| { | |
| "best_metric": 0.7136163115501404, | |
| "best_model_checkpoint": "./checkpoint-1000", | |
| "epoch": 624.0016, | |
| "global_step": 5000, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 3.0, | |
| "learning_rate": 4.4e-07, | |
| "loss": 1.9805, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "learning_rate": 9.400000000000001e-07, | |
| "loss": 1.1395, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "learning_rate": 1.44e-06, | |
| "loss": 0.3496, | |
| "step": 75 | |
| }, | |
| { | |
| "epoch": 12.0, | |
| "learning_rate": 1.94e-06, | |
| "loss": 0.1296, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 15.0, | |
| "learning_rate": 2.4400000000000004e-06, | |
| "loss": 0.0407, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 18.0, | |
| "learning_rate": 2.9400000000000002e-06, | |
| "loss": 0.008, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 21.0, | |
| "learning_rate": 3.44e-06, | |
| "loss": 0.0039, | |
| "step": 175 | |
| }, | |
| { | |
| "epoch": 24.0, | |
| "learning_rate": 3.94e-06, | |
| "loss": 0.0028, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 28.0, | |
| "learning_rate": 4.440000000000001e-06, | |
| "loss": 0.0016, | |
| "step": 225 | |
| }, | |
| { | |
| "epoch": 31.0, | |
| "learning_rate": 4.94e-06, | |
| "loss": 0.0011, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 34.0, | |
| "learning_rate": 5.4400000000000004e-06, | |
| "loss": 0.0009, | |
| "step": 275 | |
| }, | |
| { | |
| "epoch": 37.0, | |
| "learning_rate": 5.94e-06, | |
| "loss": 0.0007, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 40.0, | |
| "learning_rate": 6.440000000000001e-06, | |
| "loss": 0.0006, | |
| "step": 325 | |
| }, | |
| { | |
| "epoch": 43.0, | |
| "learning_rate": 6.9400000000000005e-06, | |
| "loss": 0.0005, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 46.0, | |
| "learning_rate": 7.440000000000001e-06, | |
| "loss": 0.0004, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 49.0, | |
| "learning_rate": 7.94e-06, | |
| "loss": 0.0004, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 53.0, | |
| "learning_rate": 8.44e-06, | |
| "loss": 0.0003, | |
| "step": 425 | |
| }, | |
| { | |
| "epoch": 56.0, | |
| "learning_rate": 8.94e-06, | |
| "loss": 0.0003, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 59.0, | |
| "learning_rate": 9.440000000000001e-06, | |
| "loss": 0.0003, | |
| "step": 475 | |
| }, | |
| { | |
| "epoch": 62.0, | |
| "learning_rate": 9.940000000000001e-06, | |
| "loss": 0.0003, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 65.0, | |
| "learning_rate": 9.951111111111111e-06, | |
| "loss": 0.0002, | |
| "step": 525 | |
| }, | |
| { | |
| "epoch": 68.0, | |
| "learning_rate": 9.895555555555557e-06, | |
| "loss": 0.0002, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 71.0, | |
| "learning_rate": 9.84e-06, | |
| "loss": 0.0002, | |
| "step": 575 | |
| }, | |
| { | |
| "epoch": 74.0, | |
| "learning_rate": 9.784444444444445e-06, | |
| "loss": 0.0002, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 78.0, | |
| "learning_rate": 9.72888888888889e-06, | |
| "loss": 0.0002, | |
| "step": 625 | |
| }, | |
| { | |
| "epoch": 81.0, | |
| "learning_rate": 9.673333333333334e-06, | |
| "loss": 0.0002, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 84.0, | |
| "learning_rate": 9.617777777777778e-06, | |
| "loss": 0.0002, | |
| "step": 675 | |
| }, | |
| { | |
| "epoch": 87.0, | |
| "learning_rate": 9.562222222222223e-06, | |
| "loss": 0.0002, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 90.0, | |
| "learning_rate": 9.506666666666667e-06, | |
| "loss": 0.0002, | |
| "step": 725 | |
| }, | |
| { | |
| "epoch": 93.0, | |
| "learning_rate": 9.451111111111112e-06, | |
| "loss": 0.0002, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 96.0, | |
| "learning_rate": 9.395555555555556e-06, | |
| "loss": 0.0001, | |
| "step": 775 | |
| }, | |
| { | |
| "epoch": 99.0, | |
| "learning_rate": 9.340000000000002e-06, | |
| "loss": 0.0001, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 103.0, | |
| "learning_rate": 9.284444444444444e-06, | |
| "loss": 0.0001, | |
| "step": 825 | |
| }, | |
| { | |
| "epoch": 106.0, | |
| "learning_rate": 9.22888888888889e-06, | |
| "loss": 0.0001, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 109.0, | |
| "learning_rate": 9.173333333333334e-06, | |
| "loss": 0.0001, | |
| "step": 875 | |
| }, | |
| { | |
| "epoch": 112.0, | |
| "learning_rate": 9.117777777777778e-06, | |
| "loss": 0.0001, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 115.0, | |
| "learning_rate": 9.062222222222224e-06, | |
| "loss": 0.0001, | |
| "step": 925 | |
| }, | |
| { | |
| "epoch": 118.0, | |
| "learning_rate": 9.006666666666666e-06, | |
| "loss": 0.0001, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 121.0, | |
| "learning_rate": 8.951111111111112e-06, | |
| "loss": 0.0001, | |
| "step": 975 | |
| }, | |
| { | |
| "epoch": 124.0, | |
| "learning_rate": 8.895555555555556e-06, | |
| "loss": 0.0001, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 124.0, | |
| "eval_loss": 0.7136163115501404, | |
| "eval_runtime": 262.7101, | |
| "eval_samples_per_second": 4.709, | |
| "eval_steps_per_second": 0.148, | |
| "eval_wer": 15.492494795661225, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 128.0, | |
| "learning_rate": 8.84e-06, | |
| "loss": 0.0001, | |
| "step": 1025 | |
| }, | |
| { | |
| "epoch": 131.0, | |
| "learning_rate": 8.784444444444446e-06, | |
| "loss": 0.0001, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 134.0, | |
| "learning_rate": 8.72888888888889e-06, | |
| "loss": 0.0001, | |
| "step": 1075 | |
| }, | |
| { | |
| "epoch": 137.0, | |
| "learning_rate": 8.673333333333334e-06, | |
| "loss": 0.0001, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 140.0, | |
| "learning_rate": 8.617777777777778e-06, | |
| "loss": 0.0001, | |
| "step": 1125 | |
| }, | |
| { | |
| "epoch": 143.0, | |
| "learning_rate": 8.562222222222224e-06, | |
| "loss": 0.0001, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 146.0, | |
| "learning_rate": 8.506666666666668e-06, | |
| "loss": 0.0001, | |
| "step": 1175 | |
| }, | |
| { | |
| "epoch": 149.0, | |
| "learning_rate": 8.451111111111112e-06, | |
| "loss": 0.0001, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 153.0, | |
| "learning_rate": 8.395555555555557e-06, | |
| "loss": 0.0001, | |
| "step": 1225 | |
| }, | |
| { | |
| "epoch": 156.0, | |
| "learning_rate": 8.34e-06, | |
| "loss": 0.0001, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 159.0, | |
| "learning_rate": 8.284444444444446e-06, | |
| "loss": 0.0001, | |
| "step": 1275 | |
| }, | |
| { | |
| "epoch": 162.0, | |
| "learning_rate": 8.22888888888889e-06, | |
| "loss": 0.0001, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 165.0, | |
| "learning_rate": 8.173333333333334e-06, | |
| "loss": 0.0001, | |
| "step": 1325 | |
| }, | |
| { | |
| "epoch": 168.0, | |
| "learning_rate": 8.11777777777778e-06, | |
| "loss": 0.0001, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 171.0, | |
| "learning_rate": 8.062222222222222e-06, | |
| "loss": 0.0001, | |
| "step": 1375 | |
| }, | |
| { | |
| "epoch": 174.0, | |
| "learning_rate": 8.006666666666667e-06, | |
| "loss": 0.0001, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 178.0, | |
| "learning_rate": 7.951111111111111e-06, | |
| "loss": 0.0001, | |
| "step": 1425 | |
| }, | |
| { | |
| "epoch": 181.0, | |
| "learning_rate": 7.895555555555557e-06, | |
| "loss": 0.0001, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 184.0, | |
| "learning_rate": 7.840000000000001e-06, | |
| "loss": 0.0001, | |
| "step": 1475 | |
| }, | |
| { | |
| "epoch": 187.0, | |
| "learning_rate": 7.784444444444445e-06, | |
| "loss": 0.0001, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 190.0, | |
| "learning_rate": 7.72888888888889e-06, | |
| "loss": 0.0001, | |
| "step": 1525 | |
| }, | |
| { | |
| "epoch": 193.0, | |
| "learning_rate": 7.673333333333333e-06, | |
| "loss": 0.0001, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 196.0, | |
| "learning_rate": 7.617777777777778e-06, | |
| "loss": 0.0001, | |
| "step": 1575 | |
| }, | |
| { | |
| "epoch": 199.0, | |
| "learning_rate": 7.562222222222223e-06, | |
| "loss": 0.0001, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 203.0, | |
| "learning_rate": 7.506666666666668e-06, | |
| "loss": 0.0001, | |
| "step": 1625 | |
| }, | |
| { | |
| "epoch": 206.0, | |
| "learning_rate": 7.451111111111111e-06, | |
| "loss": 0.0001, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 209.0, | |
| "learning_rate": 7.395555555555556e-06, | |
| "loss": 0.0001, | |
| "step": 1675 | |
| }, | |
| { | |
| "epoch": 212.0, | |
| "learning_rate": 7.340000000000001e-06, | |
| "loss": 0.0001, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 215.0, | |
| "learning_rate": 7.284444444444445e-06, | |
| "loss": 0.0001, | |
| "step": 1725 | |
| }, | |
| { | |
| "epoch": 218.0, | |
| "learning_rate": 7.22888888888889e-06, | |
| "loss": 0.0001, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 221.0, | |
| "learning_rate": 7.173333333333335e-06, | |
| "loss": 0.0001, | |
| "step": 1775 | |
| }, | |
| { | |
| "epoch": 224.0, | |
| "learning_rate": 7.117777777777778e-06, | |
| "loss": 0.0001, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 228.0, | |
| "learning_rate": 7.062222222222223e-06, | |
| "loss": 0.0001, | |
| "step": 1825 | |
| }, | |
| { | |
| "epoch": 231.0, | |
| "learning_rate": 7.006666666666667e-06, | |
| "loss": 0.0001, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 234.0, | |
| "learning_rate": 6.951111111111112e-06, | |
| "loss": 0.0001, | |
| "step": 1875 | |
| }, | |
| { | |
| "epoch": 237.0, | |
| "learning_rate": 6.8955555555555565e-06, | |
| "loss": 0.0001, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 240.0, | |
| "learning_rate": 6.8400000000000014e-06, | |
| "loss": 0.0001, | |
| "step": 1925 | |
| }, | |
| { | |
| "epoch": 243.0, | |
| "learning_rate": 6.784444444444445e-06, | |
| "loss": 0.0001, | |
| "step": 1950 | |
| }, | |
| { | |
| "epoch": 246.0, | |
| "learning_rate": 6.7288888888888895e-06, | |
| "loss": 0.0001, | |
| "step": 1975 | |
| }, | |
| { | |
| "epoch": 249.0, | |
| "learning_rate": 6.6733333333333335e-06, | |
| "loss": 0.0001, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 249.0, | |
| "eval_loss": 0.8531960844993591, | |
| "eval_runtime": 264.7965, | |
| "eval_samples_per_second": 4.672, | |
| "eval_steps_per_second": 0.147, | |
| "eval_wer": 17.00449216610058, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 253.0, | |
| "learning_rate": 6.617777777777778e-06, | |
| "loss": 0.0, | |
| "step": 2025 | |
| }, | |
| { | |
| "epoch": 256.0, | |
| "learning_rate": 6.562222222222223e-06, | |
| "loss": 0.0, | |
| "step": 2050 | |
| }, | |
| { | |
| "epoch": 259.0, | |
| "learning_rate": 6.5066666666666665e-06, | |
| "loss": 0.0, | |
| "step": 2075 | |
| }, | |
| { | |
| "epoch": 262.0, | |
| "learning_rate": 6.451111111111111e-06, | |
| "loss": 0.0, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 265.0, | |
| "learning_rate": 6.395555555555556e-06, | |
| "loss": 0.0, | |
| "step": 2125 | |
| }, | |
| { | |
| "epoch": 268.0, | |
| "learning_rate": 6.34e-06, | |
| "loss": 0.0, | |
| "step": 2150 | |
| }, | |
| { | |
| "epoch": 271.0, | |
| "learning_rate": 6.284444444444445e-06, | |
| "loss": 0.0, | |
| "step": 2175 | |
| }, | |
| { | |
| "epoch": 274.0, | |
| "learning_rate": 6.22888888888889e-06, | |
| "loss": 0.0, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 278.0, | |
| "learning_rate": 6.173333333333333e-06, | |
| "loss": 0.0, | |
| "step": 2225 | |
| }, | |
| { | |
| "epoch": 281.0, | |
| "learning_rate": 6.117777777777778e-06, | |
| "loss": 0.0, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 284.0, | |
| "learning_rate": 6.062222222222223e-06, | |
| "loss": 0.0, | |
| "step": 2275 | |
| }, | |
| { | |
| "epoch": 287.0, | |
| "learning_rate": 6.006666666666667e-06, | |
| "loss": 0.0, | |
| "step": 2300 | |
| }, | |
| { | |
| "epoch": 290.0, | |
| "learning_rate": 5.951111111111112e-06, | |
| "loss": 0.0, | |
| "step": 2325 | |
| }, | |
| { | |
| "epoch": 293.0, | |
| "learning_rate": 5.895555555555557e-06, | |
| "loss": 0.0, | |
| "step": 2350 | |
| }, | |
| { | |
| "epoch": 296.0, | |
| "learning_rate": 5.84e-06, | |
| "loss": 0.0, | |
| "step": 2375 | |
| }, | |
| { | |
| "epoch": 299.0, | |
| "learning_rate": 5.784444444444445e-06, | |
| "loss": 0.0, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 303.0, | |
| "learning_rate": 5.72888888888889e-06, | |
| "loss": 0.0, | |
| "step": 2425 | |
| }, | |
| { | |
| "epoch": 306.0, | |
| "learning_rate": 5.673333333333334e-06, | |
| "loss": 0.0, | |
| "step": 2450 | |
| }, | |
| { | |
| "epoch": 309.0, | |
| "learning_rate": 5.617777777777779e-06, | |
| "loss": 0.0, | |
| "step": 2475 | |
| }, | |
| { | |
| "epoch": 312.0, | |
| "learning_rate": 5.562222222222222e-06, | |
| "loss": 0.0, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 315.0, | |
| "learning_rate": 5.506666666666667e-06, | |
| "loss": 0.0, | |
| "step": 2525 | |
| }, | |
| { | |
| "epoch": 318.0, | |
| "learning_rate": 5.451111111111112e-06, | |
| "loss": 0.0, | |
| "step": 2550 | |
| }, | |
| { | |
| "epoch": 321.0, | |
| "learning_rate": 5.3955555555555565e-06, | |
| "loss": 0.0, | |
| "step": 2575 | |
| }, | |
| { | |
| "epoch": 324.0, | |
| "learning_rate": 5.3400000000000005e-06, | |
| "loss": 0.0, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 328.0, | |
| "learning_rate": 5.2844444444444454e-06, | |
| "loss": 0.0, | |
| "step": 2625 | |
| }, | |
| { | |
| "epoch": 331.0, | |
| "learning_rate": 5.228888888888889e-06, | |
| "loss": 0.0, | |
| "step": 2650 | |
| }, | |
| { | |
| "epoch": 334.0, | |
| "learning_rate": 5.1733333333333335e-06, | |
| "loss": 0.0, | |
| "step": 2675 | |
| }, | |
| { | |
| "epoch": 337.0, | |
| "learning_rate": 5.117777777777778e-06, | |
| "loss": 0.0, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 340.0, | |
| "learning_rate": 5.062222222222222e-06, | |
| "loss": 0.0, | |
| "step": 2725 | |
| }, | |
| { | |
| "epoch": 343.0, | |
| "learning_rate": 5.006666666666667e-06, | |
| "loss": 0.0, | |
| "step": 2750 | |
| }, | |
| { | |
| "epoch": 346.0, | |
| "learning_rate": 4.951111111111111e-06, | |
| "loss": 0.0, | |
| "step": 2775 | |
| }, | |
| { | |
| "epoch": 349.0, | |
| "learning_rate": 4.895555555555556e-06, | |
| "loss": 0.0, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 353.0, | |
| "learning_rate": 4.84e-06, | |
| "loss": 0.0, | |
| "step": 2825 | |
| }, | |
| { | |
| "epoch": 356.0, | |
| "learning_rate": 4.784444444444445e-06, | |
| "loss": 0.0, | |
| "step": 2850 | |
| }, | |
| { | |
| "epoch": 359.0, | |
| "learning_rate": 4.728888888888889e-06, | |
| "loss": 0.0, | |
| "step": 2875 | |
| }, | |
| { | |
| "epoch": 362.0, | |
| "learning_rate": 4.673333333333333e-06, | |
| "loss": 0.0, | |
| "step": 2900 | |
| }, | |
| { | |
| "epoch": 365.0, | |
| "learning_rate": 4.617777777777778e-06, | |
| "loss": 0.0, | |
| "step": 2925 | |
| }, | |
| { | |
| "epoch": 368.0, | |
| "learning_rate": 4.562222222222222e-06, | |
| "loss": 0.0, | |
| "step": 2950 | |
| }, | |
| { | |
| "epoch": 371.0, | |
| "learning_rate": 4.506666666666667e-06, | |
| "loss": 0.0, | |
| "step": 2975 | |
| }, | |
| { | |
| "epoch": 374.0, | |
| "learning_rate": 4.451111111111112e-06, | |
| "loss": 0.0, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 374.0, | |
| "eval_loss": 0.9251218438148499, | |
| "eval_runtime": 248.0815, | |
| "eval_samples_per_second": 4.986, | |
| "eval_steps_per_second": 0.157, | |
| "eval_wer": 19.097184178810124, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 378.0, | |
| "learning_rate": 4.395555555555556e-06, | |
| "loss": 0.0, | |
| "step": 3025 | |
| }, | |
| { | |
| "epoch": 381.0, | |
| "learning_rate": 4.34e-06, | |
| "loss": 0.0, | |
| "step": 3050 | |
| }, | |
| { | |
| "epoch": 384.0, | |
| "learning_rate": 4.284444444444445e-06, | |
| "loss": 0.0, | |
| "step": 3075 | |
| }, | |
| { | |
| "epoch": 387.0, | |
| "learning_rate": 4.228888888888889e-06, | |
| "loss": 0.0, | |
| "step": 3100 | |
| }, | |
| { | |
| "epoch": 390.0, | |
| "learning_rate": 4.173333333333334e-06, | |
| "loss": 0.0, | |
| "step": 3125 | |
| }, | |
| { | |
| "epoch": 393.0, | |
| "learning_rate": 4.117777777777779e-06, | |
| "loss": 0.0, | |
| "step": 3150 | |
| }, | |
| { | |
| "epoch": 396.0, | |
| "learning_rate": 4.062222222222223e-06, | |
| "loss": 0.0, | |
| "step": 3175 | |
| }, | |
| { | |
| "epoch": 399.0, | |
| "learning_rate": 4.006666666666667e-06, | |
| "loss": 0.0, | |
| "step": 3200 | |
| }, | |
| { | |
| "epoch": 403.0, | |
| "learning_rate": 3.951111111111112e-06, | |
| "loss": 0.0, | |
| "step": 3225 | |
| }, | |
| { | |
| "epoch": 406.0, | |
| "learning_rate": 3.895555555555556e-06, | |
| "loss": 0.0, | |
| "step": 3250 | |
| }, | |
| { | |
| "epoch": 409.0, | |
| "learning_rate": 3.8400000000000005e-06, | |
| "loss": 0.0, | |
| "step": 3275 | |
| }, | |
| { | |
| "epoch": 412.0, | |
| "learning_rate": 3.784444444444445e-06, | |
| "loss": 0.0, | |
| "step": 3300 | |
| }, | |
| { | |
| "epoch": 415.0, | |
| "learning_rate": 3.728888888888889e-06, | |
| "loss": 0.0, | |
| "step": 3325 | |
| }, | |
| { | |
| "epoch": 418.0, | |
| "learning_rate": 3.673333333333334e-06, | |
| "loss": 0.0, | |
| "step": 3350 | |
| }, | |
| { | |
| "epoch": 421.0, | |
| "learning_rate": 3.617777777777778e-06, | |
| "loss": 0.0, | |
| "step": 3375 | |
| }, | |
| { | |
| "epoch": 424.0, | |
| "learning_rate": 3.5622222222222224e-06, | |
| "loss": 0.0, | |
| "step": 3400 | |
| }, | |
| { | |
| "epoch": 428.0, | |
| "learning_rate": 3.5066666666666673e-06, | |
| "loss": 0.0, | |
| "step": 3425 | |
| }, | |
| { | |
| "epoch": 431.0, | |
| "learning_rate": 3.4511111111111113e-06, | |
| "loss": 0.0, | |
| "step": 3450 | |
| }, | |
| { | |
| "epoch": 434.0, | |
| "learning_rate": 3.3955555555555558e-06, | |
| "loss": 0.0, | |
| "step": 3475 | |
| }, | |
| { | |
| "epoch": 437.0, | |
| "learning_rate": 3.3400000000000006e-06, | |
| "loss": 0.0, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 440.0, | |
| "learning_rate": 3.2844444444444447e-06, | |
| "loss": 0.0, | |
| "step": 3525 | |
| }, | |
| { | |
| "epoch": 443.0, | |
| "learning_rate": 3.228888888888889e-06, | |
| "loss": 0.0, | |
| "step": 3550 | |
| }, | |
| { | |
| "epoch": 446.0, | |
| "learning_rate": 3.173333333333334e-06, | |
| "loss": 0.0, | |
| "step": 3575 | |
| }, | |
| { | |
| "epoch": 449.0, | |
| "learning_rate": 3.117777777777778e-06, | |
| "loss": 0.0, | |
| "step": 3600 | |
| }, | |
| { | |
| "epoch": 453.0, | |
| "learning_rate": 3.0622222222222225e-06, | |
| "loss": 0.0, | |
| "step": 3625 | |
| }, | |
| { | |
| "epoch": 456.0, | |
| "learning_rate": 3.0066666666666674e-06, | |
| "loss": 0.0, | |
| "step": 3650 | |
| }, | |
| { | |
| "epoch": 459.0, | |
| "learning_rate": 2.9511111111111114e-06, | |
| "loss": 0.0, | |
| "step": 3675 | |
| }, | |
| { | |
| "epoch": 462.0, | |
| "learning_rate": 2.895555555555556e-06, | |
| "loss": 0.0, | |
| "step": 3700 | |
| }, | |
| { | |
| "epoch": 465.0, | |
| "learning_rate": 2.84e-06, | |
| "loss": 0.0, | |
| "step": 3725 | |
| }, | |
| { | |
| "epoch": 468.0, | |
| "learning_rate": 2.784444444444445e-06, | |
| "loss": 0.0, | |
| "step": 3750 | |
| }, | |
| { | |
| "epoch": 471.0, | |
| "learning_rate": 2.7288888888888893e-06, | |
| "loss": 0.0, | |
| "step": 3775 | |
| }, | |
| { | |
| "epoch": 474.0, | |
| "learning_rate": 2.6733333333333333e-06, | |
| "loss": 0.0, | |
| "step": 3800 | |
| }, | |
| { | |
| "epoch": 478.0, | |
| "learning_rate": 2.617777777777778e-06, | |
| "loss": 0.0, | |
| "step": 3825 | |
| }, | |
| { | |
| "epoch": 481.0, | |
| "learning_rate": 2.5622222222222226e-06, | |
| "loss": 0.0, | |
| "step": 3850 | |
| }, | |
| { | |
| "epoch": 484.0, | |
| "learning_rate": 2.5066666666666667e-06, | |
| "loss": 0.0, | |
| "step": 3875 | |
| }, | |
| { | |
| "epoch": 487.0, | |
| "learning_rate": 2.451111111111111e-06, | |
| "loss": 0.0, | |
| "step": 3900 | |
| }, | |
| { | |
| "epoch": 490.0, | |
| "learning_rate": 2.3955555555555556e-06, | |
| "loss": 0.0, | |
| "step": 3925 | |
| }, | |
| { | |
| "epoch": 493.0, | |
| "learning_rate": 2.3400000000000005e-06, | |
| "loss": 0.0, | |
| "step": 3950 | |
| }, | |
| { | |
| "epoch": 496.0, | |
| "learning_rate": 2.2844444444444445e-06, | |
| "loss": 0.0, | |
| "step": 3975 | |
| }, | |
| { | |
| "epoch": 499.0, | |
| "learning_rate": 2.228888888888889e-06, | |
| "loss": 0.0, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 499.0, | |
| "eval_loss": 0.9786567687988281, | |
| "eval_runtime": 252.9235, | |
| "eval_samples_per_second": 4.891, | |
| "eval_steps_per_second": 0.154, | |
| "eval_wer": 21.595266790840366, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 503.0, | |
| "learning_rate": 2.1733333333333334e-06, | |
| "loss": 0.0, | |
| "step": 4025 | |
| }, | |
| { | |
| "epoch": 506.0, | |
| "learning_rate": 2.117777777777778e-06, | |
| "loss": 0.0, | |
| "step": 4050 | |
| }, | |
| { | |
| "epoch": 509.0, | |
| "learning_rate": 2.0622222222222223e-06, | |
| "loss": 0.0, | |
| "step": 4075 | |
| }, | |
| { | |
| "epoch": 512.0, | |
| "learning_rate": 2.006666666666667e-06, | |
| "loss": 0.0, | |
| "step": 4100 | |
| }, | |
| { | |
| "epoch": 515.0, | |
| "learning_rate": 1.9511111111111113e-06, | |
| "loss": 0.0, | |
| "step": 4125 | |
| }, | |
| { | |
| "epoch": 518.0, | |
| "learning_rate": 1.8955555555555557e-06, | |
| "loss": 0.0, | |
| "step": 4150 | |
| }, | |
| { | |
| "epoch": 521.0, | |
| "learning_rate": 1.8400000000000002e-06, | |
| "loss": 0.0, | |
| "step": 4175 | |
| }, | |
| { | |
| "epoch": 524.0, | |
| "learning_rate": 1.7844444444444444e-06, | |
| "loss": 0.0, | |
| "step": 4200 | |
| }, | |
| { | |
| "epoch": 528.0, | |
| "learning_rate": 1.728888888888889e-06, | |
| "loss": 0.0, | |
| "step": 4225 | |
| }, | |
| { | |
| "epoch": 531.0, | |
| "learning_rate": 1.6733333333333335e-06, | |
| "loss": 0.0, | |
| "step": 4250 | |
| }, | |
| { | |
| "epoch": 534.0, | |
| "learning_rate": 1.6177777777777778e-06, | |
| "loss": 0.0, | |
| "step": 4275 | |
| }, | |
| { | |
| "epoch": 537.0, | |
| "learning_rate": 1.5622222222222225e-06, | |
| "loss": 0.0, | |
| "step": 4300 | |
| }, | |
| { | |
| "epoch": 540.0, | |
| "learning_rate": 1.506666666666667e-06, | |
| "loss": 0.0, | |
| "step": 4325 | |
| }, | |
| { | |
| "epoch": 543.0, | |
| "learning_rate": 1.4511111111111112e-06, | |
| "loss": 0.0, | |
| "step": 4350 | |
| }, | |
| { | |
| "epoch": 546.0, | |
| "learning_rate": 1.3955555555555556e-06, | |
| "loss": 0.0, | |
| "step": 4375 | |
| }, | |
| { | |
| "epoch": 549.0, | |
| "learning_rate": 1.34e-06, | |
| "loss": 0.0, | |
| "step": 4400 | |
| }, | |
| { | |
| "epoch": 553.0, | |
| "learning_rate": 1.2844444444444445e-06, | |
| "loss": 0.0, | |
| "step": 4425 | |
| }, | |
| { | |
| "epoch": 556.0, | |
| "learning_rate": 1.228888888888889e-06, | |
| "loss": 0.0, | |
| "step": 4450 | |
| }, | |
| { | |
| "epoch": 559.0, | |
| "learning_rate": 1.1733333333333335e-06, | |
| "loss": 0.0, | |
| "step": 4475 | |
| }, | |
| { | |
| "epoch": 562.0, | |
| "learning_rate": 1.117777777777778e-06, | |
| "loss": 0.0, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 565.0, | |
| "learning_rate": 1.0622222222222222e-06, | |
| "loss": 0.0, | |
| "step": 4525 | |
| }, | |
| { | |
| "epoch": 568.0, | |
| "learning_rate": 1.0066666666666668e-06, | |
| "loss": 0.0, | |
| "step": 4550 | |
| }, | |
| { | |
| "epoch": 571.0, | |
| "learning_rate": 9.511111111111111e-07, | |
| "loss": 0.0, | |
| "step": 4575 | |
| }, | |
| { | |
| "epoch": 574.0, | |
| "learning_rate": 8.955555555555557e-07, | |
| "loss": 0.0, | |
| "step": 4600 | |
| }, | |
| { | |
| "epoch": 578.0, | |
| "learning_rate": 8.400000000000001e-07, | |
| "loss": 0.0, | |
| "step": 4625 | |
| }, | |
| { | |
| "epoch": 581.0, | |
| "learning_rate": 7.844444444444445e-07, | |
| "loss": 0.0, | |
| "step": 4650 | |
| }, | |
| { | |
| "epoch": 584.0, | |
| "learning_rate": 7.28888888888889e-07, | |
| "loss": 0.0, | |
| "step": 4675 | |
| }, | |
| { | |
| "epoch": 587.0, | |
| "learning_rate": 6.733333333333334e-07, | |
| "loss": 0.0, | |
| "step": 4700 | |
| }, | |
| { | |
| "epoch": 590.0, | |
| "learning_rate": 6.177777777777778e-07, | |
| "loss": 0.0, | |
| "step": 4725 | |
| }, | |
| { | |
| "epoch": 593.0, | |
| "learning_rate": 5.622222222222223e-07, | |
| "loss": 0.0, | |
| "step": 4750 | |
| }, | |
| { | |
| "epoch": 596.0, | |
| "learning_rate": 5.066666666666667e-07, | |
| "loss": 0.0, | |
| "step": 4775 | |
| }, | |
| { | |
| "epoch": 599.0, | |
| "learning_rate": 4.511111111111111e-07, | |
| "loss": 0.0, | |
| "step": 4800 | |
| }, | |
| { | |
| "epoch": 603.0, | |
| "learning_rate": 3.9555555555555557e-07, | |
| "loss": 0.0, | |
| "step": 4825 | |
| }, | |
| { | |
| "epoch": 606.0, | |
| "learning_rate": 3.4000000000000003e-07, | |
| "loss": 0.0, | |
| "step": 4850 | |
| }, | |
| { | |
| "epoch": 609.0, | |
| "learning_rate": 2.844444444444445e-07, | |
| "loss": 0.0, | |
| "step": 4875 | |
| }, | |
| { | |
| "epoch": 612.0, | |
| "learning_rate": 2.2888888888888892e-07, | |
| "loss": 0.0, | |
| "step": 4900 | |
| }, | |
| { | |
| "epoch": 615.0, | |
| "learning_rate": 1.7333333333333335e-07, | |
| "loss": 0.0, | |
| "step": 4925 | |
| }, | |
| { | |
| "epoch": 618.0, | |
| "learning_rate": 1.1777777777777778e-07, | |
| "loss": 0.0, | |
| "step": 4950 | |
| }, | |
| { | |
| "epoch": 621.0, | |
| "learning_rate": 6.222222222222223e-08, | |
| "loss": 0.0, | |
| "step": 4975 | |
| }, | |
| { | |
| "epoch": 624.0, | |
| "learning_rate": 6.666666666666667e-09, | |
| "loss": 0.0, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 624.0, | |
| "eval_loss": 0.9920965433120728, | |
| "eval_runtime": 252.4621, | |
| "eval_samples_per_second": 4.9, | |
| "eval_steps_per_second": 0.154, | |
| "eval_wer": 21.463788758628247, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 624.0, | |
| "step": 5000, | |
| "total_flos": 3.163875213312e+20, | |
| "train_loss": 0.01835221478660824, | |
| "train_runtime": 48453.2215, | |
| "train_samples_per_second": 6.604, | |
| "train_steps_per_second": 0.103 | |
| } | |
| ], | |
| "max_steps": 5000, | |
| "num_train_epochs": 9223372036854775807, | |
| "total_flos": 3.163875213312e+20, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |