{ "best_metric": 9.200601844614663, "best_model_checkpoint": "./checkpoint-40000", "epoch": 401.0025062656642, "eval_steps": 1000, "global_step": 40000, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.25, "learning_rate": 5.000000000000001e-07, "loss": 1.2555, "step": 25 }, { "epoch": 0.5, "learning_rate": 1.0000000000000002e-06, "loss": 0.8832, "step": 50 }, { "epoch": 0.75, "learning_rate": 1.5e-06, "loss": 0.6051, "step": 75 }, { "epoch": 1.0, "learning_rate": 2.0000000000000003e-06, "loss": 0.4886, "step": 100 }, { "epoch": 1.25, "learning_rate": 2.5e-06, "loss": 0.404, "step": 125 }, { "epoch": 1.5, "learning_rate": 3e-06, "loss": 0.3569, "step": 150 }, { "epoch": 1.75, "learning_rate": 3.5e-06, "loss": 0.3149, "step": 175 }, { "epoch": 2.01, "learning_rate": 4.000000000000001e-06, "loss": 0.2838, "step": 200 }, { "epoch": 2.26, "learning_rate": 4.5e-06, "loss": 0.2272, "step": 225 }, { "epoch": 2.51, "learning_rate": 5e-06, "loss": 0.2075, "step": 250 }, { "epoch": 2.76, "learning_rate": 5.500000000000001e-06, "loss": 0.2031, "step": 275 }, { "epoch": 3.01, "learning_rate": 6e-06, "loss": 0.1886, "step": 300 }, { "epoch": 3.26, "learning_rate": 6.5000000000000004e-06, "loss": 0.1328, "step": 325 }, { "epoch": 3.51, "learning_rate": 7e-06, "loss": 0.1279, "step": 350 }, { "epoch": 3.76, "learning_rate": 7.500000000000001e-06, "loss": 0.1257, "step": 375 }, { "epoch": 4.01, "learning_rate": 8.000000000000001e-06, "loss": 0.1241, "step": 400 }, { "epoch": 4.26, "learning_rate": 8.5e-06, "loss": 0.0741, "step": 425 }, { "epoch": 4.51, "learning_rate": 9e-06, "loss": 0.0748, "step": 450 }, { "epoch": 4.76, "learning_rate": 9.5e-06, "loss": 0.0759, "step": 475 }, { "epoch": 5.01, "learning_rate": 1e-05, "loss": 0.0711, "step": 500 }, { "epoch": 5.26, "learning_rate": 9.99367088607595e-06, "loss": 0.0373, "step": 525 }, { "epoch": 5.51, "learning_rate": 9.9873417721519e-06, "loss": 0.0376, "step": 550 }, { "epoch": 5.76, "learning_rate": 9.981012658227849e-06, "loss": 0.0395, "step": 575 }, { "epoch": 6.02, "learning_rate": 9.974683544303799e-06, "loss": 0.04, "step": 600 }, { "epoch": 6.27, "learning_rate": 9.968354430379748e-06, "loss": 0.02, "step": 625 }, { "epoch": 6.52, "learning_rate": 9.962025316455697e-06, "loss": 0.02, "step": 650 }, { "epoch": 6.77, "learning_rate": 9.955696202531647e-06, "loss": 0.0204, "step": 675 }, { "epoch": 7.02, "learning_rate": 9.949367088607596e-06, "loss": 0.02, "step": 700 }, { "epoch": 7.27, "learning_rate": 9.943037974683544e-06, "loss": 0.0106, "step": 725 }, { "epoch": 7.52, "learning_rate": 9.936708860759493e-06, "loss": 0.0106, "step": 750 }, { "epoch": 7.77, "learning_rate": 9.930379746835443e-06, "loss": 0.0104, "step": 775 }, { "epoch": 8.02, "learning_rate": 9.924050632911392e-06, "loss": 0.0101, "step": 800 }, { "epoch": 8.27, "learning_rate": 9.917721518987342e-06, "loss": 0.0068, "step": 825 }, { "epoch": 8.52, "learning_rate": 9.911392405063291e-06, "loss": 0.0059, "step": 850 }, { "epoch": 8.77, "learning_rate": 9.90506329113924e-06, "loss": 0.0081, "step": 875 }, { "epoch": 9.02, "learning_rate": 9.89873417721519e-06, "loss": 0.0079, "step": 900 }, { "epoch": 9.27, "learning_rate": 9.892405063291141e-06, "loss": 0.0055, "step": 925 }, { "epoch": 9.52, "learning_rate": 9.88607594936709e-06, "loss": 0.0055, "step": 950 }, { "epoch": 9.77, "learning_rate": 9.87974683544304e-06, "loss": 0.0058, "step": 975 }, { "epoch": 10.03, "learning_rate": 9.87341772151899e-06, "loss": 0.0055, "step": 1000 }, { "epoch": 10.03, "eval_loss": 0.2462760955095291, "eval_runtime": 1423.839, "eval_samples_per_second": 8.949, "eval_steps_per_second": 0.28, "eval_wer": 11.84246521807097, "step": 1000 }, { "epoch": 10.28, "learning_rate": 9.867088607594937e-06, "loss": 0.0036, "step": 1025 }, { "epoch": 10.53, "learning_rate": 9.860759493670887e-06, "loss": 0.0037, "step": 1050 }, { "epoch": 10.78, "learning_rate": 9.854430379746836e-06, "loss": 0.0037, "step": 1075 }, { "epoch": 11.03, "learning_rate": 9.848101265822785e-06, "loss": 0.004, "step": 1100 }, { "epoch": 11.28, "learning_rate": 9.841772151898735e-06, "loss": 0.0033, "step": 1125 }, { "epoch": 11.53, "learning_rate": 9.835443037974684e-06, "loss": 0.0029, "step": 1150 }, { "epoch": 11.78, "learning_rate": 9.829113924050634e-06, "loss": 0.0032, "step": 1175 }, { "epoch": 12.03, "learning_rate": 9.822784810126583e-06, "loss": 0.0034, "step": 1200 }, { "epoch": 12.28, "learning_rate": 9.816455696202533e-06, "loss": 0.0022, "step": 1225 }, { "epoch": 12.53, "learning_rate": 9.810126582278482e-06, "loss": 0.0024, "step": 1250 }, { "epoch": 12.78, "learning_rate": 9.803797468354431e-06, "loss": 0.003, "step": 1275 }, { "epoch": 13.03, "learning_rate": 9.79746835443038e-06, "loss": 0.0036, "step": 1300 }, { "epoch": 13.28, "learning_rate": 9.79113924050633e-06, "loss": 0.0034, "step": 1325 }, { "epoch": 13.53, "learning_rate": 9.78481012658228e-06, "loss": 0.0036, "step": 1350 }, { "epoch": 13.78, "learning_rate": 9.778481012658229e-06, "loss": 0.0039, "step": 1375 }, { "epoch": 14.04, "learning_rate": 9.772151898734179e-06, "loss": 0.0047, "step": 1400 }, { "epoch": 14.29, "learning_rate": 9.765822784810128e-06, "loss": 0.0039, "step": 1425 }, { "epoch": 14.54, "learning_rate": 9.759493670886077e-06, "loss": 0.0038, "step": 1450 }, { "epoch": 14.79, "learning_rate": 9.753164556962025e-06, "loss": 0.0039, "step": 1475 }, { "epoch": 15.04, "learning_rate": 9.746835443037975e-06, "loss": 0.0029, "step": 1500 }, { "epoch": 15.29, "learning_rate": 9.740506329113924e-06, "loss": 0.0021, "step": 1525 }, { "epoch": 15.54, "learning_rate": 9.734177215189873e-06, "loss": 0.0024, "step": 1550 }, { "epoch": 15.79, "learning_rate": 9.727848101265823e-06, "loss": 0.0026, "step": 1575 }, { "epoch": 16.04, "learning_rate": 9.721518987341772e-06, "loss": 0.0027, "step": 1600 }, { "epoch": 16.29, "learning_rate": 9.715189873417722e-06, "loss": 0.0025, "step": 1625 }, { "epoch": 16.54, "learning_rate": 9.708860759493671e-06, "loss": 0.0022, "step": 1650 }, { "epoch": 16.79, "learning_rate": 9.70253164556962e-06, "loss": 0.0025, "step": 1675 }, { "epoch": 17.04, "learning_rate": 9.69620253164557e-06, "loss": 0.0025, "step": 1700 }, { "epoch": 17.29, "learning_rate": 9.68987341772152e-06, "loss": 0.0021, "step": 1725 }, { "epoch": 17.54, "learning_rate": 9.68354430379747e-06, "loss": 0.0032, "step": 1750 }, { "epoch": 17.79, "learning_rate": 9.677215189873418e-06, "loss": 0.0029, "step": 1775 }, { "epoch": 18.05, "learning_rate": 9.670886075949368e-06, "loss": 0.0027, "step": 1800 }, { "epoch": 18.3, "learning_rate": 9.664556962025317e-06, "loss": 0.0022, "step": 1825 }, { "epoch": 18.55, "learning_rate": 9.658227848101267e-06, "loss": 0.0038, "step": 1850 }, { "epoch": 18.8, "learning_rate": 9.651898734177216e-06, "loss": 0.0029, "step": 1875 }, { "epoch": 19.05, "learning_rate": 9.645569620253165e-06, "loss": 0.0027, "step": 1900 }, { "epoch": 19.3, "learning_rate": 9.639240506329115e-06, "loss": 0.0025, "step": 1925 }, { "epoch": 19.55, "learning_rate": 9.632911392405064e-06, "loss": 0.0026, "step": 1950 }, { "epoch": 19.8, "learning_rate": 9.626582278481014e-06, "loss": 0.0025, "step": 1975 }, { "epoch": 20.05, "learning_rate": 9.620253164556963e-06, "loss": 0.003, "step": 2000 }, { "epoch": 20.05, "eval_loss": 0.2637926936149597, "eval_runtime": 1408.7067, "eval_samples_per_second": 9.045, "eval_steps_per_second": 0.283, "eval_wer": 11.317805221197437, "step": 2000 }, { "epoch": 20.3, "learning_rate": 9.613924050632913e-06, "loss": 0.0023, "step": 2025 }, { "epoch": 20.55, "learning_rate": 9.607594936708862e-06, "loss": 0.0021, "step": 2050 }, { "epoch": 20.8, "learning_rate": 9.601265822784811e-06, "loss": 0.0025, "step": 2075 }, { "epoch": 21.05, "learning_rate": 9.59493670886076e-06, "loss": 0.0023, "step": 2100 }, { "epoch": 21.3, "learning_rate": 9.58860759493671e-06, "loss": 0.0017, "step": 2125 }, { "epoch": 21.55, "learning_rate": 9.58227848101266e-06, "loss": 0.0016, "step": 2150 }, { "epoch": 21.8, "learning_rate": 9.575949367088609e-06, "loss": 0.003, "step": 2175 }, { "epoch": 22.06, "learning_rate": 9.569620253164559e-06, "loss": 0.0028, "step": 2200 }, { "epoch": 22.31, "learning_rate": 9.563291139240506e-06, "loss": 0.0023, "step": 2225 }, { "epoch": 22.56, "learning_rate": 9.556962025316456e-06, "loss": 0.0027, "step": 2250 }, { "epoch": 22.81, "learning_rate": 9.550632911392405e-06, "loss": 0.0033, "step": 2275 }, { "epoch": 23.06, "learning_rate": 9.544303797468355e-06, "loss": 0.0035, "step": 2300 }, { "epoch": 23.31, "learning_rate": 9.537974683544304e-06, "loss": 0.0033, "step": 2325 }, { "epoch": 23.56, "learning_rate": 9.531645569620253e-06, "loss": 0.0039, "step": 2350 }, { "epoch": 23.81, "learning_rate": 9.525316455696203e-06, "loss": 0.0043, "step": 2375 }, { "epoch": 24.06, "learning_rate": 9.518987341772152e-06, "loss": 0.0041, "step": 2400 }, { "epoch": 24.31, "learning_rate": 9.512658227848102e-06, "loss": 0.0038, "step": 2425 }, { "epoch": 24.56, "learning_rate": 9.506329113924051e-06, "loss": 0.0029, "step": 2450 }, { "epoch": 24.81, "learning_rate": 9.5e-06, "loss": 0.0036, "step": 2475 }, { "epoch": 25.06, "learning_rate": 9.49367088607595e-06, "loss": 0.0033, "step": 2500 }, { "epoch": 25.31, "learning_rate": 9.4873417721519e-06, "loss": 0.0025, "step": 2525 }, { "epoch": 25.56, "learning_rate": 9.481012658227849e-06, "loss": 0.0026, "step": 2550 }, { "epoch": 25.81, "learning_rate": 9.474683544303798e-06, "loss": 0.0029, "step": 2575 }, { "epoch": 26.07, "learning_rate": 9.468354430379748e-06, "loss": 0.0032, "step": 2600 }, { "epoch": 26.32, "learning_rate": 9.462025316455697e-06, "loss": 0.0022, "step": 2625 }, { "epoch": 26.57, "learning_rate": 9.455696202531647e-06, "loss": 0.0019, "step": 2650 }, { "epoch": 26.82, "learning_rate": 9.449367088607596e-06, "loss": 0.0016, "step": 2675 }, { "epoch": 27.07, "learning_rate": 9.443037974683545e-06, "loss": 0.002, "step": 2700 }, { "epoch": 27.32, "learning_rate": 9.436708860759495e-06, "loss": 0.0013, "step": 2725 }, { "epoch": 27.57, "learning_rate": 9.430379746835444e-06, "loss": 0.001, "step": 2750 }, { "epoch": 27.82, "learning_rate": 9.424050632911394e-06, "loss": 0.0011, "step": 2775 }, { "epoch": 28.07, "learning_rate": 9.417721518987343e-06, "loss": 0.0012, "step": 2800 }, { "epoch": 28.32, "learning_rate": 9.411392405063293e-06, "loss": 0.0008, "step": 2825 }, { "epoch": 28.57, "learning_rate": 9.405063291139242e-06, "loss": 0.0011, "step": 2850 }, { "epoch": 28.82, "learning_rate": 9.398734177215191e-06, "loss": 0.0012, "step": 2875 }, { "epoch": 29.07, "learning_rate": 9.39240506329114e-06, "loss": 0.0011, "step": 2900 }, { "epoch": 29.32, "learning_rate": 9.38607594936709e-06, "loss": 0.0007, "step": 2925 }, { "epoch": 29.57, "learning_rate": 9.379746835443038e-06, "loss": 0.0011, "step": 2950 }, { "epoch": 29.82, "learning_rate": 9.373417721518987e-06, "loss": 0.0014, "step": 2975 }, { "epoch": 30.08, "learning_rate": 9.367088607594937e-06, "loss": 0.0018, "step": 3000 }, { "epoch": 30.08, "eval_loss": 0.2836576998233795, "eval_runtime": 1408.2375, "eval_samples_per_second": 9.048, "eval_steps_per_second": 0.283, "eval_wer": 10.958261685164922, "step": 3000 }, { "epoch": 30.33, "learning_rate": 9.360759493670886e-06, "loss": 0.0015, "step": 3025 }, { "epoch": 30.58, "learning_rate": 9.354430379746836e-06, "loss": 0.0013, "step": 3050 }, { "epoch": 30.83, "learning_rate": 9.348101265822785e-06, "loss": 0.0017, "step": 3075 }, { "epoch": 31.08, "learning_rate": 9.341772151898735e-06, "loss": 0.002, "step": 3100 }, { "epoch": 31.33, "learning_rate": 9.335443037974684e-06, "loss": 0.0025, "step": 3125 }, { "epoch": 31.58, "learning_rate": 9.329113924050633e-06, "loss": 0.0018, "step": 3150 }, { "epoch": 31.83, "learning_rate": 9.322784810126583e-06, "loss": 0.0017, "step": 3175 }, { "epoch": 32.08, "learning_rate": 9.316455696202532e-06, "loss": 0.0018, "step": 3200 }, { "epoch": 32.33, "learning_rate": 9.310126582278482e-06, "loss": 0.0015, "step": 3225 }, { "epoch": 32.58, "learning_rate": 9.303797468354431e-06, "loss": 0.0016, "step": 3250 }, { "epoch": 32.83, "learning_rate": 9.29746835443038e-06, "loss": 0.0022, "step": 3275 }, { "epoch": 33.08, "learning_rate": 9.29113924050633e-06, "loss": 0.0026, "step": 3300 }, { "epoch": 33.33, "learning_rate": 9.28481012658228e-06, "loss": 0.0017, "step": 3325 }, { "epoch": 33.58, "learning_rate": 9.278481012658229e-06, "loss": 0.0024, "step": 3350 }, { "epoch": 33.83, "learning_rate": 9.272151898734178e-06, "loss": 0.0024, "step": 3375 }, { "epoch": 34.09, "learning_rate": 9.265822784810128e-06, "loss": 0.0021, "step": 3400 }, { "epoch": 34.34, "learning_rate": 9.259493670886077e-06, "loss": 0.0016, "step": 3425 }, { "epoch": 34.59, "learning_rate": 9.253164556962027e-06, "loss": 0.0021, "step": 3450 }, { "epoch": 34.84, "learning_rate": 9.246835443037976e-06, "loss": 0.0019, "step": 3475 }, { "epoch": 35.09, "learning_rate": 9.240506329113925e-06, "loss": 0.0023, "step": 3500 }, { "epoch": 35.34, "learning_rate": 9.234177215189875e-06, "loss": 0.0022, "step": 3525 }, { "epoch": 35.59, "learning_rate": 9.227848101265824e-06, "loss": 0.002, "step": 3550 }, { "epoch": 35.84, "learning_rate": 9.221518987341774e-06, "loss": 0.0017, "step": 3575 }, { "epoch": 36.09, "learning_rate": 9.215189873417723e-06, "loss": 0.0014, "step": 3600 }, { "epoch": 36.34, "learning_rate": 9.208860759493673e-06, "loss": 0.0026, "step": 3625 }, { "epoch": 36.59, "learning_rate": 9.202531645569622e-06, "loss": 0.0012, "step": 3650 }, { "epoch": 36.84, "learning_rate": 9.196202531645571e-06, "loss": 0.0018, "step": 3675 }, { "epoch": 37.09, "learning_rate": 9.189873417721519e-06, "loss": 0.0016, "step": 3700 }, { "epoch": 37.34, "learning_rate": 9.183544303797469e-06, "loss": 0.0009, "step": 3725 }, { "epoch": 37.59, "learning_rate": 9.177215189873418e-06, "loss": 0.001, "step": 3750 }, { "epoch": 37.84, "learning_rate": 9.170886075949367e-06, "loss": 0.0012, "step": 3775 }, { "epoch": 38.1, "learning_rate": 9.164556962025317e-06, "loss": 0.0013, "step": 3800 }, { "epoch": 38.35, "learning_rate": 9.158227848101266e-06, "loss": 0.0007, "step": 3825 }, { "epoch": 38.6, "learning_rate": 9.151898734177216e-06, "loss": 0.0008, "step": 3850 }, { "epoch": 38.85, "learning_rate": 9.145569620253165e-06, "loss": 0.0008, "step": 3875 }, { "epoch": 39.1, "learning_rate": 9.139240506329115e-06, "loss": 0.0007, "step": 3900 }, { "epoch": 39.35, "learning_rate": 9.132911392405064e-06, "loss": 0.0006, "step": 3925 }, { "epoch": 39.6, "learning_rate": 9.126582278481013e-06, "loss": 0.0005, "step": 3950 }, { "epoch": 39.85, "learning_rate": 9.120253164556963e-06, "loss": 0.0008, "step": 3975 }, { "epoch": 40.1, "learning_rate": 9.113924050632912e-06, "loss": 0.0009, "step": 4000 }, { "epoch": 40.1, "eval_loss": 0.27675050497055054, "eval_runtime": 1409.2926, "eval_samples_per_second": 9.041, "eval_steps_per_second": 0.283, "eval_wer": 10.44141785211818, "step": 4000 }, { "epoch": 40.35, "learning_rate": 9.107594936708862e-06, "loss": 0.0012, "step": 4025 }, { "epoch": 40.6, "learning_rate": 9.101265822784811e-06, "loss": 0.0013, "step": 4050 }, { "epoch": 40.85, "learning_rate": 9.09493670886076e-06, "loss": 0.0017, "step": 4075 }, { "epoch": 41.1, "learning_rate": 9.08860759493671e-06, "loss": 0.0022, "step": 4100 }, { "epoch": 41.35, "learning_rate": 9.08227848101266e-06, "loss": 0.0018, "step": 4125 }, { "epoch": 41.6, "learning_rate": 9.075949367088607e-06, "loss": 0.0015, "step": 4150 }, { "epoch": 41.85, "learning_rate": 9.069620253164557e-06, "loss": 0.002, "step": 4175 }, { "epoch": 42.11, "learning_rate": 9.063291139240506e-06, "loss": 0.0018, "step": 4200 }, { "epoch": 42.36, "learning_rate": 9.056962025316457e-06, "loss": 0.0021, "step": 4225 }, { "epoch": 42.61, "learning_rate": 9.050632911392407e-06, "loss": 0.0019, "step": 4250 }, { "epoch": 42.86, "learning_rate": 9.044303797468356e-06, "loss": 0.0027, "step": 4275 }, { "epoch": 43.11, "learning_rate": 9.037974683544305e-06, "loss": 0.0029, "step": 4300 }, { "epoch": 43.36, "learning_rate": 9.031645569620255e-06, "loss": 0.0025, "step": 4325 }, { "epoch": 43.61, "learning_rate": 9.025316455696204e-06, "loss": 0.0022, "step": 4350 }, { "epoch": 43.86, "learning_rate": 9.018987341772154e-06, "loss": 0.0023, "step": 4375 }, { "epoch": 44.11, "learning_rate": 9.012658227848103e-06, "loss": 0.0024, "step": 4400 }, { "epoch": 44.36, "learning_rate": 9.00632911392405e-06, "loss": 0.0023, "step": 4425 }, { "epoch": 44.61, "learning_rate": 9e-06, "loss": 0.0024, "step": 4450 }, { "epoch": 44.86, "learning_rate": 8.99367088607595e-06, "loss": 0.0023, "step": 4475 }, { "epoch": 45.11, "learning_rate": 8.987341772151899e-06, "loss": 0.002, "step": 4500 }, { "epoch": 45.36, "learning_rate": 8.981012658227849e-06, "loss": 0.0016, "step": 4525 }, { "epoch": 45.61, "learning_rate": 8.974683544303798e-06, "loss": 0.0019, "step": 4550 }, { "epoch": 45.86, "learning_rate": 8.968354430379747e-06, "loss": 0.0017, "step": 4575 }, { "epoch": 46.12, "learning_rate": 8.962025316455697e-06, "loss": 0.0016, "step": 4600 }, { "epoch": 46.37, "learning_rate": 8.955696202531646e-06, "loss": 0.0015, "step": 4625 }, { "epoch": 46.62, "learning_rate": 8.949367088607596e-06, "loss": 0.0012, "step": 4650 }, { "epoch": 46.87, "learning_rate": 8.943037974683545e-06, "loss": 0.0015, "step": 4675 }, { "epoch": 47.12, "learning_rate": 8.936708860759495e-06, "loss": 0.0015, "step": 4700 }, { "epoch": 47.37, "learning_rate": 8.930379746835444e-06, "loss": 0.0014, "step": 4725 }, { "epoch": 47.62, "learning_rate": 8.924050632911393e-06, "loss": 0.0017, "step": 4750 }, { "epoch": 47.87, "learning_rate": 8.917721518987343e-06, "loss": 0.0015, "step": 4775 }, { "epoch": 48.12, "learning_rate": 8.911392405063292e-06, "loss": 0.0015, "step": 4800 }, { "epoch": 48.37, "learning_rate": 8.905063291139242e-06, "loss": 0.0013, "step": 4825 }, { "epoch": 48.62, "learning_rate": 8.898734177215191e-06, "loss": 0.0016, "step": 4850 }, { "epoch": 48.87, "learning_rate": 8.892405063291139e-06, "loss": 0.0011, "step": 4875 }, { "epoch": 49.12, "learning_rate": 8.886075949367088e-06, "loss": 0.0012, "step": 4900 }, { "epoch": 49.37, "learning_rate": 8.879746835443038e-06, "loss": 0.0011, "step": 4925 }, { "epoch": 49.62, "learning_rate": 8.873417721518987e-06, "loss": 0.0011, "step": 4950 }, { "epoch": 49.87, "learning_rate": 8.867088607594937e-06, "loss": 0.0009, "step": 4975 }, { "epoch": 50.13, "learning_rate": 8.860759493670886e-06, "loss": 0.0008, "step": 5000 }, { "epoch": 50.13, "eval_loss": 0.28795620799064636, "eval_runtime": 1407.4493, "eval_samples_per_second": 9.053, "eval_steps_per_second": 0.283, "eval_wer": 10.17762232296389, "step": 5000 }, { "epoch": 50.38, "learning_rate": 8.854430379746835e-06, "loss": 0.0008, "step": 5025 }, { "epoch": 50.63, "learning_rate": 8.848101265822786e-06, "loss": 0.0006, "step": 5050 }, { "epoch": 50.88, "learning_rate": 8.841772151898736e-06, "loss": 0.0007, "step": 5075 }, { "epoch": 51.13, "learning_rate": 8.835443037974685e-06, "loss": 0.0011, "step": 5100 }, { "epoch": 51.38, "learning_rate": 8.829113924050635e-06, "loss": 0.0007, "step": 5125 }, { "epoch": 51.63, "learning_rate": 8.822784810126584e-06, "loss": 0.0009, "step": 5150 }, { "epoch": 51.88, "learning_rate": 8.816455696202532e-06, "loss": 0.0006, "step": 5175 }, { "epoch": 52.13, "learning_rate": 8.810126582278481e-06, "loss": 0.0005, "step": 5200 }, { "epoch": 52.38, "learning_rate": 8.80379746835443e-06, "loss": 0.0003, "step": 5225 }, { "epoch": 52.63, "learning_rate": 8.79746835443038e-06, "loss": 0.0006, "step": 5250 }, { "epoch": 52.88, "learning_rate": 8.79113924050633e-06, "loss": 0.0006, "step": 5275 }, { "epoch": 53.13, "learning_rate": 8.784810126582279e-06, "loss": 0.0005, "step": 5300 }, { "epoch": 53.38, "learning_rate": 8.778481012658229e-06, "loss": 0.0006, "step": 5325 }, { "epoch": 53.63, "learning_rate": 8.772151898734178e-06, "loss": 0.0005, "step": 5350 }, { "epoch": 53.88, "learning_rate": 8.765822784810127e-06, "loss": 0.0005, "step": 5375 }, { "epoch": 54.14, "learning_rate": 8.759493670886077e-06, "loss": 0.0005, "step": 5400 }, { "epoch": 54.39, "learning_rate": 8.753164556962026e-06, "loss": 0.0005, "step": 5425 }, { "epoch": 54.64, "learning_rate": 8.746835443037976e-06, "loss": 0.0005, "step": 5450 }, { "epoch": 54.89, "learning_rate": 8.740506329113925e-06, "loss": 0.0008, "step": 5475 }, { "epoch": 55.14, "learning_rate": 8.734177215189874e-06, "loss": 0.0007, "step": 5500 }, { "epoch": 55.39, "learning_rate": 8.727848101265824e-06, "loss": 0.0007, "step": 5525 }, { "epoch": 55.64, "learning_rate": 8.721518987341773e-06, "loss": 0.0007, "step": 5550 }, { "epoch": 55.89, "learning_rate": 8.715189873417723e-06, "loss": 0.0007, "step": 5575 }, { "epoch": 56.14, "learning_rate": 8.708860759493672e-06, "loss": 0.0006, "step": 5600 }, { "epoch": 56.39, "learning_rate": 8.70253164556962e-06, "loss": 0.0005, "step": 5625 }, { "epoch": 56.64, "learning_rate": 8.69620253164557e-06, "loss": 0.0008, "step": 5650 }, { "epoch": 56.89, "learning_rate": 8.689873417721519e-06, "loss": 0.0013, "step": 5675 }, { "epoch": 57.14, "learning_rate": 8.683544303797468e-06, "loss": 0.0017, "step": 5700 }, { "epoch": 57.39, "learning_rate": 8.677215189873418e-06, "loss": 0.0014, "step": 5725 }, { "epoch": 57.64, "learning_rate": 8.670886075949367e-06, "loss": 0.0015, "step": 5750 }, { "epoch": 57.89, "learning_rate": 8.664556962025317e-06, "loss": 0.0018, "step": 5775 }, { "epoch": 58.15, "learning_rate": 8.658227848101266e-06, "loss": 0.0018, "step": 5800 }, { "epoch": 58.4, "learning_rate": 8.651898734177215e-06, "loss": 0.0019, "step": 5825 }, { "epoch": 58.65, "learning_rate": 8.645569620253166e-06, "loss": 0.002, "step": 5850 }, { "epoch": 58.9, "learning_rate": 8.639240506329116e-06, "loss": 0.0017, "step": 5875 }, { "epoch": 59.15, "learning_rate": 8.632911392405064e-06, "loss": 0.0011, "step": 5900 }, { "epoch": 59.4, "learning_rate": 8.626582278481013e-06, "loss": 0.0009, "step": 5925 }, { "epoch": 59.65, "learning_rate": 8.620253164556963e-06, "loss": 0.0009, "step": 5950 }, { "epoch": 59.9, "learning_rate": 8.613924050632912e-06, "loss": 0.0014, "step": 5975 }, { "epoch": 60.15, "learning_rate": 8.607594936708861e-06, "loss": 0.0012, "step": 6000 }, { "epoch": 60.15, "eval_loss": 0.29029330611228943, "eval_runtime": 1406.9815, "eval_samples_per_second": 9.056, "eval_steps_per_second": 0.284, "eval_wer": 10.052563701735188, "step": 6000 }, { "epoch": 60.4, "learning_rate": 8.60126582278481e-06, "loss": 0.0007, "step": 6025 }, { "epoch": 60.65, "learning_rate": 8.59493670886076e-06, "loss": 0.0009, "step": 6050 }, { "epoch": 60.9, "learning_rate": 8.58860759493671e-06, "loss": 0.0011, "step": 6075 }, { "epoch": 61.15, "learning_rate": 8.582278481012659e-06, "loss": 0.0014, "step": 6100 }, { "epoch": 61.4, "learning_rate": 8.575949367088608e-06, "loss": 0.001, "step": 6125 }, { "epoch": 61.65, "learning_rate": 8.569620253164558e-06, "loss": 0.0011, "step": 6150 }, { "epoch": 61.9, "learning_rate": 8.563291139240507e-06, "loss": 0.0012, "step": 6175 }, { "epoch": 62.16, "learning_rate": 8.556962025316457e-06, "loss": 0.0013, "step": 6200 }, { "epoch": 62.41, "learning_rate": 8.550632911392406e-06, "loss": 0.0009, "step": 6225 }, { "epoch": 62.66, "learning_rate": 8.544303797468356e-06, "loss": 0.0014, "step": 6250 }, { "epoch": 62.91, "learning_rate": 8.537974683544305e-06, "loss": 0.0019, "step": 6275 }, { "epoch": 63.16, "learning_rate": 8.531645569620254e-06, "loss": 0.0016, "step": 6300 }, { "epoch": 63.41, "learning_rate": 8.525316455696204e-06, "loss": 0.0017, "step": 6325 }, { "epoch": 63.66, "learning_rate": 8.518987341772152e-06, "loss": 0.0024, "step": 6350 }, { "epoch": 63.91, "learning_rate": 8.512658227848101e-06, "loss": 0.0021, "step": 6375 }, { "epoch": 64.16, "learning_rate": 8.50632911392405e-06, "loss": 0.0022, "step": 6400 }, { "epoch": 64.41, "learning_rate": 8.5e-06, "loss": 0.0017, "step": 6425 }, { "epoch": 64.66, "learning_rate": 8.49367088607595e-06, "loss": 0.0018, "step": 6450 }, { "epoch": 64.91, "learning_rate": 8.487341772151899e-06, "loss": 0.0021, "step": 6475 }, { "epoch": 65.16, "learning_rate": 8.481012658227848e-06, "loss": 0.0014, "step": 6500 }, { "epoch": 65.41, "learning_rate": 8.474683544303798e-06, "loss": 0.0016, "step": 6525 }, { "epoch": 65.66, "learning_rate": 8.468354430379747e-06, "loss": 0.0011, "step": 6550 }, { "epoch": 65.91, "learning_rate": 8.462025316455696e-06, "loss": 0.0012, "step": 6575 }, { "epoch": 66.17, "learning_rate": 8.455696202531646e-06, "loss": 0.0008, "step": 6600 }, { "epoch": 66.42, "learning_rate": 8.449367088607595e-06, "loss": 0.0006, "step": 6625 }, { "epoch": 66.67, "learning_rate": 8.443037974683545e-06, "loss": 0.0005, "step": 6650 }, { "epoch": 66.92, "learning_rate": 8.436708860759494e-06, "loss": 0.0007, "step": 6675 }, { "epoch": 67.17, "learning_rate": 8.430379746835444e-06, "loss": 0.0005, "step": 6700 }, { "epoch": 67.42, "learning_rate": 8.424050632911393e-06, "loss": 0.0008, "step": 6725 }, { "epoch": 67.67, "learning_rate": 8.417721518987342e-06, "loss": 0.0008, "step": 6750 }, { "epoch": 67.92, "learning_rate": 8.411392405063292e-06, "loss": 0.0014, "step": 6775 }, { "epoch": 68.17, "learning_rate": 8.405063291139241e-06, "loss": 0.0012, "step": 6800 }, { "epoch": 68.42, "learning_rate": 8.39873417721519e-06, "loss": 0.0012, "step": 6825 }, { "epoch": 68.67, "learning_rate": 8.39240506329114e-06, "loss": 0.0009, "step": 6850 }, { "epoch": 68.92, "learning_rate": 8.38607594936709e-06, "loss": 0.0009, "step": 6875 }, { "epoch": 69.17, "learning_rate": 8.379746835443039e-06, "loss": 0.0008, "step": 6900 }, { "epoch": 69.42, "learning_rate": 8.373417721518988e-06, "loss": 0.0005, "step": 6925 }, { "epoch": 69.67, "learning_rate": 8.367088607594938e-06, "loss": 0.0004, "step": 6950 }, { "epoch": 69.92, "learning_rate": 8.360759493670887e-06, "loss": 0.0004, "step": 6975 }, { "epoch": 70.18, "learning_rate": 8.354430379746837e-06, "loss": 0.0002, "step": 7000 }, { "epoch": 70.18, "eval_loss": 0.29092082381248474, "eval_runtime": 1406.0159, "eval_samples_per_second": 9.062, "eval_steps_per_second": 0.284, "eval_wer": 9.83566515554166, "step": 7000 }, { "epoch": 70.43, "learning_rate": 8.348101265822786e-06, "loss": 0.0002, "step": 7025 }, { "epoch": 70.68, "learning_rate": 8.341772151898736e-06, "loss": 0.0003, "step": 7050 }, { "epoch": 70.93, "learning_rate": 8.335443037974685e-06, "loss": 0.0003, "step": 7075 }, { "epoch": 71.18, "learning_rate": 8.329113924050633e-06, "loss": 0.0002, "step": 7100 }, { "epoch": 71.43, "learning_rate": 8.322784810126582e-06, "loss": 0.0004, "step": 7125 }, { "epoch": 71.68, "learning_rate": 8.316455696202532e-06, "loss": 0.0004, "step": 7150 }, { "epoch": 71.93, "learning_rate": 8.310126582278481e-06, "loss": 0.0003, "step": 7175 }, { "epoch": 72.18, "learning_rate": 8.30379746835443e-06, "loss": 0.0003, "step": 7200 }, { "epoch": 72.43, "learning_rate": 8.29746835443038e-06, "loss": 0.0003, "step": 7225 }, { "epoch": 72.68, "learning_rate": 8.29113924050633e-06, "loss": 0.0005, "step": 7250 }, { "epoch": 72.93, "learning_rate": 8.284810126582279e-06, "loss": 0.0005, "step": 7275 }, { "epoch": 73.18, "learning_rate": 8.278481012658228e-06, "loss": 0.0005, "step": 7300 }, { "epoch": 73.43, "learning_rate": 8.272151898734178e-06, "loss": 0.0003, "step": 7325 }, { "epoch": 73.68, "learning_rate": 8.265822784810127e-06, "loss": 0.0002, "step": 7350 }, { "epoch": 73.93, "learning_rate": 8.259493670886076e-06, "loss": 0.0002, "step": 7375 }, { "epoch": 74.19, "learning_rate": 8.253164556962026e-06, "loss": 0.0002, "step": 7400 }, { "epoch": 74.44, "learning_rate": 8.246835443037975e-06, "loss": 0.0002, "step": 7425 }, { "epoch": 74.69, "learning_rate": 8.240506329113925e-06, "loss": 0.0002, "step": 7450 }, { "epoch": 74.94, "learning_rate": 8.234177215189874e-06, "loss": 0.0003, "step": 7475 }, { "epoch": 75.19, "learning_rate": 8.227848101265824e-06, "loss": 0.0003, "step": 7500 }, { "epoch": 75.44, "learning_rate": 8.221518987341773e-06, "loss": 0.0003, "step": 7525 }, { "epoch": 75.69, "learning_rate": 8.215189873417722e-06, "loss": 0.0003, "step": 7550 }, { "epoch": 75.94, "learning_rate": 8.208860759493672e-06, "loss": 0.0004, "step": 7575 }, { "epoch": 76.19, "learning_rate": 8.202531645569621e-06, "loss": 0.0008, "step": 7600 }, { "epoch": 76.44, "learning_rate": 8.19620253164557e-06, "loss": 0.0006, "step": 7625 }, { "epoch": 76.69, "learning_rate": 8.18987341772152e-06, "loss": 0.0009, "step": 7650 }, { "epoch": 76.94, "learning_rate": 8.18354430379747e-06, "loss": 0.0008, "step": 7675 }, { "epoch": 77.19, "learning_rate": 8.177215189873419e-06, "loss": 0.0009, "step": 7700 }, { "epoch": 77.44, "learning_rate": 8.170886075949368e-06, "loss": 0.0016, "step": 7725 }, { "epoch": 77.69, "learning_rate": 8.164556962025318e-06, "loss": 0.0019, "step": 7750 }, { "epoch": 77.94, "learning_rate": 8.158227848101267e-06, "loss": 0.0026, "step": 7775 }, { "epoch": 78.2, "learning_rate": 8.151898734177217e-06, "loss": 0.0037, "step": 7800 }, { "epoch": 78.45, "learning_rate": 8.145569620253164e-06, "loss": 0.0028, "step": 7825 }, { "epoch": 78.7, "learning_rate": 8.139240506329114e-06, "loss": 0.0026, "step": 7850 }, { "epoch": 78.95, "learning_rate": 8.132911392405063e-06, "loss": 0.0026, "step": 7875 }, { "epoch": 79.2, "learning_rate": 8.126582278481013e-06, "loss": 0.0029, "step": 7900 }, { "epoch": 79.45, "learning_rate": 8.120253164556962e-06, "loss": 0.0025, "step": 7925 }, { "epoch": 79.7, "learning_rate": 8.113924050632912e-06, "loss": 0.0019, "step": 7950 }, { "epoch": 79.95, "learning_rate": 8.107594936708861e-06, "loss": 0.0017, "step": 7975 }, { "epoch": 80.2, "learning_rate": 8.10126582278481e-06, "loss": 0.0013, "step": 8000 }, { "epoch": 80.2, "eval_loss": 0.27660274505615234, "eval_runtime": 1408.4166, "eval_samples_per_second": 9.047, "eval_steps_per_second": 0.283, "eval_wer": 9.939229326246677, "step": 8000 }, { "epoch": 80.45, "learning_rate": 8.09493670886076e-06, "loss": 0.0014, "step": 8025 }, { "epoch": 80.7, "learning_rate": 8.08860759493671e-06, "loss": 0.001, "step": 8050 }, { "epoch": 80.95, "learning_rate": 8.082278481012659e-06, "loss": 0.0009, "step": 8075 }, { "epoch": 81.2, "learning_rate": 8.075949367088608e-06, "loss": 0.0006, "step": 8100 }, { "epoch": 81.45, "learning_rate": 8.069620253164558e-06, "loss": 0.0007, "step": 8125 }, { "epoch": 81.7, "learning_rate": 8.063291139240507e-06, "loss": 0.0008, "step": 8150 }, { "epoch": 81.95, "learning_rate": 8.056962025316456e-06, "loss": 0.0007, "step": 8175 }, { "epoch": 82.21, "learning_rate": 8.050632911392406e-06, "loss": 0.0012, "step": 8200 }, { "epoch": 82.46, "learning_rate": 8.044303797468355e-06, "loss": 0.0005, "step": 8225 }, { "epoch": 82.71, "learning_rate": 8.037974683544305e-06, "loss": 0.0005, "step": 8250 }, { "epoch": 82.96, "learning_rate": 8.031645569620252e-06, "loss": 0.0006, "step": 8275 }, { "epoch": 83.21, "learning_rate": 8.025316455696202e-06, "loss": 0.0004, "step": 8300 }, { "epoch": 83.46, "learning_rate": 8.018987341772153e-06, "loss": 0.0003, "step": 8325 }, { "epoch": 83.71, "learning_rate": 8.012658227848102e-06, "loss": 0.0005, "step": 8350 }, { "epoch": 83.96, "learning_rate": 8.006329113924052e-06, "loss": 0.0005, "step": 8375 }, { "epoch": 84.21, "learning_rate": 8.000000000000001e-06, "loss": 0.0004, "step": 8400 }, { "epoch": 84.46, "learning_rate": 7.99367088607595e-06, "loss": 0.0005, "step": 8425 }, { "epoch": 84.71, "learning_rate": 7.9873417721519e-06, "loss": 0.0003, "step": 8450 }, { "epoch": 84.96, "learning_rate": 7.98101265822785e-06, "loss": 0.0003, "step": 8475 }, { "epoch": 85.21, "learning_rate": 7.974683544303799e-06, "loss": 0.0002, "step": 8500 }, { "epoch": 85.46, "learning_rate": 7.968354430379748e-06, "loss": 0.0002, "step": 8525 }, { "epoch": 85.71, "learning_rate": 7.962025316455698e-06, "loss": 0.0003, "step": 8550 }, { "epoch": 85.96, "learning_rate": 7.955696202531646e-06, "loss": 0.0002, "step": 8575 }, { "epoch": 86.22, "learning_rate": 7.949367088607595e-06, "loss": 0.0001, "step": 8600 }, { "epoch": 86.47, "learning_rate": 7.943037974683544e-06, "loss": 0.0001, "step": 8625 }, { "epoch": 86.72, "learning_rate": 7.936708860759494e-06, "loss": 0.0001, "step": 8650 }, { "epoch": 86.97, "learning_rate": 7.930379746835443e-06, "loss": 0.0001, "step": 8675 }, { "epoch": 87.22, "learning_rate": 7.924050632911393e-06, "loss": 0.0001, "step": 8700 }, { "epoch": 87.47, "learning_rate": 7.917721518987342e-06, "loss": 0.0001, "step": 8725 }, { "epoch": 87.72, "learning_rate": 7.911392405063292e-06, "loss": 0.0001, "step": 8750 }, { "epoch": 87.97, "learning_rate": 7.905063291139241e-06, "loss": 0.0001, "step": 8775 }, { "epoch": 88.22, "learning_rate": 7.89873417721519e-06, "loss": 0.0, "step": 8800 }, { "epoch": 88.47, "learning_rate": 7.89240506329114e-06, "loss": 0.0001, "step": 8825 }, { "epoch": 88.72, "learning_rate": 7.88607594936709e-06, "loss": 0.0, "step": 8850 }, { "epoch": 88.97, "learning_rate": 7.879746835443039e-06, "loss": 0.0, "step": 8875 }, { "epoch": 89.22, "learning_rate": 7.873417721518988e-06, "loss": 0.0, "step": 8900 }, { "epoch": 89.47, "learning_rate": 7.867088607594938e-06, "loss": 0.0, "step": 8925 }, { "epoch": 89.72, "learning_rate": 7.860759493670887e-06, "loss": 0.0, "step": 8950 }, { "epoch": 89.97, "learning_rate": 7.854430379746836e-06, "loss": 0.0001, "step": 8975 }, { "epoch": 90.23, "learning_rate": 7.848101265822786e-06, "loss": 0.0001, "step": 9000 }, { "epoch": 90.23, "eval_loss": 0.3110061585903168, "eval_runtime": 1406.8374, "eval_samples_per_second": 9.057, "eval_steps_per_second": 0.284, "eval_wer": 9.300257933406284, "step": 9000 }, { "epoch": 90.48, "learning_rate": 7.841772151898734e-06, "loss": 0.0, "step": 9025 }, { "epoch": 90.73, "learning_rate": 7.835443037974683e-06, "loss": 0.0, "step": 9050 }, { "epoch": 90.98, "learning_rate": 7.829113924050632e-06, "loss": 0.0, "step": 9075 }, { "epoch": 91.23, "learning_rate": 7.822784810126582e-06, "loss": 0.0, "step": 9100 }, { "epoch": 91.48, "learning_rate": 7.816455696202531e-06, "loss": 0.0, "step": 9125 }, { "epoch": 91.73, "learning_rate": 7.810126582278482e-06, "loss": 0.0, "step": 9150 }, { "epoch": 91.98, "learning_rate": 7.803797468354432e-06, "loss": 0.0, "step": 9175 }, { "epoch": 92.23, "learning_rate": 7.797468354430381e-06, "loss": 0.0, "step": 9200 }, { "epoch": 92.48, "learning_rate": 7.79113924050633e-06, "loss": 0.0, "step": 9225 }, { "epoch": 92.73, "learning_rate": 7.78481012658228e-06, "loss": 0.0, "step": 9250 }, { "epoch": 92.98, "learning_rate": 7.77848101265823e-06, "loss": 0.0, "step": 9275 }, { "epoch": 93.23, "learning_rate": 7.772151898734179e-06, "loss": 0.0, "step": 9300 }, { "epoch": 93.48, "learning_rate": 7.765822784810127e-06, "loss": 0.0, "step": 9325 }, { "epoch": 93.73, "learning_rate": 7.759493670886076e-06, "loss": 0.0, "step": 9350 }, { "epoch": 93.98, "learning_rate": 7.753164556962026e-06, "loss": 0.0, "step": 9375 }, { "epoch": 94.24, "learning_rate": 7.746835443037975e-06, "loss": 0.0, "step": 9400 }, { "epoch": 94.49, "learning_rate": 7.740506329113924e-06, "loss": 0.0, "step": 9425 }, { "epoch": 94.74, "learning_rate": 7.734177215189874e-06, "loss": 0.0, "step": 9450 }, { "epoch": 94.99, "learning_rate": 7.727848101265823e-06, "loss": 0.0, "step": 9475 }, { "epoch": 95.24, "learning_rate": 7.721518987341773e-06, "loss": 0.0, "step": 9500 }, { "epoch": 95.49, "learning_rate": 7.715189873417722e-06, "loss": 0.0, "step": 9525 }, { "epoch": 95.74, "learning_rate": 7.708860759493672e-06, "loss": 0.0, "step": 9550 }, { "epoch": 95.99, "learning_rate": 7.702531645569621e-06, "loss": 0.0, "step": 9575 }, { "epoch": 96.24, "learning_rate": 7.69620253164557e-06, "loss": 0.0, "step": 9600 }, { "epoch": 96.49, "learning_rate": 7.68987341772152e-06, "loss": 0.0, "step": 9625 }, { "epoch": 96.74, "learning_rate": 7.68354430379747e-06, "loss": 0.0, "step": 9650 }, { "epoch": 96.99, "learning_rate": 7.677215189873419e-06, "loss": 0.0, "step": 9675 }, { "epoch": 97.24, "learning_rate": 7.670886075949368e-06, "loss": 0.0, "step": 9700 }, { "epoch": 97.49, "learning_rate": 7.664556962025318e-06, "loss": 0.0, "step": 9725 }, { "epoch": 97.74, "learning_rate": 7.658227848101265e-06, "loss": 0.0, "step": 9750 }, { "epoch": 97.99, "learning_rate": 7.651898734177215e-06, "loss": 0.0, "step": 9775 }, { "epoch": 98.25, "learning_rate": 7.645569620253164e-06, "loss": 0.0, "step": 9800 }, { "epoch": 98.5, "learning_rate": 7.639240506329114e-06, "loss": 0.0, "step": 9825 }, { "epoch": 98.75, "learning_rate": 7.632911392405063e-06, "loss": 0.0, "step": 9850 }, { "epoch": 99.0, "learning_rate": 7.6265822784810125e-06, "loss": 0.0, "step": 9875 }, { "epoch": 99.25, "learning_rate": 7.620253164556962e-06, "loss": 0.0, "step": 9900 }, { "epoch": 99.5, "learning_rate": 7.613924050632911e-06, "loss": 0.0, "step": 9925 }, { "epoch": 99.75, "learning_rate": 7.607594936708861e-06, "loss": 0.0, "step": 9950 }, { "epoch": 100.0, "learning_rate": 7.601265822784811e-06, "loss": 0.0, "step": 9975 }, { "epoch": 100.25, "learning_rate": 7.5949367088607605e-06, "loss": 0.0, "step": 10000 }, { "epoch": 100.25, "eval_loss": 0.32776471972465515, "eval_runtime": 1406.5972, "eval_samples_per_second": 9.059, "eval_steps_per_second": 0.284, "eval_wer": 9.331522588713458, "step": 10000 }, { "epoch": 100.5, "learning_rate": 7.58860759493671e-06, "loss": 0.0, "step": 10025 }, { "epoch": 100.75, "learning_rate": 7.582278481012659e-06, "loss": 0.0, "step": 10050 }, { "epoch": 101.0, "learning_rate": 7.575949367088609e-06, "loss": 0.0, "step": 10075 }, { "epoch": 101.25, "learning_rate": 7.569620253164558e-06, "loss": 0.0, "step": 10100 }, { "epoch": 101.5, "learning_rate": 7.563291139240508e-06, "loss": 0.0, "step": 10125 }, { "epoch": 101.75, "learning_rate": 7.556962025316457e-06, "loss": 0.0, "step": 10150 }, { "epoch": 102.01, "learning_rate": 7.550632911392406e-06, "loss": 0.0, "step": 10175 }, { "epoch": 102.26, "learning_rate": 7.544303797468355e-06, "loss": 0.0, "step": 10200 }, { "epoch": 102.51, "learning_rate": 7.5379746835443045e-06, "loss": 0.0, "step": 10225 }, { "epoch": 102.76, "learning_rate": 7.531645569620254e-06, "loss": 0.0, "step": 10250 }, { "epoch": 103.01, "learning_rate": 7.525316455696203e-06, "loss": 0.0, "step": 10275 }, { "epoch": 103.26, "learning_rate": 7.518987341772153e-06, "loss": 0.0, "step": 10300 }, { "epoch": 103.51, "learning_rate": 7.512658227848102e-06, "loss": 0.0, "step": 10325 }, { "epoch": 103.76, "learning_rate": 7.506329113924052e-06, "loss": 0.0, "step": 10350 }, { "epoch": 104.01, "learning_rate": 7.500000000000001e-06, "loss": 0.0, "step": 10375 }, { "epoch": 104.26, "learning_rate": 7.49367088607595e-06, "loss": 0.0, "step": 10400 }, { "epoch": 104.51, "learning_rate": 7.487341772151899e-06, "loss": 0.0, "step": 10425 }, { "epoch": 104.76, "learning_rate": 7.4810126582278485e-06, "loss": 0.0, "step": 10450 }, { "epoch": 105.01, "learning_rate": 7.474683544303798e-06, "loss": 0.0, "step": 10475 }, { "epoch": 105.26, "learning_rate": 7.468354430379747e-06, "loss": 0.0, "step": 10500 }, { "epoch": 105.51, "learning_rate": 7.462025316455697e-06, "loss": 0.0, "step": 10525 }, { "epoch": 105.76, "learning_rate": 7.455696202531646e-06, "loss": 0.0, "step": 10550 }, { "epoch": 106.02, "learning_rate": 7.449367088607596e-06, "loss": 0.0, "step": 10575 }, { "epoch": 106.27, "learning_rate": 7.443037974683544e-06, "loss": 0.0, "step": 10600 }, { "epoch": 106.52, "learning_rate": 7.436708860759494e-06, "loss": 0.0, "step": 10625 }, { "epoch": 106.77, "learning_rate": 7.430379746835443e-06, "loss": 0.0, "step": 10650 }, { "epoch": 107.02, "learning_rate": 7.4240506329113925e-06, "loss": 0.0, "step": 10675 }, { "epoch": 107.27, "learning_rate": 7.417721518987342e-06, "loss": 0.0, "step": 10700 }, { "epoch": 107.52, "learning_rate": 7.411392405063291e-06, "loss": 0.0, "step": 10725 }, { "epoch": 107.77, "learning_rate": 7.405063291139241e-06, "loss": 0.0, "step": 10750 }, { "epoch": 108.02, "learning_rate": 7.39873417721519e-06, "loss": 0.0, "step": 10775 }, { "epoch": 108.27, "learning_rate": 7.3924050632911405e-06, "loss": 0.0, "step": 10800 }, { "epoch": 108.52, "learning_rate": 7.38607594936709e-06, "loss": 0.0, "step": 10825 }, { "epoch": 108.77, "learning_rate": 7.379746835443039e-06, "loss": 0.0, "step": 10850 }, { "epoch": 109.02, "learning_rate": 7.373417721518989e-06, "loss": 0.0, "step": 10875 }, { "epoch": 109.27, "learning_rate": 7.367088607594937e-06, "loss": 0.0, "step": 10900 }, { "epoch": 109.52, "learning_rate": 7.360759493670887e-06, "loss": 0.0, "step": 10925 }, { "epoch": 109.77, "learning_rate": 7.354430379746836e-06, "loss": 0.0, "step": 10950 }, { "epoch": 110.03, "learning_rate": 7.348101265822786e-06, "loss": 0.0, "step": 10975 }, { "epoch": 110.28, "learning_rate": 7.341772151898735e-06, "loss": 0.0, "step": 11000 }, { "epoch": 110.28, "eval_loss": 0.3393121063709259, "eval_runtime": 1406.9189, "eval_samples_per_second": 9.057, "eval_steps_per_second": 0.284, "eval_wer": 9.308074097233078, "step": 11000 }, { "epoch": 110.53, "learning_rate": 7.3354430379746845e-06, "loss": 0.0, "step": 11025 }, { "epoch": 110.78, "learning_rate": 7.329113924050634e-06, "loss": 0.0, "step": 11050 }, { "epoch": 111.03, "learning_rate": 7.322784810126583e-06, "loss": 0.0, "step": 11075 }, { "epoch": 111.28, "learning_rate": 7.316455696202533e-06, "loss": 0.0, "step": 11100 }, { "epoch": 111.53, "learning_rate": 7.310126582278481e-06, "loss": 0.0, "step": 11125 }, { "epoch": 111.78, "learning_rate": 7.303797468354431e-06, "loss": 0.0, "step": 11150 }, { "epoch": 112.03, "learning_rate": 7.29746835443038e-06, "loss": 0.0, "step": 11175 }, { "epoch": 112.28, "learning_rate": 7.29113924050633e-06, "loss": 0.0, "step": 11200 }, { "epoch": 112.53, "learning_rate": 7.284810126582279e-06, "loss": 0.0, "step": 11225 }, { "epoch": 112.78, "learning_rate": 7.2784810126582285e-06, "loss": 0.0, "step": 11250 }, { "epoch": 113.03, "learning_rate": 7.272151898734178e-06, "loss": 0.0, "step": 11275 }, { "epoch": 113.28, "learning_rate": 7.265822784810127e-06, "loss": 0.0, "step": 11300 }, { "epoch": 113.53, "learning_rate": 7.259493670886077e-06, "loss": 0.0, "step": 11325 }, { "epoch": 113.78, "learning_rate": 7.253164556962025e-06, "loss": 0.0, "step": 11350 }, { "epoch": 114.04, "learning_rate": 7.246835443037975e-06, "loss": 0.0, "step": 11375 }, { "epoch": 114.29, "learning_rate": 7.240506329113924e-06, "loss": 0.0, "step": 11400 }, { "epoch": 114.54, "learning_rate": 7.234177215189874e-06, "loss": 0.0, "step": 11425 }, { "epoch": 114.79, "learning_rate": 7.227848101265823e-06, "loss": 0.0, "step": 11450 }, { "epoch": 115.04, "learning_rate": 7.2215189873417725e-06, "loss": 0.0, "step": 11475 }, { "epoch": 115.29, "learning_rate": 7.215189873417722e-06, "loss": 0.0, "step": 11500 }, { "epoch": 115.54, "learning_rate": 7.208860759493671e-06, "loss": 0.0, "step": 11525 }, { "epoch": 115.79, "learning_rate": 7.202531645569621e-06, "loss": 0.0, "step": 11550 }, { "epoch": 116.04, "learning_rate": 7.196202531645569e-06, "loss": 0.0, "step": 11575 }, { "epoch": 116.29, "learning_rate": 7.189873417721519e-06, "loss": 0.0, "step": 11600 }, { "epoch": 116.54, "learning_rate": 7.18354430379747e-06, "loss": 0.0, "step": 11625 }, { "epoch": 116.79, "learning_rate": 7.1772151898734185e-06, "loss": 0.0, "step": 11650 }, { "epoch": 117.04, "learning_rate": 7.170886075949368e-06, "loss": 0.0, "step": 11675 }, { "epoch": 117.29, "learning_rate": 7.164556962025317e-06, "loss": 0.0, "step": 11700 }, { "epoch": 117.54, "learning_rate": 7.158227848101267e-06, "loss": 0.0, "step": 11725 }, { "epoch": 117.79, "learning_rate": 7.151898734177216e-06, "loss": 0.0, "step": 11750 }, { "epoch": 118.05, "learning_rate": 7.145569620253166e-06, "loss": 0.0, "step": 11775 }, { "epoch": 118.3, "learning_rate": 7.139240506329115e-06, "loss": 0.0, "step": 11800 }, { "epoch": 118.55, "learning_rate": 7.1329113924050645e-06, "loss": 0.0, "step": 11825 }, { "epoch": 118.8, "learning_rate": 7.126582278481014e-06, "loss": 0.0, "step": 11850 }, { "epoch": 119.05, "learning_rate": 7.1202531645569625e-06, "loss": 0.0, "step": 11875 }, { "epoch": 119.3, "learning_rate": 7.113924050632912e-06, "loss": 0.0, "step": 11900 }, { "epoch": 119.55, "learning_rate": 7.107594936708861e-06, "loss": 0.0, "step": 11925 }, { "epoch": 119.8, "learning_rate": 7.101265822784811e-06, "loss": 0.0, "step": 11950 }, { "epoch": 120.05, "learning_rate": 7.09493670886076e-06, "loss": 0.0, "step": 11975 }, { "epoch": 120.3, "learning_rate": 7.08860759493671e-06, "loss": 0.0, "step": 12000 }, { "epoch": 120.3, "eval_loss": 0.3508254885673523, "eval_runtime": 1408.2039, "eval_samples_per_second": 9.048, "eval_steps_per_second": 0.283, "eval_wer": 9.299280912927935, "step": 12000 }, { "epoch": 120.55, "learning_rate": 7.082531645569621e-06, "loss": 0.0, "step": 12025 }, { "epoch": 120.8, "learning_rate": 7.0762025316455705e-06, "loss": 0.0, "step": 12050 }, { "epoch": 121.05, "learning_rate": 7.06987341772152e-06, "loss": 0.0, "step": 12075 }, { "epoch": 121.3, "learning_rate": 7.063544303797469e-06, "loss": 0.0, "step": 12100 }, { "epoch": 121.55, "learning_rate": 7.057215189873419e-06, "loss": 0.0, "step": 12125 }, { "epoch": 121.8, "learning_rate": 7.050886075949368e-06, "loss": 0.0, "step": 12150 }, { "epoch": 122.06, "learning_rate": 7.044556962025318e-06, "loss": 0.0, "step": 12175 }, { "epoch": 122.31, "learning_rate": 7.038227848101267e-06, "loss": 0.0, "step": 12200 }, { "epoch": 122.56, "learning_rate": 7.031898734177216e-06, "loss": 0.0, "step": 12225 }, { "epoch": 122.81, "learning_rate": 7.025569620253165e-06, "loss": 0.0, "step": 12250 }, { "epoch": 123.06, "learning_rate": 7.0192405063291145e-06, "loss": 0.0, "step": 12275 }, { "epoch": 123.31, "learning_rate": 7.012911392405064e-06, "loss": 0.0, "step": 12300 }, { "epoch": 123.56, "learning_rate": 7.006582278481013e-06, "loss": 0.0, "step": 12325 }, { "epoch": 123.81, "learning_rate": 7.000253164556963e-06, "loss": 0.0, "step": 12350 }, { "epoch": 124.06, "learning_rate": 6.993924050632912e-06, "loss": 0.0, "step": 12375 }, { "epoch": 124.31, "learning_rate": 6.987594936708862e-06, "loss": 0.0, "step": 12400 }, { "epoch": 124.56, "learning_rate": 6.981265822784811e-06, "loss": 0.0, "step": 12425 }, { "epoch": 124.81, "learning_rate": 6.97493670886076e-06, "loss": 0.0, "step": 12450 }, { "epoch": 125.06, "learning_rate": 6.968607594936709e-06, "loss": 0.0, "step": 12475 }, { "epoch": 125.31, "learning_rate": 6.9622784810126585e-06, "loss": 0.0, "step": 12500 }, { "epoch": 125.56, "learning_rate": 6.955949367088608e-06, "loss": 0.0, "step": 12525 }, { "epoch": 125.81, "learning_rate": 6.949620253164557e-06, "loss": 0.0, "step": 12550 }, { "epoch": 126.07, "learning_rate": 6.943291139240507e-06, "loss": 0.0, "step": 12575 }, { "epoch": 126.32, "learning_rate": 6.936962025316456e-06, "loss": 0.0, "step": 12600 }, { "epoch": 126.57, "learning_rate": 6.930632911392406e-06, "loss": 0.0, "step": 12625 }, { "epoch": 126.82, "learning_rate": 6.924303797468355e-06, "loss": 0.0, "step": 12650 }, { "epoch": 127.07, "learning_rate": 6.917974683544304e-06, "loss": 0.0, "step": 12675 }, { "epoch": 127.32, "learning_rate": 6.911645569620253e-06, "loss": 0.0, "step": 12700 }, { "epoch": 127.57, "learning_rate": 6.9053164556962025e-06, "loss": 0.0, "step": 12725 }, { "epoch": 127.82, "learning_rate": 6.898987341772152e-06, "loss": 0.0, "step": 12750 }, { "epoch": 128.07, "learning_rate": 6.892658227848101e-06, "loss": 0.0, "step": 12775 }, { "epoch": 128.32, "learning_rate": 6.886329113924051e-06, "loss": 0.0, "step": 12800 }, { "epoch": 128.57, "learning_rate": 6.88e-06, "loss": 0.0, "step": 12825 }, { "epoch": 128.82, "learning_rate": 6.8736708860759505e-06, "loss": 0.0, "step": 12850 }, { "epoch": 129.07, "learning_rate": 6.8673417721519e-06, "loss": 0.0, "step": 12875 }, { "epoch": 129.32, "learning_rate": 6.861012658227849e-06, "loss": 0.0, "step": 12900 }, { "epoch": 129.57, "learning_rate": 6.854683544303799e-06, "loss": 0.0, "step": 12925 }, { "epoch": 129.82, "learning_rate": 6.848354430379748e-06, "loss": 0.0, "step": 12950 }, { "epoch": 130.08, "learning_rate": 6.842025316455697e-06, "loss": 0.0, "step": 12975 }, { "epoch": 130.33, "learning_rate": 6.835696202531646e-06, "loss": 0.0, "step": 13000 }, { "epoch": 130.33, "eval_loss": 0.3617466986179352, "eval_runtime": 1406.254, "eval_samples_per_second": 9.061, "eval_steps_per_second": 0.284, "eval_wer": 9.321752383929967, "step": 13000 }, { "epoch": 130.58, "learning_rate": 6.829367088607596e-06, "loss": 0.0, "step": 13025 }, { "epoch": 130.83, "learning_rate": 6.823037974683545e-06, "loss": 0.0, "step": 13050 }, { "epoch": 131.08, "learning_rate": 6.8167088607594945e-06, "loss": 0.0, "step": 13075 }, { "epoch": 131.33, "learning_rate": 6.810379746835444e-06, "loss": 0.0, "step": 13100 }, { "epoch": 131.58, "learning_rate": 6.804050632911393e-06, "loss": 0.0, "step": 13125 }, { "epoch": 131.83, "learning_rate": 6.797721518987343e-06, "loss": 0.0, "step": 13150 }, { "epoch": 132.08, "learning_rate": 6.791392405063292e-06, "loss": 0.0, "step": 13175 }, { "epoch": 132.33, "learning_rate": 6.785063291139241e-06, "loss": 0.0, "step": 13200 }, { "epoch": 132.58, "learning_rate": 6.77873417721519e-06, "loss": 0.0, "step": 13225 }, { "epoch": 132.83, "learning_rate": 6.77240506329114e-06, "loss": 0.0, "step": 13250 }, { "epoch": 133.08, "learning_rate": 6.766075949367089e-06, "loss": 0.0, "step": 13275 }, { "epoch": 133.33, "learning_rate": 6.7597468354430385e-06, "loss": 0.0, "step": 13300 }, { "epoch": 133.58, "learning_rate": 6.753417721518988e-06, "loss": 0.0, "step": 13325 }, { "epoch": 133.83, "learning_rate": 6.747088607594937e-06, "loss": 0.0, "step": 13350 }, { "epoch": 134.09, "learning_rate": 6.740759493670887e-06, "loss": 0.0, "step": 13375 }, { "epoch": 134.34, "learning_rate": 6.734430379746836e-06, "loss": 0.0, "step": 13400 }, { "epoch": 134.59, "learning_rate": 6.728101265822785e-06, "loss": 0.0, "step": 13425 }, { "epoch": 134.84, "learning_rate": 6.721772151898734e-06, "loss": 0.0, "step": 13450 }, { "epoch": 135.09, "learning_rate": 6.715443037974684e-06, "loss": 0.0, "step": 13475 }, { "epoch": 135.34, "learning_rate": 6.709113924050633e-06, "loss": 0.0, "step": 13500 }, { "epoch": 135.59, "learning_rate": 6.7027848101265825e-06, "loss": 0.0, "step": 13525 }, { "epoch": 135.84, "learning_rate": 6.696455696202532e-06, "loss": 0.0, "step": 13550 }, { "epoch": 136.09, "learning_rate": 6.690126582278481e-06, "loss": 0.0, "step": 13575 }, { "epoch": 136.34, "learning_rate": 6.683797468354431e-06, "loss": 0.0, "step": 13600 }, { "epoch": 136.59, "learning_rate": 6.67746835443038e-06, "loss": 0.0, "step": 13625 }, { "epoch": 136.84, "learning_rate": 6.671139240506329e-06, "loss": 0.0, "step": 13650 }, { "epoch": 137.09, "learning_rate": 6.66481012658228e-06, "loss": 0.0, "step": 13675 }, { "epoch": 137.34, "learning_rate": 6.6584810126582285e-06, "loss": 0.0, "step": 13700 }, { "epoch": 137.59, "learning_rate": 6.652151898734178e-06, "loss": 0.0, "step": 13725 }, { "epoch": 137.84, "learning_rate": 6.645822784810127e-06, "loss": 0.0, "step": 13750 }, { "epoch": 138.1, "learning_rate": 6.639493670886077e-06, "loss": 0.0, "step": 13775 }, { "epoch": 138.35, "learning_rate": 6.633164556962026e-06, "loss": 0.0, "step": 13800 }, { "epoch": 138.6, "learning_rate": 6.626835443037976e-06, "loss": 0.0, "step": 13825 }, { "epoch": 138.85, "learning_rate": 6.620506329113925e-06, "loss": 0.0, "step": 13850 }, { "epoch": 139.1, "learning_rate": 6.6141772151898745e-06, "loss": 0.0, "step": 13875 }, { "epoch": 139.35, "learning_rate": 6.607848101265824e-06, "loss": 0.0, "step": 13900 }, { "epoch": 139.6, "learning_rate": 6.6015189873417725e-06, "loss": 0.0, "step": 13925 }, { "epoch": 139.85, "learning_rate": 6.595189873417722e-06, "loss": 0.0, "step": 13950 }, { "epoch": 140.1, "learning_rate": 6.588860759493671e-06, "loss": 0.0, "step": 13975 }, { "epoch": 140.35, "learning_rate": 6.582531645569621e-06, "loss": 0.0, "step": 14000 }, { "epoch": 140.35, "eval_loss": 0.37321582436561584, "eval_runtime": 1407.2213, "eval_samples_per_second": 9.055, "eval_steps_per_second": 0.284, "eval_wer": 9.335430670626858, "step": 14000 }, { "epoch": 140.6, "learning_rate": 6.57620253164557e-06, "loss": 0.0, "step": 14025 }, { "epoch": 140.85, "learning_rate": 6.5701265822784825e-06, "loss": 0.0, "step": 14050 }, { "epoch": 141.1, "learning_rate": 6.563797468354431e-06, "loss": 0.0, "step": 14075 }, { "epoch": 141.35, "learning_rate": 6.5574683544303805e-06, "loss": 0.0, "step": 14100 }, { "epoch": 141.6, "learning_rate": 6.55113924050633e-06, "loss": 0.0, "step": 14125 }, { "epoch": 141.85, "learning_rate": 6.544810126582279e-06, "loss": 0.0, "step": 14150 }, { "epoch": 142.11, "learning_rate": 6.538481012658229e-06, "loss": 0.0, "step": 14175 }, { "epoch": 142.36, "learning_rate": 6.532151898734178e-06, "loss": 0.0, "step": 14200 }, { "epoch": 142.61, "learning_rate": 6.525822784810128e-06, "loss": 0.0, "step": 14225 }, { "epoch": 142.86, "learning_rate": 6.519493670886077e-06, "loss": 0.0, "step": 14250 }, { "epoch": 143.11, "learning_rate": 6.5131645569620265e-06, "loss": 0.0, "step": 14275 }, { "epoch": 143.36, "learning_rate": 6.506835443037975e-06, "loss": 0.0, "step": 14300 }, { "epoch": 143.61, "learning_rate": 6.5005063291139245e-06, "loss": 0.0, "step": 14325 }, { "epoch": 143.86, "learning_rate": 6.494177215189874e-06, "loss": 0.0, "step": 14350 }, { "epoch": 144.11, "learning_rate": 6.487848101265823e-06, "loss": 0.0, "step": 14375 }, { "epoch": 144.36, "learning_rate": 6.481518987341773e-06, "loss": 0.0, "step": 14400 }, { "epoch": 144.61, "learning_rate": 6.475189873417722e-06, "loss": 0.0, "step": 14425 }, { "epoch": 144.86, "learning_rate": 6.468860759493672e-06, "loss": 0.0, "step": 14450 }, { "epoch": 145.11, "learning_rate": 6.462531645569621e-06, "loss": 0.0, "step": 14475 }, { "epoch": 145.36, "learning_rate": 6.4562025316455705e-06, "loss": 0.0, "step": 14500 }, { "epoch": 145.61, "learning_rate": 6.449873417721519e-06, "loss": 0.0, "step": 14525 }, { "epoch": 145.86, "learning_rate": 6.4435443037974685e-06, "loss": 0.0, "step": 14550 }, { "epoch": 146.12, "learning_rate": 6.437215189873418e-06, "loss": 0.0, "step": 14575 }, { "epoch": 146.37, "learning_rate": 6.430886075949367e-06, "loss": 0.0, "step": 14600 }, { "epoch": 146.62, "learning_rate": 6.424556962025317e-06, "loss": 0.0, "step": 14625 }, { "epoch": 146.87, "learning_rate": 6.418227848101266e-06, "loss": 0.0, "step": 14650 }, { "epoch": 147.12, "learning_rate": 6.411898734177216e-06, "loss": 0.0, "step": 14675 }, { "epoch": 147.37, "learning_rate": 6.405569620253165e-06, "loss": 0.0, "step": 14700 }, { "epoch": 147.62, "learning_rate": 6.3992405063291145e-06, "loss": 0.0, "step": 14725 }, { "epoch": 147.87, "learning_rate": 6.392911392405063e-06, "loss": 0.0, "step": 14750 }, { "epoch": 148.12, "learning_rate": 6.3865822784810125e-06, "loss": 0.0, "step": 14775 }, { "epoch": 148.37, "learning_rate": 6.380253164556962e-06, "loss": 0.0, "step": 14800 }, { "epoch": 148.62, "learning_rate": 6.373924050632911e-06, "loss": 0.0, "step": 14825 }, { "epoch": 148.87, "learning_rate": 6.367594936708861e-06, "loss": 0.0, "step": 14850 }, { "epoch": 149.12, "learning_rate": 6.361265822784811e-06, "loss": 0.0, "step": 14875 }, { "epoch": 149.37, "learning_rate": 6.3549367088607605e-06, "loss": 0.0, "step": 14900 }, { "epoch": 149.62, "learning_rate": 6.34860759493671e-06, "loss": 0.0, "step": 14925 }, { "epoch": 149.87, "learning_rate": 6.342278481012659e-06, "loss": 0.0, "step": 14950 }, { "epoch": 150.13, "learning_rate": 6.335949367088609e-06, "loss": 0.0, "step": 14975 }, { "epoch": 150.38, "learning_rate": 6.329620253164558e-06, "loss": 0.0, "step": 15000 }, { "epoch": 150.38, "eval_loss": 0.3848804533481598, "eval_runtime": 1407.7638, "eval_samples_per_second": 9.051, "eval_steps_per_second": 0.283, "eval_wer": 9.373534469282477, "step": 15000 }, { "epoch": 150.63, "learning_rate": 6.323291139240507e-06, "loss": 0.0, "step": 15025 }, { "epoch": 150.88, "learning_rate": 6.316962025316456e-06, "loss": 0.0, "step": 15050 }, { "epoch": 151.13, "learning_rate": 6.310632911392406e-06, "loss": 0.0, "step": 15075 }, { "epoch": 151.38, "learning_rate": 6.304303797468355e-06, "loss": 0.0, "step": 15100 }, { "epoch": 151.63, "learning_rate": 6.2979746835443045e-06, "loss": 0.0, "step": 15125 }, { "epoch": 151.88, "learning_rate": 6.291645569620254e-06, "loss": 0.0, "step": 15150 }, { "epoch": 152.13, "learning_rate": 6.285316455696203e-06, "loss": 0.0, "step": 15175 }, { "epoch": 152.38, "learning_rate": 6.278987341772153e-06, "loss": 0.0, "step": 15200 }, { "epoch": 152.63, "learning_rate": 6.272658227848102e-06, "loss": 0.0, "step": 15225 }, { "epoch": 152.88, "learning_rate": 6.266329113924051e-06, "loss": 0.0, "step": 15250 }, { "epoch": 153.13, "learning_rate": 6.26e-06, "loss": 0.0, "step": 15275 }, { "epoch": 153.38, "learning_rate": 6.25367088607595e-06, "loss": 0.0008, "step": 15300 }, { "epoch": 153.63, "learning_rate": 6.247594936708862e-06, "loss": 0.0118, "step": 15325 }, { "epoch": 153.88, "learning_rate": 6.241265822784811e-06, "loss": 0.0138, "step": 15350 }, { "epoch": 154.14, "learning_rate": 6.234936708860761e-06, "loss": 0.01, "step": 15375 }, { "epoch": 154.39, "learning_rate": 6.228607594936709e-06, "loss": 0.0084, "step": 15400 }, { "epoch": 154.64, "learning_rate": 6.222278481012659e-06, "loss": 0.0057, "step": 15425 }, { "epoch": 154.89, "learning_rate": 6.215949367088608e-06, "loss": 0.0057, "step": 15450 }, { "epoch": 155.14, "learning_rate": 6.209620253164558e-06, "loss": 0.0037, "step": 15475 }, { "epoch": 155.39, "learning_rate": 6.203291139240507e-06, "loss": 0.0027, "step": 15500 }, { "epoch": 155.64, "learning_rate": 6.1969620253164565e-06, "loss": 0.0022, "step": 15525 }, { "epoch": 155.89, "learning_rate": 6.190632911392406e-06, "loss": 0.0015, "step": 15550 }, { "epoch": 156.14, "learning_rate": 6.184303797468355e-06, "loss": 0.0008, "step": 15575 }, { "epoch": 156.39, "learning_rate": 6.177974683544305e-06, "loss": 0.0005, "step": 15600 }, { "epoch": 156.64, "learning_rate": 6.171645569620253e-06, "loss": 0.0004, "step": 15625 }, { "epoch": 156.89, "learning_rate": 6.165316455696203e-06, "loss": 0.0003, "step": 15650 }, { "epoch": 157.14, "learning_rate": 6.158987341772152e-06, "loss": 0.0003, "step": 15675 }, { "epoch": 157.39, "learning_rate": 6.152658227848102e-06, "loss": 0.0001, "step": 15700 }, { "epoch": 157.64, "learning_rate": 6.146329113924051e-06, "loss": 0.0004, "step": 15725 }, { "epoch": 157.89, "learning_rate": 6.1400000000000005e-06, "loss": 0.0001, "step": 15750 }, { "epoch": 158.15, "learning_rate": 6.13367088607595e-06, "loss": 0.0001, "step": 15775 }, { "epoch": 158.4, "learning_rate": 6.127341772151899e-06, "loss": 0.0001, "step": 15800 }, { "epoch": 158.65, "learning_rate": 6.121265822784811e-06, "loss": 0.0002, "step": 15825 }, { "epoch": 158.9, "learning_rate": 6.11493670886076e-06, "loss": 0.0001, "step": 15850 }, { "epoch": 159.15, "learning_rate": 6.10860759493671e-06, "loss": 0.0001, "step": 15875 }, { "epoch": 159.4, "learning_rate": 6.102278481012659e-06, "loss": 0.0001, "step": 15900 }, { "epoch": 159.65, "learning_rate": 6.0959493670886085e-06, "loss": 0.0, "step": 15925 }, { "epoch": 159.9, "learning_rate": 6.089620253164558e-06, "loss": 0.0001, "step": 15950 }, { "epoch": 160.15, "learning_rate": 6.083291139240507e-06, "loss": 0.0, "step": 15975 }, { "epoch": 160.4, "learning_rate": 6.076962025316456e-06, "loss": 0.0, "step": 16000 }, { "epoch": 160.4, "eval_loss": 0.30726706981658936, "eval_runtime": 1407.2848, "eval_samples_per_second": 9.054, "eval_steps_per_second": 0.284, "eval_wer": 9.333476629670159, "step": 16000 }, { "epoch": 160.65, "learning_rate": 6.070632911392405e-06, "loss": 0.0, "step": 16025 }, { "epoch": 160.9, "learning_rate": 6.064303797468355e-06, "loss": 0.0, "step": 16050 }, { "epoch": 161.15, "learning_rate": 6.057974683544304e-06, "loss": 0.0, "step": 16075 }, { "epoch": 161.4, "learning_rate": 6.051645569620254e-06, "loss": 0.0, "step": 16100 }, { "epoch": 161.65, "learning_rate": 6.045316455696203e-06, "loss": 0.0, "step": 16125 }, { "epoch": 161.9, "learning_rate": 6.0389873417721525e-06, "loss": 0.0, "step": 16150 }, { "epoch": 162.16, "learning_rate": 6.032658227848102e-06, "loss": 0.0, "step": 16175 }, { "epoch": 162.41, "learning_rate": 6.0263291139240505e-06, "loss": 0.0, "step": 16200 }, { "epoch": 162.66, "learning_rate": 6.02e-06, "loss": 0.0, "step": 16225 }, { "epoch": 162.91, "learning_rate": 6.013670886075949e-06, "loss": 0.0, "step": 16250 }, { "epoch": 163.16, "learning_rate": 6.007341772151899e-06, "loss": 0.0, "step": 16275 }, { "epoch": 163.41, "learning_rate": 6.001012658227848e-06, "loss": 0.0, "step": 16300 }, { "epoch": 163.66, "learning_rate": 5.994683544303798e-06, "loss": 0.0, "step": 16325 }, { "epoch": 163.91, "learning_rate": 5.988354430379747e-06, "loss": 0.0, "step": 16350 }, { "epoch": 164.16, "learning_rate": 5.9820253164556965e-06, "loss": 0.0, "step": 16375 }, { "epoch": 164.41, "learning_rate": 5.975696202531646e-06, "loss": 0.0, "step": 16400 }, { "epoch": 164.66, "learning_rate": 5.9693670886075945e-06, "loss": 0.0, "step": 16425 }, { "epoch": 164.91, "learning_rate": 5.963037974683544e-06, "loss": 0.0, "step": 16450 }, { "epoch": 165.16, "learning_rate": 5.956708860759493e-06, "loss": 0.0, "step": 16475 }, { "epoch": 165.41, "learning_rate": 5.950379746835444e-06, "loss": 0.0, "step": 16500 }, { "epoch": 165.66, "learning_rate": 5.944050632911393e-06, "loss": 0.0, "step": 16525 }, { "epoch": 165.91, "learning_rate": 5.9377215189873425e-06, "loss": 0.0, "step": 16550 }, { "epoch": 166.17, "learning_rate": 5.931392405063292e-06, "loss": 0.0, "step": 16575 }, { "epoch": 166.42, "learning_rate": 5.925063291139241e-06, "loss": 0.0, "step": 16600 }, { "epoch": 166.67, "learning_rate": 5.918734177215191e-06, "loss": 0.0, "step": 16625 }, { "epoch": 166.92, "learning_rate": 5.91240506329114e-06, "loss": 0.0, "step": 16650 }, { "epoch": 167.17, "learning_rate": 5.90607594936709e-06, "loss": 0.0, "step": 16675 }, { "epoch": 167.42, "learning_rate": 5.899746835443039e-06, "loss": 0.0, "step": 16700 }, { "epoch": 167.67, "learning_rate": 5.893417721518988e-06, "loss": 0.0, "step": 16725 }, { "epoch": 167.92, "learning_rate": 5.887088607594937e-06, "loss": 0.0, "step": 16750 }, { "epoch": 168.17, "learning_rate": 5.8807594936708865e-06, "loss": 0.0, "step": 16775 }, { "epoch": 168.42, "learning_rate": 5.874430379746836e-06, "loss": 0.0, "step": 16800 }, { "epoch": 168.67, "learning_rate": 5.868101265822785e-06, "loss": 0.0, "step": 16825 }, { "epoch": 168.92, "learning_rate": 5.861772151898735e-06, "loss": 0.0, "step": 16850 }, { "epoch": 169.17, "learning_rate": 5.855443037974684e-06, "loss": 0.0, "step": 16875 }, { "epoch": 169.42, "learning_rate": 5.849113924050634e-06, "loss": 0.0, "step": 16900 }, { "epoch": 169.67, "learning_rate": 5.842784810126583e-06, "loss": 0.0, "step": 16925 }, { "epoch": 169.92, "learning_rate": 5.836455696202532e-06, "loss": 0.0, "step": 16950 }, { "epoch": 170.18, "learning_rate": 5.830126582278481e-06, "loss": 0.0, "step": 16975 }, { "epoch": 170.43, "learning_rate": 5.8237974683544305e-06, "loss": 0.0, "step": 17000 }, { "epoch": 170.43, "eval_loss": 0.33195701241493225, "eval_runtime": 1406.9323, "eval_samples_per_second": 9.057, "eval_steps_per_second": 0.284, "eval_wer": 9.356925121150539, "step": 17000 }, { "epoch": 170.68, "learning_rate": 5.81746835443038e-06, "loss": 0.0, "step": 17025 }, { "epoch": 170.93, "learning_rate": 5.811139240506329e-06, "loss": 0.0, "step": 17050 }, { "epoch": 171.18, "learning_rate": 5.804810126582279e-06, "loss": 0.0, "step": 17075 }, { "epoch": 171.43, "learning_rate": 5.798481012658228e-06, "loss": 0.0, "step": 17100 }, { "epoch": 171.68, "learning_rate": 5.792151898734178e-06, "loss": 0.0, "step": 17125 }, { "epoch": 171.93, "learning_rate": 5.785822784810127e-06, "loss": 0.0, "step": 17150 }, { "epoch": 172.18, "learning_rate": 5.779493670886076e-06, "loss": 0.0, "step": 17175 }, { "epoch": 172.43, "learning_rate": 5.773164556962025e-06, "loss": 0.0, "step": 17200 }, { "epoch": 172.68, "learning_rate": 5.7668354430379745e-06, "loss": 0.0, "step": 17225 }, { "epoch": 172.93, "learning_rate": 5.760506329113924e-06, "loss": 0.0, "step": 17250 }, { "epoch": 173.18, "learning_rate": 5.754177215189873e-06, "loss": 0.0, "step": 17275 }, { "epoch": 173.43, "learning_rate": 5.747848101265823e-06, "loss": 0.0, "step": 17300 }, { "epoch": 173.68, "learning_rate": 5.741518987341773e-06, "loss": 0.0, "step": 17325 }, { "epoch": 173.93, "learning_rate": 5.7351898734177225e-06, "loss": 0.0, "step": 17350 }, { "epoch": 174.19, "learning_rate": 5.728860759493672e-06, "loss": 0.0, "step": 17375 }, { "epoch": 174.44, "learning_rate": 5.722531645569621e-06, "loss": 0.0, "step": 17400 }, { "epoch": 174.69, "learning_rate": 5.716202531645571e-06, "loss": 0.0, "step": 17425 }, { "epoch": 174.94, "learning_rate": 5.70987341772152e-06, "loss": 0.0, "step": 17450 }, { "epoch": 175.19, "learning_rate": 5.703544303797469e-06, "loss": 0.0, "step": 17475 }, { "epoch": 175.44, "learning_rate": 5.697215189873418e-06, "loss": 0.0, "step": 17500 }, { "epoch": 175.69, "learning_rate": 5.690886075949368e-06, "loss": 0.0, "step": 17525 }, { "epoch": 175.94, "learning_rate": 5.684556962025317e-06, "loss": 0.0, "step": 17550 }, { "epoch": 176.19, "learning_rate": 5.6782278481012665e-06, "loss": 0.0, "step": 17575 }, { "epoch": 176.44, "learning_rate": 5.671898734177216e-06, "loss": 0.0, "step": 17600 }, { "epoch": 176.69, "learning_rate": 5.665569620253165e-06, "loss": 0.0, "step": 17625 }, { "epoch": 176.94, "learning_rate": 5.659240506329115e-06, "loss": 0.0, "step": 17650 }, { "epoch": 177.19, "learning_rate": 5.652911392405064e-06, "loss": 0.0, "step": 17675 }, { "epoch": 177.44, "learning_rate": 5.646582278481013e-06, "loss": 0.0, "step": 17700 }, { "epoch": 177.69, "learning_rate": 5.640253164556962e-06, "loss": 0.0, "step": 17725 }, { "epoch": 177.94, "learning_rate": 5.633924050632912e-06, "loss": 0.0, "step": 17750 }, { "epoch": 178.2, "learning_rate": 5.627594936708861e-06, "loss": 0.0, "step": 17775 }, { "epoch": 178.45, "learning_rate": 5.6212658227848105e-06, "loss": 0.0, "step": 17800 }, { "epoch": 178.7, "learning_rate": 5.61493670886076e-06, "loss": 0.0, "step": 17825 }, { "epoch": 178.95, "learning_rate": 5.608607594936709e-06, "loss": 0.0, "step": 17850 }, { "epoch": 179.2, "learning_rate": 5.602278481012659e-06, "loss": 0.0, "step": 17875 }, { "epoch": 179.45, "learning_rate": 5.595949367088608e-06, "loss": 0.0, "step": 17900 }, { "epoch": 179.7, "learning_rate": 5.589620253164557e-06, "loss": 0.0, "step": 17925 }, { "epoch": 179.95, "learning_rate": 5.583291139240506e-06, "loss": 0.0, "step": 17950 }, { "epoch": 180.2, "learning_rate": 5.576962025316456e-06, "loss": 0.0, "step": 17975 }, { "epoch": 180.45, "learning_rate": 5.570632911392405e-06, "loss": 0.0, "step": 18000 }, { "epoch": 180.45, "eval_loss": 0.3452655076980591, "eval_runtime": 1406.6392, "eval_samples_per_second": 9.058, "eval_steps_per_second": 0.284, "eval_wer": 9.302211974362983, "step": 18000 }, { "epoch": 180.7, "learning_rate": 5.5643037974683545e-06, "loss": 0.0, "step": 18025 }, { "epoch": 180.95, "learning_rate": 5.557974683544304e-06, "loss": 0.0, "step": 18050 }, { "epoch": 181.2, "learning_rate": 5.551645569620253e-06, "loss": 0.0, "step": 18075 }, { "epoch": 181.45, "learning_rate": 5.545316455696203e-06, "loss": 0.0, "step": 18100 }, { "epoch": 181.7, "learning_rate": 5.538987341772151e-06, "loss": 0.0, "step": 18125 }, { "epoch": 181.95, "learning_rate": 5.5326582278481025e-06, "loss": 0.0, "step": 18150 }, { "epoch": 182.21, "learning_rate": 5.526329113924052e-06, "loss": 0.0, "step": 18175 }, { "epoch": 182.46, "learning_rate": 5.5200000000000005e-06, "loss": 0.0, "step": 18200 }, { "epoch": 182.71, "learning_rate": 5.51367088607595e-06, "loss": 0.0, "step": 18225 }, { "epoch": 182.96, "learning_rate": 5.507341772151899e-06, "loss": 0.0, "step": 18250 }, { "epoch": 183.21, "learning_rate": 5.501012658227849e-06, "loss": 0.0, "step": 18275 }, { "epoch": 183.46, "learning_rate": 5.494683544303798e-06, "loss": 0.0, "step": 18300 }, { "epoch": 183.71, "learning_rate": 5.488354430379748e-06, "loss": 0.0, "step": 18325 }, { "epoch": 183.96, "learning_rate": 5.482025316455697e-06, "loss": 0.0, "step": 18350 }, { "epoch": 184.21, "learning_rate": 5.4756962025316465e-06, "loss": 0.0, "step": 18375 }, { "epoch": 184.46, "learning_rate": 5.469367088607596e-06, "loss": 0.0, "step": 18400 }, { "epoch": 184.71, "learning_rate": 5.4630379746835445e-06, "loss": 0.0, "step": 18425 }, { "epoch": 184.96, "learning_rate": 5.456708860759494e-06, "loss": 0.0, "step": 18450 }, { "epoch": 185.21, "learning_rate": 5.450379746835443e-06, "loss": 0.0, "step": 18475 }, { "epoch": 185.46, "learning_rate": 5.444050632911393e-06, "loss": 0.0, "step": 18500 }, { "epoch": 185.71, "learning_rate": 5.437721518987342e-06, "loss": 0.0, "step": 18525 }, { "epoch": 185.96, "learning_rate": 5.431392405063292e-06, "loss": 0.0, "step": 18550 }, { "epoch": 186.22, "learning_rate": 5.425063291139241e-06, "loss": 0.0, "step": 18575 }, { "epoch": 186.47, "learning_rate": 5.4187341772151905e-06, "loss": 0.0, "step": 18600 }, { "epoch": 186.72, "learning_rate": 5.41240506329114e-06, "loss": 0.0, "step": 18625 }, { "epoch": 186.97, "learning_rate": 5.4060759493670885e-06, "loss": 0.0, "step": 18650 }, { "epoch": 187.22, "learning_rate": 5.399746835443038e-06, "loss": 0.0, "step": 18675 }, { "epoch": 187.47, "learning_rate": 5.393417721518987e-06, "loss": 0.0, "step": 18700 }, { "epoch": 187.72, "learning_rate": 5.387088607594937e-06, "loss": 0.0, "step": 18725 }, { "epoch": 187.97, "learning_rate": 5.380759493670886e-06, "loss": 0.0, "step": 18750 }, { "epoch": 188.22, "learning_rate": 5.374430379746836e-06, "loss": 0.0, "step": 18775 }, { "epoch": 188.47, "learning_rate": 5.368101265822785e-06, "loss": 0.0, "step": 18800 }, { "epoch": 188.72, "learning_rate": 5.3617721518987345e-06, "loss": 0.0, "step": 18825 }, { "epoch": 188.97, "learning_rate": 5.355443037974684e-06, "loss": 0.0, "step": 18850 }, { "epoch": 189.22, "learning_rate": 5.3491139240506325e-06, "loss": 0.0, "step": 18875 }, { "epoch": 189.47, "learning_rate": 5.342784810126582e-06, "loss": 0.0, "step": 18900 }, { "epoch": 189.72, "learning_rate": 5.336455696202531e-06, "loss": 0.0, "step": 18925 }, { "epoch": 189.97, "learning_rate": 5.330126582278481e-06, "loss": 0.0, "step": 18950 }, { "epoch": 190.23, "learning_rate": 5.323797468354431e-06, "loss": 0.0, "step": 18975 }, { "epoch": 190.48, "learning_rate": 5.3174683544303805e-06, "loss": 0.0, "step": 19000 }, { "epoch": 190.48, "eval_loss": 0.35610973834991455, "eval_runtime": 1405.7127, "eval_samples_per_second": 9.064, "eval_steps_per_second": 0.284, "eval_wer": 9.30709707675473, "step": 19000 }, { "epoch": 190.73, "learning_rate": 5.31113924050633e-06, "loss": 0.0, "step": 19025 }, { "epoch": 190.98, "learning_rate": 5.304810126582279e-06, "loss": 0.0, "step": 19050 }, { "epoch": 191.23, "learning_rate": 5.298481012658229e-06, "loss": 0.0, "step": 19075 }, { "epoch": 191.48, "learning_rate": 5.292151898734178e-06, "loss": 0.0, "step": 19100 }, { "epoch": 191.73, "learning_rate": 5.285822784810128e-06, "loss": 0.0, "step": 19125 }, { "epoch": 191.98, "learning_rate": 5.279493670886077e-06, "loss": 0.0, "step": 19150 }, { "epoch": 192.23, "learning_rate": 5.273164556962026e-06, "loss": 0.0, "step": 19175 }, { "epoch": 192.48, "learning_rate": 5.266835443037975e-06, "loss": 0.0, "step": 19200 }, { "epoch": 192.73, "learning_rate": 5.2605063291139245e-06, "loss": 0.0, "step": 19225 }, { "epoch": 192.98, "learning_rate": 5.254177215189874e-06, "loss": 0.0, "step": 19250 }, { "epoch": 193.23, "learning_rate": 5.247848101265823e-06, "loss": 0.0, "step": 19275 }, { "epoch": 193.48, "learning_rate": 5.241518987341773e-06, "loss": 0.0, "step": 19300 }, { "epoch": 193.73, "learning_rate": 5.235189873417722e-06, "loss": 0.0, "step": 19325 }, { "epoch": 193.98, "learning_rate": 5.228860759493672e-06, "loss": 0.0, "step": 19350 }, { "epoch": 194.24, "learning_rate": 5.222531645569621e-06, "loss": 0.0, "step": 19375 }, { "epoch": 194.49, "learning_rate": 5.21620253164557e-06, "loss": 0.0, "step": 19400 }, { "epoch": 194.74, "learning_rate": 5.209873417721519e-06, "loss": 0.0, "step": 19425 }, { "epoch": 194.99, "learning_rate": 5.2035443037974685e-06, "loss": 0.0, "step": 19450 }, { "epoch": 195.24, "learning_rate": 5.197215189873418e-06, "loss": 0.0, "step": 19475 }, { "epoch": 195.49, "learning_rate": 5.190886075949367e-06, "loss": 0.0, "step": 19500 }, { "epoch": 195.74, "learning_rate": 5.184556962025317e-06, "loss": 0.0, "step": 19525 }, { "epoch": 195.99, "learning_rate": 5.178227848101266e-06, "loss": 0.0, "step": 19550 }, { "epoch": 196.24, "learning_rate": 5.171898734177216e-06, "loss": 0.0, "step": 19575 }, { "epoch": 196.49, "learning_rate": 5.165569620253165e-06, "loss": 0.0, "step": 19600 }, { "epoch": 196.74, "learning_rate": 5.159240506329114e-06, "loss": 0.0, "step": 19625 }, { "epoch": 196.99, "learning_rate": 5.152911392405063e-06, "loss": 0.0, "step": 19650 }, { "epoch": 197.24, "learning_rate": 5.1465822784810125e-06, "loss": 0.0, "step": 19675 }, { "epoch": 197.49, "learning_rate": 5.140253164556962e-06, "loss": 0.0, "step": 19700 }, { "epoch": 197.74, "learning_rate": 5.133924050632911e-06, "loss": 0.0, "step": 19725 }, { "epoch": 197.99, "learning_rate": 5.127594936708861e-06, "loss": 0.0, "step": 19750 }, { "epoch": 198.25, "learning_rate": 5.12126582278481e-06, "loss": 0.0, "step": 19775 }, { "epoch": 198.5, "learning_rate": 5.1149367088607605e-06, "loss": 0.0, "step": 19800 }, { "epoch": 198.75, "learning_rate": 5.10860759493671e-06, "loss": 0.0, "step": 19825 }, { "epoch": 199.0, "learning_rate": 5.102278481012659e-06, "loss": 0.0, "step": 19850 }, { "epoch": 199.25, "learning_rate": 5.095949367088609e-06, "loss": 0.0, "step": 19875 }, { "epoch": 199.5, "learning_rate": 5.089620253164557e-06, "loss": 0.0, "step": 19900 }, { "epoch": 199.75, "learning_rate": 5.083291139240507e-06, "loss": 0.0, "step": 19925 }, { "epoch": 200.0, "learning_rate": 5.076962025316456e-06, "loss": 0.0, "step": 19950 }, { "epoch": 200.25, "learning_rate": 5.070632911392406e-06, "loss": 0.0, "step": 19975 }, { "epoch": 200.5, "learning_rate": 5.064303797468355e-06, "loss": 0.0, "step": 20000 }, { "epoch": 200.5, "eval_loss": 0.36604076623916626, "eval_runtime": 1406.3835, "eval_samples_per_second": 9.06, "eval_steps_per_second": 0.284, "eval_wer": 9.298303892449585, "step": 20000 }, { "epoch": 200.75, "learning_rate": 5.0579746835443045e-06, "loss": 0.0, "step": 20025 }, { "epoch": 201.0, "learning_rate": 5.051645569620254e-06, "loss": 0.0, "step": 20050 }, { "epoch": 201.25, "learning_rate": 5.045316455696203e-06, "loss": 0.0, "step": 20075 }, { "epoch": 201.5, "learning_rate": 5.038987341772153e-06, "loss": 0.0, "step": 20100 }, { "epoch": 201.75, "learning_rate": 5.032658227848101e-06, "loss": 0.0, "step": 20125 }, { "epoch": 202.01, "learning_rate": 5.026329113924051e-06, "loss": 0.0, "step": 20150 }, { "epoch": 202.26, "learning_rate": 5.02e-06, "loss": 0.0, "step": 20175 }, { "epoch": 202.51, "learning_rate": 5.01367088607595e-06, "loss": 0.0, "step": 20200 }, { "epoch": 202.76, "learning_rate": 5.007341772151899e-06, "loss": 0.0, "step": 20225 }, { "epoch": 203.01, "learning_rate": 5.0010126582278485e-06, "loss": 0.0, "step": 20250 }, { "epoch": 203.26, "learning_rate": 4.994683544303798e-06, "loss": 0.0, "step": 20275 }, { "epoch": 203.51, "learning_rate": 4.988354430379747e-06, "loss": 0.0, "step": 20300 }, { "epoch": 203.76, "learning_rate": 4.982025316455697e-06, "loss": 0.0, "step": 20325 }, { "epoch": 204.01, "learning_rate": 4.975696202531645e-06, "loss": 0.0, "step": 20350 }, { "epoch": 204.26, "learning_rate": 4.969367088607595e-06, "loss": 0.0, "step": 20375 }, { "epoch": 204.51, "learning_rate": 4.963037974683545e-06, "loss": 0.0, "step": 20400 }, { "epoch": 204.76, "learning_rate": 4.9567088607594945e-06, "loss": 0.0, "step": 20425 }, { "epoch": 205.01, "learning_rate": 4.950379746835444e-06, "loss": 0.0, "step": 20450 }, { "epoch": 205.26, "learning_rate": 4.944050632911393e-06, "loss": 0.0, "step": 20475 }, { "epoch": 205.51, "learning_rate": 4.937721518987342e-06, "loss": 0.0, "step": 20500 }, { "epoch": 205.76, "learning_rate": 4.931392405063291e-06, "loss": 0.0, "step": 20525 }, { "epoch": 206.02, "learning_rate": 4.925063291139241e-06, "loss": 0.0, "step": 20550 }, { "epoch": 206.27, "learning_rate": 4.91873417721519e-06, "loss": 0.0, "step": 20575 }, { "epoch": 206.52, "learning_rate": 4.91240506329114e-06, "loss": 0.0, "step": 20600 }, { "epoch": 206.77, "learning_rate": 4.906075949367089e-06, "loss": 0.0, "step": 20625 }, { "epoch": 207.02, "learning_rate": 4.8997468354430385e-06, "loss": 0.0, "step": 20650 }, { "epoch": 207.27, "learning_rate": 4.893417721518988e-06, "loss": 0.0, "step": 20675 }, { "epoch": 207.52, "learning_rate": 4.887088607594937e-06, "loss": 0.0, "step": 20700 }, { "epoch": 207.77, "learning_rate": 4.880759493670886e-06, "loss": 0.0, "step": 20725 }, { "epoch": 208.02, "learning_rate": 4.874430379746835e-06, "loss": 0.0, "step": 20750 }, { "epoch": 208.27, "learning_rate": 4.868101265822785e-06, "loss": 0.0, "step": 20775 }, { "epoch": 208.52, "learning_rate": 4.861772151898734e-06, "loss": 0.0, "step": 20800 }, { "epoch": 208.77, "learning_rate": 4.8554430379746845e-06, "loss": 0.0, "step": 20825 }, { "epoch": 209.02, "learning_rate": 4.849113924050634e-06, "loss": 0.0, "step": 20850 }, { "epoch": 209.27, "learning_rate": 4.8427848101265825e-06, "loss": 0.0, "step": 20875 }, { "epoch": 209.52, "learning_rate": 4.836455696202532e-06, "loss": 0.0, "step": 20900 }, { "epoch": 209.77, "learning_rate": 4.830126582278481e-06, "loss": 0.0, "step": 20925 }, { "epoch": 210.03, "learning_rate": 4.823797468354431e-06, "loss": 0.0, "step": 20950 }, { "epoch": 210.28, "learning_rate": 4.81746835443038e-06, "loss": 0.0, "step": 20975 }, { "epoch": 210.53, "learning_rate": 4.81113924050633e-06, "loss": 0.0, "step": 21000 }, { "epoch": 210.53, "eval_loss": 0.37553900480270386, "eval_runtime": 1405.7397, "eval_samples_per_second": 9.064, "eval_steps_per_second": 0.284, "eval_wer": 9.287556667187745, "step": 21000 }, { "epoch": 210.78, "learning_rate": 4.804810126582279e-06, "loss": 0.0, "step": 21025 }, { "epoch": 211.03, "learning_rate": 4.7984810126582285e-06, "loss": 0.0, "step": 21050 }, { "epoch": 211.28, "learning_rate": 4.792151898734178e-06, "loss": 0.0, "step": 21075 }, { "epoch": 211.53, "learning_rate": 4.7858227848101265e-06, "loss": 0.0, "step": 21100 }, { "epoch": 211.78, "learning_rate": 4.779493670886076e-06, "loss": 0.0, "step": 21125 }, { "epoch": 212.03, "learning_rate": 4.773164556962025e-06, "loss": 0.0, "step": 21150 }, { "epoch": 212.28, "learning_rate": 4.766835443037975e-06, "loss": 0.0, "step": 21175 }, { "epoch": 212.53, "learning_rate": 4.760506329113924e-06, "loss": 0.0, "step": 21200 }, { "epoch": 212.78, "learning_rate": 4.7541772151898745e-06, "loss": 0.0, "step": 21225 }, { "epoch": 213.03, "learning_rate": 4.747848101265823e-06, "loss": 0.0, "step": 21250 }, { "epoch": 213.28, "learning_rate": 4.7415189873417725e-06, "loss": 0.0, "step": 21275 }, { "epoch": 213.53, "learning_rate": 4.735189873417722e-06, "loss": 0.0, "step": 21300 }, { "epoch": 213.78, "learning_rate": 4.728860759493671e-06, "loss": 0.0, "step": 21325 }, { "epoch": 214.04, "learning_rate": 4.722531645569621e-06, "loss": 0.0, "step": 21350 }, { "epoch": 214.29, "learning_rate": 4.71620253164557e-06, "loss": 0.0, "step": 21375 }, { "epoch": 214.54, "learning_rate": 4.70987341772152e-06, "loss": 0.0, "step": 21400 }, { "epoch": 214.79, "learning_rate": 4.703544303797469e-06, "loss": 0.0, "step": 21425 }, { "epoch": 215.04, "learning_rate": 4.6972151898734185e-06, "loss": 0.0, "step": 21450 }, { "epoch": 215.29, "learning_rate": 4.690886075949367e-06, "loss": 0.0, "step": 21475 }, { "epoch": 215.54, "learning_rate": 4.6845569620253165e-06, "loss": 0.0, "step": 21500 }, { "epoch": 215.79, "learning_rate": 4.678227848101266e-06, "loss": 0.0, "step": 21525 }, { "epoch": 216.04, "learning_rate": 4.671898734177215e-06, "loss": 0.0, "step": 21550 }, { "epoch": 216.29, "learning_rate": 4.665569620253165e-06, "loss": 0.0, "step": 21575 }, { "epoch": 216.54, "learning_rate": 4.659240506329114e-06, "loss": 0.0, "step": 21600 }, { "epoch": 216.79, "learning_rate": 4.652911392405064e-06, "loss": 0.0, "step": 21625 }, { "epoch": 217.04, "learning_rate": 4.646582278481013e-06, "loss": 0.0, "step": 21650 }, { "epoch": 217.29, "learning_rate": 4.6402531645569625e-06, "loss": 0.0, "step": 21675 }, { "epoch": 217.54, "learning_rate": 4.633924050632912e-06, "loss": 0.0, "step": 21700 }, { "epoch": 217.79, "learning_rate": 4.627594936708861e-06, "loss": 0.0, "step": 21725 }, { "epoch": 218.05, "learning_rate": 4.621265822784811e-06, "loss": 0.0, "step": 21750 }, { "epoch": 218.3, "learning_rate": 4.61493670886076e-06, "loss": 0.0, "step": 21775 }, { "epoch": 218.55, "learning_rate": 4.60860759493671e-06, "loss": 0.0, "step": 21800 }, { "epoch": 218.8, "learning_rate": 4.602278481012658e-06, "loss": 0.0, "step": 21825 }, { "epoch": 219.05, "learning_rate": 4.595949367088608e-06, "loss": 0.0, "step": 21850 }, { "epoch": 219.3, "learning_rate": 4.589620253164557e-06, "loss": 0.0, "step": 21875 }, { "epoch": 219.55, "learning_rate": 4.5832911392405065e-06, "loss": 0.0, "step": 21900 }, { "epoch": 219.8, "learning_rate": 4.576962025316456e-06, "loss": 0.0, "step": 21925 }, { "epoch": 220.05, "learning_rate": 4.570632911392405e-06, "loss": 0.0, "step": 21950 }, { "epoch": 220.3, "learning_rate": 4.564303797468355e-06, "loss": 0.0, "step": 21975 }, { "epoch": 220.55, "learning_rate": 4.557974683544304e-06, "loss": 0.0, "step": 22000 }, { "epoch": 220.55, "eval_loss": 0.3847161531448364, "eval_runtime": 1410.8906, "eval_samples_per_second": 9.031, "eval_steps_per_second": 0.283, "eval_wer": 9.497616070032828, "step": 22000 }, { "epoch": 220.8, "learning_rate": 4.551645569620254e-06, "loss": 0.0, "step": 22025 }, { "epoch": 221.05, "learning_rate": 4.545316455696203e-06, "loss": 0.0, "step": 22050 }, { "epoch": 221.3, "learning_rate": 4.5389873417721525e-06, "loss": 0.0, "step": 22075 }, { "epoch": 221.55, "learning_rate": 4.532658227848102e-06, "loss": 0.0, "step": 22100 }, { "epoch": 221.8, "learning_rate": 4.526329113924051e-06, "loss": 0.0, "step": 22125 }, { "epoch": 222.06, "learning_rate": 4.520000000000001e-06, "loss": 0.0, "step": 22150 }, { "epoch": 222.31, "learning_rate": 4.51367088607595e-06, "loss": 0.0, "step": 22175 }, { "epoch": 222.56, "learning_rate": 4.507341772151899e-06, "loss": 0.0, "step": 22200 }, { "epoch": 222.81, "learning_rate": 4.501012658227848e-06, "loss": 0.0, "step": 22225 }, { "epoch": 223.06, "learning_rate": 4.494683544303798e-06, "loss": 0.0, "step": 22250 }, { "epoch": 223.31, "learning_rate": 4.488354430379747e-06, "loss": 0.0, "step": 22275 }, { "epoch": 223.56, "learning_rate": 4.4820253164556965e-06, "loss": 0.0, "step": 22300 }, { "epoch": 223.81, "learning_rate": 4.475696202531646e-06, "loss": 0.0, "step": 22325 }, { "epoch": 224.06, "learning_rate": 4.469367088607595e-06, "loss": 0.0, "step": 22350 }, { "epoch": 224.31, "learning_rate": 4.463037974683545e-06, "loss": 0.0, "step": 22375 }, { "epoch": 224.56, "learning_rate": 4.456708860759494e-06, "loss": 0.0, "step": 22400 }, { "epoch": 224.81, "learning_rate": 4.450379746835443e-06, "loss": 0.0, "step": 22425 }, { "epoch": 225.06, "learning_rate": 4.444050632911392e-06, "loss": 0.0, "step": 22450 }, { "epoch": 225.31, "learning_rate": 4.4377215189873425e-06, "loss": 0.0, "step": 22475 }, { "epoch": 225.56, "learning_rate": 4.431392405063292e-06, "loss": 0.0, "step": 22500 }, { "epoch": 225.81, "learning_rate": 4.425063291139241e-06, "loss": 0.0, "step": 22525 }, { "epoch": 226.07, "learning_rate": 4.418734177215191e-06, "loss": 0.0, "step": 22550 }, { "epoch": 226.32, "learning_rate": 4.412405063291139e-06, "loss": 0.0, "step": 22575 }, { "epoch": 226.57, "learning_rate": 4.406075949367089e-06, "loss": 0.0, "step": 22600 }, { "epoch": 226.82, "learning_rate": 4.399746835443038e-06, "loss": 0.0, "step": 22625 }, { "epoch": 227.07, "learning_rate": 4.393417721518988e-06, "loss": 0.0, "step": 22650 }, { "epoch": 227.32, "learning_rate": 4.387088607594937e-06, "loss": 0.0, "step": 22675 }, { "epoch": 227.57, "learning_rate": 4.3807594936708865e-06, "loss": 0.0, "step": 22700 }, { "epoch": 227.82, "learning_rate": 4.374430379746836e-06, "loss": 0.0, "step": 22725 }, { "epoch": 228.07, "learning_rate": 4.368101265822785e-06, "loss": 0.0, "step": 22750 }, { "epoch": 228.32, "learning_rate": 4.361772151898735e-06, "loss": 0.0, "step": 22775 }, { "epoch": 228.57, "learning_rate": 4.355443037974683e-06, "loss": 0.0, "step": 22800 }, { "epoch": 228.82, "learning_rate": 4.349113924050633e-06, "loss": 0.0, "step": 22825 }, { "epoch": 229.07, "learning_rate": 4.342784810126582e-06, "loss": 0.0, "step": 22850 }, { "epoch": 229.32, "learning_rate": 4.3364556962025325e-06, "loss": 0.0, "step": 22875 }, { "epoch": 229.57, "learning_rate": 4.330126582278482e-06, "loss": 0.0, "step": 22900 }, { "epoch": 229.82, "learning_rate": 4.323797468354431e-06, "loss": 0.0, "step": 22925 }, { "epoch": 230.08, "learning_rate": 4.31746835443038e-06, "loss": 0.0, "step": 22950 }, { "epoch": 230.33, "learning_rate": 4.311139240506329e-06, "loss": 0.0, "step": 22975 }, { "epoch": 230.58, "learning_rate": 4.304810126582279e-06, "loss": 0.0, "step": 23000 }, { "epoch": 230.58, "eval_loss": 0.39401647448539734, "eval_runtime": 1412.4195, "eval_samples_per_second": 9.021, "eval_steps_per_second": 0.282, "eval_wer": 9.505432233859622, "step": 23000 }, { "epoch": 230.83, "learning_rate": 4.298481012658228e-06, "loss": 0.0, "step": 23025 }, { "epoch": 231.08, "learning_rate": 4.292151898734178e-06, "loss": 0.0, "step": 23050 }, { "epoch": 231.33, "learning_rate": 4.285822784810127e-06, "loss": 0.0, "step": 23075 }, { "epoch": 231.58, "learning_rate": 4.2794936708860765e-06, "loss": 0.0, "step": 23100 }, { "epoch": 231.83, "learning_rate": 4.273164556962026e-06, "loss": 0.0, "step": 23125 }, { "epoch": 232.08, "learning_rate": 4.266835443037975e-06, "loss": 0.0, "step": 23150 }, { "epoch": 232.33, "learning_rate": 4.260506329113924e-06, "loss": 0.0, "step": 23175 }, { "epoch": 232.58, "learning_rate": 4.254177215189873e-06, "loss": 0.0, "step": 23200 }, { "epoch": 232.83, "learning_rate": 4.247848101265823e-06, "loss": 0.0, "step": 23225 }, { "epoch": 233.08, "learning_rate": 4.241518987341772e-06, "loss": 0.0, "step": 23250 }, { "epoch": 233.33, "learning_rate": 4.2351898734177225e-06, "loss": 0.0, "step": 23275 }, { "epoch": 233.58, "learning_rate": 4.228860759493671e-06, "loss": 0.0, "step": 23300 }, { "epoch": 233.83, "learning_rate": 4.2225316455696205e-06, "loss": 0.0, "step": 23325 }, { "epoch": 234.09, "learning_rate": 4.21620253164557e-06, "loss": 0.0, "step": 23350 }, { "epoch": 234.34, "learning_rate": 4.209873417721519e-06, "loss": 0.0, "step": 23375 }, { "epoch": 234.59, "learning_rate": 4.203544303797469e-06, "loss": 0.0, "step": 23400 }, { "epoch": 234.84, "learning_rate": 4.197215189873418e-06, "loss": 0.0, "step": 23425 }, { "epoch": 235.09, "learning_rate": 4.190886075949368e-06, "loss": 0.0, "step": 23450 }, { "epoch": 235.34, "learning_rate": 4.184556962025317e-06, "loss": 0.0, "step": 23475 }, { "epoch": 235.59, "learning_rate": 4.1782278481012665e-06, "loss": 0.0, "step": 23500 }, { "epoch": 235.84, "learning_rate": 4.171898734177215e-06, "loss": 0.0, "step": 23525 }, { "epoch": 236.09, "learning_rate": 4.1655696202531645e-06, "loss": 0.0, "step": 23550 }, { "epoch": 236.34, "learning_rate": 4.159240506329114e-06, "loss": 0.0, "step": 23575 }, { "epoch": 236.59, "learning_rate": 4.152911392405063e-06, "loss": 0.0, "step": 23600 }, { "epoch": 236.84, "learning_rate": 4.146582278481013e-06, "loss": 0.0, "step": 23625 }, { "epoch": 237.09, "learning_rate": 4.140253164556962e-06, "loss": 0.0, "step": 23650 }, { "epoch": 237.34, "learning_rate": 4.133924050632912e-06, "loss": 0.0, "step": 23675 }, { "epoch": 237.59, "learning_rate": 4.127594936708861e-06, "loss": 0.0, "step": 23700 }, { "epoch": 237.84, "learning_rate": 4.1212658227848105e-06, "loss": 0.0, "step": 23725 }, { "epoch": 238.1, "learning_rate": 4.11493670886076e-06, "loss": 0.0, "step": 23750 }, { "epoch": 238.35, "learning_rate": 4.108607594936709e-06, "loss": 0.0, "step": 23775 }, { "epoch": 238.6, "learning_rate": 4.102278481012659e-06, "loss": 0.0, "step": 23800 }, { "epoch": 238.85, "learning_rate": 4.09620253164557e-06, "loss": 0.0, "step": 23825 }, { "epoch": 239.1, "learning_rate": 4.08987341772152e-06, "loss": 0.0, "step": 23850 }, { "epoch": 239.35, "learning_rate": 4.083544303797469e-06, "loss": 0.0, "step": 23875 }, { "epoch": 239.6, "learning_rate": 4.077215189873418e-06, "loss": 0.0, "step": 23900 }, { "epoch": 239.85, "learning_rate": 4.070886075949367e-06, "loss": 0.0, "step": 23925 }, { "epoch": 240.1, "learning_rate": 4.0645569620253165e-06, "loss": 0.0, "step": 23950 }, { "epoch": 240.35, "learning_rate": 4.058227848101266e-06, "loss": 0.0, "step": 23975 }, { "epoch": 240.6, "learning_rate": 4.051898734177215e-06, "loss": 0.0, "step": 24000 }, { "epoch": 240.6, "eval_loss": 0.4021439850330353, "eval_runtime": 1410.5501, "eval_samples_per_second": 9.033, "eval_steps_per_second": 0.283, "eval_wer": 9.47025949663905, "step": 24000 }, { "epoch": 240.85, "learning_rate": 4.045569620253165e-06, "loss": 0.0, "step": 24025 }, { "epoch": 241.1, "learning_rate": 4.039240506329114e-06, "loss": 0.0, "step": 24050 }, { "epoch": 241.35, "learning_rate": 4.032911392405064e-06, "loss": 0.0, "step": 24075 }, { "epoch": 241.6, "learning_rate": 4.026582278481013e-06, "loss": 0.0, "step": 24100 }, { "epoch": 241.85, "learning_rate": 4.0202531645569625e-06, "loss": 0.0, "step": 24125 }, { "epoch": 242.11, "learning_rate": 4.013924050632912e-06, "loss": 0.0, "step": 24150 }, { "epoch": 242.36, "learning_rate": 4.007594936708861e-06, "loss": 0.0, "step": 24175 }, { "epoch": 242.61, "learning_rate": 4.001265822784811e-06, "loss": 0.0, "step": 24200 }, { "epoch": 242.86, "learning_rate": 3.99493670886076e-06, "loss": 0.0, "step": 24225 }, { "epoch": 243.11, "learning_rate": 3.98860759493671e-06, "loss": 0.0, "step": 24250 }, { "epoch": 243.36, "learning_rate": 3.982278481012658e-06, "loss": 0.0, "step": 24275 }, { "epoch": 243.61, "learning_rate": 3.975949367088608e-06, "loss": 0.0, "step": 24300 }, { "epoch": 243.86, "learning_rate": 3.969620253164557e-06, "loss": 0.0, "step": 24325 }, { "epoch": 244.11, "learning_rate": 3.9632911392405065e-06, "loss": 0.0, "step": 24350 }, { "epoch": 244.36, "learning_rate": 3.956962025316456e-06, "loss": 0.0, "step": 24375 }, { "epoch": 244.61, "learning_rate": 3.950632911392405e-06, "loss": 0.0, "step": 24400 }, { "epoch": 244.86, "learning_rate": 3.944303797468355e-06, "loss": 0.0, "step": 24425 }, { "epoch": 245.11, "learning_rate": 3.937974683544304e-06, "loss": 0.0, "step": 24450 }, { "epoch": 245.36, "learning_rate": 3.931645569620254e-06, "loss": 0.0, "step": 24475 }, { "epoch": 245.61, "learning_rate": 3.925316455696203e-06, "loss": 0.0, "step": 24500 }, { "epoch": 245.86, "learning_rate": 3.9189873417721525e-06, "loss": 0.0, "step": 24525 }, { "epoch": 246.12, "learning_rate": 3.912658227848102e-06, "loss": 0.0, "step": 24550 }, { "epoch": 246.37, "learning_rate": 3.906329113924051e-06, "loss": 0.0, "step": 24575 }, { "epoch": 246.62, "learning_rate": 3.900000000000001e-06, "loss": 0.0, "step": 24600 }, { "epoch": 246.87, "learning_rate": 3.89367088607595e-06, "loss": 0.0, "step": 24625 }, { "epoch": 247.12, "learning_rate": 3.887341772151899e-06, "loss": 0.0, "step": 24650 }, { "epoch": 247.37, "learning_rate": 3.881012658227848e-06, "loss": 0.0, "step": 24675 }, { "epoch": 247.62, "learning_rate": 3.874683544303798e-06, "loss": 0.0, "step": 24700 }, { "epoch": 247.87, "learning_rate": 3.868354430379747e-06, "loss": 0.0, "step": 24725 }, { "epoch": 248.12, "learning_rate": 3.8620253164556965e-06, "loss": 0.0, "step": 24750 }, { "epoch": 248.37, "learning_rate": 3.855696202531646e-06, "loss": 0.0, "step": 24775 }, { "epoch": 248.62, "learning_rate": 3.849367088607595e-06, "loss": 0.0, "step": 24800 }, { "epoch": 248.87, "learning_rate": 3.843037974683545e-06, "loss": 0.0, "step": 24825 }, { "epoch": 249.12, "learning_rate": 3.836708860759493e-06, "loss": 0.0, "step": 24850 }, { "epoch": 249.37, "learning_rate": 3.830379746835443e-06, "loss": 0.0, "step": 24875 }, { "epoch": 249.62, "learning_rate": 3.824050632911392e-06, "loss": 0.0, "step": 24900 }, { "epoch": 249.87, "learning_rate": 3.8177215189873425e-06, "loss": 0.0, "step": 24925 }, { "epoch": 250.13, "learning_rate": 3.8113924050632915e-06, "loss": 0.0, "step": 24950 }, { "epoch": 250.38, "learning_rate": 3.805063291139241e-06, "loss": 0.0, "step": 24975 }, { "epoch": 250.63, "learning_rate": 3.7987341772151903e-06, "loss": 0.0, "step": 25000 }, { "epoch": 250.63, "eval_loss": 0.4126415252685547, "eval_runtime": 1409.5997, "eval_samples_per_second": 9.039, "eval_steps_per_second": 0.283, "eval_wer": 9.453650148507114, "step": 25000 }, { "epoch": 250.88, "learning_rate": 3.7924050632911398e-06, "loss": 0.0, "step": 25025 }, { "epoch": 251.13, "learning_rate": 3.786075949367089e-06, "loss": 0.0, "step": 25050 }, { "epoch": 251.38, "learning_rate": 3.779746835443038e-06, "loss": 0.0, "step": 25075 }, { "epoch": 251.63, "learning_rate": 3.7734177215189876e-06, "loss": 0.0, "step": 25100 }, { "epoch": 251.88, "learning_rate": 3.767088607594937e-06, "loss": 0.0, "step": 25125 }, { "epoch": 252.13, "learning_rate": 3.7607594936708865e-06, "loss": 0.0, "step": 25150 }, { "epoch": 252.38, "learning_rate": 3.7544303797468355e-06, "loss": 0.0, "step": 25175 }, { "epoch": 252.63, "learning_rate": 3.748101265822785e-06, "loss": 0.0, "step": 25200 }, { "epoch": 252.88, "learning_rate": 3.7417721518987344e-06, "loss": 0.0, "step": 25225 }, { "epoch": 253.13, "learning_rate": 3.7354430379746838e-06, "loss": 0.0, "step": 25250 }, { "epoch": 253.38, "learning_rate": 3.7291139240506328e-06, "loss": 0.0, "step": 25275 }, { "epoch": 253.63, "learning_rate": 3.722784810126582e-06, "loss": 0.0, "step": 25300 }, { "epoch": 253.88, "learning_rate": 3.716455696202532e-06, "loss": 0.0, "step": 25325 }, { "epoch": 254.14, "learning_rate": 3.7103797468354435e-06, "loss": 0.0003, "step": 25350 }, { "epoch": 254.39, "learning_rate": 3.704303797468355e-06, "loss": 0.0016, "step": 25375 }, { "epoch": 254.64, "learning_rate": 3.6979746835443044e-06, "loss": 0.0049, "step": 25400 }, { "epoch": 254.89, "learning_rate": 3.6916455696202534e-06, "loss": 0.0051, "step": 25425 }, { "epoch": 255.14, "learning_rate": 3.685316455696203e-06, "loss": 0.003, "step": 25450 }, { "epoch": 255.39, "learning_rate": 3.6789873417721523e-06, "loss": 0.0024, "step": 25475 }, { "epoch": 255.64, "learning_rate": 3.6726582278481017e-06, "loss": 0.0022, "step": 25500 }, { "epoch": 255.89, "learning_rate": 3.6663291139240507e-06, "loss": 0.0018, "step": 25525 }, { "epoch": 256.14, "learning_rate": 3.66e-06, "loss": 0.0008, "step": 25550 }, { "epoch": 256.39, "learning_rate": 3.6536708860759495e-06, "loss": 0.0004, "step": 25575 }, { "epoch": 256.64, "learning_rate": 3.647341772151899e-06, "loss": 0.0003, "step": 25600 }, { "epoch": 256.89, "learning_rate": 3.6410126582278484e-06, "loss": 0.0003, "step": 25625 }, { "epoch": 257.14, "learning_rate": 3.6346835443037974e-06, "loss": 0.0002, "step": 25650 }, { "epoch": 257.39, "learning_rate": 3.628354430379747e-06, "loss": 0.0002, "step": 25675 }, { "epoch": 257.64, "learning_rate": 3.6220253164556967e-06, "loss": 0.0001, "step": 25700 }, { "epoch": 257.89, "learning_rate": 3.615696202531646e-06, "loss": 0.0002, "step": 25725 }, { "epoch": 258.15, "learning_rate": 3.6093670886075955e-06, "loss": 0.0001, "step": 25750 }, { "epoch": 258.4, "learning_rate": 3.603037974683545e-06, "loss": 0.0001, "step": 25775 }, { "epoch": 258.65, "learning_rate": 3.596708860759494e-06, "loss": 0.0001, "step": 25800 }, { "epoch": 258.9, "learning_rate": 3.5903797468354434e-06, "loss": 0.0, "step": 25825 }, { "epoch": 259.15, "learning_rate": 3.584050632911393e-06, "loss": 0.0, "step": 25850 }, { "epoch": 259.4, "learning_rate": 3.5777215189873422e-06, "loss": 0.0, "step": 25875 }, { "epoch": 259.65, "learning_rate": 3.5713924050632913e-06, "loss": 0.0, "step": 25900 }, { "epoch": 259.9, "learning_rate": 3.5650632911392407e-06, "loss": 0.0, "step": 25925 }, { "epoch": 260.15, "learning_rate": 3.55873417721519e-06, "loss": 0.0, "step": 25950 }, { "epoch": 260.4, "learning_rate": 3.5524050632911395e-06, "loss": 0.0, "step": 25975 }, { "epoch": 260.65, "learning_rate": 3.546075949367089e-06, "loss": 0.0, "step": 26000 }, { "epoch": 260.65, "eval_loss": 0.31743550300598145, "eval_runtime": 1406.6303, "eval_samples_per_second": 9.059, "eval_steps_per_second": 0.284, "eval_wer": 9.275832421447554, "step": 26000 }, { "epoch": 260.9, "learning_rate": 3.539746835443038e-06, "loss": 0.0, "step": 26025 }, { "epoch": 261.15, "learning_rate": 3.5334177215189874e-06, "loss": 0.0, "step": 26050 }, { "epoch": 261.4, "learning_rate": 3.527088607594937e-06, "loss": 0.0, "step": 26075 }, { "epoch": 261.65, "learning_rate": 3.5207594936708862e-06, "loss": 0.0, "step": 26100 }, { "epoch": 261.9, "learning_rate": 3.514430379746836e-06, "loss": 0.0, "step": 26125 }, { "epoch": 262.16, "learning_rate": 3.5081012658227855e-06, "loss": 0.0, "step": 26150 }, { "epoch": 262.41, "learning_rate": 3.5017721518987345e-06, "loss": 0.0, "step": 26175 }, { "epoch": 262.66, "learning_rate": 3.495443037974684e-06, "loss": 0.0, "step": 26200 }, { "epoch": 262.91, "learning_rate": 3.4891139240506334e-06, "loss": 0.0, "step": 26225 }, { "epoch": 263.16, "learning_rate": 3.482784810126583e-06, "loss": 0.0, "step": 26250 }, { "epoch": 263.41, "learning_rate": 3.476455696202532e-06, "loss": 0.0, "step": 26275 }, { "epoch": 263.66, "learning_rate": 3.4701265822784812e-06, "loss": 0.0, "step": 26300 }, { "epoch": 263.91, "learning_rate": 3.4637974683544307e-06, "loss": 0.0, "step": 26325 }, { "epoch": 264.16, "learning_rate": 3.45746835443038e-06, "loss": 0.0, "step": 26350 }, { "epoch": 264.41, "learning_rate": 3.451139240506329e-06, "loss": 0.0, "step": 26375 }, { "epoch": 264.66, "learning_rate": 3.4448101265822785e-06, "loss": 0.0, "step": 26400 }, { "epoch": 264.91, "learning_rate": 3.438481012658228e-06, "loss": 0.0, "step": 26425 }, { "epoch": 265.16, "learning_rate": 3.4321518987341774e-06, "loss": 0.0, "step": 26450 }, { "epoch": 265.41, "learning_rate": 3.425822784810127e-06, "loss": 0.0, "step": 26475 }, { "epoch": 265.66, "learning_rate": 3.419493670886076e-06, "loss": 0.0, "step": 26500 }, { "epoch": 265.91, "learning_rate": 3.4131645569620257e-06, "loss": 0.0, "step": 26525 }, { "epoch": 266.17, "learning_rate": 3.406835443037975e-06, "loss": 0.0, "step": 26550 }, { "epoch": 266.42, "learning_rate": 3.4005063291139245e-06, "loss": 0.0, "step": 26575 }, { "epoch": 266.67, "learning_rate": 3.394177215189874e-06, "loss": 0.0, "step": 26600 }, { "epoch": 266.92, "learning_rate": 3.3878481012658234e-06, "loss": 0.0, "step": 26625 }, { "epoch": 267.17, "learning_rate": 3.3815189873417724e-06, "loss": 0.0, "step": 26650 }, { "epoch": 267.42, "learning_rate": 3.375189873417722e-06, "loss": 0.0, "step": 26675 }, { "epoch": 267.67, "learning_rate": 3.3688607594936712e-06, "loss": 0.0, "step": 26700 }, { "epoch": 267.92, "learning_rate": 3.3625316455696207e-06, "loss": 0.0, "step": 26725 }, { "epoch": 268.17, "learning_rate": 3.3562025316455697e-06, "loss": 0.0, "step": 26750 }, { "epoch": 268.42, "learning_rate": 3.349873417721519e-06, "loss": 0.0, "step": 26775 }, { "epoch": 268.67, "learning_rate": 3.3435443037974685e-06, "loss": 0.0, "step": 26800 }, { "epoch": 268.92, "learning_rate": 3.337215189873418e-06, "loss": 0.0, "step": 26825 }, { "epoch": 269.17, "learning_rate": 3.3308860759493674e-06, "loss": 0.0, "step": 26850 }, { "epoch": 269.42, "learning_rate": 3.3245569620253164e-06, "loss": 0.0, "step": 26875 }, { "epoch": 269.67, "learning_rate": 3.318227848101266e-06, "loss": 0.0, "step": 26900 }, { "epoch": 269.92, "learning_rate": 3.3118987341772152e-06, "loss": 0.0, "step": 26925 }, { "epoch": 270.18, "learning_rate": 3.305569620253165e-06, "loss": 0.0, "step": 26950 }, { "epoch": 270.43, "learning_rate": 3.2992405063291145e-06, "loss": 0.0, "step": 26975 }, { "epoch": 270.68, "learning_rate": 3.292911392405064e-06, "loss": 0.0, "step": 27000 }, { "epoch": 270.68, "eval_loss": 0.34435129165649414, "eval_runtime": 1407.0553, "eval_samples_per_second": 9.056, "eval_steps_per_second": 0.284, "eval_wer": 9.262154134750665, "step": 27000 }, { "epoch": 270.93, "learning_rate": 3.286582278481013e-06, "loss": 0.0, "step": 27025 }, { "epoch": 271.18, "learning_rate": 3.2802531645569624e-06, "loss": 0.0, "step": 27050 }, { "epoch": 271.43, "learning_rate": 3.273924050632912e-06, "loss": 0.0, "step": 27075 }, { "epoch": 271.68, "learning_rate": 3.2675949367088612e-06, "loss": 0.0, "step": 27100 }, { "epoch": 271.93, "learning_rate": 3.2612658227848102e-06, "loss": 0.0, "step": 27125 }, { "epoch": 272.18, "learning_rate": 3.2549367088607597e-06, "loss": 0.0, "step": 27150 }, { "epoch": 272.43, "learning_rate": 3.248607594936709e-06, "loss": 0.0, "step": 27175 }, { "epoch": 272.68, "learning_rate": 3.2422784810126585e-06, "loss": 0.0, "step": 27200 }, { "epoch": 272.93, "learning_rate": 3.2359493670886075e-06, "loss": 0.0, "step": 27225 }, { "epoch": 273.18, "learning_rate": 3.229620253164557e-06, "loss": 0.0, "step": 27250 }, { "epoch": 273.43, "learning_rate": 3.2232911392405064e-06, "loss": 0.0, "step": 27275 }, { "epoch": 273.68, "learning_rate": 3.216962025316456e-06, "loss": 0.0, "step": 27300 }, { "epoch": 273.93, "learning_rate": 3.2106329113924052e-06, "loss": 0.0, "step": 27325 }, { "epoch": 274.19, "learning_rate": 3.204303797468355e-06, "loss": 0.0, "step": 27350 }, { "epoch": 274.44, "learning_rate": 3.197974683544304e-06, "loss": 0.0, "step": 27375 }, { "epoch": 274.69, "learning_rate": 3.1916455696202535e-06, "loss": 0.0, "step": 27400 }, { "epoch": 274.94, "learning_rate": 3.185316455696203e-06, "loss": 0.0, "step": 27425 }, { "epoch": 275.19, "learning_rate": 3.1789873417721524e-06, "loss": 0.0, "step": 27450 }, { "epoch": 275.44, "learning_rate": 3.172658227848102e-06, "loss": 0.0, "step": 27475 }, { "epoch": 275.69, "learning_rate": 3.166329113924051e-06, "loss": 0.0, "step": 27500 }, { "epoch": 275.94, "learning_rate": 3.1600000000000002e-06, "loss": 0.0, "step": 27525 }, { "epoch": 276.19, "learning_rate": 3.1536708860759497e-06, "loss": 0.0, "step": 27550 }, { "epoch": 276.44, "learning_rate": 3.147341772151899e-06, "loss": 0.0, "step": 27575 }, { "epoch": 276.69, "learning_rate": 3.141012658227848e-06, "loss": 0.0, "step": 27600 }, { "epoch": 276.94, "learning_rate": 3.1346835443037975e-06, "loss": 0.0, "step": 27625 }, { "epoch": 277.19, "learning_rate": 3.128354430379747e-06, "loss": 0.0, "step": 27650 }, { "epoch": 277.44, "learning_rate": 3.1220253164556964e-06, "loss": 0.0, "step": 27675 }, { "epoch": 277.69, "learning_rate": 3.115696202531646e-06, "loss": 0.0, "step": 27700 }, { "epoch": 277.94, "learning_rate": 3.109367088607595e-06, "loss": 0.0, "step": 27725 }, { "epoch": 278.2, "learning_rate": 3.1030379746835442e-06, "loss": 0.0, "step": 27750 }, { "epoch": 278.45, "learning_rate": 3.096708860759494e-06, "loss": 0.0, "step": 27775 }, { "epoch": 278.7, "learning_rate": 3.0903797468354435e-06, "loss": 0.0, "step": 27800 }, { "epoch": 278.95, "learning_rate": 3.084050632911393e-06, "loss": 0.0, "step": 27825 }, { "epoch": 279.2, "learning_rate": 3.0777215189873424e-06, "loss": 0.0, "step": 27850 }, { "epoch": 279.45, "learning_rate": 3.0713924050632914e-06, "loss": 0.0, "step": 27875 }, { "epoch": 279.7, "learning_rate": 3.065063291139241e-06, "loss": 0.0, "step": 27900 }, { "epoch": 279.95, "learning_rate": 3.0587341772151902e-06, "loss": 0.0, "step": 27925 }, { "epoch": 280.2, "learning_rate": 3.0524050632911397e-06, "loss": 0.0, "step": 27950 }, { "epoch": 280.45, "learning_rate": 3.0460759493670887e-06, "loss": 0.0, "step": 27975 }, { "epoch": 280.7, "learning_rate": 3.039746835443038e-06, "loss": 0.0, "step": 28000 }, { "epoch": 280.7, "eval_loss": 0.35884568095207214, "eval_runtime": 1406.9421, "eval_samples_per_second": 9.057, "eval_steps_per_second": 0.284, "eval_wer": 9.208418008441457, "step": 28000 }, { "epoch": 280.95, "learning_rate": 3.0334177215189875e-06, "loss": 0.0, "step": 28025 }, { "epoch": 281.2, "learning_rate": 3.027088607594937e-06, "loss": 0.0, "step": 28050 }, { "epoch": 281.45, "learning_rate": 3.0207594936708864e-06, "loss": 0.0, "step": 28075 }, { "epoch": 281.7, "learning_rate": 3.0144303797468354e-06, "loss": 0.0, "step": 28100 }, { "epoch": 281.95, "learning_rate": 3.008101265822785e-06, "loss": 0.0, "step": 28125 }, { "epoch": 282.21, "learning_rate": 3.0017721518987342e-06, "loss": 0.0, "step": 28150 }, { "epoch": 282.46, "learning_rate": 2.995443037974684e-06, "loss": 0.0, "step": 28175 }, { "epoch": 282.71, "learning_rate": 2.9891139240506335e-06, "loss": 0.0, "step": 28200 }, { "epoch": 282.96, "learning_rate": 2.9827848101265825e-06, "loss": 0.0, "step": 28225 }, { "epoch": 283.21, "learning_rate": 2.976455696202532e-06, "loss": 0.0, "step": 28250 }, { "epoch": 283.46, "learning_rate": 2.9701265822784814e-06, "loss": 0.0, "step": 28275 }, { "epoch": 283.71, "learning_rate": 2.963797468354431e-06, "loss": 0.0, "step": 28300 }, { "epoch": 283.96, "learning_rate": 2.9574683544303802e-06, "loss": 0.0, "step": 28325 }, { "epoch": 284.21, "learning_rate": 2.9511392405063292e-06, "loss": 0.0, "step": 28350 }, { "epoch": 284.46, "learning_rate": 2.9448101265822787e-06, "loss": 0.0, "step": 28375 }, { "epoch": 284.71, "learning_rate": 2.938481012658228e-06, "loss": 0.0, "step": 28400 }, { "epoch": 284.96, "learning_rate": 2.9321518987341775e-06, "loss": 0.0, "step": 28425 }, { "epoch": 285.21, "learning_rate": 2.9258227848101265e-06, "loss": 0.0, "step": 28450 }, { "epoch": 285.46, "learning_rate": 2.919493670886076e-06, "loss": 0.0, "step": 28475 }, { "epoch": 285.71, "learning_rate": 2.9131645569620254e-06, "loss": 0.0, "step": 28500 }, { "epoch": 285.96, "learning_rate": 2.906835443037975e-06, "loss": 0.0, "step": 28525 }, { "epoch": 286.22, "learning_rate": 2.9005063291139242e-06, "loss": 0.0, "step": 28550 }, { "epoch": 286.47, "learning_rate": 2.8941772151898732e-06, "loss": 0.0, "step": 28575 }, { "epoch": 286.72, "learning_rate": 2.887848101265823e-06, "loss": 0.0, "step": 28600 }, { "epoch": 286.97, "learning_rate": 2.8815189873417725e-06, "loss": 0.0, "step": 28625 }, { "epoch": 287.22, "learning_rate": 2.875189873417722e-06, "loss": 0.0, "step": 28650 }, { "epoch": 287.47, "learning_rate": 2.8688607594936714e-06, "loss": 0.0, "step": 28675 }, { "epoch": 287.72, "learning_rate": 2.862531645569621e-06, "loss": 0.0, "step": 28700 }, { "epoch": 287.97, "learning_rate": 2.85620253164557e-06, "loss": 0.0, "step": 28725 }, { "epoch": 288.22, "learning_rate": 2.8498734177215192e-06, "loss": 0.0, "step": 28750 }, { "epoch": 288.47, "learning_rate": 2.8435443037974687e-06, "loss": 0.0, "step": 28775 }, { "epoch": 288.72, "learning_rate": 2.837215189873418e-06, "loss": 0.0, "step": 28800 }, { "epoch": 288.97, "learning_rate": 2.830886075949367e-06, "loss": 0.0, "step": 28825 }, { "epoch": 289.22, "learning_rate": 2.8245569620253165e-06, "loss": 0.0, "step": 28850 }, { "epoch": 289.47, "learning_rate": 2.818227848101266e-06, "loss": 0.0, "step": 28875 }, { "epoch": 289.72, "learning_rate": 2.8118987341772154e-06, "loss": 0.0, "step": 28900 }, { "epoch": 289.97, "learning_rate": 2.805569620253165e-06, "loss": 0.0, "step": 28925 }, { "epoch": 290.23, "learning_rate": 2.799240506329114e-06, "loss": 0.0, "step": 28950 }, { "epoch": 290.48, "learning_rate": 2.7929113924050632e-06, "loss": 0.0, "step": 28975 }, { "epoch": 290.73, "learning_rate": 2.786582278481013e-06, "loss": 0.0, "step": 29000 }, { "epoch": 290.73, "eval_loss": 0.36975371837615967, "eval_runtime": 1411.5711, "eval_samples_per_second": 9.027, "eval_steps_per_second": 0.283, "eval_wer": 9.347154916367046, "step": 29000 }, { "epoch": 290.98, "learning_rate": 2.7802531645569625e-06, "loss": 0.0, "step": 29025 }, { "epoch": 291.23, "learning_rate": 2.773924050632912e-06, "loss": 0.0, "step": 29050 }, { "epoch": 291.48, "learning_rate": 2.767594936708861e-06, "loss": 0.0, "step": 29075 }, { "epoch": 291.73, "learning_rate": 2.7612658227848104e-06, "loss": 0.0, "step": 29100 }, { "epoch": 291.98, "learning_rate": 2.75493670886076e-06, "loss": 0.0, "step": 29125 }, { "epoch": 292.23, "learning_rate": 2.7486075949367092e-06, "loss": 0.0, "step": 29150 }, { "epoch": 292.48, "learning_rate": 2.7422784810126587e-06, "loss": 0.0, "step": 29175 }, { "epoch": 292.73, "learning_rate": 2.7359493670886077e-06, "loss": 0.0, "step": 29200 }, { "epoch": 292.98, "learning_rate": 2.729620253164557e-06, "loss": 0.0, "step": 29225 }, { "epoch": 293.23, "learning_rate": 2.7232911392405065e-06, "loss": 0.0, "step": 29250 }, { "epoch": 293.48, "learning_rate": 2.716962025316456e-06, "loss": 0.0, "step": 29275 }, { "epoch": 293.73, "learning_rate": 2.710632911392405e-06, "loss": 0.0, "step": 29300 }, { "epoch": 293.98, "learning_rate": 2.7043037974683544e-06, "loss": 0.0, "step": 29325 }, { "epoch": 294.24, "learning_rate": 2.697974683544304e-06, "loss": 0.0, "step": 29350 }, { "epoch": 294.49, "learning_rate": 2.6916455696202532e-06, "loss": 0.0, "step": 29375 }, { "epoch": 294.74, "learning_rate": 2.685316455696203e-06, "loss": 0.0, "step": 29400 }, { "epoch": 294.99, "learning_rate": 2.6789873417721525e-06, "loss": 0.0, "step": 29425 }, { "epoch": 295.24, "learning_rate": 2.6726582278481015e-06, "loss": 0.0, "step": 29450 }, { "epoch": 295.49, "learning_rate": 2.666329113924051e-06, "loss": 0.0, "step": 29475 }, { "epoch": 295.74, "learning_rate": 2.6600000000000004e-06, "loss": 0.0, "step": 29500 }, { "epoch": 295.99, "learning_rate": 2.65367088607595e-06, "loss": 0.0, "step": 29525 }, { "epoch": 296.24, "learning_rate": 2.6473417721518992e-06, "loss": 0.0, "step": 29550 }, { "epoch": 296.49, "learning_rate": 2.6410126582278482e-06, "loss": 0.0, "step": 29575 }, { "epoch": 296.74, "learning_rate": 2.6346835443037977e-06, "loss": 0.0, "step": 29600 }, { "epoch": 296.99, "learning_rate": 2.628354430379747e-06, "loss": 0.0, "step": 29625 }, { "epoch": 297.24, "learning_rate": 2.6220253164556965e-06, "loss": 0.0, "step": 29650 }, { "epoch": 297.49, "learning_rate": 2.6156962025316455e-06, "loss": 0.0, "step": 29675 }, { "epoch": 297.74, "learning_rate": 2.609367088607595e-06, "loss": 0.0, "step": 29700 }, { "epoch": 297.99, "learning_rate": 2.6030379746835444e-06, "loss": 0.0, "step": 29725 }, { "epoch": 298.25, "learning_rate": 2.596708860759494e-06, "loss": 0.0, "step": 29750 }, { "epoch": 298.5, "learning_rate": 2.5903797468354432e-06, "loss": 0.0, "step": 29775 }, { "epoch": 298.75, "learning_rate": 2.5840506329113922e-06, "loss": 0.0, "step": 29800 }, { "epoch": 299.0, "learning_rate": 2.577721518987342e-06, "loss": 0.0, "step": 29825 }, { "epoch": 299.25, "learning_rate": 2.5713924050632915e-06, "loss": 0.0, "step": 29850 }, { "epoch": 299.5, "learning_rate": 2.565063291139241e-06, "loss": 0.0, "step": 29875 }, { "epoch": 299.75, "learning_rate": 2.5587341772151904e-06, "loss": 0.0, "step": 29900 }, { "epoch": 300.0, "learning_rate": 2.5524050632911394e-06, "loss": 0.0, "step": 29925 }, { "epoch": 300.25, "learning_rate": 2.546075949367089e-06, "loss": 0.0, "step": 29950 }, { "epoch": 300.5, "learning_rate": 2.5397468354430382e-06, "loss": 0.0, "step": 29975 }, { "epoch": 300.75, "learning_rate": 2.5334177215189876e-06, "loss": 0.0, "step": 30000 }, { "epoch": 300.75, "eval_loss": 0.37864601612091064, "eval_runtime": 1411.619, "eval_samples_per_second": 9.027, "eval_steps_per_second": 0.283, "eval_wer": 9.342269813975301, "step": 30000 }, { "epoch": 301.0, "learning_rate": 2.527088607594937e-06, "loss": 0.0, "step": 30025 }, { "epoch": 301.25, "learning_rate": 2.520759493670886e-06, "loss": 0.0, "step": 30050 }, { "epoch": 301.5, "learning_rate": 2.5144303797468355e-06, "loss": 0.0, "step": 30075 }, { "epoch": 301.75, "learning_rate": 2.508101265822785e-06, "loss": 0.0, "step": 30100 }, { "epoch": 302.01, "learning_rate": 2.5017721518987344e-06, "loss": 0.0, "step": 30125 }, { "epoch": 302.26, "learning_rate": 2.495443037974684e-06, "loss": 0.0, "step": 30150 }, { "epoch": 302.51, "learning_rate": 2.4891139240506332e-06, "loss": 0.0, "step": 30175 }, { "epoch": 302.76, "learning_rate": 2.4827848101265826e-06, "loss": 0.0, "step": 30200 }, { "epoch": 303.01, "learning_rate": 2.4764556962025317e-06, "loss": 0.0, "step": 30225 }, { "epoch": 303.26, "learning_rate": 2.470126582278481e-06, "loss": 0.0, "step": 30250 }, { "epoch": 303.51, "learning_rate": 2.4637974683544305e-06, "loss": 0.0, "step": 30275 }, { "epoch": 303.76, "learning_rate": 2.45746835443038e-06, "loss": 0.0, "step": 30300 }, { "epoch": 304.01, "learning_rate": 2.4511392405063294e-06, "loss": 0.0, "step": 30325 }, { "epoch": 304.26, "learning_rate": 2.444810126582279e-06, "loss": 0.0, "step": 30350 }, { "epoch": 304.51, "learning_rate": 2.4384810126582282e-06, "loss": 0.0, "step": 30375 }, { "epoch": 304.76, "learning_rate": 2.4321518987341776e-06, "loss": 0.0, "step": 30400 }, { "epoch": 305.01, "learning_rate": 2.4258227848101266e-06, "loss": 0.0, "step": 30425 }, { "epoch": 305.26, "learning_rate": 2.419493670886076e-06, "loss": 0.0, "step": 30450 }, { "epoch": 305.51, "learning_rate": 2.4131645569620255e-06, "loss": 0.0, "step": 30475 }, { "epoch": 305.76, "learning_rate": 2.406835443037975e-06, "loss": 0.0, "step": 30500 }, { "epoch": 306.02, "learning_rate": 2.400506329113924e-06, "loss": 0.0, "step": 30525 }, { "epoch": 306.27, "learning_rate": 2.394177215189874e-06, "loss": 0.0, "step": 30550 }, { "epoch": 306.52, "learning_rate": 2.3878481012658232e-06, "loss": 0.0, "step": 30575 }, { "epoch": 306.77, "learning_rate": 2.3815189873417722e-06, "loss": 0.0, "step": 30600 }, { "epoch": 307.02, "learning_rate": 2.3751898734177216e-06, "loss": 0.0, "step": 30625 }, { "epoch": 307.27, "learning_rate": 2.368860759493671e-06, "loss": 0.0, "step": 30650 }, { "epoch": 307.52, "learning_rate": 2.3625316455696205e-06, "loss": 0.0, "step": 30675 }, { "epoch": 307.77, "learning_rate": 2.35620253164557e-06, "loss": 0.0, "step": 30700 }, { "epoch": 308.02, "learning_rate": 2.349873417721519e-06, "loss": 0.0, "step": 30725 }, { "epoch": 308.27, "learning_rate": 2.3435443037974688e-06, "loss": 0.0, "step": 30750 }, { "epoch": 308.52, "learning_rate": 2.337215189873418e-06, "loss": 0.0, "step": 30775 }, { "epoch": 308.77, "learning_rate": 2.3308860759493672e-06, "loss": 0.0, "step": 30800 }, { "epoch": 309.02, "learning_rate": 2.3248101265822787e-06, "loss": 0.0, "step": 30825 }, { "epoch": 309.27, "learning_rate": 2.318481012658228e-06, "loss": 0.0, "step": 30850 }, { "epoch": 309.52, "learning_rate": 2.3121518987341775e-06, "loss": 0.0, "step": 30875 }, { "epoch": 309.77, "learning_rate": 2.3058227848101265e-06, "loss": 0.0, "step": 30900 }, { "epoch": 310.03, "learning_rate": 2.299493670886076e-06, "loss": 0.0, "step": 30925 }, { "epoch": 310.28, "learning_rate": 2.293164556962026e-06, "loss": 0.0, "step": 30950 }, { "epoch": 310.53, "learning_rate": 2.286835443037975e-06, "loss": 0.0, "step": 30975 }, { "epoch": 310.78, "learning_rate": 2.2805063291139242e-06, "loss": 0.0, "step": 31000 }, { "epoch": 310.78, "eval_loss": 0.38683274388313293, "eval_runtime": 1411.8913, "eval_samples_per_second": 9.025, "eval_steps_per_second": 0.283, "eval_wer": 9.316867281538222, "step": 31000 }, { "epoch": 311.03, "learning_rate": 2.2741772151898737e-06, "loss": 0.0, "step": 31025 }, { "epoch": 311.28, "learning_rate": 2.267848101265823e-06, "loss": 0.0, "step": 31050 }, { "epoch": 311.53, "learning_rate": 2.261518987341772e-06, "loss": 0.0, "step": 31075 }, { "epoch": 311.78, "learning_rate": 2.2551898734177215e-06, "loss": 0.0, "step": 31100 }, { "epoch": 312.03, "learning_rate": 2.248860759493671e-06, "loss": 0.0, "step": 31125 }, { "epoch": 312.28, "learning_rate": 2.2425316455696204e-06, "loss": 0.0, "step": 31150 }, { "epoch": 312.53, "learning_rate": 2.23620253164557e-06, "loss": 0.0, "step": 31175 }, { "epoch": 312.78, "learning_rate": 2.2298734177215192e-06, "loss": 0.0, "step": 31200 }, { "epoch": 313.03, "learning_rate": 2.2235443037974687e-06, "loss": 0.0, "step": 31225 }, { "epoch": 313.28, "learning_rate": 2.217215189873418e-06, "loss": 0.0, "step": 31250 }, { "epoch": 313.53, "learning_rate": 2.210886075949367e-06, "loss": 0.0, "step": 31275 }, { "epoch": 313.78, "learning_rate": 2.2045569620253165e-06, "loss": 0.0, "step": 31300 }, { "epoch": 314.04, "learning_rate": 2.198227848101266e-06, "loss": 0.0, "step": 31325 }, { "epoch": 314.29, "learning_rate": 2.1918987341772154e-06, "loss": 0.0, "step": 31350 }, { "epoch": 314.54, "learning_rate": 2.185569620253165e-06, "loss": 0.0, "step": 31375 }, { "epoch": 314.79, "learning_rate": 2.1792405063291142e-06, "loss": 0.0, "step": 31400 }, { "epoch": 315.04, "learning_rate": 2.1729113924050637e-06, "loss": 0.0, "step": 31425 }, { "epoch": 315.29, "learning_rate": 2.1665822784810127e-06, "loss": 0.0, "step": 31450 }, { "epoch": 315.54, "learning_rate": 2.160253164556962e-06, "loss": 0.0, "step": 31475 }, { "epoch": 315.79, "learning_rate": 2.1539240506329115e-06, "loss": 0.0, "step": 31500 }, { "epoch": 316.04, "learning_rate": 2.147594936708861e-06, "loss": 0.0, "step": 31525 }, { "epoch": 316.29, "learning_rate": 2.1412658227848104e-06, "loss": 0.0, "step": 31550 }, { "epoch": 316.54, "learning_rate": 2.13493670886076e-06, "loss": 0.0, "step": 31575 }, { "epoch": 316.79, "learning_rate": 2.1286075949367092e-06, "loss": 0.0, "step": 31600 }, { "epoch": 317.04, "learning_rate": 2.1222784810126582e-06, "loss": 0.0, "step": 31625 }, { "epoch": 317.29, "learning_rate": 2.1159493670886077e-06, "loss": 0.0, "step": 31650 }, { "epoch": 317.54, "learning_rate": 2.109620253164557e-06, "loss": 0.0, "step": 31675 }, { "epoch": 317.79, "learning_rate": 2.1032911392405065e-06, "loss": 0.0, "step": 31700 }, { "epoch": 318.05, "learning_rate": 2.096962025316456e-06, "loss": 0.0, "step": 31725 }, { "epoch": 318.3, "learning_rate": 2.090632911392405e-06, "loss": 0.0, "step": 31750 }, { "epoch": 318.55, "learning_rate": 2.084303797468355e-06, "loss": 0.0, "step": 31775 }, { "epoch": 318.8, "learning_rate": 2.0779746835443042e-06, "loss": 0.0, "step": 31800 }, { "epoch": 319.05, "learning_rate": 2.0716455696202532e-06, "loss": 0.0, "step": 31825 }, { "epoch": 319.3, "learning_rate": 2.0653164556962027e-06, "loss": 0.0, "step": 31850 }, { "epoch": 319.55, "learning_rate": 2.058987341772152e-06, "loss": 0.0, "step": 31875 }, { "epoch": 319.8, "learning_rate": 2.0526582278481015e-06, "loss": 0.0, "step": 31900 }, { "epoch": 320.05, "learning_rate": 2.0463291139240505e-06, "loss": 0.0, "step": 31925 }, { "epoch": 320.3, "learning_rate": 2.04e-06, "loss": 0.0, "step": 31950 }, { "epoch": 320.55, "learning_rate": 2.03367088607595e-06, "loss": 0.0, "step": 31975 }, { "epoch": 320.8, "learning_rate": 2.027341772151899e-06, "loss": 0.0, "step": 32000 }, { "epoch": 320.8, "eval_loss": 0.394780695438385, "eval_runtime": 1412.1761, "eval_samples_per_second": 9.023, "eval_steps_per_second": 0.283, "eval_wer": 9.32859152727841, "step": 32000 }, { "epoch": 321.05, "learning_rate": 2.0210126582278482e-06, "loss": 0.0, "step": 32025 }, { "epoch": 321.3, "learning_rate": 2.0146835443037977e-06, "loss": 0.0, "step": 32050 }, { "epoch": 321.55, "learning_rate": 2.008354430379747e-06, "loss": 0.0, "step": 32075 }, { "epoch": 321.8, "learning_rate": 2.0020253164556965e-06, "loss": 0.0, "step": 32100 }, { "epoch": 322.06, "learning_rate": 1.9956962025316455e-06, "loss": 0.0, "step": 32125 }, { "epoch": 322.31, "learning_rate": 1.989367088607595e-06, "loss": 0.0, "step": 32150 }, { "epoch": 322.56, "learning_rate": 1.9830379746835444e-06, "loss": 0.0, "step": 32175 }, { "epoch": 322.81, "learning_rate": 1.976708860759494e-06, "loss": 0.0, "step": 32200 }, { "epoch": 323.06, "learning_rate": 1.9703797468354432e-06, "loss": 0.0, "step": 32225 }, { "epoch": 323.31, "learning_rate": 1.9640506329113927e-06, "loss": 0.0, "step": 32250 }, { "epoch": 323.56, "learning_rate": 1.957721518987342e-06, "loss": 0.0, "step": 32275 }, { "epoch": 323.81, "learning_rate": 1.951392405063291e-06, "loss": 0.0, "step": 32300 }, { "epoch": 324.06, "learning_rate": 1.9450632911392405e-06, "loss": 0.0, "step": 32325 }, { "epoch": 324.31, "learning_rate": 1.93873417721519e-06, "loss": 0.0, "step": 32350 }, { "epoch": 324.56, "learning_rate": 1.9324050632911394e-06, "loss": 0.0, "step": 32375 }, { "epoch": 324.81, "learning_rate": 1.926075949367089e-06, "loss": 0.0, "step": 32400 }, { "epoch": 325.06, "learning_rate": 1.9197468354430382e-06, "loss": 0.0, "step": 32425 }, { "epoch": 325.31, "learning_rate": 1.9134177215189877e-06, "loss": 0.0, "step": 32450 }, { "epoch": 325.56, "learning_rate": 1.9070886075949369e-06, "loss": 0.0, "step": 32475 }, { "epoch": 325.81, "learning_rate": 1.900759493670886e-06, "loss": 0.0, "step": 32500 }, { "epoch": 326.07, "learning_rate": 1.8944303797468355e-06, "loss": 0.0, "step": 32525 }, { "epoch": 326.32, "learning_rate": 1.888101265822785e-06, "loss": 0.0, "step": 32550 }, { "epoch": 326.57, "learning_rate": 1.8817721518987344e-06, "loss": 0.0, "step": 32575 }, { "epoch": 326.82, "learning_rate": 1.8754430379746838e-06, "loss": 0.0, "step": 32600 }, { "epoch": 327.07, "learning_rate": 1.8691139240506332e-06, "loss": 0.0, "step": 32625 }, { "epoch": 327.32, "learning_rate": 1.8627848101265824e-06, "loss": 0.0, "step": 32650 }, { "epoch": 327.57, "learning_rate": 1.8564556962025319e-06, "loss": 0.0, "step": 32675 }, { "epoch": 327.82, "learning_rate": 1.850126582278481e-06, "loss": 0.0, "step": 32700 }, { "epoch": 328.07, "learning_rate": 1.8437974683544305e-06, "loss": 0.0, "step": 32725 }, { "epoch": 328.32, "learning_rate": 1.8374683544303797e-06, "loss": 0.0, "step": 32750 }, { "epoch": 328.57, "learning_rate": 1.8311392405063292e-06, "loss": 0.0, "step": 32775 }, { "epoch": 328.82, "learning_rate": 1.8248101265822788e-06, "loss": 0.0, "step": 32800 }, { "epoch": 329.07, "learning_rate": 1.818481012658228e-06, "loss": 0.0, "step": 32825 }, { "epoch": 329.32, "learning_rate": 1.8121518987341774e-06, "loss": 0.0, "step": 32850 }, { "epoch": 329.57, "learning_rate": 1.8058227848101267e-06, "loss": 0.0, "step": 32875 }, { "epoch": 329.82, "learning_rate": 1.799493670886076e-06, "loss": 0.0, "step": 32900 }, { "epoch": 330.08, "learning_rate": 1.7931645569620253e-06, "loss": 0.0, "step": 32925 }, { "epoch": 330.33, "learning_rate": 1.7868354430379747e-06, "loss": 0.0, "step": 32950 }, { "epoch": 330.58, "learning_rate": 1.7805063291139242e-06, "loss": 0.0, "step": 32975 }, { "epoch": 330.83, "learning_rate": 1.7741772151898736e-06, "loss": 0.0, "step": 33000 }, { "epoch": 330.83, "eval_loss": 0.40184545516967773, "eval_runtime": 1411.5048, "eval_samples_per_second": 9.027, "eval_steps_per_second": 0.283, "eval_wer": 9.333476629670159, "step": 33000 }, { "epoch": 331.08, "learning_rate": 1.767848101265823e-06, "loss": 0.0, "step": 33025 }, { "epoch": 331.33, "learning_rate": 1.7615189873417724e-06, "loss": 0.0, "step": 33050 }, { "epoch": 331.58, "learning_rate": 1.7551898734177217e-06, "loss": 0.0, "step": 33075 }, { "epoch": 331.83, "learning_rate": 1.748860759493671e-06, "loss": 0.0, "step": 33100 }, { "epoch": 332.08, "learning_rate": 1.7425316455696203e-06, "loss": 0.0, "step": 33125 }, { "epoch": 332.33, "learning_rate": 1.7362025316455697e-06, "loss": 0.0, "step": 33150 }, { "epoch": 332.58, "learning_rate": 1.729873417721519e-06, "loss": 0.0, "step": 33175 }, { "epoch": 332.83, "learning_rate": 1.7235443037974686e-06, "loss": 0.0, "step": 33200 }, { "epoch": 333.08, "learning_rate": 1.717215189873418e-06, "loss": 0.0, "step": 33225 }, { "epoch": 333.33, "learning_rate": 1.7108860759493672e-06, "loss": 0.0, "step": 33250 }, { "epoch": 333.58, "learning_rate": 1.7045569620253167e-06, "loss": 0.0, "step": 33275 }, { "epoch": 333.83, "learning_rate": 1.6982278481012659e-06, "loss": 0.0, "step": 33300 }, { "epoch": 334.09, "learning_rate": 1.6918987341772153e-06, "loss": 0.0, "step": 33325 }, { "epoch": 334.34, "learning_rate": 1.6855696202531645e-06, "loss": 0.0, "step": 33350 }, { "epoch": 334.59, "learning_rate": 1.679240506329114e-06, "loss": 0.0, "step": 33375 }, { "epoch": 334.84, "learning_rate": 1.6729113924050636e-06, "loss": 0.0, "step": 33400 }, { "epoch": 335.09, "learning_rate": 1.6665822784810128e-06, "loss": 0.0, "step": 33425 }, { "epoch": 335.34, "learning_rate": 1.6602531645569622e-06, "loss": 0.0, "step": 33450 }, { "epoch": 335.59, "learning_rate": 1.6539240506329117e-06, "loss": 0.0, "step": 33475 }, { "epoch": 335.84, "learning_rate": 1.6475949367088609e-06, "loss": 0.0, "step": 33500 }, { "epoch": 336.09, "learning_rate": 1.6412658227848103e-06, "loss": 0.0, "step": 33525 }, { "epoch": 336.34, "learning_rate": 1.6349367088607595e-06, "loss": 0.0, "step": 33550 }, { "epoch": 336.59, "learning_rate": 1.628607594936709e-06, "loss": 0.0, "step": 33575 }, { "epoch": 336.84, "learning_rate": 1.6222784810126582e-06, "loss": 0.0, "step": 33600 }, { "epoch": 337.09, "learning_rate": 1.6159493670886078e-06, "loss": 0.0, "step": 33625 }, { "epoch": 337.34, "learning_rate": 1.6096202531645572e-06, "loss": 0.0, "step": 33650 }, { "epoch": 337.59, "learning_rate": 1.6032911392405064e-06, "loss": 0.0, "step": 33675 }, { "epoch": 337.84, "learning_rate": 1.5969620253164559e-06, "loss": 0.0, "step": 33700 }, { "epoch": 338.1, "learning_rate": 1.590632911392405e-06, "loss": 0.0, "step": 33725 }, { "epoch": 338.35, "learning_rate": 1.5843037974683545e-06, "loss": 0.0, "step": 33750 }, { "epoch": 338.6, "learning_rate": 1.577974683544304e-06, "loss": 0.0, "step": 33775 }, { "epoch": 338.85, "learning_rate": 1.5716455696202532e-06, "loss": 0.0, "step": 33800 }, { "epoch": 339.1, "learning_rate": 1.5653164556962028e-06, "loss": 0.0, "step": 33825 }, { "epoch": 339.35, "learning_rate": 1.558987341772152e-06, "loss": 0.0, "step": 33850 }, { "epoch": 339.6, "learning_rate": 1.5526582278481014e-06, "loss": 0.0, "step": 33875 }, { "epoch": 339.85, "learning_rate": 1.5463291139240509e-06, "loss": 0.0, "step": 33900 }, { "epoch": 340.1, "learning_rate": 1.54e-06, "loss": 0.0, "step": 33925 }, { "epoch": 340.35, "learning_rate": 1.5336708860759495e-06, "loss": 0.0, "step": 33950 }, { "epoch": 340.6, "learning_rate": 1.5273417721518987e-06, "loss": 0.0, "step": 33975 }, { "epoch": 340.85, "learning_rate": 1.5210126582278481e-06, "loss": 0.0, "step": 34000 }, { "epoch": 340.85, "eval_loss": 0.4080604016780853, "eval_runtime": 1414.6983, "eval_samples_per_second": 9.007, "eval_steps_per_second": 0.282, "eval_wer": 9.32859152727841, "step": 34000 }, { "epoch": 341.1, "learning_rate": 1.5146835443037978e-06, "loss": 0.0, "step": 34025 }, { "epoch": 341.35, "learning_rate": 1.508354430379747e-06, "loss": 0.0, "step": 34050 }, { "epoch": 341.6, "learning_rate": 1.5020253164556964e-06, "loss": 0.0, "step": 34075 }, { "epoch": 341.85, "learning_rate": 1.4956962025316456e-06, "loss": 0.0, "step": 34100 }, { "epoch": 342.11, "learning_rate": 1.489367088607595e-06, "loss": 0.0, "step": 34125 }, { "epoch": 342.36, "learning_rate": 1.4830379746835443e-06, "loss": 0.0, "step": 34150 }, { "epoch": 342.61, "learning_rate": 1.4767088607594937e-06, "loss": 0.0, "step": 34175 }, { "epoch": 342.86, "learning_rate": 1.4703797468354431e-06, "loss": 0.0, "step": 34200 }, { "epoch": 343.11, "learning_rate": 1.4640506329113926e-06, "loss": 0.0, "step": 34225 }, { "epoch": 343.36, "learning_rate": 1.457721518987342e-06, "loss": 0.0, "step": 34250 }, { "epoch": 343.61, "learning_rate": 1.4513924050632912e-06, "loss": 0.0, "step": 34275 }, { "epoch": 343.86, "learning_rate": 1.4450632911392406e-06, "loss": 0.0, "step": 34300 }, { "epoch": 344.11, "learning_rate": 1.43873417721519e-06, "loss": 0.0, "step": 34325 }, { "epoch": 344.36, "learning_rate": 1.4324050632911393e-06, "loss": 0.0, "step": 34350 }, { "epoch": 344.61, "learning_rate": 1.4260759493670887e-06, "loss": 0.0, "step": 34375 }, { "epoch": 344.86, "learning_rate": 1.419746835443038e-06, "loss": 0.0, "step": 34400 }, { "epoch": 345.11, "learning_rate": 1.4134177215189876e-06, "loss": 0.0, "step": 34425 }, { "epoch": 345.36, "learning_rate": 1.407088607594937e-06, "loss": 0.0, "step": 34450 }, { "epoch": 345.61, "learning_rate": 1.4007594936708862e-06, "loss": 0.0, "step": 34475 }, { "epoch": 345.86, "learning_rate": 1.3944303797468356e-06, "loss": 0.0, "step": 34500 }, { "epoch": 346.12, "learning_rate": 1.3881012658227849e-06, "loss": 0.0, "step": 34525 }, { "epoch": 346.37, "learning_rate": 1.3817721518987343e-06, "loss": 0.0, "step": 34550 }, { "epoch": 346.62, "learning_rate": 1.3754430379746835e-06, "loss": 0.0, "step": 34575 }, { "epoch": 346.87, "learning_rate": 1.369113924050633e-06, "loss": 0.0, "step": 34600 }, { "epoch": 347.12, "learning_rate": 1.3627848101265824e-06, "loss": 0.0, "step": 34625 }, { "epoch": 347.37, "learning_rate": 1.3564556962025318e-06, "loss": 0.0, "step": 34650 }, { "epoch": 347.62, "learning_rate": 1.3501265822784812e-06, "loss": 0.0, "step": 34675 }, { "epoch": 347.87, "learning_rate": 1.3437974683544304e-06, "loss": 0.0, "step": 34700 }, { "epoch": 348.12, "learning_rate": 1.3374683544303799e-06, "loss": 0.0, "step": 34725 }, { "epoch": 348.37, "learning_rate": 1.3311392405063293e-06, "loss": 0.0, "step": 34750 }, { "epoch": 348.62, "learning_rate": 1.3248101265822785e-06, "loss": 0.0, "step": 34775 }, { "epoch": 348.87, "learning_rate": 1.318481012658228e-06, "loss": 0.0, "step": 34800 }, { "epoch": 349.12, "learning_rate": 1.3121518987341771e-06, "loss": 0.0, "step": 34825 }, { "epoch": 349.37, "learning_rate": 1.3060759493670888e-06, "loss": 0.0, "step": 34850 }, { "epoch": 349.62, "learning_rate": 1.2997468354430382e-06, "loss": 0.0, "step": 34875 }, { "epoch": 349.87, "learning_rate": 1.2934177215189875e-06, "loss": 0.0, "step": 34900 }, { "epoch": 350.13, "learning_rate": 1.2870886075949369e-06, "loss": 0.0, "step": 34925 }, { "epoch": 350.38, "learning_rate": 1.280759493670886e-06, "loss": 0.0, "step": 34950 }, { "epoch": 350.63, "learning_rate": 1.2744303797468355e-06, "loss": 0.0, "step": 34975 }, { "epoch": 350.88, "learning_rate": 1.2681012658227847e-06, "loss": 0.0, "step": 35000 }, { "epoch": 350.88, "eval_loss": 0.41380730271339417, "eval_runtime": 1417.2551, "eval_samples_per_second": 8.991, "eval_steps_per_second": 0.282, "eval_wer": 9.336407691105206, "step": 35000 }, { "epoch": 351.13, "learning_rate": 1.2617721518987342e-06, "loss": 0.0, "step": 35025 }, { "epoch": 351.38, "learning_rate": 1.2554430379746838e-06, "loss": 0.0, "step": 35050 }, { "epoch": 351.63, "learning_rate": 1.249113924050633e-06, "loss": 0.0, "step": 35075 }, { "epoch": 351.88, "learning_rate": 1.2427848101265824e-06, "loss": 0.0, "step": 35100 }, { "epoch": 352.13, "learning_rate": 1.2364556962025317e-06, "loss": 0.0, "step": 35125 }, { "epoch": 352.38, "learning_rate": 1.230126582278481e-06, "loss": 0.0, "step": 35150 }, { "epoch": 352.63, "learning_rate": 1.2237974683544305e-06, "loss": 0.0, "step": 35175 }, { "epoch": 352.88, "learning_rate": 1.21746835443038e-06, "loss": 0.0, "step": 35200 }, { "epoch": 353.13, "learning_rate": 1.2111392405063292e-06, "loss": 0.0, "step": 35225 }, { "epoch": 353.38, "learning_rate": 1.2048101265822786e-06, "loss": 0.0, "step": 35250 }, { "epoch": 353.63, "learning_rate": 1.1984810126582278e-06, "loss": 0.0, "step": 35275 }, { "epoch": 353.88, "learning_rate": 1.1921518987341774e-06, "loss": 0.0, "step": 35300 }, { "epoch": 354.14, "learning_rate": 1.1858227848101267e-06, "loss": 0.0, "step": 35325 }, { "epoch": 354.39, "learning_rate": 1.179493670886076e-06, "loss": 0.0, "step": 35350 }, { "epoch": 354.64, "learning_rate": 1.1731645569620253e-06, "loss": 0.0, "step": 35375 }, { "epoch": 354.89, "learning_rate": 1.1668354430379747e-06, "loss": 0.0, "step": 35400 }, { "epoch": 355.14, "learning_rate": 1.1605063291139242e-06, "loss": 0.0, "step": 35425 }, { "epoch": 355.39, "learning_rate": 1.1541772151898736e-06, "loss": 0.0, "step": 35450 }, { "epoch": 355.64, "learning_rate": 1.1478481012658228e-06, "loss": 0.0, "step": 35475 }, { "epoch": 355.89, "learning_rate": 1.1415189873417722e-06, "loss": 0.0, "step": 35500 }, { "epoch": 356.14, "learning_rate": 1.1351898734177217e-06, "loss": 0.0, "step": 35525 }, { "epoch": 356.39, "learning_rate": 1.1288607594936709e-06, "loss": 0.0, "step": 35550 }, { "epoch": 356.64, "learning_rate": 1.1225316455696203e-06, "loss": 0.0, "step": 35575 }, { "epoch": 356.89, "learning_rate": 1.1162025316455697e-06, "loss": 0.0, "step": 35600 }, { "epoch": 357.14, "learning_rate": 1.1098734177215192e-06, "loss": 0.0, "step": 35625 }, { "epoch": 357.39, "learning_rate": 1.1035443037974684e-06, "loss": 0.0, "step": 35650 }, { "epoch": 357.64, "learning_rate": 1.0972151898734178e-06, "loss": 0.0, "step": 35675 }, { "epoch": 357.89, "learning_rate": 1.0908860759493672e-06, "loss": 0.0, "step": 35700 }, { "epoch": 358.15, "learning_rate": 1.0845569620253167e-06, "loss": 0.0, "step": 35725 }, { "epoch": 358.4, "learning_rate": 1.0782278481012659e-06, "loss": 0.0, "step": 35750 }, { "epoch": 358.65, "learning_rate": 1.0718987341772153e-06, "loss": 0.0, "step": 35775 }, { "epoch": 358.9, "learning_rate": 1.0655696202531647e-06, "loss": 0.0, "step": 35800 }, { "epoch": 359.15, "learning_rate": 1.059240506329114e-06, "loss": 0.0, "step": 35825 }, { "epoch": 359.4, "learning_rate": 1.0529113924050634e-06, "loss": 0.0, "step": 35850 }, { "epoch": 359.65, "learning_rate": 1.0465822784810128e-06, "loss": 0.0, "step": 35875 }, { "epoch": 359.9, "learning_rate": 1.040253164556962e-06, "loss": 0.0, "step": 35900 }, { "epoch": 360.15, "learning_rate": 1.0339240506329114e-06, "loss": 0.0, "step": 35925 }, { "epoch": 360.4, "learning_rate": 1.0275949367088609e-06, "loss": 0.0, "step": 35950 }, { "epoch": 360.65, "learning_rate": 1.02126582278481e-06, "loss": 0.0, "step": 35975 }, { "epoch": 360.9, "learning_rate": 1.0149367088607595e-06, "loss": 0.0, "step": 36000 }, { "epoch": 360.9, "eval_loss": 0.4191250205039978, "eval_runtime": 1415.5907, "eval_samples_per_second": 9.001, "eval_steps_per_second": 0.282, "eval_wer": 9.34324683445365, "step": 36000 }, { "epoch": 361.15, "learning_rate": 1.008607594936709e-06, "loss": 0.0, "step": 36025 }, { "epoch": 361.4, "learning_rate": 1.0022784810126584e-06, "loss": 0.0, "step": 36050 }, { "epoch": 361.65, "learning_rate": 9.959493670886076e-07, "loss": 0.0, "step": 36075 }, { "epoch": 361.9, "learning_rate": 9.89620253164557e-07, "loss": 0.0, "step": 36100 }, { "epoch": 362.16, "learning_rate": 9.832911392405064e-07, "loss": 0.0, "step": 36125 }, { "epoch": 362.41, "learning_rate": 9.769620253164559e-07, "loss": 0.0, "step": 36150 }, { "epoch": 362.66, "learning_rate": 9.70632911392405e-07, "loss": 0.0, "step": 36175 }, { "epoch": 362.91, "learning_rate": 9.643037974683545e-07, "loss": 0.0, "step": 36200 }, { "epoch": 363.16, "learning_rate": 9.57974683544304e-07, "loss": 0.0, "step": 36225 }, { "epoch": 363.41, "learning_rate": 9.516455696202533e-07, "loss": 0.0, "step": 36250 }, { "epoch": 363.66, "learning_rate": 9.453164556962026e-07, "loss": 0.0, "step": 36275 }, { "epoch": 363.91, "learning_rate": 9.389873417721519e-07, "loss": 0.0, "step": 36300 }, { "epoch": 364.16, "learning_rate": 9.326582278481014e-07, "loss": 0.0, "step": 36325 }, { "epoch": 364.41, "learning_rate": 9.263291139240508e-07, "loss": 0.0, "step": 36350 }, { "epoch": 364.66, "learning_rate": 9.200000000000001e-07, "loss": 0.0, "step": 36375 }, { "epoch": 364.91, "learning_rate": 9.136708860759494e-07, "loss": 0.0, "step": 36400 }, { "epoch": 365.16, "learning_rate": 9.073417721518988e-07, "loss": 0.0, "step": 36425 }, { "epoch": 365.41, "learning_rate": 9.010126582278482e-07, "loss": 0.0, "step": 36450 }, { "epoch": 365.66, "learning_rate": 8.946835443037975e-07, "loss": 0.0, "step": 36475 }, { "epoch": 365.91, "learning_rate": 8.883544303797469e-07, "loss": 0.0, "step": 36500 }, { "epoch": 366.17, "learning_rate": 8.820253164556963e-07, "loss": 0.0, "step": 36525 }, { "epoch": 366.42, "learning_rate": 8.756962025316457e-07, "loss": 0.0, "step": 36550 }, { "epoch": 366.67, "learning_rate": 8.69367088607595e-07, "loss": 0.0, "step": 36575 }, { "epoch": 366.92, "learning_rate": 8.630379746835443e-07, "loss": 0.0, "step": 36600 }, { "epoch": 367.17, "learning_rate": 8.567088607594938e-07, "loss": 0.0, "step": 36625 }, { "epoch": 367.42, "learning_rate": 8.503797468354432e-07, "loss": 0.0, "step": 36650 }, { "epoch": 367.67, "learning_rate": 8.440506329113925e-07, "loss": 0.0, "step": 36675 }, { "epoch": 367.92, "learning_rate": 8.377215189873418e-07, "loss": 0.0, "step": 36700 }, { "epoch": 368.17, "learning_rate": 8.313924050632912e-07, "loss": 0.0, "step": 36725 }, { "epoch": 368.42, "learning_rate": 8.250632911392407e-07, "loss": 0.0, "step": 36750 }, { "epoch": 368.67, "learning_rate": 8.1873417721519e-07, "loss": 0.0, "step": 36775 }, { "epoch": 368.92, "learning_rate": 8.124050632911393e-07, "loss": 0.0, "step": 36800 }, { "epoch": 369.17, "learning_rate": 8.060759493670886e-07, "loss": 0.0, "step": 36825 }, { "epoch": 369.42, "learning_rate": 7.99746835443038e-07, "loss": 0.0, "step": 36850 }, { "epoch": 369.67, "learning_rate": 7.934177215189874e-07, "loss": 0.0, "step": 36875 }, { "epoch": 369.92, "learning_rate": 7.870886075949367e-07, "loss": 0.0, "step": 36900 }, { "epoch": 370.18, "learning_rate": 7.807594936708861e-07, "loss": 0.0, "step": 36925 }, { "epoch": 370.43, "learning_rate": 7.744303797468355e-07, "loss": 0.0, "step": 36950 }, { "epoch": 370.68, "learning_rate": 7.681012658227849e-07, "loss": 0.0, "step": 36975 }, { "epoch": 370.93, "learning_rate": 7.617721518987342e-07, "loss": 0.0, "step": 37000 }, { "epoch": 370.93, "eval_loss": 0.4233819842338562, "eval_runtime": 1416.2908, "eval_samples_per_second": 8.997, "eval_steps_per_second": 0.282, "eval_wer": 9.331522588713458, "step": 37000 }, { "epoch": 371.18, "learning_rate": 7.554430379746835e-07, "loss": 0.0, "step": 37025 }, { "epoch": 371.43, "learning_rate": 7.49113924050633e-07, "loss": 0.0, "step": 37050 }, { "epoch": 371.68, "learning_rate": 7.430379746835444e-07, "loss": 0.0, "step": 37075 }, { "epoch": 371.93, "learning_rate": 7.367088607594937e-07, "loss": 0.0, "step": 37100 }, { "epoch": 372.18, "learning_rate": 7.303797468354431e-07, "loss": 0.0, "step": 37125 }, { "epoch": 372.43, "learning_rate": 7.240506329113925e-07, "loss": 0.0, "step": 37150 }, { "epoch": 372.68, "learning_rate": 7.177215189873418e-07, "loss": 0.0, "step": 37175 }, { "epoch": 372.93, "learning_rate": 7.113924050632912e-07, "loss": 0.0, "step": 37200 }, { "epoch": 373.18, "learning_rate": 7.050632911392406e-07, "loss": 0.0, "step": 37225 }, { "epoch": 373.43, "learning_rate": 6.9873417721519e-07, "loss": 0.0, "step": 37250 }, { "epoch": 373.68, "learning_rate": 6.924050632911393e-07, "loss": 0.0, "step": 37275 }, { "epoch": 373.93, "learning_rate": 6.860759493670886e-07, "loss": 0.0, "step": 37300 }, { "epoch": 374.19, "learning_rate": 6.797468354430379e-07, "loss": 0.0, "step": 37325 }, { "epoch": 374.44, "learning_rate": 6.734177215189875e-07, "loss": 0.0, "step": 37350 }, { "epoch": 374.69, "learning_rate": 6.670886075949368e-07, "loss": 0.0, "step": 37375 }, { "epoch": 374.94, "learning_rate": 6.607594936708861e-07, "loss": 0.0, "step": 37400 }, { "epoch": 375.19, "learning_rate": 6.544303797468354e-07, "loss": 0.0, "step": 37425 }, { "epoch": 375.44, "learning_rate": 6.481012658227849e-07, "loss": 0.0, "step": 37450 }, { "epoch": 375.69, "learning_rate": 6.417721518987343e-07, "loss": 0.0, "step": 37475 }, { "epoch": 375.94, "learning_rate": 6.354430379746836e-07, "loss": 0.0, "step": 37500 }, { "epoch": 376.19, "learning_rate": 6.291139240506329e-07, "loss": 0.0, "step": 37525 }, { "epoch": 376.44, "learning_rate": 6.227848101265824e-07, "loss": 0.0, "step": 37550 }, { "epoch": 376.69, "learning_rate": 6.164556962025317e-07, "loss": 0.0, "step": 37575 }, { "epoch": 376.94, "learning_rate": 6.10126582278481e-07, "loss": 0.0, "step": 37600 }, { "epoch": 377.19, "learning_rate": 6.037974683544304e-07, "loss": 0.0, "step": 37625 }, { "epoch": 377.44, "learning_rate": 5.974683544303797e-07, "loss": 0.0, "step": 37650 }, { "epoch": 377.69, "learning_rate": 5.911392405063292e-07, "loss": 0.0, "step": 37675 }, { "epoch": 377.94, "learning_rate": 5.848101265822785e-07, "loss": 0.0, "step": 37700 }, { "epoch": 378.2, "learning_rate": 5.784810126582279e-07, "loss": 0.0, "step": 37725 }, { "epoch": 378.45, "learning_rate": 5.721518987341772e-07, "loss": 0.0, "step": 37750 }, { "epoch": 378.7, "learning_rate": 5.658227848101267e-07, "loss": 0.0, "step": 37775 }, { "epoch": 378.95, "learning_rate": 5.59493670886076e-07, "loss": 0.0, "step": 37800 }, { "epoch": 379.2, "learning_rate": 5.531645569620254e-07, "loss": 0.0, "step": 37825 }, { "epoch": 379.45, "learning_rate": 5.468354430379747e-07, "loss": 0.0, "step": 37850 }, { "epoch": 379.7, "learning_rate": 5.405063291139241e-07, "loss": 0.0, "step": 37875 }, { "epoch": 379.95, "learning_rate": 5.341772151898735e-07, "loss": 0.0, "step": 37900 }, { "epoch": 380.2, "learning_rate": 5.278481012658228e-07, "loss": 0.0, "step": 37925 }, { "epoch": 380.45, "learning_rate": 5.215189873417721e-07, "loss": 0.0, "step": 37950 }, { "epoch": 380.7, "learning_rate": 5.151898734177216e-07, "loss": 0.0, "step": 37975 }, { "epoch": 380.95, "learning_rate": 5.088607594936709e-07, "loss": 0.0, "step": 38000 }, { "epoch": 380.95, "eval_loss": 0.4270497262477875, "eval_runtime": 1414.6397, "eval_samples_per_second": 9.007, "eval_steps_per_second": 0.282, "eval_wer": 9.340315773018602, "step": 38000 }, { "epoch": 381.2, "learning_rate": 5.025316455696203e-07, "loss": 0.0, "step": 38025 }, { "epoch": 381.45, "learning_rate": 4.962025316455696e-07, "loss": 0.0, "step": 38050 }, { "epoch": 381.7, "learning_rate": 4.898734177215191e-07, "loss": 0.0, "step": 38075 }, { "epoch": 381.95, "learning_rate": 4.835443037974684e-07, "loss": 0.0, "step": 38100 }, { "epoch": 382.21, "learning_rate": 4.772151898734178e-07, "loss": 0.0, "step": 38125 }, { "epoch": 382.46, "learning_rate": 4.7088607594936713e-07, "loss": 0.0, "step": 38150 }, { "epoch": 382.71, "learning_rate": 4.645569620253165e-07, "loss": 0.0, "step": 38175 }, { "epoch": 382.96, "learning_rate": 4.582278481012659e-07, "loss": 0.0, "step": 38200 }, { "epoch": 383.21, "learning_rate": 4.518987341772152e-07, "loss": 0.0, "step": 38225 }, { "epoch": 383.46, "learning_rate": 4.455696202531646e-07, "loss": 0.0, "step": 38250 }, { "epoch": 383.71, "learning_rate": 4.3924050632911395e-07, "loss": 0.0, "step": 38275 }, { "epoch": 383.96, "learning_rate": 4.3291139240506333e-07, "loss": 0.0, "step": 38300 }, { "epoch": 384.21, "learning_rate": 4.2658227848101265e-07, "loss": 0.0, "step": 38325 }, { "epoch": 384.46, "learning_rate": 4.202531645569621e-07, "loss": 0.0, "step": 38350 }, { "epoch": 384.71, "learning_rate": 4.139240506329114e-07, "loss": 0.0, "step": 38375 }, { "epoch": 384.96, "learning_rate": 4.075949367088608e-07, "loss": 0.0, "step": 38400 }, { "epoch": 385.21, "learning_rate": 4.0126582278481015e-07, "loss": 0.0, "step": 38425 }, { "epoch": 385.46, "learning_rate": 3.949367088607595e-07, "loss": 0.0, "step": 38450 }, { "epoch": 385.71, "learning_rate": 3.8860759493670885e-07, "loss": 0.0, "step": 38475 }, { "epoch": 385.96, "learning_rate": 3.822784810126583e-07, "loss": 0.0, "step": 38500 }, { "epoch": 386.22, "learning_rate": 3.759493670886076e-07, "loss": 0.0, "step": 38525 }, { "epoch": 386.47, "learning_rate": 3.69620253164557e-07, "loss": 0.0, "step": 38550 }, { "epoch": 386.72, "learning_rate": 3.6329113924050634e-07, "loss": 0.0, "step": 38575 }, { "epoch": 386.97, "learning_rate": 3.569620253164557e-07, "loss": 0.0, "step": 38600 }, { "epoch": 387.22, "learning_rate": 3.506329113924051e-07, "loss": 0.0, "step": 38625 }, { "epoch": 387.47, "learning_rate": 3.4430379746835447e-07, "loss": 0.0, "step": 38650 }, { "epoch": 387.72, "learning_rate": 3.379746835443038e-07, "loss": 0.0, "step": 38675 }, { "epoch": 387.97, "learning_rate": 3.316455696202532e-07, "loss": 0.0, "step": 38700 }, { "epoch": 388.22, "learning_rate": 3.2531645569620254e-07, "loss": 0.0, "step": 38725 }, { "epoch": 388.47, "learning_rate": 3.1898734177215197e-07, "loss": 0.0, "step": 38750 }, { "epoch": 388.72, "learning_rate": 3.126582278481013e-07, "loss": 0.0, "step": 38775 }, { "epoch": 388.97, "learning_rate": 3.0632911392405066e-07, "loss": 0.0, "step": 38800 }, { "epoch": 389.22, "learning_rate": 3.0000000000000004e-07, "loss": 0.0, "step": 38825 }, { "epoch": 389.47, "learning_rate": 2.9367088607594936e-07, "loss": 0.0, "step": 38850 }, { "epoch": 389.72, "learning_rate": 2.8734177215189874e-07, "loss": 0.0, "step": 38875 }, { "epoch": 389.97, "learning_rate": 2.810126582278481e-07, "loss": 0.0, "step": 38900 }, { "epoch": 390.23, "learning_rate": 2.746835443037975e-07, "loss": 0.0, "step": 38925 }, { "epoch": 390.48, "learning_rate": 2.6835443037974686e-07, "loss": 0.0, "step": 38950 }, { "epoch": 390.73, "learning_rate": 2.6202531645569623e-07, "loss": 0.0, "step": 38975 }, { "epoch": 390.98, "learning_rate": 2.556962025316456e-07, "loss": 0.0, "step": 39000 }, { "epoch": 390.98, "eval_loss": 0.429427832365036, "eval_runtime": 1408.8936, "eval_samples_per_second": 9.044, "eval_steps_per_second": 0.283, "eval_wer": 9.215257151789903, "step": 39000 }, { "epoch": 391.23, "learning_rate": 2.4936708860759493e-07, "loss": 0.0, "step": 39025 }, { "epoch": 391.48, "learning_rate": 2.430379746835443e-07, "loss": 0.0, "step": 39050 }, { "epoch": 391.73, "learning_rate": 2.3670886075949368e-07, "loss": 0.0, "step": 39075 }, { "epoch": 391.98, "learning_rate": 2.306329113924051e-07, "loss": 0.0, "step": 39100 }, { "epoch": 392.23, "learning_rate": 2.2430379746835445e-07, "loss": 0.0, "step": 39125 }, { "epoch": 392.48, "learning_rate": 2.1797468354430383e-07, "loss": 0.0, "step": 39150 }, { "epoch": 392.73, "learning_rate": 2.116455696202532e-07, "loss": 0.0, "step": 39175 }, { "epoch": 392.98, "learning_rate": 2.0531645569620258e-07, "loss": 0.0, "step": 39200 }, { "epoch": 393.23, "learning_rate": 1.989873417721519e-07, "loss": 0.0, "step": 39225 }, { "epoch": 393.48, "learning_rate": 1.9265822784810127e-07, "loss": 0.0, "step": 39250 }, { "epoch": 393.73, "learning_rate": 1.8632911392405062e-07, "loss": 0.0, "step": 39275 }, { "epoch": 393.98, "learning_rate": 1.8e-07, "loss": 0.0, "step": 39300 }, { "epoch": 394.24, "learning_rate": 1.7367088607594937e-07, "loss": 0.0, "step": 39325 }, { "epoch": 394.49, "learning_rate": 1.6734177215189875e-07, "loss": 0.0, "step": 39350 }, { "epoch": 394.74, "learning_rate": 1.610126582278481e-07, "loss": 0.0, "step": 39375 }, { "epoch": 394.99, "learning_rate": 1.546835443037975e-07, "loss": 0.0, "step": 39400 }, { "epoch": 395.24, "learning_rate": 1.4835443037974684e-07, "loss": 0.0, "step": 39425 }, { "epoch": 395.49, "learning_rate": 1.4202531645569622e-07, "loss": 0.0, "step": 39450 }, { "epoch": 395.74, "learning_rate": 1.3569620253164557e-07, "loss": 0.0, "step": 39475 }, { "epoch": 395.99, "learning_rate": 1.2936708860759494e-07, "loss": 0.0, "step": 39500 }, { "epoch": 396.24, "learning_rate": 1.2303797468354432e-07, "loss": 0.0, "step": 39525 }, { "epoch": 396.49, "learning_rate": 1.1670886075949368e-07, "loss": 0.0, "step": 39550 }, { "epoch": 396.74, "learning_rate": 1.1037974683544304e-07, "loss": 0.0, "step": 39575 }, { "epoch": 396.99, "learning_rate": 1.0405063291139241e-07, "loss": 0.0, "step": 39600 }, { "epoch": 397.24, "learning_rate": 9.772151898734178e-08, "loss": 0.0, "step": 39625 }, { "epoch": 397.49, "learning_rate": 9.139240506329115e-08, "loss": 0.0, "step": 39650 }, { "epoch": 397.74, "learning_rate": 8.506329113924051e-08, "loss": 0.0, "step": 39675 }, { "epoch": 397.99, "learning_rate": 7.873417721518989e-08, "loss": 0.0, "step": 39700 }, { "epoch": 398.25, "learning_rate": 7.240506329113925e-08, "loss": 0.0, "step": 39725 }, { "epoch": 398.5, "learning_rate": 6.607594936708861e-08, "loss": 0.0, "step": 39750 }, { "epoch": 398.75, "learning_rate": 5.974683544303797e-08, "loss": 0.0, "step": 39775 }, { "epoch": 399.0, "learning_rate": 5.3417721518987346e-08, "loss": 0.0, "step": 39800 }, { "epoch": 399.25, "learning_rate": 4.708860759493671e-08, "loss": 0.0, "step": 39825 }, { "epoch": 399.5, "learning_rate": 4.0759493670886076e-08, "loss": 0.0, "step": 39850 }, { "epoch": 399.75, "learning_rate": 3.4430379746835444e-08, "loss": 0.0, "step": 39875 }, { "epoch": 400.0, "learning_rate": 2.8101265822784812e-08, "loss": 0.0, "step": 39900 }, { "epoch": 400.25, "learning_rate": 2.177215189873418e-08, "loss": 0.0, "step": 39925 }, { "epoch": 400.5, "learning_rate": 1.5443037974683545e-08, "loss": 0.0, "step": 39950 }, { "epoch": 400.75, "learning_rate": 9.113924050632912e-09, "loss": 0.0, "step": 39975 }, { "epoch": 401.0, "learning_rate": 2.7848101265822787e-09, "loss": 0.0, "step": 40000 }, { "epoch": 401.0, "eval_loss": 0.43030694127082825, "eval_runtime": 1408.5225, "eval_samples_per_second": 9.046, "eval_steps_per_second": 0.283, "eval_wer": 9.200601844614663, "step": 40000 }, { "epoch": 401.0, "step": 40000, "total_flos": 1.0429712599473193e+22, "train_loss": 0.004462987120796424, "train_runtime": 517255.1062, "train_samples_per_second": 19.797, "train_steps_per_second": 0.077 } ], "logging_steps": 25, "max_steps": 40000, "num_input_tokens_seen": 0, "num_train_epochs": 405, "save_steps": 1000, "total_flos": 1.0429712599473193e+22, "train_batch_size": 64, "trial_name": null, "trial_params": null }