |
| }, |
| { |
| "epoch": 1.6538227540052728, |
| "grad_norm": 0.53125, |
| "learning_rate": 0.00023986097176659755, |
| "loss": 0.107, |
| "step": 81550 |
| }, |
| { |
| "epoch": 1.6548367471101195, |
| "grad_norm": 0.51171875, |
| "learning_rate": 0.0002397804044124752, |
| "loss": 0.1022, |
| "step": 81600 |
| }, |
| { |
| "epoch": 1.6558507402149667, |
| "grad_norm": 0.45703125, |
| "learning_rate": 0.00023969979667838907, |
| "loss": 0.1086, |
| "step": 81650 |
| }, |
| { |
| "epoch": 1.6568647333198134, |
| "grad_norm": 0.443359375, |
| "learning_rate": 0.00023961914860059358, |
| "loss": 0.1044, |
| "step": 81700 |
| }, |
| { |
| "epoch": 1.6578787264246602, |
| "grad_norm": 0.46484375, |
| "learning_rate": 0.00023953846021536134, |
| "loss": 0.1145, |
| "step": 81750 |
| }, |
| { |
| "epoch": 1.6588927195295073, |
| "grad_norm": 0.51171875, |
| "learning_rate": 0.00023945773155898306, |
| "loss": 0.1097, |
| "step": 81800 |
| }, |
| { |
| "epoch": 1.659906712634354, |
| "grad_norm": 0.4296875, |
| "learning_rate": 0.00023937696266776746, |
| "loss": 0.1062, |
| "step": 81850 |
| }, |
| { |
| "epoch": 1.6609207057392008, |
| "grad_norm": 0.546875, |
| "learning_rate": 0.0002392961535780416, |
| "loss": 0.1034, |
| "step": 81900 |
| }, |
| { |
| "epoch": 1.661934698844048, |
| "grad_norm": 0.490234375, |
| "learning_rate": 0.00023921530432615036, |
| "loss": 0.1035, |
| "step": 81950 |
| }, |
| { |
| "epoch": 1.6629486919488947, |
| "grad_norm": 0.47265625, |
| "learning_rate": 0.00023913441494845684, |
| "loss": 0.0979, |
| "step": 82000 |
| }, |
| { |
| "epoch": 1.6639626850537417, |
| "grad_norm": 0.50390625, |
| "learning_rate": 0.00023905348548134212, |
| "loss": 0.1251, |
| "step": 82050 |
| }, |
| { |
| "epoch": 1.6649766781585886, |
| "grad_norm": 0.494140625, |
| "learning_rate": 0.00023897251596120535, |
| "loss": 0.1029, |
| "step": 82100 |
| }, |
| { |
| "epoch": 1.6659906712634354, |
| "grad_norm": 0.455078125, |
| "learning_rate": 0.0002388915064244637, |
| "loss": 0.1129, |
| "step": 82150 |
| }, |
| { |
| "epoch": 1.6670046643682823, |
| "grad_norm": 0.5859375, |
| "learning_rate": 0.00023881045690755224, |
| "loss": 0.1063, |
| "step": 82200 |
| }, |
| { |
| "epoch": 1.6680186574731293, |
| "grad_norm": 0.462890625, |
| "learning_rate": 0.00023872936744692418, |
| "loss": 0.1019, |
| "step": 82250 |
| }, |
| { |
| "epoch": 1.669032650577976, |
| "grad_norm": 0.45703125, |
| "learning_rate": 0.0002386482380790506, |
| "loss": 0.1066, |
| "step": 82300 |
| }, |
| { |
| "epoch": 1.670046643682823, |
| "grad_norm": 0.6484375, |
| "learning_rate": 0.0002385670688404205, |
| "loss": 0.1154, |
| "step": 82350 |
| }, |
| { |
| "epoch": 1.67106063678767, |
| "grad_norm": 0.46875, |
| "learning_rate": 0.00023848585976754088, |
| "loss": 0.0969, |
| "step": 82400 |
| }, |
| { |
| "epoch": 1.6720746298925167, |
| "grad_norm": 0.5703125, |
| "learning_rate": 0.00023840461089693664, |
| "loss": 0.1085, |
| "step": 82450 |
| }, |
| { |
| "epoch": 1.6730886229973636, |
| "grad_norm": 0.490234375, |
| "learning_rate": 0.00023832332226515057, |
| "loss": 0.1046, |
| "step": 82500 |
| }, |
| { |
| "epoch": 1.6730886229973636, |
| "eval_cer": 0.035738996082756325, |
| "eval_loss": 0.1605072170495987, |
| "eval_runtime": 143.5413, |
| "eval_samples_per_second": 19.562, |
| "eval_steps_per_second": 0.153, |
| "eval_wer": 0.10792876972036021, |
| "step": 82500 |
| }, |
| { |
| "epoch": 1.6741026161022106, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.0002382419939087433, |
| "loss": 0.1057, |
| "step": 82550 |
| }, |
| { |
| "epoch": 1.6751166092070573, |
| "grad_norm": 0.51171875, |
| "learning_rate": 0.00023816062586429343, |
| "loss": 0.108, |
| "step": 82600 |
| }, |
| { |
| "epoch": 1.6761306023119042, |
| "grad_norm": 0.490234375, |
| "learning_rate": 0.00023807921816839728, |
| "loss": 0.0969, |
| "step": 82650 |
| }, |
| { |
| "epoch": 1.6771445954167512, |
| "grad_norm": 0.490234375, |
| "learning_rate": 0.00023799777085766917, |
| "loss": 0.0997, |
| "step": 82700 |
| }, |
| { |
| "epoch": 1.678158588521598, |
| "grad_norm": 0.5078125, |
| "learning_rate": 0.00023791628396874106, |
| "loss": 0.1044, |
| "step": 82750 |
| }, |
| { |
| "epoch": 1.679172581626445, |
| "grad_norm": 0.546875, |
| "learning_rate": 0.00023783475753826283, |
| "loss": 0.1015, |
| "step": 82800 |
| }, |
| { |
| "epoch": 1.6801865747312918, |
| "grad_norm": 0.462890625, |
| "learning_rate": 0.00023775319160290208, |
| "loss": 0.1015, |
| "step": 82850 |
| }, |
| { |
| "epoch": 1.6812005678361386, |
| "grad_norm": 0.5, |
| "learning_rate": 0.0002376715861993442, |
| "loss": 0.106, |
| "step": 82900 |
| }, |
| { |
| "epoch": 1.6822145609409858, |
| "grad_norm": 0.4609375, |
| "learning_rate": 0.0002375899413642924, |
| "loss": 0.0996, |
| "step": 82950 |
| }, |
| { |
| "epoch": 1.6832285540458325, |
| "grad_norm": 0.51953125, |
| "learning_rate": 0.00023750825713446752, |
| "loss": 0.1062, |
| "step": 83000 |
| }, |
| { |
| "epoch": 1.6842425471506792, |
| "grad_norm": 0.431640625, |
| "learning_rate": 0.0002374265335466081, |
| "loss": 0.1042, |
| "step": 83050 |
| }, |
| { |
| "epoch": 1.6852565402555264, |
| "grad_norm": 0.478515625, |
| "learning_rate": 0.00023734477063747054, |
| "loss": 0.1, |
| "step": 83100 |
| }, |
| { |
| "epoch": 1.6862705333603731, |
| "grad_norm": 0.50390625, |
| "learning_rate": 0.00023726296844382876, |
| "loss": 0.1002, |
| "step": 83150 |
| }, |
| { |
| "epoch": 1.68728452646522, |
| "grad_norm": 0.478515625, |
| "learning_rate": 0.00023718112700247445, |
| "loss": 0.107, |
| "step": 83200 |
| }, |
| { |
| "epoch": 1.688298519570067, |
| "grad_norm": 0.462890625, |
| "learning_rate": 0.00023709924635021687, |
| "loss": 0.0996, |
| "step": 83250 |
| }, |
| { |
| "epoch": 1.6893125126749138, |
| "grad_norm": 0.453125, |
| "learning_rate": 0.000237017326523883, |
| "loss": 0.1053, |
| "step": 83300 |
| }, |
| { |
| "epoch": 1.6903265057797607, |
| "grad_norm": 0.447265625, |
| "learning_rate": 0.0002369353675603174, |
| "loss": 0.1013, |
| "step": 83350 |
| }, |
| { |
| "epoch": 1.6913404988846077, |
| "grad_norm": 0.4140625, |
| "learning_rate": 0.00023685336949638225, |
| "loss": 0.1002, |
| "step": 83400 |
| }, |
| { |
| "epoch": 1.6923544919894544, |
| "grad_norm": 0.48046875, |
| "learning_rate": 0.00023677133236895724, |
| "loss": 0.1074, |
| "step": 83450 |
| }, |
| { |
| "epoch": 1.6933684850943014, |
| "grad_norm": 0.494140625, |
| "learning_rate": 0.00023668925621493975, |
| "loss": 0.0999, |
| "step": 83500 |
| }, |
| { |
| "epoch": 1.6943824781991483, |
| "grad_norm": 0.490234375, |
| "learning_rate": 0.00023660714107124464, |
| "loss": 0.099, |
| "step": 83550 |
| }, |
| { |
| "epoch": 1.695396471303995, |
| "grad_norm": 0.447265625, |
| "learning_rate": 0.0002365249869748043, |
| "loss": 0.1107, |
| "step": 83600 |
| }, |
| { |
| "epoch": 1.696410464408842, |
| "grad_norm": 0.4453125, |
| "learning_rate": 0.00023644279396256863, |
| "loss": 0.1086, |
| "step": 83650 |
| }, |
| { |
| "epoch": 1.697424457513689, |
| "grad_norm": 0.462890625, |
| "learning_rate": 0.0002363605620715051, |
| "loss": 0.1003, |
| "step": 83700 |
| }, |
| { |
| "epoch": 1.6984384506185357, |
| "grad_norm": 0.5, |
| "learning_rate": 0.00023627829133859865, |
| "loss": 0.0973, |
| "step": 83750 |
| }, |
| { |
| "epoch": 1.6994524437233827, |
| "grad_norm": 0.421875, |
| "learning_rate": 0.0002361959818008516, |
| "loss": 0.1023, |
| "step": 83800 |
| }, |
| { |
| "epoch": 1.7004664368282296, |
| "grad_norm": 0.431640625, |
| "learning_rate": 0.00023611363349528388, |
| "loss": 0.0998, |
| "step": 83850 |
| }, |
| { |
| "epoch": 1.7014804299330764, |
| "grad_norm": 0.54296875, |
| "learning_rate": 0.0002360312464589327, |
| "loss": 0.104, |
| "step": 83900 |
| }, |
| { |
| "epoch": 1.7024944230379233, |
| "grad_norm": 0.48046875, |
| "learning_rate": 0.00023594882072885273, |
| "loss": 0.1024, |
| "step": 83950 |
| }, |
| { |
| "epoch": 1.7035084161427703, |
| "grad_norm": 0.44140625, |
| "learning_rate": 0.00023586635634211618, |
| "loss": 0.1024, |
| "step": 84000 |
| }, |
| { |
| "epoch": 1.7035084161427703, |
| "eval_cer": 0.029870933843840874, |
| "eval_loss": 0.16048672795295715, |
| "eval_runtime": 127.4658, |
| "eval_samples_per_second": 22.029, |
| "eval_steps_per_second": 0.173, |
| "eval_wer": 0.08886857607150112, |
| "step": 84000 |
| }, |
| { |
| "epoch": 1.704522409247617, |
| "grad_norm": 0.466796875, |
| "learning_rate": 0.0002357838533358124, |
| "loss": 0.0991, |
| "step": 84050 |
| }, |
| { |
| "epoch": 1.705536402352464, |
| "grad_norm": 0.44921875, |
| "learning_rate": 0.0002357013117470483, |
| "loss": 0.1014, |
| "step": 84100 |
| }, |
| { |
| "epoch": 1.706550395457311, |
| "grad_norm": 0.6875, |
| "learning_rate": 0.00023561873161294811, |
| "loss": 0.1084, |
| "step": 84150 |
| }, |
| { |
| "epoch": 1.7075643885621576, |
| "grad_norm": 0.44140625, |
| "learning_rate": 0.00023553611297065337, |
| "loss": 0.1031, |
| "step": 84200 |
| }, |
| { |
| "epoch": 1.7085783816670048, |
| "grad_norm": 0.482421875, |
| "learning_rate": 0.00023545345585732292, |
| "loss": 0.1072, |
| "step": 84250 |
| }, |
| { |
| "epoch": 1.7095923747718516, |
| "grad_norm": 0.49609375, |
| "learning_rate": 0.00023537076031013296, |
| "loss": 0.1091, |
| "step": 84300 |
| }, |
| { |
| "epoch": 1.7106063678766983, |
| "grad_norm": 0.38671875, |
| "learning_rate": 0.00023528802636627687, |
| "loss": 0.1064, |
| "step": 84350 |
| }, |
| { |
| "epoch": 1.7116203609815455, |
| "grad_norm": 0.466796875, |
| "learning_rate": 0.00023520525406296549, |
| "loss": 0.1147, |
| "step": 84400 |
| }, |
| { |
| "epoch": 1.7126343540863922, |
| "grad_norm": 0.466796875, |
| "learning_rate": 0.00023512244343742666, |
| "loss": 0.1, |
| "step": 84450 |
| }, |
| { |
| "epoch": 1.7136483471912392, |
| "grad_norm": 0.443359375, |
| "learning_rate": 0.00023503959452690566, |
| "loss": 0.0954, |
| "step": 84500 |
| }, |
| { |
| "epoch": 1.714662340296086, |
| "grad_norm": 0.49609375, |
| "learning_rate": 0.00023495670736866498, |
| "loss": 0.1194, |
| "step": 84550 |
| }, |
| { |
| "epoch": 1.7156763334009328, |
| "grad_norm": 0.427734375, |
| "learning_rate": 0.00023487378199998414, |
| "loss": 0.0977, |
| "step": 84600 |
| }, |
| { |
| "epoch": 1.7166903265057798, |
| "grad_norm": 0.4375, |
| "learning_rate": 0.00023479081845816002, |
| "loss": 0.0976, |
| "step": 84650 |
| }, |
| { |
| "epoch": 1.7177043196106268, |
| "grad_norm": 0.48046875, |
| "learning_rate": 0.00023470781678050658, |
| "loss": 0.1088, |
| "step": 84700 |
| }, |
| { |
| "epoch": 1.7187183127154735, |
| "grad_norm": 0.451171875, |
| "learning_rate": 0.000234624777004355, |
| "loss": 0.1156, |
| "step": 84750 |
| }, |
| { |
| "epoch": 1.7197323058203204, |
| "grad_norm": 0.46484375, |
| "learning_rate": 0.00023454169916705347, |
| "loss": 0.1037, |
| "step": 84800 |
| }, |
| { |
| "epoch": 1.7207462989251674, |
| "grad_norm": 0.466796875, |
| "learning_rate": 0.0002344585833059675, |
| "loss": 0.108, |
| "step": 84850 |
| }, |
| { |
| "epoch": 1.7217602920300141, |
| "grad_norm": 0.455078125, |
| "learning_rate": 0.00023437542945847948, |
| "loss": 0.1128, |
| "step": 84900 |
| }, |
| { |
| "epoch": 1.722774285134861, |
| "grad_norm": 0.52734375, |
| "learning_rate": 0.00023429223766198906, |
| "loss": 0.1165, |
| "step": 84950 |
| }, |
| { |
| "epoch": 1.723788278239708, |
| "grad_norm": 0.474609375, |
| "learning_rate": 0.00023420900795391282, |
| "loss": 0.1013, |
| "step": 85000 |
| }, |
| { |
| "epoch": 1.7248022713445548, |
| "grad_norm": 0.412109375, |
| "learning_rate": 0.00023412574037168457, |
| "loss": 0.0963, |
| "step": 85050 |
| }, |
| { |
| "epoch": 1.7258162644494017, |
| "grad_norm": 0.51171875, |
| "learning_rate": 0.00023404243495275486, |
| "loss": 0.1073, |
| "step": 85100 |
| }, |
| { |
| "epoch": 1.7268302575542487, |
| "grad_norm": 0.462890625, |
| "learning_rate": 0.00023395909173459163, |
| "loss": 0.1034, |
| "step": 85150 |
| }, |
| { |
| "epoch": 1.7278442506590954, |
| "grad_norm": 0.44140625, |
| "learning_rate": 0.00023387571075467954, |
| "loss": 0.1085, |
| "step": 85200 |
| }, |
| { |
| "epoch": 1.7288582437639424, |
| "grad_norm": 0.498046875, |
| "learning_rate": 0.00023379229205052035, |
| "loss": 0.0986, |
| "step": 85250 |
| }, |
| { |
| "epoch": 1.7298722368687893, |
| "grad_norm": 0.5078125, |
| "learning_rate": 0.00023370883565963275, |
| "loss": 0.1028, |
| "step": 85300 |
| }, |
| { |
| "epoch": 1.730886229973636, |
| "grad_norm": 0.380859375, |
| "learning_rate": 0.00023362534161955236, |
| "loss": 0.1011, |
| "step": 85350 |
| }, |
| { |
| "epoch": 1.731900223078483, |
| "grad_norm": 0.435546875, |
| "learning_rate": 0.00023354180996783185, |
| "loss": 0.1068, |
| "step": 85400 |
| }, |
| { |
| "epoch": 1.73291421618333, |
| "grad_norm": 0.5, |
| "learning_rate": 0.00023345824074204061, |
| "loss": 0.1006, |
| "step": 85450 |
| }, |
| { |
| "epoch": 1.7339282092881767, |
| "grad_norm": 0.53515625, |
| "learning_rate": 0.00023337463397976518, |
| "loss": 0.1035, |
| "step": 85500 |
| }, |
| { |
| "epoch": 1.7339282092881767, |
| "eval_cer": 0.030047808060262086, |
| "eval_loss": 0.16027864813804626, |
| "eval_runtime": 127.051, |
| "eval_samples_per_second": 22.101, |
| "eval_steps_per_second": 0.173, |
| "eval_wer": 0.09066287494075428, |
| "step": 85500 |
| }, |
| { |
| "epoch": 1.7349422023930239, |
| "grad_norm": 0.5, |
| "learning_rate": 0.00023329098971860867, |
| "loss": 0.0958, |
| "step": 85550 |
| }, |
| { |
| "epoch": 1.7359561954978706, |
| "grad_norm": 0.4453125, |
| "learning_rate": 0.0002332073079961914, |
| "loss": 0.1067, |
| "step": 85600 |
| }, |
| { |
| "epoch": 1.7369701886027173, |
| "grad_norm": 0.54296875, |
| "learning_rate": 0.00023312358885015024, |
| "loss": 0.1067, |
| "step": 85650 |
| }, |
| { |
| "epoch": 1.7379841817075645, |
| "grad_norm": 0.388671875, |
| "learning_rate": 0.0002330398323181391, |
| "loss": 0.1022, |
| "step": 85700 |
| }, |
| { |
| "epoch": 1.7389981748124113, |
| "grad_norm": 0.4375, |
| "learning_rate": 0.0002329560384378286, |
| "loss": 0.102, |
| "step": 85750 |
| }, |
| { |
| "epoch": 1.7400121679172582, |
| "grad_norm": 0.52734375, |
| "learning_rate": 0.00023287220724690615, |
| "loss": 0.0998, |
| "step": 85800 |
| }, |
| { |
| "epoch": 1.7410261610221052, |
| "grad_norm": 0.49609375, |
| "learning_rate": 0.00023278833878307602, |
| "loss": 0.098, |
| "step": 85850 |
| }, |
| { |
| "epoch": 1.742040154126952, |
| "grad_norm": 0.44921875, |
| "learning_rate": 0.00023270443308405917, |
| "loss": 0.1163, |
| "step": 85900 |
| }, |
| { |
| "epoch": 1.7430541472317989, |
| "grad_norm": 0.49609375, |
| "learning_rate": 0.00023262049018759335, |
| "loss": 0.0991, |
| "step": 85950 |
| }, |
| { |
| "epoch": 1.7440681403366458, |
| "grad_norm": 0.53125, |
| "learning_rate": 0.00023253651013143304, |
| "loss": 0.1046, |
| "step": 86000 |
| }, |
| { |
| "epoch": 1.7450821334414925, |
| "grad_norm": 0.46875, |
| "learning_rate": 0.0002324524929533494, |
| "loss": 0.1031, |
| "step": 86050 |
| }, |
| { |
| "epoch": 1.7460961265463395, |
| "grad_norm": 0.6484375, |
| "learning_rate": 0.00023236843869113032, |
| "loss": 0.1051, |
| "step": 86100 |
| }, |
| { |
| "epoch": 1.7471101196511865, |
| "grad_norm": 0.44921875, |
| "learning_rate": 0.0002322843473825803, |
| "loss": 0.1011, |
| "step": 86150 |
| }, |
| { |
| "epoch": 1.7481241127560332, |
| "grad_norm": 0.447265625, |
| "learning_rate": 0.00023220021906552059, |
| "loss": 0.0971, |
| "step": 86200 |
| }, |
| { |
| "epoch": 1.7491381058608801, |
| "grad_norm": 0.48046875, |
| "learning_rate": 0.00023211605377778906, |
| "loss": 0.1072, |
| "step": 86250 |
| }, |
| { |
| "epoch": 1.750152098965727, |
| "grad_norm": 0.38671875, |
| "learning_rate": 0.0002320318515572402, |
| "loss": 0.1097, |
| "step": 86300 |
| }, |
| { |
| "epoch": 1.7511660920705738, |
| "grad_norm": 0.46484375, |
| "learning_rate": 0.0002319476124417451, |
| "loss": 0.1016, |
| "step": 86350 |
| }, |
| { |
| "epoch": 1.7521800851754208, |
| "grad_norm": 0.4765625, |
| "learning_rate": 0.00023186333646919142, |
| "loss": 0.1063, |
| "step": 86400 |
| }, |
| { |
| "epoch": 1.7531940782802677, |
| "grad_norm": 0.458984375, |
| "learning_rate": 0.00023177902367748347, |
| "loss": 0.1061, |
| "step": 86450 |
| }, |
| { |
| "epoch": 1.7542080713851145, |
| "grad_norm": 0.474609375, |
| "learning_rate": 0.00023169467410454213, |
| "loss": 0.1031, |
| "step": 86500 |
| }, |
| { |
| "epoch": 1.7552220644899614, |
| "grad_norm": 0.462890625, |
| "learning_rate": 0.00023161028778830465, |
| "loss": 0.1073, |
| "step": 86550 |
| }, |
| { |
| "epoch": 1.7562360575948084, |
| "grad_norm": 0.48828125, |
| "learning_rate": 0.00023152586476672506, |
| "loss": 0.1099, |
| "step": 86600 |
| }, |
| { |
| "epoch": 1.7572500506996551, |
| "grad_norm": 0.55859375, |
| "learning_rate": 0.00023144140507777363, |
| "loss": 0.1016, |
| "step": 86650 |
| }, |
| { |
| "epoch": 1.758264043804502, |
| "grad_norm": 0.416015625, |
| "learning_rate": 0.00023135690875943736, |
| "loss": 0.0989, |
| "step": 86700 |
| }, |
| { |
| "epoch": 1.759278036909349, |
| "grad_norm": 0.4609375, |
| "learning_rate": 0.00023127237584971962, |
| "loss": 0.0986, |
| "step": 86750 |
| }, |
| { |
| "epoch": 1.7602920300141958, |
| "grad_norm": 0.50390625, |
| "learning_rate": 0.0002311878063866402, |
| "loss": 0.1043, |
| "step": 86800 |
| }, |
| { |
| "epoch": 1.761306023119043, |
| "grad_norm": 0.4609375, |
| "learning_rate": 0.00023110320040823536, |
| "loss": 0.1043, |
| "step": 86850 |
| }, |
| { |
| "epoch": 1.7623200162238897, |
| "grad_norm": 0.56640625, |
| "learning_rate": 0.00023101855795255787, |
| "loss": 0.0959, |
| "step": 86900 |
| }, |
| { |
| "epoch": 1.7633340093287364, |
| "grad_norm": 0.423828125, |
| "learning_rate": 0.00023093387905767675, |
| "loss": 0.0968, |
| "step": 86950 |
| }, |
| { |
| "epoch": 1.7643480024335836, |
| "grad_norm": 0.447265625, |
| "learning_rate": 0.00023084916376167754, |
| "loss": 0.102, |
| "step": 87000 |
| }, |
| { |
| "epoch": 1.7643480024335836, |
| "eval_cer": 0.02896575403039115, |
| "eval_loss": 0.15883292257785797, |
| "eval_runtime": 128.449, |
| "eval_samples_per_second": 21.861, |
| "eval_steps_per_second": 0.171, |
| "eval_wer": 0.08903784954973255, |
| "step": 87000 |
| }, |
| { |
| "epoch": 1.7653619955384303, |
| "grad_norm": 0.51953125, |
| "learning_rate": 0.0002307644121026621, |
| "loss": 0.0996, |
| "step": 87050 |
| }, |
| { |
| "epoch": 1.7663759886432773, |
| "grad_norm": 0.458984375, |
| "learning_rate": 0.00023067962411874862, |
| "loss": 0.1088, |
| "step": 87100 |
| }, |
| { |
| "epoch": 1.7673899817481242, |
| "grad_norm": 0.486328125, |
| "learning_rate": 0.00023059479984807165, |
| "loss": 0.097, |
| "step": 87150 |
| }, |
| { |
| "epoch": 1.768403974852971, |
| "grad_norm": 0.421875, |
| "learning_rate": 0.00023050993932878213, |
| "loss": 0.1053, |
| "step": 87200 |
| }, |
| { |
| "epoch": 1.769417967957818, |
| "grad_norm": 0.470703125, |
| "learning_rate": 0.00023042504259904718, |
| "loss": 0.0989, |
| "step": 87250 |
| }, |
| { |
| "epoch": 1.7704319610626649, |
| "grad_norm": 0.5, |
| "learning_rate": 0.00023034010969705024, |
| "loss": 0.1057, |
| "step": 87300 |
| }, |
| { |
| "epoch": 1.7714459541675116, |
| "grad_norm": 0.435546875, |
| "learning_rate": 0.00023025514066099108, |
| "loss": 0.0951, |
| "step": 87350 |
| }, |
| { |
| "epoch": 1.7724599472723586, |
| "grad_norm": 0.482421875, |
| "learning_rate": 0.0002301701355290857, |
| "loss": 0.0984, |
| "step": 87400 |
| }, |
| { |
| "epoch": 1.7734739403772055, |
| "grad_norm": 0.466796875, |
| "learning_rate": 0.00023008509433956631, |
| "loss": 0.1055, |
| "step": 87450 |
| }, |
| { |
| "epoch": 1.7744879334820522, |
| "grad_norm": 0.412109375, |
| "learning_rate": 0.00023000001713068135, |
| "loss": 0.1107, |
| "step": 87500 |
| }, |
| { |
| "epoch": 1.7755019265868992, |
| "grad_norm": 0.515625, |
| "learning_rate": 0.00022991490394069542, |
| "loss": 0.0969, |
| "step": 87550 |
| }, |
| { |
| "epoch": 1.7765159196917462, |
| "grad_norm": 0.45703125, |
| "learning_rate": 0.00022982975480788937, |
| "loss": 0.1072, |
| "step": 87600 |
| }, |
| { |
| "epoch": 1.777529912796593, |
| "grad_norm": 0.47265625, |
| "learning_rate": 0.00022974456977056026, |
| "loss": 0.1, |
| "step": 87650 |
| }, |
| { |
| "epoch": 1.7785439059014398, |
| "grad_norm": 0.470703125, |
| "learning_rate": 0.00022965934886702108, |
| "loss": 0.1103, |
| "step": 87700 |
| }, |
| { |
| "epoch": 1.7795578990062868, |
| "grad_norm": 0.455078125, |
| "learning_rate": 0.00022957409213560112, |
| "loss": 0.0997, |
| "step": 87750 |
| }, |
| { |
| "epoch": 1.7805718921111335, |
| "grad_norm": 0.486328125, |
| "learning_rate": 0.0002294887996146459, |
| "loss": 0.0987, |
| "step": 87800 |
| }, |
| { |
| "epoch": 1.7815858852159805, |
| "grad_norm": 0.5, |
| "learning_rate": 0.00022940347134251673, |
| "loss": 0.0994, |
| "step": 87850 |
| }, |
| { |
| "epoch": 1.7825998783208274, |
| "grad_norm": 0.466796875, |
| "learning_rate": 0.0002293181073575913, |
| "loss": 0.1015, |
| "step": 87900 |
| }, |
| { |
| "epoch": 1.7836138714256742, |
| "grad_norm": 0.470703125, |
| "learning_rate": 0.00022923270769826314, |
| "loss": 0.1024, |
| "step": 87950 |
| }, |
| { |
| "epoch": 1.7846278645305214, |
| "grad_norm": 0.53125, |
| "learning_rate": 0.0002291472724029419, |
| "loss": 0.1145, |
| "step": 88000 |
| }, |
| { |
| "epoch": 1.785641857635368, |
| "grad_norm": 0.5703125, |
| "learning_rate": 0.0002290618015100534, |
| "loss": 0.1031, |
| "step": 88050 |
| }, |
| { |
| "epoch": 1.7866558507402148, |
| "grad_norm": 0.5234375, |
| "learning_rate": 0.00022897629505803918, |
| "loss": 0.1004, |
| "step": 88100 |
| }, |
| { |
| "epoch": 1.787669843845062, |
| "grad_norm": 0.4765625, |
| "learning_rate": 0.00022889075308535703, |
| "loss": 0.099, |
| "step": 88150 |
| }, |
| { |
| "epoch": 1.7886838369499087, |
| "grad_norm": 0.482421875, |
| "learning_rate": 0.00022880517563048062, |
| "loss": 0.0972, |
| "step": 88200 |
| }, |
| { |
| "epoch": 1.7896978300547555, |
| "grad_norm": 0.474609375, |
| "learning_rate": 0.00022871956273189957, |
| "loss": 0.0992, |
| "step": 88250 |
| }, |
| { |
| "epoch": 1.7907118231596026, |
| "grad_norm": 0.427734375, |
| "learning_rate": 0.0002286339144281194, |
| "loss": 0.1057, |
| "step": 88300 |
| }, |
| { |
| "epoch": 1.7917258162644494, |
| "grad_norm": 0.458984375, |
| "learning_rate": 0.0002285482307576617, |
| "loss": 0.1115, |
| "step": 88350 |
| }, |
| { |
| "epoch": 1.7927398093692963, |
| "grad_norm": 0.47265625, |
| "learning_rate": 0.00022846251175906378, |
| "loss": 0.0936, |
| "step": 88400 |
| }, |
| { |
| "epoch": 1.7937538024741433, |
| "grad_norm": 0.482421875, |
| "learning_rate": 0.00022837675747087896, |
| "loss": 0.096, |
| "step": 88450 |
| }, |
| { |
| "epoch": 1.79476779557899, |
| "grad_norm": 0.54296875, |
| "learning_rate": 0.00022829096793167643, |
| "loss": 0.1007, |
| "step": 88500 |
| }, |
| { |
| "epoch": 1.79476779557899, |
| "eval_cer": 0.028596399049040978, |
| "eval_loss": 0.1581563800573349, |
| "eval_runtime": 132.2778, |
| "eval_samples_per_second": 21.228, |
| "eval_steps_per_second": 0.166, |
| "eval_wer": 0.08693885841966281, |
| "step": 88500 |
| }, |
| { |
| "epoch": 1.795781788683837, |
| "grad_norm": 0.470703125, |
| "learning_rate": 0.00022820514318004122, |
| "loss": 0.1084, |
| "step": 88550 |
| }, |
| { |
| "epoch": 1.796795781788684, |
| "grad_norm": 0.48046875, |
| "learning_rate": 0.00022811928325457408, |
| "loss": 0.1006, |
| "step": 88600 |
| }, |
| { |
| "epoch": 1.7978097748935307, |
| "grad_norm": 0.515625, |
| "learning_rate": 0.00022803338819389184, |
| "loss": 0.1016, |
| "step": 88650 |
| }, |
| { |
| "epoch": 1.7988237679983776, |
| "grad_norm": 0.53515625, |
| "learning_rate": 0.00022794745803662687, |
| "loss": 0.0935, |
| "step": 88700 |
| }, |
| { |
| "epoch": 1.7998377611032246, |
| "grad_norm": 0.4609375, |
| "learning_rate": 0.00022786149282142743, |
| "loss": 0.1056, |
| "step": 88750 |
| }, |
| { |
| "epoch": 1.8008517542080713, |
| "grad_norm": 0.478515625, |
| "learning_rate": 0.0002277754925869576, |
| "loss": 0.1, |
| "step": 88800 |
| }, |
| { |
| "epoch": 1.8018657473129183, |
| "grad_norm": 0.53125, |
| "learning_rate": 0.00022768945737189716, |
| "loss": 0.1013, |
| "step": 88850 |
| }, |
| { |
| "epoch": 1.8028797404177652, |
| "grad_norm": 0.494140625, |
| "learning_rate": 0.00022760338721494158, |
| "loss": 0.1087, |
| "step": 88900 |
| }, |
| { |
| "epoch": 1.803893733522612, |
| "grad_norm": 0.5078125, |
| "learning_rate": 0.00022751728215480215, |
| "loss": 0.1056, |
| "step": 88950 |
| }, |
| { |
| "epoch": 1.804907726627459, |
| "grad_norm": 0.462890625, |
| "learning_rate": 0.00022743114223020572, |
| "loss": 0.1004, |
| "step": 89000 |
| }, |
| { |
| "epoch": 1.8059217197323059, |
| "grad_norm": 0.455078125, |
| "learning_rate": 0.00022734496747989496, |
| "loss": 0.1072, |
| "step": 89050 |
| }, |
| { |
| "epoch": 1.8069357128371526, |
| "grad_norm": 0.384765625, |
| "learning_rate": 0.0002272587579426281, |
| "loss": 0.1114, |
| "step": 89100 |
| }, |
| { |
| "epoch": 1.8079497059419996, |
| "grad_norm": 0.490234375, |
| "learning_rate": 0.0002271725136571791, |
| "loss": 0.1071, |
| "step": 89150 |
| }, |
| { |
| "epoch": 1.8089636990468465, |
| "grad_norm": 0.64453125, |
| "learning_rate": 0.00022708623466233748, |
| "loss": 0.1177, |
| "step": 89200 |
| }, |
| { |
| "epoch": 1.8099776921516932, |
| "grad_norm": 0.46484375, |
| "learning_rate": 0.00022699992099690843, |
| "loss": 0.1003, |
| "step": 89250 |
| }, |
| { |
| "epoch": 1.8109916852565404, |
| "grad_norm": 0.73046875, |
| "learning_rate": 0.0002269135726997126, |
| "loss": 0.0984, |
| "step": 89300 |
| }, |
| { |
| "epoch": 1.8120056783613872, |
| "grad_norm": 0.5390625, |
| "learning_rate": 0.00022682718980958645, |
| "loss": 0.0969, |
| "step": 89350 |
| }, |
| { |
| "epoch": 1.8130196714662339, |
| "grad_norm": 0.462890625, |
| "learning_rate": 0.00022674077236538182, |
| "loss": 0.1103, |
| "step": 89400 |
| }, |
| { |
| "epoch": 1.814033664571081, |
| "grad_norm": 0.451171875, |
| "learning_rate": 0.00022665432040596604, |
| "loss": 0.0958, |
| "step": 89450 |
| }, |
| { |
| "epoch": 1.8150476576759278, |
| "grad_norm": 0.50390625, |
| "learning_rate": 0.00022656783397022223, |
| "loss": 0.0938, |
| "step": 89500 |
| }, |
| { |
| "epoch": 1.8160616507807745, |
| "grad_norm": 0.546875, |
| "learning_rate": 0.00022648131309704872, |
| "loss": 0.0957, |
| "step": 89550 |
| }, |
| { |
| "epoch": 1.8170756438856217, |
| "grad_norm": 0.4609375, |
| "learning_rate": 0.00022639475782535945, |
| "loss": 0.0991, |
| "step": 89600 |
| }, |
| { |
| "epoch": 1.8180896369904684, |
| "grad_norm": 0.458984375, |
| "learning_rate": 0.00022630816819408395, |
| "loss": 0.0988, |
| "step": 89650 |
| }, |
| { |
| "epoch": 1.8191036300953154, |
| "grad_norm": 0.466796875, |
| "learning_rate": 0.000226221544242167, |
| "loss": 0.1055, |
| "step": 89700 |
| }, |
| { |
| "epoch": 1.8201176232001623, |
| "grad_norm": 0.546875, |
| "learning_rate": 0.00022613488600856885, |
| "loss": 0.1056, |
| "step": 89750 |
| }, |
| { |
| "epoch": 1.821131616305009, |
| "grad_norm": 0.484375, |
| "learning_rate": 0.0002260481935322654, |
| "loss": 0.0986, |
| "step": 89800 |
| }, |
| { |
| "epoch": 1.822145609409856, |
| "grad_norm": 0.49609375, |
| "learning_rate": 0.00022596146685224762, |
| "loss": 0.0908, |
| "step": 89850 |
| }, |
| { |
| "epoch": 1.823159602514703, |
| "grad_norm": 0.46484375, |
| "learning_rate": 0.0002258747060075221, |
| "loss": 0.0998, |
| "step": 89900 |
| }, |
| { |
| "epoch": 1.8241735956195497, |
| "grad_norm": 0.458984375, |
| "learning_rate": 0.00022578791103711069, |
| "loss": 0.0994, |
| "step": 89950 |
| }, |
| { |
| "epoch": 1.8251875887243967, |
| "grad_norm": 0.47265625, |
| "learning_rate": 0.0002257010819800506, |
| "loss": 0.1117, |
| "step": 90000 |
| }, |
| { |
| "epoch": 1.8251875887243967, |
| "eval_cer": 0.030026999328918415, |
| "eval_loss": 0.15934033691883087, |
| "eval_runtime": 129.703, |
| "eval_samples_per_second": 21.649, |
| "eval_steps_per_second": 0.17, |
| "eval_wer": 0.09049360146252285, |
| "step": 90000 |
| } |
| ], |
| "logging_steps": 50, |
| "max_steps": 246550, |
| "num_input_tokens_seen": 0, |
| "num_train_epochs": 5, |
| "save_steps": 10000, |
| "stateful_callbacks": { |
| "TrainerControl": { |
| "args": { |
| "should_epoch_stop": false, |
| "should_evaluate": false, |
| "should_log": false, |
| "should_save": true, |
| "should_training_stop": false |
| }, |
| "attributes": {} |
| } |
| }, |
| "total_flos": 0.0, |
| "train_batch_size": 16, |
| "trial_name": null, |
| "trial_params": null |
| } |
|
|
