| { | |
| "best_metric": 43.233499722684414, | |
| "best_model_checkpoint": "./whisper-tiny-it-3/checkpoint-4000", | |
| "epoch": 3.816793893129771, | |
| "global_step": 4000, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 4.0000000000000003e-07, | |
| "loss": 3.8, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.000000000000001e-07, | |
| "loss": 3.277, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.4000000000000001e-06, | |
| "loss": 2.4419, | |
| "step": 75 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.9000000000000002e-06, | |
| "loss": 1.7858, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 2.4000000000000003e-06, | |
| "loss": 1.2439, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 2.9e-06, | |
| "loss": 0.9568, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 3.4000000000000005e-06, | |
| "loss": 0.9208, | |
| "step": 175 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 3.900000000000001e-06, | |
| "loss": 0.8419, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 4.4e-06, | |
| "loss": 0.7544, | |
| "step": 225 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 4.9000000000000005e-06, | |
| "loss": 0.7858, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 5.400000000000001e-06, | |
| "loss": 0.7416, | |
| "step": 275 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 5.9e-06, | |
| "loss": 0.7453, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 6.4000000000000006e-06, | |
| "loss": 0.7462, | |
| "step": 325 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 6.9e-06, | |
| "loss": 0.7031, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 7.4e-06, | |
| "loss": 0.737, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 7.9e-06, | |
| "loss": 0.6862, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 8.400000000000001e-06, | |
| "loss": 0.7026, | |
| "step": 425 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 8.900000000000001e-06, | |
| "loss": 0.6702, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 9.4e-06, | |
| "loss": 0.6382, | |
| "step": 475 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 9.9e-06, | |
| "loss": 0.669, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 9.942857142857145e-06, | |
| "loss": 0.6747, | |
| "step": 525 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.871428571428572e-06, | |
| "loss": 0.6479, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 9.800000000000001e-06, | |
| "loss": 0.6645, | |
| "step": 575 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 9.72857142857143e-06, | |
| "loss": 0.6539, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 9.657142857142859e-06, | |
| "loss": 0.6598, | |
| "step": 625 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 9.585714285714286e-06, | |
| "loss": 0.6486, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 9.514285714285715e-06, | |
| "loss": 0.5952, | |
| "step": 675 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 9.442857142857144e-06, | |
| "loss": 0.6305, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 9.371428571428572e-06, | |
| "loss": 0.6345, | |
| "step": 725 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 9.3e-06, | |
| "loss": 0.5889, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 9.22857142857143e-06, | |
| "loss": 0.6237, | |
| "step": 775 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 9.157142857142857e-06, | |
| "loss": 0.587, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 9.085714285714286e-06, | |
| "loss": 0.5985, | |
| "step": 825 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 9.014285714285715e-06, | |
| "loss": 0.6102, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 8.942857142857142e-06, | |
| "loss": 0.5944, | |
| "step": 875 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 8.871428571428571e-06, | |
| "loss": 0.6003, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 8.8e-06, | |
| "loss": 0.6335, | |
| "step": 925 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 8.72857142857143e-06, | |
| "loss": 0.5661, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 8.657142857142858e-06, | |
| "loss": 0.5851, | |
| "step": 975 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 8.585714285714286e-06, | |
| "loss": 0.5837, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "eval_loss": 0.790373682975769, | |
| "eval_runtime": 269.4198, | |
| "eval_samples_per_second": 5.568, | |
| "eval_steps_per_second": 0.698, | |
| "eval_wer": 50.29811425402107, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 8.514285714285715e-06, | |
| "loss": 0.6186, | |
| "step": 1025 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 8.442857142857144e-06, | |
| "loss": 0.5641, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 8.371428571428573e-06, | |
| "loss": 0.4524, | |
| "step": 1075 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 8.3e-06, | |
| "loss": 0.4612, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 8.22857142857143e-06, | |
| "loss": 0.4643, | |
| "step": 1125 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 8.157142857142858e-06, | |
| "loss": 0.4931, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 8.085714285714287e-06, | |
| "loss": 0.4509, | |
| "step": 1175 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 8.014285714285715e-06, | |
| "loss": 0.4374, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 7.942857142857144e-06, | |
| "loss": 0.466, | |
| "step": 1225 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 7.871428571428573e-06, | |
| "loss": 0.4515, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 7.800000000000002e-06, | |
| "loss": 0.4551, | |
| "step": 1275 | |
| }, | |
| { | |
| "epoch": 1.24, | |
| "learning_rate": 7.72857142857143e-06, | |
| "loss": 0.45, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 1.26, | |
| "learning_rate": 7.657142857142858e-06, | |
| "loss": 0.4336, | |
| "step": 1325 | |
| }, | |
| { | |
| "epoch": 1.29, | |
| "learning_rate": 7.585714285714286e-06, | |
| "loss": 0.4587, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 1.31, | |
| "learning_rate": 7.514285714285715e-06, | |
| "loss": 0.4315, | |
| "step": 1375 | |
| }, | |
| { | |
| "epoch": 1.34, | |
| "learning_rate": 7.442857142857144e-06, | |
| "loss": 0.4806, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 1.36, | |
| "learning_rate": 7.371428571428571e-06, | |
| "loss": 0.4556, | |
| "step": 1425 | |
| }, | |
| { | |
| "epoch": 1.38, | |
| "learning_rate": 7.3e-06, | |
| "loss": 0.4247, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 1.41, | |
| "learning_rate": 7.2285714285714294e-06, | |
| "loss": 0.4582, | |
| "step": 1475 | |
| }, | |
| { | |
| "epoch": 1.43, | |
| "learning_rate": 7.1571428571428584e-06, | |
| "loss": 0.462, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 1.46, | |
| "learning_rate": 7.085714285714286e-06, | |
| "loss": 0.4532, | |
| "step": 1525 | |
| }, | |
| { | |
| "epoch": 1.48, | |
| "learning_rate": 7.014285714285715e-06, | |
| "loss": 0.482, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 1.5, | |
| "learning_rate": 6.942857142857144e-06, | |
| "loss": 0.4561, | |
| "step": 1575 | |
| }, | |
| { | |
| "epoch": 1.53, | |
| "learning_rate": 6.871428571428572e-06, | |
| "loss": 0.4214, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 1.55, | |
| "learning_rate": 6.800000000000001e-06, | |
| "loss": 0.4621, | |
| "step": 1625 | |
| }, | |
| { | |
| "epoch": 1.57, | |
| "learning_rate": 6.728571428571429e-06, | |
| "loss": 0.4175, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 1.6, | |
| "learning_rate": 6.657142857142857e-06, | |
| "loss": 0.4539, | |
| "step": 1675 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "learning_rate": 6.585714285714286e-06, | |
| "loss": 0.4262, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 1.65, | |
| "learning_rate": 6.514285714285715e-06, | |
| "loss": 0.4304, | |
| "step": 1725 | |
| }, | |
| { | |
| "epoch": 1.67, | |
| "learning_rate": 6.442857142857143e-06, | |
| "loss": 0.425, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 1.69, | |
| "learning_rate": 6.371428571428572e-06, | |
| "loss": 0.4332, | |
| "step": 1775 | |
| }, | |
| { | |
| "epoch": 1.72, | |
| "learning_rate": 6.300000000000001e-06, | |
| "loss": 0.4386, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 1.74, | |
| "learning_rate": 6.22857142857143e-06, | |
| "loss": 0.4311, | |
| "step": 1825 | |
| }, | |
| { | |
| "epoch": 1.77, | |
| "learning_rate": 6.157142857142858e-06, | |
| "loss": 0.3991, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 1.79, | |
| "learning_rate": 6.085714285714286e-06, | |
| "loss": 0.4074, | |
| "step": 1875 | |
| }, | |
| { | |
| "epoch": 1.81, | |
| "learning_rate": 6.014285714285715e-06, | |
| "loss": 0.4405, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 1.84, | |
| "learning_rate": 5.942857142857143e-06, | |
| "loss": 0.4205, | |
| "step": 1925 | |
| }, | |
| { | |
| "epoch": 1.86, | |
| "learning_rate": 5.871428571428572e-06, | |
| "loss": 0.4327, | |
| "step": 1950 | |
| }, | |
| { | |
| "epoch": 1.88, | |
| "learning_rate": 5.8e-06, | |
| "loss": 0.4209, | |
| "step": 1975 | |
| }, | |
| { | |
| "epoch": 1.91, | |
| "learning_rate": 5.7285714285714285e-06, | |
| "loss": 0.4183, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 1.91, | |
| "eval_loss": 0.7300998568534851, | |
| "eval_runtime": 262.8118, | |
| "eval_samples_per_second": 5.708, | |
| "eval_steps_per_second": 0.715, | |
| "eval_wer": 45.41735995562951, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 1.93, | |
| "learning_rate": 5.6571428571428576e-06, | |
| "loss": 0.4236, | |
| "step": 2025 | |
| }, | |
| { | |
| "epoch": 1.96, | |
| "learning_rate": 5.5857142857142866e-06, | |
| "loss": 0.3987, | |
| "step": 2050 | |
| }, | |
| { | |
| "epoch": 1.98, | |
| "learning_rate": 5.514285714285714e-06, | |
| "loss": 0.4259, | |
| "step": 2075 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 5.442857142857143e-06, | |
| "loss": 0.3942, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 5.371428571428572e-06, | |
| "loss": 0.3139, | |
| "step": 2125 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 5.300000000000001e-06, | |
| "loss": 0.3156, | |
| "step": 2150 | |
| }, | |
| { | |
| "epoch": 2.08, | |
| "learning_rate": 5.22857142857143e-06, | |
| "loss": 0.3385, | |
| "step": 2175 | |
| }, | |
| { | |
| "epoch": 2.1, | |
| "learning_rate": 5.157142857142857e-06, | |
| "loss": 0.3217, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 2.12, | |
| "learning_rate": 5.085714285714286e-06, | |
| "loss": 0.3263, | |
| "step": 2225 | |
| }, | |
| { | |
| "epoch": 2.15, | |
| "learning_rate": 5.014285714285715e-06, | |
| "loss": 0.3189, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 2.17, | |
| "learning_rate": 4.9428571428571435e-06, | |
| "loss": 0.3441, | |
| "step": 2275 | |
| }, | |
| { | |
| "epoch": 2.19, | |
| "learning_rate": 4.871428571428572e-06, | |
| "loss": 0.338, | |
| "step": 2300 | |
| }, | |
| { | |
| "epoch": 2.22, | |
| "learning_rate": 4.800000000000001e-06, | |
| "loss": 0.332, | |
| "step": 2325 | |
| }, | |
| { | |
| "epoch": 2.24, | |
| "learning_rate": 4.728571428571429e-06, | |
| "loss": 0.3465, | |
| "step": 2350 | |
| }, | |
| { | |
| "epoch": 2.27, | |
| "learning_rate": 4.657142857142857e-06, | |
| "loss": 0.332, | |
| "step": 2375 | |
| }, | |
| { | |
| "epoch": 2.29, | |
| "learning_rate": 4.585714285714286e-06, | |
| "loss": 0.3374, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 2.31, | |
| "learning_rate": 4.514285714285714e-06, | |
| "loss": 0.3124, | |
| "step": 2425 | |
| }, | |
| { | |
| "epoch": 2.34, | |
| "learning_rate": 4.442857142857143e-06, | |
| "loss": 0.3238, | |
| "step": 2450 | |
| }, | |
| { | |
| "epoch": 2.36, | |
| "learning_rate": 4.371428571428572e-06, | |
| "loss": 0.3116, | |
| "step": 2475 | |
| }, | |
| { | |
| "epoch": 2.39, | |
| "learning_rate": 4.3e-06, | |
| "loss": 0.3416, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 2.41, | |
| "learning_rate": 4.228571428571429e-06, | |
| "loss": 0.3302, | |
| "step": 2525 | |
| }, | |
| { | |
| "epoch": 2.43, | |
| "learning_rate": 4.1571428571428575e-06, | |
| "loss": 0.326, | |
| "step": 2550 | |
| }, | |
| { | |
| "epoch": 2.46, | |
| "learning_rate": 4.0857142857142865e-06, | |
| "loss": 0.3306, | |
| "step": 2575 | |
| }, | |
| { | |
| "epoch": 2.48, | |
| "learning_rate": 4.014285714285715e-06, | |
| "loss": 0.3355, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 2.5, | |
| "learning_rate": 3.942857142857143e-06, | |
| "loss": 0.3314, | |
| "step": 2625 | |
| }, | |
| { | |
| "epoch": 2.53, | |
| "learning_rate": 3.871428571428572e-06, | |
| "loss": 0.3286, | |
| "step": 2650 | |
| }, | |
| { | |
| "epoch": 2.55, | |
| "learning_rate": 3.8000000000000005e-06, | |
| "loss": 0.3295, | |
| "step": 2675 | |
| }, | |
| { | |
| "epoch": 2.58, | |
| "learning_rate": 3.7285714285714286e-06, | |
| "loss": 0.3247, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 2.6, | |
| "learning_rate": 3.6571428571428576e-06, | |
| "loss": 0.3185, | |
| "step": 2725 | |
| }, | |
| { | |
| "epoch": 2.62, | |
| "learning_rate": 3.5857142857142862e-06, | |
| "loss": 0.33, | |
| "step": 2750 | |
| }, | |
| { | |
| "epoch": 2.65, | |
| "learning_rate": 3.5142857142857144e-06, | |
| "loss": 0.3317, | |
| "step": 2775 | |
| }, | |
| { | |
| "epoch": 2.67, | |
| "learning_rate": 3.4428571428571434e-06, | |
| "loss": 0.3166, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 2.7, | |
| "learning_rate": 3.3714285714285716e-06, | |
| "loss": 0.3162, | |
| "step": 2825 | |
| }, | |
| { | |
| "epoch": 2.72, | |
| "learning_rate": 3.3000000000000006e-06, | |
| "loss": 0.3383, | |
| "step": 2850 | |
| }, | |
| { | |
| "epoch": 2.74, | |
| "learning_rate": 3.2285714285714288e-06, | |
| "loss": 0.3029, | |
| "step": 2875 | |
| }, | |
| { | |
| "epoch": 2.77, | |
| "learning_rate": 3.1571428571428573e-06, | |
| "loss": 0.3287, | |
| "step": 2900 | |
| }, | |
| { | |
| "epoch": 2.79, | |
| "learning_rate": 3.085714285714286e-06, | |
| "loss": 0.3265, | |
| "step": 2925 | |
| }, | |
| { | |
| "epoch": 2.81, | |
| "learning_rate": 3.0142857142857145e-06, | |
| "loss": 0.3415, | |
| "step": 2950 | |
| }, | |
| { | |
| "epoch": 2.84, | |
| "learning_rate": 2.9428571428571427e-06, | |
| "loss": 0.3287, | |
| "step": 2975 | |
| }, | |
| { | |
| "epoch": 2.86, | |
| "learning_rate": 2.8714285714285717e-06, | |
| "loss": 0.3147, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 2.86, | |
| "eval_loss": 0.7131519913673401, | |
| "eval_runtime": 268.7699, | |
| "eval_samples_per_second": 5.581, | |
| "eval_steps_per_second": 0.699, | |
| "eval_wer": 44.315030504714365, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 2.89, | |
| "learning_rate": 2.8000000000000003e-06, | |
| "loss": 0.3319, | |
| "step": 3025 | |
| }, | |
| { | |
| "epoch": 2.91, | |
| "learning_rate": 2.728571428571429e-06, | |
| "loss": 0.3273, | |
| "step": 3050 | |
| }, | |
| { | |
| "epoch": 2.93, | |
| "learning_rate": 2.6571428571428575e-06, | |
| "loss": 0.3221, | |
| "step": 3075 | |
| }, | |
| { | |
| "epoch": 2.96, | |
| "learning_rate": 2.5857142857142856e-06, | |
| "loss": 0.3347, | |
| "step": 3100 | |
| }, | |
| { | |
| "epoch": 2.98, | |
| "learning_rate": 2.5142857142857147e-06, | |
| "loss": 0.313, | |
| "step": 3125 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "learning_rate": 2.442857142857143e-06, | |
| "loss": 0.3172, | |
| "step": 3150 | |
| }, | |
| { | |
| "epoch": 3.03, | |
| "learning_rate": 2.371428571428572e-06, | |
| "loss": 0.2681, | |
| "step": 3175 | |
| }, | |
| { | |
| "epoch": 3.05, | |
| "learning_rate": 2.3000000000000004e-06, | |
| "loss": 0.2608, | |
| "step": 3200 | |
| }, | |
| { | |
| "epoch": 3.08, | |
| "learning_rate": 2.228571428571429e-06, | |
| "loss": 0.2605, | |
| "step": 3225 | |
| }, | |
| { | |
| "epoch": 3.1, | |
| "learning_rate": 2.157142857142857e-06, | |
| "loss": 0.2785, | |
| "step": 3250 | |
| }, | |
| { | |
| "epoch": 3.12, | |
| "learning_rate": 2.0857142857142858e-06, | |
| "loss": 0.2652, | |
| "step": 3275 | |
| }, | |
| { | |
| "epoch": 3.15, | |
| "learning_rate": 2.0142857142857144e-06, | |
| "loss": 0.283, | |
| "step": 3300 | |
| }, | |
| { | |
| "epoch": 3.17, | |
| "learning_rate": 1.942857142857143e-06, | |
| "loss": 0.268, | |
| "step": 3325 | |
| }, | |
| { | |
| "epoch": 3.2, | |
| "learning_rate": 1.8714285714285715e-06, | |
| "loss": 0.2584, | |
| "step": 3350 | |
| }, | |
| { | |
| "epoch": 3.22, | |
| "learning_rate": 1.8000000000000001e-06, | |
| "loss": 0.2786, | |
| "step": 3375 | |
| }, | |
| { | |
| "epoch": 3.24, | |
| "learning_rate": 1.7285714285714287e-06, | |
| "loss": 0.2733, | |
| "step": 3400 | |
| }, | |
| { | |
| "epoch": 3.27, | |
| "learning_rate": 1.657142857142857e-06, | |
| "loss": 0.26, | |
| "step": 3425 | |
| }, | |
| { | |
| "epoch": 3.29, | |
| "learning_rate": 1.5857142857142857e-06, | |
| "loss": 0.2675, | |
| "step": 3450 | |
| }, | |
| { | |
| "epoch": 3.32, | |
| "learning_rate": 1.5142857142857145e-06, | |
| "loss": 0.2562, | |
| "step": 3475 | |
| }, | |
| { | |
| "epoch": 3.34, | |
| "learning_rate": 1.442857142857143e-06, | |
| "loss": 0.2713, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 3.36, | |
| "learning_rate": 1.3714285714285717e-06, | |
| "loss": 0.2699, | |
| "step": 3525 | |
| }, | |
| { | |
| "epoch": 3.39, | |
| "learning_rate": 1.3e-06, | |
| "loss": 0.2722, | |
| "step": 3550 | |
| }, | |
| { | |
| "epoch": 3.41, | |
| "learning_rate": 1.2285714285714286e-06, | |
| "loss": 0.2802, | |
| "step": 3575 | |
| }, | |
| { | |
| "epoch": 3.44, | |
| "learning_rate": 1.1571428571428572e-06, | |
| "loss": 0.255, | |
| "step": 3600 | |
| }, | |
| { | |
| "epoch": 3.46, | |
| "learning_rate": 1.0857142857142858e-06, | |
| "loss": 0.2795, | |
| "step": 3625 | |
| }, | |
| { | |
| "epoch": 3.48, | |
| "learning_rate": 1.0142857142857144e-06, | |
| "loss": 0.2771, | |
| "step": 3650 | |
| }, | |
| { | |
| "epoch": 3.51, | |
| "learning_rate": 9.42857142857143e-07, | |
| "loss": 0.2728, | |
| "step": 3675 | |
| }, | |
| { | |
| "epoch": 3.53, | |
| "learning_rate": 8.714285714285716e-07, | |
| "loss": 0.2705, | |
| "step": 3700 | |
| }, | |
| { | |
| "epoch": 3.55, | |
| "learning_rate": 8.000000000000001e-07, | |
| "loss": 0.2824, | |
| "step": 3725 | |
| }, | |
| { | |
| "epoch": 3.58, | |
| "learning_rate": 7.285714285714287e-07, | |
| "loss": 0.2668, | |
| "step": 3750 | |
| }, | |
| { | |
| "epoch": 3.6, | |
| "learning_rate": 6.571428571428571e-07, | |
| "loss": 0.2702, | |
| "step": 3775 | |
| }, | |
| { | |
| "epoch": 3.63, | |
| "learning_rate": 5.857142857142857e-07, | |
| "loss": 0.2648, | |
| "step": 3800 | |
| }, | |
| { | |
| "epoch": 3.65, | |
| "learning_rate": 5.142857142857143e-07, | |
| "loss": 0.266, | |
| "step": 3825 | |
| }, | |
| { | |
| "epoch": 3.67, | |
| "learning_rate": 4.4285714285714286e-07, | |
| "loss": 0.2585, | |
| "step": 3850 | |
| }, | |
| { | |
| "epoch": 3.7, | |
| "learning_rate": 3.7142857142857145e-07, | |
| "loss": 0.2738, | |
| "step": 3875 | |
| }, | |
| { | |
| "epoch": 3.72, | |
| "learning_rate": 3.0000000000000004e-07, | |
| "loss": 0.2844, | |
| "step": 3900 | |
| }, | |
| { | |
| "epoch": 3.75, | |
| "learning_rate": 2.285714285714286e-07, | |
| "loss": 0.2605, | |
| "step": 3925 | |
| }, | |
| { | |
| "epoch": 3.77, | |
| "learning_rate": 1.5714285714285717e-07, | |
| "loss": 0.2739, | |
| "step": 3950 | |
| }, | |
| { | |
| "epoch": 3.79, | |
| "learning_rate": 8.571428571428573e-08, | |
| "loss": 0.2606, | |
| "step": 3975 | |
| }, | |
| { | |
| "epoch": 3.82, | |
| "learning_rate": 1.4285714285714288e-08, | |
| "loss": 0.267, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 3.82, | |
| "eval_loss": 0.7116732001304626, | |
| "eval_runtime": 265.6181, | |
| "eval_samples_per_second": 5.647, | |
| "eval_steps_per_second": 0.708, | |
| "eval_wer": 43.233499722684414, | |
| "step": 4000 | |
| } | |
| ], | |
| "max_steps": 4000, | |
| "num_train_epochs": 4, | |
| "total_flos": 1.57509165109248e+18, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |