diff --git "a/trainer_state.json" "b/trainer_state.json" --- "a/trainer_state.json" +++ "b/trainer_state.json" @@ -21,16290 +21,16290 @@ }, { "epoch": 0.0, - "learning_rate": 6e-07, + "learning_rate": 6.000000000000001e-08, "loss": 9.1561, "step": 3 }, { "epoch": 0.0, - "learning_rate": 6e-07, + "learning_rate": 6.000000000000001e-08, "loss": 9.6171, "step": 4 }, { "epoch": 0.01, - "learning_rate": 1.2e-06, + "learning_rate": 1.2000000000000002e-07, "loss": 9.0606, "step": 5 }, { "epoch": 0.01, - "learning_rate": 1.8e-06, - "loss": 11.2556, + "learning_rate": 1.8e-07, + "loss": 11.2568, "step": 6 }, { "epoch": 0.01, - "learning_rate": 1.8e-06, - "loss": 18.1972, + "learning_rate": 1.8e-07, + "loss": 18.2062, "step": 7 }, { "epoch": 0.01, - "learning_rate": 2.4e-06, - "loss": 8.9491, + "learning_rate": 2.4000000000000003e-07, + "loss": 8.9525, "step": 8 }, { "epoch": 0.01, - "learning_rate": 2.9999999999999997e-06, - "loss": 9.0679, + "learning_rate": 3.0000000000000004e-07, + "loss": 9.0775, "step": 9 }, { "epoch": 0.01, - "learning_rate": 3.6e-06, - "loss": 8.8046, + "learning_rate": 3.6e-07, + "loss": 8.8305, "step": 10 }, { "epoch": 0.01, - "learning_rate": 4.2e-06, - "loss": 8.6018, + "learning_rate": 4.2e-07, + "loss": 8.6673, "step": 11 }, { "epoch": 0.01, - "learning_rate": 4.8e-06, - "loss": 10.8441, + "learning_rate": 4.800000000000001e-07, + "loss": 10.9806, "step": 12 }, { "epoch": 0.01, - "learning_rate": 4.8e-06, - "loss": 28.7373, + "learning_rate": 4.800000000000001e-07, + "loss": 30.1335, "step": 13 }, { "epoch": 0.02, - "learning_rate": 5.399999999999999e-06, - "loss": 8.4937, + "learning_rate": 5.4e-07, + "loss": 8.6737, "step": 14 }, { "epoch": 0.02, - "learning_rate": 5.999999999999999e-06, - "loss": 8.3715, + "learning_rate": 6.000000000000001e-07, + "loss": 8.5914, "step": 15 }, { "epoch": 0.02, - "learning_rate": 6.599999999999999e-06, - "loss": 8.605, + "learning_rate": 6.599999999999999e-07, + "loss": 8.8991, "step": 16 }, { "epoch": 0.02, - "learning_rate": 7.2e-06, - "loss": 8.8051, + "learning_rate": 7.2e-07, + "loss": 9.2613, "step": 17 }, { "epoch": 0.02, - "learning_rate": 7.799999999999998e-06, - "loss": 10.5933, + "learning_rate": 7.799999999999999e-07, + "loss": 11.3974, "step": 18 }, { "epoch": 0.02, - "learning_rate": 7.799999999999998e-06, - "loss": 29.5409, + "learning_rate": 7.799999999999999e-07, + "loss": 35.2593, "step": 19 }, { "epoch": 0.02, - "learning_rate": 8.4e-06, - "loss": 8.5578, + "learning_rate": 8.4e-07, + "loss": 9.2454, "step": 20 }, { "epoch": 0.02, - "learning_rate": 8.999999999999999e-06, - "loss": 8.195, + "learning_rate": 9e-07, + "loss": 9.0118, "step": 21 }, { "epoch": 0.02, - "learning_rate": 9.6e-06, - "loss": 7.7195, + "learning_rate": 9.600000000000001e-07, + "loss": 8.6126, "step": 22 }, { "epoch": 0.03, - "learning_rate": 1.02e-05, - "loss": 8.1096, + "learning_rate": 1.0200000000000002e-06, + "loss": 9.2191, "step": 23 }, { "epoch": 0.03, - "learning_rate": 1.0799999999999998e-05, - "loss": 7.866, + "learning_rate": 1.08e-06, + "loss": 9.249, "step": 24 }, { "epoch": 0.03, - "learning_rate": 1.14e-05, - "loss": 10.9233, + "learning_rate": 1.14e-06, + "loss": 14.3693, "step": 25 }, { "epoch": 0.03, - "learning_rate": 1.1999999999999999e-05, - "loss": 7.1087, + "learning_rate": 1.2000000000000002e-06, + "loss": 8.9806, "step": 26 }, { "epoch": 0.03, - "learning_rate": 1.26e-05, - "loss": 6.8374, + "learning_rate": 1.26e-06, + "loss": 9.0391, "step": 27 }, { "epoch": 0.03, - "learning_rate": 1.3199999999999997e-05, - "loss": 6.3112, + "learning_rate": 1.3199999999999999e-06, + "loss": 8.5903, "step": 28 }, { "epoch": 0.03, - "learning_rate": 1.3799999999999998e-05, - "loss": 6.1265, + "learning_rate": 1.38e-06, + "loss": 9.0014, "step": 29 }, { "epoch": 0.03, - "learning_rate": 1.44e-05, - "loss": 5.7044, + "learning_rate": 1.44e-06, + "loss": 9.3066, "step": 30 }, { "epoch": 0.03, - "learning_rate": 1.4999999999999999e-05, - "loss": 5.7833, + "learning_rate": 1.5e-06, + "loss": 12.2976, "step": 31 }, { "epoch": 0.04, - "learning_rate": 1.5599999999999996e-05, - "loss": 8.5004, + "learning_rate": 1.5599999999999999e-06, + "loss": 23.0703, "step": 32 }, { "epoch": 0.04, - "learning_rate": 1.6199999999999997e-05, - "loss": 4.0818, + "learning_rate": 1.62e-06, + "loss": 8.6068, "step": 33 }, { "epoch": 0.04, - "learning_rate": 1.68e-05, - "loss": 3.7474, + "learning_rate": 1.68e-06, + "loss": 8.8712, "step": 34 }, { "epoch": 0.04, - "learning_rate": 1.74e-05, - "loss": 3.5497, + "learning_rate": 1.74e-06, + "loss": 8.628, "step": 35 }, { "epoch": 0.04, - "learning_rate": 1.7999999999999997e-05, - "loss": 3.4765, + "learning_rate": 1.8e-06, + "loss": 9.0748, "step": 36 }, { "epoch": 0.04, - "learning_rate": 1.8599999999999998e-05, - "loss": 3.4897, + "learning_rate": 1.86e-06, + "loss": 10.7939, "step": 37 }, { "epoch": 0.04, - "learning_rate": 1.92e-05, - "loss": 4.5035, + "learning_rate": 1.9200000000000003e-06, + "loss": 24.4837, "step": 38 }, { "epoch": 0.04, - "learning_rate": 1.98e-05, - "loss": 3.4165, + "learning_rate": 1.98e-06, + "loss": 8.2996, "step": 39 }, { "epoch": 0.04, - "learning_rate": 2.04e-05, - "loss": 3.3678, + "learning_rate": 2.0400000000000004e-06, + "loss": 8.486, "step": 40 }, { "epoch": 0.05, - "learning_rate": 2.1e-05, - "loss": 3.2659, + "learning_rate": 2.1000000000000002e-06, + "loss": 8.2839, "step": 41 }, { "epoch": 0.05, - "learning_rate": 2.1599999999999996e-05, - "loss": 3.1835, + "learning_rate": 2.16e-06, + "loss": 9.0285, "step": 42 }, { "epoch": 0.05, - "learning_rate": 2.2199999999999998e-05, - "loss": 3.2212, + "learning_rate": 2.22e-06, + "loss": 10.2284, "step": 43 }, { "epoch": 0.05, - "learning_rate": 2.28e-05, - "loss": 3.8846, + "learning_rate": 2.28e-06, + "loss": 25.0483, "step": 44 }, { "epoch": 0.05, - "learning_rate": 2.34e-05, - "loss": 3.0881, + "learning_rate": 2.34e-06, + "loss": 8.8377, "step": 45 }, { "epoch": 0.05, - "learning_rate": 2.3999999999999997e-05, - "loss": 3.0452, + "learning_rate": 2.4000000000000003e-06, + "loss": 8.33, "step": 46 }, { "epoch": 0.05, - "learning_rate": 2.4599999999999998e-05, - "loss": 3.0327, + "learning_rate": 2.46e-06, + "loss": 9.0151, "step": 47 }, { "epoch": 0.05, - "learning_rate": 2.52e-05, - "loss": 3.0034, + "learning_rate": 2.52e-06, + "loss": 8.0929, "step": 48 }, { "epoch": 0.05, - "learning_rate": 2.5799999999999997e-05, - "loss": 2.9872, + "learning_rate": 2.58e-06, + "loss": 8.9703, "step": 49 }, { "epoch": 0.06, - "learning_rate": 2.6399999999999995e-05, - "loss": 3.6616, + "learning_rate": 2.6399999999999997e-06, + "loss": 16.031, "step": 50 }, { "epoch": 0.06, - "learning_rate": 2.6999999999999996e-05, - "loss": 2.9641, + "learning_rate": 2.7e-06, + "loss": 8.2522, "step": 51 }, { "epoch": 0.06, - "learning_rate": 2.7599999999999997e-05, - "loss": 2.9352, + "learning_rate": 2.76e-06, + "loss": 8.0357, "step": 52 }, { "epoch": 0.06, - "learning_rate": 2.8199999999999998e-05, - "loss": 2.9209, + "learning_rate": 2.82e-06, + "loss": 8.0506, "step": 53 }, { "epoch": 0.06, - "learning_rate": 2.88e-05, - "loss": 2.9176, + "learning_rate": 2.88e-06, + "loss": 8.3815, "step": 54 }, { "epoch": 0.06, - "learning_rate": 2.94e-05, - "loss": 2.9116, + "learning_rate": 2.9400000000000002e-06, + "loss": 8.1559, "step": 55 }, { "epoch": 0.06, - "learning_rate": 2.9999999999999997e-05, - "loss": 3.0188, + "learning_rate": 3e-06, + "loss": 11.5317, "step": 56 }, { "epoch": 0.06, - "learning_rate": 3.06e-05, - "loss": 3.7112, + "learning_rate": 3e-06, + "loss": 18.6265, "step": 57 }, { "epoch": 0.07, - "learning_rate": 3.119999999999999e-05, - "loss": 2.8784, + "learning_rate": 3.06e-06, + "loss": 7.8839, "step": 58 }, { "epoch": 0.07, - "learning_rate": 3.1799999999999994e-05, - "loss": 2.8898, + "learning_rate": 3.1199999999999998e-06, + "loss": 7.3116, "step": 59 }, { "epoch": 0.07, - "learning_rate": 3.2399999999999995e-05, - "loss": 2.8928, + "learning_rate": 3.18e-06, + "loss": 8.0086, "step": 60 }, { "epoch": 0.07, - "learning_rate": 3.2999999999999996e-05, - "loss": 2.8816, + "learning_rate": 3.24e-06, + "loss": 7.6336, "step": 61 }, { "epoch": 0.07, - "learning_rate": 3.36e-05, - "loss": 2.9159, + "learning_rate": 3.3e-06, + "loss": 9.2959, "step": 62 }, { "epoch": 0.07, - "learning_rate": 3.42e-05, - "loss": 3.006, + "learning_rate": 3.36e-06, + "loss": 15.4261, "step": 63 }, { "epoch": 0.07, - "learning_rate": 3.48e-05, - "loss": 2.8547, + "learning_rate": 3.4200000000000003e-06, + "loss": 7.1528, "step": 64 }, { "epoch": 0.07, - "learning_rate": 3.539999999999999e-05, - "loss": 2.8681, + "learning_rate": 3.48e-06, + "loss": 6.985, "step": 65 }, { "epoch": 0.07, - "learning_rate": 3.5999999999999994e-05, - "loss": 2.8717, + "learning_rate": 3.54e-06, + "loss": 7.0016, "step": 66 }, { "epoch": 0.08, - "learning_rate": 3.6599999999999995e-05, - "loss": 2.9001, + "learning_rate": 3.6e-06, + "loss": 7.1214, "step": 67 }, { "epoch": 0.08, - "learning_rate": 3.7199999999999996e-05, - "loss": 2.8589, + "learning_rate": 3.66e-06, + "loss": 7.3454, "step": 68 }, { "epoch": 0.08, - "learning_rate": 3.78e-05, - "loss": 2.9983, + "learning_rate": 3.72e-06, + "loss": 11.7396, "step": 69 }, { "epoch": 0.08, - "learning_rate": 3.84e-05, - "loss": 2.8826, + "learning_rate": 3.7800000000000002e-06, + "loss": 6.2014, "step": 70 }, { "epoch": 0.08, - "learning_rate": 3.9e-05, - "loss": 2.8817, + "learning_rate": 3.8400000000000005e-06, + "loss": 6.1452, "step": 71 }, { "epoch": 0.08, - "learning_rate": 3.96e-05, - "loss": 2.8707, + "learning_rate": 3.9e-06, + "loss": 6.14, "step": 72 }, { "epoch": 0.08, - "learning_rate": 4.02e-05, - "loss": 2.8787, + "learning_rate": 3.96e-06, + "loss": 5.9955, "step": 73 }, { "epoch": 0.08, - "learning_rate": 4.08e-05, - "loss": 2.9377, + "learning_rate": 4.0200000000000005e-06, + "loss": 6.1978, "step": 74 }, { "epoch": 0.08, - "learning_rate": 4.14e-05, - "loss": 3.1176, + "learning_rate": 4.080000000000001e-06, + "loss": 7.0407, "step": 75 }, { "epoch": 0.09, - "learning_rate": 4.2e-05, - "loss": 2.863, + "learning_rate": 4.14e-06, + "loss": 5.4833, "step": 76 }, { "epoch": 0.09, - "learning_rate": 4.259999999999999e-05, - "loss": 2.8484, + "learning_rate": 4.2000000000000004e-06, + "loss": 5.3788, "step": 77 }, { "epoch": 0.09, - "learning_rate": 4.319999999999999e-05, - "loss": 2.8717, + "learning_rate": 4.26e-06, + "loss": 5.2096, "step": 78 }, { "epoch": 0.09, - "learning_rate": 4.3799999999999994e-05, - "loss": 2.8582, + "learning_rate": 4.32e-06, + "loss": 5.0098, "step": 79 }, { "epoch": 0.09, - "learning_rate": 4.4399999999999995e-05, - "loss": 2.8632, + "learning_rate": 4.3799999999999996e-06, + "loss": 4.4992, "step": 80 }, { "epoch": 0.09, - "learning_rate": 4.4999999999999996e-05, - "loss": 2.895, + "learning_rate": 4.44e-06, + "loss": 4.5938, "step": 81 }, { "epoch": 0.09, - "learning_rate": 4.56e-05, - "loss": 3.0682, + "learning_rate": 4.5e-06, + "loss": 5.7692, "step": 82 }, { "epoch": 0.09, - "learning_rate": 4.62e-05, - "loss": 2.8878, + "learning_rate": 4.56e-06, + "loss": 3.9145, "step": 83 }, { "epoch": 0.09, - "learning_rate": 4.68e-05, - "loss": 2.8604, + "learning_rate": 4.62e-06, + "loss": 3.8004, "step": 84 }, { "epoch": 0.1, - "learning_rate": 4.7399999999999993e-05, - "loss": 2.8689, + "learning_rate": 4.68e-06, + "loss": 3.6052, "step": 85 }, { "epoch": 0.1, - "learning_rate": 4.7999999999999994e-05, - "loss": 2.8964, + "learning_rate": 4.74e-06, + "loss": 3.5676, "step": 86 }, { "epoch": 0.1, - "learning_rate": 4.8599999999999995e-05, - "loss": 2.9494, + "learning_rate": 4.800000000000001e-06, + "loss": 3.6966, "step": 87 }, { "epoch": 0.1, - "learning_rate": 4.9199999999999997e-05, - "loss": 2.9488, + "learning_rate": 4.86e-06, + "loss": 5.7616, "step": 88 }, { "epoch": 0.1, - "learning_rate": 4.98e-05, - "loss": 2.8596, + "learning_rate": 4.92e-06, + "loss": 3.4516, "step": 89 }, { "epoch": 0.1, - "learning_rate": 5.04e-05, - "loss": 2.8545, + "learning_rate": 4.980000000000001e-06, + "loss": 3.443, "step": 90 }, { "epoch": 0.1, - "learning_rate": 5.1e-05, - "loss": 2.8467, + "learning_rate": 5.04e-06, + "loss": 3.4251, "step": 91 }, { "epoch": 0.1, - "learning_rate": 5.1599999999999994e-05, - "loss": 2.8689, + "learning_rate": 5.1e-06, + "loss": 3.392, "step": 92 }, { "epoch": 0.1, - "learning_rate": 5.2199999999999995e-05, - "loss": 2.8921, + "learning_rate": 5.16e-06, + "loss": 3.3725, "step": 93 }, { "epoch": 0.11, - "learning_rate": 5.279999999999999e-05, - "loss": 3.046, + "learning_rate": 5.22e-06, + "loss": 4.7644, "step": 94 }, { "epoch": 0.11, - "learning_rate": 5.339999999999999e-05, - "loss": 2.8741, + "learning_rate": 5.279999999999999e-06, + "loss": 3.2849, "step": 95 }, { "epoch": 0.11, - "learning_rate": 5.399999999999999e-05, - "loss": 2.8586, + "learning_rate": 5.34e-06, + "loss": 3.2496, "step": 96 }, { "epoch": 0.11, - "learning_rate": 5.459999999999999e-05, - "loss": 2.8565, + "learning_rate": 5.4e-06, + "loss": 3.2185, "step": 97 }, { "epoch": 0.11, - "learning_rate": 5.519999999999999e-05, - "loss": 2.8545, + "learning_rate": 5.46e-06, + "loss": 3.1935, "step": 98 }, { "epoch": 0.11, - "learning_rate": 5.5799999999999994e-05, - "loss": 2.8755, + "learning_rate": 5.52e-06, + "loss": 3.2266, "step": 99 }, { "epoch": 0.11, - "learning_rate": 5.6399999999999995e-05, - "loss": 2.9558, + "learning_rate": 5.58e-06, + "loss": 3.4274, "step": 100 }, { "epoch": 0.11, - "eval_loss": 2.9238102436065674, - "eval_runtime": 49.2733, - "eval_samples_per_second": 54.857, - "eval_steps_per_second": 0.873, + "eval_loss": 4.141883850097656, + "eval_runtime": 48.2643, + "eval_samples_per_second": 56.004, + "eval_steps_per_second": 0.891, "eval_wer": 1.0, "step": 100 }, { "epoch": 0.11, - "learning_rate": 5.6999999999999996e-05, - "loss": 2.8881, + "learning_rate": 5.64e-06, + "loss": 3.1286, "step": 101 }, { "epoch": 0.11, - "learning_rate": 5.76e-05, - "loss": 2.8645, + "learning_rate": 5.7000000000000005e-06, + "loss": 3.1, "step": 102 }, { "epoch": 0.12, - "learning_rate": 5.82e-05, - "loss": 2.8544, + "learning_rate": 5.76e-06, + "loss": 3.0865, "step": 103 }, { "epoch": 0.12, - "learning_rate": 5.88e-05, - "loss": 2.8582, + "learning_rate": 5.82e-06, + "loss": 3.0593, "step": 104 }, { "epoch": 0.12, - "learning_rate": 5.94e-05, - "loss": 2.859, + "learning_rate": 5.8800000000000005e-06, + "loss": 3.0541, "step": 105 }, { "epoch": 0.12, - "learning_rate": 5.9999999999999995e-05, - "loss": 2.8868, + "learning_rate": 5.940000000000001e-06, + "loss": 3.1007, "step": 106 }, { "epoch": 0.12, - "learning_rate": 6.0599999999999996e-05, - "loss": 2.948, + "learning_rate": 6e-06, + "loss": 3.8774, "step": 107 }, { "epoch": 0.12, - "learning_rate": 6.12e-05, - "loss": 2.9054, + "learning_rate": 6.0600000000000004e-06, + "loss": 3.0093, "step": 108 }, { "epoch": 0.12, - "learning_rate": 6.18e-05, - "loss": 2.9458, + "learning_rate": 6.12e-06, + "loss": 2.9985, "step": 109 }, { "epoch": 0.12, - "learning_rate": 6.239999999999999e-05, - "loss": 2.9522, + "learning_rate": 6.18e-06, + "loss": 3.0035, "step": 110 }, { "epoch": 0.12, - "learning_rate": 6.299999999999999e-05, - "loss": 2.887, + "learning_rate": 6.2399999999999995e-06, + "loss": 2.988, "step": 111 }, { "epoch": 0.13, - "learning_rate": 6.359999999999999e-05, - "loss": 2.9583, + "learning_rate": 6.3e-06, + "loss": 3.0032, "step": 112 }, { "epoch": 0.13, - "learning_rate": 6.419999999999999e-05, - "loss": 3.0663, + "learning_rate": 6.36e-06, + "loss": 3.6984, "step": 113 }, { "epoch": 0.13, - "learning_rate": 6.479999999999999e-05, - "loss": 3.2776, + "learning_rate": 6.42e-06, + "loss": 2.9702, "step": 114 }, { "epoch": 0.13, - "learning_rate": 6.539999999999999e-05, - "loss": 3.1604, + "learning_rate": 6.48e-06, + "loss": 2.9481, "step": 115 }, { "epoch": 0.13, - "learning_rate": 6.599999999999999e-05, - "loss": 3.0539, + "learning_rate": 6.54e-06, + "loss": 2.96, "step": 116 }, { "epoch": 0.13, - "learning_rate": 6.659999999999999e-05, - "loss": 2.894, + "learning_rate": 6.6e-06, + "loss": 2.9263, "step": 117 }, { "epoch": 0.13, - "learning_rate": 6.72e-05, - "loss": 2.8832, + "learning_rate": 6.660000000000001e-06, + "loss": 2.9427, "step": 118 }, { "epoch": 0.13, - "learning_rate": 6.78e-05, - "loss": 3.6111, + "learning_rate": 6.72e-06, + "loss": 3.4858, "step": 119 }, { "epoch": 0.13, - "learning_rate": 6.84e-05, - "loss": 3.0338, + "learning_rate": 6.78e-06, + "loss": 2.9144, "step": 120 }, { "epoch": 0.14, - "learning_rate": 6.9e-05, - "loss": 3.1392, + "learning_rate": 6.840000000000001e-06, + "loss": 2.9067, "step": 121 }, { "epoch": 0.14, - "learning_rate": 6.96e-05, - "loss": 3.1065, + "learning_rate": 6.900000000000001e-06, + "loss": 2.9029, "step": 122 }, { "epoch": 0.14, - "learning_rate": 7.02e-05, - "loss": 2.9959, + "learning_rate": 6.96e-06, + "loss": 2.9001, "step": 123 }, { "epoch": 0.14, - "learning_rate": 7.079999999999999e-05, - "loss": 2.9141, + "learning_rate": 7.0200000000000006e-06, + "loss": 2.8996, "step": 124 }, { "epoch": 0.14, - "learning_rate": 7.139999999999999e-05, - "loss": 3.1355, + "learning_rate": 7.08e-06, + "loss": 2.9885, "step": 125 }, { "epoch": 0.14, - "learning_rate": 7.199999999999999e-05, - "loss": 2.9686, + "learning_rate": 7.14e-06, + "loss": 2.8884, "step": 126 }, { "epoch": 0.14, - "learning_rate": 7.259999999999999e-05, - "loss": 2.9157, + "learning_rate": 7.2e-06, + "loss": 2.8877, "step": 127 }, { "epoch": 0.14, - "learning_rate": 7.319999999999999e-05, - "loss": 2.8561, + "learning_rate": 7.26e-06, + "loss": 2.8755, "step": 128 }, { "epoch": 0.14, - "learning_rate": 7.379999999999999e-05, - "loss": 2.8984, + "learning_rate": 7.32e-06, + "loss": 2.8803, "step": 129 }, { "epoch": 0.15, - "learning_rate": 7.439999999999999e-05, - "loss": 2.9527, + "learning_rate": 7.3800000000000005e-06, + "loss": 2.8804, "step": 130 }, { "epoch": 0.15, - "learning_rate": 7.5e-05, - "loss": 2.9885, + "learning_rate": 7.44e-06, + "loss": 2.9005, "step": 131 }, { "epoch": 0.15, - "learning_rate": 7.56e-05, - "loss": 3.7413, + "learning_rate": 7.5e-06, + "loss": 3.1553, "step": 132 }, { "epoch": 0.15, - "learning_rate": 7.62e-05, - "loss": 2.8833, + "learning_rate": 7.5600000000000005e-06, + "loss": 2.8638, "step": 133 }, { "epoch": 0.15, - "learning_rate": 7.68e-05, - "loss": 2.8906, + "learning_rate": 7.62e-06, + "loss": 2.8755, "step": 134 }, { "epoch": 0.15, - "learning_rate": 7.74e-05, - "loss": 2.8718, + "learning_rate": 7.680000000000001e-06, + "loss": 2.8612, "step": 135 }, { "epoch": 0.15, - "learning_rate": 7.8e-05, - "loss": 2.8609, + "learning_rate": 7.74e-06, + "loss": 2.8595, "step": 136 }, { "epoch": 0.15, - "learning_rate": 7.86e-05, - "loss": 2.9341, + "learning_rate": 7.8e-06, + "loss": 2.8934, "step": 137 }, { "epoch": 0.15, - "learning_rate": 7.92e-05, - "loss": 3.563, + "learning_rate": 7.860000000000001e-06, + "loss": 3.0252, "step": 138 }, { "epoch": 0.16, - "learning_rate": 7.98e-05, - "loss": 2.8933, + "learning_rate": 7.92e-06, + "loss": 2.8963, "step": 139 }, { "epoch": 0.16, - "learning_rate": 8.04e-05, - "loss": 2.925, + "learning_rate": 7.98e-06, + "loss": 2.8941, "step": 140 }, { "epoch": 0.16, - "learning_rate": 8.1e-05, - "loss": 2.9263, + "learning_rate": 8.040000000000001e-06, + "loss": 2.89, "step": 141 }, { "epoch": 0.16, - "learning_rate": 8.16e-05, - "loss": 2.8969, + "learning_rate": 8.1e-06, + "loss": 2.8698, "step": 142 }, { "epoch": 0.16, - "learning_rate": 8.22e-05, - "loss": 2.8666, + "learning_rate": 8.160000000000001e-06, + "loss": 2.8627, "step": 143 }, { "epoch": 0.16, - "learning_rate": 8.28e-05, - "loss": 3.0954, + "learning_rate": 8.220000000000001e-06, + "loss": 2.9767, "step": 144 }, { "epoch": 0.16, - "learning_rate": 8.34e-05, - "loss": 2.8754, + "learning_rate": 8.28e-06, + "loss": 2.8814, "step": 145 }, { "epoch": 0.16, - "learning_rate": 8.4e-05, - "loss": 2.8502, + "learning_rate": 8.340000000000001e-06, + "loss": 2.8599, "step": 146 }, { "epoch": 0.16, - "learning_rate": 8.459999999999998e-05, - "loss": 2.86, + "learning_rate": 8.400000000000001e-06, + "loss": 2.8633, "step": 147 }, { "epoch": 0.17, - "learning_rate": 8.519999999999998e-05, - "loss": 2.8728, + "learning_rate": 8.459999999999999e-06, + "loss": 2.8555, "step": 148 }, { "epoch": 0.17, - "learning_rate": 8.579999999999998e-05, - "loss": 2.8918, + "learning_rate": 8.52e-06, + "loss": 2.8747, "step": 149 }, { "epoch": 0.17, - "learning_rate": 8.639999999999999e-05, - "loss": 2.966, + "learning_rate": 8.58e-06, + "loss": 2.9209, "step": 150 }, { "epoch": 0.17, - "learning_rate": 8.699999999999999e-05, - "loss": 2.9058, + "learning_rate": 8.64e-06, + "loss": 2.8598, "step": 151 }, { "epoch": 0.17, - "learning_rate": 8.759999999999999e-05, - "loss": 2.9453, + "learning_rate": 8.7e-06, + "loss": 2.8579, "step": 152 }, { "epoch": 0.17, - "learning_rate": 8.819999999999999e-05, - "loss": 2.9213, + "learning_rate": 8.759999999999999e-06, + "loss": 2.8576, "step": 153 }, { "epoch": 0.17, - "learning_rate": 8.879999999999999e-05, - "loss": 2.8648, + "learning_rate": 8.82e-06, + "loss": 2.8477, "step": 154 }, { "epoch": 0.17, - "learning_rate": 8.939999999999999e-05, - "loss": 2.8692, + "learning_rate": 8.88e-06, + "loss": 2.8686, "step": 155 }, { "epoch": 0.17, - "learning_rate": 8.999999999999999e-05, - "loss": 3.0057, + "learning_rate": 8.939999999999999e-06, + "loss": 2.8905, "step": 156 }, { "epoch": 0.18, - "learning_rate": 9.059999999999999e-05, - "loss": 3.012, + "learning_rate": 9e-06, + "loss": 2.8926, "step": 157 }, { "epoch": 0.18, - "learning_rate": 9.12e-05, - "loss": 2.9408, + "learning_rate": 9.06e-06, + "loss": 2.8477, "step": 158 }, { "epoch": 0.18, - "learning_rate": 9.18e-05, - "loss": 2.9086, + "learning_rate": 9.12e-06, + "loss": 2.8372, "step": 159 }, { "epoch": 0.18, - "learning_rate": 9.24e-05, - "loss": 2.8869, + "learning_rate": 9.18e-06, + "loss": 2.8458, "step": 160 }, { "epoch": 0.18, - "learning_rate": 9.3e-05, - "loss": 2.8712, + "learning_rate": 9.24e-06, + "loss": 2.8603, "step": 161 }, { "epoch": 0.18, - "learning_rate": 9.36e-05, - "loss": 2.9554, + "learning_rate": 9.3e-06, + "loss": 2.8877, "step": 162 }, { "epoch": 0.18, - "learning_rate": 9.419999999999999e-05, - "loss": 4.1712, + "learning_rate": 9.36e-06, + "loss": 2.9358, "step": 163 }, { "epoch": 0.18, - "learning_rate": 9.479999999999999e-05, - "loss": 2.8689, + "learning_rate": 9.42e-06, + "loss": 2.8634, "step": 164 }, { "epoch": 0.18, - "learning_rate": 9.539999999999999e-05, - "loss": 2.8324, + "learning_rate": 9.48e-06, + "loss": 2.835, "step": 165 }, { "epoch": 0.19, - "learning_rate": 9.599999999999999e-05, - "loss": 2.8584, + "learning_rate": 9.54e-06, + "loss": 2.8496, "step": 166 }, { "epoch": 0.19, - "learning_rate": 9.659999999999999e-05, - "loss": 2.8735, + "learning_rate": 9.600000000000001e-06, + "loss": 2.8657, "step": 167 }, { "epoch": 0.19, - "learning_rate": 9.719999999999999e-05, - "loss": 2.8972, + "learning_rate": 9.66e-06, + "loss": 2.8915, "step": 168 }, { "epoch": 0.19, - "learning_rate": 9.779999999999999e-05, - "loss": 3.5018, + "learning_rate": 9.72e-06, + "loss": 2.9453, "step": 169 }, { "epoch": 0.19, - "learning_rate": 9.839999999999999e-05, - "loss": 2.8614, + "learning_rate": 9.780000000000001e-06, + "loss": 2.8537, "step": 170 }, { "epoch": 0.19, - "learning_rate": 9.9e-05, - "loss": 2.8616, + "learning_rate": 9.84e-06, + "loss": 2.8585, "step": 171 }, { "epoch": 0.19, - "learning_rate": 9.96e-05, - "loss": 2.8358, + "learning_rate": 9.9e-06, + "loss": 2.8401, "step": 172 }, { "epoch": 0.19, - "learning_rate": 0.0001002, - "loss": 2.85, + "learning_rate": 9.960000000000001e-06, + "loss": 2.8295, "step": 173 }, { "epoch": 0.2, - "learning_rate": 0.0001008, - "loss": 2.8914, + "learning_rate": 1.002e-05, + "loss": 2.8425, "step": 174 }, { "epoch": 0.2, - "learning_rate": 0.0001014, - "loss": 2.9718, + "learning_rate": 1.008e-05, + "loss": 2.9424, "step": 175 }, { "epoch": 0.2, - "learning_rate": 0.000102, - "loss": 2.8523, + "learning_rate": 1.0140000000000001e-05, + "loss": 2.8608, "step": 176 }, { "epoch": 0.2, - "learning_rate": 0.0001026, - "loss": 2.8565, + "learning_rate": 1.02e-05, + "loss": 2.8358, "step": 177 }, { "epoch": 0.2, - "learning_rate": 0.00010319999999999999, - "loss": 2.9154, + "learning_rate": 1.0260000000000002e-05, + "loss": 2.8661, "step": 178 }, { "epoch": 0.2, - "learning_rate": 0.00010379999999999999, - "loss": 2.8821, + "learning_rate": 1.032e-05, + "loss": 2.8527, "step": 179 }, { "epoch": 0.2, - "learning_rate": 0.00010439999999999999, - "loss": 2.8496, + "learning_rate": 1.0379999999999999e-05, + "loss": 2.8485, "step": 180 }, { "epoch": 0.2, - "learning_rate": 0.00010499999999999999, - "loss": 2.8642, + "learning_rate": 1.044e-05, + "loss": 2.8796, "step": 181 }, { "epoch": 0.2, - "learning_rate": 0.00010559999999999998, - "loss": 3.0432, + "learning_rate": 1.05e-05, + "loss": 2.9178, "step": 182 }, { "epoch": 0.21, - "learning_rate": 0.00010619999999999998, - "loss": 2.8936, + "learning_rate": 1.0559999999999999e-05, + "loss": 2.8379, "step": 183 }, { "epoch": 0.21, - "learning_rate": 0.00010679999999999998, - "loss": 2.8616, + "learning_rate": 1.062e-05, + "loss": 2.8425, "step": 184 }, { "epoch": 0.21, - "learning_rate": 0.00010739999999999998, - "loss": 2.8334, + "learning_rate": 1.068e-05, + "loss": 2.8448, "step": 185 }, { "epoch": 0.21, - "learning_rate": 0.00010799999999999998, - "loss": 2.8687, + "learning_rate": 1.074e-05, + "loss": 2.8529, "step": 186 }, { "epoch": 0.21, - "learning_rate": 0.00010859999999999998, - "loss": 2.9295, + "learning_rate": 1.08e-05, + "loss": 2.8523, "step": 187 }, { "epoch": 0.21, - "learning_rate": 0.00010919999999999998, - "loss": 2.9511, + "learning_rate": 1.086e-05, + "loss": 2.9041, "step": 188 }, { "epoch": 0.21, - "learning_rate": 0.00010979999999999999, - "loss": 2.87, + "learning_rate": 1.092e-05, + "loss": 2.855, "step": 189 }, { "epoch": 0.21, - "learning_rate": 0.00011039999999999999, - "loss": 2.841, + "learning_rate": 1.098e-05, + "loss": 2.8735, "step": 190 }, { "epoch": 0.21, - "learning_rate": 0.00011099999999999999, - "loss": 2.8369, + "learning_rate": 1.104e-05, + "loss": 2.8531, "step": 191 }, { "epoch": 0.22, - "learning_rate": 0.00011159999999999999, - "loss": 2.8713, + "learning_rate": 1.11e-05, + "loss": 2.8455, "step": 192 }, { "epoch": 0.22, - "learning_rate": 0.00011219999999999999, - "loss": 2.8801, + "learning_rate": 1.116e-05, + "loss": 2.8465, "step": 193 }, { "epoch": 0.22, - "learning_rate": 0.00011279999999999999, - "loss": 3.0974, + "learning_rate": 1.1220000000000001e-05, + "loss": 2.9433, "step": 194 }, { "epoch": 0.22, - "learning_rate": 0.00011339999999999999, - "loss": 2.8318, + "learning_rate": 1.128e-05, + "loss": 2.867, "step": 195 }, { "epoch": 0.22, - "learning_rate": 0.00011399999999999999, - "loss": 2.8437, + "learning_rate": 1.134e-05, + "loss": 2.8726, "step": 196 }, { "epoch": 0.22, - "learning_rate": 0.0001146, - "loss": 2.8527, + "learning_rate": 1.1400000000000001e-05, + "loss": 2.8659, "step": 197 }, { "epoch": 0.22, - "learning_rate": 0.0001152, - "loss": 2.8549, + "learning_rate": 1.146e-05, + "loss": 2.8491, "step": 198 }, { "epoch": 0.22, - "learning_rate": 0.0001158, - "loss": 2.8576, + "learning_rate": 1.152e-05, + "loss": 2.856, "step": 199 }, { "epoch": 0.22, - "learning_rate": 0.0001164, - "loss": 2.8978, + "learning_rate": 1.1580000000000001e-05, + "loss": 2.9657, "step": 200 }, { "epoch": 0.22, - "eval_loss": 2.9613687992095947, - "eval_runtime": 48.443, - "eval_samples_per_second": 55.797, - "eval_steps_per_second": 0.888, + "eval_loss": 3.1203415393829346, + "eval_runtime": 47.2842, + "eval_samples_per_second": 57.165, + "eval_steps_per_second": 0.909, "eval_wer": 1.0, "step": 200 }, { "epoch": 0.23, - "learning_rate": 0.000117, - "loss": 2.843, + "learning_rate": 1.164e-05, + "loss": 2.8505, "step": 201 }, { "epoch": 0.23, - "learning_rate": 0.0001176, - "loss": 2.8997, + "learning_rate": 1.1700000000000001e-05, + "loss": 2.8814, "step": 202 }, { "epoch": 0.23, - "learning_rate": 0.0001182, - "loss": 2.8765, + "learning_rate": 1.1760000000000001e-05, + "loss": 2.8671, "step": 203 }, { "epoch": 0.23, - "learning_rate": 0.0001188, - "loss": 2.8628, + "learning_rate": 1.182e-05, + "loss": 2.8535, "step": 204 }, { "epoch": 0.23, - "learning_rate": 0.0001194, - "loss": 2.839, + "learning_rate": 1.1880000000000001e-05, + "loss": 2.8443, "step": 205 }, { "epoch": 0.23, - "learning_rate": 0.00011999999999999999, - "loss": 2.8732, + "learning_rate": 1.1940000000000001e-05, + "loss": 2.8618, "step": 206 }, { "epoch": 0.23, - "learning_rate": 0.00012059999999999999, - "loss": 2.9547, + "learning_rate": 1.2e-05, + "loss": 2.8921, "step": 207 }, { "epoch": 0.23, - "learning_rate": 0.00012119999999999999, - "loss": 2.986, + "learning_rate": 1.2060000000000001e-05, + "loss": 2.8482, "step": 208 }, { "epoch": 0.23, - "learning_rate": 0.00012179999999999999, - "loss": 3.0205, + "learning_rate": 1.2120000000000001e-05, + "loss": 2.8446, "step": 209 }, { "epoch": 0.24, - "learning_rate": 0.0001224, - "loss": 2.9825, + "learning_rate": 1.2180000000000002e-05, + "loss": 2.8358, "step": 210 }, { "epoch": 0.24, - "learning_rate": 0.00012299999999999998, - "loss": 2.9385, + "learning_rate": 1.224e-05, + "loss": 2.8546, "step": 211 }, { "epoch": 0.24, - "learning_rate": 0.0001236, - "loss": 2.8995, + "learning_rate": 1.2299999999999999e-05, + "loss": 2.8722, "step": 212 }, { "epoch": 0.24, - "learning_rate": 0.00012419999999999998, - "loss": 2.9027, + "learning_rate": 1.236e-05, + "loss": 2.9072, "step": 213 }, { "epoch": 0.24, - "learning_rate": 0.00012479999999999997, - "loss": 2.9209, + "learning_rate": 1.242e-05, + "loss": 2.8363, "step": 214 }, { "epoch": 0.24, - "learning_rate": 0.00012539999999999999, - "loss": 3.0299, + "learning_rate": 1.2479999999999999e-05, + "loss": 2.8555, "step": 215 }, { "epoch": 0.24, - "learning_rate": 0.00012599999999999997, - "loss": 2.9825, + "learning_rate": 1.254e-05, + "loss": 2.8306, "step": 216 }, { "epoch": 0.24, - "learning_rate": 0.0001266, - "loss": 2.9288, + "learning_rate": 1.26e-05, + "loss": 2.8512, "step": 217 }, { "epoch": 0.24, - "learning_rate": 0.00012719999999999997, - "loss": 2.9049, + "learning_rate": 1.2659999999999999e-05, + "loss": 2.8592, "step": 218 }, { "epoch": 0.25, - "learning_rate": 0.0001278, - "loss": 3.4437, + "learning_rate": 1.272e-05, + "loss": 2.8846, "step": 219 }, { "epoch": 0.25, - "learning_rate": 0.00012839999999999998, - "loss": 2.8304, + "learning_rate": 1.278e-05, + "loss": 2.8348, "step": 220 }, { "epoch": 0.25, - "learning_rate": 0.000129, - "loss": 2.8716, + "learning_rate": 1.284e-05, + "loss": 2.8395, "step": 221 }, { "epoch": 0.25, - "learning_rate": 0.00012959999999999998, - "loss": 2.8629, + "learning_rate": 1.29e-05, + "loss": 2.8253, "step": 222 }, { "epoch": 0.25, - "learning_rate": 0.0001302, - "loss": 2.8708, + "learning_rate": 1.296e-05, + "loss": 2.8549, "step": 223 }, { "epoch": 0.25, - "learning_rate": 0.00013079999999999998, - "loss": 2.8501, + "learning_rate": 1.302e-05, + "loss": 2.8474, "step": 224 }, { "epoch": 0.25, - "learning_rate": 0.0001314, - "loss": 3.0454, + "learning_rate": 1.308e-05, + "loss": 2.8839, "step": 225 }, { "epoch": 0.25, - "learning_rate": 0.00013199999999999998, - "loss": 2.8276, + "learning_rate": 1.314e-05, + "loss": 2.8358, "step": 226 }, { "epoch": 0.25, - "learning_rate": 0.0001326, - "loss": 2.8319, + "learning_rate": 1.32e-05, + "loss": 2.8391, "step": 227 }, { "epoch": 0.26, - "learning_rate": 0.00013319999999999999, - "loss": 2.8372, + "learning_rate": 1.326e-05, + "loss": 2.8433, "step": 228 }, { "epoch": 0.26, - "learning_rate": 0.0001338, - "loss": 2.8405, + "learning_rate": 1.3320000000000001e-05, + "loss": 2.8476, "step": 229 }, { "epoch": 0.26, - "learning_rate": 0.0001344, - "loss": 2.8471, + "learning_rate": 1.338e-05, + "loss": 2.8387, "step": 230 }, { "epoch": 0.26, - "learning_rate": 0.000135, - "loss": 2.8856, + "learning_rate": 1.344e-05, + "loss": 2.877, "step": 231 }, { "epoch": 0.26, - "learning_rate": 0.0001356, - "loss": 2.99, + "learning_rate": 1.3500000000000001e-05, + "loss": 2.8735, "step": 232 }, { "epoch": 0.26, - "learning_rate": 0.0001362, - "loss": 2.8487, + "learning_rate": 1.356e-05, + "loss": 2.8756, "step": 233 }, { "epoch": 0.26, - "learning_rate": 0.0001368, - "loss": 2.824, + "learning_rate": 1.362e-05, + "loss": 2.8453, "step": 234 }, { "epoch": 0.26, - "learning_rate": 0.0001374, - "loss": 2.8503, + "learning_rate": 1.3680000000000001e-05, + "loss": 2.852, "step": 235 }, { "epoch": 0.26, - "learning_rate": 0.000138, - "loss": 2.8399, + "learning_rate": 1.374e-05, + "loss": 2.8408, "step": 236 }, { "epoch": 0.27, - "learning_rate": 0.0001386, - "loss": 2.9885, + "learning_rate": 1.3800000000000002e-05, + "loss": 2.8637, "step": 237 }, { "epoch": 0.27, - "learning_rate": 0.0001392, - "loss": 2.9268, + "learning_rate": 1.3860000000000001e-05, + "loss": 2.8792, "step": 238 }, { "epoch": 0.27, - "learning_rate": 0.00013979999999999998, - "loss": 2.8381, + "learning_rate": 1.392e-05, + "loss": 2.8389, "step": 239 }, { "epoch": 0.27, - "learning_rate": 0.0001404, - "loss": 2.8293, + "learning_rate": 1.3980000000000002e-05, + "loss": 2.8317, "step": 240 }, { "epoch": 0.27, - "learning_rate": 0.00014099999999999998, - "loss": 2.831, + "learning_rate": 1.4040000000000001e-05, + "loss": 2.8409, "step": 241 }, { "epoch": 0.27, - "learning_rate": 0.00014159999999999997, - "loss": 2.8452, + "learning_rate": 1.4099999999999999e-05, + "loss": 2.8499, "step": 242 }, { "epoch": 0.27, - "learning_rate": 0.0001422, - "loss": 2.8426, + "learning_rate": 1.416e-05, + "loss": 2.8395, "step": 243 }, { "epoch": 0.27, - "learning_rate": 0.00014279999999999997, - "loss": 3.7653, + "learning_rate": 1.422e-05, + "loss": 2.8938, "step": 244 }, { "epoch": 0.27, - "learning_rate": 0.0001434, - "loss": 2.8247, + "learning_rate": 1.428e-05, + "loss": 2.8663, "step": 245 }, { "epoch": 0.28, - "learning_rate": 0.00014399999999999998, - "loss": 2.8207, + "learning_rate": 1.434e-05, + "loss": 2.8694, "step": 246 }, { "epoch": 0.28, - "learning_rate": 0.0001446, - "loss": 2.8202, + "learning_rate": 1.44e-05, + "loss": 2.8746, "step": 247 }, { "epoch": 0.28, - "learning_rate": 0.00014519999999999998, - "loss": 2.8326, + "learning_rate": 1.446e-05, + "loss": 2.8734, "step": 248 }, { "epoch": 0.28, - "learning_rate": 0.0001458, - "loss": 2.8299, + "learning_rate": 1.452e-05, + "loss": 2.8565, "step": 249 }, { "epoch": 0.28, - "learning_rate": 0.00014639999999999998, - "loss": 2.8964, + "learning_rate": 1.458e-05, + "loss": 2.8875, "step": 250 }, { "epoch": 0.28, - "learning_rate": 0.000147, - "loss": 2.8463, + "learning_rate": 1.464e-05, + "loss": 2.8562, "step": 251 }, { "epoch": 0.28, - "learning_rate": 0.00014759999999999998, - "loss": 2.8226, + "learning_rate": 1.47e-05, + "loss": 2.8689, "step": 252 }, { "epoch": 0.28, - "learning_rate": 0.0001482, - "loss": 2.8467, + "learning_rate": 1.4760000000000001e-05, + "loss": 2.8496, "step": 253 }, { "epoch": 0.28, - "learning_rate": 0.00014879999999999998, - "loss": 2.8479, + "learning_rate": 1.482e-05, + "loss": 2.8576, "step": 254 }, { "epoch": 0.29, - "learning_rate": 0.0001494, - "loss": 2.8608, + "learning_rate": 1.488e-05, + "loss": 2.8487, "step": 255 }, { "epoch": 0.29, - "learning_rate": 0.00015, - "loss": 2.8715, + "learning_rate": 1.4940000000000001e-05, + "loss": 2.8671, "step": 256 }, { "epoch": 0.29, - "learning_rate": 0.00015059999999999997, - "loss": 2.9296, + "learning_rate": 1.5e-05, + "loss": 2.8844, "step": 257 }, { "epoch": 0.29, - "learning_rate": 0.0001512, - "loss": 2.8995, + "learning_rate": 1.506e-05, + "loss": 2.8548, "step": 258 }, { "epoch": 0.29, - "learning_rate": 0.00015179999999999998, - "loss": 2.8288, + "learning_rate": 1.5120000000000001e-05, + "loss": 2.849, "step": 259 }, { "epoch": 0.29, - "learning_rate": 0.0001524, - "loss": 2.8411, + "learning_rate": 1.518e-05, + "loss": 2.838, "step": 260 }, { "epoch": 0.29, - "learning_rate": 0.00015299999999999998, - "loss": 2.8481, + "learning_rate": 1.524e-05, + "loss": 2.8299, "step": 261 }, { "epoch": 0.29, - "learning_rate": 0.0001536, - "loss": 2.8344, + "learning_rate": 1.53e-05, + "loss": 2.8598, "step": 262 }, { "epoch": 0.29, - "learning_rate": 0.00015419999999999998, - "loss": 2.961, + "learning_rate": 1.5360000000000002e-05, + "loss": 2.8897, "step": 263 }, { "epoch": 0.3, - "learning_rate": 0.0001548, - "loss": 2.8883, + "learning_rate": 1.542e-05, + "loss": 2.8452, "step": 264 }, { "epoch": 0.3, - "learning_rate": 0.00015539999999999998, - "loss": 2.8662, + "learning_rate": 1.548e-05, + "loss": 2.8323, "step": 265 }, { "epoch": 0.3, - "learning_rate": 0.000156, - "loss": 2.8105, + "learning_rate": 1.554e-05, + "loss": 2.8238, "step": 266 }, { "epoch": 0.3, - "learning_rate": 0.00015659999999999998, - "loss": 2.8375, + "learning_rate": 1.56e-05, + "loss": 2.8357, "step": 267 }, { "epoch": 0.3, - "learning_rate": 0.0001572, - "loss": 2.9222, + "learning_rate": 1.5660000000000003e-05, + "loss": 2.8551, "step": 268 }, { "epoch": 0.3, - "learning_rate": 0.0001578, - "loss": 3.3082, + "learning_rate": 1.5720000000000002e-05, + "loss": 2.9061, "step": 269 }, { "epoch": 0.3, - "learning_rate": 0.0001584, - "loss": 2.8263, + "learning_rate": 1.578e-05, + "loss": 2.8298, "step": 270 }, { "epoch": 0.3, - "learning_rate": 0.000159, - "loss": 2.8183, + "learning_rate": 1.584e-05, + "loss": 2.8373, "step": 271 }, { "epoch": 0.3, - "learning_rate": 0.0001596, - "loss": 2.8278, + "learning_rate": 1.59e-05, + "loss": 2.8252, "step": 272 }, { "epoch": 0.31, - "learning_rate": 0.0001602, - "loss": 2.8521, + "learning_rate": 1.596e-05, + "loss": 2.8456, "step": 273 }, { "epoch": 0.31, - "learning_rate": 0.0001608, - "loss": 2.8361, + "learning_rate": 1.6020000000000002e-05, + "loss": 2.8302, "step": 274 }, { "epoch": 0.31, - "learning_rate": 0.0001614, - "loss": 2.9221, + "learning_rate": 1.6080000000000002e-05, + "loss": 2.8762, "step": 275 }, { "epoch": 0.31, - "learning_rate": 0.000162, - "loss": 2.9084, + "learning_rate": 1.614e-05, + "loss": 2.8303, "step": 276 }, { "epoch": 0.31, - "learning_rate": 0.0001626, - "loss": 2.9352, + "learning_rate": 1.62e-05, + "loss": 2.8298, "step": 277 }, { "epoch": 0.31, - "learning_rate": 0.0001632, - "loss": 2.9647, + "learning_rate": 1.626e-05, + "loss": 2.849, "step": 278 }, { "epoch": 0.31, - "learning_rate": 0.0001638, - "loss": 2.8901, + "learning_rate": 1.6320000000000003e-05, + "loss": 2.8458, "step": 279 }, { "epoch": 0.31, - "learning_rate": 0.0001644, - "loss": 2.8285, + "learning_rate": 1.6380000000000002e-05, + "loss": 2.8335, "step": 280 }, { "epoch": 0.32, - "learning_rate": 0.000165, - "loss": 2.8505, + "learning_rate": 1.6440000000000002e-05, + "loss": 2.8573, "step": 281 }, { "epoch": 0.32, - "learning_rate": 0.0001656, - "loss": 2.9298, + "learning_rate": 1.65e-05, + "loss": 2.8805, "step": 282 }, { "epoch": 0.32, - "learning_rate": 0.0001662, - "loss": 2.8998, + "learning_rate": 1.656e-05, + "loss": 2.8313, "step": 283 }, { "epoch": 0.32, - "learning_rate": 0.0001668, - "loss": 2.8872, + "learning_rate": 1.6620000000000004e-05, + "loss": 2.8463, "step": 284 }, { "epoch": 0.32, - "learning_rate": 0.0001674, - "loss": 2.866, + "learning_rate": 1.6680000000000003e-05, + "loss": 2.8485, "step": 285 }, { "epoch": 0.32, - "learning_rate": 0.000168, - "loss": 2.8298, + "learning_rate": 1.6740000000000002e-05, + "loss": 2.8405, "step": 286 }, { "epoch": 0.32, - "learning_rate": 0.0001686, - "loss": 3.2442, + "learning_rate": 1.6800000000000002e-05, + "loss": 2.8564, "step": 287 }, { "epoch": 0.32, - "learning_rate": 0.00016919999999999997, - "loss": 2.9634, + "learning_rate": 1.686e-05, + "loss": 2.9336, "step": 288 }, { "epoch": 0.32, - "learning_rate": 0.00016979999999999998, - "loss": 3.0217, + "learning_rate": 1.6919999999999997e-05, + "loss": 2.9363, "step": 289 }, { "epoch": 0.33, - "learning_rate": 0.00017039999999999997, - "loss": 2.9834, + "learning_rate": 1.698e-05, + "loss": 2.9251, "step": 290 }, { "epoch": 0.33, - "learning_rate": 0.00017099999999999998, - "loss": 2.95, + "learning_rate": 1.704e-05, + "loss": 2.881, "step": 291 }, { "epoch": 0.33, - "learning_rate": 0.00017159999999999997, - "loss": 2.8882, + "learning_rate": 1.71e-05, + "loss": 2.8648, "step": 292 }, { "epoch": 0.33, - "learning_rate": 0.00017219999999999998, - "loss": 3.0177, + "learning_rate": 1.716e-05, + "loss": 2.8463, "step": 293 }, { "epoch": 0.33, - "learning_rate": 0.00017279999999999997, - "loss": 2.9559, + "learning_rate": 1.7219999999999998e-05, + "loss": 2.8791, "step": 294 }, { "epoch": 0.33, - "learning_rate": 0.00017339999999999996, - "loss": 2.8421, + "learning_rate": 1.728e-05, + "loss": 2.8457, "step": 295 }, { "epoch": 0.33, - "learning_rate": 0.00017399999999999997, - "loss": 2.8295, + "learning_rate": 1.734e-05, + "loss": 2.838, "step": 296 }, { "epoch": 0.33, - "learning_rate": 0.00017459999999999996, - "loss": 2.8274, + "learning_rate": 1.74e-05, + "loss": 2.8493, "step": 297 }, { "epoch": 0.33, - "learning_rate": 0.00017519999999999998, - "loss": 2.8284, + "learning_rate": 1.746e-05, + "loss": 2.8292, "step": 298 }, { "epoch": 0.34, - "learning_rate": 0.00017579999999999996, - "loss": 2.9028, + "learning_rate": 1.7519999999999998e-05, + "loss": 2.8691, "step": 299 }, { "epoch": 0.34, - "learning_rate": 0.00017639999999999998, - "loss": 2.9267, + "learning_rate": 1.758e-05, + "loss": 2.9069, "step": 300 }, { "epoch": 0.34, - "eval_loss": 3.3049137592315674, - "eval_runtime": 48.4007, - "eval_samples_per_second": 55.846, - "eval_steps_per_second": 0.888, + "eval_loss": 3.01072096824646, + "eval_runtime": 46.9646, + "eval_samples_per_second": 57.554, + "eval_steps_per_second": 0.916, "eval_wer": 1.0, "step": 300 }, { "epoch": 0.34, - "learning_rate": 0.00017699999999999997, - "loss": 2.8334, + "learning_rate": 1.764e-05, + "loss": 2.837, "step": 301 }, { "epoch": 0.34, - "learning_rate": 0.00017759999999999998, - "loss": 2.8245, + "learning_rate": 1.77e-05, + "loss": 2.8405, "step": 302 }, { "epoch": 0.34, - "learning_rate": 0.00017819999999999997, - "loss": 2.8371, + "learning_rate": 1.776e-05, + "loss": 2.8475, "step": 303 }, { "epoch": 0.34, - "learning_rate": 0.00017879999999999998, - "loss": 2.8279, + "learning_rate": 1.782e-05, + "loss": 2.8319, "step": 304 }, { "epoch": 0.34, - "learning_rate": 0.00017939999999999997, - "loss": 2.8175, + "learning_rate": 1.7879999999999998e-05, + "loss": 2.8274, "step": 305 }, { "epoch": 0.34, - "learning_rate": 0.00017999999999999998, - "loss": 2.8434, + "learning_rate": 1.794e-05, + "loss": 2.8463, "step": 306 }, { "epoch": 0.34, - "learning_rate": 0.00018059999999999997, - "loss": 3.2567, + "learning_rate": 1.8e-05, + "loss": 2.8677, "step": 307 }, { "epoch": 0.35, - "learning_rate": 0.00018119999999999999, - "loss": 2.8222, + "learning_rate": 1.806e-05, + "loss": 2.8167, "step": 308 }, { "epoch": 0.35, - "learning_rate": 0.00018179999999999997, - "loss": 2.8051, + "learning_rate": 1.812e-05, + "loss": 2.8126, "step": 309 }, { "epoch": 0.35, - "learning_rate": 0.0001824, - "loss": 2.8087, + "learning_rate": 1.818e-05, + "loss": 2.8218, "step": 310 }, { "epoch": 0.35, - "learning_rate": 0.00018299999999999998, - "loss": 2.8121, + "learning_rate": 1.824e-05, + "loss": 2.8162, "step": 311 }, { "epoch": 0.35, - "learning_rate": 0.0001836, - "loss": 2.8193, + "learning_rate": 1.83e-05, + "loss": 2.8347, "step": 312 }, { "epoch": 0.35, - "learning_rate": 0.00018419999999999998, - "loss": 3.4214, + "learning_rate": 1.836e-05, + "loss": 2.8858, "step": 313 }, { "epoch": 0.35, - "learning_rate": 0.0001848, - "loss": 2.8145, + "learning_rate": 1.842e-05, + "loss": 2.8335, "step": 314 }, { "epoch": 0.35, - "learning_rate": 0.00018539999999999998, - "loss": 2.8173, + "learning_rate": 1.848e-05, + "loss": 2.8337, "step": 315 }, { "epoch": 0.35, - "learning_rate": 0.000186, - "loss": 2.8237, + "learning_rate": 1.854e-05, + "loss": 2.8362, "step": 316 }, { "epoch": 0.36, - "learning_rate": 0.00018659999999999998, - "loss": 2.8222, + "learning_rate": 1.86e-05, + "loss": 2.8388, "step": 317 }, { "epoch": 0.36, - "learning_rate": 0.0001872, - "loss": 2.8202, + "learning_rate": 1.866e-05, + "loss": 2.8384, "step": 318 }, { "epoch": 0.36, - "learning_rate": 0.00018779999999999998, - "loss": 3.1603, + "learning_rate": 1.872e-05, + "loss": 2.9301, "step": 319 }, { "epoch": 0.36, - "learning_rate": 0.00018839999999999997, - "loss": 2.8342, + "learning_rate": 1.878e-05, + "loss": 2.8394, "step": 320 }, { "epoch": 0.36, - "learning_rate": 0.00018899999999999999, - "loss": 2.819, + "learning_rate": 1.884e-05, + "loss": 2.8264, "step": 321 }, { "epoch": 0.36, - "learning_rate": 0.00018959999999999997, - "loss": 2.8102, + "learning_rate": 1.8900000000000002e-05, + "loss": 2.8198, "step": 322 }, { "epoch": 0.36, - "learning_rate": 0.0001902, - "loss": 2.8122, + "learning_rate": 1.896e-05, + "loss": 2.8185, "step": 323 }, { "epoch": 0.36, - "learning_rate": 0.00019079999999999998, - "loss": 2.8259, + "learning_rate": 1.902e-05, + "loss": 2.8411, "step": 324 }, { "epoch": 0.36, - "learning_rate": 0.0001914, - "loss": 2.8794, + "learning_rate": 1.908e-05, + "loss": 2.8773, "step": 325 }, { "epoch": 0.37, - "learning_rate": 0.00019199999999999998, - "loss": 2.8276, + "learning_rate": 1.914e-05, + "loss": 2.8161, "step": 326 }, { "epoch": 0.37, - "learning_rate": 0.0001926, - "loss": 2.8293, + "learning_rate": 1.9200000000000003e-05, + "loss": 2.8193, "step": 327 }, { "epoch": 0.37, - "learning_rate": 0.00019319999999999998, - "loss": 2.8148, + "learning_rate": 1.9260000000000002e-05, + "loss": 2.8256, "step": 328 }, { "epoch": 0.37, - "learning_rate": 0.0001938, - "loss": 2.8369, + "learning_rate": 1.932e-05, + "loss": 2.8378, "step": 329 }, { "epoch": 0.37, - "learning_rate": 0.00019439999999999998, - "loss": 2.845, + "learning_rate": 1.938e-05, + "loss": 2.8286, "step": 330 }, { "epoch": 0.37, - "learning_rate": 0.000195, - "loss": 2.8906, + "learning_rate": 1.944e-05, + "loss": 2.8421, "step": 331 }, { "epoch": 0.37, - "learning_rate": 0.00019559999999999998, - "loss": 2.8926, + "learning_rate": 1.95e-05, + "loss": 2.8685, "step": 332 }, { "epoch": 0.37, - "learning_rate": 0.0001962, - "loss": 2.7992, + "learning_rate": 1.9560000000000002e-05, + "loss": 2.8088, "step": 333 }, { "epoch": 0.37, - "learning_rate": 0.00019679999999999999, - "loss": 2.8, + "learning_rate": 1.9620000000000002e-05, + "loss": 2.8125, "step": 334 }, { "epoch": 0.38, - "learning_rate": 0.0001974, - "loss": 2.8075, + "learning_rate": 1.968e-05, + "loss": 2.8099, "step": 335 }, { "epoch": 0.38, - "learning_rate": 0.000198, - "loss": 2.8211, + "learning_rate": 1.974e-05, + "loss": 2.8241, "step": 336 }, { "epoch": 0.38, - "learning_rate": 0.0001986, - "loss": 2.8378, + "learning_rate": 1.98e-05, + "loss": 2.8452, "step": 337 }, { "epoch": 0.38, - "learning_rate": 0.0001992, - "loss": 2.9229, + "learning_rate": 1.9860000000000003e-05, + "loss": 2.8623, "step": 338 }, { "epoch": 0.38, - "learning_rate": 0.0001998, - "loss": 2.8148, + "learning_rate": 1.9920000000000002e-05, + "loss": 2.8073, "step": 339 }, { "epoch": 0.38, - "learning_rate": 0.0002004, - "loss": 2.8336, + "learning_rate": 1.9980000000000002e-05, + "loss": 2.8263, "step": 340 }, { "epoch": 0.38, - "learning_rate": 0.000201, - "loss": 2.8382, + "learning_rate": 2.004e-05, + "loss": 2.8333, "step": 341 }, { "epoch": 0.38, - "learning_rate": 0.0002016, - "loss": 2.8089, + "learning_rate": 2.01e-05, + "loss": 2.821, "step": 342 }, { "epoch": 0.38, - "learning_rate": 0.0002022, - "loss": 2.8516, + "learning_rate": 2.016e-05, + "loss": 2.8655, "step": 343 }, { "epoch": 0.39, - "learning_rate": 0.0002028, - "loss": 2.9696, + "learning_rate": 2.0220000000000003e-05, + "loss": 3.1069, "step": 344 }, { "epoch": 0.39, - "learning_rate": 0.00020339999999999998, - "loss": 2.819, + "learning_rate": 2.0280000000000002e-05, + "loss": 2.814, "step": 345 }, { "epoch": 0.39, - "learning_rate": 0.000204, - "loss": 2.8316, + "learning_rate": 2.0340000000000002e-05, + "loss": 2.8224, "step": 346 }, { "epoch": 0.39, - "learning_rate": 0.00020459999999999999, - "loss": 2.8014, + "learning_rate": 2.04e-05, + "loss": 2.8044, "step": 347 }, { "epoch": 0.39, - "learning_rate": 0.0002052, - "loss": 2.7978, + "learning_rate": 2.046e-05, + "loss": 2.8103, "step": 348 }, { "epoch": 0.39, - "learning_rate": 0.0002058, - "loss": 2.7958, + "learning_rate": 2.0520000000000003e-05, + "loss": 2.8055, "step": 349 }, { "epoch": 0.39, - "learning_rate": 0.00020639999999999998, - "loss": 2.8616, + "learning_rate": 2.0580000000000003e-05, + "loss": 2.8575, "step": 350 }, { "epoch": 0.39, - "learning_rate": 0.00020699999999999996, - "loss": 2.7944, + "learning_rate": 2.064e-05, + "loss": 2.8075, "step": 351 }, { "epoch": 0.39, - "learning_rate": 0.00020759999999999998, - "loss": 2.8103, + "learning_rate": 2.07e-05, + "loss": 2.8186, "step": 352 }, { "epoch": 0.4, - "learning_rate": 0.00020819999999999996, - "loss": 2.8276, + "learning_rate": 2.0759999999999998e-05, + "loss": 2.8327, "step": 353 }, { "epoch": 0.4, - "learning_rate": 0.00020879999999999998, - "loss": 2.8329, + "learning_rate": 2.082e-05, + "loss": 2.839, "step": 354 }, { "epoch": 0.4, - "learning_rate": 0.00020939999999999997, - "loss": 2.8322, + "learning_rate": 2.088e-05, + "loss": 2.8366, "step": 355 }, { "epoch": 0.4, - "learning_rate": 0.00020999999999999998, - "loss": 2.8693, + "learning_rate": 2.094e-05, + "loss": 2.8463, "step": 356 }, { "epoch": 0.4, - "learning_rate": 0.00021059999999999997, - "loss": 2.9903, + "learning_rate": 2.1e-05, + "loss": 2.8815, "step": 357 }, { "epoch": 0.4, - "learning_rate": 0.00021119999999999996, - "loss": 2.8673, + "learning_rate": 2.1059999999999998e-05, + "loss": 2.8166, "step": 358 }, { "epoch": 0.4, - "learning_rate": 0.00021179999999999997, - "loss": 2.9037, + "learning_rate": 2.1119999999999998e-05, + "loss": 2.8178, "step": 359 }, { "epoch": 0.4, - "learning_rate": 0.00021239999999999996, - "loss": 2.8661, + "learning_rate": 2.118e-05, + "loss": 2.8246, "step": 360 }, { "epoch": 0.4, - "learning_rate": 0.00021299999999999997, - "loss": 2.8238, + "learning_rate": 2.124e-05, + "loss": 2.817, "step": 361 }, { "epoch": 0.41, - "learning_rate": 0.00021359999999999996, - "loss": 2.8444, + "learning_rate": 2.13e-05, + "loss": 2.8278, "step": 362 }, { "epoch": 0.41, - "learning_rate": 0.00021419999999999998, - "loss": 2.9498, + "learning_rate": 2.136e-05, + "loss": 2.871, "step": 363 }, { "epoch": 0.41, - "learning_rate": 0.00021479999999999996, - "loss": 2.8278, + "learning_rate": 2.1419999999999998e-05, + "loss": 2.8253, "step": 364 }, { "epoch": 0.41, - "learning_rate": 0.00021539999999999998, - "loss": 2.8073, + "learning_rate": 2.148e-05, + "loss": 2.8104, "step": 365 }, { "epoch": 0.41, - "learning_rate": 0.00021599999999999996, - "loss": 2.7822, + "learning_rate": 2.154e-05, + "loss": 2.7893, "step": 366 }, { "epoch": 0.41, - "learning_rate": 0.00021659999999999998, - "loss": 2.8097, + "learning_rate": 2.16e-05, + "loss": 2.8132, "step": 367 }, { "epoch": 0.41, - "learning_rate": 0.00021719999999999997, - "loss": 2.8281, + "learning_rate": 2.166e-05, + "loss": 2.8303, "step": 368 }, { "epoch": 0.41, - "learning_rate": 0.00021779999999999998, - "loss": 3.293, + "learning_rate": 2.172e-05, + "loss": 3.0313, "step": 369 }, { "epoch": 0.41, - "learning_rate": 0.00021839999999999997, - "loss": 2.9146, + "learning_rate": 2.178e-05, + "loss": 2.8162, "step": 370 }, { "epoch": 0.42, - "learning_rate": 0.00021899999999999998, - "loss": 2.9651, + "learning_rate": 2.184e-05, + "loss": 2.8166, "step": 371 }, { "epoch": 0.42, - "learning_rate": 0.00021959999999999997, - "loss": 2.9193, + "learning_rate": 2.19e-05, + "loss": 2.8189, "step": 372 }, { "epoch": 0.42, - "learning_rate": 0.00022019999999999999, - "loss": 2.8383, + "learning_rate": 2.196e-05, + "loss": 2.8241, "step": 373 }, { "epoch": 0.42, - "learning_rate": 0.00022079999999999997, - "loss": 2.8849, + "learning_rate": 2.202e-05, + "loss": 2.8208, "step": 374 }, { "epoch": 0.42, - "learning_rate": 0.0002214, - "loss": 2.9737, + "learning_rate": 2.208e-05, + "loss": 2.873, "step": 375 }, { "epoch": 0.42, - "learning_rate": 0.00022199999999999998, - "loss": 2.8657, + "learning_rate": 2.214e-05, + "loss": 2.8071, "step": 376 }, { "epoch": 0.42, - "learning_rate": 0.0002226, - "loss": 2.8098, + "learning_rate": 2.22e-05, + "loss": 2.8349, "step": 377 }, { "epoch": 0.42, - "learning_rate": 0.00022319999999999998, - "loss": 2.8999, + "learning_rate": 2.226e-05, + "loss": 2.8263, "step": 378 }, { "epoch": 0.42, - "learning_rate": 0.0002238, - "loss": 2.8585, + "learning_rate": 2.232e-05, + "loss": 2.8164, "step": 379 }, { "epoch": 0.43, - "learning_rate": 0.00022439999999999998, - "loss": 2.8549, + "learning_rate": 2.238e-05, + "loss": 2.8236, "step": 380 }, { "epoch": 0.43, - "learning_rate": 0.000225, - "loss": 2.8741, + "learning_rate": 2.2440000000000002e-05, + "loss": 2.8631, "step": 381 }, { "epoch": 0.43, - "learning_rate": 0.00022559999999999998, - "loss": 2.9457, + "learning_rate": 2.25e-05, + "loss": 2.8685, "step": 382 }, { "epoch": 0.43, - "learning_rate": 0.00022619999999999997, - "loss": 2.8229, + "learning_rate": 2.256e-05, + "loss": 2.8026, "step": 383 }, { "epoch": 0.43, - "learning_rate": 0.00022679999999999998, - "loss": 2.8535, + "learning_rate": 2.262e-05, + "loss": 2.8143, "step": 384 }, { "epoch": 0.43, - "learning_rate": 0.00022739999999999997, - "loss": 2.8305, + "learning_rate": 2.268e-05, + "loss": 2.8268, "step": 385 }, { "epoch": 0.43, - "learning_rate": 0.00022799999999999999, - "loss": 3.0675, + "learning_rate": 2.274e-05, + "loss": 2.8218, "step": 386 }, { "epoch": 0.43, - "learning_rate": 0.00022859999999999997, - "loss": 2.8456, + "learning_rate": 2.2800000000000002e-05, + "loss": 2.845, "step": 387 }, { "epoch": 0.43, - "learning_rate": 0.0002292, - "loss": 2.9505, + "learning_rate": 2.286e-05, + "loss": 2.9958, "step": 388 }, { "epoch": 0.44, - "learning_rate": 0.00022979999999999997, - "loss": 2.8488, + "learning_rate": 2.292e-05, + "loss": 2.8129, "step": 389 }, { "epoch": 0.44, - "learning_rate": 0.0002304, - "loss": 2.8644, + "learning_rate": 2.298e-05, + "loss": 2.8117, "step": 390 }, { "epoch": 0.44, - "learning_rate": 0.00023099999999999998, - "loss": 2.8272, + "learning_rate": 2.304e-05, + "loss": 2.79, "step": 391 }, { "epoch": 0.44, - "learning_rate": 0.0002316, - "loss": 2.8623, + "learning_rate": 2.3100000000000002e-05, + "loss": 2.8426, "step": 392 }, { "epoch": 0.44, - "learning_rate": 0.00023219999999999998, - "loss": 2.8532, + "learning_rate": 2.3160000000000002e-05, + "loss": 2.8221, "step": 393 }, { "epoch": 0.44, - "learning_rate": 0.0002328, - "loss": 2.9274, + "learning_rate": 2.322e-05, + "loss": 2.939, "step": 394 }, { "epoch": 0.44, - "learning_rate": 0.00023339999999999998, - "loss": 2.8394, + "learning_rate": 2.328e-05, + "loss": 2.809, "step": 395 }, { "epoch": 0.44, - "learning_rate": 0.000234, - "loss": 2.8419, + "learning_rate": 2.334e-05, + "loss": 2.8262, "step": 396 }, { "epoch": 0.45, - "learning_rate": 0.00023459999999999998, - "loss": 2.8232, + "learning_rate": 2.3400000000000003e-05, + "loss": 2.8177, "step": 397 }, { "epoch": 0.45, - "learning_rate": 0.0002352, - "loss": 2.8148, + "learning_rate": 2.3460000000000002e-05, + "loss": 2.8049, "step": 398 }, { "epoch": 0.45, - "learning_rate": 0.00023579999999999999, - "loss": 2.833, + "learning_rate": 2.3520000000000002e-05, + "loss": 2.8066, "step": 399 }, { "epoch": 0.45, - "learning_rate": 0.0002364, - "loss": 3.1351, + "learning_rate": 2.358e-05, + "loss": 2.8666, "step": 400 }, { "epoch": 0.45, - "eval_loss": 2.9246013164520264, - "eval_runtime": 48.0626, - "eval_samples_per_second": 56.239, - "eval_steps_per_second": 0.895, + "eval_loss": 2.8960044384002686, + "eval_runtime": 46.9146, + "eval_samples_per_second": 57.615, + "eval_steps_per_second": 0.917, "eval_wer": 1.0, "step": 400 }, { "epoch": 0.45, - "learning_rate": 0.000237, - "loss": 2.8196, + "learning_rate": 2.364e-05, + "loss": 2.8094, "step": 401 }, { "epoch": 0.45, - "learning_rate": 0.0002376, - "loss": 2.8058, + "learning_rate": 2.37e-05, + "loss": 2.8021, "step": 402 }, { "epoch": 0.45, - "learning_rate": 0.0002382, - "loss": 2.8241, + "learning_rate": 2.3760000000000003e-05, + "loss": 2.8031, "step": 403 }, { "epoch": 0.45, - "learning_rate": 0.0002388, - "loss": 2.844, + "learning_rate": 2.3820000000000002e-05, + "loss": 2.8188, "step": 404 }, { "epoch": 0.45, - "learning_rate": 0.0002394, - "loss": 2.8428, + "learning_rate": 2.3880000000000002e-05, + "loss": 2.8253, "step": 405 }, { "epoch": 0.46, - "learning_rate": 0.00023999999999999998, - "loss": 2.972, + "learning_rate": 2.394e-05, + "loss": 2.8382, "step": 406 }, { "epoch": 0.46, - "learning_rate": 0.0002406, - "loss": 3.8781, + "learning_rate": 2.4e-05, + "loss": 2.8423, "step": 407 }, { "epoch": 0.46, - "learning_rate": 0.00024119999999999998, - "loss": 2.8535, + "learning_rate": 2.4060000000000003e-05, + "loss": 2.8126, "step": 408 }, { "epoch": 0.46, - "learning_rate": 0.0002418, - "loss": 2.8144, + "learning_rate": 2.4120000000000003e-05, + "loss": 2.7891, "step": 409 }, { "epoch": 0.46, - "learning_rate": 0.00024239999999999998, - "loss": 2.8464, + "learning_rate": 2.4180000000000002e-05, + "loss": 2.8263, "step": 410 }, { "epoch": 0.46, - "learning_rate": 0.000243, - "loss": 2.8177, + "learning_rate": 2.4240000000000002e-05, + "loss": 2.7988, "step": 411 }, { "epoch": 0.46, - "learning_rate": 0.00024359999999999999, - "loss": 2.8823, + "learning_rate": 2.43e-05, + "loss": 2.8235, "step": 412 }, { "epoch": 0.46, - "learning_rate": 0.00024419999999999997, - "loss": 4.3712, + "learning_rate": 2.4360000000000004e-05, + "loss": 2.8708, "step": 413 }, { "epoch": 0.46, - "learning_rate": 0.0002448, - "loss": 2.8584, + "learning_rate": 2.442e-05, + "loss": 2.7999, "step": 414 }, { "epoch": 0.47, - "learning_rate": 0.00024539999999999995, - "loss": 2.8633, + "learning_rate": 2.448e-05, + "loss": 2.7887, "step": 415 }, { "epoch": 0.47, - "learning_rate": 0.00024599999999999996, - "loss": 2.8579, + "learning_rate": 2.454e-05, + "loss": 2.7979, "step": 416 }, { "epoch": 0.47, - "learning_rate": 0.0002466, - "loss": 2.8447, + "learning_rate": 2.4599999999999998e-05, + "loss": 2.8142, "step": 417 }, { "epoch": 0.47, - "learning_rate": 0.0002472, - "loss": 2.8883, + "learning_rate": 2.4659999999999998e-05, + "loss": 2.8189, "step": 418 }, { "epoch": 0.47, - "learning_rate": 0.00024779999999999995, - "loss": 4.3444, + "learning_rate": 2.472e-05, + "loss": 2.8827, "step": 419 }, { "epoch": 0.47, - "learning_rate": 0.00024839999999999997, - "loss": 2.7984, + "learning_rate": 2.478e-05, + "loss": 2.7723, "step": 420 }, { "epoch": 0.47, - "learning_rate": 0.000249, - "loss": 2.8209, + "learning_rate": 2.484e-05, + "loss": 2.7859, "step": 421 }, { "epoch": 0.47, - "learning_rate": 0.00024959999999999994, - "loss": 2.8281, + "learning_rate": 2.49e-05, + "loss": 2.7983, "step": 422 }, { "epoch": 0.47, - "learning_rate": 0.00025019999999999996, - "loss": 2.8106, + "learning_rate": 2.4959999999999998e-05, + "loss": 2.774, "step": 423 }, { "epoch": 0.48, - "learning_rate": 0.00025079999999999997, - "loss": 2.8412, + "learning_rate": 2.502e-05, + "loss": 2.8019, "step": 424 }, { "epoch": 0.48, - "learning_rate": 0.0002514, - "loss": 3.195, + "learning_rate": 2.508e-05, + "loss": 2.8276, "step": 425 }, { "epoch": 0.48, - "learning_rate": 0.00025199999999999995, - "loss": 2.8781, + "learning_rate": 2.514e-05, + "loss": 2.7691, "step": 426 }, { "epoch": 0.48, - "learning_rate": 0.00025259999999999996, - "loss": 2.8914, + "learning_rate": 2.52e-05, + "loss": 2.7737, "step": 427 }, { "epoch": 0.48, - "learning_rate": 0.0002532, - "loss": 2.8957, + "learning_rate": 2.526e-05, + "loss": 2.7527, "step": 428 }, { "epoch": 0.48, - "learning_rate": 0.0002538, - "loss": 2.8708, + "learning_rate": 2.5319999999999998e-05, + "loss": 2.7753, "step": 429 }, { "epoch": 0.48, - "learning_rate": 0.00025439999999999995, - "loss": 2.841, + "learning_rate": 2.538e-05, + "loss": 2.749, "step": 430 }, { "epoch": 0.48, - "learning_rate": 0.00025499999999999996, - "loss": 3.177, + "learning_rate": 2.544e-05, + "loss": 2.8065, "step": 431 }, { "epoch": 0.48, - "learning_rate": 0.0002556, - "loss": 4.1109, + "learning_rate": 2.55e-05, + "loss": 2.7861, "step": 432 }, { "epoch": 0.49, - "learning_rate": 0.0002562, - "loss": 2.8666, + "learning_rate": 2.556e-05, + "loss": 2.7497, "step": 433 }, { "epoch": 0.49, - "learning_rate": 0.00025679999999999995, - "loss": 2.8387, + "learning_rate": 2.562e-05, + "loss": 2.7122, "step": 434 }, { "epoch": 0.49, - "learning_rate": 0.00025739999999999997, - "loss": 2.8312, + "learning_rate": 2.568e-05, + "loss": 2.7291, "step": 435 }, { "epoch": 0.49, - "learning_rate": 0.000258, - "loss": 2.837, + "learning_rate": 2.574e-05, + "loss": 2.7313, "step": 436 }, { "epoch": 0.49, - "learning_rate": 0.0002586, - "loss": 2.8642, + "learning_rate": 2.58e-05, + "loss": 2.7357, "step": 437 }, { "epoch": 0.49, - "learning_rate": 0.00025919999999999996, - "loss": 3.7805, + "learning_rate": 2.586e-05, + "loss": 2.7855, "step": 438 }, { "epoch": 0.49, - "learning_rate": 0.00025979999999999997, - "loss": 2.8728, + "learning_rate": 2.592e-05, + "loss": 2.6667, "step": 439 }, { "epoch": 0.49, - "learning_rate": 0.0002604, - "loss": 2.8819, + "learning_rate": 2.5980000000000002e-05, + "loss": 2.6526, "step": 440 }, { "epoch": 0.49, - "learning_rate": 0.000261, - "loss": 2.888, + "learning_rate": 2.604e-05, + "loss": 2.6484, "step": 441 }, { "epoch": 0.5, - "learning_rate": 0.00026159999999999996, - "loss": 2.8765, + "learning_rate": 2.61e-05, + "loss": 2.6548, "step": 442 }, { "epoch": 0.5, - "learning_rate": 0.0002622, - "loss": 2.8623, + "learning_rate": 2.616e-05, + "loss": 2.6448, "step": 443 }, { "epoch": 0.5, - "learning_rate": 0.0002628, - "loss": 4.2702, + "learning_rate": 2.622e-05, + "loss": 2.7107, "step": 444 }, { "epoch": 0.5, - "learning_rate": 0.00026339999999999995, - "loss": 2.8202, + "learning_rate": 2.628e-05, + "loss": 2.6078, "step": 445 }, { "epoch": 0.5, - "learning_rate": 0.00026399999999999997, - "loss": 2.8055, + "learning_rate": 2.6340000000000002e-05, + "loss": 2.5855, "step": 446 }, { "epoch": 0.5, - "learning_rate": 0.0002646, - "loss": 2.825, + "learning_rate": 2.64e-05, + "loss": 2.5755, "step": 447 }, { "epoch": 0.5, - "learning_rate": 0.0002652, - "loss": 2.8291, + "learning_rate": 2.646e-05, + "loss": 2.5595, "step": 448 }, { "epoch": 0.5, - "learning_rate": 0.00026579999999999996, - "loss": 2.842, + "learning_rate": 2.652e-05, + "loss": 2.5494, "step": 449 }, { "epoch": 0.5, - "learning_rate": 0.00026639999999999997, - "loss": 3.1618, + "learning_rate": 2.658e-05, + "loss": 2.6012, "step": 450 }, { "epoch": 0.51, - "learning_rate": 0.000267, - "loss": 2.8592, + "learning_rate": 2.6640000000000002e-05, + "loss": 2.5592, "step": 451 }, { "epoch": 0.51, - "learning_rate": 0.0002676, - "loss": 2.8664, + "learning_rate": 2.6700000000000002e-05, + "loss": 2.5106, "step": 452 }, { "epoch": 0.51, - "learning_rate": 0.00026819999999999996, - "loss": 2.8623, + "learning_rate": 2.676e-05, + "loss": 2.4966, "step": 453 }, { "epoch": 0.51, - "learning_rate": 0.0002688, - "loss": 2.8454, + "learning_rate": 2.682e-05, + "loss": 2.4648, "step": 454 }, { "epoch": 0.51, - "learning_rate": 0.0002694, - "loss": 2.8344, + "learning_rate": 2.688e-05, + "loss": 2.4497, "step": 455 }, { "epoch": 0.51, - "learning_rate": 0.00027, - "loss": 2.9286, + "learning_rate": 2.6940000000000003e-05, + "loss": 2.4467, "step": 456 }, { "epoch": 0.51, - "learning_rate": 0.00027059999999999996, - "loss": 3.9234, + "learning_rate": 2.7000000000000002e-05, + "loss": 2.506, "step": 457 }, { "epoch": 0.51, - "learning_rate": 0.0002712, - "loss": 2.844, + "learning_rate": 2.7060000000000002e-05, + "loss": 2.4349, "step": 458 }, { "epoch": 0.51, - "learning_rate": 0.0002718, - "loss": 2.8245, + "learning_rate": 2.712e-05, + "loss": 2.3871, "step": 459 }, { "epoch": 0.52, - "learning_rate": 0.0002724, - "loss": 2.8309, + "learning_rate": 2.718e-05, + "loss": 2.3904, "step": 460 }, { "epoch": 0.52, - "learning_rate": 0.00027299999999999997, - "loss": 2.826, + "learning_rate": 2.724e-05, + "loss": 2.3374, "step": 461 }, { "epoch": 0.52, - "learning_rate": 0.0002736, - "loss": 2.8548, + "learning_rate": 2.7300000000000003e-05, + "loss": 2.3567, "step": 462 }, { "epoch": 0.52, - "learning_rate": 0.0002742, - "loss": 3.6627, + "learning_rate": 2.7360000000000002e-05, + "loss": 2.4863, "step": 463 }, { "epoch": 0.52, - "learning_rate": 0.0002748, - "loss": 2.859, + "learning_rate": 2.7420000000000002e-05, + "loss": 2.3544, "step": 464 }, { "epoch": 0.52, - "learning_rate": 0.00027539999999999997, - "loss": 2.8703, + "learning_rate": 2.748e-05, + "loss": 2.3528, "step": 465 }, { "epoch": 0.52, - "learning_rate": 0.000276, - "loss": 2.843, + "learning_rate": 2.754e-05, + "loss": 2.2629, "step": 466 }, { "epoch": 0.52, - "learning_rate": 0.0002766, - "loss": 2.8191, + "learning_rate": 2.7600000000000003e-05, + "loss": 2.2426, "step": 467 }, { "epoch": 0.52, - "learning_rate": 0.0002772, - "loss": 2.8468, + "learning_rate": 2.7660000000000003e-05, + "loss": 2.2433, "step": 468 }, { "epoch": 0.53, - "learning_rate": 0.0002778, - "loss": 4.8544, + "learning_rate": 2.7720000000000002e-05, + "loss": 2.539, "step": 469 }, { "epoch": 0.53, - "learning_rate": 0.0002784, - "loss": 2.8308, + "learning_rate": 2.778e-05, + "loss": 2.177, "step": 470 }, { "epoch": 0.53, - "learning_rate": 0.000279, - "loss": 2.8354, + "learning_rate": 2.784e-05, + "loss": 2.1367, "step": 471 }, { "epoch": 0.53, - "learning_rate": 0.00027959999999999997, - "loss": 2.8151, + "learning_rate": 2.79e-05, + "loss": 2.1351, "step": 472 }, { "epoch": 0.53, - "learning_rate": 0.0002802, - "loss": 2.8449, + "learning_rate": 2.7960000000000003e-05, + "loss": 2.145, "step": 473 }, { "epoch": 0.53, - "learning_rate": 0.0002808, - "loss": 2.8305, + "learning_rate": 2.8020000000000003e-05, + "loss": 2.0954, "step": 474 }, { "epoch": 0.53, - "learning_rate": 0.00028139999999999996, - "loss": 3.0622, + "learning_rate": 2.8080000000000002e-05, + "loss": 2.1236, "step": 475 }, { "epoch": 0.53, - "learning_rate": 0.00028199999999999997, - "loss": 2.8426, + "learning_rate": 2.8139999999999998e-05, + "loss": 2.0367, "step": 476 }, { "epoch": 0.53, - "learning_rate": 0.0002826, - "loss": 2.8197, + "learning_rate": 2.8199999999999998e-05, + "loss": 1.9793, "step": 477 }, { "epoch": 0.54, - "learning_rate": 0.00028319999999999994, - "loss": 2.82, + "learning_rate": 2.826e-05, + "loss": 1.9567, "step": 478 }, { "epoch": 0.54, - "learning_rate": 0.00028379999999999996, - "loss": 2.827, + "learning_rate": 2.832e-05, + "loss": 1.9561, "step": 479 }, { "epoch": 0.54, - "learning_rate": 0.0002844, - "loss": 2.8159, + "learning_rate": 2.838e-05, + "loss": 1.9105, "step": 480 }, { "epoch": 0.54, - "learning_rate": 0.000285, - "loss": 3.0608, + "learning_rate": 2.844e-05, + "loss": 1.9232, "step": 481 }, { "epoch": 0.54, - "learning_rate": 0.00028559999999999995, - "loss": 4.047, + "learning_rate": 2.8499999999999998e-05, + "loss": 2.0018, "step": 482 }, { "epoch": 0.54, - "learning_rate": 0.00028619999999999996, - "loss": 2.8332, + "learning_rate": 2.856e-05, + "loss": 1.8231, "step": 483 }, { "epoch": 0.54, - "learning_rate": 0.0002868, - "loss": 2.8641, + "learning_rate": 2.862e-05, + "loss": 1.7783, "step": 484 }, { "epoch": 0.54, - "learning_rate": 0.00028739999999999994, - "loss": 2.8633, + "learning_rate": 2.868e-05, + "loss": 1.752, "step": 485 }, { "epoch": 0.54, - "learning_rate": 0.00028799999999999995, - "loss": 2.8655, + "learning_rate": 2.874e-05, + "loss": 1.7328, "step": 486 }, { "epoch": 0.55, - "learning_rate": 0.00028859999999999997, - "loss": 2.8786, + "learning_rate": 2.88e-05, + "loss": 1.7199, "step": 487 }, { "epoch": 0.55, - "learning_rate": 0.0002892, - "loss": 4.3636, + "learning_rate": 2.8859999999999998e-05, + "loss": 1.8317, "step": 488 }, { "epoch": 0.55, - "learning_rate": 0.00028979999999999994, - "loss": 2.8189, + "learning_rate": 2.892e-05, + "loss": 1.6713, "step": 489 }, { "epoch": 0.55, - "learning_rate": 0.00029039999999999996, - "loss": 2.8267, + "learning_rate": 2.898e-05, + "loss": 1.6053, "step": 490 }, { "epoch": 0.55, - "learning_rate": 0.00029099999999999997, - "loss": 2.8121, + "learning_rate": 2.904e-05, + "loss": 1.5713, "step": 491 }, { "epoch": 0.55, - "learning_rate": 0.0002916, - "loss": 2.8202, + "learning_rate": 2.91e-05, + "loss": 1.5445, "step": 492 }, { "epoch": 0.55, - "learning_rate": 0.00029219999999999995, - "loss": 2.8691, + "learning_rate": 2.916e-05, + "loss": 1.5447, "step": 493 }, { "epoch": 0.55, - "learning_rate": 0.00029279999999999996, - "loss": 4.8131, + "learning_rate": 2.922e-05, + "loss": 1.8523, "step": 494 }, { "epoch": 0.55, - "learning_rate": 0.0002934, - "loss": 2.8204, + "learning_rate": 2.928e-05, + "loss": 1.4661, "step": 495 }, { "epoch": 0.56, - "learning_rate": 0.000294, - "loss": 2.8324, + "learning_rate": 2.934e-05, + "loss": 1.419, "step": 496 }, { "epoch": 0.56, - "learning_rate": 0.00029459999999999995, - "loss": 2.8223, + "learning_rate": 2.94e-05, + "loss": 1.4155, "step": 497 }, { "epoch": 0.56, - "learning_rate": 0.00029519999999999997, - "loss": 2.8245, + "learning_rate": 2.946e-05, + "loss": 1.3735, "step": 498 }, { "epoch": 0.56, - "learning_rate": 0.0002958, - "loss": 2.8275, + "learning_rate": 2.9520000000000002e-05, + "loss": 1.3335, "step": 499 }, { "epoch": 0.56, - "learning_rate": 0.0002964, - "loss": 3.4365, + "learning_rate": 2.958e-05, + "loss": 1.4535, "step": 500 }, { "epoch": 0.56, - "eval_loss": 4.279839515686035, - "eval_runtime": 48.1123, - "eval_samples_per_second": 56.181, - "eval_steps_per_second": 0.894, - "eval_wer": 1.0, + "eval_loss": 1.4062193632125854, + "eval_runtime": 47.768, + "eval_samples_per_second": 56.586, + "eval_steps_per_second": 0.9, + "eval_wer": 0.8664387338700783, "step": 500 }, { "epoch": 0.56, - "learning_rate": 0.00029699999999999996, - "loss": 2.8231, + "learning_rate": 2.964e-05, + "loss": 1.2605, "step": 501 }, { "epoch": 0.56, - "learning_rate": 0.00029759999999999997, - "loss": 2.8342, + "learning_rate": 2.97e-05, + "loss": 1.2826, "step": 502 }, { "epoch": 0.56, - "learning_rate": 0.0002982, - "loss": 2.8121, + "learning_rate": 2.976e-05, + "loss": 1.1953, "step": 503 }, { "epoch": 0.57, - "learning_rate": 0.0002988, - "loss": 2.8413, + "learning_rate": 2.982e-05, + "loss": 1.2011, "step": 504 }, { "epoch": 0.57, - "learning_rate": 0.00029939999999999996, - "loss": 2.8457, + "learning_rate": 2.9880000000000002e-05, + "loss": 1.2051, "step": 505 }, { "epoch": 0.57, - "learning_rate": 0.0003, - "loss": 3.0245, + "learning_rate": 2.994e-05, + "loss": 1.2353, "step": 506 }, { "epoch": 0.57, - "learning_rate": 0.00029986213235294113, - "loss": 3.6726, + "learning_rate": 3e-05, + "loss": 1.2317, "step": 507 }, { "epoch": 0.57, - "learning_rate": 0.0002997242647058823, - "loss": 2.8574, + "learning_rate": 2.9986213235294118e-05, + "loss": 1.0997, "step": 508 }, { "epoch": 0.57, - "learning_rate": 0.0002995863970588235, - "loss": 2.8785, + "learning_rate": 2.9972426470588235e-05, + "loss": 1.1052, "step": 509 }, { "epoch": 0.57, - "learning_rate": 0.0002994485294117647, - "loss": 2.8836, + "learning_rate": 2.9958639705882355e-05, + "loss": 1.0738, "step": 510 }, { "epoch": 0.57, - "learning_rate": 0.00029931066176470587, - "loss": 2.8391, + "learning_rate": 2.9944852941176472e-05, + "loss": 1.0767, "step": 511 }, { "epoch": 0.57, - "learning_rate": 0.000299172794117647, - "loss": 2.8773, + "learning_rate": 2.993106617647059e-05, + "loss": 1.0856, "step": 512 }, { "epoch": 0.58, - "learning_rate": 0.00029903492647058823, - "loss": 3.624, + "learning_rate": 2.9917279411764706e-05, + "loss": 1.1641, "step": 513 }, { "epoch": 0.58, - "learning_rate": 0.0002988970588235294, - "loss": 2.8199, + "learning_rate": 2.9903492647058827e-05, + "loss": 1.0905, "step": 514 }, { "epoch": 0.58, - "learning_rate": 0.00029875919117647055, - "loss": 2.8148, + "learning_rate": 2.9889705882352944e-05, + "loss": 1.0416, "step": 515 }, { "epoch": 0.58, - "learning_rate": 0.0002986213235294117, - "loss": 2.8157, + "learning_rate": 2.987591911764706e-05, + "loss": 0.9573, "step": 516 }, { "epoch": 0.58, - "learning_rate": 0.0002984834558823529, - "loss": 2.8068, + "learning_rate": 2.9862132352941178e-05, + "loss": 0.9841, "step": 517 }, { "epoch": 0.58, - "learning_rate": 0.0002983455882352941, - "loss": 2.8241, + "learning_rate": 2.9848345588235295e-05, + "loss": 1.0002, "step": 518 }, { "epoch": 0.58, - "learning_rate": 0.0002982077205882353, - "loss": 4.323, + "learning_rate": 2.983455882352941e-05, + "loss": 1.0605, "step": 519 }, { "epoch": 0.58, - "learning_rate": 0.00029806985294117644, - "loss": 2.8387, + "learning_rate": 2.982077205882353e-05, + "loss": 0.9282, "step": 520 }, { "epoch": 0.58, - "learning_rate": 0.00029793198529411765, - "loss": 2.8282, + "learning_rate": 2.9806985294117646e-05, + "loss": 0.9483, "step": 521 }, { "epoch": 0.59, - "learning_rate": 0.0002977941176470588, - "loss": 2.8378, + "learning_rate": 2.9793198529411766e-05, + "loss": 0.8976, "step": 522 }, { "epoch": 0.59, - "learning_rate": 0.00029765624999999996, - "loss": 2.826, + "learning_rate": 2.9779411764705883e-05, + "loss": 0.8773, "step": 523 }, { "epoch": 0.59, - "learning_rate": 0.0002975183823529411, - "loss": 2.8493, + "learning_rate": 2.9765625e-05, + "loss": 0.9194, "step": 524 }, { "epoch": 0.59, - "learning_rate": 0.00029738051470588233, - "loss": 3.2449, + "learning_rate": 2.9751838235294117e-05, + "loss": 0.9902, "step": 525 }, { "epoch": 0.59, - "learning_rate": 0.00029724264705882354, - "loss": 2.8175, + "learning_rate": 2.9738051470588234e-05, + "loss": 0.8459, "step": 526 }, { "epoch": 0.59, - "learning_rate": 0.0002971047794117647, - "loss": 2.8211, + "learning_rate": 2.9724264705882355e-05, + "loss": 0.8455, "step": 527 }, { "epoch": 0.59, - "learning_rate": 0.00029696691176470585, - "loss": 2.8162, + "learning_rate": 2.971047794117647e-05, + "loss": 0.8104, "step": 528 }, { "epoch": 0.59, - "learning_rate": 0.000296829044117647, - "loss": 2.8194, + "learning_rate": 2.969669117647059e-05, + "loss": 0.855, "step": 529 }, { "epoch": 0.59, - "learning_rate": 0.0002966911764705882, - "loss": 2.8505, + "learning_rate": 2.9682904411764706e-05, + "loss": 0.8748, "step": 530 }, { "epoch": 0.6, - "learning_rate": 0.0002965533088235294, - "loss": 2.9977, + "learning_rate": 2.9669117647058826e-05, + "loss": 0.8495, "step": 531 }, { "epoch": 0.6, - "learning_rate": 0.00029641544117647053, - "loss": 3.8876, + "learning_rate": 2.9655330882352943e-05, + "loss": 0.8263, "step": 532 }, { "epoch": 0.6, - "learning_rate": 0.00029627757352941174, - "loss": 2.8168, + "learning_rate": 2.964154411764706e-05, + "loss": 0.7833, "step": 533 }, { "epoch": 0.6, - "learning_rate": 0.00029613970588235295, - "loss": 2.8217, + "learning_rate": 2.9627757352941177e-05, + "loss": 0.7775, "step": 534 }, { "epoch": 0.6, - "learning_rate": 0.0002960018382352941, - "loss": 2.8192, + "learning_rate": 2.9613970588235297e-05, + "loss": 0.713, "step": 535 }, { "epoch": 0.6, - "learning_rate": 0.00029586397058823527, - "loss": 2.8152, + "learning_rate": 2.9600183823529414e-05, + "loss": 0.7537, "step": 536 }, { "epoch": 0.6, - "learning_rate": 0.0002957261029411764, - "loss": 2.9017, + "learning_rate": 2.958639705882353e-05, + "loss": 0.7946, "step": 537 }, { "epoch": 0.6, - "learning_rate": 0.00029558823529411763, - "loss": 4.2157, + "learning_rate": 2.957261029411765e-05, + "loss": 0.809, "step": 538 }, { "epoch": 0.6, - "learning_rate": 0.0002954503676470588, - "loss": 2.8329, + "learning_rate": 2.9558823529411766e-05, + "loss": 0.7656, "step": 539 }, { "epoch": 0.61, - "learning_rate": 0.00029531249999999995, - "loss": 2.8295, + "learning_rate": 2.9545036764705883e-05, + "loss": 0.7569, "step": 540 }, { "epoch": 0.61, - "learning_rate": 0.00029517463235294116, - "loss": 2.815, + "learning_rate": 2.953125e-05, + "loss": 0.734, "step": 541 }, { "epoch": 0.61, - "learning_rate": 0.0002950367647058823, - "loss": 2.8183, + "learning_rate": 2.9517463235294117e-05, + "loss": 0.7277, "step": 542 }, { "epoch": 0.61, - "learning_rate": 0.0002948988970588235, - "loss": 2.8543, + "learning_rate": 2.9503676470588234e-05, + "loss": 0.727, "step": 543 }, { "epoch": 0.61, - "learning_rate": 0.0002947610294117647, - "loss": 4.4142, + "learning_rate": 2.9489889705882354e-05, + "loss": 0.8248, "step": 544 }, { "epoch": 0.61, - "learning_rate": 0.00029462316176470584, - "loss": 2.8215, + "learning_rate": 2.947610294117647e-05, + "loss": 0.7076, "step": 545 }, { "epoch": 0.61, - "learning_rate": 0.000294485294117647, - "loss": 2.8207, + "learning_rate": 2.9462316176470588e-05, + "loss": 0.6686, "step": 546 }, { "epoch": 0.61, - "learning_rate": 0.0002943474264705882, - "loss": 2.8122, + "learning_rate": 2.9448529411764705e-05, + "loss": 0.6483, "step": 547 }, { "epoch": 0.61, - "learning_rate": 0.00029420955882352936, - "loss": 2.8012, + "learning_rate": 2.9434742647058825e-05, + "loss": 0.6448, "step": 548 }, { "epoch": 0.62, - "learning_rate": 0.0002940716911764706, - "loss": 2.851, + "learning_rate": 2.9420955882352942e-05, + "loss": 0.6437, "step": 549 }, { "epoch": 0.62, - "learning_rate": 0.00029393382352941173, - "loss": 3.2928, + "learning_rate": 2.940716911764706e-05, + "loss": 0.7524, "step": 550 }, { "epoch": 0.62, - "learning_rate": 0.00029379595588235294, - "loss": 2.8249, + "learning_rate": 2.9393382352941176e-05, + "loss": 0.6681, "step": 551 }, { "epoch": 0.62, - "learning_rate": 0.0002936580882352941, - "loss": 2.8, + "learning_rate": 2.9379595588235297e-05, + "loss": 0.6339, "step": 552 }, { "epoch": 0.62, - "learning_rate": 0.00029352022058823525, - "loss": 2.8158, + "learning_rate": 2.9365808823529414e-05, + "loss": 0.6611, "step": 553 }, { "epoch": 0.62, - "learning_rate": 0.0002933823529411764, - "loss": 2.8129, + "learning_rate": 2.935202205882353e-05, + "loss": 0.6574, "step": 554 }, { "epoch": 0.62, - "learning_rate": 0.0002932444852941176, - "loss": 2.826, + "learning_rate": 2.9338235294117648e-05, + "loss": 0.6234, "step": 555 }, { "epoch": 0.62, - "learning_rate": 0.00029310661764705883, - "loss": 2.977, + "learning_rate": 2.9324448529411768e-05, + "loss": 0.6944, "step": 556 }, { "epoch": 0.62, - "learning_rate": 0.00029296875, - "loss": 3.6756, + "learning_rate": 2.9310661764705885e-05, + "loss": 0.6815, "step": 557 }, { "epoch": 0.63, - "learning_rate": 0.00029283088235294115, - "loss": 2.843, + "learning_rate": 2.9296875000000002e-05, + "loss": 0.616, "step": 558 }, { "epoch": 0.63, - "learning_rate": 0.0002926930147058823, - "loss": 2.8824, + "learning_rate": 2.9283088235294116e-05, + "loss": 0.6228, "step": 559 }, { "epoch": 0.63, - "learning_rate": 0.0002925551470588235, - "loss": 2.8727, + "learning_rate": 2.9269301470588233e-05, + "loss": 0.6144, "step": 560 }, { "epoch": 0.63, - "learning_rate": 0.00029241727941176467, - "loss": 2.8611, + "learning_rate": 2.9255514705882353e-05, + "loss": 0.598, "step": 561 }, { "epoch": 0.63, - "learning_rate": 0.0002922794117647058, - "loss": 2.9041, + "learning_rate": 2.924172794117647e-05, + "loss": 0.6714, "step": 562 }, { "epoch": 0.63, - "learning_rate": 0.00029214154411764704, - "loss": 4.0101, + "learning_rate": 2.9227941176470587e-05, + "loss": 0.7081, "step": 563 }, { "epoch": 0.63, - "learning_rate": 0.00029200367647058825, - "loss": 2.8464, + "learning_rate": 2.9214154411764704e-05, + "loss": 0.6172, "step": 564 }, { "epoch": 0.63, - "learning_rate": 0.0002918658088235294, - "loss": 2.8281, + "learning_rate": 2.9200367647058825e-05, + "loss": 0.6027, "step": 565 }, { "epoch": 0.63, - "learning_rate": 0.00029172794117647056, - "loss": 2.8163, + "learning_rate": 2.9186580882352942e-05, + "loss": 0.6181, "step": 566 }, { "epoch": 0.64, - "learning_rate": 0.0002915900735294117, - "loss": 2.8316, + "learning_rate": 2.917279411764706e-05, + "loss": 0.6028, "step": 567 }, { "epoch": 0.64, - "learning_rate": 0.00029145220588235293, - "loss": 2.8989, + "learning_rate": 2.9159007352941176e-05, + "loss": 0.6063, "step": 568 }, { "epoch": 0.64, - "learning_rate": 0.0002913143382352941, - "loss": 4.4138, + "learning_rate": 2.9145220588235296e-05, + "loss": 0.6727, "step": 569 }, { "epoch": 0.64, - "learning_rate": 0.00029117647058823524, - "loss": 2.8334, + "learning_rate": 2.9131433823529413e-05, + "loss": 0.5491, "step": 570 }, { "epoch": 0.64, - "learning_rate": 0.00029103860294117645, - "loss": 2.8072, + "learning_rate": 2.911764705882353e-05, + "loss": 0.5514, "step": 571 }, { "epoch": 0.64, - "learning_rate": 0.00029090073529411766, - "loss": 2.8041, + "learning_rate": 2.9103860294117647e-05, + "loss": 0.5414, "step": 572 }, { "epoch": 0.64, - "learning_rate": 0.0002907628676470588, - "loss": 2.8239, + "learning_rate": 2.9090073529411768e-05, + "loss": 0.5573, "step": 573 }, { "epoch": 0.64, - "learning_rate": 0.000290625, - "loss": 2.8314, + "learning_rate": 2.9076286764705885e-05, + "loss": 0.5387, "step": 574 }, { "epoch": 0.64, - "learning_rate": 0.00029048713235294113, - "loss": 3.2916, + "learning_rate": 2.90625e-05, + "loss": 0.6273, "step": 575 }, { "epoch": 0.65, - "learning_rate": 0.0002903492647058823, - "loss": 2.8419, + "learning_rate": 2.904871323529412e-05, + "loss": 0.5639, "step": 576 }, { "epoch": 0.65, - "learning_rate": 0.0002902113970588235, - "loss": 2.8398, + "learning_rate": 2.9034926470588236e-05, + "loss": 0.5556, "step": 577 }, { "epoch": 0.65, - "learning_rate": 0.00029007352941176466, - "loss": 2.8428, + "learning_rate": 2.9021139705882356e-05, + "loss": 0.5785, "step": 578 }, { "epoch": 0.65, - "learning_rate": 0.00028993566176470587, - "loss": 2.8196, + "learning_rate": 2.9007352941176473e-05, + "loss": 0.5122, "step": 579 }, { "epoch": 0.65, - "learning_rate": 0.000289797794117647, - "loss": 2.8333, + "learning_rate": 2.8993566176470587e-05, + "loss": 0.5431, "step": 580 }, { "epoch": 0.65, - "learning_rate": 0.00028965992647058823, - "loss": 3.0719, + "learning_rate": 2.8979779411764704e-05, + "loss": 0.5917, "step": 581 }, { "epoch": 0.65, - "learning_rate": 0.0002895220588235294, - "loss": 3.5945, + "learning_rate": 2.8965992647058824e-05, + "loss": 0.6315, "step": 582 }, { "epoch": 0.65, - "learning_rate": 0.00028938419117647055, - "loss": 2.8249, + "learning_rate": 2.895220588235294e-05, + "loss": 0.5509, "step": 583 }, { "epoch": 0.65, - "learning_rate": 0.0002892463235294117, - "loss": 2.8315, + "learning_rate": 2.8938419117647058e-05, + "loss": 0.4966, "step": 584 }, { "epoch": 0.66, - "learning_rate": 0.0002891084558823529, - "loss": 2.8437, + "learning_rate": 2.8924632352941175e-05, + "loss": 0.5453, "step": 585 }, { "epoch": 0.66, - "learning_rate": 0.00028897058823529407, - "loss": 2.8207, + "learning_rate": 2.8910845588235296e-05, + "loss": 0.5631, "step": 586 }, { "epoch": 0.66, - "learning_rate": 0.0002888327205882353, - "loss": 2.9229, + "learning_rate": 2.8897058823529413e-05, + "loss": 0.5115, "step": 587 }, { "epoch": 0.66, - "learning_rate": 0.00028869485294117644, - "loss": 4.381, + "learning_rate": 2.888327205882353e-05, + "loss": 0.5831, "step": 588 }, { "epoch": 0.66, - "learning_rate": 0.00028855698529411765, - "loss": 2.8163, + "learning_rate": 2.8869485294117647e-05, + "loss": 0.5101, "step": 589 }, { "epoch": 0.66, - "learning_rate": 0.0002884191176470588, - "loss": 2.8182, + "learning_rate": 2.8855698529411767e-05, + "loss": 0.52, "step": 590 }, { "epoch": 0.66, - "learning_rate": 0.00028828124999999996, - "loss": 2.8046, + "learning_rate": 2.8841911764705884e-05, + "loss": 0.5265, "step": 591 }, { "epoch": 0.66, - "learning_rate": 0.0002881433823529411, - "loss": 2.8259, + "learning_rate": 2.8828125e-05, + "loss": 0.5385, "step": 592 }, { "epoch": 0.66, - "learning_rate": 0.00028800551470588233, - "loss": 2.843, + "learning_rate": 2.8814338235294118e-05, + "loss": 0.4868, "step": 593 }, { "epoch": 0.67, - "learning_rate": 0.0002878676470588235, - "loss": 4.3121, + "learning_rate": 2.8800551470588235e-05, + "loss": 0.6107, "step": 594 }, { "epoch": 0.67, - "learning_rate": 0.0002877297794117647, - "loss": 2.8182, + "learning_rate": 2.8786764705882356e-05, + "loss": 0.4818, "step": 595 }, { "epoch": 0.67, - "learning_rate": 0.00028759191176470585, - "loss": 2.8525, + "learning_rate": 2.8772977941176473e-05, + "loss": 0.4721, "step": 596 }, { "epoch": 0.67, - "learning_rate": 0.000287454044117647, - "loss": 2.8452, + "learning_rate": 2.875919117647059e-05, + "loss": 0.5144, "step": 597 }, { "epoch": 0.67, - "learning_rate": 0.0002873161764705882, - "loss": 2.833, + "learning_rate": 2.8745404411764707e-05, + "loss": 0.4871, "step": 598 }, { "epoch": 0.67, - "learning_rate": 0.0002871783088235294, - "loss": 2.8806, + "learning_rate": 2.8731617647058827e-05, + "loss": 0.5195, "step": 599 }, { "epoch": 0.67, - "learning_rate": 0.00028704044117647054, - "loss": 3.1861, + "learning_rate": 2.8717830882352944e-05, + "loss": 0.6821, "step": 600 }, { "epoch": 0.67, - "eval_loss": 4.074018478393555, - "eval_runtime": 48.1611, - "eval_samples_per_second": 56.124, - "eval_steps_per_second": 0.893, - "eval_wer": 1.0, + "eval_loss": 0.5529720783233643, + "eval_runtime": 47.278, + "eval_samples_per_second": 57.172, + "eval_steps_per_second": 0.91, + "eval_wer": 0.492978199330907, "step": 600 }, { "epoch": 0.67, - "learning_rate": 0.00028690257352941175, - "loss": 2.8081, + "learning_rate": 2.8704044117647058e-05, + "loss": 0.4821, "step": 601 }, { "epoch": 0.67, - "learning_rate": 0.00028676470588235296, - "loss": 2.8004, + "learning_rate": 2.8690257352941175e-05, + "loss": 0.4726, "step": 602 }, { "epoch": 0.68, - "learning_rate": 0.0002866268382352941, - "loss": 2.8188, + "learning_rate": 2.8676470588235295e-05, + "loss": 0.471, "step": 603 }, { "epoch": 0.68, - "learning_rate": 0.00028648897058823527, - "loss": 2.8281, + "learning_rate": 2.8662683823529412e-05, + "loss": 0.4906, "step": 604 }, { "epoch": 0.68, - "learning_rate": 0.0002863511029411764, - "loss": 2.8099, + "learning_rate": 2.864889705882353e-05, + "loss": 0.5093, "step": 605 }, { "epoch": 0.68, - "learning_rate": 0.00028621323529411764, - "loss": 3.0387, + "learning_rate": 2.8635110294117646e-05, + "loss": 0.5461, "step": 606 }, { "epoch": 0.68, - "learning_rate": 0.0002860753676470588, - "loss": 3.5862, + "learning_rate": 2.8621323529411766e-05, + "loss": 0.5271, "step": 607 }, { "epoch": 0.68, - "learning_rate": 0.00028593749999999995, - "loss": 2.8218, + "learning_rate": 2.8607536764705883e-05, + "loss": 0.4817, "step": 608 }, { "epoch": 0.68, - "learning_rate": 0.00028579963235294116, - "loss": 2.8394, + "learning_rate": 2.859375e-05, + "loss": 0.4552, "step": 609 }, { "epoch": 0.68, - "learning_rate": 0.0002856617647058823, - "loss": 2.8202, + "learning_rate": 2.8579963235294117e-05, + "loss": 0.4357, "step": 610 }, { "epoch": 0.68, - "learning_rate": 0.00028552389705882353, - "loss": 2.8192, + "learning_rate": 2.8566176470588234e-05, + "loss": 0.4859, "step": 611 }, { "epoch": 0.69, - "learning_rate": 0.0002853860294117647, - "loss": 2.8476, + "learning_rate": 2.8552389705882355e-05, + "loss": 0.4829, "step": 612 }, { "epoch": 0.69, - "learning_rate": 0.00028524816176470584, - "loss": 3.482, + "learning_rate": 2.8538602941176472e-05, + "loss": 0.5268, "step": 613 }, { "epoch": 0.69, - "learning_rate": 0.000285110294117647, - "loss": 2.8299, + "learning_rate": 2.852481617647059e-05, + "loss": 0.4924, "step": 614 }, { "epoch": 0.69, - "learning_rate": 0.0002849724264705882, - "loss": 2.8233, + "learning_rate": 2.8511029411764706e-05, + "loss": 0.4551, "step": 615 }, { "epoch": 0.69, - "learning_rate": 0.00028483455882352937, - "loss": 2.8255, + "learning_rate": 2.8497242647058826e-05, + "loss": 0.4705, "step": 616 }, { "epoch": 0.69, - "learning_rate": 0.0002846966911764706, - "loss": 2.8274, + "learning_rate": 2.8483455882352943e-05, + "loss": 0.4634, "step": 617 }, { "epoch": 0.69, - "learning_rate": 0.00028455882352941173, - "loss": 2.8287, + "learning_rate": 2.846966911764706e-05, + "loss": 0.4862, "step": 618 }, { "epoch": 0.69, - "learning_rate": 0.00028442095588235294, - "loss": 4.3124, + "learning_rate": 2.8455882352941177e-05, + "loss": 0.5746, "step": 619 }, { "epoch": 0.7, - "learning_rate": 0.0002842830882352941, - "loss": 2.8196, + "learning_rate": 2.8442095588235298e-05, + "loss": 0.4531, "step": 620 }, { "epoch": 0.7, - "learning_rate": 0.00028414522058823526, - "loss": 2.807, + "learning_rate": 2.842830882352941e-05, + "loss": 0.4626, "step": 621 }, { "epoch": 0.7, - "learning_rate": 0.0002840073529411764, - "loss": 2.8075, + "learning_rate": 2.841452205882353e-05, + "loss": 0.4457, "step": 622 }, { "epoch": 0.7, - "learning_rate": 0.0002838694852941176, - "loss": 2.8218, + "learning_rate": 2.8400735294117645e-05, + "loss": 0.474, "step": 623 }, { "epoch": 0.7, - "learning_rate": 0.0002837316176470588, - "loss": 2.8199, + "learning_rate": 2.8386948529411766e-05, + "loss": 0.449, "step": 624 }, { "epoch": 0.7, - "learning_rate": 0.00028359375, - "loss": 3.1865, + "learning_rate": 2.8373161764705883e-05, + "loss": 0.4694, "step": 625 }, { "epoch": 0.7, - "learning_rate": 0.00028345588235294115, - "loss": 2.8079, + "learning_rate": 2.8359375e-05, + "loss": 0.4535, "step": 626 }, { "epoch": 0.7, - "learning_rate": 0.0002833180147058823, - "loss": 2.8316, + "learning_rate": 2.8345588235294117e-05, + "loss": 0.4342, "step": 627 }, { "epoch": 0.7, - "learning_rate": 0.0002831801470588235, - "loss": 2.8257, + "learning_rate": 2.8331801470588234e-05, + "loss": 0.4268, "step": 628 }, { "epoch": 0.71, - "learning_rate": 0.00028304227941176467, - "loss": 2.8183, + "learning_rate": 2.8318014705882354e-05, + "loss": 0.4602, "step": 629 }, { "epoch": 0.71, - "learning_rate": 0.00028290441176470583, - "loss": 2.8689, + "learning_rate": 2.830422794117647e-05, + "loss": 0.4692, "step": 630 }, { "epoch": 0.71, - "learning_rate": 0.00028276654411764704, - "loss": 3.2119, + "learning_rate": 2.8290441176470588e-05, + "loss": 0.469, "step": 631 }, { "epoch": 0.71, - "learning_rate": 0.0002826286764705882, - "loss": 3.9641, + "learning_rate": 2.8276654411764705e-05, + "loss": 0.4539, "step": 632 }, { "epoch": 0.71, - "learning_rate": 0.0002824908088235294, - "loss": 2.8135, + "learning_rate": 2.8262867647058826e-05, + "loss": 0.4317, "step": 633 }, { "epoch": 0.71, - "learning_rate": 0.00028235294117647056, - "loss": 2.7965, + "learning_rate": 2.8249080882352943e-05, + "loss": 0.4035, "step": 634 }, { "epoch": 0.71, - "learning_rate": 0.0002822150735294117, - "loss": 2.8331, + "learning_rate": 2.823529411764706e-05, + "loss": 0.4406, "step": 635 }, { "epoch": 0.71, - "learning_rate": 0.00028207720588235293, - "loss": 2.81, + "learning_rate": 2.8221507352941177e-05, + "loss": 0.4291, "step": 636 }, { "epoch": 0.71, - "learning_rate": 0.0002819393382352941, - "loss": 2.8587, + "learning_rate": 2.8207720588235297e-05, + "loss": 0.4751, "step": 637 }, { "epoch": 0.72, - "learning_rate": 0.00028180147058823524, - "loss": 4.1234, + "learning_rate": 2.8193933823529414e-05, + "loss": 0.5036, "step": 638 }, { "epoch": 0.72, - "learning_rate": 0.00028166360294117645, - "loss": 2.8348, + "learning_rate": 2.818014705882353e-05, + "loss": 0.4227, "step": 639 }, { "epoch": 0.72, - "learning_rate": 0.0002815257352941176, - "loss": 2.8475, + "learning_rate": 2.8166360294117648e-05, + "loss": 0.4026, "step": 640 }, { "epoch": 0.72, - "learning_rate": 0.0002813878676470588, - "loss": 2.8167, + "learning_rate": 2.815257352941177e-05, + "loss": 0.3661, "step": 641 }, { "epoch": 0.72, - "learning_rate": 0.00028125, - "loss": 2.8399, + "learning_rate": 2.8138786764705882e-05, + "loss": 0.4305, "step": 642 }, { "epoch": 0.72, - "learning_rate": 0.00028111213235294114, - "loss": 2.8664, + "learning_rate": 2.8125e-05, + "loss": 0.429, "step": 643 }, { "epoch": 0.72, - "learning_rate": 0.0002809742647058823, - "loss": 4.078, + "learning_rate": 2.8111213235294116e-05, + "loss": 0.4873, "step": 644 }, { "epoch": 0.72, - "learning_rate": 0.0002808363970588235, - "loss": 2.8063, + "learning_rate": 2.8097426470588233e-05, + "loss": 0.4256, "step": 645 }, { "epoch": 0.72, - "learning_rate": 0.00028069852941176466, - "loss": 2.8144, + "learning_rate": 2.8083639705882354e-05, + "loss": 0.4355, "step": 646 }, { "epoch": 0.73, - "learning_rate": 0.00028056066176470587, - "loss": 2.8141, + "learning_rate": 2.806985294117647e-05, + "loss": 0.4006, "step": 647 }, { "epoch": 0.73, - "learning_rate": 0.000280422794117647, - "loss": 2.8228, + "learning_rate": 2.8056066176470588e-05, + "loss": 0.4009, "step": 648 }, { "epoch": 0.73, - "learning_rate": 0.00028028492647058824, - "loss": 2.8379, + "learning_rate": 2.8042279411764705e-05, + "loss": 0.4652, "step": 649 }, { "epoch": 0.73, - "learning_rate": 0.0002801470588235294, - "loss": 3.2831, + "learning_rate": 2.8028492647058825e-05, + "loss": 0.5123, "step": 650 }, { "epoch": 0.73, - "learning_rate": 0.00028000919117647055, - "loss": 2.8274, + "learning_rate": 2.8014705882352942e-05, + "loss": 0.4193, "step": 651 }, { "epoch": 0.73, - "learning_rate": 0.0002798713235294117, - "loss": 2.8309, + "learning_rate": 2.800091911764706e-05, + "loss": 0.408, "step": 652 }, { "epoch": 0.73, - "learning_rate": 0.0002797334558823529, - "loss": 2.8468, + "learning_rate": 2.7987132352941176e-05, + "loss": 0.4053, "step": 653 }, { "epoch": 0.73, - "learning_rate": 0.0002795955882352941, - "loss": 2.8193, + "learning_rate": 2.7973345588235297e-05, + "loss": 0.4329, "step": 654 }, { "epoch": 0.73, - "learning_rate": 0.0002794577205882353, - "loss": 2.8409, + "learning_rate": 2.7959558823529414e-05, + "loss": 0.4188, "step": 655 }, { "epoch": 0.74, - "learning_rate": 0.00027931985294117644, - "loss": 3.0539, + "learning_rate": 2.794577205882353e-05, + "loss": 0.4378, "step": 656 }, { "epoch": 0.74, - "learning_rate": 0.00027918198529411765, - "loss": 3.8312, + "learning_rate": 2.7931985294117648e-05, + "loss": 0.4631, "step": 657 }, { "epoch": 0.74, - "learning_rate": 0.0002790441176470588, - "loss": 2.8279, + "learning_rate": 2.7918198529411768e-05, + "loss": 0.4003, "step": 658 }, { "epoch": 0.74, - "learning_rate": 0.00027890624999999997, - "loss": 2.8283, + "learning_rate": 2.7904411764705885e-05, + "loss": 0.4331, "step": 659 }, { "epoch": 0.74, - "learning_rate": 0.0002787683823529411, - "loss": 2.8411, + "learning_rate": 2.7890625000000002e-05, + "loss": 0.3867, "step": 660 }, { "epoch": 0.74, - "learning_rate": 0.00027863051470588233, - "loss": 2.8132, + "learning_rate": 2.787683823529412e-05, + "loss": 0.3655, "step": 661 }, { "epoch": 0.74, - "learning_rate": 0.0002784926470588235, - "loss": 2.8871, + "learning_rate": 2.7863051470588236e-05, + "loss": 0.4028, "step": 662 }, { "epoch": 0.74, - "learning_rate": 0.0002783547794117647, - "loss": 4.2448, + "learning_rate": 2.7849264705882353e-05, + "loss": 0.499, "step": 663 }, { "epoch": 0.74, - "learning_rate": 0.00027821691176470586, - "loss": 2.8285, + "learning_rate": 2.783547794117647e-05, + "loss": 0.3902, "step": 664 }, { "epoch": 0.75, - "learning_rate": 0.000278079044117647, - "loss": 2.8308, + "learning_rate": 2.7821691176470587e-05, + "loss": 0.3751, "step": 665 }, { "epoch": 0.75, - "learning_rate": 0.0002779411764705882, - "loss": 2.8466, + "learning_rate": 2.7807904411764704e-05, + "loss": 0.4024, "step": 666 }, { "epoch": 0.75, - "learning_rate": 0.0002778033088235294, - "loss": 2.8236, + "learning_rate": 2.7794117647058824e-05, + "loss": 0.3939, "step": 667 }, { "epoch": 0.75, - "learning_rate": 0.00027766544117647054, - "loss": 2.8398, + "learning_rate": 2.778033088235294e-05, + "loss": 0.4713, "step": 668 }, { "epoch": 0.75, - "learning_rate": 0.00027752757352941175, - "loss": 4.8878, + "learning_rate": 2.776654411764706e-05, + "loss": 0.5055, "step": 669 }, { "epoch": 0.75, - "learning_rate": 0.0002773897058823529, - "loss": 2.8042, + "learning_rate": 2.7752757352941176e-05, + "loss": 0.3821, "step": 670 }, { "epoch": 0.75, - "learning_rate": 0.0002772518382352941, - "loss": 2.8322, + "learning_rate": 2.7738970588235296e-05, + "loss": 0.3911, "step": 671 }, { "epoch": 0.75, - "learning_rate": 0.00027711397058823527, - "loss": 2.8255, + "learning_rate": 2.7725183823529413e-05, + "loss": 0.3993, "step": 672 }, { "epoch": 0.75, - "learning_rate": 0.00027697610294117643, - "loss": 2.8068, + "learning_rate": 2.771139705882353e-05, + "loss": 0.3727, "step": 673 }, { "epoch": 0.76, - "learning_rate": 0.00027683823529411764, - "loss": 2.811, + "learning_rate": 2.7697610294117647e-05, + "loss": 0.399, "step": 674 }, { "epoch": 0.76, - "learning_rate": 0.0002767003676470588, - "loss": 3.0901, + "learning_rate": 2.7683823529411767e-05, + "loss": 0.4778, "step": 675 }, { "epoch": 0.76, - "learning_rate": 0.00027656249999999995, - "loss": 2.8249, + "learning_rate": 2.7670036764705884e-05, + "loss": 0.363, "step": 676 }, { "epoch": 0.76, - "learning_rate": 0.00027642463235294116, - "loss": 2.826, + "learning_rate": 2.765625e-05, + "loss": 0.401, "step": 677 }, { "epoch": 0.76, - "learning_rate": 0.0002762867647058823, - "loss": 2.8103, + "learning_rate": 2.764246323529412e-05, + "loss": 0.3416, "step": 678 }, { "epoch": 0.76, - "learning_rate": 0.00027614889705882353, - "loss": 2.8153, + "learning_rate": 2.7628676470588235e-05, + "loss": 0.3849, "step": 679 }, { "epoch": 0.76, - "learning_rate": 0.0002760110294117647, - "loss": 2.8621, + "learning_rate": 2.7614889705882356e-05, + "loss": 0.4183, "step": 680 }, { "epoch": 0.76, - "learning_rate": 0.00027587316176470584, - "loss": 3.0923, + "learning_rate": 2.7601102941176473e-05, + "loss": 0.4373, "step": 681 }, { "epoch": 0.76, - "learning_rate": 0.000275735294117647, - "loss": 4.1167, + "learning_rate": 2.758731617647059e-05, + "loss": 0.4197, "step": 682 }, { "epoch": 0.77, - "learning_rate": 0.0002755974264705882, - "loss": 2.8132, + "learning_rate": 2.7573529411764707e-05, + "loss": 0.437, "step": 683 }, { "epoch": 0.77, - "learning_rate": 0.00027545955882352937, - "loss": 2.8315, + "learning_rate": 2.7559742647058824e-05, + "loss": 0.3953, "step": 684 }, { "epoch": 0.77, - "learning_rate": 0.0002753216911764706, - "loss": 2.8058, + "learning_rate": 2.754595588235294e-05, + "loss": 0.3545, "step": 685 }, { "epoch": 0.77, - "learning_rate": 0.00027518382352941174, - "loss": 2.8276, + "learning_rate": 2.7532169117647058e-05, + "loss": 0.3413, "step": 686 }, { "epoch": 0.77, - "learning_rate": 0.00027504595588235295, - "loss": 2.9565, + "learning_rate": 2.7518382352941175e-05, + "loss": 0.3978, "step": 687 }, { "epoch": 0.77, - "learning_rate": 0.0002749080882352941, - "loss": 4.231, + "learning_rate": 2.7504595588235295e-05, + "loss": 0.4487, "step": 688 }, { "epoch": 0.77, - "learning_rate": 0.00027477022058823526, - "loss": 2.8207, + "learning_rate": 2.7490808823529412e-05, + "loss": 0.3637, "step": 689 }, { "epoch": 0.77, - "learning_rate": 0.0002746323529411764, - "loss": 2.8067, + "learning_rate": 2.747702205882353e-05, + "loss": 0.3719, "step": 690 }, { "epoch": 0.77, - "learning_rate": 0.0002744944852941176, - "loss": 2.8378, + "learning_rate": 2.7463235294117646e-05, + "loss": 0.4103, "step": 691 }, { "epoch": 0.78, - "learning_rate": 0.0002743566176470588, - "loss": 2.8217, + "learning_rate": 2.7449448529411767e-05, + "loss": 0.3861, "step": 692 }, { "epoch": 0.78, - "learning_rate": 0.00027421875, - "loss": 2.8747, + "learning_rate": 2.7435661764705884e-05, + "loss": 0.373, "step": 693 }, { "epoch": 0.78, - "learning_rate": 0.00027408088235294115, - "loss": 4.8488, + "learning_rate": 2.7421875e-05, + "loss": 0.4853, "step": 694 }, { "epoch": 0.78, - "learning_rate": 0.0002739430147058823, - "loss": 2.8247, + "learning_rate": 2.7408088235294118e-05, + "loss": 0.3786, "step": 695 }, { "epoch": 0.78, - "learning_rate": 0.0002738051470588235, - "loss": 2.81, + "learning_rate": 2.7394301470588235e-05, + "loss": 0.3743, "step": 696 }, { "epoch": 0.78, - "learning_rate": 0.0002736672794117647, - "loss": 2.8309, + "learning_rate": 2.7380514705882355e-05, + "loss": 0.342, "step": 697 }, { "epoch": 0.78, - "learning_rate": 0.00027352941176470583, - "loss": 2.8156, + "learning_rate": 2.7366727941176472e-05, + "loss": 0.3738, "step": 698 }, { "epoch": 0.78, - "learning_rate": 0.00027339154411764704, - "loss": 2.9112, + "learning_rate": 2.735294117647059e-05, + "loss": 0.3617, "step": 699 }, { "epoch": 0.78, - "learning_rate": 0.0002732536764705882, - "loss": 2.914, + "learning_rate": 2.7339154411764706e-05, + "loss": 0.4827, "step": 700 }, { "epoch": 0.78, - "eval_loss": 3.6876211166381836, - "eval_runtime": 48.0292, - "eval_samples_per_second": 56.278, - "eval_steps_per_second": 0.895, - "eval_wer": 1.0, + "eval_loss": 0.41219812631607056, + "eval_runtime": 47.274, + "eval_samples_per_second": 57.177, + "eval_steps_per_second": 0.91, + "eval_wer": 0.36296459688982025, "step": 700 }, { "epoch": 0.79, - "learning_rate": 0.0002731158088235294, - "loss": 2.8387, + "learning_rate": 2.7325367647058827e-05, + "loss": 0.348, "step": 701 }, { "epoch": 0.79, - "learning_rate": 0.00027297794117647057, - "loss": 2.8485, + "learning_rate": 2.7311580882352944e-05, + "loss": 0.3513, "step": 702 }, { "epoch": 0.79, - "learning_rate": 0.0002728400735294117, - "loss": 2.8542, + "learning_rate": 2.729779411764706e-05, + "loss": 0.3614, "step": 703 }, { "epoch": 0.79, - "learning_rate": 0.00027270220588235293, - "loss": 2.8543, + "learning_rate": 2.7284007352941178e-05, + "loss": 0.374, "step": 704 }, { "epoch": 0.79, - "learning_rate": 0.0002725643382352941, - "loss": 2.8552, + "learning_rate": 2.7270220588235295e-05, + "loss": 0.4034, "step": 705 }, { "epoch": 0.79, - "learning_rate": 0.00027242647058823525, - "loss": 2.9024, + "learning_rate": 2.7256433823529412e-05, + "loss": 0.3486, "step": 706 }, { "epoch": 0.79, - "learning_rate": 0.00027228860294117646, - "loss": 3.8765, + "learning_rate": 2.724264705882353e-05, + "loss": 0.4164, "step": 707 }, { "epoch": 0.79, - "learning_rate": 0.0002721507352941176, - "loss": 2.8125, + "learning_rate": 2.7228860294117646e-05, + "loss": 0.3293, "step": 708 }, { "epoch": 0.79, - "learning_rate": 0.0002720128676470588, - "loss": 2.836, + "learning_rate": 2.7215073529411766e-05, + "loss": 0.356, "step": 709 }, { "epoch": 0.8, - "learning_rate": 0.000271875, - "loss": 2.8317, + "learning_rate": 2.7201286764705883e-05, + "loss": 0.3839, "step": 710 }, { "epoch": 0.8, - "learning_rate": 0.00027173713235294114, - "loss": 2.8239, + "learning_rate": 2.71875e-05, + "loss": 0.3559, "step": 711 }, { "epoch": 0.8, - "learning_rate": 0.0002715992647058823, - "loss": 2.9014, + "learning_rate": 2.7173713235294117e-05, + "loss": 0.3932, "step": 712 }, { "epoch": 0.8, - "learning_rate": 0.0002714613970588235, - "loss": 3.9069, + "learning_rate": 2.7159926470588234e-05, + "loss": 0.4496, "step": 713 }, { "epoch": 0.8, - "learning_rate": 0.00027132352941176466, - "loss": 2.8391, + "learning_rate": 2.7146139705882355e-05, + "loss": 0.3224, "step": 714 }, { "epoch": 0.8, - "learning_rate": 0.00027118566176470587, - "loss": 2.8728, + "learning_rate": 2.713235294117647e-05, + "loss": 0.3614, "step": 715 }, { "epoch": 0.8, - "learning_rate": 0.00027104779411764703, - "loss": 2.8638, + "learning_rate": 2.711856617647059e-05, + "loss": 0.3597, "step": 716 }, { "epoch": 0.8, - "learning_rate": 0.00027090992647058824, - "loss": 2.8222, + "learning_rate": 2.7104779411764706e-05, + "loss": 0.3215, "step": 717 }, { "epoch": 0.8, - "learning_rate": 0.0002707720588235294, - "loss": 2.9042, + "learning_rate": 2.7090992647058826e-05, + "loss": 0.3507, "step": 718 }, { "epoch": 0.81, - "learning_rate": 0.00027063419117647055, - "loss": 4.7086, + "learning_rate": 2.7077205882352943e-05, + "loss": 0.4947, "step": 719 }, { "epoch": 0.81, - "learning_rate": 0.0002704963235294117, - "loss": 2.8282, + "learning_rate": 2.706341911764706e-05, + "loss": 0.3594, "step": 720 }, { "epoch": 0.81, - "learning_rate": 0.0002703584558823529, - "loss": 2.8171, + "learning_rate": 2.7049632352941177e-05, + "loss": 0.3777, "step": 721 }, { "epoch": 0.81, - "learning_rate": 0.0002702205882352941, - "loss": 2.7947, + "learning_rate": 2.7035845588235297e-05, + "loss": 0.3382, "step": 722 }, { "epoch": 0.81, - "learning_rate": 0.0002700827205882353, - "loss": 2.8056, + "learning_rate": 2.7022058823529414e-05, + "loss": 0.3308, "step": 723 }, { "epoch": 0.81, - "learning_rate": 0.00026994485294117644, - "loss": 2.8598, + "learning_rate": 2.700827205882353e-05, + "loss": 0.3491, "step": 724 }, { "epoch": 0.81, - "learning_rate": 0.00026980698529411766, - "loss": 3.3177, + "learning_rate": 2.6994485294117645e-05, + "loss": 0.5004, "step": 725 }, { "epoch": 0.81, - "learning_rate": 0.0002696691176470588, - "loss": 2.7975, + "learning_rate": 2.6980698529411766e-05, + "loss": 0.2959, "step": 726 }, { "epoch": 0.82, - "learning_rate": 0.00026953124999999997, - "loss": 2.8296, + "learning_rate": 2.6966911764705883e-05, + "loss": 0.3375, "step": 727 }, { "epoch": 0.82, - "learning_rate": 0.0002693933823529411, - "loss": 2.8097, + "learning_rate": 2.6953125e-05, + "loss": 0.3627, "step": 728 }, { "epoch": 0.82, - "learning_rate": 0.00026925551470588234, - "loss": 2.8162, + "learning_rate": 2.6939338235294117e-05, + "loss": 0.3515, "step": 729 }, { "epoch": 0.82, - "learning_rate": 0.0002691176470588235, - "loss": 2.8327, + "learning_rate": 2.6925551470588234e-05, + "loss": 0.3662, "step": 730 }, { "epoch": 0.82, - "learning_rate": 0.0002689797794117647, - "loss": 3.0546, + "learning_rate": 2.6911764705882354e-05, + "loss": 0.4085, "step": 731 }, { "epoch": 0.82, - "learning_rate": 0.00026884191176470586, - "loss": 4.7525, + "learning_rate": 2.689797794117647e-05, + "loss": 0.3357, "step": 732 }, { "epoch": 0.82, - "learning_rate": 0.000268704044117647, - "loss": 2.802, + "learning_rate": 2.6884191176470588e-05, + "loss": 0.3456, "step": 733 }, { "epoch": 0.82, - "learning_rate": 0.0002685661764705882, - "loss": 2.8237, + "learning_rate": 2.6870404411764705e-05, + "loss": 0.3477, "step": 734 }, { "epoch": 0.82, - "learning_rate": 0.0002684283088235294, - "loss": 2.8093, + "learning_rate": 2.6856617647058825e-05, + "loss": 0.3374, "step": 735 }, { "epoch": 0.83, - "learning_rate": 0.00026829044117647054, - "loss": 2.8204, + "learning_rate": 2.6842830882352942e-05, + "loss": 0.3321, "step": 736 }, { "epoch": 0.83, - "learning_rate": 0.00026815257352941175, - "loss": 2.8852, + "learning_rate": 2.682904411764706e-05, + "loss": 0.363, "step": 737 }, { "epoch": 0.83, - "learning_rate": 0.0002680147058823529, - "loss": 4.4256, + "learning_rate": 2.6815257352941176e-05, + "loss": 0.411, "step": 738 }, { "epoch": 0.83, - "learning_rate": 0.0002678768382352941, - "loss": 2.8156, + "learning_rate": 2.6801470588235297e-05, + "loss": 0.3327, "step": 739 }, { "epoch": 0.83, - "learning_rate": 0.0002677389705882353, - "loss": 2.8291, + "learning_rate": 2.6787683823529414e-05, + "loss": 0.3452, "step": 740 }, { "epoch": 0.83, - "learning_rate": 0.00026760110294117643, - "loss": 2.8506, + "learning_rate": 2.677389705882353e-05, + "loss": 0.3833, "step": 741 }, { "epoch": 0.83, - "learning_rate": 0.00026746323529411764, - "loss": 2.8297, + "learning_rate": 2.6760110294117648e-05, + "loss": 0.3251, "step": 742 }, { "epoch": 0.83, - "learning_rate": 0.0002673253676470588, - "loss": 2.8687, + "learning_rate": 2.6746323529411768e-05, + "loss": 0.332, "step": 743 }, { "epoch": 0.83, - "learning_rate": 0.00026718749999999996, - "loss": 3.8608, + "learning_rate": 2.6732536764705885e-05, + "loss": 0.3983, "step": 744 }, { "epoch": 0.84, - "learning_rate": 0.00026704963235294117, - "loss": 2.815, + "learning_rate": 2.6718750000000002e-05, + "loss": 0.3701, "step": 745 }, { "epoch": 0.84, - "learning_rate": 0.0002669117647058823, - "loss": 2.8278, + "learning_rate": 2.6704963235294116e-05, + "loss": 0.378, "step": 746 }, { "epoch": 0.84, - "learning_rate": 0.00026677389705882353, - "loss": 2.8201, + "learning_rate": 2.6691176470588233e-05, + "loss": 0.3201, "step": 747 }, { "epoch": 0.84, - "learning_rate": 0.0002666360294117647, - "loss": 2.8414, + "learning_rate": 2.6677389705882353e-05, + "loss": 0.3012, "step": 748 }, { "epoch": 0.84, - "learning_rate": 0.00026649816176470585, - "loss": 2.8381, + "learning_rate": 2.666360294117647e-05, + "loss": 0.4314, "step": 749 }, { "epoch": 0.84, - "learning_rate": 0.000266360294117647, - "loss": 3.0428, + "learning_rate": 2.6649816176470587e-05, + "loss": 0.4148, "step": 750 }, { "epoch": 0.84, - "learning_rate": 0.0002662224264705882, - "loss": 2.8352, + "learning_rate": 2.6636029411764704e-05, + "loss": 0.3164, "step": 751 }, { "epoch": 0.84, - "learning_rate": 0.00026608455882352937, - "loss": 2.8417, + "learning_rate": 2.6622242647058825e-05, + "loss": 0.3253, "step": 752 }, { "epoch": 0.84, - "learning_rate": 0.0002659466911764706, - "loss": 2.8426, + "learning_rate": 2.6608455882352942e-05, + "loss": 0.3203, "step": 753 }, { "epoch": 0.85, - "learning_rate": 0.00026580882352941174, - "loss": 2.8534, + "learning_rate": 2.659466911764706e-05, + "loss": 0.3536, "step": 754 }, { "epoch": 0.85, - "learning_rate": 0.00026567095588235295, - "loss": 2.8482, + "learning_rate": 2.6580882352941176e-05, + "loss": 0.328, "step": 755 }, { "epoch": 0.85, - "learning_rate": 0.0002655330882352941, - "loss": 3.005, + "learning_rate": 2.6567095588235296e-05, + "loss": 0.3224, "step": 756 }, { "epoch": 0.85, - "learning_rate": 0.00026539522058823526, - "loss": 3.6811, + "learning_rate": 2.6553308823529413e-05, + "loss": 0.4244, "step": 757 }, { "epoch": 0.85, - "learning_rate": 0.0002652573529411764, - "loss": 2.8072, + "learning_rate": 2.653952205882353e-05, + "loss": 0.355, "step": 758 }, { "epoch": 0.85, - "learning_rate": 0.00026511948529411763, - "loss": 2.8098, + "learning_rate": 2.6525735294117647e-05, + "loss": 0.3062, "step": 759 }, { "epoch": 0.85, - "learning_rate": 0.0002649816176470588, - "loss": 2.837, + "learning_rate": 2.6511948529411768e-05, + "loss": 0.3744, "step": 760 }, { "epoch": 0.85, - "learning_rate": 0.00026484375, - "loss": 2.8162, + "learning_rate": 2.6498161764705885e-05, + "loss": 0.332, "step": 761 }, { "epoch": 0.85, - "learning_rate": 0.00026470588235294115, - "loss": 2.865, + "learning_rate": 2.6484375000000002e-05, + "loss": 0.3474, "step": 762 }, { "epoch": 0.86, - "learning_rate": 0.0002645680147058823, - "loss": 4.245, + "learning_rate": 2.647058823529412e-05, + "loss": 0.4157, "step": 763 }, { "epoch": 0.86, - "learning_rate": 0.0002644301470588235, - "loss": 2.8288, + "learning_rate": 2.6456801470588236e-05, + "loss": 0.311, "step": 764 }, { "epoch": 0.86, - "learning_rate": 0.0002642922794117647, - "loss": 2.8234, + "learning_rate": 2.6443014705882356e-05, + "loss": 0.315, "step": 765 }, { "epoch": 0.86, - "learning_rate": 0.00026415441176470583, - "loss": 2.8109, + "learning_rate": 2.6429227941176473e-05, + "loss": 0.3005, "step": 766 }, { "epoch": 0.86, - "learning_rate": 0.00026401654411764704, - "loss": 2.8187, + "learning_rate": 2.6415441176470587e-05, + "loss": 0.3327, "step": 767 }, { "epoch": 0.86, - "learning_rate": 0.0002638786764705882, - "loss": 2.8301, + "learning_rate": 2.6401654411764704e-05, + "loss": 0.3596, "step": 768 }, { "epoch": 0.86, - "learning_rate": 0.0002637408088235294, - "loss": 4.1032, + "learning_rate": 2.6387867647058824e-05, + "loss": 0.4152, "step": 769 }, { "epoch": 0.86, - "learning_rate": 0.00026360294117647057, - "loss": 2.8131, + "learning_rate": 2.637408088235294e-05, + "loss": 0.3332, "step": 770 }, { "epoch": 0.86, - "learning_rate": 0.0002634650735294117, - "loss": 2.8127, + "learning_rate": 2.6360294117647058e-05, + "loss": 0.3003, "step": 771 }, { "epoch": 0.87, - "learning_rate": 0.00026332720588235294, - "loss": 2.8267, + "learning_rate": 2.6346507352941175e-05, + "loss": 0.3397, "step": 772 }, { "epoch": 0.87, - "learning_rate": 0.0002631893382352941, - "loss": 2.831, + "learning_rate": 2.6332720588235296e-05, + "loss": 0.3361, "step": 773 }, { "epoch": 0.87, - "learning_rate": 0.00026305147058823525, - "loss": 2.8475, + "learning_rate": 2.6318933823529413e-05, + "loss": 0.3108, "step": 774 }, { "epoch": 0.87, - "learning_rate": 0.00026291360294117646, - "loss": 3.4994, + "learning_rate": 2.630514705882353e-05, + "loss": 0.3785, "step": 775 }, { "epoch": 0.87, - "learning_rate": 0.0002627757352941176, - "loss": 2.8266, + "learning_rate": 2.6291360294117647e-05, + "loss": 0.3352, "step": 776 }, { "epoch": 0.87, - "learning_rate": 0.00026263786764705883, - "loss": 2.8103, + "learning_rate": 2.6277573529411767e-05, + "loss": 0.3207, "step": 777 }, { "epoch": 0.87, - "learning_rate": 0.0002625, - "loss": 2.8086, + "learning_rate": 2.6263786764705884e-05, + "loss": 0.3181, "step": 778 }, { "epoch": 0.87, - "learning_rate": 0.00026236213235294114, - "loss": 2.8206, + "learning_rate": 2.625e-05, + "loss": 0.2929, "step": 779 }, { "epoch": 0.87, - "learning_rate": 0.0002622242647058823, - "loss": 2.8187, + "learning_rate": 2.6236213235294118e-05, + "loss": 0.3183, "step": 780 }, { "epoch": 0.88, - "learning_rate": 0.0002620863970588235, - "loss": 2.8939, + "learning_rate": 2.6222426470588235e-05, + "loss": 0.3258, "step": 781 }, { "epoch": 0.88, - "learning_rate": 0.00026194852941176466, - "loss": 3.6248, + "learning_rate": 2.6208639705882356e-05, + "loss": 0.3722, "step": 782 }, { "epoch": 0.88, - "learning_rate": 0.0002618106617647059, - "loss": 2.83, + "learning_rate": 2.6194852941176473e-05, + "loss": 0.2754, "step": 783 }, { "epoch": 0.88, - "learning_rate": 0.00026167279411764703, - "loss": 2.845, + "learning_rate": 2.618106617647059e-05, + "loss": 0.3379, "step": 784 }, { "epoch": 0.88, - "learning_rate": 0.00026153492647058824, - "loss": 2.8432, + "learning_rate": 2.6167279411764707e-05, + "loss": 0.3096, "step": 785 }, { "epoch": 0.88, - "learning_rate": 0.0002613970588235294, - "loss": 2.8523, + "learning_rate": 2.6153492647058827e-05, + "loss": 0.3431, "step": 786 }, { "epoch": 0.88, - "learning_rate": 0.00026125919117647056, - "loss": 2.8589, + "learning_rate": 2.613970588235294e-05, + "loss": 0.3362, "step": 787 }, { "epoch": 0.88, - "learning_rate": 0.0002611213235294117, - "loss": 3.7905, + "learning_rate": 2.6125919117647058e-05, + "loss": 0.3671, "step": 788 }, { "epoch": 0.88, - "learning_rate": 0.0002609834558823529, - "loss": 2.8113, + "learning_rate": 2.6112132352941175e-05, + "loss": 0.3195, "step": 789 }, { "epoch": 0.89, - "learning_rate": 0.0002608455882352941, - "loss": 2.8163, + "learning_rate": 2.6098345588235295e-05, + "loss": 0.2988, "step": 790 }, { "epoch": 0.89, - "learning_rate": 0.0002607077205882353, - "loss": 2.8196, + "learning_rate": 2.6084558823529412e-05, + "loss": 0.3269, "step": 791 }, { "epoch": 0.89, - "learning_rate": 0.00026056985294117645, - "loss": 2.8409, + "learning_rate": 2.607077205882353e-05, + "loss": 0.3348, "step": 792 }, { "epoch": 0.89, - "learning_rate": 0.00026043198529411766, - "loss": 2.9266, + "learning_rate": 2.6056985294117646e-05, + "loss": 0.3628, "step": 793 }, { "epoch": 0.89, - "learning_rate": 0.0002602941176470588, - "loss": 4.5281, + "learning_rate": 2.6043198529411766e-05, + "loss": 0.4583, "step": 794 }, { "epoch": 0.89, - "learning_rate": 0.00026015624999999997, - "loss": 2.8223, + "learning_rate": 2.6029411764705883e-05, + "loss": 0.2975, "step": 795 }, { "epoch": 0.89, - "learning_rate": 0.00026001838235294113, - "loss": 2.8116, + "learning_rate": 2.6015625e-05, + "loss": 0.303, "step": 796 }, { "epoch": 0.89, - "learning_rate": 0.00025988051470588234, - "loss": 2.8179, + "learning_rate": 2.6001838235294117e-05, + "loss": 0.352, "step": 797 }, { "epoch": 0.89, - "learning_rate": 0.0002597426470588235, - "loss": 2.8134, + "learning_rate": 2.5988051470588234e-05, + "loss": 0.2781, "step": 798 }, { "epoch": 0.9, - "learning_rate": 0.0002596047794117647, - "loss": 2.8758, + "learning_rate": 2.5974264705882355e-05, + "loss": 0.3272, "step": 799 }, { "epoch": 0.9, - "learning_rate": 0.00025946691176470586, - "loss": 3.0777, + "learning_rate": 2.5960477941176472e-05, + "loss": 0.4485, "step": 800 }, { "epoch": 0.9, - "eval_loss": 3.7421417236328125, - "eval_runtime": 47.8097, - "eval_samples_per_second": 56.537, - "eval_steps_per_second": 0.899, - "eval_wer": 1.0, + "eval_loss": 0.35965707898139954, + "eval_runtime": 47.2832, + "eval_samples_per_second": 57.166, + "eval_steps_per_second": 0.909, + "eval_wer": 0.32434469320980847, "step": 800 }, { "epoch": 0.9, - "learning_rate": 0.000259329044117647, - "loss": 2.8261, + "learning_rate": 2.594669117647059e-05, + "loss": 0.3064, "step": 801 }, { "epoch": 0.9, - "learning_rate": 0.00025919117647058823, - "loss": 2.8294, + "learning_rate": 2.5932904411764706e-05, + "loss": 0.3144, "step": 802 }, { "epoch": 0.9, - "learning_rate": 0.0002590533088235294, - "loss": 2.8353, + "learning_rate": 2.5919117647058826e-05, + "loss": 0.3242, "step": 803 }, { "epoch": 0.9, - "learning_rate": 0.00025891544117647054, - "loss": 2.8106, + "learning_rate": 2.5905330882352943e-05, + "loss": 0.3075, "step": 804 }, { "epoch": 0.9, - "learning_rate": 0.00025877757352941175, - "loss": 2.8244, + "learning_rate": 2.589154411764706e-05, + "loss": 0.3153, "step": 805 }, { "epoch": 0.9, - "learning_rate": 0.0002586397058823529, - "loss": 2.9765, + "learning_rate": 2.5877757352941177e-05, + "loss": 0.3706, "step": 806 }, { "epoch": 0.9, - "learning_rate": 0.0002585018382352941, - "loss": 4.0711, + "learning_rate": 2.5863970588235298e-05, + "loss": 0.339, "step": 807 }, { "epoch": 0.91, - "learning_rate": 0.0002583639705882353, - "loss": 2.8253, + "learning_rate": 2.585018382352941e-05, + "loss": 0.2937, "step": 808 }, { "epoch": 0.91, - "learning_rate": 0.00025822610294117643, - "loss": 2.8, + "learning_rate": 2.583639705882353e-05, + "loss": 0.2751, "step": 809 }, { "epoch": 0.91, - "learning_rate": 0.00025808823529411764, - "loss": 2.8054, + "learning_rate": 2.5822610294117645e-05, + "loss": 0.3114, "step": 810 }, { "epoch": 0.91, - "learning_rate": 0.0002579503676470588, - "loss": 2.8132, + "learning_rate": 2.5808823529411766e-05, + "loss": 0.2923, "step": 811 }, { "epoch": 0.91, - "learning_rate": 0.00025781249999999996, - "loss": 2.8857, + "learning_rate": 2.5795036764705883e-05, + "loss": 0.3018, "step": 812 }, { "epoch": 0.91, - "learning_rate": 0.00025767463235294117, - "loss": 3.9406, + "learning_rate": 2.578125e-05, + "loss": 0.4102, "step": 813 }, { "epoch": 0.91, - "learning_rate": 0.0002575367647058823, - "loss": 2.8219, + "learning_rate": 2.5767463235294117e-05, + "loss": 0.3353, "step": 814 }, { "epoch": 0.91, - "learning_rate": 0.00025739889705882354, - "loss": 2.8078, + "learning_rate": 2.5753676470588234e-05, + "loss": 0.2906, "step": 815 }, { "epoch": 0.91, - "learning_rate": 0.0002572610294117647, - "loss": 2.8076, + "learning_rate": 2.5739889705882354e-05, + "loss": 0.2779, "step": 816 }, { "epoch": 0.92, - "learning_rate": 0.00025712316176470585, - "loss": 2.8251, + "learning_rate": 2.572610294117647e-05, + "loss": 0.2918, "step": 817 }, { "epoch": 0.92, - "learning_rate": 0.000256985294117647, - "loss": 2.878, + "learning_rate": 2.571231617647059e-05, + "loss": 0.3549, "step": 818 }, { "epoch": 0.92, - "learning_rate": 0.0002568474264705882, - "loss": 3.6558, + "learning_rate": 2.5698529411764705e-05, + "loss": 0.3621, "step": 819 }, { "epoch": 0.92, - "learning_rate": 0.0002567095588235294, - "loss": 2.8178, + "learning_rate": 2.5684742647058826e-05, + "loss": 0.3077, "step": 820 }, { "epoch": 0.92, - "learning_rate": 0.0002565716911764706, - "loss": 2.8352, + "learning_rate": 2.5670955882352943e-05, + "loss": 0.3334, "step": 821 }, { "epoch": 0.92, - "learning_rate": 0.00025643382352941174, - "loss": 2.8225, + "learning_rate": 2.565716911764706e-05, + "loss": 0.2928, "step": 822 }, { "epoch": 0.92, - "learning_rate": 0.00025629595588235295, - "loss": 2.8065, + "learning_rate": 2.5643382352941177e-05, + "loss": 0.2924, "step": 823 }, { "epoch": 0.92, - "learning_rate": 0.0002561580882352941, - "loss": 2.8442, + "learning_rate": 2.5629595588235297e-05, + "loss": 0.2899, "step": 824 }, { "epoch": 0.92, - "learning_rate": 0.00025602022058823526, - "loss": 3.312, + "learning_rate": 2.5615808823529414e-05, + "loss": 0.4031, "step": 825 }, { "epoch": 0.93, - "learning_rate": 0.0002558823529411764, - "loss": 2.8276, + "learning_rate": 2.560202205882353e-05, + "loss": 0.3077, "step": 826 }, { "epoch": 0.93, - "learning_rate": 0.00025574448529411763, - "loss": 2.8119, + "learning_rate": 2.5588235294117648e-05, + "loss": 0.3115, "step": 827 }, { "epoch": 0.93, - "learning_rate": 0.0002556066176470588, - "loss": 2.8105, + "learning_rate": 2.557444852941177e-05, + "loss": 0.3135, "step": 828 }, { "epoch": 0.93, - "learning_rate": 0.00025546875, - "loss": 2.8137, + "learning_rate": 2.5560661764705882e-05, + "loss": 0.2564, "step": 829 }, { "epoch": 0.93, - "learning_rate": 0.00025533088235294116, - "loss": 2.8187, + "learning_rate": 2.5546875e-05, + "loss": 0.2936, "step": 830 }, { "epoch": 0.93, - "learning_rate": 0.0002551930147058823, - "loss": 2.943, + "learning_rate": 2.5533088235294116e-05, + "loss": 0.3773, "step": 831 }, { "epoch": 0.93, - "learning_rate": 0.0002550551470588235, - "loss": 3.5419, + "learning_rate": 2.5519301470588233e-05, + "loss": 0.334, "step": 832 }, { "epoch": 0.93, - "learning_rate": 0.0002549172794117647, - "loss": 2.8052, + "learning_rate": 2.5505514705882354e-05, + "loss": 0.3005, "step": 833 }, { "epoch": 0.93, - "learning_rate": 0.00025477941176470584, - "loss": 2.8134, + "learning_rate": 2.549172794117647e-05, + "loss": 0.2818, "step": 834 }, { "epoch": 0.94, - "learning_rate": 0.00025464154411764705, - "loss": 2.8156, + "learning_rate": 2.5477941176470588e-05, + "loss": 0.301, "step": 835 }, { "epoch": 0.94, - "learning_rate": 0.0002545036764705882, - "loss": 2.8022, + "learning_rate": 2.5464154411764705e-05, + "loss": 0.286, "step": 836 }, { "epoch": 0.94, - "learning_rate": 0.0002543658088235294, - "loss": 2.9061, + "learning_rate": 2.5450367647058825e-05, + "loss": 0.3242, "step": 837 }, { "epoch": 0.94, - "learning_rate": 0.00025422794117647057, - "loss": 4.1307, + "learning_rate": 2.5436580882352942e-05, + "loss": 0.3986, "step": 838 }, { "epoch": 0.94, - "learning_rate": 0.00025409007352941173, - "loss": 2.811, + "learning_rate": 2.542279411764706e-05, + "loss": 0.2718, "step": 839 }, { "epoch": 0.94, - "learning_rate": 0.00025395220588235294, - "loss": 2.8139, + "learning_rate": 2.5409007352941176e-05, + "loss": 0.2825, "step": 840 }, { "epoch": 0.94, - "learning_rate": 0.0002538143382352941, - "loss": 2.8203, + "learning_rate": 2.5395220588235297e-05, + "loss": 0.2983, "step": 841 }, { "epoch": 0.94, - "learning_rate": 0.00025367647058823525, - "loss": 2.8007, + "learning_rate": 2.5381433823529414e-05, + "loss": 0.3036, "step": 842 }, { "epoch": 0.95, - "learning_rate": 0.00025353860294117646, - "loss": 2.8275, + "learning_rate": 2.536764705882353e-05, + "loss": 0.3012, "step": 843 }, { "epoch": 0.95, - "learning_rate": 0.0002534007352941176, - "loss": 5.1937, + "learning_rate": 2.5353860294117648e-05, + "loss": 0.3767, "step": 844 }, { "epoch": 0.95, - "learning_rate": 0.00025326286764705883, - "loss": 2.8176, + "learning_rate": 2.5340073529411768e-05, + "loss": 0.2919, "step": 845 }, { "epoch": 0.95, - "learning_rate": 0.000253125, - "loss": 2.8201, + "learning_rate": 2.5326286764705885e-05, + "loss": 0.3085, "step": 846 }, { "epoch": 0.95, - "learning_rate": 0.00025298713235294114, - "loss": 2.8241, + "learning_rate": 2.5312500000000002e-05, + "loss": 0.3421, "step": 847 }, { "epoch": 0.95, - "learning_rate": 0.0002528492647058823, - "loss": 2.8199, + "learning_rate": 2.529871323529412e-05, + "loss": 0.3015, "step": 848 }, { "epoch": 0.95, - "learning_rate": 0.0002527113970588235, - "loss": 2.8141, + "learning_rate": 2.5284926470588236e-05, + "loss": 0.2985, "step": 849 }, { "epoch": 0.95, - "learning_rate": 0.00025257352941176467, - "loss": 3.1744, + "learning_rate": 2.5271139705882353e-05, + "loss": 0.3934, "step": 850 }, { "epoch": 0.95, - "learning_rate": 0.0002524356617647059, - "loss": 2.8049, + "learning_rate": 2.525735294117647e-05, + "loss": 0.288, "step": 851 }, { "epoch": 0.96, - "learning_rate": 0.00025229779411764703, - "loss": 2.8085, + "learning_rate": 2.5243566176470587e-05, + "loss": 0.2814, "step": 852 }, { "epoch": 0.96, - "learning_rate": 0.00025215992647058824, - "loss": 2.8171, + "learning_rate": 2.5229779411764704e-05, + "loss": 0.3107, "step": 853 }, { "epoch": 0.96, - "learning_rate": 0.0002520220588235294, - "loss": 2.8089, + "learning_rate": 2.5215992647058824e-05, + "loss": 0.2872, "step": 854 }, { "epoch": 0.96, - "learning_rate": 0.00025188419117647056, - "loss": 2.8282, + "learning_rate": 2.520220588235294e-05, + "loss": 0.3135, "step": 855 }, { "epoch": 0.96, - "learning_rate": 0.0002517463235294117, - "loss": 3.0224, + "learning_rate": 2.518841911764706e-05, + "loss": 0.3231, "step": 856 }, { "epoch": 0.96, - "learning_rate": 0.0002516084558823529, - "loss": 4.1633, + "learning_rate": 2.5174632352941176e-05, + "loss": 0.3271, "step": 857 }, { "epoch": 0.96, - "learning_rate": 0.0002514705882352941, - "loss": 2.8104, + "learning_rate": 2.5160845588235296e-05, + "loss": 0.2821, "step": 858 }, { "epoch": 0.96, - "learning_rate": 0.0002513327205882353, - "loss": 2.8179, + "learning_rate": 2.5147058823529413e-05, + "loss": 0.287, "step": 859 }, { "epoch": 0.96, - "learning_rate": 0.00025119485294117645, - "loss": 2.8114, + "learning_rate": 2.513327205882353e-05, + "loss": 0.2581, "step": 860 }, { "epoch": 0.97, - "learning_rate": 0.00025105698529411766, - "loss": 2.8201, + "learning_rate": 2.5119485294117647e-05, + "loss": 0.2871, "step": 861 }, { "epoch": 0.97, - "learning_rate": 0.0002509191176470588, - "loss": 2.8938, + "learning_rate": 2.5105698529411767e-05, + "loss": 0.296, "step": 862 }, { "epoch": 0.97, - "learning_rate": 0.00025078125, - "loss": 4.1448, + "learning_rate": 2.5091911764705884e-05, + "loss": 0.3581, "step": 863 }, { "epoch": 0.97, - "learning_rate": 0.00025064338235294113, - "loss": 2.8197, + "learning_rate": 2.5078125e-05, + "loss": 0.2903, "step": 864 }, { "epoch": 0.97, - "learning_rate": 0.00025050551470588234, - "loss": 2.8255, + "learning_rate": 2.506433823529412e-05, + "loss": 0.3437, "step": 865 }, { "epoch": 0.97, - "learning_rate": 0.0002503676470588235, - "loss": 2.7987, + "learning_rate": 2.5050551470588235e-05, + "loss": 0.2686, "step": 866 }, { "epoch": 0.97, - "learning_rate": 0.0002502297794117647, - "loss": 2.8415, + "learning_rate": 2.5036764705882356e-05, + "loss": 0.3311, "step": 867 }, { "epoch": 0.97, - "learning_rate": 0.00025009191176470586, - "loss": 2.8747, + "learning_rate": 2.5022977941176473e-05, + "loss": 0.3459, "step": 868 }, { "epoch": 0.97, - "learning_rate": 0.000249954044117647, - "loss": 4.7102, + "learning_rate": 2.500919117647059e-05, + "loss": 0.4307, "step": 869 }, { "epoch": 0.98, - "learning_rate": 0.00024981617647058823, - "loss": 2.8103, + "learning_rate": 2.4995404411764703e-05, + "loss": 0.2759, "step": 870 }, { "epoch": 0.98, - "learning_rate": 0.0002496783088235294, - "loss": 2.8107, + "learning_rate": 2.4981617647058824e-05, + "loss": 0.2955, "step": 871 }, { "epoch": 0.98, - "learning_rate": 0.00024954044117647055, - "loss": 2.81, + "learning_rate": 2.496783088235294e-05, + "loss": 0.2553, "step": 872 }, { "epoch": 0.98, - "learning_rate": 0.00024940257352941176, - "loss": 2.8111, + "learning_rate": 2.4954044117647058e-05, + "loss": 0.2854, "step": 873 }, { "epoch": 0.98, - "learning_rate": 0.0002492647058823529, - "loss": 2.8632, + "learning_rate": 2.4940257352941175e-05, + "loss": 0.3036, "step": 874 }, { "epoch": 0.98, - "learning_rate": 0.0002491268382352941, - "loss": 3.0939, + "learning_rate": 2.4926470588235295e-05, + "loss": 0.3396, "step": 875 }, { "epoch": 0.98, - "learning_rate": 0.0002489889705882353, - "loss": 2.8082, + "learning_rate": 2.4912683823529412e-05, + "loss": 0.3216, "step": 876 }, { "epoch": 0.98, - "learning_rate": 0.00024885110294117644, - "loss": 2.8065, + "learning_rate": 2.489889705882353e-05, + "loss": 0.2962, "step": 877 }, { "epoch": 0.98, - "learning_rate": 0.00024871323529411765, - "loss": 2.8354, + "learning_rate": 2.4885110294117646e-05, + "loss": 0.2602, "step": 878 }, { "epoch": 0.99, - "learning_rate": 0.0002485753676470588, - "loss": 2.824, + "learning_rate": 2.4871323529411767e-05, + "loss": 0.2626, "step": 879 }, { "epoch": 0.99, - "learning_rate": 0.00024843749999999996, - "loss": 2.8248, + "learning_rate": 2.4857536764705884e-05, + "loss": 0.2822, "step": 880 }, { "epoch": 0.99, - "learning_rate": 0.00024829963235294117, - "loss": 2.9677, + "learning_rate": 2.484375e-05, + "loss": 0.3266, "step": 881 }, { "epoch": 0.99, - "learning_rate": 0.00024816176470588233, - "loss": 3.2993, + "learning_rate": 2.4829963235294118e-05, + "loss": 0.312, "step": 882 }, { "epoch": 0.99, - "learning_rate": 0.00024802389705882354, - "loss": 2.8499, + "learning_rate": 2.4816176470588235e-05, + "loss": 0.2325, "step": 883 }, { "epoch": 0.99, - "learning_rate": 0.0002478860294117647, - "loss": 2.9018, + "learning_rate": 2.4802389705882355e-05, + "loss": 0.2846, "step": 884 }, { "epoch": 0.99, - "learning_rate": 0.00024774816176470585, - "loss": 2.9094, + "learning_rate": 2.4788602941176472e-05, + "loss": 0.2544, "step": 885 }, { "epoch": 0.99, - "learning_rate": 0.000247610294117647, - "loss": 2.8661, + "learning_rate": 2.477481617647059e-05, + "loss": 0.2784, "step": 886 }, { "epoch": 0.99, - "learning_rate": 0.0002474724264705882, - "loss": 2.8773, + "learning_rate": 2.4761029411764706e-05, + "loss": 0.2704, "step": 887 }, { "epoch": 1.0, - "learning_rate": 0.0002473345588235294, - "loss": 3.718, + "learning_rate": 2.4747242647058827e-05, + "loss": 0.3243, "step": 888 }, { "epoch": 1.0, - "learning_rate": 0.0002471966911764706, - "loss": 2.8392, + "learning_rate": 2.4733455882352944e-05, + "loss": 0.2741, "step": 889 }, { "epoch": 1.0, - "learning_rate": 0.00024705882352941174, - "loss": 2.8326, + "learning_rate": 2.471966911764706e-05, + "loss": 0.3045, "step": 890 }, { "epoch": 1.0, - "learning_rate": 0.00024692095588235295, - "loss": 2.8275, + "learning_rate": 2.4705882352941174e-05, + "loss": 0.31, "step": 891 }, { "epoch": 1.0, - "learning_rate": 0.0002467830882352941, - "loss": 4.1512, + "learning_rate": 2.4692095588235295e-05, + "loss": 0.3449, "step": 892 }, { "epoch": 1.0, - "learning_rate": 0.00024664522058823527, - "loss": 2.8263, + "learning_rate": 2.4678308823529412e-05, + "loss": 0.3129, "step": 893 }, { "epoch": 1.0, - "learning_rate": 0.0002465073529411764, - "loss": 2.8333, + "learning_rate": 2.466452205882353e-05, + "loss": 0.2506, "step": 894 }, { "epoch": 1.0, - "learning_rate": 0.00024636948529411763, - "loss": 2.8188, + "learning_rate": 2.4650735294117646e-05, + "loss": 0.2811, "step": 895 }, { "epoch": 1.0, - "learning_rate": 0.0002462316176470588, - "loss": 2.7962, + "learning_rate": 2.4636948529411766e-05, + "loss": 0.2733, "step": 896 }, { "epoch": 1.01, - "learning_rate": 0.00024609375, - "loss": 2.8377, + "learning_rate": 2.4623161764705883e-05, + "loss": 0.2693, "step": 897 }, { "epoch": 1.01, - "learning_rate": 0.00024595588235294116, - "loss": 3.0776, + "learning_rate": 2.4609375e-05, + "loss": 0.3042, "step": 898 }, { "epoch": 1.01, - "learning_rate": 0.0002458180147058823, - "loss": 4.4842, + "learning_rate": 2.4595588235294117e-05, + "loss": 0.3386, "step": 899 }, { "epoch": 1.01, - "learning_rate": 0.0002456801470588235, - "loss": 2.8181, + "learning_rate": 2.4581801470588234e-05, + "loss": 0.2666, "step": 900 }, { "epoch": 1.01, - "eval_loss": 3.7824625968933105, - "eval_runtime": 48.0266, - "eval_samples_per_second": 56.281, - "eval_steps_per_second": 0.895, - "eval_wer": 1.0, + "eval_loss": 0.3104283809661865, + "eval_runtime": 48.3749, + "eval_samples_per_second": 55.876, + "eval_steps_per_second": 0.889, + "eval_wer": 0.2789970956950112, "step": 900 }, { "epoch": 1.01, - "learning_rate": 0.0002455422794117647, - "loss": 2.8201, + "learning_rate": 2.4568014705882355e-05, + "loss": 0.2535, "step": 901 }, { "epoch": 1.01, - "learning_rate": 0.00024540441176470584, - "loss": 2.8182, + "learning_rate": 2.455422794117647e-05, + "loss": 0.2481, "step": 902 }, { "epoch": 1.01, - "learning_rate": 0.00024526654411764705, - "loss": 2.8302, + "learning_rate": 2.454044117647059e-05, + "loss": 0.2578, "step": 903 }, { "epoch": 1.01, - "learning_rate": 0.0002451286764705882, - "loss": 2.8796, + "learning_rate": 2.4526654411764706e-05, + "loss": 0.3242, "step": 904 }, { "epoch": 1.01, - "learning_rate": 0.0002449908088235294, - "loss": 4.5008, + "learning_rate": 2.4512867647058826e-05, + "loss": 0.2917, "step": 905 }, { "epoch": 1.02, - "learning_rate": 0.0002448529411764706, - "loss": 2.8077, + "learning_rate": 2.4499080882352943e-05, + "loss": 0.2736, "step": 906 }, { "epoch": 1.02, - "learning_rate": 0.00024471507352941173, - "loss": 2.8081, + "learning_rate": 2.448529411764706e-05, + "loss": 0.2874, "step": 907 }, { "epoch": 1.02, - "learning_rate": 0.00024457720588235294, - "loss": 2.8306, + "learning_rate": 2.4471507352941177e-05, + "loss": 0.2827, "step": 908 }, { "epoch": 1.02, - "learning_rate": 0.0002444393382352941, - "loss": 2.8154, + "learning_rate": 2.4457720588235297e-05, + "loss": 0.2466, "step": 909 }, { "epoch": 1.02, - "learning_rate": 0.00024430147058823525, - "loss": 2.8187, + "learning_rate": 2.4443933823529414e-05, + "loss": 0.2509, "step": 910 }, { "epoch": 1.02, - "learning_rate": 0.00024416360294117646, - "loss": 3.74, + "learning_rate": 2.443014705882353e-05, + "loss": 0.3205, "step": 911 }, { "epoch": 1.02, - "learning_rate": 0.00024402573529411765, - "loss": 2.8217, + "learning_rate": 2.4416360294117645e-05, + "loss": 0.2516, "step": 912 }, { "epoch": 1.02, - "learning_rate": 0.0002438878676470588, - "loss": 2.8194, + "learning_rate": 2.4402573529411766e-05, + "loss": 0.2549, "step": 913 }, { "epoch": 1.02, - "learning_rate": 0.00024375, - "loss": 2.8138, + "learning_rate": 2.4388786764705883e-05, + "loss": 0.2774, "step": 914 }, { "epoch": 1.03, - "learning_rate": 0.00024361213235294115, - "loss": 2.814, + "learning_rate": 2.4375e-05, + "loss": 0.2679, "step": 915 }, { "epoch": 1.03, - "learning_rate": 0.00024347426470588233, - "loss": 2.8299, + "learning_rate": 2.4361213235294117e-05, + "loss": 0.2923, "step": 916 }, { "epoch": 1.03, - "learning_rate": 0.0002433363970588235, - "loss": 3.2741, + "learning_rate": 2.4347426470588234e-05, + "loss": 0.3599, "step": 917 }, { "epoch": 1.03, - "learning_rate": 0.0002431985294117647, - "loss": 2.8186, + "learning_rate": 2.4333639705882354e-05, + "loss": 0.2639, "step": 918 }, { "epoch": 1.03, - "learning_rate": 0.00024306066176470585, - "loss": 2.8205, + "learning_rate": 2.431985294117647e-05, + "loss": 0.2644, "step": 919 }, { "epoch": 1.03, - "learning_rate": 0.00024292279411764704, - "loss": 2.7997, + "learning_rate": 2.4306066176470588e-05, + "loss": 0.2583, "step": 920 }, { "epoch": 1.03, - "learning_rate": 0.00024278492647058822, - "loss": 2.8126, + "learning_rate": 2.4292279411764705e-05, + "loss": 0.2793, "step": 921 }, { "epoch": 1.03, - "learning_rate": 0.0002426470588235294, - "loss": 2.8354, + "learning_rate": 2.4278492647058825e-05, + "loss": 0.2487, "step": 922 }, { "epoch": 1.03, - "learning_rate": 0.00024250919117647056, - "loss": 3.0096, + "learning_rate": 2.4264705882352942e-05, + "loss": 0.3074, "step": 923 }, { "epoch": 1.04, - "learning_rate": 0.00024237132352941174, - "loss": 3.6619, + "learning_rate": 2.425091911764706e-05, + "loss": 0.2945, "step": 924 }, { "epoch": 1.04, - "learning_rate": 0.00024223345588235293, - "loss": 2.8328, + "learning_rate": 2.4237132352941176e-05, + "loss": 0.262, "step": 925 }, { "epoch": 1.04, - "learning_rate": 0.0002420955882352941, - "loss": 2.8325, + "learning_rate": 2.4223345588235297e-05, + "loss": 0.2696, "step": 926 }, { "epoch": 1.04, - "learning_rate": 0.00024195772058823527, - "loss": 2.8435, + "learning_rate": 2.4209558823529414e-05, + "loss": 0.2464, "step": 927 }, { "epoch": 1.04, - "learning_rate": 0.00024181985294117645, - "loss": 2.8096, + "learning_rate": 2.419577205882353e-05, + "loss": 0.2404, "step": 928 }, { "epoch": 1.04, - "learning_rate": 0.00024168198529411764, - "loss": 2.8516, + "learning_rate": 2.4181985294117648e-05, + "loss": 0.2682, "step": 929 }, { "epoch": 1.04, - "learning_rate": 0.00024154411764705882, - "loss": 3.9277, + "learning_rate": 2.416819852941177e-05, + "loss": 0.2796, "step": 930 }, { "epoch": 1.04, - "learning_rate": 0.00024140624999999998, - "loss": 2.819, + "learning_rate": 2.4154411764705885e-05, + "loss": 0.2505, "step": 931 }, { "epoch": 1.04, - "learning_rate": 0.00024126838235294116, - "loss": 2.8146, + "learning_rate": 2.4140625e-05, + "loss": 0.2481, "step": 932 }, { "epoch": 1.05, - "learning_rate": 0.00024113051470588232, - "loss": 2.8151, + "learning_rate": 2.4126838235294116e-05, + "loss": 0.2648, "step": 933 }, { "epoch": 1.05, - "learning_rate": 0.00024099264705882353, - "loss": 2.8212, + "learning_rate": 2.4113051470588233e-05, + "loss": 0.2577, "step": 934 }, { "epoch": 1.05, - "learning_rate": 0.00024085477941176468, - "loss": 2.8763, + "learning_rate": 2.4099264705882353e-05, + "loss": 0.242, "step": 935 }, { "epoch": 1.05, - "learning_rate": 0.00024071691176470587, - "loss": 4.1219, + "learning_rate": 2.408547794117647e-05, + "loss": 0.268, "step": 936 }, { "epoch": 1.05, - "learning_rate": 0.00024057904411764702, - "loss": 2.8044, + "learning_rate": 2.4071691176470587e-05, + "loss": 0.2405, "step": 937 }, { "epoch": 1.05, - "learning_rate": 0.00024044117647058823, - "loss": 2.8199, + "learning_rate": 2.4057904411764704e-05, + "loss": 0.2809, "step": 938 }, { "epoch": 1.05, - "learning_rate": 0.0002403033088235294, - "loss": 2.8045, + "learning_rate": 2.4044117647058825e-05, + "loss": 0.239, "step": 939 }, { "epoch": 1.05, - "learning_rate": 0.00024016544117647057, - "loss": 2.8191, + "learning_rate": 2.4030330882352942e-05, + "loss": 0.2603, "step": 940 }, { "epoch": 1.05, - "learning_rate": 0.00024002757352941173, - "loss": 2.8846, + "learning_rate": 2.401654411764706e-05, + "loss": 0.2876, "step": 941 }, { "epoch": 1.06, - "learning_rate": 0.00023988970588235294, - "loss": 3.1311, + "learning_rate": 2.4002757352941176e-05, + "loss": 0.4034, "step": 942 }, { "epoch": 1.06, - "learning_rate": 0.0002397518382352941, - "loss": 2.8208, + "learning_rate": 2.3988970588235296e-05, + "loss": 0.2721, "step": 943 }, { "epoch": 1.06, - "learning_rate": 0.00023961397058823528, - "loss": 2.8012, + "learning_rate": 2.3975183823529413e-05, + "loss": 0.2434, "step": 944 }, { "epoch": 1.06, - "learning_rate": 0.00023947610294117644, - "loss": 2.8089, + "learning_rate": 2.396139705882353e-05, + "loss": 0.2471, "step": 945 }, { "epoch": 1.06, - "learning_rate": 0.00023933823529411765, - "loss": 2.8125, + "learning_rate": 2.3947610294117647e-05, + "loss": 0.2583, "step": 946 }, { "epoch": 1.06, - "learning_rate": 0.0002392003676470588, - "loss": 2.8139, + "learning_rate": 2.3933823529411768e-05, + "loss": 0.254, "step": 947 }, { "epoch": 1.06, - "learning_rate": 0.0002390625, - "loss": 3.1338, + "learning_rate": 2.3920036764705885e-05, + "loss": 0.2825, "step": 948 }, { "epoch": 1.06, - "learning_rate": 0.00023892463235294115, - "loss": 3.9068, + "learning_rate": 2.3906250000000002e-05, + "loss": 0.2881, "step": 949 }, { "epoch": 1.07, - "learning_rate": 0.0002387867647058823, - "loss": 2.8216, + "learning_rate": 2.389246323529412e-05, + "loss": 0.2647, "step": 950 }, { "epoch": 1.07, - "learning_rate": 0.00023864889705882351, - "loss": 2.8304, + "learning_rate": 2.3878676470588236e-05, + "loss": 0.2627, "step": 951 }, { "epoch": 1.07, - "learning_rate": 0.0002385110294117647, - "loss": 2.8281, + "learning_rate": 2.3864889705882356e-05, + "loss": 0.2306, "step": 952 }, { "epoch": 1.07, - "learning_rate": 0.00023837316176470585, - "loss": 2.8117, + "learning_rate": 2.385110294117647e-05, + "loss": 0.2296, "step": 953 }, { "epoch": 1.07, - "learning_rate": 0.000238235294117647, - "loss": 2.9299, + "learning_rate": 2.3837316176470587e-05, + "loss": 0.2788, "step": 954 }, { "epoch": 1.07, - "learning_rate": 0.00023809742647058822, - "loss": 3.7643, + "learning_rate": 2.3823529411764704e-05, + "loss": 0.3637, "step": 955 }, { "epoch": 1.07, - "learning_rate": 0.0002379595588235294, - "loss": 2.8242, + "learning_rate": 2.3809742647058824e-05, + "loss": 0.2765, "step": 956 }, { "epoch": 1.07, - "learning_rate": 0.00023782169117647056, - "loss": 2.8375, + "learning_rate": 2.379595588235294e-05, + "loss": 0.2592, "step": 957 }, { "epoch": 1.07, - "learning_rate": 0.00023768382352941175, - "loss": 2.8138, + "learning_rate": 2.3782169117647058e-05, + "loss": 0.2521, "step": 958 }, { "epoch": 1.08, - "learning_rate": 0.00023754595588235293, - "loss": 2.8402, + "learning_rate": 2.3768382352941175e-05, + "loss": 0.2243, "step": 959 }, { "epoch": 1.08, - "learning_rate": 0.0002374080882352941, - "loss": 2.8533, + "learning_rate": 2.3754595588235296e-05, + "loss": 0.2261, "step": 960 }, { "epoch": 1.08, - "learning_rate": 0.00023727022058823527, - "loss": 4.2321, + "learning_rate": 2.3740808823529413e-05, + "loss": 0.2732, "step": 961 }, { "epoch": 1.08, - "learning_rate": 0.00023713235294117645, - "loss": 2.8264, + "learning_rate": 2.372702205882353e-05, + "loss": 0.2449, "step": 962 }, { "epoch": 1.08, - "learning_rate": 0.00023699448529411764, - "loss": 2.8155, + "learning_rate": 2.3713235294117647e-05, + "loss": 0.2826, "step": 963 }, { "epoch": 1.08, - "learning_rate": 0.00023685661764705882, - "loss": 2.8334, + "learning_rate": 2.3699448529411767e-05, + "loss": 0.2687, "step": 964 }, { "epoch": 1.08, - "learning_rate": 0.00023671874999999998, - "loss": 2.8, + "learning_rate": 2.3685661764705884e-05, + "loss": 0.2456, "step": 965 }, { "epoch": 1.08, - "learning_rate": 0.00023658088235294116, - "loss": 2.8266, + "learning_rate": 2.3671875e-05, + "loss": 0.2585, "step": 966 }, { "epoch": 1.08, - "learning_rate": 0.00023644301470588232, - "loss": 3.1199, + "learning_rate": 2.3658088235294118e-05, + "loss": 0.3058, "step": 967 }, { "epoch": 1.09, - "learning_rate": 0.00023630514705882353, - "loss": 2.8182, + "learning_rate": 2.3644301470588235e-05, + "loss": 0.2314, "step": 968 }, { "epoch": 1.09, - "learning_rate": 0.00023616727941176468, - "loss": 2.8015, + "learning_rate": 2.3630514705882356e-05, + "loss": 0.2251, "step": 969 }, { "epoch": 1.09, - "learning_rate": 0.00023602941176470587, - "loss": 2.8331, + "learning_rate": 2.3616727941176473e-05, + "loss": 0.2519, "step": 970 }, { "epoch": 1.09, - "learning_rate": 0.00023589154411764703, - "loss": 2.8218, + "learning_rate": 2.360294117647059e-05, + "loss": 0.2469, "step": 971 }, { "epoch": 1.09, - "learning_rate": 0.00023575367647058824, - "loss": 2.827, + "learning_rate": 2.3589154411764707e-05, + "loss": 0.2225, "step": 972 }, { "epoch": 1.09, - "learning_rate": 0.0002356158088235294, - "loss": 3.0293, + "learning_rate": 2.3575367647058827e-05, + "loss": 0.2731, "step": 973 }, { "epoch": 1.09, - "learning_rate": 0.00023547794117647058, - "loss": 3.9145, + "learning_rate": 2.356158088235294e-05, + "loss": 0.2908, "step": 974 }, { "epoch": 1.09, - "learning_rate": 0.00023534007352941173, - "loss": 2.8178, + "learning_rate": 2.3547794117647058e-05, + "loss": 0.2678, "step": 975 }, { "epoch": 1.09, - "learning_rate": 0.00023520220588235294, - "loss": 2.8056, + "learning_rate": 2.3534007352941175e-05, + "loss": 0.2346, "step": 976 }, { "epoch": 1.1, - "learning_rate": 0.0002350643382352941, - "loss": 2.8254, + "learning_rate": 2.3520220588235295e-05, + "loss": 0.2534, "step": 977 }, { "epoch": 1.1, - "learning_rate": 0.00023492647058823528, - "loss": 2.8209, + "learning_rate": 2.3506433823529412e-05, + "loss": 0.2763, "step": 978 }, { "epoch": 1.1, - "learning_rate": 0.00023478860294117644, - "loss": 2.9854, + "learning_rate": 2.349264705882353e-05, + "loss": 0.3046, "step": 979 }, { "epoch": 1.1, - "learning_rate": 0.00023465073529411765, - "loss": 4.3129, + "learning_rate": 2.3478860294117646e-05, + "loss": 0.2603, "step": 980 }, { "epoch": 1.1, - "learning_rate": 0.0002345128676470588, - "loss": 2.8257, + "learning_rate": 2.3465073529411766e-05, + "loss": 0.2363, "step": 981 }, { "epoch": 1.1, - "learning_rate": 0.000234375, - "loss": 2.8334, + "learning_rate": 2.3451286764705883e-05, + "loss": 0.2848, "step": 982 }, { "epoch": 1.1, - "learning_rate": 0.00023423713235294115, - "loss": 2.8243, + "learning_rate": 2.34375e-05, + "loss": 0.254, "step": 983 }, { "epoch": 1.1, - "learning_rate": 0.0002340992647058823, - "loss": 2.8288, + "learning_rate": 2.3423713235294118e-05, + "loss": 0.2538, "step": 984 }, { "epoch": 1.1, - "learning_rate": 0.00023396139705882352, - "loss": 2.8774, + "learning_rate": 2.3409926470588235e-05, + "loss": 0.2791, "step": 985 }, { "epoch": 1.11, - "learning_rate": 0.0002338235294117647, - "loss": 5.0103, + "learning_rate": 2.3396139705882355e-05, + "loss": 0.3154, "step": 986 }, { "epoch": 1.11, - "learning_rate": 0.00023368566176470586, - "loss": 2.816, + "learning_rate": 2.3382352941176472e-05, + "loss": 0.2656, "step": 987 }, { "epoch": 1.11, - "learning_rate": 0.000233547794117647, - "loss": 2.8262, + "learning_rate": 2.336856617647059e-05, + "loss": 0.2764, "step": 988 }, { "epoch": 1.11, - "learning_rate": 0.00023340992647058822, - "loss": 2.8218, + "learning_rate": 2.3354779411764706e-05, + "loss": 0.272, "step": 989 }, { "epoch": 1.11, - "learning_rate": 0.0002332720588235294, - "loss": 2.837, + "learning_rate": 2.3340992647058826e-05, + "loss": 0.2748, "step": 990 }, { "epoch": 1.11, - "learning_rate": 0.00023313419117647056, - "loss": 2.8627, + "learning_rate": 2.3327205882352943e-05, + "loss": 0.2816, "step": 991 }, { "epoch": 1.11, - "learning_rate": 0.00023299632352941172, - "loss": 3.0469, + "learning_rate": 2.331341911764706e-05, + "loss": 0.3058, "step": 992 }, { "epoch": 1.11, - "learning_rate": 0.00023285845588235293, - "loss": 2.837, + "learning_rate": 2.3299632352941177e-05, + "loss": 0.2429, "step": 993 }, { "epoch": 1.11, - "learning_rate": 0.00023272058823529411, - "loss": 2.8464, + "learning_rate": 2.3285845588235294e-05, + "loss": 0.2367, "step": 994 }, { "epoch": 1.12, - "learning_rate": 0.00023258272058823527, - "loss": 2.8608, + "learning_rate": 2.327205882352941e-05, + "loss": 0.2617, "step": 995 }, { "epoch": 1.12, - "learning_rate": 0.00023244485294117643, - "loss": 2.8276, + "learning_rate": 2.325827205882353e-05, + "loss": 0.2639, "step": 996 }, { "epoch": 1.12, - "learning_rate": 0.00023230698529411764, - "loss": 2.8159, + "learning_rate": 2.3244485294117645e-05, + "loss": 0.256, "step": 997 }, { "epoch": 1.12, - "learning_rate": 0.00023216911764705882, - "loss": 2.9018, + "learning_rate": 2.3230698529411766e-05, + "loss": 0.2889, "step": 998 }, { "epoch": 1.12, - "learning_rate": 0.00023203124999999998, - "loss": 3.5101, + "learning_rate": 2.3216911764705883e-05, + "loss": 0.2883, "step": 999 }, { "epoch": 1.12, - "learning_rate": 0.00023189338235294113, - "loss": 2.8211, + "learning_rate": 2.3203125e-05, + "loss": 0.2378, "step": 1000 }, { "epoch": 1.12, - "eval_loss": 3.9629738330841064, - "eval_runtime": 48.0612, - "eval_samples_per_second": 56.241, - "eval_steps_per_second": 0.895, - "eval_wer": 1.0, + "eval_loss": 0.2912798225879669, + "eval_runtime": 47.4101, + "eval_samples_per_second": 57.013, + "eval_steps_per_second": 0.907, + "eval_wer": 0.26129554060512483, "step": 1000 }, { "epoch": 1.12, - "learning_rate": 0.00023175551470588232, - "loss": 2.8258, + "learning_rate": 2.3189338235294117e-05, + "loss": 0.2346, "step": 1001 }, { "epoch": 1.12, - "learning_rate": 0.00023161764705882353, - "loss": 2.8384, + "learning_rate": 2.3175551470588234e-05, + "loss": 0.261, "step": 1002 }, { "epoch": 1.12, - "learning_rate": 0.00023147977941176469, - "loss": 2.8134, + "learning_rate": 2.3161764705882354e-05, + "loss": 0.2189, "step": 1003 }, { "epoch": 1.13, - "learning_rate": 0.00023134191176470587, - "loss": 2.9699, + "learning_rate": 2.314797794117647e-05, + "loss": 0.2699, "step": 1004 }, { "epoch": 1.13, - "learning_rate": 0.00023120404411764703, - "loss": 4.5316, + "learning_rate": 2.313419117647059e-05, + "loss": 0.3512, "step": 1005 }, { "epoch": 1.13, - "learning_rate": 0.00023106617647058824, - "loss": 2.8339, + "learning_rate": 2.3120404411764705e-05, + "loss": 0.2981, "step": 1006 }, { "epoch": 1.13, - "learning_rate": 0.0002309283088235294, - "loss": 2.8401, + "learning_rate": 2.3106617647058826e-05, + "loss": 0.2598, "step": 1007 }, { "epoch": 1.13, - "learning_rate": 0.00023079044117647058, - "loss": 2.8422, + "learning_rate": 2.3092830882352943e-05, + "loss": 0.2071, "step": 1008 }, { "epoch": 1.13, - "learning_rate": 0.00023065257352941173, - "loss": 2.8396, + "learning_rate": 2.307904411764706e-05, + "loss": 0.2345, "step": 1009 }, { "epoch": 1.13, - "learning_rate": 0.00023051470588235294, - "loss": 2.8329, + "learning_rate": 2.3065257352941177e-05, + "loss": 0.2501, "step": 1010 }, { "epoch": 1.13, - "learning_rate": 0.0002303768382352941, - "loss": 4.012, + "learning_rate": 2.3051470588235297e-05, + "loss": 0.3078, "step": 1011 }, { "epoch": 1.13, - "learning_rate": 0.00023023897058823528, - "loss": 2.8013, + "learning_rate": 2.3037683823529414e-05, + "loss": 0.2385, "step": 1012 }, { "epoch": 1.14, - "learning_rate": 0.00023010110294117644, - "loss": 2.8108, + "learning_rate": 2.302389705882353e-05, + "loss": 0.2278, "step": 1013 }, { "epoch": 1.14, - "learning_rate": 0.00022996323529411765, - "loss": 2.8035, + "learning_rate": 2.3010110294117648e-05, + "loss": 0.2444, "step": 1014 }, { "epoch": 1.14, - "learning_rate": 0.0002298253676470588, - "loss": 2.8292, + "learning_rate": 2.2996323529411765e-05, + "loss": 0.2852, "step": 1015 }, { "epoch": 1.14, - "learning_rate": 0.0002296875, - "loss": 2.8494, + "learning_rate": 2.2982536764705882e-05, + "loss": 0.2469, "step": 1016 }, { "epoch": 1.14, - "learning_rate": 0.00022954963235294115, - "loss": 3.0759, + "learning_rate": 2.296875e-05, + "loss": 0.255, "step": 1017 }, { "epoch": 1.14, - "learning_rate": 0.0002294117647058823, - "loss": 2.8072, + "learning_rate": 2.2954963235294116e-05, + "loss": 0.2599, "step": 1018 }, { "epoch": 1.14, - "learning_rate": 0.00022927389705882352, - "loss": 2.8102, + "learning_rate": 2.2941176470588233e-05, + "loss": 0.2375, "step": 1019 }, { "epoch": 1.14, - "learning_rate": 0.0002291360294117647, - "loss": 2.8323, + "learning_rate": 2.2927389705882354e-05, + "loss": 0.2224, "step": 1020 }, { "epoch": 1.14, - "learning_rate": 0.00022899816176470586, - "loss": 2.83, + "learning_rate": 2.291360294117647e-05, + "loss": 0.2172, "step": 1021 }, { "epoch": 1.15, - "learning_rate": 0.000228860294117647, - "loss": 2.8217, + "learning_rate": 2.2899816176470588e-05, + "loss": 0.2715, "step": 1022 }, { "epoch": 1.15, - "learning_rate": 0.00022872242647058822, - "loss": 2.9414, + "learning_rate": 2.2886029411764705e-05, + "loss": 0.3156, "step": 1023 }, { "epoch": 1.15, - "learning_rate": 0.0002285845588235294, - "loss": 3.8028, + "learning_rate": 2.2872242647058825e-05, + "loss": 0.2919, "step": 1024 }, { "epoch": 1.15, - "learning_rate": 0.00022844669117647056, - "loss": 2.8098, + "learning_rate": 2.2858455882352942e-05, + "loss": 0.2352, "step": 1025 }, { "epoch": 1.15, - "learning_rate": 0.00022830882352941172, - "loss": 2.8036, + "learning_rate": 2.284466911764706e-05, + "loss": 0.2315, "step": 1026 }, { "epoch": 1.15, - "learning_rate": 0.00022817095588235293, - "loss": 2.827, + "learning_rate": 2.2830882352941176e-05, + "loss": 0.2687, "step": 1027 }, { "epoch": 1.15, - "learning_rate": 0.00022803308823529412, - "loss": 2.8147, + "learning_rate": 2.2817095588235297e-05, + "loss": 0.2516, "step": 1028 }, { "epoch": 1.15, - "learning_rate": 0.00022789522058823527, - "loss": 2.9508, + "learning_rate": 2.2803308823529414e-05, + "loss": 0.2776, "step": 1029 }, { "epoch": 1.15, - "learning_rate": 0.00022775735294117643, - "loss": 4.8558, + "learning_rate": 2.278952205882353e-05, + "loss": 0.2598, "step": 1030 }, { "epoch": 1.16, - "learning_rate": 0.00022761948529411764, - "loss": 2.8181, + "learning_rate": 2.2775735294117648e-05, + "loss": 0.244, "step": 1031 }, { "epoch": 1.16, - "learning_rate": 0.00022748161764705882, - "loss": 2.8223, + "learning_rate": 2.2761948529411768e-05, + "loss": 0.2221, "step": 1032 }, { "epoch": 1.16, - "learning_rate": 0.00022734374999999998, - "loss": 2.8263, + "learning_rate": 2.2748161764705885e-05, + "loss": 0.2291, "step": 1033 }, { "epoch": 1.16, - "learning_rate": 0.00022720588235294114, - "loss": 2.8214, + "learning_rate": 2.2734375000000002e-05, + "loss": 0.264, "step": 1034 }, { "epoch": 1.16, - "learning_rate": 0.00022706801470588232, - "loss": 2.8323, + "learning_rate": 2.272058823529412e-05, + "loss": 0.2486, "step": 1035 }, { "epoch": 1.16, - "learning_rate": 0.00022693014705882353, - "loss": 4.4247, + "learning_rate": 2.2706801470588233e-05, + "loss": 0.2695, "step": 1036 }, { "epoch": 1.16, - "learning_rate": 0.0002267922794117647, - "loss": 2.7966, + "learning_rate": 2.2693014705882353e-05, + "loss": 0.2152, "step": 1037 }, { "epoch": 1.16, - "learning_rate": 0.00022665441176470584, - "loss": 2.8172, + "learning_rate": 2.267922794117647e-05, + "loss": 0.1923, "step": 1038 }, { "epoch": 1.16, - "learning_rate": 0.00022651654411764703, - "loss": 2.8055, + "learning_rate": 2.2665441176470587e-05, + "loss": 0.2313, "step": 1039 }, { "epoch": 1.17, - "learning_rate": 0.00022637867647058824, - "loss": 2.7938, + "learning_rate": 2.2651654411764704e-05, + "loss": 0.2538, "step": 1040 }, { "epoch": 1.17, - "learning_rate": 0.0002262408088235294, - "loss": 2.861, + "learning_rate": 2.2637867647058825e-05, + "loss": 0.2416, "step": 1041 }, { "epoch": 1.17, - "learning_rate": 0.00022610294117647055, - "loss": 3.1593, + "learning_rate": 2.262408088235294e-05, + "loss": 0.2661, "step": 1042 }, { "epoch": 1.17, - "learning_rate": 0.00022596507352941174, - "loss": 2.8154, + "learning_rate": 2.261029411764706e-05, + "loss": 0.2571, "step": 1043 }, { "epoch": 1.17, - "learning_rate": 0.00022582720588235295, - "loss": 2.7972, + "learning_rate": 2.2596507352941176e-05, + "loss": 0.248, "step": 1044 }, { "epoch": 1.17, - "learning_rate": 0.0002256893382352941, - "loss": 2.822, + "learning_rate": 2.2582720588235296e-05, + "loss": 0.2975, "step": 1045 }, { "epoch": 1.17, - "learning_rate": 0.00022555147058823526, - "loss": 2.8063, + "learning_rate": 2.2568933823529413e-05, + "loss": 0.215, "step": 1046 }, { "epoch": 1.17, - "learning_rate": 0.00022541360294117644, - "loss": 2.8125, + "learning_rate": 2.255514705882353e-05, + "loss": 0.2243, "step": 1047 }, { "epoch": 1.17, - "learning_rate": 0.00022527573529411765, - "loss": 2.9599, + "learning_rate": 2.2541360294117647e-05, + "loss": 0.2256, "step": 1048 }, { "epoch": 1.18, - "learning_rate": 0.0002251378676470588, - "loss": 3.6394, + "learning_rate": 2.2527573529411767e-05, + "loss": 0.2775, "step": 1049 }, { "epoch": 1.18, - "learning_rate": 0.000225, - "loss": 2.8053, + "learning_rate": 2.2513786764705884e-05, + "loss": 0.2321, "step": 1050 }, { "epoch": 1.18, - "learning_rate": 0.00022486213235294115, - "loss": 2.8225, + "learning_rate": 2.25e-05, + "loss": 0.2333, "step": 1051 }, { "epoch": 1.18, - "learning_rate": 0.0002247242647058823, - "loss": 2.8058, + "learning_rate": 2.248621323529412e-05, + "loss": 0.256, "step": 1052 }, { "epoch": 1.18, - "learning_rate": 0.00022458639705882352, - "loss": 2.8089, + "learning_rate": 2.2472426470588235e-05, + "loss": 0.2208, "step": 1053 }, { "epoch": 1.18, - "learning_rate": 0.0002244485294117647, - "loss": 2.8857, + "learning_rate": 2.2458639705882356e-05, + "loss": 0.2295, "step": 1054 }, { "epoch": 1.18, - "learning_rate": 0.00022431066176470586, - "loss": 3.9711, + "learning_rate": 2.2444852941176473e-05, + "loss": 0.2797, "step": 1055 }, { "epoch": 1.18, - "learning_rate": 0.00022417279411764701, - "loss": 2.8173, + "learning_rate": 2.243106617647059e-05, + "loss": 0.2311, "step": 1056 }, { "epoch": 1.18, - "learning_rate": 0.00022403492647058823, - "loss": 2.8139, + "learning_rate": 2.2417279411764703e-05, + "loss": 0.2273, "step": 1057 }, { "epoch": 1.19, - "learning_rate": 0.0002238970588235294, - "loss": 2.8219, + "learning_rate": 2.2403492647058824e-05, + "loss": 0.224, "step": 1058 }, { "epoch": 1.19, - "learning_rate": 0.00022375919117647057, - "loss": 2.8242, + "learning_rate": 2.238970588235294e-05, + "loss": 0.2212, "step": 1059 }, { "epoch": 1.19, - "learning_rate": 0.00022362132352941172, - "loss": 2.8316, + "learning_rate": 2.2375919117647058e-05, + "loss": 0.232, "step": 1060 }, { "epoch": 1.19, - "learning_rate": 0.00022348345588235293, - "loss": 4.0957, + "learning_rate": 2.2362132352941175e-05, + "loss": 0.311, "step": 1061 }, { "epoch": 1.19, - "learning_rate": 0.00022334558823529412, - "loss": 2.819, + "learning_rate": 2.2348345588235295e-05, + "loss": 0.2265, "step": 1062 }, { "epoch": 1.19, - "learning_rate": 0.00022320772058823527, - "loss": 2.8442, + "learning_rate": 2.2334558823529412e-05, + "loss": 0.2546, "step": 1063 }, { "epoch": 1.19, - "learning_rate": 0.00022306985294117643, - "loss": 2.8292, + "learning_rate": 2.232077205882353e-05, + "loss": 0.2628, "step": 1064 }, { "epoch": 1.19, - "learning_rate": 0.00022293198529411764, - "loss": 2.8207, + "learning_rate": 2.2306985294117646e-05, + "loss": 0.2245, "step": 1065 }, { "epoch": 1.2, - "learning_rate": 0.00022279411764705882, - "loss": 2.8315, + "learning_rate": 2.2293198529411767e-05, + "loss": 0.241, "step": 1066 }, { "epoch": 1.2, - "learning_rate": 0.00022265624999999998, - "loss": 3.3618, + "learning_rate": 2.2279411764705884e-05, + "loss": 0.2979, "step": 1067 }, { "epoch": 1.2, - "learning_rate": 0.00022251838235294114, - "loss": 2.8103, + "learning_rate": 2.2265625e-05, + "loss": 0.2192, "step": 1068 }, { "epoch": 1.2, - "learning_rate": 0.00022238051470588232, - "loss": 2.8254, + "learning_rate": 2.2251838235294118e-05, + "loss": 0.2313, "step": 1069 }, { "epoch": 1.2, - "learning_rate": 0.00022224264705882353, - "loss": 2.8238, + "learning_rate": 2.2238051470588235e-05, + "loss": 0.2606, "step": 1070 }, { "epoch": 1.2, - "learning_rate": 0.0002221047794117647, - "loss": 2.8175, + "learning_rate": 2.2224264705882355e-05, + "loss": 0.2109, "step": 1071 }, { "epoch": 1.2, - "learning_rate": 0.00022196691176470585, - "loss": 2.8489, + "learning_rate": 2.2210477941176472e-05, + "loss": 0.2594, "step": 1072 }, { "epoch": 1.2, - "learning_rate": 0.00022182904411764703, - "loss": 2.9228, + "learning_rate": 2.219669117647059e-05, + "loss": 0.275, "step": 1073 }, { "epoch": 1.2, - "learning_rate": 0.00022169117647058824, - "loss": 3.5095, + "learning_rate": 2.2182904411764706e-05, + "loss": 0.2321, "step": 1074 }, { "epoch": 1.21, - "learning_rate": 0.0002215533088235294, - "loss": 2.8216, + "learning_rate": 2.2169117647058827e-05, + "loss": 0.2442, "step": 1075 }, { "epoch": 1.21, - "learning_rate": 0.00022141544117647055, - "loss": 2.8394, + "learning_rate": 2.2155330882352944e-05, + "loss": 0.2007, "step": 1076 }, { "epoch": 1.21, - "learning_rate": 0.00022127757352941174, - "loss": 2.8622, + "learning_rate": 2.2141544117647057e-05, + "loss": 0.2426, "step": 1077 }, { "epoch": 1.21, - "learning_rate": 0.00022113970588235295, - "loss": 2.8253, + "learning_rate": 2.2127757352941174e-05, + "loss": 0.2231, "step": 1078 }, { "epoch": 1.21, - "learning_rate": 0.0002210018382352941, - "loss": 2.9237, + "learning_rate": 2.2113970588235295e-05, + "loss": 0.2558, "step": 1079 }, { "epoch": 1.21, - "learning_rate": 0.00022086397058823526, - "loss": 4.3803, + "learning_rate": 2.2100183823529412e-05, + "loss": 0.3026, "step": 1080 }, { "epoch": 1.21, - "learning_rate": 0.00022072610294117644, - "loss": 2.8123, + "learning_rate": 2.208639705882353e-05, + "loss": 0.2013, "step": 1081 }, { "epoch": 1.21, - "learning_rate": 0.00022058823529411765, - "loss": 2.8145, + "learning_rate": 2.2072610294117646e-05, + "loss": 0.2405, "step": 1082 }, { "epoch": 1.21, - "learning_rate": 0.0002204503676470588, - "loss": 2.8093, + "learning_rate": 2.2058823529411766e-05, + "loss": 0.1908, "step": 1083 }, { "epoch": 1.22, - "learning_rate": 0.00022031249999999997, - "loss": 2.8008, + "learning_rate": 2.2045036764705883e-05, + "loss": 0.2009, "step": 1084 }, { "epoch": 1.22, - "learning_rate": 0.00022017463235294115, - "loss": 2.9014, + "learning_rate": 2.203125e-05, + "loss": 0.2734, "step": 1085 }, { "epoch": 1.22, - "learning_rate": 0.0002200367647058823, - "loss": 5.0639, + "learning_rate": 2.2017463235294117e-05, + "loss": 0.2668, "step": 1086 }, { "epoch": 1.22, - "learning_rate": 0.00021989889705882352, - "loss": 2.8234, + "learning_rate": 2.2003676470588234e-05, + "loss": 0.2279, "step": 1087 }, { "epoch": 1.22, - "learning_rate": 0.00021976102941176468, - "loss": 2.8099, + "learning_rate": 2.1989889705882355e-05, + "loss": 0.2026, "step": 1088 }, { "epoch": 1.22, - "learning_rate": 0.00021962316176470586, - "loss": 2.8047, + "learning_rate": 2.197610294117647e-05, + "loss": 0.2478, "step": 1089 }, { "epoch": 1.22, - "learning_rate": 0.00021948529411764702, - "loss": 2.7992, + "learning_rate": 2.196231617647059e-05, + "loss": 0.2038, "step": 1090 }, { "epoch": 1.22, - "learning_rate": 0.00021934742647058823, - "loss": 2.8302, + "learning_rate": 2.1948529411764706e-05, + "loss": 0.3116, "step": 1091 }, { "epoch": 1.22, - "learning_rate": 0.00021920955882352938, - "loss": 3.1337, + "learning_rate": 2.1934742647058826e-05, + "loss": 0.2413, "step": 1092 }, { "epoch": 1.23, - "learning_rate": 0.00021907169117647057, - "loss": 2.8205, + "learning_rate": 2.1920955882352943e-05, + "loss": 0.2521, "step": 1093 }, { "epoch": 1.23, - "learning_rate": 0.00021893382352941172, - "loss": 2.8086, + "learning_rate": 2.190716911764706e-05, + "loss": 0.234, "step": 1094 }, { "epoch": 1.23, - "learning_rate": 0.00021879595588235293, - "loss": 2.8257, + "learning_rate": 2.1893382352941177e-05, + "loss": 0.2407, "step": 1095 }, { "epoch": 1.23, - "learning_rate": 0.0002186580882352941, - "loss": 2.8185, + "learning_rate": 2.1879595588235297e-05, + "loss": 0.218, "step": 1096 }, { "epoch": 1.23, - "learning_rate": 0.00021852022058823527, - "loss": 2.8128, + "learning_rate": 2.1865808823529415e-05, + "loss": 0.2188, "step": 1097 }, { "epoch": 1.23, - "learning_rate": 0.00021838235294117643, - "loss": 3.0954, + "learning_rate": 2.1852022058823528e-05, + "loss": 0.2826, "step": 1098 }, { "epoch": 1.23, - "learning_rate": 0.00021824448529411764, - "loss": 4.0505, + "learning_rate": 2.1838235294117645e-05, + "loss": 0.3093, "step": 1099 }, { "epoch": 1.23, - "learning_rate": 0.00021810661764705883, - "loss": 2.8209, + "learning_rate": 2.1824448529411766e-05, + "loss": 0.2516, "step": 1100 }, { "epoch": 1.23, - "eval_loss": 3.9604644775390625, - "eval_runtime": 48.134, - "eval_samples_per_second": 56.156, - "eval_steps_per_second": 0.893, - "eval_wer": 1.0, + "eval_loss": 0.2701849043369293, + "eval_runtime": 47.3157, + "eval_samples_per_second": 57.127, + "eval_steps_per_second": 0.909, + "eval_wer": 0.24515642807249732, "step": 1100 }, { "epoch": 1.23, - "learning_rate": 0.00021796874999999998, - "loss": 2.8131, + "learning_rate": 2.1810661764705883e-05, + "loss": 0.2001, "step": 1101 }, { "epoch": 1.24, - "learning_rate": 0.00021783088235294114, - "loss": 2.814, + "learning_rate": 2.1796875e-05, + "loss": 0.2361, "step": 1102 }, { "epoch": 1.24, - "learning_rate": 0.00021769301470588232, - "loss": 2.8251, + "learning_rate": 2.1783088235294117e-05, + "loss": 0.2515, "step": 1103 }, { "epoch": 1.24, - "learning_rate": 0.00021755514705882353, - "loss": 2.9788, + "learning_rate": 2.1769301470588234e-05, + "loss": 0.2761, "step": 1104 }, { "epoch": 1.24, - "learning_rate": 0.0002174172794117647, - "loss": 4.9852, + "learning_rate": 2.1755514705882354e-05, + "loss": 0.3218, "step": 1105 }, { "epoch": 1.24, - "learning_rate": 0.00021727941176470585, - "loss": 2.8101, + "learning_rate": 2.174172794117647e-05, + "loss": 0.2119, "step": 1106 }, { "epoch": 1.24, - "learning_rate": 0.00021714154411764703, - "loss": 2.8127, + "learning_rate": 2.1727941176470588e-05, + "loss": 0.2282, "step": 1107 }, { "epoch": 1.24, - "learning_rate": 0.00021700367647058824, - "loss": 2.8233, + "learning_rate": 2.1714154411764705e-05, + "loss": 0.2319, "step": 1108 }, { "epoch": 1.24, - "learning_rate": 0.0002168658088235294, - "loss": 2.8335, + "learning_rate": 2.1700367647058825e-05, + "loss": 0.2311, "step": 1109 }, { "epoch": 1.24, - "learning_rate": 0.00021672794117647055, - "loss": 2.8434, + "learning_rate": 2.1686580882352942e-05, + "loss": 0.2428, "step": 1110 }, { "epoch": 1.25, - "learning_rate": 0.00021659007352941174, - "loss": 3.6762, + "learning_rate": 2.167279411764706e-05, + "loss": 0.2379, "step": 1111 }, { "epoch": 1.25, - "learning_rate": 0.00021645220588235295, - "loss": 2.8188, + "learning_rate": 2.1659007352941176e-05, + "loss": 0.2169, "step": 1112 }, { "epoch": 1.25, - "learning_rate": 0.0002163143382352941, - "loss": 2.8235, + "learning_rate": 2.1645220588235297e-05, + "loss": 0.247, "step": 1113 }, { "epoch": 1.25, - "learning_rate": 0.00021617647058823526, - "loss": 2.8122, + "learning_rate": 2.1631433823529414e-05, + "loss": 0.242, "step": 1114 }, { "epoch": 1.25, - "learning_rate": 0.00021603860294117645, - "loss": 2.822, + "learning_rate": 2.161764705882353e-05, + "loss": 0.219, "step": 1115 }, { "epoch": 1.25, - "learning_rate": 0.00021590073529411766, - "loss": 2.8153, + "learning_rate": 2.1603860294117648e-05, + "loss": 0.2054, "step": 1116 }, { "epoch": 1.25, - "learning_rate": 0.0002157628676470588, - "loss": 3.3701, + "learning_rate": 2.159007352941177e-05, + "loss": 0.2543, "step": 1117 }, { "epoch": 1.25, - "learning_rate": 0.00021562499999999997, - "loss": 2.8099, + "learning_rate": 2.1576286764705885e-05, + "loss": 0.2266, "step": 1118 }, { "epoch": 1.25, - "learning_rate": 0.00021548713235294115, - "loss": 2.8291, + "learning_rate": 2.15625e-05, + "loss": 0.2346, "step": 1119 }, { "epoch": 1.26, - "learning_rate": 0.0002153492647058823, - "loss": 2.8419, + "learning_rate": 2.1548713235294116e-05, + "loss": 0.1964, "step": 1120 }, { "epoch": 1.26, - "learning_rate": 0.00021521139705882352, - "loss": 2.8303, + "learning_rate": 2.1534926470588233e-05, + "loss": 0.2639, "step": 1121 }, { "epoch": 1.26, - "learning_rate": 0.00021507352941176468, - "loss": 2.8122, + "learning_rate": 2.1521139705882353e-05, + "loss": 0.189, "step": 1122 }, { "epoch": 1.26, - "learning_rate": 0.00021493566176470586, - "loss": 3.085, + "learning_rate": 2.150735294117647e-05, + "loss": 0.2992, "step": 1123 }, { "epoch": 1.26, - "learning_rate": 0.00021479779411764702, - "loss": 3.5107, + "learning_rate": 2.1493566176470587e-05, + "loss": 0.2458, "step": 1124 }, { "epoch": 1.26, - "learning_rate": 0.00021465992647058823, - "loss": 2.8383, + "learning_rate": 2.1479779411764704e-05, + "loss": 0.2158, "step": 1125 }, { "epoch": 1.26, - "learning_rate": 0.00021452205882352938, - "loss": 2.8524, + "learning_rate": 2.1465992647058825e-05, + "loss": 0.2144, "step": 1126 }, { "epoch": 1.26, - "learning_rate": 0.00021438419117647057, - "loss": 2.832, + "learning_rate": 2.1452205882352942e-05, + "loss": 0.2321, "step": 1127 }, { "epoch": 1.26, - "learning_rate": 0.00021424632352941172, - "loss": 2.8436, + "learning_rate": 2.143841911764706e-05, + "loss": 0.2616, "step": 1128 }, { "epoch": 1.27, - "learning_rate": 0.00021410845588235294, - "loss": 2.8484, + "learning_rate": 2.1424632352941176e-05, + "loss": 0.2422, "step": 1129 }, { "epoch": 1.27, - "learning_rate": 0.0002139705882352941, - "loss": 4.0795, + "learning_rate": 2.1410845588235296e-05, + "loss": 0.2533, "step": 1130 }, { "epoch": 1.27, - "learning_rate": 0.00021383272058823528, - "loss": 2.794, + "learning_rate": 2.1397058823529413e-05, + "loss": 0.2091, "step": 1131 }, { "epoch": 1.27, - "learning_rate": 0.00021369485294117643, - "loss": 2.813, + "learning_rate": 2.138327205882353e-05, + "loss": 0.2032, "step": 1132 }, { "epoch": 1.27, - "learning_rate": 0.00021355698529411764, - "loss": 2.8287, + "learning_rate": 2.1369485294117647e-05, + "loss": 0.2012, "step": 1133 }, { "epoch": 1.27, - "learning_rate": 0.0002134191176470588, - "loss": 2.8393, + "learning_rate": 2.1355698529411768e-05, + "loss": 0.2379, "step": 1134 }, { "epoch": 1.27, - "learning_rate": 0.00021328124999999998, - "loss": 2.8642, + "learning_rate": 2.1341911764705885e-05, + "loss": 0.2459, "step": 1135 }, { "epoch": 1.27, - "learning_rate": 0.00021314338235294114, - "loss": 4.2219, + "learning_rate": 2.1328125000000002e-05, + "loss": 0.2756, "step": 1136 }, { "epoch": 1.27, - "learning_rate": 0.00021300551470588232, - "loss": 2.8102, + "learning_rate": 2.131433823529412e-05, + "loss": 0.2147, "step": 1137 }, { "epoch": 1.28, - "learning_rate": 0.0002128676470588235, - "loss": 2.804, + "learning_rate": 2.1300551470588236e-05, + "loss": 0.2117, "step": 1138 }, { "epoch": 1.28, - "learning_rate": 0.0002127297794117647, - "loss": 2.8069, + "learning_rate": 2.1286764705882353e-05, + "loss": 0.2163, "step": 1139 }, { "epoch": 1.28, - "learning_rate": 0.00021259191176470585, - "loss": 2.824, + "learning_rate": 2.127297794117647e-05, + "loss": 0.2279, "step": 1140 }, { "epoch": 1.28, - "learning_rate": 0.00021245404411764703, - "loss": 2.8258, + "learning_rate": 2.1259191176470587e-05, + "loss": 0.2172, "step": 1141 }, { "epoch": 1.28, - "learning_rate": 0.00021231617647058822, - "loss": 3.4254, + "learning_rate": 2.1245404411764704e-05, + "loss": 0.2457, "step": 1142 }, { "epoch": 1.28, - "learning_rate": 0.0002121783088235294, - "loss": 2.822, + "learning_rate": 2.1231617647058824e-05, + "loss": 0.2281, "step": 1143 }, { "epoch": 1.28, - "learning_rate": 0.00021204044117647056, - "loss": 2.8082, + "learning_rate": 2.121783088235294e-05, + "loss": 0.2293, "step": 1144 }, { "epoch": 1.28, - "learning_rate": 0.00021190257352941174, - "loss": 2.8096, + "learning_rate": 2.1204044117647058e-05, + "loss": 0.2228, "step": 1145 }, { "epoch": 1.28, - "learning_rate": 0.00021176470588235295, - "loss": 2.8373, + "learning_rate": 2.1190257352941175e-05, + "loss": 0.2126, "step": 1146 }, { "epoch": 1.29, - "learning_rate": 0.0002116268382352941, - "loss": 2.8306, + "learning_rate": 2.1176470588235296e-05, + "loss": 0.2816, "step": 1147 }, { "epoch": 1.29, - "learning_rate": 0.00021148897058823526, - "loss": 2.8804, + "learning_rate": 2.1162683823529413e-05, + "loss": 0.2882, "step": 1148 }, { "epoch": 1.29, - "learning_rate": 0.00021135110294117645, - "loss": 3.564, + "learning_rate": 2.114889705882353e-05, + "loss": 0.2213, "step": 1149 }, { "epoch": 1.29, - "learning_rate": 0.00021121323529411766, - "loss": 2.828, + "learning_rate": 2.1135110294117647e-05, + "loss": 0.2148, "step": 1150 }, { "epoch": 1.29, - "learning_rate": 0.00021107536764705881, - "loss": 2.8212, + "learning_rate": 2.1121323529411767e-05, + "loss": 0.2166, "step": 1151 }, { "epoch": 1.29, - "learning_rate": 0.00021093749999999997, - "loss": 2.8308, + "learning_rate": 2.1107536764705884e-05, + "loss": 0.2396, "step": 1152 }, { "epoch": 1.29, - "learning_rate": 0.00021079963235294115, - "loss": 2.8401, + "learning_rate": 2.109375e-05, + "loss": 0.2141, "step": 1153 }, { "epoch": 1.29, - "learning_rate": 0.0002106617647058823, - "loss": 2.9003, + "learning_rate": 2.1079963235294118e-05, + "loss": 0.2739, "step": 1154 }, { "epoch": 1.29, - "learning_rate": 0.00021052389705882352, - "loss": 4.804, + "learning_rate": 2.1066176470588235e-05, + "loss": 0.2982, "step": 1155 }, { "epoch": 1.3, - "learning_rate": 0.00021038602941176468, - "loss": 2.8057, + "learning_rate": 2.1052389705882356e-05, + "loss": 0.212, "step": 1156 }, { "epoch": 1.3, - "learning_rate": 0.00021024816176470586, - "loss": 2.8042, + "learning_rate": 2.1038602941176473e-05, + "loss": 0.2094, "step": 1157 }, { "epoch": 1.3, - "learning_rate": 0.00021011029411764702, - "loss": 2.8102, + "learning_rate": 2.102481617647059e-05, + "loss": 0.2012, "step": 1158 }, { "epoch": 1.3, - "learning_rate": 0.00020997242647058823, - "loss": 2.8016, + "learning_rate": 2.1011029411764707e-05, + "loss": 0.2113, "step": 1159 }, { "epoch": 1.3, - "learning_rate": 0.00020983455882352939, - "loss": 2.8615, + "learning_rate": 2.0997242647058824e-05, + "loss": 0.259, "step": 1160 }, { "epoch": 1.3, - "learning_rate": 0.00020969669117647057, - "loss": 4.1654, + "learning_rate": 2.098345588235294e-05, + "loss": 0.3951, "step": 1161 }, { "epoch": 1.3, - "learning_rate": 0.00020955882352941173, - "loss": 2.8088, + "learning_rate": 2.0969669117647058e-05, + "loss": 0.1926, "step": 1162 }, { "epoch": 1.3, - "learning_rate": 0.00020942095588235294, - "loss": 2.8153, + "learning_rate": 2.0955882352941175e-05, + "loss": 0.2213, "step": 1163 }, { "epoch": 1.3, - "learning_rate": 0.0002092830882352941, - "loss": 2.803, + "learning_rate": 2.0942095588235295e-05, + "loss": 0.2073, "step": 1164 }, { "epoch": 1.31, - "learning_rate": 0.00020914522058823528, - "loss": 2.8238, + "learning_rate": 2.0928308823529412e-05, + "loss": 0.2234, "step": 1165 }, { "epoch": 1.31, - "learning_rate": 0.00020900735294117643, - "loss": 2.8434, + "learning_rate": 2.091452205882353e-05, + "loss": 0.2465, "step": 1166 }, { "epoch": 1.31, - "learning_rate": 0.00020886948529411764, - "loss": 3.262, + "learning_rate": 2.0900735294117646e-05, + "loss": 0.2859, "step": 1167 }, { "epoch": 1.31, - "learning_rate": 0.0002087316176470588, - "loss": 2.8233, + "learning_rate": 2.0886948529411766e-05, + "loss": 0.2013, "step": 1168 }, { "epoch": 1.31, - "learning_rate": 0.00020859374999999998, - "loss": 2.8239, + "learning_rate": 2.0873161764705883e-05, + "loss": 0.2099, "step": 1169 }, { "epoch": 1.31, - "learning_rate": 0.00020845588235294114, - "loss": 2.8285, + "learning_rate": 2.0859375e-05, + "loss": 0.2619, "step": 1170 }, { "epoch": 1.31, - "learning_rate": 0.00020831801470588232, - "loss": 2.8126, + "learning_rate": 2.0845588235294118e-05, + "loss": 0.2116, "step": 1171 }, { "epoch": 1.31, - "learning_rate": 0.0002081801470588235, - "loss": 2.8216, + "learning_rate": 2.0831801470588235e-05, + "loss": 0.1998, "step": 1172 }, { "epoch": 1.32, - "learning_rate": 0.0002080422794117647, - "loss": 3.0288, + "learning_rate": 2.0818014705882355e-05, + "loss": 0.225, "step": 1173 }, { "epoch": 1.32, - "learning_rate": 0.00020790441176470585, - "loss": 3.6273, + "learning_rate": 2.0804227941176472e-05, + "loss": 0.2537, "step": 1174 }, { "epoch": 1.32, - "learning_rate": 0.00020776654411764703, - "loss": 2.8208, + "learning_rate": 2.079044117647059e-05, + "loss": 0.1997, "step": 1175 }, { "epoch": 1.32, - "learning_rate": 0.00020762867647058822, - "loss": 2.8206, + "learning_rate": 2.0776654411764706e-05, + "loss": 0.2372, "step": 1176 }, { "epoch": 1.32, - "learning_rate": 0.0002074908088235294, - "loss": 2.8153, + "learning_rate": 2.0762867647058826e-05, + "loss": 0.229, "step": 1177 }, { "epoch": 1.32, - "learning_rate": 0.00020735294117647056, - "loss": 2.7922, + "learning_rate": 2.0749080882352943e-05, + "loss": 0.1787, "step": 1178 }, { "epoch": 1.32, - "learning_rate": 0.00020721507352941174, - "loss": 2.8682, + "learning_rate": 2.073529411764706e-05, + "loss": 0.2361, "step": 1179 }, { "epoch": 1.32, - "learning_rate": 0.00020707720588235292, - "loss": 3.7011, + "learning_rate": 2.0721507352941177e-05, + "loss": 0.3212, "step": 1180 }, { "epoch": 1.32, - "learning_rate": 0.0002069393382352941, - "loss": 2.8166, + "learning_rate": 2.0707720588235294e-05, + "loss": 0.2089, "step": 1181 }, { "epoch": 1.33, - "learning_rate": 0.00020680147058823526, - "loss": 2.8197, + "learning_rate": 2.069393382352941e-05, + "loss": 0.2475, "step": 1182 }, { "epoch": 1.33, - "learning_rate": 0.00020666360294117645, - "loss": 2.808, + "learning_rate": 2.068014705882353e-05, + "loss": 0.2598, "step": 1183 }, { "epoch": 1.33, - "learning_rate": 0.00020652573529411763, - "loss": 2.8246, + "learning_rate": 2.0666360294117645e-05, + "loss": 0.1899, "step": 1184 }, { "epoch": 1.33, - "learning_rate": 0.00020638786764705882, - "loss": 2.8503, + "learning_rate": 2.0652573529411766e-05, + "loss": 0.194, "step": 1185 }, { "epoch": 1.33, - "learning_rate": 0.00020624999999999997, - "loss": 4.4275, + "learning_rate": 2.0638786764705883e-05, + "loss": 0.2289, "step": 1186 }, { "epoch": 1.33, - "learning_rate": 0.00020611213235294116, - "loss": 2.8028, + "learning_rate": 2.0625e-05, + "loss": 0.204, "step": 1187 }, { "epoch": 1.33, - "learning_rate": 0.0002059742647058823, - "loss": 2.8164, + "learning_rate": 2.0611213235294117e-05, + "loss": 0.2289, "step": 1188 }, { "epoch": 1.33, - "learning_rate": 0.00020583639705882352, - "loss": 2.8147, + "learning_rate": 2.0597426470588234e-05, + "loss": 0.2027, "step": 1189 }, { "epoch": 1.33, - "learning_rate": 0.00020569852941176468, - "loss": 2.8055, + "learning_rate": 2.0583639705882354e-05, + "loss": 0.2343, "step": 1190 }, { "epoch": 1.34, - "learning_rate": 0.00020556066176470586, - "loss": 2.8396, + "learning_rate": 2.056985294117647e-05, + "loss": 0.2247, "step": 1191 }, { "epoch": 1.34, - "learning_rate": 0.00020542279411764702, - "loss": 3.5156, + "learning_rate": 2.055606617647059e-05, + "loss": 0.2771, "step": 1192 }, { "epoch": 1.34, - "learning_rate": 0.00020528492647058823, - "loss": 2.8027, + "learning_rate": 2.0542279411764705e-05, + "loss": 0.207, "step": 1193 }, { "epoch": 1.34, - "learning_rate": 0.0002051470588235294, - "loss": 2.8086, + "learning_rate": 2.0528492647058826e-05, + "loss": 0.229, "step": 1194 }, { "epoch": 1.34, - "learning_rate": 0.00020500919117647057, - "loss": 2.8187, + "learning_rate": 2.0514705882352943e-05, + "loss": 0.2033, "step": 1195 }, { "epoch": 1.34, - "learning_rate": 0.00020487132352941173, - "loss": 2.8059, + "learning_rate": 2.050091911764706e-05, + "loss": 0.2032, "step": 1196 }, { "epoch": 1.34, - "learning_rate": 0.00020473345588235294, - "loss": 2.8344, + "learning_rate": 2.0487132352941177e-05, + "loss": 0.2348, "step": 1197 }, { "epoch": 1.34, - "learning_rate": 0.0002045955882352941, - "loss": 2.9513, + "learning_rate": 2.0473345588235297e-05, + "loss": 0.2957, "step": 1198 }, { "epoch": 1.34, - "learning_rate": 0.00020445772058823528, - "loss": 3.6013, + "learning_rate": 2.0459558823529414e-05, + "loss": 0.2668, "step": 1199 }, { "epoch": 1.35, - "learning_rate": 0.00020431985294117643, - "loss": 2.8304, + "learning_rate": 2.044577205882353e-05, + "loss": 0.2456, "step": 1200 }, { "epoch": 1.35, - "eval_loss": 3.7005317211151123, - "eval_runtime": 48.3685, - "eval_samples_per_second": 55.883, - "eval_steps_per_second": 0.889, - "eval_wer": 1.0, + "eval_loss": 0.26194095611572266, + "eval_runtime": 47.271, + "eval_samples_per_second": 57.181, + "eval_steps_per_second": 0.91, + "eval_wer": 0.23377816992022352, "step": 1200 }, { "epoch": 1.35, - "learning_rate": 0.00020418198529411765, - "loss": 2.8484, + "learning_rate": 2.0431985294117648e-05, + "loss": 0.2261, "step": 1201 }, { "epoch": 1.35, - "learning_rate": 0.0002040441176470588, - "loss": 2.8309, + "learning_rate": 2.0418198529411765e-05, + "loss": 0.2368, "step": 1202 }, { "epoch": 1.35, - "learning_rate": 0.00020390624999999999, - "loss": 2.8048, + "learning_rate": 2.0404411764705882e-05, + "loss": 0.203, "step": 1203 }, { "epoch": 1.35, - "learning_rate": 0.00020376838235294114, - "loss": 2.8672, + "learning_rate": 2.0390625e-05, + "loss": 0.2512, "step": 1204 }, { "epoch": 1.35, - "learning_rate": 0.00020363051470588233, - "loss": 3.752, + "learning_rate": 2.0376838235294116e-05, + "loss": 0.2392, "step": 1205 }, { "epoch": 1.35, - "learning_rate": 0.0002034926470588235, - "loss": 2.789, + "learning_rate": 2.0363051470588233e-05, + "loss": 0.1894, "step": 1206 }, { "epoch": 1.35, - "learning_rate": 0.0002033547794117647, - "loss": 2.818, + "learning_rate": 2.0349264705882354e-05, + "loss": 0.2178, "step": 1207 }, { "epoch": 1.35, - "learning_rate": 0.00020321691176470585, - "loss": 2.804, + "learning_rate": 2.033547794117647e-05, + "loss": 0.2433, "step": 1208 }, { "epoch": 1.36, - "learning_rate": 0.00020307904411764703, - "loss": 2.8128, + "learning_rate": 2.0321691176470588e-05, + "loss": 0.2162, "step": 1209 }, { "epoch": 1.36, - "learning_rate": 0.00020294117647058822, - "loss": 2.8448, + "learning_rate": 2.0307904411764705e-05, + "loss": 0.225, "step": 1210 }, { "epoch": 1.36, - "learning_rate": 0.0002028033088235294, - "loss": 4.4379, + "learning_rate": 2.0294117647058825e-05, + "loss": 0.2792, "step": 1211 }, { "epoch": 1.36, - "learning_rate": 0.00020266544117647056, - "loss": 2.8143, + "learning_rate": 2.0280330882352942e-05, + "loss": 0.2303, "step": 1212 }, { "epoch": 1.36, - "learning_rate": 0.00020252757352941174, - "loss": 2.8057, + "learning_rate": 2.026654411764706e-05, + "loss": 0.198, "step": 1213 }, { "epoch": 1.36, - "learning_rate": 0.00020238970588235293, - "loss": 2.8164, + "learning_rate": 2.0252757352941176e-05, + "loss": 0.2013, "step": 1214 }, { "epoch": 1.36, - "learning_rate": 0.0002022518382352941, - "loss": 2.8178, + "learning_rate": 2.0238970588235297e-05, + "loss": 0.2334, "step": 1215 }, { "epoch": 1.36, - "learning_rate": 0.00020211397058823527, - "loss": 2.8642, + "learning_rate": 2.0225183823529414e-05, + "loss": 0.2451, "step": 1216 }, { "epoch": 1.36, - "learning_rate": 0.00020197610294117645, - "loss": 3.1984, + "learning_rate": 2.021139705882353e-05, + "loss": 0.2438, "step": 1217 }, { "epoch": 1.37, - "learning_rate": 0.00020183823529411763, - "loss": 2.7987, + "learning_rate": 2.0197610294117648e-05, + "loss": 0.2326, "step": 1218 }, { "epoch": 1.37, - "learning_rate": 0.00020170036764705882, - "loss": 2.8239, + "learning_rate": 2.0183823529411768e-05, + "loss": 0.2188, "step": 1219 }, { "epoch": 1.37, - "learning_rate": 0.00020156249999999997, - "loss": 2.8188, + "learning_rate": 2.0170036764705885e-05, + "loss": 0.2262, "step": 1220 }, { "epoch": 1.37, - "learning_rate": 0.00020142463235294116, - "loss": 2.8259, + "learning_rate": 2.0156250000000002e-05, + "loss": 0.1947, "step": 1221 }, { "epoch": 1.37, - "learning_rate": 0.0002012867647058823, - "loss": 2.8147, + "learning_rate": 2.014246323529412e-05, + "loss": 0.2398, "step": 1222 }, { "epoch": 1.37, - "learning_rate": 0.00020114889705882352, - "loss": 2.9994, + "learning_rate": 2.0128676470588233e-05, + "loss": 0.3286, "step": 1223 }, { "epoch": 1.37, - "learning_rate": 0.00020101102941176468, - "loss": 3.5207, + "learning_rate": 2.0114889705882353e-05, + "loss": 0.2325, "step": 1224 }, { "epoch": 1.37, - "learning_rate": 0.00020087316176470586, - "loss": 2.8066, + "learning_rate": 2.010110294117647e-05, + "loss": 0.202, "step": 1225 }, { "epoch": 1.37, - "learning_rate": 0.00020073529411764702, - "loss": 2.8253, + "learning_rate": 2.0087316176470587e-05, + "loss": 0.2125, "step": 1226 }, { "epoch": 1.38, - "learning_rate": 0.00020059742647058823, - "loss": 2.8231, + "learning_rate": 2.0073529411764704e-05, + "loss": 0.2257, "step": 1227 }, { "epoch": 1.38, - "learning_rate": 0.0002004595588235294, - "loss": 2.8066, + "learning_rate": 2.0059742647058825e-05, + "loss": 0.2349, "step": 1228 }, { "epoch": 1.38, - "learning_rate": 0.00020032169117647057, - "loss": 2.8979, + "learning_rate": 2.004595588235294e-05, + "loss": 0.2462, "step": 1229 }, { "epoch": 1.38, - "learning_rate": 0.00020018382352941173, - "loss": 3.7534, + "learning_rate": 2.003216911764706e-05, + "loss": 0.2964, "step": 1230 }, { "epoch": 1.38, - "learning_rate": 0.00020004595588235294, - "loss": 2.8221, + "learning_rate": 2.0018382352941176e-05, + "loss": 0.2062, "step": 1231 }, { "epoch": 1.38, - "learning_rate": 0.0001999080882352941, - "loss": 2.8484, + "learning_rate": 2.0004595588235296e-05, + "loss": 0.2384, "step": 1232 }, { "epoch": 1.38, - "learning_rate": 0.00019977022058823528, - "loss": 2.832, + "learning_rate": 1.9990808823529413e-05, + "loss": 0.2105, "step": 1233 }, { "epoch": 1.38, - "learning_rate": 0.00019963235294117644, - "loss": 2.83, + "learning_rate": 1.997702205882353e-05, + "loss": 0.2004, "step": 1234 }, { "epoch": 1.38, - "learning_rate": 0.00019949448529411765, - "loss": 2.8338, + "learning_rate": 1.9963235294117647e-05, + "loss": 0.2405, "step": 1235 }, { "epoch": 1.39, - "learning_rate": 0.0001993566176470588, - "loss": 4.9529, + "learning_rate": 1.9949448529411767e-05, + "loss": 0.2881, "step": 1236 }, { "epoch": 1.39, - "learning_rate": 0.00019921875, - "loss": 2.7981, + "learning_rate": 1.9935661764705884e-05, + "loss": 0.1769, "step": 1237 }, { "epoch": 1.39, - "learning_rate": 0.00019908088235294114, - "loss": 2.7971, + "learning_rate": 1.9921875e-05, + "loss": 0.1961, "step": 1238 }, { "epoch": 1.39, - "learning_rate": 0.00019894301470588233, - "loss": 2.8032, + "learning_rate": 1.990808823529412e-05, + "loss": 0.212, "step": 1239 }, { "epoch": 1.39, - "learning_rate": 0.0001988051470588235, - "loss": 2.8231, + "learning_rate": 1.9894301470588235e-05, + "loss": 0.1889, "step": 1240 }, { "epoch": 1.39, - "learning_rate": 0.0001986672794117647, - "loss": 2.852, + "learning_rate": 1.9880514705882356e-05, + "loss": 0.2682, "step": 1241 }, { "epoch": 1.39, - "learning_rate": 0.00019852941176470585, - "loss": 3.1133, + "learning_rate": 1.9866727941176473e-05, + "loss": 0.3147, "step": 1242 }, { "epoch": 1.39, - "learning_rate": 0.00019839154411764704, - "loss": 2.8316, + "learning_rate": 1.9852941176470586e-05, + "loss": 0.2073, "step": 1243 }, { "epoch": 1.39, - "learning_rate": 0.00019825367647058822, - "loss": 2.8069, + "learning_rate": 1.9839154411764704e-05, + "loss": 0.2112, "step": 1244 }, { "epoch": 1.4, - "learning_rate": 0.0001981158088235294, - "loss": 2.7946, + "learning_rate": 1.9825367647058824e-05, + "loss": 0.2206, "step": 1245 }, { "epoch": 1.4, - "learning_rate": 0.00019797794117647056, - "loss": 2.807, + "learning_rate": 1.981158088235294e-05, + "loss": 0.1941, "step": 1246 }, { "epoch": 1.4, - "learning_rate": 0.00019784007352941174, - "loss": 2.813, + "learning_rate": 1.9797794117647058e-05, + "loss": 0.2834, "step": 1247 }, { "epoch": 1.4, - "learning_rate": 0.00019770220588235293, - "loss": 2.9537, + "learning_rate": 1.9784007352941175e-05, + "loss": 0.3134, "step": 1248 }, { "epoch": 1.4, - "learning_rate": 0.0001975643382352941, - "loss": 3.564, + "learning_rate": 1.9770220588235295e-05, + "loss": 0.2419, "step": 1249 }, { "epoch": 1.4, - "learning_rate": 0.00019742647058823527, - "loss": 2.8241, + "learning_rate": 1.9756433823529412e-05, + "loss": 0.2122, "step": 1250 }, { "epoch": 1.4, - "learning_rate": 0.00019728860294117645, - "loss": 2.8326, + "learning_rate": 1.974264705882353e-05, + "loss": 0.1999, "step": 1251 }, { "epoch": 1.4, - "learning_rate": 0.00019715073529411763, - "loss": 2.8195, + "learning_rate": 1.9728860294117646e-05, + "loss": 0.2248, "step": 1252 }, { "epoch": 1.4, - "learning_rate": 0.00019701286764705882, - "loss": 2.8252, + "learning_rate": 1.9715073529411767e-05, + "loss": 0.2505, "step": 1253 }, { "epoch": 1.41, - "learning_rate": 0.00019687499999999997, - "loss": 2.9116, + "learning_rate": 1.9701286764705884e-05, + "loss": 0.2512, "step": 1254 }, { "epoch": 1.41, - "learning_rate": 0.00019673713235294116, - "loss": 4.1353, + "learning_rate": 1.96875e-05, + "loss": 0.2873, "step": 1255 }, { "epoch": 1.41, - "learning_rate": 0.00019659926470588231, - "loss": 2.7972, + "learning_rate": 1.9673713235294118e-05, + "loss": 0.2305, "step": 1256 }, { "epoch": 1.41, - "learning_rate": 0.00019646139705882353, - "loss": 2.8084, + "learning_rate": 1.9659926470588235e-05, + "loss": 0.2035, "step": 1257 }, { "epoch": 1.41, - "learning_rate": 0.00019632352941176468, - "loss": 2.823, + "learning_rate": 1.9646139705882355e-05, + "loss": 0.2242, "step": 1258 }, { "epoch": 1.41, - "learning_rate": 0.00019618566176470587, - "loss": 2.8326, + "learning_rate": 1.9632352941176472e-05, + "loss": 0.2055, "step": 1259 }, { "epoch": 1.41, - "learning_rate": 0.00019604779411764702, - "loss": 3.0079, + "learning_rate": 1.961856617647059e-05, + "loss": 0.27, "step": 1260 }, { "epoch": 1.41, - "learning_rate": 0.00019590992647058823, - "loss": 5.3265, + "learning_rate": 1.9604779411764706e-05, + "loss": 0.348, "step": 1261 }, { "epoch": 1.41, - "learning_rate": 0.0001957720588235294, - "loss": 2.8093, + "learning_rate": 1.9590992647058827e-05, + "loss": 0.2026, "step": 1262 }, { "epoch": 1.42, - "learning_rate": 0.00019563419117647057, - "loss": 2.8333, + "learning_rate": 1.9577205882352944e-05, + "loss": 0.1813, "step": 1263 }, { "epoch": 1.42, - "learning_rate": 0.00019549632352941173, - "loss": 2.8125, + "learning_rate": 1.9563419117647057e-05, + "loss": 0.1777, "step": 1264 }, { "epoch": 1.42, - "learning_rate": 0.00019535845588235294, - "loss": 2.8193, + "learning_rate": 1.9549632352941174e-05, + "loss": 0.2077, "step": 1265 }, { "epoch": 1.42, - "learning_rate": 0.0001952205882352941, - "loss": 2.8349, + "learning_rate": 1.9535845588235295e-05, + "loss": 0.2303, "step": 1266 }, { "epoch": 1.42, - "learning_rate": 0.00019508272058823528, - "loss": 3.1616, + "learning_rate": 1.9522058823529412e-05, + "loss": 0.2436, "step": 1267 }, { "epoch": 1.42, - "learning_rate": 0.00019494485294117644, - "loss": 2.8072, + "learning_rate": 1.950827205882353e-05, + "loss": 0.2, "step": 1268 }, { "epoch": 1.42, - "learning_rate": 0.00019480698529411765, - "loss": 2.8128, + "learning_rate": 1.9494485294117646e-05, + "loss": 0.2149, "step": 1269 }, { "epoch": 1.42, - "learning_rate": 0.0001946691176470588, - "loss": 2.7967, + "learning_rate": 1.9480698529411766e-05, + "loss": 0.178, "step": 1270 }, { "epoch": 1.42, - "learning_rate": 0.00019453125, - "loss": 2.807, + "learning_rate": 1.9466911764705883e-05, + "loss": 0.2007, "step": 1271 }, { "epoch": 1.43, - "learning_rate": 0.00019439338235294115, - "loss": 2.8636, + "learning_rate": 1.9453125e-05, + "loss": 0.2483, "step": 1272 }, { "epoch": 1.43, - "learning_rate": 0.00019425551470588233, - "loss": 3.0599, + "learning_rate": 1.9439338235294117e-05, + "loss": 0.2116, "step": 1273 }, { "epoch": 1.43, - "learning_rate": 0.0001941176470588235, - "loss": 3.9443, + "learning_rate": 1.9425551470588234e-05, + "loss": 0.2462, "step": 1274 }, { "epoch": 1.43, - "learning_rate": 0.0001939797794117647, - "loss": 2.8197, + "learning_rate": 1.9411764705882355e-05, + "loss": 0.2308, "step": 1275 }, { "epoch": 1.43, - "learning_rate": 0.00019384191176470585, - "loss": 2.8357, + "learning_rate": 1.939797794117647e-05, + "loss": 0.2281, "step": 1276 }, { "epoch": 1.43, - "learning_rate": 0.00019370404411764704, - "loss": 2.8116, + "learning_rate": 1.938419117647059e-05, + "loss": 0.2043, "step": 1277 }, { "epoch": 1.43, - "learning_rate": 0.00019356617647058822, - "loss": 2.8604, + "learning_rate": 1.9370404411764706e-05, + "loss": 0.2314, "step": 1278 }, { "epoch": 1.43, - "learning_rate": 0.0001934283088235294, - "loss": 2.8473, + "learning_rate": 1.9356617647058826e-05, + "loss": 0.3141, "step": 1279 }, { "epoch": 1.43, - "learning_rate": 0.00019329044117647056, - "loss": 3.8609, + "learning_rate": 1.9342830882352943e-05, + "loss": 0.2971, "step": 1280 }, { "epoch": 1.44, - "learning_rate": 0.00019315257352941174, - "loss": 2.8823, + "learning_rate": 1.932904411764706e-05, + "loss": 0.2107, "step": 1281 }, { "epoch": 1.44, - "learning_rate": 0.00019301470588235293, - "loss": 2.8855, + "learning_rate": 1.9315257352941177e-05, + "loss": 0.2169, "step": 1282 }, { "epoch": 1.44, - "learning_rate": 0.0001928768382352941, - "loss": 2.9027, + "learning_rate": 1.9301470588235298e-05, + "loss": 0.1919, "step": 1283 }, { "epoch": 1.44, - "learning_rate": 0.00019273897058823527, - "loss": 2.9149, + "learning_rate": 1.9287683823529415e-05, + "loss": 0.2322, "step": 1284 }, { "epoch": 1.44, - "learning_rate": 0.00019260110294117645, - "loss": 2.8766, + "learning_rate": 1.9273897058823528e-05, + "loss": 0.2423, "step": 1285 }, { "epoch": 1.44, - "learning_rate": 0.00019246323529411764, - "loss": 3.9012, + "learning_rate": 1.9260110294117645e-05, + "loss": 0.2524, "step": 1286 }, { "epoch": 1.44, - "learning_rate": 0.00019232536764705882, - "loss": 2.8727, + "learning_rate": 1.9246323529411766e-05, + "loss": 0.2162, "step": 1287 }, { "epoch": 1.44, - "learning_rate": 0.00019218749999999998, - "loss": 2.8366, + "learning_rate": 1.9232536764705883e-05, + "loss": 0.2005, "step": 1288 }, { "epoch": 1.45, - "learning_rate": 0.00019204963235294116, - "loss": 2.843, + "learning_rate": 1.921875e-05, + "loss": 0.2249, "step": 1289 }, { "epoch": 1.45, - "learning_rate": 0.00019191176470588232, - "loss": 2.8218, + "learning_rate": 1.9204963235294117e-05, + "loss": 0.1993, "step": 1290 }, { "epoch": 1.45, - "learning_rate": 0.00019177389705882353, - "loss": 2.8402, + "learning_rate": 1.9191176470588234e-05, + "loss": 0.238, "step": 1291 }, { "epoch": 1.45, - "learning_rate": 0.00019163602941176468, - "loss": 3.0862, + "learning_rate": 1.9177389705882354e-05, + "loss": 0.2362, "step": 1292 }, { "epoch": 1.45, - "learning_rate": 0.00019149816176470587, - "loss": 2.8275, + "learning_rate": 1.916360294117647e-05, + "loss": 0.1923, "step": 1293 }, { "epoch": 1.45, - "learning_rate": 0.00019136029411764702, - "loss": 2.8096, + "learning_rate": 1.9149816176470588e-05, + "loss": 0.2313, "step": 1294 }, { "epoch": 1.45, - "learning_rate": 0.00019122242647058823, - "loss": 2.8225, + "learning_rate": 1.9136029411764705e-05, + "loss": 0.2038, "step": 1295 }, { "epoch": 1.45, - "learning_rate": 0.0001910845588235294, - "loss": 2.8054, + "learning_rate": 1.9122242647058825e-05, + "loss": 0.2158, "step": 1296 }, { "epoch": 1.45, - "learning_rate": 0.00019094669117647057, - "loss": 2.8221, + "learning_rate": 1.9108455882352942e-05, + "loss": 0.2099, "step": 1297 }, { "epoch": 1.46, - "learning_rate": 0.00019080882352941173, - "loss": 3.1525, + "learning_rate": 1.909466911764706e-05, + "loss": 0.271, "step": 1298 }, { "epoch": 1.46, - "learning_rate": 0.00019067095588235294, - "loss": 4.109, + "learning_rate": 1.9080882352941176e-05, + "loss": 0.22, "step": 1299 }, { "epoch": 1.46, - "learning_rate": 0.0001905330882352941, - "loss": 2.85, + "learning_rate": 1.9067095588235297e-05, + "loss": 0.2392, "step": 1300 }, { "epoch": 1.46, - "eval_loss": 3.508472204208374, - "eval_runtime": 48.5093, - "eval_samples_per_second": 55.721, - "eval_steps_per_second": 0.886, - "eval_wer": 1.0, + "eval_loss": 0.24657748639583588, + "eval_runtime": 48.1105, + "eval_samples_per_second": 56.183, + "eval_steps_per_second": 0.894, + "eval_wer": 0.2194588434248741, "step": 1300 }, { "epoch": 1.46, - "learning_rate": 0.00019039522058823528, - "loss": 2.8686, + "learning_rate": 1.9053308823529414e-05, + "loss": 0.2591, "step": 1301 }, { "epoch": 1.46, - "learning_rate": 0.00019025735294117644, - "loss": 2.8743, + "learning_rate": 1.903952205882353e-05, + "loss": 0.1786, "step": 1302 }, { "epoch": 1.46, - "learning_rate": 0.00019011948529411765, - "loss": 2.869, + "learning_rate": 1.9025735294117648e-05, + "loss": 0.2252, "step": 1303 }, { "epoch": 1.46, - "learning_rate": 0.0001899816176470588, - "loss": 2.8682, + "learning_rate": 1.901194852941177e-05, + "loss": 0.2571, "step": 1304 }, { "epoch": 1.46, - "learning_rate": 0.00018984375, - "loss": 4.2041, + "learning_rate": 1.8998161764705882e-05, + "loss": 0.3722, "step": 1305 }, { "epoch": 1.46, - "learning_rate": 0.00018970588235294115, - "loss": 2.8464, + "learning_rate": 1.8984375e-05, + "loss": 0.2042, "step": 1306 }, { "epoch": 1.47, - "learning_rate": 0.00018956801470588233, - "loss": 2.819, + "learning_rate": 1.8970588235294116e-05, + "loss": 0.2231, "step": 1307 }, { "epoch": 1.47, - "learning_rate": 0.00018943014705882351, - "loss": 2.8187, + "learning_rate": 1.8956801470588233e-05, + "loss": 0.2023, "step": 1308 }, { "epoch": 1.47, - "learning_rate": 0.0001892922794117647, - "loss": 2.8143, + "learning_rate": 1.8943014705882353e-05, + "loss": 0.2134, "step": 1309 }, { "epoch": 1.47, - "learning_rate": 0.00018915441176470585, - "loss": 2.8596, + "learning_rate": 1.892922794117647e-05, + "loss": 0.2399, "step": 1310 }, { "epoch": 1.47, - "learning_rate": 0.00018901654411764704, - "loss": 4.4044, + "learning_rate": 1.8915441176470587e-05, + "loss": 0.2524, "step": 1311 }, { "epoch": 1.47, - "learning_rate": 0.00018887867647058822, - "loss": 2.8066, + "learning_rate": 1.8901654411764704e-05, + "loss": 0.2399, "step": 1312 }, { "epoch": 1.47, - "learning_rate": 0.0001887408088235294, - "loss": 2.8088, + "learning_rate": 1.8887867647058825e-05, + "loss": 0.1807, "step": 1313 }, { "epoch": 1.47, - "learning_rate": 0.00018860294117647056, - "loss": 2.8096, + "learning_rate": 1.8874080882352942e-05, + "loss": 0.2297, "step": 1314 }, { "epoch": 1.47, - "learning_rate": 0.00018846507352941175, - "loss": 2.8058, + "learning_rate": 1.886029411764706e-05, + "loss": 0.1956, "step": 1315 }, { "epoch": 1.48, - "learning_rate": 0.00018832720588235293, - "loss": 2.8613, + "learning_rate": 1.8846507352941176e-05, + "loss": 0.2325, "step": 1316 }, { "epoch": 1.48, - "learning_rate": 0.0001881893382352941, - "loss": 3.0864, + "learning_rate": 1.8832720588235296e-05, + "loss": 0.3321, "step": 1317 }, { "epoch": 1.48, - "learning_rate": 0.00018805147058823527, - "loss": 2.8688, + "learning_rate": 1.8818933823529413e-05, + "loss": 0.2272, "step": 1318 }, { "epoch": 1.48, - "learning_rate": 0.00018791360294117645, - "loss": 2.8676, + "learning_rate": 1.880514705882353e-05, + "loss": 0.1706, "step": 1319 }, { "epoch": 1.48, - "learning_rate": 0.00018777573529411764, - "loss": 2.8408, + "learning_rate": 1.8791360294117647e-05, + "loss": 0.2209, "step": 1320 }, { "epoch": 1.48, - "learning_rate": 0.00018763786764705882, - "loss": 2.855, + "learning_rate": 1.8777573529411768e-05, + "loss": 0.2099, "step": 1321 }, { "epoch": 1.48, - "learning_rate": 0.00018749999999999998, - "loss": 2.8446, + "learning_rate": 1.8763786764705885e-05, + "loss": 0.2071, "step": 1322 }, { "epoch": 1.48, - "learning_rate": 0.00018736213235294116, - "loss": 2.9301, + "learning_rate": 1.8750000000000002e-05, + "loss": 0.211, "step": 1323 }, { "epoch": 1.48, - "learning_rate": 0.00018722426470588232, - "loss": 3.8478, + "learning_rate": 1.873621323529412e-05, + "loss": 0.1903, "step": 1324 }, { "epoch": 1.49, - "learning_rate": 0.00018708639705882353, - "loss": 2.7932, + "learning_rate": 1.8722426470588236e-05, + "loss": 0.2025, "step": 1325 }, { "epoch": 1.49, - "learning_rate": 0.00018694852941176468, - "loss": 2.8297, + "learning_rate": 1.8708639705882353e-05, + "loss": 0.1825, "step": 1326 }, { "epoch": 1.49, - "learning_rate": 0.00018681066176470587, - "loss": 2.8149, + "learning_rate": 1.869485294117647e-05, + "loss": 0.1986, "step": 1327 }, { "epoch": 1.49, - "learning_rate": 0.00018667279411764702, - "loss": 2.8082, + "learning_rate": 1.8681066176470587e-05, + "loss": 0.169, "step": 1328 }, { "epoch": 1.49, - "learning_rate": 0.00018653492647058824, - "loss": 2.8736, + "learning_rate": 1.8667279411764704e-05, + "loss": 0.2199, "step": 1329 }, { "epoch": 1.49, - "learning_rate": 0.0001863970588235294, - "loss": 3.8851, + "learning_rate": 1.8653492647058824e-05, + "loss": 0.2651, "step": 1330 }, { "epoch": 1.49, - "learning_rate": 0.00018625919117647058, - "loss": 2.8531, + "learning_rate": 1.863970588235294e-05, + "loss": 0.2143, "step": 1331 }, { "epoch": 1.49, - "learning_rate": 0.00018612132352941173, - "loss": 2.8327, + "learning_rate": 1.8625919117647058e-05, + "loss": 0.2167, "step": 1332 }, { "epoch": 1.49, - "learning_rate": 0.00018598345588235294, - "loss": 2.8564, + "learning_rate": 1.8612132352941175e-05, + "loss": 0.2101, "step": 1333 }, { "epoch": 1.5, - "learning_rate": 0.0001858455882352941, - "loss": 2.8291, + "learning_rate": 1.8598345588235296e-05, + "loss": 0.1816, "step": 1334 }, { "epoch": 1.5, - "learning_rate": 0.00018570772058823528, - "loss": 2.8321, + "learning_rate": 1.8584558823529413e-05, + "loss": 0.1871, "step": 1335 }, { "epoch": 1.5, - "learning_rate": 0.00018556985294117644, - "loss": 3.9564, + "learning_rate": 1.857077205882353e-05, + "loss": 0.2779, "step": 1336 }, { "epoch": 1.5, - "learning_rate": 0.00018543198529411765, - "loss": 2.8076, + "learning_rate": 1.8556985294117647e-05, + "loss": 0.1967, "step": 1337 }, { "epoch": 1.5, - "learning_rate": 0.0001852941176470588, - "loss": 2.7971, + "learning_rate": 1.8543198529411767e-05, + "loss": 0.209, "step": 1338 }, { "epoch": 1.5, - "learning_rate": 0.00018515625, - "loss": 2.8086, + "learning_rate": 1.8529411764705884e-05, + "loss": 0.2063, "step": 1339 }, { "epoch": 1.5, - "learning_rate": 0.00018501838235294115, - "loss": 2.8323, + "learning_rate": 1.8515625e-05, + "loss": 0.1942, "step": 1340 }, { "epoch": 1.5, - "learning_rate": 0.00018488051470588233, - "loss": 2.8535, + "learning_rate": 1.8501838235294118e-05, + "loss": 0.1965, "step": 1341 }, { "epoch": 1.5, - "learning_rate": 0.00018474264705882351, - "loss": 3.3502, + "learning_rate": 1.8488051470588235e-05, + "loss": 0.3053, "step": 1342 }, { "epoch": 1.51, - "learning_rate": 0.0001846047794117647, - "loss": 2.8069, + "learning_rate": 1.8474264705882356e-05, + "loss": 0.2166, "step": 1343 }, { "epoch": 1.51, - "learning_rate": 0.00018446691176470586, - "loss": 2.8252, + "learning_rate": 1.8460477941176473e-05, + "loss": 0.2418, "step": 1344 }, { "epoch": 1.51, - "learning_rate": 0.00018432904411764704, - "loss": 2.814, + "learning_rate": 1.844669117647059e-05, + "loss": 0.1994, "step": 1345 }, { "epoch": 1.51, - "learning_rate": 0.00018419117647058822, - "loss": 2.8021, + "learning_rate": 1.8432904411764707e-05, + "loss": 0.1982, "step": 1346 }, { "epoch": 1.51, - "learning_rate": 0.0001840533088235294, - "loss": 2.8209, + "learning_rate": 1.8419117647058824e-05, + "loss": 0.2175, "step": 1347 }, { "epoch": 1.51, - "learning_rate": 0.00018391544117647056, - "loss": 2.986, + "learning_rate": 1.840533088235294e-05, + "loss": 0.245, "step": 1348 }, { "epoch": 1.51, - "learning_rate": 0.00018377757352941175, - "loss": 3.6202, + "learning_rate": 1.8391544117647058e-05, + "loss": 0.2431, "step": 1349 }, { "epoch": 1.51, - "learning_rate": 0.00018363970588235293, - "loss": 2.8158, + "learning_rate": 1.8377757352941175e-05, + "loss": 0.1817, "step": 1350 }, { "epoch": 1.51, - "learning_rate": 0.00018350183823529411, - "loss": 2.8262, + "learning_rate": 1.8363970588235295e-05, + "loss": 0.1948, "step": 1351 }, { "epoch": 1.52, - "learning_rate": 0.00018336397058823527, - "loss": 2.8214, + "learning_rate": 1.8350183823529412e-05, + "loss": 0.1676, "step": 1352 }, { "epoch": 1.52, - "learning_rate": 0.00018322610294117645, - "loss": 2.8197, + "learning_rate": 1.833639705882353e-05, + "loss": 0.1921, "step": 1353 }, { "epoch": 1.52, - "learning_rate": 0.00018308823529411764, - "loss": 2.8659, + "learning_rate": 1.8322610294117646e-05, + "loss": 0.2341, "step": 1354 }, { "epoch": 1.52, - "learning_rate": 0.00018295036764705882, - "loss": 3.9313, + "learning_rate": 1.8308823529411766e-05, + "loss": 0.2471, "step": 1355 }, { "epoch": 1.52, - "learning_rate": 0.00018281249999999998, - "loss": 2.82, + "learning_rate": 1.8295036764705884e-05, + "loss": 0.1857, "step": 1356 }, { "epoch": 1.52, - "learning_rate": 0.00018267463235294116, - "loss": 2.799, + "learning_rate": 1.828125e-05, + "loss": 0.2182, "step": 1357 }, { "epoch": 1.52, - "learning_rate": 0.00018253676470588232, - "loss": 2.8153, + "learning_rate": 1.8267463235294118e-05, + "loss": 0.2042, "step": 1358 }, { "epoch": 1.52, - "learning_rate": 0.00018239889705882353, - "loss": 2.8065, + "learning_rate": 1.8253676470588235e-05, + "loss": 0.1896, "step": 1359 }, { "epoch": 1.52, - "learning_rate": 0.00018226102941176469, - "loss": 2.8495, + "learning_rate": 1.8239889705882355e-05, + "loss": 0.2983, "step": 1360 }, { "epoch": 1.53, - "learning_rate": 0.00018212316176470587, - "loss": 4.718, + "learning_rate": 1.8226102941176472e-05, + "loss": 0.3222, "step": 1361 }, { "epoch": 1.53, - "learning_rate": 0.00018198529411764703, - "loss": 2.8249, + "learning_rate": 1.821231617647059e-05, + "loss": 0.2213, "step": 1362 }, { "epoch": 1.53, - "learning_rate": 0.00018184742647058824, - "loss": 2.8398, + "learning_rate": 1.8198529411764706e-05, + "loss": 0.2371, "step": 1363 }, { "epoch": 1.53, - "learning_rate": 0.0001817095588235294, - "loss": 2.8318, + "learning_rate": 1.8184742647058826e-05, + "loss": 0.1916, "step": 1364 }, { "epoch": 1.53, - "learning_rate": 0.00018157169117647058, - "loss": 2.8282, + "learning_rate": 1.8170955882352943e-05, + "loss": 0.186, "step": 1365 }, { "epoch": 1.53, - "learning_rate": 0.00018143382352941173, - "loss": 2.8382, + "learning_rate": 1.815716911764706e-05, + "loss": 0.2239, "step": 1366 }, { "epoch": 1.53, - "learning_rate": 0.00018129595588235294, - "loss": 3.3375, + "learning_rate": 1.8143382352941177e-05, + "loss": 0.2316, "step": 1367 }, { "epoch": 1.53, - "learning_rate": 0.0001811580882352941, - "loss": 2.8104, + "learning_rate": 1.8129595588235294e-05, + "loss": 0.2061, "step": 1368 }, { "epoch": 1.53, - "learning_rate": 0.00018102022058823528, - "loss": 2.8076, + "learning_rate": 1.811580882352941e-05, + "loss": 0.1984, "step": 1369 }, { "epoch": 1.54, - "learning_rate": 0.00018088235294117644, - "loss": 2.8073, + "learning_rate": 1.810202205882353e-05, + "loss": 0.2262, "step": 1370 }, { "epoch": 1.54, - "learning_rate": 0.00018074448529411765, - "loss": 2.8407, + "learning_rate": 1.8088235294117645e-05, + "loss": 0.2077, "step": 1371 }, { "epoch": 1.54, - "learning_rate": 0.0001806066176470588, - "loss": 2.8526, + "learning_rate": 1.8074448529411766e-05, + "loss": 0.183, "step": 1372 }, { "epoch": 1.54, - "learning_rate": 0.00018046875, - "loss": 3.0407, + "learning_rate": 1.8060661764705883e-05, + "loss": 0.2708, "step": 1373 }, { "epoch": 1.54, - "learning_rate": 0.00018033088235294115, - "loss": 3.8336, + "learning_rate": 1.8046875e-05, + "loss": 0.2194, "step": 1374 }, { "epoch": 1.54, - "learning_rate": 0.00018019301470588233, - "loss": 2.8096, + "learning_rate": 1.8033088235294117e-05, + "loss": 0.2173, "step": 1375 }, { "epoch": 1.54, - "learning_rate": 0.00018005514705882352, - "loss": 2.8069, + "learning_rate": 1.8019301470588234e-05, + "loss": 0.1841, "step": 1376 }, { "epoch": 1.54, - "learning_rate": 0.0001799172794117647, - "loss": 2.8147, + "learning_rate": 1.8005514705882354e-05, + "loss": 0.2471, "step": 1377 }, { "epoch": 1.54, - "learning_rate": 0.00017977941176470586, - "loss": 2.8188, + "learning_rate": 1.799172794117647e-05, + "loss": 0.2213, "step": 1378 }, { "epoch": 1.55, - "learning_rate": 0.00017964154411764704, - "loss": 2.9063, + "learning_rate": 1.797794117647059e-05, + "loss": 0.1597, "step": 1379 }, { "epoch": 1.55, - "learning_rate": 0.00017950367647058822, - "loss": 4.597, + "learning_rate": 1.7964154411764705e-05, + "loss": 0.373, "step": 1380 }, { "epoch": 1.55, - "learning_rate": 0.0001793658088235294, - "loss": 2.8106, + "learning_rate": 1.7950367647058826e-05, + "loss": 0.1995, "step": 1381 }, { "epoch": 1.55, - "learning_rate": 0.00017922794117647056, - "loss": 2.8145, + "learning_rate": 1.7936580882352943e-05, + "loss": 0.2243, "step": 1382 }, { "epoch": 1.55, - "learning_rate": 0.00017909007352941175, - "loss": 2.8249, + "learning_rate": 1.792279411764706e-05, + "loss": 0.1829, "step": 1383 }, { "epoch": 1.55, - "learning_rate": 0.00017895220588235293, - "loss": 2.8151, + "learning_rate": 1.7909007352941177e-05, + "loss": 0.1802, "step": 1384 }, { "epoch": 1.55, - "learning_rate": 0.00017881433823529412, - "loss": 2.8397, + "learning_rate": 1.7895220588235297e-05, + "loss": 0.2376, "step": 1385 }, { "epoch": 1.55, - "learning_rate": 0.00017867647058823527, - "loss": 4.603, + "learning_rate": 1.7881433823529414e-05, + "loss": 0.2887, "step": 1386 }, { "epoch": 1.55, - "learning_rate": 0.00017853860294117646, - "loss": 2.8287, + "learning_rate": 1.786764705882353e-05, + "loss": 0.2193, "step": 1387 }, { "epoch": 1.56, - "learning_rate": 0.00017840073529411764, - "loss": 2.8244, + "learning_rate": 1.7853860294117645e-05, + "loss": 0.2289, "step": 1388 }, { "epoch": 1.56, - "learning_rate": 0.00017826286764705882, - "loss": 2.812, + "learning_rate": 1.7840073529411765e-05, + "loss": 0.2028, "step": 1389 }, { "epoch": 1.56, - "learning_rate": 0.00017812499999999998, - "loss": 2.8164, + "learning_rate": 1.7826286764705882e-05, + "loss": 0.2127, "step": 1390 }, { "epoch": 1.56, - "learning_rate": 0.00017798713235294116, - "loss": 2.8421, + "learning_rate": 1.78125e-05, + "loss": 0.2111, "step": 1391 }, { "epoch": 1.56, - "learning_rate": 0.00017784926470588232, - "loss": 3.3066, + "learning_rate": 1.7798713235294116e-05, + "loss": 0.2689, "step": 1392 }, { "epoch": 1.56, - "learning_rate": 0.00017771139705882353, - "loss": 2.8133, + "learning_rate": 1.7784926470588233e-05, + "loss": 0.213, "step": 1393 }, { "epoch": 1.56, - "learning_rate": 0.0001775735294117647, - "loss": 2.8159, + "learning_rate": 1.7771139705882354e-05, + "loss": 0.1841, "step": 1394 }, { "epoch": 1.56, - "learning_rate": 0.00017743566176470587, - "loss": 2.7978, + "learning_rate": 1.775735294117647e-05, + "loss": 0.1933, "step": 1395 }, { "epoch": 1.57, - "learning_rate": 0.00017729779411764703, - "loss": 2.8326, + "learning_rate": 1.7743566176470588e-05, + "loss": 0.2259, "step": 1396 }, { "epoch": 1.57, - "learning_rate": 0.00017715992647058824, - "loss": 2.8324, + "learning_rate": 1.7729779411764705e-05, + "loss": 0.196, "step": 1397 }, { "epoch": 1.57, - "learning_rate": 0.0001770220588235294, - "loss": 3.07, + "learning_rate": 1.7715992647058825e-05, + "loss": 0.2101, "step": 1398 }, { "epoch": 1.57, - "learning_rate": 0.00017688419117647058, - "loss": 3.6662, + "learning_rate": 1.7702205882352942e-05, + "loss": 0.2386, "step": 1399 }, { "epoch": 1.57, - "learning_rate": 0.00017674632352941173, - "loss": 2.8509, + "learning_rate": 1.768841911764706e-05, + "loss": 0.2117, "step": 1400 }, { "epoch": 1.57, - "eval_loss": 3.6156930923461914, - "eval_runtime": 47.7788, - "eval_samples_per_second": 56.573, - "eval_steps_per_second": 0.9, - "eval_wer": 1.0, + "eval_loss": 0.23789431154727936, + "eval_runtime": 47.9843, + "eval_samples_per_second": 56.331, + "eval_steps_per_second": 0.896, + "eval_wer": 0.20920186757839784, "step": 1400 }, { "epoch": 1.57, - "learning_rate": 0.00017660845588235295, - "loss": 2.8383, + "learning_rate": 1.7674632352941176e-05, + "loss": 0.1621, "step": 1401 }, { "epoch": 1.57, - "learning_rate": 0.0001764705882352941, - "loss": 2.8649, + "learning_rate": 1.7660845588235297e-05, + "loss": 0.2035, "step": 1402 }, { "epoch": 1.57, - "learning_rate": 0.00017633272058823529, - "loss": 2.8348, + "learning_rate": 1.7647058823529414e-05, + "loss": 0.1902, "step": 1403 }, { "epoch": 1.57, - "learning_rate": 0.00017619485294117644, - "loss": 2.8627, + "learning_rate": 1.763327205882353e-05, + "loss": 0.2601, "step": 1404 }, { "epoch": 1.58, - "learning_rate": 0.00017605698529411765, - "loss": 4.2212, + "learning_rate": 1.7619485294117648e-05, + "loss": 0.3174, "step": 1405 }, { "epoch": 1.58, - "learning_rate": 0.0001759191176470588, - "loss": 2.8415, + "learning_rate": 1.7605698529411768e-05, + "loss": 0.1889, "step": 1406 }, { "epoch": 1.58, - "learning_rate": 0.00017578125, - "loss": 2.8369, + "learning_rate": 1.7591911764705885e-05, + "loss": 0.198, "step": 1407 }, { "epoch": 1.58, - "learning_rate": 0.00017564338235294115, - "loss": 2.8245, + "learning_rate": 1.7578125000000002e-05, + "loss": 0.202, "step": 1408 }, { "epoch": 1.58, - "learning_rate": 0.00017550551470588233, - "loss": 2.8212, + "learning_rate": 1.7564338235294116e-05, + "loss": 0.2384, "step": 1409 }, { "epoch": 1.58, - "learning_rate": 0.00017536764705882352, - "loss": 2.8252, + "learning_rate": 1.7550551470588233e-05, + "loss": 0.2232, "step": 1410 }, { "epoch": 1.58, - "learning_rate": 0.0001752297794117647, - "loss": 4.6676, + "learning_rate": 1.7536764705882353e-05, + "loss": 0.255, "step": 1411 }, { "epoch": 1.58, - "learning_rate": 0.00017509191176470586, - "loss": 2.8011, + "learning_rate": 1.752297794117647e-05, + "loss": 0.1943, "step": 1412 }, { "epoch": 1.58, - "learning_rate": 0.00017495404411764704, - "loss": 2.8029, + "learning_rate": 1.7509191176470587e-05, + "loss": 0.201, "step": 1413 }, { "epoch": 1.59, - "learning_rate": 0.00017481617647058823, - "loss": 2.8013, + "learning_rate": 1.7495404411764704e-05, + "loss": 0.1893, "step": 1414 }, { "epoch": 1.59, - "learning_rate": 0.0001746783088235294, - "loss": 2.8073, + "learning_rate": 1.7481617647058825e-05, + "loss": 0.2482, "step": 1415 }, { "epoch": 1.59, - "learning_rate": 0.00017454044117647057, - "loss": 2.8786, + "learning_rate": 1.746783088235294e-05, + "loss": 0.2356, "step": 1416 }, { "epoch": 1.59, - "learning_rate": 0.00017440257352941175, - "loss": 3.3093, + "learning_rate": 1.745404411764706e-05, + "loss": 0.2897, "step": 1417 }, { "epoch": 1.59, - "learning_rate": 0.00017426470588235293, - "loss": 2.8256, + "learning_rate": 1.7440257352941176e-05, + "loss": 0.1949, "step": 1418 }, { "epoch": 1.59, - "learning_rate": 0.00017412683823529412, - "loss": 2.8223, + "learning_rate": 1.7426470588235296e-05, + "loss": 0.1752, "step": 1419 }, { "epoch": 1.59, - "learning_rate": 0.00017398897058823527, - "loss": 2.8007, + "learning_rate": 1.7412683823529413e-05, + "loss": 0.1919, "step": 1420 }, { "epoch": 1.59, - "learning_rate": 0.00017385110294117646, - "loss": 2.8176, + "learning_rate": 1.739889705882353e-05, + "loss": 0.2098, "step": 1421 }, { "epoch": 1.59, - "learning_rate": 0.00017371323529411764, - "loss": 2.8421, + "learning_rate": 1.7385110294117647e-05, + "loss": 0.2219, "step": 1422 }, { "epoch": 1.6, - "learning_rate": 0.00017357536764705882, - "loss": 2.895, + "learning_rate": 1.7371323529411767e-05, + "loss": 0.2873, "step": 1423 }, { "epoch": 1.6, - "learning_rate": 0.00017343749999999998, - "loss": 3.7592, + "learning_rate": 1.7357536764705884e-05, + "loss": 0.2335, "step": 1424 }, { "epoch": 1.6, - "learning_rate": 0.00017329963235294116, - "loss": 2.8121, + "learning_rate": 1.734375e-05, + "loss": 0.185, "step": 1425 }, { "epoch": 1.6, - "learning_rate": 0.00017316176470588232, - "loss": 2.8284, + "learning_rate": 1.732996323529412e-05, + "loss": 0.188, "step": 1426 }, { "epoch": 1.6, - "learning_rate": 0.00017302389705882353, - "loss": 2.8209, + "learning_rate": 1.7316176470588235e-05, + "loss": 0.1757, "step": 1427 }, { "epoch": 1.6, - "learning_rate": 0.0001728860294117647, - "loss": 2.8066, + "learning_rate": 1.7302389705882356e-05, + "loss": 0.171, "step": 1428 }, { "epoch": 1.6, - "learning_rate": 0.00017274816176470587, - "loss": 2.9088, + "learning_rate": 1.7288602941176473e-05, + "loss": 0.2436, "step": 1429 }, { "epoch": 1.6, - "learning_rate": 0.00017261029411764703, - "loss": 4.3231, + "learning_rate": 1.7274816176470587e-05, + "loss": 0.2662, "step": 1430 }, { "epoch": 1.6, - "learning_rate": 0.00017247242647058824, - "loss": 2.8151, + "learning_rate": 1.7261029411764704e-05, + "loss": 0.2023, "step": 1431 }, { "epoch": 1.61, - "learning_rate": 0.0001723345588235294, - "loss": 2.8136, + "learning_rate": 1.7247242647058824e-05, + "loss": 0.2109, "step": 1432 }, { "epoch": 1.61, - "learning_rate": 0.00017219669117647058, - "loss": 2.8128, + "learning_rate": 1.723345588235294e-05, + "loss": 0.1559, "step": 1433 }, { "epoch": 1.61, - "learning_rate": 0.00017205882352941174, - "loss": 2.8259, + "learning_rate": 1.7219669117647058e-05, + "loss": 0.2008, "step": 1434 }, { "epoch": 1.61, - "learning_rate": 0.00017192095588235295, - "loss": 2.8645, + "learning_rate": 1.7205882352941175e-05, + "loss": 0.2028, "step": 1435 }, { "epoch": 1.61, - "learning_rate": 0.0001717830882352941, - "loss": 4.3276, + "learning_rate": 1.7192095588235295e-05, + "loss": 0.2721, "step": 1436 }, { "epoch": 1.61, - "learning_rate": 0.0001716452205882353, - "loss": 2.8033, + "learning_rate": 1.7178308823529412e-05, + "loss": 0.1897, "step": 1437 }, { "epoch": 1.61, - "learning_rate": 0.00017150735294117644, - "loss": 2.8305, + "learning_rate": 1.716452205882353e-05, + "loss": 0.1915, "step": 1438 }, { "epoch": 1.61, - "learning_rate": 0.00017136948529411765, - "loss": 2.8269, + "learning_rate": 1.7150735294117646e-05, + "loss": 0.1987, "step": 1439 }, { "epoch": 1.61, - "learning_rate": 0.0001712316176470588, - "loss": 2.813, + "learning_rate": 1.7136948529411767e-05, + "loss": 0.2044, "step": 1440 }, { "epoch": 1.62, - "learning_rate": 0.00017109375, - "loss": 2.8341, + "learning_rate": 1.7123161764705884e-05, + "loss": 0.1909, "step": 1441 }, { "epoch": 1.62, - "learning_rate": 0.00017095588235294115, - "loss": 3.278, + "learning_rate": 1.7109375e-05, + "loss": 0.2913, "step": 1442 }, { "epoch": 1.62, - "learning_rate": 0.00017081801470588234, - "loss": 2.816, + "learning_rate": 1.7095588235294118e-05, + "loss": 0.1885, "step": 1443 }, { "epoch": 1.62, - "learning_rate": 0.00017068014705882352, - "loss": 2.8151, + "learning_rate": 1.7081801470588235e-05, + "loss": 0.1966, "step": 1444 }, { "epoch": 1.62, - "learning_rate": 0.0001705422794117647, - "loss": 2.7981, + "learning_rate": 1.7068014705882355e-05, + "loss": 0.1804, "step": 1445 }, { "epoch": 1.62, - "learning_rate": 0.00017040441176470586, - "loss": 2.8415, + "learning_rate": 1.7054227941176472e-05, + "loss": 0.1875, "step": 1446 }, { "epoch": 1.62, - "learning_rate": 0.00017026654411764704, - "loss": 2.8007, + "learning_rate": 1.704044117647059e-05, + "loss": 0.2226, "step": 1447 }, { "epoch": 1.62, - "learning_rate": 0.00017012867647058823, - "loss": 2.9488, + "learning_rate": 1.7026654411764706e-05, + "loss": 0.225, "step": 1448 }, { "epoch": 1.62, - "learning_rate": 0.0001699908088235294, - "loss": 3.8687, + "learning_rate": 1.7012867647058827e-05, + "loss": 0.2607, "step": 1449 }, { "epoch": 1.63, - "learning_rate": 0.00016985294117647057, - "loss": 2.8153, + "learning_rate": 1.699908088235294e-05, + "loss": 0.2, "step": 1450 }, { "epoch": 1.63, - "learning_rate": 0.00016971507352941175, - "loss": 2.8012, + "learning_rate": 1.6985294117647057e-05, + "loss": 0.2001, "step": 1451 }, { "epoch": 1.63, - "learning_rate": 0.00016957720588235293, - "loss": 2.8093, + "learning_rate": 1.6971507352941174e-05, + "loss": 0.195, "step": 1452 }, { "epoch": 1.63, - "learning_rate": 0.00016943933823529412, - "loss": 2.8162, + "learning_rate": 1.6957720588235295e-05, + "loss": 0.2056, "step": 1453 }, { "epoch": 1.63, - "learning_rate": 0.00016930147058823527, - "loss": 2.9138, + "learning_rate": 1.6943933823529412e-05, + "loss": 0.1825, "step": 1454 }, { "epoch": 1.63, - "learning_rate": 0.00016916360294117646, - "loss": 3.9884, + "learning_rate": 1.693014705882353e-05, + "loss": 0.2294, "step": 1455 }, { "epoch": 1.63, - "learning_rate": 0.00016902573529411764, - "loss": 2.817, + "learning_rate": 1.6916360294117646e-05, + "loss": 0.2317, "step": 1456 }, { "epoch": 1.63, - "learning_rate": 0.00016888786764705883, - "loss": 2.8077, + "learning_rate": 1.6902573529411766e-05, + "loss": 0.2043, "step": 1457 }, { "epoch": 1.63, - "learning_rate": 0.00016874999999999998, - "loss": 2.8123, + "learning_rate": 1.6888786764705883e-05, + "loss": 0.2005, "step": 1458 }, { "epoch": 1.64, - "learning_rate": 0.00016861213235294117, - "loss": 2.8143, + "learning_rate": 1.6875e-05, + "loss": 0.1956, "step": 1459 }, { "epoch": 1.64, - "learning_rate": 0.00016847426470588232, - "loss": 2.8521, + "learning_rate": 1.6861213235294117e-05, + "loss": 0.1854, "step": 1460 }, { "epoch": 1.64, - "learning_rate": 0.00016833639705882353, - "loss": 4.4464, + "learning_rate": 1.6847426470588234e-05, + "loss": 0.2703, "step": 1461 }, { "epoch": 1.64, - "learning_rate": 0.0001681985294117647, - "loss": 2.8163, + "learning_rate": 1.6833639705882355e-05, + "loss": 0.1791, "step": 1462 }, { "epoch": 1.64, - "learning_rate": 0.00016806066176470587, - "loss": 2.8189, + "learning_rate": 1.681985294117647e-05, + "loss": 0.1666, "step": 1463 }, { "epoch": 1.64, - "learning_rate": 0.00016792279411764703, - "loss": 2.8166, + "learning_rate": 1.680606617647059e-05, + "loss": 0.1955, "step": 1464 }, { "epoch": 1.64, - "learning_rate": 0.00016778492647058824, - "loss": 2.8204, + "learning_rate": 1.6792279411764706e-05, + "loss": 0.2102, "step": 1465 }, { "epoch": 1.64, - "learning_rate": 0.0001676470588235294, - "loss": 2.809, + "learning_rate": 1.6778492647058826e-05, + "loss": 0.1787, "step": 1466 }, { "epoch": 1.64, - "learning_rate": 0.00016750919117647058, - "loss": 3.1769, + "learning_rate": 1.6764705882352943e-05, + "loss": 0.2892, "step": 1467 }, { "epoch": 1.65, - "learning_rate": 0.00016737132352941174, - "loss": 2.8016, + "learning_rate": 1.675091911764706e-05, + "loss": 0.2236, "step": 1468 }, { "epoch": 1.65, - "learning_rate": 0.00016723345588235295, - "loss": 2.7884, + "learning_rate": 1.6737132352941177e-05, + "loss": 0.1724, "step": 1469 }, { "epoch": 1.65, - "learning_rate": 0.0001670955882352941, - "loss": 2.8003, + "learning_rate": 1.6723345588235298e-05, + "loss": 0.1954, "step": 1470 }, { "epoch": 1.65, - "learning_rate": 0.0001669577205882353, - "loss": 2.8098, + "learning_rate": 1.670955882352941e-05, + "loss": 0.2083, "step": 1471 }, { "epoch": 1.65, - "learning_rate": 0.00016681985294117644, - "loss": 2.8174, + "learning_rate": 1.6695772058823528e-05, + "loss": 0.205, "step": 1472 }, { "epoch": 1.65, - "learning_rate": 0.00016668198529411766, - "loss": 2.8856, + "learning_rate": 1.6681985294117645e-05, + "loss": 0.2066, "step": 1473 }, { "epoch": 1.65, - "learning_rate": 0.0001665441176470588, - "loss": 3.4919, + "learning_rate": 1.6668198529411766e-05, + "loss": 0.1989, "step": 1474 }, { "epoch": 1.65, - "learning_rate": 0.00016640625, - "loss": 2.7923, + "learning_rate": 1.6654411764705883e-05, + "loss": 0.2061, "step": 1475 }, { "epoch": 1.65, - "learning_rate": 0.00016626838235294115, - "loss": 2.8229, + "learning_rate": 1.6640625e-05, + "loss": 0.2072, "step": 1476 }, { "epoch": 1.66, - "learning_rate": 0.0001661305147058823, - "loss": 2.8206, + "learning_rate": 1.6626838235294117e-05, + "loss": 0.187, "step": 1477 }, { "epoch": 1.66, - "learning_rate": 0.00016599264705882352, - "loss": 2.8272, + "learning_rate": 1.6613051470588234e-05, + "loss": 0.2019, "step": 1478 }, { "epoch": 1.66, - "learning_rate": 0.0001658547794117647, - "loss": 2.8423, + "learning_rate": 1.6599264705882354e-05, + "loss": 0.1662, "step": 1479 }, { "epoch": 1.66, - "learning_rate": 0.00016571691176470586, - "loss": 3.7986, + "learning_rate": 1.658547794117647e-05, + "loss": 0.2508, "step": 1480 }, { "epoch": 1.66, - "learning_rate": 0.00016557904411764704, - "loss": 2.8237, + "learning_rate": 1.6571691176470588e-05, + "loss": 0.1872, "step": 1481 }, { "epoch": 1.66, - "learning_rate": 0.00016544117647058823, - "loss": 2.8132, + "learning_rate": 1.6557904411764705e-05, + "loss": 0.1916, "step": 1482 }, { "epoch": 1.66, - "learning_rate": 0.0001653033088235294, - "loss": 2.8237, + "learning_rate": 1.6544117647058825e-05, + "loss": 0.1612, "step": 1483 }, { "epoch": 1.66, - "learning_rate": 0.00016516544117647057, - "loss": 2.8309, + "learning_rate": 1.6530330882352942e-05, + "loss": 0.1881, "step": 1484 }, { "epoch": 1.66, - "learning_rate": 0.00016502757352941175, - "loss": 2.845, + "learning_rate": 1.651654411764706e-05, + "loss": 0.2127, "step": 1485 }, { "epoch": 1.67, - "learning_rate": 0.00016488970588235294, - "loss": 4.3636, + "learning_rate": 1.6502757352941177e-05, + "loss": 0.2503, "step": 1486 }, { "epoch": 1.67, - "learning_rate": 0.00016475183823529412, - "loss": 2.8127, + "learning_rate": 1.6488970588235297e-05, + "loss": 0.1894, "step": 1487 }, { "epoch": 1.67, - "learning_rate": 0.00016461397058823528, - "loss": 2.8094, + "learning_rate": 1.6475183823529414e-05, + "loss": 0.179, "step": 1488 }, { "epoch": 1.67, - "learning_rate": 0.00016447610294117646, - "loss": 2.8102, + "learning_rate": 1.646139705882353e-05, + "loss": 0.186, "step": 1489 }, { "epoch": 1.67, - "learning_rate": 0.00016433823529411764, - "loss": 2.8071, + "learning_rate": 1.6447610294117648e-05, + "loss": 0.2083, "step": 1490 }, { "epoch": 1.67, - "learning_rate": 0.00016420036764705883, - "loss": 2.843, + "learning_rate": 1.643382352941177e-05, + "loss": 0.2015, "step": 1491 }, { "epoch": 1.67, - "learning_rate": 0.00016406249999999998, - "loss": 3.1001, + "learning_rate": 1.6420036764705882e-05, + "loss": 0.2427, "step": 1492 }, { "epoch": 1.67, - "learning_rate": 0.00016392463235294117, - "loss": 2.8004, + "learning_rate": 1.640625e-05, + "loss": 0.2147, "step": 1493 }, { "epoch": 1.67, - "learning_rate": 0.00016378676470588232, - "loss": 2.81, + "learning_rate": 1.6392463235294116e-05, + "loss": 0.2036, "step": 1494 }, { "epoch": 1.68, - "learning_rate": 0.00016364889705882353, - "loss": 2.8079, + "learning_rate": 1.6378676470588233e-05, + "loss": 0.1779, "step": 1495 }, { "epoch": 1.68, - "learning_rate": 0.0001635110294117647, - "loss": 2.8388, + "learning_rate": 1.6364889705882353e-05, + "loss": 0.2044, "step": 1496 }, { "epoch": 1.68, - "learning_rate": 0.00016337316176470587, - "loss": 2.858, + "learning_rate": 1.635110294117647e-05, + "loss": 0.2058, "step": 1497 }, { "epoch": 1.68, - "learning_rate": 0.00016323529411764703, - "loss": 3.0182, + "learning_rate": 1.6337316176470587e-05, + "loss": 0.233, "step": 1498 }, { "epoch": 1.68, - "learning_rate": 0.00016309742647058824, - "loss": 3.6434, + "learning_rate": 1.6323529411764704e-05, + "loss": 0.1929, "step": 1499 }, { "epoch": 1.68, - "learning_rate": 0.0001629595588235294, - "loss": 2.8643, + "learning_rate": 1.6309742647058825e-05, + "loss": 0.1837, "step": 1500 }, { "epoch": 1.68, - "eval_loss": 3.5116281509399414, - "eval_runtime": 48.0897, - "eval_samples_per_second": 56.207, - "eval_steps_per_second": 0.894, - "eval_wer": 1.0, + "eval_loss": 0.22951027750968933, + "eval_runtime": 47.5344, + "eval_samples_per_second": 56.864, + "eval_steps_per_second": 0.905, + "eval_wer": 0.20291533399507372, "step": 1500 }, { "epoch": 1.68, - "learning_rate": 0.00016282169117647058, - "loss": 2.8781, + "learning_rate": 1.6295955882352942e-05, + "loss": 0.1976, "step": 1501 }, { "epoch": 1.68, - "learning_rate": 0.00016268382352941174, - "loss": 2.8752, + "learning_rate": 1.628216911764706e-05, + "loss": 0.1951, "step": 1502 }, { "epoch": 1.68, - "learning_rate": 0.00016254595588235295, - "loss": 2.8548, + "learning_rate": 1.6268382352941176e-05, + "loss": 0.1977, "step": 1503 }, { "epoch": 1.69, - "learning_rate": 0.0001624080882352941, - "loss": 2.8439, + "learning_rate": 1.6254595588235296e-05, + "loss": 0.2239, "step": 1504 }, { "epoch": 1.69, - "learning_rate": 0.0001622702205882353, - "loss": 3.5817, + "learning_rate": 1.6240808823529413e-05, + "loss": 0.2256, "step": 1505 }, { "epoch": 1.69, - "learning_rate": 0.00016213235294117645, - "loss": 2.8369, + "learning_rate": 1.622702205882353e-05, + "loss": 0.1852, "step": 1506 }, { "epoch": 1.69, - "learning_rate": 0.00016199448529411766, - "loss": 2.8463, + "learning_rate": 1.6213235294117647e-05, + "loss": 0.1898, "step": 1507 }, { "epoch": 1.69, - "learning_rate": 0.00016185661764705881, - "loss": 2.8403, + "learning_rate": 1.6199448529411768e-05, + "loss": 0.1826, "step": 1508 }, { "epoch": 1.69, - "learning_rate": 0.00016171875, - "loss": 2.8281, + "learning_rate": 1.6185661764705885e-05, + "loss": 0.2111, "step": 1509 }, { "epoch": 1.69, - "learning_rate": 0.00016158088235294115, - "loss": 2.8606, + "learning_rate": 1.6171875000000002e-05, + "loss": 0.2066, "step": 1510 }, { "epoch": 1.69, - "learning_rate": 0.0001614430147058823, - "loss": 4.5665, + "learning_rate": 1.615808823529412e-05, + "loss": 0.2986, "step": 1511 }, { "epoch": 1.7, - "learning_rate": 0.00016130514705882352, - "loss": 2.8239, + "learning_rate": 1.6144301470588236e-05, + "loss": 0.1949, "step": 1512 }, { "epoch": 1.7, - "learning_rate": 0.0001611672794117647, - "loss": 2.809, + "learning_rate": 1.6130514705882353e-05, + "loss": 0.1933, "step": 1513 }, { "epoch": 1.7, - "learning_rate": 0.00016102941176470586, - "loss": 2.8242, + "learning_rate": 1.611672794117647e-05, + "loss": 0.1987, "step": 1514 }, { "epoch": 1.7, - "learning_rate": 0.00016089154411764702, - "loss": 2.792, + "learning_rate": 1.6102941176470587e-05, + "loss": 0.1948, "step": 1515 }, { "epoch": 1.7, - "learning_rate": 0.00016075367647058823, - "loss": 2.839, + "learning_rate": 1.6089154411764704e-05, + "loss": 0.2268, "step": 1516 }, { "epoch": 1.7, - "learning_rate": 0.0001606158088235294, - "loss": 3.1182, + "learning_rate": 1.6075367647058824e-05, + "loss": 0.2202, "step": 1517 }, { "epoch": 1.7, - "learning_rate": 0.00016047794117647057, - "loss": 2.8167, + "learning_rate": 1.606158088235294e-05, + "loss": 0.2047, "step": 1518 }, { "epoch": 1.7, - "learning_rate": 0.00016034007352941173, - "loss": 2.8324, + "learning_rate": 1.6047794117647058e-05, + "loss": 0.2391, "step": 1519 }, { "epoch": 1.7, - "learning_rate": 0.00016020220588235294, - "loss": 2.8279, + "learning_rate": 1.6034007352941175e-05, + "loss": 0.186, "step": 1520 }, { "epoch": 1.71, - "learning_rate": 0.00016006433823529412, - "loss": 2.8236, + "learning_rate": 1.6020220588235296e-05, + "loss": 0.209, "step": 1521 }, { "epoch": 1.71, - "learning_rate": 0.00015992647058823528, - "loss": 2.8287, + "learning_rate": 1.6006433823529413e-05, + "loss": 0.1835, "step": 1522 }, { "epoch": 1.71, - "learning_rate": 0.00015978860294117643, - "loss": 2.9113, + "learning_rate": 1.599264705882353e-05, + "loss": 0.2168, "step": 1523 }, { "epoch": 1.71, - "learning_rate": 0.00015965073529411764, - "loss": 3.7218, + "learning_rate": 1.5978860294117647e-05, + "loss": 0.2081, "step": 1524 }, { "epoch": 1.71, - "learning_rate": 0.00015951286764705883, - "loss": 2.8065, + "learning_rate": 1.5965073529411767e-05, + "loss": 0.1934, "step": 1525 }, { "epoch": 1.71, - "learning_rate": 0.00015937499999999998, - "loss": 2.8007, + "learning_rate": 1.5951286764705884e-05, + "loss": 0.1769, "step": 1526 }, { "epoch": 1.71, - "learning_rate": 0.00015923713235294117, - "loss": 2.8068, + "learning_rate": 1.59375e-05, + "loss": 0.1759, "step": 1527 }, { "epoch": 1.71, - "learning_rate": 0.00015909926470588232, - "loss": 2.8001, + "learning_rate": 1.5923713235294118e-05, + "loss": 0.2154, "step": 1528 }, { "epoch": 1.71, - "learning_rate": 0.00015896139705882354, - "loss": 2.8861, + "learning_rate": 1.5909926470588235e-05, + "loss": 0.2007, "step": 1529 }, { "epoch": 1.72, - "learning_rate": 0.0001588235294117647, - "loss": 3.9605, + "learning_rate": 1.5896139705882356e-05, + "loss": 0.2332, "step": 1530 }, { "epoch": 1.72, - "learning_rate": 0.00015868566176470588, - "loss": 2.8094, + "learning_rate": 1.5882352941176473e-05, + "loss": 0.2201, "step": 1531 }, { "epoch": 1.72, - "learning_rate": 0.00015854779411764703, - "loss": 2.7899, + "learning_rate": 1.586856617647059e-05, + "loss": 0.1773, "step": 1532 }, { "epoch": 1.72, - "learning_rate": 0.00015840992647058824, - "loss": 2.8156, + "learning_rate": 1.5854779411764703e-05, + "loss": 0.1742, "step": 1533 }, { "epoch": 1.72, - "learning_rate": 0.0001582720588235294, - "loss": 2.8108, + "learning_rate": 1.5840992647058824e-05, + "loss": 0.1892, "step": 1534 }, { "epoch": 1.72, - "learning_rate": 0.00015813419117647058, - "loss": 2.8582, + "learning_rate": 1.582720588235294e-05, + "loss": 0.2012, "step": 1535 }, { "epoch": 1.72, - "learning_rate": 0.00015799632352941174, - "loss": 3.9647, + "learning_rate": 1.5813419117647058e-05, + "loss": 0.2432, "step": 1536 }, { "epoch": 1.72, - "learning_rate": 0.00015785845588235295, - "loss": 2.8095, + "learning_rate": 1.5799632352941175e-05, + "loss": 0.2496, "step": 1537 }, { "epoch": 1.72, - "learning_rate": 0.0001577205882352941, - "loss": 2.8179, + "learning_rate": 1.5785845588235295e-05, + "loss": 0.2143, "step": 1538 }, { "epoch": 1.73, - "learning_rate": 0.0001575827205882353, - "loss": 2.7974, + "learning_rate": 1.5772058823529412e-05, + "loss": 0.1959, "step": 1539 }, { "epoch": 1.73, - "learning_rate": 0.00015744485294117645, - "loss": 2.803, + "learning_rate": 1.575827205882353e-05, + "loss": 0.1837, "step": 1540 }, { "epoch": 1.73, - "learning_rate": 0.00015730698529411766, - "loss": 2.8193, + "learning_rate": 1.5744485294117646e-05, + "loss": 0.1661, "step": 1541 }, { "epoch": 1.73, - "learning_rate": 0.00015716911764705881, - "loss": 3.2058, + "learning_rate": 1.5730698529411767e-05, + "loss": 0.2454, "step": 1542 }, { "epoch": 1.73, - "learning_rate": 0.00015703125, - "loss": 2.807, + "learning_rate": 1.5716911764705884e-05, + "loss": 0.1948, "step": 1543 }, { "epoch": 1.73, - "learning_rate": 0.00015689338235294116, - "loss": 2.7974, + "learning_rate": 1.5703125e-05, + "loss": 0.1875, "step": 1544 }, { "epoch": 1.73, - "learning_rate": 0.0001567555147058823, - "loss": 2.8116, + "learning_rate": 1.5689338235294118e-05, + "loss": 0.1963, "step": 1545 }, { "epoch": 1.73, - "learning_rate": 0.00015661764705882352, - "loss": 2.8149, + "learning_rate": 1.5675551470588235e-05, + "loss": 0.1592, "step": 1546 }, { "epoch": 1.73, - "learning_rate": 0.0001564797794117647, - "loss": 2.8198, + "learning_rate": 1.5661764705882355e-05, + "loss": 0.1761, "step": 1547 }, { "epoch": 1.74, - "learning_rate": 0.00015634191176470586, - "loss": 2.9131, + "learning_rate": 1.5647977941176472e-05, + "loss": 0.2054, "step": 1548 }, { "epoch": 1.74, - "learning_rate": 0.00015620404411764702, - "loss": 3.8575, + "learning_rate": 1.563419117647059e-05, + "loss": 0.2569, "step": 1549 }, { "epoch": 1.74, - "learning_rate": 0.00015606617647058823, - "loss": 2.796, + "learning_rate": 1.5620404411764706e-05, + "loss": 0.139, "step": 1550 }, { "epoch": 1.74, - "learning_rate": 0.00015592830882352941, - "loss": 2.7967, + "learning_rate": 1.5606617647058826e-05, + "loss": 0.189, "step": 1551 }, { "epoch": 1.74, - "learning_rate": 0.00015579044117647057, - "loss": 2.8237, + "learning_rate": 1.5592830882352943e-05, + "loss": 0.2013, "step": 1552 }, { "epoch": 1.74, - "learning_rate": 0.00015565257352941173, - "loss": 2.7996, + "learning_rate": 1.557904411764706e-05, + "loss": 0.2141, "step": 1553 }, { "epoch": 1.74, - "learning_rate": 0.00015551470588235294, - "loss": 2.9246, + "learning_rate": 1.5565257352941174e-05, + "loss": 0.1827, "step": 1554 }, { "epoch": 1.74, - "learning_rate": 0.00015537683823529412, - "loss": 4.2402, + "learning_rate": 1.5551470588235294e-05, + "loss": 0.2281, "step": 1555 }, { "epoch": 1.74, - "learning_rate": 0.00015523897058823528, - "loss": 2.806, + "learning_rate": 1.553768382352941e-05, + "loss": 0.201, "step": 1556 }, { "epoch": 1.75, - "learning_rate": 0.00015510110294117643, - "loss": 2.802, + "learning_rate": 1.552389705882353e-05, + "loss": 0.1588, "step": 1557 }, { "epoch": 1.75, - "learning_rate": 0.00015496323529411765, - "loss": 2.8097, + "learning_rate": 1.5510110294117645e-05, + "loss": 0.1947, "step": 1558 }, { "epoch": 1.75, - "learning_rate": 0.00015482536764705883, - "loss": 2.8144, + "learning_rate": 1.5496323529411766e-05, + "loss": 0.1955, "step": 1559 }, { "epoch": 1.75, - "learning_rate": 0.00015468749999999999, - "loss": 2.8312, + "learning_rate": 1.5482536764705883e-05, + "loss": 0.1953, "step": 1560 }, { "epoch": 1.75, - "learning_rate": 0.00015454963235294114, - "loss": 4.1358, + "learning_rate": 1.546875e-05, + "loss": 0.2254, "step": 1561 }, { "epoch": 1.75, - "learning_rate": 0.00015441176470588233, - "loss": 2.8155, + "learning_rate": 1.5454963235294117e-05, + "loss": 0.2095, "step": 1562 }, { "epoch": 1.75, - "learning_rate": 0.00015427389705882354, - "loss": 2.8237, + "learning_rate": 1.5441176470588234e-05, + "loss": 0.1854, "step": 1563 }, { "epoch": 1.75, - "learning_rate": 0.0001541360294117647, - "loss": 2.8195, + "learning_rate": 1.5427389705882354e-05, + "loss": 0.1864, "step": 1564 }, { "epoch": 1.75, - "learning_rate": 0.00015399816176470585, - "loss": 2.8285, + "learning_rate": 1.541360294117647e-05, + "loss": 0.1949, "step": 1565 }, { "epoch": 1.76, - "learning_rate": 0.00015386029411764703, - "loss": 2.8316, + "learning_rate": 1.539981617647059e-05, + "loss": 0.1859, "step": 1566 }, { "epoch": 1.76, - "learning_rate": 0.00015372242647058824, - "loss": 3.2287, + "learning_rate": 1.5386029411764705e-05, + "loss": 0.2488, "step": 1567 }, { "epoch": 1.76, - "learning_rate": 0.0001535845588235294, - "loss": 2.8107, + "learning_rate": 1.5372242647058826e-05, + "loss": 0.2293, "step": 1568 }, { "epoch": 1.76, - "learning_rate": 0.00015344669117647056, - "loss": 2.8005, + "learning_rate": 1.5358455882352943e-05, + "loss": 0.1661, "step": 1569 }, { "epoch": 1.76, - "learning_rate": 0.00015330882352941174, - "loss": 2.8105, + "learning_rate": 1.534466911764706e-05, + "loss": 0.1968, "step": 1570 }, { "epoch": 1.76, - "learning_rate": 0.00015317095588235295, - "loss": 2.7955, + "learning_rate": 1.5330882352941177e-05, + "loss": 0.2269, "step": 1571 }, { "epoch": 1.76, - "learning_rate": 0.0001530330882352941, - "loss": 2.8125, + "learning_rate": 1.5317095588235297e-05, + "loss": 0.1974, "step": 1572 }, { "epoch": 1.76, - "learning_rate": 0.0001528952205882353, - "loss": 3.0142, + "learning_rate": 1.5303308823529414e-05, + "loss": 0.2461, "step": 1573 }, { "epoch": 1.76, - "learning_rate": 0.00015275735294117645, - "loss": 3.8451, + "learning_rate": 1.528952205882353e-05, + "loss": 0.2209, "step": 1574 }, { "epoch": 1.77, - "learning_rate": 0.00015261948529411766, - "loss": 2.8073, + "learning_rate": 1.5275735294117645e-05, + "loss": 0.1982, "step": 1575 }, { "epoch": 1.77, - "learning_rate": 0.00015248161764705882, - "loss": 2.8058, + "learning_rate": 1.5261948529411765e-05, + "loss": 0.1905, "step": 1576 }, { "epoch": 1.77, - "learning_rate": 0.00015234375, - "loss": 2.8114, + "learning_rate": 1.5248161764705884e-05, + "loss": 0.1795, "step": 1577 }, { "epoch": 1.77, - "learning_rate": 0.00015220588235294116, - "loss": 2.8341, + "learning_rate": 1.5234375000000001e-05, + "loss": 0.2289, "step": 1578 }, { "epoch": 1.77, - "learning_rate": 0.0001520680147058823, - "loss": 2.8394, + "learning_rate": 1.5220588235294118e-05, + "loss": 0.1788, "step": 1579 }, { "epoch": 1.77, - "learning_rate": 0.00015193014705882352, - "loss": 4.2367, + "learning_rate": 1.5206801470588233e-05, + "loss": 0.227, "step": 1580 }, { "epoch": 1.77, - "learning_rate": 0.0001517922794117647, - "loss": 2.7942, + "learning_rate": 1.5193014705882354e-05, + "loss": 0.1738, "step": 1581 }, { "epoch": 1.77, - "learning_rate": 0.00015165441176470586, - "loss": 2.8033, + "learning_rate": 1.517922794117647e-05, + "loss": 0.1956, "step": 1582 }, { "epoch": 1.77, - "learning_rate": 0.00015151654411764702, - "loss": 2.8037, + "learning_rate": 1.5165441176470588e-05, + "loss": 0.2114, "step": 1583 }, { "epoch": 1.78, - "learning_rate": 0.00015137867647058823, - "loss": 2.8312, + "learning_rate": 1.5151654411764705e-05, + "loss": 0.1912, "step": 1584 }, { "epoch": 1.78, - "learning_rate": 0.00015124080882352942, - "loss": 2.863, + "learning_rate": 1.5137867647058825e-05, + "loss": 0.2319, "step": 1585 }, { "epoch": 1.78, - "learning_rate": 0.00015110294117647057, - "loss": 3.8798, + "learning_rate": 1.5124080882352942e-05, + "loss": 0.3092, "step": 1586 }, { "epoch": 1.78, - "learning_rate": 0.00015096507352941173, - "loss": 2.8157, + "learning_rate": 1.511029411764706e-05, + "loss": 0.184, "step": 1587 }, { "epoch": 1.78, - "learning_rate": 0.00015082720588235294, - "loss": 2.8039, + "learning_rate": 1.5096507352941176e-05, + "loss": 0.1755, "step": 1588 }, { "epoch": 1.78, - "learning_rate": 0.00015068933823529412, - "loss": 2.8187, + "learning_rate": 1.5082720588235297e-05, + "loss": 0.1896, "step": 1589 }, { "epoch": 1.78, - "learning_rate": 0.00015055147058823528, - "loss": 2.81, + "learning_rate": 1.5068933823529412e-05, + "loss": 0.2194, "step": 1590 }, { "epoch": 1.78, - "learning_rate": 0.00015041360294117644, - "loss": 2.8455, + "learning_rate": 1.5055147058823529e-05, + "loss": 0.2246, "step": 1591 }, { "epoch": 1.78, - "learning_rate": 0.00015027573529411765, - "loss": 3.2046, + "learning_rate": 1.5041360294117646e-05, + "loss": 0.2379, "step": 1592 }, { "epoch": 1.79, - "learning_rate": 0.00015013786764705883, - "loss": 2.7994, + "learning_rate": 1.5027573529411766e-05, + "loss": 0.2199, "step": 1593 }, { "epoch": 1.79, - "learning_rate": 0.00015, - "loss": 2.8226, + "learning_rate": 1.5013786764705883e-05, + "loss": 0.1729, "step": 1594 }, { "epoch": 1.79, - "learning_rate": 0.00014986213235294114, - "loss": 2.8077, + "learning_rate": 1.5e-05, + "loss": 0.1578, "step": 1595 }, { "epoch": 1.79, - "learning_rate": 0.00014972426470588235, - "loss": 2.8097, + "learning_rate": 1.4986213235294117e-05, + "loss": 0.1737, "step": 1596 }, { "epoch": 1.79, - "learning_rate": 0.0001495863970588235, - "loss": 2.8297, + "learning_rate": 1.4972426470588236e-05, + "loss": 0.2319, "step": 1597 }, { "epoch": 1.79, - "learning_rate": 0.0001494485294117647, - "loss": 2.9845, + "learning_rate": 1.4958639705882353e-05, + "loss": 0.2856, "step": 1598 }, { "epoch": 1.79, - "learning_rate": 0.00014931066176470585, - "loss": 3.6271, + "learning_rate": 1.4944852941176472e-05, + "loss": 0.1843, "step": 1599 }, { "epoch": 1.79, - "learning_rate": 0.00014917279411764706, - "loss": 2.8265, + "learning_rate": 1.4931066176470589e-05, + "loss": 0.1757, "step": 1600 }, { "epoch": 1.79, - "eval_loss": 3.632187843322754, - "eval_runtime": 47.9962, - "eval_samples_per_second": 56.317, - "eval_steps_per_second": 0.896, - "eval_wer": 1.0, + "eval_loss": 0.224042609333992, + "eval_runtime": 47.3067, + "eval_samples_per_second": 57.138, + "eval_steps_per_second": 0.909, + "eval_wer": 0.19490092276019264, "step": 1600 }, { "epoch": 1.79, - "learning_rate": 0.00014903492647058822, - "loss": 2.8297, + "learning_rate": 1.4917279411764706e-05, + "loss": 0.216, "step": 1601 }, { "epoch": 1.8, - "learning_rate": 0.0001488970588235294, - "loss": 2.8506, + "learning_rate": 1.4903492647058823e-05, + "loss": 0.1865, "step": 1602 }, { "epoch": 1.8, - "learning_rate": 0.00014875919117647056, - "loss": 2.8499, + "learning_rate": 1.4889705882352942e-05, + "loss": 0.206, "step": 1603 }, { "epoch": 1.8, - "learning_rate": 0.00014862132352941177, - "loss": 2.8459, + "learning_rate": 1.4875919117647059e-05, + "loss": 0.2329, "step": 1604 }, { "epoch": 1.8, - "learning_rate": 0.00014848345588235293, - "loss": 3.8794, + "learning_rate": 1.4862132352941177e-05, + "loss": 0.2473, "step": 1605 }, { "epoch": 1.8, - "learning_rate": 0.0001483455882352941, - "loss": 2.8335, + "learning_rate": 1.4848345588235294e-05, + "loss": 0.175, "step": 1606 }, { "epoch": 1.8, - "learning_rate": 0.00014820772058823527, - "loss": 2.8276, + "learning_rate": 1.4834558823529413e-05, + "loss": 0.2083, "step": 1607 }, { "epoch": 1.8, - "learning_rate": 0.00014806985294117648, - "loss": 2.8012, + "learning_rate": 1.482077205882353e-05, + "loss": 0.2051, "step": 1608 }, { "epoch": 1.8, - "learning_rate": 0.00014793198529411763, - "loss": 2.7975, + "learning_rate": 1.4806985294117649e-05, + "loss": 0.1878, "step": 1609 }, { "epoch": 1.8, - "learning_rate": 0.00014779411764705882, - "loss": 2.8465, + "learning_rate": 1.4793198529411766e-05, + "loss": 0.2003, "step": 1610 }, { "epoch": 1.81, - "learning_rate": 0.00014765624999999997, - "loss": 4.7673, + "learning_rate": 1.4779411764705883e-05, + "loss": 0.2489, "step": 1611 }, { "epoch": 1.81, - "learning_rate": 0.00014751838235294116, - "loss": 2.8105, + "learning_rate": 1.4765625e-05, + "loss": 0.199, "step": 1612 }, { "epoch": 1.81, - "learning_rate": 0.00014738051470588234, - "loss": 2.8135, + "learning_rate": 1.4751838235294117e-05, + "loss": 0.1817, "step": 1613 }, { "epoch": 1.81, - "learning_rate": 0.0001472426470588235, - "loss": 2.82, + "learning_rate": 1.4738051470588235e-05, + "loss": 0.2208, "step": 1614 }, { "epoch": 1.81, - "learning_rate": 0.00014710477941176468, - "loss": 2.818, + "learning_rate": 1.4724264705882352e-05, + "loss": 0.1859, "step": 1615 }, { "epoch": 1.81, - "learning_rate": 0.00014696691176470587, - "loss": 2.8188, + "learning_rate": 1.4710477941176471e-05, + "loss": 0.2173, "step": 1616 }, { "epoch": 1.81, - "learning_rate": 0.00014682904411764705, - "loss": 3.1197, + "learning_rate": 1.4696691176470588e-05, + "loss": 0.1689, "step": 1617 }, { "epoch": 1.81, - "learning_rate": 0.0001466911764705882, - "loss": 2.8121, + "learning_rate": 1.4682904411764707e-05, + "loss": 0.1861, "step": 1618 }, { "epoch": 1.82, - "learning_rate": 0.00014655330882352942, - "loss": 2.8136, + "learning_rate": 1.4669117647058824e-05, + "loss": 0.1781, "step": 1619 }, { "epoch": 1.82, - "learning_rate": 0.00014641544117647057, - "loss": 2.8369, + "learning_rate": 1.4655330882352943e-05, + "loss": 0.1782, "step": 1620 }, { "epoch": 1.82, - "learning_rate": 0.00014627757352941176, - "loss": 2.8345, + "learning_rate": 1.4641544117647058e-05, + "loss": 0.1727, "step": 1621 }, { "epoch": 1.82, - "learning_rate": 0.0001461397058823529, - "loss": 2.8311, + "learning_rate": 1.4627757352941177e-05, + "loss": 0.2309, "step": 1622 }, { "epoch": 1.82, - "learning_rate": 0.00014600183823529412, - "loss": 2.9627, + "learning_rate": 1.4613970588235294e-05, + "loss": 0.241, "step": 1623 }, { "epoch": 1.82, - "learning_rate": 0.00014586397058823528, - "loss": 3.5565, + "learning_rate": 1.4600183823529412e-05, + "loss": 0.2122, "step": 1624 }, { "epoch": 1.82, - "learning_rate": 0.00014572610294117646, - "loss": 2.8249, + "learning_rate": 1.458639705882353e-05, + "loss": 0.2006, "step": 1625 }, { "epoch": 1.82, - "learning_rate": 0.00014558823529411762, - "loss": 2.8239, + "learning_rate": 1.4572610294117648e-05, + "loss": 0.1963, "step": 1626 }, { "epoch": 1.82, - "learning_rate": 0.00014545036764705883, - "loss": 2.8381, + "learning_rate": 1.4558823529411765e-05, + "loss": 0.2158, "step": 1627 }, { "epoch": 1.83, - "learning_rate": 0.0001453125, - "loss": 2.8274, + "learning_rate": 1.4545036764705884e-05, + "loss": 0.1935, "step": 1628 }, { "epoch": 1.83, - "learning_rate": 0.00014517463235294114, - "loss": 2.8935, + "learning_rate": 1.453125e-05, + "loss": 0.1868, "step": 1629 }, { "epoch": 1.83, - "learning_rate": 0.00014503676470588233, - "loss": 4.1261, + "learning_rate": 1.4517463235294118e-05, + "loss": 0.2067, "step": 1630 }, { "epoch": 1.83, - "learning_rate": 0.0001448988970588235, - "loss": 2.8071, + "learning_rate": 1.4503676470588237e-05, + "loss": 0.1693, "step": 1631 }, { "epoch": 1.83, - "learning_rate": 0.0001447610294117647, - "loss": 2.8113, + "learning_rate": 1.4489889705882352e-05, + "loss": 0.197, "step": 1632 }, { "epoch": 1.83, - "learning_rate": 0.00014462316176470585, - "loss": 2.8189, + "learning_rate": 1.447610294117647e-05, + "loss": 0.2121, "step": 1633 }, { "epoch": 1.83, - "learning_rate": 0.00014448529411764704, - "loss": 2.8306, + "learning_rate": 1.4462316176470588e-05, + "loss": 0.1837, "step": 1634 }, { "epoch": 1.83, - "learning_rate": 0.00014434742647058822, - "loss": 2.8366, + "learning_rate": 1.4448529411764706e-05, + "loss": 0.2163, "step": 1635 }, { "epoch": 1.83, - "learning_rate": 0.0001442095588235294, - "loss": 4.1006, + "learning_rate": 1.4434742647058823e-05, + "loss": 0.2724, "step": 1636 }, { "epoch": 1.84, - "learning_rate": 0.00014407169117647056, - "loss": 2.8032, + "learning_rate": 1.4420955882352942e-05, + "loss": 0.1829, "step": 1637 }, { "epoch": 1.84, - "learning_rate": 0.00014393382352941174, - "loss": 2.827, + "learning_rate": 1.4407169117647059e-05, + "loss": 0.192, "step": 1638 }, { "epoch": 1.84, - "learning_rate": 0.00014379595588235293, - "loss": 2.8149, + "learning_rate": 1.4393382352941178e-05, + "loss": 0.1767, "step": 1639 }, { "epoch": 1.84, - "learning_rate": 0.0001436580882352941, - "loss": 2.8174, + "learning_rate": 1.4379595588235295e-05, + "loss": 0.1571, "step": 1640 }, { "epoch": 1.84, - "learning_rate": 0.00014352022058823527, - "loss": 2.8299, + "learning_rate": 1.4365808823529413e-05, + "loss": 0.2193, "step": 1641 }, { "epoch": 1.84, - "learning_rate": 0.00014338235294117648, - "loss": 3.1108, + "learning_rate": 1.4352022058823529e-05, + "loss": 0.2366, "step": 1642 }, { "epoch": 1.84, - "learning_rate": 0.00014324448529411763, - "loss": 2.8076, + "learning_rate": 1.4338235294117647e-05, + "loss": 0.1787, "step": 1643 }, { "epoch": 1.84, - "learning_rate": 0.00014310661764705882, - "loss": 2.8104, + "learning_rate": 1.4324448529411765e-05, + "loss": 0.2117, "step": 1644 }, { "epoch": 1.84, - "learning_rate": 0.00014296874999999998, - "loss": 2.8093, + "learning_rate": 1.4310661764705883e-05, + "loss": 0.1948, "step": 1645 }, { "epoch": 1.85, - "learning_rate": 0.00014283088235294116, - "loss": 2.8408, + "learning_rate": 1.4296875e-05, + "loss": 0.1848, "step": 1646 }, { "epoch": 1.85, - "learning_rate": 0.00014269301470588234, - "loss": 2.8081, + "learning_rate": 1.4283088235294117e-05, + "loss": 0.2049, "step": 1647 }, { "epoch": 1.85, - "learning_rate": 0.0001425551470588235, - "loss": 2.9313, + "learning_rate": 1.4269301470588236e-05, + "loss": 0.2182, "step": 1648 }, { "epoch": 1.85, - "learning_rate": 0.00014241727941176468, - "loss": 3.695, + "learning_rate": 1.4255514705882353e-05, + "loss": 0.2299, "step": 1649 }, { "epoch": 1.85, - "learning_rate": 0.00014227941176470587, - "loss": 2.8163, + "learning_rate": 1.4241727941176472e-05, + "loss": 0.1735, "step": 1650 }, { "epoch": 1.85, - "learning_rate": 0.00014214154411764705, - "loss": 2.8123, + "learning_rate": 1.4227941176470589e-05, + "loss": 0.187, "step": 1651 }, { "epoch": 1.85, - "learning_rate": 0.0001420036764705882, - "loss": 2.8184, + "learning_rate": 1.4214154411764706e-05, + "loss": 0.207, "step": 1652 }, { "epoch": 1.85, - "learning_rate": 0.0001418658088235294, - "loss": 2.8148, + "learning_rate": 1.4200367647058823e-05, + "loss": 0.2156, "step": 1653 }, { "epoch": 1.85, - "learning_rate": 0.00014172794117647057, - "loss": 2.8723, + "learning_rate": 1.4186580882352941e-05, + "loss": 0.1839, "step": 1654 }, { "epoch": 1.86, - "learning_rate": 0.00014159007352941176, - "loss": 4.402, + "learning_rate": 1.4172794117647058e-05, + "loss": 0.2589, "step": 1655 }, { "epoch": 1.86, - "learning_rate": 0.00014145220588235291, - "loss": 2.8125, + "learning_rate": 1.4159007352941177e-05, + "loss": 0.1981, "step": 1656 }, { "epoch": 1.86, - "learning_rate": 0.0001413143382352941, - "loss": 2.8184, + "learning_rate": 1.4145220588235294e-05, + "loss": 0.1959, "step": 1657 }, { "epoch": 1.86, - "learning_rate": 0.00014117647058823528, - "loss": 2.8076, + "learning_rate": 1.4131433823529413e-05, + "loss": 0.1813, "step": 1658 }, { "epoch": 1.86, - "learning_rate": 0.00014103860294117647, - "loss": 2.8382, + "learning_rate": 1.411764705882353e-05, + "loss": 0.1827, "step": 1659 }, { "epoch": 1.86, - "learning_rate": 0.00014090073529411762, - "loss": 2.9122, + "learning_rate": 1.4103860294117649e-05, + "loss": 0.2048, "step": 1660 }, { "epoch": 1.86, - "learning_rate": 0.0001407628676470588, - "loss": 4.6155, + "learning_rate": 1.4090073529411766e-05, + "loss": 0.2248, "step": 1661 }, { "epoch": 1.86, - "learning_rate": 0.000140625, - "loss": 2.8456, + "learning_rate": 1.4076286764705884e-05, + "loss": 0.1869, "step": 1662 }, { "epoch": 1.86, - "learning_rate": 0.00014048713235294115, - "loss": 2.8016, + "learning_rate": 1.40625e-05, + "loss": 0.1784, "step": 1663 }, { "epoch": 1.87, - "learning_rate": 0.00014034926470588233, - "loss": 2.8145, + "learning_rate": 1.4048713235294117e-05, + "loss": 0.199, "step": 1664 }, { "epoch": 1.87, - "learning_rate": 0.0001402113970588235, - "loss": 2.806, + "learning_rate": 1.4034926470588235e-05, + "loss": 0.1747, "step": 1665 }, { "epoch": 1.87, - "learning_rate": 0.0001400735294117647, - "loss": 2.8466, + "learning_rate": 1.4021139705882352e-05, + "loss": 0.2022, "step": 1666 }, { "epoch": 1.87, - "learning_rate": 0.00013993566176470585, - "loss": 3.3109, + "learning_rate": 1.4007352941176471e-05, + "loss": 0.2249, "step": 1667 }, { "epoch": 1.87, - "learning_rate": 0.00013979779411764704, - "loss": 2.8147, + "learning_rate": 1.3993566176470588e-05, + "loss": 0.18, "step": 1668 }, { "epoch": 1.87, - "learning_rate": 0.00013965992647058822, - "loss": 2.8201, + "learning_rate": 1.3979779411764707e-05, + "loss": 0.2, "step": 1669 }, { "epoch": 1.87, - "learning_rate": 0.0001395220588235294, - "loss": 2.8091, + "learning_rate": 1.3965992647058824e-05, + "loss": 0.1583, "step": 1670 }, { "epoch": 1.87, - "learning_rate": 0.00013938419117647056, - "loss": 2.8127, + "learning_rate": 1.3952205882352942e-05, + "loss": 0.1766, "step": 1671 }, { "epoch": 1.87, - "learning_rate": 0.00013924632352941174, - "loss": 2.821, + "learning_rate": 1.393841911764706e-05, + "loss": 0.1819, "step": 1672 }, { "epoch": 1.88, - "learning_rate": 0.00013910845588235293, - "loss": 2.9206, + "learning_rate": 1.3924632352941177e-05, + "loss": 0.1941, "step": 1673 }, { "epoch": 1.88, - "learning_rate": 0.0001389705882352941, - "loss": 3.3867, + "learning_rate": 1.3910845588235294e-05, + "loss": 0.2275, "step": 1674 }, { "epoch": 1.88, - "learning_rate": 0.00013883272058823527, - "loss": 2.8198, + "learning_rate": 1.3897058823529412e-05, + "loss": 0.1757, "step": 1675 }, { "epoch": 1.88, - "learning_rate": 0.00013869485294117645, - "loss": 2.8109, + "learning_rate": 1.388327205882353e-05, + "loss": 0.1512, "step": 1676 }, { "epoch": 1.88, - "learning_rate": 0.00013855698529411764, - "loss": 2.8369, + "learning_rate": 1.3869485294117648e-05, + "loss": 0.1992, "step": 1677 }, { "epoch": 1.88, - "learning_rate": 0.00013841911764705882, - "loss": 2.8126, + "learning_rate": 1.3855698529411765e-05, + "loss": 0.1939, "step": 1678 }, { "epoch": 1.88, - "learning_rate": 0.00013828124999999998, - "loss": 2.8337, + "learning_rate": 1.3841911764705884e-05, + "loss": 0.2467, "step": 1679 }, { "epoch": 1.88, - "learning_rate": 0.00013814338235294116, - "loss": 3.4375, + "learning_rate": 1.3828125e-05, + "loss": 0.195, "step": 1680 }, { "epoch": 1.88, - "learning_rate": 0.00013800551470588234, - "loss": 2.8011, + "learning_rate": 1.3814338235294118e-05, + "loss": 0.1504, "step": 1681 }, { "epoch": 1.89, - "learning_rate": 0.0001378676470588235, - "loss": 2.8257, + "learning_rate": 1.3800551470588236e-05, + "loss": 0.1968, "step": 1682 }, { "epoch": 1.89, - "learning_rate": 0.00013772977941176468, - "loss": 2.8287, + "learning_rate": 1.3786764705882353e-05, + "loss": 0.1899, "step": 1683 }, { "epoch": 1.89, - "learning_rate": 0.00013759191176470587, - "loss": 2.827, + "learning_rate": 1.377297794117647e-05, + "loss": 0.2164, "step": 1684 }, { "epoch": 1.89, - "learning_rate": 0.00013745404411764705, - "loss": 2.8539, + "learning_rate": 1.3759191176470587e-05, + "loss": 0.1638, "step": 1685 }, { "epoch": 1.89, - "learning_rate": 0.0001373161764705882, - "loss": 3.8616, + "learning_rate": 1.3745404411764706e-05, + "loss": 0.2054, "step": 1686 }, { "epoch": 1.89, - "learning_rate": 0.0001371783088235294, - "loss": 2.8052, + "learning_rate": 1.3731617647058823e-05, + "loss": 0.2056, "step": 1687 }, { "epoch": 1.89, - "learning_rate": 0.00013704044117647058, - "loss": 2.7956, + "learning_rate": 1.3717830882352942e-05, + "loss": 0.1595, "step": 1688 }, { "epoch": 1.89, - "learning_rate": 0.00013690257352941176, - "loss": 2.8135, + "learning_rate": 1.3704044117647059e-05, + "loss": 0.1987, "step": 1689 }, { "epoch": 1.89, - "learning_rate": 0.00013676470588235292, - "loss": 2.7922, + "learning_rate": 1.3690257352941178e-05, + "loss": 0.1746, "step": 1690 }, { "epoch": 1.9, - "learning_rate": 0.0001366268382352941, - "loss": 2.8384, + "learning_rate": 1.3676470588235295e-05, + "loss": 0.2156, "step": 1691 }, { "epoch": 1.9, - "learning_rate": 0.00013648897058823528, - "loss": 3.3806, + "learning_rate": 1.3662683823529413e-05, + "loss": 0.292, "step": 1692 }, { "epoch": 1.9, - "learning_rate": 0.00013635110294117647, - "loss": 2.8199, + "learning_rate": 1.364889705882353e-05, + "loss": 0.2052, "step": 1693 }, { "epoch": 1.9, - "learning_rate": 0.00013621323529411762, - "loss": 2.8098, + "learning_rate": 1.3635110294117647e-05, + "loss": 0.1567, "step": 1694 }, { "epoch": 1.9, - "learning_rate": 0.0001360753676470588, - "loss": 2.8117, + "learning_rate": 1.3621323529411764e-05, + "loss": 0.1805, "step": 1695 }, { "epoch": 1.9, - "learning_rate": 0.0001359375, - "loss": 2.8321, + "learning_rate": 1.3607536764705883e-05, + "loss": 0.2192, "step": 1696 }, { "epoch": 1.9, - "learning_rate": 0.00013579963235294115, - "loss": 2.8236, + "learning_rate": 1.359375e-05, + "loss": 0.177, "step": 1697 }, { "epoch": 1.9, - "learning_rate": 0.00013566176470588233, - "loss": 3.0804, + "learning_rate": 1.3579963235294117e-05, + "loss": 0.1958, "step": 1698 }, { "epoch": 1.9, - "learning_rate": 0.00013552389705882351, - "loss": 4.1622, + "learning_rate": 1.3566176470588236e-05, + "loss": 0.2271, "step": 1699 }, { "epoch": 1.91, - "learning_rate": 0.0001353860294117647, - "loss": 2.8032, + "learning_rate": 1.3552389705882353e-05, + "loss": 0.1626, "step": 1700 }, { "epoch": 1.91, - "eval_loss": 4.032460689544678, - "eval_runtime": 48.1234, - "eval_samples_per_second": 56.168, - "eval_steps_per_second": 0.894, - "eval_wer": 1.0, + "eval_loss": 0.21954555809497833, + "eval_runtime": 47.8132, + "eval_samples_per_second": 56.533, + "eval_steps_per_second": 0.899, + "eval_wer": 0.19269512150288592, "step": 1700 }, { "epoch": 1.91, - "learning_rate": 0.00013524816176470585, - "loss": 2.8286, + "learning_rate": 1.3538602941176472e-05, + "loss": 0.1981, "step": 1701 }, { "epoch": 1.91, - "learning_rate": 0.00013511029411764704, - "loss": 2.805, + "learning_rate": 1.3524816176470589e-05, + "loss": 0.1884, "step": 1702 }, { "epoch": 1.91, - "learning_rate": 0.00013497242647058822, - "loss": 2.8168, + "learning_rate": 1.3511029411764707e-05, + "loss": 0.2392, "step": 1703 }, { "epoch": 1.91, - "learning_rate": 0.0001348345588235294, - "loss": 2.8851, + "learning_rate": 1.3497242647058823e-05, + "loss": 0.2114, "step": 1704 }, { "epoch": 1.91, - "learning_rate": 0.00013469669117647056, - "loss": 4.4468, + "learning_rate": 1.3483455882352941e-05, + "loss": 0.2, "step": 1705 }, { "epoch": 1.91, - "learning_rate": 0.00013455882352941175, - "loss": 2.8089, + "learning_rate": 1.3469669117647058e-05, + "loss": 0.1709, "step": 1706 }, { "epoch": 1.91, - "learning_rate": 0.00013442095588235293, - "loss": 2.8268, + "learning_rate": 1.3455882352941177e-05, + "loss": 0.1817, "step": 1707 }, { "epoch": 1.91, - "learning_rate": 0.0001342830882352941, - "loss": 2.8289, + "learning_rate": 1.3442095588235294e-05, + "loss": 0.2054, "step": 1708 }, { "epoch": 1.92, - "learning_rate": 0.00013414522058823527, - "loss": 2.8174, + "learning_rate": 1.3428308823529413e-05, + "loss": 0.1997, "step": 1709 }, { "epoch": 1.92, - "learning_rate": 0.00013400735294117645, - "loss": 2.8237, + "learning_rate": 1.341452205882353e-05, + "loss": 0.206, "step": 1710 }, { "epoch": 1.92, - "learning_rate": 0.00013386948529411764, - "loss": 4.1152, + "learning_rate": 1.3400735294117648e-05, + "loss": 0.3055, "step": 1711 }, { "epoch": 1.92, - "learning_rate": 0.00013373161764705882, - "loss": 2.805, + "learning_rate": 1.3386948529411765e-05, + "loss": 0.1756, "step": 1712 }, { "epoch": 1.92, - "learning_rate": 0.00013359374999999998, - "loss": 2.8043, + "learning_rate": 1.3373161764705884e-05, + "loss": 0.1793, "step": 1713 }, { "epoch": 1.92, - "learning_rate": 0.00013345588235294116, - "loss": 2.8091, + "learning_rate": 1.3359375000000001e-05, + "loss": 0.1691, "step": 1714 }, { "epoch": 1.92, - "learning_rate": 0.00013331801470588235, - "loss": 2.8151, + "learning_rate": 1.3345588235294116e-05, + "loss": 0.2141, "step": 1715 }, { "epoch": 1.92, - "learning_rate": 0.0001331801470588235, - "loss": 2.8391, + "learning_rate": 1.3331801470588235e-05, + "loss": 0.19, "step": 1716 }, { "epoch": 1.92, - "learning_rate": 0.00013304227941176469, - "loss": 3.1118, + "learning_rate": 1.3318014705882352e-05, + "loss": 0.2208, "step": 1717 }, { "epoch": 1.93, - "learning_rate": 0.00013290441176470587, - "loss": 2.8094, + "learning_rate": 1.3304227941176471e-05, + "loss": 0.2224, "step": 1718 }, { "epoch": 1.93, - "learning_rate": 0.00013276654411764705, - "loss": 2.8072, + "learning_rate": 1.3290441176470588e-05, + "loss": 0.2015, "step": 1719 }, { "epoch": 1.93, - "learning_rate": 0.0001326286764705882, - "loss": 2.8103, + "learning_rate": 1.3276654411764707e-05, + "loss": 0.1799, "step": 1720 }, { "epoch": 1.93, - "learning_rate": 0.0001324908088235294, - "loss": 2.8234, + "learning_rate": 1.3262867647058824e-05, + "loss": 0.2031, "step": 1721 }, { "epoch": 1.93, - "learning_rate": 0.00013235294117647058, - "loss": 2.8201, + "learning_rate": 1.3249080882352942e-05, + "loss": 0.1524, "step": 1722 }, { "epoch": 1.93, - "learning_rate": 0.00013221507352941176, - "loss": 3.1533, + "learning_rate": 1.323529411764706e-05, + "loss": 0.2048, "step": 1723 }, { "epoch": 1.93, - "learning_rate": 0.00013207720588235292, - "loss": 3.7444, + "learning_rate": 1.3221507352941178e-05, + "loss": 0.2118, "step": 1724 }, { "epoch": 1.93, - "learning_rate": 0.0001319393382352941, - "loss": 2.819, + "learning_rate": 1.3207720588235293e-05, + "loss": 0.1729, "step": 1725 }, { "epoch": 1.93, - "learning_rate": 0.00013180147058823528, - "loss": 2.8264, + "learning_rate": 1.3193933823529412e-05, + "loss": 0.1811, "step": 1726 }, { "epoch": 1.94, - "learning_rate": 0.00013166360294117647, - "loss": 2.8212, + "learning_rate": 1.3180147058823529e-05, + "loss": 0.1582, "step": 1727 }, { "epoch": 1.94, - "learning_rate": 0.00013152573529411762, - "loss": 2.8246, + "learning_rate": 1.3166360294117648e-05, + "loss": 0.2016, "step": 1728 }, { "epoch": 1.94, - "learning_rate": 0.0001313878676470588, - "loss": 2.8453, + "learning_rate": 1.3152573529411765e-05, + "loss": 0.1517, "step": 1729 }, { "epoch": 1.94, - "learning_rate": 0.00013125, - "loss": 4.4718, + "learning_rate": 1.3138786764705884e-05, + "loss": 0.242, "step": 1730 }, { "epoch": 1.94, - "learning_rate": 0.00013111213235294115, - "loss": 2.8367, + "learning_rate": 1.3125e-05, + "loss": 0.2005, "step": 1731 }, { "epoch": 1.94, - "learning_rate": 0.00013097426470588233, - "loss": 2.8376, + "learning_rate": 1.3111213235294118e-05, + "loss": 0.181, "step": 1732 }, { "epoch": 1.94, - "learning_rate": 0.00013083639705882352, - "loss": 2.8177, + "learning_rate": 1.3097426470588236e-05, + "loss": 0.1818, "step": 1733 }, { "epoch": 1.94, - "learning_rate": 0.0001306985294117647, - "loss": 2.816, + "learning_rate": 1.3083639705882353e-05, + "loss": 0.1949, "step": 1734 }, { "epoch": 1.95, - "learning_rate": 0.00013056066176470586, - "loss": 2.8865, + "learning_rate": 1.306985294117647e-05, + "loss": 0.1699, "step": 1735 }, { "epoch": 1.95, - "learning_rate": 0.00013042279411764704, - "loss": 3.8728, + "learning_rate": 1.3056066176470587e-05, + "loss": 0.2379, "step": 1736 }, { "epoch": 1.95, - "learning_rate": 0.00013028492647058822, - "loss": 2.8269, + "learning_rate": 1.3042279411764706e-05, + "loss": 0.1923, "step": 1737 }, { "epoch": 1.95, - "learning_rate": 0.0001301470588235294, - "loss": 2.8266, + "learning_rate": 1.3028492647058823e-05, + "loss": 0.2043, "step": 1738 }, { "epoch": 1.95, - "learning_rate": 0.00013000919117647056, - "loss": 2.817, + "learning_rate": 1.3014705882352942e-05, + "loss": 0.1758, "step": 1739 }, { "epoch": 1.95, - "learning_rate": 0.00012987132352941175, - "loss": 2.8145, + "learning_rate": 1.3000919117647059e-05, + "loss": 0.1948, "step": 1740 }, { "epoch": 1.95, - "learning_rate": 0.00012973345588235293, - "loss": 2.8394, + "learning_rate": 1.2987132352941177e-05, + "loss": 0.215, "step": 1741 }, { "epoch": 1.95, - "learning_rate": 0.00012959558823529411, - "loss": 3.2991, + "learning_rate": 1.2973345588235294e-05, + "loss": 0.2299, "step": 1742 }, { "epoch": 1.95, - "learning_rate": 0.00012945772058823527, - "loss": 2.8087, + "learning_rate": 1.2959558823529413e-05, + "loss": 0.1766, "step": 1743 }, { "epoch": 1.96, - "learning_rate": 0.00012931985294117646, - "loss": 2.8011, + "learning_rate": 1.294577205882353e-05, + "loss": 0.1868, "step": 1744 }, { "epoch": 1.96, - "learning_rate": 0.00012918198529411764, - "loss": 2.8045, + "learning_rate": 1.2931985294117649e-05, + "loss": 0.1882, "step": 1745 }, { "epoch": 1.96, - "learning_rate": 0.00012904411764705882, - "loss": 2.8058, + "learning_rate": 1.2918198529411764e-05, + "loss": 0.1721, "step": 1746 }, { "epoch": 1.96, - "learning_rate": 0.00012890624999999998, - "loss": 2.8166, + "learning_rate": 1.2904411764705883e-05, + "loss": 0.2085, "step": 1747 }, { "epoch": 1.96, - "learning_rate": 0.00012876838235294116, - "loss": 3.0537, + "learning_rate": 1.2890625e-05, + "loss": 0.2075, "step": 1748 }, { "epoch": 1.96, - "learning_rate": 0.00012863051470588235, - "loss": 4.5934, + "learning_rate": 1.2876838235294117e-05, + "loss": 0.2059, "step": 1749 }, { "epoch": 1.96, - "learning_rate": 0.0001284926470588235, - "loss": 2.8109, + "learning_rate": 1.2863051470588236e-05, + "loss": 0.1704, "step": 1750 }, { "epoch": 1.96, - "learning_rate": 0.0001283547794117647, - "loss": 2.8338, + "learning_rate": 1.2849264705882353e-05, + "loss": 0.2111, "step": 1751 }, { "epoch": 1.96, - "learning_rate": 0.00012821691176470587, - "loss": 2.8103, + "learning_rate": 1.2835477941176471e-05, + "loss": 0.1762, "step": 1752 }, { "epoch": 1.97, - "learning_rate": 0.00012807904411764705, - "loss": 2.7951, + "learning_rate": 1.2821691176470588e-05, + "loss": 0.2069, "step": 1753 }, { "epoch": 1.97, - "learning_rate": 0.0001279411764705882, - "loss": 3.0195, + "learning_rate": 1.2807904411764707e-05, + "loss": 0.2178, "step": 1754 }, { "epoch": 1.97, - "learning_rate": 0.0001278033088235294, - "loss": 4.3327, + "learning_rate": 1.2794117647058824e-05, + "loss": 0.2311, "step": 1755 }, { "epoch": 1.97, - "learning_rate": 0.00012766544117647058, - "loss": 2.81, + "learning_rate": 1.2780330882352941e-05, + "loss": 0.1918, "step": 1756 }, { "epoch": 1.97, - "learning_rate": 0.00012752757352941176, - "loss": 2.82, + "learning_rate": 1.2766544117647058e-05, + "loss": 0.1806, "step": 1757 }, { "epoch": 1.97, - "learning_rate": 0.00012738970588235292, - "loss": 2.8232, + "learning_rate": 1.2752757352941177e-05, + "loss": 0.1901, "step": 1758 }, { "epoch": 1.97, - "learning_rate": 0.0001272518382352941, - "loss": 2.8348, + "learning_rate": 1.2738970588235294e-05, + "loss": 0.1991, "step": 1759 }, { "epoch": 1.97, - "learning_rate": 0.00012711397058823529, - "loss": 2.9158, + "learning_rate": 1.2725183823529413e-05, + "loss": 0.1761, "step": 1760 }, { "epoch": 1.97, - "learning_rate": 0.00012697610294117647, - "loss": 4.8274, + "learning_rate": 1.271139705882353e-05, + "loss": 0.2817, "step": 1761 }, { "epoch": 1.98, - "learning_rate": 0.00012683823529411763, - "loss": 2.8135, + "learning_rate": 1.2697610294117648e-05, + "loss": 0.1916, "step": 1762 }, { "epoch": 1.98, - "learning_rate": 0.0001267003676470588, - "loss": 2.8261, + "learning_rate": 1.2683823529411765e-05, + "loss": 0.191, "step": 1763 }, { "epoch": 1.98, - "learning_rate": 0.0001265625, - "loss": 2.8151, + "learning_rate": 1.2670036764705884e-05, + "loss": 0.152, "step": 1764 }, { "epoch": 1.98, - "learning_rate": 0.00012642463235294115, - "loss": 2.8221, + "learning_rate": 1.2656250000000001e-05, + "loss": 0.1871, "step": 1765 }, { "epoch": 1.98, - "learning_rate": 0.00012628676470588233, - "loss": 2.8149, + "learning_rate": 1.2642463235294118e-05, + "loss": 0.1966, "step": 1766 }, { "epoch": 1.98, - "learning_rate": 0.00012614889705882352, - "loss": 3.1905, + "learning_rate": 1.2628676470588235e-05, + "loss": 0.2194, "step": 1767 }, { "epoch": 1.98, - "learning_rate": 0.0001260110294117647, - "loss": 2.8074, + "learning_rate": 1.2614889705882352e-05, + "loss": 0.1575, "step": 1768 }, { "epoch": 1.98, - "learning_rate": 0.00012587316176470586, - "loss": 2.8114, + "learning_rate": 1.260110294117647e-05, + "loss": 0.1845, "step": 1769 }, { "epoch": 1.98, - "learning_rate": 0.00012573529411764704, - "loss": 2.8046, + "learning_rate": 1.2587316176470588e-05, + "loss": 0.1815, "step": 1770 }, { "epoch": 1.99, - "learning_rate": 0.00012559742647058822, - "loss": 2.8082, + "learning_rate": 1.2573529411764706e-05, + "loss": 0.2047, "step": 1771 }, { "epoch": 1.99, - "learning_rate": 0.0001254595588235294, - "loss": 2.8186, + "learning_rate": 1.2559742647058823e-05, + "loss": 0.2061, "step": 1772 }, { "epoch": 1.99, - "learning_rate": 0.00012532169117647056, - "loss": 3.0392, + "learning_rate": 1.2545955882352942e-05, + "loss": 0.2309, "step": 1773 }, { "epoch": 1.99, - "learning_rate": 0.00012518382352941175, - "loss": 3.6054, + "learning_rate": 1.253216911764706e-05, + "loss": 0.2433, "step": 1774 }, { "epoch": 1.99, - "learning_rate": 0.00012504595588235293, - "loss": 2.8154, + "learning_rate": 1.2518382352941178e-05, + "loss": 0.1705, "step": 1775 }, { "epoch": 1.99, - "learning_rate": 0.00012490808823529412, - "loss": 2.8373, + "learning_rate": 1.2504595588235295e-05, + "loss": 0.172, "step": 1776 }, { "epoch": 1.99, - "learning_rate": 0.00012477022058823527, - "loss": 2.8082, + "learning_rate": 1.2490808823529412e-05, + "loss": 0.1638, "step": 1777 }, { "epoch": 1.99, - "learning_rate": 0.00012463235294117646, - "loss": 2.8197, + "learning_rate": 1.2477022058823529e-05, + "loss": 0.2014, "step": 1778 }, { "epoch": 1.99, - "learning_rate": 0.00012449448529411764, - "loss": 2.9493, + "learning_rate": 1.2463235294117648e-05, + "loss": 0.2161, "step": 1779 }, { "epoch": 2.0, - "learning_rate": 0.00012435661764705882, - "loss": 3.8987, + "learning_rate": 1.2449448529411765e-05, + "loss": 0.2463, "step": 1780 }, { "epoch": 2.0, - "learning_rate": 0.00012421874999999998, - "loss": 2.8314, + "learning_rate": 1.2435661764705883e-05, + "loss": 0.1836, "step": 1781 }, { "epoch": 2.0, - "learning_rate": 0.00012408088235294116, - "loss": 2.8604, + "learning_rate": 1.2421875e-05, + "loss": 0.1898, "step": 1782 }, { "epoch": 2.0, - "learning_rate": 0.00012394301470588235, - "loss": 2.8492, + "learning_rate": 1.2408088235294117e-05, + "loss": 0.1973, "step": 1783 }, { "epoch": 2.0, - "learning_rate": 0.0001238051470588235, - "loss": 3.294, + "learning_rate": 1.2394301470588236e-05, + "loss": 0.2502, "step": 1784 }, { "epoch": 2.0, - "learning_rate": 0.0001236672794117647, - "loss": 2.8441, + "learning_rate": 1.2380514705882353e-05, + "loss": 0.1821, "step": 1785 }, { "epoch": 2.0, - "learning_rate": 0.00012352941176470587, - "loss": 2.8639, + "learning_rate": 1.2366727941176472e-05, + "loss": 0.1845, "step": 1786 }, { "epoch": 2.0, - "learning_rate": 0.00012339154411764706, - "loss": 2.8631, + "learning_rate": 1.2352941176470587e-05, + "loss": 0.1617, "step": 1787 }, { "epoch": 2.0, - "learning_rate": 0.0001232536764705882, - "loss": 2.8272, + "learning_rate": 1.2339154411764706e-05, + "loss": 0.1465, "step": 1788 }, { "epoch": 2.01, - "learning_rate": 0.0001231158088235294, - "loss": 2.837, + "learning_rate": 1.2325367647058823e-05, + "loss": 0.1786, "step": 1789 }, { "epoch": 2.01, - "learning_rate": 0.00012297794117647058, - "loss": 2.9223, + "learning_rate": 1.2311580882352942e-05, + "loss": 0.1935, "step": 1790 }, { "epoch": 2.01, - "learning_rate": 0.00012284007352941176, - "loss": 3.4375, + "learning_rate": 1.2297794117647059e-05, + "loss": 0.1821, "step": 1791 }, { "epoch": 2.01, - "learning_rate": 0.00012270220588235292, - "loss": 2.8294, + "learning_rate": 1.2284007352941177e-05, + "loss": 0.1649, "step": 1792 }, { "epoch": 2.01, - "learning_rate": 0.0001225643382352941, - "loss": 2.811, + "learning_rate": 1.2270220588235294e-05, + "loss": 0.1739, "step": 1793 }, { "epoch": 2.01, - "learning_rate": 0.0001224264705882353, - "loss": 2.8094, + "learning_rate": 1.2256433823529413e-05, + "loss": 0.1524, "step": 1794 }, { "epoch": 2.01, - "learning_rate": 0.00012228860294117647, - "loss": 2.8142, + "learning_rate": 1.224264705882353e-05, + "loss": 0.1693, "step": 1795 }, { "epoch": 2.01, - "learning_rate": 0.00012215073529411763, - "loss": 2.8812, + "learning_rate": 1.2228860294117649e-05, + "loss": 0.1621, "step": 1796 }, { "epoch": 2.01, - "learning_rate": 0.00012201286764705882, - "loss": 4.762, + "learning_rate": 1.2215073529411766e-05, + "loss": 0.2264, "step": 1797 }, { "epoch": 2.02, - "learning_rate": 0.000121875, - "loss": 2.8233, + "learning_rate": 1.2201286764705883e-05, + "loss": 0.1718, "step": 1798 }, { "epoch": 2.02, - "learning_rate": 0.00012173713235294116, - "loss": 2.8175, + "learning_rate": 1.21875e-05, + "loss": 0.1698, "step": 1799 }, { "epoch": 2.02, - "learning_rate": 0.00012159926470588235, - "loss": 2.8053, + "learning_rate": 1.2173713235294117e-05, + "loss": 0.168, "step": 1800 }, { "epoch": 2.02, - "eval_loss": 4.012497425079346, - "eval_runtime": 48.1133, - "eval_samples_per_second": 56.18, - "eval_steps_per_second": 0.894, - "eval_wer": 1.0, + "eval_loss": 0.2136542946100235, + "eval_runtime": 47.324, + "eval_samples_per_second": 57.117, + "eval_steps_per_second": 0.909, + "eval_wer": 0.1852873056137642, "step": 1800 }, { "epoch": 2.02, - "learning_rate": 0.00012146139705882352, - "loss": 2.8164, + "learning_rate": 1.2159926470588235e-05, + "loss": 0.1876, "step": 1801 }, { "epoch": 2.02, - "learning_rate": 0.0001213235294117647, - "loss": 2.8405, + "learning_rate": 1.2146139705882353e-05, + "loss": 0.205, "step": 1802 }, { "epoch": 2.02, - "learning_rate": 0.00012118566176470587, - "loss": 4.719, + "learning_rate": 1.2132352941176471e-05, + "loss": 0.1834, "step": 1803 }, { "epoch": 2.02, - "learning_rate": 0.00012104779411764706, - "loss": 2.8096, + "learning_rate": 1.2118566176470588e-05, + "loss": 0.1862, "step": 1804 }, { "epoch": 2.02, - "learning_rate": 0.00012090992647058823, - "loss": 2.8062, + "learning_rate": 1.2104779411764707e-05, + "loss": 0.1404, "step": 1805 }, { "epoch": 2.02, - "learning_rate": 0.00012077205882352941, - "loss": 2.8185, + "learning_rate": 1.2090992647058824e-05, + "loss": 0.1496, "step": 1806 }, { "epoch": 2.03, - "learning_rate": 0.00012063419117647058, - "loss": 2.8299, + "learning_rate": 1.2077205882352943e-05, + "loss": 0.1571, "step": 1807 }, { "epoch": 2.03, - "learning_rate": 0.00012049632352941176, - "loss": 2.8388, + "learning_rate": 1.2063419117647058e-05, + "loss": 0.2073, "step": 1808 }, { "epoch": 2.03, - "learning_rate": 0.00012035845588235293, - "loss": 3.0989, + "learning_rate": 1.2049632352941177e-05, + "loss": 0.2123, "step": 1809 }, { "epoch": 2.03, - "learning_rate": 0.00012022058823529412, - "loss": 2.7982, + "learning_rate": 1.2035845588235294e-05, + "loss": 0.1565, "step": 1810 }, { "epoch": 2.03, - "learning_rate": 0.00012008272058823529, - "loss": 2.8148, + "learning_rate": 1.2022058823529412e-05, + "loss": 0.1585, "step": 1811 }, { "epoch": 2.03, - "learning_rate": 0.00011994485294117647, - "loss": 2.8127, + "learning_rate": 1.200827205882353e-05, + "loss": 0.1297, "step": 1812 }, { "epoch": 2.03, - "learning_rate": 0.00011980698529411764, - "loss": 2.8224, + "learning_rate": 1.1994485294117648e-05, + "loss": 0.1353, "step": 1813 }, { "epoch": 2.03, - "learning_rate": 0.00011966911764705882, - "loss": 2.8119, + "learning_rate": 1.1980698529411765e-05, + "loss": 0.1638, "step": 1814 }, { "epoch": 2.03, - "learning_rate": 0.00011953125, - "loss": 2.9279, + "learning_rate": 1.1966911764705884e-05, + "loss": 0.1995, "step": 1815 }, { "epoch": 2.04, - "learning_rate": 0.00011939338235294115, - "loss": 3.4056, + "learning_rate": 1.1953125000000001e-05, + "loss": 0.2075, "step": 1816 }, { "epoch": 2.04, - "learning_rate": 0.00011925551470588235, - "loss": 2.8073, + "learning_rate": 1.1939338235294118e-05, + "loss": 0.1607, "step": 1817 }, { "epoch": 2.04, - "learning_rate": 0.0001191176470588235, - "loss": 2.8127, + "learning_rate": 1.1925551470588235e-05, + "loss": 0.1768, "step": 1818 }, { "epoch": 2.04, - "learning_rate": 0.0001189797794117647, - "loss": 2.8067, + "learning_rate": 1.1911764705882352e-05, + "loss": 0.1281, "step": 1819 }, { "epoch": 2.04, - "learning_rate": 0.00011884191176470587, - "loss": 2.8231, + "learning_rate": 1.189797794117647e-05, + "loss": 0.1683, "step": 1820 }, { "epoch": 2.04, - "learning_rate": 0.00011870404411764706, - "loss": 2.9821, + "learning_rate": 1.1884191176470588e-05, + "loss": 0.1602, "step": 1821 }, { "epoch": 2.04, - "learning_rate": 0.00011856617647058823, - "loss": 4.1483, + "learning_rate": 1.1870404411764706e-05, + "loss": 0.1388, "step": 1822 }, { "epoch": 2.04, - "learning_rate": 0.00011842830882352941, - "loss": 2.8213, + "learning_rate": 1.1856617647058823e-05, + "loss": 0.1518, "step": 1823 }, { "epoch": 2.04, - "learning_rate": 0.00011829044117647058, - "loss": 2.8059, + "learning_rate": 1.1842830882352942e-05, + "loss": 0.1831, "step": 1824 }, { "epoch": 2.05, - "learning_rate": 0.00011815257352941176, - "loss": 2.8159, + "learning_rate": 1.1829044117647059e-05, + "loss": 0.1417, "step": 1825 }, { "epoch": 2.05, - "learning_rate": 0.00011801470588235293, - "loss": 2.8163, + "learning_rate": 1.1815257352941178e-05, + "loss": 0.16, "step": 1826 }, { "epoch": 2.05, - "learning_rate": 0.00011787683823529412, - "loss": 2.8483, + "learning_rate": 1.1801470588235295e-05, + "loss": 0.1709, "step": 1827 }, { "epoch": 2.05, - "learning_rate": 0.00011773897058823529, - "loss": 3.9416, + "learning_rate": 1.1787683823529413e-05, + "loss": 0.204, "step": 1828 }, { "epoch": 2.05, - "learning_rate": 0.00011760110294117647, - "loss": 2.819, + "learning_rate": 1.1773897058823529e-05, + "loss": 0.1462, "step": 1829 }, { "epoch": 2.05, - "learning_rate": 0.00011746323529411764, - "loss": 2.7943, + "learning_rate": 1.1760110294117648e-05, + "loss": 0.1321, "step": 1830 }, { "epoch": 2.05, - "learning_rate": 0.00011732536764705883, - "loss": 2.8031, + "learning_rate": 1.1746323529411765e-05, + "loss": 0.155, "step": 1831 }, { "epoch": 2.05, - "learning_rate": 0.0001171875, - "loss": 2.8145, + "learning_rate": 1.1732536764705883e-05, + "loss": 0.1755, "step": 1832 }, { "epoch": 2.05, - "learning_rate": 0.00011704963235294115, - "loss": 2.8309, + "learning_rate": 1.171875e-05, + "loss": 0.1719, "step": 1833 }, { "epoch": 2.06, - "learning_rate": 0.00011691176470588235, - "loss": 3.0531, + "learning_rate": 1.1704963235294117e-05, + "loss": 0.2112, "step": 1834 }, { "epoch": 2.06, - "learning_rate": 0.0001167738970588235, - "loss": 2.8208, + "learning_rate": 1.1691176470588236e-05, + "loss": 0.1794, "step": 1835 }, { "epoch": 2.06, - "learning_rate": 0.0001166360294117647, - "loss": 2.8238, + "learning_rate": 1.1677389705882353e-05, + "loss": 0.171, "step": 1836 }, { "epoch": 2.06, - "learning_rate": 0.00011649816176470586, - "loss": 2.8023, + "learning_rate": 1.1663602941176472e-05, + "loss": 0.1562, "step": 1837 }, { "epoch": 2.06, - "learning_rate": 0.00011636029411764706, - "loss": 2.8013, + "learning_rate": 1.1649816176470589e-05, + "loss": 0.1916, "step": 1838 }, { "epoch": 2.06, - "learning_rate": 0.00011622242647058821, - "loss": 2.8089, + "learning_rate": 1.1636029411764706e-05, + "loss": 0.164, "step": 1839 }, { "epoch": 2.06, - "learning_rate": 0.00011608455882352941, - "loss": 2.9001, + "learning_rate": 1.1622242647058823e-05, + "loss": 0.2047, "step": 1840 }, { "epoch": 2.06, - "learning_rate": 0.00011594669117647057, - "loss": 3.4248, + "learning_rate": 1.1608455882352941e-05, + "loss": 0.1641, "step": 1841 }, { "epoch": 2.07, - "learning_rate": 0.00011580882352941176, - "loss": 2.8123, + "learning_rate": 1.1594669117647058e-05, + "loss": 0.1778, "step": 1842 }, { "epoch": 2.07, - "learning_rate": 0.00011567095588235293, - "loss": 2.7929, + "learning_rate": 1.1580882352941177e-05, + "loss": 0.1501, "step": 1843 }, { "epoch": 2.07, - "learning_rate": 0.00011553308823529412, - "loss": 2.8117, + "learning_rate": 1.1567095588235294e-05, + "loss": 0.1787, "step": 1844 }, { "epoch": 2.07, - "learning_rate": 0.00011539522058823529, - "loss": 2.8143, + "learning_rate": 1.1553308823529413e-05, + "loss": 0.1647, "step": 1845 }, { "epoch": 2.07, - "learning_rate": 0.00011525735294117647, - "loss": 2.8393, + "learning_rate": 1.153952205882353e-05, + "loss": 0.1687, "step": 1846 }, { "epoch": 2.07, - "learning_rate": 0.00011511948529411764, - "loss": 4.1701, + "learning_rate": 1.1525735294117649e-05, + "loss": 0.1781, "step": 1847 }, { "epoch": 2.07, - "learning_rate": 0.00011498161764705883, - "loss": 2.7995, + "learning_rate": 1.1511948529411766e-05, + "loss": 0.146, "step": 1848 }, { "epoch": 2.07, - "learning_rate": 0.00011484375, - "loss": 2.8064, + "learning_rate": 1.1498161764705883e-05, + "loss": 0.1633, "step": 1849 }, { "epoch": 2.07, - "learning_rate": 0.00011470588235294115, - "loss": 2.823, + "learning_rate": 1.1484375e-05, + "loss": 0.1647, "step": 1850 }, { "epoch": 2.08, - "learning_rate": 0.00011456801470588235, - "loss": 2.8186, + "learning_rate": 1.1470588235294117e-05, + "loss": 0.1549, "step": 1851 }, { "epoch": 2.08, - "learning_rate": 0.0001144301470588235, - "loss": 2.8323, + "learning_rate": 1.1456801470588235e-05, + "loss": 0.1519, "step": 1852 }, { "epoch": 2.08, - "learning_rate": 0.0001142922794117647, - "loss": 3.7986, + "learning_rate": 1.1443014705882352e-05, + "loss": 0.1933, "step": 1853 }, { "epoch": 2.08, - "learning_rate": 0.00011415441176470586, - "loss": 2.8155, + "learning_rate": 1.1429227941176471e-05, + "loss": 0.1608, "step": 1854 }, { "epoch": 2.08, - "learning_rate": 0.00011401654411764706, - "loss": 2.806, + "learning_rate": 1.1415441176470588e-05, + "loss": 0.1791, "step": 1855 }, { "epoch": 2.08, - "learning_rate": 0.00011387867647058821, - "loss": 2.8107, + "learning_rate": 1.1401654411764707e-05, + "loss": 0.1491, "step": 1856 }, { "epoch": 2.08, - "learning_rate": 0.00011374080882352941, - "loss": 2.8185, + "learning_rate": 1.1387867647058824e-05, + "loss": 0.1685, "step": 1857 }, { "epoch": 2.08, - "learning_rate": 0.00011360294117647057, - "loss": 2.8464, + "learning_rate": 1.1374080882352943e-05, + "loss": 0.1767, "step": 1858 }, { "epoch": 2.08, - "learning_rate": 0.00011346507352941177, - "loss": 3.2311, + "learning_rate": 1.136029411764706e-05, + "loss": 0.1853, "step": 1859 }, { "epoch": 2.09, - "learning_rate": 0.00011332720588235292, - "loss": 2.7925, + "learning_rate": 1.1346507352941177e-05, + "loss": 0.1554, "step": 1860 }, { "epoch": 2.09, - "learning_rate": 0.00011318933823529412, - "loss": 2.8132, + "learning_rate": 1.1332720588235294e-05, + "loss": 0.162, "step": 1861 }, { "epoch": 2.09, - "learning_rate": 0.00011305147058823528, - "loss": 2.8037, + "learning_rate": 1.1318933823529412e-05, + "loss": 0.1409, "step": 1862 }, { "epoch": 2.09, - "learning_rate": 0.00011291360294117647, - "loss": 2.7933, + "learning_rate": 1.130514705882353e-05, + "loss": 0.1443, "step": 1863 }, { "epoch": 2.09, - "learning_rate": 0.00011277573529411763, - "loss": 2.7968, + "learning_rate": 1.1291360294117648e-05, + "loss": 0.129, "step": 1864 }, { "epoch": 2.09, - "learning_rate": 0.00011263786764705883, - "loss": 3.0818, + "learning_rate": 1.1277573529411765e-05, + "loss": 0.1494, "step": 1865 }, { "epoch": 2.09, - "learning_rate": 0.0001125, - "loss": 3.6346, + "learning_rate": 1.1263786764705884e-05, + "loss": 0.1562, "step": 1866 }, { "epoch": 2.09, - "learning_rate": 0.00011236213235294115, - "loss": 2.7908, + "learning_rate": 1.125e-05, + "loss": 0.1509, "step": 1867 }, { "epoch": 2.09, - "learning_rate": 0.00011222426470588235, - "loss": 2.807, + "learning_rate": 1.1236213235294118e-05, + "loss": 0.1393, "step": 1868 }, { "epoch": 2.1, - "learning_rate": 0.00011208639705882351, - "loss": 2.8126, + "learning_rate": 1.1222426470588236e-05, + "loss": 0.1779, "step": 1869 }, { "epoch": 2.1, - "learning_rate": 0.0001119485294117647, - "loss": 2.8175, + "learning_rate": 1.1208639705882352e-05, + "loss": 0.1646, "step": 1870 }, { "epoch": 2.1, - "learning_rate": 0.00011181066176470586, - "loss": 2.8753, + "learning_rate": 1.119485294117647e-05, + "loss": 0.1663, "step": 1871 }, { "epoch": 2.1, - "learning_rate": 0.00011167279411764706, - "loss": 3.8443, + "learning_rate": 1.1181066176470587e-05, + "loss": 0.1585, "step": 1872 }, { "epoch": 2.1, - "learning_rate": 0.00011153492647058821, - "loss": 2.806, + "learning_rate": 1.1167279411764706e-05, + "loss": 0.1474, "step": 1873 }, { "epoch": 2.1, - "learning_rate": 0.00011139705882352941, - "loss": 2.8181, + "learning_rate": 1.1153492647058823e-05, + "loss": 0.1316, "step": 1874 }, { "epoch": 2.1, - "learning_rate": 0.00011125919117647057, - "loss": 2.7996, + "learning_rate": 1.1139705882352942e-05, + "loss": 0.1786, "step": 1875 }, { "epoch": 2.1, - "learning_rate": 0.00011112132352941177, - "loss": 2.8288, + "learning_rate": 1.1125919117647059e-05, + "loss": 0.1848, "step": 1876 }, { "epoch": 2.1, - "learning_rate": 0.00011098345588235292, - "loss": 2.8438, + "learning_rate": 1.1112132352941178e-05, + "loss": 0.1831, "step": 1877 }, { "epoch": 2.11, - "learning_rate": 0.00011084558823529412, - "loss": 4.3241, + "learning_rate": 1.1098345588235295e-05, + "loss": 0.1817, "step": 1878 }, { "epoch": 2.11, - "learning_rate": 0.00011070772058823528, - "loss": 2.8173, + "learning_rate": 1.1084558823529413e-05, + "loss": 0.168, "step": 1879 }, { "epoch": 2.11, - "learning_rate": 0.00011056985294117647, - "loss": 2.7994, + "learning_rate": 1.1070772058823529e-05, + "loss": 0.1796, "step": 1880 }, { "epoch": 2.11, - "learning_rate": 0.00011043198529411763, - "loss": 2.8104, + "learning_rate": 1.1056985294117647e-05, + "loss": 0.1889, "step": 1881 }, { "epoch": 2.11, - "learning_rate": 0.00011029411764705883, - "loss": 2.8163, + "learning_rate": 1.1043198529411764e-05, + "loss": 0.1707, "step": 1882 }, { "epoch": 2.11, - "learning_rate": 0.00011015624999999998, - "loss": 2.8377, + "learning_rate": 1.1029411764705883e-05, + "loss": 0.1545, "step": 1883 }, { "epoch": 2.11, - "learning_rate": 0.00011001838235294115, - "loss": 3.366, + "learning_rate": 1.1015625e-05, + "loss": 0.3129, "step": 1884 }, { "epoch": 2.11, - "learning_rate": 0.00010988051470588234, - "loss": 2.8001, + "learning_rate": 1.1001838235294117e-05, + "loss": 0.1574, "step": 1885 }, { "epoch": 2.11, - "learning_rate": 0.00010974264705882351, - "loss": 2.8119, + "learning_rate": 1.0988051470588236e-05, + "loss": 0.1707, "step": 1886 }, { "epoch": 2.12, - "learning_rate": 0.00010960477941176469, - "loss": 2.8214, + "learning_rate": 1.0974264705882353e-05, + "loss": 0.1542, "step": 1887 }, { "epoch": 2.12, - "learning_rate": 0.00010946691176470586, - "loss": 2.8244, + "learning_rate": 1.0960477941176472e-05, + "loss": 0.1663, "step": 1888 }, { "epoch": 2.12, - "learning_rate": 0.00010932904411764705, - "loss": 2.8246, + "learning_rate": 1.0946691176470589e-05, + "loss": 0.1574, "step": 1889 }, { "epoch": 2.12, - "learning_rate": 0.00010919117647058822, - "loss": 2.9763, + "learning_rate": 1.0932904411764707e-05, + "loss": 0.1947, "step": 1890 }, { "epoch": 2.12, - "learning_rate": 0.00010905330882352941, - "loss": 3.8736, + "learning_rate": 1.0919117647058823e-05, + "loss": 0.2198, "step": 1891 }, { "epoch": 2.12, - "learning_rate": 0.00010891544117647057, - "loss": 2.8083, + "learning_rate": 1.0905330882352941e-05, + "loss": 0.1622, "step": 1892 }, { "epoch": 2.12, - "learning_rate": 0.00010877757352941177, - "loss": 2.8141, + "learning_rate": 1.0891544117647058e-05, + "loss": 0.1401, "step": 1893 }, { "epoch": 2.12, - "learning_rate": 0.00010863970588235292, - "loss": 2.8208, + "learning_rate": 1.0877757352941177e-05, + "loss": 0.1814, "step": 1894 }, { "epoch": 2.12, - "learning_rate": 0.00010850183823529412, - "loss": 2.8282, + "learning_rate": 1.0863970588235294e-05, + "loss": 0.1485, "step": 1895 }, { "epoch": 2.13, - "learning_rate": 0.00010836397058823528, - "loss": 2.8668, + "learning_rate": 1.0850183823529413e-05, + "loss": 0.1856, "step": 1896 }, { "epoch": 2.13, - "learning_rate": 0.00010822610294117647, - "loss": 4.3593, + "learning_rate": 1.083639705882353e-05, + "loss": 0.2152, "step": 1897 }, { "epoch": 2.13, - "learning_rate": 0.00010808823529411763, - "loss": 2.8158, + "learning_rate": 1.0822610294117648e-05, + "loss": 0.1514, "step": 1898 }, { "epoch": 2.13, - "learning_rate": 0.00010795036764705883, - "loss": 2.8403, + "learning_rate": 1.0808823529411765e-05, + "loss": 0.1521, "step": 1899 }, { "epoch": 2.13, - "learning_rate": 0.00010781249999999998, - "loss": 2.819, + "learning_rate": 1.0795036764705884e-05, + "loss": 0.168, "step": 1900 }, { "epoch": 2.13, - "eval_loss": 3.7971091270446777, - "eval_runtime": 48.2392, - "eval_samples_per_second": 56.033, - "eval_steps_per_second": 0.891, - "eval_wer": 1.0, + "eval_loss": 0.21225057542324066, + "eval_runtime": 47.43, + "eval_samples_per_second": 56.989, + "eval_steps_per_second": 0.907, + "eval_wer": 0.18387191647365905, "step": 1900 }, { "epoch": 2.13, - "learning_rate": 0.00010767463235294115, - "loss": 2.8443, + "learning_rate": 1.078125e-05, + "loss": 0.1777, "step": 1901 }, { "epoch": 2.13, - "learning_rate": 0.00010753676470588234, - "loss": 2.834, + "learning_rate": 1.0767463235294116e-05, + "loss": 0.1627, "step": 1902 }, { "epoch": 2.13, - "learning_rate": 0.00010739889705882351, - "loss": 3.9077, + "learning_rate": 1.0753676470588235e-05, + "loss": 0.1477, "step": 1903 }, { "epoch": 2.13, - "learning_rate": 0.00010726102941176469, - "loss": 2.8282, + "learning_rate": 1.0739889705882352e-05, + "loss": 0.1441, "step": 1904 }, { "epoch": 2.14, - "learning_rate": 0.00010712316176470586, - "loss": 2.8038, + "learning_rate": 1.0726102941176471e-05, + "loss": 0.1516, "step": 1905 }, { "epoch": 2.14, - "learning_rate": 0.00010698529411764705, - "loss": 2.8051, + "learning_rate": 1.0712316176470588e-05, + "loss": 0.1645, "step": 1906 }, { "epoch": 2.14, - "learning_rate": 0.00010684742647058822, - "loss": 2.8256, + "learning_rate": 1.0698529411764707e-05, + "loss": 0.1702, "step": 1907 }, { "epoch": 2.14, - "learning_rate": 0.0001067095588235294, - "loss": 2.8227, + "learning_rate": 1.0684742647058824e-05, + "loss": 0.188, "step": 1908 }, { "epoch": 2.14, - "learning_rate": 0.00010657169117647057, - "loss": 3.114, + "learning_rate": 1.0670955882352942e-05, + "loss": 0.1779, "step": 1909 }, { "epoch": 2.14, - "learning_rate": 0.00010643382352941175, - "loss": 2.8182, + "learning_rate": 1.065716911764706e-05, + "loss": 0.177, "step": 1910 }, { "epoch": 2.14, - "learning_rate": 0.00010629595588235292, - "loss": 2.7982, + "learning_rate": 1.0643382352941176e-05, + "loss": 0.1763, "step": 1911 }, { "epoch": 2.14, - "learning_rate": 0.00010615808823529411, - "loss": 2.8143, + "learning_rate": 1.0629595588235293e-05, + "loss": 0.1681, "step": 1912 }, { "epoch": 2.14, - "learning_rate": 0.00010602022058823528, - "loss": 2.8173, + "learning_rate": 1.0615808823529412e-05, + "loss": 0.1763, "step": 1913 }, { "epoch": 2.15, - "learning_rate": 0.00010588235294117647, - "loss": 2.8216, + "learning_rate": 1.0602022058823529e-05, + "loss": 0.1599, "step": 1914 }, { "epoch": 2.15, - "learning_rate": 0.00010574448529411763, - "loss": 2.955, + "learning_rate": 1.0588235294117648e-05, + "loss": 0.1817, "step": 1915 }, { "epoch": 2.15, - "learning_rate": 0.00010560661764705883, - "loss": 3.7286, + "learning_rate": 1.0574448529411765e-05, + "loss": 0.1785, "step": 1916 }, { "epoch": 2.15, - "learning_rate": 0.00010546874999999999, - "loss": 2.8229, + "learning_rate": 1.0560661764705884e-05, + "loss": 0.1509, "step": 1917 }, { "epoch": 2.15, - "learning_rate": 0.00010533088235294116, - "loss": 2.8027, + "learning_rate": 1.0546875e-05, + "loss": 0.1329, "step": 1918 }, { "epoch": 2.15, - "learning_rate": 0.00010519301470588234, - "loss": 2.8017, + "learning_rate": 1.0533088235294118e-05, + "loss": 0.1758, "step": 1919 }, { "epoch": 2.15, - "learning_rate": 0.00010505514705882351, - "loss": 2.8216, + "learning_rate": 1.0519301470588236e-05, + "loss": 0.1812, "step": 1920 }, { "epoch": 2.15, - "learning_rate": 0.00010491727941176469, - "loss": 2.9639, + "learning_rate": 1.0505514705882353e-05, + "loss": 0.1851, "step": 1921 }, { "epoch": 2.15, - "learning_rate": 0.00010477941176470586, - "loss": 4.3145, + "learning_rate": 1.049172794117647e-05, + "loss": 0.189, "step": 1922 }, { "epoch": 2.16, - "learning_rate": 0.00010464154411764705, - "loss": 2.8071, + "learning_rate": 1.0477941176470587e-05, + "loss": 0.1802, "step": 1923 }, { "epoch": 2.16, - "learning_rate": 0.00010450367647058822, - "loss": 2.8277, + "learning_rate": 1.0464154411764706e-05, + "loss": 0.1299, "step": 1924 }, { "epoch": 2.16, - "learning_rate": 0.0001043658088235294, - "loss": 2.804, + "learning_rate": 1.0450367647058823e-05, + "loss": 0.1772, "step": 1925 }, { "epoch": 2.16, - "learning_rate": 0.00010422794117647057, - "loss": 2.8172, + "learning_rate": 1.0436580882352942e-05, + "loss": 0.1749, "step": 1926 }, { "epoch": 2.16, - "learning_rate": 0.00010409007352941175, - "loss": 2.8447, + "learning_rate": 1.0422794117647059e-05, + "loss": 0.1884, "step": 1927 }, { "epoch": 2.16, - "learning_rate": 0.00010395220588235292, - "loss": 4.7787, + "learning_rate": 1.0409007352941177e-05, + "loss": 0.2303, "step": 1928 }, { "epoch": 2.16, - "learning_rate": 0.00010381433823529411, - "loss": 2.812, + "learning_rate": 1.0395220588235294e-05, + "loss": 0.1682, "step": 1929 }, { "epoch": 2.16, - "learning_rate": 0.00010367647058823528, - "loss": 2.812, + "learning_rate": 1.0381433823529413e-05, + "loss": 0.1743, "step": 1930 }, { "epoch": 2.16, - "learning_rate": 0.00010353860294117646, - "loss": 2.8175, + "learning_rate": 1.036764705882353e-05, + "loss": 0.1436, "step": 1931 }, { "epoch": 2.17, - "learning_rate": 0.00010340073529411763, - "loss": 2.8058, + "learning_rate": 1.0353860294117647e-05, + "loss": 0.1482, "step": 1932 }, { "epoch": 2.17, - "learning_rate": 0.00010326286764705882, - "loss": 2.8437, + "learning_rate": 1.0340073529411764e-05, + "loss": 0.1859, "step": 1933 }, { "epoch": 2.17, - "learning_rate": 0.00010312499999999999, - "loss": 3.1019, + "learning_rate": 1.0326286764705883e-05, + "loss": 0.1902, "step": 1934 }, { "epoch": 2.17, - "learning_rate": 0.00010298713235294116, - "loss": 2.7956, + "learning_rate": 1.03125e-05, + "loss": 0.1729, "step": 1935 }, { "epoch": 2.17, - "learning_rate": 0.00010284926470588234, - "loss": 2.8141, + "learning_rate": 1.0298713235294117e-05, + "loss": 0.1701, "step": 1936 }, { "epoch": 2.17, - "learning_rate": 0.00010271139705882351, - "loss": 2.8094, + "learning_rate": 1.0284926470588236e-05, + "loss": 0.1386, "step": 1937 }, { "epoch": 2.17, - "learning_rate": 0.0001025735294117647, - "loss": 2.8024, + "learning_rate": 1.0271139705882353e-05, + "loss": 0.144, "step": 1938 }, { "epoch": 2.17, - "learning_rate": 0.00010243566176470586, - "loss": 2.8136, + "learning_rate": 1.0257352941176471e-05, + "loss": 0.1528, "step": 1939 }, { "epoch": 2.17, - "learning_rate": 0.00010229779411764705, - "loss": 2.8987, + "learning_rate": 1.0243566176470588e-05, + "loss": 0.147, "step": 1940 }, { "epoch": 2.18, - "learning_rate": 0.00010215992647058822, - "loss": 3.4392, + "learning_rate": 1.0229779411764707e-05, + "loss": 0.2434, "step": 1941 }, { "epoch": 2.18, - "learning_rate": 0.0001020220588235294, - "loss": 2.8023, + "learning_rate": 1.0215992647058824e-05, + "loss": 0.1749, "step": 1942 }, { "epoch": 2.18, - "learning_rate": 0.00010188419117647057, - "loss": 2.8423, + "learning_rate": 1.0202205882352941e-05, + "loss": 0.1291, "step": 1943 }, { "epoch": 2.18, - "learning_rate": 0.00010174632352941175, - "loss": 2.7998, + "learning_rate": 1.0188419117647058e-05, + "loss": 0.148, "step": 1944 }, { "epoch": 2.18, - "learning_rate": 0.00010160845588235293, - "loss": 2.8237, + "learning_rate": 1.0174632352941177e-05, + "loss": 0.1548, "step": 1945 }, { "epoch": 2.18, - "learning_rate": 0.00010147058823529411, - "loss": 2.8997, + "learning_rate": 1.0160845588235294e-05, + "loss": 0.186, "step": 1946 }, { "epoch": 2.18, - "learning_rate": 0.00010133272058823528, - "loss": 4.461, + "learning_rate": 1.0147058823529413e-05, + "loss": 0.2379, "step": 1947 }, { "epoch": 2.18, - "learning_rate": 0.00010119485294117646, - "loss": 2.8286, + "learning_rate": 1.013327205882353e-05, + "loss": 0.1401, "step": 1948 }, { "epoch": 2.18, - "learning_rate": 0.00010105698529411763, - "loss": 2.825, + "learning_rate": 1.0119485294117648e-05, + "loss": 0.146, "step": 1949 }, { "epoch": 2.19, - "learning_rate": 0.00010091911764705882, - "loss": 2.821, + "learning_rate": 1.0105698529411765e-05, + "loss": 0.1542, "step": 1950 }, { "epoch": 2.19, - "learning_rate": 0.00010078124999999999, - "loss": 2.8376, + "learning_rate": 1.0091911764705884e-05, + "loss": 0.1671, "step": 1951 }, { "epoch": 2.19, - "learning_rate": 0.00010064338235294116, - "loss": 2.8161, + "learning_rate": 1.0078125000000001e-05, + "loss": 0.1506, "step": 1952 }, { "epoch": 2.19, - "learning_rate": 0.00010050551470588234, - "loss": 4.0609, + "learning_rate": 1.0064338235294116e-05, + "loss": 0.2244, "step": 1953 }, { "epoch": 2.19, - "learning_rate": 0.00010036764705882351, - "loss": 2.8243, + "learning_rate": 1.0050551470588235e-05, + "loss": 0.1626, "step": 1954 }, { "epoch": 2.19, - "learning_rate": 0.0001002297794117647, - "loss": 2.8105, + "learning_rate": 1.0036764705882352e-05, + "loss": 0.189, "step": 1955 }, { "epoch": 2.19, - "learning_rate": 0.00010009191176470586, - "loss": 2.8232, + "learning_rate": 1.002297794117647e-05, + "loss": 0.1576, "step": 1956 }, { "epoch": 2.19, - "learning_rate": 9.995404411764705e-05, - "loss": 2.8437, + "learning_rate": 1.0009191176470588e-05, + "loss": 0.1526, "step": 1957 }, { "epoch": 2.2, - "learning_rate": 9.981617647058822e-05, - "loss": 2.824, + "learning_rate": 9.995404411764706e-06, + "loss": 0.162, "step": 1958 }, { "epoch": 2.2, - "learning_rate": 9.96783088235294e-05, - "loss": 3.2027, + "learning_rate": 9.981617647058824e-06, + "loss": 0.2006, "step": 1959 }, { "epoch": 2.2, - "learning_rate": 9.954044117647057e-05, - "loss": 2.7968, + "learning_rate": 9.967830882352942e-06, + "loss": 0.1354, "step": 1960 }, { "epoch": 2.2, - "learning_rate": 9.940257352941176e-05, - "loss": 2.818, + "learning_rate": 9.95404411764706e-06, + "loss": 0.1925, "step": 1961 }, { "epoch": 2.2, - "learning_rate": 9.926470588235293e-05, - "loss": 2.8301, + "learning_rate": 9.940257352941178e-06, + "loss": 0.1573, "step": 1962 }, { "epoch": 2.2, - "learning_rate": 9.912683823529411e-05, - "loss": 2.8304, + "learning_rate": 9.926470588235293e-06, + "loss": 0.1602, "step": 1963 }, { "epoch": 2.2, - "learning_rate": 9.898897058823528e-05, - "loss": 2.8176, + "learning_rate": 9.912683823529412e-06, + "loss": 0.1825, "step": 1964 }, { "epoch": 2.2, - "learning_rate": 9.885110294117646e-05, - "loss": 3.0811, + "learning_rate": 9.898897058823529e-06, + "loss": 0.1796, "step": 1965 }, { "epoch": 2.2, - "learning_rate": 9.871323529411763e-05, - "loss": 3.7184, + "learning_rate": 9.885110294117648e-06, + "loss": 0.183, "step": 1966 }, { "epoch": 2.21, - "learning_rate": 9.857536764705882e-05, - "loss": 2.7908, + "learning_rate": 9.871323529411765e-06, + "loss": 0.1696, "step": 1967 }, { "epoch": 2.21, - "learning_rate": 9.843749999999999e-05, - "loss": 2.8028, + "learning_rate": 9.857536764705883e-06, + "loss": 0.1395, "step": 1968 }, { "epoch": 2.21, - "learning_rate": 9.829963235294116e-05, - "loss": 2.8063, + "learning_rate": 9.84375e-06, + "loss": 0.1618, "step": 1969 }, { "epoch": 2.21, - "learning_rate": 9.816176470588234e-05, - "loss": 2.8134, + "learning_rate": 9.829963235294117e-06, + "loss": 0.1475, "step": 1970 }, { "epoch": 2.21, - "learning_rate": 9.802389705882351e-05, - "loss": 2.9011, + "learning_rate": 9.816176470588236e-06, + "loss": 0.1558, "step": 1971 }, { "epoch": 2.21, - "learning_rate": 9.78860294117647e-05, - "loss": 3.9904, + "learning_rate": 9.802389705882353e-06, + "loss": 0.1796, "step": 1972 }, { "epoch": 2.21, - "learning_rate": 9.774816176470586e-05, - "loss": 2.8199, + "learning_rate": 9.788602941176472e-06, + "loss": 0.165, "step": 1973 }, { "epoch": 2.21, - "learning_rate": 9.761029411764705e-05, - "loss": 2.8096, + "learning_rate": 9.774816176470587e-06, + "loss": 0.1485, "step": 1974 }, { "epoch": 2.21, - "learning_rate": 9.747242647058822e-05, - "loss": 2.8038, + "learning_rate": 9.761029411764706e-06, + "loss": 0.1724, "step": 1975 }, { "epoch": 2.22, - "learning_rate": 9.73345588235294e-05, - "loss": 2.8181, + "learning_rate": 9.747242647058823e-06, + "loss": 0.1786, "step": 1976 }, { "epoch": 2.22, - "learning_rate": 9.719669117647057e-05, - "loss": 2.8456, + "learning_rate": 9.733455882352942e-06, + "loss": 0.1862, "step": 1977 }, { "epoch": 2.22, - "learning_rate": 9.705882352941176e-05, - "loss": 4.8086, + "learning_rate": 9.719669117647059e-06, + "loss": 0.2251, "step": 1978 }, { "epoch": 2.22, - "learning_rate": 9.692095588235293e-05, - "loss": 2.822, + "learning_rate": 9.705882352941177e-06, + "loss": 0.1758, "step": 1979 }, { "epoch": 2.22, - "learning_rate": 9.678308823529411e-05, - "loss": 2.8127, + "learning_rate": 9.692095588235294e-06, + "loss": 0.1533, "step": 1980 }, { "epoch": 2.22, - "learning_rate": 9.664522058823528e-05, - "loss": 2.817, + "learning_rate": 9.678308823529413e-06, + "loss": 0.1459, "step": 1981 }, { "epoch": 2.22, - "learning_rate": 9.650735294117646e-05, - "loss": 2.8147, + "learning_rate": 9.66452205882353e-06, + "loss": 0.1639, "step": 1982 }, { "epoch": 2.22, - "learning_rate": 9.636948529411763e-05, - "loss": 2.8458, + "learning_rate": 9.650735294117649e-06, + "loss": 0.1401, "step": 1983 }, { "epoch": 2.22, - "learning_rate": 9.623161764705882e-05, - "loss": 3.0699, + "learning_rate": 9.636948529411764e-06, + "loss": 0.2282, "step": 1984 }, { "epoch": 2.23, - "learning_rate": 9.609374999999999e-05, - "loss": 2.7943, + "learning_rate": 9.623161764705883e-06, + "loss": 0.1673, "step": 1985 }, { "epoch": 2.23, - "learning_rate": 9.595588235294116e-05, - "loss": 2.8171, + "learning_rate": 9.609375e-06, + "loss": 0.1478, "step": 1986 }, { "epoch": 2.23, - "learning_rate": 9.581801470588234e-05, - "loss": 2.7969, + "learning_rate": 9.595588235294117e-06, + "loss": 0.1542, "step": 1987 }, { "epoch": 2.23, - "learning_rate": 9.568014705882351e-05, - "loss": 2.7912, + "learning_rate": 9.581801470588236e-06, + "loss": 0.1384, "step": 1988 }, { "epoch": 2.23, - "learning_rate": 9.55422794117647e-05, - "loss": 2.8089, + "learning_rate": 9.568014705882353e-06, + "loss": 0.1712, "step": 1989 }, { "epoch": 2.23, - "learning_rate": 9.540441176470587e-05, - "loss": 3.0399, + "learning_rate": 9.554227941176471e-06, + "loss": 0.2364, "step": 1990 }, { "epoch": 2.23, - "learning_rate": 9.526654411764705e-05, - "loss": 3.8146, + "learning_rate": 9.540441176470588e-06, + "loss": 0.1954, "step": 1991 }, { "epoch": 2.23, - "learning_rate": 9.512867647058822e-05, - "loss": 2.7966, + "learning_rate": 9.526654411764707e-06, + "loss": 0.1687, "step": 1992 }, { "epoch": 2.23, - "learning_rate": 9.49908088235294e-05, - "loss": 2.8152, + "learning_rate": 9.512867647058824e-06, + "loss": 0.1663, "step": 1993 }, { "epoch": 2.24, - "learning_rate": 9.485294117647057e-05, - "loss": 2.8141, + "learning_rate": 9.499080882352941e-06, + "loss": 0.1773, "step": 1994 }, { "epoch": 2.24, - "learning_rate": 9.471507352941176e-05, - "loss": 2.835, + "learning_rate": 9.485294117647058e-06, + "loss": 0.1754, "step": 1995 }, { "epoch": 2.24, - "learning_rate": 9.457720588235293e-05, - "loss": 2.8828, + "learning_rate": 9.471507352941177e-06, + "loss": 0.1544, "step": 1996 }, { "epoch": 2.24, - "learning_rate": 9.443933823529411e-05, - "loss": 3.8194, + "learning_rate": 9.457720588235294e-06, + "loss": 0.1867, "step": 1997 }, { "epoch": 2.24, - "learning_rate": 9.430147058823528e-05, - "loss": 2.8023, + "learning_rate": 9.443933823529412e-06, + "loss": 0.1685, "step": 1998 }, { "epoch": 2.24, - "learning_rate": 9.416360294117646e-05, - "loss": 2.8097, + "learning_rate": 9.43014705882353e-06, + "loss": 0.1543, "step": 1999 }, { "epoch": 2.24, - "learning_rate": 9.402573529411763e-05, - "loss": 2.8163, + "learning_rate": 9.416360294117648e-06, + "loss": 0.1576, "step": 2000 }, { "epoch": 2.24, - "eval_loss": 3.921595811843872, - "eval_runtime": 48.0333, - "eval_samples_per_second": 56.273, - "eval_steps_per_second": 0.895, - "eval_wer": 1.0, + "eval_loss": 0.20951244235038757, + "eval_runtime": 47.5848, + "eval_samples_per_second": 56.804, + "eval_steps_per_second": 0.904, + "eval_wer": 0.18030587110767987, "step": 2000 }, { "epoch": 2.24, - "learning_rate": 9.388786764705882e-05, - "loss": 2.8309, + "learning_rate": 9.402573529411765e-06, + "loss": 0.168, "step": 2001 }, { "epoch": 2.24, - "learning_rate": 9.374999999999999e-05, - "loss": 2.827, + "learning_rate": 9.388786764705884e-06, + "loss": 0.1531, "step": 2002 }, { "epoch": 2.25, - "learning_rate": 9.361213235294116e-05, - "loss": 3.906, + "learning_rate": 9.375000000000001e-06, + "loss": 0.189, "step": 2003 }, { "epoch": 2.25, - "learning_rate": 9.347426470588234e-05, - "loss": 2.7979, + "learning_rate": 9.361213235294118e-06, + "loss": 0.1702, "step": 2004 }, { "epoch": 2.25, - "learning_rate": 9.333639705882351e-05, - "loss": 2.8099, + "learning_rate": 9.347426470588235e-06, + "loss": 0.1566, "step": 2005 }, { "epoch": 2.25, - "learning_rate": 9.31985294117647e-05, - "loss": 2.8193, + "learning_rate": 9.333639705882352e-06, + "loss": 0.1501, "step": 2006 }, { "epoch": 2.25, - "learning_rate": 9.306066176470587e-05, - "loss": 2.8128, + "learning_rate": 9.31985294117647e-06, + "loss": 0.183, "step": 2007 }, { "epoch": 2.25, - "learning_rate": 9.292279411764705e-05, - "loss": 2.8357, + "learning_rate": 9.306066176470588e-06, + "loss": 0.1747, "step": 2008 }, { "epoch": 2.25, - "learning_rate": 9.278492647058822e-05, - "loss": 3.102, + "learning_rate": 9.292279411764706e-06, + "loss": 0.2538, "step": 2009 }, { "epoch": 2.25, - "learning_rate": 9.26470588235294e-05, - "loss": 2.833, + "learning_rate": 9.278492647058823e-06, + "loss": 0.15, "step": 2010 }, { "epoch": 2.25, - "learning_rate": 9.250919117647057e-05, - "loss": 2.8106, + "learning_rate": 9.264705882352942e-06, + "loss": 0.1514, "step": 2011 }, { "epoch": 2.26, - "learning_rate": 9.237132352941176e-05, - "loss": 2.8124, + "learning_rate": 9.250919117647059e-06, + "loss": 0.1678, "step": 2012 }, { "epoch": 2.26, - "learning_rate": 9.223345588235293e-05, - "loss": 2.8063, + "learning_rate": 9.237132352941178e-06, + "loss": 0.1792, "step": 2013 }, { "epoch": 2.26, - "learning_rate": 9.209558823529411e-05, - "loss": 2.8047, + "learning_rate": 9.223345588235295e-06, + "loss": 0.1513, "step": 2014 }, { "epoch": 2.26, - "learning_rate": 9.195772058823528e-05, - "loss": 2.9635, + "learning_rate": 9.209558823529412e-06, + "loss": 0.1644, "step": 2015 }, { "epoch": 2.26, - "learning_rate": 9.181985294117647e-05, - "loss": 3.3116, + "learning_rate": 9.195772058823529e-06, + "loss": 0.1959, "step": 2016 }, { "epoch": 2.26, - "learning_rate": 9.168198529411764e-05, - "loss": 2.7884, + "learning_rate": 9.181985294117648e-06, + "loss": 0.1634, "step": 2017 }, { "epoch": 2.26, - "learning_rate": 9.154411764705882e-05, - "loss": 2.8026, + "learning_rate": 9.168198529411765e-06, + "loss": 0.1852, "step": 2018 }, { "epoch": 2.26, - "learning_rate": 9.140624999999999e-05, - "loss": 2.7997, + "learning_rate": 9.154411764705883e-06, + "loss": 0.1484, "step": 2019 }, { "epoch": 2.26, - "learning_rate": 9.126838235294116e-05, - "loss": 2.8281, + "learning_rate": 9.140625e-06, + "loss": 0.1405, "step": 2020 }, { "epoch": 2.27, - "learning_rate": 9.113051470588234e-05, - "loss": 2.936, + "learning_rate": 9.126838235294117e-06, + "loss": 0.1475, "step": 2021 }, { "epoch": 2.27, - "learning_rate": 9.099264705882351e-05, - "loss": 4.3016, + "learning_rate": 9.113051470588236e-06, + "loss": 0.1933, "step": 2022 }, { "epoch": 2.27, - "learning_rate": 9.08547794117647e-05, - "loss": 2.8077, + "learning_rate": 9.099264705882353e-06, + "loss": 0.1651, "step": 2023 }, { "epoch": 2.27, - "learning_rate": 9.071691176470587e-05, - "loss": 2.8141, + "learning_rate": 9.085477941176472e-06, + "loss": 0.163, "step": 2024 }, { "epoch": 2.27, - "learning_rate": 9.057904411764705e-05, - "loss": 2.8116, + "learning_rate": 9.071691176470589e-06, + "loss": 0.1836, "step": 2025 }, { "epoch": 2.27, - "learning_rate": 9.044117647058822e-05, - "loss": 2.7925, + "learning_rate": 9.057904411764706e-06, + "loss": 0.1807, "step": 2026 }, { "epoch": 2.27, - "learning_rate": 9.03033088235294e-05, - "loss": 2.8627, + "learning_rate": 9.044117647058823e-06, + "loss": 0.2017, "step": 2027 }, { "epoch": 2.27, - "learning_rate": 9.016544117647057e-05, - "loss": 4.7428, + "learning_rate": 9.030330882352941e-06, + "loss": 0.1754, "step": 2028 }, { "epoch": 2.27, - "learning_rate": 9.002757352941176e-05, - "loss": 2.7982, + "learning_rate": 9.016544117647058e-06, + "loss": 0.1687, "step": 2029 }, { "epoch": 2.28, - "learning_rate": 8.988970588235293e-05, - "loss": 2.8209, + "learning_rate": 9.002757352941177e-06, + "loss": 0.1619, "step": 2030 }, { "epoch": 2.28, - "learning_rate": 8.975183823529411e-05, - "loss": 2.8014, + "learning_rate": 8.988970588235294e-06, + "loss": 0.1393, "step": 2031 }, { "epoch": 2.28, - "learning_rate": 8.961397058823528e-05, - "loss": 2.8224, + "learning_rate": 8.975183823529413e-06, + "loss": 0.1622, "step": 2032 }, { "epoch": 2.28, - "learning_rate": 8.947610294117647e-05, - "loss": 2.827, + "learning_rate": 8.96139705882353e-06, + "loss": 0.1716, "step": 2033 }, { "epoch": 2.28, - "learning_rate": 8.933823529411764e-05, - "loss": 3.1191, + "learning_rate": 8.947610294117649e-06, + "loss": 0.2203, "step": 2034 }, { "epoch": 2.28, - "learning_rate": 8.920036764705882e-05, - "loss": 2.8064, + "learning_rate": 8.933823529411766e-06, + "loss": 0.1819, "step": 2035 }, { "epoch": 2.28, - "learning_rate": 8.906249999999999e-05, - "loss": 2.8205, + "learning_rate": 8.920036764705883e-06, + "loss": 0.1567, "step": 2036 }, { "epoch": 2.28, - "learning_rate": 8.892463235294116e-05, - "loss": 2.8182, + "learning_rate": 8.90625e-06, + "loss": 0.1693, "step": 2037 }, { "epoch": 2.28, - "learning_rate": 8.878676470588234e-05, - "loss": 2.8439, + "learning_rate": 8.892463235294117e-06, + "loss": 0.1532, "step": 2038 }, { "epoch": 2.29, - "learning_rate": 8.864889705882351e-05, - "loss": 2.8248, + "learning_rate": 8.878676470588235e-06, + "loss": 0.1623, "step": 2039 }, { "epoch": 2.29, - "learning_rate": 8.85110294117647e-05, - "loss": 2.8954, + "learning_rate": 8.864889705882352e-06, + "loss": 0.1845, "step": 2040 }, { "epoch": 2.29, - "learning_rate": 8.837316176470587e-05, - "loss": 3.645, + "learning_rate": 8.851102941176471e-06, + "loss": 0.1728, "step": 2041 }, { "epoch": 2.29, - "learning_rate": 8.823529411764705e-05, - "loss": 2.827, + "learning_rate": 8.837316176470588e-06, + "loss": 0.147, "step": 2042 }, { "epoch": 2.29, - "learning_rate": 8.809742647058822e-05, - "loss": 2.8102, + "learning_rate": 8.823529411764707e-06, + "loss": 0.1523, "step": 2043 }, { "epoch": 2.29, - "learning_rate": 8.79595588235294e-05, - "loss": 2.8254, + "learning_rate": 8.809742647058824e-06, + "loss": 0.1617, "step": 2044 }, { "epoch": 2.29, - "learning_rate": 8.782169117647058e-05, - "loss": 2.8244, + "learning_rate": 8.795955882352943e-06, + "loss": 0.1671, "step": 2045 }, { "epoch": 2.29, - "learning_rate": 8.768382352941176e-05, - "loss": 2.8383, + "learning_rate": 8.782169117647058e-06, + "loss": 0.1618, "step": 2046 }, { "epoch": 2.29, - "learning_rate": 8.754595588235293e-05, - "loss": 3.8393, + "learning_rate": 8.768382352941177e-06, + "loss": 0.1691, "step": 2047 }, { "epoch": 2.3, - "learning_rate": 8.740808823529411e-05, - "loss": 2.8055, + "learning_rate": 8.754595588235294e-06, + "loss": 0.1398, "step": 2048 }, { "epoch": 2.3, - "learning_rate": 8.727022058823528e-05, - "loss": 2.8071, + "learning_rate": 8.740808823529412e-06, + "loss": 0.1517, "step": 2049 }, { "epoch": 2.3, - "learning_rate": 8.713235294117647e-05, - "loss": 2.815, + "learning_rate": 8.72702205882353e-06, + "loss": 0.1574, "step": 2050 }, { "epoch": 2.3, - "learning_rate": 8.699448529411764e-05, - "loss": 2.813, + "learning_rate": 8.713235294117648e-06, + "loss": 0.1562, "step": 2051 }, { "epoch": 2.3, - "learning_rate": 8.685661764705882e-05, - "loss": 2.848, + "learning_rate": 8.699448529411765e-06, + "loss": 0.1809, "step": 2052 }, { "epoch": 2.3, - "learning_rate": 8.671874999999999e-05, - "loss": 4.3169, + "learning_rate": 8.685661764705884e-06, + "loss": 0.1816, "step": 2053 }, { "epoch": 2.3, - "learning_rate": 8.658088235294116e-05, - "loss": 2.8035, + "learning_rate": 8.671875e-06, + "loss": 0.1919, "step": 2054 }, { "epoch": 2.3, - "learning_rate": 8.644301470588234e-05, - "loss": 2.8206, + "learning_rate": 8.658088235294118e-06, + "loss": 0.181, "step": 2055 }, { "epoch": 2.3, - "learning_rate": 8.630514705882351e-05, - "loss": 2.8253, + "learning_rate": 8.644301470588236e-06, + "loss": 0.1366, "step": 2056 }, { "epoch": 2.31, - "learning_rate": 8.61672794117647e-05, - "loss": 2.8259, + "learning_rate": 8.630514705882352e-06, + "loss": 0.1779, "step": 2057 }, { "epoch": 2.31, - "learning_rate": 8.602941176470587e-05, - "loss": 2.8469, + "learning_rate": 8.61672794117647e-06, + "loss": 0.1563, "step": 2058 }, { "epoch": 2.31, - "learning_rate": 8.589154411764705e-05, - "loss": 3.3163, + "learning_rate": 8.602941176470587e-06, + "loss": 0.188, "step": 2059 }, { "epoch": 2.31, - "learning_rate": 8.575367647058822e-05, - "loss": 2.806, + "learning_rate": 8.589154411764706e-06, + "loss": 0.1778, "step": 2060 }, { "epoch": 2.31, - "learning_rate": 8.56158088235294e-05, - "loss": 2.8064, + "learning_rate": 8.575367647058823e-06, + "loss": 0.1712, "step": 2061 }, { "epoch": 2.31, - "learning_rate": 8.547794117647058e-05, - "loss": 2.8113, + "learning_rate": 8.561580882352942e-06, + "loss": 0.1314, "step": 2062 }, { "epoch": 2.31, - "learning_rate": 8.534007352941176e-05, - "loss": 2.8286, + "learning_rate": 8.547794117647059e-06, + "loss": 0.16, "step": 2063 }, { "epoch": 2.31, - "learning_rate": 8.520220588235293e-05, - "loss": 2.8303, + "learning_rate": 8.534007352941178e-06, + "loss": 0.1488, "step": 2064 }, { "epoch": 2.32, - "learning_rate": 8.506433823529411e-05, - "loss": 2.8664, + "learning_rate": 8.520220588235295e-06, + "loss": 0.1642, "step": 2065 }, { "epoch": 2.32, - "learning_rate": 8.492647058823528e-05, - "loss": 3.5573, + "learning_rate": 8.506433823529413e-06, + "loss": 0.1421, "step": 2066 }, { "epoch": 2.32, - "learning_rate": 8.478860294117647e-05, - "loss": 2.8074, + "learning_rate": 8.492647058823529e-06, + "loss": 0.1339, "step": 2067 }, { "epoch": 2.32, - "learning_rate": 8.465073529411764e-05, - "loss": 2.8289, + "learning_rate": 8.478860294117647e-06, + "loss": 0.141, "step": 2068 }, { "epoch": 2.32, - "learning_rate": 8.451286764705882e-05, - "loss": 2.8217, + "learning_rate": 8.465073529411764e-06, + "loss": 0.166, "step": 2069 }, { "epoch": 2.32, - "learning_rate": 8.437499999999999e-05, - "loss": 2.8121, + "learning_rate": 8.451286764705883e-06, + "loss": 0.1533, "step": 2070 }, { "epoch": 2.32, - "learning_rate": 8.423713235294116e-05, - "loss": 2.8812, + "learning_rate": 8.4375e-06, + "loss": 0.1534, "step": 2071 }, { "epoch": 2.32, - "learning_rate": 8.409926470588234e-05, - "loss": 3.7369, + "learning_rate": 8.423713235294117e-06, + "loss": 0.1508, "step": 2072 }, { "epoch": 2.32, - "learning_rate": 8.396139705882351e-05, - "loss": 2.8063, + "learning_rate": 8.409926470588236e-06, + "loss": 0.1596, "step": 2073 }, { "epoch": 2.33, - "learning_rate": 8.38235294117647e-05, - "loss": 2.8145, + "learning_rate": 8.396139705882353e-06, + "loss": 0.1588, "step": 2074 }, { "epoch": 2.33, - "learning_rate": 8.368566176470587e-05, - "loss": 2.8069, + "learning_rate": 8.382352941176472e-06, + "loss": 0.172, "step": 2075 }, { "epoch": 2.33, - "learning_rate": 8.354779411764705e-05, - "loss": 2.7962, + "learning_rate": 8.368566176470589e-06, + "loss": 0.1562, "step": 2076 }, { "epoch": 2.33, - "learning_rate": 8.340992647058822e-05, - "loss": 2.8244, + "learning_rate": 8.354779411764706e-06, + "loss": 0.1579, "step": 2077 }, { "epoch": 2.33, - "learning_rate": 8.32720588235294e-05, - "loss": 4.5267, + "learning_rate": 8.340992647058823e-06, + "loss": 0.1828, "step": 2078 }, { "epoch": 2.33, - "learning_rate": 8.313419117647058e-05, - "loss": 2.812, + "learning_rate": 8.327205882352941e-06, + "loss": 0.1317, "step": 2079 }, { "epoch": 2.33, - "learning_rate": 8.299632352941176e-05, - "loss": 2.814, + "learning_rate": 8.313419117647058e-06, + "loss": 0.1351, "step": 2080 }, { "epoch": 2.33, - "learning_rate": 8.285845588235293e-05, - "loss": 2.8244, + "learning_rate": 8.299632352941177e-06, + "loss": 0.1424, "step": 2081 }, { "epoch": 2.33, - "learning_rate": 8.272058823529411e-05, - "loss": 2.8125, + "learning_rate": 8.285845588235294e-06, + "loss": 0.182, "step": 2082 }, { "epoch": 2.34, - "learning_rate": 8.258272058823528e-05, - "loss": 2.8335, + "learning_rate": 8.272058823529413e-06, + "loss": 0.1713, "step": 2083 }, { "epoch": 2.34, - "learning_rate": 8.244485294117647e-05, - "loss": 3.1585, + "learning_rate": 8.25827205882353e-06, + "loss": 0.2299, "step": 2084 }, { "epoch": 2.34, - "learning_rate": 8.230698529411764e-05, - "loss": 2.8237, + "learning_rate": 8.244485294117648e-06, + "loss": 0.1517, "step": 2085 }, { "epoch": 2.34, - "learning_rate": 8.216911764705882e-05, - "loss": 2.8138, + "learning_rate": 8.230698529411765e-06, + "loss": 0.1765, "step": 2086 }, { "epoch": 2.34, - "learning_rate": 8.203124999999999e-05, - "loss": 2.8052, + "learning_rate": 8.216911764705884e-06, + "loss": 0.1572, "step": 2087 }, { "epoch": 2.34, - "learning_rate": 8.189338235294116e-05, - "loss": 2.8153, + "learning_rate": 8.203125e-06, + "loss": 0.153, "step": 2088 }, { "epoch": 2.34, - "learning_rate": 8.175551470588235e-05, - "loss": 2.8129, + "learning_rate": 8.189338235294117e-06, + "loss": 0.1961, "step": 2089 }, { "epoch": 2.34, - "learning_rate": 8.161764705882352e-05, - "loss": 3.0119, + "learning_rate": 8.175551470588235e-06, + "loss": 0.2183, "step": 2090 }, { "epoch": 2.34, - "learning_rate": 8.14797794117647e-05, - "loss": 3.8393, + "learning_rate": 8.161764705882352e-06, + "loss": 0.1634, "step": 2091 }, { "epoch": 2.35, - "learning_rate": 8.134191176470587e-05, - "loss": 2.8054, + "learning_rate": 8.147977941176471e-06, + "loss": 0.1651, "step": 2092 }, { "epoch": 2.35, - "learning_rate": 8.120404411764705e-05, - "loss": 2.8175, + "learning_rate": 8.134191176470588e-06, + "loss": 0.1562, "step": 2093 }, { "epoch": 2.35, - "learning_rate": 8.106617647058822e-05, - "loss": 2.8197, + "learning_rate": 8.120404411764707e-06, + "loss": 0.1953, "step": 2094 }, { "epoch": 2.35, - "learning_rate": 8.092830882352941e-05, - "loss": 2.8326, + "learning_rate": 8.106617647058824e-06, + "loss": 0.1331, "step": 2095 }, { "epoch": 2.35, - "learning_rate": 8.079044117647058e-05, - "loss": 2.899, + "learning_rate": 8.092830882352942e-06, + "loss": 0.1668, "step": 2096 }, { "epoch": 2.35, - "learning_rate": 8.065257352941176e-05, - "loss": 4.2006, + "learning_rate": 8.07904411764706e-06, + "loss": 0.1589, "step": 2097 }, { "epoch": 2.35, - "learning_rate": 8.051470588235293e-05, - "loss": 2.8018, + "learning_rate": 8.065257352941176e-06, + "loss": 0.1432, "step": 2098 }, { "epoch": 2.35, - "learning_rate": 8.037683823529411e-05, - "loss": 2.8263, + "learning_rate": 8.051470588235293e-06, + "loss": 0.171, "step": 2099 }, { "epoch": 2.35, - "learning_rate": 8.023897058823528e-05, - "loss": 2.8214, + "learning_rate": 8.037683823529412e-06, + "loss": 0.1756, "step": 2100 }, { "epoch": 2.35, - "eval_loss": 3.9177870750427246, - "eval_runtime": 48.0272, - "eval_samples_per_second": 56.281, + "eval_loss": 0.20747698843479156, + "eval_runtime": 48.0344, + "eval_samples_per_second": 56.272, "eval_steps_per_second": 0.895, - "eval_wer": 1.0, + "eval_wer": 0.17764052792176757, "step": 2100 }, { "epoch": 2.36, - "learning_rate": 8.010110294117647e-05, - "loss": 2.8037, + "learning_rate": 8.023897058823529e-06, + "loss": 0.1753, "step": 2101 }, { "epoch": 2.36, - "learning_rate": 7.996323529411764e-05, - "loss": 2.8345, + "learning_rate": 8.010110294117648e-06, + "loss": 0.1709, "step": 2102 }, { "epoch": 2.36, - "learning_rate": 7.982536764705882e-05, - "loss": 4.5048, + "learning_rate": 7.996323529411765e-06, + "loss": 0.2331, "step": 2103 }, { "epoch": 2.36, - "learning_rate": 7.968749999999999e-05, - "loss": 2.8271, + "learning_rate": 7.982536764705884e-06, + "loss": 0.1719, "step": 2104 }, { "epoch": 2.36, - "learning_rate": 7.954963235294116e-05, - "loss": 2.8167, + "learning_rate": 7.96875e-06, + "loss": 0.1492, "step": 2105 }, { "epoch": 2.36, - "learning_rate": 7.941176470588235e-05, - "loss": 2.8052, + "learning_rate": 7.954963235294118e-06, + "loss": 0.1556, "step": 2106 }, { "epoch": 2.36, - "learning_rate": 7.927389705882352e-05, - "loss": 2.8246, + "learning_rate": 7.941176470588236e-06, + "loss": 0.154, "step": 2107 }, { "epoch": 2.36, - "learning_rate": 7.91360294117647e-05, - "loss": 2.863, + "learning_rate": 7.927389705882352e-06, + "loss": 0.1104, "step": 2108 }, { "epoch": 2.36, - "learning_rate": 7.899816176470587e-05, - "loss": 3.2715, + "learning_rate": 7.91360294117647e-06, + "loss": 0.2572, "step": 2109 }, { "epoch": 2.37, - "learning_rate": 7.886029411764705e-05, - "loss": 2.7956, + "learning_rate": 7.899816176470587e-06, + "loss": 0.1419, "step": 2110 }, { "epoch": 2.37, - "learning_rate": 7.872242647058822e-05, - "loss": 2.8025, + "learning_rate": 7.886029411764706e-06, + "loss": 0.1745, "step": 2111 }, { "epoch": 2.37, - "learning_rate": 7.858455882352941e-05, - "loss": 2.8204, + "learning_rate": 7.872242647058823e-06, + "loss": 0.1593, "step": 2112 }, { "epoch": 2.37, - "learning_rate": 7.844669117647058e-05, - "loss": 2.8197, + "learning_rate": 7.858455882352942e-06, + "loss": 0.1665, "step": 2113 }, { "epoch": 2.37, - "learning_rate": 7.830882352941176e-05, - "loss": 2.8406, + "learning_rate": 7.844669117647059e-06, + "loss": 0.1804, "step": 2114 }, { "epoch": 2.37, - "learning_rate": 7.817095588235293e-05, - "loss": 3.0686, + "learning_rate": 7.830882352941177e-06, + "loss": 0.168, "step": 2115 }, { "epoch": 2.37, - "learning_rate": 7.803308823529412e-05, - "loss": 3.6766, + "learning_rate": 7.817095588235294e-06, + "loss": 0.1988, "step": 2116 }, { "epoch": 2.37, - "learning_rate": 7.789522058823529e-05, - "loss": 2.8043, + "learning_rate": 7.803308823529413e-06, + "loss": 0.1554, "step": 2117 }, { "epoch": 2.37, - "learning_rate": 7.775735294117647e-05, - "loss": 2.8223, + "learning_rate": 7.78952205882353e-06, + "loss": 0.1668, "step": 2118 }, { "epoch": 2.38, - "learning_rate": 7.761948529411764e-05, - "loss": 2.7984, + "learning_rate": 7.775735294117647e-06, + "loss": 0.1752, "step": 2119 }, { "epoch": 2.38, - "learning_rate": 7.748161764705882e-05, - "loss": 2.8187, + "learning_rate": 7.761948529411764e-06, + "loss": 0.1452, "step": 2120 }, { "epoch": 2.38, - "learning_rate": 7.734374999999999e-05, - "loss": 2.9287, + "learning_rate": 7.748161764705883e-06, + "loss": 0.1658, "step": 2121 }, { "epoch": 2.38, - "learning_rate": 7.720588235294116e-05, - "loss": 3.9767, + "learning_rate": 7.734375e-06, + "loss": 0.1929, "step": 2122 }, { "epoch": 2.38, - "learning_rate": 7.706801470588235e-05, - "loss": 2.8067, + "learning_rate": 7.720588235294117e-06, + "loss": 0.154, "step": 2123 }, { "epoch": 2.38, - "learning_rate": 7.693014705882352e-05, - "loss": 2.8119, + "learning_rate": 7.706801470588236e-06, + "loss": 0.1356, "step": 2124 }, { "epoch": 2.38, - "learning_rate": 7.67922794117647e-05, - "loss": 2.8092, + "learning_rate": 7.693014705882353e-06, + "loss": 0.1398, "step": 2125 }, { "epoch": 2.38, - "learning_rate": 7.665441176470587e-05, - "loss": 2.8078, + "learning_rate": 7.679227941176471e-06, + "loss": 0.1424, "step": 2126 }, { "epoch": 2.38, - "learning_rate": 7.651654411764705e-05, - "loss": 2.8314, + "learning_rate": 7.665441176470588e-06, + "loss": 0.1812, "step": 2127 }, { "epoch": 2.39, - "learning_rate": 7.637867647058822e-05, - "loss": 4.0292, + "learning_rate": 7.651654411764707e-06, + "loss": 0.1437, "step": 2128 }, { "epoch": 2.39, - "learning_rate": 7.624080882352941e-05, - "loss": 2.7955, + "learning_rate": 7.637867647058822e-06, + "loss": 0.1383, "step": 2129 }, { "epoch": 2.39, - "learning_rate": 7.610294117647058e-05, - "loss": 2.8121, + "learning_rate": 7.624080882352942e-06, + "loss": 0.1329, "step": 2130 }, { "epoch": 2.39, - "learning_rate": 7.596507352941176e-05, - "loss": 2.8342, + "learning_rate": 7.610294117647059e-06, + "loss": 0.1359, "step": 2131 }, { "epoch": 2.39, - "learning_rate": 7.582720588235293e-05, - "loss": 2.8157, + "learning_rate": 7.596507352941177e-06, + "loss": 0.1305, "step": 2132 }, { "epoch": 2.39, - "learning_rate": 7.568933823529412e-05, - "loss": 2.832, + "learning_rate": 7.582720588235294e-06, + "loss": 0.1698, "step": 2133 }, { "epoch": 2.39, - "learning_rate": 7.555147058823529e-05, - "loss": 3.0698, + "learning_rate": 7.568933823529413e-06, + "loss": 0.1665, "step": 2134 }, { "epoch": 2.39, - "learning_rate": 7.541360294117647e-05, - "loss": 2.8023, + "learning_rate": 7.55514705882353e-06, + "loss": 0.1484, "step": 2135 }, { "epoch": 2.39, - "learning_rate": 7.527573529411764e-05, - "loss": 2.8028, + "learning_rate": 7.541360294117648e-06, + "loss": 0.1548, "step": 2136 }, { "epoch": 2.4, - "learning_rate": 7.513786764705882e-05, - "loss": 2.8018, + "learning_rate": 7.5275735294117645e-06, + "loss": 0.1617, "step": 2137 }, { "epoch": 2.4, - "learning_rate": 7.5e-05, - "loss": 2.8134, + "learning_rate": 7.513786764705883e-06, + "loss": 0.1495, "step": 2138 }, { "epoch": 2.4, - "learning_rate": 7.486213235294118e-05, - "loss": 2.8076, + "learning_rate": 7.5e-06, + "loss": 0.1565, "step": 2139 }, { "epoch": 2.4, - "learning_rate": 7.472426470588235e-05, - "loss": 2.8846, + "learning_rate": 7.486213235294118e-06, + "loss": 0.1634, "step": 2140 }, { "epoch": 2.4, - "learning_rate": 7.458639705882353e-05, - "loss": 4.1729, + "learning_rate": 7.472426470588236e-06, + "loss": 0.1451, "step": 2141 }, { "epoch": 2.4, - "learning_rate": 7.44485294117647e-05, - "loss": 2.809, + "learning_rate": 7.458639705882353e-06, + "loss": 0.1506, "step": 2142 }, { "epoch": 2.4, - "learning_rate": 7.431066176470588e-05, - "loss": 2.824, + "learning_rate": 7.444852941176471e-06, + "loss": 0.1653, "step": 2143 }, { "epoch": 2.4, - "learning_rate": 7.417279411764705e-05, - "loss": 2.8039, + "learning_rate": 7.431066176470589e-06, + "loss": 0.1545, "step": 2144 }, { "epoch": 2.4, - "learning_rate": 7.403492647058824e-05, - "loss": 2.835, + "learning_rate": 7.4172794117647065e-06, + "loss": 0.1601, "step": 2145 }, { "epoch": 2.41, - "learning_rate": 7.389705882352941e-05, - "loss": 2.8424, + "learning_rate": 7.403492647058824e-06, + "loss": 0.1277, "step": 2146 }, { "epoch": 2.41, - "learning_rate": 7.375919117647058e-05, - "loss": 3.3652, + "learning_rate": 7.389705882352941e-06, + "loss": 0.1784, "step": 2147 }, { "epoch": 2.41, - "learning_rate": 7.362132352941175e-05, - "loss": 2.7962, + "learning_rate": 7.375919117647058e-06, + "loss": 0.1704, "step": 2148 }, { "epoch": 2.41, - "learning_rate": 7.348345588235293e-05, - "loss": 2.8013, + "learning_rate": 7.362132352941176e-06, + "loss": 0.1857, "step": 2149 }, { "epoch": 2.41, - "learning_rate": 7.33455882352941e-05, - "loss": 2.8145, + "learning_rate": 7.348345588235294e-06, + "loss": 0.1384, "step": 2150 }, { "epoch": 2.41, - "learning_rate": 7.320772058823529e-05, - "loss": 2.8259, + "learning_rate": 7.334558823529412e-06, + "loss": 0.1444, "step": 2151 }, { "epoch": 2.41, - "learning_rate": 7.306985294117646e-05, - "loss": 2.8165, + "learning_rate": 7.320772058823529e-06, + "loss": 0.1472, "step": 2152 }, { "epoch": 2.41, - "learning_rate": 7.293198529411764e-05, - "loss": 4.486, + "learning_rate": 7.306985294117647e-06, + "loss": 0.2185, "step": 2153 }, { "epoch": 2.41, - "learning_rate": 7.279411764705881e-05, - "loss": 2.8131, + "learning_rate": 7.293198529411765e-06, + "loss": 0.1515, "step": 2154 }, { "epoch": 2.42, - "learning_rate": 7.265625e-05, - "loss": 2.7978, + "learning_rate": 7.2794117647058826e-06, + "loss": 0.1533, "step": 2155 }, { "epoch": 2.42, - "learning_rate": 7.251838235294116e-05, - "loss": 2.8093, + "learning_rate": 7.265625e-06, + "loss": 0.1628, "step": 2156 }, { "epoch": 2.42, - "learning_rate": 7.238051470588235e-05, - "loss": 2.8228, + "learning_rate": 7.251838235294118e-06, + "loss": 0.1397, "step": 2157 }, { "epoch": 2.42, - "learning_rate": 7.224264705882352e-05, - "loss": 2.8468, + "learning_rate": 7.238051470588235e-06, + "loss": 0.1805, "step": 2158 }, { "epoch": 2.42, - "learning_rate": 7.21047794117647e-05, - "loss": 3.1293, + "learning_rate": 7.224264705882353e-06, + "loss": 0.2382, "step": 2159 }, { "epoch": 2.42, - "learning_rate": 7.196691176470587e-05, - "loss": 2.8045, + "learning_rate": 7.210477941176471e-06, + "loss": 0.1612, "step": 2160 }, { "epoch": 2.42, - "learning_rate": 7.182904411764706e-05, - "loss": 2.7949, + "learning_rate": 7.196691176470589e-06, + "loss": 0.1592, "step": 2161 }, { "epoch": 2.42, - "learning_rate": 7.169117647058824e-05, - "loss": 2.8057, + "learning_rate": 7.182904411764707e-06, + "loss": 0.1648, "step": 2162 }, { "epoch": 2.42, - "learning_rate": 7.155330882352941e-05, - "loss": 2.806, + "learning_rate": 7.169117647058824e-06, + "loss": 0.1426, "step": 2163 }, { "epoch": 2.43, - "learning_rate": 7.141544117647058e-05, - "loss": 2.8181, + "learning_rate": 7.155330882352942e-06, + "loss": 0.1311, "step": 2164 }, { "epoch": 2.43, - "learning_rate": 7.127757352941175e-05, - "loss": 3.0588, + "learning_rate": 7.141544117647059e-06, + "loss": 0.2119, "step": 2165 }, { "epoch": 2.43, - "learning_rate": 7.113970588235293e-05, - "loss": 3.7185, + "learning_rate": 7.1277573529411765e-06, + "loss": 0.1862, "step": 2166 }, { "epoch": 2.43, - "learning_rate": 7.10018382352941e-05, - "loss": 2.8179, + "learning_rate": 7.113970588235294e-06, + "loss": 0.1401, "step": 2167 }, { "epoch": 2.43, - "learning_rate": 7.086397058823529e-05, - "loss": 2.8161, + "learning_rate": 7.100183823529411e-06, + "loss": 0.1634, "step": 2168 }, { "epoch": 2.43, - "learning_rate": 7.072610294117646e-05, - "loss": 2.8094, + "learning_rate": 7.086397058823529e-06, + "loss": 0.1639, "step": 2169 }, { "epoch": 2.43, - "learning_rate": 7.058823529411764e-05, - "loss": 2.8232, + "learning_rate": 7.072610294117647e-06, + "loss": 0.1723, "step": 2170 }, { "epoch": 2.43, - "learning_rate": 7.045036764705881e-05, - "loss": 2.9881, + "learning_rate": 7.058823529411765e-06, + "loss": 0.1699, "step": 2171 }, { "epoch": 2.43, - "learning_rate": 7.03125e-05, - "loss": 4.9846, + "learning_rate": 7.045036764705883e-06, + "loss": 0.251, "step": 2172 }, { "epoch": 2.44, - "learning_rate": 7.017463235294116e-05, - "loss": 2.814, + "learning_rate": 7.03125e-06, + "loss": 0.1431, "step": 2173 }, { "epoch": 2.44, - "learning_rate": 7.003676470588235e-05, - "loss": 2.8177, + "learning_rate": 7.017463235294118e-06, + "loss": 0.1497, "step": 2174 }, { "epoch": 2.44, - "learning_rate": 6.989889705882352e-05, - "loss": 2.8257, + "learning_rate": 7.0036764705882355e-06, + "loss": 0.1345, "step": 2175 }, { "epoch": 2.44, - "learning_rate": 6.97610294117647e-05, - "loss": 2.8312, + "learning_rate": 6.989889705882353e-06, + "loss": 0.1824, "step": 2176 }, { "epoch": 2.44, - "learning_rate": 6.962316176470587e-05, - "loss": 2.869, + "learning_rate": 6.976102941176471e-06, + "loss": 0.2159, "step": 2177 }, { "epoch": 2.44, - "learning_rate": 6.948529411764706e-05, - "loss": 4.4795, + "learning_rate": 6.962316176470588e-06, + "loss": 0.1812, "step": 2178 }, { "epoch": 2.44, - "learning_rate": 6.934742647058823e-05, - "loss": 2.811, + "learning_rate": 6.948529411764706e-06, + "loss": 0.1636, "step": 2179 }, { "epoch": 2.44, - "learning_rate": 6.920955882352941e-05, - "loss": 2.8239, + "learning_rate": 6.934742647058824e-06, + "loss": 0.1573, "step": 2180 }, { "epoch": 2.45, - "learning_rate": 6.907169117647058e-05, - "loss": 2.8177, + "learning_rate": 6.920955882352942e-06, + "loss": 0.1869, "step": 2181 }, { "epoch": 2.45, - "learning_rate": 6.893382352941175e-05, - "loss": 2.813, + "learning_rate": 6.907169117647059e-06, + "loss": 0.1605, "step": 2182 }, { "epoch": 2.45, - "learning_rate": 6.879595588235293e-05, - "loss": 2.8194, + "learning_rate": 6.893382352941177e-06, + "loss": 0.1615, "step": 2183 }, { "epoch": 2.45, - "learning_rate": 6.86580882352941e-05, - "loss": 3.1379, + "learning_rate": 6.879595588235294e-06, + "loss": 0.1976, "step": 2184 }, { "epoch": 2.45, - "learning_rate": 6.852022058823529e-05, - "loss": 2.8081, + "learning_rate": 6.865808823529412e-06, + "loss": 0.1689, "step": 2185 }, { "epoch": 2.45, - "learning_rate": 6.838235294117646e-05, - "loss": 2.8314, + "learning_rate": 6.8520220588235294e-06, + "loss": 0.198, "step": 2186 }, { "epoch": 2.45, - "learning_rate": 6.824448529411764e-05, - "loss": 2.8126, + "learning_rate": 6.838235294117647e-06, + "loss": 0.1972, "step": 2187 }, { "epoch": 2.45, - "learning_rate": 6.810661764705881e-05, - "loss": 2.8201, + "learning_rate": 6.824448529411765e-06, + "loss": 0.158, "step": 2188 }, { "epoch": 2.45, - "learning_rate": 6.796875e-05, - "loss": 2.8016, + "learning_rate": 6.810661764705882e-06, + "loss": 0.1489, "step": 2189 }, { "epoch": 2.46, - "learning_rate": 6.783088235294117e-05, - "loss": 2.9334, + "learning_rate": 6.796875e-06, + "loss": 0.2164, "step": 2190 }, { "epoch": 2.46, - "learning_rate": 6.769301470588235e-05, - "loss": 3.4224, + "learning_rate": 6.783088235294118e-06, + "loss": 0.1509, "step": 2191 }, { "epoch": 2.46, - "learning_rate": 6.755514705882352e-05, - "loss": 2.8197, + "learning_rate": 6.769301470588236e-06, + "loss": 0.1514, "step": 2192 }, { "epoch": 2.46, - "learning_rate": 6.74172794117647e-05, - "loss": 2.8041, + "learning_rate": 6.755514705882354e-06, + "loss": 0.1499, "step": 2193 }, { "epoch": 2.46, - "learning_rate": 6.727941176470587e-05, - "loss": 2.8045, + "learning_rate": 6.741727941176471e-06, + "loss": 0.1618, "step": 2194 }, { "epoch": 2.46, - "learning_rate": 6.714154411764706e-05, - "loss": 2.8109, + "learning_rate": 6.7279411764705885e-06, + "loss": 0.1693, "step": 2195 }, { "epoch": 2.46, - "learning_rate": 6.700367647058823e-05, - "loss": 2.849, + "learning_rate": 6.714154411764706e-06, + "loss": 0.1549, "step": 2196 }, { "epoch": 2.46, - "learning_rate": 6.686580882352941e-05, - "loss": 4.2739, + "learning_rate": 6.700367647058824e-06, + "loss": 0.214, "step": 2197 }, { "epoch": 2.46, - "learning_rate": 6.672794117647058e-05, - "loss": 2.8044, + "learning_rate": 6.686580882352942e-06, + "loss": 0.1456, "step": 2198 }, { "epoch": 2.47, - "learning_rate": 6.659007352941175e-05, - "loss": 2.8299, + "learning_rate": 6.672794117647058e-06, + "loss": 0.1743, "step": 2199 }, { "epoch": 2.47, - "learning_rate": 6.645220588235293e-05, - "loss": 2.8116, + "learning_rate": 6.659007352941176e-06, + "loss": 0.1467, "step": 2200 }, { "epoch": 2.47, - "eval_loss": 3.960448980331421, - "eval_runtime": 48.1529, - "eval_samples_per_second": 56.134, - "eval_steps_per_second": 0.893, - "eval_wer": 1.0, + "eval_loss": 0.2048598676919937, + "eval_runtime": 47.2627, + "eval_samples_per_second": 57.191, + "eval_steps_per_second": 0.91, + "eval_wer": 0.17541634498731665, "step": 2200 }, { "epoch": 2.47, - "learning_rate": 6.63143382352941e-05, - "loss": 2.8381, + "learning_rate": 6.645220588235294e-06, + "loss": 0.1516, "step": 2201 }, { "epoch": 2.47, - "learning_rate": 6.617647058823529e-05, - "loss": 2.9052, + "learning_rate": 6.631433823529412e-06, + "loss": 0.1602, "step": 2202 }, { "epoch": 2.47, - "learning_rate": 6.603860294117646e-05, - "loss": 4.627, + "learning_rate": 6.61764705882353e-06, + "loss": 0.1514, "step": 2203 }, { "epoch": 2.47, - "learning_rate": 6.590073529411764e-05, - "loss": 2.8091, + "learning_rate": 6.603860294117647e-06, + "loss": 0.1641, "step": 2204 }, { "epoch": 2.47, - "learning_rate": 6.576286764705881e-05, - "loss": 2.8155, + "learning_rate": 6.5900735294117646e-06, + "loss": 0.1565, "step": 2205 }, { "epoch": 2.47, - "learning_rate": 6.5625e-05, - "loss": 2.8122, + "learning_rate": 6.576286764705882e-06, + "loss": 0.1612, "step": 2206 }, { "epoch": 2.47, - "learning_rate": 6.548713235294117e-05, - "loss": 2.8087, + "learning_rate": 6.5625e-06, + "loss": 0.1359, "step": 2207 }, { "epoch": 2.48, - "learning_rate": 6.534926470588235e-05, - "loss": 2.8226, + "learning_rate": 6.548713235294118e-06, + "loss": 0.1781, "step": 2208 }, { "epoch": 2.48, - "learning_rate": 6.521139705882352e-05, - "loss": 3.1379, + "learning_rate": 6.534926470588235e-06, + "loss": 0.1826, "step": 2209 }, { "epoch": 2.48, - "learning_rate": 6.50735294117647e-05, - "loss": 2.821, + "learning_rate": 6.521139705882353e-06, + "loss": 0.162, "step": 2210 }, { "epoch": 2.48, - "learning_rate": 6.493566176470587e-05, - "loss": 2.7921, + "learning_rate": 6.507352941176471e-06, + "loss": 0.1352, "step": 2211 }, { "epoch": 2.48, - "learning_rate": 6.479779411764706e-05, - "loss": 2.8071, + "learning_rate": 6.493566176470589e-06, + "loss": 0.1462, "step": 2212 }, { "epoch": 2.48, - "learning_rate": 6.465992647058823e-05, - "loss": 2.8215, + "learning_rate": 6.479779411764707e-06, + "loss": 0.1674, "step": 2213 }, { "epoch": 2.48, - "learning_rate": 6.452205882352941e-05, - "loss": 2.8486, + "learning_rate": 6.4659926470588244e-06, + "loss": 0.1638, "step": 2214 }, { "epoch": 2.48, - "learning_rate": 6.438419117647058e-05, - "loss": 2.9086, + "learning_rate": 6.4522058823529415e-06, + "loss": 0.184, "step": 2215 }, { "epoch": 2.48, - "learning_rate": 6.424632352941175e-05, - "loss": 3.4843, + "learning_rate": 6.4384191176470585e-06, + "loss": 0.2116, "step": 2216 }, { "epoch": 2.49, - "learning_rate": 6.410845588235294e-05, - "loss": 2.8187, + "learning_rate": 6.424632352941176e-06, + "loss": 0.1559, "step": 2217 }, { "epoch": 2.49, - "learning_rate": 6.39705882352941e-05, - "loss": 2.8087, + "learning_rate": 6.410845588235294e-06, + "loss": 0.1713, "step": 2218 }, { "epoch": 2.49, - "learning_rate": 6.383272058823529e-05, - "loss": 2.8084, + "learning_rate": 6.397058823529412e-06, + "loss": 0.1417, "step": 2219 }, { "epoch": 2.49, - "learning_rate": 6.369485294117646e-05, - "loss": 2.8245, + "learning_rate": 6.383272058823529e-06, + "loss": 0.1685, "step": 2220 }, { "epoch": 2.49, - "learning_rate": 6.355698529411764e-05, - "loss": 2.8604, + "learning_rate": 6.369485294117647e-06, + "loss": 0.1745, "step": 2221 }, { "epoch": 2.49, - "learning_rate": 6.341911764705881e-05, - "loss": 3.8816, + "learning_rate": 6.355698529411765e-06, + "loss": 0.1922, "step": 2222 }, { "epoch": 2.49, - "learning_rate": 6.328125e-05, - "loss": 2.81, + "learning_rate": 6.341911764705883e-06, + "loss": 0.1792, "step": 2223 }, { "epoch": 2.49, - "learning_rate": 6.314338235294117e-05, - "loss": 2.8204, + "learning_rate": 6.3281250000000005e-06, + "loss": 0.1534, "step": 2224 }, { "epoch": 2.49, - "learning_rate": 6.300551470588235e-05, - "loss": 2.8125, + "learning_rate": 6.3143382352941175e-06, + "loss": 0.1449, "step": 2225 }, { "epoch": 2.5, - "learning_rate": 6.286764705882352e-05, - "loss": 2.8286, + "learning_rate": 6.300551470588235e-06, + "loss": 0.1539, "step": 2226 }, { "epoch": 2.5, - "learning_rate": 6.27297794117647e-05, - "loss": 2.7982, + "learning_rate": 6.286764705882353e-06, + "loss": 0.13, "step": 2227 }, { "epoch": 2.5, - "learning_rate": 6.259191176470587e-05, - "loss": 3.9117, + "learning_rate": 6.272977941176471e-06, + "loss": 0.1835, "step": 2228 }, { "epoch": 2.5, - "learning_rate": 6.245404411764706e-05, - "loss": 2.8135, + "learning_rate": 6.259191176470589e-06, + "loss": 0.1427, "step": 2229 }, { "epoch": 2.5, - "learning_rate": 6.231617647058823e-05, - "loss": 2.796, + "learning_rate": 6.245404411764706e-06, + "loss": 0.1437, "step": 2230 }, { "epoch": 2.5, - "learning_rate": 6.217830882352941e-05, - "loss": 2.8041, + "learning_rate": 6.231617647058824e-06, + "loss": 0.155, "step": 2231 }, { "epoch": 2.5, - "learning_rate": 6.204044117647058e-05, - "loss": 2.8317, + "learning_rate": 6.217830882352942e-06, + "loss": 0.1557, "step": 2232 }, { "epoch": 2.5, - "learning_rate": 6.190257352941175e-05, - "loss": 2.8177, + "learning_rate": 6.204044117647059e-06, + "loss": 0.1749, "step": 2233 }, { "epoch": 2.5, - "learning_rate": 6.176470588235294e-05, - "loss": 3.2776, + "learning_rate": 6.1902573529411766e-06, + "loss": 0.1751, "step": 2234 }, { "epoch": 2.51, - "learning_rate": 6.16268382352941e-05, - "loss": 2.8119, + "learning_rate": 6.176470588235294e-06, + "loss": 0.1514, "step": 2235 }, { "epoch": 2.51, - "learning_rate": 6.148897058823529e-05, - "loss": 2.8011, + "learning_rate": 6.1626838235294114e-06, + "loss": 0.1435, "step": 2236 }, { "epoch": 2.51, - "learning_rate": 6.135110294117646e-05, - "loss": 2.8039, + "learning_rate": 6.148897058823529e-06, + "loss": 0.1676, "step": 2237 }, { "epoch": 2.51, - "learning_rate": 6.121323529411764e-05, - "loss": 2.8131, + "learning_rate": 6.135110294117647e-06, + "loss": 0.1351, "step": 2238 }, { "epoch": 2.51, - "learning_rate": 6.107536764705881e-05, - "loss": 2.811, + "learning_rate": 6.121323529411765e-06, + "loss": 0.1633, "step": 2239 }, { "epoch": 2.51, - "learning_rate": 6.09375e-05, - "loss": 3.0463, + "learning_rate": 6.107536764705883e-06, + "loss": 0.1601, "step": 2240 }, { "epoch": 2.51, - "learning_rate": 6.0799632352941174e-05, - "loss": 3.6621, + "learning_rate": 6.09375e-06, + "loss": 0.207, "step": 2241 }, { "epoch": 2.51, - "learning_rate": 6.066176470588235e-05, - "loss": 2.8049, + "learning_rate": 6.079963235294118e-06, + "loss": 0.1534, "step": 2242 }, { "epoch": 2.51, - "learning_rate": 6.052389705882353e-05, - "loss": 2.8103, + "learning_rate": 6.066176470588236e-06, + "loss": 0.1787, "step": 2243 }, { "epoch": 2.52, - "learning_rate": 6.0386029411764705e-05, - "loss": 2.8179, + "learning_rate": 6.0523897058823535e-06, + "loss": 0.1592, "step": 2244 }, { "epoch": 2.52, - "learning_rate": 6.024816176470588e-05, - "loss": 2.8051, + "learning_rate": 6.038602941176471e-06, + "loss": 0.1807, "step": 2245 }, { "epoch": 2.52, - "learning_rate": 6.011029411764706e-05, - "loss": 2.8514, + "learning_rate": 6.024816176470588e-06, + "loss": 0.142, "step": 2246 }, { "epoch": 2.52, - "learning_rate": 5.9972426470588236e-05, - "loss": 3.5701, + "learning_rate": 6.011029411764706e-06, + "loss": 0.2165, "step": 2247 }, { "epoch": 2.52, - "learning_rate": 5.983455882352941e-05, - "loss": 2.8073, + "learning_rate": 5.997242647058824e-06, + "loss": 0.1596, "step": 2248 }, { "epoch": 2.52, - "learning_rate": 5.9696691176470576e-05, - "loss": 2.8204, + "learning_rate": 5.983455882352942e-06, + "loss": 0.1225, "step": 2249 }, { "epoch": 2.52, - "learning_rate": 5.955882352941175e-05, - "loss": 2.7926, + "learning_rate": 5.969669117647059e-06, + "loss": 0.1658, "step": 2250 }, { "epoch": 2.52, - "learning_rate": 5.9420955882352936e-05, - "loss": 2.8066, + "learning_rate": 5.955882352941176e-06, + "loss": 0.1113, "step": 2251 }, { "epoch": 2.52, - "learning_rate": 5.928308823529411e-05, - "loss": 2.8375, + "learning_rate": 5.942095588235294e-06, + "loss": 0.1657, "step": 2252 }, { "epoch": 2.53, - "learning_rate": 5.914522058823529e-05, - "loss": 4.1591, + "learning_rate": 5.928308823529412e-06, + "loss": 0.1998, "step": 2253 }, { "epoch": 2.53, - "learning_rate": 5.900735294117647e-05, - "loss": 2.8098, + "learning_rate": 5.9145220588235295e-06, + "loss": 0.1726, "step": 2254 }, { "epoch": 2.53, - "learning_rate": 5.8869485294117644e-05, - "loss": 2.824, + "learning_rate": 5.900735294117647e-06, + "loss": 0.1925, "step": 2255 }, { "epoch": 2.53, - "learning_rate": 5.873161764705882e-05, - "loss": 2.8197, + "learning_rate": 5.886948529411764e-06, + "loss": 0.1619, "step": 2256 }, { "epoch": 2.53, - "learning_rate": 5.859375e-05, - "loss": 2.8156, + "learning_rate": 5.873161764705882e-06, + "loss": 0.1671, "step": 2257 }, { "epoch": 2.53, - "learning_rate": 5.8455882352941175e-05, - "loss": 2.8674, + "learning_rate": 5.859375e-06, + "loss": 0.1869, "step": 2258 }, { "epoch": 2.53, - "learning_rate": 5.831801470588235e-05, - "loss": 2.9537, + "learning_rate": 5.845588235294118e-06, + "loss": 0.1696, "step": 2259 }, { "epoch": 2.53, - "learning_rate": 5.818014705882353e-05, - "loss": 2.8107, + "learning_rate": 5.831801470588236e-06, + "loss": 0.1394, "step": 2260 }, { "epoch": 2.53, - "learning_rate": 5.8042279411764705e-05, - "loss": 2.7926, + "learning_rate": 5.818014705882353e-06, + "loss": 0.1718, "step": 2261 }, { "epoch": 2.54, - "learning_rate": 5.790441176470588e-05, - "loss": 2.8112, + "learning_rate": 5.804227941176471e-06, + "loss": 0.1332, "step": 2262 }, { "epoch": 2.54, - "learning_rate": 5.776654411764706e-05, - "loss": 2.8158, + "learning_rate": 5.7904411764705886e-06, + "loss": 0.1735, "step": 2263 }, { "epoch": 2.54, - "learning_rate": 5.7628676470588236e-05, - "loss": 2.8101, + "learning_rate": 5.7766544117647064e-06, + "loss": 0.1633, "step": 2264 }, { "epoch": 2.54, - "learning_rate": 5.749080882352941e-05, - "loss": 2.9359, + "learning_rate": 5.762867647058824e-06, + "loss": 0.1578, "step": 2265 }, { "epoch": 2.54, - "learning_rate": 5.7352941176470576e-05, - "loss": 3.5398, + "learning_rate": 5.749080882352941e-06, + "loss": 0.1465, "step": 2266 }, { "epoch": 2.54, - "learning_rate": 5.721507352941175e-05, - "loss": 2.8087, + "learning_rate": 5.735294117647058e-06, + "loss": 0.1693, "step": 2267 }, { "epoch": 2.54, - "learning_rate": 5.707720588235293e-05, - "loss": 2.8001, + "learning_rate": 5.721507352941176e-06, + "loss": 0.122, "step": 2268 }, { "epoch": 2.54, - "learning_rate": 5.693933823529411e-05, - "loss": 2.8169, + "learning_rate": 5.707720588235294e-06, + "loss": 0.1433, "step": 2269 }, { "epoch": 2.54, - "learning_rate": 5.6801470588235284e-05, - "loss": 2.8241, + "learning_rate": 5.693933823529412e-06, + "loss": 0.1998, "step": 2270 }, { "epoch": 2.55, - "learning_rate": 5.666360294117646e-05, - "loss": 2.8812, + "learning_rate": 5.68014705882353e-06, + "loss": 0.1486, "step": 2271 }, { "epoch": 2.55, - "learning_rate": 5.652573529411764e-05, - "loss": 3.9079, + "learning_rate": 5.666360294117647e-06, + "loss": 0.1566, "step": 2272 }, { "epoch": 2.55, - "learning_rate": 5.6387867647058815e-05, - "loss": 2.8096, + "learning_rate": 5.652573529411765e-06, + "loss": 0.1315, "step": 2273 }, { "epoch": 2.55, - "learning_rate": 5.625e-05, - "loss": 2.8268, + "learning_rate": 5.6387867647058825e-06, + "loss": 0.1621, "step": 2274 }, { "epoch": 2.55, - "learning_rate": 5.6112132352941175e-05, - "loss": 2.8053, + "learning_rate": 5.625e-06, + "loss": 0.1522, "step": 2275 }, { "epoch": 2.55, - "learning_rate": 5.597426470588235e-05, - "loss": 2.814, + "learning_rate": 5.611213235294118e-06, + "loss": 0.1491, "step": 2276 }, { "epoch": 2.55, - "learning_rate": 5.583639705882353e-05, - "loss": 2.8583, + "learning_rate": 5.597426470588235e-06, + "loss": 0.1777, "step": 2277 }, { "epoch": 2.55, - "learning_rate": 5.5698529411764706e-05, - "loss": 4.1541, + "learning_rate": 5.583639705882353e-06, + "loss": 0.1851, "step": 2278 }, { "epoch": 2.55, - "learning_rate": 5.556066176470588e-05, - "loss": 2.817, + "learning_rate": 5.569852941176471e-06, + "loss": 0.1788, "step": 2279 }, { "epoch": 2.56, - "learning_rate": 5.542279411764706e-05, - "loss": 2.8093, + "learning_rate": 5.556066176470589e-06, + "loss": 0.1642, "step": 2280 }, { "epoch": 2.56, - "learning_rate": 5.528492647058824e-05, - "loss": 2.817, + "learning_rate": 5.542279411764707e-06, + "loss": 0.1275, "step": 2281 }, { "epoch": 2.56, - "learning_rate": 5.5147058823529414e-05, - "loss": 2.8174, + "learning_rate": 5.528492647058824e-06, + "loss": 0.187, "step": 2282 }, { "epoch": 2.56, - "learning_rate": 5.500919117647058e-05, - "loss": 2.8325, + "learning_rate": 5.5147058823529415e-06, + "loss": 0.1796, "step": 2283 }, { "epoch": 2.56, - "learning_rate": 5.4871323529411754e-05, - "loss": 3.2584, + "learning_rate": 5.5009191176470586e-06, + "loss": 0.1569, "step": 2284 }, { "epoch": 2.56, - "learning_rate": 5.473345588235293e-05, - "loss": 2.8029, + "learning_rate": 5.487132352941176e-06, + "loss": 0.1411, "step": 2285 }, { "epoch": 2.56, - "learning_rate": 5.459558823529411e-05, - "loss": 2.8091, + "learning_rate": 5.473345588235294e-06, + "loss": 0.1365, "step": 2286 }, { "epoch": 2.56, - "learning_rate": 5.4457720588235285e-05, - "loss": 2.8175, + "learning_rate": 5.459558823529411e-06, + "loss": 0.1562, "step": 2287 }, { "epoch": 2.57, - "learning_rate": 5.431985294117646e-05, - "loss": 2.81, + "learning_rate": 5.445772058823529e-06, + "loss": 0.1463, "step": 2288 }, { "epoch": 2.57, - "learning_rate": 5.418198529411764e-05, - "loss": 2.8191, + "learning_rate": 5.431985294117647e-06, + "loss": 0.1554, "step": 2289 }, { "epoch": 2.57, - "learning_rate": 5.4044117647058815e-05, - "loss": 2.9835, + "learning_rate": 5.418198529411765e-06, + "loss": 0.1751, "step": 2290 }, { "epoch": 2.57, - "learning_rate": 5.390624999999999e-05, - "loss": 3.8741, + "learning_rate": 5.404411764705883e-06, + "loss": 0.2315, "step": 2291 }, { "epoch": 2.57, - "learning_rate": 5.376838235294117e-05, - "loss": 2.7994, + "learning_rate": 5.390625e-06, + "loss": 0.1742, "step": 2292 }, { "epoch": 2.57, - "learning_rate": 5.3630514705882346e-05, - "loss": 2.8046, + "learning_rate": 5.376838235294118e-06, + "loss": 0.1689, "step": 2293 }, { "epoch": 2.57, - "learning_rate": 5.349264705882352e-05, - "loss": 2.8162, + "learning_rate": 5.3630514705882355e-06, + "loss": 0.1739, "step": 2294 }, { "epoch": 2.57, - "learning_rate": 5.33547794117647e-05, - "loss": 2.8089, + "learning_rate": 5.349264705882353e-06, + "loss": 0.17, "step": 2295 }, { "epoch": 2.57, - "learning_rate": 5.321691176470588e-05, - "loss": 2.8674, + "learning_rate": 5.335477941176471e-06, + "loss": 0.1652, "step": 2296 }, { "epoch": 2.58, - "learning_rate": 5.3079044117647054e-05, - "loss": 3.9751, + "learning_rate": 5.321691176470588e-06, + "loss": 0.1362, "step": 2297 }, { "epoch": 2.58, - "learning_rate": 5.294117647058824e-05, - "loss": 2.8251, + "learning_rate": 5.307904411764706e-06, + "loss": 0.162, "step": 2298 }, { "epoch": 2.58, - "learning_rate": 5.2803308823529414e-05, - "loss": 2.8197, + "learning_rate": 5.294117647058824e-06, + "loss": 0.1399, "step": 2299 }, { "epoch": 2.58, - "learning_rate": 5.266544117647058e-05, - "loss": 2.81, + "learning_rate": 5.280330882352942e-06, + "loss": 0.1702, "step": 2300 }, { "epoch": 2.58, - "eval_loss": 3.9278817176818848, - "eval_runtime": 48.2911, - "eval_samples_per_second": 55.973, - "eval_steps_per_second": 0.89, - "eval_wer": 1.0, + "eval_loss": 0.2013147622346878, + "eval_runtime": 47.2411, + "eval_samples_per_second": 57.217, + "eval_steps_per_second": 0.91, + "eval_wer": 0.17216278813278924, "step": 2300 }, { "epoch": 2.58, - "learning_rate": 5.2527573529411755e-05, - "loss": 2.8052, + "learning_rate": 5.266544117647059e-06, + "loss": 0.1411, "step": 2301 }, { "epoch": 2.58, - "learning_rate": 5.238970588235293e-05, - "loss": 2.8915, + "learning_rate": 5.252757352941177e-06, + "loss": 0.1585, "step": 2302 }, { "epoch": 2.58, - "learning_rate": 5.225183823529411e-05, - "loss": 5.3169, + "learning_rate": 5.238970588235294e-06, + "loss": 0.3671, "step": 2303 }, { "epoch": 2.58, - "learning_rate": 5.2113970588235285e-05, - "loss": 2.8156, + "learning_rate": 5.2251838235294115e-06, + "loss": 0.1509, "step": 2304 }, { "epoch": 2.58, - "learning_rate": 5.197610294117646e-05, - "loss": 2.8082, + "learning_rate": 5.211397058823529e-06, + "loss": 0.1703, "step": 2305 }, { "epoch": 2.59, - "learning_rate": 5.183823529411764e-05, - "loss": 2.8234, + "learning_rate": 5.197610294117647e-06, + "loss": 0.1787, "step": 2306 }, { "epoch": 2.59, - "learning_rate": 5.1700367647058816e-05, - "loss": 2.8197, + "learning_rate": 5.183823529411765e-06, + "loss": 0.1679, "step": 2307 }, { "epoch": 2.59, - "learning_rate": 5.156249999999999e-05, - "loss": 2.8448, + "learning_rate": 5.170036764705882e-06, + "loss": 0.1616, "step": 2308 }, { "epoch": 2.59, - "learning_rate": 5.142463235294117e-05, - "loss": 3.1339, + "learning_rate": 5.15625e-06, + "loss": 0.1856, "step": 2309 }, { "epoch": 2.59, - "learning_rate": 5.128676470588235e-05, - "loss": 2.8052, + "learning_rate": 5.142463235294118e-06, + "loss": 0.1483, "step": 2310 }, { "epoch": 2.59, - "learning_rate": 5.1148897058823524e-05, - "loss": 2.8143, + "learning_rate": 5.128676470588236e-06, + "loss": 0.1144, "step": 2311 }, { "epoch": 2.59, - "learning_rate": 5.10110294117647e-05, - "loss": 2.8032, + "learning_rate": 5.1148897058823536e-06, + "loss": 0.1456, "step": 2312 }, { "epoch": 2.59, - "learning_rate": 5.087316176470588e-05, - "loss": 2.8122, + "learning_rate": 5.1011029411764706e-06, + "loss": 0.1873, "step": 2313 }, { "epoch": 2.59, - "learning_rate": 5.0735294117647054e-05, - "loss": 2.8278, + "learning_rate": 5.0873161764705884e-06, + "loss": 0.1381, "step": 2314 }, { "epoch": 2.6, - "learning_rate": 5.059742647058823e-05, - "loss": 3.0408, + "learning_rate": 5.073529411764706e-06, + "loss": 0.1466, "step": 2315 }, { "epoch": 2.6, - "learning_rate": 5.045955882352941e-05, - "loss": 3.6024, + "learning_rate": 5.059742647058824e-06, + "loss": 0.1658, "step": 2316 }, { "epoch": 2.6, - "learning_rate": 5.032169117647058e-05, - "loss": 2.8126, + "learning_rate": 5.045955882352942e-06, + "loss": 0.1792, "step": 2317 }, { "epoch": 2.6, - "learning_rate": 5.0183823529411755e-05, - "loss": 2.8224, + "learning_rate": 5.032169117647058e-06, + "loss": 0.1739, "step": 2318 }, { "epoch": 2.6, - "learning_rate": 5.004595588235293e-05, - "loss": 2.8246, + "learning_rate": 5.018382352941176e-06, + "loss": 0.1621, "step": 2319 }, { "epoch": 2.6, - "learning_rate": 4.990808823529411e-05, - "loss": 2.829, + "learning_rate": 5.004595588235294e-06, + "loss": 0.1554, "step": 2320 }, { "epoch": 2.6, - "learning_rate": 4.9770220588235286e-05, - "loss": 2.85, + "learning_rate": 4.990808823529412e-06, + "loss": 0.179, "step": 2321 }, { "epoch": 2.6, - "learning_rate": 4.963235294117646e-05, - "loss": 4.1999, + "learning_rate": 4.97702205882353e-06, + "loss": 0.2191, "step": 2322 }, { "epoch": 2.6, - "learning_rate": 4.949448529411764e-05, - "loss": 2.8144, + "learning_rate": 4.963235294117647e-06, + "loss": 0.1505, "step": 2323 }, { "epoch": 2.61, - "learning_rate": 4.935661764705882e-05, - "loss": 2.8073, + "learning_rate": 4.9494485294117645e-06, + "loss": 0.1267, "step": 2324 }, { "epoch": 2.61, - "learning_rate": 4.9218749999999994e-05, - "loss": 2.8102, + "learning_rate": 4.935661764705882e-06, + "loss": 0.1292, "step": 2325 }, { "epoch": 2.61, - "learning_rate": 4.908088235294117e-05, - "loss": 2.8194, + "learning_rate": 4.921875e-06, + "loss": 0.1568, "step": 2326 }, { "epoch": 2.61, - "learning_rate": 4.894301470588235e-05, - "loss": 2.8092, + "learning_rate": 4.908088235294118e-06, + "loss": 0.1506, "step": 2327 }, { "epoch": 2.61, - "learning_rate": 4.8805147058823524e-05, - "loss": 4.0504, + "learning_rate": 4.894301470588236e-06, + "loss": 0.1597, "step": 2328 }, { "epoch": 2.61, - "learning_rate": 4.86672794117647e-05, - "loss": 2.8162, + "learning_rate": 4.880514705882353e-06, + "loss": 0.1877, "step": 2329 }, { "epoch": 2.61, - "learning_rate": 4.852941176470588e-05, - "loss": 2.8051, + "learning_rate": 4.866727941176471e-06, + "loss": 0.1594, "step": 2330 }, { "epoch": 2.61, - "learning_rate": 4.8391544117647055e-05, - "loss": 2.8222, + "learning_rate": 4.852941176470589e-06, + "loss": 0.1494, "step": 2331 }, { "epoch": 2.61, - "learning_rate": 4.825367647058823e-05, - "loss": 2.8186, + "learning_rate": 4.8391544117647065e-06, + "loss": 0.1151, "step": 2332 }, { "epoch": 2.62, - "learning_rate": 4.811580882352941e-05, - "loss": 2.8218, + "learning_rate": 4.825367647058824e-06, + "loss": 0.1458, "step": 2333 }, { "epoch": 2.62, - "learning_rate": 4.797794117647058e-05, - "loss": 3.2412, + "learning_rate": 4.811580882352941e-06, + "loss": 0.2275, "step": 2334 }, { "epoch": 2.62, - "learning_rate": 4.7840073529411756e-05, - "loss": 2.7931, + "learning_rate": 4.797794117647058e-06, + "loss": 0.1295, "step": 2335 }, { "epoch": 2.62, - "learning_rate": 4.770220588235293e-05, - "loss": 2.799, + "learning_rate": 4.784007352941176e-06, + "loss": 0.1552, "step": 2336 }, { "epoch": 2.62, - "learning_rate": 4.756433823529411e-05, - "loss": 2.7955, + "learning_rate": 4.770220588235294e-06, + "loss": 0.1386, "step": 2337 }, { "epoch": 2.62, - "learning_rate": 4.7426470588235287e-05, - "loss": 2.8114, + "learning_rate": 4.756433823529412e-06, + "loss": 0.1484, "step": 2338 }, { "epoch": 2.62, - "learning_rate": 4.7288602941176463e-05, - "loss": 2.8334, + "learning_rate": 4.742647058823529e-06, + "loss": 0.1494, "step": 2339 }, { "epoch": 2.62, - "learning_rate": 4.715073529411764e-05, - "loss": 3.1722, + "learning_rate": 4.728860294117647e-06, + "loss": 0.2636, "step": 2340 }, { "epoch": 2.62, - "learning_rate": 4.701286764705882e-05, - "loss": 3.953, + "learning_rate": 4.715073529411765e-06, + "loss": 0.1846, "step": 2341 }, { "epoch": 2.63, - "learning_rate": 4.6874999999999994e-05, - "loss": 2.7997, + "learning_rate": 4.7012867647058826e-06, + "loss": 0.1898, "step": 2342 }, { "epoch": 2.63, - "learning_rate": 4.673713235294117e-05, - "loss": 2.8088, + "learning_rate": 4.6875000000000004e-06, + "loss": 0.1621, "step": 2343 }, { "epoch": 2.63, - "learning_rate": 4.659926470588235e-05, - "loss": 2.8061, + "learning_rate": 4.6737132352941174e-06, + "loss": 0.1668, "step": 2344 }, { "epoch": 2.63, - "learning_rate": 4.6461397058823525e-05, - "loss": 2.8417, + "learning_rate": 4.659926470588235e-06, + "loss": 0.1822, "step": 2345 }, { "epoch": 2.63, - "learning_rate": 4.63235294117647e-05, - "loss": 2.9424, + "learning_rate": 4.646139705882353e-06, + "loss": 0.1864, "step": 2346 }, { "epoch": 2.63, - "learning_rate": 4.618566176470588e-05, - "loss": 4.6347, + "learning_rate": 4.632352941176471e-06, + "loss": 0.1708, "step": 2347 }, { "epoch": 2.63, - "learning_rate": 4.6047794117647056e-05, - "loss": 2.816, + "learning_rate": 4.618566176470589e-06, + "loss": 0.1404, "step": 2348 }, { "epoch": 2.63, - "learning_rate": 4.590992647058823e-05, - "loss": 2.8034, + "learning_rate": 4.604779411764706e-06, + "loss": 0.1476, "step": 2349 }, { "epoch": 2.63, - "learning_rate": 4.577205882352941e-05, - "loss": 2.819, + "learning_rate": 4.590992647058824e-06, + "loss": 0.152, "step": 2350 }, { "epoch": 2.64, - "learning_rate": 4.563419117647058e-05, - "loss": 2.8228, + "learning_rate": 4.577205882352942e-06, + "loss": 0.1671, "step": 2351 }, { "epoch": 2.64, - "learning_rate": 4.5496323529411756e-05, - "loss": 2.8587, + "learning_rate": 4.563419117647059e-06, + "loss": 0.1547, "step": 2352 }, { "epoch": 2.64, - "learning_rate": 4.5358455882352933e-05, - "loss": 4.4024, + "learning_rate": 4.5496323529411765e-06, + "loss": 0.2104, "step": 2353 }, { "epoch": 2.64, - "learning_rate": 4.522058823529411e-05, - "loss": 2.814, + "learning_rate": 4.535845588235294e-06, + "loss": 0.1557, "step": 2354 }, { "epoch": 2.64, - "learning_rate": 4.508272058823529e-05, - "loss": 2.8219, + "learning_rate": 4.522058823529411e-06, + "loss": 0.1618, "step": 2355 }, { "epoch": 2.64, - "learning_rate": 4.4944852941176464e-05, - "loss": 2.8092, + "learning_rate": 4.508272058823529e-06, + "loss": 0.185, "step": 2356 }, { "epoch": 2.64, - "learning_rate": 4.480698529411764e-05, - "loss": 2.8044, + "learning_rate": 4.494485294117647e-06, + "loss": 0.1567, "step": 2357 }, { "epoch": 2.64, - "learning_rate": 4.466911764705882e-05, - "loss": 2.8204, + "learning_rate": 4.480698529411765e-06, + "loss": 0.1448, "step": 2358 }, { "epoch": 2.64, - "learning_rate": 4.4531249999999995e-05, - "loss": 3.1125, + "learning_rate": 4.466911764705883e-06, + "loss": 0.1766, "step": 2359 }, { "epoch": 2.65, - "learning_rate": 4.439338235294117e-05, - "loss": 2.8245, + "learning_rate": 4.453125e-06, + "loss": 0.1462, "step": 2360 }, { "epoch": 2.65, - "learning_rate": 4.425551470588235e-05, - "loss": 2.8161, + "learning_rate": 4.439338235294118e-06, + "loss": 0.1715, "step": 2361 }, { "epoch": 2.65, - "learning_rate": 4.4117647058823526e-05, - "loss": 2.8153, + "learning_rate": 4.4255514705882355e-06, + "loss": 0.1559, "step": 2362 }, { "epoch": 2.65, - "learning_rate": 4.39797794117647e-05, - "loss": 2.8048, + "learning_rate": 4.411764705882353e-06, + "loss": 0.177, "step": 2363 }, { "epoch": 2.65, - "learning_rate": 4.384191176470588e-05, - "loss": 2.8201, + "learning_rate": 4.397977941176471e-06, + "loss": 0.1341, "step": 2364 }, { "epoch": 2.65, - "learning_rate": 4.3704044117647056e-05, - "loss": 2.9787, + "learning_rate": 4.384191176470588e-06, + "loss": 0.2233, "step": 2365 }, { "epoch": 2.65, - "learning_rate": 4.356617647058823e-05, - "loss": 3.3826, + "learning_rate": 4.370404411764706e-06, + "loss": 0.1718, "step": 2366 }, { "epoch": 2.65, - "learning_rate": 4.342830882352941e-05, - "loss": 2.8045, + "learning_rate": 4.356617647058824e-06, + "loss": 0.1446, "step": 2367 }, { "epoch": 2.65, - "learning_rate": 4.329044117647058e-05, - "loss": 2.8133, + "learning_rate": 4.342830882352942e-06, + "loss": 0.1631, "step": 2368 }, { "epoch": 2.66, - "learning_rate": 4.315257352941176e-05, - "loss": 2.8001, + "learning_rate": 4.329044117647059e-06, + "loss": 0.1602, "step": 2369 }, { "epoch": 2.66, - "learning_rate": 4.3014705882352934e-05, - "loss": 2.8172, + "learning_rate": 4.315257352941176e-06, + "loss": 0.1586, "step": 2370 }, { "epoch": 2.66, - "learning_rate": 4.287683823529411e-05, - "loss": 2.8552, + "learning_rate": 4.301470588235294e-06, + "loss": 0.1558, "step": 2371 }, { "epoch": 2.66, - "learning_rate": 4.273897058823529e-05, - "loss": 3.8376, + "learning_rate": 4.287683823529412e-06, + "loss": 0.14, "step": 2372 }, { "epoch": 2.66, - "learning_rate": 4.2601102941176465e-05, - "loss": 2.808, + "learning_rate": 4.2738970588235295e-06, + "loss": 0.1389, "step": 2373 }, { "epoch": 2.66, - "learning_rate": 4.246323529411764e-05, - "loss": 2.8102, + "learning_rate": 4.260110294117647e-06, + "loss": 0.1399, "step": 2374 }, { "epoch": 2.66, - "learning_rate": 4.232536764705882e-05, - "loss": 2.8109, + "learning_rate": 4.246323529411764e-06, + "loss": 0.137, "step": 2375 }, { "epoch": 2.66, - "learning_rate": 4.2187499999999995e-05, - "loss": 2.812, + "learning_rate": 4.232536764705882e-06, + "loss": 0.1491, "step": 2376 }, { "epoch": 2.66, - "learning_rate": 4.204963235294117e-05, - "loss": 2.8218, + "learning_rate": 4.21875e-06, + "loss": 0.1442, "step": 2377 }, { "epoch": 2.67, - "learning_rate": 4.191176470588235e-05, - "loss": 4.1445, + "learning_rate": 4.204963235294118e-06, + "loss": 0.1838, "step": 2378 }, { "epoch": 2.67, - "learning_rate": 4.1773897058823526e-05, - "loss": 2.8221, + "learning_rate": 4.191176470588236e-06, + "loss": 0.1506, "step": 2379 }, { "epoch": 2.67, - "learning_rate": 4.16360294117647e-05, - "loss": 2.7977, + "learning_rate": 4.177389705882353e-06, + "loss": 0.1341, "step": 2380 }, { "epoch": 2.67, - "learning_rate": 4.149816176470588e-05, - "loss": 2.8109, + "learning_rate": 4.163602941176471e-06, + "loss": 0.1574, "step": 2381 }, { "epoch": 2.67, - "learning_rate": 4.136029411764706e-05, - "loss": 2.8215, + "learning_rate": 4.1498161764705885e-06, + "loss": 0.1575, "step": 2382 }, { "epoch": 2.67, - "learning_rate": 4.1222426470588234e-05, - "loss": 2.8219, + "learning_rate": 4.136029411764706e-06, + "loss": 0.1818, "step": 2383 }, { "epoch": 2.67, - "learning_rate": 4.108455882352941e-05, - "loss": 3.3193, + "learning_rate": 4.122242647058824e-06, + "loss": 0.235, "step": 2384 }, { "epoch": 2.67, - "learning_rate": 4.094669117647058e-05, - "loss": 2.7994, + "learning_rate": 4.108455882352942e-06, + "loss": 0.1307, "step": 2385 }, { "epoch": 2.67, - "learning_rate": 4.080882352941176e-05, - "loss": 2.8118, + "learning_rate": 4.094669117647058e-06, + "loss": 0.1307, "step": 2386 }, { "epoch": 2.68, - "learning_rate": 4.0670955882352935e-05, - "loss": 2.801, + "learning_rate": 4.080882352941176e-06, + "loss": 0.1393, "step": 2387 }, { "epoch": 2.68, - "learning_rate": 4.053308823529411e-05, - "loss": 2.799, + "learning_rate": 4.067095588235294e-06, + "loss": 0.1521, "step": 2388 }, { "epoch": 2.68, - "learning_rate": 4.039522058823529e-05, - "loss": 2.8309, + "learning_rate": 4.053308823529412e-06, + "loss": 0.1588, "step": 2389 }, { "epoch": 2.68, - "learning_rate": 4.0257352941176465e-05, - "loss": 2.9482, + "learning_rate": 4.03952205882353e-06, + "loss": 0.1816, "step": 2390 }, { "epoch": 2.68, - "learning_rate": 4.011948529411764e-05, - "loss": 3.8104, + "learning_rate": 4.025735294117647e-06, + "loss": 0.2346, "step": 2391 }, { "epoch": 2.68, - "learning_rate": 3.998161764705882e-05, - "loss": 2.814, + "learning_rate": 4.0119485294117646e-06, + "loss": 0.1502, "step": 2392 }, { "epoch": 2.68, - "learning_rate": 3.9843749999999996e-05, - "loss": 2.8044, + "learning_rate": 3.9981617647058824e-06, + "loss": 0.1791, "step": 2393 }, { "epoch": 2.68, - "learning_rate": 3.970588235294117e-05, - "loss": 2.8209, + "learning_rate": 3.984375e-06, + "loss": 0.1587, "step": 2394 }, { "epoch": 2.68, - "learning_rate": 3.956801470588235e-05, - "loss": 2.8109, + "learning_rate": 3.970588235294118e-06, + "loss": 0.1482, "step": 2395 }, { "epoch": 2.69, - "learning_rate": 3.943014705882353e-05, - "loss": 3.0238, + "learning_rate": 3.956801470588235e-06, + "loss": 0.1477, "step": 2396 }, { "epoch": 2.69, - "learning_rate": 3.9292279411764704e-05, - "loss": 5.6123, + "learning_rate": 3.943014705882353e-06, + "loss": 0.2539, "step": 2397 }, { "epoch": 2.69, - "learning_rate": 3.915441176470588e-05, - "loss": 2.8211, + "learning_rate": 3.929227941176471e-06, + "loss": 0.147, "step": 2398 }, { "epoch": 2.69, - "learning_rate": 3.901654411764706e-05, - "loss": 2.8063, + "learning_rate": 3.915441176470589e-06, + "loss": 0.1726, "step": 2399 }, { "epoch": 2.69, - "learning_rate": 3.8878676470588234e-05, - "loss": 2.8051, + "learning_rate": 3.901654411764707e-06, + "loss": 0.177, "step": 2400 }, { "epoch": 2.69, - "eval_loss": 3.9737062454223633, - "eval_runtime": 47.8031, - "eval_samples_per_second": 56.544, - "eval_steps_per_second": 0.9, - "eval_wer": 1.0, + "eval_loss": 0.19929687678813934, + "eval_runtime": 47.305, + "eval_samples_per_second": 57.14, + "eval_steps_per_second": 0.909, + "eval_wer": 0.17012242196978053, "step": 2400 }, { "epoch": 2.69, - "learning_rate": 3.874080882352941e-05, - "loss": 2.8195, + "learning_rate": 3.887867647058824e-06, + "loss": 0.1375, "step": 2401 }, { "epoch": 2.69, - "learning_rate": 3.860294117647058e-05, - "loss": 2.8602, + "learning_rate": 3.8740808823529415e-06, + "loss": 0.1774, "step": 2402 }, { "epoch": 2.69, - "learning_rate": 3.846507352941176e-05, - "loss": 3.992, + "learning_rate": 3.8602941176470585e-06, + "loss": 0.1894, "step": 2403 }, { "epoch": 2.7, - "learning_rate": 3.8327205882352935e-05, - "loss": 2.8173, + "learning_rate": 3.846507352941176e-06, + "loss": 0.1304, "step": 2404 }, { "epoch": 2.7, - "learning_rate": 3.818933823529411e-05, - "loss": 2.7999, + "learning_rate": 3.832720588235294e-06, + "loss": 0.1608, "step": 2405 }, { "epoch": 2.7, - "learning_rate": 3.805147058823529e-05, - "loss": 2.8183, + "learning_rate": 3.818933823529411e-06, + "loss": 0.1396, "step": 2406 }, { "epoch": 2.7, - "learning_rate": 3.7913602941176466e-05, - "loss": 2.8093, + "learning_rate": 3.8051470588235295e-06, + "loss": 0.1467, "step": 2407 }, { "epoch": 2.7, - "learning_rate": 3.777573529411764e-05, - "loss": 2.8361, + "learning_rate": 3.791360294117647e-06, + "loss": 0.171, "step": 2408 }, { "epoch": 2.7, - "learning_rate": 3.763786764705882e-05, - "loss": 3.0944, + "learning_rate": 3.777573529411765e-06, + "loss": 0.184, "step": 2409 }, { "epoch": 2.7, - "learning_rate": 3.75e-05, - "loss": 2.7977, + "learning_rate": 3.7637867647058822e-06, + "loss": 0.1083, "step": 2410 }, { "epoch": 2.7, - "learning_rate": 3.7362132352941174e-05, - "loss": 2.8267, + "learning_rate": 3.75e-06, + "loss": 0.1544, "step": 2411 }, { "epoch": 2.7, - "learning_rate": 3.722426470588235e-05, - "loss": 2.8219, + "learning_rate": 3.736213235294118e-06, + "loss": 0.1659, "step": 2412 }, { "epoch": 2.71, - "learning_rate": 3.708639705882353e-05, - "loss": 2.8196, + "learning_rate": 3.7224264705882354e-06, + "loss": 0.1606, "step": 2413 }, { "epoch": 2.71, - "learning_rate": 3.6948529411764704e-05, - "loss": 2.8205, + "learning_rate": 3.7086397058823533e-06, + "loss": 0.1423, "step": 2414 }, { "epoch": 2.71, - "learning_rate": 3.6810661764705874e-05, - "loss": 2.9686, + "learning_rate": 3.6948529411764707e-06, + "loss": 0.1847, "step": 2415 }, { "epoch": 2.71, - "learning_rate": 3.667279411764705e-05, - "loss": 3.6836, + "learning_rate": 3.681066176470588e-06, + "loss": 0.1725, "step": 2416 }, { "epoch": 2.71, - "learning_rate": 3.653492647058823e-05, - "loss": 2.799, + "learning_rate": 3.667279411764706e-06, + "loss": 0.1319, "step": 2417 }, { "epoch": 2.71, - "learning_rate": 3.6397058823529405e-05, - "loss": 2.8196, + "learning_rate": 3.6534926470588234e-06, + "loss": 0.1316, "step": 2418 }, { "epoch": 2.71, - "learning_rate": 3.625919117647058e-05, - "loss": 2.8232, + "learning_rate": 3.6397058823529413e-06, + "loss": 0.1398, "step": 2419 }, { "epoch": 2.71, - "learning_rate": 3.612132352941176e-05, - "loss": 2.8231, + "learning_rate": 3.625919117647059e-06, + "loss": 0.1539, "step": 2420 }, { "epoch": 2.71, - "learning_rate": 3.5983455882352936e-05, - "loss": 2.8395, + "learning_rate": 3.6121323529411766e-06, + "loss": 0.1462, "step": 2421 }, { "epoch": 2.72, - "learning_rate": 3.584558823529412e-05, - "loss": 3.9424, + "learning_rate": 3.5983455882352944e-06, + "loss": 0.1394, "step": 2422 }, { "epoch": 2.72, - "learning_rate": 3.570772058823529e-05, - "loss": 2.8096, + "learning_rate": 3.584558823529412e-06, + "loss": 0.1616, "step": 2423 }, { "epoch": 2.72, - "learning_rate": 3.556985294117647e-05, - "loss": 2.8175, + "learning_rate": 3.5707720588235293e-06, + "loss": 0.1664, "step": 2424 }, { "epoch": 2.72, - "learning_rate": 3.5431985294117644e-05, - "loss": 2.8165, + "learning_rate": 3.556985294117647e-06, + "loss": 0.1414, "step": 2425 }, { "epoch": 2.72, - "learning_rate": 3.529411764705882e-05, - "loss": 2.8391, + "learning_rate": 3.5431985294117646e-06, + "loss": 0.1594, "step": 2426 }, { "epoch": 2.72, - "learning_rate": 3.515625e-05, - "loss": 2.8676, + "learning_rate": 3.5294117647058825e-06, + "loss": 0.188, "step": 2427 }, { "epoch": 2.72, - "learning_rate": 3.5018382352941174e-05, - "loss": 4.614, + "learning_rate": 3.515625e-06, + "loss": 0.2155, "step": 2428 }, { "epoch": 2.72, - "learning_rate": 3.488051470588235e-05, - "loss": 2.8223, + "learning_rate": 3.5018382352941178e-06, + "loss": 0.1427, "step": 2429 }, { "epoch": 2.72, - "learning_rate": 3.474264705882353e-05, - "loss": 2.8131, + "learning_rate": 3.4880514705882356e-06, + "loss": 0.1626, "step": 2430 }, { "epoch": 2.73, - "learning_rate": 3.4604779411764705e-05, - "loss": 2.82, + "learning_rate": 3.474264705882353e-06, + "loss": 0.1529, "step": 2431 }, { "epoch": 2.73, - "learning_rate": 3.4466911764705875e-05, - "loss": 2.8173, + "learning_rate": 3.460477941176471e-06, + "loss": 0.1445, "step": 2432 }, { "epoch": 2.73, - "learning_rate": 3.432904411764705e-05, - "loss": 2.8358, + "learning_rate": 3.4466911764705884e-06, + "loss": 0.1918, "step": 2433 }, { "epoch": 2.73, - "learning_rate": 3.419117647058823e-05, - "loss": 3.1063, + "learning_rate": 3.432904411764706e-06, + "loss": 0.2398, "step": 2434 }, { "epoch": 2.73, - "learning_rate": 3.4053308823529406e-05, - "loss": 2.8149, + "learning_rate": 3.4191176470588237e-06, + "loss": 0.1908, "step": 2435 }, { "epoch": 2.73, - "learning_rate": 3.391544117647058e-05, - "loss": 2.7989, + "learning_rate": 3.405330882352941e-06, + "loss": 0.1754, "step": 2436 }, { "epoch": 2.73, - "learning_rate": 3.377757352941176e-05, - "loss": 2.8091, + "learning_rate": 3.391544117647059e-06, + "loss": 0.166, "step": 2437 }, { "epoch": 2.73, - "learning_rate": 3.3639705882352937e-05, - "loss": 2.7984, + "learning_rate": 3.377757352941177e-06, + "loss": 0.1413, "step": 2438 }, { "epoch": 2.73, - "learning_rate": 3.3501838235294113e-05, - "loss": 2.8064, + "learning_rate": 3.3639705882352942e-06, + "loss": 0.1482, "step": 2439 }, { "epoch": 2.74, - "learning_rate": 3.336397058823529e-05, - "loss": 2.8773, + "learning_rate": 3.350183823529412e-06, + "loss": 0.1834, "step": 2440 }, { "epoch": 2.74, - "learning_rate": 3.322610294117647e-05, - "loss": 3.347, + "learning_rate": 3.336397058823529e-06, + "loss": 0.1788, "step": 2441 }, { "epoch": 2.74, - "learning_rate": 3.3088235294117644e-05, - "loss": 2.7912, + "learning_rate": 3.322610294117647e-06, + "loss": 0.1495, "step": 2442 }, { "epoch": 2.74, - "learning_rate": 3.295036764705882e-05, - "loss": 2.7899, + "learning_rate": 3.308823529411765e-06, + "loss": 0.1448, "step": 2443 }, { "epoch": 2.74, - "learning_rate": 3.28125e-05, - "loss": 2.8245, + "learning_rate": 3.2950367647058823e-06, + "loss": 0.1764, "step": 2444 }, { "epoch": 2.74, - "learning_rate": 3.2674632352941175e-05, - "loss": 2.8134, + "learning_rate": 3.28125e-06, + "loss": 0.1498, "step": 2445 }, { "epoch": 2.74, - "learning_rate": 3.253676470588235e-05, - "loss": 2.8565, + "learning_rate": 3.2674632352941176e-06, + "loss": 0.1591, "step": 2446 }, { "epoch": 2.74, - "learning_rate": 3.239889705882353e-05, - "loss": 3.9857, + "learning_rate": 3.2536764705882354e-06, + "loss": 0.187, "step": 2447 }, { "epoch": 2.74, - "learning_rate": 3.2261029411764706e-05, - "loss": 2.8031, + "learning_rate": 3.2398897058823533e-06, + "loss": 0.1356, "step": 2448 }, { "epoch": 2.75, - "learning_rate": 3.2123161764705876e-05, - "loss": 2.8173, + "learning_rate": 3.2261029411764707e-06, + "loss": 0.1574, "step": 2449 }, { "epoch": 2.75, - "learning_rate": 3.198529411764705e-05, - "loss": 2.8161, + "learning_rate": 3.212316176470588e-06, + "loss": 0.1331, "step": 2450 }, { "epoch": 2.75, - "learning_rate": 3.184742647058823e-05, - "loss": 2.8065, + "learning_rate": 3.198529411764706e-06, + "loss": 0.1444, "step": 2451 }, { "epoch": 2.75, - "learning_rate": 3.1709558823529406e-05, - "loss": 2.8519, + "learning_rate": 3.1847426470588235e-06, + "loss": 0.1738, "step": 2452 }, { "epoch": 2.75, - "learning_rate": 3.157169117647058e-05, - "loss": 4.0457, + "learning_rate": 3.1709558823529413e-06, + "loss": 0.2171, "step": 2453 }, { "epoch": 2.75, - "learning_rate": 3.143382352941176e-05, - "loss": 2.7921, + "learning_rate": 3.1571691176470588e-06, + "loss": 0.1554, "step": 2454 }, { "epoch": 2.75, - "learning_rate": 3.129595588235294e-05, - "loss": 2.8096, + "learning_rate": 3.1433823529411766e-06, + "loss": 0.1621, "step": 2455 }, { "epoch": 2.75, - "learning_rate": 3.1158088235294114e-05, - "loss": 2.8078, + "learning_rate": 3.1295955882352945e-06, + "loss": 0.1361, "step": 2456 }, { "epoch": 2.75, - "learning_rate": 3.102022058823529e-05, - "loss": 2.8143, + "learning_rate": 3.115808823529412e-06, + "loss": 0.1497, "step": 2457 }, { "epoch": 2.76, - "learning_rate": 3.088235294117647e-05, - "loss": 2.85, + "learning_rate": 3.1020220588235294e-06, + "loss": 0.1673, "step": 2458 }, { "epoch": 2.76, - "learning_rate": 3.0744485294117645e-05, - "loss": 3.079, + "learning_rate": 3.088235294117647e-06, + "loss": 0.2209, "step": 2459 }, { "epoch": 2.76, - "learning_rate": 3.060661764705882e-05, - "loss": 2.8061, + "learning_rate": 3.0744485294117646e-06, + "loss": 0.1635, "step": 2460 }, { "epoch": 2.76, - "learning_rate": 3.046875e-05, - "loss": 2.81, + "learning_rate": 3.0606617647058825e-06, + "loss": 0.1321, "step": 2461 }, { "epoch": 2.76, - "learning_rate": 3.0330882352941176e-05, - "loss": 2.804, + "learning_rate": 3.046875e-06, + "loss": 0.1384, "step": 2462 }, { "epoch": 2.76, - "learning_rate": 3.0193014705882352e-05, - "loss": 2.8272, + "learning_rate": 3.033088235294118e-06, + "loss": 0.1953, "step": 2463 }, { "epoch": 2.76, - "learning_rate": 3.005514705882353e-05, - "loss": 2.8113, + "learning_rate": 3.0193014705882357e-06, + "loss": 0.1616, "step": 2464 }, { "epoch": 2.76, - "learning_rate": 2.9917279411764706e-05, - "loss": 2.9848, + "learning_rate": 3.005514705882353e-06, + "loss": 0.2191, "step": 2465 }, { "epoch": 2.76, - "learning_rate": 2.9779411764705876e-05, - "loss": 3.5449, + "learning_rate": 2.991727941176471e-06, + "loss": 0.1633, "step": 2466 }, { "epoch": 2.77, - "learning_rate": 2.9641544117647057e-05, - "loss": 2.8032, + "learning_rate": 2.977941176470588e-06, + "loss": 0.1452, "step": 2467 }, { "epoch": 2.77, - "learning_rate": 2.9503676470588234e-05, - "loss": 2.817, + "learning_rate": 2.964154411764706e-06, + "loss": 0.1736, "step": 2468 }, { "epoch": 2.77, - "learning_rate": 2.936580882352941e-05, - "loss": 2.8165, + "learning_rate": 2.9503676470588237e-06, + "loss": 0.134, "step": 2469 }, { "epoch": 2.77, - "learning_rate": 2.9227941176470587e-05, - "loss": 2.8377, + "learning_rate": 2.936580882352941e-06, + "loss": 0.1439, "step": 2470 }, { "epoch": 2.77, - "learning_rate": 2.9090073529411764e-05, - "loss": 2.9186, + "learning_rate": 2.922794117647059e-06, + "loss": 0.1611, "step": 2471 }, { "epoch": 2.77, - "learning_rate": 2.895220588235294e-05, - "loss": 3.8674, + "learning_rate": 2.9090073529411764e-06, + "loss": 0.1999, "step": 2472 }, { "epoch": 2.77, - "learning_rate": 2.8814338235294118e-05, - "loss": 2.8031, + "learning_rate": 2.8952205882352943e-06, + "loss": 0.153, "step": 2473 }, { "epoch": 2.77, - "learning_rate": 2.8676470588235288e-05, - "loss": 2.796, + "learning_rate": 2.881433823529412e-06, + "loss": 0.1678, "step": 2474 }, { "epoch": 2.77, - "learning_rate": 2.8538602941176465e-05, - "loss": 2.8042, + "learning_rate": 2.867647058823529e-06, + "loss": 0.1412, "step": 2475 }, { "epoch": 2.78, - "learning_rate": 2.8400735294117642e-05, - "loss": 2.8097, + "learning_rate": 2.853860294117647e-06, + "loss": 0.1486, "step": 2476 }, { "epoch": 2.78, - "learning_rate": 2.826286764705882e-05, - "loss": 2.9156, + "learning_rate": 2.840073529411765e-06, + "loss": 0.159, "step": 2477 }, { "epoch": 2.78, - "learning_rate": 2.8125e-05, - "loss": 4.4031, + "learning_rate": 2.8262867647058823e-06, + "loss": 0.2414, "step": 2478 }, { "epoch": 2.78, - "learning_rate": 2.7987132352941176e-05, - "loss": 2.8063, + "learning_rate": 2.8125e-06, + "loss": 0.143, "step": 2479 }, { "epoch": 2.78, - "learning_rate": 2.7849264705882353e-05, - "loss": 2.8139, + "learning_rate": 2.7987132352941176e-06, + "loss": 0.1521, "step": 2480 }, { "epoch": 2.78, - "learning_rate": 2.771139705882353e-05, - "loss": 2.8103, + "learning_rate": 2.7849264705882355e-06, + "loss": 0.1954, "step": 2481 }, { "epoch": 2.78, - "learning_rate": 2.7573529411764707e-05, - "loss": 2.8095, + "learning_rate": 2.7711397058823533e-06, + "loss": 0.1543, "step": 2482 }, { "epoch": 2.78, - "learning_rate": 2.7435661764705877e-05, - "loss": 2.815, + "learning_rate": 2.7573529411764708e-06, + "loss": 0.1793, "step": 2483 }, { "epoch": 2.78, - "learning_rate": 2.7297794117647054e-05, - "loss": 3.1989, + "learning_rate": 2.743566176470588e-06, + "loss": 0.2234, "step": 2484 }, { "epoch": 2.79, - "learning_rate": 2.715992647058823e-05, - "loss": 2.8183, + "learning_rate": 2.7297794117647056e-06, + "loss": 0.1637, "step": 2485 }, { "epoch": 2.79, - "learning_rate": 2.7022058823529408e-05, - "loss": 2.8016, + "learning_rate": 2.7159926470588235e-06, + "loss": 0.1538, "step": 2486 }, { "epoch": 2.79, - "learning_rate": 2.6884191176470585e-05, - "loss": 2.7905, + "learning_rate": 2.7022058823529414e-06, + "loss": 0.1719, "step": 2487 }, { "epoch": 2.79, - "learning_rate": 2.674632352941176e-05, - "loss": 2.814, + "learning_rate": 2.688419117647059e-06, + "loss": 0.1639, "step": 2488 }, { "epoch": 2.79, - "learning_rate": 2.660845588235294e-05, - "loss": 2.7968, + "learning_rate": 2.6746323529411767e-06, + "loss": 0.1142, "step": 2489 }, { "epoch": 2.79, - "learning_rate": 2.647058823529412e-05, - "loss": 3.2659, + "learning_rate": 2.660845588235294e-06, + "loss": 0.2368, "step": 2490 }, { "epoch": 2.79, - "learning_rate": 2.633272058823529e-05, - "loss": 4.1254, + "learning_rate": 2.647058823529412e-06, + "loss": 0.1768, "step": 2491 }, { "epoch": 2.79, - "learning_rate": 2.6194852941176466e-05, - "loss": 2.7971, + "learning_rate": 2.6332720588235294e-06, + "loss": 0.1507, "step": 2492 }, { "epoch": 2.79, - "learning_rate": 2.6056985294117643e-05, - "loss": 2.8145, + "learning_rate": 2.619485294117647e-06, + "loss": 0.1695, "step": 2493 }, { "epoch": 2.8, - "learning_rate": 2.591911764705882e-05, - "loss": 2.8057, + "learning_rate": 2.6056985294117647e-06, + "loss": 0.1398, "step": 2494 }, { "epoch": 2.8, - "learning_rate": 2.5781249999999996e-05, - "loss": 2.8142, + "learning_rate": 2.5919117647058825e-06, + "loss": 0.2021, "step": 2495 }, { "epoch": 2.8, - "learning_rate": 2.5643382352941173e-05, - "loss": 2.9577, + "learning_rate": 2.578125e-06, + "loss": 0.2019, "step": 2496 }, { "epoch": 2.8, - "learning_rate": 2.550551470588235e-05, - "loss": 4.7235, + "learning_rate": 2.564338235294118e-06, + "loss": 0.1306, "step": 2497 }, { "epoch": 2.8, - "learning_rate": 2.5367647058823527e-05, - "loss": 2.8093, + "learning_rate": 2.5505514705882353e-06, + "loss": 0.1355, "step": 2498 }, { "epoch": 2.8, - "learning_rate": 2.5229779411764704e-05, - "loss": 2.8216, + "learning_rate": 2.536764705882353e-06, + "loss": 0.1566, "step": 2499 }, { "epoch": 2.8, - "learning_rate": 2.5091911764705878e-05, - "loss": 2.8179, + "learning_rate": 2.522977941176471e-06, + "loss": 0.1417, "step": 2500 }, { "epoch": 2.8, - "eval_loss": 3.9725117683410645, - "eval_runtime": 48.3673, - "eval_samples_per_second": 55.885, - "eval_steps_per_second": 0.889, - "eval_wer": 1.0, + "eval_loss": 0.19831597805023193, + "eval_runtime": 47.4021, + "eval_samples_per_second": 57.023, + "eval_steps_per_second": 0.907, + "eval_wer": 0.16876217786110806, "step": 2500 }, { "epoch": 2.8, - "learning_rate": 2.4954044117647055e-05, - "loss": 2.8233, + "learning_rate": 2.509191176470588e-06, + "loss": 0.1793, "step": 2501 }, { "epoch": 2.8, - "learning_rate": 2.481617647058823e-05, - "loss": 2.8468, + "learning_rate": 2.495404411764706e-06, + "loss": 0.1633, "step": 2502 }, { "epoch": 2.81, - "learning_rate": 2.467830882352941e-05, - "loss": 4.2228, + "learning_rate": 2.4816176470588233e-06, + "loss": 0.1987, "step": 2503 }, { "epoch": 2.81, - "learning_rate": 2.4540441176470585e-05, - "loss": 2.826, + "learning_rate": 2.467830882352941e-06, + "loss": 0.1462, "step": 2504 }, { "epoch": 2.81, - "learning_rate": 2.4402573529411762e-05, - "loss": 2.793, + "learning_rate": 2.454044117647059e-06, + "loss": 0.1395, "step": 2505 }, { "epoch": 2.81, - "learning_rate": 2.426470588235294e-05, - "loss": 2.801, + "learning_rate": 2.4402573529411765e-06, + "loss": 0.1371, "step": 2506 }, { "epoch": 2.81, - "learning_rate": 2.4126838235294116e-05, - "loss": 2.8093, + "learning_rate": 2.4264705882352943e-06, + "loss": 0.1635, "step": 2507 }, { "epoch": 2.81, - "learning_rate": 2.398897058823529e-05, - "loss": 2.8456, + "learning_rate": 2.412683823529412e-06, + "loss": 0.1379, "step": 2508 }, { "epoch": 2.81, - "learning_rate": 2.3851102941176466e-05, - "loss": 3.1946, + "learning_rate": 2.398897058823529e-06, + "loss": 0.2191, "step": 2509 }, { "epoch": 2.81, - "learning_rate": 2.3713235294117643e-05, - "loss": 2.8238, + "learning_rate": 2.385110294117647e-06, + "loss": 0.1524, "step": 2510 }, { "epoch": 2.82, - "learning_rate": 2.357536764705882e-05, - "loss": 2.826, + "learning_rate": 2.3713235294117645e-06, + "loss": 0.162, "step": 2511 }, { "epoch": 2.82, - "learning_rate": 2.3437499999999997e-05, - "loss": 2.8026, + "learning_rate": 2.3575367647058824e-06, + "loss": 0.1227, "step": 2512 }, { "epoch": 2.82, - "learning_rate": 2.3299632352941174e-05, - "loss": 2.8223, + "learning_rate": 2.3437500000000002e-06, + "loss": 0.1491, "step": 2513 }, { "epoch": 2.82, - "learning_rate": 2.316176470588235e-05, - "loss": 2.8258, + "learning_rate": 2.3299632352941177e-06, + "loss": 0.168, "step": 2514 }, { "epoch": 2.82, - "learning_rate": 2.3023897058823528e-05, - "loss": 2.8847, + "learning_rate": 2.3161764705882355e-06, + "loss": 0.2052, "step": 2515 }, { "epoch": 2.82, - "learning_rate": 2.2886029411764705e-05, - "loss": 3.6641, + "learning_rate": 2.302389705882353e-06, + "loss": 0.2448, "step": 2516 }, { "epoch": 2.82, - "learning_rate": 2.2748161764705878e-05, - "loss": 2.8068, + "learning_rate": 2.288602941176471e-06, + "loss": 0.1796, "step": 2517 }, { "epoch": 2.82, - "learning_rate": 2.2610294117647055e-05, - "loss": 2.8245, + "learning_rate": 2.2748161764705882e-06, + "loss": 0.1812, "step": 2518 }, { "epoch": 2.82, - "learning_rate": 2.2472426470588232e-05, - "loss": 2.8112, + "learning_rate": 2.2610294117647057e-06, + "loss": 0.1758, "step": 2519 }, { "epoch": 2.83, - "learning_rate": 2.233455882352941e-05, - "loss": 2.8122, + "learning_rate": 2.2472426470588235e-06, + "loss": 0.1212, "step": 2520 }, { "epoch": 2.83, - "learning_rate": 2.2196691176470586e-05, - "loss": 2.8815, + "learning_rate": 2.2334558823529414e-06, + "loss": 0.1426, "step": 2521 }, { "epoch": 2.83, - "learning_rate": 2.2058823529411763e-05, - "loss": 4.4423, + "learning_rate": 2.219669117647059e-06, + "loss": 0.2083, "step": 2522 }, { "epoch": 2.83, - "learning_rate": 2.192095588235294e-05, - "loss": 2.7993, + "learning_rate": 2.2058823529411767e-06, + "loss": 0.1518, "step": 2523 }, { "epoch": 2.83, - "learning_rate": 2.1783088235294117e-05, - "loss": 2.8119, + "learning_rate": 2.192095588235294e-06, + "loss": 0.1472, "step": 2524 }, { "epoch": 2.83, - "learning_rate": 2.164522058823529e-05, - "loss": 2.8124, + "learning_rate": 2.178308823529412e-06, + "loss": 0.1634, "step": 2525 }, { "epoch": 2.83, - "learning_rate": 2.1507352941176467e-05, - "loss": 2.8095, + "learning_rate": 2.1645220588235294e-06, + "loss": 0.1426, "step": 2526 }, { "epoch": 2.83, - "learning_rate": 2.1369485294117644e-05, - "loss": 2.8306, + "learning_rate": 2.150735294117647e-06, + "loss": 0.1694, "step": 2527 }, { "epoch": 2.83, - "learning_rate": 2.123161764705882e-05, - "loss": 4.2139, + "learning_rate": 2.1369485294117647e-06, + "loss": 0.1785, "step": 2528 }, { "epoch": 2.84, - "learning_rate": 2.1093749999999998e-05, - "loss": 2.8036, + "learning_rate": 2.123161764705882e-06, + "loss": 0.1603, "step": 2529 }, { "epoch": 2.84, - "learning_rate": 2.0955882352941175e-05, - "loss": 2.8322, + "learning_rate": 2.109375e-06, + "loss": 0.1546, "step": 2530 }, { "epoch": 2.84, - "learning_rate": 2.081801470588235e-05, - "loss": 2.8223, + "learning_rate": 2.095588235294118e-06, + "loss": 0.165, "step": 2531 }, { "epoch": 2.84, - "learning_rate": 2.068014705882353e-05, - "loss": 2.8277, + "learning_rate": 2.0818014705882353e-06, + "loss": 0.1302, "step": 2532 }, { "epoch": 2.84, - "learning_rate": 2.0542279411764705e-05, - "loss": 2.8696, + "learning_rate": 2.068014705882353e-06, + "loss": 0.1613, "step": 2533 }, { "epoch": 2.84, - "learning_rate": 2.040441176470588e-05, - "loss": 3.0984, + "learning_rate": 2.054227941176471e-06, + "loss": 0.1783, "step": 2534 }, { "epoch": 2.84, - "learning_rate": 2.0266544117647056e-05, - "loss": 2.8233, + "learning_rate": 2.040441176470588e-06, + "loss": 0.1752, "step": 2535 }, { "epoch": 2.84, - "learning_rate": 2.0128676470588233e-05, - "loss": 2.797, + "learning_rate": 2.026654411764706e-06, + "loss": 0.1474, "step": 2536 }, { "epoch": 2.84, - "learning_rate": 1.999080882352941e-05, - "loss": 2.8226, + "learning_rate": 2.0128676470588234e-06, + "loss": 0.1764, "step": 2537 }, { "epoch": 2.85, - "learning_rate": 1.9852941176470586e-05, - "loss": 2.8191, + "learning_rate": 1.9990808823529412e-06, + "loss": 0.1599, "step": 2538 }, { "epoch": 2.85, - "learning_rate": 1.9715073529411763e-05, - "loss": 2.836, + "learning_rate": 1.985294117647059e-06, + "loss": 0.1651, "step": 2539 }, { "epoch": 2.85, - "learning_rate": 1.957720588235294e-05, - "loss": 3.1812, + "learning_rate": 1.9715073529411765e-06, + "loss": 0.1995, "step": 2540 }, { "epoch": 2.85, - "learning_rate": 1.9439338235294117e-05, - "loss": 4.0503, + "learning_rate": 1.9577205882352944e-06, + "loss": 0.1922, "step": 2541 }, { "epoch": 2.85, - "learning_rate": 1.930147058823529e-05, - "loss": 2.8247, + "learning_rate": 1.943933823529412e-06, + "loss": 0.1618, "step": 2542 }, { "epoch": 2.85, - "learning_rate": 1.9163602941176468e-05, - "loss": 2.8198, + "learning_rate": 1.9301470588235292e-06, + "loss": 0.1324, "step": 2543 }, { "epoch": 2.85, - "learning_rate": 1.9025735294117645e-05, - "loss": 2.7984, + "learning_rate": 1.916360294117647e-06, + "loss": 0.128, "step": 2544 }, { "epoch": 2.85, - "learning_rate": 1.888786764705882e-05, - "loss": 2.8095, + "learning_rate": 1.9025735294117648e-06, + "loss": 0.1283, "step": 2545 }, { "epoch": 2.85, - "learning_rate": 1.875e-05, - "loss": 2.8837, + "learning_rate": 1.8887867647058824e-06, + "loss": 0.1778, "step": 2546 }, { "epoch": 2.86, - "learning_rate": 1.8612132352941175e-05, - "loss": 4.0216, + "learning_rate": 1.875e-06, + "loss": 0.1672, "step": 2547 }, { "epoch": 2.86, - "learning_rate": 1.8474264705882352e-05, - "loss": 2.8024, + "learning_rate": 1.8612132352941177e-06, + "loss": 0.135, "step": 2548 }, { "epoch": 2.86, - "learning_rate": 1.8336397058823526e-05, - "loss": 2.815, + "learning_rate": 1.8474264705882353e-06, + "loss": 0.1519, "step": 2549 }, { "epoch": 2.86, - "learning_rate": 1.8198529411764703e-05, - "loss": 2.8032, + "learning_rate": 1.833639705882353e-06, + "loss": 0.1673, "step": 2550 }, { "epoch": 2.86, - "learning_rate": 1.806066176470588e-05, - "loss": 2.8121, + "learning_rate": 1.8198529411764706e-06, + "loss": 0.1613, "step": 2551 }, { "epoch": 2.86, - "learning_rate": 1.792279411764706e-05, - "loss": 2.8521, + "learning_rate": 1.8060661764705883e-06, + "loss": 0.1926, "step": 2552 }, { "epoch": 2.86, - "learning_rate": 1.7784926470588233e-05, - "loss": 4.4576, + "learning_rate": 1.792279411764706e-06, + "loss": 0.1401, "step": 2553 }, { "epoch": 2.86, - "learning_rate": 1.764705882352941e-05, - "loss": 2.8049, + "learning_rate": 1.7784926470588236e-06, + "loss": 0.174, "step": 2554 }, { "epoch": 2.86, - "learning_rate": 1.7509191176470587e-05, - "loss": 2.8023, + "learning_rate": 1.7647058823529412e-06, + "loss": 0.1468, "step": 2555 }, { "epoch": 2.87, - "learning_rate": 1.7371323529411764e-05, - "loss": 2.8151, + "learning_rate": 1.7509191176470589e-06, + "loss": 0.1533, "step": 2556 }, { "epoch": 2.87, - "learning_rate": 1.7233455882352938e-05, - "loss": 2.7987, + "learning_rate": 1.7371323529411765e-06, + "loss": 0.1314, "step": 2557 }, { "epoch": 2.87, - "learning_rate": 1.7095588235294114e-05, - "loss": 2.8671, + "learning_rate": 1.7233455882352942e-06, + "loss": 0.1679, "step": 2558 }, { "epoch": 2.87, - "learning_rate": 1.695772058823529e-05, - "loss": 3.4517, + "learning_rate": 1.7095588235294118e-06, + "loss": 0.2388, "step": 2559 }, { "epoch": 2.87, - "learning_rate": 1.6819852941176468e-05, - "loss": 2.8025, + "learning_rate": 1.6957720588235295e-06, + "loss": 0.1375, "step": 2560 }, { "epoch": 2.87, - "learning_rate": 1.6681985294117645e-05, - "loss": 2.8006, + "learning_rate": 1.6819852941176471e-06, + "loss": 0.1383, "step": 2561 }, { "epoch": 2.87, - "learning_rate": 1.6544117647058822e-05, - "loss": 2.8148, + "learning_rate": 1.6681985294117646e-06, + "loss": 0.1757, "step": 2562 }, { "epoch": 2.87, - "learning_rate": 1.640625e-05, - "loss": 2.8113, + "learning_rate": 1.6544117647058824e-06, + "loss": 0.1486, "step": 2563 }, { "epoch": 2.87, - "learning_rate": 1.6268382352941176e-05, - "loss": 2.8321, + "learning_rate": 1.640625e-06, + "loss": 0.1481, "step": 2564 }, { "epoch": 2.88, - "learning_rate": 1.6130514705882353e-05, - "loss": 3.1591, + "learning_rate": 1.6268382352941177e-06, + "loss": 0.1724, "step": 2565 }, { "epoch": 2.88, - "learning_rate": 1.5992647058823526e-05, - "loss": 4.0667, + "learning_rate": 1.6130514705882354e-06, + "loss": 0.2019, "step": 2566 }, { "epoch": 2.88, - "learning_rate": 1.5854779411764703e-05, - "loss": 2.8078, + "learning_rate": 1.599264705882353e-06, + "loss": 0.1536, "step": 2567 }, { "epoch": 2.88, - "learning_rate": 1.571691176470588e-05, - "loss": 2.8155, + "learning_rate": 1.5854779411764707e-06, + "loss": 0.1621, "step": 2568 }, { "epoch": 2.88, - "learning_rate": 1.5579044117647057e-05, - "loss": 2.8056, + "learning_rate": 1.5716911764705883e-06, + "loss": 0.1485, "step": 2569 }, { "epoch": 2.88, - "learning_rate": 1.5441176470588234e-05, - "loss": 2.8215, + "learning_rate": 1.557904411764706e-06, + "loss": 0.1613, "step": 2570 }, { "epoch": 2.88, - "learning_rate": 1.530330882352941e-05, - "loss": 2.8971, + "learning_rate": 1.5441176470588234e-06, + "loss": 0.1647, "step": 2571 }, { "epoch": 2.88, - "learning_rate": 1.5165441176470588e-05, - "loss": 4.2977, + "learning_rate": 1.5303308823529413e-06, + "loss": 0.2288, "step": 2572 }, { "epoch": 2.88, - "learning_rate": 1.5027573529411765e-05, - "loss": 2.8097, + "learning_rate": 1.516544117647059e-06, + "loss": 0.1532, "step": 2573 }, { "epoch": 2.89, - "learning_rate": 1.4889705882352938e-05, - "loss": 2.8214, + "learning_rate": 1.5027573529411766e-06, + "loss": 0.1756, "step": 2574 }, { "epoch": 2.89, - "learning_rate": 1.4751838235294117e-05, - "loss": 2.8225, + "learning_rate": 1.488970588235294e-06, + "loss": 0.1704, "step": 2575 }, { "epoch": 2.89, - "learning_rate": 1.4613970588235294e-05, - "loss": 2.8123, + "learning_rate": 1.4751838235294118e-06, + "loss": 0.1504, "step": 2576 }, { "epoch": 2.89, - "learning_rate": 1.447610294117647e-05, - "loss": 2.8174, + "learning_rate": 1.4613970588235295e-06, + "loss": 0.162, "step": 2577 }, { "epoch": 2.89, - "learning_rate": 1.4338235294117644e-05, - "loss": 4.3284, + "learning_rate": 1.4476102941176471e-06, + "loss": 0.2519, "step": 2578 }, { "epoch": 2.89, - "learning_rate": 1.4200367647058821e-05, - "loss": 2.8043, + "learning_rate": 1.4338235294117646e-06, + "loss": 0.1833, "step": 2579 }, { "epoch": 2.89, - "learning_rate": 1.40625e-05, - "loss": 2.8089, + "learning_rate": 1.4200367647058824e-06, + "loss": 0.152, "step": 2580 }, { "epoch": 2.89, - "learning_rate": 1.3924632352941177e-05, - "loss": 2.8057, + "learning_rate": 1.40625e-06, + "loss": 0.1543, "step": 2581 }, { "epoch": 2.89, - "learning_rate": 1.3786764705882353e-05, - "loss": 2.8202, + "learning_rate": 1.3924632352941177e-06, + "loss": 0.161, "step": 2582 }, { "epoch": 2.9, - "learning_rate": 1.3648897058823527e-05, - "loss": 2.8354, + "learning_rate": 1.3786764705882354e-06, + "loss": 0.1929, "step": 2583 }, { "epoch": 2.9, - "learning_rate": 1.3511029411764704e-05, - "loss": 3.3643, + "learning_rate": 1.3648897058823528e-06, + "loss": 0.1841, "step": 2584 }, { "epoch": 2.9, - "learning_rate": 1.337316176470588e-05, - "loss": 2.8025, + "learning_rate": 1.3511029411764707e-06, + "loss": 0.1789, "step": 2585 }, { "epoch": 2.9, - "learning_rate": 1.323529411764706e-05, - "loss": 2.8005, + "learning_rate": 1.3373161764705883e-06, + "loss": 0.1527, "step": 2586 }, { "epoch": 2.9, - "learning_rate": 1.3097426470588233e-05, - "loss": 2.801, + "learning_rate": 1.323529411764706e-06, + "loss": 0.1518, "step": 2587 }, { "epoch": 2.9, - "learning_rate": 1.295955882352941e-05, - "loss": 2.8073, + "learning_rate": 1.3097426470588234e-06, + "loss": 0.1577, "step": 2588 }, { "epoch": 2.9, - "learning_rate": 1.2821691176470587e-05, - "loss": 2.8056, + "learning_rate": 1.2959558823529413e-06, + "loss": 0.1983, "step": 2589 }, { "epoch": 2.9, - "learning_rate": 1.2683823529411764e-05, - "loss": 3.1023, + "learning_rate": 1.282169117647059e-06, + "loss": 0.1509, "step": 2590 }, { "epoch": 2.9, - "learning_rate": 1.2545955882352939e-05, - "loss": 4.2717, + "learning_rate": 1.2683823529411766e-06, + "loss": 0.1566, "step": 2591 }, { "epoch": 2.91, - "learning_rate": 1.2408088235294116e-05, - "loss": 2.7966, + "learning_rate": 1.254595588235294e-06, + "loss": 0.151, "step": 2592 }, { "epoch": 2.91, - "learning_rate": 1.2270220588235293e-05, - "loss": 2.8075, + "learning_rate": 1.2408088235294117e-06, + "loss": 0.1628, "step": 2593 }, { "epoch": 2.91, - "learning_rate": 1.213235294117647e-05, - "loss": 2.8203, + "learning_rate": 1.2270220588235295e-06, + "loss": 0.1444, "step": 2594 }, { "epoch": 2.91, - "learning_rate": 1.1994485294117645e-05, - "loss": 2.8031, + "learning_rate": 1.2132352941176472e-06, + "loss": 0.1486, "step": 2595 }, { "epoch": 2.91, - "learning_rate": 1.1856617647058822e-05, - "loss": 2.8754, + "learning_rate": 1.1994485294117646e-06, + "loss": 0.1573, "step": 2596 }, { "epoch": 2.91, - "learning_rate": 1.1718749999999999e-05, - "loss": 4.3035, + "learning_rate": 1.1856617647058822e-06, + "loss": 0.2088, "step": 2597 }, { "epoch": 2.91, - "learning_rate": 1.1580882352941175e-05, - "loss": 2.7922, + "learning_rate": 1.1718750000000001e-06, + "loss": 0.1504, "step": 2598 }, { "epoch": 2.91, - "learning_rate": 1.1443014705882352e-05, - "loss": 2.8148, + "learning_rate": 1.1580882352941178e-06, + "loss": 0.1682, "step": 2599 }, { "epoch": 2.91, - "learning_rate": 1.1305147058823528e-05, - "loss": 2.8098, + "learning_rate": 1.1443014705882354e-06, + "loss": 0.1302, "step": 2600 }, { "epoch": 2.91, - "eval_loss": 3.9591329097747803, - "eval_runtime": 47.9456, - "eval_samples_per_second": 56.376, - "eval_steps_per_second": 0.897, - "eval_wer": 1.0, + "eval_loss": 0.1976867914199829, + "eval_runtime": 47.6743, + "eval_samples_per_second": 56.697, + "eval_steps_per_second": 0.902, + "eval_wer": 0.16778794897246424, "step": 2600 }, { "epoch": 2.92, - "learning_rate": 1.1167279411764704e-05, - "loss": 2.8253, + "learning_rate": 1.1305147058823528e-06, + "loss": 0.1599, "step": 2601 }, { "epoch": 2.92, - "learning_rate": 1.1029411764705881e-05, - "loss": 2.844, + "learning_rate": 1.1167279411764707e-06, + "loss": 0.1691, "step": 2602 }, { "epoch": 2.92, - "learning_rate": 1.0891544117647058e-05, - "loss": 4.689, + "learning_rate": 1.1029411764705884e-06, + "loss": 0.1868, "step": 2603 }, { "epoch": 2.92, - "learning_rate": 1.0753676470588234e-05, - "loss": 2.8038, + "learning_rate": 1.089154411764706e-06, + "loss": 0.1182, "step": 2604 }, { "epoch": 2.92, - "learning_rate": 1.061580882352941e-05, - "loss": 2.7987, + "learning_rate": 1.0753676470588234e-06, + "loss": 0.1469, "step": 2605 }, { "epoch": 2.92, - "learning_rate": 1.0477941176470587e-05, - "loss": 2.8122, + "learning_rate": 1.061580882352941e-06, + "loss": 0.1449, "step": 2606 }, { "epoch": 2.92, - "learning_rate": 1.0340073529411764e-05, - "loss": 2.8179, + "learning_rate": 1.047794117647059e-06, + "loss": 0.1786, "step": 2607 }, { "epoch": 2.92, - "learning_rate": 1.020220588235294e-05, - "loss": 2.8057, + "learning_rate": 1.0340073529411766e-06, + "loss": 0.1218, "step": 2608 }, { "epoch": 2.92, - "learning_rate": 1.0064338235294116e-05, - "loss": 3.3286, + "learning_rate": 1.020220588235294e-06, + "loss": 0.1691, "step": 2609 }, { "epoch": 2.93, - "learning_rate": 9.926470588235293e-06, - "loss": 2.8072, + "learning_rate": 1.0064338235294117e-06, + "loss": 0.169, "step": 2610 }, { "epoch": 2.93, - "learning_rate": 9.78860294117647e-06, - "loss": 2.8282, + "learning_rate": 9.926470588235295e-07, + "loss": 0.1465, "step": 2611 }, { "epoch": 2.93, - "learning_rate": 9.650735294117645e-06, - "loss": 2.8141, + "learning_rate": 9.788602941176472e-07, + "loss": 0.1638, "step": 2612 }, { "epoch": 2.93, - "learning_rate": 9.512867647058822e-06, - "loss": 2.812, + "learning_rate": 9.650735294117646e-07, + "loss": 0.1838, "step": 2613 }, { "epoch": 2.93, - "learning_rate": 9.375e-06, - "loss": 2.8375, + "learning_rate": 9.512867647058824e-07, + "loss": 0.1855, "step": 2614 }, { "epoch": 2.93, - "learning_rate": 9.237132352941176e-06, - "loss": 2.9144, + "learning_rate": 9.375e-07, + "loss": 0.2117, "step": 2615 }, { "epoch": 2.93, - "learning_rate": 9.099264705882351e-06, - "loss": 4.1624, + "learning_rate": 9.237132352941177e-07, + "loss": 0.1791, "step": 2616 }, { "epoch": 2.93, - "learning_rate": 8.96139705882353e-06, - "loss": 2.803, + "learning_rate": 9.099264705882353e-07, + "loss": 0.1507, "step": 2617 }, { "epoch": 2.93, - "learning_rate": 8.823529411764705e-06, - "loss": 2.8094, + "learning_rate": 8.96139705882353e-07, + "loss": 0.154, "step": 2618 }, { "epoch": 2.94, - "learning_rate": 8.685661764705882e-06, - "loss": 2.8026, + "learning_rate": 8.823529411764706e-07, + "loss": 0.149, "step": 2619 }, { "epoch": 2.94, - "learning_rate": 8.547794117647057e-06, - "loss": 2.8107, + "learning_rate": 8.685661764705883e-07, + "loss": 0.167, "step": 2620 }, { "epoch": 2.94, - "learning_rate": 8.409926470588234e-06, - "loss": 2.8893, + "learning_rate": 8.547794117647059e-07, + "loss": 0.1663, "step": 2621 }, { "epoch": 2.94, - "learning_rate": 8.272058823529411e-06, - "loss": 4.1489, + "learning_rate": 8.409926470588236e-07, + "loss": 0.1743, "step": 2622 }, { "epoch": 2.94, - "learning_rate": 8.134191176470588e-06, - "loss": 2.8061, + "learning_rate": 8.272058823529412e-07, + "loss": 0.1671, "step": 2623 }, { "epoch": 2.94, - "learning_rate": 7.996323529411763e-06, - "loss": 2.7953, + "learning_rate": 8.134191176470589e-07, + "loss": 0.1388, "step": 2624 }, { "epoch": 2.94, - "learning_rate": 7.85845588235294e-06, - "loss": 2.8004, + "learning_rate": 7.996323529411765e-07, + "loss": 0.1843, "step": 2625 }, { "epoch": 2.94, - "learning_rate": 7.720588235294117e-06, - "loss": 2.8041, + "learning_rate": 7.858455882352942e-07, + "loss": 0.1535, "step": 2626 }, { "epoch": 2.95, - "learning_rate": 7.582720588235294e-06, - "loss": 2.8191, + "learning_rate": 7.720588235294117e-07, + "loss": 0.1321, "step": 2627 }, { "epoch": 2.95, - "learning_rate": 7.444852941176469e-06, - "loss": 4.066, + "learning_rate": 7.582720588235295e-07, + "loss": 0.2189, "step": 2628 }, { "epoch": 2.95, - "learning_rate": 7.306985294117647e-06, - "loss": 2.7988, + "learning_rate": 7.44485294117647e-07, + "loss": 0.1727, "step": 2629 }, { "epoch": 2.95, - "learning_rate": 7.169117647058822e-06, - "loss": 2.8129, + "learning_rate": 7.306985294117647e-07, + "loss": 0.1405, "step": 2630 }, { "epoch": 2.95, - "learning_rate": 7.03125e-06, - "loss": 2.7986, + "learning_rate": 7.169117647058823e-07, + "loss": 0.1229, "step": 2631 }, { "epoch": 2.95, - "learning_rate": 6.893382352941177e-06, - "loss": 2.8118, + "learning_rate": 7.03125e-07, + "loss": 0.1334, "step": 2632 }, { "epoch": 2.95, - "learning_rate": 6.755514705882352e-06, - "loss": 2.8003, + "learning_rate": 6.893382352941177e-07, + "loss": 0.164, "step": 2633 }, { "epoch": 2.95, - "learning_rate": 6.61764705882353e-06, - "loss": 3.3194, + "learning_rate": 6.755514705882353e-07, + "loss": 0.1755, "step": 2634 }, { "epoch": 2.95, - "learning_rate": 6.479779411764705e-06, - "loss": 2.8132, + "learning_rate": 6.61764705882353e-07, + "loss": 0.1603, "step": 2635 }, { "epoch": 2.96, - "learning_rate": 6.341911764705882e-06, - "loss": 2.7908, + "learning_rate": 6.479779411764706e-07, + "loss": 0.1507, "step": 2636 }, { "epoch": 2.96, - "learning_rate": 6.204044117647058e-06, - "loss": 2.828, + "learning_rate": 6.341911764705883e-07, + "loss": 0.1543, "step": 2637 }, { "epoch": 2.96, - "learning_rate": 6.066176470588235e-06, - "loss": 2.8241, + "learning_rate": 6.204044117647058e-07, + "loss": 0.1328, "step": 2638 }, { "epoch": 2.96, - "learning_rate": 5.928308823529411e-06, - "loss": 2.8089, + "learning_rate": 6.066176470588236e-07, + "loss": 0.1446, "step": 2639 }, { "epoch": 2.96, - "learning_rate": 5.790441176470588e-06, - "loss": 2.953, + "learning_rate": 5.928308823529411e-07, + "loss": 0.1955, "step": 2640 }, { "epoch": 2.96, - "learning_rate": 5.652573529411764e-06, - "loss": 4.2951, + "learning_rate": 5.790441176470589e-07, + "loss": 0.1898, "step": 2641 }, { "epoch": 2.96, - "learning_rate": 5.514705882352941e-06, - "loss": 2.799, + "learning_rate": 5.652573529411764e-07, + "loss": 0.143, "step": 2642 }, { "epoch": 2.96, - "learning_rate": 5.376838235294117e-06, - "loss": 2.8144, + "learning_rate": 5.514705882352942e-07, + "loss": 0.1487, "step": 2643 }, { "epoch": 2.96, - "learning_rate": 5.238970588235294e-06, - "loss": 2.8121, + "learning_rate": 5.376838235294117e-07, + "loss": 0.1326, "step": 2644 }, { "epoch": 2.97, - "learning_rate": 5.10110294117647e-06, - "loss": 2.8206, + "learning_rate": 5.238970588235295e-07, + "loss": 0.1192, "step": 2645 }, { "epoch": 2.97, - "learning_rate": 4.963235294117647e-06, - "loss": 2.8724, + "learning_rate": 5.10110294117647e-07, + "loss": 0.1931, "step": 2646 }, { "epoch": 2.97, - "learning_rate": 4.825367647058823e-06, - "loss": 4.3045, + "learning_rate": 4.963235294117648e-07, + "loss": 0.1678, "step": 2647 }, { "epoch": 2.97, - "learning_rate": 4.6875e-06, - "loss": 2.8077, + "learning_rate": 4.825367647058823e-07, + "loss": 0.1587, "step": 2648 }, { "epoch": 2.97, - "learning_rate": 4.549632352941176e-06, - "loss": 2.814, + "learning_rate": 4.6875e-07, + "loss": 0.1516, "step": 2649 }, { "epoch": 2.97, - "learning_rate": 4.4117647058823526e-06, - "loss": 2.8035, + "learning_rate": 4.5496323529411766e-07, + "loss": 0.1454, "step": 2650 }, { "epoch": 2.97, - "learning_rate": 4.273897058823529e-06, - "loss": 2.8285, + "learning_rate": 4.411764705882353e-07, + "loss": 0.1259, "step": 2651 }, { "epoch": 2.97, - "learning_rate": 4.1360294117647055e-06, - "loss": 2.8343, + "learning_rate": 4.2738970588235296e-07, + "loss": 0.1604, "step": 2652 }, { "epoch": 2.97, - "learning_rate": 3.998161764705882e-06, - "loss": 4.0238, + "learning_rate": 4.136029411764706e-07, + "loss": 0.1872, "step": 2653 }, { "epoch": 2.98, - "learning_rate": 3.8602941176470585e-06, - "loss": 2.8154, + "learning_rate": 3.9981617647058825e-07, + "loss": 0.174, "step": 2654 }, { "epoch": 2.98, - "learning_rate": 3.7224264705882345e-06, - "loss": 2.7956, + "learning_rate": 3.8602941176470585e-07, + "loss": 0.1494, "step": 2655 }, { "epoch": 2.98, - "learning_rate": 3.584558823529411e-06, - "loss": 2.8108, + "learning_rate": 3.722426470588235e-07, + "loss": 0.1456, "step": 2656 }, { "epoch": 2.98, - "learning_rate": 3.4466911764705884e-06, - "loss": 2.8041, + "learning_rate": 3.5845588235294115e-07, + "loss": 0.1499, "step": 2657 }, { "epoch": 2.98, - "learning_rate": 3.308823529411765e-06, - "loss": 2.8261, + "learning_rate": 3.4466911764705885e-07, + "loss": 0.1475, "step": 2658 }, { "epoch": 2.98, - "learning_rate": 3.170955882352941e-06, - "loss": 3.2979, + "learning_rate": 3.308823529411765e-07, + "loss": 0.164, "step": 2659 }, { "epoch": 2.98, - "learning_rate": 3.0330882352941174e-06, - "loss": 2.8182, + "learning_rate": 3.1709558823529414e-07, + "loss": 0.1736, "step": 2660 }, { "epoch": 2.98, - "learning_rate": 2.895220588235294e-06, - "loss": 2.813, + "learning_rate": 3.033088235294118e-07, + "loss": 0.1358, "step": 2661 }, { "epoch": 2.98, - "learning_rate": 2.7573529411764703e-06, - "loss": 2.7971, + "learning_rate": 2.8952205882352944e-07, + "loss": 0.1609, "step": 2662 }, { "epoch": 2.99, - "learning_rate": 2.619485294117647e-06, - "loss": 2.8172, + "learning_rate": 2.757352941176471e-07, + "loss": 0.1507, "step": 2663 }, { "epoch": 2.99, - "learning_rate": 2.4816176470588233e-06, - "loss": 2.8076, + "learning_rate": 2.6194852941176474e-07, + "loss": 0.147, "step": 2664 }, { "epoch": 2.99, - "learning_rate": 2.34375e-06, - "loss": 2.9311, + "learning_rate": 2.481617647058824e-07, + "loss": 0.1869, "step": 2665 }, { "epoch": 2.99, - "learning_rate": 2.2058823529411763e-06, - "loss": 3.6914, + "learning_rate": 2.34375e-07, + "loss": 0.15, "step": 2666 }, { "epoch": 2.99, - "learning_rate": 2.0680147058823528e-06, - "loss": 2.8134, + "learning_rate": 2.2058823529411765e-07, + "loss": 0.1559, "step": 2667 }, { "epoch": 2.99, - "learning_rate": 1.9301470588235292e-06, - "loss": 2.8046, + "learning_rate": 2.068014705882353e-07, + "loss": 0.1662, "step": 2668 }, { "epoch": 2.99, - "learning_rate": 1.7922794117647055e-06, - "loss": 2.8137, + "learning_rate": 1.9301470588235292e-07, + "loss": 0.1379, "step": 2669 }, { "epoch": 2.99, - "learning_rate": 1.6544117647058824e-06, - "loss": 2.8205, + "learning_rate": 1.7922794117647057e-07, + "loss": 0.1686, "step": 2670 }, { "epoch": 2.99, - "learning_rate": 1.5165441176470587e-06, - "loss": 2.8325, + "learning_rate": 1.6544117647058825e-07, + "loss": 0.189, "step": 2671 }, { "epoch": 3.0, - "learning_rate": 1.3786764705882352e-06, - "loss": 3.8884, + "learning_rate": 1.516544117647059e-07, + "loss": 0.1676, "step": 2672 }, { "epoch": 3.0, - "learning_rate": 1.2408088235294117e-06, - "loss": 2.7972, + "learning_rate": 1.3786764705882354e-07, + "loss": 0.1692, "step": 2673 }, { "epoch": 3.0, - "learning_rate": 1.1029411764705881e-06, - "loss": 2.8215, + "learning_rate": 1.240808823529412e-07, + "loss": 0.1816, "step": 2674 }, { "epoch": 3.0, - "learning_rate": 9.650735294117646e-07, - "loss": 2.8421, + "learning_rate": 1.1029411764705883e-07, + "loss": 0.1517, "step": 2675 }, { "epoch": 3.0, - "learning_rate": 8.272058823529412e-07, - "loss": 3.8869, + "learning_rate": 9.650735294117646e-08, + "loss": 0.1309, "step": 2676 }, { "epoch": 3.0, "step": 2676, "total_flos": 2.024684214493066e+19, - "train_loss": 3.0757908573421484, - "train_runtime": 3886.8119, - "train_samples_per_second": 22.028, - "train_steps_per_second": 0.688 + "train_loss": 0.9408533566301446, + "train_runtime": 3518.1837, + "train_samples_per_second": 24.336, + "train_steps_per_second": 0.761 } ], "max_steps": 2676,