diff --git "a/trainer_state.json" "b/trainer_state.json" new file mode 100644--- /dev/null +++ "b/trainer_state.json" @@ -0,0 +1,7338 @@ +{ + "best_metric": 20.1760839127649, + "best_model_checkpoint": "./hamsa-small-finetuned-qasr/checkpoint-20000", + "epoch": 0.3007649199490722, + "eval_steps": 2500, + "global_step": 30001, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.0, + "learning_rate": 4.4e-07, + "loss": 0.8113, + "step": 25 + }, + { + "epoch": 0.0, + "learning_rate": 9.400000000000001e-07, + "loss": 0.761, + "step": 50 + }, + { + "epoch": 0.0, + "learning_rate": 1.44e-06, + "loss": 0.6789, + "step": 75 + }, + { + "epoch": 0.0, + "learning_rate": 1.94e-06, + "loss": 0.5567, + "step": 100 + }, + { + "epoch": 0.0, + "learning_rate": 2.4400000000000004e-06, + "loss": 0.57, + "step": 125 + }, + { + "epoch": 0.0, + "learning_rate": 2.9400000000000002e-06, + "loss": 0.5093, + "step": 150 + }, + { + "epoch": 0.0, + "learning_rate": 3.44e-06, + "loss": 0.5382, + "step": 175 + }, + { + "epoch": 0.0, + "learning_rate": 3.94e-06, + "loss": 0.4284, + "step": 200 + }, + { + "epoch": 0.0, + "learning_rate": 4.440000000000001e-06, + "loss": 0.4742, + "step": 225 + }, + { + "epoch": 0.0, + "learning_rate": 4.94e-06, + "loss": 0.4024, + "step": 250 + }, + { + "epoch": 0.0, + "learning_rate": 5.4400000000000004e-06, + "loss": 0.4384, + "step": 275 + }, + { + "epoch": 0.0, + "learning_rate": 5.94e-06, + "loss": 0.3865, + "step": 300 + }, + { + "epoch": 0.0, + "learning_rate": 6.440000000000001e-06, + "loss": 0.372, + "step": 325 + }, + { + "epoch": 0.0, + "learning_rate": 6.9400000000000005e-06, + "loss": 0.3847, + "step": 350 + }, + { + "epoch": 0.0, + "learning_rate": 7.440000000000001e-06, + "loss": 0.4258, + "step": 375 + }, + { + "epoch": 0.0, + "learning_rate": 7.94e-06, + "loss": 0.3677, + "step": 400 + }, + { + "epoch": 0.0, + "learning_rate": 8.44e-06, + "loss": 0.433, + "step": 425 + }, + { + "epoch": 0.0, + "learning_rate": 8.94e-06, + "loss": 0.3937, + "step": 450 + }, + { + "epoch": 0.0, + "learning_rate": 9.440000000000001e-06, + "loss": 0.3592, + "step": 475 + }, + { + "epoch": 0.01, + "learning_rate": 9.940000000000001e-06, + "loss": 0.3627, + "step": 500 + }, + { + "epoch": 0.01, + "learning_rate": 9.997541899441341e-06, + "loss": 0.3692, + "step": 525 + }, + { + "epoch": 0.01, + "learning_rate": 9.994748603351955e-06, + "loss": 0.3787, + "step": 550 + }, + { + "epoch": 0.01, + "learning_rate": 9.991955307262571e-06, + "loss": 0.3698, + "step": 575 + }, + { + "epoch": 0.01, + "learning_rate": 9.989162011173185e-06, + "loss": 0.3563, + "step": 600 + }, + { + "epoch": 0.01, + "learning_rate": 9.9863687150838e-06, + "loss": 0.4017, + "step": 625 + }, + { + "epoch": 0.01, + "learning_rate": 9.983575418994414e-06, + "loss": 0.3663, + "step": 650 + }, + { + "epoch": 0.01, + "learning_rate": 9.980782122905028e-06, + "loss": 0.4053, + "step": 675 + }, + { + "epoch": 0.01, + "learning_rate": 9.977988826815644e-06, + "loss": 0.3679, + "step": 700 + }, + { + "epoch": 0.01, + "learning_rate": 9.975195530726259e-06, + "loss": 0.4012, + "step": 725 + }, + { + "epoch": 0.01, + "learning_rate": 9.972402234636871e-06, + "loss": 0.3551, + "step": 750 + }, + { + "epoch": 0.01, + "learning_rate": 9.969608938547487e-06, + "loss": 0.3755, + "step": 775 + }, + { + "epoch": 0.01, + "learning_rate": 9.9668156424581e-06, + "loss": 0.3693, + "step": 800 + }, + { + "epoch": 0.01, + "learning_rate": 9.964022346368716e-06, + "loss": 0.3555, + "step": 825 + }, + { + "epoch": 0.01, + "learning_rate": 9.96122905027933e-06, + "loss": 0.4021, + "step": 850 + }, + { + "epoch": 0.01, + "learning_rate": 9.958435754189944e-06, + "loss": 0.3644, + "step": 875 + }, + { + "epoch": 0.01, + "learning_rate": 9.95564245810056e-06, + "loss": 0.3162, + "step": 900 + }, + { + "epoch": 0.01, + "learning_rate": 9.952849162011175e-06, + "loss": 0.3966, + "step": 925 + }, + { + "epoch": 0.01, + "learning_rate": 9.950055865921789e-06, + "loss": 0.3304, + "step": 950 + }, + { + "epoch": 0.01, + "learning_rate": 9.947262569832403e-06, + "loss": 0.3919, + "step": 975 + }, + { + "epoch": 0.01, + "learning_rate": 9.944469273743017e-06, + "loss": 0.3778, + "step": 1000 + }, + { + "epoch": 0.01, + "learning_rate": 9.941675977653632e-06, + "loss": 0.3526, + "step": 1025 + }, + { + "epoch": 0.01, + "learning_rate": 9.938882681564248e-06, + "loss": 0.3938, + "step": 1050 + }, + { + "epoch": 0.01, + "learning_rate": 9.936089385474861e-06, + "loss": 0.3311, + "step": 1075 + }, + { + "epoch": 0.01, + "learning_rate": 9.933296089385475e-06, + "loss": 0.3767, + "step": 1100 + }, + { + "epoch": 0.01, + "learning_rate": 9.930502793296091e-06, + "loss": 0.368, + "step": 1125 + }, + { + "epoch": 0.01, + "learning_rate": 9.927709497206705e-06, + "loss": 0.3693, + "step": 1150 + }, + { + "epoch": 0.01, + "learning_rate": 9.92491620111732e-06, + "loss": 0.3699, + "step": 1175 + }, + { + "epoch": 0.01, + "learning_rate": 9.922122905027934e-06, + "loss": 0.3641, + "step": 1200 + }, + { + "epoch": 0.01, + "learning_rate": 9.919329608938548e-06, + "loss": 0.3404, + "step": 1225 + }, + { + "epoch": 0.01, + "learning_rate": 9.916536312849164e-06, + "loss": 0.3496, + "step": 1250 + }, + { + "epoch": 0.01, + "learning_rate": 9.913743016759777e-06, + "loss": 0.3785, + "step": 1275 + }, + { + "epoch": 0.01, + "learning_rate": 9.910949720670391e-06, + "loss": 0.3413, + "step": 1300 + }, + { + "epoch": 0.01, + "learning_rate": 9.908156424581007e-06, + "loss": 0.3179, + "step": 1325 + }, + { + "epoch": 0.01, + "learning_rate": 9.90536312849162e-06, + "loss": 0.3595, + "step": 1350 + }, + { + "epoch": 0.01, + "learning_rate": 9.902569832402236e-06, + "loss": 0.3528, + "step": 1375 + }, + { + "epoch": 0.01, + "learning_rate": 9.89977653631285e-06, + "loss": 0.3581, + "step": 1400 + }, + { + "epoch": 0.01, + "learning_rate": 9.896983240223464e-06, + "loss": 0.3382, + "step": 1425 + }, + { + "epoch": 0.01, + "learning_rate": 9.89418994413408e-06, + "loss": 0.4115, + "step": 1450 + }, + { + "epoch": 0.01, + "learning_rate": 9.891396648044693e-06, + "loss": 0.363, + "step": 1475 + }, + { + "epoch": 0.02, + "learning_rate": 9.888603351955309e-06, + "loss": 0.3411, + "step": 1500 + }, + { + "epoch": 0.02, + "learning_rate": 9.885810055865923e-06, + "loss": 0.3286, + "step": 1525 + }, + { + "epoch": 0.02, + "learning_rate": 9.883016759776537e-06, + "loss": 0.3721, + "step": 1550 + }, + { + "epoch": 0.02, + "learning_rate": 9.880223463687152e-06, + "loss": 0.3447, + "step": 1575 + }, + { + "epoch": 0.02, + "learning_rate": 9.877430167597766e-06, + "loss": 0.3646, + "step": 1600 + }, + { + "epoch": 0.02, + "learning_rate": 9.874636871508381e-06, + "loss": 0.3285, + "step": 1625 + }, + { + "epoch": 0.02, + "learning_rate": 9.871843575418995e-06, + "loss": 0.3691, + "step": 1650 + }, + { + "epoch": 0.02, + "learning_rate": 9.869050279329609e-06, + "loss": 0.3433, + "step": 1675 + }, + { + "epoch": 0.02, + "learning_rate": 9.866256983240225e-06, + "loss": 0.3747, + "step": 1700 + }, + { + "epoch": 0.02, + "learning_rate": 9.863463687150839e-06, + "loss": 0.3377, + "step": 1725 + }, + { + "epoch": 0.02, + "learning_rate": 9.860670391061452e-06, + "loss": 0.3449, + "step": 1750 + }, + { + "epoch": 0.02, + "learning_rate": 9.857877094972068e-06, + "loss": 0.3684, + "step": 1775 + }, + { + "epoch": 0.02, + "learning_rate": 9.855083798882682e-06, + "loss": 0.334, + "step": 1800 + }, + { + "epoch": 0.02, + "learning_rate": 9.852290502793297e-06, + "loss": 0.3461, + "step": 1825 + }, + { + "epoch": 0.02, + "learning_rate": 9.849497206703911e-06, + "loss": 0.3135, + "step": 1850 + }, + { + "epoch": 0.02, + "learning_rate": 9.846703910614525e-06, + "loss": 0.3661, + "step": 1875 + }, + { + "epoch": 0.02, + "learning_rate": 9.84391061452514e-06, + "loss": 0.3266, + "step": 1900 + }, + { + "epoch": 0.02, + "learning_rate": 9.841117318435754e-06, + "loss": 0.3661, + "step": 1925 + }, + { + "epoch": 0.02, + "learning_rate": 9.83832402234637e-06, + "loss": 0.2896, + "step": 1950 + }, + { + "epoch": 0.02, + "learning_rate": 9.835530726256984e-06, + "loss": 0.3318, + "step": 1975 + }, + { + "epoch": 0.02, + "learning_rate": 9.832737430167598e-06, + "loss": 0.3284, + "step": 2000 + }, + { + "epoch": 0.02, + "learning_rate": 9.829944134078213e-06, + "loss": 0.3404, + "step": 2025 + }, + { + "epoch": 0.02, + "learning_rate": 9.827150837988829e-06, + "loss": 0.3802, + "step": 2050 + }, + { + "epoch": 0.02, + "learning_rate": 9.824357541899443e-06, + "loss": 0.3754, + "step": 2075 + }, + { + "epoch": 0.02, + "learning_rate": 9.821564245810056e-06, + "loss": 0.3526, + "step": 2100 + }, + { + "epoch": 0.02, + "learning_rate": 9.81877094972067e-06, + "loss": 0.3903, + "step": 2125 + }, + { + "epoch": 0.02, + "learning_rate": 9.815977653631286e-06, + "loss": 0.3734, + "step": 2150 + }, + { + "epoch": 0.02, + "learning_rate": 9.8131843575419e-06, + "loss": 0.3331, + "step": 2175 + }, + { + "epoch": 0.02, + "learning_rate": 9.810391061452514e-06, + "loss": 0.3546, + "step": 2200 + }, + { + "epoch": 0.02, + "learning_rate": 9.807597765363129e-06, + "loss": 0.361, + "step": 2225 + }, + { + "epoch": 0.02, + "learning_rate": 9.804804469273745e-06, + "loss": 0.3747, + "step": 2250 + }, + { + "epoch": 0.02, + "learning_rate": 9.802011173184359e-06, + "loss": 0.2949, + "step": 2275 + }, + { + "epoch": 0.02, + "learning_rate": 9.799217877094972e-06, + "loss": 0.3596, + "step": 2300 + }, + { + "epoch": 0.02, + "learning_rate": 9.796424581005586e-06, + "loss": 0.3211, + "step": 2325 + }, + { + "epoch": 0.02, + "learning_rate": 9.793631284916202e-06, + "loss": 0.3527, + "step": 2350 + }, + { + "epoch": 0.02, + "learning_rate": 9.790837988826817e-06, + "loss": 0.3643, + "step": 2375 + }, + { + "epoch": 0.02, + "learning_rate": 9.788044692737431e-06, + "loss": 0.3279, + "step": 2400 + }, + { + "epoch": 0.02, + "learning_rate": 9.785251396648045e-06, + "loss": 0.3237, + "step": 2425 + }, + { + "epoch": 0.02, + "learning_rate": 9.78245810055866e-06, + "loss": 0.3128, + "step": 2450 + }, + { + "epoch": 0.02, + "learning_rate": 9.779664804469274e-06, + "loss": 0.3576, + "step": 2475 + }, + { + "epoch": 0.03, + "learning_rate": 9.77687150837989e-06, + "loss": 0.3207, + "step": 2500 + }, + { + "epoch": 0.03, + "eval_loss": 0.3457668423652649, + "eval_runtime": 1955.1642, + "eval_samples_per_second": 2.56, + "eval_steps_per_second": 0.16, + "eval_wer": 24.49275018391514, + "step": 2500 + }, + { + "epoch": 0.03, + "learning_rate": 9.774078212290504e-06, + "loss": 0.3304, + "step": 2525 + }, + { + "epoch": 0.03, + "learning_rate": 9.771284916201118e-06, + "loss": 0.3589, + "step": 2550 + }, + { + "epoch": 0.03, + "learning_rate": 9.768491620111733e-06, + "loss": 0.3662, + "step": 2575 + }, + { + "epoch": 0.03, + "learning_rate": 9.765698324022347e-06, + "loss": 0.3785, + "step": 2600 + }, + { + "epoch": 0.03, + "learning_rate": 9.762905027932963e-06, + "loss": 0.3259, + "step": 2625 + }, + { + "epoch": 0.03, + "learning_rate": 9.760111731843576e-06, + "loss": 0.324, + "step": 2650 + }, + { + "epoch": 0.03, + "learning_rate": 9.75731843575419e-06, + "loss": 0.3435, + "step": 2675 + }, + { + "epoch": 0.03, + "learning_rate": 9.754525139664806e-06, + "loss": 0.3414, + "step": 2700 + }, + { + "epoch": 0.03, + "learning_rate": 9.75173184357542e-06, + "loss": 0.3546, + "step": 2725 + }, + { + "epoch": 0.03, + "learning_rate": 9.748938547486034e-06, + "loss": 0.3692, + "step": 2750 + }, + { + "epoch": 0.03, + "learning_rate": 9.746145251396649e-06, + "loss": 0.3405, + "step": 2775 + }, + { + "epoch": 0.03, + "learning_rate": 9.743351955307263e-06, + "loss": 0.3608, + "step": 2800 + }, + { + "epoch": 0.03, + "learning_rate": 9.740558659217879e-06, + "loss": 0.3806, + "step": 2825 + }, + { + "epoch": 0.03, + "learning_rate": 9.737765363128492e-06, + "loss": 0.321, + "step": 2850 + }, + { + "epoch": 0.03, + "learning_rate": 9.734972067039106e-06, + "loss": 0.3465, + "step": 2875 + }, + { + "epoch": 0.03, + "learning_rate": 9.732178770949722e-06, + "loss": 0.2844, + "step": 2900 + }, + { + "epoch": 0.03, + "learning_rate": 9.729385474860336e-06, + "loss": 0.3555, + "step": 2925 + }, + { + "epoch": 0.03, + "learning_rate": 9.726592178770951e-06, + "loss": 0.3118, + "step": 2950 + }, + { + "epoch": 0.03, + "learning_rate": 9.723798882681565e-06, + "loss": 0.374, + "step": 2975 + }, + { + "epoch": 0.03, + "learning_rate": 9.721005586592179e-06, + "loss": 0.3427, + "step": 3000 + }, + { + "epoch": 0.03, + "learning_rate": 9.718212290502794e-06, + "loss": 0.3054, + "step": 3025 + }, + { + "epoch": 0.03, + "learning_rate": 9.715418994413408e-06, + "loss": 0.3738, + "step": 3050 + }, + { + "epoch": 0.03, + "learning_rate": 9.712625698324024e-06, + "loss": 0.3525, + "step": 3075 + }, + { + "epoch": 0.03, + "learning_rate": 9.709832402234638e-06, + "loss": 0.3514, + "step": 3100 + }, + { + "epoch": 0.03, + "learning_rate": 9.707039106145251e-06, + "loss": 0.3291, + "step": 3125 + }, + { + "epoch": 0.03, + "learning_rate": 9.704245810055867e-06, + "loss": 0.2935, + "step": 3150 + }, + { + "epoch": 0.03, + "learning_rate": 9.701452513966481e-06, + "loss": 0.352, + "step": 3175 + }, + { + "epoch": 0.03, + "learning_rate": 9.698659217877095e-06, + "loss": 0.403, + "step": 3200 + }, + { + "epoch": 0.03, + "learning_rate": 9.69586592178771e-06, + "loss": 0.328, + "step": 3225 + }, + { + "epoch": 0.03, + "learning_rate": 9.693072625698324e-06, + "loss": 0.3691, + "step": 3250 + }, + { + "epoch": 0.03, + "learning_rate": 9.69027932960894e-06, + "loss": 0.3313, + "step": 3275 + }, + { + "epoch": 0.03, + "learning_rate": 9.687486033519554e-06, + "loss": 0.3481, + "step": 3300 + }, + { + "epoch": 0.03, + "learning_rate": 9.684692737430167e-06, + "loss": 0.358, + "step": 3325 + }, + { + "epoch": 0.03, + "learning_rate": 9.681899441340783e-06, + "loss": 0.3522, + "step": 3350 + }, + { + "epoch": 0.03, + "learning_rate": 9.679106145251398e-06, + "loss": 0.3674, + "step": 3375 + }, + { + "epoch": 0.03, + "learning_rate": 9.676312849162012e-06, + "loss": 0.3277, + "step": 3400 + }, + { + "epoch": 0.03, + "learning_rate": 9.673519553072626e-06, + "loss": 0.3194, + "step": 3425 + }, + { + "epoch": 0.03, + "learning_rate": 9.67072625698324e-06, + "loss": 0.3617, + "step": 3450 + }, + { + "epoch": 0.03, + "learning_rate": 9.667932960893856e-06, + "loss": 0.3334, + "step": 3475 + }, + { + "epoch": 0.04, + "learning_rate": 9.665139664804471e-06, + "loss": 0.3728, + "step": 3500 + }, + { + "epoch": 0.04, + "learning_rate": 9.662346368715085e-06, + "loss": 0.3464, + "step": 3525 + }, + { + "epoch": 0.04, + "learning_rate": 9.659553072625699e-06, + "loss": 0.343, + "step": 3550 + }, + { + "epoch": 0.04, + "learning_rate": 9.656759776536314e-06, + "loss": 0.3496, + "step": 3575 + }, + { + "epoch": 0.04, + "learning_rate": 9.653966480446928e-06, + "loss": 0.3262, + "step": 3600 + }, + { + "epoch": 0.04, + "learning_rate": 9.651173184357542e-06, + "loss": 0.3127, + "step": 3625 + }, + { + "epoch": 0.04, + "learning_rate": 9.648379888268156e-06, + "loss": 0.3329, + "step": 3650 + }, + { + "epoch": 0.04, + "learning_rate": 9.645586592178771e-06, + "loss": 0.3329, + "step": 3675 + }, + { + "epoch": 0.04, + "learning_rate": 9.642793296089387e-06, + "loss": 0.3137, + "step": 3700 + }, + { + "epoch": 0.04, + "learning_rate": 9.640000000000001e-06, + "loss": 0.3186, + "step": 3725 + }, + { + "epoch": 0.04, + "learning_rate": 9.637206703910615e-06, + "loss": 0.3448, + "step": 3750 + }, + { + "epoch": 0.04, + "learning_rate": 9.63441340782123e-06, + "loss": 0.3238, + "step": 3775 + }, + { + "epoch": 0.04, + "learning_rate": 9.631620111731844e-06, + "loss": 0.3527, + "step": 3800 + }, + { + "epoch": 0.04, + "learning_rate": 9.62882681564246e-06, + "loss": 0.3538, + "step": 3825 + }, + { + "epoch": 0.04, + "learning_rate": 9.626033519553074e-06, + "loss": 0.3252, + "step": 3850 + }, + { + "epoch": 0.04, + "learning_rate": 9.623240223463687e-06, + "loss": 0.3241, + "step": 3875 + }, + { + "epoch": 0.04, + "learning_rate": 9.620446927374303e-06, + "loss": 0.3334, + "step": 3900 + }, + { + "epoch": 0.04, + "learning_rate": 9.617653631284917e-06, + "loss": 0.3489, + "step": 3925 + }, + { + "epoch": 0.04, + "learning_rate": 9.614860335195532e-06, + "loss": 0.3697, + "step": 3950 + }, + { + "epoch": 0.04, + "learning_rate": 9.612067039106146e-06, + "loss": 0.3522, + "step": 3975 + }, + { + "epoch": 0.04, + "learning_rate": 9.60927374301676e-06, + "loss": 0.3276, + "step": 4000 + }, + { + "epoch": 0.04, + "learning_rate": 9.606480446927376e-06, + "loss": 0.3478, + "step": 4025 + }, + { + "epoch": 0.04, + "learning_rate": 9.603798882681566e-06, + "loss": 0.3646, + "step": 4050 + }, + { + "epoch": 0.04, + "learning_rate": 9.60100558659218e-06, + "loss": 0.3981, + "step": 4075 + }, + { + "epoch": 0.04, + "learning_rate": 9.598212290502794e-06, + "loss": 0.3265, + "step": 4100 + }, + { + "epoch": 0.04, + "learning_rate": 9.59541899441341e-06, + "loss": 0.3344, + "step": 4125 + }, + { + "epoch": 0.04, + "learning_rate": 9.592625698324023e-06, + "loss": 0.3179, + "step": 4150 + }, + { + "epoch": 0.04, + "learning_rate": 9.589832402234637e-06, + "loss": 0.3201, + "step": 4175 + }, + { + "epoch": 0.04, + "learning_rate": 9.587039106145251e-06, + "loss": 0.3569, + "step": 4200 + }, + { + "epoch": 0.04, + "learning_rate": 9.584245810055866e-06, + "loss": 0.3668, + "step": 4225 + }, + { + "epoch": 0.04, + "learning_rate": 9.581452513966482e-06, + "loss": 0.3108, + "step": 4250 + }, + { + "epoch": 0.04, + "learning_rate": 9.578659217877096e-06, + "loss": 0.301, + "step": 4275 + }, + { + "epoch": 0.04, + "learning_rate": 9.57586592178771e-06, + "loss": 0.3305, + "step": 4300 + }, + { + "epoch": 0.04, + "learning_rate": 9.573072625698325e-06, + "loss": 0.328, + "step": 4325 + }, + { + "epoch": 0.04, + "learning_rate": 9.570279329608939e-06, + "loss": 0.3639, + "step": 4350 + }, + { + "epoch": 0.04, + "learning_rate": 9.567486033519555e-06, + "loss": 0.3616, + "step": 4375 + }, + { + "epoch": 0.04, + "learning_rate": 9.564692737430169e-06, + "loss": 0.3035, + "step": 4400 + }, + { + "epoch": 0.04, + "learning_rate": 9.561899441340782e-06, + "loss": 0.3191, + "step": 4425 + }, + { + "epoch": 0.04, + "learning_rate": 9.559106145251398e-06, + "loss": 0.3329, + "step": 4450 + }, + { + "epoch": 0.04, + "learning_rate": 9.556312849162012e-06, + "loss": 0.3168, + "step": 4475 + }, + { + "epoch": 0.05, + "learning_rate": 9.553519553072627e-06, + "loss": 0.3678, + "step": 4500 + }, + { + "epoch": 0.05, + "learning_rate": 9.550726256983241e-06, + "loss": 0.3078, + "step": 4525 + }, + { + "epoch": 0.05, + "learning_rate": 9.547932960893855e-06, + "loss": 0.3177, + "step": 4550 + }, + { + "epoch": 0.05, + "learning_rate": 9.54513966480447e-06, + "loss": 0.3319, + "step": 4575 + }, + { + "epoch": 0.05, + "learning_rate": 9.542346368715084e-06, + "loss": 0.3178, + "step": 4600 + }, + { + "epoch": 0.05, + "learning_rate": 9.539553072625698e-06, + "loss": 0.3618, + "step": 4625 + }, + { + "epoch": 0.05, + "learning_rate": 9.536759776536314e-06, + "loss": 0.3413, + "step": 4650 + }, + { + "epoch": 0.05, + "learning_rate": 9.533966480446928e-06, + "loss": 0.3385, + "step": 4675 + }, + { + "epoch": 0.05, + "learning_rate": 9.531173184357543e-06, + "loss": 0.344, + "step": 4700 + }, + { + "epoch": 0.05, + "learning_rate": 9.528379888268157e-06, + "loss": 0.3472, + "step": 4725 + }, + { + "epoch": 0.05, + "learning_rate": 9.525586592178771e-06, + "loss": 0.3189, + "step": 4750 + }, + { + "epoch": 0.05, + "learning_rate": 9.522793296089386e-06, + "loss": 0.3208, + "step": 4775 + }, + { + "epoch": 0.05, + "learning_rate": 9.52e-06, + "loss": 0.3359, + "step": 4800 + }, + { + "epoch": 0.05, + "learning_rate": 9.517206703910616e-06, + "loss": 0.336, + "step": 4825 + }, + { + "epoch": 0.05, + "learning_rate": 9.51441340782123e-06, + "loss": 0.3512, + "step": 4850 + }, + { + "epoch": 0.05, + "learning_rate": 9.511620111731844e-06, + "loss": 0.3432, + "step": 4875 + }, + { + "epoch": 0.05, + "learning_rate": 9.508826815642459e-06, + "loss": 0.315, + "step": 4900 + }, + { + "epoch": 0.05, + "learning_rate": 9.506033519553073e-06, + "loss": 0.3419, + "step": 4925 + }, + { + "epoch": 0.05, + "learning_rate": 9.503240223463689e-06, + "loss": 0.323, + "step": 4950 + }, + { + "epoch": 0.05, + "learning_rate": 9.500446927374302e-06, + "loss": 0.358, + "step": 4975 + }, + { + "epoch": 0.05, + "learning_rate": 9.497653631284916e-06, + "loss": 0.3264, + "step": 5000 + }, + { + "epoch": 0.05, + "eval_loss": 0.3315780758857727, + "eval_runtime": 1884.7604, + "eval_samples_per_second": 2.656, + "eval_steps_per_second": 0.166, + "eval_wer": 23.408244144379317, + "step": 5000 + }, + { + "epoch": 0.05, + "learning_rate": 9.494860335195532e-06, + "loss": 0.3394, + "step": 5025 + }, + { + "epoch": 0.05, + "learning_rate": 9.492067039106146e-06, + "loss": 0.3042, + "step": 5050 + }, + { + "epoch": 0.05, + "learning_rate": 9.48927374301676e-06, + "loss": 0.3489, + "step": 5075 + }, + { + "epoch": 0.05, + "learning_rate": 9.486480446927375e-06, + "loss": 0.3268, + "step": 5100 + }, + { + "epoch": 0.05, + "learning_rate": 9.483687150837989e-06, + "loss": 0.3115, + "step": 5125 + }, + { + "epoch": 0.05, + "learning_rate": 9.480893854748604e-06, + "loss": 0.3127, + "step": 5150 + }, + { + "epoch": 0.05, + "learning_rate": 9.478100558659218e-06, + "loss": 0.3652, + "step": 5175 + }, + { + "epoch": 0.05, + "learning_rate": 9.475307262569832e-06, + "loss": 0.2886, + "step": 5200 + }, + { + "epoch": 0.05, + "learning_rate": 9.472513966480448e-06, + "loss": 0.3148, + "step": 5225 + }, + { + "epoch": 0.05, + "learning_rate": 9.469720670391063e-06, + "loss": 0.2918, + "step": 5250 + }, + { + "epoch": 0.05, + "learning_rate": 9.466927374301677e-06, + "loss": 0.2931, + "step": 5275 + }, + { + "epoch": 0.05, + "learning_rate": 9.464134078212291e-06, + "loss": 0.3023, + "step": 5300 + }, + { + "epoch": 0.05, + "learning_rate": 9.461340782122905e-06, + "loss": 0.3384, + "step": 5325 + }, + { + "epoch": 0.05, + "learning_rate": 9.45854748603352e-06, + "loss": 0.343, + "step": 5350 + }, + { + "epoch": 0.05, + "learning_rate": 9.455754189944136e-06, + "loss": 0.3318, + "step": 5375 + }, + { + "epoch": 0.05, + "learning_rate": 9.45296089385475e-06, + "loss": 0.2998, + "step": 5400 + }, + { + "epoch": 0.05, + "learning_rate": 9.450167597765364e-06, + "loss": 0.3328, + "step": 5425 + }, + { + "epoch": 0.05, + "learning_rate": 9.447374301675979e-06, + "loss": 0.3258, + "step": 5450 + }, + { + "epoch": 0.05, + "learning_rate": 9.444581005586593e-06, + "loss": 0.3201, + "step": 5475 + }, + { + "epoch": 0.06, + "learning_rate": 9.441787709497209e-06, + "loss": 0.3189, + "step": 5500 + }, + { + "epoch": 0.06, + "learning_rate": 9.438994413407822e-06, + "loss": 0.3042, + "step": 5525 + }, + { + "epoch": 0.06, + "learning_rate": 9.436201117318436e-06, + "loss": 0.3058, + "step": 5550 + }, + { + "epoch": 0.06, + "learning_rate": 9.433407821229052e-06, + "loss": 0.3335, + "step": 5575 + }, + { + "epoch": 0.06, + "learning_rate": 9.430614525139666e-06, + "loss": 0.2892, + "step": 5600 + }, + { + "epoch": 0.06, + "learning_rate": 9.42782122905028e-06, + "loss": 0.3198, + "step": 5625 + }, + { + "epoch": 0.06, + "learning_rate": 9.425027932960895e-06, + "loss": 0.3164, + "step": 5650 + }, + { + "epoch": 0.06, + "learning_rate": 9.422234636871509e-06, + "loss": 0.3167, + "step": 5675 + }, + { + "epoch": 0.06, + "learning_rate": 9.419441340782124e-06, + "loss": 0.3411, + "step": 5700 + }, + { + "epoch": 0.06, + "learning_rate": 9.416648044692738e-06, + "loss": 0.3157, + "step": 5725 + }, + { + "epoch": 0.06, + "learning_rate": 9.413854748603352e-06, + "loss": 0.3082, + "step": 5750 + }, + { + "epoch": 0.06, + "learning_rate": 9.411061452513968e-06, + "loss": 0.2975, + "step": 5775 + }, + { + "epoch": 0.06, + "learning_rate": 9.408268156424581e-06, + "loss": 0.3462, + "step": 5800 + }, + { + "epoch": 0.06, + "learning_rate": 9.405474860335197e-06, + "loss": 0.3326, + "step": 5825 + }, + { + "epoch": 0.06, + "learning_rate": 9.402681564245811e-06, + "loss": 0.304, + "step": 5850 + }, + { + "epoch": 0.06, + "learning_rate": 9.399888268156425e-06, + "loss": 0.304, + "step": 5875 + }, + { + "epoch": 0.06, + "learning_rate": 9.39709497206704e-06, + "loss": 0.2968, + "step": 5900 + }, + { + "epoch": 0.06, + "learning_rate": 9.394301675977654e-06, + "loss": 0.3245, + "step": 5925 + }, + { + "epoch": 0.06, + "learning_rate": 9.39150837988827e-06, + "loss": 0.3071, + "step": 5950 + }, + { + "epoch": 0.06, + "learning_rate": 9.388715083798884e-06, + "loss": 0.3481, + "step": 5975 + }, + { + "epoch": 0.06, + "learning_rate": 9.385921787709497e-06, + "loss": 0.346, + "step": 6000 + }, + { + "epoch": 0.06, + "learning_rate": 9.383128491620113e-06, + "loss": 0.2907, + "step": 6025 + }, + { + "epoch": 0.06, + "learning_rate": 9.380335195530727e-06, + "loss": 0.3435, + "step": 6050 + }, + { + "epoch": 0.06, + "learning_rate": 9.37754189944134e-06, + "loss": 0.335, + "step": 6075 + }, + { + "epoch": 0.06, + "learning_rate": 9.374748603351956e-06, + "loss": 0.3284, + "step": 6100 + }, + { + "epoch": 0.06, + "learning_rate": 9.37195530726257e-06, + "loss": 0.344, + "step": 6125 + }, + { + "epoch": 0.06, + "learning_rate": 9.369162011173186e-06, + "loss": 0.3286, + "step": 6150 + }, + { + "epoch": 0.06, + "learning_rate": 9.3663687150838e-06, + "loss": 0.3186, + "step": 6175 + }, + { + "epoch": 0.06, + "learning_rate": 9.363575418994413e-06, + "loss": 0.3291, + "step": 6200 + }, + { + "epoch": 0.06, + "learning_rate": 9.360782122905029e-06, + "loss": 0.3162, + "step": 6225 + }, + { + "epoch": 0.06, + "learning_rate": 9.357988826815643e-06, + "loss": 0.2876, + "step": 6250 + }, + { + "epoch": 0.06, + "learning_rate": 9.355195530726258e-06, + "loss": 0.339, + "step": 6275 + }, + { + "epoch": 0.06, + "learning_rate": 9.352402234636872e-06, + "loss": 0.3686, + "step": 6300 + }, + { + "epoch": 0.06, + "learning_rate": 9.349608938547486e-06, + "loss": 0.2804, + "step": 6325 + }, + { + "epoch": 0.06, + "learning_rate": 9.346815642458101e-06, + "loss": 0.3308, + "step": 6350 + }, + { + "epoch": 0.06, + "learning_rate": 9.344022346368717e-06, + "loss": 0.3588, + "step": 6375 + }, + { + "epoch": 0.06, + "learning_rate": 9.341229050279331e-06, + "loss": 0.3046, + "step": 6400 + }, + { + "epoch": 0.06, + "learning_rate": 9.338435754189945e-06, + "loss": 0.3511, + "step": 6425 + }, + { + "epoch": 0.06, + "learning_rate": 9.335642458100559e-06, + "loss": 0.3157, + "step": 6450 + }, + { + "epoch": 0.06, + "learning_rate": 9.332849162011174e-06, + "loss": 0.2966, + "step": 6475 + }, + { + "epoch": 0.07, + "learning_rate": 9.330055865921788e-06, + "loss": 0.2805, + "step": 6500 + }, + { + "epoch": 0.07, + "learning_rate": 9.327262569832402e-06, + "loss": 0.2846, + "step": 6525 + }, + { + "epoch": 0.07, + "learning_rate": 9.324469273743017e-06, + "loss": 0.3633, + "step": 6550 + }, + { + "epoch": 0.07, + "learning_rate": 9.321675977653633e-06, + "loss": 0.3094, + "step": 6575 + }, + { + "epoch": 0.07, + "learning_rate": 9.318882681564247e-06, + "loss": 0.3063, + "step": 6600 + }, + { + "epoch": 0.07, + "learning_rate": 9.31608938547486e-06, + "loss": 0.328, + "step": 6625 + }, + { + "epoch": 0.07, + "learning_rate": 9.313296089385474e-06, + "loss": 0.3322, + "step": 6650 + }, + { + "epoch": 0.07, + "learning_rate": 9.31050279329609e-06, + "loss": 0.3157, + "step": 6675 + }, + { + "epoch": 0.07, + "learning_rate": 9.307709497206706e-06, + "loss": 0.3237, + "step": 6700 + }, + { + "epoch": 0.07, + "learning_rate": 9.30491620111732e-06, + "loss": 0.2976, + "step": 6725 + }, + { + "epoch": 0.07, + "learning_rate": 9.302122905027933e-06, + "loss": 0.2819, + "step": 6750 + }, + { + "epoch": 0.07, + "learning_rate": 9.299329608938549e-06, + "loss": 0.3415, + "step": 6775 + }, + { + "epoch": 0.07, + "learning_rate": 9.296536312849163e-06, + "loss": 0.3351, + "step": 6800 + }, + { + "epoch": 0.07, + "learning_rate": 9.293743016759778e-06, + "loss": 0.3308, + "step": 6825 + }, + { + "epoch": 0.07, + "learning_rate": 9.290949720670392e-06, + "loss": 0.328, + "step": 6850 + }, + { + "epoch": 0.07, + "learning_rate": 9.288156424581006e-06, + "loss": 0.3255, + "step": 6875 + }, + { + "epoch": 0.07, + "learning_rate": 9.285363128491621e-06, + "loss": 0.2979, + "step": 6900 + }, + { + "epoch": 0.07, + "learning_rate": 9.282569832402235e-06, + "loss": 0.3163, + "step": 6925 + }, + { + "epoch": 0.07, + "learning_rate": 9.279776536312851e-06, + "loss": 0.3376, + "step": 6950 + }, + { + "epoch": 0.07, + "learning_rate": 9.276983240223465e-06, + "loss": 0.3421, + "step": 6975 + }, + { + "epoch": 0.07, + "learning_rate": 9.274189944134079e-06, + "loss": 0.3158, + "step": 7000 + }, + { + "epoch": 0.07, + "learning_rate": 9.271396648044694e-06, + "loss": 0.3216, + "step": 7025 + }, + { + "epoch": 0.07, + "learning_rate": 9.268603351955308e-06, + "loss": 0.292, + "step": 7050 + }, + { + "epoch": 0.07, + "learning_rate": 9.265810055865922e-06, + "loss": 0.3269, + "step": 7075 + }, + { + "epoch": 0.07, + "learning_rate": 9.263016759776537e-06, + "loss": 0.3164, + "step": 7100 + }, + { + "epoch": 0.07, + "learning_rate": 9.260223463687151e-06, + "loss": 0.3315, + "step": 7125 + }, + { + "epoch": 0.07, + "learning_rate": 9.257430167597767e-06, + "loss": 0.2825, + "step": 7150 + }, + { + "epoch": 0.07, + "learning_rate": 9.25463687150838e-06, + "loss": 0.3319, + "step": 7175 + }, + { + "epoch": 0.07, + "learning_rate": 9.251843575418994e-06, + "loss": 0.3279, + "step": 7200 + }, + { + "epoch": 0.07, + "learning_rate": 9.24905027932961e-06, + "loss": 0.3571, + "step": 7225 + }, + { + "epoch": 0.07, + "learning_rate": 9.246256983240224e-06, + "loss": 0.3131, + "step": 7250 + }, + { + "epoch": 0.07, + "learning_rate": 9.24346368715084e-06, + "loss": 0.3135, + "step": 7275 + }, + { + "epoch": 0.07, + "learning_rate": 9.240670391061453e-06, + "loss": 0.3303, + "step": 7300 + }, + { + "epoch": 0.07, + "learning_rate": 9.237877094972067e-06, + "loss": 0.3256, + "step": 7325 + }, + { + "epoch": 0.07, + "learning_rate": 9.235083798882683e-06, + "loss": 0.3299, + "step": 7350 + }, + { + "epoch": 0.07, + "learning_rate": 9.232290502793296e-06, + "loss": 0.3054, + "step": 7375 + }, + { + "epoch": 0.07, + "learning_rate": 9.229497206703912e-06, + "loss": 0.2909, + "step": 7400 + }, + { + "epoch": 0.07, + "learning_rate": 9.226703910614526e-06, + "loss": 0.3113, + "step": 7425 + }, + { + "epoch": 0.07, + "learning_rate": 9.22391061452514e-06, + "loss": 0.3183, + "step": 7450 + }, + { + "epoch": 0.07, + "learning_rate": 9.221117318435755e-06, + "loss": 0.3228, + "step": 7475 + }, + { + "epoch": 0.08, + "learning_rate": 9.218324022346369e-06, + "loss": 0.3223, + "step": 7500 + }, + { + "epoch": 0.08, + "eval_loss": 0.3239099085330963, + "eval_runtime": 2021.9396, + "eval_samples_per_second": 2.475, + "eval_steps_per_second": 0.155, + "eval_wer": 25.105009611049145, + "step": 7500 + }, + { + "epoch": 0.08, + "learning_rate": 9.215530726256983e-06, + "loss": 0.3243, + "step": 7525 + }, + { + "epoch": 0.08, + "learning_rate": 9.212737430167599e-06, + "loss": 0.3268, + "step": 7550 + }, + { + "epoch": 0.08, + "learning_rate": 9.209944134078212e-06, + "loss": 0.3337, + "step": 7575 + }, + { + "epoch": 0.08, + "learning_rate": 9.207262569832403e-06, + "loss": 0.3308, + "step": 7600 + }, + { + "epoch": 0.08, + "learning_rate": 9.204469273743017e-06, + "loss": 0.288, + "step": 7625 + }, + { + "epoch": 0.08, + "learning_rate": 9.201675977653632e-06, + "loss": 0.2891, + "step": 7650 + }, + { + "epoch": 0.08, + "learning_rate": 9.198882681564246e-06, + "loss": 0.3397, + "step": 7675 + }, + { + "epoch": 0.08, + "learning_rate": 9.196089385474862e-06, + "loss": 0.3192, + "step": 7700 + }, + { + "epoch": 0.08, + "learning_rate": 9.193296089385476e-06, + "loss": 0.2998, + "step": 7725 + }, + { + "epoch": 0.08, + "learning_rate": 9.19050279329609e-06, + "loss": 0.3535, + "step": 7750 + }, + { + "epoch": 0.08, + "learning_rate": 9.187709497206705e-06, + "loss": 0.3422, + "step": 7775 + }, + { + "epoch": 0.08, + "learning_rate": 9.184916201117319e-06, + "loss": 0.2789, + "step": 7800 + }, + { + "epoch": 0.08, + "learning_rate": 9.182122905027934e-06, + "loss": 0.2941, + "step": 7825 + }, + { + "epoch": 0.08, + "learning_rate": 9.179329608938548e-06, + "loss": 0.2866, + "step": 7850 + }, + { + "epoch": 0.08, + "learning_rate": 9.176536312849162e-06, + "loss": 0.2951, + "step": 7875 + }, + { + "epoch": 0.08, + "learning_rate": 9.173743016759778e-06, + "loss": 0.2932, + "step": 7900 + }, + { + "epoch": 0.08, + "learning_rate": 9.170949720670392e-06, + "loss": 0.3058, + "step": 7925 + }, + { + "epoch": 0.08, + "learning_rate": 9.168156424581005e-06, + "loss": 0.3455, + "step": 7950 + }, + { + "epoch": 0.08, + "learning_rate": 9.165363128491621e-06, + "loss": 0.3048, + "step": 7975 + }, + { + "epoch": 0.08, + "learning_rate": 9.162569832402235e-06, + "loss": 0.3181, + "step": 8000 + }, + { + "epoch": 0.08, + "learning_rate": 9.15977653631285e-06, + "loss": 0.3227, + "step": 8025 + }, + { + "epoch": 0.08, + "learning_rate": 9.156983240223464e-06, + "loss": 0.3032, + "step": 8050 + }, + { + "epoch": 0.08, + "learning_rate": 9.154189944134078e-06, + "loss": 0.3138, + "step": 8075 + }, + { + "epoch": 0.08, + "learning_rate": 9.151396648044694e-06, + "loss": 0.2907, + "step": 8100 + }, + { + "epoch": 0.08, + "learning_rate": 9.148603351955307e-06, + "loss": 0.3406, + "step": 8125 + }, + { + "epoch": 0.08, + "learning_rate": 9.145810055865923e-06, + "loss": 0.3234, + "step": 8150 + }, + { + "epoch": 0.08, + "learning_rate": 9.143016759776537e-06, + "loss": 0.34, + "step": 8175 + }, + { + "epoch": 0.08, + "learning_rate": 9.14022346368715e-06, + "loss": 0.309, + "step": 8200 + }, + { + "epoch": 0.08, + "learning_rate": 9.137430167597766e-06, + "loss": 0.3291, + "step": 8225 + }, + { + "epoch": 0.08, + "learning_rate": 9.134636871508382e-06, + "loss": 0.3026, + "step": 8250 + }, + { + "epoch": 0.08, + "learning_rate": 9.131843575418996e-06, + "loss": 0.2752, + "step": 8275 + }, + { + "epoch": 0.08, + "learning_rate": 9.12905027932961e-06, + "loss": 0.3305, + "step": 8300 + }, + { + "epoch": 0.08, + "learning_rate": 9.126256983240223e-06, + "loss": 0.3345, + "step": 8325 + }, + { + "epoch": 0.08, + "learning_rate": 9.123463687150839e-06, + "loss": 0.3081, + "step": 8350 + }, + { + "epoch": 0.08, + "learning_rate": 9.120670391061454e-06, + "loss": 0.3226, + "step": 8375 + }, + { + "epoch": 0.08, + "learning_rate": 9.117877094972068e-06, + "loss": 0.3195, + "step": 8400 + }, + { + "epoch": 0.08, + "learning_rate": 9.115083798882682e-06, + "loss": 0.3366, + "step": 8425 + }, + { + "epoch": 0.08, + "learning_rate": 9.112290502793298e-06, + "loss": 0.3295, + "step": 8450 + }, + { + "epoch": 0.08, + "learning_rate": 9.109497206703911e-06, + "loss": 0.3311, + "step": 8475 + }, + { + "epoch": 0.09, + "learning_rate": 9.106703910614525e-06, + "loss": 0.3352, + "step": 8500 + }, + { + "epoch": 0.09, + "learning_rate": 9.10391061452514e-06, + "loss": 0.2861, + "step": 8525 + }, + { + "epoch": 0.09, + "learning_rate": 9.101117318435755e-06, + "loss": 0.3098, + "step": 8550 + }, + { + "epoch": 0.09, + "learning_rate": 9.09832402234637e-06, + "loss": 0.3104, + "step": 8575 + }, + { + "epoch": 0.09, + "learning_rate": 9.095530726256984e-06, + "loss": 0.2831, + "step": 8600 + }, + { + "epoch": 0.09, + "learning_rate": 9.092737430167598e-06, + "loss": 0.3028, + "step": 8625 + }, + { + "epoch": 0.09, + "learning_rate": 9.089944134078214e-06, + "loss": 0.2974, + "step": 8650 + }, + { + "epoch": 0.09, + "learning_rate": 9.087150837988827e-06, + "loss": 0.3027, + "step": 8675 + }, + { + "epoch": 0.09, + "learning_rate": 9.084357541899443e-06, + "loss": 0.3133, + "step": 8700 + }, + { + "epoch": 0.09, + "learning_rate": 9.081564245810057e-06, + "loss": 0.311, + "step": 8725 + }, + { + "epoch": 0.09, + "learning_rate": 9.07877094972067e-06, + "loss": 0.3004, + "step": 8750 + }, + { + "epoch": 0.09, + "learning_rate": 9.075977653631286e-06, + "loss": 0.349, + "step": 8775 + }, + { + "epoch": 0.09, + "learning_rate": 9.0731843575419e-06, + "loss": 0.3342, + "step": 8800 + }, + { + "epoch": 0.09, + "learning_rate": 9.070391061452516e-06, + "loss": 0.2863, + "step": 8825 + }, + { + "epoch": 0.09, + "learning_rate": 9.06759776536313e-06, + "loss": 0.3346, + "step": 8850 + }, + { + "epoch": 0.09, + "learning_rate": 9.064804469273743e-06, + "loss": 0.3319, + "step": 8875 + }, + { + "epoch": 0.09, + "learning_rate": 9.062011173184359e-06, + "loss": 0.3283, + "step": 8900 + }, + { + "epoch": 0.09, + "learning_rate": 9.059217877094973e-06, + "loss": 0.2886, + "step": 8925 + }, + { + "epoch": 0.09, + "learning_rate": 9.056424581005587e-06, + "loss": 0.3101, + "step": 8950 + }, + { + "epoch": 0.09, + "learning_rate": 9.053631284916202e-06, + "loss": 0.3044, + "step": 8975 + }, + { + "epoch": 0.09, + "learning_rate": 9.050837988826816e-06, + "loss": 0.2963, + "step": 9000 + }, + { + "epoch": 0.09, + "learning_rate": 9.048044692737431e-06, + "loss": 0.318, + "step": 9025 + }, + { + "epoch": 0.09, + "learning_rate": 9.045251396648045e-06, + "loss": 0.3037, + "step": 9050 + }, + { + "epoch": 0.09, + "learning_rate": 9.04245810055866e-06, + "loss": 0.3057, + "step": 9075 + }, + { + "epoch": 0.09, + "learning_rate": 9.039664804469275e-06, + "loss": 0.3049, + "step": 9100 + }, + { + "epoch": 0.09, + "learning_rate": 9.036871508379889e-06, + "loss": 0.3148, + "step": 9125 + }, + { + "epoch": 0.09, + "learning_rate": 9.034078212290504e-06, + "loss": 0.3168, + "step": 9150 + }, + { + "epoch": 0.09, + "learning_rate": 9.031284916201118e-06, + "loss": 0.307, + "step": 9175 + }, + { + "epoch": 0.09, + "learning_rate": 9.028491620111732e-06, + "loss": 0.3061, + "step": 9200 + }, + { + "epoch": 0.09, + "learning_rate": 9.025698324022347e-06, + "loss": 0.3041, + "step": 9225 + }, + { + "epoch": 0.09, + "learning_rate": 9.022905027932961e-06, + "loss": 0.3068, + "step": 9250 + }, + { + "epoch": 0.09, + "learning_rate": 9.020111731843577e-06, + "loss": 0.3263, + "step": 9275 + }, + { + "epoch": 0.09, + "learning_rate": 9.01731843575419e-06, + "loss": 0.3192, + "step": 9300 + }, + { + "epoch": 0.09, + "learning_rate": 9.014525139664804e-06, + "loss": 0.3072, + "step": 9325 + }, + { + "epoch": 0.09, + "learning_rate": 9.01173184357542e-06, + "loss": 0.2932, + "step": 9350 + }, + { + "epoch": 0.09, + "learning_rate": 9.008938547486034e-06, + "loss": 0.3333, + "step": 9375 + }, + { + "epoch": 0.09, + "learning_rate": 9.006145251396648e-06, + "loss": 0.2981, + "step": 9400 + }, + { + "epoch": 0.09, + "learning_rate": 9.003351955307263e-06, + "loss": 0.3416, + "step": 9425 + }, + { + "epoch": 0.09, + "learning_rate": 9.000558659217877e-06, + "loss": 0.3331, + "step": 9450 + }, + { + "epoch": 0.09, + "learning_rate": 8.997765363128493e-06, + "loss": 0.3253, + "step": 9475 + }, + { + "epoch": 0.1, + "learning_rate": 8.994972067039107e-06, + "loss": 0.3231, + "step": 9500 + }, + { + "epoch": 0.1, + "learning_rate": 8.99217877094972e-06, + "loss": 0.3151, + "step": 9525 + }, + { + "epoch": 0.1, + "learning_rate": 8.989385474860336e-06, + "loss": 0.299, + "step": 9550 + }, + { + "epoch": 0.1, + "learning_rate": 8.986592178770951e-06, + "loss": 0.311, + "step": 9575 + }, + { + "epoch": 0.1, + "learning_rate": 8.983798882681565e-06, + "loss": 0.309, + "step": 9600 + }, + { + "epoch": 0.1, + "learning_rate": 8.98100558659218e-06, + "loss": 0.3092, + "step": 9625 + }, + { + "epoch": 0.1, + "learning_rate": 8.978212290502793e-06, + "loss": 0.3281, + "step": 9650 + }, + { + "epoch": 0.1, + "learning_rate": 8.975418994413409e-06, + "loss": 0.3227, + "step": 9675 + }, + { + "epoch": 0.1, + "learning_rate": 8.972625698324024e-06, + "loss": 0.3242, + "step": 9700 + }, + { + "epoch": 0.1, + "learning_rate": 8.969832402234638e-06, + "loss": 0.2984, + "step": 9725 + }, + { + "epoch": 0.1, + "learning_rate": 8.967039106145252e-06, + "loss": 0.3086, + "step": 9750 + }, + { + "epoch": 0.1, + "learning_rate": 8.964245810055867e-06, + "loss": 0.2793, + "step": 9775 + }, + { + "epoch": 0.1, + "learning_rate": 8.961452513966481e-06, + "loss": 0.3247, + "step": 9800 + }, + { + "epoch": 0.1, + "learning_rate": 8.958659217877097e-06, + "loss": 0.3279, + "step": 9825 + }, + { + "epoch": 0.1, + "learning_rate": 8.955865921787709e-06, + "loss": 0.3273, + "step": 9850 + }, + { + "epoch": 0.1, + "learning_rate": 8.953072625698324e-06, + "loss": 0.3243, + "step": 9875 + }, + { + "epoch": 0.1, + "learning_rate": 8.95027932960894e-06, + "loss": 0.3218, + "step": 9900 + }, + { + "epoch": 0.1, + "learning_rate": 8.947486033519554e-06, + "loss": 0.3473, + "step": 9925 + }, + { + "epoch": 0.1, + "learning_rate": 8.944692737430168e-06, + "loss": 0.3192, + "step": 9950 + }, + { + "epoch": 0.1, + "learning_rate": 8.941899441340783e-06, + "loss": 0.3296, + "step": 9975 + }, + { + "epoch": 0.1, + "learning_rate": 8.939106145251397e-06, + "loss": 0.3121, + "step": 10000 + }, + { + "epoch": 0.1, + "eval_loss": 0.31428924202919006, + "eval_runtime": 2085.6517, + "eval_samples_per_second": 2.4, + "eval_steps_per_second": 0.15, + "eval_wer": 23.455706115474975, + "step": 10000 + }, + { + "epoch": 0.1, + "learning_rate": 8.936312849162013e-06, + "loss": 0.3209, + "step": 10025 + }, + { + "epoch": 0.1, + "learning_rate": 8.933519553072626e-06, + "loss": 0.2865, + "step": 10050 + }, + { + "epoch": 0.1, + "learning_rate": 8.93072625698324e-06, + "loss": 0.313, + "step": 10075 + }, + { + "epoch": 0.1, + "learning_rate": 8.927932960893856e-06, + "loss": 0.2982, + "step": 10100 + }, + { + "epoch": 0.1, + "learning_rate": 8.92513966480447e-06, + "loss": 0.2759, + "step": 10125 + }, + { + "epoch": 0.1, + "learning_rate": 8.922346368715085e-06, + "loss": 0.3019, + "step": 10150 + }, + { + "epoch": 0.1, + "learning_rate": 8.919553072625699e-06, + "loss": 0.3208, + "step": 10175 + }, + { + "epoch": 0.1, + "learning_rate": 8.916759776536313e-06, + "loss": 0.2993, + "step": 10200 + }, + { + "epoch": 0.1, + "learning_rate": 8.913966480446929e-06, + "loss": 0.3264, + "step": 10225 + }, + { + "epoch": 0.1, + "learning_rate": 8.911173184357542e-06, + "loss": 0.3102, + "step": 10250 + }, + { + "epoch": 0.1, + "learning_rate": 8.908379888268158e-06, + "loss": 0.3288, + "step": 10275 + }, + { + "epoch": 0.1, + "learning_rate": 8.905586592178772e-06, + "loss": 0.2933, + "step": 10300 + }, + { + "epoch": 0.1, + "learning_rate": 8.902793296089386e-06, + "loss": 0.3253, + "step": 10325 + }, + { + "epoch": 0.1, + "learning_rate": 8.900000000000001e-06, + "loss": 0.3391, + "step": 10350 + }, + { + "epoch": 0.1, + "learning_rate": 8.897206703910615e-06, + "loss": 0.3203, + "step": 10375 + }, + { + "epoch": 0.1, + "learning_rate": 8.894413407821229e-06, + "loss": 0.3123, + "step": 10400 + }, + { + "epoch": 0.1, + "learning_rate": 8.891620111731844e-06, + "loss": 0.3193, + "step": 10425 + }, + { + "epoch": 0.1, + "learning_rate": 8.888826815642458e-06, + "loss": 0.3056, + "step": 10450 + }, + { + "epoch": 0.11, + "learning_rate": 8.886033519553074e-06, + "loss": 0.3146, + "step": 10475 + }, + { + "epoch": 0.11, + "learning_rate": 8.883240223463688e-06, + "loss": 0.2976, + "step": 10500 + }, + { + "epoch": 0.11, + "learning_rate": 8.880446927374302e-06, + "loss": 0.3261, + "step": 10525 + }, + { + "epoch": 0.11, + "learning_rate": 8.877653631284917e-06, + "loss": 0.3244, + "step": 10550 + }, + { + "epoch": 0.11, + "learning_rate": 8.874860335195531e-06, + "loss": 0.2843, + "step": 10575 + }, + { + "epoch": 0.11, + "learning_rate": 8.872067039106146e-06, + "loss": 0.2894, + "step": 10600 + }, + { + "epoch": 0.11, + "learning_rate": 8.86927374301676e-06, + "loss": 0.3314, + "step": 10625 + }, + { + "epoch": 0.11, + "learning_rate": 8.866480446927374e-06, + "loss": 0.3001, + "step": 10650 + }, + { + "epoch": 0.11, + "learning_rate": 8.86368715083799e-06, + "loss": 0.3292, + "step": 10675 + }, + { + "epoch": 0.11, + "learning_rate": 8.860893854748605e-06, + "loss": 0.3168, + "step": 10700 + }, + { + "epoch": 0.11, + "learning_rate": 8.858100558659219e-06, + "loss": 0.2887, + "step": 10725 + }, + { + "epoch": 0.11, + "learning_rate": 8.855307262569833e-06, + "loss": 0.3184, + "step": 10750 + }, + { + "epoch": 0.11, + "learning_rate": 8.852625698324024e-06, + "loss": 0.2771, + "step": 10775 + }, + { + "epoch": 0.11, + "learning_rate": 8.849832402234637e-06, + "loss": 0.3221, + "step": 10800 + }, + { + "epoch": 0.11, + "learning_rate": 8.847039106145251e-06, + "loss": 0.2883, + "step": 10825 + }, + { + "epoch": 0.11, + "learning_rate": 8.844245810055867e-06, + "loss": 0.2801, + "step": 10850 + }, + { + "epoch": 0.11, + "learning_rate": 8.84145251396648e-06, + "loss": 0.3501, + "step": 10875 + }, + { + "epoch": 0.11, + "learning_rate": 8.838659217877096e-06, + "loss": 0.3047, + "step": 10900 + }, + { + "epoch": 0.11, + "learning_rate": 8.83586592178771e-06, + "loss": 0.3613, + "step": 10925 + }, + { + "epoch": 0.11, + "learning_rate": 8.833072625698324e-06, + "loss": 0.3149, + "step": 10950 + }, + { + "epoch": 0.11, + "learning_rate": 8.83027932960894e-06, + "loss": 0.2902, + "step": 10975 + }, + { + "epoch": 0.11, + "learning_rate": 8.827486033519553e-06, + "loss": 0.2972, + "step": 11000 + }, + { + "epoch": 0.11, + "learning_rate": 8.824692737430169e-06, + "loss": 0.3305, + "step": 11025 + }, + { + "epoch": 0.11, + "learning_rate": 8.821899441340783e-06, + "loss": 0.3036, + "step": 11050 + }, + { + "epoch": 0.11, + "learning_rate": 8.819106145251397e-06, + "loss": 0.2998, + "step": 11075 + }, + { + "epoch": 0.11, + "learning_rate": 8.816312849162012e-06, + "loss": 0.2994, + "step": 11100 + }, + { + "epoch": 0.11, + "learning_rate": 8.813519553072628e-06, + "loss": 0.2905, + "step": 11125 + }, + { + "epoch": 0.11, + "learning_rate": 8.810726256983241e-06, + "loss": 0.3005, + "step": 11150 + }, + { + "epoch": 0.11, + "learning_rate": 8.807932960893855e-06, + "loss": 0.3063, + "step": 11175 + }, + { + "epoch": 0.11, + "learning_rate": 8.80513966480447e-06, + "loss": 0.3447, + "step": 11200 + }, + { + "epoch": 0.11, + "learning_rate": 8.802346368715085e-06, + "loss": 0.2897, + "step": 11225 + }, + { + "epoch": 0.11, + "learning_rate": 8.7995530726257e-06, + "loss": 0.3349, + "step": 11250 + }, + { + "epoch": 0.11, + "learning_rate": 8.796759776536314e-06, + "loss": 0.3262, + "step": 11275 + }, + { + "epoch": 0.11, + "learning_rate": 8.793966480446928e-06, + "loss": 0.314, + "step": 11300 + }, + { + "epoch": 0.11, + "learning_rate": 8.791173184357542e-06, + "loss": 0.3116, + "step": 11325 + }, + { + "epoch": 0.11, + "learning_rate": 8.788379888268157e-06, + "loss": 0.2874, + "step": 11350 + }, + { + "epoch": 0.11, + "learning_rate": 8.785586592178771e-06, + "loss": 0.2885, + "step": 11375 + }, + { + "epoch": 0.11, + "learning_rate": 8.782793296089385e-06, + "loss": 0.3308, + "step": 11400 + }, + { + "epoch": 0.11, + "learning_rate": 8.78e-06, + "loss": 0.3471, + "step": 11425 + }, + { + "epoch": 0.11, + "learning_rate": 8.777206703910616e-06, + "loss": 0.3322, + "step": 11450 + }, + { + "epoch": 0.12, + "learning_rate": 8.77441340782123e-06, + "loss": 0.3143, + "step": 11475 + }, + { + "epoch": 0.12, + "learning_rate": 8.771620111731844e-06, + "loss": 0.3146, + "step": 11500 + }, + { + "epoch": 0.12, + "learning_rate": 8.768826815642458e-06, + "loss": 0.3034, + "step": 11525 + }, + { + "epoch": 0.12, + "learning_rate": 8.766033519553073e-06, + "loss": 0.3306, + "step": 11550 + }, + { + "epoch": 0.12, + "learning_rate": 8.763240223463689e-06, + "loss": 0.3084, + "step": 11575 + }, + { + "epoch": 0.12, + "learning_rate": 8.760446927374303e-06, + "loss": 0.3323, + "step": 11600 + }, + { + "epoch": 0.12, + "learning_rate": 8.757653631284917e-06, + "loss": 0.3295, + "step": 11625 + }, + { + "epoch": 0.12, + "learning_rate": 8.754860335195532e-06, + "loss": 0.2913, + "step": 11650 + }, + { + "epoch": 0.12, + "learning_rate": 8.752067039106146e-06, + "loss": 0.3257, + "step": 11675 + }, + { + "epoch": 0.12, + "learning_rate": 8.749273743016761e-06, + "loss": 0.3017, + "step": 11700 + }, + { + "epoch": 0.12, + "learning_rate": 8.746480446927375e-06, + "loss": 0.2702, + "step": 11725 + }, + { + "epoch": 0.12, + "learning_rate": 8.74368715083799e-06, + "loss": 0.3075, + "step": 11750 + }, + { + "epoch": 0.12, + "learning_rate": 8.740893854748605e-06, + "loss": 0.2901, + "step": 11775 + }, + { + "epoch": 0.12, + "learning_rate": 8.738100558659219e-06, + "loss": 0.3004, + "step": 11800 + }, + { + "epoch": 0.12, + "learning_rate": 8.735307262569832e-06, + "loss": 0.2878, + "step": 11825 + }, + { + "epoch": 0.12, + "learning_rate": 8.732513966480448e-06, + "loss": 0.3208, + "step": 11850 + }, + { + "epoch": 0.12, + "learning_rate": 8.729720670391062e-06, + "loss": 0.2778, + "step": 11875 + }, + { + "epoch": 0.12, + "learning_rate": 8.726927374301677e-06, + "loss": 0.3073, + "step": 11900 + }, + { + "epoch": 0.12, + "learning_rate": 8.724134078212291e-06, + "loss": 0.2768, + "step": 11925 + }, + { + "epoch": 0.12, + "learning_rate": 8.721340782122905e-06, + "loss": 0.2819, + "step": 11950 + }, + { + "epoch": 0.12, + "learning_rate": 8.71854748603352e-06, + "loss": 0.2921, + "step": 11975 + }, + { + "epoch": 0.12, + "learning_rate": 8.715754189944134e-06, + "loss": 0.3119, + "step": 12000 + }, + { + "epoch": 0.12, + "learning_rate": 8.71296089385475e-06, + "loss": 0.3071, + "step": 12025 + }, + { + "epoch": 0.12, + "learning_rate": 8.710167597765364e-06, + "loss": 0.3107, + "step": 12050 + }, + { + "epoch": 0.12, + "learning_rate": 8.707374301675978e-06, + "loss": 0.3041, + "step": 12075 + }, + { + "epoch": 0.12, + "learning_rate": 8.704581005586593e-06, + "loss": 0.3027, + "step": 12100 + }, + { + "epoch": 0.12, + "learning_rate": 8.701787709497207e-06, + "loss": 0.2921, + "step": 12125 + }, + { + "epoch": 0.12, + "learning_rate": 8.698994413407823e-06, + "loss": 0.2605, + "step": 12150 + }, + { + "epoch": 0.12, + "learning_rate": 8.696201117318437e-06, + "loss": 0.2899, + "step": 12175 + }, + { + "epoch": 0.12, + "learning_rate": 8.69340782122905e-06, + "loss": 0.3086, + "step": 12200 + }, + { + "epoch": 0.12, + "learning_rate": 8.690614525139666e-06, + "loss": 0.2926, + "step": 12225 + }, + { + "epoch": 0.12, + "learning_rate": 8.68782122905028e-06, + "loss": 0.2943, + "step": 12250 + }, + { + "epoch": 0.12, + "learning_rate": 8.685027932960894e-06, + "loss": 0.3505, + "step": 12275 + }, + { + "epoch": 0.12, + "learning_rate": 8.68223463687151e-06, + "loss": 0.2825, + "step": 12300 + }, + { + "epoch": 0.12, + "learning_rate": 8.679441340782123e-06, + "loss": 0.3179, + "step": 12325 + }, + { + "epoch": 0.12, + "learning_rate": 8.676648044692739e-06, + "loss": 0.3298, + "step": 12350 + }, + { + "epoch": 0.12, + "learning_rate": 8.673854748603352e-06, + "loss": 0.2815, + "step": 12375 + }, + { + "epoch": 0.12, + "learning_rate": 8.671061452513966e-06, + "loss": 0.3511, + "step": 12400 + }, + { + "epoch": 0.12, + "learning_rate": 8.668268156424582e-06, + "loss": 0.3209, + "step": 12425 + }, + { + "epoch": 0.12, + "learning_rate": 8.665474860335196e-06, + "loss": 0.2993, + "step": 12450 + }, + { + "epoch": 0.13, + "learning_rate": 8.662681564245811e-06, + "loss": 0.3082, + "step": 12475 + }, + { + "epoch": 0.13, + "learning_rate": 8.659888268156425e-06, + "loss": 0.3103, + "step": 12500 + }, + { + "epoch": 0.13, + "eval_loss": 0.30788692831993103, + "eval_runtime": 1992.3986, + "eval_samples_per_second": 2.512, + "eval_steps_per_second": 0.157, + "eval_wer": 23.42960203137236, + "step": 12500 + }, + { + "epoch": 0.13, + "learning_rate": 8.657094972067039e-06, + "loss": 0.2993, + "step": 12525 + }, + { + "epoch": 0.13, + "learning_rate": 8.654301675977654e-06, + "loss": 0.2673, + "step": 12550 + }, + { + "epoch": 0.13, + "learning_rate": 8.65150837988827e-06, + "loss": 0.3003, + "step": 12575 + }, + { + "epoch": 0.13, + "learning_rate": 8.648715083798884e-06, + "loss": 0.2941, + "step": 12600 + }, + { + "epoch": 0.13, + "learning_rate": 8.645921787709498e-06, + "loss": 0.3006, + "step": 12625 + }, + { + "epoch": 0.13, + "learning_rate": 8.643128491620112e-06, + "loss": 0.3078, + "step": 12650 + }, + { + "epoch": 0.13, + "learning_rate": 8.640335195530727e-06, + "loss": 0.3108, + "step": 12675 + }, + { + "epoch": 0.13, + "learning_rate": 8.637541899441343e-06, + "loss": 0.3329, + "step": 12700 + }, + { + "epoch": 0.13, + "learning_rate": 8.634748603351955e-06, + "loss": 0.3166, + "step": 12725 + }, + { + "epoch": 0.13, + "learning_rate": 8.63195530726257e-06, + "loss": 0.2964, + "step": 12750 + }, + { + "epoch": 0.13, + "learning_rate": 8.629162011173186e-06, + "loss": 0.3357, + "step": 12775 + }, + { + "epoch": 0.13, + "learning_rate": 8.6263687150838e-06, + "loss": 0.3159, + "step": 12800 + }, + { + "epoch": 0.13, + "learning_rate": 8.623575418994414e-06, + "loss": 0.3186, + "step": 12825 + }, + { + "epoch": 0.13, + "learning_rate": 8.620782122905027e-06, + "loss": 0.3497, + "step": 12850 + }, + { + "epoch": 0.13, + "learning_rate": 8.617988826815643e-06, + "loss": 0.2811, + "step": 12875 + }, + { + "epoch": 0.13, + "learning_rate": 8.615195530726259e-06, + "loss": 0.315, + "step": 12900 + }, + { + "epoch": 0.13, + "learning_rate": 8.612402234636872e-06, + "loss": 0.2971, + "step": 12925 + }, + { + "epoch": 0.13, + "learning_rate": 8.609608938547486e-06, + "loss": 0.307, + "step": 12950 + }, + { + "epoch": 0.13, + "learning_rate": 8.606815642458102e-06, + "loss": 0.2879, + "step": 12975 + }, + { + "epoch": 0.13, + "learning_rate": 8.604022346368716e-06, + "loss": 0.2958, + "step": 13000 + }, + { + "epoch": 0.13, + "learning_rate": 8.601229050279331e-06, + "loss": 0.2966, + "step": 13025 + }, + { + "epoch": 0.13, + "learning_rate": 8.598435754189945e-06, + "loss": 0.2977, + "step": 13050 + }, + { + "epoch": 0.13, + "learning_rate": 8.595642458100559e-06, + "loss": 0.3158, + "step": 13075 + }, + { + "epoch": 0.13, + "learning_rate": 8.592849162011174e-06, + "loss": 0.2983, + "step": 13100 + }, + { + "epoch": 0.13, + "learning_rate": 8.590055865921788e-06, + "loss": 0.2791, + "step": 13125 + }, + { + "epoch": 0.13, + "learning_rate": 8.587262569832404e-06, + "loss": 0.3084, + "step": 13150 + }, + { + "epoch": 0.13, + "learning_rate": 8.584469273743018e-06, + "loss": 0.3114, + "step": 13175 + }, + { + "epoch": 0.13, + "learning_rate": 8.581675977653632e-06, + "loss": 0.2815, + "step": 13200 + }, + { + "epoch": 0.13, + "learning_rate": 8.578882681564247e-06, + "loss": 0.3106, + "step": 13225 + }, + { + "epoch": 0.13, + "learning_rate": 8.576089385474861e-06, + "loss": 0.299, + "step": 13250 + }, + { + "epoch": 0.13, + "learning_rate": 8.573296089385475e-06, + "loss": 0.314, + "step": 13275 + }, + { + "epoch": 0.13, + "learning_rate": 8.57050279329609e-06, + "loss": 0.3224, + "step": 13300 + }, + { + "epoch": 0.13, + "learning_rate": 8.567709497206704e-06, + "loss": 0.3119, + "step": 13325 + }, + { + "epoch": 0.13, + "learning_rate": 8.56491620111732e-06, + "loss": 0.3702, + "step": 13350 + }, + { + "epoch": 0.13, + "learning_rate": 8.562122905027934e-06, + "loss": 0.2737, + "step": 13375 + }, + { + "epoch": 0.13, + "learning_rate": 8.559329608938547e-06, + "loss": 0.3291, + "step": 13400 + }, + { + "epoch": 0.13, + "learning_rate": 8.556536312849163e-06, + "loss": 0.2593, + "step": 13425 + }, + { + "epoch": 0.13, + "learning_rate": 8.553743016759777e-06, + "loss": 0.2511, + "step": 13450 + }, + { + "epoch": 0.14, + "learning_rate": 8.550949720670392e-06, + "loss": 0.2917, + "step": 13475 + }, + { + "epoch": 0.14, + "learning_rate": 8.548268156424581e-06, + "loss": 0.3156, + "step": 13500 + }, + { + "epoch": 0.14, + "learning_rate": 8.545474860335197e-06, + "loss": 0.2705, + "step": 13525 + }, + { + "epoch": 0.14, + "learning_rate": 8.54268156424581e-06, + "loss": 0.2826, + "step": 13550 + }, + { + "epoch": 0.14, + "learning_rate": 8.539888268156426e-06, + "loss": 0.2864, + "step": 13575 + }, + { + "epoch": 0.14, + "learning_rate": 8.53709497206704e-06, + "loss": 0.3085, + "step": 13600 + }, + { + "epoch": 0.14, + "learning_rate": 8.534301675977654e-06, + "loss": 0.326, + "step": 13625 + }, + { + "epoch": 0.14, + "learning_rate": 8.53150837988827e-06, + "loss": 0.2829, + "step": 13650 + }, + { + "epoch": 0.14, + "learning_rate": 8.528715083798883e-06, + "loss": 0.3028, + "step": 13675 + }, + { + "epoch": 0.14, + "learning_rate": 8.525921787709497e-06, + "loss": 0.2901, + "step": 13700 + }, + { + "epoch": 0.14, + "learning_rate": 8.523240223463688e-06, + "loss": 0.3134, + "step": 13725 + }, + { + "epoch": 0.14, + "learning_rate": 8.520446927374303e-06, + "loss": 0.302, + "step": 13750 + }, + { + "epoch": 0.14, + "learning_rate": 8.517653631284917e-06, + "loss": 0.2903, + "step": 13775 + }, + { + "epoch": 0.14, + "learning_rate": 8.514860335195531e-06, + "loss": 0.3192, + "step": 13800 + }, + { + "epoch": 0.14, + "learning_rate": 8.512067039106145e-06, + "loss": 0.3075, + "step": 13825 + }, + { + "epoch": 0.14, + "learning_rate": 8.50927374301676e-06, + "loss": 0.3314, + "step": 13850 + }, + { + "epoch": 0.14, + "learning_rate": 8.506480446927376e-06, + "loss": 0.3004, + "step": 13875 + }, + { + "epoch": 0.14, + "learning_rate": 8.50368715083799e-06, + "loss": 0.3236, + "step": 13900 + }, + { + "epoch": 0.14, + "learning_rate": 8.500893854748604e-06, + "loss": 0.2836, + "step": 13925 + }, + { + "epoch": 0.14, + "learning_rate": 8.49810055865922e-06, + "loss": 0.3133, + "step": 13950 + }, + { + "epoch": 0.14, + "learning_rate": 8.495307262569833e-06, + "loss": 0.2978, + "step": 13975 + }, + { + "epoch": 0.14, + "learning_rate": 8.492513966480449e-06, + "loss": 0.334, + "step": 14000 + }, + { + "epoch": 0.14, + "learning_rate": 8.489720670391062e-06, + "loss": 0.3161, + "step": 14025 + }, + { + "epoch": 0.14, + "learning_rate": 8.486927374301676e-06, + "loss": 0.2711, + "step": 14050 + }, + { + "epoch": 0.14, + "learning_rate": 8.484134078212292e-06, + "loss": 0.2785, + "step": 14075 + }, + { + "epoch": 0.14, + "learning_rate": 8.481340782122906e-06, + "loss": 0.3238, + "step": 14100 + }, + { + "epoch": 0.14, + "learning_rate": 8.478547486033521e-06, + "loss": 0.272, + "step": 14125 + }, + { + "epoch": 0.14, + "learning_rate": 8.475754189944135e-06, + "loss": 0.2765, + "step": 14150 + }, + { + "epoch": 0.14, + "learning_rate": 8.472960893854749e-06, + "loss": 0.2938, + "step": 14175 + }, + { + "epoch": 0.14, + "learning_rate": 8.470167597765364e-06, + "loss": 0.337, + "step": 14200 + }, + { + "epoch": 0.14, + "learning_rate": 8.467374301675978e-06, + "loss": 0.2747, + "step": 14225 + }, + { + "epoch": 0.14, + "learning_rate": 8.464581005586592e-06, + "loss": 0.3116, + "step": 14250 + }, + { + "epoch": 0.14, + "learning_rate": 8.461787709497208e-06, + "loss": 0.3087, + "step": 14275 + }, + { + "epoch": 0.14, + "learning_rate": 8.458994413407822e-06, + "loss": 0.3135, + "step": 14300 + }, + { + "epoch": 0.14, + "learning_rate": 8.456201117318437e-06, + "loss": 0.2814, + "step": 14325 + }, + { + "epoch": 0.14, + "learning_rate": 8.453407821229051e-06, + "loss": 0.3104, + "step": 14350 + }, + { + "epoch": 0.14, + "learning_rate": 8.450614525139665e-06, + "loss": 0.2835, + "step": 14375 + }, + { + "epoch": 0.14, + "learning_rate": 8.44782122905028e-06, + "loss": 0.2804, + "step": 14400 + }, + { + "epoch": 0.14, + "learning_rate": 8.445027932960894e-06, + "loss": 0.2999, + "step": 14425 + }, + { + "epoch": 0.14, + "learning_rate": 8.44223463687151e-06, + "loss": 0.2989, + "step": 14450 + }, + { + "epoch": 0.15, + "learning_rate": 8.439441340782124e-06, + "loss": 0.3174, + "step": 14475 + }, + { + "epoch": 0.15, + "learning_rate": 8.436648044692737e-06, + "loss": 0.2969, + "step": 14500 + }, + { + "epoch": 0.15, + "learning_rate": 8.433854748603353e-06, + "loss": 0.3178, + "step": 14525 + }, + { + "epoch": 0.15, + "learning_rate": 8.431061452513967e-06, + "loss": 0.2931, + "step": 14550 + }, + { + "epoch": 0.15, + "learning_rate": 8.428268156424582e-06, + "loss": 0.326, + "step": 14575 + }, + { + "epoch": 0.15, + "learning_rate": 8.425474860335196e-06, + "loss": 0.2643, + "step": 14600 + }, + { + "epoch": 0.15, + "learning_rate": 8.42268156424581e-06, + "loss": 0.2951, + "step": 14625 + }, + { + "epoch": 0.15, + "learning_rate": 8.419888268156426e-06, + "loss": 0.3197, + "step": 14650 + }, + { + "epoch": 0.15, + "learning_rate": 8.41709497206704e-06, + "loss": 0.3097, + "step": 14675 + }, + { + "epoch": 0.15, + "learning_rate": 8.414301675977653e-06, + "loss": 0.3232, + "step": 14700 + }, + { + "epoch": 0.15, + "learning_rate": 8.411508379888269e-06, + "loss": 0.295, + "step": 14725 + }, + { + "epoch": 0.15, + "learning_rate": 8.408715083798883e-06, + "loss": 0.3089, + "step": 14750 + }, + { + "epoch": 0.15, + "learning_rate": 8.405921787709498e-06, + "loss": 0.3033, + "step": 14775 + }, + { + "epoch": 0.15, + "learning_rate": 8.403128491620112e-06, + "loss": 0.2882, + "step": 14800 + }, + { + "epoch": 0.15, + "learning_rate": 8.400335195530726e-06, + "loss": 0.3307, + "step": 14825 + }, + { + "epoch": 0.15, + "learning_rate": 8.397541899441342e-06, + "loss": 0.2912, + "step": 14850 + }, + { + "epoch": 0.15, + "learning_rate": 8.394748603351957e-06, + "loss": 0.3066, + "step": 14875 + }, + { + "epoch": 0.15, + "learning_rate": 8.391955307262571e-06, + "loss": 0.2995, + "step": 14900 + }, + { + "epoch": 0.15, + "learning_rate": 8.389162011173185e-06, + "loss": 0.3067, + "step": 14925 + }, + { + "epoch": 0.15, + "learning_rate": 8.386368715083799e-06, + "loss": 0.3014, + "step": 14950 + }, + { + "epoch": 0.15, + "learning_rate": 8.383575418994414e-06, + "loss": 0.3218, + "step": 14975 + }, + { + "epoch": 0.15, + "learning_rate": 8.38078212290503e-06, + "loss": 0.3041, + "step": 15000 + }, + { + "epoch": 0.15, + "eval_loss": 0.303312212228775, + "eval_runtime": 1965.6467, + "eval_samples_per_second": 2.546, + "eval_steps_per_second": 0.159, + "eval_wer": 23.211276964332328, + "step": 15000 + }, + { + "epoch": 0.15, + "learning_rate": 8.377988826815644e-06, + "loss": 0.3262, + "step": 15025 + }, + { + "epoch": 0.15, + "learning_rate": 8.375195530726257e-06, + "loss": 0.3196, + "step": 15050 + }, + { + "epoch": 0.15, + "learning_rate": 8.372402234636873e-06, + "loss": 0.274, + "step": 15075 + }, + { + "epoch": 0.15, + "learning_rate": 8.369608938547487e-06, + "loss": 0.3029, + "step": 15100 + }, + { + "epoch": 0.15, + "learning_rate": 8.3668156424581e-06, + "loss": 0.2883, + "step": 15125 + }, + { + "epoch": 0.15, + "learning_rate": 8.364022346368715e-06, + "loss": 0.2914, + "step": 15150 + }, + { + "epoch": 0.15, + "learning_rate": 8.36122905027933e-06, + "loss": 0.3186, + "step": 15175 + }, + { + "epoch": 0.15, + "learning_rate": 8.358435754189946e-06, + "loss": 0.3596, + "step": 15200 + }, + { + "epoch": 0.15, + "learning_rate": 8.35564245810056e-06, + "loss": 0.3164, + "step": 15225 + }, + { + "epoch": 0.15, + "learning_rate": 8.352849162011173e-06, + "loss": 0.2753, + "step": 15250 + }, + { + "epoch": 0.15, + "learning_rate": 8.350055865921787e-06, + "loss": 0.2767, + "step": 15275 + }, + { + "epoch": 0.15, + "learning_rate": 8.347262569832403e-06, + "loss": 0.3053, + "step": 15300 + }, + { + "epoch": 0.15, + "learning_rate": 8.344469273743018e-06, + "loss": 0.2999, + "step": 15325 + }, + { + "epoch": 0.15, + "learning_rate": 8.341675977653632e-06, + "loss": 0.2854, + "step": 15350 + }, + { + "epoch": 0.15, + "learning_rate": 8.338882681564246e-06, + "loss": 0.3122, + "step": 15375 + }, + { + "epoch": 0.15, + "learning_rate": 8.336089385474862e-06, + "loss": 0.2993, + "step": 15400 + }, + { + "epoch": 0.15, + "learning_rate": 8.333296089385475e-06, + "loss": 0.2938, + "step": 15425 + }, + { + "epoch": 0.15, + "learning_rate": 8.330502793296091e-06, + "loss": 0.3102, + "step": 15450 + }, + { + "epoch": 0.16, + "learning_rate": 8.327709497206705e-06, + "loss": 0.2904, + "step": 15475 + }, + { + "epoch": 0.16, + "learning_rate": 8.324916201117319e-06, + "loss": 0.2854, + "step": 15500 + }, + { + "epoch": 0.16, + "learning_rate": 8.322122905027934e-06, + "loss": 0.3141, + "step": 15525 + }, + { + "epoch": 0.16, + "learning_rate": 8.319329608938548e-06, + "loss": 0.2863, + "step": 15550 + }, + { + "epoch": 0.16, + "learning_rate": 8.316536312849164e-06, + "loss": 0.345, + "step": 15575 + }, + { + "epoch": 0.16, + "learning_rate": 8.313743016759777e-06, + "loss": 0.3363, + "step": 15600 + }, + { + "epoch": 0.16, + "learning_rate": 8.310949720670391e-06, + "loss": 0.3084, + "step": 15625 + }, + { + "epoch": 0.16, + "learning_rate": 8.308156424581007e-06, + "loss": 0.3029, + "step": 15650 + }, + { + "epoch": 0.16, + "learning_rate": 8.30536312849162e-06, + "loss": 0.3159, + "step": 15675 + }, + { + "epoch": 0.16, + "learning_rate": 8.302569832402235e-06, + "loss": 0.3164, + "step": 15700 + }, + { + "epoch": 0.16, + "learning_rate": 8.29977653631285e-06, + "loss": 0.3086, + "step": 15725 + }, + { + "epoch": 0.16, + "learning_rate": 8.296983240223464e-06, + "loss": 0.2947, + "step": 15750 + }, + { + "epoch": 0.16, + "learning_rate": 8.29418994413408e-06, + "loss": 0.3027, + "step": 15775 + }, + { + "epoch": 0.16, + "learning_rate": 8.291396648044693e-06, + "loss": 0.2759, + "step": 15800 + }, + { + "epoch": 0.16, + "learning_rate": 8.288603351955307e-06, + "loss": 0.2862, + "step": 15825 + }, + { + "epoch": 0.16, + "learning_rate": 8.285810055865923e-06, + "loss": 0.3359, + "step": 15850 + }, + { + "epoch": 0.16, + "learning_rate": 8.283016759776537e-06, + "loss": 0.3034, + "step": 15875 + }, + { + "epoch": 0.16, + "learning_rate": 8.280223463687152e-06, + "loss": 0.2885, + "step": 15900 + }, + { + "epoch": 0.16, + "learning_rate": 8.277430167597766e-06, + "loss": 0.3088, + "step": 15925 + }, + { + "epoch": 0.16, + "learning_rate": 8.27463687150838e-06, + "loss": 0.3001, + "step": 15950 + }, + { + "epoch": 0.16, + "learning_rate": 8.271843575418995e-06, + "loss": 0.29, + "step": 15975 + }, + { + "epoch": 0.16, + "learning_rate": 8.269050279329611e-06, + "loss": 0.2975, + "step": 16000 + }, + { + "epoch": 0.16, + "learning_rate": 8.266256983240225e-06, + "loss": 0.3203, + "step": 16025 + }, + { + "epoch": 0.16, + "learning_rate": 8.263463687150839e-06, + "loss": 0.2684, + "step": 16050 + }, + { + "epoch": 0.16, + "learning_rate": 8.260670391061452e-06, + "loss": 0.2704, + "step": 16075 + }, + { + "epoch": 0.16, + "learning_rate": 8.257877094972068e-06, + "loss": 0.3068, + "step": 16100 + }, + { + "epoch": 0.16, + "learning_rate": 8.255083798882682e-06, + "loss": 0.3122, + "step": 16125 + }, + { + "epoch": 0.16, + "learning_rate": 8.252290502793296e-06, + "loss": 0.3004, + "step": 16150 + }, + { + "epoch": 0.16, + "learning_rate": 8.249497206703911e-06, + "loss": 0.3121, + "step": 16175 + }, + { + "epoch": 0.16, + "learning_rate": 8.246703910614527e-06, + "loss": 0.2894, + "step": 16200 + }, + { + "epoch": 0.16, + "learning_rate": 8.24391061452514e-06, + "loss": 0.2744, + "step": 16225 + }, + { + "epoch": 0.16, + "learning_rate": 8.241117318435755e-06, + "loss": 0.2918, + "step": 16250 + }, + { + "epoch": 0.16, + "learning_rate": 8.238324022346368e-06, + "loss": 0.3032, + "step": 16275 + }, + { + "epoch": 0.16, + "learning_rate": 8.235530726256984e-06, + "loss": 0.2526, + "step": 16300 + }, + { + "epoch": 0.16, + "learning_rate": 8.2327374301676e-06, + "loss": 0.2634, + "step": 16325 + }, + { + "epoch": 0.16, + "learning_rate": 8.229944134078213e-06, + "loss": 0.3097, + "step": 16350 + }, + { + "epoch": 0.16, + "learning_rate": 8.227150837988827e-06, + "loss": 0.2896, + "step": 16375 + }, + { + "epoch": 0.16, + "learning_rate": 8.224357541899441e-06, + "loss": 0.2734, + "step": 16400 + }, + { + "epoch": 0.16, + "learning_rate": 8.221564245810057e-06, + "loss": 0.2945, + "step": 16425 + }, + { + "epoch": 0.16, + "learning_rate": 8.218770949720672e-06, + "loss": 0.2981, + "step": 16450 + }, + { + "epoch": 0.17, + "learning_rate": 8.215977653631286e-06, + "loss": 0.293, + "step": 16475 + }, + { + "epoch": 0.17, + "learning_rate": 8.2131843575419e-06, + "loss": 0.2814, + "step": 16500 + }, + { + "epoch": 0.17, + "learning_rate": 8.210391061452515e-06, + "loss": 0.3024, + "step": 16525 + }, + { + "epoch": 0.17, + "learning_rate": 8.20759776536313e-06, + "loss": 0.2682, + "step": 16550 + }, + { + "epoch": 0.17, + "learning_rate": 8.204804469273743e-06, + "loss": 0.3134, + "step": 16575 + }, + { + "epoch": 0.17, + "learning_rate": 8.202011173184357e-06, + "loss": 0.3174, + "step": 16600 + }, + { + "epoch": 0.17, + "learning_rate": 8.199217877094972e-06, + "loss": 0.3177, + "step": 16625 + }, + { + "epoch": 0.17, + "learning_rate": 8.196424581005588e-06, + "loss": 0.3074, + "step": 16650 + }, + { + "epoch": 0.17, + "learning_rate": 8.193631284916202e-06, + "loss": 0.2895, + "step": 16675 + }, + { + "epoch": 0.17, + "learning_rate": 8.190837988826816e-06, + "loss": 0.252, + "step": 16700 + }, + { + "epoch": 0.17, + "learning_rate": 8.188044692737431e-06, + "loss": 0.3087, + "step": 16725 + }, + { + "epoch": 0.17, + "learning_rate": 8.185251396648045e-06, + "loss": 0.2436, + "step": 16750 + }, + { + "epoch": 0.17, + "learning_rate": 8.18245810055866e-06, + "loss": 0.2949, + "step": 16775 + }, + { + "epoch": 0.17, + "learning_rate": 8.179664804469274e-06, + "loss": 0.2923, + "step": 16800 + }, + { + "epoch": 0.17, + "learning_rate": 8.176871508379888e-06, + "loss": 0.3105, + "step": 16825 + }, + { + "epoch": 0.17, + "learning_rate": 8.174078212290504e-06, + "loss": 0.3077, + "step": 16850 + }, + { + "epoch": 0.17, + "learning_rate": 8.171284916201118e-06, + "loss": 0.2723, + "step": 16875 + }, + { + "epoch": 0.17, + "learning_rate": 8.168491620111733e-06, + "loss": 0.3138, + "step": 16900 + }, + { + "epoch": 0.17, + "learning_rate": 8.165698324022347e-06, + "loss": 0.3269, + "step": 16925 + }, + { + "epoch": 0.17, + "learning_rate": 8.162905027932961e-06, + "loss": 0.3129, + "step": 16950 + }, + { + "epoch": 0.17, + "learning_rate": 8.160111731843577e-06, + "loss": 0.3319, + "step": 16975 + }, + { + "epoch": 0.17, + "learning_rate": 8.15731843575419e-06, + "loss": 0.2923, + "step": 17000 + }, + { + "epoch": 0.17, + "learning_rate": 8.154525139664804e-06, + "loss": 0.3022, + "step": 17025 + }, + { + "epoch": 0.17, + "learning_rate": 8.15173184357542e-06, + "loss": 0.2951, + "step": 17050 + }, + { + "epoch": 0.17, + "learning_rate": 8.148938547486034e-06, + "loss": 0.2674, + "step": 17075 + }, + { + "epoch": 0.17, + "learning_rate": 8.14614525139665e-06, + "loss": 0.2988, + "step": 17100 + }, + { + "epoch": 0.17, + "learning_rate": 8.143351955307263e-06, + "loss": 0.2918, + "step": 17125 + }, + { + "epoch": 0.17, + "learning_rate": 8.140558659217877e-06, + "loss": 0.2862, + "step": 17150 + }, + { + "epoch": 0.17, + "learning_rate": 8.137765363128492e-06, + "loss": 0.2902, + "step": 17175 + }, + { + "epoch": 0.17, + "learning_rate": 8.134972067039106e-06, + "loss": 0.3034, + "step": 17200 + }, + { + "epoch": 0.17, + "learning_rate": 8.132178770949722e-06, + "loss": 0.2693, + "step": 17225 + }, + { + "epoch": 0.17, + "learning_rate": 8.129385474860336e-06, + "loss": 0.2604, + "step": 17250 + }, + { + "epoch": 0.17, + "learning_rate": 8.12659217877095e-06, + "loss": 0.3255, + "step": 17275 + }, + { + "epoch": 0.17, + "learning_rate": 8.123798882681565e-06, + "loss": 0.3016, + "step": 17300 + }, + { + "epoch": 0.17, + "learning_rate": 8.12100558659218e-06, + "loss": 0.2923, + "step": 17325 + }, + { + "epoch": 0.17, + "learning_rate": 8.118212290502794e-06, + "loss": 0.2911, + "step": 17350 + }, + { + "epoch": 0.17, + "learning_rate": 8.115418994413408e-06, + "loss": 0.3096, + "step": 17375 + }, + { + "epoch": 0.17, + "learning_rate": 8.112625698324022e-06, + "loss": 0.2944, + "step": 17400 + }, + { + "epoch": 0.17, + "learning_rate": 8.109832402234638e-06, + "loss": 0.3034, + "step": 17425 + }, + { + "epoch": 0.17, + "learning_rate": 8.107039106145253e-06, + "loss": 0.2685, + "step": 17450 + }, + { + "epoch": 0.18, + "learning_rate": 8.104245810055867e-06, + "loss": 0.322, + "step": 17475 + }, + { + "epoch": 0.18, + "learning_rate": 8.101452513966481e-06, + "loss": 0.3077, + "step": 17500 + }, + { + "epoch": 0.18, + "eval_loss": 0.29980674386024475, + "eval_runtime": 1882.7047, + "eval_samples_per_second": 2.658, + "eval_steps_per_second": 0.166, + "eval_wer": 21.709105579154702, + "step": 17500 + }, + { + "epoch": 0.18, + "learning_rate": 8.098659217877095e-06, + "loss": 0.3134, + "step": 17525 + }, + { + "epoch": 0.18, + "learning_rate": 8.09586592178771e-06, + "loss": 0.2648, + "step": 17550 + }, + { + "epoch": 0.18, + "learning_rate": 8.093072625698324e-06, + "loss": 0.2645, + "step": 17575 + }, + { + "epoch": 0.18, + "learning_rate": 8.090279329608938e-06, + "loss": 0.293, + "step": 17600 + }, + { + "epoch": 0.18, + "learning_rate": 8.087486033519554e-06, + "loss": 0.2645, + "step": 17625 + }, + { + "epoch": 0.18, + "learning_rate": 8.08469273743017e-06, + "loss": 0.2967, + "step": 17650 + }, + { + "epoch": 0.18, + "learning_rate": 8.081899441340783e-06, + "loss": 0.2914, + "step": 17675 + }, + { + "epoch": 0.18, + "learning_rate": 8.079106145251397e-06, + "loss": 0.286, + "step": 17700 + }, + { + "epoch": 0.18, + "learning_rate": 8.07631284916201e-06, + "loss": 0.294, + "step": 17725 + }, + { + "epoch": 0.18, + "learning_rate": 8.073519553072626e-06, + "loss": 0.2902, + "step": 17750 + }, + { + "epoch": 0.18, + "learning_rate": 8.070726256983242e-06, + "loss": 0.3132, + "step": 17775 + }, + { + "epoch": 0.18, + "learning_rate": 8.067932960893856e-06, + "loss": 0.3193, + "step": 17800 + }, + { + "epoch": 0.18, + "learning_rate": 8.06513966480447e-06, + "loss": 0.2573, + "step": 17825 + }, + { + "epoch": 0.18, + "learning_rate": 8.062346368715085e-06, + "loss": 0.2911, + "step": 17850 + }, + { + "epoch": 0.18, + "learning_rate": 8.059553072625699e-06, + "loss": 0.2882, + "step": 17875 + }, + { + "epoch": 0.18, + "learning_rate": 8.056759776536314e-06, + "loss": 0.2684, + "step": 17900 + }, + { + "epoch": 0.18, + "learning_rate": 8.053966480446928e-06, + "loss": 0.3059, + "step": 17925 + }, + { + "epoch": 0.18, + "learning_rate": 8.051284916201117e-06, + "loss": 0.3025, + "step": 17950 + }, + { + "epoch": 0.18, + "learning_rate": 8.048491620111733e-06, + "loss": 0.278, + "step": 17975 + }, + { + "epoch": 0.18, + "learning_rate": 8.045698324022347e-06, + "loss": 0.2978, + "step": 18000 + }, + { + "epoch": 0.18, + "learning_rate": 8.04290502793296e-06, + "loss": 0.2887, + "step": 18025 + }, + { + "epoch": 0.18, + "learning_rate": 8.040111731843576e-06, + "loss": 0.2795, + "step": 18050 + }, + { + "epoch": 0.18, + "learning_rate": 8.037318435754192e-06, + "loss": 0.2991, + "step": 18075 + }, + { + "epoch": 0.18, + "learning_rate": 8.034525139664805e-06, + "loss": 0.2733, + "step": 18100 + }, + { + "epoch": 0.18, + "learning_rate": 8.03173184357542e-06, + "loss": 0.3165, + "step": 18125 + }, + { + "epoch": 0.18, + "learning_rate": 8.028938547486033e-06, + "loss": 0.2888, + "step": 18150 + }, + { + "epoch": 0.18, + "learning_rate": 8.026145251396649e-06, + "loss": 0.2808, + "step": 18175 + }, + { + "epoch": 0.18, + "learning_rate": 8.023351955307264e-06, + "loss": 0.2982, + "step": 18200 + }, + { + "epoch": 0.18, + "learning_rate": 8.020558659217878e-06, + "loss": 0.3258, + "step": 18225 + }, + { + "epoch": 0.18, + "learning_rate": 8.017765363128492e-06, + "loss": 0.3045, + "step": 18250 + }, + { + "epoch": 0.18, + "learning_rate": 8.014972067039107e-06, + "loss": 0.2899, + "step": 18275 + }, + { + "epoch": 0.18, + "learning_rate": 8.012178770949721e-06, + "loss": 0.3117, + "step": 18300 + }, + { + "epoch": 0.18, + "learning_rate": 8.009385474860337e-06, + "loss": 0.304, + "step": 18325 + }, + { + "epoch": 0.18, + "learning_rate": 8.00659217877095e-06, + "loss": 0.302, + "step": 18350 + }, + { + "epoch": 0.18, + "learning_rate": 8.003798882681565e-06, + "loss": 0.2899, + "step": 18375 + }, + { + "epoch": 0.18, + "learning_rate": 8.00100558659218e-06, + "loss": 0.2875, + "step": 18400 + }, + { + "epoch": 0.18, + "learning_rate": 7.998212290502794e-06, + "loss": 0.3105, + "step": 18425 + }, + { + "epoch": 0.18, + "learning_rate": 7.99541899441341e-06, + "loss": 0.298, + "step": 18450 + }, + { + "epoch": 0.19, + "learning_rate": 7.992625698324022e-06, + "loss": 0.306, + "step": 18475 + }, + { + "epoch": 0.19, + "learning_rate": 7.989832402234637e-06, + "loss": 0.3143, + "step": 18500 + }, + { + "epoch": 0.19, + "learning_rate": 7.987039106145253e-06, + "loss": 0.2671, + "step": 18525 + }, + { + "epoch": 0.19, + "learning_rate": 7.984245810055867e-06, + "loss": 0.2854, + "step": 18550 + }, + { + "epoch": 0.19, + "learning_rate": 7.98145251396648e-06, + "loss": 0.2786, + "step": 18575 + }, + { + "epoch": 0.19, + "learning_rate": 7.978659217877096e-06, + "loss": 0.2808, + "step": 18600 + }, + { + "epoch": 0.19, + "learning_rate": 7.97586592178771e-06, + "loss": 0.2853, + "step": 18625 + }, + { + "epoch": 0.19, + "learning_rate": 7.973072625698325e-06, + "loss": 0.3062, + "step": 18650 + }, + { + "epoch": 0.19, + "learning_rate": 7.97027932960894e-06, + "loss": 0.2828, + "step": 18675 + }, + { + "epoch": 0.19, + "learning_rate": 7.967486033519553e-06, + "loss": 0.2966, + "step": 18700 + }, + { + "epoch": 0.19, + "learning_rate": 7.964692737430169e-06, + "loss": 0.2787, + "step": 18725 + }, + { + "epoch": 0.19, + "learning_rate": 7.961899441340782e-06, + "loss": 0.2853, + "step": 18750 + }, + { + "epoch": 0.19, + "learning_rate": 7.959106145251398e-06, + "loss": 0.2844, + "step": 18775 + }, + { + "epoch": 0.19, + "learning_rate": 7.956312849162012e-06, + "loss": 0.282, + "step": 18800 + }, + { + "epoch": 0.19, + "learning_rate": 7.953519553072626e-06, + "loss": 0.2994, + "step": 18825 + }, + { + "epoch": 0.19, + "learning_rate": 7.950726256983241e-06, + "loss": 0.3017, + "step": 18850 + }, + { + "epoch": 0.19, + "learning_rate": 7.947932960893855e-06, + "loss": 0.2652, + "step": 18875 + }, + { + "epoch": 0.19, + "learning_rate": 7.94513966480447e-06, + "loss": 0.2799, + "step": 18900 + }, + { + "epoch": 0.19, + "learning_rate": 7.942346368715085e-06, + "loss": 0.3076, + "step": 18925 + }, + { + "epoch": 0.19, + "learning_rate": 7.939553072625698e-06, + "loss": 0.2935, + "step": 18950 + }, + { + "epoch": 0.19, + "learning_rate": 7.936759776536314e-06, + "loss": 0.3062, + "step": 18975 + }, + { + "epoch": 0.19, + "learning_rate": 7.933966480446928e-06, + "loss": 0.2924, + "step": 19000 + }, + { + "epoch": 0.19, + "learning_rate": 7.931173184357542e-06, + "loss": 0.2917, + "step": 19025 + }, + { + "epoch": 0.19, + "learning_rate": 7.928379888268157e-06, + "loss": 0.2713, + "step": 19050 + }, + { + "epoch": 0.19, + "learning_rate": 7.925586592178771e-06, + "loss": 0.3005, + "step": 19075 + }, + { + "epoch": 0.19, + "learning_rate": 7.922793296089387e-06, + "loss": 0.3162, + "step": 19100 + }, + { + "epoch": 0.19, + "learning_rate": 7.92e-06, + "loss": 0.2537, + "step": 19125 + }, + { + "epoch": 0.19, + "learning_rate": 7.917206703910614e-06, + "loss": 0.2903, + "step": 19150 + }, + { + "epoch": 0.19, + "learning_rate": 7.91441340782123e-06, + "loss": 0.2932, + "step": 19175 + }, + { + "epoch": 0.19, + "learning_rate": 7.911620111731845e-06, + "loss": 0.3054, + "step": 19200 + }, + { + "epoch": 0.19, + "learning_rate": 7.90882681564246e-06, + "loss": 0.2779, + "step": 19225 + }, + { + "epoch": 0.19, + "learning_rate": 7.906033519553073e-06, + "loss": 0.317, + "step": 19250 + }, + { + "epoch": 0.19, + "learning_rate": 7.903240223463687e-06, + "loss": 0.2724, + "step": 19275 + }, + { + "epoch": 0.19, + "learning_rate": 7.900446927374302e-06, + "loss": 0.2802, + "step": 19300 + }, + { + "epoch": 0.19, + "learning_rate": 7.897653631284918e-06, + "loss": 0.3068, + "step": 19325 + }, + { + "epoch": 0.19, + "learning_rate": 7.894860335195532e-06, + "loss": 0.3022, + "step": 19350 + }, + { + "epoch": 0.19, + "learning_rate": 7.892067039106146e-06, + "loss": 0.2838, + "step": 19375 + }, + { + "epoch": 0.19, + "learning_rate": 7.889273743016761e-06, + "loss": 0.2696, + "step": 19400 + }, + { + "epoch": 0.19, + "learning_rate": 7.886480446927375e-06, + "loss": 0.2813, + "step": 19425 + }, + { + "epoch": 0.19, + "learning_rate": 7.883687150837989e-06, + "loss": 0.2808, + "step": 19450 + }, + { + "epoch": 0.2, + "learning_rate": 7.880893854748603e-06, + "loss": 0.2822, + "step": 19475 + }, + { + "epoch": 0.2, + "learning_rate": 7.878100558659218e-06, + "loss": 0.2923, + "step": 19500 + }, + { + "epoch": 0.2, + "learning_rate": 7.875307262569834e-06, + "loss": 0.2835, + "step": 19525 + }, + { + "epoch": 0.2, + "learning_rate": 7.872513966480448e-06, + "loss": 0.3168, + "step": 19550 + }, + { + "epoch": 0.2, + "learning_rate": 7.869720670391062e-06, + "loss": 0.2682, + "step": 19575 + }, + { + "epoch": 0.2, + "learning_rate": 7.866927374301675e-06, + "loss": 0.2593, + "step": 19600 + }, + { + "epoch": 0.2, + "learning_rate": 7.864134078212291e-06, + "loss": 0.2558, + "step": 19625 + }, + { + "epoch": 0.2, + "learning_rate": 7.861340782122907e-06, + "loss": 0.2639, + "step": 19650 + }, + { + "epoch": 0.2, + "learning_rate": 7.85854748603352e-06, + "loss": 0.3232, + "step": 19675 + }, + { + "epoch": 0.2, + "learning_rate": 7.855754189944134e-06, + "loss": 0.2657, + "step": 19700 + }, + { + "epoch": 0.2, + "learning_rate": 7.85296089385475e-06, + "loss": 0.264, + "step": 19725 + }, + { + "epoch": 0.2, + "learning_rate": 7.850167597765364e-06, + "loss": 0.2612, + "step": 19750 + }, + { + "epoch": 0.2, + "learning_rate": 7.84737430167598e-06, + "loss": 0.2847, + "step": 19775 + }, + { + "epoch": 0.2, + "learning_rate": 7.844581005586593e-06, + "loss": 0.277, + "step": 19800 + }, + { + "epoch": 0.2, + "learning_rate": 7.841787709497207e-06, + "loss": 0.2768, + "step": 19825 + }, + { + "epoch": 0.2, + "learning_rate": 7.838994413407822e-06, + "loss": 0.288, + "step": 19850 + }, + { + "epoch": 0.2, + "learning_rate": 7.836201117318436e-06, + "loss": 0.2881, + "step": 19875 + }, + { + "epoch": 0.2, + "learning_rate": 7.83340782122905e-06, + "loss": 0.3224, + "step": 19900 + }, + { + "epoch": 0.2, + "learning_rate": 7.830614525139666e-06, + "loss": 0.3185, + "step": 19925 + }, + { + "epoch": 0.2, + "learning_rate": 7.82782122905028e-06, + "loss": 0.2736, + "step": 19950 + }, + { + "epoch": 0.2, + "learning_rate": 7.825027932960895e-06, + "loss": 0.3149, + "step": 19975 + }, + { + "epoch": 0.2, + "learning_rate": 7.822346368715084e-06, + "loss": 0.2867, + "step": 20000 + }, + { + "epoch": 0.2, + "eval_loss": 0.2943369448184967, + "eval_runtime": 1824.6732, + "eval_samples_per_second": 2.743, + "eval_steps_per_second": 0.172, + "eval_wer": 20.1760839127649, + "step": 20000 + }, + { + "epoch": 0.2, + "learning_rate": 7.819553072625698e-06, + "loss": 0.3079, + "step": 20025 + }, + { + "epoch": 0.2, + "learning_rate": 7.816759776536313e-06, + "loss": 0.2947, + "step": 20050 + }, + { + "epoch": 0.2, + "learning_rate": 7.813966480446929e-06, + "loss": 0.3144, + "step": 20075 + }, + { + "epoch": 0.2, + "learning_rate": 7.811173184357543e-06, + "loss": 0.2547, + "step": 20100 + }, + { + "epoch": 0.2, + "learning_rate": 7.808379888268157e-06, + "loss": 0.2933, + "step": 20125 + }, + { + "epoch": 0.2, + "learning_rate": 7.805586592178772e-06, + "loss": 0.2718, + "step": 20150 + }, + { + "epoch": 0.2, + "learning_rate": 7.802793296089386e-06, + "loss": 0.3352, + "step": 20175 + }, + { + "epoch": 0.2, + "learning_rate": 7.800000000000002e-06, + "loss": 0.3105, + "step": 20200 + }, + { + "epoch": 0.2, + "learning_rate": 7.797206703910615e-06, + "loss": 0.2733, + "step": 20225 + }, + { + "epoch": 0.2, + "learning_rate": 7.79441340782123e-06, + "loss": 0.2908, + "step": 20250 + }, + { + "epoch": 0.2, + "learning_rate": 7.791620111731845e-06, + "loss": 0.293, + "step": 20275 + }, + { + "epoch": 0.2, + "learning_rate": 7.788826815642459e-06, + "loss": 0.3129, + "step": 20300 + }, + { + "epoch": 0.2, + "learning_rate": 7.786033519553074e-06, + "loss": 0.2922, + "step": 20325 + }, + { + "epoch": 0.2, + "learning_rate": 7.783240223463688e-06, + "loss": 0.2763, + "step": 20350 + }, + { + "epoch": 0.2, + "learning_rate": 7.780446927374302e-06, + "loss": 0.3337, + "step": 20375 + }, + { + "epoch": 0.2, + "learning_rate": 7.777653631284917e-06, + "loss": 0.2925, + "step": 20400 + }, + { + "epoch": 0.2, + "learning_rate": 7.774860335195531e-06, + "loss": 0.2538, + "step": 20425 + }, + { + "epoch": 0.21, + "learning_rate": 7.772067039106145e-06, + "loss": 0.2645, + "step": 20450 + }, + { + "epoch": 0.21, + "learning_rate": 7.76927374301676e-06, + "loss": 0.3107, + "step": 20475 + }, + { + "epoch": 0.21, + "learning_rate": 7.766480446927375e-06, + "loss": 0.2914, + "step": 20500 + }, + { + "epoch": 0.21, + "learning_rate": 7.76368715083799e-06, + "loss": 0.3234, + "step": 20525 + }, + { + "epoch": 0.21, + "learning_rate": 7.760893854748604e-06, + "loss": 0.285, + "step": 20550 + }, + { + "epoch": 0.21, + "learning_rate": 7.758100558659218e-06, + "loss": 0.291, + "step": 20575 + }, + { + "epoch": 0.21, + "learning_rate": 7.755307262569833e-06, + "loss": 0.2749, + "step": 20600 + }, + { + "epoch": 0.21, + "learning_rate": 7.752513966480447e-06, + "loss": 0.2857, + "step": 20625 + }, + { + "epoch": 0.21, + "learning_rate": 7.749720670391063e-06, + "loss": 0.2689, + "step": 20650 + }, + { + "epoch": 0.21, + "learning_rate": 7.746927374301677e-06, + "loss": 0.3055, + "step": 20675 + }, + { + "epoch": 0.21, + "learning_rate": 7.74413407821229e-06, + "loss": 0.2785, + "step": 20700 + }, + { + "epoch": 0.21, + "learning_rate": 7.741340782122906e-06, + "loss": 0.2723, + "step": 20725 + }, + { + "epoch": 0.21, + "learning_rate": 7.73854748603352e-06, + "loss": 0.302, + "step": 20750 + }, + { + "epoch": 0.21, + "learning_rate": 7.735754189944135e-06, + "loss": 0.2766, + "step": 20775 + }, + { + "epoch": 0.21, + "learning_rate": 7.73296089385475e-06, + "loss": 0.273, + "step": 20800 + }, + { + "epoch": 0.21, + "learning_rate": 7.730167597765363e-06, + "loss": 0.2911, + "step": 20825 + }, + { + "epoch": 0.21, + "learning_rate": 7.727374301675979e-06, + "loss": 0.3225, + "step": 20850 + }, + { + "epoch": 0.21, + "learning_rate": 7.724581005586592e-06, + "loss": 0.2923, + "step": 20875 + }, + { + "epoch": 0.21, + "learning_rate": 7.721787709497206e-06, + "loss": 0.2939, + "step": 20900 + }, + { + "epoch": 0.21, + "learning_rate": 7.718994413407822e-06, + "loss": 0.3003, + "step": 20925 + }, + { + "epoch": 0.21, + "learning_rate": 7.716201117318436e-06, + "loss": 0.3042, + "step": 20950 + }, + { + "epoch": 0.21, + "learning_rate": 7.713407821229051e-06, + "loss": 0.295, + "step": 20975 + }, + { + "epoch": 0.21, + "learning_rate": 7.710614525139665e-06, + "loss": 0.3168, + "step": 21000 + }, + { + "epoch": 0.21, + "learning_rate": 7.707821229050279e-06, + "loss": 0.3117, + "step": 21025 + }, + { + "epoch": 0.21, + "learning_rate": 7.705027932960895e-06, + "loss": 0.2583, + "step": 21050 + }, + { + "epoch": 0.21, + "learning_rate": 7.70223463687151e-06, + "loss": 0.27, + "step": 21075 + }, + { + "epoch": 0.21, + "learning_rate": 7.699441340782124e-06, + "loss": 0.2793, + "step": 21100 + }, + { + "epoch": 0.21, + "learning_rate": 7.696648044692738e-06, + "loss": 0.2689, + "step": 21125 + }, + { + "epoch": 0.21, + "learning_rate": 7.693854748603352e-06, + "loss": 0.2794, + "step": 21150 + }, + { + "epoch": 0.21, + "learning_rate": 7.691061452513967e-06, + "loss": 0.3204, + "step": 21175 + }, + { + "epoch": 0.21, + "learning_rate": 7.688268156424583e-06, + "loss": 0.2865, + "step": 21200 + }, + { + "epoch": 0.21, + "learning_rate": 7.685474860335197e-06, + "loss": 0.2756, + "step": 21225 + }, + { + "epoch": 0.21, + "learning_rate": 7.68268156424581e-06, + "loss": 0.2967, + "step": 21250 + }, + { + "epoch": 0.21, + "learning_rate": 7.679888268156426e-06, + "loss": 0.2881, + "step": 21275 + }, + { + "epoch": 0.21, + "learning_rate": 7.67709497206704e-06, + "loss": 0.3097, + "step": 21300 + }, + { + "epoch": 0.21, + "learning_rate": 7.674301675977655e-06, + "loss": 0.3265, + "step": 21325 + }, + { + "epoch": 0.21, + "learning_rate": 7.671508379888268e-06, + "loss": 0.2805, + "step": 21350 + }, + { + "epoch": 0.21, + "learning_rate": 7.668715083798883e-06, + "loss": 0.293, + "step": 21375 + }, + { + "epoch": 0.21, + "learning_rate": 7.665921787709499e-06, + "loss": 0.287, + "step": 21400 + }, + { + "epoch": 0.21, + "learning_rate": 7.663128491620112e-06, + "loss": 0.2933, + "step": 21425 + }, + { + "epoch": 0.22, + "learning_rate": 7.660335195530726e-06, + "loss": 0.2731, + "step": 21450 + }, + { + "epoch": 0.22, + "learning_rate": 7.65754189944134e-06, + "loss": 0.2871, + "step": 21475 + }, + { + "epoch": 0.22, + "learning_rate": 7.654748603351956e-06, + "loss": 0.2692, + "step": 21500 + }, + { + "epoch": 0.22, + "learning_rate": 7.651955307262571e-06, + "loss": 0.2979, + "step": 21525 + }, + { + "epoch": 0.22, + "learning_rate": 7.649162011173185e-06, + "loss": 0.2851, + "step": 21550 + }, + { + "epoch": 0.22, + "learning_rate": 7.646368715083799e-06, + "loss": 0.3049, + "step": 21575 + }, + { + "epoch": 0.22, + "learning_rate": 7.643575418994415e-06, + "loss": 0.267, + "step": 21600 + }, + { + "epoch": 0.22, + "learning_rate": 7.640782122905028e-06, + "loss": 0.2746, + "step": 21625 + }, + { + "epoch": 0.22, + "learning_rate": 7.637988826815644e-06, + "loss": 0.2861, + "step": 21650 + }, + { + "epoch": 0.22, + "learning_rate": 7.635195530726258e-06, + "loss": 0.3183, + "step": 21675 + }, + { + "epoch": 0.22, + "learning_rate": 7.632402234636872e-06, + "loss": 0.2826, + "step": 21700 + }, + { + "epoch": 0.22, + "learning_rate": 7.629608938547487e-06, + "loss": 0.3346, + "step": 21725 + }, + { + "epoch": 0.22, + "learning_rate": 7.626815642458101e-06, + "loss": 0.3218, + "step": 21750 + }, + { + "epoch": 0.22, + "learning_rate": 7.624022346368716e-06, + "loss": 0.3007, + "step": 21775 + }, + { + "epoch": 0.22, + "learning_rate": 7.62122905027933e-06, + "loss": 0.278, + "step": 21800 + }, + { + "epoch": 0.22, + "learning_rate": 7.618435754189944e-06, + "loss": 0.3223, + "step": 21825 + }, + { + "epoch": 0.22, + "learning_rate": 7.61564245810056e-06, + "loss": 0.2794, + "step": 21850 + }, + { + "epoch": 0.22, + "learning_rate": 7.612849162011174e-06, + "loss": 0.3169, + "step": 21875 + }, + { + "epoch": 0.22, + "learning_rate": 7.610055865921788e-06, + "loss": 0.304, + "step": 21900 + }, + { + "epoch": 0.22, + "learning_rate": 7.607262569832403e-06, + "loss": 0.273, + "step": 21925 + }, + { + "epoch": 0.22, + "learning_rate": 7.604469273743017e-06, + "loss": 0.2929, + "step": 21950 + }, + { + "epoch": 0.22, + "learning_rate": 7.601675977653632e-06, + "loss": 0.3161, + "step": 21975 + }, + { + "epoch": 0.22, + "learning_rate": 7.598882681564247e-06, + "loss": 0.2971, + "step": 22000 + }, + { + "epoch": 0.22, + "learning_rate": 7.596089385474861e-06, + "loss": 0.2866, + "step": 22025 + }, + { + "epoch": 0.22, + "learning_rate": 7.593296089385476e-06, + "loss": 0.2983, + "step": 22050 + }, + { + "epoch": 0.22, + "learning_rate": 7.5905027932960895e-06, + "loss": 0.303, + "step": 22075 + }, + { + "epoch": 0.22, + "learning_rate": 7.587709497206704e-06, + "loss": 0.291, + "step": 22100 + }, + { + "epoch": 0.22, + "learning_rate": 7.58491620111732e-06, + "loss": 0.2746, + "step": 22125 + }, + { + "epoch": 0.22, + "learning_rate": 7.582122905027934e-06, + "loss": 0.2934, + "step": 22150 + }, + { + "epoch": 0.22, + "learning_rate": 7.579329608938548e-06, + "loss": 0.2774, + "step": 22175 + }, + { + "epoch": 0.22, + "learning_rate": 7.576536312849163e-06, + "loss": 0.3043, + "step": 22200 + }, + { + "epoch": 0.22, + "learning_rate": 7.573743016759777e-06, + "loss": 0.3318, + "step": 22225 + }, + { + "epoch": 0.22, + "learning_rate": 7.570949720670392e-06, + "loss": 0.3015, + "step": 22250 + }, + { + "epoch": 0.22, + "learning_rate": 7.5681564245810054e-06, + "loss": 0.2771, + "step": 22275 + }, + { + "epoch": 0.22, + "learning_rate": 7.565363128491621e-06, + "loss": 0.3009, + "step": 22300 + }, + { + "epoch": 0.22, + "learning_rate": 7.562569832402236e-06, + "loss": 0.2873, + "step": 22325 + }, + { + "epoch": 0.22, + "learning_rate": 7.5597765363128495e-06, + "loss": 0.2871, + "step": 22350 + }, + { + "epoch": 0.22, + "learning_rate": 7.556983240223464e-06, + "loss": 0.2789, + "step": 22375 + }, + { + "epoch": 0.22, + "learning_rate": 7.554189944134079e-06, + "loss": 0.3126, + "step": 22400 + }, + { + "epoch": 0.22, + "learning_rate": 7.551396648044693e-06, + "loss": 0.2563, + "step": 22425 + }, + { + "epoch": 0.23, + "learning_rate": 7.548603351955308e-06, + "loss": 0.3063, + "step": 22450 + }, + { + "epoch": 0.23, + "learning_rate": 7.545810055865922e-06, + "loss": 0.2766, + "step": 22475 + }, + { + "epoch": 0.23, + "learning_rate": 7.543016759776537e-06, + "loss": 0.265, + "step": 22500 + }, + { + "epoch": 0.23, + "eval_loss": 0.29214224219322205, + "eval_runtime": 1963.8182, + "eval_samples_per_second": 2.549, + "eval_steps_per_second": 0.159, + "eval_wer": 21.65215121383991, + "step": 22500 + }, + { + "epoch": 0.23, + "learning_rate": 7.540223463687152e-06, + "loss": 0.282, + "step": 22525 + }, + { + "epoch": 0.23, + "learning_rate": 7.5374301675977654e-06, + "loss": 0.2904, + "step": 22550 + }, + { + "epoch": 0.23, + "learning_rate": 7.534636871508381e-06, + "loss": 0.2779, + "step": 22575 + }, + { + "epoch": 0.23, + "learning_rate": 7.531843575418996e-06, + "loss": 0.254, + "step": 22600 + }, + { + "epoch": 0.23, + "learning_rate": 7.5290502793296095e-06, + "loss": 0.2842, + "step": 22625 + }, + { + "epoch": 0.23, + "learning_rate": 7.526256983240224e-06, + "loss": 0.3148, + "step": 22650 + }, + { + "epoch": 0.23, + "learning_rate": 7.523463687150838e-06, + "loss": 0.3017, + "step": 22675 + }, + { + "epoch": 0.23, + "learning_rate": 7.520670391061453e-06, + "loss": 0.2641, + "step": 22700 + }, + { + "epoch": 0.23, + "learning_rate": 7.517877094972068e-06, + "loss": 0.2789, + "step": 22725 + }, + { + "epoch": 0.23, + "learning_rate": 7.515083798882682e-06, + "loss": 0.3007, + "step": 22750 + }, + { + "epoch": 0.23, + "learning_rate": 7.512290502793297e-06, + "loss": 0.2637, + "step": 22775 + }, + { + "epoch": 0.23, + "learning_rate": 7.509497206703911e-06, + "loss": 0.2989, + "step": 22800 + }, + { + "epoch": 0.23, + "learning_rate": 7.506703910614525e-06, + "loss": 0.2949, + "step": 22825 + }, + { + "epoch": 0.23, + "learning_rate": 7.503910614525141e-06, + "loss": 0.3081, + "step": 22850 + }, + { + "epoch": 0.23, + "learning_rate": 7.501229050279331e-06, + "loss": 0.3087, + "step": 22875 + }, + { + "epoch": 0.23, + "learning_rate": 7.4984357541899446e-06, + "loss": 0.284, + "step": 22900 + }, + { + "epoch": 0.23, + "learning_rate": 7.495642458100559e-06, + "loss": 0.2782, + "step": 22925 + }, + { + "epoch": 0.23, + "learning_rate": 7.492849162011174e-06, + "loss": 0.2913, + "step": 22950 + }, + { + "epoch": 0.23, + "learning_rate": 7.490055865921788e-06, + "loss": 0.286, + "step": 22975 + }, + { + "epoch": 0.23, + "learning_rate": 7.487262569832403e-06, + "loss": 0.3133, + "step": 23000 + }, + { + "epoch": 0.23, + "learning_rate": 7.484469273743017e-06, + "loss": 0.3088, + "step": 23025 + }, + { + "epoch": 0.23, + "learning_rate": 7.481675977653632e-06, + "loss": 0.3023, + "step": 23050 + }, + { + "epoch": 0.23, + "learning_rate": 7.478882681564247e-06, + "loss": 0.2573, + "step": 23075 + }, + { + "epoch": 0.23, + "learning_rate": 7.4760893854748605e-06, + "loss": 0.2677, + "step": 23100 + }, + { + "epoch": 0.23, + "learning_rate": 7.473296089385475e-06, + "loss": 0.2876, + "step": 23125 + }, + { + "epoch": 0.23, + "learning_rate": 7.470502793296091e-06, + "loss": 0.2744, + "step": 23150 + }, + { + "epoch": 0.23, + "learning_rate": 7.4677094972067045e-06, + "loss": 0.2716, + "step": 23175 + }, + { + "epoch": 0.23, + "learning_rate": 7.464916201117319e-06, + "loss": 0.3055, + "step": 23200 + }, + { + "epoch": 0.23, + "learning_rate": 7.462122905027933e-06, + "loss": 0.3037, + "step": 23225 + }, + { + "epoch": 0.23, + "learning_rate": 7.459329608938548e-06, + "loss": 0.2763, + "step": 23250 + }, + { + "epoch": 0.23, + "learning_rate": 7.456536312849163e-06, + "loss": 0.2719, + "step": 23275 + }, + { + "epoch": 0.23, + "learning_rate": 7.453743016759777e-06, + "loss": 0.3171, + "step": 23300 + }, + { + "epoch": 0.23, + "learning_rate": 7.450949720670392e-06, + "loss": 0.2814, + "step": 23325 + }, + { + "epoch": 0.23, + "learning_rate": 7.448156424581007e-06, + "loss": 0.2957, + "step": 23350 + }, + { + "epoch": 0.23, + "learning_rate": 7.4453631284916204e-06, + "loss": 0.2962, + "step": 23375 + }, + { + "epoch": 0.23, + "learning_rate": 7.442569832402235e-06, + "loss": 0.2735, + "step": 23400 + }, + { + "epoch": 0.23, + "learning_rate": 7.439776536312849e-06, + "loss": 0.2845, + "step": 23425 + }, + { + "epoch": 0.24, + "learning_rate": 7.4369832402234645e-06, + "loss": 0.3007, + "step": 23450 + }, + { + "epoch": 0.24, + "learning_rate": 7.434189944134079e-06, + "loss": 0.2498, + "step": 23475 + }, + { + "epoch": 0.24, + "learning_rate": 7.431396648044693e-06, + "loss": 0.3054, + "step": 23500 + }, + { + "epoch": 0.24, + "learning_rate": 7.428603351955308e-06, + "loss": 0.2996, + "step": 23525 + }, + { + "epoch": 0.24, + "learning_rate": 7.425810055865922e-06, + "loss": 0.2656, + "step": 23550 + }, + { + "epoch": 0.24, + "learning_rate": 7.423016759776537e-06, + "loss": 0.2559, + "step": 23575 + }, + { + "epoch": 0.24, + "learning_rate": 7.420223463687152e-06, + "loss": 0.2999, + "step": 23600 + }, + { + "epoch": 0.24, + "learning_rate": 7.417430167597766e-06, + "loss": 0.2848, + "step": 23625 + }, + { + "epoch": 0.24, + "learning_rate": 7.4146368715083804e-06, + "loss": 0.2584, + "step": 23650 + }, + { + "epoch": 0.24, + "learning_rate": 7.411843575418995e-06, + "loss": 0.2918, + "step": 23675 + }, + { + "epoch": 0.24, + "learning_rate": 7.409050279329609e-06, + "loss": 0.3117, + "step": 23700 + }, + { + "epoch": 0.24, + "learning_rate": 7.4062569832402245e-06, + "loss": 0.286, + "step": 23725 + }, + { + "epoch": 0.24, + "learning_rate": 7.403463687150838e-06, + "loss": 0.2934, + "step": 23750 + }, + { + "epoch": 0.24, + "learning_rate": 7.400670391061453e-06, + "loss": 0.2734, + "step": 23775 + }, + { + "epoch": 0.24, + "learning_rate": 7.397877094972068e-06, + "loss": 0.2902, + "step": 23800 + }, + { + "epoch": 0.24, + "learning_rate": 7.395083798882682e-06, + "loss": 0.2604, + "step": 23825 + }, + { + "epoch": 0.24, + "learning_rate": 7.392290502793296e-06, + "loss": 0.2731, + "step": 23850 + }, + { + "epoch": 0.24, + "learning_rate": 7.389497206703912e-06, + "loss": 0.2729, + "step": 23875 + }, + { + "epoch": 0.24, + "learning_rate": 7.386703910614526e-06, + "loss": 0.2927, + "step": 23900 + }, + { + "epoch": 0.24, + "learning_rate": 7.38391061452514e-06, + "loss": 0.2537, + "step": 23925 + }, + { + "epoch": 0.24, + "learning_rate": 7.381117318435754e-06, + "loss": 0.3067, + "step": 23950 + }, + { + "epoch": 0.24, + "learning_rate": 7.378324022346369e-06, + "loss": 0.3143, + "step": 23975 + }, + { + "epoch": 0.24, + "learning_rate": 7.3755307262569845e-06, + "loss": 0.3046, + "step": 24000 + }, + { + "epoch": 0.24, + "learning_rate": 7.372737430167598e-06, + "loss": 0.2616, + "step": 24025 + }, + { + "epoch": 0.24, + "learning_rate": 7.369944134078213e-06, + "loss": 0.3409, + "step": 24050 + }, + { + "epoch": 0.24, + "learning_rate": 7.367150837988828e-06, + "loss": 0.3058, + "step": 24075 + }, + { + "epoch": 0.24, + "learning_rate": 7.364357541899442e-06, + "loss": 0.279, + "step": 24100 + }, + { + "epoch": 0.24, + "learning_rate": 7.361564245810056e-06, + "loss": 0.2882, + "step": 24125 + }, + { + "epoch": 0.24, + "learning_rate": 7.35877094972067e-06, + "loss": 0.2705, + "step": 24150 + }, + { + "epoch": 0.24, + "learning_rate": 7.355977653631286e-06, + "loss": 0.3175, + "step": 24175 + }, + { + "epoch": 0.24, + "learning_rate": 7.3531843575419e-06, + "loss": 0.2601, + "step": 24200 + }, + { + "epoch": 0.24, + "learning_rate": 7.350391061452514e-06, + "loss": 0.267, + "step": 24225 + }, + { + "epoch": 0.24, + "learning_rate": 7.347597765363129e-06, + "loss": 0.291, + "step": 24250 + }, + { + "epoch": 0.24, + "learning_rate": 7.3448044692737445e-06, + "loss": 0.2628, + "step": 24275 + }, + { + "epoch": 0.24, + "learning_rate": 7.342011173184358e-06, + "loss": 0.2462, + "step": 24300 + }, + { + "epoch": 0.24, + "learning_rate": 7.339217877094973e-06, + "loss": 0.2995, + "step": 24325 + }, + { + "epoch": 0.24, + "learning_rate": 7.336424581005587e-06, + "loss": 0.2804, + "step": 24350 + }, + { + "epoch": 0.24, + "learning_rate": 7.333631284916202e-06, + "loss": 0.2876, + "step": 24375 + }, + { + "epoch": 0.24, + "learning_rate": 7.330837988826816e-06, + "loss": 0.2638, + "step": 24400 + }, + { + "epoch": 0.24, + "learning_rate": 7.32804469273743e-06, + "loss": 0.2629, + "step": 24425 + }, + { + "epoch": 0.25, + "learning_rate": 7.325251396648046e-06, + "loss": 0.2737, + "step": 24450 + }, + { + "epoch": 0.25, + "learning_rate": 7.32245810055866e-06, + "loss": 0.2764, + "step": 24475 + }, + { + "epoch": 0.25, + "learning_rate": 7.319664804469274e-06, + "loss": 0.2397, + "step": 24500 + }, + { + "epoch": 0.25, + "learning_rate": 7.316871508379889e-06, + "loss": 0.3003, + "step": 24525 + }, + { + "epoch": 0.25, + "learning_rate": 7.314078212290503e-06, + "loss": 0.2797, + "step": 24550 + }, + { + "epoch": 0.25, + "learning_rate": 7.3112849162011175e-06, + "loss": 0.2933, + "step": 24575 + }, + { + "epoch": 0.25, + "learning_rate": 7.308491620111733e-06, + "loss": 0.2648, + "step": 24600 + }, + { + "epoch": 0.25, + "learning_rate": 7.305698324022347e-06, + "loss": 0.3133, + "step": 24625 + }, + { + "epoch": 0.25, + "learning_rate": 7.302905027932962e-06, + "loss": 0.293, + "step": 24650 + }, + { + "epoch": 0.25, + "learning_rate": 7.3001117318435755e-06, + "loss": 0.2909, + "step": 24675 + }, + { + "epoch": 0.25, + "learning_rate": 7.29731843575419e-06, + "loss": 0.2804, + "step": 24700 + }, + { + "epoch": 0.25, + "learning_rate": 7.294525139664806e-06, + "loss": 0.282, + "step": 24725 + }, + { + "epoch": 0.25, + "learning_rate": 7.2917318435754195e-06, + "loss": 0.2836, + "step": 24750 + }, + { + "epoch": 0.25, + "learning_rate": 7.288938547486034e-06, + "loss": 0.2879, + "step": 24775 + }, + { + "epoch": 0.25, + "learning_rate": 7.286145251396649e-06, + "loss": 0.254, + "step": 24800 + }, + { + "epoch": 0.25, + "learning_rate": 7.283351955307263e-06, + "loss": 0.2751, + "step": 24825 + }, + { + "epoch": 0.25, + "learning_rate": 7.2805586592178775e-06, + "loss": 0.2575, + "step": 24850 + }, + { + "epoch": 0.25, + "learning_rate": 7.277765363128491e-06, + "loss": 0.2635, + "step": 24875 + }, + { + "epoch": 0.25, + "learning_rate": 7.274972067039107e-06, + "loss": 0.2934, + "step": 24900 + }, + { + "epoch": 0.25, + "learning_rate": 7.272178770949722e-06, + "loss": 0.2753, + "step": 24925 + }, + { + "epoch": 0.25, + "learning_rate": 7.2693854748603354e-06, + "loss": 0.287, + "step": 24950 + }, + { + "epoch": 0.25, + "learning_rate": 7.26659217877095e-06, + "loss": 0.2688, + "step": 24975 + }, + { + "epoch": 0.25, + "learning_rate": 7.263798882681566e-06, + "loss": 0.3096, + "step": 25000 + }, + { + "epoch": 0.25, + "eval_loss": 0.2893700897693634, + "eval_runtime": 1899.9716, + "eval_samples_per_second": 2.634, + "eval_steps_per_second": 0.165, + "eval_wer": 22.150501910344335, + "step": 25000 + }, + { + "epoch": 0.25, + "learning_rate": 7.2610055865921795e-06, + "loss": 0.2716, + "step": 25025 + }, + { + "epoch": 0.25, + "learning_rate": 7.258212290502794e-06, + "loss": 0.2785, + "step": 25050 + }, + { + "epoch": 0.25, + "learning_rate": 7.255418994413408e-06, + "loss": 0.2964, + "step": 25075 + }, + { + "epoch": 0.25, + "learning_rate": 7.252625698324023e-06, + "loss": 0.3212, + "step": 25100 + }, + { + "epoch": 0.25, + "learning_rate": 7.2498324022346375e-06, + "loss": 0.2649, + "step": 25125 + }, + { + "epoch": 0.25, + "learning_rate": 7.247039106145251e-06, + "loss": 0.2751, + "step": 25150 + }, + { + "epoch": 0.25, + "learning_rate": 7.244245810055867e-06, + "loss": 0.304, + "step": 25175 + }, + { + "epoch": 0.25, + "learning_rate": 7.241452513966482e-06, + "loss": 0.2748, + "step": 25200 + }, + { + "epoch": 0.25, + "learning_rate": 7.2386592178770954e-06, + "loss": 0.2611, + "step": 25225 + }, + { + "epoch": 0.25, + "learning_rate": 7.23586592178771e-06, + "loss": 0.2491, + "step": 25250 + }, + { + "epoch": 0.25, + "learning_rate": 7.233072625698324e-06, + "loss": 0.2587, + "step": 25275 + }, + { + "epoch": 0.25, + "learning_rate": 7.230279329608939e-06, + "loss": 0.315, + "step": 25300 + }, + { + "epoch": 0.25, + "learning_rate": 7.227486033519554e-06, + "loss": 0.2623, + "step": 25325 + }, + { + "epoch": 0.25, + "learning_rate": 7.224692737430168e-06, + "loss": 0.3033, + "step": 25350 + }, + { + "epoch": 0.25, + "learning_rate": 7.221899441340783e-06, + "loss": 0.3289, + "step": 25375 + }, + { + "epoch": 0.25, + "learning_rate": 7.2191061452513975e-06, + "loss": 0.2518, + "step": 25400 + }, + { + "epoch": 0.25, + "learning_rate": 7.216312849162011e-06, + "loss": 0.2771, + "step": 25425 + }, + { + "epoch": 0.26, + "learning_rate": 7.213519553072627e-06, + "loss": 0.3256, + "step": 25450 + }, + { + "epoch": 0.26, + "learning_rate": 7.210726256983241e-06, + "loss": 0.2805, + "step": 25475 + }, + { + "epoch": 0.26, + "learning_rate": 7.207932960893855e-06, + "loss": 0.2808, + "step": 25500 + }, + { + "epoch": 0.26, + "learning_rate": 7.20513966480447e-06, + "loss": 0.2556, + "step": 25525 + }, + { + "epoch": 0.26, + "learning_rate": 7.202346368715084e-06, + "loss": 0.3013, + "step": 25550 + }, + { + "epoch": 0.26, + "learning_rate": 7.199553072625699e-06, + "loss": 0.2808, + "step": 25575 + }, + { + "epoch": 0.26, + "learning_rate": 7.196759776536314e-06, + "loss": 0.2559, + "step": 25600 + }, + { + "epoch": 0.26, + "learning_rate": 7.193966480446928e-06, + "loss": 0.2839, + "step": 25625 + }, + { + "epoch": 0.26, + "learning_rate": 7.191173184357543e-06, + "loss": 0.2411, + "step": 25650 + }, + { + "epoch": 0.26, + "learning_rate": 7.188379888268157e-06, + "loss": 0.2878, + "step": 25675 + }, + { + "epoch": 0.26, + "learning_rate": 7.185586592178771e-06, + "loss": 0.3006, + "step": 25700 + }, + { + "epoch": 0.26, + "learning_rate": 7.182793296089387e-06, + "loss": 0.3088, + "step": 25725 + }, + { + "epoch": 0.26, + "learning_rate": 7.180000000000001e-06, + "loss": 0.2883, + "step": 25750 + }, + { + "epoch": 0.26, + "learning_rate": 7.177206703910615e-06, + "loss": 0.2761, + "step": 25775 + }, + { + "epoch": 0.26, + "learning_rate": 7.17441340782123e-06, + "loss": 0.2722, + "step": 25800 + }, + { + "epoch": 0.26, + "learning_rate": 7.171620111731844e-06, + "loss": 0.2515, + "step": 25825 + }, + { + "epoch": 0.26, + "learning_rate": 7.168826815642459e-06, + "loss": 0.2927, + "step": 25850 + }, + { + "epoch": 0.26, + "learning_rate": 7.1660335195530725e-06, + "loss": 0.3226, + "step": 25875 + }, + { + "epoch": 0.26, + "learning_rate": 7.163240223463688e-06, + "loss": 0.2818, + "step": 25900 + }, + { + "epoch": 0.26, + "learning_rate": 7.160446927374303e-06, + "loss": 0.2687, + "step": 25925 + }, + { + "epoch": 0.26, + "learning_rate": 7.157653631284917e-06, + "loss": 0.2536, + "step": 25950 + }, + { + "epoch": 0.26, + "learning_rate": 7.154860335195531e-06, + "loss": 0.2981, + "step": 25975 + }, + { + "epoch": 0.26, + "learning_rate": 7.152067039106145e-06, + "loss": 0.2632, + "step": 26000 + }, + { + "epoch": 0.26, + "learning_rate": 7.14927374301676e-06, + "loss": 0.2694, + "step": 26025 + }, + { + "epoch": 0.26, + "learning_rate": 7.146480446927375e-06, + "loss": 0.2544, + "step": 26050 + }, + { + "epoch": 0.26, + "learning_rate": 7.143687150837989e-06, + "loss": 0.2542, + "step": 26075 + }, + { + "epoch": 0.26, + "learning_rate": 7.140893854748604e-06, + "loss": 0.2951, + "step": 26100 + }, + { + "epoch": 0.26, + "learning_rate": 7.138100558659219e-06, + "loss": 0.3033, + "step": 26125 + }, + { + "epoch": 0.26, + "learning_rate": 7.1353072625698325e-06, + "loss": 0.2908, + "step": 26150 + }, + { + "epoch": 0.26, + "learning_rate": 7.132513966480448e-06, + "loss": 0.3144, + "step": 26175 + }, + { + "epoch": 0.26, + "learning_rate": 7.129720670391062e-06, + "loss": 0.3125, + "step": 26200 + }, + { + "epoch": 0.26, + "learning_rate": 7.126927374301677e-06, + "loss": 0.3257, + "step": 26225 + }, + { + "epoch": 0.26, + "learning_rate": 7.124134078212291e-06, + "loss": 0.2592, + "step": 26250 + }, + { + "epoch": 0.26, + "learning_rate": 7.121340782122905e-06, + "loss": 0.3107, + "step": 26275 + }, + { + "epoch": 0.26, + "learning_rate": 7.11854748603352e-06, + "loss": 0.2937, + "step": 26300 + }, + { + "epoch": 0.26, + "learning_rate": 7.115754189944135e-06, + "loss": 0.2735, + "step": 26325 + }, + { + "epoch": 0.26, + "learning_rate": 7.112960893854749e-06, + "loss": 0.2786, + "step": 26350 + }, + { + "epoch": 0.26, + "learning_rate": 7.110167597765364e-06, + "loss": 0.2754, + "step": 26375 + }, + { + "epoch": 0.26, + "learning_rate": 7.107374301675978e-06, + "loss": 0.2895, + "step": 26400 + }, + { + "epoch": 0.26, + "learning_rate": 7.1045810055865925e-06, + "loss": 0.2805, + "step": 26425 + }, + { + "epoch": 0.27, + "learning_rate": 7.101787709497208e-06, + "loss": 0.262, + "step": 26450 + }, + { + "epoch": 0.27, + "learning_rate": 7.098994413407822e-06, + "loss": 0.2826, + "step": 26475 + }, + { + "epoch": 0.27, + "learning_rate": 7.096201117318437e-06, + "loss": 0.2807, + "step": 26500 + }, + { + "epoch": 0.27, + "learning_rate": 7.093407821229051e-06, + "loss": 0.2671, + "step": 26525 + }, + { + "epoch": 0.27, + "learning_rate": 7.090614525139665e-06, + "loss": 0.2924, + "step": 26550 + }, + { + "epoch": 0.27, + "learning_rate": 7.08782122905028e-06, + "loss": 0.2849, + "step": 26575 + }, + { + "epoch": 0.27, + "learning_rate": 7.085027932960894e-06, + "loss": 0.2708, + "step": 26600 + }, + { + "epoch": 0.27, + "learning_rate": 7.082234636871509e-06, + "loss": 0.2795, + "step": 26625 + }, + { + "epoch": 0.27, + "learning_rate": 7.079441340782124e-06, + "loss": 0.2711, + "step": 26650 + }, + { + "epoch": 0.27, + "learning_rate": 7.076648044692738e-06, + "loss": 0.2816, + "step": 26675 + }, + { + "epoch": 0.27, + "learning_rate": 7.0738547486033525e-06, + "loss": 0.2871, + "step": 26700 + }, + { + "epoch": 0.27, + "learning_rate": 7.071061452513967e-06, + "loss": 0.2914, + "step": 26725 + }, + { + "epoch": 0.27, + "learning_rate": 7.068268156424581e-06, + "loss": 0.2835, + "step": 26750 + }, + { + "epoch": 0.27, + "learning_rate": 7.065474860335197e-06, + "loss": 0.3025, + "step": 26775 + }, + { + "epoch": 0.27, + "learning_rate": 7.0626815642458104e-06, + "loss": 0.2999, + "step": 26800 + }, + { + "epoch": 0.27, + "learning_rate": 7.059888268156425e-06, + "loss": 0.2882, + "step": 26825 + }, + { + "epoch": 0.27, + "learning_rate": 7.05709497206704e-06, + "loss": 0.2944, + "step": 26850 + }, + { + "epoch": 0.27, + "learning_rate": 7.054301675977654e-06, + "loss": 0.318, + "step": 26875 + }, + { + "epoch": 0.27, + "learning_rate": 7.051508379888269e-06, + "loss": 0.2736, + "step": 26900 + }, + { + "epoch": 0.27, + "learning_rate": 7.048715083798884e-06, + "loss": 0.3086, + "step": 26925 + }, + { + "epoch": 0.27, + "learning_rate": 7.045921787709498e-06, + "loss": 0.2803, + "step": 26950 + }, + { + "epoch": 0.27, + "learning_rate": 7.0431284916201125e-06, + "loss": 0.2619, + "step": 26975 + }, + { + "epoch": 0.27, + "learning_rate": 7.040335195530726e-06, + "loss": 0.2935, + "step": 27000 + }, + { + "epoch": 0.27, + "learning_rate": 7.037541899441341e-06, + "loss": 0.2684, + "step": 27025 + }, + { + "epoch": 0.27, + "learning_rate": 7.0347486033519566e-06, + "loss": 0.2635, + "step": 27050 + }, + { + "epoch": 0.27, + "learning_rate": 7.032067039106146e-06, + "loss": 0.2927, + "step": 27075 + }, + { + "epoch": 0.27, + "learning_rate": 7.02927374301676e-06, + "loss": 0.277, + "step": 27100 + }, + { + "epoch": 0.27, + "learning_rate": 7.026480446927375e-06, + "loss": 0.2992, + "step": 27125 + }, + { + "epoch": 0.27, + "learning_rate": 7.023687150837989e-06, + "loss": 0.2656, + "step": 27150 + }, + { + "epoch": 0.27, + "learning_rate": 7.020893854748604e-06, + "loss": 0.284, + "step": 27175 + }, + { + "epoch": 0.27, + "learning_rate": 7.018100558659219e-06, + "loss": 0.2572, + "step": 27200 + }, + { + "epoch": 0.27, + "learning_rate": 7.015307262569833e-06, + "loss": 0.2614, + "step": 27225 + }, + { + "epoch": 0.27, + "learning_rate": 7.0125139664804475e-06, + "loss": 0.2583, + "step": 27250 + }, + { + "epoch": 0.27, + "learning_rate": 7.009720670391062e-06, + "loss": 0.2694, + "step": 27275 + }, + { + "epoch": 0.27, + "learning_rate": 7.006927374301676e-06, + "loss": 0.2753, + "step": 27300 + }, + { + "epoch": 0.27, + "learning_rate": 7.004134078212292e-06, + "loss": 0.266, + "step": 27325 + }, + { + "epoch": 0.27, + "learning_rate": 7.0013407821229054e-06, + "loss": 0.2565, + "step": 27350 + }, + { + "epoch": 0.27, + "learning_rate": 6.99854748603352e-06, + "loss": 0.2772, + "step": 27375 + }, + { + "epoch": 0.27, + "learning_rate": 6.995754189944135e-06, + "loss": 0.2738, + "step": 27400 + }, + { + "epoch": 0.27, + "learning_rate": 6.992960893854749e-06, + "loss": 0.2907, + "step": 27425 + }, + { + "epoch": 0.28, + "learning_rate": 6.990167597765363e-06, + "loss": 0.2624, + "step": 27450 + }, + { + "epoch": 0.28, + "learning_rate": 6.987374301675979e-06, + "loss": 0.2901, + "step": 27475 + }, + { + "epoch": 0.28, + "learning_rate": 6.984581005586593e-06, + "loss": 0.2813, + "step": 27500 + }, + { + "epoch": 0.28, + "eval_loss": 0.28631410002708435, + "eval_runtime": 1920.3451, + "eval_samples_per_second": 2.606, + "eval_steps_per_second": 0.163, + "eval_wer": 22.499347397897434, + "step": 27500 + }, + { + "epoch": 0.28, + "learning_rate": 6.9817877094972075e-06, + "loss": 0.3067, + "step": 27525 + }, + { + "epoch": 0.28, + "learning_rate": 6.978994413407821e-06, + "loss": 0.308, + "step": 27550 + }, + { + "epoch": 0.28, + "learning_rate": 6.976201117318436e-06, + "loss": 0.2688, + "step": 27575 + }, + { + "epoch": 0.28, + "learning_rate": 6.973407821229052e-06, + "loss": 0.2738, + "step": 27600 + }, + { + "epoch": 0.28, + "learning_rate": 6.9706145251396654e-06, + "loss": 0.3074, + "step": 27625 + }, + { + "epoch": 0.28, + "learning_rate": 6.96782122905028e-06, + "loss": 0.2585, + "step": 27650 + }, + { + "epoch": 0.28, + "learning_rate": 6.965027932960895e-06, + "loss": 0.2638, + "step": 27675 + }, + { + "epoch": 0.28, + "learning_rate": 6.962234636871509e-06, + "loss": 0.2795, + "step": 27700 + }, + { + "epoch": 0.28, + "learning_rate": 6.959441340782123e-06, + "loss": 0.2609, + "step": 27725 + }, + { + "epoch": 0.28, + "learning_rate": 6.956648044692737e-06, + "loss": 0.2802, + "step": 27750 + }, + { + "epoch": 0.28, + "learning_rate": 6.953854748603353e-06, + "loss": 0.276, + "step": 27775 + }, + { + "epoch": 0.28, + "learning_rate": 6.9510614525139675e-06, + "loss": 0.2654, + "step": 27800 + }, + { + "epoch": 0.28, + "learning_rate": 6.948268156424581e-06, + "loss": 0.2462, + "step": 27825 + }, + { + "epoch": 0.28, + "learning_rate": 6.945474860335196e-06, + "loss": 0.2796, + "step": 27850 + }, + { + "epoch": 0.28, + "learning_rate": 6.94268156424581e-06, + "loss": 0.2553, + "step": 27875 + }, + { + "epoch": 0.28, + "learning_rate": 6.9398882681564254e-06, + "loss": 0.2987, + "step": 27900 + }, + { + "epoch": 0.28, + "learning_rate": 6.93709497206704e-06, + "loss": 0.2957, + "step": 27925 + }, + { + "epoch": 0.28, + "learning_rate": 6.934301675977654e-06, + "loss": 0.2966, + "step": 27950 + }, + { + "epoch": 0.28, + "learning_rate": 6.931508379888269e-06, + "loss": 0.2994, + "step": 27975 + }, + { + "epoch": 0.28, + "learning_rate": 6.928715083798883e-06, + "loss": 0.2884, + "step": 28000 + }, + { + "epoch": 0.28, + "learning_rate": 6.925921787709497e-06, + "loss": 0.2776, + "step": 28025 + }, + { + "epoch": 0.28, + "learning_rate": 6.923128491620113e-06, + "loss": 0.2618, + "step": 28050 + }, + { + "epoch": 0.28, + "learning_rate": 6.920335195530727e-06, + "loss": 0.252, + "step": 28075 + }, + { + "epoch": 0.28, + "learning_rate": 6.917541899441341e-06, + "loss": 0.2979, + "step": 28100 + }, + { + "epoch": 0.28, + "learning_rate": 6.914748603351956e-06, + "loss": 0.27, + "step": 28125 + }, + { + "epoch": 0.28, + "learning_rate": 6.91195530726257e-06, + "loss": 0.2719, + "step": 28150 + }, + { + "epoch": 0.28, + "learning_rate": 6.9091620111731846e-06, + "loss": 0.2766, + "step": 28175 + }, + { + "epoch": 0.28, + "learning_rate": 6.9063687150838e-06, + "loss": 0.2741, + "step": 28200 + }, + { + "epoch": 0.28, + "learning_rate": 6.903575418994414e-06, + "loss": 0.3138, + "step": 28225 + }, + { + "epoch": 0.28, + "learning_rate": 6.900782122905029e-06, + "loss": 0.2622, + "step": 28250 + }, + { + "epoch": 0.28, + "learning_rate": 6.8979888268156425e-06, + "loss": 0.2803, + "step": 28275 + }, + { + "epoch": 0.28, + "learning_rate": 6.895195530726257e-06, + "loss": 0.3161, + "step": 28300 + }, + { + "epoch": 0.28, + "learning_rate": 6.892402234636873e-06, + "loss": 0.2744, + "step": 28325 + }, + { + "epoch": 0.28, + "learning_rate": 6.889608938547487e-06, + "loss": 0.2949, + "step": 28350 + }, + { + "epoch": 0.28, + "learning_rate": 6.886815642458101e-06, + "loss": 0.301, + "step": 28375 + }, + { + "epoch": 0.28, + "learning_rate": 6.884022346368716e-06, + "loss": 0.2722, + "step": 28400 + }, + { + "epoch": 0.28, + "learning_rate": 6.88122905027933e-06, + "loss": 0.2806, + "step": 28425 + }, + { + "epoch": 0.29, + "learning_rate": 6.8784357541899446e-06, + "loss": 0.2702, + "step": 28450 + }, + { + "epoch": 0.29, + "learning_rate": 6.875642458100558e-06, + "loss": 0.2837, + "step": 28475 + }, + { + "epoch": 0.29, + "learning_rate": 6.872849162011174e-06, + "loss": 0.2661, + "step": 28500 + }, + { + "epoch": 0.29, + "learning_rate": 6.870055865921789e-06, + "loss": 0.2797, + "step": 28525 + }, + { + "epoch": 0.29, + "learning_rate": 6.8672625698324025e-06, + "loss": 0.2719, + "step": 28550 + }, + { + "epoch": 0.29, + "learning_rate": 6.864469273743017e-06, + "loss": 0.2465, + "step": 28575 + }, + { + "epoch": 0.29, + "learning_rate": 6.861675977653633e-06, + "loss": 0.2569, + "step": 28600 + }, + { + "epoch": 0.29, + "learning_rate": 6.858882681564247e-06, + "loss": 0.2721, + "step": 28625 + }, + { + "epoch": 0.29, + "learning_rate": 6.856089385474861e-06, + "loss": 0.2459, + "step": 28650 + }, + { + "epoch": 0.29, + "learning_rate": 6.853296089385475e-06, + "loss": 0.28, + "step": 28675 + }, + { + "epoch": 0.29, + "learning_rate": 6.85050279329609e-06, + "loss": 0.2746, + "step": 28700 + }, + { + "epoch": 0.29, + "learning_rate": 6.8477094972067046e-06, + "loss": 0.2725, + "step": 28725 + }, + { + "epoch": 0.29, + "learning_rate": 6.844916201117318e-06, + "loss": 0.2385, + "step": 28750 + }, + { + "epoch": 0.29, + "learning_rate": 6.842122905027934e-06, + "loss": 0.2476, + "step": 28775 + }, + { + "epoch": 0.29, + "learning_rate": 6.839329608938549e-06, + "loss": 0.2633, + "step": 28800 + }, + { + "epoch": 0.29, + "learning_rate": 6.8365363128491625e-06, + "loss": 0.2403, + "step": 28825 + }, + { + "epoch": 0.29, + "learning_rate": 6.833743016759777e-06, + "loss": 0.2691, + "step": 28850 + }, + { + "epoch": 0.29, + "learning_rate": 6.830949720670391e-06, + "loss": 0.2593, + "step": 28875 + }, + { + "epoch": 0.29, + "learning_rate": 6.828156424581006e-06, + "loss": 0.2645, + "step": 28900 + }, + { + "epoch": 0.29, + "learning_rate": 6.825363128491621e-06, + "loss": 0.2633, + "step": 28925 + }, + { + "epoch": 0.29, + "learning_rate": 6.822569832402235e-06, + "loss": 0.2919, + "step": 28950 + }, + { + "epoch": 0.29, + "learning_rate": 6.81977653631285e-06, + "loss": 0.3, + "step": 28975 + }, + { + "epoch": 0.29, + "learning_rate": 6.816983240223464e-06, + "loss": 0.2945, + "step": 29000 + }, + { + "epoch": 0.29, + "learning_rate": 6.814189944134078e-06, + "loss": 0.2612, + "step": 29025 + }, + { + "epoch": 0.29, + "learning_rate": 6.811396648044694e-06, + "loss": 0.299, + "step": 29050 + }, + { + "epoch": 0.29, + "learning_rate": 6.808603351955308e-06, + "loss": 0.2729, + "step": 29075 + }, + { + "epoch": 0.29, + "learning_rate": 6.8058100558659225e-06, + "loss": 0.3038, + "step": 29100 + }, + { + "epoch": 0.29, + "learning_rate": 6.803016759776537e-06, + "loss": 0.3107, + "step": 29125 + }, + { + "epoch": 0.29, + "learning_rate": 6.800223463687151e-06, + "loss": 0.3022, + "step": 29150 + }, + { + "epoch": 0.29, + "learning_rate": 6.797430167597766e-06, + "loss": 0.2762, + "step": 29175 + }, + { + "epoch": 0.29, + "learning_rate": 6.79463687150838e-06, + "loss": 0.2863, + "step": 29200 + }, + { + "epoch": 0.29, + "learning_rate": 6.791843575418995e-06, + "loss": 0.2754, + "step": 29225 + }, + { + "epoch": 0.29, + "learning_rate": 6.78905027932961e-06, + "loss": 0.3279, + "step": 29250 + }, + { + "epoch": 0.29, + "learning_rate": 6.786256983240224e-06, + "loss": 0.288, + "step": 29275 + }, + { + "epoch": 0.29, + "learning_rate": 6.783463687150838e-06, + "loss": 0.2988, + "step": 29300 + }, + { + "epoch": 0.29, + "learning_rate": 6.780670391061454e-06, + "loss": 0.2827, + "step": 29325 + }, + { + "epoch": 0.29, + "learning_rate": 6.777877094972068e-06, + "loss": 0.2588, + "step": 29350 + }, + { + "epoch": 0.29, + "learning_rate": 6.7750837988826825e-06, + "loss": 0.2379, + "step": 29375 + }, + { + "epoch": 0.29, + "learning_rate": 6.772290502793296e-06, + "loss": 0.2543, + "step": 29400 + }, + { + "epoch": 0.29, + "learning_rate": 6.769497206703911e-06, + "loss": 0.2646, + "step": 29425 + }, + { + "epoch": 0.3, + "learning_rate": 6.766703910614526e-06, + "loss": 0.2595, + "step": 29450 + }, + { + "epoch": 0.3, + "learning_rate": 6.764022346368716e-06, + "loss": 0.282, + "step": 29475 + }, + { + "epoch": 0.3, + "learning_rate": 6.76122905027933e-06, + "loss": 0.2965, + "step": 29500 + }, + { + "epoch": 0.3, + "learning_rate": 6.758435754189945e-06, + "loss": 0.2667, + "step": 29525 + }, + { + "epoch": 0.3, + "learning_rate": 6.7556424581005596e-06, + "loss": 0.2575, + "step": 29550 + }, + { + "epoch": 0.3, + "learning_rate": 6.752849162011173e-06, + "loss": 0.2858, + "step": 29575 + }, + { + "epoch": 0.3, + "learning_rate": 6.750055865921788e-06, + "loss": 0.2632, + "step": 29600 + }, + { + "epoch": 0.3, + "learning_rate": 6.747262569832402e-06, + "loss": 0.2864, + "step": 29625 + }, + { + "epoch": 0.3, + "learning_rate": 6.7444692737430175e-06, + "loss": 0.3084, + "step": 29650 + }, + { + "epoch": 0.3, + "learning_rate": 6.741675977653632e-06, + "loss": 0.2393, + "step": 29675 + }, + { + "epoch": 0.3, + "learning_rate": 6.738882681564246e-06, + "loss": 0.2778, + "step": 29700 + }, + { + "epoch": 0.3, + "learning_rate": 6.736089385474861e-06, + "loss": 0.2682, + "step": 29725 + }, + { + "epoch": 0.3, + "learning_rate": 6.733296089385476e-06, + "loss": 0.2481, + "step": 29750 + }, + { + "epoch": 0.3, + "learning_rate": 6.73050279329609e-06, + "loss": 0.2822, + "step": 29775 + }, + { + "epoch": 0.3, + "learning_rate": 6.727709497206705e-06, + "loss": 0.2375, + "step": 29800 + }, + { + "epoch": 0.3, + "learning_rate": 6.724916201117319e-06, + "loss": 0.2936, + "step": 29825 + }, + { + "epoch": 0.3, + "learning_rate": 6.722122905027933e-06, + "loss": 0.2855, + "step": 29850 + }, + { + "epoch": 0.3, + "learning_rate": 6.719329608938548e-06, + "loss": 0.2717, + "step": 29875 + }, + { + "epoch": 0.3, + "learning_rate": 6.716536312849162e-06, + "loss": 0.2697, + "step": 29900 + }, + { + "epoch": 0.3, + "learning_rate": 6.7137430167597775e-06, + "loss": 0.3121, + "step": 29925 + }, + { + "epoch": 0.3, + "learning_rate": 6.710949720670391e-06, + "loss": 0.2606, + "step": 29950 + }, + { + "epoch": 0.3, + "learning_rate": 6.708156424581006e-06, + "loss": 0.2606, + "step": 29975 + }, + { + "epoch": 0.3, + "learning_rate": 6.705363128491621e-06, + "loss": 0.2805, + "step": 30000 + }, + { + "epoch": 0.3, + "eval_loss": 0.28317904472351074, + "eval_runtime": 1932.087, + "eval_samples_per_second": 2.59, + "eval_steps_per_second": 0.162, + "eval_wer": 21.011414604048507, + "step": 30000 + }, + { + "epoch": 0.3, + "step": 30001, + "total_flos": 1.3852560952000512e+20, + "train_loss": 1.1846045168919276e-05, + "train_runtime": 3.611, + "train_samples_per_second": 33231.537, + "train_steps_per_second": 2076.971 + } + ], + "logging_steps": 25, + "max_steps": 7500, + "num_input_tokens_seen": 0, + "num_train_epochs": 1, + "save_steps": 2500, + "total_flos": 1.3852560952000512e+20, + "train_batch_size": 16, + "trial_name": null, + "trial_params": null +}