{ "best_metric": 0.35683760683760685, "best_model_checkpoint": "/content/drive/MyDrive/model_ft/checkpoint-60000", "epoch": 49.01959583588487, "global_step": 60000, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.08, "learning_rate": 8e-08, "loss": 17.0818, "step": 100 }, { "epoch": 0.16, "learning_rate": 1.6333333333333334e-07, "loss": 15.2157, "step": 200 }, { "epoch": 0.24, "learning_rate": 2.466666666666667e-07, "loss": 15.0857, "step": 300 }, { "epoch": 0.33, "learning_rate": 3.3e-07, "loss": 15.4181, "step": 400 }, { "epoch": 0.41, "learning_rate": 4.133333333333334e-07, "loss": 15.3673, "step": 500 }, { "epoch": 0.49, "learning_rate": 4.958333333333334e-07, "loss": 15.3838, "step": 600 }, { "epoch": 0.57, "learning_rate": 5.791666666666667e-07, "loss": 15.0048, "step": 700 }, { "epoch": 0.65, "learning_rate": 6.625000000000001e-07, "loss": 13.8472, "step": 800 }, { "epoch": 0.73, "learning_rate": 7.458333333333334e-07, "loss": 14.4646, "step": 900 }, { "epoch": 0.82, "learning_rate": 8.291666666666667e-07, "loss": 11.767, "step": 1000 }, { "epoch": 0.9, "learning_rate": 9.125e-07, "loss": 10.0359, "step": 1100 }, { "epoch": 0.98, "learning_rate": 9.958333333333335e-07, "loss": 8.9849, "step": 1200 }, { "epoch": 1.06, "learning_rate": 1.0791666666666668e-06, "loss": 7.6915, "step": 1300 }, { "epoch": 1.14, "learning_rate": 1.1625e-06, "loss": 7.4909, "step": 1400 }, { "epoch": 1.23, "learning_rate": 1.2458333333333334e-06, "loss": 6.3382, "step": 1500 }, { "epoch": 1.31, "learning_rate": 1.3291666666666667e-06, "loss": 6.2147, "step": 1600 }, { "epoch": 1.39, "learning_rate": 1.4125e-06, "loss": 5.7939, "step": 1700 }, { "epoch": 1.47, "learning_rate": 1.4958333333333336e-06, "loss": 5.4857, "step": 1800 }, { "epoch": 1.55, "learning_rate": 1.5791666666666669e-06, "loss": 5.3803, "step": 1900 }, { "epoch": 1.63, "learning_rate": 1.6625000000000002e-06, "loss": 5.2082, "step": 2000 }, { "epoch": 1.72, "learning_rate": 1.7458333333333335e-06, "loss": 4.7779, "step": 2100 }, { "epoch": 1.8, "learning_rate": 1.8291666666666668e-06, "loss": 4.7639, "step": 2200 }, { "epoch": 1.88, "learning_rate": 1.9125000000000003e-06, "loss": 4.516, "step": 2300 }, { "epoch": 1.96, "learning_rate": 1.9958333333333334e-06, "loss": 4.5625, "step": 2400 }, { "epoch": 2.04, "learning_rate": 2.079166666666667e-06, "loss": 4.5228, "step": 2500 }, { "epoch": 2.12, "learning_rate": 2.1625e-06, "loss": 4.2219, "step": 2600 }, { "epoch": 2.21, "learning_rate": 2.2458333333333336e-06, "loss": 4.1294, "step": 2700 }, { "epoch": 2.29, "learning_rate": 2.3291666666666667e-06, "loss": 4.2024, "step": 2800 }, { "epoch": 2.37, "learning_rate": 2.4125e-06, "loss": 3.9283, "step": 2900 }, { "epoch": 2.45, "learning_rate": 2.4958333333333333e-06, "loss": 3.8443, "step": 3000 }, { "epoch": 2.53, "learning_rate": 2.5791666666666672e-06, "loss": 3.8079, "step": 3100 }, { "epoch": 2.61, "learning_rate": 2.6625e-06, "loss": 3.6979, "step": 3200 }, { "epoch": 2.7, "learning_rate": 2.745833333333334e-06, "loss": 3.7727, "step": 3300 }, { "epoch": 2.78, "learning_rate": 2.8291666666666665e-06, "loss": 3.5884, "step": 3400 }, { "epoch": 2.86, "learning_rate": 2.9125000000000005e-06, "loss": 3.6409, "step": 3500 }, { "epoch": 2.94, "learning_rate": 2.995833333333333e-06, "loss": 3.5832, "step": 3600 }, { "epoch": 3.02, "learning_rate": 3.079166666666667e-06, "loss": 3.5055, "step": 3700 }, { "epoch": 3.1, "learning_rate": 3.1625000000000002e-06, "loss": 3.5141, "step": 3800 }, { "epoch": 3.19, "learning_rate": 3.2458333333333337e-06, "loss": 3.4218, "step": 3900 }, { "epoch": 3.27, "learning_rate": 3.3291666666666673e-06, "loss": 3.4604, "step": 4000 }, { "epoch": 3.35, "learning_rate": 3.4125000000000004e-06, "loss": 3.4134, "step": 4100 }, { "epoch": 3.43, "learning_rate": 3.495833333333334e-06, "loss": 3.3811, "step": 4200 }, { "epoch": 3.51, "learning_rate": 3.579166666666667e-06, "loss": 3.4506, "step": 4300 }, { "epoch": 3.59, "learning_rate": 3.6625000000000005e-06, "loss": 3.3525, "step": 4400 }, { "epoch": 3.68, "learning_rate": 3.7458333333333336e-06, "loss": 3.3783, "step": 4500 }, { "epoch": 3.76, "learning_rate": 3.829166666666667e-06, "loss": 3.3088, "step": 4600 }, { "epoch": 3.84, "learning_rate": 3.9125e-06, "loss": 3.3303, "step": 4700 }, { "epoch": 3.92, "learning_rate": 3.995833333333334e-06, "loss": 3.3213, "step": 4800 }, { "epoch": 4.0, "learning_rate": 4.0791666666666664e-06, "loss": 3.3047, "step": 4900 }, { "epoch": 4.08, "learning_rate": 4.1625e-06, "loss": 3.2773, "step": 5000 }, { "epoch": 4.17, "learning_rate": 4.2458333333333335e-06, "loss": 3.2719, "step": 5100 }, { "epoch": 4.25, "learning_rate": 4.3291666666666674e-06, "loss": 3.2904, "step": 5200 }, { "epoch": 4.33, "learning_rate": 4.4125000000000005e-06, "loss": 3.2518, "step": 5300 }, { "epoch": 4.41, "learning_rate": 4.495833333333334e-06, "loss": 3.2756, "step": 5400 }, { "epoch": 4.49, "learning_rate": 4.5791666666666676e-06, "loss": 3.3096, "step": 5500 }, { "epoch": 4.57, "learning_rate": 4.662500000000001e-06, "loss": 3.2517, "step": 5600 }, { "epoch": 4.66, "learning_rate": 4.745833333333334e-06, "loss": 3.2298, "step": 5700 }, { "epoch": 4.74, "learning_rate": 4.829166666666667e-06, "loss": 3.2291, "step": 5800 }, { "epoch": 4.82, "learning_rate": 4.912500000000001e-06, "loss": 3.2146, "step": 5900 }, { "epoch": 4.9, "learning_rate": 4.995833333333334e-06, "loss": 3.1858, "step": 6000 }, { "epoch": 4.98, "learning_rate": 5e-06, "loss": 3.1731, "step": 6100 }, { "epoch": 5.07, "learning_rate": 5e-06, "loss": 3.1668, "step": 6200 }, { "epoch": 5.15, "learning_rate": 5e-06, "loss": 3.1453, "step": 6300 }, { "epoch": 5.23, "learning_rate": 5e-06, "loss": 3.0692, "step": 6400 }, { "epoch": 5.31, "learning_rate": 5e-06, "loss": 3.0427, "step": 6500 }, { "epoch": 5.39, "learning_rate": 5e-06, "loss": 2.9999, "step": 6600 }, { "epoch": 5.47, "learning_rate": 5e-06, "loss": 2.9585, "step": 6700 }, { "epoch": 5.56, "learning_rate": 5e-06, "loss": 2.8819, "step": 6800 }, { "epoch": 5.64, "learning_rate": 5e-06, "loss": 2.8528, "step": 6900 }, { "epoch": 5.72, "learning_rate": 5e-06, "loss": 2.8175, "step": 7000 }, { "epoch": 5.8, "learning_rate": 5e-06, "loss": 2.735, "step": 7100 }, { "epoch": 5.88, "learning_rate": 5e-06, "loss": 2.6801, "step": 7200 }, { "epoch": 5.96, "learning_rate": 5e-06, "loss": 2.6716, "step": 7300 }, { "epoch": 6.05, "learning_rate": 5e-06, "loss": 2.5997, "step": 7400 }, { "epoch": 6.13, "learning_rate": 5e-06, "loss": 2.5214, "step": 7500 }, { "epoch": 6.21, "learning_rate": 5e-06, "loss": 2.5189, "step": 7600 }, { "epoch": 6.29, "learning_rate": 5e-06, "loss": 2.4448, "step": 7700 }, { "epoch": 6.37, "learning_rate": 5e-06, "loss": 2.3884, "step": 7800 }, { "epoch": 6.45, "learning_rate": 5e-06, "loss": 2.3592, "step": 7900 }, { "epoch": 6.54, "learning_rate": 5e-06, "loss": 2.2958, "step": 8000 }, { "epoch": 6.62, "learning_rate": 5e-06, "loss": 2.2593, "step": 8100 }, { "epoch": 6.7, "learning_rate": 5e-06, "loss": 2.2337, "step": 8200 }, { "epoch": 6.78, "learning_rate": 5e-06, "loss": 2.2003, "step": 8300 }, { "epoch": 6.86, "learning_rate": 5e-06, "loss": 2.1803, "step": 8400 }, { "epoch": 6.94, "learning_rate": 5e-06, "loss": 2.1439, "step": 8500 }, { "epoch": 7.03, "learning_rate": 5e-06, "loss": 2.1152, "step": 8600 }, { "epoch": 7.11, "learning_rate": 5e-06, "loss": 2.0329, "step": 8700 }, { "epoch": 7.19, "learning_rate": 5e-06, "loss": 2.0435, "step": 8800 }, { "epoch": 7.27, "learning_rate": 5e-06, "loss": 2.0207, "step": 8900 }, { "epoch": 7.35, "learning_rate": 5e-06, "loss": 1.9882, "step": 9000 }, { "epoch": 7.43, "learning_rate": 5e-06, "loss": 2.0032, "step": 9100 }, { "epoch": 7.52, "learning_rate": 5e-06, "loss": 1.9383, "step": 9200 }, { "epoch": 7.6, "learning_rate": 5e-06, "loss": 1.9368, "step": 9300 }, { "epoch": 7.68, "learning_rate": 5e-06, "loss": 1.8718, "step": 9400 }, { "epoch": 7.76, "learning_rate": 5e-06, "loss": 1.838, "step": 9500 }, { "epoch": 7.84, "learning_rate": 5e-06, "loss": 1.8233, "step": 9600 }, { "epoch": 7.92, "learning_rate": 5e-06, "loss": 1.8229, "step": 9700 }, { "epoch": 8.01, "learning_rate": 5e-06, "loss": 1.8271, "step": 9800 }, { "epoch": 8.09, "learning_rate": 5e-06, "loss": 1.8069, "step": 9900 }, { "epoch": 8.17, "learning_rate": 5e-06, "loss": 1.7105, "step": 10000 }, { "epoch": 8.17, "eval_cer": 0.2669406109474337, "eval_loss": 1.2349891662597656, "eval_runtime": 8.338, "eval_samples_per_second": 35.98, "eval_steps_per_second": 17.99, "eval_wer": 0.8188034188034188, "step": 10000 }, { "epoch": 8.25, "learning_rate": 5e-06, "loss": 1.7295, "step": 10100 }, { "epoch": 8.33, "learning_rate": 5e-06, "loss": 1.6777, "step": 10200 }, { "epoch": 8.41, "learning_rate": 5e-06, "loss": 1.6999, "step": 10300 }, { "epoch": 8.5, "learning_rate": 5e-06, "loss": 1.7222, "step": 10400 }, { "epoch": 8.58, "learning_rate": 5e-06, "loss": 1.6903, "step": 10500 }, { "epoch": 8.66, "learning_rate": 5e-06, "loss": 1.6686, "step": 10600 }, { "epoch": 8.74, "learning_rate": 5e-06, "loss": 1.675, "step": 10700 }, { "epoch": 8.82, "learning_rate": 5e-06, "loss": 1.63, "step": 10800 }, { "epoch": 8.9, "learning_rate": 5e-06, "loss": 1.6014, "step": 10900 }, { "epoch": 8.99, "learning_rate": 5e-06, "loss": 1.5622, "step": 11000 }, { "epoch": 9.07, "learning_rate": 5e-06, "loss": 1.6179, "step": 11100 }, { "epoch": 9.15, "learning_rate": 5e-06, "loss": 1.5449, "step": 11200 }, { "epoch": 9.23, "learning_rate": 5e-06, "loss": 1.5691, "step": 11300 }, { "epoch": 9.31, "learning_rate": 5e-06, "loss": 1.5252, "step": 11400 }, { "epoch": 9.4, "learning_rate": 5e-06, "loss": 1.5056, "step": 11500 }, { "epoch": 9.48, "learning_rate": 5e-06, "loss": 1.5231, "step": 11600 }, { "epoch": 9.56, "learning_rate": 5e-06, "loss": 1.4915, "step": 11700 }, { "epoch": 9.64, "learning_rate": 5e-06, "loss": 1.5377, "step": 11800 }, { "epoch": 9.72, "learning_rate": 5e-06, "loss": 1.4515, "step": 11900 }, { "epoch": 9.8, "learning_rate": 5e-06, "loss": 1.4831, "step": 12000 }, { "epoch": 9.89, "learning_rate": 5e-06, "loss": 1.4679, "step": 12100 }, { "epoch": 9.97, "learning_rate": 5e-06, "loss": 1.4193, "step": 12200 }, { "epoch": 10.05, "learning_rate": 5e-06, "loss": 1.4637, "step": 12300 }, { "epoch": 10.13, "learning_rate": 5e-06, "loss": 1.4281, "step": 12400 }, { "epoch": 10.21, "learning_rate": 5e-06, "loss": 1.4774, "step": 12500 }, { "epoch": 10.29, "learning_rate": 5e-06, "loss": 1.3995, "step": 12600 }, { "epoch": 10.38, "learning_rate": 5e-06, "loss": 1.3925, "step": 12700 }, { "epoch": 10.46, "learning_rate": 5e-06, "loss": 1.3734, "step": 12800 }, { "epoch": 10.54, "learning_rate": 5e-06, "loss": 1.4023, "step": 12900 }, { "epoch": 10.62, "learning_rate": 5e-06, "loss": 1.3599, "step": 13000 }, { "epoch": 10.7, "learning_rate": 5e-06, "loss": 1.3517, "step": 13100 }, { "epoch": 10.78, "learning_rate": 5e-06, "loss": 1.3845, "step": 13200 }, { "epoch": 10.87, "learning_rate": 5e-06, "loss": 1.3561, "step": 13300 }, { "epoch": 10.95, "learning_rate": 5e-06, "loss": 1.3288, "step": 13400 }, { "epoch": 11.03, "learning_rate": 5e-06, "loss": 1.329, "step": 13500 }, { "epoch": 11.11, "learning_rate": 5e-06, "loss": 1.2921, "step": 13600 }, { "epoch": 11.19, "learning_rate": 5e-06, "loss": 1.2805, "step": 13700 }, { "epoch": 11.27, "learning_rate": 5e-06, "loss": 1.3231, "step": 13800 }, { "epoch": 11.36, "learning_rate": 5e-06, "loss": 1.3033, "step": 13900 }, { "epoch": 11.44, "learning_rate": 5e-06, "loss": 1.3202, "step": 14000 }, { "epoch": 11.52, "learning_rate": 5e-06, "loss": 1.289, "step": 14100 }, { "epoch": 11.6, "learning_rate": 5e-06, "loss": 1.2955, "step": 14200 }, { "epoch": 11.68, "learning_rate": 5e-06, "loss": 1.2818, "step": 14300 }, { "epoch": 11.76, "learning_rate": 5e-06, "loss": 1.2486, "step": 14400 }, { "epoch": 11.85, "learning_rate": 5e-06, "loss": 1.2817, "step": 14500 }, { "epoch": 11.93, "learning_rate": 5e-06, "loss": 1.2654, "step": 14600 }, { "epoch": 12.01, "learning_rate": 5e-06, "loss": 1.312, "step": 14700 }, { "epoch": 12.09, "learning_rate": 5e-06, "loss": 1.2222, "step": 14800 }, { "epoch": 12.17, "learning_rate": 5e-06, "loss": 1.2101, "step": 14900 }, { "epoch": 12.25, "learning_rate": 5e-06, "loss": 1.2481, "step": 15000 }, { "epoch": 12.34, "learning_rate": 5e-06, "loss": 1.2697, "step": 15100 }, { "epoch": 12.42, "learning_rate": 5e-06, "loss": 1.2074, "step": 15200 }, { "epoch": 12.5, "learning_rate": 5e-06, "loss": 1.2031, "step": 15300 }, { "epoch": 12.58, "learning_rate": 5e-06, "loss": 1.2189, "step": 15400 }, { "epoch": 12.66, "learning_rate": 5e-06, "loss": 1.2007, "step": 15500 }, { "epoch": 12.74, "learning_rate": 5e-06, "loss": 1.2208, "step": 15600 }, { "epoch": 12.83, "learning_rate": 5e-06, "loss": 1.2302, "step": 15700 }, { "epoch": 12.91, "learning_rate": 5e-06, "loss": 1.1985, "step": 15800 }, { "epoch": 12.99, "learning_rate": 5e-06, "loss": 1.197, "step": 15900 }, { "epoch": 13.07, "learning_rate": 5e-06, "loss": 1.2294, "step": 16000 }, { "epoch": 13.15, "learning_rate": 5e-06, "loss": 1.1481, "step": 16100 }, { "epoch": 13.24, "learning_rate": 5e-06, "loss": 1.1518, "step": 16200 }, { "epoch": 13.32, "learning_rate": 5e-06, "loss": 1.2009, "step": 16300 }, { "epoch": 13.4, "learning_rate": 5e-06, "loss": 1.1421, "step": 16400 }, { "epoch": 13.48, "learning_rate": 5e-06, "loss": 1.1966, "step": 16500 }, { "epoch": 13.56, "learning_rate": 5e-06, "loss": 1.1785, "step": 16600 }, { "epoch": 13.64, "learning_rate": 5e-06, "loss": 1.1549, "step": 16700 }, { "epoch": 13.73, "learning_rate": 5e-06, "loss": 1.1844, "step": 16800 }, { "epoch": 13.81, "learning_rate": 5e-06, "loss": 1.1471, "step": 16900 }, { "epoch": 13.89, "learning_rate": 5e-06, "loss": 1.1748, "step": 17000 }, { "epoch": 13.97, "learning_rate": 5e-06, "loss": 1.1463, "step": 17100 }, { "epoch": 14.05, "learning_rate": 5e-06, "loss": 1.154, "step": 17200 }, { "epoch": 14.13, "learning_rate": 5e-06, "loss": 1.1551, "step": 17300 }, { "epoch": 14.22, "learning_rate": 5e-06, "loss": 1.1271, "step": 17400 }, { "epoch": 14.3, "learning_rate": 5e-06, "loss": 1.1019, "step": 17500 }, { "epoch": 14.38, "learning_rate": 5e-06, "loss": 1.1212, "step": 17600 }, { "epoch": 14.46, "learning_rate": 5e-06, "loss": 1.0808, "step": 17700 }, { "epoch": 14.54, "learning_rate": 5e-06, "loss": 1.171, "step": 17800 }, { "epoch": 14.62, "learning_rate": 5e-06, "loss": 1.157, "step": 17900 }, { "epoch": 14.71, "learning_rate": 5e-06, "loss": 1.0964, "step": 18000 }, { "epoch": 14.79, "learning_rate": 5e-06, "loss": 1.1006, "step": 18100 }, { "epoch": 14.87, "learning_rate": 5e-06, "loss": 1.1505, "step": 18200 }, { "epoch": 14.95, "learning_rate": 5e-06, "loss": 1.1402, "step": 18300 }, { "epoch": 15.03, "learning_rate": 5e-06, "loss": 1.0922, "step": 18400 }, { "epoch": 15.11, "learning_rate": 5e-06, "loss": 1.1458, "step": 18500 }, { "epoch": 15.2, "learning_rate": 5e-06, "loss": 1.0835, "step": 18600 }, { "epoch": 15.28, "learning_rate": 5e-06, "loss": 1.0964, "step": 18700 }, { "epoch": 15.36, "learning_rate": 5e-06, "loss": 1.0833, "step": 18800 }, { "epoch": 15.44, "learning_rate": 5e-06, "loss": 1.0897, "step": 18900 }, { "epoch": 15.52, "learning_rate": 5e-06, "loss": 1.0544, "step": 19000 }, { "epoch": 15.6, "learning_rate": 5e-06, "loss": 1.0939, "step": 19100 }, { "epoch": 15.69, "learning_rate": 5e-06, "loss": 1.0333, "step": 19200 }, { "epoch": 15.77, "learning_rate": 5e-06, "loss": 1.1071, "step": 19300 }, { "epoch": 15.85, "learning_rate": 5e-06, "loss": 1.0865, "step": 19400 }, { "epoch": 15.93, "learning_rate": 5e-06, "loss": 1.0961, "step": 19500 }, { "epoch": 16.01, "learning_rate": 5e-06, "loss": 1.0696, "step": 19600 }, { "epoch": 16.09, "learning_rate": 5e-06, "loss": 1.0574, "step": 19700 }, { "epoch": 16.18, "learning_rate": 5e-06, "loss": 1.0373, "step": 19800 }, { "epoch": 16.26, "learning_rate": 5e-06, "loss": 1.0526, "step": 19900 }, { "epoch": 16.34, "learning_rate": 5e-06, "loss": 1.0707, "step": 20000 }, { "epoch": 16.34, "eval_cer": 0.14808497441463792, "eval_loss": 0.6507723331451416, "eval_runtime": 8.0755, "eval_samples_per_second": 37.149, "eval_steps_per_second": 18.575, "eval_wer": 0.5153846153846153, "step": 20000 }, { "epoch": 16.42, "learning_rate": 5e-06, "loss": 1.0709, "step": 20100 }, { "epoch": 16.5, "learning_rate": 5e-06, "loss": 1.0897, "step": 20200 }, { "epoch": 16.58, "learning_rate": 5e-06, "loss": 1.0179, "step": 20300 }, { "epoch": 16.67, "learning_rate": 5e-06, "loss": 1.0557, "step": 20400 }, { "epoch": 16.75, "learning_rate": 5e-06, "loss": 1.0734, "step": 20500 }, { "epoch": 16.83, "learning_rate": 5e-06, "loss": 1.0456, "step": 20600 }, { "epoch": 16.91, "learning_rate": 5e-06, "loss": 1.0186, "step": 20700 }, { "epoch": 16.99, "learning_rate": 5e-06, "loss": 1.0345, "step": 20800 }, { "epoch": 17.08, "learning_rate": 5e-06, "loss": 1.093, "step": 20900 }, { "epoch": 17.16, "learning_rate": 5e-06, "loss": 1.0239, "step": 21000 }, { "epoch": 17.24, "learning_rate": 5e-06, "loss": 1.0453, "step": 21100 }, { "epoch": 17.32, "learning_rate": 5e-06, "loss": 1.01, "step": 21200 }, { "epoch": 17.4, "learning_rate": 5e-06, "loss": 1.0602, "step": 21300 }, { "epoch": 17.48, "learning_rate": 5e-06, "loss": 1.0047, "step": 21400 }, { "epoch": 17.57, "learning_rate": 5e-06, "loss": 1.0379, "step": 21500 }, { "epoch": 17.65, "learning_rate": 5e-06, "loss": 1.0277, "step": 21600 }, { "epoch": 17.73, "learning_rate": 5e-06, "loss": 1.0434, "step": 21700 }, { "epoch": 17.81, "learning_rate": 5e-06, "loss": 0.9741, "step": 21800 }, { "epoch": 17.89, "learning_rate": 5e-06, "loss": 0.9905, "step": 21900 }, { "epoch": 17.97, "learning_rate": 5e-06, "loss": 1.0115, "step": 22000 }, { "epoch": 18.06, "learning_rate": 5e-06, "loss": 0.9891, "step": 22100 }, { "epoch": 18.14, "learning_rate": 5e-06, "loss": 1.0262, "step": 22200 }, { "epoch": 18.22, "learning_rate": 5e-06, "loss": 1.0048, "step": 22300 }, { "epoch": 18.3, "learning_rate": 5e-06, "loss": 1.0176, "step": 22400 }, { "epoch": 18.38, "learning_rate": 5e-06, "loss": 1.01, "step": 22500 }, { "epoch": 18.46, "learning_rate": 5e-06, "loss": 0.9899, "step": 22600 }, { "epoch": 18.55, "learning_rate": 5e-06, "loss": 0.9553, "step": 22700 }, { "epoch": 18.63, "learning_rate": 5e-06, "loss": 0.9849, "step": 22800 }, { "epoch": 18.71, "learning_rate": 5e-06, "loss": 0.9745, "step": 22900 }, { "epoch": 18.79, "learning_rate": 5e-06, "loss": 0.9761, "step": 23000 }, { "epoch": 18.87, "learning_rate": 5e-06, "loss": 1.0133, "step": 23100 }, { "epoch": 18.95, "learning_rate": 5e-06, "loss": 0.9874, "step": 23200 }, { "epoch": 19.04, "learning_rate": 5e-06, "loss": 0.9607, "step": 23300 }, { "epoch": 19.12, "learning_rate": 5e-06, "loss": 0.9738, "step": 23400 }, { "epoch": 19.2, "learning_rate": 5e-06, "loss": 0.9543, "step": 23500 }, { "epoch": 19.28, "learning_rate": 5e-06, "loss": 0.9533, "step": 23600 }, { "epoch": 19.36, "learning_rate": 5e-06, "loss": 1.0028, "step": 23700 }, { "epoch": 19.44, "learning_rate": 5e-06, "loss": 0.937, "step": 23800 }, { "epoch": 19.53, "learning_rate": 5e-06, "loss": 0.9979, "step": 23900 }, { "epoch": 19.61, "learning_rate": 5e-06, "loss": 0.9872, "step": 24000 }, { "epoch": 19.69, "learning_rate": 5e-06, "loss": 0.9519, "step": 24100 }, { "epoch": 19.77, "learning_rate": 5e-06, "loss": 0.9624, "step": 24200 }, { "epoch": 19.85, "learning_rate": 5e-06, "loss": 1.0209, "step": 24300 }, { "epoch": 19.93, "learning_rate": 5e-06, "loss": 0.9874, "step": 24400 }, { "epoch": 20.02, "learning_rate": 5e-06, "loss": 0.9655, "step": 24500 }, { "epoch": 20.1, "learning_rate": 5e-06, "loss": 0.9492, "step": 24600 }, { "epoch": 20.18, "learning_rate": 5e-06, "loss": 0.9851, "step": 24700 }, { "epoch": 20.26, "learning_rate": 5e-06, "loss": 0.9355, "step": 24800 }, { "epoch": 20.34, "learning_rate": 5e-06, "loss": 0.972, "step": 24900 }, { "epoch": 20.42, "learning_rate": 5e-06, "loss": 0.9426, "step": 25000 }, { "epoch": 20.51, "learning_rate": 5e-06, "loss": 0.9908, "step": 25100 }, { "epoch": 20.59, "learning_rate": 5e-06, "loss": 0.9182, "step": 25200 }, { "epoch": 20.67, "learning_rate": 5e-06, "loss": 0.9191, "step": 25300 }, { "epoch": 20.75, "learning_rate": 5e-06, "loss": 0.9784, "step": 25400 }, { "epoch": 20.83, "learning_rate": 5e-06, "loss": 0.944, "step": 25500 }, { "epoch": 20.91, "learning_rate": 5e-06, "loss": 0.9475, "step": 25600 }, { "epoch": 21.0, "learning_rate": 5e-06, "loss": 0.9254, "step": 25700 }, { "epoch": 21.08, "learning_rate": 5e-06, "loss": 0.9553, "step": 25800 }, { "epoch": 21.16, "learning_rate": 5e-06, "loss": 0.9452, "step": 25900 }, { "epoch": 21.24, "learning_rate": 5e-06, "loss": 0.9666, "step": 26000 }, { "epoch": 21.32, "learning_rate": 5e-06, "loss": 0.9389, "step": 26100 }, { "epoch": 21.4, "learning_rate": 5e-06, "loss": 0.971, "step": 26200 }, { "epoch": 21.49, "learning_rate": 5e-06, "loss": 0.9212, "step": 26300 }, { "epoch": 21.57, "learning_rate": 5e-06, "loss": 0.9039, "step": 26400 }, { "epoch": 21.65, "learning_rate": 5e-06, "loss": 0.8837, "step": 26500 }, { "epoch": 21.73, "learning_rate": 5e-06, "loss": 0.9587, "step": 26600 }, { "epoch": 21.81, "learning_rate": 5e-06, "loss": 0.9101, "step": 26700 }, { "epoch": 21.89, "learning_rate": 5e-06, "loss": 0.928, "step": 26800 }, { "epoch": 21.98, "learning_rate": 5e-06, "loss": 0.889, "step": 26900 }, { "epoch": 22.06, "learning_rate": 5e-06, "loss": 0.9088, "step": 27000 }, { "epoch": 22.14, "learning_rate": 5e-06, "loss": 0.9235, "step": 27100 }, { "epoch": 22.22, "learning_rate": 5e-06, "loss": 0.9353, "step": 27200 }, { "epoch": 22.3, "learning_rate": 5e-06, "loss": 0.9046, "step": 27300 }, { "epoch": 22.39, "learning_rate": 5e-06, "loss": 0.9106, "step": 27400 }, { "epoch": 22.47, "learning_rate": 5e-06, "loss": 0.9404, "step": 27500 }, { "epoch": 22.55, "learning_rate": 5e-06, "loss": 0.9081, "step": 27600 }, { "epoch": 22.63, "learning_rate": 5e-06, "loss": 0.9042, "step": 27700 }, { "epoch": 22.71, "learning_rate": 5e-06, "loss": 0.8971, "step": 27800 }, { "epoch": 22.79, "learning_rate": 5e-06, "loss": 0.9116, "step": 27900 }, { "epoch": 22.88, "learning_rate": 5e-06, "loss": 0.9135, "step": 28000 }, { "epoch": 22.96, "learning_rate": 5e-06, "loss": 0.9011, "step": 28100 }, { "epoch": 23.04, "learning_rate": 5e-06, "loss": 0.9025, "step": 28200 }, { "epoch": 23.12, "learning_rate": 5e-06, "loss": 0.9275, "step": 28300 }, { "epoch": 23.2, "learning_rate": 5e-06, "loss": 0.9086, "step": 28400 }, { "epoch": 23.28, "learning_rate": 5e-06, "loss": 0.9075, "step": 28500 }, { "epoch": 23.37, "learning_rate": 5e-06, "loss": 0.9022, "step": 28600 }, { "epoch": 23.45, "learning_rate": 5e-06, "loss": 0.9067, "step": 28700 }, { "epoch": 23.53, "learning_rate": 5e-06, "loss": 0.895, "step": 28800 }, { "epoch": 23.61, "learning_rate": 5e-06, "loss": 0.891, "step": 28900 }, { "epoch": 23.69, "learning_rate": 5e-06, "loss": 0.8775, "step": 29000 }, { "epoch": 23.77, "learning_rate": 5e-06, "loss": 0.8691, "step": 29100 }, { "epoch": 23.86, "learning_rate": 5e-06, "loss": 0.9149, "step": 29200 }, { "epoch": 23.94, "learning_rate": 5e-06, "loss": 0.9142, "step": 29300 }, { "epoch": 24.02, "learning_rate": 5e-06, "loss": 0.9338, "step": 29400 }, { "epoch": 24.1, "learning_rate": 5e-06, "loss": 0.8882, "step": 29500 }, { "epoch": 24.18, "learning_rate": 5e-06, "loss": 0.8561, "step": 29600 }, { "epoch": 24.26, "learning_rate": 5e-06, "loss": 0.8855, "step": 29700 }, { "epoch": 24.35, "learning_rate": 5e-06, "loss": 0.8879, "step": 29800 }, { "epoch": 24.43, "learning_rate": 5e-06, "loss": 0.9106, "step": 29900 }, { "epoch": 24.51, "learning_rate": 5e-06, "loss": 0.902, "step": 30000 }, { "epoch": 24.51, "eval_cer": 0.12203442394169639, "eval_loss": 0.547027051448822, "eval_runtime": 8.3805, "eval_samples_per_second": 35.798, "eval_steps_per_second": 17.899, "eval_wer": 0.4239316239316239, "step": 30000 }, { "epoch": 24.59, "learning_rate": 4.985666666666667e-06, "loss": 0.9302, "step": 30100 }, { "epoch": 24.67, "learning_rate": 4.969e-06, "loss": 0.8792, "step": 30200 }, { "epoch": 24.75, "learning_rate": 4.952333333333334e-06, "loss": 0.8448, "step": 30300 }, { "epoch": 24.84, "learning_rate": 4.935666666666667e-06, "loss": 0.8973, "step": 30400 }, { "epoch": 24.92, "learning_rate": 4.919000000000001e-06, "loss": 0.8425, "step": 30500 }, { "epoch": 25.0, "learning_rate": 4.902333333333334e-06, "loss": 0.8707, "step": 30600 }, { "epoch": 25.08, "learning_rate": 4.885666666666667e-06, "loss": 0.8867, "step": 30700 }, { "epoch": 25.16, "learning_rate": 4.869e-06, "loss": 0.8502, "step": 30800 }, { "epoch": 25.24, "learning_rate": 4.8523333333333335e-06, "loss": 0.8851, "step": 30900 }, { "epoch": 25.33, "learning_rate": 4.835666666666667e-06, "loss": 0.858, "step": 31000 }, { "epoch": 25.41, "learning_rate": 4.819e-06, "loss": 0.8725, "step": 31100 }, { "epoch": 25.49, "learning_rate": 4.802333333333334e-06, "loss": 0.8324, "step": 31200 }, { "epoch": 25.57, "learning_rate": 4.785666666666667e-06, "loss": 0.866, "step": 31300 }, { "epoch": 25.65, "learning_rate": 4.769e-06, "loss": 0.8658, "step": 31400 }, { "epoch": 25.73, "learning_rate": 4.752333333333334e-06, "loss": 0.8725, "step": 31500 }, { "epoch": 25.82, "learning_rate": 4.735666666666668e-06, "loss": 0.8853, "step": 31600 }, { "epoch": 25.9, "learning_rate": 4.719e-06, "loss": 0.8961, "step": 31700 }, { "epoch": 25.98, "learning_rate": 4.702333333333334e-06, "loss": 0.8206, "step": 31800 }, { "epoch": 26.06, "learning_rate": 4.685666666666667e-06, "loss": 0.8634, "step": 31900 }, { "epoch": 26.14, "learning_rate": 4.6690000000000005e-06, "loss": 0.9202, "step": 32000 }, { "epoch": 26.23, "learning_rate": 4.652333333333334e-06, "loss": 0.8333, "step": 32100 }, { "epoch": 26.31, "learning_rate": 4.635666666666667e-06, "loss": 0.8525, "step": 32200 }, { "epoch": 26.39, "learning_rate": 4.619e-06, "loss": 0.8417, "step": 32300 }, { "epoch": 26.47, "learning_rate": 4.602333333333333e-06, "loss": 0.8567, "step": 32400 }, { "epoch": 26.55, "learning_rate": 4.5856666666666674e-06, "loss": 0.8276, "step": 32500 }, { "epoch": 26.63, "learning_rate": 4.569e-06, "loss": 0.8267, "step": 32600 }, { "epoch": 26.72, "learning_rate": 4.552333333333334e-06, "loss": 0.8312, "step": 32700 }, { "epoch": 26.8, "learning_rate": 4.535666666666667e-06, "loss": 0.8272, "step": 32800 }, { "epoch": 26.88, "learning_rate": 4.519e-06, "loss": 0.9122, "step": 32900 }, { "epoch": 26.96, "learning_rate": 4.5023333333333335e-06, "loss": 0.862, "step": 33000 }, { "epoch": 27.04, "learning_rate": 4.485666666666667e-06, "loss": 0.882, "step": 33100 }, { "epoch": 27.12, "learning_rate": 4.469000000000001e-06, "loss": 0.8624, "step": 33200 }, { "epoch": 27.21, "learning_rate": 4.452333333333333e-06, "loss": 0.8011, "step": 33300 }, { "epoch": 27.29, "learning_rate": 4.435666666666667e-06, "loss": 0.8589, "step": 33400 }, { "epoch": 27.37, "learning_rate": 4.419e-06, "loss": 0.8521, "step": 33500 }, { "epoch": 27.45, "learning_rate": 4.402333333333334e-06, "loss": 0.8436, "step": 33600 }, { "epoch": 27.53, "learning_rate": 4.385666666666667e-06, "loss": 0.8348, "step": 33700 }, { "epoch": 27.61, "learning_rate": 4.369000000000001e-06, "loss": 0.847, "step": 33800 }, { "epoch": 27.7, "learning_rate": 4.352333333333334e-06, "loss": 0.8742, "step": 33900 }, { "epoch": 27.78, "learning_rate": 4.335666666666667e-06, "loss": 0.8275, "step": 34000 }, { "epoch": 27.86, "learning_rate": 4.3190000000000005e-06, "loss": 0.8436, "step": 34100 }, { "epoch": 27.94, "learning_rate": 4.302333333333334e-06, "loss": 0.8261, "step": 34200 }, { "epoch": 28.02, "learning_rate": 4.285666666666667e-06, "loss": 0.7982, "step": 34300 }, { "epoch": 28.1, "learning_rate": 4.269e-06, "loss": 0.8012, "step": 34400 }, { "epoch": 28.19, "learning_rate": 4.2525e-06, "loss": 0.8682, "step": 34500 }, { "epoch": 28.27, "learning_rate": 4.235833333333334e-06, "loss": 0.8125, "step": 34600 }, { "epoch": 28.35, "learning_rate": 4.219166666666667e-06, "loss": 0.8077, "step": 34700 }, { "epoch": 28.43, "learning_rate": 4.202500000000001e-06, "loss": 0.8447, "step": 34800 }, { "epoch": 28.51, "learning_rate": 4.185833333333333e-06, "loss": 0.8205, "step": 34900 }, { "epoch": 28.59, "learning_rate": 4.169166666666667e-06, "loss": 0.8297, "step": 35000 }, { "epoch": 28.68, "learning_rate": 4.1525000000000005e-06, "loss": 0.8311, "step": 35100 }, { "epoch": 28.76, "learning_rate": 4.135833333333334e-06, "loss": 0.818, "step": 35200 }, { "epoch": 28.84, "learning_rate": 4.119166666666667e-06, "loss": 0.8653, "step": 35300 }, { "epoch": 28.92, "learning_rate": 4.1025e-06, "loss": 0.8136, "step": 35400 }, { "epoch": 29.0, "learning_rate": 4.085833333333334e-06, "loss": 0.8276, "step": 35500 }, { "epoch": 29.08, "learning_rate": 4.0691666666666665e-06, "loss": 0.7686, "step": 35600 }, { "epoch": 29.17, "learning_rate": 4.052500000000001e-06, "loss": 0.856, "step": 35700 }, { "epoch": 29.25, "learning_rate": 4.035833333333334e-06, "loss": 0.803, "step": 35800 }, { "epoch": 29.33, "learning_rate": 4.019166666666667e-06, "loss": 0.827, "step": 35900 }, { "epoch": 29.41, "learning_rate": 4.0025e-06, "loss": 0.8293, "step": 36000 }, { "epoch": 29.49, "learning_rate": 3.9858333333333334e-06, "loss": 0.8098, "step": 36100 }, { "epoch": 29.57, "learning_rate": 3.969166666666667e-06, "loss": 0.8037, "step": 36200 }, { "epoch": 29.66, "learning_rate": 3.9525e-06, "loss": 0.8306, "step": 36300 }, { "epoch": 29.74, "learning_rate": 3.935833333333334e-06, "loss": 0.7893, "step": 36400 }, { "epoch": 29.82, "learning_rate": 3.919166666666667e-06, "loss": 0.8476, "step": 36500 }, { "epoch": 29.9, "learning_rate": 3.9025e-06, "loss": 0.8068, "step": 36600 }, { "epoch": 29.98, "learning_rate": 3.8858333333333335e-06, "loss": 0.8298, "step": 36700 }, { "epoch": 30.07, "learning_rate": 3.869333333333334e-06, "loss": 0.8144, "step": 36800 }, { "epoch": 30.15, "learning_rate": 3.852666666666667e-06, "loss": 0.7807, "step": 36900 }, { "epoch": 30.23, "learning_rate": 3.836e-06, "loss": 0.7881, "step": 37000 }, { "epoch": 30.31, "learning_rate": 3.819333333333333e-06, "loss": 0.8269, "step": 37100 }, { "epoch": 30.39, "learning_rate": 3.8026666666666666e-06, "loss": 0.8161, "step": 37200 }, { "epoch": 30.47, "learning_rate": 3.7860000000000003e-06, "loss": 0.8432, "step": 37300 }, { "epoch": 30.56, "learning_rate": 3.769333333333334e-06, "loss": 0.7769, "step": 37400 }, { "epoch": 30.64, "learning_rate": 3.752666666666667e-06, "loss": 0.79, "step": 37500 }, { "epoch": 30.72, "learning_rate": 3.7360000000000003e-06, "loss": 0.8203, "step": 37600 }, { "epoch": 30.8, "learning_rate": 3.7193333333333335e-06, "loss": 0.7892, "step": 37700 }, { "epoch": 30.88, "learning_rate": 3.702666666666667e-06, "loss": 0.8422, "step": 37800 }, { "epoch": 30.96, "learning_rate": 3.686e-06, "loss": 0.7969, "step": 37900 }, { "epoch": 31.05, "learning_rate": 3.6693333333333336e-06, "loss": 0.8289, "step": 38000 }, { "epoch": 31.13, "learning_rate": 3.6526666666666672e-06, "loss": 0.8019, "step": 38100 }, { "epoch": 31.21, "learning_rate": 3.636e-06, "loss": 0.8314, "step": 38200 }, { "epoch": 31.29, "learning_rate": 3.6193333333333336e-06, "loss": 0.8357, "step": 38300 }, { "epoch": 31.37, "learning_rate": 3.6026666666666673e-06, "loss": 0.7961, "step": 38400 }, { "epoch": 31.45, "learning_rate": 3.586e-06, "loss": 0.7876, "step": 38500 }, { "epoch": 31.54, "learning_rate": 3.5695000000000003e-06, "loss": 0.775, "step": 38600 }, { "epoch": 31.62, "learning_rate": 3.5528333333333335e-06, "loss": 0.8121, "step": 38700 }, { "epoch": 31.7, "learning_rate": 3.536166666666667e-06, "loss": 0.7963, "step": 38800 }, { "epoch": 31.78, "learning_rate": 3.5195e-06, "loss": 0.7908, "step": 38900 }, { "epoch": 31.86, "learning_rate": 3.5028333333333336e-06, "loss": 0.8165, "step": 39000 }, { "epoch": 31.94, "learning_rate": 3.486166666666667e-06, "loss": 0.765, "step": 39100 }, { "epoch": 32.03, "learning_rate": 3.4695e-06, "loss": 0.7875, "step": 39200 }, { "epoch": 32.11, "learning_rate": 3.4528333333333336e-06, "loss": 0.8024, "step": 39300 }, { "epoch": 32.19, "learning_rate": 3.4361666666666673e-06, "loss": 0.7563, "step": 39400 }, { "epoch": 32.27, "learning_rate": 3.4195e-06, "loss": 0.8206, "step": 39500 }, { "epoch": 32.35, "learning_rate": 3.4028333333333337e-06, "loss": 0.7858, "step": 39600 }, { "epoch": 32.43, "learning_rate": 3.386166666666667e-06, "loss": 0.7956, "step": 39700 }, { "epoch": 32.52, "learning_rate": 3.3695000000000005e-06, "loss": 0.7966, "step": 39800 }, { "epoch": 32.6, "learning_rate": 3.3528333333333333e-06, "loss": 0.8083, "step": 39900 }, { "epoch": 32.68, "learning_rate": 3.336166666666667e-06, "loss": 0.7696, "step": 40000 }, { "epoch": 32.68, "eval_cer": 0.11156768491238951, "eval_loss": 0.5117331147193909, "eval_runtime": 8.1077, "eval_samples_per_second": 37.002, "eval_steps_per_second": 18.501, "eval_wer": 0.3871794871794872, "step": 40000 }, { "epoch": 32.76, "learning_rate": 3.3195000000000006e-06, "loss": 0.7772, "step": 40100 }, { "epoch": 32.84, "learning_rate": 3.3028333333333334e-06, "loss": 0.7818, "step": 40200 }, { "epoch": 32.92, "learning_rate": 3.286166666666667e-06, "loss": 0.8184, "step": 40300 }, { "epoch": 33.01, "learning_rate": 3.2695000000000006e-06, "loss": 0.8117, "step": 40400 }, { "epoch": 33.09, "learning_rate": 3.2528333333333334e-06, "loss": 0.7759, "step": 40500 }, { "epoch": 33.17, "learning_rate": 3.236166666666667e-06, "loss": 0.8274, "step": 40600 }, { "epoch": 33.25, "learning_rate": 3.2195000000000003e-06, "loss": 0.7923, "step": 40700 }, { "epoch": 33.33, "learning_rate": 3.2028333333333335e-06, "loss": 0.7858, "step": 40800 }, { "epoch": 33.41, "learning_rate": 3.1861666666666667e-06, "loss": 0.7894, "step": 40900 }, { "epoch": 33.5, "learning_rate": 3.169666666666667e-06, "loss": 0.7956, "step": 41000 }, { "epoch": 33.58, "learning_rate": 3.1530000000000006e-06, "loss": 0.7691, "step": 41100 }, { "epoch": 33.66, "learning_rate": 3.1363333333333333e-06, "loss": 0.7619, "step": 41200 }, { "epoch": 33.74, "learning_rate": 3.119666666666667e-06, "loss": 0.7676, "step": 41300 }, { "epoch": 33.82, "learning_rate": 3.1030000000000006e-06, "loss": 0.8083, "step": 41400 }, { "epoch": 33.9, "learning_rate": 3.0863333333333334e-06, "loss": 0.7691, "step": 41500 }, { "epoch": 33.99, "learning_rate": 3.069666666666667e-06, "loss": 0.746, "step": 41600 }, { "epoch": 34.07, "learning_rate": 3.0530000000000002e-06, "loss": 0.7611, "step": 41700 }, { "epoch": 34.15, "learning_rate": 3.0363333333333335e-06, "loss": 0.7673, "step": 41800 }, { "epoch": 34.23, "learning_rate": 3.0196666666666667e-06, "loss": 0.7686, "step": 41900 }, { "epoch": 34.31, "learning_rate": 3.0030000000000003e-06, "loss": 0.7694, "step": 42000 }, { "epoch": 34.4, "learning_rate": 2.9863333333333335e-06, "loss": 0.815, "step": 42100 }, { "epoch": 34.48, "learning_rate": 2.9696666666666667e-06, "loss": 0.7531, "step": 42200 }, { "epoch": 34.56, "learning_rate": 2.9530000000000004e-06, "loss": 0.753, "step": 42300 }, { "epoch": 34.64, "learning_rate": 2.936333333333334e-06, "loss": 0.7748, "step": 42400 }, { "epoch": 34.72, "learning_rate": 2.9196666666666668e-06, "loss": 0.7828, "step": 42500 }, { "epoch": 34.8, "learning_rate": 2.9030000000000004e-06, "loss": 0.7849, "step": 42600 }, { "epoch": 34.89, "learning_rate": 2.8863333333333336e-06, "loss": 0.8083, "step": 42700 }, { "epoch": 34.97, "learning_rate": 2.869666666666667e-06, "loss": 0.7838, "step": 42800 }, { "epoch": 35.05, "learning_rate": 2.853e-06, "loss": 0.8093, "step": 42900 }, { "epoch": 35.13, "learning_rate": 2.8363333333333337e-06, "loss": 0.7689, "step": 43000 }, { "epoch": 35.21, "learning_rate": 2.8196666666666665e-06, "loss": 0.7665, "step": 43100 }, { "epoch": 35.29, "learning_rate": 2.803e-06, "loss": 0.7172, "step": 43200 }, { "epoch": 35.38, "learning_rate": 2.7863333333333337e-06, "loss": 0.7646, "step": 43300 }, { "epoch": 35.46, "learning_rate": 2.7696666666666674e-06, "loss": 0.7874, "step": 43400 }, { "epoch": 35.54, "learning_rate": 2.753e-06, "loss": 0.8095, "step": 43500 }, { "epoch": 35.62, "learning_rate": 2.7363333333333338e-06, "loss": 0.7498, "step": 43600 }, { "epoch": 35.7, "learning_rate": 2.719666666666667e-06, "loss": 0.7713, "step": 43700 }, { "epoch": 35.78, "learning_rate": 2.703e-06, "loss": 0.7965, "step": 43800 }, { "epoch": 35.87, "learning_rate": 2.6863333333333334e-06, "loss": 0.7848, "step": 43900 }, { "epoch": 35.95, "learning_rate": 2.669666666666667e-06, "loss": 0.7648, "step": 44000 }, { "epoch": 36.03, "learning_rate": 2.653e-06, "loss": 0.757, "step": 44100 }, { "epoch": 36.11, "learning_rate": 2.6363333333333335e-06, "loss": 0.7863, "step": 44200 }, { "epoch": 36.19, "learning_rate": 2.619666666666667e-06, "loss": 0.7656, "step": 44300 }, { "epoch": 36.27, "learning_rate": 2.603e-06, "loss": 0.7338, "step": 44400 }, { "epoch": 36.36, "learning_rate": 2.5863333333333335e-06, "loss": 0.7977, "step": 44500 }, { "epoch": 36.44, "learning_rate": 2.569666666666667e-06, "loss": 0.7555, "step": 44600 }, { "epoch": 36.52, "learning_rate": 2.5530000000000004e-06, "loss": 0.7268, "step": 44700 }, { "epoch": 36.6, "learning_rate": 2.5363333333333336e-06, "loss": 0.7737, "step": 44800 }, { "epoch": 36.68, "learning_rate": 2.519666666666667e-06, "loss": 0.7716, "step": 44900 }, { "epoch": 36.76, "learning_rate": 2.5030000000000004e-06, "loss": 0.7961, "step": 45000 }, { "epoch": 36.85, "learning_rate": 2.4863333333333336e-06, "loss": 0.73, "step": 45100 }, { "epoch": 36.93, "learning_rate": 2.4698333333333335e-06, "loss": 0.7859, "step": 45200 }, { "epoch": 37.01, "learning_rate": 2.4531666666666667e-06, "loss": 0.7968, "step": 45300 }, { "epoch": 37.09, "learning_rate": 2.4365000000000003e-06, "loss": 0.7851, "step": 45400 }, { "epoch": 37.17, "learning_rate": 2.4198333333333335e-06, "loss": 0.7777, "step": 45500 }, { "epoch": 37.25, "learning_rate": 2.403166666666667e-06, "loss": 0.7303, "step": 45600 }, { "epoch": 37.34, "learning_rate": 2.3865000000000004e-06, "loss": 0.7871, "step": 45700 }, { "epoch": 37.42, "learning_rate": 2.3698333333333336e-06, "loss": 0.7566, "step": 45800 }, { "epoch": 37.5, "learning_rate": 2.3531666666666668e-06, "loss": 0.7616, "step": 45900 }, { "epoch": 37.58, "learning_rate": 2.3365e-06, "loss": 0.7324, "step": 46000 }, { "epoch": 37.66, "learning_rate": 2.3198333333333336e-06, "loss": 0.7485, "step": 46100 }, { "epoch": 37.74, "learning_rate": 2.303166666666667e-06, "loss": 0.7578, "step": 46200 }, { "epoch": 37.83, "learning_rate": 2.2865e-06, "loss": 0.7522, "step": 46300 }, { "epoch": 37.91, "learning_rate": 2.2698333333333337e-06, "loss": 0.7478, "step": 46400 }, { "epoch": 37.99, "learning_rate": 2.253166666666667e-06, "loss": 0.7818, "step": 46500 }, { "epoch": 38.07, "learning_rate": 2.2365e-06, "loss": 0.7424, "step": 46600 }, { "epoch": 38.15, "learning_rate": 2.2198333333333337e-06, "loss": 0.7343, "step": 46700 }, { "epoch": 38.24, "learning_rate": 2.203166666666667e-06, "loss": 0.779, "step": 46800 }, { "epoch": 38.32, "learning_rate": 2.1865e-06, "loss": 0.7509, "step": 46900 }, { "epoch": 38.4, "learning_rate": 2.1698333333333334e-06, "loss": 0.7452, "step": 47000 }, { "epoch": 38.48, "learning_rate": 2.1531666666666666e-06, "loss": 0.7771, "step": 47100 }, { "epoch": 38.56, "learning_rate": 2.1365000000000002e-06, "loss": 0.7662, "step": 47200 }, { "epoch": 38.64, "learning_rate": 2.1198333333333334e-06, "loss": 0.7476, "step": 47300 }, { "epoch": 38.73, "learning_rate": 2.103166666666667e-06, "loss": 0.7692, "step": 47400 }, { "epoch": 38.81, "learning_rate": 2.0865000000000003e-06, "loss": 0.7256, "step": 47500 }, { "epoch": 38.89, "learning_rate": 2.0698333333333335e-06, "loss": 0.7455, "step": 47600 }, { "epoch": 38.97, "learning_rate": 2.053166666666667e-06, "loss": 0.7437, "step": 47700 }, { "epoch": 39.05, "learning_rate": 2.0365000000000003e-06, "loss": 0.7618, "step": 47800 }, { "epoch": 39.13, "learning_rate": 2.0198333333333335e-06, "loss": 0.731, "step": 47900 }, { "epoch": 39.22, "learning_rate": 2.0031666666666667e-06, "loss": 0.7739, "step": 48000 }, { "epoch": 39.3, "learning_rate": 1.9865e-06, "loss": 0.7263, "step": 48100 }, { "epoch": 39.38, "learning_rate": 1.9698333333333336e-06, "loss": 0.7434, "step": 48200 }, { "epoch": 39.46, "learning_rate": 1.953166666666667e-06, "loss": 0.7491, "step": 48300 }, { "epoch": 39.54, "learning_rate": 1.9365e-06, "loss": 0.7346, "step": 48400 }, { "epoch": 39.62, "learning_rate": 1.9198333333333336e-06, "loss": 0.7308, "step": 48500 }, { "epoch": 39.71, "learning_rate": 1.9031666666666669e-06, "loss": 0.7484, "step": 48600 }, { "epoch": 39.79, "learning_rate": 1.8865000000000003e-06, "loss": 0.7621, "step": 48700 }, { "epoch": 39.87, "learning_rate": 1.8698333333333335e-06, "loss": 0.7626, "step": 48800 }, { "epoch": 39.95, "learning_rate": 1.8531666666666667e-06, "loss": 0.7763, "step": 48900 }, { "epoch": 40.03, "learning_rate": 1.8365000000000003e-06, "loss": 0.7559, "step": 49000 }, { "epoch": 40.11, "learning_rate": 1.8198333333333335e-06, "loss": 0.7356, "step": 49100 }, { "epoch": 40.2, "learning_rate": 1.8031666666666668e-06, "loss": 0.7245, "step": 49200 }, { "epoch": 40.28, "learning_rate": 1.7866666666666668e-06, "loss": 0.7425, "step": 49300 }, { "epoch": 40.36, "learning_rate": 1.77e-06, "loss": 0.7435, "step": 49400 }, { "epoch": 40.44, "learning_rate": 1.7535000000000002e-06, "loss": 0.7487, "step": 49500 }, { "epoch": 40.52, "learning_rate": 1.7368333333333334e-06, "loss": 0.7614, "step": 49600 }, { "epoch": 40.6, "learning_rate": 1.7201666666666669e-06, "loss": 0.7309, "step": 49700 }, { "epoch": 40.69, "learning_rate": 1.7035e-06, "loss": 0.7304, "step": 49800 }, { "epoch": 40.77, "learning_rate": 1.6868333333333333e-06, "loss": 0.7716, "step": 49900 }, { "epoch": 40.85, "learning_rate": 1.670166666666667e-06, "loss": 0.7319, "step": 50000 }, { "epoch": 40.85, "eval_cer": 0.10707086369979842, "eval_loss": 0.4964157044887543, "eval_runtime": 7.9633, "eval_samples_per_second": 37.673, "eval_steps_per_second": 18.836, "eval_wer": 0.3645299145299145, "step": 50000 }, { "epoch": 40.93, "learning_rate": 1.6535000000000001e-06, "loss": 0.7407, "step": 50100 }, { "epoch": 41.01, "learning_rate": 1.6368333333333333e-06, "loss": 0.759, "step": 50200 }, { "epoch": 41.09, "learning_rate": 1.6201666666666668e-06, "loss": 0.7313, "step": 50300 }, { "epoch": 41.18, "learning_rate": 1.6035e-06, "loss": 0.7404, "step": 50400 }, { "epoch": 41.26, "learning_rate": 1.5868333333333336e-06, "loss": 0.7065, "step": 50500 }, { "epoch": 41.34, "learning_rate": 1.5701666666666668e-06, "loss": 0.734, "step": 50600 }, { "epoch": 41.42, "learning_rate": 1.5535e-06, "loss": 0.7326, "step": 50700 }, { "epoch": 41.5, "learning_rate": 1.5368333333333335e-06, "loss": 0.7268, "step": 50800 }, { "epoch": 41.58, "learning_rate": 1.5201666666666667e-06, "loss": 0.7687, "step": 50900 }, { "epoch": 41.67, "learning_rate": 1.5035000000000003e-06, "loss": 0.7397, "step": 51000 }, { "epoch": 41.75, "learning_rate": 1.4868333333333335e-06, "loss": 0.7426, "step": 51100 }, { "epoch": 41.83, "learning_rate": 1.4701666666666667e-06, "loss": 0.7418, "step": 51200 }, { "epoch": 41.91, "learning_rate": 1.4535000000000001e-06, "loss": 0.7789, "step": 51300 }, { "epoch": 41.99, "learning_rate": 1.4368333333333334e-06, "loss": 0.7412, "step": 51400 }, { "epoch": 42.08, "learning_rate": 1.4203333333333336e-06, "loss": 0.7638, "step": 51500 }, { "epoch": 42.16, "learning_rate": 1.4036666666666668e-06, "loss": 0.7291, "step": 51600 }, { "epoch": 42.24, "learning_rate": 1.387e-06, "loss": 0.7531, "step": 51700 }, { "epoch": 42.32, "learning_rate": 1.3703333333333334e-06, "loss": 0.7164, "step": 51800 }, { "epoch": 42.4, "learning_rate": 1.3536666666666667e-06, "loss": 0.7621, "step": 51900 }, { "epoch": 42.48, "learning_rate": 1.3370000000000003e-06, "loss": 0.7719, "step": 52000 }, { "epoch": 42.57, "learning_rate": 1.3203333333333335e-06, "loss": 0.7192, "step": 52100 }, { "epoch": 42.65, "learning_rate": 1.3036666666666667e-06, "loss": 0.73, "step": 52200 }, { "epoch": 42.73, "learning_rate": 1.2870000000000001e-06, "loss": 0.756, "step": 52300 }, { "epoch": 42.81, "learning_rate": 1.2703333333333333e-06, "loss": 0.7576, "step": 52400 }, { "epoch": 42.89, "learning_rate": 1.2536666666666666e-06, "loss": 0.7161, "step": 52500 }, { "epoch": 42.97, "learning_rate": 1.2370000000000002e-06, "loss": 0.7123, "step": 52600 }, { "epoch": 43.06, "learning_rate": 1.2203333333333334e-06, "loss": 0.7612, "step": 52700 }, { "epoch": 43.14, "learning_rate": 1.2036666666666668e-06, "loss": 0.7216, "step": 52800 }, { "epoch": 43.22, "learning_rate": 1.187e-06, "loss": 0.7363, "step": 52900 }, { "epoch": 43.3, "learning_rate": 1.1705e-06, "loss": 0.7175, "step": 53000 }, { "epoch": 43.38, "learning_rate": 1.1538333333333335e-06, "loss": 0.7358, "step": 53100 }, { "epoch": 43.46, "learning_rate": 1.1371666666666667e-06, "loss": 0.7449, "step": 53200 }, { "epoch": 43.55, "learning_rate": 1.1205000000000001e-06, "loss": 0.7414, "step": 53300 }, { "epoch": 43.63, "learning_rate": 1.1038333333333333e-06, "loss": 0.7333, "step": 53400 }, { "epoch": 43.71, "learning_rate": 1.0871666666666667e-06, "loss": 0.7042, "step": 53500 }, { "epoch": 43.79, "learning_rate": 1.0705000000000002e-06, "loss": 0.7239, "step": 53600 }, { "epoch": 43.87, "learning_rate": 1.0538333333333334e-06, "loss": 0.7493, "step": 53700 }, { "epoch": 43.95, "learning_rate": 1.0371666666666668e-06, "loss": 0.7588, "step": 53800 }, { "epoch": 44.04, "learning_rate": 1.0205e-06, "loss": 0.7311, "step": 53900 }, { "epoch": 44.12, "learning_rate": 1.0038333333333334e-06, "loss": 0.7177, "step": 54000 }, { "epoch": 44.2, "learning_rate": 9.871666666666666e-07, "loss": 0.7187, "step": 54100 }, { "epoch": 44.28, "learning_rate": 9.705e-07, "loss": 0.7348, "step": 54200 }, { "epoch": 44.36, "learning_rate": 9.538333333333335e-07, "loss": 0.7477, "step": 54300 }, { "epoch": 44.44, "learning_rate": 9.371666666666668e-07, "loss": 0.7348, "step": 54400 }, { "epoch": 44.53, "learning_rate": 9.205000000000001e-07, "loss": 0.7133, "step": 54500 }, { "epoch": 44.61, "learning_rate": 9.038333333333333e-07, "loss": 0.7707, "step": 54600 }, { "epoch": 44.69, "learning_rate": 8.871666666666668e-07, "loss": 0.7422, "step": 54700 }, { "epoch": 44.77, "learning_rate": 8.705000000000001e-07, "loss": 0.7397, "step": 54800 }, { "epoch": 44.85, "learning_rate": 8.538333333333335e-07, "loss": 0.7233, "step": 54900 }, { "epoch": 44.93, "learning_rate": 8.371666666666667e-07, "loss": 0.7672, "step": 55000 }, { "epoch": 45.02, "learning_rate": 8.205e-07, "loss": 0.7235, "step": 55100 }, { "epoch": 45.1, "learning_rate": 8.038333333333334e-07, "loss": 0.7442, "step": 55200 }, { "epoch": 45.18, "learning_rate": 7.871666666666668e-07, "loss": 0.7244, "step": 55300 }, { "epoch": 45.26, "learning_rate": 7.705e-07, "loss": 0.7272, "step": 55400 }, { "epoch": 45.34, "learning_rate": 7.538333333333334e-07, "loss": 0.7283, "step": 55500 }, { "epoch": 45.42, "learning_rate": 7.371666666666667e-07, "loss": 0.7126, "step": 55600 }, { "epoch": 45.51, "learning_rate": 7.205000000000001e-07, "loss": 0.7342, "step": 55700 }, { "epoch": 45.59, "learning_rate": 7.038333333333334e-07, "loss": 0.7334, "step": 55800 }, { "epoch": 45.67, "learning_rate": 6.871666666666667e-07, "loss": 0.7709, "step": 55900 }, { "epoch": 45.75, "learning_rate": 6.705000000000001e-07, "loss": 0.7267, "step": 56000 }, { "epoch": 45.83, "learning_rate": 6.538333333333334e-07, "loss": 0.7248, "step": 56100 }, { "epoch": 45.91, "learning_rate": 6.371666666666667e-07, "loss": 0.7289, "step": 56200 }, { "epoch": 46.0, "learning_rate": 6.205e-07, "loss": 0.7192, "step": 56300 }, { "epoch": 46.08, "learning_rate": 6.038333333333333e-07, "loss": 0.7347, "step": 56400 }, { "epoch": 46.16, "learning_rate": 5.871666666666668e-07, "loss": 0.7493, "step": 56500 }, { "epoch": 46.24, "learning_rate": 5.705e-07, "loss": 0.7034, "step": 56600 }, { "epoch": 46.32, "learning_rate": 5.538333333333334e-07, "loss": 0.7668, "step": 56700 }, { "epoch": 46.4, "learning_rate": 5.371666666666667e-07, "loss": 0.7516, "step": 56800 }, { "epoch": 46.49, "learning_rate": 5.205e-07, "loss": 0.7294, "step": 56900 }, { "epoch": 46.57, "learning_rate": 5.038333333333334e-07, "loss": 0.7343, "step": 57000 }, { "epoch": 46.65, "learning_rate": 4.871666666666667e-07, "loss": 0.6948, "step": 57100 }, { "epoch": 46.73, "learning_rate": 4.7050000000000004e-07, "loss": 0.7239, "step": 57200 }, { "epoch": 46.81, "learning_rate": 4.5383333333333335e-07, "loss": 0.7533, "step": 57300 }, { "epoch": 46.89, "learning_rate": 4.371666666666667e-07, "loss": 0.7433, "step": 57400 }, { "epoch": 46.98, "learning_rate": 4.205e-07, "loss": 0.737, "step": 57500 }, { "epoch": 47.06, "learning_rate": 4.0383333333333336e-07, "loss": 0.7363, "step": 57600 }, { "epoch": 47.14, "learning_rate": 3.871666666666667e-07, "loss": 0.7304, "step": 57700 }, { "epoch": 47.22, "learning_rate": 3.7050000000000004e-07, "loss": 0.7017, "step": 57800 }, { "epoch": 47.3, "learning_rate": 3.538333333333334e-07, "loss": 0.7194, "step": 57900 }, { "epoch": 47.39, "learning_rate": 3.371666666666667e-07, "loss": 0.7061, "step": 58000 }, { "epoch": 47.47, "learning_rate": 3.2050000000000004e-07, "loss": 0.7565, "step": 58100 }, { "epoch": 47.55, "learning_rate": 3.04e-07, "loss": 0.7279, "step": 58200 }, { "epoch": 47.63, "learning_rate": 2.8733333333333334e-07, "loss": 0.748, "step": 58300 }, { "epoch": 47.71, "learning_rate": 2.706666666666667e-07, "loss": 0.7418, "step": 58400 }, { "epoch": 47.79, "learning_rate": 2.54e-07, "loss": 0.7147, "step": 58500 }, { "epoch": 47.88, "learning_rate": 2.3733333333333334e-07, "loss": 0.7313, "step": 58600 }, { "epoch": 47.96, "learning_rate": 2.2066666666666666e-07, "loss": 0.7291, "step": 58700 }, { "epoch": 48.04, "learning_rate": 2.0400000000000003e-07, "loss": 0.7145, "step": 58800 }, { "epoch": 48.12, "learning_rate": 1.8733333333333337e-07, "loss": 0.6993, "step": 58900 }, { "epoch": 48.2, "learning_rate": 1.7066666666666669e-07, "loss": 0.706, "step": 59000 }, { "epoch": 48.28, "learning_rate": 1.5400000000000003e-07, "loss": 0.7443, "step": 59100 }, { "epoch": 48.37, "learning_rate": 1.3733333333333335e-07, "loss": 0.7421, "step": 59200 }, { "epoch": 48.45, "learning_rate": 1.2066666666666666e-07, "loss": 0.7562, "step": 59300 }, { "epoch": 48.53, "learning_rate": 1.04e-07, "loss": 0.7263, "step": 59400 }, { "epoch": 48.61, "learning_rate": 8.733333333333333e-08, "loss": 0.7286, "step": 59500 }, { "epoch": 48.69, "learning_rate": 7.066666666666666e-08, "loss": 0.7011, "step": 59600 }, { "epoch": 48.77, "learning_rate": 5.400000000000001e-08, "loss": 0.7223, "step": 59700 }, { "epoch": 48.86, "learning_rate": 3.733333333333334e-08, "loss": 0.6973, "step": 59800 }, { "epoch": 48.94, "learning_rate": 2.066666666666667e-08, "loss": 0.7321, "step": 59900 }, { "epoch": 49.02, "learning_rate": 4e-09, "loss": 0.7392, "step": 60000 }, { "epoch": 49.02, "eval_cer": 0.10614048689719337, "eval_loss": 0.49394938349723816, "eval_runtime": 8.0445, "eval_samples_per_second": 37.292, "eval_steps_per_second": 18.646, "eval_wer": 0.35683760683760685, "step": 60000 } ], "max_steps": 60000, "num_train_epochs": 50, "total_flos": 6.280569561121633e+19, "trial_name": null, "trial_params": null }