| { |
| "best_metric": 18.47345132743363, |
| "best_model_checkpoint": "logdir/checkpoint-15000", |
| "epoch": 0.812248710555172, |
| "global_step": 20000, |
| "is_hyper_param_search": false, |
| "is_local_process_zero": true, |
| "is_world_process_zero": true, |
| "log_history": [ |
| { |
| "epoch": 0.0, |
| "learning_rate": 3.8e-07, |
| "loss": 15.1656, |
| "step": 25 |
| }, |
| { |
| "epoch": 0.0, |
| "learning_rate": 8.8e-07, |
| "loss": 9.1511, |
| "step": 50 |
| }, |
| { |
| "epoch": 0.0, |
| "learning_rate": 1.3800000000000001e-06, |
| "loss": 4.725, |
| "step": 75 |
| }, |
| { |
| "epoch": 0.0, |
| "learning_rate": 1.8800000000000002e-06, |
| "loss": 2.4974, |
| "step": 100 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 2.38e-06, |
| "loss": 1.6609, |
| "step": 125 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 2.88e-06, |
| "loss": 1.4332, |
| "step": 150 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 3.3800000000000007e-06, |
| "loss": 1.3122, |
| "step": 175 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 3.88e-06, |
| "loss": 1.1897, |
| "step": 200 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 4.38e-06, |
| "loss": 1.0063, |
| "step": 225 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 4.880000000000001e-06, |
| "loss": 0.7627, |
| "step": 250 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 5.380000000000001e-06, |
| "loss": 0.5165, |
| "step": 275 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 5.8800000000000005e-06, |
| "loss": 0.3999, |
| "step": 300 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 6.380000000000001e-06, |
| "loss": 0.3533, |
| "step": 325 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 6.88e-06, |
| "loss": 0.3189, |
| "step": 350 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 7.3800000000000005e-06, |
| "loss": 0.2946, |
| "step": 375 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 7.88e-06, |
| "loss": 0.2715, |
| "step": 400 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 8.380000000000001e-06, |
| "loss": 0.258, |
| "step": 425 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 8.880000000000001e-06, |
| "loss": 0.2434, |
| "step": 450 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 9.38e-06, |
| "loss": 0.233, |
| "step": 475 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 9.88e-06, |
| "loss": 0.2207, |
| "step": 500 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 9.992244897959185e-06, |
| "loss": 0.2159, |
| "step": 525 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 9.982040816326532e-06, |
| "loss": 0.2039, |
| "step": 550 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 9.971836734693878e-06, |
| "loss": 0.1958, |
| "step": 575 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 9.961632653061225e-06, |
| "loss": 0.1885, |
| "step": 600 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 9.951428571428571e-06, |
| "loss": 0.1865, |
| "step": 625 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 9.941224489795919e-06, |
| "loss": 0.1721, |
| "step": 650 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 9.931020408163266e-06, |
| "loss": 0.1744, |
| "step": 675 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 9.920816326530614e-06, |
| "loss": 0.1745, |
| "step": 700 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 9.91061224489796e-06, |
| "loss": 0.1629, |
| "step": 725 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 9.900408163265307e-06, |
| "loss": 0.1628, |
| "step": 750 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 9.890204081632654e-06, |
| "loss": 0.1634, |
| "step": 775 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 9.88e-06, |
| "loss": 0.1573, |
| "step": 800 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 9.869795918367348e-06, |
| "loss": 0.149, |
| "step": 825 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 9.859591836734694e-06, |
| "loss": 0.1517, |
| "step": 850 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 9.849387755102041e-06, |
| "loss": 0.1499, |
| "step": 875 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 9.839183673469389e-06, |
| "loss": 0.1454, |
| "step": 900 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 9.828979591836736e-06, |
| "loss": 0.1467, |
| "step": 925 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 9.818775510204083e-06, |
| "loss": 0.1487, |
| "step": 950 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 9.80857142857143e-06, |
| "loss": 0.1425, |
| "step": 975 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 9.798367346938777e-06, |
| "loss": 0.1427, |
| "step": 1000 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 9.788163265306123e-06, |
| "loss": 0.1389, |
| "step": 1025 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 9.77795918367347e-06, |
| "loss": 0.1343, |
| "step": 1050 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 9.767755102040818e-06, |
| "loss": 0.1366, |
| "step": 1075 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 9.757551020408163e-06, |
| "loss": 0.1339, |
| "step": 1100 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 9.74734693877551e-06, |
| "loss": 0.1323, |
| "step": 1125 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 9.737142857142858e-06, |
| "loss": 0.1289, |
| "step": 1150 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 9.726938775510206e-06, |
| "loss": 0.1288, |
| "step": 1175 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 9.716734693877552e-06, |
| "loss": 0.1278, |
| "step": 1200 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 9.706530612244899e-06, |
| "loss": 0.1237, |
| "step": 1225 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 9.696326530612245e-06, |
| "loss": 0.1242, |
| "step": 1250 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 9.686122448979592e-06, |
| "loss": 0.1277, |
| "step": 1275 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 9.67591836734694e-06, |
| "loss": 0.1211, |
| "step": 1300 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 9.665714285714286e-06, |
| "loss": 0.1229, |
| "step": 1325 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 9.655510204081635e-06, |
| "loss": 0.1167, |
| "step": 1350 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 9.64530612244898e-06, |
| "loss": 0.1229, |
| "step": 1375 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 9.635102040816328e-06, |
| "loss": 0.1151, |
| "step": 1400 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 9.624897959183674e-06, |
| "loss": 0.1164, |
| "step": 1425 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 9.614693877551021e-06, |
| "loss": 0.1184, |
| "step": 1450 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 9.604489795918369e-06, |
| "loss": 0.114, |
| "step": 1475 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 9.594285714285715e-06, |
| "loss": 0.112, |
| "step": 1500 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 9.584081632653062e-06, |
| "loss": 0.1132, |
| "step": 1525 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 9.573877551020408e-06, |
| "loss": 0.1103, |
| "step": 1550 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 9.563673469387757e-06, |
| "loss": 0.1119, |
| "step": 1575 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 9.553469387755103e-06, |
| "loss": 0.1106, |
| "step": 1600 |
| }, |
| { |
| "epoch": 0.07, |
| "learning_rate": 9.54326530612245e-06, |
| "loss": 0.111, |
| "step": 1625 |
| }, |
| { |
| "epoch": 0.07, |
| "learning_rate": 9.533061224489796e-06, |
| "loss": 0.1095, |
| "step": 1650 |
| }, |
| { |
| "epoch": 0.07, |
| "learning_rate": 9.522857142857144e-06, |
| "loss": 0.1091, |
| "step": 1675 |
| }, |
| { |
| "epoch": 0.07, |
| "learning_rate": 9.512653061224491e-06, |
| "loss": 0.1134, |
| "step": 1700 |
| }, |
| { |
| "epoch": 0.07, |
| "learning_rate": 9.502448979591837e-06, |
| "loss": 0.1101, |
| "step": 1725 |
| }, |
| { |
| "epoch": 0.07, |
| "learning_rate": 9.492244897959184e-06, |
| "loss": 0.1065, |
| "step": 1750 |
| }, |
| { |
| "epoch": 0.07, |
| "learning_rate": 9.48204081632653e-06, |
| "loss": 0.1051, |
| "step": 1775 |
| }, |
| { |
| "epoch": 0.07, |
| "learning_rate": 9.47183673469388e-06, |
| "loss": 0.1077, |
| "step": 1800 |
| }, |
| { |
| "epoch": 0.07, |
| "learning_rate": 9.461632653061225e-06, |
| "loss": 0.1052, |
| "step": 1825 |
| }, |
| { |
| "epoch": 0.08, |
| "learning_rate": 9.451428571428573e-06, |
| "loss": 0.1016, |
| "step": 1850 |
| }, |
| { |
| "epoch": 0.08, |
| "learning_rate": 9.44122448979592e-06, |
| "loss": 0.1011, |
| "step": 1875 |
| }, |
| { |
| "epoch": 0.08, |
| "learning_rate": 9.431020408163266e-06, |
| "loss": 0.1022, |
| "step": 1900 |
| }, |
| { |
| "epoch": 0.08, |
| "learning_rate": 9.420816326530613e-06, |
| "loss": 0.1019, |
| "step": 1925 |
| }, |
| { |
| "epoch": 0.08, |
| "learning_rate": 9.41061224489796e-06, |
| "loss": 0.101, |
| "step": 1950 |
| }, |
| { |
| "epoch": 0.08, |
| "learning_rate": 9.400408163265307e-06, |
| "loss": 0.0972, |
| "step": 1975 |
| }, |
| { |
| "epoch": 0.08, |
| "learning_rate": 9.390204081632654e-06, |
| "loss": 0.1039, |
| "step": 2000 |
| }, |
| { |
| "epoch": 0.08, |
| "learning_rate": 9.38e-06, |
| "loss": 0.0956, |
| "step": 2025 |
| }, |
| { |
| "epoch": 0.08, |
| "learning_rate": 9.369795918367347e-06, |
| "loss": 0.0991, |
| "step": 2050 |
| }, |
| { |
| "epoch": 0.08, |
| "learning_rate": 9.359591836734695e-06, |
| "loss": 0.1015, |
| "step": 2075 |
| }, |
| { |
| "epoch": 0.09, |
| "learning_rate": 9.349387755102042e-06, |
| "loss": 0.098, |
| "step": 2100 |
| }, |
| { |
| "epoch": 0.09, |
| "learning_rate": 9.339183673469388e-06, |
| "loss": 0.1011, |
| "step": 2125 |
| }, |
| { |
| "epoch": 0.09, |
| "learning_rate": 9.328979591836736e-06, |
| "loss": 0.0962, |
| "step": 2150 |
| }, |
| { |
| "epoch": 0.09, |
| "learning_rate": 9.318775510204082e-06, |
| "loss": 0.0993, |
| "step": 2175 |
| }, |
| { |
| "epoch": 0.09, |
| "learning_rate": 9.308571428571429e-06, |
| "loss": 0.0938, |
| "step": 2200 |
| }, |
| { |
| "epoch": 0.09, |
| "learning_rate": 9.298367346938776e-06, |
| "loss": 0.0959, |
| "step": 2225 |
| }, |
| { |
| "epoch": 0.09, |
| "learning_rate": 9.288163265306122e-06, |
| "loss": 0.0992, |
| "step": 2250 |
| }, |
| { |
| "epoch": 0.09, |
| "learning_rate": 9.27795918367347e-06, |
| "loss": 0.095, |
| "step": 2275 |
| }, |
| { |
| "epoch": 0.09, |
| "learning_rate": 9.267755102040817e-06, |
| "loss": 0.0939, |
| "step": 2300 |
| }, |
| { |
| "epoch": 0.09, |
| "learning_rate": 9.257551020408165e-06, |
| "loss": 0.0952, |
| "step": 2325 |
| }, |
| { |
| "epoch": 0.1, |
| "learning_rate": 9.24734693877551e-06, |
| "loss": 0.0952, |
| "step": 2350 |
| }, |
| { |
| "epoch": 0.1, |
| "learning_rate": 9.237142857142858e-06, |
| "loss": 0.0934, |
| "step": 2375 |
| }, |
| { |
| "epoch": 0.1, |
| "learning_rate": 9.226938775510205e-06, |
| "loss": 0.0952, |
| "step": 2400 |
| }, |
| { |
| "epoch": 0.1, |
| "learning_rate": 9.216734693877551e-06, |
| "loss": 0.0908, |
| "step": 2425 |
| }, |
| { |
| "epoch": 0.1, |
| "learning_rate": 9.206530612244899e-06, |
| "loss": 0.0864, |
| "step": 2450 |
| }, |
| { |
| "epoch": 0.1, |
| "learning_rate": 9.196326530612245e-06, |
| "loss": 0.093, |
| "step": 2475 |
| }, |
| { |
| "epoch": 0.1, |
| "learning_rate": 9.186122448979594e-06, |
| "loss": 0.0949, |
| "step": 2500 |
| }, |
| { |
| "epoch": 0.1, |
| "learning_rate": 9.17591836734694e-06, |
| "loss": 0.0891, |
| "step": 2525 |
| }, |
| { |
| "epoch": 0.1, |
| "learning_rate": 9.165714285714287e-06, |
| "loss": 0.0913, |
| "step": 2550 |
| }, |
| { |
| "epoch": 0.1, |
| "learning_rate": 9.155510204081633e-06, |
| "loss": 0.0884, |
| "step": 2575 |
| }, |
| { |
| "epoch": 0.11, |
| "learning_rate": 9.14530612244898e-06, |
| "loss": 0.0889, |
| "step": 2600 |
| }, |
| { |
| "epoch": 0.11, |
| "learning_rate": 9.135102040816328e-06, |
| "loss": 0.0909, |
| "step": 2625 |
| }, |
| { |
| "epoch": 0.11, |
| "learning_rate": 9.124897959183674e-06, |
| "loss": 0.0832, |
| "step": 2650 |
| }, |
| { |
| "epoch": 0.11, |
| "learning_rate": 9.114693877551021e-06, |
| "loss": 0.0838, |
| "step": 2675 |
| }, |
| { |
| "epoch": 0.11, |
| "learning_rate": 9.104489795918367e-06, |
| "loss": 0.0852, |
| "step": 2700 |
| }, |
| { |
| "epoch": 0.11, |
| "learning_rate": 9.094285714285716e-06, |
| "loss": 0.0866, |
| "step": 2725 |
| }, |
| { |
| "epoch": 0.11, |
| "learning_rate": 9.084081632653062e-06, |
| "loss": 0.0869, |
| "step": 2750 |
| }, |
| { |
| "epoch": 0.11, |
| "learning_rate": 9.07387755102041e-06, |
| "loss": 0.0876, |
| "step": 2775 |
| }, |
| { |
| "epoch": 0.11, |
| "learning_rate": 9.063673469387755e-06, |
| "loss": 0.086, |
| "step": 2800 |
| }, |
| { |
| "epoch": 0.11, |
| "learning_rate": 9.053469387755103e-06, |
| "loss": 0.0854, |
| "step": 2825 |
| }, |
| { |
| "epoch": 0.12, |
| "learning_rate": 9.04326530612245e-06, |
| "loss": 0.0841, |
| "step": 2850 |
| }, |
| { |
| "epoch": 0.12, |
| "learning_rate": 9.033061224489796e-06, |
| "loss": 0.0839, |
| "step": 2875 |
| }, |
| { |
| "epoch": 0.12, |
| "learning_rate": 9.022857142857143e-06, |
| "loss": 0.0805, |
| "step": 2900 |
| }, |
| { |
| "epoch": 0.12, |
| "learning_rate": 9.01265306122449e-06, |
| "loss": 0.0855, |
| "step": 2925 |
| }, |
| { |
| "epoch": 0.12, |
| "learning_rate": 9.002448979591838e-06, |
| "loss": 0.0865, |
| "step": 2950 |
| }, |
| { |
| "epoch": 0.12, |
| "learning_rate": 8.992244897959184e-06, |
| "loss": 0.0881, |
| "step": 2975 |
| }, |
| { |
| "epoch": 0.12, |
| "learning_rate": 8.982040816326532e-06, |
| "loss": 0.088, |
| "step": 3000 |
| }, |
| { |
| "epoch": 0.12, |
| "learning_rate": 8.971836734693879e-06, |
| "loss": 0.084, |
| "step": 3025 |
| }, |
| { |
| "epoch": 0.12, |
| "learning_rate": 8.961632653061225e-06, |
| "loss": 0.0846, |
| "step": 3050 |
| }, |
| { |
| "epoch": 0.12, |
| "learning_rate": 8.951428571428572e-06, |
| "loss": 0.0835, |
| "step": 3075 |
| }, |
| { |
| "epoch": 0.13, |
| "learning_rate": 8.941224489795918e-06, |
| "loss": 0.0817, |
| "step": 3100 |
| }, |
| { |
| "epoch": 0.13, |
| "learning_rate": 8.931020408163266e-06, |
| "loss": 0.0846, |
| "step": 3125 |
| }, |
| { |
| "epoch": 0.13, |
| "learning_rate": 8.920816326530613e-06, |
| "loss": 0.0798, |
| "step": 3150 |
| }, |
| { |
| "epoch": 0.13, |
| "learning_rate": 8.91061224489796e-06, |
| "loss": 0.0787, |
| "step": 3175 |
| }, |
| { |
| "epoch": 0.13, |
| "learning_rate": 8.900408163265306e-06, |
| "loss": 0.0825, |
| "step": 3200 |
| }, |
| { |
| "epoch": 0.13, |
| "learning_rate": 8.890204081632654e-06, |
| "loss": 0.0813, |
| "step": 3225 |
| }, |
| { |
| "epoch": 0.13, |
| "learning_rate": 8.880000000000001e-06, |
| "loss": 0.0839, |
| "step": 3250 |
| }, |
| { |
| "epoch": 0.13, |
| "learning_rate": 8.869795918367347e-06, |
| "loss": 0.0833, |
| "step": 3275 |
| }, |
| { |
| "epoch": 0.13, |
| "learning_rate": 8.859591836734695e-06, |
| "loss": 0.0807, |
| "step": 3300 |
| }, |
| { |
| "epoch": 0.14, |
| "learning_rate": 8.84938775510204e-06, |
| "loss": 0.0801, |
| "step": 3325 |
| }, |
| { |
| "epoch": 0.14, |
| "learning_rate": 8.839183673469388e-06, |
| "loss": 0.0807, |
| "step": 3350 |
| }, |
| { |
| "epoch": 0.14, |
| "learning_rate": 8.828979591836735e-06, |
| "loss": 0.0792, |
| "step": 3375 |
| }, |
| { |
| "epoch": 0.14, |
| "learning_rate": 8.818775510204083e-06, |
| "loss": 0.0822, |
| "step": 3400 |
| }, |
| { |
| "epoch": 0.14, |
| "learning_rate": 8.80857142857143e-06, |
| "loss": 0.0788, |
| "step": 3425 |
| }, |
| { |
| "epoch": 0.14, |
| "learning_rate": 8.798367346938776e-06, |
| "loss": 0.0774, |
| "step": 3450 |
| }, |
| { |
| "epoch": 0.14, |
| "learning_rate": 8.788163265306124e-06, |
| "loss": 0.0786, |
| "step": 3475 |
| }, |
| { |
| "epoch": 0.14, |
| "learning_rate": 8.77795918367347e-06, |
| "loss": 0.0826, |
| "step": 3500 |
| }, |
| { |
| "epoch": 0.14, |
| "learning_rate": 8.767755102040817e-06, |
| "loss": 0.0774, |
| "step": 3525 |
| }, |
| { |
| "epoch": 0.14, |
| "learning_rate": 8.757551020408164e-06, |
| "loss": 0.0808, |
| "step": 3550 |
| }, |
| { |
| "epoch": 0.15, |
| "learning_rate": 8.74734693877551e-06, |
| "loss": 0.0799, |
| "step": 3575 |
| }, |
| { |
| "epoch": 0.15, |
| "learning_rate": 8.737142857142858e-06, |
| "loss": 0.0768, |
| "step": 3600 |
| }, |
| { |
| "epoch": 0.15, |
| "learning_rate": 8.726938775510205e-06, |
| "loss": 0.0752, |
| "step": 3625 |
| }, |
| { |
| "epoch": 0.15, |
| "learning_rate": 8.716734693877553e-06, |
| "loss": 0.0763, |
| "step": 3650 |
| }, |
| { |
| "epoch": 0.15, |
| "learning_rate": 8.706530612244898e-06, |
| "loss": 0.078, |
| "step": 3675 |
| }, |
| { |
| "epoch": 0.15, |
| "learning_rate": 8.696326530612246e-06, |
| "loss": 0.0766, |
| "step": 3700 |
| }, |
| { |
| "epoch": 0.15, |
| "learning_rate": 8.686122448979592e-06, |
| "loss": 0.0771, |
| "step": 3725 |
| }, |
| { |
| "epoch": 0.15, |
| "learning_rate": 8.67591836734694e-06, |
| "loss": 0.0757, |
| "step": 3750 |
| }, |
| { |
| "epoch": 0.15, |
| "learning_rate": 8.665714285714287e-06, |
| "loss": 0.0751, |
| "step": 3775 |
| }, |
| { |
| "epoch": 0.15, |
| "learning_rate": 8.655510204081633e-06, |
| "loss": 0.0742, |
| "step": 3800 |
| }, |
| { |
| "epoch": 0.16, |
| "learning_rate": 8.64530612244898e-06, |
| "loss": 0.073, |
| "step": 3825 |
| }, |
| { |
| "epoch": 0.16, |
| "learning_rate": 8.635102040816327e-06, |
| "loss": 0.0755, |
| "step": 3850 |
| }, |
| { |
| "epoch": 0.16, |
| "learning_rate": 8.624897959183675e-06, |
| "loss": 0.0762, |
| "step": 3875 |
| }, |
| { |
| "epoch": 0.16, |
| "learning_rate": 8.61469387755102e-06, |
| "loss": 0.078, |
| "step": 3900 |
| }, |
| { |
| "epoch": 0.16, |
| "learning_rate": 8.604489795918368e-06, |
| "loss": 0.0733, |
| "step": 3925 |
| }, |
| { |
| "epoch": 0.16, |
| "learning_rate": 8.594285714285716e-06, |
| "loss": 0.0751, |
| "step": 3950 |
| }, |
| { |
| "epoch": 0.16, |
| "learning_rate": 8.584081632653062e-06, |
| "loss": 0.0745, |
| "step": 3975 |
| }, |
| { |
| "epoch": 0.16, |
| "learning_rate": 8.573877551020409e-06, |
| "loss": 0.0743, |
| "step": 4000 |
| }, |
| { |
| "epoch": 0.16, |
| "learning_rate": 8.563673469387755e-06, |
| "loss": 0.0746, |
| "step": 4025 |
| }, |
| { |
| "epoch": 0.16, |
| "learning_rate": 8.553469387755102e-06, |
| "loss": 0.0743, |
| "step": 4050 |
| }, |
| { |
| "epoch": 0.17, |
| "learning_rate": 8.54326530612245e-06, |
| "loss": 0.0729, |
| "step": 4075 |
| }, |
| { |
| "epoch": 0.17, |
| "learning_rate": 8.533061224489797e-06, |
| "loss": 0.0731, |
| "step": 4100 |
| }, |
| { |
| "epoch": 0.17, |
| "learning_rate": 8.522857142857143e-06, |
| "loss": 0.0719, |
| "step": 4125 |
| }, |
| { |
| "epoch": 0.17, |
| "learning_rate": 8.51265306122449e-06, |
| "loss": 0.0771, |
| "step": 4150 |
| }, |
| { |
| "epoch": 0.17, |
| "learning_rate": 8.502448979591838e-06, |
| "loss": 0.0762, |
| "step": 4175 |
| }, |
| { |
| "epoch": 0.17, |
| "learning_rate": 8.492244897959184e-06, |
| "loss": 0.0717, |
| "step": 4200 |
| }, |
| { |
| "epoch": 0.17, |
| "learning_rate": 8.482040816326531e-06, |
| "loss": 0.0711, |
| "step": 4225 |
| }, |
| { |
| "epoch": 0.17, |
| "learning_rate": 8.471836734693877e-06, |
| "loss": 0.0719, |
| "step": 4250 |
| }, |
| { |
| "epoch": 0.17, |
| "learning_rate": 8.461632653061225e-06, |
| "loss": 0.0723, |
| "step": 4275 |
| }, |
| { |
| "epoch": 0.17, |
| "learning_rate": 8.451428571428572e-06, |
| "loss": 0.0738, |
| "step": 4300 |
| }, |
| { |
| "epoch": 0.18, |
| "learning_rate": 8.44122448979592e-06, |
| "loss": 0.0716, |
| "step": 4325 |
| }, |
| { |
| "epoch": 0.18, |
| "learning_rate": 8.431020408163267e-06, |
| "loss": 0.0719, |
| "step": 4350 |
| }, |
| { |
| "epoch": 0.18, |
| "learning_rate": 8.420816326530613e-06, |
| "loss": 0.0733, |
| "step": 4375 |
| }, |
| { |
| "epoch": 0.18, |
| "learning_rate": 8.41061224489796e-06, |
| "loss": 0.0697, |
| "step": 4400 |
| }, |
| { |
| "epoch": 0.18, |
| "learning_rate": 8.400408163265306e-06, |
| "loss": 0.0743, |
| "step": 4425 |
| }, |
| { |
| "epoch": 0.18, |
| "learning_rate": 8.390204081632654e-06, |
| "loss": 0.0714, |
| "step": 4450 |
| }, |
| { |
| "epoch": 0.18, |
| "learning_rate": 8.380000000000001e-06, |
| "loss": 0.0673, |
| "step": 4475 |
| }, |
| { |
| "epoch": 0.18, |
| "learning_rate": 8.369795918367347e-06, |
| "loss": 0.0705, |
| "step": 4500 |
| }, |
| { |
| "epoch": 0.18, |
| "learning_rate": 8.359591836734694e-06, |
| "loss": 0.0682, |
| "step": 4525 |
| }, |
| { |
| "epoch": 0.18, |
| "learning_rate": 8.349387755102042e-06, |
| "loss": 0.0702, |
| "step": 4550 |
| }, |
| { |
| "epoch": 0.19, |
| "learning_rate": 8.33918367346939e-06, |
| "loss": 0.0713, |
| "step": 4575 |
| }, |
| { |
| "epoch": 0.19, |
| "learning_rate": 8.328979591836735e-06, |
| "loss": 0.0705, |
| "step": 4600 |
| }, |
| { |
| "epoch": 0.19, |
| "learning_rate": 8.318775510204083e-06, |
| "loss": 0.0724, |
| "step": 4625 |
| }, |
| { |
| "epoch": 0.19, |
| "learning_rate": 8.308571428571428e-06, |
| "loss": 0.0663, |
| "step": 4650 |
| }, |
| { |
| "epoch": 0.19, |
| "learning_rate": 8.298367346938776e-06, |
| "loss": 0.0666, |
| "step": 4675 |
| }, |
| { |
| "epoch": 0.19, |
| "learning_rate": 8.288163265306123e-06, |
| "loss": 0.0672, |
| "step": 4700 |
| }, |
| { |
| "epoch": 0.19, |
| "learning_rate": 8.27795918367347e-06, |
| "loss": 0.0697, |
| "step": 4725 |
| }, |
| { |
| "epoch": 0.19, |
| "learning_rate": 8.267755102040818e-06, |
| "loss": 0.0663, |
| "step": 4750 |
| }, |
| { |
| "epoch": 0.19, |
| "learning_rate": 8.257551020408164e-06, |
| "loss": 0.068, |
| "step": 4775 |
| }, |
| { |
| "epoch": 0.19, |
| "learning_rate": 8.247346938775512e-06, |
| "loss": 0.0715, |
| "step": 4800 |
| }, |
| { |
| "epoch": 0.2, |
| "learning_rate": 8.237142857142857e-06, |
| "loss": 0.0677, |
| "step": 4825 |
| }, |
| { |
| "epoch": 0.2, |
| "learning_rate": 8.226938775510205e-06, |
| "loss": 0.0697, |
| "step": 4850 |
| }, |
| { |
| "epoch": 0.2, |
| "learning_rate": 8.216734693877552e-06, |
| "loss": 0.0672, |
| "step": 4875 |
| }, |
| { |
| "epoch": 0.2, |
| "learning_rate": 8.206530612244898e-06, |
| "loss": 0.0637, |
| "step": 4900 |
| }, |
| { |
| "epoch": 0.2, |
| "learning_rate": 8.196326530612246e-06, |
| "loss": 0.0685, |
| "step": 4925 |
| }, |
| { |
| "epoch": 0.2, |
| "learning_rate": 8.186122448979591e-06, |
| "loss": 0.0705, |
| "step": 4950 |
| }, |
| { |
| "epoch": 0.2, |
| "learning_rate": 8.17591836734694e-06, |
| "loss": 0.0673, |
| "step": 4975 |
| }, |
| { |
| "epoch": 0.2, |
| "learning_rate": 8.165714285714286e-06, |
| "loss": 0.0655, |
| "step": 5000 |
| }, |
| { |
| "epoch": 0.2, |
| "eval_loss": 0.020669255405664444, |
| "eval_runtime": 82.1606, |
| "eval_samples_per_second": 1.558, |
| "eval_steps_per_second": 0.049, |
| "eval_wer": 23.788049605411498, |
| "step": 5000 |
| }, |
| { |
| "epoch": 0.2, |
| "learning_rate": 8.155510204081634e-06, |
| "loss": 0.0706, |
| "step": 5025 |
| }, |
| { |
| "epoch": 0.21, |
| "learning_rate": 8.14530612244898e-06, |
| "loss": 0.0629, |
| "step": 5050 |
| }, |
| { |
| "epoch": 0.21, |
| "learning_rate": 8.135102040816327e-06, |
| "loss": 0.0631, |
| "step": 5075 |
| }, |
| { |
| "epoch": 0.21, |
| "learning_rate": 8.124897959183675e-06, |
| "loss": 0.0592, |
| "step": 5100 |
| }, |
| { |
| "epoch": 0.21, |
| "learning_rate": 8.11469387755102e-06, |
| "loss": 0.0632, |
| "step": 5125 |
| }, |
| { |
| "epoch": 0.21, |
| "learning_rate": 8.104489795918368e-06, |
| "loss": 0.0637, |
| "step": 5150 |
| }, |
| { |
| "epoch": 0.21, |
| "learning_rate": 8.094285714285714e-06, |
| "loss": 0.0641, |
| "step": 5175 |
| }, |
| { |
| "epoch": 0.21, |
| "learning_rate": 8.084081632653061e-06, |
| "loss": 0.0622, |
| "step": 5200 |
| }, |
| { |
| "epoch": 0.21, |
| "learning_rate": 8.073877551020409e-06, |
| "loss": 0.0627, |
| "step": 5225 |
| }, |
| { |
| "epoch": 0.21, |
| "learning_rate": 8.063673469387756e-06, |
| "loss": 0.0618, |
| "step": 5250 |
| }, |
| { |
| "epoch": 0.21, |
| "learning_rate": 8.053469387755104e-06, |
| "loss": 0.0601, |
| "step": 5275 |
| }, |
| { |
| "epoch": 0.22, |
| "learning_rate": 8.04326530612245e-06, |
| "loss": 0.0593, |
| "step": 5300 |
| }, |
| { |
| "epoch": 0.22, |
| "learning_rate": 8.033061224489797e-06, |
| "loss": 0.0631, |
| "step": 5325 |
| }, |
| { |
| "epoch": 0.22, |
| "learning_rate": 8.022857142857143e-06, |
| "loss": 0.0583, |
| "step": 5350 |
| }, |
| { |
| "epoch": 0.22, |
| "learning_rate": 8.01265306122449e-06, |
| "loss": 0.0616, |
| "step": 5375 |
| }, |
| { |
| "epoch": 0.22, |
| "learning_rate": 8.002448979591838e-06, |
| "loss": 0.0594, |
| "step": 5400 |
| }, |
| { |
| "epoch": 0.22, |
| "learning_rate": 7.992244897959184e-06, |
| "loss": 0.064, |
| "step": 5425 |
| }, |
| { |
| "epoch": 0.22, |
| "learning_rate": 7.982040816326531e-06, |
| "loss": 0.06, |
| "step": 5450 |
| }, |
| { |
| "epoch": 0.22, |
| "learning_rate": 7.971836734693878e-06, |
| "loss": 0.061, |
| "step": 5475 |
| }, |
| { |
| "epoch": 0.22, |
| "learning_rate": 7.961632653061226e-06, |
| "loss": 0.0619, |
| "step": 5500 |
| }, |
| { |
| "epoch": 0.22, |
| "learning_rate": 7.951428571428572e-06, |
| "loss": 0.0633, |
| "step": 5525 |
| }, |
| { |
| "epoch": 0.23, |
| "learning_rate": 7.94122448979592e-06, |
| "loss": 0.0595, |
| "step": 5550 |
| }, |
| { |
| "epoch": 0.23, |
| "learning_rate": 7.931020408163265e-06, |
| "loss": 0.063, |
| "step": 5575 |
| }, |
| { |
| "epoch": 0.23, |
| "learning_rate": 7.920816326530613e-06, |
| "loss": 0.061, |
| "step": 5600 |
| }, |
| { |
| "epoch": 0.23, |
| "learning_rate": 7.91061224489796e-06, |
| "loss": 0.0595, |
| "step": 5625 |
| }, |
| { |
| "epoch": 0.23, |
| "learning_rate": 7.900408163265306e-06, |
| "loss": 0.0617, |
| "step": 5650 |
| }, |
| { |
| "epoch": 0.23, |
| "learning_rate": 7.890204081632655e-06, |
| "loss": 0.0608, |
| "step": 5675 |
| }, |
| { |
| "epoch": 0.23, |
| "learning_rate": 7.88e-06, |
| "loss": 0.0595, |
| "step": 5700 |
| }, |
| { |
| "epoch": 0.23, |
| "learning_rate": 7.869795918367348e-06, |
| "loss": 0.0612, |
| "step": 5725 |
| }, |
| { |
| "epoch": 0.23, |
| "learning_rate": 7.859591836734694e-06, |
| "loss": 0.0606, |
| "step": 5750 |
| }, |
| { |
| "epoch": 0.23, |
| "learning_rate": 7.849387755102042e-06, |
| "loss": 0.0572, |
| "step": 5775 |
| }, |
| { |
| "epoch": 0.24, |
| "learning_rate": 7.839183673469389e-06, |
| "loss": 0.0598, |
| "step": 5800 |
| }, |
| { |
| "epoch": 0.24, |
| "learning_rate": 7.828979591836735e-06, |
| "loss": 0.0562, |
| "step": 5825 |
| }, |
| { |
| "epoch": 0.24, |
| "learning_rate": 7.818775510204082e-06, |
| "loss": 0.0606, |
| "step": 5850 |
| }, |
| { |
| "epoch": 0.24, |
| "learning_rate": 7.808571428571428e-06, |
| "loss": 0.0591, |
| "step": 5875 |
| }, |
| { |
| "epoch": 0.24, |
| "learning_rate": 7.798367346938777e-06, |
| "loss": 0.0578, |
| "step": 5900 |
| }, |
| { |
| "epoch": 0.24, |
| "learning_rate": 7.788163265306123e-06, |
| "loss": 0.0596, |
| "step": 5925 |
| }, |
| { |
| "epoch": 0.24, |
| "learning_rate": 7.77795918367347e-06, |
| "loss": 0.0582, |
| "step": 5950 |
| }, |
| { |
| "epoch": 0.24, |
| "learning_rate": 7.767755102040816e-06, |
| "loss": 0.0606, |
| "step": 5975 |
| }, |
| { |
| "epoch": 0.24, |
| "learning_rate": 7.757551020408164e-06, |
| "loss": 0.0609, |
| "step": 6000 |
| }, |
| { |
| "epoch": 0.24, |
| "learning_rate": 7.747346938775511e-06, |
| "loss": 0.0621, |
| "step": 6025 |
| }, |
| { |
| "epoch": 0.25, |
| "learning_rate": 7.737142857142857e-06, |
| "loss": 0.0592, |
| "step": 6050 |
| }, |
| { |
| "epoch": 0.25, |
| "learning_rate": 7.726938775510205e-06, |
| "loss": 0.0579, |
| "step": 6075 |
| }, |
| { |
| "epoch": 0.25, |
| "learning_rate": 7.71673469387755e-06, |
| "loss": 0.0576, |
| "step": 6100 |
| }, |
| { |
| "epoch": 0.25, |
| "learning_rate": 7.7065306122449e-06, |
| "loss": 0.0611, |
| "step": 6125 |
| }, |
| { |
| "epoch": 0.25, |
| "learning_rate": 7.696326530612245e-06, |
| "loss": 0.058, |
| "step": 6150 |
| }, |
| { |
| "epoch": 0.25, |
| "learning_rate": 7.686122448979593e-06, |
| "loss": 0.0608, |
| "step": 6175 |
| }, |
| { |
| "epoch": 0.25, |
| "learning_rate": 7.67591836734694e-06, |
| "loss": 0.0596, |
| "step": 6200 |
| }, |
| { |
| "epoch": 0.25, |
| "learning_rate": 7.665714285714286e-06, |
| "loss": 0.0561, |
| "step": 6225 |
| }, |
| { |
| "epoch": 0.25, |
| "learning_rate": 7.655510204081634e-06, |
| "loss": 0.057, |
| "step": 6250 |
| }, |
| { |
| "epoch": 0.25, |
| "learning_rate": 7.64530612244898e-06, |
| "loss": 0.0587, |
| "step": 6275 |
| }, |
| { |
| "epoch": 0.26, |
| "learning_rate": 7.635102040816327e-06, |
| "loss": 0.0609, |
| "step": 6300 |
| }, |
| { |
| "epoch": 0.26, |
| "learning_rate": 7.6248979591836735e-06, |
| "loss": 0.0577, |
| "step": 6325 |
| }, |
| { |
| "epoch": 0.26, |
| "learning_rate": 7.614693877551022e-06, |
| "loss": 0.0559, |
| "step": 6350 |
| }, |
| { |
| "epoch": 0.26, |
| "learning_rate": 7.6044897959183685e-06, |
| "loss": 0.0604, |
| "step": 6375 |
| }, |
| { |
| "epoch": 0.26, |
| "learning_rate": 7.594285714285715e-06, |
| "loss": 0.0585, |
| "step": 6400 |
| }, |
| { |
| "epoch": 0.26, |
| "learning_rate": 7.584081632653062e-06, |
| "loss": 0.0604, |
| "step": 6425 |
| }, |
| { |
| "epoch": 0.26, |
| "learning_rate": 7.5738775510204084e-06, |
| "loss": 0.0552, |
| "step": 6450 |
| }, |
| { |
| "epoch": 0.26, |
| "learning_rate": 7.563673469387756e-06, |
| "loss": 0.0596, |
| "step": 6475 |
| }, |
| { |
| "epoch": 0.26, |
| "learning_rate": 7.5534693877551025e-06, |
| "loss": 0.0595, |
| "step": 6500 |
| }, |
| { |
| "epoch": 0.26, |
| "learning_rate": 7.543265306122449e-06, |
| "loss": 0.0577, |
| "step": 6525 |
| }, |
| { |
| "epoch": 0.27, |
| "learning_rate": 7.533061224489796e-06, |
| "loss": 0.0555, |
| "step": 6550 |
| }, |
| { |
| "epoch": 0.27, |
| "learning_rate": 7.522857142857144e-06, |
| "loss": 0.0575, |
| "step": 6575 |
| }, |
| { |
| "epoch": 0.27, |
| "learning_rate": 7.512653061224491e-06, |
| "loss": 0.0559, |
| "step": 6600 |
| }, |
| { |
| "epoch": 0.27, |
| "learning_rate": 7.5024489795918374e-06, |
| "loss": 0.0579, |
| "step": 6625 |
| }, |
| { |
| "epoch": 0.27, |
| "learning_rate": 7.492244897959184e-06, |
| "loss": 0.0556, |
| "step": 6650 |
| }, |
| { |
| "epoch": 0.27, |
| "learning_rate": 7.4820408163265316e-06, |
| "loss": 0.059, |
| "step": 6675 |
| }, |
| { |
| "epoch": 0.27, |
| "learning_rate": 7.471836734693878e-06, |
| "loss": 0.0554, |
| "step": 6700 |
| }, |
| { |
| "epoch": 0.27, |
| "learning_rate": 7.461632653061225e-06, |
| "loss": 0.0618, |
| "step": 6725 |
| }, |
| { |
| "epoch": 0.27, |
| "learning_rate": 7.4514285714285715e-06, |
| "loss": 0.0575, |
| "step": 6750 |
| }, |
| { |
| "epoch": 0.28, |
| "learning_rate": 7.441224489795918e-06, |
| "loss": 0.0579, |
| "step": 6775 |
| }, |
| { |
| "epoch": 0.28, |
| "learning_rate": 7.4310204081632665e-06, |
| "loss": 0.0575, |
| "step": 6800 |
| }, |
| { |
| "epoch": 0.28, |
| "learning_rate": 7.420816326530613e-06, |
| "loss": 0.0545, |
| "step": 6825 |
| }, |
| { |
| "epoch": 0.28, |
| "learning_rate": 7.41061224489796e-06, |
| "loss": 0.0541, |
| "step": 6850 |
| }, |
| { |
| "epoch": 0.28, |
| "learning_rate": 7.400408163265307e-06, |
| "loss": 0.0526, |
| "step": 6875 |
| }, |
| { |
| "epoch": 0.28, |
| "learning_rate": 7.390204081632654e-06, |
| "loss": 0.0552, |
| "step": 6900 |
| }, |
| { |
| "epoch": 0.28, |
| "learning_rate": 7.3800000000000005e-06, |
| "loss": 0.0569, |
| "step": 6925 |
| }, |
| { |
| "epoch": 0.28, |
| "learning_rate": 7.369795918367347e-06, |
| "loss": 0.0578, |
| "step": 6950 |
| }, |
| { |
| "epoch": 0.28, |
| "learning_rate": 7.359591836734694e-06, |
| "loss": 0.0539, |
| "step": 6975 |
| }, |
| { |
| "epoch": 0.28, |
| "learning_rate": 7.349387755102041e-06, |
| "loss": 0.0562, |
| "step": 7000 |
| }, |
| { |
| "epoch": 0.29, |
| "learning_rate": 7.339183673469389e-06, |
| "loss": 0.0572, |
| "step": 7025 |
| }, |
| { |
| "epoch": 0.29, |
| "learning_rate": 7.328979591836735e-06, |
| "loss": 0.0588, |
| "step": 7050 |
| }, |
| { |
| "epoch": 0.29, |
| "learning_rate": 7.318775510204083e-06, |
| "loss": 0.0575, |
| "step": 7075 |
| }, |
| { |
| "epoch": 0.29, |
| "learning_rate": 7.3085714285714295e-06, |
| "loss": 0.053, |
| "step": 7100 |
| }, |
| { |
| "epoch": 0.29, |
| "learning_rate": 7.298367346938776e-06, |
| "loss": 0.0559, |
| "step": 7125 |
| }, |
| { |
| "epoch": 0.29, |
| "learning_rate": 7.288163265306123e-06, |
| "loss": 0.057, |
| "step": 7150 |
| }, |
| { |
| "epoch": 0.29, |
| "learning_rate": 7.277959183673469e-06, |
| "loss": 0.0545, |
| "step": 7175 |
| }, |
| { |
| "epoch": 0.29, |
| "learning_rate": 7.267755102040817e-06, |
| "loss": 0.059, |
| "step": 7200 |
| }, |
| { |
| "epoch": 0.29, |
| "learning_rate": 7.2575510204081635e-06, |
| "loss": 0.0508, |
| "step": 7225 |
| }, |
| { |
| "epoch": 0.29, |
| "learning_rate": 7.247346938775511e-06, |
| "loss": 0.0521, |
| "step": 7250 |
| }, |
| { |
| "epoch": 0.3, |
| "learning_rate": 7.2371428571428585e-06, |
| "loss": 0.0536, |
| "step": 7275 |
| }, |
| { |
| "epoch": 0.3, |
| "learning_rate": 7.226938775510205e-06, |
| "loss": 0.056, |
| "step": 7300 |
| }, |
| { |
| "epoch": 0.3, |
| "learning_rate": 7.216734693877552e-06, |
| "loss": 0.0545, |
| "step": 7325 |
| }, |
| { |
| "epoch": 0.3, |
| "learning_rate": 7.2065306122448984e-06, |
| "loss": 0.053, |
| "step": 7350 |
| }, |
| { |
| "epoch": 0.3, |
| "learning_rate": 7.196326530612245e-06, |
| "loss": 0.0569, |
| "step": 7375 |
| }, |
| { |
| "epoch": 0.3, |
| "learning_rate": 7.1861224489795926e-06, |
| "loss": 0.0563, |
| "step": 7400 |
| }, |
| { |
| "epoch": 0.3, |
| "learning_rate": 7.175918367346939e-06, |
| "loss": 0.0533, |
| "step": 7425 |
| }, |
| { |
| "epoch": 0.3, |
| "learning_rate": 7.165714285714286e-06, |
| "loss": 0.0522, |
| "step": 7450 |
| }, |
| { |
| "epoch": 0.3, |
| "learning_rate": 7.155510204081634e-06, |
| "loss": 0.0516, |
| "step": 7475 |
| }, |
| { |
| "epoch": 0.3, |
| "learning_rate": 7.145306122448981e-06, |
| "loss": 0.0548, |
| "step": 7500 |
| }, |
| { |
| "epoch": 0.31, |
| "learning_rate": 7.1351020408163274e-06, |
| "loss": 0.0556, |
| "step": 7525 |
| }, |
| { |
| "epoch": 0.31, |
| "learning_rate": 7.124897959183674e-06, |
| "loss": 0.0537, |
| "step": 7550 |
| }, |
| { |
| "epoch": 0.31, |
| "learning_rate": 7.114693877551021e-06, |
| "loss": 0.0546, |
| "step": 7575 |
| }, |
| { |
| "epoch": 0.31, |
| "learning_rate": 7.104489795918368e-06, |
| "loss": 0.0536, |
| "step": 7600 |
| }, |
| { |
| "epoch": 0.31, |
| "learning_rate": 7.094285714285715e-06, |
| "loss": 0.0555, |
| "step": 7625 |
| }, |
| { |
| "epoch": 0.31, |
| "learning_rate": 7.0840816326530615e-06, |
| "loss": 0.0555, |
| "step": 7650 |
| }, |
| { |
| "epoch": 0.31, |
| "learning_rate": 7.073877551020408e-06, |
| "loss": 0.0528, |
| "step": 7675 |
| }, |
| { |
| "epoch": 0.31, |
| "learning_rate": 7.0636734693877565e-06, |
| "loss": 0.054, |
| "step": 7700 |
| }, |
| { |
| "epoch": 0.31, |
| "learning_rate": 7.053469387755103e-06, |
| "loss": 0.0498, |
| "step": 7725 |
| }, |
| { |
| "epoch": 0.31, |
| "learning_rate": 7.04326530612245e-06, |
| "loss": 0.0539, |
| "step": 7750 |
| }, |
| { |
| "epoch": 0.32, |
| "learning_rate": 7.033061224489796e-06, |
| "loss": 0.0522, |
| "step": 7775 |
| }, |
| { |
| "epoch": 0.32, |
| "learning_rate": 7.022857142857144e-06, |
| "loss": 0.0505, |
| "step": 7800 |
| }, |
| { |
| "epoch": 0.32, |
| "learning_rate": 7.0126530612244905e-06, |
| "loss": 0.053, |
| "step": 7825 |
| }, |
| { |
| "epoch": 0.32, |
| "learning_rate": 7.002448979591837e-06, |
| "loss": 0.0544, |
| "step": 7850 |
| }, |
| { |
| "epoch": 0.32, |
| "learning_rate": 6.992244897959184e-06, |
| "loss": 0.054, |
| "step": 7875 |
| }, |
| { |
| "epoch": 0.32, |
| "learning_rate": 6.98204081632653e-06, |
| "loss": 0.052, |
| "step": 7900 |
| }, |
| { |
| "epoch": 0.32, |
| "learning_rate": 6.971836734693879e-06, |
| "loss": 0.0542, |
| "step": 7925 |
| }, |
| { |
| "epoch": 0.32, |
| "learning_rate": 6.961632653061225e-06, |
| "loss": 0.0536, |
| "step": 7950 |
| }, |
| { |
| "epoch": 0.32, |
| "learning_rate": 6.951428571428572e-06, |
| "loss": 0.0518, |
| "step": 7975 |
| }, |
| { |
| "epoch": 0.32, |
| "learning_rate": 6.9412244897959195e-06, |
| "loss": 0.0532, |
| "step": 8000 |
| }, |
| { |
| "epoch": 0.33, |
| "learning_rate": 6.931020408163266e-06, |
| "loss": 0.0503, |
| "step": 8025 |
| }, |
| { |
| "epoch": 0.33, |
| "learning_rate": 6.920816326530613e-06, |
| "loss": 0.0535, |
| "step": 8050 |
| }, |
| { |
| "epoch": 0.33, |
| "learning_rate": 6.9106122448979594e-06, |
| "loss": 0.0519, |
| "step": 8075 |
| }, |
| { |
| "epoch": 0.33, |
| "learning_rate": 6.900408163265306e-06, |
| "loss": 0.0552, |
| "step": 8100 |
| }, |
| { |
| "epoch": 0.33, |
| "learning_rate": 6.8902040816326536e-06, |
| "loss": 0.0521, |
| "step": 8125 |
| }, |
| { |
| "epoch": 0.33, |
| "learning_rate": 6.88e-06, |
| "loss": 0.0521, |
| "step": 8150 |
| }, |
| { |
| "epoch": 0.33, |
| "learning_rate": 6.869795918367348e-06, |
| "loss": 0.0512, |
| "step": 8175 |
| }, |
| { |
| "epoch": 0.33, |
| "learning_rate": 6.859591836734695e-06, |
| "loss": 0.0478, |
| "step": 8200 |
| }, |
| { |
| "epoch": 0.33, |
| "learning_rate": 6.849387755102042e-06, |
| "loss": 0.0534, |
| "step": 8225 |
| }, |
| { |
| "epoch": 0.34, |
| "learning_rate": 6.8391836734693884e-06, |
| "loss": 0.0544, |
| "step": 8250 |
| }, |
| { |
| "epoch": 0.34, |
| "learning_rate": 6.828979591836735e-06, |
| "loss": 0.0488, |
| "step": 8275 |
| }, |
| { |
| "epoch": 0.34, |
| "learning_rate": 6.818775510204082e-06, |
| "loss": 0.052, |
| "step": 8300 |
| }, |
| { |
| "epoch": 0.34, |
| "learning_rate": 6.808571428571429e-06, |
| "loss": 0.0507, |
| "step": 8325 |
| }, |
| { |
| "epoch": 0.34, |
| "learning_rate": 6.798367346938776e-06, |
| "loss": 0.0494, |
| "step": 8350 |
| }, |
| { |
| "epoch": 0.34, |
| "learning_rate": 6.7881632653061225e-06, |
| "loss": 0.0524, |
| "step": 8375 |
| }, |
| { |
| "epoch": 0.34, |
| "learning_rate": 6.77795918367347e-06, |
| "loss": 0.0533, |
| "step": 8400 |
| }, |
| { |
| "epoch": 0.34, |
| "learning_rate": 6.7677551020408175e-06, |
| "loss": 0.0523, |
| "step": 8425 |
| }, |
| { |
| "epoch": 0.34, |
| "learning_rate": 6.757551020408164e-06, |
| "loss": 0.0535, |
| "step": 8450 |
| }, |
| { |
| "epoch": 0.34, |
| "learning_rate": 6.747346938775511e-06, |
| "loss": 0.0522, |
| "step": 8475 |
| }, |
| { |
| "epoch": 0.35, |
| "learning_rate": 6.737142857142857e-06, |
| "loss": 0.0506, |
| "step": 8500 |
| }, |
| { |
| "epoch": 0.35, |
| "learning_rate": 6.726938775510205e-06, |
| "loss": 0.0481, |
| "step": 8525 |
| }, |
| { |
| "epoch": 0.35, |
| "learning_rate": 6.7167346938775515e-06, |
| "loss": 0.0494, |
| "step": 8550 |
| }, |
| { |
| "epoch": 0.35, |
| "learning_rate": 6.706530612244898e-06, |
| "loss": 0.0529, |
| "step": 8575 |
| }, |
| { |
| "epoch": 0.35, |
| "learning_rate": 6.696326530612245e-06, |
| "loss": 0.0492, |
| "step": 8600 |
| }, |
| { |
| "epoch": 0.35, |
| "learning_rate": 6.686122448979593e-06, |
| "loss": 0.0505, |
| "step": 8625 |
| }, |
| { |
| "epoch": 0.35, |
| "learning_rate": 6.67591836734694e-06, |
| "loss": 0.0548, |
| "step": 8650 |
| }, |
| { |
| "epoch": 0.35, |
| "learning_rate": 6.665714285714286e-06, |
| "loss": 0.0512, |
| "step": 8675 |
| }, |
| { |
| "epoch": 0.35, |
| "learning_rate": 6.655510204081633e-06, |
| "loss": 0.0518, |
| "step": 8700 |
| }, |
| { |
| "epoch": 0.35, |
| "learning_rate": 6.6453061224489805e-06, |
| "loss": 0.0489, |
| "step": 8725 |
| }, |
| { |
| "epoch": 0.36, |
| "learning_rate": 6.635102040816327e-06, |
| "loss": 0.0496, |
| "step": 8750 |
| }, |
| { |
| "epoch": 0.36, |
| "learning_rate": 6.624897959183674e-06, |
| "loss": 0.0515, |
| "step": 8775 |
| }, |
| { |
| "epoch": 0.36, |
| "learning_rate": 6.6146938775510204e-06, |
| "loss": 0.049, |
| "step": 8800 |
| }, |
| { |
| "epoch": 0.36, |
| "learning_rate": 6.604489795918367e-06, |
| "loss": 0.0492, |
| "step": 8825 |
| }, |
| { |
| "epoch": 0.36, |
| "learning_rate": 6.594285714285715e-06, |
| "loss": 0.0489, |
| "step": 8850 |
| }, |
| { |
| "epoch": 0.36, |
| "learning_rate": 6.584081632653062e-06, |
| "loss": 0.0513, |
| "step": 8875 |
| }, |
| { |
| "epoch": 0.36, |
| "learning_rate": 6.573877551020409e-06, |
| "loss": 0.0493, |
| "step": 8900 |
| }, |
| { |
| "epoch": 0.36, |
| "learning_rate": 6.563673469387755e-06, |
| "loss": 0.0504, |
| "step": 8925 |
| }, |
| { |
| "epoch": 0.36, |
| "learning_rate": 6.553469387755103e-06, |
| "loss": 0.0505, |
| "step": 8950 |
| }, |
| { |
| "epoch": 0.36, |
| "learning_rate": 6.5432653061224494e-06, |
| "loss": 0.0502, |
| "step": 8975 |
| }, |
| { |
| "epoch": 0.37, |
| "learning_rate": 6.533061224489796e-06, |
| "loss": 0.0505, |
| "step": 9000 |
| }, |
| { |
| "epoch": 0.37, |
| "learning_rate": 6.522857142857143e-06, |
| "loss": 0.0489, |
| "step": 9025 |
| }, |
| { |
| "epoch": 0.37, |
| "learning_rate": 6.51265306122449e-06, |
| "loss": 0.0489, |
| "step": 9050 |
| }, |
| { |
| "epoch": 0.37, |
| "learning_rate": 6.502448979591838e-06, |
| "loss": 0.0502, |
| "step": 9075 |
| }, |
| { |
| "epoch": 0.37, |
| "learning_rate": 6.492244897959184e-06, |
| "loss": 0.0499, |
| "step": 9100 |
| }, |
| { |
| "epoch": 0.37, |
| "learning_rate": 6.482040816326531e-06, |
| "loss": 0.0543, |
| "step": 9125 |
| }, |
| { |
| "epoch": 0.37, |
| "learning_rate": 6.4718367346938785e-06, |
| "loss": 0.0503, |
| "step": 9150 |
| }, |
| { |
| "epoch": 0.37, |
| "learning_rate": 6.461632653061225e-06, |
| "loss": 0.0518, |
| "step": 9175 |
| }, |
| { |
| "epoch": 0.37, |
| "learning_rate": 6.451428571428572e-06, |
| "loss": 0.0498, |
| "step": 9200 |
| }, |
| { |
| "epoch": 0.37, |
| "learning_rate": 6.441224489795918e-06, |
| "loss": 0.0488, |
| "step": 9225 |
| }, |
| { |
| "epoch": 0.38, |
| "learning_rate": 6.431020408163266e-06, |
| "loss": 0.0503, |
| "step": 9250 |
| }, |
| { |
| "epoch": 0.38, |
| "learning_rate": 6.4208163265306125e-06, |
| "loss": 0.0472, |
| "step": 9275 |
| }, |
| { |
| "epoch": 0.38, |
| "learning_rate": 6.41061224489796e-06, |
| "loss": 0.0491, |
| "step": 9300 |
| }, |
| { |
| "epoch": 0.38, |
| "learning_rate": 6.400408163265307e-06, |
| "loss": 0.05, |
| "step": 9325 |
| }, |
| { |
| "epoch": 0.38, |
| "learning_rate": 6.390204081632654e-06, |
| "loss": 0.0499, |
| "step": 9350 |
| }, |
| { |
| "epoch": 0.38, |
| "learning_rate": 6.380000000000001e-06, |
| "loss": 0.0474, |
| "step": 9375 |
| }, |
| { |
| "epoch": 0.38, |
| "learning_rate": 6.369795918367347e-06, |
| "loss": 0.0525, |
| "step": 9400 |
| }, |
| { |
| "epoch": 0.38, |
| "learning_rate": 6.359591836734694e-06, |
| "loss": 0.0513, |
| "step": 9425 |
| }, |
| { |
| "epoch": 0.38, |
| "learning_rate": 6.349387755102041e-06, |
| "loss": 0.0455, |
| "step": 9450 |
| }, |
| { |
| "epoch": 0.38, |
| "learning_rate": 6.339183673469388e-06, |
| "loss": 0.0504, |
| "step": 9475 |
| }, |
| { |
| "epoch": 0.39, |
| "learning_rate": 6.328979591836735e-06, |
| "loss": 0.0511, |
| "step": 9500 |
| }, |
| { |
| "epoch": 0.39, |
| "learning_rate": 6.318775510204082e-06, |
| "loss": 0.0467, |
| "step": 9525 |
| }, |
| { |
| "epoch": 0.39, |
| "learning_rate": 6.30857142857143e-06, |
| "loss": 0.0469, |
| "step": 9550 |
| }, |
| { |
| "epoch": 0.39, |
| "learning_rate": 6.298367346938776e-06, |
| "loss": 0.0505, |
| "step": 9575 |
| }, |
| { |
| "epoch": 0.39, |
| "learning_rate": 6.288163265306123e-06, |
| "loss": 0.0511, |
| "step": 9600 |
| }, |
| { |
| "epoch": 0.39, |
| "learning_rate": 6.27795918367347e-06, |
| "loss": 0.0486, |
| "step": 9625 |
| }, |
| { |
| "epoch": 0.39, |
| "learning_rate": 6.267755102040816e-06, |
| "loss": 0.0488, |
| "step": 9650 |
| }, |
| { |
| "epoch": 0.39, |
| "learning_rate": 6.257551020408164e-06, |
| "loss": 0.0494, |
| "step": 9675 |
| }, |
| { |
| "epoch": 0.39, |
| "learning_rate": 6.2473469387755104e-06, |
| "loss": 0.0486, |
| "step": 9700 |
| }, |
| { |
| "epoch": 0.39, |
| "learning_rate": 6.237142857142857e-06, |
| "loss": 0.0482, |
| "step": 9725 |
| }, |
| { |
| "epoch": 0.4, |
| "learning_rate": 6.226938775510205e-06, |
| "loss": 0.046, |
| "step": 9750 |
| }, |
| { |
| "epoch": 0.4, |
| "learning_rate": 6.216734693877552e-06, |
| "loss": 0.0466, |
| "step": 9775 |
| }, |
| { |
| "epoch": 0.4, |
| "learning_rate": 6.206530612244899e-06, |
| "loss": 0.0476, |
| "step": 9800 |
| }, |
| { |
| "epoch": 0.4, |
| "learning_rate": 6.196326530612245e-06, |
| "loss": 0.0508, |
| "step": 9825 |
| }, |
| { |
| "epoch": 0.4, |
| "learning_rate": 6.186122448979592e-06, |
| "loss": 0.0473, |
| "step": 9850 |
| }, |
| { |
| "epoch": 0.4, |
| "learning_rate": 6.1759183673469395e-06, |
| "loss": 0.0485, |
| "step": 9875 |
| }, |
| { |
| "epoch": 0.4, |
| "learning_rate": 6.165714285714286e-06, |
| "loss": 0.0501, |
| "step": 9900 |
| }, |
| { |
| "epoch": 0.4, |
| "learning_rate": 6.155510204081633e-06, |
| "loss": 0.0489, |
| "step": 9925 |
| }, |
| { |
| "epoch": 0.4, |
| "learning_rate": 6.145306122448979e-06, |
| "loss": 0.0491, |
| "step": 9950 |
| }, |
| { |
| "epoch": 0.41, |
| "learning_rate": 6.135102040816328e-06, |
| "loss": 0.0464, |
| "step": 9975 |
| }, |
| { |
| "epoch": 0.41, |
| "learning_rate": 6.124897959183674e-06, |
| "loss": 0.0449, |
| "step": 10000 |
| }, |
| { |
| "epoch": 0.41, |
| "eval_loss": 0.014047445729374886, |
| "eval_runtime": 87.8176, |
| "eval_samples_per_second": 1.458, |
| "eval_steps_per_second": 0.046, |
| "eval_wer": 20.8779443254818, |
| "step": 10000 |
| }, |
| { |
| "epoch": 0.41, |
| "learning_rate": 6.114693877551021e-06, |
| "loss": 0.0509, |
| "step": 10025 |
| }, |
| { |
| "epoch": 0.41, |
| "learning_rate": 6.104489795918368e-06, |
| "loss": 0.0483, |
| "step": 10050 |
| }, |
| { |
| "epoch": 0.41, |
| "learning_rate": 6.094693877551021e-06, |
| "loss": 0.0485, |
| "step": 10075 |
| }, |
| { |
| "epoch": 0.41, |
| "learning_rate": 6.084489795918368e-06, |
| "loss": 0.0464, |
| "step": 10100 |
| }, |
| { |
| "epoch": 0.41, |
| "learning_rate": 6.0742857142857145e-06, |
| "loss": 0.0489, |
| "step": 10125 |
| }, |
| { |
| "epoch": 0.41, |
| "learning_rate": 6.064081632653062e-06, |
| "loss": 0.0464, |
| "step": 10150 |
| }, |
| { |
| "epoch": 0.41, |
| "learning_rate": 6.053877551020409e-06, |
| "loss": 0.0454, |
| "step": 10175 |
| }, |
| { |
| "epoch": 0.41, |
| "learning_rate": 6.043673469387755e-06, |
| "loss": 0.0481, |
| "step": 10200 |
| }, |
| { |
| "epoch": 0.42, |
| "learning_rate": 6.033469387755102e-06, |
| "loss": 0.045, |
| "step": 10225 |
| }, |
| { |
| "epoch": 0.42, |
| "learning_rate": 6.02326530612245e-06, |
| "loss": 0.0462, |
| "step": 10250 |
| }, |
| { |
| "epoch": 0.42, |
| "learning_rate": 6.013061224489797e-06, |
| "loss": 0.05, |
| "step": 10275 |
| }, |
| { |
| "epoch": 0.42, |
| "learning_rate": 6.0028571428571435e-06, |
| "loss": 0.0459, |
| "step": 10300 |
| }, |
| { |
| "epoch": 0.42, |
| "learning_rate": 5.99265306122449e-06, |
| "loss": 0.0442, |
| "step": 10325 |
| }, |
| { |
| "epoch": 0.42, |
| "learning_rate": 5.982448979591838e-06, |
| "loss": 0.044, |
| "step": 10350 |
| }, |
| { |
| "epoch": 0.42, |
| "learning_rate": 5.972244897959184e-06, |
| "loss": 0.0449, |
| "step": 10375 |
| }, |
| { |
| "epoch": 0.42, |
| "learning_rate": 5.962040816326531e-06, |
| "loss": 0.046, |
| "step": 10400 |
| }, |
| { |
| "epoch": 0.42, |
| "learning_rate": 5.9518367346938775e-06, |
| "loss": 0.0457, |
| "step": 10425 |
| }, |
| { |
| "epoch": 0.42, |
| "learning_rate": 5.941632653061224e-06, |
| "loss": 0.0427, |
| "step": 10450 |
| }, |
| { |
| "epoch": 0.43, |
| "learning_rate": 5.9314285714285725e-06, |
| "loss": 0.0466, |
| "step": 10475 |
| }, |
| { |
| "epoch": 0.43, |
| "learning_rate": 5.921224489795919e-06, |
| "loss": 0.046, |
| "step": 10500 |
| }, |
| { |
| "epoch": 0.43, |
| "learning_rate": 5.911020408163266e-06, |
| "loss": 0.0459, |
| "step": 10525 |
| }, |
| { |
| "epoch": 0.43, |
| "learning_rate": 5.900816326530613e-06, |
| "loss": 0.0465, |
| "step": 10550 |
| }, |
| { |
| "epoch": 0.43, |
| "learning_rate": 5.89061224489796e-06, |
| "loss": 0.0443, |
| "step": 10575 |
| }, |
| { |
| "epoch": 0.43, |
| "learning_rate": 5.8804081632653065e-06, |
| "loss": 0.0416, |
| "step": 10600 |
| }, |
| { |
| "epoch": 0.43, |
| "learning_rate": 5.870204081632653e-06, |
| "loss": 0.0429, |
| "step": 10625 |
| }, |
| { |
| "epoch": 0.43, |
| "learning_rate": 5.86e-06, |
| "loss": 0.0449, |
| "step": 10650 |
| }, |
| { |
| "epoch": 0.43, |
| "learning_rate": 5.849795918367347e-06, |
| "loss": 0.0483, |
| "step": 10675 |
| }, |
| { |
| "epoch": 0.43, |
| "learning_rate": 5.839591836734695e-06, |
| "loss": 0.0449, |
| "step": 10700 |
| }, |
| { |
| "epoch": 0.44, |
| "learning_rate": 5.829387755102041e-06, |
| "loss": 0.0467, |
| "step": 10725 |
| }, |
| { |
| "epoch": 0.44, |
| "learning_rate": 5.819183673469389e-06, |
| "loss": 0.045, |
| "step": 10750 |
| }, |
| { |
| "epoch": 0.44, |
| "learning_rate": 5.8089795918367355e-06, |
| "loss": 0.0464, |
| "step": 10775 |
| }, |
| { |
| "epoch": 0.44, |
| "learning_rate": 5.798775510204082e-06, |
| "loss": 0.0448, |
| "step": 10800 |
| }, |
| { |
| "epoch": 0.44, |
| "learning_rate": 5.788571428571429e-06, |
| "loss": 0.0471, |
| "step": 10825 |
| }, |
| { |
| "epoch": 0.44, |
| "learning_rate": 5.7783673469387755e-06, |
| "loss": 0.0442, |
| "step": 10850 |
| }, |
| { |
| "epoch": 0.44, |
| "learning_rate": 5.768163265306123e-06, |
| "loss": 0.0435, |
| "step": 10875 |
| }, |
| { |
| "epoch": 0.44, |
| "learning_rate": 5.7579591836734696e-06, |
| "loss": 0.0431, |
| "step": 10900 |
| }, |
| { |
| "epoch": 0.44, |
| "learning_rate": 5.747755102040817e-06, |
| "loss": 0.0466, |
| "step": 10925 |
| }, |
| { |
| "epoch": 0.44, |
| "learning_rate": 5.7375510204081645e-06, |
| "loss": 0.0464, |
| "step": 10950 |
| }, |
| { |
| "epoch": 0.45, |
| "learning_rate": 5.727346938775511e-06, |
| "loss": 0.0461, |
| "step": 10975 |
| }, |
| { |
| "epoch": 0.45, |
| "learning_rate": 5.717142857142858e-06, |
| "loss": 0.0428, |
| "step": 11000 |
| }, |
| { |
| "epoch": 0.45, |
| "learning_rate": 5.7069387755102045e-06, |
| "loss": 0.0461, |
| "step": 11025 |
| }, |
| { |
| "epoch": 0.45, |
| "learning_rate": 5.696734693877551e-06, |
| "loss": 0.0465, |
| "step": 11050 |
| }, |
| { |
| "epoch": 0.45, |
| "learning_rate": 5.686530612244899e-06, |
| "loss": 0.0441, |
| "step": 11075 |
| }, |
| { |
| "epoch": 0.45, |
| "learning_rate": 5.676326530612245e-06, |
| "loss": 0.0426, |
| "step": 11100 |
| }, |
| { |
| "epoch": 0.45, |
| "learning_rate": 5.666122448979592e-06, |
| "loss": 0.0465, |
| "step": 11125 |
| }, |
| { |
| "epoch": 0.45, |
| "learning_rate": 5.655918367346939e-06, |
| "loss": 0.0453, |
| "step": 11150 |
| }, |
| { |
| "epoch": 0.45, |
| "learning_rate": 5.645714285714287e-06, |
| "loss": 0.0449, |
| "step": 11175 |
| }, |
| { |
| "epoch": 0.45, |
| "learning_rate": 5.6355102040816335e-06, |
| "loss": 0.0441, |
| "step": 11200 |
| }, |
| { |
| "epoch": 0.46, |
| "learning_rate": 5.62530612244898e-06, |
| "loss": 0.0447, |
| "step": 11225 |
| }, |
| { |
| "epoch": 0.46, |
| "learning_rate": 5.615102040816327e-06, |
| "loss": 0.0487, |
| "step": 11250 |
| }, |
| { |
| "epoch": 0.46, |
| "learning_rate": 5.604897959183674e-06, |
| "loss": 0.0456, |
| "step": 11275 |
| }, |
| { |
| "epoch": 0.46, |
| "learning_rate": 5.594693877551021e-06, |
| "loss": 0.0488, |
| "step": 11300 |
| }, |
| { |
| "epoch": 0.46, |
| "learning_rate": 5.5844897959183675e-06, |
| "loss": 0.0423, |
| "step": 11325 |
| }, |
| { |
| "epoch": 0.46, |
| "learning_rate": 5.574285714285714e-06, |
| "loss": 0.0423, |
| "step": 11350 |
| }, |
| { |
| "epoch": 0.46, |
| "learning_rate": 5.5640816326530625e-06, |
| "loss": 0.0467, |
| "step": 11375 |
| }, |
| { |
| "epoch": 0.46, |
| "learning_rate": 5.553877551020409e-06, |
| "loss": 0.0438, |
| "step": 11400 |
| }, |
| { |
| "epoch": 0.46, |
| "learning_rate": 5.543673469387756e-06, |
| "loss": 0.0436, |
| "step": 11425 |
| }, |
| { |
| "epoch": 0.47, |
| "learning_rate": 5.533469387755102e-06, |
| "loss": 0.0433, |
| "step": 11450 |
| }, |
| { |
| "epoch": 0.47, |
| "learning_rate": 5.52326530612245e-06, |
| "loss": 0.0457, |
| "step": 11475 |
| }, |
| { |
| "epoch": 0.47, |
| "learning_rate": 5.5130612244897965e-06, |
| "loss": 0.0428, |
| "step": 11500 |
| }, |
| { |
| "epoch": 0.47, |
| "learning_rate": 5.502857142857143e-06, |
| "loss": 0.0411, |
| "step": 11525 |
| }, |
| { |
| "epoch": 0.47, |
| "learning_rate": 5.49265306122449e-06, |
| "loss": 0.0445, |
| "step": 11550 |
| }, |
| { |
| "epoch": 0.47, |
| "learning_rate": 5.4824489795918364e-06, |
| "loss": 0.0438, |
| "step": 11575 |
| }, |
| { |
| "epoch": 0.47, |
| "learning_rate": 5.472244897959184e-06, |
| "loss": 0.0455, |
| "step": 11600 |
| }, |
| { |
| "epoch": 0.47, |
| "learning_rate": 5.462040816326531e-06, |
| "loss": 0.0437, |
| "step": 11625 |
| }, |
| { |
| "epoch": 0.47, |
| "learning_rate": 5.451836734693878e-06, |
| "loss": 0.0415, |
| "step": 11650 |
| }, |
| { |
| "epoch": 0.47, |
| "learning_rate": 5.4416326530612255e-06, |
| "loss": 0.0442, |
| "step": 11675 |
| }, |
| { |
| "epoch": 0.48, |
| "learning_rate": 5.431428571428572e-06, |
| "loss": 0.0421, |
| "step": 11700 |
| }, |
| { |
| "epoch": 0.48, |
| "learning_rate": 5.421224489795919e-06, |
| "loss": 0.0429, |
| "step": 11725 |
| }, |
| { |
| "epoch": 0.48, |
| "learning_rate": 5.4110204081632655e-06, |
| "loss": 0.0455, |
| "step": 11750 |
| }, |
| { |
| "epoch": 0.48, |
| "learning_rate": 5.400816326530612e-06, |
| "loss": 0.0434, |
| "step": 11775 |
| }, |
| { |
| "epoch": 0.48, |
| "learning_rate": 5.39061224489796e-06, |
| "loss": 0.0438, |
| "step": 11800 |
| }, |
| { |
| "epoch": 0.48, |
| "learning_rate": 5.380408163265306e-06, |
| "loss": 0.0444, |
| "step": 11825 |
| }, |
| { |
| "epoch": 0.48, |
| "learning_rate": 5.370204081632654e-06, |
| "loss": 0.042, |
| "step": 11850 |
| }, |
| { |
| "epoch": 0.48, |
| "learning_rate": 5.36e-06, |
| "loss": 0.0448, |
| "step": 11875 |
| }, |
| { |
| "epoch": 0.48, |
| "learning_rate": 5.349795918367348e-06, |
| "loss": 0.0438, |
| "step": 11900 |
| }, |
| { |
| "epoch": 0.48, |
| "learning_rate": 5.3395918367346945e-06, |
| "loss": 0.0417, |
| "step": 11925 |
| }, |
| { |
| "epoch": 0.49, |
| "learning_rate": 5.329387755102041e-06, |
| "loss": 0.0404, |
| "step": 11950 |
| }, |
| { |
| "epoch": 0.49, |
| "learning_rate": 5.319183673469388e-06, |
| "loss": 0.044, |
| "step": 11975 |
| }, |
| { |
| "epoch": 0.49, |
| "learning_rate": 5.308979591836735e-06, |
| "loss": 0.0452, |
| "step": 12000 |
| }, |
| { |
| "epoch": 0.49, |
| "learning_rate": 5.298775510204082e-06, |
| "loss": 0.0435, |
| "step": 12025 |
| }, |
| { |
| "epoch": 0.49, |
| "learning_rate": 5.2885714285714285e-06, |
| "loss": 0.0413, |
| "step": 12050 |
| }, |
| { |
| "epoch": 0.49, |
| "learning_rate": 5.278367346938776e-06, |
| "loss": 0.0447, |
| "step": 12075 |
| }, |
| { |
| "epoch": 0.49, |
| "learning_rate": 5.2681632653061235e-06, |
| "loss": 0.0429, |
| "step": 12100 |
| }, |
| { |
| "epoch": 0.49, |
| "learning_rate": 5.25795918367347e-06, |
| "loss": 0.0426, |
| "step": 12125 |
| }, |
| { |
| "epoch": 0.49, |
| "learning_rate": 5.247755102040817e-06, |
| "loss": 0.042, |
| "step": 12150 |
| }, |
| { |
| "epoch": 0.49, |
| "learning_rate": 5.237551020408163e-06, |
| "loss": 0.0416, |
| "step": 12175 |
| }, |
| { |
| "epoch": 0.5, |
| "learning_rate": 5.227346938775511e-06, |
| "loss": 0.0441, |
| "step": 12200 |
| }, |
| { |
| "epoch": 0.5, |
| "learning_rate": 5.2171428571428575e-06, |
| "loss": 0.043, |
| "step": 12225 |
| }, |
| { |
| "epoch": 0.5, |
| "learning_rate": 5.206938775510204e-06, |
| "loss": 0.0441, |
| "step": 12250 |
| }, |
| { |
| "epoch": 0.5, |
| "learning_rate": 5.196734693877551e-06, |
| "loss": 0.0437, |
| "step": 12275 |
| }, |
| { |
| "epoch": 0.5, |
| "learning_rate": 5.186938775510204e-06, |
| "loss": 0.0445, |
| "step": 12300 |
| }, |
| { |
| "epoch": 0.5, |
| "learning_rate": 5.176734693877551e-06, |
| "loss": 0.0423, |
| "step": 12325 |
| }, |
| { |
| "epoch": 0.5, |
| "learning_rate": 5.1665306122448985e-06, |
| "loss": 0.0405, |
| "step": 12350 |
| }, |
| { |
| "epoch": 0.5, |
| "learning_rate": 5.156326530612246e-06, |
| "loss": 0.0415, |
| "step": 12375 |
| }, |
| { |
| "epoch": 0.5, |
| "learning_rate": 5.146122448979593e-06, |
| "loss": 0.0427, |
| "step": 12400 |
| }, |
| { |
| "epoch": 0.5, |
| "learning_rate": 5.135918367346939e-06, |
| "loss": 0.043, |
| "step": 12425 |
| }, |
| { |
| "epoch": 0.51, |
| "learning_rate": 5.125714285714286e-06, |
| "loss": 0.0414, |
| "step": 12450 |
| }, |
| { |
| "epoch": 0.51, |
| "learning_rate": 5.1155102040816325e-06, |
| "loss": 0.0432, |
| "step": 12475 |
| }, |
| { |
| "epoch": 0.51, |
| "learning_rate": 5.10530612244898e-06, |
| "loss": 0.0439, |
| "step": 12500 |
| }, |
| { |
| "epoch": 0.51, |
| "learning_rate": 5.095102040816327e-06, |
| "loss": 0.0432, |
| "step": 12525 |
| }, |
| { |
| "epoch": 0.51, |
| "learning_rate": 5.084897959183673e-06, |
| "loss": 0.0448, |
| "step": 12550 |
| }, |
| { |
| "epoch": 0.51, |
| "learning_rate": 5.074693877551022e-06, |
| "loss": 0.041, |
| "step": 12575 |
| }, |
| { |
| "epoch": 0.51, |
| "learning_rate": 5.064489795918368e-06, |
| "loss": 0.0426, |
| "step": 12600 |
| }, |
| { |
| "epoch": 0.51, |
| "learning_rate": 5.054285714285715e-06, |
| "loss": 0.0431, |
| "step": 12625 |
| }, |
| { |
| "epoch": 0.51, |
| "learning_rate": 5.0440816326530615e-06, |
| "loss": 0.0425, |
| "step": 12650 |
| }, |
| { |
| "epoch": 0.51, |
| "learning_rate": 5.033877551020408e-06, |
| "loss": 0.0433, |
| "step": 12675 |
| }, |
| { |
| "epoch": 0.52, |
| "learning_rate": 5.023673469387756e-06, |
| "loss": 0.0396, |
| "step": 12700 |
| }, |
| { |
| "epoch": 0.52, |
| "learning_rate": 5.013469387755102e-06, |
| "loss": 0.0432, |
| "step": 12725 |
| }, |
| { |
| "epoch": 0.52, |
| "learning_rate": 5.003265306122449e-06, |
| "loss": 0.0434, |
| "step": 12750 |
| }, |
| { |
| "epoch": 0.52, |
| "learning_rate": 4.9930612244897964e-06, |
| "loss": 0.042, |
| "step": 12775 |
| }, |
| { |
| "epoch": 0.52, |
| "learning_rate": 4.982857142857143e-06, |
| "loss": 0.0414, |
| "step": 12800 |
| }, |
| { |
| "epoch": 0.52, |
| "learning_rate": 4.97265306122449e-06, |
| "loss": 0.0444, |
| "step": 12825 |
| }, |
| { |
| "epoch": 0.52, |
| "learning_rate": 4.962448979591837e-06, |
| "loss": 0.0416, |
| "step": 12850 |
| }, |
| { |
| "epoch": 0.52, |
| "learning_rate": 4.952244897959184e-06, |
| "loss": 0.0416, |
| "step": 12875 |
| }, |
| { |
| "epoch": 0.52, |
| "learning_rate": 4.942040816326531e-06, |
| "loss": 0.0425, |
| "step": 12900 |
| }, |
| { |
| "epoch": 0.52, |
| "learning_rate": 4.931836734693878e-06, |
| "loss": 0.0437, |
| "step": 12925 |
| }, |
| { |
| "epoch": 0.53, |
| "learning_rate": 4.921632653061225e-06, |
| "loss": 0.0414, |
| "step": 12950 |
| }, |
| { |
| "epoch": 0.53, |
| "learning_rate": 4.911428571428572e-06, |
| "loss": 0.0429, |
| "step": 12975 |
| }, |
| { |
| "epoch": 0.53, |
| "learning_rate": 4.901224489795919e-06, |
| "loss": 0.0411, |
| "step": 13000 |
| }, |
| { |
| "epoch": 0.53, |
| "learning_rate": 4.891020408163265e-06, |
| "loss": 0.0397, |
| "step": 13025 |
| }, |
| { |
| "epoch": 0.53, |
| "learning_rate": 4.880816326530613e-06, |
| "loss": 0.0423, |
| "step": 13050 |
| }, |
| { |
| "epoch": 0.53, |
| "learning_rate": 4.8706122448979595e-06, |
| "loss": 0.039, |
| "step": 13075 |
| }, |
| { |
| "epoch": 0.53, |
| "learning_rate": 4.860408163265307e-06, |
| "loss": 0.0399, |
| "step": 13100 |
| }, |
| { |
| "epoch": 0.53, |
| "learning_rate": 4.850204081632654e-06, |
| "loss": 0.0405, |
| "step": 13125 |
| }, |
| { |
| "epoch": 0.53, |
| "learning_rate": 4.84e-06, |
| "loss": 0.0414, |
| "step": 13150 |
| }, |
| { |
| "epoch": 0.54, |
| "learning_rate": 4.829795918367347e-06, |
| "loss": 0.0417, |
| "step": 13175 |
| }, |
| { |
| "epoch": 0.54, |
| "learning_rate": 4.819591836734694e-06, |
| "loss": 0.0408, |
| "step": 13200 |
| }, |
| { |
| "epoch": 0.54, |
| "learning_rate": 4.809387755102041e-06, |
| "loss": 0.0415, |
| "step": 13225 |
| }, |
| { |
| "epoch": 0.54, |
| "learning_rate": 4.799183673469388e-06, |
| "loss": 0.0402, |
| "step": 13250 |
| }, |
| { |
| "epoch": 0.54, |
| "learning_rate": 4.788979591836735e-06, |
| "loss": 0.0426, |
| "step": 13275 |
| }, |
| { |
| "epoch": 0.54, |
| "learning_rate": 4.778775510204083e-06, |
| "loss": 0.0438, |
| "step": 13300 |
| }, |
| { |
| "epoch": 0.54, |
| "learning_rate": 4.768571428571429e-06, |
| "loss": 0.042, |
| "step": 13325 |
| }, |
| { |
| "epoch": 0.54, |
| "learning_rate": 4.758367346938776e-06, |
| "loss": 0.0432, |
| "step": 13350 |
| }, |
| { |
| "epoch": 0.54, |
| "learning_rate": 4.7481632653061225e-06, |
| "loss": 0.0395, |
| "step": 13375 |
| }, |
| { |
| "epoch": 0.54, |
| "learning_rate": 4.737959183673469e-06, |
| "loss": 0.0423, |
| "step": 13400 |
| }, |
| { |
| "epoch": 0.55, |
| "learning_rate": 4.727755102040817e-06, |
| "loss": 0.0402, |
| "step": 13425 |
| }, |
| { |
| "epoch": 0.55, |
| "learning_rate": 4.717551020408163e-06, |
| "loss": 0.0425, |
| "step": 13450 |
| }, |
| { |
| "epoch": 0.55, |
| "learning_rate": 4.707346938775511e-06, |
| "loss": 0.0392, |
| "step": 13475 |
| }, |
| { |
| "epoch": 0.55, |
| "learning_rate": 4.6971428571428574e-06, |
| "loss": 0.0409, |
| "step": 13500 |
| }, |
| { |
| "epoch": 0.55, |
| "learning_rate": 4.686938775510205e-06, |
| "loss": 0.0383, |
| "step": 13525 |
| }, |
| { |
| "epoch": 0.55, |
| "learning_rate": 4.6767346938775516e-06, |
| "loss": 0.0413, |
| "step": 13550 |
| }, |
| { |
| "epoch": 0.55, |
| "learning_rate": 4.666530612244898e-06, |
| "loss": 0.0438, |
| "step": 13575 |
| }, |
| { |
| "epoch": 0.55, |
| "learning_rate": 4.656326530612245e-06, |
| "loss": 0.0407, |
| "step": 13600 |
| }, |
| { |
| "epoch": 0.55, |
| "learning_rate": 4.646122448979592e-06, |
| "loss": 0.0408, |
| "step": 13625 |
| }, |
| { |
| "epoch": 0.55, |
| "learning_rate": 4.635918367346939e-06, |
| "loss": 0.0412, |
| "step": 13650 |
| }, |
| { |
| "epoch": 0.56, |
| "learning_rate": 4.6257142857142864e-06, |
| "loss": 0.0441, |
| "step": 13675 |
| }, |
| { |
| "epoch": 0.56, |
| "learning_rate": 4.615510204081633e-06, |
| "loss": 0.042, |
| "step": 13700 |
| }, |
| { |
| "epoch": 0.56, |
| "learning_rate": 4.60530612244898e-06, |
| "loss": 0.0401, |
| "step": 13725 |
| }, |
| { |
| "epoch": 0.56, |
| "learning_rate": 4.595102040816327e-06, |
| "loss": 0.0434, |
| "step": 13750 |
| }, |
| { |
| "epoch": 0.56, |
| "learning_rate": 4.584897959183674e-06, |
| "loss": 0.0396, |
| "step": 13775 |
| }, |
| { |
| "epoch": 0.56, |
| "learning_rate": 4.5746938775510205e-06, |
| "loss": 0.0408, |
| "step": 13800 |
| }, |
| { |
| "epoch": 0.56, |
| "learning_rate": 4.564489795918368e-06, |
| "loss": 0.0409, |
| "step": 13825 |
| }, |
| { |
| "epoch": 0.56, |
| "learning_rate": 4.554285714285715e-06, |
| "loss": 0.0391, |
| "step": 13850 |
| }, |
| { |
| "epoch": 0.56, |
| "learning_rate": 4.544081632653062e-06, |
| "loss": 0.0408, |
| "step": 13875 |
| }, |
| { |
| "epoch": 0.56, |
| "learning_rate": 4.533877551020409e-06, |
| "loss": 0.0411, |
| "step": 13900 |
| }, |
| { |
| "epoch": 0.57, |
| "learning_rate": 4.523673469387755e-06, |
| "loss": 0.0397, |
| "step": 13925 |
| }, |
| { |
| "epoch": 0.57, |
| "learning_rate": 4.513469387755102e-06, |
| "loss": 0.0413, |
| "step": 13950 |
| }, |
| { |
| "epoch": 0.57, |
| "learning_rate": 4.5032653061224495e-06, |
| "loss": 0.0394, |
| "step": 13975 |
| }, |
| { |
| "epoch": 0.57, |
| "learning_rate": 4.493061224489796e-06, |
| "loss": 0.0417, |
| "step": 14000 |
| }, |
| { |
| "epoch": 0.57, |
| "learning_rate": 4.482857142857144e-06, |
| "loss": 0.0392, |
| "step": 14025 |
| }, |
| { |
| "epoch": 0.57, |
| "learning_rate": 4.47265306122449e-06, |
| "loss": 0.0409, |
| "step": 14050 |
| }, |
| { |
| "epoch": 0.57, |
| "learning_rate": 4.462448979591837e-06, |
| "loss": 0.0403, |
| "step": 14075 |
| }, |
| { |
| "epoch": 0.57, |
| "learning_rate": 4.452244897959184e-06, |
| "loss": 0.0424, |
| "step": 14100 |
| }, |
| { |
| "epoch": 0.57, |
| "learning_rate": 4.442040816326531e-06, |
| "loss": 0.0394, |
| "step": 14125 |
| }, |
| { |
| "epoch": 0.57, |
| "learning_rate": 4.431836734693878e-06, |
| "loss": 0.0395, |
| "step": 14150 |
| }, |
| { |
| "epoch": 0.58, |
| "learning_rate": 4.421632653061224e-06, |
| "loss": 0.0398, |
| "step": 14175 |
| }, |
| { |
| "epoch": 0.58, |
| "learning_rate": 4.411428571428572e-06, |
| "loss": 0.0419, |
| "step": 14200 |
| }, |
| { |
| "epoch": 0.58, |
| "learning_rate": 4.401224489795919e-06, |
| "loss": 0.0407, |
| "step": 14225 |
| }, |
| { |
| "epoch": 0.58, |
| "learning_rate": 4.391020408163266e-06, |
| "loss": 0.0403, |
| "step": 14250 |
| }, |
| { |
| "epoch": 0.58, |
| "learning_rate": 4.3808163265306125e-06, |
| "loss": 0.0403, |
| "step": 14275 |
| }, |
| { |
| "epoch": 0.58, |
| "learning_rate": 4.370612244897959e-06, |
| "loss": 0.0413, |
| "step": 14300 |
| }, |
| { |
| "epoch": 0.58, |
| "learning_rate": 4.360408163265307e-06, |
| "loss": 0.0419, |
| "step": 14325 |
| }, |
| { |
| "epoch": 0.58, |
| "learning_rate": 4.350204081632653e-06, |
| "loss": 0.04, |
| "step": 14350 |
| }, |
| { |
| "epoch": 0.58, |
| "learning_rate": 4.34e-06, |
| "loss": 0.0406, |
| "step": 14375 |
| }, |
| { |
| "epoch": 0.58, |
| "learning_rate": 4.3297959183673474e-06, |
| "loss": 0.0405, |
| "step": 14400 |
| }, |
| { |
| "epoch": 0.59, |
| "learning_rate": 4.319591836734694e-06, |
| "loss": 0.0401, |
| "step": 14425 |
| }, |
| { |
| "epoch": 0.59, |
| "learning_rate": 4.3093877551020416e-06, |
| "loss": 0.0396, |
| "step": 14450 |
| }, |
| { |
| "epoch": 0.59, |
| "learning_rate": 4.299183673469388e-06, |
| "loss": 0.0388, |
| "step": 14475 |
| }, |
| { |
| "epoch": 0.59, |
| "learning_rate": 4.288979591836735e-06, |
| "loss": 0.0431, |
| "step": 14500 |
| }, |
| { |
| "epoch": 0.59, |
| "learning_rate": 4.2787755102040815e-06, |
| "loss": 0.0414, |
| "step": 14525 |
| }, |
| { |
| "epoch": 0.59, |
| "learning_rate": 4.268979591836735e-06, |
| "loss": 0.0409, |
| "step": 14550 |
| }, |
| { |
| "epoch": 0.59, |
| "learning_rate": 4.258775510204082e-06, |
| "loss": 0.0423, |
| "step": 14575 |
| }, |
| { |
| "epoch": 0.59, |
| "learning_rate": 4.248571428571429e-06, |
| "loss": 0.0415, |
| "step": 14600 |
| }, |
| { |
| "epoch": 0.59, |
| "learning_rate": 4.238367346938776e-06, |
| "loss": 0.0414, |
| "step": 14625 |
| }, |
| { |
| "epoch": 0.59, |
| "learning_rate": 4.2281632653061224e-06, |
| "loss": 0.0397, |
| "step": 14650 |
| }, |
| { |
| "epoch": 0.6, |
| "learning_rate": 4.21795918367347e-06, |
| "loss": 0.0392, |
| "step": 14675 |
| }, |
| { |
| "epoch": 0.6, |
| "learning_rate": 4.2077551020408166e-06, |
| "loss": 0.04, |
| "step": 14700 |
| }, |
| { |
| "epoch": 0.6, |
| "learning_rate": 4.197551020408164e-06, |
| "loss": 0.0393, |
| "step": 14725 |
| }, |
| { |
| "epoch": 0.6, |
| "learning_rate": 4.187346938775511e-06, |
| "loss": 0.0388, |
| "step": 14750 |
| }, |
| { |
| "epoch": 0.6, |
| "learning_rate": 4.177142857142857e-06, |
| "loss": 0.039, |
| "step": 14775 |
| }, |
| { |
| "epoch": 0.6, |
| "learning_rate": 4.166938775510204e-06, |
| "loss": 0.0379, |
| "step": 14800 |
| }, |
| { |
| "epoch": 0.6, |
| "learning_rate": 4.1567346938775514e-06, |
| "loss": 0.0382, |
| "step": 14825 |
| }, |
| { |
| "epoch": 0.6, |
| "learning_rate": 4.146530612244898e-06, |
| "loss": 0.0388, |
| "step": 14850 |
| }, |
| { |
| "epoch": 0.6, |
| "learning_rate": 4.1363265306122456e-06, |
| "loss": 0.0414, |
| "step": 14875 |
| }, |
| { |
| "epoch": 0.61, |
| "learning_rate": 4.126122448979592e-06, |
| "loss": 0.0404, |
| "step": 14900 |
| }, |
| { |
| "epoch": 0.61, |
| "learning_rate": 4.115918367346939e-06, |
| "loss": 0.0398, |
| "step": 14925 |
| }, |
| { |
| "epoch": 0.61, |
| "learning_rate": 4.105714285714286e-06, |
| "loss": 0.0379, |
| "step": 14950 |
| }, |
| { |
| "epoch": 0.61, |
| "learning_rate": 4.095510204081633e-06, |
| "loss": 0.0383, |
| "step": 14975 |
| }, |
| { |
| "epoch": 0.61, |
| "learning_rate": 4.08530612244898e-06, |
| "loss": 0.0405, |
| "step": 15000 |
| }, |
| { |
| "epoch": 0.61, |
| "eval_loss": 0.009663587436079979, |
| "eval_runtime": 87.9896, |
| "eval_samples_per_second": 1.455, |
| "eval_steps_per_second": 0.045, |
| "eval_wer": 18.47345132743363, |
| "step": 15000 |
| }, |
| { |
| "epoch": 0.61, |
| "learning_rate": 4.075102040816326e-06, |
| "loss": 0.0443, |
| "step": 15025 |
| }, |
| { |
| "epoch": 0.61, |
| "learning_rate": 4.064897959183674e-06, |
| "loss": 0.0437, |
| "step": 15050 |
| }, |
| { |
| "epoch": 0.61, |
| "learning_rate": 4.054693877551021e-06, |
| "loss": 0.0412, |
| "step": 15075 |
| }, |
| { |
| "epoch": 0.61, |
| "learning_rate": 4.044489795918368e-06, |
| "loss": 0.0388, |
| "step": 15100 |
| }, |
| { |
| "epoch": 0.61, |
| "learning_rate": 4.0342857142857145e-06, |
| "loss": 0.0414, |
| "step": 15125 |
| }, |
| { |
| "epoch": 0.62, |
| "learning_rate": 4.024081632653061e-06, |
| "loss": 0.0422, |
| "step": 15150 |
| }, |
| { |
| "epoch": 0.62, |
| "learning_rate": 4.013877551020409e-06, |
| "loss": 0.0382, |
| "step": 15175 |
| }, |
| { |
| "epoch": 0.62, |
| "learning_rate": 4.003673469387755e-06, |
| "loss": 0.039, |
| "step": 15200 |
| }, |
| { |
| "epoch": 0.62, |
| "learning_rate": 3.993469387755102e-06, |
| "loss": 0.0427, |
| "step": 15225 |
| }, |
| { |
| "epoch": 0.62, |
| "learning_rate": 3.983265306122449e-06, |
| "loss": 0.0389, |
| "step": 15250 |
| }, |
| { |
| "epoch": 0.62, |
| "learning_rate": 3.973061224489797e-06, |
| "loss": 0.0396, |
| "step": 15275 |
| }, |
| { |
| "epoch": 0.62, |
| "learning_rate": 3.9628571428571435e-06, |
| "loss": 0.036, |
| "step": 15300 |
| }, |
| { |
| "epoch": 0.62, |
| "learning_rate": 3.95265306122449e-06, |
| "loss": 0.0379, |
| "step": 15325 |
| }, |
| { |
| "epoch": 0.62, |
| "learning_rate": 3.942448979591837e-06, |
| "loss": 0.0383, |
| "step": 15350 |
| }, |
| { |
| "epoch": 0.62, |
| "learning_rate": 3.9322448979591834e-06, |
| "loss": 0.0416, |
| "step": 15375 |
| }, |
| { |
| "epoch": 0.63, |
| "learning_rate": 3.922040816326531e-06, |
| "loss": 0.0386, |
| "step": 15400 |
| }, |
| { |
| "epoch": 0.63, |
| "learning_rate": 3.9118367346938776e-06, |
| "loss": 0.0376, |
| "step": 15425 |
| }, |
| { |
| "epoch": 0.63, |
| "learning_rate": 3.901632653061225e-06, |
| "loss": 0.0396, |
| "step": 15450 |
| }, |
| { |
| "epoch": 0.63, |
| "learning_rate": 3.891428571428572e-06, |
| "loss": 0.0407, |
| "step": 15475 |
| }, |
| { |
| "epoch": 0.63, |
| "learning_rate": 3.881224489795919e-06, |
| "loss": 0.0391, |
| "step": 15500 |
| }, |
| { |
| "epoch": 0.63, |
| "learning_rate": 3.871020408163266e-06, |
| "loss": 0.0394, |
| "step": 15525 |
| }, |
| { |
| "epoch": 0.63, |
| "learning_rate": 3.8608163265306124e-06, |
| "loss": 0.039, |
| "step": 15550 |
| }, |
| { |
| "epoch": 0.63, |
| "learning_rate": 3.850612244897959e-06, |
| "loss": 0.0391, |
| "step": 15575 |
| }, |
| { |
| "epoch": 0.63, |
| "learning_rate": 3.8404081632653066e-06, |
| "loss": 0.0389, |
| "step": 15600 |
| }, |
| { |
| "epoch": 0.63, |
| "learning_rate": 3.830204081632653e-06, |
| "loss": 0.0381, |
| "step": 15625 |
| }, |
| { |
| "epoch": 0.64, |
| "learning_rate": 3.820000000000001e-06, |
| "loss": 0.0413, |
| "step": 15650 |
| }, |
| { |
| "epoch": 0.64, |
| "learning_rate": 3.8097959183673473e-06, |
| "loss": 0.0387, |
| "step": 15675 |
| }, |
| { |
| "epoch": 0.64, |
| "learning_rate": 3.799591836734694e-06, |
| "loss": 0.039, |
| "step": 15700 |
| }, |
| { |
| "epoch": 0.64, |
| "learning_rate": 3.7893877551020415e-06, |
| "loss": 0.0396, |
| "step": 15725 |
| }, |
| { |
| "epoch": 0.64, |
| "learning_rate": 3.779183673469388e-06, |
| "loss": 0.0403, |
| "step": 15750 |
| }, |
| { |
| "epoch": 0.64, |
| "learning_rate": 3.768979591836735e-06, |
| "loss": 0.0383, |
| "step": 15775 |
| }, |
| { |
| "epoch": 0.64, |
| "learning_rate": 3.758775510204082e-06, |
| "loss": 0.042, |
| "step": 15800 |
| }, |
| { |
| "epoch": 0.64, |
| "learning_rate": 3.7485714285714284e-06, |
| "loss": 0.0399, |
| "step": 15825 |
| }, |
| { |
| "epoch": 0.64, |
| "learning_rate": 3.738367346938776e-06, |
| "loss": 0.0371, |
| "step": 15850 |
| }, |
| { |
| "epoch": 0.64, |
| "learning_rate": 3.728163265306123e-06, |
| "loss": 0.04, |
| "step": 15875 |
| }, |
| { |
| "epoch": 0.65, |
| "learning_rate": 3.7179591836734696e-06, |
| "loss": 0.0378, |
| "step": 15900 |
| }, |
| { |
| "epoch": 0.65, |
| "learning_rate": 3.7077551020408163e-06, |
| "loss": 0.0376, |
| "step": 15925 |
| }, |
| { |
| "epoch": 0.65, |
| "learning_rate": 3.6975510204081638e-06, |
| "loss": 0.0364, |
| "step": 15950 |
| }, |
| { |
| "epoch": 0.65, |
| "learning_rate": 3.687346938775511e-06, |
| "loss": 0.0406, |
| "step": 15975 |
| }, |
| { |
| "epoch": 0.65, |
| "learning_rate": 3.6771428571428575e-06, |
| "loss": 0.0383, |
| "step": 16000 |
| }, |
| { |
| "epoch": 0.65, |
| "learning_rate": 3.666938775510204e-06, |
| "loss": 0.0389, |
| "step": 16025 |
| }, |
| { |
| "epoch": 0.65, |
| "learning_rate": 3.656734693877551e-06, |
| "loss": 0.0375, |
| "step": 16050 |
| }, |
| { |
| "epoch": 0.65, |
| "learning_rate": 3.6465306122448986e-06, |
| "loss": 0.0392, |
| "step": 16075 |
| }, |
| { |
| "epoch": 0.65, |
| "learning_rate": 3.6363265306122453e-06, |
| "loss": 0.0412, |
| "step": 16100 |
| }, |
| { |
| "epoch": 0.65, |
| "learning_rate": 3.626122448979592e-06, |
| "loss": 0.0366, |
| "step": 16125 |
| }, |
| { |
| "epoch": 0.66, |
| "learning_rate": 3.615918367346939e-06, |
| "loss": 0.038, |
| "step": 16150 |
| }, |
| { |
| "epoch": 0.66, |
| "learning_rate": 3.6057142857142865e-06, |
| "loss": 0.0375, |
| "step": 16175 |
| }, |
| { |
| "epoch": 0.66, |
| "learning_rate": 3.595510204081633e-06, |
| "loss": 0.0386, |
| "step": 16200 |
| }, |
| { |
| "epoch": 0.66, |
| "learning_rate": 3.5853061224489797e-06, |
| "loss": 0.0376, |
| "step": 16225 |
| }, |
| { |
| "epoch": 0.66, |
| "learning_rate": 3.575102040816327e-06, |
| "loss": 0.0403, |
| "step": 16250 |
| }, |
| { |
| "epoch": 0.66, |
| "learning_rate": 3.5648979591836734e-06, |
| "loss": 0.042, |
| "step": 16275 |
| }, |
| { |
| "epoch": 0.66, |
| "learning_rate": 3.554693877551021e-06, |
| "loss": 0.0364, |
| "step": 16300 |
| }, |
| { |
| "epoch": 0.66, |
| "learning_rate": 3.5444897959183676e-06, |
| "loss": 0.0359, |
| "step": 16325 |
| }, |
| { |
| "epoch": 0.66, |
| "learning_rate": 3.5342857142857146e-06, |
| "loss": 0.0369, |
| "step": 16350 |
| }, |
| { |
| "epoch": 0.67, |
| "learning_rate": 3.5240816326530613e-06, |
| "loss": 0.0393, |
| "step": 16375 |
| }, |
| { |
| "epoch": 0.67, |
| "learning_rate": 3.5138775510204088e-06, |
| "loss": 0.0409, |
| "step": 16400 |
| }, |
| { |
| "epoch": 0.67, |
| "learning_rate": 3.5036734693877554e-06, |
| "loss": 0.0374, |
| "step": 16425 |
| }, |
| { |
| "epoch": 0.67, |
| "learning_rate": 3.4934693877551025e-06, |
| "loss": 0.0381, |
| "step": 16450 |
| }, |
| { |
| "epoch": 0.67, |
| "learning_rate": 3.483265306122449e-06, |
| "loss": 0.0376, |
| "step": 16475 |
| }, |
| { |
| "epoch": 0.67, |
| "learning_rate": 3.473061224489796e-06, |
| "loss": 0.0351, |
| "step": 16500 |
| }, |
| { |
| "epoch": 0.67, |
| "learning_rate": 3.4628571428571432e-06, |
| "loss": 0.0373, |
| "step": 16525 |
| }, |
| { |
| "epoch": 0.67, |
| "learning_rate": 3.4526530612244903e-06, |
| "loss": 0.0364, |
| "step": 16550 |
| }, |
| { |
| "epoch": 0.67, |
| "learning_rate": 3.442448979591837e-06, |
| "loss": 0.0398, |
| "step": 16575 |
| }, |
| { |
| "epoch": 0.67, |
| "learning_rate": 3.43265306122449e-06, |
| "loss": 0.0382, |
| "step": 16600 |
| }, |
| { |
| "epoch": 0.68, |
| "learning_rate": 3.422448979591837e-06, |
| "loss": 0.0394, |
| "step": 16625 |
| }, |
| { |
| "epoch": 0.68, |
| "learning_rate": 3.4122448979591838e-06, |
| "loss": 0.0388, |
| "step": 16650 |
| }, |
| { |
| "epoch": 0.68, |
| "learning_rate": 3.4020408163265312e-06, |
| "loss": 0.039, |
| "step": 16675 |
| }, |
| { |
| "epoch": 0.68, |
| "learning_rate": 3.391836734693878e-06, |
| "loss": 0.0382, |
| "step": 16700 |
| }, |
| { |
| "epoch": 0.68, |
| "learning_rate": 3.381632653061225e-06, |
| "loss": 0.039, |
| "step": 16725 |
| }, |
| { |
| "epoch": 0.68, |
| "learning_rate": 3.3714285714285716e-06, |
| "loss": 0.0372, |
| "step": 16750 |
| }, |
| { |
| "epoch": 0.68, |
| "learning_rate": 3.3612244897959182e-06, |
| "loss": 0.0372, |
| "step": 16775 |
| }, |
| { |
| "epoch": 0.68, |
| "learning_rate": 3.3510204081632657e-06, |
| "loss": 0.0373, |
| "step": 16800 |
| }, |
| { |
| "epoch": 0.68, |
| "learning_rate": 3.3408163265306128e-06, |
| "loss": 0.0408, |
| "step": 16825 |
| }, |
| { |
| "epoch": 0.68, |
| "learning_rate": 3.3306122448979594e-06, |
| "loss": 0.0371, |
| "step": 16850 |
| }, |
| { |
| "epoch": 0.69, |
| "learning_rate": 3.320408163265306e-06, |
| "loss": 0.0368, |
| "step": 16875 |
| }, |
| { |
| "epoch": 0.69, |
| "learning_rate": 3.310204081632653e-06, |
| "loss": 0.0402, |
| "step": 16900 |
| }, |
| { |
| "epoch": 0.69, |
| "learning_rate": 3.3000000000000006e-06, |
| "loss": 0.0377, |
| "step": 16925 |
| }, |
| { |
| "epoch": 0.69, |
| "learning_rate": 3.2897959183673472e-06, |
| "loss": 0.0382, |
| "step": 16950 |
| }, |
| { |
| "epoch": 0.69, |
| "learning_rate": 3.279591836734694e-06, |
| "loss": 0.038, |
| "step": 16975 |
| }, |
| { |
| "epoch": 0.69, |
| "learning_rate": 3.269387755102041e-06, |
| "loss": 0.0378, |
| "step": 17000 |
| }, |
| { |
| "epoch": 0.69, |
| "learning_rate": 3.2591836734693884e-06, |
| "loss": 0.0393, |
| "step": 17025 |
| }, |
| { |
| "epoch": 0.69, |
| "learning_rate": 3.248979591836735e-06, |
| "loss": 0.0354, |
| "step": 17050 |
| }, |
| { |
| "epoch": 0.69, |
| "learning_rate": 3.2387755102040817e-06, |
| "loss": 0.0362, |
| "step": 17075 |
| }, |
| { |
| "epoch": 0.69, |
| "learning_rate": 3.2285714285714288e-06, |
| "loss": 0.038, |
| "step": 17100 |
| }, |
| { |
| "epoch": 0.7, |
| "learning_rate": 3.2183673469387754e-06, |
| "loss": 0.0399, |
| "step": 17125 |
| }, |
| { |
| "epoch": 0.7, |
| "learning_rate": 3.208163265306123e-06, |
| "loss": 0.0386, |
| "step": 17150 |
| }, |
| { |
| "epoch": 0.7, |
| "learning_rate": 3.1979591836734695e-06, |
| "loss": 0.0364, |
| "step": 17175 |
| }, |
| { |
| "epoch": 0.7, |
| "learning_rate": 3.1877551020408166e-06, |
| "loss": 0.0369, |
| "step": 17200 |
| }, |
| { |
| "epoch": 0.7, |
| "learning_rate": 3.1775510204081632e-06, |
| "loss": 0.0393, |
| "step": 17225 |
| }, |
| { |
| "epoch": 0.7, |
| "learning_rate": 3.1673469387755107e-06, |
| "loss": 0.0383, |
| "step": 17250 |
| }, |
| { |
| "epoch": 0.7, |
| "learning_rate": 3.1571428571428573e-06, |
| "loss": 0.0392, |
| "step": 17275 |
| }, |
| { |
| "epoch": 0.7, |
| "learning_rate": 3.1469387755102044e-06, |
| "loss": 0.0365, |
| "step": 17300 |
| }, |
| { |
| "epoch": 0.7, |
| "learning_rate": 3.136734693877551e-06, |
| "loss": 0.0392, |
| "step": 17325 |
| }, |
| { |
| "epoch": 0.7, |
| "learning_rate": 3.126530612244898e-06, |
| "loss": 0.0358, |
| "step": 17350 |
| }, |
| { |
| "epoch": 0.71, |
| "learning_rate": 3.116326530612245e-06, |
| "loss": 0.035, |
| "step": 17375 |
| }, |
| { |
| "epoch": 0.71, |
| "learning_rate": 3.1061224489795922e-06, |
| "loss": 0.0361, |
| "step": 17400 |
| }, |
| { |
| "epoch": 0.71, |
| "learning_rate": 3.095918367346939e-06, |
| "loss": 0.0378, |
| "step": 17425 |
| }, |
| { |
| "epoch": 0.71, |
| "learning_rate": 3.085714285714286e-06, |
| "loss": 0.0368, |
| "step": 17450 |
| }, |
| { |
| "epoch": 0.71, |
| "learning_rate": 3.075510204081633e-06, |
| "loss": 0.0367, |
| "step": 17475 |
| }, |
| { |
| "epoch": 0.71, |
| "learning_rate": 3.06530612244898e-06, |
| "loss": 0.0384, |
| "step": 17500 |
| }, |
| { |
| "epoch": 0.71, |
| "learning_rate": 3.0551020408163267e-06, |
| "loss": 0.0376, |
| "step": 17525 |
| }, |
| { |
| "epoch": 0.71, |
| "learning_rate": 3.0448979591836738e-06, |
| "loss": 0.0394, |
| "step": 17550 |
| }, |
| { |
| "epoch": 0.71, |
| "learning_rate": 3.0346938775510204e-06, |
| "loss": 0.0341, |
| "step": 17575 |
| }, |
| { |
| "epoch": 0.71, |
| "learning_rate": 3.024489795918368e-06, |
| "loss": 0.0364, |
| "step": 17600 |
| }, |
| { |
| "epoch": 0.72, |
| "learning_rate": 3.0142857142857145e-06, |
| "loss": 0.0366, |
| "step": 17625 |
| }, |
| { |
| "epoch": 0.72, |
| "learning_rate": 3.0040816326530616e-06, |
| "loss": 0.0371, |
| "step": 17650 |
| }, |
| { |
| "epoch": 0.72, |
| "learning_rate": 2.9938775510204082e-06, |
| "loss": 0.0369, |
| "step": 17675 |
| }, |
| { |
| "epoch": 0.72, |
| "learning_rate": 2.9836734693877557e-06, |
| "loss": 0.0374, |
| "step": 17700 |
| }, |
| { |
| "epoch": 0.72, |
| "learning_rate": 2.9734693877551024e-06, |
| "loss": 0.0384, |
| "step": 17725 |
| }, |
| { |
| "epoch": 0.72, |
| "learning_rate": 2.9632653061224494e-06, |
| "loss": 0.0365, |
| "step": 17750 |
| }, |
| { |
| "epoch": 0.72, |
| "learning_rate": 2.953061224489796e-06, |
| "loss": 0.0387, |
| "step": 17775 |
| }, |
| { |
| "epoch": 0.72, |
| "learning_rate": 2.9428571428571427e-06, |
| "loss": 0.0354, |
| "step": 17800 |
| }, |
| { |
| "epoch": 0.72, |
| "learning_rate": 2.93265306122449e-06, |
| "loss": 0.0349, |
| "step": 17825 |
| }, |
| { |
| "epoch": 0.72, |
| "learning_rate": 2.9224489795918372e-06, |
| "loss": 0.0359, |
| "step": 17850 |
| }, |
| { |
| "epoch": 0.73, |
| "learning_rate": 2.912244897959184e-06, |
| "loss": 0.0377, |
| "step": 17875 |
| }, |
| { |
| "epoch": 0.73, |
| "learning_rate": 2.9020408163265305e-06, |
| "loss": 0.0388, |
| "step": 17900 |
| }, |
| { |
| "epoch": 0.73, |
| "learning_rate": 2.891836734693878e-06, |
| "loss": 0.04, |
| "step": 17925 |
| }, |
| { |
| "epoch": 0.73, |
| "learning_rate": 2.881632653061225e-06, |
| "loss": 0.0375, |
| "step": 17950 |
| }, |
| { |
| "epoch": 0.73, |
| "learning_rate": 2.8714285714285717e-06, |
| "loss": 0.0373, |
| "step": 17975 |
| }, |
| { |
| "epoch": 0.73, |
| "learning_rate": 2.8612244897959183e-06, |
| "loss": 0.0366, |
| "step": 18000 |
| }, |
| { |
| "epoch": 0.73, |
| "learning_rate": 2.8510204081632654e-06, |
| "loss": 0.0372, |
| "step": 18025 |
| }, |
| { |
| "epoch": 0.73, |
| "learning_rate": 2.8408163265306125e-06, |
| "loss": 0.0374, |
| "step": 18050 |
| }, |
| { |
| "epoch": 0.73, |
| "learning_rate": 2.8306122448979595e-06, |
| "loss": 0.0371, |
| "step": 18075 |
| }, |
| { |
| "epoch": 0.74, |
| "learning_rate": 2.820408163265306e-06, |
| "loss": 0.0338, |
| "step": 18100 |
| }, |
| { |
| "epoch": 0.74, |
| "learning_rate": 2.8102040816326532e-06, |
| "loss": 0.0383, |
| "step": 18125 |
| }, |
| { |
| "epoch": 0.74, |
| "learning_rate": 2.8000000000000003e-06, |
| "loss": 0.0368, |
| "step": 18150 |
| }, |
| { |
| "epoch": 0.74, |
| "learning_rate": 2.7897959183673474e-06, |
| "loss": 0.0356, |
| "step": 18175 |
| }, |
| { |
| "epoch": 0.74, |
| "learning_rate": 2.779591836734694e-06, |
| "loss": 0.0376, |
| "step": 18200 |
| }, |
| { |
| "epoch": 0.74, |
| "learning_rate": 2.769387755102041e-06, |
| "loss": 0.0378, |
| "step": 18225 |
| }, |
| { |
| "epoch": 0.74, |
| "learning_rate": 2.7591836734693877e-06, |
| "loss": 0.0384, |
| "step": 18250 |
| }, |
| { |
| "epoch": 0.74, |
| "learning_rate": 2.748979591836735e-06, |
| "loss": 0.0352, |
| "step": 18275 |
| }, |
| { |
| "epoch": 0.74, |
| "learning_rate": 2.738775510204082e-06, |
| "loss": 0.0355, |
| "step": 18300 |
| }, |
| { |
| "epoch": 0.74, |
| "learning_rate": 2.728571428571429e-06, |
| "loss": 0.036, |
| "step": 18325 |
| }, |
| { |
| "epoch": 0.75, |
| "learning_rate": 2.7183673469387755e-06, |
| "loss": 0.0365, |
| "step": 18350 |
| }, |
| { |
| "epoch": 0.75, |
| "learning_rate": 2.708163265306123e-06, |
| "loss": 0.0373, |
| "step": 18375 |
| }, |
| { |
| "epoch": 0.75, |
| "learning_rate": 2.6979591836734697e-06, |
| "loss": 0.0371, |
| "step": 18400 |
| }, |
| { |
| "epoch": 0.75, |
| "learning_rate": 2.6877551020408167e-06, |
| "loss": 0.035, |
| "step": 18425 |
| }, |
| { |
| "epoch": 0.75, |
| "learning_rate": 2.6775510204081634e-06, |
| "loss": 0.0353, |
| "step": 18450 |
| }, |
| { |
| "epoch": 0.75, |
| "learning_rate": 2.6673469387755104e-06, |
| "loss": 0.0358, |
| "step": 18475 |
| }, |
| { |
| "epoch": 0.75, |
| "learning_rate": 2.6571428571428575e-06, |
| "loss": 0.0341, |
| "step": 18500 |
| }, |
| { |
| "epoch": 0.75, |
| "learning_rate": 2.6469387755102045e-06, |
| "loss": 0.0361, |
| "step": 18525 |
| }, |
| { |
| "epoch": 0.75, |
| "learning_rate": 2.636734693877551e-06, |
| "loss": 0.036, |
| "step": 18550 |
| }, |
| { |
| "epoch": 0.75, |
| "learning_rate": 2.626530612244898e-06, |
| "loss": 0.0354, |
| "step": 18575 |
| }, |
| { |
| "epoch": 0.76, |
| "learning_rate": 2.616326530612245e-06, |
| "loss": 0.0373, |
| "step": 18600 |
| }, |
| { |
| "epoch": 0.76, |
| "learning_rate": 2.6061224489795924e-06, |
| "loss": 0.036, |
| "step": 18625 |
| }, |
| { |
| "epoch": 0.76, |
| "learning_rate": 2.595918367346939e-06, |
| "loss": 0.0365, |
| "step": 18650 |
| }, |
| { |
| "epoch": 0.76, |
| "learning_rate": 2.5857142857142856e-06, |
| "loss": 0.0369, |
| "step": 18675 |
| }, |
| { |
| "epoch": 0.76, |
| "learning_rate": 2.5755102040816327e-06, |
| "loss": 0.0347, |
| "step": 18700 |
| }, |
| { |
| "epoch": 0.76, |
| "learning_rate": 2.56530612244898e-06, |
| "loss": 0.0362, |
| "step": 18725 |
| }, |
| { |
| "epoch": 0.76, |
| "learning_rate": 2.555102040816327e-06, |
| "loss": 0.0357, |
| "step": 18750 |
| }, |
| { |
| "epoch": 0.76, |
| "learning_rate": 2.5448979591836735e-06, |
| "loss": 0.0355, |
| "step": 18775 |
| }, |
| { |
| "epoch": 0.76, |
| "learning_rate": 2.5346938775510205e-06, |
| "loss": 0.0374, |
| "step": 18800 |
| }, |
| { |
| "epoch": 0.76, |
| "learning_rate": 2.524489795918367e-06, |
| "loss": 0.0349, |
| "step": 18825 |
| }, |
| { |
| "epoch": 0.77, |
| "learning_rate": 2.5142857142857147e-06, |
| "loss": 0.0355, |
| "step": 18850 |
| }, |
| { |
| "epoch": 0.77, |
| "learning_rate": 2.5040816326530613e-06, |
| "loss": 0.0375, |
| "step": 18875 |
| }, |
| { |
| "epoch": 0.77, |
| "learning_rate": 2.4938775510204084e-06, |
| "loss": 0.0366, |
| "step": 18900 |
| }, |
| { |
| "epoch": 0.77, |
| "learning_rate": 2.4836734693877554e-06, |
| "loss": 0.0359, |
| "step": 18925 |
| }, |
| { |
| "epoch": 0.77, |
| "learning_rate": 2.473469387755102e-06, |
| "loss": 0.0346, |
| "step": 18950 |
| }, |
| { |
| "epoch": 0.77, |
| "learning_rate": 2.463265306122449e-06, |
| "loss": 0.0349, |
| "step": 18975 |
| }, |
| { |
| "epoch": 0.77, |
| "learning_rate": 2.453061224489796e-06, |
| "loss": 0.0361, |
| "step": 19000 |
| }, |
| { |
| "epoch": 0.77, |
| "learning_rate": 2.442857142857143e-06, |
| "loss": 0.0376, |
| "step": 19025 |
| }, |
| { |
| "epoch": 0.77, |
| "learning_rate": 2.43265306122449e-06, |
| "loss": 0.035, |
| "step": 19050 |
| }, |
| { |
| "epoch": 0.77, |
| "learning_rate": 2.422448979591837e-06, |
| "loss": 0.0355, |
| "step": 19075 |
| }, |
| { |
| "epoch": 0.78, |
| "learning_rate": 2.412244897959184e-06, |
| "loss": 0.0374, |
| "step": 19100 |
| }, |
| { |
| "epoch": 0.78, |
| "learning_rate": 2.4020408163265306e-06, |
| "loss": 0.0375, |
| "step": 19125 |
| }, |
| { |
| "epoch": 0.78, |
| "learning_rate": 2.3918367346938777e-06, |
| "loss": 0.0374, |
| "step": 19150 |
| }, |
| { |
| "epoch": 0.78, |
| "learning_rate": 2.3816326530612248e-06, |
| "loss": 0.035, |
| "step": 19175 |
| }, |
| { |
| "epoch": 0.78, |
| "learning_rate": 2.371428571428572e-06, |
| "loss": 0.0368, |
| "step": 19200 |
| }, |
| { |
| "epoch": 0.78, |
| "learning_rate": 2.3612244897959185e-06, |
| "loss": 0.035, |
| "step": 19225 |
| }, |
| { |
| "epoch": 0.78, |
| "learning_rate": 2.3510204081632655e-06, |
| "loss": 0.0333, |
| "step": 19250 |
| }, |
| { |
| "epoch": 0.78, |
| "learning_rate": 2.3408163265306126e-06, |
| "loss": 0.0332, |
| "step": 19275 |
| }, |
| { |
| "epoch": 0.78, |
| "learning_rate": 2.3306122448979592e-06, |
| "loss": 0.0327, |
| "step": 19300 |
| }, |
| { |
| "epoch": 0.78, |
| "learning_rate": 2.3204081632653063e-06, |
| "loss": 0.0348, |
| "step": 19325 |
| }, |
| { |
| "epoch": 0.79, |
| "learning_rate": 2.3102040816326534e-06, |
| "loss": 0.0345, |
| "step": 19350 |
| }, |
| { |
| "epoch": 0.79, |
| "learning_rate": 2.3000000000000004e-06, |
| "loss": 0.0361, |
| "step": 19375 |
| }, |
| { |
| "epoch": 0.79, |
| "learning_rate": 2.289795918367347e-06, |
| "loss": 0.034, |
| "step": 19400 |
| }, |
| { |
| "epoch": 0.79, |
| "learning_rate": 2.279591836734694e-06, |
| "loss": 0.035, |
| "step": 19425 |
| }, |
| { |
| "epoch": 0.79, |
| "learning_rate": 2.269387755102041e-06, |
| "loss": 0.0359, |
| "step": 19450 |
| }, |
| { |
| "epoch": 0.79, |
| "learning_rate": 2.259183673469388e-06, |
| "loss": 0.0383, |
| "step": 19475 |
| }, |
| { |
| "epoch": 0.79, |
| "learning_rate": 2.249387755102041e-06, |
| "loss": 0.0359, |
| "step": 19500 |
| }, |
| { |
| "epoch": 0.79, |
| "learning_rate": 2.239183673469388e-06, |
| "loss": 0.0347, |
| "step": 19525 |
| }, |
| { |
| "epoch": 0.79, |
| "learning_rate": 2.228979591836735e-06, |
| "loss": 0.0357, |
| "step": 19550 |
| }, |
| { |
| "epoch": 0.79, |
| "learning_rate": 2.2187755102040817e-06, |
| "loss": 0.0353, |
| "step": 19575 |
| }, |
| { |
| "epoch": 0.8, |
| "learning_rate": 2.2085714285714288e-06, |
| "loss": 0.0347, |
| "step": 19600 |
| }, |
| { |
| "epoch": 0.8, |
| "learning_rate": 2.1983673469387754e-06, |
| "loss": 0.0377, |
| "step": 19625 |
| }, |
| { |
| "epoch": 0.8, |
| "learning_rate": 2.188163265306123e-06, |
| "loss": 0.0376, |
| "step": 19650 |
| }, |
| { |
| "epoch": 0.8, |
| "learning_rate": 2.1779591836734695e-06, |
| "loss": 0.0352, |
| "step": 19675 |
| }, |
| { |
| "epoch": 0.8, |
| "learning_rate": 2.1677551020408166e-06, |
| "loss": 0.0321, |
| "step": 19700 |
| }, |
| { |
| "epoch": 0.8, |
| "learning_rate": 2.1575510204081632e-06, |
| "loss": 0.0344, |
| "step": 19725 |
| }, |
| { |
| "epoch": 0.8, |
| "learning_rate": 2.1473469387755103e-06, |
| "loss": 0.036, |
| "step": 19750 |
| }, |
| { |
| "epoch": 0.8, |
| "learning_rate": 2.1371428571428574e-06, |
| "loss": 0.0349, |
| "step": 19775 |
| }, |
| { |
| "epoch": 0.8, |
| "learning_rate": 2.126938775510204e-06, |
| "loss": 0.0342, |
| "step": 19800 |
| }, |
| { |
| "epoch": 0.81, |
| "learning_rate": 2.116734693877551e-06, |
| "loss": 0.0337, |
| "step": 19825 |
| }, |
| { |
| "epoch": 0.81, |
| "learning_rate": 2.106530612244898e-06, |
| "loss": 0.0349, |
| "step": 19850 |
| }, |
| { |
| "epoch": 0.81, |
| "learning_rate": 2.096326530612245e-06, |
| "loss": 0.0352, |
| "step": 19875 |
| }, |
| { |
| "epoch": 0.81, |
| "learning_rate": 2.086122448979592e-06, |
| "loss": 0.035, |
| "step": 19900 |
| }, |
| { |
| "epoch": 0.81, |
| "learning_rate": 2.075918367346939e-06, |
| "loss": 0.0338, |
| "step": 19925 |
| }, |
| { |
| "epoch": 0.81, |
| "learning_rate": 2.065714285714286e-06, |
| "loss": 0.0356, |
| "step": 19950 |
| }, |
| { |
| "epoch": 0.81, |
| "learning_rate": 2.0555102040816326e-06, |
| "loss": 0.033, |
| "step": 19975 |
| }, |
| { |
| "epoch": 0.81, |
| "learning_rate": 2.0453061224489797e-06, |
| "loss": 0.0366, |
| "step": 20000 |
| }, |
| { |
| "epoch": 0.81, |
| "eval_loss": 0.012057455256581306, |
| "eval_runtime": 73.1245, |
| "eval_samples_per_second": 1.75, |
| "eval_steps_per_second": 0.055, |
| "eval_wer": 19.385593220338983, |
| "step": 20000 |
| } |
| ], |
| "max_steps": 25000, |
| "num_train_epochs": 2, |
| "total_flos": 1.3063742951815905e+21, |
| "trial_name": null, |
| "trial_params": null |
| } |
|
|