{ "best_metric": 6.867476942316711, "best_model_checkpoint": "./checkpoint-37000", "epoch": 401.50564617314933, "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": 0.887, "step": 25 }, { "epoch": 0.5, "learning_rate": 1.0000000000000002e-06, "loss": 0.5318, "step": 50 }, { "epoch": 0.75, "learning_rate": 1.5e-06, "loss": 0.349, "step": 75 }, { "epoch": 1.0, "learning_rate": 2.0000000000000003e-06, "loss": 0.2908, "step": 100 }, { "epoch": 1.25, "learning_rate": 2.5e-06, "loss": 0.2495, "step": 125 }, { "epoch": 1.51, "learning_rate": 3e-06, "loss": 0.227, "step": 150 }, { "epoch": 1.76, "learning_rate": 3.5e-06, "loss": 0.2084, "step": 175 }, { "epoch": 2.01, "learning_rate": 4.000000000000001e-06, "loss": 0.1904, "step": 200 }, { "epoch": 2.26, "learning_rate": 4.5e-06, "loss": 0.1543, "step": 225 }, { "epoch": 2.51, "learning_rate": 5e-06, "loss": 0.147, "step": 250 }, { "epoch": 2.76, "learning_rate": 5.500000000000001e-06, "loss": 0.141, "step": 275 }, { "epoch": 3.01, "learning_rate": 6e-06, "loss": 0.1351, "step": 300 }, { "epoch": 3.26, "learning_rate": 6.5000000000000004e-06, "loss": 0.0963, "step": 325 }, { "epoch": 3.51, "learning_rate": 7e-06, "loss": 0.0932, "step": 350 }, { "epoch": 3.76, "learning_rate": 7.500000000000001e-06, "loss": 0.0977, "step": 375 }, { "epoch": 4.02, "learning_rate": 8.000000000000001e-06, "loss": 0.0942, "step": 400 }, { "epoch": 4.27, "learning_rate": 8.5e-06, "loss": 0.0607, "step": 425 }, { "epoch": 4.52, "learning_rate": 9e-06, "loss": 0.0603, "step": 450 }, { "epoch": 4.77, "learning_rate": 9.5e-06, "loss": 0.0646, "step": 475 }, { "epoch": 5.02, "learning_rate": 1e-05, "loss": 0.063, "step": 500 }, { "epoch": 5.27, "learning_rate": 9.99367088607595e-06, "loss": 0.0376, "step": 525 }, { "epoch": 5.52, "learning_rate": 9.9873417721519e-06, "loss": 0.0401, "step": 550 }, { "epoch": 5.77, "learning_rate": 9.981012658227849e-06, "loss": 0.0411, "step": 575 }, { "epoch": 6.02, "learning_rate": 9.974683544303799e-06, "loss": 0.0404, "step": 600 }, { "epoch": 6.27, "learning_rate": 9.968354430379748e-06, "loss": 0.025, "step": 625 }, { "epoch": 6.52, "learning_rate": 9.962025316455697e-06, "loss": 0.0242, "step": 650 }, { "epoch": 6.78, "learning_rate": 9.955696202531647e-06, "loss": 0.0261, "step": 675 }, { "epoch": 7.03, "learning_rate": 9.949367088607596e-06, "loss": 0.0235, "step": 700 }, { "epoch": 7.28, "learning_rate": 9.943037974683544e-06, "loss": 0.0147, "step": 725 }, { "epoch": 7.53, "learning_rate": 9.936708860759493e-06, "loss": 0.0161, "step": 750 }, { "epoch": 7.78, "learning_rate": 9.930379746835443e-06, "loss": 0.0158, "step": 775 }, { "epoch": 8.03, "learning_rate": 9.924050632911392e-06, "loss": 0.0154, "step": 800 }, { "epoch": 8.28, "learning_rate": 9.917721518987342e-06, "loss": 0.0098, "step": 825 }, { "epoch": 8.53, "learning_rate": 9.911392405063291e-06, "loss": 0.0101, "step": 850 }, { "epoch": 8.78, "learning_rate": 9.90506329113924e-06, "loss": 0.011, "step": 875 }, { "epoch": 9.03, "learning_rate": 9.89873417721519e-06, "loss": 0.0112, "step": 900 }, { "epoch": 9.28, "learning_rate": 9.892405063291141e-06, "loss": 0.007, "step": 925 }, { "epoch": 9.54, "learning_rate": 9.88607594936709e-06, "loss": 0.0074, "step": 950 }, { "epoch": 9.79, "learning_rate": 9.87974683544304e-06, "loss": 0.0076, "step": 975 }, { "epoch": 10.04, "learning_rate": 9.87341772151899e-06, "loss": 0.0095, "step": 1000 }, { "epoch": 10.04, "eval_loss": 0.20232713222503662, "eval_runtime": 2823.9879, "eval_samples_per_second": 4.512, "eval_steps_per_second": 0.282, "eval_wer": 9.680318899484133, "step": 1000 }, { "epoch": 10.29, "learning_rate": 9.867088607594937e-06, "loss": 0.0068, "step": 1025 }, { "epoch": 10.54, "learning_rate": 9.860759493670887e-06, "loss": 0.0062, "step": 1050 }, { "epoch": 10.79, "learning_rate": 9.854430379746836e-06, "loss": 0.0065, "step": 1075 }, { "epoch": 11.04, "learning_rate": 9.848101265822785e-06, "loss": 0.0059, "step": 1100 }, { "epoch": 11.29, "learning_rate": 9.841772151898735e-06, "loss": 0.0041, "step": 1125 }, { "epoch": 11.54, "learning_rate": 9.835443037974684e-06, "loss": 0.0043, "step": 1150 }, { "epoch": 11.79, "learning_rate": 9.829113924050634e-06, "loss": 0.0043, "step": 1175 }, { "epoch": 12.05, "learning_rate": 9.822784810126583e-06, "loss": 0.0046, "step": 1200 }, { "epoch": 12.3, "learning_rate": 9.816455696202533e-06, "loss": 0.0042, "step": 1225 }, { "epoch": 12.55, "learning_rate": 9.810126582278482e-06, "loss": 0.004, "step": 1250 }, { "epoch": 12.8, "learning_rate": 9.803797468354431e-06, "loss": 0.004, "step": 1275 }, { "epoch": 13.05, "learning_rate": 9.79746835443038e-06, "loss": 0.0039, "step": 1300 }, { "epoch": 13.3, "learning_rate": 9.79113924050633e-06, "loss": 0.0031, "step": 1325 }, { "epoch": 13.55, "learning_rate": 9.78481012658228e-06, "loss": 0.0029, "step": 1350 }, { "epoch": 13.8, "learning_rate": 9.778481012658229e-06, "loss": 0.003, "step": 1375 }, { "epoch": 14.05, "learning_rate": 9.772151898734179e-06, "loss": 0.0032, "step": 1400 }, { "epoch": 14.3, "learning_rate": 9.765822784810128e-06, "loss": 0.003, "step": 1425 }, { "epoch": 14.55, "learning_rate": 9.759493670886077e-06, "loss": 0.0029, "step": 1450 }, { "epoch": 14.81, "learning_rate": 9.753164556962025e-06, "loss": 0.0031, "step": 1475 }, { "epoch": 15.06, "learning_rate": 9.746835443037975e-06, "loss": 0.004, "step": 1500 }, { "epoch": 15.31, "learning_rate": 9.740506329113924e-06, "loss": 0.0034, "step": 1525 }, { "epoch": 15.56, "learning_rate": 9.734177215189873e-06, "loss": 0.0037, "step": 1550 }, { "epoch": 15.81, "learning_rate": 9.727848101265823e-06, "loss": 0.0038, "step": 1575 }, { "epoch": 16.06, "learning_rate": 9.721518987341772e-06, "loss": 0.0045, "step": 1600 }, { "epoch": 16.31, "learning_rate": 9.715189873417722e-06, "loss": 0.0039, "step": 1625 }, { "epoch": 16.56, "learning_rate": 9.708860759493671e-06, "loss": 0.0038, "step": 1650 }, { "epoch": 16.81, "learning_rate": 9.70253164556962e-06, "loss": 0.0043, "step": 1675 }, { "epoch": 17.06, "learning_rate": 9.69620253164557e-06, "loss": 0.0046, "step": 1700 }, { "epoch": 17.31, "learning_rate": 9.68987341772152e-06, "loss": 0.0041, "step": 1725 }, { "epoch": 17.57, "learning_rate": 9.68354430379747e-06, "loss": 0.0034, "step": 1750 }, { "epoch": 17.82, "learning_rate": 9.677215189873418e-06, "loss": 0.004, "step": 1775 }, { "epoch": 18.07, "learning_rate": 9.670886075949368e-06, "loss": 0.0034, "step": 1800 }, { "epoch": 18.32, "learning_rate": 9.664556962025317e-06, "loss": 0.0029, "step": 1825 }, { "epoch": 18.57, "learning_rate": 9.658227848101267e-06, "loss": 0.0031, "step": 1850 }, { "epoch": 18.82, "learning_rate": 9.651898734177216e-06, "loss": 0.0033, "step": 1875 }, { "epoch": 19.07, "learning_rate": 9.645569620253165e-06, "loss": 0.0032, "step": 1900 }, { "epoch": 19.32, "learning_rate": 9.639240506329115e-06, "loss": 0.0026, "step": 1925 }, { "epoch": 19.57, "learning_rate": 9.632911392405064e-06, "loss": 0.0031, "step": 1950 }, { "epoch": 19.82, "learning_rate": 9.626582278481014e-06, "loss": 0.0039, "step": 1975 }, { "epoch": 20.08, "learning_rate": 9.620253164556963e-06, "loss": 0.0032, "step": 2000 }, { "epoch": 20.08, "eval_loss": 0.21532024443149567, "eval_runtime": 2792.5976, "eval_samples_per_second": 4.563, "eval_steps_per_second": 0.285, "eval_wer": 9.052094731905582, "step": 2000 }, { "epoch": 20.33, "learning_rate": 9.613924050632913e-06, "loss": 0.0026, "step": 2025 }, { "epoch": 20.58, "learning_rate": 9.607594936708862e-06, "loss": 0.0022, "step": 2050 }, { "epoch": 20.83, "learning_rate": 9.601265822784811e-06, "loss": 0.0028, "step": 2075 }, { "epoch": 21.08, "learning_rate": 9.59493670886076e-06, "loss": 0.0025, "step": 2100 }, { "epoch": 21.33, "learning_rate": 9.58860759493671e-06, "loss": 0.0022, "step": 2125 }, { "epoch": 21.58, "learning_rate": 9.58227848101266e-06, "loss": 0.002, "step": 2150 }, { "epoch": 21.83, "learning_rate": 9.575949367088609e-06, "loss": 0.0026, "step": 2175 }, { "epoch": 22.08, "learning_rate": 9.569620253164559e-06, "loss": 0.0027, "step": 2200 }, { "epoch": 22.33, "learning_rate": 9.563291139240506e-06, "loss": 0.0018, "step": 2225 }, { "epoch": 22.58, "learning_rate": 9.556962025316456e-06, "loss": 0.002, "step": 2250 }, { "epoch": 22.84, "learning_rate": 9.550632911392405e-06, "loss": 0.003, "step": 2275 }, { "epoch": 23.09, "learning_rate": 9.544303797468355e-06, "loss": 0.0022, "step": 2300 }, { "epoch": 23.34, "learning_rate": 9.537974683544304e-06, "loss": 0.002, "step": 2325 }, { "epoch": 23.59, "learning_rate": 9.531645569620253e-06, "loss": 0.0019, "step": 2350 }, { "epoch": 23.84, "learning_rate": 9.525316455696203e-06, "loss": 0.0026, "step": 2375 }, { "epoch": 24.09, "learning_rate": 9.518987341772152e-06, "loss": 0.0022, "step": 2400 }, { "epoch": 24.34, "learning_rate": 9.512658227848102e-06, "loss": 0.0019, "step": 2425 }, { "epoch": 24.59, "learning_rate": 9.506329113924051e-06, "loss": 0.0025, "step": 2450 }, { "epoch": 24.84, "learning_rate": 9.5e-06, "loss": 0.0027, "step": 2475 }, { "epoch": 25.09, "learning_rate": 9.49367088607595e-06, "loss": 0.0023, "step": 2500 }, { "epoch": 25.35, "learning_rate": 9.4873417721519e-06, "loss": 0.0019, "step": 2525 }, { "epoch": 25.6, "learning_rate": 9.481012658227849e-06, "loss": 0.0018, "step": 2550 }, { "epoch": 25.85, "learning_rate": 9.474683544303798e-06, "loss": 0.0021, "step": 2575 }, { "epoch": 26.1, "learning_rate": 9.468354430379748e-06, "loss": 0.0022, "step": 2600 }, { "epoch": 26.35, "learning_rate": 9.462025316455697e-06, "loss": 0.0025, "step": 2625 }, { "epoch": 26.6, "learning_rate": 9.455696202531647e-06, "loss": 0.0022, "step": 2650 }, { "epoch": 26.85, "learning_rate": 9.449367088607596e-06, "loss": 0.0024, "step": 2675 }, { "epoch": 27.1, "learning_rate": 9.443037974683545e-06, "loss": 0.0028, "step": 2700 }, { "epoch": 27.35, "learning_rate": 9.436708860759495e-06, "loss": 0.0025, "step": 2725 }, { "epoch": 27.6, "learning_rate": 9.430379746835444e-06, "loss": 0.0026, "step": 2750 }, { "epoch": 27.85, "learning_rate": 9.424050632911394e-06, "loss": 0.0026, "step": 2775 }, { "epoch": 28.11, "learning_rate": 9.417721518987343e-06, "loss": 0.0024, "step": 2800 }, { "epoch": 28.36, "learning_rate": 9.411392405063293e-06, "loss": 0.0025, "step": 2825 }, { "epoch": 28.61, "learning_rate": 9.405063291139242e-06, "loss": 0.0023, "step": 2850 }, { "epoch": 28.86, "learning_rate": 9.398734177215191e-06, "loss": 0.0027, "step": 2875 }, { "epoch": 29.11, "learning_rate": 9.39240506329114e-06, "loss": 0.0025, "step": 2900 }, { "epoch": 29.36, "learning_rate": 9.38607594936709e-06, "loss": 0.0026, "step": 2925 }, { "epoch": 29.61, "learning_rate": 9.379746835443038e-06, "loss": 0.0021, "step": 2950 }, { "epoch": 29.86, "learning_rate": 9.373417721518987e-06, "loss": 0.0025, "step": 2975 }, { "epoch": 30.11, "learning_rate": 9.367088607594937e-06, "loss": 0.0023, "step": 3000 }, { "epoch": 30.11, "eval_loss": 0.22344575822353363, "eval_runtime": 2793.5917, "eval_samples_per_second": 4.561, "eval_steps_per_second": 0.285, "eval_wer": 8.86450680006253, "step": 3000 }, { "epoch": 30.36, "learning_rate": 9.360759493670886e-06, "loss": 0.0017, "step": 3025 }, { "epoch": 30.61, "learning_rate": 9.354430379746836e-06, "loss": 0.0018, "step": 3050 }, { "epoch": 30.87, "learning_rate": 9.348101265822785e-06, "loss": 0.0019, "step": 3075 }, { "epoch": 31.12, "learning_rate": 9.341772151898735e-06, "loss": 0.0017, "step": 3100 }, { "epoch": 31.37, "learning_rate": 9.335443037974684e-06, "loss": 0.0013, "step": 3125 }, { "epoch": 31.62, "learning_rate": 9.329113924050633e-06, "loss": 0.0016, "step": 3150 }, { "epoch": 31.87, "learning_rate": 9.322784810126583e-06, "loss": 0.0018, "step": 3175 }, { "epoch": 32.12, "learning_rate": 9.316455696202532e-06, "loss": 0.0016, "step": 3200 }, { "epoch": 32.37, "learning_rate": 9.310126582278482e-06, "loss": 0.0017, "step": 3225 }, { "epoch": 32.62, "learning_rate": 9.303797468354431e-06, "loss": 0.0014, "step": 3250 }, { "epoch": 32.87, "learning_rate": 9.29746835443038e-06, "loss": 0.0016, "step": 3275 }, { "epoch": 33.12, "learning_rate": 9.29113924050633e-06, "loss": 0.0013, "step": 3300 }, { "epoch": 33.38, "learning_rate": 9.28481012658228e-06, "loss": 0.0012, "step": 3325 }, { "epoch": 33.63, "learning_rate": 9.278481012658229e-06, "loss": 0.0011, "step": 3350 }, { "epoch": 33.88, "learning_rate": 9.272151898734178e-06, "loss": 0.0018, "step": 3375 }, { "epoch": 34.13, "learning_rate": 9.265822784810128e-06, "loss": 0.0019, "step": 3400 }, { "epoch": 34.38, "learning_rate": 9.259493670886077e-06, "loss": 0.0015, "step": 3425 }, { "epoch": 34.63, "learning_rate": 9.253164556962027e-06, "loss": 0.002, "step": 3450 }, { "epoch": 34.88, "learning_rate": 9.246835443037976e-06, "loss": 0.0023, "step": 3475 }, { "epoch": 35.13, "learning_rate": 9.240506329113925e-06, "loss": 0.0025, "step": 3500 }, { "epoch": 35.38, "learning_rate": 9.234177215189875e-06, "loss": 0.0023, "step": 3525 }, { "epoch": 35.63, "learning_rate": 9.227848101265824e-06, "loss": 0.0026, "step": 3550 }, { "epoch": 35.88, "learning_rate": 9.221518987341774e-06, "loss": 0.0029, "step": 3575 }, { "epoch": 36.14, "learning_rate": 9.215189873417723e-06, "loss": 0.0022, "step": 3600 }, { "epoch": 36.39, "learning_rate": 9.208860759493673e-06, "loss": 0.0024, "step": 3625 }, { "epoch": 36.64, "learning_rate": 9.202531645569622e-06, "loss": 0.0021, "step": 3650 }, { "epoch": 36.89, "learning_rate": 9.196202531645571e-06, "loss": 0.0021, "step": 3675 }, { "epoch": 37.14, "learning_rate": 9.189873417721519e-06, "loss": 0.0017, "step": 3700 }, { "epoch": 37.39, "learning_rate": 9.183544303797469e-06, "loss": 0.0013, "step": 3725 }, { "epoch": 37.64, "learning_rate": 9.177215189873418e-06, "loss": 0.0024, "step": 3750 }, { "epoch": 37.89, "learning_rate": 9.170886075949367e-06, "loss": 0.0029, "step": 3775 }, { "epoch": 38.14, "learning_rate": 9.164556962025317e-06, "loss": 0.0019, "step": 3800 }, { "epoch": 38.39, "learning_rate": 9.158227848101266e-06, "loss": 0.0016, "step": 3825 }, { "epoch": 38.64, "learning_rate": 9.151898734177216e-06, "loss": 0.0019, "step": 3850 }, { "epoch": 38.9, "learning_rate": 9.145569620253165e-06, "loss": 0.0028, "step": 3875 }, { "epoch": 39.15, "learning_rate": 9.139240506329115e-06, "loss": 0.003, "step": 3900 }, { "epoch": 39.4, "learning_rate": 9.132911392405064e-06, "loss": 0.0028, "step": 3925 }, { "epoch": 39.65, "learning_rate": 9.126582278481013e-06, "loss": 0.0027, "step": 3950 }, { "epoch": 39.9, "learning_rate": 9.120253164556963e-06, "loss": 0.0029, "step": 3975 }, { "epoch": 40.15, "learning_rate": 9.113924050632912e-06, "loss": 0.0023, "step": 4000 }, { "epoch": 40.15, "eval_loss": 0.22782479226589203, "eval_runtime": 2785.8527, "eval_samples_per_second": 4.574, "eval_steps_per_second": 0.286, "eval_wer": 8.436571830545567, "step": 4000 }, { "epoch": 40.4, "learning_rate": 9.107594936708862e-06, "loss": 0.0026, "step": 4025 }, { "epoch": 40.65, "learning_rate": 9.101265822784811e-06, "loss": 0.0032, "step": 4050 }, { "epoch": 40.9, "learning_rate": 9.09493670886076e-06, "loss": 0.0028, "step": 4075 }, { "epoch": 41.15, "learning_rate": 9.08860759493671e-06, "loss": 0.0023, "step": 4100 }, { "epoch": 41.41, "learning_rate": 9.08227848101266e-06, "loss": 0.0016, "step": 4125 }, { "epoch": 41.66, "learning_rate": 9.075949367088607e-06, "loss": 0.002, "step": 4150 }, { "epoch": 41.91, "learning_rate": 9.069620253164557e-06, "loss": 0.002, "step": 4175 }, { "epoch": 42.16, "learning_rate": 9.063291139240506e-06, "loss": 0.0013, "step": 4200 }, { "epoch": 42.41, "learning_rate": 9.056962025316457e-06, "loss": 0.0016, "step": 4225 }, { "epoch": 42.66, "learning_rate": 9.050632911392407e-06, "loss": 0.002, "step": 4250 }, { "epoch": 42.91, "learning_rate": 9.044303797468356e-06, "loss": 0.0018, "step": 4275 }, { "epoch": 43.16, "learning_rate": 9.037974683544305e-06, "loss": 0.0012, "step": 4300 }, { "epoch": 43.41, "learning_rate": 9.031645569620255e-06, "loss": 0.0015, "step": 4325 }, { "epoch": 43.66, "learning_rate": 9.025316455696204e-06, "loss": 0.0012, "step": 4350 }, { "epoch": 43.91, "learning_rate": 9.018987341772154e-06, "loss": 0.0024, "step": 4375 }, { "epoch": 44.17, "learning_rate": 9.012658227848103e-06, "loss": 0.0019, "step": 4400 }, { "epoch": 44.42, "learning_rate": 9.00632911392405e-06, "loss": 0.0013, "step": 4425 }, { "epoch": 44.67, "learning_rate": 9e-06, "loss": 0.0014, "step": 4450 }, { "epoch": 44.92, "learning_rate": 8.99367088607595e-06, "loss": 0.0018, "step": 4475 }, { "epoch": 45.17, "learning_rate": 8.987341772151899e-06, "loss": 0.0018, "step": 4500 }, { "epoch": 45.42, "learning_rate": 8.981012658227849e-06, "loss": 0.0014, "step": 4525 }, { "epoch": 45.67, "learning_rate": 8.974683544303798e-06, "loss": 0.0018, "step": 4550 }, { "epoch": 45.92, "learning_rate": 8.968354430379747e-06, "loss": 0.0018, "step": 4575 }, { "epoch": 46.17, "learning_rate": 8.962025316455697e-06, "loss": 0.0012, "step": 4600 }, { "epoch": 46.42, "learning_rate": 8.955696202531646e-06, "loss": 0.0013, "step": 4625 }, { "epoch": 46.68, "learning_rate": 8.949367088607596e-06, "loss": 0.0013, "step": 4650 }, { "epoch": 46.93, "learning_rate": 8.943037974683545e-06, "loss": 0.0011, "step": 4675 }, { "epoch": 47.18, "learning_rate": 8.936708860759495e-06, "loss": 0.0008, "step": 4700 }, { "epoch": 47.43, "learning_rate": 8.930379746835444e-06, "loss": 0.0009, "step": 4725 }, { "epoch": 47.68, "learning_rate": 8.924050632911393e-06, "loss": 0.0009, "step": 4750 }, { "epoch": 47.93, "learning_rate": 8.917721518987343e-06, "loss": 0.001, "step": 4775 }, { "epoch": 48.18, "learning_rate": 8.911392405063292e-06, "loss": 0.0014, "step": 4800 }, { "epoch": 48.43, "learning_rate": 8.905063291139242e-06, "loss": 0.001, "step": 4825 }, { "epoch": 48.68, "learning_rate": 8.898734177215191e-06, "loss": 0.0013, "step": 4850 }, { "epoch": 48.93, "learning_rate": 8.892405063291139e-06, "loss": 0.0009, "step": 4875 }, { "epoch": 49.18, "learning_rate": 8.886075949367088e-06, "loss": 0.0007, "step": 4900 }, { "epoch": 49.44, "learning_rate": 8.879746835443038e-06, "loss": 0.0008, "step": 4925 }, { "epoch": 49.69, "learning_rate": 8.873417721518987e-06, "loss": 0.0007, "step": 4950 }, { "epoch": 49.94, "learning_rate": 8.867088607594937e-06, "loss": 0.0013, "step": 4975 }, { "epoch": 50.19, "learning_rate": 8.860759493670886e-06, "loss": 0.0012, "step": 5000 }, { "epoch": 50.19, "eval_loss": 0.2259608805179596, "eval_runtime": 2783.4754, "eval_samples_per_second": 4.578, "eval_steps_per_second": 0.286, "eval_wer": 7.991050492418321, "step": 5000 }, { "epoch": 50.44, "learning_rate": 8.854430379746835e-06, "loss": 0.0012, "step": 5025 }, { "epoch": 50.69, "learning_rate": 8.848101265822786e-06, "loss": 0.0014, "step": 5050 }, { "epoch": 50.94, "learning_rate": 8.841772151898736e-06, "loss": 0.0012, "step": 5075 }, { "epoch": 51.19, "learning_rate": 8.835443037974685e-06, "loss": 0.0011, "step": 5100 }, { "epoch": 51.44, "learning_rate": 8.829113924050635e-06, "loss": 0.0013, "step": 5125 }, { "epoch": 51.69, "learning_rate": 8.822784810126584e-06, "loss": 0.0013, "step": 5150 }, { "epoch": 51.94, "learning_rate": 8.816455696202532e-06, "loss": 0.0015, "step": 5175 }, { "epoch": 52.2, "learning_rate": 8.810126582278481e-06, "loss": 0.0013, "step": 5200 }, { "epoch": 52.45, "learning_rate": 8.80379746835443e-06, "loss": 0.001, "step": 5225 }, { "epoch": 52.7, "learning_rate": 8.79746835443038e-06, "loss": 0.0015, "step": 5250 }, { "epoch": 52.95, "learning_rate": 8.79113924050633e-06, "loss": 0.0017, "step": 5275 }, { "epoch": 53.2, "learning_rate": 8.784810126582279e-06, "loss": 0.0024, "step": 5300 }, { "epoch": 53.45, "learning_rate": 8.778481012658229e-06, "loss": 0.0027, "step": 5325 }, { "epoch": 53.7, "learning_rate": 8.772151898734178e-06, "loss": 0.0023, "step": 5350 }, { "epoch": 53.95, "learning_rate": 8.765822784810127e-06, "loss": 0.0023, "step": 5375 }, { "epoch": 54.2, "learning_rate": 8.759493670886077e-06, "loss": 0.0023, "step": 5400 }, { "epoch": 54.45, "learning_rate": 8.753164556962026e-06, "loss": 0.0017, "step": 5425 }, { "epoch": 54.71, "learning_rate": 8.746835443037976e-06, "loss": 0.0014, "step": 5450 }, { "epoch": 54.96, "learning_rate": 8.740506329113925e-06, "loss": 0.0018, "step": 5475 }, { "epoch": 55.21, "learning_rate": 8.734177215189874e-06, "loss": 0.0013, "step": 5500 }, { "epoch": 55.46, "learning_rate": 8.727848101265824e-06, "loss": 0.0017, "step": 5525 }, { "epoch": 55.71, "learning_rate": 8.721518987341773e-06, "loss": 0.0018, "step": 5550 }, { "epoch": 55.96, "learning_rate": 8.715189873417723e-06, "loss": 0.0015, "step": 5575 }, { "epoch": 56.21, "learning_rate": 8.708860759493672e-06, "loss": 0.001, "step": 5600 }, { "epoch": 56.46, "learning_rate": 8.70253164556962e-06, "loss": 0.0009, "step": 5625 }, { "epoch": 56.71, "learning_rate": 8.69620253164557e-06, "loss": 0.001, "step": 5650 }, { "epoch": 56.96, "learning_rate": 8.689873417721519e-06, "loss": 0.0015, "step": 5675 }, { "epoch": 57.21, "learning_rate": 8.683544303797468e-06, "loss": 0.0016, "step": 5700 }, { "epoch": 57.47, "learning_rate": 8.677215189873418e-06, "loss": 0.0014, "step": 5725 }, { "epoch": 57.72, "learning_rate": 8.670886075949367e-06, "loss": 0.0013, "step": 5750 }, { "epoch": 57.97, "learning_rate": 8.664556962025317e-06, "loss": 0.0015, "step": 5775 }, { "epoch": 58.22, "learning_rate": 8.658227848101266e-06, "loss": 0.0016, "step": 5800 }, { "epoch": 58.47, "learning_rate": 8.651898734177215e-06, "loss": 0.0017, "step": 5825 }, { "epoch": 58.72, "learning_rate": 8.645569620253166e-06, "loss": 0.0015, "step": 5850 }, { "epoch": 58.97, "learning_rate": 8.639240506329116e-06, "loss": 0.0012, "step": 5875 }, { "epoch": 59.22, "learning_rate": 8.632911392405064e-06, "loss": 0.0013, "step": 5900 }, { "epoch": 59.47, "learning_rate": 8.626582278481013e-06, "loss": 0.0008, "step": 5925 }, { "epoch": 59.72, "learning_rate": 8.620253164556963e-06, "loss": 0.0007, "step": 5950 }, { "epoch": 59.97, "learning_rate": 8.613924050632912e-06, "loss": 0.0007, "step": 5975 }, { "epoch": 60.23, "learning_rate": 8.607594936708861e-06, "loss": 0.0005, "step": 6000 }, { "epoch": 60.23, "eval_loss": 0.24347658455371857, "eval_runtime": 2781.2525, "eval_samples_per_second": 4.581, "eval_steps_per_second": 0.287, "eval_wer": 7.906049710801938, "step": 6000 }, { "epoch": 60.48, "learning_rate": 8.60126582278481e-06, "loss": 0.0006, "step": 6025 }, { "epoch": 60.73, "learning_rate": 8.59493670886076e-06, "loss": 0.0006, "step": 6050 }, { "epoch": 60.98, "learning_rate": 8.58860759493671e-06, "loss": 0.0005, "step": 6075 }, { "epoch": 61.23, "learning_rate": 8.582278481012659e-06, "loss": 0.0004, "step": 6100 }, { "epoch": 61.48, "learning_rate": 8.575949367088608e-06, "loss": 0.0005, "step": 6125 }, { "epoch": 61.73, "learning_rate": 8.569620253164558e-06, "loss": 0.0005, "step": 6150 }, { "epoch": 61.98, "learning_rate": 8.563291139240507e-06, "loss": 0.0006, "step": 6175 }, { "epoch": 62.23, "learning_rate": 8.556962025316457e-06, "loss": 0.0004, "step": 6200 }, { "epoch": 62.48, "learning_rate": 8.550632911392406e-06, "loss": 0.0003, "step": 6225 }, { "epoch": 62.74, "learning_rate": 8.544303797468356e-06, "loss": 0.0002, "step": 6250 }, { "epoch": 62.99, "learning_rate": 8.537974683544305e-06, "loss": 0.0003, "step": 6275 }, { "epoch": 63.24, "learning_rate": 8.531645569620254e-06, "loss": 0.0001, "step": 6300 }, { "epoch": 63.49, "learning_rate": 8.525316455696204e-06, "loss": 0.0002, "step": 6325 }, { "epoch": 63.74, "learning_rate": 8.518987341772152e-06, "loss": 0.0001, "step": 6350 }, { "epoch": 63.99, "learning_rate": 8.512658227848101e-06, "loss": 0.0002, "step": 6375 }, { "epoch": 64.24, "learning_rate": 8.50632911392405e-06, "loss": 0.0004, "step": 6400 }, { "epoch": 64.49, "learning_rate": 8.5e-06, "loss": 0.0003, "step": 6425 }, { "epoch": 64.74, "learning_rate": 8.49367088607595e-06, "loss": 0.0004, "step": 6450 }, { "epoch": 64.99, "learning_rate": 8.487341772151899e-06, "loss": 0.0004, "step": 6475 }, { "epoch": 65.24, "learning_rate": 8.481012658227848e-06, "loss": 0.0007, "step": 6500 }, { "epoch": 65.5, "learning_rate": 8.474683544303798e-06, "loss": 0.0007, "step": 6525 }, { "epoch": 65.75, "learning_rate": 8.468354430379747e-06, "loss": 0.001, "step": 6550 }, { "epoch": 66.0, "learning_rate": 8.462025316455696e-06, "loss": 0.0012, "step": 6575 }, { "epoch": 66.25, "learning_rate": 8.455696202531646e-06, "loss": 0.0014, "step": 6600 }, { "epoch": 66.5, "learning_rate": 8.449367088607595e-06, "loss": 0.0025, "step": 6625 }, { "epoch": 66.75, "learning_rate": 8.443037974683545e-06, "loss": 0.0024, "step": 6650 }, { "epoch": 67.0, "learning_rate": 8.436708860759494e-06, "loss": 0.0025, "step": 6675 }, { "epoch": 67.25, "learning_rate": 8.430379746835444e-06, "loss": 0.0024, "step": 6700 }, { "epoch": 67.5, "learning_rate": 8.424050632911393e-06, "loss": 0.0035, "step": 6725 }, { "epoch": 67.75, "learning_rate": 8.417721518987342e-06, "loss": 0.0028, "step": 6750 }, { "epoch": 68.01, "learning_rate": 8.411392405063292e-06, "loss": 0.0033, "step": 6775 }, { "epoch": 68.26, "learning_rate": 8.405063291139241e-06, "loss": 0.0026, "step": 6800 }, { "epoch": 68.51, "learning_rate": 8.39873417721519e-06, "loss": 0.0027, "step": 6825 }, { "epoch": 68.76, "learning_rate": 8.39240506329114e-06, "loss": 0.0029, "step": 6850 }, { "epoch": 69.01, "learning_rate": 8.38607594936709e-06, "loss": 0.0029, "step": 6875 }, { "epoch": 69.26, "learning_rate": 8.379746835443039e-06, "loss": 0.0022, "step": 6900 }, { "epoch": 69.51, "learning_rate": 8.373417721518988e-06, "loss": 0.0015, "step": 6925 }, { "epoch": 69.76, "learning_rate": 8.367088607594938e-06, "loss": 0.0018, "step": 6950 }, { "epoch": 70.01, "learning_rate": 8.360759493670887e-06, "loss": 0.002, "step": 6975 }, { "epoch": 70.26, "learning_rate": 8.354430379746837e-06, "loss": 0.0013, "step": 7000 }, { "epoch": 70.26, "eval_loss": 0.2253977358341217, "eval_runtime": 2781.6008, "eval_samples_per_second": 4.581, "eval_steps_per_second": 0.287, "eval_wer": 7.848405502579333, "step": 7000 }, { "epoch": 70.51, "learning_rate": 8.348101265822786e-06, "loss": 0.0015, "step": 7025 }, { "epoch": 70.77, "learning_rate": 8.341772151898736e-06, "loss": 0.0014, "step": 7050 }, { "epoch": 71.02, "learning_rate": 8.335443037974685e-06, "loss": 0.0014, "step": 7075 }, { "epoch": 71.27, "learning_rate": 8.329113924050633e-06, "loss": 0.001, "step": 7100 }, { "epoch": 71.52, "learning_rate": 8.322784810126582e-06, "loss": 0.0017, "step": 7125 }, { "epoch": 71.77, "learning_rate": 8.316455696202532e-06, "loss": 0.0015, "step": 7150 }, { "epoch": 72.02, "learning_rate": 8.310126582278481e-06, "loss": 0.0013, "step": 7175 }, { "epoch": 72.27, "learning_rate": 8.30379746835443e-06, "loss": 0.0012, "step": 7200 }, { "epoch": 72.52, "learning_rate": 8.29746835443038e-06, "loss": 0.001, "step": 7225 }, { "epoch": 72.77, "learning_rate": 8.29113924050633e-06, "loss": 0.0014, "step": 7250 }, { "epoch": 73.02, "learning_rate": 8.284810126582279e-06, "loss": 0.0015, "step": 7275 }, { "epoch": 73.27, "learning_rate": 8.278481012658228e-06, "loss": 0.0015, "step": 7300 }, { "epoch": 73.53, "learning_rate": 8.272151898734178e-06, "loss": 0.001, "step": 7325 }, { "epoch": 73.78, "learning_rate": 8.265822784810127e-06, "loss": 0.0011, "step": 7350 }, { "epoch": 74.03, "learning_rate": 8.259493670886076e-06, "loss": 0.0013, "step": 7375 }, { "epoch": 74.28, "learning_rate": 8.253164556962026e-06, "loss": 0.0007, "step": 7400 }, { "epoch": 74.53, "learning_rate": 8.246835443037975e-06, "loss": 0.0006, "step": 7425 }, { "epoch": 74.78, "learning_rate": 8.240506329113925e-06, "loss": 0.0007, "step": 7450 }, { "epoch": 75.03, "learning_rate": 8.234177215189874e-06, "loss": 0.0007, "step": 7475 }, { "epoch": 75.28, "learning_rate": 8.227848101265824e-06, "loss": 0.0008, "step": 7500 }, { "epoch": 75.53, "learning_rate": 8.221518987341773e-06, "loss": 0.0007, "step": 7525 }, { "epoch": 75.78, "learning_rate": 8.215189873417722e-06, "loss": 0.0005, "step": 7550 }, { "epoch": 76.04, "learning_rate": 8.208860759493672e-06, "loss": 0.0006, "step": 7575 }, { "epoch": 76.29, "learning_rate": 8.202531645569621e-06, "loss": 0.0005, "step": 7600 }, { "epoch": 76.54, "learning_rate": 8.19620253164557e-06, "loss": 0.0005, "step": 7625 }, { "epoch": 76.79, "learning_rate": 8.18987341772152e-06, "loss": 0.0004, "step": 7650 }, { "epoch": 77.04, "learning_rate": 8.18354430379747e-06, "loss": 0.0003, "step": 7675 }, { "epoch": 77.29, "learning_rate": 8.177215189873419e-06, "loss": 0.0004, "step": 7700 }, { "epoch": 77.54, "learning_rate": 8.170886075949368e-06, "loss": 0.0005, "step": 7725 }, { "epoch": 77.79, "learning_rate": 8.164556962025318e-06, "loss": 0.0007, "step": 7750 }, { "epoch": 78.04, "learning_rate": 8.158227848101267e-06, "loss": 0.0007, "step": 7775 }, { "epoch": 78.29, "learning_rate": 8.151898734177217e-06, "loss": 0.0004, "step": 7800 }, { "epoch": 78.54, "learning_rate": 8.145569620253164e-06, "loss": 0.0005, "step": 7825 }, { "epoch": 78.8, "learning_rate": 8.139240506329114e-06, "loss": 0.0004, "step": 7850 }, { "epoch": 79.05, "learning_rate": 8.132911392405063e-06, "loss": 0.0004, "step": 7875 }, { "epoch": 79.3, "learning_rate": 8.126582278481013e-06, "loss": 0.0005, "step": 7900 }, { "epoch": 79.55, "learning_rate": 8.120253164556962e-06, "loss": 0.0003, "step": 7925 }, { "epoch": 79.8, "learning_rate": 8.113924050632912e-06, "loss": 0.0005, "step": 7950 }, { "epoch": 80.05, "learning_rate": 8.107594936708861e-06, "loss": 0.0006, "step": 7975 }, { "epoch": 80.3, "learning_rate": 8.10126582278481e-06, "loss": 0.0004, "step": 8000 }, { "epoch": 80.3, "eval_loss": 0.23669902980327606, "eval_runtime": 2781.0884, "eval_samples_per_second": 4.582, "eval_steps_per_second": 0.287, "eval_wer": 7.482999843676723, "step": 8000 }, { "epoch": 80.55, "learning_rate": 8.09493670886076e-06, "loss": 0.0007, "step": 8025 }, { "epoch": 80.8, "learning_rate": 8.08860759493671e-06, "loss": 0.0012, "step": 8050 }, { "epoch": 81.05, "learning_rate": 8.082278481012659e-06, "loss": 0.0011, "step": 8075 }, { "epoch": 81.3, "learning_rate": 8.075949367088608e-06, "loss": 0.0011, "step": 8100 }, { "epoch": 81.56, "learning_rate": 8.069620253164558e-06, "loss": 0.0009, "step": 8125 }, { "epoch": 81.81, "learning_rate": 8.063291139240507e-06, "loss": 0.001, "step": 8150 }, { "epoch": 82.06, "learning_rate": 8.056962025316456e-06, "loss": 0.0008, "step": 8175 }, { "epoch": 82.31, "learning_rate": 8.050632911392406e-06, "loss": 0.0006, "step": 8200 }, { "epoch": 82.56, "learning_rate": 8.044303797468355e-06, "loss": 0.0004, "step": 8225 }, { "epoch": 82.81, "learning_rate": 8.037974683544305e-06, "loss": 0.0005, "step": 8250 }, { "epoch": 83.06, "learning_rate": 8.031645569620252e-06, "loss": 0.0005, "step": 8275 }, { "epoch": 83.31, "learning_rate": 8.025316455696202e-06, "loss": 0.0005, "step": 8300 }, { "epoch": 83.56, "learning_rate": 8.018987341772153e-06, "loss": 0.0006, "step": 8325 }, { "epoch": 83.81, "learning_rate": 8.012658227848102e-06, "loss": 0.0005, "step": 8350 }, { "epoch": 84.07, "learning_rate": 8.006329113924052e-06, "loss": 0.0005, "step": 8375 }, { "epoch": 84.32, "learning_rate": 8.000000000000001e-06, "loss": 0.0005, "step": 8400 }, { "epoch": 84.57, "learning_rate": 7.99367088607595e-06, "loss": 0.0005, "step": 8425 }, { "epoch": 84.82, "learning_rate": 7.9873417721519e-06, "loss": 0.0009, "step": 8450 }, { "epoch": 85.07, "learning_rate": 7.98101265822785e-06, "loss": 0.0008, "step": 8475 }, { "epoch": 85.32, "learning_rate": 7.974683544303799e-06, "loss": 0.0006, "step": 8500 }, { "epoch": 85.57, "learning_rate": 7.968354430379748e-06, "loss": 0.0008, "step": 8525 }, { "epoch": 85.82, "learning_rate": 7.962025316455698e-06, "loss": 0.0014, "step": 8550 }, { "epoch": 86.07, "learning_rate": 7.955696202531646e-06, "loss": 0.0011, "step": 8575 }, { "epoch": 86.32, "learning_rate": 7.949367088607595e-06, "loss": 0.001, "step": 8600 }, { "epoch": 86.57, "learning_rate": 7.943037974683544e-06, "loss": 0.0014, "step": 8625 }, { "epoch": 86.83, "learning_rate": 7.936708860759494e-06, "loss": 0.0017, "step": 8650 }, { "epoch": 87.08, "learning_rate": 7.930632911392405e-06, "loss": 0.0015, "step": 8675 }, { "epoch": 87.33, "learning_rate": 7.924303797468355e-06, "loss": 0.0019, "step": 8700 }, { "epoch": 87.58, "learning_rate": 7.917974683544304e-06, "loss": 0.0023, "step": 8725 }, { "epoch": 87.83, "learning_rate": 7.911645569620254e-06, "loss": 0.0023, "step": 8750 }, { "epoch": 88.08, "learning_rate": 7.905316455696203e-06, "loss": 0.0017, "step": 8775 }, { "epoch": 88.33, "learning_rate": 7.898987341772153e-06, "loss": 0.0017, "step": 8800 }, { "epoch": 88.58, "learning_rate": 7.892658227848102e-06, "loss": 0.0014, "step": 8825 }, { "epoch": 88.83, "learning_rate": 7.886329113924051e-06, "loss": 0.0018, "step": 8850 }, { "epoch": 89.08, "learning_rate": 7.88e-06, "loss": 0.0016, "step": 8875 }, { "epoch": 89.34, "learning_rate": 7.87367088607595e-06, "loss": 0.0015, "step": 8900 }, { "epoch": 89.59, "learning_rate": 7.8673417721519e-06, "loss": 0.001, "step": 8925 }, { "epoch": 89.84, "learning_rate": 7.861012658227849e-06, "loss": 0.0009, "step": 8950 }, { "epoch": 90.09, "learning_rate": 7.854683544303798e-06, "loss": 0.0009, "step": 8975 }, { "epoch": 90.34, "learning_rate": 7.848354430379748e-06, "loss": 0.0008, "step": 9000 }, { "epoch": 90.34, "eval_loss": 0.22892530262470245, "eval_runtime": 2781.5915, "eval_samples_per_second": 4.581, "eval_steps_per_second": 0.287, "eval_wer": 7.441964983586057, "step": 9000 }, { "epoch": 90.59, "learning_rate": 7.842025316455697e-06, "loss": 0.0004, "step": 9025 }, { "epoch": 90.84, "learning_rate": 7.835696202531647e-06, "loss": 0.0005, "step": 9050 }, { "epoch": 91.09, "learning_rate": 7.829367088607596e-06, "loss": 0.001, "step": 9075 }, { "epoch": 91.34, "learning_rate": 7.823037974683546e-06, "loss": 0.0009, "step": 9100 }, { "epoch": 91.59, "learning_rate": 7.816708860759495e-06, "loss": 0.0006, "step": 9125 }, { "epoch": 91.84, "learning_rate": 7.810379746835443e-06, "loss": 0.0007, "step": 9150 }, { "epoch": 92.1, "learning_rate": 7.804050632911392e-06, "loss": 0.0006, "step": 9175 }, { "epoch": 92.35, "learning_rate": 7.797721518987342e-06, "loss": 0.0006, "step": 9200 }, { "epoch": 92.6, "learning_rate": 7.791392405063291e-06, "loss": 0.0007, "step": 9225 }, { "epoch": 92.85, "learning_rate": 7.78506329113924e-06, "loss": 0.0009, "step": 9250 }, { "epoch": 93.1, "learning_rate": 7.77873417721519e-06, "loss": 0.0005, "step": 9275 }, { "epoch": 93.35, "learning_rate": 7.77240506329114e-06, "loss": 0.0003, "step": 9300 }, { "epoch": 93.6, "learning_rate": 7.766075949367089e-06, "loss": 0.0003, "step": 9325 }, { "epoch": 93.85, "learning_rate": 7.759746835443038e-06, "loss": 0.0005, "step": 9350 }, { "epoch": 94.1, "learning_rate": 7.753417721518988e-06, "loss": 0.0005, "step": 9375 }, { "epoch": 94.35, "learning_rate": 7.747088607594937e-06, "loss": 0.0004, "step": 9400 }, { "epoch": 94.6, "learning_rate": 7.740759493670887e-06, "loss": 0.0005, "step": 9425 }, { "epoch": 94.86, "learning_rate": 7.734430379746836e-06, "loss": 0.0004, "step": 9450 }, { "epoch": 95.11, "learning_rate": 7.728101265822785e-06, "loss": 0.0004, "step": 9475 }, { "epoch": 95.36, "learning_rate": 7.721772151898735e-06, "loss": 0.0004, "step": 9500 }, { "epoch": 95.61, "learning_rate": 7.715443037974684e-06, "loss": 0.0003, "step": 9525 }, { "epoch": 95.86, "learning_rate": 7.709113924050634e-06, "loss": 0.0004, "step": 9550 }, { "epoch": 96.11, "learning_rate": 7.702784810126583e-06, "loss": 0.0002, "step": 9575 }, { "epoch": 96.36, "learning_rate": 7.696455696202532e-06, "loss": 0.0003, "step": 9600 }, { "epoch": 96.61, "learning_rate": 7.690126582278482e-06, "loss": 0.0007, "step": 9625 }, { "epoch": 96.86, "learning_rate": 7.683797468354431e-06, "loss": 0.0005, "step": 9650 }, { "epoch": 97.11, "learning_rate": 7.67746835443038e-06, "loss": 0.0005, "step": 9675 }, { "epoch": 97.37, "learning_rate": 7.67113924050633e-06, "loss": 0.0006, "step": 9700 }, { "epoch": 97.62, "learning_rate": 7.66481012658228e-06, "loss": 0.0009, "step": 9725 }, { "epoch": 97.87, "learning_rate": 7.658481012658229e-06, "loss": 0.0006, "step": 9750 }, { "epoch": 98.12, "learning_rate": 7.652151898734178e-06, "loss": 0.0009, "step": 9775 }, { "epoch": 98.37, "learning_rate": 7.645822784810128e-06, "loss": 0.0008, "step": 9800 }, { "epoch": 98.62, "learning_rate": 7.639493670886077e-06, "loss": 0.0012, "step": 9825 }, { "epoch": 98.87, "learning_rate": 7.633164556962027e-06, "loss": 0.0015, "step": 9850 }, { "epoch": 99.12, "learning_rate": 7.626835443037975e-06, "loss": 0.0012, "step": 9875 }, { "epoch": 99.37, "learning_rate": 7.620506329113925e-06, "loss": 0.0008, "step": 9900 }, { "epoch": 99.62, "learning_rate": 7.614177215189874e-06, "loss": 0.0007, "step": 9925 }, { "epoch": 99.87, "learning_rate": 7.607848101265824e-06, "loss": 0.0008, "step": 9950 }, { "epoch": 100.13, "learning_rate": 7.601518987341773e-06, "loss": 0.0008, "step": 9975 }, { "epoch": 100.38, "learning_rate": 7.5951898734177225e-06, "loss": 0.0007, "step": 10000 }, { "epoch": 100.38, "eval_loss": 0.23848006129264832, "eval_runtime": 2777.6649, "eval_samples_per_second": 4.587, "eval_steps_per_second": 0.287, "eval_wer": 7.531850867594185, "step": 10000 }, { "epoch": 100.63, "learning_rate": 7.588860759493671e-06, "loss": 0.0008, "step": 10025 }, { "epoch": 100.88, "learning_rate": 7.5825316455696205e-06, "loss": 0.0007, "step": 10050 }, { "epoch": 101.13, "learning_rate": 7.57620253164557e-06, "loss": 0.0006, "step": 10075 }, { "epoch": 101.38, "learning_rate": 7.569873417721519e-06, "loss": 0.0005, "step": 10100 }, { "epoch": 101.63, "learning_rate": 7.563544303797469e-06, "loss": 0.0009, "step": 10125 }, { "epoch": 101.88, "learning_rate": 7.557215189873418e-06, "loss": 0.0006, "step": 10150 }, { "epoch": 102.13, "learning_rate": 7.550886075949368e-06, "loss": 0.0008, "step": 10175 }, { "epoch": 102.38, "learning_rate": 7.544556962025317e-06, "loss": 0.0007, "step": 10200 }, { "epoch": 102.63, "learning_rate": 7.538227848101266e-06, "loss": 0.0006, "step": 10225 }, { "epoch": 102.89, "learning_rate": 7.531898734177215e-06, "loss": 0.0007, "step": 10250 }, { "epoch": 103.14, "learning_rate": 7.5255696202531645e-06, "loss": 0.001, "step": 10275 }, { "epoch": 103.39, "learning_rate": 7.519240506329114e-06, "loss": 0.0007, "step": 10300 }, { "epoch": 103.64, "learning_rate": 7.512911392405063e-06, "loss": 0.0008, "step": 10325 }, { "epoch": 103.89, "learning_rate": 7.506582278481013e-06, "loss": 0.0012, "step": 10350 }, { "epoch": 104.14, "learning_rate": 7.500253164556963e-06, "loss": 0.0009, "step": 10375 }, { "epoch": 104.39, "learning_rate": 7.4939240506329125e-06, "loss": 0.0008, "step": 10400 }, { "epoch": 104.64, "learning_rate": 7.487594936708862e-06, "loss": 0.0005, "step": 10425 }, { "epoch": 104.89, "learning_rate": 7.481265822784811e-06, "loss": 0.0007, "step": 10450 }, { "epoch": 105.14, "learning_rate": 7.474936708860761e-06, "loss": 0.0007, "step": 10475 }, { "epoch": 105.4, "learning_rate": 7.46860759493671e-06, "loss": 0.0007, "step": 10500 }, { "epoch": 105.65, "learning_rate": 7.462278481012659e-06, "loss": 0.0006, "step": 10525 }, { "epoch": 105.9, "learning_rate": 7.455949367088608e-06, "loss": 0.0005, "step": 10550 }, { "epoch": 106.15, "learning_rate": 7.449620253164558e-06, "loss": 0.0002, "step": 10575 }, { "epoch": 106.4, "learning_rate": 7.443291139240507e-06, "loss": 0.0002, "step": 10600 }, { "epoch": 106.65, "learning_rate": 7.4369620253164565e-06, "loss": 0.0003, "step": 10625 }, { "epoch": 106.9, "learning_rate": 7.430632911392406e-06, "loss": 0.0003, "step": 10650 }, { "epoch": 107.15, "learning_rate": 7.424303797468355e-06, "loss": 0.0003, "step": 10675 }, { "epoch": 107.4, "learning_rate": 7.417974683544305e-06, "loss": 0.0003, "step": 10700 }, { "epoch": 107.65, "learning_rate": 7.411645569620254e-06, "loss": 0.0003, "step": 10725 }, { "epoch": 107.9, "learning_rate": 7.405316455696203e-06, "loss": 0.0007, "step": 10750 }, { "epoch": 108.16, "learning_rate": 7.398987341772152e-06, "loss": 0.0004, "step": 10775 }, { "epoch": 108.41, "learning_rate": 7.392658227848102e-06, "loss": 0.0002, "step": 10800 }, { "epoch": 108.66, "learning_rate": 7.386329113924051e-06, "loss": 0.0002, "step": 10825 }, { "epoch": 108.91, "learning_rate": 7.3800000000000005e-06, "loss": 0.0005, "step": 10850 }, { "epoch": 109.16, "learning_rate": 7.37367088607595e-06, "loss": 0.0007, "step": 10875 }, { "epoch": 109.41, "learning_rate": 7.367341772151899e-06, "loss": 0.0009, "step": 10900 }, { "epoch": 109.66, "learning_rate": 7.361012658227849e-06, "loss": 0.0011, "step": 10925 }, { "epoch": 109.91, "learning_rate": 7.354683544303798e-06, "loss": 0.0013, "step": 10950 }, { "epoch": 110.16, "learning_rate": 7.348354430379747e-06, "loss": 0.0009, "step": 10975 }, { "epoch": 110.41, "learning_rate": 7.342025316455696e-06, "loss": 0.001, "step": 11000 }, { "epoch": 110.41, "eval_loss": 0.22928227484226227, "eval_runtime": 2789.8685, "eval_samples_per_second": 4.567, "eval_steps_per_second": 0.286, "eval_wer": 7.632483976864155, "step": 11000 }, { "epoch": 110.66, "learning_rate": 7.335696202531646e-06, "loss": 0.0007, "step": 11025 }, { "epoch": 110.92, "learning_rate": 7.329367088607595e-06, "loss": 0.001, "step": 11050 }, { "epoch": 111.17, "learning_rate": 7.3230379746835445e-06, "loss": 0.001, "step": 11075 }, { "epoch": 111.42, "learning_rate": 7.316708860759494e-06, "loss": 0.001, "step": 11100 }, { "epoch": 111.67, "learning_rate": 7.310379746835443e-06, "loss": 0.0009, "step": 11125 }, { "epoch": 111.92, "learning_rate": 7.304050632911393e-06, "loss": 0.0011, "step": 11150 }, { "epoch": 112.17, "learning_rate": 7.297721518987342e-06, "loss": 0.0007, "step": 11175 }, { "epoch": 112.42, "learning_rate": 7.2913924050632925e-06, "loss": 0.0007, "step": 11200 }, { "epoch": 112.67, "learning_rate": 7.285063291139242e-06, "loss": 0.0007, "step": 11225 }, { "epoch": 112.92, "learning_rate": 7.278734177215191e-06, "loss": 0.0008, "step": 11250 }, { "epoch": 113.17, "learning_rate": 7.27240506329114e-06, "loss": 0.0008, "step": 11275 }, { "epoch": 113.43, "learning_rate": 7.266075949367089e-06, "loss": 0.0006, "step": 11300 }, { "epoch": 113.68, "learning_rate": 7.259746835443039e-06, "loss": 0.0006, "step": 11325 }, { "epoch": 113.93, "learning_rate": 7.253417721518988e-06, "loss": 0.0007, "step": 11350 }, { "epoch": 114.18, "learning_rate": 7.247088607594938e-06, "loss": 0.0004, "step": 11375 }, { "epoch": 114.43, "learning_rate": 7.240759493670887e-06, "loss": 0.0004, "step": 11400 }, { "epoch": 114.68, "learning_rate": 7.2344303797468365e-06, "loss": 0.0004, "step": 11425 }, { "epoch": 114.93, "learning_rate": 7.228101265822786e-06, "loss": 0.0008, "step": 11450 }, { "epoch": 115.18, "learning_rate": 7.221772151898735e-06, "loss": 0.001, "step": 11475 }, { "epoch": 115.43, "learning_rate": 7.215443037974684e-06, "loss": 0.0008, "step": 11500 }, { "epoch": 115.68, "learning_rate": 7.209113924050633e-06, "loss": 0.0007, "step": 11525 }, { "epoch": 115.93, "learning_rate": 7.202784810126583e-06, "loss": 0.0007, "step": 11550 }, { "epoch": 116.19, "learning_rate": 7.196455696202532e-06, "loss": 0.0007, "step": 11575 }, { "epoch": 116.44, "learning_rate": 7.190126582278482e-06, "loss": 0.0009, "step": 11600 }, { "epoch": 116.69, "learning_rate": 7.183797468354431e-06, "loss": 0.001, "step": 11625 }, { "epoch": 116.94, "learning_rate": 7.1774683544303805e-06, "loss": 0.0006, "step": 11650 }, { "epoch": 117.19, "learning_rate": 7.17113924050633e-06, "loss": 0.0008, "step": 11675 }, { "epoch": 117.44, "learning_rate": 7.164810126582279e-06, "loss": 0.0008, "step": 11700 }, { "epoch": 117.69, "learning_rate": 7.158481012658228e-06, "loss": 0.0005, "step": 11725 }, { "epoch": 117.94, "learning_rate": 7.152151898734177e-06, "loss": 0.0004, "step": 11750 }, { "epoch": 118.19, "learning_rate": 7.145822784810127e-06, "loss": 0.0003, "step": 11775 }, { "epoch": 118.44, "learning_rate": 7.139493670886076e-06, "loss": 0.0002, "step": 11800 }, { "epoch": 118.7, "learning_rate": 7.133164556962026e-06, "loss": 0.0002, "step": 11825 }, { "epoch": 118.95, "learning_rate": 7.126835443037975e-06, "loss": 0.0003, "step": 11850 }, { "epoch": 119.2, "learning_rate": 7.1205063291139245e-06, "loss": 0.0008, "step": 11875 }, { "epoch": 119.45, "learning_rate": 7.114177215189874e-06, "loss": 0.0003, "step": 11900 }, { "epoch": 119.7, "learning_rate": 7.107848101265823e-06, "loss": 0.0001, "step": 11925 }, { "epoch": 119.95, "learning_rate": 7.101518987341772e-06, "loss": 0.0003, "step": 11950 }, { "epoch": 120.2, "learning_rate": 7.095189873417721e-06, "loss": 0.0003, "step": 11975 }, { "epoch": 120.45, "learning_rate": 7.088860759493671e-06, "loss": 0.0001, "step": 12000 }, { "epoch": 120.45, "eval_loss": 0.2473139613866806, "eval_runtime": 2824.4756, "eval_samples_per_second": 4.511, "eval_steps_per_second": 0.282, "eval_wer": 7.142996717211194, "step": 12000 }, { "epoch": 120.7, "learning_rate": 7.082531645569621e-06, "loss": 0.0003, "step": 12025 }, { "epoch": 120.95, "learning_rate": 7.0762025316455705e-06, "loss": 0.0003, "step": 12050 }, { "epoch": 121.2, "learning_rate": 7.06987341772152e-06, "loss": 0.0002, "step": 12075 }, { "epoch": 121.46, "learning_rate": 7.063544303797469e-06, "loss": 0.0002, "step": 12100 }, { "epoch": 121.71, "learning_rate": 7.057215189873419e-06, "loss": 0.0001, "step": 12125 }, { "epoch": 121.96, "learning_rate": 7.050886075949368e-06, "loss": 0.0001, "step": 12150 }, { "epoch": 122.21, "learning_rate": 7.044556962025318e-06, "loss": 0.0003, "step": 12175 }, { "epoch": 122.46, "learning_rate": 7.038227848101267e-06, "loss": 0.0003, "step": 12200 }, { "epoch": 122.71, "learning_rate": 7.031898734177216e-06, "loss": 0.0004, "step": 12225 }, { "epoch": 122.96, "learning_rate": 7.025569620253165e-06, "loss": 0.0009, "step": 12250 }, { "epoch": 123.21, "learning_rate": 7.0192405063291145e-06, "loss": 0.0007, "step": 12275 }, { "epoch": 123.46, "learning_rate": 7.012911392405064e-06, "loss": 0.0007, "step": 12300 }, { "epoch": 123.71, "learning_rate": 7.006582278481013e-06, "loss": 0.0005, "step": 12325 }, { "epoch": 123.96, "learning_rate": 7.000253164556963e-06, "loss": 0.0007, "step": 12350 }, { "epoch": 124.22, "learning_rate": 6.993924050632912e-06, "loss": 0.0008, "step": 12375 }, { "epoch": 124.47, "learning_rate": 6.987594936708862e-06, "loss": 0.0007, "step": 12400 }, { "epoch": 124.72, "learning_rate": 6.981265822784811e-06, "loss": 0.0007, "step": 12425 }, { "epoch": 124.97, "learning_rate": 6.97493670886076e-06, "loss": 0.0007, "step": 12450 }, { "epoch": 125.22, "learning_rate": 6.968607594936709e-06, "loss": 0.0006, "step": 12475 }, { "epoch": 125.47, "learning_rate": 6.9622784810126585e-06, "loss": 0.0004, "step": 12500 }, { "epoch": 125.72, "learning_rate": 6.955949367088608e-06, "loss": 0.0005, "step": 12525 }, { "epoch": 125.97, "learning_rate": 6.949620253164557e-06, "loss": 0.0011, "step": 12550 }, { "epoch": 126.22, "learning_rate": 6.943291139240507e-06, "loss": 0.001, "step": 12575 }, { "epoch": 126.47, "learning_rate": 6.936962025316456e-06, "loss": 0.0009, "step": 12600 }, { "epoch": 126.73, "learning_rate": 6.930632911392406e-06, "loss": 0.0008, "step": 12625 }, { "epoch": 126.98, "learning_rate": 6.924303797468355e-06, "loss": 0.0011, "step": 12650 }, { "epoch": 127.23, "learning_rate": 6.917974683544304e-06, "loss": 0.0006, "step": 12675 }, { "epoch": 127.48, "learning_rate": 6.911645569620253e-06, "loss": 0.0005, "step": 12700 }, { "epoch": 127.73, "learning_rate": 6.9053164556962025e-06, "loss": 0.0008, "step": 12725 }, { "epoch": 127.98, "learning_rate": 6.898987341772152e-06, "loss": 0.0006, "step": 12750 }, { "epoch": 128.23, "learning_rate": 6.892658227848101e-06, "loss": 0.0004, "step": 12775 }, { "epoch": 128.48, "learning_rate": 6.886329113924051e-06, "loss": 0.0005, "step": 12800 }, { "epoch": 128.73, "learning_rate": 6.88e-06, "loss": 0.0004, "step": 12825 }, { "epoch": 128.98, "learning_rate": 6.8736708860759505e-06, "loss": 0.0003, "step": 12850 }, { "epoch": 129.23, "learning_rate": 6.8673417721519e-06, "loss": 0.0001, "step": 12875 }, { "epoch": 129.49, "learning_rate": 6.861012658227849e-06, "loss": 0.0002, "step": 12900 }, { "epoch": 129.74, "learning_rate": 6.854683544303799e-06, "loss": 0.0003, "step": 12925 }, { "epoch": 129.99, "learning_rate": 6.848354430379748e-06, "loss": 0.0002, "step": 12950 }, { "epoch": 130.24, "learning_rate": 6.842025316455697e-06, "loss": 0.0002, "step": 12975 }, { "epoch": 130.49, "learning_rate": 6.835696202531646e-06, "loss": 0.0001, "step": 13000 }, { "epoch": 130.49, "eval_loss": 0.24884945154190063, "eval_runtime": 2827.5774, "eval_samples_per_second": 4.506, "eval_steps_per_second": 0.282, "eval_wer": 7.186962638736907, "step": 13000 }, { "epoch": 130.74, "learning_rate": 6.829367088607596e-06, "loss": 0.0001, "step": 13025 }, { "epoch": 130.99, "learning_rate": 6.823037974683545e-06, "loss": 0.0001, "step": 13050 }, { "epoch": 131.24, "learning_rate": 6.8167088607594945e-06, "loss": 0.0001, "step": 13075 }, { "epoch": 131.49, "learning_rate": 6.810379746835444e-06, "loss": 0.0002, "step": 13100 }, { "epoch": 131.74, "learning_rate": 6.804050632911393e-06, "loss": 0.0003, "step": 13125 }, { "epoch": 131.99, "learning_rate": 6.797974683544304e-06, "loss": 0.0007, "step": 13150 }, { "epoch": 132.25, "learning_rate": 6.791645569620253e-06, "loss": 0.0003, "step": 13175 }, { "epoch": 132.5, "learning_rate": 6.785316455696203e-06, "loss": 0.0002, "step": 13200 }, { "epoch": 132.75, "learning_rate": 6.778987341772153e-06, "loss": 0.0005, "step": 13225 }, { "epoch": 133.0, "learning_rate": 6.7726582278481025e-06, "loss": 0.0005, "step": 13250 }, { "epoch": 133.25, "learning_rate": 6.766329113924052e-06, "loss": 0.0006, "step": 13275 }, { "epoch": 133.5, "learning_rate": 6.760000000000001e-06, "loss": 0.0012, "step": 13300 }, { "epoch": 133.75, "learning_rate": 6.75367088607595e-06, "loss": 0.0011, "step": 13325 }, { "epoch": 134.0, "learning_rate": 6.747594936708861e-06, "loss": 0.0009, "step": 13350 }, { "epoch": 134.25, "learning_rate": 6.741265822784811e-06, "loss": 0.0008, "step": 13375 }, { "epoch": 134.5, "learning_rate": 6.734936708860759e-06, "loss": 0.0007, "step": 13400 }, { "epoch": 134.76, "learning_rate": 6.728607594936709e-06, "loss": 0.0008, "step": 13425 }, { "epoch": 135.01, "learning_rate": 6.722278481012658e-06, "loss": 0.0011, "step": 13450 }, { "epoch": 135.26, "learning_rate": 6.715949367088608e-06, "loss": 0.0015, "step": 13475 }, { "epoch": 135.51, "learning_rate": 6.709620253164557e-06, "loss": 0.0014, "step": 13500 }, { "epoch": 135.76, "learning_rate": 6.7032911392405065e-06, "loss": 0.0012, "step": 13525 }, { "epoch": 136.01, "learning_rate": 6.696962025316456e-06, "loss": 0.0012, "step": 13550 }, { "epoch": 136.26, "learning_rate": 6.690632911392405e-06, "loss": 0.001, "step": 13575 }, { "epoch": 136.51, "learning_rate": 6.684303797468355e-06, "loss": 0.0008, "step": 13600 }, { "epoch": 136.76, "learning_rate": 6.677974683544305e-06, "loss": 0.0007, "step": 13625 }, { "epoch": 137.01, "learning_rate": 6.6716455696202545e-06, "loss": 0.0007, "step": 13650 }, { "epoch": 137.26, "learning_rate": 6.665316455696204e-06, "loss": 0.0004, "step": 13675 }, { "epoch": 137.52, "learning_rate": 6.6589873417721525e-06, "loss": 0.0005, "step": 13700 }, { "epoch": 137.77, "learning_rate": 6.652658227848102e-06, "loss": 0.0004, "step": 13725 }, { "epoch": 138.02, "learning_rate": 6.646329113924051e-06, "loss": 0.0004, "step": 13750 }, { "epoch": 138.27, "learning_rate": 6.640000000000001e-06, "loss": 0.0002, "step": 13775 }, { "epoch": 138.52, "learning_rate": 6.63367088607595e-06, "loss": 0.0002, "step": 13800 }, { "epoch": 138.77, "learning_rate": 6.6273417721519e-06, "loss": 0.0002, "step": 13825 }, { "epoch": 139.02, "learning_rate": 6.621012658227849e-06, "loss": 0.0003, "step": 13850 }, { "epoch": 139.27, "learning_rate": 6.6146835443037985e-06, "loss": 0.0005, "step": 13875 }, { "epoch": 139.52, "learning_rate": 6.608354430379748e-06, "loss": 0.0002, "step": 13900 }, { "epoch": 139.77, "learning_rate": 6.6020253164556965e-06, "loss": 0.0003, "step": 13925 }, { "epoch": 140.03, "learning_rate": 6.595696202531646e-06, "loss": 0.0002, "step": 13950 }, { "epoch": 140.28, "learning_rate": 6.589367088607595e-06, "loss": 0.0001, "step": 13975 }, { "epoch": 140.53, "learning_rate": 6.583037974683545e-06, "loss": 0.0004, "step": 14000 }, { "epoch": 140.53, "eval_loss": 0.23976704478263855, "eval_runtime": 2824.3974, "eval_samples_per_second": 4.511, "eval_steps_per_second": 0.282, "eval_wer": 7.183054556823511, "step": 14000 }, { "epoch": 140.78, "learning_rate": 6.576708860759494e-06, "loss": 0.0003, "step": 14025 }, { "epoch": 141.03, "learning_rate": 6.570379746835444e-06, "loss": 0.0004, "step": 14050 }, { "epoch": 141.28, "learning_rate": 6.564050632911393e-06, "loss": 0.0002, "step": 14075 }, { "epoch": 141.53, "learning_rate": 6.5577215189873425e-06, "loss": 0.0001, "step": 14100 }, { "epoch": 141.78, "learning_rate": 6.551392405063292e-06, "loss": 0.0002, "step": 14125 }, { "epoch": 142.03, "learning_rate": 6.5450632911392405e-06, "loss": 0.0003, "step": 14150 }, { "epoch": 142.28, "learning_rate": 6.53873417721519e-06, "loss": 0.0005, "step": 14175 }, { "epoch": 142.53, "learning_rate": 6.532405063291139e-06, "loss": 0.0003, "step": 14200 }, { "epoch": 142.79, "learning_rate": 6.526075949367089e-06, "loss": 0.0004, "step": 14225 }, { "epoch": 143.04, "learning_rate": 6.519746835443038e-06, "loss": 0.0003, "step": 14250 }, { "epoch": 143.29, "learning_rate": 6.513417721518988e-06, "loss": 0.0002, "step": 14275 }, { "epoch": 143.54, "learning_rate": 6.507088607594937e-06, "loss": 0.0003, "step": 14300 }, { "epoch": 143.79, "learning_rate": 6.5007594936708865e-06, "loss": 0.0003, "step": 14325 }, { "epoch": 144.04, "learning_rate": 6.494430379746836e-06, "loss": 0.0003, "step": 14350 }, { "epoch": 144.29, "learning_rate": 6.4881012658227845e-06, "loss": 0.0004, "step": 14375 }, { "epoch": 144.54, "learning_rate": 6.481772151898734e-06, "loss": 0.0003, "step": 14400 }, { "epoch": 144.79, "learning_rate": 6.475443037974683e-06, "loss": 0.0004, "step": 14425 }, { "epoch": 145.04, "learning_rate": 6.469113924050634e-06, "loss": 0.0004, "step": 14450 }, { "epoch": 145.29, "learning_rate": 6.462784810126583e-06, "loss": 0.0002, "step": 14475 }, { "epoch": 145.55, "learning_rate": 6.4564556962025325e-06, "loss": 0.0002, "step": 14500 }, { "epoch": 145.8, "learning_rate": 6.450126582278482e-06, "loss": 0.0002, "step": 14525 }, { "epoch": 146.05, "learning_rate": 6.443797468354431e-06, "loss": 0.0003, "step": 14550 }, { "epoch": 146.3, "learning_rate": 6.437468354430381e-06, "loss": 0.0003, "step": 14575 }, { "epoch": 146.55, "learning_rate": 6.43113924050633e-06, "loss": 0.0002, "step": 14600 }, { "epoch": 146.8, "learning_rate": 6.42481012658228e-06, "loss": 0.0002, "step": 14625 }, { "epoch": 147.05, "learning_rate": 6.418481012658228e-06, "loss": 0.0001, "step": 14650 }, { "epoch": 147.3, "learning_rate": 6.412151898734178e-06, "loss": 0.0001, "step": 14675 }, { "epoch": 147.55, "learning_rate": 6.405822784810127e-06, "loss": 0.0001, "step": 14700 }, { "epoch": 147.8, "learning_rate": 6.3994936708860765e-06, "loss": 0.0003, "step": 14725 }, { "epoch": 148.06, "learning_rate": 6.393164556962026e-06, "loss": 0.0002, "step": 14750 }, { "epoch": 148.31, "learning_rate": 6.386835443037975e-06, "loss": 0.0001, "step": 14775 }, { "epoch": 148.56, "learning_rate": 6.380506329113925e-06, "loss": 0.0001, "step": 14800 }, { "epoch": 148.81, "learning_rate": 6.374177215189874e-06, "loss": 0.0, "step": 14825 }, { "epoch": 149.06, "learning_rate": 6.367848101265824e-06, "loss": 0.0001, "step": 14850 }, { "epoch": 149.31, "learning_rate": 6.361518987341772e-06, "loss": 0.0, "step": 14875 }, { "epoch": 149.56, "learning_rate": 6.355189873417722e-06, "loss": 0.0, "step": 14900 }, { "epoch": 149.81, "learning_rate": 6.348860759493671e-06, "loss": 0.0001, "step": 14925 }, { "epoch": 150.06, "learning_rate": 6.3425316455696205e-06, "loss": 0.0001, "step": 14950 }, { "epoch": 150.31, "learning_rate": 6.33620253164557e-06, "loss": 0.0001, "step": 14975 }, { "epoch": 150.56, "learning_rate": 6.329873417721519e-06, "loss": 0.0, "step": 15000 }, { "epoch": 150.56, "eval_loss": 0.2619960606098175, "eval_runtime": 2808.3991, "eval_samples_per_second": 4.537, "eval_steps_per_second": 0.284, "eval_wer": 7.058972956073159, "step": 15000 }, { "epoch": 150.82, "learning_rate": 6.323544303797469e-06, "loss": 0.0002, "step": 15025 }, { "epoch": 151.07, "learning_rate": 6.317215189873418e-06, "loss": 0.0, "step": 15050 }, { "epoch": 151.32, "learning_rate": 6.310886075949368e-06, "loss": 0.0, "step": 15075 }, { "epoch": 151.57, "learning_rate": 6.304556962025316e-06, "loss": 0.0001, "step": 15100 }, { "epoch": 151.82, "learning_rate": 6.298227848101266e-06, "loss": 0.0, "step": 15125 }, { "epoch": 152.07, "learning_rate": 6.291898734177215e-06, "loss": 0.0, "step": 15150 }, { "epoch": 152.32, "learning_rate": 6.2855696202531645e-06, "loss": 0.0, "step": 15175 }, { "epoch": 152.57, "learning_rate": 6.279240506329114e-06, "loss": 0.0003, "step": 15200 }, { "epoch": 152.82, "learning_rate": 6.272911392405063e-06, "loss": 0.0002, "step": 15225 }, { "epoch": 153.07, "learning_rate": 6.266582278481013e-06, "loss": 0.001, "step": 15250 }, { "epoch": 153.32, "learning_rate": 6.260253164556963e-06, "loss": 0.0012, "step": 15275 }, { "epoch": 153.58, "learning_rate": 6.2539240506329125e-06, "loss": 0.0017, "step": 15300 }, { "epoch": 153.83, "learning_rate": 6.247594936708862e-06, "loss": 0.0018, "step": 15325 }, { "epoch": 154.08, "learning_rate": 6.241265822784811e-06, "loss": 0.0011, "step": 15350 }, { "epoch": 154.33, "learning_rate": 6.234936708860761e-06, "loss": 0.0011, "step": 15375 }, { "epoch": 154.58, "learning_rate": 6.228607594936709e-06, "loss": 0.0019, "step": 15400 }, { "epoch": 154.83, "learning_rate": 6.222278481012659e-06, "loss": 0.0018, "step": 15425 }, { "epoch": 155.08, "learning_rate": 6.215949367088608e-06, "loss": 0.0016, "step": 15450 }, { "epoch": 155.33, "learning_rate": 6.209620253164558e-06, "loss": 0.001, "step": 15475 }, { "epoch": 155.58, "learning_rate": 6.203291139240507e-06, "loss": 0.0011, "step": 15500 }, { "epoch": 155.83, "learning_rate": 6.1969620253164565e-06, "loss": 0.001, "step": 15525 }, { "epoch": 156.09, "learning_rate": 6.190632911392406e-06, "loss": 0.001, "step": 15550 }, { "epoch": 156.34, "learning_rate": 6.184303797468355e-06, "loss": 0.0007, "step": 15575 }, { "epoch": 156.59, "learning_rate": 6.177974683544305e-06, "loss": 0.0008, "step": 15600 }, { "epoch": 156.84, "learning_rate": 6.171898734177215e-06, "loss": 0.0008, "step": 15625 }, { "epoch": 157.09, "learning_rate": 6.165569620253165e-06, "loss": 0.0006, "step": 15650 }, { "epoch": 157.34, "learning_rate": 6.159240506329115e-06, "loss": 0.0002, "step": 15675 }, { "epoch": 157.59, "learning_rate": 6.1529113924050645e-06, "loss": 0.0002, "step": 15700 }, { "epoch": 157.84, "learning_rate": 6.146582278481014e-06, "loss": 0.0003, "step": 15725 }, { "epoch": 158.09, "learning_rate": 6.140253164556963e-06, "loss": 0.0004, "step": 15750 }, { "epoch": 158.34, "learning_rate": 6.133924050632912e-06, "loss": 0.0002, "step": 15775 }, { "epoch": 158.59, "learning_rate": 6.127594936708861e-06, "loss": 0.0002, "step": 15800 }, { "epoch": 158.85, "learning_rate": 6.121265822784811e-06, "loss": 0.0003, "step": 15825 }, { "epoch": 159.1, "learning_rate": 6.11493670886076e-06, "loss": 0.0003, "step": 15850 }, { "epoch": 159.35, "learning_rate": 6.10860759493671e-06, "loss": 0.0002, "step": 15875 }, { "epoch": 159.6, "learning_rate": 6.102278481012659e-06, "loss": 0.0002, "step": 15900 }, { "epoch": 159.85, "learning_rate": 6.0959493670886085e-06, "loss": 0.0001, "step": 15925 }, { "epoch": 160.1, "learning_rate": 6.089620253164558e-06, "loss": 0.0001, "step": 15950 }, { "epoch": 160.35, "learning_rate": 6.083291139240507e-06, "loss": 0.0001, "step": 15975 }, { "epoch": 160.6, "learning_rate": 6.076962025316456e-06, "loss": 0.0001, "step": 16000 }, { "epoch": 160.6, "eval_loss": 0.2546827495098114, "eval_runtime": 2784.4693, "eval_samples_per_second": 4.576, "eval_steps_per_second": 0.286, "eval_wer": 7.196732843520401, "step": 16000 }, { "epoch": 160.85, "learning_rate": 6.070632911392405e-06, "loss": 0.0005, "step": 16025 }, { "epoch": 161.1, "learning_rate": 6.064303797468355e-06, "loss": 0.0005, "step": 16050 }, { "epoch": 161.36, "learning_rate": 6.057974683544304e-06, "loss": 0.0004, "step": 16075 }, { "epoch": 161.61, "learning_rate": 6.051645569620254e-06, "loss": 0.0001, "step": 16100 }, { "epoch": 161.86, "learning_rate": 6.045316455696203e-06, "loss": 0.0001, "step": 16125 }, { "epoch": 162.11, "learning_rate": 6.0389873417721525e-06, "loss": 0.0001, "step": 16150 }, { "epoch": 162.36, "learning_rate": 6.032658227848102e-06, "loss": 0.0001, "step": 16175 }, { "epoch": 162.61, "learning_rate": 6.0263291139240505e-06, "loss": 0.0001, "step": 16200 }, { "epoch": 162.86, "learning_rate": 6.02e-06, "loss": 0.0001, "step": 16225 }, { "epoch": 163.11, "learning_rate": 6.013670886075949e-06, "loss": 0.0001, "step": 16250 }, { "epoch": 163.36, "learning_rate": 6.007341772151899e-06, "loss": 0.0, "step": 16275 }, { "epoch": 163.61, "learning_rate": 6.001012658227848e-06, "loss": 0.0, "step": 16300 }, { "epoch": 163.86, "learning_rate": 5.994683544303798e-06, "loss": 0.0001, "step": 16325 }, { "epoch": 164.12, "learning_rate": 5.988354430379747e-06, "loss": 0.0, "step": 16350 }, { "epoch": 164.37, "learning_rate": 5.9820253164556965e-06, "loss": 0.0, "step": 16375 }, { "epoch": 164.62, "learning_rate": 5.975696202531646e-06, "loss": 0.0002, "step": 16400 }, { "epoch": 164.87, "learning_rate": 5.9693670886075945e-06, "loss": 0.0, "step": 16425 }, { "epoch": 165.12, "learning_rate": 5.963037974683544e-06, "loss": 0.0, "step": 16450 }, { "epoch": 165.37, "learning_rate": 5.956708860759493e-06, "loss": 0.0, "step": 16475 }, { "epoch": 165.62, "learning_rate": 5.950379746835444e-06, "loss": 0.0, "step": 16500 }, { "epoch": 165.87, "learning_rate": 5.944050632911393e-06, "loss": 0.0, "step": 16525 }, { "epoch": 166.12, "learning_rate": 5.9377215189873425e-06, "loss": 0.0, "step": 16550 }, { "epoch": 166.37, "learning_rate": 5.931392405063292e-06, "loss": 0.0, "step": 16575 }, { "epoch": 166.62, "learning_rate": 5.925063291139241e-06, "loss": 0.0, "step": 16600 }, { "epoch": 166.88, "learning_rate": 5.918734177215191e-06, "loss": 0.0, "step": 16625 }, { "epoch": 167.13, "learning_rate": 5.91240506329114e-06, "loss": 0.0, "step": 16650 }, { "epoch": 167.38, "learning_rate": 5.90607594936709e-06, "loss": 0.0, "step": 16675 }, { "epoch": 167.63, "learning_rate": 5.899746835443039e-06, "loss": 0.0, "step": 16700 }, { "epoch": 167.88, "learning_rate": 5.893417721518988e-06, "loss": 0.0, "step": 16725 }, { "epoch": 168.13, "learning_rate": 5.887088607594937e-06, "loss": 0.0, "step": 16750 }, { "epoch": 168.38, "learning_rate": 5.8807594936708865e-06, "loss": 0.0, "step": 16775 }, { "epoch": 168.63, "learning_rate": 5.874430379746836e-06, "loss": 0.0, "step": 16800 }, { "epoch": 168.88, "learning_rate": 5.868101265822785e-06, "loss": 0.0, "step": 16825 }, { "epoch": 169.13, "learning_rate": 5.861772151898735e-06, "loss": 0.0, "step": 16850 }, { "epoch": 169.39, "learning_rate": 5.855443037974684e-06, "loss": 0.0, "step": 16875 }, { "epoch": 169.64, "learning_rate": 5.849113924050634e-06, "loss": 0.0, "step": 16900 }, { "epoch": 169.89, "learning_rate": 5.842784810126583e-06, "loss": 0.0, "step": 16925 }, { "epoch": 170.14, "learning_rate": 5.836455696202532e-06, "loss": 0.0, "step": 16950 }, { "epoch": 170.39, "learning_rate": 5.830126582278481e-06, "loss": 0.0, "step": 16975 }, { "epoch": 170.64, "learning_rate": 5.8237974683544305e-06, "loss": 0.0, "step": 17000 }, { "epoch": 170.64, "eval_loss": 0.27676019072532654, "eval_runtime": 2784.6513, "eval_samples_per_second": 4.576, "eval_steps_per_second": 0.286, "eval_wer": 7.0736282632483976, "step": 17000 }, { "epoch": 170.89, "learning_rate": 5.81746835443038e-06, "loss": 0.0, "step": 17025 }, { "epoch": 171.14, "learning_rate": 5.811139240506329e-06, "loss": 0.0, "step": 17050 }, { "epoch": 171.39, "learning_rate": 5.804810126582279e-06, "loss": 0.0, "step": 17075 }, { "epoch": 171.64, "learning_rate": 5.798481012658228e-06, "loss": 0.0, "step": 17100 }, { "epoch": 171.89, "learning_rate": 5.792151898734178e-06, "loss": 0.0, "step": 17125 }, { "epoch": 172.15, "learning_rate": 5.785822784810127e-06, "loss": 0.0, "step": 17150 }, { "epoch": 172.4, "learning_rate": 5.779493670886076e-06, "loss": 0.0, "step": 17175 }, { "epoch": 172.65, "learning_rate": 5.773164556962025e-06, "loss": 0.0, "step": 17200 }, { "epoch": 172.9, "learning_rate": 5.7668354430379745e-06, "loss": 0.0, "step": 17225 }, { "epoch": 173.15, "learning_rate": 5.760506329113924e-06, "loss": 0.0, "step": 17250 }, { "epoch": 173.4, "learning_rate": 5.754177215189873e-06, "loss": 0.0, "step": 17275 }, { "epoch": 173.65, "learning_rate": 5.747848101265823e-06, "loss": 0.0, "step": 17300 }, { "epoch": 173.9, "learning_rate": 5.741518987341773e-06, "loss": 0.0, "step": 17325 }, { "epoch": 174.15, "learning_rate": 5.7351898734177225e-06, "loss": 0.0, "step": 17350 }, { "epoch": 174.4, "learning_rate": 5.728860759493672e-06, "loss": 0.0, "step": 17375 }, { "epoch": 174.65, "learning_rate": 5.722531645569621e-06, "loss": 0.0, "step": 17400 }, { "epoch": 174.91, "learning_rate": 5.716202531645571e-06, "loss": 0.0, "step": 17425 }, { "epoch": 175.16, "learning_rate": 5.70987341772152e-06, "loss": 0.0, "step": 17450 }, { "epoch": 175.41, "learning_rate": 5.703544303797469e-06, "loss": 0.0, "step": 17475 }, { "epoch": 175.66, "learning_rate": 5.697215189873418e-06, "loss": 0.0, "step": 17500 }, { "epoch": 175.91, "learning_rate": 5.690886075949368e-06, "loss": 0.0, "step": 17525 }, { "epoch": 176.16, "learning_rate": 5.684556962025317e-06, "loss": 0.0, "step": 17550 }, { "epoch": 176.41, "learning_rate": 5.6782278481012665e-06, "loss": 0.0, "step": 17575 }, { "epoch": 176.66, "learning_rate": 5.671898734177216e-06, "loss": 0.0, "step": 17600 }, { "epoch": 176.91, "learning_rate": 5.665569620253165e-06, "loss": 0.0, "step": 17625 }, { "epoch": 177.16, "learning_rate": 5.659240506329115e-06, "loss": 0.0, "step": 17650 }, { "epoch": 177.42, "learning_rate": 5.652911392405064e-06, "loss": 0.0, "step": 17675 }, { "epoch": 177.67, "learning_rate": 5.646582278481013e-06, "loss": 0.0, "step": 17700 }, { "epoch": 177.92, "learning_rate": 5.640253164556962e-06, "loss": 0.0, "step": 17725 }, { "epoch": 178.17, "learning_rate": 5.633924050632912e-06, "loss": 0.0, "step": 17750 }, { "epoch": 178.42, "learning_rate": 5.627594936708861e-06, "loss": 0.0, "step": 17775 }, { "epoch": 178.67, "learning_rate": 5.6212658227848105e-06, "loss": 0.0, "step": 17800 }, { "epoch": 178.92, "learning_rate": 5.61493670886076e-06, "loss": 0.0, "step": 17825 }, { "epoch": 179.17, "learning_rate": 5.608607594936709e-06, "loss": 0.0, "step": 17850 }, { "epoch": 179.42, "learning_rate": 5.602278481012659e-06, "loss": 0.0, "step": 17875 }, { "epoch": 179.67, "learning_rate": 5.595949367088608e-06, "loss": 0.0, "step": 17900 }, { "epoch": 179.92, "learning_rate": 5.589620253164557e-06, "loss": 0.0, "step": 17925 }, { "epoch": 180.18, "learning_rate": 5.583291139240506e-06, "loss": 0.0, "step": 17950 }, { "epoch": 180.43, "learning_rate": 5.576962025316456e-06, "loss": 0.0, "step": 17975 }, { "epoch": 180.68, "learning_rate": 5.570632911392405e-06, "loss": 0.0, "step": 18000 }, { "epoch": 180.68, "eval_loss": 0.28781425952911377, "eval_runtime": 2780.8898, "eval_samples_per_second": 4.582, "eval_steps_per_second": 0.287, "eval_wer": 7.000351727372206, "step": 18000 }, { "epoch": 180.93, "learning_rate": 5.5643037974683545e-06, "loss": 0.0, "step": 18025 }, { "epoch": 181.18, "learning_rate": 5.557974683544304e-06, "loss": 0.0, "step": 18050 }, { "epoch": 181.43, "learning_rate": 5.551645569620253e-06, "loss": 0.0, "step": 18075 }, { "epoch": 181.68, "learning_rate": 5.545316455696203e-06, "loss": 0.0, "step": 18100 }, { "epoch": 181.93, "learning_rate": 5.538987341772151e-06, "loss": 0.0, "step": 18125 }, { "epoch": 182.18, "learning_rate": 5.5326582278481025e-06, "loss": 0.0, "step": 18150 }, { "epoch": 182.43, "learning_rate": 5.526329113924052e-06, "loss": 0.0, "step": 18175 }, { "epoch": 182.69, "learning_rate": 5.5200000000000005e-06, "loss": 0.0, "step": 18200 }, { "epoch": 182.94, "learning_rate": 5.51367088607595e-06, "loss": 0.0, "step": 18225 }, { "epoch": 183.19, "learning_rate": 5.507341772151899e-06, "loss": 0.0, "step": 18250 }, { "epoch": 183.44, "learning_rate": 5.501012658227849e-06, "loss": 0.0, "step": 18275 }, { "epoch": 183.69, "learning_rate": 5.494683544303798e-06, "loss": 0.0, "step": 18300 }, { "epoch": 183.94, "learning_rate": 5.488354430379748e-06, "loss": 0.0, "step": 18325 }, { "epoch": 184.19, "learning_rate": 5.482025316455697e-06, "loss": 0.0, "step": 18350 }, { "epoch": 184.44, "learning_rate": 5.4756962025316465e-06, "loss": 0.0, "step": 18375 }, { "epoch": 184.69, "learning_rate": 5.469367088607596e-06, "loss": 0.0, "step": 18400 }, { "epoch": 184.94, "learning_rate": 5.4630379746835445e-06, "loss": 0.0, "step": 18425 }, { "epoch": 185.19, "learning_rate": 5.456708860759494e-06, "loss": 0.0, "step": 18450 }, { "epoch": 185.45, "learning_rate": 5.450379746835443e-06, "loss": 0.0, "step": 18475 }, { "epoch": 185.7, "learning_rate": 5.444050632911393e-06, "loss": 0.0, "step": 18500 }, { "epoch": 185.95, "learning_rate": 5.437721518987342e-06, "loss": 0.0, "step": 18525 }, { "epoch": 186.2, "learning_rate": 5.431392405063292e-06, "loss": 0.0, "step": 18550 }, { "epoch": 186.45, "learning_rate": 5.425063291139241e-06, "loss": 0.0, "step": 18575 }, { "epoch": 186.7, "learning_rate": 5.4187341772151905e-06, "loss": 0.0, "step": 18600 }, { "epoch": 186.95, "learning_rate": 5.41240506329114e-06, "loss": 0.0, "step": 18625 }, { "epoch": 187.2, "learning_rate": 5.4060759493670885e-06, "loss": 0.0, "step": 18650 }, { "epoch": 187.45, "learning_rate": 5.399746835443038e-06, "loss": 0.0, "step": 18675 }, { "epoch": 187.7, "learning_rate": 5.393417721518987e-06, "loss": 0.0, "step": 18700 }, { "epoch": 187.95, "learning_rate": 5.387088607594937e-06, "loss": 0.0, "step": 18725 }, { "epoch": 188.21, "learning_rate": 5.380759493670886e-06, "loss": 0.0, "step": 18750 }, { "epoch": 188.46, "learning_rate": 5.374430379746836e-06, "loss": 0.0, "step": 18775 }, { "epoch": 188.71, "learning_rate": 5.368101265822785e-06, "loss": 0.0, "step": 18800 }, { "epoch": 188.96, "learning_rate": 5.3617721518987345e-06, "loss": 0.0, "step": 18825 }, { "epoch": 189.21, "learning_rate": 5.355443037974684e-06, "loss": 0.0, "step": 18850 }, { "epoch": 189.46, "learning_rate": 5.3491139240506325e-06, "loss": 0.0, "step": 18875 }, { "epoch": 189.71, "learning_rate": 5.342784810126582e-06, "loss": 0.0, "step": 18900 }, { "epoch": 189.96, "learning_rate": 5.336455696202531e-06, "loss": 0.0, "step": 18925 }, { "epoch": 190.21, "learning_rate": 5.330126582278481e-06, "loss": 0.0, "step": 18950 }, { "epoch": 190.46, "learning_rate": 5.323797468354431e-06, "loss": 0.0, "step": 18975 }, { "epoch": 190.72, "learning_rate": 5.3174683544303805e-06, "loss": 0.0, "step": 19000 }, { "epoch": 190.72, "eval_loss": 0.2962440252304077, "eval_runtime": 2778.8571, "eval_samples_per_second": 4.585, "eval_steps_per_second": 0.287, "eval_wer": 6.946615601062998, "step": 19000 }, { "epoch": 190.97, "learning_rate": 5.31113924050633e-06, "loss": 0.0, "step": 19025 }, { "epoch": 191.22, "learning_rate": 5.304810126582279e-06, "loss": 0.0, "step": 19050 }, { "epoch": 191.47, "learning_rate": 5.298481012658229e-06, "loss": 0.0, "step": 19075 }, { "epoch": 191.72, "learning_rate": 5.292151898734178e-06, "loss": 0.0, "step": 19100 }, { "epoch": 191.97, "learning_rate": 5.285822784810128e-06, "loss": 0.0, "step": 19125 }, { "epoch": 192.22, "learning_rate": 5.279493670886077e-06, "loss": 0.0, "step": 19150 }, { "epoch": 192.47, "learning_rate": 5.273164556962026e-06, "loss": 0.0, "step": 19175 }, { "epoch": 192.72, "learning_rate": 5.266835443037975e-06, "loss": 0.0, "step": 19200 }, { "epoch": 192.97, "learning_rate": 5.2605063291139245e-06, "loss": 0.0, "step": 19225 }, { "epoch": 193.22, "learning_rate": 5.254177215189874e-06, "loss": 0.0, "step": 19250 }, { "epoch": 193.48, "learning_rate": 5.247848101265823e-06, "loss": 0.0, "step": 19275 }, { "epoch": 193.73, "learning_rate": 5.241518987341773e-06, "loss": 0.0, "step": 19300 }, { "epoch": 193.98, "learning_rate": 5.235189873417722e-06, "loss": 0.0, "step": 19325 }, { "epoch": 194.23, "learning_rate": 5.228860759493672e-06, "loss": 0.0, "step": 19350 }, { "epoch": 194.48, "learning_rate": 5.222531645569621e-06, "loss": 0.0, "step": 19375 }, { "epoch": 194.73, "learning_rate": 5.21620253164557e-06, "loss": 0.0, "step": 19400 }, { "epoch": 194.98, "learning_rate": 5.209873417721519e-06, "loss": 0.0, "step": 19425 }, { "epoch": 195.23, "learning_rate": 5.2035443037974685e-06, "loss": 0.0, "step": 19450 }, { "epoch": 195.48, "learning_rate": 5.197215189873418e-06, "loss": 0.0, "step": 19475 }, { "epoch": 195.73, "learning_rate": 5.190886075949367e-06, "loss": 0.0, "step": 19500 }, { "epoch": 195.98, "learning_rate": 5.184556962025317e-06, "loss": 0.0, "step": 19525 }, { "epoch": 196.24, "learning_rate": 5.178227848101266e-06, "loss": 0.0, "step": 19550 }, { "epoch": 196.49, "learning_rate": 5.171898734177216e-06, "loss": 0.0, "step": 19575 }, { "epoch": 196.74, "learning_rate": 5.165569620253165e-06, "loss": 0.0, "step": 19600 }, { "epoch": 196.99, "learning_rate": 5.159240506329114e-06, "loss": 0.0, "step": 19625 }, { "epoch": 197.24, "learning_rate": 5.152911392405063e-06, "loss": 0.0, "step": 19650 }, { "epoch": 197.49, "learning_rate": 5.1465822784810125e-06, "loss": 0.0, "step": 19675 }, { "epoch": 197.74, "learning_rate": 5.140253164556962e-06, "loss": 0.0, "step": 19700 }, { "epoch": 197.99, "learning_rate": 5.133924050632911e-06, "loss": 0.0, "step": 19725 }, { "epoch": 198.24, "learning_rate": 5.127594936708861e-06, "loss": 0.0, "step": 19750 }, { "epoch": 198.49, "learning_rate": 5.12126582278481e-06, "loss": 0.0, "step": 19775 }, { "epoch": 198.75, "learning_rate": 5.1149367088607605e-06, "loss": 0.0, "step": 19800 }, { "epoch": 199.0, "learning_rate": 5.10860759493671e-06, "loss": 0.0, "step": 19825 }, { "epoch": 199.25, "learning_rate": 5.102278481012659e-06, "loss": 0.001, "step": 19850 }, { "epoch": 199.5, "learning_rate": 5.095949367088609e-06, "loss": 0.0056, "step": 19875 }, { "epoch": 199.75, "learning_rate": 5.089620253164557e-06, "loss": 0.0061, "step": 19900 }, { "epoch": 200.0, "learning_rate": 5.083291139240507e-06, "loss": 0.0042, "step": 19925 }, { "epoch": 200.25, "learning_rate": 5.076962025316456e-06, "loss": 0.0026, "step": 19950 }, { "epoch": 200.5, "learning_rate": 5.070886075949367e-06, "loss": 0.002, "step": 19975 }, { "epoch": 200.75, "learning_rate": 5.064556962025316e-06, "loss": 0.0013, "step": 20000 }, { "epoch": 200.75, "eval_loss": 0.2354433238506317, "eval_runtime": 2781.1722, "eval_samples_per_second": 4.582, "eval_steps_per_second": 0.287, "eval_wer": 7.604150382992028, "step": 20000 }, { "epoch": 201.0, "learning_rate": 5.058227848101266e-06, "loss": 0.0016, "step": 20025 }, { "epoch": 201.25, "learning_rate": 5.051898734177215e-06, "loss": 0.0008, "step": 20050 }, { "epoch": 201.51, "learning_rate": 5.0455696202531645e-06, "loss": 0.0005, "step": 20075 }, { "epoch": 201.76, "learning_rate": 5.039240506329114e-06, "loss": 0.0007, "step": 20100 }, { "epoch": 202.01, "learning_rate": 5.032911392405063e-06, "loss": 0.0004, "step": 20125 }, { "epoch": 202.26, "learning_rate": 5.026582278481013e-06, "loss": 0.0002, "step": 20150 }, { "epoch": 202.51, "learning_rate": 5.020253164556963e-06, "loss": 0.0001, "step": 20175 }, { "epoch": 202.76, "learning_rate": 5.0139240506329125e-06, "loss": 0.0002, "step": 20200 }, { "epoch": 203.01, "learning_rate": 5.007594936708862e-06, "loss": 0.0002, "step": 20225 }, { "epoch": 203.26, "learning_rate": 5.001265822784811e-06, "loss": 0.0001, "step": 20250 }, { "epoch": 203.51, "learning_rate": 4.99493670886076e-06, "loss": 0.0001, "step": 20275 }, { "epoch": 203.76, "learning_rate": 4.988607594936709e-06, "loss": 0.0001, "step": 20300 }, { "epoch": 204.02, "learning_rate": 4.982278481012659e-06, "loss": 0.0002, "step": 20325 }, { "epoch": 204.27, "learning_rate": 4.975949367088607e-06, "loss": 0.0001, "step": 20350 }, { "epoch": 204.52, "learning_rate": 4.969620253164557e-06, "loss": 0.0001, "step": 20375 }, { "epoch": 204.77, "learning_rate": 4.963291139240507e-06, "loss": 0.0001, "step": 20400 }, { "epoch": 205.02, "learning_rate": 4.9569620253164565e-06, "loss": 0.0001, "step": 20425 }, { "epoch": 205.27, "learning_rate": 4.950632911392406e-06, "loss": 0.0001, "step": 20450 }, { "epoch": 205.52, "learning_rate": 4.944303797468355e-06, "loss": 0.0001, "step": 20475 }, { "epoch": 205.77, "learning_rate": 4.937974683544304e-06, "loss": 0.0, "step": 20500 }, { "epoch": 206.02, "learning_rate": 4.931645569620253e-06, "loss": 0.0001, "step": 20525 }, { "epoch": 206.27, "learning_rate": 4.925316455696203e-06, "loss": 0.0, "step": 20550 }, { "epoch": 206.52, "learning_rate": 4.918987341772152e-06, "loss": 0.0, "step": 20575 }, { "epoch": 206.78, "learning_rate": 4.912658227848102e-06, "loss": 0.0, "step": 20600 }, { "epoch": 207.03, "learning_rate": 4.906329113924051e-06, "loss": 0.0, "step": 20625 }, { "epoch": 207.28, "learning_rate": 4.9000000000000005e-06, "loss": 0.0, "step": 20650 }, { "epoch": 207.53, "learning_rate": 4.89367088607595e-06, "loss": 0.0, "step": 20675 }, { "epoch": 207.78, "learning_rate": 4.887341772151899e-06, "loss": 0.0, "step": 20700 }, { "epoch": 208.03, "learning_rate": 4.881012658227848e-06, "loss": 0.0, "step": 20725 }, { "epoch": 208.28, "learning_rate": 4.874683544303797e-06, "loss": 0.0, "step": 20750 }, { "epoch": 208.53, "learning_rate": 4.868354430379747e-06, "loss": 0.0, "step": 20775 }, { "epoch": 208.78, "learning_rate": 4.862025316455697e-06, "loss": 0.0, "step": 20800 }, { "epoch": 209.03, "learning_rate": 4.8556962025316465e-06, "loss": 0.0, "step": 20825 }, { "epoch": 209.28, "learning_rate": 4.849367088607596e-06, "loss": 0.0, "step": 20850 }, { "epoch": 209.54, "learning_rate": 4.8430379746835445e-06, "loss": 0.0, "step": 20875 }, { "epoch": 209.79, "learning_rate": 4.836708860759494e-06, "loss": 0.0, "step": 20900 }, { "epoch": 210.04, "learning_rate": 4.830379746835443e-06, "loss": 0.0, "step": 20925 }, { "epoch": 210.29, "learning_rate": 4.824050632911393e-06, "loss": 0.0, "step": 20950 }, { "epoch": 210.54, "learning_rate": 4.817721518987342e-06, "loss": 0.0, "step": 20975 }, { "epoch": 210.79, "learning_rate": 4.811392405063292e-06, "loss": 0.0, "step": 21000 }, { "epoch": 210.79, "eval_loss": 0.2719791829586029, "eval_runtime": 2780.4356, "eval_samples_per_second": 4.583, "eval_steps_per_second": 0.287, "eval_wer": 6.89483351571049, "step": 21000 }, { "epoch": 211.04, "learning_rate": 4.805063291139241e-06, "loss": 0.0, "step": 21025 }, { "epoch": 211.29, "learning_rate": 4.7987341772151905e-06, "loss": 0.0, "step": 21050 }, { "epoch": 211.54, "learning_rate": 4.79240506329114e-06, "loss": 0.0, "step": 21075 }, { "epoch": 211.79, "learning_rate": 4.7860759493670885e-06, "loss": 0.0, "step": 21100 }, { "epoch": 212.05, "learning_rate": 4.779746835443038e-06, "loss": 0.0, "step": 21125 }, { "epoch": 212.3, "learning_rate": 4.773417721518987e-06, "loss": 0.0, "step": 21150 }, { "epoch": 212.55, "learning_rate": 4.767088607594937e-06, "loss": 0.0, "step": 21175 }, { "epoch": 212.8, "learning_rate": 4.760759493670886e-06, "loss": 0.0, "step": 21200 }, { "epoch": 213.05, "learning_rate": 4.754430379746836e-06, "loss": 0.0, "step": 21225 }, { "epoch": 213.3, "learning_rate": 4.748101265822785e-06, "loss": 0.0, "step": 21250 }, { "epoch": 213.55, "learning_rate": 4.7417721518987345e-06, "loss": 0.0, "step": 21275 }, { "epoch": 213.8, "learning_rate": 4.735443037974684e-06, "loss": 0.0, "step": 21300 }, { "epoch": 214.05, "learning_rate": 4.729113924050633e-06, "loss": 0.0, "step": 21325 }, { "epoch": 214.3, "learning_rate": 4.722784810126583e-06, "loss": 0.0, "step": 21350 }, { "epoch": 214.55, "learning_rate": 4.716455696202532e-06, "loss": 0.0, "step": 21375 }, { "epoch": 214.81, "learning_rate": 4.710126582278482e-06, "loss": 0.0, "step": 21400 }, { "epoch": 215.06, "learning_rate": 4.703797468354431e-06, "loss": 0.0, "step": 21425 }, { "epoch": 215.31, "learning_rate": 4.69746835443038e-06, "loss": 0.0, "step": 21450 }, { "epoch": 215.56, "learning_rate": 4.691139240506329e-06, "loss": 0.0, "step": 21475 }, { "epoch": 215.81, "learning_rate": 4.6848101265822785e-06, "loss": 0.0, "step": 21500 }, { "epoch": 216.06, "learning_rate": 4.678481012658228e-06, "loss": 0.0, "step": 21525 }, { "epoch": 216.31, "learning_rate": 4.672151898734177e-06, "loss": 0.0, "step": 21550 }, { "epoch": 216.56, "learning_rate": 4.665822784810127e-06, "loss": 0.0, "step": 21575 }, { "epoch": 216.81, "learning_rate": 4.659493670886076e-06, "loss": 0.0, "step": 21600 }, { "epoch": 217.06, "learning_rate": 4.653164556962026e-06, "loss": 0.0, "step": 21625 }, { "epoch": 217.31, "learning_rate": 4.646835443037975e-06, "loss": 0.0, "step": 21650 }, { "epoch": 217.57, "learning_rate": 4.6405063291139245e-06, "loss": 0.0, "step": 21675 }, { "epoch": 217.82, "learning_rate": 4.634177215189874e-06, "loss": 0.0, "step": 21700 }, { "epoch": 218.07, "learning_rate": 4.627848101265823e-06, "loss": 0.0, "step": 21725 }, { "epoch": 218.32, "learning_rate": 4.621518987341773e-06, "loss": 0.0, "step": 21750 }, { "epoch": 218.57, "learning_rate": 4.615189873417722e-06, "loss": 0.0, "step": 21775 }, { "epoch": 218.82, "learning_rate": 4.608860759493672e-06, "loss": 0.0, "step": 21800 }, { "epoch": 219.07, "learning_rate": 4.60253164556962e-06, "loss": 0.0, "step": 21825 }, { "epoch": 219.32, "learning_rate": 4.59620253164557e-06, "loss": 0.0, "step": 21850 }, { "epoch": 219.57, "learning_rate": 4.589873417721519e-06, "loss": 0.0, "step": 21875 }, { "epoch": 219.82, "learning_rate": 4.5835443037974685e-06, "loss": 0.0, "step": 21900 }, { "epoch": 220.08, "learning_rate": 4.577215189873418e-06, "loss": 0.0, "step": 21925 }, { "epoch": 220.33, "learning_rate": 4.570886075949367e-06, "loss": 0.0, "step": 21950 }, { "epoch": 220.58, "learning_rate": 4.564556962025317e-06, "loss": 0.0, "step": 21975 }, { "epoch": 220.83, "learning_rate": 4.558227848101266e-06, "loss": 0.0, "step": 22000 }, { "epoch": 220.83, "eval_loss": 0.2865145206451416, "eval_runtime": 2791.2624, "eval_samples_per_second": 4.565, "eval_steps_per_second": 0.286, "eval_wer": 6.898741597623887, "step": 22000 }, { "epoch": 221.08, "learning_rate": 4.551898734177216e-06, "loss": 0.0, "step": 22025 }, { "epoch": 221.33, "learning_rate": 4.545569620253165e-06, "loss": 0.0, "step": 22050 }, { "epoch": 221.58, "learning_rate": 4.5392405063291145e-06, "loss": 0.0, "step": 22075 }, { "epoch": 221.83, "learning_rate": 4.532911392405064e-06, "loss": 0.0, "step": 22100 }, { "epoch": 222.08, "learning_rate": 4.526582278481013e-06, "loss": 0.0, "step": 22125 }, { "epoch": 222.33, "learning_rate": 4.520253164556963e-06, "loss": 0.0, "step": 22150 }, { "epoch": 222.58, "learning_rate": 4.513924050632912e-06, "loss": 0.0, "step": 22175 }, { "epoch": 222.84, "learning_rate": 4.507594936708861e-06, "loss": 0.0, "step": 22200 }, { "epoch": 223.09, "learning_rate": 4.50126582278481e-06, "loss": 0.0, "step": 22225 }, { "epoch": 223.34, "learning_rate": 4.49493670886076e-06, "loss": 0.0, "step": 22250 }, { "epoch": 223.59, "learning_rate": 4.488607594936709e-06, "loss": 0.0, "step": 22275 }, { "epoch": 223.84, "learning_rate": 4.4822784810126585e-06, "loss": 0.0, "step": 22300 }, { "epoch": 224.09, "learning_rate": 4.475949367088608e-06, "loss": 0.0, "step": 22325 }, { "epoch": 224.34, "learning_rate": 4.469620253164557e-06, "loss": 0.0, "step": 22350 }, { "epoch": 224.59, "learning_rate": 4.463291139240507e-06, "loss": 0.0, "step": 22375 }, { "epoch": 224.84, "learning_rate": 4.456962025316456e-06, "loss": 0.0, "step": 22400 }, { "epoch": 225.09, "learning_rate": 4.450632911392405e-06, "loss": 0.0, "step": 22425 }, { "epoch": 225.35, "learning_rate": 4.444303797468355e-06, "loss": 0.0, "step": 22450 }, { "epoch": 225.6, "learning_rate": 4.4379746835443045e-06, "loss": 0.0, "step": 22475 }, { "epoch": 225.85, "learning_rate": 4.431645569620254e-06, "loss": 0.0, "step": 22500 }, { "epoch": 226.1, "learning_rate": 4.425316455696203e-06, "loss": 0.0, "step": 22525 }, { "epoch": 226.35, "learning_rate": 4.418987341772153e-06, "loss": 0.0, "step": 22550 }, { "epoch": 226.6, "learning_rate": 4.412658227848101e-06, "loss": 0.0, "step": 22575 }, { "epoch": 226.85, "learning_rate": 4.406329113924051e-06, "loss": 0.0, "step": 22600 }, { "epoch": 227.1, "learning_rate": 4.4e-06, "loss": 0.0, "step": 22625 }, { "epoch": 227.35, "learning_rate": 4.39367088607595e-06, "loss": 0.0, "step": 22650 }, { "epoch": 227.6, "learning_rate": 4.387341772151899e-06, "loss": 0.0, "step": 22675 }, { "epoch": 227.85, "learning_rate": 4.3810126582278485e-06, "loss": 0.0, "step": 22700 }, { "epoch": 228.11, "learning_rate": 4.374683544303798e-06, "loss": 0.0, "step": 22725 }, { "epoch": 228.36, "learning_rate": 4.368354430379747e-06, "loss": 0.0, "step": 22750 }, { "epoch": 228.61, "learning_rate": 4.362025316455697e-06, "loss": 0.0, "step": 22775 }, { "epoch": 228.86, "learning_rate": 4.355696202531645e-06, "loss": 0.0, "step": 22800 }, { "epoch": 229.11, "learning_rate": 4.349367088607595e-06, "loss": 0.0, "step": 22825 }, { "epoch": 229.36, "learning_rate": 4.343037974683545e-06, "loss": 0.0, "step": 22850 }, { "epoch": 229.61, "learning_rate": 4.3367088607594945e-06, "loss": 0.0, "step": 22875 }, { "epoch": 229.86, "learning_rate": 4.330379746835444e-06, "loss": 0.0, "step": 22900 }, { "epoch": 230.11, "learning_rate": 4.3240506329113925e-06, "loss": 0.0, "step": 22925 }, { "epoch": 230.36, "learning_rate": 4.317721518987342e-06, "loss": 0.0, "step": 22950 }, { "epoch": 230.61, "learning_rate": 4.311392405063291e-06, "loss": 0.0, "step": 22975 }, { "epoch": 230.87, "learning_rate": 4.305063291139241e-06, "loss": 0.0, "step": 23000 }, { "epoch": 230.87, "eval_loss": 0.2953989505767822, "eval_runtime": 2779.9715, "eval_samples_per_second": 4.584, "eval_steps_per_second": 0.287, "eval_wer": 6.888971392840395, "step": 23000 }, { "epoch": 231.12, "learning_rate": 4.29873417721519e-06, "loss": 0.0, "step": 23025 }, { "epoch": 231.37, "learning_rate": 4.29240506329114e-06, "loss": 0.0, "step": 23050 }, { "epoch": 231.62, "learning_rate": 4.286075949367089e-06, "loss": 0.0, "step": 23075 }, { "epoch": 231.87, "learning_rate": 4.2797468354430385e-06, "loss": 0.0, "step": 23100 }, { "epoch": 232.12, "learning_rate": 4.273417721518988e-06, "loss": 0.0, "step": 23125 }, { "epoch": 232.37, "learning_rate": 4.2670886075949365e-06, "loss": 0.0, "step": 23150 }, { "epoch": 232.62, "learning_rate": 4.260759493670886e-06, "loss": 0.0, "step": 23175 }, { "epoch": 232.87, "learning_rate": 4.254430379746835e-06, "loss": 0.0, "step": 23200 }, { "epoch": 233.12, "learning_rate": 4.248101265822785e-06, "loss": 0.0, "step": 23225 }, { "epoch": 233.38, "learning_rate": 4.241772151898734e-06, "loss": 0.0, "step": 23250 }, { "epoch": 233.63, "learning_rate": 4.2354430379746845e-06, "loss": 0.0, "step": 23275 }, { "epoch": 233.88, "learning_rate": 4.229113924050633e-06, "loss": 0.0, "step": 23300 }, { "epoch": 234.13, "learning_rate": 4.2227848101265825e-06, "loss": 0.0, "step": 23325 }, { "epoch": 234.38, "learning_rate": 4.216455696202532e-06, "loss": 0.0, "step": 23350 }, { "epoch": 234.63, "learning_rate": 4.210126582278481e-06, "loss": 0.0, "step": 23375 }, { "epoch": 234.88, "learning_rate": 4.203797468354431e-06, "loss": 0.0, "step": 23400 }, { "epoch": 235.13, "learning_rate": 4.19746835443038e-06, "loss": 0.0, "step": 23425 }, { "epoch": 235.38, "learning_rate": 4.19113924050633e-06, "loss": 0.0, "step": 23450 }, { "epoch": 235.63, "learning_rate": 4.184810126582279e-06, "loss": 0.0, "step": 23475 }, { "epoch": 235.88, "learning_rate": 4.1784810126582285e-06, "loss": 0.0, "step": 23500 }, { "epoch": 236.14, "learning_rate": 4.172151898734177e-06, "loss": 0.0, "step": 23525 }, { "epoch": 236.39, "learning_rate": 4.1658227848101265e-06, "loss": 0.0, "step": 23550 }, { "epoch": 236.64, "learning_rate": 4.159493670886076e-06, "loss": 0.0, "step": 23575 }, { "epoch": 236.89, "learning_rate": 4.153164556962025e-06, "loss": 0.0, "step": 23600 }, { "epoch": 237.14, "learning_rate": 4.146835443037975e-06, "loss": 0.0, "step": 23625 }, { "epoch": 237.39, "learning_rate": 4.140506329113924e-06, "loss": 0.0, "step": 23650 }, { "epoch": 237.64, "learning_rate": 4.134177215189874e-06, "loss": 0.0, "step": 23675 }, { "epoch": 237.89, "learning_rate": 4.127848101265823e-06, "loss": 0.0, "step": 23700 }, { "epoch": 238.14, "learning_rate": 4.1215189873417725e-06, "loss": 0.0, "step": 23725 }, { "epoch": 238.39, "learning_rate": 4.115189873417722e-06, "loss": 0.0, "step": 23750 }, { "epoch": 238.64, "learning_rate": 4.108860759493671e-06, "loss": 0.0, "step": 23775 }, { "epoch": 238.9, "learning_rate": 4.102531645569621e-06, "loss": 0.0, "step": 23800 }, { "epoch": 239.15, "learning_rate": 4.09620253164557e-06, "loss": 0.0, "step": 23825 }, { "epoch": 239.4, "learning_rate": 4.08987341772152e-06, "loss": 0.0, "step": 23850 }, { "epoch": 239.65, "learning_rate": 4.083544303797469e-06, "loss": 0.0, "step": 23875 }, { "epoch": 239.9, "learning_rate": 4.077215189873418e-06, "loss": 0.0, "step": 23900 }, { "epoch": 240.15, "learning_rate": 4.070886075949367e-06, "loss": 0.0, "step": 23925 }, { "epoch": 240.4, "learning_rate": 4.0645569620253165e-06, "loss": 0.0, "step": 23950 }, { "epoch": 240.65, "learning_rate": 4.058227848101266e-06, "loss": 0.0, "step": 23975 }, { "epoch": 240.9, "learning_rate": 4.051898734177215e-06, "loss": 0.0, "step": 24000 }, { "epoch": 240.9, "eval_loss": 0.3031490743160248, "eval_runtime": 2780.7351, "eval_samples_per_second": 4.582, "eval_steps_per_second": 0.287, "eval_wer": 6.88213224949195, "step": 24000 }, { "epoch": 241.15, "learning_rate": 4.045569620253165e-06, "loss": 0.0, "step": 24025 }, { "epoch": 241.41, "learning_rate": 4.039240506329114e-06, "loss": 0.0, "step": 24050 }, { "epoch": 241.66, "learning_rate": 4.032911392405064e-06, "loss": 0.0, "step": 24075 }, { "epoch": 241.91, "learning_rate": 4.026582278481013e-06, "loss": 0.0, "step": 24100 }, { "epoch": 242.16, "learning_rate": 4.0202531645569625e-06, "loss": 0.0, "step": 24125 }, { "epoch": 242.41, "learning_rate": 4.013924050632912e-06, "loss": 0.0, "step": 24150 }, { "epoch": 242.66, "learning_rate": 4.007594936708861e-06, "loss": 0.0, "step": 24175 }, { "epoch": 242.91, "learning_rate": 4.001265822784811e-06, "loss": 0.0, "step": 24200 }, { "epoch": 243.16, "learning_rate": 3.99493670886076e-06, "loss": 0.0, "step": 24225 }, { "epoch": 243.41, "learning_rate": 3.98860759493671e-06, "loss": 0.0, "step": 24250 }, { "epoch": 243.66, "learning_rate": 3.982278481012658e-06, "loss": 0.0, "step": 24275 }, { "epoch": 243.91, "learning_rate": 3.975949367088608e-06, "loss": 0.0, "step": 24300 }, { "epoch": 244.17, "learning_rate": 3.969620253164557e-06, "loss": 0.0, "step": 24325 }, { "epoch": 244.42, "learning_rate": 3.9632911392405065e-06, "loss": 0.0, "step": 24350 }, { "epoch": 244.67, "learning_rate": 3.956962025316456e-06, "loss": 0.0, "step": 24375 }, { "epoch": 244.92, "learning_rate": 3.950632911392405e-06, "loss": 0.0, "step": 24400 }, { "epoch": 245.17, "learning_rate": 3.944303797468355e-06, "loss": 0.0, "step": 24425 }, { "epoch": 245.42, "learning_rate": 3.937974683544304e-06, "loss": 0.0, "step": 24450 }, { "epoch": 245.67, "learning_rate": 3.931645569620254e-06, "loss": 0.0, "step": 24475 }, { "epoch": 245.92, "learning_rate": 3.925316455696203e-06, "loss": 0.0, "step": 24500 }, { "epoch": 246.17, "learning_rate": 3.9189873417721525e-06, "loss": 0.0, "step": 24525 }, { "epoch": 246.42, "learning_rate": 3.912658227848102e-06, "loss": 0.0, "step": 24550 }, { "epoch": 246.68, "learning_rate": 3.906329113924051e-06, "loss": 0.0, "step": 24575 }, { "epoch": 246.93, "learning_rate": 3.900000000000001e-06, "loss": 0.0, "step": 24600 }, { "epoch": 247.18, "learning_rate": 3.89367088607595e-06, "loss": 0.0, "step": 24625 }, { "epoch": 247.43, "learning_rate": 3.887341772151899e-06, "loss": 0.0, "step": 24650 }, { "epoch": 247.68, "learning_rate": 3.881012658227848e-06, "loss": 0.0, "step": 24675 }, { "epoch": 247.93, "learning_rate": 3.874683544303798e-06, "loss": 0.0, "step": 24700 }, { "epoch": 248.18, "learning_rate": 3.868354430379747e-06, "loss": 0.0, "step": 24725 }, { "epoch": 248.43, "learning_rate": 3.8620253164556965e-06, "loss": 0.0, "step": 24750 }, { "epoch": 248.68, "learning_rate": 3.855696202531646e-06, "loss": 0.0, "step": 24775 }, { "epoch": 248.93, "learning_rate": 3.849367088607595e-06, "loss": 0.0, "step": 24800 }, { "epoch": 249.18, "learning_rate": 3.843037974683545e-06, "loss": 0.0, "step": 24825 }, { "epoch": 249.44, "learning_rate": 3.836708860759493e-06, "loss": 0.0, "step": 24850 }, { "epoch": 249.69, "learning_rate": 3.830379746835443e-06, "loss": 0.0, "step": 24875 }, { "epoch": 249.94, "learning_rate": 3.824050632911392e-06, "loss": 0.0, "step": 24900 }, { "epoch": 250.19, "learning_rate": 3.8177215189873425e-06, "loss": 0.0, "step": 24925 }, { "epoch": 250.44, "learning_rate": 3.8113924050632915e-06, "loss": 0.0, "step": 24950 }, { "epoch": 250.69, "learning_rate": 3.805063291139241e-06, "loss": 0.0, "step": 24975 }, { "epoch": 250.94, "learning_rate": 3.7987341772151903e-06, "loss": 0.0, "step": 25000 }, { "epoch": 250.94, "eval_loss": 0.3101833760738373, "eval_runtime": 2782.1977, "eval_samples_per_second": 4.58, "eval_steps_per_second": 0.286, "eval_wer": 6.877247147100203, "step": 25000 }, { "epoch": 251.19, "learning_rate": 3.7924050632911398e-06, "loss": 0.0, "step": 25025 }, { "epoch": 251.44, "learning_rate": 3.786075949367089e-06, "loss": 0.0, "step": 25050 }, { "epoch": 251.69, "learning_rate": 3.779746835443038e-06, "loss": 0.0, "step": 25075 }, { "epoch": 251.94, "learning_rate": 3.7734177215189876e-06, "loss": 0.0, "step": 25100 }, { "epoch": 252.2, "learning_rate": 3.767088607594937e-06, "loss": 0.0, "step": 25125 }, { "epoch": 252.45, "learning_rate": 3.7607594936708865e-06, "loss": 0.0, "step": 25150 }, { "epoch": 252.7, "learning_rate": 3.7544303797468355e-06, "loss": 0.0, "step": 25175 }, { "epoch": 252.95, "learning_rate": 3.748101265822785e-06, "loss": 0.0, "step": 25200 }, { "epoch": 253.2, "learning_rate": 3.7417721518987344e-06, "loss": 0.0, "step": 25225 }, { "epoch": 253.45, "learning_rate": 3.7354430379746838e-06, "loss": 0.0, "step": 25250 }, { "epoch": 253.7, "learning_rate": 3.7291139240506328e-06, "loss": 0.0, "step": 25275 }, { "epoch": 253.95, "learning_rate": 3.722784810126582e-06, "loss": 0.0, "step": 25300 }, { "epoch": 254.2, "learning_rate": 3.716455696202532e-06, "loss": 0.0, "step": 25325 }, { "epoch": 254.45, "learning_rate": 3.7101265822784815e-06, "loss": 0.0, "step": 25350 }, { "epoch": 254.71, "learning_rate": 3.703797468354431e-06, "loss": 0.0, "step": 25375 }, { "epoch": 254.96, "learning_rate": 3.6974683544303803e-06, "loss": 0.0, "step": 25400 }, { "epoch": 255.21, "learning_rate": 3.6911392405063293e-06, "loss": 0.0, "step": 25425 }, { "epoch": 255.46, "learning_rate": 3.6848101265822788e-06, "loss": 0.0, "step": 25450 }, { "epoch": 255.71, "learning_rate": 3.678481012658228e-06, "loss": 0.0, "step": 25475 }, { "epoch": 255.96, "learning_rate": 3.6721518987341776e-06, "loss": 0.0, "step": 25500 }, { "epoch": 256.21, "learning_rate": 3.665822784810127e-06, "loss": 0.0, "step": 25525 }, { "epoch": 256.46, "learning_rate": 3.659493670886076e-06, "loss": 0.0, "step": 25550 }, { "epoch": 256.71, "learning_rate": 3.6531645569620255e-06, "loss": 0.0, "step": 25575 }, { "epoch": 256.96, "learning_rate": 3.646835443037975e-06, "loss": 0.0, "step": 25600 }, { "epoch": 257.21, "learning_rate": 3.6405063291139243e-06, "loss": 0.0, "step": 25625 }, { "epoch": 257.47, "learning_rate": 3.6341772151898734e-06, "loss": 0.0, "step": 25650 }, { "epoch": 257.72, "learning_rate": 3.6278481012658228e-06, "loss": 0.0, "step": 25675 }, { "epoch": 257.97, "learning_rate": 3.621518987341772e-06, "loss": 0.0, "step": 25700 }, { "epoch": 258.22, "learning_rate": 3.6151898734177216e-06, "loss": 0.0, "step": 25725 }, { "epoch": 258.47, "learning_rate": 3.6088607594936715e-06, "loss": 0.0, "step": 25750 }, { "epoch": 258.72, "learning_rate": 3.602531645569621e-06, "loss": 0.0, "step": 25775 }, { "epoch": 258.97, "learning_rate": 3.59620253164557e-06, "loss": 0.0, "step": 25800 }, { "epoch": 259.22, "learning_rate": 3.5898734177215193e-06, "loss": 0.0, "step": 25825 }, { "epoch": 259.47, "learning_rate": 3.5835443037974688e-06, "loss": 0.0, "step": 25850 }, { "epoch": 259.72, "learning_rate": 3.577215189873418e-06, "loss": 0.0, "step": 25875 }, { "epoch": 259.97, "learning_rate": 3.5708860759493676e-06, "loss": 0.0, "step": 25900 }, { "epoch": 260.23, "learning_rate": 3.5645569620253166e-06, "loss": 0.0, "step": 25925 }, { "epoch": 260.48, "learning_rate": 3.558227848101266e-06, "loss": 0.0, "step": 25950 }, { "epoch": 260.73, "learning_rate": 3.5518987341772155e-06, "loss": 0.0, "step": 25975 }, { "epoch": 260.98, "learning_rate": 3.545569620253165e-06, "loss": 0.0, "step": 26000 }, { "epoch": 260.98, "eval_loss": 0.3166370987892151, "eval_runtime": 2780.7731, "eval_samples_per_second": 4.582, "eval_steps_per_second": 0.287, "eval_wer": 6.889948413318743, "step": 26000 }, { "epoch": 261.23, "learning_rate": 3.539240506329114e-06, "loss": 0.0, "step": 26025 }, { "epoch": 261.48, "learning_rate": 3.5329113924050633e-06, "loss": 0.0, "step": 26050 }, { "epoch": 261.73, "learning_rate": 3.5265822784810128e-06, "loss": 0.0, "step": 26075 }, { "epoch": 261.98, "learning_rate": 3.520253164556962e-06, "loss": 0.0, "step": 26100 }, { "epoch": 262.23, "learning_rate": 3.5139240506329116e-06, "loss": 0.0, "step": 26125 }, { "epoch": 262.48, "learning_rate": 3.5075949367088615e-06, "loss": 0.0, "step": 26150 }, { "epoch": 262.74, "learning_rate": 3.5012658227848105e-06, "loss": 0.0, "step": 26175 }, { "epoch": 262.99, "learning_rate": 3.49493670886076e-06, "loss": 0.0, "step": 26200 }, { "epoch": 263.24, "learning_rate": 3.4886075949367093e-06, "loss": 0.0, "step": 26225 }, { "epoch": 263.49, "learning_rate": 3.4822784810126588e-06, "loss": 0.0, "step": 26250 }, { "epoch": 263.74, "learning_rate": 3.4759493670886078e-06, "loss": 0.0, "step": 26275 }, { "epoch": 263.99, "learning_rate": 3.469620253164557e-06, "loss": 0.0, "step": 26300 }, { "epoch": 264.24, "learning_rate": 3.4632911392405066e-06, "loss": 0.0, "step": 26325 }, { "epoch": 264.49, "learning_rate": 3.456962025316456e-06, "loss": 0.0, "step": 26350 }, { "epoch": 264.74, "learning_rate": 3.4506329113924055e-06, "loss": 0.0, "step": 26375 }, { "epoch": 264.99, "learning_rate": 3.4443037974683545e-06, "loss": 0.0, "step": 26400 }, { "epoch": 265.24, "learning_rate": 3.437974683544304e-06, "loss": 0.0, "step": 26425 }, { "epoch": 265.5, "learning_rate": 3.4316455696202533e-06, "loss": 0.0, "step": 26450 }, { "epoch": 265.75, "learning_rate": 3.4253164556962028e-06, "loss": 0.0, "step": 26475 }, { "epoch": 266.0, "learning_rate": 3.4189873417721518e-06, "loss": 0.0, "step": 26500 }, { "epoch": 266.25, "learning_rate": 3.412658227848101e-06, "loss": 0.0, "step": 26525 }, { "epoch": 266.5, "learning_rate": 3.4063291139240506e-06, "loss": 0.0, "step": 26550 }, { "epoch": 266.75, "learning_rate": 3.4000000000000005e-06, "loss": 0.0, "step": 26575 }, { "epoch": 267.0, "learning_rate": 3.39367088607595e-06, "loss": 0.0, "step": 26600 }, { "epoch": 267.25, "learning_rate": 3.3873417721518993e-06, "loss": 0.0, "step": 26625 }, { "epoch": 267.5, "learning_rate": 3.3810126582278483e-06, "loss": 0.0, "step": 26650 }, { "epoch": 267.75, "learning_rate": 3.3746835443037978e-06, "loss": 0.0, "step": 26675 }, { "epoch": 268.01, "learning_rate": 3.368354430379747e-06, "loss": 0.0, "step": 26700 }, { "epoch": 268.26, "learning_rate": 3.3620253164556966e-06, "loss": 0.0, "step": 26725 }, { "epoch": 268.51, "learning_rate": 3.355696202531646e-06, "loss": 0.0, "step": 26750 }, { "epoch": 268.76, "learning_rate": 3.349367088607595e-06, "loss": 0.0, "step": 26775 }, { "epoch": 269.01, "learning_rate": 3.3430379746835445e-06, "loss": 0.0, "step": 26800 }, { "epoch": 269.26, "learning_rate": 3.336708860759494e-06, "loss": 0.0, "step": 26825 }, { "epoch": 269.51, "learning_rate": 3.3303797468354433e-06, "loss": 0.0, "step": 26850 }, { "epoch": 269.76, "learning_rate": 3.3240506329113923e-06, "loss": 0.0, "step": 26875 }, { "epoch": 270.01, "learning_rate": 3.3177215189873418e-06, "loss": 0.0, "step": 26900 }, { "epoch": 270.26, "learning_rate": 3.311392405063291e-06, "loss": 0.0, "step": 26925 }, { "epoch": 270.51, "learning_rate": 3.3050632911392406e-06, "loss": 0.0, "step": 26950 }, { "epoch": 270.77, "learning_rate": 3.2987341772151905e-06, "loss": 0.0, "step": 26975 }, { "epoch": 271.02, "learning_rate": 3.29240506329114e-06, "loss": 0.0, "step": 27000 }, { "epoch": 271.02, "eval_loss": 0.3233168423175812, "eval_runtime": 2786.2635, "eval_samples_per_second": 4.573, "eval_steps_per_second": 0.286, "eval_wer": 6.891902454275442, "step": 27000 }, { "epoch": 271.27, "learning_rate": 3.286075949367089e-06, "loss": 0.0, "step": 27025 }, { "epoch": 271.52, "learning_rate": 3.2797468354430383e-06, "loss": 0.0, "step": 27050 }, { "epoch": 271.77, "learning_rate": 3.2734177215189878e-06, "loss": 0.0, "step": 27075 }, { "epoch": 272.02, "learning_rate": 3.267088607594937e-06, "loss": 0.0, "step": 27100 }, { "epoch": 272.27, "learning_rate": 3.260759493670886e-06, "loss": 0.0, "step": 27125 }, { "epoch": 272.52, "learning_rate": 3.2544303797468356e-06, "loss": 0.0, "step": 27150 }, { "epoch": 272.77, "learning_rate": 3.248101265822785e-06, "loss": 0.0, "step": 27175 }, { "epoch": 273.02, "learning_rate": 3.2417721518987345e-06, "loss": 0.0, "step": 27200 }, { "epoch": 273.27, "learning_rate": 3.235443037974684e-06, "loss": 0.0, "step": 27225 }, { "epoch": 273.53, "learning_rate": 3.229113924050633e-06, "loss": 0.0, "step": 27250 }, { "epoch": 273.78, "learning_rate": 3.2227848101265823e-06, "loss": 0.0, "step": 27275 }, { "epoch": 274.03, "learning_rate": 3.2164556962025318e-06, "loss": 0.0, "step": 27300 }, { "epoch": 274.28, "learning_rate": 3.210126582278481e-06, "loss": 0.0, "step": 27325 }, { "epoch": 274.53, "learning_rate": 3.20379746835443e-06, "loss": 0.0, "step": 27350 }, { "epoch": 274.78, "learning_rate": 3.1974683544303805e-06, "loss": 0.0, "step": 27375 }, { "epoch": 275.03, "learning_rate": 3.1911392405063295e-06, "loss": 0.0, "step": 27400 }, { "epoch": 275.28, "learning_rate": 3.184810126582279e-06, "loss": 0.0, "step": 27425 }, { "epoch": 275.53, "learning_rate": 3.1784810126582283e-06, "loss": 0.0, "step": 27450 }, { "epoch": 275.78, "learning_rate": 3.1721518987341778e-06, "loss": 0.0, "step": 27475 }, { "epoch": 276.04, "learning_rate": 3.1658227848101268e-06, "loss": 0.0, "step": 27500 }, { "epoch": 276.29, "learning_rate": 3.159493670886076e-06, "loss": 0.0, "step": 27525 }, { "epoch": 276.54, "learning_rate": 3.1531645569620256e-06, "loss": 0.0, "step": 27550 }, { "epoch": 276.79, "learning_rate": 3.146835443037975e-06, "loss": 0.0, "step": 27575 }, { "epoch": 277.04, "learning_rate": 3.1405063291139245e-06, "loss": 0.0, "step": 27600 }, { "epoch": 277.29, "learning_rate": 3.1341772151898735e-06, "loss": 0.0, "step": 27625 }, { "epoch": 277.54, "learning_rate": 3.127848101265823e-06, "loss": 0.0, "step": 27650 }, { "epoch": 277.79, "learning_rate": 3.1215189873417723e-06, "loss": 0.0, "step": 27675 }, { "epoch": 278.04, "learning_rate": 3.1151898734177218e-06, "loss": 0.0, "step": 27700 }, { "epoch": 278.29, "learning_rate": 3.1088607594936708e-06, "loss": 0.0, "step": 27725 }, { "epoch": 278.54, "learning_rate": 3.10253164556962e-06, "loss": 0.0, "step": 27750 }, { "epoch": 278.8, "learning_rate": 3.0962025316455696e-06, "loss": 0.0, "step": 27775 }, { "epoch": 279.05, "learning_rate": 3.0898734177215195e-06, "loss": 0.0, "step": 27800 }, { "epoch": 279.3, "learning_rate": 3.083544303797469e-06, "loss": 0.0, "step": 27825 }, { "epoch": 279.55, "learning_rate": 3.0772151898734183e-06, "loss": 0.0, "step": 27850 }, { "epoch": 279.8, "learning_rate": 3.0708860759493673e-06, "loss": 0.0, "step": 27875 }, { "epoch": 280.05, "learning_rate": 3.0645569620253168e-06, "loss": 0.0, "step": 27900 }, { "epoch": 280.3, "learning_rate": 3.058227848101266e-06, "loss": 0.0, "step": 27925 }, { "epoch": 280.55, "learning_rate": 3.0518987341772156e-06, "loss": 0.0, "step": 27950 }, { "epoch": 280.8, "learning_rate": 3.0455696202531646e-06, "loss": 0.0, "step": 27975 }, { "epoch": 281.05, "learning_rate": 3.039240506329114e-06, "loss": 0.0, "step": 28000 }, { "epoch": 281.05, "eval_loss": 0.32482120394706726, "eval_runtime": 2800.9161, "eval_samples_per_second": 4.549, "eval_steps_per_second": 0.285, "eval_wer": 6.891902454275442, "step": 28000 }, { "epoch": 281.3, "learning_rate": 3.0329113924050635e-06, "loss": 0.0, "step": 28025 }, { "epoch": 281.56, "learning_rate": 3.026835443037975e-06, "loss": 0.0, "step": 28050 }, { "epoch": 281.81, "learning_rate": 3.0205063291139244e-06, "loss": 0.0, "step": 28075 }, { "epoch": 282.06, "learning_rate": 3.0141772151898734e-06, "loss": 0.0, "step": 28100 }, { "epoch": 282.31, "learning_rate": 3.0078481012658228e-06, "loss": 0.0, "step": 28125 }, { "epoch": 282.56, "learning_rate": 3.0015189873417722e-06, "loss": 0.0, "step": 28150 }, { "epoch": 282.81, "learning_rate": 2.9951898734177216e-06, "loss": 0.0, "step": 28175 }, { "epoch": 283.06, "learning_rate": 2.9888607594936715e-06, "loss": 0.0, "step": 28200 }, { "epoch": 283.31, "learning_rate": 2.982531645569621e-06, "loss": 0.0, "step": 28225 }, { "epoch": 283.56, "learning_rate": 2.97620253164557e-06, "loss": 0.0, "step": 28250 }, { "epoch": 283.81, "learning_rate": 2.9698734177215194e-06, "loss": 0.0, "step": 28275 }, { "epoch": 284.07, "learning_rate": 2.9635443037974688e-06, "loss": 0.0, "step": 28300 }, { "epoch": 284.32, "learning_rate": 2.957215189873418e-06, "loss": 0.0, "step": 28325 }, { "epoch": 284.57, "learning_rate": 2.950886075949367e-06, "loss": 0.0, "step": 28350 }, { "epoch": 284.82, "learning_rate": 2.9445569620253166e-06, "loss": 0.0, "step": 28375 }, { "epoch": 285.07, "learning_rate": 2.938227848101266e-06, "loss": 0.0, "step": 28400 }, { "epoch": 285.32, "learning_rate": 2.9318987341772155e-06, "loss": 0.0, "step": 28425 }, { "epoch": 285.57, "learning_rate": 2.925569620253165e-06, "loss": 0.0, "step": 28450 }, { "epoch": 285.82, "learning_rate": 2.919240506329114e-06, "loss": 0.0, "step": 28475 }, { "epoch": 286.07, "learning_rate": 2.9129113924050634e-06, "loss": 0.0, "step": 28500 }, { "epoch": 286.32, "learning_rate": 2.9065822784810128e-06, "loss": 0.0, "step": 28525 }, { "epoch": 286.57, "learning_rate": 2.900253164556962e-06, "loss": 0.0, "step": 28550 }, { "epoch": 286.83, "learning_rate": 2.8939240506329112e-06, "loss": 0.0, "step": 28575 }, { "epoch": 287.08, "learning_rate": 2.887594936708861e-06, "loss": 0.0, "step": 28600 }, { "epoch": 287.33, "learning_rate": 2.8812658227848105e-06, "loss": 0.0, "step": 28625 }, { "epoch": 287.58, "learning_rate": 2.87493670886076e-06, "loss": 0.0, "step": 28650 }, { "epoch": 287.83, "learning_rate": 2.8686075949367093e-06, "loss": 0.0, "step": 28675 }, { "epoch": 288.08, "learning_rate": 2.8622784810126588e-06, "loss": 0.0, "step": 28700 }, { "epoch": 288.33, "learning_rate": 2.8559493670886078e-06, "loss": 0.0, "step": 28725 }, { "epoch": 288.58, "learning_rate": 2.849620253164557e-06, "loss": 0.0, "step": 28750 }, { "epoch": 288.83, "learning_rate": 2.8432911392405066e-06, "loss": 0.0, "step": 28775 }, { "epoch": 289.08, "learning_rate": 2.836962025316456e-06, "loss": 0.0, "step": 28800 }, { "epoch": 289.34, "learning_rate": 2.830632911392405e-06, "loss": 0.0, "step": 28825 }, { "epoch": 289.59, "learning_rate": 2.8243037974683545e-06, "loss": 0.0, "step": 28850 }, { "epoch": 289.84, "learning_rate": 2.817974683544304e-06, "loss": 0.0, "step": 28875 }, { "epoch": 290.09, "learning_rate": 2.8116455696202533e-06, "loss": 0.0, "step": 28900 }, { "epoch": 290.34, "learning_rate": 2.8053164556962028e-06, "loss": 0.0, "step": 28925 }, { "epoch": 290.59, "learning_rate": 2.7989873417721518e-06, "loss": 0.0, "step": 28950 }, { "epoch": 290.84, "learning_rate": 2.792658227848101e-06, "loss": 0.0, "step": 28975 }, { "epoch": 291.09, "learning_rate": 2.7863291139240506e-06, "loss": 0.0, "step": 29000 }, { "epoch": 291.09, "eval_loss": 0.33633747696876526, "eval_runtime": 2783.0499, "eval_samples_per_second": 4.578, "eval_steps_per_second": 0.286, "eval_wer": 6.902649679537283, "step": 29000 }, { "epoch": 291.34, "learning_rate": 2.7800000000000005e-06, "loss": 0.0, "step": 29025 }, { "epoch": 291.59, "learning_rate": 2.77367088607595e-06, "loss": 0.0, "step": 29050 }, { "epoch": 291.84, "learning_rate": 2.7673417721518993e-06, "loss": 0.0, "step": 29075 }, { "epoch": 292.1, "learning_rate": 2.7610126582278483e-06, "loss": 0.0, "step": 29100 }, { "epoch": 292.35, "learning_rate": 2.7546835443037978e-06, "loss": 0.0, "step": 29125 }, { "epoch": 292.6, "learning_rate": 2.748354430379747e-06, "loss": 0.0, "step": 29150 }, { "epoch": 292.85, "learning_rate": 2.7420253164556966e-06, "loss": 0.0, "step": 29175 }, { "epoch": 293.1, "learning_rate": 2.7356962025316456e-06, "loss": 0.0, "step": 29200 }, { "epoch": 293.35, "learning_rate": 2.729367088607595e-06, "loss": 0.0, "step": 29225 }, { "epoch": 293.6, "learning_rate": 2.7230379746835445e-06, "loss": 0.0, "step": 29250 }, { "epoch": 293.85, "learning_rate": 2.716708860759494e-06, "loss": 0.0, "step": 29275 }, { "epoch": 294.1, "learning_rate": 2.7103797468354433e-06, "loss": 0.0, "step": 29300 }, { "epoch": 294.35, "learning_rate": 2.7040506329113923e-06, "loss": 0.0, "step": 29325 }, { "epoch": 294.6, "learning_rate": 2.6977215189873418e-06, "loss": 0.0, "step": 29350 }, { "epoch": 294.86, "learning_rate": 2.691392405063291e-06, "loss": 0.0, "step": 29375 }, { "epoch": 295.11, "learning_rate": 2.6850632911392406e-06, "loss": 0.0, "step": 29400 }, { "epoch": 295.36, "learning_rate": 2.6787341772151905e-06, "loss": 0.0, "step": 29425 }, { "epoch": 295.61, "learning_rate": 2.6724050632911395e-06, "loss": 0.0, "step": 29450 }, { "epoch": 295.86, "learning_rate": 2.666075949367089e-06, "loss": 0.0, "step": 29475 }, { "epoch": 296.11, "learning_rate": 2.6597468354430383e-06, "loss": 0.0, "step": 29500 }, { "epoch": 296.36, "learning_rate": 2.6534177215189878e-06, "loss": 0.0, "step": 29525 }, { "epoch": 296.61, "learning_rate": 2.647088607594937e-06, "loss": 0.0, "step": 29550 }, { "epoch": 296.86, "learning_rate": 2.640759493670886e-06, "loss": 0.0, "step": 29575 }, { "epoch": 297.11, "learning_rate": 2.6344303797468356e-06, "loss": 0.0, "step": 29600 }, { "epoch": 297.37, "learning_rate": 2.628101265822785e-06, "loss": 0.0, "step": 29625 }, { "epoch": 297.62, "learning_rate": 2.6217721518987345e-06, "loss": 0.0, "step": 29650 }, { "epoch": 297.87, "learning_rate": 2.6154430379746835e-06, "loss": 0.0, "step": 29675 }, { "epoch": 298.12, "learning_rate": 2.609113924050633e-06, "loss": 0.0, "step": 29700 }, { "epoch": 298.37, "learning_rate": 2.6027848101265823e-06, "loss": 0.0, "step": 29725 }, { "epoch": 298.62, "learning_rate": 2.5964556962025318e-06, "loss": 0.0, "step": 29750 }, { "epoch": 298.87, "learning_rate": 2.590126582278481e-06, "loss": 0.0, "step": 29775 }, { "epoch": 299.12, "learning_rate": 2.58379746835443e-06, "loss": 0.0, "step": 29800 }, { "epoch": 299.37, "learning_rate": 2.5774683544303796e-06, "loss": 0.0, "step": 29825 }, { "epoch": 299.62, "learning_rate": 2.5711392405063295e-06, "loss": 0.0, "step": 29850 }, { "epoch": 299.87, "learning_rate": 2.564810126582279e-06, "loss": 0.0, "step": 29875 }, { "epoch": 300.13, "learning_rate": 2.5584810126582283e-06, "loss": 0.0, "step": 29900 }, { "epoch": 300.38, "learning_rate": 2.5521518987341778e-06, "loss": 0.0, "step": 29925 }, { "epoch": 300.63, "learning_rate": 2.5458227848101268e-06, "loss": 0.0, "step": 29950 }, { "epoch": 300.88, "learning_rate": 2.539493670886076e-06, "loss": 0.0, "step": 29975 }, { "epoch": 301.13, "learning_rate": 2.5331645569620256e-06, "loss": 0.0, "step": 30000 }, { "epoch": 301.13, "eval_loss": 0.34187453985214233, "eval_runtime": 2782.5975, "eval_samples_per_second": 4.579, "eval_steps_per_second": 0.286, "eval_wer": 6.908511802407379, "step": 30000 }, { "epoch": 301.38, "learning_rate": 2.526835443037975e-06, "loss": 0.0, "step": 30025 }, { "epoch": 301.63, "learning_rate": 2.520506329113924e-06, "loss": 0.0, "step": 30050 }, { "epoch": 301.88, "learning_rate": 2.5144303797468355e-06, "loss": 0.0, "step": 30075 }, { "epoch": 302.13, "learning_rate": 2.508101265822785e-06, "loss": 0.0, "step": 30100 }, { "epoch": 302.38, "learning_rate": 2.5017721518987344e-06, "loss": 0.0, "step": 30125 }, { "epoch": 302.63, "learning_rate": 2.495443037974684e-06, "loss": 0.0, "step": 30150 }, { "epoch": 302.89, "learning_rate": 2.4891139240506332e-06, "loss": 0.0, "step": 30175 }, { "epoch": 303.14, "learning_rate": 2.4827848101265826e-06, "loss": 0.0, "step": 30200 }, { "epoch": 303.39, "learning_rate": 2.4764556962025317e-06, "loss": 0.0, "step": 30225 }, { "epoch": 303.64, "learning_rate": 2.470126582278481e-06, "loss": 0.0, "step": 30250 }, { "epoch": 303.89, "learning_rate": 2.4637974683544305e-06, "loss": 0.0, "step": 30275 }, { "epoch": 304.14, "learning_rate": 2.45746835443038e-06, "loss": 0.0, "step": 30300 }, { "epoch": 304.39, "learning_rate": 2.4511392405063294e-06, "loss": 0.0, "step": 30325 }, { "epoch": 304.64, "learning_rate": 2.444810126582279e-06, "loss": 0.0, "step": 30350 }, { "epoch": 304.89, "learning_rate": 2.4384810126582282e-06, "loss": 0.0, "step": 30375 }, { "epoch": 305.14, "learning_rate": 2.4321518987341776e-06, "loss": 0.0, "step": 30400 }, { "epoch": 305.4, "learning_rate": 2.4258227848101266e-06, "loss": 0.0, "step": 30425 }, { "epoch": 305.65, "learning_rate": 2.419493670886076e-06, "loss": 0.0, "step": 30450 }, { "epoch": 305.9, "learning_rate": 2.4131645569620255e-06, "loss": 0.0, "step": 30475 }, { "epoch": 306.15, "learning_rate": 2.406835443037975e-06, "loss": 0.0, "step": 30500 }, { "epoch": 306.4, "learning_rate": 2.400506329113924e-06, "loss": 0.0, "step": 30525 }, { "epoch": 306.65, "learning_rate": 2.394177215189874e-06, "loss": 0.0, "step": 30550 }, { "epoch": 306.9, "learning_rate": 2.3878481012658232e-06, "loss": 0.0, "step": 30575 }, { "epoch": 307.15, "learning_rate": 2.3815189873417722e-06, "loss": 0.0, "step": 30600 }, { "epoch": 307.4, "learning_rate": 2.3751898734177216e-06, "loss": 0.0, "step": 30625 }, { "epoch": 307.65, "learning_rate": 2.368860759493671e-06, "loss": 0.0, "step": 30650 }, { "epoch": 307.9, "learning_rate": 2.3625316455696205e-06, "loss": 0.0, "step": 30675 }, { "epoch": 308.16, "learning_rate": 2.35620253164557e-06, "loss": 0.0, "step": 30700 }, { "epoch": 308.41, "learning_rate": 2.349873417721519e-06, "loss": 0.0, "step": 30725 }, { "epoch": 308.66, "learning_rate": 2.3435443037974688e-06, "loss": 0.0, "step": 30750 }, { "epoch": 308.91, "learning_rate": 2.337215189873418e-06, "loss": 0.0, "step": 30775 }, { "epoch": 309.16, "learning_rate": 2.3308860759493672e-06, "loss": 0.0, "step": 30800 }, { "epoch": 309.41, "learning_rate": 2.3245569620253166e-06, "loss": 0.0, "step": 30825 }, { "epoch": 309.66, "learning_rate": 2.318227848101266e-06, "loss": 0.0, "step": 30850 }, { "epoch": 309.91, "learning_rate": 2.3118987341772155e-06, "loss": 0.0, "step": 30875 }, { "epoch": 310.16, "learning_rate": 2.3055696202531645e-06, "loss": 0.0, "step": 30900 }, { "epoch": 310.41, "learning_rate": 2.299240506329114e-06, "loss": 0.0, "step": 30925 }, { "epoch": 310.66, "learning_rate": 2.2929113924050638e-06, "loss": 0.0, "step": 30950 }, { "epoch": 310.92, "learning_rate": 2.286582278481013e-06, "loss": 0.0, "step": 30975 }, { "epoch": 311.17, "learning_rate": 2.2802531645569622e-06, "loss": 0.0, "step": 31000 }, { "epoch": 311.17, "eval_loss": 0.34709784388542175, "eval_runtime": 2784.8814, "eval_samples_per_second": 4.575, "eval_steps_per_second": 0.286, "eval_wer": 6.885063310926998, "step": 31000 }, { "epoch": 311.42, "learning_rate": 2.2739240506329116e-06, "loss": 0.0, "step": 31025 }, { "epoch": 311.67, "learning_rate": 2.267594936708861e-06, "loss": 0.0, "step": 31050 }, { "epoch": 311.92, "learning_rate": 2.26126582278481e-06, "loss": 0.0, "step": 31075 }, { "epoch": 312.17, "learning_rate": 2.2549367088607595e-06, "loss": 0.0, "step": 31100 }, { "epoch": 312.42, "learning_rate": 2.248607594936709e-06, "loss": 0.0, "step": 31125 }, { "epoch": 312.67, "learning_rate": 2.2422784810126584e-06, "loss": 0.0, "step": 31150 }, { "epoch": 312.92, "learning_rate": 2.2359493670886078e-06, "loss": 0.0, "step": 31175 }, { "epoch": 313.17, "learning_rate": 2.2296202531645572e-06, "loss": 0.0, "step": 31200 }, { "epoch": 313.43, "learning_rate": 2.2232911392405066e-06, "loss": 0.0, "step": 31225 }, { "epoch": 313.68, "learning_rate": 2.216962025316456e-06, "loss": 0.0, "step": 31250 }, { "epoch": 313.93, "learning_rate": 2.210632911392405e-06, "loss": 0.0, "step": 31275 }, { "epoch": 314.18, "learning_rate": 2.2043037974683545e-06, "loss": 0.0, "step": 31300 }, { "epoch": 314.43, "learning_rate": 2.197974683544304e-06, "loss": 0.0, "step": 31325 }, { "epoch": 314.68, "learning_rate": 2.1916455696202534e-06, "loss": 0.0, "step": 31350 }, { "epoch": 314.93, "learning_rate": 2.1853164556962028e-06, "loss": 0.0, "step": 31375 }, { "epoch": 315.18, "learning_rate": 2.1789873417721522e-06, "loss": 0.0, "step": 31400 }, { "epoch": 315.43, "learning_rate": 2.1726582278481016e-06, "loss": 0.0, "step": 31425 }, { "epoch": 315.68, "learning_rate": 2.1663291139240506e-06, "loss": 0.0, "step": 31450 }, { "epoch": 315.93, "learning_rate": 2.16e-06, "loss": 0.0, "step": 31475 }, { "epoch": 316.19, "learning_rate": 2.1536708860759495e-06, "loss": 0.0, "step": 31500 }, { "epoch": 316.44, "learning_rate": 2.147341772151899e-06, "loss": 0.0, "step": 31525 }, { "epoch": 316.69, "learning_rate": 2.1410126582278484e-06, "loss": 0.0, "step": 31550 }, { "epoch": 316.94, "learning_rate": 2.1346835443037978e-06, "loss": 0.0, "step": 31575 }, { "epoch": 317.19, "learning_rate": 2.128354430379747e-06, "loss": 0.0, "step": 31600 }, { "epoch": 317.44, "learning_rate": 2.1220253164556966e-06, "loss": 0.0, "step": 31625 }, { "epoch": 317.69, "learning_rate": 2.1156962025316456e-06, "loss": 0.0, "step": 31650 }, { "epoch": 317.94, "learning_rate": 2.109367088607595e-06, "loss": 0.0, "step": 31675 }, { "epoch": 318.19, "learning_rate": 2.1030379746835445e-06, "loss": 0.0, "step": 31700 }, { "epoch": 318.44, "learning_rate": 2.096708860759494e-06, "loss": 0.0, "step": 31725 }, { "epoch": 318.7, "learning_rate": 2.090379746835443e-06, "loss": 0.0, "step": 31750 }, { "epoch": 318.95, "learning_rate": 2.0840506329113928e-06, "loss": 0.0, "step": 31775 }, { "epoch": 319.2, "learning_rate": 2.077721518987342e-06, "loss": 0.0, "step": 31800 }, { "epoch": 319.45, "learning_rate": 2.0713924050632912e-06, "loss": 0.0, "step": 31825 }, { "epoch": 319.7, "learning_rate": 2.0650632911392406e-06, "loss": 0.0, "step": 31850 }, { "epoch": 319.95, "learning_rate": 2.05873417721519e-06, "loss": 0.0, "step": 31875 }, { "epoch": 320.2, "learning_rate": 2.0524050632911395e-06, "loss": 0.0, "step": 31900 }, { "epoch": 320.45, "learning_rate": 2.0460759493670885e-06, "loss": 0.0, "step": 31925 }, { "epoch": 320.7, "learning_rate": 2.039746835443038e-06, "loss": 0.0, "step": 31950 }, { "epoch": 320.95, "learning_rate": 2.0334177215189878e-06, "loss": 0.0, "step": 31975 }, { "epoch": 321.2, "learning_rate": 2.0270886075949368e-06, "loss": 0.0, "step": 32000 }, { "epoch": 321.2, "eval_loss": 0.3526245951652527, "eval_runtime": 2793.8043, "eval_samples_per_second": 4.561, "eval_steps_per_second": 0.285, "eval_wer": 6.870408003751758, "step": 32000 }, { "epoch": 321.46, "learning_rate": 2.020759493670886e-06, "loss": 0.0, "step": 32025 }, { "epoch": 321.71, "learning_rate": 2.0144303797468356e-06, "loss": 0.0, "step": 32050 }, { "epoch": 321.96, "learning_rate": 2.008101265822785e-06, "loss": 0.0, "step": 32075 }, { "epoch": 322.21, "learning_rate": 2.0017721518987345e-06, "loss": 0.0, "step": 32100 }, { "epoch": 322.46, "learning_rate": 1.9954430379746835e-06, "loss": 0.0, "step": 32125 }, { "epoch": 322.71, "learning_rate": 1.989367088607595e-06, "loss": 0.0, "step": 32150 }, { "epoch": 322.96, "learning_rate": 1.9830379746835444e-06, "loss": 0.0, "step": 32175 }, { "epoch": 323.21, "learning_rate": 1.976708860759494e-06, "loss": 0.0, "step": 32200 }, { "epoch": 323.46, "learning_rate": 1.9703797468354432e-06, "loss": 0.0, "step": 32225 }, { "epoch": 323.71, "learning_rate": 1.9640506329113927e-06, "loss": 0.0, "step": 32250 }, { "epoch": 323.96, "learning_rate": 1.957721518987342e-06, "loss": 0.0, "step": 32275 }, { "epoch": 324.22, "learning_rate": 1.951392405063291e-06, "loss": 0.0, "step": 32300 }, { "epoch": 324.47, "learning_rate": 1.9450632911392405e-06, "loss": 0.0, "step": 32325 }, { "epoch": 324.72, "learning_rate": 1.93873417721519e-06, "loss": 0.0, "step": 32350 }, { "epoch": 324.97, "learning_rate": 1.9324050632911394e-06, "loss": 0.0, "step": 32375 }, { "epoch": 325.22, "learning_rate": 1.926075949367089e-06, "loss": 0.0, "step": 32400 }, { "epoch": 325.47, "learning_rate": 1.9197468354430382e-06, "loss": 0.0, "step": 32425 }, { "epoch": 325.72, "learning_rate": 1.9134177215189877e-06, "loss": 0.0, "step": 32450 }, { "epoch": 325.97, "learning_rate": 1.9070886075949369e-06, "loss": 0.0, "step": 32475 }, { "epoch": 326.22, "learning_rate": 1.900759493670886e-06, "loss": 0.0, "step": 32500 }, { "epoch": 326.47, "learning_rate": 1.8944303797468355e-06, "loss": 0.0, "step": 32525 }, { "epoch": 326.73, "learning_rate": 1.888101265822785e-06, "loss": 0.0, "step": 32550 }, { "epoch": 326.98, "learning_rate": 1.8817721518987344e-06, "loss": 0.0, "step": 32575 }, { "epoch": 327.23, "learning_rate": 1.8754430379746838e-06, "loss": 0.0, "step": 32600 }, { "epoch": 327.48, "learning_rate": 1.8691139240506332e-06, "loss": 0.0, "step": 32625 }, { "epoch": 327.73, "learning_rate": 1.8627848101265824e-06, "loss": 0.0, "step": 32650 }, { "epoch": 327.98, "learning_rate": 1.8564556962025319e-06, "loss": 0.0, "step": 32675 }, { "epoch": 328.23, "learning_rate": 1.850126582278481e-06, "loss": 0.0, "step": 32700 }, { "epoch": 328.48, "learning_rate": 1.8437974683544305e-06, "loss": 0.0, "step": 32725 }, { "epoch": 328.73, "learning_rate": 1.8374683544303797e-06, "loss": 0.0, "step": 32750 }, { "epoch": 328.98, "learning_rate": 1.8311392405063292e-06, "loss": 0.0, "step": 32775 }, { "epoch": 329.23, "learning_rate": 1.8248101265822788e-06, "loss": 0.0, "step": 32800 }, { "epoch": 329.49, "learning_rate": 1.818481012658228e-06, "loss": 0.0, "step": 32825 }, { "epoch": 329.74, "learning_rate": 1.8121518987341774e-06, "loss": 0.0, "step": 32850 }, { "epoch": 329.99, "learning_rate": 1.8058227848101267e-06, "loss": 0.0, "step": 32875 }, { "epoch": 330.24, "learning_rate": 1.799493670886076e-06, "loss": 0.0, "step": 32900 }, { "epoch": 330.49, "learning_rate": 1.7931645569620253e-06, "loss": 0.0, "step": 32925 }, { "epoch": 330.74, "learning_rate": 1.7868354430379747e-06, "loss": 0.0, "step": 32950 }, { "epoch": 330.99, "learning_rate": 1.7805063291139242e-06, "loss": 0.0, "step": 32975 }, { "epoch": 331.24, "learning_rate": 1.7741772151898736e-06, "loss": 0.0, "step": 33000 }, { "epoch": 331.24, "eval_loss": 0.3569923937320709, "eval_runtime": 2780.874, "eval_samples_per_second": 4.582, "eval_steps_per_second": 0.287, "eval_wer": 6.883109269970299, "step": 33000 }, { "epoch": 331.49, "learning_rate": 1.767848101265823e-06, "loss": 0.0, "step": 33025 }, { "epoch": 331.74, "learning_rate": 1.7615189873417724e-06, "loss": 0.0, "step": 33050 }, { "epoch": 331.99, "learning_rate": 1.7551898734177217e-06, "loss": 0.0, "step": 33075 }, { "epoch": 332.25, "learning_rate": 1.748860759493671e-06, "loss": 0.0, "step": 33100 }, { "epoch": 332.5, "learning_rate": 1.7425316455696203e-06, "loss": 0.0, "step": 33125 }, { "epoch": 332.75, "learning_rate": 1.7362025316455697e-06, "loss": 0.0, "step": 33150 }, { "epoch": 333.0, "learning_rate": 1.729873417721519e-06, "loss": 0.0, "step": 33175 }, { "epoch": 333.25, "learning_rate": 1.7235443037974686e-06, "loss": 0.0, "step": 33200 }, { "epoch": 333.5, "learning_rate": 1.717215189873418e-06, "loss": 0.0, "step": 33225 }, { "epoch": 333.75, "learning_rate": 1.7108860759493672e-06, "loss": 0.0, "step": 33250 }, { "epoch": 334.0, "learning_rate": 1.7045569620253167e-06, "loss": 0.0, "step": 33275 }, { "epoch": 334.25, "learning_rate": 1.6982278481012659e-06, "loss": 0.0, "step": 33300 }, { "epoch": 334.5, "learning_rate": 1.6918987341772153e-06, "loss": 0.0, "step": 33325 }, { "epoch": 334.76, "learning_rate": 1.6855696202531645e-06, "loss": 0.0, "step": 33350 }, { "epoch": 335.01, "learning_rate": 1.679493670886076e-06, "loss": 0.0, "step": 33375 }, { "epoch": 335.26, "learning_rate": 1.6731645569620256e-06, "loss": 0.0, "step": 33400 }, { "epoch": 335.51, "learning_rate": 1.6668354430379748e-06, "loss": 0.0, "step": 33425 }, { "epoch": 335.76, "learning_rate": 1.6605063291139242e-06, "loss": 0.0, "step": 33450 }, { "epoch": 336.01, "learning_rate": 1.6541772151898735e-06, "loss": 0.0, "step": 33475 }, { "epoch": 336.26, "learning_rate": 1.6478481012658229e-06, "loss": 0.0, "step": 33500 }, { "epoch": 336.51, "learning_rate": 1.6415189873417723e-06, "loss": 0.0, "step": 33525 }, { "epoch": 336.76, "learning_rate": 1.6351898734177215e-06, "loss": 0.0, "step": 33550 }, { "epoch": 337.01, "learning_rate": 1.628860759493671e-06, "loss": 0.0, "step": 33575 }, { "epoch": 337.26, "learning_rate": 1.6225316455696206e-06, "loss": 0.0, "step": 33600 }, { "epoch": 337.52, "learning_rate": 1.6162025316455698e-06, "loss": 0.0, "step": 33625 }, { "epoch": 337.77, "learning_rate": 1.6098734177215192e-06, "loss": 0.0, "step": 33650 }, { "epoch": 338.02, "learning_rate": 1.6035443037974685e-06, "loss": 0.0, "step": 33675 }, { "epoch": 338.27, "learning_rate": 1.5972151898734179e-06, "loss": 0.0, "step": 33700 }, { "epoch": 338.52, "learning_rate": 1.590886075949367e-06, "loss": 0.0, "step": 33725 }, { "epoch": 338.77, "learning_rate": 1.5845569620253165e-06, "loss": 0.0, "step": 33750 }, { "epoch": 339.02, "learning_rate": 1.5782278481012657e-06, "loss": 0.0, "step": 33775 }, { "epoch": 339.27, "learning_rate": 1.5718987341772154e-06, "loss": 0.0, "step": 33800 }, { "epoch": 339.52, "learning_rate": 1.5655696202531648e-06, "loss": 0.0, "step": 33825 }, { "epoch": 339.77, "learning_rate": 1.559240506329114e-06, "loss": 0.0, "step": 33850 }, { "epoch": 340.03, "learning_rate": 1.5529113924050635e-06, "loss": 0.0, "step": 33875 }, { "epoch": 340.28, "learning_rate": 1.5465822784810127e-06, "loss": 0.0, "step": 33900 }, { "epoch": 340.53, "learning_rate": 1.540253164556962e-06, "loss": 0.0, "step": 33925 }, { "epoch": 340.78, "learning_rate": 1.5339240506329115e-06, "loss": 0.0, "step": 33950 }, { "epoch": 341.03, "learning_rate": 1.5275949367088607e-06, "loss": 0.0, "step": 33975 }, { "epoch": 341.28, "learning_rate": 1.5212658227848104e-06, "loss": 0.0, "step": 34000 }, { "epoch": 341.28, "eval_loss": 0.36138084530830383, "eval_runtime": 2780.6897, "eval_samples_per_second": 4.582, "eval_steps_per_second": 0.287, "eval_wer": 6.885063310926998, "step": 34000 }, { "epoch": 341.53, "learning_rate": 1.5149367088607598e-06, "loss": 0.0, "step": 34025 }, { "epoch": 341.78, "learning_rate": 1.508607594936709e-06, "loss": 0.0, "step": 34050 }, { "epoch": 342.03, "learning_rate": 1.5022784810126585e-06, "loss": 0.0, "step": 34075 }, { "epoch": 342.28, "learning_rate": 1.4959493670886077e-06, "loss": 0.0, "step": 34100 }, { "epoch": 342.53, "learning_rate": 1.489620253164557e-06, "loss": 0.0, "step": 34125 }, { "epoch": 342.79, "learning_rate": 1.4832911392405063e-06, "loss": 0.0, "step": 34150 }, { "epoch": 343.04, "learning_rate": 1.4769620253164557e-06, "loss": 0.0, "step": 34175 }, { "epoch": 343.29, "learning_rate": 1.470632911392405e-06, "loss": 0.0, "step": 34200 }, { "epoch": 343.54, "learning_rate": 1.4643037974683546e-06, "loss": 0.0, "step": 34225 }, { "epoch": 343.79, "learning_rate": 1.457974683544304e-06, "loss": 0.0, "step": 34250 }, { "epoch": 344.04, "learning_rate": 1.4516455696202532e-06, "loss": 0.0, "step": 34275 }, { "epoch": 344.29, "learning_rate": 1.4453164556962027e-06, "loss": 0.0, "step": 34300 }, { "epoch": 344.54, "learning_rate": 1.4389873417721519e-06, "loss": 0.0, "step": 34325 }, { "epoch": 344.79, "learning_rate": 1.4326582278481013e-06, "loss": 0.0, "step": 34350 }, { "epoch": 345.04, "learning_rate": 1.4263291139240507e-06, "loss": 0.0, "step": 34375 }, { "epoch": 345.29, "learning_rate": 1.42e-06, "loss": 0.0, "step": 34400 }, { "epoch": 345.55, "learning_rate": 1.4136708860759496e-06, "loss": 0.0, "step": 34425 }, { "epoch": 345.8, "learning_rate": 1.407341772151899e-06, "loss": 0.0, "step": 34450 }, { "epoch": 346.05, "learning_rate": 1.4010126582278482e-06, "loss": 0.0, "step": 34475 }, { "epoch": 346.3, "learning_rate": 1.3946835443037977e-06, "loss": 0.0, "step": 34500 }, { "epoch": 346.55, "learning_rate": 1.3883544303797469e-06, "loss": 0.0, "step": 34525 }, { "epoch": 346.8, "learning_rate": 1.3820253164556963e-06, "loss": 0.0, "step": 34550 }, { "epoch": 347.05, "learning_rate": 1.3756962025316455e-06, "loss": 0.0, "step": 34575 }, { "epoch": 347.3, "learning_rate": 1.369367088607595e-06, "loss": 0.0, "step": 34600 }, { "epoch": 347.55, "learning_rate": 1.3630379746835446e-06, "loss": 0.0, "step": 34625 }, { "epoch": 347.8, "learning_rate": 1.3567088607594938e-06, "loss": 0.0, "step": 34650 }, { "epoch": 348.06, "learning_rate": 1.3503797468354432e-06, "loss": 0.0, "step": 34675 }, { "epoch": 348.31, "learning_rate": 1.3440506329113925e-06, "loss": 0.0, "step": 34700 }, { "epoch": 348.56, "learning_rate": 1.3377215189873419e-06, "loss": 0.0, "step": 34725 }, { "epoch": 348.81, "learning_rate": 1.3313924050632913e-06, "loss": 0.0, "step": 34750 }, { "epoch": 349.06, "learning_rate": 1.3250632911392405e-06, "loss": 0.0, "step": 34775 }, { "epoch": 349.31, "learning_rate": 1.31873417721519e-06, "loss": 0.0, "step": 34800 }, { "epoch": 349.56, "learning_rate": 1.3124050632911394e-06, "loss": 0.0, "step": 34825 }, { "epoch": 349.81, "learning_rate": 1.3060759493670888e-06, "loss": 0.0, "step": 34850 }, { "epoch": 350.06, "learning_rate": 1.2997468354430382e-06, "loss": 0.0, "step": 34875 }, { "epoch": 350.31, "learning_rate": 1.2934177215189875e-06, "loss": 0.0, "step": 34900 }, { "epoch": 350.56, "learning_rate": 1.2870886075949369e-06, "loss": 0.0, "step": 34925 }, { "epoch": 350.82, "learning_rate": 1.280759493670886e-06, "loss": 0.0, "step": 34950 }, { "epoch": 351.07, "learning_rate": 1.2744303797468355e-06, "loss": 0.0, "step": 34975 }, { "epoch": 351.32, "learning_rate": 1.2681012658227847e-06, "loss": 0.0, "step": 35000 }, { "epoch": 351.32, "eval_loss": 0.36445754766464233, "eval_runtime": 2782.2502, "eval_samples_per_second": 4.58, "eval_steps_per_second": 0.286, "eval_wer": 6.878224167578552, "step": 35000 }, { "epoch": 351.57, "learning_rate": 1.2617721518987342e-06, "loss": 0.0, "step": 35025 }, { "epoch": 351.82, "learning_rate": 1.2554430379746838e-06, "loss": 0.0, "step": 35050 }, { "epoch": 352.07, "learning_rate": 1.249113924050633e-06, "loss": 0.0, "step": 35075 }, { "epoch": 352.32, "learning_rate": 1.2427848101265824e-06, "loss": 0.0, "step": 35100 }, { "epoch": 352.57, "learning_rate": 1.2364556962025317e-06, "loss": 0.0, "step": 35125 }, { "epoch": 352.82, "learning_rate": 1.230126582278481e-06, "loss": 0.0, "step": 35150 }, { "epoch": 353.07, "learning_rate": 1.2237974683544305e-06, "loss": 0.0, "step": 35175 }, { "epoch": 353.32, "learning_rate": 1.21746835443038e-06, "loss": 0.0, "step": 35200 }, { "epoch": 353.58, "learning_rate": 1.2111392405063292e-06, "loss": 0.0, "step": 35225 }, { "epoch": 353.83, "learning_rate": 1.2048101265822786e-06, "loss": 0.0, "step": 35250 }, { "epoch": 354.08, "learning_rate": 1.1984810126582278e-06, "loss": 0.0, "step": 35275 }, { "epoch": 354.33, "learning_rate": 1.1921518987341774e-06, "loss": 0.0, "step": 35300 }, { "epoch": 354.58, "learning_rate": 1.1858227848101267e-06, "loss": 0.0, "step": 35325 }, { "epoch": 354.83, "learning_rate": 1.179493670886076e-06, "loss": 0.0, "step": 35350 }, { "epoch": 355.08, "learning_rate": 1.1731645569620253e-06, "loss": 0.0, "step": 35375 }, { "epoch": 355.33, "learning_rate": 1.1668354430379747e-06, "loss": 0.0, "step": 35400 }, { "epoch": 355.58, "learning_rate": 1.1605063291139242e-06, "loss": 0.0, "step": 35425 }, { "epoch": 355.83, "learning_rate": 1.1541772151898736e-06, "loss": 0.0, "step": 35450 }, { "epoch": 356.09, "learning_rate": 1.1478481012658228e-06, "loss": 0.0, "step": 35475 }, { "epoch": 356.34, "learning_rate": 1.1415189873417722e-06, "loss": 0.0, "step": 35500 }, { "epoch": 356.59, "learning_rate": 1.1351898734177217e-06, "loss": 0.0, "step": 35525 }, { "epoch": 356.84, "learning_rate": 1.1288607594936709e-06, "loss": 0.0, "step": 35550 }, { "epoch": 357.09, "learning_rate": 1.1225316455696203e-06, "loss": 0.0, "step": 35575 }, { "epoch": 357.34, "learning_rate": 1.1162025316455697e-06, "loss": 0.0, "step": 35600 }, { "epoch": 357.59, "learning_rate": 1.1098734177215192e-06, "loss": 0.0, "step": 35625 }, { "epoch": 357.84, "learning_rate": 1.1035443037974684e-06, "loss": 0.0, "step": 35650 }, { "epoch": 358.09, "learning_rate": 1.0972151898734178e-06, "loss": 0.0, "step": 35675 }, { "epoch": 358.34, "learning_rate": 1.0908860759493672e-06, "loss": 0.0, "step": 35700 }, { "epoch": 358.59, "learning_rate": 1.0845569620253167e-06, "loss": 0.0, "step": 35725 }, { "epoch": 358.85, "learning_rate": 1.0782278481012659e-06, "loss": 0.0, "step": 35750 }, { "epoch": 359.1, "learning_rate": 1.0718987341772153e-06, "loss": 0.0, "step": 35775 }, { "epoch": 359.35, "learning_rate": 1.0655696202531647e-06, "loss": 0.0, "step": 35800 }, { "epoch": 359.6, "learning_rate": 1.059240506329114e-06, "loss": 0.0, "step": 35825 }, { "epoch": 359.85, "learning_rate": 1.0529113924050634e-06, "loss": 0.0, "step": 35850 }, { "epoch": 360.1, "learning_rate": 1.0465822784810128e-06, "loss": 0.0, "step": 35875 }, { "epoch": 360.35, "learning_rate": 1.040253164556962e-06, "loss": 0.0, "step": 35900 }, { "epoch": 360.6, "learning_rate": 1.0339240506329114e-06, "loss": 0.0, "step": 35925 }, { "epoch": 360.85, "learning_rate": 1.0275949367088609e-06, "loss": 0.0, "step": 35950 }, { "epoch": 361.1, "learning_rate": 1.02126582278481e-06, "loss": 0.0, "step": 35975 }, { "epoch": 361.36, "learning_rate": 1.0149367088607595e-06, "loss": 0.0, "step": 36000 }, { "epoch": 361.36, "eval_loss": 0.3663193881511688, "eval_runtime": 2812.4519, "eval_samples_per_second": 4.531, "eval_steps_per_second": 0.283, "eval_wer": 6.871385024230108, "step": 36000 }, { "epoch": 361.61, "learning_rate": 1.008607594936709e-06, "loss": 0.0, "step": 36025 }, { "epoch": 361.86, "learning_rate": 1.0022784810126584e-06, "loss": 0.0, "step": 36050 }, { "epoch": 362.11, "learning_rate": 9.959493670886076e-07, "loss": 0.0, "step": 36075 }, { "epoch": 362.36, "learning_rate": 9.89620253164557e-07, "loss": 0.0, "step": 36100 }, { "epoch": 362.61, "learning_rate": 9.832911392405064e-07, "loss": 0.0, "step": 36125 }, { "epoch": 362.86, "learning_rate": 9.769620253164559e-07, "loss": 0.0, "step": 36150 }, { "epoch": 363.11, "learning_rate": 9.70632911392405e-07, "loss": 0.0, "step": 36175 }, { "epoch": 363.36, "learning_rate": 9.643037974683545e-07, "loss": 0.0, "step": 36200 }, { "epoch": 363.61, "learning_rate": 9.57974683544304e-07, "loss": 0.0, "step": 36225 }, { "epoch": 363.86, "learning_rate": 9.516455696202533e-07, "loss": 0.0, "step": 36250 }, { "epoch": 364.12, "learning_rate": 9.453164556962026e-07, "loss": 0.0, "step": 36275 }, { "epoch": 364.37, "learning_rate": 9.389873417721519e-07, "loss": 0.0, "step": 36300 }, { "epoch": 364.62, "learning_rate": 9.326582278481014e-07, "loss": 0.0, "step": 36325 }, { "epoch": 364.87, "learning_rate": 9.263291139240508e-07, "loss": 0.0, "step": 36350 }, { "epoch": 365.12, "learning_rate": 9.200000000000001e-07, "loss": 0.0, "step": 36375 }, { "epoch": 365.37, "learning_rate": 9.136708860759494e-07, "loss": 0.0, "step": 36400 }, { "epoch": 365.62, "learning_rate": 9.073417721518988e-07, "loss": 0.0, "step": 36425 }, { "epoch": 365.87, "learning_rate": 9.010126582278482e-07, "loss": 0.0, "step": 36450 }, { "epoch": 366.12, "learning_rate": 8.946835443037975e-07, "loss": 0.0, "step": 36475 }, { "epoch": 366.37, "learning_rate": 8.883544303797469e-07, "loss": 0.0, "step": 36500 }, { "epoch": 366.62, "learning_rate": 8.820253164556963e-07, "loss": 0.0, "step": 36525 }, { "epoch": 366.88, "learning_rate": 8.756962025316457e-07, "loss": 0.0, "step": 36550 }, { "epoch": 367.13, "learning_rate": 8.69367088607595e-07, "loss": 0.0, "step": 36575 }, { "epoch": 367.38, "learning_rate": 8.630379746835443e-07, "loss": 0.0, "step": 36600 }, { "epoch": 367.63, "learning_rate": 8.567088607594938e-07, "loss": 0.0, "step": 36625 }, { "epoch": 367.88, "learning_rate": 8.503797468354432e-07, "loss": 0.0, "step": 36650 }, { "epoch": 368.13, "learning_rate": 8.440506329113925e-07, "loss": 0.0, "step": 36675 }, { "epoch": 368.38, "learning_rate": 8.377215189873418e-07, "loss": 0.0, "step": 36700 }, { "epoch": 368.63, "learning_rate": 8.313924050632912e-07, "loss": 0.0, "step": 36725 }, { "epoch": 368.88, "learning_rate": 8.250632911392407e-07, "loss": 0.0, "step": 36750 }, { "epoch": 369.13, "learning_rate": 8.1873417721519e-07, "loss": 0.0, "step": 36775 }, { "epoch": 369.39, "learning_rate": 8.124050632911393e-07, "loss": 0.0, "step": 36800 }, { "epoch": 369.64, "learning_rate": 8.060759493670886e-07, "loss": 0.0, "step": 36825 }, { "epoch": 369.89, "learning_rate": 7.99746835443038e-07, "loss": 0.0, "step": 36850 }, { "epoch": 370.14, "learning_rate": 7.934177215189874e-07, "loss": 0.0, "step": 36875 }, { "epoch": 370.39, "learning_rate": 7.870886075949367e-07, "loss": 0.0, "step": 36900 }, { "epoch": 370.64, "learning_rate": 7.807594936708861e-07, "loss": 0.0, "step": 36925 }, { "epoch": 370.89, "learning_rate": 7.744303797468355e-07, "loss": 0.0, "step": 36950 }, { "epoch": 371.14, "learning_rate": 7.681012658227849e-07, "loss": 0.0, "step": 36975 }, { "epoch": 371.39, "learning_rate": 7.617721518987342e-07, "loss": 0.0, "step": 37000 }, { "epoch": 371.39, "eval_loss": 0.3676894009113312, "eval_runtime": 2780.5773, "eval_samples_per_second": 4.583, "eval_steps_per_second": 0.287, "eval_wer": 6.867476942316711, "step": 37000 }, { "epoch": 371.64, "learning_rate": 7.554430379746835e-07, "loss": 0.0, "step": 37025 }, { "epoch": 371.89, "learning_rate": 7.49113924050633e-07, "loss": 0.0, "step": 37050 }, { "epoch": 372.15, "learning_rate": 7.427848101265824e-07, "loss": 0.0, "step": 37075 }, { "epoch": 372.4, "learning_rate": 7.364556962025317e-07, "loss": 0.0, "step": 37100 }, { "epoch": 372.65, "learning_rate": 7.30126582278481e-07, "loss": 0.0, "step": 37125 }, { "epoch": 372.9, "learning_rate": 7.237974683544304e-07, "loss": 0.0, "step": 37150 }, { "epoch": 373.15, "learning_rate": 7.174683544303799e-07, "loss": 0.0, "step": 37175 }, { "epoch": 373.4, "learning_rate": 7.111392405063292e-07, "loss": 0.0, "step": 37200 }, { "epoch": 373.65, "learning_rate": 7.048101265822785e-07, "loss": 0.0, "step": 37225 }, { "epoch": 373.9, "learning_rate": 6.984810126582279e-07, "loss": 0.0, "step": 37250 }, { "epoch": 374.15, "learning_rate": 6.921518987341773e-07, "loss": 0.0, "step": 37275 }, { "epoch": 374.4, "learning_rate": 6.858227848101266e-07, "loss": 0.0, "step": 37300 }, { "epoch": 374.65, "learning_rate": 6.79493670886076e-07, "loss": 0.0, "step": 37325 }, { "epoch": 374.91, "learning_rate": 6.731645569620254e-07, "loss": 0.0, "step": 37350 }, { "epoch": 375.16, "learning_rate": 6.668354430379748e-07, "loss": 0.0, "step": 37375 }, { "epoch": 375.41, "learning_rate": 6.605063291139241e-07, "loss": 0.0, "step": 37400 }, { "epoch": 375.66, "learning_rate": 6.541772151898734e-07, "loss": 0.0, "step": 37425 }, { "epoch": 375.91, "learning_rate": 6.478481012658229e-07, "loss": 0.0, "step": 37450 }, { "epoch": 376.16, "learning_rate": 6.415189873417723e-07, "loss": 0.0, "step": 37475 }, { "epoch": 376.41, "learning_rate": 6.351898734177216e-07, "loss": 0.0, "step": 37500 }, { "epoch": 376.66, "learning_rate": 6.288607594936709e-07, "loss": 0.0, "step": 37525 }, { "epoch": 376.91, "learning_rate": 6.225316455696203e-07, "loss": 0.0, "step": 37550 }, { "epoch": 377.16, "learning_rate": 6.162025316455696e-07, "loss": 0.0, "step": 37575 }, { "epoch": 377.42, "learning_rate": 6.098734177215191e-07, "loss": 0.0, "step": 37600 }, { "epoch": 377.67, "learning_rate": 6.035443037974684e-07, "loss": 0.0, "step": 37625 }, { "epoch": 377.92, "learning_rate": 5.972151898734177e-07, "loss": 0.0, "step": 37650 }, { "epoch": 378.17, "learning_rate": 5.908860759493671e-07, "loss": 0.0, "step": 37675 }, { "epoch": 378.42, "learning_rate": 5.845569620253165e-07, "loss": 0.0, "step": 37700 }, { "epoch": 378.67, "learning_rate": 5.782278481012658e-07, "loss": 0.0, "step": 37725 }, { "epoch": 378.92, "learning_rate": 5.718987341772152e-07, "loss": 0.0, "step": 37750 }, { "epoch": 379.17, "learning_rate": 5.655696202531645e-07, "loss": 0.0, "step": 37775 }, { "epoch": 379.42, "learning_rate": 5.59240506329114e-07, "loss": 0.0, "step": 37800 }, { "epoch": 379.67, "learning_rate": 5.529113924050633e-07, "loss": 0.0, "step": 37825 }, { "epoch": 379.92, "learning_rate": 5.465822784810127e-07, "loss": 0.0, "step": 37850 }, { "epoch": 380.18, "learning_rate": 5.40253164556962e-07, "loss": 0.0, "step": 37875 }, { "epoch": 380.43, "learning_rate": 5.339240506329115e-07, "loss": 0.0, "step": 37900 }, { "epoch": 380.68, "learning_rate": 5.275949367088608e-07, "loss": 0.0, "step": 37925 }, { "epoch": 380.93, "learning_rate": 5.212658227848102e-07, "loss": 0.0, "step": 37950 }, { "epoch": 381.18, "learning_rate": 5.149367088607595e-07, "loss": 0.0, "step": 37975 }, { "epoch": 381.43, "learning_rate": 5.08607594936709e-07, "loss": 0.0, "step": 38000 }, { "epoch": 381.43, "eval_loss": 0.3681415617465973, "eval_runtime": 2779.0085, "eval_samples_per_second": 4.585, "eval_steps_per_second": 0.287, "eval_wer": 6.880178208535251, "step": 38000 }, { "epoch": 381.68, "learning_rate": 5.022784810126583e-07, "loss": 0.0, "step": 38025 }, { "epoch": 381.93, "learning_rate": 4.959493670886076e-07, "loss": 0.0, "step": 38050 }, { "epoch": 382.18, "learning_rate": 4.896202531645569e-07, "loss": 0.0, "step": 38075 }, { "epoch": 382.43, "learning_rate": 4.832911392405064e-07, "loss": 0.0, "step": 38100 }, { "epoch": 382.69, "learning_rate": 4.769620253164557e-07, "loss": 0.0, "step": 38125 }, { "epoch": 382.94, "learning_rate": 4.706329113924051e-07, "loss": 0.0, "step": 38150 }, { "epoch": 383.19, "learning_rate": 4.6430379746835443e-07, "loss": 0.0, "step": 38175 }, { "epoch": 383.44, "learning_rate": 4.5797468354430386e-07, "loss": 0.0, "step": 38200 }, { "epoch": 383.69, "learning_rate": 4.516455696202532e-07, "loss": 0.0, "step": 38225 }, { "epoch": 383.94, "learning_rate": 4.453164556962026e-07, "loss": 0.0, "step": 38250 }, { "epoch": 384.19, "learning_rate": 4.3898734177215193e-07, "loss": 0.0, "step": 38275 }, { "epoch": 384.44, "learning_rate": 4.326582278481013e-07, "loss": 0.0, "step": 38300 }, { "epoch": 384.69, "learning_rate": 4.263291139240507e-07, "loss": 0.0, "step": 38325 }, { "epoch": 384.94, "learning_rate": 4.2000000000000006e-07, "loss": 0.0, "step": 38350 }, { "epoch": 385.19, "learning_rate": 4.136708860759494e-07, "loss": 0.0, "step": 38375 }, { "epoch": 385.45, "learning_rate": 4.073417721518988e-07, "loss": 0.0, "step": 38400 }, { "epoch": 385.7, "learning_rate": 4.010126582278481e-07, "loss": 0.0, "step": 38425 }, { "epoch": 385.95, "learning_rate": 3.9468354430379755e-07, "loss": 0.0, "step": 38450 }, { "epoch": 386.2, "learning_rate": 3.883544303797469e-07, "loss": 0.0, "step": 38475 }, { "epoch": 386.45, "learning_rate": 3.8202531645569625e-07, "loss": 0.0, "step": 38500 }, { "epoch": 386.7, "learning_rate": 3.7569620253164557e-07, "loss": 0.0, "step": 38525 }, { "epoch": 386.95, "learning_rate": 3.69367088607595e-07, "loss": 0.0, "step": 38550 }, { "epoch": 387.2, "learning_rate": 3.630379746835443e-07, "loss": 0.0, "step": 38575 }, { "epoch": 387.45, "learning_rate": 3.5670886075949364e-07, "loss": 0.0, "step": 38600 }, { "epoch": 387.7, "learning_rate": 3.5037974683544307e-07, "loss": 0.0, "step": 38625 }, { "epoch": 387.95, "learning_rate": 3.440506329113924e-07, "loss": 0.0, "step": 38650 }, { "epoch": 388.21, "learning_rate": 3.377215189873418e-07, "loss": 0.0, "step": 38675 }, { "epoch": 388.46, "learning_rate": 3.3139240506329114e-07, "loss": 0.0, "step": 38700 }, { "epoch": 388.71, "learning_rate": 3.250632911392405e-07, "loss": 0.0, "step": 38725 }, { "epoch": 388.96, "learning_rate": 3.187341772151899e-07, "loss": 0.0, "step": 38750 }, { "epoch": 389.21, "learning_rate": 3.1240506329113927e-07, "loss": 0.0, "step": 38775 }, { "epoch": 389.46, "learning_rate": 3.0607594936708864e-07, "loss": 0.0, "step": 38800 }, { "epoch": 389.71, "learning_rate": 2.99746835443038e-07, "loss": 0.0, "step": 38825 }, { "epoch": 389.96, "learning_rate": 2.934177215189874e-07, "loss": 0.0, "step": 38850 }, { "epoch": 390.21, "learning_rate": 2.8708860759493677e-07, "loss": 0.0, "step": 38875 }, { "epoch": 390.46, "learning_rate": 2.807594936708861e-07, "loss": 0.0, "step": 38900 }, { "epoch": 390.72, "learning_rate": 2.7443037974683546e-07, "loss": 0.0, "step": 38925 }, { "epoch": 390.97, "learning_rate": 2.6810126582278484e-07, "loss": 0.0, "step": 38950 }, { "epoch": 391.22, "learning_rate": 2.617721518987342e-07, "loss": 0.0, "step": 38975 }, { "epoch": 391.47, "learning_rate": 2.554430379746836e-07, "loss": 0.0, "step": 39000 }, { "epoch": 391.47, "eval_loss": 0.36862120032310486, "eval_runtime": 2786.3432, "eval_samples_per_second": 4.573, "eval_steps_per_second": 0.286, "eval_wer": 6.887994372362044, "step": 39000 }, { "epoch": 391.72, "learning_rate": 2.491139240506329e-07, "loss": 0.0, "step": 39025 }, { "epoch": 391.97, "learning_rate": 2.427848101265823e-07, "loss": 0.0, "step": 39050 }, { "epoch": 392.22, "learning_rate": 2.3645569620253166e-07, "loss": 0.0, "step": 39075 }, { "epoch": 392.47, "learning_rate": 2.30126582278481e-07, "loss": 0.0, "step": 39100 }, { "epoch": 392.72, "learning_rate": 2.2379746835443038e-07, "loss": 0.0, "step": 39125 }, { "epoch": 392.97, "learning_rate": 2.1746835443037976e-07, "loss": 0.0, "step": 39150 }, { "epoch": 393.22, "learning_rate": 2.1113924050632913e-07, "loss": 0.0, "step": 39175 }, { "epoch": 393.48, "learning_rate": 2.0481012658227848e-07, "loss": 0.0, "step": 39200 }, { "epoch": 393.73, "learning_rate": 1.9848101265822785e-07, "loss": 0.0, "step": 39225 }, { "epoch": 393.98, "learning_rate": 1.9215189873417723e-07, "loss": 0.0, "step": 39250 }, { "epoch": 394.23, "learning_rate": 1.858227848101266e-07, "loss": 0.0, "step": 39275 }, { "epoch": 394.48, "learning_rate": 1.7949367088607595e-07, "loss": 0.0, "step": 39300 }, { "epoch": 394.73, "learning_rate": 1.7316455696202533e-07, "loss": 0.0, "step": 39325 }, { "epoch": 394.98, "learning_rate": 1.668354430379747e-07, "loss": 0.0, "step": 39350 }, { "epoch": 395.23, "learning_rate": 1.6050632911392405e-07, "loss": 0.0, "step": 39375 }, { "epoch": 395.48, "learning_rate": 1.5417721518987342e-07, "loss": 0.0, "step": 39400 }, { "epoch": 395.73, "learning_rate": 1.4810126582278482e-07, "loss": 0.0, "step": 39425 }, { "epoch": 395.98, "learning_rate": 1.417721518987342e-07, "loss": 0.0, "step": 39450 }, { "epoch": 396.24, "learning_rate": 1.3544303797468354e-07, "loss": 0.0, "step": 39475 }, { "epoch": 396.49, "learning_rate": 1.2911392405063292e-07, "loss": 0.0, "step": 39500 }, { "epoch": 396.74, "learning_rate": 1.227848101265823e-07, "loss": 0.0, "step": 39525 }, { "epoch": 396.99, "learning_rate": 1.1645569620253166e-07, "loss": 0.0, "step": 39550 }, { "epoch": 397.24, "learning_rate": 1.1012658227848102e-07, "loss": 0.0, "step": 39575 }, { "epoch": 397.49, "learning_rate": 1.0379746835443039e-07, "loss": 0.0, "step": 39600 }, { "epoch": 397.74, "learning_rate": 9.746835443037975e-08, "loss": 0.0, "step": 39625 }, { "epoch": 397.99, "learning_rate": 9.113924050632913e-08, "loss": 0.0, "step": 39650 }, { "epoch": 398.24, "learning_rate": 8.481012658227849e-08, "loss": 0.0, "step": 39675 }, { "epoch": 398.49, "learning_rate": 7.848101265822786e-08, "loss": 0.0, "step": 39700 }, { "epoch": 398.75, "learning_rate": 7.215189873417721e-08, "loss": 0.0, "step": 39725 }, { "epoch": 399.0, "learning_rate": 6.582278481012659e-08, "loss": 0.0, "step": 39750 }, { "epoch": 399.25, "learning_rate": 5.9493670886075955e-08, "loss": 0.0, "step": 39775 }, { "epoch": 399.5, "learning_rate": 5.3164556962025324e-08, "loss": 0.0, "step": 39800 }, { "epoch": 399.75, "learning_rate": 4.683544303797469e-08, "loss": 0.0, "step": 39825 }, { "epoch": 400.0, "learning_rate": 4.050632911392405e-08, "loss": 0.0, "step": 39850 }, { "epoch": 400.25, "learning_rate": 3.417721518987342e-08, "loss": 0.0, "step": 39875 }, { "epoch": 400.5, "learning_rate": 2.7848101265822786e-08, "loss": 0.0, "step": 39900 }, { "epoch": 400.75, "learning_rate": 2.1518987341772154e-08, "loss": 0.0, "step": 39925 }, { "epoch": 401.0, "learning_rate": 1.518987341772152e-08, "loss": 0.0, "step": 39950 }, { "epoch": 401.25, "learning_rate": 8.860759493670887e-09, "loss": 0.0, "step": 39975 }, { "epoch": 401.51, "learning_rate": 2.5316455696202533e-09, "loss": 0.0, "step": 40000 }, { "epoch": 401.51, "eval_loss": 0.36875849962234497, "eval_runtime": 2798.7548, "eval_samples_per_second": 4.553, "eval_steps_per_second": 0.285, "eval_wer": 6.887994372362044, "step": 40000 }, { "epoch": 401.51, "step": 40000, "total_flos": 3.476313472275579e+22, "train_loss": 0.003312818900368211, "train_runtime": 987758.8877, "train_samples_per_second": 10.367, "train_steps_per_second": 0.04 } ], "logging_steps": 25, "max_steps": 40000, "num_input_tokens_seen": 0, "num_train_epochs": 405, "save_steps": 1000, "total_flos": 3.476313472275579e+22, "train_batch_size": 32, "trial_name": null, "trial_params": null }