| { |
| "best_metric": 31.733787724874663, |
| "best_model_checkpoint": "./LMT2\\checkpoint-4000", |
| "epoch": 0.062114694783918505, |
| "global_step": 4000, |
| "is_hyper_param_search": false, |
| "is_local_process_zero": true, |
| "is_world_process_zero": true, |
| "log_history": [ |
| { |
| "epoch": 0.0, |
| "learning_rate": 4.2000000000000006e-07, |
| "loss": 4.0064, |
| "step": 25 |
| }, |
| { |
| "epoch": 0.0, |
| "learning_rate": 9.200000000000001e-07, |
| "loss": 3.2941, |
| "step": 50 |
| }, |
| { |
| "epoch": 0.0, |
| "learning_rate": 1.42e-06, |
| "loss": 2.3704, |
| "step": 75 |
| }, |
| { |
| "epoch": 0.0, |
| "learning_rate": 1.9200000000000003e-06, |
| "loss": 1.3666, |
| "step": 100 |
| }, |
| { |
| "epoch": 0.0, |
| "learning_rate": 2.42e-06, |
| "loss": 0.9502, |
| "step": 125 |
| }, |
| { |
| "epoch": 0.0, |
| "learning_rate": 2.92e-06, |
| "loss": 0.8216, |
| "step": 150 |
| }, |
| { |
| "epoch": 0.0, |
| "learning_rate": 3.4200000000000007e-06, |
| "loss": 0.7506, |
| "step": 175 |
| }, |
| { |
| "epoch": 0.0, |
| "learning_rate": 3.920000000000001e-06, |
| "loss": 0.7069, |
| "step": 200 |
| }, |
| { |
| "epoch": 0.0, |
| "learning_rate": 4.42e-06, |
| "loss": 0.6066, |
| "step": 225 |
| }, |
| { |
| "epoch": 0.0, |
| "learning_rate": 4.92e-06, |
| "loss": 0.5391, |
| "step": 250 |
| }, |
| { |
| "epoch": 0.0, |
| "learning_rate": 5.420000000000001e-06, |
| "loss": 0.4676, |
| "step": 275 |
| }, |
| { |
| "epoch": 0.0, |
| "learning_rate": 5.92e-06, |
| "loss": 0.3422, |
| "step": 300 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 6.42e-06, |
| "loss": 0.3114, |
| "step": 325 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 6.92e-06, |
| "loss": 0.2746, |
| "step": 350 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 7.420000000000001e-06, |
| "loss": 0.2893, |
| "step": 375 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 7.92e-06, |
| "loss": 0.2989, |
| "step": 400 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 8.42e-06, |
| "loss": 0.2677, |
| "step": 425 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 8.920000000000001e-06, |
| "loss": 0.3219, |
| "step": 450 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 9.42e-06, |
| "loss": 0.3024, |
| "step": 475 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 9.920000000000002e-06, |
| "loss": 0.2838, |
| "step": 500 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 9.940000000000001e-06, |
| "loss": 0.2785, |
| "step": 525 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 9.86857142857143e-06, |
| "loss": 0.2667, |
| "step": 550 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 9.797142857142858e-06, |
| "loss": 0.296, |
| "step": 575 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 9.725714285714287e-06, |
| "loss": 0.2913, |
| "step": 600 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 9.654285714285716e-06, |
| "loss": 0.3022, |
| "step": 625 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 9.582857142857143e-06, |
| "loss": 0.2617, |
| "step": 650 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 9.511428571428572e-06, |
| "loss": 0.2747, |
| "step": 675 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 9.440000000000001e-06, |
| "loss": 0.289, |
| "step": 700 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 9.368571428571428e-06, |
| "loss": 0.2664, |
| "step": 725 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 9.297142857142857e-06, |
| "loss": 0.2704, |
| "step": 750 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 9.225714285714286e-06, |
| "loss": 0.2659, |
| "step": 775 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 9.154285714285715e-06, |
| "loss": 0.2679, |
| "step": 800 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 9.082857142857143e-06, |
| "loss": 0.2833, |
| "step": 825 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 9.011428571428572e-06, |
| "loss": 0.2651, |
| "step": 850 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 8.94e-06, |
| "loss": 0.2936, |
| "step": 875 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 8.86857142857143e-06, |
| "loss": 0.2829, |
| "step": 900 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 8.797142857142857e-06, |
| "loss": 0.2706, |
| "step": 925 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 8.725714285714286e-06, |
| "loss": 0.2938, |
| "step": 950 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 8.654285714285715e-06, |
| "loss": 0.2975, |
| "step": 975 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 8.582857142857144e-06, |
| "loss": 0.254, |
| "step": 1000 |
| }, |
| { |
| "epoch": 0.02, |
| "eval_loss": 0.3774094581604004, |
| "eval_runtime": 6578.4668, |
| "eval_samples_per_second": 2.489, |
| "eval_steps_per_second": 0.311, |
| "eval_wer": 38.74037421764918, |
| "step": 1000 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 8.511428571428571e-06, |
| "loss": 0.276, |
| "step": 1025 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 8.44e-06, |
| "loss": 0.2429, |
| "step": 1050 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 8.36857142857143e-06, |
| "loss": 0.2897, |
| "step": 1075 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 8.297142857142859e-06, |
| "loss": 0.25, |
| "step": 1100 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 8.225714285714288e-06, |
| "loss": 0.2531, |
| "step": 1125 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 8.154285714285715e-06, |
| "loss": 0.2671, |
| "step": 1150 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 8.082857142857144e-06, |
| "loss": 0.2584, |
| "step": 1175 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 8.011428571428573e-06, |
| "loss": 0.2791, |
| "step": 1200 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 7.94e-06, |
| "loss": 0.2686, |
| "step": 1225 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 7.86857142857143e-06, |
| "loss": 0.2511, |
| "step": 1250 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 7.797142857142858e-06, |
| "loss": 0.2821, |
| "step": 1275 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 7.725714285714286e-06, |
| "loss": 0.2645, |
| "step": 1300 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 7.654285714285715e-06, |
| "loss": 0.2608, |
| "step": 1325 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 7.5828571428571444e-06, |
| "loss": 0.2745, |
| "step": 1350 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 7.511428571428572e-06, |
| "loss": 0.2865, |
| "step": 1375 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 7.440000000000001e-06, |
| "loss": 0.2767, |
| "step": 1400 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 7.36857142857143e-06, |
| "loss": 0.2775, |
| "step": 1425 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 7.297142857142858e-06, |
| "loss": 0.2997, |
| "step": 1450 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 7.225714285714286e-06, |
| "loss": 0.2723, |
| "step": 1475 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 7.154285714285715e-06, |
| "loss": 0.2512, |
| "step": 1500 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 7.082857142857143e-06, |
| "loss": 0.2803, |
| "step": 1525 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 7.011428571428572e-06, |
| "loss": 0.2905, |
| "step": 1550 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 6.9400000000000005e-06, |
| "loss": 0.2778, |
| "step": 1575 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 6.868571428571429e-06, |
| "loss": 0.2503, |
| "step": 1600 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 6.797142857142858e-06, |
| "loss": 0.2909, |
| "step": 1625 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 6.725714285714287e-06, |
| "loss": 0.2437, |
| "step": 1650 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 6.654285714285716e-06, |
| "loss": 0.2407, |
| "step": 1675 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 6.582857142857143e-06, |
| "loss": 0.2376, |
| "step": 1700 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 6.511428571428572e-06, |
| "loss": 0.2798, |
| "step": 1725 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 6.440000000000001e-06, |
| "loss": 0.2616, |
| "step": 1750 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 6.368571428571429e-06, |
| "loss": 0.2834, |
| "step": 1775 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 6.297142857142857e-06, |
| "loss": 0.26, |
| "step": 1800 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 6.225714285714286e-06, |
| "loss": 0.2766, |
| "step": 1825 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 6.1542857142857145e-06, |
| "loss": 0.2877, |
| "step": 1850 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 6.0828571428571435e-06, |
| "loss": 0.2387, |
| "step": 1875 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 6.011428571428572e-06, |
| "loss": 0.2561, |
| "step": 1900 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 5.94e-06, |
| "loss": 0.2661, |
| "step": 1925 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 5.868571428571429e-06, |
| "loss": 0.2718, |
| "step": 1950 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 5.797142857142858e-06, |
| "loss": 0.2819, |
| "step": 1975 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 5.725714285714287e-06, |
| "loss": 0.2612, |
| "step": 2000 |
| }, |
| { |
| "epoch": 0.03, |
| "eval_loss": 0.35927361249923706, |
| "eval_runtime": 5393.364, |
| "eval_samples_per_second": 3.036, |
| "eval_steps_per_second": 0.38, |
| "eval_wer": 34.29825998623718, |
| "step": 2000 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 5.654285714285714e-06, |
| "loss": 0.2286, |
| "step": 2025 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 5.582857142857143e-06, |
| "loss": 0.2463, |
| "step": 2050 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 5.511428571428572e-06, |
| "loss": 0.268, |
| "step": 2075 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 5.4400000000000004e-06, |
| "loss": 0.2583, |
| "step": 2100 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 5.368571428571429e-06, |
| "loss": 0.2887, |
| "step": 2125 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 5.297142857142858e-06, |
| "loss": 0.2491, |
| "step": 2150 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 5.225714285714286e-06, |
| "loss": 0.235, |
| "step": 2175 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 5.154285714285715e-06, |
| "loss": 0.2838, |
| "step": 2200 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 5.082857142857144e-06, |
| "loss": 0.253, |
| "step": 2225 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 5.011428571428571e-06, |
| "loss": 0.2565, |
| "step": 2250 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 4.94e-06, |
| "loss": 0.2459, |
| "step": 2275 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 4.868571428571429e-06, |
| "loss": 0.2722, |
| "step": 2300 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 4.797142857142857e-06, |
| "loss": 0.2533, |
| "step": 2325 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 4.725714285714286e-06, |
| "loss": 0.2543, |
| "step": 2350 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 4.6542857142857145e-06, |
| "loss": 0.255, |
| "step": 2375 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 4.5828571428571435e-06, |
| "loss": 0.2366, |
| "step": 2400 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 4.511428571428572e-06, |
| "loss": 0.2674, |
| "step": 2425 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 4.440000000000001e-06, |
| "loss": 0.2538, |
| "step": 2450 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 4.368571428571429e-06, |
| "loss": 0.2612, |
| "step": 2475 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 4.297142857142858e-06, |
| "loss": 0.2386, |
| "step": 2500 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 4.225714285714286e-06, |
| "loss": 0.2496, |
| "step": 2525 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 4.154285714285714e-06, |
| "loss": 0.273, |
| "step": 2550 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 4.082857142857143e-06, |
| "loss": 0.2424, |
| "step": 2575 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 4.011428571428571e-06, |
| "loss": 0.2719, |
| "step": 2600 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 3.94e-06, |
| "loss": 0.2584, |
| "step": 2625 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 3.8685714285714286e-06, |
| "loss": 0.27, |
| "step": 2650 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 3.7971428571428576e-06, |
| "loss": 0.2698, |
| "step": 2675 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 3.7257142857142857e-06, |
| "loss": 0.2584, |
| "step": 2700 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 3.6542857142857148e-06, |
| "loss": 0.2359, |
| "step": 2725 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 3.582857142857143e-06, |
| "loss": 0.2294, |
| "step": 2750 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 3.511428571428572e-06, |
| "loss": 0.267, |
| "step": 2775 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 3.44e-06, |
| "loss": 0.233, |
| "step": 2800 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 3.3685714285714287e-06, |
| "loss": 0.256, |
| "step": 2825 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 3.2971428571428577e-06, |
| "loss": 0.2619, |
| "step": 2850 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 3.225714285714286e-06, |
| "loss": 0.2632, |
| "step": 2875 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 3.154285714285715e-06, |
| "loss": 0.2671, |
| "step": 2900 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 3.082857142857143e-06, |
| "loss": 0.2621, |
| "step": 2925 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 3.0114285714285716e-06, |
| "loss": 0.2423, |
| "step": 2950 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 2.9400000000000002e-06, |
| "loss": 0.2379, |
| "step": 2975 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 2.868571428571429e-06, |
| "loss": 0.237, |
| "step": 3000 |
| }, |
| { |
| "epoch": 0.05, |
| "eval_loss": 0.34778067469596863, |
| "eval_runtime": 4475.7535, |
| "eval_samples_per_second": 3.658, |
| "eval_steps_per_second": 0.457, |
| "eval_wer": 32.0188747255628, |
| "step": 3000 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 2.797142857142857e-06, |
| "loss": 0.2485, |
| "step": 3025 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 2.725714285714286e-06, |
| "loss": 0.2483, |
| "step": 3050 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 2.654285714285714e-06, |
| "loss": 0.2453, |
| "step": 3075 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 2.582857142857143e-06, |
| "loss": 0.2582, |
| "step": 3100 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 2.5114285714285718e-06, |
| "loss": 0.2556, |
| "step": 3125 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 2.4400000000000004e-06, |
| "loss": 0.2425, |
| "step": 3150 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 2.3685714285714285e-06, |
| "loss": 0.2826, |
| "step": 3175 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 2.297142857142857e-06, |
| "loss": 0.2603, |
| "step": 3200 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 2.2257142857142857e-06, |
| "loss": 0.2592, |
| "step": 3225 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 2.1542857142857147e-06, |
| "loss": 0.2603, |
| "step": 3250 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 2.0828571428571433e-06, |
| "loss": 0.2402, |
| "step": 3275 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 2.0114285714285715e-06, |
| "loss": 0.2401, |
| "step": 3300 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 1.94e-06, |
| "loss": 0.2606, |
| "step": 3325 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 1.8685714285714289e-06, |
| "loss": 0.2367, |
| "step": 3350 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 1.7971428571428572e-06, |
| "loss": 0.2352, |
| "step": 3375 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 1.7257142857142858e-06, |
| "loss": 0.2322, |
| "step": 3400 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 1.6542857142857144e-06, |
| "loss": 0.2629, |
| "step": 3425 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 1.582857142857143e-06, |
| "loss": 0.2329, |
| "step": 3450 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 1.5114285714285714e-06, |
| "loss": 0.2731, |
| "step": 3475 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 1.44e-06, |
| "loss": 0.2523, |
| "step": 3500 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 1.3685714285714286e-06, |
| "loss": 0.2294, |
| "step": 3525 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 1.2971428571428574e-06, |
| "loss": 0.2583, |
| "step": 3550 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 1.2257142857142857e-06, |
| "loss": 0.2407, |
| "step": 3575 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 1.1542857142857143e-06, |
| "loss": 0.2425, |
| "step": 3600 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 1.082857142857143e-06, |
| "loss": 0.2302, |
| "step": 3625 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 1.0114285714285715e-06, |
| "loss": 0.2302, |
| "step": 3650 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 9.400000000000001e-07, |
| "loss": 0.2472, |
| "step": 3675 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 8.685714285714286e-07, |
| "loss": 0.261, |
| "step": 3700 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 7.971428571428572e-07, |
| "loss": 0.2217, |
| "step": 3725 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 7.257142857142857e-07, |
| "loss": 0.2451, |
| "step": 3750 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 6.542857142857144e-07, |
| "loss": 0.2395, |
| "step": 3775 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 5.82857142857143e-07, |
| "loss": 0.2501, |
| "step": 3800 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 5.114285714285714e-07, |
| "loss": 0.2365, |
| "step": 3825 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 4.4e-07, |
| "loss": 0.2612, |
| "step": 3850 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 3.685714285714286e-07, |
| "loss": 0.2492, |
| "step": 3875 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 2.9714285714285715e-07, |
| "loss": 0.2571, |
| "step": 3900 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 2.2571428571428574e-07, |
| "loss": 0.2442, |
| "step": 3925 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 1.542857142857143e-07, |
| "loss": 0.2513, |
| "step": 3950 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 8.285714285714285e-08, |
| "loss": 0.2402, |
| "step": 3975 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 1.142857142857143e-08, |
| "loss": 0.2618, |
| "step": 4000 |
| }, |
| { |
| "epoch": 0.06, |
| "eval_loss": 0.34070348739624023, |
| "eval_runtime": 4483.294, |
| "eval_samples_per_second": 3.652, |
| "eval_steps_per_second": 0.457, |
| "eval_wer": 31.733787724874663, |
| "step": 4000 |
| } |
| ], |
| "max_steps": 4000, |
| "num_train_epochs": 1, |
| "total_flos": 1.846946562048e+19, |
| "trial_name": null, |
| "trial_params": null |
| } |
|
|
